No title

WATER RESEARCH A Journal of the International Water Association

Editor-in-Chief Mark van Loosdrecht Delft University of Technology Department of Biochemical Engineering Julianalaan 67 2628 BC Delft The Netherlands Tel: +31 15 27 81618 E-mail:[email protected]

Editors J. Block Université H. Poincaré, Nancy I France David Dixon University of Melbourne Australia Hiroaki Furumai The University of Tokyo Japan Xiaodi Hao Beijing University of Civil Engineering and Architecture China Gregory Korshin University of Washington USA Anna Ledin Formas Sweden Eberhard Morgenroth Swiss Federal Institute of Aquatic Science and Technology (EAWAG) Switzerland W. Rauch University Innsbruck Austria Maria Reis Universidade Nova de Lisboa/FCT Portugal Hang-Shik Shin Korea Advanced Institute of Science and Technology Korea Thomas Ternes Bundesanstalt für Gewässerkunde Germany Stefan Wuertz Univ. of California, Davis USA

Associate Editors Andrew Baker University of New South Wales Australia

Damien Batstone The University of Queensland Australia G-H. Chen The Hong Kong University of Science & Technology Hong Kong China Tom Curtis Univ. of Newcastle upon Tyne UK Ana Deletic Monash University USA Francis de los Reyes III North Carolina State University USA Rob Eldridge The University of Melbourne Australia Rosina Girones University of Barcelona Spain Stephen Gray Victoria University Australia Kate Grudpan Chiang Mai University Thailand E.E. Herricks University of Illinois - Urbana USA Peter Hillis United Utilities Plc UK H-Y. Hu Tsinghua University China P.M. Huck University of Waterloo Canada Bruce Jefferson Cranfield University UK Ulf Jeppsson Lund University Sweden Sergey Kalyuzhnyi Moscow State University Russian Federation Jaehong Kim Georgia Institute of Technology USA Jes La Cour Jansen Lund Institute of Technology Sweden G. Langergraber BOKU/Univ. of Natural Res. and Applied Life Scs. Austria S-L. Lo National Taiwan University Taiwan Dionisis Mantzavinos Technical University of Crete Greece

Y. Matsui Hokkaido University Japan A. Maul Université Paul Verlaine-Metz France Max Maurer EAWAG Switzerland How Yong Ng National University of Singapore Singapore Satoshi Okabe Hokkaido University Japan S.L. Ong National University of Singapore Singapore Jong M. Park Pohang University of Science & Technology Korea Susan Richardson U.S. Environmental Protection Agency USA Miguel Salgot University of Barcelona Spain David Sedlak University of California, Berkeley USA Jean-Philippe Steyer LBE-INRA France M. Takahashi Hokkaido University Japan Kai Udert EAWAG Switzerland V.P. Venugopalan BARC Facilities India E. von Sperling Federal University of Minas Gerais Brazil Hanqing Yu University of Science & Technology of China China J. Zilles University of Illinois Urbana USA A.I. Zouboulis Aristotle University Greece

Editorial Office Mailing address Water Research Radarweg 29 1043 NX Amsterdam The Netherlands Tel: 31 20 485 2701 Fax: 31 20 485 2521; 31 20 485 3264 E-mail: [email protected]

Publication information: Water Research (ISSN 0043-1354). For 2011, volume 45 is scheduled for publication. Subscription prices are available upon ­request from the publisher or from the Elsevier Customer Service Department nearest you or from this journal’s website (http://www.elsevier.com/locate/watres). Further information is available on this journal and other Elsevier products through Elsevier’s website (http://www.elsevier.com). Subscriptions are accepted on a prepaid basis only and are entered on a calendar year basis. Issues are sent by standard mail (surface within Europe, air delivery outside Europe). Priority rates are available upon request. Claims for missing issues should be made within six months of the date of dispatch. Author Enquiries: Please visit this journal’s homepage at http://www.elsevier.com/locate/watres. You can track accepted articles at http://www.elsevier.com/ trackarticle and set up e-mail alerts to inform you of when an article’s status has changed. Also accessible from here is information on copyright, frequently asked questions and more. Contact details for questions arising after acceptance of an article, especially those relating to proofs, will be provided by the publisher. Orders, claims, and journal enquiries: Please contact the Elsevier Customer Service Department nearest you: St. Louis: Elsevier Customer Service Department, 11830 Westline Industrial Drive, St. Louis, MO 63146, USA; phone: (877) 8397126 [toll free within the USA]; (+1) (314) 4537076 [outside the USA]; fax: (+1) (314) 5235153; e-mail: [email protected] Oxford: Elsevier Customer Service Department, The Boulevard, Langford Lane, Kidlington OX5 1GB, UK; phone: (+44) (1865) 843434; fax: (+44) (1865) 843970; e-mail: [email protected] Tokyo: Elsevier Customer Service Department, 4F Higashi-Azabu, 1-Chome Bldg, 1-9-15 Higashi-Azabu, Minato-ku, Tokyo 106-0044, Japan; phone: (+81) (3) 5561 5037; fax: (+81) (3) 5561 5047; e-mail: [email protected] Singapore: Elsevier Customer Service Department, 3 Killiney Road, #08-01 Winsland House I, Singapore 239519; phone: (+65) 63490222; fax: (+65) 67331510; e-mail: [email protected] Application for membership of International Water Association should be made to: Executive Director, IWA, Alliance House, 12 Caxton Street, London SW1H 0QS, U.K. (Tel.: +44 207 654 5500; Fax: +44 207 654 5555; e-mail: [email protected]; website: http://www.IWAhq.org.uk). Registered Charity (England) No. 289269. Individual membership is available from £30 upwards. For details contact IWA.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 1 9 e1 5 2 8

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Feasibility of a two-stage reduction/subsequent oxidation for treating Tetrabromobisphenol A in aqueous solutions Si Luo, Shao-gui Yang*, Cheng Sun*, Xiao-dong Wang State Key Laboratory of Pollution Control and Resource Reuse, School of the Environment, Nanjing University, Nanjing 210093, PR China

article info

abstract

Article history:

A “two-stage reduction/subsequent oxidation” (T-SRO) process consists of FeeAg reduction

Received 3 July 2010

and Fenton-like oxidation under ultrasound (US) radiation. Due to the refractory oxidation

Received in revised form

of brominated flame retardant, T-SRO was employed to remove Tetrabromobisphenol

22 October 2010

A (TBBPA) by the combination of first debromination and succeeding oxidation. It indicated

Accepted 31 October 2010

that the T-SRO process resulted in a complete decrease in TBBPA concentration and

Available online 10 December 2010

a 99.2% decrease in BPA concentration. The T-SRO process for the removal of TBBPA is much effective than Fenton-like oxidation of TBBPA alone. The result showed that US

Keywords:

radiation improved the Fenton-like oxidation rate of BPA solutions. The addition of dis-

Two-stage

solved iron into the Fenton-like oxidation system could accelerate the first 2 min reaction,

Debromination

but had little effect on the following process. The main intermediate products resulting

Fenton-like oxidation

from TBBPA reduction and BPA oxidation were identified by GCeMS and LC-MS/MS. On the

Tetrabromobisphenol A

basis of this analysis, reactions with
OH radical were identified as the major chemical

Bimetallic nanoparticles

pathways during BPA oxidation. ª 2010 Published by Elsevier Ltd.

Ultrasound radiation

1.

Introduction

Tetrabromobisphenol A (TBBPA) is one of the most widely used brominated flame retardant around the world. It can be covalently bound to the polymer in the manufacturing process (de Wit, 2002). TBBPA and its dimethylated derivative have been detected in various environmental matrices, and they negatively affect various aspects of mammalian and human physiology (Sellstro¨m and Jansson, 1995; Helleday ¨ berg et al., 2002). Conseet al., 1999; Meerts et al., 2000; O quently, removal of TBBPA in the contaminated environment is necessary and significant. The reported treatment mainly includes biotransformation, photochemical transformations and thermal decomposition (Mackenzie and Kopinke, 1996; Barontini et al., 2004; Eriksson et al., 2004). In addition, it also indicated that removal of the halogen substituent is a key step in the degradation of halogenated aromatic compounds.

This may occur as an initial step via reductive, hydrolytic, or oxygenolytic mechanisms or may occur after ring cleavage at a later stage of degradation (Monserrate and Haggblom, 1997). Zero valent iron (ZVI) and bimetallic particles have been used for degradation of halogen-containing organic substance (Orth and Gillham, 1996; Cwiertny et al., 2006). In our previous work (Luo et al., 2010), we reported that TBBPA was reductively debrominated to bisphenol A (BPA) over FeeAg bimetallic nanoparticles under US radiation. However, it is well known that BPA exhibits estrogenic activity, which increases the proliferation rate of breast cancer cells and induces the acute toxicity to freshwater and marine species (Pulgar et al., 1998; Kaiser, 2000). Therefore, the debromination of TBBPA in FeeAg/US system is incomplete, BPA must be further degraded. An effective method for BPA mineralizing is the application of Fenton (Fenton-like) oxidation technologies (Go¨zmen et al., 2003; Ioan et al., 2007). On the other hand,

* Corresponding authors. Tel./fax: þ86 25 89680580. E-mail addresses: [email protected] (S.-g. Yang), [email protected] (C. Sun). 0043-1354/$ e see front matter ª 2010 Published by Elsevier Ltd. doi:10.1016/j.watres.2010.10.039

1520

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 1 9 e1 5 2 8

ZVI can be used to substitute ferrous salts in the Fenton-like oxidation, and it seems to have similar degradation rates to homogeneous ferrous catalyst. It has confirmed that phenol (Bremner et al., 2006) and 4-chlorophenol (Zhou et al., 2008) could be rapidly degraded in ZVI/H2O2 system. Since iron can reductively transform the electron-withdrawing moieties and render recalcitrant compounds more amenable to subsequent oxidation processes, several researchers presented the ZVI reduction for the pretreatment of wastewater (Mantha et al., 2001, 2002; Oh et al., 2005). Oh et al. (2003) reported the enhanced Fenton oxidation of TNT and RDX through pretreatment with ZVI. Thus, in consideration of the complete treatment of TBBPA, ZVI-based reductive debromination followed by Fenton-like oxidation is proposed, where FeeAg bimetallic nanoparticles are used to debrominate TBBPA because of its higher catalytic activity relative to ZVI. This paper evaluates the effectiveness and feasibility of a T-SRO treatment of TBBPA. Experiments are conducted to examine separately the performance of the FeeAg nanoparticles reductive and Fenton-like oxidative systems. The effect of US radiation in Fenton-like oxidation process is discussed; the influence of dissolved iron (ferrous and ferric ions) on the oxidation kinetics of ZVI/H2O2 system is also investigated. On the basis of identifying intermediate and final products, the reaction pathways are proposed.

2.

Experimental section

2.1.

Materials

2.3.

The reduction of TBBPA (5 mg L1) was conducted by FeeAg nanoparticles (0.8 g L1) under US radiation (40 kHz and 100 W). Debromination experiments were performed in a chamber as shown in Fig. SM-1 (a) attached in Supplemental Material (SM). The detailed procedure of reduction experiment was reported in the literature (Luo et al., 2010).

2.4.

2.2. Synthesis and characterization of FeeAg bimetallic nanoparticles FeeAg bimetallic nanoparticles with core-shell structure were synthesized by reductive deposition of Ag on ZVI nanoparticles as described in the literature (Luo et al., 2010). Various analytical techniques including XRD, XPS and XRF were used to characterize the fresh and reacted (after reduction process) FeeAg bimetallic samples. X-ray diffraction (XRD) analyses of the samples were performed using ˚ ). Switzerland ARL X’TRA X-ray diffractometer (l ¼ 1.5418 A The metal oxidation states and surface atomic composition of FeeAg samples was examined via X-ray photoelectron spectroscopy (XPS, Thermo VG Scientific ESCALAB 250). X-ray fluorescence (XRF, Switzerland ARL Corporation) was used to measure the mass of Ag deposited on the surface of nanoiron.

Fenton-like oxidation experiment

To keep a constant temperature (25  1  C), the Fenton-like process was conducted in a chamber as presented in Fig. SM-1 (b). The reduction and oxidation experiments were carried out in the same vessel. In each bottle, the solution contained BPA and FeeAg nanoparticles after reduction. Its initial pH was adjusted to 3.0  0.1 with 0.1 M H2SO4 and 0.1 M NaOH solutions. The oxidation experiments were started by dropping H2O2 solutions into the mixture by a separatory funnel. The flow rate was controlled at 2 mg L1 min1 and lasted 10 min in the whole oxidation process. At the given reaction time intervals, 1 mL sample was withdrawn. 10 mL 1 M tert-butanol was immediately added into the sample as reaction inhibitor. Then the samples were filtered by a syringe filled with a little silanized glass wool. The concentrations of BPA and intermediates in the filtrate were measured by high-performance liquid chromatography (HPLC). If no specific instructions are given, initial pH of Fenton-like oxidation is 3.0  0.1, nanoparticles loading is 0.8 g L1 and Ag content in FeeAg composite material is 1 wt.%.

2.5. Tetrabromobisphenol A, bisphenol A and tert-butanol were obtained from SigmaeAldrich Company. H2O2 (30%, v/v) was purchased from Fisher Company. AgCl, FeSO4$7H2O, Fe2(SO4)3, H2SO4, NaOH, Na2SO3, 1,10-phenanthroline and ferrous ammonium sulfate were provided by Nanjing Chemical Company. HPLC-grade methanol and dichloromethane were purchased from Tedian Company and used without further purification. Milli-Q water was used throughout this study. The zero valent iron used was iron powder (Shenzhen Junye Nano Material Co., Ltd, >99.9％, <60 nm).

Reduction experiment of TBBPA

Analytical methods

The concentrations of TBBPA were analyzed via HPLC (Agilent 1200, USA), with a C18 reversed-phase column (150 mm  4.6 mm, 5 mm particles, Agilent, USA). Identification of reductive debromination products was performed by LC-ESI-MS/MS (Thermo LCQ Advantages, QuestLCQ Duo, USA) equipped with electrospray ionization with Beta Basic-C18 HPLC column (150 mm  2.1 mm id, 5 mm Thermo, USA). The specific operation conditions of HPLC and LC-ESI-MS/MS were provided in the literature (Luo et al., 2010). The HPLC operation conditions of BPA were as follows: the mobile phase was 30% water in methanol. At the detection wavelength of 226 nm and flow rate of 1.0 mL min1, the BPA retention time was 3.4 min. Since the oxidation products were so complex, a Thermo Finnigan Trace gas chromatography interfaced with a Polaris Q ion trap mass spectrometer (GC/ MS, Thermo, Finnigen, USA) equipped with DB-5 fused-silica capillary column (30 m  0.32 mm i.d, 0.25 mm film thickness) was used for analyzing the samples. Prior to GCeMS analysis, the samples were extracted with dichloromethane for three times. The extracted solution was dehydrated using anhydrous sodium sulfate and concentrated to 1 mL by rotary evaporation. After the solvent was blown away by the gentle nitrogen, trimethylsilylation was carried out at 50  C for 30 min using 0.2 mL of bis(trimethylsilyl)trifluoroacetamide (BSTFA). The initial temperature of the column oven was 40  C and following 1 min hold at this temperature, and then increased up to 300  C with a heating rate of 6  C min1.

1521

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 1 9 e1 5 2 8

Helium was used as the carrier gas. Mass spectrometric detection was operated with 70 eV electron impact (EI) mode. The degradation products were also analyzed by LC-ESI-MS/ MS. The injection volume was 20 mL, and the mobile phase was methanol to water (70:30). The MS analysis was conducted with negative-ionization mode with electrospray interface (ESI) source. Full scanning analyses were performed by scanning m/z range from 50 to 600 in profile mode. The dissolved bromide ion was determined using a Dionex ion chromatograph (IC, Dionex model ICS 1000) equipped with a dual-piston (in series) pump, a Dionex IonPac AS11-HC analytical column (4 mm  250 mm) and a Dionex DS6 conductivity detector. Suppression of the eluent was achieved with a Dionex anion ASRS 300 electrolytic suppressor (4 mm) in the auto suppression external water mode. The concentration of dissolved iron and ferrous ion was measured by the o-phenanthroline colorimetric method (l ¼ 510 nm, e ¼ 1.1  104 M1 cm1). Ferric concentration calculated by subtracting dissolved iron with ferrous concentration.

3.

Results and discussion

3.1.

Characterization

Detailed morphology and structure characterization of the FeeAg particles were presented in the literature (Luo et al., 2010). Fig. 1 is the XRD spectra of the fresh and reduced FeeAg particles. The XRD patterns of aged samples show peaks associated with iron oxides, indicating that surface of the used reductant is covered by a layer of oxide film which may form during the reduction process. The XPS survey scans

Fig. 1 e XRD of Fe-Ag bimetallic nanoparticles. (a) Fresh samples, (b) Reacted samples.

Table 1 e Composition of surface elements. Element

Ag 3d Fe 2p O 1s C 1s

XPS surface composition (atom %) Before

Reacted

0.57 40.59 36.11 22.73

0.53 24.69 53.88 20.91

of Fe and Ag over the surface of the FeeAg nanoparticles are described in Fig. SM-2. Table 1 shows elements (iron, silver and oxygen) composition of the samples before and after reaction. After reduction, iron oxide layer is adopted a larger amount of oxygen which indicated the structure of the particles with iron oxide in the outer of the sphere and ZVI in the inner.

3.2. Fenton-like oxidation of TBBPA in a heterogeneous FeeAg/H2O2 system In order to investigate the Fenton-like oxidation effect of TBBPA in a heterogeneous FeeAg/H2O2/US system, experiments were conducted in an 150 mL conical bottle with 5 mg L1 TBBPA and 0.8 g L1 FeeAg nanoparticles. The other conditions were as the same as the Fenton-like oxidation of BPA, which was mentioned in the Section 2.4. A typical reaction profile of TBBPA in FeeAg/H2O2/US system is given in Fig. 2. The result shows that more than 30% of TBBPA was removed in 2 min and then the degradation rate decreased. Up to 40% of TBBPA disappeared as the reaction was prolonged to 30 min. To gain more chemical structure information on the reaction products, reaction solution was subsequently extracted and possible polar products were silylated with BSFTA, and then subjected to GCeMS analysis. The detailed data are provided in the Supplemental Material. The results show that the main TBBPA oxidation products are brominated phenol species. In fact, there is strong

Fig. 2 e Degradation of TBBPA in aqueous solution by Fe-Ag/H2O2/US. [Fe-Ag bimetallic] [ 0.8 g LL1; flow rate of H2O2 [ 2 mg LL1 minL1, lasted 10 min; [TBBPA]0 [ 5 mg LL1; US conditions: 40 kHz, 100 W.

1522

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 1 9 e1 5 2 8

Fig. 3 e (a) Temporal disappearance of TBBPA and appearance of by-products in aqueous solution during FeeAgeUS reduction treatment. [Fe-Ag bimetallic] [ 0.8 g LL1, Ag content (wt %) [ 1%, and [TBBPA]0 [ 5 mg LL1. (b) Temporal change of BPA and its hydroxylated products concentration in the solution during Fe-Ag/H2O2/US oxidation treatment. [Fe-Ag bimetallic] [ 0.8 g LL1; flow rate of H2O2 [ 2 mg LL1 minL1, lasted 10 min; [BPA]0 [ 4.64 mg LL1; US conditions: 40 kHz, 100 W.

evidence that halogen, when placed in the organic moleculecarbon group, could increase the bio-toxicity of the organic compound. Bromine, due to its weaker electronegativity, could be considered better leaving groups and hence should be more toxic (Lag et al., 1994; DeWeese and Schultz, 2001; Huang et al., 2007). Several studies have presented the results from ecotoxicologic investigations of the brominated aromatic compounds, including brominated benzenes, brominated phenols and indoles, 2-bromo- hydroquinone and  ska, 1998; Bruchajzer et al., 2002; Reineke et al., so on (Szyman 2006). Therefore, the Fenton-like oxidation of TBBPA in a heterogeneous FeeAg/H2O2 system is unsuitable and inadequate. Considering that the primary step in the degradation of halogenated compounds is the removal of halides from the

aromatic ring, such a T-SRO process mentioned above is proposed for the treatment of TBBPA.

Fig. 4 e Degradation of BPA in aqueous solution by Fe-Ag/ H2O2 with and without ultrasonic radiation. [Fe-Ag bimetallic] [ 0.8 g LL1; flow rate of H2O2 [ 2 mg LL1 minL1, lasted 10 min; [BPA]0 [ 5 mg LL1; US conditions: 40 kHz, 100 W.

Fig. 5 e Degradation of BPA in aqueous solution by Fe-Ag/ H2O2/US with and without addition of dissolved iron. [Fe-Ag bimetallic] [ 0.8 g LL1; flow rate of H2O2 [ 2 mg LL1 minL1, lasted 10 min; [BPA]0 [ 5 mg LL1; [Fe2D] [ 1.26 mg LL1, [Fe3D] [ 0.37 mg LL1; US conditions: 40 kHz, 100 W.

3.3. TBBPA debromination by FeeAg bimetallic nanoparticles Fig. 3 (a) shows the time course profiles for disappearance of TBBPA and appearance of by-products in aqueous solution over FeeAg bimetallic nanoparticles coupled with US radiation. The initial TBBPA concentrations and catalysts amount were chosen based on results of previous reduction experiments (Luo et al., 2010). In neutral conditions (pH ¼ 6.94), FeeAg bimetallic nanoparticles completely degraded TBBPA (5 mg L1) in 20 min coupled with US radiation.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 1 9 e1 5 2 8

Along with the decrease of TBBPA concentration, almost 100% of TBBPA was gradually debrominated to lowly brominated compounds. The released bromine was in hydrogen bromine form. TBBPA might be transformed into tri-BBPA and di-BBPA, then tri-BBPA and di-BBPA were dehalogenated to

1523

mono-BBPA, at last, BPA was generated by the debromination reaction of these three intermediates. The BPA concentration increased stably and continuously in the all 70 min and reached 4.64 mg L1 at the end of reduction process. The carbon mass balance at the end of the experiment (calculated

Fig. 6 e (a) Total ions chromatogram of BPA in the solution during Fe-Ag/H2O2/US oxidation treatment. (b) LC-MS/MS analysis of BPA degradation intermediates.

1524

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 1 9 e1 5 2 8

as the sum of all organic species measured) was approximately 94% of the calculated initial TBBPA concentration. We suspect that minor losses may have occurred during filtering through the silanized glass wool.

3.4. Fenton-like oxidation of BPA in a heterogeneous FeeAg/H2O2/US system Experiments of BPA control, FeeAg nanoparticles alone and H2O2 alone were also carried out in 150 mL conical bottles under US radiation, and the initial concentration of BPA was 5 mg L1. The results in Fig. 3 (b) indicate that no obvious removal of BPA was observed in the control and FeeAg nanoparticles alone experiments throughout 30 min. In the H2O2 alone degradation experiment, 15% degradation of BPA was observed after 30 min reaction. It means that the hydroxyl radicals generated by ultrasounds radiation could destroy the BPA with adding H2O2. The effect of US radiation will be discussed in detail in Section 3.4.1 As shown in Fig. 3 (b), after 2 mg L1 min1 of H2O2 (lasted 10 min) was added to the BPA solution debrominated from TBBPA, BPA concentration rapidly declined from 4.64 to 0.11 mg L1 over 30 min with removal efficiency of 99.2%. Concentration change profile of BPA derivatives is also given in Fig. 3 (b) in terms of the HPLC peak areas of the corresponding

compounds versus the charge applied to the system. The typical chromatogram of BPA during the oxidation in 30 min is depicted in Fig. SM-4. It can be seen that peak I corresponded to the parent compound BPA and two main intermediates peaks appeared (II and III). Peak II had a retention time of 2.7 min, which increased at first but decreased quickly after radiation for 6 min. Peak III had a retention time of 2.5 min, which appeared after reaction for 8 min and decreased afterward. These two intermediates were the products for the Fenton-like oxidation of BPA, whose decrease implied that further oxidation of BPA continued in solution.

3.4.1. Effect of ultrasonic radiation on BPA Fenton-like oxidation It is all well known that introduction of US can evidently improve pollutant degradation in Fenton system (Ioan et al., 2007; Namkung et al., 2008). When US were introduced into Fenton-like reaction, the mechanism would be more complex. The results of oxidation experiments in the presence and absence of US radiation are compared and presented in Fig. 4. Under US radiation, an obvious increase of BPA degradation efficiency in the solution was observed. Thus, there was a synergistic effect between US and Fenton-like reactant, which could improve the degradation of BPA in the heterogeneous FeeAg/H2O2/US system.

Fig. 7 e GC/MS analysis of BPA degradation intermediates and final products.

1525

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 1 9 e1 5 2 8

In general, introduction of US radiation into a heterogeneous system could produce both chemical and physical effect. The possible reasons for synergistic effect between US and Fenton-like reactant are as follows. Firstly, US could increase the
OH radical concentration by US cavitation. Guo and Feng (2009) described that the US elimination of BPA in aqueous solution is mainly attributed to the
OH radical oxidation. BPA was decomposed by
OH radicals principally at the bubble-bulk solution interface in absence of FeeAg nanoparticles. In FeeAg/H2O2/US Fenton-like oxidation experiments, US cavitation increased the amount of
OH radical and therefore accelerated the degradation reaction. Secondly, US shock wave could enhance dissolution of ferrous ion from iron surface. The main reason was ascribed to the removal or destruction of passivation films on the metal surface by cavitation effects (Tomlinson, 1990; Namkung et al., 2008). Thus, the increase of iron concentration in the solution might lead to enhance the oxidation of BPA. In addition, US radiation can transform BPA slightly with H2O2, as shown in Fig. 3 (b).

3.4.2. Effect of dissolved iron (Fe2þ and Fe3þ) on BPA Fenton-like oxidation To demonstrate the role of dissolved iron, ferrous ion and ferric ion were added to the FeeAg/US system with 5 mg L1 BPA. The concentrations of Fe2þ and Fe3þ in the solution after debromination were 1.26 mg L1 and 0.37 mg L1, respectively. These values were taken as the dissolved iron amount used for addition in the oxidation experiments for comparison. Fig. 5 shows that BPA decayed during the treatment by only FeeAg and FeeAg with dissolved iron Fenton-like process. In the two cases, BPA concentration was under the detection limit after 20 min, indicating that the oxidation capacity of the two systems (FeeAg and FeeAg þ dissolved iron) showed less difference throughout the reaction. However, in the first 2 min, 56% decomposition of BPA was achieved in the presence of Fe2þ and Fe3þ ion while only 22% BPA removal was observed in the absence of Fe2þ and Fe3þ ion. Zhou et al. (2008) reported that the two-stage kinetic of 4chlorophenol degradation in a heterogeneous ZVI/H2O2 system

Table 2 e Identification of the intermediates of BPA during the Fenton-like oxidation by GC/MS. Product

Rt (min)

m/z

Name

A

3.28

90

Oxalic acid

B

4.89

116

Maleic acid

C

5.67

108

p-Benzoquinone

D

6.97

94

phenol

E

7.45

90

2-hydroxypropanoic acid

Molecular structure HO

O

O

OH

O O HO

OH

O

O

OH

OH OH

F

13.01

92

propane-1,2,3-triol

G

14.72

134

4-isopropenylphenol

H

15.75

110

p-hydroquinone

I

19.20

136

4-hydroxyacetophenone

J

30.52

228

BPA

K

35.53

244

BPA-o-catechol.

OH

OH

HO

O

OH

OH

1526

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 1 9 e1 5 2 8

Scheme 1 e Main Fenton-like oxidation reaction pathway of BPA.

was composed of an initial slow degradation stage and a followed rapid degradation stage. In the first stage, the decomposition of H2O2 occurred on/near the ZVI surface, dissolving iron to ferrous, and Fenton reaction occurred concomitantly on/near surface in the presence of ferrous and H2O2. Based on this, the addition of dissolved irons species could increase the degradation rates of BPA, especially in the initial stage. As the reaction progresses, ferrous ion and ferric ion leached from iron surface and their concentrations increased gradually. Accordingly, addition of dissolved iron would not affect the removal efficiency in the entire Fenton-like process.

3.5. Identification of Fenton-like oxidation products of BPA The analyses of main BPA products formed during the Fentonlike oxidation were carried out using LC-MS/MS and GCeMS. The intermediates and products were identified by the analysis of mass spectra obtained from LC-MS/MS and/or GCeMS applications and by the comparison with the library data of NIST. Fig. 6 (a) is the total ion chromatogram of BPA solution throughout oxidation reaction, which shows three peaks with retention time at 12.69 min, 9.78 min and 8.23 min, respectively. Fig. 6 (b) is the mass spectra for them. It can be seen that the compound at 12.69 min was BPA (I) with m/z 227. The peaks labeled with II (9.78 min, m/z ¼ 243) and III (8.23 min, m/z ¼ 257) corresponded to the monohydroxylated BPA product and the dihydroxylated BPA product, respectively. The former gave base ion at 243 m/z on ESI negative mode detection, was BPA-ocatechol. This compound also showed other ion at 241 m/z, suggesting that BPA could be transformed via the following reversible equilibration between BPA-o-catechol (MW ¼ 244) and BPA-o-quinone (MW ¼ 242). These products obviously have a greater hydrophilicity than BPA, so their retention times were shorter under the HPLC condition in our experiment.

Other products in the Fenton-like oxidation, including 4isopropenylphenol and the open ring products, were detected by GCeMS (Fig. 7). This result is similar to that reported by Go¨zmen et al. (2003) in their studies on the electro-Fenton oxidation of BPA. After all the peaks in the chromatograms were carefully examined, up to 10 compounds were identified as possible intermediates. Detailed data are listed in Table 2. The intermediates with benzene ring could be further oxidized into small organic molecules, with the retention times at 3.28, 4.89, 7.47 and 13.01 min, most of which were organic acids, such as oxalic acid, and so on. However, some products such as HCOOH were not observed due to their low levels and unavoidable loss during the sample preparation course. Based on the intermediates and the results obtained by other researchers (Katsumata et al., 2004; Guo and Feng, 2009), the possible pathways of BPA oxidation were represented in Scheme 1. The expatiation of these pathways is presented in the Supplemental Material. In addition, the brominated compounds were not present in the Fenton-like degradation products. After reduction and oxidation processes, the amount of dissolved Br measured by IC was 2.5 mg L1 and 2.3 mg L1, respectively, which showed that the variation of Br concentration in the solution was negligible in Fenton-like experiments. Based on analysis described above, we referred that the debromination of TBBPA was to release HBr and the bromination of decomposition products by bromine that was not formed even in the present of
OH radicals. In other words, the bromine generated from TBBPA reduction did not lead to the formation of secondary brominated products by interaction with the primary oxidation products of BPA.

4.

Conclusions

The degradation of TBBPA in aqueous solution was investigated by Fe-Ag reduction and Fenton-like oxidation under US

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 1 9 e1 5 2 8

radiation. TBBPA was completely debrominated to BPA after 70 min over FeeAg bimetallic nanoparticles. For BPA, removal efficiency of 99.2% was achieved after 30 min under the optimum conditions. The US radiation enhanced the degradation rate of BPA as compared to Fenton-like only. In addition, dissolved iron was found to increase the rate of the first 2 min oxidation reactions, however, the existence of Fe2þ and Fe3þ did not affect entire Fenton-like degradation of BPA. This work shows that it is feasible for T-SRO treatment to completely decompose TBBPA in aqueous solutions. GCeMS and LC-MS/MS techniques were used to identify the oxidation intermediates in order to gain a deeper insight of the reaction mechanism. The main intermediates include monohydroxylated BPA, 4-isopropenylphenol, p-hydroquinone, 4-(1-hydroxy-1-methyl-ethyl)-phenol, 4-hydroxyacetophenone and phenol. Besides,
OH radical-mediated oxidation was found to be the major destruction pathway during BPA decomposition. Future work will need to be done to improve the potential mineralization of BPA, as well as the fate of the organic molecules to confirm mineralization.

Acknowledgements The authors greatly acknowledge the National Natural Science Foundation of China (20707009), National Major Project of Science & Technology Ministry of China (NO. 20082X07421-002) for financial support, and National Natural Science Foundation of China (50938004).

Appendix. Supplementary data Supplementary data associated with the article can be found in online version, at doi:10.1016/j.watres.2010.10.039.

references

Barontini, F., Cozzani, V., Marsanich, K., Raffa, V., Petarca, L., 2004. An experimental investigation of Tetrabromobisphenol A decomposition pathways. Journal of Analytical and Applied Pyrolysis 72 (1), 41e53. Bremner, D.H., Burgess, A.E., Houllemare, D., Namkung, K.C., 2006. Phenol degradation using hydroxyl radicals generated from zero-valent iron and hydrogen peroxide. Applied Catalysis B-Environmental 63 (1e2), 15e19. Bruchajzer, E., Szymanska, J.A., Piotrowski, J.K., 2002. Acute and subacute nephrotoxicity of 2-bromophenol in rats. Toxicology Letters 134 (1e3), 245e252. Cwiertny, D.M., Bransfield, S.J., Livi, K.J.T., Fairbrother, D.H., Roberts, A.L., 2006. Exploring the influence of granular iron additives on 1,1,1-trichloroethane reduction. Environmental Science & Technology 40 (21), 6837e6843. DeWeese, A.D., Schultz, T.W., 2001. Structure-activity relationships for aquatic toxicity to Tetrahymena: halogensubstituted aliphatic esters. Environmental Toxicology 16 (1), 54e60. de Wit, C.A., 2002. An overview of brominated flame retardants in the environment. Chemosphere 46 (5), 583e624.

1527

˚ , Jakobsson, E., Eriksson, J., Rahm, S., Green, N., Bergman, A 2004. Photochemical transformations of Tetrabromobisphenol A and related phenols in water. Chemosphere 54 (1), 117e126. Go¨zmen, B., Oturan, M.A., Oturan, N., Erbatur, O., 2003. Indirect electrochemical treatment of bisphenol A in water via electrochemically generated Fenton’s reagent. Environmental Science & Technology 37 (16), 3716e3723. Guo, Z.b., Feng, R., 2009. Ultrasonic irradiation-induced degradation of low-concentration bisphenol A in aqueous solution. Journal of Hazardous Materials 163 (2e3), 855e860. ˚ , Jenssen, D., 1999. Helleday, T., Tuominen, K.L., Bergman, A Brominated flame retardants induce intragenic recombination in mammalian cells. Mutation Research 439 (2), 137e147. Huang, C.P., Wang, Y.J., Chen, C.Y., 2007. Toxicity and quantitative structure-activity relationships of nitriles based on Pseudokirchneriella subcapitata. Ecotoxicology and Environmental Safety 67 (3), 439e446. Ioan, I., Wilson, S., Lundanes, E., Neculai, A., 2007. Comparison of Fenton and sono-Fenton bisphenol A degradation. Journal of Hazardous Materials 142 (1e2), 559e563. Kaiser, J., 2000. Endocrine disrupters: panel cautiously confirms low-dose effects. Science 290 (5492), 695e697. Katsumata, H., Kawabe, S., Kaneco, S., Suzuki, T., Ohta, K., 2004. Degradation of bisphenol A in water by the photo-Fenton reaction. Journal of Photochemistry and Photobiology A: Chemistry 162 (2e3), 297e305. Lag, M., Omichinski, J.G., Dybing, E., Nelson, S.D., Soderlund, E.J., 1994. Mutagenic activity of halogenated propanes and propenes: effect of bromine and chlorine positioning. Chemico-Biological Interactions 93 (1), 73e84. Luo, S., Yang, S.G., Wang, X.D., Sun, C., 2010. Reductive degradation of Tetrabromobisphenol A over iron-silver bimetallic nanoparticles under ultrasound radiation. Chemosphere 79 (6), 672e678. Mackenzie, K., Kopinke, F.D., 1996. Debromination of duroplastic flame-retarded polymers. Chemosphere 33 (12), 2423e2430. Mantha, R., Biswas, N., Taylor, K.E., Bewtra, J.K., 2002. Removal of nitroaromatics from synthetic wastewater using twostep zero-valent iron reduction and peroxidase-catalyzed oxidative polymerization. Water Environmental Research 74 (3), 280e287. Mantha, R., Taylor, K.E., Biswas, N., Bewtra, J.K., 2001. A continuous system for Fe0 reduction of nitrobenzene in synthetic wastewater. Environmental Science & Technology 35 (15), 3231e3236. Meerts, I.A.T.M., van Zanden, J.J., Luijks, E.A.C., van Leeuwen˚ , Brouwer, A., 2000. Bol, I., Marsh, G., Jakobsson, E., Bergman, A Potent competitive interactions of some brominated flame retardants and related compounds with human transthyretin in Vitro. Toxicological Sciences 56 (1), 95e104. Monserrate, E., Haggblom, M.M., 1997. Dehalogenation and biodegradation of brominated phenols and benzoic acids under iron-reducing, sulfidogenic, and methanogenic conditions. Applied and Environmental Microbiology 63 (10), 3911e3915. Namkung, K.C., Burgess, A.E., Bremner, D.H., Staines, H., 2008. Advanced Fenton processing of aqueous phenol solutions: a continuous system study including sonication effects. Ultrasonics Sonochemistry 15 (3), 171e176. ¨ berg, K., Warman, K., O ¨ berg, T., 2002. Distribution and levels of O brominated flame retardants in sewage sludge. Chemosphere 48 (8), 805e809. Oh, S.Y., Chiu, P.C., Kim, B.J., Cha, D.K., 2003. Enhancing Fenton oxidation of TNT and RDX through pretreatment with zerovalent iron. Water Research 37 (17), 4275e4283.

1528

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 1 9 e1 5 2 8

Oh, S.Y., Chiu, P.C., Kim, B.J., Cha, D.K., 2005. Zero-valent iron pretreatment for enhancing the biodegradability of RDX. Water Research 39 (20), 5027e5032. Orth, W.S., Gillham, R.W., 1996. Dechlorination of trichloroethene in aqueous solution using Fe0. Environmental Science & Technology 30 (1), 66e71. Pulgar, R., Olea-Serrano, F., Villalobos, M., Rivas, A., Metzler, M., Pedraza, V., Olea, N., 1998. The estrogenicity of bisphenol A-related diphenylalkanes with various substituents at the central carbon and the hydroxy groups. Environmental Health Perspectives 106 (3), 167e174. Reineke, N., Biselli, S., Franke, S., Francke, W., Heinzel, N., Hu¨hnerfuss, H., Iznaguen, H., Kammann, U., Theobald, N., Vobach, M., Wosniok, W., 2006. Brominated indoles and phenols in marine sediment and water extracts from the north and baltic seas-concentrations and effects. Archives of

Environmental Contamination and Toxicology 51 (2), 186e196. Sellstro¨m, U., Jansson, B., 1995. Analysis of Tetrabromobisphenol A in a product and environmental samples. Chemosphere 31 (4), 3085e3092.  ska, J.A., 1998. Hepatotoxicity of brominated benzenes: Szyman relationship between chemical structure and hepatotoxic effects in acute intoxication of mice. Archives of Toxicology 72 (2), 97e103. Tomlinson, W.J., 1990. Effect of ultrasonically induced cavitation on corrosion. In: Mason, T.J. (Ed.), Advances in Sonochemistry, Vol. 1. Jai Press, London, pp. 173e195. Zhou, T., Li, Y.Z., Ji, J., Wong, F.S., Lua, X.H., 2008. Oxidation of 4-chlorophenol in a heterogeneous zero valent iron/H2O2 Fenton-like system: kinetic, pathway and effect factors. Separation and Purification Technology 62 (3), 551e558.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 2 9 e1 5 3 8

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Extracellular polymeric substances diversity of biofilms grown under contrasted environmental conditions Monique Ras a, Dominique Lefebvre a, Nicolas Derlon b,c,d, Etienne Paul b,c,d, Elisabeth Girbal-Neuhauser a,* a

LBAE, Laboratoire de Biologie applique´e a` l’Agro-alimentaire et a` l’Environnement, Institut Universitaire de Technologie, Universite´ Paul Sabatier Toulouse III, 24 Rue d’Embaque`s, 32000 Auch, France b Universite´ de Toulouse; INSA, UPS, INP; LISBP, 135 Avenue de Rangueil, F-31077 Toulouse, France c INRA, UMR792 Inge´nierie des Syste`mes Biologiques et des Proce´de´s, F-31400 Toulouse, France d CNRS, UMR5504, F-31400 Toulouse, France

article info

abstract

Article history:

Extracellular Polymeric Substances (EPS) analysis was undertaken on three biofilms

Received 2 September 2010

grown under different feeding conditions and offering diverging microbial activities and

Received in revised form

structural characteristics. EPS were extracted by a multi-method protocol including soni-

15 November 2010

cation, Tween and EDTA treatments and were characterized by size exclusion chroma-

Accepted 15 November 2010

tography (SEC). Tween and sonication extracts presented higher EPS size diversity

Available online 24 November 2010

compared to EDTA extracts. EPS size diversity also increased with microbial functions

Keywords:

biofilms presenting autotrophic activity. Another specific size cluster (180 kDa) occurred in

Microbial biofilm

Tween extracts provided from the mechanically stable biofilms. Such specific EPS appear

Autotrophic

as potential indicators for describing microbial and structural properties of biofilms.

Extracellular polymeric substances

This study brings new elements for designing EPS fractionation and shows that size

Size distribution

distribution analysis is an interesting tool to relate EPS diversity with macro-scale char-

Extraction strategy

acteristics of biofilms.

within the biofilms and a specific 25e50 kDa cluster was identified only in extracts from

ª 2010 Elsevier Ltd. All rights reserved.

1.

Introduction

Biofilms are described in literature as fixed micro-organisms on an interface and immobilized in a matrix of extracellular polymeric substances (or EPS) of microbial origin. The stable environment offered by the EPS matrix cradles the development of a large span of microbial communities of which several can be deleterious. The microbial heterogeneity of biofilms can also be of great interest in the environmental sector since such concentrated and diversified microbial activities can be beneficially exploited for treating organic and inorganic water pollutants. However, municipal wastewater

treatment facilities generally use suspended floc forming biomasses which are often washed out from the system (Liu et al., 2004; Matsumoto et al., 2007) and hence experience low microbial diversity functions. Fixed biomass such as biofilms can prevent such losses by retaining bacterial diversity inside the system and particularly slow-growing bacterial populations, such as nitrifiers. Such configurations can hence increase the treatment efficiency. The EPS matrix is often stated as consolidating material for the entire biofilm. Indeed, the extracellular compartment can reach 98% of the total organic carbon fraction of biofilms (Jahn and Nielsen, 1998). EPS compounds are excreted by the

* Corresponding author. Tel.: þ33 5 62 61 28 13; fax: þ33 5 62 61 63 01. E-mail address: [email protected] (E. Girbal-Neuhauser). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.11.021

1530

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 2 9 e1 5 3 8

microbial population, but can also result from natural cell lysis or from hydrolytic activities. A wide variety of polymers are reported within the matrix, where a major proportion is attributed to proteins and polysaccharides, while lipids and nucleic acids are rather found in minor proportions (Azeredo et al., 1999; Jahn and Nielsen, 1995). The influence of environmental conditions on the composition of EPS compounds has already been suggested in literature (Branda et al., 2005). Regarding the use of carbon and nitrogen elements for EPS production, the carbon/nitrogen ratio of the influent is liable to impact the type of produced EPS, i.e. carbohydrate and protein production (Durmaz and Sanin, 2001; Li et al., 2008). In addition, the carbon/nitrogen ratio can specify the microbial ecology of the biofilm (Ohashi et al., 1995), by promoting either heterotrophic growth (high ratio) or autotrophic microorganisms (low ratio). Regarding the biochemical responses to environmental and microbial parameters, characterizing the EPS fraction of biofilms could thus be a relevant procedure for describing relations between EPS and biofilm structure. In literature, studies undertaken on molecular characterization of EPS in biofilms are few. Under axenic conditions, exopolysaccharides and more specifically uronic acids containing polymers extracted from biofilms are described as essential for providing the matrix framework through strong anionic interactions (Chen and Stewart, 2002; Davies et al., 1993). For multi-species biofilms, the conditions are even more complex since a wide range of other molecular interactions have to be considered. Proteins are characterized by ionic, hydrophobic and neutral amino-acids and a large range of chemical interactions (electrostatic, hydrophobic and low energy hydrogen bonds) are able to link proteins to the biological matrix (Mayer et al., 1999). Proteins also include functionally active enzymes which take part in the production and degradation of the matrix. Therefore, inherent chemical properties of EPS and especially proteins can offer qualitative information on both physical and dynamic properties of the biofilm. However, the structural heterogeneity and complex functional properties in environmental associated biofilms make EPS characterization somewhat difficult. Several analytical methods including physical and chemical techniques are used (Denkhaus et al., 2007) but with care depending on the aim of investigation as well as the type of studied biofilm. Microscopic and optical methods which involve EPS staining techniques are not always appropriate for visualizing these components in thick and complex biofilms due to light attenuation or probe penetration problems. Infrared spectroscopy is also widely used in biofilm analysis with similar limitations relative to the penetration capacity of the IR radiations (Boualam et al., 2002). Considering these technical restrictions, molecular characterization of complex biofilms can be achieved by extracting the EPS from the biofilm and then characterizing the soluble extract by chromatography or electrophoresis separation methods. Although widely used on activated sludge samples (Comte et al., 2007; Garnier et al., 2005), this molecular scale investigation strategy was never applied for biofilm EPS characterization. Molecular weight (MW) distributions of extracted EPS can offer global characteristics of the sample and has been suggested as a useful tool for fingerprint identification. Garnier

et al. (2005) evidenced different MW profiles depending on the origin of activated sludge. Authors showed that proteins where generally found in the high MW fractions (10e600 kDa) while sugars were rather found in the lower MW fractions (1 kDa). However, studying the size distribution of EPS in complex bacterial aggregates reveals to be tricky since such analysis implies prior extraction methods which can affect not only the proportion of extracted EPS (Ras et al., 2008a; Zhang et al., 1999) but also the qualitative aspect of these polymers (Comte et al., 2007; Simon et al., 2009). The present paper explores EPS size distributions within biofilms in order to figure out specific molecular characteristics which could explain particular biofilm biological and/or physical properties. In order to validate such an approach, the investigated biofilms were grown under contrasted environmental conditions to promote diverging microbial activities within each biofilm. A multi-method protocol previously described for extracting EPS from activated sludge (Ras et al., 2008a) was used to sample EPS compounds from the biofilms. This protocol, based on mechanical, hydrophobic and ionic extraction methods, offers a globally diversified EPS extract which can be consistent of the studied biofilms. The distribution of EPS contents as well as EPS molecular weight profiles were investigated in order to relate specific molecular EPS characteristics to biofilm growth conditions and/or microbial populations. The impact of EPS extraction procedures on this molecular fingerprint diagnosis was also considered. According results are expected to help improve knowledge on biofilm growth control which is lacking in the wastewater and water distribution sectors.

2.

Methods

2.1.

Experimental setup

Three biofilms were grown in hydrodynamic controlled Couette Taylor reactors as described by Coufort et al. (2007). For a fixed gap between the two concentric cylinders, the rotational speed of the inner cylinder was fixed in order to have a wall shear stress of 0.5 Pa during the growth period. Biofilms grew on 25 polyethylene plastic plates (100  50  5 mm) distributed around the external cylinder.

2.2.

Biofilm growth conditions

A mixed carbon source composed of ethanol, propionic acid, glucose and sodium acetate was used as organic substrate for the development of the biofilms. Reactors were inoculated with conventional activated sludge sampled from the aeration tank of a local municipal wastewater treatment plant. Two biofilms were developed under organic substrate-limiting conditions and with a constant surface loading rate of 2.5 g COD m2 d1 (COD: Chemical Oxygen Demand). In order to obtain either a heterotrophic biofilm (B1) or a mixed autotrophic/heterotrophic biofilm (B2), the feed diverged in COD/ NH4eN ratios. The feed for B1 was fixed at 73 g COD g1 NeNH4 (9.5 mg NH4eN L1) and the feed for B2 at 4 g COD g1 NH4eN (175 mg NH4eN L1). For these two cases, the oxygen concentration in the bulk liquid was kept constant at a value

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 2 9 e1 5 3 8

closed of the oxygen saturation concentration. A third biofilm (B3) was grown under a high substrate loading rate of 25 g COD m2 d1 and with a ratio of 4 g COD g1 NH4eN (175 mg NH4eN L1). In this case, the oxygen concentration was kept constant between 6 and 7 mg O2 L1, inducing oxygen-limiting growth conditions. During the overall characterization period, ammonia   (NHþ 4 ), nitrite (NO2 ), nitrate (NO3 ) and COD were measured daily in the inlet and in the outlet of the Couette Taylor reactors. The ammonia concentration was measured using the Nessler method, nitrite and nitrate concentrations were determined by spectrometry and the COD was obtained with the method based on the oxidation by potassium dichromate (Standard Methods, 1995). Biofilm at steady state was defined as a biofilm characterized by stable COD removal, nitrification and denitrification rates. Stable COD removal, nitrification and denitrification and thus steady state were usually reached after 60 days of biofilm development.

2.3.

Biofilm characterization

The average biofilm thickness was measured by image analysis as described in Coufort et al. (2007). Mean accumulated mass was measured after biofilm detachment from the polyethylene plastic plates by gentle scraping and suspension in the liquid reactor. Detached biomass was then recovered by centrifugation (1500g; 15 min) and measured in terms of Suspended Solids and Volatile Suspended Solids concentration (g VSS L1) according to the standard procedures (Standard Methods, 1995). Total COD removal, nitrification efficiency and denitrification efficiency were evaluated by comparing the inlet and outlet values of COD, ammonium, nitrate and nitrite concentrations.

2.4.

EPS extraction by the multi-method protocol

Bound EPS were extracted according to the previously described multi-method protocol validated on activated sludge samples (Ras et al., 2008a). Biofilm samples were centrifuged (10 000g; 20 min) and pellets were washed twice in Phosphate Buffer Saline (PBS) pH 7. Each biofilm sample was subdivided in three 10 mL aliquots containing around 5 g VSS L1 for triplicate extractions. One protocol involved three extraction methods in sequence: sonication (3  2 min in PBS), Tween (0.25% in PBS, 1 h) and then EDTA (2% in Tris-HCl 0.3 mol L1, pH 8.5, 1 h), with intermediate centrifugation steps (10 000g; 20 min). EPS extracts were measured the same day for protein and polysaccharide contents as well as for G6P-DH activity, and then stored at 20  C for further analysis. The protocol extraction efficiency was evaluated after repeating three times the protocol sequence on the same biofilm sample. The decrease of the protein content recovered after each protocol sequence fitted an exponential curve as described in Ras et al. (2008a). The total protein content in biofilm extracts obtained by repeating the extraction protocol reached 246 mg eq. BSA g1 VSS, with 116 mg eq. BSA g1 VSS obtained by applying the protocol only once (results not shown). The extraction yield performed on the biofilm was

1531

hence 47%, which is similar to yields obtained on activated sludge samples (Ras et al., 2008a). Protein measurements were performed on all soluble extracts from B1, B2 and B3 biofilms using the Bicinchoninic Acid (Smith et al., 1985) or BCA reagent (SigmaeAldrich), according to Ras et al. (2008b) procedure. This quantification method was chosen according to its better tolerance towards chemicals used during extraction compared to modified Lowry method (Ras et al., 2008a). Bovine Serum Albumin (BSA) was used as standard. Polysaccharide concentrations were determined using the Anthrone method (Dreywood, 1946). Glucose was used as standard. Each measurement was undertaken on duplicate samples.

2.5.

Cell lysis control

The activity of the intracellular G6P-DH was measured according to Ras et al. (2008a). Enzyme substrate solution was prepared with 0.2 M Tris-HCl pH 8.5, 0.2 M 2-mercaptoethanol (Acros), 0.0005 M Nicotine Adenine Dinucleotide (NAD, Acros) and 0.01 M D-glucose-6-phosphate (Fluka). Enzyme activity was evaluated after incubating 200 mL of sample with 800 mL of the enzyme substrate solution at room temperature and measuring NADH production at 340 nm during 30 min G6P-DH activity was expressed as units (U) per mg of VSS, one unit corresponding to the number of nmol of NADH produced per min in the assay conditions. In order to correlate the G6P-DH activities measured in the extracts or in the whole biofilms with a number of lysed cells, a preliminary calibration was performed using Cupriavidus necator DSM 545 suspensions. C. necator was cultured as previously described by Ramsay et al. (1990). Briefly, the culture was first grown for 12 h in a liquid Nutrient Broth medium (Merck) under agitation (200 rpm) and at 30  C. 10 mL of the suspension was then inoculated to 150 mL of a Mineral Medium supplemented with glucose (10 g L1) and incubated for 12 h at 30  C at 100 rpm. Every 4 h, 10 mL of a culture medium sample was filtered on a cellulose 0.2 mm filter then dried and weighed in order to determine the total biomass concentration. Bacterial population was also evaluated by serial dilution of the samples and numeration on TCA agar plates: a value of 4.08 106 g of dry biomass per 106 cells was determined. After 12 h, bacteria were harvested by centrifugation (10 000g, 10 min) and resuspended in a equal volume of TES buffer (Tris-HCl 50 mM pH 8, EDTA 0.29 g L1, saccharose 25%). Cell lysis was then induced by adding 50 mL of a lysozyme solution at 47 000 U/mg (SigmaeAldrich) to 1 mL of the TES bacterial suspension. After 1 h at 37  C, numeration was performed on the suspension and the G6P-DH measured on the supernatant. Data obtained on three independent samples indicated that 0.2 U were released per 106 disrupted cells, also corresponding to 49,020 U per g of dry cells.

2.6.

Chromatography analysis

Chromatography was performed using a high-performance liquid chromatography system (AKTA Purifier, GE Healthcare) equipped with a 1 mL injection loop, a UV detector and a conductivity cell. Size exclusion chromatography (SEC) used

1532

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 2 9 e1 5 3 8

a 24 mL sepharose gel filtration column (Superose 6, GE Healthcare). Elution was carried out at room temperature using PBS at constant 1 mL min1 flow rate. According to manufacturer information, size fractionation is performed between 5 and 5000 kDa. Calibration of the column was undertaken by injecting ten different size protein standards (high and low molecular weight calibration kits GE Healthcare: aprotinin (6500 Da), ribonuclease (13 700 Da), chymotrypsin (25 000 Da), carbonic anhydrase (29 000 Da), bovine serum albumin (67 000 Da), conalbumin (75 000 Da), aldolase (158 000 Da), catalase (232 000 Da), ferritin (440 000 Da) and thyroglobulin (669 000 Da)). The calibration curve revealed the following equation: log (MW) ¼ 0.2939V þ 9.8481 with Molecular Weight (MW) expressed in Da and the elution volume V in mL. The total exclusion volume was determined after injection of Blue Dextran 2000 (GE Healthcare, 2$106 Da) and was found at 8 mL. Chromatogram profiles were recorded with UNICORN 5.1 software (GE Healthcare). Peak retention times and peak areas were directly calculated and delivered by the program.

3.

Results

3.1.

Global characteristics of the developed biofilms

Three biofilms (B1, B2 and B3) were developed under different feeding conditions in terms of COD/NH4eN ratios as well as surface organic loading rates. These experimental conditions were chosen according to previous results which reported the influence of growth conditions on biofilm structure and biological activity (Coufort et al., 2007; Derlon et al., 2008; Wijeyekoon et al., 2004). The B1 biofilm grew under a high COD/NH4eN ratio of 73 (nitrogen limitation) while B2 and B3 biofilms grew under a low COD/NH4eN ratio of 4 (excess nitrogen). This carbon/nitrogen ratio varied by modifying ammonium concentration in the feed. Both B1 and B2 biofilms were grown under a low surface loading rate of 2.5 g COD m2 per day, while B3 biofilm received a high surface loading rate of 25 g COD m2 per day. Physical measurements (Table 1) and microscopic observations (Fig. 1) of all biofilms were undertaken when steady state COD and nitrogen removal rates were reached. B1 and B3 biofilms were characterized by a particularly thick and filamentous structure in opposition to the thin and denser aspect of

Fig. 1 e Microscopic side views of B1 (A), B2 (B) and B3 (C) biofilms. S: Substratum ; B: Biofilm.

B2 biofilm (Fig. 1). B1 and B2 biofilms were fed under a low organic load and exposed a homogeneous colonization over the surface plates. B3 biofilm, on the other hand, was fed under a high organic load and experienced sloughing events which caused partial colonization of the surface plates. Thickness measurements were difficult to proceed on B3 biofilm due to the strong surface heterogeneity (values ranging from 0.5 to 4 mm). B1 accumulated more biofilm mass (8.5 g VSS m2) compared to B2 (4.2 g VSS m2), and in spite of detachment events, B3 revealed the highest accumulated mass (16.6 g VSS m2) (Table 1).

3.2.

Biofilm microbial activities

Microbial activities were investigated in B1, B2 and B3 biofilms after reaching steady state conditions. Fig. 2 reveals that heterotrophic activity was found in all biofilms. However, carbon removal efficiencies were higher for B2 and B3 biofilms (respectively 93% and 97%) compared to B1 biofilm grown under a higher COD/NH4eN ratio (84%). Nitrogen removal activities where only be measured in the B2 and B3 biofilms grown under low COD/NH4eN ratios. Indeed, B2 and B3 biofilms performed simultaneous nitrification and denitrification activities, while B1 biofilm did not express any nitrogen removing activity. However, nitrification efficiency was found to be higher in B2 biofilm compared to B3 biofilm (85% versus 66%), and denitrification efficiency on the other hand was two fold higher in B3 biofilm compared to B2 biofilm (100% versus 50%). According to these results, B1 biofilm, fed on a high carbon/ nitrogen ratio, was identified as a single heterotrophic biofilm while both B2 and B3 biofilms, fed on low carbon/nitrogen

Table 1 e Physical and structural characteristics of B1, B2 and B3 biofilms. Growth conditions, colonization aspect, biofilm thickness, accumulated biomass and natural cell lysis measured on B1, B2 and B3 biofilms. Biofilm COD/TKN Surface loading (g COD m2 d1) Aspect Surface colonization Average biofilm thickness (mm) Mean accumulated mass (g VSS m2)

B1

B2

B3

73 2.5 Homogeneous Filamentous Complete 4.4  1.1 8.5

4 2.5 Homogeneous Dense Complete 1.6  0.4 4.2

4 25 Heterogeneousa Filamentous Partial 0.5e4a 16.6

a Heterogeneous biofilm thickness due to sloughing events.

1533

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 2 9 e1 5 3 8

Proteins (mg.gVSS )

A

500 400

US Tw EDTA TOTAL

300 200 100 0

Fig. 2 e Carbon removal (-), nitrification ( ) and denitrification (,) efficiencies measured in B1, B2 and B3 biofilms, when reached steady state conditions.

ratios, were identified as mixed autotrophic and heterotrophic biofilms performing simultaneous nitrification and denitrification. The developed biofilms appeared as mature and thick structures, potentially offering a large range of microbial populations. Biochemical properties of these three diverging biofilms were then investigated in terms of EPS contents and size characteristics.

Sugars (mg.gVSS )

B

B1

B2

B3

B1

B2

B3

150

100

50

0

C

6 5

3.3.

EPS content in biofilms

A multi-method extraction protocol, previously described for quantifying EPS from activated sludge (Ras et al., 2008a), was applied on each B1, B2 and B3 biofilm. The extraction protocol involved three different extraction methods (sonication, Tween and EDTA) applied sequentially on the same sample in order to collect a consistent fraction of EPS. Soluble extracts were harvested by intermediate centrifugation steps and quantified in terms of proteins and polysaccharides. Fig. 3 shows that protein contents in all extracts were systematically higher compared to polysaccharide contents, and thus independently of the applied extraction method (sonication, Tween or EDTA) as well as the biofilm (B1, B2 or B3). Extraction yields diverged between the applied methods, but revealed similar trends between the biofilms. Indeed, both protein and polysaccharide contents were always higher in the extracts obtained by EDTA and sonication steps, while Tween steps always appeared as the least efficient extraction method. Total EPS contents in each biofilm were defined by summing the amounts of proteins and polysaccharides obtained by each extraction method (sonication þ Tween þ EDTA). Fig. 3A and B show that B1 biofilm had the lowest amount of proteins (43 mg g1 VSS) and polysaccharides (15 mg g1 VSS) whilst protein and polysaccharide contents was four fold higher in B2 and 10 fold higher in B3 biofilms. As shown in Fig. 3C, protein/polysaccharide ratios in the various extracts varied between 1.8 and 5.4 but protein to polysaccharide ratio of the total extracted EPS were similar

P/S

4 3 2 1 0 B1

B2

B3

Fig. 3 e Protein content (A), Sugar content (B) and Protein to Sugar (P/S) ratio (C) in soluble extracts obtained by the multi-method extraction protocol. Soluble extracts were harvested after each extraction method (ultrasonic, Tween, or EDTA) and both proteins and sugars were assayed. Error bars are evaluated from doubled extractions and duplicate measurements. for both B1 and B2 biofilms (2.9  0.2) and slightly higher in B3 biofilm (3.7  0.2). In order to control potential cell lysis during the extraction procedure, G6P-DH activity was systematically measured in each soluble extract. The measured units obtained in each extract were added in order to evaluate the total released G6PDH activity per biofilm. Table 2 shows that some G6P-DH activity was detected in B2 and B3 but not in B1 biofilm extracts. G6P-DH units can be related to a number of disrupted cells and hence to a mass of organic cell compounds. This conversion is possible by using experimental correlation factors established with a C. necator culture (described in the Material and Methods). G6P-DH units measured in B2 extracts

1534

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 2 9 e1 5 3 8

G6P-DH activity (U g1 VSS) Number of eq. lysed cellsb (106 g1 VSS) Released cellular compoundsc (mg g1 VSS) Total extracted proteins and sugars (mg g1 VSS) Level of extract contamination by released cellular compoundsd (%)

B3 extract

0

82

2547

0

412

12735

0

2

52

58

217

539

0

0.8

9.7

B3 900

2

4

3.4.

Global EPS size distribution in biofilms

A global EPS fingerprint investigation was undertaken by pooling each EPS extract obtained from each extraction step of the protocol (sonication, Tween and EDTA), and this for B1, B2 and B3 biofilms individually. Fig. 4 shows the size distribution of the pooled fractions from each individual biofilm. Since proteins were predominant in all extracts (Fig. 3), the absorbance signal was chosen at 280 nm. Moreover, results are expressed in mAU g1 VSS in order to standardize the signal between each biofilm sample and to compare the relative predominance of size fractions between each other, by direct evaluation of their peak area. The column was previously calibrated by injecting standard size proteins, which led to a linear semi-logarithmic relation between molecular weight of proteins and elution volume (Fig. 4A). Chromatographic profiles obtained from the pooled extracts highlight qualitative and quantitative differences between B1, B2 and B3 biofilms. Nevertheless, three fractions occurred systematically in all biofilm profiles: (i) a high molecular weight fraction eluted inside the exclusion volume of the column (8 mL) indicating size fractions above 5000 kDa, (ii) an intermediate size fraction eluted between 20 and 22 mL, represented

100000 10000 1000

13 16 17 20 22 Elution volume (mL)

6

8 10 12 14 16 18 20 22 24 26 28 30 Elution volume (mL)

B

100% < 0.5 kDa (> 24 mL)

(5)

2 - 0.5 kDa (22 -24 mL)

(4)

60%

7 - 3 kDa (20-22 mL)

(3)

40%

25 - 20kDa (17 – 18 mL)

(2)

20%

> 5000 kDa (8 mL)

(1)

80%

0%

B1

are equivalent to 412$106 C necator disrupted cells, which is liable to the release of 1.7 mg of cellular components per g of biofilm VSS. Comparing this amount with the amount of proteins and sugars measured in the soluble extracts indicates that the multi-method protocol did not induce significant cell breakage in B2 biofilm since the level of contamination of the extracted EPS by released cellular molecules was estimated to 0.8%. However, by performing similar determination for B3 biofilm extracts results indicate a higher level of intracellular compounds that was estimated as 9.7% of the total extracted sugars and proteins.

1000000

400

-100 0

a Total G6P-DH activity as the sum of the G6P-DH units measured in sonication, Tween and EDTA extracts. b Evaluated by measurement of the G6P-DH activity released after lysis of cupriavidus necator pure suspensions: 0.2 U per 106 equivalent lysed cells. c Evaluated using the correlation factor of dry biomass per number of cells: 4.08$106 g per 106 cupriavidus necator cells. d Released cellular compounds after extraction/total extracted proteins and sugars.

B1

Molecular weight (kDa)

B2 extract

1400

B2

Quantitative EPS distribution (% peak area)

a

B1 extract

A Absorbance (mAU).g VSS

Table 2 e Controls of cell lysis during the extraction performed on B1, B2 and B3 biofilm and evaluation of the related contamination level of the EPS extracts.

B2

B3

Fig. 4 e Global size distribution profiles at 280 nm of total EPS extracted from each B1, B2 and B3 biofilm (A) by size exclusion chromatography. Linear semi-logarithmic relation between molecular weight of standard proteins and elution volume (A, insert). Five different EPS size clusters (1 to 5) were identified between 0.5 kDa and 5000 kDa and their relative distribution inside each biofilm was evaluated by peak integration of the 280 nm signal (B).

by 3e7 kDa size molecules and (iii) a range of small size molecules eluted beyond 24 mL, which corresponds to the total inclusion volume of the column. These latter small fractions are not in the optimal separation range offered by the column but are expected to be under 0.5 kDa and are grouped in one single category. Fig. 4A also shows that these three recurring size fractions compose alone the B1 biofilm profile. On the other hand, additional peaks were identified in B2 and in B3 biofilm profiles. Indeed, both B2 and B3 biofilm profiles revealed a fraction eluted at 17e18 mL (i.e. 20e25 kDa), and B3 biofilm alone revealed a fraction eluted at 22e24 mL (i.e. 0.5e2 kDa). A total of five different size clusters were identified among the three studied biofilms: cluster 1 (>5000 kDa), cluster 2 (20e25 kDa), cluster 3 (3e7 kDa), cluster 4 (0.5e2 kDa) and cluster 5 (<0.5 kDa). The relative abundance of EPS size clusters between each other and between each biofilm was undertaken by peak integration of each chromatographic profile. Fig. 4B compares size clusters between each biofilm, and highlights the predominance of the three recurring EPS size clusters (1, 3 and 5). Cluster 3 (3e7 kDa) was the most represented and with 86%, 60% and 46% occurrence of the total peak areas eluted from B1, B2 and B3 chromatograms respectively. The cluster 2 was specifically found in B2 and B3 biofilms, and in the same proportions (3%). The cluster 4 appeared in B3

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 2 9 e1 5 3 8

A

B1

180

B2

B3

160

Absorbance (mAU)

140 120 100 80 60 40 20 0 -20 0

2

4

6

8 10 12 14 16 18 20 22 24 26 28 30

elution volume (mL)

B

90 80

Absorbance (mAU)

70 60 50 40 30

1535

protocol (sonication, Tween and EDTA extracts). Fig. 5 shows size distribution of EPS obtained by each extraction method individually. Fig. 5C shows that EDTA-extract profiles were generally poor in EPS size diversity and offered similar profiles between the three biofilms. Indeed, this EDTA extraction step revealed the three recurring EPS clusters alone (clusters 1, 3 and 5) with a predominance of cluster 3, i.e. EPS belonging the 3e7 kDa fraction eluted between 20 and 22 mL. Fig. 5Aand B show that sonication and Tween extract size profiles, were more diversified and diverged between biofilm samples. Indeed, only the recurring clusters (1, 3 and 5) were found in B1 biofilm, whilst all clusters (1e5) were found in B2 and B3 biofilms. This result shows that cluster 2 was found only within the heterotrophic/autotrophic B2 and B3 biofilms independently on the extraction method. This latter result confirms the global analysis performed previously. On the other hand, cluster 4 which was identified in B3 biofilm alone in the global analysis is finally identified by this specific analysis, in the sonication and Tween extracts of B1 and B2 biofilms. Interestingly, a new size cluster not yet identified in previous profiles was only visualized in Tween extract profiles provided from B1 and B2 biofilms. This latter fraction was eluted at 15.6 mL, indicating a specific size of 180 kDa (Fig. 5B).

20 10

4.

0 -10

0

2

4

6

8 10 12 14 16 18 20 22 24 26 28 30 elution volume (mL)

C 1380 Absorbance (mAU)

1180 980 780

The aim of this study was to evaluate whether molecular diversity of EPS are potential markers for biofilm macro-scale characteristics. In order to validate such an approach, molecular investigations were undertaken on three biofilms, each being differentiated by their growth conditions, i.e. different substrate loading rates or different nitrogen content in the supply.

580

4.1.

380

The COD/NH4eN ratio was first chosen as a key parameter to promote the development of carbon or nitrogen removing micro-organisms. This ratio was decreased from 73 (nitrogen limitation) for B1 biofilm, to 4 (excess nitrogen) for B2 biofilm, by increasing the NHþ 4 content in the supply. Neither nitrification, nor denitrification activity was measured in this B1 biofilm indicating that the small amount of NHþ 4 in the feed was consumed for heterotrophic growth only. On the other hand, B2 biofilm which grew under excess nitrogen conditions, showed simultaneous autotrophic and heterotrophic activities. These observations are in agreement with other findings (Matsumoto et al., 2007) which showed that in spite of carbon deficiency heterotrophic bacteria can out-compete other communities such as autotrophic ammonium-oxidizing bacteria, and this due to their higher growth rate (Elenter et al., 2007; Morgenroth and Wilderer, 2000; Okabe et al., 1995). Whilst the heterotrophic B1 biofilm exposed a filamentous structure with a high accumulated mass, B2 biofilm grew into a dense granular type biofilm with a lower accumulated mass. This type of structure is also in agreement with theories valuable for biofilm or granule formation involving slow-growing

180 -20

0

2

4

6

8 10 12 14 16 18 20 22 24 26 28 30 elution volume (mL)

Fig. 5 e Specific size distribution profiles at 280 nm of EPS extracted from each step of the multi-method protocol, sonication (A), Tween (B) and EDTA (C) from B1 biofilm ), B2 biofilm ( ) and B3 biofilm ( ). (

biofilm alone and represented 5% of the total peak areas. These data indicate that EPS diversity was higher in the mixed autotrophic/heterotrophic biofilms (B2 and B3) compared to the simple heterotrophic biofilm.

3.5.

Discussion

EPS size distribution versus extraction methods

A more specific EPS fingerprint investigation was undertaken on the three biofilms by identifying the previously described size clusters in individual extracts provided by the extraction

Relating feed to biofilm properties

1536

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 2 9 e1 5 3 8

nitrifiers which seem to affect the mass density of biological matrixes (Derlon et al., 2008; Elenter et al., 2007; Liu et al., 2004). Changing the organic loading rate by 10 fold between B2 and B3, without modifying the COD/NH4eN ratio, also affected the structural and microbial properties of biofilms. As expected, the high organic load applied for B3 biofilm promoted bacterial growth which was confirmed by the high accumulated biomass measurements. However, this B3 biofilm presented partial filaments with sloughing events which caused heterogeneous colonization. It is probable that this thick biomass developed by B3 biofilm might have been more exposed to hydrodynamics, which could have triggered localized detachment events (Ohashi et al., 1995), compared to the thinner and homogeneous structure described for B2 biofilm. In addition, such a thick structure promoted anaerobic zones inside the B3 biofilm, which was confirmed by a two fold higher denitrification activity compared to B2 biofilm. Moreover, oxygen deficiency in B3 biofilm could have promoted bacterial mortality and nitrifiers, who often lose out when competing heterotrophic bacteria for oxygen, might have been particularly affected. This hypothesis is supported by the fact that nitrification efficiency was lower in the thick B3 biofilm (65%) compared to the thin B2 biofilm (85%) and by the detection of G6P-DH activity in B3 biofilm prior to EPS extraction (results not shown). Results clearly show that controlled environmental conditions can pilot microbial activities inside growing biofilms, and also modify their macro-scale structural properties.

4.2. Influence of environmental conditions on EPS production In order to harvest a representative pool of biofilm EPS, a multi-method protocol based on both mechanical and chemical extraction steps was applied on the three biofilms. Quantitative analysis of extracted proteins and polysaccharides suggests that excess nitrogen in the feed (B2 and B3 biofilms) triggered more EPS production than nitrogen limitation (B1 biofilm). These results do not join those reported by Miqueleto et al. (2010) who related decreasing values of soluble and bound EPS to decreasing carbon/nitrogen ratios in the feed of an anaerobic sequence batch biofilm reactor. EPS were produced only when oxygen, even at very low concentration was available, suggesting that microaerophilic micro-organisms were the main secretors. Li et al. (2008) showed that different thicknesses of membrane-aerated biofilms in which counter-gradients of oxygen and substrate existed, led to different EPS distributions. These authors reported a maximum EPS content (120e140 mg EPS g1 VSS extracted by formaldehyde and NaOH) in the aerobic region of the studied biofilm where carbon limitation occurred and autotrophic ammonia oxidizing bacteria developed. This latter content can be compared to the nitrifying B2 biofilm where the extractible EPS reached 210 mg g1 VSS. The total amount of extracted EPS was 2.5 times higher in B3 biofilm (high organic load) compared to B2 biofilm (low organic load) and several theories can be quoted for this increase in EPS content. Firstly, the higher substrate load

applied on B3 biofilm might have promoted bacterial growth rates, forming a thick and less cohesive structure as confirmed by sloughing events. This is in agreement with previous data which report that heterotrophic fast growing bacteria develop lower resistance towards either mechanical or chemical disintegration methods (Denkhaus et al., 2007). Consequently, extraction of EPS might be easier in such a fragile structure leading to a higher content of proteins and polysaccharides in extracts. Secondly, the extracted molecules were contaminated by soluble intracellular compounds but the level of contamination, estimated around 10%, was not high enough to justify the 2.5 fold increase of proteins and sugars observed in B3 compared to B2 biofilm. As stated earlier in the discussion, denitrification activity evidenced anoxic areas in the B3 biofilm and Adav et al. (2009) recently located proteolitic activities in anaerobic cores of bacterial granules that might have been responsible for the occurrence of granule breakdown. Therefore, possible proteolytic activity in anaerobic zones in B3 biofilm could partly explain the associated unstable structure and the high content in released proteins. Proteins were quantified in majority in all extracts with a global protein/polysaccharide ratio of 2.9  0.2 for B1 and B2 biofilms and of 3.7  0.2 for B3 biofilm. These data are in agreement with those of Gao et al. (2008) showing protein/ polysaccharide ratios varying between 1.3 and 3.3 along vertical profiles inside heterogeneous aerobic bio-filters. According to Durmaz and Sanin (2001), the amount of substrate converted to polymers by the cell depends on the composition of the growth medium. Indeed, substrates with low nitrogen content, as found in B1 biofilm may favor polysaccharide production, and on the other hand, substrates with excess nitrogen, as found in B2 and B3 biofilms, should promote protein production. Therefore, while high protein contents of B2 and B3 biofilms (164 mg g1 VSS and 424 mg g1 VSS) are in agreement with expectations, the polysaccharide content of B1 biofilm somehow low compared to B2 (15 mg g1 VSS versus 54 mg g1 VSS). This could be explained by the low organic load applied to B1 which seems to favor primarily cell growth and hence proteins (enzymatic material) rather than carbon storage (polysaccharides).

4.3.

EPS size fingerprinting of biofilms

In order to obtain a global molecular fingerprint of each biofilm matrixes, EPS extracts obtained from the multi-method extraction protocol were pooled for global analysis of the size distribution of the extracted EPS. Fractionation of these pooled extracts by SEC revealed a total of five different EPS size clusters. Three clusters were found in common between each biofilm, of which cluster 1 (>5000 kDa) is excluded from the column due to too high molecular weight EPS. Garnier et al. (2005) have already shown the existence of associated proteins/polysaccharides/mineral compounds in fractions eluted near the size exclusion volume when characterizing EPS extracted from activated sludge by SEC. Therefore, the EPS size cluster 1 might probably be represented by polymers eluted as a colloidal structure. Cluster 5 (<0.5 kDa) is, on the other hand, eluted in the total inclusion volume of the column, where the separation efficiency is reduced. This

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 2 9 e1 5 3 8

cluster 5 may either be effectively low molecular weight organic molecules such as amino-acids and peptides, or otherwise molecules which can interact with the column and hence be partially retained during elution. Hydrophobic retention of EPS on the sepharose column beads has already been proven by Comte et al. (2007) and Garnier et al. (2005), when eluting EPS extracted from activated sludges with 5% methanol. However, performing similar experimental conditions did not allow to evidence particular hydrophobic retention in this study (results not shown). Finally, the intermediate 3e7 kDa cluster 3 was predominant overall other clusters and in all biofilms. The recurrence of this cluster 3 in the three highly diversified biofilms suggests that the associated size molecules are either associated to the heterotrophic activity confirmed in all biofilms, or to mandatory EPS involved in bacterial aggregate consolidation and/or adhesion. The three recurrent EPS size clusters 1, 3 and 5, identified in this present study, were the only components of the global EPS fingerprint from the heterotrophic biofilm B1. By introducing nitrification and denitrification activities in B2 and B3 biofilms, another EPS size fraction appeared between 20 and 25 kDa (cluster 2). This latter fraction (cluster 2) could be associated to the presence of bacteria involved in nitrogen removal processes. This cluster was represented in similar proportions within B2 and B3 biofilms (3% of the eluted EPS). Since these latter biofilms exposed different nitrification and denitrification levels, cluster 2 cannot be specifically related to nitrification or denitrification microbial activities. Concerning B3 biofilm, G6P-DH measurements indicated natural cell lysis as well as cell breakage after EPS extraction. About 9.7% of the proteins and sugars measured in the B3 extract may originate from the intracellular compartment. However, due to their low proportion and to the fact that these intracellular compounds may be natural constituents of the biofilm matrix, size fingerprint of B3 biofilm can be considered as relevant. The global EPS size profile of B3 biofilm revealed an additional size cluster between 0.5 and 2 kDa, named cluster 4, that was not identified in the global EPS size profile of B1 and B2 biofilms. Performing a more specific EPS size fractionation focused on each soluble extract indicated that cluster 4 was finally found in all three biofilms. Such a result indicates that pooling extracts from one same biofilm sample can hide under-represented size fractions and hence bias final fingerprint profiles. Interestingly, the size cluster 2 (20e25 kDa) which was identified as specific to nitrogen removal activities measured in B2 and B3 biofilms, was also highlighted in sonication and Tween extracts of B2 and B3 biofilms whilst absent in either B1 biofilm extracts. These results suggest that a 20e25 kDa EPS size fraction can effectively be related to the presence and activity of the nitrogen removing micro-organisms, evidenced within B2 and B3 biofilms in spite of their diverging growth conditions and structural properties. Still in a specific view of EPS diversity through extraction methods, chromatographic profiles pointed out the strong diversity of EPS size fractions in sonication and Tween extracts in opposition to the EDTA extraction step. EDTA extracts showed poor size diversity, although the EPS content in these extracts were the highest compared to sonication and Tween extracts. Therefore, EDTA extracts alone would not be

1537

appropriate for a size diversity fingerprint study. On the other hand, the mechanical sonication and hydrophobic Tween methods are able to extract all size clusters (1e5) identified previously. Interestingly, Tween extracts revealed an additional size fraction of 180 kDa in B1 and B2 biofilms only, which was not visualized during the global study. Tween step thus reveals the most diversified EPS size profiles although EPS content in the extracts were the lowest compared to sonication and EDTA extracts. These results suggest that extraction method-specificity could be a relevant parameter for fingerprint diagnosis. The Tween-specific 180 kDa size fraction revealed in B1 and B2 biofilms may be associated to the low organic load applied to these two biofilms. In other words, the occurrence of this size fraction may rather be related to a biochemical response towards substrate-limiting conditions than to a specific microbial function. B1 and B2 biofilms were also characterized by stable and homogeneous structures in opposition to B3 biofilm, therefore, the occurrence of this Tween-specific size fraction might also mark the mechanical stability of both biofilms in opposition to B3 where this 180 kDa size fraction was absent. The specificity of this fraction towards Tween treatments indicate that the associated EPS have hydrophobic properties. Hydrophobic properties of EPS might hence be implicated in the mechanical stability of biofilms. Such results could be of interest for the understanding of attachment and detachment processes. Authors expect that in the future, configuration of appropriate coatings could be suitable to improve biofilm adherence or on the other hand to prevent biofilm development. Indeed, integrating these hydrophobic EPS fractions inside or on top of coatings could promote molecular interactions and hence biofilm strength. On the other hand, integrating specific enzymes which are liable to digest these hydrophobic EPS can also provide an alternative to toxic biocides in order to prevent biofilm growth (e.g. for heat exchangers, drinking water distribution systems). However, further studies are required before hand, such as characterizing hydrophobic EPS in different cohesive parts of biofilms. These latter investigations are already under progress.

5.

Conclusions

Characterization of EPS extracted from multi-species biofilms was investigated using a multi-method extraction procedure coupled with a SEC analysis. Results showed that EPS size diversity was higher in the two mixed heterotrophic/autotrophic biofilms compared to the heterotrophic biofilm. The multi-method extraction strategy provided consistent quantitative and qualitative EPS fractions. However, by focusing on each extraction steps, results showed that each method offered different quantities and different size diversity profiles. Nevertheless, the occurrence of a 25e50 kDa size fraction was systematically associated to biofilms exposing nitrogen removing activities. Moreover, a 180 kDa size fraction occurred in Tween extracts only and was associated to mechanically stable biofilms. This study has put forward the importance of methodology in qualitative investigations of EPS in biofilms. Hydrophobic

1538

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 2 9 e1 5 3 8

EPS seem to provide highly diversified size profiles with a particular size category (180 kDa) which might be a print of mechanical stability. Analysis of the hydrophobic EPS of biofilms developed under different shear stress conditions is currently under investigation.

references

Adav, S.S., Lee, D.J., Lai, J.Y., 2009. Proteolytic activity in stored aerobic granular sludge and structural integrity. Bioresour. Technol. 100, 68e73. Azeredo, J., Lazarova, V., Oliveira, R., 1999. Methods to extract the exopolymeric matrix from biofilms: a comparative study. Water Sci. Technol. 39, 243e250. Boualam, M., Quiles, F., Mathieu, L., Block, J.-C., 2002. Monitoring the effect of organic matter on biofilm growth in low nutritive waters by ATR-FT-IR spectroscopy. Biofouling 18, 73e81. Branda, S., Ashild, V., Friedman, L., Kolter, R., 2005. Biofilms: the matrix revisited. Trends Microbiol. 13, 20e26. Chen, X., Stewart, P.S., 2002. Role of electrostatic interactions in cohesion of bacterial biofilms. Appl. Microbiol. Biotechnol. 59, 718e720. Comte, S., Guibaud, G., Baudu, M., 2007. Effect of extraction method on EPS from activated sludge: an HPSEC investigation. J. Hazard. Mater. 140, 129e137. Coufort, C., Derlon, N., Ochoa-Chaves, J., Line, A., Paul, E., 2007. Cohesion and detachment in biofilm systems for different electron acceptors and donors. Water Sci. Technol. 55, 421e428. Davies, D.G., Chakrabarty, A.M., Geesy, G.G., 1993. Exopolysaccharide production in biofilms: substratum activation of alginate gene expression by Pseudomonas aeruginosa. Appl. Environ. Microbiol. 59, 1181e1186. Denkhaus, E., Meisen, S., Telgheder, U., Wingender, J., 2007. Chemical and physical methods for characterisation of biofilms. Mikrochim. Acta 158, 1e27. Derlon, N., Masse, A., Escudie´, R., Bernet, N., Paul, E., 2008. Stratification in the cohesion of biofilms grown under various environmental conditions. Water Res. 42, 2102e2110. Dreywood, R., 1946. Qualitative test for carbohydrate material. Ind. Eng. Chem. 18, 499. Durmaz, B., Sanin, F.D., 2001. Effect of carbon to nitrogen ratio on the composition of microbial extracellular polymers in activated sludges. Water Sci. Technol 44, 221e229. Elenter, D., Milferstedt, K., Zhang, W., Hausner, M., Morgenroth, E., 2007. Influence of detachment on substrate removal and microbial ecology in a heterotrophic/autotrophic biofilm. Water Res. 41, 4657e4671. Gao, B., Zhu, X., Xu, C., Yue, Q., Li, W., Wey, J., 2008. Influence of extracellular polymeric substances on microbial activity and cell hydrophobicity in biofilms. J. Chem. Technol. Biotechnol. 83, 227e232. Garnier, C., Go¨rner, T., Lartigues, B.S., Abdelouhab, S., Donato, P., 2005. Characterization of activated sludge exopolymers from various origins: a combined size-exclusion chromatography and infrared microscopy study. Water Res. 39, 3044e3054. Jahn, A., Nielsen, P.H., 1995. Extraction of extracellular polymeric substances (EPS) from biofilms using cation exchange resin. Water Sci. Technol. 32, 157e164.

Jahn, A., Nielsen, P.H., 1998. Cell biomass and exopolymer composition in sewer biofilms. Water Sci. Technol. 37, 17e24. Li, T., Bai, R., Liu, J., 2008. Distribution and composition of extracellular substances in membrane-aerated biofilm. J. Biotechnol. 135, 52e57. Liu, Y., Yang, S.F., Tay, J.H., 2004. Improved stability of aerobic granules by selecting slow-growing nitrifying bacteria. J. Biotechnol. 108, 161e169. Matsumoto, S., Terada, A., Tsuneda, S., 2007. Modeling of membrane-aerated biofilm: effects of COD/TKN ratio, biofilm thickness and surface loading of oxygen on feasibility of simultaneous nitrification and denitrification. Biochem. Eng. J. 37, 98e107. Mayer, C., Moritz, R., Kirschner, C., Borchard, W., Maibaum, R., Wingender, J., Flemming, H.C., 1999. The role of intermolecular interactions: studies on model systems for bacterial biofilms. Int. J. Biol. Macromol. 26, 3e16. Miqueleto, A., Dolosic, C., Pozzi, E., Foresti, E., Zaiat, M., 2010. Influence of carbon sources and C/N ratio on EPS production in anaerobic sequencing batch reactors for wastewater treatment. Bioresour. Technol. 101, 1324e1330. Morgenroth, E., Wilderer, P.A., 2000. Influence of detachment mechanisms on competition in biofilms. Water Res. 34, 416e426. Ohashi, A., Mobarry, B., Manem, J.A., Stahl, D.A., Rittmann, B.E., 1995. Influence of substrate COD/TKN ratio on the structure of multi-species biofilms consisting of nitrifiers and heterotrophs. Water Sci. Technol. 32, 75e84. Okabe, S., Hiratia, K., Ozawa, Y., Watanabe, Y., 1995. Spatial microbial distributions of nitrifiers and heterotrophs in mixed-population biofilms. Biotechnol. Bioeng. 50, 24e35. Ramsay, B.A., Lomaliza, K., Chavarie, C., Dube´, B., Bataille, P., Ramsay, J.A., 1990. Production of poly-(beta-hydroxybutyricco-beta-hydroxyvaleric) acids. Appl. Environ. Microbiol. 56, 2093e2098. Ras, M., Girbal-Neuhauser, E., Paul, E., Lefebvre, D., 2008a. A high yield multi-method extraction protocol for protein quantification in activated sludge. Bioresour. Technol. 99, 7465e7471. Ras, M., Girbal-Neuhauser, E., Paul, E., Spe´randio, M., Lefebvre, D., 2008b. Protein extraction from activated sludge: an analytical approach. Water Res. 42, 1867e1878. Simon, S., Paı¨ro, B., Villain, M., D’Abzac, P., Van Hullebusch, E., Lens, P., Guibaud, G., 2009. Evaluation of size exclusion chromatography (SEC) for the characterization of extracellular polymeric substances (EPS) in anaerobic granular sludges. Bioresour. Technol. 100, 6258e6268. Smith, P.K., Krohn, R.I., Hermanson, G.T., Mallia, A.K., Gartner, F.H., Provenzano, M.D., Fujimoto, E.K., Goeke, N.M., Olson, B.J., Klenk, D.C., 1985. Measurement of protein using bicinchoninic acid. Anal. Biochem. 150, 76e85. Standard Methods for the Examination of Water and Wastewater, ninetienth ed., 1995 APHA, AWWA, WPCF, Washington DC, USA. Wijeyekoon, S., Mino, T., Satoh, H., Matsuo, T., 2004. Effects of substrate loading rate on biofilm structure. Water Res. 38, 2479e2488. Zhang, X.Q., Bishop, P.L., KinKle, B.K., 1999. Comparison of extraction methods for quantifying extracellular polymers in biofilms. Water Sci. Technol. 39, 211e216.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 3 9 e1 5 4 8

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Long term case study of MIEX pre-treatment in drinking water; understanding NOM removal Mary Drikas*, Mike Dixon, Jim Morran Australian Water Quality Centre, South Australian Water Corporation, GPO Box 1751, Adelaide SA 5001, Australia

article info

abstract

Article history:

Removal of natural organic matter (NOM) is a key requirement to improve drinking water

Received 25 June 2010

quality. This study compared the removal of NOM with, and without, the patented

Received in revised form

magnetic ion exchange process for removal of dissolved organic carbon (MIEX DOC) as

18 November 2010

a pre-treatment to microfiltration or conventional coagulation treatment over a 2 year

Accepted 18 November 2010

period. A range of techniques were used to characterise the NOM of the raw and treated

Available online 24 November 2010

waters. MIEX pre-treatment produced water with lower concentration of dissolved organic carbon (DOC) and lower specific UV absorbance (SUVA). The processes incorporating MIEX

Keywords:

also produced more consistent water quality and were less affected by changes in the

MIEX

concentration and character of the raw water DOC. The very hydrophobic acid fraction

Coagulation

(VHA) was the dominant NOM component in the raw water and was best removed by MIEX

Microfiltration

pre-treatment, regardless of the raw water VHA concentration. MIEX pre-treatment also

NOM

produced water with lower weight average apparent molecular weight (AMW) and with the

Fractionation

greatest reduction in complexity and range of NOM. A strong correlation was found

Molecular weight

between the VHA content and weight average AMW confirming that the VHA fraction was a major component of the NOM for both the raw water and treated waters. ª 2010 Elsevier Ltd. All rights reserved.

1.

Introduction

Natural organic matter (NOM) has a significant impact on drinking water quality either directly, by reacting with water treatment chemicals (to form disinfection by-products), or indirectly, by impacting water treatment processes (including fouling of membranes and reducing the effectiveness of activated carbon for contaminant removal). Therefore the water industry has focussed on improving current treatment and developing new processes to increase the removal of NOM. Conventional treatment comprising coagulation/flocculation/ sedimentation/filtration is one of the most widely used methods to remove NOM. Extensive research has been undertaken to increase the extent of NOM removal by conventional treatment, including the use of increased

coagulant doses and reduced pH, referred to as enhanced coagulation (Crozes et al., 1995; White et al., 1997; Bell-Ajy et al., 2000). A more recent technology developed specifically for the removal of NOM is the patented MIEX DOC Process (Morran et al., 1996; Drikas et al., 2002). This process utilises a strong base anion-exchange resin, incorporating magnetic iron oxide particles within its core, which is applied to raw water utilising a stirred contactor. The small resin beads facilitate rapid reaction whilst the magnetic component allows separation of the resin and recycling of the resin in a continuous process. This differs from other more traditional applications where ion exchange resin is applied as the final polishing step within a filter (Brattebo et al., 1987; Baker et al., 1995). Laboratory scale testing of the MIEX resin has proven the effectiveness of the process for rapid removal of NOM, to a greater extent than that

* Corresponding author. Tel.: þ61 8 7424 2110; fax: þ61 8 7003 2110. E-mail address: [email protected] (M. Drikas). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.11.024

1540

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 3 9 e1 5 4 8

possible by coagulation, or enhanced coagulation, in a range of waters (Drikas et al., 2002, 2003a; Singer and Bilyk, 2002; Morran et al., 2004; Fearing et al., 2004; Humbert et al., 2005; Boyer and Singer, 2005). Some studies have also been conducted comparing pilot plant or full scale MIEX treatment with coagulation (Drikas et al., 2003b; Allpike et al., 2005; Boyer and Singer, 2005; Warton et al., 2007; Singer et al., 2007, 2009; Jarvis et al., 2008) although all of these studies have been conducted over a short period of time. A few studies have also assessed the effectiveness of MIEX as a means of reducing fouling of microfiltration or ultra filtration membranes (Fabris et al., 2007; Humbert et al., 2007; Dixon et al., 2010). MIEX has been shown to remove both hydrophobic and hydrophilic organic acid fractions of NOM (Singer and Bilyk, 2002; Morran et al., 2004; Fearing et al., 2004; Boyer and Singer, 2005; Mergen et al., 2008, 2009) to a greater extent than possible with coagulation alone. MIEX was also found to remove a wider range of molecular weight components than coagulation with alum (Drikas et al., 2003a, b; Morran et al., 2004; Allpike et al., 2005; Humbert et al., 2005; Singer et al., 2007). The first MIEX plant was commissioned at Mt Pleasant in South Australia in August 2001 (Drikas et al., 2003b). The Mt Pleasant Water Treatment Plant (WTP) is a small (2.5 ML/d) potable water treatment plant supplying high quality treated water to the local community. However the plant is more complex, innovative and diverse in processes than necessary to enable the MIEX DOC process to be fully evaluated. The WTP has been divided into two streams of 1.25 ML/d capacity, each incorporating the MIEX DOC process but the plant also enables comparison of two separate subsequent processes for the removal of suspended matter e conventional treatment (comprising coagulation, flocculation, sedimentation, and filtration) and submerged microfiltration (MF) (Drikas et al., 2003b). This study enabled an extended evaluation of the impact of the MIEX DOC process on NOM removal by comparing the performance of the two processes operating at the Mt Pleasant WTP with separate pilot plant installations utilising the same processes (conventional treatment and submerged MF) but without MIEX pre-treatment, over a 2 year period. A quantitative assessment of the NOM removed by all the treatment processes was undertaken together with a detailed study of the character of the remaining NOM using a rapid fractionation technique and molecular weight profiles for 16 months of this period. This study has identified novel benefits of the continuous operation of the MIEX DOC process and provided a clearer understanding of the character of the NOM removed.

2.

Materials and methods

2.1.

Treatment processes

A schematic of the treatment trains used is provided in Fig. 1. A conventional pilot plant (Conv) consisting of coagulation, flocculation, sedimentation and rapid filtration was established on site at the Mt Pleasant WTP using the same raw water that supplied the WTP. The flash mixer was a vessel of 1.5 L volume which was stirred at a rate of 200 rpm with a flat paddle agitation blade. Alum was dosed directly into the top of this vessel via a piston dosing pump incorporating a flow

dampening device. The two flocculation bays held 45 L each and were separated by a plate which the water laundered over. The first vessel was stirred at 80 rpm and the second at 40 rpm by an overhead stirrer with 25 offset flat paddles. Three 50 mm pipes delivered flocculated water into the 65 L sedimentation bay. The inverted pyramid shaped sedimentation bay allowed sludge to be collected over three days and be drained off to waste. Settled water was laundered from the top of the sedimentation bay via flexible beverage tubing which ran to a peristaltic pump. The pipe was split into three via a manifold in order to reach the desired flow rate and pumped via three peristaltic heads to the top of the filter column. The filter column consisted of a 140 mm diameter acrylic column and was filled with 600 mm of gravel (of varying grades), with 300 mm of sand of size 0.5e0.6 mm and 750 mm anthracite of size 1.0e1.1 mm. The media was of the same type and depth as the filters on Stream 1 of the Mt Pleasant WTP. The conventional pilot plant operated for three days on and four days off. The pilot plant throughput was 36 L/h which gave 2.5 h flocculation and 2 h settling time. The alum dose was 40 mg/L (as Al2(SO4)3$18H2O) over the study period. This was selected by the use of a model (van Leeuwen et al., 2005) and confirmed by regular jar tests to achieve the optimum DOC removal (defined as the point of diminishing return, where an additional 10 mg/ L alum produces <0.1 mg/L DOC reduction). The pH was not optimised but was between 6.5e6.8 throughout the study. Conventional treatment (Conv) was compared with Stream 1 at the Mt Pleasant WTP which incorporates MIEX followed by conventional treatment comprising coagulation, flocculation, sedimentation, rapid filtration (MIEX Coag) (Fig. 1). During the period July 2005 to June 2007, MIEX was applied to maintain the resin dose at or above 10 mL/L for 10 min contact followed by sedimentation and removal of the resin before entering the separate particulate removal processes. The actual resin dose varied between 8 and 16 mL/L (average 12 mL/L) over this period. The resin was recirculated in a continuous process with 10% removed for regeneration using sodium chloride. Fresh regenerated resin was returned continuously to the resin contact tank to maintain a constant resin dose while regeneration was undertaken separately on a batch process as required. Virgin makeup resin was added on an infrequent basis to compensate for resin lost due to attrition. Coagulation in Stream 1 during this period varied between 6 and 10 mg/L (average 8 mg/L) (as Al2(SO4)3$18H2O) and 0.2 mg/L poly dimethyl diallyl ammonium chloride (DADMAC) as a coagulant aid to ensure filtered water turbidity was maintained below 0.2 NTU. The throughput of Stream 1 at Mt Pleasant WTP remained steady at 0.3 ML/day which gave 3.5 h flocculation and 1.5 h settling time prior to filtration. Filter run times averaged 2 days. The second stream at the Mt Pleasant WTP incorporates MIEX followed by submerged continuous microfiltration (CMFS) with polyvinylidenefluoride (PVDF) membranes which have a nominal pore size of 0.04 mm (Memcor S10 V). However the MF pilot plant was used to provide the comparison of MF with and without MIEX pre-treatment to ensure operating conditions were identical for both operating systems. The MF pilot plant consisted of a single module CMF-S membrane, the same variety as that used in the Mt Pleasant WTP. Two separate membrane modules were used in the MF pilot plant unit in a one week on, one week off rotation. The source water for

1541

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 3 9 e1 5 4 8

Raw

Conventional Treatment Pilot plant

Stream1 WTP MIEX

Submerged Microfiltration Pilot Plant

WTP Conventional Treatment

Conv

Stream 1 WTP MIEX

Submerged Microfiltration Pilot Plant

MIEX Coag

Raw MF

MIEX MF

Fig. 1 e Schematic of Treatment Trains. Conv-conventional treatment pilot plant consisting of coagulation, flocculation, sedimentation and rapid filtration; MIEX Coag-Stream 1 at Mt Pleasant water treatment plant (WTP) consisting of MIEX followed by conventional treatment utilising coagulation, flocculation, sedimentation and rapid filtration; Raw MF-raw water followed by passage through submerged microfiltration membrane in pilot plant; MIEX MF-MIEX treated water sourced from Stream 1 at Mt Pleasant WTP followed by passage through submerged microfiltration membrane in pilot plant.

one module was raw water (Raw MF) and the other MIEX treated water sourced from Stream 1 (prior to coagulation) (MIEX MF) (Fig. 1).

2.2.

Analyses

Samples were taken from (a) the raw water (Raw), (b) after the filtration stage of both the conventional pilot plant (Conv) and the WTP Stream 1 (MIEX Coag) and (c) after passage through each of the MF membranes (Raw MF and MIEX MF). Monitoring of water quality included measuring the extent of NOM removal using dissolved organic carbon (DOC) and UV absorbance at 254 nm (UV254) three times a week. Characterization of NOM was determined from August 2005 to December 2006 by measuring molecular weight distribution and rapid resin fractionation. Characterization was undertaken monthly for the first 6 months and every 2 months for the remainder of the study. Samples for DOC and UV254 were filtered through 0.45 mm prerinsed membranes. UV254 was measured through a 1 cm quartz cell. DOC was measured using a Sievers 820 Total Organic Carbon Analyser (GE Analytical Instruments, USA). Specific UV absorbance (SUVA) was calculated as (UV254  100)/DOC in L/mg-m. Rapid fractionation analysis separates DOC into four fractions by adsorption onto different adsorbent resins in a sequential process based on hydrophobicity. After filtration through 0.45 mm pre-rinsed membranes a 500 mL sample was acidified and flowed through packed columns of Supelite DAX-8 and then Amberlite XAD-4 resin with intermediate samples taken for DOC analysis. Samples were then adjusted to pH 8 and flowed through a strong anion-exchange resin (Amberlite IRA-958). All resins were obtained from Supelco (Sigma Aldrich, USA). Fraction concentrations were obtained by calculation of DOC concentration measured before and after each resin. Fractions produced are defined as very hydrophobic acids (VHA), slightly hydrophobic acids (SHA), charged hydrophilics (CHA) and neutral hydrophilics (NEU).

Specifics of the technique and definitions have been described elsewhere (Chow et al., 2004). Apparent molecular weight (AMW) distribution profiles were determined using high performance size-exclusion chromatography (HPSEC) utilising UV detection after filtration through 0.22 mm pre-rinsed membranes. Weight and number average molecular weight (Mw and Mn respectively) were derived using the following equations (Chin et al., 1994). P ni M2i i (1) Mw ¼ P ni Mi i

and P

ni Mi Mn ¼ iP ni

(2)

i

where ni is the height of the HPSEC curve and Mi is the equivalent calculated molecular weight of an analyte eluted at volume i. The polydispersity (ratio of Mw to Mn) was also calculated (Chin et al., 1994). A polydispersity value of 1 indicates the presence of a single homogenous compound while greater values indicate a more disperse, complex mixture of compounds. HPSEC analysis was undertaken using a Waters Alliance 2690 separations module and 996 photodiode array detector at 260 nm (Waters Corporation, USA). Phosphate buffer (0.02 M) with 0.1 M sodium chloride was flowed through a Shodex KW802.5 packed silica column (Showa Denko, Japan) at 1.0 mL/min. AMW was derived by calibration with poly-styrene sulphonate molecular weight standards of 35, 18, 8 and 4.6 kDa.

3.

Results and discussion

3.1.

NOM removal

The raw water quality at the Mt Pleasant WTP varied over the study period; turbidity, 8.7e60 Nephelometric turbidity units

1542

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 3 9 e1 5 4 8

7 6

Period A

Period B

Raw

Period C

Conv MIEX Coag Raw MF

DOC (mg/L)

5

MIEX MF

4 3 2 1 0 Jul-05

Oct-05

Feb-06

May-06

Aug-06

Nov-06

Mar-07

Jun-07

Fig. 2 e DOC present after each treatment process (Period A, Auguste14 November 2005; Period B, 18 November 2005e15 June 2006; Period C, 19 June 2006e30 June 2007).

a

7 6

DOC (m g/ L)

5 4 3 2 1 0 Raw

b

Conv

MIEX Coag

Raw MF

MIEX MF

0.4 0.35

UVab s(/ cm )

0.3 0.25 0.2 0.15 0.1 0.05 0 Raw

c

Conv

MIEX Coag

Raw MF

MIEX MF

8 7 6

SUVA (L/m g-m )

(NTU), colour, 6e23 Hazen units, and DOC, 2.7e5.8 mg/L. The DOC variation is shown in detail in Fig. 2. Three distinct periods were observed; Period A, Auguste14 November 2005, (winter/spring), Period B, 18 November 2005e15 June 2006, (summer/autumn), and Period C, 19 June 2006e30 June 2007 (the full season cycle during drought conditions upstream). Period A consisted of variable DOC between 3.0 and 4.8 mg/L, generally decreasing, over the 4 month period. Period B began with an increase in DOC to 5.7 mg/L, due to rain and inflow from storages upstream, and remained reasonably stable within the band 4.5e5.8 mg/L over the 7 month period. Period C was less distinct but appeared to start mid June 2006 and showed a gradual steady decrease from 4.5 mg/L to 3 mg/L over the next year, under drought conditions and very little inflow upstream. The DOC concentration after each of the treatment processes over the periods A, B and C is also shown in Fig. 2. Pre-treatment with MIEX achieved the lowest DOC, with similar concentration obtained regardless of whether MIEX was followed by coagulation or MF. The DOC concentration after MIEX treatment remained nearly constant over the entire study period despite the variation in raw water DOC. Conventional treatment did not remove as much DOC as the MIEX pre-treatment whilst MF alone with no pre-treatment was found to consistently remove only a small amount of DOC. The greater DOC removal by MIEX alone compared with coagulation has been observed previously (Singer and Bilyk, 2002; Drikas et al., 2003a,b; Morran et al., 2004; Boyer and Singer, 2005; Warton et al., 2007; Jarvis et al., 2008). The extent of DOC remaining, as well as UV254 and SUVA, can also be compared over the period of the study using box and whisker plots as summarised in Fig. 3a. Outliers were excluded for each treatment train but the number of analyses conducted over this period and used in the calculations was similar for each process e Raw, 137; Conv, 132; MIEX Coag, 132; Raw MF, 139; MIEX MF, 134. As shown in Fig. 3a, the two processes incorporating MIEX pre-treatment achieved the lowest DOC concentrations with median of 1.7 mg/L after MIEX Coag and 1.6 mg/L after MIEX MF compared with 2.5 mg/L after Conv and 3.3 mg/L after Raw MF. This equates to greater average removal of DOC using MIEX pre-treatment (54% and 57%) compared with the Conv (33%) and the Raw MF (11%)

5 4 3 2 1 0 Raw

Conv

MIEX Coag

Raw MF

MIEX MF

Fig. 3 e Box and whisker plot of (a) DOC, (b) UVabs and (c) SUVA remaining after each treatment, for the total period of study. The bottom of the box is the 25th percentile and the top is the 75th. The whiskers represent the maximum and minimum for each treatment process.

processes. The 95 percent confidence intervals around the medians (not shown in this figure) indicate that the MIEX pretreatment trains are statistically different from those not including MIEX. The extent of removal of DOC by the processes incorporating MIEX pre-treatment was within the range previously observed for waters with similar SUVA values (Boyer and Singer, 2005; Singer et al., 2007). Fig. 3a also shows that the raw water over the period of study was skewed towards higher DOC and that this trend was continued after passage through the MF (Raw MF) and to a lesser extent after conventional treatment (Conv). However the data for the two treatments incorporating MIEX did not show any skew around the median results indicating that they produced more consistent water quality and were less affected by the raw water DOC with 50% of the remaining DOC results contained within a region of less than 0.5 mg/L DOC for the duration of the study.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 3 9 e1 5 4 8

The UV254 absorbance removal followed similar trends to the DOC removal although removals obtained were significantly higher - MIEX pre-treatment achieved 81% and 78% removal for the median data compared with the Conv (55%) and the Raw MF (25%) processes. Fig. 3b indicates that the UV254 absorbance of the raw water was also skewed towards higher values but confirms that all treatments used were more effective in removal of UV absorbing components than DOC with the whiskers after all treatments significantly reduced. This is not surprising as these types of structures are generally larger and more hydrophobic which makes them easier to remove by both coagulation and physical filtration. The unexpected removal of a small amount of DOC by the Raw MF can be attributed to removal of these larger UV absorbing compounds as evidenced by the reduced range of UV254 absorbance apparent in Fig. 3b. The removal of larger and more UV absorbing organics by MF membranes has been observed by other researchers (Fan et al., 2001; Lee et al., 2005). The reduced spread in UV254 absorbance of the Conv treatment is expected based on the known preferred removal of these compounds by coagulation (Chow et al., 1999; Archer and Singer, 2006). The MIEX treated waters had a very small spread of UV254 absorbance with 50% of data within a region of less than 0.01 Abs/cm and the total spread of data less than 0.03 Abs/cm. This data confirms previous observations that the removal of UV absorbing compounds by MIEX is very effective (Drikas et al., 2002; Singer and Bilyk, 2002; Fearing et al., 2004; Boyer and Singer, 2005; Humbert et al., 2005; Singer et al., 2007). The small spread in UV254 absorbance data also confirms the consistency in treated water quality following MIEX pre-treatment regardless of raw water changes observed with DOC removal. The change in SUVA would be expected to follow similar trends for all of the treated waters as it is a parameter derived from the DOC and UV254 absorbance data. Fig. 3c confirms that Raw MF reduced the SUVA of the raw water from a median of 2.4 to 2.0 L/mg-m, while Conv achieved a median SUVA of 1.5 L/mg-m. MIEX pre-treatment consistently produced water with the lowest SUVA because more UV absorbing organics were removed by MIEX relative to the DOC. This differed from Boyer and Singer (2006) who found that the treated water SUVA was similar to the raw water SUVA but was consistent with the findings of Singer et al. (2007) who found that SUVA at each utility studied decreased with MIEX treatment. MIEX MF had a median of 1.1 L/mg-m while MIEX Coag was lower with a median of 0.9 L/mg-m. Again the differences in SUVA observed following MIEX treatment were statistically significantly based on a 95 percent confidence interval.

3.2.

Rapid fractionation

To better understand the mechanism of NOM removal, the types of organics removed by each of the treatment processes were investigated. An example of the information obtained from rapid fractionation is shown in Fig. 4 for the sample taken in November 2005. DOC of the raw water was high in the early stages of the study, especially during Period B, and contained high concentration of VHA (1.7e2.3 mg/L). The VHA fraction, consisting predominantly of higher molecular weight humic and fulvic acids, was the dominant fraction

1543

Fig. 4 e Rapid Fractionation for all treatments for sample taken in November 2005. Fractions are defined as very hydrophobic acids (VHA), slightly hydrophobic acids (SHA), charged hydrophilics (CHA) and neutral hydrophilics (NEU). Error bars represent limit of detection.

present and was removed best by the two processes containing MIEX pre-treatment. The SHA fraction, variously described as transphilic (Boyer and Singer, 2005) or hydrophilic acid (Fearing et al., 2004), although not present to a large extent in the raw water during this study, was also preferentially removed by the MIEX pre-treatment. The high proportion of VHA in this water and its preferential removal by the MIEX processes resulted in a significant difference in the amount of DOC removed and in the resulting character of the waters after treatment, as illustrated in Fig. 4. Similar comparative removal was obtained for all samples analysed over the period of the study. Fearing et al. (2004), Sharp et al. (2006) and Mergen et al. (2008, 2009) utilised a modified fractionation process where they further separated the VHA fraction using pH adjustment into two additional fractions, which they defined as HAF (humic acid fraction) and FAF (fulvic acid fraction), and did not separate the non-adsorbed hydrophilic fraction into CHA and NEU as in this study. Fearing et al. (2004) found that MIEX removed a larger proportion of one component of the VHA fraction (the FAF fraction) than coagulation with ferric sulphate but similar amounts of the SHA and the other component of the VHA fraction (the HAF fraction) while Chow et al. (2005) showed a strong correlation of the VHA fraction with alum dose in a conventional treatment plant. The CHA fraction was removed most effectively, particularly by those processes containing coagulation (Conv and MIEX Coag) and/or ion exchange (MIEX Coag and MIEX MF). Other studies have also shown that the CHA fraction is the most amenable to removal by coagulation (Chow et al., 2004; Bolto et al., 2002; van Leeuwen et al., 2002) and ion exchange (Bolto et al., 2002; Morran et al., 2004). Fearing et al. (2004) found that MIEX performed better at removal of the nonadsorbed hydrophilic fraction (combination of CHA and NEU) than ferric coagulation. Boyer and Singer (2005) showed that the MIEX removed all fractions to a greater extent than coagulation with removal increasing with increased resin dose while Singer et al. (2007) found that MIEX removed the VHA and SHA fraction more than the non-adsorbed hydrophilic fraction. Study of the regenerant produced by MIEX treatment of four waters by Mergen et al. (2009) showed that

1544

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 3 9 e1 5 4 8

MIEX removed VHA (both HAF and FAF components), SHA and the non-adsorbed hydrophilic fraction but that the SHA fraction was the only fraction which increased in concentration in the regenerant solution for one water and after coagulation of the regenerant for the four waters studied suggesting better removal of this fraction by MIEX than by coagulation. The NEU hydrophilic fraction generally consists of lower molecular weight components such as polysaccharides and proteins and is often indicative of biologically derived material (Leenheer, 1981; Buchanan et al., 2005). Only very small amounts of the NEU were removed by any of the treatments. van Leeuwen et al. (2002) identified the NEU fraction as recalcitrant following fractionation and coagulation of two waters while Chow et al. (2004) showed that removal of all fractions, except the NEU fraction, could be improved by change in coagulation conditions. Sharp et al. (2006) also showed strong removal of the hydrophobic components by alum coagulation with poor removal of the non-adsorbed hydrophilic fraction (combination of CHA and NEU). The concentration of the dominant VHA fraction present in the raw and remaining after treatment over the period of the study is summarised in Fig. 5. Initial samples taken in Period A showed some inconsistency, particularly with the raw water sample. This was due to development and optimisation of the rapid fractionation method during this initial period. Notwithstanding the early results, it appears that the VHA content of the raw water decreased during Period A, increased at the junction of Period A and B and then steadily decreased during period B and C of the study. Fig. 5 also shows that generally the Raw MF removed little or no VHA, compared with the filtered raw water. This is despite the Raw MF achieving a measurable removal of DOC (median DOC difference of 0.4 mg/L). The lack of observation of consistent removal of VHA is attributed to the limit in sensitivity of the rapid fractionation analysis. The analysis requires measurement of the DOC of the raw sample and of each fraction after passage through the resins which results in an additive limit in detection of 0.2 mg/L for each fraction. This means that small differences in VHA concentrations are not able to be detected. It is likely that some DOC was removed by the Raw MF by adsorption onto particulates during filtration by the mechanism of cake layer formation. During the start of Period B, on the 18th November 2005, there was an increase in raw water turbidity from 30 NTU to above 50 NTU. This higher turbidity level (also associated with higher DOC) was present until late February 2006 when the

turbidity again was reduced to around 30 NTU. This higher turbidity period coincided closely with the measured removal of some VHA by the Raw MF observed at the start of Period B and provides support for this theory with enhanced cake layer formation at the higher turbidity. Conv treatment removed some VHA with the extent of removal apparently dependent on the amount of VHA in the raw water. The MIEX pre-treated waters achieved the lowest VHA concentration and this concentration remained nearly constant over the entire study period, regardless of the raw water VHA concentration. This supports the previous observation, noted with the other water quality parameters, of consistent treated water quality following MIEX pre-treatment regardless of raw water quality changes. There was some change in the NOM character of the raw water during the study. Initially the raw water consisted of 60% VHA during period A and B but this gradually decreased to 50% VHA by the end of the study period. The difference in NOM character of the waters after treatment is also apparent, particularly during period B, where the two MIEX pre-treated waters consistently had between 10 and 20% less VHA present after treatment, than either the Conv or Raw MF. This indicates that the DOC of the MIEX pre-treated waters consisted of a greater percentage of the other fractions; in particular the NEU fraction was a significant component of the remaining DOC, as this was not removed by any of the treatment processes. As stated earlier this NEU component is the most recalcitrant to all forms of treatment. The percentage of the VHA fraction removed by each of the treatment processes over the study period was also calculated (Fig. 6). This confirms that the raw MF removed little or no VHA during the study. Conv treatment removed very little VHA during Period A, a constant amount of VHA, about 35%, during Period B with decreasing removal observed during Period C. The two treatments incorporating MIEX achieved the highest percentage of VHA removal throughout the study. During Period B, when the raw water had very high VHA concentration, the two treatments incorporating MIEX pretreatment consistently removed a significantly greater proportion of the VHA fraction (w70%). During this same period, the Raw MF and the Conv treatments were not able to achieve the same removal of VHA (0% and 35% respectively), resulting in higher concentration of VHA in the treated waters (as shown in Fig. 5). The MIEX pre-treated waters continued to remove a high percentage of VHA even during Period C when the raw water VHA concentration decreased and the Conv

2.50 Raw

100 50

Oct-05

Dec-05

Feb-06

Apr-06

Jun-06

Aug-06

Oct-06

Dec-06

Fig. 5 e Concentration of VHA fraction present after each treatment process.

Fig. 6 e Percentage removal of VHA Fraction by each treatment process.

Dec-06

0

Oc t-06

Aug-05

Raw MF

150

-50

0.00

MIEX MF

200

Aug-06

0.50

MIEX Coag Conv

250

J un-06

1.00

Period C

Apr-06

1.50

Period B

Feb-06

MIEX MF

Period A

300

Aug-05

Raw MF

Dec-05

MIEX Coag

% VHA Fraction Removed

VHA Concentration (mg/L)

Conv

2.00

Nov-05

Period C

Oc t-05

Period B

Sep-05

Period A

1545

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 3 9 e1 5 4 8

0.006

Raw

0.005

Raw MF

UV Abs @ 260 nm

0.004

Conv 0.003

MIEX Coag

0.002

MIEX MF

0.001

0.000 100

1000

10000

-0.001

Apparent Molecular Weight (Daltons)

Fig. 7 e Apparent Molecular Weight profiles for all samples taken in November 2005.

3.3.

Apparent molecular weight distribution profiles

AMW distribution profiles of each of the treatment processes were also determined. It should be noted that as these profiles were obtained using UV detection only UV absorbing NOM would be detected. To illustrate the type of data that was obtained, the AMW profile of the same sample selected from November 2005 that was used in Fig. 4 is shown in Fig. 7. The Raw MF removed minimal DOC, resulting in molecular weight distribution similar to the raw water. Conventional treatment (Conv) removed predominantly higher AMW material, mainly above 1000 Da. The removal of high AMW organic matter with coagulation is well documented in the literature (Collins et al., 1986; Owen et al., 1995; White et al., 1997; Chow et al., 1999). MIEX treatment removed organics across the complete AMW range, including above and below 1000 Da, as has been observed in other studies (Drikas et al., 2003a,b; Morran et al., 2004; Allpike et al., 2005; Boyer and Singer, 2005; Humbert et al., 2005; Singer et al., 2007; Mergen et al., 2009). In the example shown in Fig. 7, MIEX Coag removed more NOM than MIEX MF. This was consistent for all samples up to June 2006 (Period A and B) after which removal in the two MIEX treated trains was virtually equivalent up to the last sample analysed in March 2007. The greater removal of organics by MIEX Coag

than MIEX MF evident in the AMW profile has been attributed to the inclusion of the coagulation step, which would be expected to remove additional organics, particularly above 1000 Da. To illustrate the impact of treatment on the AMW distribution, the weight-average MW (Mw) achieved after each treatment was calculated for the period of study and is summarised in Fig. 8. The Mw of the raw water over the period of study followed a similar, albeit less pronounced, pattern to that observed with the VHA; an initial high Mw which decreased during Period A, increased slightly at the junction of Period A and B and then steadily decreased during period B and C of the study. The variation in Mw of all the treated waters also followed similar, but reduced, trends to those observed with the VHA. Fig. 8 shows that the Raw and Raw MF had virtually identical Mw. The removal of DOC by the MF was low (of the order of 0.4 mg/L) and, as apparent from the example in Fig. 7, observed differences in AMW were small, due in part to the use of tighter membranes for filtration before HPSEC analysis (0.22 mm for filtration of samples compared with 0.45 mm for the other organic analyses). Fig. 8 confirms the trends observed previously for other parameters and shows that DOC removal across the entire AMW range by the processes incorporating MIEX pre-treatment

1600

Weight Average Molecular Weight (M ) (Daltons)

treatment had a reduced VHA removal. The consistently high VHA removal compares well with the work of Mergen et al. (2008) where waters containing more than 50% hydrophobic DOC were found to have removals around 70% DOC with MIEX. However in their work MIEX applied in batch tests to water containing over 75% VHA was found to result in reduced DOC removal with repeated resin use whereas DOC removal in water containing only 20% VHA and 60% SHA remained constant. In our study, raw water VHA varied between 50 and 65% (assuming raw water VHA for Period A approximated that obtained with Raw MF) but the DOC removal with MIEX did not show a decline with time based on repeated resin use such as would occur for a proportion of the resin in a continuous process.

Period A

Period B

Period C

Raw Conv

1400

MIEX Coag Raw MF

1200

MIEX MF 1000 800 600 400 200 0 Aug-05

Oct-05

Dec-05

Feb-06

Apr-06

Jun-06

Aug-06

Oct-06

Dec-06

Fig. 8 e Weight average AMW (Mw) for all treatment processes.

1546

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 3 9 e1 5 4 8

3 2.5

1.25

VHA (m g/L)

Polydispersity (M /M )

1.3

1.2

1.15

1.1

1.5 1 0.5 0

1.05 p Raw

p Conv

p MIEX Coag

p Raw MF

p MIEX MF

Fig. 9 e Median of polydispersity achieved after each treatment processes. Error bars denote interquartile range.

resulted in the lowest Mw, around 630 Da, nearly half that of the raw water and significantly less than that achieved by the Conv treatment. The remaining Mw in the MIEX pre-treated waters concurs with the observation of Humbert et al. (2005) that the components of NOM that are refractory to MIEX (and even more to coagulation) are the lowest MW components. The data from this study provides more detailed data on the range of recalcitrant organics than that observed in other studies. The polydispersity (ratio Mw/Mn) achieved after each treatment was also calculated (Chin et al., 1994). Polydispersity is a measure of the homogeneity of the NOM in a sample, with values above 1 indicating the presence of a diverse, complex mixture of compounds. The median of the polydispersity in the raw water and after each treatment for the twelve samples is shown in Fig. 9. The raw water had the greatest polydispersity, which was unchanged after passage through the MF membrane. It is likely that the higher AMW organics that may be removed by the MF membrane (Fan et al., 2001; Lee et al., 2005) would also be removed by filtration through 0.22 mm filters prior to HPSEC analysis of both the Raw and Raw MF samples and therefore differences between their molecular weight profiles and consequently polydispersity would not be measurable. Conventional treatment (Conv) reduced the range of organics somewhat whilst the two processes incorporating MIEX pre-treatments had the lowest polydispersity confirming that these treatments had the greatest reduction in complexity and range of AMW of the organics. The smaller interquartile range for MIEX Coag in Fig. 9 also confirms that MIEX Coag was more effective in reducing the range of organics than MIEX MF. This was attributed to the removal of additional organics above 1000 Da by the coagulation step.

3.4.

2

Correlations

The different characterization techniques employed whilst measuring different components are not mutually exclusive. For example, SUVA has been shown to provide a measure of the extent of conjugation and aromaticity of water (Chin et al., 1994; Weishaar et al., 2003) which is likely to be associated with both higher AMW and hydrophobic compounds such as those predominant in the VHA fraction. Therefore it would be expected that there would be a relationship between the observed SUVA and the Mw and VHA of the various waters

0

200

400

600

800

1000

1200

1400

1600

Weight Average Molecular Weight Mw (Daltons)

Fig. 10 e Correlation of VHA and weight average AMW (Mw) for raw and treated waters (r2 [ 0.83; n [ 53).

studied. These relationships were studied utilising a linear regression function for the raw water and all of the various waters following treatment. It was found that SUVA was not strongly correlated to the amount of VHA present (r2 ¼ 0.68; n ¼ 53) or the Mw (r2 ¼ 0.63; n ¼ 57). The SUVA of the raw water alone showed a similar poor relationship with both VHA (r2 ¼ 0.55; n ¼ 8) and Mw (r2 ¼ 0.47; n ¼ 10) most likely due to the large variation in the SUVA of the raw water (these 10 samples varied from 5.1 to 1.8). However the relationship between SUVA and Mw was significantly improved when the raw data was excluded from the data set (r2 ¼ 0.85; n ¼ 47) supporting the relationship between molar absorptivity and weightaveraged molecular weight observed by Chin et al. (1994). The improved correlation of SUVA with Mw following removal of the raw data was due to removal of the vast spread in the SUVA data for very similar Mw; attributed to a significantly higher concentration of UV absorbing compounds within the same molecular weight range in some raw water samples. The relationship between SUVA and VHA was not impacted when raw water data was excluded from this data set because the data range was not markedly affected by the removal of the raw data as the range in VHA concentrations in the raw water was not significantly different to the range observed after treatment. It also suggests that SUVA is not an accurate measure of the organic character for all the waters studied and while it could be correlated with the Mw of the treated waters, or those without high levels of UV absorbing organics, it could not be related to the Mw of all waters nor could it be attributed solely to the VHA component of waters. There was, however, a strong relationship between VHA and Mw when using data from both the raw water and the waters following treatment (r2 ¼ 0.83; n ¼ 53) as illustrated in Fig. 10. This relationship was also retained when the raw water data was excluded from the data set (r2 ¼ 0.85; n ¼ 44) and improved when only the raw water data was considered (r2 ¼ 0.90; n ¼ 9). This confirms that the VHA fraction was the major component of the molecular weight distribution for both the raw and treated waters supporting the premise that the hydrophobic fraction of NOM obtained using XAD-8 resin comprises a large component of the UV absorbing NOM in this natural water and concurs with Chin et al. (1994) that the relative amount of aromatic moieties in aquatic fulvic acid increases with increasing molecular weight. The close correlation between these two parameters also supports the

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 3 9 e1 5 4 8

relevance of these two techniques as valuable tools to characterise NOM. However it should also be noted that NOM arising from sources such as algal blooms may be more hydrophilic and that the observed relationship between VHA and Mw may not hold.

4.

Conclusion

Pre-treatment with MIEX prior to both conventional treatment and microfiltration over a 2 year period was found to result in consistently greater removal of DOC with resultant lower SUVA in the treated water than the same processes without MIEX over the entire period of study. The processes incorporating MIEX produced more consistent water quality and were less affected by changes in the raw water DOC. This would make treatment plants incorporating MIEX easier to operate and maintain good quality water even when subjected to large variations in the raw water DOC concentration and character. Characterization of the organics indicated that the VHA fraction was the dominant fraction present in the raw water and this was best removed by the processes incorporating MIEX pre-treatment. The high proportion of VHA in this water and its preferential removal by MIEX pre-treatment resulted in a significant difference in the amount of DOC removed and in the resulting character of the waters after treatment. The lowest VHA concentration was achieved after treatment with processes incorporating MIEX, regardless of the raw water VHA concentration. The processes incorporating MIEX pre-treatment also removed organics across the complete AMW range resulting in lower weight average AMW in the treated water than the other treatment processes. Greater removal of organics by MIEX Coag than MIEX MF was evident in some of the specific AMW profiles and this has been attributed to the inclusion of the coagulation step, which would be expected to remove additional organics, particularly above 1000 Da. The two processes incorporating MIEX pre-treatments had the lowest polydispersity confirming that these treatments had the greatest reduction in complexity and range of AMW of the organics. The greater reduction in concentration of NOM and range of organics following treatment incorporating MIEX results in more stable water which will be less reactive to disinfectants. A strong correlation was found between the VHA and weight average AMW suggesting that the VHA fraction was a major component of the AMW for both the raw water and treated waters. SUVA was found to correlate well with the weight average AMW of the treated waters, but it could not be related to the weight average AMW of all waters nor attributed to the VHA component of the waters.

Acknowledgements The authors would like to thank the following people: Nick Nedelkov for technical assistance at Mt Pleasant WTP and Edith Kozlik, Miriam Nedic and Leanne Biddiss for analytical support.

1547

references

Allpike, B.P., Heitz, A., Joll, C.A., Kagi, R., Abbt-Braun, G., Frimmel, F., Brinkmann, T., Her, N., Amy, G., 2005. Size exclusion chromatography to characterize DOC removal in drinking water treatment. Environ. Sci. Technol. 39 (7), 2334e2342. Archer, A.D., Singer, P., 2006. Effect of SUVA and enhanced coagulation on removal of TOX precursors. J.AWWA 98 (8), 97e107. Baker, J., Lavinder, S., Fu, P., 1995. Removal of natural organic matter with anion exchange resins. AWWA Annu. Conf. Proc., 547e564. Bell-Ajy, K., Abbaszadegan, M., Ibrahim, E., Verges, D., LeChevallier, M., 2000. Conventional and optimized coagulation for NOM removal. J.AWWA 92 (10), 44e58. Bolto, B., Dixon, D., Eldridge, R., King, S., 2002. Removal of THM precursors by coagulation or ion exchange. Water Res. 36 (20), 5066e5073. Boyer, T.H., Singer, P.C., 2005. Bench-scale testing of a magnetic ion exchange resin for removal of disinfection by-product precursors. Water Res. 39 (7), 1265e1276. Boyer, T.H., Singer, P.C., 2006. A pilot-scale evaluation of magnetic ion exchange treatment for removal of natural organic material and inorganic anions. Water Res. 40 (15), 2865e2876. Brattebo, H., Ødegaard, H., Halle, O., 1987. Ion exchange for the removal of humic acids in water treatment. Water Res. 21 (9), 1045e1052. Buchanan, W., Roddick, F., Porter, N., Drikas, M., 2005. Fractionation of UV and VUV pretreated natural organic matter from drinking water. Environ. Sci. Technol. 39 (12), 4647e4654. Chin, Y.-P., Alken, G., O’Laughlin, E., 1994. Molecular weight, polydispersity and spectroscopic properties of aquatic humic substances. Environ. Sci. Technol. 28 (11), 1853e1858. Chow, C.W.K., Fabris, R., Drikas, M., 2004. A rapid fractionation technique to characterise natural organic matter for the optimisation of water treatment processes. J. Water Supply: Res. Technol. . AQUA 53 (2), 85e92. Chow, C.W.K., Fabris, R., Drikas, M., Holmes, M., 2005. A case study of treatment performance and organic character. J. Water Supply: Res. Technol. . AQUA 54 (6), 385e395. Chow, C.W.K., van Leeuwen, J.A., Drikas, M., Fabris, R., Spark, K. M., Page, D.W., 1999. The impact of the character of natural organic matter in conventional treatment with alum. Water Sci. Technol. Water Supply 40 (9), 97e104. Collins, M.R., Amy, G.L., Steelink, C., 1986. Molecular weight distribution, carboxylic acidity, and humic substances content of aquatic organic matter: implications for removal during water treatment. Environ. Sci. Technol. 20 (10), 1028e1032. Crozes, G., White, P., Marshall, M., 1995. Enhanced coagulation: its effect on NOM removal and chemical costs. J. AWWA 87 (1), 78e89. Dixon, M.B., Morran, J.Y., Drikas, M., 2010. Extending membrane longevity by using MIEX as a pre-treatment. J. Water Supply: Res. Technol. . AQUA 59 (2), 92e99. Drikas, M., Morran, J.Y., Pelekani, C., Hepplewhite, C., Bursill, D.B., 2002. Removal of natural organic matter - a fresh approach. Water Sci. Technol. Water Supply 2 (1), 71e79. Drikas, M., Chow, C.W.K., Cook, D., 2003a. The impact of recalcitrant organic character on disinfection stability, trihalomethane formation and bacterial regrowth - an evaluation of magnetic ion exchange resin (MIEX) and alum coagulation. J. Water Supply: Res. Technol. . AQUA 52 (7), 475e487. Drikas, M., Morran, J.Y., Cook, D. and Bursill, D.B, (2003b), Operating the MIEX process with microfiltration or

1548

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 3 9 e1 5 4 8

coagulation. In: Proceedings of the AWWA Water Quality Technology Conference, Philadelphia, USA, November 2003. Fabris, R., Lee, E.K., Chow, C.W.K., Chen, V., Drikas, M., 2007. Pretreatments to reduce fouling of low pressure micro-filtration (MF) membranes. J. Memb Sci. 289 (1e2), 231e240. Fan, L., Harris, J., Roddick, F., Booker, N., 2001. Influence of the characteristics of natural organic matter on the fouling of microfiltration membranes. Water Res. 35 (18), 4455e4463. Fearing, D.A., Banks, J., Guytand, S., Eroles, C.M., Jefferson, B., Wilson, D., Hillis, P., Campbell, A.T., Parsons, S.A., 2004. Combination of ferric and MIEX for the treatment of a humic rich water. Water Res. 38 (10), 2551e2558. Humbert, H., Gallard, H., Suty, H., Croue, J.P., 2005. Performance of selected anion exchange resins for the treatment of a high DOC content surface water. Water Res. 39 (9), 1699e1708. Humbert, H., Gallard, H., Jacquemet, V., Croue, J.P., 2007. Combination of coagulation and ion exchange for the reduction of UF fouling properties of a high DOC content surface water. Water Res. 41 (17), 3803e3811. Jarvis, P., Mergen, M., Banks, J., McIntosh, B., Parson, S.A., Jefferson, B., 2008. Pilot scale comparison of enhanced coagulation with magnetic resin plus coagulation systems. Environ. Sci. Technol. 42 (4), 1276e1282. Lee, N.H., Amy, G., Lozier, J., 2005. Understanding natural organic matter fouling in low-pressure membrane filtration. Desalination 178 (1e3), 85e93. Leenheer, J.A., 1981. Comprehensive approach to preparative isolation and fractionation of dissolved organic carbon from natural waters and wastewaters. Environ. Sci. Technol. 15 (5), 578e587. Mergen, M.R.D., Jarvis, P., Jefferson, B., Parsons, S.A., 2008. Magnetic resin treatment: impact of water type and resin use. Water Res. 42 (8e9), 1977e1988. Mergen, M.R.D., Adams, B.J., Vero, G.M., Prices, T.A., Parsons, S.A., Jefferson, B., Jarvis, P., 2009. Characterisation of natural organic matter (NOM) removed by magnetic ion exchange resin (MIEX Resin). Water Sci. Technol. Water Supply 9 (2), 199e205. Morran, J.Y., Bursill, D.B., Drikas, M., Nguyen, H., (1996). A new technique for the removal of natural organic matter,

Proceedings of the AWWA WaterTECH Conference, Sydney, Australia. Morran, J.Y., Drikas, M., Cook, D., Bursill, D.B., 2004. Comparison of MIEX treatment and coagulation on NOM character. Water Sci. Technol. Water Supply 4 (4), 129e137. Owen, D.M., Amy, G.L., Chowdbury, Z.K., Paode, R., McCoy, G., Viscosil, K., 1995. NOM characterisation and treatability. J.AWWA 87 (1), 46e63. Sharp, E.L., Jarvis, P., Parsons, S.A., Jefferson, B., 2006. Impact of fractional character on the coagulation of NOM. Colloids Surf. A Physicochem Eng. Asp 286 (1e3), 104e111. Singer, P.C., Bilyk, K., 2002. Enhanced coagulation using a magnetic ion exchange resin. Water Res. 36 (16), 4009e4022. Singer, P.C., Schneider, M., Edwards-Brandt, J., Budd, G.C., 2007. MIEX for removal of DBP precursors: pilot plant findings. J. AWWA 99 (4), 128e139. Singer, P.C., Boyer, T., Holmquist, A., Morran, J., Bourke, M., 2009. Integrated analysis of NOM removal by magnetic ion exchange. J. AWWA 101 (1), 65e73. van Leeuwen, J., Chow, C., Fabris, R., Withers, N., Page, D., Drikas, M., 2002. Application of a fractionation technique for the better understanding of the removal of NOM by alum coagulation. Water Sci. Technol. Water Supply. 2 (5e6), 427e433. van Leeuwen, J., Daly, R., Holmes, M., 2005. Modelling the treatment of drinking water to maximize dissolved organic matter removal and minimize disinfection by-product formation. Desalination 176 (1e3), 81e89. Warton, B., Heitz, A., Zappia, L.R., Franzmann, P.D., Masters, D., Joll, C.A., Alessandrino, M., Allpike, B., O’Leary, B., Kagi, R.I., 2007. Magnetic ion exchange drinking water treatment in a large-scale facility. J.AWWA 99 (1), 89e101. Weishaar, J.L., Aiken, G.R., Bergamaschi, B.A., Fram, M.S., Fujii, R., Mopper, K., 2003. Evaluation of specific ultraviolet absorbance as an indicator of the chemical composition and reactivity of dissolved organic carbon. Environ. Sci. Technol. 37 (20), 4702e4708. White, M.C., Thompson, J.D., Harrington, G.W., Singer, P.C., 1997. Evaluating criteria for enhanced coagulation compliance. J. AWWA 89 (5), 64e77.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 4 9 e1 5 6 0

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

The implications of household greywater treatment and reuse for municipal wastewater flows and micropollutant loads D. Michael Revitt a,*, Eva Eriksson b, Erica Donner a,1 a b

Urban Pollution Research Centre, Middlesex University, Hendon Campus, The Burroughs, London NW4 4BT, United Kingdom Department of Environmental Engineering, Technical University of Denmark, Miljoevej B113, Kgs. Lyngby, DK-2800, Denmark

article info

abstract

Article history:

An increasing worldwide interest in water recycling technologies such as greywater

Received 10 September 2010

treatment and reuse suggests that additional research to elucidate the fate of xenobiotics

Received in revised form

during such practices would be beneficial. In this paper, scenario analyses supported by

2 November 2010

empirical data are used for highlighting the potential fate of a selection of xenobiotic

Accepted 19 November 2010

micropollutants in decentralised greywater treatment systems, and for investigation of the

Available online 24 November 2010

possible implications of greywater recycling for the wider urban water cycle. Potential potable water savings of up to 43% are predicted for greywater recycling based on Danish

Keywords:

water use statistics and priority substance monitoring at a greywater treatment plant in

Greywater treatment

Denmark. Adsorption represents an important mechanism for the removal of cadmium,

Wastewater influent

nickel, lead and nonylphenol from influent greywater and therefore the disposal route

Recycling

adopted for the generated sludge can exert a major impact on the overall efficiency and

Priority substances

environmental sustainability of greywater treatment.

Scenario analyses

ª 2010 Elsevier Ltd. All rights reserved.

Sludge disposal

1.

Introduction

With pressures on potable water supplies continuing to increase worldwide, interest in the use of alternative water sources such as recycled wastewater is also growing (Chu et al., 2004; Bixio et al., 2006). In particular, greywater treatment and reuse is receiving increasing attention (e.g. Maimon et al., 2010; Liu et al., 2010). This is because greywater generally has a lower organic pollutant and pathogen content than combined municipal wastewater which also contains toilet waste (Eriksson et al., 2002). Thus, greywater is considered particularly suitable for on-site (i.e. decentralised) treatment and

reuse. Greywater treatment and reuse schemes have already been piloted in many countries around the world and are becoming increasingly commonplace in water stressed areas such as Australia and Mediterranean countries (Friedler and Gilboa, 2010; Masi et al., 2010; Pinto and Maheswari, 2010). However, related research has largely been restricted to studies of standard water quality parameters such as total organic carbon, biological oxygen demand, chemical oxygen demand and faecal and total coliforms (e.g. Pidou et al., 2008; Paulo et al., 2009). In contrast, there has been very little greywater research investigating the loads and dynamics of micropollutants. Nevertheless, Eriksson et al. (2002, 2003) and Palmquist and

* Corresponding author. Tel.: þ44 (0)20 8411 5308; fax: þ44 (0)20 8411 6774. E-mail address: [email protected] (D.M. Revitt). 1 Present address: Centre for Environmental Risk Assessment and Remediation (CERAR), University of South Australia, Building X, Mawson Lakes Campus, Mawson Lakes, SA-5095, Australia. 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.11.027

1550

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 4 9 e1 5 6 0

Hanaeus (2005) have collectively shown that a large number of xenobiotic substances can find their way into greywater via bathroom and laundry products. Donner et al. (2010) have reported initial investigations into the fate of a range of pollutants within greywater treatment and reuse systems. However, given the increasing implementation of greywater recycling technology, it is evident that additional research to elucidate the behaviour of xenobiotic micropollutants during greywater treatment would be beneficial. It would also be useful to understand the potential implications of more widespread greywater recycling for urban wastewater loads and dynamics. Greywater treatment and reuse is a very diverse field, encompassing a wide range of potential treatment trains and spatial scales, as well as numerous reuse options (Li et al., 2009; Misra et al., 2010). Current treatment options vary widely in sophistication from simple filter systems to constructed wetlands, multi-stage biological treatment systems, and membrane bioreactors. Nevertheless, all systems are based on a combination of chemical, physical and biological processes such as adsorption, coagulation, precipitation, filtration, aeration, biodegradation, and disinfection. Reuse options cover a wide range of non-potable applications, from those involving a higher risk of human exposure such as spray irrigation and car washing, to lower risk options such as toilet flushing and sub-surface irrigation of non-food crops. Although pathogen transfer is generally considered the most pressing concern, it is nonetheless important to ensure that the lack of information regarding the chemical pollutant dynamics of greywater does not lead to the prevalence of suboptimal treatment trains or inappropriate reuse practices. This is currently being brought into focus with the development of national standards and codes of practice for both greywater treatment and specific reuse applications (e.g. in the UK and Australia). Fatta-Kassinos et al. (2010) have recently reviewed the practice of wastewater reuse for irrigation purposes and concluded that the benefits associated with improved water balances and nutritional levels need to be assessed against the current lack of knowledge relating to possible impacts on ecosystems and human health of the applied organic xenobiotics and heavy metals. In this paper, scenario analyses are used to highlight the potential fate of a selection of xenobiotics in decentralised greywater treatment systems, and to investigate the possible implications of greywater recycling for the urban water cycle. All of the substances investigated are listed under the European Water Framework Directive (WFD) (European Commission, 2000a) as ‘Priority Substances’ (PS) or ‘Priority Hazardous Substances’ (PHS) and are known to be present in greywater. A range of different greywater treatment and reuse scenarios are compared in order to ascertain the likely benefits/shortcomings of the different scenarios in terms of micropollutant persistence and fate, including the possible impacts on municipal wastewater flow dynamics and pollutant source control. Due to the limited availability of relevant data, the presented results focus on cadmium (Cd), nickel (Ni), lead (Pb), benzene and 4-nonylphenol (4-NP). Cadmium, Ni and Pb are metal pollutants of high concern in the municipal wastewater treatment process, as their tendency to accumulate in sludge can counteract its beneficial reuse for nutrient recovery and soil conditioning. For instance, national and European regulations

specify acceptable levels of metal pollutants in sludge destined for recycling to agricultural land (e.g. European Commission, 1986) and sludge not meeting those criteria must be disposed of via alternative means such as incineration or landfilling. Particular focus is given in this paper to the potential for greywater treatment to act as an emission control barrier for Cd. Recognised as a PHS under the WFD and highlighted as a major element of concern in relation to sludge quality, Cd is toxic to humans, has no known biological function and is one of the more mobile metals in soil. It is thus of particular concern in terms of crop uptake potential as it can pose health risks to humans and animals at levels well below phytotoxic concentrations (McLaughlin et al., 2000). Some sludge regulations (including the Danish national regulations) also specify acceptable levels of key organic pollutants, such as nonylphenols which have been found to accumulate in the sludge fraction during wastewater treatment (e.g. Abad et al., 2005; Koh et al., 2005). For contrast, benzene has also been included among the selected substances because being a relatively volatile substance, it tends to partition predominantly to air rather than sludge or water, and can thus be expected to demonstrate a differing behaviour during greywater treatment. Both benzene and 4-NP are resistant to biodegradation, as is typically the case for substances identified as PS/PHS. This investigation of the fate of selected greywater micropollutants facilitates a good overview of the possible implications of more widespread implementation of greywater reuse technologies.

2.

Materials and methods

2.1.

Greywater treatment at Nordhavnsga˚rden

The Nordhavnsga˚rden treatment plant is located in the basement of an apartment block in Copenhagen, Denmark, and consists of a primary settling tank, a three-stage rotating biological contactor (RBC), a secondary settling tank, a sand filter, an ultraviolet disinfection unit, and a service-water storage tank. Eighty-four one-bedroom apartments (w117 inhabitants) are connected to this facility which treats bathroom greywater for reuse as toilet flushing water and is automatic and self-cleaning.

2.2. Chemical analysis of PS and PHS in greywater and greywater treatment sludge The selected PS (benzene, Ni and Pb) and PHS (Cd and 4-NP) were measured both in hot and cold potable water, and in the influent and effluent greywater from the ‘Nordhavnsga˚rden’ greywater treatment system. Sixteen time-proportional samples of influent and effluent greywater were collected over a one-week period (29 November to 5 December 2007) using acid washed bottles. In addition, bottles used to collect samples for organic analysis were pre-heated at high temperature (220  C for 24 h). All samples (except for benzene analysis) were filtered prior to analysis (GF/A 1.6 mm for metal analysis and GF/C 1.2 mm for organics analysis). Cadmium, Ni, and Pb were analysed by Inductively Coupled Plasma - Optical Emission Spectroscopy (Varian Vista-MPX

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 4 9 e1 5 6 0

1551

Table 1 e Greywater treatment and reuse scenarios considered for this study. Scenario A B C D E F G H I J K L

Treatment system No treatment Indoor - RBC Indoor - RBC Indoor - RBC Indoor - RBC Indoor - RBC Indoor - RBC Indoor - RBC Indoor - RBC Indoor - RBC Outdoor - reedbed Outdoor - reedbed

Scenario analyses

The twelve greywater treatment and reuse scenarios investigated during this study are documented in Table 1. They range

Table 2 e Proportion of household water used for different domestic purposes (after Kjellerup and Hansen, 1994). Location/use of household water Bathrooms Laundry activities Kitchens Toilet flushing Irrigation a Average percentages in parenthesis.

Reuse of greywater

e

e Toilet Toilet þ Irrigation Toilet Toilet þ Laundry Toilet þ Irrigation Toilet þ Laundry þ Irrigation Toilet þ Laundry Toilet þ Irrigation Toilet þ Laundry þ Irrigation Groundwater recharge Groundwater recharge

Bathroom Bathroom Bathroom Bathroom Bathroom Bathroom Bathroom Bathroom Bathroom Bathroom Bathroom

CCD Simultaneous ICP-OES). Benzene was determined by purge and trap (Tekdyn Tekmar Velocity XPT Purge and Trap Sample Concentrator) and gas chromatography (Shimadzu Gas Chromatograph GC-14B, equipped with a Flame Ionization Detector). 4-nonylphenol was isolated and concentrated by solid phase extraction prior to analysis by GCeMS (Agilent 6890N GC-system with an Agilent 5973 Mass Selective Detector). All instrumental analyses were performed in triplicate. Quality control procedures included determination of detection limit, quantification limit, linearity, and precision. The detection limits for the employed analytical procedures were benzene (1.4 mg l1), 4-NP (0.005 mg l1), Cd (0.01 mg l1), Ni (0.1 mg l1) and Pb (0.03 mg l1). Internal reference materials were also included in all analyses for quality control purposes. The total greywater sludge was collected from the primary settling tank and rotating biological contactor on three occasions (separated by 4 monthly intervals) and was initially dewatered by centrifugation (4000 rpm for 20 min). The settled material was dried at 105  C for 1 h, pulverised and weighed, then acid digested (7 M nitric acid at 125  C and 2 atm for 30 min according to Danish Standards (DS259, 2003, DE/EN15586, 2004) prior to metal analysis by ICP-OES. The sludge was not analysed for benzene and 4-NP. Total solids (TS) were determined according to APHA et al. (2005) to facilitate normalisation of the sludge metal content to the concentration per unit of dry weight (DW).

2.3.

Source of greywater

Range and average percentagesa 35e37 (36) 13e15 (14) 17e25 (21) 20e27 (23) 5e7 (6)

þ Laundry þ Laundry þ Laundry þ Laundry þ Laundry þ Kitchen þ Laundry þ Kitchen þ Laundry þ Kitchen þ Laundry

from a baseline scenario of no treatment and no reuse (Scenario A) to full household greywater treatment and recycling (Scenario J; bathroom, laundry and kitchen greywater treated and reused for toilet flushing, laundry washing and irrigation). The identified scenarios differ in terms of the type of treatment plant (e.g. an indoor system using an RBC system and outdoor land-based treatment systems using reedbeds), in terms of the source of the greywater being treated (e.g. bathroom vs. bathroom þ laundry) and in terms of the enduse of the recycled water (e.g. toilet flushing vs. toilet flushing þ laundry washing). In practice, bathroom greywater is the fraction most commonly recycled and this is the reason for the relative dominance of this fraction in the selected scenarios (Table 1).

2.4. Water use statistics and input data to scenario analyses The scenario analyses reported in this paper are based on Danish water use statistics. The potential effects of greywater recycling on wastewater flows under the different scenarios (assuming that 100% implementation of greywater recycling technology is practised) have been calculated based on an average Danish potable water consumption of 119 l person1 day1 and a 43% contribution from households to the influent of municipal wastewater treatment plants (DANVA, 2007). The other major inputs to wastewater treatment plants are from industrial and commercial wastewater, stormwater and sewer infiltration. The proportion of household water used for different domestic purposes (Kjellerup and Hansen, 1994) is identified in Table 2. Similar distributions have been reported by Memon and Butler (2006) for residential properties in the UK although with an increased proportion for toilet flushing and a reduced percentage for general bathroom use.

2.5.

Pollutant fate analysis

The fate of the selected substances during greywater treatment and reuse has also been evaluated under the different scenarios. Hypothetical pollutant removal efficiencies of 10%, 50% and 90% were used for the pollutant fate calculations in order to cover a broad range of potential treatment situations. With such a broad range of treatment systems potentially available and little attention given to optimising these

1552

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 4 9 e1 5 6 0

Table 3 e Nordhavnsga˚rden monitoring data used in the scenario calculations, and other relevant data from the literature (all values in mg lL1). Cd

Ni

Pb

Benzene

4-NP

Greywater influent concentration (Danish and Swedish greywater literature data)

Range: 0.01e0.22 Mean: 0.08 Mediana: 0.07 Rangeb: 0.06e0.66 Meanb: 0.22 2.5c < 0.1d Rangee: 0.06e0.16 Meane: 0.10

Range: 5.15e26.5 Mean: 9.32 Median: 6.76 Rangeb: 3.86e10.2 Meanb: 6.2 1.3c 1.5d Rangee: 4.45e28.1 Meane: 11.0

Range: 4.89e10.2 Mean: 6.95 Median: 6.82 Rangeb: 1.1e6.9 Meanb: 3.4 1.8c <2d Rangee: 2.14e3.14 Meane: 2.52

Range: <1.4e9.85 Meana: 3.61 Mediana: 2.51 All values 1.9b

Range: 0.35e1.63 Mean: 0.90 Median: 0.90 All values <0.5b,g

Potable water concentration (Nordhavnsga˚rden) Concentration in Copenhagen potable water abstraction wellsg

Cold water: <0.01 Hot water: <0.01 Range: 0.03e0.07 Mean: 0.04

Cold water: 0.24 Hot water: 0.35 Range: 0.46e8.9 Mean: 2.21

Cold water: 7.27 Hot water:6.21 Range: <0.03e0.11 Mean: 0.22

Cold water: <1.4 Hot water: <1.4 All values <1.4

Influent concentration (Nordhavnsga˚rden) (n ¼ 8)

0.76c,g 0.9d,g Rangee: 2.85e5.95 Meane: 3.8 Rangef: 0.56e1.1 Meanf: 0.76 No data All values <0.5

a 38% of the values for benzene were below the detection limit; for the purposes of calculating mean and median values these were assumed to be equal to half of this value (i.e. 0.7 mg l1 for benzene). b BO90 (apartment block), Copenhagen, Denmark (Ledin et al., 2006). c Gals Klint (campingsite), Denmark (Nielsen and Pettersen, 2005). d Vestbadet I/S, Denmark (Andersson and Dalsgaard, 2004). e Vibya˚sen (housing area), Sollentuna, near Stockholm, Sweden (Palmquist and Hanaeus, 2005). f Gebers (apartment block), Skarpnack, near Stockholm, Sweden (Palmquist, 2004). g Indicates that a measurement includes not only 4-NP but nonylphenols collectively.

systems for micropollutant removal it is prudent to conclude that many systems may have limited effectiveness in terms of non-standard parameters. Pollutant load data used for the pollutant fate calculations have predominantly been based on the Nordhavnsga˚rden data presented in this paper. However, only bathroom greywater is recycled at the Nordhavnsga˚rden site. Thus, in order to facilitate Cd fate calculations for the full suite of scenarios (Scenarios A-L), additional data on greywater Cd loads for kitchen and laundry greywater was taken from Wall (2002) and Bergstrom (2007) and the Cd load in blackwater (i.e. toilet wastewater including faeces and urine) was taken from Palmquist and Hanaeus (2005). These studies were conducted in Swedish households. As measured data for laundry and kitchen greywater were not available for benzene, 4-NP, Ni, and Pb only those scenarios involving bathrooms as the source of greywater (Scenarios B and C) have been investigated for these pollutants but a complete scenario analysis has been conducted for Cd. The physicochemical characteristics of the different pollutants have been taken into account in assessing their removal behaviour during the greywater treatment process. For the metals and their compounds the main removal process will be adsorption with negligible removal by biodegradation and no susceptibility to volatilisation. A precise assessment of metal adsorption capability is difficult due to the variety of compounds and complexes which can exist in wastewater samples but in a review of the potential of metals to be removed from stormwater, Revitt et al. (2008) have identified the highest adsorptive removal to be associated with Pb followed by Ni and with Cd demonstrating the lowest removal potential. The behaviours of benzene and 4-NP can be correlated with the relevant physiochemical parameters such as adsorption coefficients, biodegradation

half-lives and Henry’s Law constant for volatilisation (Scholes et al., 2007). These parameters suggest equal, but limited, susceptibilities for both pollutants to aerobic biodegradation but clear differences with regard to adsorption and volatilisation. Benzene is predicted to have a high potential to be removed by volatilisation compared to the moderate removal for 4-NP and the reverse is true for adsorption although to a less exaggerated extent.

3.

Results and discussion

3.1.

Priority substances in greywater

A summary of relevant pollutant monitoring data for greywater influent to the Nordhavnsga˚rden treatment plant is given in Table 3. All of the selected PS/PHS were detected at measurable concentrations and the results are generally comparable to existing Danish and Swedish greywater monitoring data for these substances (also given in Table 3), with some exceptions such as the high concentration of Cd (2.5 ug l1) measured at the Gals Clint camping site (Nielsen and Pettersen, 2005). However, a high level of consistency is not to be expected given that greywater flows and pollutant loads are inherently variable and highly dependent on the behaviour of individuals. In addition to the concentrations of the selected PS/PHS in greywater, measured values for these substances in the potable water at Nordhavnsga˚rden, and in the abstraction wells used to supply the potable water distribution network in Copenhagen (Copenhagen Energy, 2008a, 2008b) are also presented in Table 3. The abstraction well data clearly demonstrate the low background levels of the monitored substances.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 4 9 e1 5 6 0

Fig. 1 e a and b Diagrammatic representation of water flow for Scenarios A and J (dashed borders indicate water use options which are not relevant to that particular scenario).

3.2.

Flow calculations

Based on monitored greywater inflow rates and the Danish water use statistics specified in Section 2.4, effluent flow rates (expressed as litres per person per day; 1 p1 d1) have been calculated for each of the identified scenarios. Fig. 1a and b provide diagrammatic representations of the flow pathways associated with Scenarios A and J and serve as examples of the method by which the proportional potable water savings and the proportional reductions in wastewater treatment plant effluent in columns 2 and 3, respectively of Table 4 were derived. It can be seen that under the baseline conditions represented by Scenario A (i.e. no greywater treatment followed by reuse but direct use of greywater for irrigation purposes) a daily potable water use of 119 l p1 d1 results in 111.9 l p1 d1 of household wastewater being released to the municipal wastewater system. In contrast, under Scenario J (where bathroom, laundry and kitchen greywater are treated and reused for irrigation, laundry washing and toilet flushing), the effluent volume is reduced to 60.7 l p1 d1, representing a reduction in the effluent to the municipal wastewater treatment plant (WWTP)

1553

of 20% (when the 43% contribution of households to this wastewater stream is taken into account). This scenario also achieves a potable water saving of 51.2 l1p1 d1 due to the use of greywater for toilet flushing, the continued recycling of laundry effluents through the greywater treatment system and avoidance of using potable water for irrigation. The effective water use is 67.8 l1p1 d1 which amounts to a saving of 43% compared to the baseline situation represented by Scenario A. The calculations for Scenario J (Fig. 1b) also show that 33.3 l1p1 d1 of treated greywater will be produced for which there is no identified reuse application. This would represent an inefficient use of treatment resources and the described scenario analysis approach therefore offers a route for optimising the treated volumes according to user requirements. The flow calculation results provided in Table 4 demonstrate the implications of the different scenarios in terms of both potential potable water savings and reduced wastewater influent volumes at municipal WWTPs. Significant potable water savings (up to 43% for the described scenarios) can be achieved by recycling greywater. However, subsequent reductions in wastewater flows to large-scale municipal WWTP are predicted to be more modest (up to 27% for Scenario K) as the assumption has been made that only 43% of the total WWTP influent volume is derived from households (DANVA, 2007). The most beneficial combination of potable water savings and WWTP influent reductions are achieved when the volume of recycled water is sufficient to cover the requirements for toilet flushing, laundry washing, and outdoor irrigation uses (e.g. Scenarios G and J). It is important to note however that these impacts have been calculated on the basis of 100% uptake of the relevant greywater recycling scenario. Whilst this is feasible for new developments (or large-scale refurbishments), particularly in water stressed countries where water recycling regulations on new-builds are increasingly likely to be introduced, it should be recognised that implementation of greywater reuse in more established built environments without existing dual reticulation plumbing systems is likely to remain much lower than 100%.

3.3. Micropollutant fate during greywater treatment and reuse For each indoor treatment and reuse scenario (Scenarios A-J), the fates of the pollutants have been calculated based on hypothetical greywater treatment removal efficiencies of 10%, 50% and 90%. These hypothetical removal efficiencies span the wide range anticipated for the available treatment options of varying sophistication which can be expected to differ substantially in their ability to remove micropollutants. For example, losses due to volatilisation are likely to be greater in systems incorporating rotating biological contactors, than in simple filtration systems without additional aeration and will therefore exert the greatest influence on the removal of benzene. Treatment systems also vary widely in their ability to remove suspended solids and adsorbed pollutants from greywater (Donner et al., 2010). This is a process which has been identified as being important for the removal of Pb and 4NP. The composition and condition of the microbial

1554

Table 4 e Implications of Scenarios A-L for municipal wastewater flows and Cd loads, assuming on-site greywater treatment Cd removal efficiencies of 10%, 50% and 90%. Scenario

Reduction in WWTP influent (%)

Reduction in Cd load to WWTP based on 10% removal efficiencya Assuming sludge is discharged to WWTP

A B Cb D E Fb Gb H Ib Jb K L

e 23 29 23 37 29 43 37 29 43 0 0

e 11 13 11 17 13 20 17 13 20 27 17

e 0 0.45 (2.2%) 0 0 0.82(4.1%) 1.25(6.2%) 0 0.69(3.4%) 0.84(4.2%) N/A N/A

Assuming sludge is removed from WW stream e 0.31 0.75 0.77 1.19 1.59 2.28 1.13 1.62 1.97 N/A N/A

(1.5%) (3.8%) (3.8%) (5.9%) (7.9%) (11.3%) (5.6%) (8.1%) (9.8%)

Reduction in Cd load to WWTP based on 50% removal efficiencya Assuming sludge is discharged to WWTP

Assuming sludge is removed from WW stream

e 0 0.25 (1.2%) 0 0 0.46 (2.3%) 0.60 (3.0%) 0 0.38 (1.9%) 0.42 (2.1%) N/A N/A

e 1.53 (7.6%) 1.78 (8.9%) 3.85 (19.1%) 4.56 (22.7%) 4.31 (21.5%) 5.09 (25.3%) 5.16 (25.7%) 5.02 (25.0%) 5.58 (27.8%) N/A N/A

a Main value given is the reduction in load in mg p1 d1; values in brackets show the reduction in load as a percentage of the total household load). b Values given show the reduction in load after 5 cycles of the given scenario (i.e. laundry water recycled 5 times).

Reduction in Cd load to WWTP based on 90% removal efficiencya Assuming sludge is discharged to WWTP e 0 0.05 0 0 0.09 0.11 0 0.08 0.08 N/A N/A

(0.2%)

(0.4%) (0.5%) (0.4%) (0.4%)

Assuming sludge is removed from WW stream e 2.74 2.80 6.93 7.15 7.02 7.24 8.53 8.43 8.60 N/A N/A

(13.5%) (13.9%) (34.5%) (35.6) (35.0%) (36.1%) (42.5%) (42.0%) (42.8%)

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 4 9 e1 5 6 0

Potable H2O saving (%)

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 4 9 e1 5 6 0

Scenario B Irrigation Irrigation Bathroom

0 µg p d

[A]

3.04 µg p d

Greywater Treatment Plant

Laundry 0 µg p d

3.04 µg p d

Potable water

Laundry 4.65 µg p d

Sludge 2.74 µg p d

[B]

[D]

Toilet 11.16 µg p d

Kitchen 1.58 µg p d

[C]

Surplus 0.11 µg p d

Toilet 0 µg p d

Potable H2O saving = 27 l p d (23 %) WWTP influent reduction = 11 %

Municipal [F] Wastewater Treatment Plant

[E]

20.23 µg p d

Fig. 2 e Diagrammatic representation of Scenario B and associated Cd load calculations (based on 90% removal efficiency during treatment) as described in Box 1. Letters in square brackets can be used to match with the associated calculation in Box 1.

community or biofilm in biological systems will significantly affect the biodegradation potential for organic micropollutants (Donner et al., 2010; Giri et al., 2006) and has been identified as being equally important for the removal of both benzene and 4-NP. Biological greywater treatment systems can take some time to mature and establish reliable performance and may be inhibited by pollutant shock loadings, such as a predominance of bleach or other cleaning products. Treatment efficiencies can be expected to vary over time and the use of hypothetical removal efficiencies of varying effectiveness is thus a useful approach for providing an overview of the possible impacts of different greywater treatments and reuse scenarios on the wider urban water cycle. In Table 4 the results of the Cd fate calculations for the full range of scenarios are presented. These results also demonstrate how two different hypothetical pathways for sludge disposal will influence the influent Cd load to a WWTP. One set of calculations are based on the assumption that the greywater treatment sludge will be discharged or transferred periodically to the municipal WWTP (as is in fact most commonly the case) with the second set of calculations being designed to investigate the effect of employing a separate sludge disposal route (such as disposal to land). As an example of the manner by which pollutant pathways have been evaluated for the different scenarios, the fate of household-derived Cd pollution under Scenario B (see Fig. 2) is described in detail in Box 1. The different steps in the calculation can be matched to the scenario diagram by means of the square bracketed letters in both Fig. 2 and Box 1. According to Scenario B, bathroom greywater is treated on-site using an RBC and reused for toilet flushing, and the results show that treatment and reuse according to this scenario will have no positive effect on WWTP Cd influent loads unless the sludge is removed from the wastewater stream entering the associated WWTP (Table 4). Furthermore, even under conditions of separate sludge disposal, the greatest potential decrease in Cd

1555

loading at the treatment plant will be 2.74 mg p1 d1 (assuming 90% removal efficiency during treatment and 100% implementation of Scenario B). Compared to the baseline scenario (Scenario A) which incorporates no greywater treatment and reuse, this represents a fairly minor overall reduction (13.5%) on the influent Cd load at the WWTP, as baseline calculations indicate a total household load of 20.2 mg p1 d1. It is clear that the incorporation of Cd in the sludge is a critical pathway in controlling the fate of this and similar pollutants. In those situations where the sludge from the greywater treatment process is eventually discharged or transferred to a WWTP, there will be no overall Cd removal unless the scenarios incorporate removal of some of the treated greywater from the municipal wastewater stream by using it for irrigation purposes (i.e. Scenarios C, F, G, I and J). When irrigation is practised, it is interesting to note that the impact on the WWTP load is not consistent with the increasing treatment efficiency of the greywater plant. Thus for Scenario C, it can be seen that the overall removal of Cd from the wastewater stream in terms of the decrease in total household load arriving at the WWTP decreases from 2.2% to 1.2%e0.2% as the applied greywater treatment efficiencies increase from 10% to 50%e90% (Table 4). This can be explained by the fact that the higher treatment removal efficiencies (i.e. 50% and 90%) produce treated greywater with lower Cd concentrations, and hence the proportion of Cd removed from the total WWTP system due to losses via irrigation is reduced. If it is feasible to remove the sludge produced by the greywater treatment system from the external wastewater stream, it can be seen that all scenarios (other than A, K and L) produce overall Cd removal efficiencies which are consistent with the expected results based on the applied greywater treatment values. For 10% greywater treatment efficiency, the most efficient overall Cd removal is demonstrated by Scenario G (11.3%) whereas for the higher greywater treatment performances Scenario J proves to be most efficient (27.8% and 42.8%). Scenarios G and J both involve continuous recycling of laundry greywater and the results in Table 4 are based on predictions after the completion of 5 cycles. All scenarios incorporating laundry water recycling (Scenarios E, G, H and J) involve micropollutants being continually added to the system and the wastewater being continually circulated and treated for reuse. The calculations indicate that the Cd concentration in these systems initially increases but approaches an equilibrium situation with regard to the greywater Cd loading and an optimal removal efficiency is established within 5 cycles or less. This suggests that there should not be any detrimental impact on washing machine functioning due to micropollutant build-up although the elevated pH levels during typical laundry washing may encourage the precipitation of some constituents and corrosion may occur due to increased salinity. The annual influent loads of Cd, Ni, Pb, benzene and 4-NP to the Lynetten WWTP, which services the area of Copenhagen where the Nordhavnsga˚rden greywater treatment plant is located, are 21 kg, 386 kg, 1064 kg, 12.6 kg and 178 kg (Lynettefællesskabet I/S, 2008). Because of the differences in influent flows (5.7 m3/year to Nordhavnsga˚rden greywater treatment plant compared to 74 million m3/year to the WWTP), the contributions deriving from untreated Nordhavnsga˚rden

1556

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 4 9 e1 5 6 0

Box 1 Cadmium fate calculations for greywater treatment and reuse according to Scenario B (based on 90% removal efficiency).

[A] With an estimated bathroom greywater flow rate of 42.8 l p1 d1 (based on DANVA (2007) and Kjellerup and Hansen, 1994) and a median measured Cd concentration in the Nordhavnsga˚rden bathroom greywater of 0.071 mg l1, the median Cd load in untreated bathroom greywater is 3.04 mg p1 d1. [B] Assuming a greywater treatment removal efficiency of 90%, the maximum effluent Cd loading will be 0.30 mg p1 d1. The remaining Cd (2.74 mg p1 d1) will be entrained in the sludge produced by the greywater treatment system. The greywater treatment effluent has a Cd concentration of 0.0071 mg l1 (0.30 mg p1 d1 O 42.8 l p1 d1). [C] As with most treatment systems of this type the sludge produced at the Nordhavnsga˚rden treatment plant is periodically transferred directly to the municipal WWTP without further pre-treatment. [D] The Cd loading in the treated water used for toilet flushing is 0.19 mg p1 d1 (27.4 l p1 d1  0.0071 mg l1). Additionally, Cd could be added due to the addition of faeces and urine at this stage. Based on published measurements of Cd in blackwater (Palmquist and Hanaeus, 2005) it is estimated that the concentration of Cd in toilet wastewater would be 0.4 mg l1. Therefore, in a volume of 27.4 l, the maximum Cd loading contribution from the addition of blackwater would be 10.96 mg p1 d1. Hence, the total Cd load which would be discharged to the WWTP upon toilet flushing is 11.15 mg p1 d1 (0.19 þ 10.96 mg p1 d1). [E] Under Scenario B, surplus greywater treatment effluent (i.e. treated greywater not required for toilet flushing) will be discharged directly to the WWTP. The surplus flow rate is 15.4 l p1 d1 and the Cd concentration is 0.0071 mg l1 which equates to a Cd loading of 0.11 mg p1 d1. [F] The total Cd load discharged to the WWTP after greywater treatment and reuse is 14.00 mg p1 d1 (2.74 þ 11.15 þ 0.11). The three contributing sources to this Cd load are sludge [C], reused water after toilet flushing [D] and surplus treated water [E]. Under this scenario, additional household Cd releases will also occur due to laundry washing or kitchen activities as these waste streams are discharged directly to the WWTP. The relevant Cd loads from these sources are estimated to be 4.65 mg p1 d1from the laundry greywater and 1.58 mg p1 d1 from kitchen greywater (1.16 mg p1 d1 for dishwashing þ 0.26 mg p1 d1 from sink wiping þ 0.16 mg p1 d1 from food preparation) (Wall, 2002). Therefore the total Cd load to the wastewater treatment plant would be 20.23 mg p1 d1 (14.00 þ 4.65 þ 1.58). Impact: The total household Cd load without greywater treatment (Scenario A) is estimated to be 20.23 mg p1 d1 (comprising 3.04 mg p1 d1 from bathroom greywater, 4.65 mg p1 d1 from laundry greywater, 1.58 mg p1 d1 from kitchen greywater, and 10.96 mg p1 d1 from toilet wastewater). Therefore, as expected, under Scenario B there will be no decrease in Cd loading going to the WWTP unless the greywater sludge is removed from the system and treated separately. If this was practised, it would equate to a decrease in WWTP influent Cd loading of 2.74 mg p1 d1 and a potential overall per capita Cd removal efficiency of 13.5%.

greywater are very low, typically of the order of 0.001%. Therefore, clearly in terms of assessing the benefits which could be accrued by comprehensive application of greywater treatment, it is more realistic to compare per capita pollutant reductions. On this basis, the results reveal that full implementation of the most effective scenario (i.e. Scenario J with full greywater treatment and recycling and separate sludge disposal) could lead to a calculated reduction in the Cd load to the WWTP of 8.6 mg p1 d1 which is equivalent to a reduction of 14.1% of the overall Cd influent load at the WWTP (61 mg p1 d1). Although this is relatively low, it is apparent that in areas of low industrial activity and/or with separate stormwater treatment (i.e. where household wastewater is the major contributor to the municipal WWTP influent), the introduction of greywater treatment and reuse technologies may be beneficial in terms of pollutant emission control as well as water conservation. Clearly, the magnitude of the emission control function in relation to micropollutants will be highly dependent on the greywater sludge disposal pathway. The results presented in Tables 4 and 5 show that even when greywater treatment removes a substantial proportion of micropollutants from influent

greywater, for elemental pollutants such as Cd, Ni and Pb and for hydrophobic substances such as 4-NP the resulting impact at the WWTP is highly dependent on the fate of the greywater treatment sludge. In Table 5, the results derived for the bathroom greywater reuse scenarios are presented for two metals (Ni and Pb) and two organic micropollutants (benzene and 4-NP), respectively. Both metals follow similar trends to those described for Cd although with considerably elevated loading values. The magnitude of the differences in pollutant reductions according to the disposal route of the greywater treatment sludge are indicative of the adsorption potentials of different pollutants and are clearly less significant for benzene for which volatilisation plays an important role in controlling pollutant removal from the aqueous phase. The results for benzene and 4-NP shown in Table 5 have been informed by apportioning the contributions to the different removal processes during greywater treatment according to the distribution calculated using a pollutant fate model for an activated sludge WWTP (STPWIN, EPI Suite v3.20, US EPA, 2007). As expected from a consideration of the physicochemical properties, only 1.1%

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 4 9 e1 5 6 0

Table 5 e Implications of Scenarios A-C for Ni, Pb, benzene and 4-nonylphenol loads in bathroom greywater treatment sludge and household wastewater, assuming greywater removal efficiencies of 10%, 50% and 90%. Reduction in load to WWTP (mg p1 d1)a Scenario A Scenario B Scenario C Ni

Pb

Benzene

4-NP

10% removal

e

50% removal

e

90% removal

e

10% removal

e

50% removal

e

90% removal

e

10% removal

e

50% removal

e

90% removal

e

10% removal

e

50% removal

e

90% removal

e

0 (29.1) 0 (145.3) 0 (261.6) 0 (29.3) 0 (146.6) 23.87 (170.5) 10.58 (10.8) 52.89 (53.8) 95.20 (96.8) 0.03 (3.9) 0.17 (19.4) 0.31 (34.8)

42.59 (71.7) 23.66 (169.0) 4.73 (266.3) 42.98 (72.3) 0 (263.9) 4.76 (268.7) 26.33 (26.5) 61.64 (62.5) 96.95 (98.5) 5.70 (9.5) 3.32 (22.5) 0.94 (35.6)

a Main value given is the reduction in load in mg p1 d1 assuming the greywater treatment sludge is discharged to the WWTP; values in brackets show the reduction in load assuming the greywater treatment sludge is removed from the wastewater stream. Removal due to sorption, volatilisation and biodegradation is apportioned according to the distribution calculated using STPWIN (EPI Suite v3.20, US EPA, 2007).

of benzene is predicted to be removed by adsorption to sludge with volatilisation representing the major removal route (67.8%) in an overall removal capability of 68.9%. This raises concerns regarding the overall environmental effectiveness of greywater treatment as an emission control barrier for benzene. In contrast, 4-NP which has a low volatility (<1% removal by volatilisation) is predicted to partition predominantly to the sludge (90% removal by adsorption) and therefore behaves in a similar way to the metals placing the fate of this pollutant firmly on the adopted sludge disposal route during greywater treatment. Both benzene and 4-NP are identified as possessing low potentials for removal by biodegradation (<1%). Scenarios K and L investigate the potential implications of land-based greywater treatment systems. Under these scenarios, the greywater is treated using reedbed technology resulting in advantageous overall reductions in terms of the municipal WWTP influent pollutant load, but also raising concerns regarding the possible environmental impacts. For example, under Scenario K, the removal of bathroom greywater for treatment in a reedbed equates to a decrease in Cd

1557

WWTP influent loading of 3.04 mg p1 d1. Therefore, the reduction in Cd being directed to the WWTP due to this greywater treatment scenario is 15.0%. According to Scenario L, in which both bathroom and laundry greywater are treated, the corresponding reduction in WWTP influent load is 38.4%. In both cases, it is important to consider the environmental implications. Depending on the substrate of the treatment system, Cd may build up in the sediment/soil/solid phase over time and may also leach through to the groundwater. For the Nordhavnsga˚rden greywater treatment plant the annual release of Cd to the environment would be 130.4 mg and 329.0 mg for Scenarios K and L, respectively. A median wet weather removal efficiency of 84.7% has been measured for Cd passing through a sub-surface constructed wetland (Revitt et al., 2004). If applied to Scenario K this would indicate that a discharge loading of 3.04 mg p1 d1 could be reduced to 0.46 mg p1 d1 after passing through an appropriately designed vegetated greywater treatment plant. Given the hydraulic loading rate of 42.8 l p1 d1, this corresponds to a discharge concentration of 0.011 mg l1 which is well below the proposed AA-EQS value (European Commission, 2008) for Cd for the most sensitive inland surface waters (0.08 mg l1) before any dilution has occurred within the receiving water. By contrast for Scenario L, the discharge of 7.69 mg p1 d1 at a hydraulic loading of 59.5 l p1 d1 corresponds to a discharge concentration of 0.13 mg l1. Treated greywater with this Cd concentration would require an appropriate dilution on entering the receiving water. More critically, if discharged to ground the adsorption characteristics of the soil would need to ensure that appropriate protection existed for an underlying aquifer.

3.4.

Sludge fate and pollutant loading

One of the major drivers for further reducing micropollutant influent loads to municipal WWTPs is to facilitate the beneficial reuse of sewage sludge (i.e. biosolids) for soil conditioning of agricultural land. The European Directive most pertinent to the agricultural use of sewage sludge is Directive 86/278/EEC (European Commission, 1986) which establishes concentration limits for a number of metals that are typically present within sludge. The concentration limits are effectively ceiling limits, meaning that if sludge exceeds the metal concentration limit for any of the listed metals it should not be permitted for land application. Directive 86/278/EEC is currently under revision and the working draft for the revised Directive indicates that future limits will be more conservative (European Commission, 2000b). To enable some member states to achieve the new limits, it is probable that water companies will need to further tighten trade effluent consents for industries as well as seeking further means of reducing WWTP influent loads of key pollutants. The alternative would be an unwanted reduction in land recycling of sludge and a waste of a potentially beneficial resource. Currently, some member states, including Denmark, impose more stringent requirements than those in the EC Directive. For example, the current limit for Cd in the Danish regulations is 0.8 mg/kg DW compared to 20 mg/kg DW in the EC Directive and for nonylphenols the Danish value of 10 mg/kg DW is considerably

1558

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 4 9 e1 5 6 0

Table 6 e Measured concentrations of Cd, Ni and Pb in Nordhavnsga˚rden greywater treatment sludge and Danish wastewater treatment plant sludge, together with Danish and European sludge guideline limits for the relevant substances. All values are given in mg kgL1 DW. Substance

Measured concentration in Nordhavnsga˚rden primary settling tank sludgea

Concentration in Danish WWTP sludgeb

Cd

Range: 0.7e1.2 Mean: 1.0 Median: 1.1 Range: 22e35 Mean: 27 Median: 24 Range: 34e45 Mean: 37.7 Median: 34.0 No data

1995: 1.5 (0.8e6.0) 2002: 1.3 (0.3e3.2)

Ni

Pb

Nonylphenols

0.8

1995: 25.7 (10e141) 2002: 20 (11e42) 1995: 72 (26e155) 2002: 50 (11e96) 1995: 8 (0.3e61) 2002: 4 (1e25)

Danish sludge guideline limits (mg kg1)c

European sludge guideline limitsd, f 20e40

Proposed European sludge guideline limits in working drafte, f 10

30

300e400

300

120

750e1200

750

10

N/A

50

a n ¼ 3, 1 sample was taken from the primary settling tank and 2 samples were taken from the biological treatment module. b Values given are derived from a national survey of sludge quality in Danish WWTPs and are shown as median values, with the 5th and 95th percentiles in brackets (Jensen and Jepsen, 2005). c Cited in Jensen and Jepsen (2005). d Directive 86/278/EEC (European Commission, 1986). e Working document on sludge, 3rd draft (European Commission, 2000a,b). f Limit value applies to the substances nonylphenol and nonylphenolethoxylates with 1 or 2 ethoxy groups.

lower than a proposed European sludge guideline limit of 50 mg/kg DW (Table 6). Measured concentrations in the greywater treatment plant sludge from Nordhavnsga˚rden are provided in Table 6. The measured metal concentrations in the Nordhavnsga˚rden greywater treatment sludge confirm that adsorption to suspended solids is an important removal process for these substances during treatment. With median sludge concentrations of 1.1, 24 and 34 mg kg1 DW for Cd, Ni and Pb respectively it is evident that removal of greywater treatment sludge from the WWTP influent load could assist in the reduction of metal loadings in municipal WWTP sludge. The separate treatment and disposal of greywater sludge is an attractive prospect because it is unlikely to contain a significant nutrient content, and yet does effectively concentrate unwanted substances such as metals and nonylphenols. The separation of the greywater treatment sludge from community scale treatment and reuse systems is feasible and could effectively reduce WWTP sludge metal loads without significantly impacting on sludge nutrient value. In contrast, sludge separation from single household system designs is unlikely to be practical and currently these systems are typically designed to periodically backwash or flush particulate matter to the sewerage system.

4.

Conclusions

The results of the conducted scenario analyses are important in the face of increasing pressures on potable water supplies, showing that greywater recycling can potentially save significant volumes of potable water. Within a greywater treatment plant, the dominant removal process for a particular pollutant

is heavily dependent on the physical, chemical and biological properties of that pollutant. For example, some substances will be more readily biodegraded than others, and some substances will be more susceptible to sorption or volatilisation. The potential for the greywater treatment and reuse system to act as a pollutant emission barrier is thus highly substance dependent. In general, a system such as that installed at Nordhavnsga˚rden will only act as a significant pollutant barrier for substances which are readily biodegradable (but this is not the case for most PS/PHS and certainly not for metals). Thus, on the basis of current designs, which typically do not facilitate separate treatment and disposal of greywater treatment sludge, the results indicate that the potential for extra benefits associated with the emission control of xenobiotics are likely to be quite limited. On the other hand, if greywater treatment sludge were to be removed from the wider municipal WWTP load this could potentially improve the sludge quality and hence help meet the requirements of the various national and European sludge regulations.

Acknowledgements The presented results have been obtained within the framework of the ScorePP project - “Source Control Options for Reducing Emissions of Priority Pollutants”, contract no. 037036, a project coordinated by the Department of Environmental Engineering, Technical University of Denmark, within the Energy, Environment and Sustainable Development section of the European Community’s Sixth Framework Programme for Research, Technological Development and Demonstration. COST Action 636 ‘Xenobiotics in the Urban Water Cycle’ is also acknowledged.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 4 9 e1 5 6 0

references

Abad, E., Martinez, K., Planas, C., Palacios, O., Caixach, J., Rivera, J., 2005. Priority organic pollutant assessment of sludges for agricultural purposes. Chemosphere 61 (9), 1358e1369. Andersson, M., Dalsgaard, A., 2004. Demonstrationsprojekt med genanvendelse af gra˚t spildevand fra en større gra˚vandsproducent Økologisk byfornyelse og spildevandrensning Nr. 46 (accessed 02 09 10) at: http://www2.mst.dk/Udgiv/publikationer/2004/87-7614149-7/html/default.htm (In Danish). APHA, AWWA, WEF, 2005. Standard methods for the examination of water and wastewater. American Water Works Association, and Water Environment Federation, first ed. American Public Health Association, Washington D.C. Bergstrom, A., 2007. Cadmium sources in the area connected to a sewage treatment plant in Stockholm. Poster presentation at 5th IWA Specialised Conference on Assessment and Control of Micropollutants/Hazardous Substances in Water. Frankfurt am Main/Germany 17e20 June 2007. Bixio, D., Thoeye, C., De Koning, J., Joksimovic, D., Savic, D., Wintgens, T., Melin, T., 2006. Wastewater reuse in Europe. Desalination 187 (1e3), 89e101. Chu, J., Chen, J., Wang, C., Fu, P., 2004. Wastewater reuse potential analysis: implications for China’s water resources management. Water Research 38 (11), 2746e2756. Copenhagen Energy, 2008a. The content of your drinking water: chemical investigation of treated water (accessed 05 06 09) at: http://www.ke.dk/portal/pls/portal/docs/436003.PDF. Copenhagen Energy, 2008b. The content of your drinking water: content of xenobiotic substances in drinking water (accessed 05 06 09) at: http://www.ke.dk/portal/pls/portal/docs/436005. PDF. Danish Standard, 2003. Determination of Metals in Water, Sludge and Sediments - General Guidelines for Determination by Atomic Absorption Spectrophotometry in Flame, vol. 259. DS, pp. 2003. Danish Standard, 2004. Water quality - Determination of trace elements using atomic absorption spectrometry with graphite furnace, no. DS/EN ISO 15586:2004 DANVA(Danish Water and Wastewater Association), 2007. VAND I TAL DANVA’s Benchmarking og Vandstatistik (accessed 02 09 10) at: http://www.bessy.dk (In Danish). Donner, E., Eriksson, E., Revitt, D.M., Scholes, L., HoltenLutzhof, H.-C., Ledin, A., 2010. Presence and fate of priority substances in domestic greywater treatment and reuse systems. Science of the Total Environment 408 (12), 2444e2451. Eriksson, E., Auffarth, K.P.S., Henze, M., Ledin, A., 2002. Characteristics of grey wastewater. Urban Water 4 (1), 85e104. Eriksson, E., Auffarth, K., Eilersen, A.-M., Henze, M., Ledin, A., 2003. Household chemicals and personal care products as sources for xenobiotic organic compounds in grey wastewater. Water SA 29 (2), 135e146. European Commission, 1986. Directive 86/278/EEC of 12th June 1986 on the protection of the environment and in particular of the soil, when sewage sludge is used in agriculture (accessed 02 09 10) at: http://ec.europa.eu. European Commission, 2000a. Directive 2000/60/EC of the European Parliament and of the Council establishing a framework for the Community action in the field of water policy (accessed 02 09 10) at: http://ec.europa.eu/emvironment/ water/water-framework/index_en.html 23rd October, 2000. European Commission, 2000b. Working Document on Sludge, 3rd draft, ENV.E.3/LM (accessed 02 09 10) at:. European Commission, Brussels http://europa.eu.int/comm/ environment/sludge/report10.htm 27th April, 2000.

1559

European Commission, 2008. Directive 2008/105/EC of the European Parliament and of the Council of 16 December 2008 on environmental quality standards in the field of water policy, amending and subsequently repealing Council Directives 82/176/EEC, 83/513/EEC, 84/156/EEC, 84/491/EEC, 86/ 280/EEC and amending Directive 2000/60/EC of the European Parliament and of the Council. Fatta-Kassinos, D., Kalavrouziotis, I.K., Koukoulakis, P.H., Vasquez, M.I., 2010. The risks associated with wastewater reuse and xenobiotics in the agroecological environment. Science of the Total Environment. doi:10.1016/j.scitotenv.2010. 03.036. Friedler, E., Gilboa, Y., 2010. Performance of UV disinfection and the microbial quality of greywater effluent along a reuse system for toilet flushing. Science of the Total Environment 408 (9), 2109e2117. Giri, R.R., Takeuchi, J., Ozaki, H., 2006. Biodegradation of domestic wastewater under the simulated conditions of Thailand. Water and Environment Journal 20 (3), 169e176. Jensen, J., Jepsen, S.-E., 2005. The production, use and quality of sewage sludge in Denmark. Waste Management 25, 239e247. Kjellerup, M., Hansen, A.M., 1994. Vandbesparende foranstaltninger. ISBN 87-571-1435-9. Teknisk Forlag. Koh, Y.K.K., Lester, J.N., Scrimshaw, M.D., 2005. Fate and behaviour of alkylphenols and their polyethoxylates in an activated sludge plant. Bulletin of Environmental Contamination and Toxicology 75 (6), 1098e1106. Ledin, A., Auffarth, K., Eriksson, E., Smith, M., Eilersen, A.-M., Mikkelsen, P.S., Dalsgaard, A., Henze, M., 2006. Udvikling af metode til karakterisering af gra˚t spildevand. Økologisk byfornyelse og spildevandsrensning 58 (accessed 05 06 10) at: http://www.mst.dk/Udgivelser/Publikationer/2006/07/877052-116-6.htm. Li, F.Y., Wichmann, K., Otterpohl, R., 2009. Review of the technological approaches for grey water treatment and reuses. Science of the Total Environment 407 (11), 3439e3449. Liu, S., Butler, D., Memon, F.A., Makropoulos, C., Avery, L., Jefferson, B., 2010. Impacts of residence time during storage on potential of water saving for grey water recycling system. Water Research 44 (1), 267e277. Lynettefællesskabet I/S. 2008. (accessed 12 10 10) at: http://www. lyn-is.dk/Lynettef%C3%A6llesskabet/Gr%C3%B8nt_regnskab,_ Milj%C3%B8data_og_%C3%85rsberetning.aspx. Maimon, A., Tal, A., Friedler, E., Gross, A., 2010. Safe on-site reuse of greywater for irrigationeA critical review of current guidelines. Environmental Science and Technology 44 (9), 3213e3220. Masi, F., El Hamouri, B., Shafi, H.A., Baban, A., Ghrabi, A., Regelsberger, M., 2010. Treatment of segregated black/grey domestic wastewater using constructed wetlands in the Mediterranean basin: the zer0-m experience. Water Science and Technology 61 (1), 97e105. McLaughlin, M.J., Hamon, R.E., McLaren, R.G., Speir, T.W., Rogers, S.L., 2000. A bioavailability-based rationale for controlling metal and metalloid contamination of agricultural land in Australia and New Zealand. Australian Journal of Soil Research 38 (6), 1037e1086. Review. Memon, F.A., Butler, D., 2006. Water consumption trends and demand forecasting techniques. In: Butler, D., Memon, F.A. (Eds.), Water Demand Management, vol. 2006. IWA Publishing, pp. 1e26. Misra, R.K., Patel, J.H., Baxi, V.R., 2010. Reuse potential of laundry greywater for irrigation based growth, water and nutrient use of tomato. Journal of Hydrology 386 (1e4), 95e102. Nielsen, M., Pettersen, T., 2005. Genanvendelse af gra˚t spildevand pa˚ campingpladser - Fase 2 og 3 Økologisk byfornyelse og spildevandsrensning no. 57. Report to the Danish EPA

1560

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 4 9 e1 5 6 0

(accessed 02 09 10) at: http://www2.mst.dk/Udgiv/ publikationer/2005/87-7614-899-8/html/default.htm. Palmquist, H, 2004. Hazardous substances in wastewater management, PhD thesis. Luleaˆ University of Technology, Sweden. 2004:47, ISSN: 1402e1544. http://epubl.luth.se/14021544/2004/47/LTU-DT-0447-SE.pdf, (accessed 02 09 10) Palmquist, H., Hanaeus, J., 2005. Hazardous substances in separately collected grey- and blackwater from ordinary Swedish households. Science of the Total Environmen 348 (1e3), 151e163. Paulo, P.L., Begosso, L., Pansonato, N., Shrestha, R.R., Boncz, M.A., 2009. Design and configuration criteria for wetland systems treating greywater. Water Science and Technology 60 (8), 2001e2007. Pidou, M., Avery, L., Stephenson, T., Jeffrey, P., Parsons, S.A., Liu, S., Memon, F.A., Jefferson, B., 2008. Chemical solutions for greywater recycling. Chemosphere 71 (1), 147e155. Pinto, U., Maheshwari, B.L., 2010. Reuse of greywater for irrigation around homes in Australia: understanding community views, issues and practices. Urban Water Journal 7 (2), 141e153.

Revitt, D.M., Shutes, R.B.E., Jones, R.H., Forshaw, M., Winter, B., 2004. The performances of vegetative treatment systems for highway runoff during dry and wet conditions. Science of the Total Environment 334, 261e270. Revitt, D.M., Scholes, L., Ellis, J.B., 2008. A pollutant removal prediction tool for stormwater derived diffuse pollution. Water Science and Technology 57 (8), 1257e1264. Scholes, L., Revitt, M., Gasperi, J., Donner, E., 2007. Priority pollutant behaviour in stormwater Best management practices (BMPs). Deliverable 5.1, ScorePP project (accessed 22 10 10) at: http://www.scorepp.eu/index.php?option¼com_ content&task¼view&id¼30&Itemid¼56. US EPA(United States of America Environmental Protection Agency), 2007. EPI Suite v3.20 (February 2007) (accessed 02 09 10) at: http://www.epa.gov/opptintr/exposure/pubs/episuite.htm. Wall, E., 2002. Kadmium i husha˚llsspillvatten. Stockholm Vatten (Stockholm Water) Report No. 9, April 2002. (accessed on 12 10 10) at: http://www.stockholmvatten.se/commondata/ rapporter/avlopp/Processer/Kadmium_spillvatten.pdf (In Swedish).

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 6 1 e1 5 7 0

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Investigating the decay rates of Escherichia coli relative to Vibrio parahemolyticus and Salmonella Typhi in tropical coastal waters Choon Weng Lee a,*, Angie Yee Fang Ng a, Chui Wei Bong a, Kumaran Narayanan b, Edmund Ui Hang Sim c, Ching Ching Ng a a

Laboratory of Microbial Ecology, Institute of Biological Sciences, University of Malaya, 50603 Kuala Lumpur, Malaysia School of Science, Monash University, Sunway Campus, Selangor, Malaysia c Department of Molecular Biology, Faculty of Resource Science and Technology, University Malaysia Sarawak, Malaysia b

article info

abstract

Article history:

Using the size fractionation method, we measured the decay rates of Escherichia coli, Salmo-

Received 24 June 2010

nella Typhi and Vibrio parahaemolyticus in the coastal waters of Peninsular Malaysia. The size

Received in revised form

fractions were total or unfiltered, <250 mm, <20 mm, <2 mm, <0.7 mm, <0.2 mm and <0.02 mm.

18 November 2010

We also carried out abiotic (inorganic nutrients) and biotic (bacterial abundance, production

Accepted 19 November 2010

and protistan bacterivory) measurements at Port Dickson, Klang and Kuantan. Klang had

Available online 27 November 2010

highest nutrient concentrations whereas both bacterial production and protistan bacterivory rates were highest at Kuantan. We observed signs of protistebacteria coupling via the

Keywords:

following correlations: Protistan bacterivory
Bacterial Production: r ¼ 0.773, df ¼ 11, p < 0.01;

Bacterial decay rate

Protist
Bacteria: r ¼ 0.586, df ¼ 12, p < 0.05. However none of the bacterial decay rates were

Size fractionation

correlated with the biotic variables measured. E. coli and Salmonella decay rates were

Top-down control

generally higher in the larger fraction (>0.7 mm) than in the smaller fraction (<0.7 mm) sug-

Straits of Malacca

gesting the more important role played by protists. E. coli and Salmonella also decreased in the

South China sea

<0.02 mm fraction and suggested that these non-halophilic bacteria did not survive well in seawater. In contrast, Vibrio grew well in seawater. There was usually an increase in Vibrio after one day incubation. Our results confirmed that decay or loss rates of E. coli did not match that of Vibrio, and also did not correlate with Salmonella decay rates. However E. coli showed persistence where its decay rates were generally lower than Salmonella. ª 2010 Elsevier Ltd. All rights reserved.

1.

Introduction

Coastal waters account for less than 10% of the ocean area. However they are highly productive and account for 25% of primary production in the ocean (Berger et al., 1989). Coastal waters are increasingly exploited by humans for food, recreation, transport and other needs, and at present most are in various stages of degradation (Alongi, 1998). There is also an increasing public health threat from pathogens (Hazen and

Toranzos, 1990; Moe, 1997). Disposal of inadequately treated waste is considered faecal pollution and a main source of bacterial pathogens in the sea (Solo-Gabriele et al., 2000). For faecal pollution studies, the concept of bacterial indicator is standard (Wolf, 1972). A fundamental assumption to this concept is the parity in the survival of indicator and enteric pathogens over a wide range of aquatic environments (Bonde, 1977). It is however acknowledged that these indicators are inadequate to predict the presence of pathogenic

* Corresponding author. E-mail address: [email protected] (C.W. Lee). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.11.025

1562

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 6 1 e1 5 7 0

microorganisms (Rhodes and Kator, 1988; Kay et al., 1994; Borrego and Figueras, 1997). At present, the coliform group of microorganisms or microorganisms found in the intestines of all warm blooded animals especially Escherichia coli is used as a standard indicator of faecal contamination in many countries (Hazen and Toranzos, 1990) including Malaysia (Department of Environment, 2008). The value of E. coli as an indicator microorganism is significantly enhanced by the ease with which it can be detected and cultured (Wolf, 1972) when compared with other bacterial pathogens. In order to use E. coli as an indicator bacterium in seawater, knowledge of its survival must be acquired as it is not a halophilic microorganism. Loss rates of indicator microorganisms pose an interesting problem especially in coastal waters as many factors affect the survival times of E. coli for example temperature, light, salinity, predation, nutrients and pollutants (Fujioka et al., 1981; Munro et al., 1989; Presser et al., 1998; Solo-Gabriele et al., 2000; Rozen and Belkin, 2001; Sinton et al., 2002). Since the 1950s, Carlucci and Pramer (1959; 1960a; 1960b) have studied the survival of bacteria including E. coli in seawater. Numerous studies have shown that decay rates of E. coli do not reflect Vibrio cholerae in estuarine waters and also Salmonella spp. (Colwell et al., 1981; Rhodes and Kator, 1988). However to the best of our knowledge, there is no study that compares the relative loss rates of E. coli with both halophilic and non-halophilic bacterial pathogens. In this study, we investigated whether the decay rates for E. coli were similar to a non-halophilic bacterial pathogen (i.e. Salmonella Typhi) and a halophilic bacterial pathogen (i.e. Vibrio parahaemolyticus)? Our results could help properly evaluate the risk posed by such bacteria either to the health of bathers in recreational waters or to the safety of fisheries or marine aquaculture.

2.

Materials and methods

2.1.

Sampling

We sampled at three stations located on the east (Kuantan: 03 48.40 N 103 20.60 E) and west (Klang: 03 00.10 N 101 23.40 E and Port Dickson: 02 29.50 N 101 50.30 E) of Peninsular Malaysia from April until October 2006 (Fig. 1). The stations at Kuantan and Klang were located in estuaries whereas the station at Port Dickson was sandy coast. Surface seawater samples were collected during high tide, using an acid-cleaned bucket, and in-situ measurements of temperature (0.1  C) and salinity (0.1 ppt) were carried out using a salinometer (YSI-30, US). Seawater samples were then kept in a cooler box for no more than 4 h until processing in the laboratory. In the laboratory, seawater samples for dissolved nutrient analyses were filtered through pre-combusted (450  C for 5 h) Whatman GF/F filters, and stored at
20  C until analysis.

2.2.

Environmental conditions

Dissolved inorganic nitrogen (nitrate (NO3), nitrite (NO2), ammonium (NH4)), and phosphate (PO4) concentrations were measured using a spectrophotometer (Parsons et al., 1984). All nutrient measurements above were carried out in triplicates.

Fig. 1 e Map showing the location of the sampling sites, east (Kuantan) and west (Klang and Port Dickson) of Peninsular Malaysia.

Coefficient of variation (CV) for NH4, NO2 and PO4 analyses were <5%, and <10% for NO3 analysis. Bacteria and protist were determined using the direct count method by an epifluorescence microscope (Olympus BX60, Japan) with a U-MWU filter cassette (excitor 330
385 nm, dichroic mirror 400 nm, barrier 420 nm). For protist, 10 ml sample was filtered onto a black 0.8 mm pore size Isopore filter (Millipore, Ireland), and then stained with the fluorochrome primulin (40 mg ml
1 final concentration) for 5 min (Bloem et al., 1986) whereas for bacteria, 2 ml sample was filtered onto a black 0.2 m pore size Isopore filter, and then stained with 40 6diamidino-2-phenylindole (DAPI, 1 mg ml
1 final concentration) for 10 min (Kepner and Pratt, 1994). Slides were kept frozen for < 3 days before enumeration. A minimum of 10 microscope fields or 500 cells were counted for bacteria and for protist, at least 30 microscope fields were observed.

2.3.

Bacterial decay rates

For bacterial decay experiments, seawater samples were sizefractionated as total or unfiltered, <250 mm (through a 250 mm stainless steel mesh), <20 mm (20 mm pore size nylon mesh), <2 mm (2.0 mm polycarbonate membrane filter), <0.7 mm (GF/F filter), <0.2 mm (0.2 mm polycarbonate membrane filter) and <0.02 mm (0.02 mm Whatman Anodisc). The size fractions were inoculated separately with 1% (v/v) fresh cultures of E. coli, Salmonella Typhi (hereafter referred to as Salmonella) and V. parahaemolyticus (hereafter referred to as Vibrio), and then incubated at 30  C for about three days. Both E. coli and Salmonella inocula were prepared on nutrient broth whereas

1563

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 6 1 e1 5 7 0

for Vibrio, 3% NaCl (final concentration) was added to the nutrient broth. From the inoculated size fractions, the initial E. coli, Salmonella and Vibrio concentration was determined as colony forming unit (cfu) ml
1 via spread plating. The number of cfus was observed one day after spread plating on MacConkey agar (for E. coli and Salmonella) and Thiosulfate Citrate Bile Salts Sucrose (TCBS) agar with 3% NaCl (for Vibrio). A portion of the sample was also removed every day to determine the change in E. coli, Salmonella and Vibrio concentration. Bacterial decay rate was modeled by the standard differential equation: dN/dt ¼
kN (Chick, 1908) where k ¼ decay constant, N is the cfu ml
1, and t ¼ time of incubation. Integrating the differential equation gave the following ln N ¼
kt þ C. Hence for decay rate assessment, we transformed the cfu data by natural logarithm and plotted the ln cfu ml
1 against incubation time. Linear regression analysis was then used to find the best-fit slope i.e. decay rate. We also measured separately, the concurrent bacterial production and protistan bacterivory rates by measuring bacterial growth rate in both <0.7 mm and <20 mm fractions after 12 h incubation. Bacterial growth rate in each fraction was calculated as the increase in natural logarithmic bacterial abundance over time. The bacterial growth rate in the <0.7 mm fraction was assumed without grazing (m0.7) whereas growth rate in the <20 mm fraction (m20) was the product of both growth and grazing. Bacterial production (BP) was then estimated by the following equation: BP ¼ Bacterial abundance  m0.7 (Lee et al., 2009a) whereas bacterivory rate was estimated by Bacterial abundance  (m0.7
m20) (McManus, 1993).

2.4.

Statistical analyses

Statistical tests such as coefficient of variation (CV), analysis of variance, Student’s t-test, Tukey’s test, linear regression and correlation analyses were carried out according to Zar (1999). Count data were log-transformed to meet parametric assumptions of equality of variances and normal distribution before correlation and linear regression analyses. All data, unless mentioned otherwise, were reported as mean  S.D.

3.

Results

3.1.

Abiotic measurements

Table 1 shows the physico-chemical parameters measured at the three stations. Average surface seawater temperature

observed ranged from 29 to 30  C whereas average salinity measured at the estuaries (Klang and Kuantan) was lower than at Port Dickson. Salinity fluctuated over a wider range at the estuaries (CV ¼ 7% and 46%) than Port Dickson (CV ¼ 1%). Dissolved inorganic nitrogen (DIN) measured showed that concentrations at Klang were more than two-fold higher than both Kuantan and Port Dickson. Among the nitrogen species, NH4 was the dominant species at both Klang and Kuantan, accounting for > 70% of DIN. At Port Dickson, NO3 was the dominant species (about 45% of DIN). Average PO4 concentration at Klang was also higher than at both Port Dickson and Kuantan.

3.2.

Biotic measurements

Bacterial abundance ranged from 0.85 to 3.49  106 cell ml
1, and was significantly higher at Kuantan than both Klang (q ¼ 4.77, df ¼ 12, p < 0.05) and Port Dickson (q ¼ 5.77, df ¼ 12, p < 0.01) (Table 2). For protist, abundance ranged from 0.58 to 6.64  103 cell ml
1, and was about three orders lower than bacteria. Protist abundance at Kuantan was significantly higher than Port Dickson (q ¼ 4.34, df ¼ 12, p < 0.05) but not Klang. Table 2 also shows the bacterial production measured in this study. Bacterial production at both Klang and Port Dickson ranged from 0.72 to 1.69  105 cell ml
1 h
1 whereas bacterial production at Kuantan ranged from 1.99 to 4.78  105 cell ml
1 h
1. Similar to patterns exhibited by microbial abundance data, bacterial production at Kuantan was significantly higher than both Klang (q ¼ 4.69, df ¼ 12, p < 0.05) and Port Dickson (q ¼ 5.79, df ¼ 12, p < 0.01). Protistan bacterivory at Kuantan was also higher than both Klang (q ¼ 14.48, df ¼ 9, p < 0.001) and Port Dickson (q ¼ 13.64, df ¼ 9, p < 0.001). Bacterivory rates at Kuantan ranged from 1.60 to 2.15  105 cell ml
1 h
1 whereas bacterivory rates at both Klang and Port Dickson were about one order lower, and ranged from 1.11 to 4.19  104 cell ml
1 h
1. E. coli concentration from 1990 until 2005 were also obtained from the Department of Environment Malaysia monitoring stations located near our study area. Although E. coli concentration varied over four-order at all three locations (Fig. 2), analysis of variance showed significant differences among the different locations (F ¼ 19.2, df ¼ 1095, p < 0.001). At Port Dickson and Klang, average E. coli was 1300  3600 MPN (Most Probable Number) per 100 ml, and was significantly higher than Kuantan (640  1900 MPN per 100 ml) (KlangeKuantan: q ¼ 8.52, df ¼ 1095, p < 0.001; Port DicksoneKuantan: q ¼ 6.38, df ¼ 1095, p < 0.001).

Table 1 e Physico-chemical characteristics of the sampling stations in this study. Mean (±S.D.) of surface water temperature, salinity, ammonium (NH4), nitrite (NO2), nitrate (NO3), and phosphorus (PO4). Station Klang (n ¼ 4) Port Dickson (n ¼ 5) Kuantan (n ¼ 5)

Temperature  C

Salinity ppt

NH4 mM

NO2 mM

NO3 mM

PO4 mM

30.3  0.5 30.0  1.0

27.7  1.9 28.4  0.3

17.84  24.75 0.42  0.51

2.97  0.64 0.18  0.27

1.24  0.61 0.50  0.28

1.66  1.54 0.12  0.05

29.1  0.8

22.4  10.2

2.89  2.64

0.51  0.29

0.61  0.33

0.60  0.31

1564

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 6 1 e1 5 7 0

Table 2 e Bacterial abundance, protist abundance, bacterial production and protistan bacterivory rates measured in this study. Date

Bacteria (106 cell ml
1)

Protist (103 cell ml
1)

Bacterial production (105 cell ml
1 h
1)

Protistan bacterivory (104 cell ml
1 h
1)

Klang

24-Apr-06 12-Jun-06 26-Jun-06 31-Jul-06 Average (S.D.)

1.13 0.93 1.29 1.53 1.22  0.26

2.89 4.08 2.57 1.08 2.65  1.23

1.69 0.97 1.66 1.64 1.49  0.35

1.55 1.11 3.70 2.77 2.28  1.18

Port Dickson

04-Jul-06 03-Oct-06 10-Oct-06 17-Oct-06 30-Oct-06 Average (S.D.)

1.10 0.85 0.91 1.11 0.98 0.99  0.12

2.43 1.16 2.01 0.58 1.73 1.58  0.73

e 0.72 1.25 1.36 1.08 1.10  0.28

e e 3.41 4.19 1.95 2.26  2.07

Kuantan

18-Apr-06 17-May-06 19-Jun-06 25-Jul-06 15-Aug-06 Average (S.D.)

2.91 1.31 2.19 2.40 3.49 2.46  0.82

3.02 1.76 5.42 6.64 5.51 4.47  2.01

4.78 1.99 3.60 2.50 2.84 3.14  0.11

21.45 e 16.05 e 16.00 17.83  3.13

Station

3.3.

Bacterial decay rates at different size fractions

Fig. 3 shows the change in E. coli concentration after inoculation in the different fractions of seawater collected from Port Dickson, Kuantan and Klang. There was an apparent decrease in E. coli in most of the larger fraction seawater (i.e. total, <250 mm, <20 mm, <2 mm) whereas in the smaller fraction (i.e. <0.7 mm, <0.2 mm and <0.02 mm), the concentration of E. coli often did not change significantly. Fig. 4 shows the statistically significant ( p < 0.05) decay rates measured in this study, and confirmed that E. coli decay rate was significantly higher in the larger fraction (>0.7 mm) (2.28  0.90 d
1) than the smaller fraction (<0.7 mm) (0.58  0.22 d
1) (Student’s t-test: t ¼ 12.35, df ¼ 64, p < 0.001). A similar trend was observed for Salmonella where its concentration generally decreased with time, especially in the larger fraction seawater (Fig. 5). Salmonella decay rates in the larger fraction (3.17  1.19 d
1) were significantly higher than in the smaller fraction (1.51  0.84 d
1) (Student’s t-test: t ¼ 6.65, df ¼ 59, p < 0.001) (Fig. 4). In contrast to E. coli and Salmonella, Vibrio exhibited a different trend (Fig. 6). There was often an increase in Vibrio concentration after one day incubation, and less discernible difference in Vibrio concentration between the larger and smaller fractions of seawater. Few of the experiments gave significant decay rates (Fig. 4), and available Vibrio decay rates ranged 0.90e1.78 d
1 at Klang, and 0.54e4.64 d
1 at Kuantan. At Port Dickson, only two significant decay rates were observed (1.36  0.21 d
1). When we compared E. coli decay rates in the larger fractions among the stations, E. coli decay rate was highest at Klang (1.96e4.90 d
1), followed by Kuantan (0.88e2.80 d
1) and Port Dickson (0.36e2.93 d
1) (KlangeKuantan: q ¼ 6.55, df ¼ 51, p < 0.001; KlangePort Dickson: q ¼ 9.99, df ¼ 51, p < 0.001; KuantanePort Dickson: q ¼ 3.44, df ¼ 51, p < 0.05). However

Fig. 2 e Long term E. coli counts (log MPN/100 ml) from selected Department of Environment monitoring stations near Port Dickson (PD) (n [ 578), Klang (n [ 196) and Kuantan (n [ 322).

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 6 1 e1 5 7 0

1565

Fig. 3 e E. coli decay (ln cfu mlL1) measured in different fractions at Port Dickson (PD) (n [ 5), Klang (n [ 3) and Kuantan (n [ 5).

there was no significant difference in Salmonella and Vibrio decay rates among the stations.

4.

Discussion

4.1.

Environmental conditions

Surface seawater temperature observed was typical of tropical waters. Klang had the highest nutrient concentrations, and reaffirmed earlier observations (Lee and Bong, 2006, 2008; Lee et al., 2009a). One possible reason is the rapid pace of

development and industrialization taking place upstream of Klang where the capital of Malaysia, Kuala Lumpur is located (Lee and Bong, 2006). In this study, the abundance of both bacteria and protist were within the range for coastal waters of Peninsular Malaysia (Lee et al., 2005; Lee and Bong, 2007, 2008). Protist abundance was about three orders of magnitude lower than bacteria, and this observation was consistent with the analysis by Sanders et al. (1992). Both bacterial production and protistan bacterivory rates measured in this study were also within the range previously published for tropical coastal waters (Lee et al., 2005, 2009a; Lee and Bong, 2006, 2007, 2008).

1566

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 6 1 e1 5 7 0

Fig. 4 e Summary of decay rates (P < 0.05) measured in this study at Klang, Port Dickson (PD) and Kuantan. Different symbols represent different sampling dates. Error bar denotes the standard error of decay rate and is shown except when smaller than symbol.

However, in contrast to the pattern exhibited by nutrient concentrations where Klang had the highest levels, all biotic variables were highest at Kuantan. It was not clear why biotic variables at Klang were not reflective of its nutrient concentrations as found in earlier studies (Lee and Bong, 2008; Lee et al., 2009a). However both Klang and Kuantan are estuaries, and can experience episodic nutrient inputs that can stimulate microbial biomass and productivity. Long term monitoring data showed the extent of faecal pollution via the indicator E. coli. Although coastal water quality in Malaysia has deteriorated over time (Department of Environment, 2008), there was no apparent increase in E. coli over the 15-year period for all three locations. At present, Malaysia has an interim marine water quality standard that set the criterion for E. coli at 100 MPN per 100 ml (Department of Environment, 2008). From a 15-year period of monitoring data, >60% of the samples at Klang and Port Dickson exceeded the standard whereas at Kuantan, 35% of the samples exceeded the standard. Dow (1995) had earlier reported higher E. coli concentration for stations along the Straits of Malacca. Our study showed that faecal pollution remains a problem for coastal waters here. One reason is many coastal communities in Malaysia lack proper sewage disposal systems and often discharge sewage directly into the sea (Law, 1992). The faecal pollution is accentuated for stations along the Straits of Malacca as the population density is higher along the west coast of Peninsular Malaysia. Although sewage treatment facilities have increased over time, the problem showed no sign of alleviation at both Klang and Port Dickson.

4.2.

Bacterial decay rates at different size fractions

The decay rates obtained in this study were within the range reported by Anderson et al. (2005) but relatively higher than decay rates reported for temperate waters (Lessard and Sieburth, 1983; Rhodes and Kator, 1988). Relative to temperate waters, the higher decay rates obtained in this study could be due to the fact that microbial activity (including protistan bacterivory) in tropical waters is at its optimum (Pomeroy and Wiebe, 2001). For both E. coli and Salmonella, decay rates were generally higher in the larger fraction (>0.7 mm) than in the smaller fraction (<0.7 mm). The main bacterial predators in the larger fractions are nanoflagellates and ciliates (Sanders et al., 1992) whereas the cause of bacterial mortality in the smaller fraction is mainly by viral lysis (Fuhrman, 2000). In the smaller fraction, lytic bacteria may also play a role albeit a minor one (Enzinger and Cooper, 1976). Bacterial decay rates measured in this study suggested that protistan bacterivory was more important, similar to the conclusion by Enzinger and Cooper (1976). A reason why viral lysis played a minor role in this study was because both Salmonella and E. coli are not natural seawater organisms and are not active in the sea (Carlucci and Pramer, 1960b). Bacteriophages require the host physiological activity in order to replicate (Pretorius, 1962). The decay rates in the <0.02 mm fraction is effectively not due to viral lysis or bacterivory as both viruses and protists do not pass through this pore size. The <0.02 mm fraction is therefore suitable to observe the response of halophilic and

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 6 1 e1 5 7 0

1567

Fig. 5 e Salmonella decay (ln cfu mlL1) measured in different fractions at Port Dickson (PD) (n [ 4), Klang (n [ 3) and Kuantan (n [ 4).

non-halophilic pathogens in seawater. We observed that Vibrio counts in the <0.02 mm fraction increased after day one in all the experiments, and no significant decay of Vibrio could be observed throughout this study. In contrast, the decay rates for E. coli and Salmonella in the <0.02 mm fraction were still statistically significant ( p < 0.05). Our results concurred with earlier reports that non-halophiles are usually not able to grow well in seawater (Gerba and McLeod, 1976). However relative to protistan bacterivory, the decay rates for E. coli and Salmonella in the <0.02 mm fraction were low, and accounted for <20% of total decay rates. These decay rates also did not correlate ( p > 0.15) with salinity, suggesting that bacterivory was more important than salinity for non-halophilic bacterial decay. There were obvious differences in the response of Vibrio in the different seawater fractions relative to both E. coli and Salmonella. In this study, significant Vibrio decay was seldom observed, and rates observed were usually lower than E. coli and Salmonella. One reason why Vibrio decay rates were lower was probably due to the ability of V. parahaemolyticus to grow in a salty environment (Holt et al., 1994). Increase in Vibrio was still apparent after one day incubation before Vibrio counts started decreasing. The latter reduction could be due to other stresses for example limited food availability that resulted in Vibrio growth rates falling below loss rates.

E. coli is widely used as an indicator for faecal pollution and for pathogenic microorganisms (Bonde, 1977). However the validity of E. coli as an indicator is questionable especially in coastal waters (Solo-Gabriele et al., 2000) where presence of E. coli is more likely the balance between supply and loss. This is because E. coli is non-halophilic and is not known to grow well in coastal waters (Gerba and McLeod, 1976). Our results confirmed that decay or loss rates of E. coli did not match that of the halophilic Vibrio and also did not correlate significantly with Salmonella decay rates ( p > 0.50) (Table 3) even though the survival characteristic of E. coli is presumed similar to Salmonella (Bonde, 1977). When we compared the different bacterial responses in the larger fractions, we found that at Klang, Salmonella and E. coli decay rates were significantly higher than Vibrio (SalmonellaeVibrio: q ¼ 6.58, df ¼ 33, p < 0.001; E. colieVibrio: q ¼ 6.16, df ¼ 33, p < 0.001) whereas at Kuantan, Salmonella decay rate was significantly higher than both Vibrio (q ¼ 7.70, df ¼ 46, p < 0.001) and E. coli (q ¼ 4.92, df ¼ 46, p < 0.01). Similarly at Port Dickson, Salmonella decay rate was higher than E. coli (Student’s t-test: t ¼ 4.10, df ¼ 29, p < 0.001). For Port Dickson, only decay rates from Salmonella and E. coli were compared as there were too few decay rates from Vibrio experiments. Generally, we found that Salmonella decay rates were higher

1568

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 6 1 e1 5 7 0

Fig. 6 e Vibrio decay (ln cfu mlL1) measured in different fractions at Port Dickson (PD) (n [ 3), Klang (n [ 3) and Kuantan (n [ 3).

than E. coli. The persistence of E. coli relative to Salmonella fulfilled one of the criteria for an indicator organism i.e. the indicator organism should survive longer than the pathogen itself (Cabelli, 1978; Allwood et al., 2003). Although our study provided snapshots of the environment, we showed consistently the importance of protistan bacterivory in bacterial decay. E. coli was a poor indicator for halophilic pathogens, and although E. coli decay rates did not correlate with Salmonella, E. coli persisted longer than Salmonella in coastal waters. Our results provided support to the continuous use of E. coli as an indicator organism for nonhalophilic pathogens especially Salmonella.

4.3.

Protistebacteria coupling

Multiple correlation analysis showed that E. coli, Salmonella and Vibrio decay rates did not correlate with each other or other biotic variables (Table 3). Protistan bacterivory also did not correlate with E. coli, Salmonella and Vibrio decay rates. As these pathogens form only a minor fraction of the total bacterial population in the sea (Lee et al., 2009b), their decay rates played only a minor role towards total bacterivory rates. However, we did observe evidences of significant coupling between protist

and bacteria similar to Sanders et al. (1992) and Lee et al. (2005) (Protistan bacterivoryeBacterial Production: r ¼ 0.773, df ¼ 11, p < 0.01; ProtisteBacteria: r ¼ 0.586, df ¼ 12, p < 0.05). Our study showed that protistan bacterivory accounted for 9e56% of bacterial production (Table 2), and was significantly higher at Kuantan (>44% bacterial production) than both Klang (q ¼ 9.68, df ¼ 9, p < 0.001) and Port Dickson (q ¼ 6.68, df ¼ 9, p < 0.01).

Table 3 e Pearson productemoment correlation coefficient (r) between variables measured i.e. bacterial production (BP, cell mlL1 hL1), log bacterial abundance (BA, cell mlL1), grazing (cell mlL1 hL1), log Protist (cell mlL1), decay rates of E. coli (dL1), Salmonella (dL1), and Vibrio (dL1). * is P < 0.05, ** is P < 0.01, *** is P < 0.001. BP

BA

BA 0.896*** Grazing 0.773** 0.794*** Protist 0.546* 0.586* E. coli
0.214
0.120 Salmonella 0.174 0.082 Vibrio
0.129
0.135

Grazing Protist E. coli Salmonella

0.295 0.291 0.120 0.064


0.190
0.121 0.077
0.276
0.040


0.106

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 6 1 e1 5 7 0

Bacterivory rates at both Klang and Port Dickson accounted for only 9e31% bacterial production. When bacterivory was less than bacterial production, other factors might have played a role in the removal of bacterial production as bacterial abundance is stable and does not change significantly with time (Lee and Bong, 2008). Among the factors that are important include viral lysis (Fuhrman, 2000), removal by benthic filter feeders (Strom, 2000), and sedimentation (Pedro´s-Alio´ and Mas, 1993). Sedimentation is often overlooked as a loss factor for bacteria due to the low sinking rates of microorganisms. However sinking speed can increase for bacteria attached to particles and there might be significant losses by sedimentation (Pedro´s-Alio´ and Mas, 1993). Although these loss factors were not investigated here, studies on viral ecology in tropical waters have suggested that viral lysis may not be an important loss factor (Bettarel et al., 2006; Cissoko et al., 2008). Moreover the higher suspended solids in the coastal waters of Peninsular Malaysia (Bong and Lee, 2008; Lee et al., 2009b) implied that sedimentation is probably more important. However further investigations are needed.

5.

Conclusion

1. Bacterial decay in tropical coastal waters is mainly due to protistan bacterivory. 2. Via the <0.02 mm fraction, our results showed that E. coli and Salmonella do not survive well in seawater. In contrast, Vibrio grows well in seawater. Therefore E. coli is a poor indicator for halophilic pathogens. 3. E. coli decay rates do not correlate with both Salmonella and Vibrio decay rates. However E. coli persists longer than Salmonella in coastal waters. Our results provided support to the continuous use of E. coli as an indicator organism for non-halophilic pathogens especially Salmonella. 4. There is protist‒bacteria coupling where protist counts correlated with bacterial abundance, and protistan bacterivory correlated with bacterial production.

Acknowledgements We are grateful to the Department of Environment, Malaysia for providing the monitoring data of E. coli concentration. Funding for this research was provided by University of Malaya (RG064-09SUS) and Ministry of Science, Technology & Innovation (06-01-03-SF0457). The researchers would also like to thank anonymous reviewers who helped improve the manuscript and University of Malaya for providing the research facilities.

references

Allwood, P.B., Malik, Y.S., Hedberg, C.W., Goyal, S.M., 2003. Survival of F-specific RNA coliphage, feline calicivirus and Escherichia coli in water: a comparative study. Applied and Environmental Microbiology 69 (9), 5707e5710.

1569

Alongi, D.M., 1998. Coastal Ecosystem Processes. CRC Press, Boca Raton, FL. Anderson, K.L., Whitlock, J.E., Harwood, V.J., 2005. Persistence and differential survival of fecal indicator bacteria in subtropical waters and sediments. Applied and Environmental Microbiology 71 (6), 3041e3048. Berger, W.H., Smetacek, V.S., Wefer, G., 1989. Ocean productivity and paleoproductivityean overview. In: Berger, W.H., Smetacek, V.S., Wefer, G. (Eds.), Dahlem Workshop on Productivity of the Ocean: Present and Past. John Wiley & Sons, Chichester, pp. 1e34. Bettarel, Y., Bouvy, M., Dumont, C., Sime-Ngando, T., 2006. Virusbacterium interactions in water and sediment of West African inland aquatic systems. Applied and Environmental Microbiology 72 (8), 5274e5282. Bloem, J., Ba¨r-Gilissen, M.J.B., Cappenberg, T.E., 1986. Fixation, counting, and manipulation of heterotrophic nanoflagellates. Applied and Environmental Microbiology 52, 1266e1272. Bonde, G.J., 1977. Bacterial indication of water pollution. Advances in Aquatic Microbiology 1, 273e364. Bong, C.W., Lee, C.W., 2008. Nearshore and offshore comparison of marine water quality variables measured during SESMA 1. Malaysian Journal of Science 27 (3), 25e31. Borrego, J.J., Figueras, M.J., 1997. Microbiological quality of natural waters. Microbiology SEM 1, 169e180. Cabelli, V., 1978. New standards for enteric bacteria. In: Mitchell, R. (Ed.), Water Pollution Microbiology. John Wiley & Sons, New York, pp. 233e271. Carlucci, A.F., Pramer, D., 1959. Factors affecting the survival of bacteria in sea water. Applied and Environmental Microbiology 7, 388e392. Carlucci, A.F., Pramer, D., 1960a. An evaluation of factors effecting the survival of Escherichia coli in sea water. II. Salinity, pH and nutrients. Applied and Environmental Microbiology 8, 247e250. Carlucci, A.F., Pramer, D., 1960b. An evaluation of factors effecting the survival of Escherichia coli in sea water. IV. Bacteriophages. Applied and Environmental Microbiology 8, 254e256. Chick, H., 1908. An investigation of the laws of disinfection. Journal of Hygiene 8, 92e158. Cissoko, M., Desnues, A., Bouvy, M., Sime-Ngando, T., Verling, E., Bettarel, Y., 2008. Effects of freshwater and seawater mixing on virio- and bacterioplankton in a tropical estuary. Freshwater Biology 53, 1154e1162. Colwell, R.R., Seidler, R.J., Kaper, J., Joseph, S.W., Garges, S., Lockman, H., Maneval, D., Bradford, H., Roberts, N., Remmers, E., Huq, I., Huq, A., 1981. Occurrence of Vibrio cholerae serotype O1 in Maryland and Louisiana estuaries. Applied and Environmental Microbiology 41, 555e558. Department of Environment, 2008. Malaysia Environmental Quality Report 2007. Department of Environment, Ministry of Natural Resources and Environment, Malaysia. Dow, K.M., 1995. An overview of pollution issues in the Strait of Malacca. MIMA Issue Paper 5 (95), 1e34. Enzinger, R.M., Cooper, R.C., 1976. Role of bacteria and protozoa in the removal of Escherichia coli from estuarine waters. Applied and Environmental Microbiology 31, 758e763. Fuhrman, J.A., 2000. Impact of viruses on bacterial processes. In: Kirchman, D.L. (Ed.), Microbial Ecology of the Oceans. WileyLiss, New York, pp. 327e350. Fujioka, R.S., Hashimoto, H.H., Siwak, E.B., Young, R.H.F., 1981. Effect of sunlight on survival of indicator bacteria in sea water. Applied and Environmental Microbiology 41, 690e696. Gerba, C.P., McLeod, J.S., 1976. Effects of sediments on the survival of Escherichia coli in marine waters. Applied and Environmental Microbiology 32, 114e120. Hazen, T.C., Toranzos, G.A., 1990. Tropical source water. In: McFeters, G.A. (Ed.), Drinking Water Microbiology: Progress

1570

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 6 1 e1 5 7 0

and Recent Developments. Springer-Verlag, New York, USA, pp. 32e53. Holt, J.G., Krieg, N.R., Sneath, P.H.A., Staley, J.T., Williams, S.T., 1994. Bergey’s Manual of Determinative Bacteriology, ninth ed. Williams & Wilkins, Baltimore. Kay, D., Fleischer, J.M., Salomon, R.L., Jones, F., Wyer, M.D., Goodfree, A.F., Zelenauch-Jacquotte, Z., Shore, R., 1994. Predicting likelihood of gastroenteritis from sea bathing results from randomized exposure. Lancet 344, 905e909. Kepner Jr., R.L., Pratt, J.R., 1994. Use of fluorochromes for direct enumeration of total bacteria in environmental samples: past and present. Microbiology Reviews 58, 603e615. Law, H.D., 1992. The implementation constraints in waste management in Malaysia. In: Chua, T.E., Garces, L.R. (Eds.), Waste Management in the Coastal Area of the ASEAN Region: Roles of Government, Banking Institutions, Donor Agencies, Private Sector and Communities. ICLARM Conference Proceedings 33. ICLARM, Philippines, pp. 169e172. Lee, C.W., Bong, C.W., 2006. Carbon flux through bacteria in a eutrophic tropical environment: Port Klang waters. In: Wolanski, E. (Ed.), The Environment in Asia Pacific Harbours. Springer, The Netherlands, pp. 329e345. Lee, C.W., Bong, C.W., 2007. Bacterial respiration, growth efficiency and protist grazing rates in mangrove waters in Cape Rachado, Malaysia. Asian Journal of Water Environment and Pollution 4, 11e16. Lee, C.W., Bong, C.W., 2008. Bacterial abundance and production, and their relation to primary production in tropical coastal waters of Peninsular Malaysia. Marine and Freshwater Research 59, 10e21. Lee, C.W., Bong, C.W., Hii, Y.S., 2009a. Temporal variation of bacterial respiration and growth efficiency in tropical coastal waters. Applied and Environmental Microbiology 75, 7594e7601. Lee, C.W., Bong, C.W., Mohamed Yusoff, M.A., Alias, S.A., 2005. Bacterial mediated carbon flux in mangrove waters: a Malaysian perspective. International Journal of Ecology and Environmental Sciences 31, 203e211. Lee, C.W., Ng, A.Y.F., Narayanan, K., Sim, E.U.H., Ng, C.C., 2009b. Isolation and characterization of culturable bacteria from tropical coastal waters. Ciencias Marinas 35, 153e167. Lessard, E.J., Sieburth, JMcN, 1983. Survival of natural sewage populations of enteric bacteria in diffusion and batch chambers in the marine environment. Applied and Environmental Microbiology 45 (3), 950e959. McManus, G.B., 1993. Growth rates of natural populations of heterotrophic nanoplankton. In: Kemp, P.F., Sherr, B.F., Sherr, E.B., Cole, J.J. (Eds.), Handbook of Methods in Aquatic Microbial Ecology. Lewis Publishers, Boca Raton, pp. 557e562. Moe, C.L., 1997. Waterborne transmission of infectious agents. In: Hurst, C.J., Knudsen, G.R., McInerney, M.J., Stetzenbach, L.D.,

Walter, M.V. (Eds.), Manual of Environmental Microbiology. American Society for Microbiology Press, Washington, D. C., pp. 136e152. Munro, P.M., Gauthier, M.J., Breittmayer, V.A., Bongiovanni, J., 1989. Influence of osmoregulation processes on starvation survival of Escherichia coli in seawater. Applied and Environmental Microbiology 55, 2017e2024. Parsons, T.R., Maita, Y., Lalli, C.M., 1984. A Manual of Chemical and Biological Methods for Seawater Analysis. Pergamon Press, Oxford. Pedro´s-Alio´, C., Mas, J., 1993. Bacterial sinking losses. In: Kemp, P.F., Sherr, B.F., Sherr, E.B., Cole, J.J. (Eds.), Handbook of Methods in Aquatic Microbial Ecology. Lewis Publishers, Boca Raton, pp. 677e684. Pomeroy, L.R., Wiebe, W.J., 2001. Temperature and substrates as interactive limiting factors for marine heterotrophic bacteria. Aquatic Microbial Ecology 23, 187e204. Presser, K.A., Ross, T., Ratkowsky, D.A., 1998. Modelling the growth limits (growth/no growth interface) of Escherichia coli as a function of temperature, pH, lactic acid concentration, and water activity. Applied and Environmental Microbiology 64, 1773e1779. Pretorius, W.A., 1962. Some observations on the role of coliphage on the number of Escherichia coli in oxidation ponds. Journal of Hygiene 60, 279e281. Rhodes, M.W., Kator, H., 1988. Survival of Escherichia coli and Salmonella spp. in estuarine environments. Applied and Environmental Microbiology 54, 2902e2907. Rozen, Y., Belkin, S., 2001. Survival of enteric bacteria in seawater. FEMS Microbiology Reviews 25, 513e529. Sanders, R.W., Caron, D.A., Berninger, U.G., 1992. Relationship between bacteria and heterotrophic nanoplankton in marine and freshwaters: an inter-ecosystem comparison. Marine Ecology Progress Series 86, 1e14. Sinton, L.W., Hall, C.H., Lynch, P.A., Davies-Colley, R.J., 2002. Sunlight inactivation of fecal indicator bacteria and bacteriophages from waste stabilization pond effluent in fresh and saline waters. Applied and Environmental Microbiology 68, 1122e1131. Solo-Gabriele, H.M., Wolfert, M.A., Desmarais, T.R., Palmer, C.J., 2000. Sources of Escherichia coli in a coastal subtropical environment. Applied and Environmental Microbiology 66, 230e237. Strom, S.L., 2000. Bacterivory: interactions between bacteria and their grazers. In: Kirchman, D.L. (Ed.), Microbial Ecology of the Oceans. Wiley-Liss, New York, pp. 351e386. Wolf, H.W., 1972. The coliform count as a measure of water quality. In: Mitchell, R. (Ed.), Water Pollution Microbiology. Wiley Interscience, New York, pp. 333e345. Zar, J.H., 1999. Biostatistical Analysis, fourth ed. Prentice Hall, Upper Saddle River, NJ.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 7 1 e1 5 7 8

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

The effect of real-time external resistance optimization on microbial fuel cell performance R.P. Pinto a,b, B. Srinivasan b, S.R. Guiot a, B. Tartakovsky a,b,* a b

Biotechnology Research Institute, National Research Council, 6100 Royalmount Ave., Montre´al, Que., Canada H4P 2R2 Departement de Ge´nie Chimique, E´cole Polytechnique Montre´al, C.P.6079 Succ., Centre-Ville Montre´al, Que., Canada H3C 3A7

article info

abstract

Article history:

This work evaluates the impact of the external resistance (electrical load) on the long-term

Received 23 July 2010

performance of a microbial fuel cell (MFC) and demonstrates the real-time optimization of

Received in revised form

the external resistance. For this purpose, acetate-fed MFCs were operated at external

21 November 2010

resistances, which were above, below, or equal to the internal resistance of a correspond-

Accepted 22 November 2010

ing MFC. A perturbation/observation algorithm was used for the real-time optimal selec-

Available online 30 November 2010

tion of the external resistance. MFC operation at the optimal external resistance resulted in increased power output, improved Coulombic efficiency, and low methane production.

Keywords:

Furthermore, the efficiency of the perturbation/observation algorithm for maximizing

Microbial fuel cell

long-term MFC performance was confirmed by operating an MFC fed with synthetic

External resistance

wastewater for over 40 days. In this test an average Coulombic efficiency of 29% was

Optimal control

achieved. Crown Copyright ª 2010 Published by Elsevier Ltd. All rights reserved.

1.

Introduction

The environmental impact of using fossil fuels to produce energy and their low reserves are leading to a search for renewable energy technologies. Electricity production in Microbial Fuel Cells (MFCs) from a variety of highly diluted organic matter, including wastewater, is one of such technologies (Logan and Regan, 2006; Lovley, 2008). When wastewater is used, MFCs perform waste treatment while recovering energy, thus leading to the possibility of energy-producing wastewater treatment plants. However, the low power density and the restricted output voltage of MFCs limits their industrial application (Pant et al., 2010). Therefore, intense research is now focused on improving MFC power output through the development of new anode and cathode materials (Kang et al., 2003; Logan et al., 2007; Rismani-Yazdi et al., 2008; ter Heijne et al., 2008),

better MFC design (Logan, 2010; Logan et al., 2006; Shimoyama et al., 2008), understanding of electron transfer mechanisms (Debabov, 2008; Reguera et al., 2005; Torres et al., 2010), and optimizing operational conditions (Jadhav and Ghangrekar, 2009). Furthermore, stacks of MFCs are used to increase the operating voltage (Aelterman et al., 2006; Ieropoulos et al., 2008) although challenges such as voltage reversal have been encountered (Oh and Logan, 2007), leading to significant efficiency losses. One simple alternative that is often overlooked is to enhance MFC’s power output by controlling the electrical load (i.e. external resistance) thereby always producing the maximum power output (Woodward et al., 2010). As in any electric power source, maximum power is drawn when the external resistance (Rext) equals the power source’s internal resistance (Fuel Cell Handbook, 2005). An incorrect selection of Rext, either larger or smaller than the internal resistance

* Corresponding author. Biotechnology Research Institute, National Research Council, 6100 Royalmount Ave., Montre´al, Que., Canada H4P 2R2. Tel.: þ1 514 496 2664; fax: þ1 514 496 6265. E-mail address: [email protected] (B. Tartakovsky). 0043-1354/$ e see front matter Crown Copyright ª 2010 Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.11.033

1572

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 7 1 e1 5 7 8

(Rint), may lead to large losses in power output. The Rext control is an important requirement for industrial application of MFCs, since their Rint might vary with changes in operational parameters such as temperature, pH, influent strength, influent composition, and other factors. The problem of optimizing the external load for power sources has been addressed before by on-line control, and it is often referred to as Maximum Power Point Tracking (MPPT). Woodward et al. (2010) successfully applied and compared several non-model based real-time MPPT methods for tracking optimal Rext in MFCs fed with acetate. However, only short-term performance of the tracking algorithms was evaluated, with no attempt to study long-term consequences of MFC operation at an optimal Rext. Our recent work using a model-based simulation of microbial populations in an MFC suggests a performance decrease due to changes in microbial populations when there is a significant deviation from the optimal MFC electrical load (Pinto et al., 2010a, 2010b, 2010c). In addition, an application of the MPPT technique in the case of an MFC operated on wastewater, a more complex influent, has not been reported. Therefore, the goals of this paper are (i) to study the long-term effect of Rext on electricity and methane production in an MFC and (ii) to verify the applicability of the MPPT technique for optimizing Rext of an MFC fed with wastewater.

2.

Materials and methods

2.1.

Analytical methods

Acetate, propionate, and butyrate were analyzed on an Agilent 6890 gas chromatograph (Wilmington, DE, USA) equipped with a flame ionization detector. Method details are provided in Tartakovsky et al. (2008). Chemical oxygen demand (COD) of synthetic wastewater was estimated according to Standard Methods (APHA, 1995). Both total COD (tCOD) and soluble COD (sCOD) values were analyzed. Gas production in the MFC anodic chamber was measured on-line using glass U-tube bubble counters interfaced with a data acquisition system. Gas composition was measured using a gas chromatograph (6890 Series, Hewlett Packard, Wilmington, DE) equipped with a 11 m  3.2 mm 60/80 mesh Chromosorb 102 column (Supelco, Bellefonte, PA, USA) and a thermal conductivity detector. The carrier gas was argon. A detailed description of all analytical methods used in the study can be found in Tartakovsky et al. (2008).

2.2.

Inoculum and media composition

Each MFC was inoculated with 5 mL of anaerobic sludge with volatile suspended solids (VSS) content of approximately 40e50 g L1 (Lassonde Inc, Rougemont, QC, Canada) and 20 mL of effluent from an operating MFC. The stock solution of nutrients was composed of (in g L1): yeast extract (0.8), NH4Cl (18.7), KCl (148.1), K2HPO4 (64.0), and KH2PO4 (40.7). Concentration of acetate in the stock solution was varied in order to obtain the desired organic load by adding sodium acetate (20e80 g L1). Synthetic wastewater

stock solution had a tCOD of 48 g L1 and was composed of (in g L1): pepticase (15.0), beef extract (15.0), yeast extract (9.0), NH4HCO3 (5.1), NaCl (2.8), K2HPO4 (0.5), and KH2PO4 (0.4). A stock solution of the trace elements contained (in g L1): FeCl24H2O (2), H3BO3 (0.05), ZnCl2 (50), CuCl2 (0.03), MnCl24H2O (0.5), (NH4)6Mo7O244H2O (0.5), AlCl3 (0.5), CoCl26H2O (0.5), NiCl2 (0.5), EDTA (0.5), and concentrated HCl (1 mL). One mL of the trace elements stock solution was added to 1 L of deionized water, which was fed to the MFCs (dilution water). Deionized water was used for solution preparation, and the chemicals and reagents used were of analytical grade. All acetate solutions were sterilized by filtration (0.22 mm filtration unit) and maintained at 4  C, while and synthetic wastewater solution was frozen and maintained at 6  C until use.

2.3.

MFC design, operation, and characterization

Four single-chamber membraneless air-cathode MFCs were constructed using polycarbonate plates. The anodes were made of 5 mm thick carbon felt measuring 10 cm  5 cm (SGL Canada, Kitchener, ON, Canada). For MFC-1, MFC-2 and MFC-3 the cathodes were made of a gas diffusion electrode with a Pt load of 0.5 mg cm2 (GDE LT 120 EW, E-TEK Division, PEMEAS Fuel Cell Technologies, Somerset, NJ, USA). The MFC-4 cathode was made using ClFeTMPP (TriPorTech GmbH) on carbon Vulcan XC-72R (Cabot) as a precursor and contained 0.4% Fe with a total catalyst (Fe þ C) load of 2 mg cm2 (Birry et al., 2010).The electrodes were separated by a J-cloth with a thickness of about 0.7 mm. An external recirculation loop was installed for improved mixing of the anodic liquid. The anodic chamber temperature was maintained at 25  C by a PID temperature controller (Model JCR-33A, Shinko Technos Co., Ltd., Osaka, Japan) and a heating plate (120 V-10 W, Volton Manufacturing Ltd, Montreal, Qc, Canada). The electrical load of each MFC was controlled individually by an external resistor. MFC-1 and MFC-2 had a manually controlled external resistors, while the resistors connected to MFC-3 and MFC-4 were computer controlled digital resistors (Innoray, Montreal, QC, Canada) with a resistor variation range from 2.5 U to 1000 U. The MFCs were continuously fed with the stock solutions of carbon source (acetate or synthetic wastewater) and dilution water. The carbon source and dilution streams were combined before entering the anodic chamber. MFC-1, MFC-2, and MFC-3 were operated at several influent concentrations, while maintaining a hydraulic retention time (HRT) of 6 h. MFC-4 was operated at an HRT of 13 h. To account for process variability, each mode of operation was maintained long enough to ensure a steady state performance, which was assessed based on on-line measurements of the output voltage. Table 1 summarizes the operating conditions of each MFC. Polarization tests (PTs) were periodically performed for each MFC. In each PT Rext was disconnected for 30 min, then open circuit voltage (OCV) was measured. Subsequently, the external resistance was re-connected and progressively decreased from 1000 U to 5 U every 10 min with voltage measurements at the end of each period. The resulting voltage and current values were used to construct polarization curves, i.e. voltage vs. current plots from where the MFC’s total (ohmic

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 7 1 e1 5 7 8

Therefore, a smaller ΔR can be used to decrease the distance between Rext and the optimal external resistance, but the time of convergence will increase. A detailed description of this algorithm can be found in Woodward et al. (2010).

Table 1 e MFCs operating conditions. Organic load (g-COD/La/day)

Rext setting

Acetate

2.1, 4.3 or 8.5

Synthetic wastewater

4 or 6

1000 U (above Rint) 5 U (below Rint) Optimal (Rext wRint) Optimal (Rext wRint)

MFC 1 2 3 4

Influent

and solution) Rint was estimated by the slope of the linear region (Fan et al., 2008). Also, cathode and anode open circuit potentials were measured against an Ag/AgCl reference electrode (222 mV vs. normal hydrogen electrode). The Coulombic efficiency (CE) was estimated as: CE ¼

IMFC Dt DSnF

(1)

where I*MFC is the average current produced by the MFC during Dt [A]; Dt is the time interval (typically one day) used to calculate average current and substrate consumption [s]; DS is the amount of substrate consumed [mol-S]; n is the number of electrons transferred per mol of substrate [mol-e mol-S1]; and F is the Faraday constant [A s mol-e1]. Volumetric power output (Pout, mW L1 a ) was calculated using measurements of MFC output voltage, a corresponding value of Rext, and MFC anodic chamber volume (La). Maximum volumetric power output (Pmax) was estimated from the power curves (Pout vs current) obtained during the polarization tests. Since Rext in the PTs was changed stepwise, the accuracy of Pmax estimation was improved by using a linear interpolation of the polarization curve in the region of constant voltage drop, UMFC ¼ a0 þ a1 IMFC, where UMFC is the MFC output voltage (V), IMFC is the MFC current (A), and a0, a1 are the regression coefficients. Based on this interpolation, Pmax was calculated as described in Logan (2008), p. 47 : Pmax ¼

a20 4a1 La

(2)

Notably, for an MFC with small overpotentials a0 is close to the OCV estimation and a1 corresponds to Rint (Logan, 2008).

2.4.

Maximum power point tracking (MPPT) algorithm

The perturbation observation (P/O) method is commonly used in MPPT of solar panels (Hua and Shen, 1998). It was selected for this study because of its robustness, demonstrated in short-term MFC tests (Woodward et al., 2010), and its simplicity. The P/O algorithm applied in this work modified Rext with a predetermined amplitude (ΔR) at each iteration. The direction of resistance change was selected by comparing the value of the power output with that at the previous resistance. The method can be expressed as follows:   PMFC ðk þ 1Þ  PMFC ðkÞ Rext ðk þ 2Þ ¼ Rext ðk þ 1Þ þ DRsign Rext ðk þ 1Þ  Rext ðkÞ

1573

(3)

where ΔR is the amplitude of change in Rext [U]; PMFC is the MFC power output [W]; and k is the iteration number. Once the algorithm converges to an optimum, the Rext will oscillate around this optimum with a maximum distance of ΔR.

3.

Results and discussion

3.1. The impact of external resistance on MFC performance The effect of Rext on long-term MFC performance was studied by simultaneously operating acetate-fed MFC-1, 2, and 3 at high, low, and optimal Rext settings, respectively, for 30e35 days (Table 1). The selected Rext values were significantly different (e.g high Rext corresponded to 1000 U and low Rext corresponded to 5 U, Table 1). This approach reduced the impact of MFC performance variability due to the microbiological nature of the process on the comparison of MFC power outputs and other performance parameters at each mode of operation. Throughout the tests, variations in influent acetate concentration were simultaneously imposed for all MFCs. The profile of acetate influent concentration is shown in Fig. 1a. Acetate concentration measurements in the effluent streams showed similar substrate removal in all MFCs, as can be seen from the values presented in Fig. 1a. At steady state, the effluent acetate concentration varied from 20 to 160 mg L1. At the same time, power outputs were quite different. Fig. 1b summarizes power production observed throughout the tests. This figure shows that in all MFCs power output began to increase after approximately 3 days of operation, reaching steady state values after 7e10 days. Power outputs at steady state strongly depended on Rext selection with power outputs around 6, 15, and 58 mW L1 a , observed for MFC-1 (high Rext), MFC-2 (low Rext), and MFC-3 (optimal Rext), respectively. Notably on day 13 the cathode of MFC-3 was punctured and replaced by a new cathode, made of the same material. Following the replacement, MFC-3 power output approached 135 mW L1 a for about two days before returning to its previous level. This period was excluded from Fig. 1b. Also, due to technical problems, MFC-2 voltage was not recorded between days 8e12, 14.5 to 16.5, and 28.5 to 30.5. Fig. 1c presents changes in Rext of MFC-3 imposed by the P/O algorithm over time. At startup, Rext value maintained by the algorithm was above 400 U, then after about 5 days of operation Rext sharply decreased to values below 40 U. This figure also shows Rint values estimated for MFC-3 during polarization tests. As expected, the P/O algorithm provided timely adjustment of Rext to maintain it close to Rint values, such that the Rext oscillated around the optimum value equal to MFC-3 internal resistance. The profile of Rint change in the MFC-2 test was similar to that observed for MFC-3 rapidly decreasing to 15e30 U after first 6 days of MFC operation, while Rint of MFC-1 remained at around 200 U for most of the test. While the growth of anodophilic microorganisms during the startup period resulted in relatively slow changes in Rint, the variations in operating conditions had an almost immediate impact on Rint and therefore on MFC performance. The P/O algorithm’s ability to track fast variations of operating conditions (e.g. influent composition) during MFC-3 operation

1574

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 7 1 e1 5 7 8

b

Influent MFC-1 MFC-2 MFC-3

-1

1

MFC-3

60 40

0

0 5

0

10 15 20 25 30 35

5

10 15 20 25 30 35

Time (d)

Time (d)

d R ext

75 50 25 0

0

5

50

65

Pout

40

R int Rext (Ω)

Resistance (Ω)

100

30

45

20

35

10 23.75

10 15 20 25 30 35

55

Rext

25 24.25

24

Time (d)

-1

0

c

MFC-2

MFC-1

20

Pout (mW L a )

-1 Acetate (g L )

2

Pout (mW La )

a

Time (d)

Fig. 1 e (a) Acetate concentration in the influent and effluent, and (b) power production for MFC-1 MFC-2, and MFC-3 against time; (c) Rext and Rint values for MFC-3 (MFC-1 and MFC-2 Rext values were always kept at 1000 U and at 5 U, respectively); (d ) Rext and Pout values during an increase in the MFC-3 influent concentration at t [ 23.9 days.

is illustrated in Fig. 1d. Here, the influent acetate concentration was increased from 0.5 to 2 g L1 on day 24. This increase in acetate concentration caused a decrease of Rint and, accordingly, the MPPT algorithm decreased the Rext value thus maximizing power output under new operating conditions. Fig. 2 presents estimations of current density and Coulombic efficiency (CE ) for MFC-1, MFC-2 and MFC-3. As expected, MFC-1, which was operated at a high Rext, always had a low current density and a low CE, while MFC-2 and MFC-3 showed larger values. MFC-2, which was operated at the lowest Rext, was expected to have the highest Coulombic efficiency. However, current densities (Fig. 2a) and CE (Fig. 2b) of MFC-3, which was operated at an optimal Rext, on average were slightly

b

3

100%

MFC-1 MFC-3

80%

2 1.5 MFC-1 MFC-2 MFC-3

1 0.5

2.5 2

-1

2.5

MFC-2 Influent

60%

1.5

40%

1

20%

0

0.5 0

0%

0

5

10

15

20

Time (d)

25

30

35

Acetate (g L )

3.5

CE (%)

-2

Current density (A m )

a

higher than that of MFC-2. CE calculations showed a more pronounced difference between MFC-2 and MFC-3, however this difference could be attributed to the variability of the effluent acetate measurements. Also, MFC-3 featured the shortest startup time, as can be seen from the comparison of current densities in Fig. 2a. Polarization tests provided additional information for the comparison of MFC performances. Fig. 3a shows the evolution of the cathode and anode open circuit potentials (OCP) over time. As expected, cathode OCP values were similar for all MFCs and remained constant throughout the experiment. Following the startup, anode OCP values decreased for all MFCs, with the fastest decrease observed for MFC-2 (low Rext),

0

5

10

15

20

25

30

35

Time (d)

Fig. 2 e Current density (a) and Coulombic efficiency (b) observed at various acetate loads (Fig. 1a).

1575

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 7 1 e1 5 7 8

a

100

Cathode

150

0

5

10

-150

15

20

25

30

35

MFC-1 MFC-2 MFC-3

-300 -450

Anode

-1

0

-600

b

80

Time (d)

Pmax (mW L a )

Potential vs. Ag/AgCl (mV)

300

60 MFC-1

40

MFC-2 20

MFC-3

0

0

5

10

15

20

25

30

35

Time (d) Fig. 3 e (a) The evolution of the cathode and anode OCP values over time, and (b) maximum power densities estimated using polarization curves according to Eq. (2).

followed by MFC-3 (optimal Rext). MFC-1, which was operated at a high Rext, was the last to reach a steady state value of the anode potential. Nevertheless, the anode OCP values for all MFCs were similar after 15 days of testing. It can be hypothesized that this pattern of anode OCP decrease over time was reflective of anode colonization by the anodophilic microorganisms. Indeed, lower values of Rext facilitate the electron transfer process thus providing growth advantages to the anodophilic microorganisms. Consequently MFC-2 operated at the lowest Rext featured the fastest rate of the anodophilic biofilm formation. However, MFC operation at an Rext below Rint results in low power output (Logan, 2008), thus requiring Rext optimization. When maximal power outputs were estimated from the polarization test results, it was observed that MFC-1 always had low Pmax, never exceeding 9 mW L1 a . Maximal power outputs estimated for MFC-2 and MFC-3 increased during the first 15e20 days of the experiment (Fig. 3b). A maximal power output of 95 mW L1 a was estimated based on the polarization test for MFC-3 on day 15, which was carried out shortly after the cathode replacement in this MFC. As mentioned above, the cathode replacement resulted in higher than usual power output between days 13 and 15. On average, MFC-3 maximal power output remained around 70 mW L1 a , which agreed well with the power densities measured during MFC-3 test (Fig. 1b). At the same time, power output of MFC-2 during normal operation was very low because of the Rext choice. The selection of Rext for MFC-1 (Rext ¼ 1000 U) and MFC-2 (Rext ¼ 5 U) was confirmed by the polarization tests. After the startup period, Rint in MFC-1 varied between 50 and 200 U, while in MFC-2 Rint varied between 15 and 25 U. Thus, MFC-1 was operated at Rext >> Rint, and MFC-2 was operated at Rext << Rint throughout the test, as intended. Besides power output comparison, methane production in the anodic compartment of each MFC was measured throughout the tests and was used to compare the long-term effects of Rext selection on methane production. Since the MFCs were inoculated with anaerobic sludge, methane production was observed at the beginning of the operation in all MFCs. The methane production in MFC-1 increased over time, while it decreased in MFC-2 and MFC-3. To compare electricity and methane production for MFC-1, MFC-2, and

MFC-3 steady state values were calculated using experimental results obtained between day 7 and 30 of MFC operation, when constant organic load was applied. The average methane production rates were 2.0  0.7, 0.2  0.1, 0.6  0.2 mL d1 in MFC-1, MFC-2, and MFC-3, respectively. The corresponding power outputs for these MFCs were 5.6  0.3, 14.6  7.7, and 60.8  17.2 mW L1 a . It should be acknowledged that methane flow measurements were complicated by the small volume (50 mL) of the anodic compartments and low methane production rates, in the range of several mL per day. Measurements of such small flow rates resulted in large standard deviations. Nevertheless, the overall trends were clear showing significantly higher methane production in MFC-1 operated at high Rext. In contrast, methane production in MFC-1 and MFC-2 was very low. These results agree with methane production measurements reported in Martin et al. (2010), where MFCs were operated at Rext above estimated values of Rint resulting in significant methane production. It was previously demonstrated using biomolecular methods, that a high Rext, which implies MFC operation at more positive anode potentials, decreases the amount of the anodophilic microorganisms (Torres et al., 2009). There is a significant body of evidence proving that the external resistance at which the MFC operates has an influence on the microbial communities and the long-term performance (Aelterman et al., 2008; Chae et al., 2010; Lyon et al., 2010; Pinto et al., 2010c; Torres et al., 2009). Aelterman et al. (2008) observed the impact of Rext on electricity and methane production in an MFC inoculated with a mixed anaerobic culture and concluded that low methane production and stable power output are only obtained if Rext is set close to the MFC internal resistance. Furthermore, Chae et al. (2010) compared methanogenic activity and methane production in MFCs subjected to several external perturbations (pH, temperature, oxygen exposure, addition of a methanogenesis inhibitor, and Rext variation). They also concluded that electricity production was increased and methane production was decreased only when a methanogenesis inhibitor (BES, 2-bromoethanesulfonate) was added to the anodic chamber or by setting a Rext close to the Rint values. Computer simulations using a two population MFC model (Pinto et al., 2010c) corroborated with the experimental

1576

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 7 1 e1 5 7 8

evidence above. The model was used to simulate the outcome of the competition between the anodophilic and methanogenic microorganisms for a common carbon source (acetate) in MFCs operated at different organic loads and Rext settings (Pinto et al., 2010c). It was predicted that, independent of the microbial composition of the inoculum, proliferation of the anodophilic microorganisms could only be achieved at Rext values that are equal to or less than the MFC internal resistance. The same conclusion can be inferred from the results of Lyon et al. (2010). In this study the microbial composition of biofilms sampled from MFCs inoculated with the same sludge and operated at different Rext values were analyzed. By using Ribosomal Intergenic Spacer Analysis (RISA) they demonstrated, based on the profiles observed, that the microbial community structure in MFC’s biofilm operated at Rext appreciably above Rint (1 kU and 10 kU) was significantly different from that observed in the MFC operated at low Rext (10 U, 100 U, and 470 U). As mentioned above, a low Rext promotes growth and metabolic activity of the anodophilic microorganisms since electron transport to the cathode is facilitated. However Rext, which is lower than the MFC Rint value leads to a low power output, i.e. an optimal Rext value (Rext w Rint) should be maintained. Notably, all of the tests mentioned above were carried out by manual adjustment of Rext without using a real-time algorithm, which would guarantee timely correction of Rext. The results from the citations above are in perfect agreement with the findings presented in this paper, where acetatefed MFCs were inoculated with the same anaerobic sludge and consumed comparable quantities of carbon source. They were operated at high, low, or optimal Rext settings, leading to Coulombic efficiencies remarkably different (Fig. 2b). High Rext led to, essentially, an anaerobic reactor with low Coulombic efficiency and significant methane production. At the same time, both MFC-2 (low Rext) and MFC-3 (optimal Rext) featured high Coulombic efficiency, and by the end of the 30 day test methane production in these MFCs declined to near zero values. Notably, the high power output with negligible methane production observed during MFC-3 operation are in agreement with the results of ClFeTMPP cathode tests, where an MFC was also operated using the P/O algorithm for the online optimization of Rext (Birry et al., 2010). Considering that MFC operation at Rext values below Rint leads to a sharp drop in power output (Fuel Cell Handbook, 2005), it is sufficient to maintain Rext at an optimal value in order to minimize methane production and maximize power production.

3.2.

MFC operation on synthetic wastewater

The efficiency of the P/O algorithm when operating an MFC on a complex feed was confirmed by feeding MFC-4 with synthetic wastewater. The power output observed during this test and Rext values selected by the P/O algorithm are shown in Fig. 4. This graph also shows Rint estimations obtained in the polarization tests. A sharp decrease in Rint values obtained from the polarization tests and Rext values selected by the P/O algorithm can be seen after the first 5 days of MFC operation (Fig. 4c). After this point, Rint remained between 20 U and 60 U with Rext matching these values. As one can see, Pout increased steadily during the first 15 days of the experiment (Fig. 4b),

which is longer than the 5 day startup period observed in the acetate-fed MFC-3 (Fig. 1b). A longer startup period could be attributed to the use of synthetic wastewater containing complex organic matter. We hypothesize that wastewater hydrolysis and fermentation steps, which were required because of the complex wastewater composition, limited the amount of volatile fatty acids available for growth and metabolism of the anodophilic microorganisms. To demonstrate the robustness of the P/O algorithm, MFC4 was subjected to two types of perturbations: First the organic load was increased by 50% between days 16 and 20 (Fig. 4a), then between days 23 and 33 the MFC temperature was increased from 25  C to 35  C (Fig. 4d). The increase in organic load did not result in a substantial increase of power production and Rint remained unchanged (Fig. 4b,c). It can be seen that the increased organic load did not result in an immediate increase in the effluent COD concentration (Fig. 4a), likely due to a short duration of the test. Effluent VFA analysis showed that concentrations of acetate, propionate, and butyrate always remained below 70e90 mg L1 during the test. Apparently, the biotransformation process was limited by the rate of hydrolysis rather than the rate of VFA consumption by the anodophilic microorganisms. Accordingly, Rint remained unchanged during the test. MFC response to temperature increase was more pronounced with an immediate drop in Rext from about 48 U to 30 U (40% decrease) and a 50% increase of Pout. Once MFC temperature was returned to 25  C, the P/O algorithm adjusted Rext close to its previous level. A similar response to a temperature perturbation was observed by Woodward et al. (2010). Notably, an additional external perturbation during MFC-4 test was caused by the fermentation of the synthetic wastewater solution kept at room temperature in a syringe. This feeding solution was replaced every two or three days. Fermentation of the synthetic wastewater resulted in a visible change in the feed solution appearance with an average of 0.15 g L1 of volatile fatty acids at the end of each feeding period, although tCOD concentration remained constant. The arrows in Fig. 4b represent the times when the synthetic wastewater was replenished. It can be seen that synthetic wastewater fermentation had a direct impact on the Pout. After each wastewater replenishment the MFC power output declined for about 24 h, while reaching its maximal values by the end of each feeding period. Also, abrupt drops in Pout observed on days 25, 28, and 33 are related to delays in replacing the syringe, which resulted in an interruption of the feed. These perturbations of the organic load were promptly followed by the P/O algorithm thus demonstrating its excellent robustness, as can be seen from the results presented in Fig. 4b. Polarization tests not only confirmed the choice of Rext values by the P/O algorithm, but also demonstrated timerelated changes in OCV and Pmax values resulting from MFC operation at an optimal Rext. OCV and Pmax values presented in Fig. 5 show that OCV slowly increased in about 20 days stabilizing around 500 mV, while Pmax more then doubled after day 7 and fluctuated between 17 and 35 mW L1 a with higher outputs observed by the end of the test. Overall, the Coulombic efficiency of MFC-4 varied between 14% and 41%, with an average of 29% throughout the experiment.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 7 1 e1 5 7 8

1577

Fig. 4 e MFC-4 performance: (a) total COD in the influent (tCOD-in) and soluble COD in the effluent (sCOD-out). The organic load was increased by 50% between days 16 and 20; (b) power output; (c) external and internal resistance; (d ) External resistance and power density response during an increase in temperature from 25  C to 35  C at 23.25 days. Arrows show syringe replacement times.

During MFC-3 and MFC-4 operation, real-time correction of Rext was instrumental in maintaining its value at or close to Rint. Indeed, because of the variations of organic load, composition, and MFC temperature during the tests, the internal resistance of the MFCs varied in time. Polarization tests performed after each change in operating conditions were used to estimate Rint variations, as can be seen in Fig. 1c and Fig. 4c, where Rint estimations are shown along with Rext values selected by the P/O algorithm. The P/O algorithm demonstrated excellent stability and fast convergence so that 50

600 500

OCV (mV)

-1

Pmax (mW La )

40

400 30 300 20 200 OCV

100 0

Rext always remained close Rint as illustrated in Fig. 1d and Fig. 4d. The real-time strategy for Rext control also allowed for avoiding a sharp decrease in MFC performance even when MFC feed was interrupted due to a technical problem (e.g. syringe pump malfunction or a delay in syringe replacement). These events led to a sharp short-term increase in the internal resistance, which was successfully tracked by the P/O algorithm (e.g. days 25, 28 and 33 in Fig. 4c). Notably, implementing an MPPT on-line technique for MFC’s Rext control could be used for individual control of electric loads of MFC stacks. This strategy might prevent MFC operation at Rext values below its Rint thus helping to avoid voltage reversal (Oh and Logan, 2007) after a feed disruption or another operating problem.

10

Pmax 0

10

20

30

40

0

Time (d) Fig. 5 e Open circuit voltage and maximum power output estimated from the polarization tests in MFC-4.

4.

Conclusion

In this study, a simple perturbation/observation algorithm was used to maximize MFC power output by matching Rext and Rint values. The real-time optimization of Rext was tested through long-term operation of MFCs fed with either acetate or synthetic wastewater. The P/O algorithm demonstrated an excellent performance when the MFC was subject to perturbations in operating conditions such as variations in organic load, influent composition, and temperature. A comparison of MFC performance at an optimal Rext value, with MFCs operated at either high (Rext >> Rint) or low (Rext << Rint) external resistances showed that real-time resistance optimization led to significantly higher power outputs with less methane

1578

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 7 1 e1 5 7 8

production. MFCs operated at an optimal resistance showed average Coulombic efficiencies of 57% and 29% on acetate and synthetic wastewater, respectively. It is concluded that the real-time optimal control of MFC electrical load might be instrumental in the development of stackable MFCs for combined wastewater treatment and power production.

Acknowledgement This research was supported by the National Research Council of Canada (NRC publication #53342).

references

Aelterman, P., Rabaey, K., Pham, H.T., Boon, N., Verstraete, W., 2006. Continuous electricity generation at high voltages and currents using stacked microbial fuel cells. Environmental Science and Technology 40 (10), 3388e3394. Aelterman, P., Versichele, M., Marzorati, M., Boon, N., Verstraete, W., 2008. Loading rate and external resistance control the electricity generation of microbial fuel cells with different three-dimensional anodes. Bioresource Technology 99 (18), 8895e8902. APHA, 1995. Standard Methods for the Examination of Water and Wastewater. American Public Health Association, Washington, DC. Birry, L., Mehta, P., Jaouen, F., Dodelet, J.P., Guiot, S.R., Tartakovsky, B., 2010. Application of iron-based cathode catalysts in a microbial fuel cell. Electrochimica Acta, in press. Chae, K.-J., Choi, M.-J., Kim, K.-Y., Ajayi, F.F., Park, W., Kim, C.-W., Kim, I.S., 2010. Methanogenesis control by employing various environmental stress conditions in two-chambered microbial fuel cells. Bioresource Technology 101 (14), 5350e5357. Debabov, V., 2008. Electricity from microorganisms. Microbiology 77 (2), 123e131. Fan, Y., Sharbrough, E., Liu, H., 2008. Quantification of the internal resistance distribution of microbial fuel cells. Environmental Science and Technology 42, 8101e8107. Fuel Cell Handbook, 2005. National Energy Technology Laboratory U.S. Department of Energy. University Press of the Pacific. Hua, C., Shen, C., 1998. Comparative Study of Peak Power Tracking Techniques for Solar Storage System. In: Anon (Ed.). IEEE, Anaheim, CA, USA, pp. 679e685. Ieropoulos, I., Greenman, J., Melhuish, C., 2008. Microbial fuel cells based on carbon veil electrodes: stack configuration and scalability. International Journal of Energy Research 32, 1228e1240. Jadhav, G.S., Ghangrekar, M.M., 2009. Performance of microbial fuel cell subjected to variation in pH, temperature, external load and substrate concentration. Bioresource Technology 100 (2), 717e723. Kang, K.H., Jang, J.K., Pham, T.H., Moon, H., Chang, I.S., Kim, B.H., 2003. A microbial fuel cell with improved cathode reaction as a low biochemical oxygen demand sensor. Biotechnology Letters 25 (16), 1357e1361. Logan, B., Cheng, S., Watson, V., Estadt, G., 2007. Graphite fiber brush anodes for increased power production in air-cathode microbial fuel cells. Environmental Science and Technology 41, 3341e3346. Logan, B.E., 2008. Microbial Fuel Cells. John Wiley & Sons, Hobeken, New Jersey.

Logan, B.E., 2010. Scaling up microbial fuel cells and other bioelectrochemical systems. Applied Microbiology and Biotechnology 85 (6), 1665e1671. Logan, B.E., Hamelers, B., Rozendal, R.A., Schroder, U., Keller, J., Freguia, S., Aelterman, P., Verstraete, W., Rabaey, K., 2006. Microbial fuel cells: methodology and Technology. Environmental Science and Technology 40 (17), 5181e5192. Logan, B.E., Regan, J.M., 2006. Microbial fuel cells - Challenges and applications. Environmental Science and Technology 40 (17), 5172e5180. Lovley, D.R., 2008. The microbe electric: conversion of organic matter to electricity. Current Opinion in Biotechnology 19 (6), 564e571. Lyon, D.Y., Buret, F., Vogel, T.M., Monier, J.-M., 2010. Is resistance futile? Changing external resistance does not improve microbial fuel cell performance. Bioelectrochemistry 78 (1), 2e7. Martin, E., Savadogo, O., Guiot, S.R., Tartakovsky, B., 2010. The influence of operational conditions on the performance of a microbial fuel cell seeded with mesophilic anaerobic sludge. Biochemical Engineering Journal 51 (3), 132e139. Oh, S.E., Logan, B.E., 2007. Voltage reversal during microbial fuel cell stack operation. Journal of Power Sources 167 (1), 11e17. Pant, D., Van Bogaert, G., Diels, L., Vanbroekhoven, K., 2010. A review of the substrates used in microbial fuel cells (MFCs) for sustainable energy production. Bioresource Technology 101 (6), 1533e1543. Pinto, R.P., Perrier, M., Tartakovsky, B. and Srinivasan, B., 2010a. Performance analyses of microbial fuel cells operated in series, Proceedings of 9th International Symposium on Dynamics and control of process systems, Leuven, Belgium. Pinto, R.P., Srinivasan, B., Manuel, M.F., Tartakovsky, B., 2010b. A two-population bio-electrochemical model of a microbial fuel cell. Bioresource Technology 101 (14), 5256e5265. Pinto, R.P., Tartakovsky, B., Perrier, M., Srinivasan, B., 2010c. Optimizing treatment performance of microbial fuel cells by reactor Staging. Industrial & Engineering Chemistry Research 49 (19), 9222e9229. Reguera, G., McCarthy, K.D., Mehta, T., Nicoll, J.S., Tuominen, M.T., Lovley, D.R., 2005. Extracellular electron transfer via microbial nanowires. Nature Biotechnology 435, 1098e1101. Rismani-Yazdi, H., Carver, S.M., Christy, A.D., Tuovinen, I.H., 2008. Cathodic limitations in microbial fuel cells: an overview. Journal of Power Sources 180, 683e694. Shimoyama, T., Komukai, S., Yamazawa, A., Ueno, Y., Logan, B.E., Watanabe, K., 2008. Electricity generation from model organic wastewater in a cassette-electrode microbial fuel cell. Applied Microbiology and Biotechnology 80 (2), 325e330. Tartakovsky, B., Manuel, M.F., Neburchilov, V., Wang, H., Guiot, S.R., 2008. Biocatalyzed hydrogen production in a continuous flow microbial fuel cell with a gas phase cathode. Journal of Power Sources 182 (1), 291e297. ter Heijne, A., Hamelers, H.V.M., Saakes, M., Buisman, C.J.N., 2008. Performance of non-porous graphite and titanium-based anodes in microbial fuel cells. Electrochemica Acta 53, 5697e5703. Torres, C.I., Krajmalnik-Brown, R., Parameswaran, P., Marcus, A.K., Wanger, G., Gorby, Y.A., Rittmann, B.E., 2009. Selecting anoderespiring bacteria based on anode potential: phylogenetic, electrochemical, and microscopic characterization. Environmental Science and Technology 43 (24), 9519e9524. Torres, C.I., Marcus, A.K., Lee, H., Parameswaran, P., KrajmalnikBrown, R., Rittmann, B.E., 2010. A kinetic perspective on extracellular electron transfer by anode-respiring bacteria. FEMS Microbiology Reviews 34 (1), 3e17. Woodward, L., Perrier, M., Srinivasan, B., Pinto, R.P., Tartakovsky, B., 2010. Comparison of real-time methods for maximizing power output in microbial fuel cells. AIChE Journal 56 (10), 2742e2750.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 7 9 e1 5 8 6

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Electrochemical oxidation of trace organic contaminants in reverse osmosis concentrate using RuO2/IrO2-coated titanium anodes Jelena Radjenovic*, Arseto Bagastyo, Rene´ A. Rozendal, Yang Mu, Ju¨rg Keller, Korneel Rabaey Advanced Water Management Centre, The University of Queensland, Brisbane, QLD 4072, Australia

article info

abstract

Article history:

During membrane treatment of secondary effluent from wastewater treatment plants,

Received 26 July 2010

a reverse osmosis concentrate (ROC) containing trace organic contaminants is generated.

Received in revised form

As the latter are of concern, effective and economic treatment methods are required. Here,

24 November 2010

we investigated electrochemical oxidation of ROC using Ti/Ru0.7Ir0.3O2 electrodes, focussing

Accepted 24 November 2010

on the removal of dissolved organic carbon (DOC), specific ultra-violet absorbance at

Available online 1 December 2010

254 nm (SUVA254), and 28 pharmaceuticals and pesticides frequently encountered in secondary treated effluents. The experiments were conducted in a continuously fed reactor

Keywords:

at current densities (J ) ranging from 1 to 250 A m
2 anode, and a batch reactor at

Water recycling

J ¼ 250 A m
2. Higher mineralization efficiency was observed during batch oxidation (e.g.

Reverse osmosis

25.1  2.7% DOC removal vs 0% removal in the continuous reactor after applying specific

Advanced oxidation

electrical charge, Q ¼ 437.0 A h m
3 ROC), indicating that DOC removal is depending on

Pharmaceuticals

indirect oxidation by electrogenerated oxidants that accumulate in the bulk liquid. An

Pesticides

initial increase and subsequent slow decrease in SUVA254 during batch mode suggests the

LC-MS

introduction of auxochrome substituents (e.g. eCl, NH2Cl, -Br, and -OH) into the aromatic

Microtox test

compounds. Contrarily, in the continuous reactor ring-cleaving oxidation products were generated, and SUVA254 removal correlated with applied charge. Furthermore, 20 of the target pharmaceuticals and pesticides completely disappeared in both the continuous and batch experiments when applying J  150 A m
2 (i.e. Q  461.5 A h m
3) and 437.0 A h m
3 (J ¼ 250 A m
2), respectively. Compounds that were more persistent during continuous oxidation were characterized by the presence of electrophilic groups on the aromatic ring (e.g. triclopyr) or by the absence of stronger nucleophilic substituents (e.g. ibuprofen). These pollutants were oxidized when applying higher specific electrical charge in batch mode (i.e. 1.45 kA h m
3 ROC). However, baseline toxicity as determined by Vibrio fischeri bioluminescence inhibition tests (Microtox) was increasing with higher applied charge during batch and continuous oxidation, indicating the formation of toxic oxidation products, possibly chlorinated and brominated organic compounds. ª 2010 Elsevier Ltd. All rights reserved.

* Corresponding author. Tel.: þ61 7 3346 3234; fax: þ61 7 3365 4726. E-mail address: [email protected] (J. Radjenovic). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.11.035

1580

1.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 7 9 e1 5 8 6

Introduction

Due to the growing pressure on water resources, the use of treated municipal wastewater for groundwater recharge and indirect potable reuse is increasingly considered. A number of reuse facilities worldwide currently employ microfiltration (MF) followed by reverse osmosis (RO) for the treatment of secondary treated effluent prior to aquifer or reservoir recharge. High-pressure RO membranes have gained popularity due to their outstanding performance in rejecting trace organics such as endocrine disrupting compounds, pesticides, pharmaceuticals and others (Bellona and Drewes, 2007; Radjenovic et al., 2008; Snyder et al., 2007). These compounds will be concentrated four to seven times in the waste stream (reverse osmosis concentrate, ROC) that normally comprises about 15e25% of the incoming water flow. The ROC offers an opportunity for reducing human and ecotoxicological risk by implementing brine treatment prior to environmental discharge. Several advanced oxidation treatment options (e.g. TiO2 photocatalysis, sonolysis) showed moderate performance in removing the organic matter from ROC (Dialynas et al., 2008; Westerhoff et al., 2009). In the past years, electrochemical oxidation processes received renewed interest due to several perceived advantages such as efficient control of reaction conditions, no chemical requirements, simplicity and robustness of operation at ambient temperature and pressure. Van Hege and co-workers pioneered the electrochemical oxidation of ROC (Van Hege et al., 2002). More recent work by Dialynas et al. (2008) reported moderate dissolved organic carbon (DOC) removal in electrolytic oxidation of ROC on boron-doped diamond (BDD) anodes, while Perez et al. (2010) observed an excellent performance by BDD in eliminating chemical oxygen demand (COD) and 10 pharmaceuticals and stimulant drugs from ROC. Due to recent advances in electrode stability and performance, electrochemically driven processes are becoming an attractive option for the remediation of problematic waste streams. Notably, mixed metal oxide (MMO) coated electrodes have found widespread environmental applications in recent years for the treatment of pesticide contaminated water, landfill leachate, organic petroleum wastewater and other difficult to treat waste streams (Panizza, 2010). MMO anodes such as RuO2/IrO2-coated titanium with improved electrocatalytic behaviour and stability are readily available in practical mesh geometries and have extended life-time and lower costs relative to BDD electrodes. The latter have thus far been considered as the standard for electrochemical oxidation, but suffer from high product costs. Because of their low overpotential for chlorine evolution, RuO2/IrO2-coated electrodes were effectively used for the degradation of pharmaceuticals, pesticides and other organic compounds via indirect electrolysis (Malpass et al., 2006; Carlesi Jara et al., 2007; Gallard et al., 2004). However, while on one hand in-situ generated active chlorine can effectively oxidize many pollutants, on the other hand the formation of chlorinated byproducts could lead to increased toxicity levels. Different operational strategies were proposed in literature for minimizing the formation of chlorinated organic compounds, such

as application of long residence time and activated carbon polishing treatment (Rajkumar et al., 2005), continuous operation and short residence time (Bergmann and Koparal, 2005a), and over saturation of the solution by chlorine. The latter approach increases the degradation rate of chlorinated intermediates relative to their rate of formation (Gallard et al., 2004). In this study, the electrochemical oxidation using Ti/ Ru0.7Ir0.3O2 electrode was investigated for the treatment of ROC. The process efficiency was evaluated based on the removal of DOC, specific ultra-violet absorbance at 254 nm (SUVA254), and 28 pesticides and pharmaceuticals encompassing diverse molecular structures and physico-chemical properties (Table S1), selected for their ubiquity in municipal wastewater effluents and brine streams. The influence of operational mode, current density, and applied electrical charge on the oxidation efficiency was investigated. The baseline toxicity of electrochemically oxidized ROC was evaluated in the bioluminescence inhibition tests (Microtox) using the marine bacterium Vibrio fischeri.

2.

Materials and methods

2.1.

Chemicals

All standards for pharmaceuticals and pesticides used were of analytical grade (99%) (Text S1). All solvents (methanol, acetonitrile and water) were HPLC-grade and were purchased from Merck (Germany), as well as hydrochloric acid (37%), ammonium acetate, sodium hydroxide and formic acid (98%).

2.2.

Reverse osmosis concentrate

The ROC used in the experiments was sampled at an advanced water treatment plant (AWTP) in Bundamba, 30 km west of Brisbane, Australia. This AWTP receives a mixture of secondary treated effluents from four wastewater treatment plants. After the pre-treatment of secondary effluents (addition of iron coagulants, separation of solids in the clarifier), raw water is passed through MF and RO membranes. The generated ROC is further subjected to nitrification in a moving bed biofilm reactor (MBBR), coagulation and denitrification by anoxic filters, and is finally discharged into the Brisbane River. In order to investigate the removal of trace organic contaminants during electrochemical oxidation, ROC collected prior to the nitrification stage was spiked with the concentrated solutions of target compounds prepared in water and methanol. The spiking solution of compounds soluble in water was prepared from pure standards at w1 mg L
1 concentration, and for each 10 L of ROC, 100 mL of spiking solution in water was added. The spiking solution of compounds poorly soluble in water was prepared in methanol at 1 g L
1 concentration, and for each 10 L of ROC, 100 mL of spiking solution in methanol was added. Since methanol can act as a scavenger of the generated oxidants, the amount added was minimized. In order to ensure homogeneity, the spiked ROC sample was filtered through 0.7 mm glass fibre filters (Whatman, UK) prior to the experiments. The obtained final concentrations of

1581

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 7 9 e1 5 8 6

target analytes were in the range of 7.8 (carbamazepine) to 37.4 mg L
1 (iopromide) (Table S2).

2.3.

Experimental setup

The electrochemical cell was constructed by assembling two equal rectangular polycarbonate frames with internal dimensions of 20  5  1.2 cm. The frames were bolted together between two polycarbonate square plates. The anode cell was separated from the cathode cell by a cation exchange membrane Ultrex CMI-7000 (Membranes International, U.S.A.). Sealing was ensured by a rubber o-ring inserted between the two frames. The total volume for each compartment was 114 mL. The anode used was an MMO Ti/ Ru0.7Ir0.3O2 electrode with 12 g m
2 coating on Ti mesh (dimensions: 4.8  5 cm; thickness: 1 mm; specific surface area: 1.0 m2 m
2), supplied by Magneto Special Anodes (The Netherlands). The cathode used was a stainless steel woven wire mesh (dimensions: 4.8  5 cm; 80 mm  0.050 mm wire diameter), and both the anode and cathode had a projected electrode surface area of 24 cm2. For electrochemical control, a Wenking potentiostat/galvanostat (KP07, Bank Elektronik, GmbH, Pohlheim, Germany) was used. The anodic half-cell potentials were measured by placing an Ag/AgCl reference electrode (assumed þ0.197 V vs SHE) in the anode compartment. The cathode medium in was a 0.1M HCl solution. The experiments were performed at constant liquid flowrate at room temperature (25  1  C), with galvanostatic control in: i) continuous mode, with step-wise increase in current density (J ) (J ¼ 1, 10, 30, 50, 100, 150, 200 and 250 A m
2), and samples taken after 75 min of operation at each current density, and ii) batch mode at J ¼ 250 A m
2, and samples taken after 2, 4, 7 and 23.5 h. In the continuous mode experiments, the system was operated at a hydraulic retention time (HRT) of 8.8 min (i.e. flow-rate of 13 mL min
1). In order to maintain well-mixed conditions and avoid concentration gradients, both anolyte and catholyte were recirculated internally at a rate of 162 mL min
1. In order to avoid gas trapping inside the anodic and cathodic compartment and enable stable potentials and ambient pressure, two degassers were installed as illustrated in Figure S1. The volume of ROC in the degassers was maintained at 100 mL, providing a ratio of active and total volume, VACT/VTOT of 0.51 in the continuous mode. Preparative continuous oxidation experiments were carried out in order to establish the time required to reach steady state. In the batch mode experiments, anolyte and catholyte were recirculated at a rate of 162 mL min
1. The volume of ROC used in batch experiments was 10 L, hence the ratio of VACT/VTOT was 0.011. The results of each experimental condition are given as the means of triplicates, with their corresponding standard deviations (SDs). In the continuous

experiments two different ROC samples were used, collected at the abovementioned AWTP within a one month time span and marked as ROC-1 and ROC-2 (physico-chemical characteristics of ROC-1 and ROC-2 are given in Table 1), while in batch experiments only ROC-1 was used.

2.4.

Analytical methods

130 mL samples were collected in amber glass bottles. As a variety of generated oxidants can prevent efficient sample stabilization, no quenching agent was added to the sample. Furthermore, quenching of free chlorine may lead to errors in analytical determination of trace organics due to the formation of the original compounds from their N-chloro analogues (Wulfeck-Kleier et al., 2010). Immediately after the sampling, sample pH was adjusted to pH 7.0 by adding an appropriate amount of 0.1M NaOH or 0.1M HCl, and 100 mL samples were extracted on a Visiprep manifold system (SigmaeAldrich, U.S.A.) using Oasis HLB cartridges (200 mg, 6 mL) from Waters Corporation (U.S.A.), previously conditioned with 10 mL of methanol and 10 mL of deionised water (HPLC grade). Additionally, 30 mL samples were taken for the analyses of free and combined chlorine, non-purgeable organic carbon (NPOC) and ultra-violet absorption at 254 nm (UV254). The samples were filtered prior to all measurements using 0.45 mm filters (Millipore, Ireland), thus the determined NPOC is equivalent to DOC. NPOC was calculated as the difference between the total carbon (TC) and inorganic carbon that were determined by the standard high-temperature method (APHA Standard methods, 5310B) at a TC analyser (Tekmar Dohrmann DC-190), and UV254 absorbance was measured using a Varian Cary50 spectrophotometer. Ion chromatography (IC)Dionex 2010 i was used to determine Cl
and SO24 , while concentrations of Fe2þ and Mn2þ ions were determined by ICPOES Vista Pro-CCD (Varian, Australia). Conductivity was measured using a Eutech electrical conductivity meter, while the pH was measured with a Mettler Toledo Seven easy pH meter (Mettler Toledo, Australia). Free available chlorine (FAC) and total chlorine were measured with the N,N-diethyl-pphenylenediamine (DPD) ferrous titrimetric method (APHA Standard methods, 409E). It is important to stress that oxidants other than FAC present in the solution (e.g. H2O2, ClO2) will similarly to chlorine react with DPD to form a red dye, thus interfering with the measurements. Liquid chromatography-mass spectrometry (LC-MS) analyses were performed using a Shimadzu Prominence ultra-fast liquid chromatography (UFLC) system (Shimadzu, Japan) coupled with a 4000 QTRAP hybrid triple quadrupole-linear ion trap mass spectrometer (QqLIT-MS) equipped with a Turbo Ion Spray source (Applied Biosystems-Sciex, U.S.A.). Chromatographic separation was achieved with an Alltima C18

Table 1 e Main characteristics of the two ROC samples used in the experiments.

ROC-1 ROC-2

DOC, mg L
1

SUVA254, L mg
1 m
1

Conductivity, mS cm
1

pH

[Fe2þ], mg L
1

[Mn2þ], mg L
1

[Cl
], g L
1

[SO2
4 ], mg L
1

57.1 57.2

1.6 2.3

4.25 3.97

7.5 7.7

0.22 0.35

227 234

1.5 1.2

241.5 238.7

1582

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 7 9 e1 5 8 6

Column (250  4.6 mm, particle size 5 mm) run at 40  C, supplied by Alltech Associates Inc (U.S.A.). The multi-residue method used is described in the Text S2, and Tables S3 and S4. The recoveries and method quantification limits (MQLs) are summarized in Table S2.

2.5.

Microtox bioassays

For the V. fischeri bioassays, ROC-1 spiked with the mixture of target analytes was electrochemically oxidized in continuous mode with a step-wise increase in current density (i.e., J ¼ 50, 100, 150, 200 and 250 A m
2), and in batch mode at J ¼ 250 A m
2. Sampling times and sample pre-treatment were identical to the ones described in the sections Experimental setup and Analytical methods, respectively. The results of the acute toxicity tests are expressed in baseline-toxic equivalent concentration (TEQ) units, derived from a baseline toxicity quantitative structureeactivity relationship (QSAR) model using a virtual compound with octanol-water partition coefficient (log KOW) of 3 and molecular weight (MW) of 300 g mol
1 as a reference, which equates to an EC50 of 12 mg L
1 (Escher et al., 2008).

3.

Results and discussion

3.1. Electrochemical oxidation - overall organics removal in batch and continuous mode Fig. 1 illustrates the observed removal of DOC and SUVA254, and measured free and total chlorine during continuous oxidation of ROC-1 and ROC-2. There was no DOC removal for

100

0.8

80

0.6

60

0.4

40

0.2

20

0.0

0.0

2.0

100

1.8

0.6

60

0.4

40

0.2

20

0.0

1.6 SUVA254

80

Free and total Cl , mg L

0.8

DOC/DOC0

b 1.0 ROC/ROC

Free and total Cl , mg L

ROC/ROC

a 1.0

low specific electrical charges (Qsp, expressed as A h m
3 ROC) applied (Qsp  307.7 and 153.8 A h m
3 for ROC-1 and ROC-2, respectively), while SUVA254 was relatively constant at Qsp  153.8 A h m
3 for both ROCs tested. By increasing the applied charge, a gradual decrease in DOC was observed, reaching 8.9  1.4 and 26.5  2.3% removal after 769.2 A h m
3 supplied to ROC-1 and ROC-2, respectively. Considering that the initial chloride ion concentrations as well as residual chlorine concentrations measured for the ROC-1 and ROC-2 were similar, more efficient mineralization in the latter case is possibly a consequence of the higher initial specific aromaticity of ROC-2 as expressed by its SUVA254 value (Table 1). The enhanced formation of putative electron shuttles (e.g. porphyrins, quinones) from the aromatic fraction could be responsible for the higher NPOC removal observed for ROC-2, as these species accelerate the electron transfer between the organic matter and oxidants (Nurmi and Tratnyek, 2002). The removal of SUVA254 was also enhanced with increasing the supplied charge, and 28.7  1.8% and 42  2.4% removal was observed after applying 769.2 A h m
3 to ROC-1 and ROC-2, respectively. The DOC removal achieved in batch oxidation of ROC-1 was 30.7  1.8% after passing 1.45 kA h m
3 (Fig. 2). When comparing the oxidation of ROC-1 in the two operational modes tested, the same values of Qsp (w770 Ah m
3) and J (250 A m
2) in batch and continuous and oxidation of ROC-1 rendered DOC removal of w27 and 8.9  1.4%, and SUVA254 was 1 and 0.71 (i.e. 28.7  1.8% of SUVA254 removal), respectively (Figs. 1 and 2). Thus, the removal of organic carbon depended on the accumulation of long lived oxidants (e.g. HClO/ClO
, HOBr and H2O2) in the bulk liquid. For the final DOC removal achieved in continuous (8.9  1.4%) and batch mode (30.7  1.8%) oxidation of ROC-1 at 250 A m
2, energy consumption calculated according to Bolton et al. (2001) was 0.704 kWh g
1 DOC and 0.350 kWh g
1 DOC, respectively. Thus, more efficient removal of organic matter is achieved in batch mode, and the energy input per mass unit of DOC removed is lower compared to the continuously operated reactor. Nevertheless, lower throughput of batch mode compared to a continuous mode may increase significantly the total cost of the treatment.

1.4 1.2 1.0

0.0 0

1 (3.1)

10 (30.8)

30 (92.3)

50 100 150 (153.8) (307.7) (461.5)

200 (615.4)

250 (769.2)

Current density, A m (Specific electrical charge supplied, A h m )

Fig. 1 e Removal of DOC, SUVA254 and generation of free and total chlorine during anodic oxidation in continuous mode of: a) ROC-1 and b) ROC-2 spiked with the trace organic contaminants, vs J (in A mL2) and Q (in A h m-3). ADOC/DOC0, -SUVA254/(SUVA254)0, e e e free chlorine, ————total chlorine.

0.8 0.6

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

Q, kA h m-3

Fig. 2 e DOC removal and SUVA254 vs Q in anodic oxidation in batch mode at 574 J [ 250 A mL2 of ROC-1 spiked with the SUVA254, DOC/ trace organic contaminants. DOC0.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 7 9 e1 5 8 6

The decrease in SUVA254 and the limited DOC removal observed in the continuous experiments suggested that organic intermediates were mainly generated by oxidative cleavage and opening of the aromatic moieties. On the other hand, introduction of auxochrome substituents (e.g. eCl, NH2Cl, -Br, -OH) into the aromatic rings during batch oxidation led to an initial increase in SUVA254, as DOC decay was faster than the decrease in UV254 absorbance. Much higher share of active volume (i.e. ROC oxidized inside the anodic compartment) in a continuous reactor (VACT/VTOT ¼ 0.51) compared to the batch reactor (VACT/VTOT ¼ 0.011) led to enhanced reaction of organic matter with short-lived radical oxygen species (ROS) and/or radical halogen species (RHS), which are more capable of ring-opening than more stable oxidants (e.g. FAC, O2, H2O2). However, only partial DOC removal was achieved in both continuous and batch oxidation, indicating the accumulation of oxidation intermediates.

3.2. Removal of trace organic contaminants in continuous mode Fig. 3 illustrates the removals of trace organic compounds observed in the continuous experiments conducted at higher current densities (J ¼ 100e250 A m
2). The removals obtained at lower currents (J ¼ 1e50 A m
2) are represented in Figure S2. The term “removal” is used here for the conversion of a target analyte to compounds other than the parent compound. Since at low current densities (J ¼ 1, 10 A m
2) the likely mechanism for oxidation of pollutants is direct electrolysis, the rate of direct oxidation depends on the adsorption properties of the anode surface, and concentration and nature of trace organic

1583

compounds and their degradation intermediates (Panizza, 2010). Except for sertraline, removed at 70% efficiency, for most of the target analytes no removal or very low removal (30%) was observed under these conditions (Figure S2). Considering the adsorption properties of RuO2/IrO2 anodes, very high removal of sertraline even at J ¼ 1 A m
2 can be explained by its hydrophobic nature, as sertraline has the highest log KOW value (5.29) among the selected analytes. The increase in applied charge to 92.3 and 153.8 A h m
3 (i.e. J ¼ 30 and 50 A m
2) exerted a notable effect on the removal of acetaminophen (40% and 90%), diclofenac (40% and 88%), sulfadiazine (44% and 88%), diazinon (40% and 70%) and norfloxacin (65% and 90%), and led to a complete removal of ranitidine (100%) and lincomycin (>90% and 100%, respectively). Bergmann and Koparal (2005a) reported an increase in RuO2/ IrO2 electrode activity anode potentials (EAN) higher than 1.3 V, related to the production of FAC. In accordance with this study, already at J ¼ 30 A m
2 (i.e. EAN ¼ 1.32  0.05V, Table S5) the combined chlorine concentration of 3.0 mg L
1, calculated as the difference between the total and free chlorine measured (Fig. 1) implied that chlorine had already reacted to form organic and inorganic chloramines. Considering the affinity of the abovementioned compounds towards chlorine (Bedner and MacCrehan, 2006a; Westerhoff et al., 2005; Acero et al., 2008; Dodd et al., 2005), it can be assumed that they were oxidized by free chlorine. However, under the same conditions, other compounds known to have a high affinity for FAC (Bedner and MacCrehan, 2006a, 2006b; Westerhoff et al., 2005; Acero et al., 2008; Chen and Young, 2008; Chamberlain and Adams, 2006; Gould and Richards, 1984) had low removals. Examples of these are trimethoprim (25% and 46%), gemfibrozil (11% and

Fig. 3 e Removals of target analytes during anodic oxidation in continuous mode of: a) ROC-1 and b) ROC-2 spiked with the trace organic contaminants, for the tested J in the range 100e250 A mL2. RNT-ranitidine, LNC lincomycin, ACTPacetaminophen, DCF-diclofenac, DZN diazinon, ENR-enrofloxacin, SDZ-sulfadiazine, TMP-trimethoprim, NFL-norfloxacin, SRL-sertraline, GMF-gemfibrozil, CTP-citalopram, VNF-venlafaxine, MTP-metoprolol, DIU-diuron, HCThydrochlorothiazide, CAFF-caffeine, ROX-roxithromycin, TML-tramadol, CBZ carbamazepine, ATZ-atrazine, METmetolachlor, IBU-ibuprofen, PNT-phenytoin, IPM-iopromide, 2,4-D-2,4-dichlorophenoxyacetic acid, TPR triclopyr. Values are expressed as mean with their SDs.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 7 9 e1 5 8 6

43%), caffeine (14% and 26%), diuron (14% and 33%), metoprolol (10% and 30%) at J ¼ 30 and 50 A m
2, respectively. Given that their initial concentrations were very similar, it appears unlikely that this would be a due to lower mass transfer coefficients. Moreover, at J ¼ 100 A m
2 (EAN ¼ 1.70  0.2 V) only 60% of caffeine was oxidized in both ROC-1 and ROC-2, despite its reported high reactivity with FAC in the pH range found in our experiments (e.g. 6.17 at 100 A m
2,Table S5) (Gould and Richards, 1984). At the same current density of 100 A m
2 (Qsp ¼ 307.7 A h m
3), a complete disappearance of enrofloxacin was noted. Interestingly, this compound is known to react very slowly with FAC and to be recalcitrant towards NH2Cl (Dodd et al., 2005). A complete disappearance of carbamazepine from the ROC oxidized at J ¼ 150 A m
2 was probably achieved by oxidants other than chlorine, since carbamazepine is recalcitrant even towards the much stronger oxidant ClO2 (Huber et al., 2005). The largely incomplete oxidation by FAC for some of the compounds and unexpectedly rapid disappearance of others known to react slowly with chlorine suggest that other oxidants in the bulk and/or surface reactions may play an important role. Besides reacting in the bulk, chlorine can react electrochemically at the anode forming adsorbed chloro- and oxychloro-radicals, which mediate the degradation of adsorbed organics (Bergmann and Koparal, 2005a; Papastefanakis et al., 2010). Moreover, electrogenerated O2 can indirectly oxidize the bulk organics and form organic radicals via the hydrogen abstraction mechanism (Carlesi Jara et al., 2007). Organic radicals can then further react with O2 to form organic hydroperoxides (ROOH) that are short-lived and tend to decompose, often leading to the formation of molecules with a lower number of carbon atoms. Furthermore, in the presence of Fe2þ and Mn2þ ions in ROC (Table 1) and electrogenerated H2O2, contribution of Fenton reaction to bulk oxidation can be expected. Compounds that were more recalcitrant during electrochemical oxidation were characterized by either the absence of nucleophilic substituents that have an activating effect on the aromatic ring (e.g. ibuprofen, phenytoin, metolachlor, N,N-diethyl-meta-toluamide (DEET)), or by the decreased electron density on the aromatic ring due to the presence of electrophilic halogen groups (e.g. 2,4-dichlorophenoxyacetic acid (2,4-D), atrazine, triclopyr and iopromide). Further increasing the applied current density up to 250 A m
2 (Qsp ¼ 769.2 A h m
3) led to an enhanced oxidation of atrazine (92%), metolachlor (57% and 84%), ibuprofen (46% and 67%) and phenytoin (61% and 67%) in the ROC-1 and ROC-2, respectively. Considering the EAN (i.e. 2.5  0.07 V) and the measured pH (i.e. pH 2.6, Table S5), oxidants such as ClO2, HClO/OCl
, H2O2 and other radical ROS (e.g. O2
, OH, HO2) --and RHS (Br , Br2 , Cl , Cl2 ) probably had a greater participation in the indirect oxidation at this current density. Furthermore, at acidic pH (i.e. pH3) chlorination is acidcatalysed through a mechanism involving H2OClþ (Rebenne et al., 1996). This phenomenon could be responsible for the sharp increase in atrazine removal when the current density increased from 150 to 200 A m
2, as in the latter conditions the pH rapidly dropped from 5.13 to 2.91 (Table S5). In a previous study Malpass et al. (2006) demonstrated a dependency of atrazine electrochemical oxidation on the presence of FAC. Moreover, the pH will influence the active surface sites of

MMO anodes as well as their redox properties, accelerating the direct electron transfer reactions (Rossi et al., 2009). On the other hand, only around 30e40% of 2,4-D, triclopyr and iopromide removal was observed under these conditions, while DEET could not be oxidized.

3.3. mode

Removal of trace organic contaminants in batch

To further investigate the effect of a prolonged exposure of trace organic contaminants in ROC to the generated oxidants, batch experiments with external circulation were performed at J ¼ 250 A m
2. Triclopyr, 2,4-D, ibuprofen, iopromide, metolachlor, phenytoin, atrazine, and DEET were also in the batch tests more persistent than the other analysed compounds (Fig. 4 and Table S6). The oxidation of all target compounds was more efficient in batch mode due to the accumulation of oxidants in the bulk liquid, and consequently more intense indirect oxidation (Figure S3). With triclopyr as exception, all persistent trace compounds were completely oxidized in batch mode after applying 1.45 kA h m
3. Therefore, these compounds require considerably higher electrical charge supplied per anolyte unit volume and longer residence times in order for indirect oxidation to occur.

3.4.

Bioluminescence inhibition in Vibrio fischeri

To verify the effect of electrochemical oxidation on the toxicity of ROC a number of bioluminescence assays were performed using V. fischeri. Although for specifically acting compounds the baseline toxicity is generally marginal, in a mixture of a large number of trace organic contaminants and other chemicals with a variety of specific modes of action, the baseline toxicity could dominate the overall mixture effect (Escher and Schwarzenbach, 2002). Thus, baseline toxicity provides an integrative measure of the combination of chemicals that act together in concert. The contribution of each chemical is weighted only by its hydrophobicity. In order to follow the change of baseline toxicity of the target compounds, the bioassays were performed using the samples

100 80 Removal, %

1584

60 40 20 0

0.0

0.2

0.4

0.6

0.8 1.0 Q, kA h m-3

1.2

1.4

1.6

Fig. 4 e Removal of persistent target analytes vs Q in anodic oxidation in batch mode at J [ 250 A mL2, of ROC-1 spiked TPR, with the trace organic contaminants. 2,4-D, IBU, IPM, MET, PNT, ATZ, DEET.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 7 9 e1 5 8 6

enriched by the same SPE protocol as the one previously described in the section Analytical methods. The majority of matrix components such as salts, particulates, and generated oxidants were removed by the SPE sample pre-treatment. While the sample extracts of the untreated ROC-1 spiked with target analytes exhibited a TEQ value of 3.7  0.02 mg L
1, the toxicity of sample extracts of ROC-1 oxidized in continuous mode drastically increased with increased current density (209.1  14.4 mg L
1 at J ¼ 250 A m
2) (Fig. 5a). Likewise, the toxic response of V. fischeri increased with higher applied charge for the sample extracts of ROC-1 oxidized in batch mode (Fig. 5b). The TEQ value was increased from the initial 4.3  0.1 mg L
1 to 151.0  4.9 mg L
1 for ROC-1 oxidized in the batch reactor after applying 1.45 kA h m
3. Although the relative increase in aromaticity (i.e. SUVA254) in batch oxidation suggested a higher accumulation of substituted aromatic intermediates than in the continuous mode, toxic oxidation products were formed in both reactors. It is important to note that the bioassay with V. fischeri cannot distinguish between the effects of the SPE enriched trace organic pollutants and the co-extracted organic matter. The hydrophobic (i.e. peptides and protein fragments) and/or aromatic fraction of the organic matter (fulvic- and humic-like substances) contributed to the

Toxic equivalent concentration (TEQ), mg L-1

a

250 200 150 100 50 0

Toxic equivalent concentration (TEQ), mg L-1

b

0

50 (153.8)

100 (307.7)

150 (461.5)

200 (615.4)

250 (769.2)

J, A m-2 (Q, A h m-3) 180

1585

observed increase in the toxicity, as this fraction was probably well retained on the SPE cartridge. Nevertheless, TEQ values determined for the ROC oxidized in batch and continuous mode are significantly higher than the values that could be expected owing to any potential by-products of the selected micropollutants.

4.

Conclusions

Electrochemical oxidation at various current densities was investigated for the treatment of a reverse osmosis concentrate spiked with a mixture of pharmaceuticals and pesticides. The removal of DOC depended on the accumulation of oxidants in the bulk liquid. Based on the changes in specific aromaticity, expressed as SUVA254, it appears that the formation of chloro-, bromo- and hydroxyl-substituted aromatic intermediates was enhanced due to the prolonged indirect oxidation. Furthermore, the employed Ti/Ru0.7Ir0.3O2 electrode was capable of oxidizing most of the selected pharmaceuticals and pesticides, while the most persistent compounds had electrophilic substituents at the aromatic ring (2,4-D, atrazine, triclopyr, iopromide), or an aromatic ring insufficiently activated towards nucleophilic attack (ibuprofen, phenytoin, metolachlor, DEET). Nevertheless, the results of the bioluminescence bioassay imply the formation of toxic by-products during both continuous and batch electrochemical oxidation of ROC. Although the contribution of by-products of the investigated pharmaceuticals and pesticides, and compounds formed by oxidation of other organic matter (e.g. fulvic- and humic-like substances) to the measured toxicity is uncertain, participation of oxidants such as active chlorine and bromine in indirect oxidation will likely cause transformation of organic compounds to their halogenated derivatives. This may represent an insurmountable barrier to the environmental applications of RuO2/IrO2-coated Ti electrodes as such byproducts could be persistent and/or require extreme treatment conditions (i.e. very high electrical charges applied). We are currently investigating downstream treatment options.

160 140 120

Acknowledgements

100 80 60 40 20 0

0

121.6

246.4

437.0

1455.7

Q, A h m-

Fig. 5 e Bioluminiscence inhibition of ROC-1 spiked with target contaminants on Vibrio fischeri at 30 min expressed as baseline e Toxic Equivalent Concentration (TEQ) in mg L-1 in: a) continuous experiment conducted at J [ 50, 100, 150, 200 and 250 A mL2, b) batch experiment conducted at J [ 250 A mL2. Each replicate of the sample was tested in duplicates, at eight different concentrations. Results are expressed as average of duplicates ± standard deviation.

This research was supported by the Australian Research Council (grants LP0989159 and DP0985317), Veolia Water Australia, Water Secure, Magneto Special Anodes and The Urban Water Security Research Alliance. The authors would like to thank to Dr Miroslava Macova and Prof Beate Escher from National Research Centre for Environmental Toxicology, The University of Queensland, for performing the Microtox bioassays. We would also like to thank Mr Pieter Hack (Magneto Special Anodes) and Mr Yvan Poussade (Veolia Water Australia) for valuable comments.

Appendix. Supplementary data Supplementary data related to this article can be found online at: doi:10.1016/j.watres.2010.11.035.

1586

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 7 9 e1 5 8 6

references

Acero, J.L., Benitez, F.J., Real, F.J., Gonzalez, M., 2008. Chlorination of organophosphorus pesticides in natural waters. J. Hazard. Mater 153, 320e328. APHA 409E, 1975. Standard Methods for the Examination of Water and Wastewater. American Water Works Association (AWWA), Washington, DC. APHA 5310B, 1998. Standard Methods for the Examination of Water and Wastewater. American Water Works Association (AWWA), Washington, DC. Bedner, M., MacCrehan, W.A., 2006a. Transformation of acetaminophen by chlorination produces the toxicants 1,4benzoquinone and N-acetyl-p-benzoquinone imine. Environ. Sci. Technol. 40, 516e522. Bedner, M., MacCrehan, W.A., 2006b. Reactions of the aminecontaining drugs fluoxetine and metoprolol during chlorination and dechlorination processes used in wastewater treatment. Chemosphere 65, 2130e2137. Bellona, C., Drewes, J.E., 2007. Viability of a low-pressure nanofilter in treating recycled water for water reuse applications: a pilot-scale study. Water Res. 41, 3948e3958. Bergmann, M.E.H., Koparal, A.S., 2005a. Studies on electrochemical disinfectant production using anodes containing RuO2. J. Appl. Electrochem. 35, 1321e1329. Bolton, J.R., Bircher, K.G., Tumas, W., Tolman, C.A., 2001. Figuresof-merit for the technical development and application of advanced oxidation technologies for both electric- and solardriven systems. Pure Appl. Chem. 73, 627e637. Carlesi Jara, C., Fino, D., Specchia, V., Saracco, G., Spinelli, P., 2007. Electrochemical removal of antibiotics from wastewaters. Appl. Catal. B. Environ. 70, 479e487. Chamberlain, E., Adams, C., 2006. Oxidation of sulfonamides, macrolides, and carbadox with free chlorine and monochloramine. Water Res. 40, 2517e2526. Chen, W.H., Young, T.M., 2008. NDMA formation during chlorination and chloramination of aqueous diuron solutions. Environ. Sci. Technol. 42, 1072e1077. Dialynas, E., Mantzavinos, D., Diamadopoulos, E., 2008. Advanced treatment of the reverse osmosis concentrate produced during reclamation of municipal wastewater. Water Res. 42, 4603e4608. Dodd, M.C., Shah, A.D., von Gunten, U., Huang, C.H., 2005. Interactions of fluoroquinolone antibacterial agents with aqueous chlorine: reaction kinetics, mechanisms, and transformation pathways. Environ. Sci. Technol. 39, 7065e7076. Escher, B.I., Schwarzenbach, R.P., 2002. Mechanistic studies on baseline toxicity and uncoupling of organic compounds as a basis for modeling effective membrane concentrations in aquatic organisms. Aquat. Sci. 64, 20e35. Escher, B.I., Bramaz, N., Mueller, J.F., Quayle, P., Rutishauser, S., Vermeirssen, E.L.M., 2008. Toxic equivalent concentrations (TEQs) for baseline toxicity and specific modes of action as a tool to improve interpretation of ecotoxicity testing of environmental samples. J. Environ. Monit. 10, 612e621. Gallard, H., Leclercq, A., Croue, J.P., 2004. Chlorination of bisphenol A: kinetics and by-products formation. Chemosphere 56, 465e473.

Gould, J.P., Richards, J.T., 1984. The kinetics and products of the chlorination of caffeine in aqueous solution. Water Res. 18, 1001e1009. Huber, M.M., Korhonen, S., Ternes, T.A., von Gunten, U., 2005. Oxidation of pharmaceuticals during water treatment with chlorine dioxide. Water Res. 39, 3607e3617. Malpass, G.R., Miwa, D.W., Machado, S.A., Olivi, P., Motheo, A.J., 2006. Oxidation of the pesticide atrazine at DSA electrodes. J. Hazard. Mater 137, 565e572. Nurmi, J.T., Tratnyek, P.G., 2002. Electrochemical properties of natural organic matter (NOM), fractions of NOM, and model biogeochemical electron shuttles. Environ. Sci. Technol. 36, 617e624. Panizza, M., 2010. In: Comninellis, C., Guohua, C. (Eds.), Importance of Electrode Material in the Electrochemical Treatment of Wastewater Containing Organic Pollutants in Electrochemistry for the Environment. Springer New York, New York. Papastefanakis, N., Mantzavinos, D., Katsaounis, A., 2010. DSA electrochemical treatment of olive mill wastewater on Ti/ RuO2 anode. J. Appl. Electrochem. 40, 729e737. Perez, G., Fernandez-Alba, A.R., Urtiaga, A.M., Ortiz, I., 2010. Electro-oxidation of reverse osmosis concentrates generated in tertiary treatment. Water Res. 44, 2763e2772. Radjenovic, J., Petrovic, M., Ventura, F., Barcelo, D., 2008. Rejection of pharmaceuticals in nanofiltration and reverse osmosis membrane drinking water treatment. Water Res. 42, 3601e3610. Rajkumar, D., Kim, J.G., Palanivelu, K., 2005. Indirect electrochemical oxidation of phenol in the presence of chloride for wastewater treatment. Chem. Eng. Technol. 28, 98e105. Rebenne, L.M., Gonzalez, A.G., Olson, T.M., 1996. Aqueous chlorination kinetics and mechanism of substituted dihydroxybenzenes. Environ. Sci. Technol. 30, 2235e2242. Rossi, A., Alves, V.A., Da Silva, L.A., Oliveira, M.A., Assis, D.O.S., Santos, F.A., De Miranda, R.R.S., 2009. Electrooxidation and inhibition of the antibacterial activity of oxytetracycline hydrochloride using a RuO2 electrode. J. Appl. Electrochem. 39, 329e337. Snyder, S.A., Adham, S., Redding, A.M., Cannon, F.S., DeCarolis, J., Oppenheimer, J., Wert, E.C., Yoon, Y., 2007. Role of membranes and activated carbon in the removal of endocrine disruptors and pharmaceuticals. Desalination 202, 156e181. Van Hege, K., Verhaege, M., Verstraete, W., 2002. Indirect electrochemical oxidation of reverse osmosis membrane concentrates at boron-doped diamond electrodes. Electrochem. Commun. 4, 296e300. Westerhoff, P., Moon, H., Minakata, D., Crittenden, J., 2009. Oxidation of organics in retentates from reverse osmosis wastewater reuse facilities. Water Res. 43, 3992e3998. Westerhoff, P., Yoon, Y., Snyder, S., Wert, E., 2005. Fate of endocrine-disruptor, pharmaceutical, and personal care product chemicals during simulated drinking water treatment processes. Environ. Sci. Technol. 39, 6649e6663. Wulfeck-Kleier, K.A., Ybarra, M.D., Speth, T.F., Magnuson, M.L., 2010. Factors affecting atrazine concentration and quantitative determination in chlorinated water. J. Chromatogr. A. 1217, 676e682.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 8 7 e1 5 9 6

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Transformation of the antiepileptic drug oxcarbazepine upon different water disinfection processes Zhi Li a,b, He´le`ne Fenet a, Elena Gomez a, Serge Chiron b,* a b

UMR 5569 ‘Hydrosciences Montpellier’ University of Montpellier I, 15 Avenue Ch. Flahault, BP 14491, 34093 Montpellier cedex 5, France Laboratoire Chimie Provence, Aix-Marseille Universite´s-CNRS (UMR 6264), 3 place Victor Hugo, 13331 Marseille cedex 3, France

article info

abstract

Article history:

Transformation of the pharmaceutical oxcarbazepine (OXC), a keto analogue of carbama-

Received 16 July 2010

zepine (CBZ) was investigated under different water disinfection processes (ozonation,

Received in revised form

chlorination and UV irradiation) to compare its persistence, toxicity and degradation

18 November 2010

pathways with those of CBZ. Analysis by LCeion trapeMSn allowed for the identification of

Accepted 25 November 2010

up to thirteen transformation products (TPs). The major abundant and persistent TPs

Available online 3 December 2010

(10,11-dihydro-10,11-trans-dihydroxy-carbamazepine (DiOH-CBZ), acridine (ACIN) and 1-(2benzaldehyde)-(1H, 3H)-quinazoline-2,4-dione (BQD)) were identical to those previously

Keywords:

reported during water treatment of CBZ. Only one new compound arising from an intra-

Oxidation processes

molecular cyclisation reaction was identified during UV irradiation. OXC reacted quickly

Ozone

with hydroxyl radical and relatively rapidly with free chlorine while slow reaction rates

Chlorine

were recorded in presence of ozone and upon UV irradiation. An increase of the acute

UV irradiation

toxicity of UV irradiated solutions, monitored by a Daphnia magna bioassay, was recorded,

Oxcarbazepine

probably due to the accumulation of ACIN. The formation of ACIN is of concern due to the carcinogenic properties of this chemical. ACIN was also generated during the direct UV photo transformation of DiOH-CBZ and 10-hydroxy-10,11-dihydro-carbamazepine (OHCBZ), two metabolites of OXC and CBZ widely detected in water resources. Analysis of tap water samples revealed the occurrence at ng/L levels of the major TPs detected under laboratory scale experiments, except ACIN. ª 2010 Elsevier Ltd. All rights reserved.

1.

Introduction

Contamination of water resources by micro-contaminant residues is one of the major current challenges for the preservation and sustainability of the environment. In this field, the focus for water pollution research has recently been extended from priority contaminants such as pesticides to the so-called emerging contaminants. An important group of these emerging contaminants are the pharmaceutical products (PPs) since many industralized countries have discovered drug products in their water resources often used for drinking water purposes. Detection of PPs in drinking water has been

up to date quite scarce (Monpelat et al., 2009; Benotti et al., 2009). From water resource to drinking water network, the removal efficiency of PPs in drinking water treatment plants (DWTPs) is rather well established (Vieno et al., 2007) reaching for instance, 98% for acetaminophen and 85% for carbamazepine in a classical DWTP including clarification, filtration and disinfection by chlorination (Stackelberg et al., 2007). However, there is currently a lot of concern regarding the fate of these compounds during chemical disinfection processes. Chlorine is the most widely used disinfectant and as a selective oxidant only allows for the removal of a limited number of PPs from water matrices (Acero et al., 2010). However, many

* Corresponding author. Tel.: þ33 4 91 10 85 25; fax: þ33 4 91 10 63 77. E-mail address: [email protected] (S. Chiron). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.11.038

1588

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 8 7 e1 5 9 6

drinking water facilities are changing their primary disinfectant from chlorine to alternative disinfectants (e.g., ozone, ultraviolet (UV) irradiation, chlorine dioxide, and chloramine) because they generally reduce regulated trihalomethane and haloacetic acid formation levels. The use of UV irradiation, in addition to pre- or post-chlorination is gaining interest due its effectiveness for inactivation of Cryptosporidium. At the UV-C (254 nm) drinking water fluence (dose) of 400 J m2, the degree of selected pharmaceutical elimination at pH 7 strongly depends on their chemical structure and ranges for instance from 0.4% for 17-b-ethinylestradiol to 15% for sulfamethoxazole (Canonica et al., 2008), but indicates that photo transformation should be seriously taken into account when evaluating the possibility of formation of transformation products (TPs). The application of ozone in drinking water treatment is also widespread affording higher oxidation rates of micro-contaminants than free chlorine does (Westerhoff et al., 2005). Recently, it has been shown that the oxidation of selected PPs by ozone can lead to an effective transformation of many drugs during drinking (Brose´us et al., 2009) and wastewater (West et al., 2009) treatment. Upon disinfection processes, PPs can be therefore transformed into potentially toxic TPs but TPs identification has been limited to a few cases. A wide variety of (multi)-chlorinated and hydroxylated products may be formed and can be expected for conditions typical of wastewater and drinking water chlorination (Dodd and Huang, 2007). For instance, halogenated derivatives of salicylic acid occur in drinking water after chlorination (Quintana et al., 2010) while the chlorination of paracetamol yields the N-acetyl-p-benzoquinone imine and 1,4-benzoquinone toxicants (Bedner and MacCrehan, 2006). Aldehyde moieties which are known to interact with DNA are generated from the ozonation of b-blockers (Benner and Ternes, 2009) and roxithromycin TPs generated under ozonation still possess an unmodified desosamine group and might preserve some biological activity (Radjenovic et al., 2009). The carcinogen acridine is also formed during the UV irradiation of the antiepileptic drug carbamazepine (CBZ) (Vogna et al., 2004). Consequently, to assess the overall efficiency of the elimination of PPs during water treatment, the formation of TPs, their persistence, and relative toxicity compared to the parent should be considered. This will be the main contribution of this work by taking oxcarbazepine (OXC) as a probe compound. OXC is the keto analogue of CBZ and is a more recently marketed antiepileptic drug which has chemical and therapeutic similarities to CBZ but with reduced side effects. Consequently, CBZ is frequently substituted by OXC. OXC is mainly metabolized to 10-hydroxy-10,11-dihydroxy-CBZ (OH-CBZ) and this therapeutically active metabolite further yields 10,11-dihydro-10,11-trans-dihydroxy-CBZ (DiOH-CBZ) a common metabolite of CBZ and OXC (Breton et al., 2005). OXC together with its major metabolites have been detected at concentration levels as high as those of CBZ at WWTPs outlet (Leclercq et al., 2009). Consequently, the different aims of this work are the followings: 1) To investigate kinetics and transformation pathways of OXC upon different chemical oxidation processes (ozonation, chlorination and UV irradiation), 2) To compare the fate of OXC with that of CBZ (Kosjek et al., 2009) in water treatment and 3) To screen for the occurrence of major identified TPs in tap water.

2.

Materials and methods

2.1.

Chemicals

Carbamazepine (CBZ), acridine (ACIN) and acridone (ACON) were purchased from SigmaeAldrich (Saint Quentin Fallavier, France). Oxcarbazepine (OXC) was kindly supplied by Novartis Pharma (Basel, Switzerland). 10-hydroxy-10,11-dihydro-carbamazepine (OH-CBZ), 10,11-dihydro-10,11-trans-dihydroxy-carbamazepine (DiOH-CBZ), [2H2]carbamazepine were purchased from Toronto Research Chemicals (Toronto, Canada). 10,11-dihydro-10,11-cisdihydroxy-carbamazepine was easily prepared by osmiumtetroxide oxidation of CBZ as previously reported (Baker et al., 1973). Sodium hypochlorite solution NaOCl (10% available chlorine) was purchased from Fluka. Water used in all experiments and in the preparation of buffers, was purified by a Milli-Q filtration system (Millipore). Acetonitrile and methanol (LC grade) were purchased from Merck (Darmstadt, Germany). 1-(2-benzaldehyde)-(1H, 3H)-quinazoline-2,4-dione (BQD) was synthesized as previously described (McDowell et al., 2005). Briefly, 50 mg L1 (Milli-Q-water) CBZ were ozonated. An approximate ozone dose of 400 mM was added and allowed to react for 20 min. Excess ozone was then purged with helium, and the sample was freeze-dried to yield the ozonation TPs. The freeze-dried sample was then dissolved in approximately 2 mL of acetonitrile:water (50:50, v/v). To isolate BQD from the other oxidation products, fractions were collected from 300 mL injections of the dissolved material after separation on an LCeUV (l ¼ 278 nm) Merck Lichrospher semi-preparative C-18 column (250 mm  10 mm i.d.). The LC mobile phase was isocratic using acetonitrile:water (50:50, v/v) and a flow rate of 1 mL min1. Fractions were combined and the solvent was removed under a gentle nitrogen stream (T ¼ 50  C) to yield a white crystalline residue. Ozonation of 50 mg of CBZ produced up to 10 mg of BQD with this method for use as an analytical standard. 1H and 13C NMR data were compatible with the structure and are consistent with those reported by McDowell et al. (2005).

2.2.

Treatment experiments

All experiments were carried out either in distilled water or in synthetic surface water containing 25 mg L1 as CaCO3 total alkalinity, 4.5 mg (C) L1 dissolved organic carbon (DOC) as humic acids from SigmaeAldrich, 12.5 mg L1 nitrate ions and showing 75% UV transmission. pH adjustment was done with borate buffer (10 mM) for pH value above 8 and phosphate buffer (20 mM) for pH value below 8. The measured pH never varied by more than 0.1 unit during the course of the experiments. At evently spaced time intervals, 1 mL aliquots of the reaction mixture were removed for OXC, TPs and oxidant analysis. All experiments were carried out in triplicate.

2.2.1.

UV irradiation

The experiments were carried out with an immersion-type photoreactor. A low-pressure mercury (LP Hg) lamp Heraeus Noblelight model TNN 15/32 (nominal power 15 W) emitting at 254 nm was employed as UV radiation source. The 185 nm line of the low-pressure Hg-arc was absorbed by the quartz sleeve of the photoreactor. A cooling jacket system made of quartz

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 8 7 e1 5 9 6

and pure water (with negligible light absorption at the wavelength of emitted radiation) maintained a temperature of 25  0.2  C. The whole assembly was mounted on a magnetic stirrer and wrapped with aluminum foil. Initial pH (7e8) was adjusted by dropwise addition of either 0.1 M NaOH or 0.1 M H2SO4. pH was monitored and remained stable during irradiation.

2.2.2.

Chlorination

Stock solutions of chlorine (5e20 mM) were prepared by diluting a commercial solution of sodium hypochlorite (10% active chlorine). Available chlorine in the NaOCl solutions was determined by the iodometric titration method (APHA, 1998). The kinetic experiments were performed by adding an excess of chlorine (30:1 ratio of chlorine to pharmaceutical on a molar basis). The reactions were carried out in glass vials at room temperature (20  2  C) in the presence of 20 mM buffer. Aliquots of the reaction mixture were removed and quenched immediately after sampling by adding 0.1 mL of a fresh sodium sulfite solution (24 mM). The chlorine concentration in surface water experiments was analyzed by the ABTS method (Pinkernell et al., 2000) and remained nearly constant during the experiment course.

2.2.3.

Ozonation

Ozone was produced with a Fischer 502 ozone generator and its stock solutions (w1.5mM) were produced by sparging ozone-containing oxygen through Milli-Q water that was cooled in an ice bath (Bader and Hoigne´, 1981). Stock solutions of ozone were standardized spectrophotometrically based on their molar absorption coefficient 3 ¼ 3000 M1 cm1 at 260 nm (Liu et al., 2001). Experiments were conducted with ozone in excess at pH 8 (phosphate buffer, 50 mM) with or without tertbutanol (t-BuOH, 10 mM) as hydroxyl radical scavengers. The kinetic runs were started by adding an excess of ozone (30:1 ratio of ozone to pharmaceutical on a molar basis). Ozone decay was determined by the indigo/UV method (Bader and Hoigne´, 1981). Results showed that ozone decrease is <5% and it was assumed that the ozone concentration remained constant during the experiment course. The ozone residual was quenched immediately after sampling by adding 0.1 mL of a fresh sodium sulfite solution (24 mM).

2.3.

Determination of kinetic rate constants

The values of the second-order rate constants for the reaction of OXC with chlorine and ozone were calculated from the pseudo-first-rate constants (kobs) by dividing the kobs values by the total concentration of oxidants. In contrast, competition kinetics were used to determine the fast second-order rate constants for the reaction with hydroxyl (OH) radicals according to a methodology described in details elsewhere (Huber et al., 2003). For this purpose, p-chlorobenzoic acid (pCBA, kOH ¼ 5  109 M1 s1) was selected as reference compound. The experiments were carried out at 25  C with Milli-Q water solutions containing equal concentrations of OXC and reference compound (5 mM) and the pH was kept constant at 7 using 5 mM phosphate buffer. Since OXC undergoes slow direct photolysis, OH radicals were generated by photolysis of H2O2. These experiments were carried out

1589

with a 0.5 L cylindrical immersion-type photoreactor (Heraeus TQ 150 Model, radiation path length 2 cm), equipped with a water-cooled, medium-pressure mercury lamp with maximum emission wavelengths at 313, 366, 406, 436, 546, and 578 nm. The reactor is made of Pyrex glass in order to cut off the wavelengths shorter than 290 nm and to minimize direct photolysis of OXC. The whole assembly was mounted on a magnetic stirrer and wrapped with aluminum foil. The k(OH) for OXC was calculated according the following equation: Ln(Pt/P0) ¼ Ln (Rt/R0)  (kP/kR), where P0 and Pt represent OXC concentration at the initial and at any reaction time, respectively, Rt and R0 those of the reference compound, and kP, kR are the second-order reaction rate constants of OXC and reference compound with OH radical, respectively.

2.4.

Samples

Five 2 L tap water samples were collected at five different private homes distributed along the city of Marseille (France) within a 2-h time period in April 2010. All samples were collected in silanized amber glass bottles, preserved with sodium azide and previously washed with acetone, methanol and Milli-Q water. Subsequently after sampling, they were filtered through 0.45-mm nitrocellulose filters (Millipore) and stored at -4  C until analysed (within 48 h). Ascorbic acid (0.6 mg/mL) was added to these samples in order to eliminate residual chlorine.

2.5.

Analytical methods

For kinetic studies, the concentration of OXC was followed by means of a VWR Hitachi HPLC coupled with a UV detector set at l ¼ 220 nm and a LiChrospher C-18 column (250  4 mm i.d., Merck). Elution was carried out with a 50:50 mixture of acetonitrile and aqueous H3PO4 (pH 3), at a flow rate of 1.0 mL min1. Injection volume was 50 mL. For TPs and parent compounds analysis in water samples, sample pH was adjusted to a value of 7 and 1 L samples were preconcentrated with SPE on Oasis HLB 200 mg cartridges (Waters). The analytes were eluted with 2  5 mL methanol. The extract was evaporated to dryness and then redissolved in 200 mL of a water/methanol mixture (50:50 v/v). The analysis of the extract was performed by LCeMS/MS in the multiple reaction monitoring mode (MRM) and in positive electrospray (ESI) ionization mode. Only the most abundant product ion in MS/MS mode was chosen because at ultratrace levels (ng/L), the abundance of the second transition ion was too low to be detected. The selected transition ions were 237 > 194, 253 > 236, 255 > 237, 271 > 253, 180 > 152, 196 > 167 for CBZ, OXC, OH-CBZ, DiOH-CBZ, ACIN, ACON, respectively. Moreover, qualitative analysis of compounds VII, VIII and IX (see Fig. 2) were carried out by using 267 > 249, 283 > 265 and 242 > 224 as specific transition ions. The HPLC system consisted of a Metachem C-18 column 150  2 mm i.d., 3 mm particle size (Varian), and an Esquire 6000 ion trap mass spectrometer (Bruker, Bremen, Germany). The mobile phase used in chromatographic separation consisted of a binary mixture of solvents A (acetonitrile) and B (0.1% formic acid solution) at a flow rate of 0.2 mL min1. The gradient was operated from 5% to 100% A for 25 min and then back to the initial conditions in 5 min. Identification of the target analytes

1590

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 8 7 e1 5 9 6

Fig. 1 e Typical Total Ion Chromatograms (TIC) obtained after a) 1 h UV irradiation in synthetic water, b) 2 min ozonation in distilled water without tert-butanol, c) 2 h chlorination of 30 mg LL1 OXC.

in unknown samples was based on the LC retention time compared to that of a standard (30 s) and the unique combination of a precursor-product ion. The percent absolute recovery was 93  5%, 77  6%, 93  5%, 96  5%, 96  5%, and 90  6% for CBZ, OXC, OH-CBZ, DiOH-CBZ, ACIN and ACON, respectively. Limits of quantification (LOQs) ranged from 1 ng L1 to 5 ng L1 depending on the compound. Quantification was carried out by means of an isotope labeled internal

standard (IS) procedure ([2H2]CBZ, transition ions 239 > 196) spiked at 100 ng L1 in the extracts, because this is the only effective way to overcome the matrix ion suppression effects in electrospray LCeMS/MS. Calibration curves were constructed by plotting the ratio peak area/internal standard peak area against concentration levels. This ratio remained constant whether the matrix was present or not (FeitosaFelizzola et al., 2007). For TPs identification, the mass

1591

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 8 7 e1 5 9 6

HO HO

N C O NH2

indirect photolysis or O3 decomposition

HO

.

OH

.

N C O NH2

O

OH

.

+ H

O

.

O3

X

HOCl / ClOHO O

O O

HO

OO O

HO H2O

O

O

Cl

N C O NH2

XII

Cl

Cl

N C O NH2

VI

V

O

OH

OH O O

IX

N H O

OH HO

-H2O2

O

O

N H

N O

HN N

VII

N C O NH2 O

direct photolysis

HN O

OH

N C O NH2

N C O NH2

N H O

N C O NH2

VI

N C O NH2

III

O

- H

HO

XI

IV

photolyis

N C O NH2

II

H

.

N C O NH2

OH

OH

OH

direct

O

I

.

OH

OO

N C O NH2

O VIII

N

OH O

O HN O VIII

N

OH O

HN O

N C O NH2

VII

N O

Fig. 2 e Proposed transformation pathways of OXC.

spectrometer was operated in the full scan mode with the same LC gradient as used for TPs analysis in water samples.

2.6.

Ecotoxicity assays

Acute toxicity immobilization tests on Daphnia magna were carried out according to the OECD Guideline 202. The test was performed using six concentrations of each standard substance (OXC, DiOH-CBZ and BDQ), working in triplicate in the 2.5e10 mg L1 range for pure standards. To investigate the potential toxicity of the TPs, two dilutions (1:3 and 1:6) of the sample corresponding at the end of treatment time (150e180 min) were used, working in triplicate. In addition, blank solutions were added for each inhibition curve.

3.

Results and discussion

3.1. Identification of transformation products and transformation pathways Thirteen TPs of OXC have been detected by LCeMS. TPs molecular weight (MW) was assigned on the basis of the pseudo-molecular ions [M þ H]þ. On the basis of the MS2/MS3 mass fragmentation patterns alone, a definitive assignment of TPs structure was not possible. The TPs structures were tentatively elucidated by coupling MS data with knowledges on the reactivity of investigated oxidants. Confirmatory methods were also used including the use of authentic standards and matching the TPs fragmentation patterns with those reported in published mass spectra. MS data together with Total Ion Chromatograms (TIC) under different conditions are reported

in Table 1 and Fig. 1, respectively. All MS2/MS3 spectra are provided in the Supplementary Material appendix. Proposed degradation pathways of OXC under different degradation conditions is also provided in Fig. 2.

Table 1 e Intermediates and products detected by ESI (D) LCeMS/MS upon OXC chlorination, ozonation, UV irradiation and biodegradation reactions. [M þ H]þ m/z

MS/MS ions m/z

Identification/ Confirmation method

I (OXC)

253

Authentic standard

II (DiOH-CBZ)

271

III IV V (ACIN) VI

210 226 180 269

VII (BQD)

267

VIII

283

IX

242

236a, 208, 210, 180 253a, 236, 210, 180 195, 182a 208a, 198, 182 152 252a, 226, 226, 224, 208, 196 249a, 239, 224, 196 265, 240, 237, 222a, 194 224a, 196

X

269

XI XII XIII (OH-CBZ)

287 321 255

Compound

251a, 224, 208, 195, 180 251 285 237, 220, 194a

a indicates the most abundant fragment ion.

Authentic standard Kosjek et al. (2009) Kosjek et al. (2009) Authentic standard This study Authentic standard This study This study, Lu et al. (2009) This study This study This study Authentic standard

1592

3.1.1.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 8 7 e1 5 9 6

UV irradiation

In distilled water (direct photolysis), none TPs could be detected while in synthetic surface water UV photolysis of OXC (I) is mainly due to indirect photolysis since the addition of a hydroxyl radical scavenger such as 2-propanol (2% in volume) significantly inhibited its transformation. In water, I is in equilibrium with an enol tautomer, which is problaby stabilized by resonance. UV photolysis of I in surface water is mainly due to indirect photolysis since the addition of a hydroxyl radical scavenger such as 2-propanol (2% in volume) significantly inhibited its transformation. Similarly to CBZ (Kosjek et al., 2009), UV photolysis of I proceeds through two main routes. Both are likely initiated by hydroxyl radical attacks on the a-carbon of the keto moiety due to the nucleophilic properties of this peculiar position. The first route involves the formation of a diol II (Fig. 2), which was identified as DiOH-CBZ by comparing its chromatographic and mass spectrometric behavior with that of an authentic standard. II eluted as two peaks because both the trans and cis stereoisomers were formed, the trans stereoisomer being the most abundant. The trans and cis stereoisomers were distinguished on the basis of the retention time of individual standard. The formation of II is somewhat surprising since it entails a reductive step. However, upon OH radical generation, the formation of an alpha-hydroxy ketone radical might be hypothezised (Fig. 2). This latter radical might evolve by a hydrogen radical transfer reaction between two molecules of this transient radical giving simultaneously rise to II and VI. II further underwent a ring contraction process probably upon direct photolysis as previously described in details (Chiron et al., 2006) to yield III, IV and V. V was identified as ACIN due to the availability of an authentic standard, III as (9H,10H)acridine-9-aldehyde and IV as (hydroxy-(9H,10H)-acridine-9aldehyde due to the consistency of MS2 fragmentation patterns of III and IV with previously published data (Kosjek et al., 2009). For compound IV, several structural isomers are probably possible and further investigation is needed to confirm with certainty its structure. ACON was never detected in our experiments indicating a slow phototransformation rate of ACIN into ACON upon UV irradiation. The second route involves the OXC heterocyclic ring oxidation and the formation of a new carbonyl group to yield VI. The MS2 spectrum of VI was characterized by losses of 17 (m/z 252) and 43 (m/z 226) mass units due to losses of NH3 and CONH2, respectively which supports the preservation of CONH2 lateral chain. In contrast, the lack of the product ion at m/z 180 which accounts for the unmodified acridine structure revealed that the heterocyclic ring has been modified. Additional fragment ions at m/z 224 (-CONH3) and at m/z 196 (-COCONH3), might reveal the formation of a new aldehyde moiety. A detailed mechanism of ring closure of VI by an intramolecular reaction to a quinazoline moiety (VII) is provided elsewhere (McDowell et al., 2005). VII was assigned the structure of BQD due to the availability of an authentic standard. VII is further oxidized to VIII. With an increase in MW of 16 mass units with respect to VII and fragmentation patterns similar to those of VII, VIII was suggested to derive from the oxidation of an aldehyde moiety into a carboxyl

one. VII and VIII both eluted as two peaks, exhibiting the same MW and fragmentation patterns. In each case the second compound is likely to be a stereoisomer of the first one resulting from the hindered rotation of the partial double bond character of the phenyl-N bond. IX probably arises from the cleavage of the CONH2 lateral chain (lack of losses of 17 and 43 mass units featuring the CONH2 lateral chain) together with the oxidation of one of the two aldehyde functions of VI into a carboxylic acid function and followed by an intramolecular cyclisation reaction as previously proposed by Hu et al. (2009). Finally, X with a MS2 spectrum depicting fragment ions at m/z 251 (OH loss) and m/z 208 (CONH2 loss) would probably derive from a hydroxyl radical attack on a C6 aromatic ring and was assigned the structure of hydroxycarbamazepine.

3.1.2.

Ozonation

Ozonation involves two different major oxidative species: ozone and OH radical. To monitor the influence of OH radical reactions in the oxidation process of OXC, experiments in the presence and absence of t-BuOH as a OH radical scavenger were performed.

3.1.3.

Direct oxidation by ozone

According to the results from the ozonation of CBZ (McDowell et al., 2005), the two non activated aromatic rings of OXC are not reactive against ozone attacks. The ozone reaction might be entirely due to a reaction with the enol and in presence of tBuOH only VIII was determined as a major TPs of OXC (see Fig. 1 in the Supporting Material). Ozonide could be formed by ozone cyclo-addition on unsaturated bond of the enol (Fig. 2). The cleavage of the unstable ozonide might lead to a ring opening and the formation of a zwitterionic specie. The hydroxy-hydroperoxide could then evolve to compound VIII by hydrogen peroxide release from the hydroxy-hydroperoxide form following a similar ozonolyis mechanism to that suggested for progesterone for instance (Barron et al., 2006).

3.1.4.

Indirect oxidation by OH radicals

Without t-BuOH, when OH radical and ozone molecule can be concomitantly present, II, VII and VIII were identified as the three major TPs of OXC (Fig. 1b) OH radical originating from ozone decay was probably responsible for the generation of II and VII, by following the same mechanism hypothesized for the indirect photochemical transformation of OXC.VIII might arise both from direct ozone attacks and from further oxidation of VII. Surprisingly, IX was not detected in our experimental conditions.

3.1.5.

Chlorination

Under chlorine treatment, the chlorination pathway of OXC was clearly evidenced by the occurrence of the chlorine isotopic pattern in the XI and XII mass spectra, with XI and XII exhibiting one and two chlorine atoms, respectively. Moreover, the MS2 spectrum of XI and XII showed a chlorine loss with an ion at m/z 250 and m/z 285, respectively. At least two isomers seem to be possible for XI and XII since both compounds were detected as two peaks. Ketone chlorination generally results from initial substitution reactions on the

1593

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 8 7 e1 5 9 6

a-carbon to the carbonyl group, inducing the successive replacement of hydrogen by chlorine (Deborde and von Gunten, 2008). However, in case of OXC, chlorine might also react with the C6 aromatic ring by electrophilic substitutions at the Meta position due to the electron-withdrawing properties of the keto group. In contrast, chlorine is not expected to react with amide. Surprisingly, the formation of VII (BQD) was highly preserved under chlorine treatment and VII was even the major TPs of the reaction at the end of the reaction time (180 min, Fig. 1c). This result implies that the haloform reaction with free chlorine probably prevailed. During this chlorination step, hydrolysis of mono- and dichlorinated TPs probably occurred to yield VI and VII as previously reported in case of the chlorination of ketones (Guthrie et al., 1991).

3.2. Kinetics for oxcarbazepine reaction with different oxidants and upon UV irradiation In Fig. 3 are illustrated the concentration evolution profiles of OXC and its major TPs plotted against time, taking into account that chlorination and ozone reactions were batch experiments while in UV irradiation the number of incident UV quanta increased with time. In Table 2 are reported kinetic rate constants for the reaction of OXC and CBZ with oxidants and upon UV irradiation.

3.2.1.

Direct photolysis

Similarly to CBZ (Pereira et al., 2007), OXC reacted very slowly upon UV irradition in distilled water at pH 7.6. The degradation of OXC was still less than 10% after 180 min reaction (results not shown). Despite its relatively high absorbance (decadic molar absorption coefficient 3 ¼ 7245 M1 cm1 at 254 nm), the quantum yield for transformation of OXC, determined from the low measured pseudo-first order constant, is quite low (f ¼ 8  0.8  104 mol E1). This quantum yield value was experimentally determined as the ratio between the pseudo-first order rate constant (kuv) and the specific rate of light absorbed by OXC at 254 nm (Ks (l ¼ 254 nm) ¼ 11.5 E mol1 s1).

3.2.2.

Indirect photolysis

In synthetic surface water, the degradation of OXC increased probably due to an indirect production of OH radical from reaction of the UV light with DOM and NO 3 ions (Fig. 3). A first apparent order reaction constant value of kuv ¼ 9.2  0.8  103 min1 was calculated. VII (BQD) and IX were the two major TPs of the reaction and the persistence of both compounds was observed. At the end of treatment time (180 min), the concentration of VII reached 56% relative to the initial concentration of OXC. The other two identified TPs

Table 2 e A comparison between apparent second-order rate constants for the reaction of OXC and CBZ with selected oxidants in surface water. Compound O

N CONH2

Oxidant

k at pH 8 (M1 s1)

HO O3a Cl2 UVb

8.4  0.24  109 550  45 6.2  0.2  102 9.2  0.8  103

HO

8.8  0.2  109

O3

3.0  0.3  103

Cl2

<0.1

UVb

2.2  1.3  103

Reference This This This This

study study study study

Oxcarbazepine

N CONH2

Fig. 3 e Product evolution against time of five transformation products of OXC: (B) OXC, (;) BDQ, (>) DiOH-CBZ, (-) ACIN, (6) IX (second axis), (A) VIII (second axis) during a) UV irradiation in synthetic water, b) ozonation and c) chlorination of 30 mg LL1 of OXC.

Carbamazepine

a in presence of tert-butanol. b kuv in min1.

Huber et al. (2003) Huber et al. (2003) Lee and von Gunten (2010) Pereira et al. (2007)

1594

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 8 7 e1 5 9 6

Fig. 4 e Transformation product time profiles under chlorination (pH 7, 50 mg LL1 Cl2). (B) monochloro-OXC, (,) dichloro-OXC; (;) BDQ. Results normalised, e.g. 1 is the maximum observed value in the set of experiments.

(II and V) were encountered in smaller quantities but the increase in V was observed at the expense of II, suggesting the transformation of the former into the latter, probably through a direct photolysis process as previously reported (Chiron et al., 2006). At the end of the experiment, V accounted for 12% of the initial concentration of OXC. The formation of V is of concern because of the carcinogenic properties of this chemical. However, as an end-product of the reaction, V is not expected to be generated under UV doses typically required for water disinfection (40e140 mJ/ cm2) because those latter values were reached in less than 5 min in our experimental set-up. In contrast, V might arise from the direct photo transformation of II (DiOH-CBZ) and

XIII (OH-CBZ), two metabolites of OXC and CBZ which have been detected at higher concentrations than their parent compounds at the outlet of WWTPs (Leclercq et al., 2009) and in surface waters (Miao and Metcalfe, 2003). V is the major TPs resulting from the UV irradiation of II and to a lesser extent XIII after 15 min treatment, as shown in Figs. 2 and 3 in the Supplementary Material. In contrast, II and XIII have turned out to be stable in presence of ozone and chlorine. When t-BuOH (10 mM) was added as OH radical scavengers to solutions used for measurement of O3 rate constant, OXC reacted relatively slowly with O3 (ko3 ¼ 550  45 M1 s1) at pH 8 in distilled water. The constant calculated in this work is far lower than the value reported for CBZ in spite of their similar chemical structure but probably due to the lack of the major site of reaction for ozone (e.g., olefinic carbon atoms). Without t-BuOH, ozonation is influenced by OH radical generated through reactions of O3 with specific functional moieties such as phenol and amine (Buffle and von Gunten, 2006) in DOM and from autocatalytic O3 decomposition. In this case, a pseudo-first-order kinetic constant kobs ¼ 7.7  0.1  103 s1 was determined. Accordingly, >99% transformation of OXC was achieved in 30 min (see Fig. 3b). OH radical oxidation prevails over O3 oxidation due to the high second-order rate constant of the reaction of OH radical with OXC (8.4  0.2  109 M1 s1, determined in this study). OXC is nearly quantitatively oxidized into VII (Fig. 3b) which is subsequently transformed into VIII as previously reported by McDowell et al. (2005). At the end of the reaction time (150 min), VII still accounted for 32% of the initial concentration of OXC while II remained stable during the course of the reaction with a concentration accounted for 9% of the initial concentration of OXC. In contrast to CBZ (kCl2 < 0.1 M1 s1, Lee and von Gunten, 2010), OXC reacted with free chlorine at significantly higher reaction rates in comparison to CBZ. As Deborde and von Gunten (2008) pointed out, the oxidation rate for most of the

Fig. 5 e Multiple Reaction Monitoring (MRM) LCeMS/MS chromatogram obtained after preconcentration of 1 L of tap water: II (12 ± 5 ng LL1), IS (100 ng LL1). VII, VIII and IX detected but not quantified with precision (<10 ng LL1).

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 8 7 e1 5 9 6

chlorine reactions with organic compounds can be described by second-order kinetics, first-order in the free active oxidant ([Cl2]t ¼ [HOCl] þ [OCl]) and first-order in the target compound. As expected, the reaction rates increased as the chlorine concentration increased and as the pH decreased from 10 to 6. Second-order rate constants k were found to be 3.8  0.1  101, 8.9  0.1  102, 6.2  0.2  102, 4.6  0.2  102, 4.1  0.3  102 M1 s1 at pH 6, 7, 8, 9 and 10, respectively. A decrease of one order of magnitude in the k values was recorded between pH 6 and 10 probably due to the higher reactivity of HOCl versus OCl. At pH 7.6, OXC was readily transformed into VII which accounted for 81% of the initial concentration of OXC, while chlorinated derivatives were only detected in small amounts (Fig. 3c). However, due to the potential toxicity of chlorinated derivatives, the influence of pH and chlorine concentrations on their quantitative formation was further investigated. The most important factor for their formation was the chlorine concentration. Mono- and dichlorinated TPs of OXC were quickly generated at high chlorine levels (50 mg L1), but their further degradation occurred also quickly and the only remaining TPs at long reaction times was VII (see Fig. 4).

3.3.

Ecotoxicity determined in D. magna tests

Ecotoxicity results did not show any acute toxic effects for OXC, BQD and DiOH-CBZ in the tested concentration range. For the test solutions containing final ozonation and chlorine treatment TPs, neither acute toxicity nor statistically significant growth inhibition was noted. In contrast, UV irradiated solutions of OXC prepared in synthetic surface water exhibited an increase in acute toxicity corresponding to a percentage of immobilization of 25  5% which was probably due to the formation of acridine. This latter chemical has demonstrated a high toxicity against D. magna (Parkhurst et al., 1981).

3.4.

Application to tap water samples

TPs of OXC and CBZ together with the parent compounds were screened for their occurrence in tap water samples by LCeMS/MS in the MRM mode after the SPE of the samples. A chromatogram of a tap water analysis is presented in Fig. 5. The parent compounds were never detected although they have been widely detected in surface and ground waters. Among the metabolites or TPs of OXC and CBZ, DiOH-CBZ was detected in all samples at concentration levels below 15 ng L1. VII, VIII and IX were detected in three out of five samples. Although a precise quantification of these chemicals was not possible either due to the lack of analytical standard (VIII and IX) or because the recoveries were not determined (VII), they may occur at an average concentration below 10 ng L1 in tap water, assuming a similar response for the three compounds in (þ) ESI ionization mode. These occurrence levels are consistent with those reported for other PPs in drinking water (Benotti et al., 2009). The persistence of TPs of OXC in water distribution system illustrates that longer contact with secondary disinfectants do not play a significant role in removing or lowering the concentrations of these refractory compounds. This is consistent with their resistance to chlorine and ozone oxidation as shown in this work.

4.

1595

Conclusions and outlook

Although several previous studies have investigated on the fate of CBZ in water treatment, kinetics and reaction pathways of OXC, a keto homologue of CBZ, upon different water disinfection treatments were largely unknown. The present work provides substantial new details related to these aspects of the OXC reaction with chlorine, ozone and upon UV irradiation.  OXC reacted quickly with OH radical and relatively rapidly with free chlorine while slow reaction rates were recorded in presence of either ozone or upon UV irradiation.  Product analysis indicated that abundant and persistent TPs of OXC such as DiOH-CBZ, ACIN and BQD, were identical to those previously reported during water treatment of CBZ.  An increase in acute toxicity was only observed for irradiated solutions of OXC prepared in synthetic surface water probably due to the accumulation of ACIN.  Analysis of tap water samples revealed the occurrence at ng/L levels of several metabolites/transformation products of OXC and CBZ except ACIN.

Acknowledgments This work was funded by the “Agence Franc¸aise de Se´curite´ Sanitaire de l’environnement et du travail (AFSSET), Programme de recherche Environnement-Sante´-Travail 2007.

Appendix A. Supplementary material Supplementary information for this manuscript can be found in the online version at, doi:10.1016/j.watres.2010.11.038.

references

Acero, J., Benitez, F., Real, F., Roldan, G., 2010. Kinetics of aqueous chlorination of some pharmaceuticals and their elimination from water matrices. Water Res. 44, 4158e4170. APHA, 1998. Standard Methods for the Examination of Water and Wastewater. APHA, AWWA, WPCF, Washington, DC. Bader, H., Hoigne´, J., 1981. Determination of ozone in water by the indigo method. Water Res. 15, 449e456. Baker, K.M., Csetenyl, J., Frigerio, A., Morselli, P.L., 1973. 10,11dihydro-10,11-dihydroxy-5H-dibenz[b, z]azepine-5carboxamide, a potential metabolite of carbamazepine isolated from human and rat urine. J. Med. Chem. 16, 703e708. Barron, E., Deborde, M., Rabouan, S., Mazellier, P., Legube, B., 2006. Kinetic and mechanistic investigations of progesterone reaction with ozone. Water Res. 2006 (40), 2181e2189. Bedner, M., MacCrehan, W.A., 2006. Transformation of acetaminophen by chlorination produces the toxicants 1,4benzoquinone and N-acetyl-p-benzoquinone imine. Environ. Sci. Technol. 40, 516e522. Benner, J., Ternes, T.A., 2009. Ozonation of propranolol: formation of oxidation products. Environ. Sci. Technol. 43, 5086e5093.

1596

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 8 7 e1 5 9 6

Benotti, M.J., Trenholm, R.H., Vanderford, B.J., Holady, J.C., Stanford, B.D., Snyder, S.A., 2009. Pharmaceuticals and endocrine disrupting compounds in US drinking water. Environ. Sci. Technol. 43, 597e603. Breton, H., Cociglio, M., Bressolle, F., Peyriere, H., Blayac, J.-P., Hillaire-Buys, D., 2005. Liquid chromatographyeelectrospray mass spectrometry determination of carbamazepine, oxcarbazepine and eight of their metabolites in human plasma. J. Chromatogr. B. 828, 80e90. Brose´us, R., Vincent, S., Aboulfadi, K., Daneshvar, A., Sauve´, S., Barbeau, B., Pre´vost, M., 2009. Ozone oxidation of pharmaceutical, endocrine disruptors and pesticides during drinking water treatment. Water Res. 43, 4707e4717. Buffle, M.O., von Gunten, U., 2006. Phenols and amine induced HO generation during the initial phase of natural water ozonation. Environ. Sci. Technol. 40, 3057e3063. Canonica, S., Meunier, L., von Gunten, U., 2008. Phototransformation of selected pharmaceuticals during UV treatment of drinking water. Water Res. 42, 121e128. Chiron, S., Minero, C., Vione, D., 2006. Photodegradation processes of the antiepileptic drug carbamazepine, relevant to estuarine waters. Environ. Sci. Technol. 40, 5977e5983. Deborde, M., von Gunten, U., 2008. Reactions of chlorine with inorganic and organic compounds during water treatmentdKinetics and mechanisms: a critical review. Water Res. 42, 13e51. Dodd, M., Huang, C.-H., 2007. Aqueous chlorination of the antibacterial agent trimethoprim. Reaction kinetics and pathways. Water Res. 41, 647e655. Feitosa-Felizzola, J., Temime, B., Chiron, S., 2007. Evaluating on line SPE LC/ion trap/MS for reliable quantification and confirmation of several classes of antibiotics in urban wastewaters. J. Chromatogr. A 1164, 95e104. Guthrie, J.P., Cossar, J., Lu, J., 1991. Dihydroxyacids from the chlorination of ketones: an unexpected process. Can. J. Chem. 69, 1904e1908. Hu, L., Martin, H.M., Arce-Bulted, O., Sugihara, M.N., Keating, K.A., Strathmann, T.J., 2009. Oxidation of carbamazepine by Mn(VII) and Fe(VI): reaction kinetics and mechanism. Environ. Sci. Technol. 43, 509e515. Huber, M.M., Canonica, S., Park, G.Y., von Gunten, U., 2003. Oxidation of pharmaceuticals during ozonation and advanced oxidation processes. Environ. Sci. Technol. 37, 1016e1024. Kosjek, T., Andersen, H., Kompare, B., Ledin, A., Heath, E., 2009. Fate of carbamazepine during water treatment. Environ. Sci. Technol. 43, 6256e6261. Leclercq, M., Mathieu, O., Gomez, E., Casellas, C., Fenet, H., Hillaire-Buys, D., 2009. Presence and fate of carbamazepine, oxcarbazepine and seven of their metabolites at wastewater treatment plants. Arch. Environ. Contam. Toxicol. 56, 408e415. Lee, Y., von Gunten, U., 2010. Oxidative transformation of micropollutants during municipal wastewater treatment: comparison of kinetic aspects of selective (chlorine, chlorine dioxide, ferrateVI, and ozone) and non selective oxidants (hydroxyl radicals). Water Res. 44, 555e566.

Liu, Q., Scurter, L.M., Muller, C.E., Aloisio, S., Francisco, J.S., Margerum, D.W., 2001. Kinetics and mechanisms of aqueous ozone reactions with bromide, sulfite, hydrogen sulfite, iodide and nitrite ions. Inorg. Chem. 40, 4436e4442. Lu, L., Martin, H.M., Arce-Bulted, O., Sugihara, M., Keating, K., Strathmann, T., 2009. Oxidation of carbamazepine by Mn(VII) and Fe(VI): reaction kinetics and mechanism. Environ. Sci. Technol. 43, 509e515. McDowell, D., Huber, M., Wagner, M., von Gunten, U., Ternes, T., 2005. Ozonation of carbamazepine in drinking water: identification and kinetic study of major transformation products. Environ. Sci. Technol. 39, 8014e8022. Miao, X.-S., Metcalfe, C.D., 2003. Determination of carbamazepine and its metabolites in aqueous samples using liquidchromatography-electrospray tandem mass spectrometry. Anal. Chem. 75, 3731e3738. Monpelat, S., Le Bot, B., Thomas, O., 2009. Occurrence and fate of pharmaceutical products and by-products, from resource to drinking water. Environ. Intern. 35, 803e814. Parkhurst, B., Bradshaw, A., Forte, J., Wright, G., 1981. The chronic toxicity to Daphnia magna of acridine, a representative azaarene present in synthetic fossil fuel products and wastewaters. Environ. Pollut. Ser. A 24, 21e30. Pereira, V.J., Weinberg, H.S., Linden, K.G., Singer, P.C., 2007. UV degradation of pharmaceutical compounds in surface water via direct and indirect photolysis at 254 nm. Environ. Sci. Technol. 41, 1682e1688. Pinkernell, U., Nowack, B., Gallard, H., von Gunten, U., 2000. Methods for the photometric determination of reactive bromine and chlorine species with ABTS. Water Res. 34, 4343e4350. Quintana, J.B., Rodil, R., Lopez-Mahia, P., Muniategui-Lorenzo, S., Prada-Rodriguez, D., 2010. Investigating the chlorination of acidic pharmaceuticals and by-product formation aided by an experimental design methodology. Water Res. 44, 243e255. Radjenovic, J., Godehardt, M., Petrovic, M., Hein, A., Farre´, M., Jekel, M., Barcelo, D., 2009. Evidencing generation of persistent ozonation products of antibiotics roxithromycin and trimethoprim. Environ. Sci. Technol. 43, 6808e6815. Stackelberg, P.E., Gibs, J., Furlong, E.T., Meyer, M.T., Zaugg, S.D., Lippincott, R.L., 2007. Efficiency of conventional drinking water treatment processes in removal of pharmaceuticals and other organic compounds. Sci. Total Environ. 377, 255e272. Vieno, N., Ha¨rkki, H., Tuhkanen, T., Kronberg, L., 2007. Occurrence of pharmaceuticals in river water and their elimination in a pilot-scale drinking water treatment plant. Environ. Sci. Technol. 41, 5077e5084. Vogna, D., Marotta, R., Andreozzi, R., Napolitano, A., D’Ischia, M., 2004. Kinetic and chemical assessment of the UV/H2O2 treatment of antiepileptic drug carbamazepine. Chemosphere 54, 497e505. West, E., Rosario-Ortiz, F.L., Snyder, S., 2009. Effect of ozone exposure on the oxidation of trace organic contaminants in wastewater. Water Res. 43, 1005e1014. Westerhoff, P., Yoon, Y., Snyder, S., West, E., 2005. Fate of endocrine disrupter, pharmaceutical, and personal care product chemicals during simulated drinking water treatment processes. Environ. Sci. Technol. 39, 6649e6663.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 9 7 e1 6 0 6

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Temperature phased anaerobic digestion increases apparent hydrolysis rate for waste activated sludge Huoqing Ge, Paul D. Jensen, Damien J. Batstone* Advanced Water Management Centre (AWMC), Environmental Biotechnology CRC, The University of Queensland, St Lucia, QLD 4072, Australia

article info

abstract

Article history:

It is well established that waste activated sludge with an extended sludge age is inherently

Received 28 September 2010

slow to degrade with a low extent of degradation. Pre-treatment methods can be used prior

Received in revised form

to anaerobic digestion to improve the efficiency of activated sludge digestion. Among these

26 November 2010

pre-treatment methods, temperature phased anaerobic digestion (TPAD) is one promising

Accepted 28 November 2010

method with a relatively low energy input and capital cost. In this study, an experimental

Available online 4 December 2010

thermophilic (50e70




C)emesophilic system was compared against a control meso-

philicemesophilic system. The thermophilicemesophilic system achieved 41% and 48% Keywords:

volatile solids (VS) destruction during pre-treatment of 60
C and 65
C (or 70
C) respec-

Temperature phased anaerobic

tively, compared to 37% in the mesophilicemesophilic TPAD system. Solubilisation in the

digestion

first stage was enhanced during thermophilic pre-treatment (15% at 50
C and 27% at 60
C,

Thermophilic pre-treatment

65
C and 70
C) over mesophilic pre-treatment (7%) according to a COD balance. This was

Mesophilic pre-treatment

supported by ammoniaenitrogen measurements. Model based analysis indicated that the

Waste activated sludge

mechanism for increased performance was due to an increase in hydrolysis coefficient

Hydrolysis rate

under thermophilic pre-treatment of 60
C (0.5  0.1 d1), 65
C (0.7  0.2 d1) and 70
C (0.8  0.2 d1) over mesophilic pre-treatment (0.2  0.1 d1), and thermophilic pre-treatment at 50
C (0.12  0.06 d1). ª 2010 Elsevier Ltd. All rights reserved.

1.

Introduction

Anaerobic digestion is a biological decomposition process used to treat, stabilise, and reduce the quantities of organic wastes prior to disposal or beneficial re-use. Anaerobic digestion has been used extensively in municipal wastewater treatment to stabilise primary sludge and activated sludge. Over the past decade, municipal wastewater treatment processes have adapted to meet reduced discharge limits on the effluent nitrogen concentration. Process adaptations include the removal of primary settlings and primary sludge streams; and increased retention times for biological nutrient removal (BNR) processes, resulting in increased sludge age. Increased sludge age results in waste activated sludge with

inherently low degradability, as inert materials in the influent, as well inert decay products accumulate in the activated sludge (Gossett and Belser, 1982). Adaptations of modern wastewater treatment processes have introduced new challenges for anaerobic digestion, as poor degradability of activated sludge requires long digester retention times, higher mixing costs, and also results in poor gas production. Incorporating a pre-treatment into anaerobic treatment may enhance the sludge digestion by accelerating hydrolysis, which is generally accepted as the rate-limiting step in anaerobic digestion. Pre-treatment can enhance overall digestion, and requires a minimal capital investment in comparison with methods such as aerobic digestion (Ros and  ic , 2003). Temperature phased anaerobic digestion Zupanc

* Corresponding author. Tel.: þ61 7 3346 9051; fax: þ61 7 3365 4726. E-mail address: [email protected] (D.J. Batstone). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.11.042

1598

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 9 7 e1 6 0 6

(TPAD), combines a short (1e3 days) thermophilic pre-treatment stage (50e70
C) applied prior to a conventional mesophilic anaerobic digestion (35
C, 10e20 days). TPAD is highly scalable as the process incorporates standard digestion vessels and low quality heat is the main energy input. Thermophilicemesophilic TPAD has been shown to be an effective treatment for increasing methane production and volatile solids (VS) destruction, compared with a single-stage mesophilic digestion. Han and Dague (1997) reported 39% VS destruction of primary sludge achieved in a TPAD system (55
C, 3 days hydraulic retention time (HRT) and 35
C), which was higher than 32% in a single-stage control system (35
C). A corresponding methane production in the TPAD system was also 16% higher over the control system. An improvement of methane production with a TPAD system treating activated sludge was also observed by Bolzonella et al. (2007). They found the highest specific methane production from the TPAD (70
C, 2e3 days HRT and 37
C) was 370 ml gVS1 added, 30e50% higher than that from a single-stage control system (37
C). Extending pre-treatment HRT to 5 days did not improve methane production further. Nges and Liu (2009) evaluated the effect of pre-treatment temperature, and reported that the digestion performance of mixed primary and activated sludge was not influenced by thermophilic temperature, as the same VS destruction of 42% was achieved at pre-treatment temperatures of both 50 and 70
C (2 days HRT). However, this result was still greater than the VS destruction achieved in the single-stage mesophilic control (39%). Watts et al. (2006) reported a substantial improvement of activated sludge digestion with increased thermophilic temperature. They found that a TPAD system (47
C, 2 days HRT and 37
C) achieved the similar VS destruction of 24% as a single-stage mesophilic digester (37
C). VS destruction was not improved with the thermophilic temperature increased to 54
C, but was significantly enhanced to 34% at 60
C. The majority of research has focused on achieving improved performance by varying pre-treatment conditions during TPAD, with performance comparisons against singlestage thermophilic or mesophilic anaerobic digestion. However, there is little analysis to determine the nature of the pre-treatment process; whether it improves the rate or extent of subsequent sludge degradation, or both properties. As a result, optimal pre-treatment conditions (temperature, pH and HRT) have not been established. Additionally, thermophilicemesophilic TPAD rarely has been evaluated in a parallel comparison of mesophilicemesophilic with the same retention times. This study is based on our previous investigation of primary sludge (Ge et al., 2010), and further investigates pre-treatment mechanisms of TPAD on waste activated sludge, with a direct comparison against a control mesophilicemesophilic process.

2.

Materials and methods

2.1.

Substrate

Substrate was waste activated sludge, collected from a biological nutrient removal (BNR) process with 10 days sludge age

and water temperature of approximately 20
C in the Elanora wastewater treatment plant, located at Gold Coast, Australia. The feed was prepared monthly by centrifuging the sludge to a total solids (TS) concentration of 2e3%, and subsequently stored below 4
C. Regular analysis was performed to determine the characteristics and consistency of the feed material. Table 1 shows the average characteristics of the activated sludge feed based on 14 feed collections over 15 months.

2.2.

Start-up and operation

Two identical two-stage systems were used throughout. These consisted of thermophilic pre-treatment (TP) and mesophilic pre-treatment (MP) pre-treatment stages (0.6 L, 2 days HRT), and mesophilic methanogenic stages (4.2 L, 14 days HRT), as shown in Fig. 1. The basic set-up and operation of thermophilic (TP1)emesophilic (TP2) TPAD and mesophilic (MP1)emesophilic (MP2) TPAD systems were described in Ge et al. (2010). Approximately 0.3 L per day of substrate was fed simultaneously by pumping 0.05 L through the pre-treatment stage and methanogenic stage at intervals of 4 h per day (6 times daily, also weighed daily). Gas production was measured daily from each reactor using tipping bucket gas meters, and continuously logged. Reactor pH was also recorded online from each reactor continuously. Each reactor was inoculated using methanogenic inoculum from the methanogenic second stage (35  1
C, 14 days HRT) of a lab-scale thermophilicemesophilic TPAD system (Ge et al., 2010). This provided a diverse microbial community and a common starting point for each reactor. The systems were operated in parallel for over 15 months. During this time the temperature of TP1 was altered to create different operating periods:  Period 1: 50
C (186 days), including two periods of pH 5 by dosing 1 M HCL (Day 39e48, and Day 55e77)  Period 2: 60
C (100 days)  Period 3: 65
C (67 days), including a period of HRT reduced to 12 days in TP2 (Day 330e356)  Period 4: 70
C (68 days). The temperature of TP2, MP1 and MP2 was held constant at 35
C during all periods. During periods where the pH of TP1 was reduced, the pH of MP1 was also reduced, and periods

Table 1 e Characteristics of the waste activated sludge used in this study. Measure 1

TS (g L ) VS (g L1) pH COD (g L1) VFA (g COD L1) TKN (g N L1) 1 NHþ 4 eN (g L )

Activated sludge 25.4  0.1 17.5  0.1 6.5e7.5 27.4  3.5 0.2  0.1 1.9  0.5 0.06  0.04

Error margins indicate standard deviation across 14 different feed collections used in the study over 15 months.

1599

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 9 7 e1 6 0 6

Gas to exhaust

Gas meter F Gas meter PLC

F

Water jacket temperature control

Heating coil

Feed reservior

Effluent drum Pretreatment 0.6L TP1 = 50-70°C MP1 = 35°C

Feed pump

Main Digester 4L TP2=35°C MP2=35°C

Effluent pump

Digester pump

Fig. 1 e Schematic diagram of thermophilic pre-treatment TPAD system and mesophilic pre-treatment TPAD system.

where the loading of TP2 was increased, the loading of MP2 was also increased, as a result of HRT shortened to 12 days. This was done in steps of 20% around the average in an attempt to provide better model-parameter identifiability. After each acid dosing period, the pH of TP1 and MP1 returned to their natural levels of 6.6 and 6.8, respectively.

2.3.

Chemical analysis

Gas production and composition (H2, CH4, CO2) were analysed by GCeTCD as described previously (Tait et al., 2009). Liquid samples were collected from each reactor three times per week. Analysis was performed for TS, VS, volatile fatty acid (VFA), chemical oxygen demand (COD), total Kjeldahl nitrogen Analytical (TKN) and ammoniumenitrogen (NHþ 4 eN). methods were based on Standard Methods (APHA, 1998). The preparation and measurement of VFA, soluble COD (COD(S)) and NHþ 4 eN were as described previously (Ge et al., 2010).

VS destruction% ¼

VSconc;in  VSconc;out  100 VSconc;in

where VSconc,in ¼ VS concentration of inlet; VSconc,out ¼ VS concentrations of outlet. Results of the mass balance calculation are sensitive to systematic sampling issues, which may cause dilution, while the results of the Van Kleeck calculation are influenced by accumulation of mineral inerts within the reactor (under nonsteady state conditions).

2.4.2.

Extent of solubilisation

Extent of sludge solubilisation in each pre-treatment stage was calculated using the ratio of total solubilised products (methane production and COD(S)) and particulate COD concentration in the inlet feed (Song et al., 2005). Hydrogen was not detected in either pre-treatment stage. Extent of solubilisation can be expressed as Extent of solubilisation% ¼

2.4.

Calculation

2.4.1.

VS destruction

VSfrac;in  VSfrac;out
VSfrac;in  VSfrac;in  VSfrac;out

CODCH4 þ CODðSÞo CODðSÞi  100 CODðTÞi CODðSÞi (3)

The two calculation methods used to determine VS destruction were the Van Kleeck equation and the mass balance equation. The Van Kleeck equation (1) assumes the amount of mineral solids is conserved during digestion (Switzenbaum et al., 2003), and uses the volatile fractions (VS/TS  VSfrac) in the inlet and outlet as references. VS destruction% ¼

(2)

(1)

where VSfrac,in ¼ volatile fraction (VS/TS) in the inlet solids; VSfrac,out ¼ volatile fraction (VS/TS) in the outlet solids. The mass balance equation (2) uses VS concentrations (VSconc) in the inlet and outlet, expressed as

where CODCH4 ¼ methane production as mg COD during pretreatment; COD(S)i ¼ COD(S) concentration of inlet; COD (S)o ¼ COD(S) concentration of outlet; COD(T)i ¼ total COD concentration of inlet.

2.5.

Mathematical analysis

Mathematical analysis was based on the IWA Anaerobic Digestion Model No. 1 (ADM1) (Batstone et al., 2002). Implementation of ADM1 for a TPAD process is described by Ge et al. (2010), with the input model of Nopens et al. (2009). Initial conditions were adjusted based on measurements of organic solids, organic acids, ammonia, TKN, etc. There were approx 420 input changes over 450 days used in the model. Degradability

1600

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 9 7 e1 6 0 6

extent (fd) and apparent first order hydrolysis rate coefficient (khyd) were the main parameters used to assess and compare two TPAD systems. In each system, khyd and fd were simultaneously estimated to achieve the average optimal values for the whole TPAD process, which were then set to determine the twoparameter uncertainty surface for khyd and fd based on the method of Batstone et al. (2003, 2009). In the TP system, confidence regions were estimated based on system performance at each pre-treatment temperature. For the MP system, operating conditions were constant throughout the experiment; therefore only one confidence region was estimated for comparison. A 95% confidence limit was used, with appropriate F-value (2.996) for 2 parameters and the number of degrees of freedom. Van Kleeck VS destruction was used as a measured variable, with sum of squared errors (c2) as an objective function. Mass balance VS destruction gave similar optimal parameter estimates. However, confidence regions were enlarged, and the upper limit of the region could not generally be determined. Therefore the analysis is based on the Van Kleeck data and mass balance based regions are not shown.

3.

Results

3.1.

Performance of combined TPAD systems

Fig. 2 shows VS destruction in each system calculated by mass balance (Fig. 2A) and Van Kleeck (Fig. 2B) equations. During all periods, VS destruction determined using the mass balance equation (2), was consistent with VS destruction determined using the Van Kleeck equation (1), and this applied to both TPAD systems. Consistent results from the VS calculation methods confirm systematic sampling errors and/or unexpected behaviours were minimal. During Period 1 (50
C pre-treatment), thermophilic pretreatment offered no advantage over mesophilic pre-treatment. Increasing thermophilic pre-treatment temperature to 60
C (Period 2) improved VS destruction in the TP system from 34  1% to 41  1%. A further increase of VS destruction to 48  2% was observed when thermophilic pre-treatment temperature was increased to 65
C (Period 3), but no further enhancement at 70
C (Period 4). Statistical analysis (student’s t-test, a ¼ 0.05) confirmed that VS destruction in the TP system (65
C) was significantly greater than that achieved at 60
C, which was also a significant improvement over that achieved at 50
C. VS destruction in the TP system during Periods 2e4 was also significantly better as compared to the MP system (student’s t-test, a ¼ 0.05). Additionally, pre-treatment pH was temporarily lowered to pH 5 twice during 50
C pre-treatment (Day 39e48 and Day 55e77), which did not influence VS destruction in either system compared to previously. Similarly, VS destruction in each system was maintained as previously described when HRT of methanogenic stages was shortened from 14 to 12 days during Period 3 (Day 330e356). Total methane produced from TP system was consistently higher than that from MP system except Period 1, as shown in Fig. 3 and Table 2. This was confirmed by student’s t-test (a ¼ 0.05) as a statistically significant improvement, and was consistent with enhanced VS destruction in TP system over

MP system during Periods 2e4. The methane production increase in TP system was observed when thermophilic pretreatment temperature was increased to 60
C, but not at 65
C and 70
C. This was not consistent with further enhancement of VS destruction at thermophilic pre-treatment of 65
C compared to 60
C. This is likely due to three reasons: (a) As temperature increased, different portions of VS were degraded by the microbial community resulting in different gas yields per VS destroyed, (b) the decrease in stage 1 methane production (and increase in stage 2 methane production) at higher temperatures is complicating analysis, and (c) it is possible that gas leaks (approx. 10% losses overall) were occurring despite our best efforts. Thus, gas flows have not been used in the detailed model based analysis below, but have been used in solubilisation analysis, and quantitatively compared against model outputs in Section 4.3. During Periods 1e2, approximately 50e70% of methane generated from the TP system was produced in the thermophilic pre-treatment stage, with small amounts from the subsequent mesophilic digestion stage. Increasing temperature of TP1 to 65
C and 70
C caused a substantial decrease in methane production in the first stage. A correspondingly larger amount of methane was produced from TP2, especially during Period 4. In contrary, in MP system, the methane production from MP1 was less compared to that from MP2 in all periods. Moreover, acid dosing was used to lower pH to 5 in TP1 at 50
C, in order to reduce the activity of methanogens and cause washout. As expected, methane production was severely decreased in TP1, producing a corresponding increase in production from TP2. Once the acid dosing stopped, methane production rapidly returned to previous levels.

3.2.

Model based analysis

Fig. 4 shows the 95% confidence regions for degradability (fd) (x-axis) and apparent hydrolysis rate (khyd) ( y-axis) in MP system and TP system at 50
C, 60
C, 65
C and 70
C, respectively. The MP system confidence regions overall at 35
C and TP at 50
C overlapped, indicating statistically the same apparent properties. khyd in the TP system increased significantly above both MP and TP at 50
C as temperature in TP1 was increased to 60
C, but did not increase further to 70
C. The confidence region moved to the right and upward from 60
C to 65
C, indicating a slight improvement in properties, but with full overlap between the regions at 65
C and 70
C. fd in the TP system were comparable for all tested temperatures in TP1, and statistically overlapped with the fd observed in MP. Overall, the results indicate consistent increases in hydrolysis coefficient with increased temperature, but with a relatively constant degradability of 30e60% (Table 3).

3.3.

Analysis of pre-treatment stage in TPAD systems

Extent of solubilisation of the activated sludge determined using equation (3) is shown in Fig. 5. The solubilisation in TP1 was increased from 15% to 27% with the thermophilic temperature increased from 50
C to 60
C, and was not improved further at 65
C and 70
C (Fig. 5). The main profile of solubilisation at 50
C and 60
C was methane production,

1601

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 9 7 e1 6 0 6

70 60

HRT 12 days

% VS destruction in TP system % VS destruction in TP system

pH 5

pH 5

A

50 40 30

VS destruction (%)

20 10 0

% VS destruction in TP system % VS destruction in MP system

60

B

50 40 30 20 10 Period 1

0 0

50

100

150

200

250

Period 4

Period 3

Period 2

300

350

400

450

Time in operation (days) Fig. 2 e VS destruction calculated by mass balance equation (A) and Van Kleeck equation (B) during each period in the thermophilic pre-treatment (TP) and mesophilic pre-treatment (MP) systems (% VS destruction is based on the activated sludge feed characteristics).

with a relatively lower level of VFA and other soluble products presented by COD(S). However, the majority of solubilisation profile was changed to the COD(S) with increasing thermophilic temperature to 65
C and 70
C, as the methane production dropped at higher thermophilic temperatures. For all periods, solubilisation in TP1 was higher than that in MP1, and did not appear to be affected by low pH (pH 5) at 50
C (Day 39e48 and Day 55e77). The VFA profiles were similar in TP1 and MP1 during all periods with acetate as the primary VFA produced, followed by propionate as the second major acid. Other VFAs (iso-butyrate, butyrate, iso-valerate, valerate and hexanoate) were also detected at much lower levels. The acetate concentration in TP1 was lowest during thermophilic pre-treatment of 50
C and 60
C, possibly due to the combination of poor solubilisation

and good methane production (Fig. 6). Increasing the thermophilic pre-treatment temperature to 65
C and 70
C resulted in substantial increases in acetate and propionate concentrations, which was consistent with measurements of COD(S), suggesting that most material hydrolysed was converted to organic acids. It also indicated the methanogenesis was limited at the higher thermophilic temperatures. Methane production was also suppressed under acidic conditions and resulted in increased accumulation of VFA in TP1, as hydrolysis and fermentation processes continued to producing intermediate products (VFAs). Nitrogenous organic compounds contained in the sludge (e.g. protein) is solubilised in the form of NHþ 4 eN during pretreatment process, thus NHþ 4 eN is another important indicator of solubilisation. Solubilisation according to NHþ 4 eN in TP1

1602

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 9 7 e1 6 0 6

1.2

1.0

CH4 production in MP system

-1

CH4 production (L day )

CH4 production in MP1

CH4 production in TP1 CH4 production in TP system

0.8

0.6

0.4

0.2

0.0

Period 1

Period 3

Period 2

Period 4

Fig. 3 e Average methane production during each period in the thermophilic pre-treatment (TP) and mesophilic pretreatment (MP) systems (Error bars are 95% confidence in mean methane production).

4.1.

Overall performance of TPAD systems

An increase of pre-treatment temperature from 50
C to 65
C substantially improved the overall VS destruction from 34% to 48% for activated sludge, but not for primary sludge (54% at 50e65
C) (Ge et al., 2010). It should be noted that performance on primary sludge is not diminished above 65
C, and thus mixed feed digesters should be operated at elevated temperatures. This not only decreases sludge disposal costs substantially, it also allows for much better hygenisation and pathogen removal. In addition, it provides performance of VS destruction

1.2 Degrades faster

Discussion

TP 70 C

1.0

0.8 TP 60 C

-1

4.

above 38%, which is one of the legislative levels for performance implemented by US EPA (EPA, 1994), and most Australian legislation. At higher performance levels (Period 3), VS destruction was not strongly impacted by a change from 14 to 12 days, consistent with a hydrolysis coefficient of >0.5 d1. Correspondingly, the implementation of smaller digesters or increased organic loading rates could be possible, which could substantially reduce capital costs. Another advantage of the TPAD system is the use of renewable methane to produce energy, which is used to compensate heat requirements (heat demand and heat losses) in the TPAD system, and is conventionally produced by cogeneration engines. The main heat demand from the TPAD system is to heat up the sludge to the required temperature in

khyd (d )

increased with each step increasing thermophilic pre-treatment temperature (Fig. 7). Release of NHþ 4 eN was greater from TP1 than from MP1 for all periods, which was consistent with the extent of solubilisation results. The NHþ 4 eN results indicated protein fermentation was improved under thermophilic condition, and improved with thermophilic temperature increase. Again, low pH did not have an impact on NHþ 4 eN release in TP1.

0.6 TP 65 C MP 35 C

0.4

Table 2 e A summary of methane production (L gVSfedL1 fed) in the thermophilic pre-treatment (TP) system and mesophilic pre-treatment (MP) system during each period. TP system Period Period Period Period

1 2 3 4

(50
C) (60
C) (65
C) (70
C)

0.10 0.16 0.15 0.17

   

0.03 0.02 0.02 0.03

MP system 0.07 0.11 0.10 0.09

   

0.04 0.03 0.03 0.04

Error margins indicate standard deviation across different gas measurements over each period.

TP 50 C 0.2

0.0 0.2

0.3

0.4

0.5

fd

0.6 Degrades more

Fig. 4 e 95% confidence regions for apparent hydrolysis coefficient (khyd, dL1) and degradability (fd) using Van Kleeck VS destruction as an objective function in the mesophilic pre-treatment (MP) system and thermophilic pre-treatment (TP) system at 50, 60, 65 and 70
C, respectively.

0.7

1603

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 9 7 e1 6 0 6

Table 3 e A summary of apparent hydrolysis coefficient (khyd) and degradability (fd) in the mesophilic pre-treatment (MP) system and thermophilic pre-treatment (TP) system. Parameter

TP system


1

khyd (d ) fd




MP system





50 C

60 C

65 C

70 C

0.12  0.06 0.4  0.1

0.5  0.1 0.41  0.04

0.7  0.2 0.51  0.04

0.8  0.2 0.53  0.02

0.2  0.1 0.4  0.1

Numbers after ‘’ are the linear, uncorrelated 95% confidence in parameter values.

the thermophilic stage, as well as to the mesophilic temperature in the second stage. Heat from thermophilic stage will also be used in the mesophilic stage, and minimises overall heating requirements in a TPAD process. A detailed evaluation of the heat balance, including losses and sensitivity to feed concentration is contained in the supplementary information. This analysis also considers the increased performance provided by thermophilic operation, but does not consider a smaller main digester. The heat balance for thermophilic and mesophilic systems is generally positive at a 2% feed concentration, with either supplemental heat needed, or diversion of methane from electricity to heat energy, as shown in Fig. i (Supplementary information). The heat balance became negative when increasing the feed concentration to 4% in both systems, indicating the potential heat production could fully offset heat requirements in both systems. At the feed concentration of 6%, the potential heat production was greatly in excess, especially in the TP system with thermophilic pre-treatment of 65
C and 70
C. This emphasises the need for pre-thickening, but importantly, indicates that the heat balance is very similar for standard mesophilic and TPAD systems, with TPAD processes operated above 60
C generally producing more excess energy than a mesophilic process. It should also be noted that we assume only waste heat is used, from cogeneration engines with electricity being produced as the main process.

The extent of solubilisation (%)

40

Methane VFAs Others

30

20

TP1

4.2.

Regions measurably moved upwards and to the right with increased temperature from 50
C to 75
C, reflecting the increase in performance. The regions also decreased in area as temperature increased, likely due to the increase in VS destruction. Because VS destruction is a product (or fractional) term, it can be determined with better accuracy at higher destruction levels. There are a wide range of hydrolysis rates reported in literature for mixed streams containing both primary sludge and activated sludge, between 0.1 and 1 d1 (Pfeffer, 1974; Ghosh, 1981; Bolzonella et al., 2007). However, the determination of hydrolysis rate for digestion of WAS only has been limited, and is now addressed in our study. Hydrolysis coefficient was sensitive to temperature, with no improvement at 50
C, but significant increases at 60
C and higher. This may be due to emergence of true thermophiles at the higher temperatures that do not emerge at the intermediate temperature of 50
C. However, this contrasts to our previous work on primary sludge, where hydrolysis rate was improved above 50
C, but not sensitive to temperature above this. Apparent degradability was not significantly impacted, indicating that the improved VS destruction observed in this study was due to increase in apparent hydrolysis rate, rather than an increase in degradable fraction. Analysis of degradability is consistent with our previous observations on primary sludge (Ge et al., 2010). This increase in rate rather than degradability fraction is similar to the effects of mechanical pre-treatments, e.g. sonication (Climent et al., 2007; Khanal et al., 2007). This is in contrast to high impact methods such as thermal hydrolysis, which increase rate and extent substantially (Batstone et al., 2009). In a full-scale plant, faster degradation could be utilised in either design of a smaller main digester, or intensification of an existing process.

4.3. 10

0

MP1

Period 1

Period 2

Period 3

Period 4

Fig. 5 e Extent of solubilisation during each period in the thermophilic pre-treatment stage (TP1) and the mesophilic pre-treatment stage (MP1) (% solubilisation is based on the activated sludge feed characteristics and equation (3)).

Model based analysis

Pre-treatment mechanisms

Solubilisation was enhanced in TP1 at 50
C over MP1 at 35
C (Fig. 5), even though hydrolysis coefficient remained the same (Fig. 4). Both measures were consistently enhanced at 60
C and above. The inconsistency at 50
C is likely because the apparent hydrolysis coefficient was acquired from the overall performance, and was therefore dominated by the second stage performance, whereas the information in Fig. 5 is based on the first stage (methane þ VFAs þ other products). This was further tested by simulating the first stage further, and comparing model outputs to observed results. The model simulation of

1604

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 9 7 e1 6 0 6

-1

tVFA concentration (mg L )

4000

tVFA in TP1 tVFA in MP1 pH 5

pH 5 3000

2000

1000

Period 2 (60 C)

Period 1 (50 C)

0 0

50

100

150

200

250

Period 3 (65 C) 300

350

Period 4 (70 C) 400

450

Time in operation (days) Fig. 6 e tVFA concentrations during each period in the thermophilic pre-treatment stage (TP1) and mesophilic pre-treatment stage (MP1).

valid overall), while hydrolysis coefficients estimated at increased pre-treatment temperatures were valid across both digesters. However, solubilisation results and model analysis both confirmed that the solubilisation or hydrolysis was improved in TP1 at 60e70
C compared to MP1. Therefore, improved first stage hydrolysis is a major factor contributing to enhanced performances in the TP system. The increased thermophilic temperature may improve production of extracellular enzymes to hydrolyse more complex or inert substrate materials, and have selected the specialised microbial community,

solubilisation followed the same trend of measured solubilisation in TP1, except at 50
C, where model predictions were conservative compared to solubilisation measurements. Improvements at 60
C to 65
C (or 70
C) were reflected in increased solubilisation according to both model and measurements. Additionally, fractions of solubilisation predicted by model and measurements in the TP1 were consistent, but the model could not predict the decrease in methane production at 65
C and 70
C, indicating a model limitation. This comparison suggests that the hydrolysis coefficient determined at 50
C was conservative for the first stage (but

1200 NH4+-N in TP1 NH4 -N in MP1

pH 5

pH 5

-1

NH4 -N concentration (mg L )

+

1000

800

600

+

400

200 Period 2

Period 1

0 0

50

100

150

200

250

Period 3 300

Period 4 350

400

450

Time in operation (days) Fig. 7 e NHD 4 -N concentrations during each period in the thermophilic pre-treatment stage (TP1) and mesophilic pretreatment stage (MP1).

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 9 7 e1 6 0 6

which will result in an optimized hydrolysis. All these possible improvements will increase the substrate availability for digestion in the subsequent methanogenic stage. NHþ 4 eN release in the pre-treatment stage was consistent with the extent of solubilisation data and had a significant influence on overall NHþ 4 eN release. It was exhibited as the 1 NHþ 4 eN released from TP2 (approx. 880 mg L ) and was 30% 1 higher than that from MP2 (approx. 680 mg L ). However, the enhanced NHþ 4 eN release in TP1 with increased thermophilic temperature did not influence the NHþ 4 eN release in methanogenic stage. This result was different from the similar NHþ 4 eN release observed in both methanogenic stages with individual thermophilic (50e65
C) and mesophilic pre-treatment (35
C) treating primary sludge (Ge et al., 2010). It indicated that overall conversion was improved further in the methanogenic stage.

4.4. Methanogenesis during thermophilic pre-treatment process There is some uncertainty over how best to operate TPAD systems. If the initial step is regarded purely as a hydrolytic step, methanogenesis is not required. This is not the case for our results, and no difference is seen where methanogenesis is inhibited by low pH or high temperature compared to where methanogenesis is allowed to occur. Particularly for activated sludge compared to primary sludge (Ge et al., 2010), the methane production in the pre-treatment stage is far higher, with approx 65% of total production occurring in the whole system. Methanogenesis in the pre-treatment stage is not detrimental, since it allows for increased overall methanogenic retention time, improved kinetics, and provides protection to the secondary main stage. The presence of substantial methanogenesis in this pre-treatment stage, with a very short retention time is of some interest. Methanogenesis at short HRT may be due to change in metabolic pathways from aceticlastis to acetate oxidation, under which acetate is first oxidised to hydrogen and carbon dioxide, and subsequently converted to methane. This is enhanced thermodynamically at higher temperatures 50
C to 65
C (Karakashev et al., 2006; Zinder et al., 1984), and is supported by microbial results indicating a dominance of Methanosarcinaceae, which has been found to be the dominant methanogen in acetate oxidising systems by isotopic carbon analyses (Karakashev et al., 2006). While methanogenesis in the pre-treatment stage complicates the process by the presence of two methane producing units, the action of a two-step acetate oxidation/ aceticlastic methanogenic process can provide advantages. This offers better resistance to process inhibition and toxins, since aceticlastis and acetate oxidisers are influenced by different factors. As an example, protein-rich cattle and piggery wastes, have high ammonia, to which acetate oxidisers are less susceptible than aceticlasts (Karakashev et al., 2005). In essence, the robust acetate oxidation step can act to protect the more sensitive aceticlastic methanogenic step, and provide a biological buffer. A decrease of 31% and 58% methane production in TP1 at 65
C and 70
C compared to 60
C suggests the activity of methanogenesis (presumptively acetate oxidation, or both) is decreased. Therefore, the level of methanogenesis in the pre-

1605

treatment stage can be tuned by temperature, especially at 60e70
C, without negative impact on overall VS destruction. Based on our results, it is also possible to suppress methanogenesis by decreasing the pH to 5, but doing so by changing temperature is lower cost, and easier implemented.

5.

Conclusions

The following conclusions can be drawn from this study:  VS destruction in thermophilicemesophilic TPAD was increased by thermophilic pre-treatment at 50
C to 65
C (34%e48%), which was 11e30% higher than that in mesophilicemesophilic TPAD (37%), expect thermophilic pretreatment of 50
C.  Model based analysis indicated the hydrolysis coefficient in the TP system was not improved under thermophilic pretreatment of 50
C (0.06e0.18 d1) compared to the MP system (0.1e0.3 d1), but significantly enhanced to 0.6 d1 at 60
C, up to 1 d1 at 65
C and 70
C. However, increasing thermophilic pre-treatment temperature had no impact on the overall degradability in the TP system relative to the MP system (0.30e0.55).  Solubilisation was improved during thermophilic pretreatment relative to mesophilic pre-treatment, and reached to maximum of 27% at thermophilic pre-treatment of 60
C. Further thermophilic temperature increases had no further impacts. Higher NHþ 4 eN was released during thermophilic pre-treatment over mesophilic pre-treatment, and further increased by increasing the thermophilic pre-treatment temperature from 50
C to 70
C.  A large amount of methane was produced from thermophilic pre-treatment stage between 50
C and 60
C, but started to decrease with further increase of temperature to 65
C and 70
C. Methane production from the pre-treatment stage was heavily inhibited at acidic conditions (pH 5).

Acknowledgements This work was funded by the Queensland State Government, under the Smart State Research-Industry Partnerships Program (RIPP), Meat and Livestock Australia, and Environmental Biotechnology Cooperative Research Centre (EBCRC), Australia as P23 “Small-medium scale organic solids stabilization”. Huoqing Ge and Paul Jensen are recipients of an EBCRC postgraduate scholarship and postdoctoral award, respectively. We thank Gold Coast City Council (Gold coast water) for supplying samples from their Elanora Wastewater Treatment Plant.

Appendix. Supplementary information Supplementary data related to this article can be found online at doi:10.1016/j.watres.2010.11.042.

1606

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 9 7 e1 6 0 6

references

APHA, 1998. Standard Methods for the Examination of Water and Wastewater, 20th ed. American Public Health Association, Washington, DC, USA. Batstone, D.J., Keller, J., Angelidaki, I., Kalyuzhnyi, S., Pavlostathis, S.G., Rozzi, A., Sanders, W.T.M., Siegrist, H., Vavilin, V., 2002. IWA Task Group for Mathematical Modelling of Anaerobic Digestion Processes. IWA Publishing, London, UK. Batstone, D.J., Pind, P.F., Angelidaki, I., 2003. Kinetics of thermophilic, anaerobic oxidation of straight and branched chain butyrate and valerate. Biotechnology and Bioengineering 84 (2), 195e204. Batstone, D.J., Tait, S., Starrenburg, D., 2009. Estimation of hydrolysis parameters in full-scale anaerobic digesters. Biotechnology and Bioengineering 102 (5), 1513e1520. Bolzonella, D., Pavan, P., Zanette, M., Cecchi, F., 2007. Two-phase anaerobic digestion of waste activated sludge: effect of an extreme thermophilic prefermentation. Industrial and Engineering Chemistry Research 46 (21), 6650e6655. Climent, M., Ferrer, I., Baeza, M.D., Artola, A., Vazquez, F., Font, X., 2007. Effects of thermal and mechanical pretreatments of secondary sludge on biogas production under thermophilic conditions. Chemical Engineering Journal 133 (1e3), 335e342. Ge, H.Q., Jensen, P.D., Batstone, D.J., 2010. Pre-treatment mechanisms during thermophilic-mesophilic temperature phased anaerobic digestion of primary sludge. Water Research 44 (1), 123e130. Ghosh, S., 1981. Kinetics of acid-phase fermentation in anaerobic digestion. Biotechnology and Bioengineering 11, 301e313. Gossett, J.M., Belser, R.L., 1982. Anaerobic-digestion of waster activated sludge. Journal of the Environmental Engineering Division-Asce 108 (6), 1101e1120. Han, Y., Dague, R.R., 1997. Laboratory studies on the temperaturephased anaerobic digestion of domestic primary sludge. Water Environment Research 69 (6), 1139e1143. Karakashev, D., Batstone, D.J., Angelidaki, I., 2005. Influence of environmental conditions on methanogenic compositions in anaerobic biogas reactors. Applied and Environmental Microbiology 71 (1), 331e338.

Karakashev, D., Batstone, D.J., Trably, E., Angelidaki, I., 2006. Acetate oxidation is the dominant methanogenic pathway from acetate in the absence of Methanosaetaceae. Applied and Environmental Microbiology 72 (7), 5138e5141. Khanal, S.K., Grewell, D., Sung, S., Van Leeuwen, J., 2007. Ultrasound applications in wastewater sludge pretreatment: a review. Critical Reviews in Environmental Science and Technology 37 (4), 277e313. Nges, I.A., Liu, J., 2009. Effects of anaerobic pre-treatment on the degradation of dewatered-sewage sludge. Renewable Energy 34 (7), 1795e1800. Nopens, I., Batstone, D.J., Copp, J.B., Jeppsson, U., Volcke, E., Alex, J., Vanrolleghem, P.A., 2009. An ASM/ADM model interface for dynamic plant-wide simulation. Water Research 43 (7), 1913e1923. Pfeffer, J.T., 1974. Temperature effects on anaerobic fermentation of domestic refuse. Biotechnology and Bioengineering 16 (6), 771e787.  ic , G.D., 2003. Thermophilic anaerobic digestion of Ro s, M., Zupanc waste activated sludge. Acta Chimica Slovenica 50, 359e374. Song, H., Clarke, W.P., Blackall, L.L., 2005. Concurrent microscopic observations and activity measurements of cellulose hydrolyzing and methanogenic populations during the batch anaerobic digestion of crystalline cellulose. Biotechnology and Bioengineering 91 (3), 369e378. Switzenbaum, M.S., Farrell, J.B., Pincince, A.B., 2003. Relationship between the Van Kleeck and mass-balance calculation of volatile solids loss. Water Environment Research 75 (6), 572. Tait, S., Tamis, J., Edgerton, B., Batstone, D.J., 2009. Anaerobic digestion of spent bedding from deep litter piggery housing. Bioresource Technology 100 (7), 2210e2218. US Environmental Protection Agency, 1994. A plain English Guide to the EPA Part 503 Biosolids Rule. In: Environmental protection Agency 8322/R-93-003, September Washington, DC, USA. Watts, S., Hamilton, G., Keller, J., 2006. Two-stage thermophilicmesophilic anaerobic digestion of waste activated sludge from a biological nutrient removal plant. Water Science and Technology 53 (8), 149e157. Zinder, S.H., Anguish, T., Cardwell, S.C., 1984. Effects of temperature on methanogenesis in a thermophilic (58
C) anaerobic digester. Applied and Environmental Microbiology 47 (4), 808e813.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 0 7 e1 6 1 6

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Threshold concentrations of biomass and iron for pressure drop increase in spiral-wound membrane elements W.A.M. Hijnen*, E.R. Cornelissen, D. van der Kooij KWR Watercycle Research Institute, PO Box 1072, 3430 BB Nieuwegein, The Netherlands

article info

abstract

Article history:

In a model feed channel for spiral-wound membranes the quantitative relationship of

Received 27 October 2010

biomass and iron accumulation with pressure drop development was assessed. Biofouling

Received in revised form

was stimulated by the use of tap water enriched with acetate at a range of concentrations

26 November 2010

(1e1000 mg C l
1). Autopsies were performed to quantify biomass concentrations in the

Accepted 29 November 2010

fouled feed channel at a range of Normalized Pressure Drop increase values (NPDi). Active

Available online 7 December 2010

biomass was determined with adenosinetriphosphate (ATP) and the concentration of bacterial cells with Total Direct Cell count (TDC). Carbohydrates (CH) were measured to

Keywords:

include accumulated extracellular polymeric substances (EPS). The paired ATP and CH

Spiral-wound membranes

concentrations in the biofilm samples were significantly ( p < 0.001; R2 ¼ 0.62) correlated

Biofouling

and both parameters were also significantly correlated with NPDi ( p < 0.001). TDC was not

NPD

correlated with the pressure drop in this study. The threshold concentration for an NPDi of

Biomass

100% was 3.7 ng ATP cm
2 and for CH 8.1 mg CH cm
2. Both parameters are recommended

ATP

for diagnostic membrane autopsy studies. Iron concentrations of 100e400 mg m
2 accu-

Carbohydrates

mulated in the biofilm by adsorption were not correlated with the observed NPDi, thus

Fe

indicating a minor role of Fe particulates at these concentrations in fouling of spiral-wound

Threshold values

membrane. ª 2010 Elsevier Ltd. All rights reserved.

1.

Introduction

Microbial growth (‘biofouling’) in high pressure spiral-wound (SW) membranes for nanofiltration (NF) or reverse osmosis (RO) has been identified as a major cause of operational problems such as increased feed channel pressure drop (PD), decreased mass transfer coefficient (MTC) and product quality decline. First studies on biofouling date from the 1970s and 1980s of the twentieth century (Bailey and Jones, 1974; Potts et al., 1981; Ridgway et al., 1985) and a number of reviews on this issue have been published in the 1990s (Flemming, 1997; Flemming

et al., 1993a; Ridgway and Flemming, 1996) because of the increasing application of membrane processes in water treatment and desalination. Destructive membrane sampling (autopsies) has been used to analyze the composition and structure of accumulated biofilms in order to elucidate the fundamentals of the biofouling process in spiral-wound membranes. With different microscopic techniques membrane foulants have been detected and identified as bacterial matter (Ridgway and Flemming, 1996). Still there is a lack of information on the quantitative relationship between biomass concentrations and the resulting operational problems in

Abbreviations: AOC, assimilable organic carbon; ATP, Adenosinetriphosphate; CH, carbohydrates; EPS, extracellular polymeric substances; FS, feed spacer; HPC, heterotrophic colony plate count; IPC, ion chromatography; NF, nanofiltration; NPD, normalized pressure drop; MFS, membrane fouling simulator; MTC, mass transfer coefficient; PD, pressure drop; Rf, exponential fouling rate constant; RO, reversed osmosis; SW, spiral-wound; TDC, total direct cell count. * Corresponding author. E-mail address: [email protected] (W.A.M. Hijnen). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.11.047

1608

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 0 7 e1 6 1 6

NF/RO membranes. Such information is not only needed for diagnostic purposes and improvement of pretreatment but also for assessing the efficacy of cleaning procedures to control biofouling. Microbial analysis such as Heterotrophic Plate Counts (HPC) and Total Direct Cell counts (TDC), and physical and (bio) chemical analysis including total wet weight of deposits, adenosinetriphosphate (ATP), extracellular polymeric substances (EPS) and proteins have been used to measure the amount of biomass on SW membranes (Flemming and Schaule, 1988; Griebe and Flemming, 1998; Ridgway et al., 1983; Schaule et al., 1993; Vrouwenvelder et al., 1998; Vrouwenvelder et al., 2008). Only few studies have tried to establish quantitative relationships between these biomass parameters and the pressure drop or flux decline. Flemming et al. (1993b) observed an MTC decline of 25% at total cell coverage of the membrane surface of 5  107e2  108 cells cm
2 and suggested a ‘pain level’ of bacterial cells of 108 per cm2; no correlation with pressure drop was presented. This ‘pain level’ corresponds with the amount of bacterial cells observed in SW membranes with operational problems related to biofouling (Griebe and Flemming, 1998; Hijnen et al., 2009; Schaule et al., 1993; Vrouwenvelder et al., 1998). Measuring active bacterial biomass with ATP in cell cultures or biomass samples is attractive because the analytical method is rapid, cheap and simple to perform and has a low detection level; a concentration of 1 ng l
1 ATP can be detected without concentration techniques. The proportional relationship between ATP and TDC (Magic-Knezev and Van der Kooij, 2004; Vrouwenvelder et al., 2008) indicates that ATP is a potential parameter to quantify the amount of accumulated biomass. Furthermore, autopsy results from full-scale SW membrane installations showed that the increase of the Normalized Pressure Drop (NPD) was related to ATP concentrations (Vrouwenvelder et al., 2008). However, establishment of a causal relationship between ATP and NPD requires more defined conditions to exclude effects of other deposits (dead biomass, EPS and other organic or inorganic substances). Quantification of carbohydrates (CH) with the Dubois method (Dubois et al., 1956) in autopsy studies enables to estimate biomass concentrations based on EPS which consists of polysaccharides with a large water-retention capacity resulting in voluminous deposits. The Dubois method is commonly used in membrane autopsy studies (Gabelich et al., 2004; Griebe and Flemming, 1998; Ridgway et al., 1983) and correlated with flux decline (Fonseca et al., 2007). Biofouling rarely occurs without mineral deposition (Ridgway and Flemming, 1996) and Fe was identified as a predominating foulant in SW elements (Baker and Dudley, 1998), but a causal relationship between Fe and PD increase was not reported. Hence, evaluation of the use of ATP, TDC and CH as quantitative biomass parameters in diagnostic autopsies and cleaning studies as well as elucidation of the role of Fe in pressure drop problems requires biofouling studies under well defined conditions. In a recent laboratory study using a Membrane Fouling Simulator (MFS) (Vrouwenvelder et al., 2006) a quantitative relationship between acetate as a model substrate and the pressure drop increase was demonstrated (Hijnen et al., 2009). Samples of the biofouled membranes were available for autopsy studies. The objectives of the current study were:

(i) elucidation of the quantitative relationship between biomass parameters ATP, TDC and CH and the extent of the PD increase and (ii) determination of the threshold concentrations of these parameters for a 100% increase of the normalized pressure drop (NPD) and (iii) to investigate the role of iron as the major mineral in the water under the experimental conditions. Such information enables the selection of proper biomass parameter(s) in autopsies to assess the cause of PD in membrane elements.

2.

Materials and methods

2.1.

Biofouling of an NF membrane

The Membrane Fouling Simulator (MFS) loaded with sheets (7  30 cm) of a “virgin” nanofiltration membrane sheet (Trisep 4040-TS80-TSF) was supplied with non-chlorinated tap water after filtration (10 and 1 mm poly-propyleen cartridge filtration; Van Borselen Ltd.) to exclude accumulation of suspended solids and spiked with low amounts of acetate-C to initiate biofouling. These experiments have been described in detail (Hijnen et al., 2009). Briefly, the MFS is a small scale continuous flow model of an SW feed channel (0.8x4x22.cm) filled with the matching Trisep feed spacer (0.8  4  20 cm; front 2 cm without feed spacer FS) and operated at a constant feed water flow of 16 l h
1 (cross-flow velocity of 0.14 m s
1) at a constant pressure of 1 bar without permeation. Fig. 1 depicts the experimental set up. The rate of clogging of the feed channel was measured by monitoring the pressure drop normalized (NPD) to a moderate environmental temperature of 12.5  C in the feed channel (Hijnen et al., 2009). The extent of biofouling, given by the relative NPD increase (NPDi), is calculated from the final NPD (NPDf) and the initial NPD (NPDo) by %NPDi ¼

NPDf
NPDo $100% NPDo

(1)

The MFS units were supplied with pre-filtered tap water spiked with acetate-C at concentrations (Sac) of 1, 3, 5, 10, 25, 100, 500 and 1000 mg l
1. Four blank MFS units with no acetate supply were operated with either filtered tap water or unfiltered tap water.

2.2.

Feed water quality

The feed water was non-chlorinated tap water produced from anaerobic groundwater using aeration and rapid sand filtration. The pH of the water was 7.98  0.05, dissolved organic carbon content was 2.0  0.1 mg C l
1, assimilable organic carbon 3
concentration (AOC) was 3e5 mg acetate-C eq l
1, NO
3 and PO4
1 content was 0.12  0.04 and 0.02  0.02 mg l respectively. The iron content (ion chromatography; ICP method with a lower detection limit of 0.005 mg l
1) of the filtered tap water was 0.008  0.014 mg l
1 and 0.32  0.24 mg l
1 in the unfiltered tap water. Iron was the major mineral in the tap water and visually (brown deposits) accumulated in the biofilms. The ambient water temperature was daily monitored during the experiments and ranged from 13.5 to 16.8  C (average of 15.9  0.7  C) and was 19.4  2.0  C in one experiment.

1609

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 0 7 e1 6 1 6

PI

Tap water from ground water filtered by Cartridge filters 10 and 1µm

PI

Nutrient dosage

FI

FI

QC

PR

MFS unit

a

QC

Sensitive pressure drop registration dP

b QC

Chlorine dosage

QC

FC

FC

Fig. 1 e Experimental set up for the experiments with a total of five MFS units. Unit on the left with dosage equipment and unit on the right with the mobile pressure drop monitor; PI and FI [ pressure and flow indicator; FC [ flow controller; QuC [ quick connector.

2.3.

Autopsy of the membrane sheets

Two samples (1.5  2 cm) were cut from the membrane at the inlet section of the feed channel without spacer (no FS) and five samples (2  2 cm) were cut from the membrane with spacer. The samples were transferred to sterile glass tubes with 20 ml of autoclaved tap water and sonicated with High Energy Sonication (HES, #100) using a BRANSON Digital Sonifier (Model 250 D) at optimized conditions established in a previously published study (Magic-Knezev and Van der Kooij, 2004). The sonifier tip (size 6.5 mm) was inserted into the tube (1e2 cm) containing 20 ml of autoclaved water and the membrane/spacer sample. This tube was placed in melting ice and sonicated for one minute at an amplitude of 45% (15e20 watt) to separate the biomass from the membrane/spacer samples. This treatment was repeated in 20 ml fresh sterile tap water and both suspensions were mixed to obtain a total sample volume of 40 ml. The biomass samples of the experiments with Sac of 25, 3 and 1 mg l
1 were collected by additional swabbing to enlarge the biomass recovery. The swab was treated with HES for 1 min in 20 ml autoclaved tap water and subsequently mixed with the 40 ml HES suspension.

the analytical procedure was described in detail before (Hijnen et al., 2009). The detection limit is 180 cells ml
1, which corresponds to 720 cells cm
2.

2.5.

The CH concentration in the biofilm samples was analyzed with the method described by Dubois et al. (1956) using glucose as the reference carbohydrate. The extinction/adsorption at 490 nm was measured directly in the biomass suspension after hydrolysis and complexation with sulphuric acid and phenol, respectively and expressed in glucose equivalent concentration. The detection limit of this parameter was approximately 5e10 mg cm
2 depending on the sampled membrane area.

2.6.

Iron content

The iron (Fe) content of the obtained biomass suspension was assessed with Atomic Absorption Spectrometry resulting in a lower limit of detection of approximately 1 mg cm
2 of membrane surface.

2.7. 2.4.

Carbohydrate analysis

Correlation analysis and statistics

Microbial parameters

ATP was measured to determine the amount of active bacterial biomass in the biofilm samples. The analysis is based on measuring the amount of light produced by an enzymatic reaction using the luciferineeluciferase assay in a luminometer (Celcis Ltd.) and has a lower limit of detection of 1 ng l
1, which corresponds with 0.01 ng cm
2 of membrane surface. The method has been described in detail previously (Magic-Knezev and Van der Kooij, 2004). The total direct cell count (TDC) was based on counting of fluorescing cells using epifluorescence microscopy (Hobbie et al., 1977) and

Correlation analyses was done by determining Pearson’s correlation coefficient between paired values of ATP, TDC, CH and Fe using SPSS 17.0 software with a significance level of p0.01. For the correlation of the biomass and Fe accumulated in the MFS units with the NPD increase (%), the weighted average concentrations (Cavg ) were calculated from the concentrations observed in the samples at different locations in the MFS units using Pn Cavg ¼

i¼1

ðCi þ Ciþn Þ=2  Aiþn Atot

(2)

1610

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 0 7 e1 6 1 6

where and Ci, Ciþn and Aiþn (cm
2) are the concentration and the surface area at the ith part of n parts of the feed channel surface (n ¼ 3e7) and Atot is the total surface area of the MFS unit. The linear regression analysis of the correlation between the Cavg -values of the biomass parameters and Fe concentrations obtained from membrane autopsy and the NPDi was performed with Excel software. For the correlation of the biomass parameters and Fe concentrations with the NPDi the non-parametric Spearman’s rank correlation coefficient was calculated and multi-regression analysis was conducted with SPSS 17.0 software.

3.

Results

of concentrations in the subsequent parts of the channel (Fig. 3). In the units supplied with acetate the percentage of the total amount of ATP at the inlet section (no FS) was 1.5e9%, in the first 2 cm with feed spacer 8e16% and in the last part (18e20 cm) 3e10% (Fig. 3c). At acetate concentrations of 1000, 500 and 25 mg C l
1 the ATP concentrations were higher than in the units supplied with the lower acetate concentrations (10, 5, 3 and 1 mg C l
1). In the blank MFS units without acetate dosing a lower ATP concentration was observed (Fig. 3b). The spatial distribution of parameters TDC and CH and also of Fe in the channels was similar to the distribution of ATP; a declining concentration in the section with feed spacer (no figures presented; weighted average concentrations presented in Table 1).

3.1.

NPD increase

3.3.

In the MFS units supplied with acetate-enriched tap water biofouling was observed at each concentration (Fig. 2) and the NPD increase (NPDi) was characterized as a first order process (Hijnen et al., 2009). The blank MFS unit supplied with prefiltered tap water without added acetate showed no NPDi during 28 days of operation (Fig. 2a), whereas in units supplied with 1 mg of acetate-C/l biofouling was observed (Fig. 2b). Also no fouling was observed within 100 days of operation in the two blank units supplied with unfiltered water (Fig. 2c). After 100 days the pressure drop started to increase in these units. The accumulated biofilm in the feed channels was colourless at high biofouling rates and short operation times (<20 days). At lower biofouling rates and operational times of 25 days the feed channel showed accumulation of brown coloured deposits. These observations initiated the analysis of the Fe concentrations in the fouled membrane samples.

3.2.

Spatial distribution of biomass and Fe

The units were sampled for biomass and Fe concentrations at different fouling conditions with relative NPDi values ranging from 71 to 3390% (Table 1). The MFS units supplied with acetate showed high ATP concentrations at the inlet section without spacer (no FS), further elevated concentrations in the first part of the section with feed spacer (0e2 cm) and a decline

Norm. pressure drop (kPa)

100

a

1000 µg C/l

The operational periods with acetate dosing, the final NPDi values and the exponential fouling rate constant Rf and the weighted average values (Cavg ) of the biomass parameters and Fe for the correlation analysis are presented in Table 1. The correlation analysis of the paired biomass parameters showed that the log value of the ATP concentration in the MFS units was significantly ( p < 0.01) correlated with the log value of the TDC and the CH concentrations, respectively (Table 2). The linear regression equation for the relationship with TDC was Log [ATP] ¼ 0.79 (95% CI 0.69e0.89) Log [TDC] þ 4.1 (95% CI 3.6e4.6) with a goodness of fit (R2) of 0.75. Based on this correlation 1 ng of ATP equals 3  106 (95%CI 5.3  105e1.7  107) TDC cm
2. The CH concentration ranged between 10 and 100 mg cm
2 at ATP concentrations of 10e100 ng cm
2, but the linear regression fit of paired ATP and CH values was poor (R2 ¼ 0.39). A better fit (R2 ¼ 0.62; p < 0.0001) was observed for the values of the units operated under acetate limitation conditions where the fouling rate Rf was below Rf,max (Sac  10 mg l
1). ATP and Fe concentrations were not correlated when the results of the MFS units operated at high Sac values with relatively short operation times (20 days) were included. For the MFS units operated at Sac values 10 mg l
1 with longer operational periods ATP and Fe concentrations were significantly correlated ( p < 0.001; Table 2). TDC did not correlate with CH and Fe. The latter two parameters correlated significantly ( p < 0.001) with a better 20 blank unfiltered

90

10 µg C/l

15

80

blank filtered

10

blank filtered

70

Correlation analysis of biomass parameters and iron

blank unfiltered

5

60

c

0 0

50

10 8 6 4 2 0

40 30 20 10 0 0

10

20

Operational time (days)

30

50

100

150

200

1 µg C/l 1 µg C/l

Stop

0

20

40

60

Start

b 80

100

Operational time (days)

Fig. 2 e The development of the Normalized pressure drop (NPD) in the MFS units supplied with filtered tap water enriched with different acetate concentrations (a,b) and (c) supplied with unfiltered tap water (lines in b and c are duplicates).

1611

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 0 7 e1 6 1 6

Table 1 e The fouling conditions of the MFS experiments and the weighted average concentration (Cavg ) of the biomass parameters and iron measured by autopsies. Acetate Sac (mg C/l)

Operational Final pressure % NPD Rf b time (days) drop (kPa) increase (ln NPDi d
1)

0 pre-filtered 0 unfiltered 1 3 5 10 25 500 1000

20 184; 184 146; 98 34; 152 39; 46 35; 28 35; 33 20; 15 15; 14; 8

2.9 17.8; 15.3 3.3; 7.2 18.6; 24.5 15.7; 37.9 7.9c; 81.4 37.8; 54.7 61.3; 22.2 52.1; 18.4; 7.9

S1a S2a

31 53

46.1 45.4

a b c d e

6.0 156; 157 71; 119 306; 526 376; 919 234c; 2369 1352; 993 3390; 507 1820; 445; 182 725 1231

<0.001 0.015; 0.016 0.063; 0.027 0.102; 0.109 0.128; 0.245 0.205; 0.224 0.766; 0.696 0.859; 1.144 1.126; 1.475; 1.097 100: 1.123 1000: 1.160

ATP (ng cm
2)

TDC CH Fe (cell  108 cm
2) (mg gluc. eq cm
2) (mg m
2)

0.8 3; 2 6; 3 20; 18 16; 13 28; 46 175; 158 200; 91 118; 58; 184

0.02 0.3; 0.4 Ndd; 0.1 2.0; 2.3 0.3; 0.4 0.3; 0.6 1.4; 0.5 2.2; 3.1 1.0; 1.2; 1.4

9.4 9.6; 10.0 Nde Nde 11.7; 11.9 13.0; 44.6 Nde 52.4; 20.6 21.1; 14.4; 47.1

0.32 394; 298 96; 129 83; 173 149; 212 201; 426 82; 79 11; 1 1; nd; 7

37 37

0.5 0.8

21.3 37.3

334 154

Starvation experiments with variable acetate dosages and starvation periods (S1 ¼ 100-5 and S2 ¼ 1000-1000-10) (Hijnen et al., 2009). First order fouling rate Rf values from Hijnen et al. (2009) modified as submitted in an erratum (Hijnen et al., in press). Low NPDi caused by preferential flow path in the feed channel. Nd ¼ not determined. Unreliable CH data due to the use of cotton swab.

goodness of fit for the units with acetate-C concentrations of 10 mg l
1 (Table 2).

3.4.

Correlation with NPDi

The study aimed at assessing the relationship between the biomass concentration and Fe with the NPDi at the time of the autopsy. The weighted average ATP and CH concentrations in the MFS units were both significantly ( p < 0.01) correlated to the NPDi (%) as evaluated with the non-parametric Spearman’s rank correlation coefficient (R2 of 0.71 and 0.91, respectively). The linear regression analysis also showed a significant ( p < 0.001) correlation with a high correlation coefficient for

ATP and CH (0.52, 0.70 and 0.82; Table 2 and Fig. 4). No significant correlation was observed for TDC with NPDi (Fig. 4c). The variability of ATP and CH concentrations in the MFS units is presented in Fig. 4 with the standard deviation (s.d.; n ¼ 3e7). The bars show an increased variability of both parameters at increased NPDi values which was caused by increased heterogeneity of the concentrations in the feed channel (Fig. 3). No correlation was found between the concentrations of Fe and the NPDi values (Fig. 4c), but the regression plots of ATP and CH with NPDi revealed that the low ATP and CH concentration at relatively high NPDi values contained Fe concentrations of >100e200 mg m
2. However, a multi-regression analysis in combination with either ATP or CH again showed

Fig. 3 e Distribution of ATP concentrations (error bar is s.d.) in the MFS units supplied with filtered tap water enriched with different acetate-C concentrations (mg lL1) and with unfiltered tap water (a,b) and (c) % of the total ATP amount in the membrane feed channel without feed spacer (no FS) and after 0e2 and 18e20 cm with feed spacer.

1612

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 0 7 e1 6 1 6

Table 2 e Correlation matrix of the different biomass parameters, Fe and NPDi measured in the standard MFS experiments; presented are the square of the Pearson correlation coefficient R2, for all correlations p-values were <0.001 except for values indicated by * ( p-value < 0.01), the number of observations (n). Sac values (mg C L
1)

TDC (cells cm
2)

CH (mg cm
2)

Fe (mg m
2)

NPDic (%)

1e1000 10 1e1000 1e1000 10 1e1000

0.75; 85a 0.66; 43a 1

0.39; 49 0.62; 30 nc 1

ncb ( p ¼ 0.09) 0.56; 42 nc 0.34; 30 0.65; 16 1

0.52; 19 0.70*; 9 nc 0.82; 13

ATP (ng cm
2) TDC (cells cm
2) CH (mg cm
2) Fe (mg m
2)

nc

a Log transformed values. b Nc ¼ no correlation. c Correlation with the weighted average concentrations of the parameters.

no significant correlation between Fe and NPDi. This indicates that the contribution of Fe accumulation in the MFS units to the pressure drop increase was limited. The minor effect of the Fe concentrations on the NPDi was also demonstrated by the Fe content in the MFS units supplied with unfiltered tap water (unfiltered blanks; Fig. 4c). The Fe concentrations in these unfiltered blanks with a limited NPDi of 156% were 298 and 394 mg m
2, whereas ATP and CH concentrations were low (2e3 ng cm
2 and 9.6e10 mg cm
2, respectively; Table 1). In the MFS units at Sac value of 10 mg L
1 considerably higher NPDi values (234e2369%) were observed at comparable Fe content of 201e426 mg m
2 and higher biomass concentrations of 28e46 ng ATP cm
2 and 13e44.6 mg CH cm
2. Similar observation was recorded for the MFS unit supplied with 100 and 5 mg l
1 acetate and intermediate starvation period; Fe, ATP and CH content was 334 mg m
2, 37 ng cm
2 and 21.3 mg cm
2, respectively at an NPDi of 725%. The pressure drop increase was due to a decrease of the open pore volume of the feed channel which in turn was a result of biomass accumulation. The relationship between the biofilm thickness and the NPDi has been described with hydraulic equations (Schock and Miquel, 1987) and is linear in the initial stage of biofouling but exponential in the subsequent stage (Hijnen et al., 2009). Assuming that ATP and CH 10000

concentrations were linearly related with the biofilm thickness the correlations with NPDi were also tested for an exponential relationship. Both exponential fits were significant ( p < 0.01), but the goodness of fit was lower compared to the linear regression (Fig. 4).

3.5.

The ATP concentration in the feed channels of the MFS units for an NPDi of 100% was 3.7 ng cm
2 (95% CI ¼ 1.3e10.9), calculated from the equation presented in Fig. 4a. For CH the threshold concentration for this criterion was calculated from the equation given in Fig. 4b at 8.1 mg cm
2 (95% CI of 6.1e11.7). This was around the detection limit of the analysis of 5e10 mg cm
2. For TDC and Fe no threshold concentration was calculated because of the lack of correlation with NPDi.

3.6.

NPDi (%)

4000

c

Unfiltered blank

3000 2000

1000

100

Fouling and accumulation rate

The fouling rate in the feed channel of the MFS units could be described with the exponential fouling rate constant Rf (Table 1). Formation of biofilms on surfaces initially is an exponential process that is rapidly followed by a linear phase due to diffusion limitation of the substrate flux into the

a

Fe < 100 mg Fe > 100 mg Fe > 200 mg

Threshold concentrations

1000 0

Exponential R2 = 0.40 p=0.003

0

100

Proportional relation

300 -2

400

500

b

2

R = 0.48 p=0.009

3000 10

200

Fe mg.m

4000

2000 NPDi=63.7[CH]-413

logNPDi=0.79[logATP]+1.55

R2 = 0.82 p<0.0001

1000

2

R =0.73 p<0.001

0

1 0.1

1

10

100 -2

ATP (ng.cm )

1000

0

20

40

60

80

100

-2

CH (µg.cm )

Fig. 4 e The relationship between the accumulated biomass measured with ATP and CH (a,b) and (c) the accumulated mass of Fe with the NPDi (%); error bar is s.d. (n [ 4e7).

1613

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 0 7 e1 6 1 6

biofilm (Rittmann, 1995). This was clearly demonstrated for ATP and Fe accumulation on glass (Van der Kooij et al., 2003). On base of the concentrations of ATP, CH and Fe measured in the MFS units after the different dosing periods the linear accumulation rate of these parameters was calculated and correlated with the Sac in the influent and Rf (Fig. 5a). Correlations of both biomass parameters with Sac showed a similar saturation curve and relationship with Sac as described for Rf (Fig. 5a). The ATP and CH accumulation rates were strongly correlated ( p < 0.0001) with Rf with R2 of 0.91 for ATP and 0.90 for CH. This clearly demonstrates the proportional relationship of both biomass parameters with the porosity decline in the feed channel.

4.

Discussion

4.1.

ATP and TDC as biomass parameters in autopsies

In full-scale SW membrane filtration installations where operation is hampered by fouling problems, it is common practice to carry out an autopsy to verify the cause of the fouling process. Only few studies have been published on the quantitative correlation between biomass parameters and operational problems in membranes such as pressure drop increase and flux decline (Flemming et al., 1993b; Fonseca et al., 2007; Vrouwenvelder et al., 2008). The results of the present study show that ATP is a suitable parameter to elucidate the role of biofilm formation in the pressure drop increase in such membranes. This conclusion was based on the correlation with the observed NPDi and supported by the good correlation between the biofilm formation rate (ng ATP cm
2 d
1) in the feed channel of the MFS with the acetate concentration (Fig. 5a) and the exponential fouling rate constant Rf (Fig. 5b). Additionally, this clearly shows that the assessment of the biofilm formation rate for the feed water of SW membranes which is also based on ATP measurements (Van der Kooij et al., 2003) is an appropriate parameter to assess the biofouling potential of the feed water.

100

a

Rf ATP

b

ATP CH

CH

ATP/CH (ng.cm-2.d-1)

Rf (ln NPDi.d-1); ATP/CH (ng/µg.cm-2.d-1)

100

The choice of a 100% NPDi in the current study to assess a threshold biomass concentration was based on a commonly used NPDi cleaning criterion of 15% over one stage of a series of six successive membrane elements (Graham et al., 1989; Hickman, 1991; Speth et al., 1998). The NPDi is not evenly distributed over the elements and usually is mainly located in the first element. Consequently, the NPDi in this element is higher (Vrouwenvelder et al., 2009a) and may be close to 100%. The threshold ATP concentration for 100% NPDi in the MFS units was 3.7 ng cm
2. A higher threshold ATP concentration for 100% NPDi of 30 ng cm
2 was reported for SW elements operated under field conditions (Vrouwenvelder et al., 2008). However, one would expect this the other way around: lower for the same NPDi in the field elements because of differences in biofilm conditions. MFS units of the present study contained relatively young biofilms whereas biofilms in field elements were more aged with a lower ratio between active (ATP) and total biomass (including EPS and dead cell material). This difference between threshold values might be caused by the difference in 100% NPDi over SW elements and the MFS of the current study. It can also be caused by correlating on one hand the maximum ATP concentration with the NPDi in field elements with a length of 1 m (Vrouwenvelder et al., 2008) and on the other hand the weighted average ATP concentration with the NPDi in a 0.2 m feed channel of the MFS as done in the present study. Consequently, despite the positive correlations ATP results in field autopsies must be interpreted with care and additional parameters which are more related to the total amount of the accumulated biomass are needed. The present study and also the mentioned field study (Vrouwenvelder et al., 2008) revealed that in contrast to ATP, TDC was not correlated with NPDi. The range of TDC values of 1  107e3.1  108 corresponds with a biofilm thickness of 0.1e1.6 mm (assumed bacteria cell volume of 0.5 mm3; diameter of 1 mm). Theoretically for the 100% NPDi a biofilm thickness of 60 mm was estimated (Hijnen et al., 2009) thus indicating that microscopic cell count (TDC) is not an accurate parameter for total biomass and more importantly biofilms consist of more than bacterial cells.

10

1

0,1

ATP; c = 1.75 R2 = 0.913 p<0.0001

10

CH; c = 0.96 R2 = 0.903 p<0.0001

1

0,1 Rf = 1.14x(1-exp{-0.693*Sac/20.3}) (Hijnen et al., 2009;in press)

0,01 0,1

1

10

Sac (µg C.l-1)

100

1000

0,01 0,001

0,01

0,1

1

10

Rf (Ln NPDi.d-1)

Fig. 5 e The correlation of the exponential fouling rate Rf and the ATP and CH accumulation rates with the acetate concentrations Sac (a) and (b) the correlation of the biomass (ATP/CH) accumulation rates with Rf.

1614

4.2.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 0 7 e1 6 1 6

Carbohydrates as a biomass parameter

Biofilms are established by adsorption and adherence of bacteria followed by growth due the supply of nutrients. Extracellular polymeric substances (EPS) excreted by bacteria to anchor themselves to the surface and to each other play a key role in the development of biofilms (i.e. protection against environmental stress, nutrient availability; Or et al., 2007; Flemming and Wingerden, 2010). CH are important components of EPS (Sutherland, 1999). The method of Dubois et al. (1956) is commonly applied to quantify the CH concentration in field SW elements (Gabelich et al., 2004; Griebe and Flemming, 1998; Ridgway et al., 1983, 1984) and their reported CH concentrations were in the same order of magnitude as measured in the current study. The positive correlation between the CH concentration and pressure drop in the feed channel of the MFS (Fig. 4b) and the proportional correlation of the CH accumulation rate (mg cm
2 d
1) to the exponential fouling rate the Rf. (Fig. 5b) confirms that the CH concentration in SW membranes is a valuable parameter in diagnostic autopsies. The threshold CH concentration of 8.1 mg cm
2 for the defined NPDi was around the current detection limit of the analysis, but the analysis can be optimized by sampling larger surface areas. More recently the CH parameter was correlated with flux decline in NF membranes (Fonseca et al., 2007) and they reported a decline of 30e80% when the CH concentrations increased to 50 mg cm
2. Another study presented a >50% flux decline and 100% NPD increase in SWM elements at a CH concentration of 12.4 mg cm
2 (Gabelich et al., 2004). Consequently, we propose the use of the parameters ATP and CH in membrane autopsy studies: the ATP method which is cheap and fast and reveals information on the accumulated active biomass in the feed channel and CH which represents the total amount of active and inactive biomass. Inclusion of the analysis of CH is especially of interest for studies on the effect of membrane cleaning with chemicals (Cornelissen et al., 2009).

4.3.

Fe accumulation and pressure drop increase

In the flat sheet MFS units without permeate production the process of particulate accumulation on the membrane surface was not influenced by vertical forces and particle settling which normally occur in SW elements with permeate production (Belfort and Nagata, 1985; Belfort, 1988). Thus, the observed accumulation of Fe particles in the MFS was a result of adsorption of these particles onto biomass produced on the membrane surface during the short cross-flow contact time. The results of the current study show that the accumulation of biomass has a far greater effect on NPDi in the feed channel than the accumulation of Fe particulates (Fig. 4c). In the units supplied with unfiltered tap water lower biofilm concentrations were observed than in the units supplied with 1 mg acetate-C L
1 (Fig. 3) but the Fe content was much higher at similar NPDi values (71e157%; Table 1). The significant correlation between CH and Fe (Table 2) indicates that EPS plays a substantial role in the adsorption of Fe onto charged biopolymers which is related to the presence of negatively charged carboxylic and phosphate groups (Wuertz et al., 2001). The dominant role of biomass in the NPDi is explained by the high water-retention capacity of EPS. Water-retention curves

show that certain polysaccharides hold more than 50e70 g of water per gram while maintaining structural coherence (Or et al., 2007; Chenu, 1993). No studies on the role of Fe particulates in SW membranes on pressure drop are known to the authors. A recent study presented the correlation of the mass deposit of Fe micro- and nanoparticles in a porous sand column (208 cm3; empty space volume of 102 cm3 and porosity of 0.49) with the pressure drop increase (Vecchia et al., 2009). In this study an Fe concentration of 2.6 mg cm
3 resulted in a PDi in the sand column of 1 kPa. The Fe concentration in the MFS supplied with unfiltered water was 300e400 mg m
2 (Table 2) which equals a volumetric concentration of 0.41e0.54 mg cm
3 (channel height of 0.0008 m and porosity of 92%). The %NPDi in this MFS was 156e157% with higher ATP and TDC values than in the pre-filtered blank (Table 1). These calculations show that Fe accumulation in the feed channel of SW elements at a level of 100 mg m
2 (10 mg cm
2) has no effect on the NPD. Further studies under field conditions are required, however, to collect additional data on the relationship of particulate accumulation and biofouling.

4.4.

Feed spacer enhances biofilm accumulation

The spacer in the feed channel enhanced biomass accumulation (Fig. 3) which is consistent with observations in other studies (Picioreanu et al., 2009; Vrouwenvelder et al., 2009b). Possible explanations for this observation are: an increase in attachment area or/and enhanced mass transfer of nutrients to the biofilm due to increased turbulence. Based on a specific surface area of the feed spacer of 7700 m2 m
3 (Picioreanu et al., 2009) it can be estimated that the spacer contributes with around 25% to the attachment surface in the feed channel. An earlier autopsy study on SWM elements from field locations showed more accumulation of biomass (ATP) on the membrane (38e90%) than on the feed spacer (5e62% of the total amount) (Vrouwenvelder et al., 2008). Preferential flow paths shown by Computational Fluid dynamics and filamentous streamers at the spacer junctions (Picioreanu et al., 2009; Vrouwenvelder et al., 2009b) were not observed in the present study. Verification of the role of the feed spacer in biofouling of SWM elements requires further research.

5.

Conclusions

The effect of biomass accumulation in spiral-wound membranes on pressure drop increase can be elucidated by measuring concentrations of active biomass with adenosinetriphosphate (ATP) and of total biomass with carbohydrates (CH; Dubois, method) in membrane autopsies. There was a significant correlation ( p < 0.001) between these parameters in the current study. This study also showed a significant ( p < 0.001) and causal relationship between both parameters and the NPDi in a model feed channel. Furthermore, the calculated ATP and CH accumulation rates were highly correlated with the observed exponential fouling rate. Threshold concentrations for 100% NPDi were 3.7 ng ATP cm
2 and 8.1 mg CH cm
2. Because ATP is related to active biomass and CH to the total biomass, monitoring both parameters in autopsies will reveal further information on the metabolic

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 0 7 e1 6 1 6

state of the accumulated biofilm. Iron accumulation in the feed channel was enhanced by the biofilm growth as demonstrated by the significant correlation between CH and Fe concentrations ( p < 0.001). Iron concentrations of 100 mg m
2 (10 mg cm
2) of membrane surface did not contribute to pressure drop increase in spiral-wound membranes. The high impact of accumulation of low biomass concentrations on pressure drop increase is attributed to the high water-retention characteristics of polysaccharides in biofilms.

Acknowledgements The research was conducted as part of the Joint Research Program of the Dutch Water Supply Companies and in the MEDINA project co-funded by the European Commission under contract number 036997. The excellent technical support by Nanda Berg, Anke Hanzens-Brouwer and Meindert de Graaf from KWR and Amandine Balthazard and David Biraud from the Ecole Nationale Supe´rieure de Chimie de Mulhouse is greatly appreciated.

references

Bailey, D.A., Jones, K.M.C., 1974. The reclamation of water from sewage effluents by reversed osmosis. Wat. Pollut. Control 74, 353e366. Baker, J.S., Dudley, L.Y., 1998. Biofouling in membrane systems e a review. Desalination 118, 81e90. Belfort, G., 1988. Membrane modules: comparison of different configurations using fluid mechanics. J. Mem. Sci. 35, 245e270. Belfort, G., Nagata, N., 1985. Fluid mechanics and cross-flow filtration: some thoughts. Desalination 53, 57e79. Chenu, C., 1993. Clay or sand polysaccharide associations as models for the interface between micro-organisms and soil: water related properties and microstructure. Geoderma 56, 143e156. Cornelissen, E.R., Hijnen, W.A.M., van der Kooij, D., 2009. Assessment of the efficiency of hydraulic and chemical cleaning for the removal of biomass from surfaces in a laboratory test, Membrane based desalination: an Integrated Approach. EU-China workshop, Qingdao, China. Dubois, M., Gilles, K.A., Hamilton, J.K., Rebers, P.A., Smith, F., 1956. Colorimetric method for determination of sugars and related substances. Anal. Chem. 23, 350e356. Flemming, H.-C., 1997. Reverse osmosis membrane biofouling. Exp. Thermal Fluid Sci. 14, 382e391. Flemming, H.-C., Schaule, G., 1988. Biofouling of membranes, a microbiological approach. Desalination 70, 95e119. Flemming, H.-C., Schaule, G., McDonough, R., 1993a. How do performance parameters respond to initial biofilm formation on separation membranes. Vom Wasser 80, 177e186. Flemming, H.-C., Schaule, G., McDonogh, R., 1993b. Permeation properties of biofilms on separation membranes and the problem of biofouling. J. Membr. Res. 87, 199e217. Flemming, H.-C., Wingender, J., 2010. The biofilm matrix. Nat. Rev. Microbiol.. doi:10.1038/nrmicro2415. Fonseca, A.C., Summers, R.S., Greenberg, A.R., Hernandez, M.T., 2007. Extra-cellular polysaccharides, soluble microbial products, and natural organic matter impact on nanofiltration membranes flux decline. Environ. Sci. Technol. 41, 2491e2497. Gabelich, C.J., Yun, T.I., Ishida, K.P., Leddy, M.B., Safarik, J., 2004. The effect of naturally occurring biopolymers on polyamide

1615

membrane fouling during surface water treatment. Desalination 161, 263e276. Graham, S.I., Reitz, R.L., Hickman, C.E., 1989. Improving reverse osmosis performance through periodic cleaning. Desalination 74, 113e124. Griebe, T., Flemming, H.-C., 1998. Biocide-free antifouling strategy to protect RO membranes from biofouling. Desalination 118, 153e156. Hickman, C.E., 1991. Membrane cleaning techniques, AWWA Seminar Proceedings, Membrane Technologies in the Water Industry, Orlando (US, Fl), March 10e13. Hijnen, W.A.M., Biraud, D., Cornelissen, E.R., van der Kooij, D., 2009. Threshold concentration of easily assimilable organic carbon in feedwater for biofouling of spiral-wound membranes. Environ. Sci. Technol. 43, 4890e4895. Hijnen, W.A.M., Biraud, D., Cornelissen, E.R., van der Kooij, D.Erratum: threshold concentration of easily assimilable organic carbon in feedwater for biofouling of spiral-wound membranes. Environ. Sci. Technol., in press. Hobbie, J.E., Daley, R.J., Jasper, S., 1977. Use of nuclepore filters for counting bacteria by fluorescence microscopy. Appl. Environ. Microbiol. 33, 1225e1228. Magic-Knezev, A., Van der Kooij, D., 2004. Optimisation and significance of ATP analysis for measuring active biomass in granular activated carbon filters used in water treatment. Water Res. 38, 3971e3979. Or, D., Phutane, S., Dechesne, A., 2007. Extracellular polymeric substances affecting pore-scale hydrologic conditions for bacterial activity in unsaturated soils. Vadose Zone J 6, 298e305. Picioreanu, C., vrouwenvelder, J.S., van Loosdrecht, M.C.M., 2009. Three-dimensional modeling of biofouling and hydrodynamics in feed spacer channels of membrane devices. J. Mem. Sci. 345, 340e354. Potts, D., Ahlert, R.C., Wang, S.S., 1981. A critical review of fouling of reversed osmosis membrane. Desalination 36. Ridgway, H.F., Flemming, H.-C., 1996. Membrane biofouling. In: Malleviall, J., Odendal, P.E., Wiesner, M.R. (Eds.), Water Treatment Membrane Processes. McGraw-Hill, New York. Ridgway, H.F., Kelly, A., Justice, C., Olson, B.H., 1983. Microbial fouling of reverse-osmosis membranes used in advanced wastewater treatment technology: chemical, bacteriological and ultrastructural analyses. Appl. Environ. Microbiol. 45, 1066e1084. Ridgway, H.F., Rigby, M.G., Argo, D.G., 1984. Adhesion of Mycobacterium sp. to cellulose diacetate membranes used for reverse osmosis. Appl. Environ. Microbiol. 47, 61e67. Ridgway, H.F., Rigby, M.G., Argo, D.G., 1985. Bacterial adhesion and fouling of reverse osmosis membranes. J. Am. Water Works Assoc. 77, 97e106. Rittmann, B.E., 1995. Fundamentals and applications of biofilm processes in drinking water treatment. In: Hrubec, J. (Ed.), The Handbook of Environmental Chemistry. Springer Verlag, Berlin, Heidelberg, New York, pp. 61e87. Schaule, G., Kern, A., Flemming, H.-C., 1993. RO treatment of dump trickling water membrane biofouling: a case history. Desalination & Water Reuse 3, 17e23. Schock, G., Miquel, A., 1987. Mass transfer and pressure loss in spiral-wound modules. Desalination 64, 339e352. Speth, T.F., Summers, R.S., Guess, A.M., 1998. Nanofiltration foulants from a treated surface water. Environ. Sci. Technol. 32, 3612e3617. Sutherland, I.W., 1999. Polysaccharases for microbial exopolysaccharides. Carbohydr. Polym. 38, 319e328. Van der Kooij, D., Vrouwenvelder, H.R., Veenendaal, H.R., 2003. Elucidation and control of biofilm formation processes in water treatment and distribution using the unified biofilm approach. Wat. Sci. Technol 47, 83e90. Vecchia, E.D., Luna, M., Sethi, R., 2009. Transport in porous media of highly concentrated iron micro- and nanoparticles in the presence of Xanthan gum. Environ. Sci. Technol. 43, 8942e8947.

1616

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 0 7 e1 6 1 6

Vrouwenvelder, H.R., Van Paassen, J.A.M., Folmer, H.C., Hofman, J.A.M.H., Nederlof, M.M., Van der Kooij, D., 1998. Biofouling of membranes for drinking water production. Desalination 118, 157e166. Vrouwenvelder, J.S., Van Paassen, J.A.M., Wessels, L.P., Van Dam, A.F., Bakker, S.M., 2006. The membrane fouling simulator: a practical tool for fouling prediction and control. J. Mem. Sci. 281, 316e324. Vrouwenvelder, J.S., Manolarakis, S.A., van der Hoek, J.P., van Paassen, J.A.M., van der Meer, W.G.J., van Agtmaal, J.M.C., Prummel, H.D.M., Kruithof, J.C., van Loosdrecht, M.C.M., 2008. Quantitative biofouling diagnosis in full scale nanofiltration and reverse osmosis installations. Water Res. 42, 4856e4868.

Vrouwenvelder, H.R., Paassen, J.A.M., Kruithof, J.C., van Loosdrecht, M.C.M., 2009a. Sensitive pressure drop measurements of individual lead membrane elements for accurate early biofouling detection. J. Mem. Sci. 338, 92e99. Vrouwenvelder, H.R., Graf von der Schulenburg, D.A., Kruithof, J.C., Johns, M.L., van Loosdrecht, M.C.M., 2009b. Biofouling of spiralwound nanofiltration and reverse osmosis membranes: a feed spacer problem. Water Res. 43, 583e594. Wuertz, S., Spaeth, R., Hinderberger, A., Griebe, T., Flemming, H.-C., Wilderer, P.A., 2001. A new method for extraction of extracellular polymeric substances from biofilms and activated sludge suitable for direct quantification of sorbed metals. Wat. Sci. Technol. 43, 25e31.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 1 7 e1 6 2 8

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Photocatalytic degradation of benzenesulfonate on colloidal titanium dioxide Erzse´bet Szabo´-Ba´rdos a, Otı´lia Markovics a, Otto´ Horva´th a,*, Norbert To¨ro} b, Gyula Kiss c a

University of Pannonia, Institute of Chemistry, Department of General and Inorganic Chemistry, H-8200 Veszpre´m, POB. 158, Hungary University of Pannonia, Institute of Environmental Sciences, H-8200 Veszpre´m, POB. 158, Hungary c Air Chemistry Group of Hungarian Academy of Sciences at University of Pannonia, H-8201 Veszpre´m, POB. 158, Hungary b

article info

abstract

Article history:

Titanium dioxide-mediated photocatalyzed degradation of benzenesulfonate (BS) was

Received 16 August 2010

investigated by monitoring chemical oxygen demand (COD), total organic carbon (TOC)

Received in revised form

content, sulfate concentration, pH as well as the absorption and emission spectral changes

26 November 2010

in both argon-saturated and aerated systems. Liquid chromatography-mass spectrometry

Accepted 29 November 2010

analysis was utilized for the detection of intermediates formed during the irradiation in the

Available online 7 December 2010

UVA range (lmax ¼ 350 nm). The results obtained by these analytical techniques indicate that the initial step of degradation is hydroxylation of the starting surfactant, resulting in

Keywords:

the production of hydroxy- and dihydroxybenzenesulfonates. These reactions were

Benzenesulfonate

accompanied by desulfonation, which increases [Hþ] in both argon-saturated and aerated

Photocatalysis

systems. In accordance with our previous theoretical calculations, the formation of ortho-

Titanium dioxide

and meta-hydroxylated derivatives is favored in the first step. The main product of the

Oxidative degradation

further oxygenation of these derivatives was 2,5-dihydroxy-benzesulfonate. No decay of

Intermediates

the hydroxy species occurred during the 8-h irradiation in the absence of dissolved oxygen.

Oxygenation

In the aerated system much more efficient desulfonation and hydroxylation, moreover, a significant decrease of TOC took place at the initial stage. Further hydroxylation led to cleavage of the aromatic system, due to the formation of polyhydroxy derivatives, followed by ring fission, resulting in the production of aldehydes and carboxylic acids. Total mineralization was achieved by the end of the 8-h photocatalysis. It has been proved that in this photocatalytic procedure the presence of dissolved oxygen is necessary for the cleavage of the aromatic ring because hydroxyl radicals photochemically formed in the deaerated system too alone are not able to break the CeC bonds. ª 2010 Elsevier Ltd. All rights reserved.

1.

Introduction

Sulfonated aromatic compounds are used in consumer products and in many industrial processes (Tully, 1997). Linear alkylbenzenesulfonates (LASs) are utilized as surfactants in laundry and cleansing products; sulfonated azo dyes are used in diverse applications including the textile industry to color natural fibers. Fluorescent whitening agents are based on

sulfonated aromatics, benzene- and naphthalenesulfonates are used mainly as intermediates for the manufacturing of azo dyestuffs, pharmaceuticals and tanning agents. While the frequently used linear alkylbenzene sulfonate surfactants have been thoroughly investigated for their pollution effects and degradation possibilities, little work has been carried out with benzene- and naphthalenesulfonates (Lange et al., 2000). In contrast to LASs, which have been found to be readily

* Corresponding author. Tel.: þ36 88 624 159; fax: þ36 88 624 548. E-mail address: [email protected] (O. Horva´th). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.11.045

1618

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 1 7 e1 6 2 8

biodegradable (Takada and Ishiwatari, 1990; Hashim et al., 1992), aromatic sulfonates without long alkyl side chains proved to be resistant to biodegradation (Cain, 1981). This is especially true for benzene- and naphthalenesulfonates with sulfo, nitro and amino groups (Brilon et al., 1981; Zu¨rrer et al., 1987; Wittich et al., 1988). Because of their low n-octanolewater partition coefficients (Greim et al., 1994) and high mobility within aquatic systems, polar aromatic sulfonates are regularly found in natural waters. These compounds have been detected in wastewater effluents, surface waters (Alonso et al., 1999; Zerbinati et al., 1999), and landfill leachates (Suter et al., 1999; Riediker et al., 2000a, 2000b). Low biodegradability and high mobility make them potentially hazardous with respect to contamination of ground water and drinking water supplies (Reemtsma, 1996). The aquatic toxicity of aromatic sulfonates appears to be small (Integral Consulting Inc., 2007). However, a toxicity study using Photobacterium phosphoreum indicated that among several naphthalene- and benzenesulfonates, the unsubstituted benzenesulfonate turned out to be the most toxic (Alonso et al., 2005). Several methods have been tested for the degradation of these pollutants over the past 15 years. Mineralization of benzenesulfonates was accomplished by contact glow discharge electrolysis (Amano et al., 2004; Amano and Tezuka, 2006). Ozonation proved to be efficient only in the presence of activated carbon, which ensures high local concentration of the reactants, due to their adsorption (Faria et al., 2008). Carboxylic acids (oxalic and formic acid) were detected as intermediates in both cases. Electrochemical oxidation of 1, 5-naphthalenedisulfonic acid proved to be promising by in situ generation of silver(II) or peroxydisulfate as mediators (Ravera et al., 2004). Recently, oxidative degradation of this pollutant was achieved in the presence of hydrogen peroxide activated by microwaves or UV irradiation (Ravera et al., 2009, 2010). Photocatalytic methods have been proved to be suitable for the treatment of water polluted with organic and inorganic contaminants, such as surfactants (Horva´th and Husza´nk, 2003; Horva´th et al., 2005), heavy metals (Kanki et al., 2004), chromium(VI) (Kajitvichyanukul et al., 2005; Gkika et al., 2006), and various pesticides (Konstantinou and Albanis, 2003; Devipriya and Yesodharan, 2005). In heterogeneous photocatalytic methods applied for the degradation of various organic pollutants the most widely used material is titanium dioxide, TiO2 (Szabo´-Ba´rdos et al., 2003, 2004; Fabbri et al., 2006; Patsoura et al., 2007). The most efficient oxidizing agent in TiO2-mediated photocatalysis is the HO radical, which can be formed in aqueous systems via the oxidation of adsorbed water by the positively charged hole (hþ vb ) formed in the valence band of the semiconductor upon excitation (Hoffmann et al., 1995).   þ TiO2 þ hv/TiO2 e cb þ hvb   þ TiO2 hvb þ H2 Oads /TiO2 þ HO þ Hþ 

(2)





2.

Experimental section

2.1.

Materials

The titanium dioxide catalyst used in all experiments was Degussa P25 (70% anatase, 30% rutile; with a surface area of 50 m2 g1). Benzenesulfonic acid as well as 4-hydroxybenzenesulfonic acid and 2,5-dihydroxybenzenesulfonic acid (as standards for the analyses) of pure reagent grade were purchased from Merck. Compressed air or argon for stirring was introduced into the reaction mixtures from gas bottles. Beside stirring, air also served as an electron acceptor (i.e., oxidizer). High purity water used in these experiments was double distilled and then purified with a Milli-Q system.

(1)

In aerated systems, electrons (e cb ) photogenerated in the conduction band can react with dissolved oxygen, resulting in the formation of superoxide and peroxide ions.
 (3) TiO2 e cb þ O2ads /TiO2 þ O2
 2 TiO2 e cb þ O2 /TiO2 þ O2

TiO2-based techniques were applied for degradation of various amino acids (Matsushita et al., 2007; Szabo´-Ba´rdos et al., 2006) and surfactants (Zhang et al., 2003; Hegyi and Horva´th, 2004). Efficient photocatalytic mineralization of 1,5naphthalenedisulfonate was achieved on colloidal titanium dioxide (Szabo´-Ba´rdos et al., 2008). In this case, HPLC/MS analysis indicated that the degradation pathway leads to the formation and subsequent decay of benzenesulfonate. Although the photocatalytic degradation of benzenesulfonate was studied in the past (Sangchakr et al., 1995), neither a detailed analysis regarding the possible degradation pathways was carried out nor was the role of dissolved oxygen investigated. Since benzenesulfonate is involved in several industrial processes, from which it can get into natural waters as a pollutant, it is important to gain more information about its photocatalytic mineralization. Elucidation of its decomposition contributes to a better understanding of the TiO2-mediated photodegradation of disulfonatonaphthalenes too, in which it is an important intermediate (Szabo´-Ba´rdos et al., 2008). The objective of this study is to investigate the TiO2-based photocatalytic treatment of benzenesulfonate (abbreviated as BS) in a laboratory-scale reactor in order to elucidate the oxidative degradation mechanisms of this surfactant. The changes in the properties of the treated solution during photocatalysis were followed by the measurement of the pH, the absorption and emission spectra, the total organic carbon (TOC) content and the chemical oxygen demand (COD) of the irradiated mixture. TOC was chosen as a mineralization index of BS, while COD is related to the average oxidation number of the carbon atoms in the system. In addition, the intermediates were identified with MS, giving vital information to determine the main steps of the degradation mechanism. Beside their theoretical importance, the results of this work can be utilized in the design of photocatalytic procedures for wastewater treatment.

(4)

2.2.

Photochemical experiments

Photochemical experiments were carried out using a laboratory-scale reactor with an effective volume of 1.6 dm3. The heterogeneous reaction mixture (TiO2 suspension) was circulated by a peristaltic pump through the reactor and the buffer vessel and by continuously bubbling air or other gases, such as Ar with a flow rate of 40 dm3 h1 within the reactor. The photon flux of the internal light source (40 W,

1619

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 1 7 e1 6 2 8

3.

Results and discussion

3.1. Change of TOC and COD in photoassisted degradation of benzenesulfonate Practically no change of TOC was observed over the entire irradiation period of 480 min in the argon-saturated system. In the aerated system, as can be seen in Fig. 1, an almost total mineralization of 7  104 M BS was achieved within a 480 min

6 5

-3

COD

150

COD/TOC

4

100

3 2

COD/TOC

For analysis, 4 mL samples were taken with a syringe from the reactors. The solid phase of samples, when necessary, was separated by filtration using Millipore Millex-LCR PTFE 0.45 mm filters. The pH of the aqueous phase of the reaction mixture was measured with SEN Tix 41 electrode. The concentration of benzenesulfonate was determined by HPLC, using calibration curves previously prepared. The absorption and emission spectra were recorded with a Specord S 100 diode array spectrophotometer and a PerkinElmer LS50B spectrofluorometer, respectively, using quartz cuvettes of various pathlengths. Mineralization was followed by measuring the total organic carbon (TOC) concentration, with a Thermo Electron Corporation TOC TN 1200 apparatus. The dichromate method was applied for the determination of the chemical oxygen demand (COD). Liquid chromatography-mass spectrometry (HPLC-MS) analyses were performed on an Agilent 1100 Series LC/MSD Trap VL System. The HPLC consisted of an Agilent 1100 binary gradient pump, an Agilent 1100 manual injector valve with a 20 ml loop and a 1100 diode array detector that recorded absorbance in the wavelength range of 190e800 nm. Isocratic separations were carried out at room temperature on a NovaPack C18 column (150 mm  3.9 mm I.D., 5 mm particle diameter) with a flow rate of 0.5 cm3/min. The eluent was composed of 1.5% methanol þ 1.5% acetonitrile þ 96.9% HPLC grade water þ 0.1% HCOOH. The chromatograms were recorded for 15 min. MS was performed with an ion-trap mass spectrometer operated in negative electrospray ionization mode with nebulizer gas pressure of 40 psi, drying gas temperature and flow rate of 305  C and 9 dm3/min, respectively. Anions were recorded from 50 m/z to 500 m/z. Sulfate and sulfite concentrations were determined with ion chromatography, using a Dionex model 2010i apparatus described elsewhere (Horva´th and Hajo´s, 2006). All samples were analyzed in triplicates with a flow rate of 1.7 cm3/min. The separator column (250 mm  4 mm) was based on a 13 mm polystyrenedivinylbenzene copolymer agglomerated with completely aminated anion exchange latex. The ion exchange capacity of the column was 20 mequiv/column. In order not to disturb the subsequent analyses, especially the HPLC-MS experiments, no buffer was used in this system. Thus, the initial pH of the reaction mixtures was the natural one without any adjustment, i.e., pH ¼ 5.0e5.2.

TOC, COD/mg dm

Analytical procedures

200 TOC

50 1 0

0

120

240

360

480

0

Irradiation time/min

B -3

2.3.

A

TOC/mg dm

lmax ¼ 350 nm, i.e., UVA range) was determined by tris(oxalato)ferrate(III) chemical actinometry (Rabek, 1982; Kirk and Namasivayam, 1983). It was estimated to be 1.45  105 E s1.

50 40 30 20 10 0

0

140

280

420

Irradiation time/min Fig. 1 e A) Change of chemical oxygen demand (COD) (B) and total organic carbon (TOC) content (C) during the photocatalytic treatment of an aerated system containing 7 3 10L4 M benzenesulfonate and 1 g dmL3 TiO2, pH [ 5.2. The ratio of COD/TOC (6) is also plotted as the function of irradiation time. B) TOC belonging to the whole system (C), the starting material (B), and the intermediates formed (6) as functions of the irradiation time in the same system.

photocatalytic treatment. The decay curves for TOC and COD are very similar, with gradually decreasing rates, but the change of COD was significantly faster (Fig. 1A). Accordingly, the COD/TOC ratio continuously decreased during the whole irradiation period. It clearly indicates that the oxidation of the original substrate and of the intermediates formed is faster than the formation of hydrogencarbonate or carbon dioxide, i.e., mineralization. Generally, in the first period of oxidation processes oxygenation is the predominant type of reaction, and subsequent cleavage steps result in mineralization. Nevertheless, the TOC in this system perceptibly decreased in the initial stage too (in the first 30 min period of time), indicating that cleavage took place rather early in this process. Then an approximately linear decrease is shown in the interval of 30e240 min. At longer irradiation times (240e480 min) the decrease of TOC gradually slows down due to the consumption of the oxidizable intermediates, approaching total mineralization. The actual concentrations of the starting material were determined from the HPLC-MS signals as shown later. Thus, the TOC values corresponding to the unreacted benzenesulfonate

1620

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 1 7 e1 6 2 8

could also be calculated. Fig. 1B displays the TOC versus time plots belonging to the overall system, the unreacted surfactant (BS), as well as the intermediates formed during the degradation process. The latter curve is the difference of the previous two. The TOC values corresponding to the intermediates show a maximum at 80e90 min where the concentration of the unreacted benzenesulfonate is still significant. At longer times (above 90 min) the TOC representing the intermediates is diminishing because the rate of the mineralization of these species exceeds that of their formation. In this period of irradiation, especially above 120e140 min, the total TOC mostly belongs to the intermediates because the predominant part of BS has already been transformed. Hence, further decrease of TOC can be practically attributed to the mineralization of the intermediates.

3.2.

Change of pH and release of sulfate ions

During the photocatalytic oxidation of BS, especially in the first 180 min, a continuous decrease of pH was observed in both the argon-saturated and the aerated systems (Fig. 2). From the initial value of pH ¼ 5.2, it decreased to pH ¼ 4.5 in the deaerated system, while in the air-saturated reaction mixture down to pH ¼ 3.4. This corresponds to an increase of [Hþ] by about 2.3  105 M in the argon-saturated and 3.9  104 M in the aerated system. Similar values can be observed for the concentration of the sulfate ions formed during the photocatalytic degradation of BS. The concentration of these species increased during the entire irradiation period, displaying a sigmoidal shape. In the case of the air-saturated system, after 8 h irradiation, the concentration of SO2 4 was close to the amount of sulfur that the initial reaction mixture contained. Similar results were observed for the formation of SO2 4 in the case of the photocatalytic degradation of 1,5-naphthalenedisulfonate (Szabo´-Ba´rdos et al., 2008), and also for benzenesulfonate although under different conditions (Sangchakr et al., 1995). This phenomenon may be attributed

5.0

0.6

4.5 0.4 4.0 0.2

0

3.5

0

120

240

360

480



3.0

Irradiation time/min Fig. 2 e Change of SO2L 4 concentration (B,C) and pH (6, :) during the photocatalytic treatment of the system the concentrations of which are given at Fig. 1. The contour symbols stand for the aerated, while the full symbols for the argon-saturated system.





 2 þ RSO 3 þ HO /HSO4 þ R 4H þ SO4 þ R

(5)

Carbon oxidation, however, does not significantly contribute to the drop of pH because neither small molecular weight organic acids, which can be formed as intermediates, nor carbon dioxide release appreciable amount of hydrogen ions: eCH] þ 3HO / CO2 þ 2H2O

(6)

Hence, in the subsequent, longer period of irradiation (from 180 min to 480 min) only a very slight increase of the Hþ concentration was observed. In argon-saturated system, in accordance with the change of pH, only a very slow but continuous release of sulfate took place. Even after 8 h irradiation about 2.2  105 M of sulfate was formed indicating that the generation of hydroxyl radicals is over one order of magnitude slower in the absence of dissolved oxygen. Under anaerobic conditions the photogenerated electrons are not captured by O2. In our system BS itself is the most probable candidate for this reaction because its electronscavenging rate constant is high (4  109 M1s1 (Buxton et al., 1988)) and it is readily adsorbed on the surface of the catalyst at pH ¼ 5.0. On the basis of our measurements, at pH ¼ 5.0 the extent of the adsorption of BS is 13% (at 1 g/dm3 TiO2 and 103 mol/dm3 BS). This efficient adsorption can be attributed to the Coulombic attraction between the positively charged surface of the catalyst (pHzpc ¼ 6.5 for anatase (Sun et al., 2005)) and the anionic benzenesulfonate (pK ¼ 2.36 (Faria et al., 2008)). Besides, using ion chromatography, we have ions in the argon-saturated system. Its detected SO2 3 concentration increased during the first 180 min of irradiation. This species is the primary product of electron-scavenging by BS. Sulfite was gradually transformed to SO2 4 at longer irradiation times.

3.3.

5.5

pH

Conc. of SO42-/mM

0.8

to the following reaction involving an HO radical, the main oxidizing agent in this photocatalytic system:

Change of the absorption and emission spectra

The absorption and emission spectra of the solution (after removal of the colloidal catalyst) were recorded for the qualitative monitoring of the chemical change in the system during the photolysis. The absorption spectrum of benzenesulfonate is characterized by strong bands in the 250e270 nm range, assigned to p*)p transitions in the aromatic system. In the argon-saturated system an apparently continuous increase of the original bands can be observed (Fig. 3A). However, this change is accompanied by even stronger increases of the absorbance below 250 nm and above 270 nm with the appearance of a new band at 277 nm and a shoulder at about 305 nm. This phenomenon indicates that new aromatic derivatives were gradually formed over the whole irradiation period. Their absorbance overcompensates the simultaneous decrease of the original bands (superposing on the increasing “baseline”). A similar spectral change is shown for the first (30e50 min) period of the irradiation in the aerated system, indicating the formation of intermediates and the disappearance of the

1621

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 1 7 e1 6 2 8

0.9

0.6

0.3

0.0 225

A

0 20 45 90 180 240 360 420

275

325

Absorbance

0.9

375

0 5 10 20 30 50 90 120 150 180 240

0.3 0.0 225

275

325

320

370

420

60

0 5

50

10 15 40 50

40 30

90 120 150

20 10

375

Wavelength/nm Fig. 3 e Change of the absorption spectrum of the bulk solution during the irradiation of argon-saturated (A) and aerated (B) systems the concentrations of which are given at Fig. 1, ([ [ 2.0 cm). The samples were taken at the indicated times in min.

starting material (Fig. 3B). The increase in the longer-wavelength region suggests that the intermediates formed in this period of time kept the aromatic structure. After the increase of absorption, a gradual decrease can be observed at irradiation times longer than 50 min (thick lines in Fig. 3B). This phenomenon suggests the degradation of the aromatic intermediates formed during the initial period of the photocatalytic process. The emission spectrum of benzenesulfonate displays a strong band at 287 nm. In accordance with the absorption changes, in both argon-saturated and aerated solutions, the luminescence intensity at this band was found to disappear (Fig. 4), indicating the decay of the starting material. Simultaneously, a longer-wavelength band with a shoulder arose and continuously increased in the deoxygenated system. While the position of the stronger band gradually shifted to the red (to 312 nm), the wavelength of the shoulder remained the same (350 nm). This is the effect of the shorter-wavelength (287 nm) emission of the starting material, also causing that the relative intensity of the 312-nm band compared to that of the shoulder gradually decreased. This phenomenon indicates that this new emission may be attributed to at least one intermediate keeping the aromatic structure. HPLC-MS analysis (further explanation follows) indicated that the 312-nm

470

Irradiation time/min

B

0.6

40

0 270

Intensity

1.2

0 10 20 30 45 60 90 120 180 240 300 360

20

Wavelength/nm

B

80 60

Intensity

Absorbance

A

0 270

180

320

370

420

470

Wavelength/nm Fig. 4 e Change of the emission spectrum of the bulk solution during the irradiation of argon-saturated (A) and aerated (B) systems the concentrations of which are given at Fig. 1, (lexc [ 262 nm, [ [ 1.0 cm, slit [ 15 nm). The samples were taken at the indicated times in min.

emission band can be attributed to hydroxy derivatives, while the 350-nm band (shoulder) to dihydroxy species. We have carried out independent experiments with standard compounds and got the same type of emission, confirming the previous assignments. After the first 100 min apparently no change in the characteristic wavelengths of the new emission can be observed, indicating that no further luminescent intermediates were formed during the subsequent period of irradiation, at least not in a significant concentration. In the aerated system, during the first 50-min irradiation a similar spectral change can be observed (thin lines in Fig. 4B), but the maximum intensity reached is lower than in the absence of dissolved oxygen, and the position of the main band is more red-shifted (to 320 nm) at the maximum intensity. Besides, the intensity difference between the emission band at 320 nm and the shoulder at 350 nm is significantly smaller too. In the subsequent period of irradiation, similarly to the corresponding absorption spectral change, the emission intensity continuously decreased in the whole range of wavelength (thick blue lines in Fig. 4B). The band at 320 nm gradually merged into the shoulder at 350 nm, indicating that the latter one belongs to other, new intermediates, the ratio of which increased. This conclusion is confirmed by the emission of these latter intermediates obtained by excitation at a different wavelength, which results in a single-band spectrum deviating

1622

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 1 7 e1 6 2 8

200

100 50

360

410

2.0E+5

2,1 2,7

1.6E+5

Intensity

Intensity

150

0 310

A

0 5 10 20 30 45 60 90 120 150 180 240 300 360

3

1.2E+5 8.0E+4 4.0E+4

460

0.0E+0

Wavelength/ nm Fig. 5 e Change of the emission spectrum of the bulk solution during the irradiation of an aerated system the concentrations in which are given at Fig. 1, (lexc [ 302 nm, [ [ 1.0 cm, slit [ 15 nm). The samples were taken at the indicated times in min.

0

100

200

300

400

Irradiation time/min

B

8.0E+3 2,1 3

from the previous mixed ones (Fig. 5). At this wavelength (302 nm) only the latter intermediates (dihydroxy derivatives as identified by HPLC-MS measurements) are excited. Since no new bands appeared during the decay of the luminescence in the period of 150e480 min, the emitting intermediates were transformed into derivatives which are not aromatic any more, and, hence, not luminescent either.

3.4.

Analysis of the intermediate products

For the detection of intermediates formed during the oxidative degradation of benzenesulfonate, HPLC-MS analyses were also carried out. Identification of the intermediates provides useful hints for the determination of the degradation pathways in this photocatalytic process. The samples for HPLC-MS analysis were taken at various times during the irradiation period. Both argon-saturated and aerated systems were studied by this method. Although most of the intermediates formed during the photocatalytic treatment could not be ideally separated by HPLC, formation and decay of several species were detected by following the intensity of the significant m/z signals as functions of time. It should be mentioned that the absolute intensities of the signals corresponding to various species are generally not comparable because the sensitivity of the detector system with regard to these ions can be significantly different. Hence, these intensity values can rarely be used for quantitative comparative analysis. Nevertheless, the change of the relative concentration of each separate species can be followed by measuring the intensity of the corresponding m/z signal. In the case of the argon-saturated system the predominant species detected, besides the starting material (m/z ¼ 157), belong to the m/z signals 173 and 189. The first value corresponds to the hydroxybenzenesulfonates, while the second one to the dihydroxy derivatives. The formation of these species can be attributed to the reaction of BS with HO radical. As Fig. 6A indicates, three isomers of hydroxybenzenesulfonate were detected at different retention times. A control run with a standard solution of the 4-hydroxy ( para-)

Intensity

6.0E+3

4.0E+3

2.0E+3

0.0E+0 0

100

200

300

Irradiation time/min Fig. 6 e Changes of mass spectrometric signal intensities observed at in the samples taken during the irradiation of argon-saturated system (A) for the m/z [ 173 (hydroxybenzenesulfonates) at 2.1, 2.7, and 3.0 min retention times, and (B) for m/z [ 189 (dihydroxybenzenesulfonates) at 2.1 and 3.0 min retention times. The concentrations of the initial system are given at Fig. 1.

derivative indicated that its retention time was 2.1 min. Thus, the para-hydroxylated isomer gave the smallest signal. The sensitivities of the MS detector for the three hydroxy isomers do not significantly deviate because they are determined by the sulfonate group. Thus, the para-hydroxylated isomer was formed in the lowest concentration. Thus, the other two more intensive signals belong to the ortho- and meta-hydroxylated isomers. Since on a C18 column, at the applied conditions, the retention order is ortho > meta > para (Lake, 2010), according to Fig. 6A, the ortho-hydroxy species was formed in the highest concentration. This observation deviates from earlier results, which indicated the exclusive formation of the para-hydroxy derivative (Sangchakr et al., 1995), and it is in agreement with our previous quantum chemical calculation, which favored the formation of the ortho-hydroxy derivative (Szabo´-Ba´rdos et al., 2008). According to the general rule, however, the electron-withdrawing sulfo group decreases the electron density at the ortho- and para- positions of the aromatic ring

1623

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 1 7 e1 6 2 8

through resonance. Hence the meta- position is favored for an electrophilic attack. Nevertheless, recently, in photocatalytic degradation of benzene with electron-withdrawing groups such as eNO2 and eCN, formation of all the three hydroxy derivatives were observed with o:m:p ratio of 29:34:37 for nitrobenzene, and 45:30:25 for cyanobenzene (Palmisano et al., 2007b, 2007a). For the dihydroxy derivatives only two signals of different retention time were observed, with lower intensities than for the hydroxy isomers (Fig. 6B). The signal at 3.0 min retention time could be assigned to the 2,5-dihydroxy derivative as confirmed by control runs with the standard compound. This seems to be the predominant dihydroxy isomer formed in this process because the intensity of the other signal is more than one order of magnitude lower. The one hydroxy substituent of the 2,5-dihydroxy derivative is at the ortho-position and the other at the meta-. The predominant formation of this isomer is in accordance with the ratio of the hydroxy derivatives (ortho > meta > para) observed in this

A

9.0E+4 2.1

system and the theoretical considerations favorizing hydroxylation at ortho and meta positions. The concentration of all of these intermediates monotonously increased, although at different rates, during the irradiation of the argonsaturated reaction mixture. In the aerated system both the formation and decay of these intermediates were observed (Fig. 7). In this case the maximum concentrations of the hydroxy derivatives are significantly lower than those in the argon-saturated system (Fig. 7A) as a consequence of their fast transformation to dihydroxy and other intermediates. Hence, the formation of dihydroxy species is much more efficient in the aerated mixture, and their maximum concentrations are about one order of magnitude higher than the corresponding values in the deoxygenated system (Fig. 7B). Fig. 8A summarizes the decay of the starting material (benzenesulfonate, m/z ¼ 157), and the formation and decay of the hydroxy (m/z ¼ 173), dihydroxy (m/z ¼ 189), and trihydroxy (m/z ¼ 205) derivatives. The values of the signals belonging to m/z ¼ 173, m/z ¼ 189, and m/z ¼ 205 are cumulative intensities, i.e., the corresponding isomers are not distinguished here. In

Intensity

2.7 3

6.0E+4

A

157 173

2.0E+5

Intensity

3.0E+4

0.0E+0 0

100

200

300

189 205

1.5E+5 1.0E+5 5.0E+4

Irradiation time/min

B

2.5E+5

1.2E+5

0.0E+0

2.1

0

Intensity

3

100

200

300

Irradiation time/min

8.0E+4

B

10000 207 239

7500

Intensity

4.0E+4

0.0E+0 0

100

200

300

273 193

5000

2500

Irradiation time/min Fig. 7 e Changes of mass spectrometric signal intensities observed in the samples taken during the irradiation of aerated system (A) for the m/z [ 173 (hydroxybenzenesulfonates) at 2.1, 2.7, and 3.0 min retention times, and (B) for m/z [ 189 (dihydroxybenzenesulfonates) at 2.1 and 3.0 min retention times. The concentrations of the initial system are given at Fig. 1.

0

0

60

120

180

240

300

Irradiation time/min Fig. 8 e Changes of the mass spectrometric signal intensities at various m/z values in the samples taken during the irradiation of the aerated system the concentrations of which are given at Fig. 1.

1624

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 1 7 e1 6 2 8

accordance with the absorption and emission data, the predominant part (over 90e95%) of the original surfactant was converted during the first 120-min treatment. As shown in Fig. 8A, the maximum concentration of the hydroxy derivatives is at about 50 min, while that of the dihydroxy intermediates at about 90 min. Further oxygenation resulted in the formation of trihydroxy derivatives with m/z value of 205 (Fig. 8A). Their maximum concentrations appeared at about 110e120 min. These intermediates still retain their aromaticity. These maximum positions in time are in good accordance with the changes of the absorption and emission spectra, confirming that aromatic luminescent derivatives were formed. With maximum concentration at about 120e150 min species with m/z ¼ 207 and 239 were also detected (Fig. 8B). According to the absorption and emission spectra in this period of time, these compounds are not aromatic, moreover they can be ring-opened intermediates, such as aldehydes and carboxylic acids. The transformation of these intermediates involves gradual hydroxylation (and saturation of the C]C bond due to addition of HO radicals) and shortening of the carbon-chain as indicated by the m/z values of 273 and 193, respectively (Fig. 8B). Desulfonation takes place with the intermediate of m/z ¼ 273 as suggested by the m/z value of 209 (Fig. 9). The intermediate of shorter carbon-chain (m/z ¼ 193) undergoes gradual hydroxylation (and saturation of the C]C bond) too and oxidation of the terminal carbon atom (m/z ¼ 243). Desulfonation and saturation of the latter species (m/z ¼ 243) is suggested by the detection of the intermediate of m/z ¼ 179. However, decarboxylation of the intermediate with longer carbon-chain (m/z ¼ 209) leads also to the species of m/z ¼ 179. These results indicate that, depending on the order of chainshortening, desulfonation, saturation, and hydroxylation of the ring-opened intermediates, degradation can take place via different parallel pathways. This conclusion is confirmed by the fact that for several saturated intermediates the m/z values of the corresponding unsaturated species can also be found in the mass spectrum, and vice versa (not shown in Figs. 8 and 9).

7000

209 243 179 123 89

6000

Intensity

5000 4000 3000 2000 1000 0 0

120

240

360

480

Irradiation time/min Fig. 9 e Changes of the mass spectrometric signal intensities at various m/z values in the samples taken during the irradiation of aerated system the concentrations of which are given at Fig. 1.

Notably, besides the addition of HO radicals, the saturation can be the consequence of the reaction with H atoms, which are formed via scavenging of the photogenerated electrons by hydrogen ions. The low pH of the system after 120e150 min of irradiation (about 3.5, see Fig. 2), and the high value of the electron-scavenging rate constant of Hþ (2.8  1010 M1s1, (Buxton et al., 1988)) confirm the formation of nascent hydrogen. Further oxidation of the intermediate of m/z ¼ 179 is accompanied by shortening of the aliphatic chain (m/z ¼ 123). During the progress of mineralization, generation of intermediates of lower m/z values was observed, such as the formation of oxalate (m/z ¼ 89). This conclusion is confirmed by the observation in the case of degradation of BS by ozonation and glow discharge electrolysis, where oxalic and maleic acids were detected (Faria et al., 2008; Amano et al., 2004; Amano and Tezuka, 2006). The efficient photocatalytic oxidation of oxalic acid on colloidal titanium dioxide was studied earlier in our group (Szabo´-Ba´rdos et al., 2003, 2004). The final products of the total mineralization are water, carbon dioxide, and sulfate ions. Scheme 1 tentatively summarizes the possible degradation pathways of BS, which were compiled by taking into consideration the succession of formation and decay of the detected intermediates. Although this is not a detailed mechanism, it contains most of the key steps in the degradation of BS. Similar steps were proposed in the oxidative degradation pathways of p-toluenesulfonic acid treated with thermally activated hydrogen peroxide (Sto¨ffler and Luft, 1999).

3.5.

The role of dissolved oxygen

As it can be seen in the previous sections, the presence of oxygen significantly affects both the rate and the pathways of the degradation of BS. Oxygen increases the initial formation rate of the meta- and para-hydroxylated derivatives, and dramatically (at least with one order of magnitude) enhances the formation rate of the dihydroxy species (see Figs. 6 and 7). This phenomenon can be partly attributed to the electronscavenging effect of dissolved oxygen, by which it hinders the recombination of the electronehole pair on the surface of the photocatalyst (Schwitzgebel et al., 1995; Dionysiou et al., 2002). Thus, it can accelerate the formation of hydroxyl radicals via   oxidation of H2O or HO with hþ vb (see Eq. (2)). Besides, O2 formed via electron-scavenging by dissolved oxygen (see Eq. (3)) can also take part in oxidation of both the starting material and the intermediates in this system. (Notably, O2 2 , which can also be formed in the presence of oxygen (see Eq. (4)), cannot directly oxidize benzenesulfonate in a thermal reaction, as indicated by our blind probes.) For the reaction with a conduction band electron, adsorption of oxygen on the surface of the catalyst particles is an important step. Since the holes react much faster with organic species in the solution or solvent molecules than electrons do with oxygen, the rate of oxygen reduction by the conduction band electron is usually the rate-limiting step in the photocatalytic process. Accordingly, the Langmuir-type of tendency of the photocatalytic rates on oxygen concentration, also called as Langmuir-Hinselwood (L-H) mechanism has been observed (Dionysiou et al., 2002). Although oxygen appears to

1625

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 1 7 e1 6 2 8

SO3

SO 3

SO3

SO 3

OH

OH

OH

HO

OH

173

157 SO 3 -

OH HO

COOCOOH

HO

HO

HO

209

OH

HO

HO

COOH

HO

COOH

CHO

273

HO

123

207

COOH CHO

193

OH

SO42-, CO2, H2O

O

OH

OH

SO 3 -

243 O

239

OH

COOH

O

-

COOH

179 OH

SO3

CHO

COOH

HO

-

COOH

SO 3 -

COO-

OH

SO3

205

COOH

OH

OH

OH

189

O

89

Scheme 1 e A possible degradation pathway of benzenesulfonate in photocatalytic oxidation. The numbers indicate the m/z values of the corresponding species.

be weakly adsorbed on TiO2, it does not compete with the organic contaminants for adsorption sites because it adsorbs at Ti3þ sites, whereas hydroxide ions and organic pollutants adsorb at Ti4þ-lattice oxygen sites (Mills et al., 2006). In the aerated system the decay of the hydroxy and dihydroxy derivatives is strongly promoted via further oxidation steps leading to ring cleavage. On the contrary, no ring cleavage could be achieved in argon-saturated reaction mixture. We also tried to achieve ring fission in the case of pyrogallol (benzene-1,2,3-triol) in the same type of deaerated photocatalytic system, but its aromatic ring could not be cleaved in spite of the trihydroxylated structure. In the aerated system, however, it can be easily degraded, even without photocatalysis as it is well-known (Marklund and Marklund, 1974). These observations clearly indicate that the presence of oxygen is indispensable for the cleavage of the aromatic ring. They also suggest that hydroxyl radicals alone are not able to cleave the aromatic ring. This can be achieved by the attack of other oxidizing agents formed in situ from dissolved oxygen,   such as for example O2 /HO2 and/or 1O2 (singlet excited-state dioxygen). The superoxide radical anion can react via a single electron transfer mechanism, while 1O2 via the formation of the dioxetane intermediate (Wahab et al., 2008). Although transient intermediates cannot be detected by the HPLC/MS technique, it is well-known that the reaction of

hydroxyl radicals with organic substrates generally leads to the formation of carbon-centered radicals (Cooper et al., 2009). In aerated solution these species can react with dissolved oxygen to form peroxyl radicals, the fate of which in water is very complex. In several cases they undergo self-reaction to form tetroxide intermediates, which decompose to form a variety of products. In order to demonstrate the crucial role of dissolved oxygen in the photocatalytic degradation of BS in this system irradiation was accompanied by alternating bubbling with argon and air. Monitoring the change of TOC and pH as the functions of irradiation time clearly shows the importance of O2 for the mineralization process (Fig. 10). During the first (120-min) irradiation period in the argonsaturated system the pH significantly decreased, due to the reaction between the sulfonate groups and the hydroxyl radicals (see Eq. (5)), while no change of TOC was observed, indicating that no mineralization, i.e., formation of CO2 or HCO 3 occurred. Introducing air into the system resulted in a further but gradually slower decrease of pH, reaching an almost constant value at the end of this 180-min period. This can be attributed to the total desulfonation of BS. In the presence of dissolved oxygen TOC decreased relatively fast as a consequence of efficient mineralization steps involving ring cleavage. Changing back to argon, the decrease of TOC

1626

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 1 7 e1 6 2 8

60

7

detected by mass spectrometry, a tentative scheme including possible pathways of degradation was also compiled. Our results can contribute to the development of photocatalytic procedures for wastewater treatment.

6 40 30

5

pH

TOC /mg dm-3

50

20 4 10 0 0

120

240

360

3 480

Irradiation time/min Fig. 10 e Change of total organic carbon (TOC) content (C, B) and pH (:, 6) during the photocatalytic treatment of a system the concentrations of which are given at Fig. 1. Bubbling with argon and air were alternated. Full symbols stand for the argon-bubbling periods, while contour symbols for the air-bubbling periods.

became slower due to the decreasing concentration and the final absence of dissolved oxygen. The slight decrease of TOC in the second half of this 120-min period of argon-bubbling can be attributed to the mineralization of open-chain intermediates formed in the previous stage. Finally, air was led into the system again, instantly causing a further, significantly fast decrease of TOC. Thus, the crucial role of dissolved oxygen in this photocatalytic mineralization could be unambiguously observed through this experiment.

4.

Conclusions

According to our experimental results gained by various analytical methods, in the degradation of BS the initial step is the hydroxylation of the starting compound, leading to the production of hydroxy then dihydroxy derivatives, along with desulfonation in both the presence and the absence of dissolved oxygen. HPLC-MS measurements indicated that the formation of ortho- and meta-hydroxy derivatives is favored in this system, which is in accordance with our earlier quantum chemical calculations. As a consequence, and also in agreement with theoretical considerations, 2,5-dihydroxybenzenesulfonate is the predominant intermediate formed in the second step. While in the argon-saturated system no decay of the hydroxy derivatives was observed, in the presence of oxygen, gradual oxidation steps resulted in the destruction of the aromatic system, followed by ring cleavage producing aldehydes and carboxylic acids. Efficient desulfonation took place both before and after the ring opening. The importance of dissolved oxygen for the ring fission, and, thus, for the complete mineralization, has been unequivocally   proved in this system, suggesting the crucial role of O2 /HO2 and/or 1O2 in this process. Considering the intermediates

Acknowledgments This work was supported by the National Development ´ MOP 4.2.2.-08/1/2008-0018, Livable environment Agency (TA and healthier people e Bioinnovation and Green Technology research at the University of Pannonia, the project is being cofinanced by the European Social Fund with the support of the European Union). The authors thank Prof. Pe´ter Hajo´s and Dr. Krisztia´n Horva´th for the determination of sulfate and sulfite with ion chromatography.

references

Alonso, M.C., Castillo, M., Barcelo, D., 1999. Solid-phase extraction procedure of polar benzene- and naphthalenesulfonates in industrial effluents followed by unequivocal determination with ion-pair chromatography/electrospray-mass spectrometry. Analytical Chemistry 71 (13), 2586e2593. Alonso, M.C., Tirapu, L., Ginebreda, A., Barcelo´, D., 2005. Monitoring and toxicity of sulfonated derivatives of benzene and naphthalene in municipal sewage treatment plants. Environmental Pollution 137 (2), 253e262. Amano, R., Tomizawa, S., Tezuka, M., 2004. Mineralization of aqueous benzenesulfonates by contact glow discharge electrolysis. Electrochemistry 72 (12), 836e838. Amano, R., Tezuka, M., 2006. Mineralization of alkylbenzenesulfonates in water by means of contact glow discharge electrolysis. Water Research 40 (9), 1857e1863. Brilon, C., Beckmann, W., Knackmuss, H.-J., 1981. Catabolism of naphthalenesulfonic acids by Pseudomonas sp. A3 and Pseudomonas sp. C22. Applied and Environmental Microbiology 42 (1), 44e55. Buxton, G.V., Greenstock, C.L., Helman, W.P., Ross, A.B., 1988. Critical review of rate constants for reactions of hydrated electrons, hydrogen atoms and hydroxyl radicals ($OH/$O ) in aqueous solution. Journal of Physical Chemistry Reference Data 17 (2), 513e886. Cain, R.B., 1981. In: Leisinger, T., Cook, A.M., Hutter, R., Nuesch, J. (Eds.), Microbial Metabolism of Xenobiotics and Recalcitrant Compounds. Academic Press, London, pp. 325e370. Cooper, W.J., Cramer, C.J., Martin, N.H., Mezyk, S.P., O’Shea, K.E., von Sonntag, C., 2009. Free radical mechanisms for the treatment of methyl tert-butyl ether (MTBE) via advanced oxidation/reductive processes in aqueous solutions. Chemical Reviews 109 (3), 1302e1345. Devipriya, S., Yesodharan, S., 2005. Photocatalytic degradation of pesticide contaminants in water. Solar Energy Matererials and Solar Cells 86 (3), 309e348. Dionysiou, D.D., Burbano, A.A., Suidan, M.T., Baudin, I., Laıˆne´, J.-M., 2002. Effect of oxygen in a thin-film rotating disk photocatalytic reactor. Environmental Science and Technology 36 (17), 3834e3843. http://pubs.acs.org/action/ showMultipleAbstracts-es0113605AF1#es0113605AF1. Fabbri, D., Bianco Prevot, A., Pramauro, E., 2006. Effect of surfactant microstructures on photocatalytic degradation of phenol and chlorophenols. Applied Catalysis B: Environmental 62 (1e2), 21e27.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 1 7 e1 6 2 8

´ rfa˜o, J.J.M., Pereira, M.F.R., 2008. Catalytic Faria, P.C.C., O ozonation of sulfonated aromatic compounds in the presence of activated carbon. Applied Catalysis B: Environmental 83 (1e2), 150e159. Gkika, E., Troupis, A., Hiskia, A., Papaconstantinou, E., 2006. Photocatalytic reduction of chromium and oxidation of organics by polyoxometalates. Applied Catalysis B: Environmental 62 (1e2), 28e34. Greim, H., Ahlers, J., Bias, R., Broecker, B., Hollander, H., Gelbke, H. P., Klimisch, H.J., Mangelsdorf, I., Paetz, A., Schong, N., Stropp, G., Vogel, R., Weber, C., Ziegler-Skylakakis, K., Bayer, E., 1994. Toxicity and ecotoxicity of sulfonic acids: structure activity relationship. Chemosphere 28 (12), 2203e2236. Hashim, M.A., Kulandai, J., Hassan, R.S., 1992. Biodegradability of branched alkybenzene sulfonates. Journal of Chemical Technology and Biotechnology 54 (3), 207e214. Hegyi, J., Horva´th, O., 2004. Photocatalytic reduction of mercury (II) and simultaneous oxidative degradation of surfactants in titanium dioxide suspension. Progress in Colloid and Polymer Science 125, 10e16. Hoffmann, M.R., Martin, S.T., Choi, W., Bahnemann, D.W., 1995. Environmental applications of semiconductor photocatalysis. Chemical Reviews 95 (1), 69e96. Horva´th, K., Hajo´s, P., 2006. Retention profiles and mechanism of anion separation on latex based pellicular ion exchanger in ion chromatography. Journal of Chromatography, A 1104 (1e2), 75e81. Horva´th, O., Husza´nk, R., 2003. Degradation of surfactants by hydroxyl radicals photogenerated from hydroxoiron(III) complexes. Photochemical and Photobiological Sciences 2 (9), 960e966. Horva´th, O., Bodna´r, E., Hegyi, J., 2005. Photoassisted oxidative degradation of surfactants and simultaneous reduction of metals in titanium dioxide dispersions. Colloids and Surfaces A: Physicochemical and Engineering Aspects 265 (1e3), 135e140. Integral Consulting Inc., 2007. Toxicological Profiles for Three Organic Acids. ICI, Annapolis, MD, pp. 3e1 e 3e4. Kajitvichyanukul, P., Ananpattarachai, J., Pongpom, S., 2005. Solegel preparation and properties study of TiO2 thin film for photocatalytic reduction of chromium(VI) in photocatalysis process. Science and Technology of Advanced Materials 6 (3e4), 352e358. Kanki, T., Yoneda, H., Sano, N., Toyoda, A., Nagai, C., 2004. Photocatalytic reduction and deposition of metallic ions in aqueous phase. Chemical Engineering Journal 97 (1), 77e81. Kirk, A.D., Namasivayam, C., 1983. Errors in ferrioxalate actinometry. Analytical Chemistry 55 (14), 2428e2429. Konstantinou, K.I., Albanis, T.A., 2003. Photocatalytic transformation of pesticides in aqueous titanium dioxide suspensions using artificial and solar light: intermediates and degradation pathways. Applied Catalysis B: Environmental 42 (4), 319e335. Lake, R., 2010. Optimized RP-HPLC Method for Hydroxybenzoic Acids. Restek Pharmaceutical Article. http://www.restek.com/ aoi_pharm_A009.asp. Lange, F.T., Furrer, R., Brauch, H.J., 2000. Polar aromatic sulfonates and their relevance to waterworks. Nieuwegein, Association of River Waterworks e RIWA, 21e60. Marklund, S., Marklund, G., 1974. Involvement of the superoxide anion radical in the autoxidation of pyrogallol and a convenient assay for superoxide dismutase. European Journal of Biochemistry 47, 469e474. Matsushita, M., Tran, T.H., Nosaka, A.Y., Nosaka, Y., 2007. Photooxidation mechanism of L-alanine in TiO2 photocatalytic systems as studied by proton NMR spectroscopy. Catalysis Today 120 (2), 240e244. Mills, A., Wang, J., Ollis, D.F., 2006. Dependence of the kinetics of liquid-phase photocatalyzed reactionson oxygen

1627

concentration and light intensity. Journal of Catalysis 243 (1), 1e6. Palmisano, G., Addamo, M., Augugliaro, V., Caronna, T., Paola, A. D., Lopez, E.G., Loddo, V., Marci, G., Palmisano, L., Schiavello, M., 2007a. Selectivity of hydroxyl radical in the partial oxidation of aromatic compounds in heterogeneous photocatalysis. Catalysis Today 122 (1e2), 118e127. Palmisano, G., Loddo, V., Augugliaro, V., Palmisano, L., 2007b. Photocatalytic oxidation of nitrobenzene and phenylamine: pathways and kinetics. AIChE 53 (4), 961e968. Patsoura, A., Kondarides, D.I., Verykios, E.X., 2007. Photocatalytic degradation of organic pollutants with simultaneous production of hydrogen. Catalysis Today 124 (3e4), 94e102. Rabek, J.F., 1982. Experimental Methods in Photochemistry and Photophysics. Wiley-Interscience Publication, John Wiley & Sons Ltd., New York, pp. 944e6. Ravera, M., Ciccarelli, C., Gianotti, V., Scorza, S., Osella, D., 2004. Electro-assisted methods for waste destruction: silver(II) and peroxydisulfate reagents in the electrochemically mediated oxidation of polyaromatic sulfonates. Chemosphere 57 (7), 587e594. Ravera, M., Buico, A., Gosetti, F., Cassino, C., Musso, D., Osella, D., 2009. Oxidative degradation of 1,5-naphthalenedisulfonic acid in aqueous solutions by microwave irradiation in the presence of H2O2. Chemosphere 74 (10), 1309e1314. Ravera, M., Musso, D., Gosetti, F., Cassino, C., Gamalero, E., Osella, D., 2010. Oxidative degradation of 1,5naphthalenedisulfonic acid in aqueous solutions: noncatalytic remediation by UV-photolysis in the absence and presence of H2O2. Chemosphere 79 (2), 144e148. Reemtsma, T., 1996. Methods of analysis of polar aromatic sulfonates from aquatic environments. Journal of Chromatography A 733 (1e2), 473e489. Riediker, S., Suter, M.J.F., Giger, W., 2000a. Benzene- and naphthalenesulfonates in leachates and plumes of landfills. Water Research 34 (7), 2069e2079. Riediker, S., Ruckstuhl, S., Suter, M.J.F., Cook, A.M., Giger, W., 2000b. Toluenesulfonate in landfill leachates: leachability from foundry sands and aerobic biodegradation. Environmental Science and Technology 34 (11), 2156e2161. Sangchakr, B., Hisanaga, T., Tanaka, K., 1995. Photocatalytic degradation of sulfonated aromatics in aqueous TiO2 suspension. Journal of Photochemistry and Photobiology A: Chemistry 85 (1e2), 187e190. Schwitzgebel, J., Ekerdt, J.G., Gerischer, H., Heller, A., 1995. Role of the oxygen molecule and of the photogenerated electron in TiO2-photocatalyzed air oxidation reactions. Journal of Physical Chemistry 99 (15), 5633e5638. Sto¨ffler, B., Luft, G., 1999. Oxidative degradation of p-toluenesulfonic acid using hydrogen peroxide. Chemosphere 38 (5), 1035e1047. Sun, B., Reddy, E.P., Smirniotis, P.G., 2005. Visible light Cr (VI) reduction and organic chemical oxidation by TiO2 photocatalysis. Environmental Science and Technology 39 (16), 6251e6259. Suter, M.J.F., Riediker, S., Giger, W., 1999. Selective determination of aromatic sulfonates in landfill leachates and groundwater using microbore liquid chromatography coupled with mass spectrometry. Analytical Chemistry 71 (4), 897e904. Szabo´-Ba´rdos, E., Czili, H., Horva´th, A., 2003. Photocatalytic oxidation of oxalic acid enhanced by silver deposition on a TiO2 surface. Journal of Photochemistry and Photobiology A: Chemistry 154 (2e3), 195e201. Szabo´-Ba´rdos, E., Czili, H., Megyery-Balog, K., Horva´th, A., 2004. Photocatalytic oxidation of oxalic acid enhanced by silver and copper deposition on TiO2 surface. Progress in Colloid and Polymer Science 125, 42e48.

1628

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 1 7 e1 6 2 8

Szabo´-Ba´rdos, E., Pe´terva´ri, E., El-Zein, V., Horva´th, A., 2006. Photocatalytic decomposition of aspartic acid over bare and silver deposited TiO2. Journal of Photochemistry and Photobiology A: Chemistry 184 (1e2), 221e227. Szabo´-Ba´rdos, E., Zsila´k, Z., Lendvay, G., Horva´th, O., Markovics, O., } , N., 2008. Photocatalytic degradation of Hoffer, A., To¨ro 1,5-naphthalenedisulfonate on colloidal titanium dioxide. Journal of Physical Chemistry B 112 (46), 14500e14508. Takada, H., Ishiwatari, R., 1990. Biodegradation experiments of linear alkylbenzenes (LABs): isomeric composition LABs as an indicator of the degree of LAB degradation in the aquatic environment. Environmental Science and Technology 24 (1), 86e91. Tully, P.S., 1997. In: Grant, M.H. (Ed.)Kirk-Othmer Encyclopedia of Chemical Technology, fourth ed., vol. 23. JohnWiley & Sons, New York, p. 194. Wahab, H.S., Bredow, T., Aliwiand, S.M., 2008. Computational investigation of the adsorption and photocleavage of chlorobenzene on anatase TiO2 surfaces. Chemical Physics

353 (1e3), 93e103. Wittich, R., Rast, H.G., Knackmuss, H.-J., 1988. Degradation of naphthalene-2,6- and naphthalene-1,6-disulfonic acid by a Moraxella sp. Applied and Environmental Microbiology 54 (7), 1842e1844. Zerbinati, O., Diana, I., Baiocchi, C., 1999. Trace naphthalenesulfonates determination in natural water sample. International Journal of Environmental Analytical Chemistry 74 (1e4), 43e54. Zhang, T., Oyama, T., Horikoshi, S., Zhao, J., Serpone, N., Hidaka, H., 2003. Photocatalytic decomposition of the sodium dodecylbenzene sulfonate surfactant in aqueous titania suspensions exposed to highly concentrated solar radiation and effects of additives. Applied Catalysis B: Environmental 42 (1), 13e24. Zu¨rrer, D., Cook, A.M., Leisinger, Th., 1987. Microbial desulfonation of substituted naphthalenesulfonic acids and benzenesulfonic acids. Applied and Environmental Microbiology 53 (7), 1459e1463.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 2 9 e1 6 3 8

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Accelerated biodegradation of pyrene and benzo[a]pyrene in the Phragmites australis rhizosphere by bacteriaeroot exudate interactions Tadashi Toyama a,*, Tetsuya Furukawa b, Noritaka Maeda c, Daisuke Inoue b, Kazunari Sei b, Kazuhiro Mori a, Shintaro Kikuchi c, Michihiko Ike b a

Department of Research, Interdisciplinary Graduate School of Medicine and Engineering, University of Yamanashi, 4-3-11 Takeda, Kofu, Yamanashi 400-8511, Japan b Division of Sustainable Energy and Environmental Engineering, Osaka University, 2-1 Yamadaoka, Suita, Osaka 565-0871, Japan c Division of Applied Sciences, Muroran Institute of Technology, 27-1 Mizumoto, Muroran 050-8585, Japan

article info

abstract

Article history:

We investigated the biodegradation of pyrene and benzo[a]pyrene in Phragmites australis

Received 12 January 2010

rhizosphere sediment. We collected P. australis plants, rhizosphere sediments, and unve-

Received in revised form

getated sediments from natural aquatic sites and conducted degradation experiments

6 September 2010

using sediments spiked with pyrene or benzo[a]pyrene. Accelerated removal of pyrene and

Accepted 30 November 2010

benzo[a]pyrene was observed in P. australis rhizosphere sediments with plants, whereas

Available online 15 December 2010

both compounds persisted in unvegetated sediments without plants and in autoclaved rhizosphere sediments with sterilized plants, suggesting that the accelerated removal

Keywords:

resulted largely from biodegradation by rhizosphere bacteria. Initial densities of pyrene-

Phragmites australis

utilizing bacteria were substantially higher in the rhizosphere than in unvegetated sedi-

Rhizosphere

ments, but benzo[a]pyrene-utilizing bacteria were not detected in rhizosphere sediments.

Pyrene

Mycobacterium gilvum strains isolated from rhizosphere sediments utilized pyrene aerobi-

Benzo[a]pyrene

cally as a sole carbon source and were able to degrade benzo[a]pyrene when induced with

Accelerated biodegradation

pyrene. Phragmites australis root exudates containing phenolic compounds supported

Root exudates

growth as a carbon source for the one Mycobacterium strain tested, and induced benzo[a] pyrene-degrading activity of the strain. The stimulatory effect on benzo[a]pyrene biodegradation and the amounts of phenolic compounds in root exudates increased when P. australis was exposed to pyrene. Our results show that Mycobacteriumeroot exudate interactions can accelerate biodegradation of pyrene and benzo[a]pyrene in P. australis rhizosphere sediments. ª 2010 Elsevier Ltd. All rights reserved.

1.

Introduction

Polycyclic aromatic hydrocarbons (PAHs) are ubiquitous pollutants in aquatic environments. PAHs are generated from anthropogenic activities such as industrial processing,

spillage of petroleum, and incomplete combustion of fossil fuels. In aquatic environments, PAHs accumulate in sediments because of their hydrophobicity (Juhasz and Naidu, 2000; Chen et al., 2006; Liang et al., 2007). In particular, high molecular weight (HMW) PAHs, such as pyrene and benzo[a]

* Corresponding author. Tel./fax: þ81 55 220 8346. E-mail address: [email protected] (T. Toyama). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.11.044

1630

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 2 9 e1 6 3 8

pyrene, are resistant to biodegradation, persist in the environment, and show toxic, mutagenic, and carcinogenic effects in humans and in animals in the wild (Cerniglia, 1992; Menzie et al., 1992; Juhasz and Naidu, 2000; Kanaly and Harayama, 2000). Thus, there is an acute need for removal of HMWPAHs from contaminated sediments to reduce HMW-PAH risks to human health and the aquatic environment. Phytoremediationdthe use of plants to enhance biodegradation and removal of pollutantsdis a cost-effective and environmentally friendly remediation technology. Previous studies have revealed that various types of terrestrial plants promote HMW-PAH degradation in soils by stimulating bacterial metabolism (Aprill and Sims, 1990; Reilley et al., 1996; Banks et al., 1999; Liste and Alexander, 2000; Yoshitomi and Shann, 2001; Mueller and Shann, 2007; Yi and Crowley, 2007; Fan et al., 2008). However, little is known about the potential of aquatic plants to promote the biodegradation of HMW-PAHs in sediments. There is only one report (Jouanneau et al., 2005) of the accelerated biodegradation of pyrene in the rhizosphere sediment of the aquatic plant Phragmites australis (reed). Because efficient and rapid PAH biodegradation in the environment generally depends on molecular oxygen availability (Cerniglia, 1992; Chung and King, 1999; Juhasz and Naidu, 2000), the accelerated biodegradation of HMW-PAHs in the rhizosphere can be attributed to the release of oxygen to the rhizosphere by plant roots (rhizosphere oxygenation), and it has been confirmed that oxygen released by P. australis roots promotes pyrene biodegradation in the rhizosphere (Jouanneau et al., 2005). Another possible important mechanism for the accelerated biodegradation of HMW-PAHs in the rhizosphere is through stimulation of bacterial metabolism by plant root exudates. Previous studies have clearly shown that the root exudates of some terrestrial plants stimulate biodegradation of organic chemicals, including phenanthrene (Miya and Firestone, 2000, 2001), pyrene (Yoshitomi and Shann, 2001; Mueller and Shann, 2007; Yi and Crowley, 2007), benzo[a]pyrene (Rentz et al., 2005; Yi and Crowley, 2007), and polychlorinated biphenyls (PCBs) (Donnelly et al., 1994; Gilbert and Crowley, 1997; Leigh et al., 2002). Some phenolic compounds released by plants might enhance biodegradation of PCBs by serving as growth substrates for PCB-degrading bacteria (Donnelly et al., 1994; Leigh et al., 2002) and inducers of PCB-degrading enzymes (Gilbert and Crowley, 1997). We have recently found that the root exudates of another aquatic plant, Spirodela polyrrhiza (giant duckweed), contain highly concentrated and diverse phenolic compounds that contribute markedly to the accelerated biodegradation of simple aromatic compounds in the rhizosphere (Toyama et al., 2009b), suggesting that the exudate-mediated enhancement of biodegradation of chemicals is possible not only with terrestrial plants but also with aquatic plants. However, the effects of aquatic plant root exudates on HMWPAH biodegradation have not been investigated. Identification of the mechanisms underlying the accelerated biodegradation of HMW-PAHs in the rhizosphere of aquatic plantsdespecially the relationship between the accelerated biodegradation and phenolic root exudates of aquatic plantsdis essential for developing efficient use of aquatic plants for remediation of HMW-PAH-contaminated sediments.

Our objectives in this study were (i) to verify the accelerated biodegradation of pyrene and benzo[a]pyrene in natural rhizosphere sediment of P. australis and (ii) to clarify the mechanisms of this accelerated biodegradation. We followed the degradation of pyrene and benzo[a]pyrene in rhizosphere and non-rhizosphere (i.e., unvegetated) sediments spiked with pyrene or benzo[a]pyrene. We isolated pyrene-degrading Mycobacterium gilvum strains from the P. australis rhizosphere sediments, analyzed the chemical nature of P. australis root exudates, and investigated the effects of the root exudates on the biodegradation of pyrene and benzo[a]pyrene. Our results provide direct experimental evidence that P. australis root exudates stimulate biodegradation of pyrene and benzo[a] pyrene by pyrene-degrading Mycobacterium in the rhizosphere.

2.

Materials and methods

2.1.

Chemicals

Pyrene was purchased from Wako Pure Chemical Industries (Osaka, Japan). Benzo[a]pyrene was purchased from AccuStandard Inc. (New Haven, Connecticut, USA). N,O-bis(trimethylsilyl)trifluoroacetamide (BSTFA) was purchased from Tokyo Chemical Industry (Tokyo, Japan).

2.2.

Plant, sediment, and water samples

Phragmites australis plant samples and sediment and surfacewater samples were obtained from 3 freshwater sites in Japan: Inukai Pond (Yamadaoka, Suita) and Yodo River (Juso, Osaka) in Osaka prefecture, and Kamo River (Shijokawaramachi, Kyoto) in Kyoto prefecture. Young P. australis plants (4e6 leaves, 40e60 cm tall) were collected from within the P. australis vegetation zone, and rhizosphere sediment samples were collected from a depth of 0e20 cm around their roots. Sediment samples from unvegetated areas (“unvegetated sediment” samples) were collected at the same depth, but at least 1 m from the nearest P. australis plant. The unvegetated sediment samples did not have root materials. Water samples were collected from the surface around the plants. To obtain sterile (bacteria-free) plants, P. australis seeds were sterilized by a 1-min wash in 70% ethanol and a 5-min wash in sodium hypochlorite solution (5% available chorine), rinsed twice with autoclaved deionized water, and germinated on sterile Hoagland’s nutrient medium (Toyama et al., 2006) solidified with 0.25% (w:v) gellan gum. Each young plant was aseptically transferred to a flask containing 200 mL of sterile Hoagland’s nutrient solution and maintained in an incubation chamber (30  C, 10,000 lx, 16:8 h light:dark).

2.3.

Culture media

Basal salts medium (BSM) (Toyama et al., 2009a) containing pyrene (PYR-BSM) or benzo[a]pyrene (BaP-BSM) as the sole carbon source was used for bacterial culture and degradation tests. Bacterial isolates were routinely maintained on a 1/10 dilution of Luria-Bertani (LB) medium. Agar solid medium was prepared with 2.0% (w:v) agar.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 2 9 e1 6 3 8

1631

2.4. Pyrene and benzo[a]pyrene degradation experiments in sediment microcosms

2.5. Enrichment, isolation, and identification of pyrenedegrading bacteria

To check for potential HMW-PAH degradation in the natural rhizosphere sediments of P. australis, degradation experiments were performed in microcosms as follows. Sediment samples (equal to about 100 g dry-weight) were placed in 150mL vials. Pyrene or benzo[a]pyrene dissolved in n-hexane was added to each vial to a final concentration of 50 mg (kg dry sediment)1, and the n-hexane was allowed to evaporate. The vial was then shaken for 3 h at room temperature. Next, a P. australis plant was replanted in a vial containing rhizosphere sediment from the same site as the plant. Finally, 10 mL of surface-water from the same site as the sediment and plant was added to the vial. Sterile control experiments using autoclaved sediment samples (121  C, 20 min; 3 replicates), 2-month-old sterile P. australis plants (4e6 leaves, about 30 cm tall), and autoclaved surface-water samples were also performed, to evaluate adsorption and uptake of HMW-PAHs by the plants. In total, 6 treatments were performed for each of 3 freshwater sites: (i) rhizosphere sediment with plant and pyrene, (ii) rhizosphere sediment with plant and benzo[a]pyrene, (iii) unvegetated sediment without plant and with pyrene, (iv) unvegetated sediment without plant and with benzo[a]pyrene, (v) autoclaved rhizosphere sediment with sterile plant and pyrene, and (vi) autoclaved rhizosphere sediment with sterile plant and benzo[a] pyrene. Nine identical microcosms were prepared for each treatment. All microcosms were incubated statically (30  C, 10,000 lx, 16:8 h light:dark). Three vials from each treatment were sampled at the start of the 28-d experimental period and at days 14 and 28 for analysis of pyrene and benzo[a]pyrene levels and determination of the numbers of pyrene and benzo[a] pyrene-degrading bacteria. The upper parts of plants (shoots and leaves) were separated from the roots, which were left in the sediment before extraction of PAHs. Both types of sediment (i.e., rhizosphere sediment with the roots and unvegetated sediment without roots) were acidified with 5 mL salting-out solution (1 N HCl, 30% [w:v] NaCl), shaken with 20 mL of 1:1 (v:v) dichloromethane:methanol at 300 rpm for 20 min, sonicated in an ultrasonic bath (20 kHz, 200 W, 5-s interval, 4  C) for 20 min, and shaken again for 20 min, after which the organic layer was collected. This extraction was performed two more times (total ¼ 3) for each sample. The extract was dried under flowing nitrogen, dissolved in 1 mL acetonitrile, and analyzed by highperformance liquid chromatography (HPLC) (see section 2.9 below). We have confirmed that almost all of the pyrene (more than 94%) and benzo[a]pyrene (more than 92%) was recovered from the pyrene and benzo[a]pyrene-amended sediments, respectively. The numbers of pyrene and benzo[a]pyrene-degrading bacteria in the microcosms were determined by most probable number (MPN) assay, as previously described (Miya and Firestone, 2000). Bacterial densities are expressed as MPN per gram of dry sediment. For the MPN assay, pyrene (100 mg L1) or benzo[a]pyrene (100 mg L1) was provided as the sole carbon source for pyrene or benzo[a]pyrene-degrading bacteria, respectively.

For enrichment of pyrene-degrading bacteria from the initial rhizosphere sediment samples (no benzo[a]pyrene-degrading bacteria were found), about 1 g wet-weight of rhizosphere sediment from each location was added to 100 mL of PYR-BSM (pyrene, 100 mg L1), and the mixture was incubated at 28  C on a rotary shaker at 120 rpm for 14 d. After confirmation of bacterial growth, 1 mL of each culture was transferred to fresh PYR-BSM (100 mg L1) and incubated for 14 d. After a third transfer, enrichment cultures were serially diluted and spread on BSM agar plates, and the plates were sprayed with a 1:1 (v:v) solution of n-hexane and acetone containing pyrene (about 10% [w:v]) and incubated at 28  C. Morphologically different colonies that produced clear zones were screened for their ability to degrade pyrene (Kiyohara et al., 1982). Subsequently, the isolated bacterial strainsddesignated IPF, KTM-1, KTM-2, and YTMdwere characterized and identified by physiological and phylogenetic analyses, as described previously (Inoue et al., 2008). The 16S rDNA sequence data of strains IPF, KTM-1, KTM-2, and YTM have been submitted to the DDBJ/EMBL/GenBank databases under accession numbers AB491971, AB491972, AB491973, and AB491974, respectively.

2.6. Pyrene and benzo[a]pyrene degradation tests using pure cultures of isolates Each isolate was grown in PYR-BSM (100 mg L1) for 5 d. Cells were harvested by centrifugation (9600  g, 4  C, 10 min), washed twice with 50 mM potassium phosphate buffer (pH 7.5), and inoculated to a final cell density (as determined by optical density at 600 nm [OD600]) of 0.02 (i.e., OD600 ¼ 0.02) in 30-mL vials containing 10 mL of PYR-BSM (100 mg L1). The cells (whole cells) were also inoculated at an OD600 of 0.2 in 30mL vials containing 10 mL of BaP-BSM (benzo[a]pyrene, 20 mg L1). We prepared 42 identical vial cultures for each test. Culture was carried out at 28  C and 120 rpm. Triplicate vials from each test were removed over the 5-d experimental period, and the cell density (OD600) and substrate concentrations (see section 2.9 below) were monitored. We also performed sterile control tests, using autoclaved (121  C, 20 min) bacterial cells of each isolate, and control tests, using the cells of each isolate pre-grown in 1/10 LB medium.

2.7.

Characterization of root exudates of P. australis

To examine changes in P. australis root exudates due to pyrene-induced stress, 2-month-old sterile plants (4e6 leaves, about 30 cm tall) were transferred to 50 mL sterile Hoagland’s nutrient solution with 1.0 mg L1 pyrene (“pyrene-exposed” plants) or without pyrene (“unexposed” plants) and grown in an incubation chamber (30  C, 10,000 lx, 16:8 h light:dark) for 1 d. Three plants were used for each of these two treatments. Each plant was then collected and gently washed twice in 50 mL sterile water for 3 min. Each plant was then transferred to a flask containing 50 mL sterile pure water and grown in the incubation chamber for 7 d. The root exudates in the bulkwater fraction and root-surface fraction were then collected

1632

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 2 9 e1 6 3 8

from each flask and analyzed separately to examine the distribution of root exudates in the rhizosphere. Fifty milliliters of water was directly collected as the bulk-water fraction. To obtain the root-surface fraction, the roots, separated from the plant, were transferred to 50 mL of sterile pure water and sonicated in an ultrasonic bath (20 kHz, 200 W, 5-s interval, 4  C) for 10 min, and the water was then collected as the rootsurface fraction. Both fractions were subjected to total organic carbon (TOC) and total phenolic compounds (TPC) quantification (Toyama et al., 2009b). Results are presented as the mean values of triplicate samples. The ability of P. australis to release organic and phenolic compounds was calculated as milligrams of TOC or TPC per gram of wet root material per day (mg TOC [g wet root]1 d1 or mg TPC [g wet root]1 d1). Furthermore, to qualitatively examine constituents in total root exudates, 100 mL combined volume of the bulk-water and root-surface root exudate fractions (1:1 [v:v] bulk-water and root-surface fractions) was analyzed by HPLC. For this analysis, the combined root exudate mixtures were concentrated by using an Oasis HLB polymeric cartridge (500 mg/6 mL, Waters, Massachusetts, USA) (Toyama et al., 2009b) and dissolved in 200 mL acetonitrile, and the sample was analyzed by HPLC (see section 2.9 below).

2.8. Pyrene and benzo[a]pyrene degradation tests in the presence or absence of root exudates Strain IPF was selected for additional testing for degradation of pyrene and benzo[a]pyrene in the presence or absence of root exudates. Strain IPF was grown in 1/10 LB medium for 2 d. The cells were harvested, washed, and added at an OD600 of 0.01e10 mL PYR-BSM (100 mg L1) or BaP-BSM (10 mg L1). Freeze-dried concentrate of the combined root exudate mixture (1:1 [v:v] bulk-water and root-surface fractions) from pyreneexposed plants or unexposed plants was added to the BSM before incubation to investigate the effect of root exudates on strain IPF (see Results section 3.5). Cells of strain IPF were also added to 1/10 LB medium containing pyrene (100 mg L1) or benzo[a]pyrene (10 mg L1) as control experiments. We prepared 27 identical vial cultures for each test. Culture was carried out in the dark at 28  C and 120 rpm. Triplicate vials from each test were removed over the 5-d experimental period and the cell density and concentrations of pyrene and benzo[a] pyrene were monitored (see section 2.9).

2.9. Analytical procedures for PAHs and phenolic compounds Pyrene and benzo[a]pyrene concentrations in sediment and cultures were determined by HPLC, and metabolites of benzo [a]pyrene were analyzed by gas chromatographyemass spectrometry (GCeMS) after HPLC separation. For HPLC and GCeMS analyses, the entire collected culture was acidified with 1 N HCl, shaken for 3 min with an equal volume of 1:1 (v:v) n-hexane:ethyl acetate, and centrifuged (9600  g, 4  C, 10 min). The organic layer was then collected. The extract was dried under flowing nitrogen, and the dry extract was dissolved in 1 mL acetonitrile and analyzed. HPLC analysis was conducted using a Shimadzu HPLC system with a UV/vis

detector and a Shim-pack VP-ODS column (150 mm  4.6 mm, particle size 5 mm; Shimadzu, Kyoto, Japan). The mobile phase for the HPLC analysis was 8:2 acetonitrile:water (v:v), with detection at a wavelength of 254 nm. GCeMS analysis was conducted using a Shimadzu GC/MS system (GCMS-QP2010) and an Rxi-5ms capillary column (30 m  0.25 mm ID, 1.00 mm df; Restek, Bellefonte, PA USA). For GCeMS analysis of metabolites from benzo[a]pyrene degradation, the fractions containing the metabolite peaks detected by HPLC were collected, dried, and trimethylsilylated with BSTFA-acetonitrile solution at 60  C for 1 h. For GCeMS analysis, the column temperature was maintained at 90  C for 1 min, increased to 150  C by 15  C min1, increased to 300  C by 5  C min1, and maintained at 300  C for 6 min. For HPLC analysis of the concentrated and re-dissolved root exudate mixtures, we applied a gradient elution starting with 20% acetonitrile for the first 3 min, after which the acetonitrile percentage was increased linearly to 100% over 27 min; this was followed by elution with 100% acetonitrile for 10 min and then with 20% acetonitrile for the final 5 min. The flow rate was 1.0 mL min1 at 40  C, and the detection wavelength was 254 nm. The HPLC analysis was conducted using a Shim-pack VP-ODS column (250 mm  4.6 mm, particle size 5 mm; Shimadzu). Results are presented as the mean values and 95% confidence intervals of triplicate experiments.

3.

Results

3.1. Degradation of pyrene and benzo[a]pyrene in P. australis rhizosphere and unvegetated sediments Detailed physicochemical analysis of the sediment samples showed them to have a pH of 7.0e7.3, low organic carbon content (ignition loss ¼ 1.0%e2.3%), and low levels of pyrene (0.0029e0.0079 mg [kg dry sediment]1) and benzo[a]pyrene (0e0.0023 mg [kg dry sediment]1). We conducted degradation experiments in both rhizosphere and unvegetated sediments spiked with pyrene (50 mg [kg dry sediment]1) or benzo[a]pyrene (50 mg [kg dry sediment]1). Pyrene and benzo[a]pyrene levels declined slightly or did not decline at all in the unvegetated sediments without plants; 2.99%e14.4% of the pyrene and 0%e14.5% of the benzo [a]pyrene disappeared from these sediments over 28 d (Fig. 1). In contrast, in rhizosphere sediments with plants, 50.2%e 61.3% of the pyrene and 40.7%e61.5% of the benzo[a]pyrene were removed from the sediments over 28 d (Fig. 1). In autoclaved sediments with sterile plants, there was no substantial decrease in pyrene or benzo[a]pyrene content over 28 d (Fig. 1). Initial densities of pyrene-degrading bacteria in the rhizosphere sediments from all 3 sites were 100e204 times those in the unvegetated sediments (Fig. 2). After the 28-d degradation experiments with Inukai Pond and Kamo River samples, the densities of pyrene-degrading bacteria in the rhizosphere sediments were 197 and 42 times, respectively, those in the unvegetated sediments. However, in the Yodo River samples, the densities of pyrene-degrading bacteria in the rhizosphere and unvegetated sediments at the end of the experiment were almost equal (Fig. 2). Although only 14.4% of the pyrene was

1633

Pyrene, Benzo[a]pyrene (mg kg–1)

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 2 9 e1 6 3 8

A

50

B

50

40

40

40

30

30

30

20

20

20

10

10

10

0

0

10

20

30

0

0

10

20

30

C

50

0

0

10

20

30

Time (days) Fig. 1 e Removal of pyrene and benzo[a]pyrene from pyrene or benzo[a]pyrene-spiked sediment microcosms from (A) Inukai Pond, (B) Yodo River, and (C) Kamo River, Japan. Closed symbols represent microcosms with rhizosphere sediment and P. australis plants; open symbols represent microcosms with unvegetated sediment and no P. australis. Dotted lines represent sterile control experiments using autoclaved sediment samples, 2-month-old sterile P. australis plants, and autoclaved surface-water samples. Squares and circles represent pyrene and benzo[a]pyrene concentrations, respectively. Data are means of triplicate experiments, and error bars indicate 95% confidence intervals.

removed from the unvegetated sediment of the Yodo River over the experimental period, the pyrene-degrading bacterial density increased substantially for still unknown reasons. We did not detect bacteria capable of utilizing benzo[a]pyrene as the sole carbon source in any of the sediment samples.

3.2. Isolation and identification of pyrene-degrading bacteria We enriched for pyrene-degrading bacteria from the initial rhizosphere sediment samples because they had been detected in these samples by the MPN assay. In total, 4 pyrenedegrading bacterial strains were isolated from the rhizosphere sediments. Strain IPF was isolated from the Inukai Pond sediment, strains KTM-1 and KTM-2 were isolated from the Kamo River sediment, and strain YTM was isolated from the Yodo River rhizosphere sediment. The 4 bacterial isolates were rod-shaped, gram- and catalase-positive, and oxidasenegative. They utilized glucose and D-mannitol as sole carbon sources, but not L-arabinose, D-mannose, N-acetyl-D-glucosamine, maltose, gluconate, n-caprate, adipate, D,L-malate, citrate, or phenylacetate. The partial 16S rRNA gene sequences of the 4 isolates showed the highest identity (99.3%) with Mycobacterium gilvum ATCC 43909T (accession no. X55599), although there were slight differences among the isolates. Thus, the 4 isolates were identified as M. gilvum.

3.3. Degradation of pyrene and benzo[a]pyrene by isolated strains The 4 isolated M. gilvum strains that were pre-grown on pyrene completely degraded 100 mg L1 pyrene within 3 d under aerobic conditions with concomitant cell growth (Fig. 3A); however, pyrene degradation and cell growth were not observed under anaerobic conditions (data not shown). No metabolites were detectable after complete pyrene disappearance. Although the 4 isolates did not use benzo[a]pyrene

for growth, whole cells pre-grown on pyrene (i.e., whole cells induced with pyrene) did degrade approximately 40% of 20 mg L1 benzo[a]pyrene within 5 d (Fig. 3B). Benzo[a]pyrene degradation was not observed in experiments using whole cells pre-grown in 1/10 LB medium (Supplementary information: Fig. S1). Three major metabolites of benzo[a]pyrene were detected by HPLC at retention times (RTs) of 2.3, 2.8, and 3.7 min, along with the decrease in benzo[a]pyrene with an RT of 11.6 min (Supplementary information: Fig. S2); these metabolites were identified by GCeMS. The trimethylsilyl (TMS)-derivatized metabolites were uniform in mass spectra (m/z of fragment ions [% relative intensity, characterization]): m/z 430 (50, Mþ), 415 (9, Mþ e CH3), 357 (10, Mþ e TMS), 341 (54, Mþ e OTMS), 284 (12, Mþ e 2TMS), 252 (16, Mþ e 2OTMS), 147 (43, adjacent TMS groups), and 73 (100, TMS). The spectral data were consistent with those of ditrimethylsilyl-benzo[a]pyrene dihydrodiol (Moody et al., 2004); therefore, the metabolites were identified as benzo[a]pyrene dihydrodiol isomers.

3.4.

Characterization of P. australis root exudates

We compared root exudates from plants with and without pyrene exposure. The TOC and TPC (mg [g wet root]1 d1) in root exudates released from sterile P. australis plants into bulk-water and root-surface fractions of the rhizosphere are shown in Table 1. Detailed chemical analyses by HPLC and GCeMS showed that the root exudates of pyrene-exposed plants had undetectable levels of pyrene (<0.001 mg L1). The percentage TPC in the root-surface fractions of both pyreneexposed and unexposed plants were about 4.0 and 14 times, respectively, those in the bulk-water fractions. In addition, the percentage TPC in the total root exudate (the combined root exudate mixture of bulk-water and root-surface fractions) of pyrene-exposed plants was about 4.0 times that of unexposed plants.

1634

8 7 6 5 4 3 2 1 0

0.30

A

120

0.25

100

0.20

80 0.15 60 0.10

40

0.05

20 0

28

days Inukai Pond

0

28

days Yodo River

0

0

28

days Kamo River

25

3.5. Degradation of pyrene and benzo[a]pyrene by M. gilvum strain IPF in the presence of root exudates Strain IPF was selected for further degradation studies because it exhibited the highest pyrene and benzo[a]pyrenedegrading abilities among the 4 isolates. We compared pyrene (100 mg L1) and benzo[a]pyrene (10 mg L1) degradation by strain IPF in the presence and absence of P. australis root exudates. We added the root exudate mixture from the pyrene-exposed plants (final concentration: 182 mg TOC L1; 7.62 mg TPC L1) or unexposed plants (final concentration: 161 mg TOC L1; 1.38 mg TPC L1) to BSM before incubation. Both pyrene degradation and cell growth were greater in the presence of root exudates from pyrene-exposed or unexposed plants than in the absence of these exudates (Fig. 5A). The rates and extents of pyrene degradation were similar, irrespective of the type of root exudate. In addition, cell growth in 1/10 LB medium without root exudates (control experiment) was greater than that in BSM without root exudates, but the rates and extents of pyrene degradation were similar in the two experiments (Fig. 5A). In the benzo[a]pyrene degradation experiments, neither cell growth nor benzo[a]pyrene degradation was observed in the absence of root exudates (Fig. 5B). However, the cell density increased in the presence of both types of root exudates, and 51% and 22% of the benzo[a]pyrene was degraded within 5 d in the presence of root exudates from pyrene-exposed or unexposed plants, respectively (Fig. 5B). The rates and extents of benzo[a]pyrene degradation were substantially higher in the presence of root exudates from

Benzo[a]pyrene (mg L–1)

Fig. 2 e Densities of pyrene-degrading bacteria in pyrenespiked rhizosphere sediment with plants (filled bars) and unvegetated sediment without plants (open bars) at the start and end of a 28-d degradation experiment. Data are means of triplicate experiments, and error bars indicate 95% confidence intervals.

HPLC profiles of the total root exudates differed for pyreneexposed and unexposed plants (Fig. 4). Nineteen distinct peaks with RTs between 10.0 and 40.0 min were detected in the root exudates of pyrene-exposed plants, whereas only 11 were detected in those of unexposed plants, indicating that there was a markedly higher number of constituents in the root exudates from pyrene-exposed plants.

0

1

2

3

4

5

2

3

4

5

Cell density (OD600)

140

Pyrene (mg L–1)

Pyrene-degrad in g bacterial d ensity (Log 10MP N g –1 sediment)

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 2 9 e1 6 3 8

0

B

20 15 10 5 0

0

1

Time (days) Fig. 3 e Pyrene and benzo[a]pyrene degradation by the isolated strains. (A) Pyrene degradation by, and cell growth of, M. glivum strains IPF (squares), KTM-1 (diamonds), KTM-2 (triangles), and YTM (circles), grown on pyrene in basal salts medium (BSM). (B) Benzo[a]pyrene degradation by whole cells of strains IPF (squares), KTM-1 (diamonds), KTM-2 (triangles), and YTM (circles), pre-grown on pyrene in BSM. Closed symbols represent concentrations of pyrene or benzo[a]pyrene, and open symbols in A represent cell densities. Dotted line in B represents data from sterile control tests. Data are means of triplicate experiments, and error bars indicate 95% confidence intervals.

pyrene-exposed plants than in the presence of root exudates from unexposed plants. Coincident with benzo[a]pyrene degradation, we detected benzo[a]pyrene dihydrodiol isomers in the presence of both root exudates. In the benzo [a]pyrene degradation control experiment in 1/10 LB medium without root exudates, we observed substantial cell growth but no substantial benzo[a]pyrene degradation (Fig. 5B).

4.

Discussion

We observed the accelerated removal of pyrene and benzo[a] pyrene in P. australis rhizosphere sediments containing plants, whereas both compounds persisted in unvegetated sediments without plants and in autoclaved rhizosphere sediments with sterile plants. Initial densities of pyrenedegrading bacteria were substantially higher in the

1635

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 2 9 e1 6 3 8

Table 1 e Characteristics of root exudates released by sterile (bacteria-free) P. australis plants into bulk-water and rootsurface fractions of the rhizosphere. Total root exudate release rate (mg [g wet root]1 d1)

Distribution of root exudates (mg [g wet root]1 d1) Bulk-water fraction TOCa

TPCb

Root-surface fraction

TPC/TOC (%)

Unexposed plants 44.5  1.48 0.0684  0.0062 Pyrene-exposed plants 47.8  3.91 0.592  0.102

TOC

TPC

TPC/TOC (%)

TOC

TPC

TPC/TOC (%)

2.10 5.02

69.4 108

0.591 3.63

0.85 3.35

24.9  2.90 0.523  0.112 60.6  5.02 3.04  0.60

0.15 1.24

a TOC, total organic carbon. Data are means  standard deviations of triplicate experiments. b TPC, total phenolic compounds. Data are means  standard deviations of triplicate experiments.

rhizosphere sediments than in the unvegetated sediments. These results suggest that the accelerated pyrene and benzo [a]pyrene removal in the P. australis rhizosphere sediments largely resulted from biodegradation by rhizosphere bacteria rather than from adsorption and uptake by plants. These findings were common to the natural rhizosphere sediments collected from 3 freshwater sites that had not been substantially contaminated with PAHs, supporting the conclusion that phytoremediation using P. australis can be useful for pyrene and benzo[a]pyrene removal from sediments. In natural environments, low molecular weight (LMW)PAHs, such as naphthalene and phenanthrene, are relatively easily biodegraded, whereas HMW-PAHs are persistent (Cerniglia, 1992; Juhasz and Naidu, 2000; Kanaly and Harayama, 2000). Even within the same environment, the bacterial group that degrades HMW-PAHs is different than the group that degrades LMW-PAHs (Zhou et al., 2008). Recently, a variety of naphthalene and phenanthrene-degrading bacteria (mostly Pseudomonas and Paenibacillus) have been isolated from hydrocarbon-contaminated rhizosphere sediments of salt marsh plants (Daane et al., 2001; Launen et al., 2008). However, HMW-PAH-degrading bacteria in the aquatic plant rhizosphere have not been investigated. Most pyrenedegrading bacteria previously reported were fast-growing

mycobacteria and were commonly isolated from HMW-PAHor oil-contaminated soils and sediments (Heitkamp and Cerniglia, 1988; Heitkamp et al., 1988; Schneider et al., 1996; Cheung and Kinkle, 2001; Dean-Ross et al., 2002; Habe et al., 2004; Miller et al., 2004; Kim et al., 2005). Pyrene-degrading mycobacteria were recently shown to be prevalent in many HMW-PAH-contaminated sediments, where they could play an important role in the natural attenuation of HMW-PAHs (DeBruyn et al., 2007, 2009). Here, we isolated pyrene-degrading bacteriadthe M. gilvum strains IPF, KTM-1, KTM-2, and YTMdfrom P. australis rhizosphere sediments. To our knowledge, our isolates are the first bacteria from the rhizosphere of aquatic plants capable of utilizing pyrene as a sole carbon source. Interestingly, the 4 isolates from the 3 different P. australis rhizosphere sediments are all M. gilvum. Our isolates also are fast-growing mycobacteria and might play a major role in pyrene degradation in the P. australis rhizosphere. Although our rhizosphere sediments were not contaminated with HMW-PAHs, the rates of pyrene degradation by our isolates were comparable to, or greater than, those by pyrene-degrading mycobacteria derived from contaminated soils (Heitkamp and Cerniglia, 1988; Heitkamp et al., 1988; Dean-Ross et al., 2002; Habe et al., 2004; Miller et al., 2004).

225000

In ten sity (

V)

200000 175000

3 12

150000

8 4 5

125000

7

9

6

14 10 11 13 12 15

100000

18

Pyrene-exposed plant

17

19

1

75000

16

50000 7 8

25000 0 0.0

16

5.0

10.0

15.0

20.0

9

25.0

14 11 13 15

30.0

Unexposed plant 17

19

35.0

40.0

Retention time (min) Fig. 4 e Constituents of P. australis root exudates. HPLC UV/vis chromatograms (detection at 254 nm) of constituents in root exudate mixtures (bulk-water D root-surface fractions; 1:1 [v:v]) from pyrene-exposed and non-exposed P. australis plants.

1636

80

Benzo[a]pyrene (mg L–1)

0.3

60 0.2 40 0.1

20 0

B

0.4

0

1

2

3

4

5

0 0.20

12 10

0.15 8 0.10

6 4

0.05 2 0

Cell density (OD600)

100

0

1

2

3

4

5

Cell density (OD600)

A Pyrene (mg L–1)

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 2 9 e1 6 3 8

0

Time (days) Fig. 5 e Effects of P. australis root exudates on pyrene and benzo[a]pyrene degradation. (A) Pyrene and (B) benzo[a] pyrene degradation by M. glivum strain IPF in BSM in the presence of root exudate mixtures from non-pyreneexposed plants (circles) or pyrene-exposed plants (squares), or in the absence of root exudates (triangles). Diamonds in A and B represent data from control experiments in 1/10 LB medium. Closed symbols represent concentrations of pyrene or benzo[a]pyrene, and open symbols represent cell densities. Data are means of triplicate experiments, and error bars indicate 95% confidence intervals.

The presence of these pyrene-degrading mycobacteria in the P. australis rhizosphere, even under non-PAH-contaminated conditions, is of some interest. Also, although we did not determine the abundances of pyrene-degrading bacteria relative to those of total heterotrophic bacteria in the sediments, the densities of pyrene-utilizing bacteria in the P. australis rhizosphere sediments were clearly higher than those in the unvegetated sediments. This increase in microbial numbers in rhizosphere sediments compared with those in non-rhizosphere sediments is known as the “rhizosphere effect” (Shaw and Burns, 2003). Although we observed accelerated benzo[a]pyrene removal in P. australis rhizosphere sediments, we did not detect benzo [a]pyrene-utilizing bacteria. However, because the isolated pyrene-degrading mycobacteria were able to degrade benzo[a] pyrene when induced with pyrene, they might also play an important role in benzo[a]pyrene degradation in the rhizosphere sediments. Benzo[a]pyrene degradation by our isolates required both growth substrate and an inducer of benzo[a]

pyrene-degrading enzymes, such as pyrene, as reported previously (Juhasz and Naidu, 2000; Rentz et al., 2005). We detected benzo[a]pyrene dihydrodiols as the benzo[a]pyrene metabolites in our isolate cultures, suggesting a benzo[a]pyrene degradation pathway similar to that proposed for Mycobacterium sp. strains PYR-1 (Moody et al., 2004) and RJGII-135 (Schneider et al., 1996). Rhizosphere oxygenation by the roots of P. australis undoubtedly contributed to the accelerated biodegradation of pyrene and benzo[a]pyrene in this study, as shown by the inability of the isolated mycobacteria to degrade pyrene under anaerobic conditions. However, the accelerated biodegradation of both pyrene and benzo[a]pyrene and the high diversity of pyrene-degrading bacteria in the rhizosphere sediments cannot be fully explained only by rhizosphere oxygenation. We suspected that the root exudates of P. australis were another key factor in these phenomena. We focused especially on phenolic root exudates, because they are known to contribute to the accelerated rhizosphere biodegradation of simple aromatic compounds (Toyama et al., 2009b) and PCBs (Gilbert and Crowley, 1997; Leigh et al., 2002). To test this hypothesis, we characterized the root exudates of P. australis. Two-month-old sterile P. australis plants released root exudates into the rhizosphere at the specific release rates of 69.4 mg TOC and 0.591 mg TPC (g wet root)1 d1, confirming that the root exudates of P. australis contain phenolic compounds. The percentage of TPC in the root-surface fraction was about 14 times that in the bulkwater fraction. This suggests that the phenolic compounds released by P. australis roots probably have the greatest effect on rhizosphere bacteria on the root-surface, and that this effect would decrease with increasing distance from the roots. It is well documented that most higher plants respond to various stresses by activating secondary metabolic pathways, such as the phenylpropanoid mechanism, and that production of phenolic compounds plays an important role in resistance to stresses (Hutzler et al., 1998; Bagniewska-Zadworna et al., 2008). A pollutant-induced stress in plants causes a change in both the quantity and quality of the root exudates (Porteous et al., 2000; Mingji et al., 2009). In this study, the amount of TPC and the diversity of constituents in the root exudates substantially increased when the roots of P. australis were exposed to pyrene. The amounts of TOC and TPC in the total root exudates of pyrene-exposed plants were about 1.6 and 6.2 times, respectively, those of unexposed plants, suggesting that exposure of the roots to pyrene affected the exudation of TPC to a greater degree than the exudation of TOC. Our experiments also showed that the root exudates of P. australis supported cell growth of the pyrene-degrading Mycobacterium strain IPF (the only strain tested) and stimulated the degradation of pyrene and benzo[a]pyrene by this strain. In particular, we found that the root exudates were essential for benzo[a]pyrene degradation by strain IPF. Thus, the root exudates undoubtedly functioned as a carbon source for growth and an inducer for benzo[a]pyrene-degrading activity in strain IPF. Another interesting finding is the increase in the stimulatory effect of the root exudates on benzo[a]pyrene biodegradation, but not on pyrene biodegradation, after the roots of P. australis were exposed to pyrene. That is, the root exudates of pyrene-stressed P. australis

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 2 9 e1 6 3 8

stimulated benzo[a]pyrene degradation to a greater degree than the root exudates of an unexposed plant, free from pyrene stress. The increase in the amount of phenolic compounds in the root exudates following pyrene stress provides supporting evidence that phenolic exudates are most probably involved in the stimulatory effect on benzo[a]pyrene degradation by the pyrene-degrading Mycobacterium strain IPF. Identification of the key compounds involved in the stimulatory effects could lead to the use of P. australis as a source of natural compounds for bacterial stimulation; these compounds could then be applied directly to bioremediation.

5.

Conclusions

We investigated the biodegradation of pyrene and benzo[a] pyrene in P. australis rhizosphere sediments. Our conclusions are summarized as follows: (1) Accelerated removal of pyrene and benzo[a]pyrene was observed in P. australis rhizosphere sediments containing plants, whereas both compounds persisted in unvegetated sediments without plants and in autoclaved rhizosphere sediments with sterilized plants. Accelerated removal of both compounds largely resulted from biodegradation by rhizosphere bacteria. (2) We isolated pyrene-degrading Mycobacterium gilvum strains from P. australis rhizosphere sediments. The strains utilized pyrene aerobically as a sole carbon and energy source and were able to degrade benzo[a]pyrene when induced with pyrene. (3) The root exudates of P. australis clearly supported cell growth of the pyrene-degrading Mycobacterium strain tested and stimulated benzo[a]pyrene degradation by the straindthat is, the induction of benzo[a]pyrene-degrading activity. This stimulatory effect and the amounts of phenolic compounds in the root exudates increased when P. australis roots were exposed to pyrene. It appears that phenolics are key candidates for explaining the stimulation of benzo[a]pyrene biodegradation in the P. australis rhizosphere. This is the first study demonstrating that P. australis root exudates stimulate biodegradation of pyrene and benzo[a] pyrene. We conclude that interactions between Mycobacterium spp. and root exudates can accelerate removal of pyrene and benzo[a]pyrene in the P. australis rhizosphere sediment. The synergetic effects of oxygen and exudates released by P. australis roots in accelerating the biodegradation of HMW-PAHs increase the potential usefulness of aquatic plants for remediation of HMW-PAH-contaminated sediments.

Acknowledgments This research was supported partly by a Grant-in-Aid for Encouragement of Young Scientists (A) (no. 21681010) and Young Scientists (B) (no. 19710060) from the Ministry of Education, Culture, Sports, Science, and Technology of Japan.

1637

Appendix. Supplementary data Supplementary data associated with this article can be found in online version at doi:10.1016/j.watres.2010.11.044.

references

Aprill, W., Sims, R., 1990. Evaluation of the use of prairie grasses for stimulating polycyclic aromatic hydrocarbon treatment in soil. Chemosphere 20, 253e265. Bagniewska-Zadworna, A., Zenkteler, E., Karolewski, P., Zadworny, M., 2008. Phenolic compound localization in Polypodium vulgare L. rhizomes after mannitol-induced dehydration and controlled desiccation. Plant Cell Rep. 27, 1251e1259. Banks, M.K., Lee, E., Schwab, A.P., 1999. Evaluation of dissipation mechanisms for benzo[a]pyrene in the rhizosphere of tall fescue. J. Environ. Qual. 28, 294e298. Cerniglia, C.E., 1992. Biodegradation of polycyclic aromatic hydrocarbons. Biodegradation 3, 351e368. Chen, S.J., Luo, X.J., Mai, B.X., Sheng, G.Y., Fu, J.M., Zeng, E.Y., 2006. Distribution and mass inventories of polycyclic aromatic hydrocarbons and organochlorine pesticides in sediments of the Pearl River Estuary and the Northern South China Sea. Environ. Sci. Technol. 40, 709e714. Cheung, P.-Y., Kinkle, B.K., 2001. Mycobacterium diversity and pyrene mineralization in petroleum-contaminated soils. Appl. Environ. Microbiol. 67, 2222e2229. Chung, W.K., King, G.M., 1999. Biogeochemical transformations and potential polyaromatic hydrocarbon degradation in macrofaunal burrow sediments. Aquat. Microb. Ecol. 19, 285e295. Daane, L.L., Harjono, I., Zylstra, G.J., Ha¨ggblom, M.M., 2001. Isolation and characterization of polycyclic aromatic hydrocarbon-degrading bacteria associated with the rhizosphere of salt marsh plants. Appl. Environ. Microbiol. 67, 2683e2691. Dean-Ross, D., Moody, J., Cerniglia, C.E., 2002. Utilization of mixtures of polycyclic aromatic hydrocarbons by bacteria isolated from contaminated sediment. FEMS Microbiol. Ecol. 41, 1e7. DeBruyn, J.M., Chewning, C.S., Sayler, G.S., 2007. Comparative quantitative prevalence of Mycobacteria and functionally abundant nidA, nahAc, and nagAc diosygenase genes in coal tar contaminated sediments. Environ. Sci. Technol. 41, 5426e5432. DeBruyn, J.M., Mead, T.J., Wilhelm, S.W., Sayler, G.S., 2009. PAH biodegradative genotypes in Lake Erie sediments: evidence for broad geographical distribution of pyrene-degrading mycobacteria. Environ. Sci. Technol. 43, 3467e3473. Donnelly, P.K., Hegde, R.S., Fletcher, J.S., 1994. Growth of PCBdegrading bacteria on compounds from photosynthetic plants. Chemosphere 28, 981e988. Fan, S., Li, P., Gong, Z., Ren, W., He, N., 2008. Promotion of pyrene degradation in rhizosphere of alfalfa (Medicago sativa L.). Chemosphere 71, 1593e1598. Gilbert, E.S., Crowley, D.E., 1997. Plant compounds that induce polychlorinated biphenyl biodegradation by Arthrobacter sp. strain B1B. Appl. Environ. Microbiol. 63, 1933e1938. Habe, H., Kanemitsu, M., Nomura, M., Takenura, T., Iwata, K., Nojiri, H., Yamane, H., 2004. Isolation and characterization of an alkaliphilic bacterium utilizing pyrene as a carbon source. J. Biosci. Bioeng. 98, 306e308. Heitkamp, M.A., Cerniglia, C.E., 1988. Mineralization of polycyclic aromatic hydrocarbons by a bacterium isolated from

1638

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 2 9 e1 6 3 8

sediment below an oil field. Appl. Environ. Microbiol. 54, 1612e1614. Heitkamp, M.A., Franklin, W., Cerniglia, C.E., 1988. Microbial metabolism of polycyclic aromatic hydrocarbons: isolation and characterization of a pyrene-degrading bacterium. Appl. Environ. Microbiol. 54, 2549e2555. Hutzler, P., Fischbach, R., Heller, W., Jungblut, T.P., Reuber, S., Schmitz, R., Veit, M., Weissenbo¨ck, G., Schnitzler, J.-P., 1998. Tissue localization of phenolic compounds in plants by confocal laser scanning microscopy. J. Exp. Bot. 49, 953e965. Inoue, D., Hara, S., Kashihara, M., Murai, Y., Danzl, E., Sei, K., Tsunoi, S., Fujita, M., Ike, M., 2008. Degradation of bis(4hydroxyphenyl)methane (bisphenol F) by Sphingobium yanoikuyae strain FM-2 isolated from river water. Appl. Environ. Microbiol. 74, 352e358. Jouanneau, Y., Willison, J.C., Meyer, C., Krivobok, S., Chevron, N., Besombes, J.-L., Blake, G., 2005. Stimulation of pyrene mineralization in freshwater sediments by bacterial and plant bioaugmentation. Environ. Sci. Technol. 39, 5729e5735. Juhasz, A.L., Naidu, R., 2000. Bioremediation of higher molecular weight polycyclic aromatic hydrocarbons: a review of the microbial degradation of benzo[a]pyrene. Int. Biodeterior. Biodegradation 45, 57e88. Kanaly, R.A., Harayama, S., 2000. Biodegradation of highmolecular-weight polycyclic aromatic hydrocarbons by bacteria. J. Bacteriol. 182, 2059e2067. Kim, Y.-H., Engesser, K.-H., Cerniglia, C.E., 2005. Numerical and genetic analysis of polycyclic aromatic hydrocarbondegrading mycobacteria. Microb. Ecol. 50, 110e119. Kiyohara, H., Nagao, K., Yana, K., 1982. Rapid screen for bacteria degrading water-insoluble, solid hydrocarbons on agar plates. Appl. Environ. Microbiol. 43, 454e457. Launen, L.A., Dutta, J., Turpeinen, R., Eastep, M.E., Dorn, R., Buggs, V.H., Leonard, J.W., Ha¨ggblom, M.M., 2008. Characterization of the indigenous PAH-degrading bacteria of Spartina dominated salt marshes in the New York/New Jersey Harbor. Biodegradation 19, 347e363. Leigh, M.B., Fletcher, J.S., Fu, X., Schmitz, F.J., 2002. Root turnover: an important source of microbial substrates in rhizosphere remediation of recalcitrant contaminants. Environ. Sci. Technol. 36, 1579e1583. Liang, Y., Tse, M.F., Young, L., Wong, M.H., 2007. Distribution patterns of polycyclic aromatic hydrocarbons (PAHs) in the sediments and fish at Mai Po marshes nature Reserve, Hong Kong. Water Res. 41, 1303e1311. Liste, H.-H., Alexander, M., 2000. Plant-promoted pyrene degradation in soil. Chemosphere 40, 7e10. Menzie, C.A., Potocki, B.B., Santodonato, J., 1992. Exposure to carcinogenic PAHs in the environment. Environ. Sci. Technol. 26, 1278e1284. Miller, C.D., Hall, K., Liang, Y.N., Nieman, K., Sorensen, D., Issa, B., Anderson, A.J., Sims, R.C., 2004. Isolation and characterization of polycyclic aromatic hydrocarbon-degrading Mycobacterium isolates from soil. Microb. Ecol. 48, 230e238. Mingji, X., Chongling, Y., Jing, Y., Lily, W., 2009. Impact of phenanthrene on organic acids secretion and accumulation by

perennial ryegrass, Lolium perenne L., root. Bull. Environ. Contam. Toxicol 83, 75e80. Miya, R.K., Firestone, M.K., 2000. Phenanthrene-degrader community dynamics in rhizosphere soil from a common annual grass. J. Environ. Qual. 29, 584e592. Miya, R.K., Firestone, M.K., 2001. Enhanced phenanthrene biodegradation in soil by slender oat root exudates and root debris. J. Environ. Qual. 30, 1911e1918. Moody, J.D., Freeman, J.P., Fu, P.P., Cerniglia, C.E., 2004. Degradation of benzo[a]pyrene by Mycobacterium vanbaalenii PYR-1. Appl. Environ. Microbiol. 70, 340e345. Mueller, K.E., Shann, J.R., 2007. Effects of tree root-derived substrates and inorganic nutrients on pyrene mineralization in rhizosphere and bulk soil. J. Environ. Qual. 36, 120e127. Porteous, F., Killham, K., Meharg, A., 2000. Use of a lux-marked rhizobacterium as a biosensor to assess changes in rhizosphere C flow due to pollutant stress. Chemosphere 41, 1549e1554. Reilley, K.A., Banks, M.K., Schwab, A.P., 1996. Dissipation of polycyclic aromatic hydrocarbons in the rhizosphere. J. Environ. Qual. 25, 212e219. Rentz, J.A., Alvarez, P.J.J., Schnoor, J.L., 2005. Benzo[a]pyrene cometabolism in the presence of plant root extracts and exudates: implications for phytoremediation. Environ. Pollut. 136, 477e484. Schneider, J., Grosser, R., Jayasimhulu, K., Xue, W., Warshawsky, D., 1996. Degradation of pyrene, benzo[a] anthracene, and benzo[a]pyrene by Mycobacterium sp. strain RJGII-135, isolated from a former coal gasification site. Appl. Environ. Microbiol. 62, 13e19. Shaw, L.J., Burns, R.G., 2003. Biodegradation of organic pollutants in the rhizosphere. Adv. Appl. Microbiol. 53, 1e60. Toyama, T., Yu, N., Kumada, H., Sei, K., Ike, M., Fujita, M., 2006. Accelerated aromatic compounds degradation in aquatic environment by use of interaction between Spirodela polyrrhiza and bacteria in its rhizosphere. J. Biosci. Bioeng. 101, 346e353. Toyama, T., Sato, Y., Inoue, D., Sei, K., Chang, Y.-C., Kikuchi, S., Ike, M., 2009a. Biodegradation of bisphenol A and bisphenol F in the rhizosphere sediment of Phragmites australis. J. Biosci. Bioeng. 108, 147e150. Toyama, T., Sei, K., Yu, N., Kumada, H., Inoue, D., Hoang, H., Soda, S., Chang, Y.-C., Kikuchi, S., Fujita, M., Ike, M., 2009b. Enrichment of bacteria possessing catechol dioxygenase genes in the rhizosphere of Spirodela polyrrhiza: a mechanism of accelerated biodegradation of phenol. Water Res. 43, 3765e3776. Yi, H., Crowley, D.E., 2007. Biostimulation of PAH degradation with plants containing high concentrations of linoleic acid. Environ. Sci. Technol. 41, 4382e4388. Yoshitomi, K.J., Shann, J.R., 2001. Corn (Zea mays L.) root exudates and their impact on 14C-pyrene mineralization. Soil Biol. Biochem. 33, 1769e1776. Zhou, H.W., Luan, T.G., Zou, F., Tam, N.F.Y., 2008. Different bacteria groups for biodegradation of three- and four-ring PAHs isolated from a Hong Kong mangrove sediment. J. Hazard. Mater. 152, 1179e1185.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 3 9 e1 6 5 0

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Comparison of MFI-UF constant pressure, MFI-UF constant flux and Crossflow Sampler-Modified Fouling Index Ultrafiltration (CFS-MFIUF) Lee Nuang Sim, Yun Ye, Vicki Chen*, Anthony G. Fane UNESCO Centre for Membrane Science and Technology, University of New South Wales, 2052 Sydney, NSW, Australia

article info

abstract

Article history:

Understanding the foulant deposition mechanism during crossflow filtration is critical in

Received 13 March 2010

developing indices to predict fouling propensity of feed water for reverse osmosis (RO).

Received in revised form

Factors affecting the performance on different fouling indices such as MFI-UF constant

30 November 2010

pressure, MFI-UF constant flux and newly proposed fouling index, CFS-MFIUF were inves-

Accepted 1 December 2010

tigated. Crossflow Sampler-Modified Fouling Index Ultrafiltration (CFS-MFIUF) utilises

Available online 9 December 2010

a typical crossflow unit to simulate the hydrodynamic conditions in the actual RO units followed by a dead-end unit to measure the fouling propensity of foulants. CFS-MFIUF was

Keywords:

found sensitive to crossflow velocity. The crossflow velocity in the crossflow sampler unit

Fouling index

influences the particle concentration and the particle size distribution in its permeate.

Colloidal fouling

CFS-MFIUF was also found sensitive to the permeate flux of both CFS and the dead-end cell.

Reverse osmosis

To closely simulate the hydrodynamic conditions of a crossflow RO unit, the flux used for

Crossflow filtration

CFS-MFIUF measurement was critical. The best option is to operate both the CFS and deadend permeate flux at flux which is normally operated at industry RO units (w20 L/m2 h), but this would prolong the test duration excessively. In this study, the dead-end flux was accelerated by reducing the dead-end membrane area while maintaining the CFS permeate flux at 20 L/m2 h. By doing so, a flux correction factor was investigated and applied to correlate the CFS-MFIUF measured at dead-end flux of 120 L/m2 h to CFS-MFIUF measured at dead-end flux of 20 L/m2 h for RO fouling rate prediction. Using this flux correction factor, the test duration of CFS-MFIUF can be shortened from 15 h to 2 h. ª 2010 Elsevier Ltd. All rights reserved.

1.

Introduction

Significant effort has been focused on developing a reliable and useful fouling index to assess fouling propensity of feed water prior to the reverse osmosis (RO) units. A fouling index can be used to predict how rapidly a given feed water will foul the RO system due to colloidal fouling. With this information in hand, appropriate pre-treatment schemes can then be selected prior to the RO system which can ultimately reduce colloidal fouling in RO.

There are several fouling indices such as the Silt Density Index (SDI), and the Modified Fouling Index (MFI0.45). Due to its simplicity and the short duration of measurement, SDI is currently the most widely used index in water industry. However, SDI has been proven unreliable by many researchers (Boerlage et al., 2003a; Lipp et al., 1990; Schippers and Verdouw, 1980; Yiantsios and Karabelas, 2002). SDI was found to have no relationship with foulant concentration and is not derived from any fouling mechanism. These downsides led to the emergence of MFI0.45 which is calculated based on cake filtration theory

* Corresponding author. Tel.: þ61 2 9385 4813; fax: þ61 2 9385 5966. E-mail address: [email protected] (V. Chen). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.12.001

1640

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 3 9 e1 6 5 0

Abbreviations 2

A filtration area (m ) solid concentration (g/L) Cb CFS-MFIUF Crossflow Sampler-Modified Fouling Index Ultrafiltration (s L2) standard reference pressure (207 kPa) DP0 DP transmembrane pressure (Pa) h circular channel diameter (m) I cake resistivity (m4) J flux (L m2 h1) cake mass per unit area (kg/m2) mc MFI-UFconst flux Modified Fouling Index-Ultrafiltration constant flux (s L2)

(Schippers and Verdouw, 1980). MFI0.45 has a linear relationship with the feed concentration, but the subsequent research carried out by Schippers et al. (1981) indicated that MFI0.45 was unable to account for small colloids (<0.45 mm). The Modified Fouling Index-Ultrafiltration (MFI-UF) and Nanofiltration-Modified Fouling Index (NF-MFI) were subsequently introduced, where ultrafiltration and nanofiltration membranes are used respectively as the test filters. MFI-UF was originally measured under constant pressure conditions. However, due to the prolonged duration of the test, Boerlage et al. (2004) proposed a MFI-UF which can be measured under constant flux mode. From their findings, the duration of the MFI-UF test could be shortened from 20 h to approximately 2 h if operated at constant flux of 75 L/m2 h using canal water. More recently, Choi et al. (2009) proposed a novel fouling index known as Combined Fouling Index (CFI) which uses a set of different membrane filters to determine the fouling potential of water. The authors suggested that no single method can be successfully used for accurate prediction of fouling potential of feed waters, but combination of different fouling indices using different types of test membranes such as hydrophobic MF, hydrophilic MF and hydrophilic UF may be possible as each of these membranes can capture different portions of foulants in a given feed. For example, hydrophobic MF is used to capture the fouling potential of hydrophobic foulants whereas hydrophilic UF is sensitive to the effects of colloidal matter and macromolecules on fouling. However, the capability of the fouling tests relied on their ability to capture the critical factor of feed water components which may contribute significantly to the fouling of RO. The above mentioned indices are carried out in a pressurized dead-end filtration cell, whereas in actual RO systems, crossflow filtration is the most widely chosen operating condition. These two operation modes have very different hydrodynamic conditions. In crossflow filtration, particles movement to and from the membrane surface is governed by the flux towards the membrane and the back transport of particles which includes Brownian diffusion, inertial lift (Green and Belfort, 1980) and shear-induced diffusion (Romero and Davis, 1988, 1991). If conditions are such that the back transport is greater than the permeate flux, then the particles are not expected to be deposited on the membrane surface. This effect is more likely for larger particles because back

MFI-UFconst.pressure Modified Fouling Index-Ultrafiltration constant pressure (s L2) Q flowrate (L h1) t filtration time (s) v crossflow velocity (m s1) V permeate volume (L) Greek symbols a specific cake resistance (m/kg) g shear force (N) 3 cake porosity m viscosity of fluid (Pa s) u compressibility factor particle density (kg m3) rp

transport velocities increase with the particle diameter (Green and Belfort, 1980; Romero and Davis, 1988, 1991). Therefore, in the crossflow process, large particles that have larger back transport velocities will tend to migrate away from the membrane surface. These hydrodynamic conditions that occur during crossflow membrane processes are neglected in the conventional dead-end MFI test. Without considering the hydrodynamics in the RO crossflow process and the mode of operation, some important issues regarding the fouling potential of the feed might be overlooked. The Crossflow Sampler-Modified Fouling Index (CFS-MFI) was introduced to incorporate the crossflow hydrodynamic behaviour during fouling index measurement (Adham and Fane, 2008). SDI and MFI0.45 constant pressure obtained after the crossflow sampler were found to be significantly lower than standard SDI and MFI0.45 for different types of feed water, emphasising the importance of crossflow hydrodynamics (Adham and Fane, 2008; Javeed et al., 2009). Our recent work has further extended this work where the crossflow sampler unit (CFS) was directly connected with the dead-end cell and the test was carried out under constant flux, known as CFSMFIUF constant flux (Sim et al., 2010). The aim of the CFS is to simulate the selective deposition of colloids during the crossflow RO process. Due to the shear, only the portion of the particles that will potentially deposit on the membrane can permeate through the CFS and enter the dead-end MFI device. These components represent the composition that is most likely to cause fouling in a RO crossflow system if the same feed was used. The fouling potential of these foulants can hence be determined through the dead-end device. In our previous study, the sensitivity of both MFI-UFconst.flux and CFSMFIUF was validated through lab scale RO experiments using synthetic silica suspension. MFI-UFconst.flux was found less sensitive when compared to CFS-MFIUF. The fouling rate prediction based on CFS-MFIUF agreed well with the actual RO fouling behaviour with the deviation of 11%, whereas MFIUFconst.flux deviated significantly from the actual trend (>30%) even with the deposition factor correction (Sim et al., 2010). However, the factors affecting the performance of this improved MFI test such as crossflow velocity on CFS-MFIUF values have not yet been presented. This paper aims to understand the particle capture and fouling mechanism in different fouling indices at which the

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 3 9 e1 6 5 0

effects of operating conditions such as flux and pressure on MFI-UFconst.pressure and MFI-UFconst.flux were investigated. In addition, a detailed study on the effects of different operating conditions such as crossflow velocity, permeate flux of CFS and dead-end flux, and the influence of spacer on CFS-MFIUF is presented.

2.

Theory

The concept for both MFI0.45 and MFI-UFconst.pressure can be derived based on cake filtration theory without cake compression (Schippers and Verdouw, 1980). Under constant pressure, both MFI0.45 and MFI-UFconst.pressure are defined as the slope of the linear portion of the t/V and V filtration curve, where t refers to the filtration time and V is the filtrate volume. They can also be represented as follows (Schippers and Verdouw, 1980): MFI  UFconst:pressure ¼

mI 2DPA2

(1)

where m is the viscosity of the feed solution, I is the resistivity, DP is the transmembrane pressure (TMP) and A is the filtration area. The resistivity I, is the product of the specific cake resistance (a) of the deposit and concentration of particles in the feed water (Cb) (Boerlage et al., 1998): I ¼ aCb

(2)

The specific cake resistance for a known concentration solution can be calculated by substituting Eq. (2) into Eq. (1): a¼

2MFI  UFconst:pressure DPA2 mCb

(3)

In many cases, the solids concentration is unknown. So, the exact value of the specific cake resistance is unlikely to be known. Eq. (3) simply shows the relationship between specific cake resistance and MFI-UFconst.pressure. High values of MFIUFconst.pressure indicate that the cake on the test membrane is high in resistance. MFI-UFconst.pressure not only acts as an indicator for water quality, but also indirectly reflects the cake structure formed during the test. For compressible cakes, Almy and Lewis (1912) developed an empirical relationship between specific cake resistance and pressure as follows: a ¼ a0 DPu

(4)

where u is the compressibility factor of the cake, DP is the operating pressure and a0 is a constant. For incompressible cakes, u is zero and the higher the compressibility factor, the more compressible the cake. For compressible cakes, MFI-UFconst.pressure has the following relationship (Boerlage et al., 2003b): MFI  UFconst:pressure

ma0 DPu Cb ¼ 2DPA2

(5)

Simplifying Eq. (5), gives MFI  UFconst:pressure ¼ Constant  DPu1 where Constant refers to ma0Cb/2A2.

(6)

1641

The derivation of MFI-UFconst.flux under constant flux is also based on cake filtration theory (Boerlage et al., 2004). MFIUFconst.flux is represented as follows: MFI  UFconst:flux ¼

mI 2DP0 A2

(7)

where cake resistivity, I is obtained from the slope of TMP against time filtration curve. DP0 refers to the standard pressure of 2 bar and A is active filtration membrane area of 0.0014 m2. Similarly, CFS-MFIUF measured at constant flux is defined as: CFS  MFIUF ¼

mI0 2DP0 A2

(8)

where I0 is the cake resistivity modified by the Crossflow Sampler.

3.

Materials and methods

3.1.

Experimental setup

3.1.1.

Constant pressure dead-end filtration setup

The feed tank which has a volume of 1.8 L was connected to a dead-end cell. The dead-end cell with active membrane area of 0.0014 m2 (diameter ¼ 42.2 mm) is equipped with a porous support at which ultrafiltration membrane was placed on. A nitrogen gas cylinder was used to pressurize the system to the desired operating pressure. When the applied pressure reached the required level, the outlet of the dead-end cell was opened and the permeate was collected. The feed pressure was remained constant and was monitored using a pressure transducer (Labom, CB1020) located at the feed line whereas the mass of the collected permeate was measured by a balance (Mettler Toledo, PB8001-S). Both pressure and mass data from the pressure transducer and balance respectively were recorded continuously by LabVIEW 8.2.

3.1.2.

Constant flux dead-end filtration setup

In constant flux filtration, the feed tank (1.8 L) was connected to a peristaltic pump (Masterflex L/S) prior to the dead-end cell (diameter ¼ 42.2 mm, active filtration area ¼ 0.0014 m2). The pump was used to ensure constant amount of feed to be fed in to the dead-end cell and the same amount of permeate would result. Even though the setup was operated under constant flux condition, a nitrogen gas cylinder was still needed to pressurize the system as the peristaltic pump was unable to cope with high pressure difference. Two pressure transducers (Labom, CB1020) were located at the upstream and downstream of the dead-end cell. The pressures and mass data were recorded continuously by LabVIEW 8.2.

3.1.3.

CFS-MFIUF setup

The CFS-MFIUF combined a crossflow filtration module and a dead-end MFI assessment device on the permeate of the CFS cell. The crossflow cell was made from acrylic glass (also known as Perspex) which has a flow channel of 111 mm  25 mm  4 mm and effective membrane area of 0.0028 m2. An important feature of the CFS was the use of very large pore membrane (Refer Section 3.2). A feed spacer of

1642

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 3 9 e1 6 5 0

approximately 0.79 mm thick was used. A centrifugal pump was used to pump the feed to the crossflow cell and the crossflow velocity was manipulated through the valve located at the retentate line. A peristaltic pump (Masterflex L/S pump) was located in the permeate line to withdraw the permeate from the crossflow cell and pump into the dead-end filtration cell (diameter ¼ 42 mm, unless specified) as shown in Fig. 1. The purpose for this pump was to maintain constant flux for both crossflow and dead-end filtration. The membranes used in the dead-end cell were 10 kDa ultrafiltration membranes. TMP of the dead-end cell was continuously measured and recorded by computer and plotted against filtration time to obtain resistivity I0 . The CFS-MFIUF value can be determined by substituting I0 into Eq. (8). For the flux correction factor experiments, dead-end filtration cell with radius of 12 mm (membrane area ¼ 4.524  104 m2) was used to determine CFS-MFIUF (6:1).

Membranes

Ultrafiltration membranes with MWCO of 10 kDa were used in the dead-end filtration cell for the purpose of capturing fine particulates. The CFS membrane is a “non-retentive” filter with straight through pores. Four different pore sizes of membrane filters namely 5 mm, 11 mm, 40 mm and 100 mm were compared and used in this study. The specifications of all types of membranes used are given in Table 1.

3.3.

Feed water

Silica was selected as a representative of colloidal foulant. Three types of silica colloidal suspensions were used as the feed solution in most of the experiments. They were monodispersed 22 nm (LUDOX TM-50, Sigma Aldrich), mixed particles sizes of 70e100 nm silica colloidal (SNOWTEX-ZL, Nissan Chemical Industries Ltd) and 3 mm (Sigma Aldrich). Silica suspensions were prepared by adding Milli-Q to the desired concentration and underwent ultrasonic treatment for 5 min to ensure that the suspension was stable and without large aggregates. In order to ensure that the pH of the feed remained constant throughout the experiment, a buffer solution of pH 8 was used as the background solution. The

4.

Results and discussion

4.1.

MFI measured at constant pressure mode

4.1.1.

Effect of membrane pore size on MFI value

Using constant pressure of 207 kPa, the effect of different membrane pore sizes on MFI values was investigated. The membranes used for comparison ranged from 0.45 mm down to 10 kDa are presented in Table 2. Meanwhile, the feed suspension used was 50 ppm of 3 mm silica. As can be observed, MFI values increased with decreasing membrane pore size. One possible reason was due to the presence of small particles in the solution. This trend was first demonstrated by Schippers et al. (1981) who showed that MFI value of the pre-treated river Rhine water increased sharply as membrane pore size decreased. They suggested that this may be attributed to the presence of fine particles in the pretreated river Rhine water. Small particles which could not be retained by larger pore size membrane were not incorporated in the MFI measurement and hence the fouling propensity of the water was underestimated. In order to verify this possibility, the particle size used in this study was analysed using Malvern Mastersizer. The particle size distribution ranged from 1 mm to 11 mm with the average particle size of 4.12 mm. The results also indicated that the smallest particle size for this solution was 1 mm, suggesting that all particles in the solution can be retained by 0.45 mm membrane. Hence, the increase in MFI value with decreasing membrane pore size in current study cannot be due to the retention of small particles.

Crossflow Sampler (A1, J1)

Computer

PG

CFS permeate (Q1)

Centrifugal Pump

Flow meter Peristaltic Pump

PT

Valve 1

Dead-end cell (A2, J2) Valve 2

Dead-end permeate (Q2)

Cooling coil

Feed

3.2.

buffer solution was a mixture of 6.81 g of KH2PO4 and 461 ml of 0.1 M NaOH. Besides colloids, humic acid was also used as a representative of organic foulant. The concentration of humic acid samples was characterised in terms of their total organic carbon (TOC) concentration using Shimadzu TOCVCSH. Two different dynamic light scattering particle distribution detectors (NanoDLS, Brookhaven Instrument Corporation, and Malvern Mastersizer) were used to investigate the particle size distribution of the suspension under different conditions.

Water bath Balance PT

Pressure Transducer

PG

Pressure Gauge

Fig. 1 e Experimental rig for CFS-MFIUF measurement.

1643

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 3 9 e1 6 5 0

Table 1 e Specifications of different types of membrane used in this study. Membrane type

Membrane pore size

Material

Net filter

5 mm 11 mm 41 mm 100 mm

Hydrophilic Hydrophilic Hydrophilic Hydrophilic

Microfiltration membrane (MF)

0.45 mm 0.22 mm

Ultrafiltration membrane (UF)

10 kDa 30 kDa 100 kDa

% Porosity 15 6 31 44

GE Osmonics Millipore Corporation Millipore Corporation Millipore Corporation

Hydrophilic PVDF Hydrophobic PVDF

70 75

Millipore Corporation Millipore Corporation

Polyethersulfone Polyethersulfone Polyethersulfone

e e e

Omega by Pall Corporation Omega by Pall Corporation Omega by Pall Corporation

Another explanation for the dependency of MFI value on membrane pore size might be the effect of membrane surface morphology upon cake resistance. Membranes with low surface porosity and irregular distribution of pores were found to promote higher cake resistances as suggested by Fane et al. (1991). In their study, a comparison of cake resistance was made between two iso-porous membranes with the same pore size but different in porosity and pore size distribution for the filtration of Escherichia coli suspension. They suggested that irregular distribution of pores on the membrane surface could cause higher local fluxes, and hence lead to a higher cake resistance. Recent studies by Boerlage et al. (2002) found similar phenomena where MFI-UF was dependent of MWCO (1e100 kDa) varying from 2000 to 13,300 s/L2 when heterogeneously porous and low surface porosity membranes were used. Boerlage et al. stated that low surface porosity would decrease the effective filtration area; hence based on Eq. (7) artificially high MFI-UF values were expected. In the same study, when Polyacrylonitrile (PAN) membranes with different MWCO (6, 13, 50 kDa) were used, MFI-UF was found independent of MWCO varying to only a small extent from 2000 to 2800 s/L2. FESEM images of these three MWCO membranes revealed that these membranes had high surface porosity and were homogeneously porous. The results tabulated in Table 2 clearly show that the MFI values for microfiltration membrane were markedly lower than ultrafiltration membrane. MF membranes which have higher surface porosity and larger pore size than UF membrane would increase the effective filtration area, and hence lead to lower MFI value. By the use of the Carman Kozeny relationship as shown in Eq. (9), cake porosity (3) for

Polycarbonate Nylon Nylon Nylon

Suppliers

the corresponding filtration can be estimated based on specific cake resistance (a), particle density (rp) and particle diameter (dp) a¼

180ð1  3Þ rp d2p 33

(9)

The results suggest that cake formed during MF is more porous than that formed on UF membrane with the porosity value lying between 0.3 and 0.4. For rigid spherical particles, the void fraction (3) of a randomly packed cake is about 0.4 (Belfort et al., 1994). Therefore, this phenomenon can be explained by the random deposition of particle at higher filtration flux during MF filtration. The much lower cake porosity for the low MWCO UF membranes could be due to two factors. Firstly, as explained above the more sparsely surface porous membranes tend to overestimate specific cake resistance. Secondly, the lower fluxes allow formation of a more regular cake and the possible ‘infiltration’ of the smallest particles in the feed into the existing cake deposit. In summary, in this set of experiments, the dependency of MFI on membrane pore size was not due to the retention of fine particles, but membrane surface morphology. However, it is believed that fine particles are often responsible for the membrane fouling in RO, thus UF membrane is often recommended as the reference membrane in MFI measurement, known as MFI-UF.

4.1.2.

Effect of operating pressure on MFI-UFconst.pressure

Colloidal silica (22 nm, Sigma Aldrich) was employed to investigate the dependency of MFI-UFconst.pressure on applied pressure. In Fig. 2, no obvious trend can be observed between MFI-UFconst.pressure and applied pressure from 25 to 200 kPa.

Table 2 e Comparison of MFI for different membranes and their corresponding cake properties. Type of membrane MF MF UF UF UF

Membrane pore size (mm)/MWCO (kDa) 0.45 mm 0.22 mm 100 kDa 30 kDa 10 kDa

Membrane material PVDF PVDF PES PES PES

PVDF e Polyvinylidene fluoride; PES e Polyethersulphone.

Clean membrane resistance (m1)

MFI (s/L2)

Specific cake resistance, a (m/kg)

Cake porosity

1.61  1010 5.21  1010 1.45  1011 5.68  1011 7.51  1011

3.90 7.25 25.49 162.79 292.80

7.25  1010 1.34  1011 4.68  1011 2.97  1012 5.33  1012

0.42 0.35 0.24 0.14 0.11

1644

5

34

a

b

4 33

3

ln

MFI-UFconst.pressure x104 (s/L2)

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 3 9 e1 6 5 0

2

y = 1.0053x + 21.017

32

1 0

31 0

100

200

10

11

12

13

ln P

Applied Pressure (kPa)

Fig. 2 e (a) Effect of applied pressure on MFI-UFconst.pressure; (b) Compressibility check for 50 ppm of mono-size 22 nm silica suspension.

3

4.2.

MFI-UF measured at constant flux mode

4.2.1.

Effect of applied flux on MFI-UFconst.flux

As most industrial RO membrane systems are operated under constant flux mode, the MFI-UF measured under constant pressure might not be able to adequately reflect the fouling potential of the feed due to the difference in hydrodynamic behaviour. Hence, the effect of applied flux on MFI-UF was studied. The feed solution used in these experiments was 50 ppm of 22 nm colloidal silica solution. The results obtained are shown in Fig. 4. MFI-UFconst.flux was found to increase as applied flux increased. At a high applied flux, due to the great permeate drag force, more particles can easily deposit on the membrane surface. These particles that deposit on the membrane surface contributed significantly to the major resistance and hence leading to the higher values of MFI-UFconst.flux. Similar trend was observed by Boerlage et al. (2004) when investigating the effect of flux on MFI-UFconst.flux using tap and canal water. Membranes tend to foul more rapidly during high applied flux operation than in low flux operation which is why low to moderate applied flux was chosen in membrane water industry. As MFI-UFconst.flux is sensitive to operating flux, MFIUFconst.flux measured under different fluxes is not directly comparable. In order to more closely simulate the fouling behaviour in RO, the operating flux of MFI-UFconst.flux should be the same as that in the real RO filtration, which is around 35.0

a

b

2.5 34.5

2 1.5

ln

MFI-UFconst.pressure x 106 (s/L2)

This observation can be explained by using Eq. (6). When compressibility u equals to 1, MFI-UFconst.pressure becomes independent of applied pressure. In order to verify this, the compressibility factor was obtained by plotting ln a against ln DP where compressibility factor is the slope of the line. Compressibility factor for the solution used is approximately 1 as demonstrated in Fig. 2b. Hence, MFI-UFconst.pressure was proven to be independent of applied pressure for this type of suspension. Khirani et al. (2006) observed similar trend, MFI measured at 2 bar was similar to that measured at 4 bar using NOM solution (Bioiberica). The authors interpreted this phenomenon by the high compressibility factor of the cake with the value close to 1. For solutions that are less compressible (u < 1), the MFI-UFconst.pressure is expected to dependent on operating pressure. A good example is represented in Fig. 3, where a less compressible suspension (Mixture of 50 ppm 22 nm and 50 ppm 70e100 nm silica) with compressibility factor of 0.54 was used. These results indicated that MFI-UFconst.pressure depends on the applied pressure as well as the compressibility of the feed. A pressure correction factor is therefore important when MFIUFconst.pressure is not measured under the reference pressure. Boerlage et al. (2003a) incorporated the compressibility factor to correct the experimentally determined MFI-UFconst.pressure at 0.5 and 1 bar to the reference pressure of 2 bar and observed good agreement between experimentally obtained MFIUFconst.pressure and the corrected MFI-UFconst.pressure with only a slight difference of 5e10%.

34.0

1

y = 0.5435x + 28.038

0.5 0 0

50

100

150

Applied Pressure x 10 3 (kPa)

200

33.5 10.5

11.0

11.5

12.0

12.5

ln P

Fig. 3 e (a) Effect of applied pressure on MFI-UFconst.pressure; (b) Compressibility check for 50 ppm 22 nm and 50 ppm 70e100 nm mixture solution.

1645

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 3 9 e1 6 5 0

foulants formed under this case should be less than that formed under constant flux of 121 L/m2 h. Thus, the cake resistance under flux of 28 kPa is expected to be less resistive than that formed under constant flux of 121 L/m2 h. This section highlighted the fact that a direct comparison between MFI-UFconst.flux and MFI-UFconst.pressure cannot be made as this may result in misleading information on the fouling propensity of the given feed.

MFI-UFconst.flux (s/L2)

8000

6000

4000

2000

4.3.

Study of CFS-MFIUF

0 0

50

100

150

Applied Flux (L/m2.h)

Fig. 4 e Effect of applied flux on MFI-UFconst.flux; Feed [ 50 ppm of 22 nm silica colloidal in buffer.

20e30 L/m2 h. However, this prolongs the duration of the test substantially. Section 4.5.3 will further consider this issue and proposed a method to streamline the test.

4.2.2. Comparison between MFI-UF constant flux and MFI-UF constant pressure From Figs. 2 and 4, it is obvious to note that MFI-UFconst.flux showed lower values than MFI-UFconst.pressure. MFI-UFconst.flux has values ranging from 1000 s/L2 to 8000 s/L2, whilst under constant pressure, MFI-UFconst.pressure lies between 30,000 s/L2 and 40,000 s/L2. It should be noticed that the values of MFIUFconst.flux have been scaled to the standard pressure of 2 bar based on Eq. (7), whilst MFI-UFconst.pressure was determined based on its operating pressure. Thus, the values of the resistivity, I were calculated. I values for constant pressure operation ranged from 3.8  1012 to 2.52  1013 m2 while, for constant flux experiments, I lies between 1.45  1012 and 6.2  1012 m2. Two experiments with similar initial fluxes and initial pressures obtained at the start of the experiments were compared and tabulated in Table 3. MFI-UFconst.flux is lower than MFI-UFconst.pressure, but the resistivity value of constant flux is higher than the one from constant pressure. The difference between the resistivity values indicates cakes formed under these two conditions varied significantly. The cake formed under constant flux of 121 L/m2 h has higher I value and therefore higher specific cake resistance than the cake formed under constant pressure of 28 kPa. The most likely explanation for this observation is that the operating flux of 121 L/m2 h lies above the critical flux zone, severe fouling could have occurred instantly at the very early stage of the filtration. Under constant pressure mode, the flux declined from 116 L/m2 h to lower a permeate flux. The amount of

RO systems are usually operated under constant flux mode, hence, our focus is an MFI-UFconst.flux but in combination with crossflow sampler (CFS), which is known as CFS-MFIUF.

4.3.1.

Effect of CFS membrane pore size on CFS-MFIUF

As mentioned earlier a crossflow sampler (CFS) with large membrane pore size is implemented before the dead-end filtration cell in order to simulate the hydrodynamic conditions of the crossflow RO. Due to the selective deposition in the crossflow cell, large particles are unlikely to migrate near to the crossflow cell membrane surface, whereas small particles do. These particles which are able to permeate through the CFS are the foulants that represent the composition that is most likely to cause fouling in a RO crossflow system when the same feed is being used. In order to ensure that the membrane in the crossflow sampler is permeable to any foulants that approach it, the membranes must have large pore, high porosity, and ideally they would be straight through pores. This leads to the investigation of the effect of CFS membrane pore sizes on CFS-MFIUF. Four types of MF membranes with pore size ranging from 5 mm to 100 mm were investigated. The difference of CFS-MFIUF values between these four types of membrane was insignificant, with values ranging from 54,000 to 57,000 s/L2 (standard deviation of 2544 s/L2). These results imply that for the feed used in this study, membranes with pores R 5 mm can be used as the CFS membrane as the resulting CFS-MFIUF values are similar. The selection of CFS membrane could be based on the pre-treatment method used prior to RO. For example, if microfiltration membrane (5 mm) was used as the pre-treatment stage, then any membrane with higher pore size than 5 mm can be used as CFS membrane as long as no cake build up occurs to block pores.

4.3.2.

Effect of crossflow velocity on CFS-MFIUF

The mechanism of particle deposition during crossflow filtration not only depends on particle size itself, but also the hydrodynamic conditions such as crossflow velocity and filtration flux (Lu and Ju, 1989; Chellam and Wiesner, 1998;

Table 3 e Flux, TMP and MFI-UF value at their corresponding operating mode. Constant flux Operating flux (L/m2 h) 121.11

Constant pressure Initial TMP (kPa)

MFI-UFconst.flux (s/L2)

Resistivity, I (m2)

Operating TMP (kPa)

Initial flux (L/m2 h)

MFI-UFconst.pressure (s/L2)

26.7

7043

6.20  1012

28

115.71

30,730

Resistivity, I (m2) 3.81  1012

1646

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 3 9 e1 6 5 0

Fischer and Raasch, 1985). As for this reason, the potential effect of crossflow velocity on CFS-MFIUF was studied. Fig. 5 shows the trend of CFS-MFIUF as a function of crossflow velocity. Clear minimum values of CFS-MFIUF for both model feed and raw seawater were observed. The results indicated the competing effects of two phenomena: (1) Increase in crossflow velocity generates surface shear which tends to reduce the load of particles reaching the CFS surface. In this case Cb in Eq. (2) is decreased. Hence CFS-MFIUF decreases as crossflow velocity increases. (2) Increase in crossflow velocity causes the average particle size in the deposit to decrease. In this case a in Eq. (2) is increased. Consequently, this leads to a more resistant cake structure and therefore CFS-MFIUF values increase. It is suggested that in the low crossflow velocity range, an increased crossflow velocity produced a lower CFS-MFIUF values, dominating by phenomenon 1. The additional surface shear generated near the CFS membrane surface prevented particulates to flow near to the CFS surface and thus reduced the load of particles that permeate through the CFS unit. In fact, the reduction in cake mass as crossflow velocity increased has already been reported. Baker et al. (1985) observed that the cake mass during the crossflow filtration process is strongly dependent on crossflow velocity at which they correlated their cake loading data (mc, kg/m2) with the crossflow velocity (v, m/s),

a

100000

CFS-MFIUF (s/L2)

80000

60000 Phenomenon 2

Phenomenon 1

40000

20000

0 0

b

0.2

0.4

0.6

0.8

3000

CFS-MFIUF (s/L2)

2500 2000 1500 Phenomenon 2

Phenomenon 1

1000 500 0 0

0.2

0.4

0.6

0.8

1

Crossflow velocity (m/s)

Fig. 5 e CFS-MFIUF at different crossflow velocities (a) Mixture of 50 ppm 22 nm and 50 ppm 70e100 nm silica colloidal in buffer (b) raw seawater. Conditions: Crossflow flux [ 22.5 L/m2 h; Dead-end flux [ 55 L/m2 h.

mc ¼

0:17 v2

(10)

Based on this correlation, the cake mass reduces as the crossflow velocity increases. Notwithstanding this, the decrease in the cake mass is expected to reach a stable cake mass above a certain velocity. No significant cake mass was observed above crossflow velocity of 4 m/s in experimental results illustrated by Baker et al. (1985). Recently, Chong et al. (2008) who studied the crossflow RO fouling behaviour using silica colloids observed that the amount of particle that deposited on the membrane surface decreased as crossflow velocity increased. They characterised the amount of particles convected to and finally deposited on the membrane surface in term of fractional deposition constant (f), which varies between 0 and 1. A value of 1 means all particles are deposited on the membrane surface during the crossflow filtration process, while f ¼ 0 means no particle deposition. In their study, the fractional deposition constant was about 1 at low crossflow velocity (shear rate < 430 s1; v < 0.1 m/s), whereas this factor dropped to about 0.1 at high shear rate (shear rate > 940 s1; v > 0.23 m/s). They observed that the fractional deposition constant remained at a value of 0.1 when the crossflow velocity reached above 0.23 m/s. This result further confirmed that the decrease in mc may reach a stable value above a certain crossflow velocity during crossflow filtration. As shown in Fig. 5a, CFS-MFIUF appears to increase above crossflow velocity of 0.37 m/s. The rise of the CFS-MFIUF can be explained by the dominating effect of phenomenon 2 as well as the diminishing effect of phenomenon 1. The influence of the crossflow velocity on the particle size distribution of the CFS permeate was investigated using NanoDLS and is illustrated in Fig. 6. As can be seen, when the crossflow velocity increased, the particle distribution in the CFS permeate suspension tended to shift to lower average particle sizes. Such decrease in average cake particle size consequently raised the values of the specific cake resistance (i.e. CFS-MFIUF value). This phenomenon has been reported consistently in several studies (Lu and Ju, 1989; Chellam and Wiesner, 1998; Fischer and Raasch, 1985; Meier et al., 2002; Tanaka et al., 2001; Baker et al., 1985). A similar trend was observed when using raw seawater, with a minimum CFS-MFIUF occurring at crossflow velocity of 0.57 m/s, indicated in Fig. 5b. The seawater was obtained from the Sydney area and has been pre-filtered with 40 mm filter. The minimum values of CFS-MFIUF for silica suspension and seawater occurred at different crossflow velocities. This is due to the difference in particle size population as well as the composition in both feeds. In summary, crossflow velocity is an important factor that may affect the measurement of CFS-MFIUF. The selection of operating crossflow velocity in CFS-MFIUF measurement relies on the feed flowrate in the actual RO system.

4.3.3.

Influence of spacers on CFS-MFIUF

To further illustrate CFS-MFIUF can successfully capture the hydrodynamic behaviour during the crossflow filtration, the influence of spacers on CFS-MFIUF was studied. CFS-MFIUF obtained without the presence of spacers has the value of

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 3 9 e1 6 5 0

capture smaller particle sizes and the cake formed is expected to be denser, the effective thickness as well as the amount of the cake is still relatively small compared to that formed in MFI-UFconst.flux measurement. This is because MFI-UFconst.flux is measuring the effect of all components of the feed, so the cake formed on MFI-UFconst.flux membrane consists of both large and small particles.

120

Multimodal size distribution

Feed 100

0.26m/s 0.37m/s 0.47m/s

80

1647

60 40

4.5. Effect of crossflow sampler and dead-end permeate flux ratio on CFS-MFIUF

20 0 0

50

100

150

200

250

Particle size (nm)

Fig. 6 e Intensity weighted particle size distribution curves at different crossflow velocities. Conditions: Flux [ 55 L/ m2 h; Feed [ Mixture of 50 ppm 22 nm and 50 ppm 70e100 nm silica colloidal in buffer. Particle size distribution of feed and at crossflow velocity of 0.37 m/s have been previously presented in Sim et al. (2010).

One of the special features of CFS-MFIUF method is that the dead-end MFI unit is directly connected with the crossflow sampler. As such, the permeate flowrates for both devices are the same. The CFS permeate flowrate, Q1 is the product of CFS permeate flux and CFS membrane area A1. Q1 ¼ J1 A1

(11)

Similarly, the permeate flowrate for dead-end unit is Q2 ¼ J2 A2

(12)

Since, Q1 ¼ Q2, 62,500 s/L2 which was about 12% higher than that obtained with spacers. This is mainly because with the presence of spacers, greater shear forces were generated compared to the channel without spacers, thus hindering the particle migration towards the CFS membrane surface. Consequently, the load of particles in the CFS permeate (Cb) is less when compared to the situation without spacers.

4.4.

Comparison of CFS-MFIUF and MFI-UFconst.flux

In order to confirm that the crossflow sampler actually has influence on the MFI-UF values, a comparison was made between CFS-MFIUF and MFI-UFconst.flux using different types of feed. Fig. 7 shows consistent trends. The CFS-MFIs were lower than the MFI-UFconst.flux with differences ranging from 20% to 30%. This trend can be anticipated. Due to the hydrodynamic conditions in the crossflow filtration, the foulants permeate through the crossflow sampler selectively, hence the feed that enters the dead-end MFI assessment device differs from the original feed, containing small particle size range components and less foulants load. Even though CFS-MFIUF is expected to

100000 Fouling Index (s/L2)

MFI-UFconst.flux 80000

CFS-MFI-UF

60000 40000 20000

J1 A2 ¼ J2 A1

(13)

As A1 and A2 are fixed parameters, the selected CFS permeate flux will affect the operating flux in subsequent dead-end unit. In our CFS-MFIUF setup, the membrane ratio (A2:A1) is 1:2. In most our experiments presented earlier, the operating flux at the dead-end unit was 55 L/m2 h, thus the respective crossflow flux was approximately 28 L/m2 h. Since CFS and dead-end cell are interconnected, varying the operating flux in J2 subsequently leads to different J1. To single out the potential effect of each unit’s flux, a series of experiments was performed with either fixed value of J1 or J2. To achieve this, the experiments in Section 4.5.1 and 4.5.2 were conducted using the setup at which dead-end cell was disconnected from the CFS unit.

4.5.1. Variation in crossflow sampler permeate flux (J1) but fixed dead-end permeate flux (J2) In this set of experiments, permeate from the crossflow cell was first collected at different crossflow fluxes, then the permeate was used to measure fouling index at the same dead-end flux (55  2 L/m2 h). The feed solution used in these experiments was the mixture solution of 50 ppm 22 nm and 50 ppm 70e100 nm silica colloidal. Crossflow velocity was maintained at 0.37 m/s and with constant temperature of 22  C. The corresponding CFS-MFIUF values at different crossflow permeate fluxes are plotted in Fig. 8. CFS-MFIUF values increased as the permeate flux of crossflow sampler increased. This is an expected trend because at higher crossflow sampler permeate flux, more particles were captured and permeate through the CFS membrane.

0 Silica only

HA only (5ppm TOC)

Silica + HA (5ppm TOC)

Silica + NaCl(1000ppm)

Fig. 7 e Comparison of CFS-MFIUF and MFI-UFconst.flux using different type of feeds. Conditions: Flux [ 55 L/m2 h; Crossflow velocity [ 0.37 m/s.

4.5.2. Variation in dead-end permeate flux (J2) but fixed crossflow sampler permeate flux (J1) In the case of constant crossflow permeate flux at 28 L/m2 h and altering the dead-end permeate flux, the results are plotted given in Fig. 8. It is evident that, CFS-MFIUF also

1648

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 3 9 e1 6 5 0

60000

40000

20000

0 0

20

40

60

80

100

120

Applied flux (L/m2.h) Crossflow permeate flux (Section 4.5.1)

Dead-end permeate flux (Section 4.5.2)

Fig. 8 e CFS-MFIUF at different crossflow permeate fluxes (:) and at different dead-end permeate fluxes ( ). Conditions: Crossflow velocity [ 0.37 m/s; Feed [ mixture of 50 ppm 22 nm and 50 ppm 70e100 nm silica colloidal in buffer.

increased as dead-end flux increased. Similar explanations as given in previous section can be made in this case.

4.5.3.

Flux correction factor

Previous results in Sections 4.5.1 and 4.5.2 not only signify that the CFS-MFIUF is highly dependent on the permeate fluxes of both units, but also highlight the importance of the choice of membrane area ratio. Since the permeate fluxes in both units are affecting the CFS-MFIUF, one option is to have the same membrane areas, so CFS and dead-end permeate flux are the same. The purpose of the CFS is to simulate the hydrodynamic conditions in the RO unit. To be comparable to the RO flux operated in the industry which is approximately 20e30 L/ m2 h, both the dead-end and CFS permeate flux should also maintain at this range of fluxes. However, such a low operating flux would lead to a very slow fouling rate in the deadend MFI assessment unit, and thus an extended period for measurement of CFS-MFIUF values. In order to accelerate MFI measurement, we proposed to reduce the dead-end filtration area, A2 to approximately 1/6 of A1. As such, when CFS is operated at flux of 20 L/m2 h, the corresponding dead-end flux is 120 L/m2 h. However, in doing so, the CFS-MFIUF obtained at high flux (120 L/m2 h) may not be able to adequately represent the fouling behaviour at flux of 20 L/m2 h. Even though the fouling potential of different feeds can still be compared through this accelerated formed of CFSMFIUF, it can only act as a fouling propensity indicator for the feed. The RO fouling rate prediction cannot be made based on this accelerated CFS-MFIUF due to the permeate flux difference. Therefore, in this paper, we intend to investigate if a flux correction factor does exist to correlate CFS-MFIUF measured under this accelerated condition with CFS-MFIUF determined under low flux condition. Different types of feed solution such as humic acid and real water were used in a series of experiments to investigate if the flux correction factor exists for specific operation conditions regardless of the types of feed. For each type of feed, two experiments with different flux ratios (J1 ¼ 20 L/m2 h and

10000 7.5ppm Humic acid

CFS-MFIUF (s/L2) [1:1]

CFS-MFIUF (s/L2)

80000

J2 ¼ 20 L/m2 h vs J1 ¼ 20 L/m2 h and J2 ¼ 120 L/m2 h) were conducted. In determination of CFS-MFIUF 6:1 (i.e. J1 ¼ 20 L/m2 h and J2 ¼ 120 L/m2 h), a new dead-end cell with membrane area of 4.524  104 m2 was used. For CFS-MFIUF 1:1 (i.e. J1 ¼ 20 L/ m2 h and J2 ¼ 20 L/m2 h), both the CFS and dead-end fluxes were operated at 20 L/m2 h. The results were plotted in Fig. 9. Fig. 9 shows good correlation (R2 ¼ 0.92) for different feeds with a wide range of MFIs. Based on these data a flux correction of about 0.67 can be applied when converting CFS-MFIUF 6:1 to CFS-MFIUF 1:1. The flux correction factor is particularly important when conducting the RO fouling rate prediction based on CFS-MFIUF. This is because CFS-MFIUF 6:1 is measured under accelerated flux of 120 L/m2 h, whereas the RO operating flux is 20 L/m2 h. When predicting fouling rate of a RO unit, the obtained CFS-MFIUF 6:1 has to be multiplied by this correction factor before applying it to the fouling prediction model discussed in our previous work (Sim et al., 2010). One of the major advantages of using this approach is the duration of the test. The test required for TMP to increase by 3 kPa took at least 15 h for raw seawater when using the CFSMFIUF 1:1 while, CFS-MFIUF 6:1 took only 2 h for the TMP to increase to 15 kPa using the same type of feed water. Another set of experiments correlating different flux ratios were also conducted, a good correlation (R2 ¼ 0.97) between these fluxes regardless of the types of feed was also observed (figure not included). Therefore, we believed that there exists a flux correction factor that can be used to correlate different fluxes. However, further work is required to confirm the more general applicability of this correction factor. In spite of the benefits mentioned, the major drawback of this approach is that the flux correction factor is hydrodynamic specific. Both the permeate flux and crossflow velocity can affect the particle deposition during the crossflow filtration. Permeate flux can affect the cake composition in the crossflow filtration system. Several researchers have indicated that lower filtration rate induces finer deposit particles permeate flux (Baker et al., 1985; Knutsen and Davis, 2006). Furthermore, as previously shown in Section 4.3.2, CFS-MFIUF is dependent on the crossflow velocity of the system. The crossflow velocity not only determines the size of particles that being captured but also the quantity. Hence the correction

J = 20L/m .h; J = 20L/m .h

100000

8000

25ppm 70-100nm silica

6000

2.5ppm Humic acid

4000 y = 0.6738x R² = 0.9202

Seawater Batch B

2000 Seawater Batch A

Dam Water

0 0

2000

4000

6000

8000

10000

12000

14000

CFS-MFIUF (s/L2) [6:1] J = 20L/m .h; J = 120L/m .h

Fig. 9 e Relationship between CFS-MFIUF obtained at different membrane area ratios [1:1 and 6:1] using different feed solutions. Conditions: crossflow velocity [ 0.37 m/s.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 3 9 e1 6 5 0

factor might vary at different crossflow velocities or various permeate fluxes. In our study, the flux correction factor was determined under crossflow velocity of 0.37 m/s and for correlating fluxes between 120 L/m2 h and 20 L/m2 h. The changes in either crossflow velocity or the flux may lead to a different value of this factor. As a result, in real RO operation, a new calibration on flux correction factor must always be conducted in accordance to the plant operating flux and crossflow velocity.

5.

Concluding remarks

Factors affecting the performance of MFI-UFconst.pressure, MFIUFconst.flux and CFS-MFIUF were discussed. The study began with the investigation of fouling index that was obtained under constant pressure mode. The dependency of MFIUFconst.pressure on the applied pressure was sensitive to the compressibility of the feed. If the compressibility factor was less than 1, then MFI-UFconst.pressure decreased as the applied pressure increased, whereas, when compressibility factor equal to 1, MFI-UFconst.pressure became independent to the applied pressure. For the case of MFI-UF measured under constant flux mode, the dependency of MFI-UFconst.flux on the operating flux was investigated. The results indicated that MFI-UFconst.flux increased as the applied flux increased. In addition to investigating different operating modes, the effect of different membrane pore sizes in dead-end cell was studied. The results indicated that MFI-UFconst.pressure was greatly dependent on the pore size due to the difference in membrane surface properties. The effects of operating conditions such as applied flux and crossflow velocity on CFS-MFIUF were also investigated. In the case of different crossflow velocities, CFS-MFIUF was found to decrease gradually as crossflow velocity increased; however when a certain crossflow velocity was reached, CFS-MFIUF started to increase and plateau off. CFS-MFIUF was found sensitive to the applied flux and hence indicated that the importance of selecting appropriate operating flux during fouling index test. The effects of varying either crossflow sampler or dead-end permeate flux on CFS-MFIUF were studied. The results indicated that permeate fluxes of both units affect the CFS-MFIUF values significantly. In this fouling index test, CFS should simulate the hydrodynamic conditions in RO, at which the permeate flux should lie between 20 L/m2 h and 30 L/m2 h. However, this flux is too low for dead-end unit to be fouled quickly. So, in order to accelerate fouling, dead-end membrane area was reduced to have high permeate flux while maintaining CFS permeate flux to be comparable to crossflow in RO units. By accelerating the fouling test, the duration of the test can be shortened from 15 h to 2 h. To quantitatively predict RO fouling rates, the accelerated fouling rate in dead-end cell may not represent the true fouling behaviour in low flux mode, CFS-MFIUF obtained at 120 L/m2 h must be corrected to CFSMFIUF obtained at 20 L/m2 h using a correction factor of 0.67. Since the correction factor used in this study is flux and crossflow velocity specific, a different value of correction factor may be required in accordance to the industry operating conditions.

1649

This study highlighted several conditions to be considered during CFS-MFIUF measurement. With these considerations, CFS-MFIUF might be a more realistic testing protocol suitable for industry in determination of fouling potential of RO feed.

Acknowledgements This project is supported by International Science Linkages (DEST-ISL:CG110188) established under the Australian Government’s innovation statement, Backing Australia’s Ability. Furthermore, this study is in collaboration with the European Union 6th Framework Program Membrane-Based Desalination: An Integrated Approach (MEDINA).

references

Adham, S.S., Fane, A.G., 2008. Crossflow Sampler Fouling Index. Report. National Water Research Institute, California, USA. Almy, C., Lewis, W.K., 1912. Factors determining the capacity of a filter press. Ind. Eng. Chem. 4 (7), 528e532. Baker, R.J., Fane, A.G., Fell, C.J.D., Yoo, B.H., 1985. Factors affecting flux in crossflow filtration. Desalination 53 (1e3), 81e93. Belfort, G., Davis, R.H., Zydney, A.L., 1994. The behavior of suspensions and macromolecular solutions in crossflow filtration. Journal of Membrane Science 96, 1e58. Boerlage, S.F.E., Kennedy, M.D., Aniye, M.P., Abogrean, E.M., Galjaard, G., Schippers, J.C., 1998. Monitoring particulate fouling in membrane systems. Desalination 118 (1e3), 131e142. Boerlage, S.F.E., Kennedy, M.D., Dickson, M.R., El-Hodali, D.E.Y., Schippers, J.C., 2002. The modified fouling index using ultrafiltration membranes (MFI-UF): characterisation, filtration mechanisms and proposed reference membrane. Journal of Membrane Science 197 (1e2), 1e21. Boerlage, S.F.E., Kennedy, M., Aniye, M.P., Schippers, J.C., 2003a. Applications of the MFI-UF to measure and predict particulate fouling in RO systems. Journal of Membrane Science 220 (1e2), 97e116. Boerlage, S.F.E., Kennedy, M.D., Aniye, M.P., Abogrean, E., Tarawneh, Z.S., Schippers, J.C., 2003b. The MFI-UF as a water quality test and monitor. Journal of Membrane Science 211 (2), 271e289. Boerlage, S.F.E., Kennedy, M., Tarawneh, Z., De Faber, R., Schippers, J.C., 2004. Development of the MFI-UF in constant flux filtration. Desalination 161 (2), 103e113. Chellam, S., Wiesner, M.R., 1998. Evaluation of crossflow filtration models based on shear-induced diffusion and particle adhesion: complications induced by feed suspension polydispersivity. Journal of Membrane Science 138 (1), 83e97. Choi, J.-S., Hwang, T.-M., Lee, S., Hong, S., 2009. A systematic approach to determine the fouling index for a RO/NF membrane process. Desalination 238 (1e3), 117e127. Chong, T.H., Wong, F.S., Fane, A.G., 2008. Implications of critical flux and cake enhanced osmotic pressure (CEOP) on colloidal fouling in reverse osmosis: experimental observations. Journal of Membrane Science 314 (1e2), 101e111. Fane, A.G., Fell, C.J.D., Hodgson, P.H., Leslie, G., Marshall, K.C., 1991. Microfiltration of biomass and biofluids: effects of membrane morphology and operating conditions. Filtration & Separation 28 (5), 332e340. Fischer, E., Raasch, J., 1985. Cross-flow filtration. German Chemical Engineering 8, 211.

1650

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 3 9 e1 6 5 0

Green, G., Belfort, G., 1980. Fouling of ultrafiltration membranes: lateral migration and the particle trajectory model. Desalination 35, 129e147. Javeed, M.A., Chinu, K., Shon, H.K., Vigneswaran, S., 2009. Effect of pre-treatment on fouling propensity of feed as depicted by the modified fouling index (MFI) and cross-flow samplermodified fouling index (CFS-MFI). Desalination 238 (1e3), 98e108. Khirani, S., Ben Aim, R., Manero, M.-H., 2006. Improving the measurement of the Modified Fouling Index using nanofiltration membranes (NF-MFI). Desalination 191 (1e3), 1e7. Knutsen, J.S., Davis, R.H., 2006. Deposition of foulant particles during tangential flow filtration. Journal of Membrane Science 271 (1e2), 101e113. Lipp, P., Gorge, B., Gimbel, R., 1990. A comparative study of fouling-index and fouling-potential of waters to be treated by reverse osmosis. Desalination 79 (2e3), 203e216. Lu, W., Ju, S., 1989. Selective particle deposition in crossflow filtration. Separation Science and Technology 24 (7), 517e540. Meier, J., Klein, G.M., Kottke, V., 2002. Crossflow filtration as a new method of wet classification of ultrafine particles. Separation and Purification Technology 26 (1), 43e50.

Romero, C.A., Davis, R.H., 1988. Global model of crossflow microfiltration based on hydrodynamic particle diffusion. Journal of Membrane Science 39 (2), 157e185. Romero, C.A., Davis, R.H., 1991. Experimental verification of the shear-induced hydrodynamic diffusion model of crossflow microfiltration. Journal of Membrane Science 62 (3), 249e273. Schippers, J.C., Verdouw, J., 1980. The modified fouling index, a method of determining the fouling characteristics of water. Desalination 32, 137e148. Schippers, J.C., Hanemaayer, J.H., Smolders, C.A., Kostense, A., 1981. Predicting flux decline of reverse osmosis membranes. Desalination 38, 339e348. Sim, L.N., Ye, Y., Chen, V., Fane, A.G., 2010. Crossflow Sampler Modified Fouling Index Ultrafiltration (CFS-MFIUF) e an alternative Fouling Index. Journal of Membrane Science 360 (1e2), 174e184. Tanaka, T., Yamagiwa, Y., Nagano, T., Taniguchi, M., Nakanishi, K., 2001. Relationship between cake structure and membrane pore size in crossflow filtration of microbial cell suspension containing fine particles. Journal of Chemical Engineering of Japan 34 (12), 1524. Yiantsios, S.G., Karabelas, A.J., 2002. An assessment of the Silt Density Index based on RO membrane colloidal fouling experiments with iron oxide particles. Desalination 151 (3), 229e238.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 5 1 e1 6 5 8

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Exploiting a new electrochemical sensor for biofilm monitoring and water treatment optimization Giovanni Pavanello a,*, Marco Faimali a, Massimiliano Pittore b, Angelo Mollica c, Alessandro Mollica c, Alfonso Mollica a a

Istituto di Scienze Marine e Consiglio Nazionale delle Ricerche (ISMAR-CNR), via De Marini 6, 16149 Genova, Italy e-magine IT Srl, via Greto di Cornigliano 6r, 16152 Genova, Italy c Newlab Snc, via Greto di Cornigliano 6r, 16152 Genova, Italy b

article info

abstract

Article history:

Bacterial biofilm development is a serious problem in many fields, and the existing biofilm

Received 29 June 2010

monitoring sensors often turn out to be inadequate. In this perspective, a new sensor

Received in revised form

(ALVIM) has been developed, exploiting the natural marine and freshwater biofilms elec-

29 September 2010

trochemical activity, proportional to surface covering. The results presented in this work,

Accepted 3 December 2010

obtained testing the ALVIM system both in laboratory and in an industrial environment,

Available online 10 December 2010

show that the sensor gives a fast and accurate response to biofilm growth, and that this response can be used to optimize cleaning treatments inside pipelines. Compared to the

Keywords:

existing biofilm sensors, the proposed system show significant technological innovations,

Biofilm monitoring

higher sensitivity and precision. ª 2010 Elsevier Ltd. All rights reserved.

Biosensor MIC prevention Electrochemically active biofilm Cathodic depolarization

1.

Introduction

Serious wide-range technological problems (corrosion, equipment failure, energy loss, reduced performance and resistance to antimicrobial treatments) can be caused by bacterial biofilm development on any artificial apparatus exposed to natural water (fluid flow systems, water distribution lines, sensors, etc.), with subsequent highly negative economic repercussions (Parr and Hanson, 1965; Whitehouse et al., 1991; Borenstein, 1994; Geesey et al., 1994; Gilbert et al., 1997; Flemming and Shaule, 1994; Schultz and Swain, 2000). In the water lines of industrial plants, for example, large amounts of disinfectants and other chemical substances are

usually employed as a countermeasure against biofilm (Wirtanen et al., 2001; Prince et al., 2002; Maxwell, 2005). Real-time, continuous monitoring of bacterial growth is extremely useful in order to optimize these (and others) biofilm hindering treatments, making possible to apply them as soon as biofilm appears; regarding chemical treatments, biofilm monitoring allows also the optimization of biocides dosage, entailing a reduction of both costs and of biocide treatments environmental impact. This led, in the past years, to the study of different biofilm sensing techniques: measurement of (a) light scattering (Flemming et al., 1998), (b) turbidity (Klahre et al., 1998), (c) electrochemical impedance (Mun˜oz-Berbel et al., 2006; Dheilly et al., 2008), (d ) vibration response of the monitored surface

* Corresponding author. Tel.: þ39 010 6475407; fax: þ39 010 6475400. E-mail address: [email protected] (G. Pavanello). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.12.003

1652

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 5 1 e1 6 5 8

(Pereira et al., 2008), (e) diffusion limitation (Foret et al., 2010). These techniques are affected by several limitations, since all of them: - cannot discriminate between biological and inorganic fouling; this is a major problem, since these two different kinds of fouling require different treatments; - have a low sensitivity, e.g. cannot detect biofilm initial colonization phases, but only thicker bacterial layers; on the other hand, many biofilm-related problems, such as Microbiologically Influenced Corrosion (MIC), start as soon as the first bacterial spots appears on a surface (Mollica and Trevis, 1976; Dexter and La Fontaine, 1998; Kimio et al., 2002). Moreover, some of the above mentioned studies stopped at the laboratory testing phase, and have never been implemented within a real sensor (c). With the aim of overcoming the limitations presented by the existing sensors, a new device (“ALVIM”) has been developed (see Section 2.1), exploiting the cathodic depolarization induced by biofilm growth on active-passive alloys exposed to natural aerated waters. This phenomenon contributes to explain the higher corrosiveness of a natural water, in comparison with a sterile one, toward the mentioned alloys and less noble materials coupled with them. The cathodic depolarization induced by biofilm growth has been largely studied in the last 20 years and has been observed in different parts of the world, both in seawater and in freshwater (Mollica and Trevis, 1976; Scotto et al., 1985; Dexter and Zhang, 1990; Mattila et al., 1997; Dexter and La Fontaine, 1998; Kimio et al., 2002; Wang et al., 2004; Acun˜a et al., 2006; Dulon et al., 2007; Little et al., 2008). Recently, the electrochemical activity of natural aquatic biofilms was proven to be proportional to the surface area covered by bacteria (Faimali et al., 2008, 2010), therefore measuring the biofilm electrochemical signal (BES, expressed as current density or potential, see Section 2.1) is possible to know, on-line and in real-time, which is the biofilm covering on a surface. The aim of this work was to evaluate the performances of this new biosensor, meant to be used for biofilm monitoring and anti-microfouling treatments optimization in industrial environments. The work followed three subsequent steps: (1) Preliminary sensor characterization, in laboratory, to study the response of this innovative probe to biofilm growth, in controlled conditions. (2) Verify the sensor response to biofilm growth in a pilot reverse-osmosis desalination plant, during ordinary working. (3) Optimization of pipeline chemical cleaning treatments, inside the above mentioned plant, basing on biofilm growth real-time data collected by the sensor. Biofilm, indeed, represents a major problem in this kind of industrial environment, both for microfiltration (MF) modules and for reverse-osmosis (RO) membranes, increasing management costs (chemical treatments, membranes cleaning) and contributing to cause cloggings which can bring to plant stop (Fritzmann et al., 2007; Vrouwenvelder et al., 2008).

2.

Materials and methods

2.1.

ALVIM working principle

As described in detail in recent papers (Mollica et al., 1997; Faimali et al., 2008, 2010), cathodic current density i (E,t), measured at a given time t on a stainless steel (SS) sample exposed to natural seawater and polarized at a fixed potential E, can be described by the relation: iðE; tÞ ¼ i1 ðEÞ þ ½ i2 ðEÞ  i1 ðEÞ  qðtÞ

(1)

where i1(E ) is the current density measured on the “clean” fraction of the SS surface and i2(E ) is the one measured on the surface fraction q (t) [0  q (t) 1 ] covered by biofilm. Fig. 1A shows, schematically, the evolution of the overall cathodic curve (i Vs. E ) during the gradual development of biofilm on the SS surface: curve 1 describes the oxygen reduction kinetics, i1(E ), measured at the beginning of the exposure to aerated seawater on a clean SS surface, whereas curve 4 shows the cathodic curve measured on an SS sample completely covered by biofilm. Curve 2 and 3 describe the trend of the cathodic current in two intermediate conditions. If, as suggested by 1), the evolution of cathodic current is only due to biofilm evolution, any technique able to signal the gradual cathodic depolarization from curve 1 to 4, in Fig. 1, can be utilized to build sensors which can provide information on biofilm growth. At least two classical techniques can be applied to this purpose: a potentiostatic technique or an intensiostatic one. Following Eq. (1), the potentiostatic technique provides information on biofilm development through the measurement of the cathodic currents on an SS sample polarized at a fixed potential E (Fig. 1B), whereas the intensiostatic technique provides similar information through the measurement of the potentials able to sustain a fixed cathodic current i during biofilm growth (Fig. 1C). The choice of the most suitable technique between the potentiostatic and the intensiostatic one, in a particular condition or environment, can depend on the specific biofilmrelated problem that has to be studied. Potentiostatic polarization was already proved to provide detailed information on the rate of biofilm development (Mollica et al., 1997; Faimali et al., 2008, 2010) from a q value less than 1% up to a complete covering of the SS surface. A possible defect of a sensor based on the potentiostatic technique is that a gradual carbonate precipitation is possible if the high cathodic current requested when biofilm is completely developed (in the order of 50 mA cm2) is sustained for a long time; it causes, in turn, a gradual decrease of the “active” surface of the sensor which must, hence, be periodically restored by acid cleaning. The intensiostatic technique, which can operate at cathodic currents lower than 1 mA cm2, avoids this inconvenient, but provides only the information that a specific biofilm covering threshold (e.g. 10% of the sensitive surface) was reached. Fig. 1C shows, in fact, that the shape of the curve potential Vs. time is similar to a sigmoidal curve which rises rapidly in a relatively narrow time range [t1 < t < t2 ; q(t1)< q < q(t2)],

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 5 1 e1 6 5 8

1653

Fig. 1 e Evolution of the overall cathodic curve (current i Vs. potential E ) during the gradual development of biofilm on a stainless steel surface [A], cathodic currents measured at a fixed potential E [B] and potentials measured at a fixed cathodic current i [C] during biofilm growth.

depending on the selected cathodic current density i. The inflection point of the curve can be used to define the threshold value of biofilm covering signaled by the sigmoidal curve obtained at a given cathodic current.

2.2.

Basing on the illustrated working principle, the ALVIM sensor can work both in potentiostatic and intensiostatic mode, but, given the previously mentioned considerations, the tests presented here have been performed using the intensiostatic mode. This is considered to be the best one for general industrial applications, since in this working mode the sensor requires less maintenance and gives a clear signal when biofilm covering exceeds the given threshold. In this case biofilm threshold was set to 1% of the working electrode, to test the device maximum sensitivity. The ALVIM probe (Fig. 2) is compact, requires little periodic maintenance, and can be adapted to fit different kinds of pipeline plug. The sensor is a three-electrodes system, in which the zinc counter-electrode (CE) plays also the role of pseudoreference (RE). Connected to the zinc and to the stainless steel working electrode (WE), on which biofilm growth is evaluated, there is an acquisition system, composed of three main parts: the first for substratum conditioning, the second for signal transduction and elaboration, the third for data transmission, over local/GSM/GPRS network. This system can be easily scaled and can manage up to several hundreds sensors at a time. The Biofilm Electrochemical Signal (BES, expressed as current density or potential), measured in real-time, at chosen time intervals, is sent to a remote database. It is possible to graphically visualize data and to raise different alarms in case of BES abrupt changes or achievement of a preset threshold value, corresponding to a defined biofilm covering percentage. The flexibility of this system allows to change electrodes dimensions and shapes, to fit different needs.

2.3.

Fig. 2 e ALVIM biosensor.

The biosensor

Biosensor preliminary characterization

Laboratory characterization has been performed at the ISMAR Marine Station (member of the European Network of Marine

1654

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 5 1 e1 6 5 8

Research Institutes and Stations e MARS), located in the Genoa harbor (Italy), in a tank of about 100 L in which seawater was constantly renewed (1.5/2 L min1) with natural water pumped directly from the sea. Three biofilm probe replications have been immersed for about 35 days, during the period SeptembereOctober (1st year), when seawater temperature ranged from 22 to 24.5  C, and BES was automatically registered every hour. After the reaching of the chosen biofilm covering threshold value (1%), probe sensitive surfaces were detached from the sensor to verify the effective biofilm covering.

2.4.

Biofilm covering evaluation

WE surface area covered by bacteria was quantified by means of epifluorescence microscopy and software analysis. After the detachment from the ALVIM probes, each sensitive surface was gently rinsed in seawater sterilized by filtration (Millipore, 0.22 mm pore size), in order to remove unattached cells, then fixed with 2% paraformaldehyde solution for 30 min, and washed in filtered phosphate buffer saline (PBS). Samples were stored at 4  C in PBS, before staining and microscopic analysis. After staining of bacterial cells with DAPI (40 -6-diamidino-2-phenylindole, Sigma) (Takata and Hirano, 1990), samples were observed at 400 magnification using an Olympus BX41 epifluorescence microscope coupled with an UV filter block for DAPI. A digital camera CAMEDIA 5060 (Olympus) was used to acquire 30 images of 67,500 mm2 each, randomly chosen on the surface of each sample. Images were converted to tiff format (RGB colour) and the surface fraction covered by bacteria was measured, on the 30 images, by means of “Image J” software (Rasband, 1997); mean  standard error (SE) was then calculated.

2.5.

Biofilm monitoring system in-plant testing

The following experiments took place in a FISIA e Italimpianti pilot reverse-osmosis (RO) desalination plant, located at the ISMAR Marine Station too. The pilot plant drew feeding water (1.7 m3/h) directly from the sea, with a first 100 mm prefiltration and a second 0.1 mm microfiltration (MF). After the MF there was a water storage tank, and then the RO. Three ALVIM probes were installed, by means of threaded locks, in the plant pipelines (Fig. 3):  the first in a newly installed pipeline, between prefiltration and microfiltration, where the biofilm was expected to grow sooner;  the second between MF and the storage tank, where the biofilm was expected to grow later or never, since all the

Fig. 3 e Scheme of FISIA e Italimpianti reverse-osmosis desalination plant and ALVIM probes disposition.

particles larger than 0.1 mm were filtered and this section was treated every few days with strong cleaning agents (NaClO, NaOH and HCl);  the third between the tank and the RO, where the biofilm was expected to grow, after an initial incubation, because the tank, positioned just before this section, represented a possible large-surface bacteria incubator, more suitable than pipeline for bacterial growth, since water flux was slower and temperature could slightly increase. These monitoring positions were therefore chosen to obtain a complete view of plant conditions, with reference to biofilm possible problems. A first 12-days trial (December of the 1st year) was conducted using all the three above mentioned probes. After some months (July of the 2nd year), a second 12-days trial was performed, employing only the first two probes, in the same plant, to better characterize the differences between biofilm growth dynamics before and after MF, verifying, at the same time, the effectiveness of this filtration against biofilm development. During the last days of this period, continuous chlorination (1 ppm) at the water intake was applied. In the course of the testing periods, pipelines pressure ranged from 0.2 to 0.8 bar, and seawater temperature from 11.6 to 16  C during the first trial, and from 23.5 to 27  C during the second trial. As for preliminary characterization, after the reaching of the chosen biofilm covering threshold value (1%), sensitive surfaces were detached from the probes to verify the effective biofilm covering.

2.6.

ALVIM system as a chlorination triggering device

The subsequent step, a few months later (December of the 2nd year), was the optimization of pipelines chemical cleaning treatments inside the above mentioned pilot plant, basing on biofilm growth real-time data collected by the ALVIM system. For this aim, biofilm growth signal from sensor no.1 was employed as a trigger to remotely start the 1 ppm chlorination at the water intake. During the 20-days testing period, pipelines pressure ranged from 0.2 to 0.8 bar, and seawater temperature from 11 to 15  C.

3.

Results and discussion

3.1.

Biosensor preliminary characterization

During preliminary testing, after just a few days of immersion in the seawater tank, biofilm probes showed an increase of the BES from around 500 mV Vs. Zn to more than 1100 mV Vs. Zn (Fig. 4), corresponding to a biofilm covering, quantified by microscopic analysis, of 3e4% of the WE surface. Considering that the sampling has been done some days after the reaching of the threshold value (marked in Fig. 4 by an asterisk), these data fit well with the chosen biofilm covering threshold (1% of the WE surface). These results are consistent both in terms of BES evolution curves (reasonably low standard error among replications and few differences among subsequent repetitions) and of biofilm covering (actual data match expected

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 5 1 e1 6 5 8

Fig. 4 e ALVIM probes BES evolution (mean ± SE, mV Vs. Zn) during preliminary testing in a tank with constantly renewed natural seawater. The asterisks mark the reaching of the chosen biofilm covering threshold (1% of the WE), the arrows mark sensitive surface sampling/ analysis (biofilm covering data are indicated in the boxes, as % ± SE, calculated on the three replications) and replacement.

values), confirming that the ALVIM sensor worked reliably over the considered time period.

3.2.

Biofilm monitoring system in-plant testing

After preliminary experiments, ALVIM testing proceeded in the pilot reverse-osmosis desalination plant (Fig. 5). After 2e5 days, BES of biosensors no.1 and no.3 started to increase, signaling that the biofilm covered more than 1% of WE surface. The BES rise occurred later in the stretch crossed by just prefiltered water (sensor no.1), but with a new and clean pipeline, than in the section after MF, between tank and RO (sensor no.3), never cleaned. This highlights the fact, suggested also by other experimental evidences (Donlan, 2002; Nikolaev and Plakunov, 2007; Sriyutha Murthy and Venkatesan, 2009), that already existing biofilm, inside

1655

pipelines, can have more influence in biofilm propagation/ development than new bacteria transported by feeding water, underlining the importance of an appropriate pipeline periodic cleaning. Sensor no.2, positioned in a section treated every few days with strong cleaning agents, did not show any biofilm growth signal, indeed. Microscopic examination showed that, after nine days of immersion, biofilm covering percentages on sensitive surfaces no.1 and no.3 were, respectively, about 1% and 2% (matching the chosen threshold of 1%). After the sampling, the plant was stopped, the section between MF and water storage tank was chemically cleaned (NaClO, NaOH); sensitive surface no.3 was replaced, while sensor no.1 was disconnected, for technical reasons related to the plant. At the end of the 12-days testing period, sensitive surfaces no.2 and no.3 (the last one replaced after the sampling on day 9) were nearly clean. The lower biofilm covering of sensitive surfaces no.1 and no.3 sampled on day 9, compared to those observed during preliminary testing (see Section 3.1), could be due to the different conditions, such as temperature, light, water filtration. The quick and precise ALVIM system response to biofilm growth, observed in the course of the preliminary characterization, was therefore confirmed during in-plant testing. Moreover, the BES showed a clear peak in correspondence to chlorination (marked in Fig. 5 by an oval), suggesting that ALVIM could also be used to monitor chlorine-based treatments. During the second testing period (Fig. 6), the signal of sensor no.1 started to increase one day earlier than in the first testing and, thereafter, grew at an higher rate. This reflects the different seasons, the different natural biological activity and, likely, possible differences of nutrients load in the water during the two tests, since the first one took place in December, while the second one in July. This observation confirms the flexibility of the employed biofilm monitoring system, underlining at the same time the impossibility of adopting a “one-for-all” cleaning treatment approach.

Fig. 5 e ALVIM probes BES evolution (mV Vs. Zn) during the first test in a pilot reverse-osmosis desalination plant. The arrow marks sensitive surfaces no.1 and no.3 sampling/analysis before plant stop and cleaning of the section between MF and water storage tank; on sensor no.3 a new sensitive surface was installed. The oval marks a chlorination in the same pipeline.

1656

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 5 1 e1 6 5 8

Fig. 6 e ALVIM probes BES evolution (mV Vs. Zn) during the second test in a pilot reverse-osmosis desalination plant. The circles mark chlorinations in the section between MF and water storage tank. The gray area marks plant stop for maintenance. The arrow on day 6 marks sensitive surface no.1 sampling, analysis and replacement. The arrow on day 9 marks the start of continuous chlorination (1 ppm) at the intake (whole pipeline treatment).

The signal of sensor no.2 did not increase over the baseline, except for the peaks observed when the signal was acquired during or nearby the chlorinations of the pipeline in which the sensor was placed. The height of those peaks was related with the proximity of the data acquisition moment with the chlorination time. The bio-electrochemical signal of sensor no.2 confirmed that the cleaning treatments carried out in this pipeline were effective against the biofilm; the bacterial covering of probe no.2 WE, at the end of the testing period, was nearly 0%, indeed. These observations sustain the effectiveness of ALVIM as monitoring device for chlorination treatments, as suggested by the data of the first testing period. During plant stop for maintenance (gray area in Fig. 6), the signal of sensor no.2 showed a drop, because the pipeline in which that sensor was placed got empty (no water). Just before plant restart, probe no.1 WE was sampled and replaced. The biofilm covering percentage was about 1%, fitting well, as the previous data, with the chosen threshold value. On day 9, a chlorination treatment (continuous, at 1 ppm) at the water intake was started, and the signal of sensor no.1, on which the biofilm had grown, nearly immediately decreased, confirming the effectiveness of that treatment. The signal remained 100e150 mV over the baseline, because of the chlorine continuous presence in the water. Chlorine was likely “consumed” before sensor no.2 by the organic matter present inside microfiltration module and pipelines, in fact this sensor did not show any signal increase. As previously mentioned, at the end of the testing period, biofilm covering on probe no.2 WE was nearly 0%.

3.3.

ALVIM system as a chlorination triggering device

During the last testing period, in line with what was observed during the above mentioned tests, after two days of incubation the BES started to grow (Fig. 7), signaling that the biofilm

surface covering, on the probe WE, reached the chosen threshold (1%). On day 4, the first ALVIM-triggered chlorination was started; the treatment time was set to 30 min. Immediately after this chemical cleaning treatment, the BES dropped to the initial value (around 700 mV Vs. Zn). In about two days, the biofilm growth signal increased again, and this time the 30-min chlorination was started in advance with respect to the previous one (the BES was around 900 mV Vs. Zn, while on day 4 it was nearly 1100 mV Vs. Zn). After the treatment, the sensor WE was immediately sampled and replaced. Biofilm covering on the sampled surface, quantified by laboratory analysis, was about 1%, matching the expected value. This evidence highlights the fact that, obviously, the applied chlorination treatment did not imply an immediate detachment of the biofilm from the surface. After WE substitution, it took four days to the BES to start growing again. Chlorination time was extended to 60 min, to verify the requested treatment frequency with a longer treatment time. Chlorination frequency continued to be about one treatment every two days, and, moreover, after the chemical cleaning performed on day 14 the BES did not return to the initial value, showing an increased after-treatment bio-electrochemical activity. This meant that the biofilm was not completely inactivated/killed by the chlorination, and implied also the need of more frequent treatments. Such information is essential to adjust timing and dosage of chemical cleaning, since even the survival of a small part of the settled bacteria is a guarantee of the fact that the biofilm will quickly grow again, thanks to the replication of the microorganisms still alive. These data suggest that, compatibly with environmental impact consideration and by law limit, higher chlorine concentrations or longer treatments had to be used, in this case, to completely remove the biofilm from the internal surfaces of plant pipelines.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 5 1 e1 6 5 8

1657

Fig. 7 e ALVIM probe BES evolution (mV Vs. Zn) during the use of ALVIM system as a chlorination triggering device. The arrows mark pipeline chemical cleaning by means of water chlorination (1 ppm) at the intake.

At the same time, the ALVIM system demonstrated its usefulness for the monitoring of biofilm growth and consequent cleaning treatments optimization inside water pipelines.

3.4. ALVIM compared to the other biofilm monitoring devices Comparing ALVIM sensor to the other biofilm monitoring devices, such as those, discussed in the introduction, based on light scattering (Flemming et al., 1998), turbidity (Klahre et al., 1998), electrochemical impedance (Mun˜oz-Berbel et al., 2006; Dheilly et al., 2008), vibration response of the monitored surface (Pereira et al., 2008) and diffusion limitation (Foret et al., 2010), it possible to see that the experimental results of this new device testing highlight significant advantages: - the possibility of monitoring just the biological fouling (microfouling), discerning it from the inorganic fouling; the above mentioned sensors, indeed, are not able to discriminate between these two different kinds of fouling; - the detection of the biofilm since its first colonization phase (i.e. the first bacterial layer); the above mentioned sensors detect only thicker (several mm) biofilms. Among the sensors based on electrochemical techniques, the most known are BIoGEORGE (Licina and Nekoska, 1993) and BIOX, born, earlier than ALVIM, from the same research activity (Cristiani et al., 1998). They usually show an higher sensitivity, with respect to those based on the previously mentioned techniques, but no clear quantitative data has been found in literature. Considering device flexibility, both BIoGEORGE and BIOX have a fixed working mode and sensitivity, while ALVIM can be set to monitor different extents of biofilm covering and can work both in intensiostatic and in potentiostatic mode. The last one, discussed only marginally in this work, will be the subject of future studies. From the technical point of view, the electronics of BIoGEORGE for the control, data acquisition and data analyses are

housed in an external box, where the readings are stored in a database (DB). In this way data are not available in real-time from remote, but has to be downloaded to a PC. The BIOX sensor needs external hardware too, moreover device control and data reading are basically analogical (Cristiani et al., 1998, 2000; Cristiani, 2005). On the other hand, ALVIM has a fully digital management, and its electronic is completely integrated within sensor housing; in industrial environments, indeed, device compactness represents a valuable advantage. About data storage, ALVIM uses a remote DB, thus the collected information can be viewed in real-time even from remote.

4.

Conclusions

Experimental results show that the ALVIM system works reliably in a real industrial environment, representing an efficient biofilm monitoring solution; it gives a fast and accurate information about the bacterial covering, even at early stages of colonization. Furthermore, the data provided by this system proved to be very useful if applied to cleaning treatments optimization, enabling to hinder biofilm growth as soon as it starts. This is a promising technology in any field affected by biofilm-related problems, prefiguring a wide application range for the ALVIM system. Next biosensor developments will concern longer trials, experiments in different conditions (e.g. freshwater, other industrial environments) and testing of new materials for biosensor components.

Acknowledgements Authors wish to thank FISIA e Italimpianti and University of Genoa e Department of Chemistry and Industrial Chemistry staff for the contribution to the ALVIM system field testing.

1658

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 5 1 e1 6 5 8

references

Acun˜a, N., Ortega Morales, B.O., Valadez-Gonzalez, A., 2006. Biofilm colonization dynamics and its influence of austenitic UNS S31603 stainless steel exposed to Gulf of Mexico seawater. Mar. Biotechnol. 8, 62e70. Borenstein, S.W., 1994. Microbiologically Influenced Corrosion Handbook. Industrial Press Inc., New York, p. 288. Cristiani, P., Mollica, A., Ventura, G., 2000. System for Monitoring Biocide Treatments. U. S. Patent 6,054,030. Cristiani, P., 2005. Solutions to fouling in power station condensers. Appl. Therm. Eng. 25, 2630e2640. Cristiani, P., Mollica, A., Ventura, G., 1998. On-line Monitoring of Biofilm and TRO by a New ENEL’s System, Proceedings of OCEANS ’98, Nice, France, pp. 1507e11. Dexter, S.C., Zhang, H.J., 1990. Effect of Biofilms on Corrosion Potential of Stainless Steel Alloys in Estuarine Waters, Proceedings of the 11th International Corrosion Congress, Florence, Italy, pp. 333e40. Dexter, S.C., La Fontaine, J.P., 1998. Effect of natural marine biofilms on galvanic corrosion. Corrosion 54, 851e861. Dheilly, A., Linossier, I., Darchen, A., Hadjiev, D., Corbel, C., Alonso, V., 2008. Monitoring of microbial adhesion and biofilm growth using electrochemical impedancemetry. Appl. Microbiol. Biotechnol. 79, 157e164. Donlan, R.M., 2002. Biofilms: microbial life on surfaces. Emerg. Infect. Dis. 8, 881e890. Dulon, S., Parot, S., Delia, M.L., Bergel, A., 2007. Electroactive biofilms: new means for electrochemistry. J. Appl. Electrochem. 37, 173e179. Faimali, M., Chelossi, E., Garaventa, F., Corra`, C., Greco, G., Mollica, A., 2008. Evolution of oxygen reduction current and biofilm on stainless steel cathodically polarized in natural aerated seawater. Electrochim. Acta 54, 148e153. Faimali, M., Chelossi, E., Pavanello, G., Benedetti, A., Vandecandelaere, I., De Vos, P., Vandamme, P., Mollica, A., 2010. Electrochemical activity and bacterial diversity of natural marine biofilm in laboratory closed-systems. Bioelectrochemistry 78, 30e38. Flemming, H.C., Shaule, G., 1994. Microbial deterioration of materials e biofilm and biofouling: biofouling (Mikrobielle Werkstoffzersto¨rung e Biofilm und Biofouling: biofouling). Mater. Corros. 45, 29e39. Flemming, H.C., Tamachkiarowa, A., Klahre, J., Schmitt, J., 1998. Monitoring of fouling and biofouling in technical systems. Water Sci. Technol. 38, 291e298. Foret, C., Merlet, N., Chaussec, G., Martemianov, S., Tribolet, B., Hater, W., 2010. Measurement of Biofilm Thickness: An Effective Legionella Risk Assessment Tool. BKG Technical Publication. http://www.bkgwater.com/clients/bkgwater/ upload/fichiers/2010-02-bkgws_publication-measurement-ofbiofilm-thickness.pdf. Fritzmann, C., Lo¨wenberg, J., Wintgens, T., Melin, T., 2007. State-ofthe-art of reverse osmosis desalination. Desalination 216, 1e76. Geesey, G.G., Lewandowski, Z., Flemming, H.C., 1994. Biofouling and Biocorrosion in Industrial Water Systems. CRC Press, Boca Raton, p. 297. Gilbert, P., Das, J., Foley, I., 1997. Biofilm susceptibility to antimicrobials. Adv. Dent. Res. 11, 160e167. Kimio, I., Matsuhashi, R., Toshiaki, K., Miki, O., Kihira, H., Watanabe, K., Baker, P., 2002. Potential Ennoblement of Stainless Steel by Marine Biofilm and Microbial Consortia Analysis Proceedings of Corrosion/2002, Denver, Colorado, Paper No. 403. Klahre, J., Lustenberger, M., Flemming, H.C., 1998. Microbial problems in paper manufacturing. Part III: monitoring (Mikrobielle Probleme bei der Papierfabrikation. Teil III: monitoring). Das Papier 52, 590e596.

Licina, G.J., Nekoska, G., 1993. An Electrochemical Method for OnLine Monitoring of Biofilm Activity Proceedings of CORROSION/93, Houston, Texas, Paper No. 403. Little, B.J., Lee, J.S., Ray, I.R., 2008. The influence of marine biofilms on corrosion: a concise review. Electrochim. Acta 54, 2e7. Mattila, K., Carpen, L., Hakkarainen, T., Salkinoja-Salonen, M.S., 1997. Biofilm development during ennoblement of stainless steel in Baltic sea water: a microscopic study. Int. Biodeterior. Biodegradation 40, 1e10. Maxwell, S., 2005. Controlling Corrosive Biofilms by the Application of Biocides SPE International Symposium on Oilfield Corrosion, Aberdeen, UK. Mollica, A., Trevis, A., 1976. Correlation Between the Formation of the Primary Film and the Change in the Cathodic Reaction on Stainless Steels in Seawater at Speeds Between 0.3 and 5.2m/s (Correlation entre la formation de la pellicule primaire et de la modification de la cathodique sur des aciers inoxydables experimente´s en eau de mer aux vitesses 0.3 a` 5.2m/s), Proceedings of the Fourth Congress on Marine Corrosion and Fouling, Antibes, France, pp. 351e365. Mollica, A., Traverso, E., Thierry, D., 1997. In: On Oxygen Reduction Depolarisation Induced by Biofilm Growth on Stainless Steels in Seawater. European Federation of Corrosion publications N 22. The Institute of Materials, London, pp. 51e65. Mun˜oz-Berbel, X., Mun˜oz, V., Vigue´s, N., Mas, J., 2006. On-chip impedance measurements to monitor biofilm formation in the drinking water distribution network. Sens. Actuators B 118, 129e134. Nikolaev, Y.A., Plakunov, V.K., 2007. Biofilms e ‘city of microbes’ or an analogue of multicellular microorganims? Microbiology 76, 125e138. Parr, J.G., Hanson, A., 1965. An introduction to stainless steel. ASM International, Metals Park, p. 147. Pereira, A., Mendes, J., Melo, L.F., 2008. Using nanovibrations to monitor biofouling. Biotechnol. Bioeng. 99, 1407e1415. Prince, E.L., Muiry, A.V.G., Thomasz, W.M., Stollardx, R.J., Sampsonx, M., Lewis, J.A., 2002. An evaluation of the efficacy of aqualox for microbiological control of industrial cooling tower systems. J. Hosp. Infect. 52, 243e249. Rasband, W.S., 1997. ImageJ. US National Institutes of Health. Bethesda, Maryland. http://rsb.info.nih.gov/ij/. Schultz, M.P., Swain, G.W., 2000. The influence of biofilms on skin friction drag. Biofouling 15, 129e139. Scotto, V., Di Cintio, R., Marcenaro, G., 1985. The influence of marine aerobic microbial film on stainless steel corrosion behaviour. Corros. Sci. 25, 185e194. Sriyutha Murthy, P., Venkatesan, R., 2009. Industrial biofilms and their control. In: Flemming, H.C., Venkatesan, R., Sriyutha Murthy, P. (Eds.), Marine and Industrial Biofouling. Springer, New York, pp. 65e101. Takata, K., Hirano, H., 1990. Use of fluorescein-phalloidin and DAPI as a counterstain for immunofluorescence microscopic studies with semithin frozen sections. Acta Histochem. Cytochem. 23, 679e683. Vrouwenvelder, J.S., Manolarakis, S.A., van der Hoek, J.P., van Paassen, J.A.M., van der Meer, W.G.J., van Agtmaal, J.M.C., Prummel, H.D.M., Kruithof, J.C., van Loosdrecht, M.C.M., 2008. Quantitative biofouling diagnosis in full scale nanofiltration and reverse osmosis installations. Water Res. 42, 4856e4868. Wang, W., Wang, J., Li, X., Xu, H., Wu, J., 2004. Influence of biofilm growth on corrosion potential of metals immersed in seawaters. Mater. Corros. 55, 30e35. Whitehouse, R.L., Peters, E., Lizotte, J., Lilge, C., 1991. Influence of biofilms on microbial contamination in dental unit water. J. Dent 19, 290e295. Wirtanen, G., Salo, S., Helander, I.M., Mattila-Sandholm, T., 2001. Microbiological methods for testing disinfectant efficiency on Pseudomonas biofilm. Colloids Surf. B 20, 37e50.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 5 9 e1 6 6 8

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Efficacy of monitoring and empirical predictive modeling at improving public health protection at Chicago beaches Meredith B. Nevers*, Richard L. Whitman U.S. Geological Survey, Great Lakes Science Center, Lake Michigan Ecological Research Station, 1100 N. Mineral Springs Road, Porter, IN 46304, USA

article info

abstract

Article history:

Efforts to improve public health protection in recreational swimming waters have focused

Received 17 August 2010

on obtaining real-time estimates of water quality. Current monitoring techniques rely on

Received in revised form

the time-intensive culturing of fecal indicator bacteria (FIB) from water samples, but

3 December 2010

rapidly changing FIB concentrations result in management errors that lead to the public

Accepted 6 December 2010

being exposed to high FIB concentrations (type II error) or beaches being closed despite

Available online 13 December 2010

acceptable water quality (type I error). Empirical predictive models may provide a rapid solution, but their effectiveness at improving health protection has not been adequately

Keywords:

assessed. We sought to determine if emerging monitoring approaches could effectively

E. coli

reduce risk of illness exposure by minimizing management errors. We examined four

Fecal indicator bacteria

monitoring approaches (inactive, current protocol, a single predictive model for all bea-

Recreational water quality

ches, and individual models for each beach) with increasing refinement at 14 Chicago

Lake Michigan

beaches using historical monitoring and hydrometeorological data and compared

Swimming

management outcomes using different standards for decision-making. Predictability (R2) of

Risk

FIB concentration improved with model refinement at all beaches but one. Predictive models did not always reduce the number of management errors and therefore the overall illness burden. Use of a Chicago-specific single-sample standarddrather than the default 235 E. coli CFU/100 ml widely useddtogether with predictive modeling resulted in the greatest number of open beach days without any increase in public health risk. These results emphasize that emerging monitoring approaches such as empirical models are not equally applicable at all beaches, and combining monitoring approaches may expand beach access. Published by Elsevier Ltd.

1.

Introduction

In recent years, efforts to improve public health protection in recreational swimming waters have focused on obtaining realtime estimates of water quality. Current monitoring techniques rely on the culturing of fecal indicator bacteria (FIB)d

such as Escherichia coli or enterococcidfrom water samples, a process that requires an incubation time often in excess of the rate of change of bacteria concentrations in the water (Boehm et al., 2002; Whitman et al., 1999). Because of the lapse in results availability, the public are often either unknowingly swimming in contaminated beach water or are prohibited from

Abbreviations: FIB, fecal indicator bacteria; CFU, colony-forming units; MPN, most probable number; IA, inactive monitoring program model; CM, current model; RM, regional predictive model for all study beaches; IM, individual beach predictive model. * Corresponding author. Tel.: þ1 219 926 8336x425; fax: þ1 219 929 5792. E-mail address: [email protected] (M.B. Nevers). 0043-1354/$ e see front matter Published by Elsevier Ltd. doi:10.1016/j.watres.2010.12.010

1660

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 5 9 e1 6 6 8

swimming in water that meets the public health criteria. Efforts have been focused on two means of correcting this shortcoming: shorten the analytical time for the current indicator or find an alternate, faster way to assess water quality. To accomplish the latter, empirical predictive models have been attempted with various levels of success and application. Predictive models have been suggested by numerous authors as a potential means for minimizing errors in beach closings (Hou et al., 2006; Kim and Grant, 2004; Nevers and Whitman, 2005). These models range from simple models that associate weather conditions with direct bacteria loadingsesuch as rainfall and associated runoff (McPhail and Stidson, 2009) e to more advanced models that integrate multiple hydrometeorological variables (Kim and Grant, 2004). Model accuracy at predicting FIB concentration depends on beach location, instrument accuracy, wealth of available data, and level of effort, but predictive models can be successfully incorporated into beach management (Nevers and Whitman, 2005). Beaches at which models have been attempted tend to be high profile beaches with heavy visitor use (Boehm, 2007; Hou et al., 2006), directly or strongly impacted by a large point source (He and He, 2008), or having frequent swimming advisories (Olyphant and Whitman, 2004). The accuracy or success of a given modeling approach has typically been assessed by analyzing the amount of variation in the target FIB explained by the model, the error explained by the model, or the specificity (the percent of false negatives, or type II errors) and sensitivity (the percent of false positives, or type I errors) of the model. The first two calculations determine the accuracy of the model at predicting all FIB concentrations. The third, error-based calculation is used due to the use of a binary approach in beach management policies: beaches are either open or closed to swimming, depending on where the FIB concentration falls relative to a designated standard (acceptable health risk); errors occur when the predicted concentration is not equal to actual concentration. Errors result in either inadvertent exposure of the public to high concentrations of FIB (type II error) or exclusion of swimmers from water that meets the exposure standard (type I error). More type II errors result in more swimmers exposed to high concentrations of FIB and therefore a higher public health risk; decreasing the instances of type II errors is necessary to increase public health protection. Current water quality standards for freshwater were developed using epidemiological studies and based on historical acceptable illness rate (Pru¨ss, 1998, US EPA, 1986). Within the monitoring guidance, however, some measure of flexibility was provided for beach managers, including choice of application of two mathematical estimates of illness risk, based on the concentration of indicator bacteria (US EPA, 1986). Generally, beach managers have applied the single-sample maximum for an individual water sample because of its ease of use and interpretation (Nevers and Whitman, 2010, US EPA, 1986), but others use the 5-day geometric mean, both of which should theoretically provide equal levels of health protection. In this paper, we examine four potential monitoring approaches with increasing refinement at 14 Chicago beaches: inactive, current monitoring model, use of one predictive model for all beaches, and use of individual predictive models for each beach. Further, we examine alternate applications of

monitoring standards under these four approaches to assess the health and management outcome possibilities. Using historical monitoring and beach attendance data we compare the accuracy of each model with several calculations and also the relative public health protection provided by each of these models. Specifically, we sought to determine whether predictive modeling at Chicago beaches could be used as a monitoring tool to increase public health protection over traditional monitoring practices.

2.

Materials and methods

2.1.

Study site

Chicago beaches in general are not impacted by a major point source of contamination. Urban sewage is regularly discharged through the Chicago River and a series of man-made or modified canals to the Mississippi River. In events of extreme precipitation, the system override leads to sewage being directed to Lake Michigan (<1 per year); all beaches are then preemptively closed to swimming. Sources of FIB at the Chicago beaches are unknown but likely include beach sand, birds, and algae (Whitman and Nevers, 2003; Whitman et al., 2003). Beaches included in the current study were (from north to south) Loyola, Albion, Hollywood, Foster, Montrose, North Avenue, Oak, 12th Street, 31st Street, 57th Street, 63rd Street, South Shore, Rainbow, and Calumet.

2.2.

Beach monitoring data

E. coli monitoring data, measured as most probable number (MPN)/100 ml of water, were obtained from the Chicago Park District for 2000e2004. Beaches were sampled at least five days a week; replicate samples (up to three) were averaged. E. coli concentrations above or below detection limits were set at detection limits after determining that occurrences were rare (Boehm et al., 2002; Whitman and Nevers, 2008). Missing data points for individual beaches were calculated; values were estimated from the nearest 6 values (average of three previous and three subsequent readings). Beach management is a binary decision: if E. coli concentration <235 MPN/100 ml, the beach is open for swimming; if E. coli >235 a swimming advisory is issued. This model assumes that E. coli concentration today ¼ E. coli concentration yesterday. Inaccurate predictions, therefore, result in a type I or type II error (Table 1). Type I errors occur when the model predicts an E. coli concentration >235 when the actual concentration is <235, resulting in an unnecessary swimming advisory. A type II error occurs when the model predicts E. coli concentration <235 when the actual concentration is >235, resulting in swimmers being exposed to high concentrations of indicator bacteria and associated pathogens. A simple characterization is that type I errors are associated with economic losses because swimmers are denied access and type II errors are associated with greater public health risk, as swimming occurs in the presence of excessive FIB concentrations (Rabinovici et al., 2004).

1661

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 5 9 e1 6 6 8

2.3.

Predictive models

Data for developing predictive models were accumulated from several existing hydrometoerological stations, as described in Whitman and Nevers (2008). Predictors included solar insolation and precipitation (24 h); air temperature, barometric pressure, and wave height (mean for 4e10 AM); and day of year. All E. coli results were log-transformed prior to analysis. Predictor parameters were transformed to z-scores prior to regression analyses. Models were developed using multiple linear regression of the available predictors on the independent variable E. coli; Mallows Cp was used to compare resulting models. For individual models, Akaike’s Information Criterion (AIC) was used to select the best fit model for each individual beach. The regional model was developed previously (Whitman and Nevers, 2008) using barometric pressure, wave height, and day of year as the predictors; here, it was applied separately to each study beach. Model comparisons were made using SPSS (Chicago, IL) and SAS (Cary, NC) software: the adjusted coefficient of determination (R2), which describes the amount of variation in E. coli; and percent of type I and type II errors. The four models compared included inactive model (IA): beach is always open to swimming, regardless of microbiological water quality; current model (CM): E. coli concentration for day 1 is used to make a management decision for day 2; regional model (RM): a universal model used for all beaches; and an individual model (IM): a separate predictive model developed for each beach. Comparisons between models for estimating excess illnesses associated with type II errors were only made on days for which there were results for all four models.

2.4.

Calculating illness rates

Overall beach attendance rates were compared for 20 years, and an average per year estimate was calculated (Chicago Park District, unpublished data). With an annual reported range of 14e31 million visitors to all Chicago beaches, 20 million was used as an average. An estimated 91% of all visits, or 18.2 million, were associated with the 14 beaches included in this study. Number of visitors per beach was calculated based on two years for which individual beach data were available (1999 and 2007). It was assumed that 50% of beach visitors had full-body contact with the water as defined by US EPA (Dufour, 1984).

Lower estimates have been published (Rabinovici et al., 2004), but based on data from several Ontario lakes (Seyfried et al., 1985), estuaries (Lepesteur et al., 2006), and marine waters (Dwight et al., 2007; Given et al., 2006), 50% may be conservative for a large freshwater lake. Illness rates for each beach were calculated based on a 100-day swimming season, which assumes equal distribution of visits across the summer (Rabinovici et al., 2004). Illness rate (Y ) was calculated from Dufour (1984): Y ¼ 11:74 þ 9:397log10 ðECÞ

(1)

where Y ¼ the rate of swimming-associated gastrointestinal illness symptoms per 1000 swimmers and EC ¼ E. coli CFU/ 100 ml water. Monitoring standards based on the these epidemiological studies recommend a geometric mean of 126 E. coli CFU/100 ml for five samples over 30 days (Dufour, 1984): an acceptable illness rate of approximately 0.008%. A singlesample limit of 235 CFU/100 ml is also provided, which is within the confidence limits of the calculated geometric mean. These calculations were developed for beaches influenced by a point source (Dufour, 1984), but the sources affecting Chicago’s beaches have not been confidently identified. Using the derived regression equation (Dufour, 1984), we calculated our acceptable illness rate for the 235 CFU as 0.01054%, following Rabinovici et al. (2004). It should be noted that the standards were established using membrane filtration analysis. Chicago uses a defined substrate technique (Colilert; IDEXX, Westbrook, Maine); and while studies have favorably compared the two outcomes (Buckalew et al., 2006), differences in confidence intervals may influence results outcome (Gronewold et al., 2008). A daily excess illness rate was calculated following Given et al. (2006), that is, the number of beach swimmers expected to exhibit symptoms of illness beyond the acceptable 0.01054%. GI ¼ ðY  Y0 Þðv=dÞf

(2)

Where Y0 ¼ acceptable illness rate within the monitoring criteria of 10.54/1000, v ¼ number of visitors in a swimming season, d ¼ number of swimming days in the season, and f ¼ percent of beach visitors estimated to have full-body contact with the beach water. The illness regression equation has a y-intercept <0, so calculated illness rates <0 were set at 0 for calculating the mean. There is conflicting research on the association of illness risk and traditional fecal indicators at beaches without a point

Table 1 e Outcome possibilities using the current beach management model: management decision based on previous day’s E. coli concentration. Management outcome

Error

Correct open Correct closed Incorrect open Incorrect closed

None None type 2 type 1

Health risk Loss of Use

Low High High Low

Low High Low High

Health protection Accurate Accurate Overly liberal Overly conservative

Range of reported outcome Range of mean illness rate frequency for for Chicago beaches Chicago beaches (swimmers/1000) 49e78% 3e14% 12e21% 9e17%

2.7e5.0 13.6e15.3 14.4e16.1 4.2e6.4

1662

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 5 9 e1 6 6 8

source (Calderon et al., 1991; Sinigalliano et al., 2010). Questions about the appropriateness of using E. coli as an indicator have also been raised (Harwood et al., 2005; Wong et al., 2009), but a review of epidemiological studies associate elevated E. coli with higher risk of gastrointestinal illness (Marion et al., 2010; Wade et al., 2003). Because alternate standards have not been established, all beaches are managed using the E. coli standard, despite its shortcomings. Chicago does not issue a swim ban until E. coli concentration exceeds 1000 MPN/100 ml. Currently, there are no estimates of the percent of swimmers who enter the water during a swimming advisory (i.e., 235 < E. coli <1000) (Hou et al., 2006). Here, we elect to use >235 to express excess illnesses and define model errors because it is the more conservative calculation. Rate of excess illness would increase by 0.0059% using the 1000 CFU standard.

2.5. Different management applications of monitoring standard In an attempt to improve model performance, the E. coli standard for the binary outcome was targeted. As described previously, monitoring standards recommended a five-day geometric mean, and a general single-sample maximum was provided (US EPA, 1986). The guidance suggested, however, that single-sample maximums be calculated for individual jurisdictions based on local log standard deviation of E. coli concentration (US EPA, 1986). The calculation provided in the ambient criteria (US EPA, 1986) is  ss ¼ 10^ log10 GM þ cl  log10 sd

14/day. Overall, for the 14 beaches, a mean of 35 illnesses/day could be expected (range: 0e659  66 SD).

3.2.

Amount of E. coli variation explained by models (R2)

Coefficient of determination was lowest for the CM, with a range of 0.049 (South Shore) to 0.135 (63rd Street) for individual beaches and an overall adjusted R2 of 0.141 (Fig. 2a). The higher R2 for 63rd Street results from the persistent high E. coli concentrations at this location with little variation. Generally, more variation in E. coli concentrations could be explained using the RM, resulting in a range of R2 from 0.111 (63rd Street) to 0.287 (North Avenue). Further improvement in R2 was seen with the IM (Fig. 2a). The model for 63rd Street included the lowest amount of variation explained (0.141); the highest adjusted R2 was at North Avenue (0.313). The general pattern was an increasing R2 with increasing model refinement, seen at all beaches except 63rd Street. The change in R2 between the CM and the RM was greatest for Hollywood and South Shore beaches, with notable improvements at Foster, North Avenue and South Shore. Change in R2 between RM and IM was not as great overall; most improvement was at the south side beaches. Precision of the predictions increased with model refinement as seen in a decreasing RMSE.

3.3. Binary water quality standard outcome (number of prediction errors) Using the CM, the beach was correctly left open to swimming 68% of the time (range ¼ 49e78%) (Fig. 3). This increased to 78%

(3)

where ss ¼ the single-sample maximum; GM ¼ 126, geometric mean E. coli concentration for acceptable illness rate of 8 per 1000; cl ¼ 0.675, the 75% calculated one-sided confidence level for a designated heavily used beach area, and sd ¼ is the calculated standard deviation for a given jurisdiction, 0.718 for Chicago beaches (mean log E. coli ¼ 1.776) Chicago’s singlesample maximum would therefore be 385 CFU/100 ml.

3.

Results

Visits to Chicago’s lakefront beaches are often in excess of 27,000,000 annually, with fully half of the visits associated with two beaches: Oak Street and North Avenue (Chicago Park District, unpublished data).

3.1.

Baseline water quality

Chicago beaches exceed the 235 single-sample maximum 14e35% of the time, with an overall rate of 20%. Given the current water quality and visitation of Chicago’s beaches and the hypothetical absence of any monitoring or associated beach closures (IA model), the majority of expected illnesses were associated with the high visitation beaches (Fig. 1). The highest illness rate was 138/day for North Avenue, and the lowest was 7/day for Rainbow. Despite having the highest mean E. coli concentration, 63rd Street illness rate averaged

Fig. 1 e Comparison of mean E. coli concentration at each beach and mean number of swimmers that can be estimated to develop gastrointestinal illness each day. Gradation of circles indicates the expected individual illness rate. Beaches include Loyola (LY), Albion (AL), Hollywood (HW), Foster (FO), Montrose (MO), North Avenue (NA), Oak (OK), 12th Street (12), 31st Street (31), 57th Street (57), 63rd Street (63), South Shore (SS), Rainbow (RB), and Calumet (CA).

1663

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 5 9 e1 6 6 8

Fig. 2 e Comparison of (a) coefficient of variations, (b) percent type I errors, and (c) percent type II errors by modeling approach.

(range ¼ 53e85%) with the RM and was 77% (range ¼ 54e84%) with the IM. Rate of correct closures under the CM is 6% (range ¼ 3e14%). That decreased to 0.8% (0.3e9) with the RM, and with the IM, the rate was 1.4% (0.3e11%). Type I errors, in which a swimming advisory is posted although E. coli concentration is lower than the single-sample maximum, were made 11.5% (9e17%) of the time using the CM for monitoring (Fig. 2b). With the application of an RM, that

inactive

rate decreased to 2.2% (0.4e4) of the time, and with the IM application, the rate was 3.2% (1e11%). Type II errors were common under the CM, occurring 14% (12e21%) (Fig. 2c). That number increased with both predictive models, to 19% (16e27) with the RM and 18.5% (14e24) with the IM. Comparison of the percent of type II errors showed highly variable results. The greatest percent change between models was an increase in percent type II errors at 12th Street and

current

outcome correct closed correct open type I error type II error

region

individual

Fig. 3 e Management outcome using the binary approach, i.e., an E. coli concentration >235 results in a beach closure.

1664

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 5 9 e1 6 6 8

Loyola between CM and RM. Minimizing type II errors between the RM and IM was limited.

3.4. Association of monitoring and modeling approaches with illness burden A comparison of the four models reveals, most noticeably at the high visitation beaches North Avenue, Montrose and Oak Street, that the leveling off of number of type II errors was mirrored in the rate of excess illnesses prevented with increasing model refinement. There was no significant difference in rate of excess illnesses associated with model approach at any of the individual beaches. In fact, model refinement between RM and IM showed little improvement in the number of excess illnesses prevented. At several beaches, cumulative illnesses were identical for all models except for CM. Eight of the beaches showed a difference in cumulative excess illnesses associated with the four models (Table 2). The CM was associated with lower cumulative excess illnesses, largely because there were fewer instances of type II errors; however, for all beaches but 63rd Street, the CM was associated with higher mean excess illnesses, indicating that with this model, the instances of very high E. coli concentrations tend to be missed. The IA model excluded type I errors because the beach would never be closed to swimming and therefore swimmers would not be prevented from water contact under any water quality conditions. Increasingly refined models were effective at reducing type I errors because they rarely predicted events when water quality exceeded the single-sample standard; most predictions were low E. coli concentrations. This tendency is apparent in the rapid decrease in number of type I errors with increasing model refinement.

3.5.

Chicago-specific monitoring standard

With the more liberal, jurisdiction-specific single-sample maximum (385 CFU/100 ml), the overall percentage of errors for all model types decreased (Fig. 4). Reduction in percent type II errors was more pronounced with the RM and IM (mean decrease by 35e36%) than for the CM (mean decrease by 22%). Dramatic reductions in percent of type II errors (>45%) occurred at some of the beaches with lower mean E. coli concentrations (North Avenue, Oak, Albion) with the RM and IM. The difference in number of illnesses should theoretically

remain unchanged because the illness rate is within the confidence limits of the original derived regression. Because the predictive models tend to predict low overalldthe tendency leading to type II errorsdthe reduction in type I errors was not as remarkable.

4.

Discussion

With recent findings that the time-intensive current beach monitoring models are not generally predictive of real-time FIB concentration (Boehm, 2007; Whitman et al., 1999), beach managers need a means of determining water quality rapidly and efficiently in order to protect public health while maximizing beach access. A suite of approaches for expanding monitoring activities and improving timeliness of monitoring results have been proposed and considered, but minimal research has explored their capacity to meet this goal. If a new method does not provide a significant improvement over current management outcomes, it is unlikely that managers will invest the necessary effort to alter the monitoring program. The first step for many jurisdictions is to implement a water quality monitoring program using the currently used model (CM). According to our results, this activity alone decreased the number of excess illnesses at all beaches, largely as a result of keeping swimmers out of the water more often, regardless of actual water quality. Beaches were closed when water quality was within acceptable standards 9e17% of the time, a scenario that can incur a social and economic burden (Rabinovici et al., 2004). With this model, accurate closures and health protection are strictly dependent on the endurance of an E. coli contamination event. With the rapidly changing nature of water quality (Boehm, 2007; Boehm et al., 2002; Whitman and Nevers, 2004), there is high potential to make a management error (Table 1). Multiple day contamination events would result in some correct swimming advisories (correct closed), and extended periods of low E. coli concentrations would result in accurate swimming permission (correct open). The vast majority of contamination events, however, last one day or less (Leecaster and Weisberg, 2001). In Chicago, only 6% of contamination events persisted for more than one sampling day, with the exception of 63rd Street: 17%. Monitoring at beaches with persistent contamination will inevitably decrease the number of type II errors and associated illnesses because swimmers are

Table 2 e Comparison of the cumulative (mean) excess illnesses as a result of the choice of model approach: inactive, current, regional, or individual beach model. Results based on days of type II errors, ranging from 38 to 103 days of the 247e292 days (2000e2004 seasons) considered in the analysis. Inactive Montrose North Ave 31st 57th 63rd South Shore Rainbow Calumet

2492.94 4540.82 366.22 912.68 704.59 551.00 240.73 1049.60

(42.25) (105.60) (5.39) (16.90) (6.84) (10.20) (3.70) (20.58)

Current 1695.49 (42.39) 3659.66 (107.64) 250.73 (5.57) 758.64 (17.64) 389.27 (6.38) 456.19 (11.13) 176.09 (3.91) 869.01 (22.87)

Regional 2407.17 4540.82 348.00 912.51 548.65 551.00 240.73 1049.32

(41.50) (105.60) (5.52) (17.55) (6.94) (10.20) (3.70) (20.99)

Individual 2442.56 4270.00 320.70 894.31 490.04 537.40 217.60 1023.04

(42.85) (101.69) (5.34) (17.54) (6.71) (10.14) (3.69) (20.88)

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 5 9 e1 6 6 8

1665

Fig. 4 e Impact on beach advisory decisions if jurisdiction-specific water quality standard is used. The two beach examples include 63rd Street Beach, which has a high mean E. coli, lower visitation rate, and frequent swimming advisories, and North Avenue Beach, which is the most highly visited Chicago beach and has low mean E. coli and infrequent swimming advisories. Use of a Chicago-specific water quality standard would result in fewer errors overall using all three monitoring model approaches.

accurately kept from the water for some percent of the high concentration events. Likewise, beaches with infrequent contamination events will have a limited number of type I errors. To overcome these inconsistencies in actual vs. presumed water quality, predictive models have been used at some beaches to provide a real-time estimate of water quality and overcome these inherent errors. The cornerstone of predictive modeling for beach management has been the individual beach model (Nevers and Whitman, 2005), but the cost of developing such models has been one of the disincentives to widespread use; many of the models use on-site instrumentation that requires installation and continuous maintenance (Francy, 2009). For this reason, regional models have been explored, which examine the hydrometeorological factors similarly affecting groups of beaches and describe background fluctuations in FIB concentrations (Nevers and Whitman, 2008). While more crude in estimations, regional models may provide a cost-effective solution for jurisdictions with numerous monitored beaches while providing insights into source behavior. Perhaps surprising, the RM results in this exercise were quite similar to results from IM, indicating a generally predictable fluctuation in FIB concentrations across Chicago. Overall, increasing refinement of monitoring approach with the use of predictive models was associated with improved accuracy of E. coli predictions. Use of the RM increased the amount of variation explained over the current monitoring approach, and the use of beach-specific IM somewhat further improved this result. With beach-specific refinement, predictive models for all of the study beaches had higher R2 and lower RMSE: more variation in individual E. coli concentrations was

described and there was lower error in this estimation. In this study, geographically widespread predictors were used, which may have limited each model’s ability to detect beach-specific variation, although the low R2 for 63rd Street was somewhat expected due to inherent high variation in E. coli and the complex circulation pattern at this enclosed beach (Ge et al., 2010; Whitman and Nevers, 2004). Models developed for this beach have depended on higher frequency and higher intensity local data than were available for this exercise (Boehm et al., 2007; Olyphant and Whitman, 2004). Predictability improved significantly with the use of the RM for many of the beaches, with the biggest improvement between CM and RM at the north-side beaches, but it was the further refinement to IM that resulted in the greatest improvement at southern beaches. This pattern supports the idea that there is more beach-specific variation at these southern beaches and E. coli concentrations perhaps recover to background concentrations more slowly than at the northern beaches (Whitman and Nevers, 2008). Use of individual predictive models at these beaches can take these factors into account, resulting in better predictability. The number of type II errors for each model was highly variable, lacking the pattern of improvement with increasing refinement shown in the R2. All of the models failed to predict the majority of high E. coli concentrations. Because many beaches have infrequent high E. coli concentration events, it is difficult to detect a pattern of associated hydrometeorological conditions; this is a problem for many predictive modeling attempts but is an important characteristic for eliminating type II errors. The CM generally had the lowest number of type II errors, likely simply because this monitoring approach

1666

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 5 9 e1 6 6 8

closes the beach more often, an action that also results in significantly more type I errors (Fig. 3). Calculated excess illnesses were far less variable between models, indicating that more advanced predictive models do not necessarily provide improved public health protection. IM had a lower number of excess illnesses at some beaches (Table 2), but the difference was not nearly as great as might be expected. The results imply that IM detect some instances of extreme high concentrations at these beaches but generally miss these events. The majority of beaches showed a pattern of identical number of estimated cumulative illnesses for all model approaches with the exception of the CM: it was associated with the lowest cumulative excess illnesses specifically because this approach limits overall exposure to the beach water. The CM was therefore associated with lowest excess illnesses due to more closed beaches (both type I errors and correct closed).

4.1.

Application of monitoring standards

The number of excess illnesses associated with a given model is affected by the calculation of the illness rate, and although jurisdictions have the option to calculate single-sample maximums specific to their beach waters, most Great Lakes states use the default 235 CFU/100 ml standard developed using data from the original epidemiological studies for all beaches statewide (Nevers and Whitman, 2010). Use of a higher Chicagospecific standard increases swimming access without impacting illness rates by expanding the range of allowable water contact and therefore the number of accurate predictions (correct open). Percent of type II errors was reduced under the CM and drastically reduced using RM or IM. This was particularly noticeable at beaches with lower mean E. coli concentrations: upwards of 40% reduction in percent type II errors for North Avenue, Montrose, and Oak Street with RM and IM. This significant reduction is mirrored in the potential to significantly reduce the number of illnesses at these high visitation beaches; illness rates are greatly elevated during a single high FIB event at a high use beach (Turbow et al., 2003). The use of the Chicagospecific standard did not greatly reduce the number of type II errors for 63rd Street because of the higher overall mean E. coli concentration. Within the framework of the original monitoring standards, leeway is provided for level of beach use, beach-specific variation in bacteria concentrations, and calculation of the overall water quality (US EPA, 1986), and health protection is assumed to be equally provided over a range of calculated concentrations. Considering this broader range of confidence could increase beach use without influencing health outcome under a variety of monitoring approaches, including the predictive models presented here. These results indicate that the use of a higher standard, along with a predictive model could maximize access at many of the Chicago beaches without increasing public health risk. The monitoring standards recommend use of the 5-day geometric mean, but most managers opt for the single-sample maximum, likely due to ease of use. An examination of Chicago beach monitoring data reveals that use of the 5-day geometric mean results in fewer errors than either the 235 single-sample maximum or the 385 Chicago-specific standard presented here. However, the number of days exceeding the

specified limits (i.e., swimming advisories) increases significantly with use of the geometric mean. The 5-day geometric mean was developed based on studies at beaches influenced by point sources (US EPA, 1986), areas more likely to have persistent high E. coli concentration events; river discharge or sewage releases may create periods of sustained high E. coli concentrations, warranting the extended swimming advisory that results from a running geometric mean. Chicago’s beaches, however, are not influenced by a major point source except during rare events of sewage overflows, during which beaches are preemptively closed for several days.

4.2.

Maximizing public health protection

Predictive modeling results indicate that this monitoring approach would not improve health protection at all Chicago beaches. The best approach for monitoring may differ between beaches, even within an individual jurisdiction such as Chicago. The threshold for level of effort associated with increased model refinement would have to be considered for each beach, perhaps while incorporating economic considerations. Hou et al. (2006) determined that different monitoring policies provided the optimal economic and public health outcomes for each of two beaches. The application of different monitoring strategies may include combining approaches or extending to rapid methods, alternate indicators, or diverse management plans. Recent research to characterize the sources, survival, fate, and transport of FIB and the applications of monitoring programs has perhaps complicated the applicability of different monitoring and management strategies by indicating that one monitoring approach does not fit all beach types. Novel management approaches have included predictive models (Frick et al., 2008; Nevers and Whitman, 2005), rapid tests (Bushon et al., 2009; Lavender and Kinzelman, 2009), new indicators, including hostspecific markers (Bacteroides, Methanobrevibacter, virulence markers etc), and gene-based detection techniques for human pathogens (Griffith et al., 2009). Reconsiderations of monitoring standards have also been explored (Kim and Grant, 2004; Nevers and Whitman, 2010). While epidemiological studies have linked illness rates with outcomes from some of these new analyses (Wade et al., 2006), care will have to be taken to consider whether environmental conditions and sources influence the results outcomes. The ideal method for beach management may differ among beaches and jurisdictions, and it may be desirable to have a variety of monitoring approaches available to beach managers that increase accuracy and reliability at given beaches. In deciphering the best management plans for different beaches, efforts should focus on improving public health protection, perhaps considering a wide variety of available monitoring options.

5.

Conclusions

 Refinement of monitoring models generally increased predictability of E. coli but did not necessarily result in fewer errors or excess illnesses  Regional models provided similar levels of accuracy as individual beach models in many locations

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 5 9 e1 6 6 8

 Use of a location-specific water quality standard, combined with empirical predictive models, provided the greatest beach access without sacrificing public health protection

Acknowledgments We thank Murulee Byappanahalli (USGS) for his careful review. Research was funded in part by the US Ocean Action Plan: USGS Ocean Research Priorities Plan and by the Great Lakes Restoration Initiative. Any use of trade, product, or firm names is for descriptive purposes only and does not imply endorsement by the U.S. Government. This article is Contribution 1627 of the USGS Great Lakes Science Center.

references

Boehm, A.B., 2007. Enterococci concentrations in diverse coastal environments exhibit extreme variability. Environmental Science & Technology 41 (24), 8227e8232. Boehm, A.B., Grant, S.B., Kim, J.H., Mowbray, S.L., McGee, C.D., Clark, C.D., Foley, D.M., Wellman, D.E., 2002. Decadal and shorter period variability of surf zone water quality at Huntington Beach, California. Environmental Science & Technology 36 (18), 3885e3892. Boehm, A.B., Whitman, R.L., Nevers, M.B., Hou, D., Weisberg, S.B., 2007. Statistical Framework for Recreational Water Quality Criteria and Monitoring. In: Wymer, L.J. (Ed.). Wiley, West Sussex, UK, pp. 179e210. Buckalew, D.W., Hartman, L.J., Grimsley, G.A., Martin, A.E., Register, K.M., 2006. A long-term study comparing membrane filtration with Colilert defined substrates in detecting fecal coliforms and Escherichia coli in natural waters. Journal of Environmental Management 80 (3), 191e197. Bushon, R.N., Likirdopulos, C.A., Brady, A.M.G., 2009. Comparison of immunomagnetic separation/adenosine triphosphate rapid method to traditional culture-based method for E. coli and enterococci enumeration in wastewater. Water Research 43 (19), 4940e4946. Calderon, R.L., Mood, E.W., Dufour, A.P., 1991. Health effects of swimmers and nonpoint sources of contaminated water. International Journal of Environmental Health Research 1, 21e31. Dufour, A.P., 1984. Health Effects Criteria for Fresh Recreational Waters EPA Report EPA-600/1-84-004, Cincinnati, Ohio. Dwight, R.H., Brinks, M.V., SharavanaKumar, G., Semenza, J.C., 2007. Beach attendance and bathing rates for Southern California beaches. Ocean & Coastal Management 50 (10), 847e858. Francy, D.S., 2009. Use of predictive models and rapid methods to nowcast bacteria levels at coastal beaches. Aquatic Ecosystem Health and Management 12 (2), 177e182. Frick, W.A., Ge, Z., Zepp, R.G., 2008. Nowcasting and forecasting concentrations of biological contaminants at beaches: a feasibility and case study. Environmental Science & Technology 42, 4218e4824. Ge, Z., Nevers, M.B., Schwab, D.J., Whitman, R.L., 2010. Coastal loading and transport of Escherichia coli at an embayed beach in Lake Michigan. Environmental Science & Technology 44, 6731e6737. Given, S., Pendleton, L.H., Boehm, A.B., 2006. Regional public health cost estimates of contaminated coastal waters: a case

1667

study of gastroenteritis at southern California beaches. Environmental Science & Technology 40 (16), 4851e4858. Griffith, J.F., Cao, Y., McGee, C.D., Weisberg, S.B., 2009. Evaluation of rapid methods and novel indicators for assessing microbiological beach water quality. Water Research 43 (19), 4900e4907. Gronewold, A.D., Borsuk, M.E., Wolpert, R.L., Reckhow, K.H., 2008. An assessment of fecal indicator bacteria-based water quality standards. Environmental Science & Technology 42 (13), 4676e4682. Harwood, V.J., Levine, A.D., Scott, T.M., Chivukula, V., Lukasik, J., Farrah, S.R., Rose, J.B., 2005. Validity of the indicator organism paradigm for pathogen reduction in reclaimed water and public health protection. Applied and Environmental Microbiology 71 (6), 3163e3170. He, L.M., He, Z.L., 2008. Water quality prediction of marine recreational beaches receiving watershed baseflow and stormwater runoff in southern California, USA. Water Research 42 (10e11), 2563e2573. Hou, D., Rabinovici, S.J.M., Boehm, A.B., 2006. Enterococci predictions from partial least squares regression models in conjunction with a single-sample standard improve the efficacy of beach management advisories. Environmental Science & Technology 40 (6), 1737e1743. Kim, J.H., Grant, S.B., 2004. Public mis-notification of coastal water quality: a probabilistic evaluation of posting errors at Huntington Beach, California. Environmental Science & Technology 38 (9), 2497e2504. Lavender, J.S., Kinzelman, J.L., 2009. A cross comparison of QPCR to agar-based or defined substrate test methods for the determination of Escherichia coli and enterococci in municipal water quality monitoring programs. Water Research 43 (19), 4967e4979. Leecaster, M.K., Weisberg, S.B., 2001. Effect of sampling frequency on shoreline microbiology assessments. Marine Pollution Bulletin 42 (11), 1150e1154. Lepesteur, M., McComb, A.J., Moore, S.A., 2006. Do we all face the same risk when bathing in the estuary? Water Research 40 (14), 2787e2795. Marion, J.W., Lee, J., Lemeshow, S., Buckley, T.J., 2010. Association of illness and recreational water exposure during advisory and non-advisory conditions at an inland US beach. Water Research 44 (16), 4796e4804. McPhail, C.D., Stidson, R.T., 2009. Bathing water signage and predictive water quality models in Scotland. Aquatic Ecosystem Health and Management 12 (2), 183e186. Nevers, M.B., Whitman, R.L., 2005. Nowcast modeling of Escherichia coli concentrations at multiple urban beaches of southern Lake Michigan. Water Research 39 (20), 5250e5260. Nevers, M.B., Whitman, R.L., 2008. Coastal strategies to predict Escherichia coli concentrations for beaches along a 35 km stretch of southern Lake Michigan. Environmental Science & Technology 42 (12), 4454e4460. Nevers, M.B., Whitman, R.L., 2010. Policies and practices of beach monitoring in the Great Lakes, USA: a critical review. Journal of Environmental Monitoring 12 (3), 581e590. Olyphant, G.A., Whitman, R.L., 2004. Elements of a predictive model for determining beach closures on a real time basis: the case of 63rd Street Beach Chicago. Environmental Monitoring and Assessment 98 (1e3), 175e190. Pru¨ss, A., 1998. Review of epidemiological studies on health effects from exposure to recreational water. International Journal of Epidemiology 27 (1), 1e9. Rabinovici, S.J.M., Bernknopf, R.L., Wein, A.M., Coursey, D.L., Whitman, R.L., 2004. Economic and health risk trade-offs of swim closures at a Lake Michigan Beach. Environmental Science & Technology 38 (10), 2737e2745.

1668

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 5 9 e1 6 6 8

Seyfried, P.L., Tobin, R.S., Brown, N.E., Ness, P.F., 1985. A prospective study of swimming-related illness I. Swimmingassociated health risk. American Journal of Public Health 75 (9), 1068e1070. Sinigalliano, C.D., Fleisher, J.M., Gidley, M.L., Solo-Gabriele, H.M., Shibata, T., Plano, L.R.W., Elmir, S.M., Wanless, D., Bartkowiak, J., Boiteau, R., Withum, K., Abdelzaher, A.M., He, G., Ortega, C., Zhu, X., Wright, M.E., Kish, J., Hollenbeck, J., Scott, T., Backer, L.C., Fleming, L.E., 2010. Traditional and molecular analyses for fecal indicator bacteria in nonpoint source subtropical recreational marine waters. Water Research 44 (10), 3763e3772. Turbow, D.J., Osgood, N.D., Jiang, S.C., 2003. Evaluation of recreational health risk in coastal waters based on enterococcus densities and bathing patterns. Environmental Health Perspectives 111 (4), 598e603. US EPA, 1986. Ambient Water Quality Criteria for Bacteria 1986, p. 18. US EPA, Office of Water Regulations and Standards, Washington DC. Wade, T.J., Calderon, R.L., Sams, E., Beach, M., Brenner, K.P., Williams, A.H., Dufour, A.P., 2006. Rapidly measured indicators of recreational water quality are predictive of swimming associated gastrointestinal illness. Environmental Health Perspectives 114 (1), 24e28. Wade, T.J., Pai, N., Eisenberg, J.N.S., Colford Jr., J.M., 2003. Do U.S. Environmental Protection Agency water quality guidelines for recreational waters prevent gastrointestinal illness? A

systematic review and meta-analysis. Environmental Health Perspectives 111 (8), 1102e1109. Whitman, R.L., Nevers, M.B., 2003. Foreshore sand as a source of Escherichia coli in nearshore water of a Lake Michigan beach. Applied and Environmental Microbiology 69 (9), 5555e5562. Whitman, R.L., Nevers, M.B., 2004. Escherichia coli sampling reliability at a frequently closed Chicago beach: monitoring and management implications. Environmental Science & Technology 38 (16), 4241e4246. Whitman, R.L., Nevers, M.B., 2008. Summer E. coli patterns and responses along 23 Chicago beaches. Environmental Science & Technology 42 (24), 9217e9224. Whitman, R.L., Nevers, M.B., Gerovac, P.J., 1999. Interaction of ambient conditions and fecal coliform bacteria in southern Lake Michigan waters: monitoring program implications. Natural Areas Journal 19, 166e171. Whitman, R.L., Shively, D.A., Pawlik, H., Nevers, M.B., Byappanahalli, M.N., 2003. Occurrence of Escherichia coli and enterococci in Cladophora (Chlorophyta) in nearshore water and beach sand of Lake Michigan. Applied and Environmental Microbiology 69 (8), 4714e4719. Wong, M., Kumar, L., Jenkins, T.M., Xagoraraki, I., Phanikumar, M.S., Rose, J.B., 2009. Evaluation of public health risks at recreational beaches in Lake Michigan via detection of enteric viruses and a human-specific bacteriological marker. Water Research 43, 1137e1149.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 6 9 e1 6 8 0

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Performance and biofilm activity of nitrifying biofilters removing trihalomethanes David G. Wahman a,*, Lynn E. Katz b, Gerald E. Speitel, Jr.b a

United States Environmental Protection Agency, Office of Research and Development, 26 W. Martin Luther King Dr., Cincinnati, OH 45268, USA b University of Texas at Austin, Department of Civil, Architectural and Environmental Engineering, Austin, TX 78712, USA

article info

abstract

Article history:

Nitrifying biofilters seeded with three different mixed-culture sources removed trichloro-

Received 13 July 2010

methane (TCM) and dibromochloromethane (DBCM) with removals reaching 18% for TCM

Received in revised form

and 75% for DBCM. In addition, resuspended biofilm removed TCM, bromodichloro-

7 December 2010

methane (BDCM), DBCM, and tribromomethane (TBM) in backwash batch kinetic tests,

Accepted 8 December 2010

demonstrating that the biofilters contained organisms capable of biotransforming the four

Available online 16 December 2010

regulated trihalomethanes (THMs) commonly found in treated drinking water. Upon the initial and subsequent increased TCM addition, total ammonia nitrogen (TOTNH3) removal

Keywords:

decreased and then reestablished, indicating an adjustment by the biofilm bacteria. In

Trihalomethanes

addition, changes in DBCM removal indicated a change in activity related to DBCM. The

Cometabolism

backwash batch kinetic tests provided a useful tool to evaluate the biofilm’s bacteria. Based

Nitrification

on these experiments, the biofilters contained bacteria with similar THM removal kinetics

Disinfection by-products

to those seen in previous batch kinetic experiments. Overall, performance or selection does

Drinking water

not seem based specifically on nutrients, source water, or source cultures and most likely results from THM product toxicity, and the use of GAC media appeared to offer benefits over anthracite for biofilter stability and long-term performance, although the reasons for this advantage are not apparent based on research to date. Published by Elsevier Ltd.

1.

Introduction

During drinking water disinfection, natural organic matter (NOM) combines with the disinfectant to produce disinfection by-products (DBPs), including haloacetic acids (HAAs) and trihalomethanes (THMs). Chlorine disinfection remains quite popular in the United States (AWWA Water Quality and Technology Division Disinfection Systems Committee, 2000a,b, 2008a,b), but as a result of the Stage 1 and Stage 2 Disinfectants and Disinfection Byproduct Rules, many utilities now use combinations of chlorine and chloramines to avoid excessive THM and HAA formation. A recent survey reported

* Corresponding author. Tel.: þ1 513 569 7733; fax: þ1 513 487 2543. E-mail address: [email protected] (D.G. Wahman). 0043-1354/$ e see front matter Published by Elsevier Ltd. doi:10.1016/j.watres.2010.12.012

that 30% of the respondents currently chloraminate to maintain distribution system residual, and other recent surveys suggest that between 8 and 12% of drinking water utilities are contemplating a future switch to chloramination (AWWA Water Quality and Technology Division Disinfection Systems Committee, 2008b; Seidel et al., 2005) with chloramination for secondary disinfection in the United States predicted to increase to 57% of all surface and 7% of all ground water treatment systems (USEPA, 2005). A typical chloramine treatment scheme consists of an initial chlorination period to help achieve disinfection goals followed by quenching with ammonia at some point in the

1670

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 6 9 e1 6 8 0

treatment train to meet DBP goals through the lower regulated DBP formation rates associated with chloramines. Substantial formation of THMs and HAAs can occur within treatment plants even during relatively short chlorination periods (Singer et al., 1999b; Symons et al., 1982). Therefore, approaches for minimizing DBP formation or for removing DBPs within treatment plants are potentially of much practical value. Once formed, one possible removal mechanism for these DBPs is biological degradation. Evidence for HAA biodegradation in drinking water environments continues to mount (Baribeau et al., 2000; Bayless and Andrews, 2008; Leach et al., 2009; McRae et al., 2004; Singer et al., 1999a; Williams et al., 1997, 1998; Xie and Zhou, 2000, 2002; Zhang et al., 2009). Of course, THMs and HAAs tend to form together, so biological DBP removal processes must be able to deal with both classes of DBPs to be of any practical value in regulatory compliance. Unfortunately, THM biodegradation is more difficult than HAA biodegradation. In contrast to HAAs, which can serve as carbon and energy sources for microbial growth, THMs require cometabolic pathways that serve neither purpose. Thus, a primary substrate is needed to sustain the bacteria. Of particular interest in drinking water treatment is the observation that ammonia-oxidizing bacteria are capable of aerobically transforming the four regulated THMs (trichloromethane (TCM) or chloroform, bromodichloromethane (BDCM), dibromochloromethane (DBCM), tribromomethane (TBM) or bromoform) commonly found in treated drinking water (Wahman et al., 2006a, 2005). Implementation of biological THM removal should involve relatively minor retrofitting of existing plants. The cometabolism would occur in granular media filters consisting of an upper layer of granular activated carbon (GAC). Utilities could carry out prechlorination followed by ammonia addition at a relatively low concentration (1e4 mg N/L) sometime before the filters. A mixture of monochloramine and ammonia will result at the typical free chlorine concentrations (e.g., 2 mg Cl2/L) used in treatment plants. When the water is applied to the upper GAC layer in the filter, monochloramine will be destroyed through a catalytic reaction with the GAC, releasing ammonia (Fairey et al., 2006, 2007; Komorita and Snoeyink, 1985). At this point, an appropriate environment for microbial growth is established (i.e., an environment devoid of a disinfectant residual). With respect to THMs, nitrifiers, specifically ammonia oxidizers, can grow using the available ammonia and cometabolize THMs. The filtered water would then be post-disinfected, presumably with chloramines, before distribution. Previous biofilter experiments provided an initial demonstration of feasibility in lab-scale biofilter experiments seeded with a single nitrifying mixed culture from Lake Austin, Texas (Wahman et al., 2006b). The current research extended process evaluation to two additional mixed-culture sources studied previously in batch experiments (Wahman et al., 2006a). In addition, further exploration of the reduced performance over time seen during biofilter operation with Lake Austin feed water (Wahman et al., 2006b) was investigated. Biofilter media was expanded to include GAC. The GAC source was from an operating drinking water treatment plant filter that was preceded by monochloramine addition (i.e., the proposed process configuration), providing insight into

cultures likely seen in practice. Furthermore, batch kinetic tests were conducted on backwash from the biofilters to provide a direct evaluation of the biofilm’s ability to remove THMs.

2.

Materials and methods

2.1.

Water collection and storage

Lake Austin water was obtained from the raw water line of the drinking water treatment facilities in Austin, Texas prior to any treatment. Water was subsequently stored in a 4  C temperature controlled room in LDPE and HDPE storage tanks until use. Lake Austin water is a typical central United States surface water with an alkaline pH (8.26e8.43), moderate alkalinity (169e190 mg CaCO3/L), and dissolved organic carbon (3.4e4.6 mg C/L) (Roalson et al., 2003).

2.2.

Nitrifier mixed cultures

Three different mixed-culture sources were used for biofilter inoculation: (1) a sample collected from the influent line of a drinking water treatment facility in Laredo, Texas (Rio Grande), (2) an enriched nitrifier culture from several distribution systems in California and Wisconsin that was dominated by Nitrosomonas oligotropha representatives (provided by Dr. D. Noguera, University of Wisconsin) (N. oligotropha enrichment), and (3) in-use Filtrasorb 400 (F400) was obtained from the granular media filters at the City of Laredo drinking water treatment plant (Laredo).

2.3.

Biofilter media

Virgin anthracite of the appropriate mesh size (30  40) was obtained by grinding in a blender and sieving with the appropriately sized sieves. To remove the fines from the ground anthracite, it was first washed on the sieves with distilled-deionized (DDI) water. Further washing was achieved by placing the ground anthracite in a glass beaker with an approximate volume ratio of 9 parts Millipore water and 1 part anthracite where the mixture was stirred and allowed to settle before decanting the water. This process was repeated approximately 30 times or until the decanted water was clear. The in-use Filtrasorb 400 (F400) GAC media was hand ground with a mortar and pestle and sieved to obtain a 30  40 mesh size media. The fines were removed as per anthracite media.

2.4.

Biofilter setup

Two feed waters (Table 1) were used in the biofilter experiments: (1) 0.2 mm-filtered Lake Austin water supplemented with defined micronutrients (i.e., iron, copper, and phosphorus) or (2) 0.2 mm-filtered DDI water feed supplemented with nutrients (i.e., calcium, magnesium, copper, and iron) based on batch Nitrosomonas europaea growth (Wahman et al., 2005) and a carbonate/phosphate buffer to simulate natural waters (approximately 200 mg CaCO3/L). THMs were added via a syringe pump, and oxygen was added to the feed water when required, raising the biofilter influent dissolved oxygen (DO) to non-limiting levels (approximately 16e20 mg/L).

1671

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 6 9 e1 6 8 0

Table 1 e Summary of nominal biofilter operating conditions. Run

1

2 3

Period

I II III IV I II I II III IV

Source water

EBCT (min.)

SLR (gpm/ft2)

LA LA LA LA LA NW NW NW NW NW

4 4 4 4 2 2 2 2 2 2

0.63 0.63 0.63 0.63 1.3 1.3 1.3 1.3 1.3 1.3

Nominal chemical additions to influent TOTNH3 (mg N/L)

TCM (mg/L)

DBCM (mg/L)

Fe (mg/L)

Cu (mg/L)

P (mg P/L)

4 4 4 4 2 2 2 2 4 2

0 75 110 110 110 110 100 100 100 100

0 0 0 0 0 0 0 25 25 25

200 200 200 200 200 5.6 5.6 5.6 5.6 5.6

15 15 15 15 15 0.41 0.41 0.41 0.41 0.41

0 0 0 1.1 1.1 1.1 1.1 1.1 1.1 1.1

EBCT ¼ Empty-bed contact time. LA ¼ Lake Austin water. NW ¼ Synthetic nutrient water. SLR ¼ Surface loading rate.

Experiments were conducted with three parallel trains (A, B, and C) with each train consisting of two biofilters in series. Trains A and B were packed wet and seeded with a continuous growth mixed-culture inoculum (See Section 2.2): Rio Grande (Train A) and N. oligotropha (Train B) (Wahman et al., 2006a). Before initiating flow, the biofilters were allowed to sit for approximately 1 h to promote nitrifier attachment to the media. Train C (Laredo) was packed wet but did not receive additional seeding beyond what was present on the media from being in operation at the water treatment plant and was placed into operation immediately upon packing. All trains received the same nominal influent total ammonia nitrogen (TOTNH3) concentration. TOTNH3 represents the sum of ammoniaenitrogen (NH3eN) and ammoniumenitrogen (NHþ 4 eN). Sampling points were at the first biofilter’s influent (sample point 0), between the two biofilters in a train corresponding to the first biofilter’s effluent and second biofilter’s influent (sample point 1), and at the second biofilter’s effluent (sample point 2). Using the biofilm scaling procedure proposed by Manem and Rittmann (1990), the experimental biofilters simulated full-scale filters operating with 4e8 min empty-bed contact times (EBCTs) and 2.5e7.0 gpm/ft2 (147e205 m/d) surface loading rates (SLRs), depending on the actual operating conditions and whether biofilm shear loss or external mass transport is chosen for scaling. These operating conditions fall into typical values reported for rapid filtration (MWH et al., 2005). A summary of additional biofilter operating conditions is provided in Table 1.

2.5.

Product toxicity

Previously, the cometabolism stability index (Csi) (Equation (1)) was derived to quantify the expected product toxicity of THMs fed during biofilter experiments (Wahman et al., 2006b): ! KsNH3 N þ a1 STOTNH3 Yk  k TOTNH3 d a1 STOTNH3 r0g (1) Csi ¼ ¼ P k1THM STHM ri TcTHM

Where rg0 is the net rate of bacterial cell growth (1/day); ri is the rate of THM bacterial inactivation (1/day); Y is the bacterial cell yield (mg total suspended solids (TSS)/mg TOTNH3); kTOTNH3 is the TOTNH3 maximum substrate utilization rate constant (mg TOTNH3/mg TSS-day); kd is the endogenous decay coefficient (1/day); KsNH3 N is the ammoniaenitrogen half-saturation coefficient (mg N/L NH3eN); a1 is the ratio of NH3eN/TOTNH3; STOTNH3 is the TOTNH3 concentration (mg N/L TOTNH3); k1THM is the THM pseudo-first-order rate constant (L/mg TSS-day); STHM is the THM concentration (mg/L THM); and TcTHM is the THM transformation capacity (mg THM/mg TSS). For bacteria to provide sustained biotransformation of THMs, the net growth rate on ammonia (based on Monod kinetics) must be greater than the inactivation rate from THM biotransformation. Equation (1) indicates that for stable biofilter operation Csi must be greater than or equal to one (i.e., the net growth rate, rg0 , must be greater than the sum of the THM inactivation rates, ri). As was done for previous biofilter experiments (Wahman et al., 2006b), Csi values for all operating periods were initially calculated using kinetic parameters determined previously for N. europaea.

2.6.

Backwash batch kinetic experiments

During certain biofilter operating periods, biofilter backwash water was collected and organisms were subsequently harvested by centrifugation, washed, centrifuged again, and resuspended in fresh buffer medium (8 mM phosphate and 10 mM carbonate, pH 8) for batch kinetic studies. The fresh buffer media was aerated with pure oxygen to increase the dissolved oxygen concentration to levels (greater than 20 mg/L) that would not be fully consumed by ammonia removal and would not adversely affect the organisms during the experiment (Wahman et al., 2005). The approach of Aziz et al. (1999) was used to conduct the experiment. Briefly, batch kinetic assays were carried out head-space-free in 500 mL, glass, gas-tight syringes. Each syringe was wrapped in aluminum foil and contained a Teflon-coated stir bar for mixing. Chemicals were injected through the syringe nose to

1672

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 6 9 e1 6 8 0

start an experiment. Samples were collected over time by depressing the syringe plunger and ejecting the samples into a smaller gas-tight syringe. Thus, head-space-free conditions were maintained throughout the experiment (150e300 min), thereby virtually eliminating volatilization loss of chemicals. Starting individual THM concentrations ranged from 80 to 100 mg/L each and TOTNH3 concentrations from 4 to 8 mg N/L. From these experiments, ammonia and THM kinetic parameters were determined.

2.7.

Determination of batch kinetic parameters

Kinetic parameters were determined as described previously (Wahman et al., 2006a, 2005). For ammonia kinetics, Monod kinetic coefficients were estimated by nonlinear regression using Excel’s Solver routine (Smith et al., 1997, 1998). The model (Equation (2)) accounts for TOTNH3 being measured experimentally, while recognizing that NH3 is the actual substrate (Suzuki et al., 1974): dSTOTNH3 kTOTNH3 XSTOTNH3 a1 ¼ dt KsNH3 N þ STOTNH3 a1

(2)

where X is the biomass concentration (mg/L TSS) and other parameters were described previously. A fourth-order RungeeKutta numerical approximation of the Monod equation was fitted to data by minimizing the normalized residual sum of squares between predicted and experimental values. Normalization was achieved by dividing the residual sum of squares by the experimental value squared, resulting in a dimensionless sum of squares error (Robinson, 1985). For THMs, a reductant model (Equation (3)) was used that accounted for two limiting reactants, THMs and ammoniaenitrogen (NH3eN). The reductant model (Arcangeli and Arvin, 1997) was selected because it was superior to several other models in kinetic studies with N. europaea and several mixed-culture sources (Wahman et al., 2006a, 2005): dSTHM k1THM STHM ¼ KsNH3 N dt 1þ STOTNH3 a1

(3)

Where all parameters were described previously. The ammonia kinetic parameters were incorporated into the THM model without adjustment, and the THM parameters were determined using the same fitting method as used for the ammonia kinetics. As a result, the only adjustable parameters for the THM kinetic model were the THM rate constant and each initial THM concentration. The nonlinear regression analysis yielded estimates of each THM rate constant ðk1THM Þ, the ammonia maximum specific rate of biotransformation ðkTOTNH3 Þ, and the ammonia half-saturation constant ðKsNH3 N Þ, as well as the initial concentrations (S0) for ammonia and each THM. Further statistical analyses permitted estimates of the approximate 95% joint confidence limit (CL) for each parameter (Smith et al., 1997, 1998; Wahman et al., 2005).

2.8.

Simplified THM cometabolism model

A simplified THM cometabolism biofilter model can be obtained by ignoring mass transport resistances and assuming an

ideal plug flow reactor with a residence time equal to the contact time in the biofilter. Using the rate equations for TOTNH3 and THM detailed previously, a closed form solution for the removal of a given THM can be obtained and is shown below as Equation (4): k

1 DTOTNH3 k THM STHMn TOTNH3 ¼e STHM0

(4)

Where STHM0 is the THM influent concentration (mg/L THM); STHMn is the THM effluent concentration from nth (1 or 2) biofilter in series (mg/L THM); DTOTNH3 is the influent TOTNH3 minus effluent TOTNH3; and other parameters were described previously. Based on Equation (4), the THM normalized effluent S n Þ, and therefore THM fractional removal, concentration ðSTHM THM0 will be (1) independent of the influent THM and TOTNH3 concentrations and (2) for a given DTOTNH3 removal in a biofilter, dependent on the THM rate constant ðk1THM Þ. Thus, Equation (4) can be used to approximate the removal of each THM species as a function of ammonia removal.

2.9.

Biofilter operational data statistical analyses

Tukey’s paired comparison method was used to compare performance of different operating periods of the biofilters (Berthouex and Brown, 2002). A two-sided 95% confidence interval of the Studentized Range Statistic was used for all paired comparisons (Harter, 1960).

2.10.

Analytical methods

THM concentrations were measured using USEPA Method 551.1 with modifications. Concentrations of individual THM species were analyzed on a Hewlett Packard 5890A gas chromatograph with liquid autosampler and J&W DB-5 column

Table 2 e Selected biofilter performance summary (mean ± standard deviation). Train Run Period No. of D0e1TOTNH3 TCM % DBCM % samples (mg N/L) Removal Removal A

B

C

1 2 3 3 1 2 3 3a 1 2 3 3

IV I II IV IV I II IV IV I II IV

4 6 5 2 4 6 7 2 4 6 8 1

4.2  2.1  2.4  2.7  4.0  2.0  2.2  2.0  3.9  2.0  2.2  3.6

0.13 0.064 0.054 0.069 0.066 0.053 0.10 0.036 0.069 0.10 0.088

18  11  4.9  14  17  8.3  9.5  8.3  2.7  4.5  1.1  8.7

5.2 5.5 3.8 10 6.5 8.5 6.7 1.1 1.1 6.0 2.2

60  1.3 17  3.1

14  7.6 14  0.52

46  5.1 75

D0e1TOTNH3 ¼ TOTNH3 removed through the first biofilter in series (mg N/L). a Complete TOTNH3 removal not occurring.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 6 9 e1 6 8 0

1673

Fig. 1 e Run 1 biofilter (A [ Rio Grande, B [ N. oligotropha enrichment, and C [ Laredo) TOTNH3 concentrations (1) and percent removals (2) for first biofilter in series. Biofilters fed Lake Austin water with nominal influent TCM additions of 0 mg/L (Period I), 75 mg/L (Period II), or 110 mg/L (Periods III and IV).

with constant pressure and splitless injection. The initial oven temperature was 32  C for 3.5 min. Then, the temperature was ramped at 20  C/min to 72  C and remained at 72  C for 3.5 min for a total analysis time of 9 min. A standard curve was developed to span the range of anticipated THM concentrations for each experiment. For quality control purposes, a blank and duplicate sample was placed every 20 vials and a standards calibration check was run every 10th vial. An ion selective electrode probe, Thermo Orion 9512, connected to an Orion Model 920A pH/ISE electrode meter was used to measure ammonia. An ammonia standard curve was developed to span the anticipated range of ammonia concentrations the experiments. All ammonia concentrations reported are in mg N/L. DO was measured with a YSI 5905 oxygen probe on a YSI Model 54ARC oxygen meter calibrated per the manufacturer’s recommendations. pH was measured using an Orion ROSSä combination pH electrode on an Orion Model 920A pH/ISE meter calibrated with pH standards of 4, 7, and 10. Total suspended solids (TSS) and volatile suspended solids (VSS) were measured to determine the biomass for batch kinetic experiments using Standard Methods 2540D and 2540E, respectively (APHA et al., 1998). The solids were measured with the buffer solution/biomass mixture remaining after batch kinetic experiments were completed, and the

volume of solution usually ranged from 50 to 100 mL. This volume was vacuum filtered through a Whatman cellulose nitrate 0.2-mm filter. Because VSS was equal to TSS measurements (data not shown), only TSS is reported herein.

3.

Results and discussion

Three biofilter trains were seeded with different mixedculture sources: (1) Rio Grande on anthracite (Train A), (2) N. oligotropha enrichment on anthracite (Train B), and (3) Laredo GAC (Train C). This experimental design provided biofilter performance data for two mixed cultures (N. oligotropha enrichment and Rio Grande) studied previously in batch (Wahman et al., 2006a), and the inclusion of a biofilter with GAC from the City of Laredo drinking water treatment plant provided insight into cultures likely seen in practice (i.e., a plant using monochloramine for disinfection preceding a GAC filter). For ease of presentation, the results are separated into three runs. To compare the performance of Runs 1e3, average values were calculated for pseudo-steady-state operation based on TOTNH3 removal (i.e., complete TOTNH3 removal through the first column in series). Table 2 summarizes selected results for each train along with their associated standard deviations.

1674

3.1.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 6 9 e1 6 8 0

Run 1

Run 1 consisted of an initial operating period to establish nitrification (Period I), explored TCM product toxicity and recovery (Periods II and III), and attempted to address issues of low TCM removal by phosphorus addition (Period IV). Lake Austin water was supplemented with iron and copper (Table 1) as these additions were previously shown to improve biofilter performance on Lake Austin water (Wahman et al., 2006b). Fig. 1 details the TOTNH3 concentrations and associated percent removals for the first biofilter in series of all three trains in Run 1. During Run 1, an initial period (Period I) was provided for the biofilters to reach a pseudo-steady-state removal of TOTNH3 without the presence of THMs. Trains A and B stabilized with an approximate 70e80% TOTNH3 removal through the first biofilter in series. During this same timeframe, Train C showed complete TOTNH3 removal through the first biofilter in series, indicating the presence of a greater and/or more active biomass as compared with Trains A and B. To simulate a Csi value close to one (1.1), an initial addition of 75 mg/L TCM (Period II) was started. This addition resulted in an immediate large decrease in TOTNH3 removal for both Trains A (13%) and B (0%). After this sharp initial decrease in TOTNH3 removal and by the end of Period II, Train A’s TOTNH3 removal improved to a level similar to Period I (w80%), but Train B’s removal only slightly improved to approximately 20%. In contrast, Train C showed a minimal effluent TOTNH3 concentration (0.5 mg N/L) after the initial TCM addition, with subsequent samples showing complete TOTNH3 removal through the first column in series. Train C contained GAC media; therefore, adsorption of TCM may have protected a portion of the biomass from the toxic effects of TCM cometabolism. Because the trains showed improved TOTNH3 removal after their initial decrease during Period II, the influent

Fig. 2 e Train A (Rio Grande) and Train B (N. oligotropha enrichment) recovery from initial TCM addition during Run 1 (Periods II and III) for first biofilter in series.

TCM concentration was increased to 110 mg/L (Period III) to decrease the Csi value (0.76) and evaluate whether recovery would continue. After showing an initial decrease in TOTNH3 removal when the TCM concentration increased, all three trains moved toward complete TOTNH3 removal during Period III with only Train B having a minimal TOTNH3 effluent concentration (0.25 mg N/L) at the end of Period III. A similar effect on TOTNH3 removal was seen upon initial THM addition in previous biofilter studies, but in these studies, the influent THMs were removed to allow recovery (Wahman et al., 2006b). The current experiment shows that the biofilters do not require the removal of TCM to recover from the initial adverse effects of TCM addition. Even with eventual nearly complete TOTNH3 removal in all three trains, TCM removal was variable and less than in previous biofilter experiments with comparable TOTNH3 removals (Wahman et al., 2006b). TCM removal for Trains A and B ranged from 10 to 21% and 3 to 19%, respectively. Train C’s TCM removal declined from initial removals of 8e10%

Fig. 3 e Train A (Rio Grande) Run 2 biofilter TOTNH3 (A) and TCM (B) concentrations and percent removals (C) for first biofilter in series. Biofilters fed Lake Austin (Period I) or synthetic nutrient (Period II) water with nominal influent TCM addition of 110 mg/L (Periods IeII).

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 6 9 e1 6 8 0

to final removals of 0e5% at the end of Period III. Train C’s initial TCM removal was most likely a result of GAC TCM adsorption. Train A and B’s recovery rates from TCM addition (Periods II and III) were different. Compared with Train A, Train B took approximately 800 h longer to reach complete TOTNH3 removal. For Trains A and B, the recovery pattern was approximately linear with TOTNH3 mass removal through the first biofilter in series (Fig. 2). Train A’s TOTNH3 removal recovery rate (95% CL 0.012  0.0040 mg N/L-h) was significantly greater than Train B’s (95% CL of 0.0043  0.00046 mg N/L-h), indicating a different response to TCM addition. Based on the Csi concept, Train B likely possessed a biomass with either a lower yield, kTOTNH3 , or transformation capacity or a higher k1THM or ksNH3 N . For both Train A and B, the biofilm nitrifying community made an obvious adjustment to respond to the addition of TCM to the biofilter and was able to recover under a continuous TCM feed. Because of the decreased TCM removal seen during Periods II and III, phosphorus was added at 1.1 mg P/L to evaluate its effect on TCM removal (Period IV). Phosphorus was not previously studied and may be limiting in the Lake Austin

1675

water. Lake Austin water’s average total phosphate concentration ranged from 0.01 to 0.05 mg P/L during 2004 and 2005 (City of Austin Water Utility n.d.). During Period IV, TOTNH3 removal remained complete for each train. For Trains A and B, TCM removal was variable but similar to the periods without phosphorus addition, ranging from 10 to 25%. For Train C, the TCM removal remained minimal at 2e4%, indicating little effect from the phosphorus addition on any of the trains.

3.2.

Run 2

For Run 2 (Train A, Fig. 3), the influent TOTNH3 concentration was decreased from 4 to 2 mg N/L to evaluate whether enzyme competition was occurring between TOTNH3 and the THMs. In addition, the EBCT was decreased from 4 to 2 min, with the goal of achieving a measurable steady-state effluent TOTNH3 concentration from the first biofilter in series. In an attempt to improve TCM removal during Run 2, the feed water was changed from Lake Austin (Period I) to synthetic nutrient water (Period II) as better performance with respect to TCM removal was seen with a nutrient water feed (Wahman et al., 2006b).

Fig. 4 e Run 3 biofilter (A [ Rio Grande, B [ N. oligotropha enrichment, and C [ Laredo) TOTNH3 (1) and DBCM (2) concentrations and percent removals (3) for first biofilter in series. Biofilters fed synthetic nutrient water (Periods IeIV) with a nominal influent TCM addition of 100 mg/L (Periods IeIV) and DBCM addition of 25 mg/L (Periods IIeIV).

1676

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 6 9 e1 6 8 0

During Period I, TOTNH3 removal remained complete through the first biofilter in series for all three trains. Trains A (Fig. 3) and B (data not shown) showed similar TCM removals, but as compared with Run 1 (Period IV), these average TCM removals (Table 2) decreased from 18 to 11% and 17 to 8% for Trains A and B, respectively. For Train C, TCM removal remained minimal, averaging 5%. The switch to nutrient water (Period II) resulted in little change in TOTNH3 or TCM removal for any train. Overall, Run 2 provided evidence that the low TCM removals were not a result of source water characteristics or competition with TOTNH3.

3.3.

Run 3

Previous research has shown that compared to TCM, the bromine-substituted THMs are removed faster but result in greater product toxicity to the bacteria (Wahman et al., 2006a, 2005). To evaluate the removal of a bromine-substituted THM, DBCM was added to the influent at 25 mg/L, and the TCM influent was reduced to 100 mg/L. These THM feed concentrations decreased the Csi from Run 2 (0.75e0.48), allowing evaluation of whether stable biofilter operation would continue at this decreased Csi and whether the biofilm community would adjust to a bromine-substituted THM as with TCM during Run 1. Fig. 4 details the TOTNH3 and DBCM concentrations and associated percent removals for each train during Run 3. After an initial period (Period I) of TCM and TOTNH3 addition, DBCM was added to the influent (Period II). Even with the decreased Csi, DBCM addition did not lead to a decreased TOTNH3 removal in any train as complete TOTNH3 removal occurred after the first biofilter in series (Fig. 4). During Period II, Train A showed a substantial 44e61% DBCM removal and a low 0e11% TCM removal. Compared with Train A, Train B’s DBCM removal was lower (1e22%) and TCM removal was similar (0e18%). Train C showed no TCM removal upon the DBCM addition with TCM effluent concentrations increasing through the first biofilter in series, indicating that competitive adsorption was occurring between TCM and DBCM. During this time, Train C’s DBCM removal was 39e56%. To evaluate if a higher TOTNH3 concentration would stimulate increased THM removal, the influent TOTNH3 was increased from 2 to 4 mg N/L TOTNH3 (Period III). Train A’s TOTNH3 removal remained complete with Train C moving toward complete removal. Train B approached process failure as TOTNH3 removal decreased during this period, which is predicted from the Csi (0.49). This provided further evidence that Train B’s slower recovery during Run 1 (Periods II and III) resulted from differences in the bacterial kinetics leading to a lower Csi than Train A.

Table 3 e Biofilter average k1THM/kTOTNH3 ratio summary. k1THM kTOTNH3

Train Fig. 5 e Train C (Laredo) PSDM simulations showing (A) TCM and DBCM breakthrough curves assuming virgin GAC and TCM (B) and DBCM (C) adjusted breakthrough curves and experimental data assuming GAC prior loading.

A B C

(L/mg TOTNH3)

TCM

DBCM initial

DBCM final

0.045  0.016 0.045  0.0019 0.017  0.0085

0.38 0.067 0.28

0.071 0.074 0.38

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 6 9 e1 6 8 0

Because of Train B’s performance, the influent was reduced to 2 mg N/L TOTNH3 for the remainder of Run 3 (Period IV) for all biofilters. In addition, the trains were operated to generate biomass during Period IV for backwash batch kinetic tests to determine the kinetic parameters of the biofilter biomass. During this time, samples were taken only before backwashing the biofilter for the backwash batch kinetic tests. All trains showed similar TCM removals (approximately 10%) with various DBCM removals (Table 2). During Period IV, Train A’s DBCM removal declined to 17%, which was similar

Fig. 6 e Backwash batch kinetic test parameter estimation and 95% joint CL summary for (A) kTOTNH3 , (B) KsNH3 LN , and (C) k1THM (A1 and A1_3.5 at 4986 h, A2 at 5658 h, and B at 5708 h).

1677

to Train B’s 14% DBCM removal. In contrast, Train C’s DBCM removal improved to 75% at the end of its operation.

3.4.

Run comparisons

TCM and DBCM cometabolism was accomplished for Train A in a biofilter seeded with a mixed culture from the Rio Grande fed both Lake Austin and synthetic nutrient water. For this train, an increase in TOTNH3 removal did not lead to a significant increase in TCM or DBCM removal. This result is in contrast to previous biofilter experiments in which an increase in TOTNH3 removal coincided with an increase in THM removal (Wahman et al., 2006b). Excluding the period immediately after DBCM addition (Run 3, Period II), TCM removal remained relatively consistent over time (13% average), and no statistically significant difference existed in TCM removal for any operating period for Train A. Average DBCM removal significantly declined over time from an initial 60% removal to a final 17% removal, which was similar to the TCM removal for this train (Table 2). Train B demonstrated that TCM and DBCM cometabolism was accomplished in a biofilter seeded with an N. oligotropha enrichment culture fed Lake Austin and synthetic nutrient water. As with Train A, TCM removal was not significantly different for any of the operating periods and did not significantly decrease upon the addition of DBCM. In addition, DBCM removal started and remained at 14% which was similar to the TCM removal seen during this same period (9%) and not significantly different from Train A’s final DBCM removal. Because Train C was packed with GAC, adsorption may have occurred in the biofilter in addition to or in lieu of biological THM removal. To provide a baseline on THM adsorption, simulated breakthrough curves for TCM and DBCM were generated from the pore surface diffusion model (PSDM) implemented into AdDesignS (Hokanson et al., 1998). A simulation was run at the operating conditions seen in this research for the first biofilter in series of Train C with TCM and DBCM additions. Adsorption isotherm parameters present in the software library were used to generate the breakthrough curves shown in Fig. 5A. TCM was predicted to completely breakthrough the biofilter at 865 h (42% through Run 1). By contrast, DBCM was predicted to show no breakthrough during the operating time of the biofilters (5700 h). Because the GAC from the City of Laredo would be at some point of exhaustion at the time the samples were collected from the top of the full-scale filters, the predicted breakthrough curves for TCM and DBCM might be shifted in time based on the extent of exhaustion. To address this, the TCM and DBCM breakthrough curves were shifted in time so that the first data point for TCM and DBCM removal matched that respective point on their breakthrough curve. These adjusted breakthrough curves are shown for TCM (Fig. 5B) and DBCM (Fig. 5C) overlaid with the experimental normalized effluent data for each THM. If GAC adsorption was only occurring in the filter, the normalized effluent concentrations would be expected to increase as shown by the predicted breakthrough curves. For the DBCM normalized effluent concentration (Fig. 5C), a short period of increasing effluent values was followed by a decline

1678

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 6 9 e1 6 8 0

Table 4 e Parameters for cometabolism stability index calculations. Variable

a1 Y KsNH3 N kTOTNH3 kd k1TCM k1DBCM TcTCM TcDBCM

Initial N. europaeaa

Definition

[NH3-N]/[TOTNH3] at pH 8.0 Bacterial cell yield Ammoniaenitrogen half-saturation coefficient TOTNH3 maximum substrate utilization rate constant Endogenous decay coefficient TCM pseudo-first-order rate constant DBCM pseudo-first-order rate constant TCM transformation capacity DBCM transformation capacity

0.049 0.33c 0.16 2.9 0.020 0.10 0.20 9.2 6.5

Revisedb Train A

Train B

0.049 0.33c 0.029 1.5 0.015d 0.28 0.28 66d 22d

0.049 0.33c 0.21 0.83 0.015d 0.064 0.078 66d 22d

Units

Dimensionless mg TSS/mg TOTNH3 mg N/L NH3-N mg TOTNH3/mg TSS-day 1/day L/mg TSS-day L/mg TSS-day mg TCM/mg TSS mg DBCM/mg TSS

DBCM e dibromochloromethane, NH3eN e ammoniaenitrogen, TCM e trichloromethane, TOTNH3 e total ammoniaenitrogen, TSS e total suspended solids. a (Wahman et al., 2006b) unless otherwise noted. b This work unless otherwise noted. c (Rittmann and McCarty, 2001) e value also assumes volatile suspended solids equals TSS in cells. d (Bayer, 2007).

that is in opposition to the proposed breakthrough curve, suggesting that the removal of DBCM was biological in nature and improving over time as TOTNH3 removal improved. Furthermore, the lack of DBCM removal in the second biofilter in series (data not shown) in which no TOTNH3 removal was occurring suggests that the removal of DBCM in the first biofilter in series was biological in nature. For TCM, the data were unclear, but because of the predicted short time to breakthrough for TCM on virgin GAC (approximately 900 h), the GAC should have reached exhaustion early in the biofilter run, suggesting removal after 900 h (Run 1, Period II) can be attributed to biotransformation and not adsorption. Ignoring the period in which adsorption may have occurred (i.e., the first 900 h), TCM removal did not exceed 9%. DBCM removal increased over time, reaching its maximum level (75%) at the end of the run, providing evidence of substantially better DBCM removal with GAC versus anthracite. Using the average performance data and Equation (4), k1THM =kTOTNH3 ratios were determined for all three trains (Table 3). Because of the significant changes in DBCM removal seen for Trains A and C during their operation, Table 3 provides two ratios for DBCM, corresponding to their initial and final removals. For this same reason, standard deviations are only provided for the TCM ratios. The kinetic rate constant ratios for Trains A and B were the same for TCM and approached each other for DBCM through time. The final DBCM ratio for Train C approached that of Train A’s initial value.

Table 5 e Cometabolism stability index (Csi) values for biofilter runs. Parameter basis (Table 4) Initial (N. europaea) Revised (Train A) Revised (Train B)

Calculated Csi for run (Period) 1 (II) 1 (IIIeIV) 2 (IeII) 3 (I) 3 (II, IV) 3 (III) 1.1 1.5 3.3

0.76 1.0 2.3

0.75 1.0 2.1

0.82 1.1 2.3

0.48 0.65 1.2

0.49 0.65 1.3

3.5.

Backwash batch kinetic tests

To evaluate the ability of the bacteria present in the biofilm to biotransform THMs, batch kinetic tests on the biofilter backwash from the first biofilter in series were performed (Run 3, Period IV). For Train C, no kinetic parameters could be determined because of interfering effects with residual GAC carried over from biofilter backwashing and the competitive adsorption seen with all four THMs present. Fig. 6(A, B, and C) details the ammonia and THM kinetic parameters determined from these experiments along with their 95% joint CLs. Two experiments were conducted on Train A at 4986 (A1, TSS ¼ 44 mg/L) and 5658 (A2, TSS ¼ 33 mg/ L) hours and one on Train B at 5708 h (B, TSS ¼ 65 mg/L). Because of the large THM mass removed during the experiments with Train A, the A1 batch kinetic test was analyzed in two ways to see if transformation capacity affected the results. A complete analysis was conducted of the data set (A1) as well as a subset of the data starting when the TOTNH3 concentration was 3.5 mg N/L (A1_3.5). No significant difference was seen between these two analyses, indicating that transformation capacity was not an issue. In addition, the results from the two different experiments on Train A (A1 and A2) showed similar kinetic parameters and were conducted approximately 700 h apart, indicating the stability of the biofilm present in the biofilter and reproducibility of the results. For the backwash kinetic experiment conducted, the ammonia kinetic parameters for Trains A and B appear different although this cannot be statistically justified, but the THM kinetic parameters were markedly different. The THM kinetic parameters for Train A were significantly larger than for Train B. For both Trains A and B, the only significant difference among THMs was for A1 and A2 where the TBM rate constant was significantly less than for the other three THMs. This result differs from previous batch kinetic tests where the TBM rate constant was significantly greater than those of the other three THMs (Wahman et al., 2006a, 2005). For Trains A and B, the similar THM kinetic parameters for TCM and DBCM coincide with the similar TCM and DBCM biofilter removals at the time of the backwash tests.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 6 9 e1 6 8 0

The calculated Csi values used initially to select biofilter operating conditions were based on previous research with the pure culture N. europaea (Wahman et al., 2005). To investigate the initial decrease and subsequent recovery of the biofilters upon initial THM addition and stable operation with Csi values less than one, the determined ammonia and THM kinetic parameters and recently determined transformation capacities for a mixed culture (Bayer, 2007) were used to recalculate Csi values. The initial and revised parameters used to calculate Csi are provided in Table 4 with the resulting calculated Csi values for each operating condition summarized in Table 5. Based on this reanalysis, the recovery of the biofilters from initial TCM exposure is predicted because the Csi values for Runs 1 and 2 were above 1.0 (Table 5) for Trains A and B. For Run 3, Train B’s Csi values remain slightly above one, but for Train A, only Phase 1 is above 1.0. These lower values correspond with the instability seen in removals during Run 3.

4.

Conclusions

Nitrifying biofilters seeded with three different mixed-culture sources removed TCM (up to 18%) and DBCM (up to 75%). In addition, biofilm material backwashed from the biofilters biotransformed TCM, BDCM, DBCM, TBM in backwash batch kinetic tests, demonstrating that the biofilters contained organisms capable of biotransforming the four regulated THMs commonly found in treated drinking water. Upon the initial and subsequently increased TCM addition to the biofilters, TOTNH3 removal decreased and then reestablished itself without the need to remove TCM from the influent, indicating an adjustment by the biofilm bacteria after presumably experiencing some initial TCM product toxicity. These results indicate that sustained removals (approximately 10e15% for 2 mg N/L TOTNH3 removal and depending on THM speciation) should be attainable in systems with persistent THM concentrations (i.e., drinking water treatment plants). In addition, temporal changes in DBCM removal (Trains A and C) indicated a change in activity related to DBCM. Interestingly, the results from Trains A and C trended in opposite directions with Train A decreasing in DBCM removal while Train C increased in DBCM removal, indicating a possible benefit of using GAC. Furthermore, Train C’s GAC media originated from a drinking water treatment plant filter, implying that the organisms present from a drinking water treatment plant using chloramination were capable of substantial DBCM removal (up to 75%) and modest TCM removal (up to 9%). These removals could benefit utilities requiring modest removals (e.g., 10e15%) to maintain compliance with existing and future THM regulations. For treatment plants requiring ammonia removals above approximately 2 mg N/L (depending on influent dissolved oxygen levels) to achieve THM removal targets, supplemental oxygen would be required. The backwash batch kinetic tests provided a useful tool to evaluate the biofilm bacteria and provided kinetic parameters to evaluate product toxicity (i.e., Csi). Based on revised Csi calculations and biofilter performance, future backwash kinetic experiments should include determination of relevant

1679

transformation capacities to evaluate fully the biofilter operation. For GAC filters, batch experiments should minimize GAC carryover when harvesting bacteria so that adsorption to residual GAC can be eliminated. Based on the batch kinetic experiments, the biofilters contained bacteria with similar THM kinetics and greater transformation capacities than previous batch grown cultures. Overall, biofilter performance was not based specifically on nutrients, source water, or source cultures, and changes in performance most likely resulted from THM product toxicity. Use of GAC as media appeared to offer benefits over anthracite for biofilter stability and long-term performance, although the reasons for this advantage are not apparent based on research to date. Further investigations should explore the increased removals associated with using GAC filter media and further characterization of biofilm transformation capacities from these same systems.

Acknowledgements This research was funded by the American Water Works Association Research Foundation (AwwaRF) and Texas Advanced Technology Research Program (ATP), which the authors thank for their financial, technical, and administrative assistance. The comments and views detailed herein may not necessarily reflect the views of AwwaRF, its officers, directors, affiliates, or agents. Any opinions expressed are those of the authors and do not necessarily reflect the views of the U.S. Environmental Protection Agency; therefore, no official endorsement should be inferred. Any mention of trade names or commercial products does not constitute endorsement or recommendation for use.

references

APHA, AWWA, WEF, 1998. Standard Methods for the Examination of Water and Wastewater. American Public Health Association: American Water Works Association: Water Environment Federation, Washington, D.C. Arcangeli, J.P., Arvin, E., 1997. Modeling of the cometabolic biodegradation of trichloroethylene by toluene-oxidizing bacteria in a biofilm system. Environmental Science & Technology 31 (11), 3044e3052. AWWA Water Quality and Technology Division Disinfection Systems Committee, 2000a. Committee report: disinfection at large and medium-size systems. Journal of American Water Works Association 92 (5), 32e43. AWWA Water Quality and Technology Division Disinfection Systems Committee, 2000b. Committee report: disinfection at small systems. Journal of American Water Works Association 92 (5), 24e31. AWWA Water Quality and Technology Division Disinfection Systems Committee, 2008a. Committee report: disinfection survey, part 1-recent changes, current practices, and water quality. Journal of American Water Works Association 100 (10), 76e90. AWWA Water Quality and Technology Division Disinfection Systems Committee, 2008b. Committee report: disinfection survey, part 2-alternatives, experiences, and future plans. Journal of American Water Works Association 100 (11), 110e124.

1680

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 6 9 e1 6 8 0

Aziz, C.E., Georgiou, G., Speitel, G.E., 1999. Cometabolism of chlorinated solvents and binary chlorinated solvent mixtures using M. trichosporium OB3b PP358. Biotechnology and Bioengineering 65 (1), 100e107. Baribeau, H., Krasner, S.W., Chinn, R., Singer, P.C., 2000. Impact of Biomass on the Stability of Haloacetic Acids and Trihalomethanes in a Simulated Distribution System. American Water Works Association. Bayer, B.M., 2007. Significance of trihalomethanes in preventing distribution system nitrification. Master’s thesis, The University of Texas at Austin, Austin. Bayless, W., Andrews, R.C., 2008. Biodegradation of six haloacetic acids in drinking water. Journal of Water and Health 6 (1), 15e22. Berthouex, P.M., Brown, L.C., 2002. Statistics for Environmental Engineers. Lewis Publishers, Boca Raton. City of Austin Water Utility (n.d.). http://www.ci.austin.tx.us/water/waterreports. htm (accessed 19.10.06). Fairey, J.L., Speitel, G.E., Katz, L.E., 2006. Impact of natural organic matter on monochloramine reduction by granular activated carbon: the role of porosity and electrostatic surface properties. Environmental Science & Technology 40 (13), 4268e4273. Fairey, J.L., Speitel, G.E., Katz, L.E., 2007. Monochloramine destruction by GAC: effect of activated carbon type and source water characteristics. Journal of American Water Works Association 99 (7), 110e120. Harter, H.L., 1960. Tables of range and studentized range. The Annals of Mathematical Statistics 31 (4), 1122e1147. Hokanson, D.R., Hand, D.W., Crittenden, J.C., Oman, E.J., 1998. Adsorption Design Software (AdDesignS). Center for Clean Industrial and Treatment Technologies (CenCITT) Michigan Technological University, Houghton, Michigan. Komorita, J.D., Snoeyink, V.L., 1985. Monochloramine removal from water by activated carbon. Journal of American Water Works Association 77 (1), 62e64. Leach, L.H., Zhang, P., LaPara, T.M., Hozalski, R.M., Camper, A.K., 2009. Detection and enumeration of haloacetic acid-degrading bacteria in drinking water distribution systems using dehalogenase genes. Journal of Applied Microbiology 107 (3), 978e988. Manem, J.A., Rittmann, B.E., 1990. Scaling procedure for biofilm processes. Water Science and Technology 22 (1/2), 329e346. McRae, B.M., LaPara, T.M., Hozalski, R.M., 2004. Biodegradation of haloacetic acids by bacterial enrichment cultures. Chemosphere 55 (6), 915e925. MWH, Crittenden, J.C., Trussell, R.R., Hand, D.W., Howe, K.J., Tchobanoglous, G., 2005. Water Treatment: Principles and Design. John Wiley & Sons, Inc., Hoboken, New Jersey. Rittmann, B.E., McCarty, P.L., 2001. Environmental Biotechnology: Principles and Applications. McGraw-Hill Companies, Inc., New York. Roalson, S.R., Kweon, J., Lawler, D.F., Speitel, G.E.J., 2003. Enhanced softening: effects of lime dose and chemical additions. Journal of American Water Works Association 95 (11), 97e109.

Robinson, J.A., 1985. Determining microbial kinetic parameters using nonlinear regression analysis: advantages and limitations in microbial ecology. Advances in Microbial Ecology 8, 61e114. Seidel, C.J., McGuire, M.J., Summers, R.S., Via, S., 2005. Have utilities switched to chloramines? Journal of American Water Works Association 97 (10), 87e97. Singer, P.C., Arora, H., Dundore, E., Brophy, K. , Weinberg, H.S. 1999a. Control of haloacetic acid concentrations by biofiltration: a case study, pp. 1771e1779. Singer, P.C., Harrington, G.W., Cowman, G.A., Smith, M.E., Schecter, D.S., Harrington, L.J., 1999b. Impacts of Ozonation on the Formation of Chlorination and Chloramination ByProducts. Awwa Research Foundation, Denver, CO. Smith, L.H., Kitandis, P.K., McCarty, P.L., 1997. Numerical modeling and uncertainties in rate coefficients for methane utilization and TCE cometabolism by a methane-oxidizing mixed culture. Biotechnology and Bioengineering 53 (3), 320e331. Smith, L.H., McCarty, P.L., Kitandis, P.K., 1998. Spreadsheet method for evaluation of biochemical reaction rate coefficients and their uncertainties by weighted nonlinear least-squares analysis of the integrated Monod equation. Applied and Environmental Microbiology 64 (6), 2044e2050. Suzuki, I., Dular, U., Kwok, S.C., 1974. Ammonia or ammonium ion as substrate for oxidation by Nitrosomonas europaea cells and extracts. Journal of Bacteriology 120 (1), 556e558. Symons, J.M., Awwa, Us Epa, 1982. Treatment Techniques for Controlling Trihalomethanes in Drinking Water. American Water Works Association, Denver, CO. USEPA, 2005. Economic Analysis for the Final Stage 2 Disinfectants and Disinfection Byproducts Rule. USEPA. Wahman, D.G., Henry, A.E., Katz, L.E., Speitel, G.E.J., 2006a. Cometabolism of trihalomethanes by mixed culture nitrifiers. Water Research 40, 3349e3358. Wahman, D.G., Katz, L.E., Speitel, G.E.J., 2005. Cometabolism of trihalomethanes by Nitrosomonas europaea. Applied and Environmental Microbiology 71 (12), 7980e7986. Wahman, D.G., Katz, L.E., Speitel, G.E.J., 2006b. Trihalomethane cometabolism by a mixed-culture nitrifying biofilter. Journal of American Water Works Association 98 (12), 2e20. Williams, S.L., Williams, R.L., Gordon, A.S., 1997. The impact of bacterial degradation of haloacetic acids (HAA) in the distribution system, pp. 461e465. Williams, S.L., Williams, R.L., Yuan, J., 1998. Bacterial degradation of haloacetic acids in the distribution system, pp. 632e637. Xie, Y., Zhou, H., 2000. Biologically active carbon for haloacetic acid control, pp. 1437e1443. Xie, Y.F.F., Zhou, H.J., 2002. Use of BAC for HAA removal: part 2, column study. Journal of American Water Works Association 94 (5), 126e134. Zhang, P., Lapara, T.M., Goslan, E.H., Xie, Y.F., Parsons, S.A., Hozalski, R.M., 2009. Biodegradation of haloacetic acids by bacterial isolates and enrichment cultures from drinking water systems. Environmental Science & Technology 43 (9), 3169e3175.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 8 1 e1 6 9 0

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Influence of tetracycline resistance on the transport of manure-derived Escherichia coli in saturated porous media Jacob J. Walczak a, Sonia L. Bardy b, Lucia Feriancikova a, Shangping Xu a,* a b

Department of Geosciences, 3209 N Maryland Ave, University of WisconsineMilwaukee, Milwaukee, WI 53211, USA Department of Biological Sciences, 3209 N Maryland Ave, University of WisconsineMilwaukee, Milwaukee, WI 53211, USA

article info

abstract

Article history:

In this research, tetracycline resistant (tetR) and tetracycline susceptible (tetS) Escherichia

Received 24 August 2010

coli isolates were retrieved from dairy manure and the influence of tetracycline resistance

Received in revised form

on the transport of E. coli in saturated porous media was investigated through laboratory

9 December 2010

column transport experiments. Experimental results showed that tetR E. coli strains had

Accepted 10 December 2010

higher mobility than the tetS strains in saturated porous media. Measurements of cell

Available online 21 December 2010

surface properties suggested that tetR E. coli strains exhibited lower zeta potentials than the tetS strains. Because the surface of clean quartz sands is negatively charged, the repulsive

Keywords:

electrostatic double layer (EDL) interaction between the tetR cells and the surface of sands

Manure

was stronger and thus facilitated the transport of the tetR cells. Although no difference was

Antibiotic resistant bacteria

observed in surface acidity, cell size, lipopolysaccharides (LPS) sugar content and cell-

Escherichia coli

bound protein levels between the tetR and tetS strains, they displayed distinct outer

Bacteria transport

membrane protein (OMP) profiles. It was likely that the difference in OMPs, some poten-

Groundwater contamination

tially related to drug efflux pumps, between the tetR and tetS strains led to alteration in cell surface properties which in turn affected cell transport in saturated porous media. Findings from this research suggested that manure-derived tetR E. coli could spread more widely in the groundwater system and pose serious public health risks. ª 2010 Elsevier Ltd. All rights reserved.

1.

Introduction

The occurrence and spread of antibiotic resistant bacteria in the environment undermines our ability to prevent and control microbial infections and is becoming a growing public health challenge both within the United States and across the world. It was estimated that antibiotic resistant pathogens are responsible for more than 2 million illnesses and 14 000 deaths each year in the United States (Pruden et al., 2006; World Health Organization, 2000). The Centers for Disease Control and Prevention (CDC) reported that antibiotic resistance cost the United States more than 4.5 billion dollars in 1990 (Institute of Medicine, 1998).

For several decades, antibiotics have been commonly used in animal farms at therapeutic levels to treat diseases and at sub-therapeutic levels for growth promotion and prophylactic purposes (Institute of Medicine, 1988; Kumar et al., 2005; Mellon et al., 2001; Teuber, 2001). The widespread use of antibiotics in animal farm environments has resulted in high levels of antibiotic resistant bacteria in animal waste (Halbert et al., 2006; Hofacre et al., 2000; Parveen et al., 2006; Ray et al., 2006, 2007; Sato et al., 2004, 2005; Varela et al., 2008; Varga et al., 2008a, 2008b, 2009). Parveen et al. (2006) reported that 85%, 81%, 91% and 80% of Escherichia coli isolates retrieved from manures produced in swine, dairy, poultry, and beef farms respectively, were resistant to at least one antibiotic drug.

* Corresponding author. Tel.: þ1 414 229 6148. E-mail address: [email protected] (S. Xu). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.12.014

1682

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 8 1 e1 6 9 0

Manure produced in animal farms is usually stored in deep pits or outdoor lagoons before being applied to agricultural fields as a source of fertilizer (Burkholder et al., 2007; Gollehon et al., 2001; Sapkota et al., 2007). Leakage from deep pits and lagoons and downward infiltration of water through manureladen soil can lead to the pollution of groundwater by antibiotic resistant bacteria (Anderson and Sobsey, 2006; Koike et al., 2007; Mackie et al., 2006; Mckeon et al., 1995; Sapkota et al., 2007; Storteboom et al., 2007). Mckeon et al. (1995) found that 100% of the non-coliforms and 87% of the coliforms isolated from rural groundwater samples were resistant to at least one of 16 antibiotics, with resistance most commonly directed toward novobiocin, cephalothin, and ampicillin. Approximately 60% of the coliforms were resistant to multiple drugs. Sapkota et al. (2007) reported that a concentrated swine feeding operation resulted in groundwater pollution by Enterococci that were resistant to erythromycin, tetracycline and clindamycin. Anderson and Sobsey (2006) found that more than 80% of the E. coli isolates from the groundwater influenced by swine farms were resistant to tetracycline and chlortetracycline. Additionally, many of the E. coli isolates were also resistant to streptomycin, trimethoprim and ampicillin. Because groundwater is the primary source of drinking water, particularly in areas where animal farms are located, contamination of groundwater by antibiotic resistant bacteria poses a direct public health threat (Ellefson et al., 2002; Solley et al., 1998). Despite the growing concerns of groundwater contamination by manure-derived antibiotic resistant bacteria, our knowledge of the transport of antibiotic resistant bacteria in the groundwater systems remains very limited. Rysz and Alvarez (2006) investigated the transport of tetracycline resistant Burkholderia cepacia and found that >46% of the bacterial cells were able to travel through a 15-cm column packed with sand and higher breakthrough concentrations were observed when the concentration of the bacterial cells was increased. The influence of tetracycline resistance on the transport of B. cepacia, however, was not specifically examined. The main goal of this research is to evaluate the impact of tetracycline resistance on the transport of manure-derived E. coli through column transport experiments. Such information is needed to assess the health risks associated with groundwater contamination by antibiotic resistant bacteria and to improve manure management practices that aim at mitigating this problem.

2.

Materials and methods

2.1. Isolation and antimicrobial susceptibility test of E. coli E. coli used in this research was isolated from manure collected from a family dairy farm (w50 milking cows) located in Ozaukee County, WI using standard protocols established by US EPA (2000). Briefly, the collected manure samples were suspended in sterile phosphate buffered saline (PBS) solution and filtered through sterile PVDF 0.45 mm membranes (Millipore). The membranes were flipped and placed onto modified mTEC agar plates (Becton Dickinson). The plates were incubated at 35  C for 2 h and then 44.5  C for 22 h. Tentative E. coli isolates retrieved from the mTEC agar plates were confirmed with

Enterotube II (Becton Dickinson) and MacConkey II agar plates containing 4-methylumbelliferyl-D-glucuronide (MUG) (Becton Dickinson). The isolated E. coli were tested for their susceptibility to 7 representative antibiotics using MuellereHinton agar plates amended with various antibiotics (Clinical and Laboratory Stardards Institute, 2006; Walczak and Xu, 2011). For each antibiotic reagent, two different concentrations were tested. No E. coli isolates were resistant to gentamicin and ciprofloxacin, while resistance to cephalothin, ampicillin, erythromycin and tetracycline was prevalent (Walczak and Xu, 2011). E. coli isolates that differed in tetracycline resistance but had otherwise similar antibiotic susceptibility patterns were selected for the column transport study (Table 1). To minimize variations that may be caused by different growth conditions and nutrient status among different cows, the E. coli isolates selected for this research were all from the same cow. The tetR strains were referred to as RES1 and RES2 and the tetS strains were denoted by SUS1 and SUS2. Polymerase chain reaction (PCR) assays were performed to determine a total of 20 tetR genes (12 efflux genes and 8 ribosome protection genes) (Hu et al., 2008). The primers used for the PCR assays and the primer-dependent annealing temperatures can be found elsewhere (Aminov et al., 2002, 2001; Hu et al., 2008; Miranda et al., 2003). Plasmid DNA was isolated from each strain by alkaline lysis (Sambrook and Russell, 2001), and served as the template for the PCR reactions. The PCR mixture also contained 0.3 mM of each primer (Invitrogen), 0.4 units of Vent polymerase (New England Biolabs), 200 mM dNTPs, 3% dimethyl sulfoxide and 1  PCR buffer. The PCR amplification was performed using a Mastercycler thermocycler (Eppendorf). The temperature program consisted of an initial denaturing step of 94  C for 5 min, followed by 30 cycles of 94  C for 30 s, annealing for 30 s, extension at 72  C for 30 s, and a final extension of 10 min at 72  C. Negative control reactions were included for each set of primers. PCR products were analyzed on 1% agarose gel that was stained with ethidium bromide.

2.2.

Column transport studies

Duplicate chromatography columns measuring 2.5 cm in diameter and 15 cm in length were used for the column transport experiments. The columns were packed with silica sands (US Silica) (size range: 0.707e0.841 mm) that were alternately cleaned with hot, concentrated nitric acids and diluted NaOH. The porosity of the sand was 0.356. Peristaltic pumps (ColeeParmer) were used to maintain a constant specific discharge value of 0.31 cm/min. Packed sand columns were equilibrated with >40 pore volumes of background electrolyte solution (1e100 mM KCl) before the bacteria transport experiment. E. coli preserved in 20% glycerol under 80  C was streaked onto Muller-Hinton (MH) agar plates. After overnight incubation at 37  C, cells from the freshly formed colonies were transferred to culture tubes containing 15 mL Luria-Bertani (LB) broth. The culture tubes were incubated at 37  C for 6 h. The starter culture was used to inoculate LB broth (1:500 dilution ratio), which was then incubated at 37  C for 18 h. The bacterial cells were harvested using centrifuge (4000 g, 10 min, 4  C). To

1683

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 8 1 e1 6 9 0

Table 1 e Antibiotic resistance pattern of the E. coli isolates used for the transport experiments. E. coli isolate

Tetracycline a

RES1 RES2 SUS1 SUS2

Ampicillin

Cephalothin

Gentamicin

Ciprofloxacin

Erythromycin

Sulfomethoxazole

4

16

8

32

8

32

4

16

1

4

3

15

38

152

þ þ e e

þ þ e e

þ þ þ e

þ þ e e

þ þ þ þ

þ þ þ e

e e e e

e e e e

e e e e

e e e e

þ þ þ þ

þ þ þ þ

e e e e

e e e e

a Unit is mg/L.

remove the growth medium, the bacterial pellet was rinsed 4 times with the appropriate electrolyte solution. The concentration of cells was then adjusted to ∼4  107 cell/ml for the column transport experiments. The pH of the background electrolyte solutions and the cell suspension was around 5.7. The effluents of the columns were connected to flow-through quartz cuvettes (NSG Precision) and the cell concentration was determined at a wavelength of 220 nm at 30-s intervals using a spectrophotometer (Shimadzu UV-1700). After 60 min of injection (w3.5 pore volumes), the columns were flushed with background electrolyte solution until the absorbance of effluent returned to the background values. The deposition kinetics of bacterial cells in saturated porous media under clean-bed conditions is commonly quantified by the deposition rate coefficient (K ), which is determined using the following equation (Kretzschmar et al., 1999)
 n C K ¼  ln qL C0

(1)

where v is approach velocity (cm/min), L is the length of the packed bed (cm), q is the porosity of the porous medium (cm3/ cm3), C is concentration of bacteria cell in the effluent under clean-bed conditions (cell/mL) and C0 is the bacteria concentration in the influent (cell/mL). The values of C/C0 were determined for each column experiment by calculating the average normalized breakthrough concentrations (i.e., C/C0) measured between 1.8 and 2 pore volumes (Walker et al., 2005).

2.3.

Determination of cell size and surface properties

The transport of bacteria cells in saturated porous media is influenced by a variety of cell surface properties such as zeta potential, surface charge and hydrophobicity which in turn were related to factors such as cell-bound proteins and LPS (Foppen and Schijven, 2006). In this research, a range of cell surface properties were determined and related to the transport of tetR and tetS E. coli cells. Freshly harvested bacterial cells were suspended in appropriate KCl solutions (1e100 mM) at a concentration of ∼107 cells/mL. The zeta potential of the bacterial cells was then measured with a Brookhaven ZetaPALS analyzer utilizing phase analysis light scattering. The hydrophobicity of the cells was determined through microbial adhesion to hydrocarbon (MATH) test (Pembrey et al., 1999). The MATH test involved the mixing of 1 mL of n-dodecane and 4 mL of cell suspension prepared using KCl solutions. The mixture was vortexed for 2 min and then allowed to stand still for 15 min. Cell concentration in the aqueous phase was determined at a wavelength

of 546 nm. The fraction of bacterial cells that partitioned into the hydrocarbon phase was calculated based on mass balance and expressed cell hydrophobicity. To quantify EDTA-extractable cell-bound proteins, fresh cell suspensions (∼3  108 cell/ml) were prepared and mixed with 2.5% EDTA solution on a 2:3 (v/v) basis (Zhang et al., 1999). Following a 30-min incubation period at 4  C, the mixture was centrifuged at 10,400 g (4  C) for 50 min. The supernatant was then decanted and filtered through 0.45 mm filters. Protein contents in the filtrates were quantified using the Coomassie brilliant blue method developed by Bradford (1976). Standard solutions prepared with human serum albumin were used to calibrate this method. Additionally, the outer membrane proteins (OMPs) of the E. coli cells were extracted and profiled using sodium dodecyl-sulfate polyacrylamide gel (SDS-PAGE) (Ben Abdallah et al., 2009; Gatewood et al., 1994; Xu et al., 2006). The bacterial cells were harvested at 4000 g for 10 min at 4  C and rinsed twice with sterile 0.15 M NaCl. The cells were suspended in 5 mL sterile 0.15 M NaCl and subsequently disrupted by intermittent sonic oscillation (50 W, 8 cycles of 15 s of sonication, VirTis Virsonic). Following centrifugation at 5000 g for 40 min to remove cellular debris, the supernatant was transferred and centrifuged at 100 000 g for 40 min at 4  C. The pellets were resuspended in 2% sodium lauryl sarcosinate (Sigma), incubated at room temperature for 1 h and then centrifuged at 100,000 g (40 min, 4  C). The pelleted OMPs were suspended in 0.25 mL sterile 0.15 M NaCl solution and resolved on 10% sodium dodecyl-sulfate polyacrylamide gel (Laemmli). The gel was fixed in 4% perchloric acid and strained using 0.1% Coomassie G250 in hot 4% perchloric acid (Faguy et al., 1996). To extract cell-bound LPS, 5 mL cell suspension was placed into 50 mL centrifuge tubes and subjected to sonic disruption (50 W, 20 s) (Liu et al., 2007). The resulting suspension was centrifuged (10 000 g, 40 min, 4  C) and filtered through 0.2 mm cellulose acetate filters. The contents of LPS in the filtrates were measured using the phenol-sulfuric acid method and xanthan gum was used as the calibration standard (Du Bois et al., 1956). Acid-base titration of cell suspensions was performed to determine the surface acidity of bacterial cells. Cell suspensions were prepared in the same fashion as those used in the column transport experiments. Two hundred milliliter of the suspension was added to a medium bottle and purged with high purity nitrogen gas for >60 min to remove CO2. After the purging step, running nitrogen gas flow was maintained right above the surface of the suspension to maintain a CO2 free environment. Sulfuric acid solution (0.1600 N, Hach Company) was then introduced into the suspension using Hach digital

1684

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 8 1 e1 6 9 0

titrator. Once the pH of the suspension dropped to w4, NaOH (0.1600 N, Hach Company) was added at small volume increments to raise the pH to ∼10. The pH of the suspension was continuously monitored and recorded using a pH probe (AccupH, Fisher Scientific). The acidity of bacterial cells was calculated based on the amount of NaOH that was consumed to raise the pH of the cell suspension from 4 to 10. The size of the bacterial cells suspended in the KCl solutions were measured by taking photos using a Nikon Eclipse 50i microscope, which was equipped with a Photometric coolsnap ES digital camera and MetaMorph software. The length and width of a minimum of 50 cells were determined using the ImageJ software and the equivalent radii of the cells were then calculated.

3.

Results and discussion

3.1.

Characterization of tetR genes

PCR assays showed that none of the tetR genes were present in the tetS strains while the efflux gene, tetB, was present in both tetR strains (Fig. 1). This observation is consistent to previously reported results which suggested that the tetB gene was among the most commonly detected tetR genes in E. coli and coliforms (Hu et al., 2008; Marshall et al., 1983). In Hu et al. (2008), the tetB gene was found in 41% of tetracycline resistant E. coli strains isolated from a river basin impaired by both human and animal farm waste.

3.2.

Transport of the bacterial cells in the sand packs

The two tetS E. coli strains differ in ampicillin and cephalothin susceptibility but the close match in their breakthrough concentrations suggested that the difference in ampicillin and

cephalothin resistance had minimal impact on their mobility (Figs. 2 and 3). The breakthrough concentrations of the tetR E. coli strains, however, were significantly higher than those of the tetS strains under all ionic strength conditions, suggesting that tetracycline resistance can enhance the mobility of manure-derived E. coli (Figs. 2 and 3). In 1 mM KCl, for instance, the breakthrough concentrations of the tetS were ∼15% lower than the breakthrough concentrations of the tetR strains. Accordingly, the values of the deposition rate coefficients (i.e., K ) for the tetS and tetR strains were 0.0325(0.0012) (SUS1), 0.0306(0.0018) (SUS2), 0.0172(0.000005) (RES1) and 0.0169 (0.0009) (RES2) min1, respectively. When the ionic strength was increased to 3 mM, there was a significant drop in the breakthrough concentrations for all 4 E. coli isolates, suggesting that within this range, higher ionic strength facilitated their deposition at the surface of quartz sand (Figs. 2 and 3). While a further increase in ionic strength led to slightly lower breakthrough concentrations for the tetS strains, the transport of the tetR strains remained virtually unchanged (Figs. 2 and 3). Overall, the tetR E. coli isolates displayed significantly higher mobility under the ionic strength conditions tested in this research (Fig. 3). Our results suggested that environmental E. coli isolates could display marked variability in mobility, which was reported in several recent publications (Bolster et al., 2010, 2009; Foppen et al., 2010; Lutterodt et al., 2009). Bolster et al. (2009) and Bolster et al. (2010) compared the transport of 12 and 8 E. coli strains isolated from different animal sources (poultry, horse, beef and dairy cattle, human and wildlife) in saturated sands, respectively, and observed large variability in their mobility. Lutterodt et al. (2009) investigated the movement of 6 E. coli strains obtained from a soil used for cattle grazing in columns packed with sands and reported that the sticking efficiencies varied by a factor of 4e10. Foppen et al. (2010) examined the transport behavior of 54 E. coli strains and found that the attachment efficiency varied by a factor of ∼6. The observed variability in E. coli transport behavior was found to be related to a range of factors such as cell surface autotransporter proteins (e.g., Ag43 protein) (Lutterodt et al., 2009), cell width (Bolster et al., 2010; Bolster et al., 2009), cell surface/zeta potential (Bolster et al., 2010), cell LPS structure (Foppen et al., 2010) as well as cell fimbriae (Foppen et al., 2010).

3.3.

Fig. 1 e PCR detection of tetB in the E. coli isolates. Lane M: 100 bp ladder (New England Biolabs). Lanes 1e4: SUS1, SUS2, RES1, and RES2. The size of the amplicon for the tetB gene was 206 bp.

Size and surface properties of bacterial cells

The transport behavior of bacterial cells in the porous media is governed by the energy interactions between the cells and the surface of the solid matrix, which depends on a range of factors such as the Lifshitzevan der Waals force, electrostatic double layer (EDL) interactions, acid-base forces, hydrophobicity interactions and steric effects, which in turn are affected by cell size as well as cell-bound LPS and proteins and so on (Lindqvist and Bengtsson, 1991; Ong et al., 1999). In this research, various cell surface properties were measured and related to the observed difference in the transport behavior of the tetR and tetS E. coli strains. The EDL interactions between bacterial cells and sand surface were closely related to their zeta potentials. The measured zeta potential values of the bacterial cells and the

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 8 1 e1 6 9 0

1685

Fig. 2 e Breakthrough concentrations of the tetracycline susceptible (SUS1 and SUS2) and tetracycline resistant (RES1 and RES2) E. coli strains in saturated porous media under ionic strength conditions of 1, 3, 10, 30 and 100 mM KCl.

sands were negative (Fig. 4A), suggesting repulsive EDL interactions. In general, the zeta potentials of the tetR strains were more negative than those of the tetS strains. As a result, the repulsive interaction between the surface of quartz sands and the tetR E. coli cells were stronger and the deposition rates

Fig. 3 e Clean-bed deposition rate coefficients (K, minL1) for the 4 E. coli strains under ionic strength conditions of 1, 3, 10, 30 and 100 mM KCl. The values of K were calculated using Eq. (1). Error bars represent standard deviation of duplicate experiments.

should thus be lower. It is noteworthy that the zeta potentials of the tetR strains remained virtually unchanged with ionic strength, while the zeta potentials of the tetS strains increased slightly with ionic strength. This is consistent with the observation that the transport of the tetR strains was less sensitive to changes in ionic strength. The tetR strains were more hydrophobic than the tetS ones (Fig. 4B). On average, slightly over 90% of the tetR E. coli cells partitioned into the hydrocarbon phase in the MATH tests, while less than 85% of the tetS E. coli cells migrated from the aqueous phase into the hydrocarbon phase. In this research we observed higher mobility for the more hydrophobic tetR strains. It was previously suggested, however, that cell hydrophobicity could enhance the attachment of bacterial cells to clean quartz sands (Bolster et al., 2006; Mccaulou et al., 1994). It thus seemed that the EDL interactions between E. coli cells and quartz sands were more significant than the hydrophobic interactions. This is consistent to recent findings which suggested that while cell zeta potential was significantly related to the transport of manure-derived E. coli strains, the relationship between cell hydrophobicity and E. coli mobility was statistically insignificant (Bolster et al., 2010). Size is another factor that could influence the transport of colloid-sized particles in saturated porous media (Bolster et al., 2010; Yao et al., 1971). In this research, the size of the

1686

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 8 1 e1 6 9 0

Fig. 5 e Size (equivalent radius) of the bacterial cells suspended in 1, 3, 10, 30 and 100q mM of KCl. The ffiffiffiffiffiffiffiffiffiffiffiffi C equivalent size was calculated as LC 3W p , where LC and WC represent the length and width of the cell, respectively (Haznedaroglu et al., 2008). Error bars represent the standard deviation of a minimum of 50 measurements.

Fig. 4 e Zeta potential (A) and MATH test results (B). The MATH test results were expressed as the fraction of bacterial cells partitioned into the hydrocarbon phase. Error bars represent the standard deviation of triplicate measurements. For zeta potential, one measurement contained a minimum of 5 runs.

bacterial cells suspended in 5 different types of electrolyte solutions was measured. Our results showed that the size of the bacterial cells was not sensitive to ionic strength and the tetR and tetS cells have practically similar sizes (Fig. 5). Through titration of cells suspended in 3 mM KCl, the acidity of the bacterial cells were measured as 6.10 (0.16), 6.54 (0.60), 7.07 (0.83) and 6.28 (0.24)  104 meq/108 cells for SUS1, SUS2, RES1 and RES2, respectively, suggesting no significant difference between the susceptible and resistant isolates. The potentiometric titration curves showed that the pKa values of the acid-base functional groups on the surfaces of both tetracycline resistant and tetracycline susceptible cells were between 4e5 as well as 9e10. These pKa values correspond to the carboxylic/phosphoric and hydroxyl/amine groups, all of which are important components of LPS, proteins and phospholipids located on the surface of bacterial cells (Hong and Brown, 2006). Under the experimental pH conditions (∼5.7), the carboxylic/phosphoric groups were deprotonated and contributed to cell surface charges. Cell-bound LPS and protein contents varied within the range of 17.4e33.1 and 4.7e13 mg/108 cells, respectively (Fig. 6).

Fig. 6 e Comparison of the LPS sugar (A) and protein contents (B) of the tetracycline resistant and susceptible cells suspended in 1, 3, 10, 30 and 100 mM KCl. The error bars represent standard deviation of triplicate extraction attempts.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 8 1 e1 6 9 0

These values were consistent with previously reported measurements using manure-derived E. coli (Haznedaroglu et al., 2008). There were significant variations among the extraction attempts of LPS and protein, and among the electrolyte solutions used to prepare cell suspensions. Overall, we did not observe a clear pattern between the tetR and tetS strains with regard to LPS sugar and protein contents. Although no difference was identified in the cell-bound protein contents, results of SDS-PAGE analysis of OMPs suggested that different proteins existed on the outer membrane of tetR and tetS strains (Fig. 7). Specifically, there were at least four proteins present in the outer membrane of the tetR strains that were absent in the tetS strains (indicated by arrows in Fig. 7). These proteins had approximate molecular masses of 54, 47, 44 and 40 kDa. Additionally, there were three proteins (indicated with arrow heads in Fig. 7) that were present in the tetS strains but were absent in the tetR strains. As an antimicrobial agent, tetracycline has an intracellular target, inhibiting bacterial protein synthesis by disrupting the interaction of aminoacyl-tRNA with the ribosome (Walsh, 2003). Mechanisms specific to tetracycline resistance include efflux pump, ribosomal protection and modification of the antibiotic (Wax, 2008). The ribosomal protection mechanisms involve soluble structural homologues (e.g., TetM proteins) of elongation factors, which can bind to the ribosome and destabilize the interaction between tetracycline and their cellular target (Burdett, 1996; Dantley et al., 1998). Resistance to tetracycline through drug destruction was relatively rare and it was recently reported that TetX, a flavin-dependent monooxygenase, could hydroxylate the tetracycline substrate into an unstable compound which subsequently underwent non-enzymatic decomposition (Yang et al., 2004). The efflux pumps that are encoded by the tetR genes (e.g., tetB) involve the transport of tetracycline from the cytoplasm to the

periplasm through proteins inserted in the cytoplasmic membrane (Walsh, 2003). Out of the 20 tetR genes examined in this research, only the tetB gene was detected in the tetR strains (Fig. 1). Because the protein involved in the efflux pump is not exposed to the outside of the bacterial cells, it was unlikely that the presence of TetB could impact cell mobility. The tetR gene family, however, does not represent all the mechanisms responsible for tetracycline resistance in E. coli. Additional efflux pumps that involve multi-protein assemblies that often traverse both the inner and outer membranes of E. coli (e.g., the AcrAB-TolC pump) could lead to tetracycline resistance (Alekshun and Levy, 2007; de Cristobal et al., 2006; Xu et al., 2006). Increased expression of proteins such as TolC, OmpC, OmpW, along with decreased amounts of LamB and NlpB proteins have been observed in tetR E. coli (Xu et al., 2006; Zhang et al., 2008). Additionally, it was observed that deletion of TolC led to increased sensitivity of E. coli to tetracycline (Zhang et al., 2008). Because processed TolC has a molecular mass of 52 kDa, it is likely that the protein extracted from the tetR strains that migrated approximately at 54 kDa was TolC (Fig. 7). The crystal structure of the TolC protein was recently resolved and it was shown that it is a trimeric, 471-residue protein that contains an a-helical barrel and a b-barrel (Koronakis et al., 2000). The a-helical domain, which forms a tunnel through the periplasm and measure 10 nm in length, is anchored to bacterial wall by the contiguous b-barrel, which has a length of 4 nm and extends to the outside of the outer membrane. In addition to the approximately 54 kDa protein enriched in the tetR cells, there were at least three other proteins that are differentially enriched in the tetR cells. It is likely that the presence of these proteins, in combination with the absence of other proteins (indicated by arrow heads, Fig. 7), contributed to alterations in cell surface properties (e.g., zeta potential), which in turn impacted the mobility of the E. coli strains. Recently, it was reported that Ag43, an outer membrane protein of E. coli, could enhance the attachment of E. coli cells to the surface of quartz sands (Lutterodt et al., 2009). It was proposed that the positive charges of the a-domain of Ag43, which extends from the cell surface, facilitated the attachment of E. coli cells to the negatively-charged quartz surfaces (Lutterodt et al., 2009). While the Ag43 and TolC proteins are structurally different and have different impacts on E. coli transport in saturated porous media (Koronakis et al., 2000; van der Woude and Henderson, 2008), the results of Lutterodt et al. (2009) and this research suggest that cell surface proteins can have strong influences on E. coli transport and more studies will be needed to elucidate the relationship between the abundance, structure and properties of cell surface proteins and bacterial transport in porous media.

3.4. Fig. 7 e Outer membrane protein profiles of SUS1, SUS2, RES1 and RES2 using SDS-PAGE. Molecular masses (kDa) are indicated on the left. Proteins that are present in the tetracycline resistant strains, but absent in the tetracycline susceptible strains, are indicated with an arrow. Proteins that are present in the tetracycline susceptible strains, but absent in the tetracycline resistant strains are indicated with arrow heads.

1687

Environmental implications

As a result of the widespread use of antibiotics in the animal farm environment, high frequencies of antibiotic resistant bacteria have been detected in animal waste (Halbert et al., 2006; Parveen et al., 2006; Ray et al., 2006; Sapkota et al., 2007; Sato et al., 2004; Varga et al., 2008a; Walczak and Xu, 2011). On the dairy farm from which the E. coli strains used in this research were isolated, 13.1%, 72.7%, 80.5% of the E. coli isolates were resistant to tetracycline, cephalothin and

1688

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 8 1 e1 6 9 0

erythromycin, respectively (Walczak and Xu, 2011). Additionally, 100% of the tetR E. coli isolates were multi-drug resistant. The observed high mobility of the tetR E. coli strains indicated that leakage from manure storage structures and application of manure as fertilizers in agricultural fields could potentially lead to the contamination of groundwater by antibiotic resistant bacteria, which in turn could pose serious public health risks when the groundwater, often untreated, is used as a source of drinking water. Drug efflux pumps that involve outer membrane proteins like TolC are also seen in pathogenic bacteria such as Salmonella to gain antibiotic resistance (Ricci et al., 2006; Virlogeux-Payant et al., 2008). It is likely that the surface properties of these pathogens can also be altered in a similar fashion. As a result, the mobility of these pathogens in the subsurface system can be enhanced. Furthermore, it has long been observed that antibiotic resistance genes, which confer bacterial antibiotic resistance, can be transferred among a diverse group of microorganisms through conjugation, transduction and transformation (Levy et al., 1976; Lorenz et al., 1992; Mckeon et al., 1995; Nikolich et al., 1994). The antibiotic resistant genes harbored by antibiotic resistant E. coli, therefore, could potentially be horizontally transferred to bacterial pathogens such as Samonella in the subsurface environment and cause additional public health risks (Hunter et al., 1992; van Essen-Zandbergen et al., 2007).

4.

Conclusion

 In this research, we observed that manure-derived, tetR E. coli strains had higher mobility than tetS E. coli strains within saturated porous media.  The tetR E. coli strains had more negative zeta potentials than the tetS strains. This led to increased repulsive EDL interaction between the tetR E. coli cells and the surface of quartz sands and could explain the observed higher mobility of the tetR strains.  The tetR and tetS E. coli strains had distinct outer membrane proteins profiles. It is likely that such difference led to alterations in cell surface properties (such as zeta potential), which in turn affected the transport of the tetR and tetS E. coli strains.

Acknowledgement SX was supported by University of Wisconsin Milwaukee and University of Wisconsin Groundwater Research Program (WR10R007). SLB is the recipient of an NIH grant (5R00GM08314704). We thank Dr. Douglas A. Steeber, Dr. Heather A. Owen and Steven E. Hardcastle for their assistance. We thank two reviewers whose comments led to improvements of our manuscript.

references

Alekshun, M.N., Levy, S.B., 2007. Molecular mechanisms of antibacterial multidrug resistance. Cell 128, 1037e1050.

Aminov, R.I., Chee-Sanford, J.C., Garrigues, N., Teferedegne, B., Krapac, I.J., White, B.A., Mackie, R.I., 2002. Development, validation, and application of PCR primers for detection of tetracycline efflux genes of gram-negative bacteria. Applied and Environmental Microbiology 68, 1786e1793. Aminov, R.I., Garrigues-Jeanjean, N., Mackie, R.I., 2001. Molecular ecology of tetracycline resistance: development and validation of primers for detection of tetracycline resistance genes encoding ribosomal protection proteins. Applied and Environmental Microbiology 67, 22e32. Anderson, M.E., Sobsey, M.D., 2006. Detection and occurrence of antimicrobially resistant E-coli in groundwater on or near swine farms in eastern North Carolina. Water Science and Technology 54, 211e218. Ben Abdallah, F., Kallel, H., Bakhrouf, A., 2009. Enzymatic, outer membrane proteins and plasmid alterations of starved Vibrio parahaemolyticus and Vibrio alginolyticus cells in seawater. Archives of Microbiology 191, 493e500. Bolster, C.H., Cook, K.L., Marcus, I.M., Haznedaroglu, B.Z., Walker, S.L., 2010. Correlating transport behavior with cell properties for Eight Porcine Escherichia coli isolates. Environmental Science & Technology 44, 5008e5014. Bolster, C.H., Haznedaroglu, B.Z., Walker, S.L., 2009. Diversity in cell properties and transport behavior among 12 different environmental Escherichia coli isolates. Journal of Environmental Quality 38, 465e472. Bolster, C.H., Walker, S.L., Cook, K.L., 2006. Comparison of Escherichia coli and Campylobacter jejuni transport in saturated porous media. Journal of Environmental Quality 35, 1018e1025. Bradford, M.M., 1976. A rapid and sensitive method for the quantitation of microgram quantities of protein utilizing the principle of protein-dye binding. Analytical Biochemistry 72, 248e254. Burdett, V., 1996. Tet(M)-promoted release of tetracycline from ribosomes is GTP dependent. Journal of Bacteriology 178, 3246e3251. Burkholder, J., Libra, B., Weyer, P., Heathcote, S., Kolpin, D., Thorne, P.S., Wichman, M., 2007. Impacts of waste from concentrated animal feeding operations on water quality. Environmental Health Perspectives 115, 308e312. Clinical and Laboratory Stardards Institute, 2006. Performance Standards for Antimicrobial Disk Susceptibility Tests; Approved Standard - Ninth Edition. CLSI, Wayne, PA. Dantley, K.A., Dannelly, H.K., Burdett, V., 1998. Binding interaction between Tet(M) and the ribosome: requirements for binding. Journal of Bacteriology 180, 4089e4092. de Cristobal, R.E., Vincent, P.A., Salomon, R.A., 2006. Multidrug resistance pump AcrAB-TolC is required for high-level, Tet(A)mediated tetracycline resistance in Escherichia coli. Journal of Antimicrobial Chemotherapy 58, 31e36. Du Bois, M., Gilles, K.A., Hamilton, J.K., Rebers, P.A., Smith, F., 1956. Colorimetric method for determination of sugars and related substances. Analytical Chemistry 28, 350e356. Ellefson, B.R., Mueller, G.D., Buchwald, C.A., 2002. Water Use in Wisconsin, 2000. U. S. Geological Survey. open-file report 02e356. Faguy, D.M., Bayley, D.P., Kostyukova, A.S., Thomas, N.A., Jarrell, K.F., 1996. Isolation and characterization of flagella and flagellin proteins from the thermoacidophilic archaea Thermoplasma volcanium and Sulfolobus shibatae. Journal of Bacteriology 178, 902e905. Foppen, J.W., Lutterodt, G., Roling, W.F.M., Uhlenbrook, S., 2010. Towards understanding inter-strain attachment variations of Escherichia coli during transport in saturated quartz sand. Water Research 44, 1202e1212. Foppen, J.W.A., Schijven, J.F., 2006. Evaluation of data from the literature on the transport and survival of Escherichia coli and

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 8 1 e1 6 9 0

thermotolerant coliforms in aquifers under saturated conditions. Water Research 40, 401e426. Gatewood, D.M., Fenwick, B.W., Chengappa, M.M., 1994. GrowthCondition dependent expression of Pasteurella-Haemolytica a1 outer-membrane proteins, Capsule, and Leukotoxin. Veterinary Microbiology 41, 221e233. Gollehon, N., Caswell, M., Ribaudo, M., Kellogg, R., Lander, C., Letson, D., 2001. Confined Animal Production and Manure Nutrients. Resource Economics Division, Economic Research Service. U.S. Department of Agriculture, Washington DC. 39. Halbert, L.W., Kaneene, J.B., Ruegg, P.L., Warnick, L.D., Wells, S.J., Mansfield, L.S., Fossler, C.R., Campbell, A.M., Geiger-Zwald, A.M., 2006. Evaluation of antimicrobial susceptibility patterns in Campylobacter spp isolated from dairy cattle and farms managed organically and conventionally in the midwestern and northeastern United States. Journal of the American Veterinary Medical Association 228, 1074e1081. Haznedaroglu, B.Z., Bolster, C.H., Walker, S.L., 2008. The role of starvation on Escherichia coli adhesion and transport in saturated porous media. Water Research 42, 1547e1554. Hofacre, C.L., ’de Cotret, A.R., Maurer, J.J., Garritty, A., Thayer, S.G. , 2000. Presence of fluoroquinolone-resistant coliforms in poultry litter. Avian Diseases 44, 963e967. Hong, Y., Brown, D.G., 2006. Cell surface acid-base properties of Escherichia coli and Bacillus brevis and variation as a function of growth phase, nitrogen source and C: N ratio. Colloids and Surfaces B-Biointerfaces 50, 112e119. Hu, J.Y., Shi, J.C., Chang, H., Li, D., Yang, M., Kamagata, Y.C., 2008. Phenotyping and genotyping of antihiotic-resistant Escherichia coli isolated from a natural river basin. Environmental Science & Technology 42, 3415e3420. Hunter, J.E.B., Shelley, J.C., Walton, J.R., Hart, C.A., Bennett, M., 1992. Apramycin resistance plasmids in Escherichia coli: possible transfer to salmonella typhimurium in calves. Epidemiology and Infection 108, 271e278. Institute of Medicine, 1988. Human Health Risks with the Subtherapeutic Use of Penicillin or Tetracyclines in Animal Feed. Institute of Medicine, 1998. Antimicrobial Resistance: Issues and Options. National Academy Press, Washington D.C. Koike, S., Krapac, I.G., Oliver, H.D., Yannarell, A.C., CheeSanford, J.C., Aminov, R.I., Mackie, R.I., 2007. Monitoring and source tracking of tetracycline resistance genes in lagoons and groundwater adjacent to swine production facilities over a 3year period. Applied and Environmental Microbiology 73, 4813e4823. Koronakis, V., Sharff, A., Koronakis, E., Luisi, B., Hughes, C., 2000. Crystal structure of the bacterial membrane protein TolC central to multidrug efflux and protein export. Nature 405, 914e919. Kretzschmar, R., Borkovec, M., Grolimund, D., Elimelech, M., 1999. Mobile subsurface colloids and their role in contaminant transport. Advances in Agronomy 66, 121e193. Kumar, K., Gupta, S.C., Chander, Y., Singh, A.K., 2005. Antibiotic use in agriculture and its impact on the terrestrial environment. Advances in Agronomy 87, 1e54. Levy, S.B., Fitzgerald, G.B., Macone, A.B., 1976. Spread of antibiotic-resistant plasmids from chicken to chicken and from chicken to man. Nature 260, 40e42. Lindqvist, R., Bengtsson, G., 1991. Dispersal Dynamics of groundwater bacteria. Microbial Ecology 21, 49e72. Liu, Y., Yang, C.H., Li, J., 2007. Influence of extracellular polymeric substances on Pseudomonas aeruginosa transport and deposition profiles in porous media. Environmental Science & Technology 41, 198e205. Lorenz, M.G., Reipschlager, K., Wackernagel, W., 1992. Plasmid transformation of naturally competent Acinetobacter calcoaceticus in nonsterile soil extract and groundwater. Archives of Microbiology 157, 355e360.

1689

Lutterodt, G., Basnet, M., Foppen, J.W.A., Uhlenbrook, S., 2009. The effect of surface characteristics on the transport of multiple Escherichia coli isolates in large scale columns of quartz sand. Water Research 43, 595e604. Mackie, R.I., Koike, S., Krapac, I., Chee-Sanford, J., Maxwell, S., Aminov, R.I., 2006. Tetracycline residues and tetracycline resistance genes in groundwater impacted by swine production facilities. Animal Biotechnology 17, 157e176. Marshall, B., Tachibana, C., Levy, S.B., 1983. Frequency of tetracycline resistance determinant classes among lactosefermenting coliforms. Antimicrobial Agents and Chemotherapy 24, 835e840. Mccaulou, D.R., Bales, R.C., Mccarthy, J.F., 1994. Use of Short-Pulse experiments to study bacteria transport through porousmedia. Journal of Contaminant Hydrology 15, 1e14. Mckeon, D.M., Calabrese, J.P., Bissonnette, G.K., 1995. Antibioticresistant Gram-negative bacteria in rural groundwater supplies. Water Research 29, 1902e1908. Mellon, M., Benbrook, C., Benbrook, K.L., 2001. Hogging It!: Estimates of Antimicrobial Abuse in Livestock. Miranda, C.D., Kehrenberg, C., Ulep, C., Schwarz, S., Roberts, M.C., 2003. Diversity of tetracycline resistance genes in bacteria from Chilean salmon farms. Antimicrobial Agents and Chemotherapy 47, 883e888. Nikolich, M.P., Hong, G., Shoemaker, N.B., Salyers, A.A., 1994. Evidence for natural horizontal transfer of Tetq between bacteria that normally colonize humans and bacteria that normally colonize livestock. Applied and Environmental Microbiology 60, 3255e3260. Ong, Y.L., Razatos, A., Georgiou, G., Sharma, M.M., 1999. Adhesion forces between E-coli bacteria and biomaterial surfaces. Langmuir 15, 2719e2725. Parveen, S., Lukasik, J., Scott, T.M., Tamplin, M.L., Portier, K.M., Sheperd, S., Braun, K., Farrah, S.R., 2006. Geographical variation in antibiotic resistance profiles of Escherichia coli isolated from swine, poultry, beef and dairy cattle farm water retention ponds in Florida. Journal of Applied Microbiology 100, 50e57. Pembrey, R.S., Marshall, K.C., Schneider, R.P., 1999. Cell surface analysis techniques: what do cell preparation protocols do to cell surface properties? Applied and Environmental Microbiology 65, 2877e2894. Pruden, A., Pei, R.T., Storteboom, H., Carlson, K.H., 2006. Antibiotic resistance genes as emerging contaminants: studies in northern Colorado. Environmental Science & Technology 40, 7445e7450. Ray, K.A., Warnick, L.D., Mitchell, R.M., Kaneene, J.B., Ruegg, P.L., Wells, S.J., Fossler, C.P., Halbert, L.W., May, K., 2006. Antimicrobial susceptibility of Salmonella from organic and conventional dairy farms. Journal of Dairy Science 89, 2038e2050. Ray, K.A., Warnick, L.D., Mitchell, R.M., Kaneene, J.B., Ruegg, P.L., Wells, S.J., Fossler, C.P., Halbert, L.W., May, K., 2007. Prevalence of antimicrobial resistance among Salmonella on midwest and northeast USA dairy farms. Preventive Veterinary Medicine 79, 204e223. Ricci, V., Tzakas, P., Buckley, A., Coldham, N.C., Piddock, L.J.V., 2006. Ciprofloxacin-resistant Salmonella enterica serovar Typhimurium strains are difficult to select in the absence of AcrB and TolC. Antimicrobial Agents and Chemotherapy 50, 38e42. Rysz, M., Alvarez, P.J.J., 2006. Transport of antibiotic-resistant bacteria and resistance-carrying plasmids through porous media. Water Science and Technology 54, 363e370. Sambrook, J., Russell, D.W., 2001. Molecular Cloning: A Laboratory Manual, third ed. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, N.Y.

1690

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 8 1 e1 6 9 0

Sapkota, A.R., Curriero, F.C., Gibson, K.E., Schwab, K.J., 2007. Antibiotic-resistant Enterococci and Fecal Indicators in Surface Water and Groundwater Impacted by a Concentrated Swine Feeding Operation. In: Environmental Health Perspectives, 115 1040e1045. Sato, K., Bartlett, P.C., Kaneene, J.B., Downes, F.P., 2004. Comparison of prevalence and antimicrobial susceptibilities of Campylobacter spp. isolates from organic and conventional dairy herds in Wisconsin. Applied and Environmental Microbiology 70, 1442e1447. Sato, K., Bartlett, P.C., Saeed, M.A., 2005. Antimicrobial susceptibility of Escherichia coli isolates from dairy farms using organic versus conventional production methods. Journal of the American Veterinary Medical Association 226, 589e594. Solley, W.B., Pierce, R.R., Perlman, H.A., 1998. Estimated Use of Water in the United States in 1995. U.S. Geological Survey Circular, 1200, p. 79. Storteboom, H.N., Kim, S.C., Doesken, K.C., Carlson, K.H., Davis, J.G., Pruden, A., 2007. Response of antibiotics and resistance genes to high-intensity and low-intensity manure management. Journal of Environmental Quality 36, 1695e1703. Teuber, M., 2001. Veterinary use and antibiotic resistance. Current Opinion in Microbiology 4, 493e499. US EPA, 2000. Improved Enumeration Methods for the Recreational Water Quality Indicators: Enterococci and Escherichia Coli. http:// www.epa.gov/nerlcwww/RecManv.pdf, p. 52. van der Woude, M.W., Henderson, I.R., 2008. Regulation and function of Ag43 (Flu). Annual Review of Microbiology 62, 153e169. van Essen-Zandbergen, A., Smith, H., Veldman, K., Mevius, D., 2007. Occurrence and characteristics of class 1, 2 and 3 integrons in Escherichia coli, Salmonella and Campylobacter spp. in the Netherlands. Journal of Antimicrobial Chemotherapy 59, 746e750. Varela, N.P., Friendship, R., Dewey, C., Valdivieso, A., 2008. Comparison of Agar Dilution and E-test for antimicrobial susceptibility testing of Campylobacter coli isolates recovered form 80 Ontario swine farms. Canadian Journal of Veterinary Research-Revue Canadienne De Recherche Veterinaire 72, 168e174. Varga, C., Rajic, A., McFall, M.E., Avery, B.P., Reid-Smith, R.J., Deckert, A., Checkley, S.L., McEwen, S.A., 2008a. Antimicrobial resistance in generic Escherichia coli isolated from swine fecal samples in 90 Alberta finishing farms. Canadian Journal of Veterinary Research-Revue Canadienne De Recherche Veterinaire 72, 175e180. Varga, C., Rajic, A., McFall, M.E., Reid-Smith, R.J., Deckert, A.E., Checkley, S.L., McEwen, S.A., 2009. Associations between reported on-farm antimicrobial use practices and observed

antimicrobial resistance in generic fecal Escherichia coli isolated from Alberta finishing swine farms. Preventive Veterinary Medicine 88, 185e192. Varga, C., Rajic, A., McFall, M.E., Reid-Smith, R.J., Deckert, A.E., Pearl, D.L., Avery, B.P., Checkley, S.L., McEwen, S.A., 2008b. Comparison of antimicrobial resistance in generic Escherichia coli and Salmonella spp. cultured from identical fecal samples in finishing swine. Canadian Journal of Veterinary Research-Revue Canadienne De Recherche Veterinaire 72, 181e187. Virlogeux-Payant, I., Baucheron, S., Pelet, J., Trotereau, J., Bottreau, E., Velge, P., Cloeckaert, A., 2008. TolC, but not AcrB, is involved in the invasiveness of multidrug-resistant Salmonella enterica serovar Typhimurium by increasing type III secretion system-1 expression. International Journal of Medical Microbiology 298, 561e569. Walczak, J., Xu, S., 2011. Manure as a source of antibiotic resistant Escherichia coli and Enterococci: a case study of a Wisconsin, USA family dairy farm. Water Air and Soil Pollution in Review. doi:10.1007/s11270-010-0729-x. Walker, S.L., Redman, J.A., Elimelech, M., 2005. Influence of growth phase on bacterial deposition: interaction mechanisms in packed-bed column and radial stagnation point flow systems. Environmental Science & Technology 39, 6405e6411. Walsh, C., 2003. Antibiotics: Actions, Origins, Resistance. ASM Press, Washington, D.C. Wax, R.G., 2008. Bacterial Resistance to Antimicrobials, second ed.. CRC Press, Boca Raton. World Health Organization, 2000. Annual Report on Infectious Disease: Overcoming Antimicrobial Resistance. Xu, C.X., Lin, X.M., Ren, H.X., Zhang, Y.L., Wang, S.Y., Peng, X.X., 2006. Analysis of outer membrane proteome of Escherichia coli related to resistance to ampicillin and tetracycline. Proteomics 6, 462e473. Yang, W.R., Moore, I.F., Koteva, K.P., Bareich, D.C., Hughes, D.W., Wright, G.D., 2004. TetX is a flavin-dependent monooxygenase conferring resistance to tetracycline antibiotics. Journal of Biological Chemistry 279, 52346e52352. Yao, K.M., Habibian, M.M., Omelia, C.R., 1971. Water and waste water filtration - concepts and applications. Environmental Science & Technology 5, 1105e1112. Zhang, D.F., Jiang, B., Xiang, Z.M., Wang, S.Y., 2008. Functional characterisation of altered outer membrane proteins for tetracycline resistance in Escherichia coli. International Journal of Antimicrobial Agents 32, 315e319. Zhang, X.Q., Bishop, P.L., Kinkle, B.K., 1999. Comparison of extraction methods for quantifying extracellular polymers in biofilms. Water Science and Technology 39, 211e218.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 9 1 e1 7 0 1

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Impact of dissolved organic matter on colloid transport in the vadose zone: Deterministic approximation of transport deposition coefficients from polymeric coating characteristics Vero´nica L. Morales a, Wei Zhang a, Bin Gao b, Leonard W. Lion c, James J. Bisogni, Jr.c, Brendan A. McDonough a, Tammo S. Steenhuis a,* a

Department of Biological and Environmental Engineering, Riley-Robb Hall, Cornell University, Ithaca, NY 14853-5701, USA Department of Agricultural and Biological Engineering, 285 Frazier Rogers Hall, University of Florida, Gainesville, FL 32611-0570, USA c School of Civil and Environmental Engineering, Hollister Hall, Cornell University, Ithaca, NY 14853-3501, USA b

article info

abstract

Article history:

Although numerous studies have been conducted to discern colloid transport and stability

Received 30 July 2010

processes, the mechanistic understanding of how dissolved organic matter (DOM) affects

Received in revised form

colloid fate in unsaturated soils (i.e., the vadose zone) remains unclear. This study aims to

25 October 2010

bridge the gap between the physicochemical responses of colloid complexes and porous

Accepted 25 October 2010

media interfaces to solution chemistry, and the effect these changes have on colloid

Available online 31 October 2010

transport and fate. Measurements of adsorbed layer thickness, density, and charge of DOM-colloid complexes and transport experiments with tandem internal process visuali-

Keywords:

zation were conducted for key constituents of DOM, humic (HA) and fulvic acids (FA), at

Humic acid

acidic, neutral and basic pH and two CaCl2 concentrations. Polymeric characteristics reveal

Fulvic acid

that, of the two tested DOM constituents, only HA electrosterically stabilizes colloids. This

Steric stabilization

stabilization is highly dependent on solution pH which controls DOM polymer adsorption

Air-water interface

affinity, and on the presence of Caþ2 which promotes charge neutralization and inter-

Hydrophobic expulsion

particle bridging. Transport experiments indicate that HA improved colloid transport significantly, while FA only marginally affected transport despite having a large effect on particle charge. A transport model with deposition and pore-exclusion parameters fit experimental breakthrough curves well. Trends in deposition coefficients are correlated to the changes in colloid surface potential for bare colloids, but must include adsorbed layer thickness and density for sterically stabilized colloids. Additionally, internal process observations with bright field microscopy reveal that, under optimal conditions for retention, experiments with FA or no DOM promoted colloid retention at solid-water interfaces, while experiments with HA enhanced colloid retention at air-water interfaces, presumably due to partitioning of HA at the air-water interface and/or increased hydrophobic characteristics of HA-colloid complexes. ª 2010 Elsevier Ltd. All rights reserved.

* Corresponding author. Tel.: þ1 607 255 2489; fax: þ1 607 255 4080. E-mail addresses: [email protected] (V.L. Morales), [email protected] (W. Zhang), [email protected] (B. Gao), [email protected] (L.W. Lion), [email protected] (J.J. Bisogni), [email protected] (B.A. McDonough), [email protected] (T.S. Steenhuis). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.10.030

1692

1.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 9 1 e1 7 0 1

Introduction

Dissolved organic matter (DOM) plays a prominent role in many soil processes and is ubiquitous in soils; high concentrations are found in manure or wastewater sludge amended lands. Humic acid (HA) and fulvic acid (FA) are principal constituents of soil, aquatic, sewage sludge, and manure DOM (Schnitzer, 1972; Thurman, 1985). The structural properties of the moieties that make up DOM in soil environments have been explored by several investigators; reporting (as general consensus) that DOM molecules behave as flexible entities that can swell and shrink in response to changes in pH and ionic strength (Avena et al., 1999; Benedetti et al., 1996; Duval et al., 2005; Hosse and Wilkinson, 2001). The amphiphilic character (i.e., presence of hydrophobic and hydrophilic moieties) of DOM has been reported by a number of studies (Guetzloff, 1994; Lenhart and Saiers, 2004; Ma et al., 2007; von Wandruszka, 2000) and used to explain its high surface reactivity and adsorptive fractionation to solid-water and airwater interfaces (Chi and Amy, 2004; Lenhart and Saiers, 2004; Ma et al., 2007). Greater affinity of larger, more hydrophobic DOM components for mineral surfaces and air-water interfaces is of primary importance for contaminant flux through the vadose zone (Lenhart and Saiers, 2004; Meier et al., 1999), as this unsaturated soil region is the critical connection that buffers deep groundwater from surface and shallow contaminants. Physicochemical interactions between DOM and contaminants have received considerable attention in recent years. Numerous investigations have demonstrated that even small amounts of DOM greatly increase the mobility of colloidassociated contaminants (e.g., radionuclide plutonium, americium, thorium, and radium; phosphorus; hydrophobic organic compounds; uranium(IV)/(VI); carbon nanotubes; and lead) (Flury and Qiu, 2008; Granger et al., 2007; Jaisi et al., 2008; Marley et al., 1993; Mibus et al., 2007; Sen and Khilar, 2006; Tang and Weisbrod, 2009) and colloid-sized pathogens (e.g., Escherichia coli, Cryptosporidium parvum oocysts, Giardia, and bacteriophage PRD1) (Abudalo et al., 2005, 2010; Bradford et al., 2006; Foppen et al., 2008) through hydrologic pathways. Laboratory batch kinetic and isotherm experiments have also been conducted to explore the interactions between DOM and colloidal particles. Results from these experiments indicate that DOM increases the stability of colloid and nanoparticle suspensions in the presence of electrolytes through electrostatic and/or steric stabilization by way of adsorption onto colloid surfaces (Akbour et al., 2002; Chen and Elimelech, 2007; Heidmann et al., 2005; Kretzschmar et al., 1998; Pefferkorn, 2006). Moreover, complexation of surface functional groups with non-indifferent ions in solution is a widely recognized process (Amirbahman and Olson, 1995; Chen and Elimelech, 2007; Chen et al., 2006) that could significantly affect the stability and therefore the transport of colloids suspended in DOM rich solutions. A number of investigations systematically examined the impact of DOM on colloid mobility in saturated porous media in terms of pore water velocity and deposition kinetics (Akbour et al., 2002; Jaisi et al., 2008; Kretzschmar et al., 1997). The effects of mono- vs. divalent cation concentrations (Jaisi

et al., 2008), and ionic strength on attachment efficiency have been evaluated (Franchi and O’Melia, 2003; Kretzschmar and Sticher, 1997), as well as charge reversal by organic matter adsorption (Kretzschmar and Sticher, 1997). Both natural and well characterized porous media have been used in research spanning acidic and neutral pore water pH ranges (Akbour et al., 2002; Franchi and O’Melia, 2003; Jaisi et al., 2008; Kretzschmar et al., 1997; Kretzschmar and Sticher, 1997). However, only limited studies have explored the effect of DOM and pH on colloid transport in unsaturated porous media (Tang and Weisbrod, 2009) and the effect of colloid transport in alkaline DOM rich conditions (Harvey et al., 2010). This range of solution chemistry is of high relevance for soils amended with lime-treated manure as a commonly used pathogen inactivation treatment. Two general consensuses about the presence of air phases are that interfaces with air in unsaturated porous media promote colloid retention, and that organic matter of hydrophobic character preferentially fractionates to the air-water interface. Thus, it is of critical importance to understand the effect that DOM has on colloid transport in unsaturated soils; particularly if adsorbed amphiphilic DOM may provide hydrophobic characteristics to the interfaces and colloid surfaces it adsorbs onto. This study aims to experimentally bridge the gap between the physicochemical changes of DOM-colloid complexes and porous media interfaces and the effect that these systematic changes have on the transport of colloids in unsaturated soils. These objectives will be achieved by: (i) directly measuring the changes in surface charge, adsorbed layer thickness and density of organic matter-colloid complexes under acidic, neutral, and basic solution pH in the presence and absence of CaCl2, (ii) assessing with column experiments the effects that HA and FA have on colloid transport at the solution chemistries listed above, (iii) simultaneous internal observation of the dominant pore scale retention sites for each set of conditions, and (iv) mathematical modeling of the transport behavior of organic matter-colloid complexes to relate deposition coefficients with changes in solution composition.

2.

Materials and methods

2.1.

Preparation of materials

In order to meet the objective of discerning the specific steric characteristics that increase the stability of organic mattercolloid complexes (e.g., thickness of adsorbed organic matter layer and uniform adsorption of DOM onto colloid surfaces), the use of uniform and spherical colloids was essential. As such, calibration grade polystyrene and carboxylated spheres of 24 nm diameter (Bangs Laboratories Inc.; Fishers, IN) were used to measure the adsorbed layer thickness. These spheres were selected because of their exceptional size uniformity that allowed the measurement of changes in size at the nanometer scale. Similar surfactant-free, red-dyed, polystyrene and carboxylated spheres of 2.6 mm diameter (Magsphere, Pasadena, CA) were used for all other measurements. The larger red colloids were chosen because they permitted excellent visualization of individual colloids against the porous medium with Bright Field Microscopy (BFM).

1693

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 9 1 e1 7 0 1

Elliott Soil HA and FA standards purchased from the International Humic Substances Society were used for this study. The individual DOM solutions were prepared by dissolving 200 mg L1 of HA and FA in deionized water, and adjusting the stock solution pH to 7 with NaOH. HA and FA stock solutions were further diluted to create solutions of 20 mg of total dissolved organic carbon L1, as measured by persulfate oxidation with an O-I-Analytical Total Organic Carbon Analyzer model 1010 (College Station, TX). Deionized water (DI) was used as the control for no DOM. Solution pH was adjusted to pre-established experimental values (e.g., 4, 6 and 9) with NaOH and HCl immediately prior to starting the experiments. For simplification, the changes in ionic strength (IS) by addition of acid and base to adjust for pH are considered small (e.g., IS of 102 mM for pH 9 and IS of 101 mM for pH 4) relative to the change from addition of CaCl2, so the solution’s IS is referred to as 0 mM and 1 mM for solutions in the absence and presence of CaCl2, respectively. Batch measurements of surface tension of DOM solutions were significantly similar, with a mean surface tension for all tested solution chemistries of 71.7  0.11 mN m1. Translucent quartz sand (Unimin Corp., Vineland, NJ) of 0.4e0.59 mm diameter was used for the porous medium. Before use, the sand was washed according to the procedure in Morales et al. (2009) in order to remove soluble organic compounds from the surface, dissolve metal oxide coatings, and obtain a constant baseline for optical density measurements during breakthrough measurements.

2.2.

Adsorbed layer characteristics

The amount of HA and FA adsorbed onto the sand, Gs (M M1), and colloid surfaces, Gc (M L2), was measured by solution depletion for each solution composition. Briefly, for Gs, the column was wet packed with 60 g of quartz sand, and 60 mL of solution were recycled through the cell with a peristaltic pump for 24 h at a rate of 0.32 mL min1. Afterward, the equilibrium concentration of the solution was measured by spectrophotometry (l ¼ 350 nm) and the difference between the initial and the equilibrium DOM concentration was used to determine the mass of adsorbed organic matter mass per gram of sand medium. For Gc, two sets of 20 mL of each solution composition were prepared and 1x108 colloids were added to the first. The suspensions and colloid-free solutions were agitated for 24 h, then centrifuged for 2 h at 8000 rpm to settle HA-colloid complexes. This centrifugation step was verified to not significantly alter the concentration of nonadsorbed HA and FA components. The supernatant equilibrium concentration was measured by spectrophotometry, and the difference between the colloid-free and the colloid equilibrated concentration was used to determine the mass of adsorbed organic matter per m2 of colloid surface. The thickness of the organic matter layer adsorbed onto the spherical colloidal particles, d (L), was directly measured (rather than fit with existing soft particle theory that is valid only for systems with symmetric and indifferent electrolytes) as the difference between the hydrodynamic radius (RH) of particles exposed to the DOM solution and that of bare particles (i.e., suspended in DI), holding pH and ionic strength constant. For the measurement of this parameter exclusively,

24 nm microspheres were used as the core particles suspended in 18 different solution compositions of varying DOM type (DI, FA, and HA), at three different pHs (4, 6, and 9), and two different ionic strengths (0 and 1 mM). This down scaling of particle size was necessary to measure significant differences in RH at the nanometer scale for colloids with and without a brush layer of adsorbed organic matter. The RH of the colloids was determined by dynamic light scattering (DLS) at each characteristic solution composition using a BIC 200 SM DLS (Long Island, NY). The density of adsorbed DOM layer, F, was estimated with Equation (1) for colloids of radius a (L), using the DOM density value (rDOM) of 1.45x1018 mg nm3 reported by Relan et al. (1984). F¼

2.3.

h

3Gc a2 3

rDOM ðd þ aÞ a3

i

(1)

Electrophoretic mobility

Electrophoretic measurements (EM) measurements were collected with a Malvern Zetasizer nano (Worcestershire, UK) for colloid suspensions in each solution composition. Eight measurements were collected for each treatment and the mean EM values were used to calculate the zeta (z) -potential of the particles via Smoluchowski’s formula (using a measured viscosity value at 25  C of 0.91  104  0.01  104 Pa s and the dielectric constant of the water medium of 78.54). Zeta potentials were converted to surface potential, jo, in mV according to van Oss (2003):  z jo ¼ z 1 þ expðkZÞ a

(2)

where z is the distance from the particle’s surface to the slipping plane (here taken to be 0.5 nm), and 1/k is the thickness of the diffuse double layer.

2.4.

Column setup

The column apparatus consisted of a transparent vertical acrylic flow cell (inside dimensions: 10-cm length, 2-cm width, 2-cm depth) with inlet and outlet tubing connected to a twochamber micropurge peristaltic pump (Masterflex, Barnat, Barrington, IL) to induce perfectly matched steady-state flow. The column apparatus is the same as that used by Morales et al. (2009) and is built with a thin acrylic front side (1.5mm thickness) to allow tandem collection of visual data of internal processes during column breakthrough experiments. Colloid transport data were collected in duplicate from 18 column experiments consisting of three background combinations of DOM (DI, 20 mg L1 FA, and 20 mg L1 HA) at three solution pHs (4, 6, 9) and two background ionic strengths. The column was wet packed with clean sand using gentle vibration to ensure uniform packing (porosity of 0.41). Then, the saturated column was conditioned by circulating colloid-free DOM background solution through the column for 24 h to allow the sand medium to equilibrate with adsorbable DOM prior to injection of the colloid pulse. Subsequently, the column was allowed to become unsaturated by increasing the outflow rate above the inflow rate until a volumetric moisture

1694

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 9 1 e1 7 0 1

content, q, of 0.24  0.03 (equivalent to 0.61  0.08 water saturation) was attained (measured gravimetrically). Upon reaching the target moisture content, the outflow and inflow rates were restored to identical and uninterrupted rates of 0.30 mL min1 to ensure steady state flow. The colloid suspension was prepared in a solution of the same composition as that of the background. It is important to note that the optical density of dissolved HA and FA was accounted for when establishing the standard curves for colloid concentration measurements with spectrophotometry. To inject a 10 mL colloid pulse the column influent was switched to the colloid suspension, after which the inflow was switched back to the colloid-free background solution to flush un-retained colloids out of the medium. The BTCs were constructed from effluent colloid concentrations assayed by spectrophotometry (l ¼ 350 nm) with a linear correlation for absorbance and colloid concentration over the concentration range tested. For mass balance comparisons, the fraction of colloids recovered (MR) from the pulse of injected colloids was calculated as: P MR ¼

JDti ðCi þ Ciþ1 Þ 2tc JCo

Transport model

A deterministic model including terms for first-order attachment kinetics and colloid-excluded volume (i.e., colloid accessible pore space responsible for early elution of colloids with respect to the conservative tracer) in homogeneous soil was utilized to determine each solution composition’s effect on colloid deposition and pore-exclusion with the following governing equation:   vðqc CÞ v vC ¼ qc D  JC  qc kd C vt vx vx

2.6.

Internal process visualization

A horizontally mounted bright field microscope was used to visualize colloid transport inside the column from a lateral view of the cell in a fashion similar to that used by Morales et al. (2009). Briefly, observation of colloid transport and retention at the pore scale was achieved by selecting random pores at heights of 9 and 5 cm from the bottom of the column, and capturing still images at 250 magnification (resolution 0.8 mm pixel1). Wherever the pores remained in clear view for the duration of the experiment video recordings were also collected to observe the progressive retention of colloids. Supplementary video related to this article can be found at doi:10.1016/j.watres.2010.10.030.

3.

Results

3.1.

Adsorbed layer characteristics

(3)

where J is the steady state flux (L3 T1), Dti is the time difference between collected effluent samples (T), Ci is the aqueous colloid concentration of the ith effluent sample (M L3), tc is the duration of the colloid or tracer pulse (T), and Co is the initial concentration of the injected pulse (M L3). In addition, colloid BTCs were compared to an independently run nonsorbing bromide tracer (Br), analyzed with a Dionex Ion Chromatography System-2000 (Sunnyvale, California) to determine the pore-water velocity, v (L T1) and dispersion coefficient, D (L2 T1).

2.5.

complexes and polymer characteristics (as discussed in section 3.4).

(4)

where kd is the colloid deposition coefficient (T1), qc is the colloid accessible volumetric moisture content (L3 L3) (obtained from the product of E and q), E is the pore-exclusion factor (unit less), C is the concentration of the colloids (M L3), x is the distance (L), and J is the specific water flux (L3 T1). The authors acknowledge the complexities of the system and the need to construct a more detailed model to discern the contribution from specific retention mechanisms involved in DOM-colloid complexes transported through unsaturated porous media. Nonetheless, the kd term accounts for the loss of particles as the sum of all participating sinks, acceptably captures the BTC shape, and allows relationships to be interpreted between deposition of sterically stabilized colloid

A summary of the adsorbed layer characteristics determined by solution depletion, DLS, and EM is listed in Table 1 for all the conditions tested in this study. Data on the amount of DOM adsorbed onto surfaces, Gs and Gc, suggest that HA has a superior affinity for solid surface sorption than FA because of its higher molecular weight, which is consistent with the literature (Ko et al., 2005). The amount of HA adsorbed onto colloid surfaces was greatest at pH of 4 and decreased with increasing pH (see Gc values in Table 1), which is in agreement with previous reports (Jada et al., 2006; Lenhart and Saiers, 2004). Moreover, the presence of CaCl2 drastically increased Gc and Gs from 25 to 107%; particularly for adsorption of FA, which was undetectable in the absence of CaCl2. The adsorption of organic matter onto sand generally followed the same trend as that for colloids (see Gs values in Table 1), with variations here attributed to the heterogeneity of the medium. The DLS data indicate that a ‘stabilizing shell’ (Fritz et al., 2002) is developed on the colloids when they are exposed to HA. The adsorbed layer thickness, d, for HA-coated colloids d was intermediate at pH 4 (32.0e24.7 nm), greatest at pH 6 (52.8e55.1 nm), and lowest at pH 9 (1.5e11.9 nm); while d for FA-coated colloids was undetectable for all conditions except for pH 6 (see Table 1). Because the diffusivity of the colloids increased greatly with the addition of salts for suspensions of FA and DI, d could not be quantified accurately in the presence of CaCl2. Consequently, already small d values for these samples (0 < d < 2.9 nm) were assumed to not be affected significantly by electrolyte addition and are reported as the same value as those measured at IS of 0 mM. Generally, d increased with IS for experiments with equal pH levels; particularly for HA suspensions. Moreover, the thickness of d for HA at pH 4 when IS is 0 mM (32.0 nm) and at pH 9 when IS is 1 mM (11.9 nm) were almost three fold different even though the mass of adsorbed HA is nearly the same (34.6  0.7 mg m2 and 33.1  0.5 mg m2, respectively in Table 1). These results suggest that the HA adsorption conformation for pH 9 is in

1695

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 9 1 e1 7 0 1

Table 1 e Characteristics of adsorbed organic layer and transport parameters. Data is organized by pH level, dissolved organic matter type (DOM), and ionic strength by addition of CaCl2 electrolyte (IS). Characteristics of adsorbed layer include: mass of adsorbed organic matter onto colloid (Gc) and sand (Gs) surfaces, thickness of adsorbed organic matter layer (d ), density of adsorbed organic matter layer (F), electrophoretic mobility of colloid suspension (EM), and colloid surface potential (jo). Column transport parameters include the fraction of colloids recovered (MR), pore-exclusion factor (E ), and deposition rate (kd). DOM/IS (/mM) pH 4

pH 6

pH 9

HA/0 HA/1 FA/0 FA/1 DI/0 DI/1 HA/0 HA/1 FA/0 FA/1 DI/0 DI/1 HA/0 HA/1 FA/0 FA/1 DI/0 DI/1

Gca (mg m2) 34.6 43.3 0.0 31.4 0.0 0.0 27.2 38.4 0.0 13.0 0.0 0.0 16.0 33.1 0.0 35.7 0.0 0.0

 0.7  0.2  0.4

1 1 2

 0.2  0.5 1

Gs (mg g1)

db (nm)

F ()

EMa (mm cm Vs1)

joa (mV)

0.0024 0.0031 0.0 0.0031 0.0 0.0 0.0008 0.0006 0.0 0.0004 0.0 0.0 0.0022 0.0025 0.0 0.004 0.0 0.0

32.0 24.7 0.0 0.0 0.0 0.0 52.8 55.1 2.90 2.90 0.0 0.0 1.50 11.9 0.0 0.0 0.0 0.0

0.73 1.2 0.0 0.0 0.0 0.0 0.34 0.46 0.0 2.5 0.0 0.0 7.2 1.9 0.0 0.0 0.0 0.0

3.43  0.20 1.60  0.24 2.04  0.43 1.49  0.19 0.229  0.083 0.273  0.080 3.58  0.11 1.62  0.14 3.33  0.14 1.57  0.065 1.66  0.076 0.427  0.013 3.95  0.19 1.78  0.16 2.47  0.66 1.73  0.029 3.03  0.031 0.856  0.054

43.9 20.3 26.0 19.0 2.92 3.48 45.7 20.7 42.5 20.0 21.2 5.59 50.5 22.7 31.5 22.1 38.8 11.0

MRa () 60.7 41.4 45.7 5.20 20.2 3.00 85.7 24.9 71.1 5.00 75.6 2.60 48.1 20.6 70.2 4.50 78.9 4.00

 0.35  0.29  1.8  1.3  2.8  1.1  0.34  1.3  3.1  0.28  1.3  2.0  3.2  2.1  1.8  1.1  4.2  1.0

E (cm3 cm3)

kd (hr1)

0.87 0.79 0.91 0.86 0.71 0.84 0.84 0.71 0.82 0.77 0.81 0.90 0.82 0.79 0.82 0.88 0.85 1.00

0.26 0.43 0.40 1.5 0.77 1.8 0.11 0.70 0.19 1.5 0.15 1.9 0.40 0.77 0.22 1.6 0.14 1.6

a Measurements collected using 2.6 mm colloids. b Measurement collected using 24 nm colloids.

extended and tight multilayers, while that at pH 4 is in loose folds. Likewise, the d for HA at pH 6 for both IS levels (52.8 nm and 55.1 nm for IS at 0 mM and 1 mM, respectively) was very similar although the mass adsorbed experienced a 40% increase in the presence of CaCl2 (Gc ¼ 27.2 mg m2 and Gc ¼ 38.4 mg m2 for IS at 0 mM and 1 mM, respectively) (see Table 1). This result indicates that the conformation of HA strands becomes more compact and therefore the adsorbed layer more dense with increasing IS. Consistently, the density of adsorbed DOM layer, F, increased with IS, but also varied with pH such that the lightest layers formed in HA solutions at pH 6 (F ¼ 0.34e0.46) and the most dense layers formed in HA at pH 9 (F ¼ 1.9e7.2) (see Table 1). Because F is an encompassing parameter that accounts for the presence of a steric layer on the surface of colloids, only those colloid complexes with F values greater than 0 were classified as being sterically stabilized. Measurements of the particle charge are listed as raw electrophoretic mobility, EM, values along with their corresponding surface potential, jo, value in Table 1. In general, our data indicate that for the conditions tested, colloids suspended in HA solutions have the most negative jo, followed by those suspended in FA, and lastly by the bare suspensions in DI. As expected, addition of CaCl2 screened the charge for all treatments, reflected in less negative EM and jo values at higher IS treatments.

3.2.

Colloid retention sites

Four distinct colloid retention sites were identified by visual observation during the column experiments (Figure S1): (a) straining at grainegrain contact regions, (b) attachment to the

solid-water interface (SWI), (c) straining at the air-water meniscusesolid interface (AWmSI), and (d) the accumulation/ attachment of a continuous layer of enmeshed colloids along the air-water interface (AWI) (i.e., bridge flocculation at the AWI). Here, the distinction between straining and attachment processes is based on the definitions provided by Bradford and Torkzaban (2008). Although the occurrence of colloid retention at the AWmSI was frequently observed in many of the treatments tested here, it appeared to be unrelated to any particular set of solution conditions; especially to the type of DOM in solution. Retention at the SWI and grainegrain straining (i.e., multiple SWIs) (Figure S1a and b) were most common for experiments when the solution chemistry created optimal conditions for retention (i.e., higher ionic strength and/or low pH) in experiments of DI and FA. In contrast, retention at the AWI by bridge flocculation (Figure S1d) under similar conditions was the dominant retention site for experiments containing HA as is evident in Video 1(http://soilandwater.bee.cornell.edu/colloids.html) at solution pH of 4 and IS of 1 mM. Here, red colloids exhibit ripening-like behavior (defined as an increasing rate of colloid attachment with time due to colloidecolloid interactions on the collector surface (Bradford and Torkzaban, 2008)) at the AWI and become immobilized at close proximity to other previously retained colloids. It is evident that retained colloids are enmeshed in organic matter partitioned at the AWI, adsorbed on the surface of individual colloids, or both, given that the colloids in the aggregates are not in direct contact with each other but move as a floc unit with the pulsing flow. This behavior was consistently observed for experiments in HA, with denser flocs formed as pH and ionic strength levels increased (Fig. 1).

1696

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 9 1 e1 7 0 1

Fig. 1 e Colloid retention in bridge flocs when suspended in solutions of dissolved Humic Acid at different ionic strengths (IS) and pH levels. Left column 0 mM IS, right column 1 mM IS by CaCl2 addition. Top row pH 4, middle row pH 6, bottom row pH 9. Scale bar length is 250 mm.

3.3.

Colloid transport in unsaturated sand

The various solution treatments used to conduct the column studies are presented and compared in terms of DOM type, changes in solution pH, and differences in CaCl2 concentration.

3.3.1.

Effect of HA and FA

The type of DOM in the solution of the column studies significantly affected the amount of colloids retained in the unsaturated quartz sand medium, as is evident in the HA (diamonds) vs. FA (triangles) vs. DI (circles) breakthrough curves in Fig. 2aec. In all but one solution (pH 9 and IS of

0 mM), experiments with HA displayed significantly greater elution (i.e., colloid mass recovery) (see MR values in Table 1) from the column than those treated with FA or DI with equal pH and ionic strength conditions. Fig. 2aec illustrate that experiments with FA behaved very similar to those treated with DI in all but one case (pH 4 and IS of 0 mM, open triangles in Fig. 2a), in which colloid mobility was enhanced, although not as much in the presence of HA.

3.3.2.

Effect of pH

The transport of colloids in the sandy medium was considerably different at solution pH values of 4, 6 and 9 (compare like symbols from Fig. 2aec). For treatments of both DI and FA

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 9 1 e1 7 0 1

1697

retention at pH 9 (48%) (see Table 1). When the solution IS was raised to 1 mM, increases in pH resulted in a decrease in MR from 41 to 25 to 21% for pH changes from 4 to 6 to 9, respectively. The reduced mobility of HA-colloid complexes at high pH can be explained by the lack of steric hindrance evident in the thinnest adsorbed layer (smallest d values) and most densely packed adsorbed organic matter (largest F values from Table 1).

3.3.3.

Effect of CaCl2 concentration

Increases in ionic strength through addition of CaCl2 to the column study solution produced the expected reduction in colloid breakthrough (open vs. solid symbols in Fig. 2aec). For example, the MR of DI at lower IS ranged from 20 to 79%, but was reduced to less than 4% in the presence of CaCl2 at IS of 1 mM (see MR values of DI in Table 1). The effect of CaCl2 addition was similar in magnitude for experiments with FA, where the MR at lower IS ranged from 46 to 71%, and was reduced to less than 5% in the presence of CaCl2 at IS of 1 mM (see MR values for FA in Table 1). While the addition of CaCl2 to the system drastically reduced the mobility of colloids through the porous medium for all DOM types, the presence of dissolved HA resisted the destabilizing effect of CaCl2, and permitted colloids to be eluted from the column in otherwise optimal conditions for retention. The addition of HA increased the amount of recovered colloids 5e13 fold over that observed in the same conditions but without the DOM addition (i.e., DI) (see MR values in Table 1).

3.4.

Fig. 2 e Breakthrough curves of column experiments and respective model fits for solutions containing dissolved HA, FA, and DI under ionic strength levels of 0 mM and 1 mM at: a) pH 4, b) pH 6, and c) pH 9.

at lower IS, colloid transport varied directly with pH as the MR of colloids increased from 46e20% at pH 4, 71e76% at pH 6, and 70e79% at pH 9 (MR values for DI and FA in Table 1). For experiments where IS was raised to 1 mM, changes in colloid elution with solution pH for DI and FA treatments were insignificant, as the fraction of colloids recovered was low (solid circular and triangular symbols in Fig. 2aec), with MR values ranging from 3 to 5% (see Table 1). Conversely, experiments with HA exhibited distinct and significant changes in colloid mobility for the three pH values tested. For HA experiments conducted at lower IS, MR was 61% at pH 4, reached a maximum at pH 6 (86%), and experienced greatest

Mathematical model

The transport model was run in inverse mode to quantify the influence of solution composition on colloid transport and retention. Values for the colloid pore-exclusion, E, and deposition rate, kd, were estimated from the BTC data using a leastsquares algorithm and the best fit values corresponding to each solution composition presented in Table 1. Although the authors recognize that retention is likely due to more than one mechanism, discerning quantitatively the various components of retention from the breakthrough data is not feasible. We therefore approach retention with the nonspecific parameter, kd, and shed light to the retention mechanism it represents with visual data. D and v were determined from separate conservative tracer experiments where the terms for E and kd equaled 0. Model fits are presented as solid and dashed lines in Fig. 2 with R2 values > 0.9 for experiments where MR was above 5%. This good fit shows that the simple transport model employed is capable of describing the conditions here tested. The average fitted value E was 0.83  0.07 (see Table 1), and accounted for the residence time reduction of 17% of colloids from that of the Br pulse. The values for kd varied inversely with MR. kd increased with IS, decreased with increasing pH for experiments in FA and DI, and was generally much lower for experiments in HA than for other DOM solutions under the same pH and IS conditions. A positive linear correlation was found between kd for nonsterically stabilized suspensions (i.e., suspensions with measured F ¼ 0) and their respective jo (see Fig. 3a) with a Pearson’s correlation coefficient and significance level (from two-tail test) of r ¼ 0.7 and P ¼ 0.01, respectively. As shown in

1698

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 9 1 e1 7 0 1

Fig. 3 e Correlation between deposition rate, kd, of colloids traveling through an unsaturated porous medium and the particle’s surface characteristics. a) Surface characteristic for non-sterically stabilized colloids is surface potential, jo. Data included are for experiments in FA (triangles) and DI (circles) at ionic strengths of 0 mM (open symbols) and 1 mM (solid symbols). b) Surface characteristics for sterically stabilized colloids are represented by a value that includes ratio of adsorbed layer thickness to colloid radius, d/a, density of polymeric layer, F, and surface potential, jo. Data included are for experiments in HA (diamonds) and FA (triangles) at ionic strengths of 0 mM (open symbols) and 1 mM (solid symbols).

Fig. 3b, sterically stabilized suspensions (i.e., suspensions with measured F > 0) have a Pearson’s correlation coefficient and significance level of r ¼ 0.9 and P < 0.01 with the product of the measured polymeric characteristics of sterically stabilized suspensions d/a, F, and jo.

4.

Discussion

Although in general, the partitioning of DOM-colloid complexes at the AWI generally increases with the presence of HA, the overall breakthrough is still greater when HA is present than when it is not. The physicochemical characteristics of the colloids analyzed in this research indicate that, although the charge of colloids suspended in DOM was more negative than those suspended in DI (i.e., absence of DOM)

(see EM and jo values in Table 1), only HA had the capacity to electrosterically stabilize colloids. This was a result of the formation of a highly charged brush layer (of measured polymeric characteristics d, Gc, and F > 0) on the colloid’s surface that promoted colloid mobility in solutions with higher ionic strength. FA adsorption affinity onto surfaces was in most cases minor (see Gc values for FA in Table 1) and thus did not significantly improve the mobility of colloids despite its effect on colloid charge (see changes in EM and jo for FA in Table 1). As has been recognized previously, the single effect of increasing a particle’s negative charge is not sufficient to justify the increased colloid stability observed in certain treatments (Elimelech et al., 2000). In this study disparate percentages of mass recovered colloids were observed for solution compositions that yielded comparable colloid surface charges. Hence, an additional physicochemical characteristic (to electrostatic charge) responsible for improving the suspension’s stability was suspected to play a prominent role in enhancing the transport of DOM-colloid complexes. Various studies have suggested that the development of steric surface structures by adsorbed DOM may be implicated in the unsubstantiated repulsive forces of colloid suspensions, as these have been observed to remain stable even under conditions of high ionic strength (Chen and Elimelech, 2007, 2008; Franchi and O’Melia, 2003; Phenrat et al., 2010). The adsorbed layer characteristics data demonstrate that the magnitude and range of steric repulsion depend on three factors: (i) density of adsorbed DOM layer, F, (ii) the extension of the adsorbed layer, d, and (iii) the particle charge here presented as surface potential, jo. Evidently, the eletrosterically stabilized particles with most negative surface potential, large adsorbed mass, and thickest brush layer experience the best transport enhancement, and the presence of all three polymeric characteristics is required to provide a suspension with steric stability. When either d is absent or F is very large, the colloids behave like hard colloidal particles and are sensitive to aggregation with changes in solution chemistry. These results reveal that structural hindrance, in addition to electrostatic repulsion, is the mechanisms by which HA, and in one extreme case FA, increases the stability of colloids in the bulk fluid. Clearly, the development of the organic matter brush layer grants soft particle functionalities to the otherwise chemically sensitive suspension, which, to the authors’ knowledge, has only been visualized by Chen and Elimelech (2007) and physically characterized with indirect measurements by Phenrat et al. (2010). This information explains the reduced particleeparticle interaction that results in greater mobility of HA-colloid complexes through porous media, and can be assumed to be analogous to the interactions between the particles and the porous medium through which they move. Internal observations of colloid transport demonstrate that for experiments in DI and FA with solution conditions conducive for high retention (e.g., higher ionic strength and/or low pH), colloid retention occurs at sites involving single or multiple SWI interfaces (Figure S1aed) and have high deposition rate values ranging from 0.4 to 2 h1 (see kd for FA and DI at IS of 1 mM and both IS values at pH 4 in Table 1). This relation suggests that retention of hard particles, such as the colloid suspensions in DI and FA, experience retention with

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 9 1 e1 7 0 1

higher mass transfer rates toward the SWI when the medium’s solution composition hinders transport. For experiments in HA with solution conditions that favor attachment, the dominant retention sites were along the AWI in bridge flocs (Fig. 1), which were characterized with lower deposition rates ranging from 0.3 to 0.8 h1 (see kd for HA at ionic strength solutions of 1 mM and both ionic strength solutions at pH 4 in Table 1). This relation suggests that irreversible retention of soft particles, like HA-colloid complexes, at the AWI is slower than the retention of hard particles, like FA-colloid complexes or colloids in the absence of DOM, at the SWI. Moreover, colloids retained at the AWI may be more susceptible for being remobilized with transient flow conditions than the colloids predominantly retained at the SWI. The dominant retention at AWIs suggests that: (i) HA partitioning to the AWI (as has been demonstrated previously (Lenhart and Saiers, 2004; Ma et al., 2007)) creates a rough surface or gelatinous layer that enhances colloid retention; (ii) HAcolloid complexation occurs in the bulk solution, where the amphiphilic character of HA augments the hydrophobic properties of the colloids resulting in hydrophobic expulsion toward the AWI (i.e., strong attractive non-electrostatic interactions); or (iii) both processes are occurring. Although the visual data in Fig. 1 and Video 1 suggest that hydrophobic interactions significantly influence retention mechanisms in HA-colloid systems, these types of interactions do not occur in isolation from other interaction effects, warranting future work on hydrophobic alterations of surfaces by DOM. Changes in pH affect colloid transport for suspensions in DI and FA by altering the level of functional group deprotonation on the surfaces of both colloids and the porous medium, which consequently affects the thickness of their respective electric double layers and resulting electrostatic interactions. For conditions where the system’s IS was raised to 1 mM, it is evident that neutralization of the colloid’s charge (from less negative jo values in Table 1) dominated over any effect that pH had for experiments with DI and FA, as the colloid mass recovery (MR in Table 1) did not exceed 5% at any pH value. Alternatively, experiments conducted with HA demonstrate that pH is a critical factor for determining colloid transport, as the thickness of the polymeric layer and adsorption affinity of HA onto colloid and medium surfaces was dictated by the solution’s pH (see d and Gc values for HA in Table 1). The addition of CaCl2 affected the colloidal system by: (i) neutralizing the charge of deprotonated HA functional groups, (ii) forming multidentate complexes between available functional groups of the HA macromolecules and Ca cations (as demonstrated by the increase in Gc with CaCl2 addition), and (iii) compressing the electric double layer of all charged surfaces (as is evident from jo data in Table 1). Although chemical bonding of HA with Ca2þ promoted greater adsorption of organic matter onto the colloid surfaces, the chemical reaction at the Ca2þ and HA concentration tested weakened the ability of HA to mobilize colloids (observed in decreases in MR with increasing IS in Table 1) by neutralizing charged functional groups of adsorbed HA strands (apparent in the less negative jo with increasing ionic strength in Table 1), and by bridging together HA-colloid complexes forming large flocs that can experience straining in the porous medium (apparent in visual data of internal colloid processes in Fig. 1, Figure S1, and Video 1). The effects of charge

1699

neutralization and EDL compression are well known to stimulate colloid retention by reducing electrostatic energy barriers that otherwise prevent colloid aggregation and colloid attachment to immobile sites within the porous medium (Kretzschmar and Sticher, 1997). The environmental relevance of this research is most apparent in the solution compositions selected to represent natural subsurface environments rich in DOM as well as agricultural settings that practice land application of alkaline-treated manure. This waste management practice is often preferred for its practicality at the farm level to inactivate zoonotic pathogens (Gerba and Smith, 2005) and control odor (Zhu, 2000). The findings of this study advance our current understanding of the governing mechanisms of colloid transport in a broad range of organic rich subsurface environments; particularly those associated with agriculture. Moreover, our current ability to predict the transport and fate of colloid-adsorbed contaminants and pathogenic microorganisms in the natural subsurface is limited by the incomplete understanding of the physicochemical processes responsible for abiotic porous media filtration. Thus, the use of microspheres as colloidal pathogen surrogates makes elucidation of porous-media transport processes possible without the complications of even less understood biotic processes (e.g., biologically induced attachment, inactivation, die off, growth, motility, chemotaxis, bioclogging, etc.).

5.

Conclusion

In summary, the effect of DOM on colloid characteristics can be directly measured through changes in surface potential, adsorbed layer thickness, and mass of adsorbed organic matter. Solution pH and CaCl2 presence strongly affected the polymeric characteristics by varying the adsorption affinity of DOM onto surfaces and protonation of functional groups, and by neutralizing surface charge and chemically bridging DOM strands together, respectively. The presence of all three parameters (surface potential, adsorbed layer thickness, and mass of adsorbed organic matter by way of density of adsorbed DOM) was established to be a requirement for particles to be endowed with electrostatic and structural stability. Failure to possess all three polymeric characteristics resulted in one of three scenarios: (i) weakened electrostatic repulsion if surface potential were neutralized, (ii) lack of steric hindrance if adsorbed layer thickness were smaller than the separation distance where van der Waals forces are in effect, or (iii) polymerecolloid complexes approach hard particles if adsorbed layer densities were too high. Thus, of the two types of DOM examined, it was established that only HA can consistently electrosterically stabilize colloids. HA was determined to enhance colloid mobility better than FA, even under the most chemically conducive conditions for retention. Retention of HA-colloid complexes along the AWI indicates that partitioning of DOM at the AWI may create a rough or gelatinous surface for colloids to become immobilized and/or that the amphiphilic character of adsorbed HA may encourage colloidal hydrophobic expulsion. Mathematical simulations of column transport data indicate that a simple convective dispersive model with a deposition term and colloid pore-exclusion correction is suitable to describe the fate and transport of

1700

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 9 1 e1 7 0 1

organic matter-colloid complexes in the conditions here tested. A positive and significant correlation was found between deposition coefficients of electrosterically stabilized suspensions and the product of three polymeric characteristics: adsorbed layer thickness, density, and surface potential. As such, measurements of these specific polymeric characteristics can effectively be used to improve deterministic predictions of colloid transport in a wide range of DOM rich solution compositions and assess the filtering function that the vadose zone serves to protect groundwater resources.

Acknowledgments This study was financed by the National Science Foundation, Project No. 0635954; the Binational Agricultural Research and Development Fund (BARD), Project No. IS-3962-07; and the Teresa Heinz Foundation for Environmental Research. The authors thank Dr. John F. McCarthy for helpful discussions and Dr. Yuanming Zhang and Dr. Claude Cohen for assistance with light scattering techniques. We also thank Mr. Doug Caveney for constructing the flow cell used for this study.

Appendix. Supplementary Data Supplementary data related to this article can be found online at doi:10.1016/j.watres.2010.10.030.

references

Abudalo, R.A., Bogatsu, Y.G., Ryan, J.N., Harvey, R.W., Metge, D.W., Elimelech, M., 2005. Effect of ferric oxyhydroxide grain coatings on the transport of bacteriophage PRD1 and Cryptosporidium parvum oocysts in saturated porous media. Environmental Science & Technology 39, 6412e6419. Abudalo, R.A., Ryan, J.N., Harvey, R.W., Metge, D.W., Landkamer, L., 2010. Influence of organic matter on the transport of Cryptosporidium parvum oocysts in a ferric oxyhydroxide-coated quartz sand saturated porous medium. Water Research 44, 1104e1113. Akbour, R.A., Douch, J., Hamdani, M., Schmitz, P., 2002. Transport of kaolinite colloids through quartz sand: influence of humic acid, Ca2þ, and trace metals. Journal of Colloid and Interface Science 253, 1e8. Amirbahman, A., Olson, T.M., 1995. Deposition kinetics of humic matter-coated hematite in porous-media in the presence of Ca2þ. Colloids and Surfaces A-Physicochemical and Engineering Aspects 99, 1e10. Avena, M.J., Vermeer, A.W.P., Koopal, L.K., 1999. Volume and structure of humic acids studied by viscometry pH and electrolyte concentration effects. Colloids and Surfaces A-Physicochemical and Engineering Aspects 151, 213e224. Benedetti, M.F., vanRiemsdik, W.H., Koopal, L.K., 1996. Humic substances considered as a heterogeneous donnan gel phase. Environmental Science & Technology 30, 1805e1813. Bradford, S.A., Tadassa, Y.E., Pachepsky, Y., 2006. Transport of Giardia and manure suspensions in saturated porous media. Journal of Environmental Quality 35, 749e757. Bradford, S.A., Torkzaban, S., 2008. Colloid transport and retention in unsaturated porous media: a review of interface-,

collector-, and pore-scale processes and models. Vadose Zone Journal 7, 667e681. Chen, K.L., Elimelech, M., 2007. Influence of humic acid on the aggregation kinetics of fullerene (C-60) nanoparticles in monovalent and divalent electrolyte solutions. Journal of Colloid and Interface Science 309, 126e134. Chen, K.L., Elimelech, M., 2008. Interaction of fullerene (C-60) nanoparticles with humic acid and alginate coated silica surfaces: measurements, mechanisms, and environmental implications. Environmental Science & Technology 42, 7607e7614. Chen, K.L., Mylon, S.E., Elimelech, M., 2006. Aggregation kinetics of alginate-coated hematite nanoparticles in monovalent and divalent electrolytes. Environmental Science & Technology 40, 1516e1523. Chi, F.H., Amy, G.L., 2004. Kinetic study on the sorption of dissolved natural organic matter onto different aquifer materials: the effects of hydrophobicity and functional groups. Journal of Colloid and Interface Science 274, 380e391. Duval, J.F.L., Wilkinson, K.J., Van Leeuwen, H.P., Buffle, J., 2005. Humic substances are soft and permeable: Evidence from their electrophoretic mobilities. Environmental Science & Technology 39, 6435e6445. Elimelech, M., Nagai, M., Ko, C.H., Ryan, J.N., 2000. Relative insignificance of mineral grain zeta potential to colloid transport in geochemically heterogeneous porous media. Environmental Science & Technology 34, 2143e2148. Flury, M., Qiu, H.X., 2008. Modeling colloid-facilitated contaminant transport in the vadose zone. Vadose Zone Journal 7, 682e697. Foppen, J.W., Liem, Y., Schijven, J., 2008. Effect of humic acid on the attachment of Escherichia coli in columns of goethitecoated sand. Water Research 42, 211e219. Franchi, A., O’Melia, C.R., 2003. Effects of natural organic matter and solution chemistry on the deposition and reentrainment of colloids in porous media. Environmental Science & Technology 37, 1122e1129. Fritz, G., Schdler, V., Willenbacher, N., Wagner, N.J., 2002. Electrosteric stabilization of colloidal dispersions. Langmuir 18, 6381e6390. Gerba, C.P., Smith, J.E., 2005. Sources of pathogenic microorganisms and their fate during land application of wastes. Journal of Environmental Quality 34, 42e48. Granger, S.J., Bol, R., Butler, P., Haygarth, P.M., Naden, P., Old, G., Owens, P.N., Smith, B.P.G., 2007. Processes affecting transfer of sediment and colloids, with associated phosphorus, from intensively farmed grasslands: tracing sediment and organic matter. Hydrological Processes 21, 417e422. Guetzloff, T.F., 1994. Does Humic Acid Form a Micelle?, vol. 152 31e35. Harvey, R.W., Metge, D.W., Barber, L.B., Aiken, G.R., 2010. Effects of altered groundwater chemistry upon the pH-dependency and magnitude of bacterial attachment during transport within an organically contaminated sandy aquifer. Water Research 44, 1062e1071. Heidmann, I., Christl, I., Kretzschmar, R., 2005. Aggregation kinetics of kaolinite-fulvic acid colloids as affected by the sorption of Cu and Pb. Environmental Science & Technology 39, 807e813. Hosse, M., Wilkinson, K.J., 2001. Determination of electrophoretic mobilities and hydrodynamic radii of three humic substances as a function of pH and ionic strength. Environmental Science & Technology 35, 4301e4306. Jada, A., Akbour, R.A., Douch, J., 2006. Surface charge and adsorption from water onto quartz sand of humic acid. Chemosphere 64, 1287e1295. Jaisi, D.P., Saleh, N.B., Blake, R.E., Elimelech, M., 2008. Transport of single-walled carbon nanotubes in porous media: filtration

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 9 1 e1 7 0 1

mechanisms and reversibility. Environmental Science & Technology 42, 8317e8323. Ko, I., Kim, J.Y., Kim, K.W., 2005. Adsorption properties of soil humic and fulvic acids by hematite. Chemical Speciation and Bioavailability 17, 41e48. Kretzschmar, R., Barmettler, K., Grolimund, D., Yan, Y.D., Borkovec, M., Sticher, H., 1997. Experimental determination of colloid deposition rates and collision efficiencies in natural porous media. Water Resources Research 33, 1129e1137. Kretzschmar, R., Holthoff, H., Sticher, H., 1998. Influence of pH and humic acid on coagulation kinetics of kaolinite: a dynamic light scattering study. Journal of Colloid and Interface Science 202, 95e103. Kretzschmar, R., Sticher, H., 1997. Transport of humic-coated iron oxide colloids in a sandy soil: influence of Ca2þ and trace metals. Environmental Science & Technology 31, 3497e3504. Lenhart, J.J., Saiers, J.E., 2004. Adsorption of natural organic matter to air-water interfaces during transport through unsaturated porous media. Environmental Science & Technology 38, 120e126. Ma, J., Jiang, J., Pang, S., Guo, J., 2007. Adsorptive fractionation of humic acid at air-water interfaces. Environmental Science & Technology 41, 4959e4964. Marley, N.A., Gaffney, J.S., Orlandini, K.A., Cunningham, M.M., 1993. Evidence for radionuclide transport and mobilization in a shallow, sandy aquifer. Environmental Science & Technology 27, 2456e2461. Meier, M., Namjesnik-Dejanovic, K., Maurice, P.A., Chin, Y.P., Aiken, G.R., 1999. Fractionation of aquatic natural organic matter upon sorption to goethite and kaolinite. Chemical Geology 157, 275e284. Mibus, J., Sachs, S., Pfingsten, W., Nebelung, C., Bernhard, G., 2007. Migration of uranium (IV)/(VI)in the presence of humic

1701

acids in quartz sand: a laboratory column study. Journal of Contaminant Hydrology 89, 199e217. Morales, V.L., Gao, B., Steenhuis, T.S., 2009. Grain surfaceroughness effects on colloidal retention in the vadose zone. Vadose Zone Journal 8, 11e20. Pefferkorn, E., 2006. Clay and oxide destabilization induced by mixed alum/macromolecular flocculation aids. Advances in Colloid and Interface Science 120, 33e45. Phenrat, T., Song, J.E., Cisneros, C.M., Schoenfelder, D.P., Tilton, R.D., Lowry, G.V., 2010. Estimating Attachment of Nano- and Submicrometer-particles Coated with Organic macromolecules in porous media: development of an empirical model. Environmental Science & Technology 44, 4531e4538. Relan, P.S., Girdhar, K.K., Khanna, S.S., 1984. Molecularconfiguration of composts humic-acid by viscometric studies. Plant and Soil 81, 203e208. Schnitzer, M., 1972. Humic Substances in the Environment. Dekker, New York. Sen, T.K., Khilar, K.C., 2006. Review on subsurface colloids and colloid-associated contaminant transport in saturated porous media. Advances in Colloid and Interface Science 119, 71e96. Tang, X.Y., Weisbrod, N., 2009. Colloid-facilitated transport of lead in natural discrete fractures. Environmental Pollution 157, 2266e2274. Thurman, E.M., 1985. Organic Geochemistry of Natural Waters. Kluwer Academic, Boston. van Oss, C.J., 2003. Long-range and short-range mechanisms of hydrophobic attraction and hydrophilic repulsion in specific and aspecific interactions. Journal of Molecular Recognition 16, 177e190. von Wandruszka, R., 2000. Humic Acids: Their Detergent Qualities And Potential Uses In Pollution Remediation, vol. 1 10e15. Zhu, J., 2000. A review of microbiology in swine manure odor control. Agriculture Ecosystems & Environment 78, 93e106.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 0 2 e1 7 1 0

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Short-term bacterial community composition dynamics in response to accumulation and breakdown of Microcystis blooms Huabing Li a,b, Peng Xing a, Meijun Chen a, Yuanqi Bian a,b, Qinglong L. Wu a,* a

State Key Laboratory of Lake Science and Environment, Nanjing Institute of Geography & Limnology, Chinese Academy of Sciences, Nanjing 210008, PR China b Graduate School of Chinese Academy of Sciences, Beijing 100049, PR China

article info

abstract

Article history:

Short-term bacterial community composition (BCC) dynamics in response to accumulation

Received 12 May 2010

and breakdown of Microcystis blooms were examined by conducting in situ mesocosm

Received in revised form

experiments with varying levels of Microcystis sp. biomass, ranging from 15 to 3217 mg/L as

27 October 2010

measured by chlorophyll-a concentration in the freshwater water column. The BCC was

Accepted 9 November 2010

assessed by means of terminal restriction fragment length polymorphism (T-RFLP) of 16S

Available online 16 November 2010

ribosomal RNA genes followed by cloning and sequencing of selected samples. The results showed that the composition of both free-living and particle-attached bacterial commu-

Keywords:

nities changed during the accumulation and breakdown phases of a Microcystis bloom, and

Microcystis

differences were also evident with different levels of Microcystis biomass. The relative

Bacterial community composition

abundance of bacteria affiliated with Micrococcineae and Legionellales increased in general

Lake

after amendment with Microcystis. Significant correlation between the relative abundance of Micrococcineae and breakdown of Microcystis biomass was also observed. Canonical correspondence analysis (CCA) showed that the changes in the free-living and particleattached bacterial community were mostly related to the changes in the concentrations of chlorophyll-a, dissolved organic carbon (DOC), dissolved oxygen (DO) and pH, which were mainly induced by the breakdown of Microcystis biomass. Overall, our study revealed the following: i) accumulation of Microcystis blooms and their breakdown have strong impacts on bacterial community composition; ii) there might be saprophytic association between Micrococcineae and decomposition of Microcystis biomass; iii) it is necessary to reveal potential associations between Legionellales organisms and Microcystis blooms in eutrophic freshwater lakes. ª 2010 Elsevier Ltd. All rights reserved.

1.

Introduction

Due to pollution and eutrophication, cyanobacterial blooms, especially Microcystis blooms, are becoming a widespread problem in the aquatic environment (e.g. Lehman, 2007; Paerl et al., 2001). In the post-bloom stage, large quantities of

Microcystis will accumulate and decompose, producing dissolved organic matter (DOC) (Cole et al., 1982), resulting in drops in pH and dissolved oxygen (DO) (Chen et al., 2010), and these physiochemical changes may have strong impacts on other aquatic organisms (e.g. Zhang et al., 2009). Recent studies indicated a strong influence of Microcystis blooms on

* Corresponding author. Tel.: þ86 25 86882107; fax: þ86 25 57714759. E-mail address: [email protected] (Q.L. Wu). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.11.011

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 0 2 e1 7 1 0

eukaryotic microorganisms (e.g. Chen et al., 2010), while little is known about their effects on bacterial community composition (BCC) (Eiler and Bertilsson, 2004). Because heterotrophic bacteria have crucial roles in biogeochemical cycling and energy flux (e.g. Azam et al., 1983), understanding the response of the BCC to accumulation and breakdown of Microcystis blooms is very important for a better understanding of the metabolic processes in eutrophic aquatic ecosystems. Meanwhile, many algae are associated with a diverse bacterial community (e.g. Maruyama et al., 2003; Grossart et al., 2005, 2006) including some potential pathogens like Legionellale pneumophila (Tison et al., 1980), and Vibrio cholerae (Ferdous, 2009). Changes of BCC may also decrease the water quality (Paerl et al., 2003; Masango et al., 2008). Thorough understanding of the BCC’s response to Microcystis blooms may provide information necessary for searching bacterial indicator, and help to develop efficient strategies for protecting the aquatic environment and human health (Paerl et al., 2003). Lake Taihu, located in eastern China (30 55 0 40" e 31 32 0 58" N and 119 520 32" e 120 360 10" E), is a large shallow eutrophic lake. The cyanobacterial blooming, most of which were Microcystis, occurred in general from March to November with the highest biomass of Microcystis observed from Jun to September in the northern part of the lake. During the last three decades, the Microcystis blooming has become a major environmental issue for the lake (Qin et al., 2007). Due to its large surface area, the Microcystis blooms drift from place to place and even form heavy scum in some bays of the lake (Qin et al., 2007). Previous investigations hinted that the seasonal dynamics of BCC might be related to Microcystis blooms (Xing and Kong, 2007; Wu et al., 2007a, b). However, this has never been confirmed directly by experimental data, and whether Microcystis blooms may be associated with some pathogens in freshwater lakes is still unclear. Therefore, to understand the response of BCC to the accumulation and breakdown of a Microcystis bloom, we conducted in situ mesocosm experiments in Lake Taihu using different levels of Microcystis biomass that encompass the natural range of Microcystis biomass in Lake Taihu.

2.

Materials and methods

2.1.

Experimental design

In LakeTaihu, an in situ mesocosm experiment with 12 polyethylene enclosures (50 L each) was carried out between 26 August and 5 September in 2008. The enclosures were enclosed with the upside open to air. All enclosures were placed in Meiliang Bay of Lake Taihu, which was about 200 m away from the shoreline. Each of the 12 enclosures was filled with lake water from Gonghu Bay, which is a part of Lake Taihu and characterized by high water transparency, dominant submersed macrophyte communities (mainly Potamogeton sp.) and very low algae biomass consisting primarily of diatoms, green algae and cyanobacteria (chlorophyll-a concentration was about 15 mg/L). Microcystis sp. was collected with a 64-mm pore net from Meiliang Bay (31 28 0 4.5" N e 120 130 34.2" E) during a Microcystis bloom period one day before the experiment, and the samples were rinsed three times with distilled

1703

water. To reflect the natural chlorophyll-a (Chl a) concentrations found in different parts of Lake Taihu in the summer, there were four treatments: A (control) enclosures were not supplemented with Microcystis, B (low biomass) enclosures were supplemented with 113 mg/L Microcystis, C (moderate biomass) enclosures were supplemented with 303 mg/L Microcystis, and D enclosures (very high biomass) were supplemented with 3217 mg/L Microcystis. Every treatment was performed in triplicate. The measurement of physical and chemical parameters and their dynamics (Table S1) have been demonstrated by Chen et al. (2010).

2.2. Bacterial counts, biomass collection, and DNA extraction Physiochemical data were used to select appropriate dates for the BCC analysis. At day 4, significant changes in the chemical parameters occurred: the concentrations of Chl a, DO and pH decreased drastically while the DOC concentration increased strongly after 4 days of incubation in all enclosures (Table S1, for details see Chen et al., 2010). This suggests that Day 4 represented the breakdown period. In order to monitor the short-term dynamic of BCC in response to accumulation and breakdown of Microcystis blooms and to decrease the complexity for analyzing the BCC, we therefore chose the samples from days 0, 1 and 4 for an in-depth analysis of short-term BCC dynamics. Because the physiochemical parameters among enclosures A, B and C were very similar throughout the experiment, we did not analyze the BCC of enclosure C. Bacterial DNA samples were taken at day 0, 1, 4 in enclosures A, B, and D. Within 2 h after sampling, about 100 ml was filtered first through a 5.0-mm polycarbonate filter membrane (47-mm diameter, Millipore) and subsequently a 0.2-mm filter membrane (47-mm diameter, Millipore) at a pressure of <20 mbar to collect particleattached and free-living bacteria, respectively. The filters (Millipore) were stored at 80  C until nucleic acids were extracted. Total nucleic acids from particle-attached and free-living bacteria on the filters were extracted and purified using proteinase K and sodium dodecyl sulfate concomitant with chloroform extraction and isopropanol precipitation. Total bacterial numbers were determined by epifluorescence microscopy after DAPI (40 ,6-diamidino-2-phenylindole) staining (Porter and Feig, 1980) as described previously (Wu et al., 2006).

2.3. Bacterial community analysis by 16S rRNA-based T-RFLP analysis Bacterial 16S rRNA genes were amplified using the universal eubacterial primers 8f (50 - AGAGTTTGATCCTGGCTCAG -30 ; labeled with 6-carboxyfluorescein on the 50 end) and 926r (50 CCGTCAATTCCTTTGAGTTT-30 ) (Liu et al., 1997). Mixed DNA from triplicate extractions were used as PCR templates. The PCR amplification was performed in an automated thermocycler (PTC 200-cycler, MJ Research) as follows: one cycle at 94  C for 3 min, 30 cycles at 94  C for 30 s, 55  C for 30 s, 72  C for 1 min, and a final extension at 72  C for 10 min. Three to four replicate PCR samples were cleaned and concentrated into a single 50-ml aliquot with an E.Z.N.A.TM Cycle-Pure Kit (Omega, USA). Purified and concentrated T-RFLP products were digested, purified, separated and analyzed sequentially

1704

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 0 2 e1 7 1 0

(Chen et al., 2010). To account for small differences in the running time among samples, we considered fragments from different profiles with less than 1 base pair difference to be the same length. The results were then expressed as the relative area compared to the total area. Peaks of less than 60 bp, longer than 600 bp, or representing less than 1% of the total peak area were discarded.

constructed from different enclosures (Schloss et al., 2009). Regression analysis was used to measure the relationship between Chl a concentration and chemical parameters as well as the relative abundance of dominant T-RFs during the decomposition of Microcystis sp. The T-test was used to assess changes in chemical parameters during the experiments, and one-way ANOVA was used to compare chemical parameters among treatments.

2.4. Cloning, sequencing, phylogenetic analysis and assignment of T-RFs

3. To identify bacteria, three clone libraries were generated with mixed bacterial 16S DNA templates in triplicate extractions retrieved from the A enclosures at Day 0 (free-living sample) and D enclosures at day 4 (free-living and particle-attached sample respectively). Bacterial amplifications for sequence analysis were generated from 16S rRNA with the primers 8f(50 -AGAGTTTGATCCTGGCTCAG-30 ) (Lane, 1991) and 926r (50 CCGTCAATTCCTTTGAGTTT-30 ) (Muyzer et al., 1993). The PCR protocols were the same as described above. The PCR products were purified with a QIAquick gel extraction kit (QIAGEN) and ligated into the pGEM-T Easy Vector (Promega) according to the instructions of the manufacturer. The presence of the 16S rRNA gene in positive colonies was checked by PCR amplification using vector primers (M13F and M13R). Positive clones were randomly selected for sequencing, which was carried out by Invitrogen Company (Shanghai, China). Raw sequence data were processed and checked with the Lasergene software package DNASTAR (Madison, USA). All sequences were screened for potential chimeric sequences using CHIMERA_CHECK from RDP (http://rdp8.cme.msu.edu/docs/chimera_ doc.html). Correct sequences were compared to sequences in public databases by using NCBI BLAST (http://www.ncbi. nlm.nih.gov) for an initial phylogenetic affiliation. Phylogenetic analyses of retrieved 16S rRNA sequences were conducted by a neighbor-joining tree using MEGA 4 (Tamura et al., 2007). To test the phylogenetic assignments based on in silico T-RF analysis, randomly selected clones were analyzed by in vitro T-RF by finding the first Hha I enzymatic digestion site downstream from 8f. The obtained 16S rRNA gene sequences were deposited in GenBank under the accession numbers HM153606eHM153700.

3.1. Impact of Microcystis addition on bacterial abundance At Day 4, significant increase in abundance of heterotrophic bacteria was observed in Enclosure D (Fig. 1) with addition of high Microcystis biomass, while no strong variations were found in Enclosure A and B in which no or less Microcystis biomass had been added.

3.2.

The T-RFLP profiles of free-living bacterial DNA extracted from the no-addition enclosures (A enclosures) showed similar patterns after 0, 1 and 4 d of incubation (Fig. 2). In Microcystisamended enclosures, by contrast, the free-living bacterial TRFLP profiles changed after 1 and 4 d (Fig. 2). Immediately before the start of the incubation (day 0), a total of 24 distinct bacterial T-RFs were identified. These T-RFs all showed a relative abundance of >1%, and those of 65, 204, 205, 208, 369, 569, and 577 base pairs (bp) in length even exhibited a relative abundance of >5%. After 1 d, in the medium-addition (B) enclosures, the relative abundance of most of these dominant T-RFs had changed very little, so that these T-RFs were still among the most abundant ones within the bacterial T-RFLP profile. However, a new T-RF at 85/86 bp appeared and became the most abundant one (relative abundance >10%; Fig. 2), and other characteristic T-RFs of 60, 92, 513, 93, 201, 203, 365 and 565 bp (the relative abundance of the former three were all >4%) were detected for the first time. At day 4, a 202-bp and the 45

106 cells/ml)

Statistical analysis

To reveal the relationships between BBC and environmental parameters, canonical correspondence analysis (CCA) was used because the length of the first DCA (detrended correspondence analysis) axis run on species data was >2. The tested environmental variables were as follows: Chl a, TP, TN, COD, NH4eN, NOx-N, PO4eP and DO. All data were log (xþ1) transformed except for pH. In the case of the T-RFLP results, a binary matrix was constructed by scoring presence (percentage of area) and absence (0) of particular T-RFs. Environmental factors best describing the most influential gradients in community composition were identified by forward selection with 499 unrestricted Monte Carlo permutations. CCA was performed with the software CANOCO 4.5 (ter Braak  and Smilauer, 2002). The LIBSHUFF program was used to statistically assess the differences between clone libraries

T-RFLP analysis of the free-living bacterial community

40

Bacterial abundance (

2.5.

Results

25

35

Day0

Day1

Day4

30

20 15 10 5 0 A

B

D

Fig. 1 e Heterotrophic bacterial abundance in different enclosures at different day.

1705

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 0 2 e1 7 1 0

A

B

D

Day 0 Day 1 Day 4

Day 1 Day 4

Day 1 Day 4

100

90

80

Relative abundance (%)

70

60

50

40

30

20

10

0

Fig. 2 e Relative abundance of free-living bacterial 16S rRNA amplicons recovered from enclosures with different biomass of Microcystis sp. Those T-RFs of less than 1% of the total were grouped together as a single group donated by others <1%.

A

B

D

100

T-R F(bp)

90 80

Relative abundance (%)

565-bp T-RFs dominated the free-living bacterial populations in these enclosures with relative abundances of 12.5% and 13.7%, respectively; also, 3 new T-RFs (213, 509, and 533 bp) were detected for the first time (Fig. 2). In the high-addition (D) enclosures at day 1, the 65-bp T-RF became the most abundant one (relative abundance >9%; Fig. 2), and the other most abundant ones were those of 60, 92, 226, 371, 513, 82 and 369 bp (relative abundances were all >4%), the first five of which were detected for the first time. Furthermore, another 3 new T-RFs (176, 203, and 412 bp) were detected for the first time. At day 4, the structure of the active bacterial community had changed again. The 359-bp T-RF dominated the free-living bacterial population with a relative abundance of 21.3% (Fig. 2). In addition, 9 (287, 291, 298, 302, 415, 421, 444, 533, and 587 bp) out of the 12 distinct bacterial T-RFs with a relative abundance of >1% were detected for the first time.

70 60 50 40 30 20 10

3.3. T-RFLP analysis of the particle-attached bacterial community Unlike those of the free-living bacteria, the T-RFLP profiles of the particle-attached bacteria in the enclosures all had changed after 1 and 4 d of incubation (Fig. 3). Immediately before the start of the incubation (day 0), a total of 14 distinct

0 D ay 0 Day 1 D ay 4

D ay 1 Day 4

Day 1 Day 4

Fig. 3 e Relative abundance of particle-attached bacterial 16S rRNA amplicons recovered from enclosures with different biomass Microcystis sp. Those T-RFs of less than 1% of the total were grouped together as a single group donated by others <1%.

1706

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 0 2 e1 7 1 0

bacterial T-RFs were identified. These T-RFs all showed a relative abundance of >1%, and those of 65, 80, 295 and 314 bp even exhibited a relative abundance of >4%. The latter two dominated bacterial populations with relative abundances of about 43.8% and 16.3%, respectively. At day 1 in the no-addition enclosures, 237-bp and 369-bp T-RFs dominated the bacterial populations with relative abundances of 25.7% and 36.7%, respectively. The latter was detected for the first time, and most of the T-RFs showing a relative abundance of >1% were newly detected (11 out of 14) (Fig. 3). At day 4, the structure of the bacterial community had changed yet again: the 85/86 bp and 134/135 bp T-RFs dominated the bacterial populations with relative abundances of 14.2% and 12.5%, respectively. Meanwhile 4 new T-RFs (111,138,151 and 168 bp) were detected for the first time (Fig. 3). In the medium-addition enclosures, a total of 19 distinct bacterial T-RFs with relative abundances of >1% were identified, of which only 5 had been previously detected on day 1. Furthermore, the bacterial populations were dominated by the newly-detected T-RFs of 85/86 and 92 bp (with relative abundances of 17.1% and 10.3%, respectively). At day 4, the relative abundance of the 134/135 bp T-RF increased from 6.5% to 13.8%, and it dominated the bacterial populations; but the 85/86 bp T-RF decreased to 10.3%. In addition, 7 new T-RFs (92,

111, 129, 138, 163, 194, and 222 bp) were detected for the first time. In the high-addition enclosures at day 1, most of the T-RFs identified also existed in the medium-addition enclosures at the same time. However, two new T-RFs of 60 and 339 bp appeared and became the most abundant (with relative abundances of 7.0% and 9.6%, respectively) in the high-addition enclosures, and other characteristic T-RFs of 94, 106, 208, 213, 325, 374, 511, 513 and 515 bp were detected for the first time. At day 4, the bacterial populations of these enclosures were dominated by T-RFs of 60, 222, 339 and 513 bp (with relative abundances of 12.8%, 13.5%, 10.7% and 9.8%, respectively, and the second one was newly detected). Another two new T-RFs (201 and 204 bp) were detected for the first time (Fig. 3).

3.4.

Phylogenetic analysis and assignment of T-RFs

To identify bacteria, three clone libraries were generated from bacterial 16S DNA templates retrieved from the A enclosures (WA0.2-0d [n ¼ 52]) and D enclosures (WD0.2-4d [n ¼ 33]; WD54d [n ¼ 50]). The phylogenetic affiliations of the various clone sequences were determined using the neighbor-joining tree method (Fig. S1) in the program MEGA 4. In general, only about half of the detected T-RFs could be assigned to defined

Table 1 e Phylogenetic affiliations of free-living bacterial 16S rRNA sequences retrieved in clone libraries generated from the samples. Characteristic T-RFs for different clone groups are given; T-RFs with relative abundance of more than 4% are indicated in bold; T-RFs detected in more than one phylogenetic group are marked with an asterisk (*); T-RFs characteristic for one time point and T-RFs with enhanced increase in their relative abundance are underlined, respectively. Phylogenetic group

T-RFs of the free-living bacteria (bp) W0.2-0d

Actinobacteria Actinomycetales Micrococcineae Bacteroidetes Flavobacteriales Sphingobacteriales Deinococcales Alphaproteobacteria Rhizobiales Rhodobacterales Sphingomonadales Unclassified Betaproteobacteria Burkholderiales Methylophilales Neisseriales Deltaproteobacteria Desulfovibrionales Unclassified Gammaproteobacteria Legionellales Unclassified Gemmatimonadales Acidobacteriales Verrucomicrobiales Planctomycetales Unidentified

WA0.2-1d

WA0.2-4d

WB0.2-1d

WB0.2-4d

65*,359 363*,369

65*,363*,369

65*,363* 365*,369

65,363*,365*,369

65,363*,369

65*,369

94* 94*, 96 80*

94* 94*,96 80*

219 96

96

226

515

60*, 515

515, 511* 511*

60*, 513 533

60*, 513,515 512

80*

80*

92

82,92

65*, 203,363*

65*, 203

155, 570

570

65*,204, 205*, 363* 155, 570

65*, 203, 205*,363* 155, 570

WD0.2-4d 359

80* 60*, 513,515

92

80* 92

65*,365, 203,363* 570

65*,203,205*, 365, 363* 570

93 374 85/86

WD0.2-1d

208* 208*

208* 208*

85/86 572 208* 208*

577 101,148,249, 344,371,541, 569,581,587

101,148,202, 344,569

201,565, 569,

93 374

93

85/86

85/86*

208* 208*

208* 208* 215

208* 208*

101,344,371, 565,569,201

202,344,509, 565, 569

176,249,344, 371,569

533

287,291,298,302, 344,415,421, 443,448,587

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 0 2 e1 7 1 0

phylogenetic groups, and many T-RFs were detected in more than one phylogenetic group (Table 1). Phylogenetic analysis of the cloned sequences revealed that both the free-living and particle-attached bacterial populations involved in breakdown process of the added Microcystis were highly diverse. The 16S rRNA sequences of free-living bacteria in the noaddition enclosures were dominated by Alphaproteobacteria (Rhizobiales and Rhodobacterales), Betaproteobacteria (Burkholderiales), Actinobacteria (Micrococcineae) and Methylophilales during the experiment (Table 1). In the medium-addition enclosures, at day 1 the bacterial populations were dominated by Legionellales (T-RF: 85/86 bp; Fig. 2; Table 1) facultative anaerobic bacteria closely related to L. birminghamiensis (91% similarity), and the 8 new T-RFs were affiliated with the Rhizobiales and Desulfovibrionales. At day 4, Micrococcineae and Betaproteobacteria (Neisseriales and another unclassified group) dominated the free-living bacterial populations in the medium-addition enclosures. Furthermore, the T-RFs of 213 bp and 533 bp that were detected for the first time were assigned to the Verrucomicrobiales and Rhodobacterales, respectively, and the 509-bp T-RF was not represented by any of the clone sequences and therefore could not be assigned to any phylogenetic group. In the high-addition enclosures, at day 1 Micrococcineae and Rhizobiales were the most abundant bacterial populations. The number of T-RFs assigned to the former group decreased from 4 to 2, and the new T-RFs were affiliated with the Rhodobacterales, Sphingobacteriales, Burkholderiales, Acidobacteriales and Gemmatimonadales. At day 4, none of the previously observed T-RFs related to Actinobacteria were observed, except for one of 359 bp that dominated the free-living bacterial populations (21.3%). Furthermore, the T-RF of 533 bp that was detected for the first time at day 4 was assigned to Rhodobacterales; while the other eight new T-RFs were not represented by any of the clone sequences and therefore could not be assigned to any phylogenetic group. In agreement with the T-RFLP results, the particleattached bacterial populations in the enclosures all had changed after 1 and 4 d of incubation (Table S3). At day 0, 16S rRNA sequences of particle-attached were dominated by Micrococcineae, Deinococcales, Sphingomonadales and Burkholderiales. At day 1, the bacterial community of the noaddition enclosure was dominated by a newly identified TRF of 369 bp that was related to Micrococcineae (36.7%). Furthermore, another two new T-RFs (96 and 226 bp) were affiliated with Sphingomonadales, the T-RF of 85/86 bp was assigned to Legionellales and the T-RF of 565 bp could not be assigned to any phylogenetic group. At day 4, Legionellales became the most abundant bacterial population (14.2%). In the medium-addition enclosures at day 1, the T-RFs of 85/86 and 92 bp that were detected for the first time and that dominated this bacterial population (their relative abundances were 13.0% and 9.7%, respectively) were assigned to L. birminghamiensis and unclassified Alphaproteobacteria; other new T-RFs were related to Flavobacteriale, Rhizobiales, Rhodobacterales, Burkholderiales, Neisseriales, Desulfovibrionales and unclassified Gammaproteobacteria. At day 4, Legionellales decreased from 17.1% to 10.3%. The 134/135 bp T-RF, which increased from 6.5% to 13.8%, was not represented by any of the clone sequences and therefore could not be assigned to

1707

any phylogenetic group. Unlike the medium-addition enclosures, the bacterial communities in the high-addition ones were dominated by L. birminghamiensis, Micrococcineae and Rhodobacterales at day 1. At day 4, the particle-attached bacterial populations were dominated by Rhizobiales, Flavobacteriales and Rhodobacterales, which were most closely related to the T-RFs of 60, 222 and 513 bp (relative abundances of 12.8%, 13.5% and 9.8%, respectively, and the second one was newly detected). Another new T-RF (201 bp) was assigned to Burkholderiales and a T-RF of 204 bp could not be assigned to any phylogenetic group (Table 2).

3.5.

Statistical analysis

The CCA model illustrated that the concentrations of Chl a, DO, DOC and pH contributed most to the variance in the freeliving bacterial community (for all canonical axes, p ¼ 0.08, Fig. 4A). The eigenvalues of the first and second axes were 0.92 and 0.45, respectively, indicating that both axes were important. The first two axes explained 54.0% of the observed variation in the composition of the free-living bacterial community, and 76.2% was explained by the full four canonical axes. The first axis showed a high canonical correlation with the concentrations of DO and DOC, while the second axis correlated with the concentrations of Chl a and DOC. Similarly, the variance in the particle-attached bacterial community composition was also attributed mostly to Chl a, DOC, pH and DO (for all canonical axes, p ¼ 0.83, Fig. 4B). The eigenvalues of the first and second axes were 0.55 and 0.40, respectively. The first two axes explained 52.4% of the observed variation in the composition of the free-living bacterial community, and 70.1% was explained by all four canonical axes. The first axis showed a high canonical correlation with the concentrations of Chl a and DOC, while the second axis correlated with DOC and pH. Regression analysis revealed that Chl a concentration was negatively correlated with pH and DO (R2 ¼ 0.42, 0.36, respectively, p < 0.05) and positively correlated with DOC (R2 ¼ 0.45, p < 0.05). A strong positive correlation between the abundance of heterotrophic bacteria and DOC was observed as well (R2 ¼ 0.77, p < 0.01). Lineal regression analysis revealed that there was a negative correlation between Chl a concentration and the relative abundance of Micrococcineae (T-RFs 365 and 565 bp) in enclosure D (R2 ¼ 0.93, p < 0.05). However, this relationship was not very significant in enclosure A (R2 ¼ 0.17, p > 0.05) and enclosure B (R2 ¼ 0.27, p > 0.05).

4.

Discussion

The objective of this study was to characterize the BCC in water columns with differing Microcystis sp. biomass, to document BCC changes following the breakdown of Microcystis and to identify physiochemical parameters in relation to BCC variability. Two culture-independent methods, T-RFLP analysis and cloning/sequencing, were used to analyze the BCC. The changes in the physiochemical parameters in the enclosures at different sampling times (Table S1) clearly indicated the breakdown of Microcystis sp. (Chen et al., 2010).

1708

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 0 2 e1 7 1 0

Table 2 e Phylogenetic affiliations of particle-attached bacterial 16S rRNA sequences retrieved in clone libraries generated from the samples. Characteristic T-RFs for different clone groups are given; T-RFs with relative abundance of more than 4% are indicated in bold; T-RFs detected in more than one phylogenetic group are marked with an asterisk (*); T-RFs characteristic for one time point and T-RFs with enhanced increase in their relative abundance are underlined, respectively. Phylogenetic group

T-RFs of the particle-attached bacteria (bp) W5-0d

Actinobacteria Actinomycetales Bacteroidetes Flavobacteriales Sphingobacteriales Deinococcales Alphaproteobacteria Rhizobiales Rhodobacterales Sphingomonadales Unclassified Betaproteobacteria Burkholderiales Methylophilales Neisseriales Deltaproteobacteria Desulfovibrionales Unclassified Gammaproteobacteria Legionellales Unclassified Gemmatimonadales Verrucomicrobiales Unidentified

65*,97

WA5-1d

WA5-4d

97,369

219

60*

WB5-1d

65*, 97,369

65*, 369

60*, 222 96

60*, 94* 94* 80*

60*, 222

60*, 155*

511*,513, 515

92

339, 511* 80* 92

60*, 513,511*, 515 339,511*

80*

155*

60*

60*, 511*, 513

80

80*

511* 80* 92

65*,151

205* 205*

85/86

74,116,134, 172,179,237, 295,314,320

74, ,90,146, 160,182,190, 237,371, 565

151, 205* 155* 205*

65*

93 374

93

93 374

93

85/86

85/86 572

85/86

103,111,122, 134,138,146, 168,172,182,190

4.1. Change of BCC and bacterial numbers induced by breakdown of Microcystis blooms Our T-RFLP analysis clearly indicate that the composition of both the free-living and particle-attached bacterial communities in Gonghu Bay, which has little Microcystis, changed after the accumulation of a Microcystis bloom, and these changes varied with the biomass of the bloom. LIBSHUFF analysis of clone libraries confirmed these changes after addition of Microcystis biomass in enclosures (Table S3). These shifts were not only detected as the disappearance of bacterial populations and the appearance of new ones but also as the replacement of the most dominant bacterial populations (Fig. 2; Tables 1 and 2). CCA analysis revealed that changes in pH and in the concentrations of DO, DOC and Chl a had potentially important impacts on BCC. This is in accordance with some previous studies (e.g. Mauricea and Leff, 2002; Lindstro¨m et al., 2005; Grossart et al., 2005, 2006; Kolehmainena et al., 2007) underlining these regulating factors of BCC. Regression analysis revealed that the former three parameters were significantly correlated with the concentration of Chl a, suggesting that these chemical parameter changes were mainly induced by Microcystis breakdown (Chen et al., 2010). Furthermore, positive correlation between the abundance of heterotrophic bacteria and DOC suggests that the breakdown of Microcystis biomass promote the growth of heterotrophic bacteria.

WD5-4d

97,65*

80

65* 155*

WD5-1d

97,65* 60* 96 80*

96, 226

WB5-4d

88,90, 103, 134,146,179, 182, 509

103,106,111,122, 129,134,138,146, 160,163,168,172, 179,182,190,194

208 213 90,103,106,116, 129,160, 325

92 204, 205* 339 205*

213 201,172,

It has been suggested that the mucilage of Microcystis contains bacteria which degrade Microcystis cellular material (Maruyama et al., 2003). So the addition of Microcystis biomass into enclosures might have introduced additional bacteria that were not present in the control mesocosm. This addition might not influence the free-living BCC immediately because the bacteria inhabited in the mucilage of Microcystis cells are in general difficult to be dispatched (Wu et al., 2007a). This is confirmed by the fact that there was no significant change of bacterial abundance between the control and enclosures amended with Microcystis (Fig. 1) at the beginning of experiment. The introduction of Microcystis blooms into enclosures may influence the particle-attached BCC. However, the significant shift of particle-attached BCC in enclosures from Day 1 to Day 4 amended with Microcystis blooms suggests that the breakdown of Microcystis blooms did impact the structure of particle-attached BCC. Thus we conclude that the BCC changes in our study were mainly induced by the breakdown of Microcystis blooms.

4.2. Dominance of Actinomycetales during the decomposition of Microcystis blooms Among the shifts in bacterial community composition that we observed, one that should have been paid more attention to is the changes in Actinomycetales in the addition enclosures, most

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 0 2 e1 7 1 0

1709

4.3. Occurrence of abundant Legionellales organisms during accumulation and crash of Microcystis blooms Another important change is the occurrence of high percentage of bacteria affiliated with Legionellales that are mostly related to Legionella birminghamiensis. Legionellales were found in freshwater environments worldwide (Fliermans et al., 1981), but not as abundant in the present study. Some studies found that anthropogenic factors (Ortiz-Roque and Hazen, 1987) such as wastewater discharge favor a diversity of Legionella species (Carvalho et al., 2007). In addition, symbiotic interactions between Legionella and free-living protozoa may be good for the colonization of aquatic habitats by these bacteria (Ortiz-Roque and Hazen, 1987). Tison et al. (1980) even found that over a pH range of 6.9e7.6., L. pneumophila was apparently using cyanobacteria extracellular products as its carbon and energy sources. We were not able to isolate these organisms for further physiological and ecological studies in the current study. Given the fact that many species belong to the Legionellales are potentially pathogenic (Wilkinson et al., 1987; Fields et al., 2002), our investigation suggests that further studies are necessary to reveal the potential associations between Legionellales and Microcystis, and risk assessment of their occurrence to human health, and their application as indicator of water quality in eutrophic freshwater lakes.

Fig. 4 e Canonical correspondence analysis (CCA) biplots showed variable composition of free-living (A) and particle-attached (B) bacteria in relation to the important environmental factors in the enclosures with different biomass of Microcystis sp.

of which were belong to Micrococcineae. In both enclosures amended with Microcystis blooms, the relative abundance of free-living Micrococcineae was higher than that in the control and increased in general from day 1 to day 4 following the decrease of Chl a. Sequences affiliated with Micrococcineae have previously been described in various freshwater habitats (Zwart et al., 2002; Allgaier and Grossart, 2006a, b; Newton et al., 2007; Wu et al., 2007a, b; Holmfeldt et al., 2009). Actinomycetales possess a wide variety of physiological and metabolic properties (Goodfellow and Williams, 1983). The negative correlation between Chl a concentration and the relative abundance of Micrococcineae in our experiment suggests that there might be saprophytic association between the breakdown of Microcystis biomass and Micrococcineae. Yamamoto and his colleagues have isolated 83 Actinomycete strains from a eutrophic lake among which half were found able to lyse cyanobacteria including Microcystis sp. (Yamamoto et al., 1998). Interestingly, some of these isolates are closely affiliated with Micrococcineae that were found in our study. Further isolation of these microbes and their eco-physiological characterizations will help to reveal their exact role in decomposition of Microcystis blooms and nutrient cycling in freshwater lakes.

Acknowledgements This work was supported by Knowledge Innovation Project of Chinese Academy of Sciences (KZCX1-YW-14-1;KZCX2-YWJC302), National Basic Research Program of China (973 program) (No. 2008CB418104), and Jiangsu Provincial Science Foundation (BK2009024). We thank Ray Zhang for the statistical analysis and Feizhou Chen for assistance in the field.

Appendix. Supplementary material Supplementary data related to this article can be found online, at doi:10.1016/j.watres.2010.11.011.

references

Allgaier, M., Grossart, H.P., 2006a. Diversity and seasonal dynamics of actinobacteria populations in four lakes in northeastern Germany. Appl. Environ. Microbiol. 72, 3489e3497. Allgaier, M., Grossart, H.P., 2006b. Seasonal dynamics and phylogenetic diversity of free-living and particle-associated bacterial communities in four lakes in northeastern Germany. Aquat. Microb. Ecol. 45, 115e128. Azam, F.T., Fenchel, J.G., Field, F.S., Gray, J.S., MeyereReil, L.A., Thingstad, F., 1983. The ecological role of water-column microbes in the sea. Mar. Ecol. Prog. Se. 10, 257e263. Carvalho, F.R.S., Vazoller, R.F., Foronda, A.S., Pellizari, V.H., 2007. Phylogenetic study of Legionella species in pristine and

1710

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 0 2 e1 7 1 0

polluted aquatic samples from a tropical Atlantic forest ecosystem. Curr. Microbiol. 55, 288e293. Chen, M., Chen, F., Xing, P., Li, H., Wu, Q.L., 2010. Microbial eukaryotic community in response to Microcystis spp. bloom, as assessed by an enclosure experiment in Lake Taihu, China. FEMS Microbiol. Ecol. 74, 19e31. Cole, J.J., Likens, G.E., Strayer, D.L., 1982. Photosynthetically produced dissolved organic carbon: an important carbon source for planktonic bacteria. Limnol. Oceanogr. 27, 1080e1090. Eiler, A., Bertilsson, S., 2004. Composition of freshwater bacterial communities associated with cyanobacterial blooms in four Swedish lakes. Environ. Microbiol. 6, 1228e1243. Ferdous, Z., 2009. Survival of Vibrio cholerae O139 in association with Anabaena variabilis in four different microcosoms. Am. J. Appl. Sci. 6, 439e444. Fields, B.S., Benson, R.F., Besser, R.E., 2002. Legionella and Legionnaires’ Disease: 25 Years of investigation. Clin. Microbiol. Rev. 15, 506e526. Fliermans, C.B., Cherry, W.B., Orrison, L.H., Smith, S.J., Tison, D.L., Pope, D.H., 1981. Ecological distribution of Legionella pneumophila. Appl. Environ. Microbiol. 41, 9e16. Goodfellow, M., Williams, S.T., 1983. Ecology of actinomycetes. Annu. Rev. Microbiol. 37, 189e216. Grossart, H.P., Czub, G., Simon, M., 2006. Algae-bacteria interactions and their effects on aggregation and organic matter flux in the sea. Environ. Microbiol. 8, 1074e1084. Grossart, H.P., Levold, F., Allgaier, M., Simon, M., Brinkhoff, T., 2005. Marine diatom species harbour distinct bacterial communities. Environ. Microbiol. 7, 860e873. Holmfeldt, K., Dziallas, C., Titelman, J., Pohlmann, K., Grossart, H.P., Riemann, L., 2009. Diversity and abundance of freshwater Actinobacteria along environmental gradients in the brackish northern Baltic Sea. Environ. Microbiol. 11, 2042e2054. Kolehmainena, R.E., Langwaldtb, J.H., Puhakka, J.A., 2007. Natural organic matter (NOM) removal and structural changes in the bacterial community during artificial groundwater recharge with humic lake water. Water Res. 41, 2715e2752. Lane, D.J., 1991. 16S/23S rRNA sequencing. In: Stackebrandt, E., Goodfellow, M. (Eds.), Nucleic Acid Techniques in Bacterial Systematics. John Wiley and Sons, Chichester, pp. 115e175. Lehman, E.M., 2007. Seasonal occurrence and toxicity of Microcystis in impoundments of the Huron River, Michigan, USA. Water Res. 41, 795e802. Lindstro¨m, E.S., Agterveld, M.P.K., Zwart, G., 2005. Distribution of typical freshwater bacterial groups is associated with pH, temperature, and lake water retention time. Appl. Environ. Microbiol. 71, 8201e8206. Liu, W.T., Marsh, T.L., Cheng, H., Forney, L.J., 1997. Characterisation of microbial diversity by determining terminal restriction length polymorphisms of genes encoding 16S rDNA. Appl. Environ. Microbiol. 63, 4516e4522. Maruyama, T., Kato, K., Yokoyama, A., Tanaka, T., Hiraishi, A., Park, H.D., 2003. Dynamics of microcystin-degrading bacteria in mucilage of Microcystis. Microb. Ecol. 46, 279e288. Masango, M., Myburgh, J., Botha, C., Labuschagne, L., Naicker, D., 2008. A comparison of in vivo and in vitro assays to assess the toxicity of algal bloom. Water Res. 42, 3241e3248. Mauricea, P.A., Leff, L.G., 2002. Hydrogeochemical controls on the organic matter and bacterial ecology of a small freshwater wetland in the New Jersey Pine Barrens. Water Res. 36, 2561e2570. Muyzer, G., Dewaal, E.C., Uitterlinden, A.G., 1993. Profiling of complex microbial populations by denaturig gradient gel electrophoresis analysis of polymerase chain reactionamplified genes coding for 16S rRNA. Appl. Environ. Microbiol. 59, 695e700. Newton, R.J., Jones, S.E., Helmus, M.R., McMahon, K.D., 2007. Phylogenetic ecology of the freshwater Actinobacteria acI Lineage. Appl. Environ. Microbiol 73, 7169e7176.

Ortiz-Roque, C.M., Hazen, T.C., 1987. Abundance and distribution of Legionellaceae in Puerto Rican waters. Appl. Environ. Microbiol. 53, 2231e2236. Paerl, H.W., Fulton, R.S., Moisander, P.H., Dyble, J., 2001. Harmful freshwater algal blooms, with an emphasis on cyanobacteria. The Scientific World 1, 76e113. Paerl, H.W., Dyble, J., Moisander, P.H., Noble, R.T., Piehler, M.F., Pinckney, J.L., Steppe, T.F., Twomey, L., Valdes, L.M., 2003. Microbial indicators of aquatic ecosystem change: current applications to eutrophication studies. FEMS Microbiol. Ecol. 46, 233e246. Porter, K.G., Feig, Y.S., 1980. The use of DAPI for identifying and counting aquatic microflora. Limnol. Oceanogr. 25, 943e948. Qin, B., Xu, P., Wu, Q., Luo, L., Zhang, Y., 2007. Environmental issues of Lake Taihu, China. Hydrobiol. 581, 3e14. Schloss, P.D., Westcott, S.L., Ryabin, T., Hall, J.R., Hartmann, M., Hollister, E.B., Lesniewski, R.A., Oakley, B.B., Parks, D.H., Robinson, C.J., Sahl, J.W., Stres, B., Thallinger, G.G., Van Horn, D.J., Weber, C.F., 2009. Introducing mothur: open-source, Platform-independent, CommunityeSupported software for describing and comparing microbial communities. Appl. Environ. Microbiol. 75, 7537e7541. Tamura, K., Dudley, J., Nei, M., Kumar, S., 2007. MEGA4: Molecular Evolutionary Genetics Analysis (MEGA) Software Version 4.0. Mol. Biol. Evol. 24, 1596e1599.  ter Braak, C.J.F., Smilauer, P., 2002. CANOCO Reference Manual and CanoDraw for Windows User’s Guide: Software for Canonical Community Ordination (Version 4.5). Microcomputer Power, Ithaca, N.Y., USA. Tison, D.L., Pope, D.H., Cherry, W.B., Fliermans, C.B., 1980. Growth of Legionella pneumophila in association with blue-green algae (cyanobacteria). Appl. Environ. Microbiol. 39, 456e459. Wilkinson, H.W., Thacker, W.L., Benson, R.F., Polt, S.S., Brookings, E., Mayberry, W.R., Brenner, D.J., Gilley, R.G., Kirklin, J.K., 1987. Legionella birminghamensis sp. nov. isolated from a cardiac transplant recipient. J. Appl. Microbiol. 25, 2120e2122. Wu, Q.L., Chen, Y., Xu, K., Liu, Z., Hahn, M.W., 2007a. Intrahabitat heterogeneity of the microbial food web structure under the regime of eutrophication and sediment resuspension in the large subtropical shallow LakeTaihu. Hydrobiol. 581, 241e254. Wu, Q.L., Zwart, G., Wu, J.F., Kamst-Van Agterveldand, M.P., Liu, S.J., Hahn, M.W., 2007b. Submersed macrophytes play an important role in structuring the bacterioplankton diversity in a large subtropical shallow Taihu Lake, China. Environ. Microbiol. 9, 2765e2774. Wu, Q.L., Zwart, G., Schauer, M., Kamst-van Agterveld, M.P., Hahn, M.W., 2006. Bacterioplankton community composition along a salinity gradient of sixteen high-mountain lakes located on the Tibetan Plateau, China. Appl. Environ. Microbiol. 72, 5478e5485. Xing, P., Kong, F., 2007. Intra-habitat heterogeneity of environmental factors regulating bacterioplankton community composition in Lake Taihu, China. Aquat. Microb. Ecol. 46, 113e122. Yamamoto, Y., Kouchiwa, T., Hodoki, Y., Hotta, K., Uchida, H., Harada, K., 1998. Distribution and identification of actinomycetes lysing cyanobacteria in a eutrophic lake. J. Appl. Phycol. 10, 391e397. Zhang, X., Warming, T.P., Hu, H., Christoffersen, K.S., 2009. Life history responses of Daphnia magna feeding on toxic Microcystis aeruginosa alone and mixed with a mixotrophic Poterioochromonas species. Water Res. 43, 5053e5062. Zwart, G., Crump, B.C., Agterveld, M.P.K., Hagen, F., Han, S., 2002. Typical freshwater bacteria: an analysis of available 16S rRNA gene sequences from plankton of lakes and rivers. Aquat. Microb. Ecol. 28, 141e155.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 1 1 e1 7 1 9

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Insight into the heavy metal binding potential of dissolved organic matter in MSW leachate using EEM quenching combined with PARAFAC analysis Jun Wu, Hua Zhang, Pin-Jing He*, Li-Ming Shao State Key Laboratory of Pollution Control and Resources Reuse, Key Laboratory of Yangtze River Water Environment, College of Environmental Science and Engineering, Tongji University, 1239 Siping Road, Shanghai 200092, PR China

article info

abstract

Article history:

Dissolved organic matter (DOM) plays an important role in heavy metal migration from

Received 4 September 2010

municipal solid waste (MSW) to aquatic environments via the leachate pathway. In this

Received in revised form

study, fluorescence excitation-emission matrix (EEM) quenching combined with parallel

13 November 2010

factor (PARAFAC) analysis was adopted to characterize the binding properties of four heavy

Accepted 15 November 2010

metals (Cu, Pb, Zn and Cd) and DOM in MSW leachate. Nine leachate samples were

Available online 24 November 2010

collected from various stages of MSW management, including collection, transportation, incineration, landfill and subsequent leachate treatment. Three humic-like components

Keywords:

and one protein-like component were identified in the MSW-derived DOM by PARAFAC.

Municipal solid waste

Significant differences in quenching effects were observed between components and metal

(MSW) leachate

ions, and a relatively consistent trend in metal quenching curves was observed among

Dissolved organic matter (DOM)

various leachate samples. Among the four heavy metals, Cu(II) titration led to fluorescence

Heavy metal

quenching of all four PARAFAC-derived components. Additionally, strong quenching effects were only observed in protein-like and fulvic acid (FA)-like components with the

Fluorescence excitation-emission

matrix

(EEM)

addition of Pb(II), which suggested that these fractions are mainly responsible for Pb(II)

quenching

binding in MSW-derived DOM. Moreover, the significant quenching effects of the FA-like

Parallel factor (PARAFAC) analysis

component by the four heavy metals revealed that the FA-like fraction in MSW-derived DOM plays an important role in heavy metal speciation; therefore, it may be useful as an indicator to assess the potential ability of heavy metal binding and migration. ª 2010 Elsevier Ltd. All rights reserved.

1.

Introduction

Leachate from municipal solid waste (MSW) has raised concerns as a potential pathway of contaminant emission (Huang et al., 2009; Michalzik et al., 2007). Heavy metals are significant pollutants in MSW leachate, and are often present in concentrations ranging from micrograms to milligrams per liter (Christensen et al., 2001). Primary investigations (Baumann et al., 2006; Baun and Christensen, 2004; Li et al., 2009) of heavy metal distribution in landfill leachate have shown that

significant fractions of heavy metals were associated with MSWderived dissolved organic matter (DOM), suggesting that DOM plays an important role in heavy metal speciation and migration. Christensen et al. (1999, 1996) and Christensen and Christensen (1999) found that the DOM derived from MSW landfill leachate had a high affinity for Cu, Pb, Cd, Zn, and Ni (especially for Cu and Pb); thus, enhancing their mobility in leachate-polluted waters. By using biological assay methods, it was observed that the heavy metal binding capacities fluctuated by more than one order of magnitude among various leachates,

* Corresponding author. Tel./fax: þ86 21 6598 6104. E-mail address: [email protected] (P.-J. He). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.11.022

1712

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 1 1 e1 7 1 9

which was attributed to their variable compositions (Ward et al., 2005). Isolation of humic substances such as humic acid (HA) or fulvic acid (FA) from natural DOM was preferable to study the relationship between DOM and heavy metals (da Silva and Oliveira, 2002; Yang and van den Berg, 2009). Humic substances have been reported to exhibit strong affinities toward metal ions due to the large number of ionizable functional groups, which are mainly carboxylic and phenolic groups (Hernandez et al., 2006; Martensson et al., 1999; Terbouche et al., 2010). However, the heterogeneity and operationally-defined extraction and purification process of humic substances via the addition of chemicals enhance the complexity of the binding potential of DOM with metals. Furthermore, metal binding properties of DOM in the leachate from MSW landfill or other stages of MSW management have seldom been investigated to the best of our knowledge. Therefore, new insight into the heavy metal binding potential of MSW-derived DOM is needed. Fluorescence excitation-emission matrix (EEM) spectroscopy is a simple, sensitive and non-destructive technique that can provide valuable information regarding the molecular structure of DOM. Recent studies have demonstrated that EEM spectroscopy combined with a quenching method can be applied as a reliable technique to enable a better understanding of the binding properties of metal ions and isolated fluorescence substances from soil and sewage sludge (Plaza et al., 2006a, b). However, the EEM spectra of in situ DOM cannot be easily identified owing to the various types of overlapping fluorophores (Henderson et al., 2009). Recent advances in parallel factor (PARAFAC) analysis, a multivariate chemometric method, have greatly improved the quantitative interpretation of EEMs (Engelen et al., 2009; Fellman et al., 2009). PARAFAC can be applied to decompose fluorescence EEMs into different independent groups of fluorescent components, which can then reduce the interference among fluorescent compounds and allow a more accurate quantification (Andersen and Bro, 2003). Ohno et al. (2008) demonstrated that the binding properties between water soluble soil organic matter and trivalent metals (Al and Fe) could be well characterized by EEM quenching combined with PARAFAC. Additionally, Yamashita and Jaffe (2008) successfully evaluated the binding properties of surface water-borne DOM and divalent metals (Cu and Hg) using the same method. The structures and compositions of MSW-derived DOM are much more complicated than the aforementioned DOMs (Christensen et al., 1998; He et al., 2006). Furthermore, a series of biochemical reactions may occur, resulting in changes in the DOM properties during integrated MSW management (Huo et al., 2009). As a result, simple quantification of the heavy metal binding potential of MSW-derived DOM has remained elusive. In this study, the feasibility of EEM quenching-PARAFAC for use in characterization of the binding properties of heavy metals and DOM in MSW leachate was investigated. Nine leachate samples from various stages in MSW management were collected, and four heavy metals (Cu, Pb, Zn and Cd) that are common in MSW leachate were used as fluorescent quenching agents for titration. Specific fractions in MSWderived DOM responsible for heavy metals binding were identified and the differences in the binding characteristics among

various PARAFAC-derived components, heavy metals and DOMs were analyzed.

2.

Materials and methods

2.1.

Sample collection and preparation

Due to the high content of moisture and putrescible organic waste, MSW leachate is produced during collection, transportation, and storage. Nine leachate samples were collected from different stages of MSW management in Shanghai, China (S1eS9 as shown in Fig. 1). S1 and S2 were collected from an MSW collection vehicle and transfer station, respectively, with waste generated within one day. S3 was collected from the refuse bunk of an incineration plant in which MSW was stored for 1e3 days before incineration. S4 was gathered from the effluent of S3 treated by a sequential batch reactor (SBR) process combined with ultrafiltration (UF). Fresh leachate (S5) and old leachate (S7) were obtained from sanitary landfills comprising refuse aged 1e2 years and 5e10 years, respectively. S6 was the effluent of S5 treated by an upflow anaerobic sludge bed (UASB) combined with a membrane bioreactor (MBR). Leachate S7 was treated by an anaerobic-anoxic-oxic (A2O) process and an SBR, and the effluent of these processes comprised S8 and S9, respectively. The leachate samples were collected in pre-cleaned (by nitric acid and Milli-Q water) brown sampling bottles, after which they were filtered through a 0.45 mm membrane filter. The filtrates were then stored at 4  C until use. The physiochemical characteristics of the filtered samples are summarized in Table 1.

2.2.

Fluorescence titration

Prior to fluorescence titration, the leachate samples were diluted with Milli-Q water to TOC < 10 mg/L to ensure that the maximum fluorescence signal was below the upper detection limit of the spectrometer. Aliquots of 25 mL of the diluted solution of DOM were titrated in 40-mL brown sealed vials with 0.01 or 0.1 mol/L Pb(NO3)2, Cu(NO3)2, Zn(NO3)2, and Cd(NO3)2 using an automatic syringe. The metal concentrations in the final solutions ranged from 0 to 100 mmol/L. To maintain a constant pH condition before and after titration, the metal titrants were adjusted to pH 4.0 for Cu(NO3)2, pH 4.5 for Pb(NO3)2, pH 6.0 for Zn(NO3)2 and pH 7.0 for Cd(NO3)2, based on the Visual MINTEQ calculation that no precipitate existed in the solution and no more than 0.025 mL of the metal titrant was added during titration. All titrated solutions were shaken for 24 h at 25  0.1  C to ensure complexation equilibrium. The fluorescence EEM spectra of the titrated samples were measured using a Cary Eclipse fluorescence spectrophotometer (Varian Inc., Palo Alto, CA, USA), and collected by subsequent scanning emission from 250 to 500 nm at 2 nm increments by varying the excitation wavelength from 200 to 450 nm at 10 nm increments. The spectra were recorded at a scan rate of 1200 nm/min, using excitation and emission slit bandwidths of 5 nm. The voltage of the photomultiplier tube was set to 800 V to enable low level light detection.

1713

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 1 1 e1 7 1 9

Incinerator S3 MBR+UF Leachate

Transfer station

Collector

S4 Effluent

landfill S1 Leachate

S5 Fresh leachate UASB+MBR

S2 Leachate

S7 Old leachate

AO SBR

S6 Effluent S8 Effluent S9 Effluent

Fig. 1 e Leachate samples collected from MSW management facilities. MBR: membrane bioreactor, UF: ultrafiltration, UASB: upflow anaerobic sludge bed, A2O: anaerobic-anoxic-oxic process, SBR: sequential batch reactor.

2.3.

PARAFAC modeling

split half analysis, residual analysis, and visual inspection (Stedmon and Bro, 2008).

The approach of PARAFAC analysis of EEMs has been described in detail elsewhere (Bro, 1997; Stedmon and Bro, 2008) and is briefly described here. PARAFAC is a generalization of bilinear principal component analysis (PCA) to higher order arrays. In other words, PARAFAC decomposes N-way arrays into N loading matrices. Therefore, if fluorescence EEMs are arranged in a three-way array X of dimensions I  J  K, where I is the number of samples, J the number of emission wavelengths, and K the number of excitation wavelengths, PARAFAC decomposes them into three matrices A (the score matrix), B and C (loading matrices) with elements aif, bjf, and ckf. In this study, PARAFAC analysis was conducted using MATLAB 7.0 (Mathworks, Natick, MA) with the DOMFluor toolbox (www.models.life.ku.dk). A non-negativity constraint was applied to the parameters allowing only chemically relevant results. Some preprocessing steps were adopted to minimize the influence of scatter lines and other attributes of the EEM landscape. The EEM of a control (Milli-Q water) was subtracted from each sample EEM and the other Rayleigh and Raman scatters were removed according to the protocol described by Bahram et al. (2006). After the removal of the Rayleigh and Raman scatters, EEMs were normalized by dividing the spectrum by the corresponding TOC concentration. This resulted in reduction of the impact of varying DOM concentrations in different leachate samples on the component score matrix. The PARAFAC models with two to eight components were computed for the EEMs. Determination of the correct number of components was primarily based on

2.4.

Complexation modeling

The complexation parameters between heavy metals and PARAFAC-derived components were determined using the single-site fluorescence quenching model proposed by Ryan and Weber (1982). This model is based on the assumption that metal ion binding occurs at identical and independent binding sites or ligands and only 1:1 metal/ligand complexes are formed. Because PARAFAC can decompose the complex mixture of DOM fluorophores into several independent fluorescence components, the application of this model may provide more appropriate information than the fluorescence intensity from the peak maxima of bulk samples. The complexation parameters were determined using nonlinear fitting of Eq. (1) 

1 1 þ KM CL þ KM CM I ¼ I0 þ ðIML  I0 Þ 2KM CL qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi  ð1 þ KM CL þ KM CM Þ2 4K2M CL CM

(1)

where, I and I0 are the fluorescence intensity at the metal concentration, CM, and at the beginning of the titration (without adding metals), respectively. IML is the limiting value below which the fluorescence intensity does not change due to the addition of metal. KM and CL are the conditional stability constant and total ligand concentration, respectively. Nonlinear fitting using the QuasieNewton algorithm was applied to estimate the IML, KM and CL.

Table 1 e Physiochemical characteristics of the leachate samples. Sample S1 S2 S3 S4 S5 S6 S7 S8 S9

TC (mg/L)

IC (mg/L)

TOC (mg/L)

TN (mg/L)

pH

Ca (mg/L)

Mg (mg/L)

Al (mg/L)

Fe (mg/L)

Cu (mg/L)

Pb (mg/L)

Zn (mg/L)

Cd (mg/L)

Mn (mg/L)

11300 19000 13300 1230 8860 1050 2350 220 1240

780 71 1090 560 753 790 1390 93 916

10520 18929 12210 670 8107 260 960 127 324

1140 1210 2030 ND 3820 61 2380 784 1050

4.03 3.92 6.06 7.99 7.17 8.45 7.48 7.76 8.36

907 805 1400 51.7 696 33.4 253 43.5 58.1

115 190 273 210 311 133 123 154 105

8.95 6.53 <0.05 0.75 0.275 ND 0.108 ND 0.715

55.7 73.1 5.53 0.795 ND ND 1.16 ND 0.205

0.033 ND ND ND ND ND 0.045 ND <0.005

ND 0.648 ND ND ND ND ND ND ND

3.74 3.35 ND ND ND ND ND ND ND

ND 0.015 ND ND ND ND ND ND ND

0.543 2.72 7.56 0.233 1.05 0.095 0.090 0.035 0.133

ND: not detected; TC: total carbon; IC: inorganic carbon; TOC: total organic carbon.

Fig. 2 e Fluorescence excitation-emission matrix spectra of the nine leachate samples (bulk sample) with or without titration of Cu(II), Pb(II), Zn(II) and Cd(II) at a total concentration of 50 mmol/L. Blank: no added heavy metals, FI/TOC: fluorescence intensity per unit total organic carbon (arbitrary units/(mg/L)).

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 1 1 e1 7 1 9

3.

Results and discussion

3.1.

EEM contours of the leachate samples

The EEM spectra of the leachate samples measured in the absence and presence of Cu(II), Pb(II), Zn(II) or Cd(II) at a total concentration of 50 mmol/L are shown in Fig. 2. The results revealed that the EEM contours of the DOM fraction in S1 and S2 were similar, being characterized by two peaks at excitation/emission (Ex/Em) values of 230/336 and 280/334 for S1, and 230/350 and 280/352 for S2, which are commonly labeled as protein-like substances (Chen et al., 2003). As the MSW storage time increased, the fluorescence intensity of DOM in leachate at longer Ex/Em wavelengths increased gradually (S3eS5eS7) which is consistent with the humification process of MSW. Comparison of the EEMs of the DOMs before (S3, S5 and S7) and after (S4, S6, S8 and S9) treatment revealed that the protein-like substances in leachate could be removed effectively whereas humic-like substances were more concentrated per unit of total organic carbon (TOC). Addition of Cu(II), Pb(II), Zn(II) or Cd(II) to DOM solutions induced various changes in the EEM spectra depending on DOM characteristics and metal species. In particular, a marked decrease in fluorescence intensity in all nine leachate samples was observed in response to the addition of Cu(II), indicating its large quenching effect on DOM in MSW leachate. Conversely, the wavelengths of Ex/Em peaks of S1, S4, S7, S8 and S9 remained nearly constant when Zn(II) or Cd(II) was added. Additionally, Pb(II) titration induced an obvious quenching

1715

effect on the DOM in fresh leachate (S1, S2, S3, S4, S5 and S6), whereas it had only a slight effect on the DOM in old leachate (S7, S8 and S9). These results supported the suggestion that the binding characteristics between Pb(II) and various fractions in DOM might be different. As expected, further analysis based on visual peaking was quite difficult owing to serious overlaps in the peaks of MSWderived DOM, particularly for S3, S5 and S7.

3.2.

PARAFAC analysis of EEM spectra

The EEM spectra of nine leachate samples titrated with four heavy metals (Cu, Pb, Zn and Cd) at ten different concentrations were analyzed by PARAFAC. Split half analysis, residual analysis and visual inspection identified that four components were appropriate (Fig. 3). All the fluorescence EEMs can be successfully decomposed by PARAFAC analysis into a fourcomponent model, despite the dissimilar fluorescence characteristics of the nine leachate samples and the different quenching effect of different metals at various concentrations. As shown in Fig. 4, the PARAFAC model identified three humic-like substances (Component 1, Component 2 and Component 4) and one protein-like substance (Component 3). Component 1 was comprised of two peaks that were similar to the hydrophobic fraction (Ex/Em 230e245/400e430) and hydrophilic fraction (Ex/Em 300e350/400e430) of surface water-borne DOM identified by Chen et al. (2003). The fluorescence characteristics of component 2 have rarely been identified. Lu¨ et al. (2009) observed similar fluorophores

Fig. 3 e EEM analysis by DOMFluor-PARAFAC model a) Split half analysis, b) Excitation and emission loadings of four PARAFAC-derived components, c) Residual analysis of components 3e5.

1716

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 1 1 e1 7 1 9

200 Component 1 Component 2 Component 3 Component 4

150 100 50 0 100 80

Percentage (%)

(Ex/Em 230,260,340/466 and 250,310,360/464) and suggested that this may have been related to the pyrene family. The peak of component 4 was similar to that of standard Suwannee River fulvic acid (214e220,440e450) reported by Chen et al. (2003). The fluorophore of component 3, which was characterized by two peaks of Ex/Em 225e237/340e381 as well as Ex/Em 275/340, exhibited EEM peaks resembling tryptophan substances derived from sewage DOM (Henderson et al., 2009). Baker and Curry (2004) also identified one similar peak (Ex/Em 220e230/340e370) in three landfill leachate samples. PARAFAC analysis provided additional quantitative information describing the distribution of the four components in the nine leachate DOMs (Fig. 5). The protein-like substances represented by component 3, which occupied a dominant proportion in the fresh leachate samples dropped gradually from 76.3% (S1), 67.9% (S2), 58.1% (S3), and 40.5% (S5) to 29.2% (S7) as the MSW storage time increased, and further decreased by 2.43 (S4), 2.95 (S6), 1.79 (S8) and 1.69 (S9) times after various treatments, whereas the humic-like substances increased accordingly.

Fluorescence intensity per unit TOC arbitrary unit/(mg/L)

Fig. 4 e Fluorescence excitation-emission matrix contours of the four components identified by the DOMFluor-PARAFAC model.

60 40 20 0

3.3. Interactions between PARAFAC-derived components and heavy metals Fig. 6 shows the fluorescence quenching curves of each component with the addition of Cu(II), Pb(II), Zn(II) and Cd(II). Although marked differences in fluorescent fluorophores were observed between component 1 and component 2

S1

S2

S3

S4

S5

S6

S7

S8

S9

Fig. 5 e Distribution of four PARAFAC-derived components in nine MSW leachate samples (no added heavy metals). The percentage of each component was calculated by dividing its fluorescence intensity per unit of TOC by that of the corresponding bulk sample.

1717

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 1 1 e1 7 1 9

C o m po nent 1

75 60 30 15 0

C o m po nent 2

160 120 80 30 15 0

C o m p o n en t 3

75 60 45 30 15 0 90 75 60 45 30 15 0

C o m p o n en t 4

Fluorescence intensity per unit TOC arbitrary units/(mg/L)

45

0 20 40 60 80 100

0 20 40 60 80 100

0 20 40 60 80 100

0 20 40 60 80 100

Cu (µ mol/L)

Pb (µ mol/L)

Zn (µ mol/L)

Cd (µ mol/L)

S1

S2

S3

S4

S6

S7

S8

S9

S5

Fig. 6 e Changes in the fluorescence intensity of four PARAFAC-derived components with the addition of Cu(II), Pb(II), Zn(II) and Cd(II).

(Fig. 4), the fluorescence quenching curves of these two components with four heavy metals were quite similar. A larger quenching effect was observed for Cu(II), whereas negligible quenching effects were found for Pb(II), Zn(II) and Cd(II) by component 1 and 2. The quenching characteristics of these two components with the addition of Cu(II), Zn(II) and

Cd(II) were quite similar to that of commercial HA (Divya et al., 2009), soil-borne HA and compost-derived HA (Hernandez et al., 2006; Plaza et al., 2006a; Provenzano et al., 2004). However, the negligible effects of Pb(II) on those two components in this study were different from those of HA derived from various sources in previous studies.

Table 2 e The log KM values for the humic-like components in MSW-derived DOM with four heavy metals (Cu, Pb, Zn and Cd) determined by Ryan and Weber model.a Symbol

S1 S2 S3 S4 S5 S6 S7 S8 S9

Cu

Pb

Comp1

Comp2

Comp4

e e e 5.94(0.96) e 3.77(0.98) 4.61(0.94) 4.27(0.99) 3.77(0.96)

e e e 4.66(0.99) FM FM FM FM FM

5.62(0.98) 5.48(0.94) 4.71(0.96) 5.58(0.98) 4.42(0.99) 5.34(0.99) 5.03(0.95) e 5.36(0.96)

Zn

Cd

Comp4 FM 4.30(0.93) 4.78(0.94) 4.31(0.98) 4.08(0.99) 3.84(0.99) FM e 5.01(0.99)

4.44(0.96) 4.81(0.95) 3.75(0.96) FM 3.99(1.00) 4.24(0.99) 4.61(0.98) e 4.08(0.99)

FM FM FM 4.64(0.94) FM FM FM e 5.10(0.99)

a The values in parentheses are R2, FM: failed to be modeled, “e”: the fluorescence intensity per unit of TOC was too low (Fig. 4) to be modeled.

1718

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 1 1 e1 7 1 9

FA-like component 4 was quenched significantly by each of the four heavy metals, which is consistent with previous reports describing the quenching effects of the heavy metals on commercial FA (Zhao and Nelson, 2005) and soil-borne FA (da Silva and Oliveira, 2002). These results strongly suggest that FA-like substances played a key role in the complexation between heavy metals and DOM in MSW leachate, especially for Zn(II) and Cd(II). In previous studies, investigators focused on the effect that elevated organic matter had on facilitating the transport of heavy metals when MSW leachate enters surface waters (Christensen et al., 1999, 1996). In this study, component 4 was found to account for more than 20% (Fig. 5) of all leachate samples except for S8 and S9. The relatively higher FA-like content compared to surface water-borne DOM may be another reason for the improved mobility of heavy metals in leachate-polluted waters. Recognizing the important role that FA-like substances in MSW-derived DOM have played in heavy metal binding, some strategies could be adopted to improve the MSW leachate treatment and management to control the migration of heavy metals. It is well-known that fluorescent protein-like substances are quenched or enhanced by complexation of metal ions. However, the fluorescence quenching method, which is widely used to determine the binding parameters between humic fluorescence substances and metal ions, has seldom been utilized to characterize the heavy metal binding potential of specific protein-like substances. In this study, larger fluctuations in the quenching curves of protein-like component 3 were observed, compared to those of humic-like components, especially for Zn(II) and Cu(II). Similar phenomena were also observed in a study conducted by Yamashita and Jaffe (2008), who employed PARAFAC-EEM quenching to explore the interactions between surface water-borne DOM and Cu(II). These results suggest that this method may not be appropriate for evaluation of the binding characteristics of protein-like substances and metal ions. The exact reason for these findings is not yet clear, but it may be related to the stability of the complex of protein-like substances and metals. Cu(II) quenched the fluorescence intensity of all four PARAFAC-derived components. It is interesting to note that the quenching effects with the addition of Pb(II) were strong for component 3 and 4, whereas they were weak for component 1 and 2. In particular, FA-like and protein-like fractions played a dominant role in Pb(II) binding, which explains the different binding characteristics between Pb(II) with fresh and aged leachate. By contrast, the quenching effects in the presence of Zn(II) and Cd(II), which could not be observed based on visual peaking of the bulk sample fluorophore, were observed in component 3 as well as component 4. These findings clearly demonstrated that PARAFAC-based results may provide additional information that may be neglected during visual peaking analysis due to overlapping fluorophores. The stability constants (log KM) calculated using the Ryan and Weber Model for PARAFAC-derived humic-like components and four heavy metals are listed in Table 2. The log KM values ranged from 3.77 to 5.94, 3.84 to 5.01 and 3.75 to 4.81 for Cu(II), Pb(II) and Zn(II), respectively, which were in the same ranges as those found for bulk DOM samples (Luster et al., 1996), physical fractionate DOM samples (de Zarruk et al., 2007) and humic substances (Plaza et al., 2006b; Provenzano et al., 2004).

There was no systematic trend of log KM values observed among the leachate samples or PARAFAC-derived components. The binding parameters between component 1 and Cu(II) could be modeled well, whereas those between component 2 and Cu (II) could not be modeled for most samples, even though similar quenching curves were observed (Fig. 6). These findings suggested that the binding mechanisms between Cu(II) and component 1 and component 2 may be different. Additionally, the binding parameters between component 4 and Cd(II) could not be modeled well by the 1:1 complexation model, which was in agreement with the binding mechanisms of Cd(II) and commercial FA (Grassi and Daquino, 2005).

4.

Conclusions

Four components with characteristic peaks at Ex/Em of (240, 330)/412, (250, 300, 360)/458, (230, 280)/340 and 220/432, were identified by the DOMFluor-PARAFAC model. The results suggested that all the fluorescence EEMs could be successfully decomposed by PARAFAC analysis into a four-component model, despite the dissimilar fluorescence characteristics of the nine leachate samples and the different quenching effects of different metals at various concentrations. The combination of EEM quenching and PARAFAC could provide additional valuable information regarding the binding properties between heavy metals and specific humic-like fluorescence components in DOM, which may be neglected during visual peaking analysis due to overlapping fluorophores. Therefore, it is a useful tool for evaluation of the interactions between DOM and heavy metals given its nondestructive nature, high sensitivity and selectivity. PARAFACbased results revealed that the FA-like fraction in MSW-derived DOM played an important role in heavy metal speciation; therefore, it may be useful as an indicator to assess the potential ability of heavy metal binding and migration.

Acknowledgement This study was financially supported by the National Basic Research Program of China (973 Program No. 2011CB201504), the National Natural Science Foundation of China (No. 20807031) and Ministry of Education (No. 20090072120068).

references

Andersen, C.M., Bro, R., 2003. Practical aspects of PARAFAC modeling of fluorescence excitation-emission data. J. Chemom. 17 (4), 200e215. Bahram, M., Bro, R., Stedmon, C., Afkhami, A., 2006. Handling of Rayleigh and Raman scatter for PARAFAC modeling of fluorescence data using interpolation. J. Chemom. 20 (3e4), 99e105. Baker, A., Curry, M., 2004. Fluorescence of leachates from three contrasting landfills. Water Res. 38 (10), 2605e2613. Baumann, T., Fruhstorfer, P., Klein, T., Niessner, R., 2006. Colloid and heavy metal transport at landfill sites in direct contact with groundwater. Water Res. 40 (14), 2776e2786.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 1 1 e1 7 1 9

Baun, D.L., Christensen, T.H., 2004. Speciation of heavy metals in landfill leachate: a review. Waste Manage. Res. 22 (1), 3e23. Bro, R., 1997. PARAFAC. Tutorial and applications. Chemometr. Intell. Lab. 38 (2), 149e171. Chen, W., Westerhoff, P., Leenheer, J.A., Booksh, K., 2003. Fluorescence excitation-emission matrix regional integration to quantify spectra for dissolved organic matter. Environ. Sci. Technol. 37 (24), 5701e5710. Christensen, J.B., Christensen, T.H., 1999. Complexation of Cd, Ni, and Zn by DOC in polluted groundwater: a comparison of approaches using resin exchange, aquifer material sorption, and computer speciation models (WHAM and MINTEQA2). Environ. Sci. Technol. 33 (21), 3857e3863. Christensen, J.B., Jensen, D.L., Christensen, T.H., 1996. Effect of dissolved organic carbon on the mobility of cadmium, nickel and zinc in leachate polluted groundwater. Water Res. 30 (12), 3037e3049. Christensen, J.B., Jensen, D.L., Gron, C., Filip, Z., Christensen, T.H., 1998. Characterization of the dissolved organic carbon in landfill leachate-polluted groundwater. Water Res. 32 (1), 125e135. Christensen, J.B., Botma, J.J., Christensen, T.H., 1999. Complexation of Cu and Pb by DOC in polluted groundwater: a comparison of experimental data and predictions by computer speciation models (WHAM and MINTEQA2). Water Res. 33 (15), 3231e3238. Christensen, T.H., Kjeldsen, P., Bjerg, P.L., Jensen, D.L., Christensen, J.B., Baun, A., Albrechtsen, H.J., Heron, C., 2001. Biogeochemistry of landfill leachate plumes. Appl. Geochem. 16 (7e8), 659e718. da Silva, J.C.G.E., Oliveira, C.J.S., 2002. Metal ion complexation properties of fulvic acids extracted from composted sewage sludge as compared to a soil fulvic acid. Water Res. 36 (13), 3404e3409. de Zarruk, K.K., Scholer, G., Dudal, Y., 2007. Fluorescence fingerprints and Cu2þ-complexing ability of individual molecular size fractions in soil- and waste-borne DOM. Chemosphere 69 (4), 540e548. Divya, O., Venkataraman, V., Mishra, A.K., 2009. Analysis of metal ion concentration in humic acid by excitationeemission matrix fluorescence and chemometric methods. J. Appl. Spectrosc. 76 (6), 864e875. Engelen, S., Frosch, S., Jorgensen, B.M., 2009. A fully robust PARAFAC method for analyzing fluorescence data. J. Chemom. 23 (3e4), 124e131. Fellman, J.B., Miller, M.P., Cory, R.M., D’Amore, D.V., White, D., 2009. Characterizing dissolved organic matter using PARAFAC modeling of fluorescence spectroscopy: a comparison of two models. Environ. Sci. Technol. 43 (16), 6228e6234. Grassi, M., Daquino, V., 2005. 113Cd NMR and fluorescence studies of multiple binding mechanisms of Cd(II) by the Suwannee river fulvic acid. Anal. Chim. 95 (7e8), 579e591. He, P.J., Xue, J.F., Shao, L.M., Li, G.J., Lee, D.J., 2006. Dissolved organic matter (DOM) in recycled leachate of bioreactor landfill. Water Res. 40 (7), 1465e1473. Henderson, R.K., Baker, A., Murphy, K.R., Hamblya, A., Stuetz, R. M., Khan, S.J., 2009. Fluorescence as a potential monitoring tool for recycled water systems: a review. Water Res. 43 (4), 863e881. Hernandez, D., Plaza, C., Senesi, N., Polo, A., 2006. Detection of copper(II) and zinc(II) binding to humic acids from pig slurry and amended soils by fluorescence spectroscopy. Environ. Pollut. 143 (2), 212e220. Huang, J.H., Ilgen, G., Vogel, D., Michalzik, B., Hantsch, S., Tennhardt, L., Bilitewski, B., 2009. Emissions of inorganic and organic arsenic compounds via the leachate pathway from

1719

pretreated municipal waste materials: a landfill reactor study. Environ. Sci. Technol. 43 (18), 7092e7097. Huo, S.L., Xi, B.D., Yu, H.C., Liu, H.L., 2009. Dissolved organic matter in leachate from different treatment processes. Water Environ. J. 23 (1), 15e22. Li, R., Yue, D.B., Liu, J.G., Nie, Y.F., 2009. Size fractionation of organic matter and heavy metals in raw and treated leachate. Waste Manage. 29 (9), 2527e2533. Lu¨, F., Chang, C.H., Lee, D.J., He, P.J., Shao, L.M., Su, A., 2009. Dissolved organic matter with multi-peak fluorophores in landfill leachate. Chemosphere 74 (4), 575e582. Luster, J., Lloyd, T., Sposito, G., Fry, I.V., 1996. Multi-wavelength molecular fluorescence spectrometry for quantitative characterization of copper(II) and aluminum(III) complexation by dissolved organic matter. Environ. Sci. Technol. 30 (5), 1565e1574. Martensson, A.M., Aulin, C., Wahlberg, O., Agren, S., 1999. Effect of humic substances on the mobility of toxic metals in a mature landfill. Waste Manage. Res. 17 (4), 296e304. Michalzik, B., Ilgen, G., Hertel, F., Hantsch, S., Bilitewski, B., 2007. Emissions of organo-metal compounds via the leachate and gas pathway from two differently pre-treated municipal waste materials e a landfill reactor study. Waste Manage. 27 (4), 497e509. Ohno, T., Amirbahman, A., Bro, R., 2008. Parallel factor analysis of excitation-emission matrix fluorescence spectra of water soluble soil organic matter as basis for the determination of conditional metal binding parameters. Environ. Sci. Technol. 42 (1), 186e192. Plaza, C., Brunetti, G., Senesi, N., Polo, A., 2006a. Fluorescence characterization of metal ion-humic acid interactions in soils amended with composted municipal solid wastes. Anal. Bioanal. Chem. 386 (7e8), 2133e2140. Plaza, C., Brunetti, G., Senesi, N., Polo, A., 2006b. Molecular and quantitative analysis of metal ion binding to humic acids from sewage sludge and sludge-amended soils by fluorescence spectroscopy. Environ. Sci. Technol. 40 (3), 917e923. Provenzano, M.R., D’Orazio, V., Jerzykiewicz, M., Senesi, N., 2004. Fluorescence behaviour of Zn and Ni complexes of humic acids from different sources. Chemosphere 55 (6), 885e892. Ryan, D.K., Weber, J.H., 1982. Fluorescence quenching titration for determination of complexing capacities and stability constants of fulvic acid. Anal. Chem. 54, 986e990. Stedmon, C.A., Bro, R., 2008. Characterizing dissolved organic matter fluorescence with parallel factor analysis: a tutorial. Limnol. Oceanogr. Methods 6, 572e579. Terbouche, A., Djebbar, S., Benali-Baitich, O., Bouet, G., 2010. Characterization and complexing capacity of humic acid extracted from Yakouren soil with heavy metals by conductimetry and quenching of fluorescence. Soil Sediment Contam. 19 (1), 21e41. Ward, M.L., Bitton, G., Townsend, T., 2005. Heavy metal binding capacity (HMBC) of municipal solid waste landfill leachates. Chemosphere 60 (2), 206e215. Yamashita, Y., Jaffe, R., 2008. Characterizing the interactions between trace metals and dissolved organic matter using excitation-emission matrix and parallel factor analysis. Environ. Sci. Technol. 42 (19), 7374e7379. Yang, R.J., van den Berg, C.M.G., 2009. Metal complexation by humic substances in seawater. Environ. Sci. Technol. 43 (19), 7192e7197. Zhao, J., Nelson, D.J., 2005. Fluorescence study of the interaction of Suwannee River fulvic acid with metal ions and Al3þ-metal ion competition. J. Inorg. Biochem. 99 (2), 383e396.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 2 0 e1 7 2 8

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Modeling high adsorption capacity and kinetics of organic macromolecules on super-powdered activated carbon Yoshihiko Matsui*, Naoya Ando, Tomoaki Yoshida, Ryuji Kurotobi, Taku Matsushita, Koichi Ohno Graduate School of Engineering, Hokkaido University, N13W8, Sapporo 060-8628, Japan

article info

abstract

Article history:

The capacity to adsorb natural organic matter (NOM) and polystyrene sulfonates (PSSs) on

Received 3 March 2010

small particle-size activated carbon (super-powdered activated carbon, SPAC) is higher

Received in revised form

than that on larger particle-size activated carbon (powdered-activated carbon, PAC).

15 November 2010

Increased adsorption capacity is likely attributable to the larger external surface area

Accepted 15 November 2010

because the NOM and PSS molecules do not completely penetrate the adsorbent particle;

Available online 24 November 2010

they preferentially adsorb near the outer surface of the particle. In this study, we propose a new isotherm equation, the Shell Adsorption Model (SAM), to explain the higher

Keywords:

adsorption capacity on smaller adsorbent particles and to describe quantitatively adsorp-

Diffusion

tion isotherms of activated carbons of different particle sizes: PAC and SPAC. The SAM was

Isotherm

verified with the experimental data of PSS adsorption kinetics as well as equilibrium. SAM

Shell

successfully characterized PSS adsorption isotherm data for SPACs and PAC simulta-

Equilibrium

neously with the same model parameters. When SAM was incorporated into an adsorption

Homogeneous

kinetic model, kinetic decay curves for PSSs adsorbing onto activated carbons of different

surface diffusion

particle sizes could be simultaneously described with a single kinetics parameter value. On

model (HSDM)

the other hand, when SAM was not incorporated into such an adsorption kinetic model and instead isotherms were described by the Freundlich model, the kinetic decay curves were not well described. The success of the SAM further supports the adsorption mechanism of PSSs preferentially adsorbing near the outer surface of activated carbon particles. ª 2010 Elsevier Ltd. All rights reserved.

1.

Introduction

It has been thought that adsorption capacity of activated carbon does not depend on particle size because adsorption occurs in internal pores of activated carbon particles (Letterman et al., 1974; Najm et al., 1990; Peel and Benedek, 1980a and Leenheer, 2007); however, the effect of adsorbent particle size on adsorption capacity has not been examined sufficiently. With decreasing activated carbon particle size, the capacity to adsorb natural organic matter (NOM) has been reported both to not change (Randtke and Snoeyink, 1983) and

to increase (Weber et al., 1983). Possible reasons for these contradictory results have been discussed, but a clear mechanism with supporting experimental evidence has not yet been presented. Recently, very fine (median particle diameters of 0.7 mm) activated carbon particles (super-powdered activated carbon, SPAC) became available through advances in pulverization technology (Matsui et al., 2004, 2005, 2007, 2009a). Thus, it has become possible to reduce adsorbent particle diameter to the submicron range, which is ten times as small as the size of powdered-activated carbon (PAC). SPAC adsorbs NOM

* Corresponding author. Tel./fax: þ81 11 706 7280. E-mail address: [email protected] (Y. Matsui). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.11.020

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 2 0 e1 7 2 8

efficiently because of its high adsorption capacity as well as its kinetic properties. The high adsorption capacity of SPAC raises the issue of adsorption capacity dependence on adsorbent particle size. In the early stages of SPAC research, increased capacity for NOM adsorption on SPAC was thought to be attributable to the mesopore volume increase caused by the fracture of ink-bottle pore structures during pulverization (Matsui et al., 2004). However, it later became apparent that structural changes in pore size distribution attributable to pulverization were not large, and the capacity increase for NOM adsorption could not be explained adequately by an increase in mesopore volume. Instead, it has been proposed that the capacity increase for NOM adsorption on SPAC is due to adsorbate not penetrating completely into the adsorbent particle and preferentially adsorbing in the particle outer region close to the surface of the particle; we also proposed a conceptualization of this concept (Ando et al., 2010). The objective of our current research is to verify and confirm this conceptualization through model analysis. To do so, we have developed adsorption isotherm and kinetic models based on the shell adsorption mechanism, and verified our findings with experimental data by using polystyrene sulfonates (PSSs) as model substances (Karanfil et al., 1996a,b; Li et al., 2003a,b; and Ando et al., 2010).

2.

Materials and methods

2.1.

Activated carbons

Commercially available PAC (Taikou-W, Futamura Chemical Industries Co., Ltd., Gifu, Japan) was used as received (PAC-T) or pulverized in a bead mill (Metawater Co., Ltd., Tokyo, Japan) to achieve four degrees of pulverization; we designated these super-powdered activated carbons as SPACa-T, SPACb-T, SPACc-T, and SPACd-T, in increasing order of particle size. The pore size distributions and the scanning electron micrographs of SPACa-T and PAC-T are presented elsewhere (Ando et al., 2010). Slurries of each activated carbon were prepared in pure water and stored at 4  C and used after dilution and placement under vacuum. Particle size distributions of the five activated carbons were determined by using a laser-light scattering instrument (LA-700, Horiba, Ltd., Kyoto, Japan).

2.2.

1721

We dissolved the PSSs in ultrapure water after the addition of inorganic ions to adjust the ionic strength, and we adjusted the constituent inorganic ions and their concentrations to match those of natural water and the PSS waters used in previous experiments (Table 1S in the supplementary information, Matsui et al., 2004; Ando et al., 2010). All water samples were adjusted to pH 7.0  0.1 by the addition of HCl or NaOH, as required; they were then filtered through 0.2-mm membrane filters (DISMIC-25HP, Toyo Roshi Kaisha, Ltd., Tokyo) before use in experiments. We determined PSS concentrations by UV absorption at a wavelength of 262 nm (UV-1700, Shimadzu Co., Kyoto, Japan).

2.3.

Batch adsorption tests

We conducted PSS-1800 and PSS-1000 adsorption equilibrium tests with all five activated carbons, but we conducted PSS4600 adsorption equilibrium tests only with SPACa-T, SPACdT, and PAC-T. The experimental procedure, described in detail elsewhere (Ando et al., 2010), briefly is as follows. SPAC and PAC slurries were diluted, placed under vacuum, and added to 300-mL solutions containing adsorbate with mixing (Table 2S in the supplementary information). Aliquots (100 mL) were transferred from the 300-mL solutions to 125-mL vials, which were agitated on a shaker for 3 weeks at a constant temperature of 20  C. Control tests were also conducted that did not contain carbon to confirm that concentration changes during long-term mixing were negligible. After filtering the water samples through a 0.2-mm membrane filter, we measured the liquid-phase adsorbate concentrations. We investigated adsorption kinetics of SPACa-T, SPACd-T, and PAC-T by means of batch tests with efficient mixing. Sample water (3 L) containing PSSs was placed in a beaker, and an aliquot (50 mL) was withdrawn from the beaker to determine the initial PSS concentration. After the addition of a specified amount of an activated carbon suspension (Table 3S in the supplementary information), aliquots (50 mL) were withdrawn at intervals and filtered immediately through a 0.2mm membrane filter for determination of PSS concentration.

3.

Results and discussion

3.1.

Shell adsorption model

Water samples

PSSs with various molecular weights (MWs) were selected as model substances instead of NOM because PSSs are chemically homogeneous compounds with known MWs and narrow MW ranges, while NOM is a complex mixture of compounds with unknown composition. Thus, our selection of PSSs makes model analysis of adsorption equilibrium and kinetics clear and unambiguous. We refer to our first PSS formulation as PSS-4600 (Polysciences, Inc., Warrington PA, USA), with a weight-average MW (Mw) of 5180 Da and a number-average MW (Mn) of 4600 Da. Our second PSS formulation, referred to as PSS-1800 (Polysciences, Inc.), had an Mw of 1430 Da and an Mn of 1200 Da. Our final PSS formulation, referred to as PSS1000 (Polymer Standard Service GmbH., Mainz, Germany), had an Mw of 1100 Da and an Mw/Mn of <1.2.

Clearly adsorption capacity for PSS-4600, PSS-1800, and PSS1000 on activated carbon (SPACa-T, SPACb-T, SPACc-T, SPACd-T, and PAC-T) increased with decreasing adsorbent particle size (Fig. 1 and Fig. 1S in the supplementary information). Adsorption sharply increased with increasing equilibrium concentration close to the initial concentration, in particular for PSS-1000. This could be due to the heterogeneity of PSS compounds (Karanfil et al., 1996a; Matsui et al., 1998), despite our assumption of homogeneity for the PSS compounds because of their small-MW ranges. Therefore, the data points for concentrations close to the initial concentration, indicated by red color in Fig. 1, were omitted from our mathematical analysis. Adsorption capacity for all three PSS formulations, as represented by q50, increased with decreasing adsorbent particle size (Fig. 2). PSS adsorption

1722

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 2 0 e1 7 2 8

100

10

100

10 0.1

1 10 Liquid-phase concentration, mg/L

1000

0.1

1 Liquid-phase concentration, mg/L

10

SPACa-T

PSS-1000

SPACb-T SPACc-T SPACd-T

100

PAC-T

Initial conc.

Solid-phase concentration, mg/g

PSS-1800

Initial conc.

Solid-phase concentration, mg/g

1000

PSS-4600

Initial conc.

Solid-phase concentration, mg/g

1000

by SAM equation by Freundlich equation

10 0.1

1 Liquid-phase concentration, mg/L

10

Fig. 1 e Adsorption isotherms of PSS-4600 (upper left panel), PSS-1800 (upper right panel), and PSS-1000 (lower panel). Lines are SAM and Freundlich fits to data (plots close to the initial concentration are red-colored) (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article).

capacity on SPACa-T (d50 ¼ 0.7 mm), which had the smallest particle size, was highest, followed by SPACb-T (d50 ¼ 1.1 mm), SPACc-T (1.9 mm), SPACd-T (3.0 mm), and PAC-T (11.8 mm), in increasing order of particle size: d50 is a volumetric median particle diameter, and q50 is defined as the amount adsorbed on activated carbon in equilibrium with 2.5 mg/L liquid-phase concentration equal to half the initial concentration (5 mg/L) in the adsorption experiment. Ando et al. (2010) hypothesize that the increase in adsorption capacity with decreasing adsorbent particle size is attributable to molecules adsorbing principally in the exterior region close to the external particle surface. The specific external surface area (surface area per unit mass) available for adsorption would be greater for

smaller adsorbent particles, and hence adsorption capacity could be larger on SPAC, which had a much smaller particle size than PAC. In adsorption isotherm model equations, such as the Freundlich equation, amount adsorbed is expressed as mass of adsorbate per unit mass of adsorbent (e.g., Sontheimer et al., 1988). This relationship implicitly assumes that adsorption surface area is proportional to mass of adsorbent and that adsorption capacity is independent of adsorbent particle size. In a previous study (Karanfil et al., 1996a), the Freundlich equation has been employed successfully to describe adsorption isotherms of PSSs, but the effect of adsorbent size was not studied. In our current research, we

Fig. 2 e PSS adsorption capacities represented by q50 against volumetric median diameters of adsorbents. q50 is defined as the amount adsorbed on activated carbon in equilibrium with 2.5 mg/L liquid-phase concentration equal to half of initial concentration (5 mg/L) in the adsorption experiment.

1723

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 2 0 e1 7 2 8

have modified the Freundlich equation, as per Eq. (1), in order to describe adsorption capacity dependence on adsorbent particle size, as follows: qE ¼ KCE1=n

where qE is the amount adsorbed in solid-phase in equilibrium with liquid-phase concentration (mg/g), CE is the liquid-phase concentration (mg/L), K is the Freundlich adsorption capacity parameter (mg/g)/(mg/L)1/n, and n is the Freundlich exponent. Beginning with the Freundlich approach, we have modeled the mechanism of Ando et al. (2010) such that the adsorption capacity parameter K is assumed to decrease with increasing distance from the adsorbent particle surface. Using radial coordinates, the Freundlich adsorption capacity parameter is a function of radial distance and particle radius; adsorption capacity of an adsorbent with radius R at radial distance r is then given by Eq. (2), as follows: 1=n

qS ðr; RÞ ¼ KS ðr; RÞCE

(2)

where r is the radial distance from the center of a PAC particle (mm), R is the adsorbent particle radius (mm), qS(r, R) is the local solid-phase concentration (mg/g) at radial distance r in an adsorbent with radius R, and KS(r, R) is the radially changing Freundlich adsorption capacity parameter (mg/g)/(mg/L)1/n as a function of radial distance r and adsorbent radius R. Spherical particles were assumed for the PAC and the SPACs, which is the conventional practice for adsorption kinetic models (e.g., Sontheimer et al., 1988). Therefore, adsorption capacity of an adsorbent with particle radius R in equilibrium with liquid-phase concentration CE is given by Eq. (3), as shown below: ZR

3r2 1=n 3 qS ðr; RÞ 3 dr ¼ CE 3 R R

0

ZR KS ðr; RÞr2 dr

(3)

0

Accordingly, when the adsorbent size is not uniform, the overall adsorption capacity of the adsorbent is given by Eq. (4), as follows qE ¼

1=n CE

3 R3

Adsorbent particle

(1)

3 2 ZN ZR 4 KS ðr; RÞr2 dr5fR ðRÞdR 0

(4)

0

where qE is the overall adsorption capacity of adsorbent (mg/ g), and fR(R) is the normalized particle size distribution function of adsorbent (mm1). As a model for KS(r, R), we adopted Eq. (5), in which adsorption capacity linearly decreases with distance from the external surface to a depth, d, but thereafter some of the adsorption capacity remains at a level, p, inward from that depth, as depicted in Fig. 3:     rRþd ; 0 ð1  pÞ þ p KS ðr; RÞ ¼ K0 max d

(5)

where K0 is the Freundlich parameter of adsorption at the external particle surface (K0 means solid-phase concentration at r ¼ R at unity equilibrium concentration, (mg/g)/(mg/L)1/n), d is the penetration depth (or thickness of the penetration shell, mm), and p is a dimensionless parameter that defines availability of internal porous structures for adsorption.

1

p Radial distance

Fig. 3 e Schematic diagram of SAM. Molecules adsorb principally in the exterior region (black region in the figure) close to the particle surface, but to some extent do diffuse into the inner region (light gray region in the figure) of an adsorbent particle. KS(r, R)/K0, normalized adsorption capacity relative to the adsorption capacity at the external surface linearly decreases with distance from the external surface to a depth d, (from black to dark gray region) in the figure and thereafter (light gray region) it remains constant as p.

Eq. (5) evolved from the following reasoning: If adsorption occurs only at external particle surface, then adsorption capacity increase is inversely proportional to adsorbent particle size (slope of log q50 vs. log d50 ¼ 1). However, slopes for data points (Fig. 2) range only from 0.34 to 0.58 (less steep than 1), thereby indicating that some of the interior region of the adsorbent particles is available for adsorption. Some adsorbate molecules probably diffuse into and adsorb onto the interior region, while other molecules adsorb onto the exterior region close to the particle outer surface (shell region). The final form of the isotherm equation, referred to hereinafter as the Shell Adsorption Model (SAM) equation, is expressed as Eq. (6) qE ¼

8 9    ZN< ZR  = rRþd max ; 0 ð1  pÞ þ p r2 dr fR ðRÞdR : ; R d

3K0 CE1=n 3

0

0

(6) We have applied this SAM equation to describe isotherm data shown in Fig. 1; in doing so, we sought a single set of isotherm parameter values for K0, n, d, and p in order to obtain the best model fit to data for PSS adsorption isotherms of SPACa-T, SPACb-T, SPACc-T, SPACd-T, and PAC-T. SAM satisfactorily described the experimental data, as shown in Figs. 1 and 2. Our SAM equation is a modified version of the Freundlich equation that is extended so that the slope in the logelog plot of solid-phase concentration vs. liquid-phase concentration is identical for each of the activated carbon preparations. However, experimentally measured slopes for SPACa-T and SPACb-T were actually slightly less steep than those for SPACd-T and PAC-T when applying the Freundlich

1724

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 2 0 e1 7 2 8

equation to data for each activated carbon (see dashed lines in Fig. 1 and Table 1). Because the change in slope after pulverization was not very marked, however, we feel that the SAM approach was successful in providing a first estimate of the dependence of adsorption capacity on particle size.

3.2.

Adsorption kinetics in the shell adsorption model

We analyzed adsorption kinetics data to determine whether incorporation of the SAM equation into an adsorption kinetic model adequately describes the kinetics data. In combining the kinetic model with SAM, we used the pore diffusion model (PDM, e.g., Sontheimer et al., 1988). Although the homogeneous surface diffusion model (HSDM) is more widely used than PDM (Sontheimer et al., 1988), we felt that it could not be applied because it assumes homogeneity inside activated carbon particles. Such homogeneity implies that adsorbed molecules have migrated into adsorbent particles by Fick’s first law of diffusion according to a local solid-phase concentration gradient, and that adsorbate molecules are ultimately distributed evenly along an adsorbent gradient such that local solid-phase concentrations become equal. Such a scenario is inconsistent with SAM. Therefore, instead of HSDM, we used PDM in which migration of molecules in the liquid-filled pores

contributes to transport of adsorbates into particles, while local solid-phase and liquid-phase concentrations in pores remain at equilibrium during the entire period of adsorption (instantaneous adsorption). At an adsorption equilibrium condition in PDM, local liquid-phase concentrations become equal, while local solid-phase concentrations do not necessarily become equal; that condition does not violate SAM. Local adsorption equilibrium is expressed by Eq. (7), as follows: cðt; r; RÞ ¼

n  qðt; r; RÞ KS ðr; RÞ

(7)

where t is adsorption time in the batch system (s); c(t, r, R) is the liquid-phase concentration in an adsorbent particle having radius R at radial distance r and time t (mg/L); and q(t, r, R) is the solid-phase concentration in an adsorbent particle having radius R at radial distance r and time t (mg/g). Diffusion of adsorbate molecules in an adsorbent particle is expressed by Eq. (8), as follows   vqðt; r; RÞ DP 1 v 2 vcðt; r; RÞ ¼ r 2 r r vr vt vr

(8)

where DP is the pore diffusion coefficient (cm2/s); and r is adsorbent particle density (g/L).

Table 1 e Equilibrium and kinetic parameters and ENS values. PSS-4600

Simulation 1

Adsorption equilibrium

SAM K0 ¼ 1.8  102 (mg/g)/(mg/L)1/n 1/n ¼ 0.10 d ¼ 0.22 mm p ¼ 0.038

Adsorption kinetics

PDM DP ¼ 2.9  1010 cm2/s 0.25

ENS

PSS-1800

Simulation 1

Adsorption equilibrium

SAM K0 ¼ 3.2  102 (mg/g)/(mg/L)1/n 1/n ¼ 0.15 d ¼ 0.20 mm p ¼ 0.095

Adsorption kinetics

PDM DP ¼ 7.6  1010 cm2/s 0.85

ENS

PSS-1000

Simulation 1

Adsorption equilibrium

SAM K0 ¼ 2.8  102 (mg/g)/(mg/L)1/n 1/n ¼ 0.21 d ¼ 0.21 mm p ¼ 0.18

Adsorption kinetics

PDM DP ¼ 11.0  1010 cm2/s 0.40

ENS

Simulation 2

Simulation 3

Freundlich K (SPACa-T) ¼ 110 (mg/g)/(mg/L)1/n 1/n (SPACa-T) ¼ 0.064 K (SPACd-T) ¼ 39 (mg/g)/(mg/L)1/n 1/n (SPACd-T) ¼ 0.26 K (PAC-T) ¼ 18 (mg/g)/(mg/L)1/n 1/n (PAC-T) ¼ 0.27 HSDM PDM DS ¼ 3.3  1013 cm2/s DP ¼ 1.7  109 cm2/s 2.1 0.49

Simulation 2

Simulation 3

Freundlich K (SPACa-T) ¼ 190 (mg/g)/(mg/L)1/n 1/n (SPACa-T) ¼ 0.11 K (SPACd-T) ¼ 85 (mg/g)/(mg/L)1/n 1/n (SPACd-T) ¼ 0.28 K (PAC-T) ¼ 45 (mg/g)/(mg/L)1/n 1/n (PAC-T) ¼ 0.28 HSDM PDM DS ¼ 2.1  1013 cm2/s DP ¼ 3.2  109 cm2/s 0.11 0.71

Simulation 2

Simulation 3

Freundlich K (SPACa-T) ¼ 190 (mg/g)/(mg/L)1/n 1/n (SPACa-T) ¼ 0.16 K (SPACd-T) ¼ 97 (mg/g)/(mg/L)1/n 1/n (SPACd-T) ¼ 0.27 K (PAC-T) ¼ 67 (mg/g)/(mg/L)1/n 1/n (PAC-T) ¼ 0.28 HSDM PDM DS ¼ 1.5  1013 cm2/s DP ¼ 3.3  109 0.84 0.083

cm2

/s

1725

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 2 0 e1 7 2 8

Cummurative percentage undersize

By particle size measurement

Discrete approximation

100 80 SPACa-T 60 SPACd-T

40

PAC-T

20 0

0.1

1

10

100

Particle diameter, µm

Fig. 4 e Particle size distributions of SPACa-T, SPACd-T, and PAC-T.

External film balance is described by equating mass balance and mass transfer from the external particle surface to inside the particle, as shown in Eq. (9): 3 2 R Z d4 1 kf 2 r qðt; r; RÞdr5 2 ¼ ½CðtÞ  cðt; R; RÞ r dt R

(9)

0

where kf is the liquid film mass transfer coefficient (cm/s), r is the adsorbent particle density (g/L), and C(t) is the adsorbate concentration in the bulk water phase as a function of time, t (mg/L). When considering adsorbent particle size distribution (Matsui et al., 2003), the mass balance equation for an adsorbate in a batch reactor is given in Eq. (10), as follows: dCðtÞ 3CC kf ¼ r dt

ZN

fR ðRÞ ½CðtÞ  cðt; R; RÞdR R

(10)

0

We approximated particle size distribution of adsorbent by a discrete density function consisting of M size classes, where M is 13, as shown in Fig. 4. We converted the set of model Eqs (5) and (7)e(10) for adsorption in a batch reactor into a set of ordinary differential equations with respect to time, t, using the method of orthogonal collocation. We took many collocation points in an attempt to describe precisely the change of solid-phase concentration in the vicinity of the particle surface (shell region in Fig. 3). When the number of collocation points was 40, the shell region of a PAC particle 11.8 mm in

2 PN  j¼1 Cobs;j  Ccal;j ENS ¼ 1  PN  2 j¼1 Cobs;j  Cave

1.0

SPACd-T Remaining ratio

0.8 0.6 0.4

0.8

Remaining ratio

PAC-T

0.6 0.4 0.2

0.2 0.0 20

40

Contact time, min

60

SPACa-T

0.8

Experimental Simulation1 Simulation2 Simulation3

0.6 0.4 0.2

0.0 0

(11)

where Cobs,j and Ccal,j are the observed and calculated concentrations (mg/L) of adsorbate, respectively; Cave is the average concentration of adsorbate (mg/L); and N is the

1.0

1.0

Remaining ratio

diameter was divided by 6 in the radial direction and that of a SPAC particle 0.7 mm in diameter was divided by 30. Resultant equations were solved as a system of ordinary differential equations by Gear’s stiff method in the IMSL Math Library, after deriving the analytical Jacobian of the equations (Matsui et al., 2009b). Mass transfer resistance across the liquid film external to the adsorbent particle surfaces was substantially neglected because it cannot be the rate-determining step in well-mixed reactors (Sontheimer et al., 1988). In model simulations, the liquid film mass transfer coefficient (kf) was set to 10 cm/s, at which value liquid film mass transfer did not control adsorbate uptake to adsorbent, because any values larger than 10 cm/s yielded the same simulation results for concentration decay curves. Finally, a single value of pore diffusion coefficient Dp, the remaining unknown model parameter in PDM, was sought by using quasi-Newton method in the IMSL Math Library and then the Dp value was determined that produced best-fits to the experimental data for SPACa-T, SPACd-T, and PAC-T under the minimum error criterion [maximizing the ENS value defined by Eq. (11)], as follows:

0.0 0

20

40

Contact time, min

60

0

20

40

60

Contact time, min

Fig. 5 e PSS-4600 adsorption kinetics data and curves fitted with three models (Initial PSS-4600 concentration was 5 mg/L. PAC-T, SPACd-T, and SPACa-T doses were 500, 200, and 100 mg/L, respectively).

1726

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 2 0 e1 7 2 8

0.8 0.6 0.4

SPACd-T

0.8 0.6 0.4 0.2

0.2

0.0

0.0 0

20 40 Contact time, min

Experimental Simulation1 Simulation2 Simulation3

0.6 0.4 0.2 0.0

0

60

SPACa-T

0.8 Remaining ratio

PAC-T Remaining ratio

Remaining ratio

1.0

1.0

1.0

20 40 Contact time, min

60

0

20

40

60

Contact time, min

Fig. 6 e PSS-1800 adsorption kinetics data and curves fitted with models (Initial PSS-1800 concentration was 5 mg/L. PAC-T, SPACd-T, and SPACa-T doses were 200, 50, and 50 mg/L, respectively).

number of data points. ENS values vary between N and 1; a value of 1.0 indicates a perfect fit. As a comparison to the SAM þ PDM model (to be called Simulation 1 hereinafter), we used the Freundlich model þ HSDM (Simulation 2) and the Freundlich model þ PDM (Simulation 3). In these cases, the Freundlich model parameters were individually determined for each activated carbon sample (SPACa-T, SPACd-T, and PAC-T) from the corresponding set of isotherm data. Then, from adsorption kinetics data, a single value for the surface diffusion coefficient DS was sought under the minimum error criterion to simulate experimental data sets for SPACa-T, SPACd-T, and PAC-T (Simulation 2) [Matsui et al., 2003]. A single value of the pore diffusion coefficient DP was sought in Simulation 3. Isotherm model parameter values determined from experimental data of Fig. 1 and the DP and DS values searched are summarized in Table 1. All PSS kinetics curves featured a sharp concentration drop in the first few minutes, followed by a subsequent slower decrease (Figs. 5e7). Experimental data for all PSS kinetics are the best described by the SAM þ PDM model, into which a single DP value was inserted. ENS value for PSS-1000, for example, was 0.40 (Table 1). Freundlich model þ HSDM simulations carried out with one DS value (Simulation 2) did not fit experimental data for activated carbons of small and large size: ENS value was 0.84. Freundlich model þ HSDM simulations underestimate solute uptake rate into large particle-size adsorbent (PAC-T) and overestimate solute uptake rate into small particle size adsorbent (SPACa-T). Simulation 3, carried out with one DP value, also did not adequately describe PSS adsorption kinetics for PAC-T and SPACa-T (ENS value was 0.083). Simulation 1 was also reasonable in terms of diffusivity and MW: the smallest 1.0

PAC-T Remaining ratio

0.8 0.6 0.4

0.8 0.6 0.4

0.2

0.2

0.0

0.0

0

20

40

Contact time, min

60

1.0

SPACd-T Remaining ratio

1.0

Remaining ratio

molecule, PSS-1000, had the highest diffusivity, followed by PSS-1800; and the largest molecule, PSS-4600, had the lowest diffusivity (the DP values of PSS-4600, -1800, and -1000 were 2.9  1010, 7.6  1010, and 11.0  1010 cm2/s, respectively; see Table 1). Such a relationship between diffusivity and MW was not observed in Simulations 2 and 3. In our simulations using the Freundlich model þ HSDM and the Freundlich model þ PDM, we employed six adjustable parameters (2 model parameters times 3 carbons) to describe adsorption isotherms for the three activated carbon preparations: that is, we determined distinct K and 1/n values for SPACa-T, SPACd-T, and PAC-T by linear regression. On the other hand, SAM has four adjustable model parameters for all the three carbons. When considering only the number of adjustable model parameters, the SAM þ PDM model should have two less degrees of freedom in describing a variety of adsorption kinetics than the Freundlich model þ HSDM or the Freundlich model þ PDM. Our results, however, show that SAM þ PDM was more accurate in describing adsorption kinetics than the Freundlich model þ HSDM or the Freundlich model þ PDM. The Freundlich model þ HSDM or Freundlich model þ PDM simulations underestimate solute uptake rate into the large particle-size adsorbent (PAC-T). Implementation of SAM contributed to the solving this underestimation problem by enhancing adsorbate uptake rate. We attribute this enhancement to the fact that most adsorbate molecules do not diffuse into the inner region of adsorbent particles before reaching adsorption equilibrium. Thus, most of the adsorption process is complete close to the exterior particle surface. Therefore, the superiority of the SAM þ PDM model is attributable to the shell adsorption mechanism, and our finding of a better data fit to the SAM þ PDM model offers further evidence that PSS molecules adsorb mostly near the adsorbent

SPACa-T

0.8

Experimental Simulation1 Simulation2 Simulation3

0.6 0.4 0.2

0

20

40

Contact time, min

60

0.0

0

20 40 Contact time, min

60

Fig. 7 e PSS-1000 adsorption kinetics data and curves fitted with models (Initial PSS-1000 concentration was 5 mg/L. PAC-T, SPACd-T, and SPACa-T doses were 200, 50, and 50 mg/L, respectively).

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 2 0 e1 7 2 8

particle surface. We believe that the shell adsorption concept provides the best mechanism for describing adsorption kinetics on activated carbons with various particle sizes. While the discrepancies between the experimental kinetics data and data obtained from Simulation 1 (SAM þ PDM) could be partly due to experimental errors, we note some trends. For most experimental data of PAC-T and SPACd-T, concentrations drop rapidly as adsorption begins. However, initial adsorptions were slower in the simulations. On the other hand, the concentrations of SPACa-T for PSS4600 and -1000 declined faster in the simulations than we observed in our experiments. This pattern of change could be related to local slow diffusion from macropore to micropore, which follows relatively rapid radial diffusion in macropore regions in the adsorbent (Sontheimer et al., 1988; Peel and Benedek, 1980a,b). Adsorption kinetics of a small-MW adsorbate, geosmin, which could adsorb on micropores of SPAC and PAC, is better described by a branched-pore kinetic model (BPKM), in which a slow diffusion mechanism is incorporated into HSDM (Matsui et al., 2009b). We have confirmed that PSS concentration decay curves are also better defined when a slow diffusion mechanism is incorporated into PDM, and the resulting simulation is more successful (data not shown). This improvement may be due to the number of adjustable model parameters for adsorption kinetics increasing from 1 to 3 with implementation of the slow diffusion mechanism. Further research will be necessary to elucidate the slow diffusion mechanism for PSS adsorption. In this study, we have focused on the fact that application of the shell adsorption mechanism, whereby the compounds we studied adsorb mostly in the vicinity of external adsorbent particle surfaces, dramatically improves modeling of adsorption kinetics as well as isotherms. To enhance our understanding of the adsorption mechanism, we have modeled the adsorption of PSSs which are homogeneous compounds with a defined structure, but with a molecular size similar to that of NOM. In future work, adsorption behavior of NOM must be modeled in order to elucidate its adsorption capacity increase with decreasing adsorbent particle size. However, applying SAM to NOM may be difficult, as NOM is an extremely polydisperse mixture, with MWs ranging from hundreds to tens of thousands. Therefore, the parameter values of d and p might vary for the various adsorbates with different properties (including MW) within each NOM solution. The results of the current research may change the paradigm of rapid small-scale column tests (RSSCTs, Crittenden et al., 1986a,b). Our simulation by SAM-PDM implies that adsorption capacity is particle size dependent but that the intraparticle diffusion coefficient is not. The paradigm of SAM-PDM is opposite to that used to scale NOM adsorption in RSSCTs. RSSCTs for NOM adsorption implicitly assume the independence of adsorption capacity from carbon particle size and the proportional diffusivity (PD, the intraparticle diffusion coefficient linearly decreases with particle size). The RSSCT method is well supported by RSSCT data for NOM removal (e.g., Crittenden et al., 1991; Summers et al., 1995). One simple way to reconcile the SAM paradigm with the RSSCT is hypothesising that PAC adsorption capacity is dependent on carbon particle size but that GAC (granular activated carbon) adsorption capacity is not, because GAC has developed

1727

macropores that enable PSS and NOM molecules to penetrate inside of carbon particles and which then equalize carbon capacity regardless of carbon particle size (Ando et al., 2010). The diffusivity issue could be resolved if the SAM-PDM would better fit our experimental data when diffusivity was treated as variable rather than constant. In addition to kinetics, adsorption capacity is a critical parameter that must be considered in the design of RSSCTs (Crittenden et al., 1986a; Sontheimer et al., 1988). Since RSSCTs are conducted on a sieved small-size fraction of crushed carbon particles instead of on the original as-received GAC, which is used in the full-scale adsorber, for proper design of RSSCTs it is essential to understand how not only the adsorption kinetics but also the adsorption capacity is affected by particle size. Actually, capacity increases with decreasing carbon particle size are reported for GACs (Randtke and Snoeyink, 1983; Weber et al., 1983). Moreover, the theoretical background is weak for the PD on which the design of RSSCTs relies. For synthetic organic chemicals (SOCs), on the other hand, isotherm capacities are not affected by carbon particle size (Letterman et al., 1974; Najm et al., 1990 and Leenheer, 2007). The independence of SOC adsorption capacity from carbon particle size is also held for SPAC and PAC (Matsui et al., 2004; Ando et al., 2010). For SOC removals, the RSSCT data well support the assumption of constant diffusivity (Crittenden et al., 1986a). We feel, therefore, that the method of RSSCT design for NOM adsorption could be improved through the study of how NOM adsorption capacity is affected by GAC particle size.

4.

Conclusions

1) We have proposed a shell adsorption mechanism by which PSS molecules are principally adsorbed in the exterior (shell) region of activated carbon particles and adsorbed less in the interior region. The increasing adsorption capacity with decreasing particle size is explained by the increase in specific external surface area (surface area per unit mass) available for adsorption with decreasing adsorbent particle size. Therefore, the PSS adsorption capacity of SPAC was higher than that of PAC. 2) We have proposed a new isotherm equation (SAM), which incorporates the shell adsorption mechanism into the Freundlich model, and we have successfully described PSS adsorption isotherms for SPACs and PAC with the same model parameters. 3) PSS adsorption kinetics were described much better by SAM incorporated into PDM than by the conventional approaches of the Freundlich model þ HSDM or the Freundlich model þ PDM, which further supports our proposed shell adsorption mechanism.

Acknowledgments This study was supported by a Grant-in-Aid for Scientific Research A (21246083) from the Ministry of Education, Science,

1728

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 2 0 e1 7 2 8

Sports, and Culture of the Government of Japan; by a research grant from the Ministry of Health, Labor, and Welfare; and by Metawater Co., Tokyo, Japan.

Appendix. Supplementary information Additional details, including Tables 1S, 2S, and 3S and Figures 1S, are available in the online version at doi:10.1016/j.watres. 2010.11.020.

references

Ando, N., Matsui, Y., Kurotobi, Y., Nakano, Y., Matsushita, T., Ohno, K., 2010. Comparison of natural organic matter adsorption capacities of super-powdered activated carbon and powdered activated carbon. Water Res. 44 (14), 4127e4136. Crittenden, J.C., Berrigan, J.K., Hand, D.W., 1986a. Design of rapid small-scale adsorption tests for a constant diffusivity. J. Water Pollut. Control Fed. 58 (4), 312e319. Crittenden, J.C., Berrigan, J.K., Hand, D.W., Lykins, B., 1986b. Design of rapid small-scale adsorption tests for nonconstant diffusivity. J. Environ. Eng.eASCE 113 (2), 243e259. Crittenden, J.C., Reddy, P.S., Arora, H., Trynoski, J., Hand, D.W., Perram, D.L., Summers, R.S., 1991. Predicting GAC performance with rapid small-scale column tests. J. Am. Water Works Assoc. 83 (1), 77e87. Karanfil, T., Kilduff, J.E., Schlautman, M.A., Weber Jr., W.J., 1996a. Adsorption of organic macromolecules by granular activated carbon. 1. Influence of molecular properties under anoxic solution conditions. Environ. Sci. Technol. 30 (7), 2187e2194. Karanfil, T., Schlautman, M.A., Kilduff, J.E., Weber Jr., W.J., 1996b. Adsorption of organic macromolecules by granular activated carbon. 2. Influence of dissolved oxygen. Environ. Sci. Technol. 30 (7), 2195e2201. Leenheer, J.A., 2007. Progression from model structures to molecular structures of natural organic matter components. Annals of Environ. Sci. 1, 57e68. Letterman, R.D., Quon, J.E., Gemmell, R.S., 1974. Film transport coefficient in agitated suspensions of activated carbon. J. Water Pollut. Control Fed. 46 (11), 2536e2547. Li, Q., Snoeyink, V.L., Maria˜as, B.J., Campos, C., 2003a. Elucidating competitive adsorption mechanisms of atrazine and NOM using model compounds. Water Res. 37 (4), 773e784. Li, Q., Snoeyink, V.L., Marin˜as, B.J., Campos, C., 2003b. Threecomponent competitive adsorption model for flow-through PAC systems. 1. Model development and verification with

a PAC/membrane system. Environ. Sci. Technol. 37 (13), 2997e3004. Matsui, Y., Yuasa, A., Li, F., 1998. Overall adsorption isotherm of natural organic matter. J. Environ. Eng.eASCE 124 (11), 1099e1107. Matsui, Y., Fukuda, Y., Inoue, T., Matsushita, T., 2003. Effect of natural organic matter on powdered activated carbon adsorption of trace contaminants: characteristics and mechanism of competitive adsorption. Water Res. 37 (18), 4413e4424. Matsui, Y., Fukuda, Y., Murase, R., Aoki, N., Mima, S., Inoue, T., Matsushita, T., 2004. Micro-ground powdered activated carbon for effective removal of natural organic matter during water treatment. Water Sci. Tech.: Water Supply 4 (4), 155e163. Matsui, Y., Murase, R., Sanogawa, T., Aoki, N., Mima, S., Inoue, T., Matsushita, T., 2005. Rapid adsorption pretreatment with submicron powdered activated carbon particles before microfiltration. Water Sci. Tech. 51 (6e7), 249e256. Matsui, Y., Aizawa, T., Kanda, F., Nigorikawa, N., Mima, S., Kawase, Y., 2007. Adsorptive removal of geosmin by ceramic membrane filtration with super-powdered activated carbon. J. Water Supply: Res. Technol.eAQUA 56 (6e7), 411e418. Matsui, Y., Hasegawa, H., Ohno, K., Matsushita, T., Mima, S., Kawase, Y., Aizawa, T., 2009a. Effects of super-powdered activated carbon pretreatment on coagulation and transmembrane pressure buildup during microfiltration. Water Res. 43 (20), 5160e5170. Matsui, Y., Ando, N., Sasaki, H., Matsushita, T., Ohno, K., 2009b. Branched pore kinetics model analysis of geosmin adsorption on super-powdered activated carbon. Water Res. 43 (12), 3095e3103. Najm, I.N., Snoeyink, V.L., Suidan, M.T., Lee, C.H., Richard, Y., 1990. Effect of particle size and background natural organics on the adsorption efficiency of PAC. J. Am. Water Works Assoc. 82 (1), 65e72. Peel, R.G., Benedek, A., 1980a. Attainment of equilibrium in activated carbon isotherm studies. Environ. Sci. Technol. 14 (1), 66e71. Peel, R.G., Benedek, A., 1980b. Dual rate kinetic model of fixed bed adsorber. J. Environ. Eng.eASCE 106 (4), 797e813. Randtke, S.J., Snoeyink, V.L., 1983. Evaluating GAC adsorptive capacity. J. Am. Water Works Assoc. 75 (8), 406e413. Sontheimer, H., Crittenden, J.C., Summers, R.S., 1988. Activated Carbon for Water Treatment, second ed. DVGWForschungsstelle, Karlsruhe, Germany. Summers, R. Scott, Hooper, Stuart M., Solarik, Gabriele, Owen, Douglas M., Hong, Seongho, 1995. Bench-scale evaluation of GAC for NOM control. J. Am. Water Works Assoc. 87 (8), 69e80. Weber Jr., W.J., Voice, T.C., Jodellah, A., 1983. Adsorption of humic substances: effects of heterogeneity and system characteristics. J. Am. Water Works Assoc. 75 (12), 612e619.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 2 9 e1 7 3 5

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Experimental and numerical investigations of sedimentation of porous wastewater sludge flocs b  M. Hribersek a,*, B. Zajdela , A. Hribernik a, M. Zadravec a a b

Faculty of Mechanical Engineering, University of Maribor, Slovenia Regional Development Agency Mura Ltd., Slovenia

article info

abstract

Article history:

The paper studies the properties and sedimentation characteristics of sludge flocs, as they

Received 28 July 2010

appear in biological wastewater treatment (BWT) plants. The flocs are described as porous

Received in revised form

and permeable bodies, with their properties defined based on conducted experimental

29 September 2010

study. The derivation is based on established geometrical properties, high-speed camera

Accepted 15 November 2010

data on settling velocities and non-linear numerical model, linking settling velocity with

Available online 20 November 2010

physical properties of porous flocs. The numerical model for derivation is based on generalized Stokes model, with permeability of the floc described by the Brinkman model.

Keywords:

As a result, correlation for flocs porosity is obtained as a function of floc diameter. This

Sludge flocs

data is used in establishing a CFD numerical model of sedimentation of flocs in test

Porous and permeable particle

conditions, as recorded during experimental investigation. The CFD model is based on

Sedimentation

EulereLagrange formulation, where the Lagrange formulation is chosen for computation of

Computational fluid dynamics (CFD)

flocs trajectories during sedimentation. The results of numerical simulations are compared

Multiphase flow

with experimental results and very good agreement is observed. ª 2010 Elsevier Ltd. All rights reserved.

EulereLagrange model

1.

Introduction

Wastewater treatment with the activated sludge is one of the most widespread processes for removing dissoluble substances, small insolubility substances and colloidal organic pollutants from wastewater. In the final phase of the process, the flocs sediment, separating the cleaned water from the accumulated impurities in the sludge. Control of this process is possible only with an in-depth understanding of the sludge floc properties. The activated sludge flocs have a specific geometrical shape, depicted in Fig. 1, and physical properties, which clearly differ from a solid body, presenting a standard geometrical shape in engineering sedimentation correlations. The rate of sedimentation depends on properties of the flocs, such as size, shape, density, permeability, number density and type of wastewater (industrial, municipal). In the

present paper, the main attention is therefore given to determination of size distribution and main geometrical parameters of sludge flocs by means of image analysis. Additionally, free settling tests in connection with empirical models for determination of the drag coefficient (CD) are used in order to evaluate the hydrodynamic properties of flocs. The paper is organized as follows. First experimental determination of size distribution and geometric properties of flocs is described, followed by experimental determination of sedimentation velocities of flocs. The iterative algorithm for derivation of hydrodynamic properties of flocs is presented next, followed by the set up of the CFD numerical model based on the EulereLagrange two-phase flow model. The paper concludes with comparison of experimental results and results of performed numerical simulations and final conclusions.

* Corresponding author. E-mail address: [email protected] (M. Hriber sek). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.11.019

1730

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 2 9 e1 7 3 5

Nomenclature Cd () g (mm/s2) dk (mm) dp (mm) k (m2) Re St vs (mm/s)

drag coefficient gravitational acceleration floc diameter primary particle diameter hydraulic permeability Reynolds number Stokes number settling velocity in a fluid predicted by Stokes’ law vk (mm/s) settling velocity of floc vt (mm/s) velocity of liquid

2.

Experiments

2.1.

Characteristics of activated sludge flocs

The most important physical parameters of sludge flocs are their size distribution, shape, porosity and density. Of these, the porosity is clearly the parameter, that was in the past rarely included in sedimentation correlations. The porosity is a consequence of formation of flocs. They are formed as aggregates of suspended solids, microflocs and primary particles, and extracellular polymers in wastewater, and due to this a floc is composed of a solid structure having a multiscale nature (Gorczyca and Ganczarczyk, 2001; Chu and Lee, 2004a,b). The multiscale structure consists of the first level structuration in the form of large pores and of the second level structuration in the form of fine pores with fractal boundaries. Chung and Lee (2003) have shown, that porosities of flocs are very high, in the range of 98%e99%. In the latter work, sample from municipal wastewater plant was used. On the other end, Gorczyca and Ganczarczyk (2001) reported porosities as low as 8%. In order to have realistic estimations of these parameters experimental analysis on several wastewater samples from a semi-industrial wastewater plant (primarily municipal and pharmaceutical wastewater) was carried out. Since flocs are aggregations of smaller e primary particles, separate experimental analysis gave the average primary particle diameter. A freshly stirred (mixed for 2 h at 600 rpm) and diluted sample (1:20) was analysed under a stereoscopic microscope at a magnification of 500
, (Zajdela et al., 2008). Larger

Fig. 1 e A typical sludge floc, grey image (left) and its binary image (center), schematic sketch of floc structure (right).

3() porosity j () sphericity floc mass mk (g) rk (g/mm3) floc density rt (g/mm3) liquid density rp (g/mm3) primary particle density ht (g/mms) dynamic viscosity of liquid Δr (g/mm3) differences in floc and liquid density U () the ratio of the hydrodynamic drag of permeable and impermeable floc b () dimensionless permeability factor

particles with characteristic fractal floc structure (dp > 5 mm) were ignored, also, all particles smaller than 1 mm were omitted due to limitations of the microscope. Thus, the resulting 70 particles ranked among primary particles and had equivalent circular diameters between 1.077 mm and 4.746 mm, with the average diameter of all particles of 2.019 mm. This value is in accordance with experiments of other authors (Li and Ganczarczyk, 1992; Huang, 1993; Jorand et al., 1995; Lee et al., 1996) (Fig. 1). For determination of other properties, measurement of sludge flocs settling velocity in diluted suspensions was in the center of investigations. The experimental set-up, illustrated in Fig. 2, consisted of a glass tower of the size 330
260
60 mm, filled with distilled water, with dosing device, the Nikon Hi Sense camera and PC computer with image analysis software. Sedimentation velocities of flocs were measured by dropping small amounts (drops) of a previously diluted wastewater sample (1:1) at 20  C into the glass tower where flocs sedimented. The camera was used to shoot flocs movement at a depth of 230 mm, where the flocs have already reached their terminal settling velocity. Image analysis system, written in Java, determined position (coordinates x and y) and sizes of each floc on the image. The smallest size of the floc included in the experimental analysis (0.15 mm) was based on the resolution of the experimental set-up and recommendations

Fig. 2 e Experimental set-up for velocity and size measurements.

1731

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 2 9 e1 7 3 5

imposing a lower limit on the number of pixels in an object (Russ, 1990). Based on a processing of successive images the paths of flocs and their corresponding velocities were calculated. The time-interval between the shots was set to 0.75 s. Using floc image analysis only the observation of the floc’s plane projection was possible. No direct information on the flocs volume necessary for the calculation of flocs diameter and sphericity was obtained. A large number of flocs were, therefore, individually analysed under the microscope at 80
magnification (Zajdela et al., 2008). Each individual floc was rotated as to obtain flocs images in three orthogonal planes. The 3D floc’s shape was approximated by an equivalent cuboid whose projections on the orthogonal planes were equal to those of the floc. In this way, the volume and surface area of the floc could be calculated as well as the sphericity, defined as the ratio of a sphere’s surface area to the surface area of a floc with the same volume, having the value of 0.796. Since during the sedimentation measurements such analysis was impossible, the available 2D projections of flocs were implemented together with the observation, that the flocs oriented themselves in a way that the projection surface in the horizontal plane corresponded to the largest projection of the cuboid, and the side projection, available during velocity measurements, was the smallest projection of the cuboid. From the known cuboid principal axis ratios dimensions of the cuboid were calculated, its surface area and consequently the equivalent circular diameter. The described procedure of flocs diameters determination was compared with a 3D technique, used in the case of geometric analysis of stationary flocs in the previous work (Zajdela et al., 2008), and no significant deviations in results could be established. This was an important conclusion, as it allowed the mutual experimental determination of both settling velocities as well as flocs diameters. Sedimentation velocities, measured in the experiment, were in the range between 0.18 and 2.09 mm/s, see Fig. 3, and the flocs had diameters between 0.15 mm and 1.74 mm, which is in accordance with results of Neale et al. (1973); Tambo and Watanabe (1979); Chien (1994); Johnson et al. (1996); Lee et al. (1996); Wu and Lee (1998) and Li and Yuan (2002). Based on obtained results, particles were grouped into nine size classes, with number fraction distribution as shown in Fig. 4. The experimental data now consisted of a group of 306 particles-flocs with known diameters and corresponding measured settling velocities. In Fig. 4 mean velocities for each

Fig. 4 e Size distribution of sludge flocs and mean velocities with depicted standard deviation.

size class together with standard deviations are depicted. The full set of data served as a starting point for determination of flocs porosity, a key parameter in deriving a set of physical parameters, used in the two-phase flow numerical model of sedimentation. Based on obtained data the Stokes number, defined as St ¼

dk vk rk 18ht

(1)

was calculated. Values for arbitrarily selected flocs are given in Table 1. They range between 0.004 for dk ¼ 0.26 mm and 0.18 for dk ¼ 1.73 mm, which means the flocs are expected to closely follow the fluid flow, when suspended in wastewater. However, in the last phase of wastewater cleaning process, sedimentation characteristics of flocs are the key factor influencing effectiveness of the process. From Table 1 an interesting fact can also be observed, namely there is no monotonic increase of sedimentation velocity with growing flocs diameter. This is a result of flocs formation process, as they are aggregates of microflocs and primary particles, and can have different history of growth, leading to different geometric and sedimentation characteristics.

2.2.

Derivation of sludge flocs porosity

When studying sedimentation characteristics of sludge flocs the main engineering approach is to use simplified force balance on a single floc. In view of the multilevel structure it is clear that the structure of the flocs is not homogeneous.

Table 1 e Stokes number values for selected flocs. dk (mm)

Fig. 3 e Measured settling velocities of sludge flocs.

0.260 0.415 0.581 0.766 0.937 1.078 1.276 1.416 1.736

vk (mm/s)

St

0.28 1.18 0.80 0.87 1.34 1.13 1.09 1.60 1.89

0.0040 0.0274 0.0257 0.0368 0.0693 0.0672 0.0767 0.1249 0.1808

1732

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 2 9 e1 7 3 5

However, when deriving a macroscopic computational approach for the numerical simulation of flocs sedimentation, a simplification in form of a homogeneous flocs structure enables the use of mathematical models based on homogeneous properties of flocs, especially when deciding on the model of the drag force on a floc. The most common model for sedimentation of particles is the Stokes model, originating from force balance of buoyancy, gravity and hydrodynamic drag (Neale et al., 1973; Huang, 1993), i.e.: vs ¼

1
4gðrk  rt Þdk 2 3rt Cd

(2)

From geometrical considerations it is evident, that approximation of a floc as a solid particle can lead to significant errors in evaluation of settling velocities. Due to a porous structure a floc allows liquid flow through its internal structure, thus effecting the drag force excerted on it. Several authors studied this phenomenon, ranging from experimental studies (Chung et al., 2003) to computational studies of the intrafloc flow by using CFD technique (Chung and Lee, 2003; Chu et al., 2005). In Chu et al. (2005) computational results were also used for estimation of flocs permeability. Since the aim of this work is to derive macroscopic particle properties, that can be used in numerical modelling of flocs sedimentation by using Lagrange multiphase model within the framework of CFD, for the case of porous particles with isotropic properties the generalized form of the Stokes model was used as the starting point, i.e.: rk  rt 3r UCd  v2s ¼13¼  t rp  rt 4g r  r dk p

(3)

t

with U being the ratio of the hydrodynamic drag of permeable and impermeable floc. The Brinkman’s extension of Darcy’s law was used due to inability of the latter to describe the viscous force in the boundary layer, which is extremely important in the case of flow through the flocs structure. This is not the only possibility, as there are also several models taking into account the fractal nature of the flocs (Gorczyca and Ganczarczyk, 2002; Chu and Lee, 2004a,b), however they were not considered in this work. In the Brinkman model (Huang, 1993; Zajdela et al., 2008) the U is expressed as, U¼

2b2 ½1  ðtanhðbÞÞ=b 2b2 þ 3½1  ðtanhðbÞÞ=b

(4)

The permeability factor b, dk b ¼ pffiffiffi 2 k

(5)

includes hydraulic permeability of a floc, which is a function of flocs porosity, and was selected according to the Brinkman model as, rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi! 4 8 k¼ $ 3þ 3 3 72 13 13 d2p

(6)

In the equation (3), the porosity of a floc (3) is introduced through the use of the mixing rule in deriving average density of a floc, consisting of primary particles (rp) and liquid (rt). The

value of rp ¼ 1059 g/cm3, (Tambo and Watanabe, 1979; Zajdela et al., 2008), and the value of rt ¼ 998 g/cm3 for the liquid phase were selected. In specifying the hydrodynamic drag force model the model of Chien (1994) for irregularly shaped particles, CD ¼ ð30=ReÞ þ 67:289$eð5:03jÞ

(7)

was used, enabling incorporation of flocs sphericity j. With known experimental data of 306 flocs (depicted in Figs. 3 and 4) and a complete set of model equations, Eqs. (3)e (7), it was possible to derive an iterative computational algorithm for determination of unknown parameter e porosity of flocs. The algorithm was as follows: 1. System with 4 unknowns 3, U, b and k is transformed into system of 2 unknowns. First a set of new parameters is defined: 3r C v2  t d k ¼ x 4g rp  rt dk

(8)

dk ¼y 2

(9)

d2p 72

¼z

(10)

allowing to transform Eqs. (3)e(7) into the following two equations: k ¼ z$ 3 þ

rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi! 4 8 3 3 xU xU

   pffiffiffi 2y2 y  ktanh pyffiffik pffiffiffiffiffi U¼ 2y2 þ 3yk  3 k3 tanhpyffiffik

(11)

(12)

2. The obtained system (11) and (12) depends only on U and k, i.e. F ¼ F(P). 3. Set initial guess for both unknowns; P ¼ P0. 4. The Jacobian (J(Pk)) is calculated. 5. Newton iteration is used in searching the new solution: J(Pk)$DP ¼ F(Pk). 6. The new intermediate solution is computed: Pkþ1 ¼ Pk þ DP. 7. Check of convergence: if converged, go to 8., if not, go to 4. 8. Converged solution for U and k, calculation of 3 with Eq. (3) and b(b ¼ y/Ok). The results of computations are depicted in Figs. 5 and 6. Permeabilities are in the range of 1013e1011 m2. Comparing the results with results of flocs from municipal wastewater of Chu et al. (2005) with values of 1012e1010 m2, values 1012e108 m2 of Chu and Lee (2004a,b), both based on CFD computations, and values 1017e108 m2 of Lee et al. (1996), it could be concluded, that in the average the flocs in our investigation had a slightly more dense internal structure, which is also evident from the obtained porosity values in Fig. 6, ranging from 47% to 96%. One of the possible explanations is the fact, that the tested samples originated from wastewater treatment plant, where wastewater system

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 2 9 e1 7 3 5

1733

Fig. 5 e Permeabilities of sludge flocs.

included also a large pharmaceutical plant, producing smaller particles, that escaped separation processes and were later collected in bioaggregates (Fig. 7). In order to have an algebraic expression for the porosity values, which would later be used for numerical computations, the following sixth order polynomial fit was derived: 3 ¼ 0:53d6k þ3:61d5k 10:01d4k þ14:52d3k 11:74d2k þ5:14dk 0:03 (13) 2

Coefficient of determination R is 0.782, and based on this expression, all other relevant model parameters e permeability, resistance ratio U and density difference rk  rt were evaluated for each individual floc during numerical simulations.

3.

Numerical model

3.1.

EulereLagrange two-phase flow model

Implementation of CFD technique for simulation of sedimentation was used in Latsa et al. (1999), where EulereLagrange model was used and the interphase drag force was applied based on non-porous particle model. In Rao et al. (2002) additional conservation equation for particle phase volume fraction was solved in order to simulate batch sedimentation and viscous resuspension. On the other end, a direct numerical simulation of sedimentation of particle agglomerates by using

Fig. 7 e Comparison of computed settling velocities (CFD) with average experimental velocities and porosity values of representative sludge flocs.

Immersed Boundary Method (Takeuchi et al., 2008) was performed, giving a deeper insight into hydrodynamics of agglomerates. The derived expression for sludge floc porosity enables determination of sedimentation velocity through the use of the Stokes expression, Eq. (2). In the context of deriving a suitable numerical model for implementation in a CFD code it is possible to use this sedimentation velocity directly in Lagrangian particle tracking of a porous sludge floc, by simply adding this velocity to the velocity of the liquid phase. The resulting model, a simplified one-way coupling model, is simple to use, however if the sedimentation velocity of particles is to be influenced by the flow field of the surrounding liquid the velocity has to depend on particle Reynolds number, as this directly influences the drag force. Additionally, in the case of two-way coupling, with consideration of the influence of particles on fluid flow, such approach is not applicable anymore. The second option, direct solution of force balance equation on a sludge floc through EulereLagrange two-phase flow model enables implicit consideration of flow dependent properties of sludge flocs. The second option was implemented through a dedicated user FORTRAN subroutine inside the Ansys-CFX software (Ansys, 2009), enabling effective incorporation of experimental findings inside the particle transport model. Although simulation in the dilute suspension regime was performed, the two-way coupling was implemented, i.e. sedimentation of particles did induce movement of the liquid phase. The numerical model consisted of the Navier-Stokes equations, describing the liquid flow in the Eulerian frame of reference, i.e. of continuity and momentum equations: vð3t rt Þ vð3t rt vti Þ ¼0 þ vt vxi 3t rt

Fig. 6 e Porosities of sludge flocs.

Dvti vp vstji ¼ 3t rt fmi  3t þ þ Fpi Dt vxi vxj

(14)

(15)

with Fpi linking continuous and particle phase. For the particle phase, the Lagrangian frame of reference was used, leading to equations for sludge floc velocity and force balance on a floc:

1734

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 2 9 e1 7 3 5

d! x ¼! vk dt

(16)

* d vk 1 1 3 . . . . ¼  prt d2k UCd jv. mk k  vt j ðvk  vt Þ þ pdk ðrk  rt Þg dt 8 6 (17) As can be seen in the force balance equation buoyancy, gravity and hydrodynamic drag forces were considered, similarly as in the derivation of the generalized Stokes model of sedimentation, Eq. (3). As the particle Reynolds numbers observed during the experiments of sedimentation were below Re < 3, no additional forces were considered. Values of U and rk were all dependent on sludge floc porosity (Eq. (13)) and Cd depended on particle Reynolds number value (Eq. (7)). For the computation of the hydrodynamic drag the particle Reynolds number was defined as: Re ¼

jvk  vt jrt dk j ht

(18)

The numerical model was set up and solved in Ansys-CFX 11.0 (Ansys, 2009).

3.2.

Computational model

In order to match experimental setting with the numerical model the computational domain was based on the geometry of the glass tower (Fig. 2), used in experiments. Dosing device was not included in the model. At the inlet velocity of flocs was set to 0.001 m/s, i.e. the particles started almost from rest. On the side and bottom walls no slip conditions were set for the velocity of the liquid phase, at the top the free slip condition for the interface air-water was imposed. The particles exited computational domain at the bottom wall, i.e. were sedimented on the bottom of the column. For temporal advancement the time step of 0.05 s was chosen, the liquid phase was at rest at the beginning of simulation. Structured grid was used with 22,680 elements. The grid density was determined on the basis of preliminary tests with sedimentation of non-porous solid particles and comparison of results with experimental ones. Since in experimental analysis sludge flocs were divided into nine size classes, the same was done also in the case of numerical simulations. Due to dilute suspension regime, present in experiments, particleeparticle interactions of flocs were neglected, therefore simulations for representative flocs in all nine size classes were performed.

3.3.

Computational results

In order to validate the numerical model one representative particle for each of the nine size classes was selected. With the average diameter of a particle class computation of flocs porosity and density as well as permeability and resistance ratio U was done inside a dedicated module of the CFD software. Time dependent analysis was performed and initial flocs velocities were set, resulting in temporal acceleration or deceleration of a floc, depending on its final settling velocity. In the case of each simulated floc, the sedimentation velocities were achieved within max. 12 time steps. As can be seen form Fig. 7, the established flocs sedimentation velocities were the

lowest for the smallest flocs having low porosity values, on the other hand the largest simulated flocs had the largest sedimentation velocities and highest porosity values, a result also evident from experimental observations (Figs. 3 and 6). The obtained results for sedimentation velocities were compared with results of experiments. Calculated deviations (relative error) between simulation based velocity and average measured sedimentation velocity of a class of flocs, defined as %Deviation ¼

Nf 1 X jvk;n  vk j
100 Nf n¼1 vk

(19)

with Nf number of flocs per size class, vk,n the computed CFD value and vk average experimental sedimentation velocity, were within the 5% limits, which is a clear indication of the correctness of the derived numerical model.

4.

Conclusions

The goal of the presented work was to describe an experimental-computational approach for modelling of settling of sludge flocs, originating from biological wastewater treatment plant. With the simplified flocs geometry model and its relation to the two dimensional camera snapshots of the flocs movement the experimental data on flocs settling velocities together with corresponding flocs equivalent diameters can be obtained. The experimental analysis and the Brinkman model based drag force, used in the generalized form of the Stokes model, resulted in calculated flocs porosity values in the range of 47e96% and hydraulic permeabilities in the range of 1013e1011 m2, indicating weak intrafloc flow. The combination of experimental determination of flocs characteristics and iterative computational procedure for the determination of flocs porosities and permeabilities allows incorporation of different models of the drag force and can serve as a basis for further investigations of settling of porous flocs. The established EulereLagrange two-phase flow CFD model valid for tracking of solid spheres in a fluid was modified by introducing the Brinkman model to the drag force computation. By implementing physical data of flocs, based on performed experiments, an accurate numerical model for the CFD simulation of flocs settling behaviour was constructed. In view of numerical modelling and computer simulations it is important to stress the importance of compatibility of implemented models in the experimental part of the work and related values of physical parameters with implemented models within the CFD framework. The derived numerical model could also serve as a starting point for simulation of settling in dense suspensions, however additional research of interparticle interactions had to be conducted and new models for deformable porous particles need to be developed.

references

Ansys, 2009. ANSYS-CFX-11.0. Chien, S.F., 1994. Settling velocity of irregularly shaped particles. SPE Drilling and Completion, 281e289.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 2 9 e1 7 3 5

Chu, C.P., Lee, D.J., 2004a. Advective flow in a sludge floc. J. Colloid Interface Science 277, 387e395. Chu, C.P., Lee, D.J., 2004b. Multiscale structures of biological flocs. Chem. Eng. Sci. 59, 1875e1883. Chu, C.P., Lee, D.J., Tay, J.H., 2005. Floc model and intrafloc flow. Chem. Engng. Sci. 60, 565e575. Chung, H.Y., Ju, S.P., Lee, D.J., 2003. Hydrodynamic drag force exerted on activated sludge floc at intermediate Reynolds number. J. Colloid Interface Science 263, 498e505. Chung, H.Y., Lee, D.J., 2003. Porosity and interior structure of flocculated activated sludge floc. J. Colloid Interface Science 267, 136e143. Gorczyca, B., Ganczarczyk, J., 2001. Fractal analysis of pore distributions in alum coagulation and activated sludge flocs. Water Qual. Res. J. Can. 36 (4), 687e700. Gorczyca, B., Ganczarczyk, J., 2002. Flow rates through alum coagulation and activated sludge flocs. Water Qual. Res. J. Can. 37 (2), 389e398. Huang, H., 1993. Porosity e size relationship of drilling mud flocs: fractal structure. Clay Miner. 41, 373e379. Johnson, C.P., Li, X., Logan, B.E., 1996. Settling velocities of fractal aggregates. Environ. Sci. Technol. 30, 1911e1918. Jorand, F., Zartarian, F., Block, J.C., Bottero, J.Y., Villemin, G., Urbain, V., Manem, J., 1995. Chemical and strutural (2D) linkage between bacteria within activated sludge flocs. Water Res. 29, 1639e1647. Latsa, M., Assimacopoulos, D., Stamou, A., Markatos, N., 1999. Two-phase modeling of batch sedimentation. Appl. Math. Model. 23, 881e897.

1735

Lee, D.J., Chen, G.W., Hsieh, C.C., 1996. On the free-settling test for estimating activated sludge floc density. Water Res. 30, 541e550. Li, D.-H., Ganczarczyk, J.J., 1992. Advective transport in activated sludge flocs. Wat. Env. Res. 64, 236e240. Li, X., Yuan, Y., 2002. Settling velocities and permeabilities of microbial aggregates. Water Res. 36, 3110e3120. Neale, G., Epstein, N., Nader, W., 1973. Creeping flow relative to permeable spheres. Chem. Engng. 28, 1865e1874. Rao, R., Mondy, L., Sun, A., Altobelli, S., 2002. A numerical and experimental study of batch sedimentation and viscous resuspension. Int. J. Num. Meth. Fluids 39, 465e483. Russ, J.C., 1990. Computer e Assisted Microscopy: The Measurement and Analysis of Images. Plenum Press, New York. Takeuchi, S., Morita, I., Kajishima, T., 2008. Motion of particle agglomerate involving interparticle force in dilute suspension. Powder Technol. 184, 232e240. Tambo, N., Watanabe, Y., 1979. Physical characteristics of flocs. 1. The floc density function and aluminum floc. Water Res. 13, 409e419. Wu, R.M., Lee, D.J., 1998. Hidrodynamic drag force exerted on a moving floc and its implication to free settling test. Water Res. 30, 760e768. Zajdela, B., Hribersek, M., Hribernik, A., 2008. Experimental investigations of porosity and permeability of flocs in the suspensions of biological water treatment plants. J. Mech. Engng 54 (7e8), 547e556.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 3 6 e1 7 4 4

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Solar photo-Fenton degradation of nalidixic acid in waters and wastewaters of different composition. Analytical assessment by LCeTOF-MS Carla Sirtori a,b,c, Ana Zapata a, Wolfgang Gernjak a,d, Sixto Malato a, Antonio Lopez e, Ana Agu¨era c,* a

Plataforma Solar de Almerı´a, CIEMAT, Carretera Sene´s, km 4, 04200 Tabernas, Almerı´a, Spain The Capes Foundation, Ministry of Education of Brazil, PO Box 365, Brası´lia DF 70359-970, Brazil c Pesticide Residue Research Group, University of Almerı´a, 04120 Almerı´a, Spain d The University of Queensland, Advanced Water Management Centre (AWMC), Qld 4072, Australia e Department of Water Research and Technology, CNR-IRSA, Via F. de Blasio 5, Bari 70123, Italy b

article info

abstract

Article history:

This work assessed the solar photo-Fenton degradation of nalidixic acid (NXA), a quinolone

Received 27 July 2010

antibacterial agent, in several different aqueous solutions. It has been proven that the

Received in revised form

composition of the water clearly affects the efficiency of the photo-Fenton process. The

16 November 2010

presence of chlorine ions induces the concurrence of different mechanisms involving Cl Cl2





Accepted 17 November 2010

and

Available online 24 November 2010

were identified and their structures characterized by accurate LCeTOF-MS mass

radicals, which slow down the process. Up to 35 transformation products (TPs)

measurements during treatment of the different model waters. Photocatalytic degradation Keywords:

was thus observed to proceed mainly through the attack of the hydroxyl radicals on the

Liquid chromatography

double bond C(2)]C(3) which induce further ring opening. All the TPs identified persisted

Time-of-flight mass spectrometry

after total degradation of NXA. NXA in real pharmaceutical effluent was treated by photo-

Nalidixic acid

Fenton as a first stage before biological treatment. As NXA has been demonstrated to be

Solar photo-Fenton

recalcitrant to biological treatment, photo-Fenton treatment of the effluent was continued

Transformation products

until its total degradation. Although NXA was efficiently degraded, LCeMS analyses

Wastewater treatment

demonstrated that some of the TPs identified after the photo-Fenton treatment were also recalcitrant to biological treatment, persisting after the combined treatment. These results show that analytical assessment of photocatalytic water treatments is essential to assure they are functioning as intended. ª 2010 Elsevier Ltd. All rights reserved.

1.

Introduction

In recent years, extensive research has focused on improving current knowledge of the presence and fate of pharmaceuticals in the environment (Fatta et al., 2007). The frequent use of pharmaceuticals, in both human and veterinary medicine has

contributed to the increased presence of these compounds in the aquatic environment (Daughton and Ternes, 1999; Kolpin et al., 2002; Tixier et al., 2003). Extremely sensitive analytical methods for determining sub-ppb levels of a great number of pharmaceuticals have been developed and applied, providing valuable information about the environmental occurrence of

* Corresponding author. Tel.: þ34 950015531; fax: þ34 950015483. E-mail address: [email protected] (A. Agu¨era). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.11.023

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 3 6 e1 7 4 4

these recently recognized pollutants (Go´mez et al., 2007a; Vieno et al., 2006; Hernando et al., 2006). Factors such as geographic location, effectiveness of wastewater treatments, proximity to sewage treatment plants and weather conditions influence the concentration levels and fate of pharmaceuticals in natural water (Kasprzyk-Hordern et al., 2008). However, even though reported concentration levels are very low, their well-established effects on the aquatic environment, such as the emergence of antibiotic-resistant bacteria, are negative (Ku¨mmerer, 2009). Many treatments, mainly based on advanced oxidation processes (AOPs), have been applied to industrial and urban wastewater effluents to reduce discharge of such pollutants into the environment (Baumgarten et al., 2007). These processes, by themselves or coupled to conventional biological treatments, have proven their ability to efficiently eliminate or improve the biodegradability of many organic pollutants (Lucas et al., 2007; De La Rochebrochard D’Auzay et al., 2007; Go´mez et al., 2007b; Essam et al., 2007; Gonza´lez et al., 2009). Nalidixic acid (NXA), a non-biodegradable quinolone antibacterial agent widely used for the treatment of urinary tract infections, is one such pollutant that can be successfully treated by photo-Fenton combined with biotreatment (Sirtori et al., 2009a). Nevertheless, very few studies analyze remediation of TPs generated during the combined photo-Fentonbiological treatment. Liquid chromatography combined with mass spectrometry (LCeMS) and related techniques can be considered the most appropriate techniques for increasing information on unknown TPs generated during the treatments and present at low concentrations (Calza et al., 2008;  et al., 2009; Kosjek and Heath, 2008). Radjenovic Analytical identification of TPs presents serious difficulties, mainly associated with the large amount of unknown compounds that may be generated, and the wide differences in their physicalechemical properties and concentration ranges. Differences in polarity affect extraction and chromatographic separation of TPs, and the absence of standards hampers the optimization of both processes, making widespectrum sorbents and generic chromatographic conditions necessary. Despite these difficulties, this kind of study also has some advantages. As TPs derive from a common structure, element-specific information, such as characteristic isotopic signatures, or the presence of common fragmentation patterns, helps to elucidate their structures (Garcı´a-Reyes and Ferna´ndez-Alba, 2007). Furthermore, knowledge of the treatment applied and the reactions that take place also helps to predict the formation of possible TPs. In view of the above, TPs must be identified by analytical techniques able to provide: i) high sensitivity in full scan mode; ii) abundant and highly specific structural information and iii) no limitations on the type or amount of compounds that can be simultaneously analyzed. LC time-of-flight mass spectrometry (LCeTOF-MS) fulfils all the analytical requirements mentioned above and has proven to be a very valuable analytical technique for the identification of unknown compounds (Go´mez et al., 2008). Combination of LCeTOF-MS with other complementary techniques, like GCeMS, has also been reported (Agu¨era et al., 2005). Triple quadrupole-MS (QqQ-MS), Ion TrapMS (IT-MS), hybrid Q-TOF-MS or Q-Linear Ion Trap-MS (QqLITMS) are also of interest for MS/MS or MS3 fragmentation, since

1737

they provide additional structural information for fragment assignment (Eichhorn and Aga, 2004). The aims of this work were to use LCeTOF-MS techniques to increase knowledge of the behaviour of NXA and its main TPs during solar photo-Fenton treatment, evaluate aspects related to water composition and the possibility of combination with biological treatment for the regeneration of real industrial effluents.

2.

Experimental

2.1.

Chemicals

The NXA standard was provided by Fluka (Germany). HPLCgrade methanol was supplied by Merck (Germany). HPLCgrade water used in the analyses was obtained from a Milli-Q ultra-pure water system from Millipore (Milford, MA, USA). Formic acid (purity, 98%) was provided by Fluka. NXA degradation was evaluated in: i) demineralised water (DW e conductivity < 10 mS/cm, Cl 0.2e0.3 mg/L, NO 3 < 0.2 mg/L, organic carbon < 0.5 mg/L) supplied by the Plataforma Solar de Almerı´a (PSA) distillation plant; ii) saline water containing 5 g/L of NaCl (DWNaCl), the same concentration than RIE; iii) simulated industrial effluent (SIE) composed of NXA-45 mg/L, NaCH3COO-2.7 g/L and NaCl-4.5 g/ L and iv) a real industrial effluent (RIE) from a pharmaceutical manufacturer that was characterized as having 775 mg/L Dissolved Organic Carbon (DOC), 3420 mg/L Chemical Oxygen Demand (dissolved and suspended COD), 7 mS/cm conductivity, 45 mg/L NXA, 0.407 g/L TSS, 2.8 g/L Cl, 0.01 g/L PO3 4 , þ 2þ 0.16 g/L SO2 4 , 2 g/L Na and 0.02 g/L Ca . All photo-Fenton experiments were performed using Panreac iron sulphate heptahydrate (FeSO4$7H2O), reagent-grade hydrogen peroxide (30% w/v) and sulphuric acid for pH adjustment. The photo-treated solutions were neutralized by NaOH (reagent-grade, also Panreac) before discharging the photo-treated sample into the bioreactor.

2.2.

Solar photochemical treatment

All solar photochemical experiments were performed in a pilot compound parabolic collector (CPC) plant designed for the purpose. This reactor is composed of two modules with 12 Pyrex glass tubes each, mounted on a fixed platform tilted 37 (local latitude). The total illuminated area is 3 m2 and the volume is 40 L, 22 L of which are irradiated volume. At the beginning of all the photo-Fenton experiments, the solutions studied were added directly to the photoreactor, and a sample was taken after 15 min of homogenization (initial concentration). Then the pH was adjusted with sulphuric acid and another sample was taken after 15 min to confirm the pH. Afterwards, iron salt was also added (FeSO4$7H2O) and wellmixed for 15 min before sampling. Finally an initial dose of hydrogen peroxide was added and samples were taken to evaluate degradation. Photo-Fenton experiments were carried out at a pH adjusted to 2.6e2.8 (H2SO4, 2 N) and Fe2þ concentration of 2 or 20 mg/L. The initial hydrogen peroxide concentration was around 300 mg/L and was kept between 200 and 400 mg/L during the experiments. Solar ultraviolet

1738

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 3 6 e1 7 4 4

radiation (UV) was measured by a global UV radiometer (KIPP & ZONEN, model CUV 4). To compare experiments carried out on different days, the time (t30W) was normalised with an equation described by Malato et al. (2003).

Biological system

An Immobilized Biomass Reactor (IBR) was used for the biological treatment. The IBR consists of a 160-L flat-bottom tank filled with 90e95 L of 15-mm polypropylene Pall Ring supports colonized by activated sludge from the conventional aerobic wastewater treatment plant at El Ejido (Almerı´a, Spain). The system is also provided with a 100-L conditioning tank with pH control connected to the IBR by a recirculation pump. The recirculation flow rate was 500 L/h and the total system operating volume was 150 L. Dissolved oxygen, pH and temperature were automatically measured and registered. The system was operated in batch mode. Total suspended solids, DOC, pH and nitrogen concentration (as ammonium and nitrate) were monitored throughout the adaptation process.

2.4.

Analytical determinations

NXA was analyzed by direct injection in a liquid chromatography HPLCeUV system (Agilent Technologies, series 1100) equipped with a C18 column (LUNA 5 mm, 3  150 mm, from Phenomenex) and operated at a flow rate of 0.5 mL/min. Formic acid 25 mM/methanol 50/50, l ¼ 254 nm used for the isocratic method. Mineralization was followed by measuring the Dissolved Organic Carbon (DOC) by direct injection of filtered samples into a Shimadzu-5050A TOC analyzer with an NDIR detector and calibrated with standard solutions of potassium phthalate. Total iron concentration was monitored by colorimetric determination with 1,10-phenanthroline, following ISO 6332, using a Unicam-2 spectrophotometer. Hydrogen peroxide was analyzed by a fast, simple spectrophotometric method using ammonium metavanadate, which allows the H2O2 concentration to be determined immediately based on a red-orange peroxovanadium cation formed during the reaction of H2O2 with metavanadate, maximum absorption of which is at 450 nm. The peroxide concentrations are 1.0

DOC/DOC

50

0.6

40 OH

30

O

20 N

0 100

t

150

(min)

200

SIE

12 10

0.6 8 6

0.4

4 2

10

0.0 50

14

0.2

N

NXA

0

0.8

DW DW

H O consumed (mM)

60

HC

1.0

H O consumed (mM)

0.8

0.2

To improve current knowledge of the influence of water composition on the efficiency of photo-Fenton treatments, degradation assays were carried out by dissolving NXA in demineralised water (DW), saline water containing 5 g/L of NaCl (DWNaCl) and a simulated industrial effluent (SIE)

70

SIE

O

Results and discussion

3.1. Influence of water content on NXA degradation efficiency

80

DW DW

0.4

3.

[NXA]/[NXA]

2.3.

calculated from absorption measurements by a ratio found by Nogueira et al. (2005). TPs for chromatographic analysis by liquid chromatographyetime-of-flight mass spectrometry (LCeTOF-MS) were extracted from saline water samples by solid-phase extraction, using Oasis HLB cartridges first to reduce the salt content in the matrix and improve intermediate detectability. Cartridges were conditioned with 4 mL of methanol and 3 mL of water, and loaded with a 50 mL aliquot of the sample. The sorbent was washed with 3 mL of Milli-Q water and then eluted with two aliquots of 2 mL of methanol. Extracted samples were diluted (90:10 v/v e water:methanol) before injection. TPs generated during photocatalysis were monitored by LCeTOF-MS (Agilent Technologies) connected to an HPLC Series 1100 system (Agilent Technologies) equipped with a 3-mm  250-mm reverse-phase C18 analytical column, 5-mm particle size (ZORBAX, SB-C18, Agilent Technologies). The mobile phase was a mixture of acetonitrile acidified by 0.1% formic acid (A) and water acidified by 0.1% formic acid (B) at a flow rate of 0.4 mL/min. A linear gradient progressed from 10% A (initial conditions) to 100% A in 50 min, and then remained stable at 100% A for 5 min. The injection volume was 20 mL. This HPLC system was connected to a TOF mass spectrometer (Agilent Technologies) equipped with an electrospray interface operating under the following conditions: capillary 4000 V, nebuliser 40 psig, drying gas 9 L/min, gas temperature 300  C, skimmer voltage 60 V, octapole rf 250 V. Data were processed with MassHunter Workstation Software.

250

0

0.0 0

10

20

t

30

40

(min)

Fig. 1 e DOC reduction (solid dots) and H2O2 consumption (open dots) (left), and NXA degradation (solid dots) and H2O2 consumption (open dots) during photo-Fenton degradation (right). DW and DWNaCl with Fe2D 2 mg/L, SIE with Fe2D 20 mg/L.

1739

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 3 6 e1 7 4 4

to DOC mineralization also rose noticeably with Cl concentration. Furthermore, less active chloride radicals formed could also have reacted with H2O2 (Reactions (4) and (5)), increasing reagent consumption (De Laat and Le, 2006).

solution, with composition as described in the Experimental Section. Different DOC reduction behaviour was observed for each water matrix studied (see Fig. 1; left). In the experiments performed in DW, 86% of the initial DOC was reduced after 92 min of illumination time (t30W), at which point 49.8 mM of H2O2 had been consumed. In the DWNaCl matrix, DOC reduction was 73% after a similar illumination time (107 min) and H2O2 consumption was 75.5 mM. On the other hand, for SIE, after a very long illumination time of 240 min, only 20% of the initial DOC had been eliminated, and 47 mM of H2O2 consumed, which is consistent with the high DOC content compared to DW and DWNaCl. In view of the composition of DWNaCl and SIE solutions, in which the chloride concentration is very high compared to the other inorganic constituents, this may be considered the main inorganic component responsible for the results observed. Photo-Fenton process efficiency (treatment time to achieve a certain goal) has been reported to be noticeably lowered in the presence of chloride ions (Le Truong et al., 2004). There are two reasons for this: (i) decreased generation of hydroxyl radicals because of the formation of chlorideeFe (III) complexes affecting the distribution and reactivity of the iron 2þ species (FeClþ 2 , FeCl ); (ii) scavenging of hydroxyl radicals and formation of Cl and inorganic radicals (Reactions (1)e(3)) less reactive than OH (De Laat and Le, 2006). Thus, as shown in Fig. 1 (left), the DOC mineralization rate was rather lower in DWNaCl than in DW. Hydrogen peroxide consumption related

Cl þ HO / ClOH 

(1)

ClOH  þ Hþ / Cl þ H2O

(2)

Cl þ Cl / Cl2

(3)



















Cl þ H2O2 / HO2 þ Cl þ Hþ

(4)

Cl2 þ H2O2 / HO2 þ 2Cl þ Hþ

(5)













It is worth mentioning that NXA degradation in DW, DWNaCl and SIE solutions did not differ as much as DOC behaviour. The DWNaCl and SIE reaction rates were quite similar, and the effect of so much higher DOC in SIE than in DWNaCl may have been compensated by the different Fe2þ dose. But the main result is related with the consumption of H2O2, which was quite similar in all the three cases (around







OH O

O

O

O

OH N

N

OH O

O

OH

N

N

O

O

O

OH

N

O

N

P2

O

P3

O

OH

O

N

N

O

P4

P34

O

OH

O N

O

O

N

HO

OH

P5

N

OH

N

NH

NXA O

HO

P7

N

OH

P8 N O

OH

N HO N

N

O N

O

OH

OH

N

P11

O

NH OH

P12

N

N

OH

N P6

N H

N

N

N OH

P1

N

NH

P13

O

O

NH

P15

O

OH

OH

P14

O

OH

O

OH P9

O

OH

OH N

N

O N

OH

P10

O

N

O

O

N

OH

H

OH OH

N

P16

P35

Scheme 1 e Proposed photo-Fenton NXA degradation pathway in DW.

NH

O

1740

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 3 6 e1 7 4 4

6e7 mM). This may be due to the action of Cl and Cl2 which are strong oxidants (EoSHE, Cl /Cl ¼ 2.41 V; EoSHE, Cl 2/ 2Cl ¼ 2.09 V) (De Laat and Le, 2006), and if present in high concentrations, could also easily oxidize NXA. Therefore, in high concentrations of Cl, chloride radicals and radicals formed from the reaction between chloride radicals and hydrogen peroxide could partly compensate OH scavenging by Cl. These experiments with DW, DWNaCl and SIE have thus demonstrated not only the difficulties encountered in designing a treatment process based only on experiments with a pure compound in DW, but also how different behaviour is from the parent compound in water containing inorganic salts or any other DOC. In any case, it is worth mentioning that in photo-Fenton, H2O2 consumption, which is easy to follow, could be a good parameter for compensating treatment time uncertainties. NXA degraded rapidly in the experiments performed with 20 mg/L of Fe2þ in SIE, and we expected that at such a high Fe2þ concentration in DW and DWNaCl, we would not be able to follow the process in detail (H2O2 consumption, transformation products). Therefore, NXA degradation experiments in DW and DWNaCl were carried out with an initial iron concentration of 2 mg/L and samples were collected more frequently. As shown in Fig. 1 (right), this procedure allowed NXA degradation, associated with a more representative analysis of the TPs, to be observed, especially in the first stages of the treatment. On the other hand, the high DOC0 in SIE solutions used for simulating RIE, justified the use of 20 mg/L of Fe2þ.





3.2. Identification of transformation products (TPs): analysis by LCeTOF-MS




Samples taken from each experiment were analyzed by LCeTOF-MS to identify the main TPs generated during the photocatalytic degradation of NXA and find out their fate. The detection of new peaks with appearanceedisappearance time profiles indicated possible formation of TPs. The analytical information from accurate mass spectra of the TPs detected, such as measured accurate masses of the protonated molecules [M þ H]þ and in-source CID fragments, the respective errors in ppm and information about the double bonds and rings present in the molecules as provided by the DBE (Double Bond Equivalent) is summarized in Table S1 (Supplementary Material). Furthermore, the small error observed (below 3 ppm in most cases) allowed the elemental composition of the protonated molecules to be assigned with a high degree of accuracy. TP identification was based on the elemental composition for the measured accurate m/z of the protonated molecules and the DBE. Whenever possible, identities were further confirmed by the in-source CID fragmentation pattern. The most common fragments correspond to the unspecific loss of a neutral molecule of H2O yielding an [M þ H  18]þ ion associated with a one-unit increase in the DBE. This loss is characteristic of compounds containing alcohol and carboxylic acid functional groups. Loss of C2H4 and C2H2O were also common and corresponded to the ethyl chain and its oxidized form (observed in compounds P11, P20, P21, P24, P26, P28 and P29). No variation in the DBE was observed in the first, but in




OH

O

O

O

O

O OH

OH

OH O

O

O

OH

O

O

O

O N

N

N

N

OH

N

N

OH

N N

HO

O

O

O

O

O

OH

O

OH N

N

HO

O

N

OH

OH

O

N

P7

N O

P5

O

P4 P34

OH N

N

O

P3

P2

Cl

P17

N

NXA

N

N

O H

N

O OH

N

P22 O

P11

OH

O

OH

O O

OH N

OH

N

P27

O

OH N

P14

NH

P9 P12

OH

O

O OH

N

O

N

N

OH

N

OH

OH N

NH

P15

O

O

OH N

N P6

N H

N

N

OH P1

Scheme 2 e Proposed photo-Fenton NXA degradation pathway in DWNaCl.

P13

NH

1741

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 3 6 e1 7 4 4

the second, the loss of the double bond in the eCH]CHOH chain resulted in a one-unit reduction in the DBE. Losses of 28 (CO) and 44 (CO2) amu (atomic mass unit) indicated the presence of aldehyde and carboxylic acid functionalities. Thus consecutive losses of CO2 and CO observed in P4 led to the loss of the eCOeCOOH side chain. The same fragmentation occurs in P25 and P32. In P23, two consecutive losses of CO confirmed the occurrence of the eCOeCOH side chain. The concurrent loss of water also corroborated the presence of a hydroxyl group, probably in the eNHeCH2OH chain. The detailed study of the chromatograms and accurate mass spectra found during the assays enabled a total of 35 TPs to be identified in the three water systems. The structures proposed are shown in Schemes 1 and 2 and Table 1. The methyl-pyridine ring remained unchanged in all of them, with the main reactions in the 1-ethyl-1,4,5,6-tetrahydro-4oxopyridine-3-carboxylic acid moiety. Compounds identified were often present in several of the water-matrices assayed, suggesting the occurrence of common degradation routes, regardless of the type of water used. For instance, compound P2, with [M þ H]þ at m/z 249.0799 (C12H13N2O4; DBE: 8), the most abundant compound detected in DW, was also detected in the DWNaCl and RIE matrices. Increase by one oxygen atom over NXA (C12H13N2O3) and maintenance in the DBE suggested attack by hydroxyl radicals at the place on the NXA molecule most susceptible to electrophilic attack, i.e., the C(2)]C(3) double bond in Fig. 1 (left). In particular, because the nitrogen atom with its lone electron pair directly bonds to it, carbon (2) is the atom most likely to undergo electrophilic attack. This intermediate (P2) was the precursor of a series of reactions involving ring opening and further oxidation, and loss in the aliphatic chains created. This was identified as the main NXA transformation route (see Fig. S1, Supplementary Material). Other parallel reactions, which involved the oxidation and loss of the ethyl chain and reduction of the carboxyl group, led to compounds P5eP6 and P14eP1, respectively (Scheme 1). Experiments performed in DWNaCl yielded fewer intermediates. Most of them had been previously identified in DW, confirming the occurrence of a similar degradation pathway (Scheme 2). However, the mass spectra showed the formation of a chlorinated by-product (P17), when NXA degradation took place in the saline medium. The elemental composition and isotopic pattern of the [M þ H]þ ion at m/z 283.0419 (C12H12ClN2O4; DBE: 8) confirmed the presence of a chlorine atom in the molecule. Assignment of the structure was proposed based on the fragmentation pattern shown in Fig. S2, Supplementary Material. Indeed, the formation of chlorinated intermediates when AOPs are applied to water containing chloride has been described before by Anipsitakis et al. (2006). According to Lim et al. (2006), in the presence of Cl, some chloroiron complexes, such as [Fe(OH2)5Cl]2þ and [Fe (OH2)4Cl2]þ, may be generated. Later, these complexes are photolysed in the charge-transfer band at 270e400 nm to form [Fe(OH2)6]2þ and Cl , as show in the Reactions (6) and (7):


Table 1 e Additional TPs identified by LCeTOF-MS in SIE solution. Proposal structure

Formula

Experimental mass DBE

O

N

N

O

C11H15N2O3

223.1002

6

C10H15N2O2

195.1056

5

OH C10H11N2O2

191.0736

7

C9H11N2O3

195.0764

6

C9H11N2O3

195.0688

6

C9H11N2O2

179.074

6

C10H13N2O5

241.0743

6

C10H13N2O4

225.0794

6

OH P18

O

N

NH OH P19

O

N

N

P20

O OH N

NH OH P21

O O

N

NH OH P23

O

N

NH OH P24

O

OH O

N

N

OH

OH P25

O OH N

N OH OH

[Fe(OH2)5Cl]2þ þ H2O / [Fe(OH2)6]2þ þ Cl




(6)

P26 (continued on next page)

1742

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 3 6 e1 7 4 4

Table 1 (continued) Proposal structure O

Formula

Experimental mass DBE

OH

HO O N

N

OH

C11H13N2O6

269.0691

7

C9H11N2O3

195.0687

6

OH P28

O OH N

NH OH P29

most abundant by-products. These compounds were detected early on during degradation, with the highest concentration at t30W ¼ 3.5 min. After this point, their presence gradually decays. This behaviour was also observed in compounds P1 or P14, but other TPs, like P7 and P13, were formed and accumulated during phototreatment. Most of the intermediates identified persisted after total degradation of NXA. During the experiments in DWNaCl, P2 abundance was also significant from the beginning of phototreatment (see Fig. S3, Supplementary Material). However, this by-product was totally degraded after 14 min of irradiation. Similar behaviour was observed for P3, which was detected at a high concentration at 8 min, followed by a sharp drop. In contrast, P7 and P13, which represent more advanced stages of oxidation, were the most abundant TPs in the DWNaCl matrix during photo-Fenton. This different behaviour with respect to DW could be explained by the concurrence of different mechanisms, which in DWNaCl, involve the action of Cl and Cl2 radicals, as explained above. Samples were taken during the experiments in SIE after total disappearance of NXA in order to find out how persistent the main intermediates generated were. Experiments in SIE did not reveal the occurrence of alternative pathways in this type of water. Moreover, additional intermediates identified and listed in Table 1 represent intermediate steps in the oxidation route proposed. Results shown in Fig. S4 (left), Supplementary Material, demonstrate that, although all the TPs could be sufficiently reduced or removed by photo-Fenton, a very long treatment time would be required.


O

N

NH

C7H9N2O2

153.0583

5

C9H13N2O3

197.0844

5

OH P30 O OH N

NH OH

3.3. Application to a real industrial effluent (RIE): solar photo-Fenton treatment combined with biological treatment

P31 O

OH

To evaluate the applicability of these studies, a real industrial pharmaceutical effluent (RIE), described in the Experimental Section, was treated. In the “first stage”, the wastewater was treated by solar photo-Fenton until total elimination of NXA (see Fig. 2). This bio-recalcitrant compound was completely degraded after 210 min of illumination time, although only 15% of the original DOC was eliminated and 72 mM of H2O2 were consumed. Such a wide difference in treatment time and H2O2 consumption for NXA degradation

O N

N O P32

C11H13N2O5

253.0746

7

C10H13N2O3

209.0839

6

OH

O O N




N

OH

photo-Fenton

1.0

biotreatment

P33 0.8

70 60 50







However Cl radicals formed from the photo-dissociation reactions are rapidly scavenged by the Cl ions generating less reactive species (Cl2 and ClOH ) and disfavouring the appearance of chlorinated derivatives, which could explain the scarcity of P17. LCeTOF-MS analyses also allowed us to determine the formation and degradation kinetics of the major intermediates generated in the different types of water, and the most important for their abundance and persistence could be identified. In DW (see Fig. S1, Supplementary Material), P2 and P15 were the


0.6

(7)

40

C/C

[Fe(OH2)4Cl2]þ þ H2O / [Fe(OH2)5Cl]þ þ Cl

30

0.4

20




0.2

H O consumption (mM)

DOC/DOC NXA/NXA HO

10 0.0

0 0

1

2

t

(h)

20

40

60

80

100

time (h)

Fig. 2 e Treatment of a real industrial effluent (RIE) during photo-Fenton combined with a biological system.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 3 6 e1 7 4 4

between SIE and RIE was really disappointing, but it is assumed that DOC in RIE was very different from SIE, which drastically reduced photo-Fenton efficiency from the point of view of treatment time required for NXA degradation and efficient use of H2O2 to achieve total elimination of NXA. Toxicity and biodegradability evolution was included in Sirtori et al. (2009b). After solar photo-Fenton treatment, the pH of the phototreated RIE was adjusted to around pH ¼ 7 and added to an Immobilized Biomass Reactor (IBR). Approximately 70% of the initial DOC was removed after 4.5 days of treatment. Inorganic forms of nitrogen analyzed during the biological treatment demonstrated satisfactory ammonium consumption and NO 3 generation, showing that nitrification had been achieved. More details about these results are included in a previous publication (Sirtori et al., 2009b). The fate of the main TPs formed during the treatment of the RIE was also studied. As described above, photo-Fenton treatment was continued until complete NXA degradation. At this moment, the intermediates identified in the water were those shown in Fig. S4 (right), Supplementary Material. When the treated effluent was subjected to biological treatment, P7, the most abundant and persistent by-product detected, was degraded very slowly. P2, P5, and P33 remained almost constant throughout the treatment and P30 was the only intermediate not detected at the end of the experiment, probably because its simpler structure makes it more readily biodegradable.

4.

Conclusions

NXA can be completely degraded in an aqueous solution using solar photo-Fenton, however, water composition can alter degradation, and especially, the mineralization rate, which is slower the more complex the matrix is (DWNaCl and SIE). This behaviour is explained, in this case, by the high concentration of chloride ions present in the water, which reduced the photo-Fenton process efficiency, and emphasizes the need to optimize degradation treatments in real wastewater. Analytical assessment of degradation in the various water-matrices assayed enabled a large number of transformation products and degradation pathways to be identified. The composition of the water does not substantially affect the type of compounds identified, but does affect their relative abundance during the process. Furthermore, experiments in saline water containing high concentrations of chloride ions revealed the possible formation of chlorinated by-products during treatment of this type of water, although in the case of NXA, their low number and abundance showed that this compound was not especially susceptible to such reactions. All the TPs identified persisted after total degradation of NXA, and some of them identified in RIE also turned out to be recalcitrant to biological treatment. Thus, analytical assessment of wastewater treatments, including identification and monitoring of the TPs generated, becomes fundamental to understanding the mechanisms governing degradation, for viable optimization of treatments and to guarantee the quality of the treated water. LCeMS techniques using accurate mass analyzers are the best approach for these purposes.

1743

Acknowledgments The authors wish to thank the European Commission for financial support for the INNOWATECH Project under the Sixth Framework Programme, within the “Global Change and Ecosystems Programme” (Contract no.: 036882). They also wish to thank Mrs. Deborah Fuldauer for English language correction. Carla Sirtori and Ana Zapata thank the Capes Foundation e Brazilian Ministry of Education and the Spanish Ministry of Science and Innovation, respectively, for their Ph.D. research grants.

Appendix. Supplementary data Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.watres.2010.11.023.

references

Agu¨era, A., Pe´rez Estrada, L.A., Ferrer, I., Thurman, E.M., Malato, S., Ferna´ndez-Alba, A.R., 2005. Application of time of flight mass spectrometry to the analysis of natural sunlight photodegradation products of diclofenac in water. Journal of Mass Spectrometry 40, 908e915. Anipsitakis, G.P., Dionysiou, D.D., Gonzalez, M.A., 2006. Cobaltmediated activation of peroxymonosulfate and sulfate radical attack on phenolic compounds. Implications of chloride ions. Environmental Science & Technology 40, 1000e1007. Baumgarten, S., Schro¨der, H.Fr., Charwath, C., Lange, M., Beier, S., Pinnekamp, J., 2007. Evaluation of advanced treatment technologies for the elimination of pharmaceutical compounds. Water Science & Technology 56, 1e8. Calza, P., Medana, C., Carbone, F., Giancotti, V., Baiocchi, C., 2008. Characterization of intermediate compounds formed upon photoinduced degradation of quinolones by highperformance liquid chromatography/high-resolution multiple-stage mass spectrometry. Rapid Communications in Mass Spectrometry 22, 1533e1552. Daughton, C.G., Ternes, T.A., 1999. Pharmaceuticals and personal care products in the environment: agents of subtle change? Environmental Health Perspectives 107, 907e938. De La Rochebrochard D’Auzay, S., Brosillon, S., Fourcade, F., Amrane, A., 2007. Integrated process for degradation of amitrole in wastewaters: photocatalysis/biodegradation. International Journal of Chemical Reactor Engineering 5, A51. De Laat, J., Le, T.G., 2006. Effects of chloride ions on the iron(III)catalyzed decomposition of hydrogen peroxide and on the efficiency of the Fenton-like oxidation process. Applied Catalysis B: Environmental 66, 137e146. Eichhorn, P., Aga, D.S., 2004. Identification of a photooxygenation product of chlortetracycline in hog lagoons using LC/ESI-ion trap-MS and LC/ESI-time-of-flight-MS. Analytical Chemistry 76, 6002e6011. Essam, T., Aminb, M.A., El Tayeb, O., Mattiassona, B., Guieysse, B., 2007. Solar-based detoxification of phenol and p-nitrophenol by sequential TiO2 photocatalysis and photosynthetically aerated biological treatment. Water Research 41, 1697e1704. Fatta, D., Achilleos, A., Nikolaou, A., Meric¸, S., 2007. Analytical methods for tracing pharmaceutical residues in water and wastewater. TrAC, Trends in Analytical Chemistry 26, 515e533. Garcı´a-Reyes, J.F., Ferna´ndez-Alba, A.R., 2007. Identification of pesticide transformation products in food by liquid

1744

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 3 6 e1 7 4 4

chromatography/time-of-flight mass spectrometry via “fragmentationedegradation” relationships. Analytical Chemistry 79, 307e321. Go´mez, M.J., Martı´nez Bueno, M.J., Lacorte, S., Ferna´ndez-Alba, A. R., Agu¨era, A., 2007a. Pilot survey monitoring pharmaceuticals and related compounds in a sewage treatment plant located on the Mediterranean coast. Chemosphere 66, 993e1002. Go´mez, C., Rodriguez, J., Freer, J., Lizama, C., Zaror, C., Mansilla, H.D., 2007b. Coupling of photocatalytic and biological reactors to remove EDTA-Fe from aqueous solution. Environmental Technology 28, 123e127. Go´mez, M.J., Sirtori, C., Mezcua, M., Ferna´ndez-Alba, A.R., Agu¨era, A., 2008. Photodegradation study of three dipyrone metabolites in various water systems: identification and toxicity of their photodegradation products. Water Research 42, 2698e2706. Gonza´lez, O., Esplugas, M., Sans, C., Torres, A., Esplugas, S., 2009. Performance of a sequencing batch biofilm reactor for the treatment of pre-oxidized sulfamethoxazole solutions. Water Research 43, 2149e2158. Hernando, M.D., Mezcua, M., Ferna´ndez-Alba, A.R., Barcelo´, D., 2006. Environmental risk assessment of pharmaceutical residues in wastewater effluents, surface waters and sediments. Talanta 69, 334e342. Kasprzyk-Hordern, B., Dinsdale, R.M., Guwy, A.J., 2008. The occurrence of pharmaceuticals, personal care products, endocrine disruptors and illicit drugs in surface water in South Wales, UK. Water Research 42, 3498e3518. Kolpin, D.W., Furlong, E.T., Meyer, M.T., Thurman, E.M., Zaugg, S. D., Barber, L.B., Buxton, H.T., 2002. Pharmaceuticals, hormones, and other organic wastewater contaminants in U.S Streams, 1999e2000: a national reconnaissance. Environmental Science & Technology 36, 1202e1211. Kosjek, T., Heath, E., 2008. Applications of mass spectrometry to identifying pharmaceutical transformation products in water treatment. TrAC, Trends in Analytical Chemistry 27, 807e820. Ku¨mmerer, K., 2009. Antibiotics in the aquatic environment e a review e part II. Chemosphere 75, 435e441.

Le Truong, G., De Laat, J., Legube, B., 2004. Effects of chloride and sulfate on the rate of oxidation of ferrous ion by H2O2. Water Research 38, 2384e2394. Lim, M., Chiang, K., Amal, R., 2006. Photochemical synthesis of chlorine gas from iron(III) and chloride solution. Journal of Photochemistry and Photobiology A: Chemistry 183, 126e132. Lucas, M., Dias, A.A., Sampaio, A., Amaral, C., Peres, J.A., 2007. Degradation of a textile reactive azo dye by a combined chemical-biological process: Fenton’s reagent yeast. Water Research 41, 1103e1109. Malato, S., Blanco, J., Vidal, A., Alarco´n, D., Maldonado, M.I., Ca´ceres, J., Gernjak, W., 2003. Applied studies in solar photocatalytic detoxification: an overview. Solar Energy 75, 329e336. Nogueira, R.F.P., Mirela, C.O., Paterlini, W.C., 2005. Simple and fast spectrophotometric determination of H2O2 in photoFenton reactions using metavanadate. Talanta 66, 86e91. , J., Petrovic , M., Barcelo´, D., 2009. Complementary Radjenovic mass spectrometry and bioassays for evaluating pharmaceutical-transformation products in treatment of drinking water and wastewater. TrAC, Trends in Analytical Chemistry 28, 562e580. Sirtori, C., Zapata, A., Oller, I., Gernjak, W., Agu¨era, A., Malato, S., 2009a. Solar photo-Fenton as finishing step for biological treatment of a pharmaceutical wastewater. Environmental Science & Technology 43, 1185e1191. Sirtori, C., Zapata, A., Oller, I., Gernjak, W., Agu¨era, A., Malato, S., 2009b. Decontamination industrial pharmaceutical wastewater by combining solar photo-Fenton and biological treatment. Water Research 43, 661e668. Tixier, C., Singer, H.P., Oellers, S., Muller, S.R., 2003. Occurrence and fate of carbamazepine, clofibric acid, diclofenac, ibuprofen, ketoprofen and naproxen in surface waters. Environmental Science & Technology 37, 1061e1068. Vieno, N.M., Tuhkanen, T., Kronberg, L., 2006. Analysis of neutral and basic pharmaceuticals in sewage treatment plants and in recipient rivers using solid phase extraction and liquid chromatographyetandem mass spectrometry detection. Journal of Chromatography B 1134, 101e111.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 4 5 e1 7 5 1

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Source water quality effects on monochloramine inactivation of adenovirus, coxsackievirus, echovirus, and murine norovirus Amy M. Kahler a,*, Theresa L. Cromeans b,c, Jacquelin M. Roberts d, Vincent R. Hill a a

Centers for Disease Control and Prevention, National Center for Emerging and Zoonotic Infectious Diseases, Division of Foodborne, Waterborne, and Environmental Diseases, 1600 Clifton Road, Mail Stop D-66, Atlanta, GA 30329, USA b Atlanta Research and Education Foundation, Decatur, GA, USA c Centers for Disease Control and Prevention, National Center for Immunization and Respiratory Diseases, Division of Viral Diseases, Atlanta, GA, USA d Centers for Disease Control and Prevention, Center for Global Health, Division of Parasitic Diseases and Malaria, Atlanta, GA, USA

article info

abstract

Article history:

There is a need for more information regarding monochloramine disinfection efficacy for

Received 3 August 2010

viruses in water. In this study, monochloramine disinfection efficacy was investigated for

Received in revised form

coxsackievirus B5 (CVB5), echovirus 11 (E11), murine norovirus (MNV), and human

15 October 2010

adenovirus 2 (HAdV2) in one untreated ground water and two partially treated surface

Accepted 19 November 2010

waters. Duplicate disinfection experiments were completed at pH 7 and 8 in source water

Available online 24 November 2010

at concentrations of 1 and 3 mg/L monochloramine at 5 and 15
C. The Efficiency Factor Hom (EFH) model was used to calculate CT values (mg-min/L) required to achieve 2-, 3-,

Keywords:

and 4-log10 reductions in viral titers. In all water types, monochloramine disinfection was

Disinfection

most effective for MNV, with 3-log10 CT values at 5
C ranging from 27 to 110. Mono-

Monochloramine

chloramine disinfection was least effective for HAdV2 and E11, depending on water type,

Water

with 3-log10 CT values at 5
C ranging from 1200 to 3300 and 810 to 2300, respectively.

Virus

Overall, disinfection proceeded faster at 15
C and pH 7 for all water types. Inactivation of the study viruses was significantly different between water types, but there was no indication that overall disinfection efficacy was enhanced or inhibited in any one water type. CT values for HAdV2 in two types of source water exceeded federal CT value recommendations in the US. The results of this study demonstrate that water quality impacts the inactivation of viruses and should be considered when developing chloramination plans. Published by Elsevier Ltd.

1.

Introduction

Disinfection is a critical step in the drinking water treatment process to reduce infectious virus concentrations, and chlorine is the most widely used disinfectant in the United States (AWWA, 2008). Monochloramine, which is formed by combining an ammonia source with free chlorine, is * Corresponding author. Tel.: þ1 404 718 4153; fax: þ1 404 718 4197. E-mail address: [email protected] (A.M. Kahler). 0043-1354/$ e see front matter Published by Elsevier Ltd. doi:10.1016/j.watres.2010.11.026

sometimes used to maintain a disinfectant residual in the distribution system because it is more stable than free chlorine and can help to minimize biofilm growth. However, monochloramine use in the US may increase as a result of the regulations set forth by the Stage 2 Disinfectants and Disinfection Byproducts Rule (Stage 2 DBPR) (USEPA, 2006b). The Stage 2 DBPR regulates the maximum levels of disinfection

1746

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 4 5 e1 7 5 1

byproducts (DBPs) in drinking water. Both chlorine and monochloramine produce DBPs as a result of their interaction with organic and inorganic matter in water, but monochloramine is less reactive than chlorine and produces fewer regulated DBPs and at lower concentrations than chlorine. In addition, the Long Term 2 Enhanced Surface Water Treatment Rule (LT2ESWTR) (USEPA, 2006a) seeks to reduce the incidence of disease associated with pathogenic microorganisms in drinking water, particularly Cryptosporidium. Because Cryptosporidium oocysts are highly resistant to free chlorine, some water utilities may need to implement alternative disinfectants in order to comply with the LT2ESWTR treatment requirements. In its Guidance Manual for Compliance with the Filtration and Disinfection Requirements for Public Water Systems Using Surface Water Sources (Guidance Manual), the US Environmental Protection Agency (USEPA) recommended CT values of 1423 and 712 to achieve 3-log10 inactivation of viruses with monochloramine at pH8 and 5 and 15
C, respectively (USEPA, 1990). The recommended CT values do not include a factor of safety, under the assumption that addition of ammonia will follow a period of free chlorine exposure, which could dramatically lower concentrations of viruses that may be resistant to monochloramine. However, CT value requirements for viruses using preformed monochloramine should be evaluated in order to determine whether they are sufficient for water treatment systems in which ammonia is added before chlorine, they are combined simultaneously, or there is an intrusion event into a distribution system where monochloramine is present. The Guidance Manual CT value recommendations were obtained from disinfection experiments conducted with monodispersed hepatitis A virus (HAV) in buffered, reagent-grade water (RGW). Previous researchers found that chlorine disinfection of viruses in natural waters was significantly different than in RGW. Haas et al. (1996) found that the inactivation rates of MS2 were lower in surface waters than in RGW. The authors suggested that increased turbidities in the surface waters may have afforded some protection to the virions and resulted in the decreased inactivation rates. Similarly, Thurston-Enriquez et al. (2003a) found that higher CT values were required for inactivation of AdV40 in treated ground water and that it may have been due to ground water constituents that protected the viral particles via adsorption or enhanced aggregation. In another study, inactivation rates for HAdV2, E11, CVB5, and MNV in three source waters were significantly different between water types and compared to RGW, but could not be correlated to any measured water quality parameters (Kahler et al., 2010). Because water quality can impact chlorine disinfection of viruses, it is possible that monochloramine inactivation rates may also be different for RGW and natural waters. While several studies have examined the disinfection efficacy of monochloramine for viruses in RGW (Baxter et al., 2007; Cromeans et al., 2010; Sirikanchana et al., 2008; Sobsey et al., 1991, 1988), there is little information in the literature regarding the disinfection efficacy of monochloramine in natural waters. The objective of this study was to examine the disinfection efficacy of monochloramine on selected viruses from USEPA’s Contaminant Candidate List (CCL2) (USEPA, 2005) in one untreated ground water and two partially treated surface

waters from distinct geographical regions. The impact of water quality was examined by comparing the inactivation rates of the study viruses in each water type, as well as to disinfection efficacy in RGW from a previous study (Cromeans et al., 2010).

2.

Materials and methods

2.1.

Virus propagation and assay

Clones of CVB5 (Faulker strain) and E1 (Farouk strain) were prepared from strains obtained from the American Type Culture Collection (ATCC, Manassas, VA) and propagated in BGM cells (Scientific Resources Program, CDC). MNV-1 was obtained from Karst et al. (2003) and propagated in RAW 264.7 cells obtained from ATCC. HAdV2 (strain 6) was obtained from CDC and propagated in A549 cells (Scientific Resources Program, CDC). Cell lines were maintained in either Eagle’s Minimum Essential Medium (EMEM) or Dulbecco’s Modified Eagle Medium (DMEM) as described previously (Cromeans et al., 2010). Viral titers were determined by plaque assay by inoculating 10-fold dilutions onto cell monolayers in 60 mm2 dishes. After 1 h adsorption at 37
C and 5% CO2, the infected cells were overlaid with 5 mL maintenance medium (2) containing 0.5% agarose. Following a 2-day incubation of MNV and enterovirus assays and a 5-day incubation of HAdV2, a second overlay containing 2% neutral red was added to visualize plaques within 4 h.

2.2.

Cell associated virus (CAV) preparation

CAVs were prepared as described previously (Cromeans et al., 2010). Cell monolayers were infected at a multiplicity of infection of 0.5e1.0 and cultured in serum free medium until maximum virus titer was obtained based on replication studies of each virus. The culture medium was removed and replaced with chlorine-demand-free Dulbecco’s PBS (CDF DPBS) before freezing at 70
C. The CAV was purified by polyethylene glycol precipitation and chloroform extraction and the purified CAV (pCAV) was used on the same day as the experimental inoculum.

2.3.

Reagents and glassware

CDF DPBS and CDF water were prepared according to Standard Method 4500-Cl C (APHA, 2005). A monochloramine stock solution was made by mixing equal volumes of 200 mg/L free chlorine and 800 mg/L ammonium chloride in pH 8 CDF water and was stored at 4
C for 2 wk. Prior to each experiment, this stock was added to the experimental waters to achieve 1 or 3 mg/L monochloramine. Monochloramine residual was measured on a Hach DR/850 colorimeter using Hach Monochlor-F reagent pillows (Hach, Loveland, CO). CDF glassware was prepared by soaking in 5 mg/L free chlorine overnight. The glassware was rinsed 5 times with CDF water, covered with clean foil, and baked at 200
C for 2 h. All glassware and water was pre-chilled at 5 or 15
C before each experiment.

2.4.

Test waters

Partially treated source water samples were obtained from Cobb County-Marietta Water Authority (CCMWA) in Marietta,

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 4 5 e1 7 5 1

Georgia, and Washington Aqueduct (WASH) in Washington D.C water treatment plants, just prior to chemical disinfection. Source water from CCMWA was collected from a reservoir/recreational lake and subjected to coagulation and filtration. Source water from WASH was collected from the Potomac River and subjected to pre-sedimentation, coagulation, flocculation, sedimentation and filtration. Ground water was obtained from Brunswick-Glynn County Joint Water and Sewer Commission (BGC) in Savannah, Georgia prior to chlorination. Source water samples were shipped to the CDC laboratory in Atlanta, Georgia in cubitainers, and stored at 20
C in 2-L aliquots. To ensure that any naturally present chlorine residual did not interfere with the monochloramine reaction, source waters were made chlorine-demand-free. The source waters were thawed, adjusted to a free chlorine residual of 2 mg/L and left overnight. Following this procedure, the water was placed under UV light until no chlorine residual remained. On the day of an experiment, enough monochloramine stock solution was added to achieve the desired monochloramine residual. Immediately prior to an experiment the water was adjusted to pH 7 or 8 by addition of 1 M NaH2PO4 or Na2HPO4, respectively. The total volume of phosphate buffer needed to adjust the pH of the water was negligible and did not significantly alter the buffering capacity of the water.

2.5.

Experimental protocol

Source water experiments were conducted in duplicate at 5 and 15
C using 1 and 3 mg/L monochloramine. Experiments were conducted in a recirculating water bath inside a biological safety cabinet. A multi-place stir plate placed under the water bath allowed for continual mixing during experiments. For experiments lasting longer than several hours, samples were moved to a refrigerator or an environmental chamber maintained at 5 and 15
C, respectively. For each pH, five 50-mL Erlenmeyer flasks were used, each containing 40 mL source water with either 1 or 3 mg/L monochloramine. Two flasks served as the experimental replicates, and one flask each was used to monitor monochloramine residual, pH, and viral titer throughout the course of the experiment. At time zero, 1 mL of a pCAV stock was inoculated into each flask. The concentration of pCAV stocks ranged from 4  107e1  1010 PFU/mL for HAdV2 experiments, 5  108e2  109 PFU/mL for CVB5 experiments, 1  106e2  107 PFU/mL for E11 experiments, and 2  107e5  108 PFU/mL for MNV experiments. At pre-designated intervals, a 5-mL sample was removed and the monochloramine residual was quenched with 50 mg/L sodium thiosulfate. Monochloramine residual was measured immediately before an experiment, immediately after virus inoculation, at the midpoint, and at the end of an experiment, at a minimum. Experiments targeting a starting monochloramine concentration of 1 mg/L had actual starting concentrations between 0.93 and 1.30 mg/L. Experiments targeting a starting monochloramine concentration of 3 mg/L had actual starting concentrations between 2.70 and 3.26 mg/L. The average drop in residual after inoculating water with HAdV2, CVB5, E11, and MNV was 0.1, 0.08, 0.08, and 0.04, respectively. After inoculation, the monochloramine level did not drop more than 0.19 mg/L during any experiment. The pH was measured

1747

at every sampling point during an experiment. Prior to virus inoculation in the viral titer control flask, 50 mg/L sodium thiosulfate was added to quench the monochloramine residual. This flask was sampled immediately after virus inoculation and at the end of the experiment to ensure that virus infectivity was stable in the source water. There were no appreciable titer losses for the study viruses in any of the source waters matrices. Samples were held at 4
C in PBS or EMEM with 1% serum until assay.

2.6.

Kinetic modeling and CT calculations

Viral inactivation was determined by calculating the survival ratio (N/N0; infectious viruses at time t divided by infectious viruses at time zero) for each experimental sample. The EFH model was used to calculate predicted survival ratios based on experimental conditions, including disinfectant decay over time using a first-order kinetic equation (Haas and Joffe, 1994). Samples were included in the EFH modeling only if the plaque assay counts averaged 10 PFU/plate. Inactivation curves were created using Microsoft Excel to compare observed versus predicted survival ratios. CT values (chlorine residual in mg/L  contact time in min) were calculated for 2-, 3-, and 4-log10 inactivation for each virus and condition through application of the EFH model and averaged to obtain one CT value per experiment. Because the source waters were not buffered, there was a tendency for the pH to drift from the starting levels of 7 and 8 during the extended course of experiments, which lasted up to 72 h. Statistical comparisons were conducted only for experiments that were maintained within an acceptable pH range (6.75e7.25 for pH 7 and 7.75e8.25 for pH 8). Quadratic regression was used to compare viral inactivation between different viruses, water types, and pH levels using SAS version 9.0. Statistical significance was set at a ¼ 0.05.

3.

Results and discussion

3.1.

Relative monochloramine efficacy for study viruses

Water quality characteristics of the source waters are presented in Table 1. As expected, specific conductance (a measure of ionic strength) was highest for BGC ground water and lowest for CCMWA lake water. While turbidity was slightly lower in WASH water compared to CCMWA and BGC waters, TOC in BGC water was substantially higher than in the other two source waters. Ionic strength, particulate matter, and organic content are several water quality factors that may affect chemical disinfection efficacy (Barbeau et al., 2005; Berg et al., 1989b; Haas et al., 1996). CT values for 3-log10 inactivation of the study viruses in each of the source waters are shown in Tables 2e4 for experiments in which viruses were exposed to 1 or 3 mg/L monochloramine. Those experiments which met the criteria for statistical comparison (i.e., pH drift did not exceed 0.25 pH units) are annotated in each table. In all water types, monochloramine disinfection was most effective for MNV ( p < 0.0001), with 3-log10 CT values ranging from 27 to 110 at 5
C. In CCMWA water, disinfection was least

1748

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 4 5 e1 7 5 1

Table 1 e Water quality characteristics of the source waters. Water Source

pH

Turbidity (NTU)

Free chlorine (mg/L)

Specific conductance (mS/cm at 25
C)

Total hardness (mg/L as CaCO3)

Alkalinity (mg/L as CaCO3)

TOC (mg/L as C)

CCMWA WASH BGC

5.6 7.1 7.9

0.55 0.17 0.60

0.03 NT BDL

92 370 500

22 140 200

16 82 120

2.2 1.9 18

NT: Not tested, BDL: Below detection limit.

effective for E11 ( p < 0.0001), with 3-log10 CT values ranging from 810 to 2300 at 5
C, and CVB5 was always inactivated more quickly than HAdV2 ( p < 0.0001). In WASH and BGC waters, disinfection was least effective for HAdV2 ( p < 0.0001), with 3-log10 CT values ranging from 1200 to 3300 at 5
C. CVB5 was always inactivated more quickly than E11 in BGC water ( p < 0.0001), but the there was no consistent indication that disinfection was more or less effective for CVB5 or E11 in WASH water. No previous research has been conducted to examine the disinfection efficacy of monochloramine in natural waters, but several studies have examined disinfection in buffered reagent-grade water. The CT values and pattern of relative efficacy of monochloramine for HAdV2, CVB5, E11, and MNV in source waters are in agreement with our earlier work in reagent-grade water (Cromeans et al., 2010). Other studies reported an approximate CT value of 265 for 1-log10 reduction of HAdV2 at pH 8 and 22e23
C (Ballester and Malley, 2004; Sirikanchana et al., 2008). This is in accord with the 1-log10 CT values of 200e800 for HAdV2 at 15
C in the present study (data not shown). In addition, the 4-log10 CT value of 1040 for CVB5 at pH 8 and 5
C can be extrapolated from the work of Sobsey et al., (1988), which is within the range of 650e1500 for 4-log10 inactivation of CVB5 in the present study (data not shown). HAdV2 was one of the most resistant of the study viruses to monochloramine disinfection, and has also been found to be highly resistant to UV disinfection (Gerba et al., 2002; Thurston-Enriquez et al., 2003b). The mechanism of inactivation of viruses by monochloramine is not well understood.

Studies on the mechanism of virus inactivation by free chlorine have suggested that viruses may be inactivated differently and that damage may occur to either viral proteins or RNA (Alvarez and O’Brien, 1982; Harakeh and Butler, 1984). One study on a single-stranded DNA virus indicated that the initial action of chlorine was on the viral capsid (Churn et al., 1983). Because UV inactivation targets the viral genome it believed that post-disinfection DNA repair mechanisms could play a role in its resistance (Eischeid et al., 2009). The increased exposure time necessary to inactivate HAdV2 with monochloramine could suggest that this inactivation mechanism also involves DNA damage. Alternatively, the stability of the viral capsid in source water may contribute to the low inactivation rates of HAdV2 by monochloramine.

3.2.

Effect of pH

Comparisons of the disinfection efficacy between pH 7 and 8 were limited to six experiments in which the pH drift was 0.25 pH units. Inactivation of CVB5 at 15
C in CCMWA water at 3 mg/L and WASH water at 1 mg/L was more effective at pH 7 than pH 8 ( p < 0.0001). Previous studies found that the disinfection efficacy of monochloramine for CVB5 decreased as the pH increased in buffered, reagent-grade water (Cromeans et al., 2010; Kelly and Sanderson, 1960). This trend has also been demonstrated for HAdV2 in buffered water (Sirikanchana et al., 2008). Inactivation of MNV at 15
C in WASH water at 1 and 3 mg/L was more effective at pH 7 than pH 8 ( p < 0.0001). However, there was no difference in the inactivation efficacy of MNV between pH 7 and 8 at 5
C in BGC water at 1 mg/L ( p ¼ 0.1414) and WASH water at 3 mg/L ( p ¼ 0.461), which is in agreement with our earlier work in buffered, reagent-grade water (Cromeans et al., 2010).

Table 2 e CT values for 3-log10 reduction of study viruses in CCMWA water using 1 and 3 mg/L monochloramine. Concentration Temp pH HAdV2 CVB5 (mg/L) (
C) 1

5 15

3

5 15

a b c d

7 8 7 8 7 8 7 8

1300 1800 720a 750 1900a,b 2600 620a 1100

780 930 290a 330 960a 1000 250a 330a

E11

MNV

1200b 1700b,d 310a,b 300b 2100a,c 1800b 950a,b 870c

53 75 25a 29 76a 97 44a 46

pH was stable enough for statistical comparisons. CT value extrapolated using EFH model. One replicate extrapolated. CT data for only one replicate.

Table 3 e CT values for 3-log10 reduction of study viruses in WASH water using 1 and 3 mg/L monochloramine. Concentration Temp pH HAdV2 CVB5 (mg/L) (
C) 1

5 15

3

5 15

7 8 7 8 7 8 7 8

1400 2900a 300 1300 1900 2000a 820 2000a,b

1100 1200a 220a 260a 870 790a 320 400a

E11

MNV

1900b 2000a,b 510b 350a,b 980 810a 550b 380a,b

88 110 33a 63a 68a 66a 23a 32a

a pH was stable enough for statistical comparisons. b CT value extrapolated using EFH model.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 4 5 e1 7 5 1

1749

Table 4 e CT values for 3-log10 reduction of study viruses in BGC water using 1 and 3 mg/L monochloramine. Concentration Temp pH HAdV2 CVB5 (mg/L) (
C) 1

5 15

3

5 15

7 8 7 8 7 8 7 8

1200 1300a 480 810b 2600 3300a 550 450a

490 450a 250 280a 670 670a 240 240a

E11

MNV

1700b 2300b 460b 330 1400b 1300a,b 570c 430a

61a 70a 27 38a 77a 27 30a 37

a pH was stable enough for statistical comparisons. b CT value extrapolated using EFH model. c One replicate extrapolated.

3.3.

Effect of temperature and disinfectant concentration

Inactivation proceeded more rapidly at 15
C than 5
C for all study viruses. Overall, the 3-log CT values calculated for CVB5 inactivation with 1 and 3 mg/L monochloramine were similar at each temperature. However, there was occasional variation between the 1 and 3 mg/L CT values calculated for HAdV2, E11, and MNV. Theoretically, CT values obtained with different chemical disinfectant concentrations should be similar, as the CT concept suggests a linear proportional effect between disinfectant concentration and exposure time. However, this theoretical relationship was not observed in some of the HAdV2, E11, and MNV experiments, where an increase in the disinfectant concentration from 1 to 3 mg/L did not result in a proportional decrease in contact time required to achieve 2-, 3-, and 4-log inactivation. In some cases the CT values obtained at 3 mg/L were up to 3 orders of magnitude higher or lower than the CT values at 1 mg/L, indicating that inactivation proceeded slower or faster, respectively, than the theoretical relationship indicates. This variation was observed in all water types, but there was no trend to indicate that it was associated with pH, temperature, or water type.

3.4.

Fig. 1 e Inactivation curves for HAdV2 at pH 7 and 5
C using 1 mg/L monochloramine in CCMWA (d-d), WASH (–B–), BGC ($$$$:$$$$) and RGW (e$$eAe$$e).

Effect of water type

Comparisons of the disinfection efficacy of monochloramine between the source waters were limited, but could be evaluated for experiments where the pH drift was 0.25 pH units. Overall, the CT values for CVB5, HAdV2, and E11 were significantly different by water type. However, no consistent trends were observed to indicate that monochloramine disinfection was more or less effective in a particular source water type. The CT values for MNV were similar in all water types. For each of the study viruses, monochloramine disinfection in source water was both more or less effective than disinfection in buffered, reagent-grade water, depending on the source water type (Cromeans et al., 2010). Researchers have speculated that water quality characteristics such as particulate matter and ionic content are important factors that can affect disinfection efficacy in natural waters. Particulate matter may act to protect viral particles by creating a disinfectant demand or by physically shielding the virions. Not all particles studied have been found to offer a protective effect, so it may be that

particle characteristics such as composition, size, and structure are more important factors than particle quantity (Templeton et al., 2008). Increased ion content has been found to enhance the effectiveness of disinfectants (Berg et al., 1989a, 1990). The mechanism for this has not been established, but several theories have been suggested. Ion pairs that form from salt cations and OCl may be more virucidal than OCl. Alternatively, salt cations may facilitate OCl access to target sites by neutralizing negative charges on viral particles that repel OCl. Because no one water type was associated with the highest or lowest inactivation rates in the present study, it is difficult to say that any one water quality parameter measured was responsible for affecting the rate of inactivation of viruses. It is likely that a combination of both water quality parameters and viral characteristics affect the efficacy of disinfection with monochloramine.

3.5.

Inactivation kinetics

Representative inactivation kinetic curves for each of the study viruses at pH 7 and 5
C are presented in Figs. 1e4 for experiments using 1 mg/L monochloramine. Inactivation

Fig. 2 e Inactivation curves for CVB5 at pH 7 and 5
C using 1 mg/L monochloramine in CCMWA (d-d), WASH (–B–), BGC ($$$$:$$$$) and RGW (e$$eAe$$e).

1750

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 4 5 e1 7 5 1

4.

Fig. 3 e Inactivation curves for E11 at pH 7 and 5
C using 1 mg/L monochloramine in CCMWA (d-d), WASH (–B–), BGC ($$$$:$$$$) and RGW (e$$eAe$$e).

kinetic curves for the study viruses at pH 7 and 5
C in RGW are also incorporated into each figure for comparison (Cromeans et al., 2010). No differences in disinfections kinetics were observed between experiments performed at 1 and 3 mg/L monochloramine. While disinfection kinetics for some viruses differed between pH 7 and 8, only pH 7 kinetics are only described in the text of this report (see Water Research Foundation project report noted in Acknowledgments for extensive presentation of study results). HAdV2 exhibited first-order kinetics at pH 8, but at pH 7 the inactivation rate was slightly higher at the beginning of experiments (Fig. 1). CVB5 and E11 exhibited first order inactivation kinetics in all source waters (Figs. 2 and 3), while inactivation curves for MNV exhibited a second-order tailing effect in all source waters (Fig. 4). Inactivation kinetics for the study viruses in the source waters were similar to inactivation kinetics in RGW (Cromeans et al., 2010; Kelly and Sanderson, 1960; Sirikanchana et al., 2008). This indicates that water quality had little impact on the mechanism of monochloramine inactivation on the study viruses.

Conclusions

Water quality had a significant impact on the inactivation rates of the CCL2 viruses HAdV2, CVB5, and E11 exposed to monochloramine, as shown by the significant differences between the CT values obtained from different source water types. However, there was no indication that virus inactivation was consistently enhanced or inhibited in any one water type compared to the others. Water quality did not affect monochloramine disinfection of MNV, as demonstrated by similar CT values in each of the source waters. The patterns of relative resistance and inactivation kinetics for the study viruses was consistent in each of the source waters, and was also in agreement with results obtained from previous experiments conducted in RGW. CT values from experiments using 1 and 3 mg/L monochloramine were not always similar. Although no trends in CT values were observed in this study between monochloramine concentration levels, the results indicate that water treatment planners should take care when extrapolating expected CT values when applied disinfectant concentrations are different than experimental concentrations. As water utilities implement new disinfection strategies in order to comply with new regulations, a comprehensive knowledge of the disinfection efficacy of preformed monochloramine in natural water is important. The results of this study can be used to ensure that current and planned chloramination systems are designed and operated to meet specific disinfection goals. CT values from this study for HAdV2 and E11 can be used in conjunction with disinfection results from other studies to guide system planning and operation. The results from this study indicate that CT values of 3300 and 2000 may be needed to achieve a 3-log10 inactivation of HAdV2 with monochloramine at pH 8 and 5 and 15
C, respectively, which is above the CT values of 1423 and 712 recommended by the USEPA Guidance Manual to achieve a 3-log10 inactivation with chloramines at pH 8. In addition, the data from this study can be used to model the survival of these viruses in distribution systems, whether associated with a treatment system breakthrough or distribution system intrusion scenario.

Acknowledgements

Fig. 4 e Inactivation curves for MNV at pH 7 and 5
C using 1 mg/L monochloramine in CCMWA (d-d), WASH (–B–), BGC ($$$$:$$$$) and RGW (e$$eAe$$e).

The authors thank Dr. Charles Humphrey (CDC) for electron microscopy analysis of virus preparations and Bonnie Mull and Carmela Smith (Atlanta Research and Education Foundation) for assistance with experiment preparation and virus plaque assays. The use of trade names and names of commercial sources is for identification only and does not imply endorsement by the Centers for Disease Control and Prevention or the U.S. Department of Health and Human Services. The findings and conclusions in this presentation are those of the authors and do not necessarily represent those of the Centers for Disease Control and Prevention. Funding for this project was provided by the Water Research Foundation (Project #3134, Contaminant Candidate List Viruses: Evaluation of Disinfection Efficacy), the Centers for Disease Control and Prevention, and the Atlanta Research and Education Foundation.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 4 5 e1 7 5 1

references

Alvarez, M.E., O’Brien, R.T., 1982. Effects of chlorine concentration on the structure of poliovirus. Applied and Environmental Microbiology 43, 237e239. APHA, 2005. Standard Methods for the Examination of Water and Wastewater, Twenty-first ed. American Public Health Association, Washington, DC. AWWA, 2008. Committee report: disinfection survey, part 1recent changes, current practices, and water quality. Journal of the American Water Works Association 100, 76e90. Ballester, N.A., Malley, J.P., 2004. Sequential disinfection of adenovirus type 2 with UV-chlorine-chloramine. Journal of the American Water Works Association 96, 97e103. Barbeau, B., Desjardins, R., Mysore, C., Prevost, M., 2005. Impacts of water quality on chlorine and chlorine dioxide efficacy in natural waters. Water Research 39, 2024e2033. Baxter, C.S., Hofmann, R., Templeton, M.R., Brown, M., Andrews, R.C., 2007. Inactivation of adenovirus types 2, 5, and 41 in drinking water by UV light, free chlorine, and monochloramine. Journal of Environmental Engineering 133, 95e103. Berg, G., Sanjaghsaz, H., Wangwongwatana, S., 1989a. Potentiation of the poliocidal effectiveness of free chlorine by a buffer. Journal of Virological Methods 23, 179e186. Berg, G., Sanjaghsaz, H., Wangwongwatana, S., 1989b. Potentiation of the virucidal effectiveness of free chlorine by substances in drinking water. Applied and Environmental Microbiology 55, 390e393. Berg, G., Sanjaghsaz, H., Wangwongwatana, S., 1990. KCl potetiation of the virucidal effectiveness of free chlorine at pH 9.0. Applied and Environmental Microbiology 56, 1571e1575. Churn, C.C., Bates, R.C., Boardman, G.D., 1983. Mechanism of chlorine inactivation of DNA-containing parvovirus H-1. Applied and Environmental Microbiology 46, 1394e1402. Cromeans, T.L., Kahler, A.M., Hill, V.R., 2010. Inactivation of adenoviruses, enteroviruses, and murine norovirus in water by free chlorine and monochloramine. Applied and Environmental Microbiology 76, 1028e1033. Eischeid, A.C., Meyer, J.N., Linden, K.G., 2009. UV disinfection of adenoviruses: molecular indications of DNA damage efficiency. Applied and Environmental Microbiology 75, 23e28. Gerba, C.P., Gramos, D.M., Nwachuku, N., 2002. Comparative inactivation of enteroviruses and adenovirus 2 by UV light. Applied and Environmental Microbiology 68, 5167e5169. Haas, C.N., Joffe, J., 1994. Disinfection under dynamic conditions e modification of Hom model for decay. Environmental Science and Technology 28, 1367e1369.

1751

Haas, C.N., Joffe, J., Anmangandla, U., Jacangelo, J.G., Heath, M., 1996. Water quality and disinfection kinetics. Journal of the American Water Works Association 88, 95e103. Harakeh, M., Butler, M., 1984. Inactivation of human rotavirus, SA11 and other enteric viruses in effluent by disinfectants. Journal of Hygiene 93, 157e163. Kahler, A.M., Cromeans, T.L., Roberts, J.M., Hill, V.R., 2010. Effects of source water quality on chlorine inactivation of adenovirus, coxsackievirus, echovirus, and murine norovirus. Applied and Environmental Microbiology 76, 5159e5164. Karst, S.M., Wobus, C.E., Lay, M., Davidson, J., Virgin, H.W., 2003. STAT1-dependent innate immunity to a Norwalk-like virus. Science 299, 1575e1578. Kelly, S.M., Sanderson, W.W., 1960. The effect of chlorine in water on enteric viruses. II. The effect of combined chlorine on poliomyelitis and coxsackie viruses. American Journal of Public Health 50, 14e20. Sirikanchana, K., Shisler, J.L., Marinas, B.J., 2008. Inactivation kinetics of adenovirus serotype 2 with monochloramine. Water Research 42, 1467e1474. Sobsey, M.D., Fuji, T., Hall, R.M., 1991. Inactivation of cellassociated and dispersed hepatitis A virus in water. Journal of the American Water Works Association 83, 64e67. Sobsey, M.D., Sobsey, M.D., Fuji, T., Shields, P.A., 1988. Inactivation of hepatitis A virus and model viruses in water by free chlorine and monochloramine. Water Science and Technology 20, 385e391. Templeton, M.R., Andrews, R.C., Hofmann, R., 2008. Particleassociated viruses in water: impacts on disinfection processes. Critical Reviews in Environmental Science and Technology 38, 137e164. Thurston-Enriquez, J.A., Haas, C.N., Jacangelo, J., Gerba, C.P., 2003a. Chlorine inactivation of adenovirus type 40 and feline calicivirus. Applied and Environmental Microbiology 69, 3979e3985. Thurston-Enriquez, J.A., Haas, C.N., Jacangelo, J., Riley, K., Gerba, C.P., 2003b. Inactivation of feline calicivirus and adenovirus type 40 by UV radiation. Applied and Environmental Microbiology 69, 577e582. USEPA, 1990. Guidance Manual for Compliance with the Filtration and Disinfection Requirements for Public Water Systems Using Surface Water Sources. U.S. Environmental Protection Agency, Office of Water, Washington, DC. USEPA, 2005. Drinking water contaminant candidate list 2; final notice. Federal Register 70, 9071e9077. USEPA, 2006a. National primary drinking water regulations: long term 2 enhanced surface water treatment rule; final rule. Federal Register 71, 653e702. USEPA, 2006b. National primary drinking water regulations: stage 2 disinfectants and disinfection byproducts rule; final rule. Federal Register 71, 388e493.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 5 2 e1 7 6 2

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Impact of urbanization and agriculture on the occurrence of bacterial pathogens and stx genes in coastal waterbodies of central California Sarah P. Walters*, Anne L. Thebo, Alexandria B. Boehm* Stanford University, Environmental and Water Studies, Department of Civil & Environmental Engineering, 473 Via Ortega, Stanford, CA 94305, USA

article info

abstract

Article history:

Fecal pollution enters coastal waters through multiple routes, many of which originate

Received 29 August 2010

from land-based activities. Runoff from pervious and impervious land surfaces transports

Received in revised form

pollutants from land to sea and can cause impairment of coastal ocean waters. To

20 November 2010

understand how land use practices and water characteristics influence concentrations of

Accepted 22 November 2010

fecal indicator bacteria (FIB) and pathogens in natural waters, fourteen coastal streams,

Available online 30 November 2010

rivers, and tidal lagoons, surrounded by variable land use and animal densities, were sampled every six weeks over two years (2008 & 2009). Fecal indicator bacteria (FIB;

Keywords:

Escherichia coli and Enterococci) and Salmonella concentrations, the occurrence of Bacteroidales

Salmonella

human, ruminant, and pig-specific fecal markers, E. coli O157:H7, and Shiga toxin (stx)

Escherichia coli O157:H7

genes present in E. coli, were measured. In addition, environmental and climatic variables

Urbanization

(e.g., temperature, salinity, rainfall), as well as human and livestock population densities

Agriculture

and land cover were quantified. Concentrations of FIB and Salmonella were correlated with

Fecal indicator bacteria

each other, but the occurrence of host-specific Bacteroidales markers did not correlate with

Bacteroidales

FIB or pathogens. FIB and Salmonella concentrations, as well as the occurrence of E. coli

Loading

harboring stx genes, were positively associated with the fraction of the surrounding sub-

Stx

watershed that was urban, while the occurrence of E. coli O157:H7 was positively associated with the agricultural fraction. FIB and Salmonella concentrations were negatively correlated to salinity and temperature, and positively correlated to rainfall. Areal loading rates of FIB, Salmonella and E. coli O157:H7 to the coastal ocean were calculated for stream and river sites and varied with land cover, salinity, temperature, and rainfall. Results suggest that FIB and pathogen concentrations are influenced, in part, by their flux from the land, which is exacerbated during rainfall; once waterborne, bacterial persistence is affected by water temperature and salinity. ª 2010 Elsevier Ltd. All rights reserved.

1.

Introduction

Fecal pollution of environmental waters compromises water quality, poses human health risks, and creates economic burdens for federal, state and local agencies. Fecal pollution

enters environmental waters through a variety of point and non-point sources. In urban areas, sources include leaky sewer systems and septic tanks, urban runoff, wastewater treatment plant outflows, deposition by wildlife and domestic animals, and soils. Sources in agricultural areas include poultry, hog,

* Corresponding authors. Tel.: þ1 (650) 724 9128; fax: þ1 (650) 723 7058. E-mail addresses: [email protected] (S.P. Walters), [email protected] (A.B. Boehm). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.11.032

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 5 2 e1 7 6 2

and cattle rearing facilities, runoff from croplands where manure has been applied, as well as grazing lands, wildlife, and soils (Jamieson et al., 2003; Edwards et al., 2008). In forested areas, septic tanks and pit toilets at public parks, wild animal feces and soils can be sources of fecal pollution (Table S1). Many fecal pathogens of public health importance are derived from humans, yet animal feces spread zoonotic pathogens such as Salmonella and Escherichia coli O157:H7, two of the most common pathogens isolated from human gastrointestinal (GI) patients (Center for Disease Control and Prevention, 2003). Salmonella is the leading cause of gastroenteritis worldwide (Bell and Kyriakides, 2002). Most salmonellosis in the US is foodborne (Heinitz et al., 2000). However, Salmonella has been isolated from fresh and marine waters (e.g., Martinez-Urtaza et al., 2004; Haley et al., 2009; Wilkes et al., 2009) including streams and wetlands of central California (Shellenbarger et al., 2008; Schriewer et al., 2010). Furthermore, Salmonella spp. have been isolated from algal wrack, benthic organisms, and marine mammal feces (Miller et al., 2006, 2009; Byappanahalli et al., 2009). Therefore, water represents a potential vehicle for Salmonella transmission. Shiga toxin-producing E. coli (STEC) encompass a broad range of strains that cause human disease. While E. coli O157:H7 is the most common STEC strain associated with disease outbreaks, the STEC group of pathogens also includes non-O157 strains. STEC infections are most commonly associated with contaminated food products; however, outbreaks have also been linked to swimming (e.g., Craun et al., 2005; Muniesa et al., 2006). One of the most important factors influencing STEC pathogenesis is expression of one or both shiga toxin (stx) genes (Muniesa et al., 2006). Most stx genes are phage encoded, providing a means of horizontal gene transfer (HGT). Because E. coli persist and survive in extra-intestinal environments (LaLiberte and Grimes, 1982), HGT of stx genes may lead to the emergence of new pathogenic E. coli. Bacteroidales host-specific fecal markers are promising alternatives to traditional FIB for detection of fecal pollution and its source(s) (Walters et al., 2007). Host population densities are greater in some locations and certain hosts are more likely to harbor particular pathogens. Accordingly, correlation of host-specific markers with particular land use activities, or certain pathogens, may augment their use in fecal source tracking, risk analysis, and remediation. The overarching goal of this study was to understand how land cover and land use practices, as well as climatic and environmental variables, control concentrations and loading of FIB, Salmonella, E. coli O157:H7, and E. coli harboring stx genes in waterbodies along the central California coast. The investigation was premised on a conceptual model where bacterial concentrations in the waterways (1) increase due to bacterial fluxes from the surrounding land which in turn are affected by land cover and use, as well as rainfall, (2) change in response to the physico-chemical environment of the waterway, and (3) can be diluted in some cases by seawater (when waterways experience inputs of seawater during rising tides). A secondary goal was to assess the feasibility of using Bacteroidales hostspecific markers as indicators of the presence of bacterial pathogens. Accordingly, fourteen waterbodies along the central California coast draining or adjacent to subwatersheds impacted by varying land use practices were sampled every six

1753

weeks from January 2008 through November 2009. Land cover, human and farm animal populations, antecedent rainfall, water quality variables, concentrations of FIB and Salmonella, and the presence of E. coli O157:H7, stx genes in E. coli, and Bacteroidales host-specific markers were measured.

2.

Methods

2.1.

Sampling

Fourteen central Californian waterbodies were sampled during the study (Fig. 1). For all waterbodies (with exception of the Salinas and Pajaro Rivers) sampling was carried out every six weeks, on average, over 24 months from January 2008 through November 2009 (n ¼ 18 for each site); exact sampling dates are provided in the Supplementary Material (SM) (Table S2). Sampling of the Salinas and Pajaro Rivers did not commence until 20 Feb 2008 and 22 Feb 2009, respectively (n ¼ 17 and 8). All samples were obtained on the same day; sampling began at 430 h and finished by 1500 h.

2.2.

Spatial analysis

The subwatershed was chosen as the unit for land use and cover analysis. Thus, an inherent assumption in the analysis was that it was the subwatershed in close proximity to the sampling site that exerted control over water quality. The hydrologic unit classification (HUC-12) dataset from United States Geological Survey (USGS) was used to define the subwatershed boundaries (Natural Resources Conservation Service, 2009). Land cover within each of the subwatersheds was determined using the 2001 NOAA Coastal Change Analysis Program (C-CAP) land cover dataset (NOAA Coastal Services Center, 2001). The land cover data were clipped to the subwatershed shape file using ARCMAP (ESRI, Redlands, CA). This classified the land area within each subwatershed into 20 land cover classifications. Land cover classes were further classified as agricultural (classes 6 and 7), forested (classes 9e12, 19, 20 which include mixed forests and shrubs), urban (classes 2e5), and other (including wetland, classes 13e18, and open water, class 21). To determine whether C-CAP land cover class 8 (grassland/herbaceous) should be classified as agricultural or forested, all grassland/herbaceous classified areas within the subwatershed were clipped with a grazing lands layer extracted from the California prime agricultural lands land cover data (California Prime Agricultural Lands, 2008) using ARCMAP. Grassland/herbaceous areas that were also classified as grazing lands were designated as agricultural while the other grassland/herbaceous areas were designated as forested. The area of impervious surface cover (ISC) in each subwatershed was computed using the 2001 National Land Cover Dataset (USGS, 2001). Note that ISC can occur within any of the land covers described above. Human, poultry, pig, and cattle populations within each subwatershed were determined using human and animal census data provided by the United States Department of Agriculture. Details are provided in the SM. Data on livestock are not readily available in the US, so numbers should be viewed as approximate. The livestock unit (LSU) has been investigated as a predictor of fecal pathogen and indicators in catchments (Schaffter and Parriaux,

1754

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 5 2 e1 7 6 2

Fig. 1 e Sampling site locations along the central California coast. Shaded areas represent subwatersheds; dark grey lines separate adjacent subwatersheds. Black lines delineate county lines.

2002). LSU data were not readily available for the present study, so this measure of livestock could not be used. Land cover and use variables were normalized by the total subwatershed area to obtain: fraction urban, fraction forested, fraction agricultural, fraction impervious surface coverage, poultry density, cattle density, hog density, and human density. The normalization of animal density to total area of the subwatershed does not reflect the true distribution of animals on the land; such data were not publicly available (see SM). While in this region it is likely that animals are contained primarily within agricultural land cover, it is also possible they are distributed in urban and forested land covers. The primary land cover within each subwatershed was classified as forested, agricultural, or urban based on the land cover that represented the greatest fraction.

2.3.

Rainfall

Rainfall data for each site were obtained from the nearest recording rain gauge from National Weather Service (2010). Rainfall accumulation was estimated for 7-day antecedent rainfall at each site, for each sampling date and this was translated into a binary variable representing the occurrence or absence of 7-day antecedent rainfall.

2.4.

Water collection and in situ measurements

Water samples were collected in 5-L acid-washed, autoclaved, triple-rinsed, high-density polyethylene bottles from a location as close to where the waterbody discharged to the coastal ocean as possible. Samples were transported to the laboratory

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 5 2 e1 7 6 2

on ice and processed immediately. A probe (Hach Hydrolab Quanta, Loveland, CO; or YSI85, YSI Inc., Yellow Springs, OH) was used to record temperature and salinity. Dissolved oxygen percent saturation (DO%) measurements were recorded using a Clark-type electrode (Hach Hydrolab Quanta) or an optical probe (ProODO, YSI Inc.). Turbidity measurements on 7 April 2008 were obtaining using the Hydrolab Quanta probe. Thereafter, turbidity was measured at the laboratory using a bench top turbidity meter (HF Scientific DRT-15CE, Fort Myers, FL). A handheld flow meter (Flow Probe Model FP100, Global Water, Instrumentation, Inc., Gold River, CA) was used to obtain a depth- and cross sectional-average flow velocity during each sampling event at streams without USGS flow gauges (San Pedro and Waddell). The velocity was multiplied by the cross sectional area of the stream to estimate volumetric flow rates. Discharge rates for the remaining streams and rivers were obtained from USGS (2010) flow gauges (Table S3). Discharge rates were not obtained for Kirby Park, Moss Landing, or Bolinas Lagoon which represent tidal driven lagoons and estuaries. Further details are in the SM.

2.5.

Dissolved inorganic nitrogen

Thirty milliliters of water were filtered through a 0.2 mm pore size PES syringe filter (Millipore, Billerica, MA) and stored at 20  C until analysis. Nitrate þ nitrite and ammonium were measured by standard methods with a nutrient autoanalyzer (Lachat QuikChem 8000, Loveland, CO). Dissolved inorganic nitrogen (DIN) is reported as the sum of the molar concentrations of nitrate þ nitrite and ammonium.

2.6.

E. coli and Enterococci

Concentrations of E. coli and Enterococci were determined according to EPA method 1604 and 1600 (USEPA 2002a, USEPA 2002b) using membrane filtration. Typically, 30 and/or 100 ml were filtered, depending on expected concentrations. The SM contains further details.

2.7.

Salmonella

Salmonella were enumerated by adapting EPA method 1682 for detection of Salmonella in biosolids (USEPA, 2006) to a threetube Most Probable Number (MPN) method for water samples (Shellenbarger et al., 2008). Further details are provided in the SM. The lowest concentration detectable by the assay, per MPN tables, is 0.75 MPN/L.

2.8.

E. coli O157:H7

Water samples were assayed for E. coli O157:H7 starting with the second sampling event according to the procedure described by Walters et al. (2007) with a few modifications which are described in the SM. The lowest possible concentration measurable with the assay, which enriches 90 mL of water in a liquid medium, is 1 MPN/90 ml.

2.9.

Detection of stx genes in E. coli

To detect stx genes in E. coli cells, membrane filters from each MI plate were removed following incubation and enumeration,

1755

placed in 25 ml of tryptic soy broth (TSB, EMD Chemicals, Gibbstown, NJ) and incubated at 37  C for 4h. One milliliter of the bacterial suspension was pelleted, washed twice with 1 ml PBS (Invitrogen, Carlsbad, CA), repelleted, and the PBS removed. Each cell pellet received 20 ml proteinase K and 180 ml buffer ATL (Qiagen, Valencia, CA). Pellets were resuspended and stored at 80  C until DNA extraction (see SM). The presence of the stx genes on each MI filter was confirmed by PCR of stx1 and stx2 (see SM). Between 30 and 100 ml of water were filtered for E. coli enumeration on MI agar, so the assay can detect as low as 1e3 CFU stx-containing E. coli/100 ml; the true detection limit is higher due to the fact that bacteria are distributed according to a Poisson distribution in a water sample.

2.10.

Host-specific Bacteroidales

One hundred milliliters of each water sample was filtered onto 0.45 mm HA membranes (Thermo Fisher Scientific Inc., Waltham, MA). Filters were stored, processed, and tested for human (HF183), ruminant (CF128 and CF193), and pig-specific (PF163) Bacteroidales fecal markers as described previously (Boehm et al., in press) (see SM).

2.11.

Areal loading

The areal loading of FIB, Salmonella, and E. coli O157:H7 from all subwatersheds except for Kirby Park, Moss Landing, and Bolinas Lagoon (which represent tidal lagoons, harbors, or estuaries) was calculated using the equation: loading ¼ C  Q/A. In this equation, C is the concentration of microorganism, Q is the volumetric flow rate of fresh water, and A is the subwatershed area. For E. coli O157:H7, it was conservatively estimated that whenever the organism was detected, the concentration was 1 MPN/90 mL. Even though calculated over a variety of flow conditions including storm events (Sprague, 2001), the loading rates should be viewed as estimates because loading is inherently difficult to measure (Smart et al., 1999). In some cases, the sampled water had brackish salinities. In these cases, no effort was made to correct C for dilution with seawater so that a conservative areal load could be obtained. If a correction had been applied, then C and thus loading would be higher.

2.12.

Statistical analysis

All statistical analyses were carried out using PASW Statistics Release 18.0.0 (SPSS Inc., Chicago, Illinois). The approach was to first test bivariate associations between bacterial concentrations, or their occurrence, and the independent land use/ cover and environmental variables. Then multivariate models were developed using independent variables that appeared to be promising predictors based on the bivariate analyses. FIB and Salmonella concentrations, as well as loadings were log10(X þ 1) transformed (Russ et al., 2005) to achieve normality. Normality of data was confirmed using QeQ plots. One-way analyses of variance (ANOVA) with Tukey’s post hoc tests were used to compare data groups. Pearson’s r (rp) was used for bivariate correlations. Contingency tables were used to test co-occurrence of binary variables. Multivariate binary logistic and linear regressions were used to model microorganisms as a function of environmental and land use

1756

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 5 2 e1 7 6 2

variables using stepwise forward methods. Because observations represent repeated measures, generalized estimating equations (GEEs) were also used to assess associations between variables (Hardin and Hilbe, 2003). When results between GEE and other methods diverge, GEE results in the form of coefficients (b) and p values are provided. Specifically, GEE results are reported to describe the associations between land cover/use variables and water quality, and also to model bacterial loading. All statistical tests were deemed significant if p < 0.05. Tests with p < 0.1 are discussed in several instances.

3.

Results

3.1.

Land cover and use

Land cover at the study sites consisted of a variety of different agricultural covers (crop and grazing lands), forests, grasslands, rural residences and regions of heavy urbanization (Table 1). The 14 sampling locations included 3, 3, and 8 sites that were primarily for urban, agricultural, and forested land use, respectively. Animal densities, as averaged over the entire subwatershed, varied; the highest density of cattle resided in Lagunitas, while the highest densities of hogs and poultry resided in Pajaro River subwatershed (Table 1). Collinearity between land use and cover variables was examined, as 2 independent variables that are collinear should not both be retained in statistical analyses. Urban fraction was highly correlated to human population density (rp ¼ 0.89, p < 0.05) and impervious surface coverage (rp ¼ 0.99, p < 0.05), and negatively correlated to fraction forested (rp ¼ 0.71, p < 0.05). Thus, the urban fraction variable alone was used to represent all four variables. Poultry and hog densities were highly, positively correlated (rp ¼ 0.89, p < 0.05), so hog density

alone was retained to represent these animals. Agricultural fraction and cattle density did not exhibit collinearity with other land use/cover variables.

3.2.

Bacterial targets

E. coli (EC) and enterococci (ENT) were detected in 92.5% (223/ 241) and 80.0% (181/227) of samples, respectively. The number of ENT measurements was lower than the EC measurements because ENT data from July 2009 were omitted due to poor agar performance. Mean and standard deviations of log10(X þ 1) transformed concentrations across site and sampling events were 1.9  0.9 and 1.4  1.0 log CFU/100 ml for EC and ENT respectively. EC and ENT concentrations varied significantly across sites (ANOVA, p < 0.05); San Francisquito Creek had the highest concentrations of EC and San Pedro Creek had the highest concentrations of ENT (Table 2). Of the Bacteroidales host-specific markers, the human HF183 marker was most frequently detected (6.6%, 15 of 227), and widely distributed; the marker was detected at all sites except Lagunitas Creek and Moss Landing Harbor (Table 2). The ruminant maker CF193 was detected in 3 of 227 samples (1.3%) and only in Waddell Creek and San Lorenzo River. The ruminant marker CF128 was not detected. The pig marker PF163 was detected in 6 of 187 (3.2%) and confined to Kirby Park, Lagunitas, Moss Landing, Pescadero, and Waddell. During two years of sampling Salmonella were isolated from 30.7% (74/241) of all samples; concentrations ranged from <0.75 to 7.25 MPN/L. The concentrations and occurrence of Salmonella were significantly different across sites (ANOVA, p < 0.05; chi-square test, p < 0.05, respectively). Salmonella was detected most frequently at San Pedro Creek (13/18 times) and this site also had the highest mean concentration (Table 2). E. coli O157:H7 was isolated from 15 of 229 water samples (6.6%) (Table 2). The 15 isolations were obtained from 6

Table 1 e Primary land cover class, subwatershed area, average salinity, fraction (frac) of the subwatershed that is urban (URB), agricultural (AG), and forested (FOR), the impervious surface coverage (ISC) fraction, human, cattle, poultry, and hog densities within each subwatershed. Salinity is dimensionless and represents the average salinity across all measurements. B is Bolinas Lagoon, KP is Kirby Park, L is Lagunitas Creek, ML is Moss Landing Harbor, N is Napa River, SA is Salinas River, Pes is Pescadero Creek, Pet is Petaluma River, PR is Pajaro River, SF is San Francisquito Creek, SL is San Lorenzo River, SO is Soquel Creek, and W is Waddell Creek. Note that the URB, AG, and FOR fractions do not sum to 1 as additional land covers exist; these have been classified as ‘other’ and include wetlands and open water (see materials and methods). The latitude and longitude for each site can be found in Table S5. Site

Primary land cover

Area (km2)

Salinity

Urb frac

Ag frac

For frac

ISC frac

Human density (#/km2)

Cattle density (#/km2)

Poultry density (#/km2)

Hog density (#/km2)

B KP L ML N PR Pes Pet SA SF SL SP SO W

FOR FOR FOR AG URB AG FOR FOR AG URB FOR URB FOR FOR

46 127 70 47 194 197 134 95 293 114 79 81 110 62

32.4 32.5 2.9 30.1 16.2 7.4 22.2 20.5 7.2 14.5 10.6 2.0 0.2 1.6

0.07 0.13 0.03 0.27 0.33 0.18 0.00 0.12 0.13 0.84 0.22 0.60 0.07 0.00

0.001 0.18 0.12 0.36 0.09 0.45 0.13 0.06 0.52 0.01 0.02 0.001 0.01 0.01

0.91 0.59 0.83 0.30 0.14 0.33 0.86 0.52 0.32 0.15 0.75 0.38 0.92 0.99

0.02 0.04 0.01 0.09 0.18 0.08 0.00 0.05 0.06 0.36 0.09 0.32 0.02 0.00

42 108 7 189 579 296 10 116 170 1992 457 2913 125 14

19 6 28 13 1 10 2 16 15 2 0 1 0 0

0.00 0.00 0.00 0.00 0.00 2.34 0.87 0.00 0.00 0.87 0.00 0.53 0.00 0.00

0.00 0.02 0.00 0.05 0.02 0.23 0.14 0.11 0.05 0.14 0.00 0.09 0.00 0.00

1757

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 5 2 e1 7 6 2

Table 2 e Summary of bacterial data by site. Sal is Salmonella. O157 is E. coli O157:H7. Bacteroidales markers are HF183 (human), CF128 (ruminant), C193 (ruminant), and PF163 (pig). Units of ENT and EC are CFU/100 ml, units of Salmonella are MPN/L, remaining columns report the number of samples positive for the target out of the total number of samples tested. Site definitions are provided in the caption of Table 1. Site

Log10 (ENT þ 1)

Log10 (EC þ 1)

Log10 (Sal þ 1)

Sal

stx

O157

HF183

CF128

CF193

PF163

B KP L ML N PR Pes Pet SA SF SL SO SP W

1.4 1.3 1.5 1.3 1.1 1.8 0.8 1.1 1.1 2.0 1.4 1.7 2.1 1.5

1.7 1.3 2.3 1.9 1.7 2.2 1.3 1.5 1.8 2.6 2.2 2.2 2.3 2.1

0.014 0 0.19 0.15 0.027 0.030 0.016 0.078 0.15 0.14 0.17 0.20 0.41 0.17

1/18 0/18 9/18 6/18 2/18 1/8 1/18 4/18 7/17 7/18 7/18 8/18 13/18 8/18

2/18 1/18 5/18 3/18 1/18 0/8 0/18 1/18 1/17 4/18 2/18 1/18 2/18 2/18

1/17 0/17 2/17 0/17 1/17 0/8 2/17 0/17 4/17 0/17 1/17 1/17 0/17 3/17

1/17 1/17 0/17 0/17 1/17 1/7 1/17 2/17 1/16 2/17 1/17 1/17 1/17 2/17

0/18 0/18 0/18 0/18 0/18 0/8 0/18 0/18 0/17 0/18 0/18 0/18 0/18 0/18

0/17 0/17 0/17 0/17 0/17 0/7 0/17 0/17 0/16 0/17 1/17 0/17 0/17 2/17

0/14 1/14 1/14 2/14 0/14 0/6 1/14 0/14 0/13 0/14 0/14 0/14 0/14 1/14

different sampling events at 8 different sites. E. coli O157:H7 was most often isolated from Salinas River (n ¼ 4). stx genes were detected in E. coli from 25 of 241 samples (10.4%). They were detected at least once at each site except Pajaro River and Pescadero Creek; stx genes were most often detected in Lagunitas Creek (n ¼ 5) (Table 2) which also was the site with the highest cattle density.

3.3.

Correlation of bacterial pathogens and indicators

The concentrations of EC, ENT, and Salmonella were significantly correlated to each other (rp ¼ 0.59 for EC and ENT; rp ¼ 0.36 for EC and Salmonella; rp ¼ 0.26 for ENT and Salmonella, all p < 0.05). If mean bacterial concentrations at each site are considered, correlations are also significant (rp > 0.6, all p < 0.05). EC, ENT, and Salmonella concentrations were not significantly different between samples that were positive or negative for E. coli O157:H7, or for samples positive or negative for the stx genes, or any of the Bacteroidales markers. There was no significant association between the occurrence of any of the binary variables (E. coli O157:H7, Bacteroidales markers, stx genes, and Salmonella occurrence). The lack of association between E. coli O157:H7 and stx genes is discussed in the SM.

3.4.

Effects of subwatershed land cover and use

FIB and Salmonella concentrations and E. coli O157:H7, stx, and Bacteroidales marker occurrence were compared between sites with subwatersheds that have the same primary land cover classification (Fig. 2). FIB and Salmonella concentrations were elevated at primarily urban sites relative to forested sites (ANOVA, post hoc comparison, p < 0.05 for EC and Salmonella, p ¼ 0.07 for ENT). There was no significant difference in the occurrence of stx genes, E. coli O157:H7 or the Bacteroidales markers. This primary land cover classification scheme is coarse. For example, sites classified as urban have between 33 and 84% urban land cover. A more refined analysis using continuous variables to characterize land use and cover follows.

Co-variation of concentrations and occurrence of pathogens and indicators with continuous land cover and use variables was examined using generalized estimating equations (GEEs) employing linear and binary logit linking functions. EC, ENT, and Salmonella concentrations were positively associated with increasing values of urban fraction (b ¼ 0.97, 0.97, and 0.22, respectively, p < 0.05 for EC and ENT, p ¼ 0.08 for Salmonella). Salmonella concentration was also negatively associated with hog fraction (b ¼ 0.005, p < 0.05). The occurrence of E. coli O157:H7 was positively associated with the agricultural fraction (b ¼ 4.28, p < 0.05), and negatively associated with the cattle and hog densities (b ¼ 0.31 and b ¼ 0.09, respectively, p < 0.05 for both), and the urban fraction (b ¼ 8.83, p < 0.05). The occurrence of stx was positively associated with the urban fraction (b ¼ 3.75, p < 0.05) and cattle density (b ¼ 0.67, p < 0.05), and negatively associated with the hog density (b ¼ 0.12, p < 0.05). The presence of the HF marker was associated with hog density (b ¼ 0.056, p < 0.05). No variables were significant in the models for the other Bacteroidales markers.

3.5.

Effect of rainfall

The effect of rainfall was investigated by considering whether or not it had rained in the 7 days prior to sampling. When there was rainfall in the previous 7 days of sampling, EC, ENT, and Salmonella concentrations were significantly higher than when there was no rain (ANOVA; p < 0.05) by 0.2, 0.5, and 0.1 log units, respectively. In contrast, E. coli O157:H7 was more likely to be isolated when there was no rain within the previous 7 days (OR: 0.26, p < 0.05). Similarly, the odds of detecting the HF marker were higher when there was no rain within the previous 7 days (OR: 0.20, p < 0.05). No other comparisons were statistically significant. When clustering by site was accounted for using GEEs, results were unchanged.

3.6.

Effect of water parameters

Relationships between bacterial targets and salinity, temperature, dissolved oxygen percent saturation (DO%), turbidity,

1758

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 5 2 e1 7 6 2

when it was not detected (ANOVA, p < 0.05). When the CF193 marker was detected, turbidity and DIN were lower than when it was not detected ( p  0.05). The PF163 marker was also detected more often when turbidity was low ( p ¼ 0.06), which may suggest that turbidity inhibits the PCR assays. No other water quality parameters were significantly different when the binary bacterial variables (E. coli O157:H7, stx gene, and Bacteroidales markers) were present versus when they were absent. The associations were also investigated using GEEs and the results were unchanged.

3.7. Multivariate models of bacterial concentrations and occurrence

Fig. 2 e The concentrations and occurrence of bacterial targets at sites in subwatersheds of different primary land use (Ag, Urb, and For). Top: Box plot of log10(EC D 1) where EC has units of CFU/100 ml. Middle: Box plot of log10(ENT D 1) where ENT has units of CFU/100 ml. Bottom: The occurrence of stx genes, E. coli O157:H7, and Salmonella. Box plots show median (horizontal line), 25th and 75th percentiles (bottom and top lines of box), 10th and 90th percentiles (bottom and top of whiskers), and mean (black circle). Pie charts show presence (solid black) and absence (white) of targets. N values vary slightly by target; n is approximately 45 for Ag, 145 for For, and 55 for Urb.

and dissolved inorganic nitrogen (DIN) were examined. These variables may impact bacterial persistence, and salinity can be used as a measure of fresh water dilution with seawater. There were no significant correlations between EC, ENT, and Salmonella concentrations, and DO%, turbidity, and DIN. However, bacterial concentrations were negatively correlated to salinity (rp ¼ 0.34, 0.20, and 0.38 for EC, ENT, and Salmonella, respectively, p < 0.05). In fact, the odds of Salmonella detection from saline water (defined as salinity more than 0.5) were nearly one-third that of fresh water (OR: 0.28, p < 0.05). In addition, ENT and Salmonella were negatively correlated to temperature (rp ¼ 0.14 and 0.17, p < 0.05, respectively). On average, water temperatures were warmer and DO% was lower when the HF marker was detected compared to

Multivariate linear and binary logistic regression models were built for FIB, Salmonella, and E. coli O157:H7 concentrations and occurrence using independent variables that proved useful in explaining their variability in the bivariate analyses: four land use/cover variables (urban fraction, agricultural fraction, hog density, and cattle density), salinity, temperature, and a binary variable for the presence or absence of rainfall within the previous 7 days (Table 3). Multivariate models not only provide a predictive model of bacterial concentrations, they also reveal which independent variables remain important bacterial predictors when controlling for the effects of other independent variables, like salinity. Overall, models explained between 11% and 23% of the variance in the dependent variables. The urban fraction and salinity were statistically significant in the FIB and Salmonella multivariate models. The occurrence of rainfall within the previous 7 days was also significant in the ENT and Salmonella models. In the E. coli O157:H7 model, the bacterium was less likely detected if there had been rainfall, if water temperature was high or urban fraction was high, and it was more likely detected if the agricultural fraction was high. All multivariate analyses were evaluated using GEEs to account for repeated measures and results were essentially unchanged.

3.8.

Areal loading

Mean areal loading of microbial targets averaged over the 2-year study, for each subwatershed is reported in Fig. 3. EC, ENT, and Salmonella loading ranged from 0 when there was no flow to the ocean to 1012 CFU/d/km2, 1011 CFU/d/km2, and 108 MPN/d/km2, respectively. Significant differences in FIB (but not Salmonella) loading were observed among primarily urban, agricultural, and forested sites; loading from urban and forested sites was significantly higher than agricultural sites (ANOVA; post hoc tests, p < 0.05). GEEs assessed relationships between loading and four land use/cover variables: urban fraction, agricultural fraction, hog density, and cattle density, as well as rainfall, salinity, and temperature (variables previous shown to affect concentrations; Table 4). EC loading was negatively associated with agricultural fraction and temperature ( p < 0.05). ENT loading was negatively associated with agricultural fraction, salinity, and temperature, and positively associated with rainfall (all p < 0.05). Salmonella loading was negatively associated with agricultural fraction and salinity (both p < 0.05), and positively associated with the

1759

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 5 2 e1 7 6 2

Table 3 e Coefficients, unstandardized b, for multiple linear regression (MLR) or binary logistic regression models of continuous (EC, ENT, and Salmonella) and binary (E. coli O157:H7) dependent variables, respectively. bo is the constant. For the binary model, exp(b) is provided in parentheses. For the MLR, standardized b is provided in parentheses. Z is the propensity score, the dependent variable of the binary model. For the MLR models, the R2 is the adjusted R2; for the binary model, R2 is Nagelkerke R2. All variables are statistically significant in the models ( p < 0.05) unless ‘ns’ is reported. Sal is Salmonella. Dependent variable Log10(EC þ 1) Log10 (ENT þ 1) Log10(Sal þ 1) Z of E. coli O157:H7

bo

Salinity

Temp

Rainfall

Urb frac

Ag frac

Cattle density

Hog density

R2

2.07 1.18 0.14 0.97 (2.6)

0.02 (0.32) 0.01 (0.14) 0.005 (0.33) ns

ns ns ns 0.19 (0.83)

ns 0.43 (0.21) 0.07 (0.17) 1.94 (0.14)

0.73 (0.20) 0.79 (0.18) 0.14 (0.15) 6.48 (0.002)

ns ns ns 4.04 (56.8)

ns ns ns ns

ns ns ns ns

0.15 0.11 0.19 0.23

cattle density and rainfall ( p < 0.05) as well as urban fraction ( p ¼ 0.07). E. coli O157:H7 loading was estimated by assuming that a positive result for E. coli O157:H7 corresponded to 1 MPN/ 90 ml. E. coli O157:H7 loading varied from 0 when no organism was detected or there was no flow to the ocean to 108 MPN/d/ km2. Loading was not different from subwatersheds with different primary land cover. Using GEE, E. coli O157:H7 loading was negatively associated with hog density, urban fraction, and temperature (all p < 0.05).

4.

Discussion

Several studies have examined the relationships among land use characteristics and FIB, pathogens, or pathogenic determinants; however, the majority of these have focused on a single watershed or land use type (Walters et al., 2007; Wilkes et al., 2009), or a single pathogen (Haley et al., 2009) or pathogenic determinant (Higgins et al., 2005; Smith et al.,

a

b

2009). This study detected FIB, bacterial pathogens (Salmonella and E. coli O157:H7), a bacterial toxin gene (stx) and molecular markers for host-specific fecal sources in waterbodies surrounded by subwatersheds with a range of land cover and use along the central California coast. Mean concentrations of FIB and Salmonella were higher in waterbodies within primarily urban subwatersheds compared to those within primarily forested subwatersheds. Similarly, their concentrations were positively associated with urban fraction in the bivariate and multivariate analyses. A similar result was obtained when FIB in urban versus nonurban waterbodies were compared in Southern California (Tiefenthaler et al., 2008). Occurrence of stx genes in E. coli was also positively associated with urban fraction, a result consistent with that reported by Higgins et al. (2005) who showed wide-spread occurrence of stx-containing E. coli in metropolitan streams. Fecal indicators, stx, and Salmonella in urban waterbodies may come from a variety of urban point and non-point sources including, but not limited to, leaking sewage infrastructure and both domesticated and wild animal feces (e.g.,

c

d

Fig. 3 e Log-mean areal loading of fecal indicator bacteria and bacterial pathogens from subwatersheds over the 2-year study (mean of log10(X D 1) where X is the areal loading in units provided); a) EC, b) ENT, c) Salmonella, d) E. coli O157:H7. Actual numbers can be found in Table S5.

1760

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 5 2 e1 7 6 2

Table 4 e Generalized estimating equation (GEE) coefficients (b) for areal loading rates. The rainfall variable is a categorical variable. Quasi likelihood under independence model criterion (QIC) is provided for each model.

Intercept Cattle density Hog density Urb fraction Ag fraction Rainfall Salinity Temperature QIC

EC load

ENT load

Sal load

O157 load

10.37a 0.38 0.08 0.29 7.68a 0.57 0.05 0.2a 1730

7.95a 0.39 0.09 0.78 7.14a 1.56a 0.09a 0.16a 2082

2.49a 0.48a 1.0 1.03b 3.63a 0.95a 0.09a 0.57 1131

1.97a 0.10 0.02a 0.66a 0.49 0.49 0.01 0.08a 374

a indicates coefficients significantly different from 0 at p < 0.05. b indicates p < 0.1.

Morabito et al., 2001) (Table S1). The findings suggest that human exposures to waterbodies in highly urbanized areas represent greater risks for salmonellosis, and possibly other enteric illnesses, than waters in less urbanized areas. In contrast to the results for FIB, Salmonella, and stx, occurrence of E. coli O157:H7 was positively associated with agricultural fraction and negatively associated with urban fraction. In the multivariate model, E. coli O157:H7 occurrence was not associated with livestock densities, including cattle densities. This is despite the fact that cattle are a documented host for E. coli O157:H7 (Van Donkersgoed et al., 1999). This suggests non-point sources within agricultural land cover, such as manure applied to cultivated crops, or non-point sources within forested land cover, such as wildlife feces, contribute E. coli O157:H7 to waterbodies in the study area (Table S1). Interestingly, the occurrence of stx-containing E. coli was associated with cattle density in a bivariate comparison, which suggests that cattle may contribute non-O157:H7 STEC to waterbodies. The role of livestock in contributing STEC to waterbodies in this region deserves further study. The presence and age of stormwater and sanitation infrastructure within the subwatersheds likely affect indicator and pathogen concentrations in the waterways, but these variables were not quantified. Watersheds were sewered with separate stormwater and sewage collection systems; however, some rural residences surely had on-site septic systems, and parks within forested land covers had pit latrines (Table S1). Publicly owned treatment works generally discharge via outfalls to the coastal ocean, not to the streams sampled herein. In this region, the number and locations of on-site wastewater treatment systems and pit latrines are not centrally recorded. Documenting the relationship between surface water quality and stormwater and sanitation infrastructure is an area for future work. Concentration is a relevant measure for assessing human exposures and health risks. Areal loading is a measure of the total contribution of contaminants to receiving waters per area of land. In the present study, areal loading represents the number of bacteria contributed to coastal waters by a discharging stream or river per unit area of the subwatershed per unit time. Because subwatershed area was used to calculate an areal load, an inherent assumption in the loading rate is that

pathogens and indicators come primarily from the subwatershed. When areal loading was examined across subwatersheds, primarily agricultural subwatersheds had lower FIB loadings than primarily urban or forested subwatersheds. This finding is supported by the multivariate areal loading models that consistently showed a negative association between FIB and Salmonella loading and the agricultural fraction. A possible explanation for this negative association is that discharge tended to be lower from subwatersheds where there was a higher agricultural fraction (GEE for discharge as a function of agricultural fraction: b ¼ 3.0, p ¼ 0.06), a possible result of the high demand for irrigation water in subwatersheds with agricultural land cover (although a more detailed characterization of water resources is needed to understand the negative association between discharge and land cover). The positive association between Salmonella loading and urban fraction, although at a p value of 0.07, and cattle density supports the previous discussion of the importance of urban inputs of Salmonella to the waterbodies, and also suggests a role for cattle as Salmonella sources. The negative association between urban fraction and E. coli O157:H7 loading suggests forested land cover contributes this bacterium to coastal waterways. The present study used animal densities, animal counts within a subwatershed normalized by the subwatershed area, to explore the associations between livestock and microbial pollution of waterbodies. Future work should consider other methods of normalizing the animal counts (e.g., by particular land covers), the use of livestock units (LSU) as measures of livestock populations, as well as clustering of animals within the subwatershed. This will be possible once more data on livestock and its geographic distribution become available. Such work will further elucidate the influence of livestock on water quality in this region. FIB and Salmonella concentrations and loadings were positively associated with the occurrence of 7-d antecedent rainfall. An increase in Salmonella and FIB detection following rainfall has been reported by others (e.g.,Gaertner et al., 2009; Haley et al., 2009; Setti et al., 2009) and is likely linked to mobilization of bacteria from sources across land covers including soils and compromised infrastructure. Improving riparian buffer strips or stormwater infrastructure will be important in reducing stormwater inputs of bacteria to receiving waters. Higher concentrations of FIB and Salmonella were observed in water samples with the lowest salinity and temperature; E. coli O157:H7 occurrence was associated with lower temperatures. Temperature is a well documented factor influencing survival and persistence of FIB and fecal pathogens in environmental waters (e.g., Noble et al., 2004; Simental and Martinez-Urtaza, 2008), with survival being greatest in colder temperatures presumably due to reduced metabolic activity and predation (Sherr et al., 1988). Lower concentrations in high salinity waters can be explained by dilution of contaminated runoff upon mixing with relatively clean coastal waters. The multivariate statistical models for concentrations and occurrence of FIB and the bacterial pathogens highlight correlative relationships between environmental, climatic, and land use/cover parameters. There predictive power (R2 between 11 and 23%) is on par with other published models of bacterial concentrations in surface waters (Dorner et al., 2006; Whitman and Nevers, 2008). The application of process-based,

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 5 2 e1 7 6 2

deterministic models, such as the Soil Water Assessment Tool (SWAT), to the subwatersheds studied here may further the understanding of the fate and transport processes affecting pathogen and indicator concentrations in runoff (Dorner et al., 2006). The data presented herein (provided in SM) can serve to further improve pathogen predictions from programs like SWAT. Detection of the host-specific Bacteroidales markers was rare and did not correspond to the presence of host animals in the subwatersheds, the occurrence or concentration of pathogens or FIB, nor with land cover. While others have observed an increase in the odds of detecting Salmonella or E. coli O157:H7 when the ruminant-specific markers were present (Walters et al., 2007; Fremaux et al., 2009), the same relationships were not observed in the present study, nor was a significant association between any of the host-specific markers and detection of Salmonella, E. coli O157:H7, or stx genes. These results could be explained by the rare occurrence of the targets overall, or by the broad range of animals that Salmonella and STEC colonize (e.g., domestic pets, livestock, birds, rodents) (Van Donkersgoed et al., 1999; Meerburg and Kijlstra, 2007). The results add to the growing understanding of the limitations and utility of the markers in assessing health risk and microbial source characterization. Interestingly, FIB and Salmonella concentrations were correlated. It appears that in this region, FIB represent good surrogates for Salmonella.

5.

Conclusions

- Fecal indicator bacteria and bacterial pathogen concentrations in 14 waterways along the central California coast are associated with land cover within adjacent subwatersheds, water temperature, water salinity, and antecedent rainfall; antecedent rainfall increases bacterial concentrations by mobilizing sources across subwatersheds. - Sources associated with urban land cover appear to exert the most influence on FIB, stx-containing E. coli, and Salmonella, while sources associated with agricultural and forested land cover have the most influence on E. coli O157:H7 occurrence. - Salinity and temperature are important in modulating bacterial concentration and occurrence; low salinity and temperature both coincided with higher concentrations and occurrence of bacterial targets. - Detection of host-specific Bacteroidales markers did not correlate to pathogen occurrence; however, FIB correlated to the concentration and occurrence of Salmonella. - Associations between Salmonella and stx-containing E. coli and cattle density point to a possible connection between livestock and bacterial pathogens in the waterways.

Acknowledgements This work was supported by the National Research Initiative of the USDA Cooperative State Research, Education and Extension Service, grant number 2007-35102-18139. Authors acknowledge the Boehm lab for sample processing and Amy Pickering for statistical support. Comments provided by two anonymous reviewers improved the manuscript.

1761

Appendix A. Supplementary material Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.watres.2010.11.032

references

Bell, C., Kyriakides, A., 2002. Salmonella: a Practical Approach to the Organism and Its Control in Foods. Blackwell Science Ltd., Oxford, United Kingdom. Boehm, A.B., Yamahara, K.M., Walters, S.P., Layton, B.A., Keymer, D.P., Strickfaden, R.M., Knee, K.L. and Rosener, M. Dissolved inorganic nitrogen, soluble reactive phosphorous, and microbial pollutant loading from tropical rural watersheds in Hawai’i to the coastal ocean during non-storm conditions. Estuaries and Coasts, in press. Byappanahalli, M.N., Sawdey, R., Ishii, S., Shively, D.A., Ferguson, J.A., Whitman, R.L., Sadowsky, M.J., 2009. Seasonal stability of Cladophora-associated Salmonella in Lake Michigan watersheds. Water Research 43 (3), 806e814. California Prime Agricultural Lands, 2008. Famland Mapping and Monitoring Program. State of California Department of Conservation, 2004e2008. http://www.conservation.ca.gov/ dlrp/FMMP/Pages/Index.aspx (accessed 28.08.10.). Center for Disease Control and Prevention, 2003. Foodborne Diseases. Active Surveillance Network (FoodNet). CDC, Atlanta, GA. Craun, G.F., Calderon, R.L., Craun, M.F., 2005. Outbreaks associated with recreational water in the United States. International Journal of Environmental Health Research 15 (4), 243e262. Dorner, S.M., Anderson, W.B., Slawson, R.M., Kouwen, N., Huck, P. M., 2006. Hydrologic modeling of pathogen fate and transport. Environmental Science & Technology 40 (15), 4746e4753. Edwards, A.C., Kay, D., McDonald, A.T., Francis, C., Watkins, J., Wilkinson, J.R., Wyer, M.D., 2008. Farmyards, an overlooked source for highly contaminated runoff. Journal of Environmental Management 87 (4), 551e559. Fremaux, B., Gritzfeld, J., Boa, T., Yost, C.K., 2009. Evaluation of host-specific Bacteroidales 16S rRNA gene markers as a complementary tool for detecting fecal pollution in a prairie watershed. Water Research 43 (19), 4838e4849. Gaertner, J.P., Garres, T., Becker, J.C., Jimenez, M.L., Forstner, M.R.J., Hahn, D., 2009. Temporal analyses of Salmonellae in a headwater spring ecosystem reveals the effects of precipitation and runoff events. Journal of Water and Health 7 (1), 115e121. Haley, B.J., Cole, D.J., Lipp, E.K., 2009. Distribution, diversity, and seasonality of waterborne salmonellae in a rural watershed. Applied and Environmental Microbiology 75 (5), 1248e1255. Hardin, J.W., Hilbe, J.M., 2003. Generalized Estimating Equations. Chapman and Hall/CRC, New York. Heinitz, M.L., Ruble, R.D., Wagner, D.E., Tatini, S.R., 2000. Incidence of Salmonella in fish and seafood. Journal of Food Protection 63 (5), 579e592. Higgins, J.A., Belt, K.T., Karns, J.S., Russell-Anelli, J., Shelton, D.R., 2005. tir- and stx-positive Escherichia coli in stream waters in a metropolitan area. Applied and Environmental Microbiology 71 (5), 2511e2519. Jamieson, R.C., Gordon, R.J., Tattrie, S.C., Stratton, G.W., 2003. Sources and persistence of fecal coliform bacteria in a rural watershed. Water Quality Research Journal of Canada 38, 33e47. LaLiberte, P., Grimes, D.J., 1982. Survival of Escherichia coli in lake bottom sediments. Applied and Environmental Microbiology 43 (3), 623e628.

1762

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 5 2 e1 7 6 2

Martinez-Urtaza, J., Liebana, E., Garcia-Migura, L., PerezPineiro, P., Saco, M., 2004. Characterization of Salmonella enterica Serovar Typhimurium from marine environments in coastal waters of Galicia (Spain). Applied and Environmental Microbiology 70 (7), 4030e4034. Meerburg, B.G., Kijlstra, A., 2007. Role of rodents in transmission of Salmonella and Campylobacter. Journal of the Science of Food and Agriculture 87 (15), 2774e2781. Miller, M.A., Byrne, B.A., Jang, S.S., Dodd, E.M., Dorfmeier, E., Harris, M.D., Ames, J., Paradies, D., Worcester, K., Jessup, D.A., Miller, W.A., 2009. Enteric bacterial pathogen detection in southern sea otters (Enhydra lutris nereis) is associated with coastal urbanization and freshwater runoff. Vetrinarian Research 41 (1). doi:10.1051/vetres/2009049. Miller, W.A., Miller, M.A., Gardner, I.A., Atwill, E.R., Byrne, B.A., Jang, S., Harris, M., Ames, J., Jessup, D., Paradies, D., Worcester, K., Melli, A., Conrad, P.A., 2006. Salmonella spp., Vibrio spp., Clostridium perfringens, and Plesiomonas shigelloides in marine and freshwater invertebrates from coastal California ecosystems. Microbial Ecology 52 (2), 198e206. Morabito, S., Dell’Omo, G., Agrimi, U., Schmidt, H., Karch, H., Cheasty, T., Caprioli, A., 2001. Detection and characterization of shiga toxin-producing Escherichia coli in feral pigeons. Veterinary Microbiology 82 (3), 275e283. Muniesa, M., Jofre, J., Garcia-Aljaro, C., Blanch, A.R., 2006. Occurrence of Escherichia coli O157:H7 and other enterohemorrhagic Escherichia coli in the Environment. Environmental Science & Technology 40 (23), 7141e7149. National Weather Service. Precipitation data, 2010. http://www. cnrfc.noaa.gov/rainfall_data.php (accessed 27.08.10). Natural Resources Conservation Service, 2009. California Hydrologic Unit Boundary Layer (12-digit). United States Department of Agriculture, Fort Worth, TX. NOAA Coastal Services Center, 2001. Coastal Change Analysis Program (C-CAP). http://www.csc.noaa.gov/digitalcoast/data/ ccapregional/index.html (accessed 27.08.10). Noble, R.T., Lee, I.M., Schiff, K.C., 2004. Inactivation of indicator micro-organisms from various sources of faecal contamination in seawater and freshwater. Journal of Applied Microbiology 96 (3), 464e472. Russ, G.R., Stockwell, B., Alcala, A.C., 2005. Inferring versus measuring rates of recovery in no-take marine reserves. Marine Ecology Progress Series 292 (1), 1e12. Schaffter, N., Parriaux, A., 2002. Pathogenic-bacterial water contamination in mountainous catchments. Water Research 36 (1), 131e139. Schriewer, A., Miller, W.A., Byrne, B.A., Miller, M., Oates, S., Conrad, P.A., Hardin, D., Yang, H.-H., Chouicha, N., Melli, A., Jessup, D., Dominik, C., Wuertz, S., 2010. Presence of Bacteroidales as a predictor of pathogens in surface waters of the central California coast. Applied and Environmental Microbiology 76 (17), 5802e5814. Setti, I., Rodriguez-Castro, A., Pata, M.P., Cadarso-Suarez, C., Yacoubi, B., Bensmael, L., Moukrim, A., Martinez-Urtaza, J., 2009. Characteristics and dynamics of Salmonella contamination along the coast of Agadir, Morocco. Applied and Environmental Microbiology 75 (24), 7700e7709.

Shellenbarger, G.G., Athearn, N.D., Takekawa, J.Y., Boehm, A.B., 2008. Fecal indicator bacteria and Salmonella in ponds managed as bird habitat, San Francisco Bay, California, USA. Water Research 42 (12), 2921e2930. Sherr, B.F., Sherr, E.B., Rassoulzadegan, F., 1988. Rates of digestion of bacteria by marine phagotrophic protozoa: temperature dependence. Applied and Environmental Microbiology 54 (5), 1091e1095. Simental, L., Martinez-Urtaza, J., 2008. Climate patterns governing the presence and permanence of salmonellae in coastal areas of Bahia de Todos Santos, Mexico. Applied and Environmental Microbiology 74 (19), 5918e5924. Smart, T.S., Hirst, D.J., Elston, D.A., 1999. Methods for estimation loads transported by rivers. Hydrology and Earth System Sciences 3 (2), 295e303. Smith, C.J., Olszewski, A.M., Mauro, S.A., 2009. Correlation of Shiga toxin gene frequency with commonly used microbial indicators of recreational water quality. Applied and Environmental Microbiology 75 (2), 316e321. Sprague, L.A., 2001. Effects of storm-sampling frequency on estimation of water-quality loads and trends in two Tributaries to Chesapeake Bay in Virginia, USGS waterresources investigations report, 01e4136. Tiefenthaler, L., Stein, E.D., Schiff, K.C., 2008. Origins and Mechanisms of Watershed and Land Use Based Sources of Fecal Indicator Bacteria in Urban Stormwater. In: Weisberg, S. B., Miller, K. (Eds.). Southern California Coastal Water Research Project, Costa Mesa, CA, pp. 153e161. USEPA, 2002a. Method 1604: Total Coliforms and Escherichia coli in Water by Membrane Filtration Using a Simultaneous Detection Technique (MI Medium) (Washington, DC). USEPA, 2002b. Method 1600: Enterococci in Water by Membrane Filtration Using Membrane-enterococcus Indoxyl-B-Dglucoside Agar (MEI) (Washington, DC). USEPA, 2006. Method 1682: Salmonella in Sewage Sludge (Biosolids) by Modified Semisolid Rappaport-vassiliadis (MSRV) Medium. In: Water, O.o. (Ed.) (Washington, DC). USGS, 2001. Multi-resolution Land Characteristics Consortium at USGS: Urban Imperviousness Raster. http://www.mrlc.gov/ nlcd_multizone_map.php (accesed). USGS, 2010. Discharge Data. http://waterdata.usgs.gov (accesed 27.08.10). Van Donkersgoed, J., Graham, T., Gannon, V., 1999. The prevalence of verotoxins, Escherichia coli O157:H7, and Salmonella in the feces and rumen of cattle at processing. The Canadian Veterinary Journal 40 (5), 332e338. Walters, S.P., Gannon, V.P.J., Field, K.G., 2007. Detection of Bacteroidales fecal indicators and the zoonotic pathogens E. coli O157:H7, Salmonella, and Campylobacter in river water. Environmental Science & Technology 41 (6), 1856e1862. Whitman, R.L., Nevers, M.B., 2008. Summer E. coli patterns and responses along 23 Chicago beaches. Environmental Science & Technology 42 (24), 9217e9224. Wilkes, G., Edge, T., Gannon, V., Jokinen, C., Lyautey, E., Medeiros, D., Neumann, N., Ruecker, N., Topp, E., Lapen, D.R., 2009. Seasonal relationships among indicator bacteria, pathogenic bacteria, Cryptosporidium oocysts, Giardia cysts, and hydrological indices for surface waters within an agricultural landscape. Water Research 43 (8), 2209e2223.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 6 3 e1 7 7 3

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Biogenic metals for the oxidative and reductive removal of pharmaceuticals, biocides and iodinated contrast media in a polishing membrane bioreactor Ilse Forrez a, Marta Carballa b, Guido Fink c, Arne Wick c, Tom Hennebel a, Lynn Vanhaecke d, Thomas Ternes c, Nico Boon a, Willy Verstraete a,* a

Laboratory of Microbial Ecology and Technology (LabMET), Faculty of Bioscience Engineering, Ghent University, Coupure Links 653, B-9000 Gent, Belgium b Department of Chemical Engineering, School of Engineering, University of Santiago de Compostela, Ru´a Lope Go´mez de Marzoa s/n, E-15782 Santiago de Compostela, Spain c Federal Institute of Hydrology (BfG), Am Mainzer Tor 1, D-56068 Koblenz, Germany d Laboratory of Chemical Analysis, Department Veterinary Public Health and Food Safety, Faculty of Veterinary Medicine, Ghent University, Salisburylaan 133, B-9820 Merelbeke, Belgium

article info

abstract

Article history:

Pharmaceutical and personal care products, biocides and iodinated contrast media (ICM)

Received 16 June 2010

are persistent compounds, which appear in ng to mg L
1 in secondary effluents of sewage

Received in revised form

treatment plants (STPs). In this work, biogenic metals manganese oxides (BioMnOx) and

17 November 2010

bio-palladium (Bio-Pd) were applied in lab-scale membrane bioreactors (MBR) as oxidative

Accepted 22 November 2010

and reductive technologies, respectively, to remove micropollutants from STP-effluent.

Available online 30 November 2010

From the 29 substances detected in the STP-effluent, 14 were eliminated in the BioMnOxMBR: ibuprofen (>95%), naproxen (>95%), diuron (>94%), codeine (>93%), N-acetyl-sulfa-

Keywords:

methoxazole (92%), chlorophene (>89%), diclofenac (86%), mecoprop (81%), triclosan

Biodegradation

(>78%), clarithromycin, (75%), iohexol (72%), iopromide (68%), iomeprol (63%) and sulfa-

Emerging contaminants

methoxazole (52%). The putative removal mechanisms were the chemical oxidation by

Micropollutants

BioMnOx and/or the biological removal by Pseudomonas putida and associated bacteria in

Nanoparticles

the enriched biofilm. Yet, the removal rates (highest value: 2.6 mg diclofenac L
1 d
1) need

Polishing techniques

to improve by a factor 10 in order to be competitive with ozonation. ICM, persistent towards oxidative techniques, were successfully dehalogenated with a novel reductive technique using Bio-Pd as a nanosized catalyst in an MBR. Iomeprol, iopromide and iohexol were removed for >97% and the more recalcitrant diatrizoate for 90%. The conditions favorable for microbial H2-production enabling the charging of the Pd catalyst, were shown to be important for the removal of ICM. Overall, the results indicate that Mn oxide and Pd coupled to microbial catalysis offer novel potential for advanced water treatment. ª 2010 Elsevier Ltd. All rights reserved.

* Corresponding author. Tel.: þ329 264 59 76; fax: þ329 264 62 48. E-mail address: [email protected] (W. Verstraete). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.11.031

1764

1.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 6 3 e1 7 7 3

Introduction

Public health facilities and general welfare increase our quality of life, but they constitute a major burden for water resources through the constant input of trace amounts of drugs, cosmetics, UV blockers, fragrances, insect repellants, etc. This problem is challenging, since thousands of different substances are being used in medicine, agriculture, personal care and industrial processes (Ternes et al., 2004). Joss et al. (2006) suggested a classification scheme to characterize the biological degradation of 35 compounds, illustrating the poor efficiency of sewage treatment plants (STP) in degrading pharmaceutical and personal care products (PPCPs). Antibiotics, anti-epileptics, antiphlogistics, iodinated contrast agent and lipid regulators are removed between 20 and 90% in nutrient removing STPs. Iodinated contrast media (ICM) are widely used in human medicine for imaging of organs or blood vessels during diagnostic tests (Pe´rez and Barcelo, 2007) and occur in secondary effluent at several mg L
1 (Ternes and Hirsch, 2000). Organic contaminants used in agriculture such as biocides and veterinary pharmaceuticals enter the environment in a more diffuse way through application of manure and water runoff. Several studies, which report the toxicity of secondary effluents (Dizer et al., 2002; Aguayo et al., 2004; Cao et al., 2009), describe sublethal effects for aquatic organisms. Therefore, the search for mitigation technologies and strategies for wastewater reclamation to protect the aquatic environment is of uttermost importance. Ozonation and advanced oxidation processes (AOPs), have been applied during the last decade for the abatement of pollution caused by the presence of residual pharmaceuticals in water and wastewaters (Klavarioti et al., 2009). Although ozonation is an effective technique to oxidize pharmaceuticals from wastewater effluents (Ternes et al., 2003), its application in wastewater management would result in an increased cost of 12% (Hollender et al., 2009). Moreover, an increase in genotoxicity may be expected after ozonation because of the formation of nitrosamines, hydroxylamines and bialdehydes as reported by Guzzella et al. (2002), Schmidt and Brauch (2008) and Benner and Ternes (2009). Recently, manganese oxides, have been applied to oxidatively remove different kinds of organic micropollutants, including antibacterials and related compounds with phenolic and fluoroquinolonic moieties, aromatic N-oxides, tetracyclines (Zhang et al., 2008a), estrogenic compounds such as the synthetic hormone 17a-ethinylestradiol (Sabirova et al., 2008) and the anti-inflammatory drug diclofenac (Forrez et al., 2010). A mechanism involving sorption of the compound to the oxide surface and subsequent electron transfer has been proposed (Zhang and Huang, 2005). In this context, biologically produced manganese oxides (BioMnOx) offer perspectives due to their characteristics, i.e. high specific surface areas (98e224 m2 g
1) (Hennebel et al., 2009a) and the ability of the manganese-oxidizing bacteria to reoxidize the formed MnII, thus increasing the reactivity of the Mn oxides with a factor 10 (Forrez et al., 2010). Highly substituted aromatic compounds such as ICM are hard to oxidize and therefore recalcitrant towards ozonation (Ternes et al., 2003; Hollender et al., 2009; Huber et al., 2005).

Recently, the reductive hydrodehalogenation of diatrizoate with supported Pd and porous Ni catalysts (Knitt et al., 2008) and with biologically produced nanopalladium catalyst (BioPd) (Hennebel et al., 2010) was reported. In the presence of a hydrogen donor (H2 or HCOOH), Pd0 nanoparticles become charged with molecular hydrogen, which can subsequently reduce halogenated compounds (Hennebel et al., 2009b). In this work, both the biogenic metals BioMnOx and Bio-Pd were evaluated for their practical feasibility to polish secondary effluent. More specifically, STP-effluent was treated in lab-scale membrane bioreactors (MBR) to remove antiinflammatory drugs, tranquilizers, anti-epileptics, lipid regulators, antibiotics, iodinated contrast agent and biocides. Special attention was paid to matrix effects and the effect of low nutrient concentrations on the reactivity of the biogenic Mn oxide and biological reoxidation of MnII. A novel reductive technology using nano-Pd on a microbial carrier (Bio-Pd) was tested for the removal of several ICM at environmental concentrations in a continuous MBR.

2.

Experimental procedures

2.1.

Biogenic metals

Biogenic manganese oxides (BioMnOx) were produced by means of an axenic culture of Pseudomonas putida MnB6 (BCCM/LMG 2322) as described by Forrez et al. (2010). Biopalladium (Bio-Pd) was produced by the metal-respiring bacterium Shewanella oneidensis (BCCM/LMG 19005) as described by De Windt et al. (2005). Bacterial cells can reduce PdII and subsequently precipitate it as Pd0 nanocrystals on their cell wall and in their periplasmatic space.

2.2.

Batch tests

To evaluate the matrix effects of sewage treatment plant (STP) effluents and the effect of low nutrient concentrations on the pollutant removal kinetics and the biological activity of the manganese-oxidizing bacteria, batch tests were performed according to Forrez et al. (2010). In brief, for matrix effects, STPeffluent buffered at pH 6.8 with phosphate (10 mM) or HEPES (0.1 mM) (4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid) was added to BioMnOx pellets up to a volume of 150 mL and spiked with diclofenac (3 mg L
1). To determine the effect of low nutrient availability, BioMnOx was stored aerobically at 28  C for 45 days. To evaluate the removal of micropollutants with BioMnOx at environmental concentrations, a batch test with non-spiked STP-effluent and BioMnOx was performed in a 30 L-glass vessel, aerated continuously with compressed air to provide dissolved oxygen levels of 6e8 mg L. A 3 L-BioMnOx-suspension was prepared at a concentration of 55 mg MnIV L
1 as described by Forrez et al. (2010) and diluted with 27 L of STPeffluent buffered at pH 6.8 with 20 mg L
1 HEPES. For micropollutant analysis, samples of 1 L were taken 3 times a week and filtered over a 0.45 mm filter (Whatman) and stored in glass bottles at 4  C prior to solid-phase extraction (SPE) and liquid chromatography - multiple mass spectrometry (LC-MS2)

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 6 3 e1 7 7 3

analysis. Dissolved Oxygen (DO) and pH were monitored daily and adjusted where necessary.

2.3.

BioMnOx-reactors

Two lab-scale membrane bioreactors (MBR) were constructed out of a Diapes BLS517G-module and an aeration vessel (0.2 L). The outer membrane compartment (177 mL) was operated with 28 mg MnIV of BioMnOx or 160 mg MnIV L
1. Because of the densely packed hollow fibres, BioMnOx was in close contact with the membrane surface and was retained physically in the reactor (Fig. 1A). The polyethersulfone hollow fibres had a mean pore diameter of 8e10 nm and total membrane surface of 1.7 m2. A recycle (100 mL min
1; Watson Marlow pump) was applied over the outer compartment and the recycle liquid was aerated to saturate it with oxygen (8.4e8.8 mg O2 L
1). Since a pH in the range of 6.0e6.8 was shown to enhance the reactivity of the biogenic Mn oxides (Forrez et al., 2010), the pH was controlled at 6.5 (Consort pH controller, 0.1 mM HCl). The reactors were operated for a period of 45 days on secondary wastewater effluent, retrieved from a full-scale conventional activated sludge STP of 175 000 Inhabitant Equivalents (IE). Parameters of the grab samples were:
1 45e65 mg COD L
1; <0.1 mg PO3
4 -P L ; ammonium and nitrate concentrations were respectively 2.8e3.4 and 7.5e9.5 mg N L
1 in the first period (day 0e14) and 1.9 and 4.5 mg N L
1 from day 18 on. The reactors were operated semi-continuously by supplying secondary wastewater effluent for 5 min h
1 at different hydraulic residence times (HRT). Different HRTs were applied by successively altering the flux from 0.5 L m
2 h
1 (HRT 5 h) to 1.0 L m
2 h
1 (HRT 2.5 h) and 0.1 L m
2 h
1 (HRT 24 h)

1765

from day 0e14, day 14e18 and day 18e45, respectively. One reactor was running on STP-effluent spiked with 3 mg L
1 diclofenac and non-spiked STP-effluent was supplied to the other MBR reactor. For diclofenac analysis, samples (2 mL) were taken every 2 days from the spiked reactor and stored at 4  C after filtration over a 0.22 mm filter (Millipore) prior to diclofenac analysis. From the non-spiked reactor, samples were taken twice a week and stored in 1 L-glass bottles at 4  C prior to SPE enrichment and LC-MS/MS analysis. Samples were taken for MnII analysis in the effluent by atomic absorption spectrophotometer.

2.4.

Bio-Pd-reactors

The setup of the Bio-Pd MBRs was similar to the BioMnOx-MBR reactors (Fig. 1B), but containing physically retained Bio-Pd (25 mg Pd0) in the outer compartment of 177 mL or 141 mg Pd0 L
1. No pH control or aeration was installed, but a hydrogen donor was dosed continuously to the influent in the form of sodium formate (25 mM) to charge the Bio-Pd-catalyst with molecular hydrogen. One reactor was running on STPeffluent spiked with 6 mg L
1 diatrizoate and non-spiked STPeffluent was supplied to the other MBR reactor. The reactors were operated for a period of 35 days and at different HRTs, i.e. 2.5 h (day 0e14) and 5 h (day 14e35). A second test run of 45 days was performed with the inoculation of a H2-gas producing bacterium (Clostridium butyricum (BCCM/LMG 1217). This bacterium was inoculated at the beginning of the test run to produce H2 in situ from formate (10 mM) under anoxic conditions (Nandi and Sengupta, 1998). The HRTs applied were 5 h (day 0e14); 2 h (day 14e19) and 24 h (day 19e45). Sampling and monitoring was performed similar to the BioMnOx reactors.

Fig. 1 e Configuration of the MBR modules with BioMnOx (A) or Bio-Pd (B) in the outer compartment.

1766

2.5.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 6 3 e1 7 7 3

Analytical methods

Diclofenac and diatrizoate at mg L
1 concentrations were analyzed with HPLC-UV-DAD as described in Forrez et al. (2010) and Hennebel et al. (2010), respectively. The acidic pharmaceuticals and ICM in STP-effluent were enriched and analyzed according to Lo¨ffler and Ternes (2003) and Ternes and Hirsch (2000), respectively. The analytical methods for the anti-epileptic and psycho-active drugs, antibiotics and biocides were described in detail in Hummel et al. (2006), Lo¨ffler and Ternes (2003) and Wick et al. (2010a), respectively. The SPE-extracts were injected into an Agilent 1100 LC system with degasser, quaternary pump and autosampler (Agilent Technologies, Waldbronn, Germany). Details are provided in the Supporting Information (Text S1). To measure the loss of manganese (MnII) and palladium (Pd0) from the reactors, 10 mL-samples were collected, acidified with 1% of concentrated HNO3 (7 M) and stored at 4  C prior to analysis in an AA-6300 atomic absorption spectrophotometer (AAS) (Schimadzu). The limits of detection were 0.05 mg Mn L
1 and 0.5 mg Pd L
1. Acetate was analyzed by gas chromatography as described by Vanhaecke et al. (2009) and formate, nitrate and phosphate by IC, in a Metrohm 761 compact Ion Chromatograph equipped with a conductivity detector. Ammonium and COD were determined colorimetrically according to standard methods (Eaton et al., 2005).

3.

Results

3.1. Fate of pharmaceuticals, biocides and ICM in a BioMnOx-MBR Table 1 and Fig. 2 show the removal of diclofenac in both MBRs operated respectively on STP-effluent spiked with 3 mg diclofenac L
1 and on non-spiked STP-effluent. The feed of the latter was not altered, resulting in a variable influent diclofenac concentration, i.e. 790  53 ng L
1 and 390  8 ng L
1 of diclofenac for the period 0e14 and 14e45 days, respectively. The removal of diclofenac in both MBRs was initially more than 90% and decreased during the subsequent 14 days. In the spiked reactor, the diclofenac removal decreased and stabilized at 30% during the first 10 days at an HRT of 5 h (Fig. 2A). When the HRT was lowered to 2.5 h, the removal efficiency decreased further, but it recovered and stabilized at 60% when the HRT increased from 2.5 h to 24 h. In the non-spiked reactor, the removal of diclofenac was much higher at an HRT of 5 h, i.e. 70% (Fig. 2B). In the subsequent period, the removal decreased due to the lower HRT and the desorption effects resulting from the lower influent concentration, but it was quickly restored at 85% when the HRT was increased to 24 h. From the group of the pharmaceuticals detected in the STP-effluent, mainly the anti-inflammatory drugs, i.e. diclofenac (67e86%), ibuprofen (95%) and naproxen (95%), and some analgesics, i.e. codeine (79e93%), dihydrocodeine (41%) and morphine (60%) were removed in the BioMnOx-MBR (Table 1). The ICM, iopromide (68%), iomeprol (63%) and iohexol (72%) were only removed at the long HRT of 24 h. The concentrations of the biocides were very low (around 50 ng L
1), but a clear removal trend could be observed from

the data because the effluent concentrations of the BioMnOxMBR were below limit of quantification representing removal efficiencies higher than 70%.

3.2.

Batch tests with BioMnOx

In the 30 L-batch test with STP-effluent and BioMnOx, a limited number of compounds was removed (Table 1). In Fig. 3, the relatively fast (>60% in 5 days) removal of triclosan and diclofenac and the slower removal of codeine and ibuprofen are presented. Because the biodegradation of the analgesic codeine by Pseudomonas putida M10 has been reported before (Lister et al., 1999), samples of the 30 L-batch test with BioMnOx were screened for codeine transformation products. Two transformation products were detected, which have previously been identified as 14-hydroxycodeinone and 14-hydroxycodeine by Wick et al. (2010b). A negative influence of the STP-effluent matrix on the reactivity of BioMnOx was observed in batch tests (Fig. 4A). Phosphate and HEPES buffered STP-effluents decreased reactivity towards diclofenac oxidation (pseudo first order reaction rate constant k: 0.0042 h
1 and 0.0121 h
1, respectively) by a factor of 6 and 2, respectively, compared to phosphate buffered distilled water (k: 0.0245 h
1). The sorbed MnII was relatively low (20 mg MnII g
1 MnIV) in all assays, except for the phosphate buffered STP-effluent (185 mg MnII g
1 MnIV). In the latter, a white precipitate was observed and characterized with EDX as a calcium phosphate. The use of BioMnOx after 45 days aerobic storage at 28  C with low nutrient (N and C) supply resulted in a decreased diclofenac oxidation (k: 0.0149 h
1) compared to fresh BioMnOx (k: 0.0245 h
1) (Fig. 4B). Moreover, taking into account the different MnIV concentrations, the reactivity of BioMnOx towards diclofenac oxidation was therefore decreased by a factor of 3 after storage for 45 days. The latter can be explained by the decreased biological reoxidation of the MnII, generated during the oxidation of diclofenac. The amount of sorbed MnII was 95 mg MnII g
1 MnIV after 45 days low nutrient operation compared to 60 mg MnII g
1 MnIV in the control. The initial concentration of diclofenac in the batch tests (3.0 mg L
1 vs 663 ng L
1) did not result in a difference in the reaction rate constant k, which was 0.0121 and 0.0137 h
1, respectively (Fig. 4C).

3.3.

Reductive dehalogenation of ICM in a Bio-Pd-MBR

Two MBR modules with Bio-Pd in the outer membrane compartment were operated with STP-effluent. The feed of one MBR was spiked with 6 mg diatrizoate L
1 and the feed of the other MBR was not altered (non-spiked STP-effluent). Table 2 shows the ICM concentrations in the STP-effluent during the experimental period. The Bio-Pd-MBR treating STP-effluent spiked with 6 mg L
1 diatrizoate, showed low and variable removal efficiencies between 0 and 50% at an HRT of 2.5 h (Fig. 5A). When the HRT was increased to 5 h, the elimination of diatrizoate improved up to 94% but subsequently declined in 14 days to 0%. This was accompanied with an increased removal of formate (from 200 to 1000 mg L
1) and subsequent production of acetate (183e194 mg L
1). A similar behaviour was observed in the

1767

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 6 3 e1 7 7 3

Table 1 e Concentrations (ng LL1) of pharmaceuticals, ICM and biocides detected in STP-effluent before (day 0) and after polishing in a batch reactor with 3.4 mg MnIV BioMnOx LL1 (day 13), respectively before (CInfluent) and after (CEffluent) continuous treatment in a BioMnOx-MBR (0.16 g MnIV LL1) during a 45 days test run. Grey shades indicates >60% removal or below limit of quantification (LOQ) in the effluent. ‘<’: below LOQ. Desorption on day 17 indicated with yes if CEffl>CInfluent. Compound

BioMnOx batch test Day 0

Day 13

BioMnOx-reactor Day 0e14 with HRT 5 h CInfluenta

CEffluent

a

Desorption on day 17: CEffl

Day 18e45 with HRT 24 h CInfluentb

CEffluentb

Anti-inflammatory drugs Diclofenac Ibuprofen Naproxen

663 158 278

31 <8 303

790  53 191  32 437  38

254 (d14) <10 19 (d14)

Yes: 506 No: <10 No: 17

390  8 50  3 185  13

56  2 52 (d42) <10

Tranquilizers Diazepam Nordiazepam Oxazepam

351 <10 146

304 <10 122

<10 17  1 158  10

<10 <10 110 (d14)

No: <10 Yes: 127

<10 <10 51  7

<10 <10 81 (d42)

Analgesics Codeine Dihydrocodeine Morphine Tramadol Methadone

247 41 16 <10 19

82 36 15 <10 16

250  4 41  1 24  9 516  27 21  1

52  6 24  1 <10 533  51 18  1

No: 52 No: 19 No: <10 Yes: 332 No: 14

151  19 20  3 <10 <10 <10

<10 <10 <10 <10 <10

Anti-epileptics Carbamazepine 10,11-Dihydroxy-10, 11-dihydrocarbamazepine 10,11-Dihydrocarbamazepine Primidone

855 1240

709 1076

979  27 1582  59

926 (d14) 1494  43

Yes: 910 Yes: 1027

275  30 494  41

477(d35)/427(d42) 339(d35)/335(d42)

20 95

19 77

23  1 84  9

18  4 84  7

Yes: 21 Yes: 54

<10 32  1

<10 37  5

Lipid regulators Bezafibrate

23

<8

22  1

<10

No: <10

<10

<10

Antibiotics Sulfamethoxazole N-Acetyl-sulfamethoxazole Trimethoprim Erythromycin Clarithromycin

259 90 71 76 259

256 79 58 55 213

318  14 91  2 75  7 91  3 317  6

151  18 60  13 71  5 63  13 85  18

Yes: 174 No: 71 Yes: 69 No: 12 No: 68

119  9 105  4 32  2 12  2 217  16

85 (d42) 13  1 26  1 13  1 86  8

Iodinated contrast media (ICM) Iopromide Iomeprol Iohexol Diatrizoate

454 1445 488 7725

322 1056 424 7725

460  12 1541  57 609  47 6852  216

354  46 1265  93 504  97 6229  267

No: 413 Yes: 1053 Yes: 349 Yes: 4393

454  41 701  30 224  32 3010  617

142  27 257  23 61  4 3427  144

Biocides Triclosan Diuron Mecoprop Chlorophene

40 59 127 55

<25 58 121 <14

41  8 67  5 135  2 80  24

<30 <4 125  20 <14

No: <30 No: 10 No: 72 No: <14

112  26 80  3 55  6 93  25

<25 84 <10 <10

a Average concentrations are calculated using data of day 3, 7, 10 and 14. Otherwise, the day is indicated between brackets (e.g. d14). b Average concentration using data of day 21, 35 and 42. Otherwise, the day is indicated between brackets (e.g. d42).

non-spiked Bio-Pd reactor. More than 90% of the ICM was removed at the start of the operation, at an HRT of 2.5 h, and after a subsequent decrease, removal was restored to >96% by increasing the HRT from 2.5 to 5 h (Fig. 5B) and simultaneous removal of formate and production of acetate (144e186 mg acetate L
1).

In the second experiment with Clostridium butyricum, a H2 producing bacterium able to produce H2 from formate, a good removal (63e97%) was obtained in the non-spiked reactor at an HRT of 5 h (Fig. 6B), but this result was lost at an HRT of 24 h (13e52%). Similar results were observed in the Bio-Pd-MBR spiked with 6 mg L
1 of diatrizoate. Variable (30e90%)

1768

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 6 3 e1 7 7 3

Fig. 2 e Continuous removal of diclofenac in BioMnOx-MBR. The HRT (h) is indicated on the graph. Diclofenac removal efficiency at spiked influent concentrations of 3 mg LL1 (A). Diclofenac removal efficiency in non-spiked STP-effluent (B) (see Table 1 for diclofenac concentrations).

diatrizoate removal occurred at an HRT of 5 h and almost no removal (10%) in the subsequent period at an HRT of 24 h (Fig. 6A). Formate was removed by more than 50%, but no acetate was detected in the effluent.

4.

Discussion

4.1.

Removal at environmental relevant concentrations

The use of biogenic metals for micropollutant removal is only valuable if they can remove several pollutants at environmentally relevant concentrations (ng-mg L
1). The diclofenac removal kinetics with BioMnOx were not influenced at 663 ng L
1 and 3 mg L
1 in STP-effluent indicating that the technology can work at low environmental relevant pollutant concentrations. In the batch test with STP-effluent, diclofenac (95%), ibuprofen (>95%), codeine (67%), bezafibrate (>65%),

Fig. 3 e Relative concentrations of diatrizoate (3), carbamazepine (B), codeine (6), ibuprofen (>), diclofenac (,) and triclosan ( ) with BioMnOx (3.4 mg MnIV LL1) in HEPES buffered STP-effluent (pH: 6.5). C0 -concentrations are reported in Table 1.

*

triclosan (>40%) and chlorophene (>75%) were removed with BioMnOx. In the BioMnOx-MBR, diclofenac was removed at environmentally relevant concentrations (400e800 ng L
1) at removal efficiencies of 68% at an HRT of 5 h and 85% at an HRT of 24 h. As a consequence of increasing the HRT, the effluent quality improved. The influence of the HRT on the continuous removal of the biocides triclosan (>78%) and chlorophene (>89%) was less obvious due to the low concentrations (50e100 ng L
1). Successful dehalogenation of diatrizoate at mg L
1 has been shown before by Knitt et al. (2008) with chemical Pd and by Hennebel et al. (2010) with Bio-Pd. In this study, we demonstrated that this is also possible in a continuous manner and for several ICM at environmentally relevant concentrations.

4.2.

Removal mechanisms with BioMnOx

In the batch test with BioMnOx, in total 6 compounds (diclofenac, triclosan, chlorophene, ibuprofen, codeine and bezafibrate) were removed by more than 60%. It is known that diclofenac, chlorophene and triclosan are susceptible to chemical oxidation by MnO2 (Zhang et al., 2008a; Forrez et al., 2010). Ibuprofen is the compound with the highest degrada
1 tion constant (Table S2, kbiol: 9e35 L g
1 SS d ) and therefore likely to be degraded by Pseudomonas putida (Kagle et al., 2009). The two codeine transformation products (14-hydroxycodeinone and 14-hydroxycodeine) detected in the 30 L-batch test with BioMnOx have been described in literature as biological metabolites of codeine upon degradation by Pseudomonas putida M10 (Lister et al., 1999). Therefore, codeine was likely to be subjected to biological removal by the manganeseoxidizing bacteria present in the BioMnOx. During continuous application in the BioMnOx-MBR, removal efficiencies of more than 60% were observed for a total of 14 compounds, which are higher than in the batch test. This is an indication that other factors contribute to the removal in a continuous system such as adsorption to

1769

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 6 3 e1 7 7 3

Fig. 4 e STP-effluent matrix effects (A), effect of low nutrient levels (B) and effect of initial diclofenac concentration (C) on the reactivity of BioMnOx towards diclofenac oxidation. (A) Phosphate buffered distilled water (-, pH: 6.5); Phosphate buffered STP-effluent (3, pH: 6.6); HEPES buffered STP-effluent (B, pH 6.5); non-buffered STP-effluent (6, pH 7.4). (B) BioMnOx after 45 days of low nutrient supply (,, 5.8 mg MnIV LL1) and control with newly made BioMnOx (-, 2.8 mg MnIV LL1) (C) HEPES buffered STP-effluent spiked with diclofenac (B, C0: 3.0 mg diclofenac LL1); HEPES buffered STP-effluent (C, C0: 663 ng diclofenac LL1).

membranes and diversification of the biofilm by enrichment with specialized bacteria. Naproxen (96% and >95%), dihydrocodeine (41% and >50%), morphine (60%), sulfamethoxazole (52% and 34%), N-acetyl-sulfamethoxazole (34% and 92%) clarithromycin (73% and 60%) and diuron (>94% and 90%) were removed at an HRT of 5 h and 24 h, respectively. At an HRT of 24 h, mecoprop (>81%), iopromide (68%), iomeprol (63%) and iohexol (72%) were removed additionally. Naproxen and diuron are structurally similar to diclofenac and triclosan, respectively, as illustrated in Table S3 in Supporting Information. Chemical oxidation by BioMnOx is therefore possible. Dihydrocodeine and morphine structurally resemble codeine, suggesting biological degradation by manganese-oxidizing bacteria of these compounds. The human metabolite of sulfamethoxazole, N-acetyl-sulfamethoxazole is more biodegradable than the other antibiotics (kbiol: 6.0e8.0, Table S2, Supporting Information). N-acetyl-sulfamethoxazole can be biologically transformed to sulfamethoxazole (Go¨bel et al., 2007), underestimating the removal of the latter. The antibiotic clarithromycin was removed at higher efficiencies (75% and 60% at HRT of 5 and 24 h) than the structural similar erythromycin (31% and 0% at HRT of 5 and 24 h). In a previous study, the macrolide antibacterial agents clarithromycin and roxithromycin were found to adsorb strongly to FeIII and MnIV oxides and adsorption was accompanied by slow oxidation (Feitosa-Felizzola et al., 2009). Because removal efficiency is not improved at higher HRT, adsorption is possibly the dominant removal mechanism for clarithryomycin.

Iodinated contrast media such as iopromide, iomeprol, iohexol and diatrizoate are known to be very recalcitrant towards oxidation due to the high degree of substitution of the benzene ring (Ternes and Hirsch, 2000). Yet, in the BioMnOxMBR, during the period with high HRT of 24 h, iomeprol, iohexol and iopromide were removed by 63, 75 and 69%, respectively. Because diatrizoate is not removed, it can be concluded that the side chains of the amide functional groups of iopromide, iomeprol and iohexol were affected (Table S3). Either this was by the oxidation action of the BioMnOx or by biological degradation. The latter has been described in literature. Dehydroxylation metabolites of iopromide were identified by Pe´rez et al. (2006) and the biologically mediated transformations of the side chains were proposed by Schulz et al. (2008).

4.3.

Challenges of BioMnOx

The oxidation activity of BioMnOx towards diclofenac removal was decreased by a factor 2 when shifting from clean water to a real effluent matrix buffered at pH 6.5. The presence of organic and inorganic ions in an effluent matrix can interfere with the active surface sites of the MnO2 (Zhang et al., 2008a). In non-buffered STP-effluent, the pH increased up to 7.5 and diclofenac oxidation by BioMnOx was decreased by a factor 6 compared to pH-buffered distilled water. The pH of STP-effluent was 6.5 with a phosphate buffer. Yet, removal of diclofenac was decreased similarly as when no buffer was applied. Phosphate ions precipitated with Ca2þ from the

Table 2 e ICM influent concentrations (ng LL1) in STP-effluent treated in Bio-Pd MBR with HCOOH as hydrogen donor and Clostridium butyricum as H2 producing bacterium. If n > 2, mean values and standard deviation are given. HCOOH þ Clostridium butyricum

HCOOH

Iopromide Iomeprol Iohexol Diatrizoate

Day 0e14 HRT 2.5 h

Day 14e35 HRT 5 h

Day 0e14 HRT 5 h

Day 21e45 HRT 24 h

361 1833 94 6883

345 277 246 4608

574  108 1530  262 475  93 9600  908

467  23 701  47 219  15 2283  204

1770

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 6 3 e1 7 7 3

Fig. 5 e Continuous removal of diatrizoate, iomeprol, iopromide and iohexol in Bio-Pd-MBR with formate to charge the catalyst for the reductive dehalogenation of ICM. The HRT (h) is indicated on the graph. Diatrizoate removal efficiency at spiked influent concentrations of 6 mg LL1 (A). ICM removal efficiency in non-spiked STP-effluent (B).

matrix. The white Ca-phosphate precipitates did not influence the oxidation activity of BioMnOx directly. However, by removing cations such as Ca2þ, binding sites became more available for MnII ions, which was reflected in the increase of the sorbed amount of MnII on the MnIV surface from 20 to 185 mg MnII g
1 MnIV. This resulted in a decreased reactivity of the MnO2 because although all positively charged ions (Mn2þ, Zn2þ, Ca2þ, Mg2þ, NHþ 4 ) can inhibit the reactivity of MnO2, only MnII resulted in a decreased amount of reactive surface sites as modeled by Zhang et al. (2008a). Moreover, values above 130 mg MnII g
1 MnIV were previously shown to strongly limit the oxidation of diclofenac by BioMnOx (Forrez et al., 2010). It remains to be investigated why the reoxidation of the sorbed MnII by the manganese-oxidizing bacteria did not occur as

efficiently as in distilled water. Interference of the reoxidation of MnII can be caused by the presence of ammonium. At
1 a concentration of more than 0.7 mg NHþ 4 -N L , manganese oxidation activity is inhibited enzymatically (Vandenabeele et al., 1995). This was not observed in the batch tests, neither in the spiked BioMnOx-reactor with extra dosage of ammonium, because the effluent Mn concentrations of the BioMnOx-MBR were always below the drinking water limit (0.05 mg L
1). Another factor that influenced the reoxidation activity of the manganese-oxidizing bacteria was the low concentration of an organic C-source in STP-effluent. The organic C-source that becomes available from the oxidation of micropollutants may not be sufficient for the maintenance of these bacteria

Fig. 6 e Continuous removal of diatrizoate, iomeprol, iopromide and iohexol in Bio-Pd-MBR inoculated with Clostrydium butyricum to produce in situ H2 from formate to charge the catalyst for the reductive dehalogenation of ICM. The HRT (h) is indicated on the graph. Diatrizoate removal efficiency at spiked influent concentrations of 6 mg LL1 (A). ICM removal efficiency in non-spiked STP-effluent (B).

1771

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 6 3 e1 7 7 3

Table 3 e Theoretical reactor volume for a polishing BioMnOx-MBR based on data for diclofenac. MEC: measured effluent concentration, IE: Inhabitant Equivalent, VMBR: Volume of a BioMnOx-MBR, VCAS: Volume of a conventional activated sludge treatment.

France Germany a b c d

Average MECa (mg L
1)

Daily excretionb (mg IE
1 d
1)

VMBRc (L) per IE

Required expansion of STP (%)d (¼VMBR:VCAS*100)

0.4 1.4

48 168

18 64

30 107

Zhang et al. (2008b). based on 120 L IE
1 d
1.
1 based on 2.6 mg L
1 MBR d . based on 0.5 gBOD L
1 and loading rate of 1 gBOD L
1 d
1.

(Mirleau et al., 2005); nor is the metabolic energy from biological manganese oxidation substantial (80 kJ mol
1 MnII compared to 2870 kJ mol
1 glucose) (De Schamphelaire et al., 2007). This has two main implications for the technology. Firstly, it decreases the manganese reoxidation capacity, resulting in a higher sorbed MnII concentration and a decrease in BioMnOx-reactivity by a factor of 3 after 45 days. Secondly, the putative biological degradation of reaction products or biodegradable micropollutants decreases when the BioMnOx ‘ages’. The latter could be an explanation for the decreased removal of ibuprofen at the end of the operation in the BioMnOx-MBR (Table 1).

4.4.

Challenges of Bio-Pd

A major challenge is the supply of the hydrogen donor to charge the catalyst. Formate as a hydrogen donor gave variable results ranging from 30 to 90% diatrizoate removal. Good removal of diatrizoate (>90%) was accompanied by the production of acetate. A biological process, the acetogenesis, could explain the production of acetate from formate. After an enrichment period of 14 days, the anaerobic community in the MBR developed systematically. At an increased HRT of 5 h, this community was able to ferment the dosed formate, generating H2 gas (Lee et al., 2009). In situ H2-production by anaerobic bacteria was an excellent alternative to supply molecular hydrogen to the Bio-Pd-catalyst, because H2-gas has been shown more performant than formate to charge the Bio-Pd-catalyst for reductive dehalogenation (Hennebel et al., 2009b) and therefore good removal of even the most recalcitrant ICM, i.e. diatrizoate, was possible. On the other hand, homoacetogenesis also competes for the produced H2 (Siriwongrungson et al., 2007), which might explain why a high HRT of 24 h had the opposite effect in this application. Therefore, the in situ H2-production remains as the main challenge. From an engineering point of view, mixed cultures are considered to be favorable because operation of the process is facilitated (Ntaikou et al., 2010). However, the mixed consortia derived from anaerobic sludge require a heat or acid/base treatment to ensure dominance of the hydrogenproducing bacteria over methanogens, homoacetogens and lactic acid bacteria (Ntaikou et al., 2010).

4.5.

Practical feasibility

If removal rates are expressed as volumetric removal rates, the time component represented as HRT is included. The

values calculated for diclofenac are 2.6 mg L
1 d
1 and 0.3 mg L
1 d
1 at influent concentrations of 790 and 390 ng L
1, respectively (Table 1). In Table 3, the volume of a polishing MBR was calculated based on these results. A sewage treatment plant equipped with a polishing BioMnOx-MBR would need to double its volume. For practical feasibility, this has to decrease by a factor of 10 before it is competitive with ozonation (Hollender et al., 2009). The use of a Bio-Pd-MBR is even more expensive due to the high cost for Bio-Pd production (Hennebel et al., 2009c). Increasing the biometal concentration and altering reactor design to a submerged plate MBR with better contact of the liquid phase and the biometal suspension could improve the removal rates and decrease reactor volume.

5.

Conclusions

Removal of more than 60% of 14 out of 29 micropollutants, detected in STP-effluent, occurred in a continuous MBR module containing BioMnOx in the outer compartment. The main putative removal mechanisms were: chemical oxidation by biogenic manganese oxides (diclofenac, triclosan, chlorophene, naproxen and diuron), biological degradation (ibuprofen, codeine, dihydrocodeine, morphine, N-acetylsulfamethoxazole, iopromide, iomeprol and iohexol) and adsorption (clarithromycin). Successful removal of the iodinated X-ray contrast media iomeprol, iopromide, iohexol and diatrizoate was demonstrated in a Bio-Pd-MBR under conditions favorable for microbial H2-production enabling the charge the Pd-catalyst with molecular hydrogen. For the latter technology, a major challenge is the engineering of the in situ H2-production at environmental temperatures. Inoculation with Clostridium butyricum is a step in this direction. Effort to enhance the survival and activity of hydrogen-producing bacteria is required.

Acknowledgement This study was part of the EU Neptune project (Contract No 036845, SUSTDEV-2005-3.II.3.2), which was financially supported by the EU Commision (FP6-2005-Global-4). Marta Carballa was supported by the Xunta de Galicia (Isidro Parga Pondal program, contract IPP-08-37) and Tom Hennebel (7741-

1772

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 6 3 e1 7 7 3

02) was supported by the Fund of Scientific Research-Flanders (Fonds voor Wetenschappelijk Onderzoek (FWO) Vlaanderen).

Appendix. Supplementary material The supplementary data associated with this article can be found in the online version at doi:10.1016/j.watres.2010.11.031.

references

Aguayo, S., Munoz, M.J., de la Torre, A., Roset, J., de la Pena, E., Carballo, M., 2004. Identification of organic compounds and ecotoxicological assessment of sewage treatment plants (STP) effluents. Science of the Total Environment 328 (1e3), 69e81. Benner, J., Ternes, T.A., 2009. Ozonation of propranolol: formation of oxidation products. Environmental Science and Technology 43 (13), 5086e5093. Cao, N., Yang, M., Zhang, Y., Hu, J.Y., Ike, M., Hirotsuji, J., Matsui, H., Inoue, D., Sei, K., 2009. Evaluation of wastewater reclamation technologies based on in vitro and in vivo bioassays. Science of the Total Environment 407 (5), 1588e1597. De Schamphelaire, L., Rabaey, K., Boon, N., Verstraete, W., Boeckx, P., 2007. Minireview: the potential of enhanced manganese redox cycling for sediment oxidation. Geomicrobiology Journal 24 (7e8), 547e558. De Windt, W., Aelterman, P., Verstraete, W., 2005. Bioreductive deposition of palladium (0) nanoparticles on Shewanella oneidensis with catalytic activity towards reductive dechlorination of polychlorinated biphenyls. Environmental Microbiology 7 (3), 314e325. Dizer, H., Wittekindt, E., Fischer, B., Hansen, P.D., 2002. The cytotoxic and genotoxic potential of surface water and wastewater effluents as determined by bioluminescence, umu-assays and selected biomarkers. Chemosphere 46 (2), 225e233. Eaton, A.D., Clesceri, L.S., Rice, E.W., Franson, M.A.H., 2005. Standard Methods for the Examination of Water and Wastewater, 21st ed. American Public Health Association, Washington DC. Feitosa-Felizzola, J., Hanna, K., Chiron, S., 2009. Adsorption and transformation of selected human-used macrolide antibacterial agents with iron(III) and manganese(IV) oxides. Environmental Pollution 157 (4), 1317e1322. Forrez, I., Carballa, M., Verbeken, K., Vanhaecke, L., Schlu¨sener, M., Ternes, T., Boon, N., verstraete, W., 2010. Diclofenac oxidation by biogenic manganese oxides. Environmental Science and Technology 44, 3449e3454. Go¨bel, A., McArdell, C.S., Joss, A., Siegrist, H., Giger, W., 2007. Fate of sulfonamides, macrolides, and trimethoprim in different wastewater treatment technologies. Science of the Total Environment 372 (2e3), 361e371. Guzzella, L., Feretti, D., Monarca, S., 2002. Advanced oxidation and adsorption technologies for organic micropollutant removal from lake water used as drinking-water supply. Water Research 36 (17), 4307e4318. Hennebel, T., De Gusseme, B., Boon, N., Verstraete, W., 2009a. Biogenic metals in advanced water treatment. Trends in Biotechnology 27 (2), 90e98. Hennebel, T., Simoen, H., De Windt, W., Verloo, M., Boon, N., Verstraete, W., 2009b. Biocatalytic dechlorination of trichloroethylene with bio-palladium in a pilot-scale membrane reactor. Biotechnology and Bioengineering 102, 995e1002.

Hennebel, T., Verhagen, P., Simoen, H., De Gusseme, B., Vlaeminck, S.E., Boon, N., Verstraete, W., 2009c. Remediation of trichloroethylene by bio-precipitated and encapsulated palladium nanoparticles in a fixed bed reactor. Chemosphere 76 (9), 1221e1225. Hennebel, T., De Corte, S., Vanhaecke, L., Vanherck, K., Forrez, I., De Gusseme, B., Verhagen, P., Verbeken, K., Van der Bruggen, B., Vankelecom, I., Boon, N., Verstraete, W., 2010. Removal of diatrizoate with catalytically active membranes incorporating microbially produced palladium nanoparticles. Water Research 44 (5), 1458e1506. Hollender, J., Zimmermann, S.G., Koepke, S., Krauss, M., McArdell, C.S., Ort, C., Singer, H., von Gunten, U., Siegrist, H., 2009. Elimination of organic micropollutants in a municipal wastewater treatment plant upgraded with a full-scale postozonation followed by sand filtration. Environmental Science and Technology 43 (20), 7862e7869. Huber, M.M., Gobel, A., Joss, A., Hermann, N., Loffler, D., McArdell, C.S., Ried, A., Siegrist, H., Ternes, T.A., von Gunten, U., 2005. Oxidation of pharmaceuticals during ozonation of municipal wastewater effluents: a pilot study. Environmental Science and Technology 39 (11), 4290e4299. Hummel, D., Lo¨ffler, D., Fink, G., Ternes, T.A., 2006. Simultaneous determination of psychoactive drugs and their metabolites in aqueous matrices by liquid chromatography mass spectrometry. Environmental Science and Technology 40, 7321e7328. Joss, A., Zabczynski, S., Gobel, A., Hoffmann, B., Loffler, D., McArdell, C.S., Ternes, T.A., Thomsen, A., Siegrist, H., 2006. Biological degradation of pharmaceuticals in municipal wastewater treatment: Proposing a classification scheme. Water Research 40 (8), 1686e1696. Kagle, J., Porter, A.W., Murdoch, R.W., Rivera-Cancel, G., Hay, A.G., 2009. Advances in Applied Mircobiology Vol 67, 65e108. Klavarioti, M., Mantzavinos, D., Kassinos, D., 2009. Removal of residual pharmaceuticals from aqueous systems by advanced oxidation processes. Environment International 35 (2), 402e417. Knitt, L.E., Shapley, J.R., Strathmann, T.J., 2008. Rapid metalcatalyzed hydrodehalogenation of iodinated X-ray contrast media. Environmental Science and Technology 42 (2), 577e583. Lo¨ffler, D., Ternes, T.A., 2003. Analysis of acidic pharmaceuticals, antibiotics and ivermectine in river seiments using LCtandem M. Journal of Chromatography A 1021, 133e144. Lee, D.Y., Li, Y.Y., Noike, T., 2009. Continuous H2 production by anaerobic mixed microflora in membrane bioreactor. Bioresource Technology 100 (2), 690e695. Lister, D.L., Kanungo, G., Rathbone, D.A., Bruce, N.C., 1999. Transformations of codeine to important semisynthetic opiate derivatives by Pseudomonas putida M10. Fems Microbiology Letters 181 (1), 137e144. Mirleau, P., Wogelius, R., Smith, A., Kertesz, M.A., 2005. Importance of organosulfur utilization for survival of Pseudomonas putida in soil and rhizosphere. Applied and Environmental Microbiology 71 (11), 6571e6577. Nandi, R., Sengupta, S., 1998. Microbial production of hydrogen: an overview. Critical Reviews in Microbiology 24 (1), 61e84. Ntaikou, I., Antaonopoulou, G., Lyberatos, G., 2010. Biohydrogen production from biomass and wastes via dark fermentation: a review. Waste Biomass Valor 1, 21e39. Pe´rez, S., Barcelo, D., 2007. Fate and occurrence of X-ray contrast media in the environment. Analytical and Bioanalytical Chemistry 387 (4), 1235e1246. Pe´rez, S., Eichhorn, P., Celiz, M.D., Aga, D.S., 2006. Structural characterization of metabolites of the X-ray contrast agent iopromide in activated sludge using ion trap mass spectrometry. Analytical Chemistry 78 (6), 1866e1874.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 6 3 e1 7 7 3

Sabirova, J.S., Cloetens, L.F.F., Vanhaecke, L., Forrez, I., Verstraete, W., Boon, N., 2008. Manganese-oxidizing bacteria mediate the degradation of 17a-ethinylestradiol. Microbial Biotechnology 1 (6), 507e512. Schmidt, C.K., Brauch, H.J., 2008. N, N-dimethosulfamide as precursor for N-nitrosodimethylamine (NDMA) formation upon ozonation and its fate during drinking water treatment. Environmental Science and Technology 42 (17), 6340e6346. Schulz, M., Loffler, D., Wagner, M., Ternes, T.A., 2008. Transformation of the X-ray contrast medium lopromide in soil and biological wastewater treatment. Environmental Science and Technology 42 (19), 7207e7217. Siriwongrungson, V., Zeny, R.J., Angelidaki, I., 2007. Homoacetogenesis as the alternative pathway for H2 sink during thermophilic anaerobic degradation of butyrate under suppressed methanogenesis. Water Research 41, 4204e4210. Ternes, T.A., Hirsch, R., 2000. Occurence and behavior of X-ray contrast media in sewage facilities and aquatic environment. Environmental Science and Technology 34, 2741e2748. Ternes, T.A., Stuber, J., Herrmann, N., McDowell, D., Ried, A., Kampmann, M., Teiser, B., 2003. Ozonation: a tool for removal of pharmaceuticals, contrast media and musk fragrances from wastewater? Water Research 37 (8), 1976e1982. Ternes, T.A., Joss, A., Siegrist, H., 2004. Scrutinizing pharmaceuticals and personal care products in wastewater treatment. Environmental Science and Technology 38 (20), 392Ae399A. Vandenabeele, J., Vandewoestyne, M., Houwen, F., Germonpre, R., Vandesande, D., Verstraete, W., 1995. Role of autotrophic

1773

nitrifiers in biological manganese removal from groundwater containing manganese and ammonium. Microbial Ecology 29 (1), 83e98. Vanhaecke, L., Grootaert, C., Verstraete, W., Van de Wiele, T., 2009. Chemopreventive effects from prebiotic inulin towards microbial 2-amino-1-methyl-6-phenylimidazo[4,5-b]pyridine bioactivation. Journal of Applied Microbiology 106 (2), 474e485. Wick, A., Fink, G., Ternes, T.A., 2010a. Comparison of electrospray ionization and atmospheric pressure chemical ionization for multi-residue analysis of biocides, UV-filters and benzothiazoles in aqueous matrices and activated sludge by liquid chromatography - tandem mass spectrometry (LC-MS/MS). Journal of Chromatography A 1217 (14), 2088e2103. Wick, A., Fink, G. and Ternes, T.A. (2010b) Biological and chemical transformation of opium alkaloids in biological wastewater treatment. (In Preparation). Zhang, H.C., Huang, C.H., 2005. Oxidative transformation of fluoroquinolone antibacterial agents and structurally related amines by manganese oxide. Environmental Science and Technology 39 (12), 4474e4483. Zhang, H.C., Chen, W.R., Huang, C.H., 2008a. Kinetic modeling of oxidation of antibacterial agents by manganese oxide. Environmental Science and Technology 42 (15), 5548e5554. Zhang, Y.J., Geissen, S.U., Gal, C., 2008b. Carbamazepine and diclofenac: removal in wastewater treatment plants and occurrence in water bodies. Chemosphere 73 (8), 1151e1161.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 7 4 e1 7 8 0

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Removal of perfluorooctanoate from surface water by polyaluminium chloride coagulation Shubo Deng a,b,*, Qin Zhou a, Gang Yu a,b, Jun Huang a, Qing Fan a a b

Department of Environmental Science and Engineering, POPs Research Center, Tsinghua University, Beijing 100084, PR China State Key Joint Laboratory of Environment Simulation and Pollution Control, Tsinghua University, Beijing, 100084, PR China

article info

abstract

Article history:

Perfluorooctanoate (PFOA) has been detected in surface water all over the world, and little

Received 20 July 2010

is known of its removal by coagulation in water treatment plants. In this study, poly-

Received in revised form

aluminium chloride (PACl) was used to remove PFOA from surface water, and the effects of

18 November 2010

coagulant dose, solution pH, temperature, and initial turbidity on the removal of both PFOA

Accepted 22 November 2010

and suspended solids (SS) from water were investigated. Since the SS had high sorption

Available online 30 November 2010

affinity for PFOA, most PFOA was adsorbed on the particles and removed via the SS removal in the coagulation process. PFOA concentrations in aqueous phase decreased with

Keywords:

increasing initial turbidity and PACl dose, while they increased with increasing solution pH

PFOA

and temperature. Other perfluorinated compounds (PFCs) with different CeF chain lengths

Coagulation

and functional groups were also compared with PFOA. It was proved that hydrophobic

Polyaluminium chloride

interaction played an important role in the adsorption of PFOA on the SS. The addition of

Powdered activated carbon

powdered activated carbon (PAC) before the coagulation process significantly enhanced the removal efficiency of PFOA in water, and the residual PFOA concentrations in water were less than 1 mg/L after the addition of 1e16 mg/L PAC and subsequent coagulation when the initial PFOA concentrations were in the range of 0.5e3 mg/L. ª 2010 Elsevier Ltd. All rights reserved.

1.

Introduction

Perfluorinated compounds (PFCs) have increasingly attracted global concerns in recent years due to their wide application and global contamination in water environments. Perfluorooctanoate (PFOA) and perfluorooctane sulfonate (PFOS) are the typical PFCs and have been used in many industries as surfactants (Moody et al., 2001), fire retardants (Moody and Field, 2000), lubricants, and polymer additives (Giesy and Kannan, 2001). In particular, PFOA is used as additives in the production of fluoropolymers such as PTFE and Teflon (Davis et al., 2007). PFOS was categorized as one of the new persistent organic pollutants (POPs) in 2009 (Wang et al., 2009), while

PFOA is still produced and used in many industries, and thus the investigation of its occurrence and control in water environments is becoming more important. PFOA has been detected in wastewater (Loganathan et al., 2007), surface water (Hansen et al., 2002), ground water (Moody et al., 2003) and even tap water (Skutlarek et al., 2006) throughout the world. Industrial wastewater has been implicated as a point source for PFOA as well as its precursors entering into natural waters (Hansen et al., 2002; Prevedouros et al., 2006). Since PFOA cannot be removed by conventional biological wastewater treatment processes (Schroder et al., 2010), it is finally discharged into water environments. It has been reported that PFOA concentrations in natural waters are

* Corresponding author. Department of Environmental Science and Engineering, POPs Research Center, Tsinghua University, Beijing 100084, PR China. Tel.: þ86 10 62792165; fax: þ86 10 62794006. E-mail address: [email protected] (S. Deng). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.11.029

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 7 4 e1 7 8 0

usually in ng/L level, but elevated concentrations of PFOA have been detected in surface and ground waters near specific point sources (Loos et al., 2008; Moody et al., 2003). PFOA has been found in relatively high concentration levels in the rivers near fluorochemical factories in the USA up to 598 ng/L in the Tennessee River (Hansen et al., 2002), in Germany up to 56 mg/L in the River Alz and 33.9 mg/L in the Creek Steinbecke (Loos et al., 2008), and in Italy up to 1.3 mg/L in the Tanaro River (Loos et al., 2008). The maximum PFOA concentration of all drinking water samples taken in the Rhine-Ruhr area in Germany was determined at 519 ng/L (Skutlarek et al., 2006). Evidently, the high concentrations of PFOA in waters in some areas threat the security of drinking water for human beings. Some physicochemical technologies such as adsorption (Deng et al., 2010; Zhou et al., 2010), membrane (Fujii et al., 2007; Rayne and Forest, 2009) and sonochemistry (Cheng et al., 2010) are effective for PFOA removal or degradation, but they are used to remove high concentrations of PFOA in wastewater. Coagulation is an important technology used in water treatment plants, which can remove some SS and soluble organic pollutants from water (Li et al., 2009; Lee and Westerhoff, 2006). PFOA is a kind of surfactant, but the CeF chain is hydrophobic and oleophobic, different from some traditional surfactants (Guo et al., 2008). Once surface water is contaminated by PFOA, can the existing water treatment process remove it? To the best of our knowledge, no study on PFCs removal from water by coagulation has been reported in the literature. The primary objective of this study is to investigate the removal efficiencies of PFOA by PACl under different conditions. Batch experiments including the influences of coagulant dose, solution pH, temperature, initial PFOA concentrations and initial turbidity were conducted to study the removal performance of PFOA, and the adsorption of PFOA on the solid phase was also discussed. To further eliminate PFOA from water, the addition of PAC before the coagulation was investigated.

2.2.

1775

Coagulation experiments

2.

Materials and methods

Different amounts of PFOA were added into water, and the mixture was stirred for 12 h for PFOA reaching the equilibrium. 1 L water was used to conduct the coagulation experiments including the effects of coagulant dose, solution pH, temperature, initial turbidity, and initial PFOA concentrations on the removal of PFOA and turbidity. The coagulation experiments were conducted in aqueous solution at pH 7.8 and 25  C containing 0.2 mg/L PFOA, after 10 mg/L PACl was added. These parameters were changed in the corresponding condition experiments. In the effect of PACl dose experiment, PACl dose was in the range of 1e15 mg/L. In the effect of solution pH experiment, solution pH was adjusted from 5.0 to 9.0. The temperature at 5  C, 15  C, 25  C, and 35  C was investigated in the effect of temperature experiment. The initial PFOA concentrations at 0.2, 0.5, 1, and 1.5 mg/L were adopted in the study of the effect of PFOA concentrations. In the investigation of different PFCs removal, PFBA, PFHxA, PFOA, PFDoA, and PFOS at the same concentration of 0.46 mmol/L were simultaneously added into surface water at pH 7.8, and the dose of PACl was 10 mg/L. In the combination of coagulation and adsorption experiments, PAC was added into surface water containing 0.5, 1 or 3 mg/L PFOA, and the coagulation experiments began after 6 h adsorption. Standard jar tests were conducted in 1-L beakers to evaluate coagulation efficiencies. An initial rapid mixing was conducted at 400 rpm for 0.5 min followed by a slow mixing at 120 rpm for 10 min, and 50 rpm for 10 min. The suspension was left undisturbed for 15 min. After settling, the residual turbidity of the supernatant was measured by a portable turbidimeter (Hach-2100P, USA). The supernatant was then filtered through a glass fiber filter (GFF, pore size 0.22 mm, Whatman, UK), and the PFCs concentrations in the filtrate were measured. The precipitate on the bottom was collected. The coagulation process was illustrated in Fig. 1, showing the PFCs in the aqueous and solid phases before and after the coagulation.

2.1.

Materials

2.3.

Surface water was obtained from a reservoir in Beijing, which is serving as the source water for some local water treatment plants. The turbidity of raw water was about 1 NTU (nephelometric turbidity unit). Since the turbidity was low, the water was settled in a container, and the concentrated water with the turbidity of 6.1 NTU (SS ¼ 98.3 mg/L) was used in the coagulation experiments. The solution pH value was 7.8, and the content of total organic carbon was 59.6 mg/L obtained by a TOC analyzer (TOC-VCPH, Shimadzu, Japan). PFOA (sodium salt) and other PFCs including perfluorobutanoic acid (PFBA), perfluorohexanoic acid (PFHxA), Perfluorododecanoic acid (PFDoA), and PFOS were purchased from Tokyo Kasei Kogyo (Japan). Ammonium acetate and HPLC-grade methanol were obtained from SigmaeAldrich Corporation and Fisher Chemical (USA), respectively. Commercial PACl product (Al2O3 ¼ 29%) was provided by Liaoyang Water Purifying Agent Company. The powdered activated carbon (PAC) below 0.1 mm had the point of zero charge (pHpzc) of 7.5, and its specific surface area was measured to be 812 m2/g (Yu et al., 2009).

PFCs determination

The GFF after filtration or precipitate was placed in a 50 mL polypropylene centrifuge tube and freeze-dried, followed by the addition of 2 mL of 0.2 M NaOH solution. After 30 min, 20 mL of methanol and mass-labeled PFOA (MPFOA) were added into the tube and shaken at 200 rpm for 30 min. 0.2 mL of 2 mol/L HCl solution was added, and the mixture was centrifuged at

Fig. 1 e Schematic diagram illustrating the coagulation process.

1776

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 7 4 e1 7 8 0

Zeta potential measurement

The zeta potentials of suspended solids in surface water were analyzed at room temperature with a zeta potential instrument (Delsa Nano C, Beckman Coulter, USA). Solution pH values were adjusted in the range of 2.0e9.0 by the addition of 0.1 M HCl or NaOH solution. All data were determined five times, and the average value was adopted. The zero point of zeta potential was obtained in the plot of zeta potentials versus solution pH.

3.

Results and discussion

3.1.

PFOA distribution in solid and aqueous phases

Different amounts of PFOA were added into the surface water, and the mixture was stirred for 12 h to allow PFOA to adsorb on the SS and reach the sorption equilibrium. The equilibrium PFOA concentrations in water were measured after water samples were filtered by a 0.22 mm filter to remove the SS, and the amount of PFOA on the SS was also measured. As shown in Fig. 2, the amount of PFOA in water was much lower than that on the SS, indicating that most PFOA molecules were adsorbed on the SS in surface water. The SS in surface water contained some natural organic matters such as humic substances, and they may adsorb PFOA via hydrophobic interaction. Previous studies indicated that hydrophobic partition was mainly responsible for PFOS sorption on organic fraction in natural sediments (Higgins and Luthy, 2006), and the sorption of PFOA on the sludge was due to its partitioning from aqueous phase to the biosolids during the sorption process (Ochoa-Herrera and Sierra-Alvarez, 2008). When the

PFOA amount ( g)

PFOA on SS PFOA in water

2500 2000 1500 1000 500 0 0.0

0.5

1.0

1.5

2.0

2.5

3.0

Initial PFOA concentration (mg/L) Fig. 2 e Effect of initial PFOA concentration on its distribution in water and SS in 1 L surface water after 12 h sorption.

initial PFOA concentrations increased from 0.2 to 3 mg/L, the equilibrium PFOA concentrations in water were in the range of 21.2e393.2 mg/L. PFOA has been found in relatively high concentration levels up to 33.9 mg/L in the Steinbecke River, and 56 mg/L in the River Alz (Skutlarek et al., 2006). Since all water samples have to pass through a membrane to remove SS before solid phase extraction in the analysis of PFOA in water, the reported values in the literature are the PFOA concentrations in aqueous phase, and the PFOA amount on the SS should be much higher than that in water according to our study. In consideration of the actual PFOA concentrations in water environments, the initial PFOA concentration of 200 mg/L (about 21.2 mg/L in water) was used in most of the coagulation experiments in our study.

3.2.

Effect of PACl dose

The common coagulant PACl was used to remove SS in surface water, and the effect of PACl dose on the removal of turbidity and PFOA is shown in Fig. 3. It can be seen that the residual

200

7

SS Water Precipitate Residual turbidity

160

6 5

120

4

80

3 2

40

1

0 0

2

4

6

8

10

12

14

Residual turbidity (NTU)

2.4.

3000

PFOA amount ( g)

3000 rpm for 20 min. The supernatant was collected and concentrated to 1 mL by nitrogen blowing. Finally, 25 mg of Envi-Carb C and 50 mL of acetic acid were added into the supernatant, followed by mixing for 1 min and centrifugation at 10 000 rpm for 20 min. The obtained supernatant was collected for PFOA determination (Ahrens et al., 2009). An ultra performance liquid chromatography-tandem mass spectrometry (UPLC-MS/MS) was used to determine the concentrations of PFCs with an UPLC system (Waters Corp., USA) equipped with a C18 column (2.1  50 mm i.d., particle size 1.7 mm, Waters Corp., USA) and MS system Quattro Premier XE tandem quadrupole mass spectrometer (Waters Corp., USA) with an electrospray ionization source. The mobile phase was a binary mixture of solvent A (2 mmol/L ammonium acetate in 100% methanol) and B (2 mmol/L ammonium acetate in 5% methanol) at a flow rate of 0.3 mL/ min. The gradient started with 25% A and 75% B, and linearly ramped to 85% A and 15% B in 5 min, and ramped to 25% A and 75% B in the following 2 min. The column was allowed to equilibrate for 3 min, and the total running time was 10 min. The injection volume was 10 mL. The tandem MS analysis was conducted using the multiple reaction monitoring (MRM) mode, and the cone voltage and collision energy were 30 V and 11 V, respectively. Other PFCs were also determined using the above mentioned method, and the detection limits were found to be 50e100 ng/L for the five PFCs.

0 16

PACl dose (mg/L) Fig. 3 e Effect of PACl dose on the removal of turbidity and PFOA in the coagulation with 10 mg/L PACl in 1 L water at pH 7.8.

1777

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 7 4 e1 7 8 0

turbidity decreased with increasing PACl dose from 0 to 10 mg/L, and then became relatively stable (a little increase within error) with further increase of coagulant dose. When PACl dose was 10 mg/L, the residual turbidity was below 1 NTU. It also can be found that the residual PFOA concentrations in water decreased from 21.2 to 5.8 mg/L with increasing PACl dose. If PACl only removed the SS in water in the coagulation process, PFOA concentrations in water should be constant before and after coagulation. The decrease of PFOA concentrations in water indicated that some PFOA transferred from aqueous phase to solid phase in the coagulation process. Since PFOA has a pKa value of 2.5 (Yu et al., 2009), it exists as anions in surface water in this study. The negative PFOA may directly adsorb on the positive PACl or the flocs via electrostatic attraction, and then were removed with the formed precipitate. The PFOA amount in the residual SS decreased with increasing dose of PACl, which is consistent with the decrease of turbidity (namely, SS). The optimal dose of 10 mg/L for SS removal was used in the following experiments.

3.3.

Effect of solution pH

Solution pH not only affects the surface charge of the SS in water, but also has a significant effect on the hydrolysis species of PACl, and thus influences the removal of SS and some pollutants in the coagulation process (Hu et al., 2006; Wang et al., 2002; Pernitsky and Edzwald, 2003). The effect of pH on the removal of PFOA and turbidity is shown in Fig. 4. The residual turbidity decreased with the increase of solution pH. Since the point of zero charge (pHpzc) of SS is about 5.8 (see Fig. S1), the SS surface is negative at pH above 5.8. The enhanced removal of SS at high solution pH is attributed to the more strong electrostatic interaction between the positive PACl and negative SS. Similar results were discovered when PACl was used to coagulate natural high turbidity water, and the different Al species at different pH directly affected the coagulation performance (Lin et al., 2008). In Fig. 4, it is obvious that the residual PFOA concentrations in water increased from 0.9 to 8.3 mg/L with increasing solution pH, which is opposite to the trend of turbidity change. The removal of PFOA in water is related to the characteristics of SS and PACl species at different solution pH. Since PFOA is

negative in solution in the pH range studied (Steinle-Darling and Reinhard, 2008), it may be adsorbed on the positively charged SS via electrostatic attraction at pH less than 5.8. At the same time, Alb content is relatively high in acidic solution (Lin et al., 2008), which may adsorb more PFOA in the coagulation process. Evidently, both PACl species and positive SS are favorable for PFOA removal at low solution pH. The PFOA amount on the SS is positively correlated with the residual turbidity, which can be further removed by the subsequent filtration process in actual water treatment plants.

3.4.

Effect of solution temperature

Temperature is an important factor which affects the coagulation efficiency. Low temperature influences the hydrolysis of coagulant and increases water viscosity, resulting in low coagulation efficiency. Fig. 5 illustrates the effect of temperature on the removal of PFOA and turbidity in the coagulation process. The residual PFOA amount in aqueous phase increased with the increase of solution temperature, while the residual turbidity decreased slightly with increasing solution temperature from 5  C to 35  C. It is reasonable that residual turbidity was high at low solution temperature due to the low coagulation efficiency, and the lower temperature lowered the collision efficiency. The low residual PFOA amount in water at low temperature indicated that more PFOA molecules were adsorbed on the SS during the coagulation, which also verified that the sorption of PFOA on the SS was an exothermic process, namely, the sorption capacity of PFOA on the SS at low temperature was higher than that at high temperature. In addition, PFOA is one kind of surfactant, and more adsorbed PFOA on the SS at low temperature may make the SS more stable in aqueous solution, making part contribution to the higher residual turbidity. The increase of solution temperature improved the collision efficiency, which is helpful for coagulation.

3.5.

Effect of initial turbidity

The concentrated SS were added into the surface water to obtain the simulated water with initial turbidity ranging from

2.0

120 1.5

80 40

1.0

0

0.5 5

6

7

8

9

pH Fig. 4 e Effect of pH on the removal of turbidity and PFOA in 1 L water.

150

1.2

120

1.0

90

0.8 0.6

Water SS Precipitate Residual turbidity

60 30

0.4 0.2

0 5

10

15

20

25

30

35

0.0

Residual turbidity (NTU)

160

2.5

PFOA amount ( g)

PFOA amount ( g)

Water SS Precipitate Residual turbidity

Residual turbidity (NTU)

1.4

200

Temperature (°C) Fig. 5 e Effect of solution temperature on the removal of turbidity and PFOA in the coagulation process with 10 mg/L PACl in 1 L water at pH 7.8.

1778

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 7 4 e1 7 8 0

3.6.

Effect of initial PFOA concentrations

PFOA amount ( g)

160

1.5

120

1.2

80

Water 0.9 SS Precipitate Residual turbidity 0.6

40

0.3

0 2

4

6

8

10

Residual turbidity (NTU)

Fig. S2 shows the removal of turbidity and the distribution of PFOA in solid and aqueous phases after the coagulation of the simulated water with different PFOA concentrations. The residual PFOA concentrations in water after the coagulation increased with the increase of initial PFOA concentrations, and the same trend was observed for the amounts of PFOA on the SS and precipitate. When the initial PFOA concentration was 1.5 mg/L, the equilibrium PFOA concentration in aqueous phase before the coagulation was 197.4 mg/L (see Fig. 2), while the residual PFOA concentration in water was 110.4 mg/L after the coagulation. It is interesting that the residual turbidity decreased with increasing initial PFOA concentrations. More

0.0 12

Initial turbidity (NTU) Fig. 6 e Effect of initial turbidity on the residual turbidity and PFOA removal from 1 L surface water in the coagulation process.

PFOA anions were adsorbed on the SS at high initial PFOA concentrations, which should make the SS more stable. The possible reason is that the positive PACl species were more favorable for approaching the negative PFOA-sorbed SS at high PFOA concentrations, which caused the effective removal of SS in the coagulation process.

3.7.

Other PFCs removal by coagulation

There are other PFCs in natural surface water, which may have different removal efficiencies in the coagulation process. The typical perfluorocarboxylic acids with different CeF chain lengths (PFBA, PFHxA, PFOA, and PFDoA) and PFOS were selected and added into surface water simultaneously at the same concentration of 0.46 mmol/L. Their distributions in water, suspended solids, and precipitate after the coagulation are shown in Fig. 7. The mass balance of different PFCs was satisfactory with all recoveries above 90%, indicating the efficient sample pretreatment and accurate determination in the analysis. The loss amount of PFCs may be on the SS and precipitate. It can be seen that the residual PFOA concentration was about 21.9 mg/L (0.05 mmol) in the mixed system, much higher than 5.2 mg/L obtained in the single PFOA system (see Fig. 3), indicating the competitive sorption of other PFCs with PFOA on the SS and PACl. Although the same initial concentrations of five PFCs were used, their residual concentrations in water were different, which followed the decreasing order of PFBA > PFHxA > PFOA > PFDoA > PFOS. Their residual concentrations in water are closely related to their CeF chain length and functional groups. Besides the electrostatic interaction between the anionic PFCs and positive PACl, the hydrophobic interaction played an important role in the adsorption of PFCs on the particles. PFDoA with long CeF chain (C12) is more hydrophobic than other PFCs, and thus it is easier to be adsorbed on the organic SS and move to precipitant in the coagulation process. PFOS has shorter CeF chain than PFDoA, but its residual concentration in water is lower than PFDoA, possible due to the more hydrophobic sulfonate group (Fujii et al., 2007). The more hydrophobic PFCs tended to partition onto organic particles, which was consistent with

Water SS Precipitate

0.6 PFCs amount ( mol)

3 to 10.5 NTU. The effect of initial turbidity on the residual turbidity and PFOA removal from aqueous solution is presented in Fig. 6. At low initial turbidity, the residual PFOA concentrations in solution were higher. In the process of coagulation, the particles in low turbidity water were not prone to the formation of aggregates, and the coagulation efficiency was low. By contrast, the high turbidity surface water contained more SS, which may adsorb more PFOA from aqueous solution, resulting in low residual concentrations of PFOA in water. When the initial turbidity was 10.5 NTU, the PFOA concentration in water decreased to 138 ng/L after the coagulation at the PACl dose of 10 mg/L. Due to the presence of more SS in water, the collision efficiency between particles during the coagulation process increased. With the help of coagulant, the SS are easy to form aggregates and flocs, and finally separated from water. However, a large number of natural organic matters in surface water may interfere with the coagulation efficiencies of PACl (Huang and Shiu, 1996). The high turbidity water contained more natural organic matters in this study, and thus the residual turbidity increased with the increase of initial turbidity in water. It should be pointed out that the amount of PFOA on the SS at low initial turbidity was higher although the residual turbidity was lower than that at high initial turbidity. The reason was that more PFOA anions were adsorbed on per gram SS at low initial turbidity.

0.5 0.4 0.3 0.2 0.1 0.0

PFBA

PFHxA

PFOA

PFDoA

PFOS

Fig. 7 e Distribution of different PFCs in aqueous and solid phases in 1 L surface water after the coagulation at PACl dose of 10 mg/L.

PFOA in water ( g/L)

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 7 4 e1 7 8 0

400 300 200 100

4. 0.5 mg/L PFOA 1 mg/L PFOA 3mg/L PFOA

8 6 4 2 0 0

5

10

15

20

PAC dose (mg/L) Fig. 8 e Effect of PAC dose on the removal of PFOA from surface water during the coagulation by the addition of 10 mg/L of PACl.

their high amounts on the SS and precipitate. Especially their amounts on the precipitate followed the order of their hydrophobicity since the precipitate contained a significant fraction of the particles.

3.8.

Combination of coagulation and adsorption

According to above results, it can be found that coagulation can remove most PFOA from water, but the residual PFOA concentrations in water are still high, even up to 110 mg/L (see Fig. S2). In some accidents, surface water may be contaminated by the high concentration of PFCs, and the conventional coagulation is not sufficient to remove them in water treatment plants. In this case, powdered activated carbon (PAC) is often added into the source water before the coagulation process. We examined the removal efficiency of PFOA in surface water by the combination of adsorption and coagulation. As shown in Fig. S3, the turbidity removal was enhanced after adding the PAC even at low dose of 1 mg/L, and kept constant till the PAC dose of 20 mg/L. The initial PFOA concentrations had little effect on the turbidity removal in the adsorption and coagulation process. Fig. 8 illustrates the effect of PAC dose on the residual PFOA concentrations in water after the adsorption and coagulation. The addition of PAC significantly decreased the residual concentrations of PFOA in water after the coagulation. When the initial PFOA concentrations were 0.5, 1, and 3 mg/L, the residual PFOA concentrations in water were below 1 mg/L after the addition of PAC at the dose of 1, 10, and 16 mg/L, respectively. Our previous study indicated that PAC could effectively remove PFOA from water by electrostatic interaction and hydrophobic interaction, and the sorption equilibrium was reached within 4 h (Yu et al., 2009). Evidently, it is a feasible method to remove the high concentration of PFOA from surface water in the case of accidents in water treatment plants by the addition of PAC, followed by the coagulation.

1779

Conclusions

Coagulation is an efficient method to remove PFOA from surface water, and its removal percents exceeded 90% at the optimal PACl dose of 10 mg/L. Most PFOA species were removed from surface water in the coagulation via the produced precipitate since the SS in water had strong sorption affinity for PFOA. The positive PACl coagulant would adsorb the negative PFOA in water in the coagulation process. More PFOA species were adsorbed on the solid phase at low solution pH and low temperature, resulting in low residual PFOA concentrations in water. Besides the electrostatic interaction, hydrophobic interaction was also involved in the sorption of PFOA on the solid phase, and the PFCs with long CeF chain were more easily adsorbed on the SS and removed by the coagulation. The residual PFOA in aqueous phase can be further removed by the addition of PAC before the coagulation, and its final concentrations were less than 1 mg/L after the addition of 1e16 mg/L PAC and subsequent coagulation when the initial PFOA concentrations were in the range of 0.5e3 mg/L. The combination of adsorption and coagulation is a feasible method to remove some PFCs at high concentrations from water in some accidents.

Acknowledgements We thank the National Nature Science Foundation of China (no. 50778095), special fund of State Key Joint Laboratory of Environment Simulation and Pollution (no. 10Y01ESPCT), and National Outstanding Youth Foundation of China (no. 50625823) for financial support, and the Program for New Century Excellent Talents in University is also appreciated. Additionally, the analytical work was supported by the Laboratory Fund of Tsinghua University.

Appendix. Supplementary data Supplementary data related to this article can be found online at doi:10.1016/j.watres.2010.11.029

references

Ahrens, L., Plassmann, M., Xie, Z.Y., Ebinghaus, R., 2009. Determination of polyfluoroalkyl compounds in water and suspended particulate matter in the river Elbe and North Sea, Germany. Front. Environ. Sci. Engin. China 3, 152e170. Cheng, J., Vecitis, C.D., Park, H., Mader, B.T., Hoffmann, M.R., 2010. Sonochemical degradation of perfluorooctane sulfonate (PFOS) and perfluorooctanoate (PFOA) in groundwater: kinetic effects of matrix inorganics. Environ. Sci. Technol. 44, 445e450. Davis, K.L., Aucoin, M.D., Larsen, B.S., Kaiser, M.A., Hartten, A.S., 2007. Transport of ammonium perfluorooctanoate in environmental media near a fluoropolymer manufacturing facility. Chemosphere 67, 2011e2019. Deng, S.B., Yu, Q., Huang, J., Yu, G., 2010. Removal of perfluorooctane sulfonate from wastewater by anion

1780

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 7 4 e1 7 8 0

exchange resins: effects of resin properties and solution chemistry. Water Res. 44, 5188–5195. Fujii, S., Polprasert, C., Tanaka, S., Lien, N.P.H., Qiu, Y., 2007. New POPs in the water environment:distribution, bioaccumulation and treatment of perfluorinated compounds-a review paper. J. Water Supply Res. T. 56, 313e326. Giesy, J.P., Kannan, K., 2001. Global distribution of perfluorooctane sulfonate in wildlife. Environ. Sci. Technol. 35, 1339e1342. Guo, J., Resnick, P., Efimenko, K., Genzer, J., DeSimone, J.M.A., 2008. Alternative fluoropolymers to avoid the challenges associated with perfluorooctanoic acid. Ind. Eng. Chem. Res. 47, 502e508. Hansen, K.J., Johnson, H.O., Eldridge, J.S., Butenhoff, J.L., Dick, L.A., 2002. Quantitative characterization of trace levels of PFOS and PFOA in the Tennessee River. Environ. Sci. Technol. 36, 1681e1685. Higgins, C.P., Luthy, R.G., 2006. Sorption of perfluorinated surfactants on sediments. Environ. Sci. Technol. 40, 7251e7256. Huang, C.P., Shiu, H.L., 1996. Interactions between alum and organics in coagulation. Colloid Surface A 113, 155e163. Hu, C.Z., Liu, H.J., Qu, J.H., Wang, D.S., Ru, J., 2006. Coagulation behavior of aluminum salts in eutrophic water: significance of Al-13 species and pH control. Environ. Sci. Technol. 40, 325e331. Lee, W., Westerhoff, P., 2006. Dissolved organic nitrogen removal during water treatment by aluminum sulfate and cationic polymer coagulation. Water Res. 40, 3767e3774. Lin, J.L., Huang, C.P., Pan, J.R., Wang, D.S., 2008. Effect of Al(III) speciation on coagulation of highly turbid water. Chemosphere 72, 189e196. Li, X.M., Peng, P.A., Zhang, S.K., Man, R., Sheng, G.Y., Fu, J.M., 2009. Removal of polychlorinated dibenzo-p-dioxins and polychlorinated dibenzofurans by three coagulants in simulated coagulation processes for drinking water treatment. J. Hazard. Mater. 162, 180e185. Loganathan, B.G., Sajwan, K.S., Sinclair, E., Kumar, K.S., Kannan, K., 2007. Perfluoroalkyl sulfonates and perfluorocarboxylates in two wastewater treatment facilities in Kentucky and Georgia. Water Res. 41, 4611e4620. Loos, R., Locoro, G., Huber, T., Wollgast, J., Christoph, E.H., De Jager, A., Gawlik, B.M., Hanke, G., Umlauf, G., Zaldivar, J.M., 2008. Analysis of perfluorooctanoate (PFOA) and other perfluorinated compounds (PFCs) in the River Po watershed in N-Italy. Chemosphere 71, 306e313. Moody, C.A., Field, J.A., 2000. Perfluorinated surfactants and the environmental implications of their use in fire-fighting foams. Environ. Sci. Technol. 34, 3864e3870. Moody, C.A., Hebert, G.N., Strauss, S.H., Field, J.A., 2003. Occurrence and persistence of perfluorooctanesulfonate and

other perfluorinated surfactants in groundwater at a firetraining area at Wurtsmith Air Force Base, Michigan, USA. J. Environ. Monit. 5, 341e345. Moody, C.A., Kwan, W.C., Martin, J.W., Muir, D.C.G., Mabury, S.A., 2001. Determination of perfluorinated surfactants in surface water samples by two independent analytical techniques: liquid chromatography/tandem mass spectrometry and 19F NMR. Anal. Chem. 73, 2200e2206. Ochoa-Herrera, V., Sierra-Alvarez, R., 2008. Removal of perfluorinated surfactants by sorption onto granular activated carbon, zeolite and sludge. Chemosphere 72, 1588e1593. Pernitsky, D.J., Edzwald, J.K., 2003. Solubility of polyaluminium coagulants. J. Water Supply Res. T. 52, 395e406. Prevedouros, K., Cousins, I., Buck, R.C., Korzeniowski, S.H., 2006. Sources, fate and transport of perfluorocarboxyates. Environ. Sci. Technol. 40, 32e44. Rayne, S., Forest, K., 2009. Perfluoroalkyl sulfonic and carboxylic acids: a critical review of physicochemical properties, levels and patterns in waters and wastewaters, and treatment methods. J. Environ. Sci. Heal. A 44, 1145e1199. Schroder, H.F., Jose, H.J., Gebhardt, W., Moreira, R.F.P.M., Pinnekamp, J., 2010. Biological wastewater treatment followed by physicochemical treatment for the removal of fluorinated surfactants. Water Sci. Technol. 61, 3208e3215. Skutlarek, D., Exner, M., Farber, H., 2006. Perfluorinated surfactants in surface and drinking water. Environ. Sci. Poll. Res. 13, 299e307. Steinle-Darling, E., Reinhard, M., 2008. Nanofiltration for trace organic contaminant removal: structure, solution, and membrane fouling effects on the rejection of perfluorochemicals. Environ. Sci. Technol. 42, 5292e5297. Wang, D.S., Tang, H.X., Gregory, J., 2002. Relative importance of charge neutralization and precipitation on coagulation of kaolin with PACI: effect of sulfate ion. Environ. Sci. Technol. 36, 1815e1820. Wang, T., Wang, Y.W., Liao, C.Y., Cai, Y.Q., Jiang, G.B., 2009. Perspectives on the inclusion of perfluorooctane sulfonate into the Stockholm convention on persistent organic pollutants. Environ. Sci. Technol. 43, 5171e5175. Yu, Q., Zhang, R.Q., Deng, S.B., Huang, J., Yu, G., 2009. Sorption of perfluorooctane sulfonate and perfluorooctanoate on activated carbons and resin: kinetic and isotherm study. Water Res. 43, 1150e1158. Zhou, Q., Deng, S.B., Yu, Q., Zhang, Q.Y., Yu, G., Huang, J., He, H.P., 2010. Sorption of perfluorooctane sulfonate on organomontmorillonites. Chemosphere 78, 688e694.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 8 1 e1 7 9 5

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Temporal trend and source apportionment of water pollution in different functional zones of Qiantang River, China Shiliang Su a, Dan Li a, Qi Zhang a, Rui Xiao b, Fang Huang a, Jiaping Wu a,* a b

College of Environment and Natural Resources, Zhejiang University, Hangzhou, China Ministry of Agriculture Key Laboratory for Non-point Pollution Control, Zhejiang University, Hangzhou, China

article info

abstract

Article history:

The increasingly serious river water pollution in developing countries poses great threat to

Received 9 April 2010

environmental health and human welfare. The assignment of river function to specific uses,

Received in revised form

known as zoning, is a useful tool to reveal variations of water environmental adaptability to

14 October 2010

human impact. Therefore, characterizing the temporal trend and identifying responsible

Accepted 23 November 2010

pollution sources in different functional zones could greatly improve our knowledge about

Available online 30 November 2010

human impacts on the river water environment. The aim of this study is to obtain a deeper understanding of temporal trends and sources of water pollution in different functional

Keywords:

zones with a case study of the Qiantang River, China. Measurement data were obtained and

River water environment

pretreated for 13 variables from 41 monitoring sites in four categories of functional zones

Temporal trend

during the period 1996e2004. An exploratory approach, which combines smoothing and

Source

non-parametric statistical tests, was applied to characterize trends of four significant

Functional zones

parameters (permanganate index, ammonia nitrogen, total cadmium and fluoride)

Water management

accounting for differences among different functional zones identified by discriminant analysis. Aided by GIS, yearly pollution index (PI) for each monitoring site was further mapped to compare the within-group variations in temporal dynamics for different functional zones. Rotated principal component analysis and receptor model (absolute principle component score-multiple linear regression, APCS-MLR) revealed that potential pollution sources and their corresponding contributions varied among the four functional zones. Variations of APCS values for each site of one functional zone as well as their annual average values highlighted the uncertainties associated with cross space-time effects in source apportionment. All these results reinforce the notion that the concept of zoning should be taken seriously in water pollution control. Being applicable to other rivers, the framework of management-oriented source apportionment is thus believed to have potentials to offer new insights into water management and advance the source apportionment framework as an operational basis for national and local governments. ª 2010 Elsevier Ltd. All rights reserved.

1.

Introduction

Urbanization, increasing land development and industrial activities in the absence of adequate wastewater treatment, has contributed to water quality deterioration (Jonnalagadda and Mhere, 2001; Gupta, 2008). The increasingly evident

water pollution problems have also led to serious ecological and environmental consequences (Ma et al., 2009). The concurrent environmental issues are fairly complex, involve scientific and engineering aspects, and have important social, legal, economic, and political ramifications (Weng, 2007). Most developing countries are now facing a daunting task of water

* Corresponding author. Tel.: þ86 13968029606; fax: þ86 57186971359. E-mail address: [email protected] (J. Wu). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.11.030

1782

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 8 1 e1 7 9 5

resources management, dealing with problems of effective water pollution control and river ecosystem health maintenance. One promising way to deal with such complex problems is to consider them as an integrated river basin management issue, and bring together all participant planners, engineers, landscape architects, social and natural scientists, local officials, and others at basin scale, for the common good (Weng, 2007). A host of factors have imposed challenge for integrated river basin management practices, especially the multi-functional use of water resources under varying hydrological, environmental, and socio-economic circumstances. The assignment of river function to specific uses, known as zoning, is a useful option to mitigate conflicts and a key prescriptive tool for visualizing human impacts on environment. Although the majority of zoning schemes were not originally designed for water pollution monitoring and management, these practices are common (Brierley et al., 2002; Sabatini et al., 2007; Cheruvelil et al., 2008). Using zoning frameworks for such purposes assumes that characteristics of river functions within zone are more similar than those across zones. And such prescriptive zoning is not a descriptive final document but an ongoing process, assuring that the management of the area is periodically updated and better planned. Despite the popular use of zoning schemes for integrated river basin management, there have been few reports to study the pollution characteristics and variations among different zones. Since the “open-door” policy in 1978, many rivers in China have been seriously polluted due to rapid urbanization and industrialization, which exceeded the capacity of river ecosystems. For example, more than 70% of the water in the Huang, Huai and Hai River basins is polluted (Wang et al., 2008). River water pollution thus has been set high on the agenda of the Chinese government in response to a variety of environmental concerns and sustainable development needs. Environmental monitoring systems have been established since 1980s and various water quality monitoring programs carried out for decades in China. Additionally, in the new round of revision in river basin planning, enough emphasis was given to the concept of river function zoning. The fundamental goal of the river function zoning in China is to distinguish different parts of the river and make a clear division of control targets on the basis of river’s natural features, social services, and ecological functions (China Ministry of Water Resources, 2007). In such circumstances, information on temporal trends of water quality and the contributing pollution sources should be helpful for managers. Multivariate statistical methods are capable of detecting similarities among variables, and therefore allow profound interpretation of data and assessment of input contributions from various sources (Singh et al., 2005; Zhou et al., 2007; Su et al., in press). In recent years, there have been many studies on spatio-temporal patterns and source apportionment using multivariate statistical methods in environmental pollution issues (Love et al., 2004; Kuppusamy and Giridhar, 2006; Zhou et al., 2007; Huang et al., 2010). Particularly, many studies focused on physical characteristics of the measurement data and paid less or no attention on the scienceepolicy integration. Despite their high-quality, they have not effectively advanced the source apportionment framework as

an operational basis for government agencies, since the results were often difficult-to-understand for managers. Given the popular use of zoning schemes for integrated river basin management worldwide, we argue that comparison of temporal trends pollution sources in different river functional zones should significantly contribute to developing optimal control strategies and determining priorities in decisionmaking procedure. However, no source apportionment study, from management prospective, has been carried out under the framework of river function zoning. There is incomplete understanding of variations in contributing sources of different functional zones to reinforce the notion that the concept of zoning should be taken seriously in water pollution control. Given the above considerations, the aim of this study is to obtain a deeper understanding of temporal trends and sources of water pollution in different functional zones with a case study of the Qiantang River, eastern coastal China. This will be done by using multiple multivariate methods, time series analysis, pollution index (PI) and geographic information systems (GIS). Specifically, our objectives are to: (1) characterize the temporal trends of water quality in different functional zones of Qiantang River between 1996 and 2004; (2) identify significant parameters and compare the pollution sources as well as their quantitative contributions in different functional zones; and (3) provide a frame of reference for policy makers to promote the river function zoning practices as well as effective water pollution control strategies in China.

2.

Background

2.1.

Study area

The Qiantang River, located in eastern coastal China (Fig. 1), is the largest one in Zhejiang Province. It plays a critical role in the sustainable development of Yangtze River Delta, due to its multiple functions: water supply, electricity generation, irrigation, tourism, fishery and shipping. Qiantang River basin covers around 40,000 km2 and has a population of about 20 million. Following the opening-up policies, this area has made rapid strides in its social-economic development. In the period between 1996 and 2003, it has doubled its GDP, with GDP increasing from RMB 1586 billion in 1994e3366 billion Chinese Yuan in 2003. While the Qiantang River basin appears as one of the most rapidly advanced economic regions in China and has now become known as the world’s workshop, it is widely acknowledged that the water quality of Qiantang River continues to deteriorate. Nevertheless, few detailed studies have been carried out to reveal temporal trends of different functional zones in the river. Systematic mapping of pollution sources is recommended with a view to assist environmental managers and public health officials to control river water pollution.

2.2.

River functional zoning

In general, the river provides various functions: drinking, shipping, power generation, sightseeing, fisheries, etc. It needs to integrate various functions, as well as define coordination among them, to stablize river ecosystem health and promote

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 8 1 e1 7 9 5

1783

Fig. 1 e Location of Qiantang River and spatial distribution of the forty-one monitoring sites in four functional zones.

sustainable development. River function zoning is to reasonably regionalize the whole river basin according to certain indicators and standards based on natural properties (resource status, environmental conditions and geographic location) and social attributes (utilization level of water resources, demand of social advancement on water quality and quantity) of water resources. The use of river function division benefits more scientific control, supervision and management of water pollution at basin scale (China Ministry of Water Resources, 2007). Acted in 2002, the river functional zoning has been put into practice in Zhejiang Province for eight years. Besides taking into consideration of the main principles, the river functional zoning designation of Zhejiang Province was made in attempt to link regional development, rural and urban planning, land use policy, and the development planning for various economic zones. To be specific, the corresponding river functional zones for Qiantang River were divided into six categories: natural reserved zone, fishing-oriented zone, industry-oriented zone, drinking water source protection

zone, landscape and recreational zone, and multi-function zone [Zhejiang Environmental Protection Bureau (ZEPB), 2006]. Natural reserved zone refers to land and water bodies needing special protection and management, where representative natural ecosystems are dominant and rare or endangered species naturally concentrate or live. Fishingoriented zone represents water bodies used for spawning, providing migration channels, cable feeding and breeding fish, shrimp, crab, shellfish, algae and other aquatic animals and plants. Industry-oriented zone is the area where intenselyused water intake points for industrial and mining production are located. Drinking water source protection zone indicates areas designated for providing the sources of drinking water. Landscape and recreational zone is the water with basic attributes for aquatic ecology protection and is thus suitable for sightseeing and entertainment. And multi-function zone refers to water bodies whose functions possess a lower requirement on water quality and can not be attributed to one specific function category.

1784

0.79 0.66 0.60 0.74 0.001 0.001 0.0000 0.001 0.0001 0.001 0.051 0.16 0.12 8.77 4.59 3.94 4.71 0.002 0.006 0.0001 0.006 0.001 0.004 0.243 1.13 0.619 5.56 1.79 1.07 0.22 0.001 0.001 0.0000 0.001 0.0001 0.002 0.005 0.12 0.048 7.67 3.06 2.12 0.75 0.0015 0.003 0.0000 0.002 0.0002 0.003 0.06 0.27 0.136 1.52 1.66 1.42 1.74 0.037 0.005 0.0000 0.014 0.001 0.003 0.120 0.44 0.09 12.58 9.49 7.49 9.21 0.323 0.029 0.0002 0.1 0.005 0.017 0.604 2.02 0.30 2.44 1.15 0.37 0.01 0.001 0.001 0.0000 0.001 0.0001 0.002 0.01 0.038 0.01 7.23 3.66 2.27 1.47 0.007 0.004 0.0000 0.011 0.0007 0.004 0.111 0.69 0.12 1.30 1.73 1.30 0.72 0.003 0.005 0.0008 0.017 0.0019 0.0059 0.564 0.315 0.053 10.71 12.60 8.41 2.98 0.019 0.02 0.0042 0.085 0.0084 0.035 3.027 1.373 0.228 3.83 1.64 0.71 0.17 0.001 0.002 0.0000 0.001 0.0001 0.002 0.008 0.14 0.017

Max Min

Fishing-oriented zone

Mean Std Max Min Mean Std Max Min Mean

7.60 3.68 2.22 0.96 0.002 0.005 0.0002 0.008 0.001 0.006 0.237 0.50 0.088 0.91 1.46 0.99 1.05 0.009 0.0018 0.0005 0.031 0.0012 0.0034 0.216 0.394 0.108 10.15 11.58 8.03 7.17 0.108 0.016 0.0042 0.25 0.0084 0.035 1.9 2.19 0.69 4.80 0.84 0.32 0.01 0.001 0.001 0.0000 0.001 0.0001 0.002 0.005 0.038 0.01

Multivariate statistics

Discriminant analysis (DA) constructs discriminant factors (DFs) to assess the variations in water quality among different river functional zones based on three different modes: standard, forward stepwise and backward stepwise. The standard DA mode constructs DFs containing all parameters. In the forward stepwise mode, variables are added one by one, beginning with the most significant until no significant changes are obtained. In the backward stepwise mode, it is opposite: variables are removed one by one, beginning with the least significant until no significant changes occur. Principal component analysis (PCA) extracts eigenvalues and related loadings from the covariance matrix of original variables to produce new orthogonal variables through varimax rotation, which are linear combinations of the original variables

7.62 2.89 1.66 0.77 0.003 0.003 0.0001 0.011 0.0006 0.0041 0.127 0.440 0.102

3.3.1.

DO CODMn BOD NH3-N V-ArOH TCN THg TPb TCd Cr6þ Petroleum F TP

Methods

Std

3.3.

Max

All the selected parameters were measured in the laboratory of Environmental Monitoring Center of Zhejiang Province according to the environmental quality standards for surface water of China (Wei et al., 1989) and the standard methods for observation and analysis in China (Huang et al., 1999). The specific method used is presented as follows: DO, electrochemical probe method; CODMn, acidic (alkaline) potassium permanganate method; V-ArOH, after distillation by means of 4-AAP spectrophotometric method; BOD, dilution and seeding method; NH3-N, spectrophotometric method with salicylic acid; TCN, pyridineebarbituric acid colorimetry (isonicotinic acidepyrazolone colorimetric method); THg, cold-vapor atomic absorption spectrophotometry; TPb and TCd, atomic absorption spectrophotometry (chelating extraction); petroleum, infrared spectrophotometry; Cr6þ, diphenylcarbohydrazide spectrophotometric method; F, ion chromatography; TP, ammonium molybdate spectrophotometric method.

Min

Laboratory analysis

Mean

3.2.

Drinking water source protection zone

Monthly data of thirteen parameters from 41 monitoring stations in Qiantang River and its tributaries were obtained from Environmental Monitoring Center of Zhejiang Province. The 13 water quality are: dissolved oxygen (DO), permanganate index (CODMn), five day biochemical oxygen demand (BOD), ammonia nitrogen (NH3-N), volatile phenol (V-ArOH), total cyanide (TCN), total mercury (THg), total lead (TPb), total cadmium (TCd), hexavalent chromium (Cr6þ), petroleum, fluoride (F) and total phosphorus (TP) (Table 1). Six TP values and two TCN values were missing, and were replaced by virtue of smoothing. Because the distributions of V-ArOH, petroleum, Cr6þ, TCd, THg and TCN were skewed, the original data were BoxeCox transformed. According to the geographical distributions of the monitoring stations (Fig. 1), the 41 sites fall into four categories of functional zones: multi-function zone (Zone A), landscape and recreational zone (Zone B), drinking water source protection zone (Zone C) and fishing-oriented zone (Zone D).

Landscape and recreational zone

Data

Multi-function zone

3.1.

Table 1 e Descriptive statistics of water quality variables in different functional zones of Qiantang River between 1996 and 2004 (unit: mg/L).

Materials and methods

Parameters

3.

Std

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 8 1 e1 7 9 5

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 8 1 e1 7 9 5

(Kaiser, 1958; Davis, 1986; Zhang et al., 2008). The new orthogonal variables allow data reduction with minimum loss of original information, and provide information on the most meaningful parameters for the description of the whole data set (Kaiser, 1958). In our study, PCA of the normalized variables (water quality data set) was performed to extract significant principal components and to further reduce the contribution of variables with minor significance; these PCs were subjected to varimax rotation (raw) generating original variables. Following Pekey et al. (2004), eigenvalues 1 were selected as the new orthogonal variables. Absolute principal component score-multiple linear regression (APCS-MLR) can be applied to estimate the contribution of each pollution source to the total, by combining MLR with the de-normalized APCS values produced by PCA and the measured concentrations of the particular pollutant (Singh et al., 2005; Zhou et al., 2007; Su et al., in press). After determination of the number and identity of possible sources influencing the river water quality in four functional zones by using PCA, source contributions were computed through APCS-MLR technique in this paper. Quantitative contributions from each source for individual parameter/contaminant were compared with their measured values. All statistical calculations were performed using the “Statistical Package for the Social Sciences Software-SPSS 16.0 for Windows” (SPSS Inc., Chicago, IL).

Whitney, 1947), were applied to confirm whether any trends in water quality during the study period existed. For the time series that do not produce visual important variations, a KruskaleWallis test can be applied to study median homogeneity. This statistical method tests the null hypothesis that the medians within each series are the same. The data from all series were combined and ranked from lowest to highest. The average rank is then computed for the data in each series. Since the P value is less than 0.05, there is a statistically significant difference amongst the medians at a 95% confidence level. The tested series was chosen for each parameter considering the results of ES analysis. ManneWhitney test is constructed by combining the two samples, sorting the data from lowest to highest value and comparing the average rank of the two series in the combined data. ManneWhitney test has been popularly used to detect a shift or a step change in median or mean of hydro-meteorological time series such as water quality, stream flow, temperature, and precipitation (Yue and Wang, 2002). Of particular interest is the P value of the two-sided test. Since the P value is less than 0.05, there is a statistically significant difference amongst the medians at a 95% confidence level. The tested series was chosen for each parameter considering the results of KruskaleWallis test, in order to identify sudden changes in the study period.

3.3.3. 3.3.2.

Time series analysis

While widely used for trend analysis, ManneKendall’s test has found its home in water science studies (Cun and Vilagines, 1997; Chang, 2008; Yang et al., 2009; Zhang et al., 2010). However, this method for trend analysis works only when the number of observed measurements exceeds ten. Besides, it can not effectively interpret the inter-level dynamics of water quality over a certain interval. To overcome the classical problems of Mann-Kendall’s test, this paper used the following approach. The first step, referred to as a graphic analysis of trend and dispersion, is to identify the major changes in the time period with box-and-whisker graphs. These plots are a graphical summary of data distribution and the outliers’ presence in the data (individual points beyond the whiskers) (Tukey, 1977). The whiskers (end points of the lines attached to the box) extend out the lower and upper value of the data series. In addition, box-and-whisker diagrams can be used to identify the lack of symmetry in the distribution of data for a defined period of time (Cun and Vilagines, 1997). The second step is to determine the presence of shifts in the mean of the water quality variables for the investigated years. Exponential smoothing (ES) is a technique that can be applied to time series data, either to produce smoothed data for presentation, or to make forecasts. It is a powerful approach to examining short-to-medium term time series up to 5 years, since no parameters have to be estimated (Temme et al., 2004). The fundamental idea is to recalculate each value of the time series by smoothing it as the weighted average of the previous observations, where the weights decrease exponentially depending on the value of the smoothing parameter (For details, see Wei, 1989). Finally, non-parametric tests, KruskaleWallis test (Kruskal and Wallis, 1952) and ManneWhitney test (Mann and

1785

Pollution index

Pollution index (Wang et al., 2008) is applied to compare temporal variations of different functional zones in Qiantang River. Eq. (1) is used to calculate the surface water quality. PI ¼

Ci ði ¼ 1; 2; .nÞ C0

(1)

where PI is the pollution index; Ci is the actual concentration of surface water, mg/l; Co is the standard concentration value of surface water, mg/l. When PI is greater than 1, the monitoring site is regarded as polluted; otherwise, non-polluted. The “Environmental Quality Standards for Surface Water” of China (GB3838-2002), divides the surface water quality into five categories (see Appendix). For Qiantang River, each category of functional zones has its corresponding standard value (ZEPB, 2006). We then use the standard value for each monitoring site in the four functional zones to calculate PI value.

4.

Results and discussion

4.1.

Temporal trend analysis

4.1.1.

Significant variables

The most significant water quality variables associated with the differences among the four functional zones were generated by DA. As shown in Table 2, the values of Wilk’lambda for each discriminant function varied from 0.598 to 0.959 and the chi-square from 9.1 to 108.2 with the p-level < 0.01, indicating that the temporal DA results were credible. Obtained from the standard, forward stepwise and backward stepwise modes of DA, the discriminant functions (DFs) and classification matrices (CMs) were shown in Table 3. DFs from the standard stepwise mode, using 13 discriminant variables, yielded CMs correctly assigning 88.4% of the cases; DFs of

1786

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 8 1 e1 7 9 5

Table 2 e Results of discriminant analysis for temporal variation in Qiantang River between 2001 and 2004. Models

DF

Wilks’lambda

Chi-square

p-level

Standard

1 2 3 1 2 3 1 2 3

0.896 0.777 0.598 0.912 0.816 0.703 0.959 0.892 0.743

108.2 53.0 23.1 86.4 39.9 16.5 64.1 24.7 9.1

0.000 0.000 0.00 0.000 0.000 0.000 0.000 0.000 0.000

Forward

Backward

the forward stepwise mode using 7 discriminant variables, yielded CMs correctly assigning 84.9% of the cases; and DFs of the backward stepwise mode produced similar result of 80.7% correct assignment with only 4 discriminant parameters: CODMn, NH3-N, TCd and F. These DA results suggested that CODMn, NH3-N, TCd and F were the most significant parameters for discrimination among the four functional zones and accounted for most of the expected variations of water quality.

4.1.2.

Temporal trends of significant variables

Box-and-whisker plots of the significant discriminant parameters recognized by DA were shown in Fig. 2. Symmetric data would have the median to lie in the middle of the rectangle and the lengths of the upper and lower whiskers would be about the same (Cun and Vilagines, 1997). Notice in Fig. 2 that, concentrations of the four variables presented asymmetrical data almost in all the period of time between 1996 and 2004. In addition, the patterns of changes for the four variables seemed irregular and significant sudden changes were not detected. Time series yearly moving averages were applied to visualize the evolution law of water quality data from apparent irregularity. Fig. 3a showed trends of CODMn for four functional zones that behaved differently. Generally speaking, CODMn values for Zone A and D remained stable during the study period, while those for Zone B and C marked

an oscillation pattern. The trend curves of NH3-N for Zone A, C and D had similar profiles, smoothly across the period 1996e2004. NH3-N concentrations in Zone B fluctuated within the interval 1998e2003, and became stable in other years (Fig. 3b). The concentrations of TCd reflected almost no significant changes in all zones except Zone C, with an upward from 1997 to 1999 and then a downward trend in the period 1999e2002 (Fig. 3c). No significant changed trends of F existed among the four functional zones (Fig. 3d). Non-parametric tests were used to confirm the above analysis (Table 4). Results denoted that stationarities exited among the time series except the CODMn time series for Zone B and C, NH3-N series for Zone B and TCd series for Zone D. Nevertheless, ManneWhitney tests marked that 1999, 2001 and 2003 were not significant drop/increase points of CODMn time series for Zone C. Similarly, 1999 was not the significant drop point of NH3-N series for Zone B. On the contrary, significant gap between 2000 and 2004 was identified for CODMn time series of Zone B, suggesting that organic pollution became more serious in this zone. In addition, 1999 and 2002 was respectively regarded as increase and drop point of TCd series for Zone C. This may be caused by increased amount of wastewater discharged from industry in 1999. Concerning about the increasing heavy metal pollution in the drinking water protection zone, the local government invested to tackle with this alarming issue. This may account for the decline in TCd concentrations in 2002.

4.1.3.

Within-group variations for different functional zones

Aiming at comparing the within-group variations for different functional zones, yearly PI for each monitoring site was calculated and mapped with GIS. Fig. 4a indicated that CODMn of most sites in Zone A met the required standards except for site 2, 14, 26 and 39. Site 2, 14 and 26 reflected a deteriorating tendency, while site 39 was in continued polluted status. The rising trend of site 29 in Zone B and site 12 in Zone C was similar to site 2, indicating worsening water quality during the nine years. In addition, site 4 in Zone D remained in an alarming status, to which we should pay much attention.

Table 3 e Classification functions coefficients for discriminant analysis of variations in functional zones of Qiantang River between 2001 and 2004. Parameters

DO CODMn BOD NH3-N V-ArOH TCN THg TPb TCd Cr6þ Petroleum F TP Constant

Standard model

Forward stepwise model

Zone A

Zone B

Zone C

Zone D

10.85 2.20 3.35 .09 32.30 170.4 1.2E4 12.12 6.3E3 210.1 2.27 2.32 38.31 49.13

11.02 2.21 4.11 .14 34.57 489.0 1.3E4 5.56 6.5E3 345.8 1.84 3.11 35.09 52.90

10.58 2.00 3.54 .19 23.98 302.1 1.2E4 18.82 6.4E3 50.3 1.55 1.23 32.14 48.11

11.44 2.29 4.13 .16 35.36 242.5 1.5E4 11.27 7.9E3 172.9 .94 4.37 43.63 54.77

Backward stepwise model

Zone A

Zone B

Zone C

Zone D

Zone A

Zone B

Zone C

Zone D

1.32 2.18 .05

1.28 3.07 .05

1.66 2.94 .13

.98 3.25 .11

.93

1.59

1.04

1.41

.02

0.3

.08

.04

4.5E3 211.2

4.9E3 395.5

4.6E3 69.2

6.1E3 170.7

2.7E3

1.6E3

4.8E3

5.8E3

.31 17.14 27.38

2.06 19.45 31.97

2.78 21.66 27.64

2.39 24.37 31.89

.64

1.19

3.27

1.19

2.78

4.35

3.99

2.97

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 8 1 e1 7 9 5

1787

Fig. 2 e Temporal variations of CODMn, NH3-H, TCd and FL in different functional zones of Qiantang River between 1996 and 2004 (units: mg/l).

As shown in Fig. 4b, NH3-N concentrations of site 2, 14, 26, 27 and 39 in Zone A, 29 in zone B, 12, 15 and 18 in Zone C, and site 4 in Zone D were very high throughout the studied period. Site 33 in Zone B had an upward trend in NH3-N concentrations, signifying more and more serious eutrophication. Most sites

were not affected by TCd pollution during 1996e2004 except some typical sites like 2, 11, 28, 35 and 39 in Zone A, 15 and 16 in Zone B, and site 40 in Zone C (Fig. 4c). Furthermore, the TCd pollution mainly occurred in 1998 and 2003 in the sites mentioned above. Fig. 4d showed that sites 14 and 28 in Zone

Fig. 3 e Temporal trend of CODMn, NH3-H, TCd and FL in each functional zone between 1996 and 2004.

1788

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 8 1 e1 7 9 5

Table 4 e KruskaleWallis test and ManneWhitney test results for discriminant variables in functional zones of Qiantang River between 1996 and 2004. Variable CODMn

NH3-N

TCd

F

Functional zone

Kruskale Wallis test

P value

Decision

Manne Whitney test

P value

Multi-function zone Landscape and recreational zone Drinking water source protection zone Fishing-oriented zone

8.12 5.46

0.43 0.03

3.89

No difference Difference identified

87

0.003

Gap in 2000 and 2004

0.00

Difference identified

96

0.24

Not a significant value

3.37

0.39

No difference

Multi-function zone Landscape and recreational zone Drinking water source protection zone Fishing-oriented zone

8.46 2.75

0.36 0.00

No difference Difference identified

195

0.37

Not a significant value

3.56

0.67

No difference

3.84

0.60

No difference

Multi-function zone Landscape and recreational zone Drinking water source protection zone

7.82 8.35

0.78 0.86

No difference No difference

1.56

0.00

Difference identified

37

0.00

Increase in 1999 Drop in 2002

Fishing-oriented zone

5.38

0.64

No difference

Multi-function zone Landscape and recreational zone Drinking water source protection zone Fishing-oriented zone

0.96 3.69

0.75 0.54

No difference No difference

2.44

0.66

No difference

1.84

0.81

No difference

A had F pollution in most years. So was the case with sites 29 and 32 in Zone B. High F concentrations of site 18 in Zone C should be given special attention as high F concentrations in drinking water can expose public to F poisoning risks.

4.2.

Identification of potential pollution sources

Before performing PCA, Kaiser-Meyer-Olkin (KMO) and Bartlett’s sphericity tests were used to examine the validity of PCA. KMO results for the four functional zones were 0.805, 0.784, 0.816 and 0.775, respectively, and those for Bartlett’s sphericity were 1973, 1433, 2297 and 2356 (P < 0.05), indicating that PCA would be useful for providing significant reductions in dimensionality. PCA with varimax rotation explained 71.2% of the total variance in Zone A, 83.3% in Zone B, 70.6% in Zone C and 81.8% in Zone D, respectively (Table 5). According to Liu et al. (2003) and Huang et al. (2010), factor loadings > 0.75, [0.5e0.75] and [0.3e0.5] were considered to be strong, moderate and weak, respectively. For Zone A, the first varifactor (VF1), accounting for 27.0% of the total variance, had strong and positive loadings on CODMn, BOD, NH3-N, and V-ArOH. High concentrations of NH3-N in surface drainage could come from various sources including natural organic matters decomposition, factories, and fertilizer applications (Hu¨lya and Hayal, 2008). V-ArOH usually originates from chemical plants discharge, since it is widely used in plastics and organic synthesis industry. Meanwhile, the average values of NH3-N and V-ArOH were relatively low while their maximum values were high, denoting that they could stem from some point pollution sources. Thus, with monosodium

Decision

glutamate (MSG) factories and chemical plants scattered in Qiantang River basin, VF1 can be identified as “chemical plants discharge”. For VF2, it explained 18.2% of the total variance, and had strong positive loadings on THg and TCd and a moderate positive loading on TPb. On the whole, the mean levels of the three variables were low, but those of some sites were indeed high. All the elements in this factor are likely to originate from industrial wastewaters discharged into the stream. Elements like Cd, Hg and Pb are known as markers of chemical and tannery plants (Owen and Sandhu, 2000) and electronic industries (Pekey, 2006), including electroplating, dyeing and printed circuit board manufacturing, many of which are present in the study area. According to China Environment Statistical Yearbook (2005) and Zhejiang Statistical Yearbook (2005), the total industrial wastewater discharge is estimated to more than 40000 tones. Thus, VF2 may be interpreted as industrial wastewater pollution. VF3, accounting for 16.2% of the total variance, had strong positive loadings on TCN and moderate positive loadings on F and TCd. Cyanide and Cd in water could be caused by metal mining, organic chemical industries, iron and steel factories as well as the public wastewater treatment plant. F is usually from cement plants, fluorine chemical factories, phosphorus fertilizer plants and smelters (Huang et al., 2010). However, the levels of TCN and TCd in Zone A, except some typical points, were both very low and corresponded to water quality class I/II;, according to GB3838-2002 (Appendix). These typical sites, associated with high F concentrations, were located around fluoride mines. Therefore, the pollution of VF3 should be summed as fluoride mining. The specific geological conditions

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 8 1 e1 7 9 5

1789

Fig. 4 e Temporal trend of PI for CODMn, NH3-H, TCd and FL in each monitoring site between 1996 and 2004.

make Qiantang River one of the important operating waterways of Yangtze River Deltas. Sand mining machinery and vessels shipping are generally distributed throughout the entire area for lack of corresponding bound and supervision mechanism (Su et al., in press). Therefore, VF4, explaining 9.8% of the total variance and with only strong and positive loadings on petroleum, was attributed to vehicle exhaust and sand mining. In Zone B, NH3-N, CODMn, DO, BOD and TP constituted VF1, explaining 29.1% of the total variance. VF1 had strong and positive loadings on NH3-N, CODMn, BOD and TP and a strong negative loading on DO. Phosphorus could originate from both point and non-point sources. Point sources comprise municipal waste treatment plants, factories, and confined livestock operations, while non-point sources mainly include soil erosion

and water runoff from cropland, lawns and gardens, urban areas, small livestock confinement operations, etc (Hu¨lya and Hayal, 2008). If we assumed a 500 m buffer around the monitoring sites in Zone B, the dominant land cover were agricultural land (45%) and forest (25%). So, in view of hydro-chemical conditions of water and land cover patterns, VF1 can be partly identified as “agricultural non-point pollution” with the presence of NH3-N and TP, which were mainly found in agricultural drainage water as mentioned above. Similarly, VF1 had strong positive loadings on both CODMn and BOD, whereas, a strong negative loading on DO. It was, thus, a zone of purely organic pollution indicator parameters from uncontrolled domestic discharges caused by rapid urbanization (Singh et al., 2005; Zhou et al., 2007). Based on the above analysis, VF1 represented nutrient pollution from strong anthropogenic impacts

1790

.198 .004 .300 .125 .013 .207 .108 .128 .113 .006 .063 .958 .205 1.19 9.2 81.8 .662 .757 .746 .804 .099 .497 .037 .117 .093 .090 .026 .003 .158 2.53 19.5 72.6 .329 .048 .226 .062 .057 .115 .073 .914 .937 .046 .299 .110 .811 3.06 23.6 53.1 .400 .234 .066 .155 .983 .716 .954 .004 .064 .988 .034 .124 .452 3.83 29.5 29.5 .034 .141 .052 .148 .025 .060 .455 .005 .106 .516 .285 .803 .110 1.16 10.2 70.6 .238 .076 .055 .419 .111 .349 .532 .764 .660 .218 .039 .210 .126 2.08 21.0 60.4 .868 .881 .932 .677 .026 .038 .071 .006 .421 .070 .742 .146 .896 5.12 39.4 39.4 .110 .264 .292 .243 .619 .284 .052 .038 .021 .441 .231 .264 .809 2.34 18.0 70.0 . 284 .208 .191 .161 .187 .178 .022 .043 .003 .080 .855 .247 .461 1.27 9.8 71.2 DO CODMn BOD NH3-N V-ArOH TCN THg TPb TCd Cr6þ Petroleum F TP Eigenvalue % Total variance Cumulative %

.207 .892 .888 .877 .817 .096 .099 .015 .164 .039 .411 .467 .200 3.51 27.0 27.0

.088 .021 .070 .011 .063 .027 .950 .746 .934 .065 .108 .027 .014 2.37 18.2 45.2

.095 .243 .112 .095 .068 .787 .004 .101 .021 .673 .199 .741 .583 2.17 16.2 61.4

.928 .830 .909 .822 .133 .304 .016 .008 .036 .029 .241 .129 .739 3.79 29.1 29.1

.032 .022 .003 .061 .074 .070 .988 .988 .990 .092 .003 .016 .140 2.98 22.9 52.0

.029 .341 .175 .218 .185 .458 .090 .030 .028 .783 .760 .304 .018 1.73 13.3 83.3

VF4 VF3 VF2 VF3 VF2 VF1 VF4 VF3 VF2 VF1 VF4 VF3 VF1

VF2

Drinking water source protection zone Landscape and recreational zone Multi-function zone Parameters

Table 5 e Loadings of 13 selected variables on VARIMAX rotated factors of different functional zones in Qiantang River.

VF1

Fishing-oriented zone

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 8 1 e1 7 9 5

such as domestic sewage and agricultural activities. As with VF2 of Zone A, VF2 (22.9% of the total variance) had strong positive loadings on TCd, THg and TPb, and was attributed to industrial wastewater discharge (Pekey, 2006). VF3 (18.0% of the total variance) had moderately positive loadings on V-ArOH and TP, representing chemical plants discharge. VF4 (13.3% of the total variance) had moderate positive loadings on petroleum, representing vehicle exhaust and sand mining (Su et al., in press). The pollution pattern in Zone C was in part similar to that in Zone B. As with VF1 of Zone B, VF1 (39.4% of the total variance) had positive loadings on NH3-N, CODMn, BOD and TP and negative loadings on DO, and was then attributed to non-point agricultural runoff, especially from nitrogenous fertilizers (Singh et al., 2005). VF2 (21.0% of the total variance) had strong positive loadings on TCd and TPb, as with VF2 of Zone B, likely representing industrial wastewater discharge (Pekey, 2006). VF3, accounting for 10.2% of the total variance, had only positive loadings on F. This zone is known for high fluoride in soils and groundwater, and the pollution source can be attributed to fluoride weathering and mining. For Zone D, VF2 accounted for 23.6% of the total variance. High loadings of TCd and TPb were displayed, the levels of which were both very low and corresponded to water quality class I, according to GB3838-2002. So, VF2 represented natural factors such as lithology and soil types (Huang et al., 2010). VF3 had positive loadings on NH3-N, likely representing agricultural runoff THg, TCN and V-ArOH were removed in the analysis because of its very low and relatively unchanged concentration (For similar issues, see Huang et al., 2010; Su et al., in press). F levels at almost all monitoring sites were below 1.0 mg/l in Zone D, denoting that there was no or very low pollution. Such a small amount of fluoride may be influenced by local soils entering the river together with the runoff (Huang et al., 2010). Therefore, VF4 could be expressed as “soil weathering”.

4.3.

Source contribution using APCS-MLR

Absolute principle component score was used to calculate source contributions after determining the number and characteristics of possible sources by PCA. Coefficients of determination (R2) in Table 6 reflected that APCS-MLR was relatively accurate. Most sites in Zone A were primarily influenced by chemical pollution (CODMn, 73.2%; BOD, 68.1%; NH3-N, 81.4%; V-ArOH, 59.3%), industrial wastewater discharge (THg, 65.8%; TPb, 73.6%; TCd, 59.2%), fluoride mining (F, 69.2%) and vehicle exhaust and sand mining (petroleum, 83.5%). For Zone B, most monitoring sites were related to domestic sewage and agricultural pollution (NH3-N, 84.4%; TP, 81.6%; CODMn, 58.6%; DO, 64.4%; BOD, 70.8%), industrial wastewater discharge (TCd, 71.6%; THg, 77.4%; TPb, 86.4%), chemical plants discharge (V-ArOH, 65.4%; TP, 71.5%) and vehicle exhaust and sand mining. Most variables of Zone C were associated with domestic sewage and agricultural pollution, industrial wastewater discharge (TPb, 68.7%; TCd, 54.2%) and fluoride weathering and mining (F, 87.5%). For Zone D, most variables were related to agricultural runoff (NH3-N, 85.3%; CODMn, 74.2%; DO, 69.9%; BOD, 70.5%) and soil weathering (F, 78.9%). Additionally, as shown in Table 6, unidentified sources (UIS) in all groups contributed to

1791

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 8 1 e1 7 9 5

Table 6 e Source contribution of each variable of different functional zones in Qiantang River. Variables

Zone A a

R2

VF1 VF2 VF3 VF4 UISe DO CODMn BOD NH3-N V-ArOH TCN THg TPb TCd Cr6þ Petroleum F TP a b c d e

e e e e 73.2 e e e 68.1 e e e 81.4 e e e 59.3 e e e e e 8.6 e e 65.8 e e e 73.6 e e e 59.2 e e e e 29.4 e e e e 83.5 e e 69.2 e e e e e

18.7 e e 1.8 e e e e e e e e 26.9

Zone B b

R2

VF1 VF2 VF3 VF4 UISe .68 .82 .74 .86 .73 .66 .75 .78 .73 .64 .80 .84 .71

64.4 e e e 58.6 e e e 70.8 e e e 84.4 e e e e e 65.4 e e e e e e 77.4 e e e 86.4 e e e 71.6 e e e e e e e e e 79.8 e e e e 81.6 e 71.5 e

10.5 e 7.8 e e 14.6 e e e 8.7 e 17.8 15.2

Zone C c

R2

VF1 VF2 VF3 UISe .72 .73 .69 .83 .64 .76 .80 .75 .71 .68 .79 .84 .85

70.0 e e 66.9 e e 71.3 e e 82.6 e e e e e e e e e e e e 68.7 e e 54.2 e e e e e e e e e 87.5 77.8 e e

7.5 e e 12.8 11.3 2.9 e e e e 19.3 e 25.8

Zone D d

R2

VF1 VF2 VF3 VF4 UISe .76 e e .74 e e .77 e e .83 e e .71 73.5 e .69 81.6 e .73 78.4 e .78 e 83.2 .74 e 76.8 .71 e e .70 e e .81 e e .83 e e

69.9 e 74.2 e 70.5 e 85.3 e e e e e e e e e e e e e e e e 78.9 e e

e e e e 7.1 e e e e 2.9 e 18.5 23.7

.75 .80 .69 .83 .64 .67 .72 .81 .78 .74 .85 .74 .69

multi-function zone. landscape and recreational zone. drinking water source protection zone. fishing-oriented zone. UIS, unidentified sources.

pollution in Qiantang River for all water quality variables, ranging from 1.8% to 26.9%. They represented another major source of pollution besides those identified above. Therefore, field survey is in need to further identify sources of the pollution (Singh et al., 2005; Zhou et al., 2007; Su et al., in press).

4.4.

Methodological prospects

4.4.1. Combination of statistical methods for trend analysis: advantage and applicability This study used an exploratory approach, which combines smoothing and non-parametric statistical tests. This approach has several advantages: (1) its conceptual simplicity (smoothing þ non-parametric tests), (2) its reproducibility with any commercial statistical software that incorporates smoothing and non-parametric tests, (3) its strong reliability (comprehensive application of non-parametric tests to conform smoothing), (4) its non-sensitivity to inequality in time record interval, non-normality in distribution, and spatial and temporal dependence of water quality data, and, (5) its internally consistency. Though this exploratory approach is designed for short time series, it also shows promising applicability for long time series. Of importance is to note that long time series usually present seasonal patterns of temporal changes. The seasonal decomposition method can be used to divide data into a trend component and a periodical component, before applying the exploratory approach. In addition, the Mann-Kendall’s test, non-parametric procedure for monotonic trend detection, can be further applied with the other tests to conform the smoothing results, when the number of observed measurements exceeds ten. In addition, it should be mentioned that internal consistence must be checked before the combination of statistical methods is applied. The exploratory approach used in this paper presents

good internal consistence given the insensitivity of non-parametric tests. However, multivariate methods employed for source apportionment are sensitive to outliers and the nonnormal distributions of geochemical data. Therefore, when we apply various multivariate methods for source apportionment, appropriate data pretreatment should be taken into consideration, including estimation of missing data, examination of normal distributions and data transformation, to make sure that the combination of statistical methods is internally consistent.

4.4.2. Pollution index: from physical numbers to practical references The numerical descriptions, spatial patterns and trends analysis of measurement data for water quality variables could assist in uncovering dynamics of river water pollution. However, these approaches mainly depend on the data set itself, and are thus prone to generate results that lack of practical significance. Putting into practice these physical results requires a higher level of technical expertise to facilitate optimal judgments and decisions. Besides, management of water quality today requires easy-to-understand and intellectual decision-support for national and local governments (WHO and UNICEF, 2005). Basically, PI is a comparatively direct parameter that exhibits practical results by comparing measured values against standards. As for river functional zoning, standards of water quality are determined based on considerations of the likely adverse impact of pollutant levels. The comparison of measured values against standards implicitly conveys the message whether or not the water quality is acceptable, from a sustainable perspective. The results of this study highlighted the usefulness of PI in trend analysis. No significant trend of TCd was identified for monitoring sites in Group A. However, site 31 and 39 presented higher PI values (>1) in 2003. If we converted the practical results into physical values, the year

1792

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 8 1 e1 7 9 5

2003 was found to be significant gap for the nine-year time series. Consequently, we argue that, when characterizing the temporal trend of water quality, periodic reports that include time series of concentrations, and a comparison of the measured values for each pollutant against standards should be both provided.

4.4.3. Variations in source apportionment: role of cross spacetime effects Studies of water pollution source apportionment generally fall into the following framework. First, monitoring/sampling sites are classified into several groups according to the similarity in water quality; then pollution sources are identified via factor analysis/PCA; finally the contributions of different sources are determined using suitable receptor models (Zhou et al., 2007; Huang et al., 2010; Su et al., in press). Different from previous studies, this paper characterized the pollution sources from management prospective. Groups subject to source apportionment were clustered under the zoning framework, rather than using the physical characteristics of measurement data. This kind of source apportionment could provide direct references for managers, however, the intra-level variations within the functional zone may be relatively high. Contrarily, the traditional approach may exhibit relatively low variations within one group, but can not offer easy-to-understand information for managers. Regardless of the advantages and disadvantages of the two approaches, the comparison suggests that source apportionment is influenced by spatial scales and temporal intervals. More specifically, we gave two cases to exemplify the role of cross space-time effects. First, we compared the results with those of Su et al. (in press). Su et al. (in press) used traditional approaches to interpret the pollution sources of Qiantang River for the period of 2001e2004. The identified pollution sources for Sites 1-3 between 2001 and 2004 were domestic sewage and agricultural pollution, industrial wastewater pollution, fluoride mineral weathering, and vehicle exhaust and sand mining. Partly different, chemical plants discharge, industrial wastewater pollution, fluoride mining, and vehicle exhaust and sand mining were interpreted as the main pollution sources in this paper. All these variations bring to light the point that the cross role of space-time effects in source apportionment can not be overestimated. Second, we employed absolute principal component score (APCS) to quantitatively and specifically check the cross space-time effects. Since factor scores are related to source contributions, higher factor scores indicate the higher contribution of the source in the samples (Guo et al., 2009). Fig. 5 shows the annual average contributions for the extracted sources for monitoring sites of multi-function zone between 1996 and 2004. The annual average contributions for individual sources were obtained by averaging the corresponding source contributions for all the monitoring sites. The significant fluctuations of contributions for the extracted sources proved the existence of temporal variations in source apportionment. Similarly, the spatio-temporal variations in contributions of extracted sources for each monitoring site of multi-function zone (Fig. 6) also signified the space-time dependent features of source apportionment. Above analysis demonstrates that role of cross space-time effects should be

Fig. 5 e Annually temporal variations in source apportionment for monitoring sites of multi-function zone.

taken seriously in source apportionment. Besides, APCS, which contributes to quantitative description of variations in source apportionment, can be applied to characterize the cross space-time effects. One limitation of this study is that the seasonal variations are not discussed. Further study related to the corresponding variations has to be carried out. Another limitation is that the monitoring sites were spatial unevenly distributed. We were not able to study other categories of functional zones, such as industry-oriented zone and natural reserved zone.

4.5.

Management implications

The present methods for comprehensive applications of different multivariable statistics in temporal trend analysis and apportionment of water pollution source could provide a technical support for governmental agencies to implement integrated river basin management. Besides, it is designed for, but not limited, its applications at large river basin scale. Firstly, the comparison and identification of pollution sources in different functional zones would help maintain ecosystem health and make holistic policies by emphasizing the variations in different zones. Secondly, based on the information extracted from PCA, we may develop innovative planning schemes or adjust the currently used management practice to a more impartial and effective manner. The source apportionment information is also helpful for the local government to explain the dissension from the towns whose responsibility for aquatic ecosystem conservation rather than overexploitation. Thirdly, the source contributions assessment, which determines the relative importance of different variables, could be applied to optimize future monitoring program by reducing sampling frequency, the number of monitoring sites and parameters, and thus, the subsequent cost (Su et al., in press). Finally, source apportionment framework along with function zoning and planning can be integrated into an Integrated River Basin Management Decision Support System

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 8 1 e1 7 9 5

1793

Fig. 6 e Spatio-temporal variations in absolute principal component score of extracted sources for each monitoring site of multi-function zone.

that provides a GIS platform to effectively manage land and water resources with the involvement of multiple stakeholders including governments and the general public.

5.

Conclusions

This study investigated the temporal trend and identified pollution sources in different functional zones of Qiantang River using a nine-year (1996e2004) data set. CODMn, NH3-N, TCd and F were discriminant variables of variations in different functional zones. An exploratory approach, which combined smoothing and non-parametric statistical tests, was used to characterize temporal trends for the four variables among different functional zones. Pollution index was further applied to analyze within-group variations in trends for different functional zones. Based on the results, we believe that this exploratory approach is not only effective for short time series but also shows promising applicability for long time series. In addition, when characterizing the temporal trend of water quality, periodic reports that include time series of concentrations, and a comparison of the measured

values for each pollutant against standards should be both provided. Sources responsible for pollution identified by PCA differed among the four functional zones. Generally, Zone A and Zone B were mainly affected by human activities, while Zone C and Zone D by natural process; point pollution was the major source for Zone A, whereas non-point source for Zone B, C, and D. Receptor-based source apportionment through APCSMLR revealed that unidentified sources were another major latent source. Therefore, field survey is needed to identify these additional sources of the pollution. Furthermore, spatiotemporal variations in source apportionment signified the influence of spatial scales and temporal intervals. These results demonstrate that role of cross space-time effects should be taken seriously in source apportionment. Besides, APCS, which contributes to quantitative description of variations in source apportionment, can be applied to characterize the cross space-time effects. Our work suggests that the management-oriented source apportionment under the framework of river function zoning, which could provide a technical support for governmental agencies to implicate integrated river basin management, is in

1794

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 8 1 e1 7 9 5

great need, especially for developing countries. Based on the insight that the links between science and management should be a vital part of water science research, we argue that the advanced study of water pollution source apportionment should be more focused on the integration of source apportionment and different management frameworks, the uncertainties associated with cross space-time effects, the comparison between physical data-based and managementoriented source apportionment, and the relevance of source apportionment to integrated river basin management. In particular, how to develop or apply approach/model that presents direct practical significance has to be understood.

Acknowledgements We thank the Environmental Monitoring center of Zhejiang Province for providing the data. We are also very grateful to the editor and two reviewers for their constructive comments and suggestions that greatly improved this paper. This work was supported in part by the National Key Project Grant (#2008ZX07101-006).

Appendix A.

Environmental guideline of national quality standards for surface waters, China (GB3838-2002) (units: mg/l) Parameters

Category of water quality standards First

NH3-N F CODMn Cr6þ V-ArOH DO BOD Petroleum TCd THg TP TPb TCN

<0.15 <1.0 <2.0 <0.01 <0.002 >90% of saturation <3 <0.05 <0.001 <0.00005 <0.02 <0.01 <0.005

Second Third Fourth Fifth 0.5 1.0 4.0 0.05 0.002 6.0 3.0 0.05 0.005 0.00005 0.1 0.01 0.05

1 1.5 2 1.0 1.5 1.5 6.0 10.0 15.0 0.05 0.05 0.1 0.005 0.01 0.1 5.0 3.0 2.0 4.0 6.0 10.0 0.05 0.5 1.0 0.005 0.005 0.01 0.0001 0.001 0.001 0.2 0.3 0.4 0.05 0.05 0.1 0.2 0.2 0.2

references

Brierley, G., Fryirs, K., Outhet, D., Massey, C., 2002. Application of the River Styles framework as a basis for river management in New South Wales, Australia. Applied Geography 22 (1), 91e122. Chang, H.J., 2008. Spatial analysis of water quality trends in the Han River basin, South Korea. Water Research 42 (13), 3285e3304. Cheruvelil, K.S., Soranno, P.A., Bremigan, M.T., Wagner, T., Martin, S.L., 2008. Grouping lakes for water quality assessment and monitoring: the roles of Regionalization and spatial scale. Environmental Management 41 (3), 425e440.

China Bureau of Statistics, 2005. China Environment Statistical Yearbook. China Statistics Press, Beijing (In Chinese). China Ministry of Water Resources, 2007. Revisions of Integrated River Basin Planning Document of Chinese Government (In Chinese). Cun, C., Vilagines, R., 1997. Time series analysis on chlorides, nitrates, ammonium and dissolved oxygen concentrations in the Seine river near Paris. Science of the Total Environment 208 (1e2), 59e69. Davis, J.C., 1986. Statistics and Data Analysis in Geology. John Wiley & Sons, Inc, New York. Guo, H., Ding, A., So, K., Ayoko, G., Li, Y., Hung, W., 2009. Receptor modeling of source apportionment of Hong Kong aerosols and the implication of urban and regional contribution. Atmospheric Environment 43 (6), 1159e1169. Gupta, D.A., 2008. Implication of environmental flows in river basin management. Physics and Chemistry of the Earth 33 (5), 298e303. Huang, F., Wang, X., Lou, L., Zhou, Z., Wu, J., 2010. Spatial variation and source apportionment of water Pollution in Qiantang River (China) using statistical techniques. Water Research 44 (5), 1562e1572. Huang, X., Chen, W., Cai, Q., 1999. Survey, observation and analysis of lake ecology. Standards Press of China, Beijing (In Chinese). Hu¨lya, B., Hayal, B., 2008. Water pollution sources assessment by multivariate statistical methods in the Tahtali Basin, Turkey. Environmental Geology 54 (2), 275e282. Jonnalagadda, S.B., Mhere, G., 2001. Water quality of the odzi river in the eastern highlands of Zimbabwe. Water Research 35 (10), 2371e2376. Kaiser, H.F., 1958. The varimax criterion for analytic rotation in factor analysis. Psychometrika 23, 187e200. Kruskal, W.H., Wallis, W.A., 1952. Use of ranks in one-criterion variance analysis. Journal of the American Statistical Association 47 (260), 583e621. Kuppusamy, M.R., Giridhar, V.V., 2006. Factor analysis of water quality characteristics including trace metal speciation in the coastal environmental system of Chennai Ennore. Environment International 32 (2), 174e179. Liu, C., Lin, K., Kuo, Y., 2003. Application of factor analysis in the assessment of groundwater quality in a Blackfoot disease area in Taiwan. Science of the Total Environment 313 (1e3), 77e89. Love, D., Hallbauer, D., Amos, A., Hranova, R., 2004. Factor analysis as a tool in groundwater quality management: two southern African case studies. Physics and Chemistry of the Earth 29 (15e18), 1135e1143. Ma, J., Ding, Z., Wei, G., Zhao, H., Huang, T., 2009. Sources of water pollution and evolution of water quality in the Wuwei basin of Shiyang river, Northwest China. Journal of Environmental Management 90 (2), 1168e1177. Mann, H.B., Whitney, D.R., 1947. On a test of whether one of two random variables is stochastically larger than the other. Annals of Mathematical Statistics 18 (1), 50e60. Owen, R.B., Sandhu, N., 2000. Heavy metal accumulation and anthropogenic impacts on Tolo Harbour, Hong Kong. Marine Pollution Bulletin 40 (2), 174e180. Pekey, H., Karakas, D., Bakoglu, M., 2004. Source apportionment of trace metals in surface waters of a polluted stream using multivariate statistical analyses. Marine Pollution Bulletin 49, 809e818. Pekey, H., 2006. The distribution and sources of heavy metals in IzmitBay surface sediments affected by a polluted stream. Marine Pollution Bulletin 52 (10), 1197e1208. Sabatini, M.C., Verdiell, V.A., Rodrı´guez Iglesias, R.M., Vidal, M., 2007. A quantitative method for zoning of protected areas and its spatial ecological implications. Journal of Environmental Management 83 (2), 198e206.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 8 1 e1 7 9 5

Singh, K.P., Malik, A., Sinha, S., 2005. Water quality assessment and apportionment of pollution sources of Gomti river (India) using multivariate statistical techniquesda case study. Analytica Chimica Acta 538 (1e2), 355e374. Su, S., Zhi, J., Lou, L., Huang, F., Chen, X., Wu, J. Spatio-temporal patterns and source apportionment of pollution in Qiantang River (China) using neural-based modeling and multivariate statistical techniques. Physics and Chemistry of the Earth. doi:10.1016/j.pce.2010.03.021, in press. Temme, C., Ebinghaus, R., Einax, J.W., Steffen, A., Schroeder, W.H., 2004. Time series analysis of long-term data sets of atmospheric mercury concentrations. Analytical and Bioanalytical Chemistry 380 (3), 493e501. Tukey, J.M., 1977. Exploratory Data Analysis. Addison-Wesley, Reading, MA. Wang, J., Da, L., Song, K., Li, B., 2008. Temporal variations of surface water quality in urban, suburban and rural areas during rapid urbanization in Shanghai, China. Environmental Pollution 152 (2), 387e393. Weng, Q., 2007. A historical perspective of river basin management in the Pearl River Delta of China. Journal of Environmental Management 85 (4), 1048e1062. Wei, F., Kou, H., Hong, S., 1989. Methods for the Examination of Water and Wastewater. China Environmental Science Press, Beijing (In Chinese). Wei, W., 1989. Time Series Analysis: Univariate and Multivariate Methods. Addison-Wesley, New York. WHO and UNICEF, 2005. WHO and UNICEF, water for Life. WHO and UNICEF, Switzerland.

1795

Yang, T., Chen, X., Xu, C., Zhang, Z., 2009. Spatio-temporal changes of hydrological processes and underlying driving forces in Guizhou region, Southwest China. Stochastic Environmental Research and Risk Assessment 23 (8), 1071e1087. Yue, S., Wang, C., 2002. Power of the ManneWhitney test for detecting a shift in median or mean of hydro-meteorological data. Stochastic Environmental Research and Risk Assessment 16 (4), 307e323. Zhang, C., Wu, L., Luo, Y., Zhang, H., Christie, P., 2008. Identifying sources of soil inorganic pollutants on a regional scale using a multivariate statistical approach: role of pollutant migration and soil physicochemical properties. Environmental Pollution 151 (3), 470e476. Zhang, Q., Xu, C., Tao, H., Jiang, T., Chen, Y., 2010. Climate changes and their impacts on water resources in the arid regions: a case study of the Tarim River basin, China. Stochastic Environmental Research and Risk Assessment 24 (3), 349e358. Zhejiang Environmental Protection Bureau, 2006. Scheme for River Function Zoning and Water Environment Zoning of Zhejiang Province Document of Zhejiang Government (In Chinese). Zhejiang Provincial Bureau of Statistics, 2005. Zhejiang Statistical Yearbook. China Statistics Press, Beijing (In Chinese). Zhou, F., Huang, G., Guo, H., Zhang, W., Hao, Z., 2007. Spatiotemporal patterns and source apportionment of coastal water pollution in eastern Hong Kong. Water Research 41 (15), 3429e3439.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 9 6 e1 8 0 6

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Colloid straining within saturated heterogeneous porous media Alexis A. Porubcan, Shangping Xu* Department of Geosciences, University of Wisconsin e Milwaukee, Milwaukee, WI 53211, USA

article info

abstract

Article history:

The transport of 0.46 mm, 2.94 mm, 5.1 mm and 6.06 mm latex particles in heterogeneous

Received 21 June 2010

porous media prepared from the mixing of 0.78 mm, 0.46 mm and 0.23 mm quartz sands

Received in revised form

was investigated through column transport experiments. It was observed that the 0.46 mm

11 October 2010

particles traveled conservatively within the heterogeneous porous media, suggesting that

Accepted 24 November 2010

under the experimental conditions employed in this research the strong repulsive inter-

Available online 4 December 2010

actions between the negatively charged latex particles and the clean quartz sands led to minimal colloid immobilization due to physicochemical filtration. The immobilization of

Keywords:

the 2.94 mm, 5.1 mm and 6.06 mm latex particles was thus attributed to colloid straining.

Colloid transport

Experimental results showed that the straining of colloidal particles within heterogeneous

Colloid straining

sand mixtures increased when the fraction of finer sands increased. The mathematical

Groundwater contamination

model that was developed and tested based on results obtained using uniform sands (Xu et al., 2006) was found to be able to describe colloid straining within heterogeneous porous media. Examination of the relationship between the best-fit values of the clean-bed straining rate coefficients (k0) and the ratio of colloid diameter (dp) and sand grain size (dg) indicated that when number-average sizes were used to represent the size of the heterogeneous porous media, there existed a consistent relationship for both uniform sands and heterogeneous sand mixtures. Similarly, the use of the number-averaged sizes for the heterogeneous porous media produced a uniform relationship between the colloid straining capacity term (l) and the ratio of dp/dg for all the sand treatments. ª 2010 Elsevier Ltd. All rights reserved.

1.

Introduction

A thorough understanding of the transport of colloids in the subsurface system is essential to the protection of groundwater resources, which account for 98% of domestic water supply in the United States (Hutson et al., 2004), from contamination by pathogenic microorganisms and colloidbound contaminants (Chu et al., 2003; Crist et al., 2005; Keller and Auset, 2007; Mccarthy and Zachara, 1989; Ryan and Elimelech, 1996; Saiers, 2002; Saiers and Hornberger, 1996; Tufenkji et al., 2002). For several decades, the study of colloid immobilization within saturated porous media has

been primarily centered on their physicochemical filtration, which refers to the attachment of particles onto the surfaces of the solid matrix (Logan et al., 1995; Ryan and Elimelech, 1996; Yao et al., 1971). There is a growing body of evidence suggesting that a variety of processes other than physicochemical filtration such as wedging (Li et al., 2006), hydrodynamic drag (Li et al., 2005), retention within relatively “immobile” regions (Torkzaban et al., 2008) and straining (Bradford et al., 2002; McDowell-Boyer et al., 1986; Tufenkji et al., 2004) can affect the transport behavior of colloid-sized particles in saturated porous media. For relatively large colloidal particles such as Crypotosporidium oocysts, colloid

* Corresponding author. Tel.: þ1 414 229 6148. E-mail address: [email protected] (S. Xu). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.11.037

1797

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 9 6 e1 8 0 6

straining, which represents the process that occurs when a pore space is too small to allow for the passage of colloidal particles, could play important and sometimes dominant roles in their immobilization in the groundwater system (Bradford et al., 2006, 2002; Bradford and Torkzaban, 2008; Foppen et al., 2005; Herzig et al., 1970; McDowell-Boyer et al., 1986; Tufenkji et al., 2004; Xu et al., 2006). Recent findings indicate that the rate of colloid straining within saturated porous media is sensitive to the ratio of colloid diameter (dp) to sand-grain diameter (dg), the shape and surface roughness of the solid matrix, colloid shape and size non-uniformity, pore-scale hydrodynamics and porewater chemistry (Auset and Keller, 2004; Bradford et al., 2003; Keller and Auset, 2007; Li et al., 2006; Shen et al., 2008; Tufenkji et al., 2004; Xu et al., 2006, 2008; Xu and Saiers, 2009). Based on experimental results obtained under the presence of both colloid attachment and straining, Bradford et al. (2003) proposed that the straining rate coefficient was related to the ratio of particle diameter and the median sand grain size through a power function. Xu et al. (2006) examined straining kinetics of colloidal particles under conditions unfavorable for colloid attachment and reported that above a threshold value of 0.008, the straining rate coefficient increases linearly with the ratio of colloid diameter and the average diameter of sand grains. This linear relationship was extended to non-spherical particles, the straining of which was found to be governed by their minor sizes due to their orientation along the flow direction (Xu et al., 2008). Results from experiments conducted with monodisperse and bidisperse colloid suspensions suggested that colloid size-nonuniforminity can have a significant impact on straining kinetics: the straining of smaller particles was enhanced by the presence of larger particles due to the blockage of pore opening by the larger particles while the straining of larger particles was reduced by the smaller particles primarily due to the faster depletion of straining capacities (Xu and Saiers, 2009). Tufenkji et al. (2004) reported that while there was little straining of latex particles and Crypotosporidium oocysts in spherical and smooth glass beads, there was significant colloid immobilization due to straining in similar transport experiments conducted with natural sand grains of similar sizes and suggested that collector shape and surface roughness can affect colloid straining. Bradford et al. (2007) indicated that the physical and chemical factors as well as flow hydrodynamics can collectively affect colloid straining within saturated porous media.

Through theoretical analysis and experimental investigations, Shen et al. (2008) showed that colloid attachment, particularly at the grain-to-grain contact, can enhance colloid straining. Previous studies on colloid straining have predominantly relied on experiments conducted with uniform sand packs (e.g., Bradford et al., 2003, 2007; Xu et al., 2006; Xu et al., 2008). Natural soil and sediments, however, are usually characterized with physical heterogeneity which may originate from the mixing of sand grains of various sizes or the inclusion of soil lenses within a solid matrix that displays different grain size distribution and hydraulic properties (i.e. structured heterogeneity) (Bradford et al., 2004; Chesnaux and Allen, 2008; Dong et al., 2002; Fuller et al., 2000; Saiers et al., 1994; Tillmann et al., 2008). Studies that explore the influence of porous media heterogeneity on colloid straining are thus critical in advancing our understanding of colloid transport within natural heterogeneous porous media. The primary objectives of this research include i) to examine colloid straining within heterogeneous porous media; ii) to evaluate the mathematical model that was developed based on results obtained using uniform sands (Xu et al., 2006); and iii) to determine the appropriate measure of the size of heterogeneous porous media that can be used to predict colloid straining kinetics. Findings from this research provide new insights into the kinetics of colloid straining within heterogeneous porous media, which is essential to our understanding of colloid transport in natural groundwater system.

2.

Materials and methods

2.1.

Porous media

The column experiments were conducted with high-purity quartz sand (US Silica). The sand was separated with stainless steel sieves (US Standard Testing) into three classes that correspond to mesh sizes of 20e25 (median size: 0.78 mm), 35e40 (0.46 mm), and 60e70 (0.23 mm), respectively. The sieved sands were cleaned through a series of hot, concentrated nitric acid (to remove surface impurities, such as iron hydroxide and organic coatings, that could promote physiochemical deposition of colloids), sodium hydroxide (to remove colloid-sized particles that were attached to the sand surfaces), and deionized water washes (Xu et al., 2006,2008).

Table 1 e Properties of the sand mixtures used in the column transport experiments. Porous media

Sand Sand Sand Sand

mixture mixture mixture mixture

1 2 3 4

Mixing ratio (mass based) 0.78 mm sand

0.46 mm sand

0.23 mm sand

30 30 30 0

70 35 0 50

0 35 70 50

Porosity

Mass-averaged size (mm)a

Number-averaged size (mm)b

0.352 0.320 0.344 0.345

0.556 0.476 0.395 0.345

0.486 0.266 0.236 0.256

P a The mass-averaged size is calculated as (Midi), where Mi is the mass fraction of sand i and di is the median size of sand i, i ¼ 1, 2 and 3. P P b The number averaged size is calculated as ðMi =d2i = Mi =d3i Þ , where Mi is the mass fraction of sand i and di is the median size of sand i, i ¼ 1, 2 and 3.

1798

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 9 6 e1 8 0 6

Once clean, the sand was dried in an oven at 80  C. The sand mixtures used in the column transport experiments were prepared by mixing the uniform, cleaned quartz sands at various ratios (Table 1). Using the bulk density method, the porosity of the 0.78 mm, 0.46 mm and 0.23 mm sands was measured as 0.356, 0.363 and 0.373, respectively. The porosity of heterogeneous sand mixtures was similarly determined (Table 1). Both the number-averaged and mass-averaged sizes of the sand mixtures were calculated and are listed in Table 1. The thoroughly-cleaned sands were pulverized and the colloid-sized quartz particles thus produced were suspended in Nanopure water to determine the zeta potential values using a Malvern zetasizer (Table 2) (Liu et al., 2009).

2.2.

Latex microspheres

Green carboxylated latex microspheres (Magsphere Inc.) were used in the column experiments. Upon receipt of the colloids from the manufacturer, surfactants used during manufacturing were removed by collecting the colloids on Durapore membrane filters (Millipore) and resuspending the colloids in Nanopure water (Barnstead, 18.2 MU cm) for a minimum of eight times. Images of the latex particles were then obtained using a scanning electron microscope (SEM, Hitachi S-4800) and the sizes of the microspheres were measured as 0.46, 2.94, 5.1 and 6.06 mm, respectively. The zeta potential of the latex particles suspended in Nanopure water was measured using a Malvern Zetasizer (Table 2).

2.3.

Column transport experiments

Glass chromatography columns (Chromaflex, Kimble/Kontes) measuring 2.5 cm in diameter and 15 cm in length were used in the column experiments. Custom-made acrylic plates with 22 1-mm openings were used as end fittings to support the sand. A stainless steel membrane with 0.051 mm openings (Spectrum Laboratories) was placed immediately above the acrylic plate to prevent the sand from moving. The columns were vertically oriented and packed by slowly pouring the uniform sand or sand mixtures into Nanopure water at the bottom of the column. Layering of sand was minimized by mixing each new addition of sand with the previous additions and maintaining the water level in the column approximately 1 cm above the sand surface. All column experiments were conducted in duplicate and were equilibrated by pumping >40 pore volumes of Nanopure water through the columns at a specific discharge of 0.306 mL/

min. This flow condition is comparable to that encountered in riverbank filtration, and is at the higher end of the flow rates under natural gradient conditions in groundwater aquifers (Harvey et al., 1993; Havelaar et al., 1995). The variation (w10%) in the porosity of the porous media used in the experiments led to a comparable variation (w10%) in the interstitial velocity. Immediately following equilibration a transport experiment was initiated by introducing colloids suspended in Nanopure water to the top of the column. The colloids were suspended in Nanopure water to maximize the electrostatic repulsion between the colloids and the silica sands and to minimize colloid attachment to the sand surfaces (Xu et al., 2006, 2008). The concentrations of the 0.46 and 2.94, 5.1, and 6.06 mm colloids were 1.86  109, 7.12  106, 2.73  106, and 2.44  106 particles/mL, respectively. Following 2 h (7e8 pore volumes) of colloid application, the column was flushed for 1 h with colloid-free Nanopure water at the same flow rate. Effluent samples were collected by a fraction collector at 3 min intervals and analyzed for colloid concentrations with a spectrophotometer (Shimadzu UV-1700 PharmaSpec) at a wavelength of 660 nm. Conservative tracer tests with nitrate were conducted separately and in the same fashion as the colloid transport experiments. The concentration of applied nitrate was 0.25 mM and effluent samples were measured with a spectrophotometer at a wavelength of 220 nm equipped with flowthrough cuvette cells (Abudalo et al., 2005). Immediately following the column experiments with the 5.1 and 6.06 mm colloids, the profile of the strained colloids was determined. The top end fitting was removed and the sand was excavated with a spatula from top to bottom under saturated conditions at 1.5 cm increments (Bradford and Bettahar, 2005; Bradford et al., 2004; Xu et al., 2006). Each increment was placed into a 100 ml media bottle with 50 ml of Nanopure water. After vigorously shaking each media bottle by hand, 5 mL of supernatant was withdrawn, and the concentration of colloids was measured with a spectrophotometer at a wavelength of 660 nm (Xu et al., 2006). The remaining supernatant was poured off and the sand samples were dried at 80  C before measuring the mass of sand in each increment. The concentration of strained colloids within each increment was expressed as the mass of colloids per unit mass of sand.

2.4.

Mathematical model of colloid straining

An empirical model was previously proposed and validated to describe the straining of spherical and non-spherical

Table 2 e Zeta potential values of the latex microspheres and quartz sands (suspended in deionized water) and energy barriersa (peak interaction energies, unit kTb) between the microspheres and the quartz sands. Colloid/sand 0.78 mm sand (62.4 mV) 0.46 mm sand (58.7 mV) 0.23 mm sand (58.9 mV)

0.46 mm (61.9 mV)

2.94 mm (60 mV)

5.1 mm (48.4 mV)

6.06 mm (58.3 mV)

1270 1190 1195

7850 7390 7420

10,250 9910 9930

15,700 14,800 14,900

a Energy barriers were computed by summation of energies associated with van der Waals attraction and electrostatic double layer repulsion (Elimelech et al., 1998). The retarded van der Waals interaction is calculated based on the expression proposed in (Gregory, 1981) using a Hamaker constant of 1  1020 J. The electrostatic double layer interaction is calculated using the expression of (Hogg et al., 1966). b k is the Boltzmann constant, and T is absolute temperature in Kelvin.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 9 6 e1 8 0 6

colloidal particles as well as non-uniform colloid mixtures in saturated porous media (Xu et al., 2006, 2008; Xu and Saiers, 2009). This model was based on the assumptions that: 1) straining is an irreversible process; 2) straining is the sole mechanisms of colloid retention and 3) straining rates will decline over time as the accumulation of colloids fills pore spaces and flow and colloid transport are diverted to neighboring pores with dimensions larger than those of the colloids. The mathematical form of this model consists of the following two equations: vC rb vS vC v2 C þ ¼ v þ D 2 n vt vt vz vz . S l rb vS ¼ ko Ce n vt

(1)

(2)

where C is the concentrations of suspended colloids (mg/L), t is time (hour), rb is the bulk density of the solid matrix (g/L), n is porosity (unitless), S is the concentration of strained colloids (mg/g), n is the average linear porewater velocity (cm/hour), z is the coordinate parallel to flow (cm), D is the hydrodynamic dispersion coefficient (cm2/hour), ko is the straining rate coefficient for clean-bed conditions (i.e., S z 0 at t ¼ 0) (hour1), and l determines how fast the straining rate will decrease with time due to particle accumulation inside pore spaces that are capable of straining colloids and, as explained in Xu et al. (2006), reflects but does not equal to colloid straining capacity (mg/g). In equation (2), S can be greater than l. For instance, when S equals to 4l, the straining rate declines to 0.018koC. The straining rate becomes vanishingly small with further increases in S. The coupled equations (1) and (2) were numerically solved by a second-order, implicit finite-difference scheme. The boundary conditions at the column inlet was first-type (i.e., constant concentration). Values of v were calculated from specific discharge and porous media porosity and values of D were estimated by least-square inversion of the tracer test data. Values of k0 and l were then estimated in an inversion fashion by minimizing the sum of squared differences between experimentally measured and model-predicted colloid breakthrough concentrations using the LevenbergeMarquardt algorithm (The Mathworks, 2006).

3.

Results and discussions

3.1.

Results of column transport experiments

When suspended in deionized water, the surfaces of the carboxylate-modified latex particles and thoroughly-cleaned quartz sands were negatively charged as suggested by the zeta potential measurements (Table 2). Calculations based on the Derjaguin-Landau-Verwey-Overbeek (DLVO) theory showed strong repulsive particle-sand interactions and the absence of secondary minimum in the particleesand interaction profiles (Table 2). Consistent with both the DLVO calculations and previously reported results which indicated minimal physicochemical filtration of latex particles under similar experimental conditions (Xu et al., 2006,2008; Xu and Saiers, 2009),

1799

the breakthrough concentrations of the 0.46 mm colloids were w100% and closely resembled those of nitrate, the conservative tracer, in columns experiments with all seven sand treatments (Fig. 1). Considering that all the latex particles used in the experiments had carboxylic surface functional groups and the repulsive interaction between latex particles and sand grains increased with particle size (Table 2), the immobilization of the larger particles within the sand packs was thus attributed to straining and physicochemical deposition of the colloidal particles was negligible (Xu et al., 2006,2008; Xu and Saiers, 2009). The breakthrough concentrations of the 2.94 mm, 5.1 mm and 6.06 mm colloidal particles in the uniform sands were lower than their influent concentrations, suggesting colloid immobilization due to straining (Fig. 2). Analysis of the breakthrough data indicated that the straining of colloidal particles of one specific size increased when the size of the sand grains decreased. The fraction of the 6.06 mm colloids that were retained was 8.7% in the 0.78 mm sand, increased slightly to 10.3% in the 0.46 mm sands, and further increased to 23.9% in 0.23 mm sands. Similarly, 5.4%, 6.6% and 9.4% of the 2.94 mm colloids were immobilized within the 0.78 mm, 0.46 mm and 0.23 mm sands, respectively. Straining of the colloids with sizes greater than 0.46 mm was apparent in the sand mixtures (Fig. 3). It was observed that the introduction of finer sands led to more colloid straining. For instance, when 70% of the 0.78 mm sand was replaced with finer 0.46 mm sand (i.e., sand mixture 1), the percentage of colloids immobilized in the sand packs increased from 8.7% to 10% based on the integration of the breakthrough curves (Figs. 2 and 3). When half of the 0.46 mm sand in sand mixture 1 was replaced with the 0.23 mm sand (i.e., sand mixture 2 that contains 35% of 0.46 mm sand and 35% of 0.23 mm sand), there was a significant drop in colloid breakthrough concentrations and the percentage of the immobilized 6.06 mm colloids increased from 10% to 17.4% (Fig. 3). Replacing the remaining 35% of the 0.46 mm sand in sand mixture 2 with the finer 0.23 mm sand (i.e., sand mixture 3, which contains 30% of 0.78 mm sand and 70% of 0.23 mm sand) led to a small drop in the breakthrough concentrations of 6.06 mm colloids and the percentage of the 6.06 mm colloids that were immobilized increased from 17.4% to 21.7% (Fig. 3). Similarly, when the 30% of the 0.78 mm sand in sand mixture 2 was replaced with 15% of 0.46 mm sand and 15% of 0.23 mm sand (i.e., sand mixture 4 which contains 50% of 0.46 mm sand and 50% of 0.23 mm sand), the fraction of the 6.06 mm colloids that were strained within the columns increased from 17.4% to 19.5%. Concentrations of the 5.1 mm and 6.06 mm colloids in the sand packs were determined upon the completion of column transport experiments. Their mass balances, expressed as the mass of particles recovered in the effluent plus particles extracted at the end of experiment normalized by the total mass of particles injected into the columns, ranged within 91.3%e94.8%. Consistent with the breakthrough concentrations, profiles of the strained colloids also suggested that finer sands facilitate colloid straining (Fig. 4). The average concentration of the 6.06 mm colloids in the 0.78 mm, 0.46 mm and 0.23 mm sands increased from 0.012 mg/g to 0.017 mg/g and further to 0.068 mg/g. As discussed previously, sand mixtures

1800

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 9 6 e1 8 0 6

Fig. 1 e Breakthrough concentrations of (A) nitrate (the conservative tracer) and (B) 0.46 mm colloids. The result of one of the duplicate experiments is shown for clarity purpose. Legends represent the porous media (uniform sand or sand mixtures) used in the column experiments.

1, 2 and 3 represented progressively adding finer sands to the 0.78 mm sand. The average concentrations of the retained 6.06 mm colloids in the 0.78 mm sand, sand mixtures 1, 2 and 3 were 0.012 mg/g, 0.015 mg/g, 0.036 mg/g and 0.053 mg/g, respectively. The depth profiles of strained colloids also displayed fairly small variability (Xu et al., 2006, 2008; Xu and Saiers, 2009). This lack of vertical variability is related to the high breakthrough concentrations (88% of colloid influent concentration) toward the end of the column experiments. In theory, when the breakthrough concentrations approach 100%, all the straining capacities are completely depleted and the concentrations of strained colloids will approach the limit defined by the straining capacities and it is expected that there

should be little variability in the concentrations of strained colloids.

3.2. Comparison of experimental and modeled results and best-fit values of k0 and l The mathematical model of equations (1) and (2) was numerically solved and model predictions compared favorably (R2  0.98) with the measured breakthrough concentrations of the 2.94 mm, 5.1 mm and 6.06 mm colloids (Figs. 2 and 3). Particularly, the model was found to be capable of simulating colloid straining within porous media composed of sand mixtures (Fig. 3). In addition, predictions from the

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 9 6 e1 8 0 6

Fig. 2 e Breakthrough curves of (A) 2.94 mm, (B) 5.1 mm and (C) 6.06 mm colloids in 0.78 mm, 0.46 mm and 0.23 mm sands. Symbols and lines represent measured and modelpredicted concentrations, respectively.

mathematical model were in reasonably good agreement with the measured depth profiles of strained colloids in both uniform sands and sand mixtures (Fig. 4). When equations (1) and (2) were numerically solved, the best-fit values of k0 and l were obtained through inverse simulation. The best-fit values of k0 ranged from 0.42 h1 to 4.52 h1. The lower values of k0 were associated with smaller

1801

Fig. 3 e Breakthrough curves of (A) 2.94 mm, (B) 5.1 mm and (C) 6.06 mm colloids in the sand mixtures. Symbols and lines represent measured and model-predicted concentrations, respectively.

colloidal particles and coarser sand, while the greater values of k0 were observed for larger particles and porous media containing substantial quantities of finer sands. The second fitting parameter, l, reflects the capacity of colloid straining. The best-fit values of l were within the range of 0.0043e0.098 mg/g. The ratio of the average concentrations of strained colloids and l (i.e., S/l) was greater than 2 for both the

1802

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 9 6 e1 8 0 6

Fig. 4 e Depth profiles of immobilized colloid concentrations. Error bars denote standard deviations for duplicate measurements. Symbols and lines represent measured and model-predicted concentrations, respectively.

5.1 mm and 6.06 mm colloids, suggesting that at the end of the column experiments, the effective straining rate was more than 87% (i.e., 1e2) lower than the clean-bed straining rates. The significant reduction in effective straining rate occurred along with the depletion of straining capacity. As a result, it is anticipated that the profile of the strained colloids became close to the straining capacity and displayed small variation. The best-fit values of k0 and l from the experiments using uniform sands in this research are in good agreement with previously reported values that were obtained under similar experimental conditions (Xu et al., 2006,2008). For instance, for the combination of 5.1 mm colloids and 0.78 mm sand, the best-fit values of k0 and l reported in (Xu et al., 2006) were 0.81(0.41) hour1 and 0.0175(0.004) mg/g, respectively. The corresponding values calculated in this research were 0.79 (0.014) hour1 and 0.01(0.001) mg/g, respectively. In Xu et al. (2008), the values of k0 and l for the straining of 6.1 mm colloids in 0.78 mm sand were 0.793(0.099) hour1 and 0.0262(0.001) mg/g. The values of k0 and l for the straining of 6.06 mm colloids in 0.78 mm sand estimated in this research were 0.835(0.021) hour1 and 0.02(0.003) mg/ g, respectively.

3.3. Relationship between the values of the best-fit parameters and properties of colloids and porous media It was previously reported that the clean-bed straining rate coefficients, k0, and the straining capacity term, l, were closely related to the ratio of particle size (dp) and sand grain size (dg) (Bradford et al., 2003,2002; Xu et al., 2006; Xu et al., 2008; Xu and Saiers, 2009). Based on results obtained using uniform sands, Xu et al. (2008) reported that both k0 and l varied linearly with the ratio of particle size and median sand grain size (dp/dg). Bradford et al. (2002,2003) performed experiments using well-sorted sands as well as sand mixtures under conditions unfavorable for colloid attachment to evaluate the kinetics of colloid straining within saturated porous media and observed a power law relationship between colloid straining rate coefficients and dp/dg. In this research, the best-fit values of k0 and l were calculated based on results obtained using both uniform and heterogeneous porous media under conditions that minimized physicochemical deposition. For the uniform sands, values of both k0 and l were found to be linearly dependent on the ratio of dp/dg (Fig. 5, closed circles). To evaluate the

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 9 6 e1 8 0 6

Fig. 5 e Relationships between the best-fit values of (A) k0 and (B) l and the ratio of dp/dg. For the sand mixtures, the sand grain diameter (dg) was calculated based on the number of sand grains in each size category. Error bars denote standard deviations for duplicate experiments.

relationship between k0 and l estimated from sand mixture experiments and the ratio of dp/dg, the average grain size of the sand mixtures was calculated based on either the mass mixing ratio or the number of sand grains from each size categories (Table 1). When the number-averaged grain size was used to represent the size of sand mixtures, there appeared to be good linear relationships between the values of k0 and l and the ratio of dp/dg (Fig. 5, open circles). In addition, when the values of k0 or l from both uniform sands and sand mixtures were pooled together, consistent relationships between either k0 and dp/dg or l and dp/dg were observed (Fig. 5). Approximately 92% and 83% of the variations in k0 and l could be explained by the ratio dp/dg. When the mass-averaged grain size was used to represent the size of the porous media, for any given value of dp/dg, values of k0 obtained using sand mixtures were higher than values of k0 estimated from uniform sand experiments (Fig. 6A). When results obtained using both uniform sands and sand mixtures were analyzed together, only 75% and 72% of the variation in observed values of k0 and l could be explained by the ratio of dp/dg (Fig. 6). It

1803

Fig. 6 e Relationships between the best-fit values of (A) k0 and (B) l and the ratio of dp/dg. For the sand mixtures, the sand grain diameter (dg) was calculated based on the mass of sands from each size category. Error bars denote standard deviations for duplicate experiments.

thus seemed that the straining of colloidal particles in heterogeneous porous media could be predicted from values of k0 or l obtained using uniform sands provided that the number-averaged sand grain size is used to represent the size of the sand mixture. Pore scale visualization experiments revealed that colloid straining occurred within selected pore spaces the sizes of which are smaller than the diameter of the particles (Xu et al., 2006). In heterogeneous porous media, finer sand grains that are appropriately mixed with sand grains of larger sizes can significantly reduce the size of the pore opening and create pore spaces that are capable of colloid straining (Fig. 7). As illustrated by Fig. 7, for the pore space created by 3 adjacent spherical sand grains (diameter is d1), the minimum size of colloidal particles that can be strained within this pore is 0.154d1 (Herzig et al., 1970). The presence of one finer sand grain with a diameter of d2 ¼ 0.154d1 can reduce the size of pore opening and colloidal particles with sizes  0.063d1 would be strained (Appendix A). The mass-averaged diameter of the 4 sand grains is 0.999d1, which was close to the size of

1804

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 9 6 e1 8 0 6

Fig. 7 e In this illustration, the size of the pore opening created by three adjacent sand grains (diameter [ d1) is significantly reduced by one finer sand grain (diameter [ d2 [ 0.154d1) (Herzig et al., 1970). Without the finer sand grain, the minimum size of colloidal particles that can be strained is 0.154d1. Under the presence of the finer sand grain, the minimum size of colloidal particles that can be strained is reduced to d3 [ 0.063d1. The massaveraged sand grain size is 0.999d1, while the number averaged sand grain size is 0.789d1.

the coarser sand grains and could not reflect the enhancement in colloid straining caused by the mixing of the finer sand grain. In comparison, the number-averaged size is 0.789d1, a value that was significantly lower than the massaverage size and could effectively reflect the increase in straining rate and capacity due to the block of pore openings by the finer sand grain. The analyses can be extended to 3 dimensions (Appendices B and C). Here we consider a tetrahedron formed by four spherical sand grains (diameter ¼ d1). The minimum size of particles that will be strained is defined by the opening between three adjacent sand grains and the calculation is similar to the 2-D case illustrated in Fig. 7. This value equals 0.154d1. As shown in Appendix B, a sand grain with a size of 0.225d1 could fill the pore space inside the tetrahedron. As a result the pore opening is substantially reduced and particles greater than 0.078d1 will be strained (Appendix C). The mass and number-averaged diameters of the 5 sand grains are 0.998d1 and 0.84d1, respectively. Considering the significant drop in the threshold particle size for straining, the number-average diameter is more appropriate to represent the “average” size of the porous media. Although natural sand grains are usually non-spherical and the sand packing can follow complex and heterogeneous patterns, the schematic illustration in Fig. 7 suggested that the number-averaged sand grain size is more appropriate to represent the important contributions from the finer sand grains.

The heterogeneous porous media used in this research were mixtures of well-sorted, uniform quartz sands. Although such sand treatments provided controlled systems that facilitated the study of colloid straining within heterogeneous porous media, it was also recognized that porous media found in the natural subsurface system often consist of sand grains the size of which may follow different distribution patterns (Chesnaux and Allen, 2008; Tillmann et al., 2008). Additional studies will be required to extend findings from this research to natural heterogeneous porous media, particularly those with high contents of clay particles, the size of which tends to be smaller than the finest sands used in this research. The sands used in this research were thoroughly cleaned to remove surface impurities which may promote the physicochemical filtration of the negatively-charged colloidal particles. As a result, the kinetics of colloid straining could be quantified and analyzed unambiguously as the contribution of physicochemical filtration on colloid immobilization was negligible. Natural soil and sediment, however, often are coated with materials such as iron oxide and iron hydroxide which, due to the positive charges they create, could facilitate the attachment of colloidal particles, which are often negatively charged, on the surface of the solid matrix (Abudalo et al., 2005). It is expected that within natural porous media, both colloid straining and physicochemical filtration will take place simultaneously. Little is known, however, about the interactions between colloid straining and physicochemical filtration under conditions that are favorable for colloid attachment onto the surface of sand grains. Further research that addresses such complexity will be needed to gain an improved understanding of the transport of pathogens and colloid-bound contaminants within the groundwater system.

4.

Conclusions

 Fine sands mixed in heterogeneous porous media could significantly enhance the straining of colloidal particles.  The mathematical model that was developed in (Xu et al., 2006) based on results obtained using uniform sands could be used to describe the kinetics of colloid straining within saturated heterogeneous porous media.  The analysis of the best-fit values of k0 or l and the ratio of dp/dg indicated that consistent linear relationships between k0 or l and the ratio of dp/dg for both uniform sands and heterogeneous sand mixtures could be achieved provided that the number-average size was used to represent the size of the sand mixtures.

Acknowledgement We are grateful to Dr. Heather Owen for her assistance with the scanning electron microscopy (SEM) and Dr. Steven Mylon of Lafayette College who measured the zeta potential values of the colloidal particles and sands. We also want to thank two anonymous reviewers whose comments led to improvements of our manuscript. This research was partially supported by National Science Foundation (CHE-0723002).

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 9 6 e1 8 0 6

Appendix. Supplementary data Supplementary data related to this article can be found online at doi:10.1016/j.watres.2010.11.037

references

Abudalo, R.A., Bogatsu, Y.G., Ryan, J.N., Harvey, R.W., Metge, D.W., Elimelech, M., 2005. Effect of ferric oxyhydroxide grain coatings on the transport of bacteriophage PRD1 and Cryptosporidium parvum oocysts in saturated porous media. Environmental Science & Technology 39, 6412e6419. Auset, M., Keller, A.A., 2004. Pore-scale processes that control dispersion of colloids in saturated porous media. Water Resources Research 40, W03503. Bradford, S.A., Bettahar, M., 2005. Straining, attachment, and detachment of Cryptosporidium oocysts in saturated porous media. Journal of Environmental Quality 34, 469e478. Bradford, S.A., Bettahar, M., Simunek, J., van Genuchten, M.T., 2004. Straining and attachment of colloids in physically heterogeneous porous media. Vadose Zone Journal 3, 384e394. Bradford, S.A., Simunek, J., Bettahar, M., Van Genuchten, M.T., Yates, S.R., 2003. Modeling colloid attachment, straining, and exclusion in saturated porous media. Environmental Science & Technology 37, 2242e2250. Bradford, S.A., Simunek, J., Bettahar, M., van Genuchten, M.T., Yates, S.R., 2006. Significance of straining in colloid deposition: evidence and implications. Water Resources Research 42, W12S15. doi:10.1029/2005WR004791. Bradford, S.A., Torkzaban, S., 2008. Colloid transport and retention in unsaturated porous media: a review of interface-, collector-, and pore-scale processes and models. Vadose Zone Journal 7, 667e681. Bradford, S.A., Torkzaban, S., Walker, S.L., 2007. Coupling of physical and chemical mechanisms of colloid straining in saturated porous media. Water Research 41, 3012e3024. Bradford, S.A., Yates, S.R., Bettahar, M., Simunek, J., 2002. Physical factors affecting the transport and fate of colloids in saturated porous media. Water Resources Research 38, 1327. doi:10. 1029/2002WR001340. Chesnaux, R., Allen, D.M., 2008. Simulating nitrate leaching profiles in a highly permeable vadose zone. Environmental Modeling & Assessment 13, 527e539. Chu, Y.J., Jin, Y., Baumann, T., Yates, M.V., 2003. Effect of soil properties on saturated and unsaturated virus transport through columns. Journal of Environmental Quality 32, 2017e2025. Crist, J.T., Zevi, Y., McCarthy, J.F., Throop, J.A., Steenhuis, T.S., 2005. Transport and retention mechanisms of colloids in partially saturated porous media. Vadose Zone Journal 4, 184e195. Dong, H.L., Onstott, T.C., Deflaun, M.F., Fuller, M.E., Scheibe, T.D., Streger, S.H., Rothmel, R.K., Mailloux, B.J., 2002. Relative dominance of physical versus chemical effects on the transport of adhesion-deficient bacteria in intact cores from South Oyster, Virginia. Environmental Science & Technology 36, 891e900. Elimelech, M., Gregory, J., Jia, X., Williams, R.A., 1998. Particle Deposition and Aggregation Measurement, Modelling, and Simulation. Butterworth-Heinemann, Oxford England. Foppen, J.W.A., Mporokoso, A., Schijven, J.F., 2005. Determining straining of Escherichia coli from breakthrough curves. Journal of Contaminant Hydrology 76, 191e210. Fuller, M.E., Dong, H.L., Mailloux, B.J., Onstott, T.C., DeFlaun, M.F., 2000. Examining bacterial transport in intact cores from

1805

Oyster, Virginia: effect of sedimentary facies type on bacterial breakthrough and retention. Water Resources Research 36, 2417e2431. Gregory, J., 1981. Approximate expressions for retarded van der Waals interaction. Journal of Colloid and Interface Science 83, 138e145. Harvey, R.W., Kinner, N.E., Macdonald, D., Metge, D.W., Bunn, A., 1993. Role of physical heterogeneity in the interpretation of small-scale laboratory and field observations of bacteria, microbial-sized microsphere, and bromide transport through aquifer sediments. Water Resources Research 29, 2713e2721. Havelaar, A.H., Vanolphen, M., Schijven, J.F., 1995. Removal and inactivation of viruses by drinking-water treatment processes under full-scale conditions. Water Science and Technology 31, 55e62. Herzig, J.P., Leclerc, D.M., Legoff, P., 1970. Flow of suspensions through porous media e application to deep filtration. Industrial and Engineering Chemistry 62, 8e35. Hogg, R., Healy, T.W., Fuersten, Dw, 1966. Mutual coagulation of colloidal dispersions. Transactions of the Faraday Society 62, 1638e1651. Hutson, S.S., Barber, N.L., Kenny, J.F., Linsey, K.S., Lumia, D.S., Maupin, M.A., 2004. Estimated Use of Water in the United States in 2000, U.S. Geological Survey Circular 1268. US Geological Survey, Reston, VA, p. 46. Keller, A.A., Auset, M., 2007. A review of visualization techniques of biocolloid transport processes at the pore scale under saturated and unsaturated conditions. Advances in Water Resources 30, 1392e1407. Li, X., Zhang, P., Lin, C.L., Johnson, W.P., 2005. Role of hydrodynamic drag on microsphere deposition and reentrainment in porous media under unfavorable conditions. Environmental Science & Technology 39, 4012e4020. Li, X.Q., Lin, C.L., Miller, J.D., Johnson, W.P., 2006. Role of grain-tograin contacts on profiles of retained colloids in porous media in the presence of an energy barrier to deposition. Environmental Science & Technology 40, 3769e3774. Liu, Y.Y., Janjaroen, D., Kuhlenschmidt, M.S., Kuhlenschmidt, T.B., Nguyen, T.H., 2009. Deposition of Cryptosporidium parvum Oocysts on natural organic matter surfaces: microscopic evidence for secondary minimum deposition in a radial stagnation point flow cell. Langmuir 25, 1594e1605. Logan, B.E., Jewett, D.G., Arnold, R.G., Bouwer, E.J., Omelia, C.R., 1995. Clarification of clean bed filtration models. Journal of Environmental Engineering-ASCE 121, 869e873. Mccarthy, J.F., Zachara, J.M., 1989. Subsurface transport of contaminants e mobile colloids in the subsurface environment may alter the transport of contaminants. Environmental Science & Technology 23, 496e502. McDowell-Boyer, L.M., Hunt, J.R., Sitar, N., 1986. Particle-transport through porous-media. Water Resources Research 22, 1901e1921. Ryan, J.N., Elimelech, M., 1996. Colloid mobilization and transport in groundwater. Colloids and Surfaces A-Physicochemical and Engineering Aspects 107, 1e56. Saiers, J.E., 2002. Laboratory observations and mathematical modeling of colloid-facilitated contaminant transport in chemically heterogeneous systems. Water Resources Research 38, 1032. Saiers, J.E., Hornberger, G.M., 1996. The role of colloidal kaolinite in the transport of cesium through laboratory sand columns. Water Resources Research 32, 33e41. Saiers, J.E., Hornberger, G.M., Harvey, C., 1994. Colloidal silica transport through structured, heterogeneous porous media. Journal of Hydrology 163, 271e288. Shen, C., Huang, Y., Li, B., Jin, Y., 2008. Effects of solution chemistry on straining of colloids in porous media under unfavorable conditions. Water Resources Research 44, W05419.

1806

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 9 6 e1 8 0 6

The Mathworks, 2006. Optimization Toolbox for Use with Matlab. MathWorks, Inc., Natick, Mass. Tillmann, A., Englert, A., Nyari, Z., Fejes, I., Vanderborght, J., Vereecken, H., 2008. Characterization of subsoil heterogeneity, estimation of grain size distribution and hydraulic conductivity at the Krauthausen test site using cone penetration test. Journal of Contaminant Hydrology 95, 57e75. Torkzaban, S., Tazehkand, S.S., Walker, S.L., Bradford, S.A., 2008. Transport and fate of bacteria in porous media: coupled effects of chemical conditions and pore space geometry. Water Resources Research 44, W04403. Tufenkji, N., Miller, G.F., Ryan, J.N., Harvey, R.W., Elimelech, M., 2004. Transport of Cryptosporidium oocysts in porous media: role of straining and physicochemical filtration. Environmental Science & Technology 38, 5932e5938.

Tufenkji, N., Ryan, J.N., Elimelech, M., 2002. The promise of bank filtration. Environmental Science & Technology 36, 422ae428a. Xu, S.P., Gao, B., Saiers, J.E., 2006. Straining of colloidal particles in saturated porous media. Water Resources Research 42, W12S16. doi:10.1029/2006WR004948. Xu, S.P., Liao, Q., Saiers, J.E., 2008. Straining of nonspherical colloids in saturated porous media. Environmental Science & Technology 42, 771e778. Xu, S.P., Saiers, J.E., 2009. Colloid straining within watersaturated porous media: effects of colloid size nonuniformity. Water Resources Research 45, W05501. doi:10.1029/ 02008WR007258. Yao, K.M., Habibian, M.M., Omelia, C.R., 1971. Water and waste water filtration e concepts and applications. Environmental Science & Technology 5, 1105e1112.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 0 7 e1 8 1 7

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Radial transport processes as a precursor to particle deposition in drinking water distribution systems P. van Thienen a,*, J.H.G. Vreeburg a,b, E.J.M. Blokker a,b a b

KWR Watercycle Research Institute, Groningenhaven 7, 3433 PE Nieuwegein, The Netherlands Delft University of Technology, Stevinweg 1, 2628 CN Delft, The Netherlands

article info

abstract

Article history:

Various particle transport mechanisms play a role in the build-up of discoloration potential

Received 5 July 2010

in drinking water distribution networks. In order to enhance our understanding of and

Received in revised form

ability to predict this build-up, it is essential to recognize and understand their role.

19 October 2010

Gravitational settling with drag has primarily been considered in this context. However,

Accepted 24 November 2010

since flow in water distribution pipes is nearly always in the turbulent regime, turbulent

Available online 4 December 2010

processes should be considered also. In addition to these, single particle effects and forces may affect radial particle transport.

Keywords:

In this work, we present an application of a previously published turbulent particle

Discoloration

deposition theory to conditions relevant for drinking water distribution systems. We

Sediment settling

predict quantitatively under which conditions turbophoresis, including the virtual mass

Turbophoresis

effect, the Saffman lift force, and the Magnus force may contribute significantly to sedi-

Turbulent diffusion

ment transport in radial direction and compare these results to experimental observations.

Saffman lift force

The contribution of turbophoresis is mostly limited to large particles (>50 mm) in transport

Magnus force

mains, and not expected to play a major role in distribution mains. The Saffman lift force

Virtual mass effect

may enhance this process to some degree. The Magnus force is not expected to play any

Drinking water distribution

significant role in drinking water distribution systems. ª 2010 Elsevier Ltd. All rights reserved.

1.

Introduction

Sampling campaigns show a pattern of accumulation of sediments in drinking water distribution systems that is difficult to explain by invoking conventional theories of gravitational settling only Blokker et al. (2010a), as will be discussed below. One of the possible explanations invoked by these authors is an additional deposition mechanism. Therefore, a more detailed evaluation of sedimentation processes in the context of water distribution systems seems necessary. Laboratory experiments in smooth walled pipes have shown that, for conditions comparable to those found in

common drinking water distribution systems, suspended particles may exhibit two distinct modes of sedimentation in horizontal pipes, either settling at the bottom half of the pipe wall, or settling on the complete circumference of the pipe wall, including the upper half (Vreeburg and Boxall, 2007). The former mode can easily be attributed to gravitational settling, and is included as such in models for particle settling (e.g. Ryan et al., 2008). The latter mode has been qualitatively attributed to the turbulent process of turbophoresis (Vreeburg and Boxall, 2007). Scaling analysis shows that it is correct to assume turbulent processes to occur under almost all conditions prevailing in drinking water distribution systems (for a typical 100 mm pipe, flow becomes turbulent at a flow

* Corresponding author. E-mail addresses: [email protected] (P. van Thienen), [email protected] (J.H.G. Vreeburg), mirjam.blokker@ kwrwater.nl (E.J.M. Blokker). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.11.034

1808

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 0 7 e1 8 1 7

velocity of 4 cm/s; for larger pipes, this minimum velocity is even lower). We can therefore expect turbulent diffusion (also included in some form in the aforementioned model of Ryan et al., 2008) to play a role as well. Both turbophoresis and turbulent diffusion are illustrated in Fig. 1. Turbulent diffusion of particles can be seen as the macroscopic result on a particle population of the separation of individual particles by the eddies of turbulent flow (Davidson, 2004). Even though arbitrary individual sets of particles may be alternatingly moving towards each other and away from each other in an erratic way (see Fig. 1a), the net effect is an increase in separation. The process results in a reduction of particle concentration gradients. Turbulent diffusion is important when particle inertia is negligible compared to the viscous forces dragging the particle along with the flow. As a result, the particles closely follow the actual flow pattern. Turbophoresis is the transport of particles down a gradient of turbulence intensity due to particle inertia (Caporaloni et al., 1975; Guha, 1997). The particles do not follow the flow pattern but move to some degree, but not completely, in an independent way. Local velocity variations in the turbulent flow field may add to the momentum of particles. In highly turbulent regions, the larger local flow velocity variations may propel a particle into a less turbulent region. The jump of a particle in the opposite direction, from a less turbulent to a more turbulent region, is less likely, however, because of the smaller local velocity variations. As a result, a net transport of particles away from the highly turbulent region occurs (Young and Leeming, 1997). The Basset history force (Zhu et al., 2007) may be important (more so than viscous drag) in aerodynamic shock regions (Thomas, 1992), which show near-discontinuous changes in the characteristics of the medium, but can be ignored in the present application. The virtual mass effect is caused by the boundary layer surrounding the particle acting to some degree as part of the particle in terms of its inertia. Its magnitude is roughly equal to half the volume of the particle multiplied by the density of

a

Table 1 e Directions of Magnus force Fm and Saffman lift force Fs as a function of linear and angular velocities of a particle relative to the surrounding fluid. Directions are based on evaluation of the vectorial forms of the relevant equations as presented in Zhu et al. (2007). Relative linear velocity of particle

Relative angular velocity of particle

Slow Fast

Slow

Fast

Center Wall Center Center

Wall Center Wall Wall

Force

Fs Fm Fs Fm

the fluid in which it is suspended and the relative acceleration of the particle in the fluid (Odar and Hamilton, 1964; Zhu et al., 2007). For particles suspended in air, this effect is often very small due to the low density of the medium. Therefore, this effect is generally not considered in such applications. In the case of water, however, the effect is much larger, adding about 40% of virtual mass to iron hydroxide flocs (slightly heavier than water, assumed density 1200 kg/m3) and about 20% to sand particles (density 2500 kg/m3). The Magnus force results from rotation of a particle (Magnus, 1853; Zhu et al., 2007) relative to the rotation of the fluid. For a particle which is moving slower than the surrounding fluid and rotates ’towards’ the wall of the pipe at a rate slower than the local fluid rotation rate (curl of the fluid velocity field), the Magnus force is directed towards the wall of the pipe (see Table 1). A rotation rate faster than the local fluid rotation rate or a velocity greater than the local fluid velocity directs the Magnus force towards the center of the pipe. When both the rotation rate and the local velocity are greater than those of the local fluid, the Magnus force points towards the wall. Clearly, it is difficult to evaluate the effect of this force on particle deposition without looking into exact movements of individual particles. However, shear induced by decreasing fluid velocity towards the wall of the pipe may be expected to result in some particle rotation towards the wall, but at a rate

b

Fig. 1 e Simplified representation of the processes of turbulent diffusion (a) and turbophoresis (b). The curved arrows indicate vortices in the flow field. In frame a), a pair of particles (two red dots connected by a dashed red line) which are initially close together (left side) is followed in time (movement towards right side, trajectories indicated in green), showing an increase in the distance between them (dashed line). In frame b), straight arrows represent instantaneous velocity vectors of individual particles. For an explanation of the processes depicted here, see the main text. Note that the static representation of the flow field shown in this figure does not do justice to the actual field which is constantly varying with time. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article).

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 0 7 e1 8 1 7

which is at most equal to the local fluid rotation rate. Similarly, the particle may be expected to move with a velocity less than or equal to the local fluid velocity. As a result, we would expect the Magnus force to be directed towards the wall of the pipe. The Saffman force (Saffman, 1965, 1968) results from shear when the particle and the surrounding fluid are not moving at the same velocity, and when the fluid has a velocity gradient perpendicular to the direction of motion of the particle. When the particle is moving faster than the fluid, the Saffman force is directed towards the wall (see Table 1). When the particle is slower, the Saffman force pushes it towards the center of the pipe. It is clear that this effect will play a role only when particle inertia is sufficiently large to allow significant differential velocities. Turbophoresis may act to provide this differential velocity, causing a significant enhancement of the particle transport towards the wall (Young and Leeming, 1997). Note that rotation of the particle also induces shear in the surrounding fluid and may also result in a lift force (Zhu et al., 2007). The pressure gradient force results from different pressures on either side of a particle in a pressure driven flow. In drinking water distribution systems, the magnitude of this force is quite small compared to other forces, and directed along the pipe axis. As such, this effect does not contribute to radial transport of particles. The effect of turbulent and other processes on the transport of suspended particles has been widely studied in a number of more or less related fields, including theoretical and numerical considerations of particle-laden gas streams (e.g. Guha, 1997, 2008; Young and Leeming, 1997; Soldati and Marchioli, 2009) industrial applications of particle-laden gas streams (e.g. Fokeer et al., 2004; Hadinoto et al., 2005) air conditioning ducts (e.g. Sippola and Nazaroff, 2002) the transport of wind-borne and river sediments (e.g. Caporaloni et al., 1975; Zou et al., 2007; Liu and Yin, 2010) and studies of particle behavior in sewer systems (e.g. Verbanck, 2000). In this work, we aim to determine the range of conditions relevant for drinking water distribution in which different processes can be expected to operate. Our main concern is the effect of these processes on the discoloration potential of water. We present theories which have been developed elsewhere, in the context of particles suspended in air, to the water community and include the virtual mass effect (Odar and Hamilton, 1964) which can be neglected in air flow but not in water. Both the applicability and the relevance of the theory are demonstrated, linking the results of computations to experimental observations.

2.

Background

2.1.

Turbidity measurements

Blokker et al. (2010a) describe turbidity measurements obtained during flushing activities in a Dutch drinking water distribution system in the town of Purmerend, following a carefully designed flushing program. This program ensured that a clear water front was used (i.e. the water used for flushing was transported through pipes which had already

1809

been flushed themselves; therefore, the flushing water should only carry sediment originating from the section which is being flushed), a flushing velocity of 1.5 m/s or more was reached and each pipe volume was refreshed two to three times or until the turbidity was below 0.6 FTU. The turbidity was measured every 5 s in a diverted fraction of the flushing water using a Dr Lange Ultraturb SC100. In case the pipe was clean after only one turnover (i.e. the total turbidity during the first turnover was more than 80% of the total turbidity) the turbidity was interpreted as locally accumulated particles that were resuspended during flushing and transported at the flushing flow rate to the flushing point (in this case a hydrant). By taking the flushing flow velocity into account it was possible to back-trace the location in the pipe where the particles had originated from. A detailed EPANET network model in combination with a set of stochastic water demand patterns from SIMDEUM (Blokker et al., 2010b) was used by Blokker et al. (2010a) to determine the maximum velocities in the pipes under normal operating conditions. Fig. 2 (modified from Blokker et al., 2010a) shows the locally accumulated sediment which was brought into suspension (in FTU) as a function of the locally occurring maximum velocities (m/s). The figure shows particle accumulation does not occur at maximum velocities above 0.25 m/s; at lower velocities, particle accumulation may or may not occur. This means that the maximum velocity is not the only determining factor for particle accumulation. The fact that there is a peak in particle accumulation at maximum velocities of 0.05e0.1 m/s suggests that velocity does play a role in particle accumulation, but additional factors are required to explain the observations.

2.2. Characteristics of particles in drinking water distribution systems Particles occur in many sizes and types in the drinking water distribution system. For a discussion of their origins, we refer to Vreeburg and Boxall (2007). Here, we limit ourselves to providing an overview of their characteristics. The most

Fig. 2 e Turbidity measurements during flushing operations in Purmerend, the Netherlands, as a function of the maximum velocity occurring under normal operation in the corresponding distribution network section. Modified from Blokker et al. (2010a).

1810

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 0 7 e1 8 1 7

relevant characteristics for the hydraulic behavior of suspended particles are their size and density. Some particle size distributions obtained from measurements during flushing actions have been published in the literature. Fig. 3 gives an overview of measurements from the UK (Boxall et al., 2001) and The Netherlands (Vreeburg, 2007, and additional measurements performed at KWR). Whereas the Dutch measurements all show a predominance of small particles less than 10 mm in diameter (Vreeburg, 2007, writes that most particles are between 3 and 12 mm), the British data show a broad peak roughly between 5 and 30 mm. Explanations for this difference may be found in differences in water sources (predominantly ground water with fewer particles for the Dutch situation versus predominantly surface water in the UK) and treatment processes. Boxall et al. (2001), report an average density of the particles of 1000e1300 kg/m3. Fractions smaller than those listed above must also be present in drinking water, but the behaviour of these so called colloids is ruled by different effects and forces than that of larger particles (Stumm, 1992) and not treated in this work.

A relation between the bulk flow velocity Ub (m/s) and the friction velocity us (m/s) based on the classical logarithmic velocity profiles (Von Ka´rma´n, 1930) for pipe flows can be obtained (Zanoun et al., 2007):   Ub 1 Rus 3  þB ¼ ln us n k 2k

with k the dimensionless Von Ka´rma´n constant, R the pipe inner radius (m), n the kinematic fluid viscosity (m2/s), and B a dimensionless additive constant. Although different researchers report somewhat different values for both k and B (Zanoun et al., 2003), we apply the standard literature values of 0.4 and 5.5, respectively, and note that the calculations described below exhibit only a small sensitivity to the values of these parameters within the ranges found in the literature. The particle relaxation time represents the ratio between inertia and drag forces and as such indicates how long it takes a particle to adjust its trajectory to changing flow conditions in its immediate vicinity. It can be computed following Young and Leeming (1997): sp ¼

3.

Methods

3.1. Turbophoresis and turbulent diffusion with virtual mass effect A simple force balance and scaling are used to estimate under which conditions particle transport by turbophoresis dominates that by turbulent diffusion and vice versa. Several authors have presented theories of the transport and deposition behavior of particles suspended in air (Young and Leeming, 1997; Guha, 1997, 2008), which can be used as a starting point for describing the behavior of particles suspended in water. Some of the basic considerations are discussed here.

(1)

rp d2p

ð1 þ 2:7KnÞ 18rw n

(2)

In this expression, rp is the particle material density (kg/m3), dp is the particle diameter (m), rw is the fluid density (kg/m3), and Kn the Knudsen number, which is the ratio of the molecular mean free path to the particle size. Considering this parameter is more meaningful in a gas than in liquid water, since the water molecules are bound to each other through hydrogen bonds, which have a typical length on the order of 1010 m, which is much smaller than anything that could reasonably be called a particle in the sense used here. Therefore, the last factor of expression (2) can be dropped. Contrary to the situation where particles are suspended in air, water-borne particles experience a virtual mass or added mass effect of significant magnitude. This effect results from the fluid boundary layer acting as part of the particle in terms of its mass and inertia. In order to include this effect, the corresponding fluid mass of half the volume of the sphere (Odar and Hamilton, 1964) needs to be added:

sp ¼

  2rp þ rw d2p 36rw n

(3)

The dimensional particle relaxation time is usually nondimensionalized using the friction velocity us and the kinematic fluid viscosity n (Guha, 2008): spþ ¼ sp u2s =n

Fig. 3 e Particle size distributions obtained during flushing operations in the UK (Boxall et al., 2001) and The Netherlands Vreeburg, 2007, and additional measurements performed at KWR). The UK data represent an average of material from different distribution networks. The Dutch data include material from distribution networks in Franeker, Den Bosch and Rosmalen and transport mains from Nuland pumping station to Rosmalen and Den Bosch.

(4)

In the behavior of suspended particles in a fluid, three different regimes can be distinguished (Guha, 1997, 2008; Young and Leeming, 1997). The diffusional deposition regime occurs at low spþ (<0.2) and is characterized by turbulent and molecular diffusion processes. At slightly higher spþ, between 0.2 and 20, the diffusion-impaction regime shows a decline in the importance of diffusion processes and an increase in the dominance of turbophoresis. For spþ>20, turbophoresis is the dominant process in the inertia-moderated regime. It is important to note that the transport of particles towards the wall by turbophoresis does not necessarily take

1811

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 0 7 e1 8 1 7

þ

y ¼ yus =n

(5)

þ

with y and y the non-dimensional and dimensioned (m) distance from the wall, respectively.

3.2.

Gravity, the Saffman lift force and the Magnus force

a 0.001

III particle diameter (m)

place everywhere, but is generally limited to the zone adjacent to the wall. This zone has a non-dimensional thickness of 40 (Young and Leeming, 1997), in which the scaling is performed according to

0.0001

II

O I

1e-05

The gravitational force experienced by a particle is a function of its size and relative density:

with g the gravitational acceleration (m/s2). The magnitude of the Saffman lift force scales linearly with the differential velocity Δv between the particle and the surrounding fluid (Saffman, 1965; Zhu et al., 2007; Zou et al., 2007):  12 1 vu Fs ¼ 1:62ðmrw Þ2 d2p Dv vy

(7)

Note that the corresponding expression provided by Zou et al. (2007) appears to contain an error, the power 1/2 having been replaced by 2. Under conditions for which the particles closely follow the flow pattern of the fluid medium (zone I in Fig. 4a), the differential velocity is near zero and therefore the Saffman lift force is expected to be insignificant. The Saffman lift force equations are relevant when all of the following conditions are met (Saffman, 1965): 1

3

1

1e-06

(6)

22 ðr=RÞ2 Re2 [1

(8)

 1 1 R 2 1 Rr 1 1  R 22 r Re2

(9)

rp  r 12 1 Re2  1 R R

(10)

b

pipe diameter (m)

  1 Fg ¼ pd3p rp  rw g 6

3

1.2 g/cm 2.5 g/cm3 0.001

0.01

0.1

1

bulk flow velocity (m/s)

10

100

1

0.001

0.1

0.01 0.1 1 10

0.01

0.001

0.01

0.1

1

10

bulk flow velocity (m/s)

In these expressions, r represents the distance of the center of a particle from the center of the pipe (m). The first condition formulates the requirement that the curvature of the velocity profile is small. The second condition implies that the particle is not too close to the wall. The last condition is that the particle Reynolds number is small. For more details about these requirements, the reader is referred to Saffman (1965). It can easily be verified that generally the hydraulic conditions in drinking water distribution systems meet these requirements. The magnitude of the Magnus force, like the Saffman lift force, is a linear function of the differential velocity between the particle and the surrounding fluid Δv (Rubinow and Keller, 1961; White and Schulz, 1977; Zou et al., 2007):   p 1 vu (11) Fm ¼ rw d3p Dv u  8 2 vy

Fig. 4 e a) Radial particle transport behavior regimes as a function of bulk flow velocity Ub and particle diameter dm for a pipe diameter of 1 m. Domain boundaries are indicated for two different particle material densities, 2.5 g/ cm3 (solid black lines), representing sand particles, and 1.2 g/cm3 (dashed black lines), corresponding to iron hydroxide flocs. Boundaries are shown between conditions where the applied formulation breaks down because of low Re (O), the diffusional deposition regime (I), the diffusion-impaction regime (II), and the inertiadominated regime (III), following Young and Leeming (1997). Note that the positions of the boundaries are quite insensitive to the pipe diameter for practical values, except for the boundary between laminar (O) and turbulent (I,II,III) flow, which moves linearly with the pipe diameter. b) Contours of the thickness of the layer in which turbophoresis drives particles towards the wall ( yD £ 40), normalized by the pipe radius.

with u the angular velocity (rad/s). Again, under conditions for which particles closely follow the fluid flow, no significant Magnus force magnitude is to be expected. As stated above, we expect the Magnus force to be directed towards the wall of the pipe because of the expected velocity and

rotation rate of the particles. The Saffman lift force is directed either towards the center (for particles which are slower than the fluid) or towards the wall (fast particles) of the pipe. For particles which have non-negligible inertia, any radial movement caused

1

22

1812

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 0 7 e1 8 1 7

by a different mechanism, e.g. turbophoresis, will result in the particle (temporarily) going faster (when moving towards the wall) or slower (when moving towards the center) than the fluid (see Fig. 5b,c). In addition to this, the particle rotation rate may also become greater than the local fluid rotation rate by radial movement towards the center of the pipe or smaller by radial movement in the opposite direction (see Fig. 5d,e). Thus a slow particle with an angular velocity consistent with the local fluid flow field will be driven towards the center

of the pipe by the Saffman lift force. As it moves away from its initial radial position, its angular velocity becomes greater than the local rotation rate of the fluid, and the Magnus force helps to drive the particle towards the center of the pipe. For a fast particle with an angular velocity consistent with the local fluid flow field, the Saffman lift force will push the particle towards the wall of the pipe. Here the particle will have a slower angular velocity than the local fluid rotation rate, and as a result experience an opposing Magnus force. Which of the two forces will prevail depends on the local conditions. Equating expressions (7) and (11) and assuming a logarithmic velocity profile (e.g. Zanoun et al., 2007) v ¼ us

  1 yus  ln þB n k

(12)

renders the boundary curve between both domains in parameter space: u¼

  12  us us 1:62m þ 1 ky 2ky pr2w dp 8

(13)

This shows that the balance between the two forces depends on the fluid viscosity m and density rw, the particle diameter dp and rotation rate u, the bulk flow velocity u0 through the friction velocity us, and the radial location y of the particle. There is no dependence on the differential velocity between the particle and the fluid. An estimate for the particle rotation rate can be made if we assume that all particle rotation is due to velocity gradients in the fluid, see Fig. 5a: u¼

DV r

(14)

By applying the logarithmic velocity profile of expression (12), the angular velocity can be predicted as a function of the particle radius, the radial location in the pipe, and the friction velocity:

u¼ Fig. 5 e Pipe flow of suspended particles with a typical time-averaged velocity profile and velocity gradient profile (horizontal axis is position in pipe, vertical axis is local flow velocity (bottom part) or radial gradient of lateral flow velocity (top part)). a) Induced rotation of a particle due to the fluid velocity gradient. b) Movement of a particle with non-negligible inertia by some process towards the wall of the pipe results in a linear velocity surplus. c) Movement of a particle with non-negligible inertia towards the center of the pipe results in a linear velocity deficit of the particle. d) Movement of a particle with non-negligible moment of inertia towards the center of the pipe results in a angular velocity surplus of the particle. e) Movement of a particle with non-negligible moment of inertia towards the wall of the pipe results in a angular velocity deficit of the particle. Note that red dotted arrows represent particle rotation, whereas green dotted arrows indicate local fluid rotation. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article).

us ln





yþr yr

kr

4.

Results

4.1.

Results of calculations for stochastic processes

(15)

Using a bisection algorithm, the set of Eqs. (1)e(4) has been solved for spþ in order to determine the ranges of conditions in which the different regimes are applicable in drinking water distribution systems. Fig. 4(a) shows particle transport behavior regimes as a function of bulk flow velocity Ub and particle diameter dp for different particle material densities. Although it was produced for a pipe diameter of 1 m, we have verified that it is only slightly different for cm-scale pipes, so the figure can be applied to the entire practical range of pipe diameters in drinking water distribution systems. Note that the position of the laminareturbulent transition does move linearly with the pipe diameter. The figure demonstrates that diffusional processes are dominant for smaller particles and lower flow

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 0 7 e1 8 1 7

velocities. In these cases, turbulent diffusion is many orders of magnitude more important than molecular diffusion. Only for relatively large particles at high flow rates can we expect turbophoresis to play a significant role in radial particle transport. The part of the pipe cross-section in which turbophoresis actually causes outward radial movement of particles is shown in Fig. 4(b) as a fraction of the pipe radius, from Eqs. (1) and (5). For small pipes and low flow velocities, this zone encompasses the entire cross section of the pipe, whereas for larger pipes and higher flow velocities, only a narrow zone adjacent to the pipe wall shows this behavior. In terms of particle movement, this means that we can expect a uniform particle concentration throughout a cross section of the pipe for smaller particles and lower bulk velocities below the boundary between regions I and II in Fig. 4, since (turbulent) diffusion tends to even out any concentration fluctuations. Gravitational settling may slowly move the particles towards the bottom of the pipe, where they may be deposited, unless the diffusion process is sufficiently active to counter this effect (see Boxall et al., 2001). On the other hand, for larger particles and higher flow velocities above the boundary between regions II and III in Fig. 4, turbophoresis drives particles towards the pipe wall, where particles may be deposited if this effect is stronger than gravity, allowing near-uniform deposition of particles everywhere on the pipe wall, including the top. For these conditions, radial particle transport by turbophoresis is only significant in the outermost region of the water inside the pipe (see Fig. 4b), causing an increase in particle concentration at the wall and a decrease at a small distance from the wall. However, turbulent diffusion will replenish this zone with particles from the interior because of the concentration gradient induced by turbophoresis, and thus contributes to the transport of particles from the central flow zone to the wall. Using a parameterization of the particle deposition velocity due to turbulent diffusion and turbophoresis as a function of the non-dimensional relaxation time spþ based on laboratory measurements such as shown in Fig. 1 of Guha (2008), the dimensional particle deposition velocity can be plotted as a function of bulk flow velocity and particle size. This has been done in Fig. 6. Note that these velocities are derived from fluxes and are thus only meaningful for a population of particles and not for individual particles. Deposition velocities range from 107 m/s for 100 mm particles in a 0.1 m/s flow to 102 m/s for the same particles in a 3 m/s flow. If we consider a 1 m diameter pipe, it would take a particle about 58 days to move from the center of the pipe to the wall in the first case, during which period the water would have traveled 500 km. In the second case, however, it would only take the particle 50 s to cross the radial distance, travelling 150 m in lateral direction in the mean time.

4.2.

Results of calculations for deterministic processes

The balance between the Magnus force and the Saffman lift force is illustrated in Fig. 7, based on expression (13). This clearly shows that small particles require a very high angular velocity for the Magnus force to become important. For larger particles at lower bulk flow velocities, the Magnus force dominates even at very small rotation rates. Following through on the ideas depicted in Fig. 5, we have calculated the effect of an instantaneous displacement of

1813

Fig. 6 e Deposition velocity due to turbophoresis and turbulent diffusion as a function of bulk flow velocity and particle diameter in a pipe of 1 m diameter. Boundaries of the transport mechanism domains of Fig. 4 are indicated by dashed yellow lines. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article).

a particle some distance towards the wall of the pipe. It is assumed (though not relevant) that this displacement is the result of turbophoresis. As has been demonstrated in this figure, this displacement would result in a linear velocity surplus relative to the surrounding fluid and an angular velocity deficit, if we assume the particle inertia to be high enough. Fig. 8 shows which of the forces discussed above would have the highest magnitude under which conditions (bulk flow velocity and particle size). Frames aed show this for two different pipe diameters (0.1 m and 1 m) and two amounts of displacement (1% and 5% of the pipe radius). The results were calculated at a distance of 25% of the pipe radius from the wall, but we have verified that they are quite similar for other radial positions. All frames show a large domain where the gravitational pull is stronger than both the Saffman lift force and the Magnus force, specifically for low flow velocities and larger (¼heavier) particles. For higher flow velocities, the Saffman lift force becomes stronger than the gravitational force. At very high bulk flow velocities beyond those which could be reasonably expected in drinking water distribution systems, the Magnus force may start to dominate the force balance of very large particles, but it appears that for our application and first-order consideration, this effect can be safely ignored. Note that the operation of the Saffman lift force really depends on the presence of a velocity difference between the particle and the fluid. Frame e of Fig. 8 demonstrates the absence of any Saffman effect when a particle is not displaced first by some other mechanism like turbophoresis.

4.3. Combined effect of deterministic and stochastic processes Fig. 8f shows the combined effect of the stochastic processes associated with turbulence (Section 3.1) and the deterministic

1814

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 0 7 e1 8 1 7

dp=1e-3 m 10000

1000

1000

100

100

10

10

1

1

0,1

0,1

0,01 0,001

0,01

0,1

1

10

100

0,01 0,001

10000

1000

1000

100

100

10

10

1

1

0,1

0,1

0,01 0,001

0,01

0,1

1

10

100

0,01 0,001

0,01

0,1

1

10

100

0,01

0,1

1

10

100

D=1.0 m

10000

y=0.10*R y=0.25*R y=0.50*R y=0.75*R

D=0.1 m

log(angular velocity)

dp=1e-6 m 10000

log(bulk flow velocity) Fig. 7 e Boundaries between domains where the Magnus force is stronger than the Saffman lift force (above curves) and domains where the Saffman lift force prevails. On the horizontal axes, the log10 of the bulk flow velocity is plotted; the vertical axes show the log10 of the angular velocity. Curves are plotted for different distances from the pipe wall, for two pipe diameters and two grain sizes spanning the ranges of interest.

forces on individual particles (Sections 3.2 and 4.2). This figure has been constructed from a phenomenological point of view, illustrating the expected observable effects on the particle population. Parameter space, spanned by the bulk flow velocity of the fluid and the particle diameter, is divided into four domains. The first, labelled DG, corresponds to conditions where gravity and/or turbulent diffusion determine the radial transport behavior of particles. In the second domain, labelled T, turbophoresis dominates radial particle transport. Conditions under which the Saffman lift force is sufficiently strong, radial displacement of particles by turbophoresis may be enhanced by this force. This is the case in the domain labelled Tþ. Note that also in domain T, the Saffman lift force may enhance the radial particle transport to some degree. However, as it is weaker than gravity here, it will not contribute to radial particle transport 360 around. Very large particles (larger than 300 mm) may be pushed towards the center of the pipe by the Magnus force at bulk flow velocities above a few meters per second. This high minimum velocity makes the process less relevant for drinking water distribution. Under these conditions, the process will be in competition with turbophoresis, which is vigorously (see Fig. 6) transporting particles towards the wall of the pipe.

5.

Discussion

5.1.

Experimental observations

Vreeburg and Boxall (2007) presented results of an experimental setup in which a ferric chloride solution was recirculated

through a perspex pipe loop for four days. Fig. 9 shows results of a repeat of their original experiment, using 20 mg/l ferric chloride. The results of the repeat experiment mirror those of the original. At low flow velocity (0.06 m/s), we observe deposition of iron hydroxide flocs on the bottom half of the pipe inner surface exclusively. However, for a higher flow velocity of 0.14 m/s, deposition on the entire wall circumference is observed. These results can now be interpreted quantitatively in the theoretical framework described above. The flocs in the experiment described above were visible to the naked eye and their size was estimated to be in the range of 100e1000 mm. Although this is on the large end of the spectrum for typical particles in drinking water distribution systems, this does provide us with relevant observations for pinpointing the position of the observed transition in parameter space. Together with the two flow velocities, the theoretical turbulent transport process domain can be found for these two experiments in Fig. 10. From the observations, the first experiment is expected to plot in the diffusion-dominated regime (I), with no particle settling on the top half of the pipe wall. Because it shows particle deposition everywhere on the pipe wall, the second experiment is expected to plot in the transitional (II) or inertia-dominated (III) regime. The large uncertainty in particle size does not allow us to put either experiment in the turbulent diffusion (I) or transitional (II) regime without reservation. However, the observed transition, somewhere between the vertical red lines ( and ) in Fig. 10, is close to the predicted one in any case, showing consistency between theory and observations. More data points are required to further verify the theoretical predictions. Therefore, additional experiments are

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 0 7 e1 8 1 7

1815

Fig. 8 e Comparison of the magnitudes of gravitational pull, Saffman lift force, and Magnus force on particles as a function of bulk flow velocity (m/s) and particle diameter (m) for particles which have been displaced towards the wall, presumably by turbophoresis, but retained the linear and angular velocity corresponding to their original location due to inertia, see Fig. 5. Labels Fg, Fs, Fm indicate where the magnitude of gravitational pull, Saffman lift force, and Magnus force, respectively, is higher than the other two. a) Pipe diameter is 1 m; the particles have been displaced 5% of the radius (2.5 cm) towards the wall. b) Pipe diameter is 1 m; the particles have been displaced 1% of the radius (0.5 cm) towards the wall. c) Pipe diameter is 0.1 m; the particles have been displaced 5% of the radius (0.25 cm) towards the wall. d) Pipe diameter is 0.1 m; the particles have been displaced 1% of the radius (0.05 cm) towards the wall. e) No radial displacement of the particle. f) Subdivision of parameter space in behavioral domains, combining (stochastic) turbulent processes acting on the particle population with (deterministic) forces acting on individual particles. DG: Turbulent diffusion and/or gravitational settling dominate radial particle transport. T: Turbophoresis is the most important radial transporting agent. TD: Turbophoresis is enhanced by the Saffman lift force, and the combined mechanism controls radial particle transport. M: Turbophoresis drives particles towards the wall and the Magnus force drives particles towards the center of the pipe. Dashed lines included in all frames are domain boundaries from Fig. 4 for a particle density of 1200 kg/m3.

Fig. 9 e Accumulation of ferric hydroxide flocs in a perspex pipe (inner diameter 100 mm) loop after three (left) and seven (right; stable situation after first three days of experiment) days of circulation, respectively. At a low velocity, accumulation takes place at the bottom half of the pipe wall (left), whereas at higher velocity (right), material is deposited everywhere. These are the results of a repeat of the original experiment by Vreeburg and Boxall (2007), mirroring their results.

1816

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 0 7 e1 8 1 7

Fig. 10 e Positioning of the experimental results of Section 5.1 in the domain diagrams of Figs. 4 and 8f. Recirculation experiment with 100e1000 mm iron hydroxide flocs (density 1.2 g/cm3 at 0.06 m/s; recirculation experiment at 0.14 m/s; common conditions in Dutch drinking water distribution systems; additional common conditions in UK drinking water distribution systems; common conditions in transport mains.

planned, performing a parameter space search (for particle size and flow velocity) to map the locations of the domain boundaries in Fig. 4.

5.2. Discussion of relevance to drinking water distribution systems The most frequent cause for customers to complain to their water company (after lack of water supply) is discoloration (Vreeburg and Boxall, 2007), caused by the presence of suspended particles. The occurrence of water discoloration is generally linked to a disturbance of the normal hydraulic regime, e.g. due to a pipe burst or flushing operations, during which higher than normal flow velocities are capable of resuspending particles which remain at the pipe walls under regular conditions. Under regular operating conditions in drinking water distribution systems, maximum flow velocities of 0.06e0.2 m/ s can be expected (see Fig. 2). Note that typical flow velocities may be somewhat higher in distribution systems of a different design, e.g. self-cleaning networks (Vreeburg et al., 2009). Observed particle sizes in drinking water distribution systems are mostly in the range of 3e12 mm for the Dutch situation and mostly up to some hundreds of mm in the UK, see Section 2.2. and in Fig. 10 correspond to these numbers, and Areas show that particle deposition is well within the diffusion dominated regime (I, DG) for regular operating conditions. In transport mains, flow velocities may be significantly higher (0.4e1.5 m/s) and particles may be much larger (hundreds of

mm). These conditions range from the diffusion dominated in regime to the diffusion-impaction regime (II and T, Fig. 10), where turbophoresis is much more important. As we have shown above, radial transport of the particles towards the wall may be aided by the Saffman lift force once turbophoresis initiates this transport, but this effect is expected to be of minor importance in region T. Boxall et al. (2001) considered the settling behavior of particle samples from UK water distribution systems in rough walled pipes. They concluded that the turbulent motions are such that even for low flow velocities typical of low demand hours and large particles, particles do not undergo gravitational resettling once suspended. However, this appears to be in conflict with the results of the lab experiment at low flow velocity described above and with the experimental results of Ryan et al. (2008). The conclusion of Boxall et al. (2001) poses the problem of the source of the resuspended particles. They have to be either formed in situ (the explanation of Boxall et al. (2001)), or deposited by a different process than gravitational settling. Generally speaking, turbophoresis may be an obvious candidate for the latter explanation if, as mentioned above, flow velocities become sufficiently high. However, it does not suffice as an explanation for the observed sediment accumulation patterns (see Fig. 2 and Blokker et al., 2010a), as the theory predicts turbophoresis to be of minor importance for the reported flow velocities. Considering turbulent processes, most of the above focuses on turbophoresis, even though we have shown that conditions in the water distribution system are generally in the diffusion dominated regime. The effect of turbulent diffusion (diffusion due to Brownian motion is many orders of magnitude weaker) is more subtle than that of turbophoresis. By reducing concentration gradients, the process can counter any mechanism which causes preferential transport of particles to certain areas, such as gravitational settling and turbophoresis. On the other hand, when particles are taken from suspension and become attached to the wall, turbulent diffusion can cause a net transport of particles towards the wall to reduce the ’concentration deficit’ caused by the deposition process. All will depend on the balance of magnitudes of the different mechanisms. The flushing data of Fig. 2 show that there is an undeniable accumulation of particles in drinking water distribution systems, which cannot be explained solely by gravitational settling. We have discussed a number of processes which may contribute to this accumulation. In addition to these, other processes are expected to play a role in the development of the observed patterns such as  temporal variations of the particle size distribution;  timescales of different settling processes;  role of internal irregularities in the distribution pipe system as catchment areas for particles. Knowledge of the conditions under which turbulent particle transport processes occur is a first step in predicting where these processes will result in an accumulation of sediment and therefore an increase in discoloration potential. Subsequent steps are formulating equations which predict the

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 0 7 e1 8 1 7

accumulation rate and experimental verification of the theoretical predictions. Both are subject of ongoing research.

6.

Conclusions

Both turbophoresis and turbulent diffusion can be expected to affect radial particle movement in drinking water distribution pipes. For larger particles and higher flow velocities, turbophoresis is predicted to result in deposition of particles on the pipe wall 360 around. This effect has been demonstrated in a lab experiment and relevant conditions are expected to occur in transport mains. For lower flow velocities and/or smaller particles, representative of distribution mains and pipes, turbulent diffusion is such that no build-up of concentration gradients by turbophoresis is possible. The Saffman lift force may enhance radial particle transport which has been initiated by turbophoresis. The Magnus force is not expected to play any significant role under conditions relevant in drinking water distribution systems.

Acknowledgments We would like to express our gratitude to Hendrik Beverloo and Melanie Tankerville for providing particle size distribution data and again to Melanie Tankerville for reviewing the English of this paper. We thank Roberto Floris for providing pictures of his particle deposition experiments. Also, thoughtful and constructive reviews by two anonymous reviewers are gratefully acknowledged.

references

Blokker, E., Vreeburg, J., Schaap, P., van Dijk, J., 2010a. The selfcleaning velocity in practice. In: Proceedings of the Water Distribution System Analysis Conference. Tucson, Arizona. Blokker, E., Vreeburg, J., van Dijk, J., 2010b. Simulating residential water demand with a stochastic end-use model. Journal of Water Resources Planning and Management 136 (1), 19e26. Boxall, J., Skipworth, P., Saul, A., 2001. A novel approach to modelling sediment movement in distribution mains based on particle characteristics. In: Proceedings of the Computing and Control in the Water Industry Conference, Water Software Systems: Theory and Applications. De Montfort University, UK. Caporaloni, M., Tampieri, F., Trombetti, F., Vittori, O., 1975. Transfer of particles in nonisotropic air turbulence. Journal of the Atmospheric Sciences 32, 565e568. Davidson, P., 2004. Turbulence. An Introduction for Scientists and Engineers. Oxford University Press. Fokeer, S., Kingman, S., Lowndes, I., Reynolds, A., 2004. Characterisation of the cross sectional particle concentration distribution in horizontal dilute flow conveying - a review. Chemical Engineering and Processing 43, 677e691. Guha, A., 1997. A unified Eulerian theory of turbulent deposition to smooth and rough surfaces. Journal of Aerosol Science 28 (8), 1517e1537.

1817

Guha, A., 2008. Transport and deposition of particles in turbulent and laminar flow. Annual Review of Fluid Mechanics 40, 311e341. Hadinoto, K., Jones, E.N., Yurteri, C., Curtis, J.S., 2005. Reynolds number dependence of gas-phase turbulence in gas-particle flows. International Journal of Multiphase Flow 31, 416e434. Liu, S., Yin, S., 2010. Turbulent flows around sand dunes in alluvial rivers. Journal of Hydrodynamics 22 (1), 103e109. Magnus, G., 1853. Ueber die Abweiching der Geschosse, und: Ueber eine auffallende Erscheinung bei rotirenden Ko¨rpern. Annalen der Physik und Chemie 88, 1e29. Odar, F., Hamilton, W., 1964. Forces on a sphere accelerating in a viscous fluid. Journal of Fluid Mechanics 18 (2), 302e314. Rubinow, S., Keller, J., 1961. The transverse force on a spinning sphere moving in a viscous fluid. Journal of Fluid Mechanics 11, 447e459. Ryan, G., Mathes, M., Haylock, G., Jayaratne, A., Wu, J., NouiMehidi, N., Grainger, C., Nguyen, B.V., 2008. Particles in Water Distribution Systems. Tech. Rep. 33. Cooperative Research Centre for Water Quality and Treatment. Saffman, P., 1965. The lift on a small sphere in a slow shear flow. Journal of Fluid Mechanics 22 (2), 385e400. Saffman, P., 1968. Corrigendum. Journal of Fluid Mechanics 31, 624. Sippola, M.R., Nazaroff, W.W., 2002. Particle Deposition from Turbulent Flow: Review of Published Research and Its Applicability to Ventilation Ducts in Commercial Buildings. Tech. Rep. LBNL-51432. Lawrence Berkeley National Laboratory. Soldati, A., Marchioli, C., 2009. Physics and modelling of turbulent particle deposition and entrainment: review of a systematic study. International Journal of Multiphase Flow 35, 827e839. Stumm, W., 1992. Chemistry of the Solid-Water Interface. John Wiley & Sons. Thomas, P., 1992. On the influence of the Basset history force on the motion of a particle through a fluid. Physics of Fluids A 4 (9), 2090e2093. Verbanck, M.A., 2000. Computing near-bed solids transport in sewers and similar sediment-carrying open-channel flows. Urban Water 2, 277e284. ¨ hnlichkeit und Turbulenz. Von Ka´rma´n, T., 1930. Mechanische A Nachrichten von der Gesellschaft der Wissenschaften zu Go¨ttingen. Fachgruppe I (Mathematik) 5, 58e76. Vreeburg, J., 2007. Discolouration in drinking water systems: a particular approach. Ph.D. thesis, Delft University of Technology. Vreeburg, J., Blokker, E., Horst, P., van Dijk, J., 2009. Velocity based self cleaning residential drinking water distribution systems. Water Science and Technology 9 (6), 635e641. Vreeburg, J., Boxall, J., 2007. Discolouration in potable water distribution systems: a review. Water Research 41, 519e529. White, B., Schulz, J., 1977. Magnus effect in saltation. Journal of Fluid Mechanics 81, 497e512. Young, J., Leeming, A., 1997. A theory of particle deposition in turbulent pipe flow. Journal of Fluid Mechanics 340, 129e159. Zanoun, E.-S., Durst, F., Bayoumy, O., Al-Salaymeh, A., 2007. Wall skin friction and mean velocity profiles for fully developed turbulent pipe flows. Experimental Thermal and Fluid Science 32, 249e261. Zanoun, E.-S., Durst, F., Nagib, H., 2003. Evaluating the law of the wall in two-dimensional fully developed turbulent channel flows. Physics of Fluids 15 (10), 3079e3088. Zhu, H., Zhou, Z., Yang, R., Yu, A., 2007. Discrete particle simulation of particulate systems: theoretical developments. Chemical Engineering Science 62, 3378e3396. Zou, X.-Y., Cheng, H., Zhang, C.-L., Zhao, Y.-Z., 2007. Effects of the Magnus and Saffman forces on the saltation trajectories of sand grain. Geomorphology 90, 11e22.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 1 8 e1 8 2 8

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Investigation of pharmaceuticals in Missouri natural and drinking water using high performance liquid chromatography-tandem mass spectrometry Chuan Wang a,b, Honglan Shi b, Craig D. Adams b,c, Sanjeewa Gamagedara a,b, Isaac Stayton a,b, Terry Timmons d, Yinfa Ma a,b,* a

Department of Chemistry, Missouri University of Science and Technology, Rolla, MO 65409, USA Environmental Research Center, Missouri University of Science and Technology, Rolla, MO 65409, USA c Department of Civil, Architectural and Environmental Engineering, University of Kansas, Lawrence, KS, USA d Missouri Department of Natural Resources, Jefferson City, MO, USA b

article info

abstract

Article history:

A comprehensive method has been developed and validated in two different water

Received 18 June 2010

matrices for the analysis of 16 pharmaceutical compounds using solid phase extraction

Received in revised form

(SPE) of water samples, followed by liquid chromatography coupled with tandem mass

23 November 2010

spectrometry. These 16 compounds include antibiotics, hormones, analgesics, stimulants,

Accepted 25 November 2010

antiepileptics, and X-ray contrast media. Method detection limits (MDLs) that were

Available online 7 December 2010

determined in both reagent water and municipal tap water ranged from 0.1 to 9.9 ng/L. Recoveries for most of the compounds were comparable to those obtained using U.S. EPA

Keywords:

methods. Treated and untreated water samples were collected from 31 different water

Pharmaceuticals

treatment facilities across Missouri, in both winter and summer seasons, and analyzed to

Natural and drinking water

assess the 16 pharmaceutical compounds. The results showed that the highest pharma-

LC-MS/MS

ceutical concentrations in untreated water were caffeine, ibuprofen, and acetaminophen, at concentrations of 224, 77.2, and 70 ng/L, respectively. Concentrations of pharmaceuticals were generally higher during the winter months, as compared to those in the summer due, presumably, to smaller water quantities in the winter, even though pharmaceutical loadings into the receiving waters were similar for both seasons. ª 2010 Elsevier Ltd. All rights reserved.

1.

Introduction

Many types of pharmaceuticals are used in this country for a wide variety of applications to benefit humans and animals. Important classes of human pharmaceuticals include: analgesics, antibiotics and antimicrobials, anticonvulsant/antiepileptics, antidiabetics, antihystamines, antipsychotics, antidepressants, antianxiety drugs, beta-blockers (b-blockers), cytostatics and antineoplastics, estrogens and

hormonal compounds, lipid-regulators, stimulants, and Xray contrast media. These compounds may be excreted unmetabolized or partially metabolized, resulting in their eventual passage into the environment. Pharmaceuticals and personal care products (PPCPs) have been detected globally in many natural water systems including rivers, lakes, and reservoirs (Benotti et al., 2009; Daughton and Ternes, 1999; Halling-Sorensen et al., 1997; Jobling et al., 1998; Jones et al., 2001; Kim et al., 2009; Kolpin

* Corresponding author. Department of Chemistry and Environmental Research Center, Missouri University of Science and Technology, 400 West 11th Street, Rolla, MO 65409, USA. Tel.: þ1 573 341 6220; fax: þ1 573 341 6033. E-mail address: [email protected] (Y. Ma). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.11.043

1819

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 1 8 e1 8 2 8

et al., 2002; Loraine and Pettigrove, 2006; Moldovan, 2006; Nakada et al., 2007; Snyder et al., 2001a,b; Yu and Chu, 2009). The trace amounts of pharmaceuticals that have been detected in natural waters have attracted more public attention and serious concern because of their potentially adverse effects on the aquatic environment (Jobling et al., 1998; Keith et al., 2000; Snyder et al., 1999). Moreover, the pharmaceuticals that enter natural waters can ultimately transfer to our drinking water and promote unknown, but disastrous impacts on human health (Benotti et al., 2009; Vanderford and Snyder, 2006; Ye et al., 2007). While the risk that low concentrations of pharmaceuticals pose to humans is still not being adequately investigated (due to their biologically-active nature), it is important to identify the concentrations of these compounds in natural and treated drinking waters. Liquid chromatography-mass spectrometry (LC-MS) (Ahrer et al., 2001; Farre et al., 2001; Lindsey et al., 2001) and liquid chromatography-tandem mass spectrometry (LC-MS/MS) (Asperger et al., 2001; Baronti et al., 2000; Bossi et al., 2002; Croley et al., 2000; Hirsch et al., 1998; Jeannot et al., 2000; Lagana et al., 2000; Sacher et al., 2001; Ternes et al., 2001, 1998; Vanderford and Snyder, 2006; Ye et al., 2007) are popular techniques currently being used in pharmaceutical analyses. Several published reviews have discussed various methods for analyzing pharmaceutical compounds that are found in water resources (Lopez de Alda and Barcelo, 2001; Richardson, 2008, 2010; Ternes, 2001). Since most of pharmaceuticals are present in low concentrations in surface waters, an extraction process (e.g., solid phase extraction (SPE)) is often needed to concentrate target pharmaceutical compounds for analysis. Comprehensive studies that focus on PPCPs in Missouri’s water resources and finished drinking water have not been conducted and reported. In this study, an LC-MS/MS method was developed and applied for simultaneous analysis of 16 pharmaceutical compounds obtained from 31 treated and untreated drinking water resources in the state of Missouri during both the winter and summer seasons. General information about the 16 pharmaceutical compounds is listed in Table 1. Water samples were taken from a broad range of sources, including the Mississippi River, Missouri River, reservoirs, lakes, unconsolidated wells, and deep wells.

2.

Experimental

2.1.

Pharmaceutical standards and reagents

All pharmaceutical standards were purchased from SigmaeAldrich (St. Louis, MO), except lincomycin (purchased from MP Biomedicals, Aurora, OH)) and iopromide (purchased from United States Pharmacopeia, Rockville, MD). Labeled pharmaceutical standards (i.e., 13C3-caffeine, 13C3-Trimethoprim, 13C2Estrone and 13C3-Ibuprofen) were purchased from Cambridge Isotope Laboratories (Andover, MA). Oasis HLB extraction cartridges were obtained from Waters Corp. (Milford, MA), and extraction was performed using a vacuum manifold (Supelco, Corp., Bellefonte, PA). All solvents (methanol, acetonitrile, etc.) were LC-MS grade from Fisher Scientific (Fairlawn, NJ, USA). Formic acid (MS grade) and ammonium acetate (99.99þ%) were

Table 1 e Sixteen pharmaceutical compounds and their general information. Compounds Acetaminophen Caffeine Carbamazepine Clofibric acid Codeine Estradiol Estriol Estrone Ethynylestradiol Ibuprofen Iopromide Lincomycin Sulfamethoxazole Triclosan Trimethoprim Tylosin

Formula

CAS #

Molecular weight

C8H9NO2 C8H10N4O2 C15H12N2O C10H11ClO3 C18H21NO3 C18H24O2 C18H24O3 C18H22O2 C20H24O2 C13H18O2 C18H24I3N3O8 C18H34N2O6S C10H11N3O3S C12H7Cl3O2 C14H18N4O3 C46H77NO17

103-90-2 58-08-2 298-46-4 882-09-07 76-57-3 50-28-2 50-27-1 53-16-7 57-63-6 15687-27-1 107793-72-6 154-21-2 723-46-6 3380-34-5 738-70-5 1401-69-0

151.2 194.2 236.3 214.7 299.4 272.4 288.4 270.4 296.4 206.3 791.1 406.5 253.3 289.5 290.3 916.1

purchased from SigmaeAldrich (St. Louis, MO). Lab reagent water was purified by the Millipore Elix 3 water purification system (Millipore, Bierica, MA).

2.2.

Sample collection and preservation

Sample collection, preservation, and storage were accomplished by following the US EPA Method Guideline (US EPA method 1694, 2007). Sample bottles used to collect water samples were new 500-mL amber glass bottles with Teflon liner screw caps. Prior to use, bottles were pre-cleaned with Mill-Q water, methanol, acetone, and Mill-Q water, and then baked at 105  C for at least 2 h. Forty mg of sodium thiosulfate were added to each bottle to reduce any residual chlorine that had been added as a disinfectant. To collect treated tap water samples, the water was allowed to flow from the tap without an aerator for about 5 min, prior to completely filling the sample bottle, with no headspace. To collect raw (untreated) water, a large pre-cleaned wide-mouth bottle or beaker was used, and then the water was carefully transferred from the original container to a sample bottle that was completely filled with no headspace remaining. Sodium thiosulfate should not be flushed out during the collection of either treated or untreated water. The bottles were then sealed and agitated by hand for about 1 min. Two bottles of untreated source water sample and two bottles of treated water samples were collected from each water treatment facility. The samples were kept on ice and shipped overnight to the laboratory for analysis. Upon arrival, the samples were immediately preserved by adjusting the pH to 5 with sulfuric acid, and then stored at 4  C until extraction (normally, within 2 days after arrival).

2.3.

Solid phase extraction

Solid phase extraction was performed by following US EPA method 1694 (US EPA, 2007), with some modification and validation. Water samples were first filtered using 0.45-mm nylon membrane filters (Whatman, England), and then acidified to pH 2.0  0.2 using HCl. One hundred twenty-five mg of

1820

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 1 8 e1 8 2 8

Na4EDTA$2H2O were then added for each 250 mL of water sample. 50 mL internal standard mixture (concentration 1 mg/ mL) was spiked in the water. Solid phase extraction was conducted using Oasis HLB 6cc (200 mg) cartridges conditioned with 6 mL of methanol, 2 mL of Milli-Q water, 2 mL of pH 2.0 Milli-Q water, and 6 mL of Milli-Q water. Next, 250 mL of each water sample was extracted, at a flow rate of 1e2 drops per second. After extraction of the sample, each cartridge was washed with 5 mL Milli-Q water to remove any EDTA residue. The cartridges were dried under vacuum for 5 min. Analytes were eluted from the cartridge by using 5 ml of methanol and then 3 mL of an acetoneemethanol (1:1) mixture were placed into a clean test tube. The combined eluent was evaporated to 100 mL by using a Turbovap LV evaporator at 50  5  C. Nine hundred mL of 20% acetonitrile in MQ water with 0.1% formic acid was then added to each sample tube and the contents were vortex mixed. Finally, the samples were transferred to 2-mL amber glass sample vials, and stored in a refrigerator until LC/MS/MS analysis.

2.4.

LC-MS/MS analysis

A 4000Q TRAP mass spectrometer (AB SCIEX, Concord, ON, CA), equipped with an Agilent 1100 series HPLC system (Agilent Technologies, Inc., Santa Clara, CA, USA), was used for all analyses. A sample volume of 10 mL was injected. The HPLC column was a Supelco C-18 column (150  2.1 mm, 5-mm particle size, SigmaeAldrich, St. Louis, MO). Different mobile phase additives were used for positive electrospray ionization (ESIþ) and negative electrospray ionization (ESI) mode analyses. The mobile phase for ESIþ compounds consisted of 0.1% formic acid (v/v) in water (A) and 0.1% formic acid (v/v) in acetonitrile (B). Elution flow rate was 0.25 mL/min. The mobile phase for ESI-mode compounds consisted of 5 mM of ammonium acetate in water (C) and 100% acetonitrile (D). Elution flow rate was 0.25 mL/min. For both ESIþ and ESI mode compounds, the gradient began with 5% solvent B or D for 1.5 min, ramping to 75% B or D at 2.5 min, 15% B or D at 13.5 min, and 100% B or D at 15 min, where it was held for 1.5 min. Mass spectrometry utilized both ESIþ and ESI modes in multiple reaction monitoring (MRM). Mass spectrometers, including mass calibration, polarity of each compound, compound-dependent parameters, and sourcedependent parameters were all optimized for each compound. After optimization, the ion transition with the most intense signal was selected as the quantification ion pair of the corresponding compound, while the ion transition with the second highest signal was selected as the confirmation ion pair of the corresponding compound. In positive mode, the ion source temperature was set at 650  C with an ion spray voltage of 5000 V. The nebulizer gas, auxiliary gas, and curtain gas were 50, 65, and 25 psi, respectively. In negative mode, the ion source temperature was set at 550  C with an ion spray voltage of 4500 V. The nebulizer gas, auxiliary gas, and curtain gas were 50, 60, and 30 psi, respectively.

2.5.

Quality control

During the method development, the liner range of calibration for each compound, method detection limit, reagent blank,

reproducibility, spike recoveries of each compound in reagent water and in drinking water matrices were all performed. During the analysis of water samples, at least one blank, one duplicate, and one spike were proceeded with sample preparation and LC-MS detection for each batch (there were generally 8 water treatment facilities in each batch). The choices of these sample matrices represented river water, lake water, well water, and reservoir water.

3.

Results and discussion

3.1.

HPLC separation and mass spectrometry detection

Each sample was injected for both ESIþ and ESI analyses. Ten of the 16 compounds were detected with ESIþ and six of the compounds were detected with the ESI mode. The retention times and compound-dependent mass spectrometry parameters of the target pharmaceuticals are listed in Table 2. Most compounds were well separated chromatographically. Several analytes, that were not well separated chromatographically, were monitored at different MRM ion pairs. All pharmaceuticals were optimized individually for maximum sensitivity. The dwell time for each compound was also optimized, depending on their peak width thereby allowing the peaks to be integrated accurately and with high sensitivity.

3.2. Method detection limit and possible interference in water sample matrix The method detection limit (MDL) for each compound was determined by following U.S. EPA method (EPA-821-R-03-005), spiking 9 replicates of 2e5 times concentration of the estimated instrumental detection limits (IDL) (before 250 concentrated by SPE) (The IDL of each PPCP compound was determined based on the PPCP concentration at signal-tonoise ratio of 3e5). The spiked PPCP standards were proceeded by SPE and then analyzed by LC-MS/MS. The MDL of each compound was calculated by multiplying the standard deviation(s) of the replicate analyses by the Student’s T-value for eight degrees of freedom at the 99% confidence level. The MDL in both reagent water and tap drinking water were determined separately. A sample matrix was analyzed first to assess the concentrations of the target pharmaceutical compounds. For the tap water matrix, since some target analytes were found to be present in the water, each pharmaceutical standard was spiked individually to make the concentrations of pharmaceuticals fall to 2e5 times the estimated IDL and proceeded to determine MDL. The method detection limits of studied pharmaceuticals are listed in Table 3.

3.3.

Spike recovery

Spike recoveries in both reagent water and tap water were determined separately. For each type of water, two concentrations (low and high) and nine replicates of each concentration were performed discretely. High and low concentrations were selected to cover the most likely range of pharmaceutical concentrations that would be expected to be

1821

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 1 8 e1 8 2 8

Table 2 e LC-MS/MS experimental conditions of the sixteen pharmaceutical compounds. Compounds

Acetaminophen Caffeine Carbamazepine Clofibric acid Codeine Estradiol Estriol Estrone Ethynylestradiol Ibuprofen Iopromide Lincomycin Sulfamethoxazole Triclosan Trimethoprim Tylosin

Retention ESI mode Precursorion Production Declustering Collision Collision time mode potential (V) energy (V) cell exit (min) potential (V) 5.7 8.2 11.7 8.1 7.9 13.1 8.8 12 13.8 9.5 3.5 8 10.2 15.8 8.3 10.6

Positive Positive Positive Negative Positive Positive Negative Negative Positive Negative Negative Positive Positive Negative Positive Positive

151.8 194.9 236.9 212.9 300 255 286.7 268.9 279 204.9 789.9 407.1 253.9 286.7 291 916.5

110 138 194 126.8 215 159 171.1 144.8 133.2 159 126.7 126.2 156 35 230.2 174.1

observed in actual treated and untreated water samples. Recoveries of all compounds were determined by adding an appropriate amount of standard solutions of the target pharmaceutical compounds to 250 mL of reagent and real water samples. The recovery study consisted of four experiments. For each experiment involved, known concentrations of target pharmaceutical compounds were prepared in nine replicate water samples (250 mL water each). The samples were proceeded by SPE and then analyzed by LC-MS/MS. The relative percent recoveries and relative standard deviations were calculated, and are reported in Table 3. Recoveries of most of the high concentration pharmaceutical compounds (200e400 ng/L) in reagent and tap water ranged between 90 and 130%. The recoveries of tylosin and

66 51 71 40 81 76 100 110 61 55 85 76 61 50 81 136

25 31 31 24 37 29 50 54 27 10 48 39 25 110 35 55

Internal s tandard 13

6 8 12 7 14 10 9 7 8 10 5 6 10 3 14 10

C3-caffeine C3-caffeine 13 C3-trimethoprim 13 C3-ibuprofen 13 C3-trimethoprim 13 C3-caffeine 13 C6-estrone 13 C6-estrone 13 C3-caffeine 13 C3-ibuprofen 13 C3-ibuprofen 13 C3-trimethoprim 13 C3-trimethoprim 13 C6-estrone 13 C3-trimethoprim 13 C3-caffeine 13

triclosan, however, were relatively low in both water matrices, while the recovery of clofibric acid was high in the tap water matrix. The results of recovery studies, in most cases, showed excellent reproducibility with the percent relative standard deviation being less than 10%. Most of the compounds at the spiked concentrations behaved similarly between the two different water matrices. Recoveries of the pharmaceutical compounds at low concentration spikes (ranging 1.0e20 ng/L, which was near the MDL for most of the compounds) ranged from 50 to 150% for most of the compounds. Majority of the compounds also showed good reproducibility, with the percent relative standard deviation being less than 15%. For most of the studied compounds, the low concentration recoveries did not show

Table 3 e Method detection limits (MDL), spike recovery and relative standard deviation (RSD) of studied pharmaceuticals in reagent water (Deionized (DI)) water and tap water matrix (n [ 9). Compound

MDL (ng/L) DI Tap water water

Acetaminophen Caffeine Carbamazepine Clofibric acid Codeine Estradiol Estriol Estrone Ethynylestradiol Ibuprofen Iopromide Lincomycin Sulfamethoxazole Triclosan Trimethoprim Tylosin

2.7 0.8 0.5 1.3 1.0 0.8 4.3 1.4 0.1 1.0 3.5 0.1 0.4 1.0 0.3 0.3

1.4 1.1 0.2 0.6 1.5 1.2 5.2 1.0 0.5 1.6 9.9 0.1 0.3 1.2 0.4 0.2

High-level spike DI water

Low-level spike

Tap water

DI water

Tap water

Spike Recovery RSD Spike Recovery RSD Spike Recovery RSD Spike Recovery RSD (ng/L) (%) (%) (ng/L) (%) (%) (ng/L) (%) (%) (ng/L) (%) (%) 200 400 200 400 400 400 200 200 200 400 200 200 400 400 400 200

95.8 109 104 122 104 95.6 128 116 92.3 119 98.1 103 90 32 123 44.6

10.4 9.1 4.9 12.9 2.6 5 4.9 2.6 6.2 1.9 8.8 9.9 11.5 14.1 11.3 14.8

200 400 200 400 400 400 200 200 200 400 200 200 400 400 400 200

100 115 121 157 122 106 115 109 102 110 90.6 97.5 102 33.6 125 48.6

6.9 8.5 5.3 7.4 4.5 4.4 2.4 1.9 2.8 2.2 7.8 5.6 4.7 15.7 4.9 30.8

10 2 1 1 5 5 20 2 1 20 20 1 1 10 1 1

106 46.4 125 72.6 145 102 120 89.2 108 106 136 12.6 90.5 39.3 91.2 18.2

8.8 29 13 28.1 4.7 5.3 6.2 26 4.4 3.5 4.4 32.8 14.8 8.7 12.2 51.2

10 2 1 1 5 5 20 2 1 20 20 1 1 10 1 1

73.9 50.7 125 56.7 150 97.9 121 123 75.8 85.9 153 7.1 92.7 29.8 91.2 15.1

6.4 37.4 5.9 37.5 6.7 8.7 7.4 13.9 22 3.3 11.2 55.2 9.3 13.2 14 37.8

1822

Table 4 e Pharmaceutical concentration of real water samples in winter season. ID #

Water source

Treatment

MS River

Free chlorine

2

MS River

Chloramines

3

MO River

Chloramines

4

Alluvial GW

Chloramines

5

Alluvial GW

Free chlorine

6

Chloramines Free chlorine

8

Deep rock wells Deep rock wells Reservoirs

9

Reservoir

Chloramines

10

MO River

Chloramines

11

MS River

Free chlorine

12

Lake

Free chlorine

13

Lake

Chloramines

14

Lake

Chloramines

15

Deep well

Free chlorine

16

Deep well

Free chlorine

17

Deep well

Free chlorine

18

Deep well

Free chlorine

19

Lake

Free chlorine

20

Unconsolidated wells Unconsolidated wells Unconsolidated wells

Free chlorine

7

21 22

Free chlorine

Free chlorine Free chlorine

Untreated Treated Untreated Treated Untreated Treated Untreated Treated Untreated Treated Untreated Treated Untreated Treated Untreated Treated Untreated Treated Untreated Treated Untreated Treated Untreated Treated Untreated Treated Untreated Treated Untreated Treated Untreated Treated Untreated Treated Untreated Treated Untreated Treated Untreated Treated Untreated Treated Untreated Treated

Concentration (ng/L) Tylosin

Lincomycin

Trimethoprim

Sulfamethoxazole

Acetaminophen

Caffeine

Carbamazepine

Codeine

Triclosan

Ibuprofen

Iopromide

<MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL ND ND <MDL <MDL <MDL <MDL <MDL

4.2 <MDL <MDL <MDL 1.8 <MDL 1.1 <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL 2.5 <MDL 2.9 <MDL 2.8 <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL ND ND <MDL <MDL <MDL <MDL <MDL

4.6 <MDL 7.7 <MDL 2.8 1.7 <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL 9.1 <MDL 4.0 <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL ND ND <MDL <MDL <MDL <MDL <MDL

13.7 <MDL 28.8 1.5 18.8 8.2 5.5 <MDL <MDL <MDL 2.8 <MDL <MDL <MDL <MDL <MDL <MDL <MDL 38.1 4.8 18.2 <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL 1.7 ND ND <MDL <MDL <MDL <MDL <MDL

5.3 <MDL 21.8 <MDL 16.0 <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL 14.2 <MDL 56.0 <MDL 11.6 28.0 <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL ND ND <MDL <MDL <MDL <MDL <MDL

47.2 13.0 106.0 54.4 49.6 36.0 14.8 2.5 50.0 43.6 11.4 3.4 18.2 4.5 10.3 8.8 157.2 18.4 224.8 180.8 39.0 32.9 20.1 8.9 21.3 4.0 29.4 4.3 6.2 6.5 3.0 4.4 2.9 2.6 2.7 3.0 11.1 ND ND 2.8 5.6 7.2 34.4 9.3

8.4 2.2 8.7 3.9 5.5 4.7 3.2 <MDL <MDL <MDL 1.9 <MDL <MDL <MDL <MDL <MDL 1.0 <MDL 8.1 6.8 8.1 3.2 <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL ND ND <MDL <MDL <MDL <MDL <MDL

<MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL 3.7 <MDL 3.0 <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL ND ND <MDL <MDL <MDL <MDL <MDL

<MDL <MDL <MDL <MDL 2.1 <MDL <MDL <MDL <MDL <MDL 9.8 2.8 <MDL <MDL 3.4 <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL ND ND <MDL <MDL <MDL <MDL <MDL

16.6 8.8 37.5 26.6 27.1 23.4 <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL 2.4 2.0 77.2 72.8 13.6 10.4 <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL ND ND <MDL <MDL <MDL <MDL <MDL

<MDL <MDL 22.4 <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL ND ND <MDL <MDL <MDL <MDL <MDL

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 1 8 e1 8 2 8

1

Sample type

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 1 8 e1 8 2 8

39.1 3.2 6.7 4.8 27.7 3.7 7.9 7.4 10.0 6.0 11.3 6.5 31.5 20.4 16.9 3.8 10.8 5.2

3.3 <MDL <MDL <MDL <MDL <MDL <MDL <MDL 4.8 3.5 <MDL <MDL 3.2 1.4 <MDL <MDL <MDL <MDL

<MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL

<MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL

<MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL 4.4 3.2 <MDL <MDL <MDL <MDL

<MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL

significant differences between the two water matrices. However, some compounds, such as lincomycin, triclosan, and tylosin, had low recovery in both reagent water and tap water matrices. Acetaminophen and ethynylestradiol had relatively lower recoveries in tap water. Lincomycin also had a lower recovery and much higher percent relative standard deviation in the tap water matrix than reagent water matrix. This phenomenon was also reported in US EPA method 1694. Ideally, an isotope-labeled internal standard should be used for each of the studied compounds. However, only four labeled internal standards were used in this study due to the limited budget. Different isotope-labeled internal standards used in this study may also have contributed to the poor recoveries of pharmaceuticals. During pharmaceutical occurrence study in Missouri water, spike recoveries were also tested with each batch of samples and each major water type (river, lake, and well). The recoveries were similar with or better than the initial recovery studies as shown above.

1.1 <MDL <MDL <MDL 7.7 <MDL <MDL <MDL <MDL <MDL 1.1 <MDL <MDL <MDL <MDL <MDL <MDL <MDL

<MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL 2.1 <MDL

7.2 <MDL <MDL <MDL <MDL <MDL <MDL <MDL 3.1 <MDL <MDL <MDL 6.1 <MDL <MDL <MDL <MDL <MDL

<MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL 70.0 <MDL <MDL <MDL 19.4 <MDL <MDL <MDL <MDL <MDL

3.4. Occurrence of pharmaceuticals in Missouri drinking water systems

Free chlorine

Free chlorine

Chloramines

Free chlorine

Chloramines

Lake

Reservoir

Lake

River

Lake

Lake

26

27

28

29

30

31

Free chlorine

Chloramines Lake 25

Free chlorine 24

ND ¼ No data; MDL ¼ Method detection limit.

<MDL <MDL 4.3 <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL Untreated Treated Untreated Treated Untreated Treated Untreated Treated Untreated Treated Untreated Treated Untreated Treated Untreated Treated Untreated Treated Free chlorine

Unconsolidated wells Lake 23

1823

Treated and untreated waters were sampled from 31 different water treatment facilities across Missouri in both the cold (winter) and warm (summer) seasons. These water treatment facilities used varied source waters, including rivers, reservoirs, lakes, unconsolidated wells, groundwater, and deep wells. The water samples from these water treatment plants represent the most common tap drinking water in Missouri. The water treatment plants on the Missouri and Mississippi Rivers utilize conventional treatment (usually including pre-sedimentation, rapid mix, flocculation, sedimentation, filtration, chlorine and/or chloramine disinfection, periodic powdered activated carbon (PAC) adsorption, and two-stage lime softening). Specifically, the treatment plants utilize ferric chloride or sulfate coagulants, chlorine as the primary disinfection, and chloramine as the residual disinfectant. All of the water facilities that participated in this study used both chlorine and chloramines as water disinfectants. Periodic powdered activated carbon addition is also used. Both untreated and treated water samples were collected from each water treatment facility at same time to evaluate the effects of the water treatment process on the studied pharmaceuticals. In the winter season, 11 pharmaceutical compounds (Table 4) were detected in at least one of the untreated source waters: Caffeine was found in all of the untreated and treated water samples, with concentration ranges of 2.5e225 ng/L, and were present at higher levels than most of the other pharmaceuticals monitored in this study. Caffeine concentrations were higher in river water than in most of the other types of water. In most cases, lower concentrations were detected in treated waters (free chlorine or chloramines) than in untreated waters, indicating some removal of this compound by the water treatment process. Carbamazepine was found in 11 of the 31 water treatment facilities studied, in both treated and untreated waters at concentrations up to 8.7 ng/L. The concentrations were also lower, however, in treated water than in untreated water. This compound was found in all river water samples. Sulfamethoxazole was

1824

Table 5 e Pharmaceutical concentration of real water samples in summer season. ID #

Water source

Treatment

MS River

Free chlorine

2

MS River

Chloramines

3

MO River

Chloramines

4

Alluvial GW

Chloramines

5

Alluvial GW

Free chlorine

6

Chloramines

8

Deep rock wells Deep rock wells Reservoirs

9

Reservoir

Chloramines

10

MO River

Chloramines

11

MS River

Free chlorine

12

Lake

Free chlorine

13

Lake

Chloramines

14

Lake

Chloramines

15

Deep well

Free chlorine

16

Deep well

Free chlorine

17

Deep well

Free chlorine

18

Deep well

Free chlorine

19

Lake

Free chlorine

20

Unconsolidated wells

Free chlorine

21

Unconsolidated wells

Free chlorine

22

Unconsolidated wells

Free chlorine

7

Free chlorine Free chlorine

Untreated Treated Untreated Treated Untreated Treated Untreated Treated Untreated Treated Untreated Treated Untreated Treated Untreated Treated Untreated Treated Untreated Treated Untreated Treated Untreated Treated Untreated Treated Untreated Treated Untreated Treated Untreated Treated Untreated Treated Untreated Treated Untreated Treated Untreated Treated Untreated Treated Untreated Treated

Concentration (ng/L) Tylosin

Lincomycin

Trimethoprim

Sulfamethoxazole

Acetaminophen

Caffeine

Carbamazepine

Codeine

Triclosan

Ibuprofen

Iopromide

<MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL 3.9 <MDL <MDL <MDL <MDL <MDL

2.6 <MDL 4.6 2.0 1.9 <MDL 1.3 <MDL <MDL <MDL <MDL <MDL <MDL <MDL 3.0 <MDL 1.3 <MDL <MDL <MDL 3.9 <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL 7.0 4.4 <MDL <MDL <MDL <MDL

<MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL 5.1 4.7 <MDL <MDL <MDL <MDL

4.2 <MDL 11.0 <MDL 4.6 4.0 4.8 1.3 3.9 <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL 20.3 1.6 4.0 <MDL <MDL <MDL 1.9 <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL 1.1 <MDL 3.2 1.5 <MDL <MDL <MDL <MDL

<MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL 9.4 6.2 46.0 9.5 32.8 9.4 <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL

15.6 9.5 134.8 22.8 27.6 12.4 32.4 8.8 7.2 7.2 18.4 20.9 14.1 6.1 56.0 17.8 6.6 3.2 46.0 35.6 25.3 11.6 59.6 16.4 6.0 3.3 75.6 14.6 1.7 1.4 1.6 1.4 14.9 6.5 14.5 6.3 16.4 14.9 13.2 8.1 5.2 1.5 152.0 1.2

5.7 1.0 8.6 5.0 4.4 1.5 8.3 3.4 <MDL <MDL 1.3 <MDL <MDL <MDL <MDL <MDL <MDL <MDL 7.3 5.0 4.2 1.2 <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL 1.6 <MDL 9.6 6.9 <MDL <MDL <MDL <MDL

<MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL

3.1 <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL 7.0 <MDL 2.9 2.2 4.1 2.6 <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL 3.8 3.5 3.1 2.6 <MDL <MDL <MDL <MDL

<MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL

<MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 1 8 e1 8 2 8

1

Sample type

ND ¼ No data; MDL ¼ Method detection limit.

Lake 31

Chloramines

Lake 30

Free chlorine

River 29

Chloramines

Lake 28

Free chlorine

Reservoir 27

Free chlorine

Lake 26

Free chlorine

Lake 25

Chloramines

Lake 24

Free chlorine

Unconsolidated wells 23

Free chlorine

Untreated Treated Untreated Treated Untreated Treated Untreated Treated Untreated Treated Untreated Treated Untreated Treated Untreated Treated Untreated Treated

<MDL <MDL <MDL <MDL <MDL <MDL ND <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL

<MDL <MDL <MDL <MDL <MDL <MDL ND <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL

<MDL <MDL <MDL <MDL <MDL <MDL ND <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL

<MDL <MDL <MDL <MDL <MDL <MDL ND <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL

<MDL <MDL <MDL <MDL <MDL <MDL ND <MDL <MDL <MDL <MDL <MDL 4.6 <MDL <MDL <MDL <MDL <MDL

111.2 9.6 8.9 2.7 24.0 4.4 ND 5.8 12.4 11.1 41.2 6.8 13.1 10.6 8.9 2.5 6.4 3.6

<MDL 1.3 <MDL <MDL <MDL <MDL ND <MDL 7.6 4.7 <MDL <MDL 17.9 3.6 <MDL <MDL <MDL <MDL

<MDL <MDL <MDL <MDL <MDL <MDL ND <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL

<MDL <MDL 5.0 4.9 4.4 4.2 ND 7.3 5.2 3.4 9.1 5.4 <MDL <MDL <MDL <MDL <MDL <MDL

<MDL <MDL <MDL <MDL <MDL <MDL ND <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL

<MDL <MDL <MDL <MDL <MDL <MDL ND <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 1 8 e1 8 2 8

1825

present at a detectable level in 11 out of the 31 water treatment facilities, primarily in untreated river waters, with a maximum concentration of 38.1 ng/L. This indicated that the water treatment effectively removed most of this compound. Ibuprofen was also found in both untreated and treated waters in 7 of the 31 water treatment facilities; 6 sources were from rivers with a highest concentration of 77.2 ng/L. The concentrations of this compound were found to be about the same before and after water treatment, indicating that the water treatment procedure used was not effective in removing this pharmaceutical. The concentrations of the rest of the target compounds were all below method detection limits. These results indicate that the concentrations of detected pharmaceutical compounds were water source dependent and were usually higher in river water and other types of surface water samples than in groundwater. In the summer season, fewer compounds were detected in water facilities (Table 5) than in winter (Table 4) including lower concentrations or detections of tylosin, lincomycin, thrimethoprim, sulfamethoxazole, acetaminophen, caffeine, carbamazepine, and triclosan. The relative concentrations of the detected compounds in different types of water sources in water samples during the summer shared a trend similar to that of water samples collected during the winter season. Higher concentrations of pharmaceuticals overall may be due to several factors. One of the possible reasons for this difference might be due to more rapid degradation of these pharmaceuticals in warmer summer temperatures. The second contributing factor may be attributed to dilution effects being less in drier winter months and the relatively lower river flowrates. To test this hypothesis, the mass flow rate of caffeine, a commonly used tracer pharmaceutical, was studied. Based on stream flow and concentration data, the caffeine mass flow rate in the Missouri River was 89, 142, 98, and 131 mg/s, respectively, for January, April, May, and June, 2009. The mean and relative standard deviation were 115 mg/s and 22%, respectively. Similarly, the caffeine mass flow rate in the Mississippi River was 281, 238, 284, and 183 mg/s, respectively, for January, April, May and June, 2009. The mean and relative standard deviation were 247 mg/s and 19%, respectively. For both rivers, there was no significant difference in mass flows between the winter and summer months, indicating that the river flow variation is sufficient to explain differences in concentrations. Similarly, results are also seen for sulfamethoxazole. Following guidance from Missouri Department of Natural Resources (MDNR), three types of water matrixes (Mississippi River water, reservoir water and Missouri River water) were sampled monthly from winter to summer for more detailed assessments of changes in the trend of applicable pharmaceutical compounds. Assessment results showed that eight pharmaceutical compounds were detected in Mississippi River and Missouri River water samples over this period (i.e., lincomycin, trimethoprim, sulfamethoxazole, acetaminophen, caffeine, carbamazepine, triclosan, and ibuprofen), with the caffeine level being the highest. Pharmaceutical compound levels found in both Mississippi River water and Missouri River water were similar. In the reservoir water samples, however, only four compounds were detected (i.e., lincomycin, acetaminophen, caffeine, and triclosan) and their

1826

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 1 8 e1 8 2 8

concentrations were much lower than those detected in Mississippi River and Missouri River water samples. The monthly monitoring also showed a decrease in the concentrations of detectable pharmaceutical compounds from February to June. Specific concentrations of all target pharmaceutical compounds in treated and untreated waters samples in this seasonal study are reported in Table 6. One of the largest studies of pharmaceutical occurrence in untreated drinking water sources from the U.S. Geological Survey (Focazio et al., 2008) concluded that the concentrations of most detected compounds were typically in the sub-mg/L range. This matched our results. Our study also indicated that the highest concentrations of the detected pharmaceuticals in Missouri were lower than the maximum concentrations in Focazio’s report. In our study, caffeine was much more frequently detected than in their national study. However, the MDL in the national study (14 ng/L) was more than 10 times higher than the MDL in this study (1.1 ng/L). The maximum concentrations of caffeine in our study were 0.22 mg/L whereas the national survey reported 0.27 mg/L.

3.5.

The removal efficiency by water treatment facilities

Since untreated and treated water samples were collected at the same time from each water treatment facility, the pharmaceutical concentration difference should represent the removal efficiency of water treatment. Free chlorine and chloramines were used as disinfectants in these selected water treatment plants. In general, the concentrations of

target pharmaceuticals in finished drinking water decreased compared with the untreated source water in this occurrence study. The decreased concentrations can be attributed to the water treatment processes, including clarification, disinfection (chlorination), and activated carbon sorption as investigated by Stackelberg et al. (2007). Both free chlorine and chloramine treatments showed effective removal of acetaminophen. Previous work showed that the phenolic functional group of acetaminophen reacts with chlorine during chlorine disinfection (Pinkston and Sedlak, 2004). Removal of carbamazepine was not effective by chlorination or chloramination treatment. Most water treatment plants failed to remove this compound to below detection limit. This result agreed with previous studies (Benotti et al., 2009; Stackelberg et al., 2007). Though carbamazepine removal using chlorine at lower pH was reported (Westerhoff et al., 2005), the water treatments in Missouri water disinfection facilities were performed at close to neutral to higher pH. The removal efficiency of caffeine was also found poor by chlorination or chloramination treatments in this study. This was generally in agreement with observations by Snyder et al. (2007). The authors reported that the removal efficiencies of caffeine by chlorination or chloramination were less than 20% after 24 h treatment at pH 7.9e8.5. Some differences in removal efficiency were also found between chlorination and chloramination treatments. Occurrence results indicated that eight of nine water treatment plants removed sulfamethoxazole to below detection limit with chlorination treatment. Only four of ten water

Table 6 e Seasonal monitoring of PPCPs from February to June 2009. Month

Feb

Water source

Treatment

MO River

Chloramines

Reservoir Chloramines

April

MS River

Free chlorine

MO River

Chloramines

Reservoir Chloramines

May

MS River

Free chlorine

MO River

Chloramines

Reservoir Chloramines

June

MS River

Free chlorine

MO River

Chloramines

Reservoir Chloramines MS River

Free chlorine

MDL ¼ Method detection limit.

Sample type

Untreated Treated Untreated Treated Untreated Treated Untreated Treated Untreated Treated Untreated Treated Untreated Treated Untreated Treated Untreated Treated Untreated Treated Untreated Treated Untreated Treated

Concentration (ng/L) Linco- Trimetho- Sulfameth- Acetamin- Caffeine Carbama- Triclo- Ibupromycin prim oxazole ophen zepine san fen 1.8 <MDL 2.5 <MDL 4.2 <MDL 1.1 <MDL 5.9 <MDL 1.8 <MDL 2.3 <MDL 6.9 <MDL 2.0 <MDL 1.9 <MDL 1.3 <MDL 2.6 <MDL

2.8 <MDL <MDL <MDL 4.6 <MDL <MDL <MDL <MDL <MDL 2.6 <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL

18.8 8.2 <MDL <MDL 13.7 <MDL 14.0 6.8 2.2 <MDL 9.4 <MDL 8.4 5.6 2.2 <MDL 6.7 <MDL 4.6 4.0 <MDL <MDL 4.2 <MDL

16.0 <MDL 14.2 <MDL 5.3 <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL

49.6 36.0 157.2 18.4 47.2 13.0 38.3 22.0 110.8 2.7 19.9 3.4 17.6 10.6 4.6 3.7 18.7 2.8 27.6 12.4 6.6 4.0 15.6 9.5

5.5 4.7 1.0 <MDL 8.4 2.2 4.2 2.1 1.3 <MDL 4.6 <MDL 3.3 <MDL 1.1 <MDL 4.0 <MDL 4.4 1.5 <MDL <MDL 5.7 1.0

2.1 <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL 4.2 3.4 5.4 <MDL 8.1 <MDL 8.9 6.9 <MDL <MDL <MDL <MDL 3.1 <MDL

27.1 23.4 2.4 <MDL 16.6 8.8 2.2 <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL <MDL

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 1 8 e1 8 2 8

treatment plants, however, removed sulfamethoxazole to below detection limit with chloramination treatment. These results agreed well with a previous study (Chamberlain and Adams, 2006), which showed that monochloramine was much less effective than free chlorine to remove sulfonamides at typical dosage concentrations of 3 mg/L. Similar results were also reported by Stackelberg et al. (2007) for pharmaceutical removal by conventional drinking water-treatment processes. Specifically, they found that sulfamethoxazole in source water was most effectively removed after chlorination treatment. Another comprehensive study on sulfamethoxazole oxidation by free chlorine and combined chlorine showed that a reaction half-life with free chlorine was just 23 s, while the reaction half-life with free chloramines was 38 h (Dodd and Huang, 2004). The distribution of these compounds was source water dependent. River water and surface water samples had higher concentrations of studied pharmaceutical compounds, although concentrations of pharmaceutical compounds in well water were much lower due to interactions and removal on soil and aquifer materials. Results also showed that most of the investigated pharmaceutical compounds were not detectable after water treatment. Specific pharmaceutical compounds including lincomycin, trimethoprim, and sulfamethoxazole, were relatively easier to remove than others including ibuprofen, carbamazepine, and caffeine. Even though some pharmaceuticals were still detectable after treatment, the concentrations were much lower than those in the untreated water, which implied that most pharmaceutical compounds studied can be effectively removed during water treatment.

4.

Conclusions

A comprehensive method has been developed and validated for quantitative analysis of 16 pharmaceutical compounds in two different water matrices, using liquid chromatography coupled with tandem mass spectrometry. Most selected compounds investigated by this method showed good recovery and reproducibility in reagent water and tap water. The results of the occurrence study indicated that the levels of selected compounds in different types of water resources across Missouri were below 80 ng/L for all of the pharmaceuticals, except caffeine, which had the highest concentration, while levels of other compounds were very low. Most of the compounds were usually below method detection limits. The levels of the studied pharmaceutical compounds were also water source dependent. Studied pharmaceutical compound concentrations were higher in surface water than those in well water. Occurrence data also showed that the levels of the most pharmaceutical compounds in summer water samples were lower than those found in the winter water samples. The treatment processes in water facilities across Missouri were found to be effective for removing most pharmaceutical compounds.

Acknowledgements This study was funded by Missouri Department of Natural Resources and Applied Biosystems, Inc.. We also thank

1827

Millipore, Inc., and the Environmental Research Center at Missouri University of Science and Technology for the valuable support.

references

Ahrer, W., Scherwenk, E., Buchberger, W., 2001. Determination of drug residues in water by the combination of liquid chromatography or capillary electrophoresis with electrospray mass spectrometry. J. Chromatogr., A 910, 69e78. Asperger, A., Efer, J., Koal, T., Engewald, W., 2001. On the signal response of various pesticides in electrospray and atmospheric pressure chemical ionization depending on the flow-rate of eluent applied in liquid chromatography-tandem mass spectrometry. J. Chromatogr., A 937, 65e72. Baronti, C., Curini, R., D’Ascenzo, G., Di Corcia, A., Gentili, A., Samperi, R., 2000. Monitoring natural and synthetic estrogens at activated sludge sewage treatment plants and in a receiving river water. Environ. Sci. Technol. 34, 5059e5066. Benotti, M.J., Trenholm, R.A., Vanderford, B.J., Holady, J.C., Stanford, B.D., Snyder, S.A., 2009. Pharmaceuticals and endocrine disrupting compounds in U.S. drinking water. Environ. Sci. Technol. 43, 597e603. Bossi, R., Vejrup, K.V., Mogensen, B.B., Asman, W.A.H., 2002. Analysis of polar pesticides in rainwater in Denmark by liquid chromatography-tandem mass spectrometry. J. Chromatogr., A 957, 27e36. Chamberlain, E., Adams, C., 2006. Oxidation of sulfonamides, macrolides, and carbadox with free chlorine and monochloramine. Water Res. 40, 2517e2526. Croley, T.R., Hughes, R.J., Koenig, B.G., Metcalfe, C.D., March, R.E., 2000. Mass spectrometry applied to the analysis of estrogens in the environment. Rapid Commun. Mass Spectrom. 14, 1087e1093. Daughton, C.G., Ternes, T.A., 1999. Pharmaceuticals and personal care products in the environment: agents of subtle change? Environ. Health Perspect. Suppl. 107, 907e938. Dodd, M.C., Huang, C., 2004. Transformation of the antibacterial agent sulfamethoxazole in reactions with chlorine: kinetics, mechanisms, and pathways. Environ. Sci. Technol. 38, 5607e5615. Farre, M., Ferrer, I., Ginebreda, A., Figueras, M., Olivella, L., Tirapu, L., Vilanova, M., Barcelo, D., 2001. Determination of drugs in surface water and wastewater samples by liquid chromatography-mass spectrometry: methods and preliminary results including toxicity studies with Vibrio fischeri. J. Chromatogr., A 938, 187e197. Focazio, M.J., Kolpin, D.W., Barnes, K.K., Furlong, E.T., Meyer, M. T., Zaugg, S.D., Barber, L.B., Thurman, M.E., 2008. A national reconnaissance for pharmaceuticals and other organic wastewater contaminants in the United States d II: untreated drinking water sources. Sci. Total Environ. 402, 201e216. Halling-Sorensen, B., Nielsen, S.N., Lanzky, P.F., Ingerslev, F., Lutzhft, H.C.H., Jorgensen, S.E., 1997. Occurrence, fate and effects of pharmaceutical substances in the environment e a review. Chemosphere 36, 357e393. Hirsch, R., Ternes, T.A., Haberer, K., Mehlich, A., Ballwanz, F., Kratz, K.-L., 1998. Determination of antibiotics in different water compartments via liquid chromatography-electrospray tandem mass spectrometry. J. Chromatogr., A 815, 213e223. Jeannot, R., Sabik, H., Sauvard, E., Genin, E., 2000. Application of liquid chromatography with mass spectrometry combined with photodiode array detection and tandem

1828

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 1 8 e1 8 2 8

mass spectrometry for monitoring pesticides in surface waters. J. Chromatogr., A 879, 51e71. Jobling, S., Nolan, M., Tyler, C.R., Brighty, G., Sumpter, J.P., 1998. Widespread sexual disruption in wild fish. Environ. Sci. Technol. 32, 2498e2506. Jones, O.A.H., Voulvoulis, N., Lester, J.N., 2001. Human pharmaceuticals in the aquatic environment. A review. Environ. Technol. 22, 1383e1394. Keith, L.H., Jones, T.L., Needham, L.L., 2000. Analysis of Environmental Endocrine Disruptors. American Chemical Society, pp. 1e173. Kim, J.-W., Jang, H.-S., Kim, J.-G., Ishibashi, H., Hirano, M., Nasu, K., Ichikawa, N., Takao, Y., Shinohara, R., Arizono, K., 2009. Occurrence of pharmaceutical and personal care products (PPCPs) in surface water from Mankyung River, South Korea. J. Health Sci. 55, 249e258. Kolpin, D.W., Furlong, E.T., Meyer, M.T., Thurman, E.M., Zaugg, S. D., Barber, L.B., Buxton, H.T., 2002. Pharmaceuticals, hormones, and other organic wastewater contaminants in U. S. streams, 1999e2000: a National reconnaissance. Environ. Sci. Technol. 36, 1202e1211. Lagana, A., Bacaloni, A., Fago, G., Marino, A., 2000. Trace analysis of estrogenic chemicals in sewage effluent using liquid chromatography combined with tandem mass spectrometry. Rapid Commun. Mass Spectrom. 14, 401e407. Lindsey, M.E., Meyer, M., Thurman, E.M., 2001. Analysis of trace levels of sulfonamide and tetracycline antimicrobials in groundwater and surface water using solid-phase extraction and liquid chromatography/mass spectrometry. Anal. Chem. 73, 4640e4646. Lopez de Alda, M.J., Barcelo, D., 2001. Review of analytical methods for the determination of estrogens and progestogens in wastewaters. Fresenius’ J. Anal. Chem. 371, 437e447. Loraine, G.A., Pettigrove, M.E., 2006. Seasonal variations in concentrations of pharmaceuticals and personal care products in drinking water and Reclaimed wastewater in Southern California. Environ. Sci. Technol. 40, 687e695. Moldovan, Z., 2006. Occurrences of pharmaceutical and personal care products as micropollutants in rivers from Romania. Chemosphere 64, 1808e1817. Nakada, N., Komori, K., Suzuki, Y., Konishi, C., Houwa, I., Tanaka, H., 2007. Occurrence of 70 pharmaceutical and personal care products in Tone River basin in Japan. Water Sci. Technol. 56, 133e140. Pinkston, K.E., Sedlak, D.L., 2004. Transformation of aromatic ether and amine-containing pharmaceuticals during chlorine disinfection. Environ. Sci. Technol. 38, 4019e4025. Richardson, S.D., 2008. Environmental mass spectrometry: emerging contaminants and current issues. Anal. Chem. 80, pp. 4373e4402.

Richardson, S.D., 2010. Environmental mass spectrometry: emerging contaminants and current issues. Anal. Chem. 82, 4742e4774. Sacher, F., Lange, F.T., Brauch, H.-J., Blankenhorn, I., 2001. Pharmaceuticals in groundwaters. Analytical methods and results of a monitoring program in Baden-Wurttemberg, Germany. J. Chromatogr., A 938, 199e210. Snyder, S.A., Keith, T.L., Verbrugge, D.A., Snyder, E.M., Gross, T.S., Kannan, K., Giesy, J.P., 1999. Analytical methods for detection of selected estrogenic compounds in aqueous mixtures. Environ. Sci. Technol. 33, 2814e2820. Snyder, S.A., Kelly, K.L., Grange, A.H., Sovocool, G.W., Snyder, E.M., Giesy, J.P., 2001a. Pharmaceuticals and personal care products in the waters of Lake Mead, Nevada. ACS Symp. Ser. 791, 116e139. Snyder, S.A., Villeneuve, D.L., Snyder, E.M., Giesy, J.P., 2001b. Identification and quantification of estrogen receptor agonists in wastewater effluents. Environ. Sci. Technol. 35, 3620e3625. Snyder, S.A., Wert, E.C., Lei, H.X., Westerhoff, P., Yoon, Y., 2007. Removal of EDCs and Pharmaceuticals in Drinking and Resuse Treatment Process. Awwa Res. Foundation, Dever, CO, pp. 118e119. Stackelberg, P.E., Gibs, J., Frulong, E.T., Meyer, M.T., Zaugg, S.D., Lippincott, R.L., 2007. Efficiency of conventional drinkingwater-treatment process in removal of pharmaceuticals and other organic compounds. Sci. Total Environ. 377, 255e272. Ternes, T., Bonerz, M., Schmidt, T., 2001. Determination of neutral pharmaceuticals in wastewater and rivers by liquid chromatography-electrospray tandem mass spectrometry. J. Chromatogr., A 938, 175e185. Ternes, T.A., 2001. Analytical methods for the determination of pharmaceuticals in aqueous environmental samples. TrAC, Trends Anal. Chem. 20, 419e434. Ternes, T.A., Hirsch, R., Mueller, J., Haberer, K., 1998. Methods for the determination of neutral drugs as well as betablockers and beta 2-sympathomimetics in aqueous matrixes using GC/MS and LC/MS/MS. Fresenius’ J. Anal. Chem. 362, 329e340. Vanderford, B.J., Snyder, S.A., 2006. Analysis of pharmaceuticals in water by isotope dilution liquid chromatography/tandem mass spectrometry. Environ. Sci. Technol. 40, 7312e7320. Westerhoff, P., Yoon, Y., Snyder, S., Wert, E., 2005. Fate of endocrine-disruptor, pharmaceutical, and personal care product chemicals during simulated drinking water treatment process. Environ. Sci. Technol. 39, 6649e6663. Ye, Z., Weinberg, H.S., Meyer, M.T., 2007. Trace analysis of trimethoprim and sulfonamide, macrolide, quinolone, and tetracycline antibiotics in chlorinated drinking water using liquid chromatography electrospray tandem mass spectrometry. Anal. Chem. 79, 1135e1144. Yu, C.-P., Chu, K.-H., 2009. Occurrence of pharmaceuticals and personal care products along the West Prong Little Pigeon River in east Tennessee, USA. Chemosphere 75, 1281e1286.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 2 9 e1 8 3 7

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Degradation of acetaminophen by Delftia tsuruhatensis and Pseudomonas aeruginosa in a membrane bioreactor Bart De Gusseme a, Lynn Vanhaecke b, Willy Verstraete a, Nico Boon a,* a

Laboratory of Microbial Ecology and Technology (LabMET), Faculty of Bioscience Engineering, Ghent University, Coupure Links 653, B-9000 Gent, Belgium b Laboratory of Chemical Analysis, Faculty of Veterinary Medicine, Ghent University, Salisburylaan 133, B-9820 Merelbeke, Belgium

article info

abstract

Article history:

The incidence and fate of pharmaceuticals in the water cycle impose a growing concern for

Received 9 September 2010

the future reuse of treated water. Because of the recurrent global use of drugs such as

Received in revised form

Acetaminophen (APAP), an analgesic and antipyretic drug, they are often detected in

17 November 2010

wastewater treatment plant (WWTP) effluents, receiving surface waters and drinking water

Accepted 25 November 2010

resources. In this study, the removal of APAP has been demonstrated in a membrane

Available online 3 December 2010

bioreactor (MBR) fed with APAP as the sole carbon source. After 16 days of operation, at a hydraulic retention time (HRT) of 5 days, more than 99.9% removal was obtained when

Keywords:

supplying a synthetic WWTP effluent with 100 mg APAP L
1. Batch experiments indicated no

Biodegradation

sorption of APAP to the biomass, no influence of the WWTP effluent matrix, and the capa-

Effluent polishing

bility of the microbial consortium to remove APAP at environmentally relevant concentra-

Emerging contaminants

tions (8.3 mg APAP L
1). Incubation with allylthiourea, an ammonia monooxygenase

Micropollutants

inhibitor, demonstrated that the APAP removal was mainly associated with heterotrophic

Paracetamol

bacteria and not with the ammonia-oxidizing bacteria. Two APAP degrading strains were

Sewage treatment

isolated from the MBR biomass and identified as Delftia tsuruhatensis and Pseudomonas aeruginosa. During incubation of the isolates, hydroquinone e a potentially toxic transformation product e was temporarily formed but further degraded and/or metabolized. These results suggest that the specific enrichment of a microbial consortium in an MBR operated at a high sludge age might be a promising strategy for post-treatment of WWTP effluents containing pharmaceuticals. ª 2010 Elsevier Ltd. All rights reserved.

1.

Introduction

In areas with growing urban populations, the reuse of treated water is considered to be the key solution for a sustainable water cycle management in the future (Verstraete et al., 2009). With regard to the recycling of wastewater, there is a growing concern about the fate of biologically active compounds such as pharmaceuticals since a large fraction of these organic

molecules and/or metabolites thereof reach wastewater treatment plants (WWTPs) after human consumption and excretion (Castiglioni et al., 2006). WWTPs have been identified as a major environmental source of these compounds because they are currently not designed for the elimination of micropollutants (Clara et al., 2005b). The presence of persistent pharmaceuticals in WWTP effluents and the receiving surface waters has been frequently reported (Daughton and

* Corresponding author. Tel.: þ32 9 264 59 76; fax: þ32 9 264 62 48. E-mail address: [email protected]. (N. Boon). URL: http://www.labmet.ugent.be. 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.11.040

1830

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 2 9 e1 8 3 7

Ternes, 1999; Heberer et al., 2002; Jones et al., 2007; Joss et al., 2005; Ternes, 1998). Moreover, because of their polar structure, several pharmaceuticals are not removed through passage in the subsoil, and may reach the groundwaters, which are the major source of drinking water (Heberer et al., 1997). Although the reported levels are much lower than those applied for therapeutic use, the related potential human health effects associated with chronic exposure to trace levels of these compounds are still poorly known, and cannot be discarded with respect to water reuse for drinking water purpose (Ku¨mmerer, 2001). Acetaminophen (¼paracetamol, abbreviated as APAP) is a heavily used analgesic and antipyretic drug all over the world. APAP is ranked as one of the top three drugs prescribed in England, and is one of the top 200 prescriptions in the US (Sebastine and Wakeman, 2003; Zhang et al., 2008). Muir et al. (1997) have demonstrated that 58e68% of APAP is excreted from the body during therapeutic use. The reported removals in WWTPs are varying from almost complete to 86% in municipal and 80% in hospital WWTPs, respectively (Gomez et al., 2007; Jones et al., 2007; Radjenovic et al., 2009; Rosal et al., 2010; Sim et al., 2010; Stackelberg et al., 2004). Yet, the removal is not complete since concentrations ranging from several hundred nanograms up to 11.3 mg L
1 have been found in European WWTP effluents (Jones et al., 2007; Rabiet et al., 2006; Ternes, 1998). In natural waters, up to 10 mg L
1 has been reported in the US, and even more than 65 mg L
1 in the Tyne River, UK (Kolpin et al., 2002; Roberts and Thomas, 2006). Rabiet et al. (2006) detected 211 ng L
1 in a well supplying drinking water. Moreover, APAP is known to exhibit virtually no sorption and no retardation in aquifer sand studies (Lorphensri et al., 2007). In our previous work, we have demonstrated the removal of the synthetic estrogen 17a-ethinylestradiol (EE2) by a nitrifier enrichment culture (NEC) (De Gusseme et al., 2009b). Other studies also report on the removal of estrogens and other phenolic micropollutants by nitrifying activated sludge (Kim et al., 2007; Shi et al., 2004; Vader et al., 2000). Especially ammonia-oxidizing bacteria (AOB) are considered to play an important role in the removal of these compounds by means of a cometabolism, since the enzyme ammonia monooxygenase (AMO) has a wide spectrum for the degradation of several substrates (Yi and Harper, 2007). However, partial degradation of pharmaceuticals was also observed in the presence of allylthiourea (ATU), an AMO inhibitor, suggesting the involvement of associated heterotrophic bacteria (Tran et al., 2009). Also Roh et al. (2009) reported the degradation of phenolic contaminants such as bisphenol A by nitrifying activated sludge during ATU addition. Yet, little is known about the autotrophic and/or heterotrophic degradation of APAP by such a NEC. The aim of this study was to examine the removal of APAP in a membrane bioreactor (MBR), inoculated with a NEC, and the role of the associated heterotrophic bacteria. By means of ATU addition, it was shown that the role of AOB in APAP removal was limited and two associated APAP degrading strains were isolated from the MBR biomass. In axenic batch incubation tests, the transformation products were examined by chromatographic analyses.

2.

Materials and methods

2.1.

Stock solutions, media and WWTP effluent

The minimal medium was based on Stanier et al. (1966) medium. Five mineral stock solutions containing macro- and micro-nutrients were prepared as previously described (De Gusseme et al., 2009b). Stock A was dosed at 10 mL L
1, stock B at 3 mL L
1, whereas stock C, D and E were dosed at 5 mL L
1, according to De Gusseme et al. (2009b). APAP, 4aminophenol and hydroquinone stock solutions (SigmaeAldrich, Bornem, Belgium) of 1 g L
1 were prepared in deionized water. WWTP effluent was sampled from the urban WWTP Ossemeersen (Ghent, Belgium), serving 175.000 IE. The levels of chemical oxygen demand (COD), total ammonia nitrogen (TAN), total organic nitrogen (TON), Kjeldahl nitrogen
(Kj-N), NO
3 eN and NO2 eN were 27, 2.2, 1.4, 3.3, 9.3 and 0.9 mg L
1, respectively.

2.2.

Nitrifying membrane bioreactor

A NEC was obtained from an industrially produced consortium grown on both ammonium and nitrite, and calcium carbonate as a carrier matrix and buffer (Grommen et al., 2002). An MBR was inoculated with 0.75 g VSS L
1 of the NEC. The MBR (Solis, the Netherlands) had a working volume of 20 L and contained 3 plate membranes (Kubota, Tokyo, Japan) with a pore size and total membrane area of 0.4 mm and 0.3 m2, respectively (Hennebel et al., 2009). The MBR was operated as previously described (De Gusseme et al., 2009b). The influent was prepared in a 30 L glass bottle containing tap water, supplemented with nutrients by dosing the five mineral stock solutions (AeE) (see above). Yet, lower amounts were added to simulate the characteristics of a WWTP effluent. Stock A was dosed at 3 mL L
1, whereas stocks C, D and E were dosed at 0.5 mL L
1. Stock B was added in different concentrations throughout the testing period, resulting in different NHþ 4 eN concentrations in the influent (Table 1). APAP was supplemented to the influent resulting in a final concentration of 100 mg L
1. During periods AeD, an HRT of 5 days was applied, resulting in a theoretical volumetric loading rate (BV) of 20 mg APAP L
1 d
1. In periods E, the HRT was 1 day, resulting in

Table 1 e Overview of the different periods of reactor operation with changes in HRT, and volumetric loading rate of ammonium (NHD 4 eN). For each period, the mean APAP volumetric loading rate, and removal efficiency measured in the effluent of the MBR are given. Concentrations are presented as mean ± standard deviation (n [ number of data points). APAP APAP n Period Time HRT NHþ 4 eN (days) (days) (mg L
1 d
1) (mg L
1 d
1) removal (%) A B C D E

1e15 16e30 31e45 46e60 61e75

5 5 5 5 1

4.0  0.2 1.7  0.3 1.2  0.1 0.2  0.1 1.2  0.3

22.9  21.6  24.0  20.7  105.7 

0.9 0.2 0.6 0.1 0.1

84.7  2.5 >99.9 >99.9 >99.9 >99.9

14 13 13 12 15

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 2 9 e1 8 3 7

a theoretical BV of 100 mg APAP L
1 d
1 (Table 1). Bulk and effluent samples were taken regularly for APAP analysis. Nitrification was examined by triplicate measurements of the

NHþ 4 eN, NO2 eN and NO3 eN concentration in influent and effluent.

2.3.

Batch incubation experiments

2.3.1.

Batch experiments in minimal medium

To examine APAP removal by the MBR biomass, a batch test was set up in triplicate in sterile 500 mL Erlenmeyer flasks containing 200 mL of minimal medium, supplemented with
1 1000 mg APAP L
1 and 52.5 mg NHþ 4 eN L . The biomass was harvested from the MBR, centrifuged (8041g, 8 min) and resuspended to a final concentration of 0.75 g VSS L
1. To examine the APAP removal at lower concentrations, similar experiments were conducted at 10 and 100 mg APAP L
1. To investigate the sorption of APAP to the biomass, a similar batch test was set up in triplicate by incubating heat-inactivated NEC biomass (121  C, 30 min) in 200 mL of minimal medium (0.75 g VSS L
1). A biomass-free control experiment was conducted in triplicate as well. The Erlenmeyer flasks were placed on a shaker (100 rpm) and incubated in the dark for 7 days at 28  C. Liquid samples for APAP and nitrification analysis were taken at regular intervals.

2.3.2.

Batch experiments in WWTP effluent

A similar set of batch incubation experiments with the MBR biomass was conducted in WWTP effluent. After heat-treatment of the effluent (121  C, 30 min), 1000 mg APAP L
1 was supplemented to it, but no other nutrients were added.

2.3.3. Batch experiments to evaluate the relation between APAP removal and nitrification A batch experiment was set up in three sterile Erlenmeyer flasks containing 200 mL of minimal medium and 0.75 g VSS L
1 of the MBR biomass. Fifty mg ATU L
1 was supplemented to specifically inhibit the enzyme AMO, responsible for the first step in the nitrification process (Hooper and Terry, 1973). In another batch experiment, 0.75 g VSS L
1 of biomass was incubated in minimal medium without addition of ammonium.

2.4.

Isolation of APAP degrading strains

Two APAP degrading strains were isolated from the MBR biomass after repeated supplementation with APAP. Therefore, six 500 mL Erlenmeyer flasks containing 200 mL minimal medium with 50 mg APAP L
1 were inoculated with 2 mL of MBR biomass, placed on a shaker (100 rpm) and incubated in the dark at 28  C. After complete removal of APAP in 5 days, 2 mL of the enriched biomass was resuspended in 200 mL fresh minimal medium (50 mg APAP L
1) every 5 days for another 20 days. Subsequently, the enriched biomass was plated on minimal medium containing 50 mg APAP L
1 and 15 g Noble Agar L
1, and incubated at 28  C for one week. The plates showed different colony morphologies. Nine different colony types were picked up and plated on LuriaeBertani (LB) medium. Finally, these isolates were subcultured in test tubes with liquid minimal medium containing 10 mg APAP L
1. Two isolates were able to degrade APAP.

1831

2.5. Phylogenetic identification of APAP degrading isolates Ten-fold dilutions of colonies of the two APAP degrading isolates were prepared. A PCR mixture was made according to Boon et al. (2000). Samples were amplified with a 9600 Thermal Cycler (PerkineElmer, Norwalk, CT), using primers P63f (El Fantroussi et al., 1999) and P1378r (Heuer et al., 1997) as follows: 94  C for 4 min, followed by 35 cycles of 94  C for 1 min, 60  C for 1 min, and 72  C for 2 min, with a final extension at 72  C for 10 min. DNA sequencing of the PCR fragments was carried out by AGOWA Genomics (Berlin, Germany). Homologies of the DNA sequences were searched with the Ribosomal Database Project Classifier (Wang et al., 2007). The 16S rRNA gene sequences of the isolates, Delftia tsuruhatensis BDG1 and Pseudomonas aeruginosa BDG2, were deposited in the EMBL database under the accession numbers FN996011 and FN996012, respectively, and are available at the Belgian Coordinated Collections of Microorganisms BCCM/ LMG (Ghent University, Ghent, Belgium) under accession number LMG25907 and LMG25908.

2.6.

Sample preparation and analytical methods

For chromatographic analysis of APAP and hydroquinone, 2 mL samples were filtered over a 0.22-mm filter (Millipore) in a glass HPLC vial and stored at 4  C in the dark prior to analysis. APAP was analyzed on a Dionex (Sunnyvale, CA) HPLC system with a P580 pump, TCC-100 column oven, UV-DAD detector (UVD340S) and Chromeleon 6.8 software. Separation was performed on a Luna C18 column (150 mm  4.6 mm, 5 mm, Phenomenex, Torrence, CA) with a guard column. Elution and detection were performed according to Zhang et al. (2008). The injection volume was 50 mL and the column temperature was fixed at 30  C. The limit of detection (LOD) was determined at 50 mg L
1, based on the criterion that the signal to noise ration (S/N) had to be at least 3. For the detection of APAP at lower concentrations, liquid chromatography coupled to tandem mass spectrometry (LCeMS/MS) was used. Chromatography was carried out on a Thermo Scientific (San Jose´, CA) Electron Accela UHPLC system comprising of a quaternary pump and autosampler, equipped with a Nucleodur C18 Pyramid column (50 mm  2.1 mm, 1.8 mm, MacheryeNagel, Bethlehem, PA). The column temperature was set at 35  C. Analytes were detected with a Thermo Scientific TSQ Vantage triple stage quadrupole mass spectrometer using electron spray ionization (ESI). Analytes were eluted at a flow rate of 0.3 mL min
1 using a gradient starting from 98% A (0.08% formic acid in ultra pure water) and 2% B (0.08% formic acid in acetonitrile) for 0.8 min, increasing to 65% B in 0.5 min, keeping at 65% B for 0.7 min, increasing to 100% B in 1 min and keeping at 100% B for another min, before returning to the initial conditions. The injection volume was 10 mL. The MS parameters were: a vaporizer temperature of 0  C, a capillary temperature of 270  C, a sheath gas flow of 25 units, an ion sweep gas flow of 2 units and an auxiliary gas flow of 5 units. The spray voltage positive polarity was 3500 V. Argon pressure in the collision cell (Q2) was set at 1.5 m Torr and the mass resolution at the first (Q1) and third (Q3) quadrupole was set at 0.7 Da at full width at half maximum (FWHM). Precursor

1832

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 2 9 e1 8 3 7

ion, S-lens RF amplitude, and collision energy (CE) in Q2 were optimized individually per compound (Table S1). Quantification and confirmation data were acquired in selected reaction monitoring (SRM) mode, the transitions followed are displayed in Table S1. Instrument control and data processing were carried out by means of Xcalibur Software (Thermo Electron, San Jose´, CA). The samples were prepared similar as described above and isobutcar was added as internal standard. The limit of quantification (LOQ) was determined at 100 ng L
1 as the lowest point of the calibration curve. For the determination of the VSS, COD, TAN, TON and Kj-N, 50 mL grab samples were analyzed according to standard methods (Greenberg et al., 1992). For the determination of

NHþ 4 eN, NO2 eN and NO3 eN, 10 mL samples were taken, filtered over a 0.22 mm filter (Millex, Millipore, Billerica, USA) and kept at 4  C for further analysis. NHþ 4 eN was determined colorimetrically and with Nessler reagent according to
Greenberg et al. (1992). NO
2 eN and NO3 eN were determined using a Methrom 761 Compact Ion Chromatograph (Methrom, Herisau, Switzerland), according to De Gusseme et al. (2009a).

3.

Results

3.1.

APAP removal in the MBR

To retain all biomass, the NEC was inoculated in an MBR system fed with APAP as the sole carbon source. During the first 60 days of operation (period A-D), a high HRT of 5 days was applied to allow adaptation of the bacteria to the APAP loading. The influent was spiked with 100 mg APAP L
1, resulting in an average loading rate of 20 mg APAP L
1 d
1 (Table 1). In the first 15 days (¼3 HRTs or period A), APAP was removed for 85% in the reactor. Afterwards, no APAP concentrations above the LOQ of 100 ng L
1 were detected in the MBR effluent. This was also the case for period E, during which an HRT of 1 d was applied with a concurrent increase in APAP loading rate to 105.7 mg APAP L
1 d
1 (Table 1). The nitrification process in the MBR was not complete (Table S2). During all operational periods, ammonium and some nitrite were still present in the effluent of the reactor. Yet, gradually decreasing the ammonium concentration from
1
1 in period A to 1.2 mg NHþ in period 19.8 mg NHþ 4 eN L 4 eN L D and E showed no influence on the APAP removal efficiency (Table 1). During operation, a decline in the VSS concentration from 0.75  0.04 g VSS L
1 to 0.15  0.06 g VSS L
1 was detected in the bulk of the reactor, indicating the establishment of a steady-state microbial community. The pH of the reactor remained constant at 7.0  0.3 and the dissolved oxygen concentration amounted to 8.2  0.5 mg O2 L
1.

3.2.

whereas no significant removal of APAP was detected in the biomass-free control or in the batch incubation experiment with heat-inactivated biomass (Fig. 1). During APAP removal, all ammonium was oxidized to nitrate. In the batch incubation experiment in WWTP effluent spiked with 1108.4  57.8 mg APAP L
1, no lag phase was observed and an instant APAP removal was detected (Fig. 2). After 24 h, the APAP concentration had decreased to 482.4  45.6 mg L
1 and after 72 h, no APAP was detected (<100 ng L
1). No extra ammonium was added to the WWTP effluent. The TAN and nitrite present in the effluent was completely oxidized to nitrate after 24 h. In the biomass-free control or in the batch incubation experiment with heat-inactivated biomass, no removal of APAP was observed. Incubation of the NEC in minimal medium with lower, more environmentally relevant APAP concentrations of 8.3  0.7 mg APAP L
1 and 82.7  10.0 mg APAP L
1, resulted in a complete removal (<100 ng L
1) after 24 h.

3.3.

Nitrification plays a minor role in the APAP removal

In order to examine the influence of nitrification on the removal of APAP, a batch experiment was set up by incubating the NEC in minimal medium to which ATU was added. The latter specifically inhibits AMO, which prevents on its turn the oxidation of ammonium. Concomitantly, no nitrite or nitrate production was observed. However, APAP removal still occurred (Fig. 3). After 24 h, the APAP concentration decreased from 1013.7  44.7 mg L
1 to 552.8  12.3 mg L
1 (45.5% removal). Yet, 156.1  7.4 mg APAP L
1 was still present after 72 h. Eventually, the APAP concentration was lower than the LOD after 120 h. On the contrary, when no ammonium was added to the minimal medium, the APAP concentration rapidly decreased from 1092.4  31.3 mg L
1 to below LOD after 24 h only (Fig. 3). No nitrite or nitrate was detected in the medium.

APAP removal in batch incubation experiments

To elucidate the biotic or abiotic nature of the APAP removal by the NEC, the biomass was harvested from the MBR and incubated in minimal medium with APAP as the sole carbon
1 source and 52.5 mg NHþ 4 eN L . After a lag phase of 6 h, the APAP concentration decreased from 1097.6  40.0 mg L
1 to 400.0  37.7 mg L
1 (64% removal) after 24 h (Fig. 1). After 72 h of incubation, the APAP removal was already more than 99%,

Fig. 1 e APAP concentration as a function of time during 7 days incubation of NEC biomass (0.75 g VSS LL1) in minimal medium supplemented with 1000 mg APAP LL1. To examine sorption of APAP to the biomass, heatinactivated biomass was incubated in a control experiment (0.75 g VSS LL1). Results from a biomass-free control in the minimal medium are also shown. Error bars represent the standard deviations of triplicate experiments.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 2 9 e1 8 3 7

Fig. 2 e APAP concentration as a function of time during 7 days incubation of NEC biomass (0.75 g VSS LL1) in WWTP effluent, spiked with 1000 mg APAP LL1. Heat-inactivated NEC (0.75 g VSS LL1) was used as a control to examine the APAP removal due to sorption. Results from a biomass-free control in the WWTP effluent are also presented. Error bars represent the standard deviations of triplicate experiments.

3.4. Isolation of APAP degrading bacteria from the MBR biomass Samples of the NEC biomass were repeatedly supplemented with APAP as described above. After five enrichment rounds, the biomass was subsequently plated on minimal medium containing 50 mg APAP L
1 and incubated at 28  C. After one week, different colonies were visible on the plates and the color of the agar had changed from white to brown (data not shown). Nine isolates were picked from the plates and plated again on LB medium to obtain single colonies. Finally, the purified isolates were inoculated in minimal medium containing 10 mg APAP L
1, and the concentration of APAP was followed as a function of time. In two test tubes, a brown coloration of the medium developed. In these tubes, a decrease of the APAP concentration was detected (data not shown), whereas no decline was observed in the other test tubes or in the noninoculated control. Further tests were conducted with the latter two APAP removing isolates. By means of DNA sequencing of the 16S rRNA gene, the two species showed the closest hit with D. tsuruhatensis (99.7% similarity with GenBank accession no. AY684785) and P. aeruginosa (100% similarity with GenBank accession no. AB109752). Repeated experiments with these two isolates have been carried out to quantify the removal of APAP in minimal medium. When a 1% inoculum was used, D. tsuruhatensis and P. aeruginosa were able to remove APAP in 48 h from 10.325  0.027 mg L
1 to 0.263  0.034 mg L
1 (97% removal) and 6.152  0.083 mg L
1 (40% removal), respectively.

3.5.

Detection of the transformation products of APAP

The samples taken in the latter experiments with the isolates, were further analyzed to examine the formation of certain

1833

degradation products. The chromatograms obtained upon HPLCeUV analysis, revealed the formation of intermediates, especially in the case of D. tsuruhatensis. In Fig. 4A, the chromatogram of a sample taken from a test tube after 24 h of inoculation with D. tsuruhatensis is depicted. Next to the APAP peak at a retention time (tR) of 2.84 min (illustrated by the chromatogram of an APAP standard in Fig. 4B), an additional peak at tR 3.46 min was observed. During incubation, this additional peak increased and decreased again over time with a maximum peak area at after 24 h. Based on the tR and UV absorption maxima of a reference standard, this second peak was identified as hydroquinone. Its reference chromatogram at 243 nm is depicted in Fig. 4C, and the spectrum taken at tR 3.42 min between 200.0 nm and 595.5 nm revealed an absorption maximum for hydroquinone at 245.5 nm in both the samples and the standards (Fig. 4D). After 144 h of incubation, no UV-absorbing transformation products were detected any longer. In the biomass-free control tube, no removal of APAP or formation of intermediates was observed.

4.

Discussion

4.1.

Continuous removal of APAP in an MBR

In this study, a nitrifying MBR was successfully applied for the continuous removal of APAP. Batch experiments with heatinactivated biomass demonstrated a negligible biosorption of APAP, which was previously reported as well (Jones et al., 2007). As a consequence, the removal of APAP can be explained by microbiological degradation. In the first 15 days

Fig. 3 e To examine the role of nitrification in the degradation of APAP, the NEC biomass was incubated for 7 days (0.75 g VSS LL1) in minimal medium, spiked with 1000 mg APAP LL1. During a first experiment 50 mg ATU LL1 was added to the minimal medium L1 to inhibit nitrification containing 52.5 mg NHD 4 eN L (“without nitrification”). During a second experiment, no ammonium was supplemented to the medium (“without ammonium”). The APAP concentration is presented as a function of time for both experiments. Standard deviations of triplicate experiments are presented by error bars but are sometimes smaller than the depicted symbols.

1834

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 2 9 e1 8 3 7

Fig. 4 e Chromatograms obtained by HPLCeUV analysis of (A) a sample with D. tsuruhatensis after 24 h incubation in minimal medium with 10 mg APAP LL1; (B) a standard of 1 mg APAP LL1 (tR [ 2.84 min); (C) a standard of 1 mg hydroquinone LL1 (tR [ 3.46 min); and (D) the UV spectrum of the latter standard taken at tR [ 3.42 min, showing an absorption maximum at 245.5 nm.

of operation, the APAP removal efficiency was on average 85%. Most probably, an adaptation period of 3 HRTs was required to allow enrichment of the MBR biomass into a microbial consortium that is able to deal with the APAP loading. From period B onwards, more than 99.9% degradation of APAP was achieved at an HRT of 5 days. Lowering the HRT to 1 day did not affect the APAP degradation efficiency. Using a similar Kubota MBR with two plate membranes, Radjenovic et al. (2007) have also observed complete APAP degradation during nitrification at an HRT of 14 h. These facts point to the major advantage of membrane technology, which makes the HRT independent from the sludge retention time (SRT). Improved removal of pharmaceuticals with increasing SRTs was indeed observed by other researchers (Clara et al., 2005a). It appears that high SRTs allow the development of a specialized community, which may be beneficial for the degradation of low concentrations of pharmaceuticals (Batt et al., 2006). In our previous work EE2 removal in a nitrifying MBR was addressed, with the estrogen as the sole carbon source (De Gusseme et al., 2009b). An increasing EE2 removal efficiency was observed when lowering of the ammonium influent concentration until an optimal removal efficiency was
1 reached at about 1 mg NHþ 4 eN L . In case of APAP however, varying the ammonium influent concentrations did not affect

the degradation efficiency. APAP was completely removed in
1 the MBR at concentrations as low as 1.2 mg NHþ 4 eN L , even at an HRT of 1 d. Moreover, complete removal was also observed in the batch experiment without addition of ammonium. Consequently, a low NHþ 4 eN input to the MBR system is sufficient to maintain the continuous degradation of APAP. This might be very advantageous if one wants to apply nitrifying MBRs for effluent post-treatment because the concomitant nitrate concentration in the final effluent will be low as well. In addition, the batch experiments with WWTP effluent showed that complete APAP degradation by the NEC was possible without additional ammonium and that the matrix did not affect the degradation.

4.2.

Role of nitrification in degradation of APAP

Further batch experiments were undertaken to examine the role of AOB in the degradation of APAP because AOB are capable of cometabolizing a variety of aromatic contaminants such as estrogens (Shi et al., 2004), aniline (Keener and Arp, 1994) and naphthalene (Chang et al., 2002). Yi and Harper (2007) demonstrated the cometabolically mediated biotransformation of EE2 by the AMO enzyme and Shi et al. (2004) suggested the involvement of other (heterotrophic) microorganisms present

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 2 9 e1 8 3 7

in nitrifying sludge, which may be responsible for the further degradation of intermediates. In this study, addition of ATU resulted in a clear inhibition of nitrification while APAP was still degraded at the same degree. Tran et al. (2009) have also observed a high degradation of ibuprofen and partial degradation of other pharmaceuticals in the presence of ATU. They suggested that this may be due to the activity of heterotrophic microorganisms in the NEC. Another possible APAP removal pathway by nitrifying activated sludge is nitration of the compound since AOB are known to produce nitric oxide, a precursor of reactive nitrogen species (RNS), and nitrating agents such as nitrogen dioxide radicals (Lipschultz et al., 1981). Recently, Chiron et al. (2010) have demonstrated the formation of nitro-APAP derivates and suggested that nitrifying bacteria may indeed play a role in this transformation process through release of RNS. However, they showed that the production of 3-nitro-APAP would only account for a few percent of the total transformation rate of APAP in WWTPs. In our batch incubation experiments conducted without ammonium, a rapid degradation of APAP has been observed, while no nitrite or nitrate was detected. These results suggest that APAP nitration is not the main degradation pathway in both the MBR and the batch experiments and reflect once more the importance of the activity of heterotrophic bacteria in the NEC.

4.3.

Heterotrophic APAP degradation by two isolates

By selective enrichment of the MBR biomass on agar plates with APAP, two bacterial strains, D. tsuruhatensis and P. aeruginosa, were isolated and identified as APAP degrading strains. To the best of our knowledge, this is the first study to report on the isolation of bacteria that can grow on APAP as sole carbon source. D. tsuruhatensis has been shown capable of degrading other aromatic compounds such as aniline, chloroaniline, catechol and p-hydroxybenzoic acid through ring cleavage (Sheludehenko et al., 2005; Zhang et al., 2010). Recently, Rangel-Garcia et al. (2010) have demonstrated the degradation of toluene, benzene and phenol by P. aeruginosa as well. It should be noted that these organisms were isolated by selective enrichment in a medium containing 50 mg APAP L
1 while the MBR system was fed with 100 mg APAP L
1. It could be possible that the two pure cultures acquired through this isolation strategy do not represent all APAP-degrading bacteria in the reactor since the high concentration of the target compound may favor fast-growing species during enrichment (Ferrari et al., 2005). For example, Gu et al. (2010) have isolated three nonylphenol ethoxylate(NPEO)-degrading species supplying 50 mg NPEO L
1 while only one of the tree species could be isolated when 500 mg NPEO L
1 was provided. Moreover, cultivation was subsequently used to purify the APAP-degrading colonies but it is known that not all bacteria can be recovered on agar plates. Culture-dependent methodologies are therefore not always reliable to determine the relationship between the function of a microbial community and the phylogeny of the organisms responsible for it. For example Manefield et al. (2002) have demonstrated that phenol degradation in an aerobic bioreactor was dominated by a member of the Thauera genus using 13C-labeled RNA analyses, which was in contradiction with the findings based

1835

on more conventional culture depend methods. The efficient degradation by the two isolated organisms in this study thus demands further confirmation by in situ verification of their role in APAP degradation.

4.4. Hydroquinone as temporary transformation product During APAP degradation by the two isolates a brown coloration was observed, which might be related to the accumulation of certain transformation products, such as polymerization products of catechol (Boon et al., 2001). Chromatographic analyses revealed the production of different UV-absorbing products, of which the most important one was identified as hydroquinone. Also during photocatalytic degradation of APAP, hydroquinone has been identified as the main transformation product (Yang et al., 2008; Zhang et al., 2008) and shown to be in equilibrium with 1,4-benzoquinone (Richard and Boule, 1994). This metabolite might impose an environmental concern since 1,4-benzoquinone is a benzene metabolite exerting genotoxic and mutagenic effects (Bedner and Maccrehan, 2006; Snyder, 2000). In our batch incubation experiments with D. tsuruhatensis however, the transformation products were further degraded since none of them were detected with UV after 144 h. This suggests that they have lost the chromophoric structure present in the original APAP molecule, probably by oxidative cleavage of the aromatic ring (Bedner and Maccrehan, 2006). In contrast to the axenic batch experiments, no formation of hydroquinone was detected in the batch experiments or during the continuous experiments with the MBR. This suggests that other bacteria (for example AOB) were present in the MBR biomass, which assisted in the further degradation of hydroquinone. The same phenomenon was reported for linuron degradation by Variovax sp. WDL1 (Dejonghe et al., 2003). WDL1 can convert linuron first into the intermediate 3,4-dichloroaniline (3,4-DCA), which transiently accumulated in the medium but was further degraded completely in 12 days by the same strain. Yet, if WDL1 was co-inoculated with other bacteria isolated from the same linuron-degrading culture, 3,4-DCA was completely removed after 2 days by a synergistic interaction between the species. In this point of view, microbiological degradation and mineralization by polycultures can be advantageous since no toxic byproducts will be produced, thus reducing the environmental impact of these treated waters on the receiving surface waters and drinking water supplies.

5.

Conclusions

 Microbiological degradation of APAP was studied using an MBR, inoculated with an enriched nitrifying culture. The MBR biomass proved capable of removing APAP continuously at an HRT of 1 day, in a synthetic minimal medium with 100 mg APAP L
1. Batch incubation experiments confirmed that APAP removal was not the result of biosorption to the biomass. A more than 99.9% removal was achieved after an adaptation period, suggesting that MBR systems operated at a high sludge age allow the development of a more specialized community to deal with certain micropollutants.

1836

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 2 9 e1 8 3 7

 Further experiments indicated that there was no influence of the matrix of a WWTP effluent on the APAP degradation, and that the biomass was capable of degrading APAP at environmentally relevant concentrations. By inhibiting nitrification, the importance of the associated heterotrophic bacteria has been demonstrated. Two new APAP degrading strains have been isolated from the microbial consortium: D. tsuruhatensis and P. aeruginosa. To the best of our knowledge, this is the first study to report on bacterial strains that can use APAP as sole carbon source. Chromatographic analyses suggested that all toxic transformation products are further degraded and/or metabolized, thus decreasing the environmental impact of the treated water.

Acknowledgements Bart De Gusseme is supported by a PhD grant (Aspirant) from the Research Foundation e Flanders (Fonds voor Wetenschappelijk Onderzoek (FWO)-Vlaanderen). L. Vanhaecke is a postdoctoral fellow with the FWO. This work was supported by the EU Biotreat project (Contract number 266039; call FP7KBBE-2010.3.5.01). We gratefully thank Tim Baetens, Tim Lacoere, Dirk Stockx and Klaas Wille for their technical support. We acknowledge Hayde´e De Clippeleir, Joachim Desloover, Suzanne Read and Pieter Verhagen for critically reviewing this manuscript and the helpful suggestions.

Appendix. Supplementary material Supplementary data related to this article can be found online at doi:10.1016/j.watres.2010.11.040.

references

Batt, A.L., Kim, S., Aga, D.S., 2006. Enhanced biodegradation of iopromide and trimethoprim in nitrifying activated sludge. Environmental Science & Technology 40 (23), 7367e7373. Bedner, M., Maccrehan, W.A., 2006. Transformation of acetaminophen by chlorination produces the toxicants 1,4benzoquinone and N-acetyl-p-benzoquinone imine. Environmental Science & Technology 40 (2), 516e522. Boon, N., Goris, J., De Vos, P., Verstraete, W., Top, E.M., 2000. Bioaugmentation of activated sludge by an indigenous 3chloroaniline-degrading Comamonas testosteroni strain, I2gfp. Applied and Environmental Microbiology 66 (7), 2906e2913. Boon, N., Goris, J., De Vos, P., Verstraete, W., Top, E.M., 2001. Genetic diversity among 3-chloroaniline- and anilinedegrading strains of the Comamonadaceae. Applied and Environmental Microbiology 67 (3), 1107e1115. Castiglioni, S., Zuccato, E., Crisci, E., Chiabrando, C., Fanelli, R., Bagnati, R., 2006. Identification and measurement of illicit drugs and their metabolites in urban wastewater by liquid chromatographyetandem mass spectrometry. Analytical Chemistry 78 (24), 8421e8429. Chang, S.W., Hyman, M.R., Williamson, K.J., 2002. Cooxidation of naphthalene and other polycyclic aromatic hydrocarbons by the nitrifying bacterium, Nitrosomonas europaea. Biodegradation 13 (6), 373e381.

Chiron, S., Gomez, E., Fenet, H., 2010. Nitration processes of acetaminophen in nitrifying activated sludge. Environmental Science & Technology 44 (1), 284e289. Clara, M., Kreuzinger, N., Strenn, B., Gans, O., Kroiss, H., 2005a. The solids retention time e a suitable design parameter to evaluate the capacity of wastewater treatment plants to remove micropollutants. Water Research 39 (1), 97e106. Clara, M., Strenn, B., Gans, O., Martinez, E., Kreuzinger, N., Kroiss, H., 2005b. Removal of selected pharmaceuticals, fragrances and endocrine disrupting compounds in a membrane bioreactor and conventional wastewater treatment plants. Water Research 39 (19), 4797e4807. Daughton, C.G., Ternes, T.A., 1999. Pharmaceuticals and personal care products in the environment: agents of subtle change? Environmental Health Perspectives 107, 907e938. De Gusseme, B., De Schryver, P., De Cooman, M., Verbeken, K., Boeckx, P., Verstraete, W., Boon, N., 2009a. Nitrate-reducing, sulfide-oxidizing bacteria as microbial oxidants for rapid biological sulfide removal. FEMS Microbiology Ecology 67 (1), 151e161. De Gusseme, B., Pycke, B., Hennebel, T., Marcoen, A., Vlaeminck, S.E., Noppe, H., Boon, N., Verstraete, W., 2009b. Biological removal of 17 alpha-ethinylestradiol by a nitrifier enrichment culture in a membrane bioreactor. Water Research 43 (9), 2493e2503. Dejonghe, W., Berteloot, E., Goris, J., Boon, N., Crul, K., Maertens, S., Hofte, M., De Vos, P., Verstraete, W., Top, E.M., 2003. Synergistic degradation of linuron by a bacterial consortium and isolation of a single linuron-degrading Variovorax strain. Applied and Environmental Microbiology 69 (3), 1532e1541. El Fantroussi, S., Verschuere, L., Verstraete, W., Top, E.M., 1999. Effect of phenylurea herbicides on soil microbial communities estimated by analysis of 16S rRNA gene fingerprints and community-level physiological profiles. Applied and Environmental Microbiology 65 (3), 982e988. Ferrari, B.C., Binnerup, S.J., Gillings, M., 2005. Microcolony cultivation on a soil substrate membrane system selects for previously uncultured soil bacteria. Applied and Environmental Microbiology 71 (12), 8714e8720. Gomez, M.J., Bueno, M.J.M., Lacorte, S., Fernandez-Alba, A.R., Aguera, A., 2007. Pilot survey monitoring pharmaceuticals and related compounds in a sewage treatment plant located on the Mediterranean coast. Chemosphere 66 (6), 993e1002. Greenberg, A.E., Clesceri, L.S., Eaton, A.D., 1992. Standard methods for the examination of waste and wastewater. American Public Health Society, Washington DC. Grommen, R., Van Hauteghem, I., Van Wambeke, M., Verstraete, W., 2002. An improved nitrifying enrichment to remove ammonium and nitrite from freshwater aquaria systems. Aquaculture 211 (1e4), 115e124. Gu, X., Zhang, Y., Zhang, J., Yang, M., Tamaki, H., Kamagata, Y., Li, D., 2010. Isolation of phylogenetically diverse nonylphenol ethoxylate-degrading bacteria and characterization of their corresponding biotransformation pathways. Chemosphere 80 (3), 216e222. Heberer, T., Dunnbier, U., Reilich, C., Stan, H.J., 1997. Detection of drugs and drug metabolites in ground water samples of a drinking water treatment plant. Fresenius Environmental Bulletin 6 (7e8), 438e443. Heberer, T., Reddersen, K., Mechlinski, A., 2002. From municipal sewage to drinking water: fate and removal of pharmaceutical residues in the aquatic environment in urban areas. Water Science and Technology 46 (3), 81e88. Hennebel, T., Simoen, H., De Windt, W., Verloo, M., Boon, N., Verstraete, W., 2009. Biocatalytic dechlorination of trichloroethylene with bio-palladium in a pilot-scale membrane reactor. Biotechnology and Bioengineering 102 (4), 995e1002.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 2 9 e1 8 3 7

Heuer, H., Krsek, M., Baker, P., Smalla, K., Wellington, E.M.H., 1997. Analysis of actinomycete communities by specific amplification of genes encoding 16S rRNA and gelelectrophoretic separation in denaturing gradients. Applied and Environmental Microbiology 63 (8), 3233e3241. Hooper, A.B., Terry, K.R., 1973. Specific inhibitors of ammonia oxidation in Nitrosomonas. Journal of Bacteriology 115 (2), 480e485. Jones, O.A.H., Green, P.G., Voulvoulis, N., Lester, J.N., 2007. Questioning the excessive use of advanced treatment to remove organic micropollutants from wastewater. Environmental Science & Technology 41 (14), 5085e5089. Joss, A., Keller, E., Alder, A.C., Gobel, A., McArdell, C.S., Ternes, T., Siegrist, H., 2005. Removal of pharmaceuticals and fragrances in biological wastewater treatment. Water Research 39 (14), 3139e3152. Keener, W.K., Arp, D.J., 1994. Transformations of aromaticcompounds by Nitrosomonas-europaea. Applied and Environmental Microbiology 60 (6), 1914e1920. Kim, J.Y., Ryu, K., Kim, E.J., Choe, W.S., Cha, G.C., Yoo, I.K., 2007. Degradation of bisphenol A and nonylphenol by nitrifying activated sludge. Process Biochemistry 42 (10), 1470e1474. Kolpin, D.W., Furlong, E.T., Meyer, M.T., Thurman, E.M., Zaugg, S.D., Barber, L.B., Buxton, H.T., 2002. Pharmaceuticals, hormones, and other organic wastewater contaminants in US streams, 1999e2000: a national reconnaissance. Environmental Science & Technology 36 (6), 1202e1211. Ku¨mmerer, K., 2001. In: Ku¨mmerer, K. (Ed.), Pharmaceuticals in the Environment: Source, Fate, Effects and Risks. Springer, Berlin, pp. 1e8. Lipschultz, F., Zafiriou, O.C., Wofsy, S.C., McElroy, M.B., Valois, F. W., Watson, S.W., 1981. Production of NO and N2O by soil nitrifying bacteria. Nature 294 (5842), 641e643. Lorphensri, O., Sabatini, D.A., Kibbey, T.C.G., Osathaphan, K., Saiwan, C., 2007. Sorption and transport of acetaminophen, 17 alpha-ethynyl estradiol, nalidixic acid with low organic content aquifer sand. Water Research 41 (10), 2180e2188. Manefield, M., Whiteley, A.S., Griffiths, R.I., Bailey, M.J., 2002. RNA stable isotope probing, a novel means of linking microbial community function to phylogeny. Applied and Environmental Microbiology 68 (11), 5367e5373. Muir, N., Nichols, J.D., Stillings, M.R., Sykes, J., 1997. Comparative bioavailability of aspirin and paracetamol following single dose administration of soluble and plain tablets. Current Medical Research and Opinion 13 (9), 491e500. Rabiet, M., Togola, A., Brissaud, F., Seidel, J.L., Budzinski, H., Elbaz-Poulichet, F., 2006. Consequences of treated water recycling as regards pharmaceuticals and drugs in surface and ground waters of a medium-sized Mediterranean catchment. Environmental Science & Technology 40 (17), 5282e5288. Radjenovic, J., Petrovic, M., Barcelo, D., 2007. Analysis of pharmaceuticals in wastewater and removal using a membrane bioreactor. Analytical and Bioanalytical Chemistry 387 (4), 1365e1377. Radjenovic, J., Petrovic, M., Barcelo, D., 2009. Fate and distribution of pharmaceuticals in wastewater and sewage sludge of the conventional activated sludge (CAS) and advanced membrane bioreactor (MBR) treatment. Water Research 43 (3), 831e841. Rangel-Garcia, M.D., Rodriguez-Martinez, J., Garza-Garcia, Y., Martinez-Hernandez, J.L., 2010. Optimization of iron ions to eliminate pyocyanine in the degradation reaction of toluene, benzene and phenol by Pseudomonas aeruginosa. Agrociencia 44 (2), 235e247. Richard, C., Boule, P., 1994. Photocatalytic oxidation of phenolic derivatives e influence of OH-center-dot and Hþ on the distribution of products. New Journal of Chemistry 18 (4), 547e552. Roberts, P.H., Thomas, K.V., 2006. The occurrence of selected pharmaceuticals in wastewater effluent and surface waters of

1837

the lower Tyne catchment. Science of the Total Environment 356 (1e3), 143e153. Roh, H., Subramanya, N., Zhao, F.M., Yu, C.P., Sandt, J., Chu, K.H., 2009. Biodegradation potential of wastewater micropollutants by ammonia-oxidizing bacteria. Chemosphere 77 (8), 1084e1089. Rosal, R., Rodriguez, A., Perdigon-Melon, J.A., Petre, A., GarciaCalvo, E., Gomez, M.J., Aguera, A., Fernandez-Alba, A.R., 2010. Occurrence of emerging pollutants in urban wastewater and their removal through biological treatment followed by ozonation. Water Research 44 (2), 578e588. Sebastine, I.M., Wakeman, R.J., 2003. Consumption and environmental hazards of pharmaceutical substances in the UK. Process Safety and Environmental Protection 81 (B4), 229e235. Sheludehenko, A.S., Kolomytseva, M.P., Travkin, V.A., Akimov, V. N., Golovleva, L.A., 2005. Degradation of aniline by Delftia tsuruhatensis 14S in batch and continuous processes. Applied Biochemistry and Microbiology 41 (5), 465e468. Shi, J., Fujisawa, S., Nakai, S., Hosomi, M., 2004. Biodegradation of natural and synthetic estrogens by nitrifying activated sludge and ammonia-oxidizing bacterium Nitrosomonas europaea. Water Research 38 (9), 2323e2330. Sim, W.J., Lee, J.W., Oh, J.E., 2010. Occurrence and fate of pharmaceuticals in wastewater treatment plants and rivers in Korea. Environmental Pollution 158 (5), 1938e1947. Snyder, R., 2000. Overview of the toxicology of benzene. Journal of Toxicology and Environmental Health-Part A 61 (5e6), 339e346. Stackelberg, P.E., Furlong, E.T., Meyer, M.T., Zaugg, S.D., Henderson, A.K., Reissman, D.B., 2004. Persistence of pharmaceutical compounds and other organic wastewater contaminants in a conventional drinking-water treatment plant. Science of the Total Environment 329 (1e3), 99e113. Stanier, R.Y., Palleron, N.J., Doudorof, M., 1966. Aerobic Pseudomonads e a taxonomic study. Journal of General Microbiology 43 (2), 159e171. Ternes, T.A., 1998. Occurrence of drugs in German sewage treatment plants and rivers. Water Research 32 (11), 3245e3260. Tran, N.H., Urase, T., Kusakabe, O., 2009. The characteristics of enriched nitrifier culture in the degradation of selected pharmaceutically active compounds. Journal of Hazardous Materials 171 (1e3), 1051e1057. Vader, J.S., van Ginkel, C.G., Sperling, F., de Jong, J., de Boer, W., de Graaf, J.S., van der Most, M., Stokman, P.G.W., 2000. Degradation of ethinyl estradiol by nitrifying activated sludge. Chemosphere 41 (8), 1239e1243. Verstraete, W., de Caveye, P.V., Diamantis, V., 2009. Maximum use of resources present in domestic “used water”. Bioresource Technology 100 (23), 5537e5545. Wang, Q., Garrity, G.M., Tiedje, J.M., Cole, J.R., 2007. Naive Bayesian classifier for rapid assignment of rRNA sequences into the new bacterial taxonomy. Applied and Environmental Microbiology 73 (16), 5261e5267. Yang, L., Yu, L.E., Ray, M.B., 2008. Degradation of paracetamol in aqueous solutions by TiO2 photocatalysis. Water Research 42 (13), 3480e3488. Yi, T., Harper, W.F., 2007. The link between nitrification and biotransformation of 17 alpha-ethinylestradiol. Environmental Science & Technology 41 (12), 4311e4316. Zhang, L.L., He, D., Chen, J.M., Liu, Y., 2010. Biodegradation of 2chloroaniline, 3-chloroaniline, and 4-chloroaniline by a novel strain Delftia tsuruhatensis H1. Journal of Hazardous Materials 179 (1e3), 875e882. Zhang, X., Wu, F., Wu, X.W., Chen, P.Y., Deng, N.S., 2008. Photodegradation of acetaminophen in TiO2 suspended solution. Journal of Hazardous Materials 157 (2e3), 300e307.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 3 8 e1 8 4 6

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Reactions of tetracycline antibiotics with chlorine dioxide and free chlorine Pei Wang a,b, Yi-Liang He a, Ching-Hua Huang b,* a b

School of Environmental Science and Engineering, Shanghai JiaoTong University, Shanghai 200240, China School of Civil and Environmental Engineering, Georgia Institute of Technology, Atlanta, GA 30332, USA

article info

abstract

Article history:

Tetracyclines (TCs) are a group of widely used antibiotics that have been frequently found

Received 13 September 2010

in the aquatic environment. The potential reactions of TCs with common water disinfec-

Received in revised form

tion oxidants such as chlorine dioxide (ClO2) and free available chlorine (FAC) have not

13 November 2010

been studied in depth and are the focus of this study. The oxidation kinetics of tetracycline,

Accepted 25 November 2010

oxytetracycline, chlorotetracycline and iso-chlorotetracycline by ClO2 and FAC are very

Available online 3 December 2010

rapid (with large apparent second-order rate constants kapp ¼ 2.24  105e1.26  106 M1 s1 with ClO2 and kapp ¼ 1.12  104e1.78  106 M1 s1 with FAC at pH 7.0) and highly

Keywords:

dependent on pH. Species-specific rate constants are obtained by kinetic modeling that

Antimicrobials

incorporates pH-speciation of TCs and the oxidants (for FAC), and reveal that TCs primarily

Emerging contaminants

react with ClO2 and FAC by their unprotonated dimethylamino group and deprotonated

Pharmaceuticals

phenolic-diketone group. The modest difference in reactivity among the four TCs toward

Water treatment

the oxidants is consistent with expectation and can be explained by structural influences

Water disinfection

on the two reactive moieties. Product evaluation shows that oxidation of TCs by ClO2 leads

Antibacterials

to (hydr)oxylation and breakage of TC molecules, while oxidation of TCs by FAC leads to chlorinated and (hydr)oxylated products without any substantial ring breakage. Results of this study indicate that rapid transformation of TCs by oxidants such as ClO2 and FAC under water and wastewater treatment conditions can be expected. ª 2010 Elsevier Ltd. All rights reserved.

1.

Introduction

Research in recent decade has shown that many pharmaceuticals and personal care products are prevalent in the aquatic environment due to increasing use of these chemicals in daily lives. Among those, tetracyclines (TCs) are a group of broad-spectrum antibiotics that have been used since the 1940s against a wide range of both Gram-negative and Grampositive bacteria (Brodersen et al., 2000). TCs have been used extensively as antibacterial agents in human and veterinary medicine and have been used at sub-therapeutic levels to prevent epidemics and increase the growth rate and weight gain in livestock and aquaculture animals (Gra¨slund and

Bengtsson, 2001). The widespread and long-term usage of TCs have resulted in the introduction of these compounds into the environment through various routes including hospital and municipal wastewater, manure from animal husbandry, and agricultural runoff. In recent years, TCs have been frequently detected at concentrations at 0.07e1.34 mg/L in surface water samples (Lindsey et al., 2001), 86e199 mg/kg in soils (Hamscher et al., 2002), 4.0 mg/kg in liquid manure (Hamscher et al., 2002), and 3 mg/L in farm lagoons (Zhu et al., 2001). As a result, the TC resistance genes have been found in waste lagoons and groundwater (Chee-Sanford et al., 2001). The frequent detection of TC residues and related microbial resistance in the environment poses threats to human

* Corresponding author. Tel.: þ1 404 8947694; fax: þ1 404 8948266. E-mail address: [email protected] (C.-H. Huang). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.11.039

1839

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 3 8 e1 8 4 6

EDTA HPLC DPD RC 4-CR UV/MSD LC/MS

Abbreviations and notations ClO2 FAC TCs TTC OTC CTC Iso-CTC

chlorine dioxide free available chlorine tetracyclines tetracycline oxytetracycline chlorotetracycline iso-chlorotetracycline

health and the ecosystem. Processes that can effectively remove or destroy TC residues in water sources and wastes are desirable to minimize human and environmental exposure. Previous studies have reported rapid reaction of tetracycline (TTC) with ozone that results in loss of antibacterial potency of the compound (Dodd et al., 2009). Significant removal of chlorotetracycline (CTC) and TTC in secondary wastewater effluent was observed by UV radiation with or without H2O2, with CTC more easily removed than TTC (Kim et al., 2009), Furthermore, TCs can undergo solar photodegradation catalyzed by semiconductor catalysts such as TiO2 and ZnO (Palominos et al., 2009). While various pharmaceutical contaminants have been shown to be reactive to common water disinfectants such as chlorine dioxide and free chlorine, such reactions have not yet been examined in detail for TCs and thus are the focus of this study in order to obtain a better understanding of the fate of TCs during water and wastewater treatment processes. Owing to its low cost, chlorine is the most used chemical oxidant for drinking water disinfection globally (Deborde and von Gunten, 2008), and is commonly applied to control pathogens in wastewater effluent. The combined concentration of hypochlorous acid (HOCl) and hypochloride (OCl) is termed free available chlorine (FAC) and plays an important role in the fate of pharmaceutical contaminants in water treatment processes. For example, prior studies have shown that various antibiotics such as sulfamethoxazole, trimethoprim, carbadox, amoxicillin and fluoroquinolones all react with FAC rapidly (Dodd and Huang, 2004, 2007; Dodd et al., 2005; Shah

ethylenediamine tetraacetic acid, ammonium salt high performance liquid chromatography N,N-diethyl-p-phenylenediamine resorcinol 4-chlororesorcinol ultraviolet/mass spectrometry detector liquid chromatography/mass spectrometry

et al., 2006; Acero et al., 2010). Results from the above studies also indicate that FAC oxidation of antibiotics commonly leads to formation of chlorinated by-products. As an oxidant, free chlorine (HOCl) typically acts as a two-electron electrophile attacking electron-rich sites of organic molecules with its Cl atom (Deborde and von Gunten, 2008). Chlorine dioxide (ClO2), while not as popular, has the advantages of comparable biocidal efficacy but less pHdependence and disinfection by-product (DBPs) formation potential compared to free chlorine (Rav-Acha, 1998). ClO2 is most commonly used in treating potable water or for the preoxidation of surface water before chlorination or monochloromination (Chen and Regli, 2002). Previous studies have reported oxidative transformation of several pharmaceutical contaminants by ClO2 for compounds such as estrogenic 17a-ethinylestradiol, analgesic diclofenac, antibiotic sulfamethoxazole, roxithromycin, b-lactams and fluoroquinolones (Huber et al., 2005; Navalon et al., 2008; Wang et al., 2010). ClO2 is a highly selective oxidant that typically undergoes an oneelectron oxidation process and transforms to chlorite ion (ClO 2 ). Thus, ClO2 is considered a stable radical that can remain in aqueous solution for long periods of time when protected from light and at low temperature (4  C) (Rav-Acha, 1998). As shown in Fig. 1, TC molecules contain connected ring systems (lettered A through D from right to left) with multiple ionizable functional groups that are associated with three macroscopic acid dissociation constants. TCs’ structures contain several electron-rich moieties such as dimethylamino group, phenolic group and conjugated double-bonds that are

pKa3 H 3C

8 9

N

C H3

HO

C H3 OH

5

4

D C

B

A

10

12

6

7

11

OH

pKa2

O

OH

C H3

H3C

H N+

OH

O

OH

TCH+

O

pKa1

O

H 3C C H3

HO

N

C H3

pKa1

Cl

OH

O

OH

OH

O

O

CH 3

HO

H3C

H N+

O

- H+ OH

O

OH

TC

O

O

CH 3 OH

OH

OH

OH

O

O

HO

CH 3

H3C

H N+

O

C H3

OH

HO

pKa3

-

O

- H+

NH2 OH

O

O

Iso-chlorotetracycline (iso-CTC)

Chlorotetracycline (CTC)

pKa2 NH2

OH

O

CH 3 -

N

O

NH 2

NH2 OH

H3C

CH 3

OH

Oxytetracycline (OTC)

C H3

N H2

Cl

OH

O

O

OH

OH

C H3

N

NH 2

2

Tetracycline (TTC)

HO

H3C OH

3

1

OH

CH 3

HO

CH3

H3C

N

CH 3 O-

- H+

NH2

NH 2 O-

O

OH

OH

O

O

TC-

Fig. 1 e Structures and speciation of tetracyclines investigated in this study.

O-

O

OH

TC2-

OH

O

O

1840

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 3 8 e1 8 4 6

likely to be susceptible to attacks by oxidants like ClO2 and FAC. The objectives of this study are to determine the reactivity of TCs toward ClO2 and FAC under conditions relevant in water treatment and to characterize the reaction kinetics, identify compounds’ reactive sites, evaluate oxidation products and provide mechanistic insight. Four TCs, TTC, oxytetracycline (OTC), CTC and iso-chlorotetracycline (iso-CTC) (Fig. 1), were selected in this study to evaluate the impact of structural variation on the oxidation of TCs. Among them, TTC, OTC and CTC are widely prescribed TC antibiotics, while iso-CTC is a common transformational isomer of CTC particularly at higher pH (Waller et al., 1952) and contains relatively low antibacterial effect (Halling-Sorensen et al., 2002).

2.

Experimental section

2.1.

Chemical reagents

Surface water sample was collected from the reservoir supplying source water to the drinking water treatment plant. Wastewater sample was collected after activated sludge secondary treatment and gravity filtration, but prior to chlorine-based disinfection process. The samples were vacuum-filtered through 0.5 mm glass-fiber filters upon arrival in the laboratory, stored at 4  C and used within two days for experiments. Available characteristics of these real water samples are listed below:  Surface water sample: pH 6.83, iron 0.048 mg/L, color 5 C.U., alkalinity 17 mg/L, dissolved oxygen 9 mg/L, calcium hardness 9 mg/L as CaCO3, total hardness 14 mg/L as CaCO3, turbidity 1.6 NTU, nitrate 0.3 mg/L as N, and chemical oxygen demand (COD) < 5 mg/L.  Wastewater sample: pH 7.35, total phosphate 0.12 mg/L, ammonia < 0.1 mg/L, TSS 6 mg/L, VSS 2 mg/L, nitrate 8.4 mg/ L as N, nitrite 0.4 mg/L as N, and COD around 30 mg/L.

2.3. TTC, OTC and CTC from Sigma and iso-CTC from Acros were at 90e98% purity and used directly without further purification. Stocks of TCs were prepared in methanol at 1.6 mM, protected from light, stored at 15  C, and used within a month of preparation. All other reagents (e.g., buffers, NaCl, acids, etc.) were obtained from Fisher Scientific, Acros or Aldrich at analytical grade. All solutions were prepared using reagent water (18.2 MU-cm at 25  C) from a Millipore Milli-Q Ultrapure Gradient A10 purification system. NaOCl was obtained from Fisher Scientific at w7% solution concentration. Free chlorine stock solutions were prepared at 100 mg/L as Cl2 and quantified by the standard iodometric titration method (APHA, 1998). The method by Rosenblatt et al. (Rosenblatt et al., 1963) was modified for lab-scale ClO2 generation. Briefly, 150 mL of 200 mM hydrochloric acid solution from a washing bottle was added gently by a peristaltic pump at 7.5e10 mL/min into another tall-form gas washing bottle through an inlet on the bottle wall. The latter bottle contained 500 mL of 400 mM sodium chlorite solution under constant stirring using a Teflon-coated magnetic stir bar and was sealed by a screw-cap with two outlets. Gentle air purged through this solution from one outlet on the cap carried the generated ClO2 out of the washing bottle through the other outlet. The ClO2carrying air was passed through a flake sodium chlorite-filled glass tube to convert the elemental chlorine, if present, in the gas stream to ClO2. The purified ClO2 was finally purged into an amber borosilicate bottle which contained 500 mL of chilled pure water and was immersed in an ice bath to maintain low temperature and minimize loss of ClO2. All the tubes and bottles during the ClO2 generation were covered by aluminum foil to block off light. The prepared stock at 15e20 mM as ClO2 measured by the N,N-diethyl-p-phenylenediamine (DPD) titration method specified in the Standard Methods (APHA, 1998) was protected from light and refrigerated at 4e5  C and was stable for experimental usage up to one month.

2.2.

Surface water and wastewater samples

Grab samples of surface water and wastewater effluent were collected from a municipal drinking water treatment plant and a wastewater treatment plant in the Southeastern United States.

Kinetic experiments

Batch experiments were used to study the reactions of TCs with ClO2 at low pH (<3.5) in 20-mL amber borosilicate bottles with screw caps and Teflon stoppers at room temperature (22  1  C). Reaction solutions with 5 mM of target TC and pH adjusted by appropriate amounts of HCl were constantly mixed using a Teflon-coated stir bar on a multi-position magnetic stir plate. Then, reaction was initiated by adding ClO2 stock that resulted in ClO2 concentration 10 times or higher than the initial TC concentration. The sample pH was monitored and did not vary by more than 0.15 at the end of the reaction. Aliquots were taken periodically from the reactor, immediately quenched by adding 0.1 mL of 10 mM sodium thiosulfate, and then analyzed by an 1100 Agilent highperformance liquid chromatography (HPLC) system with a Zorbax RX-C18 column (4.6  250 mm, 5 mm) and a diodearray UV/vis detector for the loss of target TC. The HPLC/UV method is similar to that described previously (Zhang et al., 2008; Chen and Huang, 2009). The competition kinetics method described in Shah et al. (Shah et al., 2006) was adopted to study the reactions of TCs with ClO2 at most pHs (>3.5) and the reactions of TCs with FAC in all pH range, because these reactions were too fast to follow by batch experiments. Resorcinol (RC) and 4-chlororesorcinol (4-CR) were chosen as the competitor in reactions with ClO2 and FAC, respectively. The sample pH was controlled by using 10 mM acetate (3.5 < pH < 5), phosphate (5 < pH < 8) or borate (pH > 8) buffers or was adjusted by adding pre-calculated amounts of HCl (3.0  pH < 3.5). In all experiments, the initial and final pH difference was less then 0.1. The mixing conditions and temperature were the same as the batch experiments. Reaction was initiated by adding varying sub-stoichiometric amounts of ClO2 or FAC into solutions containing 100 mM target TC and 100 mM RC or 4-CR. Sample aliquots were taken after the reaction was completed, added with thiosulfate (to be consistent with those in batch experiments), and analyzed by HPLC/UV to follow the losses of RC (or 4-CR) and TCs. The HPLC conditions for RC and 4-CR were described previously (Shah et al., 2006). Batch kinetic experiments were also conducted in real water matrices. For those, surface water and wastewater

1841

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 3 8 e1 8 4 6

effluent samples were spiked with 2 mM TTC and 1 mg/L (14.9 mM) ClO2 or 1 mg/L (19.2 mM) FAC. Sample aliquots were taken immediately and subsequently, quenched with 0.1 mL of 10 mM sodium thiosulfate, and then analyzed by HPLC to monitor TTC loss.

2.4.

LC/MS analysis

The mixtures of TC oxidation products were prepared by adding pre-determined amounts of ClO2 or FAC stock into a 10 mL solution that contained 400 mM of target TC and was buffered at pH 7.5 by 10 mM phosphate buffer. The initial concentration ratio of oxidant to TC was set at 1:2 to avoid excessive oxidation of products. The products were analyzed by an Agilent HPLC/DAD/MSD (1100 HPLC/G1956B MSD) system with a Zorbax SB-C18 column (2.1  150 mm, 5 mM), a diode-array UV/vis detector, and a mass spectrometer. Gradient elution was conducted with pure acetonitrile and 0.2% formic acid at a flow rate of 0.30 mL/min. Products were analyzed by electrospray ionization at positive mode (ESIþ) with fragmentor voltage of 120e220 V and mass scan range of m/z 50e500. The drying gas was at 10 L/min at 350  C, the nebulizer pressure 25 psig, and the capillary voltage 4000 V. The mass resolution in the scan mode was at the increment of 0.1 m/z. The MSD was regularly autotuned and checked for mass accuracy and sensitivity, and showed mass error of 0.03% or less for the typical molecular weight range of 450 in this study.

by the standard equilibrium distribution equation for a triprotic acid (eq (2)). X


 ½TCtotal ¼ TCHþ þ ½TC þ TC þ TC2 ¼ ai ½TCtotal


þ 3
2
 Ka1 Hþ Ka1 Ka2 Hþ H Ka1 Ka2 Ka3 Q Q Q ; a4 ¼ ; a3 ¼ a1 ¼ Q ; a2 ¼
þ 2
þ Q
þ 3 ¼ H þKa1 H þKa1 Ka2 H þ Ka1 Ka2 Ka3

(2-1)

where a1, a2, a3 and a4 represent the partition of cationic TC (denoted as TCHþ), neutral/zwitterionic TC (denoted as TC collectively), anionic TC (denoted as TC) and dianionic TC (denoted as TC2) species, respectively. TCs’ three acidity constants (Ka1, Ka2 and Ka3) correspond to the tricarbonyl-amide, phenolic-diketone and dimethylamino groups, respectively (Fig. 1), and are listed in Table 1. Note that the acidity constants of iso-CTC were not available from the literature and were determined in this study by spectrophotometric techniques as described in detail in the Supplementary Information Text S1 and Fig. S1. In contrast to TCs, ClO2 does not exhibit significant protonation and thus its concentration is not affected by solution pH. By incorporating equation (2) into equation (1), the degradation of TCs by ClO2 can be expressed as following: X d ½TCtotal ¼ kapp ½TCtotal ½ClO2  ¼  ki ai ½TCtotal ½ClO2  dt i¼1;2;3;4 kapp ¼

3.

(2)

i¼1;2;3;4

X

ki ai

(3)

(4)

i¼1;2;3;4

Results and discussion

3.1. Reaction kinetics and activity of TCs with chlorine dioxide On the basis that oxidation of most organic contaminants by ClO2 follows second-order reaction kinetics (e.g., (Huber et al., 2005; Wang et al., 2010)), the reaction between TCs and ClO2 is described by a second-order reaction rate equation as below: d ½TCtotal ¼ kapp ½TCtotal ½ClO2  dt

(1)

where kapp represents the apparent second-order rate constant for the overall reaction. [TC]total represents the sum of all acidebase species of a given TC, which can be modeled

where ki represents the specific second-order rate constant for reaction of ClO2 with TC species i and evaluates the contributions of each species of TC to the overall apparent rate constant kapp. Except for reactions at pH < 3.5, the reactions of TCs with ClO2 were so fast that competition kinetics was needed to measure the rate constants. In the method, equal molar concentrations of TC and the competitor RC were allowed to both react with a limited amount of ClO2 following secondorder reaction kinetics. Dividing the second-order rate equation of TC by that of RC and integrating leads to equation (5): ln

    kcompetitor ½Competitort0 ½TCt0 ¼  ln ½Competitort kTC ½TCt

(5)

Table 1 e Properties and specific rate constants of TCs with chlorine dioxide. Compound

pKa1

pKa2

pKa3

TTC OTC CTC Iso-CTC

3.30a 3.27b 3.30b 3.47c

7.68a 7.32b 7.44b 6.85c

9.30a 9.11b 9.27b 9.43c

kTC ;ClO2 (M1 s1) 7.15 3.71 1.05 3.05

(2.64) (1.52) (1.47) (2.47)

   

106 106 106 105

kTC2 ;ClO2 (M1 s1) 2.72 (0.18) 1.63 (0.13) 2.57 (0.09) 2.13 (0.25)

   

107 107 107 107

kapp (pH ¼ 7) (M1 s1) 1.26 1.24 3.16 2.24

   

106 106 105 105

kTC ;ClO2 and kTC2 ;ClO2 are the specific rate constant of anionic and dianionic species of TCs with ClO2, respectively. The uncertainty of rate constants corresponds to 95% confidence level. kapp (pH ¼ 7) refers to the calculated apparent second-order rate constant for TCs at pH 7 based on the listed species-specific rate constants and pKa values. a (Sassman and Lee, 2005). b (Stephens et al., 1956). c Determined in this study (see Supporting Information Text S1).

1842

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 3 8 e1 8 4 6

where [Competitor]t0 and [TC]t0 represent the initial concentrations of RC and TC, and [Competitor]t and [TC]t represent the final concentrations of RC and TC after reaction with substoichiometric amounts of ClO2. The kCompetitor and kTC represent the apparent second-order rate constants of RC and TC to ClO2, respectively. Plotting experimental data of equation (5) yielded good linear relationships (R2 > 0.95, Supplementary Information Fig. S2) and the ratio of kCompetitor/ kTC were obtained from the slope of each linear correlation. The second-order rate constants (kRC) of RC at the various pHs were calculated based on the reported specific rate constants of RC with ClO2 (pKa1 ¼ 9.2, kRC;ClO2 ¼ 4:0  101 M1 s1 , kRC ;ClO2 ¼ 4:8  107 M1 s1 (Tratnyek and Hoigne, 1994; Hoigne and Bader, 1994)). Note that, while RC has a second pKa2 of 11.21 (Rebenne et al., 1996), the dianionic species (RC2) was not included in the calculation because its specific rate constant with ClO2 is not available in the literature. Furthermore, based on the high pKa2 value, RC2 species accounts for <13.7% of the total RC even at the highest pH (10.33) examined in this study. Thus, neglecting the RC2 species is acceptable and similar to the approach employed in the previous studies (Tratnyek and Hoigne, 1994; Hoigne and Bader, 1994). The kapp of TCs at each pH were then obtained from the slopes and plotted in Fig. 2 for TTC and in Supplementary Information Fig. S3 for the other TCs. The kapp values of TCs at pH < 3.5 were determined in batch reactors. All TCs showed similar trends in that the reaction rate constant was strongly dependent on pH, increasing by more than 4e6 orders of magnitude from pH 2.5 to 10.5. The specific rate constants ki between ClO2 and individual TC species were calculated by least-squares regression of the experimental data to equation (4) using the SigmaPlot software and summarized in Table 1 (R2 > 0.98). The kinetic modeling indicates that contribution from the cationic and neutral TC species to the overall kapp is insignificant (both kTCHþ ;ClO2 and kTC;ClO2 < 10M1 s1 ) while the anionic and dianionic TC species are most reactive to ClO2 with large kTC ;ClO2 and kTC2 ;ClO2 values (Table 1). These results strongly suggest that ClO2 reacts with TCs predominantly at the phenolicdiketone and dimethylamino groups and deprotonation of these functional groups as pH increases considerably favors ClO2 oxidation (Fig. 2 and S3). Note that because a range of pKa values are reported for TCs in the literature, the sensitivity of 1.0

TC22.5e+7

TC TCH 0.8

TC0.6

Meas. kapp Modeled kapp

1.5e+7

0.4

1.0e+7

Mole Fraction

-1 -1 kapp (M s )

2.0e+7

0.2

5.0e+6

0.0

0.0 2

4

6

8

10

12

pH

Fig. 2 e Effect of pH and kinetic modeling for the reaction rate constants of TTC with ClO2.

the kinetic modeling to the change in pKa values was also evaluated by employing four different sets of literature pKa values for TTC in the modeling (Supplementary Information Table S1). The analysis indicates that the specific rate constants mostly do not deviate by more than 12% with the varying pKa values. The largest values of kTC2 ;ClO2 (Table 1) indicate that the unprotonated dimethylamino group is the most reactive moiety in the TC molecule to ClO2. Previous study shows that the principal rate-controlling step in the reaction of tertiary amines with ClO2 is electron abstraction, which is strongly affected by the stability of the N-centered radical formed at the first step of oxidation (Rosenblatt et al., 1967). In comparison to the literature values, the reactivity of TTC’s dimethylamino group to ClO2 (kTC2 ;ClO2 ¼ 2:72  107 M1 s1 ) is higher than that of trimethylamine (ktrimethylamine;ClO2 ¼ 1:0  107 M1 s1 ) but lower than that of N,N-dimethylaniline (kN;Ndimethylaniline;ClO2 ¼ 6:5  107 M1 s1 ) (Neta et al., 1988). This reactivity trend is reasonable since the aromatic ring of N,N-dimethylaniline offers resonance stabilization of the N-radical intermediate while trimethylamine lacks such capability. The dimethylamino group of TCs is in close proximity to the tricarbonyl-amide resonance group, which may provide medium level of stabilization for the generated radical intermediate. The comparable kTC2 ;ClO2 values among TTC, CTC and iso-CTC is also reasonable since the chlorine substitution and formation of iso-CTC at the C/D rings are remote from the dimethylamino group in the A ring (Fig. 1). However, the kTC2 ;ClO2 of OTC is discernibly lower than the others. Literature has reported hydrogen-bonding formation between C5eOH and dimethylamino group’s N on OTC (Hussar et al., 1968). Such H-bond formation may lower the reactivity of dimethylamino group’s N atom toward oxidation by ClO2. The kTC ;ClO2 corresponds to reaction between the phenolicdiketone group of TCs with ClO2. The TTC’s kTC ;ClO2 is lower than kTC2 ;ClO2 by a factor of 3.8, in the similar range of reactivity difference between phenol versus aromatic amine (kphenol;ClO2 ¼ 6:5  107 M1 s1 , kN;Ndimethylaniline;ClO2 ¼ 6:5  107 M1 s1 (Neta et al., 1988)). The lower kTC ;ClO2 (7.15  106 M1 s1) of TTC compared to phenol suggests lower reactivity of the phenolicdiketone group. This may be attributed to the polarizing effect of C11 ketone group to the aromatic ring. For example, 4-hydroxybenzaldehyde has a lower rate constant (kanion;ClO2 ¼ 6:0  106 M1 s1 (Hoigne and Bader, 1994)) to ClO2 than phenol. The lower kTC ;ClO2 of CTC than those of TTC and OTC (Table 1) is likely due to the electron-withdrawing effect of chlorine substituent that lowers reactivity for oxidation. The considerably lower kTC ;ClO2 of iso-CTC is likely caused by the disrupted resonance involving C11-OH and C12-OH groups, and thus lowered stabilization for the radical intermediate.

3.2. Reaction kinetics and activity of TCs with free chlorine The reactions of TCs with FAC at pH 3.0e10.3 were investigated by competition kinetics using 4-chlororesorcinol (4-CR) as the competitor (Supplementary Information Fig. S4). Using the equation (5) described above and the second-order rate constants (k4-CR) of 4-CR at various pHs calculated based on its

1843

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 3 8 e1 8 4 6

6

1.0 TC2-

TCH + 5

Log(kapp (M-1s-1))

4 0.6

TC3

0.4

Measured k Modeled k with H+ catalysis term Modeled k w/o H+ catalysis term

2

Mole Fraction

0.8

TC

0.2

1

0

d ½TCtotal dt

0.0 2

4

6

8

important role in the oxidation of TC. At pH below 5, an increase of rate constant was observed for all four TCs (Fig. 3 and S5). Previous studies have observed accelerated reaction rate between HOCl and aromatic systems such as phenols, methoxybenzenes and substituted pyrimidine (Rebenne et al., 1996; Gallard and von Gunten, 2002; Dodd and Huang, 2007) and have attributed it to acid catalysis in the formation of H2OClþ, which is a stronger electrophile than HOCl. Considering the speciation of TCs (eq. (2)) and FAC (eq. (7)), and the acid-catalyzed reaction, the oxidation of TCs by FAC can be expressed as below:

10

pH

Fig. 3 e Effect of pH and kinetic modeling for the reaction rate constants of TTC with free chlorine.

reported specific rate constants with FAC (pKa1 ¼ 8.09, pKa2 ¼ 10.75, k4-CR, HOCl ¼ 4.0  101 M1 s1, k4-CR, 5 1 1 s , k4-CR-2, HOCl ¼ 6.73  107 M1 s1, kHþ ¼ HOCl ¼ 1.43  10 M 1 1:19  106 M s1 (Rebenne et al., 1996)), the apparent secondorder rate constants of TCs with free chlorine at each pH were obtained and plotted in Fig. 3 for TTC and in Supplementary Information Fig. S5 for the other TCs. The significant variation in rate constant with changing pH can be attributed to the varying importance of specific reactions amongst the individual acidebase species of TCs and FAC. The acid dissociation of HOCl and corresponding speciation can be described by equations (6) and (7): ka



HOCl 4 OCl þ Hþ

(6)

i X h  bi ½FACtotal ½FACtotal ¼ ½HOCl þ OCl ¼

(7)

i¼1;2


þ H Ka ; b2 ¼
þ  b1 ¼
þ  H þ Ka H þ Ka

(7-1)

where pKa of HOCl is 7.5 at 25  C, I ¼ 0.0 M (Morris, 1966) and bj are the fractions of HOCl and OCl species. Within the pH range from 5 to 10.3, the maximum kapp at near pH 7.6 (Fig. 3) can be attributed to the increasing proportion of deprotonated TC species while decreasing concentration of HOCl species. The OCl species is generally a much weaker electrophile than HOCl (Gerritsen and Margerum, 1990) and thus played a less

¼ kapp ½TCtotal ½FACtotal X

¼




 kij ai bj þ kHþ Hþ ½TCtotal ½FACtotal

(8)

i¼1;2;3;4;j¼1;2

where kij represents the specific second-order rate constant for the reaction between species i of TC and species j of FAC, and kHþ represents the third-order rate constant (in M2 s1) for the acid-catalyzed reaction. Because OCl is a much weaker oxidant than HOCl and thus can be neglected, equation (9) was obtained: kapp ¼

X 


 ki ai b1 þ kHþ Hþ

(9)

i¼1;2;3;4

The specific rate constants ki and kHþ were determined by fitting the experimental data to equation (9) using least-square regression by SigmaPlot (Table 2, Fig. 3 and S5). To minimize the number of fitting parameters and obtain more accurate rate constants, the kinetic fitting was done first by fitting the data at pH > 6 to obtain the specific rate constants for the anionic and dianionic species. Afterwards, kinetic fitting was conducted using the obtained kTC ;HOCl and kTC2 ;HOCl values to determine kHþ for the lower pH range. The contribution from the cationic and neutral species was found to be negligible (both kTCHþ ;HOCl and kTC;HOCl < 102 M1 s1 ). The R2 values were 0.96e0.99 for TTC, OTC and CTC, and 0.93 for iso-CTC. Another possible way to explain the increased reaction rate at acidic condition is contribution from molecular chlorine (Cl2), whose formation can be described by the equilibrium below:  K

HOCl þ Hþ þ Cl 4 Cl2 þ H2 O

(10)


 
 ½Cl2  ¼ K Cl Hþ ½HOCl

(11)

Table 2 e Specific rate constants of TCs with free chlorine. Compound TTC OTC CTC Iso-CTC

kTC ;HOCl (M1 s1) 2.83 7.23 3.81 1.09

(0.24) (1.26) (0.86) (0.45)

   

106 106 105 104

kTC2 ;HOCl (M1 s1) 1.01 4.86 4.83 3.84

(4.96) (2.52) (1.98) (2.00)

   

106 106 106 106

kHþ ;HOCl (M2 s1) 1.12 1.27 1.73 3.63

(8.48) (8.84) (2.50) (2.82)

   

107 107 107 107

kCl2 (M1 s1) 6.09 6.90 9.40 1.97

   

106 106 106 107

kapp (pH ¼ 7) (M1 s1) 3.72 1.78 8.22 1.12

   

105 106 104 104

kTC ;HOCl and kTC2 ;HOCl are the specific rate constant of anionic and dianionic species of TCs with HOCl, kHþ is the third-order rate constant for the acid-catalyzed reaction of TCs with HOCl, and kCl2 is the estimated second-order rate constant for reaction of TCs with Cl2. kapp (pH ¼ 7) refers to the calculated apparent second-order rate constant for TCs at pH 7 based on the species-specific rate constants listed in this Table and the pKa values in Table 1. The uncertainty of rate constants corresponds to 95% confidence level.

1844

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 3 8 e1 8 4 6

The reported K for equation (10) is 2.3  103 M2 (Sivey et al., 2010). Clearly, Cl2 formation is most favorable at lower pH and with the presence of Cl. Chloride was not deliberately added into reaction solutions in this study; however, small amounts of HCl were added in the pH 3.0e3.3 experiments for pH adjustment and yielded approx. 8  104 M of Cl. Equation (12) is the modified mass balance of total free chlorine that includes Cl2 species, and the calculated pH-speciation of chlorine is shown in Supplementary Information Fig. S6. 

½Free Chlorinetotal ¼ ½Cl2  þ ½HOCl þ ½OCl 

  ¼ K½Cl ½Hþ  þ 1 þ Ka ½Hþ  ½HOCl

(12)

At the lowest pH 3.0 data point, the fractions of Cl2, HOCl and OCl species are 1.84  103, 9.98  101 and 3.16  105, respectively. Thus, [HOCl] can be approximated by b1[Free Chlorine]total throughout the experimental pH range of 3e10.5. The rate expression for chlorine oxidation of TCs can be revised to equation (13) and subsequently leads to equation (14): d ½TCtotal  ¼ kapp ½TCtotal ½Free Chlorinetotal dt   ¼  kHOCl ½HOCl þ kCl2 ½Cl2  ½TCtotal P  ¼ ðki ai b1 þ kCl2 b1 K½Cl ½Hþ Þ½TCtotal ½Free Chlorinetotal i¼1;2;3;4

(13) kapp ¼

X



ðki ai b1 þ kCl2 b1 K½Cl ½Hþ Þ

(14)

i¼1;2;3;4

where kHOCl and KCl2 represent the apparent rate constant with HOCl and Cl2, respectively. Kinetic modeling by equation (14) is essentially the same as that by equation (9) at acid pH  region. Thus, KCl2 is estimated from KCl2 ¼ kHþ =b1 K½Cl  (assuming b1 y 1) yielding the values of 6.09  106, 6.90  106, 9.40  106 and 1.97  107 M1 s1 for TTC, OTC, CTC and isoCTC, respectively. The magnitude of KCl2 for TCs is slightly higher than that reported for an aromatic thiene (Sivey et al., 2010). The kinetic modeling indicates that the dianionic and anionic species of TCs are most reactive to HOCl, suggesting that FAC reacts with TCs predominantly at the unprotonated dimethylamino and the deprotonated phenolic-diketone groups. On the basis of kTC2 ;HOCl , the four TCs have similar reactivity to FAC. For TTC and OTC, the reactivity of dimethylamino and phenolic-diketone groups to FAC was similar as shown by comparable values of kTC2 ;HOCl versus kTC ;HOCl . However, kTC ;HOCl was 1 and 2 orders of magnitude lower than kTC2 ;HOCl for CTC and iso-CTC, respectively. As mentioned in the earlier section, the presence of electron-withdrawing chlorine substituent on the phenol ring and the disruption of phenolicdiketone’s conjugated resonance (in iso-CTC) are likely responsible for the lower kTC ;HOCl Compared to the rate constant of phenol with FAC, (kphenol ;HOCl ¼ 3:3  105 M1 s1 (Rebenne et al., 1996)), the kTC ;HOCl of TTC and OTC are higher by about one order of magnitude; this is consistent with expectation based on the extra resonance stabilization by the phenolic-diketone group than a simple phenol for oxidation

intermediate. The dimethylamino group of TCs are more reactive than trimethylamine with FAC (kTMA;HOCl ¼ 5:0  104 M1 s1 (Deborde and von Gunten, 2008)); this may be due to the nearby tricarbonyl-amide group that can provide resonance stabilization for the cationic imine intermediate that will be generated by electrophilic attack of HOCl to the dimethylamino N.

3.3.

Reaction product evaluation

The reaction products of TTC, OTC and CTC with ClO2 and FAC at pH 7.5 were analyzed by LC/MS to evaluate important products. The chromatographic retention times, mass spectral data, and relative abundance of the parent TCs and oxidation products are listed in Supplementary Information Tables S2eS7. The molecular ion of TTC, OTC and CTC were m/z 445, m/z 461 and m/z 479, respectively. The products were short-written as M þ 32, M-166, etc., indicating the net mass loss or gain of the product from the parent compound. Due to the lack of authentic standards, a true quantification of the products is not possible. A rudimentary quantification of the products was performed by using the crude assumption that the mass spectrometric response of the products is the same as each other. The abundance of each product in percentage relative to the total products was calculated by MS/MStotal peak area ratio for the purpose to distinguish the dominant versus minor products and shown in Tables S2eS6. Overall, the mass spectral information was insufficient to identify the structures of oxidation products; however, did provide some important insight to the different product generation patterns by ClO2 versus by FAC. Oxidation of TCs by ClO2 yielded hydroxylation or oxygenation products as evident by the formation of M þ 32 (for TTC) and M þ 16 (for CTC) products (Tables S2 and S4). Loss of the dimethylamino group probably also occurred as indicated by the M-28 product in the oxidation of OTC (Tables S3). Additional chlorine substitution was not observed in any of the products, indicating that ClO2 oxidation did not lead to chlorination of TCs. Notably, ClO2 oxidation of TTC, OTC and CTC all yielded the same product with a lower molecular ion of m/z 279. This lower-molecular-weight product indicates that cleavage of TC’s ring system by ClO2 occurred and the product might contain the A ring portion of the parent compound because of the similar A ring structure among the three TCs. In contrast, oxidation of TCs by FAC mostly generated Cl- and OHsubstituted products such as M þ 34 (1Cl) and M þ 52 (1Cl, 1OH) products (Tables S5eS7) with clear isotope pattern for chlorine substitution. Cleavage of TC’s ring system did not occur by FAC oxidation based on the relatively high molecular weight of all products (m/z ¼ 431e513).

3.4.

Environmental significance

The kinetic results indicate that the reactivity of TCs with ClO2 is about an order of magnitude higher than that to FAC (Tables 1 and 2). Nevertheless, the rate constants to both oxidants are all very large, corresponding to half-lives of 0.02e1.54 s at pH 7.0 and an oxidant dosage of 1 mg/L (14.9 mM for ClO2 and 19.2 mM for FAC). Batch experiments were conducted to verify the kinetics in surface water (pH 6.83) and wastewater (pH 7.35) samples since competition kinetics could not be applied

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 3 8 e1 8 4 6

in real water samples. The real water samples were dosed with 2 mM of TCs and 1 mg/L ClO2 or 10 mg/L FAC, and no TCs remained after 30 s of reaction time. These results confirmed that rapid transformation of TCs under typical water and wastewater disinfection processes can be expected. Previous studies have reported that the C1eC4 positions in tricarbonyl-amide group and the C10eC12 positions in phenolic-diketone group are critical sites in the effectiveness of TC antibiotics, and changes at these positions will likely reduce antibiotic activity, while changes in other positions on TCs have less impact on antibiotic activity (Mitscher, 1978; Brodersen et al., 2000; Halling-Sorensen et al., 2002). Based on the above principles, ClO2 oxidation of TCs resulting in breakage of TC’s ring systems to smaller molecules most likely diminishes the antibiotic capacity of the parent compounds. The antibiotic potency of the Cl- and/or OH-substituted products is less clear and requires further studies to determine the exact substitution positions along with suitable bioassays.

4.

Conclusions

This study is among the first to investigate the reactions between the commonly used TC antibiotics and water disinfection oxidants ClO2 and FAC. Important findings about these oxidation reactions are:  TC antibiotics (TTC, OTC and CTC) react extremely rapidly with ClO2 and FAC in both reagent water and real water samples under conditions relevant in water/wastewater treatment (e.g., half-lives < 3.2 s at 1.0 mg/L oxidant dosage and pH 7.0). The reaction rate constants were about one order of magnitude higher with ClO2 than with FAC.  The reaction kinetics of TCs with ClO2 and FAC depend significantly on the deprotonation of TCs’ dimethylamino and phenolic-diketone groups. Influence of these two groups by functional group substitution (by eCl in CTC or by eOH in OTC) or alteration (formation of iso-CTC) affects the reactivity toward oxidation.  Oxidation of TCs by ClO2 leads to (hydr)oxylation and breakage of TC molecules, while oxidation of TCs by FAC leads to chlorinated and (hydr)oxylated products without any substantial ring breakage. Oxidation by ClO2 is thus more likely to diminish the antibiotic capacity of TCs.

Acknowledgements Financial support for P.W. from the China Scholarship Council is acknowledged. The materials and supplies for this study were available from a project funded by National Science Foundation 436 (CBET 0229172).

Appendix. Supplementary data Supplementary data related to this article can be found online at doi:10.1016/j.watres.2010.11.039.

1845

references

Acero, J.L., Benitez, F.J., Real, F.J., Roldan, G., 2010. Kinetics of aqueous chlorination of some pharmaceuticals and their elimination from water matrices. Water Res. 44, 4158e4170. APHA, 1998. ALPHA, AWWA, WEF Standard Methods for the Examination of Water and Wastewater. 20th edition., Washington, D.C. Brodersen, D.E., Clemons Jr., W.M., Carter, A.P., MorganWarren, R.J., Wimberly, B.T., Ramakrishnan, V., 2000. The structural basis for the Action of the antibiotics tetracycline, pactamycin, and hygromycin B on the 30S ribosomal subunit. Cell 103, 1143e1154. Chee-Sanford, J.C., Aminov, R.I., Krapac, I.J., GarriguesJeanjean, N., Mackie, R.I., 2001. Occurrence and diversity of tetracycline resistance genes in lagoons and groundwater underlying two swine production facilities. Appl. Environ. Microbiol. 67, 1494e1502. Chen, J., Regli, S., 2002. Disinfection practices and pathogen inactivation in ICR surface water plants. In: McGuire, M.J., McLain, J.L., Obolensky, A. (Eds.), Information Collection Rule Analysis Data. AWWA Research Foundation and American Water Works Association, p. 24. Chen, W.R., Huang, C.H., 2009. Transformation of tetracyclines mediated by Mn(II) and Cu(II) ions in the presence of oxygen. Environ. Sci. Technol. 43, 401e407. Deborde, M., von Gunten, U., 2008. Reactions of chlorine with inorganic and organic compounds during water treatment e kinetics and mechanisms: a critical review. Water Res. 42, 13e51. Dodd, M.C., Huang, C.-H., 2004. Transformation of the antibacterial agent sulfamethoxazole in reactions with chlorine: kinetics, mechanisms, and pathways. Environ. Sci. Technol. 38, 5607e5615. Dodd, M.C., Huang, C.-H., 2007. Aqueous chlorination of the antibacterial agent trimethoprim: reaction kinetics and pathways. Water Res. 41, 647e655. Dodd, M.C., Kohler, H.-P.E., von Gunten, U., 2009. Oxidation of antibacterial compounds by ozone and hydroxyl radical: elimination of biological activity during aqueous ozonation processes. Environ. Sci. Technol. 43, 2498e2504. Dodd, M.C., Shah, A.D., Von Gunten, U., Huang, C.H., 2005. Interactions of fluoroquinolone antibacterial agents with aqueous chlorine: reaction kinetics, mechanisms, and transformation pathways. Environ. Sci. Technol. 39, 7065e7076. Gallard, H., von Gunten, U., 2002. Chlorination of phenols: kinetics and formation of chloroform. Environ. Sci. Technol. 36, 884e890. Gerritsen, C.M., Margerum, D.W., 1990. Non-metal redox kinetics: hypochlorite and hypochlorous acid reactions with cyanide. Inorg. Chem. 29, 2757e2762. Gra¨slund, S., Bengtsson, B.-E., 2001. Chemicals and biological products used in south-east Asian shrimp farming, and their potential impact on the environment e a review. Sci. Total Environ. 280, 93e131. Halling-Sorensen, B., Sengelov, G., Tjornelund, J., 2002. Toxicity of tetracyclines and tetracycline degradation products to environmentally relevant bacteria, including selected tetracycline-resistant bacteria. Arch. Environ. Contam. Toxicol. 42, 263e271. Hamscher, G., Sczesny, S., Hoper, H., Nau, H., 2002. Determination of persistent tetracycline residues in soil fertilized with liquid manure by high-performance liquid chromatography with electrospray ionization tandem mass spectrometry. Anal. Chem. 74, 1509e1518. Hoigne, J., Bader, H., 1994. Kinetics of reactions of chlorine dioxide (OClO) in water e I. Rate constants for inorganic and organic compounds. Water Res. 28, 45e55.

1846

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 3 8 e1 8 4 6

Huber, M.M., Korhonen, S., Ternes, T.A., Von Gunten, U., 2005. Oxidation of pharmaceuticals during water treatment with chlorine dioxide. Water Res. 39, 3607e3617. Hussar, D.A., Niebergall, P.J., Sugita, E.T., Doluisio, J.T., 1968. Aspects of the epimerization of certain tetracycline derivatives. J. Pharm. Pharmacol. 20, 539e546. Kim, I., Yamashita, N., Tanaka, H., 2009. Performance of UV and UV/H2O2 processes for the removal of pharmaceuticals detected in secondary effluent of a sewage treatment plant in Japan. J. Hazard. Mater. 166, 1134e1140. Lindsey, M.E., Meyer, M., Thurman, E.M., 2001. Analysis of trace levels of sulfonamide and tetracycline antimicrobials in groundwater and surface water using solid-phase extraction and liquid chromatography/mass spectrometry. Anal. Chem. 73, 4640e4646. Mitscher, L.A., 1978. The Chemistry of the Tetracycline Antibiotics. Marcel Dekker, New York. Morris, J.C., 1966. The acid ionization constant of HOCl from 5 to 35  C. J. Phys. Chem. 70, 3798e3805. Navalon, S., Alvaro, M., Garcia, H., 2008. Reaction of chlorine dioxide with emergent water pollutants: product study of the reaction of three b-lactam antibiotics with ClO2. Water Res. 42, 1935e1942. Neta, P., Huie, R.E., Ross, A.B., 1988. Rate constants for reactions of inorganic radicals in aqueous solution. J. Phys. Chem. Ref. Data 17, 1027e1284. Palominos, R.A., Mondaca, M.A., Giraldo, A., Penuela, G., PerezMoya, M., Mansilla, H.D., 2009. Photocatalytic oxidation of the antibiotic tetracycline on TiO2 and ZnO suspensions. Catal. Today 144, 100e105. Rav-Acha, C., 1998. Transformation of aqueous pollutants by chlorine dioxide: reactions, mechanisms and products. Handb. Environ. Chem., 143e175. Rebenne, L.M., Gonzalez, A.C., Olson, T.M., 1996. Aqueous chlorination kinetics and mechanism of substituted dihydroxybenzenes. Environ. Sci. Technol. 30, 2235e2242. Rosenblatt, D.H., Hayes Jr., A.J., Harrison, B.L., Streaty, R.A., Moore, K.A., 1963. Reaction of chlorine dioxide with

triethylamine in aqueous solution. J. Org. Chem. 28, 2790e2794. Rosenblatt, D.H., Hull, L.A., DeLuca, D.C., Davis, G.T., Weglein, R. C., Williams, H.K.R., 1967. Oxidations of amines. II. Substituent effects in chlorine dioxide oxidations. J. Am. Chem. Soc. 89, 1158e1163. Sassman, S.A., Lee, L.S., 2005. Sorption of three tetracyclines by several soils: assessing the role of pH and cation exchange. Environ. Sci. Technol. 39, 7452e7459. Shah, A.D., Kim, J.-H., Huang, C.-H., 2006. Reaction kinetics and transformation of carbadox and structurally related compounds with aqueous chlorine. Environ. Sci. Technol. 40, 7228e7235. Sivey, J.D., McCullough, C.E., Roberts, A.L., 2010. Chlorine monoxide (Cl2O) and molecular chlorine (Cl2) as active chlorinating agents in reaction of dimethenamide with aqueous free chlorine. Environ. Sci. Technol. 44, 3357e3362. Stephens, C.R., Murai, K., Brunings, K.J., Woodward, R.B., 1956. Acidity constants of the tetracycline antibiotics. J. Am. Chem. Soc. 78, 4155e4158. Tratnyek, P.G., Hoigne, J., 1994. Kinetics of reactions of chlorine dioxide (OClO) in water e II. Quantitative structure-activity relationships for phenolic compounds. Water Res. 28, 57e66. Waller, C.W., Hutchings, B.L., Wolf, C.F., Goldman, A.A., Broschard, R.W., Williams, J.H., 1952. Degradation of aureomycin. VI. isoaureomycin and aureomycin. J. Am. Chem. Soc. 74, 4981. Wang, P., He, Y.-L., Huang, C.H., 2010. Oxidation of fluoroquinolone antibiotics and structurally related amines by chlorine dioxide: Reaction kinetics, product and pathway evaluation. Water Res. 44, 5989e5998. Zhang, H., Chen, W.-R., Huang, C.-H., 2008. Kinetic modeling of the oxidation of antibacterial agents by manganese Oxides. Environ. Sci. Technol. 42, 5548e5554. Zhu, J., Snow, D.D., Cassada, D.A., Monson, S.J., Spalding, R.F., 2001. Analysis of oxytetracycline, tetracycline, and chlortetracycline in water using solid-phase extraction and liquid chromatography-tandem mass spectrometry. J. Chromatogr. A 928, 177e186.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 4 7 e1 8 5 5

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Inactivation of bacteriophage MS2 upon exposure to very low concentrations of chlorine dioxide L.M. Hornstra a,*, P.W.M.H. Smeets a, G.J. Medema a,b a b

KWR Watercycle Research Institute, Groningenhaven 7, 3433 PE Nieuwegein, The Netherlands Delft University of Technology, Stevinweg 1, 2628 CN, Delft, The Netherlands

article info

abstract

Article history:

This study investigates the effects of very low concentrations of ClO2 applied in drinking

Received 31 March 2010

water practice on the inactivation of bacteriophage MS2. Concentrations of 0.5 mg/L,

Received in revised form

0.1 mg/L and 0.02 mg/L ClO2 inactivated at least 5 log units of MS2 after an exposure time of

24 November 2010

approximately 20, 50 and 300 min respectively. When the ClO2 concentration was as low as

Accepted 28 November 2010

0.005 mg/L, inactivation of 1 log unit MS2 was observed after 300 min exposure. Increasing

Available online 4 December 2010

the contact time to 24 h did not increase the inactivation any further. Non-linear inactivation kinetics (tailing) were observed for all conditions tested. Repeated addition of MS2 to

Keywords:

the reactor showed that tailing was not caused by a reduction of the biocidal effect of ClO2

Chlorine dioxide

during disinfection. The Modified Chick-Watson, the Efficiency Factor Hom (EFH) model

MS2

and the Modified Cerf model, a modification of the two-fraction Cerf model, were fitted to

Disinfection

the non-linear inactivation curves. Both the EFH and the modified Cerf model did fit

Inactivation kinetics

accurately to the inactivation data of all experiments. The good fit of the Modified Cerf

Virus

model supports the hypothesis of the presence of two subpopulations. Our study showed that ClO2 is an effective disinfectant against model organism MS2, also at the low concentrations applied in water treatment practice. The inactivation kinetics followed a biphasic pattern due to the presence of a more ClO2-resistant subpopulation of MS2 phages, either caused by population heterogeneity or aggregation/adhesion of MS2. ª 2010 Elsevier Ltd. All rights reserved.

1.

Introduction

Pathogenic enteric viruses originating from human excreta are frequently present in surface water, and because of their particle stability they can persist in this environment for weeks to months (de Roda Husman et al., 2009). Despite the application of sophisticated water treatment strategies for drinking water production, these viruses remain important sources of waterborne outbreaks and therefore a concern for human health (Carter, 2005; Lodder and de Roda Husman, 2005). In order to avoid drinking water related outbreaks, the US Environmental Protection Agency (EPA) has launched the Long Term 2 Enhanced Surface Water Treatment Rule for the

US (USEPA 2006), which requires a removal or inactivation of enteric viruses from source water by at least 4 logs (99.99%). In the Netherlands, regulations require a risk assessment to show that the risk of drinking water related infections is below 1 infection per 10,000 consumers per year (Anonymous, Dutch Drinking Water legislation, 2001). Surface water treatment generally includes mechanical purification processes, followed by a disinfection treatment to kill residual micro-organisms. Worldwide, primarily chlorine disinfection is used to achieve the required reduction of pathogenic organisms, but chlorine has negative side effects like the formation of harmful disinfection by-products (DPB) like trihalomethanes (Rook, 1974). Alternatives like chlorine

* Corresponding author. Tel.: þ31 (0)306069628; fax: þ31 (0)306061165. E-mail address: [email protected] (L.M. Hornstra). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.11.041

1848

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 4 7 e1 8 5 5

dioxide (ClO2) have been proposed to prevent DPB formation. Advantages of ClO2 disinfection include; no formation of trihalomethanes, oxidation of iron and manganese and it is a relative persistent residual. Disadvantages include formation of chlorite and chlorate and organic halides (Aieta and Berg, 1986). ClO2 is sporadically used as primary disinfectant, but at some locations in Europe ClO2 is added to finished water to maintain a disinfectant residual in the distribution network to prevent regrowth and provide protection against ingress of fecal contaminants. Under these conditions, very low concentrations of ClO2, generally not exceeding 0.1 mg/L are added to the finished water. ClO2 effectively inactivates micro-organisms in water and inactivation of several water related pathogenic viruses has been reported (Alvarez and O’Brien, 1982; Chen and Vaughn, 1990; Thurston-Enriquez et al., 2005; Berman and Hoff, 1984; Lim et al., 2010b). Most of these studies used relative high concentrations of ClO2 in combination with short contact times, but in water applications frequently low concentrations are dosed, and inactivation is achieved by long contact time. The level of disinfection is expressed as Ct values, which is the disinfectant concentration (in mg/L) multiplied by the contact time (in minutes). The Chick-Watson model, describing first order inactivation kinetics is routinely used to calculate Ct values necessary to achieve the required level of disinfection. Based on previous viral disinfection studies the EPA has listed minimal Ct values for inactivation of viruses (USEPA, 2006). However, many observations have shown that viral inactivation kinetics using disinfectants does not follow first order kinetics, but shows significant tailing. Several explanations for tailing in general have been proposed (Cerf, 1977; Xiong et al., 1999) including (i) aggregation of viruses or attachment of viruses to particles in water, herewith protecting the target organism against exposure to the disinfectant (ThurstonEnriquez, 2003; Keswick et al., 1985; Berman and Hoff, 1984) (ii) the presence of subpopulations of the target organism with differences in resistance against the disinfectant (Hiatt, 1964; Gerba et al., 2003) (iii) a decreasing biocidal effect of the inactivating agent during the disinfection treatment (Hiatt, 1964). These non-linear inactivation curves are not adequately described by the Chick-Watson model. Other models including the Efficiency Factor Hom (EFH) model have been developed to describe non-linear inactivation curves more accurately (Gyu¨re´k and Finch, 1998; Xiong et al., 1999; Haas and Joffe, 1994). These models describe non-linear inactivation fairly precise. However, none of these models reveal information about biological causes of non-linear viral inactivation kinetics. The current study investigates the inactivation of viruses by ClO2, using bacteriophage MS2 as model organism for human enteric viruses. MS2 is frequently used either as indicator organism or as model organism for enteric viruses, because it exhibits similar morphology and is relatively resistant to disinfection treatments compared to human pathogenic viruses (Sobsey, 1989, IAWPRC Study Group, 1991). It functions therefore as an appropriate candidate representing human enteric viruses present in the environment. The objectives of this study were to (i) determine the inactivation of MS2 at the low concentrations of ClO2 applied in drinking water practice (ii) to determine the lower threshold of ClO2 concentrations able to inactivate MS2 (iii) investigate if

the Ct concept is valid for very low concentrations of ClO2 and (iv) investigate the cause of tailing observed during inactivation of MS2.

2.

Materials and methods

2.1.

Virus strain and assay

The bacteriophage MS2 solution containing 1  1012 pfu/ml was obtained from GAP Enviromicrobial Services (London, Canada). Purification by GAP Enviromicrobial Services included centrifugation followed by filtration of MS2 using a 0.2 mm filter to remove particles and cell debris. The double layer agar method was used to enumerate MS2 after the disinfection treatments according (ISO 10705-1) using Salmonella typhimurium WG49 as host organism (ISO, 1995). All experiments were performed in duplicate.

2.2.

ClO2 production and measurement

ClO2 was produced using a Halox H1000SRE unit (Halox technologies, Bridgeport, CT, USA). A ClO2 solution with a concentration of approximately 2.0 g/L was produced and diluted to 800 mg/L ClO2 stock solution. This stock solution was stored at 4  C in the dark and remained stable for at least half a year. For disinfection experiments, the stock solution was further diluted to a working stock solution with concentration between 1 and 5 mg/L. The ClO2 concentration of the working stock was determined by measuring the absorbance directly at 359 nm (the molar absorptivity is 1106 L/mol cm at 359 nm) in a 1 cm quartz cuvette. ClO2 concentrations during disinfection experiments with a concentration between 0.7 mg/L and 0.2 mg/L ClO2 were determined by measuring the absorbance at 359 nm, using a 5 cm quartz cuvette. Below 0.2 mg/L ClO2 this method is not sufficiently accurate, and therefore ClO2 concentrations with a concentration between 0.2 and 0.05 mg/L were determined by using the spectrophotometric method measuring the discoloration of the indicator chlorophenol red, as described by Fletcher and Hemmings (1985). This method allows accurate determination of ClO2 concentrations from 0.05 mg/L and higher. Samples were measured at 575 nm using 5 cm cuvettes for maximum sensitivity. Buffered demand free water was used as blanco.

2.3.

Experimental conditions

All glassware was made ClO2 demand free by overnight soaking in 5 mg/L of ClO2, followed by thorough rinsing with ultrapure MQ (Millipore, Massachusetts, US) water before use. Disinfection experiments were performed in two batch reactors (1000 ml erlenmeyers), each filled with 900 ml buffered demand free water (BDF). BDF water was made by addition of 10 ml. 0.5 M sodium phosphate buffer pH 7.2e990 ml of ultrapure MQ water, resulting in 5 mM sodium phosphate buffered demand free water, pH 7.2. The reactors were kept on ice and continuously stirred to keep the temperature at 0  C. Reactor 1 was used for disinfection experiments and before starting 100 ml of a 10 ClO2 stock solution of the appropriate concentration was added. After mixing, samples (50 ml in

1849

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 4 7 e1 8 5 5

duplicate) were taken to determine the concentration of ClO2 in the reactor, resulting in 900 ml ClO2 solution in the reactor. Subsequently 1 ml of MS2 stock consisting of 1  1010 PFU ml was added (t ¼ 0) to the reactor resulting in a concentration of approximately 1  107 PFU ml infectious MS2 in the reactor. To determine the number of remaining infectious MS2 phages during disinfection, 5 ml samples were taken at appropriate time points, and immediately quenched in 0.5 ml 1% (w/v) sodium thiosulfate and stored on ice. ClO2 quenched with sodium thiosulfate did not show any inactivation of MS2 (data not shown). Furthermore samples were taken from reactor 1 during the experiment to determine the ClO2 decay. Reactor 2 was used to determine the initial concentration of MS2 at t ¼ 0 by addition of 1 ml of MS2 stock consisting of 1  1010 PFU ml to 900 ml BDF water. BDF water (ultrapure MQ water with 5 mM sodium phosphate pH 7.2) demonstrated negligible ClO2 demand. Additionally ClO2 disinfection experiments were done in quadruplicate with initial ClO2 concentration of 0.5, 0.25, 0.1 and 0.05 mg/L respectively, to determine the effect of the initial ClO2 concentration on the ClO2 decay rate. For this, ClO2 decay was followed for at least 120 min disinfection time, and the ClO2 concentration was measured at minimally 5 time points for each disinfection experiment.

2.4.

Kinetic modeling

The Solver function in Microsoft Excel 2003 was used to minimize the sum of squares of the difference between the measured ClO2 concentration and calculated result of first order kinetic equation (equation (1)), to determine the ClO2 decay constant (k0 ).
 C ¼ C0 exp k0 t

Modified Chick  Watson lnN=N0 
 ¼ kCn0 =nk0  1  exp nk0 t

h .
im 0 1  efnk t=mg  nk0 t=m

(4)

where m is Hom’s exponent. To obtain values for k, n and m, Excel solver was used to minimize the sum of squares of the difference between the observed and calculated ln N/N0 for each disinfection experiment separately. Data from single experiments were used to estimate the model parameters for all models.

3.

Results and discussion

3.1.

Inactivation of MS2 with low concentrations of ClO2

3.1.1.

ClO2 concentration during disinfection

Disinfection experiments were conducted with ClO2 concentrations of approximately 0.5 mg/L, 0.1 mg/L, 0.02 mg/L and 0.005 mg/L. For disinfection experiments with an initial ClO2 concentration of 0.02 mg/L and 0.005 mg/L, the ClO2 concentration and ClO2 decay in the reactor could not be determined due to the detection limit of 0.05 mg/L of the chlorophenol red method. Therefore the initial ClO2 concentration of these experiments was determined by measuring the concentration of the 10 working stock shortly before addition to the reactor. As to our knowledge no information exists about ClO2 decay at very low concentrations, we had to estimate a k’ value for these experiments. To make this estimation as accurate as possible, ClO2 decay was measured over time for 15 individual disinfection experiments with various initial ClO2 concentrations ranging from 0.6 mg/L to 0.05 mg/L ClO2.,and the first order disinfectant decay rate (k’) was derived (Table 1). This should reveal any relationship between the initial ClO2

(1)

where C and C0 are the ClO2 concentration (mg/L) at time t (min). k0 is the first order disinfectant decay rate constant per minute (Haas and Joffe, 1994). The inactivation kinetics of MS2 by ClO2 were described using the modified Chick-Watson equation (3), which is derived from the classical Chick-Watson model (2) but includes a factor for disinfectant decay. Chick  Watson lnN=N0 ¼ kCt

EFH lnN=N0 ¼ kCn0 tm

Table 1 e Initial ClO2 concentrations and ClO2 decay constant k0 for 15 disinfection experiments with initial concentrations between approximately 0.6 and 0.05 mg/L ClO2. Disinfection decay calculations were based on at least 5 data points for each individual experiment. ClO2 concentration Measured

(2)

(3)

where ln N/N0 is the natural log of the survival ratio (¼number of infectious MS2 phages at time t divided by the number at t ¼ 0), k is the inactivation rate constant of the target organism (MS2) and n is the coefficient of dilution which represents the average number of molecules combined with the organism necessary to cause inactivation (Gyu¨re´k and Finch, 1998). The Efficiency Factor Hom (EFH) model is frequently applied to calculate Ct values for non-linear inactivation curves (Thurston-Enriquez et al., 2003, 2005; Lim et al., 2010a), and can be considered as an acceptable approximation of the Hom model for systems following first order disinfectant decay and showing non-linear inactivation kinetics (Haas and Joffe, 1994).

Exp

ClO2a

k0 (min1)

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

0.53 0.57 0.62 0.63 0.25 0.28 0.29 0.094 0.094 0.116 0.099 0.033 0.047 0.056 0.042

0.00072 0.00086 0.00098 0.00131 0.00107 0.00119 0.00126 0.00086 0.00092 0.00119 0.00150 0.00107 0.00136 0.00175 0.00171

a Concentration in mg/L.

1850

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 4 7 e1 8 5 5

concentration and k’ in this concentration range, and extrapolation to lower ClO2 concentrations may represent the best approximation of k’ at very low concentrations. To get better insight in the relationship between the decay rate and the initial concentration, k’ versus initial concentration was plotted for each individual experiment (Fig. 1). No obvious increasing or decreasing dependence of k’ to the initial ClO2 concentration could be revealed. A Pearson’s correlation showed that no significant trendline could be derived from the average plotted data points ( p ¼ 0.085). Hence, we assume that the ClO2 decay rate is independent from the initial ClO2 concentration within the concentration range considered. With this information, it was decided to use the average decay rate of 0.001183  0.000304 for all disinfection experiments, and this number was considered describing the decay for all concentrations accurately, including the ClO2 decay for disinfection reactions with ClO2 concentrations below the detection limit. To determine the sensitivity of the models for variation in k’, the 2.5 percentile and 97.5 percentile from the obtained k’ values were determined at 0.00077 and 0.00173. These values were used to determine the sensitivity of the models for deviations from the average k’.

3.1.2.

Fig. 2 e Inactivation of MS2 after exposure to 0.43 (diamonds), 0.1 (triangles), 0.02 (squares), 0.005 (circles) and 0 (hexagons) mg/L ClO2. The inactivation curves show average inactivation of duplicate disinfection experiments. Note the break on the x-axis to show inactivation after 1440 min contact time.

Inactivation of MS2

Disinfection experiments with initial concentrations of 0.39 and 0.46 mg/L ClO2 were able to reduce the number of infectious MS2 with at least 5 log units in 20 min (Fig. 2). Disinfection with an initial ClO2 concentration of 0.10 mg/L required an exposure time of 50 min for 5 log unit reduction of MS2 phages. Also an initial concentration of 0.02 mg/L ClO2 was able to reduce the number of infectious MS2 with 5 log, although 300 min of contact time was required for this reduction. The residual ClO2 after 300 min could be calculated using the k’ derived as described in Section 3.1.1 and was 0.013 and 0.014 mg/L for the two experiments. In order to further reduce the initial amount of ClO2, a disinfection experiment with a very low initial concentration of 0.005 mg/L ClO2 was done. This concentration did reduce the infectious MS2 population, but with only 1 log unit in approximately 300 min.

Prolongation of the contact time to 24 h did not reduce the infectious population any further (Fig. 2). These results show that an initial concentration of 0.005 mg/L ClO2 is capable of reducing the MS2 population by 1 log unit in approximately 300 min, but then inactivation ends. Inactivation may have stopped either by disappearance of the total ClO2 amount after 300 min, or by reaching a ClO2 concentration that is not capable of inactivating MS2 anymore. Because it is not possible to measure these low residual ClO2 concentrations, none of these possibilities can be excluded. However, the Ct concept states that “the level of inactivation can be assigned to a given Ct value, independently of the disinfectant concentration used”. The inactivation curves versus Ct (mg/L min) of the disinfection experiments in this study (composed by various combinations of initial ClO2 concentrations multiplied by time) confirm this concept (Fig. 3). Assuming that the disinfectant rate constant k0 is concentration independent for very low ClO2 concentrations, the calculated ClO2 residual in the reactor after 300 min disinfection is 0.0034 mg/L, and 0.00089 mg/L after 24 h (1440 min). Exposure to a ClO2 concentration of 0.0034 mg/L for 1140 min (1440e300 min) including accounting for ClO2 decay results in an additional 1.83 Ct (mg/L min) for the 1140 min timeframe (Fig. 4). It can be derived from Fig. 3 that 1.83 Ct results in an inactivation of approximately 3 log, but no inactivation is observed. This suggests that, assuming the Ct concept is valid for low concentrations, inactivation ended after the 300 min time point because ClO2 had disappeared from the reactor by decay or volatilization.

Fig. 1 e A scatter plot showing the ClO2 decay rate (k’) versus the initial ClO2 concentration for 15 disinfection experiments. The average k’ value of 0.001183 is shown by a dashed line. The dotted lines show the 2.5 percentile and 97.5 percentile values of the obtained k’ values.

3.2.

Non-linear inactivation of MS2

Tailing is frequently observed in disinfection experiments of water related pathogenic viruses (Finch and Fairbairn, 1991; Simonet and Gantzer, 2006; Alvarez and O’Brien, 1982). The

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 4 7 e1 8 5 5

1851

a reduction of the biocidal effect of ClO2. To exclude this latter possibility as a cause of tailing for the experiments of this study, the biocidal effect of ClO2 was determined during the time of the disinfection experiment.

3.3.

Fig. 3 e Inactivation of MS2 after exposure to 0.5 (diamonds), 0.1 (triangles) and 0.02 mg/L (squares) ClO2, in which inactivation is being displayed versus Ct (mg/L min). First experiment (open symbols), duplo experiment (closed symbols).

inactivation curves obtained from the disinfection experiments described in the current study showed non-linear inactivation for all ClO2 concentrations applied. MS2 inactivation occurred relatively fast directly after initiation of the disinfection treatment, and slowed down when the treatment proceeded. Rapid inactivation occurred for about 3 log inactivation, followed by continued inactivation but at a lower rate. This resulted in non-linear inactivation curves, showing tailing or a biphasic shape (Figs. 2 and 3). This could be attributed to subpopulations within the MS2 phage population caused by attachment to particles or intrinsic heterogeneity of the population, or to disinfection reaction conditions that alter during the disinfection time and herewith cause

Fig. 4 e Cumulative Ct credits obtained after exposure for 1500 min, starting with an initial ClO2 concentration of 0.0034 mg/L ClO2 at 300 min. This concentration represents the calculated concentration present after 300 min disinfection with an initial concentration of 0.005 mg/L ClO2 at 0 min. The 1140 min time point reflects 24 h disinfection, resulting in an additional Ct credit of 1.83 Ct, as is shown by the dotted lines.

Changes in biocidal effect of ClO2 during disinfection

A possible decline of the biocidal effect of ClO2 during the disinfection experiments was investigated, using an experimental setup similar to the previous disinfection experiments. Considering the disinfection experiment with a concentration of 0.10 mg/L ClO2, the transfer from fast to slow inactivation occurred after approximately 10 min (Fig. 7). If the reduction in inactivation rate is caused by a diminished biocidal effect of ClO2 after approximately 10 min disinfection, fresh MS2 phages added to the disinfection reactor after 10 min disinfection are expected to follow slow inactivation kinetics. Addition of fresh MS2 after 10 min reaction time did not result in a low inactivation rate of the added population (Fig. 5). Their inactivation was similar to inactivation of the MS2 population added at t ¼ 0. Addition of fresh MS2 after 60 min again showed similar inactivation kinetics (data from single experiment not shown). These results indicate that the biocidal effect of ClO2 did not diminish during disinfection, and was identical to the situation at t ¼ 0. Similar results regarding the biocidal effect of ClO2 have been observed for ClO2 disinfection of bacterial subpopulations (Berg et al., 1988). As the inactivation conditions apparently did not change during disinfection, tailing could be caused by intrinsic resistance differences of individual MS2 phages. In that case MS2 phages representing the slow phase of inactivation (the tail) are more resistant against ClO2 and exposure to fresh ClO2 would not result in faster inactivation. To determine this, the population representing the tail was transferred to a freshly prepared reactor containing BDF with 0.1 mg/L ClO2. This population of more ClO2-resistant phages was obtained by disinfection of 1  109 PFU of the normal MS2 population for 10 min in 0.1 mg/L ClO2, resulting in approximately 3 log

Fig. 5 e Inactivation curve of an MS2 population added at t [ 0 (closed symbols) to the reactor, and inactivation of a similar MS2 population added to the same reactor after 10 min of disinfection (open symbols). Both populations show similar inactivation kinetics.

1852

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 4 7 e1 8 5 5

inactivation of the ClO2-sensitive MS2 population. 100 ml of the resulting reaction mixture containing approximately 1  106 PFU of the ClO2-“resistant” MS2 population was immediately transferred to a batch reactor containing 900 ml fresh 0.1 mg/L ClO2. Fig. 6 shows that disinfection of this population resembles the slow inactivation at the tail of the non-pretreated population. These results exclude the possibility of diminishing biocidal effect of ClO2 during disinfection, and therefore the variance in inactivation kinetics is a property of the MS2 phage population.

3.4.

The resistant subpopulation representing the tail

The current study shows that tailing is a characteristic of the MS2 population and the population consisted of at least two subpopulations that differ substantially in their resistance against ClO2. The majority of the population (approximately 99.9%) consisted of MS2 that was rapidly inactivated, while approximately 0.1% of the population required prolonged exposure to ClO2 for inactivation. Additional resistance may be an intrinsic property of a heterogeneous virus population, or can be acquired later e.g. by aggregation with other virus particles or by attachment or adhesion to particles in the environment. Several studies have concluded that adhesion to particles or aggregation formation provides protection to viruses, herewith reducing disinfection efficacy and probably introducing tailing (Berman and Hoff, 1984; Thurston-Enriquez et al., 2003; Urakami et al., 2007). The conditions in the current study used BDF water and were performed under laboratory conditions, so attaining additional resistance by particle attachment is unlikely. Furthermore, experimental conditions do not favor aggregation of MS2 as the low isoelectric point of MS2 of 3.9 results in repulsion rather than attraction at the pH used during the disinfection experiments (Gerba, 1984). However, we cannot experimentally exclude aggregation or adhesion to particles. Also intrinsic variation in the phage population, resulting in phages with increased resistance might be an important source of tailing. Synthesis

Fig. 6 e Comparison of the inactivation kinetics of MS2 phages representing the tail of the inactivation curve (open symbols) with inactivation kinetics of the initial population (closed symbols).

of MS2 in the host is a self-regulating assembly process and it is not unlikely that this process results in virions that individually differ from each other. For example variation might exist in the structure of the capsid, resulting in variance in resistance against disinfectants.

3.5.

The shape of the inactivation curve

The shape of the inactivation curves observed can be described as upward concave or biphasic (Fig. 2, 3 and 5). A variety of disinfection models have been described for these types of inactivation curves (Gyu¨re´k and Finch, 1998; Xiong et al., 1999). These mostly empirical models were designed to properly determine Ct values based on inactivation experiments of waterborne pathogens or model organisms. Because of its simplicity, the Chick-Watson model (equation (1)) is predominantly used for Ct calculations in drinking water regulatory practice, though it has been described that it does not predict non-linear inactivation curves very precise (Gyu¨re´k and Finch, 1998). The EFH (equation (4)) describes non-linear inactivation curves more adequately, by adapting the parameter m in combination with contact time. Variance of m gives more or less importance to contact time, with m < 1 describing tailing of the inactivation curve. The EFH model empirically describes inactivation curves, but does not reveal much about biological phenomena involved in tailing. For all disinfection experiments, the inactivation rate constant k was determined using the MCW model, with n ¼ 1 to come close to the traditional Chick-Watson model. For the EFH model k, n and m where determined (Table 2) based on the best fit determined by minimizing the sum of squares. Because these models include disinfectant concentration C multiplied by contact time t, the decrease of disinfectant concentration should be compensated for by an increase in time. As concentration was the only variable in the disinfection experiments described, and inactivation versus Ct showed similar inactivation for all combinations of ClO2 concentration versus time (Fig. 3), the outcome of the EFH model parameters are expected to be close to equal for each concentration. However, variation is found for the parameters n and m in case of the EFH model. This indicates that EFH model may not accurately predict inactivation of MS2 when extrapolated to other ClO2 concentrations/contact time combinations than what was used in the disinfection experiments described. Since our results suggested population heterogeneity as main cause for tailing, a biphasic inactivation curve was considered. A typical biphasic inactivation curve results when the population of the target organism consists of two subpopulations, each with a different resistance to disinfectants (Cerf, 1977). Cerf derived a mechanistic two-fraction model that specifically describes the inactivation of two populations with differing inactivation kinetics. Cerf’s model has not been used before in the field of water treatment, and is predominantly used to describe the existence of two subpopulations with different resistance characteristics against food preservation treatments (predominantly thermal treatments). This model does not contain a parameter for disinfectant concentration. Because the experiments in this study focused on low concentrations of ClO2 in combination with long contact times, ClO2 decay during disinfection time

1853

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 4 7 e1 8 5 5

Table 2 e Summary of parameter estimates for the Modified Chick-Watson model, the EFH model and the modified Cerf model, and ESS and R2 values for comparison of predicted and observed ClO2 inactivation curves. Raw data were used for estimation of the model parameters. The ESS and R2 values between brackets were calculated using the Cerf model without accounting for disinfectant decay. Modified Chick-Watson (MCW model) Exp

ClO2 mg/L

Data pointsb

k

n

ESSc

R2

0.39 0.45 0.10 0.10 0.019 0.020 0.0049 0.0051

12 12 12 12 11 11 5 5

2.245 1.581 1.731 1.951 2.903 2.825 2.095 2.709

1 1 1 1 1 1 1 1

166.942 70.857 142.693 144.854 41.785 44.670 0.953 1.951

0.861 0.851 0.880 0.885 0.886 0.913 0.817 0.849

1a 1b 2a 2b 3a 3b 4a 4b

EFH Exp 1a 1b 2a 2b 3a 3b 4a 4b

ClO2 mg/L

Data pointsb

k

n

m

ESSb

R2

0.39 0.45 0.10 0.10 0.019 0.020 0.0049 0.0051

12 12 12 12 11 11 5 5

9.771 5.978 6.441 6.316 15.249 6.659 5.002 2.663

0.732 0.827 0.372 0.416 0.800 0.544 0.757 0.400

0.335 0.438 0.375 0.405 0.537 0.496 0.593 0.396

1.516 3.754 1.763 2.976 3.187 1.557 0.384 0.025

0.988 0.970 0.992 0.987 0.984 0.991 0.912 0.995

Modified Cerf Exp 1a 1b 2a 2b 3a 3b 4a 4b

ClO2 mg/L

Data pointsb

f

k1

k2

ESSc

ESSc

R2

R2

0.39 0.45 0.10 0.10 0.019 0.020 0.0049 0.0051

12 12 12 12 11 11 5 5

0.9978 0.9942 0.9991 0.9976 0.9962 0.9905 nda nda

14.687 5.928 10.939 11.434 6.759 9.019

0.977 0.679 0.839 0.984 1.282 1.513

6.638 0.649 5.822 5.873 0.924 1.994

(6.698) (0.656) (6.380) (6.537) (1.186) (2.379)

0.954 0.995 0.976 0.981 0.995 0.990

(0.953) (0.995) (0.974) (0.976) (0.994) (0.989)

a Not determined because inactivation is not biphasic. b Number of data points used for calculation of the inactivation curve. c Error sum of squares.

should preferably not be ignored. Therefore disinfectant concentration and decay was implemented in the Cerf model. CERFs model NðtÞ=N0 ¼ f eðk1tÞ þ ð1  f Þeðk2tÞ

(5)

in which N(t)/N0 represents the survival ratio (¼number of infectious MS2 phages at time t divided by the number at t ¼ 0), k1 and k2 are the inactivation rate constants for population 1 and 2 respectively, and f is the initial proportion in the less resistant fraction. Expressed in natural logarithm this can be rewritten as:
 lnðNðtÞ=N0 Þ ¼ ln f eðk1tÞ þ ð1  f Þeðk2tÞ To adapt this model for chemical inactivation, we replaced the constant t for Ct (concentration multiplied by time) which results in:
 lnðNðtÞ=N0 Þ ¼ ln f eðk1CtÞ þ ð1  f Þeðk2CtÞ This formula describes the inactivation of two independent subpopulations both following first order kinetics, and factors k1Ct and k2Ct are similar to Chick-Watsons model

describing first order inactivation (equation (2)). Modification of the Chick-Watsons model for inactivation conditions under disinfectant demand conditions resulted in the MCW model (equation (3)). Implementation of MCW in Cerf’s model by replacing kCt with kCn0 =nk0  ½1  eðnk0 tÞ results in a modified Cerf model capable of describing two-fraction inactivation under disinfectant decay following first order kinetics:

 þ ð1  f Þe lnNðtÞ=N0 ¼ lnf e  k1Cn0 =nk0  1  e nk0 t

  k2Cn0 =nk0  1  exp nk0 t

(6)

For this modified Cerf model, the model parameters f, k1 and k2 were determined with n ¼ 1. For individual disinfection experiments the best fit was determined by minimizing the sum of squares between the observed and the calculated value. Both the Modified Cerf model and the EFH model produced an accurate fit to the data (Table 2). Based on the Error sum of squares (ESS) and correlation coefficients (R2), the EFH model produced a slightly better fit to the data compared to the Cerf model. Comparison of the results of the Cerf model

1854

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 4 7 e1 8 5 5

(without disinfectant decay) with the Modified Cerf model (including disinfectant decay) showed that the Modified Cerf model improved the fit of the model for each experiment (see ESS and correlation coefficient in Table 2). The fit of the MCW, EFH and modified Cerf model on the inactivation data of 0.39, 0.45, 0.1, 0.1, 0.019 and 0.020 mg/L ClO2 are shown in Fig. 7. To summarize, the modified Chick-Watson cannot describe the experimental data very well, as it cannot describe deviations from linear inactivation (n ¼ 1). Both the empirical EFH model and the mechanistic Modified Cerf model are well capable of describing the inactivation curves for all individual disinfection experiments. The EFH model produced the most accurate fit based on the sum of squares and correlation coefficient. However, the accurate fit of the modified Cerf model originally designed to describe inactivation of two subpopulations, suggests the presence of two subpopulations of MS2 which differ en their resistance against ClO2. To find out how sensitive the models are for variations in the ClO2 decay rate k0 the 2.5 percentile and 97.5 percentile as determined in Section 3.1.1 of 0.00077 and 0.00173, were transferred to the Chick-Watson, the EFH and Modified Cerf model. Using these values did change the inactivation curves minimally for the disinfection experiments as was determined by comparing R2 of confidence interval values with R2 of the average k0 value. For the Chick-Watson model, the maximal deviation from the R2 obtained with the average k0 value was 2.3%, specifically for the experiments with disinfection times of 300 min. The maximal deviation from R2 found for the EFH model using the values of the 95% confidence interval was below 0.8%, while incorporation of 95% confidence interval values of k0 in the Modified Cerf model only showed <0.1% deviation from the R2 obtained from the average k0 value. The modified Cerf model showed to be highly unsensitive for the observed variation in k0 .

3.6. Presence of resistant subpopulations, implications for drinking water disinfection Inactivation curves showing tailing during disinfection of viruses have been observed regularly. The EFH model, empirically designed to calculate Ct values can describe the inactivation curve accurately, but EFH parameters k, n and m are determined for specific conditions tested. Using these constants for extrapolation of the EFH model outside the tested conditions may lead to a non-accurate calculation of Ct values due to complicated inactivation kinetics (ThurstonEnriquez et al., 2003). Furthermore, the empirical basis does not aid in translating model data to biological mechanisms occurring during inactivation. The Cerf model also calculated an accurate inactivation curve, and the mechanistic background aids in understanding inactivation kinetics. The good fit of this model is an indication for the presence of two subpopulations. Our understanding regarding tailing of viruses during disinfection is limited, but heterogeneity of micro-organism populations as major cause has been discussed (Hiatt, 1964; Najm, 2006). Presence of resistant subpopulations for viruses (as intrinsic population property or acquired after synthesis in the host) may have major implications for the disinfection efficacy of these viruses. For example, it is for viral pathogens not known whether the population in the environment exists of several subpopulations, how much of the population is represented by subpopulations (e.g. 10% or 0.1%) and how much they differ in their resistance against disinfectants. It is also not known whether they may acquire additional resistance in the environment, e.g. by attachment to naturally present particles. Finally it is unknown how resistant subpopulations behave in the environment and whether the disinfectant-resistant virus population may also be able to survive better under environmental conditions. In that case, the more resistant subpopulation will represent a significant part of the infectious virus population in source water. Viruses representing the tail of the inactivation curve require additional Ct credits to be inactivated. By applying Ct values derived from the rapidly inactivating population, the ClO2-resistant subpopulation is neglected and inactivation of this population during disinfection will be lower than expected, posing a hazard for human health. For reliable Ct values it is necessary to verify whether the viral inactivation rate is constant over the range that is targeted to inactivate at the disinfectant concentration applied (Hoff, 1986).

4. Fig. 7 e Inactivation data of MS2 after exposure to an initial concentration of 0.39 and 0.45 mg/L ClO2 (diamonds), 0.1 and 0.1 mg/L ClO2 (triangles), 0.019 and 0.02 mg/L ClO2 (squares) and 0 (hexagons) mg/L ClO2. Measured data are presented as data points, and disinfection experiments were performed in duplo (closed symbols, first experiment, open symbols, duplo experiment). For each concentration, the best fit of the MCW model (dash-dot-dash line), the EFH model (dashed line) and the Modified Cerf model (dotted line) to the measured data are plotted.

Conclusions

ClO2 inactivated bacteriophage MS2 for at least 5 log units after exposure to ClO2 concentrations from 0.5 mg/L down to 0.02 mg/L. Exposure to 0.005 mg/L ClO2 resulted in the inactivation of 1 log unit in 300 min. All MS2 inactivation curves showed significant tailing, and the transfer from fast to slow inactivation occurred after 3 log units of the population had been inactivated. A diminishing biocidal effect during disinfection can be excluded as a reason for tailing. Both the EFH model and the Modified Cerf model,

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 4 7 e1 8 5 5

designed to describe non-linear curves, produced a good fit to the inactivation data. The observed tailing may be the consequence of intrinsic virus population heterogeneity, or acquired after virus synthesis by attachment to other (virus)particles. Presence of subpopulations is not taken into account when Ct values are being calculated for the inactivation of viruses from surface water sources for drinking water. This could result in passage of the disinfection train for the most resistant viruses.

Acknowledgements This study was financed by the Dutch water supply companies as part of the joint research program (BTO). The authors thank Marijan Uytewaal-Aarts for technical assistance. The authors wish to thank Charles Haas for reading of the manuscript.

references

Alvarez, M.E., O’Brien, R.T., 1982. Mechanisms of inactivation of poliovirus by chlorine dioxide and iodine. Appl. Environ. Microbiol. 44 (5), 1064e1071. Aieta, E.M., Berg, J.D., 1986. A review of chlorine dioxide in drinking water treatment. J. Am. Water Works Assoc. June, 62e72. Anonymous, 2001. Besluit van 9 januari 2001 tot wijziging van het waterleidingbesluit in verband met de richtlijn betreffende de kwaliteit van voor menselijke consumptie bestemd water (Adaptation of Dutch drinking water legislation). Staatsblad van het Koninkrijk der Nederlanden 31, 1e53. Berg, J.D., Hoff, J.C., Roberts, P.V., Matin, A., 1988. Resistance of bacterial subpopulations to disinfection by chlorine dioxide. Res. Technol. 80 (9), 115e119. Berman, D., Hoff, J.C., 1984. Inactivation of simian rotavirus SA11 by chlorine, chlorine dioxide, and monochloramine. Appl. Environ. Microbiol. 48 (2), 317e323. Carter, M.J., 2005. Enterically infecting viruses: pathogenicity, transmission and significance for food and waterborne infection. J. Appl. Microbiol. 98 (6), 1354e1380. Cerf, O., 1977. Tailing of survival curves of bacterial spores. J. Appl. Bacteriol. 42 (1), 1e19. Chen, Y.S., Vaughn, J.M., 1990. Inactivation of human and simian rotaviruses by chlorine dioxide. Appl. Environ. Microbiol. 56 (5), 1363e1366. de Roda Husman, A.M., Lodder, W.J., Rutjes, S.A., Schijven, J.F., Teunis, P.F., 2009. Long-term inactivation study of three enteroviruses in artificial surface and groundwaters, using PCR and cell culture. Appl. Environ. Microbiol. 75 (4), 1050e1057. Finch, G.R., Fairbairn, N., 1991. Comparative inactivation of poliovirus type 3 and MS2 coliphage in demand-free phosphate buffer by using ozone. Appl. Environ. Microbiol. 57 (11), 3121e3126. Fletcher, I.J., Hemmings, P., 1985. Determination of chlorine dioxide in potable waters using chlorophenol red. Analyst 110 (6), 695e699.

1855

Gerba, C.P., 1984. Applied and theoretical aspects of virus adsorption to surfaces. Adv. Appl. Microbiol. 30, 133e168. Gerba, C.P., Nwachuku, N., Riley, K.R., 2003. Disinfection resistance of waterborne pathogens on the United States Environmental Protection Agency’s Contaminant Candidate List (CCL). J. Water Supply Res. Technol.-Aqua 52, 81e94. Gyu¨re´k, L.L., Finch, G.R., 1998. Modeling water treatment chemical disinfection kinetics. J. Environ. Eng. 124, 783e793. Haas, C.N., Joffe, J., 1994. Disinfection under dynamic conditions: modification of Hom’s model for decay. Environ. Sci. Technol. 28 (7), 1367e1369. Hiatt, C.W., 1964. Kinetics of the inactivation of viruses. Bacteriol. Rev. 28, 150e163. Hoff, J., 1986. Inactivation of Microbial Agents by Chemical Desinfectants. Project Summary, EPA/600/S2e86/067. United States Environmental Protection Agency, Washington DC, USA. IAWPRC Study Group on Health Related Water Microbiology, 1991. Bacteriophages as model viruses in water quality control. Wat. Res. 25 (5), 529e545. ISO 10705e1, Enumeration of F-specific RNA Bacteriophages, 1995. International Organization for Standardization, Geneva, Switzerland. Keswick, B.H., Satterwhite, T.K., Johnson, P.C., DuPont, H.L., Secor, S.L., Bitsura, J.A., Gary, G.W., Hoff, J.C., 1985. Inactivation of Norwalk virus in drinking water by chlorine. Appl. Environ. Microbiol. 50 (2), 261e264. Lim, M.Y., Kim, J.M., Lee, J.E., Ko, G., 2010a. Characterization of ozone disinfection of murine norovirus. Appl. Environ. Microbiol. 76 (4), 1120e1124. Lim, M.Y., Kim, J.M., Lee, J.E., Ko, G., 2010b. Disinfection kinetics of murine norovirus using chlorine and chlorine dioxide. Wat. Res. 44, 3243e3251. Lodder, W.J., de Roda Husman, A.M., 2005. Presence of noroviruses and other enteric viruses in sewage and surface waters in The Netherlands. Appl. Environ. Microbiol. 71 (3), 1453e1461. Najm, I., 2006. An alternative interpretation of disinfection kinetics. J. AWWA 98 (10), 93e101. Rook, J.J., 1974. Formation of haloforms during chlorination of natural water. Water Treat. Exam 23, 234e245. Simonet, J., Gantzer, C., 2006. Degradation of the Poliovirus 1 genome by chlorine dioxide. J. Appl. Microbiol. 100 (4), 862e870. Sobsey, M.D., 1989. Inactivation of health-related microorganisms in water by disinfection processes. Water Sci. Technol. 21, 179e195. Thurston-Enriquez, J.A., Haas, C.N., Jacangelo, J., Gerba, C.P., 2003. Chlorine inactivation of adenovirus type 40 and feline calicivirus. Appl. Environ. Microbiol. 69 (7), 3979e3985. Thurston-Enriquez, J.A., Haas, C.N., Jacangelo, J., Gerba, C.P., 2005. Inactivation of enteric adenovirus and feline calicivirus by chlorine dioxide. Appl. Environ. Microbiol. 71 (6), 3100e3105. USEPA, 2006. LT2ESWTR Long Term Second Enhanced Surface Water Treatment Rule; Final Rule. United States Environmental Protection Agency, Washington DC, USA. Urakami, H., Ikarashi, K., Okamoto, K., Abe, Y., Ikarashi, T., Kono, T., Konagaya, Y., Tanaka, N., 2007. Chlorine sensitivity of feline calicivirus, a norovirus surrogate. Appl. Environ. Microbiol. 73 (17), 5679e5682. Xiong, R., Xie, G., Edmondson, A.E., Sheard, M.A., 1999. A mathematical model for bacterial inactivation. Int. J. Food Microbiol. 46 (1), 45e55.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 5 6 e1 8 6 4

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Virus disinfection in water by biogenic silver immobilized in polyvinylidene fluoride membranes Bart De Gusseme a, Tom Hennebel a, Eline Christiaens a, Hans Saveyn b, Kim Verbeken c, Jeffrey P. Fitts d, Nico Boon a, Willy Verstraete a,* a

Department of Biochemical and Microbiological Technology, Laboratory of Microbial Ecology and Technology (LabMET), Ghent University, Coupure Links 653, B-9000 Gent, Belgium b Department of Applied Analytical and Physical Chemistry, Particle and Interfacial Technology Group (PaInt), Ghent University, Coupure Links 653, B-9000 Gent, Belgium c Department of Metallurgy and Materials Science, Ghent University, Technology Park 903, B-9052 Gent, Belgium d Environmental Sciences Department, Brookhaven National Laboratory, Building 830, Upton, NY 11973, USA

article info

abstract

Article history:

The development of innovative water disinfection strategies is of utmost importance to

Received 1 September 2010

prevent outbreaks of waterborne diseases related to poor treatment of (drinking) water.

Received in revised form

Recently, the association of silver nanoparticles with the bacterial cell surface of Lactoba-

26 November 2010

cillus fermentum (referred to as biogenic silver or bio-Ag0) has been reported to exhibit

Accepted 29 November 2010

antiviral properties. The microscale bacterial carrier matrix serves as a scaffold for Ag0

Available online 7 December 2010

particles, preventing aggregation during encapsulation. In this study, bio-Ag0 was immobilized in different microporous PVDF membranes using two different pre-treatments of

Keywords:

bio-Ag0 and the immersion-precipitation method. Inactivation of UZ1 bacteriophages using

Antimicrobial

these membranes was successfully demonstrated and was most probably related to the

Ionic silver

slow release of Agþ from the membranes. At least a 3.4 log decrease of viruses was ach-

Metallic silver

ieved by application of a membrane containing 2500 mg bio-Ag0powder m2 in a submerged

X-ray absorption spectroscopy (XAS)

plate membrane reactor operated at a flux of 3.1 L m2 h1. Upon startup, the silver concentration in the effluent initially increased to 271 mg L1 but after filtration of 31 L m2, the concentration approached the drinking water limit ( ¼ 100 mg L1). A virus decline of more than 3 log was achieved at a membrane flux of 75 L m2 h1, showing the potential of this membrane technology for water disinfection on small scale. ª 2010 Elsevier Ltd. All rights reserved.

1.

Introduction

Contamination of drinking water and the subsequent outbreak of waterborne diseases are the leading cause of death in many developing nations. Therefore, virus removal during water treatment has received more attention due to the epidemiological significance of these pathogens (Rose and Gerba, 1991). Especially for drinking water purposes, the

development of innovative water disinfection strategies is of utmost importance. Recently, enteric viruses were included in the US EPA’s new Contaminant Candidate List and thus they may become subject to the drinking water quality standards (USEPA, 2009). Significant interest has arisen in the use of silver containing nanoparticles for water disinfection (Li et al., 2008). Several authors have reported on the antiviral activity of

* Corresponding author. Tel.: þ32 9 264 59 76; fax: þ32 9 264 62 48. E-mail address: [email protected] (W. Verstraete). URL: http://www.labmet.ugent.be 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.11.046

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 5 6 e1 8 6 4

silver nanoparticles (nAg0), for example against HIV-1 (Elichiguerra et al., 2005; Rogers et al., 2008). Recently, inactivation of bacteriophages and murine noroviruses was also demonstrated in disinfection assays using biogenic silver (bio-Ag0) (De Gusseme et al., 2010b). The latter material consists of silver particles of 11.2  0.9 nm, produced by reduction of ionic silver on the bacterial cell surface of Lactobacillus fermentum (Sintubin et al., 2009). The attachment of the silver particles to a bacterial carrier matrix of micrometer scale prevents them from aggregating, which is an advantage over chemically produced nAg0 particles which tend to aggregate in aqueous solutions at high concentrations or when their average particle size is lower than 40 nm (Mafune et al., 2000). Moreover, the bacterial surface serves as a scaffold that facilitates the incorporation of biogenic metals into a supporting material, thus preventing the nanoparticles from leaching into the environment. This has been demonstrated for biogenic palladium nanoparticles (Hennebel et al., 2010, 2009a, 2009b), but not yet for biogenic silver nanoparticles. In the past decade, membrane based technologies have been increasingly applied to improve the water quality. Although viruses are smaller than the pore sizes used in microfiltration processes, they are retained by biofilms that foul the membranes (Ueda and Horan, 2000; Wu et al., 2010). The formation of these biofilms is unwanted because of the related decrease in membrane flux, increasing energy costs and shorter membrane life (McDonogh et al., 1994). To cope with this problem, silver can be applied for both biofouling and virus control, by coating nAg0 directly on the surface of the membranes or by embedment in the polymer matrix of the membrane itself (Yang et al., 2009; Zodrow et al., 2009). Zodrow et al. (2009) have prepared nAg0 incorporating membranes by means of the ‘phase inversion’ technique, during which the polymer is transformed from a liquid to a solid phase in a controlled manner (Mulder, 1991; Vankelecom, 2002). More specifically, the ‘immersionprecipitation’ method was applied, in which the solvent solution containing the polymer and nAg0 (i.e. the casting dope) was casted on a support and then immersed in a water ( ¼ non-solvent) bath to effect polymer precipitation (Vankelecom, 2002). Zodrow et al. (2009) have shown a significant improvement in the removal of bacteriophage MS2 by incorporating nAg0. Moreover, the addition of silver prevented bacterial attachment and biofilm formation. The authors suggested the release of Agþ to be the main mechanism for both outcomes. The aim of this study was to incorporate silver in polymeric membranes in order to develop a disinfecting material based on immobilized biogenic silver nanoparticles. Therefore, the immersion-precipitation technique was used to prepare the membranes, and the structure and the antiviral properties of this material were examined. The goal of this research was preparing a casting dope without addition of a dispersing agent. Two pretreatments of the biogenic silver were therefore investigated: centrifugation of a bio-Ag0 suspension (‘bio-Ag0susp’) and spraydrying to a powder (‘bio-Ag0powder’). Batch and continuous disinfection assays with the different membranes were conducted in water contaminated with the bacteriophage UZ1, a model for enteric viruses (Verthe´ et al., 2004).

2.

Materials and methods

2.1.

Production of biogenic silver

1857

Bio-Ag0 was produced with L. fermentum LMG 8900 (LMG culture collection, Ghent University, Belgium) according to Sintubin et al. (2009). Briefly, after L. fermentum biomass was harvested by subsequent centrifugation and washing in deionized water, the pH was increased with NaOH to 11.5, followed by the addition of a diamine silver complex in a ratio of 1:4.6 Ag:CDW (cell dry weight). After 24 h, the produced biogenic silver was harvested by centrifugation and resuspended in deionized water. The resulting bio-Ag0 suspension in water contained 15.890 g Ag L1 as determined by atomic absorption spectroscopy (AAS) after digestion (see below), and is referred to as ‘bio-Ag0susp’. The nanoparticles had a particle size of 11.2  10.9 nm (Sintubin et al., 2009). This suspension was fed into a Production Minor spraydryer (Gea Niro, Soeberg, Denmark) with main spray chamber of 2700 mm  1400 mm. The feed flow rate used was 185 cm3 min1, and the inlet and outlet air temperatures were 200  C and 95  C. The atomization head was operated at 15,272 rpm. A dry powder (‘bio-Ag0powder’) was obtained, containing 140 mg Ag g1 as determined by AAS after digestion (see below).

2.2.

Membrane fabrication

Both bio-Ag0susp and bio-Ag0powder (defined in section 2.1) were used to produce polyvinylidene fluoride (PVDF) membranes. In the case of bio-Ag0susp, the suspension was centrifuged in 50 mL test tubes (10,000g for 15 min) and the supernatant was removed. The remaining solids in the tubes (i.e. the wet pellet) were dispersed in N,N-dimethylformamide (DMF). In the case of bio-Ag0powder, the powder was directly added to DMF. Eventually, two DMF solutions were obtained for each of the two bio-Ag0 formulations, containing 1000 and 10,000 mg Ag L1. The bio-Ag0 particles were dispersed at 50  C in an Elmasonic S30H ultrasonic bath (Elma, Singen, Germany) for 10 min. To each of the four dispersions 14 wt.% PVDF (Kynar 500, Arkema, Amsterdam, The Netherlands) was slowly added under vigorous stirring at 50  C. Finally, the dispersions were placed in an ultrasonic bath (50  C, 10 min), in order to prepare homogeneous casting dopes. A similar casting dope without bio-Ag0 was prepared as a control. Membranes were fabricated using the direct immersionprecipitation method at 25  C, as previously described (Hennebel et al., 2010). 30 mL of a casting dope was spread uniformly on a 0.12 m2 non-woven support (Novatex FO 2471, Freudenberg, Germany) by means of an automatic film applicator (Elcometer 4340, Hermalle-sous-Argenteau, Belgium). The resulting wet film thickness was 250 mm (controlled by an Elcometer 3530 casting knife). DMF was allowed to evaporate for 30 s, and subsequently, the nascent film was immersed into a deionized water bath to allow polymer precipitation. The theoretical maximum silver content in the membranes made with the 1000 mg Ag L1 and 10,000 mg Ag L1 casting dopes, was 250 mg bio-Ag0 m2 and 2500 mg bio-Ag0 m2, respectively.

1858

2.3.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 5 6 e1 8 6 4

Membrane characterization

Hydrophobicity of the membranes was determined by sessile drop contact angle measurements. After deposition of a water drop of 3 mL on the dry membrane surface, the contact angles were determined using the Kru¨ss DSA10 drop shape analysis system (Sysmex, Hoeilaart, Belgium) and DSA Software 1.80. Values are the mean of 6 measurements per membrane surface. Membrane zeta potential was determined by measuring the streaming potential with a laboratory scale filter press according to Saveyn et al. (2005). A Fluke 189 Multimeter was connected to the cathode and anode side. All experiments were conducted in a 5 mM KNO3 background solution at pH 7. Membrane morphology and bio-Ag0 localization were studied on the top surfaces and cross sections of dried membranes, by means of scanning electron microscopy (SEM). The membrane cross sections were obtained by fracturing the membrane after immersion in liquid N2. The samples were sputter-coated with an ultrathin Au layer (Baltec AG, Balzers, Liechtenstein) and analyzed under vacuum with a FEI XL30 SEM (FEI, Eindhoven, The Netherlands) equipped with a LaB6 filament and an energy dispersive X-ray spectroscope (EDAX, Tilburg, The Netherlands).

2.4.

Growth and detection of bacteriophage UZ1

The stock of bacteriophage UZ1 was prepared as previously described (De Gusseme et al., 2010b). To detect phages, the soft agar layer method described by Adams (1959) was applied, using serial tenfold dilutions of the samples in SM medium (6.1 g L1 TriseHCl, 5.8 g L1 NaCl, 1.2 g L1 MgSO4, 0.1 g L1 gelatin (Oxoid, Basingstoke, UK), pH 7.5) and a mid-log phase Enterobacter aerogenes BE1 culture LMG 22092 (LMG culture collection, Ghent University, Belgium) (De Gusseme et al., 2010a). Phages were counted as plaque forming units (pfu) and the phage concentration was expressed as pfu mL1. The limit of detection (LOD) was determined at 1.0  102 pfu mL1. When no viruses were detected in the samples, the LOD was used as a conservative estimate for the UZ1 concentration.

2.5. Virus inactivation in batch by bio-Ag0 immobilized in PVDF membranes All experiments were performed in bottled natural source water (Spa Blauw, Spadel, Brussels, Belgium) to which no free chlorine was added to avoid virus inactivation. The composition of the water was similar as previously reported (De Gusseme et al., 2010b). The batch experiments were conducted at 25  C in sterilized 500 mL flasks on a magnetic stirrer. 25 cm2 of the 250 mg bio-Ag0susp m2, the 2500 mg bio-Ag0susp m2, and the 2500 mg bio-Ag0powder m2 membranes were incubated in 200 mL water spiked with appr. 106 pfu mL1. Samples were taken at 0 h, 2 h and 24 h, and stored at 4  C for further analysis. Silver in the samples for UZ1 detection was immediately quenched with an excess of thioglycolate-thiosulfate neutralizer (Tilton and Rosenberg, 1978).

2.6. Virus inactivation by bio-Ag0 immobilized in PVDF membranes in a membrane reactor The membrane sheets were fixed on both sides of a 0.06 m2 plate cartridge with a spacer in between, rendering a total

membrane area of 0.12 m2. The plate membrane was submerged in a polycarbonate membrane reactor with a working volume of 8 L (Fig. S1). Before startup, the reactor vessel was filled with 8 L of the influent, natural source water spiked with phage UZ1 (Cinfl ¼ 106 pfu mL1). After installation of the membrane, the influent was further dosed at the bottom of the reactor and mixed by a pressurized air diffuser. The effluent pump was working semi-continuously, thus providing relaxation periods for the membrane. The airflow along the membrane and the relaxation periods facilitated the flux through the membrane (cross-flow principle). Reactor runs for each plate membrane were conducted at a flow rate and hydraulic retention time (HRT) of 0.375 L h1 and 1 d, respectively. In a subsequent reactor run, an additional 2500 mg bioAg0powder m2 plate membrane of the same size was used at a flow rate of 8 L h1, resulting in a HRT of 1 h. Influent, bulk and effluent samples were taken at regular intervals and stored at 4  C for further analysis. The filtrate sample at time 0 h was taken immediately after installation of the membranes. Silver in the samples for UZ1 detection was immediately quenched.

2.7.

Silver measurements

The concentration of silver in the bio-Ag0susp and bio-Ag0powder stock solutions was measured by atomic absorption spectroscopy (AAS) (Shimadzu AA-6300, Japan), after digestion according to Sintubin et al. (2009). The LOD of AAS was 0.1 mg L1. The concentration of silver in the samples of the batch and reactor experiments were determined by means of ICP-MS (Elan DRC-e, Perkin Elmer, MA, USA). The LOD of ICP-MS was 0.8 mg L1.

3.

Results

3.1.

Characterization of the PVDF membranes

Both formulations of biogenic silver, bio-Ag0susp and bioAg0powder, were immobilized in PVDF membranes using the direct immersion-precipitation method. Spray-drying of the bio-Ag0 did not alter the chemical oxidation state of the metallic silver particles (see X-ray absorption spectroscopy analyses in Supplementary Data and Fig. S2). The membranes impregnated with bio-Ag0powder had a similar contact angle and zeta potential as the membrane without bio-Ag0, regardless of bio-Ag0powder concentration (Table 1). This contrasted with the bio-Ag0susp membranes, which had an increased contact angle, thus making them more hydrophobic. In addition, there was a decline in the absolute value of the membrane zeta potential at both bio-Ag0susp concentrations (Table 1). SEM images of the bio-Ag0susp membrane in the backscattered electron (BSE) mode revealed the presence of silver containing structures, visible as enlightened dots at the rough and hydrophobic surface (Fig. 1A). By larger magnification, individual entities of nanoparticles could still be observed within these agglomerates. The presence of silver was confirmed by EDX (Fig. S3A) and the diameter of the agglomerates ranged from tens to hundreds of nanometers. Similar clusters were observed as bright structures throughout the cross section of the membrane in the BSE mode (Fig. 1B and Fig. S3B). The cross section revealed a uniform sponge-like structure, built by

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 5 6 e1 8 6 4

Table 1 e Hydrophobicity of the different PVDF membranes expressed as contact angle (n [ 6) and the membrane zeta potentials (n [ 3). Values represent the means and their standard deviation. Membrane 0 mg bio-Ag0 m2 250 mg bio-Ag0powder m2 2500 mg bio-Ag0powder m2 250 mg bio-Ag0susp m2 2500 mg bio-Ag0susp m2

Contact angle ( ) 74.7 75.4 73.7 123.6 125.8

 4.0  6.9  0.7  3.4  1.8

Zeta potential (mV) 2.98 3.78 2.97 0.60 0.36

 0.80  0.48  0.71  0.18  0.06

1859

densely packed polymeric crystallites with micropores in between. The thickness of the top layer of the bio-Ag0susp membrane amounted to 35 mm. No silver agglomerates were observed in the BSE mode on the top surface of the membranes with bio-Ag0powder (Fig. 1C). Although the presence of Ag on the hydrophilic surface was demonstrated by means of EDX (Fig. S3C), the amount was 3e5 times smaller compared to the membranes containing bioAg0susp. The top surface was smoother and showed micropores with a diameter between 60.9 and 338.0 nm. In the 95 mm thick cross section, the formation of a dense skin layer and a cellular sublayer was observed (Fig. 1D). The asymmetry of this type of

Fig. 1 e SEM images of the PVDF membranes. Figure (A) and (B) present the top surface and the cross section of the membrane containing 2500 mg bio-Ag0susp mL2, respectively. In (C) and (D) the top surface and the cross section of the PVDF membrane containing 2500 mg bio-Ag0powder mL2 are shown, respectively. (E) is a higher magnification picture of the pores of the latter membrane, and (F) represents the EDX spectrum of this detail. The peaks C and F relate to the PVDF matrix. White arrows indicate agglomerates of biogenic silver, visible as enlightened dots in the BSE mode. The inserts in (A) and (C) are the profile of a water droplet on the top surface of the membranes.

1860

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 5 6 e1 8 6 4

membrane was apparent: in the skin layer, pore channels with a small diameter of hundreds of nanometers were present, whereas the sublayer consisted of larger columned pores of several micrometers. In both layers, silver agglomerates were visible as bright dots, as demonstrated by larger magnification (Fig. 1E). The presence of silver throughout this membrane cross section was confirmed by EDX (Fig. 1F). Within these agglomerates, individual particles have been observed, suggesting that the larger silver clusters are in fact the bacterial cell carrier covered with silver nanoparticles.

3.2. Virus inactivation in batch by bio-Ag0 immobilized in PVDF membranes The antiviral properties of the bio-Ag0 containing membranes were first examined in batch experiments. Submersion of 25 cm2 of the 2500 mg bio-Ag0susp m2 membrane resulted in a 4.1 log decrease after 24 h of indirect contact with immobilized bio-Ag0 (Table 2). To examine the influence of the embedded silver concentration, the same surface area of the 250 mg bioAg0susp m2 membrane was submerged in contaminated water but no virus removal was observed (Table 2). Consistent with these results, analyses of the silver concentrations in the water revealed greater release of Agþ from the 2500 mg bio-Ag0susp m2 membrane (Table 2). To investigate the influence of the biogenic silver formulation, 25 cm2 of the membrane impregnated with 2500 mg bio-Ag0powder m2 was submersed in the suspension as well. The Agþ release was lower than with the same concentration of bio-Ag0susp, yet, the virus inactivation was similar, resulting in at least a 4.3 log decline after 24 h of indirect contact with the immobilized bio-Ag0 (Table 2). Moreover, immobilization of 2500 mg m2 of both types of bio-Ag0 in PVDF membranes seemed to be efficient for more than 3 log virus decrease after only 2 h under the given test conditions.

3.3. Virus inactivation by bio-Ag0 immobilized in PVDF membranes in a continuous membrane reactor Continuous disinfection experiments were conducted in a membrane reactor by filtering contaminated water through a submerged plate, holding the bio-Ag0 PVDF membranes. When the 2500 mg bio-Ag0susp m2 membrane was used in the reactor, no viruses were detected in the effluent after filtration

of 6.3 L m2, yielding at least a 3.6 log decrease (Fig. 2A). Interestingly, the virus inactivation in the bulk phase of the membrane unit was the same as in the filtrate (data not shown). Even after filtration of 225 L m2, a 2.9 log decline could still be achieved using this membrane. In contrast, the virus removal achieved with the control membrane without bio-Ag0 and the membrane containing 250 mg bio-Ag0susp m2 amounted to no more than 1 log virus removal. Similar results were achieved in the filtration experiments with the membranes containing bio-Ag0powder (Fig. 2B). No viruses were detected in the bulk liquid or the filtrate when the 2500 mg bioAg0powder membrane was used in the reactor, even after filtration of 225 L m2. In the case of the 250 mg bio-Ag0powder m2 however, a maximum virus removal of only 1.2 log was observed. In each reactor run, the transmembrane pressure (TMP) was not higher than 50 mbar. No differences in TMP between the control membrane and the silver containing membranes were observed. Leaching of Agþ from the 250 mg bio-Ag0 m2 membranes was very low, with 10.7 mg L1 being the highest concentration detected in the filtrate (data not shown). Higher concentrations of Agþ were encountered in the effluent of the filtration experiments with the 2500 mg bio-Ag0 m2 membranes (Fig. 3). At the startup of the membrane reactor, the Agþ concentrations released by the 2500 mg bio-Ag0susp m2 and the 2500 mg bio-Ag0powder membranes reached concentrations up to 236 and 271 mg L1, respectively. In both cases, the Agþ concentration in the effluent decreased when more water was filtered. Yet, after filtration of 225 L m2 through the bioAg0susp containing membrane, 102 mg Agþ L1 was still detected in the effluent. The decrease of the Agþ concentration in the effluent was much faster in the case of the bio-Ag0powder containing membrane. After filtration of 75 L m2 the Agþ release was reduced to 57 mg L1, whereas no more then 24 mg L1 was detected in the effluent of the 2500 mg bioAg0powder membrane after filtration of 225 L m2. Subsequently, another 2500 mg bio-Ag0powder m2 membrane was applied in the membrane unit and the HRT was decreased from one day to 1 h, resulting in a membrane flux of 75 L m2 h1. At startup of the reactor, no viruses were detected in the filtrate, thus yielding at least a 3.9 log decrease (Fig. 4). After filtration of 112.5 L m2 however, low concentrations of phages were detected again in the bulk phase (data not shown)

Table 2 e Virus inactivation and AgD release by 25 cm2 of the different PVDF membranes in a 24 h batch experiment. The concentration of UZ1 is expressed as the mean value of triplicate plaque assays (±standard deviation). The LOD was determined at 1.0 3 102 pfu mLL1. Membrane

250 mg bioAg0susp m2 2500 mg bioAg0susp m2 2500 mg bioAg0powder m2

0h

2h þ

UZ1 concentration (pfu mL1)

Agþ release (mg L1)

42

3.7  0.9  106

16  4

3.4  5.4  103

208  5

3.3  3.2  102

792  10

4.9  6.2  102

27  8

< LOD

95  6

UZ1 concentration (pfu mL1)

Ag release (mg L1)

UZ1 concentration (pfu mL1)

6.5  2.7  106

NDa

4.6  0.6  106

3.8  0.8  106

ND

2.4  0.5  106

ND

a ND: not detected (below LOD ¼ 0.8 mg L1).

24 h þ

Ag release (mg L1)

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 5 6 e1 8 6 4

1861

Fig. 3 e Ionic silver released to membrane reactor effluent, as a function of volume filtered through the PVDF membranes containing 2500 mg bio-Ag0susp mL2 and 2500 mg bio-Ag0powder mL2. The HRT of the membrane reactor was 1 day. Error bars represent the standard deviations of triplicate effluent sample measurements.

Fig. 2 e Virus inactivation as a function of the filtered volume through (A) PVDF membranes containing 0, 250 and 2500 mg bio-Ag0susp mL2, and (B) 250 and 2500 mg bioAg0powder mL2. The HRT of the membrane reactor was 1 day and the maximum detectable virus removal amounted to 3.4 log. Cinfluent [ concentration of UZ1 in the influent; Ceffluent [ concentration of UZ1 in the effluent. Error bars represent the standard deviations of triplicate influent and effluent sample measurements.

was demonstrated. Batch experiments showed at least a 4 log decrease of the phages by incorporating 2500 mg bio-Ag0 m2 in the membranes. Agþ was released from the membranes and its interaction with the UZ1 phages was most probably the main virus inactivation mechanism. Other researchers have demonstrated that AgNO3 was also effective at inactivating the MS2 and T2 phages and the monkeypox virus (Kim et al., 2008; Richards, 1981; Rogers et al., 2008). Although the antimicrobial action by silver nanoparticles is not fully understood, several studies indicate the importance of Agþ release and the subsequent interaction with thiol groups of bacterial

and in the filtrate (Fig. 4). Yet, a virus decline of more than 3 log was still achieved, even after filtration of 300 L m2. The pattern of Agþ release in the filtrate was similar to the experiments at an HRT of one day. At the startup of the reactor, 223 mg Agþ L1 was detected in the effluent, followed by a decrease towards 95 mg L1 after filtration of 300 L m2. At an HRT of 1 h, the TMP varied between 50 and 100 mbar.

4.

Discussion

4.1. Antiviral action of biogenic silver incorporated PVDF membranes In this study, the applicability of PVDF membranes with impregnated biogenic silver for virus inactivation in water

Fig. 4 e Virus inactivation and ionic silver release in the membrane reactor effluent, as a function of the volume filtered through the PVDF membrane containing 2500 mg bio-Ag0powder mL2. The HRT of the membrane reactor was 1 day and the maximum detectable virus removal amounted to 3.9 log. Cinfluent [ concentration of UZ1 in the influent; Ceffluent [ concentration of UZ1 in the effluent. Error bars represent the standard deviations of triplicate influent and effluent sample measurements.

1862

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 5 6 e1 8 6 4

enzymes and (glyco)proteins exposed on the bacterial or viral surface, or the reaction with phosphorus containing nucleic acids (Feng et al., 2000; Liau et al., 1997; Matsumura et al., 2003). In addition, the direct interaction of viruses with silver nanoparticles has been suggested to physically obstruct the virus-host cell binding (Elichiguerra et al., 2005; Rogers et al., 2008). However, the latter reaction mechanism seems unlikely to be responsible for the virus inactivation in the present work, since the bio-Ag0 particles remained anchored in the polymeric matrix.

4.2. The influence of the embedded concentration and the pre-treatment of biogenic silver on Agþ release by the membranes From the results of the batch experiments, it was clear that the antiviral properties of the membranes increased with a higher embedded concentration of bio-Ag0. Whereas immobilization of 250 mg bio-Ag0 m2 in the membrane did not result in a significant virus inactivation, the incorporation of 2500 mg bio-Ag0 m2 clearly enhanced virus disinfection. For both formulations, i.e. bio-Ag0susp and bio-Ag0powder, similar disinfection efficiencies were obtained at the high concentration. The main difference between the two types was reflected in the Agþ release, with 2500 mg bio-Ag0susp m2 being the membrane that produced the most ionic silver. The difference in the two types of pre-treatment resulted in different membrane structures and characteristics as well. Because the oxidation state of the silver nanoparticles themselves remained unaffected after both pre-treatments, the differences tend to be related to the preparation of the casting dopes. Fabrication of the membranes by means of a casting dope with spray-dried bio-Ag0 (‘bio-Ag0powder’) did not alter the contact angle and the zeta potential of the membranes, compared to the PVDF membrane without silver. The measured contact angles were in the range of previously reported values for PVDF based surfaces, i.e. 75 e82 (Bormashenko et al., 2006; Peng et al., 2005). Also the negative surface charge of the membranes remained unaffected by the addition of bioAg0powder, which was also demonstrated by Zodrow et al. (2009). Yet, those authors noticed a decrease in hydrophobicity due to the addition of chemically produced nAg0 nanoparticles. The structure of the membranes with bio-Ag0powder was typically asymmetric, which can be explained by the fact that no traces of water (the non-solvent) were present in the casting dope. Subsequent immersion in a water bath likely resulted in a fast liquideliquid demixing prior to polymer precipitation, as described by Cheng et al. (1999). As a result, the membranes showed an asymmetric morphology characterized by a skin layer and a cellular sublayer. On the contrary, when a wet pellet of biogenic silver particles (‘bio-Ag0susp’) was dispersed in DMF after centrifugation of the bio-Ag0 suspension, the resulting casting dope contained a fraction of water (the non-solvent). Submersion of these kinds of dopes increases the crystallization rate of PVDF and a more uniform and dense membrane is formed (Cheng et al., 1999), as observed in the cross section of the bio-Ag0susp based membrane. This process can be enhanced by submersion in a non-solvent bath with a fraction of solvent, a so-called ‘soft precipitation bath’ (Cheng et al., 1999; Peng et al., 2005). The top

surface of these sponge-like membranes is known to exhibit greatly enhanced hydrophobicity and a certain roughness (Peng et al., 2005), which was confirmed by the larger contact angles measured and SEM images in this study. The addition of bio-Ag0susp to the membrane decreased the absolute value of the zeta potential as well. Calculation of the surface charge density according to Hiemenz and Rajagopalan (1997), indicated that the amount of charges per m2 on the 2500 mg bio-Ag0susp m2 was ten times lower than on the 2500 mg bio-Ag0powder m2 (0.62  109 vs. 6.46  109 mol m2, respectively). This is likely due to the greater ionic interaction with Agþ, initially present in the pellet of bio-Ag0susp, and could explain the susbsequent greater release of Agþ by the bio-Ag0susp membrane. Moreover, the surface of the bioAg0susp membrane contained more biogenic silver, which likely also facilitated Agþ release. Future research is needed to fully understand how the membrane structure and characteristics can influence the Agþ release from incorporated biogenic silver and the subsequent disinfection of water.

4.3. Application of the membranes for continuous disinfection The membranes with 2500 mg bio-Ag0 m2 were successfully used for continuous disinfection of the UZ1 phage in a submerged plate membrane reactor. The fact that similar virus disinfection efficiencies were observed in both the bulk phase and the effluent strongly suggests that direct interaction with the silver particles during passage through the membrane is not the main antiviral mechanism. Ionic silver release was probably the main antiviral mechanism in the continuous disinfection experiments as well. Significant Agþ concentrations were found in the case of the 2500 mg bioAg0 m2 membranes while the Agþ release by the 250 mg bioAg0 m2 membrane and the concomitant virus inactivation efficiencies were low. Again, the Agþ release by the bio-Ag0susp membrane was more than what was observed for the bioAg0powder membrane. The control of the Agþ release by the membrane is important since the drinking water threshold for Agþ is limited to 0.1 mg L1 by the WHO (2004). From this point of view, the use of bio-Ag0powder to fabricate antiviral PVDF membranes seems the best approach since the Agþ release was around or below the threshold value after a certain filtration period (31 L m2). Also from an economical point of view, a controlled release of Agþ is wanted since rapid depletion of the membrane-associated silver might compromise its long-term performance. Indeed, other researchers have demonstrated that the antimicrobial activity of silver loaded membranes and hollow fibers were greatly decreased due to a decrease of the silver content (Chou et al., 2005; Zodrow et al., 2009). Plate membranes lost their antimicrobial properties after leaching from the membrane surface layer stopped, even though 90% of the added silver particles still remained in the membrane (Zodrow et al., 2009). The application of bio-Ag0powder containing membranes might overcome this problem since agglomerates of silver were also encountered at the pore walls inside the membrane, from where they can also contribute to the release of Agþ for virus disinfection. However, since Agþ is slowly released from the membranes, the disinfection effect is expected to decrease over

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 5 6 e1 8 6 4

time. This means that in practice, these membranes can be applied to treat limited volumes of contaminated water (e.g. in refugee camps) but that they cannot be used for continuous long-term and large-scale drinking water production. The membranes must be regarded as a gentle way to provide temporarily low amounts of ionic silver without the risk of releasing nanoparticles, comparable with other slow release techniques for small scale treatment such as polymer-iodine tablets (slowly releasing iodine) or sodium dichloroisocyanurate tablets (slowly producing hypochlorous acid) (Clasen and Edmondson, 2006; Mazumdar et al., 2010). The results of the batch experiments indicate that the antiviral activity of the bio-Ag0 membranes depends on both the released Agþ dose and the contact time, meaning that the HRT is likely an important operational parameter for continuous experiments. Given that > 3 log decline was achieved after only 2 h in batch, an even shorter contact time (1 h HRT) was assessed in the continuous experiments with the 2500 mg bio-Ag0powder m2 membrane. Even at a flux as high as 75 L m2 h1, high virus inactivation efficiencies were obtained at low TMP values. Yet, viruses were again detected in the effluent after filtration of 112.5 L m2, even when the Agþ concentration was as a high as 126 mg L1. This contrasts with the results from the 1 d HRT experiments in which the UZ1 concentration remained below the LOD at Agþ levels lower than 100 mg L1. More data on the combined effect of Agþ concentration and contact time are needed before one can model adequately the optimal combination between reactor residence time and silver concentration in the membranes.

5.

Conclusions

The results of this work show that a bacterial carrier matrix containing silver nanoparticles can be successfully immobilized in microporous membranes where it provides for potent antiviral activity in a submerged plate membrane reactor. Inactivation of UZ1 bacteriophages using these membranes was successfully demonstrated at low and high membrane fluxes, and was most probably related to the slow release of Agþ from the membranes. By increasing our understanding of how the membrane structure affects the release rate of Agþ from bioAg0, the Agþ concentration in the filtrate could be further decreased and the depletion of Agþ from the material should be controlled. Yet, long-term filtration experiments are required to confirm that this does not compromise the antimicrobial activity. It is expected that due to the gradual release of Agþ from the membranes, their antiviral activity will decrease over time. Therefore, it can be concluded that this technique can be applied for the treatment of limited volumes of contaminated water but not for long-term drinking water production.

Acknowledgements This work was supported by a PhD grant (B. De Gusseme) and project grant no. 7741-02 (T. Hennebel) of the Research Foundation Flanders (FWO). It was part of project no. G.0808.10N (2010-2013), funded by the FWO. K. Verbeken is

1863

a postdoctoral fellow with the FWO. We gratefully thank Griet Vermeulen, Jan Dick, Pieter Spanoghe and Peter Mast for their technical assistance and Elke De Clerck (Janssen Pharmaceutica NV, Beerse, Belgium) for kindly providing the spray-dried biogenic silver. We acknowledge Anthony Hay, Simon De Corte, Liesje Sintubin and Willem De Muynck for critically reviewing this manuscript and the many helpful suggestions.

Appendix A. Supplementary material Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.watres.2010.11.046.

references

Adams, M.H., 1959. Bacteriophages. Interscience, New York. Bormashenko, E., Stein, T., Whyman, G., Bormashenko, Y., Pogreb, R., 2006. Wetting properties of the multiscaled nanostructured polymer and metallic superhydrophobic surfaces. Langmuir 22 (24), 9982e9985. Cheng, L.P., Lin, D.J., Shih, C.H., Dwan, A.H., Gryte, C.C., 1999. PVDF membrane formation by diffusion-induced phase separation-morphology prediction based on phase behavior and mass transfer modeling. Journal of Polymer Science Part B-Polymer Physics 37 (16), 2079e2092. Chou, W.L., Yu, D.G., Yang, M.C., 2005. The preparation and characterization of silver-loading cellulose acetate hollow fiber membrane for water treatment. Polymers for Advanced Technologies 16 (8), 600e607. Clasen, T., Edmondson, P., 2006. Sodium dichloroisocyanurate (NaDCC) tablets as an alternative to sodium hypochlorite for the routine treatment of drinking water at the household level. International Journal of Hygiene and Environmental Health 209 (2), 173e181. De Gusseme, B., Du Laing, G., Hennebel, T., Renard, P., Chidambaram, D., Fitts, J.P., Bruneel, E., Van Driessche, I., Verbeken, K., Boon, N., Verstraete, W., 2010a. Virus removal by biogenic cerium. Environmental Science & Technology 44, 6350e6356. De Gusseme, B., Sintubin, L., Baert, L., Thibo, E., Hennebel, T., Vermeulen, G., Uyttendaele, M., Verstraete, W., Boon, N., 2010b. Biogenic silver for disinfection of water contaminated with viruses. Applied and Environmental Microbiology 76 (4), 1082e1087. Elichiguerra, J.L., Burt, J.L., Morones, J.R., Camacho-Bragado, A., Gao, X., Lara, H.H., Yacaman, M.J., 2005. Interaction of silver nanoparticles with HIV-1. Journal of Nanobiotechnology 3 (1). doi:10.1186/1477-3155-1183-1186. Feng, Q.L., Wu, J., Chen, G.Q., Cui, F.Z., Kim, T.N., Kim, J.O., 2000. A mechanistic study of the antibacterial effect of silver ions on Escherichia coli and Staphylococcus aureus. Journal of Biomedical Materials Research 52 (4), 662e668. Hennebel, T., De Gusseme, B., Boon, N., Verstraete, W., 2009a. Biogenic metals in advanced water treatment. Trends in Biotechnology 27 (2), 90e98. Hennebel, T., Verhagen, P., Simoen, H., De Gusseme, B., Vlaeminck, S.E., Boon, N., Verstraete, W., 2009b. Remediation of trichloroethylene by bio-precipitated and encapsulated palladium nanoparticles in a fixed bed reactor. Chemosphere 76 (9), 1221e1225. Hennebel, T., De Corte, S., Vanhaecke, L., Vanherck, K., Forrez, I., De Gusseme, B., Verhagen, P., Verbeken, K., Van der Bruggen, B., Vankelecom, I., Boon, N., Verstraete, W., 2010.

1864

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 5 6 e1 8 6 4

Removal of diatrizoate with catalytically active membranes incorporating microbially produced palladium nanoparticles. Water Research 44 (5), 1498e1506. Hiemenz, P.C., Rajagopalan, R., 1997. Principles of Colloid and Surface Chemistry. Marcel Dekker, New York. Kim, J.Y., Lee, C., Cho, M., Yoon, J., 2008. Enhanced inactivation of E. coli and MS-2 phage by silver ions combined with UV-A and visible light irradiation. Water Research 42 (1e2), 356e362. Li, Q.L., Mahendra, S., Lyon, D.Y., Brunet, L., Liga, M.V., Li, D., Alvarez, P.J.J., 2008. Antimicrobial nanomaterials for water disinfection and microbial control: potential applications and implications. Water Research 42 (18), 4591e4602. Liau, S.Y., Read, D.C., Pugh, W.J., Furr, J.R., Russell, A.D., 1997. Interaction of silver nitrate with readily identifiable groups: relationship to the antibacterial action of silver ions. Letters in Applied Microbiology 25 (4), 279e283. Mafune, F., Kohno, J., Takeda, Y., Kondow, T., Sawabe, H., 2000. Structure and stability of silver nanoparticles in aqueous solution produced by laser ablation. Journal of Physical Chemistry B 104 (35), 8333e8337. Matsumura, Y., Yoshikata, K., Kunisaki, S., Tsuchido, T., 2003. Mode of bactericidal action of silver zeolite and its comparison with that of silver nitrate. Applied and Environmental Microbiology 69 (7), 4278e4281. Mazumdar, N., Chikindas, M.L., Uhrich, K., 2010. Slow release polymer-iodine tablets for disinfection of untreated surface water. Journal of Applied Polymer Science 117 (1), 329e334. McDonogh, R., Schaule, G., Flemming, H.C., 1994. The permeability of biofouling layers on membranes. Journal of Membrane Science 87 (1e2), 199e217. Mulder, M., 1991. Basic Principles of Membrane Technology. Kluwer Academic Publishers, Dordrecht. Peng, M., Li, H.B., Wu, L.J., Zheng, Q., Chen, Y., Gu, W.F., 2005. Porous poly(vinylidene fluoride) membrane with highly hydrophobic surface. Journal of Applied Polymer Science 98 (3), 1358e1363. Richards, R.M.E., 1981. Antimicrobial action of silver-nitrate. Microbios 31 (124), 83e91. Rogers, J.V., Parkinson, C.V., Choi, Y.W., Speshock, J.L., Hussain, S.M., 2008. A preliminary assessment of silver nanoparticle inhibition of monkeypox virus plaque formation. Nanoscale Research Letters 3 (4), 129e133.

Rose, J.B., Gerba, C.P., 1991. Assessing potential health risks from viruses and parasites in reclaimed water in Arizona and Florida, USA. Water Science and Technology 23 (10e12), 2091e2098. Saveyn, H., Pauwels, G., Timmerman, R., Van der Meeren, P., 2005. Effect of polyelectrolyte conditioning on the enhanced dewatering of activated sludge by application of an electric field during the expression phase. Water Research 39 (13), 3012e3020. Sintubin, L., De Windt, W., Dick, J., Mast, J., van der Ha, D., Verstraete, W., Boon, N., 2009. Lactic acid bacteria as reducing and capping agent for the fast and efficient production of silver nanoparticles. Applied Microbiology and Biotechnology 84 (4), 741e749. Tilton, R.C., Rosenberg, B., 1978. Reversal of silver inhibition of microorganisms by agar. Applied and Environmental Microbiology 35 (6), 1116e1120. Ueda, T., Horan, N.J., 2000. Fate of indigenous bacteriophage in a membrane bioreactor. Water Research 34 (7), 2151e2159. USEPA, 2009. Drinking water contaminant candidate list 3-final. Federal Register 74 (194), 51850e51862. Vankelecom, I.F.J., 2002. Polymeric membranes in catalytic reactors. Chemical Reviews 102, 3779e3810. Verthe´, K., Possemiers, S., Boon, N., Vaneechoutte, M., Verstraete, W., 2004. Stability and activity of an Enterobacter aerogenes-specific bacteriophage under simulated gastrointestinal conditions. Applied Microbiology and Biotechnology 65 (4), 465e472. WHO, 2004. Guidelines for Drinking Water Quality. World Health Organization, Geneva, Switzerland. Wu, J.L., Li, H.T., Huang, X., 2010. Indigenous somatic coliphage removal from a real municipal wastewater by a submerged membrane bioreactor. Water Research 44 (6), 1853e1862. Yang, H.L., Lin, J.C.T., Huang, C., 2009. Application of nanosilver surface modification to RO membrane and spacer for mitigating biofouling in seawater desalination. Water Research 43 (15), 3777e3786. Zodrow, K., Brunet, L., Mahendra, S., Li, D., Zhang, A., Li, Q.L., Alvarez, P.J.J., 2009. Polysulfone ultrafiltration membranes impregnated with silver nanoparticles show improved biofouling resistance and virus removal. Water Research 43 (3), 715e723.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 6 5 e1 8 7 1

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Fouling behavior of microstructured hollow fiber membranes in submerged and aerated filtrations P.Z. C¸ulfaz a,c, M. Wessling b,c, R.G.H. Lammertink a,* a

Soft Matter, Fluidics and Interfaces, MESAþ Institute for Nanotechnology, University of Twente, P.O. Box 217, 7500 AE Enschede, The Netherlands b Chemical Process Engineering, AVT, RWTH Aachen University, Turmstr. 46, 52056 Aachen, Germany c Membrane Technology Group, University of Twente, P.O. Box 217, 7500 AE Enschede, The Netherlands

article info

abstract

Article history:

The performance of microstructured hollow fiber membranes in submerged and aerated

Received 17 October 2010

systems was investigated using colloidal silica as a model foulant. The microstructured

Received in revised form

fibers were compared to round fibers and to twisted microstructured fibers in flux-stepping

28 November 2010

experiments. The fouling resistances in the structured fibers were found to be higher than

Accepted 4 December 2010

those of round fibers. This was attributed to stagnant zones in the grooves of the structured

Available online 10 December 2010

fibers. As the bubble sizes were larger than the size of the grooves of the structured fibers, it is possible that neither the bubbles nor the secondary flow caused by the bubbles can reach

Keywords:

the bottom parts of the grooves. Twisting the structured fibers around their axes resulted in

Microstructured membrane

decreased fouling resistances. Large, cap-shaped bubbles and slugs were found to be the

MBR

most effective in fouling removal, while small bubbles of sizes similar to the convolutions

Fouling

in the structured fiber did not cause an improvement in these fibers. Modules in a vertical

Concentration polarization

orientation performed better than horizontal modules when coarse bubbling was used. For

Bubble flow

small bubbles, the difference between vertical and horizontal modules was not significant. When the structured and twisted fibers were compared to round fibers with respect to the permeate flowrate produced per fiber length instead of the actual flux through the convoluted membrane area, they showed lower fouling resistance than round fibers. This is because the enhancement in surface area is more than the increase in resistance caused by stagnant zones in the grooves of the structured fibers. From a practical point of view, although the microstructure does not promote further turbulence in submerged and aerated systems, it can still be possible to enhance productivity per module with the microstructured fibers due to their high surface area-to-volume ratio. ª 2010 Elsevier Ltd. All rights reserved.

1.

Introduction

Membrane bioreactor (MBR) technology, which combines the activated sludge process in wastewater treatment with membrane separation, offers an attractive alternative for the conventional wastewater treatment process (Shannon et al., 2008; Buer and Cumin, 2010; Judd, 2008). The most important

advantages offered by MBRs are smaller footprint, high product quality and lower waste production (Ho and Zydney, 2006; Le-Clech et al., 2006; Melin et al., 2006; Meng et al., 2009). Although the MBR process has found widespread use in wastewater treatment in the recent years, performance decline due to membrane fouling still remains the biggest challenge facing further development and application of this

* Corresponding author. E-mail address: [email protected] (R.G.H. Lammertink). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.12.007

1866

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 6 5 e1 8 7 1

technology (Ho and Zydney, 2006; Le-Clech et al., 2006; Meng et al., 2009). In membrane bioreactors, air bubbles which are supplied to the reactor to provide dissolved oxygen for the microorganisms and to maintain the solids in suspension also serve the purpose of reducing fouling. This is due to the shear created on the surface of the membranes as a result of the liquid flow caused by bubbles, the scouring action of the bubbles themselves, the secondary flows they induce in the liquid and the fiber movement induced by the passing bubbles (Cui et al., 2003). Although aeration is a very effective way of reducing fouling, it also forms the main component of the operating costs (Judd, 2008; Le-Clech et al., 2006; Sofia et al., 2004). Periodic backwashing or intermittent operation, which serves to relax the cake layer on the membranes during the time permeation is stopped, are other ways of preventing fouling. However both decrease the amount of permeate produced. In addition to these, filtration needs to be carried out below the “critical flux” or at a “sustainable flux”, at which no or little fouling occurs (Bacchin et al., 2006; Le-Clech et al., 2006). This limits the productivity in the sense that the permeate flow is kept at a low value. However, due to reduced fouling this value can be sustained for a longer time. In submerged membrane bioreactors, where the membranes are immersed directly in the bioreactor, either flat sheet or hollow fiber membranes are used. Although the operation of flat sheet membranes is simpler due to better control over the hydrodynamics in the more well-defined geometry of the modules, hollow fiber membranes are less expensive to produce, backflushable and offer higher packing density (Cui et al., 2003; Judd, 2002). Increasing the membrane area per module volume can reduce the module production cost significantly and is highly desired (Buer and Cumin, 2010). Recently we reported the fabrication of hollow fiber ultrafiltration membranes with a microstructured outer surface (C¸ulfaz et al., 2009). The membrane surface area in a given volume could be increased by up to 90%, and due to the increased area of the membranes’ skin layer this was shown to increase the productivity of the fibers. In the present study, we report the performance of these microstructured hollow fibers in a submerged system with aeration, operated in different configurations, in comparison to round fibers with the same intrinsic properties.

2.

The twisted membranes were made by twisting the structured fibers around their own axes, approximately one full turn per 10 cm. In the vertical module orientation, a steel tube was placed at the side of the fiber bundle to support the bundle. The fibers were fixed at the top and bottom ends and could move freely otherwise. The experimental setup, with drawings of both the vertical and horizontal module orientation is shown in Fig. 1. 4 and 7 fibers were used in vertical and horizontal modules, respectively. The module and fiber lengths were fixed for the structured, round and twisted membrane modules to provide the same degree of looseness and packing density. In vertical modules, the module length was 20 cm while the fiber length was 21 cm and in horizontal modules, the module length was 10.5 cm while the fiber length was 11 cm, providing 5% looseness for both kinds of modules (Wicaksana et al., 2006). Coarse bubbles were created by a perforated metal plate with 1 mm holes. Fine bubbles were created by a fritted ceramic plate with a nominal pore size of 0.1 mm. The feed solution was 2 wt% Ludox-TMA colloidal silica (SigmaeAldrich). The stock solution has 34 wt% silica in deionized water, with pH of 6e7.

2.1.

Flux-stepping experiments

In the flux-stepping method used, the permeate fluxes were stepwise increased and then decreased back in the same steps while recording the transmembrane pressure difference (TMP). Each flux step was continued for 20 min. To be able to compare whether the convolutions in the structured fiber have an

Experimental

Microstructured and round hollow fiber membranes made by the dryewet phase inversion of the polymer dope 16.68% PES, 4.91% PVP K30, 4.91% PVP K90, 7.18% H2O, 66.32% NMP were used throughout this study. Details of the fabrication can be found elsewhere (C¸ulfaz et al., 2009). The outer perimeters of the structured fiber and round fiber were 7.5 mm and 4.7 mm, respectively, resulting in 60% higher surface area per volume for the structured fiber compared to the round fiber. The inner diameters of the structured and round fiber were 0.86 mm and 0.84 mm, respectively. The pure water permeability of the fibers are 235  11 L/h m2 bar and 233  12 L/h m2 bar for structured and round fibers, respectively. The mean pore diameter of both fibers was found to be 12 nm by permporometry.

Fig. 1 e Experimental setup and the SEM images of the structured and round fibers.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 6 5 e1 8 7 1

improving effect on the hydrodynamics around the membrane, the permeate fluxes were set using the actual (convoluted, and therefore enhanced) surface area of the structured fibers. In the rest of the text, permeate flux refers to this flux calculated as the permeate flowrate through the actual surface area of the fibers. On the other hand, normalized permeate flux was calculated using the area of the circle passing through the middle of the fins for the structured fibers (C¸ulfaz et al., 2009). The latter is used to compare the fibers from a practical point of view, i.e. to compare the permeate volume that would be produced from the same module volume of round and structured fibers. The fouling resistance was calculated using Darcy’s law, expressing the total resistance during filtration as a resistance-in-series: Rf ¼

TMP  Rm hJper

(1)

where Rf is the fouling resistance which includes concentration polarization and/or particle deposition and Rm is the intrinsic membrane resistance determined by the pure water permeability.

3.

Results and discussion

Fig. 2 shows the coarse and fine bubbles produced at different aeration rates. With the coarse bubbler, the bubbles formed had diverse size and shapes, from spherical to ellipsoid and cap-shaped. At 0.026 and 0.060 m3/m2 s, gas slugs were also observed. With the fine bubbler, only spherical or ellipsoidshaped bubbles formed. At 0.002 m3/m2 s, the bubble size was between 0.3 and 4 mm, whereas at 0.008 m3/m2 s and 0.026 m3/m2 s, the bubble sizes were between 1 and 10 mm.

3.1.

1867

size of the convolutions in the structured fibers. Since both the bubbles themselves and the wakes following the bubbles are much larger than the convolutions, the effect of the bubbles cannot reach the depths of the grooves, and leave stagnant areas more susceptible to concentration polarization and subsequent particle deposition. When the structured fibers were twisted, the polarization resistance is less than the structured fibers. In these twisted fibers, the convolutions are not parallel to the bubble trajectory as in the structured fibers. This can provide that more of the area within the grooves is within reach of the rising bubbles. Similarly, in another study where these fibers were used in cross-flow filtrations, less particle deposition was observed on twisted fibers compared to structured fibers (C ¸ ulfaz et al., submitted for publication). This was attributed to secondary flows which form due to the helical grooves and fins. Fig. 5(a) shows the polarization resistances calculated using Equation (1) for the data in Fig. 4. In addition to the three modules in Fig. 4, in a fourth module (twisted-tight), the fibers were fixed more tightly, thus disabling much of the fiber movement cause by the bubbles. It was seen that the polarization resistance in this tight module was significantly higher than those of the loose modules. This supports previous observations showing that the looseness of the fibers is important in preventing fouling (Wicaksana et al., 2006; Yeo et al., 2007). Considering that at a certain flux, the permeate production from the structured fibers is 60% more than the round fibers due to the enhanced surface area, the polarization resistances were also compared for the normalized fluxes, i.e. for equal amount of permeate production per fiber length or equivalently per module volume (Fig. 5(b)). From this figure it is seen that for the same permeate production, the structured and twisted fiber modules have lower polarization resistances than the round fibers.

Comparison of round, structured and twisted fibers 3.2.

Fig. 3 shows flux-stepping experiments with the structured and round fibers. A vertical module, coarse bubbles and an aeration rate of 0.008 m3/m2 s were used. At none of the fluxes, significant cake deposition was observed, which is indicated by a stable TMP at each flux, and equal TMP for a given flux both while stepping up and stepping down. However, the TMPs at each flux deviate from the pure water permeability values, indicating the presence of concentration polarization. This deviation is plotted for the round, structured and structured-twisted membranes in Fig. 4. Concentration polarization is highest in the structured fiber, less in the structured-twisted fiber and the lowest in the round fiber. Bubbles prevent polarization and fouling through a number of different mechanisms. The scouring action of the bubbles on the membrane surface and the liquid mixing caused by the secondary flow in the bubble wakes are mechanisms effective in the pathway of the bubbles (Cui et al., 2003). Another factor that reduces particle deposition on the membrane is the fiber movement caused by the bubbles (Wicaksana et al., 2006; Yeo et al., 2007). Considering that the looseness of the fibers was similar for all three modules, this factor is not expected to create much difference between the fibers. The higher concentration polarization on the structured fibers is related to the size of the bubbles in comparison to the

Effect of aeration rate

Fig. 6 compares the polarization resistance at three aeration rates with coarse bubbles at a normalized flux of 70 L/h m2. When the aeration rate is increased from 0.008 m3/m2 s to 0.026 m3/m2 s, the polarization resistance becomes less for the structured and round fibers, while it is not effected for twisted fibers. Increasing the aeration rate further to 0.060 m3/m2 s increases the resistance to higher values than for both of the other two aeration rates. It is often seen that there exists a critical aeration rate below which severe fouling occurs. Above this aeration rate there is little or no improvement in fouling performance (Ndinisa et al., 2006; Ueda et al., 1997). A high shear rate due to extensive aeration can also have detrimental effects, as it increases the shear-induced diffusion and inertial lift forces for the large particles and causes small particles, which can induce severe pore blocking and irreversible gel formation, to become the major foulants. The size of the silica particles used in our experiments is between 10 and 40 nm, which is small and narrow enough to exclude this effect due to increased shear. In our case, the negative effect of high aeration rate is probably due to decreased contact between the feed solution and the membrane because of the over-occupation of the reactor volume with the bubbles. Assuming that the bubbles reach a terminal velocity of 0.2 m/s

1868

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 6 5 e1 8 7 1

Fig. 2 e Bubbles produced by the coarse and fine bubblers at different aeration rates.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 6 5 e1 8 7 1

1869

A

B

Fig. 3 e Flux-stepping experiments with (a) structured and (b) round fibers with vertical modules, large bubbles and an aeration rate of 0.008 m3/m2 s (0.8 L/min air flowrate).

(Fan and Tsuchiya, 1990), the residence time of a bubble in the reactor is estimated as 2 s. Then, with this aeration rate, the bubbles should occupy about 30% of the total reactor volume, which can cause such over-occupation.

Fig. 5 e Polarization resistance as a function of permeate flux for four different types of membrane modules: Round, structured, twisted and twisted-tight. Vertical modules, large bubbles. Aeration rate: 0.008 m3/m2 s (0.8 L/min air flowrate). (a) Permeate flux is the flux through the actual (convoluted) membrane area for structured fibers. (b) Permeate flux is normalized by using the perimeter of a circle passing through the middle of the fins of the structured fiber, corresponding to a similar effective membrane volume as the round fibers.

3.3.

Fig. 4 e Pure water permeability of the membranes and the effect of concentration polarization causing deviation from pure water permeability. Vertical modules, large bubbles. Aeration rate: 0.008 m3/m2 s.

Effect of module orientation and bubble size

Fig. 7 compares vertical and horizontal modules with large and small bubbles for an aeration rate of 0.026 m3/m2 s. When large bubbles were used, vertical modules performed significantly better than horizontal modules, as also reported by other researchers (Chang et al., 2002). In vertical modules, larger bubbles were more effective in depolarizing particle buildup than small bubbles. This higher depolarizing efficiency of large bubbles is attributed to the turbulent wake behind these cap-shaped bubbles and slugs (Cui et al., 2003). With small bubbles the behavior was similar in vertical and horizontal modules. In a vertical module, the bubbles can sweep the complete surface of the fibers, whereas when the bubbles move perpendicular to the fiber bundle, dead zones occur at the

1870

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 6 5 e1 8 7 1

Fig. 6 e Polarization resistance at a normalized permeate flux of 70 L/h m2 for round, structured and twisted membranes at different aeration intensities. Vertical modules, large bubbles.

back of the fiber bundle (Chang et al., 2002; Murai et al., 2005). Such dead zones are more likely to occur when large bubbles are used, since the size of these bubbles (5e20 mm) is larger than the spacings between individual fibers in the bundle. In this case, the bubbles would preferentially sweep through the periphery of the module and not penetrate through the bundle, leaving the fibers in the inner parts of the bundle as well as the back of the fiber bundle more susceptible to polarization and fouling. For small bubbles (1e10 mm), at least part of the bubbles can penetrate between the fiber bundle, causing mixing and preventing polarization in an equivalent manner to vertical modules.

As shown in Fig. 2, the bubble size was dependent on both the sparger used and the air flowrate. Using a fine bubbler and a low aeration rate of 0.002 m3/m2 s, bubbles of sizes close to the size of the corrugations of the structured fibers could be formed. It was observed that the polarization resistances under these conditions were much higher than those for higher aeration rates used with the same bubbler (Fig. 8).

Furthermore, at the highest fluxes indicated for each fiber, cake deposition started to occur, observed as an increasing TMP during the constant flux filtration. Under these conditions, although the size of the bubbles was on the order of the size of the convolutions in the structured fiber, the bubbles did not cause any enhancement in preventing polarization in these fibers. If the bubbles would have been able to reach within the grooves in the structured fibers, we would see at least a comparable resistance in the structured and round fibers. However, the resistance in the structured fibers is much higher than that of the round fiber. Yeo et al. reported that small bubbles (0.11 cm3 which corresponds to about 6 mm diameter in their experiments) seldom moved close to the fibers (Yeo et al., 2007). In addition to this, it has been shown that in a submerged membrane operation, the secondary flows caused by the bubbles are at least as effective as the scouring action of the bubbles themselves in fouling prevention (Cui et al., 2003; Le-Clech et al., 2006). Around bubbles smaller than 1 mm, there are no wake structures, and therefore no secondary flows and mixing. Instead there are laminar streamlines, which can only be as effective as liquid cross flow over the membrane.

Fig. 7 e Polarization resistance as a function of permeate flux for vertical and horizontal membrane modules, large and small bubbles for the structured membrane. Aeration rate: 0.026 m3/m2 s.

Fig. 8 e Polarization resistance as a function of permeate flux for small bubbles of 0.3e4 mm produced at an aeration rate of 0.002 m3/m2 s. Vertical modules.

3.4.

Effect of sub-mm sized bubbles

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 6 5 e1 8 7 1

4.

Conclusions

The performance of microstructured hollow fiber membranes in a submerged and aerated system was investigated using colloidal silica as a model foulant. The fouling in structured fibers was found to be less than round fibers for the same permeate production per module volume. However, when the fibers were compared at the actual permeate flux through the convoluted area of the structured fibers, the fouling resistance in the structured fibers were higher. When large bubbles were used, since bubble size was much larger than the convolutions, the bubbles and secondary flows they create were not able to reach the whole area of the structured fibers. This created stagnant areas in the grooves, which results in a higher overall resistance. On the other hand, bubbles of sizes similar to the groove dimensions also did not improve the fouling performance in the structured fibers. In general, large, cap-shaped bubbles and slugs were found to be the most effective in fouling removal, as they produce more turbulent wakes behind them and induce more fiber movement. Twisting the structured fibers around their axes decreased the fouling resistance, since by twisting, more of the area within the grooves became within reach of the bubbles passing by. Modules in a vertical orientation performed better than horizontal modules when coarse bubbles were used. On the other hand, for small bubbles, vertical and horizontal modules showed similar fouling behavior. In general, the structured fibers, in their original straight form or as twisted around their own axes, did not cause further enhancement in liquid mixing in the aerated systems. Furthermore, in most of the cases, due to stagnant zones remaining within the grooves, the fouling resistance was higher for these fibers. However, under the low-fouling conditions used in the present study, the increase in fouling resistance for the structured and twisted fibers was lower compared to the surface area enhancement due to the microstructured surface. This implies that in submerged, aerated systems structured fibers can still offer enhanced productivity per module volume for the same permeate production compared to round fibers.

references

Bacchin, P., Aimar, P., Field, R., 2006. Critical and sustainable fluxes: theory, experiments and applications. Journal of Membrane Science 281, 42e69.

1871

Buer, T., Cumin, J., 2010. MBR module design and operation. Desalination 250, 1073e1077. C¸ulfaz, P., Haddad, M., Wessling, M., Lammertink, R. Fouling behavior of microstructured hollow fibers in cross-flow filtrations: critical flux determination and direct visual observation of particle deposition. Journal of Membrane Science, submitted for publication. C¸ulfaz, P., Rolevink, E., van Rijn, C., Lammertink, R., Wessling, M., 2009. Microstructured hollow fibers for ultrafiltration. Journal of Membrane Science 347, 32e41. Chang, S., Fane, A., Vigneswaran, S., 2002. Experimental assessment of filtration of biomass with transverse and axial fibres. Chemical Engineering Journal 87, 121e127. Cui, Z., Chang, S., Fane, A., 2003. The use of gas bubbling to enhance membrane processes. Journal of Membrane Science 221, 1e35. Fan, L., Tsuchiya, K., 1990. Bubble Wake Dynamics in Liquids and LiquideSolid Suspensions. Butterworth. Ho, C.C., Zydney, A., 2006. Overview of fouling phenomena and modeling approaches for membrane bioreactors. Separation Science and Technology 41, 1231e1251. Judd, S., 2002. Submerged membrane bioreactors: flat plate or hollow fibre? Filtration and Separation 39, 30e31. Judd, S., 2008. The status of membrane bioreactor technology. Trends in Biotechnology 26, 109e116. Le-Clech, P., Chen, V., Fane, T., 2006. Fouling in membrane bioreactors used in wastewater treatment. Journal of Membrane Science 284, 17e53. Melin, T., Jefferson, B., Bixio, D., Thoeye, C., De Wilde, W., De Koning, J., van der Graaf, J., Wintgens, T., 2006. Membrane bioreactor technology for wastewater treatment and reuse. Desalination 187, 271e282. Meng, F., Chae, S.R., Drews, A., Kraume, M., Shin, H.S., Yang, F., 2009. Recent advances in membrane bioreactors (MBRs): membrane fouling and membrane material. Water Research 43, 1489e1512. Murai, Y., Sasaki, T., Ishikawa, M.A., Yamamoto, F., 2005. Bubbledriven convection around cylinders confined in a channel. Journal of Fluids Engineering, Transactions of the ASME 127, 117e123. Ndinisa, N., Fane, A., Wiley, D., 2006. Fouling control in a submerged flat sheet membrane system: Part i - bubbling and hydrodynamic effects. Separation Science and Technology 41, 1383e1409. Shannon, M., Bohn, P., Elimelech, M., Georgiadis, J., Maras, B., Mayes, A., 2008. Science and technology for water purification in the coming decades. Nature 452, 301e310. Sofia, A., Ng, W., Ong, S., 2004. Engineering design approaches for minimum fouling in submerged mbr. Desalination 160, 67e74. Ueda, T., Hata, K., Kikuoka, Y., Seino, O., 1997. Effects of aeration on suction pressure in a submerged membrane bioreactor. Water Research 31, 489e494. Wicaksana, F., Fane, A., Chen, V., 2006. Fibre movement induced by bubbling using submerged hollow fibre membranes. Journal of Membrane Science 271, 186e195. Yeo, A., Law, A., Fane, A., 2007. The relationship between performance of submerged hollow fibers and bubble-induced phenomena examined by particle image velocimetry. Journal of Membrane Science 304, 125e137.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 7 2 e1 8 7 8

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Oxidative removal of bisphenol A using zero valent aluminumeacid system Wanpeng Liu a, Honghua Zhang a, Beipei Cao a, Kunde Lin a,*, Jay Gan b a b

College of Biological and Environmental Engineering, Zhejiang University of Technology, Hangzhou 310032, China Department of Environmental Sciences, University of California, Riverside, CA 92521, USA

article info

abstract

Article history:

Bisphenol A (BPA), a controversial endocrine disruptor, is ubiquitous in the aquatic envi-

Received 7 October 2010

ronment. In this study, the oxidative degradation of BPA and its mechanism using zero

Received in revised form

valent aluminum (ZVAl)eacid system under air-equilibrated conditions was investigated.

29 November 2010

Under pH <3.5 acidic conditions, ZVAl demonstrated an excellent capacity to remove

Accepted 4 December 2010

BPA. More than 75% of BPA was eliminated within 12 h in pH 1.5 reaction solutions initially

Available online 10 December 2010

containing 4.0 g/L aluminum and 2.0 mg/L BPA at 25
1  C. The removal of BPA was further accelerated with increasing aluminum loadings. Higher temperature and lower initial pH

Keywords:

also facilitated BPA removal. The addition of Fe2þ into the ZVAleacid system significantly

Bisphenol A

accelerated the reaction likely due to the enhancing transformation of H2O2 to HO via

Zero valent aluminum

Fenton reaction. Furthermore, the primary products or intermediates including mono-

Oxidative degradation

hydroxylated BPA, hydroquinone, 2-(4-hydroxyphenyl)propane and 4-isopropenylphenol,

Fenton reaction

were identified and a possible reaction scheme was proposed. The remarkable capacity of the ZVAleacid system in removing BPA displays its potential application in the treatment of organic compoundecontaminated water. ª 2010 Elsevier Ltd. All rights reserved.

1.

Introduction

As an important industrial chemical, bisphenol A [BPA; 2,2-bis (4-hydroxyphenyl)propane] is manufactured in large quantities for the production of polycarbonate plastic and epoxy resins (Tsai, 2006). In the United States, approximately 1.7 billion pounds of BPA are synthesized and used every year (Schwartz, 2005). The principal route of BPA’s entrance into the aquatic environment is effluent from wastewater treatment plants (WWPTs) and landfill leachates (Kang et al., 2007). Bisphenol A is frequently found in effluents of domestic and industrial WWPTs because it is not completely eliminated during conventional biotic and abiotic treatments (Lee and Peart, 2000; Quinn et al., 2003; Rigol et al., 2002). The elimination rates of BPA in treatment plants ranged from 37 to 94%

(Fuerhacker, 2003; Lee and Peart, 2000). High concentrations of BPA were frequently found in leachates at landfill sites (Asakura et al., 2004; do Nascimento et al., 2003; Yamamoto et al., 2001). For example, Yamamoto et al. (2001) reported that levels of BPA in leachate at a hazardous waste landfill site ranged from 1.3 to 17,200 mg/L (average 269 mg/L). The ubiquity of BPA in the aquatic environment has greatly inspired the exploration of BPA removal methods, including physical (Pan et al., 2008), biological (Hirooka et al., 2005; Kadowaki et al., 2007; Kang et al., 2004), and chemical (Li et al., 2008; Lin et al., 2009; Ohko et al., 2001; Wang et al., 2010) techniques. For example, Lin et al. (2009) found that aqueous BPA could be efficiently removed by manganese dioxide. However, the authors demonstrated that the loss of BPA led to the formation of other organic intermediates

* Corresponding author. Tel.: þ86 571 88320778. E-mail address: [email protected] (K. Lin). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.12.004

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 7 2 e1 8 7 8

instead of mineralization to CO2. Advanced oxidation processes (AOPs) based on the formation and use of hydroxyl radical HO were recently considered for more thorough removal of aqueous BPA. For example, 175 mM of BPA in water was completely mineralized to CO2 by TiO2-photocatalyzed reactions under UV irradiation of 10 mW/cm2 for 20 h (Ohko et al., 2001). More recently, Wang et al. (2010) found that nearly 100% of BPA was removed by mesoporous Bi2WO6 under simulated solar light irradiation. Such a high efficiency of AOPs in treating BPA-containing water undoubtedly offers a promising alternative for the removal of aqueous organic contaminants. Different AOPs use different mechanisms to generate HO radicals. Like the widely known AOP zero valent iron (Joo et al., 2004, 2005; Kang and Choi, 2009; Lee et al., 2007; Noradoun and Cheng, 2005), zero valent aluminum (ZVAl)eacid system possesses high efficiency in producing reactive oxygen species because of its great thermodynamic driving force for electron transfer. The reduction potential of aluminum is 1.67 V (Al3þ/Al), much lower than that of iron (0.44 V, Fe2þ/Fe). The electron transfer capacity of ZVAl has already been exploited for the degradation of other organic compounds. For example, Lien and Wilkin (2002) showed that ZVAl treated by surface modification with sulfate groups could easily eliminate methyl tert-butyl ether. In a recent study, Bokare and Choi (2009) demonstrated that commercially available ZVAl had the desirable capacity in oxidative degradation of 4-chlorophenol, sodium dichloroacetate, phenol and nitrobenzene. The objective of the present work was to explore the degradation of BPA with ZVAleacid system by investigating removal efficacy and influencing factors such as ZVAl loading, reaction temperature, pH, and Fe2þ. The principal reaction intermediates and products were also identified, and a reaction scheme was proposed.

2.

Materials and methods

2.1.

Chemicals

Bisphenol A standard (purity >99%), N,O-bis(trimethylsilyl) trifluoroacetamide with trimethylchlorosilane (BSTFA þ TMCS, 99:1) and 2,9-dimethyl-1,10-phenanthroline (DMP) were purchased from SigmaeAldrich (St. Louis, MO, USA). Aluminum powder (purity >99%, particle size 75e150 mm, surface covered with native aluminum oxide layer) was purchased from Sinopharm Chemical Reagent (Shanghai, China). Other chemicals and solvents used in this study were of analytical grade or high performance liquid chromatography (HPLC) grade. The concentration of the purchased hydrogen peroxide solution (30 wt%) was calibrated by titration with potassium permanganate (Bader et al., 1988). All chemicals were used as received. Ultrapure water (18.2 MU cm resistivity) was prepared using a Millipore purification system. Stock solution of 5.0 g/L BPA was prepared in acetone. The 1% (w/v) DMP solution was prepared in ethanol in a brown bottle. A copper (II) sulfate (CuSO4) solution (0.01 M) was prepared by dissolving CuSO4$5H2O in ultrapure water. Phosphate buffer solutions (0.1 M) were prepared by mixing appropriate volumes of 0.5 M Na2HPO4 with 0.5 M NaH2PO4, with the final

1873

pH adjusted to pH 7.0 by NaOH (1 M). All solutions were stored at 4  C prior to use.

2.2.

Reaction setup

All reactions were carried out in 250-mL glass flasks with a total suspension volume of 100 mL under air-equilibrated conditions and in the absence of light. Unless otherwise noted, temperature was fixed at 25
1  C. The initial pH (pHi) of the solutions was adjusted to the designated value with 1 M HClO4 standard solution. A 40-mL aliquot of 5.0 g/L BPA stock solution was added to make a nominal initial concentration of 2.0 mg/L. Reaction mixtures were constantly shaken in a thermostatic mechanical shaker at 130 rpm. Reactions were initiated by adding a predetermined amount of ZVAl into the pre-equilibrated and constantly mixed solutions. Aliquots of 1.0 mL sample were periodically withdrawn and transferred to 2-mL centrifuge tubes containing 10 mL of methanol (as HO scavenger). The samples were immediately vortexed for 10 s and centrifuged at 8000 rpm for 5 min. The supernatant was transferred to 2-mL vials and subjected to HPLC analysis to determine the remaining BPA concentration. All samples were stored at 4  C and analyzed within 24 h. Except for the ZVAlfree control, all treatments were carried out in triplicates. Specific treatments were included to evaluate the effect of Fe2þ, the presence of aluminum oxide and humic acid (HA). To determine the effect of Fe2þ and HA, a predetermined amount of FeCl2 and HA were separately added into the reaction solutions. To evaluate the effect of the native aluminum oxide layer, the reaction rates were individually measured before and after the pretreatment of aluminum by mixing in a pH 1.5 acidic solution for 2 h. To quantify the formation of H2O2 in the ZVAleacid system, an additional reaction solution without BPA was also prepared. Aliquots of 1.0 mL samples were periodically transferred to 2mL centrifuge tubes and immediately centrifuged at 8000 rpm for 5 min. One half mL subsample of the supernatant was collected and used for H2O2 measurement using a method according to Kasaka et al. (1998). The determination of H2O2 is based on a spectrophotometric method via the stoichiometric reaction of H2O2 with copper (II) ion and DMP. Briefly, 0.5 mL each of DMP, CuSO4, phosphate buffer, and the sample supernatant was added to a 5-mL volumetric flask and the flask was filled up to 5 mL with water. After mixing, the solution was transferred to 1-cm cells and the absorbance was measured at 454 nm on a V-550 UV/Vis spectrophotometer (Jasco Technologies, Tokyo, Japan).

2.3.

Chemical analysis

Bisphenol A concentrations in reaction samples were determined using a reverse-phase HPLC coupled with a variablewavelength ultraviolet (UV) detector (Jasco). The detection wavelength was 196 nm. A Grace Alltima C18 column (250  4.6 mm, 5 mm, Grace, Deerfield, IL, US) was employed for the separation. The isocratic mobile phase consisted of 55% acetonitrile and 45% 40 mM acetic acid solution with a flow rate of 1.0 mL/min. The injection volume was 25 mL. Under these conditions, the typical retention time for BPA was 6.2 min.

1874

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 7 2 e1 8 7 8

3.

Results and discussion

3.1.

Efficacy of BPA removal by ZVAleacid system

100

% BPA Residue

For reaction product identification, 2.0 mg/L BPA and 4.0 g/L aluminum were reacted at pHi 1.5 (4 replicates), and the reaction was quenched at 0, 4, 8, and 24 h, respectively, by adding 50 mL methanol. The solutions were extracted three consecutive times with 20 mL methylene chloride. The combined organic phases were dehydrated by passing through a filter paper filled with anhydrous sodium sulfate, and concentrated to near dryness on a vacuum rotary evaporator. The residue was evaporated to dryness under nitrogen and redissolved in 1.0 mL acetoneehexane (1:1). The extract was then divided into two equal parts, from which one part was analyzed directly and the other part was derivatized before analysis. For derivatization, the extract was evaporated to dryness under nitrogen, and 150 mL of BSTFA þ TMCS was added to silylate the polar products. Immediately after the addition of the silylation reagent, the vial was crimped with a cap with Teflon-lined septum and kept at 60  C for 2 h. The silylated products were redissolved in 0.5 mL hexane. Aliquots (2 mL) of the silylated and non-silylated extracts were injected into an Agilent 7890A gas chromatograph (GC) coupled with an Agilent 5975C mass spectrometer (MS) (Agilent Technologies, Wilmington, DE) for chemical structural elucidation. An HP-5MS column (30 m  0.25 mm  0.25 mm; Agilent) was used for separation. The inlet temperature was 250  C, and the detector temperature was 325  C. The oven temperature was initiated at 60  C and held for 0 min, then increased to 300  C at 10  C/min and held for 20 min. The flow rate of the carrier gas (helium) was 1.0 mL/min. The injection of 2-mL sample was carried out by an Agilent 7693 autosampler in the pulsed splitless mode that was turned on after 1.0 min. The temperatures of transfer line, ion source, and MS detector were 280, 230, and 150  C, respectively. The MS detector was operated in the electron impact mode at 70 eV and the mass spectra were acquired in the full scan mode with m/z ranging from 50 to 600. The formation of formic and acetic acids was verified using a Dionex ICS-2000 iron chromatography with an IonPac AS 19 anion column (4.0  250 mm) and a 10 mM KOH eluent at a flow rate of 1.0 mL/min.

80 60

0.0 g/L 1.0 g/L 2.0 g/L 4.0 g/L

40 20 0 0

2

4

6

8

10

12

Reaction Time (h) Fig. 1 e Effect of initial aluminum loading on bisphenol A (BPA) removal at 25 ± 1  C in pH 1.5 reaction solutions initially containing 2.0 mg/L BPA. Data for aluminum-free treatment are from single measurement, while the other data points are given as means ± standard deviations (n [ 3).

of the total surface area of aluminum. As reported by Bokare and Choi (2009), hydroxyl radicals, which are the actual oxidant for the decomposition of organic compounds, were generated through direct electron transfer from the ZVAl surface to H2O2 in the ZVAleacid system under air-equilibrated conditions. However, BPA removal was not further improved when the ZVAl loading was >4.0 g/L, likely due to the fact that both dissolution of native oxide layer and corrosion of metal aluminum led to Hþ depletion. For example, after 12 h of reaction, the final pH of the reaction solutions fortified with 1.0, 2.0 and 4.0 g/L ZVAl increased to about 3.52
0.01, 3.53
0.01 and 3.56
0.01, respectively. As a result, the rate of metal corrosion decreased and the reaction became increasingly inhibited.

100

In the absence of aluminum, BPA remained fairly constant in the solution and no degradation was observed during the reaction period. However, more than 75% of BPA disappeared in 12 h when ZVAl was added into the acidic reaction solution (Fig. 1). It should be noted that the degradation proceeded after an induction period of about 1e2 h under the experimental conditions.

3.2.

Effect of ZVAl loading

The degradation rate increased with the initial aluminum loading (Fig. 1). For example, the removal of BPA at 12 h increased from 57% for a solution with 1.0 g/L ZVAl to 75% with 2.0 g/L ZVAl. The enhanced BPA removal rates with higher aluminum loadings were obviously due to the increase

% BPA Residue

80 60 40

pH 1.0 pH 1.5 pH 2.5 pH 3.5 pH 3.5 with pretreated Al pH 1.5 with pretreated Al

20 0 -20 0

2

4

6

8

10

12

Reaction Time (h) Fig. 2 e Effect of pH on bisphenol A (BPA) removal in reaction solutions initially containing 2.0 mg/L BPA and 4.0 g/L aluminum at 25 ± 1  C. The pretreated Al was commercial aluminum powder soaked in pH 1.5 solutions for 2 h prior to use. Data points are given as means ± standard deviations (n [ 3).

1875

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 7 2 e1 8 7 8

100

175

[H2O2] (uM)

150

% BPA Residue

pH 1.5 pH 3.5

125 100 75 50

0 mg/L 1 mg/L 5 mg/L 10 mg/L 50 mg/L HA only

80 60 40 20

25 0

0 0

2

4

6

8

10

0

12

2

4

Reaction Time (h) Fig. 3 e Formation of H2O2 in a bisphenol A-free zero valent aluminumeacid system initially containing 4.0 g/L aluminum at 25 ± 1  C. Data are given as means (n [ 2).

3.3.

Effect of pH

The removal of BPA in the ZVAleacid system was clearly pH-dependent. Overall, BPA removal rates decreased with increasing pHi (Fig. 2), indicating that acidic conditions facilitated BPA removal. For example, BPA removal rate at 12 h decreased from 75% for the reaction solution with pHi 1.5e25% for that with pHi 2.5. When pHi was increased to 3.5, almost no BPA loss was observed after 12 h. The oxidative capacity of ZVAl toward other organic compounds, e.g., 4-chlorophenol, was also found to be pH-dependent (Bokare and Choi, 2009). Similarly, pH was a major factor that affected the oxidative potential of ZVI/Hþ/O2 system (Bergendahl and Thies, 2004; Chen et al., 2001; Joo et al., 2004; Lee et al., 2007; Song and Carraway, 2005). The pH of the reaction solution increased as the reaction proceeded, which in turn gradually hindered the reaction. In this study, both the dissolution of alumina and the corrosion of aluminum metal depleted Hþ and thus led to an increase in suspension pH over time. For example, after 14 h of reaction, the degradation of BPA in the pHi 1.5 solution completely

% BPA Residue

100

8

10

12

14

16

Fig. 5 e Effect of humic acid (HA) on bisphenol A (BPA) removal at 25 ± 1  C in pH 1.0 solutions initially containing 2.0 mg/L BPA and 4.0 g/L aluminum. Data for the HA only treatment are from single measurement, while the other data points are given as means ± standard deviations (n [ 3).

stopped because the reaction solution pH increased to 3.98
0.03. The reaction of BPA in the ZVAleacid system consistently displayed an induction period (Fig. 2). All reactions were relatively slower in the first 2 h, a phenomenon that was also observed in Bokare and Choi (2009), which was attributed to the dissolution of native oxide layer on the aluminum powder. However, an induction period was still visible even after the pretreatment to remove the original oxide layer (Fig. 2). Moreover, H2O2 was steadily generated during the first 2 h in the BPA-free aluminumeacid systems (Fig. 3). These results together suggest that the dissolution of the native alumina layer was not the major cause for the induction period. In the ZVAleacid system, it involved two major processes (Bokare and Choi, 2009): (i) ZVAl-induced H2O2 production (reactions (1)), and (ii) the subsequent generation of HO (reaction (2)):

2Al0 þ 3O2 þ 6Hþ / 2Al3þ þ 3H2O2

(1)

CH3

2+

Fe free 1 uM 10 uM

80

6

Reaction Time (h)

HO

C

OH

CH3

60

OH

40

CH3 HO

20

C

OH

(Monohydroxylated BPA)

CH3 OH

0 0

2

4

6

8

10

12

Reaction Time (h) Fig. 4 e Effect of additional FeCl2 on bisphenol A (BPA) removal at 25 ± 1  C in pH 1.5 reaction solutions initially containing 2.0 mg/L BPA and 4.0 g/L aluminum. Data points are given as means ± standard deviations (n [ 3).

CH2

CH3 HO

C

H

HO

OH

HO

C

CH3

CH3

Fig. 6 e The proposed reaction scheme for the oxidation of bisphenol A (BPA) by zero valent aluminumeacid system.

1876

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 7 2 e1 8 7 8

Al0 þ 3H2O2 / Al3þ þ 3HO þ 3OH

(2)

Although H2O2 was generated up to 51 mM in the ZVAleacid system with pHi 3.5 (Fig. 3), no BPA was removed over the 12 h reaction (Fig. 2), suggesting that the formed H2O2 was not immediately transformed to HO via reaction (2). Therefore, the accumulation of H2O2 may be responsible for the occurrence of the induction period.

3.4.

Effect of Fe2þ

It is well known that Fe2þ easily reacts with H2O2 and generates HO via Fenton reaction. To improve the efficiency of the ZVAleacid system, various amounts of Fe2þ were added into the reaction solutions to facilitate the generation of HO. As expected, the removal of BPA was significantly enhanced by the addition of Fe2þ (Fig. 4). For example, >99% of BPA

Fig. 7 e Mass spectra and possible ion fragment assignments of the identified products or intermediates.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 7 2 e1 8 7 8

was removal in 12 h for the reaction solution with 1.0 mM Fe2þ, and >99% was removed in 8 h for the reaction solution with 10 mM Fe2þ. In contrast, only 75% BPA was removed in 12 h without the addition of Fe2þ. These results clearly suggest that electron transfer from Fe2þ to H2O2 was more efficient than that from aluminum surface to H2O2. Thus, Fe2þ accelerated the formation of HO and the subsequent degradation of BPA.

3.5.

Effect of humic acid

Humic acid, a dissolved organic matter commonly found in natural and waste water, is another factor that may influence the oxidative efficiency of ZVAleacid system. In this study, HA demonstrated an inhibitory effect on BPA removal (Fig. 5). For example, >98% of BPA was removed in 10 h for the reaction solutions without or with 1.0 mg/L of HA, while the removal decreased to 81%, 59% and 51% for reaction solutions with 5.0, 10.0, and 50.0 mg/L of HA, respectively. As Fukushima et al. (2001) showed that HA can react with hydroxyl radical at several molecular sites, the inhibitory effect is likely due to the competitive reaction of HA and BPA with hydroxyl radical.

3.6.

Reaction products and suggested scheme

Through GCeMS analysis, four primary products (or intermediates) including monohydroxylated BPA, hydroquinone, 2-(4hydroxyphenyl)propane and 4-isopropenylphenol were identified as their silylated derivatives (Fig. 6). The mass spectra and interpretation are given in Fig. 7. On the basis of the identified products, we proposed a possible reaction scheme (Fig. 6). Oxidation of BPA is initiated by the attack of hydroxyl radicals, forming a monohydroxylated BPA (m- or/ and o-). It has been well-documented that compounds with aromatic rings react with hydroxyl radicals forming an intermediate cyclohexadienyl radical, which is then rapidly oxidized to the hydroxylated derivatives in the presence of oxidizing agents, such as O2, Fe3þ and Cu2þ (Gzmen et al., 2003). The further reaction of monohydroxylated BPA with HO leads to bond cleavages and yields intermediates of 4-isopropenylphenol, hydroquinone, and 2-(4-hydroxyphenyl)propane. The four intermediates exhibited a very similar evolution profile during 24 h of reaction. An accurate quantitation for the formation of the four intermediates was impossible because of a lack of authentic standards. However, based on the total ion chromatogram in GCeMS analysis (data not shown), the peak areas of the four intermediates steadily increased during the first 8 h of reaction, suggesting an increase in their concentrations. In contrast, the peak areas of these intermediates at 24 h were all much smaller than those at 4 h or 8 h, indicating a further removal of these intermediates. In addition, formic and acetic acids were also detected in 24-h reaction solutions via an ion chromatographic analysis (data not shown). Therefore, it is expected that these intermediates will be further oxidized to shortchain carboxylic acids (e.g., formic acid, acetic acid, oxalic acid and butendionic acid) and then finally CO2 and H2O.

4.

Conclusions

This study demonstrated that ZVAleacid system possesses a relatively high oxidative capacity in removing aqueous

1877

BPA, offering a good alternative for the treatment of aqueous organic contaminants. However, the Al3þ residue from the reaction may be an important issue to consider if this reaction is employed in wastewater treatment because of the toxicity of Al3þ toward plants, aquatic organisms and human beings (Smith, 1996). The residual Al3þ may be transformed to aluminum hydroxide through pH adjustment, a practice that is widely used in the flocculation in water treatment.

Acknowledgements This work was financially supported by the National Natural Science Foundation of China (21077091, 20837002) and the National Basic Research Program of China (2009CB421603).

references

Asakura, H., Matsuto, T., Tanaka, N., 2004. Behavior of endocrine disrupting chemicals in leachate from MSW landfill sites in Japan. Waste Management 24 (6), 613e622. Bader, H., Sturzenegger, V., Hoigne´, J., 1988. Photometric method for the determination of low concentrations of hydrogen peroxide by the peroxidase catalyzed of N, N-diethyl-pphenylenediamine (DPD). Water Research 22 (9), 1109e1115. Bergendahl, J.A., Thies, T.P., 2004. Fenton’s oxidation of MTBE with zero-valent iron. Water Research 38 (2), 327e334. Bokare, A.D., Choi, W., 2009. Zero-valent aluminum for oxidative degradation of aqueous organic pollutants. Environmental Science and Technology 43 (18), 7130e7135. Chen, J.L., Al-Abed, S.R., Ryan, J.A., Li, Z.B., 2001. Effects of pH on dechlorination of trichloroethylene by zero-valent iron. Journal of Hazardous Materials 83 (3), 243e254. do Nascimento, I., von Mu¨hlen, C., Schossler, P., Carama˜o, E.B., 2003. Identification of some plasticizers compounds landfill leachate. Chemosphere 50 (5), 657e663. Fuerhacker, M., 2003. Bisphenol A emission factors from industrial sources and elimination rates in a sewage treatment plant. Water Science and Technology 47 (10), 117e122. Fukushima, M., Tasumi, K., Nagao, S., 2001. Degradation characteristics of humic acid during photo-Fenton processes. Environmental Science and Technology 35 (18), 3683e3690. Gzmen, B., Oturan, M.A., Oturan, N., Erbatur, O., 2003. Indirect electrochemical treatment of bisphenol A in water via electrochemically generated Fenton’s reagent. Environmental Science and Technology 37 (16), 3716e3723. Hirooka, T., Nagase, H., Uchida, K., Hiroshige, Y., Ehara, Y., Nishikawa, J., Nishihara, T., Miyamoto, K., Hirata, Z., 2005. Biodegradation of bisphenol A and disappearance of its estrogenic activity by the green ala Chlorella fusca var. vacuolata. Environmental Toxicology and Chemistry 24 (8), 1896e1901. Joo, S.H., Feitz, A.J., Sedlak, D.L., Waite, T.D., 2005. Quantification of the oxidizing capacity of nanoparticulate zero-valent iron. Environmental Science and Technology 39 (5), 1263e1268. Joo, S.H., Feitz, A.J., Waite, T.D., 2004. Oxidative degradation of the carbothioate herbicide, molinate, using nanoscale zero-valent iron. Environmental Science and Technology 38 (7), 2242e2247. Kadowaki, T., Wheelock, C.E., Adachi, T., Kudo, T., Okamoto, S., Tanaka, N., Tonomura, K., Tsujimoto, G., Mamitsuka, H., Goto, S., Kanehisa, M., 2007. Identification of endocrine

1878

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 7 2 e1 8 7 8

disruptor biodegradation by integration of structureeactivity relationship with pathway analysis. Environmental Science and Technology 41 (23), 7997e8003. Kang, J.H., Aasi, D., Katayama, Y., 2007. Bisphenol A in the aquatic environment and its endocrine-disruptive effects on aquatic organisms. Critical Reviews in Toxicology 37 (7), 607e625. Kang, J.H., Ri, N., Kondo, F., 2004. Streptomyces sp. strain isolated from river water has high bisphenol A degradability. Letters in Applied Microbiology 39 (2), 178e180. Kang, S.H., Choi, W., 2009. Oxidative degradation of organic compounds using zero-valent iron in the presence of natural organic matter serving as an electron shuttle. Environmental Science and Technology 43 (3), 878e883. Kasaka, K., Yamada, H., Matsui, S., Echigo, S., Shishida, K., 1998. Comparison among the methods for hydrogen peroxide measurements to evaluate advanced oxidation processes: application of a spectrophotometric method using copper(II) ion and 2,9-dimethyl-1,10-phenanthroline. Environmental Science and Technology 32 (23), 3821e3824. Lee, H.B., Peart, T.E., 2000. Bisphenol A contamination in Canadian municipal and industrial wastewater and sludge samples. Water Quality Research Journal of Canada 35 (2), 283e298. Lee, J., Kim, J., Choi, W., 2007. Oxidation on zero valent iron promoted by polyoxometalate as an electron shuttle. Environmental Science and Technology 41 (9), 3335e3340. Li, C., Li, X.Z., Grahamc, N., Gao, N.Y., 2008. The aqueous degradation of bisphenol A and steroid estrogens by ferrate. Water Research 42 (12), 109e120. Lien, H.S.L., Wilkin, R., 2002. Reductive activation of dioxygen for degradation of methyl tert-butyl ether by bifunctional aluminum. Environmental Science and Technology 36 (20), 4436e4440. Lin, K., Liu, W., Gan, J., 2009. Oxidative removal of bisphenol A by manganese dioxide: efficacy, products, and pathways. Environmental Science and Technology 43 (10), 3860e3864. Noradoun, C.E., Cheng, I.F., 2005. EDTA degradation induced by dioxygen activation in a zero-valent iron/air/water system. Environmental Science and Technology 39 (18), 7158e7163.

Ohko, Y., Ando, I., Niwa, C., Tatsuma, T., Yamamura, T., Nakashima, T., Kubota, Y., Fujishima, A., 2001. Degradation of bisphenol A in water by TiO2 photocatalyst. Environmental Science and Technology 35 (11), 2365e2368. Pan, B., Lin, D.H., Mashayekhi, H., Xing, B.S., 2008. Adsorption and hysteresis of bisphenol A and 17R-ethinyl estradiol on carbon nanomaterials. Environmental Science and Technology 42 (15), 5480e5485. Quinn, B.P., Booth, M.M., Delfino, J.J., Holm, S.E., Gross, T.S., 2003. Selected resin acid in effluent and receiving waters derived from a bleached and unbleached kraft pulp and paper mill. Environmental Toxicology and Chemistry 22 (1), 214e218. Rigol, A., Latorre, A., Lacorte, S., Barcelo, D., 2002. Determination of toxic compounds in paper-recycling process waters by gas chromatography-mass spectrometry and liquid chromatography-mass spectrometry. Journal of Chromatography A 963 (12), 265e275. Schwartz, D.A., 2005. National Toxicology Program (NTP); Center for the evaluation of risks to human reproduction (CERHR); Plans for future expert panel evaluation of bisphenol A and hydroxyurea; requests for comments and nominations of scientists qualified to serve on these expert panels. Federal Register 70, 75827e75828. Smith, R.W., 1996. Kinetic aspects of aqueous aluminum chemistry: environmental implications. Coordination Chemistry Reviews 149 (1), 81e93. Song, H., Carraway, E.R., 2005. Reduction of chlorinated ethanes by nanosized zero-valent iron: kinetics, pathways, and effects of reaction conditions. Environmental Science and Technology 39 (16), 6237e6245. Tsai, W.T., 2006. Human health risk on environmental exposure to bisphenol-A: a review. Journal of Environmental Science and Health, Part C  Environmental Carcinogenesis and Ecotoxicology Reviews 24 (2), 225e255. Wang, C.Y., Zhang, H., Li, F., Zhu, L.Y., 2010. Degradation and mineralization of bisphenol A by mesoporous Bi2WO6 under simulated solar light irradiation. Environmental Science and Technology 44 (17), 6843e6848. Yamamoto, T., Yasuhara, A., Shiraishi, H., Nakasugi, O., 2001. Bisphenol A in hazardous waste landfill leachates. Chemosphere 42 (4), 415e418.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 7 9 e1 8 8 9

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Surfactant-coated aluminum hydroxide for the rapid removal and biodegradation of hydrophobic organic pollutants in water Tohru Saitoh*, Masato Yamaguchi, Masataka Hiraide Department of Molecular Design and Engineering, Graduate School of Engineering, Nagoya University, Chikusa-ku, Nagoya 464-8603, Japan

article info

abstract

Article history:

The removal of hydrophobic organic pollutants in water to surfactant-coated aluminum

Received 11 August 2010

hydroxide [surfactant-Al(OH)3] was investigated. Anionic surfactants such as sodium

Received in revised form

dodecyl sulfate (SDS), sodium bis(2-ethylhexyl)sulfosuccinate (AOT), and sodium oleate

2 December 2010

were sorbed on positively charged aluminum hydroxide at pH 7 and formed hydrophobic

Accepted 6 December 2010

aggregates that can incorporate hydrophobic organic pollutants in water. Because of the

Available online 13 December 2010

hydrophobic interaction and decrease in the positive charge, surfactant-Al(OH)3 was coagulated into precipitates that can readily be separated from water. Hydrophobic organic

Keywords:

pollutants such as alkylphenols, polycyclic aromatic hydrocarbons, estrogens, chlorinated

Hydrophobic organic pollutants

antifungals, and pesticides were well collected to the precipitates and thus efficiently

Wastewater treatment

removed from water. The collection of hydrophobic organic pollutants was correlated to

Surfactant-coated

their aqueous-octanol distribution coefficient. The decomposition of hydrophobic organic

aluminum hydroxide

pollutants was examined using a bacterial agent (Bacillus subtilis). Hydrophobic organic

Rapid removal

compounds collected to AOT-Al(OH)3 or sodium oleate-Al(OH)3 were insufficiently

Biodegradation

decomposed. On the other hand, nonylphenol, octylphenol, and pendimethalin collected to SDS-Al(OH)3 were decomposed within 1 week. The decomposition was accelerated by the collection to SDS-Al(OH)3. ª 2010 Elsevier Ltd. All rights reserved.

1.

Introduction

Hydrophobic organic pollutants such as alkylphenols, chlorophenols, pesticides, and polycyclic aromatic hydrocarbons are ubiquitous environmental pollutants resulting from waste from paper mills, sewage treatment processes, preparation of agricultural and industrial chemicals, and the combustion of fossil fuels. Because of their hydrophobic properties, these compounds tend to bioaccumulate in the lipid stores of animals and human beings. The potentials for several physiological actions including carcinogenic, mutagenic, estrogenlike properties have been reported (Laws et al., 2000; Ohe et al.,

2004). Additionally, their fungicidal property potentially gives significant damages to the ecosystems involving bacteria and phytoplankton (Pelletier et al., 2006; DeLorenzo and Fleming, 2008). Methods for the efficient removal of such hydrophobic organic pollutants are required. Among numerous water treatment techniques, the coagulationesedimentation method is one of the most popular methods because of its simplicity, low cost, and easiness of scaling-up (Koening, 1967; The´bault et al., 1981; Randtke and Stephen, 1988). The method includes the addition of metal salt such as polyaluminum chloride or aluminum sulfate and the subsequent neutralization of water for the formation of

* Corresponding author. Tel.: þ81 52 789 3579; fax: þ81 52 789 3241. E-mail address: [email protected] (T. Saitoh). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.12.009

1880

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 7 9 e1 8 8 9

aluminum hydroxide. Polymer flocculants such as sodium polyacrylate or sodium alginate are often used for facilitating the coagulation of finely dispersed aluminum hydroxide. The resulting flocks or precipitates can rapidly be separated by filtration. Particulate materials, colorized materials, and heavy metal ions are coprecipitated and efficiently removed from water. However, the removal of uncharged hydrophobic organic pollutants is often insufficient (The´bault et al., 1981; Boyd et al., 2003; Carballa et al., 2004; Suarez et al., 2009). We have designed a simple and efficient method for collecting hydrophobic organic compounds in water using surfactant-coated solid materials (Hiraide et al., 1997; Saitoh et al., 2002a; Saitoh et al., 2004; Saitoh et al., 2005; Saitoh et al., 2005; Saitoh et al., 2007a). The method is based on the extraction of hydrophobic compounds to the surfactant aggregates (namely hemimicelles or admicelles) on the solid surfaces (Valsaraj, 1989, 1992; Sandeep et al., 1994). Several types of admicelles can be readily prepared only by mixing solid materials and appropriate surfactants in the aqueous solution. Hydrophobic organic compounds and metal chelates were well collected and highly concentrated to the admicelles. Due to high water-permeability of the admicelles, the solute incorporation was quite rapid comparing with the collection to conventional hydrophobic solid sorbents. Therefore, the admicelle-mediated separation technique has been extensively applied to preconcentrate varieties of hydrophobic organic compounds for their instrumental analyses (Merino et al., 2003; Costi et al., 2008; Moral et al., 2008; Gangula et al., 2010). In the present study, we attempted the introduction of the admicelle-mediated separation methodology to the coagulation and sedimentation technique. Since aluminum hydroxide is positively charged at neutral pHs, anionic surfactants added are expected to be sorbed on the aluminum hydroxide to form surfactant-coated aluminum hydroxide [surfactant-Al(OH)3]. The anionic surfactant molecules can aggregate and provide hydrophobic media for the incorporation of hydrophobic organic pollutants. The potential usefulness of surfactant-Al (OH)3 for rapid and efficient removal of different hydrophobic organic compounds in water was investigated. Furthermore, biodegradation of hydrophobic organic pollutants was examined using bacteria that can be activated in the presence of anionic surfactant.

2.

Meterials and methods

2.1.

Apparatus

component (Millipore, Milford, MA, USA). A rotary shaker incubator (120 rpm, MMS-310 and FMC-1000, Tokyo Rikakikai, Tokyo, Japan) was used for the experiments of biodegradation. Dissolved oxygen was monitored with a Toa MM60R water quality meter (Tokyo, Japan). A Branson Model 450 sonifier (output control 1, Danbury, CT, USA) was employed to facilitate the elution of hydrophobic organic pollutants.

2.2.

An Al(III) solution (50 g L1 as Al(III) ions) was prepared by dissolving aluminum(III) chloride hexahydrate with water. Sodium dodecyl sulfate (SDS, for biochemical, Wako Pure Chemical, Tokyo, Japan), sodium bis(2-ethylhexyl)sulfosuccinate (AOT, Docusate sodium salt, SigmaeAldrich, St. Louis, MO, USA), and sodium oleate (Tokyo Chemical, Tokyo, Japan) were used as 50 g L1 aqueous solutions. Benz[a]anthracene, benzo[a]pyrene, fluoranthene, pyrene, estrone, b-estradiol, esprocarb, methoxychlor, pendimethalin, pretilachlor, 4-noctylphenol, 4-n-nonylphenol, 3,4,4-trichlorocarbanilide (triclocarban), and 2,4,40 -trichloro-20 -hydroxydiphenyl ether (triclosan) were obtained from Wako Pure Chemical. They were employed as 1 g L1 ethanol solution. A molecular probe, Nphenyl-1-naphthylamine (PN, Wako Pure Chemical) was used as 1 mM ethanol solution. A bacterial agent, DT-5045 (Bacillus subtilis for the degradation of oil, tar, aromatic hydrocarbons, and phenols Dyna-Bio-Cultures, Environmental Dynamics Inc., Columbia, MO, USA), was supplied by Sanyugiken (Shiki, Japan). Other chemicals employed were of analytical grade.

2.3.

Removal of hydrophobic organic pollutants

To 100 mL of water containing prescribed amount of hydrophobic organic pollutants was added 200 mL of Al(III) solution. With vigorous stirring, the solution pH was adjusted to 7 by the careful addition of 4 M sodium hydroxide solution using a micropipette. It was further stirred for 15 min in order to ensure the complete formation of aluminum hydroxide. Then, the prescribed amount of surfactant solution was added to the suspension. After stirring the mixture for 15 min, the formed precipitates were separated by centrifuging the solution at 1500 rpm (300 g) for 5 min. Hydrophobic organic pollutants remaining in the supernatant were determined by introducing 20 mL-aliquot of the supernatant to the HPLC system.

2.4.

An HPLC system composing of a Jasco (Tokyo, Japan) PU-980 intelligent pump, a UV-970 intelligent ultra-violet detector, and an 807-IT integrator were used for the separation and determination of hydrophobic organic pollutants. Zetapotential of surfactant-Al(OH)3 was measured with a Microtec Nition Zeecom ZC2000 zeta-potential analyzer (Tokyo, Japan). A Jasco 4200 FT-IR spectrometer (Hachioji, Japan) was used for the FT-IR analysis of surfactant-Al(OH)3. Fluorescence spectra were measured with a PerkineElmer LS-50B luminescence spectrometer (Waltham, MA, USA) with a 1 cm quartz cell. Purified water was prepared with Elix UV3 and Milli-Q Gradient A10 Water Purification Systems having a UV irradiation

Materials

Characterization of surfactant-Al(OH)3

The amount of surfactant sorbed on aluminum hydroxide was estimated from the surfactant concentration in the bulk aqueous solution which was spectrophotometrically determined by a methylene blue method (Longwell and Maniece, 1955). For FT-IR analysis, surfactant-Al(OH)3 formed was collected by filtration and was freeze-dried for 72 h. It was well mixed with KBr (1:100) and pressed into the disk. For the evaluation of the hydrophobic properties of surfactant-Al (OH)3, it was prepared from 0.25 mM Al(III) and 1 mM surfactant. A 10-mL portion of PN solution was added to 10 mL of surfactant-Al(OH)3 solution. The emission of PN in the different surfactant-Al(OH)3 systems was measured. The wavelength of the excitation was 340 nm.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 7 9 e1 8 8 9

2.5.

Application to wastewater treatment

In order to clarify the applicability to wastewater treatment, an effluent sample was taken from an outfall of Nagoya University. Particulate materials were removed by passing the sample through an Omunipore hydrophilic PTFE membrane filter (Millipore, pore size: 0.45 mm). A 1-L aliquot of the effluent water was used for the experiment. The prescribed amounts (typically 1.0 mg L1) of hydrophobic organic pollutants to be examined were spiked into the effluent water. Except that the experimental scale was 10-fold, the removal of hydrophobic organic pollutants was performed as described above. The amounts of aluminum(III) ions and surfactant added were 100 mg and 300 mg, respectively. Chemical oxygen demands (CODs) of the waters before and after the treatment were determined by the titration of KMnO4 solution (Moore et al., 1951). Surfactant remained in the treated water was determined by a Methylene blue method (Longwell and Maniece, 1955).

2.6.

Biodegradation

The aqueous suspension (10 mL) of surfactant-Al(OH)3 containing collected hydrophobic organic pollutants was prepared by the treatment of 1 L of effluent sample. The concentration of the hydrophobic organic pollutant in the effluent samples was 1.0 mg L1. The procedures for the treatment by the proposed method were described above. Bacteria were pre-incubated by mixing the 10 g-portion of the bacterial agent (DT-5075) with 100 mL-aliquot of untreated effluent at 30  C for 4 h in a 200 mL Erlenmeyer flask. A 10-mL portion the bacterial solution was added to the aqueous suspension containing concentrated hydrophobic organic pollutants. The resulting mixture was incubated at 30  C in a rotary shaker incubator. Air (30 mL min1) was supplied to the mixture to maintain aerobic conditions. The air was precedently passed through a 0.2 mm membrane filter and water to prevent the contamination with other bacteria. The bacterial activity was determined based on the spectrophotometric measurement at 540 nm of the hydrolysis of fluorescein diacetate (Schnu¨rer and Rosswall, 1982). A 0.5 mLportion of sample was taken for the analysis of residual hydrophobic organic pollutants. To the sample containing surfactant-Al(OH)3 and bacteria were added 1 mL of 4 M hydrochloric acid and 2.5 mL of N,Ndimethylformamide for dissolving surfactant-Al(OH)3 and hydrophobic organic pollutants. Then, 1 mL of 4 M sodium hydroxide was added for the formation of aluminum hydroxide. The resulting suspension was sonicated for 5 min. It was finally passed through a membrane filter (pore size: 0.2 mm). A 20-mL aliquot of the filtrate was introduced into the HPLC system.

3.

Results and discussion

3.1. (OH)3

Formation and characterization of surfactant-Al

It is well-known that polymeric species including Al13O4(OH)7þ 24 can be produced by neutralizing aluminum(III) chloride

1881

solution with sodium hydroxide at room temperature (Bottero et al., 1980). The polymeric species are coagulated themselves to form larger sized aluminum hydroxide gels around pH 7. The formation of aluminum hydroxide gels has been extensively studied (Nail et al., 1976a; Nail et al., 1976b; Fiessinger, 1978; Pierre and Uhlmann, 1986; Duan and Gregory, 2003). The aluminum hydroxide gels have an affinity for lowmolecular weight organic molecules, especially anionic surfactants such as sodium alkylsulfonates and sodium alkanoates (Rakotonarivo et al., 1984, 1985; Streltsova et al., 2002). Similar to alumina-based admicelles, anionic surfactants sorb on the surfaces of aluminum hydroxide gels and likely form monolayered or bilayered aggregates. In the present study, finely dispersed aluminum hydroxide was coagulated to sediments by adding anionic surfactant such as SDS, AOT, or sodium oleate. Warren (1975); Koh et al. (1985) reported that shear flocculation can occur in surfactant-coated ultrafine particles. The coagulation of aluminum hydroxide is ascribable to the hydrophobic association between surfactantAl(OH)3. The sediments were readily separated from bulk aqueous solution by centrifugation, filtration, and floatation techniques. The amount of sorption of SDS, AOT, or sodium oleate on Al(OH)3 formed from 100 mg of Al(III) ions increased with increasing the amount of each surfactant added (Fig. 1A). The sorption of anionic surfactant can be assumed by the decrease in the positive zeta-potential of Al(OH)3 (Fig. 1B). The value of sodium oleate-Al(OH)3 formed by the addition of more than

Fig. 1 e The amount of surfactant sorbed (A) and zetapotential of surfactant-Al(OH)3 (A) as a function of the amount of SDS (B), AOT (:), or sodium oleate (,) added to the aqueous system of aluminum hydroxide formed from 100 mg LL1 of Al(III) ions at pH 7.

1882

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 7 9 e1 8 8 9

50 mg L1 sodium oleate could not be determined because of the formation of large coagulates. The decrease in the positive charge seems additional reason for the surfactant-induced coagulation of Al(OH)3. The FT-IR spectra of surfactants, aluminum hydroxide, and surfactant-Al(OH)3 are shown in Fig. 2. The absorption bands at 2930 and 2860 cm1 are assignable to CeH stretching, while the band at 1460 cm1 is due to CeH deformation. In the spectra of SDS- and AOT-Al(OH)3, the band at 1245 cm1 due to SeO stretching owing to SDS and AOT was observed. An intense band at 1740 cm1 is attributed to the C ¼ O stretching in the ester moiety of AOT. On the other hand, somewhat broad band at 1590 cm1 is assignable to the C ¼ O stretching in oleate. They are also shown in the spectra for AOT- and sodium oleate-Al(OH)3. These results clearly indicate the surfactant sorption to aluminum hydroxide. As described above, the sorption of surfactant molecules can result in the formation of hemimicelle- or admicelle-like aggregates providing hydrophobic media rather than the bulk aqueous solution. Comparing fluorescence spectra of certain molecular probes has extensively been performed for evaluating the microscopic solvent properties of macromolecules or micelles (Stryer, 1965; Esumi et al., 1991). The emission properties of some fluorescent probes in micellar media have also been correlated to polarity parameters (e.g., ET(30), Sengupta et al., 2000). Among the molecular probes, N-phenyl-1-naphthylamine (PN) was useful for evaluating the solvent properties of several types of admicelles (Saitoh et al., 2002b). The wavelength at the maximum emission (lmax) shifted to lower region than that in water. The solvent properties of different types of admicelles were successfully evaluated by the comparison of the lmax values with those in pure solvents. In the present study, the enhancement and shift to lower wavelength in the emission of PN were observed in the surfactant-Al(OH)3 systems. The lmax value (442 nm) of PN in

Fig. 3 e Correlation between relative permittivity and the wavelength of the maximum emission (lmax) of PN in different solvents (1: water, 2: methanol, 3: acetonitrile, 4: ethanol, 5: 1-propanol, 6: 1-butanol, 7: 1-octanol, 8: ethyl acetate, 9: diethyl ether, 10: benzene). Arrows indicate the lmax values in the respective surfactant-Al(OH)3 solutions. The relative permittivity was cited from a reference (Janz and Tomkins, 1972).

water was shifted to 416 nm in the SDS-Al(OH)3 system, 404 nm in the AOT-Al(OH)3 system, and 394 nm in the sodium oleate-Al(OH)3 system. These values are indicated with arrows in Fig. 3, in which the correlation between relative permittivity (Janz and Tomkins, 1972) at 25  C and the lmax value of PN in some pure solvents are shown. The value of SDS-Al(OH)3 suggests the formation of hydrophobic media having alcoholic solvent property. On the other hand, AOT- and sodium oleate-Al(OH)3 seem to provide more hydrophobic media corresponding to ethyl acetate or benzene. These results

Fig. 2 e FT-IR spectra of aluminum hydroxide, SDS-Al(OH)3, AOT-Al(OH)3, sodium oleate-Al(OH)3, and three surfactants.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 7 9 e1 8 8 9

1883

Fig. 4 e Effect of mixing times for the formation of aluminum hydroxide (A) and for the sorption of SDS (B) on the collection of triclosan (B) and nonylphenol (-). The concentrations of Al(III) ions and SDS are 100 and 300 mg LL1, respectively. pH 7, 25  C.

strongly suggest the applicability of surfactant-Al(OH)3 for the efficient collection of hydrophobic organic pollutants in water.

3.2.

Collection of hydrophobic organic pollutants

The collection yield, E (%), of a certain hydrophobic organic pollutant was estimated based on the difference of the initial concentration (Cin [mg L1]) from the residual concentration (Cres [mg L1]) after the removal. Eð%Þ ¼ ðCin  Cres Þ=Cin  100 The collection yield (%) were also obtained by the complete elution of the hydrophobic organic pollutants from surfactantAl(OH)3 gels. The yields were almost identical for the most hydrophobic pollutants examined in the present study. The

formers were adopted because the yield was easily and rapidly obtained with higher reproducibility. The effect of aging time for the formation of aluminum hydroxide before the sorption of SDS on the collection of triclosan and nonylphenol was shown in Fig. 4A. These organic pollutants were well collected when the aging time was 15e20 min. However, the collection yields decreased when the solution was mixed more than 30 min. Fig. 4B shows the effect of the mixing time for the sorption of SDS on the collection of triclosan and nonylphenol. Their collection yields reached to the maximum and constant value after mixing for 15 min. They were constant at least for 1 h. Almost the same results were obtained in the AOT- and sodium oleate-Al(OH)3 systems. Therefore, the times for the respective processes were set to 15 min. The collection yields of triclosan and nonylphenol as a function of the amount of anionic surfactant were depicted in Fig. 5. In the absence of anionic surfactant, they were

Fig. 5 e Effect of the amount of SDS (B), AOT (:), or sodium oleate (,) added on the collection of triclosan and nonylphenol. The concentration of Al(III) ions is 100 mg LL1 pH 7, 25  C.

1884

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 7 9 e1 8 8 9

Table 1 e Collection of hydrophobic organic pollutants to SDS-, AOT- and sodium oleate-Al(OH)3. Organic pollutant

Octylphenol Nonylphenol Triclosan Triclocarban Esprocarb Methoxychlor Pretilachlor Pendimethalin Fluoranthene Pyrene Benz[a]anthracene Benzo[a]pyrene

Collected (%) No surfactant

SDS

AOT

Sodium oleate

4 54  1 0 69  3 1 5 0 1 0 0 4 32  2

83  1 99  1 90  1 93  1 89  1 95  1 79  2 92  1 99  1 100 100 100

90  1 99  1 88  1 99  2 77  1 98  1 25  2 92  1 100 99  1 100 100

93  1 100 96  1 96  1 87  2 91  2 73  2 95  1 100 100 100 100

Al(III): 100 mg L1, surfactant: 300 mg L1, pH 7, average of 5 experiments.

insufficiently collected to aluminum hydroxide. On the other hand, the collection increased with increasing the amount of anionic surfactant owing to the gradual formation of hydrophobic media. The collection of triclosan to SDS-Al (OH)3 or sodium oleate-Al(OH)3 was almost complete by the addition of 300 mg L1 anionic surfactant. On the other hand, the amount of surfactant for the almost complete collection of nonylphenol was 100 mg L1. Nonylphenol having a long alkyl-substituent may be suitable for the incorporation to surfactant aggregates. Strangely, the collection ability of AOT-Al(OH)3 is somewhat lower than SDS-Al(OH)3, although the former provides more hydrophobic media than the latter does. The reason may be ascribed to the difference in the specific volume of the surfactant aggregates. SDS molecule has longer alkyl moiety than AOT. Additionally, the aggregation number of AOT [12e16 (Sheu et al., 1987)] in the bulk aqueous solution is smaller than that of SDS [60e62] (Coll, 1970; Turro and Yekta, 1978). SDS molecules seem to form bulky aggregates on hydroxide surfaces rather than AOT molecules do. This is advantageous for the incorporation of bulky hydrophobic organic pollutants. The collection yields of several hydrophobic organic pollutants are listed in Table 1. Nearly complete collections of hydrophobic organic pollutants except pretilachlor were achieved by the addition of 300 mg L1 of SDS, AOT, or sodium oleate. Low collection yields of pretilachlor can be ascribed to its lower hydrophobicity than other organic pollutants. Next, the effect of the hydrophobicity of organic pollutant on the collection was studied. The solute incorporation to surfactant aggregates such as hemimicelles or admicelles has been represented by binding constant (amount of a solute incorporated in the surfactant aggregates per gram of adsorbed surfactant)/(concentration of the solute in the aqueous phase) (Valsaraj, 1989; Saitoh et al., 2005). The linear relationship between the logarithmic binding constants and logarithmic aqueous-octanol distribution coefficients indicated that hydrophobic interaction is a predominant factor for solute incorporation to the admicelles.

In the present system, the binding constant, Kad, can be estimated by taking account of the solute sorption to aluminum hydroxide. Kad ¼ DE V=Xd ð100  DEÞ where DE represents the collection yield (%) of solute to surfactant aggregates on aluminum hydroxide. It was

Table 2 e Collection and log Kad of hydrophobic organic pollutants to SDS-Al(OH)3. Organic pollutant

SDS added Collected log (Average) log Kad Ko/w (mg) (%)a

Dichlorophenol

40 75 300

71 12  1 17  2

3.3 3.4 3.2

(3.3)

3.06b

Trichlorophenol

40 75 300

13  1 22  1 39  3

3.8 3.8 3.8

(3.8)

3.72b

Octylphenol

40 50 75 100 300

32 42 62 73 83

    

2 3 3 1 1

4.3 4.3 4.5 4.6 4.5

(4.4)

4.12c

Triclosan

50 75 100 300

61 71 80 90

   

1 3 3 1

4.8 4.8 4.9 e

(4.8)

4.8d

Estrone

75 100 300

31  2 42  2 65  1

4.0 4.1 4.1

(4.1)

4.01e

b-Estradiol

75 100 300

32  3 45  1 71  5

4.0 4.2 4.3

(4.2)

4.08e

Esprocarb

50 75 100 300

42 56 68 89

   

2 4 1 1

4.4 4.5 4.6 4.6

(4.5)

4.59f

Methoxyclor

50 75 100 300

77 81 86 95

   

1 3 3 1

5.0 4.9 4.9 e

(4.9)

5.08g

Pretilachlor

75 100 300

33  3 52  3 79  2

4.0 4.2 4.3

(4.2)

4.08h

Pendimethalin

40 75 100 300

56 68 79 92

   

4.8 4.7 4.8 e

(4.8)

5.18i

Al(III): 100 mg L1, pH 7. a average of 6 experiments. b (Banerjee et al., 1984). c (Neamtu et al., 2009). d (Halden and Paull, 2005). e (Ha´jkova´ et al., 2007). f (Brudenell et al., 1995).  et al., 1995). g (Sabljic h (Kawakami et al., 2006). i (Zheng and Cooper, 1996).

5 3 5 1

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 7 9 e1 8 8 9

estimated from the difference in the collection yield, E (%), to surfactant-Al(OH)3 and the collection yield, E0 (%), to uncoated aluminum hydroxide. A symbol V denotes solution volume (L), while Xd represents the amount of surfactant (kg) sorbed on aluminum hydroxide. Table 2 lists the examples of logarithmic Kad estimated from the collection yields at different amount of SDS. The log Kad values are almost independent on the amount of SDS added, indicating the validity of the equilibrium constants. Table 2 also includes aqueous-octanol distribution coefficients (log Ko/w) that are widely accepted as a measure of solute hydrophobicity (Leo et al., 1971). An almost linear relationship between log Kad and log Ko/w was obtained (log Kad ¼ 0.77 log Ko/w þ 0.95, r2 ¼ 0.94, Fig. 6A). This is consistent with the fact that hydrophobic interaction is a predominant factor for the solute incorporation to SDS aggregates on aluminum hydroxide. As shown in Fig. 6B and C, similar relationships were obtained in the AOT-Al(OH)3 (log Kad ¼ 0.85 log Ko/w þ 0.77, r2 ¼ 0.67) and sodium oleate-Al(OH)3 systems (log Kad ¼ 0.99 log Ko/w0.27, r2 ¼ 0.71). These correlations will be useful to design the proposed method for the removal of different hydrophobic organic pollutants. Previously, we reported the collection of polycyclic aromatic hydrocarbons (PAHs) to SDS-impregnated aluminum hydroxide generated by the neutralization of aqueous mixture of SDS, aluminum(III) chloride, and magnesium(II) chloride (Saitoh et al., 2007b). PAHs were collected and concentrated

1885

into the SDS-impregnated hydroxide. However, the extractabilities of PAHs were lower than those obtained in the present study. This may be attributed to the difference in the structure of SDS aggregates. It is well-known that layered hydroxide structures containing surfactant aggregates inside the layers are formed by the hydrolysis of aluminum(III) and magnesium (II) ions (Esumi and Yamamoto, 1998; Pavan et al., 1999; You et al., 2002). Hydrophobic organic compounds are stably retained inside the hydroxides. Furthermore, the composition and property of SDS-impregnated aluminum hydroxide can be significantly dependent on the ratio of Al(III) ions to SDS (Vasilescu et al., 2004). These facts are unfavorable for easy operation in the collection of hydrophobic organic pollutants and their smooth biodegradation. In the present study, we chose surfactant-Al(OH)3 in which surfactant aggregates are formed on the surfaces of aluminum hydroxide.

3.3.

Application to wastewater treatment

The applicability to the practical wastewater treatment was tested by using an effluent sample at an outfall of our university (COD: 3.0 mg L1, hardness: 26.5 mg L1). Since triclosan was not detected, 1 mg L1 of triclosan was added to the sample. The extents of removal were 79  3% in the SDS-Al (OH)3 system, 81  2% in the AOT-Al(OH)3 system, and 95  1% in the sodium oleate-Al(OH)3 system. The extents of collection

Fig. 6 e Relationship between log Kad and log Ko/w in the different surfactant-Al(OH)3 systems.

1886

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 7 9 e1 8 8 9

were slightly lower than those obtained in the treatment of sample prepared with Milli-Q water. This can be attributed to the interference of organic or inorganic components in this effluent sample. On the other hand, the removal of more hydrophobic nonylphenol was complete (>99.9%) in all systems. However, the proposed method increased the COD value of the treated water owing to the residual surfactant. When 300 mg L1 of surfactant was used, the concentrations of remained SDS, AOT, and sodium oleate were 135 mg l1, 65 mg l1, and 22 mg l1, respectively. The COD values were 18.7 mg L1 in the SDS-Al(OH)3 system, 7.7 mg L1 in the AOT-Al (OH)3 system, and 7.1 mg L1 in the sodium oleate-Al(OH)3 system. This fact is a significant problem in the practical application to wastewater treatment. The subsequent treatment of the conventional coagulation-sedimentation technique using aluminum hydroxide was effective to reduce the both of SDS concentration and COD value. When 100 mg L1 of Al(III) ions (as aluminum chloride) was used, the SDS concentration and COD value were reduced to 67 mg L1 and 8.5 mg L1, respectively. They were further reduced to 37 mg L1 and 4.7 mg L1 by the addition of 200 mg L1 of Al(III) ions.

3.4.

Biodegradation of hydrophobic organic pollutants

Finally, the biodegradation of collected hydrophobic organic pollutants was attempted using a bacterial agent, DT-5075. It is reported that synthetic surfactants can be used to enhance the production of bacterial enzymes for the degradation of organic compounds (Reese and Maguire, 1969). Therefore, anionic surfactants used in the present study are expected to accelerate the biodegradation of hydrophobic organic pollutants. The time courses in the decomposition of nonylphenol collected to the different surfactant-Al(OH)3 are shown in Fig. 7. The respective plots were obtained by the estimation from the average of four determinations of residual nonylphenol. The recoveries in the elution of nonylphenol from SDS- and AOT-Al(OH)3 systems were more than 94%, while that from sodium oleate-Al(OH)3 system was 86%. The deviation in the respective determination was within 4%. Nonylphenol collected to AOT-Al(OH)3 or sodium oleate-Al(OH)3 was not sufficiently decomposed. On the other hand, the same compound collected to SDS-Al(OH)3 was rapidly decomposed. In sodium oleate-Al(OH)3 system, the extent of bacterial growth was most remarkable. However, the concentration of

Fig. 7 e Time course of the decomposition of nonylphenol collected to different surfactant-Al(OH)3 precipitates formed from 100 mg of Al(III) ions and 300 mg of anionic surfactant in the treatment of 1 L effluent sample containing 1.0 mg LL1 of spiked pollutant.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 7 9 e1 8 8 9

Table 3 e Decomposition of hydrophobic organic pollutants. Organic pollutant

Decomposed (%) SDS

AOT

3 days 7 days 7 days Octylphenol Nonylphenol Pendimethalin Triclosan Pyrene Benzo[a]pyrene

47 42 44 20 31 39

     

13 11 8 4 6 7

97 91 91 24 53 51

     

2 5 4 5 6 5

45 40 22 21 27 19

     

7 8 6 7 8 9

Sodium oleate 7 days 66 57 47 17 24 21

     

9 12 6 3 5 4

Al(III): 100 mg L1, surfactant: 300 mg L1, average of 5 experiments.

dissolved oxygen was reduced below 0.5 mg L1. Low oxygen concentration is adverse for aerobic biodegradation of organic pollutants. In AOT-Al(OH)3 system, bacterial activity was reduced probably due to high toxicity of AOT. In contrast, the bacterial activity in the SDS-Al(OH)3 system was maintained during the biodegradation. The concentration of dissolved oxygen was kept at ca. 7.5 mg L1, being almost the same as the solubility [7.54 mg L1 or 5.28 mL L1 (Battino, 1966)] of oxygen in water at 30  C. High oxygen concentration is necessary for the decomposition in aerobic conditions. The decomposition rate of nonylphenol collected on SDS-Al(OH)3 was 0.23  0.06 mg L1 h1 (n ¼ 10). This is ca. 10-fold greater than the rate (0.02 mg L1 h1) obtained by the direct addition of the same amount of bacterial agent to the untreated water. The decomposition yields of different hydrophobic organic pollutants are summarized in Table 3. Nonylphenol, octylphenol and pendimethalin collected to SDS-Al(OH)3 were almost completely (>90%) decomposed within 1 week. Pyrene and benzo[a]pyrene were gradually decomposed. Furthermore, apparent decomposition of triclosan was also observed. This greatly contrasts to the results obtained when the bacterial agent was directly poured into the untreated water. Without the collection, the lysis of bacteria was observed in the presence of antifungals triclosan. The collection into SDSAl(OH)3 resulted in the decrease in the concentration of triclosan in the bulk aqueous phase and therefore reduced the toxicity. The application of the proposed method would be useful for the degradation of varieties of hydrophobic organic pollutants in wastewater.

4.

Conclusions

In this study, we proposed a novel design for the wastewater treatment. The hydrophobic organic pollutants that had hardly been removed by the conventional coagulationesedimentation method was well collected to surfactant-coated aluminum hydroxide and efficiently removed from water. The method was compatible to the biodegradation of the hydrophobic organic pollutants.  Surfactant-coated aluminum hydroxides [surfactant-Al (OH)3] were formed by adding anionic surfactant to the aqueous solution of aluminum hydroxide at pH 7. Finely dispersed aluminum hydroxide gels were coagulated to the

1887

precipitates that can readily be separated from water. When sodium dodecyl sulfate (SDS), sodium bis(2-ethylhexyl)sulfosuccinate (AOT), or sodium oleate was used, the precipitates provided hydrophobic media corresponding to alcoholic solvents, ethyl acetate, or benzene. This fact strongly suggests the applicability of the surfactant-Al(OH)3 to the collection of hydrophobic organic pollutants in water.  Hydrophobic organic pollutants including octylphenol, nonylphenol, triclosan, triclocarban, esprocarb, methoxychlor, pretilachlor, pendimethalin, fluoranthene, pyrene, benz[a]anthracene, benzo[a]pyrene were well collected to the surfactant-Al(OH)3 precipitates and efficiently removed from water. The collection yield was dependent on the hydrophobicity of the organic pollutants and the kind or amount of surfactant. In each surfactant-Al(OH)3 system, there was almost linear relationship between logarithmic binding constants to the surfactant on Al(OH)3 and logarithmic aqueous-octanol distribution coefficient.  The method was applied to the removal of nonylphenol and triclosan that had been spiked in an effluent sample. The removal of triclosan was in the range of 79e95%, while that of nonylphenol was almost complete (>99.9%). This result strongly suggests the applicability of the proposed method to the wastewater treatment. However, COD values of the treated water were increased owing to the remained surfactant. The COD was reduced by the subsequent coagulationesedimentation using aluminum hydroxide.  Hydrophobic organic pollutants collected to surfactant-Al (OH)3 precipitates were decomposed with a bacterial agent (Bacillus subtilis, DT-5075). Among the surfactant-Al(OH)3 systems, SDS-Al(OH)3 was the most appropriate for the biodegradation. Nonylphenol, octylphenol, and pendimethalin were almost completely (>90%) decomposed within 1 week. Triclosan, pyrene, and benzo[a]pyrene were gradually decomposed.

Acknowledgments This study was supported by a Grant-in-Aid for Science Research Japan (A) (22241019).

references

Banerjee, S., Howard, P.H., Rosenberg, A.M., Dombrowski, A.E., Sikka, H., Tullis, D.L., 1984. Development of a general kinetic model for biodegradation and its application to chlorophenols and related compounds. Environ. Sci. Technol. 18 (6), 416e422. Battino, B., 1966. Solubility of gases in liquid. Chem. Rev. 66 (4), 395e462. Bottero, J.Y., Cases, J.M., Flesslnger, F., Polrlert, J.E., 1980. Studies of hydrolyzed aluminum chloride solutions. 1. Nature of aluminum species and composition of aqueous solutions. J. Phys. Chem. 84 (22), 2933e2939. Boyd, G.R., Reemtsma, H., Grimm, D.A., Mitra, S., 2003. Pharmaceuticals and personal care products (PPCPs) in surface and treated waters of Louisiana, USA and Ontario, Canada. Sci. Total Environ. 311 (1e3), 135e149. Brudenell, A.J.P., Baker, D.A., Grayson, B., 1995. Phloem mobility of xenobiotics: tabular review of physicochemical properties

1888

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 7 9 e1 8 8 9

governing the output of the Kleier model. Plant Growth Regul. 16 (3), 215e231. Carballa, M., Omil, F., Lema, J.M., Llompart, M., Garcı´a-Jares, C., Rodrı´guez, I., Go´mez, M., Ternes, T., 2004. Behavior of pharmaceuticals, cosmetics and hormones in a sewage treatment plant. Water Res. 38 (12), 2918e2926. Coll, H., 1970. Study of ionic surfactants by membrane osmometry. J. Phys. Chem. 74 (3), 520e528. Costi, E.M., Goryacheva, I., Sicilia, M.D., Rubioa, S., Pe´rezBendito, D., 2008. Supramolecular solid-phase extraction of ibuprofen and naproxen from sewage based on the formation of mixed supramolecular aggregates prior to their liquid chromatographic/photometric determination. J. Chromatogr. A 1210 (1), 1e7. DeLorenzo, M.E., Fleming, J., 2008. Individual and mixture effects of selected pharmaceuticals and personal care products on the marine phytoplankton species Dunaliella tertiolecta. Arch. Environ. Contam. Toxicol. 54 (2), 203e210. Duan, J., Gregory, J., 2003. Coagulation by hydrolyzing metal salts. Adv. Colloid Interface Sci. 100-102, 475e502. Esumi, K., Otsuka, H., Meguro, K., 1991. Characteristics of an adsorbed layer of hydrocarbon and fluorocarbon surfactants on alumina. Langmuir 7 (10), 2313e2316. Esumi, K.S., Yamamoto, S., 1998. Adsorption of sodium dodecyl sulfate on hydrotalcite and adsolubilization of 2-naphthol. Colloids Surf. A Physicochem. Eng. Asp. 137 (1e3), 385e388. Fiessinger, F., 1978. Coagulation and flocculation. Part I. Coagulation. Int. Wat. Supply. Ass. Congr. Proc. 26, E1eE10. Gangula, S., Suen, S.-Y., Conte, E.D., 2010. Analytical applications of admicelle and hemimicelle solid phase extraction of organic analytes. Microchem. J. 95 (1), 2e4. Ha´jkova´, K., Pulkrabova´, J., Sch urek, J., Hajslova´, J., Poustka, J., Na´pravnı´kova´, M., Kocourek, V., 2007. Novel approaches to the analysis of steroid estrogens in river sediments. Anal. Bioanal. Chem. 387 (4), 1351e1363. Halden, R.U., Paull, D.H., 2005. Co-occurrence of triclocarban and triclosan in US water resources. Environ. Sci. Technol. 39 (6), 1420e1426. Hiraide, M., Iwasawa, J., Kawaguchi, H., 1997. Collection of trace heavy metals complexed with ammonium pyrrolidinedithiocarbamate on surfactant-coated alumina sorbents. Talanta 44 (2), 231e237. Janz, G.J., Tomkins, R.P.T., 1972. Nonaqueous Electrolytes Handbook, vol. 1. Academic Press, New York. Kawakami, T., Eun, H., Arao, T., Endo, S., Ueji, M., Tamura, K., Higashi, T., 2006. Concentration and loading of several pesticides in water, suspended solids and sediment during ordinary water discharge in Sugao marsh, Ibaraki Prefecture, Japan. J. Pestic. Sci. 31 (1), 6e13. Koening, L., 1967. The cost of water treatment by coagulation, sedimentation, and rapid sand filtration. J. Am. Water Works Assoc. 59 (3), 290e336. Koh, P.T.L., Uhlherr, P.T.H., Andrews, J.R.G., 1985. The effect of capillary condensation and liquid bridging on the bonding of hydrophobic particles in shear-flocculation. J. Colloid Interface Sci. 108 (1), 95e103. Laws, S.C., Carey, S.A., Ferrell, J.M., Bodman, G.J., Cooper, R.L., 2000. Estrogenic activity of octylphenol, nonylphenol, bisphenol A and methoxychlor in rats. Toxicol. Sci. 54 (1), 154e167. Leo, A., Hansch, C., Elkins, D., 1971. Partition coefficients and their uses. Chem. Rev. 71 (6), 525e616. Longwell, J., Maniece, W.D., 1955. Determination of anionic detergents in sewage, sewage effluents, and river waters. Analyst 80 (948), 167e171. Merino, F., Rubio, S., Perez-Bendito, D., 2003. Solid-phase extraction of amphiphiles based on mixed hemimicelle/

admicelle formation: application to the concentration of benzalkonium surfactants in sewage and river water. Anal. Chem. 75 (24), 6799e6899. Moore, W.A., Ludzack, F.J., Ruchhoft, C.C., 1951. Determination of oxygen-consumed values of organic wastes. Anal. Chem. 23 (9), 1297e1300. Moral, A., Sicilia, M.D., Rubio, S., Pe´rez-Bendito, D., 2008. Multifunctional sorbents for the extraction of pesticide multiresidues from natural waters. Anal. Chim. Acta 608 (1), 61e72. Nail, S.L., White, J.L., Hem, S.L., 1976a. Structure of aluminum hydroxide Gel I: initial precipitate. J. Pharmacol. Sci. 65 (8), 1188e1191. Nail, S.L., White, J.L., Hem, S.L., 1976b. Structure of aluminum hydroxide gel II: aging mechanism. J. Pharmacol. Sci. 65 (8), 1192e1195. Neamtu, M., Popa, D.-M., Frimmel, F.H., 2009. Simulated solar UVirradiation of endocrine disrupting chemical octylphenol. J. Hazard. Mater. 164 (2e3), 1561e1567. Ohe, T., Watanabe, T., Wakabayashi, K., 2004. Mutagens in surface waters: a review. Mutat. Res. Rev. Mutat.Res. 567 (2e3), 109e149. Pavan, P.C., Crepaldi, E.L., Gomes, G.D., Valim, J.B., 1999. Adsorption of sodium dodecylsulfate on a hydrotalcite-like compound. Effect of temperature, pH and ionic strength. Colloids Surf. A Physicochem. Eng. Asp. 154 (3), 399e410. Pelletier, E., Sargian, P., Payet, P.J., Demers, S., 2006. Ecotoxicological effects of combined UVB and organic contaminants in coastal waters: a review. Photochem. Photobiol. 82 (4), 981e993. Pierre, A.C., Uhlmann, D.R., 1986. Amorphous aluminum hydroxide gels. J. Non Cryst. Solids 82 (1e3), 271e276. Rakotonarivo, E., Bottero, J.Y., Cases, J.M., Fiessinger, F., 1984. Study of the adsorption of long chain alkylsulfonates from aqueous solutions on aluminum hydroxide gels. Colloids Surf. 9 (3), 273e292. Rakotonarivo, E., Bottero, J.Y., Cases, J.M., Fiessinger, F., 1985. Study of the adsorption of long chain sodium soaps from aqueous solutions on aluminum hydroxide gels. Colloids Surf. 16 (2), 153e173. Randtke, S.J., Stephen, J., 1988. Organic contaminant removal by coagulation and related process combination. J. Am. Water Works Assoc. 80 (5), 40e56. Reese, E.T., Maguire, A., 1969. Surfactants as stimulants of enzyme production by microorganisms. Appl. Microbiol. 17 (2), 242e245. , A., Gu¨sten, H., Verhaar, H., Hermens, J., 1995. QSAR Sabljic modelling of soil sorption. Improvements and systematics of log KOC vs. log KOW correlations. Chemosphere 31 (11e12), 4489e4514. Saitoh, T., Nakayama, Y., Hiraide, M., 2002a. Concentration of chlorophenols in water with sodium dodecylsulfate-galumina admicelles for high-performance liquid chromatographic analysis. J. Chromatogr. A 972 (2), 205e209. Saitoh, T., Taguchi, K., Hiraide, M., 2002b. Evaluation of hydrophobic properties of sodium dodecylsulfate/g-alumina admicelles based on fluorescence spectra of N-phenyl-1naphthylamine. Anal. Chim. Acta 454 (2), 203e208. Saitoh, T., Matsushima, S., Hiraide, M., 2004. Aerosol-OT-galumina admicelles for the concentration of hydrophobic organic compounds in water. J. Chromatogr. A 1040 (2), 185e191. Saitoh, T., Matsushima, S., Hiraide, M., 2005. Concentration of polyaromatic hydrocarbons in water to sodium dodecyl sulfateg-alumina admicelle. J. Chromatogr. A 1069 (2), 271e274.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 7 9 e1 8 8 9

Saitoh, T., Kondo, T., Hiraide, M., 2007a. Concentration of chlorophenols in water to dialkylated cationic surfactantsilica gel admicelles. J. Chromatogr. A 1164 (1e4), 40e47. Saitoh, T., Matsushima, S., Hiraide, M., 2007b. Flotation of polycyclic aromatic hydrocarbons coprecipitated with aluminum hydroxide containing sodium dodecyl sulfate and magnesium. Colloids Surf. A Physicochem. Eng. Asp. 199 (1e3), 88e92. Sandeep, P.N., Sabatini, D.A., Harwell, J.H., 1994. Surfactant adsolubilization and modified admicellar sorption of nonpolar, polar, and ionizable organic contaminants. Environ. Sci. Technol. 28 (11), 1874e1881. Schnu¨rer, J., Rosswall, T., 1982. Fluorescein diacetate hydrolysis as a measure of total microbial activity in soil and litter. Appl. Environ. Microbiol. 43 (6), 1256e1261. Sengupta, B., Guharay, J., Sengupta, P.K., 2000. Characterization of fluorescence emission properties of prodan in different reversed micellar environments. Spectrochim. Acta A 56 (7), 1431e1441. Sheu, E.Y., Chen, H.S., Huang, J.S., 1987. Structure and growth of bis(2-ethylhexyl) sulfosuccinate micelles in aqueous solutions. J. Phys. Chem. 91 (12), 3306e3310. Streltsova, E.A., Hromysheva, E.A., Tymchuk, A.F., 2002. The adsorption of anionic surfactants by iron(III) and aluminium hydroxides. Adsorpt. Sci. Technol. 20 (8), 757e765. Stryer, L., 1965. The interaction of a naphthalene dye with apomyoglobin and apohemoglobin. A fluorescent probe of non-polar binding sites. J. Mol. Biol. 13 (2), 482e495.

1889

Suarez, S., Lema, J.M., Omil, F., 2009. Pre-treatment of hospital wastewater by coagulation-flocculation and flotation. Biores. Technol. 100 (7), 2138e2146. The´bault, P., Casses, J.M., Fiessinger, F., 1981. Mechanism understanding the removal of organic micropollutants during flocculation by aluminum or iron salt. Water Res. 15 (2), 183e189. Turro, N.J., Yekta, A., 1978. Luminescent probes for detergent solutions. A simple procedure for determination of the mean aggregation number of micelles. J. Am. Chem. Soc. 100 (18), 5951e5952. Valsaraj, K.T., 1989. Partitioning of hydrophobic nonpolar volatile organics between the aqueous and surfactant aggregates phases on alumina. Sep. Sci. Technol. 24 (14), 1191e1205. Valsaraj, K.T., 1992. Adsorption of trace hydrophobic compounds from water on surfactant-coated alumina. Sep. Sci. Technol. 27 (12), 1633e1642. Vasilescu, M., Angelescu, D., Caldararu, H., Almgren, M., Khan, A., 2004. Fluorescence study on the size and shape of sodium dodecyl sulphate-aluminium salt micelles. Colloids Surf. A Physicochem. Eng. Asp. 235 (1e3), 57e64. Warren, L.J., 1975. Shear-flocculation of ultrafine scheelite in sodium oleate solutions. J. Colloid Interface Sci. 50 (2), 307e318. You, Y.W., Zhao, H.T., Vance, G.F., 2002. Surfactant-enhanced adsorption of organic compounds by layered double hydroxides. Colloids Surf. A Physicochem. Eng. Asp. 205 (3), 161e172. Zheng, S.Q., Cooper, J.F., 1996. Adsorption, desorption, and degradation of three pesticides in different soils. Arch. Environ. Contam. Toxicol 30 (1), 15e20.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 9 0 e1 8 9 6

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Assessment of the UV/Chlorine process as an advanced oxidation process Jing Jin, Mohamed Gamal El-Din, James R. Bolton* Department of Civil and Environmental Engineering University of Alberta, Edmonton, AB, Canada T6G 2W2

article info

abstract

Article history:

Several organic compounds were used as radical scavengers/reagents to investigate the

Received 19 August 2010

possibility of the UV/chlorine process being used as an advanced oxidation process (AOP) in

Received in revised form

the treatment of drinking water and wastewater. The UV/H2O2 process was selected as

2 December 2010

a reference, so that the results from the UV/chlorine process could be compared with those

Accepted 6 December 2010

of the UV/H2O2 process. Methanol was added to active chlorine solutions at both pH 5 and

Available online 13 December 2010

10 and into hydrogen peroxide samples. The photodegradation quantum yields and the


OH radical production yield factors, which are significant in evaluating AOPs, were

Keywords:

calculated for both the UV/chlorine and the UV/H2O2 processes. The yield factor for the UV/

Advanced oxidation

chlorine process at pH 5 was 0.46  0.09, which is much lower than that of the UV/H2O2

Chlorine photolysis

process, which reached 0.85  0.04. In addition to methanol, para-chlorobenzoic acid

Hydroxyl radicals

( pCBA) and cyclohexanoic acid (CHA) were added to active chlorine solutions and to H2O2

Methanol

solutions, to evaluate the efficiencies of oxidizing these organic compounds. The specific

p-Chlorobenzoic acid

first-order reaction rate constants for the oxidation of pCBA and CHA, using the UV/chlo-

Cyclohexanoic acid

rine process, were lower than those found using the UV/H2O2 process. ª 2010 Elsevier Ltd. All rights reserved.

1.

Introduction

1.1.

Brief introduction to the UV/chlorine process

Excessive consumption of chlorine under direct sunlight has been observed in swimming pools that use aqueous chlorine as a disinfectant. Nowell and Hoigne´ (1992a) confirmed that the predominant active species from the UV/chlorine process is the
OH radical. The production of
OH radicals, when aqueous chlorine solutions are exposed to UV, has enabled the UV/chlorine process to become a potential advanced oxidation process. The UV driven chlorine process, as an AOP, can be a treatment option for disinfection by-products (DBPs) that are produced during chlorine disinfection in swimming pools (Judd and Jeffrey, 1995; Kim et al., 2002), and can be used to

inactivate water-borne pathogenic microorganisms and to destroy hazardous organic compounds in drinking water and wastewater. Several reactions occur in the UV/chlorine process (Bolton, 2010; Feng et al., 2007), such as: Cl2 þ H2O / HOCl þ HCl

(1)

HOCl 4 Hþ þ OCle (equilibrium with pKa ¼ 7.6 at 20  C)

(2)

HOCl þ UV photons /
OH þ Cl


(3)

OCl þ UV photons /
Oe þ Cl


(4)

O þ H2O /
OH þ OHe

(5)




* Corresponding author. Tel.: þ1 780 439 4709. E-mail address: [email protected] (J.R. Bolton). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.12.008

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 9 0 e1 8 9 6

Due to the presence of natural organic matter (NOM) in natural waters, there are various kinds of organic compounds that can either become promoters in the production of
OH radicals, or can become radical scavengers, which could cause chain reactions in the UV/chlorine process and hence lead to further consumption of HOCl. This was noted by Oliver and Carey (1977), who carried out a series of experiments around pH 4 using radical scavengers, such as ethanol, n-butanol and benzoic acid. When methanol was added into samples at pH 5, where HOCl predominates, there occur chain reactions and oxidative reactions as follows:


OH þ CH3OH /
CH2OH þ H2O

(6)




CH2OH þ HOCl / ClCH2OH þ
OH

(7)




CH2OH þ O2 /
O2CH2OH / HCHO þ HO2


(8)

1.2.

Quantum yield (f) and yield factor (h)

The quantum yield is a parameter that quantifies the amount of active chlorine [Each Cl2 produces only one ‘active Cl’ (as one Cl2 molecule decomposes into HOCl or OCle at pH 5 or 10, which are the pH values adapted in this research)] consumed during photolysis. Thus, the quantum yield for the photodegradation of active chlorine is defined as moles of active chlorine decomposed per einstein [One einstein is 1 mol (6.023  1023) of photons] of UV photons absorbed by the sample. The photodegradation quantum yields observed by Buxton and Subhani (1972) in the photolysis of OCle ions at ambient temperature were 0.85, 0.39 and 0.60 for wavelengths of 254, 313 and 365 nm, respectively. They also defined a yield factor h to describe the fraction of decomposed active chlorine that generates
OH and is defined as: h¼

D½,OH D½active Cl

  1 C0 k‫׳‬1 ¼ ln CF F

(10)

F ¼ E0 ðWFÞðDFÞðPFÞðRFÞt ¼ E0 ðavgÞt

(11)

10k‫׳‬1 Ul lnð10Þ3C

rate with
OH radicals and that the yield factors (h) for active chlorine photolysis at 255 nm were 0.85 and 0.1 at pH 5 and 10, respectively. Eqs. (10)e(12) enable the calculation of the photodegradation quantum yields of active chlorine when oxidizing various organic compounds and allow a connection between the quantum yields and the first-order reaction rate constants.

1.3.

Objectives

This paper focuses on evaluating the potential of the UV/ chlorine process as an AOP from two aspects: the
OH radical production yield factor and the photo-oxidation rate constants of certain organic compounds. The yield factor can be used in assessing the
OH radical production ability of a certain process; and the photo-oxidation rate constants directly indicate the efficiency of a selected compound that is treated by a given process. The organic compounds chosen for study were methanol (which is known to trigger chain reactions in the UV/chlorine process), para-chlorobenzoic acid ( pCBA) [used as an
OH radical probe by Watts and Linden (2007)] and cyclohexanoic acid (CHA) [used as model compound for the toxic naphthenic acid components in oil sands tailing water (OSTW)]. Furthermore, the above two factors obtained in the UV/chlorine process were compared with those of UV/H2O2 process, which is used widely as an AOP in the water and wastewater treatment industry. This approach allows an assessment of whether or not the UV/ chlorine process could become an effective AOP in the treatment of waste streams.

2.

Methods

2.1.

UV collimated beam apparatus

(9)

Bolton and Stefan (2002) developed a protocol to calculate quantum yields from fluence (UV dose) based on rate constants using a collimated beam apparatus. In this approach, the photodegradation quantum yields for active chlorine can be calculated as follows:

FC ¼

1891

(12)

where FC is the quantum yield of substance C; k10 is the fluence based first-order rate constant (m2 J1); 3C is the molar absorption coefficient (M1 cm1) for substance C; C0 and CF are the initial and final concentrations of the substance under photolysis; F is fluence (J me2); E0 is incident fluence rate (W me2); E0(avg) is the average fluence rate in the solution; Ul is molar photon energy (J einstein1); WF, DF, PF and RP are the water factor, divergence factor, Petri factor and reflection factor, respectively, as defined by Bolton and Linden (2003). Nowell and Hoigne´ (1992b) found on using chlorobutane and nitrobenzene probes that both react at nearly the same

A quasi collimated beam UV apparatus (Model PSI-I-120, Calgon Carbon Corporation, USA) equipped with a low pressure high output (LPHO) UV lamp (LSI Inc.) was used to generate quasi-parallel UV at 254 nm. The irradiance in the UV collimated beam was measured by a UV detector (International Light, Model SED240) connected to a radiometer (International Light, Model IL 1400A). The irradiance was measured by placing the detector in the center under the collimated beam by adjusting the calibration marker of the detector to a height that was the same as the top of the solution when placed under the collimated beam. Furthermore, a calibration of the radiometer was necessary before the measurement of the irradiance from the LP UV lamp. The radiometer was calibrated by the KI/KIO3 actinometer method (Bolton et al., 2009). The irradiance, as determined in the calibration procedure, was found to be essentially the same as that given on the readout of the radiometer, with the error for the reading being only 0.1%.

2.2.

Materials and methods

Analytical grade chemicals were used for the preparation of all the samples. MilliQ water was generated by a Maxima Ultra Pure Water System and was used for the preparation of

1892

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 9 0 e1 8 9 6

2.3.

Sample analysis

The DPD colorimetric method was adopted to measure the chlorine concentration according to standard methods (APHA et al., 1995). The concentration of hydrogen peroxide was determined by measuring its absorbance at 240 nm, where 3240 ¼ 38.1 M1 cm1 (Goldstein et al., 2007). Formaldehyde, produced after methanol has been oxidized in the UV/chlorine or UV/H2O2 process, was measured by the Nash method (Nash, 1953). Reversed-phase high pressure liquid chromatography (HPLC) with a UV detector was used for the analysis of pCBA. Analysis was implemented using a 150 mm  4.6 mm Gemini C18 (2) column with a 5 mm particle size (Phenomenex, Torrance, CA, USA). The mobile phase consisted of 45% 10 mM phosphoric acid and 55% methanol. The flow rate was set to 0.7 mL/min, and the sample injection volume was 100 mL (Vanderford et al., 2007). The analysis of CHA was carried out using an HPLC connected to an ion trap mass spectrometer (Varian 50-MS) equipped with an electrospray interface operating in negative ion mode, along with unit mass resolution using an analytical Luna C8 (5 mm, 150  3 mm, and 250  3 mm) column (Phenomenex, Torrance, CA). The experimental temperature for the chromatography was about 40  C. The mobile phase consisted of 100% methanol containing 4 mM ammonium acetate and 0.1% acetic acid in aqueous solution. The concentration of methanol was ramped from 40% to 80% over 20 min. The mobile phase flow rate was 200 mL/min and the injection volume was 20 mL (Han et al., 2008).

3.

Results and discussion

3.1. Photodegradation quantum yield of active chlorine at pH 5 and 10 According to the reactions (6)e(8) as shown in Section 1.1, knowing that methanol can have effects on increasing the photodegradation quantum yield of active chlorine at pH 5 and 10, methanol was selected as the first ‘challenge’ scavenger (and also a reactant), and samples with various methanol concentrations were prepared to investigate the photodegradation quantum yields and yield factors for the UV/chlorine process.

Solutions containing 9.9e86.5 mM methanol and approximately 50 mg/L (1.41 mM) active chlorine were buffered to pH 5. Each sample was exposed to UV in a collimated beam apparatus for 300 s with a low pressure high output (LPHO) UV lamp (Light Sources Inc.) at an irradiance of 0.38 mW/cm2. According to Feng et al. (2007), the production of formaldehyde can be detected for the UV/chlorine process in the presence of methanol using the method of Sun and Bolton (1996). The results showed that the photodegradation quantum yield for 50 mg/L active chlorine at pH 5 did depend on the concentration of methanol. From Fig. 1, the quantum yield increased linearly from (1.0  0.1) up to (16.3  0.3) mol einstein1 when the methanol concentration increased from 0.0 to 86.5 mM. The quantum yields obtained in this research (at the same methanol doses) were smaller than those obtained by Feng et al. (2007). The reason is the difference in the initial chlorine concentrations: the initial active chlorine concentration was 213 mg/L in their research, which was much higher than 47 mg/L used in this research. Feng et al. (2007) showed that chlorine concentrations can result in higher quantum yields in the presence of methanol. Parallel experiments were carried out for active chlorine solutions at pH 10 in the presence of methanol to investigate whether or not the photodegradation quantum yield of OCle is affected by the addition of methanol. Samples containing approximately 50 mg/L free chlorine at pH 10 and various concentrations of methanol ranging from 9.9 to 61.8 mM were also prepared. Then the samples were exposed under the same LPHO UV lamp (at an irradiance of 0.38 mW/cm2) in a collimated beam apparatus for 600 s. The photodegradation quantum yields for the trials were calculated to be 1.15  0.08 mol E1 over the full range of methanol concentrations, indicating that the quantum yields of active chlorine at pH 10 do not depend on the methanol dose. This agrees with the findings of Feng et al. (2007).

3.2. Quantum yield of the production of
OH radicals in the UV/H2O2 process in the presence of methanol Theoretically, the photodegradation quantum yields of H2O2 solutions in the presence of methanol should still be about 1.0,

20 Quantum Yield (mol/einstein)

solutions. Chlorine samples were prepared using a 10%e15% (by weight) sodium hypochlorite solution obtained from SigmaeAldrich, and the concentration of active chlorine was measured by the DPD total chlorine reagent using 5 mL samples (Hach, Anachemia Canada Inc.). Hydrogen peroxide samples were prepared by using a 30% hydrogen peroxide solution from Fisher Scientific. Excess H2O2 was removed by using catalase (oxidoreducatase obtained from Sigma), and the remaining catalase was filtered from the solutions using a 0.2 mm nylon syringe filter (purchased from Whatman). HPLC grade methanol was purchased from Fisher Scientific. Reagent grade pCBA was obtained from Aldrich (Canada), and 98% analytical reagent grade CHA was also obtained from Aldrich (Canada).

16

12

8

4

0 0

20

40 60 [CH3OH] (mM)

80

100

Fig. 1 e Photodegradation quantum yield of active chlorine (50 mg/L) at pH 5.0 in the presence of various concentrations (mM) of methanol.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 9 0 e1 8 9 6

since the photodecomposition of H2O2 presumably cannot trigger chain reactions that could increase the quantum yield. Samples, containing 151.7 mg/L (4.5 mM) H2O2 with methanol concentrations varying from 1.0 to 49.3 mM, were prepared and exposed to the LPHO UV at an irradiance of 0.38 mW/cm2 in a collimated beam apparatus for 2700 s. Now, however, unlike the increasing photodegradation quantum yields in the UV/chlorine process with increasing methanol concentration, the photodegradation quantum yields for production of
OH radicals in the UV/H2O2 process remained essentially constant at (1.1  0.1) mol einstein1, indicating that an increase in the methanol dose does not stimulate an increase in the photodegradation quantum yields in the UV/H2O2 process.

3.3. Generation of
OH radicals in the UV/chlorine process To investigate the
OH radical production yield factor during the UV/chlorine process, the production of formaldehyde was measured for the samples that contained methanol after UV exposure. Yield factors h were calculated for the samples at pH 5 knowing that the UV/chlorine process had a higher yield factor at this pH (Nowell and Hoigne´, 1992b). For the samples containing 9.9e61.8 mM methanol and approximately 50 mg/L active chlorine, which were discussed in Section 3.1, the residual chlorine was quenched after the UV exposure by adding 50 mL of a 20 g/L sodium thiosulphate solution to eliminate the interference of residual active chlorine on the formaldehyde produced. Since the efficiency of
OH radicals reacting with methanol and forming
CH2OH in Eq. (6) is 93% (Asmus et al., 1973), the
OH radical production can be determined by the amount of formaldehyde produced. The relation between the formaldehyde production (mM) and the addition of methanol (mM) is shown in Fig. 2. Fig. 2 shows that the formaldehyde production increased as the methanol concentration increased. When the methanol concentration rose to about 25 mM, the formaldehyde

1.0

0.4

0.9 0.8 0.3 0.6 0.2

0.5 0.4 0.3

Yield Facotr

[HCHO] (mM)

0.7

0.1 0.2 0.1 0.0 0

20

40 60 [CH3OH] (mM)

80

0.0 100

Fig. 2 e Formaldehyde production (-) as a function of the methanol concentration (mM) showing a plateau region and the yield factors (D) of $OH radical production in the UV/chlorine process at pH 5.

1893

production became almost constant. At lower methanol concentrations, since methanol/chlorine ratio was lower, the competition of active chlorine with methanol for
OH radicals under such conditions decreased the production of formaldehyde. At high methanol concentrations, methanol became the dominant radical scavenger, and thus the production of formaldehyde became constant and independent of the methanol concentration. The yield factors h [calculated according to Eq. (9)] for the
OH radical production for the samples discussed in an earlier part of this section, were also determined and are shown in Fig. 2. The yield factors stayed constant at about 0.46  0.09, which were much smaller than 1.0 and were smaller than the value obtained by Nowell and Hoigne´ (1992b), which was 0.85. The lower yield factor achieved in this study could arise from two factors. First, Nowell and Hoigne´ (1992b) used chlorobutane and nitrobenzene as probes, and 1-octanol and acetate as model scavengers to measure the
OH radicals produced. Whereas in this study, methanol was used as radical scavenger and the
OH radical production was derived from the production of formaldehyde. Second, due to the chain reactions caused by methanol, extra chlorine was consumed, so the yield factors were reduced lower than 1.0 and indeed lower than the value achieved by Nowell and Hoigne´ (1992b).

3.4.

Generation of
OH radicals in the UV/H2O2 process

The production of
OH radicals was also calculated by measuring the production of formaldehyde after the H2O2 solutions were exposed to the LPHO UV. The formaldehyde concentrations were measured after the samples [containing 151.7 mg/L (4.5 mM) H2O2 with the methanol concentration varying from 1.0 to 49.3 mM] had been exposed under the UV collimated beam apparatus. The residual hydrogen peroxide was quenched using catalase, and the samples were filtered before the concentrations of formaldehyde were measured. Similar to that of UV/chlorine process, the relation between formaldehyde production and methanol concentration formed a plateau region at a methanol dose of approximately 25 mM (see Fig. 3), around which methanol becomes the dominant radical scavenger, similar to the results obtained in the UV/chlorine process. The yield factors for the samples were also calculated by ratio of the
OH produced and the H2O2 decomposed during the photolysis, and are shown in Fig. 3 together with the formaldehyde production. It can be observed from Fig. 3 that the yield factor increased as more methanol was introduced into the sample solutions. Note that the yield factor shows a similar behaviour to the yield factor of in the UV/chlorine process. When the methanol concentration reached around 25 mM, the yield factor became constant at about 0.85  0.04. It should be mentioned that at low methanol doses, such as 1.0 and 2.5 mM, although the formaldehyde production had already reached about 0.13 mM and had begun to become stabilized, a considerable consumption of hydrogen peroxide was observed. As a result, the yield factors at low methanol doses were much smaller than 1.0 and close to 0.

1894

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 9 0 e1 8 9 6

1.0

0.20

0.9 0.8 0.7 0.12

0.6 0.5

0.08

0.4

Yield Factor

[HCHO] (mM)

0.16

0.3 0.04

0.2 0.1

0.00

0.0 0

10

20

30 40 [CH3OH] (mM)

50

60

Fig. 3 e Formaldehyde production (-) as a function of the methanol concentration (mM) showing a plateau region and the yield factors (D) of $OH radical production in the UV/H2O2 process.

3.5. Degradation of para-chlorobenzoic acid in the UV/ chlorine process Samples were prepared with various pCBA concentrations to examine the possibility that the concentration of the organic compound might affect the reaction rates. Two samples containing approximately 50 mg/L active chlorine and pCBA concentrations of 7.1 and 13.7 mM were prepared and exposed to UV for 600e3000 s, yielding UV doses of 228e1140 mJ/cm2. The pH of the sample solutions was buffered to 5, since the photodegradation quantum yields for the UV/chlorine process and the
OH production rates are greater at pH 5 than at pH 10, as described in the latter part of Section 3.1. Fig. 4 shows plots of ln([pCBA]) versus the reaction time, and the fit lines show a pseudo first-order reaction behaviour for pCBA reacting with
OH radicals using the UV/chlorine process. It can be observed from Fig. 4 that, although the initial concentrations of pCBA varied, the reaction rate constant remains about the same value, which indicates that, as expected, the impact of the initial concentration of pCBA on the pseudo first-order reaction rate constant was minor. Further investigation of the oxidation of pCBA in the UV/ chlorine process was carried out by adding methanol as an

Fig. 4 e Semi-log plot of the concentration of pCBA in the UV/chlorine process.

‘interfering’ scavenger into chlorine solutions containing pCBA. This allowed a study of the impact of the methanol concentration on the reaction rate of pCBA with
OH radicals. Three samples each containing approximately 50 mg/L active chlorine around pH 5 were prepared. Methanol was added to solutions to yield concentrations of approximately 15, 40 and 99 mM, and the pCBA concentration for each sample was 12.4, 10.0, and 14.3 mM, respectively. As discussed earlier in this section, the concentration of pCBA has virtually no effect on the rate constants for pCBA reacting with
OH radicals; hence, the various pCBA initial concentrations should not be of concern. The samples so prepared were exposed under the LPHO UV lamp at an irradiance of 0.38 mW/cm2 in a collimated beam apparatus for 240e1200 s to yield UV doses of 91.2e456.0 mJ/ cm2. The degradation rate constant of pCBA was lower in the presence of methanol, and the pseudo first-order reaction rate constant decreased from 6.2  104 to 4.9 104 to 2.2  104 s1 as the methanol concentration increased from 15 to 40e99 mM, respectively. Also the rate constants were all lower than the value of (7.3  0.2)  104 s1 obtained when there was no methanol present, as illustrated in the earlier part of this section. Fig. 5 shows a better demonstration of the decrease of the pseudo first-order reaction rate constants (s1) of pCBA in the presence of methanol in the UV/chlorine process. It can be inferred from Fig. 5 that the pseudo first-order reaction rate constant (s1) decreased linearly with a slope of 5.0  103 (M1 s1 ) with increasing methanol concentrations (mM).

3.6. Degradation of para-chlorobenzoic acid in the UV/ H2O2 process The UV/H2O2 trials were designed in parallel to those of the UV/chlorine trials, and the pseudo first-order reaction rate constants of pCBA were compared to those of UV/chlorine process. Three samples were prepared containing approximately 200 mg/L H2O2 and pCBA concentrations of 5.3, 9.4 and 14.6 mM, and were labelled as Sample 1, Sample 2, and Sample 3,

Fig. 5 e The relation between the pseudo first-order reaction rate constants (sL1) for pCBA degradation and the methanol concentration (mM).

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 9 0 e1 8 9 6

Table 1 e Summary of k1(sL1) and k1(specific) (sL1) for the degradation of pCBA in UV/chlorine and UV/H2O2 Processes. k1 (s1)

k1(specific) (s1)

(7.2  0.1)  104

(5.7  0.1)  103

k1 (s1)

k1(specific) (s1)

(1.9  0.01)  103

(9.9  0.1)  103

respectively. For Sample 3, methanol was added to yield a concentration of approximately 40 mM to investigate whether methanol would interfere with the oxidation of pCBA in the UV/H2O2 process similar to that in the UV/chlorine process. Then, the samples were exposed to LPHO UV in the collimated beam apparatus for 180e900 s. The pseudo firstorder reaction rate constants for the degradation of pCBA by
OH radicals were then calculated; the decay curves and rate constants are shown in Fig. 6. It can be easily concluded from Fig. 6 that the initial concentration of pCBA had no impact on the reaction rate constants. Furthermore, the addition of methanol greatly reduced the oxidation rate of pCBA in the UV/H2O2 process by reducing the pseudo first-order rate constant of pCBA from (1.91  0.01)  103 to (3.9  0.5)  104 s1, similar to that for the UV/chlorine process.

3.7. Degradation of cyclohexanoic acid in the UV/ chlorine process Following the examination of the oxidation of pCBA, cyclohexanoic acid (CHA) was added into chlorine solutions to further investigate the UV/chlorine process, and the results were later compared to those obtained in the UV/H2O2 process. Samples were prepared containing approximately 50 mg/L active chlorine at pH 5 and 0.54 mM CHA and were exposed to the LPHO UV at an irradiance of 0.38 mW/cm2 in a collimated beam apparatus for 1200, 1800, 2400, 3000 and 3600 s. The pseudo first-order reaction rate constant for CHA was found to be only (6.3  0.3)  105 s1, which is much smaller than that (1.91  103 s1) obtained for pCBA. This means that the rate

3.000

2.000

ln [pCBA]

1.000

0.000 0

500

1000

1500

2000

-1.000

-2.000

-3.000 Time (s) 5.3 µM

9.4 µM

constant for reaction of the
OH radical with CHA must be much smaller than that for pCBA.

3.8. Comparison of the UV/chlorine and the UV/H2O2 processes using specific rate constants

UV/H2O2 Process

UV/chlorine Process

1895

14.6 µM

Fig. 6 e Semi-log plot of the concentration of pCBA for various methanol concentrations (mM) in the UV/H2O2 process.

Since the aqueous chlorine and hydrogen peroxide components have different absorbance at 254 nm when samples are exposed to LPHO UV, the fraction of UV absorbed by the samples is different in each case. Thus, when comparing the pseudo first-order reaction rate constants of pCBA in the UV/ chlorine and the UV/H2O2 processes, the rate constants should be converted to ‘specific rate constants’ defined as k1 ðspecificÞ ¼

k1 1  10A254

(13)

where k1(specific) is the ‘specific’ pseudo first-order reaction rate constant (s1) at 254 nm; k1 is the experimental reaction rate constant (s1) calculated directly from a plot of ln([pCBA]) versus time (s); and A254 is the absorbance at 254 nm of the chlorine sample containing pCBA. In effect, the specific rate constant is the rate constant that would be obtained if all of the UV would be absorbed. Table.1 summarizes the k1 (s1) and k1(specific) (s1) values for the oxidation of pCBA in the UV/chlorine and UV/H2O2 processes without the addition of methanol. Note that the ratio of the k1(specific) values for pCBA between the UV/chlorine process and the UV/H2O2 process is almost the same as the ratio of the yield factors between the two processes.

4.

Conclusions

The photodegradation quantum yields and yield factors of the UV/chlorine process and the UV/H2O2 process were investigated. Also, the pseudo first-order reaction constants of pCBA and CHA oxidized by both of the AOPs were calculated and compared in this study. The photodegradation quantum yields of active chlorine at pH 5 no longer stayed around 1.0 when the methanol was added into the samples. However, those for active chlorine at pH 10 remained constant at about 1.0 mol E1, which indicated that the chlorine photolysis at pH 10 was not affected by the addition of organic compounds. Similar to those of the UV/ chlorine process at pH 10, the photodegradation quantum yields remained constant at about 1.0 mol E1 during the UV/ H2O2 process. The formaldehyde production during UV/chlorine process at pH 5 and the UV/H2O2 process formed a plateau area and reached maximum values at a methanol dose of approximately 25 mM, which indicates that as methanol dose increased to certain points, methanol becomes the dominant radical scavenger. Most importantly, the higher maximum
OH radical production yield factors in the UV/H2O2 process (0.85  0.04) indicated that this process was more efficient in producing
OH radicals than the UV/chlorine process (0.46  0.09) in the presence of methanol. Thus, for the waste streams that contain methanol, the UV/chlorine process is not a superior process for the removal of organic compounds.

1896

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 8 9 0 e1 8 9 6

The pseudo first-order reaction rate constant for pCBA was not affected by the initial pCBA concentrations in both UV/ chlorine and UV/H2O2 processes. The addition of methanol largely decreased the reaction rate constants for the oxidation of pCBA, and the rate constants for the degradation of pCBA decreased linearly as the methanol dose increased. The UV/ H2O2 process achieved larger ‘specific’ pseudo-first order reaction rate constants for the oxidation of pCBA, indicating that the efficiency of the UV/chlorine process oxidizing pCBA was lower than that of the UV/H2O2 process. The pseudo first-order reaction rate constant for CHA in UV/chlorine process is rather small, indicating that the UV/ chlorine process might not be suitable for the degradation of CHA which is one of the compounds in the OSTW. Thus, for the OSTW treatment, the efficiency of UV/chlorine process would appear to be lower than that of the UV/H2O2 process. However, more investigation should be carried out to further evaluate the potential of the UV/chlorine process.

references

APHA, AWWA and WEF, 1995. Standard Methods for the Examination of Water and Wastewater. 1015 Fifteenth Street. In: 4500-Cl G. DPD Colorimetric Method, nineteenth ed. American Public Health Association, NW, Washington, DC 2005. Asmus, K.-D., Mockel, H., Henglein, A., 1973. Pulse radiolytic study of site of OH radical attack on aliphatic alcohols in aqueous-solution. Journal of Physical Chemistry 77 (10), 1218e1221. Bolton, J.R., 2010. Ultraviolet Applications Handbook, third ed. ICC Lifelong Learn Inc., Edmonton, Canada. Bolton, J.R., Linden, K.G., 2003. Standardized of methods for fluence (UV dose) determination in bench-scale UV experiments. Journal of Environmental Engineering 129 (3), 209e216. Bolton, J.R., Stefan, M.I., 2002. Fundamental photochemical approach to the concepts of fluence (UV dose) and electrical energy efficiency in photochemical degradation reactions. Research on Chemical Intermediates 28 (7e9), 857e870. Bolton, J.R., Stefan, M.I., Shaw, P.S., Lykke, K.R., 2009. Determination of the Quantum Yields the Ferrioxalate and KI/ KIO3 Actinometers and a Method for the Calibration of

Radiometer Detectors. CDROM Proc., 5th Congress on Ultraviolet Technologies, Amsterdam, The Netherlands. International Ultraviolet Association, Scottsdale, AZ. Buxton, G.V., Subhani, M.S., 1972. Radiation-chemistry and photochemistry of oxychlorine ions. Part 1 e radiolysis of aqueous-solutions of hypochlorite and chlorite Ions. Journal of the Chemical Society, Faraday Transactions I 68, 947e957. Feng, Y.G., Smith, D.W., Bolton, J.R., 2007. Photolysis of aqueous free chlorine species (HOCl and OCle) with 254 nm ultraviolet light. Journal of Environmental Engineering and Science 6 (3), 277e284. Goldstein, S., Aschengrau, D., Diamant, Y., Rabani, J., 2007. Photolysis of aqueous H2O2: quantum yield and applications for polychromatic UV actinometry in photoreactors. Environmental Science & Technology 41 (21), 7486e7490. Han, X., Scott, A.C., Fedorak, P.M., Bataineh, M., Martin, J.W., 2008. Influence of molecular structure on the biodegradability of naphthenic acids. Environmental Science & Technology 42 (4), 1290e1295. Judd, S.J., Jeffrey, J.A., 1995. Trihalomethane formation during swimming pool water disinfection using hypobromous and hypochlorous acid. Water Research 29 (4), 1203e1206. Kim, H., Shim, J., Lee, S., 2002. Formation of disinfection byproducts in chlorinated swimming pool water. Chemosphere 46 (1), 123e130. Nash, T., 1953. The colorimetric estimation of formaldehyde by means of the Hantzsch reaction. Biochemical Journal 55 (3), 416e421. Nowell, L.H., Hoigne´, J., 1992a. Photolysis of aqueous chlorine at sunlight and ultraviolet wavelengths. I. Degradation rates. Water Research 26 (5), 593e598. Nowell, L.H., Hoigne´, J., 1992b. Photolysis of aqueous chlorine at sunlight and ultraviolet wavelengths. II. Hydroxyl radical production. Water Research 26 (5), 599e605. Oliver, B.G., Carey, J.H., 1977. Photochemical production of chlorinated organics in aqueous-solutions containing chlorine. Environmental Science & Technology 11 (9), 893e895. Sun, L.Z., Bolton, J.R., 1996. Determination of the quantum yield for the photochemical generation of hydroxyl radicals in TiO2 suspensions. Journal of Physical Chemistry 100 (10), 4127e4134. Vanderford, B.J., Rosario-Ortiz, F.L., Snyder, S.A., 2007. Analysis of p-chlorobenzoic acid in water by liquid chromatographytandem mass spectrometry. Journal of Chromatography A 1164 (1e2), 219e223. Watts, M.J., Linden, K.G., 2007. Chlorine photolysis and subsequent OH radical production during UV treatment of chlorinated water. Water Research 41 (13), 2871e2878.