WATER RESEARCH A Journal of the International Water Association
Editor-in-Chief Mark van Loosdrecht Delft University of Technology and KWR Watercycle Research The Netherlands E-mail:
[email protected]
Editors Jean-Claude Block Université H. Poincaré, Nancy I France David Dixon University of Melbourne Australia Hiroaki Furumai 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 ETH Zürich and Eawag Swiss Federal Institute of Aquatic Science and Technology Switzerland Wolfgang Rauch University of 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 University of California, Davis USA
Associate Editors Andrew Baker University of New South Wales Australia Damien Batstone University of Queensland Australia
Guang-Hao Chen Hong Kong University of Science & Technology China Tom Curtis University of Newcastle upon Tyne UK Francis de los Reyes III North Carolina State University USA Ana Deletic Monash University USA Rob Eldridge Glen Waverley Australia Monica B. Emelko University of Waterloo Canada Rosina Girones University of Barcelona Spain Stephen Gray Victoria University Australia Kate Grudpan Chiang Mai University Thailand Edwin Herricks University of Illinois at Urbana-Champaign USA Peter Hillis United Utilities Plc UK Hong-Ying Hu Tsinghua University China Bruce Jefferson Cranfield University UK Ulf Jeppsson Lund University Sweden Sergey Kalyuzhnyi Moscow State University Russian Federation Jae-Hong Kim Georgia Institute of Technology USA Günter Langergraber University of Natural Resources and Applied Life Sciences, Vienna (BOKU) Austria
Yoshihiko Matsui Hokkaido University Japan Armand Maul Université de Metz France Max Maurer Eawag Switzerland How Yong Ng National University of Singapore Singapore Satoshi Okabe Hokkaido University Japan Jong Moon Park Pohang University of Science & Technology Korea
Shang-Lien Lo National Taiwan University Taiwan Dionisis Mantzavinos Technical University of Crete Greece
Anastasios Zouboulis Aristotle University of Thessaloniki Greece
Susan Richardson US Environmental Protection Agency USA Miguel Salgot University of Barcelona Spain David Sedlak University of California, Berkeley USA Jean-Philippe Steyer LBE-INRA France Masahiro Takahashi Hokkaido University Japan Nathalie Tufenkji McGill University Canada Kai Udert Eawag Switzerland Vayalam Venugopalan BARC Facilities India Eduardo von Sperling Federal University of Minas Gerais Brazil Julie Zilles University of Illinois at Urbana-Champaign USA
Editorial Office 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: For enquiries relating to the submission of articles (including electronic submission) please visit this journal’s homepage at http://www.elsevier.com/ locate/watres. Contact details for questions arising after acceptance of an article, especially those relating to proofs, will be provided by the publisher. You can track accepted articles at http://www.elsevier.com/trackarticle. You can also check our Author FAQs at http://www.elsevier.com/authorFAQ and/or contact Customer Support via http://support.elsevier.com. Orders, claims, and journal enquiries: Please contact the Elsevier Customer Service Department nearest you: St. Louis: Elsevier Customer Service Department, 3251 Riverport Lane, Maryland Heights, MO 63043, USA; phone: (877) 8397126 [toll free within the USA]; (+1) (314) 4478878 [outside the USA]; fax: (+1) (314) 4478077; 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 ) 4 0 6 9 e4 0 8 0
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Effect of interferences on the breakthrough of arsenic: Rapid small scale column tests Vu L. Nguyen, Wei-Hsiang Chen, Thomas Young, Jeannie Darby* Department of Civil and Environmental Engineering, University of California, Davis, One Shields Avenue, Davis, CA 95616, USA
article info
abstract
Article history:
The influences of three important interferences (silica, phosphate, and vanadate) and the
Received 27 January 2011
effect of different pH levels and initial arsenate concentrations on the breakthrough of
Received in revised form
arsenic in adsorptive media columns were examined by using the Rapid Small Scale
15 April 2011
Column Test with a 352 fractional factorial design. Three commercially available adsor-
Accepted 18 April 2011
bents used for arsenic removal (E33, GFH and Metsorb) were tested. Results indicated that
Available online 23 April 2011
GFH was more susceptible to water quality changes than Metsorb and E33 under conditions tested. GFH also adsorbed more anions than the other two media. The pH was the factor
Keywords:
that had the most impact on the performance of the columns, followed by arsenic
RSSCT
concentration and silica concentration. Lowering pH from 8.3 to 7.0 resulted in an increase
Adsorption
of the mean bed volume treated until 10 mg/L arsenic breakthrough by 40, 12 and 18
Arsenic
thousands BV treated by GFH, E33 and Metsorb columns, respectively. However, at high
Silica
silica concentration, lowering pH did not increase the performance of the media. GFH and
Phosphate
Metsorb were more sensitive to changes in arsenic concentration at low pH than at high pH.
Vanadium
Although vanadium and phosphate were previously reported to reduce arsenic adsorption in batch tests, in column mode with the presence of competitors, their effect was insignificant compared to that of pH, arsenic or silica under the conditions used in this study. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
In ground water, arsenic is typically present in one of two oxidation states: arsenite e As(III) and arsenate e As(V), with the latter form dominant under oxidizing conditions (Cullen and Reimer, 1989). Arsenic exposure is known to be associated with skin, lung, liver, kidney and prostate cancer (NRC 1999). In 2001, the maximum contaminant level (MCL) of arsenic was lowered from 50 mg/L to 10 mg/L in the U.S.; compliance was required by 2006. The Best Available Technologies for arsenic treatment recommended by the United States Environmental Protection Agency (US-EPA) include modified coagulationfiltration, modified lime softening, oxidation filtration, adsorption on activated aluminum, ion exchange, and reverse osmosis. However, this technology list was established in 2001,
before the more stringent MCL of 10 mg/L was promulgated for arsenic. Since then, a number of new adsorptive media have been introduced including titanium, zirconium and iron based adsorbents. Fixed-bed adsorbers are relatively simple to operate and these media have the potential to selectively and effectively remove arsenic (US-EPA, 2005). The removal of arsenic from water via adsorption on metal oxide media is impacted by pH as well as the presence of competing ions, including silica, vanadium, and phosphate. Other ions commonly present in ground water, including iron, manganese, nitrate, chloride, sulfate, calcium and magnesium, can also affect adsorption, although their effects have been found to be less detrimental (Zhang et al., 2007; Pokhrel and Viraraghavan, 2008; Mak et al., 2009). Arsenic adsorption capacity on metal oxides was reported to be reduced in the
* Corresponding author. Tel.: þ1 530 752 5670; fax: þ1 530 752 7872. E-mail address:
[email protected] (J. Darby). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.04.037
4070
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 6 9 e4 0 8 0
presence of silica in previous batch studies (Meng et al., 2000, 2002; Amy et al., 2005). Adsorption of silica has also been reported to reduce the surface potential of adsorbents, with the iso-electric point of E33 (a ferric oxide adsorbent) and that of GFH (Granular Ferric Hydroxide) decreased 2 units and 4 units, respectively, in the presence of 63 mg/L silica (Westerhoff et al., 2006). Westerhoff et al. (2006) reported that silica is the main competitor that occupied the adsorption surface, with silicon contributing to 1.87% and 7.35% of the total surface atoms of the spent E33 and GFH media, respectively. Lower iso-electric points and more surface site competition lead to lower arsenic adsorption capacity. Amy et al. (2005) reported a reduction of 20% in arsenic adsorption capacity on E33, 40% on GFH and 10% on Metsorb (a titanium-based metal oxide adsorbent) in the presence of 13.5 mg/L silica at pH 6 in batch tests. When pH was adjusted to 8, the adsorption capacity reduction increased to approximately 70% for all three adsorbents. Phosphate has been reported to have similar adsorption kinetics to arsenate and thus can compete with arsenic for adsorption onto iron hydroxides (Luengo et al., 2007). Amy et al. (2005) reported a reduction of 42% in arsenic adsorption capacity on E33, 38% on GFH and 8% on Metsorb in the presence of 250 mg/L phosphate at pH 6 in batch tests. When pH was adjusted to 8, the adsorption capacity reduction increased to approximately 65%, 75% and 68% for E33, GFH and Metsorb, respectively. Zeng et al. (2008a) and Hongshao and Stanforth (2001) reported the competition of phosphate with arsenic in adsorption on iron-based media and suggested that there are specific sites and common sites for both arsenic and phosphate on the surface of the adsorbents. Youngran et al. (2007) found that at a concentration of 0.5 mg/L and higher, phosphate significantly decreased arsenic removal on Fe2O3. The vanadate ion has similar characteristics to that of phosphate and was found to bind more strongly than phosphate on iron hydroxide, thus it is a competitor with arsenate in adsorption onto iron-based adsorbents (Peacock and Sherman, 2004; Blackmore et al., 1996). Youngran et al. (2007) in batch tests found that at a concentration of 10 mg/L, vanadium (V) did not impose any effect on arsenic adsorption capacity on iron oxides. However, when the concentration of vanadium (V) was greater than 50 mg/L, the adsorption capacity of arsenic was reduced. The effect of competing ions on arsenic removal in a column mode has been examined in only a few studies. Zeng et al. (2008b) conducted a column test with an iron hydroxide adsorbent using buffered water with either 20 mg/ L silica or 39 mg/L phosphate. The addition of silica reduced the number of bed volumes treated before arsenic breakthrough (10 mg/L) from approximately 30,000 to 10,000. A similar effect was observed when the water was spiked with phosphate. Westerhoff et al. (2006, 2008) monitored the breakthrough of silica, phosphate and vanadium in Rapid Small Columns Tests (RSSCT) conducted for arsenic removal from different ground water sources as supplementary data to the breakthrough of arsenic. Because the breakthrough of anions depends on the specific quality of the influent and water quality was different in each water source, the effect of each competitive anion on the breakthrough of arsenic was not examined at different levels. Speitel et al. (2010)
conducted comparative RSSCTs with E33 using water buffered by NaHCO3 and spiked with single solute (arsenic), bisolute (arsenic - vanadium, arsenic - silica, arsenic - Ca2þ) and tri-solute (arsenic-silica- Ca2þ) to investigate the effect of pH and other ions on arsenic removal. The results showed that the bed volumes treated (for arsenic removal) were reduced 50% in the presence of 70 mg/L of vanadium and more than 90% with 50 mg/L of silica. Other than those four described above, the effects of interferences with arsenic adsorption have been examined in batch experiments. Although batch studies illustrate the potential deleterious effects of silica, phosphate, and vanadium on arsenic removal via adsorption, they are insufficient for quantifying the impact on adsorbent performance in continuous flow columns. The goal of this research was to provide information useful to designers regarding media selection in the presence of competing anions and the competitive adsorption of arsenate and other anions in column mode. A variety of adsorbents for arsenic removal have emerged in recent years and are being marketed commercially; three of the most commonly used adsorbents were tested in this study. The influences of the three important interferences (silica, phosphate, and vanadate) and the effect of different pH levels and initial arsenate concentrations on the breakthrough of arsenic in adsorptive media columns were examined by using the RSSCT with a 352 fractional factorial design. The result of this study is critical to water utilities with limited operating budgets given the extreme reduction in media life possible from commonly observed levels of competing constituents.
2.
Experiment design and methods
2.1. The rapid small scale column test and column parameters The RSSCT approach was originally developed for testing the adsorption of organic compounds with granular activated carbon (Crittenden et al., 1987, 1986). It uses a smaller column loaded with an adsorbent ground to smaller particle sizes to simulate the breakthrough curve of the full size or pilot column. Because of the similarity of mass transfer processes and hydrodynamic characteristics, the breakthrough curves of the small column and pilot/full scale systems are expected to be similar (Crittenden et al., 1991). The RSSCT has been used successfully to simulate arsenic breakthrough curves by porous metal oxide adsorptive media in pilot and full scale systems (Westerhoff et al. (2005)). In the RSSCT, it is assumed that the surface diffusion coefficient of a compound is a linear function of the particle diameter and that surface diffusion is the controlling mechanism. The relationships between design parameters in an RSSCT and a full scale column are described in the following equations assuming proportional surface diffusivity (Crittenden et al., 1986, 1987, 1991; Westerhoff et al., 2005): EBCTSC dSC ¼ EBCTLC dLC
(1)
4071
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 6 9 e4 0 8 0
VSC dSC ReSC;MIN SC ¼ VLC dLC ReLC SC
Table 2 e Levels of each factor in the influent of the RSSCTs.a
(2)
Column pH Arsenic (mg/L)
where EBCT is empty bed contact time; d is particle diameter; V is hydraulic loading rate; Re is Reynolds numbers; and Sc is Schmidt number. The subscript SC and LC indicate a smallscale column (i.e., RSSCT column) and a large-scale column (i.e., pilot- or full-scale columns), respectively. The design and operating parameters of the RSSCTs are shown in Table 1. For comparison, the small scale columns were scaled to simulate large-scale columns with an EBCT of 3 min, which is in the range recommended by adsorbent vendors (Amy et al., 2005). The mean particle sizes of both E33 and GFH in full scale columns (1.16 mm) and of Metsorb (0.67 mm) were obtained from the manufacturers. Since the particle diameters of Metsorb are different than that of E33 and GSH in full scale columns, it is impossible to have both the same particle diameter and the same EBCT in the small scale columns packed with Metsorb and those packed with E33/GFH (see Eq. (1)). The same particle diameter (0.127 mm) was used in all the RSSCTs. It was reported in the literature that Metsorb can compact easily, thus leading to high head loss and potential leaking from the head of the column (Westerhoff et al., 2005), thus the flow rate used for the Metsorb columns (6 mL/min) was half of that of the columns packed with E33 and GFH (12 mL/min). The resulting product of Re number and Sc number were within the range of 200 to 200,000 recommended to minimize axial dispersion in small scale columns (Westerhoff et al., 2005).
2.2.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27
8.30 8.30 8.30 8.30 8.30 8.30 8.30 8.30 8.30 7.65 7.65 7.65 7.65 7.65 7.65 7.65 7.65 7.65 7.00 7.00 7.00 7.00 7.00 7.00 7.00 7.00 7.00
15 15 15 35 35 35 55 55 55 15 15 15 35 35 35 55 55 55 15 15 15 35 35 35 55 55 55
Silica (mg/L)
Vanadium (mg/L)
Phosphorus (mg/L)
33 53 73 33 53 73 33 53 73 33 53 73 33 53 73 33 53 73 33 53 73 33 53 73 33 53 73
21 41 61 41 61 21 61 21 41 41 61 21 61 21 41 21 41 61 61 21 41 21 41 61 41 61 21
55 55 55 155 155 155 105 105 105 105 105 105 55 55 55 155 155 155 155 155 155 105 105 105 55 55 55
a Each column test was conducted with three media (E33, GFH and Metsorb).
Experimental design
Small adsorptive columns were fed with water having different pH, silica, vanadate, phosphate and arsenate levels. Each of the five factors was selected following a 352 fractional factorial experimental design, which was adapted from Xu (2005). In total, 27 RSSCTs were conducted for each of the three media; factors were varied in accordance with Table 2. This experimental design ensures that the level of any single factor is not correlated with the level of another factor or with the product of any two factors (Xu, 2005), thus the effect of one factor is not confounded with that of another or with the interaction effect of any two factors. Only the individual effect of each factor and the interaction of any two factors were considered important to the performance of the media and all higher interaction levels among the factors on the media were either verified to be negligible or investigated further, as described below.
Table 3 shows the correlation coefficients of the single factors and interaction of factors in this experimental design. From this table, only phosphate and arsenic had a strong correlation with a threeefactor interaction. The level of phosphate had a strong correlation with a threeefactor interaction of pH, arsenic and silica (see italic values on Table 3). Because pH, silica and arsenic were found to impose a strong effect on the performance of the media (as will be discussed later), the effect of phosphate calculated from the result of this experimental design may be the result of the combined effect of pH, silica and arsenic together rather than of phosphate itself. Additional experiments were added to clarify the effect of phosphate. Also shown by italic values on Table 3, the level of arsenic was found to have a strong correlation with the interaction of pH, silica and vanadium. However, vanadium was
Table 1 e Design and operating parameters for the RSSCTs. Parameter
E33 and GFH Small column
I.D (cm) Mean particle size (mm) Bed volume (mL) Media depth (cm) EBCT (min) Flow rate (mL/min)
0.7 0.127 3.9 10.2 0.33 12
Metsorb
Large column 1.16
3
Small column 0.7 0.127 3.4 8.9 0.57 6
Large column 0.67
3
4072
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 6 9 e4 0 8 0
Table 3 e Correlation of factors, generated from the experimental design.a
pH As Si V P pHAs pHSi pHV pHP AsSi AsV AsP VP pHAsSi pHSiV pHAsV AsSiV
pH
As
Si
V
P
pHAs
pHSi
pHV
pHP
AsSi
AsV
AsP
VP
pHAsSi
pHSiV
pHAsV
AsSiV
1.00 0.00 0.00 0.00 0.00 0.15 0.22 0.17 0.18 0.00 0.00 0.18 0.00 0.27 0.12 0.11 0.02
1.00 0.00 0.00 0.00 0.99 0.00 0.00 0.04 0.81 0.73 0.74 0.00 0.72 0.80 0.72 0.63
1.00 0.00 0.00 0.00 0.97 0.00 0.00 0.54 0.00 0.00 0.00 0.00 0.53 0.00 0.42
1.00 0.00 0.00 0.00 0.98 0.00 0.00 0.62 0.00 0.69 0.00 0.00 0.62 0.54
1.00 0.04 0.00 0.00 0.98 0.00 0.00 0.61 0.67 0.61 0.03 0.03 0.04
1.00 0.03 0.03 0.11 0.79 0.72 0.77 0.03 0.77 0.81 0.73 0.62
1.00 0.04 0.04 0.52 0.00 0.04 0.00 0.06 0.55 0.01 0.40
1.00 0.03 0.00 0.61 0.03 0.68 0.05 0.01 0.63 0.52
1.00 0.03 0.03 0.66 0.66 0.67 0.09 0.08 0.06
1.00 0.59 0.59 0.03 0.58 0.99 0.58 0.78
1.00 0.54 0.43 0.53 0.58 0.99 0.87
1.00 0.41 0.99 0.62 0.57 0.51
1.00 0.41 0.05 0.45 0.42
1.00 0.63 0.57 0.50
1.00 0.58 0.77
1.00 0.85
1.00
a Numbers 0.5 indicate a strong relation. Bold values indicate a strong correlation of two terms that do not contain any common factor.
later found to not have a significant effect on the media, therefore the interactions involving vanadium are not important and the effect of arsenic was assumed clear (not confounded with any other interaction). No additional experiment was considered necessary to clarify the effect of arsenic.
2.3.
Adsorbents
Three commonly used and commercially available adsorbents were included in this research: Bayoxide E33 (iron based), Granular Ferric Hydroxide (GFH) (iron based) and Metsorb (titanium-based). E33 and Metsorb are dry media while GFH is received as a wet material. For E33 and Metsorb, the original dry media were crushed using a mortar and a pestle and serially sieved through two stainless steel sieves (100 and 140-mesh). The fraction on the 140-mesh sieve, which has an average particle size of 0.128 mm, was collected and stored in amber glass bottles until use. For GFH, crushing and sieving were conducted in water and the sieved media were stored wet until use.
2.4.
Water source and chemical spiking
Tap water from the University of California at Davis, drawn from wells located at depths of approximately 325 m, was used. The tap water receives no treatment other than chlorination. Before adjustment, the initial pH was 8.3 and the concentrations of arsenate (as As), vanadate (as V), phosphate (as P), and silica (as SiO2) were 5, 21, 55 mg/L, and 33 mg/L, respectively. The levels of each factor used in the experiments were chosen based on its initial level in the tap water and its typical range in California’s ground water. The mean (maximum) concentration of silica, vanadium and phosphate in Californa’s ground waters used for drinking water are 38 (97) mg/L, 10 (77) mg/L and 20 (888) mg/L, respectively (CDPH, 2009). The concentrations of other inorganic constituents in the water are listed in Table 4. Prior to each RSSCT, tap water was placed in 200 gallon drums. The silica concentration was adjusted to the selected levels by adding sodium meta-silicate nonahydrate. The arsenic and vanadium concentrations were adjusted to the
desired levels using 500 mg/L stock solution prepared from sodium hydrogen arsenate heptahydrate and trisodium orthovanadate, respectively. The phosphorus concentration was adjusted using a stock solution of 5 g/L phosphorus prepared from potassium phosphate monobasic. After thorough mixing, the pH was adjusted to the desired level using concentrated nitric acid. All chemicals were reagent grade. The pH of the water in the drums was monitored twice a day and adjusted as necessary. The pH of the effluent remained consistently within 0.05 units of the influent.
2.5.
Column packing and pumps
Glass chromatography columns with internal diameter of 0.7 cm and length of 20 cm (Kontes FlexColumn Economy, Kimble Chase, NJ) were used. These columns have a 4 mL reservoir at the head and a 20 micron porous plastic disk in the bed support to help retain the media. The inlet and outlet of
Table 4 e UC Davis tap water quality.a Parameter NO3-N (mg/L) SO4-S (mg/L) Ca (Soluble) (meq/L) Mg (Soluble) (meq/L) Hardness (mg/L as CaCO3) Na (meq/L) Cl(meq/L) Mn (Soluble) (mg/L) Fe (Soluble) (mg/L) HCO3(meq/L) CO32- (meq/L) TDS (mg/L) Alkalinity (meq/L)
Concentration 1.07 10.6 0.87 1.60 123 3.29 0.5 <0.05 <0.1 3.5 0.7 300 4.2
a Sample taken on 02/18/09 and analyzed by Agricultural and Natural Resources (ANR) laboratory in the University of California at Davis. Because of the well depth, concentrations do not change significantly over time.
4073
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 6 9 e4 0 8 0
2.6.
Sampling and analysis
Effluent from the columns was collected in 50 mL plastic vials twice a day until the effluent concentration of arsenic reached the influent concentration. The pH was measured immediately, after which samples were stored at 4 C for later analysis of arsenic, vanadium, phosphate. Samples for silica were taken every 2 h until full breakthrough and analyzed colorimetrically for silica (DR/890 test kit, Hach, Denver, CO). To measure the arsenic and vanadium concentration, the samples were acidified with 1% of concentrated nitric acid (trace metal grade, Fisher Scientific, PA) and analyzed with ICP-MS (Agilent 7500i, Palo Alto, CA). Phosphate was also analyzed colorimetrically (FIA QuikChem, Lachat, Denver, CO).
2.7.
Data processing and analysis
Samples were analyzed and the arsenic, silica, vanadium, and phosphate breakthrough data were obtained. The bed volumes treated until 10 mg/L arsenic breakthrough (BV10) of each column was determined by interpolation. Statistical analysis was then conducted using Statistica software (Statsoft, Tulsa, OK) to identify the effect of each factor and the interaction of the factors on arsenic breakthrough from each adsorbent. The linear term of the main effect of one factor was calculated by the difference of the mean BV10 at the highest and lowest level of that factor. The value for the quadratic effect was defined as the difference between the mean BV10 at the mid-level of a factor and the average of the mean BV10 at the high and low level of that factor. An interaction effect of two factors was present if the difference between the linear main effects of one factor at different levels of the other factor was statistically significant, meaning that one factor did not impose the same effect on the BV10 at different levels of another factor. The student’s t test and the p values were used to check the statistical significance of the effects; the more important an effect is, the larger its t value and the smaller its p value (Kutner et al., 2005). A threshold p value of less than 0.05 was used in this research.
from the 27 columns of the three adsorbents are shown in Nguyen (2011). Typical breakthrough curves of anions from the columns are shown in Fig. 1. In all columns and across all three adsorbents, silica was the first to fully breakthrough (i.e., effluent concentration equals influent concentration, and the cumulative removal reaches a plateau), after just a few thousand bed volumes (Fig. 2). In the GFH and Metsorb columns, the subsequent breakthrough order was phosphate, arsenic and then vanadium whereas in the E33 columns, the breakthrough order varied with ion concentrations. In all of the GFH and Metsorb columns, full breakthrough of phosphate occurred before that of arsenic and vanadium regardless of the anion influent concentrations and the pH of water indicating that GFH and Metsorb were more favorable for arsenic and vanadium adsorption than for phosphate under the range of concentration utilized here, which encompasses values found in California groundwaters used as drinking water sources. The arsenic loading at column exhaustion of E33, GFH and Metsorb in all 27 columns is shown in Table 5. GFH adsorbed vanadium the most, followed by Metsorb. E33 had the lowest capacity for vanadium. Badruzzaman and Westerhoff (2005) and Westerhoff et al. (2006) also reported more vanadium was removed in the GFH column than in the E33 column. The experiment was stopped when full breakthrough of arsenic occurred so the maximum vanadium loading capacity of GFH and Metsorb was not determined. However, the loading capacity is expected to be high, because at the time the experiment was terminated, the effluent vanadium concentration was only 20 mg/L, while the influent concentration was 61 mg/L. For phosphate, the solid phase loading on GFH was higher than E33 and Metsorb adsorbed phosphate the least. The phosphate adsorption capacity ranged from 1.6 to 19.1 mg-P/cm3, 1.3e9.6 mg-P/cm3 and 0.7e6.0 mg-P/cm3 for GFH, E33 and Metsorb, respectively (Table 5). Generally, the loading capacity for arsenic was highest for GFH, followed by E33 and then Metsorb (Table 5). Only at the highest pH (8.3) and the lowest influent arsenic concentration (15 mg/L) (columns 1,2,3) were the adsorption capacities of E33 and GFH comparable. When pH decreased, the difference in arsenic loading capacity of GFH and E33 increased. The cumulative removal of silica by the three media at three pH values and a silica influent concentration of 33 mg/L is shown in Fig. 2. For all adsorbents, the amount of silica adsorbed depended on pH, with higher silica removal at higher pH. These results agree with results from batch 1.2
Phosphate 1 0.8 C/C
the column have luer-locks that can be connected to Teflon tubing with a high density polyethylene barb. The sieved adsorbent was first rinsed with deionized water until the water was clear, then transferred in slurry form to the column filled with deionized water to avoid air entrapment, and then backwashed. The media in the column in the column was allowed to settle to reach equilibrium. A layer of glass wool, presoaked in deionized water, was placed on the top of the media. A 5 cm layer of 3 mm glass beads was placed on top of the glass wool layer. Piston pumps with ceramic pump heads (Fluid Metering Inc., Syosset, NY) were used to deliver water to the columns via teflon tubing.
0.6
Arsenic 0.4
3.
Results and discussion
3.1. Breakthrough order of anions and their solid phase loading capacity All of the data collected in this study as well as the resulting breakthrough curves for arsenic, vanadium and phosphate
Vanadium
0.2 0 0
10,000
20,000
30,000
40,000
50,000
60,000
70,000
Bed volumes treated
Fig. 1 e Breakthrough curves from column 1: E33 (triangle), GFH (circle) and Metsorb (square).
4074
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 6 9 e4 0 8 0
60
Set 4
a pH
Set 7
Table 5 e Solid phase loading of arsenate (mg-As/cm3), vanadium (mg-V/cm3) and phosphate (mg-P/cm3) on the media.
8.3 GFH
3
Silica removal (mg/cm )
Set 1 40
E33
20
Metsorb
E33
0
Set 10
3
Silica removal (mg/cm )
60 Set 13
Set 16
b
pH 7.65
40
GFH 20
Metsorb
E33
0
Set 19
3
Silica removal (mg/cm )
60 Set 22
Set 25
c
pH 7.0
40
Metsorb 20
E33 GFH
0 0
2,000
4,000
6,000
8,000
10,000
GFH a
Column
As
V
P
As
V
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27
1.6 0.9 0.6 1.4 1.2 0.9 2.8 2.4 2.1 1.3 1.1 0.9 4.3 3.3 3.4 4.2 2.9 2.7 1.7 1.8 1.5 5.0 3.9 3.5 6.8 5.6 4.3
2.9 2.3 3.5 2.6 3.6 1.4 4.4 1.2 6.0 4.2 4.7 2.1 6.1 2.1 2.7 1.6 2.5 3.5 4.7 2.1 4.2 3.0 3.8 5.3 4.8 5.4 2.2
2.6 1.6 1.3 5.3 3.9 2.4 6.2 2.8 2.2 5.7 4.2 4.1 3.5 2.5 2.9 7.8 5.2 4.8 9.3 7.8 9.0 7.7 6.3 6.2 4.2 3.2 4.7
1.4 0.9 0.6 2.6 1.7 1.3 3.0 2.5 2.2 2.5 1.2 1.1 5.3 4.2 3.2 7.3 4.1 3.2 4.7 4.2 2.7 11.2 7.6 6.2 14.2 11.2 10.2
5.0 7.6 7.8 8.6 10.0 4.5 10.5 4.8 6.0 14.4 12.4 5.4 14.9 4.7 7.7 4.8 7.7 11.3 26.9 9.8 14.3 8.5 12.9 18.0 16.8 19.3 7.3
Metsorb P
As
Va
P
2.7 1.7 1.6 7.0 4.4 3.1 8.8 2.5 2.2 8.4 4.9 3.4 3.9 2.3 3.4 10.2 5.2 4.7 19.1 12.2 10.2 12.8 8.1 7.0 6.0 3.2 5.2
0.6 0.3 0.2 0.9 0.5 0.3 1.3 0.6 0.5 1.0 0.7 0.3 2.4 1.4 1.2 2.6 1.3 1.2 1.5 1.5 1.2 3.8 2.9 2.4 5.1 4.3 3.4
2.6 4.1 4.4 5.8 6.1 2.5 5.3 2.9 3.5 7.9 7.6 3.4 7.7 2.3 4.3 2.4 4.9 7.3 13.9 5.2 6.7 4.1 5.8 10.6 6.4 6.0 2.4
1.1 0.7 0.7 3.7 2.0 1.2 1.8 1.3 0.8 3.4 2.2 1.2 2.0 0.9 1.2 3.9 2.4 1.8 6.0 4.0 4.6 4.8 2.9 2.5 2.6 1.8 1.5
a Actual value is greater than shown here because full breakthrough did not occur.
Bed volumes treated
Fig. 2 e Cumulative silica removal at silica influent concentration 33 mg/L and different pH values: (a) 8.3, (b) 7.65, and (c) 7.0.
experiments by Jordan et al. (2007), in which the adsorption of silicate onto goethite, hematite and magnetite increased when pH increased from 5 to 10. Silica saturated the media in the columns and full breakthrough occurred after w3000 bed volumes for GFH and w2000 bed volumes for Metsorb and E33. GFH was the most favorable media for silica adsorption and that adsorption was more sensitive to pH than was adsorption on either E33 or Metsorb. Badruzzaman and Westerhoff (2005) reported that GFH adsorbed 2e3 times more silica than E33 in RSSCTs. The concentration of other ions (arsenate, phosphate, and vanadate) had a negligible effect on silica adsorption as would be expected, based on their relative concentrations. Silica concentrations were three orders of magnitude higher than the other anions. Set 1, 4 and 7 (Fig. 2a) have the same pH (8.3) and silica concentration (33 mg/L) but different arsenate, vanadate and phosphate concentrations but all are grouped together in terms of silica removal performance. The same was found for silica removal at pH of 7.65 and 7.0 (Fig. 2b and c) and at silica concentrations of 53 and 73 mg/L (Nguyen, 2011). When the adsorbents in the columns were saturated with
silica after a few thousand bed volumes, only limited amounts of other anions had been adsorbed, and these quantities were minimal compared to the amount of silica adsorbed. Except for the first thousand bed volumes, the anions were adsorbed on a silica pre-loaded adsorbent instead of the virgin media. Westerhoff et al. (2006) found that the iso-electric point of E33 and GFH in equilibrium with 63 mg/L silica was 2 and 4 units lower than that of the virgin media. Because the surface charge of the adsorbents affect the electrostatic interactions between the anions and the surface of the adsorbents (Amy et al., 2005), a decrease can lead to a reduction in anion adsorption on the surface. In continuous flow columns, competing anions are adsorbed onto a silica-saturated adsorbent while in batch tests arsenic, silica and other anions are adsorbed simultaneously. Therefore, using batch isotherm data for arsenic in the presence of silica to predict the performance of columns of adsorption media can be misleading, because the availability of surface sites and the energetic favorability of adsorption in the two modes are different. Moreover, although Vaughan et al. (2007) found that, in the absence of silica, longer EBCTs resulted in improved removal of arsenate (as expected), Kanematsu (2011) found that, in the presence of silica, a longer EBCT did not necessarily result in improved removal of arsenic due to the silica precoating having a relatively larger impact in a longer column than in a shorter column. The impact of EBCT in the
4075
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 6 9 e4 0 8 0
Fig. 3 e BV treated until arsenic breakthrough at 10 mg/L (BV10). E33, according to literature reports, is 340 m2/cm3 and 179 m2/cm3, respectively (Westerhoff et al., 2006) and 139 m2/cm3 for Metsorb (Amy et al., 2005). The high surface area of GFH results in more anion adsorption sites and capacity. The cost of the adsorbents ranges from approximately $150
presence of interferences, and its subsequent impact on the similarity of breakthrough curves from RSSCTs and full scale columns, has not been adequately explored. Overall, in all of the cases, GFH adsorbed more anions compared to E33 and Metsorb. The surface area of GFH and 60,000
a
Mean BV10
50,000
pH
c
GFH E33
Silica
Metsorb
40,000 30,000 20,000 10,000 0 6.5
7
7.5
8
8.5
25
40
55
70
Silica concentration (mg/L)
pH 60,000
Mean BV10
50,000
b
d
Arsenic
Vanadium
40,000 30,000 20,000 10,000 0 0
15
30
45
Arsenic concentration ( g/L)
60
10
25 40 55 Vanadium concentration ( g/L)
70
Fig. 4 e Effect of factors on mean BV at 10: (a) pH, (b) Arsenic, (c) Silica and (d) Vanadium. Error bars depict 95% confidence interval.
4076
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 6 9 e4 0 8 0
to 450/ft3. However, the actual media cost is negotiated on a case-by-case basis for most systems.
3.2. Effect of factors on BV treated until 10 mg/L arsenic breakthrough The current maximum contaminant level for arsenic in potable water in the U.S. is 10 mg/L. To evaluate the effect of changes in water quality on the arsenic treatment efficiency of the adsorbents, the BV10 was interpolated from the breakthrough curve. The BV10 values for all 27 columns for all three media are summarized in Fig. 3. For all the media, the BV10 was highest (33,500, 96,000 and 38,900 for E33, GFH and Metsorb, respectively) in column 19, where the influent level of arsenic, silica and pH were the lowest. Likewise, column 9 with the highest influent level of arsenic, silica and pH had the lowest BV10 values (4000, 5300 and 1400 for E33, GFH and Metsorb, respectively). The mean BV10 was calculated at each level of the five factors. This mean BV10 was plotted versus each factor to determine whether there was any change in mean BV10 when the level of that factor changed (Fig. 4). Not unexpectedly, lower values of pH, silica, arsenic, vanadium resulted in greater values of mean BV10. The significance of each effect was determined using analysis of variance. The factors that had significant main effect and two factor interaction effect on the BV10 of the three media are highlighted in the table of ANOVA results (Table 6). The value of R2 depicts how well the effects listed in the table describe the variation in the BV10; the effects of the factors can explain more than 98% of the variation in the BV10 of the columns (Kutner et al., 2005).
From Table 6, it is apparent that all of the media were significantly impacted by pH, arsenic, phosphate and silica levels. Vanadium only imposed a statistically significant linear main effect on GFH. ANOVA results also showed strong interaction effects between pH and silica, pH and arsenic, and arsenic and silica levels on the BV10 of GFH. Only the interaction effect between pH and arsenic levels was detected in E33. In Metsorb, there were interaction effects between pH and arsenic, and arsenic and silica as well as pH and vanadium, and vanadium and phosphate. The last two interactions in Metsorb involved vanadium, which had an insignificant individual effect on the media. The interaction involving vanadium detected here may be the result of the single effect of the other factor and is omitted from further discussion. The important interactions are illustrated in Figs. 5, 6 and 7. The effects of each factor on each media are examined in detail in the next section.
3.2.1.
pH
As illustrated in Fig. 4a, the sensitivity of the media to a pH change followed the order of GFH > Metsorb > E33. When pH was reduced from 8.3 to 7.0, the mean BV10 of GFH increased by 40,000. GFH was more sensitive to a change in pH in the lower range (7e7.65) than in the higher range (7.65e8.3). In E33 and Metsorb, lowering the pH over the range of this study increased the mean BV by 12,000 and 17,900, respectively. Across all pH levels, the mean BV10 was highest for GFH and lowest for Metsorb. At pH 8.3, the performance of GFH and E33 were comparable while at pH 7 the performance of Metsorb and E33 were comparable. These data provide greater insight into previous findings in the literature. Badruzzaman and Westerhoff (2005) reported that while in general E33 and
Table 6 e ANOVA for BV10 of each media.a GFH (R2 ¼ 0.994)
Main effect pH (L) pH (Q) Arsenic (L) Arsenic (Q) Silica (L) Silica (Q) Vanadium (L) Vanadium (Q) Phosphorus (L) Phosphorus (Q) Interaction pH * Arsenic pH * Silica pH * Vanadium pH * Phosphate Arsenic * Silica Arsenic * Vanadium Silica * Vanadium Silica * Phosphate Vanadium * Phosphate
E33 (R2 ¼ 0.984)
Metsorb (R2 ¼ 0.997)
Effect
Std.Err.
t
p
Effect
Std.Err.
t
p
Effect
Std.Err.
t
p
40067 10222 26306 5408 18689 2522 4700 1772 628 6181
1666 1443 2762 1815 1666 1443 1666 1443 2762 1815
24.1 7.1 9.5 3.0 11.2 1.7 2.8 1.2 0.2 3.4
0.000 0.000 0.000 0.021 0.000 0.124 0.026 0.259 0.827 0.011
11967 772 15472 4625 6756 1789 1267 878 6761 2747
1016 880 1684 1107 1016 880 1016 880 1684 1107
11.8 0.9 9.2 4.2 6.7 2.0 1.2 1.0 4.0 2.5
0.000 0.409 0.000 0.004 0.000 0.081 0.252 0.352 0.005 0.042
17900 3983 12167 3017 7900 117 633 50 1300 3083
548 474 908 597 548 474 548 474 908 597
32.7 8.4 13.4 5.1 14.4 0.2 1.2 0.1 1.4 5.2
0.000 0.000 0.000 0.001 0.000 0.813 0.285 0.919 0.195 0.001
20833 11834 3557 122 10591 5174 811 3898 1237
4407 2246 2246 4407 2511 2511 2885 2511 2511
4.7 5.3 1.6 0.0 4.2 2.1 0.3 1.6 0.5
0.002 0.001 0.157 0.979 0.004 0.078 0.787 0.165 0.637
7633 820 598 2211 2548 415 2678 3193 237
2687 1370 1370 2687 1531 1531 1759 1531 1531
2.8 0.6 0.4 0.8 1.7 0.3 1.5 2.1 0.2
0.025 0.568 0.676 0.438 0.140 0.794 0.172 0.075 0.881
9600 1609 2142 0 5413 176 533 1357 2654
1449 738 738 1449 826 826 949 826 826
6.6 2.2 2.9 0.0 6.6 0.2 0.6 1.6 3.2
0.000 0.066 0.023 1.000 0.000 0.837 0.591 0.144 0.015
Note: (L): linear effect; (Q): quadratic effect. a Bold values indicate significant effects ( p < 0.05).
4077
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 6 9 e4 0 8 0
40,000
a
Mean BV10
30,000
E33
Arsenic
g/L) 15 35 55
20,000
10,000
80,000
Silica (mg/L) 33
60,000 Mean BV10
GFH performance was comparable, E33 out-performed GFH in some cases, due to the specific water quality, and that RSSCTs should be conducted for each source of water. We hypothesize that at a certain pH level, the performance of E33 and GFH are comparable; however, because GFH is more sensitive to pH changes than E33, a higher pH level may result in E33 outperforming GFH, while a lower pH level may lead to better performance by GFH. As shown in Fig. 4a, Metsorb was also more sensitive to pH changes than E33, although its performance was always poorer for our experimental conditions. The change in mean BV10 in GFH and Metsorb was more profound when pH changed from 7.65 to 7.0 than from 8.3 to 7.65 while the change of mean BV10 of E33 was the same across the pH levels. These patterns are consolidated by the ANOVA results in Table 6, which indicate that pH imposed both a linear and quadratic effect on GFH and Metsorb but only a linear effect on E33. The analysis of variance (Table 6) also indicated that the mean BV10 of all media was affected by the interaction of pH and arsenic concentration, meaning that the effect of pH was dependent on the arsenic levels. For example, for GFH (Fig. 5b) and Metsorb (Fig. 5c), the changes in BV10 when pH varied
53 73
40,000 20,000
0 6.8
7.2
7.6
8
8.4
pH
Fig. 6 e Interaction effects of pH and silica on GFH. Error bars depict 95% confidence interval.
from 8.3 to 7.0 were more profound at an arsenic level of 15 mg/L than at 35 or 55 mg/L. When the arsenic level was 35 mg/L or higher, changing the pH from 8.3 to 7.65 did not significantly improve the BV10 for GFH columns. The decrease in BV10 when arsenic levels increased from 15 to 55 mg/L in all three media was dependent on pH level. Although low pH is the favorable condition for arsenic removal, the BV10 at a low pH was also more sensitive to a change in the influent arsenic concentration than at a high pH. For instance, in GFH (Fig. 5b) and Metsorb (Fig. 5c), when the arsenic concentration increased from 15 to 35 and then to 55 mg/L, the mean BV10 decreased more at pH 7 than at pH 7.65 and 8.3. In fact, no significant decrease in BV10 was detected when arsenic concentration increased from 35 to 55 mg/L at pH 8.3. For E33 (Fig. 5a), at pH 7.0 and 8.3, the mean BV10 decreased more
0
100,000
80,000
b GFH
a
60,000 60,000
Mean BV10
Mean BV 10
80,000
40,000 20,000
Arsenic
GFH
g/L) 15 35 55
40,000 20,000
0
0
40,000
30,000
c
20,000
10,000
0 6.8
7.2
b
Metsorb Mean BV10
Mean BV10
30,000
7.6
8
8.4
pH
Fig. 5 e Interaction effects of pH and arsenic on: (a) E33, (b) GFH and (c) Metsorb. Error bars depict 95% confidence interval.
Metsorb
20,000
10,000
0 30
40
50
60
70
80
Silica concentration (mg/L)
Fig. 7 e Interaction effects of silica and arsenic on: (a) GFH and (b)Metsorb. Error bars depict 95% confidence interval.
4078
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 6 9 e4 0 8 0
when arsenic level changed from 15 to 35 mg/L and did not significantly change when arsenic concentration increased from 35 to 55 mg/L. Only GFH was affected by the interaction of pH and silica as shown in Table 6. The change in mean BV10 for GFH when pH changed at different silica levels is shown in Fig. 6. The slope of the three lines indicates that the sensitivity of BV10 to changes in pH varied with silica level. The BV10 increased linearly when pH decreased from 8.3 to 7.65 and then to 7.0 at a silica concentration of 33 mg/L. At a silica concentration of 53 mg/L, lowering pH from 8.3 to 7.65 did not significantly increase the BV10 whereas lowering the pH to 7.0 did. At the highest silica concentration (73 mg/L), the BV10 did not change significantly when pH changed. This finding implies that lowering the pH in a GFH column will be more effective at lower levels of silica.
3.2.2.
Arsenic
In terms of sensitivity to arsenic concentration in the influent, GFH was again the most sensitive in term of absolute BV10 changes among the three adsorbents as illustrated in Fig. 4b and Table 6. The changes of BV10 for all three adsorbents with variable arsenic influent concentration followed a second order trend, with a greater slope at lower arsenic concentrations. The interaction effect of silica and arsenic were detected in both GFH and Metsorb. The changes in mean BV10 when silica levels varied at different arsenic concentrations are shown in Fig. 7a and b, for GFH and Metsorb, respectively. The change in BV10 was more significant at lower arsenic and silica levels than at higher levels. In neither GFH nor Metsorb was there a significant difference in the BV10 when arsenic concentration increased from 35 to 55 mg/L at silica concentration of 53 and 73 mg/L. However at lower arsenic levels (15 mg/L), decreasing the silica concentration from 73 to 53 and then to 33 mg/L had a statistically significant beneficical impact on the BV10.This can be explained by the presence of different types of adsorptive sites on the surface of the media. At low concentrations of competitor and pH, sites with high adsorption energy were more available and were occupied first by competitors or were deprotonated, resulting in a disproportionately large impact on arsenate adsorption at low arsenic concentrations. Later sorbing species only have less favorable lower energy sites available to them and competition for these sites is not as significant. Therefore, the adsorption of arsenate did not vary much at higher competitor concentrations compared to at lower concentrations.
3.2.3.
Fig. 4c, the decrease in BV10 in Metsorb was linear across the silica range of study. In GFH and E33, the BV10 decreased more when silica concentration increased from 33 to 53 than from 53 to 73 mg/L. Although there were differences in the slopes at these two silica concentration ranges, the quadratic effect of silica was not statistically important (Table 6). The mean BV10 was significantly different at the three levels of silica in the case of Metsorb and GFH. However for E33, there was no difference between the mean BV10 at a silica concentration of 55 and 73 mg/L, meaning that increasing silica concentration levels above 53 mg/L imposed insignificant effect on the performance of the E33 columns.
3.2.4.
Vanadium
The effect of vanadium on E33 and Metsorb was insignificant, as shown in Table 6 and Fig. 4. Although Table 6 indicated that vanadium imposed a statistically significant linear effect on GFH, the error bars of the mean BV10 at different vanadium levels overlapped (Fig. 4d), indicating that this effect was not as strong as the effect of other system variables. In fact, the p value for the linear effect of vanadium was 0.026, much higher than for pH, silica and arsenic. In batch tests for dual solute competition, vanadium was reported as an important competitor to arsenic adsorption onto E33 and Metsorb (Amy et al., 2005) and iron oxides (Youngran et al., 2007). Speitel et al. (2010) also reported a 50% reduction in BV treated when 70 mg/L vanadium was added to buffered water containing only arsenic. The lack of a strong interference effect from vanadium in the current study is likely due to the overshadowing effect of silica. Lakshmanan et al. (2006) reported a similar trend in batch mode, where no significant changes in arsenic adsorption capacity occurred when 40 mg/L of vanadium was added to a water matrix containing 40 mg/L (as P) of phosphate and 20 mg/L of silica.
3.2.5.
Phosphate
The effect of phosphate concentration on the BV10 was calculated from the results of the original fractional factorial experimental design, but as reported in methods, the experimental design resulted in a strong correlation between phosphorous concentration and the combination of pH, silica and arsenic concentration, all of which have strong effect on the performance of the media. Therefore, the individual
Silica
The effects of silica concentration on the mean BV10 for all adsorbents followed linear trends, as shown in Fig. 4c. Again GFH was the most sensitive to changes in silica concentration, followed by Metsorb and then E33 (Table 6). When silica concentration increased from 33 to 73 mg/L, the BV10 of GFH, E33 and Metsorb decreased by 18,700, 6800 and 7900, respectively. This result can be explained by the silica adsorption capacity of the adsorbents. Because E33 had the lowest capacity for silica, and this capacity did not vary much with pH and silica influent concentration, the change in silica concentration had the least impact on the arsenic breakthrough compared to that of GFH and Metsorb. As shown in
Fig. 8 e Mean BV10 at different phosphate concentrations (silica [ 53 mg/L, arsenic [ 35 mg/L and vanadium [ 41 mg/L).
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 6 9 e4 0 8 0
effect of phosphate cannot be separated using the standard statistical method. Instead, additional experiments were conducted (as reported in methods) to directly investigate the effect of phosphate on each of the three media. In these experiments, only pH and phosphate levels were varied, whereas the concentration of arsenic, silica and vanadium were held constant at the mid-level used in the original design (35 mg/L, 53 mg/L and 41 mg/L, respectively). As illustrated in Fig. 8, when phosphate concentration increased from 55 mg/L to 155 mg/L, the mean BV10 for GFH, E33 and Metsorb decreased by only 4100, 2500 and 2700, respectively. GFH was more sensitive to a phosphate concentration change than E33 and Metsorb. Compared with the main effect of pH, arsenic and silica in Table 6, the effect of phosphate shown here is small, and, considering the error bars, the effect of phosphate on E33 and Metsorb was insignificant. As with vanadium, silica overshadowed the effect of phosphate.
3.2.6.
4.
Comparing the significance of factors
As reported in Table 6, for all three media, the effect of pH was the most important among the factors, followed by arsenic concentration, then silica concentration. In GFH, the calculated value for the interaction of pH and arsenic was higher than that of pH and silica or arsenic and silica, therefore the interaction of pH and arsenic concentration is the most important among the interactions. This interaction is also the most important in Metsorb. Overall, pH was the factor that imposed the most significant effect on the performance of GFH, Metsorb and E33 columns. Fortunately, this is also the intervention most amendable to operational control. These results offer insights to small systems regarding the design and operation of arsenic adsorption system using the three media. For example, pH adjustment to a lower level can significantly increase the bed life of the GFH column compared to E33 or Metsorb columns. However, if failure in pH adjustment occurs, the performance of the GFH column may be reduced more in terms of absolute number of bed volumes treated compared to E33 or Metsorb columns.
4.
3.
5.
6.
7.
4079
adsorption system, that cost may be off set by improved bed life, and in all cases, is worth consideration. GFH and Metsorb were more sensitive to changes in arsenic concentration at low pH than at high pH. Overall, the performance of E33 (as BV10) changed the least when the level of the factors in the water changed. The performance of GFH was the most susceptible to water quality variation. At pH 8.3, the performance of GFH and E33 were comparable while at pH 7 the performance of Metsorb and E33 were comparable. At a certain pH level, the performance of E33 and GFH are comparable; however, because GFH is more sensitive to pH changes than E33, a higher pH level may result in E33 out-performing GFH, while a lower pH level may lead to better performance by GFH. Silica was the first anion to fully breakthough from all columns of the three media. Other anions were likely adsorbed to a silica-loaded adsorbent instead of simultaneously with silica as in batch tests. The solid phase loading of anions on GFH was the highest among the media. Metsorb and GFH adsorbed much more vanadium than E33. The order of anions to fully breakthrough from GFH and Metsorb columns was silica > phosphate > arsenate > vanadate. In E33, except for silica with an early breakthrough, the breakthrough order of phosphate, vanadate and arsenate did not have a specific trend and depended on water quality. Lowering pH from 8.3 to 7.0 in a GFH column was more effective at lower silica levels. When silica concentration was high (w73 mg/L) the performance of the column was not significantly improved by adjusting the pH to a lower level. Conventional adsorption may be an inappropriate treatment method for waters at the high end of silica concentrations in ground water. Although phosphate and vanadium were previously found to impose detrimental effect on arsenic adsorption in bisolute batch experiments, in the column mode with cocontaminants their effect on the performance of the media was insignificant and much smaller than the effect of pH, arsenic or silica. Phosphate and vanadium may well be significant competitors at higher concentrations.
Conclusions
These conclusions are valid over the range of parameters tested, which encompasses the range of values found in California groundwaters currently used as drinking water sources. The results support the importance of both multiconstituent studies as well as use of column testing. Principal finding were: 1. Given the varying impact of water quality on different media, RSSCTs or pilot testing need to be conducted for each water source prior to selection of media and prediction of operational and maintenance costs. 2. The value of pH was found to be the factor that had the most impact on the BV10 followed by the effect of arsenic and silica. Lowering pH from 8.3 to 7.0 resulted in an increase of the mean BV10 by 40, 12 and 18 thousands BV treated by GFH, E33 and Metsorb columns, respectively. Although pH control adds cost and complexity to an
Acknowledgments This research was supported under Contract No. 06-55254 from the California Department of Public Health Safe Drinking Water Revolving Fund. The content is solely the responsibility of the authors and does not necessarily represent the official views of the organizations above.
references
Amy, G., Chen, H.-W., Dinzo, A., Gunten, U., Brandhuber, P., Hund, R., Chowdhury, Z., Kommeni, S., Sinha, S., Jekel, M., Banerjee, K., 2005. Adsorbent Treatment Technologies for Arsenic Removal. AWWA Research Foundation, Denver, CO.
4080
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 6 9 e4 0 8 0
Badruzzaman, M., Westerhoff, P., 2005. The Application of Rapid Small Scale Column Tests in Iron-based Packed Bed Arsenic Treatment Systems. ASC symposium series 915: Advances in arsenic research: Integration of experimental and observational studies and implications for mitigation, 268e283. American Society of Chemistry, New York. Blackmore, D.P.T., Ellis, J., Riley, P.J., 1996. Treatment of a vanadium-containing effluent by adsorption/coprecipitation with iron oxyhydroxide. Water Research 30 (10), 2512e2516. CDPH (California Department of Public Health), 2008. Water Quality Database. http://www.cdph.ca.gov/certlic/ drinkingwater/Documents/Lawbook/DWRegBook2008_03_09a. pdf, [accessed 08 04 2009]. Crittenden, J.C., Reddy, P.S., Arora, H., Trynoski, J., Hand, D.W., Perran, D.L., Summers, S., 1991. Predicting GAC performance with rapid small-scale column test. Journal of American Water Works Association 83, 77e87. Crittenden, J.C., Berrigan, J.K., Hand, D.W., Lyskin, B., 1987. Design of rapid small-scale adsorption tests for a nonconstant diffusivity. Journal of Environmental Engineering 113 (2), 243e259. Crittenden, J.C., Berrigan, J.K., Hand, D.W., 1986. Design of rapid small-scale adsorption tests for a constant diffusivity. Journal Water Pollution Control Federation 58 (4). Cullen, W.R., Reimer, K.J., 1989. Arsenic speciation in the environment. Chemical Review 89 (4), 713e764. Hongshao, Z., Stanforth, R., 2001. Competitive adsorption of phosphate and arsenate on goethite. Environmental Science and Technology 35 (24), 4753e4757. Jordan, N., Marmier, N., Lomenech, C., Giffaut, E., Ehrhardt, J.-J., 2007. Sorption of silicates on goethite, hematite, and magnetite: experiments and modelling. Journal of Colloid and Interface Science 312 (2), 224e229. Kanematsu, M., 2011. Influence of Water Quality and Empty Bed Contact Time on Arsenate Removal by a Fixed-bed Iron Oxide Adsorber. Doctoral dissertation. University of California at Davis. Kutner, M.H., Nachtsheim, C.J., Neter, J., Li, W., 2005. Applied Linear Statistical Models, fifth ed.. McGraw-Hill/Irwin, New York. 1e1396. Lakshmanan, D., Clifford, D., Samantha, G., 2006. Effect of PH, Insitu Hydroxide Formation and Competing Ions on as (III/V) Adsorption during Coagulation with Ferric Chloride (AWWA Annual Conference, San Antonio, Texas). Luengo, C., Brigante, M., Avena, M., 2007. Adsorption kinetics of phosphate and arsenate on goethite. A comparative study. Journal of Colloid and Interface Science 311 (2), 354e360. Mak, M.S.H., Rao, P.H., Lo, I.M.C., 2009. Effects of hardness and alkalinity on the removal of arsenic(V) from humic aciddeficient and humic acid-rich groundwater by zero-valent iron. Water Research 43, 4296e4304. Meng, X., Bang, S., Korfiatism, G.P., 2000. Effects of silicate, sulfate, and carbonate on arsenic removal by ferric chloride. Water Research 34 (4), 1255e1261. Meng, X.G., Korfiatis, G.P., Bang, S.B., Bang, K.W., 2002. Combined effects of anions on arsenic removal by iron hydroxides. Toxicology Letters 133, 103e111.
Nguyen, V., 2011. Effect of interferences on the breakthrough of arsenic: rapid small scale columns tests. Dissertation. University of California - Davis, Davis. NRC - National Research Council, 1999. Arsenic in Drinking Water. 83e148. The National Academies Press, Washington, DC. 1-310. Peacock, C.L., Sherman, D.M., 2004. Vanadium(V) adsorption onto goethite (a-FeOOH) at pH 1.5 to 12: a surface complexation model based on ab initio molecular geometries and EXAFS spectroscopy. Geochimica et Cosmochimica Acta 68 (8), 1723e1733. Pokhrel, D., Viraraghavan, T., 2008. Arsenic removal from aqueous solution by iron oxide-coated biomass: common ion effects and thermodynamic analysis. Separation Science and Technology 43, 3545e3562. Speitel, G.E., Katz, L.E., Chen, C.-C., Stokes, S., Westerhoff, P., Shafieian, P., 2010. Surface complexation and dynamic transport modeling of arsenic removal on adsorptive media. Water Research Foundation and Arsenic Water Technology Partnership, Denver, CO. 1e140. US-EPA (Environmental Protection Agency), 2005. Treatment Technologies for Arsenic Removal, Environmental Protection Agency EPA document # EPA/600/S-05/006. Vaughan, R.L., Reed, B.E., Smith, E.H., 2007. Modeling As(V) removal in iron oxide impregnated activated carbon columns. Journal of Environmental Engineering-ASCE 133 (1), 121e124. Westerhoff, P., Benn, T., Chen, A., Wang, L., Cumming, L., 2008. Assessing Arsenic Removal by Metal (Hydr)oxide Adsorptive Media Using Rapid Small Scale Column Tests EPA document # EPA/600/R-08/051. Westerhoff, P., Haan, M.D., Martindale, A., Badruzzaman, M., 2006. Arsenic adsorptive media technology selection strategies. Water Quality Research Journal of Canada 41 (2), 171e184. Westerhoff, P., Highfield, D., Badruzzaman, M., Yoon, Y., 2005. Rapid small scale column tests for arsenate removal in iron oxide packed bed columns. Journal of Environmental Engineering 131 (2), 262e271. Xu, H., 2005. A catalogue of three-level regular fractional factorial designs. Metrika 62 (2e3), 259e281. Youngran, J., Fan, M., Leeuwen, J.V., Belczyk, J.F., 2007. Effect of competing solutes on arsenic(V) adsorption using iron and aluminum oxides. Journal of Environmental Sciences 19 (8), 910e919. Zeng, H., Fisher, B., Giammar, D., 2008a. Individual and competitive adsorption of arsenate and phosphate to a highsurface-area iron oxide-based sorbent. Environmental Science and Technology 42 (1), 147e152. Zeng, H., Arashiro, M., Giammar, D., 2008b. Effects of water chemistry and flow rate on arsenate removal by adsorption to an iron oxide-based sorbent. Water Research 42 (18), 4629e4636. Zhang, G.S., Qu, J.H., Liu, H.J., Liu, R.P., Li, G.T., 2007. Removal mechanism of As(III) by a novel Fe-Mn binary oxide adsorbent: oxidation and sorption. Environmental Science & Technology 41, 4613e4619.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 8 1 e4 0 9 1
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Detection of the human specific Bacteroides genetic marker provides evidence of widespread sewage contamination of stormwater in the urban environment Elizabeth P. Sauer, Jessica L. VandeWalle, Melinda J. Bootsma, Sandra L. McLellan* Great Lakes WATER Institute, UW-Milwaukee, 600 E. Greenfield Ave, Milwaukee, WI 53204, USA
article info
abstract
Article history:
Human sewage contamination of surface waters is a major human health concern. We
Received 21 December 2010
found urban stormwater systems that collect and convey runoff from impervious surfaces
Received in revised form
act as a conduit for sewage originating from breeches in sanitary sewer infrastructure. A
22 April 2011
total of 828 samples at 45 stormwater outfalls were collected over a four-year period and
Accepted 30 April 2011
assessed by culture based methods, PCR, and quantitative PCR (qPCR) to test for traditional
Available online 10 May 2011
and alternative indicators of fecal pollution. All outfalls had the HF183 (human) Bacteroides genetic marker detected in at least one sample, suggesting sewage contamination is nearly
Keywords:
ubiquitous in the urban environment. However, most outfalls were intermittently positive,
Quantitative polymerase
ranging from detection in 11%e100% of the samples. Positive results did not correlate with
chain reaction
seasonality, rainfall amounts, or days since previous rainfall. Approximately two-thirds of
Stormwater
the outfalls had high (>5000 copy number, i.e. CN, per 100 ml) or moderate levels
Fecal indicators
(1000e5000 CN per 100 ml) of the human Bacteroides genetic marker. Escherichia coli (E. coli)
Water quality
and enterococci levels did not correlate to human Bacteroides. A total of 66% of all outfall samples had standard fecal indicator levels above 10,000 CFU per 100 ml. A tiered assessment using this benchmark to identify high priority sites would have failed to flag 35% of the samples that had evidence of sewage contamination. In addition, high fecal indicators would have flagged 33% of samples as priority that had low or no evidence of sewage. Enteric virus levels in one outfall with high levels of the human Bacteroides genetic marker were similar to untreated wastewater, which illustrates stormwater can serve as a pathway for pathogen contamination. The major source of fecal pollution at four of five river sites that receive stormwater discharge appeared to be from sewage sources rather than non-human sources based on the ratios of human Bacteroides to total Bacteroides spp. This study shows the feasibility and benefits of employing molecular methods to test for alternative indicators of fecal pollution to identify sewage sources and potential health risks and for prioritization of remediation efforts. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
Urban stormwater routinely contains high levels of fecal indicator bacteria and is a major contributor to degraded water quality at urban beaches, lakes and rivers (Marsalek and
Rochfort, 2004; Noble et al., 2006; USEPA, 2009). The fecal indicator bacteria found in non-point source runoff, e.g. stormwater, are assumed to be derived from animal sources including domestic pets and wildlife. However, there is growing evidence that stormwater systems can be contaminated with
* Corresponding author. Tel.: þ1 414 382 1710; fax: þ1 414 382 1705. E-mail address:
[email protected] (S.L. McLellan). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.04.049
4082
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 8 1 e4 0 9 1
sewage due to failing infrastructure and illicit cross connections between the stormwater and sewage systems (O’Shea and Field, 1992; Haile et al., 1999; Gaffield et al., 2003; Noble et al., 2003; Salmore et al., 2006; Rajal et al., 2007). High levels of human enteric viruses have also been detected in stormwater runoff from outfalls further confirming the presence of human sources of fecal pollution in stormwater (Haile et al., 1999; Noble et al., 2006; Rajal et al., 2007; Sercu et al., 2009). Because stormwater systems are designed to release collected runoff untreated to surface waters, any sewage contamination poses a risk to human health, especially when the outfalls are located on rivers or near beaches used for recreational purposes. The extent to which the urban environment is impacted by sewage contamination delivered through stormwater systems has not been widely explored (Sercu et al., 2009; Parker et al., 2010). Traditionally, culture methods for Escherichia coli (E. coli) and enterococci have been used for water quality monitoring due to low cost and ease of use. However, these standard indicators are found in both animal and human sources and vary greatly in their potential to carry human pathogens, consequently measuring their levels contributes little to our knowledge of the source of contamination (Boehm et al., 2009). Alternative indicators of fecal pollution that can be detected by PCR and quantitative PCR (qPCR) have proven to be promising for identifying human specific fecal pollution (Ahmed et al., 2007; Stewart et al., 2008; Converse et al., 2009; Sercu et al., 2009; Parker et al., 2010). Testing for traditional fecal indicators augmented with more sensitive and specific approaches would improve our ability to identify and prioritize sources that have a high likelihood of contributing pathogens to surface waters. Such approaches may be useful to investigate unrecognized sewage sources in surface waters, which may be the result of breeches in the sanitary sewer infrastructure. In many cities around the Great Lakes, urban stormwater runoff drains into tributaries leading to nearshore waters that are both a drinking water source and the location of heavily used public beaches. The metropolitan area of Milwaukee in Wisconsin is a typical Great Lakes urban center with three urban rivers that converge and discharge to Lake Michigan in close proximity to several popular beaches. Fecal indicator bacteria such as E. coli have been detected as high as 20,000 colony forming units (CFUs) per 100 ml in metropolitan Milwaukee rivers during wet weather events and have also exceeded water quality standards during dry weather events (Salmore et al., 2006; McLellan et al., 2007). Stormwater in Milwaukee has been identified as a large contributor of fecal pollution in this system and therefore plays a significant role in the number of exceedances of recreational standards and the degradation of water quality (McLellan et al., 2007; SEWRPC, 2008). In this study, we assessed the extent of human fecal contamination in stormwater outfalls in a dense urban area in metropolitan Milwaukee during wet weather flows. We examined samples collected over a four-year period from 45 stormwater outfalls, with roughly half of these fitted with inline samplers that capture both the initial water discharge during the first 60 min (i.e. the first flush) and the subsequent discharge during the remainder of the storm event (i.e. second flush). Traditional culture methods, as well as PCR and qPCR for standard and alternative indicators were used to assess the
human fecal contamination at the outfalls. We demonstrated that the human Bacteroides genetic marker was routinely detected across the study area, a strong indication that sewage contamination is a chronic source of fecal pollution in urban stormwater. This study illustrates that molecular approaches designed to test for alternative indicators of fecal pollution can be used to improve and prioritize remediation projects and provide a higher level of information toward decision making processes aimed at protecting human health.
2.
Materials and methods
2.1.
Study site and sampling methods
The study area included four watersheds within metropolitan Milwaukee with the most intensive sampling carried out in the highly urbanized Kinnickinnic and Menomonee River watersheds (Fig. 1). The Kinnickinnic River watershed encompasses 25 square miles with 53 outfalls along 31 river miles. The Menomonee River watershed and its tributaries encompass 136 square miles with 101 outfalls discharging directly to the river over a 144 mile stretch. This study focused on Milwaukee’s separated sewer area, where stormwater outfall discharges should be urban runoff as opposed to Milwaukee’s combined sewer area where the stormwater and sanitary sewer systems are combined. A total of 45 stormwater outfalls were sampled over a fouryear period in 2006e2009. Outfalls were sampled a minimum of four times during the sampling seasons (April through November). Samples were collected from 23 outfalls using automated inline samplers installed at the last manhole access point before each outfall to ensure the first flush was captured. High flow (rise in pipe flow >0.51 cm) triggered the samplers to collect. Following the first 60 min of sample collection, the stormwater is diverted to a new sample bottle to collect flow from the remainder of the storm (the second flush). Grab samples were also collected from an additional 22 outfalls to increase outfall sampling coverage. Outfall sample locations are designated with the abbreviation for the watershed or subwatershed followed by a numerical assignment. Inline samples are given an “S” prefix. Watershed and subwatershed abbreviations are as follows: Menomonee River (MN), Honey Creek (HC), Underwood Creek (UC), Lincoln Creek (LC), and Lake Michigan (LM). Quantification of the human Bacteroides genetic marker was conducted on a subset of inline and grab samples from 16 of the stormwater outfalls discharging to the Menomonee River watershed including two subwatersheds, Honey Creek and Underwood Creek. Two additional outfalls outside of this watershed with known sewage contamination were sampled for comparison. These outfalls are designated as SLC07 and SLM09 and discharge to Lincoln Creek and directly to Lake Michigan, respectively. Inline samplers were deployed at these two locations. A total of ten wastewater treatment plant influent samples were analyzed for comparison with stormwater. Samples consisted of 24-h flow weighted samples collected at the two major wastewater treatment plants servicing metropolitan Milwaukee.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 8 1 e4 0 9 1
4083
Fig. 1 e Sampling locations in the Milwaukee metropolitan area. Inline (triangle) and grab (circle) samples taken from outfalls and grab samples (arrow) taken from river locations.
Systematic sampling within the stormwater conveyance system was also conducted, with samples taken “up the pipe”, e.g. at upstream manhole access points. The terminal outfall locations (last manhole access point) were designated with an “A”, with upstream location or branch points assigned letters alphabetically from downstream to upstream (e.g. “B” to “E”, depending on the number of points sampled). River water sampling (water quality monitoring) was conducted to assess the level of the human Bacteroides genetic marker following rain events when there was no sewage overflow. River water was collected in 2008 and 2009 at sites adjacent to stormwater outfalls at three locations on the Menomonee River, one location on Underwood Creek and one location on Lincoln Creek on a subset of the outfall sample collection days. River sites were collected with a 1 L grab sampler within 100 m of the corresponding outfall. These outfalls were greater than 36 inches in diameter and deliver large volumes of stormwater to the rivers.
2.2.
Culture-based microbial analysis
All water samples were analyzed using the USEPA method for E. coli and enterococci enumeration (USEPA, 2002, 2006). The samples were filtered through a 0.45 mm pore size 47 mm
nitrocellulose filter and placed on modified m-TEC and MEI agar. The volume of sample filtered was varied according to the expected level of contamination. The plates were incubated for 18 h and CFUs were counted and recorded.
2.3.
DNA extraction
A volume of 200 ml from each water sample was filtered onto a 0.22 mm pore size 47 mm nitrocellulose filter and immediately stored at 80 C, prior to extraction. A volume of 100 ml was filtered for sewage influent samples. The frozen filters were broken into small fragments using a sterile metal spatula. DNA was extracted using the MPBIO FastDNA SPIN Kit for Soil (MP Biomedicals, Santa Anna, CA) according to the manufacturers instructions, with the exception of the lysis step in which a bead beater (Biospec, Bartlesville, OK) was used for 1 min. Extraction efficiencies were determined using enterococci BioBalls (bioMerieux, Marcy-l’Etoile, France). Briefly, 500, 5000 and 50,000 cells were added to 100 ml sterile water and extracted using the above procedure (n ¼ 10 for each concentration). Recovery was 15.3 2.7%. Crude cell extracts were also prepared by lysing cells on filters in 10 mM Tris 0.5 mM EDTA, pH 9.0 using 212e300 mm glass beads (Sigma, St. Louis, MO) with a bead beater. Recovery was on
4084
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 8 1 e4 0 9 1
average 55%, but high levels of inhibition were present and dilution of the samples to 1:10 was necessary to remove inhibition, therefore, we chose to perform DNA extractions on all samples to optimize recovery without inhibition.
2.4.
PCR inhibition assays
Prior to all PCR and qPCR reactions, all sample DNA extractions were initially diluted to a DNA template concentration of 4 ng per ul (20 ng per reaction) based on pilot studies that demonstrated samples with higher DNA concentrations displayed inhibition of PCR. For PCR gel electrophoresis assays, we used E. coli as a control reaction since previous studies in our laboratory has demonstrated that E. coli and human Bacteroides are at similar levels. Samples were tested for E. coli using primers targeting the uidA gene, uidA298F and uidA884R (Table 1). PCR was carried out as described in Bower et al. (Bower et al., 2005). Samples containing 200 CFU per 100 ml E. coli were expected to be positive based on a previous study that examined the limit of detection by PCR of E. coli in sewage influent samples (Bower et al., 2005). In this study, all stormwater samples had E. coli levels above this threshold. If the E. coli control reaction demonstrated a negative result, samples were diluted 1:5, 1:10, 1:20, 1:50, depending upon cell counts, and individual dilutions were retested for E. coli. Samples that required dilution to detect E. coli were also tested at multiple dilutions for the human Bacteroides genetic marker as described below. Samples were considered negative for the human Bacteroides genetic marker when E. coli was detected in the original sample or two samples in the dilution series and the human Bacteroides reaction was negative. For qPCR assays, additional inhibition studies were performed using a method described by Shanks et al (Shanks et al., 2009). In this case an internal amplification control (pIAC) plasmid was constructed (Integrated DNA Technologies, Skokie, Il) containing the human Bacteroides, E. coli and enterococci primer sites with a unique binding site for the IAC probe; UC1 (Zhang et al., 2003; Sivaganesan et al., 2008). In a subset of samples, qPCR was carried out in triplicate for each
of these targets as described below with samples containing 50, 100, and 500 copies of the pIAC. The remainder of samples were only tested for inhibition using the human Bacteroides primers and 50 copies of the pIAC. In almost all cases, adjustment of the DNA template to 4 ng per ul (20 ng per reaction) was adequate to remove inhibition. Less than 1% of the samples required additional dilution to remove inhibition. In all subsequent qPCR assays, samples were diluted and final concentrations were calculated from this initial dilution.
2.5. PCR detection and qPCR quantification of fecal indicator genetic markers After extraction and PCR inhibition determination, PCR was preformed for the human Bacteroides genetic marker using the HF183F and Bac708R primers (Table 1) (Bernhard and Field, 2000) according to previously published methods (Bower et al., 2005). PCR products were visualized under UV light on a 2% agarose gel after staining with ethidium bromide. Samples with weak bands were considered positive. Samples positive for the human Bacteroides genetic marker using gel electrophoresis assays were further analyzed using qPCR (n ¼ 168). qPCR assays were performed using previously published methods for the human Bacteroides genetic marker and total Bacteroides spp., (Bernhard and Field, 2000; Dick and Field, 2004; Kildare et al., 2007), Enterococcus (Behr et al., 2000) and E. coli (Li et al., 2006) (see Table 1 for details). The qPCR reactions were run with 25 ml reaction volumes and consisted of the 1X Taqman Gene Expression Master Mix (Applied Biosystem; Foster City, CA) and primers and probes at a final concentration of 1.0 mM and 80 nM, respectively. DNA template was added at 20 ng per reaction. PCR cycling conditions were as follows: 2 min at 50 C to activate the uracil-Nglycosylase (UNG), 10 min at 95 C to inactivate the UNG and activate the Taq polymerase, 40 cycles of 95 C for 15 s followed by 1 min at 60 C. Reactions were carried out on a StepOne Real Time PCR System (Applied Biosystems, Foster City, CA). Results were reported as copy number (CN) per 100 ml.
Table 1 e Primers and probes used in PCR and qPCR assays. Primer HF183F Bac708R uidA298F uidA884R HF183F BacHum241R BacHum193 (probe) BacsppF BacsppR Bacspp346 (probe) uidA1663F uidA1790R uidA1729 (probe) Entero1F Entero2R Entero1 (probe)
0
Sequence
Target
Method
0
Human Bacteroides Total Bacteroides spp. E. coli E. coli Human Bacteroides Human Bacteroides Human Bacteroides Total Bacteroides spp. Total Bacteroides spp. Total Bacteroides spp. E. coli E. coli E. coli Enterococci Enterococci Enterococci
PCR PCR PCR PCR qPCR qPCR qPCR qPCR qPCR qPCR qPCR qPCR qPCR qPCR qPCR qPCR
5 ATCATGAGTTCACATGTCCG3 50 CAATCGGAGTTCTTCGTG30 50 AATAATCAGGAAGTGATGGAGCA30 50 CGACCAAAGCCAGTAAAGTAGAA30 50 ATCATGAGTTCACATGTCCG30 50 CGTTACCCCGCCTACTATCTAATG30 50 6-FAM-TCCGGTAGACGATGGGGATGCGTT-MGB-NFQ30 50 GCTCAGGATGAACGCTAGCT30 50 CCGTCATCCTTCACGCTACT30 50 6-FAM-CAATATTCCTCACTGCTGCCTCCCGTA- MGB-NFQ30 50 GCGACCTCGCAAGGCATA30 50 GATTCATTGTTTGCCTCCCTGCTGCG 30 50 6-FAM-TGCAGCAGAAAAGCCGCCGACTTCGG-MGB-NFQ30 50 AGAAATTCCAAACGAACTTG30 50 TCAGTGCTCTACCTCCATCATT30 50 6-FAM-TGGTTCTCTCCGAAATAGCTTTAGGGCTA-MGB-NFQ30
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 8 1 e4 0 9 1
The qPCR limit of detection was 15 CN in a reaction, which is equivalent to 112 CN per 100 ml filtered sample based upon the final volume of the sample following DNA extraction and the volume of sample used in PCR. Gel-based assays were slightly less sensitive, with the limit of detection at 500 CN per 100 ml.
2.6.
Viral analysis of SLC07 stormwater discharge
Four liters of stormwater (the largest volume we could obtain from the automated sampler) was collected directly from the outfall during a rain event and shipped overnight on ice for analysis. Analysis for human enteric viruses, including enteroviruses, rotavirus group A, hepatitis A (HAV), G1 noroviruses, GII noroviruses, and adenoviruses (serotypes 1, 2, 5, 6, 40 and 41) was carried out in the laboratory of Dr. Mark Borchardt according to previously published methods (Borchardt et al., 2004; Borchardt et al., 2007; Lambertini et al., 2008).
2.7.
Statistical analysis
All statistical analyses were performed in SPSS v11.0. After statistical tests showed non-normally distributed data, all data were log10 transformed before statistical analysis. The relationship between rainfall, days since previous rainfall, and the human Bacteroides genetic marker was explored using logistic regression. The correlation between the human Bacteroides genetic marker and standard fecal indicators was tested using Pearson’s correlation coefficient. All other data was analyzed using the t-test. All tests were considered significant at p 0.05.
3.
Results
3.1. Detection of human Bacteroides genetic marker in stormwater outfalls in metropolitan Milwaukee The human Bacteroides genetic marker was detected in outfalls in all urban watersheds using a gel-based assay (Fig. 1). During the course of four years, 828 samples from 45 stormwater outfalls were collected during rain events. The stormwater outfalls were intermittently positive for the human Bacteroides genetic marker, which was detected in at least one sample from every outfall tested. Only one outfall was positive in 100% of samples. The frequency of detection for the human Bacteroides genetic marker was categorized as low (0e40%), medium (41e60%), high (61e80%), and very high (81e100%). There was very high or high detection frequency in 20 outfalls. Twenty five outfalls had either medium or low detection levels. Overall, 476 of the 828 samples (57%) contained the human Bacteroides genetic marker. We examined the relationship between rainfall amounts, days since previous rainfall, and detection of human Bacteroides genetic marker using a logistic regression on data from 18 outfalls in the Menomonee River watershed, including Honey Creek and Underwood Creek subswatersheds. There was no significant relationship between the number of outfalls positive for the human Bacteroides genetic marker
4085
across the watershed and rainfall amounts for that day, the number of days since previous rainfall, or a combination of these two parameters. We also did not observe seasonal differences in the percent of outfalls testing positive for the human Bacteroides genetic marker (spring, 67%; summer, 60%; and fall, 70%). We hypothesized that individual outfalls have sewage migrating into the stormwater flows under different conditions, accounting for the lack of a “global correlation” of rainfall parameters and human Bacteroides across the watershed. Therefore, we examined these relationships at two individual outfalls (n ¼ 19 for outfall SUC02A and n ¼ 32 for outfall SMN01A) that had been sampled intensively to determine if patterns could be determined at individual sites. In both these cases, there was no significant relationship between antecedent conditions or rainfall amounts and the detection of the human Bacteroides genetic marker. We could not assess the remaining outfalls for patterns because the human Bacteroides genetic marker was detected in nearly all of the samples (seven outfalls), or only five to six samples had been collected which reduced the power of our analysis (nine outfalls). Overall, these results suggest that complex dynamics are involved and individual sites may be influenced differently by rainfall, preceding conditions, or physical attributes of the infrastructure.
3.2. Quantification of the human Bacteroides genetic marker and total Bacteroides spp. using qPCR For the 18 Menomonee River watershed outfalls and two comparison sites, all samples that were positive for the human Bacteroides genetic marker using the gel-based assay were tested by qPCR. High levels of the human Bacteroides genetic marker were found in outfalls discharging to the Menomonee River, Lincoln Creek, and Lake Michigan (Fig. 2a). The median levels of human Bacteroides genetic marker in these watersheds were found to be at least an order of magnitude higher than outfalls located on Underwood Creek and Honey Creek. Total Bacteroides spp., which is derived from human and non-human sources and includes the human Bacteroides, was used as a measure of “total fecal pollution”. Outfalls located in the Menomonee River, Lincoln Creek, and Lake Michigan watersheds had high total Bacteroides spp., but the proportion of human Bacteroides comprising this total was variable (Fig. 2b). We examined the ratio of the human Bacteroides to total Bacteroides spp. found in untreated sewage (e.g. human sources). The average human Bacteroides genetic marker and total Bacteroides spp. levels were 4.8 107 and 9.8 108 CN per 100 ml, respectively, which corresponded to 5.1% (2.93) of the total Bacteroides spp. being accounted for as human Bacteroides. The average percent human Bacteroides genetic marker detected in individual outfalls was highly variable (Table 2) across the study area. The outfall discharging to Lincoln Creek had the highest percentage of human Bacteroides (1.45%), which suggests human sources are a predominate source of fecal pollution. The Menomonee River, Honey Creek, and Lake Michigan outfalls collectively also had a high percentage of human Bacteroides, with averages of 0.82, 0.79, 0.51%, respectively for outfalls discharging to each of these receiving waters. Outfalls discharging to Underwood Creek had a much
4086
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 8 1 e4 0 9 1
We examined the correlation of infiltration and inflow (I&I) in the sanitary sewer system to the Bacteroides genetic marker results. Rainwater can enter the sanitary system through direct connections such as downspouts (inflow) or through cracks and leaks (infiltration). I&I scores in part reflect the integrity of the sanitary sewer lines and are classified as exceeds or acceptable and range from low to high for each category. Levels of the human Bacteroides genetic marker were significantly higher ( p 0.05) in areas with I&I scores of exceeds (all levels) or acceptable (high) compared with areas with I/I scores of acceptable (medium or low).
3.4. Correlations between the human Bacteroides genetic marker and standard fecal indictors detected by qPCR and culture methods
Fig. 2 e The contribution of Human Bacteroides and total Bacteroides spp. to receiving waters in the Milwaukee Metropolitan area. Box and whisker plots A: Human Bacteroides, B: Total Bacteroides (n [ 168). The percentage of human Bacteroides to total Bacteroides spp. is listed.
lower percentage of human Bacteroides, with an average of only 0.19%.
3.3. Individual outfall human Bacteroides patterns and correlations with infiltration and inflow Levels of the human Bacteroides genetic marker at individual outfalls ranged from an average of 300 CN per 100 ml to >400,000 CN per 100 ml (Table 2). Levels were highly variable, with standard deviations nearing 30e50% of the average for several locations. These results are likely highly influenced by dilution from different amounts of rainwater. Inline samples capture the first flush of the storm, whereas grab samples are collected later in the storm. Inline samples collected at the beginning of the storm had significantly higher levels of total Bacteroides spp. than grab samples. However, overall, inline samples did not have significantly higher levels of the human Bacteroides genetic marker compared to the grab samples.
We examined the relationship between qPCR for the human Bacteroides genetic marker and total Bacteroides spp., and two standard indicators measured by qPCR and culture. The human Bacteroides genetic marker did not correlate with enterococci or E. coli culture results, which are the most commonly used water quality measures (Table 3). Human Bacteroides genetic marker also did not correlate with enterococci qPCR, but did have a very weak correlation to E. coli qPCR. We examined how stormwater outfall samples would be ranked differently based on culture results for E. coli and/or enterococci and human Bacteroides results. Table 4 shows the number of samples distributed among different strata: high, moderate and low human Bacteroides, and high, moderate and low E. coli and/or enterococci. Only 44 samples were found to have high human Bacteroides with high E. coli. A total of 37 samples with a moderate amount of human Bacteroides (100e1000 CN per 100 ml) had E. coli levels that were high. However, 49 samples had low to moderate E. coli but moderate to high levels of the human Bacteroides genetic marker so these samples would not have been flagged as a priority.
3.5.
Human virus detection
A stormwater outfall that chronically tested positive for the human Bacteroides genetic marker, located along Lincoln Creek (SLC07), was investigated for the occurrence of human derived viruses. This outfall also demonstrated high levels of the human Bacteroides genetic marker using qPCR. One sample collected during a rain event was analyzed for enteroviruses, rotavirus group A, hepatitis A (HAV), G1 noroviruses, GII noroviruses, and adenoviruses. The sample was positive for three different viruses: adenovirus at 1.3 103 genomic equivalents per L (ge/L), enterovirus at 1.9 104 ge/L and G1 norovirus at 1.5 103 ge/L. These concentrations of viruses are similar to what is found in sewage influent (M. Borchardt, personal communication) and confirm the presence of human sewage contamination in this stormwater outfall.
3.6.
Up the pipe investigations
Five outfalls discharging to the Menomonee River (Table 5) had high levels of fecal indicator bacteria and human Bacteroides and were thus chosen for up the pipe investigations. Inline samplers were placed at different branches of the
4087
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 8 1 e4 0 9 1
Table 2 e Summary of outfalls sampled in each subwatershed. Average values for culturable E. coli and enterococci are shown. qPCR data is shown as average human Bacteroides genetic marker and the percent of the human Bacteroides genetic marker to total Bacteroides spp. Infiltration and Inflow (I&I) scores are categorized as exceeds or acceptable and range from high to low for each category. Receiving Body of Water
Number Outfall Type of of Samples Sample
Honey Creek
Menomonee River
Underwood Creek Lincoln Creek Lake Michigan
5 6 6 6 6 5 5 6 5 32 6 25 14 4 13 19 24 27
HC06 HC03 HC01 HC02 HC05 HC08 HC07 HC04 SMN15A SMN01A MN06 SMN02A SMN04A MN07 SMN03A SUC02A SLC07 SLM09
Grab Grab Grab Grab Grab Grab Grab Grab Inline Inline Grab Inline Inline Grab Inline Inline Inline Inline
Average Average E. coli Enterococci (CFU/100 ml) (CFU/100 ml) 21,400 21,200 4900 11,700 4900 43,500 1120 5800 3,410,000 787,000 40,100 237,000 222,000 17,500 428,000 144,000 28,200 252,000
stormwater system upstream of the original outfall site to pinpoint the area of contamination. In four of five outfalls, at least one upstream location had approximately two fold higher levels of the human Bacteroides genetic marker suggesting that the source of contamination originated upstream. In one location, the human Bacteroides genetic marker was not detected using gel-based assays in any of the samples, which suggests the point of contamination lies between the outfall and the upstream locations. Further sampling between these segments is necessary to isolate the breech in the sanitary system.
3.7.
Impact on water quality
We sampled five rivers that serve as receiving waters for stormwater discharges from outfalls in this study. There was no significant difference in levels of E. coli and enterococci among sites (Fig. 3). On average, the levels of the human Bacteroides genetic maker showed large differences among
Table 3 e Correlations between culturable fecal indicators and qPCR targets (n [ 168). Significant correlations are flagged.*
Human Bacteroides Total Bacteroides spp. E. coli Enterococci E. coli culturable Enterococci culturable
Human Bacteroides
Total Bacteroides spp.
1 0.552* 0.158* 0.057 0.027 0.105
0.552* 1 0.464* 0.473* 0.336* 0.328*
28,800 26,000 32,600 16,500 19,900 47,600 3380 10,500 609,000 82,000 47,200 151,000 213,000 41,000 105,000 268,000 22,700 169,000
Average Average Human Percentage Bacteroides Human/Total (CN/100 ml) 326 749 3640 1330 4100 1760 804 4790 18,700 993 360,000 1400 408,000 5460 153,000 298 32,500 22,900
0.41 0.35 0.52 2.22 2.03 0.78 0.03 4.42 1.81 0.33 2.73 0.44 3.76 1.13 3.91 0.43 1.44 0.77
I&I
Ranking
Acceptable Exceeds Exceeds Exceeds Acceptable Acceptable Acceptable Acceptable Acceptable Acceptable Acceptable Acceptable Exceeds Acceptable Exceeds Acceptable Exceeds Exceeds
Medium Medium Low Low Medium Medium Medium Medium High Medium High High Medium High Medium Medium Low Low
sites, however, these differences were not statistically significant because within site variation was great. The three Menomonee River sites and the Lincoln Creek site had ratios of human Bacteroides to total Bacteroides spp. in the range of what was found with untreated sewage. Lincoln Creek has five- to ten-fold higher levels of the human Bacteroides genetic marker than Underwood Creek although they have similar levels of fecal indicator bacteria. Across all sites, the levels of the human Bacteroides genetic marker and total Bacteroides spp. in river water were similar or higher than what was found at adjacent outfalls. Among the Menomonee River sites, the middle river site and the adjacent outfall had high levels of the human Bacteroides genetic marker (20,000 and 153,000 CN per 100 ml, respectively) and a high ratio of the human Bacteroides genetic marker to total Bacteroides spp. (>4.0% for both the river and outfall) indicating that the outfall may be a direct source of human contamination. All the river sites, with the exception of Underwood Creek, were found to be heavily impacted by fecal pollution and appear to have sewage as the major source.
4.
Discussion
Human sewage contamination of surface waters is a widespread problem in the urban environment (Lipp et al., 2001; Marsalek and Rochfort, 2004; Salmore et al., 2006; Arnone and Walling, 2007), however, identifying the primary mechanisms that introduce sewage into waterways is elusive. Sewage can enter stormwater systems as a result of breeches in sanitary sewage infrastructure, cross-connections, and abandoned sewer bypass locations, which can be exacerbated by wet weather flows (O’Shea and Field, 1992; Marsalek and Rochfort, 2004). This study attempted to quantify the extent
4088
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 8 1 e4 0 9 1
Table 4 e Stormwater outfall samples stratified by levels of fecal indicator bacteria (E. coli and enterococci) and the human Bacteroides genetic marker detection (n [ 214). High Fecal Indicator Bacteria (>10,000 CFU/100 ml)
Moderate Fecal Indicator Bacteria (1000e10,000 CFU/100 ml)
Low Fecal Indicator Bacteria (<1000 CFU/100 ml)
44
33
4
28
11
2
32
12
2
39
6
1
High Human Bacteroides (>5000 CN/100 ml) Moderate Human Bacteroides (1000e5000 CN/100 ml) Low Human Bacteroides (<1000 CN/100 ml) Negative Human Bacteroides (gel based)
in which stormwater systems act as a conduit for sewage from failing sanitary sewer infrastructure. We sampled nearly one quarter of the major stormwater outfalls in two urbanized watersheds in addition to less dense sampling in other areas of metropolitan Milwaukee. Sewage contamination was intermittently detected at every site examined, attesting to the extent of occurrence for unrecognized sewage fecal pollution sources. Further, there was on average high to moderate levels of human Bacteroides (e.g. >1000 CN per 100 ml) at two-thirds of the outfalls tested. The combination of stormwater outfall and river water assessments for human Bacteroides demonstrates that sanitary sewage sources of fecal pollution are major contributors to poor water quality within our study area. Nearly half of the 18 outfalls in which qPCR was conducted had ratios of human Bacteroides to total Bacteroides spp. consistent with 25% or more sanitary sewage composition based upon our analysis of untreated sewage. River water within all of the subwatersheds had very high levels of human Bacteroides (Fig. 3) and reflected the water quality of the outfalls within the vicinity. These findings support the concept that outfall discharges directly influence receiving waters. It is difficult to estimate actual sewage loading contributions from individual outfalls because there are many variables that influence the concentration of indicator organisms and the volume of water discharge to the
river (drainage area, amount of rain runoff, timing of sampling, etc.), but such a calculation would be critical for assigning sources to fecal pollution loads in this system. River water samples consistently exceeded recreational standards for E. coli and enterococci. E. coli, one fecal coliform, exceeded a variance standard criteria of 1000 CFU per 100 ml fecal coliforms in 97% of samples for the Menomonee River and 100% of samples for Lincoln Creek and Underwood Creek. However, there were greater differences in human Bacteroides levels among river sites, an indicator that likely serves as a better benchmark of potential human health risk. Quantitative risk assessment studies have estimated that levels of >8.6 103 copies per L (860 copies per 100 ml) of human Bacteroides genetic marker may pose a health risk in recreation waters (Soller et al., 2010); Based on previous studies of total culturable viruses in Milwaukee sewage using EPA method for total culturable viruses, (Sedmak et al., 2005), concentrations averaged 2 104 total culturable viruses per L. Therefore, concentrations of 860 copies per 100 L of human Bacteroides corresponds to w0.1e3 total culturable viruses per L during summer months. More recent estimates using qPCR detected adenovirus concentrations ranging from 5 102 to 1 106, depending upon the time of year (M. Borchardt, personal communication); this would correspond to 0.01e20 genomic copies of adenovirues per L for Milwaukee sewage. In other
Table 5 e Human Bacteroides and total Bacteroides spp. for investigations moving upstream from the original discharge point. Original Site with high frequency of Human Bacteroides
Upstream Site
SMN01A
SMN01C SMN01E SMN02B SMN02C SMN03B SMN04B SMN06B SMN06C SMN06D SMN06E
SMN02A SMN03A SMN04A SMN06A
% Positive for Human Bacteroides
Human Bacteroides (CN/100 ml)
Total Bacteroides spp. (CN/100 ml)
0 0 40 0 40 93 n/ab n/ab n/ab n/ab
n/aa n/aa 12,350 n/aa 44,700 51,500 101,000 31,000 543,000 175,000
n/aa n/aa 1,920,000 n/aa 5,050,000 1,150,000 213,000,000 6,080,000 4,800,000 14,800,000
a Not applicable, qPCR is not performed when gel-based assays are negative. b Not applicable, gel-based assays were not preformed on these samples.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 8 1 e4 0 9 1
Fig. 3 e E. coli, enterococci, the human Bacteroides genetic marker, and total Bacteroides spp. levels found in river samples for metropolitan Milwaukee subwatersheds (Lincoln Creek n [ 6, Menomonee River n [ 4, Menomonee River n [ 3, Menomonee River n [ 4, and Underwood Creek n [ 2).
areas, adenovirus has been detected at average concentrations of 3 104 genomic copies per L, with a similar large seasonal variation (Bofill-Mas et al., 2006). All of our river sites exceeded 860 copies per 100 ml, and some sites were 10e25 times greater than this target. Alarmingly, these rivers discharge near Lake Michigan beaches and are used for numerous recreational activities including canoeing and fishing. Other urban areas on marine and freshwater coasts send stormwater discharges directly to public swimming beaches (Haile et al., 1999; Marsalek and Rochfort, 2004; Converse et al., 2009). Ironically, public lands are often the location of stormwater outfall discharge points and simultaneously, the location of recreational beaches. These two uses are in direct conflict with one another. Studies have documented enteroviruses in stormwater (Rajal et al., 2007; Converse et al., 2009) and increased illness associated with swimming near stormwater outfalls. Collectively, these studies highlight the widespread nature of failing urban infrastructure and the potential health risk this problem poses. Numerous studies have been conducted on outfalls with high fecal indicator bacteria with no obvious source of contamination (e.g. direct misconnections or sewage overflows) highlighting the diffuse and problematic nature of fecal pollution in stormwater discharge (Schiff and Kinney, 2001; Sercu et al., 2009; Parker et al., 2010). No correlation has been found between rainfall, magnitude of storm, or the progression of the storm to E. coli and enterococci levels (Parker et al., 2010); however, these parameters have not been previously explored in respect to relationships with alternative, host specific markers that are indicative of a source such
4089
as sanitary sewage. In our study, there was no correlation between rainfall amounts or days since previous rainfall, and detection of the human Bacteroides genetic marker. Seasonality also did not appear to influence the frequency in which sewage was detected in stormwater. Lack of correlation to rainfall and intermittent positive results for the human Bacteroides genetic marker indicate that site specific characteristics may play a large role in stormwater discharge. Other studies have also found significant variability of the human Bacteroides marker within a single outfall (Sercu et al., 2009). Further, studies have suggested specific parameters and the source of contamination (sanitary exfiltration vs. surface runoff) can affect the levels of fecal indicator bacteria in the first flush of stormwater discharge (McCarthy, 2009). We found I&I scores in the sanitary sewer system adjacent to an outfall corresponded to elevated human Bacteroides in the stormwater released from that outfall (Table 2). I&I scores, age of development or other infrastructure characteristics may be useful parameters to consider when attempting to delineate patterns of sewage contamination across a large area, such as a major metropolitan city. The high number of outfalls contaminated with human sewage along with the overall river water quality is a testament to the importance of developing an effective approach to identifying and monitoring these areas of contamination. Others (Noble et al., 2006; Converse et al., 2009; Sercu et al., 2009) have suggested a tiered approach to identifying areas of contamination. Samples are tested through a variety of methods, starting with culture based methods for traditional fecal indicators. Further analyses such as PCR, qPCR, or virus analysis are directed toward samples exceeding water quality standards or those with the highest fecal indicator levels. This may be beneficial and cost effective for areas with a known contamination source (e.g., septic tanks, agricultural runoff). However, in urban areas with significant aging infrastructure and numerous non-point sources of pollution, qPCR may be the best approach as the first tier assessment. Culturable indicators did not correlate with qPCR for the human Bacteroides genetic marker. For example, only moderate levels of E. coli or enterococci (1000 to 10,000 CFU per 100 ml) were found in nearly half the samples with high levels of human Bacteroides genetic marker, and a number of outfalls had high E. coli or enterococci with low or no human Bacteroides (Table 4). This suggests that other sources of E. coli and enterococci, besides human inputs, are in the stormwater system, which is consistent with other reports identifying urban wildlife and pets as fecal sources (Ram et al., 2007). If traditional indicators were used as a metric for identifying and prioritizing outfalls in our study system, then two-thirds of the outfalls with clear evidence of sewage contamination would not have been given a high priority. These findings illustrate the extent in which E. coli and enterococci levels may be uncoupled to evidence of sewage contamination in the urban environment. Water resource managers and regulators will ultimately need to define priorities that target either human sources that likely carry pathogens (regardless of the level) or non-human sources in runoff that contribute high fecal indicators but are not a likely source of human pathogens. Total maximum daily load (TMDL) targets are meant to reduce water quality impairments; however, they are based on standard fecal indicators
4090
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 8 1 e4 0 9 1
and not the actual cause of the impairment, pathogens. Targeting sources for mitigation is a straightforward endeavor in cases where high fecal indicator bacteria correspond to evidence of sewage sources (Sercu et al., 2009). However, our study clearly demonstrates that there is an important decision point on a second tier of assessment; whether to target high fecal indicators, which may include non-human sources, or to prioritize sewage sources. It is important to note that few studies have rigorously tested common non-human sources in urban areas for zoonotic pathogens, with a few exceptions (Schueler and Holland, 1999; Kullas et al., 2002). In contrast, sewage sources have high levels of human pathogens (Sedmak et al., 2003). If the ultimate goal is to reduce pathogens, TMDLs need to consider more precise indicators. Molecular methods offer the opportunity for a much wider variety of organisms to be used as alternative indicators (Field and Samadpour, 2007; Stewart et al., 2008; Boehm et al., 2009). The human genetic marker used in this study has been reported to be highly specific ranging from 83% to 95% (Shanks et al., 2010; Seurinck et al., 2005). Cross reactivity has been reported with 1 of 10 dogs tested in one study (Shanks et al., 2010) and 2 of 8 dogs in another study (Kildare et al., 2007). It is possible that dog waste could account for some of the positive results, but highly unlikely that this source was a major contributor to fecal pollution given the very high E. coli and enterococci levels found in many of the outfalls. Alternative indicators that are specific for fecal waste from animals commonly found in urban areas will be important for confirming the causes of high E. coli or enterococci levels. As these methods become standardized and put into general use, it will be necessary to relate historical measurements to newer approaches. In our study, there is a correlation between culturable enterococci and qPCR enterococci targets (Byappanahalli et al., 2010; Whitman et al., 2010). As host specific qPCR markers are employed for water quality monitoring due to it’s sensitivity and potential specificity toward new targets (Byappanahalli et al., 2010; Lavender and Kinzelman, 2009), research is needed to determine the prevalence of pathogens and viruses in correlation to alternative indicators and direct linkages to human health risks.
5.
Conclusions
Urbanized coastal areas are among our oldest cities and are often challenged with maintaining aging infrastructure. Recognizing and mitigating sources of sewage contamination to surface waters is a high priority. Using PCR and qPCR methods, the sources of contamination can be systematically tracked through the stormwater system and ultimately monitored after remediation. However, a watershed approach with a long-term monitoring program may be the best approach to protect human health. With over 200 outfalls in the Milwaukee metropolitan area, tracking each outfall is time and cost prohibitive. High I&I scores indicate a high probability of sewer system leaks and these areas should therefore be the primary targets of monitoring for sewage entering stormwater systems. Determining “hotspots” of contamination through a watershed approach and then testing suspicious outfalls with traditional engineering approaches such as dye or smoke
testing could be the most effective approach to protecting human health and assessing infrastructure. This approach may be particularly useful in urban areas where numerous non-human sources of fecal pollution cause standard fecal indicators to be of little use for prioritizing remediation efforts.
Acknowledgments This work was funded by a grant through the Milwaukee Metropolitan Sewerage District. We would like to thank Erika Hollis, Andrea Zimmerman, and Morgan Depas for early contributions to this study and Kim Weckerly for GIS based mapping of sites in Fig. 1. We would like to give special thanks to Chris Magruder and Mary Singer, MMSD, for insightful discussions and feedback and Dr. Mark Borchardt, USGS/ARS for providing viral analysis of stormwater.
references
Ahmed, W., Stewart, J., et al., 2007. Sourcing faecal pollution: a combination of library-dependent and library-independent methods to identify human faecal pollution in non-sewered catchments. Water Res. 41 (16), 3771e3779. Arnone, R.D., Walling, J.P., 2007. Waterborne pathogens in urban watersheds. J. Water Health 5 (1), 149e162. Behr, T., Koob, C., et al., 2000. A nested array of rRNA targeted probes for the detection and identification of Enterococci by reverse hybridization. Syst. Appl. Microbiol. 23 (4), 563e572. Bernhard, A.E., Field, K.G., 2000. Identification of nonpoint sources of fecal pollution in coastal waters by using hostspecific 16S ribosomal DNA genetic markers from fecal anaerobes. Appl. Environ. Microbiol. 66 (4), 1587e1594. Boehm, A.B., Ashbolt, N.J., et al., 2009. A sea change ahead for recreational water quality criteria. J. Water Health 7 (1), 9e20. Bofill-Mas, S., Albinana-Gimenez, N., et al., 2006. Quantification and stability of human adenoviruses and polyomavirus JCPyV in wastewater matrices. Appl. Environ. Microbiol. 72 (12), 7894e7896. Borchardt, M.A., Haas, N.L., et al., 2004. Vulnerability of drinkingwater wells in La Crosse, Wisconsin, to enteric-virus contamination from surface water contributions. Appl. Environ. Microbiol. 70 (10), 5937e5946. Borchardt, M.A., Bradbury, K.R., et al., 2007. Human enteric viruses in groundwater from a confined bedrock aquifer. Environ. Sci. Technol. 41 (18), 6606e6612. Bower, P.A., Scopel, C.O., et al., 2005. Detection of genetic markers of fecal indicator bacteria in Lake Michigan and determination of their relationship to Escherichia coli densities using standard microbiological methods. Appl. Environ. Microbiol. 71 (12), 8305e8313. Byappanahalli, M.N., Whitman, R.L., et al., 2010. Linking nonculturable (qPCR) and culturable enterococci densities with hydrometeorological conditions. Sci. Total Environ. 408 (16), 3096e3101. Converse, R.R., Blackwood, A.D., et al., 2009. Rapid QPCR-based assay for fecal Bacteroides spp. as a tool for assessing fecal contamination in recreational waters. Water Res. 43 (19), 4828e4837. Dick, L.K., Field, K.G., 2004. Rapid estimation of numbers of fecal Bacteroidetes by use of a quantitative PCR assay for 16S rRNA genes. Appl. Environ. Microbiol. 70 (9), 5695e5697.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 8 1 e4 0 9 1
Field, K.G., Samadpour, M., 2007. Fecal source tracking, the indicator paradigm, and managing water quality. Water Res. 41 (16), 3517e3538. Gaffield, S.J., Goo, R.L., et al., 2003. Public health effects of inadequately managed stormwater runoff. Am. J. Public Health 93 (9), 1527e1533. Haile, R.W., Witte, J.S., et al., 1999. The health effects of swimming in ocean water contaminated by storm drain runoff. Epidemiology 10 (4), 355e363. Kildare, B.J., Leutenegger, C.M., et al., 2007. 16S rRNA-based assays for quantitative detection of universal, human-, cow-, and dog-specific fecal Bacteroidales: a Bayesian approach. Water Res. 41 (16), 3701e3715. Kullas, H., Coles, M., et al., 2002. Prevalence of Escherichia coli serogroups and human virulence factors in faeces of urban Canada geese (Branta canadensis). Int. J. Environ. Health Res. 12 (2), 153e162. Lambertini, E., Spencer, S.K., et al., 2008. Concentration of enteroviruses, adenoviruses, and noroviruses from drinking water by use of glass wool filters. Appl. Environ. Microbiol. 74 (10), 2990e2996. 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 Res. 43 (19), 4967e4979. Li, J., McLellan, S., et al., 2006. Accumulation and fate of green fluorescent labeled Escherichia coli in laboratory-scale drinking water biofilters. Water Res. 40 (16), 3023e3028. Lipp, E.K., Farrah, S.A., et al., 2001. Assessment and impact of microbial fecal pollution and human enteric pathogens in a coastal community. Mar Pollut. Bull. 42 (4), 286e293. Marsalek, J., Rochfort, Q., 2004. Urban wet-weather flows: sources of fecal contamination impacting on recreational waters and threatening drinking-water sources. J. Toxicol. Environ. Health A 67 (20e22), 1765e1777. McCarthy, D.T., 2009. A traditional first flush assessment of E. coli in urban stormwater runoff. Water Sci. Technol. 60 (11), 2749e2757. McLellan, S.L., Hollis, E.J., et al., 2007. Distribution and fate of Escherichia coli in Lake Michigan following contamination with urban stormwater and combined sewer overflows. J Great Lakes Res 33 (3), 566e580. Noble, R.T., Allen, S.M., et al., 2003. Use of viral pathogens and indicators to differentiate between human and non-human fecal contamination in a microbial source tracking comparison study. J. Water Health 1 (4), 195e207. Noble, R.T., Griffith, J.F., et al., 2006. Multitiered approach using quantitative PCR to track sources of fecal pollution affecting Santa Monica Bay, California. Appl. Environ. Microbiol. 72 (2), 1604e1612. O’Shea, M.L., Field, R., 1992. Detection and disinfection of pathogens in storm-generated flows. Can. J. Microbiol. 38 (4), 267e276. Parker, J.K., McIntyre, D., et al., 2010. Characterizing fecal contamination in stormwater runoff in coastal North Carolina, USA. Water Res. 44 (14), 4186e4194. Rajal, V.B., McSwain, B.S., et al., 2007. Molecular quantitative analysis of human viruses in California stormwater. Water Res. 41 (19), 4287e4298. Ram, J.L., Thompson, B., et al., 2007. Identification of pets and raccoons as sources of bacterial contamination of urban storm sewers using a sequence-based bacterial source tracking method. Water Res. 41 (16), 3605e3614. SEWRPC, 2008. Regional Water Quality Plan Update, Technical Report 39.
4091
Salmore, A.K., Hollis, E.J., et al., 2006. Delineation of a chemical and biological signature for stormwater pollution in an urban river. J. Water Health 4 (2), 247e262. Schiff, K., Kinney, P., 2001. Tracking sources of bacterial contamination in stormwater discharges to Mission Bay, California. Water Environ. Res. 73 (5), 534e542. Schueler, T., Holland, H., 1999. Microbes and urban watersheds: concentrations, sources, & pathways. Watershed Prot. Tech. 3, 554e565. Sedmak, G., Bina, D., et al., 2003. Assessment of an Enterovirus sewage surveillance system by comparison of clinical isolates with sewage isolates from Milwaukee, Wisconsin, collected August 1994 to December 2002. Appl. Environ. Microbiol. 69 (12), 7181e7187. Sedmak, G., Bina, D., et al., 2005. Nine-year study of the occurrence of culturable viruses in source water for two drinking water treatment plants and the influent and effluent of a Wastewater Treatment Plant in Milwaukee, Wisconsin. Appl. Environ. Microbiol. 71 (2), 1042e1050 (August 1994 through July 2003). Sercu, B., Van De Werfhorst, L.C., et al., 2009. Storm drains are sources of human fecal pollution during dry weather in three urban southern California watersheds. Environ. Sci. Technol. 43 (2), 293e298. Seurinck, S., Defoirdt, T., et al., 2005. Detection and quantification of the human-specific HF183 Bacteroides 16S rRNA genetic marker with real-time PCR for assessment of human faecal pollution in freshwater. Environ. Microbiol. 7 (2), 249e259. Shanks, O.C., Kelty, C.A., et al., 2009. Quantitative PCR for genetic markers of human fecal pollution. Appl Environ Microbiol.. Shanks, O.C., White, K., et al., 2010. Performance of PCR-based assays targeting Bacteroidales genetic markers of human fecal pollution in sewage and fecal samples. Environ. Sci. Technol. 44 (16), 6281e6288. Sivaganesan, M., Seifring, S., et al., 2008. A Bayesian method for calculating real-time quantitative PCR calibration curves using absolute plasmid DNA standards. BMC Bioinform. 9, 120. Soller, J.A., Schoen, M.E., et al., 2010. Estimated human health risks from exposure to recreational waters impacted by human and non-human sources of faecal contamination. Water Res. 44 (16), 4674e4691. Stewart, J.R., Gast, R.J., et al., 2008. The coastal environment and human health: microbial indicators, pathogens, sentinels and reservoirs. Environ. Health 7 (Suppl. 2), S3. USEPA, 2002. Method 1603: Escherichia coli (E. coli) in Water by Membrane Filtration Using Modified Membranethermotolerant Escherichia coli Agar (Modified mTEC). US Environmental Protection Agency, Office of Water, Washington, D.C. USEPA, 2006. Method 1600: Enterococci in Water by Membrane Filtration Using Membrane-Enterococus Indoxyl-B-DGlucoside Agar (mEI). U.S. Environmental Protection Agency, Office of Water, Washington, D.C. USEPA, 2009. National Water Quality Inventory: Report to Congress. EPA 841-R-08e001. U.S. Environmental Protection Agency, Office of Water, Washington, D.C. Whitman, R.L., Ge, Z., et al., 2010. Relationship and variation of qPCR and culturable Enterococci estimates in ambient surface waters are predictable. Environ. Sci. Technol. 44 (13), 5049e5054. Zhang, Y., Zhang, D., et al., 2003. A novel real-time quantitative PCR method using attached universal template probe. Nucleic Acids Res. 31 (20), e123.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 9 2 e4 1 0 2
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Virus attenuation by microbial mechanisms during the idle time of a household slow sand filter M.A. Elliott a,*, F.A. DiGiano b, M.D. Sobsey b a
The Water Institute at UNC, Department of Environmental Sciences and Engineering, University of North Carolina at Chapel Hill, Campus Box 7431, Chapel Hill, NC 27599, USA b Department of Environmental Sciences and Engineering, University of North Carolina at Chapel Hill, University of North Carolina at Chapel Hill, Campus Box 7431, Chapel Hill, NC 27599, USA
article info
abstract
Article history:
The biosand filter (BSF) is a household slow sand filter that is operated intermittently such
Received 22 November 2010
that an idle time of typically 18e22 h occurs in between daily charges of water. Virus
Received in revised form
attenuation during the idle time was investigated over repeated daily filtration cycles to
14 April 2011
capture the effect of media aging that encompasses processes occurring throughout the
Accepted 8 May 2011
filter depth rather than restricted to the schmutzdecke at the media surface. A threshold
Available online 17 May 2011
aging period of about one to two weeks was required before virus attenuation began. The observed rates of MS2 and PRD-1 reduction were first-order and reached maxima of 0.061-
Keywords:
and 0.053-log per hr, respectively, over seven-to-ten weeks. Suppression of microbial
Biosand filter (BSF)
activity by sodium azide eliminated virus reduction during the idle time thus indicating
Household drinking
that the operative media aging process was microbially mediated. The mechanism of virus
water treatment
reduction was not modification of media surfaces by physical/chemical or microbial
Point-of-use (POU)
processes. Instead, it appears that the activity of the microbial community within the filter
Waterborne viruses
is responsible. The most likely biological pathways are production of microbial exo-
Slow sand filtration (SSF)
products such as proteolytic enzymes or grazing of bacteria and higher microorganisms on virus particles. Implications of these findings for BSF design and operation and their relevance to other biological filtration technologies are discussed. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
The intermittently operated slow sand filter, popularly known as the “biosand filter” (BSF), is a promising household-scale, point-of-use (POU) technology for removal of microbes from drinking water in developing countries. As of 2007, over 140,000 units have been installed and this is expected to increase annually by about 25,000 (Clasen, 2009). The BSF has been highly successful when measured by user-satisfaction, sustained rates of use, and reductions in diarrheal illness (Liang et al., 2010; Stauber et al., 2009; Tiwari et al., 2009; Aiken et al., in press; Fiore et al., 2010; Duke et al., 2006). Far
higher levels of sustained use have been reported for the BSF than for other POU technologies (Sobsey et al., 2008; Albert et al., 2010). A cost-benefit analysis has demonstrated that the BSF compares favorably to other health interventions, including vaccines and boreholes (Jeuland and Whittington, 2009). Like conventional slow sand filters (SSF) the BSF has no pretreatment or backwashing, filtration is by gravity and the sand bed remains wetted throughout operation. However, there are important differences in design and operation; the most important of these is the feed flow pattern. The BSF is fed intermittently rather than continuously by introducing a single charge of water
* Corresponding author. Tel.: þ1 919 966 7302. E-mail address:
[email protected] (M.A. Elliott). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.05.008
4093
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 9 2 e4 1 0 2
(typically only one per day). Accordingly, the filtration rate decreases as the water level (elevation head) declines during filtration of a single charge. An idle time, during which pores are saturated but the BSF is not actively filtering water, occurs following the filtration of a charge. A more detailed comparison of BSF and conventional SSF design parameters is available (Elliott et al., 2006). The name “biosand filter” implies that a biological mechanism contributes to treatment but direct evidence is lacking. Biological nitrification and denitrification processes occur in the BSF (Murphy et al., 2010). Improved reductions of bacteria in BSFs have been noted with an intact topmost layer (“schmutzdecke”) (Elliott et al., 2006, 2008; Stauber et al., 2006) but this is not sufficient evidence to conclude that the biological community is responsible (it could be physical straining). There is also no direct evidence that a biological community within the filter bed depth contributes to microbial reductions. Research on SSF may yield insight into the role of the biological community in the BSF. Biological activity has been suggested as at least partly responsible for microbial reductions in the SSF (Weber-Shirk and Dick, 1997a, 1998; Eighmy et al., 1994; Haarhoff and Cleasby, 1991; Wheeler et al., 1988; Huisman and Wood, 1974). An intact schmutzdecke contributes to treatment. For instance, there was a sharp decrease in reduction of bacteria upon removal of the schmutzdecke (Hijnen et al., 2004; Dullemont et al., 2006; Unger and Collins, 2008). On the other hand, despite evidence that virus reductions improve with filter maturation, removing the schmutzdecke caused no decrease in virus reductions (Hijnen et al., 2004; DeLoyde, 2007; Poynter and Slade, 1977; Unger and Collins, 2008). Therefore, filter maturation processes other than schmutzdecke development may explain improvements in SSF performance, and this may be particularly the case for virus reductions. The term “media aging” was put forth to encompass physical, chemical and/or biological mechanisms that could occur in SSF throughout the filter depth (Poynter and Slade, 1977; Wheeler et al., 1988). Microbial communities in the form of biofilms or “slime” coatings on sand grains within the filter were identified (Wheeler et al., 1988; McConnell et al., 1984; Lloyd, 1974). A physical-chemical process has also been implicated whereby microbes are removed by sorption onto naturally occurring inorganic chemicals, including aluminum, present in the feed water accumulated on the filter media (WeberShirk and Dick, 1997b; Weber-Shirk and Chan, 2007; WeberShirk, 2002). However, this removal mechanism has only been suggested for bacterial reductions. Viruses are orders of magnitude smaller than these constituents and thus, the removal mechanism may not be the same. Explanations for virus reductions in the SSF provide at least a starting point for mechanistic understanding of the BSF. Media aging was shown to improve virus reductions in SSF (Poynter and Slade, 1977; Wheeler et al., 1988; Windle-Taylor, 1970). The effect of other operating conditions on virus reductions in SSF also indicate a role for microbial activity; for instance, lower temperature operation and draining the filter bed are reported to yield less efficient treatment of viruses. (Hendricks and Bellamy, 1991; Poynter and Slade, 1977). The operating characteristics of the BSF may also produce an important dynamics in virus reduction. In contrast to
conventional SSF, the BSF filtration rate declines to zero as the daily charge is processed after which the filter often remains idle, typically for at least 18 h, before introducing the next charge. Reductions of bacteria and viruses have been shown to be much greater immediately after introduction of the next daily water charge (Elliott et al., 2008). This implies a virus attenuation mechanism within the filter depth. Although viruses are responsible for over 40% of diarrhea hospitalizations in the developing world (Ramani and Kang, 2009), viruses have been far less studied in the BSF than enteric bacteria (Elliott et al., 2008, 2009). The objectives of this research were, therefore, to characterize virus reductions in the BSF by examining: (1) the reduction rate in the BSF filter media during idle time; (2) how reduction rates change over time with media aging; and (3) the contribution of microbial mechanisms to reduction rates and media aging. It is also hoped that these results can yield insight into the mechanisms contributing to virus reductions in this and other biological filtration technologies.
2.
Materials and methods
2.1.
Design of the BSF filter columns
The filter columns used in these experiments were designed based on a full-scale BSF (Fig. 1). Two features of note are an elevated outlet tube and a diffuser plate. The outlet tube is 2to 7 cm above the height of the filter media to allow the media to remain saturated after a charge has been filtered and the filter remains idle until the next charge. The diffuser plate is positioned just above the surface of the sand and is drilled with 2-mm diameter holes. In this way, the feed water charge is evenly spread across the sand surface without disturbing the schmutzdecke. The pore volume of householdscale units typically ranges from 13-to-18 L and the typical daily charge of water to the BSF in household use is 20-to-40 L.
Flow Diffuser
17 cm
Resting Water Level 7 cm
Outlet
Filter Media Bed
76 cm
40 cm
10 cm
Filter Underdrain
Fig. 1 e Cross-section of a plastic household-scale BSF.
4094
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 9 2 e4 1 0 2
The use of bench-scale BSF filter columns, rather than fullscale units, enabled reproducibility testing with up to twelve replicate filters operating in parallel. Transparent polypropylene columns having a diameter of 4.4 cm were filled with well-characterized Accusand silica (Unimin Corp., Le Sueur, MN). Accusand was selected because of its chemical purity, low organic matter content, and low uniformity coefficient (Schroth et al., 1996) that make it an ideal choice for controlled laboratory studies of filtration. The Accusand was pre-washed by 24-h exposure to 40% concentrated HCl, followed by a deionized water rinse to pH 5 (Litton and Olson, 1993) to remove organic matter and electropositive sesquioxide metal coatings. Three sieve fractions (U.S. Standard Mesh 30/40, 40/60 and 50/70) were blended together to provide a relatively narrow range of grain size (d10 ¼ 0.27 mm; d60/d10 ¼ 1.4). The configuration of a bench-scale BSF column is shown in Fig. 2. The daily charge volume (450 mL) maintained hydraulic similitude with full-scale BSFs, both in initial application rate
and fraction of charge stored in pore volume during the idle period. The full-scale design of the diffuser and the elevated outlet tube were replicated in the bench-scale design. The columns were covered with aluminum foil to prevent algal growth. Sampling ports with rubber septa were installed at depths of 10- and 30-cm to determine the change in virus concentration during the idle period. These were used for direct withdrawal of samples from the interstitial space with a polypropylene syringe fitted with a 25-gage beveled needle. The first sample was taken after filtration charge had been completed as measured by observing that the water elevation head had declined to about zero. Subsequent samples were withdrawn periodically throughout the idle time. Three bench-scale column experiments are reported here. Each experiment was conducted at room temperature, which varied from 18 to 22 C. Characteristics of each experiment can be found in Table 1. Column Test A consisted of three columns operating in parallel for a period of up to 8 weeks. Measurements of virus reductions were made at depths of 10- and 30 cm during the idle time. However, they were only taken once the media had been aged for 42e 52 days of daily operation. Column Test A was used to test reduction rates of virus concentrations during idle time in a filter with “aged” media. Column Test B included only a single column. It was used to test whether viruses were reduced in freshly loaded Accusand and how reduction rates changed throughout 93 days of media aging. Column Test C consisted of four columns operating in parallel. In two of these columns, microbial activity was suppressed by the daily addition of 6-mM sodium azide (Section 2.4). Column Test C was used to: (1) validate results obtained in Column Test B, which was undertaken without the benefit of replicate filter columns; and (2) determine whether microbial mechanisms caused the virus reductions observed in “aged” media.
2.2.
Fig. 2 e Cross-section of the bench-scale BSF columns used in this experiment.
Feed water
Feed water from Cane Creek Reservoir and University Lake was obtained from the raw water sample taps of the Orange Water and Sewer Authority (OWASA; Carrboro, NC, USA) water treatment plant. Both sources are protected drinking water reservoirs that do not receive wastewater discharges. Sufficient water was collected before each experiment and refrigerated for the duration of the experiment (53e93 days). Feed water was stored at 4 C. The daily charge was removed and allowed to reach room temperature (approximately 20 C) overnight before use. It was then amended with pasteurized primary effluent (PE) from the OWASA wastewater treatment plant (Chapel Hill, NC, USA) to simulate the presence of wastewater in typical drinking water sources of developing countries and to accelerate the maturation process. Two bacteriophages, MS2 and PRD-1, were selected as the virus challenges. After spiking from the stock virus solutions (see Section 2.4), the resulting feed water concentrations are listed in Table 1. The total organic carbon (TOC) of the source water (Cane Creek or University Lake; Chapel Hill,
4095
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 9 2 e4 1 0 2
Fig. 3 e Reduction in fractional concentration of MS2 and PRD-1 in Column Test A during the idle time (where t [ 0 is the time at which the first sample was drawn from the pores of the media bed). Samples were drawn directly from filter columns at 10-cm or 30-cm depth. Filter columns had been in daily operation for 42e52 days. MS2 and PRD-1 reduction rates in all samples drawn from the filter media pores were significantly greater than those in Control samples. Error bars around Control samples represent one standard deviation.
NC, USA) ranged from 5 to 8 mg/L, depending on the season. Addition of 2.5% PE and challenge microbe spike solutions increased the concentration by about 2.5 mg/L. The natural loss of MS2 and PRD-1 infectivity with time in the feed water will be presented alongside reduction rates in samples taken from the filter during the idle time. The natural loss was measured from aliquots that contained the spike from previous day’s charge and that were stored at room temperature for times corresponding to withdrawals from sampling ports during the idle time. The mean loss in virus infectivity following 24 h of storage of the feed water was less than 20%.
2.3.
Suppression of microbial activity
The microbial suppressant was sodium azide (NaN3) which is known to block the cytochrome system thereby preventing oxidative phosphorylation and subsequent respiration of aerobes and denitrifiers (Weber-Shirk and Dick, 1997a; Forget
and Fredette, 1962). Therefore, it is expected that the activity of all aerobic and denitrifying microorganisms (both bacteria and eukaryotic microorganisms) residing in the filter was inhibited upon addition of sodium azide. However, sodium azide does not affect virus survival. This was confirmed prior to the experiments by exposing MS2 and PRD-1 to sodium azide concentrations up to 50-mM and observing no loss of virus infectivity (data not shown). Unlike strong oxidants such as chlorine, sodium azide suspends microbial activity reversibly and is not expected to affect filter sand surfaces or non-living organic matter such as biofilm exopolysaccharides (Weber-Shirk and Dick, 1997a). The main disadvantage is that sodium azide does not suspend anaerobic activity. However, dissolved oxygen has been shown to be present at 30-cm depth throughout the idle time (Buzunis, 1995) inhibiting methanogens and sulfate-reducers. Fermentative organisms could possibly be present but their growth rate is slow given the small amount of organic carbon in the feed water (Weber-Shirk and Dick, 1997b).
Table 1 e Characteristics and objectives of the virus challenge experiments in bench-scale BSF. Experiment Coding
Columns Pasteurized MS2 log10 PRD-1 log10 pfu/mLc pfu/mLc Backwashed PEb
Length (days)
Source watera
Column Test A
56
Cane Creek
Yes
2.50%
2.8 0.3
3.1 0.7
Column Test B
93
Univ. Lake
No
2.50%
3.0 0.9
4.0 0.9
Column Test C
52
Cane Creek
No
2.50%
3.6 0.6
3.5 0.4
a Cane Creek ¼ Cane Creek Reservoir, Carrboro, NC; Univ. Lake ¼ University Lake, Chapel Hill, NC. b Pasteurized PE ¼ pasteurized primary effluent from OWASA WWTP, Chapel Hill, NC. c Virus concentrations are mean log10 measured concentration per mL and the log10 range.
Major objective of the experiment to determine: the reduction rates of phages MS2 and PRD-1 during the idle time in "aged" filter media how reduction rates change over 93 days of media aging the contribution of microbial mechanisms to reduction rates and media aging
4096
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 9 2 e4 1 0 2
A sodium azide dose of 6-mM (390 mg/L) was introduced into the feed water to two of the four bench-scale columns. This dosage is well above the minimum of 3-mM reported to suppress microbial activity in experiments with a SSF (WeberShirk and Dick, 1997a). Another criterion for dosage selection was that it not suppress growth of the bacterial hosts used for MS2 and PRD-1 enumeration in the single agar layer (SAL) method (USEPA, 2001) as described in Section 2.2. The concentration in SAL plates after dilution of samples withdrawn from the filter ranged from 0.0005 to 0.05 mM which was far less than 0.1 mM found experimentally to cause minor growth suppression of the phage hosts.
2.4.
MS2 and PRD-1 analysis and stock preparation
The single agar layer method was used to assay MS2 and PRD-1 concentrations using Escherichia coli F-amp and Salmonella typhimurium LT2, respectively as the phage host (USEPA, 2001). Stocks of bacteriophages MS2 and PRD-1 were grown on their respective hosts in trypticase soy broth, enumerated by USEPA Method 1602 using a double agar layer assay (USEPA, 2001) and stored at 80 C. Aliquots of each stock were thawed each week, serially diluted ten-fold in phosphate buffered saline and stored at 4 C for up to 7 days. Characteristics of MS2 and PRD-1 are listed in Table 2. MS2 and PRD-1 stocks were diluted and added to feed water daily. Bacterial stocks of E. coli strain B were also added to feed water. All microbial stocks were grown in trypticase soy broth (TSB), separated from culture medium by centrifugation, and subsequently re-suspended in phosphate buffered saline (PBS). Feed water volume consisted of between 1 105 and 1 106 volume dilution of the microbial stocks. Reductions in virus concentration in the BSF were calculated by Equation (1). All log-reduction values reported are log base 10. log Reduction ¼ logðFeed Water ConcentrationÞ logðFiltered Water ConcentrationÞ
2.5.
(1)
Data analysis
A first-order rate of virus reduction during the idle period was tested by plotting the log transformation of the fractional virus concentration (log Ct/Ct¼0) against time, where Ct¼0 is the virus concentration at the beginning of the idle period. The linear regression analysis of these plots was performed with a one-way Analysis of Covariance (ANCOVA) for independent samples. The p-values for unpaired, two-tailed tests with significance level a ¼ 0.05 are reported.
Table 2 e Characteristics of viruses used in these experiments.
3.
Results and discussion
3.1. Direct measurements of MS2 and PRD-1 reduction during idle time Fractional virus concentrations (Ct/Ct¼0) are plotted in logarithmic scale against idle time (0e24 h) in Fig. 4 where Ct is the concentration at time, t, during the idle that follows introduction of the daily charge and Ct¼0 is the concentration at the start of the idle time. These data were obtained on Days 42 and 52 of Column Test A. Also plotted is the fraction of initial virus concentration for feed water samples that were stored as a control for survival under the same time, temperature and feed water conditions. The decline in log Ct/ Ct¼0 was approximately linear in both the idle time samples and the controls, indicating a first-order attenuation rate. A test for homogeneity of regression lines for reduction of either virus measured at the 10- and 30-cm depths revealed no difference ( p > 0.50). Therefore, the rate data at both depths were combined for regression analysis. The first-order rates of reduction for MS2 and PRD-1 within the filter during the idle time were not significantly different ( p ¼ 0.74). They were also much greater than in the control ( p < 0.0001). Given that rates ranged from 0.053 to 0.056-log per hour, idle times as short as 8 h can be beneficial in reduction of infectious viruses. Trendline equations, R2 values, and first-order rates of MS2 and PRD-1 reduction are included in Table 3.
3.2. time
Effect of media aging on virus reductions during idle
Fig. 4 shows that virus reduction occurs during the idle time for media that had been allowed to age for at least 42 days. However, the effect of media aging needed investigation. Previous work in conventional SSF indicates that virus reductions improve after months of operation even after removal of the schmutzdecke (Poynter and Slade, 1977). Media aging could be caused by either a physical/chemical or biological process (Wheeler et al., 1988). PRD-1 reductions during the idle time from Column Test B are shown in Fig. 4. The age of media varied from 1 to 93 days. All of the data for these plots were collected from a depth of 30 cm. PRD-1 concentration did not decrease during idle time for the first seven days of operation based on average reduction rates that were not significantly greater than in the controls ( p ¼ 0.66). The reduction rate, however, began to increase after the media had aged for 16e50 days. The highest rates were observed from 71 to 93 days of filter operation ( p < 0.0001 when compared to controls) which corresponded to the last few weeks of the experiment. Firstorder rates of reduction, trendline equations and R2 values are included in Table 3
3.3.
Suppression of microbial activity
Virus/Phage Size (nm) Isoelectric point Genetic Material MS2 PRD-1
26 62
Source: Collins et al., 2004.
3.5e3.9 4.2
ss-RNA ds-DNA
The rates of reduction for MS2 and PRD-1 during the idle time are presented in Fig. 5 (a, b, c and d) for Column Test C in which the feed water to two of the four columns operating in parallel were amended with sodium azide to suppress
4097
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 9 2 e4 1 0 2
Fig. 4 e Effect of days of operation on reduction in fractional concentration of PRD-1 in Column Test B during the idle time (where t [ 0 is the time at which the first sample was drawn from the pores of the media bed). Samples were drawn directly from a sampling port at 30-cm depth. PRD-1 reduction rates on Days 1 & 7 were no different from those in Control samples; reduction rates were significantly greater than in Control samples on Days 16 & 23 and continued to increase throughout the 93-day experiment.
microbial activity. Also included in these plots is the effect of media aging on virus reductions. Rates of reduction of MS2 and PRD-1 without microbial suppression are shown in Fig. 5a and b, respectively. These rates are first-order and increase with media aging as discussed in previous experiments. The fact that the rates were closely replicated in two parallel columns in this series of experiments provides further confidence in the results. The rates of reduction of MS2 and PRD-1 with microbial suppression are shown in Fig. 5c and d, respectively. Statistical testing showed that virus reductions were significantly greater without (Fig. 6a and b) than with microbial suppression
(Fig. 6c and d) ( p < 0.0001 for MS2 and PRD-1). With microbial suppression, virus reduction rates in the filter media bed are not significantly different from those observed in control samples ( p ¼ 0.14 for MS2 and p ¼ 0.35 for PRD-1) for any amount of media aging. First-order rates of reduction, trendline equations and R2 values for Fig. 6a and b are included in Table 3. The results presented above prove that a biological mechanism is responsible for virus reduction during idle time. The exact mechanism, however, is still uncertain. The possibilities are: (1) increased sorption to biologically modified surfaces, including biofilms; (2) grazing of virus particles by bacteria and
Table 3 e Trendlines, R2 values and first-order reduction rates (with 95% confidence intervals) for data in Figs. 3e6. Not included are samples to which sodium azide was added and control samples, none of which had reduction rates significantly different from zero. Figure
Phage
Day(s)
Trendline Equation
3
MS2 PRD-1 PRD-1 PRD-1 PRD-1 PRD-1 MS2 MS2 MS2 MS2 PRD-1 PRD-1 PRD-1 MS2 PRD-1
41, 51 41, 51 1, 7 16, 23 44, 50 71, 86, 93 4, 8 22 36 41, 52 4, 8 22 36, 41, 52 41 41
0.93 0.84 1.01 0.99 0.98 0.97 0.89 0.94 0.90 0.95 0.96 1.03 0.91 0.94 0.99
4
5a
5b
6
e0.128 t e0.123 t e0.01 t e0.037 t e0.055 t e0.083 t e0.003 t e0.027 t e0.097 t e0.138 t e0.002 t e0.047 t e0.096 t e0.104 t e0.14 t
Ct ¼ Ct¼0
R2
First-order Rate (log10 per hr) with 95% confidence interval
0.91 0.91 0.59 0.96 0.99 0.94 0.03 0.43 0.97 0.99 0.05 0.98 0.92 0.90 0.99
0.056 (0.046e0.063) 0.053 (0.038e0.071) 0.004 (0.001e0.008) 0.016 (0.012e0.023) 0.024 (0.020e0.027) 0.036 (0.029e0.043) 0.001 (0.011e0.008) 0.012 (0.002e0.025) 0.042 (0.034e0.050) 0.060 (0.056e0.064) 0.001 (0.003e0.005) 0.020 (0.018e0.023) 0.042 (0.036e0.048) 0.061 (0.054e0.067) 0.045 (0.031e0.060)
4098
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 9 2 e4 1 0 2
a
b
c
d
Fig. 5 e a, b, c & d e Fractional concentration of viruses remaining at 30-cm depth during idle time in Column Test C (where t [ 0 is the time at which the first sample was drawn from the pores of the media bed). The effect of media aging is shown by idle time data collected on six different days during the course of a 53-day experiment. Fig. 5a (MS2) and 5b (PRD-1) are for normal operation of the BSF (i.e. without sodium azide in the feed) and Fig. 5c and d are with addition of 6-mM sodium azide to suppress microbial activity to the daily feed. MS2 and PRD-1 reduction rates in the filter media pores increased over weeks of daily filter operation under normal conditions; however, reduction rates did not increase and were not significantly different than those in Control samples when sodium azide was added to the daily feed.
Fig. 6 e MS2 and PRD-1 concentration during idle time before (Day 41) and after (Day 50) interruption of microbial activity by addition of sodium azide in Column Test C. Columns were allowed to mature through Day 47 and feed water was amended with sodium azide on Days 48e50. MS2 and PRD-1 reduction rates were significantly greater than zero prior to sodium azide addition; however, reductions were no different from zero following addition of sodium azide.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 9 2 e4 1 0 2
higher microorganisms; and (3) inactivation of virus particles by anti-viral microbial exoproducts (e.g. proteolytic enzymes).
3.4. aging
Suppression of microbial activity following media
An experiment was included at the conclusion of Column Test C to distinguish between increased virus capture produced by a microbially induced modification of media surfaces and by an explicit function of the active microbial community such as grazing or generation of exoproducts that inactivate or sequester viruses. Two bench-scale BSFs were operated for 47 days without feeding sodium azide. Microbial activity was then suppressed by addition of sodium azide for three days (days 48e50 of filter operation). Virus concentration was monitored during idle times following Day 50. If the virus reduction rate was not slowed following introduction of the sodium azide, then the mechanism responsible for virus reduction during idle times would be biological modification of media surfaces that occurred during the first 47 days. On the other hand, if the reduction rate was slowed following the introduction of sodium azide, then the mechanism would be related to explicit functions of an active microbial community. The virus reduction rates for MS2 and PRD-1 shown in Fig. 6 were notably slower during idle times after (Day 50) than before suppression of microbial activity (Day 41) by addition of sodium azide; this was confirmed statistically ( p < 0.001 for both MS2 and PRD-1). Therefore, the microbial mechanism that explains virus reductions is dependent on the presence an active microbial community rather than a microbially induced modification of the media surface as a result of aging. While the exact mechanism is still unclear, a first-order rate of virus reduction during the idle time as observed throughout the bench-scale tests is consistent with findings for virus reduction attributed to either grazers (Kim and Unno, 1996; Pinheiro et al., 2007; Gonzalez and Suttle, 1993; Suttle and Chen, 1992) or microbial exoproducts such as proteolytic enzymes (Walker and Toth, 2000; Northrop, 1964). Grazers (e.g., protozoa and chrysophytes) of bacteria have been found at depth in the SSF (Weber-Shirk and Dick, 1997a; Lloyd, 1973). Predation of viruses, however, has not been investigated directly in the SSF. Nonetheless, grazing by flagellates has been implicated in virus reductions in wastewater (Kim and Unno, 1996) and marine waters (Suttle and Chen, 1992; Gonzalez and Suttle, 1993). Bacteria, including the common biofilm bacterium Pseudomonas aeruginosa, have also been documented to use virus capsids as growth substrates (Lipson and Stotzky, 1985; Cliver and Herrmann, 1972; Herrmann et al., 1974). Viruses could also be ingested by higher microorganisms, which then act as a vector to allow survival in an infectious form and later release to the filtered water. This mode of transmission has been documented for waterborne pathogens, including protozoan parasites (Bichai et al., 2010) and bacteria (Loret et al., 2008). For waterborne pathogenic viruses it has only been reported in nematodes and only for short survival times (Chang et al., 1960); however, some plant viruses are transmitted by microorganism vectors (Rochon et al., 2004). Therefore, this mechanism warrants further investigation.
4099
Microbial exoproducts provide another possible mechanism. Proteases, which are also referred to as proteolytic enzymes, are most often associated with virus inactivation through hydrolysis of peptide bonds in the protein capsids. However, susceptibility could depend on the combination of specific protease and specific viral strain (Nasser et al., 2002; Cliver and Herrmann, 1972; Northrop, 1964). MS2 and PRD-1 have been reported to be less susceptible to some proteases than some enteric viruses (Nasser et al., 2002).
3.5. Implications for BSF design, operation and further investigation Previous laboratory study of BSF (Elliott et al., 2008) revealed MS2 and PRD-1 reductions in filtered water that fall considerably short of the 4-log reduction stated in the USEPA standard for POU devices (Schaub and Gerba, 1991). However, the challenge viruses used in both the previous and present studies most likely provide worst-case estimates for treatment of waterborne viruses. MS2 and PRD-1 are more difficult to treat by granular media filtration than most enteric viruses (Schijven and Hassanizadeh, 2000). The isoelectric points listed in Table 2 show that both viruses are net-negatively charged at the feed water pH, which averaged 6.9 and ranged from 6.6 to 7.3. Because Accusand is essentially silica, the media has a net-negative charge in this pH range. Thus, sorption should be impeded by a net electrostatic repulsion between the media and virus. They also appear less susceptible to reduction by protease activity than enteric viruses such as Hepatitis A and Coxsackie A-9 (Nasser et al., 2002). Moreover, reductions of echovirus type 12 (Elliott et al., 2008, 2009) and PhiX 174 (unpublished data) in BSF are consistently 1e3 orders of magnitude greater than reductions of MS2 and PRD-1. Therefore, many enteric viral pathogens may be sufficiently reduced by the BSF to meet the USEPA standard. However, further research is needed to determine the reductions of common waterborne viral pathogens in BSF filtered water, and the impacts of both idle time and daily volume charged. Temperature, like virus type, is likely to affect reduction rates during idle time. Temperature has been reported to be positively correlated with virus inactivation by predation (Pinheiro et al., 2007), protease activity (Walker and Toth, 2000) and unspecified die-off in groundwater (Yates et al., 1985). Likewise, a positive correlation has been reported in most studies of predation rates by grazers of bacteria (Peters, 1994) and proteolytic enzyme production by bacteria (Rosso and Azam, 1987). While the general trend is toward greater virus reduction at higher temperature, rates of predation and protease production are dependent upon the metabolic rates within a community of microorganisms. Filter operation in a constant temperature environment would strongly benefit future laboratory studies and studying reduction rates across a range of temperatures and the impact of temperature change on these rates may provide avenues for future research. Regardless of the temperature and the virus type, this research has shown that virus reductions in the BSF could be significantly improved by taking greater advantage of microbial processes that occur during the idle time. Virus
4100
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 9 2 e4 1 0 2
reductions during idle time, and the near-plug flow conditions (Elliott et al., 2008), indicate that filtered water quality after resuming operation would be higher for a considerable period. Therefore, enlarging the filter volume or decreasing the maximum charge volume would increase the fraction of water stored during the idle time and subsequently released as product water during the next daily filtration cycle. The optimal design is for the pore volume to be equal to or larger than the maximum charge volume. A factor of safety could be added to account for short-circuiting by either further decreasing of the daily charge volume or by increasing the filter volume. Beyond design implications, operation of the BSF could be improved, even if the total daily charge volume has to be larger than the filter volume to meet household water needs. Additionally, users could be trained to operate the BSF in a way that maximizes the daily idle time and to reserve the water filtered after the idle time up to about one pore volume for human consumption. The pore volumes for most full-scale BSF models range from about 13 L to 18 L, enough to supply a significant portion of daily drinking water needs for most families. Chlorination of feed water and use of chlorinated water should be discouraged if the BSF is used because this prevents development of the microbial community. While, as suggested by Manz (2009), chlorination can reduce the concentration of viruses entering the BSF, many water sources in developing countries are chlorinated inconsistently. Such conditions will not protect consumers from viruses and moreover will undermine media aging (and thus virus reductions) by frequently destroying the microbial community within the filter. Other POU devices should be considered as alternatives to the BSF if chlorination is to be a recommended treatment prior to filtration. Based on the results of this study, only post-filtration chlorination is recommended when treating water using the BSF.
4.
Conclusions
Viruses are attenuated at quantifiable rates during filter idle time within the pores of the BSF, but only after 2-to-3 weeks of media aging. The observed rate of MS2 and PRD-1 inactivation was first-order, increased significantly during the first two-tothree weeks of media aging and continued to increase during the subsequent three-to-seven weeks of operation. The media aging process is mediated by the microbial community within the BSF as evidenced by experiments in which suppression of microbial activity eliminated virus reduction during the idle time. The microbial mechanism was not related to modification of media surfaces by microbes, but rather to the activity of the microbial community within the filter. The improvement in virus reductions by media aging in the BSF in periods of zero pore velocity (i.e., during the idle period) is consistent with similar observations in continuous flow operation of the SSF. The mechanistic explanation revealed for the BSF, therefore, may also be applicable to the SSF. The specific biological pathway could be either by production of microbial exoproducts such as proteolytic enzymes or by grazing of bacteria and higher microorganisms
on virus particles as a source of food. Differentiating their importance, however, would require a well-designed experimental protocol. Proteases can be physically associated with the bacteria that produce them and/or differ in their lifetimes in natural waters (Confer and Logan, 1998; Ward et al., 1986; Rego et al., 1985; Shuval et al., 1971). Therefore, an experimental technique is needed that either removes bacteria and higher microbes while retaining protease activity or suppresses protease activity while allowing for microbial grazing. Protease inhibitors may be a promising tool. The design and operation of the BSF could be easily modified to improve virus reductions. The ratio of maximum volume in a single charge to pore volume of the filter media bed should preferably be no greater than 1:1 to maximize virus reductions; moreover, a lower ratio would provide a factor of safety to account for longitudinal dispersion. However, any design change should be subject to keeping the cost and size of the BSF reasonable and the assuring that both the volume produced by a single charge and the flow rate are still sufficient to make the filter appealing to users.
Acknowledgments This research was carried out with the financial support of International Aid, Centre for Affordable Water and Sanitation Technology (CAWST), Samaritan’s Purse USA, Samaritan’s Purse Canada and the USEPA P3 Award (SU831831 and SU832463). The primary author’s graduate study was supported by the USEPA STAR Fellowship and the University of North Carolina Society of Fellows Ross and Charlotte Johnson Family Dissertation Fellowship. Douglas Wait, Randy Goodman and Glenn Walters, research staff at the University of North Carolina, are also acknowledged for lending their expertise to the logistics and experimental design stages of this work. Special thanks to Christine Stauber for leading the first UNC laboratory studies on the BSF, and to Alice Wang, Patty Chuang, and Lily Clark for assistance in the laboratory. We would also like to thank two anonymous reviewers for their helpful comments.
references
Aiken, B.A., Stauber, C.E., Ortiz, G.M., Sobsey, M.D., An assessment of continued use and health impact of the concrete biosand filter in Bonao, Dominican Republic. Am. J. Top. Med Hyg. in press. Albert, J., Luoto, J., Levine, D., 2010. End-user preferences for and performance of competing POU water treatment technologies among the rural poor of Kenya. Environ. Sci. Technol. 44 (12), 4426e4432. Bichai, F.B., Barbeau, Y., Dullemont, Hijnen, W., 2010. Role of predation by zooplankton in transport and fate of protozoan (oo)cysts in granular activated carbon filtration. Water Res. 44, 1072e1081. Buzunis, B.J. (1995) Intermittently Operated Slow Sand Filtration: A New Water Treatment Process, Master’s Thesis, University of Calgary, Department of Civil Engineering. Calgary, Alberta. Chang, S.L., Berg, G., Clarke, N.A., Kabler, P.W., 1960. Survival and protection against chlorination of human enteric pathogens in
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 9 2 e4 1 0 2
free-living nematodes isolated from water supplies. Am. J. Trop. Med. Hyg. 9, 136e142. Clasen, T., 2009. Scaling up Household Water Treatment Among Low-Income Populations. (WHO/HSE/WSH/09.02). World Health Organization, Geneva. Cliver, D.O., Herrmann, J.E., 1972. Proteolytic and microbial inactivation of enteroviruses. Water Res. 6, 797e805. Collins, K.E., Cronin, A.A., Rueedi, J., Pedley, S., Joyce, E., Humble, P.J., Tellam, J.H., 2004. Fate and transport of bacteriophage in UK aquifers as surrogates for pathogenic viruses. In: Raymond, N., Yong, H.R. (Eds.), Geoenvironmental Engineering: Integrated Management of Groundwater and Contamination Land. Thomas Telford, London. Confer, D.R., Logan, B.E., 1998. Location of protein and polysaccharide hydrolytic activity in suspended and biofilm wastewater cultures. Water Res. 32 (1), 31e38. DeLoyde, J.L. (2007) Removal of MS2 Bacteriophage, Cryptosporidium, Giardia and Turbidity by Pilot-Scale Multistage Slow Sand Filtration. Master’s Thesis, University of Waterloo, Department of Civil Engineering. Waterloo, Ontario. Duke, W.F., Nordin, R.N., Baker, D., Mazumder, A., 2006. The use and performance of biosand filters in the artibonite valley of Haiti: a field study of 107 households. Rural & Remote Health 6 (3), 570. Dullemont, Y., Schijven, J. F., Hijnen, W.A.M., Colin, M. and Magic-Knezev, A., (2006) Removal of microorganisms by slow sand filtration. In Recent Progress in Slow Sand and Alternative Biofiltration Processes. Edited by R. Gimbel, N.J.D. Graham and M.R. Collins. International Water Association. London. Eighmy, T.T., Spanos, S.K., Royce, J., Collins, M.R., Malley Jr., J.P., 1994. Microbial activity in slow sand filters. In: Collins, M.R., Graham, N.J.D. (Eds.), Slow Sand Filtration: An International Compilation of Recent Scientific and Operational Developments. American Water Works Association, Denver. Elliott, M.A., Stauber, C.E., Koksal, F., Liang, K.R., Huslage, D.K., DiGiano, F.A., Sobsey, M.D., 2006. The operation, flow conditions and microbial reductions of an intermittently operated, household-scale slow sand filter. In: Gimbel, R., Graham, N.J.D., Collins, M.R. (Eds.), Recent Progress in Slow Sand and Alternative Biofiltration Processes. International Water Association, London. Elliott, M.A., Stauber, C.E., Koksal, F., DiGiano, F.A., Sobsey, M.D., 2008. Reductions of E. coli, echovirus type 12 and bacteriophages in an intermittently operated household-scale slow sand filter. Water Res. 42 (10e11), 2662. Elliott, M.A., F.A. DiGiano, A.M. Fabiszewski, P. Chuang, S. Kuhn, M.D. Sobsey (2009) Bench-scale Investigation of Biosand Household Drinking Water Filter Performance. Proceedings of the Carolina Environmental Program’s Environmental Symposium: " Sustainable and Safe Drinking Water in Developing and Developed Countries: Where Science Meets Policy". University of North Carolina at Chapel Hill. Fiore, M.M., Minnings, K., Fiore, L., 2010. Assessment of biosand filter performance in rural communities in southern coastal nicaragua: an evaluation of 199 households. Remote & Rural Health 10, 1483. Forget, A., Fredette, V., 1962. Sodium azide selective medium for the primary isolation of anaerobic bacteria. J. Bacteriol. 83 (6), 1217e1223. Gonzalez, J.M., Suttle, C.A., 1993. Grazing by Marine Nanoflagellates on viruses and virus-sized particles: ingestion and digestion. Marine Ecol. Prog. Ser. MESEDT 94 (1), 1e10. Haarhoff, J., Cleasby, J.L., 1991. Biological and physical mechanisms. In: Logsdon, G.S. (Ed.), Slow Sand Filtration. American Society of Civil Engineers, New York. Hendricks, D.W., Bellamy, W.D., 1991. Microorganisms removals by slow sand filtration. In: Logsdon, G.S. (Ed.), Slow Sand Filtration. American Society of Civil Engineers, New York.
4101
Herrmann, J.E., Kostenbader, K.D., Cliver, D.O., 1974. Persistence of enteroviruses in lake water. Appl. Microbiol. 28 (5), 895e896. Hijnen, W.A.M., Schijven, J.F., Bonne´, P., Visser, A., Medema, G.J., 2004. Elimination of viruses, bacteria and protozoan oocysts by slow sand filtration. Wat Sci Tech 50 (1), 147e154. Huisman, L., Wood, W.E., 1974. Slow Sand Filtration. World Health Organization, Geneva. Jeuland, M., Whittington, D., 2009. Cost-benefit comparisons of investments in improved water supply and cholera vaccination programs. Vaccine 27 (23), 3109e3120. Kim, T.D., Unno, H., 1996. The roles of microbes in the removal and inactivation of viruses in a biological wastewater treatment system. Water Sci. Technol. 33 (11), 243e250. Liang, K., Sobsey, M., Stauber, C., 2010. Improving Household Drinking Water: Use of BioSand Filters in Cambodia. Water and Sanitation Program Field Note, World Bank. http://www. wsp.org/wsp/sites/wsp.org/files/publications/WSP_biosand_ cambodia.pdf. Lipson, S.M., Stotzky, G., 1985. Effect of bacteria on the inactivation and adsorption of clay minerals of reovirus. Can. J. Microbiol. 31 (8), 730e735. Litton, G.M., Olson, T.M., 1993. Colloid deposition rates on silica bed media and artifacts related to collector surface preparation methods. Am. J. Trop. Med. Hyg. 25, 185e193. Lloyd, B.J., 1973. The construction of a sand profile sampler: its use in the study of the Vorticella populations and the general interstitial microfauna of slow sand filters. Water Res. 7, 963e973. Lloyd, B.J. (1974) The Functional Microbial Ecology of Slow Sand Filters. PhD Dissertation, University of Surry, Guildford, UK. Loret, J.F., Jousset, M., Robert, S., Saucedo, G., Ribas, F., Thomas, V., Greub, G., 2008. Amoebae-resisting bacteria in drinking water: risk assessment and management. Water Sci. Technol. 58 (3), 571e577. Manz, D., 2009. BSF Guidance Manual #3 Basic Operation of the Concrete BioSand Water Filter. http://www.manzwaterinfo. ca/gmans.htm Calgary, AB. (Accessed 10.20.09.). McConnell, L.K., Sims, R.C., Barnett, B.B., 1984. Reovirus removal and inactivation by slow-rate sand filtration. Appl. Environ. Microbiol. 48 (4), 818e825. Murphy, H.M., McBean, E.A., Farahbakhsh, K., 2010. Nitrification, denitrification and ammonification in point-of-use biosand filters in rural Cambodia. J. Water Hlth 8 (4), 804e817. Nasser, A.M., Glozman, R., Nitzan, Y., 2002. Contribution of microbial activity to virus reduction in saturated soil. Water Res. 36, 2589e2595. Northrop, J.H., 1964. The effect of proteolytic enzymes on E. coli phages and native proteins. J. Gen. Physiol. 48, 73e78. Peters, F., 1994. Prediction of planktonic protistan grazing rates. Limnol. Oceanogr. Vol. 39, 195e206. Pinheiro, M.D.O., Power, M.E., Butler, B.J., Dayeh, V.R., Slawson, R., Lee, L.E.J., Lynn, D.H., Bols, N.C., 2007. Use of Tetrahymena thermophila to study the role of protozoa in inactivation of viruses in water. Appl. Environ. Microbiol. 73 (2), 643e649. Poynter, S.F.B., Slade, J.S., 1977. The removal of viruses by slow sand filtration. Prog. Water Technol. 9 (1), 75e88. Ramani, S., Kang, G., 2009. Viruses causing childhood diarrhoea in the developing world. Curr Opin Infect Dis. 22 (5), 477e482. Rego, J.V., Billen, G., Fontigny, A., Somville, M., 1985. Free and attached proteolytic activity in water environments. Mar. Ecol. Prog. Ser. 21, 245e249. Rochon, D., Kakani, K., Robbins, M., Reade, R., 2004. Molecular aspects of plant virus transmission by olpidium and plasmodiophorid vectors. Annu. Rev. Phytopathol. 42, 211e241. Rosso, A.L., Azam, F., 1987. Proteolytic activity in coastal oceanic waters: depth distribution and relationship to bacterial populations. Mar. Ecol. Prog. Ser. Vol. 41, 231e240.
4102
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 9 2 e4 1 0 2
Schaub, S.A., Gerba, C.P., 1991. Guide standard and protocol for testing microbiological water purifiers. In: U.S. Environmental Protection Agency and American Water Works Association (Ed.), Point-of-Use/Entry Treatment of Drinking Water. William Andrew, Norwich, USA, pp. 37e43. Schijven, J.F., Hassanizadeh, S.M., 2000. Removal of viruses by soil passage: overview of modeling, processes, and parameters. Crit. Rev. Environ. Sci. Technol. 30 (1), 49e127. Schroth, M.H., Ahearn, S.J., Selker, J.S., Istok, J.D., 1996. Characterization of miller similar silica sands for laboratory hydrologic studies. Soil Sci. Soc. Am. J. 60, 1331e1339. Shuval, H.I., Thompson, A., Fattal, B., Cymbalista, S., Wiener, Y., 1971. Natural virus inactivation processes in seawater. J. Sanit. Eng. Div. Proc. Am. Soc. Civ. Eng. 97, 587e600. Sobsey, M.D., Stauber, C.E., Casanova, L.M., Brown, J.M., Elliott, M. A., 2008. Point of use household drinking water filtration: a practical, effective solution for providing sustained access to safe drinking water in the developing world. J. Bacteriol. 42 (12), 4261e4267. Stauber, C.E., Elliott, M.A., Koksal, F., Ortiz, G.M., DiGiano, F.A., Sobsey, M.D., 2006. Characterisation of the biosand filter for E. coli reductions from household drinking water under controlled laboratory conditions and field use conditions. Water Sci. Technol. 54 (3), 1e7. Stauber, C.E., Ortiz, G.M., Sobsey, M.D., Loomis, D., 2009. Randomized controlled trial of the concrete biosand filter and its impact on diarrheal disease in Bonao, Dominican Republic. Am. J. Trop. Med. Hyg. 80 (2), 286e293. Suttle, C.A., Chen, F., 1992. Mechanisms and rates of decay of marine viruses in seawater. Appl. Environ. Microbiol. 58 (11), 3721e3729. Tiwari, S.K., Schmidt, W.P., Darby, J., Kariuki, Z.G., Jenkins, M.W., 2009. Intermittent slow sand filtration for preventing diarrhoea among children in Kenyan households using
unimproved water sources: randomized controlled trial. Trop. Med. Int. Health 14 (11), 1374e1382. Unger, M., Collins, M.R., 2008. Assessing Escherichia coli removal in the schmutzdecke of slow-rate biofilters. JAWWA 100 (12), 60e73. USEPA, 2001. Method 1602: Male-specific (Fþ) and Somatic Coliphage in Water by Single Agar Layer (SAL) Procedure Washington D.C., EPA 821-R-01e029. Walker, S.C., Toth, T.E., 2000. Proteolytic inactivation of simian-11 rotavirus: a pilot study. Vet. Microbiol. 74, 195e206. Ward, R.L., Knowlton, D.R., Winston, P.E., 1986. Mechanism of inactivation of enteric viruses in fresh water. Appl. Environ. Microbiol. 52, 450e459. Weber-Shirk, M.L., 2002. Enhancing slow sand filter performance with an acid-soluble seston extract. Water Res. 365 (19), 1350e1354. Weber-Shirk, M.L., Chan, K.L., 2007. The role of aluminum in slow sand filtration. Water Res. 41 (6), 4753e4756. Weber-Shirk, M.L., Dick, R.I., 1997a. Physical-chemical mechanisms in slow sand filters. Jour. AWWA 89 (1), 87e100. Weber-Shirk, M.L., Dick, R.I., 1997b. Biological Mechanisms in slow sand filters. Jour. AWWA 89 (2), 72e83. Weber-Shirk, M.L., Dick, R.I., 1998. Bacterivory by a chrysophyte in slow sand filters. Water Res. 33 (3), 631e638. Wheeler, D., Bartram, J., Lloyd, B.J., 1988. The removal of viruses by filtration through sand. In: Graham, N.J.D. (Ed.), Slow Sand Filtration: Recent Developments in Water Treatment Technology. Ellis Horwood Ltd., Chichester, England, pp. 207e229. Windle-Taylor, E., 1970. The removal of viruses by slow sand filtration. Rep. Results Bact. Chem. Biol. Exam. Land. Waters 44 (52), 1969e1970. Yates, M.V., Gerba, C.P., Kelley, L.M., 1985. Virus persistence in groundwater. Appl. Environ. Microbiol. Vol. 49, 778e781.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 0 3 e4 1 1 8
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Enhanced immunological and detoxification responses in Pacific oysters, Crassostrea gigas, exposed to chemically dispersed oil A. Luna-Acosta a,*, R. Kanan b, S. Le Floch b, V. Huet a, P. Pineau a, P. Bustamante a, H. Thomas-Guyon a,* a
Littoral Environnement et Socie´te´s (LIENSs), UMR 6250, CNRS-Universite´ de La Rochelle, 2 rue Olympe de Gouges, F-17042 La Rochelle Cedex 01, France b Centre de Documentation, de Recherche et d’Expe´rimentations sur les Pollutions Accidentelles des Eaux (Cedre), 715 Rue Alain-Colas, CS41836-F-29218 Brest Cedex 2, France
article info
abstract
Article history:
The aim of this study was to evaluate the effects of chemically dispersed oil on an
Received 20 November 2010
economically and ecologically important species inhabiting coasts and estuaries, the
Received in revised form
Pacific oyster Crassostrea gigas. Studies were carried out with juveniles, known to generally
6 May 2011
be more sensitive to environmental stress than adults. A set of enzyme activities involved
Accepted 11 May 2011
in immune defence mechanisms and detoxification processes, i.e. superoxide dismutase
Available online 20 May 2011
(SOD), catalase, glutathione peroxidase (GPx), catecholase-type phenoloxidase (PO), laccase-type PO and lysozyme were analysed in different oyster tissues, i.e. the gills, digestive
Keywords:
gland and mantle, and in the plasma and the haemoycte lysate supernatant (HLS) of the
Chemical dispersion
haemolymph. Results indicated that total PAH body burdens were 2.7 times higher in the
Oil spill
presence than in the absence of the chemical dispersant. After 2 days of exposure to
Bivalve
chemically dispersed oil, alkylated naphthalenes accounted for 55% of the total PAH body
Defence mechanisms
burden, whereas alkylated fluorenes and alkylated dibenzothiophenes accounted for 80%
Tissue-dependent response
when the chemical dispersant was absent. Importantly, a higher number of enzyme activities were modified when oil was chemically dispersed, especially in the plasma and gills. Moreover, independently of the presence or absence of chemical dispersant, oil exposure generally inhibited enzyme activities in the gills and plasma, while they were generally activated in the mantle and haemocytes. These results suggest that the gills and plasma constitute sensitive compartments in C. gigas, and that the mantle and haemocytes may play an important role in protection against xenobiotics. Among the six enzyme activities that were analysed in these body compartments, five were modulated in the chemical dispersion (CD) treatment while only half of the enzyme activities were modulated in the mechanical dispersion treatment. Furthermore, CD treatment effects were often observed following exposure, but also during depuration periods. These results suggest that immune and/or detoxification responses are likely to be affected when dispersants are used to treat oil spills in shallow waters. ª 2011 Elsevier Ltd. All rights reserved.
* Corresponding authors. Tel.: þ33 5 46 50 76 23; fax: þ33 5 46 50 76 63. E-mail addresses:
[email protected] (A. Luna-Acosta),
[email protected] (H. Thomas-Guyon). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.05.011
4104
1.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 0 3 e4 1 1 8
Introduction
The biological and economical consequences of numerous accidental oil spills that have occurred during the last 40 years are concrete proof of the need to develop reliable oil spill countermeasures. Large oil spills (>100 tonnes) represent only 6% of the total introduction of oil into the environment (NCR, 2003). Indeed, most oil spills are relatively small (w1 m3). They occur close to the shore and oil slicks hit shorelines relatively quickly. As a result of the Torrey Canyon incident, where large amounts of toxic dispersants were used and caused severe long-term biological impacts, research efforts have focused on the mechanical clean-up and containment of floating oil. However, these types of techniques in special areas, such as estuaries and closed bays, can also cause additional harm to oil-impacted zones (Marchand, 1998). During the same period, chemical dispersants evolved from first generation products, effective but highly toxic, to third generation products, with low toxicity and high biodegradability. Chemical dispersants are complex mixtures, primarily containing surfactants and solvents, which reduce the interfacial tension at the oilewater interface, and therefore facilitate the downward mixing of oil into the water (Canevari, 1973; Li and Garrett, 1998). In this way, oil can be dispersed to concentrations below toxicity threshold limits (Lessard and DeMarco, 2000; Page et al., 2000), become more accessible to hydrocarbon-degrading bacteria (Venosa and Zhu, 2003), and have a lower impact on shorelines. Moreover, third generation dispersants are considered non-toxic and biodegradable. Unfortunately, little is known about the behaviour and effects of dispersed oil in the nearshore environment (ASTM Committee, 1984). Indeed, because of the higher exposure of organisms to petroleum compounds that may be triggered by the use of chemical dispersants in zones with low water-column depth, e.g. coastal areas and estuaries, most countries that allow the use of dispersants have strict rules: minimum water depth (normally 20 m), minimum current speed (normally 1 m per second), and minimum distance from the shore or from offshore islands (normally 2 km) (Ramachandran et al., 2004). Recently, a Net Environmental Benefit Analysis (NEBA) carried out by Baca et al. (2006) on a 20-year field study, revealed the lack of longterm environmental impact of dispersed oil in nearshore tropical areas. However, dispersants are a wide-ranging group of chemicals with varying chemical properties and related toxicities. Therefore, many questions remain unanswered about the possible direct or indirect effects of their use in other nearshore zones, in response to oil spills, where human activities are heavily concentrated, such as in the case of the Transocean Deepwater Horizon oil rig explosion. Coasts and estuaries are considered to be sensitive areas since they provide habitats for a large number of organisms, constitute nursery grounds for juveniles of several commercially important species, and bear very high productivity. Sessile and filter-feeder organisms inhabiting these zones, such as the Pacific oyster Crassostrea gigas (Thunberg, 1753), are constantly in contact with various chemical molecules. Environmental stress from pollutants is likely to be an important
determining factor in weakening defence mechanisms in these organisms and therefore promoting the occurrence or increase in diseases, particularly at early life stages, i.e. larvae and juveniles (Perdue et al., 1981; Lacoste et al., 2001). Among immune defence mechanisms in bivalves, antioxidant enzyme activities, e.g. superoxide dismutase (SOD), catalase, glutathione peroxidase (GPx), and enzyme activities involved in humoral innate defences, e.g. phenoloxidase (PO) and lysozyme, have been shown to be modulated by the presence of several types of pollutants (Bado-Nilles et al., 2008; Stabili and Pagliara, 2009; Verlecar et al., 2007). In this respect, these responses have been shown to be tissue-dependent (Cheung et al., 2001; Luna-Acosta et al., 2010a). In this general context, the aim of this study was to experimentally assess 1) the bioaccumulation and 2) the effects of chemically dispersed oil on immune defence and/or detoxification mechanisms, i.e. SOD, catalase, GPx, catecholase- and laccase-type PO, and lysozyme activities of Pacific oyster C. gigas juveniles. For this purpose, enzyme activities were determined in different tissues, i.e. gills, digestive gland, mantle, and in the haemocytes and plasma, or acellular fraction, of the haemolymph. To this end, comparisons were made between oysters exposed to oil subjected to chemical dispersion (CD) or mechanical dispersion (MD), to the water soluble fraction of the oil (WSF) and to the dispersant alone (D).
2.
Material and methods
2.1.
Chemicals
Oil A Brut Arabian Light crude oil (BAL 110) was used for this study. The crude oil was topped at 110 C to remove the most volatile components, in order to simulate the natural weathering of the oil after its release at sea (evaporation of most volatile components), before it reaches coastal zones. BAL 110 possesses the following physico-chemical characteristics, similar to the oil spilled by the Amoco Cadiz in 1978: density of 0.860 at 20 C, viscosity of 60 mPa s at 15 C, 12% polar compounds, 34% aromatic hydrocarbons and 54% saturated hydrocarbons. Dispersant The chemical dispersant used in this study was selected following an evaluation carried out by the Centre of Documentation, Research and Experimentation on Accidental Water Pollution (Cedre), which defines it as 1) effective for use in the marine environment, 2) non-toxic at the concentration recommended by the manufacturer Total Fluides (i.e. 5% v/v) and (3) biodegradable. Its physico-chemical characteristics were not available for reasons of confidentiality.
2.2.
Biological material
Pacific oyster C. gigas juveniles (3e4 cm in height, less than 1 year old) were purchased from the hatchery France Naissain, located in Bouin (France). The oysters were acclimatised in the laboratory at 15 1 C for two weeks before starting the experiments. They were fed daily with an algal diet (5 104 cell ml1) composed of
4105
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 0 3 e4 1 1 8
Heteroskeletonema sp. (Bacillariophyceae) purchased from the hatchery SATMAR (Normandie, France).
2.3.
Experimental design
The experiment, consisting of an exposure period followed by a depuration period, was carried out three times to provide analysis replicates. Experimental system The experimental system consisted of 300 L static water tanks. Because of the hydrophobic character of the oil, a funnel (at the surface of each tank) was connected to a submersed Johnson L450 water pump (at the bottom of the tank), in order to maintain the mixture of oil and dispersant as a homogenous solution. Preliminary tests confirmed that total petroleum hydrocarbon concentrations in the water column were depth-independent, suggesting that small petroleum droplets were homogeneously dispersed in the water column (data not shown). The oxygen saturation of water in each tank was maintained at around 96% by a compressor that injected air via an air stone. The experimental system was placed in a thermoregulated greenhouse, in order to maintain the temperature of the exposure studies at 15 1 C. Exposure One experimental system was used per condition, making a total of five experimental systems. Thirty oysters were transferred into each experimental system. The different exposure conditions are summarized in Table 1. The exposure period lasted 2 days. The seawater used for this study (pH: 7.95 0.05, salinity: 35.2 0.1 p.s.u.) was provided by Oceanopolis, Brest (France). This seawater was treated by UV-light and filtrated at 0.45 mm before use. In the first tank, the control exposure condition was set up with clean seawater. For the amount of BAL selected, the aim of this study was to obtain a concentration of dispersed oil in the same rank than those reported in situ, following an oil spill, such as reported by Lunel (1995) following the Braer oil spill. Thus, in the second tank, the chemical dispersion (CD) condition was set up by pouring 20 g of BAL 110 and the volume of dispersant recommended by the manufacturer, i.e. 1.2 g of dispersant, into the funnel of the experimental system. In the third tank, the mechanical dispersion (MD) condition was set up by pouring 20 g of BAL 110 into the funnel. In the fourth tank, the toxicity of molecules that naturally dissolve in seawater was tested by exposing oysters to a water-soluble fraction (WSF) of BAL 110. To obtain the WSF, a plastic circle was set on the surface of the seawater in the experimental system. The BAL 110 (20 g) was
then poured into the plastic circle, in order to contain the oil slick at the surface, without mixing. The oysters were therefore only exposed to the soluble fraction of the oil, i.e. free of particles of bulk material, and, contrary to the wateraccommodated fraction (WAF), free of soluble and volatile compounds that can naturally evaporate (Anderson et al., 1974; Singer et al., 2000). In the fifth tank, an internal control for the CD condition was set up by pouring 1.2 g of chemical dispersant into the funnel (i.e., dispersant condition, D). All experimental systems contained a funnel connected to a submersed water pump. All exposure media were added to the tanks 13 h before adding the oysters, the time needed to obtain a relatively stable oil concentration in the water column. The oysters were not fed during the exposure period. Depuration Ten oysters per treatment condition were placed in a decontamination tank, located in the thermoregulated greenhouse (T ¼ 15 1 C) and containing clean seawater, for a recovery period of 15 days. The oysters were fed daily with an algal diet (5 104 cell ml1) composed of Heteroskeletonema sp. (Bacillariophyceae).
2.4.
Sampling procedure
Pooled gills, digestive glands, mantles, haemocyte fraction and plasma of ten oysters were used for each replicate sample, and three replicates were prepared per treatment. After opening the oyster shells by cutting off the adductor muscle, approx. 0.3e0.5 ml of haemolymph was withdrawn from the pericardial cavity using a 1-ml syringe equipped with a needle (0.9 25 mm). Haemolymph samples were centrifuged at 260 g for 10 min at 4 C in order to separate the cellular (haemocytes) fraction from the plasma. The gills, digestive gland and mantle were removed from the soft tissues and homogenized at 4 C in 0.1 M Tris HCl buffer pH 7.0 (0.45 M NaCl, 26 mM MgCl2, 10 mM CaCl2; 0.5 ml of buffer.g1 of fresh weight for the gills and the mantle, and 1 ml g1 of fresh weight for the digestive gland), using an Ultra Turrax (T25 basic, IKA-WERKE) and a Thomas-Potter homogenizer (IKALabortechnik RW 20.n, size 0.13e0.18 mm, BB). The homogenates were centrifuged at 10 000 g for 10 min at 4 C. The resulting supernatant was collected for enzymatic studies.
2.5.
Biochemical analysis
Superoxide dismutase assay SOD was determined as described previously (Luna-Acosta et al., 2010a) based on competition of
Table 1 e Experimental conditions used in the study. CD: chemical dispersion, MD: mechanical dispersion, WSF: water soluble fraction and D: dispersant. BAL 110: Brut Arabian Light crude oil topped at 110 C. Quantity of the product added to the water column (mg l1)a
Control CD MD WSF D (internal control of CD)
BAL 110
Dispersant
0 67 67 67 0
0 4 0 0 4
a In order to have a stable concentration of the products in the water column, the products were added 13 h before introducing animals in the tanks.
4106
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 0 3 e4 1 1 8
SOD with iodonitrotetrazolium (INT) for dismutation of superoxide anion (O 2 ). In the presence of O2 , INT is reduced into a red formazan dye that can be measured at 505 nm at 25 C (Kit Ransod SD 125, Randox, France). One unit of SOD is defined as the amount of enzyme that promotes a 50% decrease in the rate of INT reduction. Glutathione peroxidase assay GPx activity was determined as described previously (Luna-Acosta et al., 2010a). In the presence of glutathione reductase and substrates (i.e. reduced glutathione and cumene hydroperoxide), the decrease of absorbance at 340 nm was proportional to the reduction of the oxidised glutathione by NADPH, Hþ (Kit Ransel RS 504, Randox, France). One unit of GPx oxidises 1 mmol of NADPH (eNADPH ¼ 6.22 mM1 cm1) per minute. Catalase assay Catalase activity was determined according to the method of Fossati et al. (1980). This method is based on the measurement of the hydrogen peroxide substrate remaining after the action of catalase. First, catalase converts hydrogen peroxide into water and oxygen and then this enzymatic reaction is stopped with sodium azide. An aliquot of the reaction mix was then assayed for the amount of hydrogen peroxide remaining using a colourimetric method (Kit Catalase CAT-100, Sigma). Catecholase-type and laccase-type phenoloxidase assay Catecholase- and laccase-type phenoloxidase (PO) activities were determined according to the method described previously (Luna-Acosta et al., 2010b). For PO assays, 100 mM of dopamine or 50 mM of PPD were used as substrates and the increase of absorbance at 490 and 420 nm was monitored for 4 and 2 h for catecholase- and laccase-type PO activity, respectively. Nonenzymatic oxidation by the substrate was monitored in wells without oyster sample and subtracted from oxidation of the substrate with oyster sample. One unit of catecholase- and laccase-type PO activities corresponds to the amount of enzyme that catalyzes the production of 1 mmole of product (e ¼ 3300 M1 cm1 and 43 160 M1 cm1, respectively), per minute. Lysozyme assay Lysozyme was determined as described previously (Luna-Acosta et al., 2010a). The lysozyme assay was done in triplicate for each sample and compared to hen egg white lysozyme standards (2.5e20 mg ml1), in the presence of Micrococcus lysodeikticus (SigmaeAldrich, France). One unit of lysozyme corresponds to the amount of enzyme that diminishes absorbance at 450 nm by 0.001 per minute at pH 7.0, at 25 C. Protein assay All activities were expressed in relation to the protein concentration measured according to the Lowry method with slight modifications, using bicinchoninic acid and copper sulphate 4% (Smith et al., 1985). Serum albumin was used as protein standard (SigmaeAldrich, France). All enzyme activities were measured in the gills, digestive gland, mantle, plasma and HLS, except for lysozyme that was not measured in the HLS, making a total of 29 enzymatic analyses.
2.6.
Chemical analysis in oysters
The levels of polycyclic aromatic hydrocarbons (PAHs) in oysters were determined with a Varian Saturn 2100 T GCeMS device, using the procedure of Baumard et al. (1997) with some modifications. Three pools of five oysters were analysed per
treatment. Prior to extraction, each oyster sample was homogenized using an Ultra Turrax (Janke and Kunkel, IKALabortechnik). 150 ml of perdeuterated internal standards (CUS-7249, Ultra Scientific, Analytical solutions) were added to 3 g of homogenized oyster samples and the mixtures obtained were digested for 4 h under reflux in 50 ml of an ethanolic solution of potassium hydroxide (2 M, Fisher Chemicals). After cooling, settling and addition of 20 ml of demineralised water, the digest was extracted in a 250 ml funnel twice with 20 ml of pentane (Carlo Erba Reactifs, SDS). The extract was evaporated with a TurboVap 500 concentrator (Zyman, Hopkinton, MA, USA, at 880 mbar and 50 C) to obtain 1 ml of concentrated extract. The purification of the extract was performed by transfer to a silica column (5 g of silica). Hydrocarbons were eluted with 50 ml of pentane:dichloromethane (80:20, v:v, SDS) and concentrated to 200 ml by means of a TurboVap 500 concentrator (Zyman, 880 mbar, 50 C). Aromatic compounds were analysed by GCeMS, with a detection limit of 0.005 mg g1 of dry weight, and PAHs were quantified relative to the perdeuterated internal standards introduced at the beginning of the sample preparation (Roy et al., 2005). Five perdeuterated standard compounds, i.e. Naphthalene d8, Biphenyl d10, Phenanthrene d10, Chrysene d12, and Benzo[a]pyrene d12, representative of all the PAHs analysed, were used as internal standards. A total of 20 parent PAHs and 25 alkylated compounds were quantified. The PAH recovery was >60% and the relative standard deviation (RSD, i.e. (standard error/mean 100)) was <15%.
2.7.
Statistical analysis
All values are reported as mean standard deviation (SD). Statistical analysis was carried out with STATISTICA 7.0. Values were tested for normality (Shapiro test) and homogeneity of variances (Bartlett test). In some cases, logarithmic transformations (Log10) were used to meet the underlying assumptions of normality and homogeneity of variances. Two-way nested MANOVA were used to analyse results, with treatment and period as fixed factors, and pool as a random factor. The period factor corresponds to the exposure period and the depuration period. Pool was nested within each combination of treatment and time (Zar, 1984). When the null hypothesis (H0: no difference between treatments or within treatment at different time intervals) was rejected, significant differences were tested using Tukey’s HSD test. For non normal values, i.e. to compare oysters’ PAH content between different treatments, a KruskaleWallis test was used, followed by a Dunn’s multiple comparisons test. Statistical significance was determined as being at the level of p < 0.05. The relationships between body burdens for the different PAH categories (HMW PAHs, LMW PAHs, parent PAHs and alkylated PAHs) and the responses of the enzyme activities were analysed using principal component analysis (PCA) and redundancy analysis (RDA) (Leps and Smilauer, 1999). Separate analyses were performed for the different organs. Data for enzyme activities were centred and standardized before analysis. PCA and RDA were performed using CANOCO for Windows software package, Version 4.5 (Centre for Biometry, Wageningen, The Netherlands). For
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 0 3 e4 1 1 8
RDA analyses, the significance of the relationship between enzyme activities and PAH body burdens were tested using a global Monte Carlo permutation test.
3.
Results
3.1.
Chemical analyses
After 2 days of exposure, PAHs had been efficiently bioaccumulated in the soft tissues of C. gigas. The total PAH P content in oysters ( PAHs) was equal to 7 1, 530 75, 198 22, 56 50 and 8 3 mg g1 dry weight for the control, the chemically dispersed oil (CD), the mechanically dispersed oil (MD), the water soluble fraction (WSF) and the dispersant (D) treatments, respectively. Significant differences were found between the control and the CD or the MD but not with the WSF conditions (F4,15 ¼ 12.83, p ¼ 0.01). This lack of significant difference could be explained by the high variation in total PAH levels in the WSF condition. Importantly, the total PAH content in oysters in the CD condition was almost 3 times higher than in the MD condition. P After 15 days of depuration, the PAHs was equal to 3 0, 1 15 6, 14 6, 2 0 and 2 2 mg g dry weight in control, CD, MD, WSF and D treatments, respectively. Even if the PAH content in oysters had decreased by 97 and 93% in CD and MD conditions, respectively, it remained significantly higher than the control condition (F4,15 ¼ 10.27, p ¼ 0.03). Independently of time and treatment, light PAHs (3 rings) and their alkylated homologues accounted for at least 90% of the total PAHs in oyster tissues (Fig. 1). After 2 days of exposure to the CD condition, alkylated naphthalenes (NaF), alkylated dibenzothiophenes (DBT) and alkylated fluorenes (Fl) accounted for most of the PAH content in oyster soft tissues with 55 6, 19 1, 18 8% for alkylated NaF, alkylated DBT and alkylated Fl, respectively (Fig. 1). In the MD condition, alkylated DBT and alkylated Fl accounted for most of the PAH content in the oysters’ soft tissues, representing 38 2 and 40 2% of the total PAHs, respectively. Alkylated DBT (22 0%) and alkylated Fl (53 0%) also accounted for a large proportion of the PAH content in the oysters’ soft tissues in the WSF condition. For the D condition, alkylated DBT (45 6%) and anthracene (30 7%) accounted for most of the PAH content in oyster soft tissues. After 15 days of depuration, alkylated DBT were the predominant PAH compounds in oyster tissues of the CD, MD, WSF and D conditions, representing 56 19%, 48 7%, 40 5% and 50 16% of the total PAHs, respectively (Fig. 1).
3.2.
4107
significant effect was observed for treatment for lysozyme activity in the gills and plasma (data not shown). After 2 days of exposure for the CD condition, and relative to the control condition, catalase (Fig. 3) and GPx (Fig. 4) activities were completely inhibited and laccase activity (Fig. 6) was 1.4-fold higher in the gills. SOD (Fig. 2) and GPx (Fig. 4) activities were 30% and 50% lower in the digestive gland, respectively. Catalase activity (Fig. 3) was 75% lower in the mantle. Interestingly, results from the HLS and the plasma differed. While no significant effect was observed in the HLS, SOD activity (Fig. 2) was 2-fold higher and catalase (Fig. 3) and laccase (Fig. 6) activities were completely inhibited in the plasma. For the MD condition, SOD (Fig. 2) and laccase (Fig. 6) activities were 2.5-fold and 1.4-fold higher in the gills, respectively. GPx activity (Fig. 4) was 1.5fold higher in the mantle. While SOD (Fig. 2), GPx (Fig. 4), catecholase (Fig. 5) and laccase (Fig. 6) activities were 2e4fold higher in the HLS, catalase activity (Fig. 3) was completely inhibited in the plasma. For the WSF condition, SOD (Fig. 2) and laccase (Fig. 6) activities were 2.5- and 1.7fold higher in the gills, respectively. SOD (Fig. 2) and catecholase (Fig. 5) activities were 4- and 2-fold higher in the HLS, respectively. It is important to notice that the dispersant alone (D condition), which is considered as a harmless product, also modulated different enzyme activities in C. gigas: laccase activity (Fig. 6) was 1.4-fold higher in the gills; SOD (Fig. 2) and lysozyme (Fig. 7) activities were 70 and 55% lower in the digestive gland, respectively, and SOD (Fig. 2), catecholase (Fig. 5) and laccase (Fig. 6) activities were 2.5- to 3-fold higher in the HLS. After 15 days of depuration, some activities returned to control levels, depending on the tissue and the treatment. In some tissues and for some activities, effects were observed only at this period, e.g. SOD activity (Fig. 2) was 1.3-fold higher in the mantle for the MD condition; catalase activity (Fig. 3) was 2- to 2.5-fold higher in the digestive gland for the CD, WSF and D conditions; SOD activity (Fig. 2) was 1.3-fold higher for the MD condition; lysozyme activity (Fig. 7) was 1.4- to 1.6-fold higher for the CD, WSF and D conditions in the mantle; catecholase activity (Fig. 5) was inhibited by 20e30% in the plasma for the CD, MD and WSF conditions, and 1.4-fold higher in the D condition. In addition to these results, a significant treatment, but not period, effect was observed in the mantle for catecholase (Fig. 5) and laccase (Fig. 6) activities. In this tissue, catecholase activity was inhibited by 20% for the CD and MD conditions and laccase activity was 1.2-fold higher for the CD and D conditions.
3.3. Relationships between PAH contents in oyster soft tissues and enzyme activities
Enzymatic analysis
Overall, no significant differences in enzymatic activities were observed between the control conditions from both exposure and depuration periods, whatever the tissue (Figs. 2e7). In contrast, all enzyme activities were affected by both treatment and period, independently of the considered tissue, except for catecholase and laccase activities in the mantle, which were only affected by the treatment (Figs. 5 and 6). No
At the end of the exposure period, significant correlations were observed between different enzyme activities and PAH contents in oyster soft tissues and responses varied according to the tissue that was analysed. The results of PCA and RDA are presented in Fig. 8 and Table 2, respectively. RDA indicated PAH body burdens as the significant variable explaining 49%, 43%, 43%, 39%, 44% of total variation in enzyme activities in the gills, digestive gland, mantle, plasma and HLS, respectively.
4108
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 0 3 e4 1 1 8
Fig. 1 e Proportion (%, mean ± SD, n [ 3) of the main PAHs contained in oyster tissues after 2 days of exposure to chemical dispersion (CD), mechanical dispersion (MD), water soluble fraction (WSF) and dispersant (D) treatments, followed by 15 days of depuration. PAHs: polycyclic aromatic hydrocarbons; NaF: naphthalene; Ac: acenaphtylene; Fl: fluorene; Fen: phenanthrene; Ant: anthracene; DBT: dibenzothiophene.
Enzyme activities in the gills, plasma and mantle were significantly correlated to parent PAHs. Enzyme activities in the plasma were also significantly correlated to LMW PAHs. Four out of six enzyme activities analysed in the gills and
plasma were negatively correlated to body burdens of the various PAH categories: catalase, GPx, lysozyme, and catecholase or laccase, in the gills and plasma, respectively (Fig. 8, Table 2). SOD, GPx, catecholase and lysozyme in the
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 0 3 e4 1 1 8
4109
Fig. 2 e Superoxide dismutase activity in the gills, digestive gland, mantle, plasma and haemocyte lysate supernatant (HLS) in C. gigas after 2 days of exposure to chemical dispersion (CD), mechanical dispersion (MD), water soluble fraction (WSF) and dispersant (D) treatments, followed by 15 days of depuration. Data are expressed as mean ± SD U.mg protL1, n [ 9 (i.e. 3 sample replicates from 3 experiment replicates). For clarity of results, only significant differences between control and treatment and for a given time (after 2 d of exposure or after 15 d of depuration) are shown; *p < 0.05, **p < 0.01, ***p < 0.001.
mantle were positively correlated to body burdens of the various PAH categories (Fig. 8, Table 2). Four out of five enzyme activities analysed in the HLS, i.e. SOD, GPx, catecholase and laccase, were negatively correlated to body burdens of the various PAH categories, but were positively correlated to parent PAHs (Fig. 8, Table 2). Enzyme activities in the digestive gland were significantly correlated to HMW PAHs. Five out of six enzyme activities analysed in the digestive gland were positively correlated to body burdens of the various PAH categories: SOD, catalase, catecholase, laccase and lysozyme (Fig. 8, Table 2).
4.
Discussion
4.1. PAH bioaccumulation and depuration in oyster tissues The aim of this study was to assess 1) the bioaccumulation and 2) the effects of chemically dispersed hydrocarbons on a species inhabiting coastal and estuarine zones, the Pacific oyster C.
gigas. This marine bivalve is a good indicator of the presence and bioavailability of oil in the water column and benthic sediments. In this species, bioconcentration factors (BCF) of petroleum hydrocarbons range from 10 to 50 000 (Michel and Henry, 1997), so even low levels of exposure are likely to be detectable in oyster tissues. PAHs, with high octanolewater partition coefficients (log Kow > 3.5), are readily taken up by organisms (Meador, 2003). Thus, evaluating PAH bioaccumulation plays an important part in assessing the risk that chemical dispersants are likely to pose to marine organisms inhabiting coasts and estuaries, especially carcinogenic, mutagenic or teratogenic PAHs. Because no PAH measurements were carried out in seawater, we were not able to calculate BCF, but PAH body burden analysis was conducted in oyster soft tissues. Results showed that, independently of the treatment (i.e. CD, MD, WSF or D), heavy PAHs were poorly accumulated in oyster tissues (0e3% of the total PAH content), while light PAHs and mainly their alkylated homologues, which are generally more toxic than the parent compounds, were present in large proportions in all the treatment conditions. Indeed, PAHs with low molecular weight (178.2 g mol1)
4110
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 0 3 e4 1 1 8
Fig. 3 e Catalase activity in the gills, digestive gland, mantle, plasma and haemocyte lysate supernatant (HLS) in C. gigas after 2 days of exposure to chemical dispersion (CD), mechanical dispersion (MD), water soluble fraction (WSF) and dispersant (D) treatments, followed by 15 days of depuration. Data are expressed as mean ± SD U.mg protL1, n [ 9 (i.e. 3 sample replicates from 3 experiment replicates). For clarity of results, only significant differences between control and treatment and for a given time (after 2 d of exposure or after 15 d of depuration) are shown; *p < 0.05, **p < 0.01, ***p < 0.001.
such as NaF, DBT and Fl are more likely to be bioaccumulated due to high water solubility and higher bioavailability for organisms (Neff, 2002). A high heterogeneity in PAH body burden was observed for the WSF treatment. The WSF corresponds to hydrocarbon molecules that are liable to naturally dissolve, meaning that the solution can be considered as homogeneous. Therefore the heterogeneity of results for the WSF treatment suggests that 1) organisms depurate at different rates and/or that 2) organisms accumulate at different rates (Neff, 2002). Interestingly, for the same amount of dispersed oil with (CD condition) or without (MD condition) a chemical dispersant, and for the same exposure time, i.e. 2 days, oysters bioaccumulated approximately 2.7 times more PAHs in the presence of the chemical dispersant, showing that the dispersant increased the bioavailability of PAHs and thus their uptake rate in C. gigas. The bioaccumulated PAH compounds were completely different between both CD and MD conditions. In the CD condition, 55% of the total PAH content was composed of high alkylated NaF. These compounds were poorly bioaccumulated in the MD and WSF conditions (Fig. 1), whereas
alkylated Fl and alkylated DBT represented between 70 and 80% of the total PAH content (Fig. 1). Our results for the CD condition but not for the MD condition are in agreement with a recent study carried out on the Iceland scallop Chlamys islandica where C3-NaF accounted for a large proportion of the PAHs that were accumulated in the tissues after 15 days of exposure to mechanically dispersed oil (Hannam et al., 2009). In the present study, the exposure experiment was carried out for only 2 days, and, since uptake rate constants of PAHs in bivalves generally increase with molecular weight or molecular complexity (McLeese and Burridge, 1987), it cannot be excluded that a longer exposure period would result in a higher alkylated NaF uptake in oyster tissues. Nonetheless, our results clearly demonstrated that the presence of the chemical dispersant increased the bioaccumulation rate of alkylated NaF in oyster soft tissues. However, the processes that could alter 1) the bioconcentration and/or 2) the type of components accumulated, when oil is chemically dispersed, have been poorly described in the scientific literature. A possible contributing factor for bioaccumulation of some components is that dispersing spilled oil converts the oil from
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 0 3 e4 1 1 8
4111
Fig. 4 e Glutathione peroxidase activity in the gills, digestive gland, mantle, plasma and haemocyte lysate supernatant (HLS) in C. gigas after 2 days of exposure to chemical dispersion (CD), mechanical dispersion (MD), water soluble fraction (WSF) and dispersant (D) treatments, followed by 15 days of depuration. Data are expressed as mean ± SD U mg protL1, n [ 9 (i.e. 3 sample replicates from 3 experiment replicates). For clarity of results, only significant differences between control and treatment and for a given time (after 2 d of exposure or after 15 d of depuration) are shown; *p < 0.05, ** p < 0.01, ***p < 0.001.
a surface slick to a plume of small oil droplets dispersed in the water column. These oil droplets might be more easily ingested by filter-feeding organisms, such as oysters (Baussant et al., 2001). Additionally, particles trapped on the oysters’ gills are directed into the gut and later incorporated into endocytic vacuoles in the digestive gland. Oil particles retained in intracellular vacuoles can also be assimilated to the tissue lipids (Baussant et al., 2001). Thus, in the CD treatment, a significant amount of the larger PAHs bound to particles could be retained and accumulated in intracellular vacuoles or in tissue lipids during the exposure period. After 15 days of depuration, more than 90% of the PAH body burden had been depurated from oysters’ soft tissues. Even if PAH contents in CD and MD conditions were still significantly higher than the control condition, this result clearly shows that oysters can eliminate high quantities of PAHs very rapidly. Interestingly, no differences in the total PAH content between CD and MD were found for the depuration period. This result is in agreement with a previous study carried out on mussels (Gilfillan et al., 1984). Among PAHs, alkylated DBT were the most persistent PAHs in all
treatments (Fig. 1). This result is in agreement with Berthou et al. (1987) who report that DBT persist in oyster tissues for at least one year. As in the present study, C3-DBT was among the most persistent PAHs (Berthou et al., 1987), raising questions about its potential toxic effects. Further studies are therefore needed in order to assess the longterm toxicity of this compound.
4.2.
Effect on enzyme activities
Disease emergence and organism survival are determined partly by the condition of the immune system. Therefore, the measurement of defence mechanisms can provide important early warning signals of the sub-lethal effects of exposure to contaminants and the susceptibility of animals to infectious diseases (Hannam et al., 2009). Bivalve molluscs possess two types of innate responses: 1) cellular, i.e. phagocytosis and encapsulation; and 2) humoral, e.g. PO and lysozyme enzyme activities (Tryphonas et al., 2005). During phagocytosis, reactive oxygen species (ROS), such as the superoxide anion (O2) and hydrogen peroxide (H2O2),
4112
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 0 3 e4 1 1 8
Fig. 5 e Catecholase-type phenoloxidase activity in the gills, digestive gland, mantle, plasma and haemocyte lysate supernatant (HLS) in C. gigas after 2 days of exposure to chemical dispersion (CD), mechanical dispersion (MD), water soluble fraction (WSF) and dispersant (D) treatments, followed by 15 days of depuration. Data are expressed as mean ± SD U mg protL1, n [ 9 (i.e. 3 sample replicates from 3 experiment replicates). For clarity of results, only significant differences between control and treatment and for a given time (after 2 d of exposure or after 15 d of depuration) are shown. For the mantle, the enzyme activity was not affected by both treatment and period, but only by treatment. Therefore, results represent the mean of overall data from exposure and depuration periods. *p < 0.05, **p < 0.01, ***p < 0.001.
play an important role in protection against pathogens. However, ROS can also interact with hosts’ biological macromolecules, which can result in enzyme inactivation, lipid peroxidation, DNA damage or cell death (Cazenave et al., 2006). The extent to which oxyradical generation produces biological damage depends on the effectiveness of antioxidant defences, such as SOD, which participates in the transformation of O2 into H2O2 and H2O, and of catalase and GPx, which participate in the transformation of H2O2 into O2 and H2O (Neumann et al., 2001). Environmental contaminants, such as PAHs, can lead to the formation of ROS and enhance oxidative stress in aquatic organisms (Winston, 1991). Results from the present study showed a positive correlation between PAH contents and SOD activities. Such a relationship has already been reported for different bivalve species exposed to hydrocarbons (Sole et al., 1995; Orbea et al., 2002; Richardson et al., 2008) and
suggests that hydrocarbons induce oxidative stress by producing ROS such as O 2 . However, in the present study, catalase and GPx activities were generally negatively correlated with different PAH body burdens. This may be due to the inhibition of enzyme synthesis by PAHs or to enzyme inactivation caused by high tissue contaminant concentrations (Borg and Schaich, 1983). Moreover, since GPx and catalase catalyze the transformation of H2O2 into H2O, they may act on common substrates, and thus competition may exist for the same group of substrates (Kappus, 1985). This may explain the positive correlations of catalase activity and negative correlations of GPx activity with PAH content for the same tissues in the present study (Table 2). Among enzymes involved in humoral immune defences in bivalves, POs are the key enzymes of melanization, participating in the entrapment of foreign material in a melanin capsule or in the direct killing of microbes by the toxic quinone
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 0 3 e4 1 1 8
4113
Fig. 6 e Laccase-type phenoloxidase activity in the gills, digestive gland, mantle, plasma and haemocyte lysate supernatant (HLS) in C. gigas after 2 days of exposure to chemical dispersion (CD), mechanical dispersion (MD), water soluble fraction (WSF) and dispersant (D) treatments, followed by 15 days of depuration. Data are expressed as mean ± SD U mg protL1, n [ 9 (i.e. 3 sample replicates from 3 experiment replicates). For clarity of results, only significant differences between control and treatment and for a given time (after 2 d of exposure or after 15 d of depuration) are shown. For the mantle, the enzyme activity was not affected by both treatment and period, but only by treatment. Therefore, results represent the mean of overall data from exposure and depuration periods. *p < 0.05, **p < 0.01, ***p < 0.001.
intermediates produced during the melanin production cascade (So¨derha¨ll and Cerenius, 1998). Tyrosinases, catecholases, and laccases belong to the family of POs. While tyrosinase possesses monophenolase (hydroxylation of monophenols) and catecholase (oxidation of o-diphenols) activities, catecholase only possesses catecholase activities and laccase possesses both catecholase and laccase activities (oxidation of o-, p- and m-diphenols, and other non phenolic substrates such as diamines). In a recent study, two types of PO activity, catecholase and laccase, were detected in C. gigas (Luna-Acosta et al., 2010b). In the present study, catecholase activities were generally inhibited in the presence of hydrocarbons, except for HLS catecholase activity. These results are in agreement with other studies on C. gigas where exposure to hydrocarbons induced an inhibitory effect in PO activities (Bado-Nilles et al., 2008). Moreover, when using PPD as a substrate for PO assays in this study, laccase activities were generally stimulated in the presence of hydrocarbons, except for laccase activity in plasma. These results are in agreement with previous studies where the exposure of C. gigas to a light cycle oil (LCO) soluble fraction induced an
increase in the mRNA expression of a laccase (multicopper oxidase) gene in haemocytes after 7 days of exposure (BadoNilles et al., 2010). Since laccases can catalyze oxidation of aromatics, such as PAHs, by an indirect mechanism involving the participation of an oxidative mediator (Dodor et al., 2004), an increase in laccase gene transcription in C. gigas may enable protection against bioaccumulated PAHs. Laccase could therefore be a potential defence biomarker candidate in ecotoxicological studies. Lysozyme is one of the most important bacteriolytic agents against several species of Gram-positive and Gram-negative bacteria, and has been recorded for various bivalve species (McHenery et al., 1986). During phagocytosis, the release of lysosomal enzymes, such as lysozymes, participates in the inactivation of invading pathogens. In the present study, inhibition of lysozyme activity was observed in the digestive gland in the D condition. Previous studies have shown inhibition of lysozyme activity or lysozyme gene expression in organisms exposed to hydrocarbons (Boutet et al., 2004; Gopalakrishnan et al., 2009). However, in this study no lysozyme inhibition was observed in the presence of
4114
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 0 3 e4 1 1 8
Fig. 7 e Lysozyme activity in the digestive gland and the mantle in C. gigas after 2 days of exposure to chemical dispersion (CD), mechanical dispersion (MD), water soluble fraction (WSF) and dispersant (D) treatments, followed by 15 days of depuration. Data are expressed as mean ± SD U mg protL1, n [ 9 (i.e. 3 sample replicates from 3 experiment replicates). For clarity of results, only significant differences between control and treatment and for a given time (after 2 d of exposure or after 15 d of depuration) are shown; *p < 0.05, **p < 0.01, ***p < 0.001.
hydrocarbons, except in the digestive gland for the D condition. Activation of mantle lysozyme activity was observed during the depuration period. Although generally pollutant exposure has shown to inhibit lysozyme activity in bivalves (for review, see Giron-Perez, 2010), some studies have shown an induction of lysozyme activity (e.g., Oliver et al., 2003; Ordas et al., 2003; Hannam et al., 2009). However, to our knowledge, no studies have shown the modulation of lysozyme activity during the depuration period. The activation of this enzyme activity suggests a potential immunostimulation during the depuration period. Uptake of xenobiotics from seawater is generally accomplished by filtration over the gill structure, although diffusion through the tegument may also contribute to tissue concentrations (Landrum and Stubblefield, 1991). As the organism directs seawater over the gill surface to extract oxygen, hydrophobic contaminants are very efficiently taken up because of the large surface area and lipid-rich membranes. Moreover, bivalves possess an open circulatory system and are thus continually exposed to fluctuations of environmental factors including contaminants (Cheng, 1981). As a general trend, enzyme activities in the plasma and gills were generally inhibited, especially in the CD condition (Figs. 2e7), suggesting that compartments that are the more exposed to the marine environment, such as the gills and plasma, are likely to be more affected by the presence of hydrocarbons, in comparison to tissues less exposed to the marine environment, such as the digestive gland, mantle and HLS (Cheng, 1981; Frouin et al., 2007). Enzyme activities in the digestive gland were less modulated than in the other tissues but were strongly correlated with PAH contents in this tissue, highlighting its role in xenobiotic detoxification (Chafai-El Alaoui, 1994; Chu et al., 2003).
Enzyme activities in the mantle and haemocytes, which are known to play an important role in defence mechanisms in oysters (Cheng, 1981), were generally activated in comparison to the control. Among tissues that were analysed, haemolymph can be considered as a key tissue because this fluid irrigates the whole body and, therefore, it can distribute contaminants and/or their metabolites throughout the organism (Cheng, 1981). Interestingly, enzyme activities differed notably from the plasma and HLS of the haemolymph. These differences could be due to alterations in the membrane integrity by PAHs and/or their metabolites, as suggested by in vitro and in vivo investigations in the blue mussel Mytilus edulis (Grundy et al., 1996). The extent of membrane alterations could be dependent on the physical (e.g. linear versus angular or branched configuration of isomers) and/or chemical (e.g. low molecular weight with high solubility versus high molecular weight with low solubility) properties of the compounds. In addition, modulation of enzyme activities in the plasma can be attributed to 1) normal mechanisms, such as secretion or 2) pathological features, such as cell lysis. Indeed, PAHs can cause cytolysis in haemocytes (McCormick-Ray, 1987), due to 1) a depletion or stimulation of metabolites or coenzymes, 2) an inhibition or stimulation of enzymes and other specific proteins, 3) an activation of a xenobiotic to a more toxic molecular species, or 4) membrane disturbances (for review, see Moore, 1985). Cytolysis may lead to 1) an increase in haemocyte number indicating compensation for cell lysis and/or 2) the release of cell contents in the plasma and consequently, a significant increase in plasma enzyme activities. Alternatively, direct induction or repression by PAHs of humoral factors, such as lysozyme (Luna-Gonzalez et al., 2004) and pro-phenoloxidase (So¨derha¨ll and Cerenius, 1998), may contribute to the modulation of enzyme activities found in the plasma.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 0 3 e4 1 1 8
4115
Fig. 8 e Principal component analysis (PCA) of enzyme activities and PAH body burdens for the gills, digestive gland, mantle, plasma and HLS. Eigen values of the first two axes are 5.54 and 1.57, for the gills, 5.30 and 2.30, for the digestive gland, 4.45 and 2.19, for the mantle, 4.99 and 2.03, for the plasma, and, 4.31 and 2.88, for the HLS, respectively.
4.3. Effect of chemical dispersion (CD) and dispersant (D) conditions Increased activity of an enzyme involved in defence mechanisms can be interpreted as a response of the organism to protect itself against a non-self molecule, e.g. xenobiotics. Persistent or excessive activation can lead to overstimulation of the immune system, which can be detrimental to the organism. Inhibition of the enzyme activity can be interpreted as saturation of the system because of the presence of a xenobiotic, overpowering of antioxidant enzymes by oxygen radicals, or an immunodeficiency resulting in direct or an indirect inhibition of mechanisms that modulate this enzyme activity (Huggett et al., 1992). In both cases, changes in defencive enzymes may affect the survival of the organisms when challenged with infectious pathogens (Thiagarajan et al., 2006). Twenty nine enzymatic analyses were carried out in this study. As a general trend and relative to the control, CD modulated a higher number of enzyme activities than D following the exposure period. Indeed, an equal or greater effect was observed in 13/18 (i.e. 13 out of 18) enzyme activities modulated by CD and/or D: 5/5, 2/3, 3/3, 3/3 and 0/4 enzyme
activities modulated by CD and/or D in the gills, digestive gland, mantle, plasma and HLS, respectively (Figs. 2e7). When comparing CD and MD conditions, following the exposure period, CD modulated a higher number of enzyme activities than MD, relative to the control condition, with an equal or greater effect observed in 13/19 activities modulated by CD and/or MD: 4/5, 2/2, 3/4, 4/5 and 0/4 enzyme activities modulated by CD and/or MD in the gills, digestive gland, mantle, plasma and HLS, respectively (Figs. 2e6). Moreover, CD exerted an equal or greater effect in a higher number of enzyme activities in the gills and plasma in comparison to other tissues, i.e. 4/5 enzyme activities modulated by CD and/or MD (Figs. 2e6), suggesting that the gills and plasma are sensitive compartments in C. gigas. Moreover, some enzyme activities that were modulated by the CD treatment following the exposure period were also modulated following the depuration period, e.g. SOD activity in the gills, mantle and plasma (Fig. 2), or laccase activity in the plasma (Fig.6), suggesting that CD may exert long-term effects. Importantly, D also modulated enzyme activities, especially at the end of the depuration period, e.g. SOD activity in the plasma and HLS (Fig. 2), catalase activity in the digestive gland (Fig. 3), GPx activity in the plasma (Fig. 4), catecholase activity in the
4116
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 0 3 e4 1 1 8
Table 2 e Relationships between enzyme activities in tissues, plasma and haemocyte lysate supernatant (HLS), and body burdens of the various PAH categories measured in the whole organism. Gills
Digestive gland
Mantle
Plasma
Significance of enzyme activities-PAH body burden relationship (RDA; Monte Carlo test, p-value) 0.002 0.009 0.001 0.021 Amount of total variation in enzyme activities explained by PAH body burdens (%) 48.6 42.5 43.3 38.7 Correlation with RDA 1st axis HMW PAHs 0.778 0.822 0.382 0.673 LMW PAHs 0.631 0.756 0.620 0.776 Parent PAHs 0.871 0.618 0.873 0.623 Alkylated PAHs 0.630 0.758 0.380 0.774 Trend with increasing body burden of the various PAH categories, i.e. HMW, LMW, parent and alkylated PAHs SOD b b b b Catalase a b a a GPx a a b a Catecholase a b b b Laccase b b a a Lysozyme a b b a
HLS 0.025 43.5 0.183 0.369 0.216 0.373 a (b)a b (a) a (b) a (b) a (b) No data
a The arrows between parenthesis show the trend associated with body burden of parent PAHs.
plasma and HLS (Fig. 5), laccase activity in the gills and digestive gland (Fig. 6) and lysozyme activity in the mantle (Fig. 7), suggesting that the dispersant used in the present study could also induce long-term effects, i.e. following 15 days of depuration. Our results are thus in agreement with previous studies that have shown the effects of other third generation dispersants on biological functions in marine invertebrates (Gilfillan et al., 1984; Shafir et al., 2007). Differences between oysters from the control and the treatment conditions resided only in the presence of oil and/ or dispersant in the water column (i.e. experimental devices and acclimatization conditions were identical in the control and the treatment conditions). Additionally, the D treatment (i.e. in the absence of oil) induced effects in enzyme activities such as laccase in the gills, SOD, catalase and lysozyme in the digestive gland, and SOD in the plasma, following the exposure or the depuration periods. Therefore, significant differences in biological responses between the control and the different treatments could not be considered as specific to contamination by hydrocarbons, but rather as indicators of unspecific stress in oysters, induced by chemicals (oil and/or dispersant) that have entered the organism from the water column. It is important to notice that transient effects were observed for different enzyme activities and thus longer periods of exposure should be studied. Moreover, not all the enzyme activities returned to control levels after the 15-day depuration period and thus longer depuration periods could give better insight into long-term effects.
with multiple enzymes involved in crucial biological responses. Enzyme activities were generally inhibited in the gills and plasma, while they were generally activated in the mantle and haemocytes, suggesting that the gills and plasma are sensitive compartments in C. gigas and that the mantle and haemocytes are likely to play an important role in protection against PAHs. Among the various parameters and during the exposure period, enzyme activities in the digestive gland, mantle and haemocytes were generally positively correlated with PAH body burdens. Enzyme activities in the gills and plasma were generally negatively correlated with PAH body burdens, suggesting potential suppressive effects of pollutants in immune and/or detoxification mechanisms through the inhibition of enzyme activities involved in these biological responses. Finally, the presence of dispersants or of chemically dispersed oil modulates different biological responses in C. gigas. These results raise questions as to the potential effects of chemically dispersed oil in nearshore areas on immune and/or detoxification responses for this estuarine species, such as in the case of the Deepwater Horizon oil spill, in which a high percentage of oil was dispersed in the water column, with a large number of unknowns on the long-term impact of the oiledispersant association.
Acknowledgements 5.
Conclusions
Responses of enzyme activities involved in immune and detoxification mechanisms in juveniles of the Pacific oyster C. gigas were highly variable depending on the treatment, the time and the tissue that was studied, highlighting the importance of carrying out studies in different tissues and
This study was supported by a PhD grant for A. Luna-Acosta from the Conseil Ge´ne´ral de la Charente-Maritime. The Programme Ecotechnologies et De´veloppement Durable (PRECODD) of the Agence Nationale de la Recherche (ANR) and especially M. Girin and G. Le Lann are acknowledged for financial support for the project ‘DISCOBIOL’ (“Dispersants et technique de lutte en milieux coˆtiers : effets biologiques et
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 0 3 e4 1 1 8
apports a` la re´glementation”), managed by F. X. Merlin. The authors also acknowledge Total Fluides for providing chemicals. Special thanks go to M. Menguy and M. Pommier for their help and assistance during the study and during the experimental procedures. The authors thank also S. Ferguson (Alba Traduction, Quimper, France) for her revision of the English language. Finally, the authors specially acknowledge Dr. T. Caquet and Dr. P-G. Sauriau for their help and assistance for statistical analysis.
Appendix. Supplementary material Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.watres.2011.05.011.
references
Anderson, J., Neff, J., Cox, B., Tatum, H., Hightower, G., 1974. Characteristics of dispersions and water-soluble extracts of crude and refined oils and their toxicity to estuarine crustaceans and fish. Marine Biology 27, 75e88. ASTM Committee, 1984. Oil spill chemical dispersants: Research, experience, and recommendations: A symposium. ASTM International, p. 465. Baca, B., Ward, G., Lane, C., Schuler, P., 2006. Net Environmental Benefit Analysis (NEBA) of dispersed oil on near shore tropical ecosystems derived from the 20 years ’tropics’ field study. In: Proceedings of the International Oil Spill Conference, London. Bado-Nilles, A., Quentel, C., Auffret, M., Le Floch, S., Gagnaire, B., Renault, T., Thomas-Guyon, H., 2008. Immune effects of HFO on European sea bass, Dicentrarchus labrax, and Pacific oyster, Crassostrea gigas. Ecotoxicology and Environmental Safety 72, 1446e1454. Bado-Nilles, A., Renault, T., Faury, N., Le Floch, S., Quentel, C., Auffret, M., Thomas-Guyon, H., 2010. In vivo effects of LCO soluble fraction on immune-related functions and gene transcription in the Pacific oyster, Crassostrea gigas (Thunberg). Aquatic Toxicology 97 (3), 196e203. Baumard, P., Budzinski, H., Garrigues, P., 1997. Analytical procedure for the analysis of PAHs in biological tissues by gas chromatography coupled to mass spectrometry: Application to mussels. Fresenius’ Journal of Analytical Chemistry 359 (6), 502e509. Baussant, T., Sanni, S., Skadsheim, A., Jonsson, G., Borseth, J., Gaudebert, B., 2001. Bioaccumulation of polycyclic aromatic compounds: 2. Modeling bioaccumulation in marine organisms chronically exposed to dispersed oil. Environmental Toxicology and Chemistry 20 (6), 1185e1195. Berthou, F., Baloue¨t, G., Bodennec, G., Marchand, M., 1987. The occurrence of hydrocarbons and histopathological abnormalities in oysters for seven years following the wreck of the Amoco Cadiz in Brittany (France). Marine Environmental Research 23 (2), 103e133. Borg, D., Schaich, K., 1983. Cytotoxicity from coupled redox cycling of autoxidizing xenobiotics and metals. Israel Journal of Chemistry 24 (1), 38e53. Boutet, I., Tanguy, A., Moraga, D., 2004. Response of the Pacific oyster Crassostrea gigas to hydrocarbon contamination under experimental conditions. Gene 329 (1), 147e157. Canevari, G., 1973. Development of the ’Next Generation’ Chemical Dispersants, Conference on Prevention and Control of Oil Spills, Washington DC, pp. 231e240.
4117
Cazenave, J., Bistoni, M., Zwirnmann, E., Wunderlin, D., Wiegand, C., 2006. Attenuating effects of natural organic matter on microcystin toxicity in zebra fish (Danio rerio) embryos - benefits and costs of microcystin detoxication. Environmental Toxicology 21 (1), 22e32. Chafai-El Alaoui, S., 1994. Etude de quelques aspects de la contamination me´tallique chez certains mollusques et poissons marins: Etude environnementale (coˆte atlantique marocaine) et approche expe´rimentale. Faculte´ des Sciences Semlalia, Cadi Ayyad, Marrakech, p. 155. Cheng, T., 1981. Invertebrate Blood Cells. Academic Press, London, pp. 233e300. Cheung, C.C.C., Zheng, G.J., Li, A.M.Y., Richardson, B.J., Lam, P.K.S., 2001. Relationships between tissue concentrations of polycyclic aromatic hydrocarbons and antioxidative responses of marine mussels, Perna viridis. Aquatic Toxicology 52 (3e4), 189e203. Chu, F., Soudant, P., Hale, R., 2003. Relationship between PCB accumulation and reproductive output in conditioned oysters Crassostrea virginica fed a contaminated algal diet. Aquatic Toxicology, 65293e65307. Dodor, D.E., Hwang, H.-M., Ekunwe, S.I.N., 2004. Oxidation of anthracene and benzo[a]pyrene by immobilized laccase from Trametes versicolor. Enzyme and Microbial Technology 35 (2e3), 210e217. Fossati, P., Prencipe, L., Berti, G., 1980. Use of 3,5-dichloro-2hydroxybenzenesulfonic acid/4-aminophenazone chromogenic system in direct enzymic assay of uric acid in serum and urine. Clinical Chemistry 26, 227e231. Frouin, H., Pellerin, J., Fournier, M., Pelletier, E., Richard, P., Pichaud, N., Rouleau, C., Garnerot, F., 2007. Physiological effects of polycyclic aromatic hydrocarbons on soft-shell clam Mya arenaria. Aquatic Toxicology 82 (2), 120e134. Gilfillan, E., Page, D., Hanson, S., Foster, J., Hotham, J., Vallas, D., Gerber, R., 1984. Effects of test spills of chemically dispersed and nondispersed oil on the activity of aspartate amino-transferase and glucose-6-phosphate dehydrogenase in two intertidal bivalves, Mya arenaria and Mytilus edulis. In: Oil Spill Chemical Dispersants: Research, Experience, and Recommendations: A Symposium. ASTM International, pp. 299e313. Giron-Perez, M.I., 2010. Relationships between innate immunity in bivalve molluscs and environmental pollution. Invertebrate Survival Journal 7, 149e156. Gopalakrishnan, S., Thilagam, H., Huang, W.-B., Wang, K.-J., 2009. Immunomodulation in the marine gastropod Haliotis diversicolor exposed to benzo(a)pyrene. Chemosphere 75 (3), 389e397. Grundy, M., Moore, M., Howell, S., Ratcliffe, N., 1996. Phagocytic reduction and effects on lysosomal membranes by polycylic aromatic hydrocarbons, in haemocytes of Mytilus edulis. Aquatic Toxicology 34, 273e290. Hannam, M.L., Bamber, S.D., Moody, J.A., Galloway, T.S., Jones, M.B., 2009. Immune function in the Arctic scallop, Chlamys islandica, following dispersed oil exposure. Aquatic Toxicology 92 (3), 187e194. Huggett, R., Kimerle, R., Mehrle Jr., P., 1992. Biomarkers: Biochemical, physiological and histological markers of anthropogenic stress. Lewis, Boca Raton, p. 347. Kappus, H., 1985. Lipid peroxidation: mechanisms, analysis, enzymology and biological relevance. In: Sies, H. (Ed.), Oxidative stress. Academic Press, London, pp. 273e310. Lacoste, A., Jalabert, F., Malham, S.K., Cueff, A., Poulet, S.A., 2001. Stress and stress-induced neuroendocrine changes increase the susceptibility of juvenile oysters (Crassostrea gigas) to Vibrio splendidus. General and Comparative Endocrinology 67 (5), 2304e2309. Landrum, P., Stubblefield, C., 1991. Role of respiration in the accumulation of organic xenobiotics by the amphipod, Pontoporeia hoyi. Environmental Toxicology and Chemistry 10 (1), 1019e1028.
4118
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 0 3 e4 1 1 8
Leps, J.S., Smilauer, P., 1999. Multivariate analysis of ecological data: Course material, Ceske Budejovice, p. 102. Lessard, R.R., DeMarco, G., 2000. The significance of oil spill dispersants. Spill Science and Technology Bulletin 6 (1), 59e68. Li, M., Garrett, C., 1998. The relationship between oil droplet size and upper ocean turbulence. Marine Pollution Bulletin 36 (12), 961e970. Luna-Acosta, A., Bustamante, P., Godefroy, J., FruitierArnaudin, I., Thomas-Guyon, H., 2010a. Seasonal variation of pollution biomarkers to assess the impact on health status of juvenile Pacific oysters Crassostrea gigas exposed in situ. Environmental Science and Pollution Research 17 (4), 999e1008. Luna-Acosta, A., Rosenfeld, E., Amari, M., Fruitier-Arnaudin, I., Bustamante, P., Thomas-Guyon, H., 2010b. First evidence of laccase activity in the Pacific oyster Crassostrea gigas. Fish and Shellfish Immunology 28 (4), 719e726. Luna-Gonzalez, A., Maeda-Martinez, A., Ascencio-Valle, F., Robles-Mungaray, M., 2004. Ontogenic variations of hydrolytic enzymes in the Pacific oyster Crassostrea gigas. Fish and Shellfish Immunology 16, 287e294. Lunel, T., 1995. The Braer Oil Spill: Oil Fate Governed by Dispersion. International Oil Spill Conference, Long Beach, California, USA. Marchand, R., 1998. Marketing of solid waste management services in Tingloy, the Philippines UWEP Working Document 9. McCormick-Ray, M., 1987. Hemocytes of Mytilus edulis affected by Prudhoe Bay crude oil emulsion. Marine Environmental Research 22 (2), 107e122. McHenery, J.G., Allen, J.A., Birkbeck, T.H., 1986. Distribution of lysozyme-like activity in 30 bivalve species. Comparative Biochemistry and Physiology B 85 (3), 581e584. McLeese, D.W., Burridge, L.E., 1987. Comparative accumulation of PAHs in four marine invertebrates. In: CapuzzoJ., KesterD (Eds.), Oceanic Processes in Marine Pollution. Robert E. Krieger Publishing Company, Malobar, Florida, pp. 109e118. Meador, J., 2003. Bioacccumulation of PAHs in marine invertebrates. In: Douben, P. (Ed.), PAHs: An ecological perspective. John Wiley and son, NY, USA, p. 392. Michel, J., Henry, C.B., 1997. Oil uptake and depuration in oysters after use of dispersants in shallow water in El Salvador. Spill Science and Technology Bulletin 4 (2), 57e70. Moore, M.N., 1985. Cellular responses to pollutants. Marine Pollution Bulletin 16 (4), 134e139. NCR (National Research Council), 2003. Oil in the Sea III: Inputs, Fates and Effects. The National Academies Press, Washington DC, p. 265. Neff, J.M., 2002. Bioaccumulation in marine organisms: Effect of contaminants from oil well produced water. Elsevier Science, Oxford, p. 452. Neumann, N.F., Stafford, J.L., Barreda, D., Ainsworth, A.J., Belosevic, M., 2001. Antimicrobial mechanisms of fish phagocytes and their role in host defense. Developmental and Comparative Immunology 25 (8e9), 807e825. Oliver, L.M., Fisher, W.S., Volety, A.K., Malaeb, Z., 2003. Greater hemocyte bactericidal activity in oysters (Crassostrea virginica) from a relatively contaminated site in Pensacola Bay, Florida. Aquatic Toxicology 64, 363e373. Orbea, A., Ortiz-Zarragoitia, M., Sole´, M., Porte, C., Cajaraville, M.P., 2002. Antioxidant enzymes and peroxisome proliferation in relation to contaminant body burdens of PAHs and PCBs in bivalve molluscs, crabs and fish from the Urdaibai and Plentzia estuaries (Bay of Biscay). Aquatic Toxicology 58 (1e2), 75e98. Ordas, M.C., Albaiges, J., Bayona, J.M., Ordas, A., Figueras, A., 2003. Assessment of in vivo effects of the Prestige fuel oil spill on
the Mediterranean mussel immune system. Archives of Environmental Contamination and Toxicology 52, 200e206. Page, C.A., Bonner, J.S., Sumner, P.L., McDonald, T.J., Autenrieth, R.L., Fuller, C.B., 2000. Behavior of a chemicallydispersed oil and a whole oil on a near-shore environment. Water Research 34 (9), 2507e2516. Perdue, J., Beattie, J., Chew, K., 1981. Some relationships between gametogenic cycle and summer mortality phenomenon in the Pacific oyster (Crassostrea gigas) in Washington State. Journal of Shellfish Research 1 (1), 9e16. Ramachandran, S.D., Hodson, P.V., Khan, C.W., Lee, K., 2004. Oil dispersant increases PAH uptake by fish exposed to crude oil. Ecotoxicology and Environmental Safety 59 (3), 300e308. Richardson, B.J., Mak, E., De Luca-Abbott, S.B., Martin, M., McClellan, K., Lam, P.K.S., 2008. Antioxidant responses to polycyclic aromatic hydrocarbons and organochlorine pesticides in green-lipped mussels (Perna viridis): do mussels "integrate" biomarker responses? Marine Pollution Bulletin 57 (6e12), 503e514. Roy, G., Vuillemin, R., Guyomarch, J., 2005. On-site determination of polynuclear aromatic hydrocarbons in seawater by stir bar sorptive extraction (SBSE) and thermal desorption GC-MS. Talanta 66, 540e546. Shafir, S., Van Rijn, J., Rinkevich, B., 2007. Short- and long-term toxicity of crude oil and oil dispersants to two representative coral species. Environmental Science and Technology 41 (15), 5571e5574. Singer, M., Aurand, D., Bragin, G., Clark, J., Coelho, G., Sowby, M., Tjeerdema, R., 2000. Standardization of the preparation and quantitation of water-accommodated fractions of petroleum for toxicity testing. Marine Pollution Bulletin 40 (11), 1007e1016. Smith, P., Krohn, R., Hermanson, G., Mallia, A., Gartner, F., Provenzano, M., Fujimoto, E., Goeke, N., Olson, B., Klenk, D., 1985. Measurement of protein using bicinchoninic acid. Analytical Biochemistry, 15076e15085. So¨derha¨ll, K., Cerenius, L., 1998. Role of the prophenoloxidaseactivating system in invertebrate immunity. Current Opinion in Immunology 10 (1), 23e28. Sole, M., Porte, C., Albaiges, J., 1995. The use of biomarkers for assessing the effects of organic pollution in mussels. Science of the Total Environment 159 (2e3), 147e153. Stabili, L., Pagliara, P., 2009. Effect of zinc on lysozyme-like activity of the seastar Marthasterias glacialis (Echinodermata, Asteroidea) mucus. Journal of Invertebrate Pathology 100 (3), 189e192. Thiagarajan, R., Gopalakrishnan, S., Thilagam, H., 2006. Immunomodulation in the marine green mussel Perna viridis exposed to sub-lethal concentrations of Cu and Hg. Archives of Environmental Contamination and Toxicology 51 (3), 392e399. Tryphonas, H., Fournier, M., Blakley, B., Smits, J., Brousseau, P., 2005. Investigative Immunotoxicology. Informa Healthcare, London, p. 480. Venosa, A.D., Zhu, X., 2003. Biodegradation of crude oil contaminating marine shorelines and freshwater wetlands. Spill Science and Technology Bulletin 8 (2), 163e178. Verlecar, X.N., Jena, K.B., Chainy, G.B.N., 2007. Biochemical markers of oxidative stress in Perna viridis exposed to mercury and temperature. Chemico-Biological Interactions 167 (3), 219e226. Winston, G.W., 1991. Oxidants and antioxidants in aquatic animals. Comparative Biochemistry and Physiology C 100 (1e2), 173e176. Zar, J., 1984. Biostatistical Analysis. Prentice-Hall, New Jersey, p. 718.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 1 9 e4 1 3 0
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Degradation of pharmaceutical beta-blockers by electrochemical advanced oxidation processes using a flow plant with a solar compound parabolic collector Eloy Isarain-Cha´vez, Rosa Marı´a Rodrı´guez, Pere Lluı´s Cabot, Francesc Centellas, Conchita Arias, Jose´ Antonio Garrido, Enric Brillas* Laboratori d’Electroquı´mica de Materials i del Medi Ambient, Departament de Quı´mica Fı´sica, Facultat de Quı´mica, Universitat de Barcelona, Martı´ i Franque`s 1-11, 08028 Barcelona, Spain
article info
abstract
Article history:
The degradation of the beta-blockers atenolol, metoprolol tartrate and propranolol
Received 24 February 2011
hydrochloride was studied by electro-Fenton (EF) and solar photoelectro-Fenton (SPEF).
Received in revised form
Solutions of 10 L of 100 mg L1 of total organic carbon of each drug in 0.1 M Na2SO4 with
9 May 2011
0.5 mM Fe2þ of pH 3.0 were treated in a recirculation flow plant with an electrochemical
Accepted 21 May 2011
reactor coupled with a solar compound parabolic collector. Single Pt/carbon felt (CF) and
Available online 31 May 2011
boron-doped diamond (BDD)/air-diffusion electrode (ADE) cells and combined Pt/ADEePt/ CF and BDD/ADEePt/CF cells were used. SPEF treatments were more potent with the latter
Keywords:
cell, yielding 95e97% mineralization with 100% of maximum current efficiency and energy
Atenolol
consumptions of about 0.250 kWh g TOC1. However, the Pt/ADEePt/CF cell gave much
Metoprolol
lower energy consumptions of about 0.080 kWh g TOC1 with slightly lower mineralization
Propranolol
of 88e93%, then being more useful for its possible application at industrial level. The EF
Electro-Fenton
method led to a poorer mineralization and was more potent using the combined cells by
Photoelectro-Fenton
the additional production of hydroxyl radicals (OH) from Fenton’s reaction from the fast
Solar photo-assisted process
Fe2þ regeneration at the CF cathode. Organics were also more rapidly destroyed at BDD than at Pt anode. The decay kinetics of beta-blockers always followed a pseudo first-order reaction, although in SPEF, it was accelerated by the additional production of OH from the action of UV light of solar irradiation. Aromatic intermediates were also destroyed by hydroxyl radicals. Ultimate carboxylic acids like oxalic and oxamic remained in the treated solutions by EF, but their Fe(III) complexes were photolyzed by solar irradiation in SPEF, thus explaining its higher oxidation power. NO3 was the predominant inorganic ion lost in EF, whereas the SPEF process favored the production of NH4þ ion and volatile N-derivatives. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
Indirect electro-oxidation methods based on H2O2 electrogeneration such as electro-Fenton (EF) and photoelectroFenton (PEF) have recently received great attention for the
remediation of acidic wastewaters with toxic and/or bio¨ zcan et al., 2008; Brillas et al., refractory organic pollutants (O 2009). The EF process involves the continuous supply of H2O2 to the acidic contaminated water of the cell from the twoelectron reduction of injected O2 at the cathode by reaction
* Corresponding author. Tel.: þ34 93 4021223; fax: þ34 93 4021231. E-mail address:
[email protected] (E. Brillas). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.05.026
4120
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 1 9 e4 1 3 0
(1), which is very efficient using carbonaceous cathodes like carbon felt (CF) (Oturan and Oturan, 2005; Diagne et al., 2007; Hammami et al., 2008) and carbon-PTFE (Sire´s et al., 2006; Flox et al., 2007; Skoumal et al., 2008) and carbon nanotubesPTFE (Khataee et al., 2010; Zarei et al., 2010) gas (O2 or air) diffusion electrodes: O2 þ 2Hþ þ 2e /H2 O2
(1)
Addition of a small amount of Fe2þ to the medium as catalyst enhances the oxidation power of H2O2, because they form Fe(OH)2þ, the predominant Fe3þ species at pH 2.8e3.5, and hydroxyl radical (OH) from Fenton’s reaction (2) (Sun and Pignatello, 1993): Fe2þ þ H2 O2 /FeðOHÞ2þ þ OH
(2)
OH is a powerful oxidant (E ( OH/H2O) ¼ 2.80 V/SHE) that non-selectively reacts with organic pollutants giving dehydrogenated or hydroxylated derivatives until mineralization, i.e., transformation into CO2, water and inorganic ions (Boye et al., 2006; Balci et al., 2009). An advantage of EF compared with the chemical Fenton treatment is the regeneration of Fe2þ from Fe3þ reduction at the cathode by reaction (3), propagating Fenton’s reaction (2) (Diagne et al., 2007; Oturan et al., 2008): Fe3þ þ e /Fe2þ
(3)
When an undivided cell is employed with a high O2 overpotential anode such as boron-doped diamond (BDD), pollutants are oxidized by OH formed in the bulk from Fenton’s reaction (2) and heterogeneous hydroxyl radical (BDD(OH)) produced from water oxidation at the anode surface (Marselli et al., 2003; Can˜izares et al., 2008; Panizza and Cerisola, 2009): BDD þ H2 O/BDDð OHÞ þ Hþ þ e
(4)
BDD is the anode with higher oxidation power known because it interacts very weakly with generated OH promoting a much greater O2 overpotential than conventional anodes like PbO2 and Pt and enhancing organic removal with reactive BDD(OH) (Sire´s et al., 2008; Hamza et al., 2009; Panizza and Cerisola, 2009; Skoumal et al., 2009). In PEF, the electrolyzed solution is simultaneously submitted to an UV irradiation supplied by an artificial lamp (Sire´s et al., 2007; Skoumal et al., 2008; Khataee et al., 2010; Zarei et al., 2010). The complex synergistic action of UV light can be ascribed to: (i) the greater Fe2þ regeneration and OH production by photolysis of Fe(OH)2þ from reaction (5) (Sun and Pignatello, 1993; Brillas et al., 2009) and (ii) the photodecarboxylation of complexes of Fe(III) with most generated carboxylic acids by the general reaction (6) (Zuo and Hoigne´, 1992): FeðOHÞ2þ þhv/Fe2þ þ OH
(5)
FeðOOCRÞ2þ þhv/Fe2þ þ CO2 þ R
(6)
The excessive energy cost of the artificial UV light limits the possible application of PEF to wastewater treatment. To solve this problem, our group has recently proposed the alternative use of the solar photoelectro-Fenton (SPEF) process in which the UV irradiation of sunlight (l > 300 nm) is utilized as
inexpensive and renewable energy source (Flox et al., 2007; Skoumal et al., 2009; Ruiz et al., 2011). However, more fundamental research is still needed to show that SPEF can be useful for wastewater remediation. Recently, low contents of a high number of pharmaceutical drugs have been detected in surface, ground and drinking waters as a result of their inefficient destruction in sewage treatment plants (Maurer et al., 2007; Vieno et al., 2007; Ku¨mmerer, 2009; Ramil et al., 2010). In the aquatic environment, drugs like beta-blockers are dangerous because they affect the endocrine system of fishes and exert toxic effects on algae and invertebrates (Andreozzi et al., 2003; Cleuvers, 2005; Fent et al., 2006; Owen et al., 2007) and hence, powerful oxidation methods are needed to be developed to ensure the total removal of these contaminants and their metabolites from wastewaters. Several authors have described the destruction of beta-blockers by ozonation in neutral and alkaline media (Benner et al., 2008; Rosal et al., 2008; Benı´tez et al., 2009), O3/H2O2 (Rosal et al., 2008), radiolysis (Song et al., 2008), UV and UV/H2O2 (Kim et al., 2009) and a biological Fenton-like system mediated by the white-rot fungus Trametes versicolor (Marco-Urrea et al., 2010). Sire´s et al. (2010) reported the mineralization of a mixture with 0.15 mM of beta-blockers like atenolol, metoprolol and propranolol at pH 3 by EF with a Pt/carbon felt (CF) cell and high current. In our laboratory, we have tested the degradation of 100 mL of atenolol (Isarain-Cha´vez et al., 2010a), metoprolol (IsarainCha´vez et al., 2011b) and propranolol (Isarain-Cha´vez et al., 2010b, 2011a) solutions by EF and PEF. The formula of these beta-blockers is shown in Fig. 1. A novel configuration of two cells combined in parallel with a BDD/air-diffusion electrode (ADE) and a Pt/CF pair, was found more efficient for both EAOPs than single Pt/ADE or BDD/ADE cells, because of the larger generation of OH from Fenton’s reaction (2) by the quick Fe2þ regeneration from Fe3þ reduction at the CF cathode by reaction (3). From these results, we have designed and built-up a 10 L recirculation flow plant containing a filter-press reactor with the above two or four electrodes, coupled with a compound parabolic collector (CPC) directly exposed to sunlight. Our aim was to show the viability of SPEF for the
Fig. 1 e Chemical structure of (a) atenolol, (b) metoprolol and (c) propranolol.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 1 9 e4 1 3 0
possible treatment of wastewaters with beta-blockers at industrial level by determining its oxidation ability, mineralization current efficiency and energy consumption in such plant. This paper presents a study on the SPEF treatments of 10 L of solutions with 100 mg L1 of total organic carbon (TOC) of atenolol, metoprolol tartrate (2:1) and propranolol hydrochloride using a recirculation flow plant with single or combined cells coupled with the CPC photoreactor. Comparative EF treatments in the dark were also made to clarify the oxidation power of the different cells tested. The decay kinetics of the drugs, as well as the evolution of aromatic intermediates, generated carboxylic acids and released inorganic ions, was determined by chromatographic techniques to clarify the effect of solar irradiation.
2.
Materials and methods
2.1.
Chemicals
Atenolol, metoprolol tartrate and propranolol hydrochloride, all them of 99% purity, were supplied by the pharmaceutical AstraZeneca Espan˜a (Madrid, Spain). p-Benzoquinone, 4-hydroxyphenylacetamide, 4-(2-methoxyethyl)phenol and phthalic acid were of reagent grade from SigmaeAldrich. Reactive reagent 1-naphthol was from BDH Chemical Ltd. Oxalic and oxamic acids were of analytical grade from Avocado. Reagent grade 33% (w/v) H2O2 was supplied by Panreac. Anhydrous sodium sulfate, used as background electrolyte, and ferrous sulfate heptahydrate, used as catalyst, were of analytical grade from Merck and Fluka, respectively. The solutions were prepared with deionized water and their pH was adjusted to 3.0 with analytical grade sulfuric acid from Merck. Organic solvents and other chemicals used were either of HPLC or analytical grade from Merck, Fluka and Avocado.
2.2.
Apparatus
The solution pH was determined with a Crison GLP 22 pHmeter. Colorimetric measurements were conducted with a Unicam UV4 UV/vis Prisma spectrophotometer thermostated at 25 C. TOC of solutions was obtained with a Shimadzu VCSN TOC analyzer. The decay of the three betablockers and the evolution of their aromatic products were monitored by reversed-phase HPLC using a Waters 600 liquid chromatograph fitted with a Spherisorb ODS2 5 mm, 150 mm 4.6 mm (i.d.), column at 35 C, coupled with a Waters 996 photodiode array detector selected at the maximum wavelength found for the UV spectrum of each compound. Carboxylic acids were detected and quantified by ion-exclusion HPLC using the above system fitted with a BioRad Aminex HPX 87H, 300 mm 7.8 mm (i.d.), column at 35 C and the photodiode array selected at l ¼ 210 nm. Ionic chromatography was carried out with a Shimadzu 10 Avp liquid chromatograph coupled with a Shimadzu CDD 10 Avp conductivity detector, using a Shodex IC YK-421, 125 mm 4.6 mm (i.d.), cation column at 40 C, for NH4þ and a ShimPack IC-A1S, 100 mm 4.6 mm (i.d.), anion column at 40 C for NO3.
2.3.
4121
Flow plant
Fig. 2a presents a scheme of the 10 L recirculation flow plant used for the EF and SPEF degradations of the beta-blockers in batch mode operating with a combined cell. The solution was introduced in the reservoir and continuously circulated through the system by a peristaltic pump at a liquid flow rate of 250 L h1 adjusted by a rotameter. The temperature was regulated at 35 C by two heat exchangers. Samples for analysis were withdrawn from the effluent in the reservoir. The electrochemical cell was a filter-press reactor with 12 cm 18 cm components, separated with Viton gaskets to avoid leakages. Four cell configurations with monopolar connection, containing either two or four alternated electrodes of 10 cm 10 cm in dimension, were utilized: (i) Pt/ADE, (ii) BDD/ ADE, (iii) combined Pt/ADEePt/CF and (iv) combined BDD/ ADEePt/CF. A scheme of the latter configuration is shown in Fig. 2b. The anodes were Pt sheets of 99.99% purity from SEMPSA and/or a BDD thin film from Adamant Technologies. The cathodes were a CF from Sofacel or a carbon-PTFE ADE from E-TEK. PVC liquid compartments with a central window of 9.5 cm 9.5 cm (90.25 cm2) allowed contacting the effluent with the outer faces of each pair of electrodes with 1.2 cm of separation. The inner face of the ADE cathode was pressed to a Ni mesh as electrical connector in contact with a PVC gas chamber circulating compressed air with a pump at a flow rate of 150 mL min1 regulated with a back-pressure gauge. The independent constant current applied to each pair of electrodes was imposed by Grelco GDL3020 and/or GVD310 power sources, which directly displayed the applied potential. The solar CPC photoreactor with an area of 0.4 m2 and concentration factor of 1 was composed of 12 borosilicate-glass tubes of 50.5 cm length 1.82 cm inner diameter (irradiated volume 1.57 L), with connecting tubing and valves mounted in an aluminum frame on a platform tilted 41 to better collect the sun rays in our laboratory of Barcelona (latitude: 41 210 N, longitude: 2 100 E). The solar trials were performed during the summer of 2009, with an average UV incident radiation of about 19 W m2, determined by the meteorological center of the Universitat de Barcelona. The EF experiments were performed in the dark by covering the plant with a black cloth. Solutions containing 100 mg L1 TOC of each beta-blocker in 0.1 M Na2SO4 with 0.5 mM Fe2þ of pH 3.0 were comparatively degraded by EF and SPEF for 360 min with the different cell configurations. Before using the plant, the BDD anode and ADE cathode were polarized under electrolysis of 10 L of 0.1 M Na2SO4 at pH 3.0 and 3 A for 240 min to remove their impurities and activate them.
2.4.
Analytical procedures
Before the analysis of aliquots withdrawn from treated solutions, they were filtered with 0.45 mm PTFE filters from Whatman. The H2O2 concentration was determined from the light absorption of its Ti(IV) colored complex at l ¼ 409 nm (Welcher, 1975). Reproducible TOC values with an accuracy of 1% were determined by injecting 50 mL aliquots into the TOC analyzer. These data allowed calculating the mineralization current efficiency (MCE) for electrolyzed solutions from Eq. (7) (Hamza et al., 2009):
4122
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 1 9 e4 1 3 0
Fig. 2 e (a) Experimental set up of the recirculation flow plant. (1) Flow electrochemical cell, (2) reservoir, (3) sampling, (4) peristaltic pump, (5) flowmeter, (6) heat exchanger, (7) solar CPC photoreactor, (8) power supply and (9) air pump. (b) Scheme of a combined filter-press cell. (1) End plate, (2) gasket, (3) air inlet, (4) air outlet, (5) air chamber, (6) boron-doped diamond (BDD) anode, (7) air-diffusion (ADE) cathode, (8) carbon felt (CF) cathode, (9) Pt anode, (10) liquid compartment, (11) liquid inlet in the cell, (12) liquid outlet of the Pt/CF pair connected to 13, (13) liquid inlet in the BDD/ADE pair and (14) liquid outlet of the cell.
MCEð%Þ ¼
n F Vs DðTOCÞexp 4:32 107 m I t
100
(7) 1
where F is the Faraday constant (96487 C mol ), Vs is the solution volume (L), DðTOCÞexp is the experimental TOC decay (mg L1), 4.32 107 is a conversion factor to homogenize units (3600 s h1 12000 mg mol1), m is the number of carbon atoms of each beta-blocker, I is the total applied current (A) and t is the electrolysis time (h). The number of electrons consumed (n) was taken as 66 for atenolol, 162 for metoprolol tartrate and 78 for propranolol hydrochloride, assuming that their overall mineralization involves the release of NH4þ ion as
main inorganic species according to reactions (8), (9) and (10), respectively: þ C14 H22 N2 O3 þ 25 H2 O/14 CO2 þ 2 NHþ 4 þ 64 H þ 66e
(8)
þ ðC15 H25 NO3 Þ2 C4 H6 O6 þ56H2 O/34CO2 þ2NHþ 4 þ160H þ162e
(9) C16 H22 NOþ 2 Cl
þ 30 H2 O/16
CO2 NHþ 4
þ
þ Cl þ 78 H þ 78 e (10)
From TOC removal, the energy consumption per unit TOC mass (EC) was determined for single cells from Eq. (11) (Ruiz et al., 2011):
4123
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 1 9 e4 1 3 0
EC kWh g TOC1
¼
Ecell I t Vs DðTOCÞexp
100
where Ecell is the average applied potential (V). For combined cells, a similar equation was applied considering the addition of the electrical potency (Ecell I ) of each pair. Reversed-phase HPLC was carried out with 20 mL aliquots and mobile phase of: (a) a 30:70 (v/v) methanol/phosphate buffer (pH 3.0), with 1.1 g of sodium heptanesulfonate and 2 mL of di-n-butylamine per liter, at 0.6 mL min1 for atenolol and (b) a 36:36:28 (v/v/v) acetonitrile/methanol/water mixture, with 2 g L1 sodium dodecyl sulfate and 0.15 M phosphoric acid of pH 3.0, at 1.5 mL min1 for the other betablockers. For ion-exclusion HPLC, 4 mM H2SO4 at 0.6 mL min1 was used. Ionic chromatography was performed with 25 mL aliquots and mobile phases of 5.0 mM tartaric acid, 1.0 mM dipicolinic acid, 24.2 mM boric acid and 1.5 mM corona ether at 1.0 mL min1 for NH4þ and 2.5 mM phthalic acid and 2.4 mM tris(hydroxymethyl)aminomethane of pH 4.0 at 1.5 mL min1 for NO3.
80
a
TOC / mg L
-1
(11)
60 40 20 0
b
175
% MCE
150 125 100 75 50 25
3.
Results and discussion
3.1. Optimization of the SPEF treatment with a single BDD/ADE cell A preliminary study was performed to check the ability of the flow plant with a single BDD/ADE cell to electrogenerate and accumulate H2O2. To do this, 10 L of 0.1 M Na2SO4 solutions of pH 3.0 and 35 C were electrolyzed for 420 min in the dark at different currents. A progressive increase in H2O2 concentration with electrolysis time and current was always observed, giving a large accumulation of this species at the end of trials. For example, 13.8 and 19.5 mM H2O2 were determined for 3.0 and 5.0 A, respectively. This is indicative of the high efficiency of the ADE cathode to continuously produce H2O2 from reaction (1), which can then be used for the removal of organic matter in the EAOPs. The oxidation power of the tested cells is expected to strongly depend on the applied current (Brillas et al., 2009; Panizza and Cerisola, 2009). To find the best operative current for the treatment of the beta-blockers by EAOPs, a series of SPEF experiments was made by electrolyzing 10 L of a 0.594 mM atenolol solution with 0.5 mM Fe2þ at pH 3.0 between 1.5 and 5.0 A for 360 min using the flow plant with a BDD/ADE cell. This reactor was chosen because previous work with 100 mL solutions showed that it had a higher oxidation ability to destroy organic pollutants than the Pt/ADE one (Isarain-Cha´vez et al., 2010a). In these trials, the solution pH remained practically constant, decaying to final values of 2.6e2.8. Fig. 3a evidences a fast and gradual TOC decay in all cases. Increasing final mineralization of 81%, 87% and 89% was achieved for 1.5, 3.0 and 5.0 A, indicating a higher oxidation power of the system as current increases. This behavior can be accounted for by the production of more amounts of BDD(OH) from reaction (4) and OH from Fenton’s reaction (2) owing to the larger generation of H2O2, as stated above. However, TOC varies slightly from 3.0 to 5.0 A, suggesting a loss in efficiency of the SPEF process. This can be
0
0
60
120
180
240
300
360
420
time / min
Fig. 3 e Effect of applied current on (a) TOC decay and (b) mineralization current efficiency calculated from Eq. (7) vs. electrolysis time for the solar photoelectro-Fenton (SPEF) degradation of 10 L of 0.594 mM atenolol in 0.1 M Na2SO4 with 0.5 mM Fe2D at pH 3.0 and 35 C using the flow plant with a single BDD/ADE cell coupled with the CPC photoreactor at a liquid flow rate of 250 L hL1. Current: (;) 1.5 A, (:) 3.0 A and (A) 5.0 A.
confirmed in Fig. 3b, where the corresponding MCE values obtained from Eq. (7) considering that atenolol mineralization follows reaction (8) are presented. As can be seen, SPEF is so potent that yields a maximum efficiency of 180% at 120 min of 1.5 A, a value that falls to 104% and 68% for 3.0 and 5.0 A, respectively. The existence of a maximum in all MCE vs. time plots suggests the initial mineralization of easily oxidizable byproducts of atenolol, followed by the destruction of other by-products that are more difficultly destroyed with hydroxyl radicals and/or UV light of solar irradiation. The decrease in efficiency with rising current shown in Fig. 3b can then be ascribed to the acceleration of non-oxidant reactions of hydroxyl radicals giving a relatively smaller quantity of organic oxidation events. These waste reactions involve mainly the primary oxidation of BDD(OH) to O2 by reaction (12), as well as the dimerization of OH to H2O2 by reaction (13) or its destruction with H2O2 by reaction (14) (Marselli et al., 2003; Sire´s et al., 2006). The relative amount of generated BDD(OH) can also be reduced by the formation of weaker oxidants such as S2O82 ion from SO42 ion of the background electrolyte by reaction (15) and ozone by reaction (16) (Flox et al., 2007; Hamza et al., 2009; Panizza and Cerisola, 2009):
4124
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 1 9 e4 1 3 0
2BDDð OHÞ/2BDD þ O2ðgÞ þ 2Hþ þ 2e
2 OH/H2 O2
H2 O2 þ
OH/HO2
(12) (13)
þ H2 O
a
100 80
(14) 60
2 2SO2 4 /S2 O8 þ 2e
(15)
3H2 O/O3 þ 6Hþ þ 6e
(16)
40 20 0
b
100
-1
80 TOC / mg L
Taking into account that the TOC abatement for the SPEF experiments of Fig. 3 is not substantially improved from 3.0 A, this current was applied to the Pt/ADE and BDD/ADE pairs in the single and combined cells for all the treatments of betablockers. On the other hand, previous work (Isarain-Cha´vez et al., 2010a) showed that a small Pt/CF cell containing 100 mL of a 0.05 M Na2SO4 and 0.5 mM Fe2þ solution at pH 3.0 was able to regenerate Fe2þ at the CF cathode from the reduction of Fe3þ up to 4 mA cm2. From this, 0.4 A was supplied to the Pt/CF pair of the combined cells in the flow plant.
60 40 20
3.2. SPEF
Comparative degradation of beta-blockers by EF and
0
c
100 1
The degradation of 100 mg L TOC of atenolol, metoprolol tartrate and propranolol hydrochloride was comparatively studied by EF and SPEF in the recirculation flow plant using single Pt/ADE and BDD/ADE cells at 3.0 A and combined Pt/ ADEePt/CF and BDD/ADEePt/CF cells at 3.0e0.4 A. Fig. 4a,b and c depict the corresponding TOC abatements for 360 min of such treatments. The percentages of TOC removal at the end of these experiments are collected in Table 1. An inspection of these data reveals that the SPEF treatments allow up to 97% mineralization of beta-blockers and are much more potent than the EF ones, where no more than 66% decontamination is reached. This evidences that UV irradiation of sunlight strongly enhances the destruction of organic matter. Fig. 4a,b and c show that the oxidation power of the EF processes increases when the electrochemical cells are used in the sequence Pt/ADE < BDD/ADE < Pt/ADEePt/CF < BDD/ ADEePt/CF. The poorest mineralization achieved for the Pt/ ADE cell (see Table 1) indicates a quite slow destruction of organics with both, Pt(OH) formed from the anodic oxidation of water at the anode and OH produced from Fenton’s reaction(2) between added Fe2þ and generated H2O2. The higher oxidation ability of BDD(OH) than Pt(OH) to attack organics (Sire´s et al., 2006; Hamza et al., 2009; Panizza and Cerisola, 2009) explains the slightly greater mineralization found for the BDD/ADE system. This behavior can also justify the superiority of the BDD/ADEePt/CF cell in relation to the Pt/ ADEePt/CF one. The much greater mineralization reached for both combined systems compared with the single ones can be associated with the generation of much higher amounts of oxidant OH in the bulk as a result of the fast regeneration of Fe2þ from Fe3þ reduction at the CF cathode, thereby strongly accelerating the destruction rate of organics. Note that Fe3þ is quickly and continuously formed from Fenton’s reaction (2) and from the oxidation of Fe2þ at the anode. These results demonstrate that the enhancement of the cathodic reduction of Fe3þ to Fe2þ from reaction (3) increases the oxidation power
80 60 40 20 0
0
60
120
180
240
300
360
420
time / min
Fig. 4 e TOC removal with electrolysis time for the electroFenton (EF) and SPEF treatments of 10 L of (a) 0.594 mM atenolol, (b) 0.246 mM metoprolol tartrate and (c) 0.521 mM propranolol hydrochloride in 0.1 M Na2SO4 with 0.5 mM Fe2D at pH 3.0 and 35 C in the recirculation flow plant. () EF in Pt/ADE cell at 3.0 A, (-) EF in Pt/ADEePt/CF cell at 3.0e0.4 A, (B) EF in BDD/ADE cell at 3.0 A, (C) EF in BDD/ ADEePt/CF cell at 3.0e0.4 A, (6) SPEF in Pt/ADEePt/CF cell at 3.0e0.4 A and (:) SPEF in BDD/ADEePt/CF cell at 3.0e0.4 A.
of the EF treatment of beta-blockers, which becomes more potent with the BDD/ADEePt/CF cell. The same trend for the cells tested was also found for the SPEF treatments of all beta-blockers, although in this case, smaller variations of the mineralization degree were obtained due to the potent degradative action of sunlight. Comparison of Figs. 3a and 4a, for example, evidences that TOC of the atenolol solution is reduced by 87%, 88% and 94% for the BDD/ ADE, Pt/ADEePt/CF and BDD/ADEePt/CF cells, respectively. From this behavior, only the results for the higher degradations obtained with the combined cells are given in Fig. 4 and Table 1. Again, the most potent SPEF method was found using
4125
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 1 9 e4 1 3 0
Table 1 e Kinetic parameters, percentages of TOC removal and energy consumptions determined for the degradation of 10 L of beta-blocker solutions in 0.1 M Na2SO4 with 0.5 mM Fe2D at pH 3.0 and 35 C by different EAOPs using the recirculation flow plant at a liquid flow rate of 250 L hL1. The current applied to the single and combined electrochemical cells was 3.0 A and 3.0e0.4 A, respectively. Compound Atenolol
Process EF
SPEF Metoprolol
EF
SPEF Propranolol
EF
SPEF
Cell
k1 103a/s1
R2b
% TOC removal at 360 min
ECc/kWh g TOC1
Pt/ADE Pt/ADEePt/CF BDD/ADE BDD/ADEePt/CF Pt/ADEePt/CF BDD/ADEePt/CF Pt/ADE Pt/ADEePt/CF BDD/ADE BDD/ADEePt/CF Pt/ADEePt/CF BDD/ADEePt/CF Pt/ADE Pt/ADEePt/CF BDD/ADE BDD/ADEePt/CF Pt/ADEePt/CF BDD/ADEePt/CF
0.31 0.36 2.24 2.43 2.48 2.46 0.13 0.18 0.22 0.74 1.91 1.53 0.09 0.15 0.13 0.91 2.37 2.14
0.992 0.999 0.996 0.997 0.988 0.996 0.991 0.988 0.989 0.985 0.985 0.988 0.992 0.980 0.981 0.983 0.992 0.980
22 26 54 61 88 94 9 22 54 66 90 97 15 35 56 65 93 95
0.333 0.280 0.412 0.386 0.084 0.250 0.754 0.291 0.421 0.359 0.072 0.244 0.463 0.224 0.405 0.354 0.085 0.240
a Pseudo first-order rate constant. b Square of regression coefficient. c Energy consumption per unit TOC mass calculated from Eq. (11) at 360 min of electrolysis.
the BDD/ADEePt/CF cell, where almost total mineralization (94e97% TOC removal) was reached for all beta-blocker solutions owing to the synergistic oxidation action of BDD(OH) at the anode, OH in the bulk and UV light supplied by solar irradiation. The relative oxidation power of the EAOPs tested to degrade the solutions of atenolol, metoprolol tartrate and propranolol hydrochloride is also reflected in the MCE values shown in Fig. 5a,b and c, respectively. As can be seen, high maximum efficiencies of 100e110% are obtained after 90e120 min of all SPEF processes, which drastically decay to about 50% at the end of treatment due to the loss of organic matter and the formation of more hardly oxidizable byproducts. In contrast, MCE values <45% can be observed for the EF treatments with combined cells, whereas the use of single cells yields much lower efficiencies (<20%) by the lower generation of OH in the bulk from Fenton’s reaction (2). For most of EF processes, however, no significant changes in efficiency occurs during treatment, suggesting that byproducts formed undergo a similar mineralization rate. The energy consumption calculated from Eq. (11) at the end of the above experiments is listed in the last column of Table 1. These data show much lower costs for the SPEF degradations than the EF ones, as expected by the greater mineralization attained in the former method. The Pt/ADE pairs yielded much lower energy costs because average potentials as low as 3.2e4.1 V were needed for 3.0 A, whereas much greater average potentials of 12.2e12.8 V were required for the BDD/ADE pairs. The average potential supplied to the Pt/CF pair in combined cells at 0.4 A varied between 2.4 and 2.8 V. So, in SPEF energy consumptions of 0.072e0.085 kWh g TOC1 (6.4e7.9 kWh m3) are determined for the Pt/ADEePt/CF cell,
which rise to 0.240e0.250 kWh g TOC1 (22.5e23.7 kWh m3) for the BDD/ADEePt/CF one (see Table 1). The above findings demonstrate that the SPEF process is potent enough to yield almost total mineralization of all betablocker solutions with very high efficiencies. Although a BDD/ ADEePt/CF cell has greater oxidation power, the alternative use of a Pt/ADEePt/CF cell is much more economic and from this point of view, more viable for the application of this method to wastewater remediation.
3.3.
Decay kinetics of beta-blockers
To better understand the synergistic action of generated oxidants (OH, Pt(OH) and/or BDD(OH)) and UV light from solar irradiation in the above trials, the decay kinetics of the three beta-blockers was analyzed by reversed-phase HPLC. These chromatograms displayed well-defined absorption peaks at retention time (tr) of 5.40 min for atenolol, 3.59 min for metoprolol and 6.02 min for propranolol. These compounds were neither attacked by generated H2O2 nor directly photolyzed by sunlight, because they underwent insignificant removal in blank chemical experiments carried out in the flow plant by recirculating the beta-blockers solutions with 20 mM H2O2 under 1 h of solar irradiation. Fig. 6a exemplifies the comparative concentration decay of metoprolol under the experimental conditions of Fig. 4b. Similar relative trends for the EAOPs tested were found for atenolol and propranolol. Note that the initial metoprolol concentration in the 0.246 mM metoprolol tartrate solution was C0 ¼ 0.492 mM since this (2:1) complex was dissociated when dissolved, as confirmed by the determination of 0.246 mM tartaric acid in such a solution by ion-exclusion
4126
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 1 9 e4 1 3 0
120
a
0.5
a [metoprolol] / mM
100 80 60 40
0.3 0.2 0.1
20
0.0
0 120
b
b 3
ln (C / C )
100 80
0
% MCE
0.4
60 40
2
1
20 0
0 100
c
0
60
120
180
240
300
360
420
time / min
Fig. 6 e (a) Decay of metoprolol concentration for the degradation of 10 L of 0.492 mM of the drug in the recirculation flow plant under the same conditions of Fig. 4. (b) Kinetic analysis assuming that metoprolol follows a pseudo first-order reaction.
80 60 40 20 0
0
60
120
180
240
300
360
420
time / min
Fig. 5 e Variation of the mineralization current efficiency calculated from Eq. (7) with electrolysis time for the EF and SPEF treatments of 10 L of (a) 0.594 mM atenolol, (b) 0.246 mM metoprolol tartrate and (c) 0.521 mM propranolol hydrochloride in 0.1 M Na2SO4 with 0.5 mM Fe2D at pH 3.0 and 35 C in the recirculation flow plant under the same conditions of Fig. 4.
chromatography. Fig. 6a shows that the EF process gives a faster removal of metoprolol using the BDD/ADE pair instead the Pt/ADE one in both single and combined cells. Although this tendency seems analogous to that of TOC removal in Fig. 4, there are significant differences due to the fast attack of BDD(OH) on the beta-blocker. This can be easily deduced from the quite similar decay of the beta-blocker in both Pt/ADEePt/CF and BDD/ADE cells, disappearing in 240 min, much more rapidly than ca. 360 min needed for the Pt/ADE one. This compound then reacts practically at the same rate with BDD(OH) in the BDD/ADE reactor as with Pt(OH) plus the additional OH produced in the Pt/ADEePt/CF one. In the combined BDD/ADEePt/CF cell, however, the metoprolol removal is strongly accelerated to disappear in only 90 min. This corroborates the large enhancement of Fenton’s reaction
(2) by the continuous Fe2þ regeneration from Fe3þ reduction at the CF cathode in the combined cells. In contrast, when the SPEF treatments are considered, the beta-blocker falls even more rapidly, but at similar rate for both combined cells, being completely removed in about 50 min. This can be accounted for by the production of high amounts of oxidant OH induced from reaction (5) under the action of UV light of solar irradiation. Comparison of metoprolol removal for Pt/ADE cell in EF and for Pt/ADEePt/CF cell in EF and SPEF indicates that OH generation is even more effective with sunlight than accelerating Fe2þ regeneration at the CF cathode. The above concentration decays were analyzed by kinetic equations related to simple reaction orders and good linear correlations were only obtained considering a pseudo firstorder reaction, as can be seen in Fig. 6b for the experiments of metoprolol of Fig. 6a. This behavior evidences that a constant amount of each generated hydroxyl radical attacks the beta-blockers during their abatement. The pseudo firstorder rate constant (k1), along with the corresponding square of regression coefficient (R2), thus determined for atenolol, metoprolol and propranolol removal by the EAOPs tested are collected in Table 1. Similar k1 values were always obtained for the two latter beta-blockers, whereas atenolol was oxidized more rapidly in all cases. Note that the attack of BDD(OH) on atenolol is so fast that the maximum oxidation power of the system is practically achieved for the BDD/ADE cell in EF, being only slightly enhanced using the combined cells in SPEF.
4127
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 1 9 e4 1 3 0
3.4. Evolution of aromatic intermediates, carboxylic acids and inorganic ions Reversed-phase chromatograms of treated solutions exhibited additional peaks related to aromatic intermediates. In each trial, all these by-products disappeared practically at the same electrolysis time, as expected if they react with OH, Pt(OH) and/or BDD(OH). The evolution of selected aromatics like 4hydroxyphenylacetamide (tr ¼ 3.81 min) from atenolol, 4-(2methoxyethyl)phenol (tr ¼ 1.45 min) from metoprolol and phthalic acid (tr ¼ 1.12 min) from propranolol is depicted in Fig. 7a,b and c, respectively. These compounds were unequivocally identified by comparing their retention times and UVevis spectra with those of pure standards. Comparison of Figs. 6a and 7b indicates that for each EAOP,
[4-hydroxyphenylacetamide] / mM
0.03
a
0.02
0.01
0.00 0
30
60
90
120
150
4-(2-methoxyethyl)phenol is destroyed at time slightly longer than metoprolol, following in EF the same increasing oxidation sequence for the electrochemical cells used, i.e., Pt/ADE < Pt/ ADEePt/CF BDD/ADE < BDD/ADEePt/CF, and disappearing more rapidly and at analogous rate for both SPEF treatments. Fig. 7a and c show similar trends for the other two intermediates, although 4-hydroxyphenylacetamide is more quickly formed and destroyed for all cells because of the higher removal rate of atenolol than metoprolol (see k1 values in Table 1), whereas phthalic acid disappears more slowly using the BDD/ADE cell compared with the Pt/ADEePt/CF one, suggesting its faster reaction with OH than with BDD(OH). All these findings confirm the much quicker oxidation of aromatics with BDD(OH) than with Pt(OH) and its fast attack with OH in the bulk from Fenton’s reaction (2), accelerated by Fe2þ regeneration from reaction (3) at the CF cathode in the combined cells and even induced in larger extent from reaction (5) under the potent action of the incident UV light of solar irradiation at the CPC photoreactor. A different behavior was found for generated carboxylic acids by ion-exclusion chromatography, since the UV irradiation photolyzes their Fe(III) complexes from reaction (6). Tartaric acid (tr ¼ 8.4 min) detected in the metoprolol solution, as well as other acids such as maleic (tr ¼ 7.6 min) and formic (tr ¼ 13.8 min) coming from the oxidative breaking of the aromatic moieties and lateral groups of beta-blockers (Sire´s et al., 2007; Skoumal et al., 2008; Isarain-Cha´vez et al.,
180
0.8
b
a
0.7
[oxalic acid] / mM
[4-(2-methoxyethyl)phenol] / mM
0.03
0.02
0.01
0.6 0.5 0.4 0.3 0.2 0.1
0.00
0.0 0
60
120
180
240
300
360
420
b
0.07
[phthalic acid] / mM
[oxamic acid] / mM
c
0.06 0.05 0.04 0.03
0.15
0.10
0.05
0.02
0.00 0
0.01 0.00
60
120
180
240
300
360
420
time / min 0
60
120
180
240
300
360
420
time /min
Fig. 7 e Evolution of the aromatic intermediates: (a) 4hydroxyphenylacetamide from atenolol, (b) 4-(2methoxyethyl)phenol from metoprolol and (c) phthalic acid from propranolol, detected during the degradation of the drugs in the conditions of Fig. 4.
Fig. 8 e Time-course of (a) oxalic and (b) oxamic acids during the degradation of beta-blocker solutions in the recirculation flow plant with a combined BDD/ADEePt/CF cell operating in the same conditions of Fig. 4. EF process: (B) atenolol, (6) metoprolol tartrate and () propranolol hydrochloride. SPEF process: (C) atenolol, (:) metoprolol tartrate and (-) propranolol hydrochloride.
4128
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 1 9 e4 1 3 0
Table 2 e Concentration of ammonium and nitrate ions obtained after 360 min of EF and SPEF treatments of 10 L of betablockers solutions in 0.1 M Na2SO4 with 0.5 mM Fe2D at pH 3.0 and 35 C using the recirculation flow plant with a BDD/ ADEePt/CF cell at 3.0e0.4 A and liquid flow rate of 250 L hL1. Process
Compound
[NH4þ]/mM
% Initial N as NH4þ
[NO3]/mM
% Initial N as NO3
EF
Atenolol Metoprolol Propranolol Atenolol Metoprolol Propranolol
0.330 0.124 0.117 0.561 0.214 0.193
28 25 22 47 44 37
0.551 0.335 0.329 0.027 0.025 0.100
46 68 63 2 5 19
SPEF
2011a,b), disappear in short time for all treated solutions, whereas oxalic (tr ¼ 6.4 min) and oxamic (tr ¼ 9.0 min) acids remained as the more persistent ultimate carboxylic acids, which are directly mineralized to CO2. Fig. 8a shows that oxalic acid is largely accumulated up to 0.58e0.74 mM after 360 min of all EF treatments with a BDD/ ADEePt/CF cell, since its Fe(III) complexes are hardly attacked with BDD(OH) and OH (Brillas et al., 2009; Ruiz et al., 2011). In contrast, the homologous SPEF processes yield maximum contents <0.24 mM of this acid at 60 min that drastically decay to <0.01e0.03 mM at 360 min, as a consequence of the fast photodecomposition of Fe(III)-oxalate complexes (Zuo and Hoigne´, 1992). On the other hand, Fig. 8b evidences a much slower photolysis of Fe(III)-oxamate complexes. In the case of atenolol, for example, oxamic acid attains a steady content of 0.15 mM in EF, but the photolysis of its Fe(III) complexes only reduces its content to 0.078 mM at the end of the SPEF process, corresponding to ca. 2 mg L1 TOC, a value lower than 6 mg L1 TOC determined for the final electrolyzed solution (see Table 1). This indicates the formation of other undetected and more refractory compounds during the EAOPs, which can be confirmed by a simple mass balance for the EF degradation of atenolol with a BDD/ADEePt/CF cell. So, from Fig. 8a and b, one infers that oxalic and oxamic acids contribute in 21 mg L1 to the 39 mg L1 of TOC of the final treated solution (see Table 1), indicating that 18 mg L1 TOC are due to other refractory compounds that are accumulated in large extent. Note that the SPEF process mineralizes strongly these compounds, probably composed of ironcomplexes that are photodecomposed by UV irradiation. Table 2 summarizes the concentrations of NH4þ and NO3 ions lost during the degradation of the three beta-blockers in the flow plant with a BDD/ADEePt/CF reactor. The EF process leads to a much larger release of NO3 ion in all cases, which becomes as high as 68% of the initial N for metoprolol. For the SPEF treatment, however, the conversion into NH4þ is strongly enhanced in 1.7-fold, but the formation of NO3 ion is drastically inhibited to practically disappear for atenolol and metoprolol. That means that UV irradiation favors the mineralization of a larger proportion of N-intermediates via NH4þ ion and the production of volatile N-derivatives, probably NOx, prior to NO3 generation.
4.
Conclusions
It has been demonstrated that SPEF can be a viable method for the remediation of wastewaters with beta-blocker. When an
optimized current of 3.0e0.4 A was applied to the most potent BDD/ADEePt/CF cell coupled with a CPC photoreactor in the 10 L recirculation flow plant, 95e97% mineralization with about 100% of maximum current efficiency was obtained after 360 min of treatment of 100 mg L1 TOC of all drugs. However, the energy consumption of this treatment was of about 0.250 kWh g TOC1, a value much higher than about 0.080 kWh g TOC1 found for the Pt/ADEePt/CF cell under comparable conditions. The latter cell is then preferable for SPEF from an economic point of view, although it yields a slightly lower mineralization of 88e93%. The comparative EF treatments of the same beta-blocker solutions gave poorer mineralization, becoming superior for the combined than single cells due to the faster oxidation of organics with the higher amounts of OH produced from Fenton’s reaction (2) from the fast Fe2þ regeneration at the CF cathode. The organics were also more rapidly destroyed by BDD(OH) than Pt(OH) in both single and combined cells. The higher oxidation power of BDD and the additional formation of OH in combined cells were confirmed from the quicker decay determined for the three beta-blockers in EF by reversedphase HPLC. A larger acceleration in the removal of these compounds was found in SPEF by the additional production of OH in the bulk induced by the photolysis of Fe(OH)2þ species from the potent action of UV light of solar irradiation at the CPC photoreactor. The decay kinetics of beta-blockers in all treatments always followed a pseudo first-order reaction. In each trial, aromatic intermediates were destroyed with the same generated hydroxyl radicals. In contrast, ultimate carboxylic acids like oxalic and oxamic and other undetected recalcitrant by-products remained in the final treated solutions by EF, but their Fe(III) complexes were photolyzed by UV light in SPEF, thus explaining the much greater mineralization reached by this method. While the EF process yielded a much larger loss of NO3 than NH4þ ion, the UV irradiation of sunlight in SPEF favored the mineralization of N-intermediates via NH4þ ion and the production of volatile N-derivatives.
Acknowledgment The authors thank the financial support from MEC (Ministerio de Educacio´n y Ciencia, Spain) under the project CTQ200760708/BQU, cofinanced with FEDER funds. The grant awarded to E. Isarain-Cha´vez from CONACYT (Consejo Nacional de Ciencia y Tecnologı´a, Mexico) to do this work is also acknowledged.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 1 9 e4 1 3 0
references
Andreozzi, R., Marotta, R., Nicklas, P., 2003. Pharmaceuticals in STP effluents and their solar photodegradation in aquatic environment. Chemosphere 50 (10), 1319e1330. Balci, B., Oturan, N., Cherrier, R., Oturan, M.A., 2009. Degradation of atrazine in aqueous medium by electrocatalytically generated hydroxyl radicals. A kinetic and mechanistic study. Water Res. 43 (7), 1924e1934. Benı´tez, F.J., Acero, J.L., Real, F.J., Rolda´n, G., 2009. Ozonation of pharmaceutical compounds: rate constants and elimination in various water matrices. Chemosphere 77 (1), 53e59. Benner, J., Salhi, E., Ternes, T., von Gunten, U., 2008. Ozonation of reverse osmosis concentrate: kinetics and efficiency of beta blocker oxidation. Water Res. 42 (12), 3003e3012. Boye, B., Brillas, E., Marselli, B., Michaud, P.A., Comninellis, Ch., Farnia, G., Sandona`, G., 2006. Electrochemical incineration of chloromethylphenoxy herbicides in acid medium by anodic oxidation with boron-doped diamond electrode. Electrochim. Acta 51 (14), 2872e2880. Brillas, E., Sire´s, I., Oturan, M.A., 2009. Electro-Fenton process and related electrochemical technologies based on Fenton’s reaction chemistry. Chem. Rev. 109 (12), 6570e6631. Can˜izares, P., Beteta, A., Saez, C., Rodrı´guez, L., Rodrigo, M.A., 2008. Use of electrochemical technology to increase the quality of the effluents of bio-oxidation processes. A case studied. Chemosphere 72 (7), 1080e1085. Cleuvers, M., 2005. Initial risk assessment for three b-blockers found in the aquatic environment. Chemosphere 59 (2), 199e205. Diagne, M., Oturan, N., Oturan, M.A., 2007. Removal of methyl parathion from water by electrochemically generated Fenton’s reagent. Chemosphere 66 (5), 841e848. Fent, K., Weston, A.A., Caminada, D., 2006. Ecotoxicology of human pharmaceuticals. Aquat. Toxicol. 76 (2), 122e159. Flox, C., Cabot, P.L., Centellas, F., Garrido, J.A., Rodrı´guez, R.M., Arias, C., Brillas, E., 2007. Solar photoelectro-Fenton degradation of cresols using a flow reactor with a boron-doped diamond anode. Appl. Catal. B: Environ. 75 (1e2), 17e28. Hammami, S., Bellakhal, N., Oturan, N., Oturan, M.A., Dachraoui, M., 2008. Degradation of acid orange 7 by electrochemically generated OH radicals in acidic aqueous medium using a boron-doped diamond or platinum anode. A mechanistic study. Chemosphere 73 (5), 678e684. Hamza, M., Abdelhedi, R., Brillas, E., Sire´s, I., 2009. Comparative electrochemical degradation of the triphenylmethane dye methyl violet with boron-doped diamond and Pt anodes. J. Electroanal. Chem. 627 (1e2), 41e50. Isarain-Cha´vez, E., Arias, C., Cabot, P.L., Centellas, F., Rodrı´guez, R.M., Garrido, J.A., Brillas, E., 2010a. Mineralization of the drug b-blocker atenolol by electro-Fenton and photoelectro-Fenton using an air-diffusion cathode for H2O2 electrogeneration combined with a carbon-felt cathode for Fe2þ regeneration. Appl. Catal. B: Environ. 96 (3e4), 361e369. Isarain-Cha´vez, E., Rodrı´guez, R.M., Garrido, J.A., Arias, C., Centellas, F., Cabot, P.L., Brillas, E., 2010b. Degradation of the beta-blocker propranolol by electrochemical advanced oxidation processes based on Fenton’s reaction chemistry using a borondoped diamond anode. Electrochim. Acta 56 (1), 215e221. Isarain-Cha´vez, E., Cabot, P.L., Centellas, F., Rodrı´guez, R.M., Arias, C., Garrido, J.A., Brillas, E., 2011a. Electro-Fenton and photoelectro-Fenton degradations of the drug beta-blocker propranolol using a Pt anode: identification and evolution of oxidation products. J. Hazard. Mater. 185 (2e3), 1228e1235. Isarain-Cha´vez, E., Garrido, J.A., Rodrı´guez, R.M., Centellas, F., Arias, C., Cabot, P.L., Brillas, E., 2011b. Mineralization of metoprolol by electro-Fenton and photoelectro-Fenton processes. J. Phys. Chem. A 115 (7), 1234e1242.
4129
Khataee, A.R., Zarei, M., Asl, S.K., 2010. Photocatalytic treatment of a dye solution using immobilized TiO2 nanoparticles combined with photoelectro-Fenton process: optimization of operational parameters. J. Electroanal. Chem. 648 (2), 143e150. Kim, I., Yamashita, N., Tanaka, H., 2009. Photodegradation of pharmaceuticals and personal care products during UV and UV/H2O2 treatments. Chemosphere 77 (4), 518e525. Ku¨mmerer, K., 2009. Antibiotics in the aquatic environmenta review-Part I. Chemosphere 75 (4), 417e434. , J., Caminal, G., Petrovic , M., Marco-Urrea, E., Radjenovic Vicent, T., Barcelo´, D., 2010. Oxidation of atenolol, propranolol, carbamazepine and clofibric acid by a biological Fenton-like system mediated by the white-rot fungus Trametes versicolor. Water Res. 44 (2), 521e532. Marselli, B., Garcı´a-Go´mez, J., Michaud, P.A., Rodrigo, M.A., Comninellis, Ch, 2003. Electrogeneration of hydroxyl radicals on boron-doped diamond electrodes. J. Electrochem. Soc. 150 (3), D79eD83. Maurer, M., Escher, B.I., Richle, P., Schaffner, C., Alder, A.C., 2007. Elimination of b-blockers in sewage treatment plants. Water Res. 41 (7), 1614e1622. Oturan, N., Oturan, M.A., 2005. Degradation of three pesticides used in viticulture by electrogenerated Fenton’s reagent. Agron. Sustain. Dev. 25 (2), 267e270. Oturan, M.A., Pimentel, M., Oturan, N., Sire´s, I., 2008. Reaction sequence for the mineralization of the short-chain carboxylic acids usually formed upon cleavage of aromatics during electrochemical Fenton treatment. Electrochim. Acta 54 (2), 173e182. Owen, S.F., Giltrow, E., Huggett, D.B., Hutchinson, T.H., Saye, J., Winter, M.J., Sumpter, J.P., 2007. Comparative physiology, pharmacology and toxicology of b-blockers: mammals versus fish. Aquat. Toxicol. 82 (3), 145e162. ¨ zcan, A., S‚ahin, Y., Savas‚ Koparal, A., Oturan, M.A., 2008. O Degradation of picloram by the electro-Fenton process. J. Hazard. Mater. 153 (1e2), 718e727. Panizza, M., Cerisola, G., 2009. Direct and mediated anodic oxidation of organic pollutants. Chem. Rev. 109 (12), 6541e6569. Ramil, M., El Aref, T., Fink, G., Scheurer, M., Ternes, T., 2010. Fate of beta blockers in aquatic-sediment systems: sorption and biotransformation. Environ. Sci. Technol. 44 (3), 962e970. Rosal, R., Rodrı´guez, A., Perdigo´n-Melo´n, J.A., Mezcua, M., Hernando, M.D., Leto´n, P., Garcı´a-Calvo, E., Agu¨era, A., Ferna´ndez-Alba, A.R., 2008. Removal of pharmaceuticals and kinetics of mineralization by O3/H2O2 in a biotreated municipal wastewater. Water Res. 42 (14), 3719e3728. Ruiz, E.J., Arias, C., Brillas, E., Herna´ndez-Ramı´rez, A., PeraltaHerna´ndez, J.M., 2011. Mineralization of acid yellow 36 azo dye by electro-Fenton and solar photoelectro-Fenton processes with a boron-doped diamond anode. Chemosphere 82 (4), 495e501. Sire´s, I., Cabot, P.L., Centellas, F., Garrido, J.A., Rodrı´guez, R.M., Arias, C., Brillas, E., 2006. Electrochemical degradation of clofibric acid in water by anodic oxidation. Comparative study with platinum and boron-doped diamond electrodes. Electrochim. Acta 52 (1), 75e85. Sire´s, I., Centellas, F., Garrido, J.A., Rodrı´guez, R.M., Arias, C., Cabot, P.L., Brillas, E., 2007. Mineralization of clofibric acid by electrochemical advanced oxidation processes using a borondoped diamond anode and Fe2þ and UVA light as catalysts. Appl. Catal. B: Environ. 72 (3e4), 373e381. Sire´s, I., Brillas, E., Cerisola, G., Panizza, M., 2008. Comparative depollution of mecoprop aqueous solutions by electrochemical incineration using BDD and PbO2 as high oxidation power anodes. J. Electroanal. Chem. 613 (2), 151e159. Sire´s, I., Oturan, N., Oturan, M.A., 2010. Electrochemical degradation of b-blockers. Studies on single and multicomponent synthetic aqueous solutions. Water Res. 44 (10), 3109e3120.
4130
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 1 9 e4 1 3 0
Skoumal, M., Arias, C., Cabot, P.L., Centellas, F., Garrido, J.A., Rodrı´guez, R.M., Brillas, E., 2008. Mineralization of the biocide chloroxylenol by electrochemical advanced oxidation processes. Chemosphere 71 (9), 1718e1729. Skoumal, M., Rodrı´guez, R.M., Cabot, P.L., Centellas, F., Garrido, J.A., Arias, C., Brillas, E., 2009. Electro-Fenton, UVA photoelectroFenton and solar photoelectro-Fenton degradation of the drug ibuprofen in acid aqueous medium using platinum and borondoped diamond. Electrochim. Acta 54 (7), 2077e2085. Song, W., Cooper, W.J., Mezyk, S.P., Greaves, J., Peake, B.M., 2008. Free radical destruction of b-blockers in aqueous solution. Environ. Sci. Technol. 42 (4), 1256e1261. Sun, Y., Pignatello, J.J., 1993. Photochemical-reactions involved in the total mineralization of 2,4-D by Fe3þ/H2O2/UV. Environ. Sci. Technol. 27 (2), 304e310.
Vieno, N., Tuhkanen, T., Kronberg, L., 2007. Elimination of pharmaceuticals in sewage treatment plants in Finland. Water Res. 41 (5), 1001e1012. Welcher, F.J. (Ed.), 1975. Standard methods of chemical analysis, Sixth ed., vol. 2. R.E. Krieger Publishers, Huntington, New York, p. 1827. Part B. Zarei, M., Khataee, A.R., Ordikhani-Seyedlar, R., Fathinia, M., 2010. Photoelectro-Fenton combined with photocatalytic process for degradation of an azo dye using supported TiO2 nanoparticles and carbon nanotube cathode: neural network modeling. Electrochim. Acta 55 (24), 7259e7265. Zuo, Y., Hoigne´, J., 1992. Formation of hydrogen peroxide and depletion of oxalic acid in atmospheric water by photolysis of iron (III)-oxalato complexes. Environ. Sci. Technol. 26 (5), 1014e1022.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 3 1 e4 1 4 0
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Effects of titanate nanotubes synthesized by a microwave hydrothermal method on photocatalytic decomposition of perfluorooctanoic acid Ying-Chu Chen a,b, Shang-Lien Lo a,*, Jeff Kuo c a
Graduate Institute of Environmental Engineering, National Taiwan University, 71, Chou-Shan Road, Taipei 106, Taiwan, ROC Department of Environmental Monitoring & Information Management, Environmental Protection Agency, Taiwan, ROC c Department of Civil and Environmental Engineering, California State University, Fullerton, CA, USA b
article info
abstract
Article history:
Titanate nanotubes (TNTs) were used to remove perfluorooctanoic acid (PFOA) from
Received 13 December 2010
aqueous solutions in this study. Direct photolysis of PFOA by a 254-nm UV light (400 W) was
Received in revised form
found effective to decompose PFOA without presence of photocatalysts. Shorter-chain
7 May 2011
perfluorocarboxylic acids (PFCAs) and fluoride ions were formed during photodecomposi-
Accepted 21 May 2011
tion. Addition of TNTs as photocatalysts did not greatly enhance photocatalytic decom-
Available online 31 May 2011
position of PFOA. TNTs mainly act as adsorbents to adsorb PFOA and form TNTePFOA complexes. It suggested that sodium ions and oxygen atoms on the surfaces of TNTs play
Keywords:
important roles in PFOA adsorption. X-ray Photoelectron Spectroscopy (XPS) and Fourier
Microwave
Transform Infrared Spectroscopy (FTIR) analyses indicated that ion-exchange, electrostatic
Perfluorooctanoic acid
interaction, and hydrophobic interaction all participated in the photocatalytic reaction of
Photocatalysis
PFOA by TNTs.
Titanate nanotube
Crown Copyright ª 2011 Published by Elsevier Ltd. All rights reserved.
X-ray photoelectron spectroscopy
1.
Introduction
Perfluorocarboxylic acids (CnF2nþ1COOH, PFCAs) and their derivatives are anthropogenic organic compounds that have a wide range of applications. PFCAs can be used as industrial surfactants, additives, firefighting foams, and lubricants (Lin et al., 2010a). Some perfluorinated acids, especially perfluorooctanoic acid (PFOA, C7H15COOH), have garnered concerns because of their ubiquitous usage, persistence, and bioaccumulation in the environment. PFOA is exceptionally inert and persists indefinitely in the environment (Musijowski et al., 2007), even in remote polar areas (Qu et al., 2010). Some evidences show that PFOA can accumulate in organisms and
pose a potential human health risk (Hinderliter et al., 2006; Potera, 2009). The Science Advisory Board of the United States Environmental Protection Agency (U.S. EPA) reported that PFOA is likely a carcinogen (Hogue, 2006). Furthermore, U.S. EPA launched a program to reduce 95% PFOA emissions from manufacturers by 2010 and to eliminate the use of the chemical by 2015 (Qu et al., 2009). Development of effective treatment methods to convert PFOA into harmless species is desirable. Due to its inherent resistance to chemical and microbiological treatment, many technologies have been developed to decompose PFOA, including advanced oxidation processes (AOPs), such as UV-H2O2, persulfate, photo-Fenton, ozonation
* Corresponding author. Tel.: þ886 2 2362 5373; fax: þ886 2 2392 8821. E-mail address:
[email protected] (S.-L. Lo). 0043-1354/$ e see front matter Crown Copyright ª 2011 Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.05.020
4132
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 3 1 e4 1 4 0
and sonolysis (Cheng et al., 2008; Hori et al., 2004, 2008b; Moriwaki et al., 2005; Vecitis et al., 2008). Direct photolysis (Cao et al., 2010; Chen and Zhang, 2006; Chen et al., 2007; Hori et al., 2004) and photocatalysis (Qu et al., 2010) in aqueous solutions are alternatives that are being researched in recent years. PFOA is sensitive to light with wavelengths from deep UV-region to 200 nm (Cao et al., 2010), thus can be efficiently decomposed by direct UV irradiation. TiO2-mediated heterogeneous photocatalysis of PFOA was found to involve generation of electron/hole pairs that led to better photocatalytic efficiencies (Dillert et al., 2008; Doll and Frimmel, 2005; Estrellan et al., 2010; Panchangam et al., 2009; Ravichandran et al., 2006, 2009). The b-Gallium oxide (Zhao and Zhang, 2009) and ZnO (Ravichandran et al., 2007) were also used as photocatalysts to assist photodecomposition of PFOA. Decomposition of PFOA into fluoride ions, which can readily react with Ca2þ to form CaF2, a raw material in global demand for production of hydrofluoric acid (Hori et al., 2009, 2010; Ravichandran et al., 2006). In addition to direct photolysis and photocatalysis, adsorption has been demonstrated to be an effective and economical method to remove polar organic pollutants from aqueous solutions (Senevirathna et al., 2010; Yu et al., 2009), but only a few papers were found in our literature survey that studied the removal of PFOA by adsorbents such as active carbon (Lampert et al., 2007; Ochoa-Herrera and Sierra-Alvarez, 2008; Senevirathna et al., 2010; Yu et al., 2009; Qu et al., 2009), polymers (Senevirathna et al., 2010), zeolite and sludge (Ochoa-Herrera and Sierra-Alvarez, 2008). PFOA is moderately adsorbable with active carbon, and the adsorption efficiency is greatly influenced by particle sizes of adsorbents. The powdered activated carbon (PAC) achieved an adsorption equilibrium within 4 h (Senevirathna et al., 2010; Yu et al., 2009), while granular activated carbon (GAC) required over 168 h to reach the equilibrium (Yu et al., 2009). Titanate nanotubes (TNTs) synthesized by microwave hydrothermal (M-H) methods have some desired properties, such as high specific surface area, photocatalytic properties, and ion-exchangeable capabilities (Ou and Lo, 2007). In our previous study, the specific surface area of microwaveassisted TNTs was 150 m2 g1 and could adsorb 2000 mg Pb(II) per gram of TNTs at pH of 4 (Chen et al., 2010a). Compositing TNTs with semiconductors, such as cadmium sulfide (CdS), can enhance their photocatalytic capability to eliminate 52.3% of ammonia in water (Chen et al., in press). TNTs synthesized by M-H methods have great potentials to be photcatalysts and adsorbents in various applications. This study aimed at investigating adsorption and photocatalytic behaviors of TNTs synthesized by a M-H method for their removal of PFOA from aqueous solutions. The sorption isotherms were developed and effects of solution pH were evaluated; while effects of UV light, loading concentrations, and solution pH on photocatalysis were also studied. A detailed reaction mechanism was proposed based on the experimental results and the instrumental (X-ray Photoelectron Spectroscopy (XPS) and Fourier Transform Infrared Spectroscopy (FTIR)) analyses. Findings of this study provide insights of using TNTs to remediate environmental media that are contaminated by PFOA.
2.
Materials and methods
2.1.
Materials
All chemicals used in the experiments were of reagent grade. Perfluorooctanoic acid (PFOA, C7F15COOH, 96% purity) was from Aldrich (USA). Perfluoroheptanoic acid (PFHpA, C6F13COOH, 98% purity), perfluoropentanoic acid (PFPeA, C4F9COOH, 97% purity), and heptafluorobutyric acid (HFBA, C3F7COOH, 99% purity) were from Alfa Aesar (USA). Perfluorohexanoic acid (PFHxA, C5F11COOH, 97% purity) was purchased from Fluka (USA). Methyl alcohol anhydrous (CH3OH, 99.9% purity, Mallinckrodt Chemicals, USA), boric acid (H3BO3, 99.5% purity, Nacalai Tesque Inc., Japan), and sodium hydroxide (NaOH, 99.7% purity, Alps Chem Co. Ltd., Taiwan) were used to prepare mobile phases for high performance liquid chromatography (HPLC) analyses. Sulfuric acid (H2SO4, 97% purity, Nacalai Tesque Inc., Japan) was used to regenerate the suppressor column of HPLC. Also, sodium bicarbonate (NaHCO3, 99.6% purity, Nacalai Tesque Inc., Japan), sodium carbonate (Na2CO3, 99.0% purity, Nacalai Tesque Inc., Japan), and H2SO4 were used to prepare the mobile phase for ion chromatography. Titanium dioxide (TiO2 Degussa P25, 99.5% purity, Simakyu Pure Chemicals, Japan), NaOH, and hydrochloric acid (HCl, 70% purity, Fisher Scientific, USA) were used to fabricate TNTs. Nitric acid (HNO3, 60.0% purity, Yakuri Pure Chemicals Co. Ltd., Japan) and NaOH were used for pH adjustment. All solutions were prepared with Milli-Q ultrapure water (18 MUcm resistivity).
2.2.
Synthesis of TNTs
Titanate nanotubes were synthesized by the M-H method described elsewhere (Chen et al., 2010a). The microwave digestion system (Ethos Touch Control, Milestone Corporation, Italy) consists of a double-walled vessel with an inner ˆ Teflon liner and an outer shell of high strength ULTEMA polyetherimide. The Teflon liner can resist reaction conditions employed in this study and no cross-contamination from the Teflon liner was observed. Briefly, a mixture of 600 mg TiO2 and 70 mL NaOH (10 N) was stirred for 40 min in a Teflon container which was then sealed before the microwave heating process. The container was then placed into the microwave digestion system under 400-W irradiation at 403 K for 3 h. After the treatment, the solution was washed three times with 0.5 N HCl and four times with deionized water, followed by centrifugation at 800 rpm for 5 min (Kubota 6800, Kubota Co., Japan), and vacuum dried at 214 K for 24 h (FD312P, Kingmech Co., Ltd., Taiwan).
2.3.
Batch experiments
All the experiments were carried out in polypropylene (PP) bottles to avoid potential interferences from sample containers. The experimental apparatus for photocatalysis contains a cooling water jacket to maintain the temperature at 298 K (B204, Firstek Scientific Co., Ltd., Taiwan) and a UV lamp (254 nm, 400 W, Philips, Holland). Known amounts of
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 3 1 e4 1 4 0
4133
CF 100% C0 15
(2)
photocatalysts (TNTs or TiO2) were dispersed into 2-L PFOA solutions (50 mg L1). During UV irradiation, 20 mL aliquots were taken at 0, 1, 2, 3, 6, 9, 24 h, and photocatalysts were immediately removed by filtration through 0.22-mm filters (MillexGS, Millipore, Ireland) for subsequent analyses. The experiments were carried out twice and mean values were reported here. Adsorption experiments were performed under ambient conditions. Analytical grade PFOA was used to prepare a stock solution of 1000 mg L1, which was further diluted with deionized water to different concentrations for adsorption experiments. Adsorption isotherm studies were conducted with initial PFOA concentration ranging from 5 to 100 mg L1 together with 20 mg TNTs or TiO2 in 100 mL solutions. Solutions were then shaken at 298 K for 24 h at a speed of 150 rpm in a reciprocal shaker bath (BT-350, Bersing Co., Ltd., Taiwan). The pH values of initial solutions were adjusted by using HNO3 or NaOH to reach the desired values. Samples were taken from each solution and then analyzed by HPLC. All experiments were conducted twice and average values were used in the data analyses.
2.4.
Analytical methods
Concentrations of PFOA and shorter-chain PFCAs were analyzed by HPLC (Dionex, UltiMate 3000, USA) coupled with a conductivity detector (ED-50, Dionex, USA) and an anion self-regenerating suppressor (ASRS 300 2-mm, USA). The conductivity detector was used in conjunction with a Dionex micromembrane suppressor column. The suppressor column was constantly regenerated using 0.5 mM H2SO4 at a flow rate of 1 mL min1. The PFCAs were extracted by a 150 2.1 mm, 3.5 mm column (AcclainPolar AdvantageII C18, Dionex, USA) maintained at 303 K. A trinary gradient was employed with mobile phase A containing 100% methanol, mobile phase B containing Milli-Q water, and mobile phase C containing 9 mM NaOH and 100 mM H3BO3 in deionized water. The flow rate of the solvents was 0.3 mL min1. The gradient was operated at 20% phase A, 40% phase B for the initial 5 min; 20% phase A increased to 60% and 40% phase B decreased to 0% for the next 10 min (5e15 min); Maintaining 60% phase A, 0% phase B during 15e20 min; 60% phase A decreased to 20%, 0% phase B increased to 40% for final 5 min. The phase C was maintained at 40% during the total running time, 25 min. All calibration curves for PFCAs were linear over the 0.5e50 mg L1 range. The limits of detection (LODs) using 50 mL samples, based on a signal-to-noise (S/N) ratio of 3, were 1 mg L1 for PFCAs. Degradation ratios were calculated as follows: R¼
C0 C 100% C0
(1)
where C is concentration of PFOA (mM) and C0 is initial concentration of PFOA (mM). Concentrations of aqueous fluoride ions were determined by an ion chromatography (Metrohm 790 Personal IC, Metrohm Ltd., Switzerland) with a column of CH-9101 Herisan (Metrohm Ltd., Switzerland). A mixture of 1.8 M Na2CO3/1.7 M NaHCO3 (1:1) was used as the mobile phase at a flow rate of 1.0 mL min1. Defluorination ratios were calculated as follows:
RF ¼
Where CF is concentration of fluoride ions (mM), C0 is initial concentration of PFOA (mM), and the factor of 15 corresponds to the number of fluorine atoms in one PFOA molecule.
2.5.
Characterization methods for TNTs
The TNTs used in this study were fabricated in our research lab. The physicochemcial properties, such as BET surface area and zeta potential, of TNTs synthesized by the M-H method can be found in Chen et al. (2010a). Surface functional groups of TNTs were characterized by FTIR and XPS. The IR scanning range was 4000e650 cm1 with 4 cm1 resolution using a mercury-cadmium-telluride detector in a Nexus 470 spectrometer (Thermo Nicolet, USA). KBr powder was used to record the reference spectrum. Chemical bonding of PFCAs onto surfaces of TNTs/TiO2 was investigated by XPS using an ESCA PHI 1600 (Physical Electronics, USA). The XPS spectra were recorded with a monochromatized Mg (Ka) source of 1253.6 eV energy (15 kV, 400 W). The bonding energy scale for final calibration was corrected by the C1s peak of 284.6 eV.
3.
Results and discussion
3.1.
Direct photolysis
Aqueous PFOA solutions (50 mg L1) were irradiated with a 254-nm UV light (400 W), and the extents of photolysis were shown in Fig. 1(a). Control tests (without UV irradiation) were carried out and no significant changes in PFOA concentration were observed. The degradation rate was the fastest at pH 4, followed by pH 7, and then pH 10; 98% of initial PFOA was photodecomposed at pH 4 after 48-h irradiation. It was reported that PFOA has a weak absorption of UV light longer than 200 nm, and only 9% of initial PFOA was decomposed with a 254-nm UV light after 2-h irradiation (23 W lowpressure mercury lamp) (Cao et al., 2010), and 89.5% of initial PFOA was decomposed by a xenon-mercury lamp after 72-h irradiation (wavelengths were mainly in the range of 220e460 nm) (Hori et al., 2004). It implies that irradiation intensity of lights plays an important role in affecting photolytic efficiency of PFOA. PFOA can be efficiently photodecomposed under a light emitting higher irradiation intensity than that with weaker irradiation intensity. Photodecomposition of PFOA and formation of shorter-chain PFCAs, bearing C6eC7 perfluoroalkyl groups were quantified by HPLC. The C5eC4 perfluoroalkyl groups were hardly detected within 24-h irradiation, while C3eC2 perfluoroalkyl groups were known to be easily evaporated under ambient conditions. At pH 4, the concentration of PFHpA increased with irradiation up to 24 h and then decreased, concentration of PFHxA increased up to 36 h and then decreased, while fluoride concentrations increased continuously with extended irradiation time (Fig. 1(b)). It indicates that photodecomposition of PFOA was achieved by a step-by-step removal of the CF2 units. Besides irradiating with a 254-nm UV light, photolysis of PFOA was also effective in VUV system (185 nm) (Cao et al., 2010;
4134
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 3 1 e4 1 4 0
a 1.0
pH 4 pH 7 pH 10 No UV
0.8
C/C0
0.6
0.4
Degradation Defluorination
100
0.2
Percentage
80 60 40 20
0.0
0
0
b
pH4
pH7
pH10
10
No UV
20
Time (h)
30
40
50
6
5
Fluoride ion (mM)
4
3 Fluoride ion PFOA PFHpA PFHxA Total fluoride ion
2
1
0 0
10
20
30
40
50
Time (h) Fig. 1 e (a) Direct photolysis of PFOA at different solution pH by 254 nm UV irradiation (b) and formed shorter-chain PFCAs during photodecomposition. The reaction conditions are [PFOA]initial [ 50 mg LL1, T [ 298 K in a 2 L solution.
Chen and Zhang, 2006; Chen et al., 2007). Therefore, direct photolysis of PFOA was mainly depended on irradiation intensity of lights when the wavelength were the same; otherwise, longer wavelength of lights needs higher irradiation intensity to assist photolytic efficiency of PFOA without catalysts.
After 24 h of irradiation, defluorination of PFOA at pH 4, 7, and 10 were 85%, 68%, and 55%, respectively. The acid condition (pH 4) was found more favorable for photodegradation (98%) and defluorination (85%) of PFOA. Although initial pH value of solutions were adjusted by HNO3 and/or NaOH to reach the desired values, pH value of solutions would
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 3 1 e4 1 4 0
gradually decrease during 24-h UV irradiation and finally reached the equilibrium at pH 4, which is the natural pH of 50 mg L1 PFOA under ambient conditions. HNO3 has been demonstrated to have minor effects on degradation and defluorination of PFOA (Ravichandran et al., 2006), whereas presume of sodium ions may have retardation effects on defluorination and decomposition of PFOA since dissolved metal ions were found to affect photocatalytic efficiencies of PFCAs (Ravichandran et al., 2006). Therefore, using NaOH to increase initial pH value of solutions to 7 or 10 might not be favorable in photodecomposition of PFOA. Released fluoride ions can bind with sodium ions to form sodium fluoride (NaF), causing a lower defluorination at pH 7 and 10. In addition, more anionic fluoride ions released into solutions would also decrease pH value of solutions during photolysis. Solution pH can be an indicator to observe if experiments reached the equilibrium when it becomes steady. Consequently, this study suggests not to adjust pH value of the solutions and maintain it at pH 4 (the highest photolytic efficiency) in the subsequent experiments.
3.2.
Adsorption of PFOA onto TNTs
Adsorption isotherms of PFOA onto TNT or TiO2 are shown in Fig. 2. TiO2 has a much lower adsorption capacity for PFOA than TNTs at pH 4; it is as expected be due to the fact that TNTs have a surface area that is three times larger, 140 v.s. 50 m2 g1 (Chen et al., 2010a). The length of PFOA molecules is 1 nm (Yu et al., 2009), thus PFOA can easily diffuse into inner pores of TNTs (pore diameters of TNTs were 9 nm (Chen et al., 2010a)), and adsorb onto inner as well as outside walls of TNTs. Fig. 2 also shows the extents of adsorption decrease with an increase in solution pH. Due to the low pKa of PFOA, 2.5 (US EPA, 2002), more PFOA will be in ionic form as solution pH
50 40
-1
Qe (mg g )
30 20 10
TiO2 (pH 4) TNTs (pH 4) TNTs (pH 7) TNTs (pH 10)
0 -10 20
40
60
80
-1
Ce (mg L ) Fig. 2 e Adsorption isotherms of PFOA onto TNTs/TiO2. The reaction conditions are [PFOA]initial [ 5e100 mg LL1, T [ 298 K, reacted with 20 mg of TNTs or TiO2 in a 100 mL solution. Ce is the equilibrium concentration in solutions and qe is the amount of PFOA adsorbed per unit weight of the adsorbent.
4135
increases. Electrostatic attraction between negatively charged PFOA and positively charged TNT surfaces would promote surface reactions with valence band holes (hþ), which are favorable in PFOA photodecomposition (Bahnemann et al., 1997; Estrellan et al., 2010). Higher adsorption efficiency achieved at lower solution pH was due to electrostatic forces of attraction (Panchangam et al., 2009). Besides electrostatic interaction, hydrophobic interaction was also important for adsorption of PFOA (Higgins et al., 2005; Higgins and Luthy, 2006; Yu et al., 2009). Hydrophobic PFOA and TNTs had a hydrophobic interaction during the adsorption process (Yu et al., 2009). The formation of PFOA-TNTs complexes could enhance oxidative power of TNTs (Dillert et al., 2008).
3.3.
Photocatalysis induced by TNTs
First-order kinetics were used to simulate photodecomposition of PFOA at pH 4 by a 254-nm UV light, and the results were shown in Fig. 3. The pseudo first-order rate constants were calculated to be 0.0230 h1, 0.0337 h1 and 0.0271 h1, for 0.05, 0.125 and 0.25 g L1 TNTs, respectively. The remaining PFOA concentration at 24-h photocatalytic reactions decreased as amounts of TNTs increased from 0.05 to 0.125 g L1; when amounts of TNTs further increased 0.25 g L1, the remaining PFOA concentration started to increase. It is apparent that amounts of TNTs have great influences on photodecomposition of PFOA and a suggested loading amount was 0.125 g L1. An adequate loading concentration increases the generation rate of electron/hole pairs that promote photodecomposition of PFOA, while excessive amounts of TNTs in solutions would decrease the light penetration (Zhu et al., 2005). The superior photocatalytic performance of TNTs than TiO2 might be due to the larger surface area of TNTs to receive more UV irradiation to excite PFOA adsorbed on the surfaces than TiO2 particles (Estrellan et al., 2010; Panchangam et al., 2009). Consequently, TNTs synthesized by raw materials of TiO2 can enhance their photocatalytic efficiency to decompose PFOA. Concentrations of remaining PFOA gradually decreased while those of fluoride ions increased as the photocatalytic reaction proceeded; however, the reaction rates with the presence of TNTs were much slower than that of direct photolysis. In the presence of 0.125 g L1 TNTs, only 59% of the initial PFOA was degraded, whereas the corresponding yield for direct photolysis was almost 100% after 24-h irradiation. Thus, an addition TNTs or TiO2 as photocatalysts would retard photodecomposition of PFOA. This phenomenon can be explained in terms of heterogeneous reactions. TNTs, which are intended as photocatalysts, act as adsorbents to adsorb PFOA before mineralization occurs. Photocatalysts were also found to scatter UV lights (Zhu et al., 2005) and suppress formation of shorter-chain PFCAs during photocatalytic reactions (Hori et al., 2004). Generating active hydroxyl radicals ($OH) in TiO2 system has poor photoreactivity for PFCAs (Cao et al., 2010; Hori et al., 2004). Slower degradation rates of PFOA in a TNT system may also be attributed to release sodium ions, that were reported to affect photodecomposition of PFCAs. Multi-walled TNTs were synthesized by scrolling up four layers of TiO6 octahedrons (Scheme S1 in the online Supplementary Material) (Chen et al., 2002), thus the inner surface
4136
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 3 1 e4 1 4 0
1.0
60
Degradation Defluorination
0.8
ln (C0/C)
0.6
Percentage
50
59%
44%
49% 2
R =0.9754
40 30 20
19% 2
R =0.9655
10 0
0.50g TiO2 0.50g TNTs 0.25g TNTs 0.10g TNTs 2
R =0.9598
0.4
2
R =0.9939
0.2
TiO2 (0.50 g) TNTs (0.50 g) TNTs (0.25 g) TNTs (0.10 g)
0.0 0
5
10
Time (h)
15
20
25
Fig. 3 e Pseudo first-order kinetics of photocatalytic decomposition of PFOA by TNTs/TiO2 by 254 nm UV irradiation. The reaction conditions are [PFOA]initial [ 50 mg LL1, pH 4, T [ 298 K in a 2 L solution. The first-order rate reaction can be expressed as ln½Ct [LktDln½C0 , where [C]t is the concentration of the chemical of interest at a particular time and [C]0 is the initial concentration.
area of TNTs is much larger than outside walls. Researchers found that the non-planar structure provided a scattering center for incident light, thus elongated the path to promote the photocatalytic capability of photocatalysts (Lin et al., 2010b). Outer surfaces of TNTs could harvest the lights from any directions in the surrounding that could be helpful for increasing the photocatalytic efficiency; therefore, when PFOA diffused into pores and adsorbed onto inner walls of TNTs, its opportunity to receive UV lights was greatly reduced. Defluorination ratios of PFOA were much smaller than the PFOA decomposition ratios. It implies that intermediate products formed during photocatalysis (Chen and Zhang, 2006). Fluoride ions can adsorb onto surfaces of photocatalysts, leading to less available surface active sites and smaller defluorination ratios (Estrellan et al., 2010; Panchangam et al., 2009). TNTs have a high affinity to adsorb fluoride ions (Fig. A1 in the online Supplementary Material). Positive charges on TNT surfaces under acidic conditions (pH 4) cause more electrostatic attractions of fluoride ions. From the above results, it can be deduced that smaller defluorination ratios of photocatalytic reactions was caused by adsorption of fluoride ions onto TNT surfaces since fluoride ions were continuously released during photodecomposition of PFOA; the other possibility was that PFOA was directly adsorbed onto inner surfaces of TNTs without mineralization.
3.4.
FTIR and XPS analyses of PFOA-TNTs/TiO2
Fig. 4 shows XPS results of TNT or TiO2 before and after photocatalytic reactions. The O1s, C1s, F1s, Na1s, and Ti2p3 peaks
were investigated in XPS wide-scans (Fig. 4(a)). Intensities O1s peaks at 532 eV were obviously decreased after photocatalytic reactions (Fig. 4(b)). It reveals that oxygen-terminated functional groups on TNT surfaces react with PFOA during photocatalytic reactions that led to a decrease in oxygen concentrations after photocatalytic reactions (Guan et al., 2007). The exciting peaks at 287 eV and 292 eV corresponded to bulk carbon signals (Fig. 4(c)) (Guan et al., 2007). Enhanced intensities of C1s peaks suggest that PFOA and shorter-chain PFCAs were directly adsorbed onto TNTs/TiO2 surfaces. If mineralization of PFOA continued, peaks corresponded to fluoride ions should appear at 689 eV (Fig. 4(d)) (Guan et al., 2007; Hori et al., 2006, 2008a); however, no peaks corresponded to fluoride species were found. Most of PFOA molecules were directly adsorbed onto surfaces of TNTs without mineralization during UV irradiation since the adsorption rate of TNTs is very fast (Chen et al., 2010a). Intensities of Na1s peaks apparently decreased after photocatalysis (Fig. 4(e)). These results support the previous ones that TNTs mainly act as adsorbents to adsorb PFOAs rather than as photocatalysts. The adsorption mechanism of TNTs is that sodium ions on TNT surfaces react with PFOA and less sodium ions would be detected on the surfaces of TNTs. No sodium ions were detected while the reactions were photocatalyzed by TiO2. The binding energies for Ti(IV) (467 eV) and Ti(III) (461 eV) on TNTs or TiO2 were similar after photocatalytic reactions (Fig. 4(f)). The XPS spectra support the proposed photocatalytic mechanisms that PFOA firstly adsorbed onto surfaces of TNTs, and the remaining PFOA in solutions or PFOA adsorbed on the outside walls of TNTs are mineralized by UV irradiation and formed shorter-chain PFCAs.
4137
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 3 1 e4 1 4 0
a
b
TiO 2 + PFOA
O1s
TiO 2 + PFOA
TNTs + PFOA TNTs + PFOA TNTs
TNTs
1000
800
600
400
200
c
C1s
0
545
540
535
530
d
525 F1s
Intensity (a.u.)
TiO 2 + PFOA TiO 2 + PFOA TNTs + PFOA TNTs + PFOA TNTs TNTs
300
295
290
e
285
280 700 Na1s
TiO 2 + PFOA
695
690
685
f
680 Ti2p3
TiO 2 + PFOA
TNTs + PFOA TNTs + PFOA
TNTs
1085
TNTs
1080
1075
1070
475
470
465
460
455
Binding Energy (eV) Fig. 4 e XPS spectra of virgin TNTs and PFOAeTNTs/TiO2 complexes after photocatalysis.
Fig. 5 shows FTIR spectra of TNTs/TiO2 photocatalysts and PFOAeTNTs/TiO2 complexes. The band at 1138 cm1 was attributed to CeF bonds stretching vibrations in eOOCC7F15 groups on the surfaces of PFOAeTNTs/TiO2 complexes (Kutsuna and Hori, 2008; Men et al., 2008; Xu et al., 2009). The bands at 1438 cm1 and 1647 cm1 were attributed to eCOOe asymmetrical stretching vibrations and symmetrical stretching vibrations, respectively (Xu et al., 2009). The wide band at 3400 cm1 region was attributed to presence of hydroxyl groups ($OH) and traces of water in the KBr pellet (Men et al., 2008). Intensities of hydroxyl groups ($OH) after photocatalytic reactions indicated that catalysts were excited by UV lights to form hydroxyl radicals ($OH), which can degrade PFOA and prevent rapid recombination with electron/hole pairs (Estrellan et al., 2010). However, intensities of hydroxyl groups ($OH) after photocatalytic reaction by TiO2 were ascribed to be the occurrence of oxygen vacancies on the surfaces, which were also found to be beneficial to its photocatalytic activity (Ismail et al., 2007). The above FTIR and XPS results indicated
that the removal of PFOA mainly occurred with the functional groups on the surface of photocatalysts.
3.5.
Mechanisms of PFOA decomposition
Based on our experimental results, it can propose the photodecomposition mechanisms of PFOA by TNTs under UV irradiation (Fig. 6). PFOA exists as an anionic compound when emerging in solutions (Eq. (3)) (Burns et al., 2008). C7 F15 COOH þ H2 O/C7 F15 COO þ H3 Oþ
(3)
Positive charges on TNT surfaces are favorable to adsorb anionic PFOA with or without UV irradiation (Eq. (4)). hy TNTs /TNTs$ /TNTsþþC7 F15 COO /TNTsC7 F15 COO
(4)
PFOA adsorbed on the outer surfaces of TNTs and the remaining PFOA in solutions were excited by UV irradiation to
4138
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 3 1 e4 1 4 0
2.0
C7F15COOH
a
Ionization
TNTs
1.5
C7F15COO -
TNTs+PFOA
1.0
Photocatalysis
TNTs Adsorption 0.5
Ionization
TNT+ + C7F15COO-
0.0
-1.0 4000
3500
3000
2000
1500
C7F15OH H+ F-
1000
C6F13COF
b
1.5
2500
H 2O
Photocatalysis
TNT - C7F15COO TNT - C7F15
-0.5
Absorbence
C7F15 + COO
Adsorption
TNT - PFOA
TiO2+PFOA
1.0
TiO2
0.5
Mineralization
FC6F13COOH CF2 PFHpA CF2
0.0 -0.5 -1.0 4000
Fig. 6 e Proposed PFOA decomposition mechanism by TNTs. 3500
3000
2500
2000
1500
1000
-1
Wavenumber (cm ) Fig. 5 e The Fourier-transformed infrared spectra of virgin TNTs/TiO2 and PFOAeTNTs/TiO2 complexes after photocatalysis. form excited state of PFOA and photolyzed into C7F15$ and COOH$ radicals (Eq. (5)). hy
hy
C7 F15 COOH/C7 F15 COOH$/C7 F15 $ þ COOH$
hy
Conclusions
Titanate nanotubes (TNTs) synthesized by the M-H method with large surface specific surface areas make them good candidates for removal of PFOA from aqueous solutions. Direct photolysis, photocatalysis were conducted under different solution pH, coupled with XPS and FTIR spectroscopies. The findings obtained in this study have demonstrated the following:
(5)
The C7F15$ radical continue adsorb onto surfaces of excited TNTs to reach the maximum adsorbility (Eq. (6)). TNTs/TNTs$/TNTs$ þ C7 F15 $/TNTs C7 F15
4.
(6)
The C7F15 radicals also react with water (water acts as a nucleophilic reagent to attract the end carbon atom) to form C7F15OH (Chen et al., 2007), an alcohol, which undergoes HF elimination to form C6F13COF (Eq. (7)) (Chen and Zhang, 2006; Hori et al., 2004). C6F13COF was then hydrolyzed to form PFHpA with one less the CF2 unit (Hori et al., 2004) and fluoride ions were releases into aqueous solutions (Eq. (8)). C7 F15 OH/C6 F13 COF þ Hþ þ F
(7)
C6 F13 COF/C6 F13 COOH þ Hþ þ F
(8)
The CeC bond between C6F13 and COOH can be further cleaved to form C6F13 radicals, which undergoes similar reaction routes as C7F15 radicals to give PFHxA. The bond cleavage routes of PFOA were similar to the photo-Kolbe mechanism (Dillert et al., 2008; Ravichandran et al., 2006). In the same manner, PFOA bearing shorter-chain PFCAs, such as PFHpA, PFHxA, PFPeA, and HFBA, were formed in a stepwise manner as time progressed.
(1) PFOA could efficiently be photodecomposed by a 254-nm UV light (400 W) within 48 h without presence of photocatalysts. Also, shorter-chain perfluorocarboxylic acids (PFCAs) and released fluoride ions were formed during photodecomposition. (2) TNTs are good adsorbents than TiO2 for PFOA removal. The maximum adsorption capacity can be as high as 50 mg PFOA/g TNT at acidic pH of 4. (3) XPS and FTIR analyses indicated that interactions for the removal of PFOA mainly occurred on the surfaces of TNTs.
Acknowledgments The authors would like to thank the National Science Council of the Republic of China, for their financial support under Contract No. NSC 98-2221-E-002-040-MY3.
Appendix. Supplementary data Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.watres.2011.05.020.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 3 1 e4 1 4 0
references
Bahnemann, D.W., Hilgendorff, M., Memming, R., 1997. Charge carrier dynamics at TiO2 particles: reactivity of free and trapped holes. J. Phys. Chem. B 101 (21), 4265e4275. Burns, D.C., Ellis, D.A., Li, H., Mcmurdo, C.J., Webster, E., 2008. Experimental pKa determination for perfluorooctanoic acid (PFOA) and the potential impact of pKa concentration dependence on laboratory-measured partitioning phenomena and environmental modeling. Environ. Sci. Technol. 42 (24), 9283e9288. Cao, M.H., Wang, B.B., Yu, H.S., Wang, L.L., Yuan, S.H., Chen, J., 2010. Photochemical decomposition of perfluorooctanoic acid in aqueous periodate with VUV and UV light irradiation. J. Hazard. Mater. 179 (1e3), 1143e1146. Chen, J., Zhang, P., 2006. Photodegradation of perfluorooctanoic acid in water under irradiation of 254 nm and 185 nm light by use of persulfate. Water Sci. Technol. 54 (11e12), 317e325. Chen, J., Zhang, P., Liu, J., 2007. Photodegradation of perfluorooctanoic acid by 185 nm vacuum ultraviolet light. J. Environ. Sci. 19 (4), 387e390. Chen, Q., Du, G.H., Zhang, S., Peng, L.M., 2002. The structure of trititanate nanotubes. Acta Cryst. Sect. B 58 (4), 587e593. Chen, Y.C., Lo, S.L., Kuo, J., 2010a. Pb(II) adsorption capacity and behavior of titanate nanotubes made by microwave hydrothermal method. Colloids Surf. A: Physicochem. Eng. Asp. 361 (1e3), 126e131. Chen, Y.C., Lo, S.L., Ou, H.H., and Chen, C.H. Photocatalytic oxidation of ammonia by cadmium sulfide/titanate nanotubes synthesized by microwave hydrothermal method. Water Sci. Technol. in press. Cheng, J., Vecitis, C.D., Park, H., Mader, B.T., Hoffmann, M.R., 2008. Sonochemical degradation of perfluorooctane sulfonate (PFOS) and perfluorooctanoate (PFOA) in landfill groundwater: environmental matrix effects. Environ. Sci. Technol. 42 (21), 8057e8063. Dillert, R., Bahnemann, D., Hidaka, H., 2008. Light-induced degradation of perfluorocarboxylic acids in the presence of titanium dioxide. Chemosphere 67 (4), 785e792. Doll, T.E., Frimmel, F.H., 2005. Cross-flow microfiltration with periodical back-washing for photocatalytic degradation of pharmaceutical and diagnostic residues-evaluation of the long-term stability of the photocatalytic activity of TiO2. Water Res. 39 (5), 847e854. Estrellan, C.R., Salim, C., Hinode, H., 2010. Photocatalytic decomposition of perfluorooctanoic acid by iron and niobium co-doped titanium dioxide. J. Hazard. Mater. 179 (1e3), 79e83. Guan, B., Zhi, J., Zhang, X., Murakami, T., Fujishima, A., 2007. Electrochemical route for fluorinated modification of borondoped diamond surface with perfluorooctanoic acid. Electrochem. Commun. 9 (12), 2817e2821. Higgins, C.P., Luthy, R.G., 2006. Sorption of perfluorinated surfactants on sediments. Environ. Sci. Technol. 40 (23), 7251e7256. Higgins, C.P., Field, J.A., Criddle, C.S., Luthy, R.G., 2005. Quantitative determination of perfluorochemicals in sediments and domestic sludge. Environ. Sci. Technol. 39 (11), 3946e3956. Hinderliter, P.M., DeLorme, M.P., Kennedy, G.L., 2006. Perfluorooctanoic acid: relationship between repeated inhalation exposures and plasma PFOA concentration in the rat. Toxicology 222 (1e2), 80e85. Hogue, C., 2006. POFA called likely human carcinogen. Chem. Eng. News 83, 5. Hori, H., Hayakawa, E., Einaga, H., Kutsuna, S., Koike, K., Ibusuki, T., et al., 2004. Decomposition of environmentally persistent perfluorooctanoic acid in water by photochemical approaches. Environ. Sci. Technol. 38 (22), 6118e6124.
4139
Hori, H., Murayama, M., Kutsuna, S., 2009. Oxygen-induced efficient mineralization of perfluoroalkylether sulfonates in subcritical water. Chemosphere 77 (10), 1400e1405. Hori, H., Murayama, M., Inoue, N., Ishida, K., Kutsuna, S., 2010. Efficient mineralization of hydroperfluorocarboxylic acids with persulfate in hot water. Catal. Today 151 (1e2), 131e136. Hori, H., Nagaoka, Y., Sano, T., Kutsuna, S., 2008a. Iron-induced decomposition of perfluorohexanesulfonate in sub- and supercritical water. Chemosphere 70 (5), 800e806. Hori, H., Nagaoka, Y., Yamamoto, A., Sano, T., Yamashita, N., Taniyasu, S., et al., 2006. Efficient decomposition of environmentally persistent perfluorooctanesulfonate and related fluorochemicals using zerovalent iron in subcritical water. Environ. Sci. Technol. 40 (3), 1049e1054. Hori, H., Yamamoto, A., Koike, K., Kutsuna, S., Murayama, M., Yoshimoto, A., et al., 2008b. Photocatalytic decomposition of a perfluoroether carboxylic acid by tungstic heteropolyacids in water. Appl. Catal. B: Environ. 82 (1e2), 58e66. Ismail, A.A., Abboudi, M., Holloway, P., 2007. Photoluminence from terbium doped silica-titania prepared by a sol-gel method. Mater. Res. Bull. 42 (1), 137e142. Kutsuna, S., Hori, H., 2008. Experimental determination of Henry’s law constant of perfluorooctanoic acid (PFOA) at 298 K by means of an inert-gas stripping method with a helical plate. Atmos. Environ. 42 (39), 8883e8892. Lampert, D.J., Frisch, M.A., Speitel Jr., G.E., 2007. Removal of perfluorooctanoic acid and perfluorooctane sulfonate from wastewater by ion exchange. Period Haz., Tox. Radioac. Waste Manag. 11 (1), 60e68. Lin, A.Y.C., Panchangam, S.C., Ciou, P.S., 2010a. High levels of perfluorochemicals in Taiwan’s wastewater treatment plants and downstream rivers pose great risk to local aquatic ecosystems. Chemosphere 80 (10), 1167e1174. Lin, C.J., Chen, S.Y., Liou, Y.H., 2010b. Wire-shaped electrode of CdSe-sensitized ZnO nanowire arrays for photoelectrochemical hydrogen generation. Electrochem. Commun. 12 (11), 1513e1516. Men, X.H., Zhang, Z.Z., Song, H.J., Wang, K., Jiang, W., 2008. Fabrication of superhydrophobic surfaces with poly (furfuryl alcohol)/multi-walled carbon nanotubes composites. Appl. Surf. Sci. 254 (9), 2563e2568. Moriwaki, H., Takagi, Y., Tanaka, M., Tsuruho, K., Okitsu, K., Maeda, Y., 2005. Sonochemical decomposition of perfluorooctane sulfonate and perfluorooctanoic acid. Eviron. Sci. Technol. 39 (9), 3388e3392. Musijowski, J., Trojanowicz, M., Szostek, B., Lima, J.L.F.D.C., Lapa, R., Yamashita, H., et al., 2007. Flow-injection determination of total organic fluorine with off-line defluorination reaction on a solid sorbent bed. Anal. Chim. Acta 600 (1e2), 147e154. Ochoa-Herrera, V., Sierra-Alvarez, R., 2008. Removal of perfluorinated surfactants by sorption onto granular activated carbon, zeolite, and sludge. Chemosphere 72 (10), 1588e1593. Ou, H.H., Lo, S.L., 2007. Review of titania nanotubes synthesized via the hydrothermal treatment: fabrication, modification, and application. Sep. Purif. Technol. 58 (1), 179e191. Panchangam, S.C., Lin, A.Y.C., Tsai, J.H., Lin, C.F., 2009. Sonication-assisted photocatalytic decomposition of perfluorooctanoic acid. Chemosphere 75 (5), 654e660. Potera, C., 2009. Reproductive toxicology: study associates PFOS and PFOA with impaired fertility. Environ. Health Perspect. 117 (4), A148. Qu, Y., Zhang, C., Li, F., Bo, X., Liu, G., Zhou, Q., 2009. Equilibrium and kinetics study on the adsorption of perfluorooctanoic acid from aqueous solution onto powdered activated carbon. J. Hazard. Mater. 169 (1e3), 146e152.
4140
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 3 1 e4 1 4 0
Qu, Y., Zhang, C., Li, F., Chen, J., Zhou, Q., 2010. Photo-reductive defluorination of perfluorooctanoic acid in water. Water Res. 44 (9), 2939e2947. Ravichandran, L., Selvam, K., Swaminathan, M., 2007. Effect of oxidants and metal ions on photodefluoridation of pentafluorobenzoic acid with ZnO. Sep. Purif. Technol. 56 (2), 192e198. Ravichandran, L., Selvam, K., Krishnakumar, B., Swaminathan, M., 2009. Photovalorisation of pentafluorobenzoic acid with platinum doped TiO2. J. Hazard. Mater. 167 (1e3), 763e769. Ravichandran, L., Selvam, K., Muruganandham, M., Swaminathan, M., 2006. Photocatalytic cleavage of C-F bond in pentafluorobenzoic acid with titanium dioxide-P25. J. Fluorine Chem. 127 (9), 1204e1210. Senevirathna, S.T.M.L.D., Tanaka, S., Fujii, S., Kunacheva, C., Harada, H., Shivakoti, B.R., et al., 2010. A comparative study of adsorption of perfluorooctane sulfonate (PFOS) onto granular activated carbon, ion-exchange polymers and non-ionexchange polymers. Chemosphere 80 (6), 647e651. US EPA, 2002. Revised Draft Hazard Assessment of Perfluorooctanoic Acid and Its Salts. United States Environmental Protection Agency Office of Pollution Prevention and Toxics Risk Assessment Division, Washington,
DC. Available from. http://www.ewg.org/files/EPA_PFOA_ 110402.pdf. Vecitis, C.D., Park, H., Cheng, J., Mader, B.T., Hoffmann, M.R., 2008. Kinetics and mechanism of the sonolytic conversion of the aqueous perfluorinated surfactants, perfluorooctanoate (PFOA), and perfluorooctane sulfonate (PFOS) into inorganic products. J. Phys. Chem. 112 (18), 4261e4270. Xu, X.H., Zhang, Z.Z., Liu, W., 2009. Fabrication of superhydrophobic surfaces with perfluorooctanoic acid modified TiO2/polystyrene nanocomposites coating. Colloid Surf. A: Phys. Eng. Asp. 341 (1e3), 21e26. Yu, Q., Zhang, R., Deng, S., Huang, J., Yu, G., 2009. Sorption of perfluorooctane sulfonate and perfluorooctanoate on active carbons and resin: kinetic and isotherm study. Water Res. 43 (4), 1150e1158. Zhao, B., Zhang, P., 2009. Photocatalytic decomposition of perfluorooctanoic acid with b-Ga2O3 wide bandgap photocatalyst. Catal. Commun. 10 (8), 1184e1187. Zhu, X., Castleberry, S.R., Nanny, M.A., Butler, E.C., 2005. Effects of pH and catalyst concentration on photocatalytic oxidation of aqueous ammonia and nitrite in titanium dioxide suspersions. Environ. Sci. Technol. 39 (10), 3784e3791.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 4 1 e4 1 5 1
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
A comparison of different approaches for measuring denitrification rates in a nitrate removing bioreactor So¨ren Warneke a,*, Louis A. Schipper a, Denise A. Bruesewitz b, W. Troy Baisden c a
Department of Earth and Ocean Sciences, University of Waikato, Hamilton, New Zealand Department of Biological Sciences, University of Waikato, Hamilton, New Zealand c National Isotope Centre, GNS Science, Lower Hutt, New Zealand b
article info
abstract
Article history:
Denitrifying woodchip bioreactors (denitrification beds) are increasingly used to remove
Received 15 January 2011
excess nitrate (NO 3 ) from point-sources such as wastewater effluent or subsurface drains
Received in revised form
from agricultural fields. NO 3 removal in these beds is assumed to be due to microbial
15 May 2011
denitrification but direct measurements of denitrification are lacking. Our objective was to
Accepted 22 May 2011
test four different approaches for measuring denitrification rates in a denitrification bed
Available online 31 May 2011
that treated effluent discharged from a glasshouse. We compared these denitrification
Keywords:
8.73 1.45 g m3 d1. In vitro acetylene inhibition assays resulted in highly variable deni-
rates with the rate of NO 3 removal along the length of the bed. The NO3 removal rate was
Denitrification
trification rates (DRAI) along the length of the bed and generally 5 times greater than the
15
measured (NO 3 dN removal rate. An in situ pushepull test, where enriched
N
15
NdNO 3 was
Pushepull
injected into 2 locations along the bed, resulted in rates of 23.2 1.43 g N m3 d1 and
Stable isotopes
8.06 1.64 g N m3 d1. The denitrification rate calculated from the increase in dissolved N2
Acetylene block
and N2O concentrations (DRN2) along the length of the denitrification bed was
Natural abundance
6.7 1.61 g N m3 d1. Lastly, denitrification rates calculated from changes in natural abundance measurements of d15NeN2 and d15NdNO 3 along the length of the bed yielded
a denitrification rate (DRNA) of 6.39 2.07 g m3 d1. Based on our experience, DRN2
measurements were the easiest and most efficient approach for determining the denitrification rate and N2O production of a denitrification bed. However, the other approaches were useful for testing other hypotheses such as factors limiting denitrification or may be applied to determine denitrification rates in environmental systems different to our study site. DRN2 does require very careful sampling to avoid atmospheric N2 contamination but could be used to rapidly determine denitrification rates in a variety of aquatic systems with high N2 production and even water flows. These measurements demonstrated that the majority of NO 3 removal was due to heterotrophic denitrification. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
The global production of anthropogenic nitrogen (N) is increasing due to food and energy production (Vitousek et al., 1997; Canfield et al., 2010). This N also has lasting adverse
effects on the environment, including increased greenhouse gas emissions, stratospheric ozone depletion, pollution of drinking water, and eutrophication of streams, lakes and coastal waters (Galloway et al., 2004, 2008; Canfield et al., 2010). There are a range of strategies to reduce the N load to
* Corresponding author. Tel.: þ64 7 858 3700; fax: þ64 7 858 4964. E-mail address:
[email protected] (S. Warneke). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.05.027
4142
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 4 1 e4 1 5 1
aquatic ecosystems from agricultural practices, such as the construction or preservation of wetlands and riparian buffers, and installation of denitrification beds or walls (Dinnes et al., 2002; Vymazal et al., 2006; Schipper et al., 2010). Denitrification beds are large containers filled with wood by-products that act as a carbon source to support heterotrophic denitrification, which converts nitrate (NO 3 ) to nitrous oxide (N2O) and N2 gases (Seitzinger et al., 2006). These beds are increasingly being installed to remove NO 3 from point-source discharges such as effluent streams and drainage systems (Schipper et al., 2010). It is generally presumed that microbial denitrification is predominantly responsible for the NO 3 dN removal in these beds (Schipper et al., 2010) and that other NO 3 removal processes such as dissimilatory NO 3 reduction to ammonium (DNRA), anammox and microbial/plant uptake are relatively low. For example, Greenan et al. (2006, 2009) showed that less than 4% of NO 3 dN removal in woodchip columns was due to DNRA and that microbial uptake only accounted for 2e3.5% of NO 3 dN removed. Isotopic enrichment of natural abundance of 15N in NO 3 was measured in the outflow of a denitrification bed and in a column study while NHþ 4 concentrations were low, was also suggestive of microbial denitrification (Robertson et al., 2000; Robertson, 2010). However, there are various processes beside heterotrophic denitrification that can account for 15 NdNO 3 increase in natural systems (Bedard-Haughn et al., 2001). Therefore, measurement of the products of denitrification (N2, N2O), is critical to establish that denitrification is responsible for NO 3 removal. Our previous work suggested that denitrification is the primary pathway for NO 3 removal in denitrification beds because we measured very high potential rates of denitrification using the acetylene inhibition method. Anammox and DNRA were likely negligible due to low NHþ 4 concentrations and the lack of plant/algae growth on the denitrification bed ruled out biotic uptake of NO 3 (Warneke et al., 2011). However, there are no direct measurements of denitrification rates in operating denitrification beds to demonstrate that denitrification dominates other NO 3 removal processes. Developing a method to directly measure denitrification rates would also allow reliable determination of NO 3 removal rates in denitrification beds and potentially in other similar aquatic systems because determining NO 3 removal via measurement of inflow and outflow NO 3 concentrations is difficult in many of these systems due to high temporal variability in NO 3 concentrations and flow rates at inflow and outflow (Schipper et al., 2010). A number of different techniques may be used to measure denitrification rates in terrestrial and aquatic environments (Groffman et al., 2006). The acetylene inhibition method has probably been the most commonly used approach for measuring denitrification (Groffman et al., 2006). Acetylene inhibits the reduction of N2O to N2 and accumulated N2O can be measured using gas chromatography. However, the acetylene block technique can lead to inaccurate measurements of denitrification rates because acetylene has a number of other unwanted effects on microbial populations e.g., acting as an inhibitor of nitrifiers or as a carbon source (Groffman et al., 2006). Denitrification rates measured in soils using acetylene inhibition technique are generally an underestimate of actual rates (Groffman et al., 2006).
Denitrification rates in water-saturated environments (e.g., groundwater or wetlands) can also be estimated using the pushepull method (Addy et al., 2002) where a slug of 15Nlabelled NO 3 is added into the denitrifying environment and the accumulation of 15NeN2 and 15NeN2O is measured with time (Hauck and Melsted, 1956; Addy et al., 2002; Baker and Vervier, 2004). Direct quantification of denitrification by measuring N2 emissions from soils has also been attempted, although it is technically challenging due to the high atmospheric background concentration of N2 (Butterbach-Bahl et al., 2002). Similarly, in aquatic environments, increases in dissolved N2 can be measured but are also confounded by high background levels of dissolved N2 concentrations derived from the atmosphere (Groffman et al., 2006). However, conditions for measuring denitrification in rivers via dissolved N2 concentrations described by Laursen and Seitzinger (2005) suggested that it may be possible to directly measure increases in dissolved N2 concentrations along the length of denitrification beds due to their turbulent-free water flow and their potentially high production of N2 through denitrification. A final approach that could be used to demonstrate denitrification as the main mechanism for NO 3 removal in denitrification beds is the measurement of changes in the 15N/14N natural abundance of NO 3 and nitrogen gases along the length of the bed. If denitrification was the main mechanism of NO 3 removal then there should be increases in natural abundance 15 15 N/14N in NO 3 , observed as d NeNO3, due to the strong 15 discrimination against N during denitrification (Mariotti et al., 1981) and a negative congruent decrease in the 15N/14N of N2 gas produced, reported as d15NeN2. The main objectives of this study were to determine whether denitrification rates were high enough to account for the observed NO 3 removal in an operational denitrification bed and to compare different methods for measuring denitrification rates in denitrification beds. A range of the techniques were trialled for accuracy, ease, and expense of measurement, including measuring changes in the dissolved nitrogen gases and natural abundance stable isotope (15NeN2 and 15 NdNO 3 ) along the length of the bed, acetylene inhibition assays, and accumulation of 15N-labelled N2 and N2O following introduction of an 15N-labelled NO 3 spike.
2.
Materials and methods
2.1.
Study site
This study was performed at a large denitrification bed (176 m 5 m 1.5 m) constructed in 2006 and filled with a mixture of woodchips and sawdust (Warneke et al., 2011). The bed treated effluent from a glasshouse, which grew hydroponic cucumbers, tomatoes and capsicums at Karaka, New Zealand. The effluent from the glasshouse was pumped into one end of the denitrification bed through a PVC pipe 1 m below the surface of the woodchips and was discharged from the other end of the bed into a drainage ditch. Twelve fully screened PVC wells (2 m long; diameter 0.05 m) were installed along the length of the bed at 16 m intervals for effluent sampling. Mechanical water metres (LXLG-80, Bil, China) at
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 4 1 e4 1 5 1
the inflow and at the outflow of the bed measured the flow rate of the incoming and outcoming effluent of the bed.
2.2.
Nitrate-N removal rate
Effluent samples for NO 3 concentration were taken from each well along the length of the bed using a simple pump and stored in plastic tubes (50 mL) before being frozen at 16 C until analysis. After filtering (0.45 mm disposable membrane filters), NO 3 dN concentrations of effluent samples were analysed with a flow injection analyser (Lachat Instruments; Loveland, Colorado, USA; APHA, 1992). NO 3 dN expresses the ). NO removal rates (g N m3 nitrogen (N) of the nitrate (NO 3 3 1 bed material d ) were calculated from the linear decline of 3 the NO 3 dN concentrations (g m ) along the length of the bed (DNO3 ): NO3 dN removal rate ¼ DNO 3 dN FR/Vbed, where FR was the flow rate of the effluent (m3 d1) and Vbed was the volume of the bed (m3).
2.3.
Denitrification rates
2.3.1. In vitro denitrification measurement using acetylene inhibition technique e DRAI A modified version of the acetylene inhibition technique (Tiedje et al., 1989) was used to measure in vitro denitrification rates (DRAI) (see Warneke et al., 2011). Woodchips (effluentsaturated bed material) were sampled along the bed from 0.2 m below the bed surface using a shovel and stored in plastic bags at 4 C until analysis within 2 days. Woodchips (100 g) and effluent (60 g) from each sampling location (12) was placed into airtight glass jars (600 ml). After addition of acetylene (10% of the headspace volume) the increase in N2O concentration was measured using a gas chromatograph (Varian; Palo Alto, USA) equipped with an electron capture detector, and Hayesep D column (3.6 m 1/800 2.0 mm). The column oven temperature was 80 C, the ECD detector temperature was 300 C and the flow rate of the argon/ methane-carrier gas was 40 mL min1. The DRAI was calculated from linear increase in N2O concentration with time. This rate was adjusted for the difference in temperature between the bed at the time of sampling (19 C)and laboratory incubation temperature (27 C) using a Q10 of 2.1 (Warneke et al., 2011). Q10 is the factor of the reaction rate increase with every 10 C rise in temperature. The temperature of the bed effluent was measured in each well using an InLab 605 O2Sensor (Mettler Toledo, Switzerland).
4143
Tygontubing (internal diameter 4 mm) directly into a Tedlar gas bag (25 L) to avoid contact with the atmosphere and to maintain anoxic water conditions (Fig. 1). The extracted 15 NO effluent was amended with 15 NdNO 3 (95.0% 3 ) to about 20% above background NO3 concentration. Sodium bromide (NaBr) was also added (10 times background Br concentration) to allow calculation of the dilution of added 15 NdNO 3 and 15N gas during subsequent sampling. Then the amended effluent was pumped back into the bed through the same pipe. Prior to amendment, background effluent and gas samples were collected to determine background concentrations of 15 NeN2O and 15NeN2 (t1). 15NeN2O and 15NeN2 NO 3 dN, express the nitrogen isotope 15N of the nitrous oxide (N2O) and nitrogen (N2) gas respectively Immediately after the enriched effluent was pumped back in the bed, effluent samples were taken at time 0 (t0) and subsequently every 30 min for 3 h (t1et6). For each sampling, effluent was pumped out of the stainless steel pipe with the first litre of the effluent being discarded. Subsequent effluent samples were stored in 50 ml PP tubes (Greiner Bio-one, Germany) at 4 C until 15 NdNO analysis of Br, NO 3 , and 3 concentrations. To obtain samples for analysis of dissolved 15NeN2O and 15 NeN2, the effluent was pumped through the pipe and tubing into a 50-ml plastic syringe fitted with a 3 way stopcock. This sampling procedure was performed in a transparent plastic container (20 L) filled water to avoid atmospheric contact (Fig. 1). The first outlet of the stopcock was connected to the syringe, the second outlet of the stopcock was connected to the tubing of the ultra-high purity helium tank (99.999% purity, BOC Ltd., New Zealand) and the third outlet was connected after the helium purging step (see below) to the tubing of the stainless steel pipe (effluent) or later on to a needle (Fig. 2aeg). Before collecting the effluent with the syringe, the stopcock and the syringe was generously purged (five times) with helium to remove traces of background N2 gas (Fig. 2aeb). The effluent tube from the pipe was connected to the stopcock while continuously pumping out effluent from the bed. Exactly 35 ml of effluent was pumped into the syringe (Fig. 2c).
2.3.2. In situ denitrification measurement using pushepull technique e DRPP In situ denitrification rates (DRPP) were measured using the pushepull 15 NdNO 3 technique (Addy et al., 2002; Baker and Vervier, 2004) at two locations (location A at 48 m bed length; location B at 128 m bed length) in the denitrification bed within 2 days. 15 NdNO 3 expresses the nitrogen isotope 15 N of the nitrate (NO ). Effluent flow in the bed was stopped 3 at the inlet 6 days before sampling to reduce movement of the introduced tracers away from the sampling locations. Effluent from 1.2 m depth from the bed (20 L) was pulled (400 mL min1) using a peristaltic pump and a stainless steel pipe fitted with a screen at the base which was connected to
Fig. 1 e Overview of experimental setup to sample effluent and their dissolved gases (gas headspaces) from 1.20 m depth of the denitrification bed without atmosphere contact.
4144
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 4 1 e4 1 5 1
Fig. 2 e Detailed view on the sampling steps under water in the transparent container of Fig. 1 to obtain headspace gas samples from the effluent for analysis of dissolved N2 and N2O. a) and b) He purging; c) effluent sampling; d) Filling with He; e) Transport to headspace equilibrium procedure; f) Purging needle with He; g) Transfer of headspace gases into evacuated exetainer; h) Storage of exetainer in water filled PP tube. See further explanation in text.
The stopcock was switched to the helium connection to add 15 mL of helium into the syringe (Fig. 2d). The stopcock was turned to close the syringe connection and the He and effluent tubing were removed (Fig. 2e) before starting the headspace equilibrium procedure (see below). For each sampling, duplicate syringes were filled with effluent. To achieve equilibrium between dissolved gases with the helium headspace, the two syringes of the transparent water filled container were placed into a small water filled container which was closed carefully and was placed onto a shaker table. The syringes were shaken for 10 min at 450 rpm to equilibrate gases dissolved in the effluent with the He headspace. After equilibration, the container with the syringes was returned to the large transparent container filled with water (Fig. 1). The syringes were removed from the small container under water. A needle and the helium tube were connected to the 3 way stopcock valve (Fig. 2f). After purging the needle with helium, the headspace gas in the syringe (w12 ml of 15 ml) was transferred to a evacuated exetainer (12 mL Labco, UK) (Fig. 2g). The exetainers were then stored in water filled PP tubes (50 ml)
(Fig. 2h). These samples were analysed for 15N2O and 15N2 at the stable isotope facility at the University of California, Davis, USA using a SerCon Cryoprep trace gas concentration system interfaced to a PDZ Europa 20e20 isotope ratio mass spectrometer (Sercon Ltd., Cheshire, UK). Gas samples were analysed within 3 weeks of collection. Prior to sample collection, the exetainers were evacuated with an Edwards RV3 roughing pump and stored in helium-sparged water. The DRpp (R; mg l1 min1) from this experiment was determined using the calculations after Clilverd et al. (2008) as followed: 1 1 )t , DRpp ¼ ((15Ngas(t)eb15Ngas(t0 )) (15 NO 3 %b) 15
Ngas(t) was the concentration of dissolved 15NeN2 and dissolved 15NeN2O at each sampling time (t). 15Ngas(t0 ) was the concentration of dissolved 15NeN2 and dissolved 15NeN2O at time 0 of each assay. To calculate the dissolved 15NeN2 and 15 NeN2O concentrations, total dissolved N2eN and N2OeN concentrations were multiplied by the respective atom % 15N. The respective atom % 15N expresses the fraction of 15N in %
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 4 1 e4 1 5 1
(15N (15N þ 14N)1) of the N gas (N2 or N2O). Dissolved N2eN and N2OeN concentrations were calculated using the Bunsen solubility coefficient from Weiss (1970) and Weiss and Price (1980). b was the dilution factor for each sampling. Dilution was caused by water movement and diffusion. b was calculated by dividing Br concentration at time (t) by the Br 15 concentration at time 0. 15 NO NO 3 % was the percent 3 of total NO3 added to the well at the start of the pushepull test (20%) and t was sampling time in minutes. Lastly, denitrification rates were converted to bed volume by multiplying the effluent volume with the determined tracer porosity of 7 years old woodchips (60%) by Robertson (2010).
2.3.3. In situ denitrification measurement using N2 and N2O concentrations e DRN2 In situ denitrification rates were measured by collecting effluent samples along the length of the bed using the equipment described above and analysing collected effluent for total dissolved N2eN and N2OeN concentrations (DRN2 ) (Figs. 1 and 2). The headspaces of these effluent samples (duplicates) were collected from 8 locations along the bed (0 m, 16 m, 32 m, 48 m, 64 m, 96 m, 128 m, 160 m) and sent to the stable isotope facility at the University of California, Davis, USA for analysis of N2eN and N2OeN concentrations using a SerCon Cryoprep trace gas concentration system interfaced to a PDZ Europa 20e20 isotope ratio mass spectrometer (Sercon Ltd., Cheshire, UK). The denitrification rate (mg N L1 effluent m1 bed length) was calculated from the slope of a linear regression of the dissolved N2eN and N2OeN along the length of the bed. The denitrification rate per unit time and bed volume (g N m3 bed volume d1) was calculated by multiplication of the effluent flow velocity through the bed and a wood chip porosity of 60% (Robertson, 2010). The theoretical flow velocity of the effluent through the bed was determined using Darcy’s law (1856) as followed: v ¼ (V (A p)1)1 where v was the effluent velocity (min m1) through the bioreactor, V was the inlet volume (m3 min1) of the effluent, A(m2) was the cross-sectional area of the bioreactor and p is the tracer porosity of the bed material. The inlet volume of 170 m3 d1, the cross-sectional area of 7.5 m2 and a tracer porosity of 60% (Robertson, 2010) yielded an effluent velocity of 38.1 min m1. 15 2.3.4. Comparison of in situ d15 NdNO 3 increase to d NeN2 decline e DRNA
We compared the natural enrichment of 15N in NO 3 with the depletion of 15N in dissolved N2 in the effluent along the length of the bed. We hypothesised that if denitrification was the main mechanism for NO 3 removal that the rate of change in natural abundance 15N along the length of the bed for NO 3 and N2 would be of similar value but with opposite sign. The natural abundance 15NeN2, expressed as d15NeN2 was determined by collecting equilibrated headspace samples at eight locations along the length of the bed using the equipment described above (Figs. 1 and 2). d15N describes the difference between the 15N/14N ratio of the sample (Rsample) and the 15N/14N ratio (Rstandard) of the natural abundance
4145
standard (air) in & ((RsampleeRstandard) Rstandard1) to allow the detections of small differences in natural abundance of 15N in dissolved gases (d15NeN2; d15NeN2O) and nitrate (d15 NdNO 3 ). The equilibrated headspace samples of duplicate were analysed at the stable isotope facility at the University of California, Davis, USA using a SerCon Cryoprep trace gas concentration system interfaced to a PDZ Europa 20e20 isotope ratio mass spectrometer (Sercon Ltd., Cheshire, UK). Additionally, d15NeN2 was measured at 12 locations along the length of the bed via vacuum extraction of the dissolved gases. This measurement was performed during a different season 3 bed with a similar NO 3 dN removal rate of the bed (7.2 g N m 1 15 material d ) than in the time when d NdNO3 was measured (8.7 g N m3 bed material d1). Samples of water from 1.20 m depth of the bed were withdrawn through a tube into a 50 mL plastic syringe taking care to keep all air out of the sample. A two way tap was placed on the outlet of the syringe and excess water sample purged through the outlet to ensure all parts were filled with the water sample. With the tap turned off, the plunger of the syringe was pulled down to create a vacuum in the syringe and the syringe was shaken vigorously. The created vacuum extracted dissolved gases from the water and these gases formed a bubble in the syringe. Tests with airsaturated water and with isotopically-labelled gases dissolved in water showed that we could obtain at least 90% of the theoretical amount of dissolved gases by this technique and the recovered gases contained close to appropriate gas ratios of air (data not shown). The bubble was displaced into a 1 mL, zero dead space, insulin syringe fitted to the two way tap and previously purged with sample water to eliminate any air contamination. The 1 mL syringe while attached to the apparatus was tipped up so that the small amount of water in it formed a seal and it was withdrawn, the water expelled and needle of syringe pushed into a rubber bung for storage until analysis. This gas sample, normally of 0.3e0.4 mL, was injected into a port in the continuous flow mass spectrometer (PDZ Europa 20-20) and isotopic abundance of N measured. Furthermore, d15 NdNO 3 of water samples was determined using chemical reduction via cadmium and azide to convert NO3 to NO2 and then N2O (McIlvin and Altabet, 2005) and measured on an IsoPrime isotope spectrometer at New Zealand’s National Isotope Centre. This measurement includes nitrite, typically in negligible amounts. The ratio of the rate of 15 increase of d15 NdNO 3 and the decrease of d NeN2 multiplied by the measured NO3 dN removal gives the denitrification rate e DRNA. The negative congruence of the slopes shows the proportion of NO 3 dN removed by microbial denitrification. The fractionation factor associated with denitrification was estimated using the Rayleigh equation by plotting d15 NdNO 3 versus the natural logarithm of the ratio of NO3 dN concentrations along (Ct) and at the inlet of the bed (C0) (Fig. 3). The slope of the linear regression curve through the data points expressed the isotope fractionation factor in &.
2.4.
Statistical analysis
Standard errors (SE) for each DR approach, nitrate removal and fractionation factor were calculated with linear regressions analysis using SAS software version 8.2 (SAS Institute Inc., Cary, USA). Results are shown as DR SE and as SE of the
4146
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 4 1 e4 1 5 1
Fig. 3 e d15 NdNOL 3 measured along the length of the bed versus ln(Ct/C0), where Ct is the NOL 3 dN concentration along the length and C0 at the inlet of the bed). Linear regression line indicates isotopic fractionation factor of L19&.
slope in equations. The mean DRAI and the belonging SE of the DRAI approach were calculated including DRAI estimates resulting of n ¼ 12 measuring points along the length of the bed. Furthermore we compared the different DR approaches by calculating the 95% Wald confidence intervals with the SAS software for each approach as shown in Fig. 6. Bounds express the range of the confidence intervals in Fig. 6. The experimental errors of DRN2 , DRPP and d15NeN2 measurements during sampling, transport and analysis processes are expressed as experimental standard errors (SEexp) and are shown for the sampling points in the Figs. 7e9.
3.
Results
3.1.
Nitrate removal rate
Fig. 5 e In vitro via acetylene inhibition method measured DRAI at 27 C along the length of the bed.
mgL1; x ¼ length of the bed in m) in NO 3 dN concentration along the length of the bed (Fig. 4). This yielded a NO 3 dN removal rate of 8.73 1.45 g N m3 bed material d1 or 11.5 1.9 kg N bed1 d1.
3.2.
Denitrification rates
DRAI ranged from 9.07 mg N h1 g1 dried media of the bed to 20.88 mg N h1 g1 (Fig. 5) averaging 13.9 1 mg N h1 g1. These analysis were conducted at 27 C and when corrected to the bed temperature at the time of sampling (19 C) using a Q10 of 2.1, resulted in a DRAI of 7.94 0.58 mg N h1 g1 (13.9 1 associated factor 1.75) or divided by the Q10 40.02 2.9 g N m3 d1 based on a measured bulk density of 210 kg dried media m3 (volume of bed) (Fig. 6). The in situ pushepull technique with enriched 15 NdNO 3 measured concentrations of dissolved 15N gas (15NeN2 and 15 NeN2O) at locations A (48 m) and B (128 m). There were linear increases with time (15NeN2-gas concentration at
There was a significant linear decline ( y ¼ 0.39 þ173; R2 ¼ 0.78; p < 0.001; SE ¼ 0.064; y ¼ NO 3 dN concentration in
Fig. 4 e Concentrations of NOL 3 dN along the length of the denitrification bed (April). Linear regression line was fitted.
Fig. 6 e Denitrification rates (bars) of different approaches compared with the measured NOL 3 dN removal rate. Bounds express the range of the Wald confidence interval (95%).
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 4 1 e4 1 5 1
Fig. 7 e 15N-Nitrogen gas concentration (top) and 15NeN2O concentration (bottom) at different times determined via pushepull technique at location A and B. Error bar is one experimental standard error (n [ 2).
location A and B: y ¼ 26.9x þ 201; R2 ¼ 0.99; p < 0.001; SE ¼ 1.65 and y ¼ 9.3x429; R2 ¼ 0.86; p ¼ 0.008; SE ¼ 1.90 respectively; where y ¼ 15NeN2-gas concentration in mg L1 and x ¼ time in min) (15NeN2O at location A and B: y ¼ 0.1x0.7; R2 ¼ 0.99; p < 0.001; SE ¼ 0.002 and y ¼ 0.11x4; R2 ¼ 0.97; p < 0.001; SE ¼ 0.01 respectively; y ¼ 15NeN2O-gas concentration in mg L1; x ¼ time in min) (Fig. 7). From these increases, DRpp were calculated as 26.8 1.65 mg N L1 min1 at location A and 9.3 1.9 mg N L1 min1 at location B. Assuming a bed material porosity of 60% (Robertson, 2010) the DRs were converted to 23.2 1.43 g N m3 d1 for site A and 8.06 1.64 g N m3 d1 for site B (Fig. 6). The dissolved N2OeN release was 0.4% and 1.55% respectively of the potentially removed NO 3 dN (Fig. 7). In situ measured dissolved N2eN and N2OeN concentrations increased significantly linear along the length of the denitrification bed ( y1 ¼ 0.27x þ 22.74; R2 ¼ 0.7; p ¼ 0.01; SE ¼ 0.072 and y2 ¼ 0.03x þ 0.16; R2 ¼ 0.99; p < 0.001; SE ¼ 0.001 respectively; y1 ¼ N2-N gas concentration in mg L1; x ¼ length of bed in m; y2 ¼ N2OeN gas concentration in mg L1; x ¼ length of bed in m) (Fig. 8). A DRN2 of 295 71 mg N L1 m1 or 6.7 1.61 g N m3 d1 was calculated based on a theoretical flow velocity of 38.1 min m1 and a woodchip porosity of 60% (Fig. 6). The natural abundance of d15 NdNO 3 of the remaining NO3 increased linearly along the length of the bed ( y ¼ 0.07x þ 6.18; R2 ¼ 0.99; p < 0.001; SE ¼ 0.002; y ¼ d15 NdNO 3 ; x ¼ length in m).
4147
Fig. 8 e Dissolved N2 gas concentration (top) and dissolved N2OeN gas concentration along the length of the denitrification bed. Error bar is one experimental standard error (n [ 2).
15 Fig. 9 e d15 NdNOL 3 , d NeN2 measured via headspace 15 equilibrium and d NeN2 measured via vacuum extraction along the length of the denitrification bed. Error bar is one experimental standard error (n [ 2).
4148
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 4 1 e4 1 5 1
As expected the d15NeN2 decreased along the length of the bed for both the vacuum extracted 15NeN2 ( y ¼ 0.05x2.31; R2 ¼ 0.89; p < 0.001; SE ¼ 0.005; y ¼ d15NeN2; x ¼ length in m) and the via headspace equilibration method gained 15NeN2 ( y ¼ 0.06x2.02; R2 ¼ 0.69; p ¼ 0.011; SE ¼ 0.016; y ¼ d15NeN2; x ¼ length in m) (Fig. 9). A comparison of the slopes of 15 d15 NdNO 3 and the average of the slopes of d NeN2 indicated that about 73.2 23.8% of removed NO3 dN was converted via denitrification to N2 gas. d15 NdNO 3 measured along the length of the bed decreased linearly ( y ¼ 19.4 þ8.2; R2 ¼ 0.78; p < 0.001; SE ¼ 3.2; y ¼ d15 NdNO 3 ; x ¼ ln(Ct/C0)) with the natural logarithm of the dN concentrations along (Ct) and at the inlet of ratio of NO 3 the bed (C0). The slope of this linear regression indicated an of isotopic fractionation of natural abunded 15 NdNO 3 19.4 3.2& (Fig. 3).
4.
Discussion
3 The NO bed material d1) of this study 3 removal rate (g N m was similar to NO3 removal previously measured in this denitrification bed (Warneke et al., 2011), and in the range of NO 3 dN removal of other denitrification beds (Schipper et al., 2010). The majority of NO 3 dN removal rates measured in denitrification beds have been calculated from the difference in total NO 3 dN inputs (concentration multiplied by inlet flow rates) and total NO 3 dN at the outlet divided by bed volume. It is likely that these measured NO 3 dN removal rates are imprecise estimates of the average NO 3 dN removal due to variability in NO 3 inlet concentrations and time taken for effluent to move to the outlet. Consequently, the measured outflow NO 3 dN concentration may not reflect the NO3 dN removal of the measured inlet NO3 dN concentration. Therefore, it was necessary to develop a method to obtain reliable rates of NO 3 removal and to determine whether this NO 3 dN removal was due to denitrification or other removal processes, such as DNRA, precipitation, absorption and/or biotic uptake. These other processes lead to temporary removal of NO 3 from the effluent and can be subsequently released whereas denitrification represents a permanent N sink. A substantial advantage of this study was that we could obtain reliable results for the NO 3 dN removal rate due to only one effluent inlet, an almost constant inlet flow rate (149 8 m3 d1; 7 measurements year1) and constant inlet 1 solute concentration (146 13 mg NO 3 dN L ; 7 measurements year1), the construction of the bioreactor (only open to the top) and the number of NO 3 dN concentration measurements along the length of the bed (n ¼ 12). Therefore we were able to compare the rate of NO 3 dN removal with the measured denitrification rates using different techniques.
4.1. Microbial denitrification was the main NO 3 dN removing process in the bed There have been measurements of NO 3 removal in denitrification beds, and it has often been assumed that microbial denitrification was the main mechanism for NO 3 dN removal (Greenan et al., 2006, 2009; Robertson et al., 2000; Robertson, 2010; Schipper et al., 2010; Moorman et al., 2010; Warneke
et al., 2011), but none of these studies have directly measured denitrification rates via products of the denitrification process (N2 and N2O). Using four different approaches for assessing the importance of denitrification for NO 3 removal we found persuasive evidence that NO 3 removal was largely due to denitrification. Both in situ measurements of dissolved 15NeN2 via pushepull test (DRPP) and the direct measurement of dissolved N2 gas concentrations DRN2 along the length of the bed resulted in denitrification rates that were sufficient to account for the actual NO 3 dN removal of the bed (Fig. 6). Furthermore, the increase of dissolved d15 NdNO 3 and the decline of dissolved d15NeN2 along the length of the bed were almost negatively congruent (Fig. 9) suggesting that at least 73% of NO 3 dN removal was due to production of dissolved N2. This was likely an underestimate of the proportion of NO 3 removal due to denitrification because N2O production rates could be not included in this calculation. Anammox could be eliminated as þ a significant NO 3 dN removal process as NH4 concentrations were always low (Warneke et al., 2011). The fractionation factors of 15 NO 3 caused by denitrification (enrichment of 15 NdNO 3 in the system) are highly variable and depend on various conditions in a denitrifying system e.g., NO 3 concentration, concentration and microbial availability of carbon (electron donor), temperature, denitrification rate, species of the denitrifying bacterium (Mariotti et al., 1982; Bryan et al., 1983; Macko et al., 1987). The calculated fractionation factors of 15 NdNO 3 along the length of our study site was 19.4 3.2&, similar to the fractionation factors measured in laboratory studies under optimal denitrifying conditions by Blackmer and Bremner (1977) (17&) and by Mariotti et al. (1981) (29.4&). Our fractionation factor is greater than the fractionation factors calculated by Robertson et al. (2000) for two denitrification bed (7.8 and 5.7&) and calculated by Robertson (2010) in a woodchip column study (13&). Furthermore, fractionation factors determined in other field studies were also lower due to nitrogen transformations other than denitrification, e.g., Spalding and Parrott (1994) determined a fractionation factor of 9.6& in a groundwater study and Lund et al. (2000) a fractionation factor of 2.5& in a large constructed wetland. This is further evidence that microbial denitrification was responsible for the NO 3 removal in our study. The higher fractionation factor in our study may be due available carbon limitation of the denitrification process (electron donor) and the consistently high concentration of electron acceptor (NO 3) (Mariotti et al., 1981).
4.2. Evaluation of different approaches to determine denitrification rates of the bed Our second objective was to determine the advantages and disadvantages of these different approaches for measuring denitrification rates in denitrification beds. This is the first study that has compared denitrification rates measured using different techniques with known rate of NO 3 removal in a denitrification bed. The DRN2 measurement technique resulted in a denitrification rate close to the measured NO 3 dN removal rate in the denitrification bed, whereas the DRPP at location A and the
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 4 1 e4 1 5 1
DRAI approach were significantly higher than the measured NO 3 dN removal rate (Fig. 6). The DRN2 method is based on measuring the increase of dissolved N2 and N2O along the length of the bed and required the assumption that the N2 emission was constant along the length of the bed to calculate a reliable denitrification rate. N2O gas emissions measured in chambers along the length of the bed showed no trend and consequently support this assumption (Warneke et al., 2011). The results suggest that the DRN2 method was accurate, inexpensive and easy to use; however, the water sampling must be done with conservable care to avoid N2 contamination from the atmosphere. Furthermore this method allows determination of dissolved N2O production, which was about 2 times greater than the annual average total N2O release from this denitrification bed (Warneke et al., 2011) and could be in the range of N2O variations over months. The temperature adjusted DRAI was highly variable along the length of the bed (ranging from 26.1 to 60.1 g N m3 bed material d1) similar to previous measurements at this site (Warneke et al., 2011) and generally 5 times greater than the 3 bed material d1; calculated NO 3 removal rate (8.7 g N m Fig. 6). This corresponds with the findings of Bernot et al. (2003) who measured significantly lower denitrification rates using the membrane inlet mass spectrometry technique than using the acetylene-block technique without adding chloramphenicol in aquatic systems. Although the acetylene inhibition technique is simple to perform, allows for a high degree of replication, and is not influenced by atmospheric N2 contamination, it is likely to overestimate denitrification due to favourable incubation conditions (constant shaking of samples, constant temperature) and the addition of acetylene could act as a carbon source (Kanner and Bartha, 1979; De Bont and Peck, 1980; Tam et al., 1983; Schink, 1985). Furthermore, this method does not distinguish between N2 and N2O production. Consequently DRAI has limited utility for estimating denitrification rates in denitrification beds, but can be used to determine whether C or N is limiting denitrification (e.g., Warneke et al., 2011) and it is also useful as a comparative index of denitrification activity across different sites or seasons (Groffman et al., 1992, 2006). The pushepull technique was a reasonable method for determining the denitrification rate at a specific location within the bed; however, there was considerable variability of DRPP between the two sampling locations. DRPP was almost 3 times greater at location A than at the location B. This magnitude of variability between location A and B was similar to the variability of DRAI at the same measurement locations (Fig. 5), but average DRAI was more than 2 times greater than DRPP. In contrast, Moorman et al. (2010) measured DRPP in a denitrifcation wall in the same range as measurements of DRAI. The Wald confidence interval (95%) of average DRPP of the bed ranged from 13.5 g N m3 to 17.8 g N m3 and was significantly greater than the measured NO 3 removal rate. Multiple pushepull tests along the bed would be required to obtain an average denitrification rate of the entire bed. This approach required substantial effort and resources to minimize N2 contamination risk and to obtain sufficient samples of dissolved gases at each location (minimum 16 per site), which can be costly. Furthermore, the effluent flow through the bed had to be stopped during each experiment, which may have
4149
altered the denitrifying conditions of the bed. Denitrification in the bed was previously shown to be C limited (Warneke et al., 2011), so the injection of additional 15 NdNO 3 was unlikely to alter the denitrification process in our experiment. However, addition of NO 3 would be an important consideration in systems where NO 3 concentrations were much lower and potentially limiting. Dissolved N2O production also could be determined via pushepull technique and was in the range of the measured N2O emission via chambers by Warneke et al. (2011) at the denitrification bed (around 1% of the removed NO 3 dN). Summing up, the DRPP method yielded reliable rates, but was a time consuming and expensive approach for measuring denitrification rates in denitrification beds particularly in comparison to the DRN2 approach. Measurements of natural abundance d15N-gas and 15 d NdNO 3 along the length of the bed measured a DRNA that accounted for 73.2 23.8% of NO 3 dN removal. This method was also susceptible to contamination by atmospheric N2 due to very small increases in dissolved 15NeN2-gas. Additionally, the NO 3 dN removal rate based on inflow and outflow NO3 concentration was needed to calculate DRNA using this method. However, this approach was useful for determining whether the main pathway of NO 3 dN removal was microbial denitrification.
5.
Conclusions
All four approaches for measuring denitrification rates supported the hypothesis that microbial denitrification was the main mechanism for NO 3 dN removal in the denitrification bed. Our results suggest that the DRN2 approach was useful for obtaining reliable NO 3 removal rates and was superior to the other investigated techniques for determining denitrification rates. This method may allow rapid measurement of denitrification rates for a variety of beds constructed in different locations with different carbon substrates and NO 3 loadings and for similar aquatic systems, so long as care is taken to avoid atmospheric contamination of samples. In situ pushepull experiments with enriched 15 NdNO 3 (DRPP) were more intensive and only useful for measuring denitrification rates at specific locations when NO 3 was non-limiting. Denitrification rates determined via changes in natural abundance 15 of d15 NdNO 3 and d NeN2 (DRNA) required data on NO3 dN removal rates to determine the proportion of NO3 removed by denitrification and was prone to atmospheric N2 contamination. The DRAI method largely overestimated the denitrification rate of our system, but was useful for determining whether denitrifcation was C or N limited in the bioreactor. Future denitrification rate measurements can also identify where the highest DR activity is in a bed (depth and length of the bed), and may lead to recommendations about the construction size of the bed.
Role of the funding source WaikatoLink Ltd (New Zealand) and Hans-Sauer-Foundation (Germany) funded this study. They were not involved in project design, sampling, data analysis/interpretation and writing the manuscript.
4150
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 4 1 e4 1 5 1
Acknowledgements This research was funded by WaikatoLink Ltd. (New Zealand) and Hans-Sauer-Foundation (Germany). We thank Yvonne Tay, Jacinta Parenzee, Chris McKinnon, Peter Jarman, Annette Rodgers, Craig Hosking, Gerhard Bartzke, Rafael Guedes, Dr Susanna Rutledge, Dr Ralph Boch, Marco Rothfeld and Carl Ebbers for their support. Furthermore, comments by anonymous reviewers and the handling editor improved the manuscript.
references
Addy, K., Kellogg, D.Q., Gold, A.J., Groffman, P.M., Ferendo, G., Sawyer, C., 2002. In situ pushepull method to determine ground water denitrification in riparian zones. J. Environ. Qual. 31, 1017e1024. American Public Health Association, 1992. Standard Methods for the Examination of Water and Wastewater, 18 ed. USA. Baker, M.A., Vervier, P., 2004. Hydrological variability, organic matter supply and denitrification in the Garonne River ecosystem. Freshw. Biol. 49, 181e190. Bedard-Haughn, A., van Groenigen, J.W., van Kessel, C., 2001. Tracing 15N through landscapes: potential uses and precautions. J. Hydrology 272, 175e190. Bernot, M.J., Dodds, W.K., Gardner, W.S., McCarthy, M.J., Sobolev, D., Tank, J.L., 2003. Comparing denitrification estimates for a Texas estuary by using acetylene inhibition and membrane inlet mass spectrometry. Appl. Environ. Microb. 69 (10), 5950e5956. Blackmer, A.M., Bremner, J.M., 1977. Nitrogen isotope discrimination in denitrification of nitrate in soils. Soil Biol. Biochem. 9, 73e77. Butterbach-Bahl, K., Willibald, G., Papen, H., 2002. Soil core method for direct simultaneous determination of N2 and N2O emissions from forest soils. Plant Soil 240, 105e116. Bryan, B.A., Shearer, G., Skeeters, J.L., Kohl, D.H., 1983. Variable expression of the nitrogen isotope effect associated with denitrification of nitrite. J. Biol. Chem. 258, 8613e8617. Canfield, D.E., Glazer, A.N., Falkowski, P.G., 2010. The Evolution and future of earth’s nitrogen cycle. Science 330, 192e196. Clilverd, H.M., Jones, J.B., Kielland, K., 2008. Nitrogen retention in the hyporheic zone of a glacial river in interior Alaska. Biogeochemistry 88, 31e46. Darcy, H., 1856. Les Fontaines Publiques de la Ville de Dijon. Dalmont, Paris. De Bont, J.A.M., Peck, M.W., 1980. Metabolism of acetylene by Rhodococcus A1. Arch. Microbiol. 127, 99e104. Dinnes, D.L., Karlen, K.L., Jaynes, D.B., Kaspar, T.C., Hatfield, J.L., Colvin, T.S., Cambardella, C.A., 2002. Nitrogen management strategies to reduce nitrate leaching in tile-drained Midwestern soils. Agron. J. 94, 153e171. Galloway, J.N., Townsend, A.R., Erisman, J.W., Bekunda, M., Cai, Z., Freney, J.R., Martinelli, L.A., Seitzinger, S.P., Sutton, M.A., 2008. Transformation of the nitrogen cycle: recent trends, questions and potential solutions. Science 320, 889e892. Galloway, J.N., Dentener, F.J., Capone, D.G., Boyer, E.W., Howarth, R.W., Seitzinger, S.P., Asner, G.P., Cleveland, C.C., Green, P.A., Holland, E.A., Karl, D.M., Michaels, A.F., Porter, J.H., Townsend, A.R., Vorosmarty, C.J., 2004. Nitrogen cycles: past, present, and future. Biogeochemistry 70, 153e226. Greenan, C.M., Moorman, T.B., Kaspar, T.C., Parkin, T.B., Jaynes, D.B., 2006. Comparing carbon substrates for denitrification of subsurface drainage water. J. Environ. Qual. 35, 824e829.
Greenan, C.M., Moorman, T.B., Parkin, T.B., Kaspar, T.C., Jaynes, D.B., 2009. Denitrification in wood chip bioreactors at different water flows. J. Environ. Qual. 38, 1664e1671. Groffman, P.M., Altabet, M.A., Bohlke, J.K., Butterbach-Bahl, K., David, M.B., Firestone, M.K., Giblin, A.E., Kana, T.M., Nielsen, L. P., Voytek, M.A., 2006. Methods for measuring denitrification: diverse approaches to a difficult problem. Ecol. Appl. 16, 2091e2122. Groffman, P.M., Gold, A.J., Simmons, R.C., 1992. Nitrate dynamics in riparian forests: microbial studies. J. Environ. Qual. 21, 666e671. Hauck, R.D. and Melsted, S.W, (1956) Some aspects of the problem of evaluating denitrification in soils. Soil Science Society of America Proceedings 20, pp. 361e364. Kanner, D., Bartha, R., 1979. Growth of Nocardia rhodochrous on acetylene gas. J. Bacteriol. 139, 225e230. Laursen, A.E., Seitzinger, S.P., 2005. Limitations to measuring riverine denitrification at the whole reach scale: effects of channel geometry, wind velocity, sampling interval, and temperature inputs of N2-enriched groundwater. Hydrobiologia 545, 225e236. Lund, L.J., Horne, A.J., Williams, A.E., 2000. Estimating denitrification in a large constructed wetland using stable nitrogen isotope ratios. Ecol. Eng. 14, 67e76. Macko, S.A., Fogel, M.L., Hare, P.E., Hoering, T.C., 1987. Isotopic fractionation of nitrogen and carbon in the synthesis of amino acids by microorganisms. Chem. Geol. 65, 79e92. Mariotti, A., Germon, J.C., Hubert, P., Kaiser, P., Letolle, R., Tardieux, A., Tardieux, P., 1981. Experimental determination of nitrogen kinetic isotope fractionation: some principles: illustration for the denitrification and nitrification processes. Plant Soil 62, 413e430. Mariotti, A., Germon, J.C., Leclerc, A., 1982. Nitrogen isotope fractionation with the NO2/N2O step of denitrification in soils. Can. J. Soil Sci. 62, 227e241. McIlvin, M.R., Altabet, M.A., 2005. Chemical conversion of nitrate and nitrite to nitrous oxide for nitrogen and oxygen isotopic analysis in freshwater and seawater. Anal. Chem. 77 (17), 5589e5595. Moorman, T.B., Parkin, T.B., Kaspar, T.C., Jaynes, D.B., 2010. Denitrification activity, wood loss, and N2O emissions over 9 years from a wood chip bioreactor. Ecol. Eng. 36, 1567e1574. Robertson, W.D., 2010. Nitrate removal of woodchip media of varying age. Ecol. Eng. 36, 1581e1587. Robertson, W.D., Blowes, D.W., Ptacek, C.J., Cherry, J.A., 2000. Long-term performance of in situ reactive barriers for nitrate remediation. Ground Water 38 (5), 689e695. Schink, B., 1985. Fermentation of acetylene by an obligate anaerobe, Pelobacter acetylenicus sp. nov. Arch. Microbiol. 142, 295e301. Schipper, L.A., Robertson, W.D., Gold, A.J., Jaynes, D.B., Cameron, S.C., 2010. Denitrifying bioreactors e an approach for reducing nitrate loads to receiving waters. Ecol. Eng. 36, 1532e1543. Seitzinger, S., Harrison, J.A., Bo¨hlke, J.K., Bouwman, A.F., Lowrance, R., Peterson, B., Tobias, C., Van Drecht, G., 2006. Denitrification across landscapes and waterscapes: a synthesis. Ecol. Appl. 16, 2064e2090. Spalding, R.F., Parrott, J.D., 1994. Shallow groundwater denitrification. Sci. Total Environ. 141, 17e25. Tam, T.Y., Mayfield, C.I., Inniss, W.E., 1983. Aerobic acetylene utilization by stream sediment and isolated bacteria. Curr. Microbiol. 8, 165e168. Tiedje, J.M., Simkins, S., Groffman, P.M., 1989. Perspectives on measurement of denitrification in the field including recommended protocols for acetylene based methods. Plant Soil 115, 261e284. Vitousek, P.M., Aber, J., Howarth, R.W., Likens, G.E., Matson, P.A., Schindler, D.W., Schlesinger, W.H., Tilman, G.D., 1997. Human
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 4 1 e4 1 5 1
alteration of the global nitrogen cycle: causes and consequences. Ecol. Appl. 7, 737e750. Vymazal, J., Greenway, M., Tonderski, K., Brix, H., Mander, U., 2006. Constructed wetlands for wastewater treatment. In: Verhoeven, J.T.A., Beltman, B., Bobbink, R., Whigham, D.F. (Eds.), Wetlands and Natural Resource Management, pp. 69e96.
4151
Warneke, S., Schipper, L.A., Bruesewitz, D.A., McDonald, I., Cameron, S., 2011. Rates, controls and potential adverse effects of nitrate removal in a denitrification bed. Ecol. Eng. 37, 511e522. Weiss, R.F., 1970. The solubility of nitrogen, oxygen and argon in water and seawater. Deep Sea Res. 17, 721e735. Weiss, R.F., Price, B.A., 1980. Nitrous oxide solubility in water and seawater. Mar. Chem. 8, 347e359.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 5 2 e4 1 6 8
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Variability of fresh- and salt-water marshes characteristics on the west coast of France: A spatio-temporal assessment Se´bastien Tortajada a,*, Vale´rie David a, Amel Brahmia a, Christine Dupuy a, Thomas Laniesse a, Bernard Parinet b, Frederic Pouget a, Frederic Rousseau a, Benoit Simon-Bouhet a, Franc¸ois-Xavier Robin c a
LIENSs, UMR 6250 Universite´ de La Rochelle e CNRS, 2 rue Olympe de Gouge, 17000 La Rochelle, France LCME, UMR 6008 Universite´ de Poitiers e CNRS, 15 rue de l’Hoˆtel Dieu, 86034 Poitiers Cedex, France c Union des marais de Charente-Maritime, rue Jacques de Vaucanson Zone Industrielle de Pe´rigny, 17180 Pe´rigny, France b
article info
abstract
Article history:
The degradation of water quality and the multiple conflicts of interest between users make
Received 1 December 2010
marsh restoration very important. A Water Quality Evaluation System (WQES) was developed
Received in revised form
for river systems by the European Water Framework Directive (WFD). Some form of
14 March 2011
biologically-based, habitat-specific reference standard seems absolutely essential for wise
Accepted 22 May 2011
management and stewardship of marsh ecosystems. The goal of this study was to develop
Available online 31 May 2011
a statistical method to define and to characterize a water body typology for drained marshes of the Charente-Maritime wetlands on the French Atlantic coast, placing particular emphasis
Keywords:
on environmental factors as hydraulic functioning, human activities and pedological
Marshes
substratum. The Charente-Maritime marshes represent a good field study because of his high
Typology
diversity of types of marshes and of anthropogenic activities in a restrictive area thus erasing
Hydrodynamic
spatial climatic effect (latitude effect). The statistical method developed here had permitted
Eutrophication
to define and characterize 12 different water bodies, 7 in freshwater (F1 to F7) and 5 in salt
Nitrate removing
water marshes for the Charente-Maritime area. This typology demonstrated an important
Catchment basin
link between the size catchment area, nitrate concentrations, and leaching of precipitation from cultured soils. Even though the Charente-Maritime marshes are strongly impacted by humans, they may still retain the ability to remove nitrate. The increasing gradient of water renewal in the freshwater marshes from F1 to F7 explained the decreasing gradient of eutrophication. A better management of the hydrodynamic of the marshes can avoid eutrophication risk on the coastal sea area. Reliance on the WFD parameter set necessarily placed limits on the kinds of interpretations that could be made and on the study’s potential contribution to the basic science of marshes. Ecologically-based insights regarding both external flows (links between ecosystems, meta-ecosystem theory) and internal flows (structure of the planktonic food web) seem an essential prerequisite for further advances in the study of marsh ecosystems. ª 2011 Elsevier Ltd. All rights reserved.
* Corresponding author. Tel.: þ33 5 46 50 76 54; fax: þ33 5 46 50 76 63. E-mail addresses:
[email protected] (S. Tortajada),
[email protected] (V. David),
[email protected] (A. Brahmia),
[email protected] (C. Dupuy),
[email protected] (T. Laniesse),
[email protected] (B. Parinet), frederic.
[email protected] (F. Pouget),
[email protected] (F. Rousseau),
[email protected] (B. Simon-Bouhet), fx.robin@ unima.fr (F.-X. Robin). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.05.024
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 5 2 e4 1 6 8
1.
Introduction
Increases in human population and levels of industrialization have produced growing demands for more water of better quality. Over time, requirements for water quantity and quality have emerged to provide water for drinking, personal hygiene, agriculture, industry, energy production and many additional purposes related to essential human needs (Meybeck and Helmer, 1996). Unfortunately, anthropogenic activities also impact water quantity and quality (Lotze et al., 2006) and therefore determine human access to potable water (Cominelli et al., 2009). In recognition of this basic human need, industrialized nations have in the past few years developed programs to restore the quality and the quantity of their natural freshwater and saltwater resources. The Ecological Society of America’s research priorities for ecology have been in effect since 1988 (Lubchenco et al., 1991). Through its Water Framework Directive (WFD), the European Union has provided a holistic framework since 2000 for management and protection of all water bodies under its jurisdiction (No˜ges et al., 2009). Ecosystems provide goods and services to human populations. Wetlands serve very effectively to restore water quality (Millenium Ecosystem Assessment, 2005). Compared with other ecosystems, wetlands rank first for water supply and as habitat and refuge for organisms. They also rank first for waste treatment or nutrient cycling. They rank second for water regulation (Costanza, 1997). Sixty-seven percent of wetland surface area has been lost during the last 150 years. The result has been an inevitable degradation of water quality (Lotze et al., 2006), yet restoration of only 5% of a wetland area will purify 40% of nitrates introduced by agricultural activity (Verhoeven et al., 2006). Marshes represent 50% of wetland area and provide 75% of the total services furnished by these systems (Costanza, 1997). Drainage of wetlands for human use through history has left few pristine freshwater or saltwater marshes. Today, the former marshlands are human-controlled ecosystems. Their hydrological functions depend now on human activities and decisions. The degradation of water quality and the multiple conflicts of interest between users make marsh restoration very important. In Europe, the Water Quality Evaluation System (WQES) developed for river systems by the WFD is also applied to marshes because standard was not been developed for marshes. Majority of the study on freshwater marshes focused on macrofauna (birds and mammals) and on macrophytes diversity (Weller, 1978; Benoit and Askins, 1999; Lougheed et al., 2001). Freshwater marshes water quality is poorly studied all around the world and considered, in most cases, only one type of marshes (Rozas and Odum, 1988; Ahn and Mitsch, 2002; Mitsch et al., 1995; Rojo et al., 2010). Some form of biologically-based, habitat-specific reference standard seems absolutely essential for wise management and stewardship of marsh ecosystems and needs to be based on scientific knowledge of marshes functioning (No˜ges et al., 2009). The first step in this understanding is to realize a water body typology presenting different functioning like the WFD recommended it (No˜ges et al., 2009). A good field
4153
study with a high diversity of marshes types and a high diversity of anthropogenic activities in a restrictive area without latitudinal effect (latitude effect), is needed to develop standard. The Charente-Maritime marshes (Atlantic Coast, France) are thus appropriate models to develop this standard since they present different types of environments (fresh, brackish or salt water) having a large range of hydrological controls (presence or absence of tidal effects, presence or absence of anthropogenic replenishment). The large range of human activities that they support include shellfish farming, agriculture, water purification facilities and cattle husbandry. Indeed, this area (including the ‘Marennes-Ole´ron’ Bay) is one of France’s chief oyster-producing areas (Goulletquer and Heral, 1997). These contrasting human activities have produced several conflicts of interest over water quantity and quality (e.g., conflict between shellfish aquaculture and cereal farming). For example, oyster maturing are realized in ponds, downstream cereal production. The intensive irrigation and the use of pesticides provoked a deterioration of the water quality and quantity used to refill oyster ponds. This deterioration had strong impact on the oyster maturing (Gagnaire et al., 2007, 2006). To realize a water body typology presenting different functioning on Charente-Maritime marshes, a strong statistical approach is proposed in this study in the perspective to be applicable at every kind of wetland zone. To better understand the functioning of each type of marshes, the statistical approach comprised a characterization step relating water body typology and environmental factors (i.e. anthropogenic activities, hydraulic functioning.). Our data source is a fiveyear survey database collected by stakeholders based on the standard set of physico-chemical and biological parameters proposed by the WQES. Our results will suggest meaningful assumptions about the biological functions of these marshes. Additional parameters will also be proposed whose clever use might well yield enhanced scientific understanding of the biological, bio-hydrological and eco-ethological functions of marshland ecosystems.
2.
Materials and methods
2.1.
Study site
The Charente-Maritime marshes of the French Atlantic coast (46 100 North, 01 120 West) are the second-largest French wetland zone (over 100 000 ha). These marshes have been used for salt production since Roman times (Talureau, 1965). From the 12th to the 14th century these impoundments, progressively isolated from the sea, were invaded by freshwater and drained for salt production. From the 17th to the 19th century, further drainage allowed a major expansion of agriculture in the former marshlands (Talureau, 1965; Billaud, 1984). The coastal salt industry prospered until the 17th century. Since the 18th century the historic coastal salt culture has been replaced by oyster culture. Today the department of Charente-Maritime is the premier oyster producer in all of Europe (Lemonnier, 1980). This geographical area exhibits
4154
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 5 2 e4 1 6 8
a high diversity of marsh types and anthropogenic uses. For example, one may find both tidal and non-tidal freshwater or brackish marshes. The area includes drained floodplain and drained marshes, both replenished and unreplenished. Tidal saltwater marshes are divided into multiple ponds. These ponds support a variety of human activities and facilities including oyster and fish farming, cattle husbandry, industrial plants, water purification facilities and saltern ponds. Each form of development produces different sorts of impacts on the water quality of the main channel. The drained marshes constitute a very significant artificial hydrographic network of channels and ditches (linear stretches of several thousand km). Locks control this network so as to prevent drying and flooding throughout the year. Drying of replenished marshes is also limited by a replenishment channel that brings water from the Charente River during the summer. From North to South, marshes occupy six geographical zones (Fig. 1): the Se`vre Niortaise River marshes (unreplenished, drained marshes), the Re´ Island marshes (tidal marshes), the Rochefort marshes (replenished, drained marshes), the Seudre Estuary (tidal marshes), Ole´ron Island (tidal marshes) and the Gironde Estuary (non-tidal marshes).
2.2.
Sampling strategy
A water quality monitoring program has been in place since spring 2003 on 51 stations located in all sections of the six geographical zones of marshes (Fig. 1). The stations were selected to represent (i) different types of marshes (tidal or non-tidal, drained or not drained, fresh, brackish or salt water); (ii) different soil uses (e.g., aquaculture, agriculture, or urban uses); (iii) different outlets (littoral, river, or channel); (iv) different anthropogenic impacts (e.g., farming or swimming); or (v) potential replenishment during summer. The methodology is based on protocols recommended by the WQES that were defined for the European WFD. From the overall collection of potential indicators proposed by the WQES, stakeholders selected several physicochemical and biological parameters to consider in the present study. Physicochemical parameters included temperature (T C), conductivity/salinity (cond), dissolved oxygen concentrations, dissolved oxygen saturation (O2%), biological oxygen demand (BOD, standard NF EN 1899-2), concentrations of dissolved organic carbon (DOC standard NF EN 1484), suspended particulate matter (SPM, standard NF EN 872), nitrates (NO3, standard NF EN ISO 13395), nitrites (NO2, standard NF EN ISO 13395), phosphates (PO4, standard ISO 15681-2), and ammonium (NH4, (Aminot and Ke´rouel, 2004)). Biological parameters included chlorophyll a (Chl a) (Lorenzen 1967), pheopigment (Pheo) (Lorenzen 1967) and fecal bacteria (Escherichia coli and Enterococcus, standards NF EN ISO 9308-3 and 7899-1) concentrations. All parameters were measured six times per year (‘winter’ period -November, January, March- and ‘summer’ period - June, August, September)., except for the pigment concentrations (Chl a and Pheo), which were sampled only during ‘summer’ periods (three samples a year).
2.3.
Statistical analyses
Three databases were available. Two annual databases (6 months a year, 5 years, 51 stations) did not include pigment concentrations: one for fresh to brackish water marshes (AFM database) and one for salt marshes (ASM database). The summer database or S database (3 months a year, 5 years, 51 stations) included pigment concentrations. The statistical method was summarized in Fig. 2.
2.3.1.
Fig. 1 e Map representing the 6 geographical zones of the marshes of Charente Maritime department and the stations at which stakeholders collected samples (1, 2, 3.71).
Step 1: water body typology (Fig. 2)
The regionalization method, adapted from (Souissi et al., 2000), was used to address spatio-temporal heterogeneity. This numerical analysis method for time series is based on successive Principal Component Analyses (PCA) and cluster analyses. A final cluster was obtained by using a similarity matrix that was used for subsequent PCA and cluster analyses. Correlations were prescreened before applying the time-series cluster method to eliminate redundancy among variables that would otherwise have ascribed excessive importance to multiple variables statistically associated with the same physical or biological quantity. For each node in the final cluster, an Approximately Unbiased (AU) p-value was calculated from successive random resampling (pvclust package for R software) (Suzuki and Shimodaira, 2006) to evaluate the uncertainty associated with the cluster. The
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 5 2 e4 1 6 8
4155
Fig. 2 e Diagram of the analysis steps that led to identify and characterize the water body typology.
regionalization method was applied on the AFM and the ASM databases.
2.3.2.
2.3.3.
Step 3: characterization of each water body (Fig. 2)
The characterization of each water body was done with the discriminant parameters found thanks to the FDA.
Step 2: discriminant parameters (Fig. 2)
The regionalization method allowed us to discriminate among stations of the AFM and ASM databases that belonged to different water body groups without identifying the actual discriminating parameters (or variables). A Factorial Discriminant Analysis (FDA) was thus performed on the AFM and ASM databases modified to erased the temporal variability (mean annual values for each station) and to include groups found by the regionalization method (MFMG: mean freshwater marshes including groups; MSMG: mean saltwater marshes including groups); an FDA is a constrained Principal Component Analysis (PCA) in which groups are predefined.
2.3.3.1. Step 3a: characterization of each source of explanatory variation. To evaluate the relative importance of each source of spatio-temporal variability in explaining the fluctuations of physicochemical characteristics, a nested ANOVA design (Sokal and Rohlf, 1995) was applied to each parameter using the AFM and the ASM databases including groups. Three hierarchical levels (fixed factors) were considered. The nested model was structured according to groups (discriminated by the water body typology) within months within years. Interactions among factors were assumed negligible. This method allowed us to determine the overall significance of the factors
4156
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 5 2 e4 1 6 8
as well as the relative importance of each factor in explaining fluctuations of the discriminant parameters (Dagne´lie, 1975). Only these parameters were treated in the analysis.
2.3.3.2. Step 3b: spatial variability. For each discriminating parameter and each group (MFMG, MSMG and MSG databases), a boxplot displaying the mean and the standard deviation was constructed to identify the chief differences between groups. The boxplots provided a schematic overview of the variability of the different parameters for each group. Unfortunately, the method used to calculate the statistics displayed on the boxplots included an inherent bias. To compare different stations in the same group, we calculated a mean for each station using the entire database and therefore lost certain information (temporality). The strength of the regionalization method is that it accounts for temporal variability. A better understanding of the functional dynamics of each water body had thus been obtained from an additional study of temporal variability. 2.3.3.3. Step 3c: temporal variability. Year-to-year and seasonal variations of physicochemical parameters were derived for each group using the seasonal partition CENSUS 1 (additive model) (David et al., 2005). A Three-Mode Principal Component Analysis (PCA) (Beaugrand et al., 2000, Goberville et al., 2010) was
performed independently on each time series (year-to-year and seasonal series). This method is based on three different PCA applied to three different tables/modes: parameter mode (variables discriminating groups), spatial mode (water body typology, i.e., Fig. 2) and temporal mode (seasonal or year-to-year variability). This method allowed us to identify the principal trends in the time series and to compare them to the precipitation information to which the seasonal partition CENSUS 1 was applied (Me´te´o France data). For each parameter and each group a two-way ANOVA was performed to analyze seasonal variation (months, first factor) and intra-group variability (stations, second factor).
2.3.4. Step 4: environmental factors generating the water body typology (Fig. 2) An environmental factor is defined in this study as any natural or anthropogenic parameter having a potential impact on water quality. Such factors were relative to 1) hydraulic functioning, 2) human activities and 3) pedological substratum (Table 1). A factor can be measured using different metrics that represent the quantifiable values describing the factor state (Table 1). The first task in the analysis was to quantify all these metrics for each marsh using GIS (Geographic Information System) software (ArcGis). Two databases were compiled, one for freshwater marshes and one for salt marshes. Data were available for
Table 1 e List of environmental factor taken into account. yes: Available data, no: no available data. Sources of data: (1) Union des Marais de Charente Maritime, (2) Syndicat des eaux de Charente Maritime, (3) Institut Ge´ographique National, (4) Direction De´partementale de l’Agriculture et des Foreˆts, (5) Institut Franc¸ais de Recherche pour l’Exploitation de la Mer (6) Agence de l’Eau Loire-Bretagne and Agence de l’Eau Adour-Garonne (7) Bureau de recherches ge´ologiques et minie`res. Factors Hydraulic functionning Web hydrographic structure
Refeeding
Human activities Land cover
Water purification plant Pedological substratum Pedological substrateum Nature
Metrics
Fresh marshes
Salt marshes
Surface catchment basin (ha)(1) Ratio between marsh surface and catchment basin surface (M/BC)(1) Channels density (m ha1)(1) Distance to the see in meter(1) Position on the web: Primary channel or Secondary channel(1) Percentage of primary channels(1) Ground water replenished(2) (10: High replenishement, 6.6: middle replenishement, 3.3: slow replenishement, 0: no replenishement) Charente replenished(1)(0: no replenishement, 5: replenishement, 10: high replenishement) NO3 concentration on ground water(2)
Yes Yes
Yes Yes
Yes No No Yes Yes
No Yes Yes No No
Yes
No
Yes
No
Yes Yes Yes Yes No No No Yes
Yes Yes Yes Yes Yes Yes Yes Yes
Yes Yes Yes Yes
No No No No
Percentage of construction(3) Percentage of wood(3) Percentage of cultur(4) Percentage of meadow(4) Percentage of saltern ponds(5) Percentage of shellfish culture ponds(5) Percentage of non exploited ponds(5) Number of water purification plant by ha(6) Percentage Percentage Percentage Percentage
of limestone(7) of peat(7) of alluvium(7) of silt(7)
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 5 2 e4 1 6 8
32 on 38 fresh marshes and for all salt marsh stations. An FDA was then performed for each database using ‘Groups’ as a constraint in order to determine which factors discriminated our group of stations. An ANOVA followed by a Post-Hoc test was then compiled on the different discriminant factor. The Post-Hoc test allowed classifying the different water body groups in homogenous classes: groups being in the same class did not present significant differences.
3.
Results
3.1.
Water body typology
We calculated (pairwise) correlations between all physicochemical parameters to eliminate redundant information, Significant correlations were found between salinity and conductivity ( p < 0.05, R ¼ 0.99) as well as between dissolved oxygen concentration and dissolved oxygen saturation ( p < 0.05, R ¼ 0.92) for the ‘annual’ database and between pheopigment and chlorophyll concentrations ( p < 0.05, R ¼ 0.99) for the ‘summer’ database. Salinity, dissolved oxygen concentration and pheopigment were thus eliminated from the database because of their close associations with conductivity, dissolved oxygen saturation and chlorophyll concentrations, respectively. Only the member of each correlated pair that would furnish more information about water conditions was retained. The regionalization method was applied to the freshwater database. This analysis identified seven different groups of stations (Fig. 3 A) associated in part with geographical zones: North Aunis marshes for F1 (stations 1-2-7-8), Gironde Estuary marshes for F3 (stations 47-63-64-65), North Aunis marshes for F4 (stations 3-4-5-6) and North Rochefort marshes for F7 (stations 22-23-24-34-37-38-40-41); these four groups are in contrast to three other groups, F2 (stations 9-13-19-20-44-6869), F5 (stations 10-21-45-4-59-70) and F6 (stations 36-48-6658), that contained stations located across the entire study area. Five different groups of stations were discriminated for saltwater marshes (Fig. 3 B). These groups were clearly associated with particular geographical zones: Re´ Island for S1, S4 and S5 (stations 15, 16, 17, 18 and 71); Ole´ron Island for S3 (stations 51-52-53-55); and Ole´ron Island and Seudre Estuary for S2 (stations 60-61-62).
3.2.
4157
Discriminant variables
The method of regionalization classified the 51 stations of the database into 12 different water body groups. FDA identified the parameters responsible for discrimination by region. The first and second axes of the FDA contributed 81% of the discriminant function for the freshwater database. The first axis was significantly explained by E. coli (14% of the contribution), DOC (12.5%), phosphate (10.5%) and conductivity (6%). The groups defined by the regionalization method (part 1) ranged from freshwater (low conductivity) with low E. coli, DOC and PO4 concentrations (groups F3, F6 and F7) to brackish water with high concentrations (group 1) (data not shown). The second axis was significantly explained by DOC (15%), E. coli (13%) and nitrate (11%) concentrations discriminating groups with low concentrations (groups F2 and F4) from groups with high concentrations (group F5). The saltwater body types were not discriminated by the same physicochemical parameters that discriminated the fresh and brackish types. The first two axes contributed to 98% of the discriminant function. The first axis explained 96% of the group discrimination with 35% of this contribution explained by conductivity, 17% by temperature, 15% by SPM, and 7% by dissolved oxygen saturation. The second axis explained only 2% of the group discrimination, with 42% of the contribution explained by BOD and 23% by DOC.
3.3. Water body typology: characterization of each source of explanatory variation A nested ANOVA was conducted to identify the principal source of variation for the significant parameters of the FDA (Fig. 4). According to the nested ANOVA, the group effect was always significant ( p < 0.05) for each parameter (except for Temperature) and explained most of the variation. The seasonal variability (month) was significant for temperature (87% of the total variance), SPM (50%), dissolved oxygen saturation (49%), NO3 (39%) and BOD (35%). Monthly variability was also significant for PO4 and salinity, but this variability contributed only slightly to the total variability for these parameters respectively 6% and 1.3% (Fig. 4). The year-to-year variance was significant for only four parameters: BOD, NO3, dissolved oxygen saturation and temperature, but this variability explained no more than 16% of the total variance (for NO3).
Fig. 3 e Final cluster resulting from the regionalization method applied to the annual database containing: A- fresh and brackish water and B- salt water. 89: AU p-value.
4158
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 5 2 e4 1 6 8
Fig. 4 e Percentage of estimated variance according to the nested ANOVA design (group within month within year) applied on the discriminant parameters. DOC: dissolved organic carbon, Cond: conductivity, PO4: phosphate, E. coli: Escherichia coli concentration, BOD: biological oxygen demand, NO3: nitrate, SPM: suspended matter, O2%: oxygen saturation, T C: temperature. *: significant differences p < 0.05.
3.4.
Spatial variability
Groups were classified according to decreasing summer chl a concentration (F1 to F7 for freshwater groups and S1 to S5 for saltwater groups, Fig. 4A). The gradient of chl a was positively correlated with gradients of DOC and of BOD (Fig. 5 A, B, C). The chl a gradient was also positively correlated to PO4 but only for freshwater groups (F1 to F7, Fig. 5D). The highest values of PO4 were observed for the two Re´-Island groups (S1 and S5). Freshwater marshes (F1 to F7) were effectively discriminated by NO3 concentrations, with low NO3 for F1, F3 and F6, intermediate concentrations for F4 and high concentrations for F2, F5 and F7 (Fig. 5E). Saltwater marshes exhibited very low NO3 concentrations and high intragroup variability. F1 and F3 exhibited higher conductivity than the other freshwater marshes (Fig. 5F). S1 and S2 were less saline than the other saltwater marshes. Intragroup variability was very high for SPM and for oxygen saturation. However, these parameters were only effective discriminant for saltwater marshes (Fig. 5G,H). F1 and F2 have higher concentrations of E. coli than do other freshwater marshes (Fig. 5F).
efficiently isolated by the two first axes of the PCA. Considering the three axes, our three groups are well clustered, but the 2-D representation did not allow seeing it. For the temporal mode, three seasons were separated using a HAC: winter (November to January), spring (March to June) and summer (August to September) (result of the HAC reported in Fig. 6B). An interpolation of the first axis of the PCA in the parameter mode was constructed to determine the seasonal evolution of the different groups of marshes (Fig. 6B). The groups that were most correlated with axis 1 will exert the greatest influence on the formation of axis 1. Higher curvature of the lines thus corresponds to greater differences between periods. The results indicated that the transition between the different periods was gradual and gentle. Variation was higher for the contrast of freshwater marshes than for salt marshes and exhibited a gradual shift from F1 to S5 (Fig. 6B). Groups F6 and F7 seemed to be transitional. Their properties looked like those of the other freshwater marshes during rainy months (November and January) and those of saltwater marshes during dry months (August and September).
3.5.
3.5.2.
3.5.1.
Temporal variability Seasonal variability
Significant seasonal variability was observed for temperature, DOC, NO3 and E.coli (Table 2). Some groups presented a higher amount of seasonal variability than others. In groups F1, F7 and S3 67% of the variables exhibited seasonal variability. 55% of the variables exhibited seasonal variability for S2 and S4. For the three-mode PCA applied to the seasonal series, the temporal mode (99.9%) represented the greatest amount of variability, followed by the group mode (98.5%) and the parameter mode (52%, Fig. 6A). Three clusters were identified by a Hierarchical Ascendency Classification (HAC, Euclidean distance, Ward method). The first of these clusters regrouped freshwater marshes F6 and F7, the second regrouped freshwater marshes F1 to F5 and the last regrouped saltwater marshes S1 to S5 (Fig. 6A). The two latter clusters were not
Year-to-year variability
For the three-mode PCA applied to the year-to-year series, the temporal mode (99.9%) exhibited the greatest variability, followed by the group mode (98.1%) and the parameter mode (60.8%). Three clusters were identified by the HAC. The first cluster regrouped freshwater marshes F6 and F7, the second cluster regrouped freshwater marshes F1 to F5 and the third cluster regrouped saltwater marshes S1 to S5 (Fig. 7A). In an analysis similar to that used for seasonal variability (Fig. 7B), an interpolation of axis 1 of the PCA of the parameter mode was constructed to evaluate year-to-year variability. The results of this analysis were compared with those for the year-to-year evolution of precipitation (Fig. 7B). This comparison revealed a correlation of year-to-year evolution with precipitation. The transitions correspond to alternating dry and rainy periods with gentle, gradual transitions between the different periods. Variation was higher for freshwater
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 5 2 e4 1 6 8
4159
Fig. 5 e Boxplot displaying the mean per group for the following parameters (the temporal variability was erased): A- chlorophyll a (chl a) concentration, B- dissolved organic carbon (DOC) concentration, C- biological oxygen demand (BOD), D- phosphate (PO4) concentration, E- nitrate (NO3) concentration, F- conductivity, G- suspended matter (SPM) concentration, H- dissolved oxygen saturation and I- Escherichia coli concentration. F1 to S5: Groups of marshes found by the regionalization method.
marshes than for saltwater marshes (Fig. 7B). As previously found for seasonal variability, groups F6 and F7 seemed to be transitional between freshwater and saltwater marshes in accordance with the amount of precipitation.
3.6. Relation between typology and environmental factors The groups identified by the regionalization method were used to constrain two discriminant factorial analyses (FDA) that included environmental factors (hydraulic functioning, human activities and pedological substratum): one for salt marshes and one for freshwater marshes. C
For the freshwater marshes, the first two axes of the FDA explained most of the variability (86%) (Fig. 8A). Seven environmental factors significantly discriminated the freshwater groups. The extent of Charente
replenishment, the NO3 concentrations in ground water, the number of WPP and the percentage of silt were strongly correlated with the first axis (49%). The catchment basin index, the degree of replenishment by ground water, the channel density and the percentage of soil culture were correlated with the second axis (37%). Most of these discriminant factors are involved with the hydrographical functioning of the network of marshes. Significant differences between groups (ANOVA) were detected for all the environmental discriminant factors found by the FDA (Table 3) with the exception of the number of WPP and the percentage of silt. However, only two groups (F2 and F4) exhibited WPP in their marshes. The combined replenishment factors (replenishment by the Charente river and replenishment by ground water) seem to imply a water renewal gradient from F1 (no replenishment) to F7 (replenishment by the Charente). Perturbations of this gradient
4160
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 5 2 e4 1 6 8
Table 2 e Results of the two-way ANOVA comparing seasonal variability (‘month’) and intragroup variability (‘station’) and their ‘interactions’, given as p-level (***: p < 0.0001, **: p < 0.001, *: p < 0.01, ns: p > 0.05). T C: temperature, DOC: dissolved organic carbon, NO3: nitrate, E. coli: Escherichia coli, Cond: conductivity, BOD: biological oxygen demand, SPM: suspended matter, O2%: oxygen saturation, PO4: phosphate. Groups
ANOVA results
F1
Factor ’Month’ Factor ’Station’ Interactions Factor ’Month’ Factor ’Station’ Interactions Factor ’Month’ Factor ’Station’ Interactions Factor ’Month’ Factor ’Station’ Interactions Factor ’Month’ Factor ’Station’ Interactions Factor ’Month’ Factor ’Station’ Interactions Factor ’Month’ Factor ’Station’ Interactions Factor ’Month’ Factor ’Station’ Interactions Factor ’Month’ Factor ’Station’ Interactions Factor ’Month’ Factor ’Station’ Interactions
F2
F3
F4
F5
F6
F7
S2
S3
S4
T C
DOC
NO3
E.coli
Cond
BOD
SPM
O2 %
PO4
*** ** ns *** *** ns *** ns * *** ns ns *** ns ns *** ns ns *** ns ns *** ns ns *** ns ns *** ns ns
*** *** *** ns ** * *** ns ns *** *** ns *** *** ns *** *** ns *** *** ns *** *** ns *** *** ns *** *** ns
*** * ns *** ns ** ns ns ns *** *** ns *** *** ns *** *** ns *** *** ns *** *** ns *** *** ns *** *** ns
ns ns ns ns *** ns ns * ns * *** ns * *** ns * *** ns * *** ns * *** ns * *** ns * *** ns
** *** ns ns ns ns *** ** ns ns ns ns ns ns ns ns ns ns ** *** ns ** *** ns ** *** ns ** *** ns
* ** ns * *** ns ns ns ns ns ns ns ns ns ns ns ns ns ns ns ns ns ns ns ** ns ns ** ns ns
ns ** ns ns *** ns ns ns ns ns ns ns ns ns ns ns ns ns *** ns ns ns * ns ns * ns ns * ns
* ns ns *** ns * ns ns ns ns ns ns ns ns ns ns ns ns ns ns ns ns ns ns ns ns ns ns ns ns
ns ns ns ns ns ns ns ns ns ns ns ns ns ns ns ns ns ns ns ns ns ns ns ns ns ns ns ns ns ns
reflect the importance marsh surface area relative to catchment basin surface area (M/CB, the more the indices value is low, the more the catchment basin surface is important; F2, F4, F5 and F7; Table 3). In contrast, groups F1 and F3 were defined by a very small catchment basin (Table 3). The groundwater NO3 concentration was very high for group F4 (nearly 70 mg l1) and very low for groups F1 and F7 (Table 3). Other groups exhibited intermediate values. Channel densities were high for F2, F3, F5 and F6 and low for F1, F4 and F7. Percentages of culture were high for F1, F3, F4 and F7 and low for F2, F5 and F6. Only the group F3 presents a low percentage of substratum silt (Table 3). C
For the saltwater marshes, the first two axes of the FDA captured essentially all of the variability (100%) represented by the discriminating factors (Fig. 8B). Four environmental factors significantly discriminated the saltwater groups. Natural marshes, percentage of wood and shellfish culture were strongly correlated with the first axis whereas salt culture was correlated with the second axis. All the discriminant factors involved anthropogenic activities.
Significant differences between groups were found only for the following factors: number of unexploited ponds, saltern
ponds and the area of the catchment basin (Table 4). Groups S1 and S5 were characterized by a very low percentage of shellfish culture ponds (1.5% and 3.3%, respectively). Most of the surface area of the S1 and S5 marshes was unexploited (75.8% and 66.1%, respectively; Table 4). The percentage of saltern ponds differed between these two groups: high value for S5 had a (11.9%), and low value for S1 (1.5%). The chief activity for the other groups was shellfish culture ponds (Table 4). S3 has a large catchment basin compared to the other groups (Table 4), whereas S4 is characterized by high salt culture activity (Table 4).
4.
Discussion
4.1.
Statistical approach
To understand the functioning of each type of marshes, a statistical approach was developed comprising a characterization step relating water body typology and environmental factors (i.e. anthropogenic activities, hydraulic functioning.) (Fig. 2). Such approach can now be applied to other areas. The first step of this approach was to define this typology using the five-year survey. The first challenge was to find a statistical approach clustering a high number of stations having the
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 5 2 e4 1 6 8
4161
Fig. 6 e A- PCA representing the correlations between groups and the first and second principal components for the analysis of parameters and years. Percentage of variance explained by the components is given in parentheses. Groups were clustered into three zones, B- Variability of the first principal component for the analysis of months and groups derived using Parameters mode PCA. The gray shading indicates the intensity of the first component. The groups determined from the cluster analysis are indicated for month on the ordinate and for groups on the abscissa. (Jan: January, Mar: March, Jun: June, Aug: August, Sep: September, Nov: November).
same patterns in terms of spatio-temporal variability. Souissi et al. (2000) had developed the ‘regionalization method’ in this way. Some amelioration of the method was done, in particular to test the significant of the groups found by the final cluster using the pv-clust method. This method is commonly used in phylogeny to cluster closed genetic species (Shimodaira, 2002). The ‘regionalization method’ is a powerful tool to cluster stations with a long-time series, however, it do not permit to find discriminant parameters. To have an idea of the functioning of the different water body, statistical tests were necessary to both study their spatial and temporal variability. To study the spatial variability, the temporal variability needed to be erased by computing a mean value on the five years for each station and each parameter. A Factorial Discriminant Analysis (FDA) was performed on this database. FDA allowed finding and classifying the parameters having the main effect on the typology. The Nested ANOVA had
demonstrated the importance of seasonal and year-to-year variability. Each temporal variability was then extracted thanks to the seasonal partition CENSUS 1. To have a global vision of this temporal variability, 3-modes PCA were performed on these two new databases. The results of this method had permit to show same evolution across groups. Uses of FDA and 3-modes PCA had allowed describing this variability in a global view. The second challenge was to analyze relationships linking this typology with such environmental factors as hydraulic functioning, human activities and pedological substratum. The majority of the similar studies focused on the effect of one environmental factor on the water quality: i.e. anthropogenic land used (Tong and Chen, 2002; Hussenot, 1998). The statistical method used here was developed in order to identify and classify the main environmental factors implied on the typology. The first step of this work was to realize the more
Fig. 7 e A- Group Mode: PCA representing the correlations between groups and the first and second principal components for the analysis of parameters and years. Percentage of variance explained by the components is given in parentheses. Groups were clustered into three zones, B- Parameters mode: Variability of the first principal component analysis of years and groups derived using Parameters mode PCA. The gray shading indicates the intensity of the first component. The groups determined from the cluster analysis are indicated for years on the ordinate and for groups on the abscissa. (M: March, A: August, N: November). The white curve represents the time course of precipitation between November 2003 and September 2007.
4162
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 5 2 e4 1 6 8
Fig. 8 e Factorial plane of DFA axis 1 and 2 for A- groups of freshwater marshes and B- groups of saltwater marshes. M/CB: ratio between surface marsh and surface catchment basin, GW: ground water, R: replenishement, [NO3]: nitrate concentration, WPP: water purification plant.
exhaustive list of environmental factor having potentially an impact on hydrobiological parameters and identify their availability. The complexity and the multiplicity of the environmental factors having a potential effect on the water quality had needed a global approach. The second step is to digitize these factors on a geographic information system (GIS) in order to define metrics (i.e. number of purification plant per ha, marshes and catchment basin surface area, channel densities inside the marsh.). This step allowed obtaining a database for each march with the different factors. An FDA was then computed in order to find and classify factors explaining the typology. The last challenge was to link these factors and the hydrobiological parameters thanks to scientific knowledge of other ecosystems. These analyses yielded specific proposals regarding the biological functioning of these marshes. The results of our study also suggest that the use of further parameters might lead to an improved scientific understanding of marsh dynamics and function.
4.2. Importance of hydrological functioning and water renewal Twelve groups of marshes were found to differ in both average values and temporal evolution of hydrological parameters (Fig. 2). Seven of these groups consisted of freshwater marshes and the five groups for saltwater marshes. Different discriminant parameters were found for the two types of marshes. Intergroup variability was consistently significant for all the discriminant hydrological parameters. Temporal variability, in particular seasonal fluctuations, was significant for most of the parameters as well (Fig. 3). Our considerations of seasonal variability and of year-to-year variability yielded the same three clusters of groups (Fig. 5A and Fig. 6A). The first cluster contained freshwater marshes (F1 to F5), and the second cluster contained saltwater marshes (S1 to S5). The third cluster contained freshwater marshes (F6 and F7) and was transitional between the two other groups. The functional hydrology of these three groups seems to be linked to
precipitation with different behavior during dry periods and during rainy periods. Behavior of the transitional cluster during rainy periods is similar to that of freshwater marshes and during dry periods to that of saltwater marshes. The apparent difference in hydrological functioning among the three clusters seems to result from water renewal in the marshes. A gradient of water renewal was observed among the three clusters: (i) little or no renewal due to the low replenishment rates for the freshwater cluster (F1 to F5), (ii) high renewal for the transitional cluster (F6 and F7) and (iii) very high for the saltwater cluster due to the effect of the tide (S1 to S5). The amount of water renewal in freshwater marshes depends on the degree of replenishment from fluvial sources (the result of human control) and on the degree of replenishment from groundwater sources during summer (Table 3). Differences between rainy and dry periods were more marked for low-replenishment marshes (Fig. 5B and Fig. 6B). During winter, the gates of the sea locks for freshwater marshes are often opened to flush excess water resulting from precipitation and thereby limit the actual flooding of the marshes. These gates remain closed during summer to prevent desiccation that would otherwise results from low precipitation. Marshes were thus much more stagnant during summer than during winter, with the exception of freshwater marshes characterized by high human control replenishment rates (Group F7). Consequently, hydrological parameters exhibit their greatest temporal fluctuations in low-replenishment marshes (Fig. 5B and Fig. 6B). In this context the transitional position of group F6 and F7 is evidently the result of (i) a water renewal rate similar to that of other freshwater marshes during the rainy period, and (ii) a high water renewal rate during the dry period due to high human-control replenishment, a pattern similar to that found for saltwater groups. Tidal effects caused saltwater groups to exhibit high rates of water renewal in channel areas. In saltwater marshes, the refill channel serves to return water during high tides to ponds located on the marshes. The types of activities (e.g., shellfish culture, saltern ponds, and fish aquaculture) carried out on these ponds seem to exert
4163
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 5 2 e4 1 6 8
Table 3 e Mean (±standard deviation) of the discriminant environmental factors for freshwater marshes. p-values (bold face) indicate significant differences among groups according to ANOVA (a [ 0.05), same letters (A,B,C and D) indicate homogenous classes (Cl) of groups according to LSD POST HOC test (a [ 0.05). M/CB: ratio between surface marsh and surface catchment basin, CD: channel density, GW: ground water, [NO3]: nitrate concentration, %: percentage, Ch: Charente River. Metrics for each factor were explained in Table 1. Groups
CD (m ha1) GW [NO3] (mg.l1)
M/CB (%) Mean
F1 97.1 5.5 F2 23.4 17 F3 66 23.6 F4 46 21 F5 32 10 F6 31.12 16 F7 10.82 Anova <0.0001 p-value
Cuture (%)
Cl
Mean
Cl
Mean
Cl
Mean
A D B C B B D
60 23 114 21 125 40 75 9.5 113 45 110 21 69.67 0.030
A B B A B B A
00 40 0 22.5 20.6 70 0 32.5 15 26 23 0.00 <0.0001
A B C D B B A
68.1 7 35.6 20.6 45 33 68 4.6 22.7 18 33 8 51.81 0.020
marked influence on the discrimination results for the salt marshes (Fig. 7B). Even if one activity dominates others within a salt marsh, the marsh still resembles a patchwork of several different activities (Table 4). However, the water residence time within the ponds varies between a few hours and 15 day depending on the nature of ongoing aquaculture activities (Bel Hassen, 2001) and also on the season of the year. For example, oyster culture ponds are never emptied into the refill channel during winter (L. Anras, pers. comm.). The composition of the water in a given refill channel will therefore vary from day to day depending on the particular ponds that empty into the channel at a given time. Differences in water residence times between ponds can influence water quality significantly. Hussenot (1998) showed that semi-intensive aquaculture (fish aquaculture) produced more pollution than did extensive aquaculture (shellfish culture). Semi-intensive aquaculture requires massive renewals of water. The massive discharges that result then carry untransformed dissolved substances, particularly nitrogen compounds (NO3, NO2, NH3, etc.) and phosphate, introduced by aquaculture activity. In shellfish culture ponds, on the other hand, water residence times are comparatively longer. These ponds exhibit high primary production resulting from the assimilation of nitrogen compounds and phosphates. The physicochemical parameters can also be influenced strongly by the quality of coastal seawater. The S2 group was slightly less saline than the other salt marshes (Fig. 4F), and this decreased salinity is likely due to higher input of terrestrial freshwater in the S2 group
Cl
Silt (%) Mean
A 100 0 B 84.7 26.7 A 56 42 A 98 2.3 B 70 27 B 95 9 A 98.90 0.156
Ch replenishement GW replenishement Cl
Mean
Cl
Mean
Cl
A A B A A A A
00 00 00 00 1.25 0.5 3.33 0.6 10 <0.0001
A A A A A B C
00 6.19 3 6.25 2.8 6.33 0 8.33 1.9 6.4 0.4 0 0.010
A B B B D B A
(Seudre Estuary marshes). Similarly, seawater can produce dilution effects that influence the various parameters. The sampling approach used in this study (samples collected at low tide at the outlet of the main refill channel) did not allow us to portray the full hydrodynamic complexity of the salt marshes, nor did it allow us to define a clear pattern linking the principal activity on the marshes and the physicochemical characteristics of the different saltwater bodies found there. It would be very interesting to monitor water quality simultaneously at the outlets of different ponds that supported different aquaculture activities as well as in the refill channel.
4.3.
Water renewal and eutrophication
The increasing gradient of water renewal in the freshwater marshes from F1 to F7 described before could explain the decreasing gradient observed for hydrological and biological parameters, particularly for the chl a, BOD, DOC and PO4 concentrations (Fig. 4). The link between these four parameters may indicate a gradient of eutrophication from F1 (high eutrophication) to F7 (little or no eutrophication). Eutrophication is the process by which water bodies are made more nutrient-rich as the result of anthropogenic activities. Phytoplankton blooms are the usual consequence of such nutrient enrichment (Smith et al., 1999). In such eutrophic environments, the production of algae exceeds the consumption of grazers. The resulting accumulation of algal detritus
Table 4 e Mean (±standard deviation) of the discriminant environmental factors for salt marshes (selected by the FDA). pvalues in bold indicate significant differences among groups according to ANOVA (a [ 0.05), same letters indicate homogenous classes (Cl) of groups according to LSD post hoc test (a [ 0.05). CB: catchment basin, %: percentage. Groups
S1 S2 S3 S4 S5 ANOVA p-value
Wood (%)
Saltern ponds (%)
Shellfish Culture ponds (%)
No exploited ponds (%)
Surface CB (ha)
Mean
Cl
Mean
Cl
Mean
Cl
Mean
Cl
Mean
Cl
1.9 1 1.7 3.4 2.4 1.4 0.3 1.8 0.421
NS NS NS NS NS
1.5 0.2 0.6 0.03 0.06 13.4 12.7 11.9 0.066
A A A B B
1.5 23.9 30.2 32.2 12.8 22.2 14.8 3.3 0.680
NS NS NS NS NS
75.8 16.6 54.3 11.3 50.2 19.8 66.1 0.9 0.006
A B C C A
826 1268.4 661.6 3801.8 1308.5 775 332.6 1557 0.008
A A B A A
4164
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 5 2 e4 1 6 8
stimulates bacterial activity which then produces oxygen depletion in the water column (Strain and Yeats, 1999). Phosphorus, rather than nitrates, demonstrably limits primary production in most freshwater ecosystems (Schindler, 1977). In stagnant waters, more phosphorus can be released to the water column from the sediment or from particulate organic phosphorus (POP). Sulfate accumulation resulting from agricultural activity and low water oxygen saturation facilitate this process by anion exchange (Correll, 1998; Lucassen et al., 2004). The measures of PO4 concentrations in this survey are not sufficient to explain the PO4 variation in water. It would seem necessary to consider both PO4 and POP concentrations. Nevertheless, the phosphorus gradient resulting from water stagnation can explain the chl a gradient (Fig. 4). Consistent with this supposition, the concentration of NO3 varied substantially along the water renewal gradient (Fig. 4). For example, groups F1 and F2 exhibited the highest values of PO4 and chl a concentration, whereas group F1 had less NO3 than did group F2 (Fig. 4E). Furthermore, a low N:P ratio may induce blooms of potentially-toxic nitrogen-fixing cyanobacteria. Under such environmental conditions, these cyanobacteria will outcompete other phytoplankton groups (Havens et al., 2003). Rooted macrophytes and phytoplankton can also compete for nutrients. Many rooted macrophytes can extract nutrients from water as well as from sediment. Nutrient limitation is therefore less important for macrophytes than for phytoplankton, thus allowing them to outcompete phytoplankton (Sand-Jensen and Borum, 1991). Intense nutrient loading stimulates epiphyte growth on macrophyte leaves and stem and reduces macrophyte growth through shading and nutrient competition effects and favor phytoplankton growth. These processes will lead to eutrophication (Bro¨nmark and Weisner, 1992). Analyses of the macrophyte community and phytoplankton diversity are essential to understanding the eutrophication mechanism in freshwater marshes, and such analyses could lead to better discrimination of the freshwater marshes. Nevertheless, along this gradient, a highly stagnant system such as F1 may be more sensitive to eutrophication than a system like F7. In such stagnant systems, algae lysates and extracellular products from algae can cause DOC to increase. DOC may then accumulate as its release rates exceed the DOC consuming by microorganisms (Cheng and Chi, 2003). The DOC gradient and chlorophyll a concentration gradient may therefore be linked. The simultaneous maxima that we observed for DOC and for chlorophyll a concentrations are consistent with this proposal. The same gradient was observed for biological oxygen demand (BOD). The BOD measures the consumption of oxygen by microorganisms through physiological respiration, particularly in relation to organic matter. The correlation observed between chlorophyll a concentration and BOD may suggest that detrital algal carbon is the primary contributor to BOD, as is the case in other ecosystems (Volkmar and Dahlgren, 2006). The availability of DOC produced by algae may also stimulate the growth of heterotrophic bacteria in eutrophic systems (Pinckney et al., 2001). A possible complication is that heterotrophic bacteria and phytoplankton may or may not compete for NH4, depending on the concentration of dissolved organic nitrogen (DON) (Legendre and Rassoulzadegan, 1995). If the DON pool is low,
competition occur between production by phytoplankton and decomposition by bacteria (Legendre and Rassoulzadegan, 1995). Production will amplify eutrophication, but excessive decomposition will oppose it. When bacteria consume oxygen and DOC they increase the BOD in the water column. Their metabolic activity induces hypoxia, as we observed during summer for groups F1 and F2 (dissolved oxygen saturation reaching value between 6 and 25% during summer period). This depletion of oxygen may in turn inhibit decomposition and remineralization of nutrients locked in biomass, thus inhibiting further phytoplankton production and halting growth. Measurements of DON concentration could therefore be extremely informative concerning the nitrogen cycle.
4.4.
Catchment basins and land (soil) use
Freshwater marshes characterized by high NO3 concentrations exhibited large catchment basin areas relative to their marsh surface (groups F2, F4, F5 and F7; Fig. 4E, Table 3). The types of development (culture) that occupied these catchment basins tended to use nitrogen fertilizer, the main source of nitrates in water (Vitousek et al., 1997). A large catchment basin area relative to the marsh surface will be characterized by increases in NO3 concentration at the marsh sample point that are the consequences of leaching precipitation farmed soils (Wiesler and Horst, 1993). However, this pattern was not observed for groups F6 (marshes with large catchment basins but with low NO3 concentrations) and F7 (small catchment basins but high NO3 concentrations) (Fig. 4E, Table 3). The catchment basins of group F6 were occupied chiefly by meadow or were replenished by groundwater that contained essentially no nitrates. Group F7 is a cluster of stations associated with the Charente replenishment channel. The nitrate concentrations observed for F7 can be explained by the fact that the channel’s nutrient concentration is close to nutrient concentration of the Charente River which drains a large catchment. Atmospheric precipitation following the use of nitrogen fertilizers may be a second source of nitrate. The amount of this input will depend on the surface area of the catchment basin (Morales-Baquero et al., 1999). NO3 concentrations found for salt marshes were very low compared to those found for freshwater marshes (Fig. 4E). These concentrations reflect phytoplankton nutrient limitation. Such limitation results in most cases from limited nitrate availability in marine ecosystems (Arrigo, 2005). Nevertheless, S2 marshes exhibited the greatest NO3 concentrations among the salt marshes. The physical locations of these marshes seem to explain this result. Some of the S2 marshes are located along the Seudre Estuary, which may supply NO3 to these marshes. Other S2 marshes are located above Ole´ron Island. Their characteristics may therefore reflect strong influence by the input that feeds the catchment basin. This effect would not apply to the other salt marshes because they are located at the outlet of the main channel. Nitrates showed strong seasonality for all fresh- and saltwater marshes except for those of S1. The seasonal nitrate maximum occurred during winter (maximum value 71.76 mg l1 for fresh marshes and 9.03 mg l1 for salt marshes). The lack of NO3 seasonal variation for S1 may be the result of regular nitrate input from a fish farm located near the
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 5 2 e4 1 6 8
sample point (Tovar et al., 2000). The seasonality of the other groups appears to be the result of high precipitation during winter. This precipitation leaches soil and brings NO3 from the catchment basin or from the farmlands on the marsh. During spring and to an even greater extent during summer, NO3 concentrations reached their lowest values (>0.05 mg l1). These values were comparable to those observed in lakes (Andersen, 1982) or in marine ecosystems (Del Amo et al., 1997). This depletion of nitrates may result from phytoplankton assimilation during spring or summer blooms (Andersen, 1982), by denitrification (Seitzinger, 1988) or, in freshwater marshes, from uptake by riparian vegetation and by hydrophytes (Verhoeven et al., 2006). Indeed, the lack of plant growth during the winter due to limited sunlight may lead to the accumulation of nitrate during this season. In salt marshes, uptake of nitrate by phytoplankton and microphytobenthos may result in NO3 depletion. At the autotrophic base of the food web, we may find competition between phytoplankton, microphytobenthos and hydrophytes for light, nutrient and inorganic carbon (Sand-Jensen and Borum, 1991). Microphytobenthos can actually develop on their banks and resuspension can occur due to tidal activity and specific meteorological events which caused a benthicpelagic coupling (Guarini et al., 2008). Further study of this ecosystem component may lead to improved scientific understanding of NO3 depletion mechanisms. In addition, Verhoeven et al. (2006) argued that wetlands contribute significantly to NO3 removal if they remove at least 30% of the NO3 load. For each group, we observed levels of NO3 removal that always exceeded 30%. We found >80% for F1, F4, F6, S1 and S2, between 70% and 80% for F2, F7 and S4 and between 60% and 70% for F3, F5 and S3 (ratio between the highest value of NO3 in winter and the lowest in summer). This conclusion may imply that even though the Charente-Maritime marshes are strongly impacted by humans, they still retain the ability to remove nitrate, an important characteristic of wetlands. This suggestion needs to be viewed with some caution because eutrophication may occur in groups F1 and F2 and because of the bias inherent to the calculation. The calculation method (ratio between winter and summer value) imply winter water stagnation but like it is described before, during high precipitation event, water marshes are evacuate to the sea. Eutrophication also removes NO3, but the consequences of such eutrophication are harmful for the ecosystem and may include shifts in phytoplankton species composition to taxa that may be toxic or inedible (e.g., bloom-forming cyanobacteria), and dissolved oxygen depletion in the water column that induces shifts in fish species composition towards less desirable species and, in the worst case, increased fish kills (Smith et al., 1999). Bloom-forming cyanobacteria were recently observed in several CharenteMaritime marshes in late summer during eutrophication phenomena (no published data).
4.5. Which additional parameters could improve water quality monitoring? Twelve different water body types were defined and classified. This typology made sense in terms of anthropogenic hydraulic controls and in terms of soil use. The resulting interpretations
4165
were complex. The chief source of this complexity was that overall ecological understanding was restricted by the parameters available for analysis. The parameters that were available to us from stakeholder surveys were those used by the WFD to evaluate water quality in rivers. Reliance on the WFD parameter set necessarily placed limits on the kinds of interpretations that could be made and on the study’s potential contribution to the basic science of marshes. Rivers and marshes differ in significant ways. For example, different hydrodynamic regimes could occur because water is more stagnant in marshes than in rivers. The particular parameters used for rivers by WQES might well be incomplete for describing and for understanding marsh function. The need remains for appropriate scientific studies that will facilitate interpretation of measured WFD parameters and lead to better scientific understanding of the ecological patterns and functional mechanisms underlying our marsh typology. Available measurements of the WFD parameters are generally expressed as concentrations. However, to better understand the complexity of marsh ecosystems, we need to go beyond isolated point measurements and to consider external flows (energy or matter exchanges between ecosystems) and internal flows (energy or matter exchanges within the ecosystem). In fact, the Charente-Maritime salt and freshwater marshes are connected hydraulically to adjacent ecosystems. The connections of the catchment basin or of the river/groundwater will influence water quality and quantity in the marshes. Marsh water quality and quantity will influence coastal seawater when sea lock gates are opened during periods of high precipitation and water drains from the marshes to the sea. The connection between saltwater marshes and the coastal zone is a two-way street, an important link in a complex, dynamic and still poorly understood network that includes ecological causation, interactions and relationships ecomplexity mainly due to the difference of water residence time within the ponds according to the activity. Interactions between differently-functioning ecosystems produce a ‘meta-ecosystem’: an ecosystem connected by spatio-temporal flows of energy, materials and organisms across ecosystem boundaries (Loreau et al., 2003). Improved understanding of the origin of the organic matter involved in all these flows (allochthonous/autochthonous, phytoplankton/ microphytobenthos/bacteria and marine/terrestrial) should allow better definition of such exchanges and the interactions they produce. The origin of the organic matter could be identified by using different methods as isotopic ratio of the POM (Galois et al., 2000) and/or 3-D fluorescence of the DOM. Furthermore, the results of this study highlight the overall importance of the biota and the role of the planktonic community in eutrophication and in NO3 removal. Of equal overall importance is recognition that benthic-pelagic coupling in salt marshes can result from resuspension of microphytobenthos. To achieve better scientific understanding of the functioning of the different water bodies, it seems crucial to study the linkages and the temporal evolution of the different planktonic compartments (from bacteria to mesozooplankton). Such studies necessarily require detailed examination of the planktonic food web (i.e. grazing activity, bacterial and phytoplanktonic production) and of the patterns that characterize its changing dynamics in space and
4166
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 5 2 e4 1 6 8
time. Appropriate consideration of the unique functional properties of the planktonic food web, as defined by Legendre and Rassoulzadegan (1995) and Sintes et al. (2004), could shed new light on the following functional characteristics of marshes: ‘purification’ by the recycling of material components, exportation of matter to adjacent ecosystems or ecotrophic efficiency.
5.
Conclusions
A statistical method was developed to define and characterize a marsh water body typology. For the CharenteMaritime area, twelve different water bodies were defined: 7 in freshwater and 5 in saltwater marshes). This statistical method should be applied in other marshes area to improve the sensibility of this typology. This typology made sense in terms of natural and anthropogenic hydraulic controls and in terms of soil use. However, for saltwater marshes, the sampling approach used in this study did not allow us to define a clear pattern linking the principal activity and the physicochemical characteristics of the different saltwater bodies found there because of the complexity of the hydrodynamic features in those marshes. The size of the catchment basin and particularly the soil use (culture) influence strongly nitrates concentration inside the freshwater marshes. Even though the CharenteMaritime marshes are strongly impacted by humans, they may still retain the ability to remove nitrate, an important characteristic of wetlands. The increasing gradient of water renewal in the freshwater marshes from F1 to F7 explained the decreasing gradient of eutrophication. Freshwater marshes are at the interface between the continent and the sea: catchment basin water quality will influence marshes and water quality marshes will influence water coastal area. A better management of the hydrodynamic of the marshes mainly human control can avoid eutrophication risk on the coastal sea area which sustains a great oyster culture production. The European Water Framework Directive (WFD) furnished a common frame of reference that supported the analysis of marshes and is likely suitable for comparisons of other complex aquatic ecosystems. However, we feel that improved scientific understanding of the resulting typologies can also stem from a broader ecological approach. In particular, ecologically-based insights regarding both external flows (links between ecosystems, meta-ecosystem theory) and internal flows (structure of the planktonic food web) seem an essential prerequisite for further advances in the study of marsh ecosystems.
Acknowledgments This study was supported by the « Ministe`re de l’Enseignement Supe´rieur et de la Recherche », the water agencies Loire-
Bretagne and Adour-Garonne, conseil ge´ne´ral de CharenteMaitime, European Union. Thanks are extended to Christian Pointillard and Serena Como for their helpful suggestions.
references
Ahn, C., Mitsch, W.J., 2002. Scaling considerations of mesocosm wetlands in simulating large created freshwater marshes. Ecological Engineering 18 (3), 327e342. Aminot, A., Ke´rouel, R., 2004. In: Hydrologie des e´cosyste`mes marins: parame`tres et analyses. IFREMER Publishing, Plouzane´, France. Arrigo, K.R., 2005. Marine micoorganisms and global nutrient cycles. Nature 437 (7057), 349e355. Andersen, J.M., 1982. Effect of nitrate concentration in lake water on phosphate release from the sediment. Water Research 16 (7), 1119e1126. Beaugrand, G., Ibanez, F., Reid, P.C., 2000. Spatial, seasonal and long-term fluctuations of plankton in relation to hydroclimatic features in the English Channel, Celtic Sea and Bay of Biscay. Marine Ecology Progress Series 200, 93e102. Bel Hassen, M., 2001. Spatial and temporal variability in nutrients and suspended material processing in the Fier d’Ars Bay (France). Estuarine. Coastal and Shelf Science 52 (4), 457e469. Benoit, L., Askins, R., 1999. Impact Impact of the spread of Phragmites on the distribution of birds in Connecticut tidal marshes. Wetlands 19 (1), 194e208. Billaud, J.-P., 1984. Marais Poitevin: rencontres de la terre et de l’eau. In: Mendras, H., Jollivet, M. (Eds.), Ecrits et travaux du groupe de sociologie rurale du CNRS. L’Harmattan Publishing, Paris, France, p. 265. Bro¨nmark, C., Weisner, S.E.B., 1992. Indirect effects of fish community structure on submerged vegetation in shallow, eutrophic lakes: an alternative mechanism. Hydrobiologia 243e244 (1), 293e301. Cheng, W.P., Chi, F.-H., 2003. Influence of eutrophication on the coagulation efficiency in reservoir water. Chemosphere 53 (7), 773e778. Cominelli, E., Galbiati, M., Tonelli, C., Bowler, C., 2009. Water: the invisible problem. EMBO Reports 10 (7), 671e676. Correll, D.L., 1998. The role of phosphorus in the eutrophication of receiving waters: a review. Journal of Environmental Quality 27 (2), 261e266. Costanza, R., d’Arge, R., de Groot, R., Farber, S., Grasso, M., Hannon, B., Limburg, K., Naeem, S., O’Neill, R.V., Paruelo, J., Raskin, R.G., Sutton, P., van den Belt, M., 1997. The value of the world’s ecosystem services and natural capital. Nature 387 (6630), 253e260. Dagne´lie, P. (Ed.), 1975. The´orie et me´thodes statistiques: Applications Agronomiques. Les presses agronomique de Gembloux Publishing, Gembloux, Belgica, p. 464. David, V., Sautour, B., Chardy, P., Leconte, M., 2005. Long-term changes of the zooplankton variability in a turbid environment: the Gironde estuary (France). Estuarine, Coastal and Shelf Science 64 (2e3), 171e184. Del Amo, Y., Le Pape, O., Treguer, P., Queguiner, B., Menesguen, A., Aminot, A., 1997. Impacts of high-nitrate freshwater inputs on macrotidal ecosystems. I. Seasonal evolution of nutrient limitation for the diatom-dominated phytoplankton of the Bay of Brest (France). Marine Ecology-Progress Series 161, 213e224. Gagnaire, B., Thomas-Guyon, H., Burgeot, T., Renault, T., 2006. Pollutant effects on Pacific oyster, Crassostrea gigas (Thunberg), hemocytes: screening of 23 molecules using flow cytometry. Cell Biology and Toxicology 22 (1), 1e14. Gagnaire, B., Gay, M., Huvet, A., Daniel, J.-Y., Saulnier, D., Renault, T., 2007. Combination of a pesticide exposure and
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 5 2 e4 1 6 8
a bacterial challenge: in vivo effects on immune response of Pacific oyster, Crassostrea gigas (Thunberg). Aquatic Toxicology 84 (1), 92e102. Galois, R., Blanchard, G., Seguignes, M., Huet, V., Joassard, L., 2000. Spatial distribution of sediment particulate organic matter on two estuarine intertidal mudflats: a comparison between Marennes-Ole´ron Bay (France) and the Humber Estuary (UK). Continental Shelf Research 20 (10e11), 1199e1217. Goberville, E., Beaugrand, G., Sautour, B., Treguer, P., Team, S., 2010. Climate-driven changes in coastal marine systems of western Europe. Marine Ecology-Progress Series 408, 129eU159. Goulletquer, P., Heral, M., 1997. The history, present condition and future of the molluscan fisheries of North and Central America and Europe. In: MackenzieJr., C.L., Jr.Burrell., V.G. (Eds.), NOAA Technical Report. NMFS, Europe, pp. 137e164. Guarini, J.-M., Sari, N., Moritz, C., 2008. Modelling the dynamics of the microalgal biomass in semi-enclosed shallow-water ecosystems. Ecological Modelling 211 (3e4), 267e278. Havens, K.E., James, R.T., East, T.L., Smith, V.H., 2003. N: P ratios, light limitation, and cyanobacterial dominance in a subtropical lake impacted by non-point source nutrient pollution. Environmental Pollution 122 (3), 379e390. Hussenot, J., 1998. In: Hussenot, J., Buchet, V. (Eds.), Marais Maritimes et aquaculture. Activite´ durable pour la pre´servation et l’exploitation des zones humides littorales. IFREMER, Rochefort, p. 278. Legendre, L., Rassoulzadegan, F., 1995. Plankton and nutrient dynamics in marine waters. Ophelia 41, 153e172. Lemonnier, P., 1980. Les Salines de l’Ouest. Logique technique, logique sociale. E´ditions de la Maison des Sciences de l’Homme/PUL, Paris, France, pp 222. Loreau, M., Mouquet, N., Holt, R.D., 2003. Meta-ecosystems: a theoretical framework for a spatial ecosystem ecology. Ecology Letters 6 (8), 673e679. Lorenzen, C.J., 1967. Determination of chlorophyll and phaeopigments: spectrophotometric equations. Limnology and Oceanography 12, 343e346. Lotze, H.K., Lenihan, H.S., Bourque, B.J., Bradbury, R.H., Cooke, R.G., Kay, M.C., Kidwell, S.M., Kirby, M.X., Peterson, C.H., Jackson, J.B. C., 2006. Depletion, degradation, and Recovery potential of estuaries and coastal Seas. Science 312 (5781), 1806e1809. Lougheed, V.L., Crosbie, B., Chow-Fraser, P., 2001. Primary determinants of macrophyte community structure in 62 marshes across the Great Lakes basin: latitude, land use, and water quality effects. Canadian Journal of Fisheries Aquatic Sciences 58 (8), 1603e1612. Lubchenco, J., Olson, A.M., Brubaker, L.B., Carpenter, S.R., Holland, M.M., Hubbell, S.P., Simon, A.L., MacMahon, J.A., Matson, P.A., Melillo, J.M., Mooney, H.A., Peterson, C.H., Pulliam, H.R., Real, L.A., Regal, P.J., Risser, P.G., 1991. The Sustainable Biosphere Initiative: an ecological research Agenda: a Report from the ecological Society of America. Ecology 72 (2), 371e412. Lucassen, E., Smolders, A.J.P., van de Crommenacker, J., Roelofs, J. G.M., 2004. Effects of stagnating sulphate-rich groundwater on the mobility of phosphate in freshwater wetlands: a field experiment. Archiv Fur Hydrobiologie 160 (1), 117e131. Meybeck, M., Helmer, R., 1996. In: Chapman, D. (Ed.), Water quality assessments - A guide to use of biota, sediments and water in environmental monitpring. E&FN Spon, London, p. 651. Millennium Ecosystem Assessment, 2005. Ecosystems and Human Well-Being: Wetlands and Water Synthesis. World Resources Institute, Washington, DC, p. 68. Mitsch, W.J., Cronk, J.K., Wu, X., Nairn, R.W., Hey, D.L., 1995. Phosphorus Retention in constructed freshwater riparian marshes. Ecological Applications 5 (3), 830e845.
4167
Morales-Baquero, R., Carrillo, R., Reche, I., Sanchez-Castillo, P., 1999. Nitrogen-phosphorus relationship in high mountain lakes: effects of the size of catchment basins. Canadian Journal of Fisheries and Aquatic Sciences 56 (10), 1809e1817. No˜ges, P., van de Bund, W., Cardoso, A., Solimini, A., Heiskanen, A.-S., 2009. Assessment of the ecological status of European surface waters: a work in progress. Hydrobiologia 633 (1), 197e211. Pinckney, J.L., Paerl, H.W., Tester, P., Richardson, T.L., 2001. The role of nutrient loading and eutrophication in estuarine ecology. Environmental Health Perspectives 109, 699e706. Rojo, C., Baron-Rodriguez, M.M., Alvarez-Cobelas, M., Rodrigo, M. A., 2010. Sustained primary production with changing phytoplankton assemblages in a semiarid wetland. Hydrobiologia 639, 55e62. Rozas, L.P., Odum, W.E., 1988. The roˆle of submerged aquatic vegetation in influencing the abundance of nekton on contiguous tidal fresh-water marshes. Journal of Experimental Marine Biology and Ecology 114 (2e3), 289e300. Sand-Jensen, K., Borum, J., 1991. Interactions among phytoplankton, periphyton, and macrophytes in temperate freshwaters and estuaries. Aquatic Botany 41 (1e3), 137e175. Schindler, D.W., 1977. Evolution of phosphorous limitaion in lakes. Science 195 (4275), 260e262. Seitzinger, S.P., 1988. Denitrification in freshwater and coastal marine ecosystems: ecological and Geochemical significance. Part 2: comparative ecology of freshwater and marine ecosystems. Limnology and Oceanography 33 (4), 702e724. Shimodaira, H., 2002. An Approximately Unbiased test of Phylogenetic Tree Selection. Systematic Biology 51 (3), 492e508. Sintes, E., Martinez-Taberner, A., Moya, G., Ramon, G., 2004. Dissecting the microbial food web: structure and function in the absence of autotrophs. Aquatic Microbial Ecology 37 (3), 283e293. Smith, V.H., Tilman, G.D., Nekola, J.C., 1999. Eutrophication: impacts of excess nutrient inputs on freshwater, marine, and terrestrial ecosystems. Environmental Pollution 100 (1e3), 179e196. Sokal, R., Rohlf, F., 1995. Biometry San Fransisco. Souissi, S., Yahia-Kefi, O.D., Yahia, M.N.D., 2000. Spatial characterization of nutrient dynamics in the Bay of Tunis (south-western Mediterranean) using multivariate analyses: consequences for phyto- and zooplankton distribution. Journal of Plankton Research 22 (11), 2039e2059. Strain, P.M., Yeats, P.A., 1999. The relationships between chemical measures and potential Predictors of the eutrophication status of Inlets. Marine Pollution Bulletin 38 (12), 1163e1170. Suzuki, R., Shimodaira, H., 2006. Pvclust: an R package for assessing the uncertainty in hierarchical clustering. Bioinformatics 22 (12), 1540e1542. Talureau, R., 1965. L’ame´nagement des marais et l’avenir de l’Ouest Atlantique, Paris. Tong, S.T.Y., Chen, W., 2002. Modeling the relationship between land use and surface water quality. Journal of Environmental Management 66 (4), 377e393. Tovar, A., Moreno, C., Ma´nuel-Vez, M.P., Garcı´a-Vargas, M., 2000. Environmental Implications of intensive marine aquaculture in Earthen ponds. Marine Pollution Bulletin 40 (11), 981e988. Verhoeven, J.T.A., Arheimer, B., Yin, C., Hefting, M.M., 2006. Regional and global concerns over wetlands and water quality. Trends in Ecology & Evolution 21 (2), 96e103. Vitousek, P.M., Aber, J.D., Howarth, R.W., Likens, G.E., Matson, P.A., Schindler, D.W., Schlesinger, W.H., Tilman, G.D., 1997. Human alteration of the global nitrogen cycle: sources and consequences. Ecological Applications 7 (3), 737e750.
4168
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 5 2 e4 1 6 8
Volkmar, E.C., Dahlgren, R.A., 2006. Biological oxygen demand dynamics in the Lower San Joaquin River, California. Environmental Science & Technology 40 (18), 5653e5660. Weller, M., 1978. In: Good, R.E., Whigham, D.G., Simpson, R.L. (Eds.), Freshwater Wetlands: Ecological Processes
and Management Potential. Academic Press, New York, pp. 267e284. Wiesler, F., Horst, W.J., 1993. Differences among maize cultivars in the utilization of soil nitrate and the related losses of nitrate through leaching. Plant and Soil 151 (2), 193e203.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 6 9 e4 1 8 2
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Phytoplankton community succession shaping bacterioplankton community composition in Lake Taihu, China Yuan Niu a, Hong Shen a, Jun Chen a, Ping Xie a,*, Xi Yang a, Min Tao a, Zhimei Ma a, Min Qi b a
Donghu Experimental Station of Lake Ecosystems, State Key Laboratory of Freshwater Ecology and Biotechnology of China, Institute of Hydrobiology, Chinese Academy of Sciences, Wuhan 430072, PR China b College of Fishery, Huazhong Agricultural University, Wuhan, Hubei, 430070, PR China
article info
abstract
Article history:
PCR-denaturing gradient gel electrophoresis (DGGE) and canonical correspondence anal-
Received 6 December 2010
ysis (CCA) were used to explore the relationship between succession of phytoplankton
Received in revised form
community and temporal variation of bacterioplankton community composition (BCC) in
15 March 2011
the eutrophic Lake Taihu. Serious Microcystis bloom was observed in JulyeDecember 2008
Accepted 22 May 2011
and Bacillariophyta and Cryptophyta dominated in JanuaryeJune 2009. BCC was charac-
Available online 31 May 2011
terized by DGGE of 16S rRNA gene with subsequent sequencing. The DGGE banding patterns revealed a remarkable seasonality which was closely related to phytoplankton
Keywords:
community succession. Variation trend of ShannoneWiener diversity index in bacter-
Bacterioplankton community
ioplankton community was similar to that of phytoplankton community. CCA revealed
composition (BCC)
that temperature and phytoplankton played key roles in structuring BCC. Sequencing of
Phytoplankton community
DGGE bands suggested that the majority of the sequences were affiliated with common
succession
phylogenetic groups in freshwater: Alphaproteobacteria, Betaproteobacteria, Bacteroidetes and
Denaturing gradient gel
Actinobacteria. The cluster STA2-30 (affiliated with Actinobacteria) was found almost across
electrophoresis (DGGE)
the sampling time at the two study sites. We observed that the family Flavobacteriaceae
Canonical correspondence
(affiliated with Bacteroidetes) tightly coupled to diatom bloom and the cluster ML-5-51.2
analysis (CCA)
(affiliated with Actinobacteria) dominated the bacterioplankton communities during Micro-
Flavobacteriaceae
cystis bloom. These results were quite similar at the two sampling sites, indicating that BCC
ML-5-51.2
changes were not random but with fixed pattern. Our study showed insights into rela-
Lake Taihu
tionships between phytoplankton and bacterioplankton communities at species level, facilitating a better understanding of microbial loop and ecosystem functioning in the lake. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
In natural freshwater systems, bacterioplankton plays a key role in the breakdown of organic matter and nutrients cycling (Cole et al., 1988; Caron, 1994). Despite recent advances in the
characterization of freshwater bacterioplankton diversity, our knowledge about the factors regulating the bacterioplankton community composition (BCC) is far from holistic understanding (Lindstro¨m et al., 2005; Hahn, 2006). Several environmental factors have been suggested to be related to the
* Corresponding author. Tel./fax: þ86 27 68780622. E-mail address:
[email protected] (P. Xie). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.05.022
4170
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 6 9 e4 1 8 2
temporal variation in BCC in experimental and natural systems, which include water temperature (Crump and Hobbie, 2005; Kan et al., 2007; Shade et al., 2007), phytoplankton succession (Ho¨fle et al., 1999; Pinhassi et al., 2004; Rooney-Varga et al., 2005) and predation (van Hannen et al., 1999b; Langenheder and Ju¨rgens, 2001; Muylaert et al., 2002). A previous study has shown that in eutrophic shallow lakes, seasonality of bacterial community structure is dependent on the dominant substrate source as well as on the food web structure (Muylaert et al., 2002). Rooney-Varga et al. (2005) also suggest that changes in phytoplankton community composition may influence the composition of bacterial communities that function as part of the microbial loop. Differences in the quality of organic matter produced by different phytoplankton communities probably result in changes in the composition of bacterioplankton community utilizing this organic matter (van Hannen et al., 1999a). Despite the fact that bacterioplankton and phytoplankton are thought to be closely linked in aquatic ecosystem, our knowledge about how these communities interact with each other at the species composition level is still limited especially in shallow eutrophic lakes. Lake Taihu, the third largest freshwater lake in China, is a typical shallow eutrophic subtropical lake located in east China (surface area: 2338 km2, mean depth: 1.9 m). In general, like most eutrophic lakes (Kalff and Knoechel, 1978), diatoms dominate in spring phytoplankton, while heavy cyanobacterial blooms are characteristic for late-summer on its northern part (including Zhushan Bay, Meiliang Bay and Gonghu Bay) each year. Microbial communities are key players in ecological processes related to water quality, so the detailed knowledge of their diversity and function in the system is an essential prerequisite for the sustainable management of freshwater resources (Hahn, 2006). Although a few studies have focused on bacterioplankton community diversity and its regulating factors including submersed macrophytes, phytoplankton community, nutrients, etc. in Lake Taihu (Wu et al., 2007a, 2007b; Xing and Kong, 2007; Tian et al., 2009), our understanding of the relationships between phytoplankton and specific bacterioplankton taxa, and how bacterioplankton are affected by environmental change are far from complete. We hypothesized that phytoplankton community succession might be crucial for bacterioplankton community composition and the specific bacterioplankton taxa could change in accordance with the succession of predominant phytoplankton genus. To this end, links between phytoplankton community succession and BCC variation at the species composition level are badly needed. In addition, this was the first time to explore the BCC in Gonghu Bay, which provides the bulk of drinking water of Wuxi city for millions of residents. Since the first publication by Muyzer et al. (1993) the PCR-denaturing gradient gel electrophoresis (DGGE) fingerprinting has been widely used in environmental microbiology, and has been recognized to give an acceptable view on differences and similarities of the dominant populations of microbial communities (Muyzer, 1999; Muyzer and Smalla, 1998; Lindstro¨m, 2000, 2001; Dorigo et al., 2005). In the present study, the DGGE and canonical correspondence analysis (CCA) methods were used to explore whether there was a close relationship between phytoplankton community succession
and temporal variation in BCC in Lake Taihu. Our results clearly demonstrated the above hypothesis, namely, we observed that the family Flavobacteriaceae (affiliated with Bacteroidetes) might be in close relation to diatom bloom, while the cluster ML-5-51.2 (affiliated with Actinobacteria, Wu et al., 2007b) was in association with Microcystis blooms. These may facilitate to understand the relationships between phytoplankton and bacterioplankton communities in eutrophic lake.
2.
Materials and methods
2.1.
Site and sampling
GonghuBay, where a terrible drinking water malodor incident occurred in the early summer of 2007, is located in the northeastern part of Lake Taihu and has a surface area of approximately 120 km2. The two sampling sites (Site A 31 250 11.51"N, 120 150 35.04"E; Site B 31 26’49.16"N, 120 220 11.10"E) are located near the Gonghu Waterworks and Xidong Waterworks, respectively (Fig. 1). Lake Taihu has become increasingly eutrophic over the past three decades (Qin et al., 2007) and the ratio of area in hypertrophic state had risen to 77% (Wu et al., 2007b). According to the results of our field survey at 30 sites distributed throughout the lake from 2008e2009 (unpublished data), the succession of predominant phytoplankton and zooplankton communities presented similar patterns at most sites. So the two sampling sites could represent the majority of cases in the entire lake. Water samples were collected monthly from July 2008 to June 2009, from surface waters (0e0.5 m) with a 5 l Schindler sampler. Water temperature, pH and dissolved oxygen (DO) were measured on location by an YSI 6600 Multi-Parameter Water Quality Sonde. Bacterioplankton samples (200e300 ml of water) for BCC analysis were collected on 0.2 mm-pore-size filters after pre-filtration through 5.0 mm-pore-size filters (diameter 45 mm; Whatmann, UK). Filters were stored at 80 C until analysis. Water samples (50 ml) were fixed with glutaraldehyde to a final concentration of 2% for determination of total bacterial abundance. 1 l water samples for identification and counts of phytoplankton were preserved with 1% Lugol’s iodine solution. Zooplankton were collected by straining 10 l water samples through a 64 mm plankton net and fixed with formalin.
2.2.
Chemical analysis
Total nitrogen (TN), ammonium (NH4-N), nitrate (NO3-N), total phosphorus (TP), ortho-phosphorus (PO4-P) and chlorophyll a (chl a) were analyzed according to standard methods (Jin and Tu, 1990).
2.3. Identification and counting of phytoplankton and zooplankton Phytoplankton samples were concentrated to 50 ml after sedimentation for 48 h. Then 0.1 ml concentrated samples were counted on a microscope under 400 magnification after mixing. Colonial Microcystis spp. cells were separated using an ultrasonic device and then counted. Phytoplankton species
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 6 9 e4 1 8 2
4171
Fig. 1 e Locations of the two sampling sites (A and B) in Gonghu Bay of Lake Taihu.
were identified according to Hu and Wei (2006). Biomass was estimated from cell numbers and cell size measurements, assuming that 1 mm3 of algal volume equals 1 mg of fresh weight biomass. Zooplankton were identified according to Chiang and Du (1979) and Sheng (1979) and counted using a microscope at a magnification of 100. To estimate changes in phytoplankton community, we used the ShannoneWiener Index (H0p , bits) H0p ¼
s X ni i¼1
n
ln
ni n
where ni is biomass of the ith genus, n is the total biomass of all genera, and s is the total number of genera.
2.4.
Total bacterial abundance
2 ml pre-fixed water samples were stained with 40 ,60 diamidino-2-phenolindole (DAPI, Sigma, final concentration 1 mg ml1) for 15 min (Poter and Feig, 1980) and filtered onto 0.2 mm pore size black polycarbonate filters (Whatman). Total bacterial cells were enumerated under an epifluorescence microscope (Zeiss Axioshop 20). Atleast 1000 cells and a minimum 10 fields of view were counted for each sample.
2.5.
DNA extraction and PCR amplification
The filters containing microbes were cut into small pieces with a sterile scalpel and then bacterial genomic DNA was extracted using a Bacterial DNA kit (Omega) following the protocol, followed by purification on QIAamp DNA Kit (Qiagen,Valencia, CA, USA). The purified DNA was used as template to amplify the 16S rRNA gene fragment with the Bacteria-specific primers 357F (50 -CCTACGGGAGGCAGCAG-30 )
with a 40-bp GC clamp attached to its 50 end and the universal primer 518R (50 -ATTACCGCGGCTGCTGG-30 ) (Muyzer et al., 1993). PCR mixtures of 50 ml contained 1 PCR buffer, 1.5 mM MgCl2, 200 mM of each dNTP, 0.2 mM of each primer, 2.5 U of Taq DNA polymerase (Takara) and 50 ng of template DNA. 5 min initial denaturation at 94 C was followed by a thermal cycling program as follows: 1 min denaturation at 94 C; 1 min annealing at an initial 65 C, decreasing 1 C every cycle to a final of 55 C; 1 min extension at 72 C. 30 cycles were run followed by a final 10 min extension at 72 C. A negative control, in which the template was replaced by an equivalent volume of sterile deionized water, was included. PCR products were confirmed by 1.5% agarose gel electrophoresis.
2.6.
Denaturating gradient gel electrophoresis (DGGE)
A total of 800 ng of PCR product for each sample was loaded on a 8% (w/v) polyacrylamide gel (37.5:1 acrylamide: bisacrylamide) with a denaturing gradient that ranged from 40 to 60%, where 100% denaturant is defined as 7M urea and 40% deionized formamide. DGGE was performed with a Dcode system (Bio-Rad Laboratories, USA) using 1 TAE running buffer (20 mM Tris, 10 mM acetic acid, 0.5 mM EDTA, pH 8.0) at 60 C for 7 h at 150V. The gel was stained in 1:10000 diluted GelRed (Biotium, USA) nucleic acid staining solution for 30 min and photographed using a Bio Image System (Gene Com.) under UV light. DGGE were performed separately for Sites A and B.
2.7.
Sequencing of DGGE bands
All visible DGGE bands were excised with a sterile razor blade and eluted overnight at 4 C in 40 ml MilliQ water. 2 ml of the supernatant was used as template for reamplification with the
4172
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 6 9 e4 1 8 2
same primer set as described above (without a GC clamp). The amplicons were electrophoresed again on a DGGE gel to check the position of the original band and then purified with Gel Recovery Purification Kit (AxyPrep) and ligated into pMD18-T plasmid vector (Takara, Japan) following the manufacturer’s instruction. The ligated DNA was transformed into Escherichia coli DH5a-competent cells. The recombinant clones were selected and then submitted for sequencing using an M13 primers and an automated ABI sequencer at the Genomics Company (Wuhan, China).
2.8.
n X
Pi ln Pi
i¼1
where Pi is the relative intensity of each band and n is the total number of bands in each lane. To reveal relationships between BCC and environmental variables (including the physical and chemical parameters, abundances of crustacean zooplankton and biomasses of different taxonomic groups of phytoplankton), canonical correspondence analysis (CCA) was carried out with the software CANOCO (version 4.5, Microcomputer Power, Ithaca, New York, USA) as the length of the first DCA axis run on species data was >4 (ter Braak, 1987; ter Braak and Verdonschot, 1995). By scoring presence (1) and absence (0) of particular DGGE bands, the DGGE data were used to compile a binary matrix by scoring presence (1) and absence (0) of particular DGGE bands. The environmental variables significantly related to BCC were tested by forward selection and Monte Carlo permutation test. CCA was run separately for data sets of Sites A and B. Statistical analysis of the physicalechemical parameters and abundances of crustacean zooplankton and total bacterial was conducted by SPSS software, version 13.0 for Windows (Chicago, USA). Significance was determined at an alpha level of 0.05 ( p < 0.05).
2.9.
2.10.
Nucleotide sequences accession numbers
The 16S rRNA gene sequences have been deposited to GenBank under the Accession No. HQ259136 to HQ259213.
Cluster analysis of DGGE profiles and Statistics
Cluster analyses of DGGE profiles were performed with the NTSYS program version 2.10e (Exeter software, Setauket, NY, USA). A binary matrix was constructed by scoring presence or absence of DGGE bands. Pairwise similarities between gel banding patterns were quantified using the Dice coefficient as: SD ¼ (2NAB)/(NA þNB), where NAB is the number of bands common to the samples A and B, and NA and NB are the number of bands in samples A and B, respectively. The similarity coefficients were then used to construct a dendrogram using unweighted pair group method with arithmetic average (UPGMA) through the sequential, hierarchical, agglomerative, and nested clustering (SHAN) routine of the NTSYS program. Gel images were analyzed using Gel-Pro Analyzer (version 4.5). A densitometric curve was calculated for each lane and the relative intensities of all bands were obtained. The ShannoneWiener Index (H0b , bits) was calculated to estimate changes in bacterioplankton community composition. H0b ¼
CLUSTAL W program (Thompson et al., 1994) and the phylogenetic analyses were performed with the MEGA4.0 (Tamura et al., 2007) software package using neighbor-joining methods (Saitou and Nei, 1987). The sequences with similarities greater than 97% were grouped in one operational taxonomic unit (OTU).
Phylogenetic analysis
The obtained 16S rRNA gene sequences were aligned to known sequences using BLAST. Sequences were aligned with the
3.
Results
3.1.
Environmental parameters
The nutrient concentrations (TN, NH4-N, NO3-N, TP, TN:TP mass ratios and PO4eP), physical parameters (water temperature, pH, dissolved oxygen, water depth and secchi depth), chlorophyll a (chl a) and abundances of crustacean zooplankton of each site are shown in Table 1. The highest concentrations of TN and TP at Site A both occurred in October, whereas the maximum TN and TP at Site B were detected in May and September, respectively (Fig. 2b and c). There were no significant differences ( p > 0.05) in nutrient concentrations (TN, NH4-N, NO3-N, TP, TN:TP mass ratio and PO4eP) and pH between the two sites. The level of chlorophyll
Table 1 e Environmental parameters (mean and range) of the two sampling sites located in Gonghu Bay of Lake Taihu from July 2008 to June 2009. Variable
Site A Mean
Total nitrogen (TN, mg L1) Total phosphorus (TP, mg L1) TN:TP NH4eN (mg L1) NO3eN (mg L1) PO4eP (mg L1) Dissolved oxygen (mg L1) Water temperature ( C) Water depth (m) Secchi depth (cm) chlorophyll a (mg L1) pH Cladoceran abundance (ind. L1) Copepods abundance (ind. L1) Total bacterial abundance (106 cells ml1)
Range
Site B Mean
Range
2.12
0.21e5.48
1.50
0.36e2.72
0.138
0.033e0.620
0.088
0.039e0.156
21 0.23 0.51 0.008 10.2
6e52 0.12e0.58 0.04e1.34 0.001e0.018 6.4e19.0
18 0.31 0.50 0.010 9.8
7e37 0.07e0.95 0.03e1.92 0.004e0.022 7.2e14.0
18.7
3.1e31.7
18.1
2.5e30.3
1.6 36 168.8
1.3e1.8 9e70 12.8e940.0
2.0 54 17.4
1.9e2.4 25e103 2.3e46.2
8.4 241
7.8e9.6 0.1e1054
8.5 67
7.9e8.8 1e313
44
0.7e339
37
0.2e180
5.93
3.10e12.00
5.08
2.10e7.80
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 6 9 e4 1 8 2
4173
Fig. 2 e Seasonal variations in (a) water temperature, (b, c, d) total nitrogen (TN), total phosphorus (TP) and TN:TP ratio, (e) chlorophyll a (chl a), (f) abundance of crustacean zooplankton, (g) total bacterial abundance and (h) Shannone-Wiener diversity index of phytoplankton and bacterioplankton at the two sampling sites from July 2008 to June 2009.
4174
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 6 9 e4 1 8 2
a during the cyanobacterial bloom was far in excess of the yearly average and the highest concentrations occurred in October at both sites. The chlorophyll a concentrations were significantly higher ( p < 0.05) at Site A (with more serious cyanobacterial bloom) than at Site B. Additionally, Site B had higher water depth and secchi depth than Site A ( p < 0.05). The seasonal variations in abundance of crustacean zooplankton are shown in Fig. 2f. There were no significant differences ( p > 0.05) between the two sites in abundance of both cladoceran and copepod. The peaks of cladocerans at Sites A and B appeared in August and September, respectively. Ceriodaphnia cornuta (only in summer and autumn) and Bosmina coregoni were the dominant species at both sites. The peaks of copepod appeared in July and Limnoithona sinensis as well as Sinocalanus dorrii were the dominant species at both sites.
3.2.
Total bacterial abundance
The seasonal variations in total bacterial abundance are shown in Fig. 2g. There was no significant difference ( p > 0.05)
between the two sites. The peaks of total bacterial abundance at Sites A and B appeared in October and August, respectively. There was significant correlation between the total bacterial abundance and biomass of Cyanophyta (r ¼ 0.757, p ¼ 0.000), indicating that high total bacterial abundance was always companied with the occurrence of cyanobacterial bloom.
3.3.
Succession of phytoplankton community
Variations in composition of phytoplankton community during the sampling period (from July 2008 to June 2009) are shown in Fig. 3. We can clearly see that Cyanophyta and Bacillariophyta were the two major dominant phyla at both sites in Gonghu Bay (Fig. 3a). Generally, Cyanophyta was the absolutely dominant phylum in JulyeDecember 2008 (cyanobacterial-bloom-period), whereas Bacillariophyta dominated in JanuaryeJune 2009 (diatom-dominated period). In addition, Cryptophyta dominated in March and April at Site B. Microcystis was the dominant genus in the phylum of Cyanophyta (>99% at both sites) and in the phylum of Bacillariophyta
Fig. 3 e Variations of biomass in phytoplankton community composition at the two sampling sites: (a) Relative ratio of different phyla; (b) Cyanophyta and Bacillariophyta from July 2008 to June 2009.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 6 9 e4 1 8 2
Cyclotella was the main genus (>40% and >80% at Sites A and B, respectively). Fig. 3b displays the variations in biomass of Cyanophyta and Bacillariophyta at both sites during the sampling period. Biomass of Cyanophyta maintained at a high level (>2 mg L1) from July 2008 to January 2009 at both Sites A and B, and the peak values occurred in October with 129.7 mg L1 and November with 17.1 mg L1, respectively. From February 2009 biomass of Cyanophyta dropped quickly and remained at a low level (<1 mg L1) until June. In general, more serious cyanobacterial bloom occurred at Site A than at Site B. At Site A, biomass of Bacillariophyta ranged from 0 to 1.6 mg L1 and reached a peak in early spring. While at Site B, biomass of Bacillariophyta remained at a low level from July to November, followed by a sharp increase in December and gradually decreased until April. Finally, the highest values occurred in May with 6.5 mg L1 and dropped to 2.5 mg L1 in June.
3.4. Seasonal variations in diversity of phytoplankton and bacterioplankton communities Seasonal variations in the ShannoneWiener diversity index of phytoplankton and bacterioplankton community are shown
4175
in Fig. 2h. Variations in the ShannoneWiener divers ity index of phytoplankton community revealed a similar pattern at both sites. The index remained at a low level during most time of the cyanobacterial blooms and remarkably increased from December. It reached a peak in February and gradually decreased afterwards. The variation trend in ShannoneWiener diversity index in bacterioplankton community was somewhat similar to that of phytoplankton community. At Site A, the value gradually increased from August, reached its peak in December, and then continuously declined until June. The value was relatively stable and high at Site B ( p < 0.05).
3.5.
Composition of bacterioplankton community
DGGE profiles of 16S rRNA gene fragments from the two sites (24 samples in total) show visible changes in relative brightness and position of DGGE banding pattern in various seasons (Fig. 4a). The cluster analysis (UPGMA) dendrograms of BCC also revealed remarkable seasonality (Fig. 4b). The bacterioplankton communities in various months from Site A (Fig. 4b) were grouped into 5 defined clusters. The October sample (Microcystis biomass reached a peak with 128.7 mg L1)
Fig. 4 e (a) DGGE profiles of 16S rRNA gene fragments from the two sampling sites (b) Cluster analysis of BCC based on DGGE profiles of the two sampling sites from July 2008 to June 2009. The first band obtained from July at Site A in Gonghu Bay was indicated by GH-A-JUL1.The other bands were named in the same manner.
4176
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 6 9 e4 1 8 2
at Site A separately formed cluster 5 (Fig. 4b) since it was very dissimilar to the other samples. Samples from Site B distinctly formed two separate clusters (Fig. 4b). Cluster 1 included the samples taken in 2008 and Cluster 2 was composed of samples collected from 2009. Combined with Fig. 3, we found that BCC was in close association with succession of phytoplankton community. Bands from different months at the same vertical position in a gel were assumed to have identical sequences (Riemann et al., 1999). Thirty-four and forty-four bands from different vertical positions in the DGGE profiles from Sites A and B, respectively, were excised, reamplified, purified and sequenced, of which 4 sequences belonging to Cyanobacteria (bands GH-A-JUL5, GH-B-JUL4, GH-B-FEB3 and GH-B-JUN1) were rejected from further analysis since we focused on heterotrophic bacteria. In the DGGE profiles, the first band obtained from July at Site A was indicated by GH-A-JUL1.The other bands were named in the same manner. Phylogenetic affiliation of clones obtained from the DGGE profiles at Sites A and B is shown in Table 2. The majority of the sequences were affiliated with phylogenetic groups commonly found in freshwater: Alphaproteobacteria, Betaproteobacteria, Bacteroidetes and Actinobacteria. A total of 78 clones were classified into 50 OTUs，where 12 OTUs were belonging to Bacteroidetes, followed by Actinobacteria (8 OTUs), Alphaproteobacteria (7 OTUs) and Betaproteobacteria (6 OTUs). Taxonomic descriptions of the 78 bands obtained from the DGGE profiles of the two sampling sites are shown in Table S1. It had been suggested that band intensity was related to relative abundance of the corresponding phylotype in the template mixture (Murray et al., 1996; Riemann et al., 1999; Fromin et al., 2002), thus bands with relatively high intensities in a lane were assumed to be dominated. During the Micro cystis bloom phase (from July to December), Actinoba cteria (bands GH-A-JUL2 and GH-B-SEP1), Alphaproteobacteria (band GH-A-AUG1), Betaproteobacteria (bands GH-B-JUL1 and GH-B-JUL8) and Bacteroidetes (band GH-B-JUL2) were the most prominent. October sample from Site A was quite diffe rent with dominance of Firmicutes (bands GH-A-OCT2 and GH-A-OCT3), Bacteroidetes (band GH-A-OCT1) and
Table 2 e Phylogenetic affiliation of clones obtained from the DGGE profiles at the two sampling sites. Taxon
Alphaproteobacteria Betaproteobacteria Gammaproteobacteria Deltaproteobacteria Bacteroidetes Actinobacteria Firmicutes Thermomicrobia Fibrobacteres Verrucomicrobia Cyanobacteria Total
No. of clones Site A
Site B
6 4 1 1 9 4 6 0 0 2 1 34
8 6 2 1 9 10 1 1 1 2 3 44
Alphaproteobacteria (band GH-A-OCT4). During most of the diatom-dominated period, samples were more diverse and there was no obviously predominant band. Remarkably, a considerable number of bands affiliated with the family Flavobacteriaceae, Bacteroidetes (bands GH-A-DEC5, GH-A-FEB1, GH-B-JAN1, GH-B-FEB1, GH-B-FEB4, GH-B-APR1, GH-B-APR2 and GH-B-APR3, see Fig. S1d) appeared in this period. May and June samples were dominated by Alphaproteobacteria (bands GH-A-APR1, GH-A-APR2, GH-A-APR4 and GH-B-MAY2). In addition, cluster STA2-30, Actinobacteria (bands GH-A-JAN3 and GH-B-JUL6) was seen almost throughout the whole sampling stage at both sites.
3.6.
CCA
Correspondence canonical analysis (CCA) based on DGGE data and environmental variables were carried out separately for the two sites. According to the results of CCA (Fig. 5a), the differences in BCC are related to the two most important environmental variables (biomass of Cyanophyta and water temperature, p < 0.05) at Site A. The two axes explained 33% of the observed variation in BCC. The first axis was positively related with the biomass of Cyanophyta (r ¼ 0.98) and the second axis was positively related with water temperature (r ¼ 0.89). Similarly, results of CCA (Fig. 5b) illustrated that the differences in BCC were related to the three most important environmental variables (biomass of Cyanophyta and Bacillariophyta, and water temperature, p < 0.05) at Site B. The three variables and the two axes explained 53% and 42% of the observed variation in BCC, respectively. The first axis was positively related with the biomass of Cyanophyta and water temperature (r ¼ 0.92 and 0.47 respectively), and the second axis was positively related with the biomass of Bacillariophyta (r ¼ 0.93). CCA biplot revealed that samples from Sites A and B formed three and four clusters respectively (Fig. 5). Samples collected at Site A from July to September 2008 and from April to June 2009 formed one cluster, while the remaining samples (except October) formed the second cluster. October sample was quite different from the others due to its high biomass of Cyanophyta. Samples collected at Site B were grouped into 4 clusters. Cluster 1 consisted of samples from January to April 2009. Cluster 2 contained samples from July to November 2008. Cluster 3 was composed of samples from May and June 2009 whereas December sample of 2008 formed a separate cluster. These results are similar to cluster analysis of BCC based on DGGE profiles (Fig. 4b).
4.
Discussion
In recent years, the Microcystis blooms have expanded from the northern bays to the center and the northern as well as western regions of the Lake Taihu are now regularly covered by thick blooms from late spring into autumn (Xu et al., 2010), whereas East Bay is covered by a submersed macrophyte community composed mainly of Potamogeton spp. (Wu et al., 2007a). Previous studies indicated that nutrients (Xing and Kong, 2007), submersed macrophytes (Wu et al., 2007a) and phytoplankton community (Tian et al., 2009) have been
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 6 9 e4 1 8 2
Fig. 5 e Correspondence canonical analysis (CCA) biplot based on DGGE data and environmental variables of Sites A (a) and B (b) from July 2008 to June 2009.
suggested to play important roles in shaping BCC in Lake Taihu. However, our results further explained the relationship between predominant phytoplankton groups and specific bacterioplankton taxa. DGGE banding patterns and subsequent cluster analysis intuitively revealed that BCC had a remarkable seasonality which was closely related to phytoplankton community succession. CCA analyses confirmed the key role of phytoplankton community in shaping BCC. By sequencing all visible DGGE bands with phylogenetic analysis, we also found specific bacterioplankton taxa (cluster ML-5-51.2 and family Flavobacteriaceae) that coupled to predominant phytoplankton taxonomic groups (Microcystis spp. and diatom, respectively) at different stages.
4.1. Environmental (abiotic and biotic) factors regulating BCC The CCA results of the present study revealed that biomass of Cyanophyta and Bacillariophyta, as well as water temperat
4177
ure, were the three most influential factors on BCC, responsible for a major part of the observed variation in BCC. Our results agree with the observation of several studies that temperature can significantly influence the seasonal variation in BCC (Crump and Hobbie, 2005; Kan et al., 2007; Shade et al., 2007). As one of the main seasonal factors, water temperature may exert direct and indirect influences on abundance and composition of bacterioplankton community. First, it is always a potentially limiting factor since every bacterial phylotype has minimal, maximal and optimal temperature characteristics (Pomeroy and Wiebe, 2001). Consequently, abundance of each bacterial phylotype probably changes as the water temperature fluctuates, which results in variation of BCC. Second, in particular, temperature should be viewed as an interactive factor because it affects all chemical and biochemical processes (Pomeroy and Wiebe, 2001). It is well known that water temperature has a pronounced impact on phytoplankton and zooplankton compositions, and several studies also indicated that phytoplankton and zooplankton strongly influence bacterioplankton diversity via bottom-up and top-down regulations (Langenheder and Ju¨rgens, 2001; Muylaert et al., 2002; Pinhassi et al., 2004; Rooney-Varga et al., 2005). As a consequence, temperature-mediated succession of plankton community likely contributed to the variations in BCC. Nutrient concentration and composition may directly influence BCC through affecting on growth of bacteria (Pinhassi and Hagstrom, 2000; Haukka et al., 2006). However, none of them was significantly related to BCC ( p > 0.05) in our study. TN:TP ratio is very important in the development of phytoplankton populations and cyanobacteria may dominate in the phytoplankton under low TN:TP ratio conditions (Smith, 1983; Sekar et al., 2002). It is possible that they may have an indirect impact on BCC via influencing the phytoplankton populations. Our study indicated that succession of phytoplankton community was the most important influential factor on BCC. Specially, the biomass of Cyanophyta explained over 20% of the observed variation in BCC independently at both sampling sites, indicating that cyanobacterial blooms may be the major shaping force in Gonghu Bay, Lake Taihu. In addition, the biomass of Bacillariophyta at Site B was significantly related to BCC ( p < 0.05) and explained 17% of the observed variation in BCC independently. It has been reported that phytoplankton plays a key role in regulating BCC in the mesocosms and natural systems (Ho¨fle et al., 1999; Pinhassi et al., 2004; Rooney-Varga et al., 2005; Tian et al., 2009). In some mesocosms experiments, specific bacterioplankton taxa that were in association with different phytoplankton taxonomic groups were also found (Pinhassi et al., 2004; Li et al., 2011). The release of dissolved organic matter (DOM) by phytoplankton has long been recognized as an important source of high quality carbon to bacterial (Cole et al., 1982) and it has been demonstrated that changes in dissolved organic carbon (DOC) can drive changes in BCC (Eiler et al., 2003; Jones et al., 2009). The DOM is rapidly consumed and remineralized by the bacterial community (Jensen, 1983; Obernosterer and Herndle, 1995). Thus, differences in the quality of organic matter produced by different dominant phytoplankton communities probably result in changes in the composition of bacterioplankton community utilizing this organic matter (van
4178
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 6 9 e4 1 8 2
Table 3 e Distribution of three specific taxons (clusters STA2-30, ML-5-51.2 and family Flavobacteriaceae) in the present study. Taxon
STA2-30 ML-5-51.2 Flavobacteriaceae
Site
A B A B A B
Sampling Month Jul
Aug
Sep
Oct
Nov
Dec
Jan
Feb
Mar
Apr
May
Jun
þ þ þ þ e e
þ þ þ þ e e
þ þ þ þ e e
e þ e þ e e
þ þ þ þ e e
e þ e þ þ e
þ þ e e þ þ
þ þ e e þ þ
þ þ e e þ þ
e þ e e e þ
þ þ e e e þ
þ þ e e e þ
Hannen et al., 1999a). van Hannen et al. (1999a) showed that the green algal detritus and cyanobacterial detritus can affect the structure of the microbial community. Combining these results, we can infer that the diverse DOM originated from cyanobacterial detritus, diatomic detritus and Cryptophyta detritus (three dominant phyla in our study, see Fig. 3) might have markedly influenced the composition of bacterioplankton community. High total bacterial abundance was
always companied with the occurrence of cyanobacterial bloom, indicating that since the strengthened photosynthesis of cyanobacteria can increase the amount of organic substrates for bacteria, excessive primary production accelerates the growth of heterotrophic bacteria (Wu et al., 2007b). To this end, different phytoplankton regimens might support development of specific bacterial groups or species in Lake Taihu.
Table 4 e Information of the cyanobacterial bloom-related sequences that were 100 identical to bands GH-A-JUL2 or GH-BSEP1 (both affiliated with cluster ML-5-51.2). Clone Uncultured bacterium clone xyhfb1-22a Uncultured bacterium clone xyhfb1-19a Uncultured bacterium clone dcfb4-60a Uncultured bacterium clone dcfb4-48a Uncultured bacterium clone chfb1-50a Uncultured bacterium clone chfb1-35a Uncultured bacterium clone dcfb4-80b Uncultured actinobacterium clone WA0.2-0d-46a Uncultured actinobacterium clone WA0.2-0d-80b Uncultured actinobacterium clone WA0.2-0d-75b Uncultured actinobacterium clone WA0.2-0d-33b Uncultured actinobacterium clone WA0.2-0d-31b Uncultured actinobacterium clone WA0.2-0d-26b Uncultured actinobacterium clone WA0.2-0d-20b Uncultured actinobacterium clone TH1-97a Uncultured bacterium clone ML-9-97.2a Uncultured bacterium clone ML-5-51.2b Uncultured actinobacterium clone CYN-1-50a
GenBank accession no.
Location
Reference
HM050912
Lake Xingyunhu, China
Bacterial community composition in four eutrophic shallow lakes (unpublished data)
HM050909 HM050642
Lake Dianchi, China
HM050630 HM050479
Lake Chaohu, China
HM050467 HM050661
Lake Dianchi, China
HM153628
Lake Taihu, China
Li et al., 2011
HM153640 HM153638 HM153619 HM153618 HM153615 HM153611 AM690889
Wu et al., 2007a
DQ520197
Wu et al., 2007b
DQ520165 EF158354
a sequences that were 100 identical to band GH-A-JUL2. b sequences that were 100 identical to band GH-B-SEP1.
Lake Samsonvale, Australia
Pope and Patel, 2008
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 6 9 e4 1 8 2
Muylaert et al. (2002) showed that composition of bacterial community was related to phytoplankton biomass, and found no evidence for top-down regulation of bacterial community composition in the turbid lakes characterized by the occurrence of phytoplankton blooms and high-nutrient. Similarly, in our study, abundance of crustacean zooplankton was not significantly related to BCC ( p > 0.05). It is possible that bottomup (resources) regulation could be more important compared to top-down (predation) regulation in the eutrophic lake.
4.2. Taxonomic groups and their distribution associated with phytoplankton community In the present study, the majority of the non-cyanobacterial sequences obtained from the two sampling sites were affiliated with the divisions Bacteroidetes, Proteobacteria, Actinobacteria, and Firmicutes (Table 2). Clones affliated with Thermomicrobia, Fibrobacteres, and Verrucomicrobia were also found in low numbers (<5 sequences per division). Actinobacteria were regarded as one of the most significant divisions in this study. In various freshwater habitats Actinobacteria comprise large fractions of the total bacterioplankton (Glo¨ckner et al., 2000; Warnecke et al., 2004; Allgaier and Grossart, 2006). A few studies have suggested that Actinobacteria are resistant to flagellate predation (Pernthaler et al., 2001; Hahn et al., 2003), which could be one possible explanation for their wide distribution and high abundances in freshwater systems. Clusters ACK-M1 and STA2-30, which are detected in almost every freshwater habitat (Zwart et al., 2003; Eiler and Bertilsson, 2004), were also identified in the present study. Interestingly, cluster STA2-30 (bands GH-A-JAN3 and GH-B-JUL6 at Sites A and B, respectively) was found as a dominant phylotype almost across the sampling time at the two study sites (Table 3). In accordance with previous studies in Lake Taihu, cluster STA2-30 was detected across all five open water areas (Wu et al., 2007a) and it was the only one of the two groups containing OTUs from all the four libraries (March, May, July, and September) obtained from Meiliang Bay (Wu et al., 2007b). Consequently, cluster STA2-30 may be one of the most important dominant phylotypes in the shallow eutrophic Lake Taihu. In the present study, we found that the cluster ML-5-51.2 (includes bands GH-A-JUL2, GH-A-JUL3, GHB-SEP1 and GH-B-OCT1) appeared as the most prominent phylotype only during the Microcystis bloom period (Table 3; Fig. 4a). According to the BLAST analyses of bands GH-A-JUL2 and GH-B-SEP1 (a single-base difference between them), considerable part of the sequences that were 100 identical to bands GH-A-JUL2 or GH-B-SEP1 were closely related to the cyanobacterial bloom (Table 4). Most clones in Table 4 were originated from the four eutrophic shallow lakes in China (Lake Xingyunhu, Dianchi, Chaohu and Taihu), whereas the uncultured actinobacterium clone CYN-1-50 was detected in Lake Samsonvale, Australia, where the toxin-producing cyanobacteria dominate the planktonic blooms (Pope and Patel, 2008). Specially, of the eighteen clones listed in Table 4, three were found in the September libraries (Clone TH1-97, Wu et al., 2007a, ; Clone ML-9-97.2, Wu et al., 2007b) or May library (Clone ML-5-51.2, Wu et al., 2007b) in Meiliang Bay, Lake Taihu during the cyanobacterial bloom period and seven were detected when Li et al. (2011) studied the short-term dynamics of
4179
bacterial community composition in response to Microcystis bloom in Lake Taihu. As a consequence, it is conceivable that cluster ML-5-51.2 is a typical phylotype associated with the occurrence of Microcystis bloom and may regard as an indicator in eutrophic shallow lakes at least in Lake Taihu. It was reported that Bacteriodetes was the most representative division in the clone libraries of four Swedish lakes associated with cyanobacterial blooms (Eiler and Bertilsson, 2004). This division was also present with high frequency in this study (Fig. S1d). Particularly, the family Flavobacteriaceae appeared with a higher proportion in the Bacteroidetes phylum (9 of 18) in the upper region of the DGGE profiles. Pinhassi et al. (2004) found that Flavobacteriaceae phylotypes were characteristic of the diatom-dominated community in their seawater microcosm experiments and the consistent response to diatom bloom conditions of a few Flavobacteriaceae phylotypes could indicate that some bacteria in this family harbor physiological or ecological traits that make them tightly coupled to diatom species. Subsequent phylogenetic analyses of published Bacteroidetes 16S rRNA gene sequences confirmed that members of the Flavobacteriaceae are remarkably responsive to algal blooms in seawater ecosystems (Pinhassi et al., 2004). According to a pervious study in Meiliang Bay, Lake Taihu (Wu et al., 2007b), as much as 70% of the Flavobacteriaceae phylotypes in the four clone libraries (March, May, July, and September) in Meiliang Bay (among a total of 10 sequences) were detected in March library (diatom-dominated period) whereas no clone was found in July and September when the Microcystis bloom happened. In accordance with the abovementioned studies, in our field study the family Flavobacteriaceae was only detected from December 2008 to July 2009, when diatom began to flourish and dominated phytoplankton communities (Table 3). Our results indicated that some Flavobacteriaceae phylotypes may be closely linked to the diatom bloom and play particularly important roles in the processing of organic matter during the diatom blooms in Lake Taihu. Therefore, we suggest that there is coupling between members of the Flavobacteriaceae phylotypes and diatom blooms both in seawater and freshwater ecosystems. In the present study, a total of 29 clones were affliated with the Proteobacteria, and most of these clones were from the subdivisions a-Proteobacteria and b-Proteobacteria (Fig. S1a and b). The majority of the a-Proteobacteria sequences fell into two distinct families, Sphingomonadaceae and Rhodobacteraceae. In this study, 4 typical freshwater clusters of a-Proteobacteria defined by Zwart et al. (2002) were detected, GOBB3-C201, Brevundimonas intermedia, CR-FL11, and LD12. In addition, a-Proteobacteria (bands GH-A-MAY2, GH-A-JUN1 and GH-AJUN2) seemed to be predominant in early summer at Site A. Clones affiliated with the genus Rhodoferax of the b-Proteobacteria (BAL47 freshwater cluster; Zwart et al., 2003; Fig. 6b), which were abundant in all four cyanobacterial bloom libraries in Sweden (Eiler and Bertilsson, 2004), were also detected in our study. A mere five clones grouped in theg- Proteobacteria anddProteobacteria subdivisions (Fig. S1e). Firmicutes are seldom discussed in previous studies of microbial ecology due to their low frequencies in the libraries (Eiler and Bertilsson, 2004; Wu et al., 2007b). Nevertheless, 6 and 1 clones were detected in our study at Sites A and B, respectively, and they all fell into the class Bacilli (Fig. S1f).
4180
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 6 9 e4 1 8 2
Three clones (bands GH-A-APR1, GH-A-APR2 and GH-A-APR4) were detected in April at Site A and they seemed to dominate the bacterioplankton community reflected by the DGGE profile (Fig. 4a). Bands GH-A-OCT2 and GH-A-OCT3 appeared in October, when the Microcystis bloom was very severe (Microcystis biomass reached a peak with 128.7 mg L1). The left 6 of the 74 non-cyanobacterial clones were not affliated with the relatively abundant divisions discussed above. These clones were grouped into Thermomicrobia, Fibrobacteres and Verrucomicrobia (Fig. S1g).
4.3. Diversity of phytoplankton and bacterioplankton community In the present study, variation trend of the ShannoneWiener diversity index in bacterioplankton community was similar to that of phytoplankton community at both sites. The values were relatively low in cyanobacterial bloom period and high in winter or spring (Fig. 2h). Ho¨fle et al. (1999) speculated that the diverse spring phytoplankton bloom releases more different substrates that provide increased niches for the more diverse bacterioplankton observed than does the phytoplankton bloom of a single alga in late-summer. We can conclude that the heavy cyanobacterial bloom consisted of Microcystis spp. resulted in low bacterial diversity while the diverse spring phytoplankton bloom (mainly consisted of Bacillariophyta and Cryptophyta) led to more diverse bacterioplankton community.
5.
Conclusion
A detailed field survey on relationships between specific bacterioplankton taxa and phytoplankton communities shows that biomass of Cyanophyta and Bacillariophyta, as well as water temperature, were the three most influential factors on variation of BCC in Lake Taihu of China. Microcystis bloom resulted in low bacterial diversity while the diverse spring Bacillariophyta and Cryptophyta blooms led to more diverse bacterioplankton community. The family Flavobacteriaceae tightly coupled to diatom bloom while the cluster ML-5-51.2 dominated the bacterioplankton communities during Microcystis bloom. Conclusively, succession of phytoplankton community played key roles in shaping BCC. These may facilitate to understand the microbial loop and ecosystem functioning in eutrophic lake, although the mechanisms on interaction between phytoplankton and bacterioplankton communities need further investigations.
Acknowledgments We thank the editor and two anonymous reviewers for their constructive suggestions and professional editing. This study was jointly supported by National Basic Research Program of China (2008CB418101), Lake water quality, water quantity and biological resources investigation in China (2006FY110600) and State key laboratory of freshwater ecology and biotechnology (2008FBZ01). We are grateful to Dr Xiaoxue Sun, Jing Zhang for zooplankton identification, Meng Zhang for the help on sample collection.
Appendix. Supplementary material Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.watres.2011.05.022.
references
Allgaier, M., Grossart, H.P., 2006. Diversity and seasonal dynamics of actinobacteria populations in four lakes in northeastern Germany. Applied and Environmental Microbiology 72 (5), 3489e3497. Chiang, S.C., Du, N.S., 1979. Fauan Sinica, Crustacea, Freshwater Cladocera. Science Press, Academia Sinica, Beijing (In Chinese). Caron, D.A., 1994. Inorganic nutrients, bacteria, and the microbial loop. Microbial Ecology 28 (2), 295e298. Cole, J.J., Likens, G.E., Strayer, D.L., 1982. Photosynthetically produced dissolved organic carbon: an important carbon source for planktonic bacteria. Limnology and Oceanography 27 (6), 1080e1090. Cole, J.J., Findlay, S., Pace, M.L., 1988. Bacterial production in fresh and saltwater: a cross-system overview. Marine EcologyProgress Series 43 (1e2), 1e10. Crump, B.C., Hobbie, J.E., 2005. Synchrony and seasonality in bacterioplankton communities of two temperate rivers. Limnology and Oceanography 50 (6), 1718e1729. Dorigo, U., Volatier, L., Humbert, J.F., 2005. Molecular approaches to the assessment of biodiversity in aquatic microbial communities. Water Research 39 (11), 2207e2218. Eiler, A., Langenheder, S., Bertilsson, S., Tranvik, L.J., 2003. Heterotrophic bacterial growth efficiency and community structure at different natural organic carbon concentrations. Applied and Environmental Microbiology 69 (7), 3701e3709. Eiler, A., Bertilsson, S., 2004. Composition of freshwater bacterial communities associated with cyanobacterial blooms in four Swedish lakes. Environmental Microbiology 6 (12), 1228e1243. Fromin, N., Hamelin, J., Tarnawski, S., Roesti, D., JourdainMiserez, K., Forestier, N., Teyssier-Cuvelle, S., Gillet, F., Aragno, M., Rossi, P., 2002. Statistical analysis of denaturing gel electrophoresis (DGE) fingerprinting patterns. Environmental Microbiology 4 (11), 634e643. Glo¨ckner, F.O., Zaichikov, E., Belkova, N., Denissova, L., Pernthaler, J., Pernthaler, A., Amann, R., 2000. Comparative 16S rRNA analysis of lake bacterioplankton reveals globally distributed phylogenetic clusters including an abundant group of actinobacteria. Applied and Environmental Microbiology 66 (11), 5053e5065. Hahn, M.W., 2006. The microbial diversity of inland waters. Current Opinion in Biotechnology 17 (3), 256e261. Hahn, M.W., Lu¨nsdorf, H., Wu, Q.L., Schauer, M., Ho¨fle, M.G., Boenigk, J., Stadler, P., 2003. Isolation of novel ultramicrobacteria classified as Actinobacteria from five freshwater habitats in Europe and Asia. Applied and Environmental Microbiology 69 (3), 1442e1451. Haukka, K., Kolmonen, E., Hyder, R., Hietala, J., Vakkiainen, K., Kairesalo, T., Haario, H., Sivonen, K., 2006. Effect of nutrient loading on bacterioplankton community composition in lake mesocosms. Microbial Ecology 51 (2), 137e146. Ho¨fle, M.G., Haas, H., Dominik, K., 1999. Seasonal dynamics of bacterioplankton community structure in a eutrophic lake as determined by 5S rRNA analysis. Applied and Environmental Microbiology 65 (7), 3164e3174. Hu, H.J., Wei, Y.X., 2006. The Freshwater Algae of China. Science and Technology Press, Beijing (In Chinese).
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 6 9 e4 1 8 2
Jensen, L.M., 1983. Phytoplankton release of extracellular organic carbon, molecular weight composition, and bacterial assimilation. Marine Ecology-Progress Series 11 (1), 39e48. Jin, X.C., Tu, Q.Y., 1990. The Standard Methods for Observation and Analysis of Lake Eutrophication, second ed. China Environmental Science Press, Beijing, China (In Chinese). Jones, S.E., Newton, R.J., McMahon, K.D., 2009. Evidence for structuring of bacterial community composition by organic carbon source in temperate lakes. Environmental Microbiology 11 (9), 2463e2472. Kalff, J., Knoechel, R., 1978. Phytoplankton and their dynamics in oligotrophic and eutrophic lakes. Annual Review of Ecology and Systematics 9, 475e495. Kan, J., Suzuki, M.T., Wang, K., Evans, S.E., Chen, F., 2007. High temporal but low spatial heterogeneity of bacterioplankton in the Chesapeake Bay. Applied and Environmental Microbiology 73 (21), 6776e6789. Langenheder, S., Ju¨rgens, K., 2001. Regulation of bacterial biomass and community structure by metazoan and protozoan predation. Limnology and Oceanography 46 (1), 121e134. Li, H.B., Xing, P., Chen, M.J., Bian, Y.Q., Wu, Q.L., 2011. Short-term bacterial community composition dynamics in response to accumulation and breakdown of Microcystis blooms. Water Research 45 (4), 1702e1710. Lindstro¨m, E.S., 2000. Bacterioplankton community composition in five lakes differing in trophic status and humic content. Microbial Ecology 40 (2), 104e113. Lindstro¨m, E.S., 2001. Investigating influential factors on bacterioplankton community composition: results from a field study of five mesotrophic lakes. Microbial Ecology 42 (4), 598e605. Lindstro¨m, E.S., Kamst-van Agterveld, M.P., Zwart, G., 2005. Distribution of typical freshwater bacterial groups is associated with pH, temperature, and lake water retention time. Applied and Environmental Microbiology 71 (12), 8201e8206. Murray, A.E., Hollibaugh, J.T., Orrego, C., 1996. Phylogenetic compositions of bacterioplankton from two California estuaries compared by denaturing gradient gel electrophoresis of 16S rDNA fragments. Applied and Environmental Microbiology 62 (7), 2676e2680. Muylaert, K., Van der Gucht, K., Vloemans, N., Meester, L.D., Gillis, M., Vyverman, W., 2002. Relationship between bacterial community composition and bottom-up versus top-down variables in four eutrophic shallow lakes. Applied and Environmental Microbiology 68 (10), 4740e4750. Muyzer, G., De Waal, E.C., Uitterlinden, A.G., 1993. Profiling of complex microbial populations by denaturing gradient gel electrophoresis analysis of polymerase chain reaction amplified genes coding for 16S rRNA. Applied and Environmental Microbiology 59 (3), 695e700. Muyzer, G., Smalla, K., 1998. Application of denaturing gradient gel electrophoresis (DGGE) and temperature gradient gel electrophoresis (TGGE) in microbial ecology. Antonie van Leeuwenhoek 73 (1), 127e141. Muyzer, G., 1999. DGGE/TGGE a method for identifying genes from natural ecosystems. Current Opinion in Microbiology 2 (3), 317e322. Obernosterer, I., Herndle, G.J., 1995. Phytoplankton extracellular release and bacterial growth: dependence on the inorganic N: P ratio. Marine Ecology-Progress Series 116 (1e3), 247e257. Pernthaler, J., Posch, T., Simek, K., Vrba, J., Pernthaler, A., Glo¨ckner, F.O., Nubel, U., Psenner, R., Amann, R., 2001. Predator-specific enrichment of Actinobacteria from a cosmopolitan freshwater clade in mixed continuous culture. Applied and Environmental Microbiology 67 (5), 2145e2155. Pinhassi, J., Hagstrom, A., 2000. Seasonal succession in marine bacterioplankton. Aquatic Microbial Ecology 21 (3), 245e256. ` ., Pinhassi, J., Sala, M.M., Havskum, H., Peters, F., Guadayol, O Malits, A., Marrase´, C., 2004. Changes in bacterioplankton
4181
composition under different phytoplankton regimens. Applied and Environmental Microbiology 70 (11), 6753e6766. Pomeroy, L.R., Wiebe, W.J., 2001. Temperature and substrates as interactive limiting factors for marine heterotrophic bacteria. Aquatic Microbial Ecology 23 (2), 187e204. Pope, P.B., Patel, B.K.C., 2008. Metagenomic analysis of a freshwater toxic cyanobacteria bloom. FEMS Microbiology Ecology 64 (1), 9e27. Poter, K.G., Feig, Y.S., 1980. The use of DAPI for identifying and counting aquatic microflora. Limnology and Oceanography 25 (5), 943e948. Qin, B.Q., Xu, P.Z., Wu, Q.L., Luo, L.C., Zhang, Y.L., 2007. Environmental issues of Lake Taihu, China. Hydrobiologia 581, 3e14. Riemann, L., Steward, G.F., Fandino, L.B., Campbell, L., Landry, M.R., Azam, F., 1999. Bacterial community composition during two consecutive NE monsoon periods in the Arabian Sea studied by denaturing gradient gel electrophoresis (DGGE) of rRNA genes. Deep Sea Research 46 (8e9), 1791e1811. Rooney-Varga, J.N., Giewat, M.W., Savin, M.C., Sood, S., LeGresley, M., Martin, J.L., 2005. Links between phytoplankton and bacterial community dynamics in a coastal marine environment. Microbial Ecology 49 (1), 163e175. Saitou, N., Nei, M., 1987. The Neighbor-Joining methodda new method for reconstructing phylogenetic trees. Molecular Biology and Evolution 4 (4), 406e425. Sekar, R., Nair, K.V.K., Rao, V.N.R., Venugopalan, V.P., 2002. Nutrient dynamics and successional changes in a lentic freshwater biofilm. Freshwater Biology 47 (10), 1893e1907. Sheng, J.R. (Ed.), 1979. Fauna Sinica, Crustacea, Freshwater Copepoda. Science Press, Academia Sinica, Beijing In Chinese. Shade, A., Kent, A.D., Jones, S.E., Newton, R.J., Triplett, E.W., McMahon, K.D., 2007. Interannual dynamics and phenology of bacterial communities in a eutrophic lake. Limnology and Oceanography 52 (2), 487e494. Smith, V.H., 1983. Low nitrogen to phosphorus ratios favor dominance by blue-green algae in lake phytoplankton. Science 221 (4611), 669e671. Tamura, K., Dudley, J., Nei, M., Kumar, S., 2007. MEGA4: molecular Evolutionary Genetics Analysis (MEGA) software version 4.0. Molecular Biology and Evolution 24 (8), 1596e1599. ter Braak, C.J.F., 1987. The analysis of vegetationeenvironment relationships by canonical correspondence analysis. Vegetatio 69 (1e3), 69e77. ter Braak, C.J.F., Verdonschot, P.F.M., 1995. Canonical correspondence analysis and related multivariate methods in aquatic ecology. Aquatic Sciences 57 (3), 255e289. Thompson, J.D., Higgins, D.G., Gibson, T.J., 1994. CLUSTAL W: improving the sensitivity of progressive multiple sequence alignment through sequence weighting, positions-specific gap penalties and weight matrix choice. Nucleic Acids Research 22 (22), 4673e4680. Tian, C., Tan, J., Wu, X., Ye, W.J., Liu, X.L., Li, D.T., Yang, H., 2009. Spatiotemporal transition of bacterioplankton diversity in a large shallow hypertrophic freshwater lake, as determined by denaturing gradient gel electrophoresis. Journal of Plankton Research 31 (8), 885e897. van Hannen, E.J., Mooij, W., Kamst-van Agterveld, M.P., Gons, H.J., Laanbroek, H.J., 1999a. Detritus-dependent development of the microbial community in an experimental system: qualitative analysis by denaturing gradient gel electrophoresis. Applied and Environmental Microbiology 65 (6), 2478e2484. van Hannen, E.J., Veninga, M., Bloem, J., Gons, H.J., Laanbroek, H.J. , 1999b. Genetic changes in the bacterial community structure associated with protistan grazers. Archiv fu¨r Hydrobiologie 145 (1), 25e38.
4182
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 6 9 e4 1 8 2
Warnecke, F., Amann, R., Pernthaler, J., 2004. Actinobacterial 16S rRNA genes from freshwater habitats cluster in four distinct lineages. Environmental Microbiology 6 (3), 242e253. Wu, Q.L., Zwart, G., Wu, J.F., Kamst-van Agterveld, M.P., Liu, S.J., Hahn, M.W., 2007a. Submersed macrophytes play a key role in structuring bacterioplankton community composition in the large, shallow, subtropical Taihu Lake, China. Environmental Microbiology 9 (11), 2765e2774. Wu, X., Xi, W.Y., Ye, W.J., Yang, H., 2007b. Bacterial community composition of a shallow hypertrophic freshwater lake in China, revealed by 16S rRNA gene sequences. FEMS Microbiology Ecology 61 (1), 85e96. Xing, P., Kong, F.X., 2007. Intra-habitat heterogeneity of environmental factors regulating bacterioplankton
community composition in Lake Taihu, China. Aquatic Microbial Ecology 48 (2), 113e122. Xu, H., Paerl, H.W., Qin, B.Q., Zhu, G.W., Gao, G., 2010. Nitrogen and phosphorus inputs control phytoplankton growth in eutrophic Lake Taihu, China. Limnology and Oceanography 55 (1), 420e432. Zwart, G., Crump, B.C., Kamst-van Agterveld, M.P., Hagen, F., Han, S.K., 2002. Typical freshwater bacteria: an analysis of available 16S rRNA gene sequences from plankton of lakes and rivers. Aquatic Microbial Ecology 28 (2), 141e155. Zwart, G., van Hannen, E.J., Kamst-van Agterfeld, M.P., Van der Gucht, K., Lindstro¨m, E.S., Van Wichelen, J., Lauridsen, T., Crump, B.C., Han, S.K., Declerck, S., 2003. Rapid screening of freshwater bacterial groups by using reverse line blot hybridization. Applied and Environmental Microbiolog 69 (10), 5875e5883.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 8 3 e4 1 9 7
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Advancing assessment and design of stormwater monitoring programs using a self-organizing map: Characterization of trace metal concentration profiles in stormwater runoff Seo Jin Ki a, Joo-Hyon Kang b, Seung Won Lee a, Yun Seok Lee a, Kyung Hwa Cho a, Kwang-Guk An c, Joon Ha Kim a,d,* a
School of Environmental Science and Engineering, Gwangju Institute of Science and Technology (GIST), Gwangju 500 712, Republic of Korea Department of Civil and Environmental Engineering, Dongguk University-Seoul, Seoul 100 715, Republic of Korea c School of Bioscience and Biotechnology, Chungnam National University, Daejeon 305 764, Republic of Korea d Sustainable Water Resource Technology Center, GIST, Gwangju 500 712, Republic of Korea b
article info
abstract
Article history:
Stormwater runoff poses a great challenge to the scientific assessment of the effects of
Received 20 December 2010
diffuse pollution sources on receiving waters. In this study, a self-organizing map (SOM),
Received in revised form
a research tool for analyzing specific patterns in a large array of data, was applied to the
21 April 2011
monitoring data obtained from a stormwater monitoring survey to acquire new insights
Accepted 22 May 2011
into stream water quality profiles under different rainfall conditions. The components of
Available online 31 May 2011
the input data vectors used by the SOM included concentrations of 10 metal elements, river discharge, and rainfall amount which were collected at the inlet and endpoint of an urban
Keywords:
segment of the Yeongsan River, Korea. From the study, it was found that the SOM dis-
Self-organizing map
played significant variability in trace metal concentrations for different monitoring sites
Urban stormwater runoff
and rainfall events, with a greater impact of stormwater runoff on stream water quality at
Stormwater monitoring programs
the upstream site than at the downstream site, except under low rainfall conditions
Effective sample size
(4 mm). In addition, the SOM clearly determined the water quality characteristics for
Event mean concentration
“non-storm” and “storm” data, where the parameters nickel and arsenic and the param-
Trace metal concentrations
eters chromium, cadmium, and lead played an important role in reflecting the spatial and temporal water quality, respectively. When the SOM was used to examine the efficacy of stormwater quality monitoring programs, between 34 and 64% of the sample size in the current data set was shown to be sufficient for estimating the stormwater pollutant loads. The observed errors were small, generally being below 10, 6, and 20% for load estimation, map resolution, and clustering accuracy, respectively. Thus, the method recommended may be used to minimize monitoring costs if both the efficiency and accuracy are further determined by examining a large existing data set. ª 2011 Elsevier Ltd. All rights reserved.
* Corresponding author. School of Environmental Science and Engineering, Gwangju Institute of Science and Technology (GIST), Gwangju 500 712, Republic of Korea. Tel.: þ82 62 715 3277; fax: þ82 62 715 2434. E-mail address:
[email protected] (J.H. Kim). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.05.021
4184
1.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 8 3 e4 1 9 7
Introduction
In recent years, wet weather pollutant discharge (i.e., stormwater pollution) has been given particular emphasis in surface water quality management, due mainly to its various impacts on environmental degradation, including (inland and coastal) water quality impairment, an increase of public health risks, and alterations or destruction of ecosystems (US EPA, 1994; Surbeck et al., 2006; Kim et al., 2007a; 2007b; Lee et al., 2007; Kang et al., 2009; Chun et al., 2010). Primary sources of pollution include all types of pollutants (e.g., suspended solids, nutrients, organic compounds, bacteria, metals, etc.; US EPA, 1994; Grout et al., 1999; Gobel et al., 2007; Kayhanian et al., 2007; Chun et al., 2010), which are diffused by surface runoff from many potential hot spots, particularly in mining zones, highway roads, agricultural lands, and urban areas (Lee et al., 2004; Kim et al., 2007a, 2007b). Depending on the properties of various pollutants, they typically experiencedonce distributeddmultiple fate and transport processes in the environment (e.g., sorption, decay, deposition, etc.), resulting in a wide range of water quality variability in space and time that current stormwater quality monitoring programs or regulations may not fully address (Surbeck et al., 2006; Lee et al., 2007). The dynamic nature of stormwater quality has been ascribed to a number of factors that may contribute to water quality variation in different ways. Previous studies have shown that the observed pollutant concentration matrix of stormwater was highly sensitive to local environmental conditions, including: i) season (e.g., seasonal first flush, rainfall pattern, antecedent dry day, etc.; Brezonik and Stadelmann, 2002; Lee et al., 2004), ii) location (e.g., land use activities, drainage area, etc.; Kim et al., 2005; Gobel et al., 2007; Kayhanian et al., 2007), iii) water quality conditions (e.g., pH, guen and Dominik, 2003; salinity, suspended solids, etc.; Guec Nguyen et al., 2005), and iv) pollutants of interest (e.g., mass loading or decay rates, composition or morphology of colloids, etc.; Grout et al., 1999; Surbeck et al., 2006). Along with the intrinsic randomness of nature, most of the above studies also showed that many complex interactions among these factors occurred in various dimensions (Kim et al., 2005). In such conditions, while conventional statistical methods, i.e., multivariate (e.g., principal component analysis and multiple linear regression) and uni/bivariate approaches (e.g., correlation and ANOVA), have been routinely employed to describe the relationship between water quality and various influence factors, the results were often prone to various sources of statistical bias unless all the requirements in the analysis (e.g., parametric assumptions, multicollinearity, interaction terms, model selection criteria, etc.; Ge and Frick, 2007; Cho et al., 2009) were rigorously addressed or adjusted. Accordingly, there appears to be an inherent difficulty in assessing the variability of stormwater quality, because all these issues currently hinder a complete understanding of water quality changes and their various effects on water resources in response to storm events. Differences in the accuracy of estimating stormwater pollutant loads are another hurdle to overcome (Hicks et al., 1997; Leecaster et al., 2002; Lee et al., 2007; Ackerman et al.,
2010). An indicator popularly used in computing mass emissions is the event mean concentration (EMC), which allows for a comparison of the contribution of pollutant loads in different local areas over any time period (Kim et al., 2005; Lee et al., 2007). However, the EMCs estimated from different stormwater monitoring programs may not be directly comparable, because various sampling schemes (e.g., grab, flow-weighted, and time-weighted composite samples) significantly affect the variability and uncertainty of the load estimations (Lee et al., 2007). For this reason, the use of EMC in assessing the efficacy of the monitoring programs from different sampling schemes is not desirable under various environmental conditions (e.g., seasonal variation in pollution sources under different antecedent conditions), possibly introducing much error in a variable pollutograph over time (Ackerman et al., 2010). In addition, there is less scientific consensus on substitution of analytical detection limits (DLs) for data evaluation (Hicks et al., 1997; Clarke, 1998), which again lessens the accuracy of EMCs. In the absence of reliable data, a deterministic modeling approach seems to be an excellent candidate for identifying the effect of sampling methods (Ackerman et al., 2010). However, the quality of such an assessment is not always certain, as the performance of the model again depends on the quality of field data, available resources, and the assumptions made. Therefore, designing effective stormwater monitoring programs may be challenging yet greatly rewarding. Compared to previous research, the present study describes a novel method for analyzing stormwater monitoring data based on the characteristics involved in water quality profiles, thereby providing an advanced understanding of the dynamic nature in stormwater quality. For this study, a self-organizing map (SOM), which was shown to have a wide range of applications in the field of data analysis in complex dimensions (Vesanto, 2002; Bieroza et al., 2009, 2010), was applied to a data set that included 10 trace metal elements in an attempt to characterize concentration profiles in stormwater runoff. Specifically, this study used the SOM: i) to investigate relationship between elements and their variability in relation to different locations and rainfall amounts, ii) to identify dominant parameters reflecting spatial and temporal stream water quality, and iii) to develop a tool for assessing and designing stormwater monitoring programs with special needs in reducing non-beneficial samples from the data set. Note that although a few case studies in stormwater monitoring survey are introduced in this study, the results drawn are sufficient for exploring such ideas in question; it is recognized that the effectiveness and accuracy of the recommended methodology can be greatly enhanced by a subsequent analysis of a large existing data set (as a future study). In addition, the study raises a question on basic functions and algorithms implemented in the SOM (e.g., the role of active/inactive neurons, map quality, and clustering accuracy), showing how such tools can be used to address the research questions. From this study, we expect that the findings not only provide valuable information on developing enforceable policies for diffuse source control, but also strongly broaden the potential applications of the SOM in the field of water resource engineering.
4185
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 8 3 e4 1 9 7
2.
Materials and methods
2.1.
Field site description
The Yeongsan River, located in the southwest region of the Korean Peninsula, is home to 1.7 million people and associated ecosystems in the South Jeolla Province. The river stretches over 135 km along the western sector of the province, and the drainage basin includes the territories of three cities and fifteen districts, with a total area of about 3500 km2. In the basin, agriculture and farming mainly dominate the land use, along with forest coverage, and there are many artificial structures (e.g., dams, dykes, and weirs) to support such activities. Although no major commercial or industrial activities have occurred in the basin in recent decades, the river is subject to pressure from physical barriers and diffuse pollution due to agricultural and urban land uses (Ki et al., 2007; Cho et al., 2009; Kang et al., 2009), routinely calling for action to reduce their effects on the receiving water body. Along the mainstream, we selected two monitoring sites, Uchi (UC) and Gwangsan (GS), located at the upper and lower sides of Gwangju City, respectively, to better characterize the water quality dynamics in an urban area during storm events. In particular, GS has been identified as a local hot spot leading to chronic water quality deterioration since the Gwangju Tributary, which merged into the mainstream in the center of the city, was found to significantly degrade the mainstream water quality (Ki et al., 2007). Around 80% of the population in the basin lived in the city, where the load capacity reached 14,160, 11,483, and 4131 kg/day for biochemical oxygen demand, total nitrogen, and total phosphorus in 2005, respectively. Around the city, the average rainfall amount and air temperature recorded 1368 mm and 13.5 C over the last thirty years (1971e2000).
2.2.
Water sampling and chemical analysis
In the monitoring sites, grab water samples were obtained using 500 mL plastic bottles during five sampling events; one non-storm and four storm events (Table 1). Over three years, each sampling event was randomly selected before a long rainy season (i.e., monsoon season) to avoid drastic changes in the river water quality and hydrological conditions. While
the sampling interval was occasionally reduced from every hour to half an hour depending on the rainfall intensity, the period was also increased to examine the amount of samples required to determine stormwater quality characteristics. Antecedent dry days ranged from 3 to 20 d during storm events that recorded low to moderate rainfall intensity. After collecting the water samples, they were all delivered to a laboratory at the Gwangju Institute of Science and Technology (GIST) within 6 h for further analysis. In the laboratory, we analyzed 20 water quality parameters, including 10 dissolved and total trace metals (aluminum (Al), chromium (Cr), iron (Fe), cobalt (Co), nickel (Ni), copper (Cu), zinc (Zn), arsenic (As), cadmium (Cd), and lead (Pb)). Dissolved and total trace elements in the water column were analyzed using inductively-coupled plasma mass spectrometry (ICP-MS; 7500ce, Agilent Technologies Inc., Palo Alto, USA) and flame atomic absorption spectrometry (FAAS; 5100, PerkineElmer Inc., Uberlingen, Germany), respectively, after membrane filtering (0.45 mm) and acid treatment (HNO3 and HCl) of the samples prior to measurement. Both standard reference materials and continuing calibration verifications analyzed at the specified frequency were within established acceptance criteria (i.e., 15%), though there were a few missing observations due to detection failures (Table 1). In cases when the concentrations of the samples fell below the minimum boundaries, they were replaced with the DLs. Note here that the DLs are proposed as a method of choice because of the following reasons: 1) there has been still active debate on substitution of the analytical DLs in various studies (e.g., fixed values, a sliding scale procedure, regression on order statistics, maximum likelihood approaches, etc.; Hicks et al., 1997; Clarke, 1998), 2) a fraction of DLs in our data set is less than 3% except for Cd (48%) and Pb (23%), 3) the performance of SOM (or neural network) is not strongly influenced by outliers and extreme values (the highest and lowest values, Astel et al., 2007; Bieroza et al., 2009, 2010), and 4) stormwater pollutant loads can be underestimated under a fixed value of zero where the true value is still unknown, thus likely producing a type II, “false-negative”, error. In fact, the difference between the two extreme cases (i.e., replacement of the values by zero and the DL) is found to be small enough that the root mean square error between values trained by SOM is less than 0.07 for all parameters. More details on sample handling,
Table 1 e Basic characteristics of the five monitoring (one non-storm and four storm) events employed in this study. Event
Non-storm Storm 1 Storm 2 Storm 3 Storm 4
Average Antecedent Sampling Date na Rainfall Storm amount duration intensity dry daysb intervals (mm/dd/yyyy) (d) (h) (mm) (h) (mm/h) 6/15/2005 1/13/2006 3/23/2007 4/19/2005 6/27/2005
28 33 31 38 58
0 4 21 24 41.5
0 6 13 8 44
0 0.7 1.6 3.0 0.9
3 20 14 8 15
1 0.5e1 1e4 0.5e1 0.5e1
Parameters analyzedc dMs, dMs, dMs, dMs, dMs,
R, R, R, R, R,
Missing datad (%)
and D e and D Cre12% (UC) D, and tMs e and D e and D Nie16% (UC), Nie2% (GS)
a Number of samples. b Antecedent dry days are counted based on the days after the most recent event exceeds 1 mm of rainfall amount. c dMs- 10 dissolved trace metals; R- rainfall amount; D- river discharge; tMs- 10 total trace metals (see the water quality analysis for a detailed list of parameters). d UC- Uchi monitoring site; GS- Gwangsan monitoring site.
4186
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 8 3 e4 1 9 7
preparation of calibration curves and DLs in the trace metal analyses can be found in Kang et al. (2009). Discharge and rainfall data were obtained from the Ministry of Land, Transport, and Maritime Affairs and the Meteorological Administration in Korea, respectively. In brief, a tipping bucket rain gauge consisting of a receiving funnel, a thermal heater, two liquid collection buckets under the funnel, and a remote recorder was used to automatically measure the amount of precipitation in the study area. Depending upon the amount of rainfall, the funnel collectors with 0.1 and 0.5 mm in diameter were used by rotation to capture both large and small rainwater, where the accuracy observed from both collectors fell below 5%. In addition, a float-type level gauge that transmitted the vertical movement of the float to the rotation of a wire drum was employed to obtain an electronic recording of the water level at stream gauging stations at 10 min sampling intervals. The stream flow rate was then estimated based on a rating curve, an equation that established the relationship between the water level and discharge for different water levels, in the gauging stations. The observed accuracy in measuring the water level and stream flow rate were less than 10 mm and 13% (12.26% for UC and 6.45% for GS), respectively, during the study period (MLTMA, 2008).
2.3.
Data analysis, SOM and statistical tests
In this study, a SOM was applied to time intensive monitoring data to better understand water quality variations under storm event conditions. The SOM, as a neural network based on an unsupervised learning algorithm (i.e., capable of learning with no prior information of data or structure), is an excellent research tool for drawing a profile of highdimensional matrix patterns in a two-dimensional domain (Astel et al., 2007; Bieroza et al., 2009, 2010). It has a wide range of engineering applications for handling complex real world problems (e.g., non-linear modeling or optimization) and is typically used for classification, clustering, prediction, modeling, and data mining (Vesanto, 2002). Topology conserving mapping (i.e., preservation of the topology present in raw data) and vector quantization (i.e., dimensionality reduction) are the main algorithms in the SOM, where four different processes (vector initialization, competition, cooperation, and adaptive processes) are implemented to eventually determine the coordinates for each corresponding observation (i.e., the locations of neurons) in the map. More information and guidance regarding the SOM is well documented in literature (Vesanto et al., 2000; Vesanto, 2002; Bieroza et al., 2009, 2010). In this study, input data vectors mainly consisted of 10 dissolved metal concentrations, river flow rate, and rainfall amount (i.e., in a 12-dimensional vector space). We normalized each data vector component in a scale from 0 to 1 after a log transformation (i.e.,log (x þ 1)), to avoid the potential effect of magnitude difference among data vectors on the final map determination. Then, the map size was automatically determined based on the number of samples in the input data pffiffiffi vectors (z 5 n). This is because the quantification error, which presents the accuracy of data represented on the map, is highly sensitive to the map size. Note also that although an increase of the size guarantees a small error (i.e., a good
quantification of data set), it is not always desirable since redundant or less significant neurons (i.e., inactive neurons) are also increased in parallel (see Section 3.3.1). Once the map size was determined, a linear initialization and a batch training algorithm were then applied to train input data vectors as the default option. Finally, prototype (or codebook) vectors representing a weighted average of the data samples were arranged on a series of low-dimensional grids, i.e., component planes of the SOM. For a qualitative correlation analysis among the different parameters, the component planes were visualized using a jet color scale. When we examined sensitivity in map resolution and clustering accuracy (see Section 3.3.2), the individual data samples that overlapped in a neuron, or more specifically the samples in an active neuron, were subsequently removed in rotation to determine the contribution of redundant samples on SOM. Such samples are, in fact, highly similar to each other in terms of the data characteristics; thus, they group together in the same neuron and can be reduced to one or very few samples to minimize the monitoring cost. For convenience, a precedent in the time series (i.e., a sample collected at an earlier time) from multiple samples was excluded from this analysis since one of them was supposed to have unique characteristicsdlike a single data sample in a neuron. Note that this example of introducing a new concept of sensitivity to the data set, which varies in size to measure the SOM performance in terms of sensitivity in map resolution or clustering accuracy, is the novel contribution of this paper, to the best of our knowledge. For consistency, the clustering accuracy was also measured under three partitioned clusters, no matter how many clusters were recommended as an optimal number of clusters in each data set (i.e., at monitoring sites over different amounts of rainfall). However, assessment of the accuracy can be better examined by optimal clustering structures depending on the objective(s) of a particular study. Here, all SOM analyses were performed by implementing the SOM toolbox in MATLAB (MATLAB 7.6, MathWorks Inc., Natick, MA, USA). In addition, two non-parametric tests, the ManneWhitney U and the KruskaleWallis tests, were employed to detect statistical differences between the two clusters and among multiple clusters, respectively, using statistical software (SPSS 15.0, SPSS Inc., Chicago, IL, USA).
3.
Results and discussion
3.1.
Variation in water quality during storm events
3.1.1.
Overview of trace metal concentration
Table 2 shows the mean and coefficient of variation (CV) values for the concentrations of the 10 trace metal elements in dissolved phase and the river discharge at UC and GS during the five monitoring events. In the table, a large CV value was typically observed during precipitation for each metal element, reflecting a great variability in stream water quality for different sampling sites and rainfall amounts. In particular, the water quality at UC during storm events was more variable than at GS (compare the CV values for metal elements), indicating that UC had quite different distribution patterns of trace metal concentrations with GS. In fact, comparatively high
4187
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 8 3 e4 1 9 7
Table 2 e Summary statistics (means and coefficients of variation in parentheses) for water quality parameters (10 dissolved trace metals) and river discharge at two monitoring sites for the five monitoring events. Al
Cr
Fe
Co
Ni
Cu
Zn
As
Cd
Pb
m3/s
mg/L Uchi 0 mm 4 mm 21 mm 24 mm 41.5 mm Gwangsan 0 mm 4 mm 21 mm 24 mm 41.5 mm Criteriaa
Discharge
14.18 (0.43) 16.76 (0.42) 9.86 (0.45) 38.50 (0.86) 8.18 (0.99)
0.22 (0.18) 0.19 (1.07) 4.45 (0.19) 0.60 (0.55) 0.90 (1.37)
7.23 (0.14) 61.39 (0.20) 25.24 (0.36) 17.58 (0.31) 19.83 (1.01)
0.11 (0.16) 0.08 (0.32) 0.19 (0.36) 0.15 (0.24) 0.45 (0.30)
0.69 (0.22) 0.26 (0.38) 3.37 (0.17) 0.76 (1.34) 20.06 (0.85)
2.87 (0.28) 1.02 (0.29) 1.96 (0.58) 3.76 (0.55) 1.91 (0.34)
8.34 (0.68) 9.99 (0.75) 4.68 (0.99) 4.04 (0.50) 9.51 (2.48)
1.09 (0.11) 0.36 (0.21) 0.47 (0.26) 0.77 (0.53) 1.50 (0.27)
0.39 (0.81) 0.04 (0.35) 0.09 (0.75) 0.05 (0.61) 0.15 (0.04)
0.14 (0.62) 0.18 (0.68) 0.37 (0.27) 0.17 (0.62) 0.06 (2.40)
5.00 (0.00) 4.09 (0.07) 6.87 (0.34) 4.95 (0.24) 12.14 (0.49)
33.36 (0.85) 31.46 (0.18) 18.65 (0.62) 12.67 (0.68) 12.04 (1.38) 340 150
0.21 (0.20) 0.60 (0.32) 0.44 (0.46) 0.35 (0.29) 0.37 (0.61) 570b (16) 74 (11)
9.65 (0.28) 69.13 (0.12) 11.90 (0.41) 14.40 (0.37) 17.39 (1.05) e e
0.13 (0.07) 0.28 (0.21) 0.21 (0.60) 0.16 (0.14) 0.13 (0.18) e e
3.32 (0.07) 6.06 (0.09) 2.89 (0.48) 3.87 (0.15) 2.22 (0.21) 470 52
3.76 (0.57) 2.67 (0.06) 2.09 (0.44) 2.63 (0.28) 2.14 (0.24) 13 9
8.66 (0.26) 17.29 (0.23) 6.15 (0.46) 9.95 (0.29) 8.83 (0.44) 120 120
1.98 (0.07) 0.99 (0.11) 0.87 (0.39) 1.54 (0.09) 1.91 (0.10) e e
0.37 (1.01) 0.03 (0.25) 0.07 (0.58) 0.03 (0.41) 0.03 (0.38) 2.0 0.25
0.39 (0.82) 0.23 (0.22) 0.19 (0.46) 0.14 (0.51) 0.07 (1.19) 65 2.5
20.71 (0.05) 17.42 (0.21) 24.10 (0.40) 36.32 (0.63) 61.93 (0.55) e e
a National acute (upper row) or chronic (lower row) criteria for the protection of freshwater aquatic life recommended by US EPA (US EPA, 2006). b The values indicate the criteria for the chromium (III) and chromium (VI, in parentheses) forms.
concentrations of trace metals were routinely measured at GS during the non-storm and 4-mm storm events, implying that the urban area (Gwangju City) served as a distributed source of many metal elements under low flow conditions (Kang et al., 2009; Ki et al., 2007). However, the average metal concentrations were comparable between the two sites, being within levels considered acceptable by the US EPA (US EPA, 2006). Here, it is also important to note that the river discharge generally increases as rainfall amount increases, whereas high amount of rainfall does not always lead to a discharge greater than the flow at low amount of rainfall. For instance, the discharge at UC during the 24-mm precipitation event was lower than the flow at 0 mm of rainfall during the rainy season, suggesting that the discharge at a specific location was strongly modulated by various factors, such as physical (e.g., land use and soil properties), hydrological (e.g., flow geometry), and meteorological (e.g., intensity and duration of rainfall and antecedent dry days) characteristics in the river basin. Such variability, in turn, produces great uncertainty in river water quality, requiring a more complicated analysis in order to provide strong connections to sources and variations of trace metals in response to storm events.
3.1.2. Relationship between water quality and other relevant parameters The potential relationship between dissolved metals and discharge during storm events was re-examined using the SOM, as illustrated in Fig. 1. In the figure, variation of the input vectors corresponding to the parameters of interest is displayed by the indexed colors in the map, where the logarithmic
values are used to reduce the impact of a large variation in the parameters. The figure showed that the metals exhibited a ubiquitous occurrence in the river, in which only As appeared to have a good correlation with the discharge (see similar distribution patterns of the component planes between As and discharge). Note that although Al, Fe, Cu, and Zn were not directly related to the discharge (or rainfall), they had a close relationship with each other and these metals except for Cu were mainly generated from the city (see wide variation for metal concentrations in the map). It should be also noted that the gradient of the map was the lowest for rainfall because the data at the two sites were only collected using a single rain gauge in the study area. The use of a single gauge prevents any meaningful analysis of spatial variability and makes the result almost entirely dependent on the influence of unquantified factors. Therefore, more study on spatial variability is still warranted to precisely support this assessment.
3.1.3.
Relationship between dissolved and total trace metals
Correlation among the parameters observed for the Storm 2 (at 21 mm of rainfall) event was further investigated in order to describe the effect of stormwater runoff on the concentrations of different forms of metals in the river (Fig. 2). Note that only a few selected metal elements (i.e., Cr, Ni, As, and Cd) were present with the rainfall and discharge since distribution of other metals in component planes was sufficiently similar to these parameters. It was found that a high degree of correlation was observed between the concentrations of total metals except for Cd, whereas the correlation was weak between the metals in dissolved form (this was much the same for the other
4188
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 8 3 e4 1 9 7
Fig. 1 e Visualization of component planes in a SOM during five monitoring (i.e., one non-storm and four storm) events. In the figure, vertical bars in the jet colormap present log-transformed values, i.e., gradients of trace metal concentrations, rainfall amount, and river discharge.
metals; data not shown). Interestingly, the total metals, again excluding Cd, presented a strong, positive correlation to the river discharge, indicating that stormwater runoff (or potentially suspended solids and colloids) was the main driver of
increases in the concentrations of the total metals (Nguyen et al., 2005). Conversely, low or negative correlations were shown between the metals in the dissolved fraction and river discharge, possibly reflecting the fact that the dissolved metals
Fig. 2 e Relationship among concentrations of 10 dissolved and total trace metals, rainfall amount, and river discharge during the Storm 2 (at 21 mm of rainfall) event. For simplicity in presentation, only a few representative trace elements (i.e., Cr, Ni, As, and Cd) are shown. In the figure, variation of parameters is illustrated in each component plane with an indexed color, and the prefixes d- and t- signify dissolved and total metal elements, respectively. (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 ) 4 1 8 3 e4 1 9 7
experienced intense, macro- and microscopic processes in the aquatic environment, such as various discharge regimes, solution chemistry, and surface interactions. For instance, it was previously shown that depending on the conditions, the dissolved metals bound to carriers were either rapidly transported into the sediment or remained free metals or metalguen and Dominik, eligand complexes in river water (Guec 2003). Furthermore, the figure showed a significant, positive correlation between the different forms of As and river discharge, again implying the strong dependence of As in both forms (here, even the total metal of As) on the river discharge. Overall, the results not only show that stormwater runoff strongly modulates the relationship between river discharge and the concentrations of metals in various forms, but also illustrates an interesting advantage of SOM over the conventional correlation analysis in that it provides a detailed description of the correlation in both space and time. A further elucidation of these benefits is in the next section.
3.2. Parameters dominating stormwater quality variation Details of how each parameter affects the variation of river water quality during storm events are of great interest. For
4189
this purpose, the entire data set was divided into two subsets, one for non-storm event (Non-storm) and the other for storm events (Storm 1e4), in which a series of graphical tools (i.e., a unified distance matrix, hit histogram, and pie chart) in the SOM toolbox were employed to assess the degree of influence of individual parameters (i.e., 10 trace metals, river discharge, and rainfall amount) on the spatial and temporal variation of water quality, respectively (Vesanto et al., 2000). Here, we believe that such an approach enlarges the potential application of the SOM, in addition to providing considerable insight into the water quality variabilitydmainly changes in river water quality between non-storm and storm events.
3.2.1.
Spatial variability
Fig. 3 illustrates the spatial difference between UC and GS for non-storm events through which the parameters affecting the difference are clearly described. In Fig. 3A, multiple hit histograms present the distribution of the data set on the map, for the corresponding monitoring sites on a unified distance matrix (U-matrix). In the figure, the data corresponding to UC and GS are shown in red and green circles, respectively, where a larger circle indicates a higher density of the data in the hexagonal map unit (or neuron), and vice versa. Also shown in grayscale is the U-matrix that represents the
Fig. 3 e Spatial difference of water quality parameters between UC (red) and GS (green) monitoring sites during the nonstorm event (0 mm): (A) hit histogram (i.e., data density indicated by the circle size) in a U-matrix, (B) relative importance of parameters indicated in each pie chart via different sizes and colors, and (C) rank of significant parameters from the highest rank in the top to the lowest in the bottom. The U-matrix represents the distance between the SOM units (or neurons), where the dark color corresponding to a large distance indicates a cluster border. Percentage statistics (i.e., means and standard deviations in error bars) for the rank are averaged based on the four large map units (see solid box in (B)). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
4190
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 8 3 e4 1 9 7
Euclidean distance among data vectors in each map unit (see units superimposed with green or red circles) as well as the distance between the neighboring map units (see hexagons in grayscale located at an intermediate position between the units). Light regions in the map indicate areas that show a high degree of similarity between the neurons (i.e., codebook vectors are close to each other), whereas dark areas corresponding to a large distance signify the borders separating them. Fig. 3A showed that the monitoring sites readily separated into two distinctive clusters (see dark areas in the middle of the map), but the samples belonging to each cluster were evenly distributed across the map. This result implies that not only do UC and GS have very different water quality characteristics during non-storm event, but also that the SOM may effectively discriminate between similar and dissimilar samples in the distance metric space. To better elucidate the dissimilarity in water quality between the upstream (UC) and downstream (GS) sites, we also investigated which parameters played an important role in determining the clusters and how individual parameters were associated with the clusters (Vesanto et al., 2000). Fig. 3B shows pie charts for the 12 parameters in the map units, where the relative importance of each parameter is described by the different sizes and colors of the charts. In the figure, individual parameters are displayed as a jet colormap, and the relative size of the charts is determined according to the Euclidean distance of a particular neuron normalized by the maximum distance. Note also that the numbers in the pie charts in Fig. 3B are smaller than the number of neurons in Fig. 3A because the U-matrix, as discussed above, includes additional neurons obtained from the distance between the adjacent units. In any case, an important aspect of the figure is the four largest pie charts, indicated by a solid black line, which specify the location of the cluster borders. These four charts were determined based on the criterion that the size of charts normalized between 0 and 1 was greater than the
median Euclidean distance of the U-matrix (see grayscale bar in Fig. 3A). This is attributed the fact that the borders, as discussed above, separate data into groups containing different water quality characteristics (i.e., UC vs GS). Since the parameters of importance were found to be largely responsible for the recommended cluster borders (Vesanto et al., 2000), we further evaluated the contributions of individual parameters to the formation of the boundaries of the four largest pie charts. Fig. 3C shows the significance of parameters on the borders, at which the relative contributions of individual parameters to the size of the pie chart (i.e., the percentage of a particular slice in the pie) are averaged and arranged in rank order. Here, error bars represent the standard deviation of the mean (n ¼ 4). From the figure, it was determined that river discharge, Ni, and As had a significant effect on the boundaries constructed by the SOM (also see the colorbar in Fig. 3B), whereas the impact on the borders was very low for the following parameters: rainfall amount, Cr, Pb, Cd, Cu, and Zn. This finding was further confirmed by an analysis that examined the statistically significant difference between the two clusters (see the MeW test results for the non-storm event in Table 3), where the parameters with little or no effect on the borders identified by the SOM (see the parameters at the bottom of Fig. 3C) displayed no measurable difference. Here, no significant difference in Pb concentration was observed between the clusters by the SOM although the MeW test did show a significant difference. This disparity can be attributed to a difference in the number of samples available for the computation (i.e., raw data for the statistical estimation and a reduced number of samples for the SOM) as well as the way they are calculated (i.e., the difference between the statistical calculations and the training algorithm in the SOM toolbox used to produce the prototype vectors). Accordingly, these results reveal that Ni and As are the major metals generated from an urban area during non-storm periods and the site-to-site variations can be largely explained by variations of these two metal concentrations.
Table 3 e Spatial and temporal differences (means and standard deviations in parentheses) of water quality and other relevant parameters among clusters during one non-storm and four storm events. Units
Non-storm event (0 mm) Cluster 1 (n ¼ 20)
Al Cr Fe Co Ni Cu Zn As Cd Pb Rainfall Discharge
mg/L mg/L mg/L mg/L mg/L mg/L mg/L mg/L mg/L mg/L mm m3/s
Cluster 2 (n ¼ 16)
14.19 (3.76) 27.84 (14.38) 0.21 (0.02) 0.21 (0.02) 7.25 (0.32) 9.30 (0.88) 0.11 (0.01) 0.13 (0.00) 0.91(0.41) 3.19 (0.23) 2.90 (0.43) 3.53 (1.10) 8.21 (2.79) 8.72 (1.23) 1.18 (0.15) 1.95 (0.06) 0.31 (0.17) 0.33 (0.20) 0.16 (0.06) 0.36 (0.18) 0.00 (0.00) 0.00 (0.00) 6.23 (2.10) 19.91 (1.40)
Four storm events (4, 21, 24, and 41.5 mm) a
MeW test (significance) 0.00b 0.84 0.00 0.00 0.00 0.09 0.16 0.00 0.94 0.00 1.00 0.00
Cluster 1 (n ¼ 30)
Cluster 2 Cluster 3 Cluster 4 (n ¼ 7) (n ¼ 12) (n ¼ 14)
11.51 (5.05) 9.43 (1.13) 9.26 (5.30) 8.75 0.39 (0.07) 3.63 (0.39) 0.85 (0.50) 0.77 14.88 (6.34) 21.61 (2.59) 7.66 (2.78) 15.67 0.18 (0.03) 0.19 (0.04) 0.24 (0.08) 0.40 3.19 (0.85) 3.03 (0.29) 1.92 (0.30) 14.21 2.30 (0.27) 1.87 (0.30) 1.57 (0.38) 2.14 8.27 (1.91) 3.46 (0.36) 2.71 (1.44) 5.92 1.55 (0.32) 0.50 (0.04) 0.77 (0.20) 1.55 0.04 (0.01) 0.09 (0.04) 0.10 (0.03) 0.13 0.12 (0.06) 0.36 (0.06) 0.15 (0.05) 0.07 0.35 (0.47) 0.50 (0.20) 0.87 (0.41) 0.67 35.80 (19.51) 6.60 (0.92) 9.89 (3.04) 12.35
Cluster 5 (n ¼ 25)
(1.67) 24.52 (6.66) (0.20) 0.62 (0.39) (5.36) 33.83 (17.01) (0.08) 0.16 (0.06) (9.49) 1.77 (1.72) (0.16) 2.44 (0.67) (3.15) 7.42 (4.29) (0.17) 0.70 (0.20) (0.02) 0.04 (0.01) (0.04) 0.19 (0.03) (0.35) 0.21 (0.33) (3.44) 8.65 (5.36)
K-W test (significance) 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
a MeW and K-W tests represent ManneWhitney (for two independent samples) and KruskaleWallis (for more than two independent samples) non-parametric tests, respectively. b Bold letters indicate a significance level of less than 0.05.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 8 3 e4 1 9 7
3.2.2.
Temporal variability
To examine temporal variation in water quality during storm events, we applied an identical procedure, as employed in the previous analysis, to four storm events (Storm 1e4). Fig. 4 illustrates the temporal difference of parameters among a series of storm events, specifically, multiple hit histograms in the U-matrix, the relative importance of parameters on the cluster borders, and their rank order (Fig. 4A, B, and C, respectively). In Fig. 4A, the four storm events (4, 21, 24, and 41.5 mm of rainfall), indicated by red, green, blue, and yellow colors, respectively, are overlaid over the U-matrixdwith various circles used to represent the data density in the map. From the figure, the water quality was found to be highly variable among the storm events (see wide variation of circles in the map) and the samples appeared to be partitioned into several clusters (see dark spots in the U-matrix). Interestingly, the samples corresponding to 4 mm of rainfall amount were concentrated in a more compact region than in the other samples (compare red vs green, blue, and yellow circles), indicating that the water quality was less variable under low rainfall (or discharge) conditions. The combined effect of a long period of antecedent dry days (20 days), low amount of rainfall (4 mm), and increased accumulation of pollutants in the ground surface during the dry season (i.e., seasonal variation in pollution sources) could be the reason for the observed stable water quality, resulting in less variable
4191
pollutant discharge to the river. Therefore, stormwater sampling needs to be carefully designed based on various properties in the system to ensure that their reflections on the monitoring program better enhance the effectiveness of Best Management Practices (BMPs) control. In such conditions, for instance, a low frequency of water sampling can be recommended to reduce monitoring cost and levels of analysis (or computational time) as long as a significant bias is not introduced in the load estimates (i.e., EMCs, see Section 3.3.1 and 2). To determine the parameters dominantly characterizing temporal variability in stormwater, we also investigated the effect of individual parameters on the borders based on the border selection criteria used in Fig. 3 (solid black line in Fig. 4B), and then ranked them in order of importance (Fig. 4C). It became apparent that Cr, Cd, and Pb, in contrast to the spatial changes in stream water quality, played an important role in the borders constructed by the SOM, whereas negligible or smaller effects were observed in the As and river discharge. However, the statistical analysis showed a significant difference in all parameters among a series of clusters, again due to the difference in computational schemes and the number of samples taken (Table 3), revealing that stormwater quality exhibiting different temporal profiles fell into distinct clusters. Consequently, the results show that the SOM has an additional benefit for describing dynamic patterns of water quality and identifying important parameters such as Cr, Cd, and Pb,
Fig. 4 e Temporal difference of water quality parameters among four storm events, 4 (red), 21 (green), 24 (blue), and 41.5 mm (yellow): (A) hit histogram (i.e., data density indicated by relative size of a circle) in a U-matrix, (B) relative importance of parameters indicated in each pie chart by different sizes and colors, and (C) rank of significant parameters from the highest rank in the top to the lowest in the bottom. The U-matrix represents the distance between the SOM units (or neurons), where a dark color corresponding to a large distance indicates a cluster border. Percentage statistics (i.e., means and standard deviations in error bars) for the rank are averaged based on the four large map units (see solid box in (B)). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
4192
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 8 3 e4 1 9 7
Fig. 5 e Classification of storm water samples (in rainfall amounts of 4, 21, 24, and 41.5 mm) by training the SOM with 12 parameters (i.e., 10 dissolved metal concentrations, river discharge, and rainfall amount): (A) dendrogram computed based on the Ward linkage method in a hierarchical cluster analysis, and (B) labels in SOM units. Different clusters are presented by Roman numerals.
by being able to characterize temporal variation in stormwater quality. Since stormwater runoff was often accompanied by an intense pollution load and variability over a short time span (Kim et al., 2005), we further analyzed temporal changes in stormwater quality using a dendrogram and labels specified in map units (Fig. 5A and B). The dendrogram was created using
the Ward linkage method in a hierarchical cluster analysis, where k-means clustering was applied to assign the optimal number of clusters via a Davies-Bouldin (DB) index, one of the indices for cluster validity (Astel et al., 2007). The DB index recorded the lowest value among a series of clusters (0.84 at k ¼ 5; dotted line in Fig. 5A), which allowed us to divide the map units into different subgroups (Roman numerals in
Table 4 e Distribution of data samples in SOM units (i.e., neurons) at two monitoring sites for the four storm eventsa. n Map sizeb
Qe
DB
Distribution of data samples on the map c
Uchi 4 mm 33 21 mm 31 24 mm 38 42 mm 58 Gwangsan 4 mm 33 21 mm 31 24 mm 38 42 mm 58
32 (8 32 (8 30 (6 40 (8
0
1
2
3
4
5
6
7
Sum of neurons Total sum Sample size (2 data samples) of neuronsd requirede (%)
4) 4) 5) 5)
0.351 0.425 0.425 0.315
0.917 1.059 0.864 0.812
12 15 10 19
8 8 11 4
11 4 5 6
1 5 1 7
0 0 1 1
0 0 2 2
0 0 0 0
0 0 0 1
12 9 9 17
20 17 20 21
61 (20/33) 55 (17/31) 53 (20/38) 36 (21/58)
30 (6 5) 32 (8 4) 32 (8 4) 40 (8 5)
0.432 0.389 0.414 0.334
0.730 0.949 0.985 1.020
12 14 7 13
7 9 17 12
7 5 3 7
4 4 5 4
0 0 0 2
0 0 0 1
0 0 0 0
0 0 0 1
11 9 8 15
18 18 25 27
55 (18/33) 58 (18/31) 66 (25/38) 47 (27/58)
a n- number of samples; Qe- quantification error; DB- Davies-Bouldin index. pffiffiffi b Map size (i.e., the number of map units) is automatically determined based on the number of samples in input data vectors (z 5 n). c 0 indicates that no data sample is included in a neuron (i.e., “inactive” neuron), whereas values 1 through 7 signify that a neuron (i.e., “active” neuron) has one or multiple data samples. d The total sum of neurons is estimated from sum of neurons which have only one data sample and those of multiple data samples (2 data samples). e The sample size required is calculated by dividing the total sum of neurons by the number of samples collected in each storm event and is represented as a percentage.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 8 3 e4 1 9 7
4193
Fig. 6 e Characterization of data samples in an active neuron for different amounts of rainfall by dimensionless time at (A) Uchi and (B) Gwangsan. In the figure, the ones that have one, two, and more than three data samples in a single map unit are indicated by white, gray, and black circles, respectively. The dimensionless time is normalized by the duration of each storm event (see Table 1), where a discrete number of time intervals indicate an elapsed time from the beginning of the storm event.
Fig. 5B). In Fig. 5B, the labels are classified based on monitoring events, sites, and frequency (i.e., data density) and each cluster is again summarized by the representative rainfall events and monitoring sites (see bold letters at the right). From the figures, it was observed that while the monitoring site UC was widely dispersed over the map depending on the amount of rainfall, the samples from GS were clustered in a confined region of the map. In particular, the monitoring site GS showed distinctive water quality patterns in response to storm events, with moderate (or high) amounts of rainfall (compare cluster 1 vs clusters 2e5), probably reflecting the fact that GS led to a chronic deterioration in water quality. However, the average concentrations of trace metals in cluster 1 were not always as high as those in the other clusters (see Table 3), which implied that the urban area was not a single driver of elevated levels of trace metals in the river. Accordingly, these results show that the monitoring site UC is highly
susceptible to storm eventsdin accordance with the results from the previous sectiondthough the impact on water quality is found to be similar between the sites under low rainfall conditions, as described above.
3.3. Advanced assessment of stormwater monitoring programs 3.3.1.
Design of effective stormwater sampling
As cost and/or effectiveness are another key issue in most water quality monitoring programs, we further expand potential areas for SOM application in designing stormwater monitoring programs. Table 4 illustrates a method of how we reduce the number of samples in storm events, where each data set with respect to different rainfall amounts and monitoring sites is separately introduced and trained by the SOM. Map sizes varied with the unique characteristics and the number of samples for
4194
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 8 3 e4 1 9 7
Table 5 e Comparison between the EMCs of three metal elements (i.e., actual loads) and those calculated from a reduced number of samples (i.e., predicted loads) at two monitoring sites for the four storm events. Uchi Cu
Gwangsan
Zn
Pb
Cu
Zn
mg/L Actual pollutant loads 4 mm 1.02 21 mm 1.73 24 mm 4.21 41.5 mm 2.00 Predicted pollutant loadsa 4 mm 0.96 (6.17%)b 21 mm 1.66 (4.54%) 24 mm 3.58 (14.86%) 41.5 mm 1.86 (6.67%)
9.08 4.37 3.76 6.45 8.44 3.67 3.76 3.99
mg/L 0.17 0.31 0.16 0.06
(7.02%) (16.01%) (0.05%) (38.14%)
Pb
0.16 0.35 0.16 0.06
2.63 1.88 2.84 2.35 (4.33%) (13.90%) (3.01%) (1.50%)
2.58 1.76 2.65 2.15
(1.76%) (6.32%) (6.74%) (8.39%)
17.01 5.12 10.04 8.76
0.23 0.16 0.15 0.07
16.43 (3.38%) 4.73 (7.79%) 9.56 (4.79%) 8.18 (6.68%)
0.22 (2.10%) 0.15 (6.57%) 0.15 (3.45%) 0.06 (14.98%)
a The predicted loads are computed from the effective sample size, which is determined in advance for the corresponding storm event (see the total sum of neurons in Table 4). b The values in parentheses indicate the percent changes of EMCs for three metal elements, which are calculated by dividing the difference between the predicted and actual loads by the actual loads and then multiplying it by 100.
each data set, and no consistent trends in Qe and DB index were observed over the different amounts of rainfall again due to their dependence on the data characteristics. When we then investigated the distribution of data samples over the map, in which a considerable number of “inactive” neurons that had no data sample in a single neuron were created along with “active” neurons in order to address a required pattern in the data set (for example, see Fig. 5B).
Most active neurons had 1e3 data samples in a single map unit, though some neurons occasionally included a much high number of samples for high amounts of rainfall (i.e., 24 and 41 mm). Because multiple samples in a neuron indicate a high degree of similarity in terms of data properties (see Section 2.3), they can be reduced to one or very few samples to decrease the overall monitoring cost. Accordingly, the total sum of neurons required for characterization of stormwater
Fig. 7 e Effects of multiple data samples overlapped in an active neuron on map resolution (Qe) and clustering accuracy (DB index) at (A and C) Uchi and (B and D) Gwangsan over a range of rainfall amounts. The distribution of data that describes the effects of such samples is indicated by a box (i.e., 25th, 50th, and 75th percentiles) and whisker (i.e., the lowest and highest values) plot, at which the number of data in different amounts of rainfall corresponds to a reduced number of samples (i.e., n e total sum of neurons, see Table 4).
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 8 3 e4 1 9 7
quality, referred to hereafter as “effective sample size”, was determined based on the sums of neurons that had a single data sample and those of multiple samples (2 data samples). The last column of the table shows the fraction of the effective samples in the total number of samples collected, elaborately indicating a method for how the sample size can be reduced for different amounts of rainfall. The proportion of the nonoverlapping (or effective) samples fluctuated between 36 and 66% over both monitoring sites and rainfall amounts due mainly to the non-uniform sampling strategies employed in this study (i.e., irregular sampling intervals and sample sizes, see Table 1). However, the greatest proportion of the overlapping samples was observed for the highest amount of rainfall, implying that the number of samples to be removed increased as the total number of samples taken highly increased. From the table, the current sample size was found to be reducible to between 34 and 64% of the total sample size (i.e., 100 e a fraction of the effective sample size). Therefore, the question becomes how these overlapping and non-overlapping samples are distributed over time. Fig. 6 illustrates the distribution of one and multiple samples in an active neuron at both monitoring sites and various rainfall amounts for a dimensionless time. In the figure, different colors (i.e., white, gray, and black circles) indicate the ones that include different amounts of data samples in a neuron, where the dimensionless time normalized by the storm duration is employed to characterize relative frequency of such samples under an identical time scale. From the figure, it was shown that non-overlapping samples were mainly distributed at an early dimensionless time (0.6). However, there were some exceptions prevailing in late dimensionless time also (see the distribution of one data sample in 24 and 41.5 mm at UC and that in 41.5 mm at GS), where overlapping samples appeared to be more distributed towards the beginning of the dimensionless time. Accordingly, these results indicate that the pattern for the frequency of overlapping and non-overlapping samples over time is not yet clear due to the dynamic nature of stormwater quality, and thus, “effective sampling points” in time were hard to determine from the current analysis, especially when considering that only a few storm events were examined.
3.3.2. Stormwater pollutant loads (EMCs) with effective sample size When the sample size is reduced in the data set (from the intensive monitoring data to an effective sample size), its effect on describing the average concentrations of parameters during storm events (i.e., EMCs) should be examined. Table 5 presents the EMCs of three metal elements (i.e., the actual loads) and those computed from the effective sample size (i.e., the predicted loads) at two monitoring sites for the four storm events. Here, Cu, Zn, and Pb are selected as representative metals because they are priority trace elements designated by the Nationwide Urban Runoff Program along with other organic pollutants (e.g., BOD, COD, TN, TKN, etc.; US EPA, 1983). For the assessment, priority is given to a precedent in time series among multiple samples by simply assuming that the samples collected at later time points may be omitted. From the table, it was determined that the actual loads of the three metal elements largely varied with rainfall amounts
4195
rather than monitoring sites, which appeared to be primarily attributable to seasonal variation in pollution sources. The observed pollutant loads were the highest for Zn, followed by Cu and Pb. When a comparison was made between the actual and predicted loads, the predicted loads computed from the effective sample size mostly underestimated the actual loads (see many negative fractions). The difference between the two loads was greater at UC than at GS and the largest for Zn. The percent changes of the loads for three metal elements ranged from 0 to 38% over both monitoring sites and rainfall amounts in terms of absolute value, but generally fell below 10%dexcept for a few extreme cases. This result indicates that reduction of the sample size in the data set does not significantly affect the accuracy in assessing stormwater pollutant loads, and thus, the recommended approach can be employed as a method of choice either to minimize the monitoring cost or to trade it off against the accuracy. Therefore, the question again arises as to whether reduction of the sample size also influences sensitivity in map resolution and clustering accuracy (see Fig. 7). In the figure, the effects of size reduction in the data set on map resolution (Qe) and clustering accuracy (DB index) is quantified by a box and whisker plot, where the data samples overlapped in a neuron are removed by rotation with respect to data in the time series; again, except for the precedent in multiple samples. In the figure, no obvious pattern of changes in map resolution and clustering accuracy was detected over a range of rainfall amounts at both monitoring sites. The observed variations of map resolution and clustering accuracy were within the range of 6% and 20% in both directions, respectively, implying that the size reduction was not a significant concern of the SOM, in particular for the map resolution, which indicated the accuracy of data on the map. However, further investigation pertaining to the selection of an effective sample size is highly recommended in order to verify all these effects and to more accurately support such an assessment.
4.
Conclusions
In the present study, a SOM was applied to a set of monitoring data collected during non-storm and storm events in an attempt to gain new insights into stormwater quality, i.e., its dynamic patterns across time and space and their impacts on description of stormwater pollutant loads. The main conclusions drawn from this study are as follows. Stomwater quality is highly variable depending on the monitoring site and rainfall amount, although only a few storm events were examined in this study. Variability in river discharge in response to storm events makes it difficult to directly spot the impact of storms on stream water quality (here, the concentrations of metals), where conventional approaches are not likely to describe the nonlinear relationship between the parameters well. Conversely, the SOM has the ability to successfully illustrate an overview of the changes in stormwater quality. However, application of the SOM to more relevant parameters in the time series (e.g., spatial variations of precipitation) is
4196
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 8 3 e4 1 9 7
strongly recommended in order to provide additional information and/or clarification on stormwater quality. The SOM clearly distinguishes samples resulting from the dissimilarity between all pairs of attributes in a distance metric space. Indeed, the dynamic nature of stormwater quality becomes evident when the SOM is applied to different subsets of monitoring events (i.e., “non-storm” and “storm” data), where the parameters Ni and As and the parameters Cr, Cd, and Pb are specified as major metals that reflect spatial (i.e., between the monitoring sites) and temporal changes (i.e., between the storm events) in water quality, respectively. The influence of storms on water quality is higher for the upstream site UC than for the downstream site GS, except under low rainfall conditions (<4 mm). The SOM also sheds new light on the efficacy of stormwater monitoring programs. Specifically, the SOM presents a method of how an effective sample size may be determined, where between 34 and 64% of the current sample size is shown to be removable from the data set. However, the timing of effective sampling is hard to determine from the analysis of only a few case studies, again due to high variability of stormwater quality. The SOM with its modified algorithm also elucidates the effects of (reducing) sample size on stormwater pollutant loads assessment (i.e., EMCs) and data characterization (i.e., map resolution and clustering accuracy). The observed effects are small enough that the variations of the loads, map resolution, and clustering accuracy are generally less than 10, 6, and 20%, respectively. Therefore, the recommended approach may be employed to minimize monitoring cost, once their accuracy and efficiency are adequately determined by further examination of a large data set.
Acknowledgments This research was supported by Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Education, Science and Technology (No. 2010-0011822). We also acknowledge the high quality public services provided by the Ministry of Land, Transport, and Maritime Affairs and Meteorological Administration in Korea in supplying the monitoring data.
references
Ackerman, D., Stein, E., Ritter, K., 2010. Evaluating performance of stormwater sampling approaches using a dynamic watershed model. Environmental Monitoring and Assessment. doi:10. 1007/s10661-010-1788-6. Astel, A., Tsakouski, S., Barbieri, P., Simeonov, V., 2007. Comparison of self-organizing maps classification approach with cluster and principal components analysis for large environmental data sets. Water Research 41 (19), 4566e4578. Bieroza, M., Baker, A., Bridgeman, J., 2009. Exploratory analysis of excitation-emission matrix fluorescence spectra with selforganizing maps as a basis for determination of organic
matter removal efficiency at water treatment works. Journal of Geophysical Research-Biogeosciences 114. doi:10.1029/ 2009JG000940 G00F07. Bieroza, M., Baker, A., Bridgeman, J., 2010. Classification and calibration of organic matter fluorescence data with multiway analysis methods and artificial neural networks: an operational tool for improved drinking water treatment. Environmetrics. doi:10.1002/env.1045. Brezonik, P.L., Stadelmann, T.H., 2002. Analysis and predictive models of stormwater runoff volumes, loads, and pollutant concentrations from watersheds in the twin cities metropolitan area, Minnesota, USA. Water Research 36 (7), 1743e1757. Cho, K.H., Kang, J.H., Ki, S.J., Park, Y., Cha, S.M., Kim, J.H., 2009. Determination of the optimal parameters in regression models for the prediction of chlorophyll-a: a case study of the Yeongsan reservoir, Korea. Science of the Total Environment 407 (8), 2536e2545. Chun, J.A., Cooke, R.A., Kang, M.S., Choi, M., Timlin, D., Park, S.W., 2010. Runoff losses of suspended sediment, nitrogen, and phosphorus from a small watershed in Korea. Journal of Environmental Quality 39 (3), 981e990. Clarke, J.U., 1998. Evaluation of censored data methods to allow statistical comparisons among very small samples with below detection limit observations. Environmental Science & Technology 32 (1), 177e183. Ge, Z.F., Frick, W.E., 2007. Some statistical issues related to multiple linear regression modeling of beach bacteria concentrations. Environmental Research 103 (3), 358e364. Gobel, P., Dierkes, C., Coldewey, W.C., 2007. Storm water runoff concentration matrix for urban areas. Journal of Contaminant Hydrology 91 (1e2), 26e42. Grout, H., Wiesner, M.R., Bottero, J.Y., 1999. Analysis of colloidal phases in urban stormwater runoff. Environmental Science & Technology 33 (6), 831e839. guen, C., Dominik, J., 2003. Partitioning of trace metals Guec between particulate, colloidal and truly dissolved fractions in a polluted river: the Upper Vistula River (Poland). Applied Geochemistry 18 (3), 457e470. Hicks, P., McBean, E., Quigley, S., Polan, B., 1997. Issues regarding the application of a mass-balance equation to an urban Creek. In: James, W. (Ed.), (Chapter 20) in Advances in Modeling the Management of Stormwater Impacts, vol. 5. Computational Hydraulics International, Guelph, Ontario, Canada, pp. 327e350. Kang, J.H., Lee, Y.S., Ki, S.J., Lee, Y.G., Cha, S.M., Cho, K.H., Kim, J. H., 2009. Characteristics of wet and dry weather heavy metal discharges in the Yeongsan Watershed, Korea. Science of the Total Environment 407 (11), 3482e3493. Kayhanian, M., Suverkropp, C., Ruby, A., Tsay, K., 2007. Characterization and prediction of highway runoff constituent event mean concentration. Journal of Environmental Management 85 (2), 279e295. Ki, S.J., Lee, Y.G., Kim, S.W., Lee, Y.J., Kim, J.H., 2007. Spatial and temporal pollutant budget analyses toward the total maximum daily loads management for the Yeongsan watershed in Korea. Water Science & Technology 55 (1e2), 367e374. Kim, G., Chung, S., Lee, C., 2007a. Water quality of runoff from agricultural-forestry watersheds in the Geum River Basin, Korea. Environmental Monitoring and Assessment 134 (1e3), 441e452. Kim, L.H., Kayhanian, M., Zoh, K.D., Stenstrom, M.K., 2005. Modeling of highway stormwater runoff. Science of the Total Environment 348 (1e3), 1e18. Kim, L.H., Ko, S.O., Jeong, S., Yoon, J., 2007b. Characteristics of washed-off pollutants and dynamic EMCs in parking lots and bridges during a storm. Science of the Total Environment 376 (1e3), 178e184.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 8 3 e4 1 9 7
Lee, H., Lau, S.L., Kayhanian, M., Stenstrom, M.K., 2004. Seasonal first flush phenomenon of urban stormwater discharges. Water Research 38 (19), 4153e4163. Lee, H., Swamikannu, X., Radulescu, D., Kim, S.J., Stenstrom, M.K. , 2007. Design of stormwater monitoring programs. Water Research 41 (18), 4186e4196. Leecaster, M.K., Schiff, K., Tiefenthaler, L.L., 2002. Assessment of efficient sampling designs for urban stormwater monitoring. Water Research 36 (6), 1556e1564. Ministry of Land, Transport, and Maritime Affairs, Korea (MLTMA), 2008. The 2007 annual stream flow report: VI. Discharge measurements at gaging stations in the Yeongsan river. Government Publication Registration Number 111611000-000034-10. MLTMA, Gyeonggi Province, Korea. Nguyen, H.L., Leermakers, M., Elskens, M., De Ridder, F., Doan, T.H., Baeyens, W., 2005. Correlations, partitioning and bioaccumulation of heavy metals between different compartments of Lake Balaton. Science of the Total Environment 341 (1e3), 211e226. Surbeck, C.Q., Jiang, S.C., Ahn, J.H., Grant, S.B., 2006. Flow fingerprinting fecal pollution and suspended solids in
4197
stormwater runoff from an urban coastal watershed. Environmental Science & Technology 40 (14) , 4435e4441. U.S. Environmental Protection Agency (EPA), 1983. Results of the Nationwide Urban Runoff Program. Volume I-Final report. NTIS access number PB84e185552. Water Planning Division, Washington, DC, USA. U.S. Environmental Protection Agency (EPA), 1994. Nonpoint sources pollution control program. Report number 841-F-94005. Office of Water, Washington, DC, USA. U.S. Environmental Protection Agency (EPA), 2006. National recommended water quality criteria. Report number 4304T. Office of Water, Washington, DC, USA. Vesanto, J., 2002. Data exploration process based on the selforganizing map. Ph.D. dissertation. Available at:. Helsinki University of Technology, Espoo, Finland http://lib.tkk.fi/Diss/ 2002/isbn9512258978/isbn9512258978.pdf. Vesanto, J., Himberg, J., Alhoniemi, E., Parhankangas, J., 2000. SOM toolbox for Matlab 5. Report A57. Available at:. Helsinki University of Technology, Espoo, Finland http://www.cis.hut. fi/somtoolbox/package/papers/techrep.pdf.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 9 8 e4 2 1 0
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Synergistic effect of nickel ions on the coupled dechlorination of trichloroethylene and 2,4-dichlorophenol by Fe/TiO2 nanocomposites in the presence of UV light under anoxic conditions Ganesh K. Parshetti, Ruey-an Doong* Department of Biomedical Engineering and Environmental Sciences, National Tsing Hua University, 101, sec. 2, Kuang-Fu Road, Hsinchu 30013, Taiwan
article info
abstract
Article history:
The coupled removal of priority pollutants by nanocomposite materials has recently been
Received 25 December 2010
receiving much attention. In this study, trichloroethylene (TCE) and 2,4-dichlorophenol
Received in revised form
(DCP) in aqueous solutions were simultaneously removed by Fe/TiO2 nanocomposites
12 May 2011
under anoxic conditions in the presence of nickel ions and UV light at 365 nm. Both TCE
Accepted 22 May 2011
and DCP were effectively dechlorinated by Fe/TiO2 nanocomposites, and the pseudo-first-
Available online 1 June 2011
order rate constants (kobs) for TCE and DCP dechlorination were (1.39 0.05)102 and
Keywords:
iron alone. In addition, the kobs for DCP dechlorination was enhanced by a factor of 77 when
Fe/TiO2 nanocomposites
Fe/TiO2 was illuminated with UV light for 2 h. Hydrodechlorination was found to be the
Simultaneous photodechlorination
major reaction pathway for TCE dechlorination, while DCP could undergo reductive
Trichloroethylene
dechlorination or react with hydroxyl radicals to produce 1,4-benzoquinone and phenol.
2,4-Dichlorophenol
TCE was a stronger electron acceptor than DCP, which could inhibit the dechlorination
Ni
efficiency and rate of DCP during simultaneous removal processes. The addition of nickel
(1.08 0.05)102 h1, respectively, which were higher than that by nanoscale zerovalent
ions significantly enhanced the simultaneous photodechlorination efficiency of TCE and DCP under the illumination of UV light. The kobs values for DCP and TCE photodechlorination by Fe/TiO2 in the presence of 20e100 mM Ni(II) were 30.4e136 and 13.2e192 times greater, respectively, when compared with those in the dark. Electron spin resonance analysis showed that the photo-generated electron-hole pairs could be effectively separated through Ni ions cycling, leading to the improvement of electron transfer efficiency of TCE and DCP by Fe/TiO2. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
Chlorinated compounds such as tetrachloroethylene, trichloroethylene (TCE), and chlorophenols are the most common priority pollutants in the aquatic environment. These
chemicals often coexist in the contaminated environments and present potential threats to the health of the environment and human beings (Riley et al., 1992). Several technologies employing physical, chemical and biological methods have been used to effectively remove priority pollutants in the
* Corresponding author. Tel.: þ886 3 5726785; fax: þ886 3 5718649. E-mail address:
[email protected] (R.-a. Doong). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.05.019
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 9 8 e4 2 1 0
aquatic environment (Shen et al., 2005; Choi et al., 2008; Lee and Doong, 2011). However, these methods were developed for treatment of a single pollutant. Therefore, the development of novel techniques for coupled removal of mixed contaminants is urgently demanded. Nanoscale zerovalent irons (NZVI) have been successfully used for the treatment of various contaminants including chlorinated hydrocarbons, nitroaromatics, arsenic and chromate (Dries et al., 2005; Doong and Lai, 2005; Li and Zhang., 2007). Several studies have depicted that chlorinated compounds including TCE and DCP can undergo reductive dechlorination by NZVI to produce less-chlorinated homologues (Zheng et al., 2008; Sunkara et al., 2010; Hou et al., 2010). In addition, bimetallic nanoparticles including Pd/Fe, Ni/Fe, Cu/Al and Ni/Si have been found to be promising materials for the degradation of a wide variety of priority pollutants (Zhu and Lim, 2007; He and Zhao, 2008; Choi et al., 2008; Lee and Doong, 2008; Tee et al., 2009). Hydrodechlorination is believed to be the major reaction pathways for the dechlorination of chlorinated ethylenes by NZVI in the presence of second catalytic metal ions, leading to the formation of non-halogenated compounds such as ethane and ethylene (Doong and Lai, 2006; He and Zhao, 2008; Parshetti and Doong, 2009). The immobilization of NZVI onto a support is also a plausible strategy for enhancing the dechlorination efficiency and rate of priority pollutants under anoxic conditions because of the decreased agglomeration of nanoparticles and the protection of nanoparticles from oxidation (Xu and Bhattacharyya, 2005; Parshetti and Doong, 2009). Several materials such as granular activated carbon, metal oxides, membranes, and polymer nanofibers have been employed as the supports to prevent the agglomeration of NZVI and to accelerate the dechlorination efficiency and rate of chlorinated hydrocarbon (Schrick et al., 2004; Darab et al., 2007; Choi et al., 2008; Liu et al., 2009), showing that the combination of NZVI with support is an attractive strategy for the enhancement of NZVI activity towards dechlorination of chlorinated compounds. Titanium dioxide (TiO2) is one of the most often used photocatalysts for removal of organic pollutants in the environment because of the advantages of the low cost, high stability, nontoxicity, and high photocatalytic activity (Chang et al., 2009; Han et al., 2009; Doong et al., 2010). TiO2 can form electron-hole pairs under the illumination of near-UV light, which encompasses energies higher than the corresponding band gap. When heterogeneous photocatalysis is applied to the oxidation of pollutants, holes and electrons would react with water and oxygen, respectively, to produce O-centered radical adducts for catalytic oxidation of DCP and TCE (Ou and Lo, 2007; Liu et al., 2009). In addition, the photodechlorination of TCE by TiO2, wherein the chlorine atoms are displaced effectively by hydrogen atoms, was also reported when exposure to UV light (Chu and Choy, 2000). However, the high degree of recombination of electrons and holes is a ratelimiting factor controlling the photocatalytic efficiency (Yu et al., 2009a). The addition of metal ions such as Fe, Pt, Mo, V, and Ag as dopants to reduce the recombination rate is one of the promising methods to improve the degradation efficiency and rate of organic pollutants in aqueous solutions by TiO2 (Chang and Doong, 2006; Doong et al., 2009, 2010; Yu
4199
et al., 2009b). More recently, the combination of NZVI with TiO2 (Fe/TiO2 nanocomposites) has been reported to effectively degrade pollutants (Huang et al., 2007; Liu et al., 2009; Hsieh et al., 2010). Huang et al. (2007) depicted that the sol-gel-derived Fe/TiO2 nanoparticles could enhance the photodegradation efficiency of azo dye under the illumination of UV light. In addition, a composite membrane with NZVI, TiO2 and activated carbon fiber was prepared for photocatalytic degradation of 2,4-dichloropheol (DCP) (Liu et al., 2009). Since NZVI has a strong reducing power for reduction of chlorinated hydrocarbons and TiO2 is an effective photocatalyst for degradation of organic pollutants, the Fe/TiO2 nanocomposites can retain both the dechlorination and photocatalytic activity for removal of chlorinated compounds under anoxic conditions. More recently, our study showed that the presence of nickel ions could enhance the dechlorination efficiency and rate of TCE by Fe/TiO2 nanocomposites (Parshetti and Doong, 2010). However, the use of Fe/TiO2 nanocomposites for coupled removal of chlorinated aliphatic and aromatic compounds in the presence of nickel ions as the dopant as well as catalytic metal ion has rarely been reported. In addition, the role of nickel ions in dechlorination of TCE and DCP in the presence of UV light remains unclear. In this study, the coupled dechlorination of TCE and DCP by Fe/TiO2 nanocomposites in the presence of nickel ions and UV light was investigated under anoxic conditions. The morphology, crystallinity, and chemical species of Fe/TiO2 nanocomposites were characterized by scanning electron microscopy (SEM), X-ray diffraction (XRD), and X-ray photoelectron spectroscopy (XPS), respectively. The Fe/TiO2 nanocomposites were then used for dechlorination and photocatalytic degradation of TCE and DCP in the presence and absence of 365 nm UV light. It is noted that the bandgap of P-25 TiO2 is 3.2 eV, which can absorb the wavelength lower than 380 nm. Therefore, 365 nm of UV light was used to illuminate P-25 TiO2 for minimizing the possible interference of direct photolysis. In addition, the influence of Ni loading on the photodechlorination rates of TCE and DCP was evaluated. The reaction kinetics as well as the mechanisms for TCE and DCP photodechlorination were also proposed and examined.
2.
Materials and methods
2.1.
Chemicals
Trichloroethylene (TCE) (>99.8%, GC grade), HEPES (N[2-Hydroxyethyl] piperazine-N 0 -[2-ethanesulfonic acid]) (C8H18N2O4S), and poly(ethylene glycol) (PEG, MW 8400) were purchased from SigmaeAldrich Co. (Milwaukee, WI). Degussa P-25 titanium dioxide (P-25 TiO2), a mixture of 80% anatase and 20% rutile, was obtained from Degussa Co. All other chemicals were of analytical grade and were used as received without further purification. Deoxygenated solutions were prepared by purging deionized water (Millipore, 18.3 MU cm) with high-purity nitrogen gas (99.995%) in vacuum-sealed bottles. This process was repeated 4e5 times to remove trace amount of oxygen in the solution (Maithreepala and Doong, 2004; Doong and Lai, 2006).
4200
2.2.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 9 8 e4 2 1 0
Synthesis of Fe/TiO2 nanocomposite
Nanoscale Fe/TiO2 nanoparticles were prepared by dissolving 1.19 M of PEG into 50 mL of ethanol solution containing 50 mg P-25 TiO2 and stirred for 30 min to form a homogeneous sol solution of PEG-TiO2. Then 50 mL of 0.33 M FeSO4,7H2O was added into the PEG-TiO2 sol solution and Fe/TiO2 nanoparticles were obtained by dropwise addition of 0.5M NaBH4 into the solution under stirring at room temperature. The formation of the Fe/TiO2 nanoparticles was visually confirmed when the solution turned slightly black in color. An excess of borohydride was applied to accelerate the synthesis reaction and then the solution was centrifuged at 10,000 g for 10 min to harvest the precipitates. The precipitate, consisting of Fe/TiO2 nanoparticles, was washed with ethanol several times and was completely dried under a gentle flow of nitrogen gas. The synthesis of pure NZVI nanoparticles was carried out under anoxic conditions by dissolving 0.2 M FeSO4$7H2O in deionized water at pH 8.0 followed by dropwise addition of 0.5 M NaBH4 to form the black precipitate. The precipitates were then washed with ethanol and dried under a N2 atmosphere.
2.3.
2.4.
Photodechlorination of TCE and DCP
The photocatalytic activity of Fe/TiO2 nanocomposites towards DCP and TCE dechlorination was conducted using batch reactor. The solutions were continuously purged with high-purity nitrogen gas to maintain anoxic conditions during the experiment. Four 8 W mercury lamps at 365 nm were positioned within the central part of the photoreactor and cooling water was circulated through a Pyrex jacket surrounding the lamps. The Fe/TiO2 nanocomposites at 3.5 g/L were suspended in a 20-mL quartz tube containing 10 mg/L DCP. In addition, magnetic stirring with a Teflon-coated stir bar was used to well-mix the suspension. For the photodechlorination of TCE, a 50-mL Pyrex serum bottle containing 20 mL of HEPES buffer solution and 10 mM TCE was used. After Fe/TiO2 nanoparticles were added to the solutions, samples were sonicated for 20 min under anoxic conditions to obtain good dispersion of the nanoparticles. Immediately after the sonication, the sample was magnetically stirred and light was turned on. A 2-mL sample was withdrawn from the solution at various time intervals for analysis after centrifugation at 14000 rpm for 10 min.
Dechlorination of TCE and DCP
The dechlorination of TCE and/or DCP by Fe/TiO2 nanocomposites was carried out using 50-mL serum bottles sealed with Teflon-lined rubber septa and aluminum crimp caps. N2purged serum bottles were filled with 20 mL of deoxygenated buffer solutions and 3.5 g/L of Fe/TiO2 nanoparticles. 10 mM HEPES buffer solution was used to control pH at 7.2 0.1. After being capped with rubber septa and aluminum caps, the serum bottles were purged with N2 immediately to maintain the anoxic conditions. The deoxygenated stock solutions of TCE and DCP were then introduced into the system to get the final concentrations of 1.3 mg/L (10 mM) and 10 mg/L, respectively. The selection of a nearly order-of-magnitude difference in the DCP and TCE concentrations is mainly attributed to that the TCE concentration in contaminated environments can be up to the level of mg/L (Sloto, 2010), while the DCP concentration in industrial wastewater is often at least one order of magnitude higher than that of TCE. In this study, the addition of catalytic metal ions including Co(II), Ni(II), and Cu(II) for accelerating the dechlorination efficiency and rate of TCE was examined. The preliminary results are shown in Fig. S1 (see supplementary materials). An incubation period of 24 h was required to achieve 50% and 95% dechlorination of the original TCE when 1 mM Co(II) and Cu(II), respectively, were added. In contrast, a nearly complete dechlorination of TCE was obtained within 30 min when 1 mM Ni(II) was added. Therefore, low concentrations of Ni(II) ranging from 20 to 100 mM were added to evaluate the synergistic effect of second metal ion on simultaneous photodechlorination of TCE and DCP by Fe/TiO2 nanocomposites. The deoxygenated stock solutions of Ni(II) ions were delivered into serum bottles by gas-tight syringes. The bottles were incubated in an orbital shaker at 100 rpm and at 25 1 C in the dark. Control experiments were also preformed in the absence of Fe/TiO2 nanoparticles. All experiments were carried out in duplicate or triplicate and the average values were reported.
2.5.
Analysis of TCE and DCP
The headspace analytical technique was used for the determination of chlorinated and non-chlorinated hydrocarbons. Concentrations of TCE and the byproducts in the headspace of the test bottles were monitored by withdrawing 40 mL of gas from the headspace using a 50-mL gas-tight syringe. The headspace sample was immediately injected into a gas chromatograph (GC) equipped with an electron capture detector (ECD) and a flame ionization detector (FID) (PerkineElmer, Autosystem, Norwark, CT). A 60-m VOCOL fused-silica megabore capillary column (0.545 mm 3.0 mm, Supelco Co.) was employed to separate the organic compounds. The column was connected to the FID and ECD simultaneously by a Y-splitter; and optimal sensitivity for chlorinated hydrocarbons was achieved with 40% of the flow (1.85 ml/min) to the ECD. The column temperature was isothermally maintained at 120 C using ultra-high-purity nitrogen (>99.9995%) as the carrier gas. The temperatures of the ECD and FID were maintained at 325 and 250 C, respectively. The relative standard deviations (RSD) of the ECD analyses were within 10% and those for the FID analyses were within 5%. Control samples were also used to check for the possible leakage of target compounds during the incubation process. The concentrations of chlorinated hydrocarbons in aqueous solutions were calculated using the external standard method by preparing the known concentrations of chlorinated hydrocarbons in aqueous solutions. The aqueous concentration of DCP was determined by high performance liquid chromatography (HPLC) equipped with a variable wavelength UV detector and an autosampler (Agilent technologies 1200 series). A Supelcosil LC-18 column (25 cm 0.46 mm 5.0 m) was employed for determining the DCP. The mobile phase was a mixture of 80/20 (v/v) ethanolewater mixture. The eluent was delivered at a rate of 1.0 mL/min and the wavelength for detection was 280 nm.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 9 8 e4 2 1 0
2.6.
GC/MS analysis
The intermediates and end products of DCP after photodechlorination were identified by GCeMS. After photodechlorination of DCP by Fe/TiO2 nanocomposites, the mixture was centrifuged at 10,000 rpm for 20 min. The reaction products in the supernatant were extracted using dichloroethane. One gram of anhydrous Na2SO4 was added to the extract for removal of trace amounts of water, and then concentrated to about 1 mL on a rotary evaporator (EYELAOSB-2100). The separation and identification of intermediates and end products were carried out by a HewlettePackard 6890 GC equipped with a 5973 MSD. A 30-m HP-5MS column (0.25 mm 0.25 mm) was used to separate the organics. The column temperature was initially set at 45 C for 1 min, increased to 280 C at a rate of 24 C min1, then ramped up by 5 C min1 to 310 C, and held at that temperature for 1 min. Helium at a flow rate of 1 mL min1 was used as the carrier gas, and the temperature of injector was maintained at 250 C. The ionization was carried out in the electron impact mode (70 eV). The electron multiplier voltage and automatic gain control target were set automatically. The transfer line and ion trap mainfold were set at 280 and 230 C, respectively. The mass range scanned was from 50 to 550 amu under full scan acquisition mode.
2.7.
Surface characterization of Fe/TiO2 nanocomposite
The surface morphology of the Fe/TiO2 nanocomposites was determined by SEM (JEOL JSM-6700F Oxford Inca Energy 400).
4201
All the samples were Au-coated using Ion Sputter e-1030 (Hitachi, Japan). The accelerating voltage and the applied current were 5 kV and 20 nA, respectively. After coating with Au, samples were placed under high vacuum (103e 107 mbar) conditions. The scanning electron image (SEI) resolution is 1.5 nm at 15 kV and 5 nm at 1.0 kV. The crystallinity of the Fe/TiO2 nanocomposites was determined by X-ray diffraction (XRD) using Rigaku (Rigaku, D-Max III VC, Japan) diffractometer with Cu Ka radiation. The XPS measurement was performed by an ESCA PHI 1600 photoelectron spectrometer (Physical Electronics, Eden Prairie, MN) using Al Ka X-ray source (1486.6 eV photon energy). The spherical capacitor analyzer with a multichannel detector had a takeoff angle of 70 related to the horizontal plane of the sample. Data were recorded digitally, and all peak scans were signal averaged until an acceptable signal-to-noise ratio was obtained. The pressure in the sample chamber was maintained below 2.5 108 torr, and the binding energies of the photoelectrons were determined under the assumption that carbon has a binding energy of 284.8 eV. The production of free radicals was identified using an electron paramagnetic resonance (EPR) spectrometer (Bruker, EMX-10, Germany) equipped with an NMR Gaussmeter. One mL of anoxic solution was withdrawn using a N2-purged plastic syringe and immediately injected into the flat-cell under a continuous purge of nitrogen gas to prevent oxidation of the samples. The flat-cell was designed for aqueous solutions to minimize the interference of water during EPR analysis. The organic radical anions present in solutions were
Fig. 1 e SEM images of (a) P-25 TiO2 nanoparticles (b) Fe/TiO2 nanocomposites. Figure (c) and (d) are the EDS analysis of P-25 TiO2 and Fe/TiO2 nanocomposite, respectively.
4202
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 9 8 e4 2 1 0
analyzed immediately at 20 C by EPR following the transfer of the flat-cell to the cavity (microwave frequency: X-band, 9.78 GHz; microwave power: 6.331 and 0.632 mW; modulation frequency: 100 kHz; modulation amplitude: 0.1 G).
3.
Results and discussion
3.1.
Characteristics of Fe/TiO2 nanocomposites
The SEM was first used to examine the morphology of Fe/TiO2 nanocomposites. The SEM images shown in Fig. 1 indicates that the particle size of pristine TiO2 was 28 7 nm, which is close to the reported diameter of P-25 TiO2. After addition of ferrous ion and NaBH4, the NZVI with particle size of 58 12 nm was obtained. The TiO2 nanoparticles were embedded onto the NZVI surfaces to form Fe/TiO2 nanocomposites. The EDS analysis from SEM images clearly showed peaks of Ti and Fe in the Fe/TiO2 nanocomposites. However, the Fe/TiO2 nanocomposites cannot be generated in the absence of PEG, showing that the water-soluble PEG serves as a cross-linking agent for combination of TiO2 with NZVI. PEG is a bifunctional polymer with reactive OH groups on both ends, and can immobilize NZVI onto different supports including membranes, polymers and metal oxides (Parshetti and Doong, 2009). The presence of PEG increases the surface hydroxyl content of the inside nanocomposites, which is beneficial for surface modification to form Fe/TiO2 nanoparticles. The XRD patterns of P-25 TiO2 and Fe/TiO2 nanocomposites were also examined to understand the crystallinity of the nanocomposites (Fig. S2, see supplementary materials). For P-25 TiO2, peaks centered at 25.37 , 38.02 , and 48.07 2q were clearly observed, which can be assigned as (101), (004), and (200) orientations of the anatase phase. In addition, a small rutile peak centered at 27.44 2q was also observed, which was in good agreement with the reference patterns for P-25 TiO2. For Fe/TiO2 nanocomposites, several additional peaks at 37.02 , 45.14 and 65.46 2q were present, depicting the presence of NZVI. The XRD patterns of the Fe/TiO2 matched well with those corresponding JCPDS standards of Fe and TiO2 (JCPDS 06-0696 for Fe; JCPDS 21-1272 for anatase; JCPDS 211276 for rutile), clearly indicating the nature of composite structures of Fe/TiO2 nanoparticles. It is noted that the XRD peaks of TiO2 became weak and broad after the incorporation of NZVI into TiO2 nanostructures, presumably attributed to the deposition of NZVI onto the TiO2 surface. The XPS spectra of Fe/TiO2 nanocomposites were examined to understand the chemical speciation of the nanocomposites (Fig. 2). The Ti2p spectra showed peaks at 459.0 and 464.6 eV, which could be assigned as Ti2p3/2 and Ti2p1/2, respectively. The doublet separation between the 2p1/2 and 2p3/2 peaks was found to be 5.6 eV, which is characteristic of TiO2. In addition, two photoelectron peaks at 707.8 and 720.2 eV were observed, depicting the presence of NZVI. It is noteworthy that peaks centered at 710.8 and 723.6 eV indicate the presence of ferric oxides on the surface of NZVI (Li and Zhang, 2006, 2007). The oxide layer is thought to form instantaneously upon NZVI synthesis to passivate the highly reactive NZVI core, resulting in the preservation of the reducing power of NZVI.
a
b
Fig. 2 e X-Ray photoelectron spectra of (a) Ti2p and (b) Fe2p of Fe/TiO2 nanocomposites.
3.2. Dechlorination of TCE and DCP by Fe/TiO2 nanocomposites The dechlorination of chlorinated compounds by Fe/TiO2 nanocomposites in the presence and absence of UV light was described by pseudo-first-order rate equation: d½CCC ¼ kobs ½CCC dt
(1)
where kobs is the pseudo-first-order rate constant (h1), t is the reaction time, and [CCC] is the aqueous phase concentration of TCE or DCP (mM or mg/L). The kobs can be determined from the linear relationship of ln ([CCC]t/[CCC]0) versus time. The reactivity of Fe/TiO2 nanocomposites, P-25TiO2 particles and pure NZVI in the dark was evaluated by the dechlorination of 1.3 mg/L TCE or 10 mg/L of DCP. As depicted in Fig. 3a, less than 20% of the original TCE was dechlorinated by P-25 TiO2 after 145 h of the incubation, while a dechlorination efficiency of 66% was observed when NZVI was used for TCE dechlorination. The combination of TiO2 with NZVI exhibited good dechlorination ability, and 87% of the original TCE was dechlorinated by Fe/TiO2 within 145 h. The dechlorination of TCE followed the pseudo-first-order kinetics. As shown in Table 1, the kobs values for TCE dechlorination by NZVI and Fe/
4203
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 9 8 e4 2 1 0
Similar dechlorination patterns of DCP by Fe/TiO2 nanocomposites were observed. 54% and 77% of the original DCP were dechlorinated after the incubation of 145 h in the presence of NZVI and Fe/TiO2, respectively (Fig. 3b). The kobs for DCP dechlorination was (6.1 0.3)103 h1 for NZVI and (1.08 0.1)102 h1 for Fe/TiO2 nanocomposites (Table 1). The reason for the increased rate constants for TCE and DCP dechlorination by Fe/TiO2 compared to those with NZVI only may be attributed to the relatively homogeneous dispersion of NZVI in the nanocomposites, resulting in the increase in active reactive sites for dechlorination. This result also reflects that Fe/TiO2 is an effective nanocomposite for dechlorination of TCE and DCP. In addition, the dechlorination efficiency of TCE is higher than that of DCP, suggesting that TCE is a more readily degradable compound than DCP by Fe/TiO2.
a
b
3.3.
Fig. 3 e Dechlorination of (a) TCE and (b) DCP by NZVI, P-25 TiO2 and Fe/TiO2 nanocomposites in the dark under anoxic conditions.
TiO2 nanocomposites were (7.8 0.4) 103 and (1.39 0.05) 102 h1, respectively, clearly showing that the Fe/TiO2 nanocomposites had a higher reactivity towards TCE dechlorination than that of NZVI. In addition, a total of 92% of ethane was recovered and no less-chlorinated homologues such as dichloroethylene and vinyl chloride were found, indicating that hydrodechlorination is the major reaction pathway for TCE dechlorination by Fe/TiO2 nanocomposites.
Photodechlorination of TCE and DCP by UV light
The photodechlorination of organic pollutants in aqueous solutions using TiO2 nanoparticles is an interesting issue. In this study, the photocatalytic activity of Fe/TiO2 nanocomposites was evaluated by the photodechlorination of DCP and TCE. Fig. 4 shows the photodechlorination of DCP and TCE by Fe/TiO2 in the presence of 365 nm UV light. No obvious photodechlorination of DCP and TCE was observed after UV illumination of 100 min without the addition of P-25 TiO2 or Fe/TiO2 nanocomposites (direct photolysis). In the presence of Fe/TiO2 nanocomposites or P-25 TiO2, TCE also showed little photodechlorination within 100 min, while 75% and >99% of the original DCP was photodechlorinated by Fe/TiO2 and P-25 TiO2, respectively, indicating the good photocatalytic activity of P-25 TiO2 and Fe/TiO2 towards DCP dechlorination. The photodechlorination of DCP followed the pseudo-firstorder kinetics and the kobs for DCP photodechlorination by Fe/ TiO2 was 0.83 0.02 h1 (Table 1). This value is 77 times higher than that under dark conditions ((1.08 0.51) 102 h1), clearly showing the excellent photocatalytic activity of Fe/ TiO2 towards DCP dechlorination. It is noteworthy that the photodechlorination of DCP by P-25 TiO2 is more efficient than that by Fe/TiO2 and the kobs for DCP photodechlorination was 2.08 0.12 h1 (Table 1), which is 2.5 times higher than that by Fe/TiO2. The relatively low kobs for DCP photodechlorination by Fe/TiO2 is probably attributed to the masking effect of light penetration because the color of Fe/TiO2 nanocomposites is
Table 1 e The pseudo-first-order rate constants (kobs) for dechlorination of TCE and DCP alone by different nanomaterials in the presence and absence of UV lights under anoxic conditions. The nanomaterials used in this study were NZVI, P-25 TiO2 and Fe/TiO2 nanocomposites. Materials
Dark reaction TCE
Blank NZVI P-25 TiO2 Fe/TiO2
Photodechlorination DCP
TCE
DCP
kobs (h1)
r2
kobs (h1)
r2
kobs (h1)
r2
kobs (h1)
r2
nm 0.0078 0.0004 0.002 0.0001 0.0139 0.0005
nm 0.993 0.997 0.995
nm 0.006 0.0003 nm 0.0108 0.001
nm 0.996 nm 0.981
nm nm nm nm
nm nm nm nm
nm nm 2.08 0.12 0.83 0.02
nm nm 0.996 0.989
nm: No measureable degradation.
4204
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 9 8 e4 2 1 0
3.4. Possible reaction mechanism for TCE and DCP photodechlorination The reaction pathway for TCE dechlorination by NZVI has been well-investigated under anoxic conditions (He and Zhao, 2008; Parshetti and Doong, 2009). Reductive dechlorination and hydrodechlorination are the two major reaction mechanisms for dechlorination of chlorinated hydrocarbon. In this study, 90e94% of ethane, based on the stoichiometric equation for the dechlorination of TCE, was recovered during the dechlorination of TCE by Fe/TiO2 under dark conditions, clearly showing that hydrodechlorination is the major pathway for TCE dechlorination. The reduction pathways of DCP by NZVI were discussed by Cheng et al. (2007). They found that the mechanism included adsorption, dechlorination and cleavage of the benzene ring and several intermediates including 2-chlorophenol, 4-chloropheol, phenol, and chlorocatechol had been detected (Cheng et al., 2007; Liu et al., 2009; Hou et al., 2010). To elucidate the reaction pathways of DCP by Fe/TiO2, GC/MS was used to identify the intermediates and end products. As depicted in Fig. S3 and Table S1 (see supplementary materials), degradation intermediates including 2-chlorohydroquione, 4-chlorophenol and catechol were identified, while 1,4-benzoquine and phenol were detected as the end products during the course of DCP photodechlorination. Therefore, a possible reaction mechanism for DCP dechlorination by Fe/TiO2 in the presence of UV light is proposed. As shown in Fig. 5, DCP can react with photo-generated hydroxyl radicals to form 2-chlorohydroquinone and then to 1,4-benzoquinone. In addition, DCP can undergo reductive dechlorination to generate 4-chlorophenol first and then reacts with hydroxyl radicals, leading to the formation of catechol and phenol. Although DCP cannot be completely dechlorinated by Fe/TiO2,
Fig. 4 e Photodechlorination of TCE and DCP by Fe/TiO2 nanocomposites and P-25 TiO2 in the presence of 365 nm UV light under anoxic conditions.
black. In addition, the difference in the added amounts of TiO2 between P-25 TiO2 and Fe/TiO2 is also the plausible reason for the decreased rate constant. The concentration of nanomaterial used in this study was 3.5 g/L, and the EDS analysis showed that Fe/TiO2 contained 16.2 wt% Fe, which means that only 2.9 g/L TiO2 was used for photodechlorination. Several studies have depicted that photocatalytic degradation was a surface-mediated reaction and the degradation efficiency and rate increased with the increase in TiO2 amounts (Chong et al., 2009; Doong et al., 2009), which means the photodechlorination of DCP by Fe/TiO2 is lower than that by P-25 TiO2.
OH
O Cl
OH
OH
O
Cl
OH
OH
OH OH
Cl
Cl
Fig. 5 e The reaction pathway for DCP photodechlorination by Fe/TiO2 after illumination of 365 nm UV light under anoxic conditions.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 9 8 e4 2 1 0
the degradation intermediates as well as end products are less toxic than that of DCP and can be readily biodegraded into harmless compounds by microorganisms under both aerobic and anaerobic conditions (Shen et al., 2005).
3.5.
Simultaneous dechlorination of TCE and DCP
The feasibility of using Fe/TiO2 nanocomposites for simultaneous removal of TCE and DCP was further evaluated. Fig. 6 shows the simultaneous dechlorination of TCE and DCP by Fe/TiO2 in the absence and presence of 365 nm UV light. In the absence of UV light, the co-existence of DCP has little effect on the dechlorination efficiency of TCE, and 90% of the TCE was dechlorinated within 170 h. The kobs for TCE dechlorination was (1.3 0.06) 102 h1, which is in good agreement with the result of single TCE system ((1.39 0.05)102 h1). On the contrary, a significant inhibition of DCP dechlorination was observed when 10 mM TCE was present in the solution. Less than 25% of the original DCP was dechlorinated by Fe/TiO2 nanocomposites within the first 100 h. However, the dechlorination efficiency and rate of DCP was enhanced when the TCE concentration was lower than about 3 mM. It is known the redox reaction is a sequential reaction which the species with highest reduction potential in the solution dominates the reaction. In this study, TCE is a stronger electron acceptor
4205
than DCP. This means that TCE would react with most of the electrons produced from the anaerobic corrosion of NZVI, and subsequently inhibits the dechlorination efficiency and rate of DCP. Similar to the photodechlorination behavior of TCE only, no obvious TCE was dechlorinated by Fe/TiO2 in the presence
a
b
a
c b
Fig. 6 e Simultaneous dechlorination of TCE and DCP by Fe/ TiO2 nanocomposites (a) in the dark and (b) under UV illumination.
Fig. 7 e Simultaneous photodechlorination of TCE and DCP by Fe/TiO2 nanocomposites in the presence of (a) 20, (b) 50 and (c) 100 mM nickel ions under UV illumination. The concentrations of TCE and DCP were 10 mM and 10 mg/L, respectively.
4206
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 9 8 e4 2 1 0
of 10 mg/L DCP, indicating that TCE is not easily photodechlorinated within 2 h in aqueous solutions. This result is consistent with the finding of Ou and Lo (2007). However, 68% of the original DCP was dechlorinated when Fe/TiO2 was illuminated with UV light for 120 min in the presence of 10 mM TCE. The kobs for DCP photodechlorination was 0.588 0.05 h1, which was lower than that for DCP alone, presumably attributed to that TCE is a strong electron acceptor and could compete for electrons generated from the anaerobic corrosion of NZVI with DCP. This also means that the coexistence of other strong electron acceptors may influence the photodechlorination efficiency and rate of DCP by Fe/TiO2.
agreement with the results obtained in simultaneous removal of DCP and TCE under dark conditions, clearly showing that the presence of TCE inhibited the photodechlorination efficiency and rate of DCP under UV illuminated conditions. The simultaneous dechlorination of TCE and DCP in the presence of nickel ions under dark conditions was also investigated. The dechlorination efficiency of TCE and DCP increased with increasing Ni(II) concentrations, and the dechlorination efficiency and rate of TCE was higher than that of DCP in the presence of Ni(II) in the dark (Fig. S4, see supplementary materials). Table 2 shows the kobs values for simultaneous dechlorination of TCE and DCP by Fe/TiO2 in the presence and absence of nickel ions under dark and UV-illumination conditions. The kobs for TCE dechlorination by Fe/TiO2 in the dark increased from 0.013 0.001 h1 in the absence of Ni(II) to 0.389 0.003 h1 at 100 mM Ni(II). In addition, a 28.6-time increase in the kobs for DCP dechlorination was obtained when Ni(II) concentration increased from 20 to 100 mM. The synergistic effect of Ni(II) on the simultaneous photodechlorination of TCE and DCP in the dark is mainly attributed to the highly catalytic ability of Ni(II) in the hydrodechlorination reaction. The nickel ion can be electrochemically converted to zerovalent nickel by NZVI to form bimetallic Ni/Fe systems (Li and Zhang, 2006, 2007). In addition, the corrosion of NZVI with water resulted in the formation of Fe(II) ions and hydrogen gas. The hydrogen gas would then adsorb onto the metal surface to generate and store hydride-like species (Bransfield et al., 2006), subsequently enhancing the hydrodechlorination rate of TCE and DCP by bimetallic Ni/Fe nanoparticles. Several studies have depicted that the dechlorination rate of TCE by bimetallic Ni/Fe nanoparticles was 3.2e18.5 times higher than that by NZVI only (Tee et al., 2009). In this study, the kobs for TCE dechlorination by Fe/TiO2 at 100 mM Ni(II) increased by a factor of 29.9 in the dark, clearly showing that Fe/TiO2 nanocomposites is a promising material for dechlorination of chlorinated compounds in the presence of Ni(II) ions. Similar to the simultaneous dechlorination behaviors of TCE and DCP in the dark, the kobs for TCE photodechlorination under UV illumination increased from 0.90 0.12 h1 at 20 mM Ni(II) to 11.8 0.42 h1 at 100 mM Ni(II), clearly depicting that addition of Ni(II) has a synergistic effect on Fe/TiO2. In addition, the kobs for DCP photodechlorination by Fe/TiO2 increased from 0.59 0.05 h1 in the absence of Ni(II) to 2.36 0.18 h1 at 50 mM NI(II) and then slightly decreased to
3.6. Synergistic effect of Ni(II) on simultaneous dechlorination In this study, the Fe/TiO2 nanocomposites have been shown to have a good catalytic ability for rapid dechlorination of DCP under UV illumination. No obvious photodechlorination of TCE, however, was observed after UV illumination of 120 min. Several studies have shown that the deposition of small amounts of a second catalytic metal ion such as Ni, Cu and Pd onto zerovalent metal surface could enhance the dechlorination efficiency and rate of chlorinated hydrocarbons (Doong and Lai, 2006; He and Zhao, 2008; Lee and Doong, 2008; Tee et al., 2009). The simultaneous dechlorination of TCE and DCP by Fe/TiO2 as a function of Ni(II) concentration in the presence of UV light is shown in Fig. 7. The addition of Ni(II) has a considerable effect on the enhancement of photodechlorination efficiency and rate of TCE and DCP. The photodechlorination efficiency of TCE was 52% at 20 mM Ni(II) after UV illumination of 60 min, while a nearly complete photodechlorination of TCE was observed within 30 min when the Ni(II) concentration was higher than 50 mM. No obvious photodechlorination of TCE was observed when the solution only contained TiO2 and 100 mM Ni(II), clearly indicating that NZVI plays a crucial role in photocatalytic dechlorination of TCE. In addition, the photodechlorination efficiency of DCP increased with increasing Ni(II) concentrations between the range of 20 and 50 mM, and then slightly decreased when the Ni(II) concentration was increased to 100 mM. It is noteworthy that the photodechlorination efficiency of TCE was slightly lower than that of DCP at 20 mM Ni(II), while the rapid photodechlorination of TCE was observed when Ni(II) concentration was higher than 50 mM. This phenomenon is in good
Table 2 e The pseudo-first-order rate constants ( kobs) for simultaneous dechlorination of TCE and DCP by Fe/TiO2 nanocomposites in the presence of Ni(II) under anoxic conditions. Ni concentration (mM)
Dark reaction TCE
0 20 50 100
Photodechlorination DCP
TCE
DCP
kobs (h1)
r2
kobs (h1)
r2
kobs (h1)
r2
0.988 0.997 0.968 0.996
nm 0.005 0.0003 0.013 0.001 0.143 0.003
nm 0.992 0.988 0.987
nm 0.90 0.12 6.12 0.23 11.8 0.4
nm 0.998 0.983 0.999
0.013 0.014 0.045 0.389
nm: No measureable degradation.
0.001 0.001 0.006 0.003
kobs (h1)
r2
0.994 0.988 0.996 0.992
0.59 0.96 2.36 1.88
0.05 0.06 0.18 0.14
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 9 8 e4 2 1 0
1.88 0.14 h1 when further increased the Ni(II) concentration to 100 mM. In general, the kobs values for TCE and DCP photodechlorination in the presence of UV light are 30.4e136 and 13.4e192 times higher than those in the dark, respectively. The production of free radicals is the plausible reason for acceleration of photodechlorination efficiency and rate of TCE and DCP in the presence of Fe/TiO2 and Ni(II) ions under UV illumination. Fig. 8 shows the change in the EPR intensity of free radicals produced from the UV illumination of aqueous solution containing Fe/TiO2 and mixed DCP and TCE in the presence of DMPO and various concentrations of nickel ions. Addition of DMPO produced six-line EPR spectra after 5 min of UV illumination, suggesting the formation of O-centered radical adducts such as OH, OOH, and ROO (Brezova et al., 2007; Ghorai et al., 2008). The EPR signal of the radical spin adducts was very weak when no nickel ion was added to the photoreaction of chlorinated compounds by Fe/TiO2 nanocomposites. However, the signal intensity of free radicals increased obviously upon increasing aqueous nickel ion concentrations from 20 to 100 mM. In addition, the EPR intensity could be maintained from 10 min at 20 mM Ni(II) to 20 min at 100 mM Ni(II), clearly indicating the enhanced effect of nickel ions on the production of radicals. Several studies have depicted that Ni(II) ions could react with photo-generated holes to form Ni(III) ions under anoxic conditions, and the formed Ni(III) ions converted back to Ni(II) ions again when reacted with electrons (Devi et al., 2010; Parshetti and Doong, 2010). The EPR spectrum also showed the photogeneration of Ni(III) ion after UV illumination of 10 min (Figure S5, see supplementary materials). Therefore, the photo-generated electron-hole pairs can thus be separated through Ni ions cycling, leading to the improvement of electron transfer efficiency, and thereby accelerate the formation of hydroxyl radicals to enhance the photocatalytic activity of TCE and DCP by Fe/TiO2. In this study, we have demonstrated the first report on the synergistic effect of nickel ion on the coupled photodechlorination of TCE and DCP by Fe/TiO2 nanocomposites in the presence of UV light under anoxic conditions. Chlorinated hydrocarbons as well as aromatic compounds are frequently found in the aqueous environments, and NZVI and TiO2 are the most often used materials for removal of priority pollutants. NZVI is typically a powerful reductant to reduce organic compounds, while TiO2 has strong oxidation ability towards the photodegradation of organic pollutants under UV illumination. Therefore, the combination of NZVI with TiO2 makes this nanocomposite an ideal platform to decompose a wide variety of pollutants in the impaired water. In this study, Fe/ TiO2 nanocomposites have been found to effectively dechlorinate TCE and DCP in the dark. The addition of nickel ions in the concentration range 50e100 mM significantly enhanced the efficiency and rate of simultaneous dechlorination of TCE and DCP, and the kobs for TCE and DCP dechlorination was enhanced by factors of 29.9 and 28.6, respectively, when compared with that in the absence of nickel ions, clearly indicating the synergistic effect on nickel ion on dechlorination of chlorinated compounds under anoxic conditions in the dark. The presence of nickel ions also has a synergistic effect on the photodechlorination of TCE and DCP by Fe/TiO2 under UV
4207
illumination. The reaction mechanism for enhanced photodechlorination of chlorinated compounds by Fe/TiO2 is proposed in Scheme 1. The TiO2 photocatalysts can be photoexcited by UV light to generate electron-hole pairs, while the
Fig. 8 e The EPR signals produced from the photodegradation of TCE and DCP by the illuminated Fe/ TiO2 nanocomposites in the presence of (a) 20, (b) 50, and (c) 100 mM nickel ions and DMPO under anoxic conditions.
4208
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 9 8 e4 2 1 0
Scheme 1 e The enhanced photodechlorination efficiency of DCP and TCE by Fe/TiO2 nanocomposites in the presence of Ni(II) ions and 365 nm UV light under anoxic conditions.
metallization of TiO2 with Fe prevents the recombination of holes with electrons, leading to the enhancement of the oxidizing capability of TiO2 (Hsieh et al., 2010). In addition, the Ni(II) and produced Fe(II) ions from the anaerobic corrosion of NZVI can react with photo-generated holes to form Ni(III)/ Fe(III) ions and then be converted back to Ni(II)/Fe(II) ions again when reacted with electrons or hydroxyl anions, resulting in the prevention of hole-electron recombination and the increase in the total amounts of hydroxyl radicals (Ghorai et al., 2008; Devi et al., 2010; Parshetti and Doong, 2010). TiO2 þ hv/ e þ hþ þ
H2 O þ h /, OH 2þ
þ
3þ
(2)
4.
(3)
In this study, TCE and DCP were simultaneously dechlorinated by Fe/TiO2 nanocomposites under anoxic conditions in the presence of nickel ions and UV light. Both TCE and DCP were effectively dechlorinated by Fe/TiO2 nanocomposites, and the dechlorination efficiency and rate of TCE and DCP were higher than by NZVI alone. Although TCE could not be photodegraded after UV illumination of 2 h, the rate constant for DCP photodechlorination was enhanced by a factor of 50. Hydrodechlorination and hydroxyl radical chain reaction were found to be the major pathways for TCE and DCP dechlorination. The co-existence of TCE inhibits the dechlorination efficiency and rate of DCP during simultaneous photodechlorination processes. In addition, nickel ions have a significant effect on enhancing the simultaneous photodechlorination efficiency of TCE and DCP under the illumination of UV light. The kobs for DCP and TCE photodechlorination by Fe/TiO2 at 20e100 mM Ni(II) enhanced 30.4e136 and 13.2e192 times, respectively, when compared with those in the dark, presumably due primarily to the separation of the photo-generated electron-hole pairs through Ni ions cycling. Results obtained in this study clearly show that the Fe/TiO2 nanocomposite is an ideal platform to accelerate the simultaneous photodechlorination rates of chlorinated compounds, and would be helpful in facilitating the development of processes that could be useful for the enhanced degradation of co-contaminants in the aquatic environment.
Ni =Fe2þ þ h / Ni =Fe3þ
(4)
Ni =Fe2þ þ e / Ni =Fe2þ
(5)
Ni =Fe3þ þ OH þ hv/ Ni =Fe2þ þ , OH
(6)
3þ
3þ
2þ
2þ
C2 HCl3 ðTCEÞ þ 5Hþ þ 8 e / C2 H6 ðethaneÞ þ 3Cl
C6 H4 Cl2 O ðDCPÞ þ e /C6 H4 ClOð4chlorophenolÞþ Cl
(7)
by Fe/TiO2 in the presence and absence of UV light. NZVI has been proven to effectively adsorb nickel ions first and then partially convert to the zerovalent forms of nickel via the electron transfer from Fe. This process would significantly accelerate the simultaneous photodechlorination rates of chlorinated compounds, and would be helpful in facilitating the development of processes that could be useful for the enhanced degradation of co-contaminants in the aquatic environment.
(8)
C6 H4 Cl2 O ðDCPÞ þ , OH / C6 H5 ClO2 ð2 chlorohydroquinoneÞ (9) Therefore, the photo-generated electron-hole pairs can thus be separated effectively through Fe and Ni ion cycling, leading to the improvement of electron transfer efficiency and rapid dechlorination rate of TCE and DCP simultaneously. In addition, a low concentration of Ni(II) at 50 mM has been proven to be effective in enhancing the efficiency and rate of TCE and DCP dechlorination. This gives great impetus to coupled removal of heavy metals and chlorinated compounds
Conclusions
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 9 8 e4 2 1 0
Acknowledgments The authors thank the National Science Council, Taiwan (NSC98-2221-E-007-030-MY3) and National Tsing Hua University (99N2452E1) for financial support.
Appendix. Supplementary material Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.watres.2011.05.019.
references
Bransfield, S.J., Cwiertny, D.M., Roberts, A.L., Fairbrother, D.H., 2006. Influence of copper loading and surface coverage on the reactivity of granular iron toward 1,1,1-trichloroethane. Environmental Science Technology 40 (5), 1485e1490. Brezova, V., Dvoranova, D., Stasko, A., 2007. Characterization of titanium dioxide photoactivity following the formation of radicals by EPR spectroscopy. Research on Chemical Intermediates 33 (3e5), 251e268. Chang, S.M., Doong, R.A., 2006. Characterization of Zr-doped TiO2 nanocrystals prepared by a nonhydrolytic sol-gel method at high temperatures. Journal of Physical Chemistry B 110 (42), 20808e20814. Chang, S.M., Hou, C.Y., Lo, P.H., Chang, C.T., 2009. Preparation of phosphated Zr-doped TiO2 exhibiting high photocatalytic activity through calcination of ligand-capped nanocrystals. Applied Catalysis B: Environmental 90 (1e2), 233e241. Cheng, R., Wang, J.L., Zhang, W.X., 2007. Reductive dechlorination of 2,4-dichlorophenol using nanoscale Fe-0: influencing factors and possible mechanism. Science in China Series B: Chemistry 50 (4), 574e579. Choi, H., Al-Abed, S.R., Agarwal, S., Dionysiou, D.D., 2008. Synthesis of reactive nano-Fe/Pd bimetallic systemimpregnated activated carbon for the simultaneous adsorption and dechlorination of PCBs. Chemistry of Materials 20 (11), 3649e3655. Chong, M.N., Jin, B., Zhu, H.Y., Chow, C.W.K., Saint, C., 2009. Application of H-titanate nanofibers for degradation of Congo Red in an annular slurry photoreactor. Chemical Engineering Journal 150 (1), 49e54. Chu, W., Choy, W.K., 2000. The study of lag phase and rate improvement of TCE decay in UV/surfactant systems. Chemosphere 41 (8), 1199e1204. Darab, J.G., Amonette, A.B., Burke, D.S.D., Orr, R.D., Ponder, S.M., Schrick, B., Mallouk, T.E., Lukens, W.W., Caulder, D.L., Shuh, D.K., 2007. Removal of pertechnetate from simulated nuclear waste streams using supported zerovalent iron. Chemistry of Materials 19 (23), 5703e5713. Devi, L.G., Kottam, N., Murthy, B.N., Kumar, S.G., 2010. Enhanced photocatalytic activity of transition metal ions Mn2þ, Ni2þ and Zn2þ doped polycrystalline titania for the degradation of Aniline Blue under UV/solar light. Journal of Molecular Catalysis A: Chemical 328 (1e2), 44e52. Doong, R.A., Chang, P.Y., Huang, C.H., 2009. Microstructural and photocatalytic properties of solegel-derived vanadium-doped mesoporous titanium dioxide nanoparticles. Journal of NonCrystalline Solids 355 (45e47), 2302e2308. Doong, R.A., Hsieh, T.C., Huang, C.P., 2010. Photoassisted reduction of metal ions and organic dye by titanium dioxide nanoparticles in aqueous solution under anoxic conditions. Science of the Total Environment 408 (16), 3334e3341.
4209
Doong, R.A., Lai, Y.J., 2005. Dechlorination of tetrachloroethylene by palladized iron in the presence of humic acid. Water Research 39 (11), 2309e2318. Doong, R.A., Lai, Y.L., 2006. Effect of metal ions and humic acid on the dechlorination of tetrachloroethylene by zerovalent iron. Chemosphere 64 (3), 371e378. Dries, J., Bastiaens, L., Springael, D., Agathos, S.N., Diels, L., 2005. Combined removal of chlorinated ethenes and heavy metals by zerovalent iron in batch and continuous flow column systems. Environmental Science Technology 39 (21), 8460e8465. Ghorai, T.K., Dhak, D., Dalai, S., Pramanik, P., 2008. Preparation and photocatalytic activity of nano-sized nickel molybdate (NiMoO4) doped bismuth titanate (Bi2Ti4O11) (NMBT) composite. Journal of Alloys and Compounds 463 (1e2), 390e397. Han, X.G., Kuang, Q., Jin, M.S., Xie, Z.X., Zheng, L.S., 2009. Synthesis of titania nanosheets with a high percentage of exposed (001) facets and related photocatalytic properties. Journal of the American Chemical Society 131 (9), 3152e3153. He, F., Zhao, D.Y., 2008. Hydrodechlorination of trichloroethene using stabilized Fe-Pd nanoparticles: reaction mechanism and effects of stabilizers, catalysts and reaction conditions. Applied Catalysis B: Environmental 84 (3e4), 533e540. Hou, C.F., Ge, X.P., Zhou, Y.M., Li, Y., Wang, D.S., Tang, H.X., 2010. Characterization of nanoscale iron and its degradation of 2,4-dichlorophenol. Chinese Science Bulletin 55 (4e5), 350e357. Hsieh, W.P., Pan, J.R., Huang, C.P., Su, Y.C., Juang, Y.J., 2010. Enhance the photocatalytic activity for the degradation of organic contaminants in water by incorporating TiO2 with zero-valent iron. Science of the Total Environment 408 (3), 672e679. Huang, C.P., Hsieh, W.P., Pan, J.R.S., Chang, S.M., 2007. Characteristic of an innovative TiO2/Fe(0) composite for treatment of azo dye. Separation and Purification Technology 58 (1), 152e158. Lee, C.C., Doong, R.A., 2008. Dechlorination of tetrachloroethylene in aqueous solutions using metal-modified zerovalent silicon. Environmental Science Technology 42 (13), 4752e4757. Lee, C.C., Doong, R.A., 2011. Enhanced dechlorination of tetrachloroethylene by zerovalent silicon in the presence of polyethylene glycol under anoxic conditions. Environmental Science Technology 45, 2301e2307. Li, X.Q., Zhang, W.X., 2006. Iron nanoparticles: the core-shell structure and unique properties for Ni(II) sequestration. Langmuir 22 (10), 4638e4642. Li, X.Q., Zhang, W.X., 2007. Sequestration of metal cations with zerovalent iron nanoparticles - A study with high resolution X-ray photoelectron spectroscopy (HR-XPS). Journal of Physical Chemistry C 111 (19), 6939e6946. Liu, L.F., Chen, F., Yang, F.L., 2009. Stable photocatalytic activity of immobilized Fe/TiO2/ACF on composite membrane in degradation of 2,4-dichlorophenol. Separation and Purification Technology 70 (2), 173e178. Maithreepala, R.A., Doong, R.A., 2004. Synergistic effect of copper ion on the reductive dechlorination of carbon tetrachloride by surface-bound Fe(II) associated with goethite. Environmental Science Technology 38 (1), 260e268. Ou, H.H., Lo, S.L., 2007. Photocatalysis of gaseous trichloroethylene (TCE) over TiO2: the effect of oxygen and relative humidity on the generation of dichloroacetyl chloride (DCAC) and phosgene. Journal of Hazardous Materials 146 (1e2), 302e308. Parshetti, G.K., Doong, R.A., 2009. Dechlorination of trichloroethylene by Ni/Fe nanoparticles immobilized in PEG/ PVDF and PEG/nylon 66 membranes. Water Research 43 (12), 3086e3094.
4210
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 1 9 8 e4 2 1 0
Parshetti, G.K., Doong, R.A., 2010. Dechlorination and photodegradation of trichloroethylene by Fe/TiO2 nanocomposites in the presence of nickel ions under anoxic conditions. Applied Catalysis B: Environmental 100 (1e2), 116e123. Riley, R.G., Zachara, J.M., Wobber, F.J., 1992. Chemical Contaminants on DOE Lands and Selection of Contaminant Mixtures for Subsurface Science Research. DOE/ER-0547T, U. S.. Department of Energy, Washington, D. C. Schrick, B., Hydutsky, B.W., Blough, J.L., Mallouk, T.E., 2004. Delivery vehicles for zerovalent metal nanoparticles in soil and groundwater. Chemistry of Materials 16 (11), 2187e2193. Shen, D.S., Liu, X.W., Feng, H.J., 2005. Effect of easily degradable substrate on anaerobic degradation of pentachlorophenol in an up-flow anaerobic sludge blanket (UASB) reactor. Journal of Hazardous Materials 119 (1e3), 239e243. Sloto, R.A., 2010. Changes in Groundwater Flow and Volatile Organic Compound Concentrations at the Fischer and Porter Superfund Site, Warminster Township, Bucks County, Pennsylvania, 1993e2009. Scientific Investigations Report 2010e5054, U.S. Geological Survey, Reston, Virginia. Sunkara, B., Zhan, J., He, J.B., McPherson, G.L., Piringer, G., John, V. T., 2010. Nanoscale zerovalent iron supported on uniform carbon microspheres for the in situ remediation of chlorinated hydrocarbons. ACS Applied Materials Interfaces 2 (10), 2854e2862.
Tee, Y.H., Bachas, L., Bhattacharyya, D., 2009. Degradation of trichloroethylene and dichlorobiphenyls by iron-based bimetallic nanoparticles. Journal of Physical Chemistry C 113 (22), 9454e9464. Xu, J., Bhattacharyya, D., 2005. Membrane-based bimetallic nanoparticles for environmental remediation: synthesis and reactive properties. Environmental Progress 24 (4), 358e366. Yu, J.G., Xiang, Q.J., Zhou, M.H., 2009a. Preparation, characterization and visible-light-driven photocatalytic activity of Fe-doped titania nanorods and first-principles study for electronic structures. Applied Catalysis B: Environmental 90 (3e4), 595e602. Yu, J.G., Yue, L., Liu, S.W., Huang, B.B., Zhang, X.Y., 2009b. Hydrothermal preparation and photocatalytic activity of mesoporous Au-TiO2 nanocomposite microspheres. Journal of Colloid and Interface Science 334 (1), 58e64. Zheng, T.H., Zhan, J.J., He, J.B., Day, C., Lu, Y.F., Mcpherson, G.L., Piringer, G., John, V.T., 2008. Reactivity characteristics of nanoscale zerovalent iron-silica composites for trichloroethylene remediation. Environmental Science Technology 42 (12), 4494e4499. Zhu, B.W., Lim, T.T., 2007. Catalytic reduction of Chlorobenzenes with Pd/Fe nanoparticles: reactive sites, catalyst stability, particle aging, and regeneration. Environmental Science Technology 41 (21), 7523e7529.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 1 1 e4 2 2 6
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Soil aquifer treatment of artificial wastewater under saturated conditions H.M.K. Essandoh a, C. Tizaoui b, M.H.A. Mohamed a,*, G. Amy c,d, D. Brdjanovic c a
School of Engineering Design and Technology, University of Bradford, BD7 1DP, UK College of Engineering, Swansea University, SA2 8PP, UK c UNESCO-IHE, Westvest 7, 2611 AX Delft, The Netherlands d Water Desalination and Reuse Center, King Abdullah University of Science and Technology, Thuwal 23955-6900, Saudi Arabia b
article info
abstract
Article history:
A 2000 mm long saturated laboratory soil column was used to simulate soil aquifer
Received 23 November 2010
treatment under saturated conditions to assess the removal of chemical and biochemical
Received in revised form
oxygen demand (COD and BOD), dissolved organic carbon (DOC), nitrogen and phos-
18 May 2011
phate, using high strength artificial wastewater. The removal rates were determined
Accepted 23 May 2011
under a combination of constant hydraulic loading rates (HLR) and variable COD
Available online 31 May 2011
concentrations as well as variable HLR under a constant COD. Within the range of COD concentrations considered (42 mg L 1e135 mg L1) it was found that at fixed hydraulic
Keywords:
loading rate, a decrease in the influent concentrations of dissolved organic carbon (DOC),
Chemical oxygen demand
biochemical oxygen demand (BOD), total nitrogen and phosphate improved their
Hydraulic loading rate
removal efficiencies. At the high COD concentrations applied residence times influenced
Mass loading rate
the redox conditions in the soil column. Long residence times were detrimental to the
Removal efficiency
removal process for COD, BOD and DOC as anoxic processes and sulphate reduction
Soil aquifer treatment
played an important role as electron acceptors. It was found that total COD mass loading within the range of 911 mg d1e1780 mg d1 applied as low COD wastewater infiltrated coupled with short residence times would provide better effluent quality than the same mass applied as a COD with higher concentration at long residence times. The opposite was true for organic nitrogen where relatively high concentrations coupled with long residence time gave better removal efficiency. Crown Copyright ª 2011 Published by Elsevier Ltd. All rights reserved.
1.
Introduction
The importance of adequate sanitation as well as the availability of enough quantities of fresh water for human consumption and industrial and agricultural use cannot be underestimated as they play a vital role in maintaining a healthy livelihood and in the development of nations. As populations continue to increase with their associated problems of waste generation and increased contamination
of surface and ground waters, pressure on available water resources is increasing. This, coupled with uneven distribution of water resources and periodic droughts around the world, has brought about the need for innovative sources of water supply and local conservation. Highly treated wastewater effluents from municipal wastewater treatment plants are therefore now increasingly being considered as a reliable source of water supply (Metcalf and Eddy et al., 2003).
* Corresponding author. Tel.: þ44 (0)1274 233856; fax: þ44 (0)1274 234111. E-mail address:
[email protected] (M.H.A. Mohamed). 0043-1354/$ e see front matter Crown Copyright ª 2011 Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.05.017
4212
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 1 1 e4 2 2 6
Soil aquifer treatment (SAT) has been found to be a lowcost sustainable tertiary wastewater treatment technology, which has the ability to generate high quality effluent from secondary treated wastewater for potable and non-potable uses (Cha et al., 2006; Fox et al., 2006). During SAT, the saturated and unsaturated zones of the natural soil and groundwater aquifer act as the medium in which physicochemical and biological reactions occur (Cha et al., 2006). These reactions substantially reduce the levels of organic and inorganic compounds including nitrogen, phosphorus, suspended solids, pathogens and heavy metals leading to an improvement in water quality (Pescod, 1992; Bdour et al., 2009). Mixing of the infiltrated wastewater with the groundwater and the slow movement through the aquifer increases the contact time with the aquifer material leading to further purification of the water (Asano and Cotruvo, 2004; Dillon et al., 2006). Besides treatment, SAT offers the opportunity of aquifer recharge (Droste, 1997) thus seasonal or long-term storage of water can be achieved (Fox et al., 2006), which is especially beneficial in arid areas. The unsaturated zone is characterized by availability of oxygen as well as increased ability of flow of air during the drying period of the SAT treatment cycle. Existence of oxygen in the unsaturated zone is highly important in promoting aerobic biodegradation processes and nitrification. Factors influencing the efficiency of SAT include characteristics of treatment site, soil and wastewater characteristics, climate and infiltration rate (TanIk and C ¸ omakoglu, 1996). Redox conditions and residence time can have a significant influence on the kinetics of dissolved organic carbon (DOC) degradation (Gru¨nheid et al., 2005) and may affect the removal efficiency. In the saturated zone, where a greater portion of the residence time occurs (Fox and Makam, 2009), dissolved oxygen is limited and the level of contaminants in the infiltrating wastewater and associated oxygen demand may have a major impact on the efficiency of the removal process. Although a large number of SAT systems exist, most of them involve well-treated effluents of low organic content and utilize the vadose zone. So far, limited work has been done to demonstrate the applicability and practicality of using SAT in treating poorly treated effluents or even primary effluents. In addition, complete reliance on the saturated zone without utilisation of the vadose zone in the treatment has never been explored. Earlier studies carried out found a correlation between the organic and hydraulic loading rates and effluent quality (Nema et al., 2001). It was observed that effluent biochemical oxygen demand (BOD), chemical oxygen demand (COD), suspended solids (SS), total kjeldahl nitrogen (TKN), ammonia nitrogen (NH3) and phosphorus concentrations increased linearly with an increase in cumulative mass loading. Effluent quality with respect to these parameters was also found to deteriorate linearly with increase in cumulative hydraulic loading (Nema et al., 2001). These results were however found to be contradictory to previous studies indicating that the hydraulic loading was a more important operating parameter than the organic loading in determining the effluent quality (Carlson et al., 1982). These studies involved an unsaturated zone for the treatment and in recent years no further studies have been carried out. This paper presents results of the experimental investigation using a 2 m
long soil column with a particular focus on the removal of BOD, DOC, COD, nitrogen and phosphate in the saturated zone. In addition, the influence of hydraulic loading rates on the treatment efficiency of wastewater of much higher COD than normally encountered in secondary or tertiary effluents applied in SAT systems was studied.
2.
Materials and methods
2.1.
Soil column description and setup
Fig. 1 shows a schematic drawing of the soil column setup. The column used for this study was made of acrylic tube with an inner diameter of 140 mm and length 2000 mm. Flanges were fitted to the top and bottom of the column for attachment of the top and bottom end caps. Two holes were provided in the top cap. One served as the column exit and the other was fitted with a valve for bleeding air out of the column. A 10 mm thick PVC distributor cut out in the form of a labyrinth was mounted on the inner surface of bottom cap to facilitate even distribution of water over the entire cross section of the column. To maintain a watertight seal, a gasket was placed between the tube and the end caps before securing them together. Water sampling points consisting of 3.2 mm inner diameter stainless steel tubes were provided at ten points (100 mm, 200 mm, 300 mm, 400 mm, 500 mm, 600 mm, 800 mm, 1100 mm, 1400 mm, and 1700 mm) from the bottom of the column as shown in Fig. 1. These sampling tubes extended to the centre of the column’s cross section. The sampling ports were closed by means of flexible tubing and a clip. CONMARK 314 stainless steel digital pocket thermometers were inserted at 170 mm and 1830 mm from the bottom of the column to monitor the column temperature. The column was mounted in a steel frame and a funnel and tubing arrangement provided on the frame at the same height as the top of the column for discharge of the column effluent. The column was packed to a density of 1.55 g cm3 under saturated conditions with uniform silica sand of effective diameter of 0.51 mm and average diameter of 0.75 mm, obtained from WBB Minerals. The uniformity coefficient of the sand and porosity of the packing were found to be 1.6 and 0.41 respectively. Water saturated condition in the column was achieved by ensuring that the water level in the column was always above the surface of the sand during packing. After filling, the column was wrapped with aluminium foil to shut out light and discourage the growth of algae during operation. A variable speed peristaltic pump was used to deliver wastewater to the column through soft tygon tubing and a flow meter, which was fully opened and used for monitoring of flow to the column. An injection port consisting of a T-shaped glass tube with a septum stopper was provided in the influent line at the column entry for injection of a tracer during residence time studies.
2.2.
Column start-up and general operation
The sand column was set up in a controlled temperature room set at 20 C 0.5 C. Synthetic wastewater was prepared by dilution in tap water of a stock solution containing 9.6 g
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 1 1 e4 2 2 6
4213
Fig. 1 e Soil column setup.
peptone from meat, 6.6 g meat extract, 12 g glucose, 0.42 g sodium chloride, 3.42 g ammonium chloride, 0.24 g calcium chloride, 0.12 g magnesium sulphate and 1.68 g potassium dihydrogen phosphate in 1 L of deionised water. This preparation is a modified version (Prochaska et al., 2007) of the
OECD standard sewage (OECD, 1996) which is often used as a recipe for synthetic sewage. Peptone, which is commonly used as a culture medium for microorganisms, and meat extract were added to glucose to obtain a rich source of carbon and nutrients for microbial growth. All chemicals used were of
4214
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 1 1 e4 2 2 6
analytical grade and were obtained from VWR International Ltd., Poole, UK. The prepared stock solution was kept in the fridge at 4 C and dilutions were prepared daily and aerated continuously with laboratory air for application to the column. Aeration was required to maintain high levels of dissolved oxygen (DO) in the influent wastewater. After three to four days the solution was discarded and a fresh one was prepared. As the wastewater to be passed through the column contained no microorganisms that are usually encountered in wastewater treatment plants, it was found essential to initially seed the influent wastewater with microorganisms. A non-pathogenic microbial seeding solution prepared from capsules containing a blend of microorganisms normally found in wastewater and used for biochemical oxygen demand (BOD) determination in wastewater analysis was used for this purpose. These capsules were obtained from ColeeParmer, UK. The seed solution was introduced regularly into the influent line via the injection port (see Fig. 1) at a volume of 5 mL for every 500 mL of synthetic wastewater diluted to obtain a chemical oxygen demand (COD) of 135 mg L1. Wastewater was supplied to the column at a rate of 10 mL min1 during seeding. Seeding was stopped after about twice the volume of voids of wastewater had been passed through the column and only the synthetic wastewater fed to the column afterwards at a rate of 5 mL min1. The pore volume of the column packing, determined as the product of the volume of sand in the column and the porosity, was 12.4 L. The soil column was fed from bottom to top throughout the experiments to maintain fully saturated conditions. Three months were allowed for growth of microorganisms in the soil and establishment of steady state conditions as evidenced by near constant concentrations of wastewater parameters in column wastewater samples. Routine cleaning and sterilizing of the tubing and flow meter was carried out twice a week with chlorine-based sterilizing tablets. It was found that less frequent cleaning led to an appreciable reduction in the dissolved oxygen available in the influent, especially in the case where the high concentration wastewater was fed to the column at a slow flow rate. Influent samples were collected at the entry to the soil column thus eliminating any degradation effects that may occur in the supply tubing and flow meter.
2.3.
SAT simulations
The column experiments were run at three hydraulic loading rates (HLR) and three COD concentrations. The wastewaters have been classified according to strength as high concentration (HC), medium concentration (MC) and low concentration (LC). Table 1 gives the average characteristics of the wastewater applied. Of note the pH of the influent was measured and found to be about 7.2 in all experiments. The experimental conditions and sample numbers are summarised in Table 2. Residence time distribution (RTD) tests were carried out at each hydraulic loading rate using 25 g L1 fluorescein sodium salt purchased from SigmaeAldrich, UK as a tracer. The tests were carried out in the clean column sand before the seeding process and wastewater infiltration. A volume of 2 mL of this solution was rapidly injected through the injection port and
Table 1 e Influent Characteristics. Parameter
Chemical oxygen demand (COD) Biochemical oxygen demand (BOD) Dissolved organic carbon (DOC) Total kjeldahl nitrogen (TKN) Ammonia nitrogen (NH3-N) Organic nitrogen (Org-N) Nitrate nitrogen (NO3-N) Total nitrogen (TN) Phosphate (PO4) Sulphate (SO4)
Average concentration (mg/L) HC
MC
LC
135 88 50 15 4 11 2.5 17.5 6.7 50
61 43 26.6 6.7 1.7 5 2.3 9 4.4 56
42 26 16 4.1 1.4 2.7 2.4 6.5 4 50
the tracer concentration at the effluent end was measured by a UV/Visible spectrophotometer set to a wavelength of 489 nm. This wavelength was chosen based on the result of a preliminary spectra test, which showed that maximum absorbance for fluorescein occurred at this wavelength. The spectrophotometer was connected to a computer through a data logger set at a sampling rate of 1 sample/min for rapid recording of the measurements. The suitability of fluorescein as a tracer for the tests was verified by comparing the mass of tracer injected to the mass recovered by integration of the concentrations over the test duration. Tracer recovery was between 96% and 98% for the three RTD studies carried out. For each SAT experiment wastewater was sampled from water sampling ports provided at depths of 100 mm, 600 mm, 1100 mm and 1700 mm along the column (see, Fig. 1). Samples were not taken from all the ports provided as it was realized after initial analysis of test samples that the column wastewater quality did not vary much after the 100 mm sampling point. The influent and final effluent from the column was also sampled. All collected samples were analyzed for BOD, COD, DOC, TKN, total nitrogen, organic nitrogen, ammonia nitrogen, nitrate, nitrite, phosphate and sulphate according to established methods. Total nitrogen was determined by calculation as the sum of the TKN, nitrate and nitrite. Organic nitrogen was obtained by subtraction of ammonia nitrogen of the sample from the TKN. Edge effects in the soil column were minimal as the column diameter was greater than 30 times the average grain diameter (Relyea, 1982). Besides, soil column sampling points extended to the centre of the soil column cross section. A water bath was used to raise the temperature of the influent wastewater to 20 C when wastewater was being fed to the column at a hydraulic loading rate of 169 cm d1. This was achieved by immersing a short portion of the influent tubing in the bath.
2.4.
Analytical methods
Dissolved organic carbon (DOC) was determined on filtered samples by low temperature oxidation using an ISCO Total organic carbon (TOC) analyzer. COD analysis was carried out by the addition of 2 mL of sample to COD vials from Hach
4215
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 1 1 e4 2 2 6
Table 2 e Experimental conditions. Experiment
HC-5 HC-10 HC-20 MC-20 LC-20
Wastewater concentration
Hydraulic loading rate (cm d1)
Hydraulic residence time (h)
Average COD mass loading rate (mg d1)
Number of samples
Standard error
High concentrationa High concentrationa High concentrationa Medium concentrationb Low concentrationc
44 88 169 169 169
51.8 23.8 12.2 12.2 12.2
911 1781 3553 1586 1084
9 6 6 3 3
30.8 42.5 18.3 31.8 31.2
a HC concentrations from Table 1. b MC concentrations from Table 1. c LC concentrations from Table 1.
Lange Limited. The vials were heated at 150 C for 2 h in a Hach COD reactor, allowed to cool and the COD determined colorimetrically on a DR/2400 spectrophotometer. BOD was measured according to the Standard Methods for the Examination of Water and Wastewater (Clesceri et al., 1998). Total kjeldahl nitrogen (TKN) was carried out using the digestion method developed by FOSS (2009) with some modification to the distillation step, which was carried out using indicating boric acid (Clesceri et al., 1998). The procedure was first tested on a peptone solution prepared to a known concentration. The results obtained from the digestion and distillation agreed closely with the TKN content of the prepared solution as obtained from calculations based on manufacturer’s information. The digestion was carried out in a Tecator 2006 digestor. Digested samples were diluted with deionised water and distilled in a Tecator Kjeltec System 1002 distilling unit after the addition of sufficient 40% sodium hydroxide solution (FOSS, 2009) to raise the pH to 11. Distillate was collected into 50 mL of indicating boric acid and back titrated with 0.25 N sulphuric acid (Clesceri et al., 1998). Ammonia-N analysis involved distilling undigested samples in the same manner. Nitrate, phosphate and sulphate were determined by ion chromatography on a Dionex series 4000i Ion chromatograph with an eluent made up of 1.8 mM sodium carbonate and 1.7 mM sodium bicarbonate, and a 0.025 N H2SO4 regenerant. Anion separation was carried out using an IonPac AG4A-SC 4-mm analytical column. An IonPac AS4A-SC 4-mm guard column was used to protect the analytical column from contamination. Samples were injected into the instrument through 0.2 mm syringe filters. Dissolved oxygen was measured with a Hach Sension 6 dissolved oxygen meter and pH by an Oakton pH/mV/ C pH 11 series meter.
3.
Results and discussions
This section presents findings from the three experiments performed at constant hydraulic loading rate and varying wastewater COD concentration as well as at three different hydraulic loading rates at fixed COD of much higher concentration than that encountered in secondary effluents. These saturated zone simulations may be likened to wastewater flow through shallow sandy aquifers where there is high likelihood of hydraulic continuity of the ponded wastewater
in the SAT infiltration basin with the saturated zone of the aquifer. During the simulations soil column wastewater sampling was carried out approximately every three days. About two weeks was allowed after each change in experimental condition to allow the microorganisms to get acclimated to the system. For each experiment sampling was carried out at least on three separate occasions to confirm the results obtained.
3.1.
Mass loadings on the soil column
The mass loadings on the soil column under the different conditions simulated are shown in Fig. 2aec. Error bars in all figures are based on the standard error. These loadings were determined for each experiment by taking the product of the influent concentration, the HLR and the cross sectional area of the soil in the column. The mass of the parameters existing at the sampling points along the soil column was also determined in a similar fashion. The COD, DOC and BOD mass loadings to the soil column increased with an increase in HLR with HC-20 having the highest loading. Mass loadings for HC-5 were close to LC-20 and that for HC-10 comparable to MC-20. The differences were less than 20% in the former and 10% for the latter. The same ratios of dissolved oxygen (DO) to COD, BOD and DOC were maintained when the HLR was changed and were 0.06, 0.09 and 0.16 respectively. MC-20 and LC-20 had ratios of 0.16 and 0.26 for COD, 0.22 and 0.41 for BOD and 0.36 and 0.67 for DOC. Higher dissolved oxygen to COD, BOD and DOC ratios resulting from decrease in substrate concentration means that there is greater dissolved oxygen availability per gram of substrate for aerobic degradation processes. The ratios of nitrate to COD, BOD and COD were respectively 0.02, 0.02 and 0.05 during infiltration of HC wastewater and sulphate, 0.4, 0.6 and 1.0.
3.2.
Redox reactions in soil column
Fig. 3aec shows respectively the mass of dissolved oxygen, nitrate and sulphate consumed per day in the sand column using HC influent at the three different HLRs of 44 cm d1, 88 cm d1 and 169 cm d1 and Fig. 3def, the mass removed when the COD was reduced. At the loading rates applied, hydraulic residence times in the soil column were 51.8 h (2.15
4216
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 1 1 e4 2 2 6
Fig. 2 e Mass loadings of: (a) Dissolved oxygen, nitrate and sulphate; (b) COD, DOC and BOD; (c) Organic nitrogen, ammonia nitrogen, total nitrogen and phosphate.
days), 23.8 h (0.99 days) and 12.2 h (0.5 days). In all cases, mass removal rate of oxygen used increased with an increase in HLR (Fig. 3a) due to higher organic mass loadings. Oxygen utilisation rates were thus highest in HC-20. At 100 mm, its consumption as a function of the amount available initially increased with corresponding increase in COD mass loading. 57%, 67% and 69% dissolved oxygen was already consumed by 100 mm for HC-5, HC-10 and HC-20 respectively. Beyond that depth, as most of the oxygen had already been consumed and concentrations in the wastewater were only 3 mg L1, removals measured were not found to reduce consistently and in some cases, slight increases in concentrations were measured. The inconsistencies in measured concentrations may be due to some entrainment of oxygen during sample collection and as the concentrations were already low, any little addition would not be buffered and would be obvious. Overall oxygen was utilised at 48 mg d1, 83 mg d1 and 182 mg d1 for HC-5, HC-10 and HC-20 respectively.
The results shown in Fig. 3b and c indicate that denitrification and sulphate reduction begun within the first 100 mm of the soil column as shown by the removal of nitrate and sulphate. These reductions occurred simultaneously with dissolved oxygen removal in the soil column. Denitrification proceeds by the reduction of nitrate to nitrogen gas through nitrite, nitric and nitrous oxide intermediaries. It is carried out in wastewater by microorganisms belonging to the Pseudomonas, Bacillus, Spirillum, Hyphomicrobium, Agrobacterium, Acinetobacter, Propionibacterium, Rhizobium, Corynebacterium, Cytophaga, Thiobacillus and Alcaligenes genera, with the latter often found in soils (Bitton, 1999). Sulphate reducing bacteria are responsible for sulphate reduction and belong to bacteria genera such as Desulfovibrio, Desulfotomaculum, Desulfobulbus, Desulfomonas, Desulfobacter, Desulfococcus, Desulfonema, Desulfosarcina, Desulfobacterium and Thermodesulfobacterium (Bitton, 1999). These bacteria are strict anaerobes, which utilize sulphate in wastewater as the terminal electron acceptor when oxygen and nitrate are not present. They are however able to tolerate oxygen in their environment. Carbon sources of low molecular weight such as fermentation products of carbohydrates and proteins are used as electron donors (Bitton, 1999). Due to the slow rate of oxygen replenishment to the soil column at reduced HLR, the other electron acceptors were increasingly relied on when HLR and thus mass loadings were increased. This occurred especially within the first 100 mm of the soil column where highest organic loading pertains. It can be seen from Fig. 3b and c that nitrate and sulphate, being the next electron acceptors after oxygen in that order were also consumed mostly within the first 100 mm of the soil column. In all soil column samples nitrate was completely consumed within the column by the 1100 mm depth. No nitrate was detected at 600 mm in HC-10 and HC-20 soil column samples. Within the whole column depth, denitrification occurred at a rate of 17 mg d1, 27 mg d1 and 39 mg d1 for HC-5, HC-10 and HC-20 respectively. As HLR and thus the mass loadings increased, the mass of sulphate removed also increased. At HLR of 44 cm d1, 88 cm d1 and 169 cm d1, sulphate reduction rates were respectively 334 mg d1, 613 mg d1 and 1406 mg d1, representing 98%, 89% and 85% reduction within the first 100 mm of the column. By the column exit, sulphate reduction had reached 660 mg d1 and 1599 mg d1 for HC-10 and HC-20 respectively. No further increase occurred for HC-5 since most of the sulphate had already been consumed. From mass balance analysis 1 g of oxygen is consumed per gram of COD removed (Metcalf and Eddy et al., 2003). It is therefore not expected that all the COD applied would be removed by aerobic degradation processes as all the DO/COD ratios were below the requirement. Fig. 3def depicts the mass removal profiles of dissolved oxygen, nitrate and sulphate along the soil column at different wastewater concentrations. Again, at 100 mm, dissolved oxygen concentrations in samples were 3 mg L1 on the average, representing a significant utilization of about 70%. In all samples, overall, greater than 80% of available dissolved oxygen was utilised. Removal rates for MC-20 and LC-20 were respectively 203 mg d1 and 230 mg d1. By the 600 mm depth, nitrate was completely depleted in all samples. In the LC influent, nitrate was again detected at the 1100 mm depth. This
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 1 1 e4 2 2 6
4217
Fig. 3 e Mass removal rate against soil column depth at 135 mg LL1 COD for: (a) dissolved oxygen (b) nitrate (c) sulphate; Mass removal rate at HLR of 169 cm dL1 for: (d) dissolved oxygen, (e) nitrate and (f) sulphate.
may be due to nitrification of ammonia taking place as a result of the low wastewater concentration. A corresponding reduction in ammonia was not detected probably due to the continued degradation of organic nitrogen in the infiltrating wastewater. Nitrate was once again reduced by denitrification further along the column. Sulphate reduction occurred at all the wastewater concentrations (Fig. 3c and f). During infiltration of HC wastewater, a large proportion of the sulphate applied (97%) was broken down (due to a high demand for oxygen for the degradation of organic carbon. Removal rate was 1599 mg d1. The characteristic rotten egg smell of sulphide emanated from all column samples, with it being strongest in the HC wastewater and lowest in the LC wastewater where less than 40% of influent sulphate was reduced. Removal rates dropped to 1029 mg d1 and 507 mg d1 in MC
and LC wastewaters respectively (Fig. 3f). At constant hydraulic loading rate therefore, sulphate reduction was found to be clearly dependent on the COD mass loading with its breakdown reducing with a corresponding reduction in COD loading. Aerobic respiration proceeds rapidly and has high biochemically efficiency compared to anaerobic respiration, which is slow and often produces malodorous chemically complex by-products (Gray, 2004). The free energy produced from microbial respiration depends on the terminal electron acceptor used and is highest for oxygen, and lowest from sulphate reduction (Bitton, 1999). The amount of adenosine triphosphate (ATP) formed during aerobic oxidative phosphorylation depends on the difference between the electron donor and electron acceptor redox potentials. Oxygen has lower redox potential compared to nitrate and sulphate and
4218
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 1 1 e4 2 2 6
thus more ATP is released. Greater microbial assimilation of organic carbon as cell material thus results from the greater energy utilisation. In mixed microbial cultures, the microorganisms pursue the route with the highest energy yield so as to attain maximum cell synthesis. Aerobic and facultative bacteria first oxidise organics in the wastewater, depleting the dissolved oxygen. After the oxygen is used up, facultative and anaerobic bacteria utilize oxygen bound in nitrates and sulphates for the breakdown of any remaining organic matter (Gray, 2004). Oxygen therefore competes with nitrate and sulphate as the final electron acceptor. This however does not preclude the occurrence of denitrification and sulphate reduction in cases where there are relatively high oxygen concentrations existing in the bulk wastewater. Denitrification processes and sulphate reduction have also been observed to occur in aerobic wastewater treatment processes such as in the biofilms of trickling filters (Bitton, 1999). Contrary to the widely accepted fact that denitrification is only possible in pure anoxic environments several studies have shown that denitrification can occur under aerobic conditions, even at oxygen concentrations that are close to or surpass air saturation values (Lloyd et al., 1987; Gao et al., 2010). Specific bacteria have also been isolated (Carter et al., 1995; Patureau et al., 2000; Kim et al., 2008) that are able to carry out aerobic denitrification as it is so called. Nitrate respiration under aerobic conditions is facilitated by active nitrate reductase sites within the periplasmic parts of the bacteria (Carter et al., 1995). Some fungi are also capable of denitrification in aerobic soils (Hayatsu et al., 2008). Sulphate reducing bacteria have also been found to exist in activated sludge flocs and their presence in this aerobic environment as well as in biofilms of rotating biological contactors and trickling filters has been attributed to the development of anoxic microsites in their environment and physiological adaptability of the anaerobic microorganisms (Lens et al., 1995). Oxygen concentration gradients across the thickness of biofilms contribute to the formation of anoxic and anaerobic zones in the deeper layers of the biofilm. Sulphate reduction has also been observed to occur consistently in well-oxygenated biomats (Canfield and Des Marais, 1991). As aerobic digestion, denitrification and anaerobic digestion of organics require the use of oxygen, nitrate and sulphate respectively as the electron acceptor, the simultaneous reduction of dissolved oxygen, nitrate and sulphate in the soil column suggests that these three processes occurred at the same time in the column and thus the substances in the wastewater are removed simultaneously by aerobic, anoxic and anaerobic processes.
3.3.
DOC, BOD and COD removal
3.3.1.
Effects of hydraulic loading rates
Fig. 4aec shows the mass removal of DOC, BOD and COD in the column at constant COD and variable HLR. There was a corresponding increase in mass removed with mass loading rate. This is because aerobic bacteria conversion of organic material, assessed by BOD, is typically a first-order reaction and is a function of the substrate concentration remaining at any time (Metcalf and Eddy et al., 2003). Even though more mass was removed as the mass loading was increased, it did not necessarily translate into improved removal efficiencies for
BOD and COD. Although for DOC, COD and BOD the lowest loading rate performed very well initially, in terms of its efficiency of removal, it did not give the best overall removal due to its slower rate of removal beyond the 100 mm depth. This may have been contributed by the high retention time in the column, necessitating high oxygen replenishment for aerobic processes and limitations on the availability of the other electron acceptors. Within the first 100 mm as dissolved oxygen levels were high, longer residence times allowed sufficient time for effective aerobic degradation of the DOC, BOD and COD. However, as dissolved oxygen was being depleted, its low rate of replenishment from the influent wastewater became an important factor as removal could not be sustained deeper within the soil column. This is evident from the flattening of the curves beyond 100 mm as the HLR is reduced (Fig. 4aec). Most of the available dissolved oxygen was consumed within the first portions of the soil column and on the average only 3 mg L1 remained after 100 mm. Besides, nitrate and sulphate had already been considerably used up within the first 100 mm with less than 0.5 mg L1 nitrate and 2 mg L1 sulphate remaining. There were thus limitations on the availability of electron acceptors. The low oxygen concentrations and yet high oxygen demand of the wastewater therefore promoted the onset of anoxic and anaerobic degradation processes. Thus by the 600 mm depth, long residence time was no longer the most efficient operating condition. Due to better oxygen replenishment upon increase of the HLR, DOC, BOD and COD removal could be sustained better within deeper layers of the soil column although the rate was considerably slowed down. A balance between residence times and the rate of replenishment of electron acceptors was thus found crucial to the efficiency of the removal process in the soil column. HC-10 therefore gave the highest and HC-5 the lowest overall removal of BOD and COD, with HC-20 being the most efficient condition for DOC. This behaviour suggests that a certain critical hydraulic loading rate may exist for efficient removal of each parameter based on the influent concentration and that long residence times do not necessarily improve performance.
3.3.2.
Effects of influent concentration
Fig. 4def shows profiles of the mass reductions in COD, BOD and DOC. At constant hydraulic loading rate, the mass loadings of COD, DOC and BOD applied to the soil column decrease with reduction in COD concentration. Mass of DOC, COD and BOD removed decreased with a corresponding decrease in mass applied. The removal efficiency however increased. COD mass removal rate was 1024 mg d1 for HC-20, decreasing to 582 mg d1 for LC-20. The corresponding increase in removal efficiency was from 29% to 54%. Increase in removal efficiency with reduction in COD under constant HLR has also been found to occur in horizontal flow constructed wetlands (Ojeda et al., 2008). Regardless of the residence times or influent concentration, the first 100 mm depth of the soil column accounted for a greater proportion of DOC, COD and BOD removal obtained. These results are in good agreement with published literature confirming the important role of the first few cm of the soil in the treatment process (Quanrud et al., 2003; Cha et al., 2004; Gru¨nheid et al., 2005). In field studies, about 75% of the DOC removal obtained after 37 m infiltration occurred within the first
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 1 1 e4 2 2 6
4219
Fig. 4 e Mass removal rate with soil column depth at 135 mg LL1 COD for: (a) DOC (b) BOD (c) COD; Mass removal rate at HLR 169 cm dL1 for: (d) DOC, (e) BOD and (f) COD.
1.5 m of soil (Quanrud et al., 2003) and in unsaturated soil columns 44% DOC removal was realised within the top 8 cm of the soil (Quanrud et al., 1996). Removal is rapid near the soil/ water interface in SAT systems because organic matter and dissolved oxygen are at their highest concentrations resulting in high biological activity (Fox et al., 2006). The prominence of the role of this depth was observed to increase with a decrease in influent concentration, accounting for 79%, 91% and then 92% of the overall removal achieved respectively. These percentages have been determined by expressing the amount of DOC removed at the 100 mm depth as a percentage of the total amount removed in the whole 2000 mm depth of the column. Overall DOC mass removals per day achieved over the 2000 mm column for HC-20, MC-20 and LC-20 were 716 mg (53%), 449 mg (65%) and 309 mg (74%) respectively.
Biodegradation as measured by the BOD (Fig. 4e) also followed a similar pattern of reduction as the DOC. Increases in the values of BOD and COD within the sand profile were observed in some samples depicted by a drop in mass removed in removal efficiency as shown in Fig. 4e and f respectively. The increases may be attributed to the generation of soluble extracellular by-products by the microorganisms in the column or desorption of organics from the sand as has been observed to occur for DOC (Drewes and Jekel, 1996; Reemtsma et al., 2000; Quanrud et al., 2003). BOD removed per day was 555 mg (28%), 456 mg (41%) and 458 mg (67%). Even though higher mass loadings corresponded with larger mass removals, the percentage removal was better at lower mass loading rates due to higher dissolved oxygen to substrate mass loading ratios. The decrease in removal of
4220
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 1 1 e4 2 2 6
COD, BOD and DOC with increase in influent concentration may be attributed to the corresponding increase in anaerobic conditions in the soil column. Magdoff et al. (1974) found in their study on soil infiltration with septic tank effluent that COD removal decreased when the soil surface became permanently ponded due to encrustation and the subsequent onset of anaerobic conditions. The pattern of dissolved oxygen, nitrate and sulphate removal (Fig. 3def) suggests the predominance of aerobic degradation during infiltration with LC wastewater. The percentage utilisation of dissolved oxygen was found to increase with a reduction in mass loadings, whilst the opposite effect occurred on sulphate. Consequently, sulphate reduction processes dropped from overall utilisation of 97% for a COD loading of 3552 mg d1 down to 70% at 1586 mg d1 and a further reduction to 40% when a COD loading of 1084 mg d1 was applied. Aerobic degradation which proceeds more rapidly and has much higher biochemical efficiency than anoxic and anaerobic degradations (Gray, 2004) thus increasingly dominated the removal process resulting in improved removal efficiencies. Dissolved oxygen (DO) mass removal rates were correlated with COD, DOC, nitrate and sulphate removal rates when the influent concentration was held constant and the HLR varied and also for constant HLR and variable COD (Fig. 5). In both cases there was a strong positive linear correlation between DO removal and the removal of the parameters considered, except nitrate. Nitrate however correlated well with COD removal under constant COD giving a rate of 0.04 mg/mg COD removed. Correlation of sulphate reduction with COD removal showed sulphate reduction occurred at a rate of 1.45 mg/mg of COD removed per day in both cases. There was a close agreement between dissolved oxygen and COD removal rates at constant COD and at constant HLR. The rates were 6.85 mg and 6.61 mg respectively, the former being chosen because the correlation was relatively stronger. Results obtained from running the soil column at condition LC-20 after sterilisation showed only slight reductions in COD and DOC occurring within the first 100 mm of the soil column, which in the biotic experiments was the most active removal zone. COD and DOC removal rates realised were only 52 mg d1 and 2.3 mg d1 respectively. Abiotic removal such as physical adsorption thus represented only a small proportion of the removals obtained in the soil column. COD, DOC and BOD removal in the soil column could therefore be attributed primarily to biological removal. This may occur through the mechanisms of biodegradation or biosorption onto biomass structures such as the biofilms (Carlson and Silverstein, 1998) developed around the sand grains.
3.3.3.
Electron donor and acceptor balance
From stoichiometric considerations, 1 g of COD would require an equal amount of dissolved oxygen for aerobic oxidation. Due to the low dissolved oxygen to COD ratios in the wastewater, which were respectively 0.06, 0.16 and 0.26 at HC, MC and LC loadings, it was not expected that the available dissolved oxygen would adequately serve as an electron acceptor for all the electrons to be donated by the organics, measured by the COD. Predicted COD removals calculated based on the oxygen available in the influent were respectively 48, 83, 182, 203 and 230 mg d1 for conditions HC-5, HC-10, HC-20, MC-20
and LC-20. COD removal attained in the soil column however exceeded these values, dissolved oxygen accounting for only 13e15% for HC, 18% for MC and 32% for LC, confirming that further COD removal occurred by some other removal mechanisms. The extra removal could not be attributed to sorption as abiotic soil column tests showed negligible COD removal efficiencies and has been attributed to denitrification and sulphate reduction processes. Using stoichiometric relations (Sarfaraz et al., 2004; Henze et al., 2008; Velasco et al., 2008), expected COD removal based on consumption of electron acceptors was estimated. The oxygen equivalent of nitrate is 2.86 mg O2/mg NO3-N. Therefore 1 mg NO3-N denitrified to nitrogen gas would have the same electron accepting capacity as 2.86 mg of oxygen. Likewise, during COD oxidation, oxygen accepts 4 electrons whilst sulphate accepts 8. Thus 2 mol of oxygen is equivalent to 1 mol of sulphate and 1 g of sulphate would have the electron accepting capability of 0.67 g of oxygen. These factors were used to convert the respective masses of electron acceptors consumed in the soil column to the equivalent COD removed and summed up to obtain a predicted value for COD removal to occur in the soil column. Actual COD removal that occurred in the soil column was found to be less than the predicted value by about 20%. HC-10 however gave only a 5% deviation (Table 3). Besides experimental errors, the deviation between actual and predicted COD removal could be attributed to incomplete denitrification to nitrogen gas. Although nitrite was not detected in any of the samples, some nitrate may have been converted to the other denitrification intermediaries being nitric and nitrous oxide. Also as the soil column environment was neither strictly aerobic, anoxic nor anaerobic, the stoichiometric conversion factors used to predict COD removal may not have been accurate.
3.4.
Phosphate removal
Phosphate was poorly removed in the soil column under all the experimental conditions. Concentrations above the influent phosphate concentration were measured in most column samples especially at the 100 mm depth. The removal was observed to be dependent on the hydraulic loading rate. The best removal occurred at the lowest HLRs (HC-5), where the mass loading was 45 mg d1. Here phosphate removal occurred after an initial increase in concentration at the 100 mm depth. Overall removal rate achieved was only 5 mg d1. For a COD of 135 mg L1, an increase in HLR led to a drop in the overall removal, from 20% to 12% for a loading rate of 88 cm d1 and then no removal is observed for a loading rate of 169 cm d1. These percentages represent removals occurring after the initial increase in concentration at 100 mm. At HLR of 169 cm d1 no phosphate removal was observed to occur in the column except within the first 100 mm of the sand when the influent of low concentration was infiltrated. The concentration however rose again in the column by about 45% after a travel distance of 600 mm. Phosphate is known to be poorly retained by sandy soils and increasing concentrations of phosphate in the soil during infiltration have also been observed by other researchers (Stuyfzand, 1989; Reemtsma et al., 2000) and may be due to the release of phosphates previously assimilated by bacterial
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 1 1 e4 2 2 6
4221
Fig. 5 e Correlations between (a) dissolved oxygen and COD and DOC, (b) dissolved oxygen and nitrate and sulphate, (c) nitrate and sulphate and COD mass removal rates at constant 135 mg LL1 COD and varying HLR, (d) dissolved oxygen and COD and DOC (e) dissolved oxygen and nitrate and sulphate, (f) Nitrate and sulphate and COD mass removal rates, at HLR of 169 cm dL1.
biofilms in the column solution following their death and subsequent degradation under anoxic conditions.
3.5.
Nitrogen removal
3.5.1.
Effect of hydraulic loading rate
The production of ammonia nitrogen and organic nitrogen and total nitrogen removal rates are shown in Fig. 6aec. There was a high conversion of organic nitrogen to ammonia at all the loading rates applied. Ammonium was the main form of
nitrogen in the column effluent. Organic nitrogen in the wastewater was transformed to ammonium by hydrolysis. At a loading rate of 44 cm d1, 55 mg d1 (representing 74%) organic nitrogen was removed within the first 100 mm of the sand (Fig. 6a). When the mass loading was increased to 130 mg d1 the removal dropped to 65%. A further increase in loading rate up to 169 cm d1 did not yield any difference in removal. Overall organic nitrogen removal was 71%, 74% and 76% for loadings of 74 mg d1 (HC-5), 130 mg d1 (HC-10) and 263 mg d1 (HC-20) respectively. The rate of removal as shown
4222
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 1 1 e4 2 2 6
Table 3 e Comparison of predicted and actual COD removal in 2000 mm of the soil column. Experimental condition
HC-5 HC-10 HC-20 LC-20
Nitrate (mg d1)
Dissolved oxygen (mg d1)
Sulphate (mg d1)
Total COD removed (mg d1)
DO used
Predicted COD removal
NO3 used
Predicted COD removal
SO4 used
Predicted COD removal
Predicted
Actual
Error (%)
48 83 182 230
48 83 182 230
17 27 39 57
48 77 112 163
332 660 1599 507
221 440 1066 338
316 600 1360 730
249 568 1118 582
21 5 18 20
Fig. 6 e Mass against column depth at 135 mg LL1 COD for (a) organic nitrogen removal, (b) ammonia production, (c) total nitrogen removal; Mass against column depth at HLR 169 cm dL1 for (d) organic nitrogen removal, (e) ammonia production, (f) total nitrogen removal.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 1 1 e4 2 2 6
by the slope of the graph after the first 100 mm was however lowest at loading rate 44 cm d1 and some increases in concentration were observed after passage through 1100 mm of the sand. After the transformation of organic nitrogen to ammonia in the lower parts of the column, the hydraulic loading rate was observed to influence the amount of ammonia present further along the column (Fig. 6b). At 44 cm d1, a decrease in the mass of ammonia in the soil column begun after the 600 mm depth. This behaviour was also observed after the 1100 mm depth at 88 cm d1, but at a slightly lower rate. The reduction in ammonia at the lower hydraulic loading rate could be attributed to biosorption onto the biofilms coating the sand grains. No decrease in ammonia concentration occurred at 169 cm d1. This may be due to the continued hydrolysis of organic nitrogen to ammonia. Total nitrogen removal was most efficient at 44 cm d1 (Fig. 6c). Overall however a removal rate of only 28 mg d1 (25% removal) was attained due to the accumulation of ammonia in the column. Removal rates at HLR of 88 cm d1 and 169 cm d1 were 41 mg d1 and 74 mg d1 being 20% and 19% removal respectively. Although a higher removal rate is obtained at higher mass loadings, the efficiency of the removal process is lowered.
3.5.2.
Effect of influent concentration
A decrease in influent concentration resulted in a decrease in the nitrogen mass loadings to the soil column. There was a corresponding reduction in the organic nitrogen removal and the conversion rate to ammonia varied with wastewater concentration, with greater percentage reduction occurring at higher influent concentrations. Removal rates were 199 mg d1 (achieving 76% removal efficiency), 88 mg d1 (69% removal efficiency) and 41 mg d1 (58% removal efficiency) in HC, MC and LC wastewaters respectively. Ammonium production was thus lowest in the LC column (Fig. 6e). Resulting total nitrogen removal was therefore found to be dependent on influent concentration with highest removal occurring in the LC wastewater (Fig. 6f). The removal of organic nitrogen under varying HLR and COD did not cause an equal increase in ammonia. As nitrification hardly occurred in the soil column, this difference could mainly be due to the use of ammonia as a nutrient source by microorganisms during cell synthesis (Gray, 2004) and also a result of adsorption, which is the main removal mechanism for ammonia during SAT (Fox et al., 2006). In the sterile column, no increase in ammonia occurred along the column. As opposed to the biotic soil column conditions, the concentration decreased along the soil column confirming the occurrence of adsorption. The removal rate obtained was 33 mg d1, representing a 69% reduction.
3.6.
Soil column reaction kinetics
The kinetics of the reactions occurring within the soil column for the removal of the wastewater parameters was determined from experiments carried out in a 300 mm long soil column filled with silica sand of the same characteristics as for the long columns, seeded similarly and fed with synthetic wastewater prepared with the same recipe as used in the current experiments. The kinetics was assessed by
4223
considering zero-, first- and second-order reactions and fitting the COD concentration data to the respective rate equations. These reaction orders did not yield a straight line. A straight line was obtained upon fitting the data to a saturation type or mixed-order reaction. Mixed-order reaction is expected to prevail within the soil column as substrate travels across biofilm layers developed around the sand grains by diffusion. A portion of the substrate is consumed within each layer, leading to reductions in substrate concentrations with the depth of the biofilm. In biofilms, saturation type reactions describe the rate of substrate utilisation at any point within the biofilm (Metcalf and Eddy et al., 2003). To compare the removal profiles obtained from the experiments with the soil column microbial concentration profile, sand was sampled from the soil column at 100 mm, 300 mm, 600 mm, 1050 mm, 1400 mm and 1700 mm for phospholipid fatty acid (PLFA) analysis. Briefly, the PLFA analysis (Findlay et al., 1989) involved extracting the PLFA from 3 g of sand samples using a one phase extraction mixture consisting of chloroform, methanol and 50 mM phosphate buffer in the ratio 1:2:0.8. The phases were broken and phosphates released from the lipid containing chloroform layer of the extraction mixture by potassium persulphate digestion. The digested extract was reacted with ammonium molybdate followed by malachite green solution and the absorbance read at a wavelength 610 nm on a UV/Vis spectrophotometer. Absorbance readings were converted to phosphate concentrations using a calibration line developed by the digestion of known concentrations of glycerol phosphate. Results of the PLFA analysis showed that the soil column had a microbial concentration profile that correlated well with the removal profile of the parameters studied. The highest concentration of phospholipids (1207 n mol/g of sand) was measured at the first sampling point. There was a marked decrease in the concentration of phospholipids beyond that point. At 300 mm, the phospholipid concentration was only 17% of that at 100 mm. The lowest concentration of 129 n mol/ g of sand, being approximately 10% of the concentration at 100 mm, was measured at 1400 mm. Phospholipids were highest at the 100 mm point because from the column entry to this point, organic carbon, nutrients and electron acceptors were at their highest concentration. This pattern is expected because in SAT systems microorganisms grow quickly and have high activity at the soil water interface due to abundance of biodegradable organic matter and dissolved oxygen (Fox et al., 2006). The removal profile correlated well with the amount of phospholipids in the soil column because biodegradation was the main removal mechanism in the soil column. Beyond the 100 mm depth of the column the bacterial consortium is likely to be composed mainly of facultative and anaerobic types, which grow more slowly than aerobes and are less biochemically efficient (Gray, 2004), contributing to lower removal of COD. Thus besides relatively lower availability of electron acceptors and less effective redox conditions pertaining beyond 100 mm, the concentration and type of microorganisms are thought to have limited the removal within the deeper layers of the soil column. Besides, it is possible that most or all of more readily biodegradable organic carbon was depleted within the first 100 mm leaving relatively slower biodegradable forms.
4224
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 1 1 e4 2 2 6
Fig. 7 e Comparison of mass removal under investigated experimental conditions.
3.7. Comparison of removal efficiencies under varying experimental conditions of COD and HLR To determine the significance of varying HLR and influent COD on the removal efficiencies obtained for dissolved oxygen, sulphate, phosphate, nitrogen, COD, BOD, and DOC, the data were analysed using the analysis of variance (ANOVA) statistical test at a significance level, p of 0.05. The experimental conditions were to be deemed to have a significant effect if the p-value obtained is less than 0.05. For all the wastewater parameters investigated, the p values were less than 0.05 and therefore the differences in removal efficiencies resulting from changes in HLR and influent COD were significant. The removal efficiencies of COD, BOD, DOC and organic nitrogen were compared under experimental conditions of HC-5 and HC-10 against LC-20 and MC-20 respectively. These conditions were chosen for comparison because the mass loadings applied for HC-5 and HC-10 were close to that of LC20 and MC-20 respectively. Fig. 7 compares the mass removal rates. The removal at 100 mm has been included in the analysis because removals beyond that point sometimes did not decrease consistently. However it can be seen that generally LC-20 and MC-20 achieved better removals than HC5 and HC-10 respectively for approximately the same COD, BOD and DOC mass applied. These differences could be attributed to dissolved oxygen limitations in the wastewater of higher concentration. Dissolved oxygen limitation in the saturated zone is an important consideration as the provision of oxygen for degradation processes is only from the regional
groundwater (Fox et al., 2006). The results suggest that the application of a low concentration substrate at short residence time would be more effective in the removal process than applying the mass in the form of a more highly concentrated substrate with long residence times. The opposite holds for organic nitrogen, where longer residence times improve its removal.
4.
Conclusions
The HLR and influent COD influenced the removal efficiencies obtained along the depth of the soil column as confirmed by the p values obtained from the ANOVA analysis. Under all experimental conditions investigated, the first 100 mm of the soil column was responsible for most of the removal or transformation of the wastewater parameters that occurred. Dissolved oxygen availability and its rate of replenishment to the soil column played a key role in the removal process. Relatively high DO to DOC, COD and BOD mass loadings improved the efficiency of the removal process. Anoxic (denitrification) and anaerobic (sulphate reduction) processes occurred in the soil column in addition to aerobic degradation. At high mass loadings, they played a more active role as electron acceptors. Higher mass loadings achieved higher mass removal rates, however COD, DOC and BOD removal efficiency is better when the mass applied has lower COD concentration. Comparison of removal efficiencies of HC-5 versus LC-20 and HC-10 against MC-20 suggests that within the range of mass loadings investigated lowering influent COD
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 1 1 e4 2 2 6
concentrations, not increasing residence times would improve removal efficiencies of COD, BOD and DOC in the saturated zone of SAT systems while high concentrations and long residence times would improve nitrogen removal. It is therefore expected that in cases where dissolved oxygen is a limiting factor, high permeability soils infiltrated with effluent of low concentration would achieve better removal efficiencies than infiltrating a relatively higher concentrated effluent through soils of low permeability allowing longer residence times. It should be noted however that the organic concentrations used in this study (even that classified as low concentration) exceed that normally applied in SAT systems.
Acknowledgement The study was carried out with support from The Netherlands Organisation for International Cooperation in Higher Education (Nuffic).
references
Asano, T., Cotruvo, J.A., 2004. Groundwater recharge with reclaimed municipal wastewater: health and regulatory considerations. Water Research 38 (8), 1941e1951. Bdour, A.N., Hamdi, M.R., Tarawneh, Z., 2009. Perspectives on sustainable wastewater treatment technologies and reuse options in the urban areas of the Mediterranean region. Desalination 237 (1e3), 162e174. Bitton, G., 1999. Wastewater Microbiology, second ed. Wiley-Liss, Inc., New York. Canfield, D., Des Marais, D., 1991. Aerobic sulfate reduction in microbial mats. Science 251 (5000), 1471e1473. Carlson, G., Silverstein, J., 1998. Effect of molecular size and charge on biofilm sorption of organic matter. Water Research 32 (5), 1580e1592. Carlson, R.R., Linstedt, K.D., Bennett, E.R., Hartman, R.B., 1982. Rapid infiltration treatment of primary and secondary effluent. Journal Water Pollution Control Federation 54 (3), 270e280. Carter, J.P., Hsiao, Y.S., Spiro, S., Richardson, D.J., 1995. Soil and sediment bacteria capable of aerobic nitrate respiration. Applied and Environmental Microbiology 61 (8), 2852e2858. Cha, W., Choi, H., Kim, J., Kim, I.S., 2004. Evaluation of wastewater effluents for soil aquifer treatment in South Korea. Water Science and Technology 50 (2), 315e322. Cha, W., Kim, J., Choi, H., 2006. Evaluation of steel slag for organic and inorganic removals in soil aquifer treatment. Water Research 40 (5), 1034e1042. Clesceri, L.S., Greenberg, A.E., Eaton, A.D., 1998. Standard Methods for the Examination of Water and Wastewater, twentieth ed. American Public Health Association, Washington. Dillon, P., Pavelic, P., Toze, S., Rinck-Pfeiffer, S., Martin, R., Knapton, A., Pidsley, D., 2006. Role of aquifer storage in water reuse. Desalination 188 (1e3), 123e134. Drewes, J.E., Jekel, M., 1996. Simulation of groundwater recharge with advanced treated wastewater. Water Science and Technology 33 (10e11), 409e418. Droste, R.L., 1997. Theory and Practice of Water and Wastewater Treatment. John Wiley & Sons Inc., New York.
4225
Findlay, R.H., King, G.M., Watling, L., 1989. Efficacy of phospholipid analysis in determining microbial biomass in sediments. Applied and Environmental Microbiology 55 (11), 2888e2893. FOSS, 2009. The determination of nitrogen according to kjeldahl using block digestion and steam distillation. Application Note 300 (8). Fox, P., Houston, S., Westerhoff, P., Nellor, M., Yanko, W., Baird, R. , Rincon, M., Gully, J., Carr, S., Arnold, R., Lancey, K., Quanrud, D., Ela, W., Amy, G., Reinhard, M., Drewes, J.E., 2006. Advances in Soil Aquifer Treatment Research for Sustainable Water Reuse. AWWA Research Foundation and American Water Works Association, Denver, CO. Fox, P., Makam, R., 2009. Surface area and travel time relationships in aquifer treatment systems. Water Environment Research 81 (11), 2337e2343. Gao, H., Schreiber, F., Collins, G., Jensen, M.M., Kostka, J.E., Lavik, G., de Beer, D., Zhou, H.Y., Kuypers, M.M.M., 2010. Aerobic denitrification in permeable Wadden Sea sediments. ISME Journal 4 (3), 417e426. Gray, N.F., 2004. Biology of Wastewater Treatment, second ed. Imperial College Press, London. Gru¨nheid, S., Amy, G., Jekel, M., 2005. Removal of bulk dissolved organic carbon (DOC) and trace organic compounds by bank filtration and artificial recharge. Water Research 39 (14), 3219e3228. Hayatsu, M., Tago, K., Saito, M., 2008. Various players in the nitrogen cycle: diversity and functions of the microorganisms involved in nitrification and denitrification. Soil Science and Plant Nutrition 54 (1), 33e45. Henze, M., van Loosdrecht, M.C.M., Ekama, G., Brdjanovic, D., 2008. Biological Wastewater Treatment: Principles, Modelling and Design. IWA Publishing, London. Kim, M., Jeong, S.Y., Yoon, S.J., Cho, S.J., Kim, Y.H., Kim, M.J., Ryu, E.Y., Lee, S.J., 2008. Aerobic denitrification of Pseudomonas putida AD-21 at different C/N ratios. Journal of Bioscience and Bioengineering 106 (5), 498e502. Lens, P.N., De Poorter, M.P., Cronenberg, C.C., Verstraete, W.H., 1995. Sulfate reducing and methane producing bacteria in aerobic wastewater treatment systems. Water Research 29 (3), 871e880. Lloyd, D., Boddy, L., Davies, K.J.P., 1987. Persistence of bacterial denitrification capacity under aerobic conditions: the rule rather than the exception. FEMS Microbiology Letters 45 (3), 185e190. Magdoff, F.R., Keeney, D.R., Bouma, J., Ziebell, W.A., 1974. Columns representing mound-type disposal systems for septic tank effluent: II. Nutrient transformations and bacterial populations. Journal of Environmental Quality 3 (3), 228e234. Metcalf and Eddy, Tchobanoglous G., Burton, F.L., Stensel, H.D., 2003. Wastewater Engineering: Treatment and Reuse, fourth ed. McGraw-Hill Inc., New York. Nema, P., Ojha, C.S.P., Kumar, A., Khanna, P., 2001. Technoeconomic evaluation of soil-aquifer treatment using primary effluent at Ahmedabad, India. Water Research 35 (9), 2179e2190. OECD, 1996. Guideline for Testing of Chemicals Simulation TestAerobic Sewage Treatment. Technical Report. Organisation for Economic Co-operation and Development (OECD), Paris, France. Ojeda, E., Caldentey, J., Saaltink, M.W., Garcia, J., 2008. Evaluation of relative importance of different microbial reactions on organic matter removal in horizontal subsurface-flow constructed wetlands using a 2D simulation model. Ecological Engineering 34 (1), 65e75. Patureau, D., Zumstein, E., Delgenes, J.P., Moletta, R., 2000. Aerobic denitrifiers isolated from diverse natural and managed ecosystems. Microbial Ecology 39, 145e152.
4226
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 1 1 e4 2 2 6
Pescod, M.B., 1992. Wastewater Treatment and Use in Agriculture e FAO Irrigation and Drainage Paper , 47 ed., vol. 47. Food and Agriculture Organization of the United Nations, Rome. Prochaska, C.A., Zouboulis, A.I., Eskridge, K.M., 2007. Performance of pilot-scale vertical-flow constructed wetlands, as affected by season, substrate, hydraulic load and frequency of application of simulated urban sewage. Ecological Engineering 31 (1), 57e66. Quanrud, D.M., Arnold, R.G., Wilson, L.G., Gordon, H.J., Graham, D.W., Amy, G.L., 1996. Fate of organics during column studies of soil aquifer treatment. Journal of Environmental Engineering 122 (4), 314e321. Quanrud, D.M., Hafer, J., Karpiscak, M.M., Zhang, J., Lansey, K.E., Arnold, R.G., 2003. Fate of organics during soil-aquifer treatment: sustainability of removals in the field. Water Research 37 (14), 3401e3411. Reemtsma, T., Gnirß, R., Jekel, M., 2000. Infiltration of combined sewer overflow and tertiary municipal wastewater: an
integrated laboratory and field study on nutrients and dissolved organics. Water Research 34 (4), 1179e1186. Relyea, J.F., 1982. Theoretical and experimental considerations for the use of the column method. Radioactive Waste Management and Nuclear Fuel Cycle, 3151e3166. Sarfaraz, S., Thomas, S., Tewari, U.K., Iyengar, L., 2004. Anoxic treatment of phenolic wastewater in sequencing batch reactor. Water Research 38 (4), 965e971. Stuyfzand, P.J., 1989. Hydrology and water-quality aspects of Rhine bank groundwater in the Netherlands. Journal of Hydrology 106 (3e4), 341e363. TanIk, A., C¸omakoglu, B., 1996. Nutrient removal from domestic wastewater by rapid infiltration system. Journal of Arid Environments 34 (3), 379e390. Velasco, A., Ramı´rez, M., Volke-Sepu´lveda, T., Gonza´lezSa´nchez, A., Revah, S., 2008. Evaluation of feed COD/sulfate ratio as a control criterion for the biological hydrogen sulfide production and lead precipitation. Journal of Hazardous Materials 151 (2e3), 407e413.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 2 7 e4 2 3 7
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Natural versus wastewater derived dissolved organic carbon: Implications for the environmental fate of organic micropollutants Peta A. Neale a,*, Alice Antony b, Wolfgang Gernjak c, Greg Leslie b, Beate I. Escher a a
The University of Queensland, National Research Centre for Environmental Toxicology (Entox), Brisbane QLD 4108, Australia UNESCO Centre for Membrane Science and Technology, The University of New South Wales, Sydney NSW 2033, Australia c The University of Queensland, Advanced Water Management Centre (AWMC), Brisbane QLD 4072, Australia b
article info
abstract
Article history:
The interaction of organic micropollutants with dissolved organic carbon (DOC) can
Received 21 February 2011
influence their transport, degradation and bioavailability. While this has been well estab-
Received in revised form
lished for natural organic carbon, very little is known regarding the influence of DOC on the
27 May 2011
fate of micropollutants during wastewater treatment and water recycling. Dissolved
Accepted 27 May 2011
organic carbonewater partition coefficients (KDOC) for wastewater derived and reference
Available online 7 June 2011
DOC were measured for a range of micropollutants using a depletion method with poly-
Keywords:
(log KOW) greater than 4 there was a significant difference in KDOC between reference and
Dissolved organic carbon
wastewater derived DOC, with partitioning to wastewater derived DOC over 1000 times
Micropollutants
lower for the most hydrophobic micropollutants. The interaction of nonylphenol with
Water recycling
wastewater derived DOC from different stages of a wastewater and advanced water
dimethylsiloxane disks. For micropollutants with an octanolewater partition coefficient
Partition coefficient
treatment train was studied, but little difference in KDOC was observed. Organic carbon characterisation revealed that reference and wastewater derived DOC had very different properties due to their different origins. Consequently, the reduced sorption capacity of wastewater derived DOC may be related to their microbial origin which led to reduced aromaticity and lower molecular weight. This study suggests that for hydrophobic micropollutants (log KOW > 4) a higher concentration of freely dissolved and thus bioavailable micropollutants is expected in the presence of wastewater derived DOC than predicted using KDOC values quantified using reference DOC. The implication is that naturally derived DOC may not be an appropriate surrogate for wastewater derived DOC as a matrix for assessing the fate of micropollutants in engineered systems. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
Organic micropollutants can be defined as natural and synthetic organic compounds found in the environment at picogramme per litre (pg/L) to microgram per litre (mg/L) concentrations (Schwarzenbach et al., 2006). Due to variable
removal by wastewater treatment processes micropollutants are often detected at low concentrations in secondary treated effluent (e.g. Miao et al., 2004; Ying et al., 2009) as well as surface waters (e.g. Kolpin et al., 2002; Yoon et al., 2010). The implications of micropollutants in the environment are wide ranging and can include feminisation of male fish by
* Corresponding author. Tel.: þ61 7 3274 9221; fax: þ61 7 3274 9003. E-mail address:
[email protected] (P.A. Neale). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.05.038
4228
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 2 7 e4 2 3 7
steroidal hormones (Jobling et al., 1998), increased bacterial resistance by antibiotics (Reinthaler et al., 2003) and significant risks for human health and the environment. The fate and behaviour of micropollutants in the aquatic environment can be influenced by their interaction with dissolved organic carbon (DOC), which is ubiquitous in natural waters. Bioavailability and hence toxicity of micropollutants can decrease when bound to organic carbon (e.g. Burgess et al., 2005; Qiao and Farrell, 2002). In contrast, studies have also shown that the presence of DOC can reduce micropollutant sorption to soil, thus potentially increasing the mobility of micropollutants in the environment (Huang and Lee, 2001). This interaction can also influence the degradation of micropollutants by photodegradation (Lam and Mabury, 2005; Latch and McNeill, 2006) and hydroxyl radicals (Lindsey and Tarr, 2000). While many studies have observed reduced degradation in the presence of DOC, Lam and Mabury (2005) found increased degradation of carbamazepine and atorvastatin with DOC and attributed this to the increased formation of reactive oxygen species upon irradiation. The interaction of micropollutants with DOC can be quantified via dissolved organic carbonewater partition coefficients (KDOC) which represent the equilibrium distribution of a micropollutant between the two phases. Quantification of KDOC can be difficult as most analytical techniques cannot differentiate between micropollutants that are sorbed to DOC and those that are freely dissolved. However, through the use of a third phase, such as a polymer fibre, this limitation can be overcome (ter Laak et al., 2005). The majority of studies have focused on quantifying micropollutant interaction with reference or natural DOC (e.g. Chefetz and Xing, 2009), with little known regarding micropollutant interaction with wastewater derived DOC. Wastewater derived DOC contains a range of components including natural organic matter, microbially derived material and organic micropollutants, and the properties can vary significantly with season and location, while treatment processes can modify both the quality and quantity of wastewater derived DOC (Shon et al., 2006). An understanding of micropollutant interaction with wastewater derived DOC is important as many streams and rivers, particularly in arid or semi-arid climates, can be dominated by discharges from wastewater treatment plants (WWTP) (Brooks et al., 2006). Further, given the increased use of secondary treated effluent as the feed water for advanced water treatment processes in non-potable and indirect potable applications (Hawker et al., 2011), it is important to monitor the fate and behaviour of micropollutants through the secondary treatment and subsequent advanced treatment processes. There are few studies which have attempted to quantify micropollutant interaction with domestic wastewater derived DOC, though the importance of this interaction for micropollutant fate during the secondary treatment stage has been identified by Katsoyiannis and Samara (2007). This study found decreased micropollutant sorption to wastewater solids with increasing DOC concentration, suggesting that the micropollutant-DOC interaction could interfere with the micropollutant removal efficiency of the secondary treatment process. The majority of studies fail to consider the dissolved phase, instead only focus on the particulate and water phases,
which will contain both freely dissolved and DOC-bound micropollutants (e.g. Arditsoglou and Voutsa, 2010). The lack of studies is related to the difficulty associated with measuring the freely dissolved fraction (Barret et al., 2010). Quantification techniques, such as equilibrium dialysis and solubility enhancement, have been applied to measure partitioning of micropollutants, including pesticides, antibiotics and fluorotelomer alcohols, to wastewater derived DOC (Carmosini and Lee, 2009, 2008; Ilani et al., 2005; Seol and Lee, 2000). In the majority of studies KDOC for wastewater derived DOC was significantly lower than reference or natural DOC, while KDOC could not be measured for the antibiotic ciprofloxacin suggesting it had no detectable affinity for wastewater effluent (Carmosini and Lee, 2009). From the literature, it appears that micropollutants interact differently with wastewater derived DOC compared to reference or natural DOC, however, this interaction is poorly understood. The aim of this study was to assess micropollutant partitioning to DOC taken from different stages of the wastewater and advanced water treatment train and compare with reference and natural DOC. The studied DOC was characterised with liquid chromatography-organic carbon detection (LC-OCD) to understand how composition and size distribution influence partitioning. KDOC was measured using polydimethylsiloxane (PDMS) disks which act as a third phase, with desorption of micropollutants from preloaded disks in the presence and absence of DOC allowing for the derivation of KDOC. The proposed PDMS disk method was developed to measure partitioning of proteins and lipid vesicles (Kwon et al., 2009) and was recently applied to DOC (Kim et al., 2010).
2.
Materials and methods
2.1.
Dissolved organic carbon
Water samples were collected from Bundamba Advanced Water Treatment Plant (AWTP) and South Caboolture WWTP, Queensland, Australia. Bundamba AWTP receives primarily domestic secondary treated effluent from four WWTPs including Bundamba, Oxley, Goodna and Wacol (Queensland, Australia). The treatment processes used at Bundamba AWTP includes pre-treatment with coagulation and clarification, followed by microfiltration, reverse osmosis and advanced oxidation, while South Caboolture WWTP applies biological nutrient removal. Wastewater derived DOC was collected from the WWTP influent (South Caboolture), secondary treated effluent (Bundamba and South Caboolture), reverse osmosis feed (ROF) and reverse osmosis concentrate (ROC) (both from Bundamba). Sodium thiosulphate was added to ROF and ROC to quench chloramines. All samples were filtered using 0.45 mm nylon filters to remove particulate matter. The non-purgeable DOC concentration in the samples was measured using an Analytik Jena multi N/C 3100 instrument (Jena, Germany) and the concentration ranged from 9 to 70 mg of carbon per litre (mgC/L). All samples were concentrated to 2 mgC/mL by freeze drying after freezing with liquid nitrogen. Aldrich humic acid (HA) sodium salt (Castle Hill, Australia), Suwannee River standard HA (2S101H) and fulvic
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 2 7 e4 2 3 7
acid (FA) (2S101F) (International Humic Substance Society, St. Paul, US) were selected as the reference DOC as they are commonly used in the literature.
2.2.
Chemicals
All chemicals were of analytical grade. The 100 mM phosphate buffer at pH 7.8 was composed of potassium phosphate (KH2PO4 and K2HPO4). The studied micropollutants included pharmaceuticals, pesticides, endocrine disrupting chemicals and polycyclic aromatic hydrocarbons (PAH). Specifically, these were 4-n-nonylphenol (Alfa Aesar, Heysham, UK), irgarol, terbutryn, pyrene, metolachlor (Fluka, Buchs, Switzerland), methoxychlor (Riedel-de Hae¨n, Seelze, Germany), chlorpyrifos (Dow Chemical Company, Midland, US), benzo(a)pyrene, carbamazepine (Sigma Aldrich, Castle Hill, Australia) and dibenzo(ah)anthracene (Supleco, Bellefonte, US). The chemicals were selected as they represent a wide range of octanolewater partition coefficients (KOW) covering more than four orders of magnitude (log KOW 2.3e6.75). All chemicals were neutral at the studied pH. All chemicals, except for chlorpyrifos and methoxychlor, were analysed using a Shimadzu High Performance Liquid Chromatography (HPLC) system with an LC-20AD pump and a SIL-20AHT auto sampler (Rydalmere, Australia). PAHs were analysed using a Supelcosil LC-PAH column (150 mm 4.6 mm, 5 mm) (Supleco, Bellefonte, US) at 40 C and detected using an RF-10AXL fluorescence detector. Pyrene had excitation and emission wavelengths of 330 and 375 nm, respectively, while benzo(a)pyrene and dibenzo(ah)anthracene both had excitation and emission wavelengths of 290 and 430 nm, respectively. The other chemicals were analysed using a Nucleodur C18 Gravity column (125 mm 4.6 mm, 5 mm) (MachereyeNagel, Du¨ren, Germany) at 40 C and detected using an SPD-M20A diode array detector. For all chemicals the flow rate was 1 mL/min. The mobile phase consisted of MilliQ grade water and methanol, though a phosphate buffer (20 mM K2HPO4 pH 3) was used for nonylphenol instead of water. Methoxychlor and chlorpyrifos were analysed using a Hewlett Packard 5890 Gas Chromatography-Electron Capture Detector (GC-ECD) Series II with an HP-7673A auto sampler (Palo Alto, US). For methoxychlor the column temperature started at 150 C and increased to 220 C at a rate of 30 C/min followed by 10 C/min until 270 C. The column temperature for chlorpyrifos also started at 150 C and increased to 220 C at a rate of 30 C/min followed by 10 C/min until 250 C and 30 C/min until 300 C, which was then held for 1 min. Both chemicals were analysed using a DB-5 column (30 m 0.25 mm i.d.) (J&W Scientific, Folsom, US).
4229
were loaded to the disks in methanol:water solutions (60:40) for 4 days with concentrations ranging from 0.0002 to 4 mg/ mL, depending on detection limit. Consequently, the initial concentrations in the PDMS disks were in the mg/L PDMS range (3e3500 mg/L PDMS). In 2 mL HPLC vials, a preloaded disk was added to a suspension containing 100 mM phosphate buffer (pH 7.8) and DOC at concentrations ranging from 1 to 2 mgC/mL. Sodium azide (0.05%) was added for preservation. The vials were shaken for 96 h in an incubator at 25 C. While equilibrium between DOC and water is expected to be reached instantaneously, equilibrium between the PDMS and DOC suspension was only reached at 96 h for the more hydrophobic compounds, such as nonylphenol (Fig. 1). This was due to rate-limited desorption from the PDMS caused by the aqueous diffusion layer around the disk (ter Laak et al., 2008). Preloaded disks were added to vials containing phosphate buffer only for 96 h controls. After 96 h the disks from the DOC suspension (CPDMS t ¼ 96, with DOC) and t ¼ 96 h controls (CPDMS t ¼ 96, without DOC) were removed and added to vials containing 500 mL of methanol or 200 mL of hexane and desorbed by shaking for 2 h in an incubator at 25 C. Given the high solvent volume to disk volume ratio, the extraction efficiency should be exhaustive. Preloaded disks were also added directly to vials containing 500 mL of methanol (HPLC) or 200 mL of hexane (GC-ECD) for time zero (t ¼ 0) controls (CPDMS t ¼ 0). The methanol extracts were analysed using HPLC and the hexane extracts were analysed using GC-ECD. The disks were dried and weighed to determine CPDMS t ¼ 0, CPDMS t ¼ 96, without DOC and CPDMS t ¼ 96, with DOC. All experiments were repeated in triplicate. The PDMS-water partition coefficient (KPDMS-w) represents the equilibrium distribution of a micropollutant between the PDMS disk and water in the absence of DOC. KPDMS-w was measured independently for hydrophobic micropollutants (log KOW > 4) using the aqueous boundary layer (ABL)
2.3. Dissolved organic carbonewater partition coefficient Partitioning between DOC and water for the studied micropollutants was measured using a depletion method developed initially by ter Laak et al. (2005) using solid-phase microextraction (SPME) fibres and adapted to PDMS disks by Kwon et al. (2009). Prior to the experiment, disks with a volume of approximately 1.4 mL were cut and cleaned by soxhleting with hexane and methanol for 2 h each. The studied chemicals
Fig. 1 e Concentration in PDMS in the presence of DOC relative to the initial concentration in PDMS (CPDMS t [ 96, with DOC/CPDMS t [ 0) as a function of time with 95% confidence intervals (pH 7.8, 100 mM phosphate buffer, average CPDMS t [ 0 3711 mg/L PDMS, Aldrich HA concentration 2 mgC/mL).
4230
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 2 7 e4 2 3 7
permeation method (Kwon et al., 2007). For hydrophilic micropollutants (log KOW < 4), KPDMS-w was measured in situ using a full mass balance as there was significant depletion from the disks in the presence of buffer alone (20e70%). The full mass balance is shown in Equation (1), where ntot was the total amount in the disk at time zero, nPDMS1 was the amount in the disk after 96 h in phosphate buffer and nw1 was the freely dissolved amount in water (Fig. 2). KPDMS-w was then determined using Equation (2) where Vw was the solution volume (L) and VPDMS was the PDMS volume (L). All KPDMS-w values used in this study are shown in Table 1. ntot ¼ nPDMS1 þ nw1 KPDMSw ¼
(1)
nPDMS1 Vw $ nw1 VPDMS
(2)
As the presence of DOC reduced the amount of freely dissolved micropollutants in water (nw2), a new mass balance equation was required (Equation (3)) where nPDMS2 was the amount in the disk after 96 h in DOC suspension and nDOC was the amount sorbed to DOC (Fig. 2). In both Equations (1) and (3), 100% mass balance was assumed, though sorption to glass vials and volatilisation was possible. Such losses were minimised by the high sorptive capacity of the PDMS and the small headspace. KDOC (L/kg) was determined using Equation (4) where mDOC was the mass of DOC in suspension (kg). ntot ¼ nPDMS2 þ nw2 þ nDOC
KDOC
(3)
ntot $KPDMSw ,VPDMS Vw ðKPDMSw ,VPDMS Þ nPDMS2 ¼ mDOC
(4)
As KDOC is a concentration ratio, the fraction of micropollutant sorbed to DOC ( fDOC) can change with changing DOC concentration. This is particularly relevant to wastewater derived DOC as the quantity of DOC can be altered throughout the treatment train. fDOC was calculated using Equation (5). fDOC
2.4.
1 ¼ Vw 1þ ðmDOC ,KDOC Þ
3.
Results and discussion
3.1.
Isotherms
To assess the influence of the preload concentration on partitioning the freely dissolved (Cw) and sorbed (CDOC) concentrations were studied over a 10 fold concentration range for nonylphenol. Within the literature, nonlinear isotherms have been observed for DOC with increasing micropollutant concentration (Laor and Rebhun, 2002) and this could be a potential limitation for the chemicals with a higher detection limit, such as nonylphenol. Using the Freundlich equation, the slope of the log regression was close to 1 which suggests that sorption was linear on a nonlogarithmic scale over the studied concentration range (Fig. 3). Consequently, it was a partitioning process and the sorption sites were not yet saturated indicating that it was still acceptable to measure partitioning at higher concentrations.
3.2. Dissolved organic carbonewater partition coefficients
(5)
Dissolved organic carbon characterisation
The studied DOC was characterised using liquid chromatography combined with an organic carbon detector (LC-OCD) (DOC-Labour, Karlsruhe, Germany). This technique combines size exclusion chromatography with organic carbon detection to separate DOC into different fractions, such as biopolymers,
nPDMS1 Phosphate buffer
humic substances, building blocks (degraded humic substances) and low molecular weight (LMW) neutrals and acids. DOC is separated via steric interactions with the size exclusion chromatography resin, while the LMW organic acid fraction is separated by amphiphilic elution (Ciputra et al., 2010). LC-OCD can also provide information on humic substance molecular weight and aromaticity, as indicated by specific UV absorbance (SUVA) at 254 nm. A size exclusion column (HW-50S) (Tosoh, Stuttgart, Germany) with a particle size of 30 mm was used. The mobile phase was 28 mM phosphate buffer (pH 6.58). For each sample 1000 mL was injected and each sample ran for 150 min. The chromatograms were interpreted using DOC-Labor ChromCALC. Further information on the LC-OCD method used and instrument calibration can be found in Ciputra et al. (2010) and Huber et al. (2011).
nPDMS2
DOC Suspension
nDOC
nw1
nw2
ntot = ntot Fig. 2 e Full mass balance in the absence and presence of DOC.
To compare micropollutant interaction with reference and wastewater derived DOC KDOC was measured for a range of micropollutants with Aldrich HA and ROC (Table 1). Given the increased interest in water recycling using advanced water treatment processes, such as membrane filtration, ROC was selected as a representative wastewater derived DOC. As well as being rich in DOC (up to 70 mgC/L) and salts (conductivity around 5.55 mS/cm), it can also contain elevated levels of micropollutants (Watkinson et al., 2007). Prior to being disposed in the estuarine Brisbane River, ROC is treated using nitrifying and denitrifying processes to reduce nutrient levels, however, the presence of micropollutants in treated ROC may pose an environmental hazard to the receiving waters. A strong relationship was observed between log KOW and log KDOC for Aldrich HA (Fig. 4). The slope and intercept were not statistically different from 1 and 0, respectively (Table 2). The correlation suggests that octanol was a perfect surrogate for Aldrich HA. The quantitative structureeactivity relationship (QSAR) obtained here was similar to previous studies with HA and hydrophobic micropollutants (Table 2), and indicates that partitioning was driven by non-specific
4231
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 2 7 e4 2 3 7
Table 1 e Octanolewater partition coefficients (KOW), PDMS-water partition coefficients (KPDMS-w) and dissolved organic carbonewater partition coefficients (KDOC) for a range of chemicals with Aldrich HA and reverse osmosis concentrate from Bundamba Advanced Water Treatment Plant. log KOWa
KPDMS-w
Carbamazepine Metolachlor Irgarol Terbutryn Methoxychlor Chlorpyrifos Pyrene
2.30 3.13 3.38 3.74 4.95 4.96 5.00
0.23 0.52 0.34 0.27
179e 657e 3188e 2254e 30903f 22863g 22909h
Nonylphenol Benzo(a)pyrene Dibenzo(ah) anthracene
5.76 6.35 0.33 6.75 0.34
32359f 123027h 295121h
0.35 0.32
log KDOC Aldrich HAb 2.87 2.95 3.21 3.31 4.99 4.36 5.15
log KDOC ROCb
Fraction sorbed to ROCc
Modelled log KOCd
8.85% 4.85% 8.48% 4.25% 7.47% 8.48% 9.22%
2.23 2.39 2.63 2.78 4.43 3.86 4.74
11.11% 28.33% 44.10%
4.58 5.32 5.68
0.06 0.03 0.06 0.07 0.19 0.12 0.08
5.25 0.15 6.84 0.15 6.96 0.09
3.14 2.86 3.12 2.80 3.06 3.12 3.16
0.24 0.24 0.41 0.36 0.53 0.28 0.37
3.25 0.13 3.75 0.26 4.05 0.62
Literature log KOC (Aldrich HA) e e e e e 4.28i 5.18,j 5.02,k 5.36,l 5.55,m 5.51n 4.83 6.28m7.16n6.31p 7.56n
a Recommended experimental octanolewater partition coefficient (log KOW) with standard deviation (Sangster, 2006). b L/kg. c Calculated using Equation (5). d Organic carbonewater partition coefficient (KOC) modelled using KOCWIN (estimated using log KOW) (US EPA, 2008). e Measured in situ using Equation (2). f Measured using the ABL permeation method (see Kwon et al. (2007) for further details). g van der Voet (2008). h Kwon et al. (2007). i Huang and Lee (2001). j Chin et al. (1997). k Gauthier et al. (1987). l Perminova et al. (1999). m ter Laak et al. (2005). n Kim and Kwon (2010). o Yamamoto et al. (2003). p McCarthy and Jimenez (1985).
interactions, such as Van der Waals forces. In contrast, a weak correlation was observed between log KOW and log KDOC for ROC (r2 ¼ 0.58), with a slope of 0.20 and an intercept of 2.30 (Table 2). Previous studies have attributed such changes in slope to the hydrophobicity of organic carbon, with Schwarzenbach and Westall (1981) finding a reduction in
Fig. 3 e Nonylphenol linear isotherm with Cw as the concentration freely dissolved in water (mol/L) and CDOC as the concentration sorbed to DOC (pH 7.8, 100 mM phosphate buffer, average CPDMS t [ 0 2776e23013 mg/L PDMS, Aldrich HA concentration 2 mgC/mL).
slope as the organic carbon became more hydrophilic. A number of studies have indicated that wastewater derived DOC contains more hydrophilic carbon than reference DOC as certain treatment processes, such as ozonation and membrane filtration, can significantly reduce the hydrophobic fraction in wastewater derived DOC (Imai et al., 2002). The different slopes for Aldrich HA and ROC may also indicate different intermolecular interactions between the studied DOC and micropollutants (Niederer et al., 2007). There was no significant difference between KDOC for Aldrich HA and ROC for micropollutants with a log KOW less than 4. For these compounds minimal depletion from the PDMS disk was observed in the presence of both ROC and Aldrich HA. These micropollutants are more soluble than the other studied compounds and previous work by Chiou et al. (1986) has shown that DOC concentration and properties can have little influence on the solubility enhancement of such micropollutants which is related to partitioning. Consequently, it appears that the difference in DOC properties have minimal influence on the sorption of these more soluble micropollutants. In contrast, partitioning of the most hydrophobic micropollutants, such as benzo(a)pyrene, to Aldrich HA was over 1000 times greater than ROC. These compounds are non-polar and have a planar conformation, promoting strong interactions with the hydrophobic Aldrich HA, with depletion from the PDMS disk up to 99% for the most hydrophobic micropollutants. Using Equation (5), the fraction of micropollutants sorbed to ROC was estimated (Table 1). For the majority of the
4232
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 2 7 e4 2 3 7
2008). Above a log KOW of 4, the modelled KOC values generally fit better with the Aldrich HA KDOC values compared to ROC. As a result, such modelled values are not suitable for the prediction of micropollutant interaction with wastewater derived DOC. Within the literature, the interaction of some of the studied chemicals, including chlorpyrifos, nonylphenol and dibenzo(ah)anthracene, have been quantified with Aldrich HA using a variety of techniques (e.g. Huang and Lee, 2001; Kim and Kwon, 2010; Yamamoto et al., 2003). The literature KDOC values fit well with the Aldrich HA KDOC values in this study (Table 1). The potential for DOC uptake to the disks was explored using the method described in Section 2.3 with clean PDMS disks. In the presence of wastewater derived DOC, particularly ROC, some small peaks were observed at the beginning of the HPLC chromatograms and it was assumed that these were hydrophilic micropollutants which were poorly removed by conventional treatment processes. However, the concentrations of these compounds on the disks were insignificant compared to the concentrations of the studied micropollutants. No changes in HPLC chromatograms were observed for reference DOC suggesting that DOC was not bound to the disks. Further, no visible fouling, such as colour change, was observed indicating that DOC uptake to the disks was not significant.
Fig. 4 e Relationship between octanolewater partition coefficients (KOW) and dissolved organic carbonewater partition coefficients (KDOC) for studied micropollutants for Aldrich HA and Bundamba reverse osmosis concentrate (pH 7.8, 100 mM phosphate buffer, average CPDMS t [ 0 3e3500 mg/L PDMS, DOC concentration 1e2 mgC/mL). The error bars represent standard deviation, with some error bars, particularly for Aldrich HA, smaller than the symbol size.
3.3. Nonylphenol sorption throughout the treatment train compounds, the fraction sorbed to DOC was less than 10% despite the high carbon concentration of ROC. Sorption to DOC in ROC was more significant for the highly hydrophobic micropollutants, such as benzo(a)pyrene and dibenzo(ah) anthracene (28% and 44%, respectively). However, as most micropollutants present in secondary treated effluent are expected to be more hydrophilic, as the more hydrophobic compounds are already removed by sorption to biosolids during secondary treatment, the majority can be considered freely dissolved. In Table 1 experimental KDOC values for Aldrich HA and ROC were compared with modelled organic carbonewater partition coefficients (KOC) predicted using KOCWIN (US EPA,
During water treatment processes the quality and quantity of DOC can be altered, and this is expected to have implication for micropollutant fate. The interaction of nonylphenol with influent and secondary treated effluent from South Caboolture WWTP and secondary treated effluent, ROF and ROC from Bundamba AWTP is shown in Fig. 5A and compared to partitioning to reference DOC, including Aldrich HA and Suwannee River HA and FA. Nonylphenol was selected for study as it has been found in concentrations up to 0.069 mg/L in purified recycled samples taken from Bundamba AWTP (Hawker et al., 2011), indicating that it was not removed effectively during conventional wastewater treatment processes and persists
Table 2 e Quantitative activityestructure relationships (QSAR) between dissolved organic carbonewater partition coefficients (KDOC) and octanolewater partition coefficients (KOW) from the current study and the literature (log KDOC [ slope 3 log KOW D intercept). Dissolved organic carbon Bundamba ROC Aldrich HA Aldrich HAa Aldrich HAb Suwannee River FAb Roth HAc Roth HAc Aldrich HAd a b c d
Durjava et al. (2007). Kim and Kwon (2010). Poerschmann and Kopinke (2001). ter Laak et al. (2005).
Slope std. error 0.20 1.01 0.76 1.23 0.82 0.92 0.98 1.19
0.06 0.01 0.08 0.13 0.09 0.04 0.06 0.07
Intercept std. error 2.30 0.07 1.55 0.82 0.31 0.47 0.39 0.62
0.29 0.48 0.55 0.75 0.53 0.26 0.25 0.40
r2
Studied micropollutants
0.58 0.93 0.94 0.94 0.93 0.99 0.99 0.99
Current study Current study PCBs PAHs PAHs PCBs PAHs PAHs
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 2 7 e4 2 3 7
4233
Fig. 5 e A) Dissolved organic carbonewater partition coefficients (KDOC) for nonylphenol for reference and wastewater derived DOC and B) DOC concentration and fraction of nonylphenol sorbed to wastewater derived DOC (fDOC) (pH 7.8, 100 mM phosphate buffer, average CPDMS t [ 0 3500 mg/L PDMS; DOC concentration 1e2 mgC/mL).
through the advanced water treatment train despite being highly hydrophobic (log KOW 5.76). Fig. 5A indicates that there was no significant difference in partitioning for the wastewater derived DOC, despite treatment processes, such as coagulation and microfiltration, expected to alter the quality of DOC. The quantity of the DOC decreased throughout the treatment train, for example, the influent at South Caboolture WWTP had a DOC concentration of 49 mgC/L which decreased to 15 mgC/L in the secondary treated effluent (Fig. 5B). The change in DOC concentration will influence the fraction of nonylphenol sorbed to wastewater DOC. For example, approximately 10e15% of nonylphenol was expected to be sorbed to DOC in the WWTP influent and ROC, compared to only 1e2% sorbed DOC in the secondary treated effluent and ROF, despite similar KDOC values. For the reference DOC, KDOC for Aldrich HA was an order of magnitude larger than Suwannee River HA, while KDOC for Suwannee River FA was similar to the wastewater derived DOC (Fig. 5A). A similar order of partitioning was observed previously in the literature (e.g. Chin et al., 1997; Niederer et al., 2007) and the differences may be related to the different origins and properties of the reference DOC. It has been suggested previously that commercial HA, such as Aldrich HA, are not representative of naturally occurring DOC (Malcolm and MacCarthy, 1986), however it was still included in this study as it has been used widely in the literature and served to validate the experimental method. The decreased partitioning of Suwannee River FA compared to Suwannee River HA may be related to the higher content of carboxyl groups (Ritchie and Perdue, 2003). The carboxyl groups were deprotonated at the studied pH, making FA more polar than HA, which consequently reduced its sorption capacity. The low depletion of nonylphenol from the disk in the presence of Suwannee River FA added increased uncertainty to the results. However, it was not possible to increase the volume of suspension as this would lead to the freely dissolved fraction no longer being insignificant.
3.4.
Dissolved organic carbon characterisation
To understand why micropollutants have a lower affinity for wastewater derived DOC compared to reference DOC, the DOC was characterised using LC-OCD. This technique revealed that the reference DOC contained a higher fraction of humic substances compared to wastewater derived DOC, which contained more biogenic organic carbon, including biopolymers and LMW neutrals (Table 3). Consequently, wastewater derived DOC had a lower weight-averaged molecular weight (MW) compared to reference DOC (578e800 Vs. 928e1469 g/mol). Compared to previous studies, such as Chin et al. (1994), the MW of Aldrich HA is low (1092 V 4100), however it is important to note that this is the MW of the humic substance fraction only, not the whole sample. This also explains why there is little difference in polydispersity between wastewater derived and reference DOC (Table 3). Wastewater derived DOC also had a lower SUVA value which suggests that wastewater derived DOC is less aromatic than reference DOC. Low SUVA values have been previously found in effluent impacted waters and this was attributed to the microbial or autochthonous origin of wastewater derived DOC (Rosario-Ortiz et al., 2007). Further, the biopolymer fraction of wastewater derived DOC contained a significant fraction of proteins which is also an indicator of microbial activity (Drewes and Croue´, 2002). The different properties of the reference and wastewater derived DOC reflect their different origins.
3.5. Influence of dissolved organic carbon properties on micropollutant partitioning To improve understanding of micropollutant interaction with DOC many studies have focused on the relationship between KDOC and DOC properties, such as MW and polarity (Chiou et al., 1986). Using the LC-OCD results in Table 3, the relationship between KDOC for nonylphenol and weight-averaged
4234
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 2 7 e4 2 3 7
Table 3 e Liquid chromatography-organic carbon detector analysis of reference and wastewater derived DOC. Biopolymersb (%)
Humic Substances (%)
Building Blocksc (%)
LMW Neutrals (%)
e e
e 0.2
61.2 78.3
9.6 9.7
18.8 14.9
1092 1469
730 1000
1.50 1.47
9.80 7.74
e e
e
e
79.0
8.9
11.8
928
613
1.51
5.89
e
9.2
4.4
35.1
17.3
28.1
710
495
1.43
2.20
48.58
2.0
5.0
34.1
16.3
35.0
800
636
1.26
3.99
14.68
12.3
16.2
33.7
12.6
20.1
634
473
1.34
2.12
9.71
0.4
1.8
44.5
18.6
24.9
601
455
1.32
1.97
9.11
1.4
1.9
46.6
18.6
24.4
578
451
1.28
2.22
70.30
Biopolymersa (%) Aldrich HA Suwannee River HA Suwannee River FA South Caboolture WWTP influent South Caboolture secondary effluent Bundamba secondary effluent Bundamba ROF Bundamba ROC
MWd Mne MW/ (g/ (g/ Mnf mol) mol)
SUVA-HS 254 nmg (L/ (mg∙m))
DOCh (mgC/L)
NB: Remaining DOC fraction was non-chromatographic DOC which was retained on the column. a Protein biopolymers. b Polysaccharides and aminosugars biopolymers. c Humic acid breakdown products. d Weight-averaged molecular weight of humic substances. e Number-averaged molecular weight of humic substances. f Polydispersity of humic substances. g Specific UV absorbance of humic substances at 254 nm. h DOC concentration in the studied wastewater and advanced water treatment plants.
MW and SUVA for all reference and wastewater derived DOC samples was studied (Fig. 6). The majority of studies have focused on non-polar micropollutants, particularly PAHs. While nonylphenol has a high log KOW value, it also contains a bipolar functional group, allowing it to interact with DOC through hydrogen bonding in addition to Van der Waals forces.
A weakly positive relationship was observed between KDOC and MW (Fig. 6A). Chin et al. (1997) found a strong positive relationship between the increasing MW and KDOC for pyrene and suggested that the additional aromatic functional groups in the larger DOC molecules contributed to stronger sorption. Hur and Schlautman (2003) also observed a similar relationship between KDOC and MW for pyrene, but warned that partitioning
Fig. 6 e Relationship between dissolved organic carbonewater partition coefficients (KDOC) for nonylphenol and A) weightaveraged molecular weight (MW) and B) specific UV absorbance (SUVA) of humic substances for the studied DOC.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 2 7 e4 2 3 7
was not only dependent on physical properties but also on structure and origin. Given the different sources of the reference and wastewater derived DOC this may explain the weak relationship observed in the current study. Further, the fact that the measured MW only represents the humic substance fraction may also contribute to the weak relationship. A moderately positive relationship was observed between KDOC and SUVA (Fig. 6B). Gauthier et al. (1987) found increasing organic carbon aromaticity led to increased interaction with pyrene, and suggested that this was due to increased polarizability of the organic matter. Increased polarizability can increase non-specific molecular interactions through induced dipole interactions (Schwarzenbach et al., 2003) and this may contribute to stronger partitioning for hydrophobic compounds. A correlation between aromaticity and KDOC for reference DOC has also been observed in several other studies (e.g. Chin et al., 1997; Perminova et al., 1999). In contrast, Carmosini and Lee (2008) found no relationship between aromaticity and KDOC for fluorotelomer alcohols with both reference and wastewater derived DOC. Therefore, similar to MW, the variability may be related to the studied DOC and micropollutant. While Suwannee River FA was smaller and less aromatic than the other reference DOC, Fig. 6 cannot fully explain why its sorption capacity was so similar to wastewater derived DOC for nonylphenol. To improve understanding and prediction of micropollutant sorption to reference and wastewater derived DOC in future studies, polyparameter linear free energy relationships (pp-LFER) can be applied. pp-LFERs can take into account specific and non-specific interactions, as well as cavity formation in DOC, and have been successfully applied by Niederer et al. (2007) to predict natural organic matter-air and -water partition coefficients.
4.
Conclusions
The fate of micropollutants in the aquatic environment and engineered systems can be influenced by the properties of DOC. Within the literature the majority of studies focus on reference or natural DOC, with little known regarding micropollutant interaction with wastewater derived DOC. Given the different properties, reference DOC, particularly Aldrich HA, was not an appropriate surrogate for wastewater derived DOC. This is because KDOC measured using reference DOC will underestimate the freely dissolved and thus bioavailable fraction of moderately hydrophobic micropollutants (log KOW > 4) in water recycling or water bodies receiving significant wastewater effluent discharges. These findings also have relevance to other wastewater applications including use of biosolids in agriculture. As minimal sorption of micropollutants to wastewater derived DOC is expected this may lead to more sorption to biosolids and thus higher micropollutant release during land application than predicted. This study also illustrated the importance of DOC concentration for micropollutant fate, with micropollutants present in secondary treated effluent expected to be more bioavailable than in DOC rich waste streams, such as ROC. Suwannee River FA had a similar KDOC to wastewater derived DOC for nonylphenol, but further research is required to understand its
4235
sorption capacity and interaction with other micropollutants before it can be used as a model for wastewater derived DOC.
Acknowledgements The National Research Centre for Environmental Toxicology (Entox) is a joint venture of The University of Queensland and Queensland Health Forensic and Scientific Services (QHFSS). This study was supported under the Australian Research Council (ARC) Linkage Project funding scheme (LP100200276) with industry partners WaterSecure, Water Quality Research Australia Limited (WQRA) and Veolia Water Australia. Julien Reungoat (AWMC, UQ) is thanked for sample collection and Ben Mewburn and Sibylle Rutishauser (Entox, UQ) are acknowledged for laboratory assistance. Jo¨rg Drewes (Colorado School of Mines) is thanked for helpful discussions, while Yvan Poussade (Veolia Water Australia) and Cedric Robillot (WaterSecure) are acknowledged for providing access to the Bundamba AWTP, as well as useful discussions.
references
Arditsoglou, A., Voutsa, D., 2010. Partitioning of endocrine disrupting compounds in inland waters and wastewaters discharged into the coastal area of Thessaloniki, Northern Greece. Environmental Science and Pollution Research 17, 529e538. Barret, M., Patureau, D., Latrille, E., Carre`re, H., 2010. A threecompartment model for micropollutants sorption in sludge: methodological approach and insights. Water Research 44, 616e624. Brooks, B., Riley, T., Taylor, R., 2006. Water quality of effluentdominated ecosystems: ecotoxicological, hydrological, and management considerations. Hydrobiologia 556, 365e379. Burgess, R.M., Pelletier, M.C., Gundersen, J.L., Perron, M.M., Ryba, S.A., 2005. Effects of different forms of organic carbon on the partitioning and bioavailability of nonylphenol. Environmental Toxicology and Chemistry 24, 1609e1617. Carmosini, N., Lee, L.S., 2009. Ciprofloxacin sorption by dissolved organic carbon from reference and bio-waste materials. Chemosphere 77, 813e820. Carmosini, N., Lee, L.S., 2008. Partitioning of fluorotelomer alcohols to octanol and different sources of dissolved organic carbon. Environmental Science and Technology 42, 6559e6565. Chefetz, B., Xing, B., 2009. Relative role of aliphatic and aromatic moieties as sorption domains for organic compounds: a review. Environmental Science and Technology 43, 1680e1688. Chin, Y.-P., Aiken, G., Danielsen, K.M., 1997. Binding of pyrene to aquatic and commercial humic substances: the role of molecular weight and aromaticity. Environmental Science and Technology 31, 1630e1635. Chin, Y.-P., Aiken, G., O’Loughlin, E., 1994. Molecular weight, polydispersity and spectroscopic properties of aquatic humic substances. Environmental Science and Technology 28, 1853e1858. Chiou, C., Malcolm, R.L., Brinton, T.I., Kile, D.E., 1986. Water solubility enhancement of some organic pollutants and pesticides by dissolved humic and fulvic acids. Environmental Science and Technology 20, 502e508. Ciputra, S., Antony, A., Phillips, R., Richardson, D., Leslie, G., 2010. Comparison of treatment options for removal of recalcitrant dissolved organic matter from paper mill effluent. Chemosphere 81, 86e91.
4236
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 2 7 e4 2 3 7
Drewes, J.E., Croue´, J.P., 2002. New approaches for structural characterization of organic matter in drinking water and wastewater effluents. Water Science and Technology: Water Supply 2, 1e10. Durjava, M.K., ter Laak, T.L., Hermens, J.L.M., Struijs, J., 2007. Distribution of PAHs and PCBs to dissolved organic matter: high distribution coefficients with consequences for environmental fate modeling. Chemosphere 67, 990e997. Gauthier, T.D., Seitz, W.R., Grant, C.L., 1987. Effects of structural and compositional variations of dissolved humic materials on pyrene Koc values. Environmental Science and Technology 21, 243e248. Hawker, D.W., Cumming, J.L., Neale, P.A., Bartkow, M.E., Escher, B.I., 2011. A screening level fate model of organic contaminants from advanced water treatment in a potable water supply reservoir. Water Research 45, 768e780. Huang, X., Lee, L.S., 2001. Effects of dissolved organic matter from animal waste effluent on chlorpyrifos sorption by soils. Journal of Environmental Quality 30, 1258e1265. Huber, S.A., Balz, A., Abert, M., Pronk, W., 2011. Characterisation of aquatic humic and non-humic matter with size-exclusion chromatography - organic carbon detection - organic nitrogen detection (LC-OCD-OND). Water Research 45, 879e885. Hur, J., Schlautman, M.A., 2003. Using selected operational descriptors to examine the heterogeneity within a bulk humic substance. Environmental Science and Technology 37, 880e887. Ilani, T., Schulz, E., Chefetz, B., 2005. Interactions of organic compounds with wastewater dissolved organic matter: role of hydrophobic fractions. Journal of Environmental Quality 34, 552e562. Imai, A., Fukushima, T., Matsushige, K., Kim, Y.-H., Choi, K., 2002. Characterization of dissolved organic matter in effluents from wastewater treatment plants. Water Research 36, 859e870. Jobling, S., Nolan, M., Tyler, C.R., Brighty, G., Sumpter, J.P., 1998. Widespread sexual disruption in wild fish. Environmental Science and Technology 32, 2498e2506. Katsoyiannis, A., Samara, C., 2007. The fate of dissolved organic carbon (DOC) in the wastewater treatment process and its importance in the removal of wastewater contaminants. Environmental Science and Pollution Research 14, 284e292. Kim, H.-J., Lee, D.S., Kwon, J.-H., 2010. Sorption of benzimidazole anthelmintics to dissolved organic matter surrogates and sewage sludge. Chemosphere 80, 256e262. Kim, S.-J., Kwon, J.-H., 2010. Determination of partition coefficients for selected PAHs between water and dissolved organic matter. CLEAN e Soil, Air, Water 38, 797e802. 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, 1999-2000: a national reconnaissance. Environmental Science and Technology 36, 1202e1211. Kwon, J.-H., Wuethrich, T., Mayer, P., Escher, B.I., 2009. Development of a dynamic delivery method for in vitro bioassays. Chemosphere 76, 83e90. Kwon, J.-H., Wuethrich, T., Mayer, P., Escher, B.I., 2007. Dynamic permeation method to determine partition coefficients of highly hydrophobic chemicals between poly(dimethylsiloxane) and water. Analytical Chemistry 79, 6816e6822. Lam, M.W., Mabury, S.A., 2005. Photodegradation of the pharmaceuticals atorvastatin, carbamazepine, levofloxacin, and sulfamethoxazole in natural waters. Aquatic Sciences 67, 177e188. Laor, Y., Rebhun, M., 2002. Evidence for nonlinear binding of PAHs to dissolved humic acid. Environmental Science and Technology 36, 955e961.
Latch, D.E., McNeill, K., 2006. Microheterogeneity of singlet oxygen distributions in irradiated humic acid solutions. Science 311, 1743e1747. Lindsey, M.E., Tarr, M.A., 2000. Inhibition of hydroxyl radical reaction with aromatics by dissolved natural organic matter. Environmental Science and Technology 34, 444e449. Malcolm, R.L., MacCarthy, P., 1986. Limitations in the use of commercial humic acids in water and soil research. Environmental Science and Technology 20, 904e911. McCarthy, J.F., Jimenez, B.D., 1985. Interactions between polycyclic aromatic hydrocarbons and dissolved humic material: binding and dissociation. Environmental Science and Technology 19, 1072e1076. Miao, X.-S., Bishay, F., Chen, M., Metcalfe, C.D., 2004. Occurrence of antimicrobials in the final effluents of wastewater treatment plants in Canada. Environmental Science and Technology 38, 3533e3541. Niederer, C., Schwarzenbach, R.P., Goss, K.-U., 2007. Elucidating differences in the sorption properties of 10 humic and fulvic acids for polar and nonpolar organic chemicals. Environmental Science and Technology 41, 6711e6717. Perminova, I.V., Grechishcheva, N.Y., Petrosyan, V.S., 1999. Relationships between structure and binding affinity of humic substances for polycyclic aromatic hydrocarbons: relevance of molecular descriptors. Environmental Science and Technology 33, 3781e3787. Poerschmann, J., Kopinke, F.D., 2001. Sorption of very hydrophobic organic compounds (VHOCs) on dissolved humic organic matter (DOM) 2. Measurement of sorption and application of a Flory-Huggins concept to interpret the data. Environmental Science and Technology 35, 1142e1148. Qiao, P., Farrell, A.P., 2002. Influence of dissolved humic acid on hydrophobic chemical uptake in juvenile rainbow trout. Comparative Biochemistry and Physiology Part C: Toxicology and Pharmacology 133, 575e585. Reinthaler, F.F., Posch, J., Feierl, G., Wu¨st, G., Haas, D., Ruckenbauer, G., Mascher, F., Marth, E., 2003. Antibiotic resistance of E. coli in sewage and sludge. Water Research 37, 1685e1690. Ritchie, J.D., Perdue, E.M., 2003. Proton-binding study of standard and reference fulvic acids, humic acids, and natural organic matter. Geochimica et Cosmochimica Acta 67, 85e96. Rosario-Ortiz, F.L., Snyder, S.A., Suffet, I.H., 2007. Characterization of dissolved organic matter in drinking water sources impacted by multiple tributaries. Water Research 41, 4115e4128. Sangster, J., 2006. LOGKOW: A databank of evaluated octanolwater partition coefficients (LogP) (accessed 08.11.2010). http:// logkow.cisti.nrc.ca/logkow. Schwarzenbach, R.P., Escher, B.I., Fenner, K., Hofstetter, T.B., Johnson, C.A., von Gunten, U., Wehrli, B., 2006. The challenge of micropollutants in aquatic systems. Science 313, 1072e1077. Schwarzenbach, R.P., Gschwend, P.W., Imboden, D.M., 2003. Environmental Organic Chemistry. John Wiley & Sons, Inc, Hobeken. Schwarzenbach, R.P., Westall, J., 1981. Transport of nonpolar organic compounds from surface water to groundwater. Laboratory sorption studies. Environmental Science and Technology 15, 1360e1367. Seol, Y., Lee, L.S., 2000. Effect of dissolved organic matter in treated effluents on sorption of atrazine and prometryn by soils. Soil Science Society of America Journal 64, 1976e1983. Shon, H.K., Vigneswaran, S., Snyder, S.A., 2006. Effluent organic matter (EfOM) in wastewater: constituents, effects, and treatment. Critical Reviews in Environmental Science and Technology 36, 327e374.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 2 7 e4 2 3 7
ter Laak, T.L., Busser, F.J.M., Hermens, J.L.M., 2008. Poly(dimethylsiloxane) as passive sampler material for hydrophobic chemicals: effect of chemical properties and sampler characteristics on partitioning and equilibration times. Analytical Chemistry 80, 3859e3866. ter Laak, T.L., Durjava, M., Struijs, J., Hermens, J.L.M., 2005. Solid phase dosing and sampling technique to determine partition coefficients of hydrophobic chemicals in complex matrixes. Environmental Science and Technology 39, 3736e3742. US EPA, 2008. Estimation Programs Interface Suite for Microsoft Windows, v4.0. United States Environmental Protection Agency, Washington, DC, USA. van der Voet, J. (2008) Passive sampling of polar compounds: development of a performance reference system, MSc Thesis, Swiss Federal Institute of Aquatic Science and Technology (Eawag), Du¨bendorf.
4237
Watkinson, A.J., Murby, E.J., Costanzo, S.D., 2007. Removal of antibiotics in conventional and advanced wastewater treatment: implications for environmental discharge and wastewater recycling. Water Research 41, 4164e4176. Yamamoto, H., Liljestrand, H.M., Shimizu, Y., Morita, M., 2003. Effects of physical-chemical characteristics on the sorption of selected endocrine disruptors by dissolved organic matter surrogates. Environmental Science and Technology 37, 2646e2657. Ying, G.-G., Kookana, R.S., Kolpin, D.W., 2009. Occurrence and removal of pharmaceutically active compounds in sewage treatment plants with different technologies. Journal of Environmental Monitoring 11, 1498e1505. Yoon, Y., Ryu, J., Oh, J., Choi, B.-G., Snyder, S.A., 2010. Occurrence of endocrine disrupting compounds, pharmaceuticals, and personal care products in the Han River (Seoul, South Korea). Science of the Total Environment 408, 636e643.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 3 8 e4 2 4 7
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Bioanalytical tools for the evaluation of organic micropollutants during sewage treatment, water recycling and drinking water generation Miroslava Macova a, Simon Toze b,d, Leonie Hodgers b, Jochen F. Mueller a, Michael Bartkow c, Beate I. Escher a,* a
The University of Queensland, National Research Centre for Environmental Toxicology (Entox), 39 Kessels Rd, Brisbane, Qld 4108, Australia CSIRO, Water for a Healthy Country, Queensland Ecosciences Precinct, Dutton Park, Qld 4102, Australia c Queensland Bulk Water Supply Authority trading as Seqwater, 240 Margaret St, Brisbane City, Qld 4000, Australia d The University of Queensland, School of Population Health, Herston Rd, Herston, Brisbane, Qld 4006, Australia b
article info
abstract
Article history:
A bioanalytical test battery was used for monitoring organic micropollutants across an
Received 15 December 2010
indirect potable reuse scheme testing sites across the complete water cycle from sewage to
Received in revised form
drinking water to assess the efficacy of different treatment barriers. The indirect potable
20 March 2011
reuse scheme consists of seven treatment barriers: (1) source control, (2) wastewater
Accepted 30 May 2011
treatment plant, (3) microfiltration, (4) reverse osmosis, (5) advanced oxidation, (6) natural
Available online 7 June 2011
environment in a reservoir and (7) drinking water treatment plant. Bioanalytical results provide complementary information to chemical analysis on the sum of micropollutants
Keywords:
acting together in mixtures. Six endpoints targeting the groups of chemicals with modes of
Bioassays
toxic action of particular relevance for human and environmental health were included in
In-vitro
the evaluation: genotoxicity, estrogenicity (endocrine disruption), neurotoxicity, phyto-
Treatment barriers
toxicity, dioxin-like activity and non-specific cell toxicity. The toxicity of water samples
Micropollutants
was expressed as toxic equivalent concentrations (TEQ), a measure that translates the
Toxicity
effect of the mixtures of unknown and potentially unidentified chemicals in a water
Water recycling
sample to the effect that a known reference compound would cause. For each bioassay
Indirect potable reuse
a different representative reference compound was selected. In this study, the TEQ concept was applied for the first time to the umuC test indicative of genotoxicity using 4nitroquinoline as the reference compound for direct genotoxicity and benzo[a]pyrene for genotoxicity after metabolic activation. The TEQ were observed to decrease across the seven treatment barriers in all six selected bioassays. Each bioassay showed a differentiated picture representative for a different group of chemicals and their mixture effect. The TEQ of the samples across the seven barriers were in the same order of magnitude as seen during previous individual studies in wastewater and advanced water treatment plants and reservoirs. For the first time a benchmarking was performed that allows direct comparison of different treatment technologies and covers several orders of magnitude of TEQ from highly contaminated sewage to drinking water with TEQ close or below the limit of detection. Detection limits of the bioassays were decreased in comparison to earlier studies by optimizing sample preparation and test protocols, and were comparable to or lower than the quantification limits of the routine chemical analysis, which allowed monitoring of the presence and
* Corresponding author. Tel.: þ61 7 3274 9180; fax: þ61 7 3274 9003. E-mail address:
[email protected] (B.I. Escher). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.05.032
4239
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 3 8 e4 2 4 7
removal of micropollutants post Barrier 2 and in drinking water. The results obtained by bioanalytical tools were reproducible, robust and consistent with previous studies assessing the effectiveness of the wastewater and advanced water treatment plants. The results of this study indicate that bioanalytical results expressed as TEQ are useful to assess removal efficiency of micropollutants throughout all treatment steps of water recycling. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
The indirect potable reuse scheme (IPR, http://www. westerncorridor.com.au) investigated in this study is the largest potable water recycling scheme in Australia and one of the largest in the Southern Hemisphere (Freeman et al., 2008; Traves et al., 2008). The scheme consists of seven treatment barriers: 1 e source control; 2 e wastewater treatment plant (WWTP); 3 e microfiltration; 4 e reverse osmosis; 5 e advanced oxidation (combining hydrogen peroxide and UV irradiation); 6 e natural environment; and 7 e drinking water treatment plant. It takes treated wastewater from five of the largest wastewater treatment plants in the greater Brisbane area and treats this water to potable standards via three advanced water treatment plants. The resulting purified recycled water (PRW) can then be piped to Lake Wivenhoe, the largest of the freshwater reservoirs in South East Queensland (SEQ). This reservoir supplies greater than 60% of the freshwater resources for the city of Brisbane. The production of the PRW is based on international experiences of other recycling plants such as Water Factory 21 in California, Singapore’s NEWater, and the Torreele project in Belgium. All of these schemes use a similar treatment process of water treatment plants followed by membrane and reverse osmosis filtration and at least UV disinfection. The Torreele and Water Factory 21 schemes then add the purified recycled water to a local aquifer prior to recovery and addition to the drinking water system. Singapore’s NEWater is the same as the indirect potable reuse scheme in SEQ in that the purified water is added to a reservoir. The water produced in the studied PRW scheme meets potable standards, but is presently only used for industrial purposes and has not yet been introduced to Lake Wivenhoe. Supplementation of drinking water storage reservoirs is envisaged only after the combined level of water in the three major SEQ reservoirs falls below 40%. Water at all stages of the treatment process is subject to quality monitoring to assess the efficacy of the treatment barriers and to ensure the water meets health and safety requirements. A number of organic and inorganic micropollutants have been monitored during the last two years in PRW (Queensland Water Commission, 2009; WaterSecure, 2010; Hawker et al., 2011). Toxicity testing may provide complementary information to chemical analysis on the sum of micropollutants present during water treatment. Therefore, a bioanalytical “mode of action” test battery, developed or optimized at Entox in collaboration with colleagues from the Swiss Federal Institute of Aquatic Science and Technology, has been included in water recycling projects to support water quality assessment. Bioanalytical techniques have been selected to target the
groups of chemicals of particular relevance for human and environmental health including genotoxicity, endocrine activity, neurotoxicity, dioxin-like activity and non-specific cell toxicity (Escher et al., 2008, 2009; Macova et al., 2010). For better comparability, the results in all toxicity tests were expressed as toxic equivalent concentrations that give an account of the concentration of a reference chemical that would elicit the same effect as the sample does (Villeneuve et al., 2000). The TEQ concept was previously established for five of the bioassays used (Escher et al., 2008; Macova et al., 2010) and was newly developed for the umuC assay for genotoxicity (International Organization for Standardization, 2000) in the present study. The goal of this study was to evaluate the applicability of this bioanalytical test battery for monitoring the micropollutants across all seven barriers of the indirect potable reuse scheme and to obtain a benchmark of water quality that may serve in the future for classification of water samples from emerging technologies and for alternative source water such as stormwater and bore water. To achieve these goals, the existing and validated bioassay test battery was further optimized to achieve lower detection limits and a testing strategy was developed to allow the assessment of samples with a wide range of chemical contamination level.
2.
Materials and methods
2.1.
Samples and sites
Grab samples were collected at 21 sites across the seven barriers of the indirect potable reuse scheme (Table 1, for a map see http://www.westerncorridor.com.au/resources/ factsheets, select South East Queensland Water Grid): raw wastewater and tractor effluent at the Oxley Creek wastewater treatment plant (WWTP) (Barrier 1e2), product water from microfiltration, reverse osmosis and advanced oxidation at the Bundamba advanced water treatment plant (AWTP), PRW from the Bundamba off-take, Lowood and Caboonbah Pipeline (Barrier 3e5), water from the Swanbank Power Station lake, Lake Wivenhoe and mid-Brisbane river representing the natural environment (Barrier 6), as well as samples from the inlet and outlet of the Mt. Crosby drinking water treatment plant (DWTP) and the drinking water distribution system (DWS) (Barrier 7). While PRW has not been introduced into the drinking water reservoir Lake Wivenhoe, the Power Station lake is a reservoir that receives PRW. Sampling was complemented by three additional samples collected at Caboolture WWTP and Caboolture enhanced
4240
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 3 8 e4 2 4 7
Table 1 e Sample description. Barrier Barrier 1e2
Barrier 3e5
Barrier 6a
Barrier 7
Others
Sample location
Sampling date
Sample volume (L)
Oxley Ck WWTP Inlet Oxley Ck WWTP Activated Sludge Oxley Ck WWTP post Clarifiers Oxley Ck WWTP post UV Bundamba AWTP pre MF (Inlet) Bundamba AWTP post MF Bundamba AWTP post RO Bundamba AWTP post AO Bundamba AWTP RO concentrate PRW pipeline (Bundamba off-take) PRW pipeline (Lowood) PRW pipeline (Caboonbah) Power Station lake Lake Wivenhoe e Logan’s Inlet Lake Wivenhoe e Dam Wall Mid-Brisbane e Lowood Mid-Brisbane e Burton’s Bridge Mt Crosby DWTP Intake (raw) Mt Crosby DWTP Outlet Drinking Water System e mid way on distribution line Drinking Water System e towards end of distribution line South Caboolture WWTP Influent (raw) South Caboolture EWTP Influent (Effluent from WWTP) South Caboolture EWTP Effluent Bottled Water type 1b Bottled water type 2b Hinze Dam (Lake Advancetown) Gold Coast Water Distribution system
28.10.2009 28.10.2009 28.10.2009 28.10.2009 29.10.2009 29.10.2009 29.10.2009 29.10.2009 29.10.2009 08.12.2009 08.12.2009 08.12.2009 08.12.2009 13.10.2009 13.10.2009 13.10.2009 13.10.2009 13.10.2009 13.10.2009 19.11.2009 19.11.2009 22.10.2009 22.10.2009 22.10.2009 08.12.2009 08.12.2009 28.10.2009 06.11.2009
0.5 1.0 2.0 2.0 2.0 4.0 4.0 4.0 1.0 4.0 4.0 4.0 4.0 4.0 4.0 4.0 4.0 4.0 4.0 4.0 4.0 0.5 2.0 4.0 4.0 4.0 4.0 4.0
a There is a discontinuation between the AWTPs and the natural environment. The resulting PRW is currently used for industrial purposes and has not been reintroduced to the Lake Wivenhoe at this time but is supplementing the Power Station lake. b Purchased in a Brisbane supermarket on 29.10.2009.
water treatment plant (EWTP) fed with the Caboolture WWTP effluent, to compare the treatment/removal efficiency of the membrane processes with ozonation followed by biologically activated carbon. Two samples were collected at the Gold Coast Hinze Dam (Lake Advancetown) and water distribution system to compare drinking water sourced from different catchments. Two types of bottled water were purchased at a local supermarket to compare with the quality of the purified recycled water (Table 1). Based on previous studies (Macova et al., 2010), collected sample volumes ranged from 0.5 Le4 L depending on the expected toxicity of the samples (Table 1). Different sampling volume allows us to test the sample extracts across the bioanalytical test battery without pre-dilution of the extract and to achieve low limits of detection in the assay. Samples were kept on ice during transport and until processing. Samples were extracted by solid phase extraction within 24 h of collection.
2.2.
Solid phase extraction
Immediately after sampling, 1 mL of 0.1% sodium thiosulphate was added per 1 L of sample to neutralise the presence of chlorine and concentrated HCl (36%) was added to a final concentration of 5 mM for preservation. It was demonstrated in earlier work that a pharmaceutical cocktail in a wastewater matrix had highest recoveries for HLB at pH 3 (Escher et al., 2005).
Samples were extracted using 1 g OASIS HLB solid phase material in 20 mL cartridges (Waters, Australia) following filtration with a glass fibre filter. After conditioning the cartridges with 10 mL methanol and 20 mL of 5 mM HCl in MilliQ water, a known volume of sample was percolated under vacuum. Cartridges were sealed and kept at 20 C until elution with the solvent mixture. Immediately before elution, the cartridges were dried for 2e3 h under vacuum and were eluted with 10 mL methanol and 10 mL hexane:acetone (1:1). All eluates were evaporated to approximately 1 mL under purified nitrogen gas and were solvent exchanged to methanol at a final volume of 500 mL.
2.3.
Bioanalytical tools
An overview on the bioanalytical test battery comprising six endpoints is given in Table 2. The phytotoxicity assay with the green algae Pseudokirchneriella subcapitata was performed as described by (Escher et al., 2008). The detailed methodology of the remaining bioanalytical techniques was described in (Macova et al., 2010). Bioanalytical results were reported in terms of toxic equivalent concentrations (TEQ) (Villeneuve et al., 2000; Escher et al., 2008; Macova et al., 2010) using a corresponding reference compound representing the group of targeted chemicals in a given assay (Table 2). In previous work, we had not used the TEQ concept for the umuC assay for genotoxicity. Here we tested a series of potential reference compounds (see
4241
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 3 8 e4 2 4 7
Table 2 e Overview of the bioanalytical test battery (adapted from Macova et al. (2010)). Assay
Targeted chemicals
Bioluminescence inhibition in Vibrio fischeri Neurotoxicity e AChE Phytotoxicity e Max-I-PAM Estrogenicity e E-SCREEN Ah-Receptor e AhR-CAFLUX
Genotoxicity e UmuC
Reference compound
Result expression
Literature reference (International Organization for Standardization, 1998; Johnson, 2005; Farre´ et al., 2006) (Ellman et al., 1961; Hamers et al., 2000) (Schreiber et al., 2002; Schreiber et al., 2007) (Soto et al., 1995; Korner et al., 1999) (Nagy et al., 2002; Zhao and Denison, 2004)
All chemicals
Virtual baseline toxicant; Phenola
Baseline toxicity equivalent concentrations (baseline-TEQ)
Organophosphates and carbamate insecticides Triazine and phenylurea herbicides Estrogens, estrogenic industrial chemicals Polychlorinated dibenzodioxins/furans and biphenyls, polycyclic aromatic hydrocarbons Chlorinated byproducts, aromatic amines, polycyclic aromatic hydrocarbons
Parathion
Parathion equivalent concentrations (PTEQ) Diuron equivalent concentrations (DEQ) Estradiol equivalent concentrations (EEQ); TCDD equivalent concentrations (TCDDEQ)
Diuron 17b-Estradiol (E2) 2,3,7,8Tetrachlorodibenzo dioxin (TCDD) (S9) 4-nitroquinoline -N-oxide (4NQO) (þS9) Benzo[a]pyrene (BaP)
4NQO and BaP equivalent concentrations (4NQOEQ and BaPEQ)
(Oda et al., 1985; Reifferscheid et al., 1991; International Organization for Standardization, 2000)
a Phenol was only used as positive control, not as reference compound.
Supplementary Information) and validated 4NQO as the reference compound without prior metabolic activation by a rat liver S9 enzyme extract and benzo[a]pyrene as the reference compound for genotoxicity after metabolic activation.
2.4.
QA/QC
For quality and assurance purposes, all samples were collected in duplicates and were extracted and analysed across the bioanalytical test battery to assess the repeatability of the SPE and the bioassay. Both replicates of the sample extract were tested in duplicates or triplicates per run depending on the assay, with the standard error typically between 10 and 15%. To assess the day-to-day variation of the assays, a second replicate of the sample extract was analysed in bioassays on a different day than the first replicate. Final TEQs were expressed as the average standard deviation of two independent replicates reflecting day-to-day variation of the assays. Another QA/QC parameter of the bioanalytical results was the long term record of the EC50 values of the reference compounds. If the EC50 value of the reference compound of a given run varied more than three times the standard deviation of the long term average, the results were not included and the run was repeated. To assess any effect associated with the extraction process or with the solvent, MilliQ water was processed the same way as the samples and assessed in all bioassays as a procedural blank. There are two aspects that influence the limit of detection: the variability of the response in a given assay assessed by the concentration-effect curve of the reference compound and the maximum enrichment of the sample that could be achieved in the assay. The detection limits of all assays were defined as three times standard deviation of the control response. For example, if the average of effect of the control was 2.3 3.1%, then the LOD was assigned to the concentration of sample that produced 3 3.1% ¼ 9.3% effect. For the bacteria Vibrio fischeri the control diluted in the medium was used to derive this standard deviation, for the other five
bioassays the response using the lowest concentration of the reference compound that induced an effect significantly different from the control was used. Typically, the thus derived standard deviation was in the range of an effect level of 8e10% effect. Since four different volumes of the samples were extracted depending on the expected toxicity (0.5, 1, 2 and 4 L), there are four different LODs for each assay summarised in Table 3. Analysis of variance (ANOVA, GraphPad Prism, San Diego, CA, USA) was used to analyse statistical differences among the average TEQs of the samples.
3.
Results and discussion
3.1. Baseline toxicity e bioluminescence inhibition in Vibrio fischeri The baselineeTEQs were observed to decrease by 94% after treatment in the Oxley Creek WWTP (Barrier 2), from 25.6 mg/ L in the influent to 1.26 mg/L after activated sludge treatment with no further decrease post clarifiers or UV treatment (Table 4, Fig. 1). Microfiltration at Bundamba AWTP (Barrier 3) significantly increased baseline toxicity from 0.91 to 2.66 mg/L. This increase may be caused by chloramination that is preceeding microfiltration or by removal of particulate matter by microfiltration and release of micropollutants into the dissolved phase, in which they are bioavailable. Reverse osmosis and advanced oxidation at the Bundamba AWTP (Barrier 4e5) decreased the baseline toxicity to 0.42 and 0.12 mg/L representing 44% and 13%, respectively, of the original activity in the Bundamba AWTP inlet; the latter not being significantly different from the blank (0.08 mg/L, t-test, p ¼ 0.304). Baseline toxicity of the sample after Barrier 2 treatment of the indirect potable reuse scheme was comparable with the effluent of the Caboolture WWTP of 1.0 mg/L, representing 11% of the toxicity equivalents in the influent. Effluent from the conventional Caboolture WWTP was further treated in the Caboolture enhanced water treatment plant (EWTP) with
4242
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 3 8 e4 2 4 7
Table 3 e Limits of detection of the assays for individual sample volume. Assay
Result expression
Baseline Toxicity e Bioluminescence inhibition in Vibrio fischeri Neurotoxicity e AChE Phytotoxicity e Max-I-PAM Estrogenicity e E-SCREENa Ah-Receptor e AhR-CAFLUX Genotoxicity e UmuC
Baseline-TEQ PTEQ DEQ EEQ TCDDEQ (S9) 4NQOEQ (þS9) BaPEQ
LOD for different sample volume 0.5 L
1L
2L
4L
0.1 mg/L 0.4 mg/L 0.05 mg/L 0.08 ng/L 0.09 ng/L 0.4 mg/L 6.4 mg/L
0.05 mg/L 0.2 mg/L 0.02 mg/L 0.04 ng/L 0.05 ng/L 0.2 mg/L 3.2 mg/L
0.02 mg/L 0.1 mg/L 0.01 mg/L 0.02 ng/L 0.02 ng/L 0.1 mg/L 1.6 mg/L
0.01 mg/L 0.06 mg/L 0.005 mg/L 0.01 ng/L 0.01 ng/L 0.05 mg/L 0.8 mg/L
a If the maximum proliferation of the sample did not reach 50% of the 17b-estradiol, the sample was not classified as estrogenic, therefore EEQ was not quantified and the results were reported as below quantification limit of the assay (Soto et al., 1995).
ozonation and activated carbon treatment (van Leeuwen et al., 2003; Reungoat et al., 2010). Baseline toxicity in the final effluent was decreased to 0.56 mg/L, a level not significantly different from the blank. Results are in agreement with the previous study at the Caboolture EWTP, where the baseline toxicity was reduced throughout the enhanced treatment chain from 2.3 mg/L to 0.52 mg/L in the final effluent (Macova et al., 2010). Baseline toxicity of the Caboolture EWTP final effluent was comparable with the baseline toxicity post Barrier 4 of the IPR, reverse osmosis at Bundamba AWTP. Apart from the Mt. Crosby outlet sample, the baseline toxicity of samples collected after Barrier 5 was on average elevated by a factor of two as compared to the blank and not significantly different from the bottled water (t-test, p ¼ 0.25) (Table 4, Fig. 1). The observed increased baseline toxicity of the Mt. Crosby outlet sample as compared to the inlet of the drinking water plant was reproducible in a second sampling campaign but the level is of no concern relating to potential health impacts because they are below permissible effect levels modelled for this endpoint under the assumption that all chemical concentrations are present at or below their drinking water guideline values using methods detailed in Vermeirssen et al. (2010) and model input parameters from Hawker et al. (2011). Furthermore this effect decreased significantly in the drinking water supply pipeline and might be related to intermittent production of chlorinated disinfection byproducts or assimilable organic carbon.
3.2.
Estrogenic activity e E-SCREEN assay
The estrogenic effect of the samples, expressed as estradiol equivalent concentration (EEQ), decreased by 86% during activated sludge treatment after Barrier 2 from 3.2 ng/L in Oxley Creek WWTP influent to 0.44 ng/L in the effluent (Table 4, Fig. 1). EEQ was further reduced after the clarifiers but UV treatment did not alter the EEQ. Microfiltration at the Bundamba AWTP (Barrier 3) reduced estrogenic effect to the level below the detection limit of the assay (<0.01 ng/L). The reverse osmosis concentrate, where micropollutants are enriched by a factor of six to eight times, did show appreciable estrogenic activity, indicating that despite the EEQ being below the LOD after microfiltration, there were still residual estrogenic compounds but they were rejected by reverse osmosis. No further alteration in the estrogenicity was observed in any
sample collected after Barrier 3. Results are in agreement with the previous findings where no estrogenic response (EEQ <0.01 ng/L) was observed in any sample collected at Lake Wivenhoe (Seqwater project 2007, unpublished data). The estrogenic effects of the samples collected at Hinze Dam (Lake Advancetown), the Gold Coast Water distribution system and bottled water were also below the LOD of the assay (<0.01 ng/L for 4 L water samples). Estrogenicity of the influent to the Oxley Creek WWTP was lower than typically seen in the raw sewage samples. EEQ in the raw sewage sample from the Caboolture WWTP in a previous study ranged from 68 to 91 ng/L in three different samples collected in the course of one month (2008, unpublished results). This is comparable with the EEQ reported in the raw sewage in the Brisbane area using the same bioassay and with other bioassays for estrogenicity (reviewed in Macova et al. (2010)). The EEQ of the additional sample collected at the Caboolture WWTP was 18.5 ng/L, which is again in the lower range of what is typically seen in raw sewage. Surprisingly, the EEQ was reduced to below the LOD of the assay of 0.02 ng/L (for 2 L sample) already by the treatment in the conventional WWTP. In a previous study, this WWTP effluent, which was an influent to the enhanced treatment plant, exhibited EEQ of 6 ng/L (Macova et al., 2010). Estrogenic effect was markedly decreased by ozonation and further reduced to below LOD (<0.02 ng/L) by activated carbon treatment, key steps of the enhanced treatment chain that are effective in the removal of the estrogenic effect (Macova et al., 2010). High removal efficiency of ozonation on the estrogenic activity is consistent with literature (Lee et al., 2008; Escher et al., 2009).
3.3. assay
Neurotoxicity e acetylcholinesterase inhibition
Results of the bioassay targeting organophosphates and carbamate insecticides, expressed as parathion equivalent concentration (PTEQ), showed 78% decrease in the toxicity post Barrier 2, from 4.36 mg/L in the Oxley Creek WWTP inlet to 0.94 mg/L after activated sludge treatment, with no further decrease post clarifiers or UV treatment (Table 4, Fig. 1). Barrier 3 (microfiltration) did not alter PTEQ, while Barrier 4 (reverse osmosis) reduced PTEQ from 1.96 mg/L to the level below the LOD (<0.06 mg/L). Results are in agreement with a previous study at the Bundamba AWTP where microfiltration did not affect PTEQ (unpublished data).
Table 4 e Summary of the bioanalytical results expressed as the average ± standard deviation (sd) of two replicates. Treatment Barrier
1e2
6
7
Others
Bioluminescence Inhibition Baseline e TEQ (mg/L)
Oxley Ck WWTP Inlet Oxley Ck WWTP Activated Sludge Oxley Ck WWTP Post Clarifiers Oxley Ck WWTP Post UV Bundamba AWTP Pre MF (Inlet) Bundamba AWTP Post MF Bundamba AWTP post RO Bundamba AWTP Post AO Bundamba AWTP RO Concentrate PRW Pipeline (Bundamba off-take) PRW Pipeline (Lowood) PRW Pipeline (Caboonbah) Power Station Lake Lake Wivenhoe e Logan’s Inlet Lake Wivenhoe - Dam Wall Mid-Brisbane e Lowood Mid-Brisbane - Burton’s Bridge Mt Crosby DWTP Intake (raw) Mt Crosby DWTP Outlet DWS e Mid Way on Distribution Line SEW e Towards End of Distribution Line South Caboolture WWTP Influent South Caboolture EWTP Influent (Effluent from WWTP) South Caboolture EWTP Effluent Bottled Water type 1 Bottled Water Type 2 Hinze Dam (Lake Advancetown) Gold Coast Water Distribution System MilliQ Water (Negative Control)
AChE
I-PAM
E-SCREEN AhR e CAFLUX
umuC S9a
umuC þS9b
PTEQ (mg/L) DEQ (mg/L) EEQ (ng/L) TCDDEQ (ng/L) 4NQOEQ (mg/L) BaPEQ (mg/L)
Avg
sd
Avg
sd
Avg
sd
Avg
sd
Avg
sd
Avg
sd
Avg
sd
25.6 1.26 1.26 1.25 0.91 2.66 0.42 0.12 3.25 0.65 0.27 0.83 0.19 0.16 0.14 0.14 0.17 0.23 1.68 0.17 0.34 9.17 1.00 0.56 0.18 0.19 0.20 0.70 0.08
17.4 0.09 0.47 0.44 0.22 0.58 0.07 0.04 1.04 0.15 0.18 0.18 0.21 0.11 0.14 0.13 0.13 0.12 0.64 0.13 0.20 4.56 0.52 0.44 0.03 0.01 0.05 0.27 0.01
4.36 0.94 1.00 1.00 1.50 1.96 <0.06 <0.06 7.45 <0.06 <0.06 <0.06 0.17 0.16 0.11 0.19 0.24 0.21 0.28 0.68 0.31 5.98 0.67 0.10 <0.06 <0.06 0.12 0.30 <0.06
0.12 0.56 0.41 0.34 0.57 0.34
2.15 0.98 1.36 1.31 0.26 0.20 0.04 0.05 0.8 0.08 0.1 0.08 0.05 0.02 0.02 0.02 0.01 0.01 0.05 0.08 0.06 0.26 0.09 0.11 0.06 0.05 0.04 0.07 0.04
0.21 0.18 0.35 0.34 0.06 0.04 0.01 0.01 0.11 0.02 0.06 0.04 0.01 0.01 0.01 0.01 0.003 0.003 0.01 0.01 0.002 0.05 0.02 0.06 0.01 0.01 0.01 0.01 0.01
3.15 0.44 0.25 0.31 0.34 <0.01 <0.01 <0.01 0.75 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 18.53 <0.02 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01
0.2 0.2 0.1 0.1 0.2
1.13 0.85 0.77 0.56 1.15 0.33 0.11 0.08 2.00 0.11 0.11 0.15 0.17 0.15 0.15 0.19 0.19 0.17 0.17 0.19 0.28 1.80 0.21 0.16 0.15 0.13 0.17 0.20 0.20
0.51 0.43 0.13 0.36 0.54 0.18 0.08 0.06 0.73 0.08 0.08 0.14 0.08 0.02 0.05 0.02 0.01 0.07 0.06 0.05 0.06 0.54 0.04 0.04 0.04 0.01 0.01 0.02 0.03
1.52 0.25 0.19 0.17 0.24 0.44 <0.05 <0.05 3.07 <0.05 <0.05 <0.05 <0.05 <0.05 <0.05 <0.05 <0.05 <0.05 <0.05 <0.05 <0.05 0.48 0.21 <0.05 <0.05 <0.05 <0.05 0.06 <0.05
0.70
32.19 3.42 4.89 4.87 5.95 10.31 <0.8 <0.8 32.04 <0.8 <0.8 <0.8 <0.8 <0.8 <0.8 <0.8 <0.8 <0.8 <0.8 <0.8 <0.8 < 6.4 2.61 <0.8 <0.8 <0.8 <0.8 <0.8 <0.8
11.09
2.87
0.03 0.04 0.03 0.03 0.03 0.03 0.03 0.08 0.32 1.10 0.25 0.02
0.03 0.09
0.7
9.3
0.03 0.05 0.05 0.08
2.04
0.06
0.004
0.72 0.79 1.41 2.60
6.27
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 3 8 e4 2 4 7
3e5
Sample Location
0.24
a (S9) without exogenous metabolic activation. b (þS9) with exogenous metabolic activation.
4243
4244
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 3 8 e4 2 4 7
Fig. 1 e Reduction of TEQ across seven treatment barriers in selected bioassays expressed relative to the TEQ of the Oxley Creek WWTP Inlet. Data represent the average of two replicates. Error bars indicate the standard deviation. Missing bars represent data below LOD. *There is a discontinuation between the AWTPs and the natural environment. The resulting purified recycled water is currently used for industrial purposes (represented by Power Station Lake) and has not been reintroduced to the Lake Wivenhoe at the time of this publication.
PTEQ of the samples collected across Barrier 6 and 7 ranged between 0.11 and 0.31 mg/L, with a slight increase in the drinking water system (mid-way along the distribution line) not exceeding 0.7 mg/L but clearly above the blanks. This increase of PTEQ in the drinking water pipeline might be an artefact caused by disinfection with chlorine. Approximately 0.3 mg/L Cl equivalents are used in the drinking water plant for final disinfection but this dose is increased up to 0.8 mg/L in the distribution system (personal communication, Water Grid Manager Technical Committee, 2010). While chlorine should have been quenched prior to extraction of the sample, there might be residual chlorine still present. This should not affect cell-based bioassays but the AChE inhibition assay is done with purified enzyme, which can be denatured or looses activity. Due to the nature of the assay, a non-specific denaturation of the enzyme cannot be differentiated from a specific inhibition. This issue will require further investigation and it is advisable that the AChE inhibition assay be replaced in the future by a cell-based assay indicative of the same endpoint. PTEQ of the samples collected at Hinze Dam (0.12 mg/L) and the Gold Coast Water distribution system (0.3 mg/L) was comparable with the PTEQ of the samples collected across Barrier 6. PTEQ of both types of bottled water was below the LOD (<0.06 mg/L for 4 L water samples). Similar to the Oxley Creek WWTP (Barrier 2), PTEQ was decreased by 89% in the Caboolture WWTP from 5.98 mg/L in the inlet to 0.67 mg/L in the WWTP outlet (Table 4). A further decrease to 0.1 mg/L was observed after treatment throughout the treatment chain of the Caboolture EWTP. Results are in
agreement with a previous study at the Caboolture EWTP, where PTEQ were significantly decreased during enhanced treatment to 0.36 mg/L (Macova et al., 2010).
3.4.
Phytotoxicity e PSII inhibition I-PAM assay
A similar decrease of the toxicity across the seven barriers was observed in the I-PAM assay targeting triazine and phenylurea herbicides, expressed as diuron equivalent concentration (DEQ). Phytotoxicity of the samples was reduced from the inlet to post Barrier 2 from 2.15 mg/L to 1.31 mg/L of DEQ (Table 4, Fig. 1). The relatively high DEQ after Barrier 2 treatment is in agreement with the literature data, showing lower treatment efficiency of biological treatment than ozonation in the removal of herbicides (Escher et al., 2009). DEQ was not altered by microfiltration (Barrier 3) but was five-fold reduced by reverse osmosis (Barrier 4) from 0.2 to 0.04 mg/L. DEQ of the samples collected after Barrier 5 were not significantly different from the blank (0.04 mg/L, t-test, p ¼ 0.61). Similar results were observed in the previous study at Lake Wivenhoe with the DEQ levels of 0.020 mg/L and 0.025 mg/L at the dam wall and Logan’s Inlet respectively (Seqwater project 2007, unpublished data). DEQ was also reduced by treatment in the Caboolture WWTP from 0.26 to 0.09 mg/L. No further reduction in DEQ was observed after enhanced treatment (Table 4). These results are not in agreement with a previous study at the Caboolture EWTP where the enhanced treatment (particularly preeozonation) reduced the DEQ effect to a level below LOD
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 3 8 e4 2 4 7
4245
(Macova et al., 2010) but since herbicide input types and levels are highly variable, which is also evident from the almost tenfold difference in WWTP inlets between Oxley WWT and Caboolture WWTP, the observed difference might reflect rather differences in input than differences in treatment efficiency. The MilliQ blank also showed an effect slightly higher (0.04 mg/L) than the LOD of the bioassay but of no environmental concern.
3.5. assay
Arylhydrocarbon receptor response e AhR-CAFLUX
In this study, sample extracts were tested in the AhR-CAFLUX assay without acid silica gel clean up, which destroys all but persistent compounds such as polychlorinated dibenzodioxins/furans and biphenyls. Therefore, the TCDDEQ reported in this study represented the sum of all compounds present in the sample that can bind to the arylhydrocarbon receptor, not only the persistent compounds. The less persistent group includes, for example, polycyclic aromatic hydrocarbons. The response of arylhydrocarbon receptor (AhR) compounds targeted in the AhR-CAFLUX assay and expressed as 2,3,7,8-Tetrachlorodibenzo-dioxin (TCDD) equivalent concentration (TCDDEQ) was reduced across the seven treatment barriers (Table 4). The first decrease was observed post Barrier 2 from an original level of 1.13 ng/L of TCDDEQ to 0.56 ng/L after UV treatment at the Oxley Creek WWTP. TCDDEQ was also decreased throughout the advanced water treatment chain in the Bundamba AWTP, from 1.15 ng/L to 0.33 ng/L after microfiltration (Barrier 3) and further to 0.11 ng/ L after reverse osmosis (Barrier 4), a level not significantly different from the blank (0.2 ng/L). In a previous study at the Bundamba WWTP, the TCDDEQ was reduced after reverse osmosis from 1.72 ng/L to 0.12 ng/L (2008, unpublished data). No alterations in the TCDDEQ were observed after Barrier 4 and the effect of all samples was not significantly different from the blank. A decrease in TCDDEQ was observed also after treatment in the South Caboolture WWTP from 1.8 ng/L in the influent to 0.21 ng/L in the WWTP effluent.
3.6.
Genotoxicity e UmuC assay
To enable detection of progenotoxins that require metabolic activation to become genotoxic, samples were tested both with and without the inclusion of rat liver supernatant fraction (S9). Response in the umuC assay was defined as the induction ratio (IR), the ratio of the sample response to the control, and an effect concentration inducing an IR of 1.5 (ECIR1.5) was deduced from linear concentration-response curves. The ECIR1.5 in umuC assay were expressed as 4nitroquinoline-oxide (4NQO) equivalent concentration (4NQOEQ) for the assay without metabolic activation (eS9) and benzo[a]pyrene (BaP) equivalent concentration (BaPEQ) for the assay with metabolic activation (þS9) (see Supplementary Information). Similar to other assays, where the TEQs were calculated as the ratio of EC50 of the corresponding reference compound to EC50 of the sample, the TEQs in umuC assay were calculated using the effective concentration ECIR1.5 that induces the induction ratio of 1.5 defined by
Fig. 2 e Removal of TEQ in the two advanced water treatment chains investigated in this study.
the International Standard Organisation (ISO) guideline as the threshold of genotoxic effect (International Organization for Standardization, 2000), providing the sample was not cytotoxic (growth < 0.5) (Table 4, Fig. 1, and Supplementary information). TEQs were calculated using the average ECIR1.5 (4NQO) of 0.008 mg/L and ECIR1.5 (BaP) of 0.12 mg/L (see Supplementary Information). Similar to all other bioassays, the genotoxic effect was significantly decreased post Barrier 2. Activated sludge at the Oxley Creek WWTP reduced the genotoxic effects both without and with metabolic activation by 90% and 87%, respectively (Table 4). No further decrease was observed post clarifiers and after UV treatment. Barrier 3 treatment (microfiltration) at the Bundamba AWTP did not significantly alter the genotoxic effect of the microfiltration feed. However, the genotoxic effect was markedly reduced by reverse osmosis treatment (Barrier 4) to a level below the detection limits of the assay (<0.05 mg/L of 4NQOEQ and <0.8 mg/L of BaPEQ). No genotoxic effect was observed in any sample collected post Barrier 4. An unexpectedly low genotoxic effect was observed in the Inlet of the Caboolture WWTP compared to the results of a previous study (2008, unpublished data) and in comparison to genotoxic effect at the Oxley Creek WWTP inlet e 0.48 mg/L of 4NQOEQ and no genotoxic effect in the assay with metabolic activation. The genotoxic effect was reduced throughout the enhanced water treatment chain of the Caboolture EWTP from 0.21 mg/L of 4NQOEQ and 2.61 mg/L of BaPEQ in the outlet of the Caboolture WWTP, to a level below the LOD at the outlet of the EWTP (<0.05 mg/L of 4NQOEQ and <0.8 mg/L of BaPEQ).
4.
Conclusions
This paper comprises the first study that monitored mixture toxicity in various bioassays across all steps of a water recycling scheme from sewage to drinking water. This will allow benchmarking of water quality levels in the future, when data from alternative source water, e.g. stormwater or bore water and from alternative treatment processes, e.g. treatment of coal seam gas water, become available. Fig. 1 represents the relative TEQ in comparison to raw sewage across one (albeit discontinuous) water cycle and subsequent environmental buffers and drinking water
4246
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 3 8 e4 2 4 7
treatment and clearly demonstrates how TEQ were reduced across the seven treatment barriers in all six selected bioassays. In all cases, the micropollutant burden was reduced by one order of magnitude or more, but to a different extent, in Barriers 2 to 5. The six bioassays showed a differentiated picture, each one representative for a different group of chemicals and their mixture effect, and the reduction in the toxicity for certain groups of chemicals could be related to their physicochemical properties and behaviour during the various treatment processes. For example, the fact that herbicides are recalcitrant to biodegradation in wastewater treatment plant but rejected by reverse osmosis and well degraded by ozonation is reflected in the DEQ pattern seen over the treatment chain. Another example refers to the EEQ that are brought down to low level by biodegradation already, so they were below detection limits already after Barrier 3. The results obtained in this study are also useful to compare different treatment options. Fig. 2 compares the removal of TEQ of the advanced water treatment (Barrier 3e5 with membrane filtration, reverse osmosis and UV/H2O2) with an alternative advanced treatment chain composed of ozonation and biological activated carbon filtration (BAC). Both chains performed equally well for the removal of non-specific toxicity and while the membrane technology was slightly superior to the chemical treatment with O3 followed by BAC, the differences were not large, indicating the comparable efficiency of both treatment technologies. While micropollutant removal is not the only indicator for the selection of a treatment technology, the proposed bioanalytical water quality assessment might prove useful in the future for benchmarking treatment technologies. The effects in Barrier 6, 7 and in drinking water were very low for many endpoints, typically falling below the detection limit. A similar very low detection was observed in the Power Station lake sample, which represents a reservoir where natural water and indirect purified recycled water (PRW) are mixed. ANOVA of the TEQ in the purified recycled water (Bundamba after AOP), drinking water (two samples in the drinking water distribution line) and bottled water (two samples) revealed no statistically significant differences for baseline-TEQ ( p ¼ 0.75), DEQ ( p ¼ 0.12), and TCDDEQ ( p ¼ 0.24) and the EEQ, BaPEQ and 4NQOEQ were all below detection limit in these samples. Only the PTEQ were slightly increased in the drinking water pipeline as discussed in Section 3.3. Despite these promising results that indicate that bioassays can be used to assess the fate of mixtures of chemicals across a wide range of water matrices, it is advisable to expand the battery of bioanalytical tools in the future to further endpoints relevant to human health, e.g., oxidative stress, additional genotoxicity endpoints and additional hormonal effects to include a wider variety of chemical groups in the investigation. It has also been demonstrated in a parallel study (Escher et al., 2011) that baseline-TEQ derived from the Microtox are not ideal as they are biased towards compounds of low hydrophobicity. Thus cytotoxicity endpoints based on 24 h (or longer) growth of mammalian cells or even bacteria would be beneficial to complement the bioassay battery as non-specific measure of all chemicals present in a given sample. Further, this study shows only a snapshot in time, given that grab samples were taken,
therefore the treatment efficiency given in Fig. 1 is not absolute but will vary in space and time. The novelty of this study is that it connects previous and future studies on individual steps of the water cycle constituting a first benchmark of the entire water cycle and clearly demonstrates that bioanalytical tools are applicable for a wide range of matrices. Detection limits of the bioassays were comparable or lower than the quantification limits of the routine chemical analysis, and allowed monitoring of the presence and removal of micropollutants post Barrier 2. The results obtained by the bioanalytical tools were found to be reproducible, robust and consistent with findings from previous studies assessing the effectiveness of the wastewater and advanced water treatment plants. The results of this study indicate that bioanalytical results expressed as toxic equivalent concentration (TEQ) provide valuable complementary information on mixture effects of groups of chemicals with a common mode of toxic action that help identify potential issues or to predict potential exposure/ risks of micropollutants to humans or the environment and are useful for continuous monitoring of the removal efficiencies of the various treatment processes and natural attenuation.
Acknowledgements This research was undertaken as part of the South East Queensland Urban Water Security Research Alliance, a scientific collaboration between the Queensland Government, CSIRO, The University of Queensland and Griffith University. We thank Seqwater for financial support of the development of the TEQ concept for the umuC assay. Particular thanks go to Hanne Thoen, Haipu Bi, Ben Mewburn and Kristie Lee Chue (Entox) for experimental assistance, Peter Sillivan (CSIRO) for collecting and extracting the samples and Julien Reungoat (AWMC, UQ) and Cathy Moore (Seqwater) for their help during sampling. We thank Greg Jackson of Queensland Health for helpful discussions. The authors acknowledge Queensland Urban Utilities for access to Oxley Creek Wastewater Treatment Plant and to the water distribution system; WaterSecure and Veolia Water Australia for access to the advanced treatment plants and PRW pipeline; Seqwater for samples from Lake Wivenhoe, mid-Brisbane River and the Mt Crosby Water Treatment Plant; Moreton Bay Water for access to the South Caboolture Water Reclamation Plant; and Gold Coast Water (now Allconnex Water) for access and assistance in samples from the Gold Coast water distribution system. We would like to also thank Michael Denison (University of California Davis, USA) for providing the H4G1.1c2 cells; Georg Reifferscheid (German Federal Institute of Hydrology, Germany) for providing the bacteria Salmonella typhimurium TA1535/pSK1002; and Ana Soto (Tufts University, USA) for providing the MCF7-BOS cells.
Appendix. Supplementary material Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.watres.2011.05.032.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 3 8 e4 2 4 7
references
Ellman, G.L., Courtney, K.D., Andres, V., Featherstone, R.M., 1961. A new and rapid colorimetric determination of acetylcholinesterase activity. Biochemical Pharmacology 7, 88e95. Escher, B.I., Bramaz, N., Maurer, M., Richter, M., Sutter, D., von Ka¨nel, C., Zschokke, M., 2005. Screening test battery for pharmaceuticals in urine and wastewater. Environmental Toxicology and Chemistry 24, 750e758. Escher, B.I., Bramaz, N., Mueller, J.F., Quayle, P., Rutishauser, S., Vermeirssen, E., 2008. Toxic equivalent concentrations (TEQs) for baseline toxicity and specific modes of action as a tool to improve evaluation of ecotoxicity tests on environmental samples. Journal Environmental Monitoring 10, 612e621. Escher, B.I., Bramaz, N., Ort, C., 2009. JEM Spotlight: monitoring the treatment efficiency of a full scale ozonation on a sewage treatment plant with a mode-of-action based test battery. Journal Environmental Monitoring 11, 1836e1846. Escher, B.I., Lawrence, M.G., Macova, M., Mueller, J.F., Poussade, Y., Robillot, C., Roux, A., Gernjak, A., 2011. Evaluation of contaminant removal efficiency of reverse osmosis and advanced oxidation in full-scale operation by combining passive sampling with chemical analysis and bioanalytical tools. Environmental Science and Technology 45, 5387e5394. Farre´, M., Martinez, E., Hernando, M., Fernandez-Alba, A., Fritz, J.I., Unruh, E., Mihail, O., Sakkas, V., Morbey, A., Albanis, T.A., Brito, F., Hansen, P.D., Barcelo, D., 2006. European Ring Exercise on Water Toxicity Using Different Bioluminescence Inhibition Tests Based on Vibrio fischeri, in Support to the Implementation of the Water Framework Directive. Talanta, 323e333. Freeman, S., Bates, J., Wallis-Lage, C., McEvoy, J., 2008. Drought relief in Soth-East Queensland, Australia provided by membrane-reclaimed water. Journal of American Water Works Association 100, 40e52. Hamers, T., Molin, K.R.J., Koeman, J.H., Murk, A.J., 2000. A smallvolume bioassay for quantification of the esterase inhibiting potency of mixtures of organophosphates and carbamate insceticides in rainwater: development and optimization. Toxicological Sciences 58, 60e67. Hawker, D.W., Cumming, J.L., Neale, P.A., Bartkow, M.E., Escher, B.I., 2011. A screening level fate model of organic contaminants from advanced water treatment in a potable water supply reservoir. Water Research 45, 768e780. International Organization for Standardization, 1998. Water Qualityedetermination of the Inhibitory Effect of Water Samples on the Light Emission of Vibrio fischeri (Luminescent Bacteria Test) EN ISO 11348e3, Geneva, Switzerland. International Organization for Standardization, 2000. Water Quality e Determination of the Genotoxicity of Water and Waste Water Using the Umu-test EN ISO 13829(2000) and 38415e3 (1996), Geneva, Switzerland. Johnson, B., 2005. Microtox acute toxicity test. In: Blaise, C., Ferard, J.-F. (Eds.), Small-scale Freshwater Toxicity Investigations. Springer. Korner, W., Hanf, V., Schuller, W., Kempter, C., Metzger, J., Hagenmaier, H., 1999. Development of a sensitive E-SCREEN assay for quantitative analysis of estrogenic activity in municipal sewage plant effluents. Science of the Total Environment 225, 33e48. Lee, Y., Escher, B.I., von Gunten, U., 2008. Efficient removal of estrogenic activity during oxidative treatment of waters
4247
containing steroid estrogens. Environmental Science & Technology 42, 6333e6339. Macova, M., Escher, B.I., Reungoat, J., Carswell, S., Lee Chue, K., Keller, J., Mueller, J.F., 2010. Monitoring the biological activity of micropollutants during advanced wastewater treatment with ozonation and activated carbon filtration. Water Research 44, 477e492. Nagy, S., Sanborn, J., Hammock, B., Denison, M., 2002. Development of a green fluorescent protein-based cell bioassay for the rapid and inexpensive detection and characterization of Ah receptor agonists. Toxicological Sciences 65, 200e210. Oda, J., Nakamura, S.I., Oki, I., Kato, T., Shinagawa, H., 1985. Evaluation of the new system (umu-test) for the detection of environmental mutagens and carcinogens. Mutation Research 147, 219e229. Queensland Water Commission, 2009. Interim Water Quality Report February 2009. http://www.qwc.qld.gov.au/ Interimþwaterþqualityþreport accessed 24.02.10), Queensland Government, Brisbane, Australia. Reifferscheid, G., Heil, J., Oda, Y., Zahn, R.K., 1991. A microplate version of the SOS/umu-test for rapid detection of genotoxins and genotoxic potentials of environmental samples. Mutation Research 253, 215e222. Reungoat, J., Macova, M., Escher, B.I., Carswell, S., Mueller, J.F., Keller, J., 2010. Removal of micropollutants and reduction of biological adverse effects in a full scale reclamation plant using ozonation and activated carbon filtration. Water Research 44, 625e637. Schreiber, U., Mu¨ller, J., Haugg, A., Gademann, R., 2002. New type of dual-channel PAM chlorophyll fluorometer for highly sensitive water toxicity biotests. Photosynthesis Research 74, 317e330. Schreiber, U., Quayle, P., Schmidt, S., Escher, B.I., Mueller, J., 2007. Methodology and evaluation of a highly sensitive algae toxicity test based on multiwell chlorophyll fluorescence imaging. Biosensors and Bioelectronics 22, 2554e2563. Soto, A.M., Sonnenschein, C., Chung, K.L., Fernandez, M.F., Olea, N., Serrano, F.O., 1995. The E-SCREEN assay as a tool to identify estrogens: an update on estrogenic environmental pollutants. Environ. Health Perspect 103 (Suppl. 7), 113e122. Traves, W.H., Gardner, E.A., Dennien, B., Spiller, D., 2008. Towards indirect potable reuse in South East Queensland. Water Science & Technology 58, 153e161. van Leeuwen, J., Pipe-Martin, C., Lehmann, R.M., 2003. Water reclamation at South Caboolture, Queensland, Australia. Ozone: Science & Engineering 25, 107e120. Vermeirssen, E.L.M., Hollender, J., Bramaz, N., von der Voet, J., Escher, B.I., 2010. Linking toxicity in algal and bacterial assays with chemical analysis in passive samplers exposed to treated sewage effluent. Environmental Toxicology and Chemistry 29, 2575e2582. Villeneuve, D.L., Blankenship, A.L., Giesy, J.P., 2000. Derivation and application of relative potency estimates based on in-vitro bioassays. Environmental Toxicology Chemistry 19, 2835e2843. WaterSecure, 2010. Water Quality Report October 2010. http:// www.westerncorridor.com.au/latest-news/water-qualityreport-december-2008-to-june-2010 (accessed 20.12.10.), State of Queensland, Australia, Brisbane, Australia. Zhao, B., Denison, M., 2004. Development and characterization of a green fluorescence protein-based rat cell bioassay system for detection of Ah receptor ligands. Organohalogen Compounds 66, 3332e3337.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 4 8 e4 2 5 9
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Kinetic modelling and microbial community assessment of anaerobic biphasic fixed film bioreactor treating distillery spent wash Bhavik K. Acharya a, Hilor Pathak a, Sarayu Mohana a, Yogesh Shouche b, Vasdev Singh c, Datta Madamwar a,* a
BRD School of Bioscience, Sardar Patel Maidan, Vadtal Road, Satellite Campus, PO Box 39, Sardar Patel University, Vallabh Vidyanagar 388 120, Gujarat, India b National Centre for Cell Sciences, Pune University Campus, Ganeshkhind, Pune 411 007, Maharashtra, India c G.H. Patel College of Engineering and Technology, Sardar Patel University, Vallabh Vidyanagar 388 120, Gujarat, India
article info
abstract
Article history:
Anaerobic digestion, microbial community structure and kinetics were studied in a biphasic
Received 17 January 2011
continuously fed, upflow anaerobic fixed film reactor treating high strength distillery waste-
Received in revised form
water. Treatment efficiency of the bioreactor was investigated at different hydraulic retention
14 May 2011
times (HRT) and organic loading rates (OLR 5e20 kg COD m3 d1). Applying the modified
Accepted 31 May 2011
StovereKincannon model to the reactor, the maximum removal rate constant (Umax) and
Available online 12 June 2011
saturation value constant (KB) were found to be 2 kg m3 d1 and 1.69 kg m3 d1 respectively. Bacterial community structures of acidogenic and methanogenic reactors were assessed using
Keywords:
culture-independent analyses. Sequencing of 16S rRNA genes exhibited a total of 123 distinct
Distillery spent wash
operational taxonomic units (OTUs) comprising 49 from acidogenic reactor and 74 (28 of
Biphasic reactor
eubacteria and 46 of archaea) from methanogenic reactor. The findings reveal the role of
Kinetic modelling
Lactobacillus sp. (Firmicutes) as dominant acid producing organisms in acidogenic reactor and
Bacterial community
Methanoculleus sp. (Euryarchaeotes) as foremost methanogens in methanogenic reactor.
16S rRNA gene
1.
Introduction
Anaerobic digestion has been considered as an attractive biotechnological method for degrading a variety of polluting organic wastes. However, it is a complex process requiring the concerted activity of multiple microbial populations interacting in a trophic web. The breakdown of organic chemicals in an anaerobic reactor usually involves several consequent degradation phases such as hydrolysis, acidogenesis, and then methanogenesis (Gujer and Zehnder, 1983). In the first two phases, organic pollutants are hydrolyzed and/or fermented into intermediate short-chain fatty acids (e.g., lactate, butyrate
ª 2011 Elsevier Ltd. All rights reserved.
and propionate), which are further degraded to acetate and H2/CO2. In the last phase, acetate and H2/CO2 are converted into methane. However, the inherently low growth rate of methanogenic bacteria can make the anaerobic systems sensitive to environmental changes (Xing et al., 1997). Disturbances in populations of one trophic level may affect the entire community and thus reducing the efficiency of the bioreactor (Raskin et al., 1996; Fernandez et al., 1999). Moreover, a singlestage anaerobic digester is susceptible to upset by rapid increase in volatile fatty acids and decrease in the pH of bulk solution, subsequently inhibiting the methanogenesis and leading to process failure. To eliminate such a common
* Corresponding author. Tel.: þ91 2692 229380; fax: þ91 2692 231042. E-mail addresses:
[email protected] (B.K. Acharya),
[email protected] (D. Madamwar). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.05.048
4249
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 4 8 e4 2 5 9
operational problem, two-stage anaerobic processes have been introduced and investigated as an alternative. The physical separation of the acidogenic and methanogenic phases increases stability as overloading of the methane reactor is prevented by proper control of the acidification step. Besides this, phase separation allows the maintenance of appropriate densities of the acid and methane formers in separate reactors and enables maximization of the rates of acidification and methanogenesis by applying optimal operational conditions. The acidogenic reactor can also serve as a buffer system when the composition of the wastewater is variable and can help in the removal of compounds that are toxic to methanogens. And finally, the acidogenic reactor provides a constant substrate for the methanogens, which are known to adapt slowly to varying substrate content and composition (Dinopoulou et al., 1988). In the present study, the anaerobic biphasic fixed film reactor has been favourably considered for treatment of distillery spent wash due to the advantages over other systems. Mathematical models based on process kinetics are used to design specific unit operations, optimize and control of wastewater treatment processes, and to understand the underlying biological and transport mechanisms within the reactor. There are numerous mathematical models cited in the literature for studying biological processes, such as Monod model (Williamson and McCarty, 1976), first order model with respect to the substrate (Pfeffer, 1968) and Contois model (Chen and Hashimoto, 1980). In the anaerobic fixed film reactor, the mass transfer limitations and their effect on the intrinsic kinetics are also important aspects to be considered. However, the complexity and uneven distribution of anaerobic biofilms in reactors make the kinetics very complex (Henze and Harremoes, 1983). There are several kinetic models developed for organic substance removal in continuously operated anaerobic reactors. StovereKincannon is one of most widely used mathematical model for determining the kinetic constants in immobilized systems (Kapdan, 2005). The model has been applied to continuously operated mesophilic and thermophilic upflow anaerobic filters for treatment of paperpulp liquors (Ahn and Forster, 2002), simulated starch wastewater (Ahn and Forster, 2000), soybean wastewater treatment (Yu et al., 1998), and anaerobic hybrid reactor (Buyukkamaci and Filibeli, 2002). Stover and Kincannon (Kincannon and Stover, 1982) established a kinetic model for biofilm reactors where the substrate utilization rate is expressed as function of the organic loading rate by monomolecular kinetics. The application of anaerobic technology in wastewater treatment requires careful operation and monitoring the conventional parameters such as pH, alkalinity, temperature, etc. Generally, little attention has been paid to the composition and activity of the microbial community as compared to the conventional parameters during the functioning of anaerobic reactors. However, an interdependent microbial community in anaerobic reactors respond very sensitively to sudden changes in environmental conditions and any imposed stress may lead to change in species types, their relative population levels and their activity, which is ultimately reflected in the reactor performance. Therefore, maintenance of active methanogenic populations in an anaerobic reactor is critical for stable performance (Harper and Pohland, 1986; Ince et al., 2001). An understanding of both the microbial ecology and their activity
are essential to operate the anaerobic reactors effectively. It is, therefore, necessary to determine the amount of active methanogenic populations in anaerobic reactors. Cultivation-dependent methods are known to be insufficient to characterize in situ structure and dynamics of microbial assemblages. As a consequence of developments in molecular ecology, the application of both qualitative and quantitative molecular methods such as, microbial community fingerprinting, cloning and sequence analysis, and fluorescence in situ hybridization, have led to new insights into microbial processes in biological reactors. In anaerobic bioreactors, where stability and performance are strongly dependent on complex microbial interactions, this information provides an opportunity to couple the microbial community structure and the functional characteristics of the anaerobic reactors (Pereira et al., 2002). Determining the underlying principles of the structure and function of these complex microbial communities may help to design optimized biological treatment systems; thus process failure may be avoided. This may facilitate the ability to design more efficient anaerobic reactors in terms of loading capacity with high methane yield (Akarsubasi et al., 2006). The aim of this study was to provide kinetic analysis of the biodegradation process and microbial community profile of an anaerobic biphasic fixed film bioreactor treating distillery spent wash.
2.
Materials and methods
2.1.
Substrate
Distillery spent wash was collected from a distillery situated at Ankleshwar, GIDC, Gujarat, India. The characteristics of the wastewater are shown in Table 1.
2.2. Construction and working of biphasic fixed film reactor Laboratory scale anaerobic biphasic fixed film reactor as shown in Fig. 1 was used in the present study. Both the
Table 1 e Characteristics of distillery spent wash. Parameters pH BOD5(mg L1) COD (mg L1) Total solid (TS) (mg L1) Total volatile solid (TVS) (mg L1) Total suspended solid (TSS) (mg L1) Total dissolved solids (TDS) (mg L1) Chlorides (mg L1) Phenols (mg L1) Sulfate (mg L1) Phosphate (mg L1) Total nitrogen (mg L1)
Values for distillery spent wash 3.0e4.5 50,000e60,000 110,000e190,000 110,000e190,000 80,000e120,000 13,000e15,000 90,000e150,000 8000e8500 8000e10,000 7500e9000 2500e2700 5000e7000
4250
1250 15.50 950 15.75 700 16.22 480 18.41 ND ND 4150 30.50 4150 25.75 4000 26.22 2580 25.45 2500 25.25
3000 20.25
5.75 3.25 0.45 004 85.50 52.95 6.90 010 78.75 5.25 45.95 5.50 7.20 0.40 02 003 72.50 5.10 40.10 5.25 8.20 0.35 059 002 69.90 2.25 37.90 4.40 8.22 0.3 060 002 64.40 5.10 31.40 4.10 8.35 0.2 063 003 65.50 4.25 30.20 3.25 8.30 0.2 064 002 179.50 10.25 155.65 10.10 5.0 0.2 178 005 178.95 10.25 141.95 9.35 5.50 0.5 165 007 178.90 12.50 140.80 10.15 5.50 0.5 150 005 166.90 12.10 134.80 10.25 5.60 0.5 145 005 158.90 10.60 141.50 9.60 5.80 0.4 125 004 156 10.50 140 9.50 5.80 0.5 135 005
52.55 2.35 65 3.35 72 3.40 79.25 3.15 84 2.25 89.51 2.20 e e e e e e
51.00 3.35 60.50 2.25 68.75 3.65 73.39 3.50 74.5 2.20 81.31 2.25 4.10 2.10 8.75 3.50 10 4.05 20 3.15
15 3.50
8 19 12.35 0.25 10 17 20.25 0.30 12 15 23.79 0.25 15 11.33 27.64 0.20 2.5 80
20 8 26.92 0.15
18 8.88 26.78 0.18
MR
3.0 66.66 3.5 57.14 e 4.5 44.44 e
20 3.21
ND e Not detected.
Various physicoechemical parameters were analyzed to study the biodegradation profile of the distillery spent wash by anaerobic biphasic reactor. All the analyses were carried out according to Standard Methods (APHA) as reported earlier (Acharya et al., 2008).
5.5 36.33 e
Analytical methods
6 33.33 e
2.3.
HRT (days) OLR (kgm-3 d1) Methane (m3m-3 d1) COD reduction (%) BOD reduction (%) TS (g L1) TDS (g L1) pH O/R potential (mV) TVA (mg L1)
reactors, acidogenic reactor (AR) and methanogenic reactor (MR) were constructed of a glass column. AR with a working volume of 3 L was having the following specifications: reactor inner diameter 9 cm; reactor height 134 cm; media height 75 cm; total volume (without bedding material) 4.8 L. MR with a working volume of 10 L having the following specifications: reactor inner diameter 18 cm; reactor height 152 cm; media height 75 cm; total volume (without bedding material) 18.2 L. The reactors were packed with uniform pieces of charcoal (2 kg in AR and 8 kg in MR). Distillery spent wash was fed into the reactor in upward direction at the required rate using a peristaltic pump (Gilson Miniplus 3, France). After the establishment of active biofilm in both the reactors, the acidogenic reactor was being operated by continuous feeding of raw distillery spent wash with an initial 6 days HRT and the methanogenic reactor was being operated at 20 days HRT and allowed to reach a steady state. Steady state condition was judged by stable gas production and constant COD and BOD of the effluent. Reactor performance data were recorded during the steady state condition of respective HRT and reported with standard deviation. Table 2 summarizes the steady state performance of biphasic reactor operated at 37 C in a temperature control chamber under different OLR.
AR
Fig. 1 e Schematic diagram of lab scale anaerobic biphasic fixed film reactor. A, untreated dsw; B, treated effluent; C, peristaltic pump; D1&D2, glass column reactor; E1eE4, gas collection system; F1&F2, outlet for effluent from respective reactor; G1&G2, gas outlet.
Table 2 e Values of an upflow anaerobic biphasic reactor effluent operated at 37 C with charcoal as packing material at varying hydraulic retention times under steady state conditions.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 4 8 e4 2 5 9
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 4 8 e4 2 5 9
4251
2.4. DNA extraction from acidogenic and methanogenic reactor
2.6. 16S rRNA gene sequencing and phylogenetic analysis
For the extraction of metagenomic DNA, 100 mL of the liquid sample was taken and concentrated by centrifugation at 10,000 g for 10 min. The pellet were mixed with 3 mm glass beads (1 g) in 30 mL of extraction buffer (10% (w/v) sucrose, 1% (w/v) cetyltrimethylammonium bromide (CTAB), 1.5 M NaCl, 100 mM TriseCl pH 8.0, 100 mM EDTA pH 8.0, 100 mM sodium phosphate buffer pH 8.0, containing 100 mL of 20 mg/mL proteinase K), and agitated manually for 15 min. Lysozyme (25 mg) was added and mixture was incubated at 37 C for 30 min. SDS (3 mL of 20%) was then added to the mixture and following incubation at 60 C for 30 min with intermittent mixing, 1 g powdered activated charcoal (Sigma) was added followed by further incubation under same condition for 30 min. The mixture was centrifuged at 10,000 g for 30 min and the pellets were discarded. The supernatant was treated with an equal volume of (24:1) (v/v) chloroform:isoamylalcohol and centrifuged at 10,000 g for 30 min. The aqueous phase, was removed and re-extracted with equal volume of chloroform. For DNA precipitation, isopropanol 0.8% (v/v) was added to the aqueous phase incubating overnight at room temperature and centrifuged at 10,000 g for 30 min. The crude DNA pellet was washed twice with 70% ethanol, air dried and resuspended in 2 mL TE buffer (10 mM TriseCl, 1 mM EDTA, pH 8.0).The extracted DNA was finally cleaned by gel permeation chromatography. For gel permeation column chromatography, a glass column (bed volume of 5 mL) was packed with freshly activated (0.5 g) Sephadex G150 (Pharmacia Fine Chemicals, Uppsala, Sweden) and equilibrated with 100 mM TriseCl (pH 8.0).
Sequensing of 16S rRNA genes cloned were obtained using the automated DNA Analyzer 3730 and ABI Prism BigDye terminator cycle sequencing v3.1 chemistry (Applied Biosystems, Foster City, CA, USA). The sequencing reactions were performed using vector specific M13 primers. Gene sequences were compared to database of rRNA gene sequences from GeneBank and reference sequences were searched using BLASTN programme (http://www.ncbi.nlm.nih.gov/BLAST/). The sequences were also compared to current database at the Ribosomal Database Project-II (RDP-II) using the RDP query programme SEQMATCH (http://www.simo.marsci.uga.edu/ publicdb/rdp_query.htm). All the sequences were checked for chimera formation with the CHECK_CHIMERA software of the RDP-II, and were aligned using Clustal W 1.6 programme (http://www.ebi.ac.uk/clustalw/). The alignment files obtained were analyzed and edited using DAMBE (Data Analysis in Molecular Biology and Evolution) software package (Xia and Xie, 2001). Phylogenetic analysis was performed using aligned sequences by the Neighbor-Joining algorithm with Kimura 2-parameter distance and bootstrapping with more than 1000 replicates in Molecular Evolutionary Genetics Analysis MEGA 4.0 software (Tamura et al., 2007).
2.5.
16S rRNA gene library construction
Three libraries were constructed, two of which were of Eubacterial 16S rRNA gene (one from AR and MR each) and one of Archaebacterial 16S rRNA gene. Eubacterial 16S rRNA genes were amplified using universal primers, 8F (50 -AGA GTT TGA TCC TGG CTC AG-30 ) and 1492R (50 -GGT TAC CTT GTT ACG ACT T-30 ) whereas Archaebacterial 16S rRNA genes were amplified using universal primers 21F (50 -TTCCGGTTGATCCYGCCGGA-30 ) and 958R (50 -YCCGGCGTTGAMTCCAATT-30 ). The amplification program was performed with initial denaturation of 5 min at 95 C; followed by 35 cycles of 2 min at 94 C, 1 min at 55 C and 1.5 min at 72 C; and a final extention at 72 C for 15 min. The PCR products were gel-purified using the Wizard PCR Clean-up kit (Promega, Madison, WI, USA), transformed into Escherichia coli DH5a competent cells via pGEM-T vector (Promega, Madison, WI, USA) and subjected to blue-white screening. White colonies on Luria-Bertani agar plates supplemented with ampicillin (50mgmL1), isopropyl-bD-thiogalactopyranoside and 5-bromo-4-chloro-3-indolyl-b-Dgalactoside were selected. Plasmid of positive clones was isolated using the ABI Prism MiniPrep kit (Applied Biosystems, Foster City, CA). A total of 900 16S rRNA gene clones (300 from each library) were randomly selected and subjected to 16S rRNA gene sequencing.
2.7.
Diversity indices and rarefaction analysis
All the sequences were grouped in to operational taxonomic units (OTUs) using programme of distance based OTU and richness determination (DOTUR) (Schloss and Handelsman, 2005). Based on grouping of unique OTUs, the frequency data for each distance level was used to construct rarefaction curves and the diversity indices like Shannon’s evenness index for general diversity (H0 ¼ -SPilnPi) and Simpson’s dominance index (D ¼ S [ni (ni-1)/N (N-1)]) (Hill et al., 2003; Wani et al., 2006; Kapley et al., 2007; Desai et al., 2009). In the equations, N is total number of OTUs in ith 16S rRNA gene sequence phylogroup, Pi is proportion of OTUs in ith 16S rRNA gene sequence phylogroup, ni is the number of OTUs in the ith 16S rRNA gene sequence phylogroup.
2.8.
Nucleotide sequence accession numbers
The nucleotide sequence data reported in this paper will appear in the EMBL, GenBank and DDBJ nucleotide sequence databases under accession no. HM218830 e HM218952 with the generic name of DMAR, DMER and DMMR.
3.
Results
3.1. Biomethanation of distillery spent wash by anaerobic biphasic reactor After the completion of start-up period, acidogenic reactor (AR) and methanogenic reactor (MR) were operated at 6 d and 20 d HRT respectively by considering the void volume of each reactor. Acidogenic reactor was initially started with very high
4252
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 4 8 e4 2 5 9
OLR of 33.33 kg COD m3 d1 to forbid the methanogenic population within the reactor. The hydraulic retention time was gradually decreased up to 2.5 d where the OLR was 80 kg COD m3 d1. Total volatile acid production was found to increase with decrease in the HRT and increase in organic loading rate in the AR due to high metabolic activity of acidogens. The highest TVA concentration (4150 mg L1) was observed at 3 and 2.5 d HRT. High acid production also resulted in the drop in pH from 6.0 to 5.0. Methanogenic reactor was started at initial HRT of 20 days with OLR of 8 kg COD m3 d1 of the effluent from AR. The TVA in the effluent of MR was below detection level during 20 d and 18 d HRT indicating high activity of methanogens within MR. During this period of operation, maximum COD and BOD reduction were 81.31% and 89.51% respectively. The reactor performance was stable between 18 and 12 d HRT resulting in around 70e75% COD reduction. However, with further decrease in HRT to 8 d (OLR of 19 kg COD m3 d1) COD removal efficiency of the reactor declined sharply as shown in Fig. 2. Biogas production is the foremost advantage in the anaerobic treatment of organic wastes. High volume/yield of biogas was observed during the treatment process indicating the presence of large number of methanogens in the MR. Fig. 2 shows that with change in OLR of MR from 8 kg COD m3 d1 (20 d HRT) to 11.33 kg COD m3 d1 (15 d HRT) gas production increased from 38.92 to 41.29 m3 m3 d1. However, with further increase in OLR, sharp decline in gas production was observed. Observed methane yield was more than 0.4 m3 per kg of COD removed at higher HRT, and then gradually it decreased upto 0.3 m3 per kg COD at 12 d HRT. During 10 d and 8 d HRT, the yield sharply declined due to the higher organic loading and gradual loss of methanogens from the reactor.
3.2.
Stover and Kincannon model
The StovereKincannon model considers the organic substance removal rate as a function of organic loading rate at steady state in Eq. (1). dS Q ¼ ðSi Se Þ dt V
(1)
The original StovereKincannon model for rotating biofilm reactor is described as in Eq. (2).
Fig. 3 e Stover and Kincannon model plot showing the effect of organic loading rate on rate of COD removal. QSi dS QðSi Se Þ A ¼ ¼ QSi dt V KB þ A Umax
(2)
Rearrangement of Eq. (2) gives the relationship
dS dt
1 ¼
V KB V 1 ¼ þ QðSi Se Þ Umax QSi Umax
(3)
The model applied to the methanogenic reactor (MR) and the substrate is expressed as COD (kg m3) in the Eq. (3), where dS/dt is the substrate removal rate (kg m3 d1), S is the reactor substrate concentration (kg m3); Umax is the maximum removal rate constant (kg m3 d1) and KB is a saturation value constant (kg m3 d1). If (dS/dt)1 is taken as V/Q(SiSe) which is the inverse of the loading removal rate and this is plotted against the inverse of the total loading rate V/(QSi), a straight line results. The intercept and slope of the line are 1/Umax and KB/Umax respectively. Fig. 3, illustrates the result of plotting the model graph i.e., inverse of specific substrate removal rate Q(SieSe)/V versus inverse of total loading rate QSi/V for all the HRT studied. The major feature of the graph is the gradual loss of efficiency with increase in loading rate. The kinetic constants KB and Umax can be estimated as 1.69 kg m3 d1 and 2 kg m3 d1 respectively from Fig. 3. The regression line had a R2 of 0.992 confirming the applicability of Eq. (2). The Umax and KB values obtained can be used to determine the volume required to decrease the influent organic concentration from Si to Se or to determine the effluent concentration for a given V and Si. If we consider a volume of an anaerobic reactor, where the mass of substrate into the volume of media is equal to the mass of substrate out of the volume of media plus the mass of substrate biodegraded, then a mass balance of substrate in and out of the volume can be made as follows: dS (4) QSi ¼ QSe þ V dt Relationship (4) for dS/dt can be substituted giving
Fig. 2 e Effect of OLR on COD removal and biogas production.
QSi ¼ QSe þ
Umax ðQSi =VÞ V KB ðQSi =VÞ
(5)
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 4 8 e4 2 5 9
This expression can be solved for either the required volume of the anaerobic reactor or the effluent concentration. V¼
QSi Umax Si KB Si Se
Se ¼ Si
Umax Si KB þ ðQSi =VÞ
(6)
(7)
Eq. (7) can be employed to calculate the effluent organic concentration at a given organic loading rate and influent concentration for the lab scale anaerobic fixed film reactor. Fig. 4 shows the satisfactory straight line relationship between effluent concentrations predicted by the Eq. (7) and observed during the experiment which appears to validate the model.
3.3. (AR)
Bacterial community structure of acidogenic reactor
Based on 16S rRNA gene sequence analysis clones were grouped in 49 OTUs on the basis of 0e2% sequence divergence of clones belonging to the same OTU. Phylogenetic analysis of the reactor AR revealed that Firmicutes (gram positive bacteria with low G þ C content) dominated the community which includes few representatives from Bacteroidetes also. Among the identified acid formers, Lactobacillus was found to dominate the overall community. Lactobacillus amylovorous was represented by the maximum number of clones along with other species like Lactobacillus mobilis, Lactobacillus gallinarum, Lactobacillus fermentum, Lactobacillus acidophilus, Lactobacillus oeni, Lactobacillus reuteri, Lactobacillus farmicimins, Lactobacillus secaphilus, Lactobacillus vaginalis etc. The phylogeny of the bacterial community reconstructed from AR is shown in Fig. 5.
3.4. Bacterial community structure of methanogenic reactor (MR) Two distinct libraries each of Eubacteria (DMER) and Archaea (DMMR) were constructed from methanogenic reactor using
Fig. 4 e Relationship between predicted effluent concentration and observed effluent concentration.
4253
16S rRNA gene approach. Sequence analysis of the library DMER revealed the existence 28 OTUs within the reactor. Phylotypes of bacteria in MR were affiliated with four major bacterial lineages namely Firmicutes (10.71%), Bacteroidetes (82.14%), d-Proteobacteria (3.57%) and Spirochaetes (3.57%). The phylogeny of the eubacterial community reconstructed from MR is shown in Fig. 6. Remarkably most of the phylotypes represent the unculturable and unidentified clones sharing common lineages as Firmicutes or Bacteroidetes. Analyses of archaeal 16S rRNA gene phylotypes disclose the presence of two major phyla Euryarchaeota (80.43%) and Crenarchaeota (19.56%). All the phylotypes belonged to phylum Crenarchaeota were found as uncultured and unidentified. However, among euryarchaeotes some identified isolates were Methanoculleus bourgensis (DMMR 315), Methanoculleus palmolei (DMMR 247), Methanoculleus sp. (DMMR 252, 219, 304). The phylogeny of the archaebacterial community reconstructed from MR is shown in Fig. 7.
3.5.
Diversity indices and rarefaction analysis
Shannon’s general diversity indices (H0 ) depicting evenness for all the three libraries DMAR, DMER and DMMR were 4.67 0.15, 4.02 0.18, 3.55 0.17 respectively, which clearly indicated the existence of highly diverse community in both the reactors. Simpson’s dominance indices (D) for DMAR, DMER and DMMR were 0.008 0.001, 0.007 0.001, 0.04 0.01 respectively, which suggest comparatively higher dominance of euryarchaeota in methanogenic reactor. As observed from Fig. 8, rarefaction curves representing bacterial community of AR (DMAR) and archaebacterial community of MR (DMMR) exhibited higher species richness along with a greater number of expected OTUs. The expected number of OTUs calculated by rarefaction analyses was congruent with the number of OTUs actually recovered from samples from each reactor.
4.
Discussion
To demonstrate the mutual effect of OLR and HRT on performance of anaerobic biphasic reactor treating distillery waste, the reactor was operated under varying HRT and OLR. The first phase AR, comprising fast growing, hydrolytic and acid producing bacteria which are less sensitive to environmental changes was smaller in volume as compared to the second phase, the methanogenic reactor. During the course of the study, COD reduction in AR was observed in the range of 4e20% which was mainly due to the hydrolysis of larger and complex molecules and settlement of the solids. The effluent of an acidogenic reactor contains mainly acetic, propionic and butyric acids, although higher fatty acids are found at lower concentration (Dinopoulou et al., 1988). Methanogens are responsible for the conversion of organic acids to methane which is also considered as rate limiting step of the process. The quality of the methane is also very important for further usage as pure methane has a heat value of 1000 Btu/ft3 but, when methane is mixed with carbon dioxide that is produced in an anaerobic digester, its heat value decreases significantly. In our study, the methane content of the biogas remained between 60 and 65% except during the 8 d HRT.
4254
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 4 8 e4 2 5 9
DMAR273 (HM218834) DMAR293 (HM218869) DMAR269 (HM218833) Uncultured bacterium clone (DQ318869) DMAR233 (HM218830) DMAR183 (HM218873) 7976 Lactobacillus amylovorus (EF120369) Lactobacillus sobrius (AY700063) DMAR227 (HM218839) 91 64 DMAR114 (HM218841) Lactobacillus hamsteri (NR_025448) Uncultured bacterium clone (AY959018) 82 Lactobacillus kefiranofaciens (AB372208) DMAR234 (HM218856) 93 DMAR180 (HM218857) Uncultured bacterium clone (AY959119) 82 DMAR199 (HM218832) 100 Uncultured bacterium clone (EF653416) DMAR245 (HM218864) 100 DMAR292 (HM218854) 62 DMAR160 (HM218878) DMAR195 (HM218844) 96 100 Uncultured bacterium clone (AF371485) Uncultured bacterium clone (AF371483) DMAR238 (HM218852) 100 Lactobacillus similis (AB282889) DMAR239 (HM218850) DMAR119 (HM218848) 100 Lactobacillus vini (AY681132) 63 DMAR100 (HM218847) 99 DMAR225 (HM218877) Bacillus subtilis (DQ301917) Rumen bacterium R2-10 (DQ393028) Clostridium stercorarium (NR_026113) DMAR272 (HM218860) DMAR136 (HM218859) 100 Uncultured bacterium clone (EU828411) 63
73
100
100
70 99 99
87 100
Firmicutes
100
0.02
Fig. 5 e The evolutionary history of bacterial clone library DMAR (from AR) was inferred using the NeighboreJoining method. The percentage of replicate trees in which the associated taxa clustered together in the bootstrap test (1000 replicates) is shown above the branches. The trees are drawn to scale, with the evolutionary distances used to infer the phylogenetic tree. The evolutionary distances were computed using the Kimura 2-parameter method and are in the units of the number of base substitutions per site. All positions containing alignment gaps and missing data were eliminated only in pair wise sequence comparisons. Phylogenetic analyses were conducted in MEGA4.
Results show that the methanogenic reactor was able to resist significant changes in the pH even at high loading of volatile acids during lower HRT (Table 2). The hydraulic retention time affects the rate and extent of methane production, which in turn is affected by environmental conditions within the digestion tank, the operating temperature, the solids concentration, and the content of volatile solids in the feed sludge. The retention time must be higher than the generation time of the slowest growing microorganisms present in the digester, to prevent them from washout. For stable operation it is recommended to keep the retention time two times higher than the generation time of the most important microorganisms (methanogens) (Dohanyos and Zabranska, 2001). Process imbalance in an anaerobic digester will normally lead to accumulation of volatile fatty acids (VFA) resulting in
a decrease in pH. Our study shows that though the acid concentration was increasing with decrease in HRT, reactor was stable in terms of pH (Table 2). The increase in acid concentration may not register as a drop in pH immediately if the buffer capacity of the material in the reactor is high. Most organic wastes often have a high content of weak acids and bases resulting in high buffering capacity. The organic acid accumulation therefore has to reach a high level before it is detected as a drop in pH (Acharya et al., 2008). A special feature of modified StovereKincanon model is the utilization of concept of total organic rate as the major parameter to describe the kinetics of an anaerobic fixed film reactor in terms of organic matter removal and methane production (Yu et al., 1998). Anaerobic fixed film reactors operate as a flow through contact processes in which waste passes through a bed of
4255
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 4 8 e4 2 5 9
DMER198 (HM218886) DMER287 (HM218887) 100 Uncultured bacterium clone (FJ825441) 79 DMER41 (HM218885) Bacteroidetes clone (GQ468584) 83 100 Uncultured bacterium clone (AB175367) 82 Uncultured bacterium clone (AB175366) DMER189 (HM218888) 100 100 100 DMER231 (HM218889) 100 Uncultured bacterium clone (EF559136) Bacterium enrichment culture clone (FJ799155) Uncultured bacterium clone (EU037335) DMER237 (HM218890) 100 Uncultured bacterium clone (FJ825451) 100 DMER229 (HM218891) DMER188 (HM218879) 100 Bacterium clone (EF559073) DMER186 (HM218883) 100 Bacterium clone (AM982594) DMER234 (HM218881) 100 Uncultured soil bacterium clone (DQ378270) DMER273 (HM218882) 98 Bacteroides nordii (EU887841) DMER261 (HM218893) 100 Spirochaeta sp. (AY800103) Bacterium clone B-25 (DQ115987) Uncultured bacterium clone (EF063613) Clostridium hastiforme (X80841) DMER194 (HM218894) 100 Bacterium clone (EF559053) DMER220 (HM218892) 100 Uncultured bacterium clone (AM500730)
100
100 80 100
98 59 50 70
Spirochaetes
Firmicutes
100
Bacteroidetes
78 82
0.05
Fig. 6 e The evolutionary history of bacterial clone library DMER (from MR) was inferred using the NeighboreJoining method. The percentage of replicate trees in which the associated taxa clustered together in the bootstrap test (1000 replicates) is shown above the branches. The trees are drawn to scale, with the evolutionary distances used to infer the phylogenetic tree. The evolutionary distances were computed using the Kimura 2-parameter method and are in the units of the number of base substitutions per site. All positions containing alignment gaps and missing data were eliminated only in pair.wise sequence comparisons. Phylogenetic analyses were conducted in MEGA4.
biomass either as a biofilm attached on the fixed media or as a mass of suspended growth solids within the bed. Previous studies by Song and Young (Song and Young, 1986) have shown that suspended biomass present within the media interstitial void spaces is a significant factor in producing high and stable removal efficiency in anaerobic fixed film reactors. Tay et al. (1996) demonstrated that the suspended biomass in anaerobic filters contributes approximately one half of the total waste removal. Therefore, the volume of the filter is used instead of the surface area of the support media for the anaerobic filter in this study. According to Kincannon and Stover (1982), both hydraulic loading rate and organic concentration exhibit definite relationship with substrate removal rate and efficiency. The results of the community analyses show clear dominance of Lactobacillus sp. in the acidogenic reactor. According to metabolism, Lactobacillus species can be divided into three groups. Obligatory homofermentative (Group I) L. acidophilus, Lactobacillus delbrueckii, Lactobacillus helveticus, Lactobacillus
salivarius; facultatively heterofermentative (Group II) Lactobacillus casei, Lactobacillus curvatus, Lactobacillus plantarum, Lactobacillus sakei and obligatory heterofermentative (Group III) Lactobacillus brevis, Lactobacillus buchneri, L. fermentum, L. reuteri (Ljungh and Wadstrom, 2009). The phylogenetic analysis shows existence of all the three groups of Lactobacillus sp. within the AR which play a major role in the bioconversion of complex molecules into acids. Components of wastewater showed a strong effect on the bacterial structure. Unlike the archaeal community, the eubacterial community in an anaerobic reactor is mostly determined by the wastes treated and would not seriously be affected by the type of reactor used. Results indicate that high rate effluent treatment was achieved in the study which is attributed to the biphasic fixed film reactor we used. OTUs classified to phyla Firmicutes and Bacteroidetes were detected in the domain bacteria. This was also consistent with the results for other reports on thermophilic reactors (Sasaki et al., 2006; Leven et al., 2007; Tang et al., 2004; Sasaki et al., 2007). The
4256
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 4 8 e4 2 5 9
DMMR398 (HM218911) DMMR75 (HM218920) Archaeon cloneGZK7 (AJ576208) 75 Archaeon cloneA05 (FJ205754) DMMR280 (HM218919) DMMR225 (HM218926) 69 82 Methanoculleus bourgensis (AY196674 50 DMMR58 (HM218921) 53 DMMR123 (HM218924) DMMR299 (HM218934) 70 Archaeon cloneSA31 (AB494249) DMMR284 (HM218923) DMMR360 (HM218927) 89 DMMR263 (HM218929) DMMR34 (HM218928) 54 97 Methanoculleus sp.dm2 (AJ550158) Uncultured euryarchaeoteB35 (EF552198) 66 Euryarchaeote cloneBSA2A (AB175350) 100 Uncultured Methanomicrobiales (CU917425) Methanoculleus palmoleiDSM4273 (NR_028253) 60 DMMR400 (HM218915) DMMR55 (HM218918) DMMR348 (HM218916) 79 DMMR179 (HM218917) DMMR363 (HM218930) 81 72 DMMR51 (HM218932) DMMR297 (HM218933) Uncultured cloneT18 (EU662691) 97 57 Archaeon clone (AB355099) Methanogenic-cloneSMPFLSS (FJ982763) DMMR356 (HM218938) 95 DMMR229 (HM218934) 100 DMMR327 (HM218935) 91 76 Crenarchaeote cloneQEEF1BH071 (CU916870) Crenarchaeote cloneQEED1AE091 (CU917175)
Crenarchaeotes
Euryarchaeotes
87 61
0.05 Fig. 7 e The evolutionary history of archaeal clone library DMMR (from MR) was inferred using the NeighboreJoining method. The percentage of replicate trees in which the associated taxa clustered together in the bootstrap test (1000 replicates) is shown above the branches. The trees are drawn to scale, with the evolutionary distances used to infer the phylogenetic tree. The evolutionary distances were computed using the Kimura 2-parameter method and are in the units of the number of base substitutions per site. All positions containing alignment gaps and missing data were eliminated only in pair.wise sequence comparisons. Phylogenetic analyses were conducted in MEGA4.
results suggest that the presence of lactose sugar in the wastewater may be responsible for large number of OTUs related to the genus Lactobacillus. However, there are reports of anaerobic conversion of lactic acid to acetic acid by Lactobacillus sp. (Elferink et al., 2001) thus explaining the dominance of the same in AR. It is suspected that clones, which are affiliated to unidentified microorganisms, could perform homoacetogenic metabolism, these clones act as bacteria with a potential for reverse homoacetogenesis under low hydrogen partial pressure established by active hydrogenotrophic methanogens
(Zinder, 1994). Irrespective of the kind of wastewater treated bacteria of the phylum Firmicutes were predominant microbes in methane reactors both at mesophilic and thermophilic temperatures, indicating high diversity of species in/of this phylum (Sasaki et al., 2006; Shigematsu et al., 2006). In an anaerobic reactor treating distillery wastewater, Tang et al. (2007) observed 60% and 30% of the total clones were classified into Firmicutes and Bacteroidetes respectively, indicating the high population of these two phyla. In our study, Clostridia and Spirochaetes were also identified as a part of the bacterial clone library from MR. Clostridia exhibit diverse metabolic
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 4 8 e4 2 5 9
Fig. 8 e Rarefaction curves for bacterial OTUs sampled from libraries DMAR, DMER and DMMR.
out to determine whether methanogenic populations present in anaerobic reactors are governed by intrinsic characteristics of the ecosystems themselves or simply arise from the operation of the bioreactors. It was hypothesized that the populations resulted from selective forces such as temperature and/or operating conditions acting on the microbes in the original inoculums over a period of time. This would lead to the development of very different microbial populations within different bioreactors (Visser et al., 1991; McHugh et al., 2003). Rarefaction curves are a concept developed by Hurlbert (1971), by plotting the cumulative number of unique OTUs versus the number of screened clones. Such type of curves indicate if the community diversity is well represented by the number of clones that have been sequenced, and reveal the approximate total number of different OTUs (or OTU richness) (Talbot et al., 2008).
5. capabilities including classic fermentation, homoacetogens or syntrophic acetate oxidation (Thiele, 1991; Drake et al., 1997; Schnurer et al., 1996). Little is known about the ecological role of synergistes or Spirochaetes found in anaerobic digestors. It has been hypothesized that these bacteria can ferment lactate, or degrade it syntrophically with hydrogenotrophs (Delbes et al., 2001). Dominance of hydrogenotrophic methanogens suggests that hydrogen-dependent methanogenesis was the favoured route of methane production in the MR compare to acetatedependent methanogenesis. Such pathways have been observed by Schnurer et al. (1999) and Hansen et al. (1999) in laboratory scale anaerobic digestors treating various effluents and two digestors fed with glucose (Raskin et al., 1996). Two mechanisms for methane formation from acetate have been described. The first one is aceticlastic, being carried out by Methanosarcinaceae or Methanosaetaceae where, Methanosarcinaceae generally have a higher acetate threshold but a higher growth rate and yield than Methanosaetaceae (Ferry, 1993). The second mechanism encompasses a two-step reaction in which acetate is first oxidized to H2 and CO2 and, with these products, subsequently converted to methane (Zinder and Koch, 1984). This reaction is performed by acetateoxidizing bacteria (often Clostridium sp.) in a syntrophic association with hydrogenotrophic methanogens (often Methanomicrobiales or Methanobacteriales) (Hattori et al., 2000; Petersen and Ahring, 1991; Schnurer et al., 1997). The probable key mechanism by which acetate degradation takes place is syntrophic acetate oxidation to carbon dioxide by a proton reducing bacterium followed by the reduction of carbon dioxide to methane by the hydrogenotrophic methanogenesis (Schnurer et al., 1999; Hansen et al., 1999). In a study by Tang et al. (2007) two OTUs, closely related to Methanosarcina thermophila and M. bourgensis respectively were detected by 16S rRNA gene clone analysis of microbial population of thermophilic upflow anaerobic filter reactor treating distillery wastewater. It is apparent that many of the sequences obtained in our study also were retained closely only to uncultured clones for which physiological and other properties remain unknown. A number of studies have been carried
4257
Conclusions
Anaerobic digesters often exhibit significant stability problems that may be avoided through appropriate bioprocess models and assessment of microbial communities involved in the complex process of pollutant removal. High quality microbiological information will provide an invaluable tool in the design and process control of anaerobic digestion application. The results show the applicability of StovereKincannon model to the anaerobic biphasic reactor which is efficient to remove 50e80% COD at different OLR. Phylogenetic analyses reveal the presence of acidogenic organisms (Firmicutes) in AR and dominance of hydrogenotrophic methanogens (Euryarchaeotes) in MR. Community profiling at every change in the HRT and OLR would be of great advantage in further understanding of complex microbial process.
references
Acharya, B.K., Mohana, S., Madamwar, D., 2008. Anaerobic treatment of distillery spent wash: a study on upflow anaerobic fixed film bioreactor. Bioresource Technology. 99 (11), 4621e4626. Ahn, J.H., Forster, C.F., 2000. Kinetic analyses of operation of mesophilic and thermophilic anaerobic filters treating simulated starch wastewater. Process Biochemistry 36 (1e2), 19e23. Ahn, J.H., Forster, C.F., 2002. A comparison of mesophilic and thermophilic anaerobic upflow filters treating paper-pulpliquors. Process Biochemistry 38 (2), 257e262. Akarsubasi, A.T., Ince, O., Oz, N.A., Kirdar, B., Ince, B.K., 2006. Evaluation of performance, acetoclastic methanogenic activity and archaeal composition of full-scale UASB reactors treating alcohol distillery wastewaters. Process Biochemistry 41 (1), 28e35. Buyukkamaci, N., Filibeli, A., 2002. Determination of kinetic constant of an anaerobic hybrid reactor. Process Biochemistry 38 (1), 73e79. Chen, Y.R., Hashimoto, A.G., 1980. Substrate utilization kinetic model for biological treatment process. Biotechnology and Bioengineering 22 (10), 2081e2095.
4258
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 4 8 e4 2 5 9
Delbes, C., Moletta, R., Godon, J.J., 2001. Bacterial and archaeal 16S rDNA and 16S rRNA dynamics during an acetate crisis in an anaerobic digestor ecosystem. FEMS Microbiology and Ecology 35 (1), 19e26. Desai, C., Parikh, R.Y., Vaishnav, T., Shouche, Y.S., Madamwar, D., 2009. Tracking the influence of long term chromium pollution on soil bacterial community structures by comparative analyses of 16S rRNA gene phylotypes. Research in Microbiology 160 (1), 1e9. Dinopoulou, G., Rudd, T., Lester, J.N., 1988. Anaerobic acidogenesis of a complex wastewater: I, the influence of operational parameters on reactor performance. Biotechnology and Bioengineering 31 (9), 958e968. Dohanyos, M., Zabranska, J., 2001. Anaerobic digestion. In: Spinosa, L., Vesilind, P.A. (Eds.), Sludge into Biosolids: Processing, Disposal, Utilization. IWA Publishing, pp. 223e241. Drake, H.L., Daniel, S.L., Kusel, K., Matthies, C., Kuhner, C., Braus-Stromeyer, S., 1997. Acetogenic bacteria: what are the insitu consequences of their diverse metabolic versatilities? Biofactors 6 (1), 13e24. Elferink, S.J.W.H., Krooneman, J., Gottschal, J.C., Spoelstra, S.F., Faber, F., Driehuis, F., 2001. Anaerobic conversion of lactic acid to acetic acid and 1,2-propanediol by Lactobacillus buchneri. Applied and Environmental Microbiology 67 (1), 125e132. Fernandez, A., Huang, S., Seston, S., Xing, J., Hickey, R., Criddle, C., Tiedje, J., 1999. How stable is stable? Function versus community composition. Applied and Environmental Microbiology 65 (8), 3697e3704. Ferry, J., 1993. Fermentation of acetate. In: Ferry, J.G. (Ed.), Methanogenesis: Ecology, Physiology, Biochemistry and Genetics. Chapman and Hall, New York, USA, pp. 305e334. Gujer, W., Zehnder, A.J.B., 1983. Conversion processes in anaerobic digestion. Water Science and Technology 15 (8e9), 127e167. Hansen, K.H., Ahring, B.K., Raskin, L., 1999. Quantification of syntrophic fatty acid-beta-oxidizing bacteria in a mesophilic biogas reactor by oligonucleotide probe hybridization. Applied and Environmental Microbiology 65 (11), 4767e4774. Harper, S.R., Pohland, F.G., 1986. Recent developments in hydrogen management during anaerobic biological wastewater treatment. Biotechnology and Bioengineering 28 (4), 585e602. Hattori, S., Kamagata, Y., Hanada, S., Shoun, H., 2000. Thermacetogenium phaeum gen. nov., sp. nov., a strictly anaerobic, thermophilic, syntrophic acetate-oxidizing bacterium. International Journal of Systemic Evolutionary Microbiology 50 (4), 1601e1604. Henze, M., Harremoes, P., 1983. Anaerobic treatment of wastewater in fixed film reactors: a literature review. Water Science and Technology 15 (8e9), 1e101. Hill, T.C.J., Walsh, K.A., Harris, J.A., Moffett, B.F., 2003. Using ecological diversity measures with bacterial communities. FEMS Microbiology and Ecology 43 (1), 1e11. Hurlbert, S.H., 1971. The non-concept of species diversity: a critique and alternative parameters. Ecology 52 (4), 577e586. Ince, O., Ince, B.K., Yenigun, O., 2001. Determination of potential methane production capacity of a granular sludge from a pilot-scale upflow anaerobic sludge blanket reactor using a specific methanogenic activity test. Journal of Chemical Technology and Biotechnology 76 (6), 573e578. Kapdan, I.K., 2005. Kinetic analysis of dyestuff and COD removal from synthetic wastewater in an anaerobic packed column reactor. Process Biochemistry 40 (7), 2545e2550. Kapley, A., Baere, T.D., Purohit, H.J., 2007. Eubacterial diversity of activated biomass from a common effluent treatment plant. Research in Microbiology 158 (6), 494e500. Kincannon, D.F., Stover, E.L., 1982. Design methodology for fixed film reaction-RBC’s and biological towers. In: Civil Engineering
for Practicing and Design Engineers, vol. 2. Paragamon Press, New York, pp. 107e124. Leven, L., Eriksson, A.R.B., Schnurer, A., 2007. Effect of process temperature on bacterial and archaeal community in two methanogenic bioreactors treating organic household waste. FEMS Microbiology and Ecology 59 (3), 683e693. Ljungh, A., Wadstrom, T., 2009. Lactobacillus Molecular Biology: from Genomics to Probiotics. Caister Academic Press, ISBN 978-1-904455-41-7. McHugh, S., Carton, M., Mahony, T., O’Flaherty, V., 2003. Methanogenic population structure in a variety of anaerobic bioreactors. FEMS Microbiology Letters 219 (2), 297e304. Pereira, M.A., Roest, K., Stams, A.J.M., Mota, M., Alves, M., Akkermans, A.D.L., 2002. Molecular monitoring of microbial diversity in expanded granular sludge bed (EGSB) reactors treating oleic acid. FEMS Microbiology and Ecology 41 (2), 95e103. Petersen, S., Ahring, B., 1991. Acetate oxidation in a thermophilic anaerobic sewage-sludge digestor: the importance of nonaceticlastic methanogenesis from acetate. FEMS Microbiology and Ecology 86 (2), 149e158. Pfeffer, J.T., 1968. Increased loading on digesters with recycle of digester solids. Journal of Water Pollution Control Federation 40 (11), 1920e1933. Raskin, L., Rittman, B.E., Stahl, D.A., 1996. Competition and coexistence of sulphate-reducing and methanogenic population in an aerobic biofilms. Applied and Environmental Microbiology 62 (10), 3847e3857. Sasaki, K., Haruta, S., Tatara, M., Yamazawa, A., Ueno, Y., Ishii, M., Igarashi, Y., 2006. Microbial community in methanogenic packed bed reactor successfully operating at short hydraulic retention time. Journal of Bioscience and Bioengineering 101 (3), 271e273. Sasaki, K., Haruta, S., Ueno, Y., Ishii, M., Igarashi, Y., 2007. Microbial population in the biomass adhering to supporting material in a packed-bed reactor degrading organic solid waste. Applied Microbiology and Biotechnology 75 (4), 941e952. Schloss, P.D., Handelsman, J., 2005. Introducing DOTUR, a computer program for defining operational taxonomic units and estimating species richness. Applied and Environmental Microbiology 71 (3), 1501e1506. Schnurer, A., Schink, B., Svensson, B.H., 1996. Clostridium ultunense sp. nov., a mesophilic bacterium oxidizing acetate in syntrophic association with a hydrogenotrophic methanogenic bacterium. International Journal of Systemic Bacteriology 46 (4), 1145e1152. Schnurer, A., Svensson, B., Schinkl, B., 1997. Enzyme activities in energetic of acetate metabolism by the mesophilic syntrophically acetate-oxidizing anaerobe Clostridium ultunense. FEMS Microbiology Letters 154 (2), 331e336. Schnurer, A., Zellner, G., Svensson, B.H., 1999. Mesophilic syntrophic acetate oxidation during methane formation in biogas reactors. FEMS Microbiology and Ecology 29 (3), 249e261. Shigematsu, T., Era, S., Mizuno, Y., Ninomiya, K., Kamegawa, Y., Morimura, S., Kida, K., 2006. Microbial community of a mesophilic propionate-degrading methanogenic consortium in chemostat cultivation analyzed based on 16S rRNA and acetate kinase genes. Applied Microbiology and Biotechnology 72 (2), 401e415. Song, K.H., Young, J.C., 1986. Media design factors for fixed bed reactors. Journal of Water Pollution Control Federation 58 (2), 115e121. Talbot, G., Toppb, E., Palina, M.F., Massea, D.I., 2008. Evaluation of molecular methods used for establishing the interactions and functions of microorganisms in anaerobic bioreactors. Water Research 42 (3), 513e537.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 4 8 e4 2 5 9
Tamura, K., Dudley, J., Nei, M., Kumar, S., 2007. MEGA4: molecular evolutionary genetics analysis (MEGA) software version 4.0. Molecular Biology and Evolution 24 (8), 1596e1599. Tang, Y., Shigematsu, Q., Ikbal, T., Morimura, S., Kida, K., 2004. The effect of micro-aeration on the phylogenetic diversity of microorganisms in a thermophilic anaerobic municipal solidwaste digester. Water Research 38 (10), 2537e2550. Tang, Y., Fujimura, Y., Shigematsu, T., Morimura, S., Kida, K., 2007. Anaerobic treatment performance and microbial population of thermophilic upflow anaerobic filter reactor treating awamori distillery wastewater. Journal of Bioscience and Bioengineering 104 (4), 281e287. Tay, J.H., Show, K.Y., Jeyaseelan, S., 1996. Effects of media characteristics on performance of upflow anaerobic packed-bed reactors. Journal of Environmental Engineering 122 (6), 469e476. Thiele, J.H., 1991. Mixed-culture interactions in methanogens. In: Zeikus, G., Johnson, E.A. (Eds.), Mixed Cultures in Biotechnology. McGraw- Hill, New York, pp. 261e292. Visser, F.A., Van Lier, J.B., Macario, A.J.L., Conway de Macario, E., 1991. Diversity and population dynamics of methanomic bacteria in a granular consortium. Applied Environmental Microbiology 57 (6), 1728e1734.
4259
Wani, A.A., Surakasi, V.P., Siddharth, J., Raghavan, R.G., Patole, M. S., Ranade, D., Shouche, Y.S., 2006. Molecular analyses of microbial diversity associated with the Lonar soda lake in India: an impact crater in a basalt area. Research in Microbiology 157 (10), 928e937. Williamson, K., McCarty, P.L., 1976. A model of substrate utilization by bacterial films. Journal of Water Pollution Control Federation 48 (1), 9e16. Xia, X., Xie, Z., 2001. DAMBE: data analysis in molecular biology and evolution. Journal of Heredity 92 (4), 371e373. Xing, J., Criddle, C., Hickey, R., 1997. Effects of a long term periodic substrate perturbation on an anaerobic community. Water Research 31 (9), 2195e2204. Yu, H., Wilson, F., Tay, J.H., 1998. Kinetic analysis of an anaerobic filter treating soybean wastewater. Water Research 32 (11), 3341e3352. Zinder, S.H., 1994. Syntrophic acetate oxidation and reversible acetogenesis. In: Drake, H. (Ed.), Acetogenesis. Chapman & Hall, NewYork, pp. 386e415. Zinder, S., Koch, M., 1984. Non-aceticlastic methanogenesis from acetate: acetate oxidation by a thermophilic syntrophic coculture. Archives in Microbiology 138 (3), 263e272.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 6 0 e4 2 6 8
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Effect of two-stage coagulant addition on coagulationultrafiltration process for treatment of humic-rich water Ting Liu a, Zhong-lin Chen a,*, Wen-zheng Yu a, Ji-min Shen a, John Gregory b a
State Key Laboratory of Urban Water Resource and Environment, School of Municipal and Environmental Engineering, Harbin Institute of Technology, Harbin 150090, China b Department of Civil, Environmental and Geomatic Engineering, University College London, Gower Street, London WC1E 6BT, UK
article info
abstract
Article history:
A novel two-stage coagulant addition strategy applied in a coagulation-ultrafiltration (UF)
Received 8 January 2011
process for treatment of humic-rich water at neutral pH was investigated in this study.
Received in revised form
When aluminum sulfate (alum) doses were set at a ratio of 3:1 added during rapid mix
26 May 2011
stage and half way through flocculation stage, the integrated process of two-stage alum
Accepted 31 May 2011
addition achieved almost the same organic matter removal as that of conventional one-
Available online 12 June 2011
stage alum addition at the same overall dose. Whereas membrane fouling could be effectively mitigated by the two-stage addition exhibited by trans-membrane pressure
Keywords:
(TMP) developments. The TMP developments were found to be primarily attributed to
Two-stage coagulant addition
external fouling on membrane surface, which was closely associated with floc character-
Coagulation-ultrafiltration
istics. The results of jar tests indicated that the average size of flocs formed in two-stage
Membrane fouling
addition mode roughly reached one half larger than that in one-stage addition mode,
Floc characteristics
which implied a beneficial effect on membrane fouling reduction. Moreover, the flocs with
Humic-rich water
more irregular structure and lower effective density resulted from the two-stage alum addition, which caused higher porosity of cake layer formed by such flocs on membrane surface. Microscopic observations of membrane surface demonstrated that internal fouling in membrane pores could be also remarkably limited by two-stage alum addition. It is likely that the freshly formed hydroxide precipitates were distinct in surface characteristics from the aged precipitates due to formation of more active groups or adsorption of more labile aluminum species. Consequently, the flocs could further connect and aggregate to contribute to preferable properties for filtration performance of the coagulation-UF process. As a simple and efficient approach, two-stage coagulant addition strategy could have great practical significance in coagulation-membrane processes. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
The application of low-pressure membrane filtration in water treatment has dramatically increased in the past decade with improvement in membrane quality and decreases in membrane costs (Katsoufidou et al., 2005; Huang et al., 2007).
Microfiltration (MF) and ultrafiltration (UF) can remove particles and colloid almost completely and exhibit significant advantages in controlling microorganisms and pathogens (Tian et al., 2010). But the critical issues of both MF and UF are ineffective removal of dissolved organic substances such as natural organic matter (NOM), and deterioration in performance due to
* Corresponding author. Tel./fax: þ86 451 86283028. E-mail address:
[email protected] (Z.-l. Chen). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.05.037
4261
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 6 0 e4 2 6 8
membrane fouling (Fabris et al., 2007; Peiris et al., 2010). Thus, it is of significance to integrate MF or UF with appropriate pretreatment processes to improve organic removal and mitigate membrane fouling. Coagulation remains the most common process for water treatment, normally for the removal of turbidity and NOM. A number of studies have reported that humic substances, as the main component of NOM, play an important role in membrane fouling (Yuan and Zydney, 2000; Costa and Pinho, 2005). For removal of humic substances by hydrolyzing coagulants, two main mechanisms have been proposed (Duan and Gregory, 2003). At pH 5 or less, neutralizing their charge by cationic metal species leads to precipitation. At higher pH, it is likely that dissolved organic substances can adsorb on the amorphous hydroxide precipitate (sweep flocculation) and hence be removed by physical separation such as sedimentation and filtration. On the one hand, membrane filtration is able to achieve the physical separation by rejecting aggregates formed in flocculation; on the other hand, coagulant addition prior to membrane process can remove the contaminants that contribute to fouling. Therefore, the integrated process composed of MF or UF and pre-coagulation, especially in-line coagulation (without settling), has become particularly attractive since it is expected to display preferable operation efficiencies (Choi et al., 2008; Bergamasco et al., 2011). Applying a coagulation step before membrane filtration can remarkably improve the permeate quality in terms of organic matter content in surface water treatment (Guigui et al., 2002; Barbot et al., 2008). Recent studies also report that MF and UF are effective for virus removal when combined with coagulation (Fiksdal and Leiknes, 2006; Shirasaki et al., 2009). However, the results of previous investigations on effect of coagulation pre-treatment on membrane permeability are inconsistent, which is probably attributed to different water qualities and experimental conditions (Judd and Hillis, 2001; Kimura et al., 2008). Coagulation conditions, such as coagulant type and dose, mix mode, etc., have a great impact on coagulation-membrane processes (Kim et al., 2006; Barbot et al., 2008). Konieczny et al. (2009) found that the higher organic removal efficiency of an inline coagulation-UF process was achieved with aluminum salt coagulant than with ferric salt. As noted by Choi and Dempsey (2004), under-dosing of coagulant could be considered for operation of UF, since it was possible to obtain good removals of contaminants and improved performance due to producing a smaller volume of waste solids. Cho et al. (2006) observed that coagulation with a rapid mix followed by a slow mix gave lower specific cake resistance than that with only a rapid mix for a submerged MF membrane, suggesting that different coagulation condition may lead to different floc filterability. It is essential to optimize the coagulation process to achieve the highest membrane filtration performance of coagulationmembrane systems. Our previous work found that a low additional dose of hydrolyzing coagulant added during floc breakage could bring about a change of floc surface characteristics and complete regrowth of broken flocs occurred (Yu et al., 2010). To take advantage of the newly precipitated material, a novel two-stage coagulant addition strategy was investigated in an in-line coagulation combined with submerged UF membrane process for water treatment. The effectiveness of such addition mode in process performance
was assessed in comparison with that of a conventional onestage addition mode. The mechanisms involved of its effect on membrane fouling were also discussed by analysis of floc properties and microscopic observation of membrane surface.
2.
Materials and methods
2.1.
Synthetic raw water and coagulant
Five grams of humic acid (Jufeng, Shanghai, China) was dissolved in 300 ml 0.1 M NaOH and mixed by a magnetic stirrer for 24 h. Then the pH was adjusted to 7.5 by adding 0.1 M HCl. The solution was diluted to 1 L in a measuring flask and was stored in the dark. The raw water used in this study was prepared by diluting the stock humic acid (40 ml) in local (Harbin, China) tap water (50 L), in a 60 L tank, to give the raw water with a humic acid concentration of 4 mg/L. Synthetic raw water has been chosen here in order to simplify the study since natural surface water would have implied reproducibility issues. The characteristics of the raw water and the tap water are listed in Table 1. Aluminum sulfate hydrate (Al2(SO4)3 18H2O; Bodi, China, ＞99%) ‘alum’ was used as a coagulant in this study. Stock alum solution was prepared at a concentration of 0.05 M (as Al) in DI water.
2.2.
Jar test
The equipment used in jar test was a beaker (capacity 1.2 L) with a flocculator (ZR4-6, Zhongrun, China), which enables mix speed and duration to be preset. Jar tests were carried out to determine the coagulant dose as well as to identify the floc characteristics.
2.2.1.
Coagulant dose
In jar test, the raw water (1 L) was stirred at 100 rpm for 1 min with the purpose of uniform mixing. Then a certain dose of alum was added, with a simultaneous increase of stirring speed to 200 rpm. The rapid mix speed of 200 rpm was maintained for 1 min, and reduced to 50 rpm afterward for more than 10 min to allow floc growth to occur. The optimum alum dose was determined by organic removal and zeta potential of coagulated water.
2.2.2.
Floc monitoring
Floc size was determined by Photometric Dispersion Analyzer (PDA-2000, Rank Brothers, UK) in jar test. The experimental
Table 1 e Characteristics of raw water and tap water. Parameters
Raw water
pH Turbidity (NTU) Conductivity (mS/cm) DOC (mg/L) UV254 (1/cm) Ca (mg/L) Al (mg/L) Temperature ( C)
7.5 1.47 86 4.36 0.158 11.4 0.077 22
0.1 0.18 4 0.32 0.007 0.2 0.027 1
Tap water 7.4 0.65 80 2.14 0.046 11.3 0.075 22
0.1 0.12 3 0.25 0.006 0.2 0.030 1
4262
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 6 0 e4 2 6 8
procedure of PDA-2000 was similar to that of Yukselen and Gregory (2002). The average transmitted light intensity (dc value) through the flowing sample and the root-mean-square value (rms) of the fluctuating component are monitored. The ratio (rms/dc) provides a sensitive measure of particle aggregation and it is often called the Flocculation Index (FI). The FI value is strongly correlated with floc size and always increases as flocs grow larger. In this work, after the FI value reached an initial steady value, coagulant was added into the solution and the FI value was recorded by a PC data acquisition system at 1 s intervals.
2.2.3.
Image analysis
Floc structure was determined in jar test by an optical microscope with a CCD camera and image analysis to derive two-dimensional fractal dimension (D2). During the slow mix period, samples of flocs were taken from below the surface of the suspension with a hollow glass tube with an inner diameter of 5 mm. One end of tube was inserted 3.0 cm below the surface and the other end was covered by a finger, and then the samples were withdrawn carefully. After transferring the sample onto a flat microscope slide, the image of flocs in the sample was captured by an optical microscope with a CCD camera (BX51, Olympus, Japan). The camera has a sensor matrix consisting of 1360 (horizontal) 1024 (vertical) pixels. Each pixel is recorded using bmp files. To interpret the image sizes correctly, a graduated microscale was photographed to determine the number of pixels corresponding to a given standard length. Images were obtained from an interrogation window of about 2370 mm 1784 mm with a resolution of 574 pixels/mm were achieved. Thus 1 pixel corresponds to about 1.7 mm. All data were recorded on the hard drive of a PC and a public domain software package, Scion-image (Scion Corporation, Frederick, MD), was used to analyze the captured images. About 50 samples of flocs in the suspension were analyzed to determine the fractal dimension. According to Stone and Krishnappan (2003), collections of similar and natural objects have been found to have area-perimeter relationships described by a function: D2
PfA 2
High Level Tank
2.3.
Coagulation-UF processes
Coagulation-UF process of one-stage coagulant addition (denoted as CUF-A) and that of two-stage coagulant addition (denoted as CUF-B) were operated in parallel. A schematic illustration of CUF-B is shown in Fig. 1. CUF-A was the same as that of CUF-B except for the absence of the second dose pump (red broken line). Raw water was fed into a constant-level tank to maintain the water head for membrane reactor. Certain doses of alum were continuously added into the mix tank and the second flocculation tank with dose pumps respectively. The rapid mix speed of 200 rpm was maintained in the mix tank with hydraulic retention time (HRT) of 1 min, and then reduced to 50 rpm in two flocculation tanks with each HRT of 5.5 min. It should be noted that the stirring conditions in the flocculation tanks of both systems were identical with the low speed, which was also applied in jar tests. Flocs can reach a steady-state size for a given shear condition. So despite the floc breakup of a lesser degree in the flocculation tanks, the floc growth processes were comparable between the two systems. Then coagulated water was directly introduced into the membrane reactor, which had an effective volume of 0.5 L. A hollow-fiber UF membrane module (Litree, China) of a total surface area of 0.1 m2 was submerged in the membrane reactor. The UF membrane was made of polyvinyl chloride (PVC), with a nominal pore size of 0.01 mm. The constant flux mode was employed for UF operation and the permeate flux was maintained at 10 L/(m2 h), corresponding to a HRT of 0.5 h. Permeate was continuously collected from the membrane module by a suction pump, with a pressure gauge monitoring trans-membrane pressure (TMP) development. Sludge accumulated at the bottom of the membrane reactor was discharged once a day.
Pressure Gauge Suction Pump
Coagulant
Permeate
Constant Level Tank
Dosage Pump
Mix Tank
Raw Water
where P is the perimeter, A is the projected area and D2 is the fractal dimension of the objects. As the object area becomes larger, the perimeter increases more rapidly than for Euclidean objects indicating that the boundary of the objects is becoming more convoluted and the shape of the objects is more irregular. The D2 can only take values between 1 and 2. If the image is circular, then D2 ¼ 1, indicating a spherical floc; while D2 ¼ 2 indicates a floc of linear structure.
Flocculation Tank
UF Membrane Module
Magnetic Stirrer Sludge Discharge
Feed Pump
Fig. 1 e Schematic diagram of CUF-B.
4263
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 6 0 e4 2 6 8
20
0.16
15
0.14
10
-1
UV254 (cm )
5
0.10
0
0.08
-5
UV254
0.06
Zeta potential
0.04
-10
Zeta potential (mV)
0.12
-15
0.02
-20 0.00 0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.16 0.18 0.20
Al (mM)
Fig. 2 e Effect of alum dose on UV254 removal and zeta potential.
2.4.
Other analytical methods
Samples were taken for zeta potential measurements in jar test immediately after 1 min rapid mix by a zeta meter (Zetasizer Model ZEN2600, Malvern, UK). UV absorbance at the wavelength of 254 nm (UV254) was determined by a spectrometer (T6, Puxi, China). Dissolved organic carbon (DOC) was determined with a total organic carbon (TOC) analyzer (TOC-VCPH, Shimadzu, Japan). Metal concentrations in water were determined with an inductively coupled plasma-atomic emission spectroscopy (ICP-AES; Optima 5300DV, PerkineElmer, USA). The fouled membrane fibers were cut down from the two membrane modules. The foulant layer attached on the membrane surface was gently washed off by DI water until the membrane surface was exposed again. The fouled membrane samples together with a clean membrane sample were then gold-coated by a sputter and observed under scanning electron microscopy (SEM; S-4700, Hitachi, Japan).
3.
Results and discussion
3.1.
Coagulant dose
Preliminary jar tests were performed to determine the appropriate coagulant dose in this study. The coagulation efficiency can be roughly reflected by the UV254 reduction in the case of NOM-rich water (Tran et al., 2006). As shown in Fig. 2, the residual UV254 decreased with the increase of alum
dose. It should be noted that UV254 was reduced dramatically in a low dose range of 0.008w0.015 mM Al, and then showed nearly constant values at doses more than 0.04 mM Al. The removal efficiency of UV254 was approximately 86% at alum dose of 0.04 mM Al, while the percent removal was only enhanced to 89% at a dose of 0.15 mM Al. In addition, zeta potential seems to be a useful control parameter for coagulation, with anionic NOM in raw water and exhibiting maximum removal in the zeta potential around zero (Bond et al., 2010; Matilainen et al., 2010). Results in Fig. 2 showed that the Al-humic flocs were negatively charged at low alum dose, and that charge reversal occurred at a dose of about 0.04 mM Al. Quite good removal of UV254 was achieved well before charge neutralization. At higher doses the zeta potential increased and did not exceed 15 mV. Thus, 0.04 mM Al was chosen as the optimum dose since it showed a similar level of UV254 removal when compared to higher doses and satisfied the zeta potential near zero. At alum dose of 0.04 mM Al, the pH of final solution was measured to decrease to w6.7, which was also in the range dominated by precipitate adsorption. Prior research (Yu et al., 2010) had indicated that just a small additional dose added during floc breakage was able to fully restore the regrowth ability of broken flocs. In this study, therefore, the two-stage alum doses were tentatively set at a ratio of 3:1 added during rapid mix stage and half way through flocculation stage of coagulation-UF operation.
3.2. Water quality of effluents of coagulation-UF systems The removal efficiencies of turbidity and dissolved organic matter by both CUF-A and CUF-B at the dose of 0.04 mM Al were listed in Table 2. Turbidity was reduced from initial 1.47 0.18 NTU in raw water to a level as low as 0.07 0.02 NTU in the effluents. It indicated that particulates were almost completely removed by both systems due to UF membrane rejection. Furthermore, there were no significant differences of average removal of DOC and UV254 between the two systems during the operation period, suggesting that the two-stage alum addition did not give any increase in organic removals. Cho et al. (2006) also reported that removals of organic matter seemed to be similar regardless of flocculation conditions including mix mode and flocculation time in a coagulation-MF system. These results implied that organic removal efficiency was probably dependent on overall coagulant dose rather than coagulation condition. As shown in Table 2, it was very interesting that residual Al concentration
Table 2 e Water quality of influent and effluents in coagulation-UF systems. Water quality indexes Turbidity (NTU) UV254 (cm1) DOC (mg/L) Residual Al (mg/L)
Influent
CUF-A Effluent
1.47 0.158 4.355 0.077
0.18 0.007 0.323 0.027
0.07 0.029 1.412 0.015
0.02 0.003 0.256 0.020
CUF-B Removal (%) 95.2 1.2 81.6 3.2 67.6 4.5 e
Note: For turbidity, UV254, and DOC, the measurements number n ¼ 9; for residual Al, n ¼ 5.
Effluent
Removal (%)
95.2 1.2 82.3 2.8 67.1 4.3 e
0.07 0.028 1.431 0.015
0.02 0.002 0.208 0.025
4264
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 6 0 e4 2 6 8
in the effluents (0.015 mg/L) was much lower than that in the influent (0.07 mg/L). It was likely because there was postprecipitation of nano-particles of Al(OH)3(s) that did not aggregate until coagulant was added. Moreover, it can be seen that the residual Al concentration in CUF-B effluent did not increase as a result of two-stage alum addition compared with that in CUF-A effluent at the same dose. Therefore, it is reasonable to infer that water quality of effluents of the two coagulation-UF systems was almost the same in organic removal and residual Al concentration.
3.3.
TMP developments of coagulation-UF systems
Membrane fouling is usually considered the most challenging factor in coagulation-membrane processes due to uncertain impact of coagulated flocs on membrane permeability (Howe et al., 2006). The TMP which was required to maintain the desired flux was used as the indicator of membrane fouling. In this study, the TMP increased with operation time while the flux of the two coagulation-UF systems was maintained constant at about 10 L/(m2 h) before the TMP exceeded 40 kPa. As shown in Fig. 3, there is an evident difference of TMP developments during continuous operation between the two systems at the dose of 0.04 mM Al. The TMP gradually increased from the initial 7.5 kPa for both systems with a similar trend within the beginning phase (0e30 h). However, there was a distinct gap of 2.0 kPa between them after 110 h. The TMP development of CUF-A exhibited a slow increase followed by a rapid one; while the TMP of CUF-B showed a steady development all through the operation period. The permeate flux declined slightly when the TMP of CUF-A exceeded 40 kPa, so this operation cycle came to an end. The final TMP of CUF-B only increased to 24 kPa, which was 16 kPa lower than that of CUF-A (40 kPa). It was thus believed that two-stage coagulant addition was an effective strategy to reduce membrane fouling in the coagulation-UF process. Membrane fouling mainly results from external fouling (i.e., cake and/or gel layer formed on membrane surface) and
40
0.02 0.02 0.04 0.04
35
TMP (kpa)
30
(CUF-A) (CUF-B) (CUF-A) (CUF-B)
25 20
internal fouling (i.e., small size substances deposited in membrane pores) (Meng et al., 2010). In order to identify the proportions of external and internal fouling resistances, the external foulants on membrane surface were carefully wiped off with a sponge at the end of the operation period, and then the filtration process was restored again using DI water. It was observed that the TMP in CUF-A and CUF-B was decreased to 13 kPa and 8.5 kPa respectively. Therefore, it could be roughly estimated that external and internal fouling resistances contributed to 83% and 17% of TMP development in CUF-A; while the contributions of external and internal fouling resistances to TMP development in CUF-B were 94% and 6% respectively. This indicated that the fouling resistance increase was dominated by foulant layer formation on the membrane surface in both systems, further confirming that the characteristics of coagulated flocs played an important role in membrane filtration process. In addition, it should be noted that the percentage of internal fouling resistance of CUF-A was much higher than that of CUF-B. This result implies that deposition in membrane pores and pore-blocking by small particles can be limited to a great degree in CUF-B, the mechanism of which will be discussed later. Furthermore, it’s been reported that under-dosing conditions sometimes obtained better performance of the coagulation with low-pressure membrane filtration process (Choi and Dempsey, 2004), so a lower dose of 0.02 mM Al (50% of the original dose) was applied in additional experiments to investigate the effect of two-stage coagulant addition. In this case, the two-stage alum doses were also set at a ratio of 3:1 added during rapid mix and half way through flocculation stage. It can be seen from Fig. 3 that TMP development of CUFB was much lower than that of CUF-A at the same constant flux of 10 L/(m2 h), which suggested that the two-stage addition was also an effective strategy to mitigate fouling at a lower coagulant dose. However, it should be noted that the TMP increase rates at a dose of 0.02 mM Al were respectively higher than those at a dose of 0.04 mM Al, as shown in Fig. 3. It was probably attributed to more microflocs that could not further aggregate and remained in coagulated water at a low dose. Then the microflocs might fill the cake layer and/or block the membrane pores, causing more severe membrane fouling. This is in consistence with the findings of Judd and Hillis (2001), who found that the low coagulant doses apparently cause incomplete aggregation of colloidal particles such that internal fouling of the membrane takes place. Leiknes et al. (2004) also reported that zeta potential which is measured close to a neutral charge is beneficial for the aggregate formation. Thus, the dose of 0.04 mM Al was chosen in this study for further investigation of floc characteristics.
3.4.
15 10 5 0 0
50
100
150
200
250
300
350
Time (h)
Fig. 3 e TMP variations at doses of 0.02 and 0.04 mM Al in CUF-A (one-stage) and CUF-B (two-stage).
Floc characteristics
Several studies reported that external fouling or fouling layer formation was the major cause of membrane fouling (Choi et al., 2008; Meng et al., 2010), which was coincident with the TMP results of this study. In coagulation-membrane processes, external fouling can be greatly affected by floc characteristics, e.g., hydrophobicity and morphological characteristics such as size and structure (Park et al., 2006; Zhao et al., 2010). The influence of floc hydrophobicity could be
4265
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 6 0 e4 2 6 8
3.4.1.
Flocculation index of flocs
Results from continuous monitoring of humic acid coagulation of the two addition modes are shown in Fig. 4. For both cases, there were short lag phases, during which small precipitate particles were formed and then aggregated, but the aggregates were not yet large enough to give a measurable increase in FI value. Soon after the lag phases a rapid rise in FI value occurred. Afterward the FI value reached a plateau value after a few minutes. These results are typical of the behavior of hydrolyzing coagulants at moderate doses and neutral pH, where hydroxide precipitates are formed (Yukselen and Gregory, 2002). As shown in Fig. 4, there was no appreciable variation in the onset or rise rate of the FI value at initial dose of 0.03 mM Al compared with that of 0.04 mM Al. However, there was a considerable difference of the steady FI values between them, when 0.01 mM Al was added about 5 min after the end of the rapid mix stage in two-stage addition mode. The plateau FI value of the two-stage addition roughly reached one half higher than that of the one-stage addition. The plateau FI values represent limiting floc sizes, determined by the effective shear rate in the stirred vessel and the strength of the flocs. The results suggested that a second alum dosage contributed to formation of stronger flocs at the same stirring speed, so the average size of stable flocs in two-stage addition mode was much larger than that in conventional addition mode. Floc size has a significant effect on membrane fouling potential in coagulation-membrane processes. Park et al. (2006) and Zhao et al. (2010) found that the flocs with smaller size contributed to higher specific cake resistance and more severe membrane fouling in coagulation-MF processes. Barbot et al. (2008) demonstrated that cake layer made of the flocs which were larger and more resistant to shear stress showed a flux advantage in a hybrid coagulation-UF system.
The results of this study compared favorably with those of the previous works. The two-stage coagulant addition resulted in larger flocs at the same overall alum dose, which had a beneficial effect on fouling control. This was probably because the new formed hydroxide precipitates were distinct in surface characteristics from the aged precipitates. It is likely that there are more active groups on the fresh precipitate particles which are capable of binding the aged precipitates together. Besides, Zhao et al. (2009) reported that polymeric Al species formed in the aged solutions were more stable than new formed ones. Thus another possibility is the adsorption of newly formed soluble Al species on the surface of the aged flocs, giving improved adhesion. This mechanism still needs to be further investigated in more detail.
3.4.2.
Fractal dimension of flocs
Different coagulation condition may lead to different evolution of floc structure. Fractal dimension (D2) of the flocs was investigated by jar test to provide information on the effect of two-stage alum addition on floc structure. As shown in Fig. 5, the fractal dimension of flocs formed in one-stage coagulant addition (D2 ¼ 1.293) was lower than that in two-stage one (D2 ¼ 1.439). It seems that the flocs formed during the first flocculation stage can more readily connect with each other
a
5.0
D2=1.293
4.8
2
R =0.93
4.6 4.4
Log As
negligible since the raw water quality was the same in this study. Therefore, the floc size and structure described by flocculation index (FI) and fractal dimension in two-stage coagulant addition mode were investigated by jar test in comparison with the case of one-stage addition mode.
4.2 4.0 3.8 3.6 3.4 3.2 3.0 2.0
2.2
2.4
2.6
2.8
3.0
3.2
3.0
3.2
3.4
3.6
Log P
b
0.30
2
R =0.934
4.6 4.4
0.20
L o g As
Flocculation Index
D2=1.439
4.8
0.25
0.15 0.10
0.04 mM (one-stage) 0.03+0.01 mM (two-stage)
0.05 0.00
5.0
4.2 4.0 3.8 3.6 3.4 3.2
0
200
400
600 800 Time (s)
1000
1200
Fig. 4 e FI values of flocs formed in one-stage and twostage alum additions. Arrow indicates the time of the second alum dosing.
3.0 2.0
2.2
2.4
2.6
2.8
3.4
3.6
Log P Fig. 5 e Fractal dimension of flocs formed in (a) one-stage and (b) two-stage alum additions.
4266
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 6 0 e4 2 6 8
with the help of newly hydroxide precipitates formed during the second alum dosing stage, so such flocs have relatively more branches and more convoluted boundaries. This process results in a higher D2 value of flocs formed during coagulation process with the two-stage alum addition. The higher fractal dimension means the more open or irregular floc structure and the lower effective density. Floc density can have a major effect on solideliquid separation processes (Gregory, 1998). In a conventional coagulation process for water treatment, the flocs with higher density are preferable because of their better settleability in the subsequent settling unit. However, it may not be the case in the coagulation-membrane processes. The cake layer on membrane surface formed by flocs with lower density may have higher porosity than that formed by higher density flocs. Choi et al. (2008) also reported that the humic flocs formed with relatively regular shape flocs resulted in more severe membrane fouling in a coagulation-MF system. Such aggregates were compressed easily by external pressure, so the cake layer built up with the compressed flocs imposed a higher specific resistance on the water flow. The flocs formed in two-stage coagulant addition mode were able to improve the permeability of water through cake layer on the membrane surface to reduce membrane fouling.
3.5.
Microscopic observations of membranes surface
Internal fouling can lead to the formation of irreversible fouling, which is harmful for long-term operation of membrane processes (Kimura et al., 2008; Meng et al., 2010). To evaluate the internal fouling degree, SEM images were taken for microscopic observations of the surface of a new and the fouled membranes (at the dose of 0.04 mM Al) with external
foulants to be removed. As shown in Fig. 6, there were a great number of large pores on the new membrane surface, and pore distribution appeared relatively uniform. At the end of this operation cycle, the number of pores on the fouled membrane surface of the two systems both decreased to some extent. It could be seen that the large pores scarcely existed on the membrane surface of CUF-A and the number of pores was also decreased to a large extent. Whereas many large pores could still be found on the membrane surface of CUF-B, and the number of pores retained on membrane surface of CUF-B was much more than that of CUF-A. The results of microscopic observations demonstrated that the internal fouling induced by deposition or blockage in membrane pores of CUF-A was much more serious than that of CUF-B, which compared favorably with the results of internal fouling resistances described by TMP of both systems. As mentioned above, the removal of humic acid was dependent on overall alum dose. That is, the residual concentration of humic acid in coagulated water of the two systems was almost the same. Thus, the less severe internal fouling in CUFB could not be attributed to less deposition of humic acid in the membrane pores. It is likely that the internal fouling was mainly caused by microflocs formed in coagulation process. The second addition of coagulant could cement the microflocs together and consequently form floc aggregates, then a smaller quantity of microflocs remained in coagulated water which was fed into membrane unit. The FI value is strongly influenced by large flocs, and microflocs do not greatly affect the FI value (Gregory and Chung, 1995). Therefore, quantity decrease of the microflocs and size increase of floc aggregates (higher FI value shown in section 3.4.1) suggested that the two-stage alum addition could contribute to larger flocs over a wide floc size distribution.
Fig. 6 e SEM images of (a)310000 and (b)350000 fold new membrane surface, and fouled membrane surface washed by DI water (350000 fold) of (c) CUF-A and (d) CUF-B.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 6 0 e4 2 6 8
4.
Conclusions
In this study, a two-stage coagulant addition strategy was investigated in an in-line coagulation combined with submerged UF membrane process, in comparison with a conventional one-stage coagulant addition for treatment of humic-rich water. When alum doses were set at a ratio of 3:1 added during rapid mix stage and half way through flocculation stage, almost the same permeate quality was achieved at the same overall dose at neutral pH, whereas membrane fouling could be effectively mitigated by the two-stage addition exhibited by TMP developments. The results of jar tests indicated that there was a significant difference of floc characteristics between the two alum addition modes. The average floc size represented by FI value of the two-stage addition mode roughly reached one half larger than that of the one-stage mode, which implied a beneficial effect on membrane fouling reduction. Moreover, the flocs with higher fractal dimension and lower effective density resulted from the two-stage alum addition, which caused higher porosity of cake layer formed on the membrane surface. It is likely that the flocs can further connect and aggregate with the help of newly hydroxide precipitates, making them have larger size and more irregular structure. Microscopic observations of membranes surface also demonstrated that internal pore fouling was remarkably alleviated by the two-stage alum addition attributed to a smaller quantity of microflocs remained in coagulated water fed into membrane unit. The present work provides a simple and effective approach to maintain relatively higher membrane permeability of a coagulation-UF process. Further work is still required to assess the effect of two-stage coagulant addition on membrane filtration performance with other kinds of contaminants in raw water.
Acknowledgment This research was funded by National High Technology Research and Development Program of China (2007AA06Z339) and State Key Laboratory of Urban Water Resource and Environment (HIT, Grant No. 2010DX12).
references
Barbot, E., Moustier, S., Bottero, J.Y., Moulin, P., 2008. Coagulation and ultrafiltration: understanding of the key parameters of the hybrid process. Journal of Membrane Science 325 (2), 520e527. Bergamasco, R., Konradt-Moraes, L.C., Vieira, M.F., FagundesKlen, M.R., Vieira, A.M.S., 2011. Performance of a coagulationeultrafiltration hybrid process for water supply treatment. Chemical Engineering Journal 166 (2), 483e489. Bond, T., Goslan, E.H., Parsons, S.A., Jefferson, B., 2010. Disinfection by-product formation of natural organic matter surrogates and treatment by coagulation, MIEX and nanofiltration. Water Research 44 (5), 1645e1653. Cho, M.H., Lee, C.H., Lee, S., 2006. Effect of flocculation conditions on membrane permeability in coagulation-microfiltration. Desalination 191 (1e3), 386e396.
4267
Choi, K.Y.J., Dempsey, B.A., 2004. In-line coagulation with low pressure membrane filtration. Water Research 38 (19), 4271e4281. Choi, Y.H., Kim, H.S., Kweon, J.H., 2008. Role of hydrophobic natural organic matter flocs on the fouling in coagulationmembrane processes. Separation and Purification Technology 62 (3), 529e534. Costa, A.R., Pinho, M.N., 2005. Effect of membrane pore size and solution chemistry on the ultrafiltration of humic substances solutions. Journal of Membrane Science 255 (1e2), 49e56. Duan, J.M., Gregory, J., 2003. Coagulation by hydrolysing metal salts. Advances in Colloid and Interface Science 100-102, 475e502. Fabris, R., Lee, E.K., Chow, C.W.K., Chen, V., Drikas, M., 2007. Pretreatments to reduce fouling of low pressure micro-flltration (MF) membranes. Journal of Membrane Science 289 (1e2), 231e240. Fiksdal, L., Leiknes, T.O., 2006. The effect of coagulation with MF/ UF membrane filtration for the removal of virus in drinking water. Journal of Membrane Science 279 (1e2), 364e371. Gregory, J., 1998. The role of floc density in solid-liquid separation. Filtration and Separation 35 (1), 571e575. Gregory, J., Chung, H.J., 1995. Continuous monitoring of floc properties in stirred suspensions. Journal of Water Supply: Research Technology, AQUA 44 (3), 125e131. Guigui, C., Rouch, J.C., Durand-Bourlier, L., Bonnelye, V., Aptel, P., 2002. Impact of coagulation conditions on the in-line coagulation/UF process for drinking water production. Desalination 147 (1e3), 95e100. Howe, K.J., Marwah, A., Chiu, K.P., Adham, S.S., 2006. Effect of coagulation on the size of MF and UF membrane foulants. Environmental Science and Technology 40 (24), 7908e7913. Huang, H.O., Lee, N.H., Young, T., Gary, A., Lozier, J.C., Jacangelo, J.G., 2007. Natural organic matter fouling of lowpressure, hollow-fiber membranes: effects of NOM source and hydrodynamic conditions. Water Research 41 (17), 3823e3832. Judd, S.J., Hillis, P., 2001. Optimization of combined coagulation and microfiltration for water treatment. Water Research 35 (12), 2895e2904. Katsoufidou, K., Yiantsios, S.G., Karabelas, A.J., 2005. A study of ultrafiltration membrane fouling by humic acids and flux recovery by backwashing: experiments and modeling. Journal of Membrane Science 266 (1e2), 40e50. Kim, H.C., Hong, J.H., Lee, S., 2006. Fouling of microfiltration membranes by natural organic matter after coagulation treatment: a comparison of different initial mixing conditions. Journal of Membrane Science 283 (1e2), 266e272. Kimura, K., Maeda, T., Yamamura, H., Watanabe, Y., 2008. Irreversible membrane fouling in microfiltration membranes filtering coagulated surface water. Journal of Membrane Science 320 (1e2), 356e362. Konieczny, K., Sa˛kol, D., P1onka, J., Rajca, M., Bodzek, M., 2009. Coagulation-ultrafiltration system for river water treatment. Desalination 240 (1e3), 151e159. Leiknes, T., Odegaard, H., Myklebust, H., 2004. Removal of natural organic matter (NOM) in drinking water treatment by coagulationemicrofiltration using metal membranes. Journal of Membrane Science 242 (1e2), 47e55. Matilainen, A., Vepsa¨la¨inen, M., Sillanpa¨a¨, M., 2010. Natural organic matter removal by coagulation during drinking water treatment: a review. Advances in Colloid and Interface Science 159 (2), 189e197. Meng, F.G., Liao, B.Q., Liang, S., Yang, F.L., Zhang, H.M., Song, L.F., 2010. Morphological visualization, componential characterization and microbiological identification of membrane fouling in membrane bioreactors (MBRs). Journal of Membrane Science 361 (1e2), 1e14. Park, P.K., Lee, C.H., Lee, S., 2006. Permeability of collapsed cakes formed by deposition of fractal aggregates upon membrane
4268
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 6 0 e4 2 6 8
filtration. Environmental Science and Technology 40 (8), 2699e2705. Peiris, R.H., Halle´, C., Budman, H., Moresoli, C., Peldszus, S., Huck, P.M., Legge, R.L., 2010. Identifying fouling events in a membrane-based drinking water treatment process using principal component analysis of fluorescence excitationemission matrices. Water Research 44 (1), 185e194. Shirasaki, N., Matsushita, T., Matsui, Y., Kobuke, M., Ohno, K., 2009. Comparison of removal performance of two surrogates for pathogenic waterborne viruses, bacteriophage Qb and MS2, in a coagulation-ceramic microfiltration system. Journal of Membrane Science 326 (2), 564e571. Stone, M., Krishnappan, B.G., 2003. Floc morphology and size distributions of cohesive sediment in steady-state flow. Water Research 37 (11), 2739e2747. Tian, J.Y., Chen, Z.L., Yang, Y.L., Liang, H., Nan, J., Li, G.B., 2010. Consecutive chemical cleaning of fouled PVC membrane using NaOH and ethanol during ultrafiltration of river water. Water Research 44 (1), 59e68.
Tran, T., Gray, S., Naughton, R., Bolto, B., 2006. Polysilicato-iron for improved NOM removal and membrane performance. Journal of Membrane Science 280 (1e2), 560e571. Yuan, W., Zydney, A.L., 2000. Humic acid fouling during ultrafiltration. Environmental Science and Technology 34 (23), 5043e5050. Yukselen, M.A., Gregory, J., 2002. Breakage and reformation of alum flocs. Environmental Engineering Science 19 (4), 229e236. Yu, W.Z., Gregory, J., Campos, L., 2010. Breakage and regrowth of Al-humic flocs - effect of additional coagulant dosage. Environmental Science and Technology 44 (16), 6371e6376. Zhao, B.Q., Wang, D.S., Li, T., Chow, C.W.K., Huang, C., 2010. Influence of floc structure on coagulationemicrofiltration performance: effect of Al speciation characteristics of PACls. Separation and Purification Technology 72 (1), 22e27. Zhao, H., Liu, H.J., Qu, J.H., 2009. Effect of pH on the aluminum salts hydrolysis during coagulation process: formation and decomposition of polymeric aluminum species. Journal of Colloid and Interface Science 330 (1), 105e112.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 6 9 e4 2 7 8
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Reduced membrane fouling in a novel bio-entrapped membrane reactor for treatment of food and beverage processing wastewater Kok-Kwang Ng a, Cheng-Fang Lin a,*, Sri Chandana Panchangam a, Pui-Kwan Andy Hong b, Ping-Yi Yang c a
Graduate Institute of Environmental Engineering, National Taiwan University, 71 Chou-Shan Rd., Taipei 106, Taiwan Department of Civil and Environmental Engineering, University of Utah, 110 South Central Campus Drive, 2068 MCE, Salt Lake City, UT 84112, USA c Department of Molecular Biosciences and Bioengineering, University of Hawaii at Manoa, Honolulu, HI 96822, USA b
article info
abstract
Article history:
A novel Bio-Entrapped Membrane Reactor (BEMR) packed with bio-ball carriers was con-
Received 24 December 2010
structed and investigated for organics removal and membrane fouling by soluble microbial
Received in revised form
products (SMP). An objective was to evaluate the stability of the filtration process in
27 May 2011
membrane bioreactors through backwashing and chemical cleaning. The novel BEMR was
Accepted 31 May 2011
compared to a conventional membrane bioreactor (CMBR) on performance, with both
Available online 7 June 2011
treating identical wastewater from a food and beverage processing plant. The new reactor has a longer sludge retention time (SRT) and lower mixed liquor suspended solids (MLSS)
Keywords:
content than does the conventional. Three different hydraulic retention times (HRTs) of
Bio-entrapped membrane reactor
6, 9, and 12 h were studied. The results show faster rise of the transmembrane pressure
Conventional membrane bioreactor
(TMP) with decreasing hydraulic retention time (HRT) in both reactors, where most
Hydraulic retention time
significant membrane fouling was associated with high SMP (consisting of carbohydrate
Sludge retention time
and protein) contents that were prevalent at the shortest HRT of 6 h. Membrane fouling
Soluble microbial products
was improved in the new reactor, which led to a longer membrane service period with the
Membrane fouling
new reactor. Rapid membrane fouling was attributed to increased production of biomass and SMP, as in the conventional reactor. SMP of 10e100 kDa from both MBRs were predominant with more than 70% of the SMP <100 kDa. Protein was the major component of SMP rather than carbohydrate in both reactors. The new reactor sustained operation at constant permeate flux that required seven times less frequent chemical cleaning than did the conventional reactor. The new BEMR offers effective organics removal while reducing membrane fouling. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
Membrane bioreactors (MBRs) have been widely adopted for secondary treatment of municipal wastewater in the past
decade, especially in developed countries. They have advantages over conventional activated sludge systems such as in high removal efficiency, stable and good effluent qualities, small footprint, short hydraulic retention time (HRT), simple
* Corresponding author. Tel.: þ886 2 3366 7427; fax: þ886 2 2392 8830. E-mail address:
[email protected] (C.-F. Lin). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.05.031
4270
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 6 9 e4 2 7 8
operation, ease in maintenance of high biomass concentrations, and enhanced nitrification (Meng et al., 2007; Malamis and Andreadakis, 2009). However, a major issue of MBRs is the rapid decline of permeate flux due to a high level of biomass in the reactor that accelerates membrane fouling (Meng et al., 2005; Chae et al., 2006). In recent years, extracellular polymeric substances (EPS) and/or soluble microbial products (SMP) have been established as a main cause of membrane fouling (Cho et al., 2005; Jarusutthirak and Amy, 2006; Liang et al., 2007; Malamis and Andreadakis, 2009; Meng et al., 2009). Bound EPS are extracellular components tightly attached to the biological flocs, whereas soluble cellular components are soluble EPS or SMP from microbial growth and decay, as well as from dissolution of bound EPS (Ramesh et al., 2006; Ng et al., 2010). EPS and SMP typically consist of polysaccharides, proteins, lipids, and nucleic acids (Le-Clech et al., 2006; Liang et al., 2007). Found in most biologically treated effluents, SMP contributes significantly to soluble organic matter and Chemical Oxygen Demand (COD) of the effluent (Barker and Stuckey, 1999; Zhou et al., 2009). In addition, polysaccharide-like and protein-like substances are found predominant in EPS and/or SMP (Rosenberger et al., 2006; Le-Clech et al., 2003; FrØlund et al., 1995; Malamis and Andreadakis, 2009), though the species of these substances have not been well defined. Depending on its nature and molecular size, SMP may form a cake layer on the membrane surface or penetrate into the membrane pores (Jarusutthirak and Amy, 2006; Rosenberger et al., 2006; Meng et al., 2009). The behavior of SMP in membrane fouling is complex because of its disparate molecular weight (MW), hydrophobicity, and zeta potential (Barker and Stuckey, 1999; Jarusutthirak and Amy, 2006; Pan et al., 2010). SMP comprises a wide range of MW ranging from <1 kDa to 0.45 mm (Barker and Stuckey, 1999; Malamis and Andreadakis, 2009; Ni et al., 2010). Further research is warranted on how MW distribution and characteristics of SMP influence membrane fouling. Studies were undertaken that investigated moving bed biofilm reactor (MBBR) coupled with membrane as an alternative to the conventional MBR (Artiga et al., 2005; Ivanonic et al., 2006; Lee et al., 2006; Leiknes and Ødegaard, 2007). Yang et al. (2009) found that moving bed membrane bioreactor (MBMBR) produced more biomass in the effluent leading to increased membrane fouling than did CMBR. SMP in MBMBR were found to be more abundant than in CMBR. The objective in these studies of MBBR was to reduce MLSS in MBR, as ours was in developing the BEMR. Reducing MLSS enables less frequent backwashing and reduces downtime to clean the membrane. Contrarily, high MLSS in MBRs may increase non-Newtonian viscosities that hamper oxygen transfer and require additional energy for pumping (Drews and Kraume, 2005; Drews et al., 2005). We have developed here a new BEMR that reduced suspended biomass and increased SRT in the reactor with the objectives to achieve high organics removal in a more facile operation with a short start-up period. As membrane fouling may differ between the new BEMR and conventional MBR, we have investigated SMP and their characteristics in membrane fouling of both reactors at various HRTs, and further evaluated membrane cleaning of reactors for comparison. An overall
study goal is to reduce membrane fouling commonly encountered in MBRs.
2.
Material and methods
2.1.
Setup of laboratory-scale BEMR and CMBR
A BEMR and a CMBR were set up in the laboratory for experiments as shown in Fig. 1. Each of the MBRs had a working volume of 50 L with a polyvinylidene fluoride (PVDF) hollow fiber UF membrane module installed in it. They were operated in parallel for about four months with configurations and operation parameters specified in Table 1. The BEMR consisted of two compartments, with the first housing the entrapped bio-balls and the second housing the membrane module. The separate compartments allowed each to be designed and operated optimally. The entrapped bio-balls, 2.5 cm in diameter, were prepared per Yang et al. (2002, 2003), and were packed in the BEMR occupying 55% of the first compartment (first compartment volume ¼ 42 L). The activated sludge immobilized in the bio-balls of the BEMR and in the CMBR was from a wastewater treatment plant of the food industry. Prior to data collection, the CMBR was operated in batch mode and the BEMR in continuous-flow for 20 days to reach a steady-state condition that attained 90% removal of COD (Yu et al., 2009). After the steady state was reached in the effluents, the membrane modules were installed into the reactors. Both BEMR and CMBR were then operated at varied hydraulic retention times (HRT) of 6, 9, or 12 h during experiments on membrane fouling. Both MBRs were allowed to run for a period longer than 10 times the test HRT prior to experimental data collection. The SRT in the BEMR was determined according to Qian et al. (2001). The average SRTs of the BEMR and the CMBR were calculated to be 500 d and 20 d, respectively. The MLSS concentration in the CMBR was maintained at 8000e9000 mg L1 by withdrawal of excess sludge from the CMBR. The hollow fiber membrane module had an effective filtration area of 0.046 m2 and nominal pore size of 0.036 mm (GE ZeeWeed-1, USA). The membrane characteristics are summarized in Table 2. The wastewater was taken from a food processing complex of the Uni-President Enterprises Corporation in Taoyuan, Taiwan which manufactured flavored and fresh milks, grain and tea beverages, fruit juices, dairy products and instant noodles. The wastewater contained 590e1350 mg L1 of COD and 77e120 mg L1 of suspended solids (SS) and was fed to both MBRs. The wastewater was introduced into the MBRs by a peristaltic pump and air into the bottom of the MBRs by an air pump to maintain an aerobic environment for the microorganisms. Dissolved oxygen (DO) concentrations were maintained at the optimal levels of 7e8 mg L1 and 2.5e3.5 mg L1 for the BEMR and the CMBR, respectively. The membrane permeate was withdrawn through the UF hollow fiber membrane by a suction pump. An average flux of 20 L m2 h1 was maintained and the increasing transmembrane pressure (TMP) was continually monitored for both MBRs performance evaluation as membrane fouling results in an increase of TMP. The water level sensor, suction pump, and backwash pump were controlled by a computer.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 6 9 e4 2 7 8
4271
Fig. 1 e Schematic of the membrane bioreactors, a) BEMR, b) CMBR.
2.2.
Membrane test configurations
Three experimental configurations, Experiments I, II, and III, were used. Experiment I was conducted at an average flux of 20 L m2 h1 through the membrane at various HRTs without backwashing. Variations of HRTs at 6, 9, and 12 h were facilitated by operating the BEMR and the CMBR with recirculation. This was studied because changes in HRT could significantly impact membrane fouling. The data was collected after 14 days of operation at each HRT to ensure establishment of the steady state in the MBRs. Once the transmembrane pressure
exceeded the maximum operating pressure, i.e., 55 kPa, the membrane modules were removed for chemical cleaning. Chemical cleaning was performed by soaking the membrane in a solution of sodium hypochlorite (NaOCl) for 30 min. After cleaning the membrane was tested by compressing air into the hollow fibers in a tank filled with distilled water; the member was then reinstalled or replaced when air bubbles were observed. In Experiment II, the two MBRs (BEMR and CMBR) were operated at the optimum HRT of 12 h while maintaining an average flux of 20 L m2 h1, the performance of the
4272
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 6 9 e4 2 7 8
Parameter Carrier diameter size (cm) Total reactor volume, (L) Void volume (L) Packing ratio (%) Dissolved oxygen (mg L1) Average SRT (d) MLSS (mg L1) Temperature ( C) PH Average permeate flux (L m2 h1)
BEMR 2.5 50 23 55 7.0e8.0 500 11000a 25 1.0 7.0e8.3 20
CMBR e 50 e e 2.5e3.5 20 8000e9000 25 1.0 7.0e8.2 20
a The biomass in the bio-carrier.
membranes was examined periodically after brief operations. Membrane conditions after 15 min of suction and 1 min of backwash, after 30 min of suction and 1 min of backwash, and after backwashing of a fouled membrane at the flow velocity of 0.1 m s1 were compared. In Experiment III, both MBRs were operated at the flux of 20 L m2 h1 with 15 min of suction and 1 min of backwash. Experiment III was carried out to determine the service duration that the MBRs could be continually operated before the TMP threshold of 55 kPa was reached when chemical cleaning became necessary.
2.3.
Analytical methods
The soluble microbial products (SMP) in the MBRs were collected by centrifugation (BOECO U-320R, Germany) of samples at 3500g for 15 min. The supernatant was filtered through a 0.22-mm filter. The filtrate thus contained the total SMP contents (total protein and total carbohydrate) in the MBRs. The SMP were then fractionated into four groups, G-1 (>100 kDa), G-2 (10e100 kDa), G-3 (1e10 kDa), and G-4 (<1 kDa), according to their apparent molecular weights (AMW) by gel filtration chromatography (GFC), as previously detailed (Lin et al., 1999; Ng et al., 2010). The total SMP and all GFC fractions were analyzed for their total protein and total carbohydrate contents, which were regarded as the most important parts of SMP materials. The total protein content
was determined using the modified Lowry method with bovine serum albumin (BSA) as the protein standard (Lowry et al., 1951; FrØlund et al., 1995; Liang et al., 2007), and the total carbohydrate content was determined using the modified Anthrone method with glucose as the standard (Raukjaer et al., 1994; FrØlund et al., 1995). Organic compounds in the MBRs, once reaching the steadystate condition, were analyzed two or three times per week throughout experimentation. Daily measurements were recorded at the start-up stage of each run. The COD was analyzed by HACH closed reflux colorimetric method with the use of a spectrophotometer (HACH DR 2800). The pH and DO values were measured with a portable dissolved oxygen/pH meter (HACH HQ20) and a HACH Conductivity Meter (secsION5), respectively. The MLSS concentration of the CMBR was measured with an MLSS meter (KRK SS-5Z, Japan). The SS concentration was determined by drying at 105 C and the residual weight.
3.
Results and discussion
3.1. Performance of BEMR (new) and CMBR (conventional) on COD removal Fig. 2 shows the influent COD ranging from 590 to 1350 mg L1 in the food processing wastewater being fed to both MBRs, the effluent COD, and the calculated COD removal in each. After the start-up period of 20 days, the removal of COD in both reactors reached a stable removal value of about 90% and from days 21 through 120 the removal was 91e98% for the CMBR and 93e98% for the BEMR. A major advantage of MBRs is their ability to achieve more than 90% removal of COD (Cho et al., 2005; Fallah et al., 2010).
3.2.
Membrane fouling in the MBRs
Both the CMBR and BEMR in Experiment I were operated at 20 L m2 h1 without backwashing. Fig. 3 illustrates the continual rise of TMP in both reactors as a function of operation time. The results further showed that the rate of rise in TMP significantly increased as HRT decreased, which could
HRT=6h
HRT=9 h
HRT=12 h
100
Table 2 e Characteristics of the PVDF ultrafiltration membrane (ZeeWeed-1).
Membrane material Module type Membrane pore size (mm) Membrane surface area (m2) Outer/inner diameter (mm) Maximum operating temperature ( C) Operating pH range Cleaning pH range Maximum operating pressure (kPa) Operating flux (L m2 h1)
Characteristics Polyvinylidene fluoride (PVDF) Hollow fiber 0.036 0.046 1.9/0.8 40 5.0e9.5 2.0e10.5 55 18e40
1500
COD (mgL-1)
Category
Influent BEMR Effluent CMBR Effluent BEMR COD Removal CMBR COD Removal
80
60
Pseudo steady-state condition for both BEMR and CMBR
1000
40
500
20
0
0 0
20
40
60
80
100
Time (d)
Fig. 2 e COD removal by BEMR and CMBR.
120
COD Removal (%)
Table 1 e Operation parameters of the lab-scale BEMR and CMBR.
4273
6 9
60
6
12
9
12
40
CMBR 6 HRT CMBR 9 HRT CMBR 12 HRT BEMR 6 HRT BEMR 9 HRT BEMR 12 HRT
20
0 0
50
100
150
200
250
Time (min)
Fig. 3 e Effects of HRT on TMP and membrane fouling of the BEMR and CMBR.
hasten membrane fouling. This was in agreement with previous studies (Cho et al., 2005; Chae et al., 2006; Meng et al., 2009). The TMP of the CMBR increased exponentially to 55 kPa within 39, 48, and 57 min at HRT of 6, 9, and 12 h, respectively, while the TMP of the BEMR reached 55 kPa in 135, 175, and
a
16
BEMR
> 100 kDa 10 - 100 kDa 1 - 10 kDa < 1 kDa
14
SMPc (mg L-1)
12
-1 Total SMPc (6 h) = 19.32 mg L -1 Total SMPc (9 h) = 12.98 mg L -1 Total SMPc (12 h) = 11.16 mg L
10 8 6
215 min at HRT of 6, 9, and 12 h, respectively. A higher MLSS content in the MBR has higher potential to cause membrane fouling because an increased amount of bound EPS will be released into the sludge flocs (Trussell et al., 2006). The BEMR sustained a longer period of operation before reaching the TMP of 55 kPa because the membrane was exposed to a much lower concentration of suspended solid (20e30 mg L1) in the separate membrane compartment. A long SRT in the BEMR allows development of slow-growing microorganisms (Ahmed et al., 2007) and specific microorganisms that could assimilate dead or inactive microorganisms (Han et al., 2005). These microorganisms are more capable of consuming macromolecules such as polysaccharides, carbohydrate, and protein as substrates and producing less biopolymers (Masse et al., 2006), key to reduce membrane fouling. Ahmed et al. (2007) also reported less membrane fouling with increasing SRT from 20 to 100 d. Besides, bacteria were found widely present on fouled membrane (Ng et al., 2006); the growth of bacteria on membrane that resulted in formation of foulants and biocake on the membrane surfaces was attributed to membrane fouling. Hijnen et al. (2009) studied the biofouling of membrane in spiral-wound membrane and the results showed biofouling of membrane occurs at very low concentrations of easily biodegradable organic compounds because of the microbial growth in the feed water.
a
Total SMPp (6 h) = 25.81 mg L Total SMPp (9 h) = 16.65 mg L
-1 -1
-1 Total SMPp (12 h) = 15.88 mg L
8 6 4
2
2 0 HRT = 9 h
CMBR
HRT = 12 h
HRT = 6 h
> 100 kDa 10 - 100 kDa 1 - 10 kDa < 1 kDa
14 12
Total SMPc (6 h) = 25.87 mg L 10
Total SMPc (9 h) = 18.18 mg L
-1
6
HRT = 9 h
16 14 12
-1
-1 Total SMPc (12 h) = 13.50 mg L
8
b
SMPp (mg L-1)
HRT = 6 h
SMPc (mg L-1 )
> 100 kDa 10 - 100 kDa 1 - 10 kDa < 1 kDa
10
4
16
BEMR
12
0
b
16 14
SMPp (mg L-1)
Trans-Membrane Pressure (kPa)
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 6 9 e4 2 7 8
CMBR
Total SMPp (6 h) = 41.59 mg L Total SMPp (9 h) = 31.91 mg L
HRT = 12 h
-1 -1
Total SMPp (12 h) = 23.90 mg L
-1
> 100 kDa 10 - 100 kDa 1 - 10 kDa < 1 kDa
10 8 6
4
4
2
2 0
0 HRT = 6 h
HRT = 9 h
HRT = 12 h
Fig. 4 e Carbohydrate components of SMP at various HRTs in a) BEMR, and b) CMBR.
HRT = 6 h
HRT = 9 h
HRT = 12 h
Fig. 5 e Protein components of SMP at various HRTs in a) BEMR, and b) CMBR.
4274
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 6 9 e4 2 7 8
Fractionation of SMP in CMBR and BEMR
Figs. 4 and 5 illustrate the SMPc (total carbohydrate) and SMPp (total protein) contents of the two MBRs at different HRTs. The total SMP (SMPc þ SMPp) was also calculated. The CMBR produced more SMP than the BEMR by 33%, 41%, and 28% at
Trans-Membrane Pressure Flux
40
40
30
30
Flux was maintained at about 20 LMH
-2
-1
Flux (L m h )
Backwashing
20
20
10
10
0 0
20
40
60
80
100
120
140
c
50
a) CMBR running 15 min and backwashing 1 min
50
50
c) BEMR running 15 min and backwashing 1 min Trans-Membrane Pressure Flux
40
0 160
30
30
Flux was maintained at about 20 LMH 20
20
10
10
0 0
20
40
60
40
Flux (L m-2 h-1)
Backwashing 30
30
Flux was maintained at about 20 LMH 20
20
10
10
0
0 0
50
100
150
Time (min)
200
250
300
50
100
120
140
0 160
50
d) BEMR running 30 min and backwashing 1 min Trans-Membrane Pressure Flux
40
40
Backwashing
Flux (L m-2 h-1)
Trans-Membrane Pressure Flux
40
d
50
b) CMBR running 30 min and backwashing 1 min
Trans-Membrane Pressure (kPa)
50
80
Time (min)
Time (min)
b
40
Backwashing
30
30
Flux was maintained at about 20 LMH 20
20
10
10
0
Trans-Membrane Pressure (kPa)
50
Trans-Membrane Pressure (kPa)
a
Flux (L m-2 h-1)
3.3.
HRT of 6, 9, and 12 h, respectively. In both MBRs, the SMP contents with the HRT of 6 h were higher than those with HRTs of 9 and 12 h. The total SMP contents in BEMR and in CMBR increased from 27 to 45 mg L1 and from 37 to 68 mg L1, respectively, when the HRT decreased from 12 to 6 h. At lower HRTs, a larger amount of SMP could be released by the abundant filamentous organisms in the bioreactors during their stationary and endogenous metabolism stage that could adversely impact the membrane (Choi et al., 2002). Fallah et al. (2010) compared the removal of styrene from synthetic wastewater by MBRs with HRTs of 18 h and 24 h and found that the SMP concentration increased when HRT was reduced to 18 h. The relationship between SRT and EPS and/or SMP has been previously studied relative to membrane fouling (Cho et al., 2005; Han et al., 2005; Liang et al., 2007). The bio-balls employed in the BEMR had longer SRT and good contact with supplied nutrients. This reduced cell death and production of EPS and SMP, resulting in enhanced membrane filterability (Ng et al., 2010). Meng et al. (2007) reported increased production of bound EPS and sludge viscosity when the foodto-microorganism ratio (F/M) and organic loading rate were increased. This may be due to the formation of bound EPS that was growth-related and produced in direct proportion to substrate utilization (Laspidou and Rittman, 2002).
Trans-Membrane Pressure (kPa)
Decreasing the HRT in CMBR might also decrease DO, which could lead to poor flocculation and filamentous bulking in the bioreactor (Liu and Liu, 2006). Meng et al. (2009) found that DO decreased as the HRT decreased from 10 to 12 h to 3e4 h, which increased growth of filamentous organisms in the flocs. Meng et al. (2007) found DO concentrations to be 3.8e6.5 mg L1 at HRT of 10e12 h and 0.2e1.5 mg L1 at HRT of 4e5 h. Microorganisms in bioreactors likely required high oxygen consumption to degrade organic matters and form new cellular materials. In addition, Meng et al. (2006) reported that overgrowth of filamentous bacteria could lead to a sharp increase of EPS, sludge viscosity, and sludge hydrophobicity, resulting in increased fouling. Jin et al. (2006) and Kim et al. (2006) reported more severe membrane fouling with low DO levels. Therefore, it is reasonable to conclude that higher DO and longer SRT conditions employed in the BEMR resulted in less membrane fouling (as measured by TMP).
0 0
50
100
150
200
250
300
Time (min)
Fig. 6 e The rise of TMP with different membrane filtration times: a) CMBR with 15-min filtration time, b) CMBR with 30-min filtration time, c) BEMR with 15-min filtration time, and d) BEMR with 30-min filtration time.
4275
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 6 9 e4 2 7 8
obstructive to membrane pores were less in the BEMR than in the CMBR. This was consistent with our previous findings on the positive effects of membrane fouling by proteins of low molecular weights. The protein likely narrowed membrane pores via adsorption, and fouling was exacerbated by denatured protein (Ng et al., 2010). With backwashing, the membrane process of BEMR could sustain a longer operation period at a higher permeate flux than could the CMBR. Operation duration and frequency of backwashing are central design parameters of MBRs. Jiang et al. (2005) found that an MBR using a 600-s membrane filtration process with 45-s backwashing was more efficient than one using a 2000-s membrane filtration process with 45-s backwashing. Therefore, to reduce fouling for the CMBR at an average flux of 20 L m2 h1, it is conceivable that operating with 15-min suction and 1-min backwash can be less fouling than with 30-min suction and 1-min backwash. The operation protocol may be varied to achieve a longer service run with a nominal backwashing interval. Fig. 7 shows results of Experiment III that contrast the performances of the two MBRs. Both MBRs were operated at 20 L m2 h1 with a service cycle of 15-min suction and 1-min backwash. In the CMBR, TMP threshold of 55 kPa was reached much sooner, necessitating chemical cleaning every 5 days (i.e., at day 10, 16, 21.5, 27, 32.5, and 38).
a
70
50
50
40
40
30
30
20
20
10
10
-2
-1
60
0 10
Membrane performance of BEMR and CMBR
20
30
0 40
Operating Time (d)
b
60
TMP Flux Chemical Cleaning
40
40
20
20
0 0
10
20
30
Trans-Membrane Pressure (kPa)
BEMR 60
Flux (L m-2 h-1)
Fig. 6 compares TMP increases in CMBR and BEMR operating at two different filtration/backwash cycles (15 min or 30 min of filtration plus 1 min of backwash) and cleaning. With the cycle of 15-min filtration plus 1-min backwash, the TMP of CMBR started at 3.6 kPa at the beginning of service and reached 11.2 kPa at the end of 10 service cycles; whereas with the cycle of 30-min filtration plus 1-min backwash, the TMP reached 23 kPa at the end (Fig. 6a and b). In contrast, the TMP of BEMR started at 2.4 kPa at the beginning of service and was stable after backwashing of each cycle; it increased much more gradually and slowly reaching 3.9 kPa and 6.2 kPa at the ends of 10 service cycles with 15 and 30 filtration durations, respectively. As the operation time increased, more colloids and SMPs with different MW particles accumulated on the membrane that made it increasingly difficult to remove by backwashing. The BEMR produced less SMP and displayed less fouling than did the CMBR. SMP <100 kDa and protein that were more
CMBR
60
0
3.4.
Chemical Cleaning
70
Trans-Membrane Pressure (kPa)
TMP Flux
Flux (L m h )
Figs. 4 and 5 also show that the SMP concentration of carbohydrate and protein gradually decreased in each MW fraction when HRTs decreased, except for the SMPp at HRT of 9 h. The results reveal that the dominating MW fraction of SMP in both MBRs was G-2 (10e100 kDa) with all tested HRTs. The G-2 fraction contributed approximately 38e54% of the total SMP concentrations in the BEMR and 40e43% in the CMBR despite the effects of HRTs. These findings were supported by studies of Pan et al. (2010) and Janga et al. (2007) proved that the bulk of SMP carbohydrate and SMP protein had MW >30 kDa and 10 kDa, respectively. However, Malamis and Andreadakis (2009) and Ng et al. (2010) stated that the SMP <1 kDa was predominant. The discrepancy might have arisen from different methods of SMP separation and analysis, operation conditions, and feed characteristics being employed. Protein was found to be the major component of SMP rather than carbohydrate in BEMR and CMBR. In BEMR and CMBR, SMPp concentrations varied from 56 to 59% and from 62 to 64%, respectively, under the studied HRTs. The results were consistent with several past studies (Le-Clech et al., 2003; Yigit et al., 2008; Malamis and Andreadakis, 2009; Ng et al., 2010) that showed protein as the main component of SMP. Further, Viero et al. (2007) reported that SMPc was the major component that formed a cake layer or a gel layer on the membrane. Owing to the major occurrence of SMPp in our study, we believe protein was the main component that caused pore blocking of the membrane. Therefore, higher DO and longer SRT in the BEMR led to less SMP and reduced membrane fouling. It is important to analyze the molecular fractions larger and smaller than 100 kDa in the BEMR and CMBR. The membrane (GE ZeeWeed-1) used in this study had a pore size of 0.036 mm approximating 100 kDa. At different HRTs, the total SMP contents (SMPp and SMPc) larger than 100 kDa in the BEMR were 15e23% while in the CMBR they were 17e22%. Lin et al. (2009) reported that particles of smaller molecular weight could easily penetrate the membrane pores by adsorption in a 100-kDa membrane. Therefore, the major fouling mechanism by SMP (more than 70%) is likely caused by pore blocking, rather than by the formation of cake layer on the membrane surface.
0 40
Operating Time (d)
Fig. 7 e The rise of TMP with filtration time in a) CMBR, and b) BEMR.
4276
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 6 9 e4 2 7 8
The TMP could be restored to the initial value of 3.5e4.0 kPa, which indicated complete removal of most foulants on the membrane surface and in the pores. Jeison and van Lier (2007) studied an anaerobic submerged membrane reactor over 200 days and found that external, physical cleaning was necessary because the consolidated cake could not be removed by back-flush cycles. Fallah et al. (2010) also reported that fouling by pore blocking could only be eliminated by chemical cleaning. In the BEMR, TMP was below 10 kPa for 10 days; it gradually increased till day 29 and then increased at a faster rate reaching 55 kPa at 39 days (Fig. 7(b)). The increase of TMP was attributed to colloids and soluble EPS being deposited on the membrane. The operation time of the BEMR before chemical cleaning needed was 7 times longer than that of the CMBR. This was attributed to less MLSS and less total SMP in the BEMR. Based on these results, BEMR appears to be less susceptible to membrane fouling due to reduced SMP and MLSS contents and therefore requires less-frequent cleaning.
4.
Conclusions
In this study, we have compared the new BEMR with a conventional CMBR on performance, SMP production, and service duration, and conclude the following: 1. In both MBRs, TMP arose faster as HRT decreased. The BEMR was less susceptible to fouling, and it sustained a longer service duration than did the CMBR (39 days vs. 5 days). 2. The BEMR produced less SMP than did CMBR (34e48% less protein and 16e29% less carbohydrate) due to slow-growing microorganisms with long SRT in the new bioreactor. 3. Both MBRs produced SMP of 10e100 kDa primarily of protein (59% in BEMR and 64% in CMBR), which likely caused membrane pores clogging. 4. BEMR appears promising in controlling membrane fouling, requiring less frequent chemical cleaning, and being more economical to operate.
Acknowledgments This work was partially funded by the National Science Council of the Republic of China (NSC98-2221-E-002-029-MY3 and NSC99-2811-E-002-042).
references
Ahmed, Z., Cho, J., Lim, B.R., Song, K.G., Ahn, K.H., 2007. Effects of sludge retention time on membrane fouling and microbial community structure in a membrane bioreactor. Journal of Membrane Science 287 (2), 211e218. Artiga, P., Oyanedel, V., Garrido, J.M., Mendez, R., 2005. An innovative biofilm-suspended biomass hybrid membrane bioreactor for wastewater treatment. Desalination 179 (1e3), 171e179.
Barker, D.J., Stuckey, D.C., 1999. A review of soluble microbial products (SMP) in wastewater treatment systems. Water Research 33 (14), 3063e3082. Chae, S.R., Ahn, Y.T., Kang, S.T., Shin, H.S., 2006. Mitigated membrane fouling in a vertical submerged membrane bioreactor (VSMBR). Journal of Membrane Science 280 (1e2), 572e581. Cho, J., Song, K.G., Lee, S.H., Anh, K.H., 2005. Sequencing anoxic/ anaerobic membrane bioreactor (SAM) pilot plant for advanced wastewater treatment. Desalination 178 (1e3), 219e225. Choi, J.G., Bae, T.H., Kim, J.H., Tak, T.M., Randall, A.A., 2002. The behavior of membrane fouling initiation on the crossflow membrane bioreactor system. Journal of Membrane Science 203 (1e2), 103e113. Drews, A., Evenblij, H., Rosenberger, S., 2005. Potential and drawbacks of microbiology-membrane interaction in membrane bioreactors. Environmental Progress 24 (4), 426e433. Drews, A., Kraume, M., 2005. Process improvement by application of membrane bioreactors. Chemical Engineering Research and Design 83 (A3), 276e284. Fallah, N., Bonakdarpour, B., Nasernejad, B., Alavi Moghadam, M. R., 2010. Long-term operation of submerged membrane bioreactor (MBR) for the treatment of synthetic wastewater containing styrene as volatile organic compound (VOC): effect of hydraulic retention time (HRT). Journal of Hazardous Materials 178 (1e3), 718e724. FrØlund, B., Palmgren, R., Keidng, K., Nielsen, P.H., 1995. Extraction of extracellular polymers from activated sludge using a cation exchange resin. Water Research 30 (8), 1749e1758. Han, S.S., Bae, T.H., Jang, G.G., Mak, T.M., 2005. Influence of sludge retention time on membrane fouling and bioactivities in membrane bioreactor system. Process Biochemistry 40, 2393e2400. 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-would membranes. Environmental Science and Technology 43, 4890e4895. Ivanonic, I., Leiknes, T., Ødegaard, H., 2006. Influence of loading rates on production on characteristics of retentate from a biofilm membrane bioreactor (BF-MBR). Desalination 199 (1e3), 490e492. Janga, N., Ren, X., Kim, G., Ahn, C., Cho, J., Kim, I.S., 2007. Characteristics of soluble microbial products and extracellular polymeric substances in the membrane bioreactor for water reuse. Desalination 202 (1e3), 90e98. Jarusutthirak, C., Amy, G., 2006. Role of soluble microbial products (SMP) in membrane fouling and flux decline. Environmental Science and Technology 40 (3), 969e974. Jeison, D., van Lier, J.B., 2007. Cake formation and consolidation: main factors governing the applicable flux in anaerobic submerged membrane bioreactors (AnSMBR) treating acidified wastewaters. Separation and Purification Technology 56 (1), 71e78. Jiang, T., Kennedy, M.D., Guinzbourg, B.F., Vanrolleghem, P.A., Schippers, J.C., 2005. Optimizing the operation of a MBR pilot plant by quantitative analysis of the membrane fouling mechanism. Water Science and Technology 51 (6e7), 19e25. Jin, Y.L., Lee, W.N., Lee, C.H., Chang, I.S., Huang, X., Swaminathan, T., 2006. Effect of DO concentration on biofilm structure and membrane filterability in submerged membrane bioreactor. Water Research 40 (15), 2829e2836. Kim, H.Y., Yeon, K.M., Lee, C.H., Lee, S., Swaminathan, T., 2006. Biofilm structure and extracellular polymer substances in low and high dissolved oxygen membrane bioreactors. Separation and Science Technology 41, 1213e1230. Laspidou, C.S., Rittman, B.E., 2002. A unified theory for extracellular polymeric substances, soluble microbial
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 6 9 e4 2 7 8
products, and active and inert biomass. Water Research 36 (11), 2711e2720. Le-Clech, P., Chen, V., Fane, T.A.G., 2006. Fouling in membrane bioreactors used in wastewater treatment. Journal of Membrane Science 284 (1e2), 17e53. Le-Clech, P., Jefferson, B., Chang, I.S., Judd, S.J., 2003. Critical flux determination by the flux-step method in a submerged membrane bioreactor. Journal of Membrane Science 227 (1e2), 81e93. Lee, W.N., Kang, I.J., Lee, C.H., 2006. Factor affecting filtration characteristics in membrane-coupled moving bed biofilm reactor. Water Research 40 (1), 1827e1835. Leiknes, T., Ødegaard, H., 2007. The development of biofilm membrane bioreactor. Desalination 202 (1e3), 135e143. Liang, S., Lui, C., Song, L., 2007. Soluble microbial products in membrane bioreactor operation: behaviors, characteristics, and fouling potential. Water Research 41 (1), 95e101. Lin, C.F., Huang, Y.J., Hao, O.J., 1999. Ultrafiltration processes for removing humic substances: effect of molecular weight fractions and PAC treatment. Water Research 33 (5), 1252e1264. Lin, C.F., Lin, A.Y.C., Panchangam, S.C., Tsai, C.Y., 2009. Effect of mass retention of dissolved organic matter and membrane pore size on membrane fouling and flux decline. Water Research 43 (2), 389e394. Liu, Y., Liu, Q.S., 2006. Causes and control of filamentous growth in aerobic granular sludge sequencing batch reactors. Biotechnology Advances 24 (1), 115e127. Lowry, O.H., Rosebrough, N.J., Farr, A.L., Randall, R.J., 1951. Protein measurement with the Folin phenol reagent. Journal of Biological Chemistry 193 (1), 265e275. Malamis, S., Andreadakis, A., 2009. Fractionation of proteins and carbohydrates of extracellular polymeric substances in a membrane bioreactor system. Bioresource Technology 100 (13), 3350e3357. Masse, A., Sperandio, M., Cabassud, C., 2006. Comparison of sludge characteristics and performance of a submerged membrane bioreactor and an active sludge process at high solids retention time. Water Research 40, 2405e2415. 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 (6), 1489e1512. Meng, F., Shi, B., Yang, F., 2007. Effect of hydraulic retention time on membrane fouling and biomass characteristics in submerged membrane bioreactors. Bioprocess and Biosystems Engineering 30 (5), 359e367. Meng, F., Zhang, H., Yang, F., Li, Y., Xiao, J., Zhang, X., 2006. Effect of filamentous bacteria on membrane fouling in submerged membrane bioreactor. Journal of Membrane Science 272 (1e2), 161e168. Meng, F., Zhang, H., Li, Y., Zhang, X., Yang, F., Xiao, J., 2005. Cake layer morphology in microfiltration of activated sludge wastewater based on fractal analysis. Separation and Purification Technology 44 (3), 250e257. Ni, B.J., Zheng, R.J., Fang, F., Xie, W.M., Sheng, Q.P., Yu, H.Q., 2010. Fractionating soluble microbial products in the activated sludge process. Water Research 44 (7), 2292e2302. Ng, K.K., Lin, C.F., Shaik, K.L., Panchangam, S.C., Hong, A.P. K., Yang, P.Y., 2010. The effect of soluble microbial products on membrane fouling in a fixed carrier biological system. Separation and Purification Technology 72 (1), 98e104. Ng, H.Y., Tan, T.W., Ong, S.L., 2006. Membrane fouling of submerged membrane bioreactors: impact of mean cell residence time and the contributing factors. Environmental Science and Technology 40 (8), 2706e2713.
4277
Pan, J.R., Su, Y., Chihpin, H., 2010. Characteristics of soluble microbial products in membrane bioreactor and its effect on membrane fouling. Desalination 250 (2), 778e780. Qian, X., Yang, P.Y., Maekawa, T., 2001. Evaluation of direct removal if nitrate with EMMC technology using ethanol as carbon source. Water Environment Research 73 (5), 584e589. Ramesh, A., Lee, D.J., Wang, M.L., Hsu, J.P., Juang, R.S., Hwang, K.J., Liu, J.C., Tseng, S.J., 2006. Biofouling in membrane bioreactor. Separation and Science Technology 41 (7), 1345e1370. Raukjaer, K., Hvitved, J., Nielsen, P.H., 1994. Measurement of pools of protein, carbohydrate and lipid in domestic wastewater. Water Research 28 (2), 251e262. Rosenberger, S., Laabs, C., Lesjean, B., Gnirss, R., Amy, G., Jekel, M., Schrotter, J.C., 2006. Impact of colloidal and soluble organic material on membrane performance in membrane bioreactors for municipal wastewater treatment. Water Research 40 (4), 710e720. Trussell, R.S., Merlo, R.P., Hermanowicz, S.W., Jenkins, D., 2006. The effect of organic loading on process performance and membrane fouling in a submerged membrane bioreactor treating municipal wastewater. Water Research 40 (14), 2675e2683. Viero, A.F., Sant’ Anna Jr., G.L., Nobrega, R., 2007. The use of polyetherimide hollow fibres in a submerged membrane bioreactor operating with air backwashing. Journal of Membrane Science 302, 127e135. Yang, P.Y., Cao, K., Kim, S.J., 2002. Entrapped mixed microbial cell process for combined secondary and tertiary wastewater treatment. Water Environment Research 74 (3), 226e234. Yang, P.Y., Su, R., Kim, S.J., 2003. EMMC process for combined removal of organics, nitrogen and odor producing substance. Journal of Environmental Management 69 (4), 381e389. Yang, S., Yang, F., Fu, Z., Lei, R., 2009. Comparison between a moving bed membrane bioreactor and a conventional membrane bioreactor on membrane fouling. Bioresource Technology 100 (24), 6655e6657. Yigit, N.O., Harman, I., Civelekoglu, G., Koseoglu, H., Cicek, N., Kitis, M., 2008. Membrane fouling in a pilot-scale submerged membrane bioreactor operated under various conditions. Desalination 231 (1e3), 124e132. Yu, T.H., Lin, A.Y.C., Shaik, K.L., Lin, C.F., Yang, P.Y., 2009. Removal of antibiotics and non-steroidal anti-inflammatory drugs by extended sludge age biological process. Chemosphere 77 (2), 175e181. Zhou, W., Wu, B., She, Q., Chi, L., Zhang, Z., 2009. Investigation of soluble microbial products in a full-scale UASB reactor running at low organic loading rate. Bioresource Technology 100 (14), 3471e3476.
Nomenclature AMW: apparent molecular weight BEMR: bio-entrapped membrane reactor BSA: bovine serum albumin COD: chemical oxygen demand (mg L1) CMBR: conventional membrane bioreactor DO: dissolved oxygen (mg L1) EPS: extracellular polymeric substances (mg L1) F/M: food to microorganisms ratio GFC: gel filtration chromatography HRT: hydraulic retention time (h) MBBR: moving bed biofilm reactor MBR: membrane bioreactor MBMBR: moving bed membrane bioreactor MLSS: mixed liquor suspended solid (mg L1) MW: molecular weight NaOCl: sodium hypochlorite PVDF: polyvinylidene fluoride
4278
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 6 9 e4 2 7 8
SMP: soluble microbial products (mg L1) SMPc: fraction of carbohydrate contained in the sludge solution (mg L1) SMPp: fraction of protein contained in the sludge solution (mg L1)
SRT: sludge retention time (day) SS: suspended solids (mg L1) TMP: trans-membrane pressure (kPa) UF: Ultrafiltration
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 7 9 e4 2 8 6
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Effect of average flow and capacity utilization on effluent water quality from US municipal wastewater treatment facilities Scott R. Weirich a,*, JoAnn Silverstein a, Balaji Rajagopalan a,b a b
Department of Civil, Environmental, and Architectural Engineering, University of Colorado, Boulder, Colorado, United States Cooperative Institute for Research in Environmental Sciences, University of Colorado, Boulder, Colorado, United States
article info
abstract
Article history:
There is increasing interest in decentralization of wastewater collection and treatment
Received 15 December 2010
systems. However, there have been no systematic studies of the performance of small
Received in revised form
treatment facilities compared with larger plants. A statistical analysis of 4 years of
13 April 2011
discharge monthly report (DMR) data from 210 operating wastewater treatment facilities
Accepted 3 June 2011
was conducted to determine the effect of average flow rate and capacity utilization on
Available online 12 June 2011
effluent biochemical oxygen demand (BOD), total suspended solids (TSS), ammonia, and fecal coliforms relative to permitted values. Relationships were quantified using general-
Keywords:
ized linear models (GLMs). Small facilities (40 m3/d) had violation rates greater than 10
Wastewater treatment
times that of the largest facilities (400,000 m3/d) for BOD, TSS, and ammonia. For facilities
Reliability
with average flows less than 40,000 m3/d, increasing capacity utilization was correlated
Risk
with increased effluent levels of BOD and TSS. Larger facilities tended to operate at flows closer to their design capacity while maintaining treatment suggesting greater efficiency. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
Centralized wastewater collection, treatment, and disposal in the US began during the latter 19th Century with an effort to protect public health in cities and to mitigate nuisance conditions brought about through lack of local disposal sites for residential waste. In addition, the availability of piped water and flush toilets enabled “wet” collection and transport of wastes. High population density in large urban centers limited land available for disposal of the untreated waste, prompting construction of sewers. Between 1850 and 1920, the number of cities with more than 50,000 people increased from just under 400 to over 2700, and the population served by combined or sanitary sewers increased from 1 million to 25 million (Burian et al., 2000). The 1972 Federal Water Pollution
Control Act (FWPCA) amendments to the Clean Water Act brought sweeping changes to wastewater treatment with a discharge permit system to protect surface water quality as its foundation. Approximately $80 billion in Federal investments through the construction grants program resulted in the construction of centralized secondary treatment plants, using the sewer infrastructure that had been built over the previous 75 years. In addition to economies of scale, the costs of compliance with National Pollutant Discharge Elimination System (NPDES) permits including treatment to higher levels and extensive monitoring may have favored larger facilities supported by fees from residential, commercial and industrial users. Decentralized systems were dominated by on-site treatment in low density rural areas (U.S. Environmental Protection
* Corresponding author. UCB 428, Boulder, CO 80309-0428, United States. E-mail address:
[email protected] (S.R. Weirich). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.06.002
4280
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 7 9 e4 2 8 6
Agency [USEPA], 1997), but recent growth of suburban and exurban areas has increased use of on-site systems in these medium density developments. Data from the 2000 US Census indicated that one-third of new homes and the majority of mobile homes are served by on-site systems (USEPA, 2002). However, on-site systems are unable to meet common discharge standards for nitrogen met by centralized systems, and as a result are major contributors of nitrogen to the aquatic environment (Oakley et al., 2010). The need to serve communities too dense for on-site systems but far from existing sewers and centralized treatment facilities and location of communities near nitrogen sensitive watersheds has brought new interest in smaller collection systems served by satellite wastewater treatment plants. Factors such as the cost of building out collection systems and pumping wastewater, improvements in small system technology, and automated operation have led organizations such as National Decentralized Wastewater Resources Capacity Development to advocate for decentralized systems and small satellite plants (National Decentralized Water Resources Capacity Development Project, 2009). Decreases in housing density affect the economies of scale derived from large centralized wastewater treatment through increased collection system costs (Carruthers and Ulfarsson, 2003). For smaller communities, the capital cost of conventional gravity sewers on average was found to be four times the cost of treatment with a similar relation for operation and maintenance costs. This is due to both the longer piping distances in lower density developments and the increased need for lift stations (WEF, 2008). Water reuse can also provide a significant incentive for smaller collection systems. In Denver, for example, pumping from a centralized facility constitutes 50% of the costs of reuse water (Good, 2006). Reducing the service area of a wastewater facility would reduce distribution system costs and energy consumption provided there were local opportunities for reuse. The EPA has estimated that if current documented needs in the US are met there will be 1552 new treatment plants of which 53% will serve communities of fewer than 10,000 people or 4000 m3/d wastewater flow (USEPA, 2008b). Many of these facilities are expected to replace failing on-site treatment systems. Total small community needs are $17 billion, 9% of the total wastewater monetary need (USEPA, 2008b). There are regulatory issues that differentially impact small wastewater systems. One is significant fixed costs to apply for a discharge permit for a new or expanded plant. These fixed costs place a higher burden on smaller systems and thus are a disincentive to redesign and improve facilities. Small systems with significant dilution of their discharges may benefit from relaxed permit standards, particularly for constituents such as nitrogen. Furthermore increasingly limited resources for enforcement and focus on large dischargers may mean less regulatory attention paid to small facilities. Lack of oversight could result in fewer facility improvements or reduced effluent quality. Overall, the potential for proliferation of small wastewater treatment facilities coupled with constraints on monitoring and enforcement provides the rationale to investigate the effect of facility size on treatment performance.
Earlier statistical evaluations of treatment reliability were done in the 1970s in the midst of the Constructions Grants program. Effluent BOD and suspended solids (SS) data from 37 US treatment plants was characterized with lognormal distributions in order to develop a coefficient of reliability (COR), the probability of a plant achieving mean effluent quality at a selected fraction of the discharge standard. For any plant, prediction of the COR relied on knowing, or estimating, the coefficient of variance for BOD and SS (Niku et al., 1979). A comparison of treatment reliability as a function of process type was performed on 166 plants in Brazil which also used a lognormal distribution of effluent data and found that activated sludge processes achieved the highest reliability while septic tanks had the lowest, although plant capacity was not explicitly considered (Oliveira and Von Sperling, 2008). Oakley et al. (2010) found that on-site treatment systems were significantly less capable of meeting limits on watershed nitrogen loads than centralized biological nutrient removal (BNR) systems. Linear regression analysis was used in a study of the effect of operations on individual treatment plant performance with the conclusion that no single or consistent group of factors, including flow, could explain individual plant variability in BOD and SS removal (Niku and Schroeder, 1981b). However, lack of identifiable factors affecting performance may have been partly due to the relatively narrow range of operating conditions at any individual plant. Niku and Schroeder (1981a) reported poor correlations between arithmetic mean annual flow and annual mean and standard deviation values for BOD and suspended solids in a sample of 43 activated sludge treatment plants ranging from 2000 to 800,000 m3/ d (0.56e209 mgd). Overall, applicability of previous studies predicting the reliability of treatment plants is limited in that the methods require prior knowledge of the inherent mean and variance of effluent quality parameters for an individual facility or process type. As such, they do not provide a general basis for comparing the reliability of a network of decentralized treatment plants to a centralized system or the risk of excess contaminant loading to a watershed. Mathematical optimization models have been developed to evaluate the effect of the degree of centralization, but the major objective function was cost with assumed economy of scale a major factor (Cunha et al., 2009; Wang and Jamieson, 2002; Voutchkov and Boulos, 1993). To address the need for predictive models of reliability, a statistical study was designed to test whether a relationship exists between average monthly flow, capacity utilization and effluent constituent levels and violation probability. The goal of the study is to provide guidance to planners, regulators, and utility managers in defining service areas and facility conditions which will provide economical treatment with reduced environmental risk.
1.1.
Generalized linear models
Effluent concentration of a constituent and relative concentration, normalized to permit limits, will have both systematic variabilities, represented as its relationship to factors like facility size and capacity utilization, and random variability
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 7 9 e4 2 8 6
arising from changes in effluent water quality within a single facility over time and inherent differences between facilities including influent characteristics, facility age and process type. Factors which may be associated with plant size such as equipment, maintenance levels, labor quality and hours could lead to variation in treatment performance (Niku and Schroeder, 1981b). Since the work of Niku et al., the generalized linear model (GLM) has been developed as a flexible statistical method of accurately modeling a wide variety of data including nonnormal distributions and discrete variables. In a GLM, the response or the dependent variable Y can be assumed to be a realization from any distribution in the exponential family with a set of parameters (McCullagh and Nelder, 1989). Thus positively-skewed, nonnegative data such as effluent concentrations of constituents can be modeled with a gamma or lognormal distribution and violation probability can be directly modeled with a binomial distribution using the same statistical method. The GLM enables simultaneous consideration of more than one independent variable without the assumption of linear relationships between independent and dependent variables. In GLM, a smooth and invertible link function transforms the conditional expectation of Y to a set of predictors. GðEðYÞÞ ¼ h ¼ f ðXÞ þ e ¼ XbT þ e
(1)
T
b is the transposed vector of model parameters, X is the set of predictors or independent variables, E(Y ) is the expected value of the response variable, e is the error, and G(.) is the link function. The ability to choose a distribution for Y and the associated link function allow GLM to model a wide variety of data. For skewed variables with a lower bound of 0 such as effluent concentrations and relative concentrations, the gamma distribution with the inverse link function is appropriate; for binary variables such as modeling probability of violations, the binomial distribution with the logit link function is appropriate; for discrete variables such as number of violations over a given period the Poisson distribution is appropriate. If a normal distribution is assumed with an identify link function it collapses to a linear model e see McCullagh and Nelder (1989) for information about other distributions and link functions. After choosing a distribution and link function, the model parameters are estimated using an iterated weighted least squares (IWLS) method that maximizes the likelihood function as opposed to an ordinary least squares method used in linear modeling. The best model is chosen based on the Akaike Information Criteria (AIC, Akaike, 1974) by comparing models fit using all possible subsets of predictors. For each model, the AIC is computed as: AIC ¼ 2k 2L
(2)
where L is the logarithm of the likelihood function of the model with the predictor subset under consideration obtained from the IWLS procedure mentioned above and k is the number of parameters to be estimated in this model. The model with the lowest AIC is taken to be the ‘best model’. Models can also be tested for significance against a null model or an appropriate subset model using a chi-squared test.
2.
4281
Methods
To analyze the effect of treatment facility size and capacity utilization on effluent quality and violation history, data from the Environmental Protection Agency’s Integrated Compliance Information System (ICIS) was used. ICIS contains enforcement and compliance information for over 10,000 wastewater facilities with NPDES permits in 28 states and US territories (USEPA, 2008a). The ICIS database is gradually replacing the older Permit Compliance System (PCS); hence it does not have data for facilities in the 22 states that have not yet switched their reporting to the newer system. For those states with data, however, the most recent 2e5 years of discharge monthly reports (DMR) are available for most facilities. The data include all required reporting for each facility, including effluent concentrations for permitted constituents, influent measurements, flow through the plant, and the permitted discharge limits each month. To reduce data processing time and storage requirements, a systematic sample consisting of 5% of the ICIS database, (629 facilities) was used for analysis. The data set was further reduced by filtering out facilities with insufficient data for analysis of each of the four constituents, BOD, TSS, ammonia, and fecal coliforms, resulting in four separate data sets. The data set for BOD contains 209 facilities, TSS has 211, ammonia has 110, and fecal coliforms has 109 with an average of 41 months of data per facility.
2.1.
Prediction of effluent constituent levels
An important criterion related to treatment performance is effluent concentration, but permit standards vary considerably between plants due to factors such as receiving water quality, dilution factor, location, and season. Plants may be designed to meet current permit levels or anticipated future permit levels. As a result of these local differences, the absolute concentration of a constituent in the effluent is expected to differ between treatment facilities of equivalent treatment performance and reliability. To account for this the relative concentration was selected to be the dependent variable for regression, where relative concentration is the reported average monthly discharge concentration for a given constituent divided by the discharge permit standard. Relative concentration is greater than zero and positivelyskewed so the gamma function and its associated canonical link function, the inverse, was selected for GLM modeling of effluent constituent levels resulting in the following equation. 1 ¼ XbT þ e EðRÞ
(3)
where E(R) > 0 is the predicted effluent concentration of the constituent (BOD, TSS, ammonia or fecal coliforms), X is the matrix of independent variables, e is the error, and bT is the transposed vector of model parameters which are estimated following the methods described above. To determine the effect of facility size on treatment performance two independent variables have been chosen. First is the logarithm of the average monthly flow rate, A. Flow rates of facilities in the data sets vary from 1 to 335,000 m3/d. It
4282
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 7 9 e4 2 8 6
was hypothesized that plant performance could also be influenced by over- or under-loading, so the second independent variable is capacity utilization, C, defined as the reported monthly average flow rate divided by the design flow rate. The product of these two variables, AC, is also included as an independent variable, and all combinations of independent variables are compared using AIC. Process type was explicitly ignored as an independent variable due to lack of data and a desire to quantify performance explicitly as related to decentralization. Analysis was performed using R, a free software package for statistical computing and graphics.
2.2.
using the gamma distribution and the inverse link function using the same procedure described in section 2.1.
3.
Results of GLM analysis for prediction of effluent BOD as a function of plant flow and capacity utilization are presented in detail. Since the same procedure was followed for TSS, ammonia and fecal coliforms, results for these constituents have been summarized to allow discussion of differences in effluent trends among the four constituents.
Probability of violation 3.1.
To quantify risk the frequency and magnitude of permit violations were modeled. Permit standards are based on scientific water quality criteria adopted to protect aquatic life and other uses of receiving waters, and therefore the probability of violations are a reasonable indicator of risk of significant adverse effects to the receiving water. The probability of a violation also can be considered a second indicator of treatment plant reliability. Permit violations subject the plant owner/operator to regulatory penalties, including fines. To model violation frequency the response variable, effluent concentration, was converted to a binomial variable where 1 represented a permit violation and 0 represented no violation. The GLM is fitted using a binomial distribution with the logit link function as follows: EðVÞ ¼ XbT þ e ln 1 EðVÞ
(4)
where E(V) is the probability of a violation that ranges between 0 and 1, and other terms are as described for equation (3). With the fitted best model the risk of violations can be estimated.
2.3.
Resuts and discussion
Violation magnitude
A second component of risk is the magnitude of violations. Large exceedances of discharge standards could have a significant effect on receiving water quality, especially in cases where there is little dilution, sensitive aquatic habitat, or proximate human use. The relative discharge values exceeding the threshold are obtained from the data and best GLM model based on the three independent variables is fitted
Effluent concentration model e BOD
Flow rates of the 209 facilities used for BOD analysis ranged from 1 m3/d (0.001 MGD) to 335,000 m3/d (100 MGD) and capacity utilization rates ranged from 5% to 180%. Interestingly, 13% of those facilities average flow rates above their permitted capacity. Because EPA regulations require a plant to begin the redesign phase when a facility averages more than 85% of its design capacity, these facilities are in violation of that portion of their permit. Exceeding the hydraulic capacity of plants turns out to be a significant factor in effluent quality for smaller plants, as will be discussed below. Using these 209 facilities, GLMs to predict relative BOD concentration were fit for all possible combinations of the independent variables (logarithm of average flow and capacity utilization), as described in section 2.1. The AIC values of each model were compared, as shown in Table 1, and CondAþCþAC was identified as the best model. This model uses both independent variables as well as the nonlinear product and was significantly better at fitting effluent data than the unconditional model and CondA and CondAþC at alpha ¼ 0.05. This model is used for subsequent analysis. A generalized linear model from the gamma family was fit to the data using the inverse link function. Thus, the expected value of the relative effluent BOD, E(R), is modeled as: 1=EðRBOD Þ ¼ 3:03 þ 0:0691A þ 0:308C þ 0:176AC
(5)
As shown in (5), positive coefficients indicate a negative correlation between the independent variable and response variable so large and highly utilized facilities have lower expected effluent BOD than small facilities operating close to
Table 1 e GLM functions, parameters, bT, and associated AIC values for prediction of relative effluent BOD. R [ predicted monthly average BOD (mg/l); A [ log (average monthly flow) m3/d; C [ fraction of hydraulic capacity utilized; se [ standard error.
1/R¼
Uncond
CondA
CondC
CondAC
b0
b0 þ b1A
b0 þ b2C
b0 þ b3AC
b0 (se) 2.62 (0.0332) 2.79 (0.0405) 2.83 (0.0714) 2.79 (0.0377) e 0.143 (0.0160) e e b1 (se) e e 0.299 (0.0869) e b2 (se) e e e 0.266 (0.0217) b3 (se) AIC 3 175 26 290 a Term not significant at a ¼ 0.05.
CondAþC
CondAþAC
CondCþAC
b0 þ b1A þ b2C b0 þ b1A þ b3AC b0 þ b2C þ b3AC 3.26 (0.0829) 0.182 (0.0177) 0.588 (0.0870) e 269
2.79 (0.0406) 0.00744 (0.0254)a e 0.259 (0.0332) 288
2.89 (0.0678) e 0.155 (0.0849)a 0.254 (0.0222) 295
CondAþCþAC b0 þ b1A þ b2C þ b3AC 3.03 (0.0989) 0.0691 (0.0343) 0.308 (0.114) 0.176 (0.0450) 302
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 7 9 e4 2 8 6
4283
or over their permitted capacity, as shown in Fig. 1. More specifically, small facilities (40 m3/d) are predicted to discharge BOD that averages 40% or more of permit limits while predicted effluent BOD from large facilities (400,000 m3/ d) is consistently 33% of permit limits. Furthermore, for facilities 40,000 m3/d and larger, capacity utilization has almost no effect on the effluent BOD while for smaller facilities increasing capacity utilization is associated with increasing relative effluent BOD.
3.1.1.
Comparison of model and actual data
A boxplot shows actual effluent data sorted by facility flow rate (Fig. 2). The whiskers of the boxplot indicate the 5th and 95th percentiles, and not the interquartile range (IQR) times 1.5 as is common for boxplots, and the black squares show the predicted effluent level for the average flow rate and capacity utilization in that size range. The predicted values consistently fall between the first and third quartiles and also follow the trend shown in the median values up to 40,000 m3/d, specifically that there is a decrease in average BOD discharged as plants increase in size. Above 40,000 m3/d, however, there is a significant increase in median effluent BOD not predicted by the model. Both the IQR and the whisker length decrease as facilities get larger, and it is especially notable that the 95th percentile effluent BOD is above the permit limit for the two smallest size categories. By contrast, the largest facilities have the highest median discharge but the 95th percentile is furthest below the permit limit, showing that median discharge is not related to permit violations in the same way for larger facilities as it is for smaller plants. The latter result suggests that large facilities discharge closer to their permit limits on average but also have fewer violations, possibly due to more consistent treatment.
Fig. 2 e Boxplot showing variation of BOD discharges by facility size where whiskers show 5% and 95% percentiles and black square is the model prediction using the average flow and capacity utilization for facilities in the size range.
The large variation in BOD discharges, both for facilities of a given size range as shown but also within individual facilities, indicates that flow and capacity utilization are not sufficient to predict a facility’s performance in any given month; however, for comparison of the BOD removal among the entire data set or prediction for a single facility over a long period of time, the GLM does provide a good estimate of effluent BOD.
3.2.
BOD permit limit violations
While average effluent levels provide one measure of treatment performance, permit violations may be a better measure of the risk of significant BOD release to receiving waters. As Fig. 2 shows, though the largest facilities have the highest median relative effluent BOD they have the fewest discharges above their permitted values. To directly model probability of permit violations the data for effluent BOD are transformed into a 1 for effluent BOD exceeding the permit limit e a violation, or a 0 indicating no violation. As described in section 2.2, a binomial GLM was fit to the data using the logit link function. The model with the best set of predictors (lowest AIC) is: EðVBOD Þ ¼ 3:35 þ 0:128A þ 0:284AC ln 1 EðVBOD Þ
Fig. 1 e Predicted relative effluent BOD concentration versus average flow and capacity utilization.
(6)
The negative coefficients indicate a negative correlation for average flow and for the combined term, meaning larger facilities and more highly utilized facilities have lower violation rates as shown in Fig. 3. Consistent with the average effluent BOD data, as facility size increases the predicted fraction of months in violation decreased, with small facilities (40 m3/d) violating their BOD permits in more than 6.6% of months while large facilities (400,000 m3/d) violate BOD limits less than 2.2% of the time.
4284
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 7 9 e4 2 8 6
The lower whisker and first quartile fall at zero violations for all facility size ranges, and even the median is zero for all except the smallest size range indicating most of the facilities had zero violations in the database and are performing reliably. The upper whisker and third quartile do show trends among the worst 5% and 25% of facilities, however. The model predictions and actual data show that increasing facility size is associated with fewer BOD violations. The worst 5% of plants smaller than 400 m3/d violate their BOD permit limits nearly half of the time, while the worst 5% of plants larger than 40,000 m3/d have violations only 11% of the time. Additionally, for the smallest facilities one-quarter violated limits at least 12% of months, or 1.4 BOD violations per year. While most facilities are reliable and have no violations, significantly more small plants violate their BOD permits more frequently than larger ones and the GLM captures this trend.
3.3.
Fig. 3 e Model prediction of BOD violations versus average flow and capacity utilization.
Second, capacity utilization has a large positive relationship with violation frequency for facilities 4000 m3/d and smaller, while for large facilities capacity utilization has almost no relationship to violation frequency.
3.2.1.
Comparison of model and actual violation data
Actual violation data are grouped by plant size and shown in a boxplot (Fig. 4) of the fraction of months in violation. As before, the whiskers indicate the 5th and 95th percentiles.
Violation magnitude and risk
To model violation magnitude, the BOD data are filtered to include only those data points in which a violation occurred. Of the 209 facilities in the original data set, there were 454 violations at 84 unique facilities with flow rates from 4 m3/d to 40,000 m3/d. A comparison of models showed that neither average flow rate nor capacity utilization was significant with respect to violation magnitude. The intercept indicates that BOD limit violations for facilities of all sizes average 1.6 times the permitted value. While the mean violation magnitude is 1.6 times the permit, the median is only 1.3 and 95% of the violations are less than 2.8. 15% of violations are serious violations, defined as effluent discharges more than double the permitted value for BOD, and there are a small number of extreme violations up to 20 times the permit. Risk is considered as the violation frequency multiplied by the relative magnitude, but because violation magnitude does not vary with facility flow rate or capacity utilization, risk is characterized by the violation frequency alone. Modeled violation probabilities translate to a violation every 8 months for the smallest facilities (40 m3/d), every 2.5 years for the medium facilities (4000 m3/d), and every 10 years at the largest facilities (400,000 m3/d). Because 15% of these violations are serious, the expected return period for serious BOD permit violations is 4.3 years at small facilities, 16 years at mediumsized facilities, and 70 years at the largest. Wastewater treatment facilities are operated for many decades so it is likely that all but the largest facilities will have several serious BOD violations during its lifetime. This is especially true for smaller facilities.
3.4. Predicted effluent TSS, ammonia, and fecal coliforms
Fig. 4 e Boxplot showing variation of violation fractions grouped by facility size where whiskers show 5% and 95% percentiles of BOD violation frequency and square is the model prediction calculated using the average flow and capacity utilization for facilities in the size range.
Using the same methods as presented previously for BOD, average relative effluent values for TSS, ammonia, and fecal coliforms were predicted using GLMs with the gamma distribution and inverse link function. 1= EðRTSS Þ ¼ 3:41 þ 0:233A þ 0:626C
(7)
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 7 9 e4 2 8 6
1= EðRNH4 Þ ¼ 3:91 þ 0:329A þ 0:423AC
(8)
1= EðRFC Þ ¼ 7:28
(9)
The best GLMs for each constituent differ somewhat: the product AC was not significant for TSS, and capacity utilization, C, was not significant for ammonia except as the product AC. However, both TSS and ammonia show similar trends as BOD. Specifically, small and highly utilized facilities are predicted to discharge higher average relative levels of these two constituents. None of the flow or capacity variables produced better prediction of fecal coliforms than the unconditional model. The difference is not surprising because disinfection is carried out in a separate process from the biological treatment processes that determine the effluent BOD, total suspended solids, and ammonia.
3.5.
Risk of TSS, NH4, and FC permit violations
Violation frequency and risk models for TSS, ammonia, and fecal coliforms using a binomial GLM are as follows: EðVTSS Þ ln ¼ 3:37 þ 0:167A þ 0:275AC 1 EðVTSS Þ
(10)
EðVNH4 Þ ln ¼ 3:95 þ 0:483A þ 0:658C 1 EðVNH4 Þ
(11)
EðVFC Þ ¼ 4:30 þ 0:212A ln 1 EðVFC Þ
(12)
value of 0.69 and three flow rates: 40, 4000, and 400,000 m3/d. The smallest facilities (40 m3/d) are estimated to violate BOD and TSS permits about 15 times more frequently than the largest facilities (400,000 m3/d), ammonia permits 75 times more frequently, and fecal coliform permits 7 times more often. Because BOD and TSS are closely related in treatment processes, it is not surprising that the models of those two constituents have very similar violation rates and magnitudes. By contrast, there are fewer fecal coliform violations for facilities of all sizes, but the magnitude of violations is much greater relative to permits levels. While disinfection processes are more reliable generally, the failures that do happen appear to be more significant. As a constituent of increasing concern (State-EPA Nutrient Innovations Task Group, 2009), the risk for ammonia violations stands out both for the predicted frequency of violations for small facilities as well as the severity of the violations. Modeled violation probabilities translate to a violation every 4.5 months for the smallest facilities (40 m3/d) and every 2.8 years for the medium facilities (4000 m3/d). 40% of these violations are double the permitted value, so the expected return period for serious ammonia permit violations is 11 months at small facilities and 7 years at medium facilities while larger facilities have significantly fewer violations. This finding indicates that care should be taken when implementing a decentralized wastewater infrastructure in watersheds sensitive to excess ammonia such as the Chesapeake Bay.
3.6.
The coefficients for average flow are negative for all four constituents, meaning that larger facilities violate their permits less frequently than smaller facilities. This trend is especially strong for ammonia while BOD actually shows the weakest trend. Like for BOD, small and highly utilized facilities have higher rates of TSS violations; however underutilized facilities are predicted to have more frequent ammonia violations than highly utilized ones. Interestingly the GLM for expected frequency of violations of fecal coliform standards is associated only with plant average flow, with higher risk at smaller plants. Average flow rate is a statistically significant predictor in more of the best-fit models than capacity utilization. Therefore, model predictions of violation rate are presented in Table 2 with capacity utilization fixed at the observed mean
BOD TSS NH4 3
Violation rate (40 m /d flow) Violation rate (4000 m3/d flow) Violation rate (400,000 m3/d flow) Mean violation magnitude (effluent/permitted) Serious violations (% of total violations)
FC
13% 3.4% 0.8% 1.63
15% 3.3% 0.7% 1.75
22% 3.0% 0.3% 2.59
3.5% 1.3% 0.5% 2.34
15%
20%
40%
37%
Implications of results
Small facilities often have fewer resources for upgrading or expansion of their treatment facilities which may be an explanation for the number of plants reporting flow rates over the design capacity that in turn affects performance. Many small wastewater facilities also may have fewer hours of attended operation than centralized plants. While large facilities can afford to have full time certified operators and engineers, small facilities can often only afford part-time contract operators. One possible result of limited oversight is that management is less responsive to process changes or upsets, resulting in increased effluent variability and a higher violation frequency. Coupled with the reduced regulatory oversight for small facilities, there is little incentive to improve their operation.
4. Table 2 e Summary of GLM-predicted violation rates and magnitudes based on average monthly flow rate with capacity utilization fixed at 0.69. Serious violations are those where the discharge concentration was twice the permit limit.
4285
Conclusion
Statistical evaluation of discharge monthly report (DMR) data for 211 wastewater treatment plants in the EPA ICIS-NPDES database using a generalized linear model (GLM) indicated significantly increased frequency of permit violations for BOD, TSS, ammonia, and fecal coliforms as plant capacity decreased. This trend was consistent over the entire range of plant capacities sampled: 1e335,000 m3/d. For facilities smaller than 40,000 m3/d, there is also a trend that increasing facility size correlates with decreasing effluent constituent concentrations relative to permitted values for BOD, TSS, and ammonia. The trend toward increasing risk of discharges for
4286
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 2 7 9 e4 2 8 6
smaller facilities exceeding permit limits was strongest for ammonia. Facilities larger than 40,000 m3/d have predicted effluent levels of constituents that are closer to permit limits but reduced violation rates, suggesting that larger plants can operate more efficiently than smaller facilities by not overtreating wastewater. For facilities smaller than 4000 m3/d, exceeding the plant design hydraulic capacity was a significant factor in decreased treatment reliability. Small facilities near or over their design flow rates had significantly more permit violations and higher relative effluent levels for BOD, TSS, and ammonia than those operating under their hydraulic capacity. The GLM approach developed in this research offers a flexible framework for modeling a suite of variables with different characteristics (skewed, binary, discrete etc.) unlike the more commonly used linear modeling methods. As demonstrated above, we obtained insights into reliability and risk associated with facility size which may guide effective management and planning of treatment plants. If networks of decentralized small facilities are to become a larger part of the wastewater treatment infrastructure in the US, planners and regulators should consider the GLM results that suggest the possibility of increased aggregate risk to surface water quality and public health from multiple small plants.
references
Akaike, H., 1974. A new look at the statistical model identification. IEEE Transactions Automation Control 19 (6), 716e723. Burian, S.J., Nix, S.J., Pitt, R.E., Durans, S.R., 2000. Urban wastewater management in the United States: past, present, and future. Journal Urban Technology 7 (3), 33e62. Carruthers, J.I., Ulfarsson, G.F., 2003. Urban sprawl and the cost of public services. Environmental Planning B 30 (4), 503e522. Cunha, M.C., Pinheiro, L., Zeferino, J., Antunes, A., Afonso, P., 2009. Optimization model for integrated regional wastewater systems planning. Journal Water Resources Planning & Management 135 (1), 23e33.
Good, B., 2006. Rocky Mountain Reuse e Barriers to Overcome. Rocky Mountain Water Reuse Workshop, Golden, CO. McCullagh, P., Nelder, J.A., 1989. Generalized Linear Models. Chapman and Hall, London. National Decentralized Water Resources Capacity Development Project, 2009. www.ndwrcdp.org/ (accessed 01/06). Niku, S., Schroeder, E.D., 1981a. Stability of activated sludge processes based on statistical measures. Journal Water Pollution Control Federation 53 (4), 457e473. Niku, S., Schroeder, E.D., 1981b. Factors affecting effluent variability from activated sludge processes. Journal Water Pollution Control Federation 53 (5), 546e559. Niku, S., Schroeder, E.D., Samaniego, F.J., 1979. Performance of activated sludge process and reliability-based design. Journal Water Pollution Control Federation 51 (12), 2841e2857. Oakley, S.M., Gold, A.J., Oczkowski, A.J., 2010. Nitrogen control through decentralized wastewater treatment: process performance and alternative management strategies. Ecological Engineering 36, 1520e1531. Oliveira, S.C., Von Sperling, M., 2008. Reliability analysis of wastewater treatment plants. Water Research 42 (4), 1182e1194. State-EPA Nutrient Innovations Task Group, 2009. An Urgent Call to Action. U.S. Environmental Protection Agency, 2008a. Enforcement & Compliance History Online (ECHO). http://www.epa-echo.gov/ echo/compliance_report_water.html (accessed 15.01.09). U.S. Environmental Protection Agency, 2008b. 2004 Clean Watersheds Needs Survey: Report to Congress. U.S. Environmental Protection Agency, 2002. Onsite Wastewater Treatment Systems Manual 625-R-00e008. U.S. Environmental Protection Agency, 1997. Response to Congress on Use of Decentralized Wastewater Treatment Systems EPA 832-R-97e001b. Voutchkov, N.S., Boulos, P.F., 1993. Heuristic screening methodology for regional wastewater-treatment planning. Journal Environmental Engineering 119 (4), 603e614. Wang, C.G., Jamieson, D.G., 2002. An objective approach to regional wastewater treatment planning. Water Resources Research 38 (3). Water Environment Foundation Alternative Sewer Systems Task Force, 2008. Alternative Sewer Systems, WEF Manual of Practice FD-12.
