WATER RESEARCH A Journal of the International Water Association
Editor-in-Chief Mark van Loosdrecht Delft University of Technology Department of Biochemical Engineering Julianalaan 67 2628 BC Delft The Netherlands Tel: +31 15 27 81618 E-mail:
[email protected]
Editors J. Block Université H. Poincaré, Nancy I France David Dixon University of Melbourne Australia Hiroaki Furumai The University of Tokyo Japan Xiaodi Hao Beijing University of Civil Engineering and Architecture China Gregory Korshin University of Washington USA Anna Ledin Formas Sweden Eberhard Morgenroth Swiss Federal Institute of Aquatic Science and Technology (EAWAG) Switzerland W. Rauch University Innsbruck Austria Maria Reis Universidade Nova de Lisboa/FCT Portugal Hang-Shik Shin Korea Advanced Institute of Science and Technology Korea Thomas Ternes Bundesanstalt für Gewässerkunde Germany Stefan Wuertz Univ. of California, Davis USA
Associate Editors Andrew Baker University of New South Wales Australia
Damien Batstone The University of Queensland Australia G-H. Chen The Hong Kong University of Science & Technology Hong Kong China Tom Curtis Univ. of Newcastle upon Tyne UK Ana Deletic Monash University USA Francis de los Reyes III North Carolina State University USA Rob Eldridge The University of Melbourne Australia Rosina Girones University of Barcelona Spain Stephen Gray Victoria University Australia Kate Grudpan Chiang Mai University Thailand E.E. Herricks University of Illinois - Urbana USA Peter Hillis United Utilities Plc UK H-Y. Hu Tsinghua University China P.M. Huck University of Waterloo Canada Bruce Jefferson Cranfield University UK Ulf Jeppsson Lund University Sweden Sergey Kalyuzhnyi Moscow State University Russian Federation Jaehong Kim Georgia Institute of Technology USA Jes La Cour Jansen Lund Institute of Technology Sweden G. Langergraber BOKU/Univ. of Natural Res. and Applied Life Scs. Austria S-L. Lo National Taiwan University Taiwan Dionisis Mantzavinos Technical University of Crete Greece
Y. Matsui Hokkaido University Japan A. Maul Université Paul Verlaine-Metz France Max Maurer EAWAG Switzerland How Yong Ng National University of Singapore Singapore Satoshi Okabe Hokkaido University Japan S.L. Ong National University of Singapore Singapore Jong M. Park Pohang University of Science & Technology Korea Susan Richardson U.S. Environmental Protection Agency USA Miguel Salgot University of Barcelona Spain David Sedlak University of California, Berkeley USA Jean-Philippe Steyer LBE-INRA France M. Takahashi Hokkaido University Japan Kai Udert EAWAG Switzerland V.P. Venugopalan BARC Facilities India E. von Sperling Federal University of Minas Gerais Brazil Hanqing Yu University of Science & Technology of China China J. Zilles University of Illinois Urbana USA A.I. Zouboulis Aristotle University Greece
Editorial Office Mailing address Water Research Radarweg 29 1043 NX Amsterdam The Netherlands 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 ) 3 8 1 1 e3 8 2 2
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Efficiency and energy requirements for the transformation of organic micropollutants by ozone, O3/H2O2 and UV/H2O2 Ioannis A. Katsoyiannis a, Silvio Canonica a, Urs von Gunten a,b,* a
Eawag, Swiss Federal Institute of Aquatic Science and Technology, Department of Water Resources and Drinking Water, Ueberlandstrasse 133, 8600 Du¨bendorf, Switzerland b School of Architecture, Civil and Environmental Engineering (ENAC), Ecole Polytechnique Fe´de´rale Lausanne (EPFL), Switzerland
article info
abstract
Article history:
The energy consumptions of conventional ozonation and the AOPs O3/H2O2 and UV/H2O2
Received 23 February 2011
for transformation of organic micropollutants, namely atrazine (ATR), sulfamethoxazole
Received in revised form
(SMX) and N-nitrosodimethylamine (NDMA) were compared. Three lake waters and
15 April 2011
a wastewater were assessed. With p-chlorobenzoic acid (pCBA) as a hydroxyl radical (OH)
Accepted 18 April 2011
probe compound, we experimentally determined the rate constants of organic matter of
Available online 7 May 2011
the selected waters for their reaction with OH (kOH,DOM), which varied from 2.0 104 to
Keywords:
various water matrices, which were in the range 6.1e20 104 s1. The varying scavenging
Oxidation
rates influenced the required oxidant dose for the same degree of micropollutant trans-
Energy
formation. In ozonation, for 90% pCBA transformation in the water with the lowest scav-
Ozonation O3/H2O2
enging rate (lake Zu¨rich water) the required O3 dose was roughly 2.3 mg/L, and in the water with the highest scavenging rate (Du¨bendorf wastewater) it was 13.2 mg/L, corresponding
UV/H2O2
to an energy consumption of 0.035 and 0.2 kWh/m3, respectively. The use of O3/H2O2
Scavenging rate
increased the rate of micropollutant transformation and reduced bromate formation by
Micropollutants
70%, but the H2O2 production increased the energy requirements by 20e25%. UV/H2O2
3.5 104 L mgC1 s1. Based on these data we calculated OH scavenging rates of the
efficiently oxidized all examined micropollutants but energy requirements were substantially higher (For 90% pCBA conversion in lake Zu¨rich water, 0.17e0.75 kWh/m3 were required, depending on the optical path length). Energy requirements between ozonation and UV/H2O2 were similar only in the case of NDMA, a compound that reacts slowly with ozone and OH but is transformed efficiently by direct photolysis. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
Ozone (O3) is widely used in water treatment as disinfectant and oxidant. Transformation of organic compounds with O3 occurs via direct reaction with O3 or with hydroxyl radicals (OH), resulting from ozone decay in water (von Gunten, 2003a). O3 reacts selectively with organic compounds and
second order rate constants vary over 10 orders of magnitude, whereas OH is a less selective oxidant and its reaction with the majority of organic compounds is nearly diffusion controlled (von Gunten, 2003a). Advanced Oxidation Processes (AOPs) are based on the enhanced formation of OH. The combined use of ozone/hydrogen peroxide (O3/H2O2) accelerates the conversion of O3 to OH, which can reduce the
* Corresponding author. Eawag, Swiss Federal Institute of Aquatic Science and Technology, Department of Water Resources and Drinking Water, Ueberlandstrasse 133, 8600 Du¨bendorf, Switzerland. Tel.: þ41 44 8235270. E-mail address:
[email protected] (U. von Gunten). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.04.038
3812
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 1 1 e3 8 2 2
reaction time required for micropollutant transformation (Acero and von Gunten, 2001). Combining ultraviolet radiation (UV) with H2O2 results in the generation of OH and represents an AOP as well. It can lead to micropollutant transformation by direct photolysis and by reaction with OH (Legrini et al., 2003). OH is a powerful oxidant but can be scavenged by dissolved organic matter (DOM) and carbonate/bicarbonate. Nitrite can also consume hydroxyl radicals, but this is typically only the case in poorly nitrified/denitrified wastewaters (Lee and von Gunten, 2010). Consequently, in real waters, only a small fraction of OH can reach the target micropollutants. To quantify the scavenging of OH by the water matrix, the pseudo-first-order rate constant kscav (s1) can be used: X kscav ¼ kOH;Si ½Si (1) i
where [Si] is the concentration of the ith- scavenger species Si and kOH;Si is the second-order rate constant for the reaction of OH with Si. The major scavengers in any water are the DOM and carbonate/bicarbonate, thus the scavenging rate in natural waters can be calculated based on the following equation: 2 kscav ¼ kOH;DOM ½DOM þ kOH;HCO3 HCO 3 þ kOH;CO2 CO3 3
(2)
The rate constant for the reaction of bicarbonate with OH is kOH;HCO3 ¼ 8:5 106 M1 s1 , and for carbonate kOH;CO2 ¼ 3 3:9 108 M1 s1 (Buxton and Elliot, 1986). For the reaction of DOM with OH an average rate constant is reported to be 2.5 104 L mgC1 s1 (Hoigne´, 1998) but the actual value depends on the nature of the DOM (Westerhoff et al., 2007; Dong et al., 2010). Therefore, diverse water matrices have different scavenging rates, because of different kOH,DOM values and varying DOM and carbonate-alkalinity concentrations. Other important factors for the efficiency of micropollutant transformation result from the second order rate constants of their reaction with O3, OH as well as direct photolysis in UV based processes (Huber et al., 2003; Canonica et al., 2008). There has been little research in comparing the energy requirements of conventional ozonation and AOPs to transform target micropollutants, taking into account the various factors affecting the efficiency of these methods. In a recent study by Rosenfeldt and co-workers (2006, 2008), some of these issues were addressed. They compared the oxidative ability of ozone or O3/H2O2 and UV/H2O2 processes and the energy required for each technology to form OH and found that under most of the tested conditions, ozone is a more energy efficient technology for production of OH in the waters tested. However, experiments were performed only by using p-chlorobenzoic acid (pCBA) as a probe compound, scavenging rate calculations were done using the average kOH,DOM values taken from the literature, and the issue of bromate formation, a by-product of ozonation (von Gunten, 2003b) was not addressed. The objective of the present study was to perform a systematic comparison of conventional ozonation with the AOPs O3/H2O2 and UV/H2O2, investigating the effect of scavenging rate of water and the type of micropollutant to be treated on the energy requirements for oxidative water treatment. In addition, these processes were evaluated in
view of bromate formation. We used 4 water matrices, consisting of 3 lake waters and a wastewater (after membrane bioreactor (MBR) treatment), covering a broad range of OH scavenging rates. DOM was characterized with Liquid Chromatography e Organic Carbon Detection (LC-OCD) and the rate constants of DOM with OH were experimentally determined by competition kinetics. We additionally investigated the transformation efficiency of 4 organic compounds, with varying rate constants for their reactions with O3, with OH and for direct photolysis. The concluding objective was to perform energy calculations for the oxidation of various micropollutants by the application of the selected technologies in waters with varying scavenging rates.
2.
Materials and methods
2.1.
Investigated waters
To simulate real water treatment conditions we performed the experiments with 4 real waters. Lake Zu¨rich (ZH) and Lake Greifensee (GF) are located in Switzerland, whereas Lake Jonsvatnet (NW) is in Norway. Du¨bendorf wastewater effluent (DW) is also from Switzerland. Lake waters were filtered through 0.45 mm cellulose filters and stored at 4 C. The wastewater effluent was collected after MBR with a cut off diameter 0.1 mm and was used without additional filtration. Major physicochemical parameters of the examined waters are shown in Table 1.
2.2.
Chemicals and investigated micropollutants
pCBA was used as the probe compound for hydroxyl radicals (Elovitz and von Gunten, 1999) because it reacts very slowly with O3 and its transformation by direct photolysis is also slow (Rosenfeldt et al., 2006). We also used atrazine (ATR), sulfamethoxazole (SMX) and N-nitrosodimethylamine (NDMA) to investigate various scenarios with respect to the efficiency of the selected oxidation processes. The basic kinetic parameters of these compounds are listed in Table 2.
2.3.
Ozonation and O3/H2O2
Ozonation experiments were performed in a 500 mL batch reactor, similar to previous studies (Huber et al., 2003). The solutions were prepared as follows: firstly we filled the reactor with the selected water, adjusted the temperature to 20 C, buffered with 5 mM borate and adjusted the pH with 1M H2SO4 or NaOH. All experiments were peroformed at pH 8. Next, pCBA and other compounds were spiked in the water to a final concentration of 0.5e1 mM and a sample was taken at time zero. Bromide was additionally spiked to a concentration of 80 mg/L, to investigate bromate formation. Ozone was injected under stirring from a stock solution of approximately 1.5 mM to achieve the desired O3 dose. Samples were taken after 24 h to measure pCBA (or other micropollutants) transformation after complete O3 consumption. The O3/H2O2 experiments were performed the same way as the ozonation experiments with the addition of H2O2 (2:1 M basis O3:H2O2) prior to O3 addition. For the kinetic experiments, samples were taken at specific time points
3813
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 1 1 e3 8 2 2
Table 1 e DOM and carbonate-alkalinity concentrations of waters examined in this study; experimentally determined rate constants (kOH,DOM) for the reaction of DOM with .OH radicals, including standard errors calculated from standard errors for values A and B displayed in Table S2; calculated scavenging rate attributable to DOM and carbonate-alkalinity and calculated total scavenging rates for the selected water matrices. For comparison, total scavenging rate values, calculated by applying an average value for kOH,DOM of 2.5 3 104 L mgCL1 sL1 are displayeda. Lake Zu¨rich water (ZH-water) [DOM] (mgC L1) kOH,DOM (L mgC1 s1) Scavenging rate from DOM (s1) Carbonate-alkalinity (mM) Scavenging rate from Carbonate-alkalinity (s1) Total scavenging rate (s1) Total scavenging rate calculated using average values from literature for kOH,DOM (s1)
1.3 2.7 3.5 2.6 2.5 6.1 5.8
(1.2) 104 104 104 104 104
Lake Jonsvatnet water (NW-water) 3 2.0 5.9 0.4 0.3 6.2 7.8
(0.2) 104 104 104 104 104
Lake Greifensee water (GF-water)
Wastewater Du¨ bendorf (DW-water)
3.1 2.1 (0.8) 104 6.5 104 4.0 3.9 104 10.4 104 11.7 104
3.9 3.5 (0.2) 104 13.7 104 6.5 6.4 104 20.0 104 16.1 104
a Experimental conditions: pH ¼ 8,T ¼ 20 C.
and the reaction was quenched by addition of the samples into an acidified indigo solution (Bader and Hoigne´, 1981).
2.4.
UV/H2O2
A merry-go-round photoreactor was used for the UV/H2O2 investigations. The methodology for irradiation kinetics experiments have been described elsewhere (Canonica et al., 1995, Canonica et al., 2008). For most irradiations a lowpressure (LP) mercury arc lamp (Heraeus Noblelight model TNN 15/32, nominal power 15 W) was used in combination with a quartz cooling jacket. Kinetic experiments were performed at pH 8. Pure water with negligible light absorption was recirculated in a temperature-controlled photoreactor. Fluence rate values were determined by chemical actinometry at low optical density (Canonica et al., 2008) using 5 mM atrazine as an actinometer (solution buffered at pH ¼ 7.0 with 5 mM phosphate). The determined photon fluence rate was 2.73 105 E m2 s1. Another photoreactor setup, with a medium-pressure (MP) mercury arc lamp (Heraeus Noblelight model TQ718, nominal power 500 700 W) and a UVW-55 glass band pass filter (l ¼ 308 410 nm) in the cooling jacket was employed for the determination of kOH,DOM (Huber et al., 2003). The application of this setup excluded any pCBA transformation by direct photolysis.
2.5.
Analytical methods
The concentrations of the selected compounds were measured by HPLC equipped with a UV detector. Eluents consisted of 10 mM phosphoric acid and methanol or acetonitrile. Depending on compounds isocratic or gradient elutions were used (column: Nucleosil 100, 5 mM C18, MachereyeNagel). Quantification limits of about 0.05e0.1 mM were achieved (Huber et al., 2003). Dissolved ozone was determined with the indigo method (Bader and Hoigne´, 1981). H2O2 was measured photometrically at l ¼ 240 nm (e ¼ 40 M1cm1) (Bader et al., 1988). Bromate was measured by ion chromatography and UV detection after post column reaction (Salhi and von Gunten, 1999). Organic matter characterization was performed by liquid chromatography coupled to an organic carbon detector (LC-OCD), as described in detail elsewhere
(Meylan et al., 2007). Carbonate/Bicarbonate concentrations were measured as alkalinity by titration with 0.1 M hydrochloric acid (endpoint pH ¼ 4.5) using a Titrando potentiometric titrator (Metrohm, Herisau, Switzerland).
3.
Results and discussion
3.1. Characterization of organic matter of the waters used in this study The results of chromatographic characterization of dissolved organic matter (DOM) of lake waters and effluent organic matter of the wastewater (EfOM) are displayed in the supplementary section (Figure S1 and Table S1), showing the different fractions of organic compounds, as described in literature (Huber and Frimmel, 1992; Meylan et al., 2007; Rosario-Ortiz et al., 2008). For simplicity, we will designate all organic matter as DOM. The results indicate that humic substances comprise the major part of DOM in the examined waters. ZH- and GF-waters have higher percentage of high molecular weight compounds (e.g., polysaccharides, proteins) than NW- and DW-waters. The sum of high molecular compounds and humic substances in the three lake waters comprise more than 60% of the DOM, whereas in DW-water it comprises only around 50%, indicating that the fraction of low molecular weight compounds is higher in the wastewater sample. Specific UVA absorbance values indicate that NW-water has the highest aromaticity within the examined waters, followed by DW-water (Table S1) (Weishaar et al., 2003).
3.2. Determination of the rate constant for the reaction of organic matter with hydroxyl radicals Competition kinetics was used to calculate the rate constant for the reaction of DOM with OH, employing pCBA as the probe compound and t-butanol (t-BuOH) as a competitive scavenger (kOH;tBuOH ¼ 6 108 M1 s1 ) (Staehelin and Hoigne´, 1982; Flyunt et al., 2003). To circumvent scavenging of OH by carbonate/bicarbonate, the water samples were pretreated by acidification with 1M H2SO4 and subsequent purging with nitrogen to remove any carbonate as carbon
3814
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 1 1 e3 8 2 2
Table 2 e Kinetic parameters for the oxidative and photochemical transformation of target micropollutants. Target compound
kO3/M1 s1
Structure
kOH/M1 s1 Direct phototransformation(l ¼ 254 nm) 4 a/mol einstein1
pechlorobenzoic acid (pCBA)
<0.1c
Sulfamethoxazole (SMX)
2.5 106 f (at pH 7)
Atrazine (ATR)
6
N-Nitrosodimethylamine (NDMA)
0.052
h
k
kE0p b/m2 einstein1
5 109
c
0.013de0.026e
5.5 109
f
0.046
g
176
g
3 109
h
0.046
i
40.9
j
4.5 108
k
0.3
l
n. a.
z 104
m
a Quantum yield. b Photon fluence-based rate constant (see Canonica et al., 2008). c from Elovitz and von Gunten, 1999. d from Rosenlfeldt and Linden, 2007. e from Chen et al., 1998. f from Huber et al., 2003. g from Canonica et al., 2008. h from Acero et al., 2000. i from Hessler et al., 1993. j Calculated using a molar absorption coefficient of 3860 M1 cm1 (Nick et al., 1992). k from Lee et al., 2007. l from Sharpless and Linden, 2003. m calculated using a molar absorption coefficient of z1500 M1 cm1 extracted from graphical data (Sharpless and Linden, 2003).
dioxide. pH was finally readjusted prior to the experiments using 20e40 mL of 1M NaOH, which was calculated to have a negligible effect in carbonate/bicarbonate concentration of the waters. The kinetic methodology consisted of measuring the time course of pCBA depletion (initial concentration: 0.5 mM) during UV radiation in an aqueous solution made up of the pre-treated natural water (90% by volume), a given amount of H2O2 (1e2 mM) and varying concentrations of
½$OHss ¼
kpCBA ¼ kOH;pCBA ½$OHss app
aOH kOH;DOM ½DOM þ kOH;tBuOH ½t BuOH þ kOH;pCBA ½pCBA þ kOH;H2 O2 ½H2 O2
t-BuOH (0 1000 mM). The calculation of kOH;DOM was done based on the following procedure: The depletion of pCBA follows pseudo-first-order kinetics, with apparent rate app constant kpCBA (s1), according to equation (3). ½pCBAt app ¼ kpCBA t ln ½pCBA0
where t is the irradiation time, [pCBA]o and [pCBA]t are the concentrations of pCBA at irradiation time zero and t, respectively. Under the applied irradiation conditions, direct phototransformation of pCBA is negligible, which leaves OH as the only species responsible for pCBA depletion. One can approximate:
(3)
(4)
(5)
with: where [OH]ss is the steady state concentration of OH, and aOH is the formation rate of OH. The denominator in equation (5) corresponds to the first-order scavenging rate of decarbonized water, including the scavenging caused by H2O2 (kOH;H2 O2 ¼ 2:7 107 M1 s1 ) (Christensen et al., 1982) and by pCBA.
3815
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 1 1 e3 8 2 2
a small fraction of CO32- is always present and this needs to be taken into consideration because kOH;CO2 is much higher than
Substituting equation (5) into equation (4) results in:
3
aOH kOH;DOM ½DOM þ kOH;tBuOH ½t BuOH þ kOH;pCBA ½pCBA þ kOH;H2 O2 ½H2 O2
Inverting both sides of equation (6) results in equation (7): kOH;DOM ½DOM þ kOH;pCBA ½pCBA þ kOH;H2 O2 ½H2 O2 1 ¼ app aOH kOH;pCBA kpCBA kOH;tBuOH þ ½t BuOH aOH kOH;pCBA
(7)
app
By plotting 1=kpCBA as a function of the concentration of the competitive scavenger (t-BuOH), a linear relationship is observed (Figure S2), with intercept A and slope B according to equations (8) and (9), respectively. A¼
kOH;DOM ½DOM þ kOH;pCBA ½pCBA þ kOH;H2 O2 ½H2 O2 aOH kOH;pCBA
(8)
B¼
kOH;tBuOH aOH kOH;pCBA
(9)
Dividing A by B and solving for kOH,DOM yields equation (10):
kOH;DOM ¼
A kOH;tBuOH kOH;pCBA ½pCBA þ kOH;H2 O2 ½H2 O2 ½DOM B ½DOM
(6)
kOH;HCO3 (Buxton and Elliot, 1986). To calculate the fraction of HCO3 and CO32- in the waters at pH 8 and 20 C we used pK1 ¼ 6.38 and pK2 ¼ 10.38 (Stumm and Morgan, 1996). Table 1 lists the scavenging rate values arising from carbonatealkalinity.
(10)
It is noted that the concentration of DOM was quantified as DOC in mgC L1. Figure S2 shows the experimental results and Table S2 the results of the linear regression analysis. The kOH,DOM values, obtained by this method, for the different waters are listed in Table 1. They vary from 2.0 104 to 3.5 104 L mgC1 s1, which agrees well with the range of values found in the literature. Brezonik and FulkersonBrekken (1998) studied 5 surface water sources and reported kOH,DOM values in the range 1.53e3.07 104 L mgC1 s1, while Westerhoff et al. (2007) studied seven DOM isolates and found that kOH,DOM ranged from 1.2 104 to 3.8 104 L mgC1 s1. The latter study and a recent study from Dong et al. (2010) showed that more polar lower molecular weight DOM isolates from wastewater have higher values of kOH,DOM, in general agreement with the present study. Applying the experimentally determined rate constants for the calculation of OH scavenging rates arising from DOM (kOH,DOM [DOM]) gives significantly different values than those obtained if an average value from the literature is applied. This shows the importance of calculating the individual rate constants for the reaction of hydroxyl radicals with DOM for each of the selected waters and greatly contributes to the correct explanation of experimental results.
3.3. Total scavenging rate of the examined waters and effect on the transformation of pCBA by conventional ozonation The total OH scavenging rates resulting from both DOM and carbonate-alkalinity for the waters examined in the present study vary from 6.1 104 to 20 104 s1 (see Table 1). To examine the effect of scavenging rate on the efficiency of conventional ozonation, we measured the pCBA removal after complete consumption of O3 for varying O3 doses in the selected waters. The results in Fig. 1 depict that in waters with higher scavenging rates, a lower extent of pCBA removal takes place for the same O3 doses. In ZH- and NW-waters, having similar OH scavenging rates, pCBA removal was similar at any O3 dose. The use of scavenging rates calculated based on the average kOH,DOM (Table 1) would lead to an incorrect prediction in this case. Despite the similar extent of pCBA removal, the kinetics of pCBA removal in these waters varied (Fig. 2). pCBA removal was faster in NW-water than in ZH-water. This is assumed to be attributable to the much higher DOM and much lower carbonate-alkalinity concentrations of NW-water compared to ZH-water. Higher DOM concentration leads to a much faster O3 decomposition, thus a more rapid conversion into OH and
100
80
% pCBA transformation
app
kpCBA ¼ kOH;pCBA
ZH-water NW-water GF-water DW-water
60
40
20
0 1
3.2.1. Calculation of the scavenging rate resulting from carbonate-alkalinity Carbonate/Bicarbonate (expressed as carbonate-alkalinity) is the other major component usually found in real waters which acts as a scavenger of OH. For typical pH values of real waters, alkalinity is mainly present as HCO3. However,
2
4
Ozone Dose (mg/L)
Fig. 1 e pCBA transformation by conventional ozonation as a function of ozone dose in waters with varying OH scavenging rates (pH [ 8, T [ 20 C). Scavenging rate of different waters: ZH: 6.1 3 104 sL1; NW: 6.2 3 104 sL1; GF: 10.4 3 104 sL1; DW: 20 3 104 sL1.
3816
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 1 1 e3 8 2 2
1.0
NW-water ZH-water NW-water dilluted (1:1) + Alkalinity
[pCBA]t/[pCBA]o
0.8
0.6
0.4
0.2
0.0 0
5
10
15
20
25
30
Time (min)
Fig. 2 e Kinetics of pCBA transformation by ozonation (1 mg/L) in ZH-, NW- and in NW-water diluted 1:1 with Milli-Q water and spiked with NaHCO3 to reach the carbonate-alkalinity and DOM concentrations of ZH-water at 20 C and pH 8.
consequently faster pCBA oxidation. This shows that the OH scavenging rate is important for the extent of transformation of pCBA, yet, the kinetics of the process is mainly dominated by the distribution of scavenging between DOM and carbonate-alkalinity. To further elucidate the role of OH scavenging rate and the distribution of scavenging on the micropollutant transformation, we examined the removal of pCBA in 1:1 diluted NW-water with Milli-Q water, spiked with NaHCO3, to achieve DOM and carbonate-alkalinity concentrations similar to those of ZH-water. Even though after this procedure the two waters have very similar DOM and carbonate-alkalinity concentrations, the extent and kinetics of pCBA removal are higher in the diluted NW-water. The faster kinetics of pCBA transformation in the diluted NW-water shows that the DOM of NW-water is more reactive than the DOM of ZH-water. Furthermore, the slower kinetics of pCBA removal in ZH-water and in the adjusted NW-water, compared to NW-water, is ascribed to the slower kinetics of O3 decay in these waters (data not shown), due to the higher carbonate-alkalinity as compared with NW-water, because carbonate/bicarbonate acts as inhibitor of O3 decomposition (von Gunten, 2003a). In waters with higher scavenging rates, like the GF and DW, higher O3 doses were required to achieve a similar micropollutant transformation level. For example in GF-water, an ozone dose of 4 mg/L leads to around 85% pCBA transformation, which is roughly twice the dose required for the same pCBA removal in ZH- and NW-waters. At this dose, in DW-water only about 50% pCBA conversion occurs.
3.4. Effect of scavenging rate on bromate formation during ozonation It is well demonstrated that ozonation of bromide-containing waters leads to bromate formation (von Gunten and Hoigne´,
1994). Bromate is a potential carcinogen and is regulated by the European Commission and USEPA, which have established a maximum contaminant level of 10 mg/L in drinking waters (EC directive 98/83, USEPA, 1998). We examined the formation of bromate in parallel to pCBA removal for the selected waters, which were spiked to a bromide concentration of 80 mg/L. Figure S3 illustrates that for an O3 dose of 2 mg/L, the highest pCBA conversion (85%) but also the highest bromate formation was observed in spiked ZH-water, much higher than the drinking water standard of 10 mg/L. In NW-water, bromate formation was quite low (<10 mg/L), while pCBA removal was as high as in ZH-water. In GF-water just exceeded the standard of 10 mg/L and pCBA transformation was around 70%. Finally, in DW-water, an O3 dose of 2 mg/L did not result in formation of significant bromate concentrations but only 20% pCBA removal was achieved. It is noteworthy that bromate formation was much higher in ZH-water than in NW-water, both having similar scavenging rates. The remarkable difference is attributable to the different contributions of DOM and carbonate/bicarbonate to the scavenging rate (Table 1). In ZH-water carbonate-alkalinity is much higher than in NW water. Carbonate/bicarbonate reacting with OH form carbonate radicals, which can convert hypobromite into the BrO radical, an important intermediate towards bromate formation (von Gunten and Hoigne´, 1994). Furthermore, in ZH-water the stability of O3 is higher leading to a higher O3 exposure. Thus, increased carbonate-alkalinity and ozone exposure enhanced bromate formation in the case of ZH-water. GF-water has a similar scavenging distribution as ZH-water, but much higher total scavenging rate. Thus bromate formation is much lower in GF- than in ZH-water, because of a smaller transient OH concentration and smaller O3 exposure.
3.5. Advanced oxidation processes (O3/H2O2 and UV/ H2O2), micropollutant transformation and bromate formation 3.5.1.
O3/H2O2
The AOP O3/H2O2 increases the kinetics of ozone decay and accelerates its transformation into OH. However, it does not affect greatly the extent of micropollutant transformation, for compounds reacting slowly with ozone (e.g., pCBA) (Acero and von Gunten, 2001). Fig. 3a shows the increased rate of O3 decay and Fig. 3b shows the increased kinetics of pCBA transformation, when H2O2 is used. In both cases (conventional ozonation and O3/H2O2) the extent of pCBA transformation was roughly the same and mainly depended on the O3 dose, being in agreement with previous studies (Acero and von Gunten, 2001). The use of H2O2 results in a substantial reduction of bromate formation, because H2O2 reduces HOBr to bromide (von Gunten and Oliveras, 1997; von Gunten and Oliveras, 1998). In the present study, O3/H2O2 reduced bromate formation by approximately 70% in ZH-water (Figure S4). Nevertheless, conversion of the conventional ozonation to the AOP reduces the disinfection efficiency of ozonation (von Gunten, 2003b). Consequently, to ensure disinfection and reduce bromate formation, H2O2 can be added in the reactor in a later stage. Fig. 3a and b show the kinetics of ozone and
3817
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 1 1 e3 8 2 2
1.0
ZH-water NW-water GF-water DW-water
[pCBA]t / [pCBA]o
0.8
0.6
0.4
0.2
0.0 0
5000
10000
15000
-2
Fluence (J m )
Fig. 4 e Transformation of pCBA in various water matrixes as a function of fluence by LP-UV/H2O2 at 20 C and pH 8, applying H2O2 concentration of 0.2 mM.
concentration (0.2 mM; 6.8 mg/L). It can be seen that the scavenging rate of water plays an important role in this case as well, since pCBA transformation is much faster in ZH-water than in the other waters, having much higher scavenging rates. However, in addition to that, the absorption of the water is also relevant. ZH- and NW-waters have quite different absorption coefficients at 254 nm (0.037 and 0.1 cm1, respectively) and despite the similar scavenging rates, the removal of pCBA is significantly faster in the case of ZH-water. Use of UV/H2O2 does not induce bromate formation (von Gunten and Oliveras, 1998), but, as it will be discussed later, it is generally more energy-intensive than ozonation or the AOP O3/H2O2. Fig. 3 e (a) O3 depletion during conventional ozonation, the AOP O3/H2O2 (molar ratio 2:1) and conventional ozonation followed by addition of H2O2. (b) Kinetics of pCBA decrease by conventional ozonation, the AOP O3/H2O2 and conventional ozonation followed by addition of H2O2. The experiments were performed in ZH-water, with 84 mM (4 mg/L) O3 and 42 mM (1.42 mg/L) H2O2, at 20 C and pH 8.
pCBA depletion in this combined experiment, where H2O2 is added 10 min after O3 addition in the water, as compared with conventional ozonation and the AOP O3/H2O2. In this case, ozone can act as disinfectant in the beginning causing also some oxidation (Fig. 3b), and afterwards oxidation can be accelerated by introducing H2O2 in the reactor, resulting in 30% less bromate formation (data not shown), which in some cases might be sufficient to keep the bromate concentration below 10 mg/L.
3.5.2.
UV/H2O2
The use of UV/H2O2 generates OH through light absorption and subsequent photolysis of H2O2 (Legrini et al., 2003). Fig. 4 shows the transformation of pCBA in the selected waters, by using a low-pressure mercury lamp and constant H2O2
3.6. Efficiency of ozonation and UV/H2O2 for the transformation of various micropollutants 3.6.1.
Ozonation
The rate constant for the reaction of micropollutants with ozone is decisive for the required ozone dose to achieve a defined level of transformation. We used three micropollutants, namely SMX, ATR, and NDMA, while pCBA was used as an OH probe compound. From these compounds, SMX reacts very fast with O3 and with OH, pCBA and ATR react relatively slowly with O3 but fast with OH, and NDMA reacts slowly with both O3 and OH (Table 2). ZH- and DW-waters were selected as the waters with the lowest and highest scavenging rates respectively, and the removal of the selected compounds was examined as a function of the O3 dose (Fig. 5). Fig. 5a shows that in ZH-water SMX, the compound with the highest kO3 and kOH, is almost completely transformed even for an O3 dose as low as 0.2 mg/L. In contrast, NDMA, the slowest reacting compound, was transformed only by 20% for an ozone dose of 4 mg/L. In DW-water (Fig. 5b), again transformation of SMX is most efficient for equal applied O3 doses. It is worth noting that for a similar transformation of SMX in the two examined waters, a 10 fold higher O3 dose is necessary in DWwater. For ATR and pCBA (reacting slowly with ozone) only
3818
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 1 1 e3 8 2 2
a
SMX ATR
ZH-water
1.00
pCBA NDMA
b
pCBA NDMA
0.75
Ct / Co
Ct / Co
0.75
SMX ATR
DW-water
1.00
0.50
0.25
0.50
0.25
0.00 0
1
2
3
0.00
4
0
2
Ozone Dose (mg/L)
4
6
8
10
Ozone Dose (mg/L)
c SMX ATR
ZH-water
1.0
pCBA NDMA
0.8
Ct / Co
0.6
0.4
0.2
0.0 0
3000
6000
9000
12000
-2
Fluence (J m )
Fig. 5 e Transformation of SMX, ATR, pCBA and NDMA as a function of the ozone dose (a) in ZH-water, (b) in DW-water and (c) as a function of fluence in ZH-water (H2O2 = 0.2 mM). Experiments were performed at pH 8 and T 20 C.
a 5 times higher O3 dose is required, indicating the important role of direct ozone consumption and scavenging of OH by DOM and the rate of the ozone reaction with the micropollutants. Plotting the data of Fig. 5 as ln(Ct/Co) versus ozone dose (example is given in Figure S5) results in a linear relationship, from which ozone dose-based rate constants can be obtained. Table 3 shows the ozone dose-based rate constants for the transformation of the compounds by conventional ozonation
in two different water matrices. In general, it can be noted that the transformation of compounds is higher in waters with lower DOM concentration and/or lower OH scavenging rates.
3.6.2.
UV/H2O2
Fig. 5c shows the transformation kinetics of the examined compounds by UV/H2O2 in ZH-water and Table 3 shows the fluence-based apparent rate constants. It is noted that the
Table 3 e Ozone dose-based and fluence-based rate constants for the transformation of micropollutants during conventional ozonation in ZH- and DW-waters and by LP-UV/H2O2 in ZH-water respectivelya. Target compound SMX pCBA ATR NDMA
O3 dose-based rate constants in ZH-water (L mgO31)
O3 dose-based rate constants in DW-water (L mgO31)
Fluence-based rate constant in ZH-water (m2 J1)
22.3 1.00 0.67 0.067
0.74 0.17 0.11 0.038
6.7 104 3.3 104 2.46 104 1.43 104
a Experimental conditions: Target compound concentration ¼ 0.5 mM, pH ¼ 8, T ¼ 20 C.
3819
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 1 1 e3 8 2 2
Table 4 e Energy requirements in kWh/m3 for 90% transformation of the selected micropollutants by conventional ozonation in ZH- and DW-waters and by using LP-UV/H2O2 (0.2 mM) for varying optical path lengths (cm) in ZH-watera. Target compound
SMX pCBA ATR NDMA
ZH-water
DW-water
ZH-water
Ozonation
Ozonation
UV/H2O2 1 cm
UV/H2O2 5 cm
UV/H2O2 10 cm
0.0015 0.035 0.05 0.5
0.045 0.2 0.3 0.9
0.39 0.75 0.98 1.62
0.15 0.23 0.28 0.44
0.11 0.17 0.2 0.3
a Experimental conditions: Target compound concentration ¼ 0.5 mM, pH ¼ 8, T ¼ 20 C.
fluence-based rate of SMX is only 5 times higher than that of NDMA, whereas by ozonation in ZH-water it was roughly 300 times higher. The chosen compounds have direct phototransformation rate constants decreasing in the order SMX > NDMA > ATR z pCBA (see Table 2). This difference affects the apparent rates of transformation, especially for NDMA, which reacts slowly with O3 and OH but can be transformed by UV irradiation relatively efficiently. This is in agreement with previous studies of Sharpless and Linden (2003), where it was shown that in simulated drinking water the rate constant of NDMA transformation increased only by 26%, when H2O2 was added to the UV treatment.
3.7.
Energy calculations
3.7.1.
Conventional ozonation and the AOP O3/H2O2
To calculate the energy requirements when applying ozonation or the AOP O3/H2O2, we assumed an average energy requirement of 15 kWh/kg for O3 and of 10 kWh/kg for H2O2 production (Rosenfeldt et al., 2006). All energy calculations are based on a 90% removal of contaminants. The results of energy calculations are shown in Table 4 and are based on the O3 dose and fluence-based rate constants shown in Table 3. The energy requirements for 90% micropollutant transformation by ozonation varied in the range from 0.0015 to 0.9 kWh/m3, depending on the water matrix and on the type of micropollutant (Table 4). The use of the AOP O3/H2O2 for the case of pCBA and ATR (data not shown) showed that it enhanced mainly the kinetics of the reaction but not the extent of micropollutant transformation. Thus, roughly the same O3 dose, as for ozonation, would be required for 90% transformation, however, the energy for H2O2 production has to be accounted for. If we assume a molar ratio 2:1 (O3:H2O2), then for 90% pCBA transformation in ZH-water, the energy consumption will be increased from 0.035 to 0.043 kWh/m3, corresponding to a 23% increase in energy requirements. However, this increase is compensated by reduced bromate formation and smaller required hydraulic residence times. In Table 5, we compare the energy requirements as affected by difference in water matrix, hence by varying scavenging rates. Treatment of NW-water requires the same energy for pCBA conversion as of ZH-water but bromate formation is much lower (Figure S3), because of a lower carbonate-alkalinity content and ozone exposure. Treatment of GF- and DW-waters (waters with higher scavenging rates) would require more energy (0.065 and 0.2 kWh/m3 respectively) than ZH-water to achieve the same degree of compound transformation.
3.7.2.
UV/H2O2
The energy required for 90% transformation of the selected micropollutants in ZH-water applying UV/H2O2 (0.2 mM) can be calculated by the fluence-based rate constants given in Table 3 and employing a photon fluence rate of 2.73 105 E m2 s1 (appropriately converted to a fluence rate of 12.9 J m2 s1) by using the Morowitz correction factor, which accounts for the absorption coefficient of the water and of H2O2 (an example of calculation is given in the supplementary information). The results are illustrated in Table 4 for different optical path length scenarios. It is worth noting that increasing the path length up to an optimum level, energy requirements decrease because of more efficient energy use, as shown in Figure S6. Treatment of all compounds except NDMA in all water matrices required more energy by UV/H2O2 than by conventional ozonation. In the case of NDMA, UV/H2O2 showed similar energy efficiencies as ozonation (0.44 kWh/m3 for UV/H2O2 (0.2 mM) and 5 cm optical path length and 0.5 kWh/m3 for ozonation), being in agreement with previous studies, which reported energy requirements for transformation of NDMA by UV from 0.3 to 0.5 kWh/m3 (Stefan and Bolton, 2002; Sharpless and Linden, 2005). The similar energy requirements for the transformation of NDMA by ozonation and UV/H2O2 are attributable to the fact that NDMA reacts slowly with ozone and OH but is transformed quite efficiently by direct photolysis (Table 2). For the required energy for UV treatment, the absorption of the water plays a significant role. For waters with comparable scavenging rates but with different absorption coefficients (e.g., ZH- and NW-water) the removal of pCBA requires substantially less energy in the less absorbing water (ZH-water, Table 5).
Table 5 e Energy requirements (kWh/m3) for 90% pCBA transformation in various water matrices by conventional ozonation, O3/H2O2 and UV/H2O2, employing 1, 5 or 10 cm path length and 0.2 mM H2O2. Water Matrix Ozonation O3/H2O2
ZH-water NW-water GF-water DW-water
0.035 0.035 0.065 0.2
a
0.043 w0.043b w0.080b w0.25b
UV/H2O2 1 cm
5 cm
10 cm
0.75 1.28 1.92 2.28
0.23 0.45 0.61 0.82
0.17 0.36 0.48 0.70
a Energy calculations based on measurements. b Energy values for O3/H2O2 are estimated based on a molar ratio of O3:H2O2/2:1 and considering that the extent of pCBA transformation by O3/H2O2 depends mainly on the ozone dose and on the water matrix.
3820
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 1 1 e3 8 2 2
a
0.2 mM / 6.8 mg/L 1 mM / 34 mg/L 5 mM / 170 mg/L
0.8
[pCBA]t / [pCBA]o
4.
[H2O2]
1.0
Conclusions C
0.6 C
0.4
0.2
0.0 0
5
10
15
20
25
30
C
Time (min)
b
[H2O2]
2.5
0.2 mM / 6.8 mg/L 1 mM / 34 mg/L 5 mM / 170 mg/L
C
3
Energy (kWh / m )
2.0
1.5
1.0 C
0.5
0.0 0
5
10
15
20
Optical Path Length (cm)
Fig. 6 e (a) Kinetics of pCBA transformation by UV/H2O2 as a function of the H2O2 concentration in DW-water, at pH 8 and T [ 20 C (b) Estimation of energy consumption for the AOP-UV/H2O2 as a function of the optical path length for varying H2O2 concentrations and for 90% pCBA transformation in DW-water (pH 8 and T [ 20 C).
In addition, the energy requirements for transformation of micropollutants by UV/H2O2 depend on the concentration of H2O2. The concentration of H2O2 can be optimized to increase the kinetics of transformation and energy requirements. Fig. 6a illustrates the kinetics of pCBA removal as a function of the H2O2 concentration in DW-water and Fig. 6b the respective energy requirements for different optical path lengths scenarios. Increasing the H2O2 concentration increases the kinetics of pCBA removal and reduces energy consumption, but only up to a certain concentration. Fig. 6b shows that the energy consumption for 90% pCBA removal is reduced when increasing H2O2 concentration from 0.2 to 1 mM but a further increase of the H2O2 concentration e.g., to 5 mM increases the energy requirements dramatically, although kinetics of pCBA conversion are further increased. This can be attributed to the increase of energy requirements for H2O2 production and to the contribution of H2O2 to the scavenging of OH.
Increasing OH scavenging rates of waters lead to increased energy consumption for micropollutant transformation. For example, 90% pCBA depletion by O3 required 0.035 and 0.2 kWh/m3 in the waters with the lowest and highest scavenging rates, respectively. Transformation of compounds, susceptible to direct ozone oxidation, such as olefins, activated aromatics and amines (e.g., SMX in our study), requires roughly 10e20 times less energy than of those which are O3-resistant (e.g., pCBA and ATR). In ZH-water, 90% SMX removal by O3 required 0.0015 kWh/m3, whereas pCBA and ATR 0.035 and 0.05 kWh/m3, respectively. Application of O3/H2O2 increased the transformation rates of compounds and contributed to reduced bromate formation, but the additional energy for the production of H2O2 increased the overall energy consumption by roughly 25%. UV/H2O2 was roughly 5e20 times more energy intensive than ozonation or O3/H2O2, depending on the optical path length, H2O2 concentration, water matrix and type of micropollutant. For 90% transformation of pCBA with UV/H2O2, energy requirements were 0.23 and 0.82 kWh/ m3 in ZH- and DW-water, respectively for a 5 cm optical path length. UV/H2O2 is a viable solution for the transformation of organic micropollutants with low O3 and OH reactivity but high photoactivity such as NDMA. This is most relevant in waters with high bromide content, because UV/H2O2 excludes bromate formation. In ZH-water a 90% NDMA transformation by UV/H2O2 required 0.44 kWh/m3 compared to 0.5 kWh/m3 for ozonation.
Acknowledgments This study was funded by the 6th Framework European Integrated Project TECHNEAU (018320). The authors are grateful to E. Sahli and J. Traber for support in the analytical part of the project.
Appendix. Supplementary material Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.watres.2011.04.038.
references
Acero, J.L., Stemmler, K., von Gunten, U., 2000. Degradation kinetics of atrazine and its degradation products with ozone and OH-radicals: a predictive tool for drinking water treatment. Environmental Science and Technology 34 (4), 591e597. Acero, J.L., von Gunten, U., 2001. Characterisation of oxidation processes: ozonation and the AOP O3/H2O2. Journal of American Water Works Association 93, 90e100.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 1 1 e3 8 2 2
Bader, H., Hoigne´, J., 1981. Determination of ozone in water by the indigo method. Water Research 15 (10), 449e456. Bader, H., Sturzenegger, V., Hoigne´, J., 1988. Photometric method for the determination of low concentrations of hydrogen peroxide by the peroxidase catalyzed oxidation of N, N-diethylp-phenylenediamine (DPD). Water Research 22, 1109e1115. Brezonik, P.L., Fulkerson-Brekken, J., 1998. Nitrate-induced photolysis in natural waters: Controls on concentrations of hydroxyl radical photo-intermediates by natural scavenging agents. Environmental Science and Technology 32 (19), 3004e3010. Buxton, G.V., Elliot, A.J., 1986. Rate constants for reaction of hydroxyl radicals with bicarbonate. Radiation Physics and Chemistry 27 (3), 241e243. Canonica, S., Jans, U., Stemmler, K., Hoigne´, J., 1995. Transformation kinetics of phenols in water: Photosensitization by dissolved natural organic material and aromatic ketones. Environmental Science and Technology 29 (7), 1822e1831. Canonica, S., Meunier, L., von Gunten, U., 2008. Phototransformation of selected pharamceuticals during UV treatment of drinking water. Water Research 42 (1e2), 121e128. Christensen, H., Sehested, K., Corfitzen, H., 1982. Reactions of hydroxyl radicals with hydrogen peroxide at ambient and elevated temperatures. Journal of Physical Chemistry 86 (9), 1588e1590. Chen, J., Peijnenburg, W.J.G.M., Quan, X., Zhao, Y., Xue, D., Yang, F., 1998. The application of quantum chemical and statistical technique in developing quantitative structureproperty relationships for the photohydrolysis quantum yields of substituted aromatic halides. Chemosphere 37 (6), 1169e1186. Dong, M.M., Mezyk, S.P., Rosario-Ortiz, F.L., 2010. Reactivity of effluent organic matter (EfOM) with hydroxyl radical as a function of molecular weight. Environmental Science and Technology 44, 5714e5720. Elovitz, M., von Gunten, U., 1999. Hydroxyl radical/ozone ratios during ozonation processes I. The Rct Concept. Ozone Science and Engineering 21 (3), 239e260. European Council Directive 98/83 of 3 November 1998 related to the quality of water intended for human consumption. Official Journal of the European Communities, 1998 Brussels, December 05. Flyunt, R., Leitzke, A., Mark, G., Mvula, E., Reisz, E., Schick, R., von Sonntag, C., 2003. Determination of OH, O2 , and hydroperoxide yields in ozone reactions in aqueous solution. Jounal of Physical Chemistry B 107 (30), 7242e7253. Hessler, D.P., Gorenflo, V., Frimmel, F.H., 1993. UV degradation of atrazine and metazachlor in the absence and presence of H "SUB 2" O "SUB 2", bicarbonate and humic substances. Aqua 42 (1), 8e12. Hoigne´, J., 1998. Chemistry of aqueous ozone and transformation of pollutants by ozonation and advanced oxidation processes. In: Huber (Ed.), The Handbook of Environmental Chemistry, Quality and Treatment of Drinking Water. Springer, Berlin. Huber, M.M., Canonica, S., von Gunten, U., 2003. Oxidation of pharmaceuticals during ozonation and advanced oxidation processes. Environmental Science and Technology 37 (5), 1016e1024. Huber, S.A., Frimmel, F.H., 1992. New method for the characterization of organic carbon in aquatic systems. International Journal of Environmental Analytical Chemistry 49 (1e2), 49e57. Lee, C., Yoon, J., von Gunten, U., 2007. Oxidative degradation of N-nitrosodimethylamine by conventional ozonation and the advanced oxidation process ozone/hydrogen peroxide. Water Research 41 (3), 581e590.
3821
Lee, Y., von Gunten, U., 2010. Oxidative transformation of micropollutants during municipal wastewater treatment: comparison of kinetic aspects of selective (chlorine, chlorine dioxide, ferrate VI and ozone) and non-selective oxidants (hydroxyl radical). Water Research 44, 555e556. Legrini, O., Oliveros, E., Braun, A.M., 2003. Photochemical processes for water treatment. Chemical Reviews 93 (2), 671e698. Meylan, S., Hammes, F., Traber, J., Sahli, E., von Gunten, U., Pronk, W., 2007. Permeability of low molecular weight organics through nanofiltration membranes. Water Research 41 (17), 3968e3976. Nick, K., Scho¨ler, H.F., Mark, G., So¨ylemez, T., Akhlaq, M.S., Schuchmann, H.-P., von Sonntag, C., 1992. Degradation of some triazine herbicides by UV radiation such as used in the UV disinfection of drinking water. Journal of Water Supply Research and Technology-Aqua 41 (2), 82e87. Rosario-Ortiz, F.L., Mezyk, S.P., Doud, D.V., Snyder, A.A., 2008. Quantitative correlation of absolute hydroxyl radical rate constants with non isolated effluent organic matter bulk properties in water. Environmental Science and Technology 42 (16), 5924e5930. Rosenfeldt, E.J., Linden, K.G., Canonica, S., von Gunten, U., 2006. Comparison of the efficiency of OH radical formation during ozonation and the advanced oxidation processes O3/H2O2 and UV/H2O2. Water Research 40 (20), 3695e3704. Rosenfeldt, E.J., Linden, K.G., Canonica, S., von Gunten, U., 2008. Erratum to “Comparison of the efficiency of OH radical formation during ozonation and the advanced oxidation processes O3/H2O2 and UV/H2O2”. Water Research 42 (10e11), 2836e2838. Rosenfeldt, E.J., Linden, K.G., 2007. The ROH, UV concept to characterize and the model UV/H2O2 process in natural waters. Environmental Science and Technology 41, 2548e2553. Salhi, E., von Gunten, U., 1999. Simultaneous determination of bromide, bromate and nitrite in low mg l1 levels by ion chromatography without sample pretreatment. Water Research 33 (15), 3229e3244. Sharpless, C.M., Linden, K.G., 2003. Experimental and model comparisons of low- and medium-pressure Hg lamps for the direct and H2O2 assisted UV photodegradation of N-nitrosodimethylamine in simulated drinking water. Environmental Science and Technology 37 (9), 1933e1940. Sharpless, C.M., Linden, K.G., 2005. Interpreting collimated beam ultraviolet photolysis rate data in terms of electrical efficiency of treatment. Journal of Environmental Engineering Science 4 (1), 19e26. Staehelin, J., Hoigne´, J., 1982. Decomposition of ozone in water: rate of initiation by hydroxide ions and hydrogen peroxide. Environmental Science and Technology 16 (10), 676e681. Stefan, M.I., Bolton, J.R., 2002. UV direct photolysis of N-nitrosodimethylamine (NDMA): kinetic and product study. Helvetica Chimica Acta 85 (5), 1416e1426. Stumm, W., Morgan, J.J., 1996. Aquatic Chemistry: Chemical Equilibria and Rates in Natural Waters, third ed. Wiley INT Publication, New York. USEPA, 1998. National Primary Drinking Water Regulations: Disinfectants and Disinfection Byproducts Final Rule. RIN 2040-AB82. 40 CFR Parts 9, 141, and 142. Section II E. December. von Gunten, U., 2003a. Ozonation of drinking water: part I: oxidation kinetics and product formation. Water Research 37 (7), 1443e1467. von Gunten, U., 2003b. Ozonation of drinking water: Part II. Disinfection and by-product formation in presence of bromide, iodide or chlorine. Water Research 37 (7), 1469e1487. von Gunten, U., Hoigne´, J., 1994. Bromate formation during ozonation of bromide-containing waters: Interaction of ozone
3822
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 1 1 e3 8 2 2
and hydroxyl radical reactions. Environmental Science and Technology 28 (7), 1234e1242. von Gunten, U., Oliveras, Y., 1997. Kinetics of the reaction between hydrogen peroxide and hypobromous acid: Implication on water treatment and natural systems. Water Research 31 (4), 900e906. von Gunten, U., Oliveras, Y., 1998. Advanced oxidation of bromide-containing waters: bromate formation mechanisms. Environmental Science and Technology 32 (1), 63e70.
Weishaar, J.L., Aiken, G.R., Bergamaschi, B.A., Fram, M.S., Fujii, R., Mopper, K., 2003. Evaluation of specific ultraviolet absorbance as an indicator of the chemical composition and reactivity. Environmental Science and Technology 37 (20), 4702e4708. Westerhoff, P., Mezyk, S.P., Cooper, W.J., Minakata, D., 2007. Electron pulse radiolysis determination of hydroxyl radical rate constants with Suwannee river fulvic acid and other dissolved organic matter isolates. Environmental Science and Technology 41 (13), 4640e4646.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 2 3 e3 8 3 5
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Uncertainty-based calibration and prediction with a stormwater surface accumulation-washoff model based on coverage of sampled Zn, Cu, Pb and Cd field data E. Lindblom a,1, S. Ahlman b,2, P.S. Mikkelsen a,* a
Department of Environmental Engineering (DTU Environment), Technical University of Denmark, Miljøvej, Building 113, DK-2800 Kongens Lyngby, Denmark b Department of Civil and Environmental Engineering, Division of Water Environment Engineering, Chalmers University of Technology, SE-412 96 Go¨teborg, Sweden
article info
abstract
Article history:
A dynamic conceptual and lumped accumulation wash-off model (SEWSYS) is uncertainty-
Received 18 December 2010
calibrated with Zn, Cu, Pb and Cd field data from an intensive, detailed monitoring
Received in revised form
campaign. We use the generalized linear uncertainty estimation (GLUE) technique in
17 March 2011
combination with the Metropolis algorithm, which allows identifying a range of behavioral
Accepted 18 April 2011
model parameter sets. The small catchment size and nearness of the rain gauge justified
Available online 4 May 2011
excluding the hydrological model parameters from the uncertainty assessment. Uniform, closed prior distributions were heuristically specified for the dry and wet removal
Keywords:
parameters, which allowed using an open not specified uniform prior for the dry deposition
Stormwater
parameter. We used an exponential likelihood function based on the sum of squared errors
Heavy metals
between observed and simulated event masses and adjusted a scaling factor to cover 95%
Dynamic conceptual model
of the observations within the empirical 95% model prediction bounds. A positive corre-
Sampled event mass
lation between the dry deposition and the dry (wind) removal rates was revealed as well as
Site mean concentration
a negative correlation between the wet removal (wash-off) rate and the ratio between the
GLUE
dry deposition and wind removal rates, which determines the maximum pool of accumulated metal available on the conceptual catchment surface. Forward Monte Carlo analysis based on the posterior parameter sets covered 95% of the observed event mean concentrations, and 95% prediction quantiles for site mean concentrations were estimated to 470 mg/l 20% for Zn, 295 mg/l 40% for Cu, 20 mg/l 80% for Pb and 0.6 mg/l 35% for Cd. This uncertainty-based calibration procedure adequately describes the prediction uncertainty conditioned on the used model and data, but seasonal and site-to-site variation is not considered, i.e. predicting metal concentrations in stormwater runoff from gauged as well as ungauged catchments with the SEWSYS model is generally more uncertain than the indicated numbers. ª 2011 Elsevier Ltd. All rights reserved.
* Corresponding author. E-mail addresses:
[email protected] (E. Lindblom),
[email protected] (S. Ahlman),
[email protected] (P.S. Mikkelsen). 1 Present address: SWECO International AB, Water and Environment, Gjo¨rwellsgatan 22, Box 34044, SE-100 26 Stockholm, Sweden. 2 Present address: Kalmar Vatten AB, Box 822, SE-391 28 Kalmar, Sweden 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.04.033
3824
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 2 3 e3 8 3 5
Nomenclature Aimp C Crain EF EMC g(qjy) h k K L(yjq) LOSS m mM,k Msurf P5 peff
1.
total impervious area [L2] simulated stormwater concentration [M L3] pollutant concentration of the rain [M L3] emission factor [M T1 L2] event mean concentration [M L3] posterior parameter distribution water level height of the reservoir [L] normalising constant reservoir coefficient [L3/5 T1] likelihood measure maximum initial loss for one event [L] vector with model outputs simulated sampled masses [M] stored mass on the impervious surfaces [M L2] maximum rain intensity [L T1] effective rain intensity [L T1]
Introduction
Heavy metal pollution originating from stormwater runoff from paved surfaces is among the most significant sources to poor environmental quality of urban water courses (e.g. Christensen et al., 2006; Eriksson et al., 2007; Karlaviciene et al., 2009; Kayhanian et al., 2008). Heavy metals are of particular interest in stormwater runoff due to their toxicity, ubiquitous feature, and the fact that metals cannot be biologically transformed. Monitoring generally reveals high variability of metal concentrations from site to site, from event to event and within events, due to the multitude of diffuse urban sources of metal pollution (Lu¨tzhøft et al., 2009; Chon et al., 2010) and the complexity of the processes leading to accumulation on urban surfaces as well as release and transport during rainfall-runoff (Bertrand-Krawjewski, 2007). This makes it difficult to model and thus to predict heavy metal concentrations and loads accurately, which is needed as part of the efforts to limit urban emissions of heavy metals to surface waters as required by the European Water Framework Directive (European Union, 2000, 2008). Several conceptual computer models have been developed for analysing water quality problems related to stormwater runoff (e.g. Achleitner and Rauch, 2005; Calabro, 2001; Obropta and Kardos, 2007; Ruan, 1999; Wong et al., 2002). If appropriately applied, these constitute tools for enhancing further understanding, for predicting flow and water quality in urban drainage systems and receiving waters and thereby for decision support in relation to implementation of monitoring programmes or pollution mitigation strategies. Past applications of such models have however focused on macro pollutants such as nutrients and organic matter characterised as COD or BOD, and only very few have focused on heavy metals and other priority substances. In a conceptual model the mechanistic details are simplified by considering empirical relationships. Thus, the parameters of a conceptual model need to be adjusted by comparing simulations with measured data, i.e. a calibration of the model has to be performed. As pointed out by e.g. Freer
Q SMC SSE T tdry tj tk Vj y yC,k yM,k yQ q q1 q2 q3 f
simulated stormwater flow [L3 T1] site mean concentration [M L3] sum of squarred errors scaling factor used in L(yjq) dry weather period [T] observed event duration [T] time period for the sample collection [T] observed stormwater event volume [L3] vector with experimental observation flow prop. concentration measurement [M L3] intra-event sample mass [M] measured stormwater flow [L3 T1] the entire model parameter vector dry deposition load [M T1 L2] rate coefficient for dry removal [T1] rate coefficient for wet removal [L1] runoff coefficient [-]
et al. (1996), the gain of finding solely one "optimally calibrated" solution is limited, because there will be many others that are almost equally good and if a second period of data is considered, then the rankings of these will change and the best solutions found for the first period will not be the best for the second. Parameter estimates are thus associated with uncertainty, which will influence the predictive ability of the model; this is true for many environmental systems in which observations to support model calibration often are relatively sparse, and in particular to the case of urban stormwater runoff quality where the dynamics exceed those of most other environmental systems. Early work on uncertainty in relation to urban drainage modelling was based on simple first order analysis or forward Monte Carlo simulation based on assumed uncertainties in input and parameters (e.g. Arnbjerg-Nielsen and Harremoe¨s, 1996; Daebel and Gujer, 2005; Hansen et al., 2005; Harremoe¨s et al., 2005; Lei and Schilling, 1996). Currently, literature tends to concentrate on general water quality parameters (TSS, organic matter, nutrients) where large uncertainties exist (e.g. Willems, 2006) and methods to condition model predictions on observed data are being applied and evaluated (e.g. Dotto et al., 2010, 2011; Freni et al., 2008, 2009; Freni and Mannina, 2010; Kleidorfer et al., 2009; Thorndahl et al., 2008), however mostly without explicitly quantifying the uncertainty in relation to how well the model predictions are able to cover the available observations. The objective of the current work is to analyse the uncertainty related with model predictions of heavy metal loads in stormwater runoff, which is considered more uncertain than modelling of general water quality parameters. The results are derived conditional on (i) a pre-defined dynamic conceptual stormwater rainfall-runoff accumulation-washoff model, (ii) a fixed set of results from a field sampling campaign, and (iii) a desire to cover 95% of the observations with the 95% empirical prediction bounds. The applied model is a re-formulation of SEWSYS that was developed for simulation of water flow and quality in urban drainage systems (Ahlman and Svensson, 2002). The experimental data include monitored rain
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 2 3 e3 8 3 5
3825
intensities and stormwater flow as well as intra-event flowproportional concentration measurements of copper (Cu), zinc (Zn), lead (Pb) and cadmium (Cd). The uncertainty was assessed for predictions of the event mean concentrations (EMCs) and the site mean concentrations (SMCs), as these are important variables in determining the total pollutant load from the area, as well as for communicating the results and comparing them with other studies.
2.
Material
2.1.
The catchment and field samples
The samples forming the experimental data were collected at the outlet of a separate sewer system in Vasastaden, an urban district in the city centre of Go¨teborg, Sweden (Fig. 1). The area is densely populated, consists mainly of older residential and commercial buildings and has a separate sewer system. The roof (1.95 ha) and road areas (1.15 ha) of the total impervious catchment area (4.83 ha) have been derived using GIS data. The percentage of Zn (5%) and Cu (3%) roofs as well as traffic activity data and other catchment-specific data have been estimated or measured as well. For further information about the catchment and field survey the reader is directed to Ahlman (2006). The sampling campaign was running during five weeks in AprileMay 2002. Fig. 2 exemplifies the characteristics and used notation for the resulting data. A flow meter (an ADS 3600 equipped with ultrasonic sensors for level and velocity, and a pressure depth sensor for backup) recorded the stormwater flow, yQ [L3 T1], with a 2-min resolution. Rainfall data (1-min resolution) was collected with a tipping bucket rain
Fig. 2 e Available data and main model outputs exemplified for event #3 and the heavy metal Zn. Lower panel: Recorded stormwater flow ( yQ, solid line), sampled stormwater volumes (light and dark grey coloured areas) and simulated stormwater flow (Q, dashed line). Upper panel: 5 simulations of the stormwater concentration (C, grey lines) for various values of the model parameters and flow-proportional concentration measurements collected during the time periods tk ( yC,k, horizontal black lines).
gauge (type HoBo/MJK), located approximately 60 m from the catchment outlet. An ISCO 6700 automatic water sampler was used to collect flow-proportional samples each representing approximately 20 m3 of runoff volume over time periods tk [T]. The concentrations of Zn, Cu, Pb and Cd in these were determined by inductively coupled plasma mass spectrometry (ICP-MS) and are denoted yC,k [M L3]. Table 1 shows a summary of the experimental results for each of the j ¼ 1,2,.,18 identified events. Each event included a number of sub-samples, which is shown furthest to the left in parenthesis. In total, 57 samples were collected. Moving to the right in the table, the event durations (tj) and the entire event volumes (Vj) are shown. Only part of the event volumes were actually sampled for further quality analysis, which is indicated as percentages in parenthesis. This can also be seen in Fig. 2 (upper panel), where the integral of the first 5.5 h of the hydrograph is shown with light and dark coloured areas for the 14 sampled volumes of event #3. The durations of the antecedent dry weather periods (tdry) and maximum rain intensities (P5) are shown in the table as well. The average rainfall for the month of May in Go¨teborg is approximately 50 mm, and the measured rainfall in May amounted to 57 mm. The latter are shown as the maximum over a 5-min period, as this was the resolution used as a model input. By combining the recorded stormwater flow with the flowproportional concentration measurements, the observed intra-event sample masses yM,k [M] are given as: Z yM;k ¼ yC;k ,
Fig. 1 e The monitored catchment Vasastaden. y and p indicates the catchment outlet where samples were taken and the location of the rain gauge.
yQ dt
(1)
tk
The EMCs shown furthest to the right in Table 1 have been calculated by summing up the sampled masses available for
3826
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 2 3 e3 8 3 5
Table 1 e A summary of the experimental data on an event-basis. See the text for details. Event j tj [h] Vj [m3] tdry [d] P5 [mm h1]
EMCa [mg l1] Zn Cu Pb
1 2() 3(14) 4(3) 5() 6(6) 7(2) 8(1) 9(1) 10(1) 11(3) 12(3) 13(2) 14(8) 15(2) 16(11) 17() 18()
13.1 2.6 6.4 7.5 6.7 3.1 1.1 3.5 2.5 1.5 3.0 3.4 1.3 7.9 2.3 14.9 4.4 15.7
145() 11() 351(96%) 108 (88%) 30() 159(85%) 35(77%) 58(81%) 28(72%) 40(61%) 85(86%) 72(94%) 21(83%) 195(85%) 40(82%) 420(55%) 75() 295()
e 1.1 0.6 0.3 0.8 0.2 0.4 6.9 0.6 2.8 2.2 1.7 3.6 4.4 0.6 1.4 1.3 0.4
2.5 0.8 8.2 4.1 0.8 4.1 3.3 9.9 3.3 5.8 6.6 2.5 1.6 4.1 3.3 4.1 1.6 3.3 SMCa
e e 343 370 e 292 288 752 1050 951 798 619 1436 688 428 294 e e 470
e e 254 258 e 181 219 887 600 736 334 344 632 345 233 169 e e 295
e e 23.4 16.0 e 9.9 10.5 103.2 35.5 44.4 23.2 11.3 17.1 18.3 8.8 4.9 e e 19.5
Cd e e 0.40 0.43 e 0.33 0.28 1.01 1.72 1.66 1.01 0.67 1.71 0.85 0.82 0.37 e e 0.59
a The shown EMCs and SMCs have been estimated based on partially sampled event volumes.
each event and by dividing this "partially" observed event mass with the corresponding "partial" event volume. The SMCs at the bottom of the table were calculated in the same manner but with all 57 samples. The measured concentrations are fairly high compared to other studies with mixed land use. Gnecco et al. (2005) e.g. report a median EMC for Zn of 408 mg/l for runoff from a roof covered with slate and with zinc gutters, and runoff from a heavily trafficked highway is reported by Pettersson et al. (2005) to have a median EMC of 290 mg/l for Zn, and 82 mg/l for Cu. Gromaire-Mertz et al. (1999) however report extremely high concentrations of Zn with a median EMC of 3200 mg/l for runoff from a roof with pure zinc sheets and gutters, and when comparing with e.g. the review by Go¨bel et al. (2007) and considering that the Vasastaden catchment has significant traffic and metal roof sources the concentrations reported in Table 1 seem within the expected range.
2.2.
The SEWSYS model
The conceptual stormwater model SEWSYS (running in MATLAB/Simulink) has been developed for simulations of substance flows in urban drainage systems and acts as a starting point for the study. The substances are modelled with different source parameters such as corrosion rates, atmospheric deposition and other material emission factors which are used to describe pollutant build-up on three different types of surfaces: roofs, roads and other impervious areas. The pollutants are accumulated in dry periods and washed off during rainfall, processes described with classical build-up and wash-off functions (e.g. Overton and Meadows, 1976). Default parameter values have been obtained from literature studies and in particular Swedish inventory reports. For further information about the SEWSYS model and the
default values the reader is directed to Ahlman and Svensson (2002), Ahlman et al. (2005) and Ahlman (2006). Main outputs of SEWSYS include stormwater volumes and pollution loads, which can be combined to yield EMCs and SMCs. The following compact continuous form of SEWSYS has previously been formulated in Lindblom et al. (2007a,b): 5=3
QðtÞ ¼ Aimp ,K,hðtÞ
(2)
dhðtÞ ¼ f,peff K,hðtÞ5=3 dt
(3)
dMsurf ðtÞ ¼ Aimp ,q1 q2 þ q3 ,peff ðtÞ ,Msurf ðtÞ dt
(4)
dCðtÞ peff ðtÞ , q3 ,Msurf ðtÞ þ f,Aimp ,ðCrain CðtÞÞ ¼ dt hðtÞ,Atot
(5)
In Equation (2), h [L] is the water level in a non-linear reservoir, K [L3/5 T1] the reservoir coefficient and Aimp [L2] the total impervious area. In Equation (3), f [-] is the runoff coefficient and peff [L T1] the effective rain intensity (the rain intensity remaining after subtraction of initial loss). The initial loss is determined by the parameter LOSS [L], which represents the maximum initial loss for one single event. In Equation (4), Msurf [M L2] is the stored pollutant mass on the total impervious surface and q1, q2 and q3 represent conceptual parameters depending on compound- and catchment-specific properties. The dry deposition load q1 [M T1 L2] can be seen as a model input variable that is calculated from compoundspecific as well as catchment-specific information. Different sources (e.g. traffic activities, surface corrosion and atmospheric deposition) are combined with catchment specific attribute data to give various emission factors, EFj [M T1 L2], which are weighted with the specific areas Aj for which they are relevant. q1 ¼
X
Aj ,EFj =Aimp
(6)
j
q2 [T1] is the rate coefficient for pollutant dry removal (removal by wind and other decay) and q3 [L1] is the rate constant for wet removal by wash-off. Finally in Equation (5), C represents the stormwater pollutant concentration [M L-3] and Crain [M L-3] the pollutant concentration of the rain. The modelled sample masses (compare Equation (1)), mM,k [M], are given from the model by: Z mM;k ¼
CðtÞ,QðtÞ dt
(7)
tk
3.
Uncertainty-based model calibration
A set of N experimental observations y¼( y1,.,yk,.,yN) will never exactly equal the associated model outputs m¼(m1,.,mk,.,mN). This is so because of a number of incorporated uncertainties e.g. model structure, model parameters, input data and measurement uncertainty, see e.g. Walker et al., 2003 and Refsgaard et al. (2007) for further details on classifications of uncertainties. As opposed to traditional
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 2 3 e3 8 3 5
model calibration where the aim commonly is to provide “as certain as possible” predictions of the observations by finding “optimal” model parameters, the aim here is to describe the predictive uncertainty in a quantitative manner. The chosen approach is based on projecting all the uncertainty on the model parameter vector q of the deterministic model, which in this paper is seen partly as random (or at least not fixed). The selection of parameters to treat as random will be described in Section 4.1. q is defined by the so called joint posterior parameter distribution g(qjy), which is obtained by updating a prior parameter distribution p(q), indicating what is known about the parameters before observing the experimental data, with a likelihood measure L(yjq), indicating how well different parameter sets from the prior distribution perform when comparing model outputs with observations: 1 gðqjyÞ ¼ ,LðyjqÞ,pðqÞ k
(9)
Here k is a normalising constant ensuring that g is a true probability distribution (the integral of g(qjy) over all q should be unity). This way of expressing model uncertainty as parameter uncertainty is closely connected to the Bayesian statistical paradigm in general and to the generalized likelihood uncertainty estimation method (GLUE) of Beven and Binley (1992) in particular. According to Beven (2008), “GLUE is a form of Bayesian model conditioning methodology without the need for defining a formal structure for the errors”, and the “Bayesian identification of models is a special case of GLUE.” (Beven et al., 2008). McIntyre et al. (2002) refers the conversion of the likelihood response surface into a “calibrated” parameter distribution like in Equation (9) as “uncertainty-based model calibration”; this notation is hereby used. Importantly, the posterior g(qjy) is assumed to include all knowledge about the statistical properties of a certain model output. The model prediction uncertainty is given by Monte Carlo propagation of the posterior through the model followed by analysis of the empirical prediction quantiles. For example, with n ¼ 1,2,.,N draws of parameter sets (q(n)) from g(qjy) the probability P that a model output mk lies in a certain region Ak is calculated by simply counting and averaging the times this happens: Pðmk ðqÞ˛Ak Þ ¼
N 1X j qðnÞ N n¼1
(10)
where the function J takes the value 0 or 1: jðqÞ ¼
1 0
if mk qðnÞ ˛Ak else
(11)
The region Ak in which 95% of the model outputs fall yields P ¼ 0.95 and the empirical 95% model prediction quantile. To form the prediction quantiles of Equations (10) and (11), draws of q from the posterior parameter distribution is required. Here, the Metropolis algorithm (Metropolis et al., 1953), the originator to what today is called Markov Chain Monte Carlo (MCMC) methods, is used. The theory behind MCMC algorithms lies beyond the scope of this paper but is well documented in statistical literature like Robert and Casella (2004) and Tanner (1996). The practical implementation implemented here is treated in Section 5.1 below.
3827
4. Model reformulation and preparation for uncertainty analysis 4.1.
Model simplification and sensitivity analysis
Before considering experimental data the model parameter sensitivities were assessed using Monte Carlo simulation in combination with multi-linear regression as proposed by Lei and Schilling (1996). For each parameter of the original SEWSYS model a log-normal distribution (to avoid negative parameter values) was defined with the default value as mean and a coefficient of variation (CV, standard deviation divided by the mean) in the range 0.2e1.0, inspired by Daebel and Gujer (2005) and Hansen et al. (2005) for the hydrological parameters and heuristically assessed (in the high range) for the accumulation-washoff parameters. Monte Carlo simulations were then conducted and used to calculate first order model parameter sensitivity coefficients, which indicate the relative contribution of a particular parameter variance to a model’s total output variance. The purpose was to identify the most sensitive parameters and to focus on these in the proceeding uncertainty-based calibration. More elaborate, Global Sensitivity Analysis methods have been advocated in recent years (e.g. Saltelli and Annoni, 2010) and even applied within stormwater modelling (e.g. Vezzaro et al., 2011), and formal methods for expert elicitation have furthermore been suggested for use in the urban water sector (e.g. Garthwaite and O’Hagan, 2000). However, there is yet no consensus on the use of these methods (see e.g. Dotto et al., 2010) and we consider the OAT (One-At-Time) method used here in combination with heuristically chosen CV’s appropriate as an initial step when preparing the model for uncertainty-based calibration.
4.1.1.
The rainfall-runoff module
To simulate the stormwater event volumes the integral of Equation (2) was calculated. The sensitivity analysis showed that the parameters f and LOSS were influential, but not K. LOSS mainly had an effect on the smaller events and does not widely affect the simulated total stormwater volume. For simulation of the EMCs, the parameters f and LOSS were both influential whereas for simulation of the SMCs they did not have a large impact. By comparing observed (Vj in Table 1) and simulated total event volumes, a set of model parameters (f ¼ 0.62 and LOSS ¼ 0.34 mm) was derived with linear regression. Fig. 3 shows the good performance of this set in simulating both the observed total event and partial event volumes. Since the event volumes are well predicted with the model and since the simulation of SMCs were not sensitive to f and LOSS, these parameters were kept fixed in, and excluded from, the proceeding uncertainty analysis.
4.1.2.
The accumulation-washoff module
The simulation of EMCs and SMCs were sensitive to the emission factors, forming the dry deposition load (q1), as well as to the dry and wet removal rates (q2 and q3). This was true for all 4 compounds. Regarding the concentration of the pollutants in the rain (Crain), this parameter also had a significant sensitivity coefficient for Pb and Cd. For Zn and Cu
3828
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 2 3 e3 8 3 5
4.2.
Fig. 3 e Simulated and observed total (dots) and partial (crosses) event volumes.
however, wet deposition is only a small part of the total load. To simplify the uncertainty analysis it was decided to keep Crain constant, cf. Table 2. The SEWSYS model assumes spatially uniform rainfall and no spatial dimensions of the physical catchment area. The samples were collected at the catchment outlet only. The emission factors forming the dry deposition load (Equation (6)) can thus not be identified from the results of the sampling campaign. Various sets of SEWSYS parameters e.g. "the emission factor for Cu surface corrosion" and "the emission factor for Cu in brake wear" give identical Cu loads in the catchment outlet. In a similar manner, the information content of the data was simply too small to allow distinguishing between dry and wet deposition. As a result it was found appropriate to replace the different areas of SEWSYS (cf. Equation (6)) with one single total impervious area, to disregard Equation (6) and consider q1 a lumped parameter (rather than an input) that represents several sources of pollution, depositing pollutants uniformly on the total impervious area. The model parameters considered as fixed are summarized in Table 2. In the following, only the parameters q1, q2 and q3 are considered random variables in the entire model parameter vector q.
Table 2 e An overview of input variables assumed to be constant and their values. The rain concentrations are those used as default values in SEWSYS, obtained from Swedish inventory reports (Notter, 1993; Stockholm Vatten, 1999). Parameter f LOSS K Crain: Cu Zn Pb Cd
Value
Unit
0.62 0.34 0.4
[e] mm m3/5$s1
2.10 11.20 2.30 0.21
mg$l1 mg$l1 mg$l1 mg$l1
Prior considerations
The function of the prior parameter distribution is to model what is known about the parameters before having considered specific data. The default values of the accumulationwashoff parameters of SEWSYS are all based on realistic considerations from literature. However, the available data do not allow for stating any specific statistical distributions for those, and it was therefore decided to consistently assume uniform prior parameter distributions throughout the paper (the use of uniform priors in a similar application has recently been evaluated and justified, see Freni and Mannina, 2010). Table 3 summarises the resulting parameter ranges for the uniform prior parameter distributions, which were derived by the following reasoning. From the structure of Equation (4) it is seen that the ratio between q1 and q2 determines the maximum mass of pollutants that can accumulate on the surface during dry weather, while the time required to achieve this mass depends on the inverse of q2. The value of q3 determines the rate, at which the accumulated pollutants are depleted during wet weather. The durations of the dry weather periods (Table 1) for the sampling campaign were in the order 0.2e6.9 days. The range for q2 was established by assuming that the time to establish pollutant equilibrium during dry weather ranges between around 10 h and 10 days. A faster pollutant build-up would mean that we move towards a different model structure where the equilibrium is obtained instantaneously. A slower build-up rate would be possible but since the available experimental data is limited the above-mentioned upper limit was chosen. Fig. 4 (left) shows two simulations of the entire sampling campaign period with the extremes of the applied prior for q2. With the high build-up rate (2.5 d-1) equilibrium occurs quickly also during the short dry periods while with the low build-up rate (0.1 d-1) the relatively long dry period for t ¼ 5e10 days is far from enough to establish a new equilibrium. The ranges for q3 were established by arguing that the time constant for pollutant depletion due to a (hypothetical, rectangular) rain with high intensity (2 mm s-1 ¼ 7.2 mm h-1) should not be smaller than 8 min whereas for a moderate rain (0.2 mm s-1 ¼ 0.72 mm h-1) it should not be longer than 1 day. In Fig. 4 (right) a simulation of event #3 with the extremes of the applied prior for q3 is shown. The "fastest" wet removal
Table 3 e An overview of the uniform prior parameter distributions. The default values of SEWSYS are included for comparison. For q1, the default values have been calculated from the default emission factors of SEWSYS together with catchment-specific data for Vasastaden. Parameter
Default
q1 Cu Zn Pb Cd q2 q3
156.71 421.54 24.70 0.47 0.40 0.60
Prior range Min 0.00 0.00 0.00 0.00 0.10 0.05
Max Inf Inf Inf Inf 2.5 2.0
Unit mg d1 m2 mg d1 m2 mg d1 m2 mg d1 m2 [d1] [mm1]
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 2 3 e3 8 3 5
3829
Fig. 4 e Left: A simulation of the wash-off process with the extremes (solid and dashed lines) of the applied prior for q2. Right: A simulation of the wash-off process with the extremes (solid and dashed lines) of the applied prior for q3.
process (2 mm-1) involves that all accumulated pollutants on the surface are washed out within an hour, despite the initially moderate rain intensity of approximately 0.5 mm s-1. Contradictory, by applying the “slowest” wet removal rate value (0.05 mm-1) only part of the surface pollutants are washed off during the 5 h rain event. With these restrictions on the rate parameters it was possible to have an open not specified uniform prior for the dry deposition parameter, q1.
application by Mailhot et al. (1997) and was found practical for this application because (i) the sum of squared errors is common measure to evaluate the goodness of fit, (ii) negative errors are weighted in the same manner as positive errors, and (iii) the likelihood decreased exponentially with the SSE at a rate (T ) that can be adjusted manually.
5.
Application and results
4.3.
5.1.
Calibration: partial event loads
Specification of the likelihood function
While the chosen "flat" structure of the prior parameter distribution is quite obvious when the available prior information is limited, there are numerous available likelihood functions proposed in literature (see Beven and Freer, 2001) and the choice is subjective and may influence the results (Freni et al., 2009). Often the model errors are assumed to be independent and normally distributed. This was done in Kanso et al. (2005) where suspended solid concentrations in stormwater runoff with a model similar to the one presented here were studied. In Freni et al. (2008) the (also frequently applied) Nash and Sutcliffe (1970) efficiency criterion was used to analyse the model predictive uncertainty of maximum peak flow, runoff volume, maximum peak BOD concentration, BOD load and maximum oxygen depletion in a down-stream river cross-section, and it has subsequently been used in several urban drainage applications (e.g. Kleidorfer et al., 2009; Dotto et al., 2010). In this paper the following likelihood function structure has however consistently been applied: LðyjqÞ ¼ expf SSEðq; yÞ=Tg
(12)
where exp denotes the exponential function and SSE the sum of squared errors between observed and simulated partially observed event masses calculated as: 12
0 SSEðq; yÞ ¼
X j ¼ 3; 4; 6; ::; 16
B B B @
X
yM;k
All k ˛Event j
X
C C mM;k C A
(13)
All k ˛Event j
The parameter set minimising the sum of squared errors is given the highest likelihood while the decrease in likelihood due to larger errors will depend on the value of the scaling factor T. Equation. (12) was previously used in a similar
Having reformulated the original SEWSYS model and defined the prior distribution and structure of the likelihood function, the next step is to generate a sequence of model parameter samples from g(qjy), as defined in Equation (9). To do so we need to tune the Metropolis algorithm and chose values for the parameter T.
5.1.1.
Tuning of the metropolis algorithm
The Metropolis algorithm implementation is inspired by the work of Thyer et al. (2002) and Kuczera and Parent (1998). The main iterations are shown in Box 1. The spread of 3, defined by the matrix V and the scalar s such that 3wN(0,s$V), was updated during the initial iterations to give an efficient acceptance rate. V was updated to be proportional to the covariance matrix of the already accepted
Box 1 The main iterations of the implemented Metropolis algorithm. 1. Starting from a parameter set q(i), a second set q(*) ¼ q(i) þ 3 is proposed, where 3 is a normally distributed (symmetric) multivariate with mean 0. 2. q(iþ1) is updated to q(*) with acceptance probability a: ! g qðÞ jy a qðiÞ ;qðÞ ¼ min 1; ðiÞ g q jy n o ! exp SSE qðÞ ;y T ,p qðÞ n ðB:1Þ ¼ min 1; o exp SSE qðiÞ ;y T ,p qðiÞ If the proposal is rejected, the chain stays at its’ previous value, e.g. q(1þ1) ¼ q(1)
3830
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 2 3 e3 8 3 5
Fig. 5 e Left: Illustration of the metropolis algorithm and the function of the scaling factor T. See text for details. Right: Observed (dots) and predicted partial event masses (95% and 50% quantiles) for Cu after calibration.
parameters and s was tuned so that approximately 25% of the proposals were accepted, a rule-of-thumb for the current type of MCMC application (Robert and Casella, 2004). When the algorithm had been tuned, the updating of s and V ended and the subsequent sequence of parameters was considered an approximate draw from g(qjy).
5.1.2.
Deciding when the uncertainty is adequately described
The influence of the proposals on the acceptance rate mentioned above is of purely "numerical" nature; a too high rate involves that the posterior will be explored slowly by the algorithm whereas a too low rate will involve that the algorithm might be "trapped" in limited parts of the entire parameter space. The second factor determining the acceptance rate is the value of T in the likelihood function. From Equation B.1 it is seen that if SSE(q(*)) < SSE(q(i)) and if q(*) is within the range defined by the prior the acceptance probability is 1, e.g. moves to areas with higher posterior probability are always accepted. In cases where q(*) lies outside the prior, both the prior and posterior probability is 0 and the proposal is rejected. Now assume the example illustrated in Fig. 5 (left) where we, from a parameter set within the prior range that gives a “good” simulation with SSE(q(i)) ¼ 75, propose a set within the prior range that is not as good yielding SSE(q(*)) ¼ 275. If the two sets are evaluated with T ¼ 100, the acceptance probability is 0.06/0.47 ¼ 0.14 (solid arrows) whereas if T ¼ 300 is applied, we accept the proposal with probability 0.40/0.78 ¼ 0.51 (dashed arrows). Thus, by changing the value of T it is "chosen" to what degree "bad" simulations will be accepted. This is naturally a highly subjective choice that is related to the degree of belief the modeller has in the data as compared to the model. In this study we presume that the observed data are representative of reality. We found it appropriate to vary T manually until the empirical 95% modelled prediction limits (calculated with Equation 10) covered 95% of the observed partial event masses (in practise 12 out of 13 partial event masses were covered, corresponding to 92%). This involves that the Metropolis algorithm shown in Box 1 was run and tuned for various values of T for Zn, Cu, Pb and Cd. To bracket the observations, T had to be increased to rather high values for all four compounds, involving that parameter sets within the entire ranges defined in the priors were
considered for the dry (q2) and wet (q3) removal rates. Fig. 5 (right) exemplifies the calibration results by showing the 13 observed and predicted partial event masses for Cu; similar results were obtained for Zn, Pb and Cd. In Fig. 6, draws from the posterior parameter distributions, which were used to construct these quantiles for Cd as well as for Zn, are shown. A positive correlation between q2 and q1 is seen as well as an inverse correlation between q3 and the q1/q2 ratio, which determines the maximum pool of accumulated metal available on the conceptual catchment surface. The results for Cu and Pb were similar in shape and are not shown here.
5.2.
Simulation: event mean concentrations
With the derived posterior parameter distributions the uncertainty of any other model output is easily computed with Equations 10 and 11. Fig. 7 shows the 13 simulated and partially observed EMCs that were included in the uncertainty-based calibration, as well as the simulated EMCs for the 5 events for which stormwater concentrations were not measured (cf. Table 1). The uncertainty-based calibration seems to adequately describe the uncertainty associated with using the model for prediction conditioned on the data from the Vasastaden area, as the forward Monte Carlo analysis output covers approximately 95% of the observed EMCs (to be precise all EMCs are covered for Zn, Cu and Cd, whereas 12 out of 13 EMCs are covered for Pb).
5.3.
Simulation: site mean concentration
The histograms of the SMCs, simulated with the posterior parameter distributions are shown in Fig. 8 together with the “partially” observed SMCs of the sampling campaign (cf. Table 1). The histograms are symmetric (the mean of the distribution is close to the median) and the upper and lower bound of the 95% quantiles are approximately 20%, 40%, 80% and 35% of the medians for Zn, Cu, Pb and Cd, respectively. It is noted that the prediction uncertainty is larger for Pb than for the other metals, although Pb is usually considered strongly attached to particles, which is what the model’s accumulation wash-out function was originally developed for. The larger uncertainty suggests that other phenomena not included in the model such as formation of Pb
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 2 3 e3 8 3 5
3831
Fig. 6 e Draws from the posterior parameter distribution generated with the Metropolis algorithm.
precipitates may play a role; this is however not further investigated due to lack of information about the chemical matrix. Since the results are conditional on the fixed hydrological model and sampling data, the model uncertainties of the pollutant load predictions are given directly from the SMC histograms.
6.
Discussion
In Lindblom et al. (2007a; 2007b) analyses similar to the one presented here were conducted for Cu only. Although the
current and the referenced analyses are not directly comparable (in the two previous publications it was the intra-event sampled masses and cumulative sampled masses that were included in the likelihood function, respectively), comparison of the results indicate some interesting features of the presented method. Following the notation of the GLUE methodology, the parameter values shown in Fig. 6 are referred to as behavioral and should be considered as plausible for simulating the Zn and Cd loads. The parameter values are derived as parameter sets and the statistical properties of the parameters alone (the marginal parameter distributions) are not directly provided. For example, for both Zn and Cd
Fig. 7 e Simulated EMCs (95% and 50% prediction quantiles) for the 18 events together with the 13 “partially” observed EMCs [mg/l].
3832
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 2 3 e3 8 3 5
Fig. 8 e Histograms of the site mean concentrations [mg/l] for the four compounds with results of the measurement campaign indicated with crosses.
behavioral values of q2 and q3 are found within the entire prior ranges, cf. Table 3. However it is seen that low values of q2 are behavioural only conditionally on low values of q1; if you chose to simulate the model with little removal by e.g. wind you should also chose a small dry deposition load. Similarly, low values of q3 are behavioural conditionally on high values of the quota q1/q2; to predict the observations with a small wet removal rate a relatively large pool of accumulated pollutants on the surfaces is required. In Lindblom et al. (2007a) different and wider prior parameter distributions were used and the derived posterior parameter distributions were wider as well. Still, the model output uncertainty was of the same order of magnitude as here. The “take home” message is that model output uncertainty not necessarily increases with parameter uncertainty and that in the present application, the parameter uncertainty to a large degree depends on the assumed prior parameter distributions. This is typical for models where the parameter correlation is high and for situations where available data is not sufficient to identify the parameter values. If parameter uncertainty is of interest it is thus recommended that that the minimum and maximum values of uniform prior distributions are, as in this paper, carefully defined.. The resulting model output uncertainties in Fig. 5 (right), Figs. 7 and 8 are obtained without having made assumptions considering the input data uncertainty or measurement (analytical) uncertainty. The fact that errors are not explicitly modelled is typical for the GLUE methodology as well. Two common arguments for not including a formal likelihood function (error model) often emphasised is (i) that it is usually impossible to verify an assumed statistical error structure due to the limited available data and (ii) that making predictions with an additive error model often involves negative e.g. runoff rates and concentrations, which are difficult to interpret. This exclusion of a formal
error model is also subject to criticism. Mantovan and Todini (2006) baptize GLUE as “pseudo-Bayesian”, arguing that the method is an inconsistent and incoherent statistical inference process and that it overestimates the parameter uncertainty. This might be true but is not crucial for the present study, since we only consider the parameter uncertainty and correlation indirectly as a means to assess the uncertainty of the model output. Furthermore, we emphasize that there is not yet consensus on the use of uncertainty assessment methods in urban drainage modelling (see e.g. Dotto et al., 2010) and that this applies in particular to modelling of heavy metal runoff as in this study where measurements are extremely scarce. It is noted that the model simplification and sensitivity analysis carried out to in section 4.1 could be done in a more elaborate manner using GSA methods (see e.g. Saltelli and Annoni, 2010) or formal methods for identifiability analysis (e.g. Freni et al., 2011), and that more than three parameters could also have been considered random in the GLUE analysis followed by inspection of dotty plots as done in some recent studies within urban drainage modelling (e.g. Dotto et al., 2011). It is likely that either of these approaches might have identified that the hydrological model parameters have little influence on the predicted SMCs, that the information content in data is too little to distinguish different emission sources in the catchment, cf. Equation (6) and that the accumulation-washoff parameters are correlated as illustrated in Fig. 6. However, we explicitly chose to base the investigation on a relatively small catchment with good flow-measurements and a close-by rain gauge to eliminate the influence of non-homogenous spatial rainfall on the hydrological model parameters, and the simplifications related to Equation (6) follow directly from its linear structure. Therefore, the simple approach combining OAT sensitivity analysis with heuristic considerations is justified. We furthermore emphasize that the model simplification and sensitivity analysis is not the main paper of this paper; it was only an initial step necessary to allow focusing the uncertainty-based calibration on a reasonable number of relevant parameters that allowed covering 95% of the observations considered. Using observation coverage as e measure of goodness in the uncertainty-based calibration adds an element of objectivity to the method that is not commonly found in GLUE applications.
7.
Conclusions
In this paper the uncertainty related with model predictions of heavy metal loads in stormwater runoff was investigated and we derived the following quantitative results: The observed SMC for zinc (470 mg/l) was predicted with 20% The observed SMC for copper (295 mg/l) was predicted with 40% The observed SMC for lead (20 mg/l) was predicted with 80% The observed SMCs for cadmium (0.6 mg/l) was predicted with 35%
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 2 3 e3 8 3 5
The results were obtained by developing and using an uncertainty-based calibration procedure, where we combined the GLUE technique with the Metropolis algorithm and used an exponential likelihood measure based on measured event masses scaled to cover 95% of the observations within the empirical 95% model prediction bounds. This allowed identifying a range of behavioural model parameter sets that adequately described the uncertainty involved with making model predictions conditional on (i) a pre-defined dynamic conceptual stormwater accumulation-washoff model and (ii) a selected set of data from an intensive field monitoring sampling. The results are fairly high compared to other studies with mixed land use and the larger uncertainty for Pb compared with the other metals suggest that processes in addition to particle transport not included in the model may play a role. Reality moreover includes a site-to-site and seasonal variation that is not considered in this study, i.e. predicting metal concentrations in stormwater runoff from gauged as well as ungauged catchments is generally more uncertain than the indicated numbers indicate. Although we have strived to make as correct assumptions as possible (e.g. the derivation of realistic accumulation-washoff parameter ranges through model performance analysis) the methodology and way towards these numeric results raised a number of subjective judgements (e.g. exclusion of the hydrological sub-model from the uncertainty assessment, declaration of a number of model parameters as constant and the definition of the likelihood function). However, it is believed that the presented method is adequate for the purpose of analyzing the uncertainty related with model based prediction of stormwater pollutant loads, since the requirement to bracket the observations adds an element of objectivity that is not always included in uncertainty assessment of environmental models. Since the entire prediction uncertainty is described with the empirically derived posterior parameter distributions and since no assumptions about parameter correlations or the statistical structure of model errors need to be made, parameter sets obtained with the developed method can be directly used for prediction and scenario analysis purposes. Using this kind of uncertainty assessment will greatly enhance the trustworthiness of using dynamic models as support in decision making related stormwater management practices.
Acknowledgement The research work of Stefan Ahlman received financial support from the Swedish Foundation for Strategic Environmental Research (MISTRA). The Go¨teborg Water and Wastewater Works is acknowledged for their financial support and help with the field measurements of stormwater. We thank Luca Vezzaro, Technical University of Denmark for assisting with calibrating the hydrological rainfall-runoff model used as a basis for this investigation.
3833
references
Achleitner, S., Rauch, W., 2005. CITY DRAIN ª - A simulation software for integrated modeling of urban drainage systems. In: Eriksson, E., et al. (Eds.), Proceedings of the 10th International Conference on Urban Drainage August 21-26 2005, Copenhagen, Denmark. Ahlman, S., Svensson, G., 2002. Modelling substance flows in urban sewer systems using MATLAB/Simulink. In: Strecker, E.W., Huber, W.C. (Eds.), Proceedings of the Ninth International Conference On: Urban Drainage September 8-13 2002, Portland, OR, USA. Ahlman, S., Malm, A., Kant, H., Svensson, G., Karlsson, P., 2005. Modelling non-structural best management practices - focus on reductions in stormwater pollution. Water Science and Technology 52 (5), 9e16. Ahlman, S., 2006. Modelling of substance flows in urban drainage systems. PhD thesis, Dept. of Civil and Environmental Engineering, Division of Water Environment Technology, Chalmers University of Technology, Go¨teborg. Arnbjerg-Nielsen, K., Harremoe¨s, P., 1996. The importance of inherent uncertainties in state-of-the-art urban storm drainage modelling for ungauged small catchments. Journal of Hydrology 179 (1e4), 305e319. Bertrand-Krawjewski, J.L., 2007. Stormwater pollutant loads modelling: epistemological aspects and case studies on the influence of field data sets on calibration and verification. Water Science and Technology 55, 1e17. Beven, K., Binley, A., 1992. Future of distributed models: model calibration and uncertainty prediction. Hydrological Processes 6, 279e298. Beven, K.J., Freer, J.E., 2001. Equifinality, data assimilation, and uncertainty estimation in mechanistic modelling of complex environmental systems using the GLUE methodology. Journal of Hydrology 249, 11e29. Beven, K.J., Smith, P., Freer, J.E., 2008. So just why should a modeller choose to be incoherent? Journal of Hydrology 354, 15e32. Beven, K., 2008. Environmental Modeling: An Uncertain Future? e An Introduction to Techniques for Uncertainty Estimation in Environmental Prediction. Routledge, London and New York. Calabro, P.S., 2001. Cosmoss: conceptual simplified model for sewer system simulation; a new model for urban runoff quality. Urban Water 3 (1e2), 33e42. Chon, H.-S., Ohandja, D.-G., Voulvoulis, N., 2010. Implementation of E.U. water framework directive: source assessment of metallic substances at catchment levels. Journal of Environmental Monitoring 12, 36e47. Christensen, A.M., Nakajima, F., Baun, A., 2006. Toxicity of water and sediment in a small urban river (Store Vejlea , Denmark). Environmental Pollution 144, 621e625. Daebel, H., Gujer, W., 2005. Uncertainty in predicting riverbed erosion caused by urban stormwater discharge. Water Science and Technology 52 (5), 77e85. Dotto, C.B.S., Kleidorfer, M., Deletic, A., Fletcher, T.D., McCarthy, D.T., Rauch, W., 2010. Stormwater quality models: performance and sensitivity analysis. Water Science and Technology 62 (4), 827e843. Dotto, C.B.S., Mannina, G., Kleidorfer, M., Vezzaro, L., Henrichs, M., McCarthy, D.T., Freni, G., Rauch, W., Deletic, A. 2011. Comparison of different uncertainty techniques in urban stormwater quantity and quality modelling. Water Research, Submitted. Eriksson, E., Baun, A., Scholes, L., Ledin, A., Ahlman, S., Revitt, M., Noutsopoulos, C., Mikkelsen, P.S., 2007. Selected stormwater priority pollutants - a European perspective. Science of the Total Environment 383, 41e51.
3834
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 2 3 e3 8 3 5
European Union, 2000. Directive 2000/60/EC of the European Parliament and of the Council of 23 October 2000 Establishing a Framework for Community Action in the Field of Water Policy 2000/60/EC ed. European Union, 2008. Directive 2008/105/EC of the European Parliament and of the Council of 16 December 2008 on environmental quality standards in the field of water policy, amending and subsequently repealing Council Directives 82/ 176/EEC, 83/513/EEC, 84/156/EEC, 84/491/EEC, 86/280/EEC and amending Directive 2000/60/EC of the European Parliament and of the Council. 2008/105/EC ed. Freer, J., Beven, K.J., Ambroise, B., 1996. Bayesian estimation of uncertainty in runoff prediction and the value of data: an application of the GLUE approach. Water Resources Research 32 (7), 2161e2173. Freni, G., Mannina, G., 2010. Bayesian approach for uncertainty quantification in water quality modelling: the influence of prior distribution. Journal of Hydrology 392, 31e39. Freni, G., Mannina, G., Gaspare, V., 2008. Uncertainty in urban stormwater quality modelling: the effect of acceptability threshold in the GLUE methodology. Water Research 42 (8e9), 2061e2072. Freni, G., Mannina, G., Viviani, G., 2009. Uncertainty in urban stormwater quality modelling: the influence of likelihood measure formulation in the GLUE methodology. The Science of the Total Environment 408, 138e145. Freni, G., Mannina, G., Viviani, G., 2011. Assessment of the integrated urban water quality model complexity through identifiability analysis. Water Research 45, 37e50. Garthwaite, P.H., O’Hagan, A., 2000. Quantifying expert opinion in the UK Water industry: an experimental study. The Statistician 49 (Part 4), 455e477. Gnecco, I., Berretta, C., Lanza, L.G., La Barbera, P., 2005. Storm water pollution in the urban environment of Genoa, Italy. Atmospheric Research 77 (1e4), 60e73. Go¨bel, P., Dierkes, C., Coldewey, W.C., 2007. Storm water runoff concentration matrix for urban areas. Journal of Contaminant Hydrology 91, 26e42. Gromaire-Mertz, M.C., Garnaud, S., Gonzalez, A., Chebbo, G., 1999. Characterisation of urban runoff pollution in Paris. Water Science and Technology 39 (2), 1e8. Hansen, A.C., Liu, L., Linde, J.J., Mark, O., Mikkelsen, P.S., 2005. Accounting for uncertainty in urban drainage system performance assessment using safety factors applied to runoff. In: Eriksson, E., et al. (Eds.), Proceedings of the 10th International Conference on Urban Drainage August 21-26 2005, Copenhagen, Denmark. Harremoe¨s, P., Petersen, C.M., Laustsen, A., Sørensen, S., Laden, B., Friis, C., Andersen, H.K., Linde, J.J., Mikkelsen, P.S., Jacobsen, C., 2005. Funktionspraksis for Afløbssystemer under Regn (Functional Requirements for Urban Drainage Systems during Rain. in Danish), Spildevandskomiteens skrift, 27. Ingeniørforeningen i Danmark IDA, Spildevandskomiteen, København (in Danish)1e66. Kanso, A., Chebbo, G., Tassin, B., 2005. Stormwater quality modelling in combined sewers: calibration and uncertainty analysis. Water Science and Technology 52 (3), 63e71. iulioniene, D.E., Randerson, P., Karlaviciene, V., Svediene, S., Marc Rimeika, M., Hogland, W., 2009. The impact of storm water runoff on a small urban stream. Journal of Soils and Sediments 9 (1), 6e12. Kayhanian, M., Stransky, C., Bay, S., Lau, S.-L., Stenstrom, M.K., 2008. Toxicity of urban highway runoff with respect to storm duration. Science of the Total Environment 389 (2e3), 386e406. Kleidorfer, M., Deletic, A., Fletcher, T.D., Rauch, W., 2009. Impact of input data uncertainties on urban stormwater model parameters. Water Science & Technology 60 (6), 1545e1554.
Kuczera, G., Parent, E., 1998. Monte Carlo assessment of parameter uncertainty in conceptual catchment models: the Metropolis algorithm. Journal of Hydrology 211 (1e4), 69e85. Lei, J.H., Schilling, W., 1996. Preliminary uncertainty analysis a prerequisite for assessing the predictive uncertainty of hydrologic models. Water Science and Technology 33 (2), 79e90. Lindblom, E., Ahlman, S., Mikkelsen, P.S., 2007a. How uncertain is model-based prediction of copper loads in stormwater runoff? Water Science and Technology 56 (11), 65e72. Lindblom, E., Madsen, H., Mikkelsen, P.S., 2007b. Comparative uncertainty analysis of copper loads in stormwater systems using GLUE and grey-box modeling. Water Science and Technology 56 (6), 11e18. Lu¨tzhøft, H.C.H., Eriksson, E., Donner, E., Wickman, T., Banovec, P., Mikkelsen, P.S., Ledin, A., 2009. Quantifying releases of priority pollutants from urban sources. In: Proceedings of WEFTEC 09, 82nd Annual Water Environment Federation Technical Exhibition and Conference, Orlando, Florida, USA, October 10e14, 2009. CD-ROM, Water Environment Federation, Alexandria, VA, pp. 5873e5891. Mailhot, A., Gaume, E., Villeneuve, J.-P., 1997. Uncertainty analysis of calibrated parameter values of an urban stormwater quality model using Metropolis Monte Carlo algorithm. Water Science and Technology 36 (5), 141e148. Mantovan, P., Todini, E., 2006. Hydrological forecasting uncertainty assessment: inherence in the GLUE methodology. Journal of Hydrology 330 (1e2), 368e381. McIntyre, N., Wheater, H., Lees, M., 2002. Estimation and propagation of parametric uncertainty in environmental models. Journal of Hydroinformatics 4 (3), 177e198. Metropolis, N., Rosenbluth, A.W., Rosenbluth, M.N., Teller, A.H., Teller, E., 1953. Equations of state calculations by fast computing mashines. Journal of Chemical Physics 21 (6), 1087e1092. Nash, J.E., Sutcliffe, J.V., 1970. River flow forecasting through conceptual models part I e A discussion of principles. Journal of Hydrology 10 (3), 282e290. Notter, M., 1993. The metals and the environment. Report 4135, Swedish EPA, Stockholm (in Swedish). Obropta, C.C., Kardos, J.S., 2007. Review of urban stormwater quality models: deterministic, stochastic, and hybrid approaches. Journal of the American Water Resources Association 43 (6), 1508e1523. Overton, D.E., Meadows, M.E., 1976. Stormwater Modeling. Academic Press, New York. Pettersson, T.J.R., Stro¨mvall, A.-M., Ahlman, S., 2005. Underground sedimentation systems for treatment of highway runoff in dense city areas. In: Eriksson, E., et al. (Eds.), Proceedings of the 10th International Conference on Urban Drainage August 21-26 2005, Copenhagen, Denmark. Refsgaard, J.C., van der Sluijs, J.P., Højbjerg, A.L., Vanrolleghrem, P.A., 2007. Uncertainty in the environmental modelling process e A framework and guidance. Environmental Modelling & Software 22, 1543e1556. Robert, P.R., Casella, G., 2004. Monte Carlo Statistical Methods, second ed. Springer, New York. Ruan, M., 1999. Continuous simulation of CSO emissions. Urban Water 1 (3), 201e205. Saltelli, A., Annoni, P., 2010. How to avoid a perfunctory sensitivity analysis. Environmental Modelling and Software 25 (12), 1508e1517. Stockholm Vatten, 1999. Pollution load to Lake Trekanten. Evalution of a Model for Estimation of Stormwater Pollution. Report 44/99, Stockholm Vatten, Stockholm (in Swedish). Tanner, M.A., 1996. Tools for Statistical Inference: Methods for the Exploration of Posterior Distributions and Likelihood Functions, third ed.. Springer-Verlag New Yourk Inc., USA. Thorndahl, S., Beven, K.J., Jensen, J.B., Schaarup-Jensen, K., 2008. Event based uncertainty assessment in urban drainage
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 2 3 e3 8 3 5
modelling, applying the GLUE methodology. Journal of Hydrology 357, 421e437. Thyer, M., Kuzcera, G., Wang, Q.J., 2002. Quantifying parameter uncertainty in stochastic models using the Box-Cox transformation. Journal of Hydrology 265, 246e257. Vezzaro, L., Eriksson, E., Ledin, A. and Mikkelsen, P.S. 2011. Modelling the fate of organic micropollutants in stormwater ponds. The Science of the Total Environment 409 (13), 2597e2606. Walker, W.E., Harremoe¨s, P., Rotmans, J., Van der Sluijs, J.P., Van Asselt, M.B.A., Janssen, P., Krayer von Krauss, M.P., 2003.
3835
Defining uncertainty e A conceptual basis for uncertainty management in model-based decision-support. Integrated Assessment 4 (1), 5e17. Willems, P., 2006. Random number generator or sewer water quality model? Water Science and Technology 54 (6e7), 387e394. Wong, T.H.F., Fletcher, T.D., Duncan, H.P., Coleman, J.R., Jenkins, G.A., 2002. A model for urban stormwater improvement conceptualisation. In: Proceedings of the Ninth International Conference On: Urban Drainage September 8-13 2002, Portland, OR, USA.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 3 6 e3 8 4 4
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Photo-dissolution of flocculent, detrital material in aquatic environments: Contributions to the dissolved organic matter pool Oliva Pisani a, Youhei Yamashita b, Rudolf Jaffe´ a,* a
Southeast Environmental Research Center, and Department of Chemistry & Biochemistry, Florida International University, 11200 SW 8th Street, Miami, FL 33199, USA b Faculty of Environmental Earth Science, Hokkaido University, Sapporo, Hokkaido 060-0810, Japan
article info
abstract
Article history:
This study shows that light exposure of flocculent material (floc) from the Florida Coastal
Received 3 February 2011
Everglades (FCE) results in significant dissolved organic matter (DOM) generation through
Received in revised form
photo-dissolution processes. Floc was collected at two sites along the Shark River Slough
6 April 2011
(SRS) and irradiated with artificial sunlight. The DOM generated was characterized using
Accepted 19 April 2011
elemental analysis and excitation emission matrix fluorescence coupled with parallel
Available online 28 April 2011
factor analysis. To investigate the seasonal variations of DOM photo-generation from floc, this experiment was performed in typical dry (April) and wet (October) seasons for the FCE.
Keywords:
Our results show that the dissolved organic carbon (DOC) for samples incubated under
Photo reactivity
dark conditions displayed a relatively small increase, suggesting that microbial processes
Everglades
and/or leaching might be minor processes in comparison to photo-dissolution for the
Dissolved organic matter
generation of DOM from floc. On the other hand, DOC increased substantially (as much as
Fluorescence
259 mgC gC1) for samples exposed to artificial sunlight, indicating the release of DOM
Detrital organic matter
through photo-induced alterations of floc. The fluorescence intensity of both humic-like and protein-like components also increased with light exposure. Terrestrial humic-like components were found to be the main contributors (up to 70%) to the chromophoric DOM (CDOM) pool, while protein-like components comprised a relatively small percentage (up to 16%) of the total CDOM. Simultaneously to the generation of DOC, both total dissolved nitrogen and soluble reactive phosphorus also increased substantially during the photo-incubation period. Thus, the photo-dissolution of floc can be an important source of DOM to the FCE environment, with the potential to influence nutrient dynamics in this system. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
Dissolved organic matter (DOM) comprises the largest pool of organic matter (OM) in a wide range of aquatic environments and plays a key role in the biogeochemical cycles affecting
processes such as metal complexation, pH buffering, light attenuation, nutrient availability, microbial and phytoplankton activity, and ecosystem productivity (Findlay and Sinsabaugh, 2003). The optical properties of chromophoric DOM (CDOM), the fraction of DOM that absorbs ultraviolet
* Corresponding author. Tel.: þ1 305 348 2456; fax: þ1 305 348 3096. E-mail address:
[email protected] (R. Jaffe´). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.04.035
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 3 6 e3 8 4 4
(UV) and visible light, have been extensively investigated in various aquatic ecosystems to determine the sources and transformations of this material in the environment (Coble, 1996, 2007). Photochemical effects on DOM dynamics have also been studied. CDOM containing numerous chromophoric moieties can undergo important photo-induced processes including photolysis of higher molecular weight to lower molecular weight compounds (Lou and Xie, 2006), generation of free radicals (Holder Sandvik et al., 2000), photo-mineralization reactions (Clark et al., 2004), and photo-bleaching (Shank et al., 2010). Photo-reactions have also been shown to help in the formation of biologically labile compounds, making the organic material more available for both autotrophic and heterotrophic biological activity (Moran and Zepp. 1997). More recently, the effects of light on the dissolution of particulate organic matter (POM) have been studied (Kieber et al., 2006; Mayer et al., 2006 , 2009a). It has been well established that POM can absorb light at similar wavelengths as DOM (Kirk, 1980; Kieber et al., 2006) allowing the particulate material to undergo similar photo-induced reactions. Such reactions can induce processes that break down larger molecules into smaller photo-products through the absorption of light (Miller and Moran, 1997). These reactions can influence the transition between the particulate and the dissolved phase of organic material (Mayer et al., 2006) and therefore the frequent exposure of particulates and sediments to light can ultimately lead to the transfer of particulate carbon to the dissolved phase. Kieber et al. (2006) irradiated sediments from the Cape Fear River estuary in North Carolina and found that on average, the dissolved organic carbon (DOC) photoproduction rate was 0.0056 mmol DOC g1 dry sediment h1, and suggested this value was larger than local riverine discharge and benthic flux sources of DOC to the ocean. Mayer et al. (2006) irradiated sediments from the Mississippi River and found that under optimal conditions two thirds of the exposed particulate organic carbon (POC) underwent photodissolution after several days. Shank et al. (2011) irradiated suspended sediments from Florida Bay and found that after 24 h of light exposure, the DOC concentration increased from 0.5 to 3.0 mgC L1. This potential generation of DOM through photo-induced mechanisms can play a significant role inorganic carbon and other biogeochemical cycles of aquatic environments, affecting both nutrient dynamics (Kieber et al., 2006; Zhang et al., 2009) and biological activity (Miller and Moran, 1997). In the Florida Coastal Everglades (FCE), the majority of the POM occurs at the sediment-water interface as flocculent detritus (floc, 0.02e1.4 mg L1). This material has been previously studied (Neto et al., 2006; Gao et al., 2007; Larsen et al., 2009; Troxler and Richards, 2009) and is known to be composed mainly of an assembly of periphyton, higher plant detritus and carbonates. With the application of molecular biomarkers, Neto et al. (2006) found that floc composition is primarily controlled by local vegetation inputs and early diagenetic transformations of OM. Using isotopic characterization, Troxler and Richards (2009) determined that detrital remains of Utricularia species comprise the primary components of floc materials found in deep sloughs of the FCE. Isoprenoid hydrocarbons known as botryococcenes, and believed
3837
to be produced by the microalga Botryococcus braunii or by filamentous green algae, have also been reported in floc from the FCE (Gao et al., 2007). However, little is still known about the biogeochemical dynamics of floc in this environment. Detritus is known to be a source of energy and nutrients to living organisms in many food webs (Moore et al., 2004). In the FCE, floc and periphyton mats have been proposed as primary energy sources driving local trophic dynamics (Williams and Trexler, 2006). For this reason alone, it is important to understand floc dynamics in the waters of this oligotrophic, subtropical wetland. In the shallow waters of the FCE, floc is naturally re-suspended through wind and bio-turbation (Larsen et al., 2009), allowing it to be exposed to intense sunlight (light penetration in FCE waters can reach 1745 mE cm2 s1; F. Tobias, personal communication). In the Everglades, floc is not entrained by water flow (entrainment threshold of 1.0 102 Pa; Larsen et al., 2009) because the flow is not sufficient for significant floc transport. However, some authors have suggested that floc is mobile enough to reach the estuarine areas of the FCE (Jaffe´ et al., 2001). With the implementation of the Comprehensive Everglades Restoration Plan (CERP) there will be an increase in water flow through the Shark River Slough (SRS) to the Gulf of Mexico (www. evergladesplan.org). This increase in water delivery can potentially increase floc transport from the freshwater marshes to the mangrove fringe and out to the Gulf, where the flocculent material will be exposed to intense sunlight. Light exposure can initiate a series of reactions and alterations in detrital OM (Kieber et al., 2006; Mayer et al., 2006, 2009a and 2009b), and therefore in floc, potentially affecting its environmental dynamics and ecosystem functions. Thus it is important to determine the photochemical reactivity of floc in the FCE and aquatic environments in general, in order to estimate the potential contribution of such processes to the DOM pool and its overall influence on the biogeochemistry of detrital rich ecosystems. The specific objectives of this study were to quantitatively assess the amount and quality of DOM that is photo-produced from floc of different composition/origin on both spatial and seasonal scales (i.e. freshwater marsh vs. mangrove fringe; wet season vs. dry season).
2.
Methods
2.1.
Site description
The Florida Coastal Everglades (FCE) is a subtropical wetland located on the southern tip of the Florida peninsula. The FCE extends west to the Gulf of Mexico and south to Florida Bay. This oligotrophic wetland is characterized by very low dissolved nutrient concentrations in the water column. There are two main drainage basins in the FCE; Shark River Slough (SRS) drains to the southwest coast of Everglades National Park (ENP) and into the Gulf of Mexico, while Taylor Slough (TS) drains to the southeast and into Florida Bay. Water discharge to the southwest coast of ENP through SRS has been shown to be substantially larger than discharge through TS (Woods, 2010). Floc samples were collected in SRS at sites that have been previously described by the on-going Florida Coastal
3838
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 3 6 e3 8 4 4
Everglades-Long Term Ecological Research program (FCELTER), namely at a freshwater marsh site (SRS2) and at an estuarine mangrove site (SRS6) (Fig. 1). The former is a long hydroperiod site characterized by peat soils where the dominant vegetation is Cladium jamaicense (sawgrass), Eleocharis cellulosa (gulfcoast spikerush) and calcareous periphyton, an assemblage of cyanobacteria, green algae, diatoms and higher plant detritus. The latter site, located on the coastal fringe, is dominated by Rhizophora mangle (red mangrove). This site may receive in addition to the dominant mangrove detritus, some marine OM inputs from seagrasses and phytoplankton (Hernandez et al., 2001) through tidal exchange. Basic water quality and floc parameters are summarized in Table 1.
2.2.
Table 1 e Natural water and floc bulk chemical parameters. Site Season
Natural water
Floc
Salinity DOC TDN SRP Density TOC TN (mgC (mgN (mgP (g mL1) (%) (%) L1) L1) L1) SRS2 Dry Wet SRS6 Dry Wet
0 0 34.4 17.1
32.8 22.3 6.7 7.9
1.33 0.23 0.31 0.25
0.008 n.a.a 0.018 n.a.
0.067 0.71 n.a.b n.a.
25.6 37.3 14.9 11.4
3.02 3.25 0.54 0.41
a n.a. ¼ not analyzed. b The density for SRS6 floc could not be measured due to low tide at the time of sample collection.
Sample collection
Floc samples were collected according to Neto et al. (2006). Briefly, floc samples were collected using a transparent plastic corer (inner diameter of 2.5 cm). The core was pushed about 10 cm below the sediment surface, capped to create suction, and retrieved. The floc layer was visible in the core and was decanted from the consolidated surface of the soil/sediment using a plunger with a smaller diameter to that of the core tube to hold the bottom layer in place. Excess water was decanted and the floc was collected in pre-rinsed 1 L Teflon jars (Nalgene). This procedure was repeated at randomly selected locations at each site enough times to obtain about 1 L of floc composite for each sampling event. Eight L of natural water were also collected at each site in Nalgene bottles sequentially pre-washed with 0.5 N HCl and 0.1 N NaOH. Water samples were kept on ice and upon return to the laboratory, they were filtered through pre-combusted (450 C for 4 h) 0.7 mm glass-fiber filters (GF/F) (Whatman International Ltd.) and 0.22 mm Durapore Membrane filters (Millipore) to remove POM from water samples. The filtrate
was passed through an activated carbon filter cartridge (Whatman) to remove much of the DOM from the natural water (%DOC removed was 46e64%; %absorbance at 254 nm removed was 57e93%). This step was needed to reduce the background DOC levels, and thus be able to better determine its photo-generation rates, as Everglades waters are commonly enriched in DOC (Table 1).
2.3.
Experimental setup
Floc samples were mixed with natural water (after DOM removal) to give solutions with a final floc concentration of about 24 g floc L1 (dry weight). Such high initial concentrations were used to simulate the floc layer in the natural environment which can reach concentrations of up to 710 g floc L1 (unpublished data). These solutions were prepared in pre-combusted glass jars (in triplicate), covered with quartz plates for light exposure, or wrapped in black plastic bags for dark controls. Light and dark controls were performed in the solar simulator’s water circulating bath (26 C), to maintain
Fig. 1 e Florida coastal Everglades map showing sampling site locations along the Shark River Slough (SRS).
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 3 6 e3 8 4 4
similar temperature conditions for all experiments. Flat top glass jars were used instead of beakers to obtain a better seal with the quartz cover plates in order to avoid sample contamination by dust particles. True dark blanks were not performed as poisoning with either mercuric chloride or sodium azide would result in fluorescence quenching. Samples were incubated for different periods of time (0, 0.5, 1, 2, 4, and 7 days) in a solar simulator (Suntest XLSþ, Atlas Material Testing Technology LLC) set at 765 W m2. These conditions correspond to about 1.2 times solar noon in South Florida (Maie et al., 2008). After photo-exposure, samples were filtered (0.7 mm GF/F) to separate the aqueous phase for DOM analysis, and to recover the detrital fraction. The filtered particulates were dried overnight in a 60 C oven and the recovered floc was ground and saved for elemental analysis. The filtrate was analyzed for DOC, total dissolved nitrogen (TDN) and soluble reactive phosphorus (SRP), and the optical properties were examined using UVevis spectroscopy and excitation emission matrix (EEM) fluorescence spectroscopy.
2.4.
Elemental analysis
About 8e10 mg of floc sample was weighed in silver cups and de-carbonated by exposure to hydrochloric acid vapors overnight (Harris et al., 2001). Samples were dried in a 60 C oven overnight and analyzed for total organic C (%TOC) and total N (%TN) concentrations. Triplicate samples were measured using a Carlo Erba NA 1500 nitrogen/carbon analyzer with a reproducibility of 1.07% for TOC and 0.09% for TN on average. DOC concentration was measured with a Shimadzu TOC-V total organic carbon (TOC) analyzer. Prior to analysis, the samples were acidified (pH < 2) and purged with CO2-free air for 5 min to remove inorganic C. Total dissolved nitrogen (TDN) was measured on an ANTEK 9000 nitrogen analyzer.
2.5.
UVevis and fluorescence spectroscopy
UVevisible absorption spectra were obtained using a Varian Cary-50 Bio spectrophotometer at wavelengths between 250 and 800 nm. Samples were measured in a 1 cm quartz cuvette using Milli-Q water as the blank. EEM Fluorescence was measured on a Horiba Jobin-Yvon Fluoromax-3 spectrofluorometer equipped with a 150-W Xenon arc lamp according to Chen et al. (2010) and Yamashita et al. (2010). Briefly, scans were acquired in a 1 cm quartz cuvette at excitation wavelengths (lex) between 260 and 455 nm at 5 nm intervals. Emission wavelengths (lem) were scanned from lex þ 10 nm to lex þ 250 nm at 2 nm intervals. The individual spectra were concatenated to form a threedimensional matrix. All spectra were acquired in S/R mode and were corrected for inner filter effects and instrument bias. Finally, fluorescence intensity values were converted to quinine sulfate units (QSU) to facilitate inter-laboratory comparisons.
2.6.
Parallel factor analysis (PARAFAC)
Parallel factor analysis (PARAFAC) is a three-way multivariate statistical method that has been used to decompose EEMs of
3839
complex mixtures into their individual fluorescent components (Stedmon et al., 2003). The EEMs of 75 incubated floc and natural water samples were fitted to an existing PARAFAC model created with ca 1400 surface water samples collected from the Everglades and Florida Bay (Chen et al., 2010). PARAFAC analysis was performed using MATLAB 7.0.4 (Mathworks, Natick, MA) with the DOMFluor toolbox (Stedmon and Bro, 2008). Obvious residual peaks were not found after fitting our samples to this eight component model, indicating that the fluorophores produced from the irradiation of floc are similar to those of surface waters from the Everglades. The spectral characteristics of the eight components are summarized below.
3.
Results & discussion
3.1.
Natural water & floc chemical characteristics
Spatial differences in the initial DOC concentration for the two water samples are summarized in Table 1. As expected, the higher DOC values were obtained for the freshwater site (SRS2) compared to the mangrove site (SRS6), where a contribution of DOM to the former derive from the abundant macrophytes, periphyton mats and organic rich soils (peat) (Yamashita et al., 2010), while the latter is mostly influenced by mangrove derived sources and diluted by tidal mixing (Jaffe´ et al., 2003). Seasonal differences were also observed; water collected at the freshwater site was found to have 32.8 mgC L1 in the dry season and 22.3 gC L1 in the wet season. The smaller DOC concentration obtained in the wet season could be indicative of a dilution effect due to an increase in rainfall. The DOC content of natural water collected at the mangrove site was found to be seasonally similar, at 6.7 mgC L1 in the dry season and 7.9 mgC L1 in the wet season. TDN was found to be higher in the natural water at the freshwater site during the dry season, indicative of a concentration effect. In addition, the abundant periphyton mats found at SRS2 contain numerous N-fixing cyanobacteria which may be contributing to the local TDN pool. SRP was higher at the mangrove site which receives phosphorus inputs from the adjacent Gulf of Mexico, while the SRS2 site is a typically Plimited FCE freshwater marsh site (Childers et al., 2006). The floc collected in the freshwater marsh had higher %TOC and %TN compared to the mangrove floc, probably due to increased accumulation of OM at the former long hydroperiod site. The mangrove site is strongly influenced by tidal activity and the floc found there may not have the opportunity for significant accumulation. In fact, the sediment accretion rate at this particular site has been estimated to be 0.30 0.03 cm year1 (Castan˜eda-Moya et al., 2009) while accretion rates in the SRS2 vicinity have been estimated at 0.50 cm year1 (Saunders et al., 2006). Floc collected at SRS2 during the wet season had higher %TOC and %TN than the floc collected in the dry season, indicative of higher inputs from increased local biomass productivity. The floc at SRS6 had a higher %TOC and %TN in the dry season, probably due to a decreased dilution effect, and higher nitrogen immobilization by bacteria associated with leaf litter decomposition (Twilley et al., 1986).
3840
3.2.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 3 6 e3 8 4 4
Photochemical production of DOC from floc
Photo-exposure of floc collected at both the freshwater and the mangrove sites caused the generation of significant amounts of DOC (Fig. 2). Normalized to the initial POC content, the floc collected at the freshwater site (SRS2) photo-generated up to 259 mgC gC1 while SRS6 floc produced up to 173 mgC gC1 with exposure to sunlight (corrected for dark treatments). This is in agreement with recent studies on the generation of DOM from re-suspended sediments in shallow aquatic environments (Shank et al., 2011). The authors found that upon solar irradiation, the sediments with the highest %OC content, exhibited the largest increases in DOC and terrestrial humic components. It should be noted that DOC in surface water samples may photo-degrade during photo-irradiation, and thus, values of photo-produced DOC reported here would be underestimated. However, it is important to mention that floc has been reported to contain some live benthic periphyton, including cyanobacteria (Neto et al., 2006). These organisms upon light exposure could generate DOM through enhanced primary productivity. However, it has been reported that microbial activity in solutions exposed to intense sunlight
Fig. 2 e Photo-dissolution induced changes in DOC concentrations normalized to initial floc POC content for floc at SRS2 (a) and SRS6 (b). Photo-produced DOC at day [ t was calculated by subtracting the DOC values at t [ 0: photo-produced DOC (t) [ DOC (t) e DOC (0). Error bars are for triplicate experiments. Open and filled symbols correspond to light and dark treatments, respectively. (C: dry and -: wet season for SRS2; A: dry and :: wet season for SRS6).
(such as in the solar simulator) is significantly inhibited (Xie et al., 2009) and therefore unlikely to make significant contributions to the DOC pool. While the overall trend is one of increasing DOC with exposure time, some variations were observed after several days of light exposure. This was particularly the case for the data from the dry season floc from both locations. While the DOC generation curve for floc from the wet season was relatively constant with time for both freshwater and mangrove floc, the data for floc from the dry season showed a fast increment in DOC generation during the first two and four days for the freshwater and mangrove samples respectively, followed by an overall decrease. These variations in DOC concentration with incubation time could be due to several mechanisms including re-adsorption onto particles, flocculation (von Wachenfeldt et al., 2009) and/or photo-mineralization of DOC to yield dissolved inorganic carbon (DIC) (Clark et al., 2004), and seem more pronounced for the dry season samples (see Fig. 2). Regardless of this trend, the difference in DOC production between the two sites suggests enrichment in photo-labile material at the freshwater site (SRS2) compared to the mangrove site (SRS6). The former is dominated by marsh vegetation (sawgrass and spikerush) and abundant periphyton mats which seem to control the main sources of OM to the floc layer (Neto et al., 2006). The organic rich, peat soils at SRS2 may also contribute OM to the floc layer. As such, floc at SRS2 is expected to be lower in lignin phenol content compared to that at SRS6 where mangrove derived detrital OM in the form of decaying leaf and root materials are likely the main OM sources to the floc (Neto et al., 2006). Consequently, the floc at SRS6 is expected to feature more biologically recalcitrant organic matter. However, lignin phenol is photodegradable (Opsahl and Benner, 1998), and sunlight intensity is not considerably different throughout the year in South Florida. Thus, considering that its lignin phenol content is larger at SRS6 floc the lower reactivity to photoexposure is somewhat unexpected. Samples that were incubated under dark conditions also produced measurable amounts of DOC (126 mgC gC1 for SRS2 floc and 34 mgC gC1 for SRS6 floc) but significantly less compared to the photo-exposed samples. While leachates from some common Everglades biomass, such as sawgrass and spikerush blades, periphyton and mangrove leaves, have been reported to be important contributors to the DOC pool, leaching between 8 and 51 mgC g1 of dry biomass during the early stages of decomposition (Maie et al., 2006), the floc from the freshwater site leached up to 79 mgC g dry floc. Such experimental results indicate that leaching from floc may be a more important source of this dissolved material than previously believed, although the photo-induced generation of DOC clearly dominates. Exposure of flocculent material to artificial sunlight also caused the production of dissolved nutrients at both sites, showing photo-generation of TDN (5.2 and 0.98 mgN g1 floc for SRS2 and SRS6, respectively) and SRP concentrations (0.07 and 0.19 mgP g1 floc for SRS2 and SRS6, respectively). Because these parameters were only measured for floc collected in the dry season, seasonal effects will not be addressed. However, the photo-generation of DOM-associated N and P can greatly affect food web dynamics and biogeochemical cycles, especially in the oligotrophic waters of the FCE where most of the
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 3 6 e3 8 4 4
dissolved nutrients are found in the organic form (Noe and Childers, 2007). Seasonal differences (wet vs. dry season) in DOC photoproduction from floc were also observed (Fig. 2). Throughout the length of the incubation period, site SRS2 floc collected in the dry season produced 97 mgC gC1 more than floc collected at the same site in the wet season. Photo-exposure of floc collected at SRS6 during the dry season also produced more DOC during the first 4 days of incubation, but fell below the levels of photo-produced DOC from the wet season floc after 7 days of exposure. Similarly, SRS2 floc from the dry season was significantly more photo-productive of DOC during the first 2 days of exposure (see above). The higher (initial) photoproduction rates of DOC for the dry season samples may be due to the presence of more degraded, aged OM in the floc layer during this period. An increase in mangrove litterfall during the wet season has been observed at SRS6 (Twilley et al., 1986), contributing fresher inputs of OM to the floc layer. Similarly, the abundant periphyton mats found at SRS2 have been shown to display an increase in primary productivity at the onset of the wet season (Ewe et al., 2006), contributing significant amounts of more labile, fresh OM to the floc layer. The older, more degraded floc present during the dry season however, seems to be more photo-reactive. This is in agreement with Mayer et al. (2009a) who showed that photo-dissolution is greatly enhanced by microbial decay, suggesting that older, more humified OM is more photo-labile. Therefore, seasonal primary productivity variations may result in changes in the floc OM quality and consequently its photo-reactivity. While Mayer et al. (2006) reported that light exposure of freshwater suspended particulates could result in a loss of 64% of the POC over a 15 d period of 6 h d1 irradiation, in the present study, the POC content did not change significantly during incubation of both the light and the dark treatments. This is likely due to an analytical artifact, since very high initial concentrations of POC (up to 3 gC L1) were used to simulate the natural floc layer conditions. As a result, the POC carbon loss through DOC photo-dissolution was a very small fraction of the total and consequently within the analytical error of the POC analysis. Thus, POC loss data and potential correlations with DOC production are not presented here. However, and in agreement with the literature (Kieber et al., 2006; Mayer et al., 2006 and 2009a) floc exposed to light generated a significant amount of DOC.
3.3.
Composition of photo-produced DOM
Fluorescence properties of natural waters have been used for determining the sources of DOM as well as its transformations in different aquatic environments and have been extensively applied for the quantification of fluorescent DOM (FDOM) in natural waters (Coble, 1996). EEM fluorescence can provide detailed information on the types of fluorescent compounds present in complex mixtures such as DOM (Coble, 1996). This fluorescence technique has been coupled with parallel factor analysis (PARAFAC), a statistical modeling approach, to decompose the EEMs into individual fluorescent components (Stedmon et al., 2003). Applying this approach, a total of eight fluorescent components had previously been obtained through
3841
PARAFAC modeling for the Everglades ecosystem (Chen et al., 2010; Yamashita et al., 2010). The fluorescence characteristics of these components were assigned to be characteristic for terrestrial humic-like (C1, 3 and 5), microbial humic-like (C4), protein-like (C7 and 8) and two unknown components (C2 and 6) which have recently been suggested to represent a humic-like component derived from soil oxidation and a ubiquitous humic-like component, respectively (Yamashita et al., 2010). In this study, the fluorescence intensity of the three humiclike and the two protein-like components were combined into two groups for simplicity reasons. The fluorescence intensity of the three humic-like components, C1 (lex ¼ 260 (345) nm, lem ¼ 462 nm), C3 (lex ¼ 260 (305) nm, lem ¼ 416 nm) and C5 (lex < 275 (405) nm, lem > 500 nm), increased for floc samples irradiated with artificial sunlight, suggesting that these components are photo-generated. These three components comprised a large portion (46e70% after 7 d of light exposure) of the total fluorescence, suggesting that the majority of the CDOM produced from irradiation of floc has humic-like optical characteristics. Two protein-like components were identified, a tyrosine-like component (C7; lex ¼ 275 nm, lem ¼ 326 nm) and a tryptophan-like component (C8; lex ¼ 300 nm, lem ¼ 342 nm) which also increased during photo-incubation. However, unlike the terrestrial humic-like components, these proteinlike components comprised a smaller portion (10e16% after 7 d of light exposure) of the CDOM produced during photoincubation of floc. The photo-generation of these protein-like components is in agreement with previous findings that tannin compounds leached from abscised mangrove leaves and other types of vegetation can form insoluble complexes with proteins, which upon photo-exposure have been shown to break up and re-release the N-containing compounds (Maie et al., 2008). Fluorescence intensity of protein-like components in DOM has also been reported to be strongly structure dependent (Mayer et al., 1999), and thus, could in part explain an increment in fluorescence intensity after photo-exposure. However, detailed EEM-PARAFAC based photo-degradation studies of Everglades DOM have not shown such effects, but instead show a decrease in intensity of protein-like fluorescence with increasing light exposure (Chen and Jaffe´, unpublished). Thus, the increase in protein-like fluorescence observed in this study is most likely the result of photodissolution of floc. The increase in TDN during these experiments seems to agree with this suggestion. However, overall, the fluorescence signature was dominated by photo-generated humic-like compounds. To look at the generation rates of the different fluorescent components we plotted the sum of the fluorescence intensity of the terrestrial humic-like components (C1, 3 and 5) and the protein-like components (C7 and 8), normalized to POC content, versus incubation time (Fig. 3). Differences in generation rates between samples, PARAFAC components and season are evidenced by significant differences in the slope of the linear correlations shown in Fig. 3 (see Table 2). When exposed to artificial sunlight, the floc collected at SRS2 produced more terrestrial humic-like material compared to the floc collected at SRS6 on a per-g POC basis. Shank et al. (2011) characterized the fluorophores generated from photoirradiation of Florida Bay suspended sediments, and found that the most organic rich sediments exhibited the largest
3842
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 3 6 e3 8 4 4
increases in terrestrial humic-like components. Similarly, photo-production of the protein-like components, where the presence of labile floc components from periphyton may be an important source of dissolved nitrogen, was higher in freshwater than in mangrove floc exposure experiments. Seasonal differences were similar to those previously described for DOC (see above), where higher initial (2e4 days) generation of humic- and protein-like components in the dry season was observed (Fig. 3), suggesting that aged floc is more photoreactive. Because the maximum photo-production of CDOM differed for the floc collected at the two sites, the slopes of the best-fit line for the linear portion of the experiment were compared (2 days for SRS2 and 4 days for SRS6). The generation of the humic-like and protein-like components was significantly different between sites and between seasons (Table 2). Humic-like components were generated at a much faster rate than the protein-like components, and during the wet season, these components were generated at a lower rate than during the dry season. This seasonal difference could be explained by the fact that unprocessed, fresher material incorporated into the floc layer during the wet season, is less photo-reactive, while older, more degraded material found in floc during the dry season is more reactive to sunlight. This is
Table 2 e Regression analysis for CDOM generation rates. Linear regressions were determined between incubation time and fluorescence intensities of terrestrial humic-like and protein-like components for 0e2 and 0e4 days for SRS2 and SRS6, respectively. Components
Site
Season
Humic-like
SRS2
Dry Wet Dry Wet Dry Wet Dry Wet
SRS6 Protein-like
SRS6
p
0.065 0.018 0.034 0.199 0.097 <0.001 0.027 0.002
3.30 1.35 1.00 0.28 1.03 0.19 0.14 0.09
0.88 0.18 0.27 0.17 0.35 0.00 0.04 0.01
in agreement with recent studies by Mayer et al. (2009a) which showed that the photochemical reactivity of algal detritus increases with increasing microbial decay and/or humification of OM. This data suggests that potentially both OM sources and degree of degradation (age) control the resulting composition of photo-generated DOM. The exact mechanism for these processes is presently not known.
4.
Fig. 3 e Photo-production of terrestrial humic-like components, C1, C3 and C5 were combined (B: Dry season, ,: Wet season) and protein-like components, C7 and C8 (>: Dry season, Δ: Wet season) from SRS2 (3a) and SRS6 (3b) floc. Fluorescence intensities were normalized to the initial floc POC content. Photo-produced fluorescent components at day [ t was calculated by subtracting the fluorescence intensity at t [ 0: photo-produced fluorescent intensity (t) [ fluorescence intensity (t) e fluorescence intensity (0).
SRS2
Generation rate (QSU gC1 d1)
Conclusions
In summary, the data presented above show that flocculent detritus in the FCE generates significant amounts of DOM as well as TDN and SRP when exposed to artificial sunlight. In the shallow waters of the FCE, floc is naturally re-suspended (Larsen et al., 2009), and can easily be exposed to intense sunlight. This is particularly critical for floc from freshwater marshes where the dominant vegetation is composed of short grasses and sedges, with minimal tree cover and consequently low shading effects. The resulting light exposure of the floc can aid in the transfer of POM into the dissolved phase through photo-dissolution processes (Kieber et al., 2006; Mayer et al., 2006) and as such fuel the microbial loop. This is especially important in the oligotrophic waters of the FCE where the concentrations of dissolved nutrients are naturally very low, but where most of the dissolved N and P are in an organic form (Noe and Childers, 2007). Regarding the composition of the photo-generated DOM, terrestrial humic-like components were the main contributors to the CDOM fluorescence, indicating a preferential photo-dissolution of humic moieties. Shank et al. (2011) reported that terrestrial humic-like components were preferentially photo-desorbed from re-suspended estuarine sediments, indicating that photo-generated materials seem to be preferentially dominated by organics with a more terrigenous character. The generation rate of the terrestrial humicand protein-like components was higher for floc collected at the freshwater site compared to the mangrove site, suggesting that there are differences in floc composition between the freshwater and mangrove sites that are reflected in differences in their photo-reactivity. Similarly, the generation rate of the terrestrial humic- and protein-like components was higher during dry season than wet season for both sites. It has
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 3 6 e3 8 4 4
recently been reported that older, partially degraded material can be significantly more photo-reactive compared to unprocessed, fresh material (Mayer et al. 2009a). Floc receives much of its OM input during the wet season when periphyton mats are more productive and mangrove litterfall increases. Consequently, floc present during the wet season is expected to be fresher, while it is more aged during the dry season, therefore increasing its potential photo-reactivity during the latter. While floc collected during the dry season clearly showed higher photo-dissolution rates during the first 2e4 days of exposure for freshwater and mangrove floc respectively, the overall DOC production after one week of exposure was not too different between wet and dry season samples. The Florida Coastal Everglades is an oligotrophic subtropical wetland, where detrital carbon pools are critical components of the food web and control to a significant extent the trophic dynamics in this system (Williams and Trexler, 2006). This study suggests that floc photo-dissolution has the potential to generate high amounts of DOC as well as TDN and SRP and thus can affect the biogeochemical cycling and productivity of this system. The efficiency of these photodissolution processes is dependent on floc quality, which seems dependent on biomass type inputs and primary productivity on both spatial and temporal scales. Potential changes, such as increased water delivery, particularly through Shark River Slough as a result of the implementation of the Comprehensive Everglades Restoration Plan may induce changes in floc dynamics in this system. A better understanding of the effects of light exposure to POM, suspended sediments or floc is needed to assess carbon dynamics in shallow and/or turbid aquatic ecosystems.
Acknowledgements We thank the Wetlands Ecosystems Lab at Florida International University for help with sample collection and the Southeast Environmental Research Center for elemental analysis. The authors also thank three anonymous reviewers for valuable comments and suggestions that helped improve the quality of this manuscript. This material is based upon work supported by the National Science Foundation through the Florida Coastal Everglades Long-Term Ecological Research program under Cooperative Agreements#DBI-0620409and #DEB-9910514. Any opinions, findings, conclusions, or recommendations expressed in the material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation. O.P thanks the FIU Graduate School for a Dissertation Year Fellowship. This is SERC contribution No. 522.
references
Castan˜eda-Moya, E., Twilley, R.R., Rivera-Monroy, V.H., Zhang, K., Davis III, S.E., Ross, M., 2009. Sediment and nutrient deposition associated with hurricane Wilma in mangroves of the Florida Coastal Everglades. Estuaries and Coasts 33 (1), 45e58.
3843
Chen, M., Price, R.M., Yamashita, Y., Jaffe´, R., 2010. Comparative study of dissolved organic matter from groundwater and surface water in the Florida coastal Everglades using multidimensional spectrofluorometry combined with multivariate statistics. Applied Geochemistry 25, 872e880. Childers, D.L., Boyer, J.N., Davis, S.E., Madden, C.J., Rudnick, D.T., Sklar, F.H., 2006. Relating precipitation and water management to nutrient concentrations in the oligotrophic “upside-down” estuaries of the Florida Everglades. Limnology and Oceanography 51 (1/2), 602e616. Clark, C.D., Hiscock, W.T., Millero, F.J., Hitchcock, G., Brand, L., Miller, W.L., Ziolkowski, L., Chen, R.F., Zika, R.G., 2004. CDOM distribution and CO2 production on the southwest Florida Shelf. Marine Chemistry 89, 145e167. Coble, P.G., 1996. Characterization of marine and terrestrial DOM in seawater using excitation-emission matrix spectroscopy. Marine Chemistry 51, 325e346. Coble, P.G., 2007. Marine optical biogeochemistry: the chemistry of ocean color. Chemical Reviews 107, 402e418. Ewe, S.M.L., Gaiser, E.E., Childers, D.L., Iwaniec, D., RiveraMonroy, V.H., Twilley, R.R., 2006. Spatial and temporal patterns of aboveground net primary productivity (ANPP) along two freshwater-estuarine transects in the Florida Coastal Everglades. Hydrobiologia 569, 459e474. Findlay, S.E.G., Sinsabaugh, R.L., 2003. Aquatic Ecosystems: Interactivity of Dissolved Organic Matter. Academic Press, San Diego. Gao, M., Simoneit, B.R.T., Gantar, M., Jaffe´, R., 2007. Occurrence and distribution of novel botryococcene hydrocarbons in freshwater wetlands of the Florida Everglades. Chemosphere 70, 224e236. Harris, D., Horwath, W.R., van Kessel, C., 2001. Acid fumigation of soils to remove carbonates prior to total organic carbon or carbon-13 isotopic analysis. Soil Science Society of America Journal 65, 1853e1856. Hernandez, M.E., Mead, R., Peralba, M.C., Jaffe´, R., 2001. Origin and transport of n-alkane-2-ones in a subtropical estuary: potential biomarkers for seagrass-derived organic matter. Organic Geochemistry 32, 21e32. Holder Sandvik, S.L., Bilski, P., Pakulski, J.D., Chignell, C.F., Coffin, R. B., 2000. Photogeneration of singlet oxygen and free radicals in dissolved organic matter isolated from the Mississippi and Atchafalaya River plumes. Marine Chemistry 69, 139e152. Jaffe´, R.J., Mead, R., Hernandez, M.E., Peralba, M.C., DiGuida, O.A., 2001. Origin and transport of sedimentary organic matter in two subtropical estuaries: a comparative, biomarker-based study. Organic Geochemistry 32, 507e526. Jaffe´, R., Boyer, J.N., Lu, X., Maie, N., Yang, C., Scully, N.M., Mock, S., 2003. Source characterization of dissolved organic matter in a subtropical mangrove-dominated estuary by fluorescence analysis. Marine Chemistry 84, 195e210. Kieber, R.J., Whitehead, R.F., Skrabal, S.A., 2006. Photochemical production of dissolved organic carbon from resuspended sediments. Limnology and Oceanography 51 (5), 2187e2195. Kirk, J.T.O., 1980. Spectral absorption properties of natural waters: contribution of the soluble and particulate fractions to light absorption some inland waters of south-eastern Australia. Australian Journal of Marine and Freshwater Research 31, 287e296. Larsen, L.G., Harvey, J.W., Noe, G.B., Crimaldi, J.P., 2009. Predicting organic floc transport dynamics in shallow aquatic ecosystems: Insights from the field, the laboratory, and numerical modeling. Water Resources Research 45, W01411. doi:10.1029/2008WR007221. Lou, T., Xie, H., 2006. Photochemical alteration of the molecular weight of dissolved organic matter. Chemosphere 65, 2333e2342.
3844
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 3 6 e3 8 4 4
Maie, N., Jaffe´, R., Miyoshi, T., Childers, D.L., 2006. Quantitative and qualitative aspects of dissolved organic carbon leached from senescent plants in an oligotrophic wetland. Biogeochemistry 78, 285e314. Maie, N., Pisani, O., Jaffe´, R., 2008. Mangrove tannins in aquatic ecosystems: their fate and possible influence on dissolved organic carbon and nitrogen cycling. Limnology and Oceanography 53 (1), 160e171. Mayer, L.M., Schick, K., Loder III, T.C., 1999. Dissolved protein fluorescence in two Maine estuaries. Marine Chemistry 64, 171e179. Mayer, L.M., Schick, L.L., Skorko, K., Boss, E., 2006. Photodissolution of particulate organic matter from sediments. Limnology and Oceanography 51 (2), 1064e1071. Mayer, L.M., Schick, L.L., Hardy, K.R., Estapa, M.L., 2009a. Photodissolution and other photochemical changes upon irradiation of algal detritus. Limnology and Oceanography 54 (5), 1688e1698. Mayer, L.M., Schick, L.L., Bianchi, T.S., Wysocki, L.A., 2009b. Photochemical changes in chemical markers of sedimentary organic matter source and age. Marine Chemistry 113, 123e128. Miller, W.L., Moran, M.A., 1997. Interaction of photochemical and microbial processes in the degradation of refractory dissolved organic matter from a coastal marine environment. Limnology and Oceanography 42 (6), 1317e1324. Moore, J.C., Berlow, E.L., Coleman, D.C., de Ruiter, P.C., Dong, Q., Hastings, A., Johnson, N.C., McCann, K.S., Melville, K., Morin, P.J., Nadelhoffer, K., Rosemond, A.D., Post, D.M., Sabo, J. L., Scow, K.M., Vanni, M.J., Wall, D.H., 2004. Detritus, trophic dynamics and biodiversity. Ecology Letters 7, 584e600. Moran, M.A., Zepp, R.G., 1997. Role of photoreactions in the formation of biologically labile compounds from dissolved organic matter. Limnology and Oceanography 42 (6), 1307e1316. Neto, R.R., Mead, R.N., Louda, W.J., Jaffe´, R., 2006. Organic biogeochemistry of detrital flocculent material (floc) in a subtropical, coastal wetland. Biogeochemistry 77, 283e304. Noe, G.B., Childers, D.L., 2007. Phosphorus budgets in Everglades wetland ecosystems: the effects of hydrology and nutrient enrichment. Wetlands Ecology and Management 15, 189e205. Opsahl, S., Benner, R., 1998. Photochemical reactivity of dissolved lignin in river and ocean waters. Limnology and Oceanography 43, 1297e1304. Saunders, C.J., Gao, M., Lynch, J.A., Jaffe´, R., Childers, D.L., 2006. Using soil profiles of seeds and molecular markers as proxies for sawgrass and wet prairie slough vegetation in Shark River Slough, Everglades National Park. Hydrobiologia 569, 475e492.
Shank, G.C., Zepp, R.G., Va¨ha¨talo, A., Lee, R., Bartels, E., 2010. Photobleaching kinetics of chromophoric dissolved organic matter derived from mangrove leaf litter and floating Sargassum colonies. Marine Chemistry 119, 162e171. Shank, G.C., A. Evans, Y. Yamashita and R. Jaffe´. Solar radiationenhanced dissolution of particulate organic matter from coastal marine sediments. Limnology and Oceanography in press. Stedmon, C.A., Bro, R., 2008. Characterizing dissolved organic matter fluorescence with parallel factor analysis: a tutorial. Limnology and Oceanography Methods 6, 572e579. Stedmon, C.A., Markager, S., Bro, R., 2003. Tracing dissolved organic matter in aquatic environments using a new approach to fluorescence spectroscopy. Marine Chemistry 82, 239e254. Troxler, T.G., Richards, J., 2009. d13C, d15N, carbon, nitrogen and phosphorus as indicators of plant ecophysiology and organic matter pathways in Everglades deep slough, Florida, USA. Aquatic Botany 91 (3), 157e165. Twilley, R.R., Lugo, A.E., Patterson-Zucca, C., 1986. Litter production and turnover in basin mangrove forests in southwest Florida. Ecology 67 (3), 670e683. von Wachenfeldt, E., Bastviken, D., Tranvik, L.J., 2009. Microbially induced flocculation of allochthonous dissolved organic carbon in lakes. Limnology and Oceanography 54 (5), 1811e1818. Williams, A.J., Trexler, J.C., 2006. A preliminary analysis of the correlation of food-web characteristics with hydrology and nutrient gradients in the southern Everglades. Hydrobiologia 569, 493e504. Woods, J. 2010. Surface Water Discharge and Salinity Monitoring of Coastal Estuaries in Everglades National Park, USA, in support of the Comprehensive Everglades Restoration Plan. In: Proceedings of the 3rd International Perspective on Current & Future State of Water Resources & the Environment, Chennai, India. Xie, H., Zafiriou, O.C., 2009. Evidence for significant photochemical production of carbon monoxide by particles in coastal and oligotrophic marine waters. Geophysical Research Letters 36, L23606. doi:10.1029/2009GL041158. Yamashita, Y., Scinto, L.J., Maie, N., Jaffe´, R., 2010. Dissolved organic matter characteristics across a subtropical wetland’s landscape: application of optical properties in the assessment of environmental dynamics. Ecosystems 13, 1006e1019. Zhang, Y., van Dijk, M.A., Liu, M., Zhu, G., Qin, B., 2009. The contribution of phytoplankton degradation to chromophoric dissolved organic matter (CDOM) in eutrophic shallow lakes: field and experimental evidence. Water Research 43, 4685e4697.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 4 5 e3 8 5 4
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Process optimization by decoupled control of key microbial populations: Distribution of activity and abundance of polyphosphate-accumulating organisms and nitrifying populations in a full-scale IFAS-EBPR plant Annalisa Onnis-Hayden a, Nehreen Majed a, Andreas Schramm b, April Z. Gu a,* a
Civil and Environmental Engineering Department, Northeastern University, 400 Snell Engineering Center, 360 Huntington Ave, Boston, MA 02115, USA b Department of Biological Sciences, Microbiology, Aarhus University, Denmark
article info
abstract
Article history:
This study investigated the abundance and distribution of key functional microbial pop-
Received 26 October 2010
ulations and their activities in a full-scale integrated fixed film activated sludgeeenhanced
Received in revised form
biological phosphorus removal (IFAS-EBPR) process. Polyphosphate accumulating organisms
10 February 2011
(PAOs) including Accumulibacter and EBPR activities were predominately associated with the
Accepted 21 April 2011
mixed liquor (>90%) whereas nitrifying populations and nitrification activity resided mostly
Available online 4 May 2011
(>70%) on the carrier media. Ammonia oxidizer bacteria (AOB) were members of the Nitrosomonas europaea/eutropha/halophila and the Nitrosomonas oligotropha lineages, while nitrite
Keywords:
oxidizer bacteria (NOB) belonged to the Nitrospira genus. Addition of the carrier media in the
BNR
hybrid activated sludge system increased the nitrification capacity and stability; this effect
IFAS
was much greater in the first IFAS stage than in the second one where the residual ammonia
EBPR
concentration becomes limiting. Our results show that IFAS-EBPR systems enable decoupling
PAOs
of solid residence time (SRT) control for nitrifiers and PAOs that require or prefer conflicting
AOB
SRT values (e.g. >15 days required for nitrifiers and <5 days preferred for PAOs). Allowing the
NOB
slow-growing nitrifiers to attach to the carrier media and the faster-growing phosphorus (P)removing organisms (and other heterotrophs, e.g. denitrifiers) to be in the suspended mixed liquor (ML), the EBPR-IFAS system facilitates separate SRT controls and overall optimization for both N and P removal processes. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
The increasingly stringent limits imposed on nitrogen (N) and phosphorus (P) discharge in wastewater effluents demand for more reliable and better optimization of Biological Nutrient Removal (BNR) processes that target for simultaneous N and P removal. Efficient and reliable N removal normally requires
relatively long solid residence time (SRT > 8e15 days) for nitrification process and sufficient carbon source for denitrification process. Fixed film systems such as Integrated FixedFilm Activated Sludge (IFAS) or moving bed biofilm reactors (MBBR) have been shown to be successful for the enhancement of nitrification and denitrification in BNR system upgrade (Azimi et al., 2007; Christensson and Welander, 2004; Ødegaard,
* Corresponding author. Tel.: þ1 617 373 3631; fax: þ1 617 373 4419. E-mail addresses:
[email protected] (A. Onnis-Hayden),
[email protected] (N. Majed),
[email protected] (A. Schramm),
[email protected] (A.Z. Gu). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.04.039
3846
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 4 5 e3 8 5 4
2006; Onnis-Hayden et al., 2007; Sen et al., 1994). IFAS allows for decoupling of the growth rate of nitrifying populations and the suspended mixed liquor phase SRT (MLSRT) and, it provides higher treatment capacity with smaller footprint due to increased solids inventory on the carrier media. In addition, less waste sludge production and improvement in robustness and resistance to load variations were observed (Maas, 2007). These advantages make fixed-film systems such as IFAS or MBBR preferable for N removal. As for phosphorus removal in a BNR process, one main challenge remaining with the enhanced biological phosphors removal (EBPR), is how to improve its reliability and stability since many EBPR plants experience unpredicted upsets and performance fluctuations (Neethling et al., 2005; Gu et al., 2008). Among the identified factors that affect the stability of EBPR process, maintaining conditions favoring the proliferation of polyphosphate accumulating organisms (PAOs) over Glycogen Accumulating Organisms (GAOs) is critical (Gu et al., 2008; Christensson et al., 1998). Shorter SRT (<3 days), higher pH (>7.25) and certain substrates (e.g. propionate) and feeding strategy seem to favor PAOs over GAOs (Filipe et al., 2001; Oehmen et al., 2005; Rodrigo et al., 1996; Whang and Park, 2006). The possibility of incorporating IFAS into an EBPR process has been recently explored by a few researchers at pilot scale (Christensson and Welander, 2004; Sriwiriyarat and Randall, 2005 and Kim et al., 2010). Effective N and P removal at a full-scale IFAS-BNR plant in Broomfield, Colorado has been recently reported (Onnis-Hayden et al., 2007; Rogalla et al., 2006). These limited number of studies demonstrated the potential of IFAS-EBPR for simultaneous N and P removal, although detailed microbial populations analysis was not carried out in any of these studies. One unrecognized and therefore not fully-investigated advantage of an IFAS-EBPR system is that it potentially enables separate SRT control for the slower-growing nitrifiers and the fastergrowing heterotrophs including PAOs and denitrifiers, by allowing the former to attach to the carrier media and the latter to be in the suspended mixed liquor (ML). This hypothesis is based on the understanding that nitrifiers usually prefer to reside on fixed-film carrier media, whereas PAOs and denitrifiers (some denitrifiers may be PAOs) would mostly reside in the circulating mixed liquor because proliferation of PAOs requires alternating anaerobic and aerobic/ anoxic conditions as provided by the circulating mixed liquor. This decoupling ability is desirable in full-scale practice since the decoupling and separate SRTs controls of key functionally relevant populations allow for simultaneous optimization for both N and P removal processes. To evaluate the validity of this hypothesis, we conducted and reported for the first time a comprehensive and integrated evaluation of the PAO populations and P removal performance, as well as nitrifying populations (ammonia-oxidizing bacteria (AOB) and nitrite-oxidizing bacteria (NOB)) and nitrification activity at a full-scale IFAS-EBPR system. The PAOs, AOB and NOB population abundance and distribution, the N and P removal activities and their distribution on the biofilm (carrier media) versus that in the suspended biomass in the IFAS-EBPR system were evaluated. The implication of the results on the IFAS-EBPR process modeling, design and operation were discussed.
2.
Material and methods
2.1.
Full scale IFAS-EBPR process
Broomfield Wastewater Treatment Plant (WWTP) located in Denver, Colorado is one of the few full-scale municipal wastewater treatment plants that were designed as an IFASBNR process in the United States (Fig. 1A). The secondary treatment process consists of a pre-anoxic zone, an anaerobic and an anoxic stage followed by two-staged IFAS aeration basins in series that contain AnoxKaldnes K1 carrier media (see Fig. 1A). K1 media is made of high-density polyethylene (density 0.95 g/cm3) with an effective surface area of 500 m2 per m3 and, the water volume displaced by the carrier media is about 6.4% with a volumetric filling of 30%. Current permit requires monthly average effluent NH4eN <5 mg/l. Although there is no P limit, the plant has an internal target of effluent TP < 1 mg/l. Operational conditions at the plant for the period of the study are reported in Table 1.
2.2. Batch tests for evaluation of EBPR activity and distribution among different forms of biomass To investigate the level and distribution of PAOs activities between the suspended mixed liquor (ML) and the biofilm media in the IFAS-EBPR processes, batch P uptake and release tests (in triplicates) were carried out with different forms of biomass drawn from aerobic stage 1 and stage 2 IFAS basins (Fig. 1A), including suspended mixed liquor (ML) alone, combination of ML and carrier media (about 30% fill in mixed liquor) and carrier media alone (30% fill in media solution). The tests were conducted on site to eliminate the potential impact of biomass transport (e.g. at least overnight) on the tests results. Details of the P uptake and release tests can be found elsewhere (Gu et al., 2008). Briefly, sodium acetate was added (80 mg/l as acetate) at the beginning of the anaerobic conditions (maintained with nitrogen gas purging for 45 min), then aerobic phase was maintained for 3 h by supplying air flow to maintain a DO level of 4e5 mg/l. Samples were collected at 15 min intervals for about 3 h, immediately filtered through two series filtration (100 mm then 0.45 mm) before they were analyzed for soluble ortho-P, volatile fatty acid (VFA), COD, NO2eN and NO3eN. Dissolved oxygen (DO), pH, temperature and Oxidation Reduction Potential (ORP) were continuously recorded. pH was controlled at values measured in the aeration basin (7e7.3) and the temperature was maintained at 20 0.5 C. MLSS and MLVSS analysis for ML were conducted at the end of each test. Determination of all parameters were performed according to standard methods (SM 4110 for anions, SM 5220 for COD, SM 5560 for VFA, SM 4500-Hþ$B for pH, SM 4500-O$G for DO, SM 2580 for ORP and SM 2540 for MLSS and MLVSS, APHA, 2005). Determination of the total solids (TS) attached to the carrier media was performed according to the procedure suggested by the manufacturer (Anox-Kaldnes, Inc. Sweden). Briefly, 10e20 media carriers were dried in a 105 C oven and the weights were measured before and after a sulfuric acid treatment, which removes the biomass attached to the carrier media. Biomass on media was then derived from the difference in the weights.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 4 5 e3 8 5 4
3847
Fig. 1 e (A) Process schematic of the Broomfield WWTP in Denver, Colorado (FEQ: flow equalization return); (B) profiles of nitrogen and phosphorus species along the BNR process at the Broomfield plant, concentration in respect to influent flow (average of three days, the vertical bars represent the standard deviations). INF: influent; AN: anaerobic; AX: anoxic; AE1: IFAS stage 1; AE2: IFAS stage 2; EFF: effluent.
2.3. Batch tests for evaluation of nitrification activities and distribution among different forms of biomass To determine the nitrification rates, batch tests (in duplicate) were conducted on site at the Broomfield WWTP for sludge samples taken from both aerobic IFAS stage 1 and stage 2
Table 1 e Operating conditions at the Broomfield WWTP during the period of study. Data are based on those collected for 5 years of operation (frequency of measurements: daily for most parameters, once a week for MLSRT). Parameter
Value range [average] 3
Flow rate (m /s) Secondary influent bCOD (mg/l)a Influent TP (mgP/l) Influent ammonia (mgN/l) Influent NOx (mgN/l) MLSRT (days) Recycle flows in relative to the influent flow Temperature [ C] C/P (mg bCOD/mgP) [b]
0.11e0.39 [0.2] 43e361 [177] 4e16.5 [8.5] 17e66 [35.2] 0e15.4 [6] 2.6e5.6 [3.78] RAS 40% MLR 160% 13e22 [17.4] 8.4e30[21] [17.5]b
a bCOD ¼ 1.6 BOD5. b Average C/P available to PAOs, the value is corrected considering the presence of NOx in influent.
(Fig. 1A). Samples were transferred in the 2-L beaker, aerated to obtain a dissolved oxygen (DO) concentration similar to the one in the aeration basin (4 mg/l) and sodium bicarbonate (alkalinity of 200 mg/L as CaCO3) and ammonium chloride (20 mg NH4eN/L) were added to the reactor. Subsequently samples were collected at 15-min intervals for about 3 h, immediately filtered through two series filtration (100 mm then 0.45 mm) and then analyzed for NH4eN, NO2eN and NO3eN. Dissolved oxygen (DO), Temperature, pH, and OxidationReduction Potential (ORP) were continuously recorded. Analytical methods used for the above mentioned parameters are the same as previously described. At the end of the test, MLTSS, MLVSS and TS attached to the carrier media were also analyzed as previously described.
2.4. Identification and quantification of candidate PAOs and GAOs, AOB and NOB Observation and quantification of candidate PAOs and GAOs residing in biomass from suspended mixed liquor and in biomass from scraped biofilm on the carrier media were investigated by Neisser, PHB and DAPI staining (Jenkins et al., 1993; Streichan et al., 1990), as well as by fluorescence in situ hybridization (FISH) targeting known PAOs and GAOs (see FISH probes in Table S1, supplementary information). The FISH protocol and hybridization conditions used were
3848
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 4 5 e3 8 5 4
previously described (He et al., 2008; Zilles et al., 2002). Large sample aggregates were avoided by mild sonication (5W, 1 min) and samples were homogenized passing them through a 27 gage syringe needle for 10e20 times. For the determination of PAO fraction, intracellular polyP was visualized by incubation with 1 ug/mL of 40 ,6-Diamidino-2-phenylindole (DAPI) for 60 min (Zilles et al., 2002). Under these conditions, cells containing a large amount of polyP are stained bright yellow while the rest of the cells are blue. The fractions of PAOs (yellow) were determined as the percentage of the total cells (blue þ yellow). On separate slides, Accumulibacterrelated organisms were detected by 16S rRNA-targeted fluorescent FISH. After hybridization, the slide was counterstained with 1 mg/mL of DAPI solution for 3 min to quantify total cells to allow the estimation of the fraction of Accumulibacter expressed as the percentage of the total cells. For some selected slides, combined DAPI and FISH were performed to visualize the overlay of Poly-P staining and Accumulibactertype FISH results, with the aim to observe the involvement of Accumulibacter-type PAOs in EBPR, and these results, however, were not used for quantification of Accumulibacter fraction. Although nitrifying bacteria are found in at least 6 different phylogenetic groups, only three major groups (AOB within Betaproteobacteria and NOB of the genera Nitrospira and Nitrobacter) are to be expected in wastewater treatment. Therefore, probes specific for these groups were chosen, and additional probes were used to resolve the various lineages within the genus Nitrosomonas (Supplementary Table S1). Initial PCRbased analysis had revealed the absence of ammoniaoxidizing Archaea (AOA) in the system (data not shown). Betaproteobacterial AOB, and NOB of the genus Nitrobacter and Nitrospira were analyzed in paraformaldehyde (4%)-fixed and homogenized samples by FISH with a suite of published probes according to standard protocols (Pernthaler et al., 2001). Hybridized cells were observed with an epifluorescent microscope (Zeiss Axioplan 2, Zeiss, Oberkochen, Germany). Quantifications of population distributions were carried out using the software DAIME (Daims et al., 2006). Around 20e25 separate randomly chosen images were evaluated with final results reflecting the cumulative biovolumetric fractions of Accumulibacter, Competibacter, total PAOs, AOB and NOB present in the corresponding samples. Microbial population fractions were expressed as percentage of EUB or DAPI stained cells.
3.
Results and discussion
3.1.
Effective P and N removals in the IFAS-EBPR system
The IFAS system at Broomfield has performed very well consistently over the past 5 years and monthly average data for influent and effluent nitrogen and phosphorus species are presented in Figure S1. The effluent ammonia concentration averaged 0.37 0.5 mg/l, indicating efficient and complete nitrification, even during the winter months at low temperature of 11e15 C at a relatively short MLSRT of 3.5e4 d. This demonstrates the advantage of IFAS system that can retain the majority of nitrifiers on the biofilm media, provides high treatment capacity and improves stability at low
temperatures. The variation in the effluent TN reflects the seasonal varying requirements for denitrification at the plant. The average effluent ortho-P for the five years of operation was found to be 0.87 0.62 mgP/l and the average TP was 1.2 0.78 mg/l, indicating good P removal at the plant with influent TP varying from 6 to 21.7 mg/l (Figure S1). One point worthy of mentioning is that this plant has relatively high nitrate and nitrite level in the influent (average influent NO3eN and NO2eN at the plant were 3.1 mg/l and 2.92 mg/l, respectively). A pre-anoxic zone was included in the plant design to reduce the introduction of nitrate to the following anaerobic stage; however, an average of 0.9 mgNO3eN/l and 7.83 NO3eNmg/l was measured in the anaerobic zone and in the anoxic zone during the time of this study (see Fig. 1B). Analysis of the samples taken in the different zones at the plant indicates P release occurring in the “anaerobic” zone despite the presence of nitrate (see Fig. 1B). Several studies have shown that nitrate has a negative effect on the EBPR performance during the anaerobic phase, due to the competition for carbon between the denitrifying population and the PAOs (Lopez-Vazquez et al., 2008; Yagci et al., 2003), and possible inhibitory effect of nitric oxide produced during denitrification (Van Niel et al., 1998). Therefore, in case where there is nitrate being introduced into the anaerobic zone, sufficient carbon source to satisfy the denitrifiers and ensure a true anaerobic carbon-rich zone is considered to be required for effective EBPR. It is interesting that effective P removal was achieved at Broomfield plant despite the relatively lower COD/P (mg/mg) ratio (17.5, see Table 1) than recommended (C/P > 20e25), (Gu et al., 2008; Randall et al., 1992) and the lack of a true anaerobic zone. The much shorter MLSRT (3.5e4 days) at this plant compared to that for a typical suspended BNR system (10e15 days) might favor the PAOs over GAOs as we hypothesized and it warrants further investigation.
3.2. Distribution of PAOs between the suspended biomass and biofilm on media Table 2 summarizes the abundance and distribution of PAOs associated with different forms/portions of the biomass in the IFAS-EBPR system. Most PAOs were found to be associated with suspended mixed liquor biomass. In contrast to the intuitive expectation that PAOs may only reside in the suspended mixed liquor that is exposed to alternating anaerobic and aerobic condition in an IFAS system, some PAOs were also found in biofilm biomass scraped from carrier media obtained from the aerobic zones in the IFAS-EBPR system. The abundance of cells containing poly-P granules in the media biofilm biomass, however, was much less than that observed in the suspended mixed liquor biomass. The relative abundance of PAOs in the mixed liquor and in biofilm biomass was estimated to be 20e30% and 3e8% of total bacterial cells, respectively. FISH was used to visualize and enumerate Accumulibacterrelated organisms in mixed liquor sludge and in biofilm scraped from the IFAS carrier media (Fig. 2A and B). Accumulibacter-like PAOs accounted for 15.8 1.4% of the total bacterial population in the mixed liquor sample, whereas they represented less than 4 1% of the total bacterial population in the biofilm sample. The abundance of Accumulibacter in ML is comparable to that observed in conventional EBPR plants in the range of 9e24% of total bacterial population (He et al.,
3849
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 4 5 e3 8 5 4
Table 2 e PAOs and EBPR activity distribution among different biomass fractions at the Broomfield WWTP. Biomass fraction
Population distribution
EBPR activity
PAOs fractions [%]a
Accumulibacter fraction [%]a
P-release/P-uptake rate (stage 1)b
Contribution to overall EBPR activity [%]c
Suspended biomass (ML)
25 5
15.8 1.4
96.5%
Attached biomass (Media)
53
41
12.2/3.8 [mgP/gMLVSS/h] 10.1/3.3 [mgP/gMLSS/h] 0.09 [gP/m2/d] 0.49 [mgP/gTS/h]
3.5%
a The numbers represent the percentages of the total bacterial population in the respective sample, determined from the cumulative area identified by quantitative image analysis (Daime) the standard deviation. b The numbers represent the average rates for stage 1. c The % of EBPR activity was calculated considering the fraction of EBPR associated with a specific biomass, respect to the overall EBPR activity of the system.
2008; Kong et al., 2004; Gu et al., 2008). Overlay of DAPI and FISH showed that the Accumulibacter-related organisms and other type of PAOs attached to the IFAS carrier media contained poly-P granules (data not shown), therefore may have been active in EBPR. However, it is difficult to determine whether these PAOs were growing on the IFAS carrier media or simply adhered to the biofilm via contact and exchange with those in the mixed liquor.
3.3. Distribution of PAOs activities between the suspended biomass and the biofilm on media To further evaluate the EBPR activities and contributions of PAOs residing on the biofilm or in the mixed liquor, a series of P release and uptake batch tests were conducted with different forms of biomass in the IFAS-EBPR systems, including suspended mixed liquor (ML), biofilm on the carrier media (media) and combination of suspended mixed liquor and media (ML þ media). Fig. 3A shows an example of the P profiles obtained during three batch tests with different forms of biomass from the aerobic stage 1 at the Broomfield WWTP. The P release amount in the test with media þ ML was slightly higher than that with ML alone, indicating some level of EBPR activity associated with the carrier media. However, for the test with biofilm media alone, there was no trend of P uptake and release
as for EBPR process, instead, there was a slight and continues P release during the test at a rate of 0.61 mgP/L/h. The reason for the absence of P uptake during the test is unclear, however, one possible explanation is that the COD dosed was left unconsumed at the end of the anaerobic phase due to the very low EBPR and denitrifying activities associated with the biofilm, which inhibited the P uptake due to competition for oxygen and/or carbon between the PAOs and other heterotrophs in the biofilm. Previous studies have demonstrated that EBPR activity can occur within fixed-film provided with alternating anaerobic/ aerobic conditions (Goncalves and Rogalla, 2000; Helness and Odegaard, 1999). In a full-scale system, often there is diurnal fluctuation (such as diurnal changes in the influent COD and NH4eN loadings) that may cause micro-scale local and periodical alternating aerobic or anaerobic condition due to DO level and diffusion depth variations, which may allow for growth of PAOs within the biofilm. However, we believe that the presence of a rather low relative abundance of PAOs in the biofilm is most likely due to attachment and detachment exchange of biomass between the mixed liquor and the biofilm because similar observations were found in our lab-scale IFAS-EBPR process that had rather consistent loading conditions (data not shown). Further investigation is therefore needed to better understand the exchange and interaction of
Fig. 2 e FISH Micrographs for samples from IFAS stage #1 of the Broomfield WWTP. (A) Suspended mixed liquor and (B) biofilm from carrier media; the samples were hybridized with Cy3-labeled PAO-mix probe and FAM-labeled EUB probe. Accumulibacter are shown in yellow and all other bacteria are shown in green. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
3850
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 4 5 e3 8 5 4
Fig. 3 e (A) Comparison of soluble ortho-P profiles during the anaerobic (AN) P release and aerobic (AE) P uptake batch testing among three different forms of biomass from IFAS aerobic stage 1 at the Broomfield WWTP. (B) P release and uptake rates for various forms of biomass samples (media plus mixed liquor (ML), mixed liquor only (ML) and media only) from IFAS at the Broomfield WWTP for stage 1 (left), and stage 2 (right). Represented are average values of the triplicate tests, while the vertical bars represent the corresponding standard deviations.
microbial cells (e.g. PAOs) in the biofilm with those in the mixed liquor in an IFAS-EBPR process. Table 2 summarizes the results of the EBPR activity for the different biomass fractions, for stage 1. The results were consistent with the population abundance observed, therefore supporting the hypothesis that most PAOs activity is associated with the suspended biomass. Over 96% of the EBPR activity was indeed associated with the mixed liquor, where the majority of PAOs were found, but only a very small percentage (less than 4%) of EBPR activity was associated with the biofilm. The aerobic P uptake rates and anaerobic P release rates with the mixed liquor biomass of stage 1 were 3.9 0.43 and 12.1 2.1 mgP/gVSS/h, respectively, which were comparable to the values found in other studies for full-scale EBPR plants (Gu et al., 2008; Neethling et al., 2005; Lopez-Vazquez et al., 2008). Similar values were also obtained, as expected, for tests using ML from stage 2 (Fig. 3B).
3.4. Distribution of nitrifying microbial populations in the IFAS-EBPR system Abundance of AOB and NOB on carrier media and those in mixed liquor were determined for biomass from aerobic stage 1. Table 3 summarizes the abundance of AOB and NOB estimated for various fractions of biomass in the IFAS-BNR system, as well as the nitrification activities obtained for those fractions. The combined results clearly demonstrate that nitrification is mainly associated with the biofilm attached to the carrier media, where nitrifiers can be maintained even at lower MLSRT and temperature. FISH analysis of biomass scraped off the carrier media showed heterogeneous distribution of AOB and NOB: in some patches (most likely from the anoxic deeper layers of the biofilm) nitrifiers were nearly absent while other areas (most likely from the oxic, nitrifying surface) were densely colonized with AOB and NOB
3851
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 4 5 e3 8 5 4
(Fig. 4). AOB were identified as members of the N. europaea/ eutropha/halophila and the N. oligotropha lineages based on an hierarchical set of probes (BET42a, NSO1225, NEU23a and Nmo218). These lineages have been previously found in sequencing batch biofilm reactors (Gieseke et al., 2001) and other systems with high ammonium/high salt environments (Juretschko et al., 1998; Koops et al., 2003) and highly fluctuating conditions (especially oxic/anoxic cycles). None of the other five probes targeting betaproteobacterial AOB (NSV443, NSE1472, NmV, NmIV, NmII) yielded any positive results, therefore indicating that the AOB community of the biofilm consists entirely of the two populations mentioned above. In the presumably oxic, nitrifying part of the biofilm, both AOB together accounted for less than 10% of the total population. In the mixed liquor only few single cells or small aggregates of AOB were detected, most likely originating from the biofilm by detachment. NOB were rather abundant in the presumably oxic, nitrifying part of the biofilm (10e20% of the total population), forming typical cell clusters with extremely small cells (<1 mm) of Nitrospira sp., as identified by probes Ntspa712 and Ntspa662 (Fig. 4). Nitrobacter sp. was not detected. Such high abundance of Nitrospira sp. has been previously reported from sequencing batch biofilm reactors (Gieseke et al., 2001; Schramm, 2003). Similar to AOB, only few NOB were detected in the mixed liquor, mostly as very small aggregates, again indicating that they originated from the biofilm by detachment.
3.5. Distribution of nitrification activity between the suspended biomass and the biofilm on media Nitrification activities associated with various forms of biomass in the IFAS-BNR system including the mixed liquor, the carrier media in the aerobic zone and the mixture of ML and media, were evaluated with batch tests. Figure S2 shows exemplary nitrate profiles obtained with the different biomass fractions. A summary of the nitrification rates obtained from the batch tests is presented in Table 3 and Table 4. The nitrification rate per carrier media surface area for the stage 1 was consistent with those obtained by others (Rusten et al., 1995; Rusten et al., 2003). It is clear that the nitrification rate associated with the biomass attached to the carrier media for stage 1 is much higher than the one obtained with only ML, indicating higher
Fig. 4 e FISH images of biomass from the nitrifying part of the biofilm from stage 1. Orange-red, Nitrospira-like NOB (hybridized with probe Ntspa662-CY3); whitish (pink)-blue, Nitrosomonas oligotropha -like AOB (hybridized with probes Nmo218-CY3, NSO1225-FITC, and BET42a-CY5); light (green-) blue, other AOB (Nitrosomonas europaea/ eutropha/halophila-like, hybridized with probes NSO1225FITC, and BET42a-CY5); dark blue, other Betaproteobacteria (hybridized with probe BET42a-CY5); green, background fluorescence of the biofilm. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
abundance and/or activities of nitrifiers associated with carrier media than those with ML. The specific nitrification rates in the tests conducted with only mixed liquor from stage 1 and stage 2 were very similar as expected since it is the same ML. Note that the specific nitrification rates obtained in the test with carrier media only from stage 1 was 145% higher than the one from stage 2. The biofilm thickness was also substantially different in the two reactors by visual inspection (thinner and darker biofilm in carrier media from stage 2 than
Table 3 e Nitrifiers and nitrification activities distribution among different biomass fractions at the Broomfield WWTP. Biomass fraction
Population distribution AOB fraction [%]a
NOB fraction [%]a
Suspended biomass (ML)
<1%
<1%
Attached biomass (Media)
74
15 6
Nitrification activity Nitrification rate (stage 1)
Contribution to overall nitrification activity [%]b
2.5 [mgN/gMLVSS/h] 2.0 [mgN/gMLSS/h] 1.1 [gN/m2/d] 5.87 [mgN/gTS/h]
25.4% 75.6%
a The numbers represent the percentages of the total bacterial population in the respective sample, determined from the cumulative area identified by quantitative image analysis Daime the corresponding standard deviation. b The % of nitrification activity was calculated considering the fraction of nitrification associated with a specific biomass, respect to the overall nitrification activity of the system.
3852
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 4 5 e3 8 5 4
stage 1) and, this was further confirmed by the higher value of biomass attached on unit media surface area in stage 1 than that of stage 2 (7.93 versus 4.97 g/m2). Nitrification rates and nitrifier presence on the fixed film are affected by dissolved oxygen (DO), organic loading and ammonium concentration. Bulk ammonium concentration can be limiting at concentration lower than 1e3 mgNH4eN/l (Ødegaard, 2006) or when the oxygen to ammonium concentration ratio is higher than 2e5 (Hem et al., 1994). In the case of Broomfield, organic loading and dissolved oxygen concentrations are similar for both stages however the average NH4eN concentrations were found to be 3.6 mg/L and 0.6 mg/L in stage 1 and 2, respectively and the DO to ammonia ratio was 1.3 for stage 1 and 7.5 for stage 2. Therefore, the differences in the nitrification rates are likely attributed to the difference in ammonium level in the two stages, resulting in ammonialimiting condition for stage 2. The results also show that the addition of the carrier media in the hybrid system increased the nitrification capacity by about 155% for stage 1 and by 25% for stage 2. This indicates that the benefits of maintaining higher nitrification activity on the fixed-film start to diminish as the residual ammonia concentration becomes really low and is limiting the reaction rates. Another point worth mentioning is the presence of nitrite (up to 2.5 mg/l) in all the tests performed with the ML only, whereas in the tests containing carrier media, the concentration of nitrite remained very low (below 0.2 mg/l); this fact can be seen as a sign of more stable and coupled AOB and NOB population on carrier media than in mixed liquor. This observation is consistent with the population analysis results, which showed that the majority of AOB and NOBs reside on the carrier media and only a very limited number of AOB and NOB are present in the mixed liquor. During the time of this study, the aerobic MLSRT was about 3.5 days and the average plant temperature was 17 C. The minimal SRT required for nitrification at 17 C was calculated to be 5.96 days (Tchobanoglous
et al., 2003), therefore the nitrification activities observed with ML in the IFAS-EBPR system is likely due to sloughing off of nitrifiers from the biofilm.
3.6. Implication of the results on design operation and modeling of IFAS-EBPR systems Our results demonstrated that for an IFAS-EBPR process, PAOs and EBPR activity is predominately associated with the mixed liquor (96% of the total P removal activity) and in contrast, most nitrifiers and nitrification activity (>75%) reside on the carrier media. These findings have several implications for BNR design and operation. First, the MLSRT that controls most of the EBPR populations can be varied and optimized to favor EBPR activities and potentially improve stabilities without being restrained by nitrifying populations. For example, SRT was suggested to be a possible factor that impacts the competition between PAOs and GAOs, with the former preferring short SRT (less than 3.5 days) (Whang and Park, 2006). Although other factors, such as substrate type, pH and temperature, affect the EBPR populations dynamics as well as previously described, these parameters are very difficult, if possible at all, to be adjusted for real practice. The flexibility of adjusting the MLSRT for optimizing the EBPR process, without affecting the nitrification performances, is therefore a valuable advantage of an IFAS-EBPR system for achieving simultaneous and optimal P and N removal. The dynamics of the PAO activity and its distribution between ML and the carrier media are not currently considered and addressed in the IFASEBPR process design and modeling. Particularly, the factors that affect the extent of PAO populations and EBPR activities associated with the biofilm, although shown to be a relatively small fraction for this study, are not fully understood and require further investigation. The contribution from the PAOs residing on the biofilm media to the overall P removal and
Table 4 e Summary of results from the nitrification batch testing. Stage 1 tests
Dissolved oxygen (mg/L) Temperature ( C) MLSS (mgTSS/l) Attached biomass (g/m2) Total biomass (mgTS/l) Nitrate formation Rate gNOx-N/m2/d Ammonia oxidation rate mgNH4-N/gTS/h Nitrate formation rate mgNOx-N/gTS/h Ammonia oxidation rate mgNH4 -N/l/h Nitrate formation rate mgNOx-N/l/h
a
Stage 2 tests b
ML only
Media only
ML þ media
S(ML þ media)
ML only
Media only
ML þ mediaa
S(ML þ media)b
3.68 20.1 1538 e 1538 e
4.20 19.9 e 7.93 1488 1.12
4.20 20.0 698 7.74 2225 e
e e e e 2225 e
3.58 20.3 1546 e 1546 e
3.98 20.0 e 4.97 1010 0.28
3.27 20.2 1425 4.46 2239 e
e e e e 2239 e
2.64
5.89
4.77
4.87
2.00
2.36
2.30
2.13
2.02
5.87
5.16
4.66
1.91
2.39
2.38
2.08
4.06
8.76
10.63
12.40
3.10
2.39
5.16
5.55
3.10
8.73
11.47
12.34
2.96
2.41
5.34
5.36
a Measured value with both media and mixed liquor (ML) in the reactor. b Calculated value using test results with either media alone or ML alone, separately.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 4 5 e3 8 5 4
their potential impact on nitrification performance on the IFAS media (e.g. competition for oxygen) need to be quantified and incorporated into IFAS-EBPR process modeling and design. For N removal in a IFAS-EBPR system, the dominant residence of nitrifying populations on the biofilm media decouples the growth of nitrifiers from the MLSRT, therefore potentially allows for a more robust and reliable nitrification process that is more resistant to hydraulic loading fluctuation and toxic shocks, as well as temperature changes. It also seems to lead to more stable and coupled AOB and NOB population on carrier media than in mixed liquor, as indicated by the higher nitrite accumulation observed during the batch testing with the ML than those with biofilm media, as previously discussed. Another possible advantage of having shorter MLSRT in IFAS-EBPR system is related to denitrification rate. Specific denitrification rate (SDNR) is affected by readily biodegradable COD, nitrate concentration and biomass SRT (F/M ratio) and at any given COD and NO3eN levels, the SDNR is reversely correlated to SRT (Tchobanoglous et al., 2003). Comparing to a conventional suspended nitrogen removal activated sludge process, the IFAS-EBPR process that allows for much shorter MLSRT (2e4 days versus 8e15 days) for the suspended mixed liquor where most of the denitrifiers would reside, would lead to higher SDNR than conventional BNR plants.
4.
Conclusions
In conclusion: 1. In the full-scale IFAS-EBPR process studied, PAOs and EBPR activity is predominately associated with the mixed liquor rather than the biofilm media. The relative abundance of PAOs and Accumulibacter-like PAOs was estimated to be 20e30% and 15.8 1.4% in the mixed liquor and, 3e8% and 4 1% in the biofilm media, respectively. 2. Abundance and distribution of nitrifying populations and their activities showed that most nitrifiers and nitrification activity (>75%) reside on the carrier media rather than in the mixed liquor. In addition, more coupled AOB and NOB populations and more stable nitrate removal (indicated by less nitrite accumulation) was observed with biofilm than with mixed liquor. 3. The addition of the carrier media in the hybrid system increased the nitrification capacity, but it was found that the benefits of maintaining higher nitrification activity on the fixed-film starts to diminish in secondary stages where the residual ammonia concentration becomes limiting. 4. The results demonstrated that in the IFAS-EBPR process, the Neremoving and Peremoving populations that require or prefer conflicting SRT values (e.g. > 15 days for slowgrowing nitrifiers and <5 days for fast-growing PAOs) can be decoupled, therefore allowing for separate SRT control and overall optimization for both N and P removal processes. 5. The results from this study contribute to the fundamental understanding and further development of comprehensive mathematical models for IFAS-EBPR process design and modeling.
3853
Acknowledgment This study was funded by AnoxKaldnes, USA. The authors gratefully acknowledge the personnel at the Broomfield WWTP, especially Mr. Ken Rutt and Mr. Anthony Tuka, for all their help during the testing period. We also acknowledge Daniel Dair, Chandler Johnson, Kruger Inc. AnoxKaldnes and Magnus Christensson and Thomas Welander of AnoxKaldnes AB, for their contribution toward this project and, Britta Poulsen from the Department of Biological Sciences, Microbiology, Aarhus University, and Jason Flowers and Trina McMahon from the Department of Civil and Environmental Engineering, University of Wisconsin-Madison, for their advice and assistance with the molecular analyses.
Appendix. Supplementary material Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.watres.2011.04.039.
references
APHA, 2005. Standard Methods for the Examination of Water and Wastewater for the Examination of Water and Wastewater. American Public Health Assoc., Washington, DC. Azimi, A.A., Hooshyari, B., Mehrdadi, N., Bidhendi, G.N., 2007. Enhanced COD and nutrient removal efficiency in a hybrid integrated fixed film-activated sludge process. Iranian Journal of Science and Technology Transaction B-Engineering 31 (B5), 523e533. Christensson, M., Lie, E., Jonsson, K., Johansson, P., Welander, T., 1998. Increasing substrate for polyphosphate-accumulating bacteria in municipal wastewater through hydrolysis and fermentation of sludge in primary clarifiers. Water Environment Research 70 (2), 138e145. Christensson, M., Welander, T., 2004. Treatment of municipal wastewater in a hybrid process using a new suspended carrier with large surface area. Water Science and Technology 49 (11e12), 207e214. Daims, H., Lucker, S., Wagner, M., 2006. Daime, a novel image analysis program for microbial ecology and biofilm research. Environmental Microbiology 8 (2), 200e213. Filipe, C.D.M., Daigger, G.T., Grady, C.P.L., 2001. Effects of pH on the rates of aerobic metabolism of phosphate-accumulating and glycogen-accumulating organisms. Water Environment Research 73 (2), 213e222. Gieseke, A., Purkhold, U., Wagner, M., Amann, R., Schramm, A., 2001. Community structure and activity dynamics of nitrifying bacteria in a phosphate-removing biofilm. Applied and Environmental Microbiology 67 (3), 1351e1362. Goncalves, R.F., Rogalla, F., 2000. Optimising the A/O cycle for phosphorus removal in a submerged biofilter under continuous feed. Water Science and Technology 41 (4e5), 503e508. Gu, A.Z., Saunders, A., Neethling, J.B., Stensel, H.D., Blackall, L.L., 2008. Functionally relevant microorganisms to enhanced biological phosphorus removal performance at full-scale wastewater treatment plants in the United States. Water Environment Research 80 (8), 688e698. He, S., Gu, A.Z., McMahon, K.D., 2008. Progress toward understanding the distribution of Accumulibacter among full-
3854
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 4 5 e3 8 5 4
scale enhanced biological phosphorus removal systems. Microbial Ecology 55 (2), 229e236. Helness, H., Odegaard, H., 1999. Biological phosphorus removal in a sequencing batch moving bed biofilm reactor. Water Science and Technology 40 (4e5), 161e168. Hem, L.J., Rusten, B., Odegaard, H., 1994. Nitrification in a movingbed biofilm reactor. Water Research 28 (6), 1425e1433. Jenkins, D., Richard, M.G., Daigger, G.T., 1993. Manual on the Causes and Control of Activated Sludge Bulking and Foaming. Lewis Publishers, New York, N.Y. Juretschko, S., Timmermann, G., Schmid, M., Schleifer, K.H., Pommerening-Roser, A., Koops, H.P., Wagner, M., 1998. Combined molecular and conventional analyses of nitrifying bacterium diversity in activated sludge: Nitrosococcus mobilis and Nitrospira-like bacteria as dominant populations. Applied and Environmental Microbiology 64 (8), 3042e3051. Kim, H.S., Gellner, J.W., Boltz, J.P., Freudenberg, R.G., Gunsch, C.K. , Schuler, A.J., 2010. Effects of integrated fixed film activated sludge media on activated sludge settling in biological nutrient removal systems. Water Research 44 (5), 1553e1561. Kong, Y.H., Nielsen, J.L., Nielsen, P.H., 2004. Microautoradiographic study of Rhodocyclus-related polyphosphate accumulating bacteria in full-scale enhanced biological phosphorus removal plants. Applied and Environmental Microbiology 70 (9), 5383e5390. Koops, H.P., Purkhold, U., Pommerening-Ro¨ ser, A., Timmermann, G., Wagner, M., 2003. The lithoautotrophic ammonia-oxidizing bacteria. In: Dworkin, M., Falkow, N., Rosenberg, H., Schleifer, K.H., Stackebrandt, E. (Eds.), The Prokaryotes: An Evolving Electronic Resource for the Microbiological Community, third ed. Springer, New York. Lopez-Vazquez, C.M., Hooijmans, C.M., Brdjanovic, D., Gijzen, H.J. , van Loosdrecht, M.C.M., 2008. Factors affecting the microbial populations at full-scale enhanced biological phosphorus removal (EBPR) wastewater treatment plants in the Netherlands. Water Research 42 (10e11), 2349e2360. Maas, C.L.A., 2007. Detachment of Solids and Nitrifiers in IFAS Systems, In: Proceedings of WEFTEC 2007, San Diego, CA. Neethling, J.B., Bakke, B., Benisch, M., Gu, A.Z., Stephens, S., Stensel, H.D., 2005. Factors Influencing the Reliability of Enhanced Biological Phosphorus Removal. Report 01-CTS-3. Water Environment Research Foundation, Alexandria, Virginia. Ødegaard, H., 2006. Innovations in wastewater treatment: the moving bed biofilm process. Water Science and Technology 53 (9), 17e33. Oehmen, A., Yuan, Z.G., Blackall, L.L., Keller, J., 2005. Comparison of acetate and propionate uptake by polyphosphate accumulating organisms and glycogen accumulating organisms. Biotechnology and Bioengineering 91 (2), 162e168. Onnis-Hayden, A., Dair, D., Johnson, C., Schramm, A., Gu, A.Z., 2007. Kinetics and nitrifying populations in nitrogen removal processes at a full-scale integrated fixed-film activated sludge (IFAS) plant. In: Proceedings of WEFTEC 2007, San Diego, CA.
Pernthaler, J., Glockner, F.O., Schonhuber, W., Amann, R., 2001. Fluorescence in situ hybridization (FISH) with rRNA-targeted oligonucleotide probes. Methods in Microbiology 30 (30), 207e226. Randall, C.W., Barnard, J.L., Stensel, D.H., 1992. Design and Retrofit of Wastewater Treatment Plants for Biological Nutrient Removal. Technomic Publishing Co, Lancaster, PA. Rodrigo, M.A., Seco, A., PenyaRoja, J.M., Ferrer, J., 1996. Influence of sludge age on enhanced phosphorus removal in biological systems. Water Science and Technology 34 (1e2), 41e48. Rogalla, F., Johnson, T.L., McQuarrie, J., 2006. Fixed film phosphorus removal e flexible enough? Water Science and Technology 53 (12), 75e81. Rusten, B., Hem, L.J., Odegaard, H., 1995. Nitrification of municipal waste-water in moving-bed biofilm reactors. Water Environment Research 67 (1), 75e86. Rusten, B., Merete Nielsen, I., Welander, T., Rasmussen, V., 2003. Increasing the capacity of activated sludge plants by using AS/ MBBR hybrid process with Kaldnes biofilm carriers. In: IWA (Ed.), 5th IWA Conference on Biofilms Systems Cape Town, South Africa. Schramm, A., 2003. In situ analysis of structure and activity of the nitrifying community in biofilms, aggregates, and sediments. Geomicrobiology Journal 20 (4), 313e333. Sen, D., Mitta, P., Randall, C.W., 1994. Performance of fixed-film media integrated in activated-sludge reactors to enhance nitrogen removal. Water Science and Technology 30 (11), 13e24. Sriwiriyarat, T., Randall, C.W., 2005. Performance of IFAS wastewater treatment processes for biological phosphorus removal. Water Research 39 (16), 3873e3884. Streichan, M., Golecki, J.R., Schon, G., 1990. Polyphosphateaccumulating bacteria from sewage plants with different processes for biological phosphorus removal. FEMS Microbiology Ecology 73 (2), 113e124. Tchobanoglous, G., Burton, F.L., Stensel, H.D., 2003. In: Eddy, Ma (Ed.), Wastewater Engineering: Treatment and Reuse. McGraw-Hill, New York, NY. Van Niel, E.W.J., Appeldoorn, K.J., Zehnder, A.J.B., Kortstee, G.J.J., 1998. Inhibition of anaerobic phosphate release by nitric oxide in activated sludge. Applied and Environmental Microbiology 64 (8), 2925e2930. Whang, L.M., Park, J.K., 2006. Competition between polyphosphateand glycogen-accumulating organisms in enhanced-biologicalphosphorus-removal systems: effect of temperature and sludge age. Water Environment Research 78 (1), 4e11. Yagci, N.O., Tasli, R., Artan, N., Orhon, D., 2003. The effect of nitrate and different substrates on enhanced biological phosphorus removal in sequencing batch reactors. Journal of Environmental Science and Health Part A-Toxic/Hazardous Substances & Environmental Engineering 38 (8), 1489e1497. Zilles, J.L., Peccia, J., Kim, M.W., Hung, C.H., Noguera, D.R., 2002. Involvement of Rhodocyclus-related organisms in phosphorus removal in full-scale wastewater treatment plants. Applied and Environmental Microbiology 68 (6), 2763e2769.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 5 5 e3 8 6 2
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
UV reactor flow visualization and mixing quantification using three-dimensional laser-induced fluorescence Varun Gandhi a, Philip J.W. Roberts a, Thorsten Stoesser a, Harold Wright b, Jae-Hong Kim a,* a b
School of Civil and Environmental Engineering, Georgia Institute of Technology, 200 Bobby Dodd Way, Atlanta, GA 30332-0373, USA Carollo Engineers, 12592 West Explorer Dr. Suite # 200, Boise, ID 83713-1596, USA
article info
abstract
Article history:
Three-dimensional laser-induced fluorescence (3DLIF) was applied to visualize and quan-
Received 28 January 2011
titatively analyze mixing in a lab-scale UV reactor consisting of one lamp sleeve placed
Received in revised form
perpendicular to flow. The recirculation zone and the von Karman vortex shedding that
15 April 2011
commonly occur in flows around bluff bodies were successfully visualized. Multiple flow
Accepted 22 April 2011
paths were analyzed by injecting the dye at various heights with respect to the lamp sleeve.
Available online 5 May 2011
A major difference in these pathways was the amount of dye that traveled close to the sleeve, i.e., a zone of higher residence time and higher UV exposure. Paths away from the
Keywords:
center height had higher velocities and hence minimal influence by the presence of sleeve.
UV disinfection
Approach length was also characterized in order to increase the probability of microbes
Hydrodynamics
entering the region around the UV lamp. The 3DLIF technique developed in this study is
Unsteady turbulence
expected to provide new insight on UV dose delivery useful for the design and optimization
3DLIF
of UV reactors. ª 2011 Elsevier Ltd. All rights reserved.
Laser-induced fluorescence
1.
Introduction
UV disinfection has been gaining popularity in drinking water treatment over the past decade due to the discovery of the efficient inactivation of Cryptosporidium parvum oocysts and Giardia lamblia cysts at relatively low doses (Clancy et al., 1998; Linden et al., 2002) with much less concern on the formation of disinfection by-products as compared to chemical disinfectants (Bellar et al., 1974; Glaze et al., 1993). However, a spatial and temporal assessment of the UV dose delivered and the reactor performance has been severely limited for utilities practicing UV disinfection. The commonly used validation method, biodosimetry, treats the UV reactor as a “black-box” and hence cannot account for the dependence of the dose delivery on the complex hydrodynamics and the
spatial variation in UV intensity. Development of proper reactor design would be ideally pursued through an understanding of the fluid behavior that determines how microorganisms accumulate UV dose as they spend varying amounts of time in regions of fluctuating light intensity (Lawryshyn and Cairns, 2003). The unsteadiness of flow in UV reactors arises mainly due to the placement of cylindrical UV lamps perpendicular to flow that leads to separation forming unsteady large-scale vortices and consequently significant fluid mixing (Williamson, 1996; Zdravkovich, 1997). These so called von Karman vortices (repeating pattern of alternating vortices produced downstream of a bluff body) compound the complexity and unsteadiness in the flow region behind the lamp, as they involve the interactions of three shear layers,
* Corresponding author. Tel.: þ1 404 894 2216; fax: þ1 404 385 7087. 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.2011.04.041
3856
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 5 5 e3 8 6 2
i.e., a boundary layer around the sleeve, a separating free shear layer, and a highly turbulent wake (Williamson, 1996). The presence of multiple lamps in a staggered configuration further complicates the flow, rendering the prediction of unsteady hydrodynamics significantly difficult. Furthermore, inlet configuration (Sozzi and Taghipour, 2006), upstream pipe bends (Zhao et al., 2009), and the presence of modifications such as baffles (Blatchley et al., 1998; Wols et al., 2010a), rings (Janex et al., 1998), or “wave-like” walls (Chiu et al., 1999a) have been found to significantly affect the hydrodynamics and reactor performance. Due to the complexity of the flow, computational fluid dynamics (CFD) has been increasingly used to model the hydrodynamics in UV reactors based on the time-averaged Reynolds Averaged Navier Stokes (RANS) approaches (Sozzi and Taghipour, 2006; Alpert et al., 2010; Wols et al., 2010b). However, the velocity distributions in the RANS simulations have been found to differ from the experimental results typically obtained using particle image velocimetry (PIV). Alpert et al. (2010) and Wols et al. (2010b) determined that RANS simulations under-predicted the flow complexity in dynamic wake regions and dead zones due to the poor capture of large vortices and turbulent motions. These studies concluded that resolving the unsteady turbulent motions is essential to provide an accurate representation of the microorganism trajectories and more significantly the UV dose received by each microbe. Dose distributions consist of both spatial and temporal components. While many past studies have focused on the former, the authors are unaware of any experimental or computational studies that considered the temporal component in the reactor. In this study, a three-dimensional laserinduced fluorescence (3DLIF) (Tian and Roberts, 2003) was applied for the first time to examine the hydrodynamics in a lab-scale model UV reactor both spatially and temporally. In the 3DLIF system, a planar monochromatic laser sheet is created and scanned across the width of the reactor to obtain three-dimensional images. The laser causes a tracer dye to
fluoresce which is captured by a high-speed CCD camera (Guiraud et al., 1991). Using this technique, instantaneous, 2D and 3D mixing characteristics in a model UV reactor were visualized and quantitatively examined.
2.
Experimental methods
2.1.
3DLIF system
Detailed information regarding the 3DLIF system used in this study is given in Kim et al. (2010). Briefly, a laser beam generated by an argon ion laser (Innova 90, Coherent, Palo Alto, CA) at a wavelength of 514 nm and an intensity of 1.5 W was used to excite a fluorescent dye tracer, Rhodamine 6G (SigmaeAldrich, St. Louis, MO). The laser beam was first directed toward a mirror which oscillated vertically at high frequency to produce a 2D laser sheet that passed through the center width of a reactor. A high-speed CCD camera (Dalsa CA-D6, Ontario, Canada) captured fluorescence images at a frequency of 200 Hz. To obtain 3D images, another mirror scanned the laser sheet horizontally across the reactor width once per 0.2 s with 40 images captured per scan. The images, attained using Video Savant Version 4 (IO Industries, Ontario, Canada), were then processed using software TFlook (Tian and Roberts, 2003) which converted the images that can be visualized in three dimensions by Tecplot (Bellevue, WA). Dye concentrations in these images were calculated from fluorescence intensities using a calibration factor found by images taken of known dye concentrations.
2.2.
Model UV reactor
A lab-scale model UV reactor (Fig. 1) was constructed, using transparent 1.27 cm (0.5 in.) thick acrylic sheets, to examine the hydrodynamics across a single 2.8 cm (1.1 in.) O.D. lamp sleeve placed perpendicular to the direction of flow. Even though it is less common than circular cross-section in full-
Fig. 1 e Schematic of the model lab-scale UV reactor (dimensions in cm).
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 5 5 e3 8 6 2
scale design, the reactor was designed with a square crosssection (7.62 cm 7.62 cm, 3 in. 3 in.) to prevent complications associated with the laser light reflecting from a curved surface. The distance between the wall and the lamp placed closest to the wall of the reactor were similar in the model reactor and full-scale rectangular reactors. In order to achieve flow characteristics that closely represent a full-scale reactor, the influent flow rate was determined by matching the Reynolds number, Re ¼ UW/n, where U ¼ velocity [LT1], W ¼ characteristic length [L], and n ¼ kinematic viscosity [L2T1]. The lab-scale model was operated at a flow velocity of approximately 17 cm/s (0.55 ft/s) or a flow rate of 59 L/min (15.5 gal/min), which corresponds to Re ¼ 4900 based on the sleeve diameter. In full-scale UV reactors, velocities range from 4 to 225 cm/s with Re between 1000 and 100,000 based on sleeve diameter. Since water was transported through a 2.54 cm (1 in.) diameter pipe to the reactor inlet, an abrupt change in the cross-sectional area would lead to significant flow separation, which is not characteristic of full-scale reactors. The inlet was, therefore, designed to gradually diverge and packed with 0.32 cm (1/8 in.) acrylic balls to evenly distribute the influent velocities and avoid flow separation. In order to avoid inlet effects on the region of interest, the upstream region was of constant cross-section for a length of more than ten cylinder diameters.
2.3.
Tracer test
Tap water, dechlorinated and filtered through 10 and 1 mm cartridge filters at room temperature (20 2 C), was fed to the model reactor at the design flow rate. Upon reaching steadystate, a solution containing 20 mg/L Rhodamine 6G was continuously injected into the reactor at 25 mL/min via a 1.27 cm (0.5 in.) long L-shaped injection port (0.32 cm (1/8 in.) O.D., 0.16 cm (1/16 in.) I.D.) aligned in the stream-wise direction. Dye was injected at the center of the lamp sleeve, and 1.27 cm and 2.54 cm (0.5 in. and 1 in., respectively) above the center of the sleeve through the injection point located 5.85 cm (1.75 in.) upstream of the sleeve center and 3.81 cm (1.5 in.) from the side wall. The tracer was conservative since the tests were performed without UV lamps; the sleeve was filled with water to minimize laser light reflection off the sleeve. Image capture was initiated prior to dye injection to completely capture its transport around the sleeve.
3.
Results and discussion
3.1.
Flow visualization using 3DLIF
Fig. 2 shows how three-dimensional flows in a UV reactor can be visualized using 3DLIF at a high resolution (corresponding to millions of sampling points), which is not possible with traditional dye tracer test techniques. These images were obtained from a 3DLIF experiment performed with dye injected at the point in the center of the y-z plane. The region from the sleeve to the outlet is dissected in the stream-wise (x-z plane), span-wise (x-y plane), and cross-stream (y-z plane) directions. The cross-sections are presented as concentration contours normalized by the initial dye concentration in false
3857
Fig. 2 e 3DLIF images showing the dye transport in the UV reactor dissected into multiple cross-sections.
color with red as the highest and blue as the lowest dye concentrations. The gray-colored semi-transparent surface is an iso-concentration surface at a level of 0.15 (arbitrary low threshold normalized value) that indicates the outer extent of the dye cloud as it moves through the reactor. The lamp sleeve is represented as a red cylinder in the 3D images and as a black circle/rectangle in the 2D slices. The 3DLIF visualization shows that the presence of the sleeve affects the dye flow in all three dimensions, however, more pronouncedly in the stream-wise direction. As the dye approached the sleeve, it underwent spreading in all directions as shown by the expansion of the iso-concentration surface which fluctuated (not shown) about the center. Upon contact with the sleeve, the dye plume spread axially as momentum carried it past the sleeve, depicted by the red contours around the sleeve in the stream-wise and crossstream slices. Lower dye concentrations in the wake region indicated increased mixing while a von Karman vortex street was observed in the stream-wise slice, albeit vaguely. Furthermore, a vertical slice in the stream-wise direction offset horizontally by 2.5 cm also showed the presence of the vortex street, lower dye concentrations in the wake region and higher concentrations around the edges of the cylinder, confirming previous observations that the flow around a circular cylinder is three-dimensional (Williamson, 1996). While the presence of the lamp certainly enhanced dispersion in all three directions, it mainly complicated the flow in the streamwise direction with the more unsymmetrical flow patterns. Based on the observations in Fig. 2, further analysis of the stream-wise plane in 2D (Fig. 3) was performed to visualize the details of the flow. Fig. 3a consists of time-series images, spanning 0.12 s, for a dye injected at the center height when the planar laser sheet was fixed at the mid-point between the walls. These 2D images depict the two main characteristics of the flow around a cylinder: the recirculation region and the von Karman vortex shedding (Williamson, 1996; Zdravkovich, 1997). The presence of the sleeve created a low pressure zone in the wake region, causing water, for example, from the top of the cylinder to rotate clockwise and fill up this region (Fig. 3a). At the next instant, a vortex on the opposite side of the lamp was formed (not shown) which rotated counterclockwise into the wake region that caused a release of the first vortex, hence vortex shedding occurred. Immediately, the
3858
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 5 5 e3 8 6 2
Fig. 3 e 2D visualization of (a) instantaneous dye movement (0.12 s) and (b) time-averaged concentration contour for the dye injected at the center. Low concentrations (less than 10% of maximum concentration), depicted as white, were cutoff. The gray outlines represent the walls of the UV reactor.
released vortex began to grow in size while simultaneously, a new vortex, on the same side of the cylinder, was created and the cycle continued (Zdravkovich, 1997). It is important to note that this study visualizes the flow around a cylinder in the presence of two additional shear layers due to boundary layers at the walls, representative of the flow in UV reactors. Very few experimental studies were focused on this flow, one at very low Re (250, based on the cylinder diameter) (Rehimi et al., 2008) and another at very high Re (¼140,000) (Braza et al., 2006). The presence of the lamp induced blockage or a reduction in the cross-sectional area leading to higher velocities around the cylinder by a factor of 1.6. This caused an increase in the length of the recirculation zone as the increase in momentum forced the dye to travel further as compared to an infinite domain (Singha and Sinhamahapatra, 2010). The length of the mean recirculation zone, which was measured as the extent of the red contour in the wake region in the concentration plot averaged over 5 s (Fig. 3b), was approximately 1.75 cm. In addition, the confinement suppressed the complete formation of vortex structures such as the von
Karman streets that are typically observed in the cross-flow across an isolated cylinder (Williamson, 1996; Zdravkovich, 1997). Finally, the vortex shedding frequency was characterized using the Strouhal number St ¼ fD/U, where f ¼ vortex shedding frequency [T1], and D ¼ diameter of the cylinder [L]. St was found to increase when a cylinder is placed near a plane wall (Price et al., 2002). Typical values for the flow in an infinite domain at Re ¼ 4900 is 0.18 St 0.22 (Zdravkovich, 1997). However, in this setup, St ¼ 0.28, calculated using Fast Fourier transform (FFT) of the transient dye concentration at a point 1.5 cm downstream of the sleeve center (data not shown). Thus, the presence of the confinement and the sleeve, due to an increase in velocities around the cylinder, increased the frequency of vortex shedding. Chiu et al. (1999b) suggested that the recirculation zone increased the residence time for some microorganisms, based on the analysis of a large-scale reactor with 25 lamps placed perpendicular to flow in a staggered pattern. According to their random walk model based on laser Doppler velocimetry (LDV) measurements, particles that were trapped in the wake region spent longer time in the reactor, thus accumulating a higher dose. This increase in residence time for some microbes caused other organisms to short-circuit through the reactor as they traveled following a higher velocity path found closer to the walls of the reactors and received a lower dose. Similarly, Sozzi and Taghipour (2006) observed regions of higher velocities along the walls of an annular reactor that reduced the UV dose delivered to the microbes. This velocity gradient would lead to minimal transverse mixing in the system, deviating from the ideal plug-flow. In order to analyze these paths, the height of dye injection was increased from the center. Fig. 4 shows the instantaneous and time-averaged flow of the dye when injected off-center at 1.27 cm (Fig. 4a and b) and 2.54 cm (Fig. 4c and d) above the center of the sleeve. The sequence of images presented is approximately 0.080 s and 0.13 s, respectively. In comparison with the center injection, the off-center injections showed less dye flow into the wake region. In Fig. 4a, dye entered the wake region, however, with declining instances as compared to the center injection. Fig. 4b showed highest concentrations directly above the lamp sleeve, while only about 60% of the dye entered the recirculation zone as indicated from the legend. Only some of the dye that rotated clockwise into the recirculation zone remained. Then, from below the cylinder, water, rotating counter-clockwise, entered into this zone causing the vortex to release at the top of the cylinder, hence the vortex was shed. Simultaneously, water entered the recirculation zone and the cycle was repeated. Finally, downstream of the cylinder region, dye was observed to be entrained into a von Karman vortex street. Hence, even though dye was injected away from the center, the presence of the cylinder affected this path of travel, entraining substantial amounts of dye into the wake region. Negligible amounts of dye entered the recirculation zone and no vortex was shed when dye was injected at 1 cm; all of the dye followed a straight path away from the recirculation zone past the sleeve. These observations were supplemented by Fig. 4d which showed all the dye flew past the cylinder close to the walls followed by minimal dye mixing occurring downstream of the UV lamp.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 5 5 e3 8 6 2
3859
Fig. 4 e 2D visualization of (a) instantaneous dye movement (0.08 s), (b) time-averaged concentration contour for the dye injected 1.3 cm off-center and (c) instantaneous dye transport (0.13 s), (d) time-averaged concentration contour for the dye injected 2.5 cm off-center.
In order to supplement the above observations, dye concentration as a function of radial distance away from the lamp (Supporting Information Fig. S1) was analyzed. Concentrations from circular discs around the cylinder were obtained from the time-averaged plots of each injection case using Tecplot, where each disc, with radial increments, Dr ¼ 0.05 cm, from the surface of the sleeve, had a single averaged dye concentration value. As expected, the highest dye concentrations were found closest to the lamp in the center injection, which would correspond with the higher UV intensities thus receiving the highest doses. For the off-center cases, the highest dye concentrations corresponded with distances away from the lamp, thus would be exposed to lower UV intensities (approximately 20% and 4% of the lamp surface intensity for 1.3 cm and 2.5 cm off-center injections, respectively, from Fig. S1 inset) and hence would receive a proportionally lower dose. Superimposed on the effect of decreased residence time at an off-center injection as discussed above, this would lead to much lower level of inactivation compared to the center injection, the quantitative evaluation of which is the focus of a future study.
3.2.
Approach section
Dye, when injected at the wall, showed increased dispersion as the length of the inlet section increased, hence a higher probability of microbes entering the region close to the UV lamp. Dye was injected at 20 mL/min at the top wall through point sources
located either at 22.9 cm (9 in.) or 32.4 cm (12.75 in.) upstream of the lamp sleeve. The resulting dye concentrations around the sleeve are presented in Fig. 5 as time-averaged normalized plots of 3000 images captured at 125 Hz. Water flows from the left while the laser enters the reactor from the right. Since the sleeve partially blocked the laser light, there was a sharp gradient in dye concentration (from green to blue) at the edges of the sleeve where the dye concentrations should be higher than it appears. Values below 0.2 were cutoff (indicated as white) to show the extent of dye dispersion. Dye injected 22.9 cm upstream of the sleeve showed that the highest dye concentrations followed a path close to the wall with very little dye entrained into the wake region. Dye dispersion was limited such that only about 30% of the initial dye concentration was detected at the top surface of the sleeve and less than 20% of the initial dye concentration immediately upstream of the sleeve. Dye injected 32.4 cm upstream of the sleeve established that the highest dye concentrations were found at the wall upstream of the sleeve. However, due to the increased distance of travel, greater dye dispersion was observed, such that 35e40% of the initial dye concentration was observed immediately upstream and downstream of the sleeve. Hence, by increasing the distance of travel upstream of the lamp array, the paths that led to short-circuiting in UV reactors became limited since more dye entered the region closer to the lamp surface. Sozzi and Taghipour (2006) and Moreira et al. (2007) analyzed annular reactors with inlets parallel and perpendicular to lamp axis,
3860
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 5 5 e3 8 6 2
3.3.
Outlet section
Mixing of the tracer dye across the reactor height at the wake and the outlet as a function of time was also examined. The degree of mixing can be inferred from a statistical analysis of the tracer concentration over the spatial and temporal variation by computing the coefficient of variation (COV) defined as:
COV ¼
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 12 0 u u X u 1 1 X @ t C x; yi ; zj ; t A C x; yi ; zj ; t n1 n i;j i;j 1 X C x; yi ; zj ; t n i;j
(1)
Fig. 5 e LIF images around the sleeve showing the extent of dye dispersion versus distance of injection at (a) 22.9 cm and (b) 32.4 cm upstream of the lamp.
where n ¼ number of data samples at a given time (n ¼ 1000), C ¼ dye concentration, i and j ¼ indices of elements in the y and z direction and t ¼ time. A lower value of COV indicates a more uniform distribution of tracer concentration or greater mixing across a given section. The temporal variations in COV for the different injection heights at the vertical wake section (Fig. 6a) suggested that as the dye was injected further away from the center, the COV increased, implying a non-uniform tracer distribution. This difference in mixing is due to the greater mixing that occurred in the recirculation zone for the center injection. The COV as a function of time at the outlet (Fig. 6b) showed the same upward trend for the various injection heights as observed at the wake region. In addition, the standard deviation of the COV from the mean, denoted next to each average line (shown in green), increased with the injection
and suggested that rearranging the inlet altered the velocity field, thus having a significant impact on the UV dose distribution and the reactor performance. Moreira et al. (2007) further examined a reactor with three inlets placed perpendicular to the lamp axis and found that when operated simultaneously, they behaved like a plug-flow with limited axial dispersion and was more efficient than a reactor with a single perpendicular to lamp inlet. Chiu et al. (1999b) established that particles that entered close to the wall had a higher probability of experiencing a low UV dose due to higher velocities and distance from the lamp. In order to reduce short-circuiting for microorganisms, the length of the approach section leading to the lamp array might be increased such that microbes entering the reactor at a specific location (e.g. close to the wall) would have an equal probability of traversing regions close to the lamp where UV intensities are higher. Zhao et al. (2009) showed that a longer straight pipe inlet placed perpendicular to an annular reactor shifted the peak of the dose distribution to higher values and increased the reduction equivalent dose (RED) predicted using microspheres. Based on the results of this study, a minimum straight inlet channel length of at least 30 cm or about 11 sleeve diameters is required to significantly increase the probability of microbes passing through higher UV intensity zones. This, however, depends on the upstream hydraulics such as the flow rate and the velocity profile which defer between reactors. The results show that 3DLIF can be utilized to optimize the length of the approach section to help improve the performance of UV reactors.
Fig. 6 e Changes in COV of dye concentration through the cross section of the (a) wake region and the (b) outlet over time.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 5 5 e3 8 6 2
height. For the off-center injections, the larger fluctuations in the outlet dye concentration would have a direct effect on the biodosimetry results and consequently the RED values. To further quantify these fluctuations, the standard deviation, s, (eq. (3)) of the tracer concentration at various vertical line sections along the length of the reactor was computed (Fig. 7), where N ¼ number of samples over time (1000 at each section) and Ci and C ¼ instantaneous and time-averaged concentrations. The standard deviation at a point in each vertical section represents the deviation in the concentration compared to other values along that section. As such, a low standard deviation indicates a lower variation in the concentration about the mean at that point, i.e., greater mixing. vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi u N u 1 X ðCi CÞ2 s¼t N 1 i1
(2)
As expected, the center injection revealed a low standard deviation in the wake region, confirming prior observations of increased mixing compared to the off-center injections. Analyzing section (i), where the flow separates around the sleeve, the center injection case showed two peaks around the lamp that corresponded to the movement of the dye around the cylinder. A smaller peak in the wake region was observed indicating the dye was well-mixed temporally as opposed to 1.3 cm injection, where a larger peak indicated lower mixing. The off-center injection cases also showed the presence of a relatively large peak at the dye injection height. At section (ii), the center injection showed two symmetrical peaks, which occurred due to the origination of the von Karman vortex streets. Furthermore, the region between the peaks
i ii Center
iii
i
1.0
ii
iii
0.8 0.6 0.4 0.2 0.0
0
5
10
15 20 25 0
5
10
15 20 25 0
5
10
15
20 25
5
10
15 20 25 0
5
10
15
20 25
5
10 15
5
10
15 20 25
1.3 cm off-center 1.0
h/H
0.8 0.6 0.4 0.2 0.0
0
5
10
15 20 25 0
2.5 cm off-center 1.0 0.8 0.6 0.4 0.2 0.0
0
5
10 15
20 25 0
20 25 0
Standard Deviation Fig. 7 e Standard deviation of tracer dye concentration at multiple line sections in the UV reactor comparing dye mixing for the various injection heights.
3861
showed a higher standard deviation than in the wake region, indicating relatively lower mixing. For the 1.3 cm injection, the standard deviation became more uniform in the central region of the reactor while the large peak close to the wall in the 2.5 cm injection indicated the tracer dye continued traveling the short-circuited paths. The outlet profile for the center injection was uniform across the reactor height. A relatively flat standard deviation profile across the reactor height indicated uniform mixing. Ignoring wall effects, i.e., where the standard deviation decreased fairly quickly close to the wall, the standard deviations were within 10% of the mean. However, the standard deviation profiles at the outlet for the off-center injection cases had a greater variation across the reactor height compared to the center injection. A few inactivated microbial samples collected at the outlet in biodosimetry would not be sufficient to capture the fluctuations observed causing deviations in dose estimations. Therefore, a minimum reactor length downstream of the final lamp array has to be defined where the fluctuations across the outlet is found to be fairly uniform.
4.
Conclusions
This study applied the 3DLIF technique for the first time to visualize and quantitatively analyze the flow across a UV lamp in a model reactor used for drinking water treatment. In addition to three-dimensional mixing, the technique successfully visualized the two-dimensional, transient mixing behaviors within the reactor, which has not been possible with traditional tracer test techniques. It is also noteworthy that the 3DLIF technique is non-intrusive, i.e., there is no disturbance in the flow due to the placement of the sampling probe. Tracer tests revealed unsteady turbulent flow characteristics such as the recirculation zone and the von Karman vortex street that are normally observed in flows around cylinders. The length of the recirculation zone and the Strouhal number were found to increase in the confined flow compared to an infinite medium. Paths away from the center height, characterized by higher velocities and less influence of the cylinder, were also analyzed. The results demonstrated that a major difference in these pathways was the decreasing amount of dye entering the recirculation zone, which has a higher residence time, as the injection height increased. The results also suggested that a longer approach length was beneficial to increase the probability of microbes entering the region around the lamp sleeve irrespective of their entrance height into the reactor. Lastly, the length of the outlet i.e., distance from the last lamp array to the reactor exit, was examined as mixing at the outlet was determined to drastically vary over time with an increase in injection height. A wellmixed outlet, i.e., when the concentration profile across the reactor height is within 10% as in the center injection case, would be desirable to improve the accuracy of the biodosimetry results. These inlet and outlet analyses were the first of its kind and aid in the optimization of the reactor design. Numerous studies that experimentally validate the hydraulics and predict spatial dose distributions in UV reactors have employed RANS based CFD simulations. However, discrepancies between the model’s calculations and the
3862
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 5 5 e3 8 6 2
actual measurements, which are mainly caused by turbulence modeling, occur near the wake region of the lamp (Liu et al., 2007; Wols et al., 2010a, 2010b). For a better prediction of the time dependent turbulent flows and UV dose distributions, advanced models such as large eddy simulation (LES) need to be used, evident from the vast difference in the instantaneous center injection flow (Fig. 3a), as opposed to the time-averaged flow (Fig. 3b). In addition, results from Wols et al. (2010a) validate the above findings that LES will result in a more physical and accurate representation of the flow velocities and the transport and mixing within UV reactors which will likely lead to a more accurate prediction of the UV dose received by microorganisms. Therefore 3DLIF is a powerful tool to fully capture the flow in UV reactors that can aid in the validation of LES results and provide information on the temporal dose distributions that has not been considered in past studies.
Acknowledgments This research was partially funded by Water Research Foundation (Project No. 4134) and Korea Water Resources Corporation (Kwater).
Appendix. Supplementary material Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.watres.2011.04.041.
references
Alpert, S.M., Knappe, D.R.U., Ducoste, J.J., 2010. Modeling the UV/ hydrogen peroxide advanced oxidation process using computational fluid dynamics. Water Research 44 (6), 1797e1808. Bellar, T.A., Lichtenberg, J.J., Kroner, R.C., 1974. Occurrence of organohalides in chlorinated drinking waters. Journal American Water Works Association 66 (12), 703e706. Blatchley, E.R., Do-Quang, Z., Janex, M.L., Laine, J.M., 1998. Process modeling of ultraviolet disinfection. Water Science and Technology 38 (6), 63e69. Braza, M., Perrin, R., Hoarau, Y., 2006. Turbulence properties in the cylinder wake at high Reynolds numbers. Journal of Fluids and Structures 22 (6e7), 757e771. Chiu, K.P., Lyn, D.A., Savoye, P., Blatchley, E.R., 1999a. Effect of UV system modifications on disinfection performance. Journal of Environmental Engineering-ASCE 125 (5), 459e469. Chiu, K., Lyn, D.A., Savoye, P., Blatchley, E.R., 1999b. Integrated UV disinfection model based on particle tracking. Journal of Environmental Engineering-ASCE 125 (1), 7e16. Clancy, J.L., Hargy, T.M., Marshall, M.M., Dyksen, J.E., 1998. UV light inactivation of Cryptosporidium oocysts. Journal American Water Works Association 90 (9), 92e102.
Glaze, W.H., Weinberg, H.S., Cavanagh, J.E., 1993. Evaluating the formation of brominated DBPs during ozonation. Journal American Water Works Association 85 (1), 96e103. Guiraud, P., Bertrand, J., Costes, J., 1991. Laser measurements of local velocity and concentration in a turbulent jet-stirred tubular reactor. Chemical Engineering Science 46 (5e6), 1289e1297. Janex, M.L., Savoye, P., Do-Quang, Z., Blatchley, E., Laine, J.M., 1998. Impact of water quality and reactor hydrodynamics on wastewater disinfection by UV, use of CFD modeling for performance optimization. Water Science and Technology 38 (6), 71e78. Kim, D., Nemlioglu, S., Roberts, P.J.W., Kim, J.H., 2010. Ozonecontactor flow visualization and quantification using threedimensional laser-induced fluorescence. Journal American Water Works Association 102 (1), 90e99. Lawryshyn, Y.A., Cairns, B., 2003. UV disinfection of water: the need for UV reactor validation. Water Science and Technology: Water Supply 3 (4), 293e300. Linden, K.G., Shin, G.A., Faubert, G., Cairns, W., Sobsey, M.D., 2002. UV disinfection of Giardia lamblia cysts in water. Environmental Science & Technology 36 (11), 2519e2522. Liu, D., Wu, C., Linden, K., Ducoste, J., 2007. Numerical simulation of UV disinfection reactors: evaluation of alternative turbulence models. Applied Mathematical Modelling 31 (9), 1753e1769. Moreira, R.M., Pinto, A.M.F., Mesnier, R., Leclerc, J.P., 2007. Influence of inlet positions on the flow behavior inside a photoreactor using radiotracers and colored tracer investigations. Applied Radiation and Isotopes 65 (4), 419e427. Price, S.J., Sumner, D., Smith, J.G., Leong, K., Paidoussis, M.P., 2002. Flow visualization around a circular cylinder near to a plane wall. Journal of Fluids and Structures 16 (2), 175e191. Rehimi, F., Aloui, F., Ben Nasrallah, S., Doubliez, L., Legrand, J., 2008. Experimental investigation of a confined flow downstream of a circular cylinder centred between two parallel walls. Journal of Fluids and Structures 24 (6), 855e882. Singha, S., Sinhamahapatra, K.P., 2010. Flow past a circular cylinder between parallel walls at low Reynolds numbers. Ocean Engineering 37 (8e9), 757e769. Sozzi, D.A., Taghipour, F., 2006. UV reactor performance modeling by Eulerian and Lagrangian methods. Environmental Science & Technology 40 (5), 1609e1615. Tian, X.D., Roberts, P.J.W., 2003. A 3DLIF system for turbulent buoyant jet flows. Experiments in Fluids 35 (6), 636e647. Williamson, C.H.K., 1996. Vortex dynamics in the cylinder wake. Annual Review of Fluid Mechanics 28, 477e539. Wols, B.A., Uijttewaal, W.S.J., Hofman, J., Rietveld, L.C., van Dijk, J.C., 2010a. The weaknesses of a k-epsilon model compared to a large-eddy simulation for the prediction of UV dose distributions and disinfection. Chemical Engineering Journal 162 (2), 528e536. Wols, B.A., Shao, L., Uijttewaal, W.S.J., Hofman, J.A.M.H., Rietveld, L.C., van Dijk, J.C., 2010b. Evaluation of experimental techniques to validate numerical computations of the hydraulics inside a UV bench-scale reactor. Chemical Engineering Science 65 (15), 4491e4502. Zdravkovich, M.M., 1997. Flow Around Circular Cylinders. In: Fundamentals, vol. 1. Oxford Science Publications, New York. Zhao, X., Alpert, S.M., Ducoste, J.J., 2009. Assessing the impact of upstream hydraulics on the dose distribution of ultraviolet reactors using fluorescence microspheres and computational fluid dynamics. Environmental Engineering Science 26 (5), 947e959.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 6 3 e3 8 7 0
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Heavy metal removal in anaerobic semi-continuous stirred tank reactors by a consortium of sulfate-reducing bacteria Hoa T.Q. Kieu*, Elizabeth Mu¨ller, Harald Horn Institute of Water Quality Control, Technische Universita¨t Mu¨nchen, Am Coulombwall, 85748 Garching, Germany
article info
abstract
Article history:
Removal of heavy metals by an enriched consortium of sulfate-reducing bacteria (SRB) was
Received 9 February 2011
evaluated through the abundance of SRB, sulfate reduction, sulfide production and heavy
Received in revised form
metal precipitation. Five parallel anaerobic semi-continuous stirred tank reactors (CSTR,
17 April 2011
V ¼ 2 L) (referred as R1eR5) were fed with synthetic wastewater containing mixtures of
Accepted 23 April 2011
Cu2þ, Zn2þ, Ni2þ, and Cr6þ in the concentrations of 30, 60, 90, 120, and 150 mg L1 of each
Available online 11 May 2011
metal and operated with a hydraulic retention time of 20 days for 12 weeks. The loading rates of each metal in R1eR5 were 1.5, 3, 4.5, 6, and 7.5 mg L1 d1, respectively. The results
Keywords:
showed that there was no inhibition of SRB growth and that heavy metal removal effi-
Sulfate-reducing bacteria (SRB)
ciencies of 94e100% for Cu2þ, Zn2þ, Ni2þ, and Cr6þ were achieved in R1eR3 throughout the
Heavy metal removal
experiment and in R4 during the first 8 weeks. The toxic effect of heavy metals on the SRB
Continuous stirred tank reactor
consortium was revealed in R5, in which no SRB could survive and almost no heavy metal
(CSTR)
precipitation was detected after four weeks of operation.
Fluorescent in situ hybridization
ª 2011 Elsevier Ltd. All rights reserved.
(FISH)
1.
Introduction
The environmental pollution caused by wastewaters containing high concentrations of dissolved heavy metals and low pH from mining and industrial processing (e.g. metallurgical, electronic, electroplating and metal finishing industries) negatively impacts to living organisms as well as humans. The toxic effects of heavy metals include ion displacement and/or substitution of essential ions from cellular sites and blocking of functional groups of important molecules, e.g. enzymes, polynucleotides, and essential nutrient transport systems. This results in denaturation and inactivation of enzymes and disruption of cell organelle membrane integrity, as well as damage to the structure of DNA, nerves, livers and bones (Sani et al., 2001). Unlike organic contaminants, which can be
degraded into harmless chemical species, heavy metals cannot be degraded. However, they can be transformed from mobile and toxic forms into their stable immobile and less toxic forms (Uhrie et al., 1996; Beyenal and Lewandowski, 2004). Many methods have been used for treatment of heavy metal contaminated wastewaters. Among them, the classical physicochemical methods were widely applied (e.g. precipitation, absorption, ion exchange and complex formation). Despite effective treatment, these methods are expensive and generate large amounts of residual sludge (Gallegos-Garcia et al., 2009; Tekerlekopoulou et al., 2010). Therefore, immobilization of heavy metals through microbial mediated reduction and precipitation is now of considerable interest. Especially, metal sulfide precipitation by SRB has promise as an attractive alternative over physico-chemical and other
Abbreviations: SRB, sulfate-reducing bacteria; CSTR, continuous stirred tank reactor; FISH, fluorescent in situ hybridization. * Corresponding author. Tel.: þ 49 0 89 289 13714; fax: þ 49 0 89 289 13718. E-mail addresses:
[email protected],
[email protected] (H.T.Q. Kieu). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.04.043
3864
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 6 3 e3 8 7 0
methods. This method involves two stages: (i) Sulfatereducing bacteria (SRB), under anaerobic conditions, oxidize simple organic compounds (e.g. lactate, acetate, propionate, butyrate, etc.) by utilizing sulfate as an electron acceptor and generating hydrogen sulfide and bicarbonate ion (Eq. (1)), and (ii) the biologically produced hydrogen sulfide reacts with dissolved heavy metals such as Cu, Zn, and Ni to form insoluble metal sulfide precipitates (Eq. (2) (Hulshof et al., 2006; Neculita et al., 2007). 2CH2 O þ SO2 4 /H2 S þ 2HCO3
(1)
Me2þ þ H2 S/MeSY þ 2Hþ ðMe ¼ MetalÞ
(2)
3þ
SRB are able to reduce indirectly Cr via the production of H2S (Eq. (3)) (Chang and Kim, 2007; Neculita et al., 2007) and by a direct enzymatic process involving hydrogenases and c3 cytochromes (Lovley and Phillips, 1994; Goulhen et al., 2006). 3HS þ 2Cr6þ /3So þ 2Cr3þ Y þ 3Hþ
(3)
In comparison with chemical precipitation of metals as hydroxides or oxyhydroxides heavy metal precipitation by using biological sulfide has advantages which include low amounts of residual sludge and more important a lower solubility of the produced sulfides even at acidic pH, highly reactive efficiency, and cost effectiveness (Tsukamoto et al., 2004; Gallegos-Garcia et al., 2009). This method not only eliminates toxic heavy metals from heavy metal contaminated wastewater, but also enables the recovery of valuable metals as metallic sulfide (Gallegos-Garcia et al., 2009). The metal resistance of microorganisms varies with the species by developing a variety of specific resistance mechanisms such as metal exclusion by permeability barrier, active transport of the metal away from the cell, intracellular sequestration of the metal by protein binding, extracellular sequestration, enzymatic detoxification of the metal to a less toxic form, and reduction in metal sensitivity of cellular targets (Bruins et al., 2000). In addition, the ability of heavy metal resistance of organisms is also dependent on the mobility, bioavailability, and toxicological effect of each heavy metal. Heavy metal treatment using anaerobic sulfate reduction is influenced by a variety of parameters such as electron donor, pH, temperature, sulfate concentration, and heavy metal species. Therefore, the use of an active treatment method as a sulfidogenic bioreactor for heavy metal removal is probably preferable to passive treatment. The advantage of this method is the ease of control during the treatment process resulting in the permanent removal of heavy metals as compared to passive process. In recent years, heavy metal removal by sulfate reduction has been studied in various bioreactor-types such as continuously stirred tank reactor (CSTR) (Nagpal et al., 2000; Sahinkaya, 2009), upflow anaerobic sludge blanket reactor (UASB) (Lenz et al., 2008), off-line sulfidogenic bioreactors (Goncalves et al., 2007), fixed bed reactor (FBR) (Kaksonen et al., 2003; Viggi et al., 2010), and permeable reactive barriers (PRB) (Bartzas et al., 2006). The aim of the present study was to investigate the heavy metal removal efficiency of selected SRB consortium in five
parallel semi-continuous stirred tank reactors spiked with different concentrations of heavy metal mixtures (Cu2þ, Ni2þ, Zn2þ, and Cr6þ).
2.
Materials and methods
2.1.
Inoculum
A heavy metal tolerant SRB consortium obtained from heavy metal contaminated sediment in Tong Xa, a settlement known for bronze, iron casting and electroplating in Nam Dinh, Vietnam was used as an inoculum. Element composition measured in the water phase of the sample used as the inoculum in this study comprised of 82 mg L1 total Cr, 64 mg L1 Ni2þ, 76 mg L1 total Fe, 18.5 mg L1 Zn2þ, 38 mg L1 Cu2þ, 4 mg L1 Mn2þ; and 202 mg L1 SO42-. This culture was cultivated under anaerobic condition using Postgate’s medium B (Postgate, 1984). All procedures during preparation of the medium and cultivation were performed according to the modified Hungate’s method for anaerobes (Miller and Wolin, 1974). To enrich the bacteria number the cultivation step was repeated three times before inoculating the bioreactors.
2.2.
Experimental set-up
The schematic diagram of semi-continuous stirred tank reactor is present in Fig. 1
2.2.1.
Bioreactors
The experiments were carried out in five glass anaerobic semicontinuous stirred tank reactors (CSTR, V ¼ 2 L). This reactor type was selected because of making the concentration uniform throughout the reactor maintains the culture in a constant average physiological state. The reactors were kept at 30 C in a heated water bath and mixed by magnetic stirrers at a speed of 400 rpm. All reactors were soaked in a 3 M HNO3 solution for 72 h and rinsed with de-ionized before use to avoid metal contamination.
2.2.2.
Feeding tanks
Synthetic wastewater (see below) was prepared aseptically every week to avoid the contamination and then fed continuously at the top of the bioreactors by a peristaltic pump (Ismatec SA, Zuerich, Switzerland) with a volumetric flow rate of 100 mL d1. To maintain the anaerobic condition, feeding tanks were purged with filter sterilized nitrogen gas (0.22 mm). Gas produced during the treatment process was trapped by gas collection bags.
2.2.3.
Synthetic wastewater composition
Most heavy metal contaminated wastewater contains low concentration of sulfate ions and relatively little dissolved organic carbon. Thus, synthetic wastewater was used to make the heavy metal removal process efficient by supplementing an adequate carbon and energy source for SRB. The composition of the synthetic wastewater was prepared as follows: KH2PO4, 0.5 g L1; NH4Cl, 1.0 g L1; Na2SO4, 3.7 g L1; Sodium lactate, 4.42 g L1; and Trisodium citrate, 0.3 g L1. The
3865
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 6 3 e3 8 7 0
Fig. 1 e Schematic diagram of semi-continuous stirred tank reactor.
synthetic wastewater was not supplemented with Fe2þ and reducing agents such as ascorbic acid, sodium thioglycolate, Na2S to allow the valuation of the precipitation of other metals under investigation. Trisodium citrate was added to prevent any initial metal precipitation. The pH was adjusted to 6.0 0.2 using HCl and NaOH. For continuous operation, synthetic wastewater was spiked with a mixture of Cu2þ, Ni2þ, Zn2þ, and Cr6þ in different concentrations as shown in Table 1. Solution for each heavy metal was prepared from the dissolution of chloride salt for Cu2þ, Ni2þ, Zn2þ and potassium dichromate (K2Cr2O7) for Cr6þ and then sterilized by membrane filtration (0.22 mm).
as electron acceptor and donor, respectively, without heavy metals. After pre-incubation (start of the experiment), five reactors were fed continuously with synthetic wastewater containing mixtures of Cu2þ, Ni2þ, Zn2þ, and Cr6þ in the influent concentrations of 30, 60, 90, 120 and 150 mg L1 each, respectively. Reactor 1 (R1) was spiked with the loading rate of 1.5 mg L1 d1 each metal. The loading rate of each metal was increased up to 3, 4.5, 6 and 7.5 mg L1 d1 for reactor 2 (R2), 3 (R3), 4 (R4), and 5 (R5), respectively. The reactors were operated semi-continuously with a hydraulic retention time (HRT) of 20 days for 12 weeks, with the exception of R5 which was only operated for 6 weeks (Table 1).
2.3.
2.4.
Experimental procedure
The bioreactors were first inoculated with 10% (v/v) of the enriched SRB consortium containing 1 108 cells mL1 and incubated for 9 days at batch operating conditions using synthetic wastewater containing sulfate and sodium lactate
Analytical methods
Samples were taken weekly to measure sulfate, dissolve sulfide, dissolved heavy metals, and pH according to standard methods (APHA, 1998). Dissolved sulfide and pH were immediately measured after collection. For sulfate and dissolved
Table 1 e Reactor operating characteristics. Reactor
Conc. of each heavy metala spiked in synthetic wastewater (mg L1)
Loading rate of each heavy metala (mg L1 d1)
Total time of operation (weeks)
30 60 90 120 150
1.5 3.0 4.5 6.0 7.5
12 12 12 12 6
R1 R2 R3 R4 R5 a Cu2þ, Zn2þ, Ni2þ and Cr6þ.
3866
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 6 3 e3 8 7 0
heavy metal analyses the samples were filtered through 0.45 mm cellulose nitrate - membrane filters (Germany) before measuring. Dissolved sulfide, sulfate and Cr6þ were analyzed by using Dr. Lange kit LCK-653, LCK-153, and LCK- 313, respectively. For the Dr. Lange analyses, a spectrophotometer Dr. 2800 was used (Hach Lange GMBH, Germany). The concentrations of Cu2þ, Zn2þ, and Ni2þ were determined by an atomic absorption spectrophotometer (AAS) after acidifying with concentrated nitric acid (pH < 2) to prevent metal precipitation and adsorption to surfaces.
2.4.1.
EDS analysis
Qualitative analysis of precipitates was realized by energy dispersive spectrometry (EDS) analysis using a Leica/Cambridge, model StereoScan 360. The precipitates were obtained after filtering the effluent sample of R2 (3 mg L1 d1) through 0.45 mm cellulose nitrate - membrane filters (Germany). The filter paper was dried in an oven for 2 h at 105 C and then thin coated by Au for EDS analysis.
2.4.2. Total cell counts and fluorescent in situ hybridization (FISH) The abundance of SRB was estimated by total cell count using 40 , 60 diamidino-2-phenylindole (DAPI) staining and FISH technique using 16S rRNA-target oligonucleotide probe (SRB385) labeled at the 50 end with indocarbocyanine (Cy3) reactive fluorescent dye. All procedures such as fixation of samples by paraformaldehyde and EtOH, DAPI staining, drying and washing were performed using modified standard procedure described previously (Amann et al., 1995). Hybridized and DAPI stained cells were visually detected by using Axioplan epifluorescence microscope (Carl Zeiss, Jena, Germany). Samples of R1-R4 were collected at the start, after two, six, and 11 weeks and of R5 at the start, after one and four weeks and immediately fixed with 4% (w/v) paraformaldehyde (PFA) and stored at 20 C until analysis.
3.
Results
The obtained results showed that Cu2þ, Zn2þ, Ni2þ, and Cr6þ were removed effectively throughout the experiment in R1, R2, and R3 with the heavy metal loading rate of 1.5, 3, and 4.5 mg L1 d1, respectively. The behavior of the SRB consortium in these three reactors was similar and marked by high levels of heavy metal removal, sulfate reduction, and sulfide production. However, the inhibition of the SRB consortium began to be detected in R4 (6 mg L1 d1) from week 9 onwards. A toxic effect on the SRB consortium was observed in R5 (7.5 mg L1 d1). No growth of SRB and almost no heavy metal precipitation were detected after four weeks of the experiment. Therefore, the results of sulfate reduction, sulfide production and heavy metal removal obtained from R2, R4, and R5 were shown representatively in Fig. 2.
3.1.
Heavy metal removal
Heavy metal removal efficiencies of 96e100% for Cu2þ, 94e100% for Zn2þ and Ni2þ, and 96e100% for Cr6þ were achieved in R1 (1.5 mg L1 d1), R2 (3 mg L1 d1), and R3 (4.5 mg L1 d1) during 12 weeks of operation. The final heavy
metal concentrations for all experiments are summarized in Table 2. Furthermore, the results for R2, R4 and R5 are shown in Fig. 2. Although Cu2þ, Zn2þ, Ni2þ, and Cr6þ were also removed effectively (98e100%) in R4 (6 mg L1 d1) during the first 8 weeks, the decrease in heavy metal removal was observed from week 9 (91e97%). The removal of heavy metals in R5 (7.5 mg L1 d1) was significantly lower than the 98e100% achieved during the first week to 78e91% at the second week and no heavy metal precipitation was detected in the effluent of R5 after four weeks of operation.
3.2. Effect of heavy metals on sulfate reduction and sulfide production Table 2 displays the final concentrations achieved in all five reactors at the end of operation. The production of sulfide correlates nicely with the removal of heavy metals. Sulfate was converted by about 43e67% of initial concentration to dissolved sulfide of 145e310 mg L1 in R1- R3 (1.5e4.5 mg L1 d1) throughout the experiment (Fig. 2b). In R4 (6 mg L1 d1) this high sulfide production could only be observed within the first 8 weeks of experiment. After that time point only 17e39% of initial sulfate concentration was reduced in R4 (Fig. 2d). This led to a gradual decrease of dissolved sulfide in R4 with a sulfide concentration of 26 mg L1 at the end of experiment (Fig. 2d). In R5 only 33% of the initial sulfate concentration was reduced (7.5 mg L1 d1) in the first week. At the end of experiment R5 nearly no sulfate reducing activity has been observed (see Fig. 2f). The results indicated that the efficiency of sulfate reduction and sulfide production greatly decreased with the increase of heavy metal loading rate up to 6 (R4) and 7.5 mg L1 d1 (R5).
3.3.
Sulfur balance
A sulfur balance was estimated from an average value of the obtained results of R2 (heavy metal loading 3 mg l1 d1) throughout 12 weeks of operation. The balance was only done for one reactor to check the general plausibility of the measured sulfate and sulfide concentrations. The balance was based on elemental sulfur (S) by assuming that the total initial sulfur as 1 sulfate ðSO2 4 Þ (w762 mg L ) was converted partly to total dis2 solved sulfide (S þ HS þ dissolved H2S). Therefore, the total final sulfur existing in R2 effluent includes unconsumed sulfate (w305 mg L1) and generated sulfide (w269 mg L1). The percent ratio of total final sulfur to total initial sulfur was only 75%. The difference between total initial and final sulfur might be explained as follows: (i) The loss of w48 mg L1 (w6.2%) of sulfur for the precipitation of Cu2þ, Zn2þ, Ni2þ, and Cr6þ in the R2 (ii) The loss of sulfur as volatile sulfide through air oxidation in transferring samples and diffusion of H2S gas.
3.4. Effect of heavy metals on sulfate-reducing bacteria population The hybridized positive cells detected by FISH using the specific probe for SRB (SRB385) and total DAPI stained cells of the enriched sulfidogenic consortium were estimated in all five reactors. The three representative reactors R2, R4, and R5 are shown in Fig. 3. In general the abundance of SRB decreased with increasing heavy metal load. The ratio of SRB detected by FISH to
3867
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 6 3 e3 8 7 0
a
b
2 1.8
Cu2+
Zn2+
1.6
Ni2+
Cr6+
350
2500
300
2000
250
1.4 1.2
200
1500
0.8
150
1000
0.6
100
1
Sulfide Sulfate
0.4
50
0.2 0
0 0
c
500
1
2
3
4
5 6 7 Time (weeks)
8
9
10
11
12
0 Inf. 0
1
2
3
4
5
6
7
8
9
10 11 12
Time (weeks)
d
5
Cu2+ Ni2+
4.5 4
Zn2+ Cr6+
2500
300 250
3.5
2000
200 1500
3
150
2.5
1000
2
100
1.5 1
Sulfide Sulfate
50
0.5 0
500
0
0 0
1
2
3
4
5
6
7
8
9
10
11
12
Inf. 0
1
2
3
Time (weeks)
e
Cu2+
Zn2+
Ni2+
Cr6+
5
6
7
8
9
10 11 12
Time (weeks)
160 140
4
f
300
120
250
100
200
2500
2000
1500
80
150
60
1000 100
40
Sulfide Sulfate
50
20
500
0
0
0
1
2
3
4
5
6
0 Inf.
0
1
2
3
4
5
6
Time (weeks)
Time (weeks)
Fig. 2 e Residual concentrations of heavy metals (Cu2D, Zn2D, Ni2D, and Cr6D), sulfate, and sulfide in R2 (3 mg LL1 dL1), R4 (6 mg LL1 dL1) and R5 (7.5 mg LL1 dL1) with time.
the total cell counts was 58% at the start. In R2 the abundance ratio of SRB increased from 58% at the start to 83e84%. In R3 a similar behavior has been detected. The SRB achieved a fraction of 83e86% at the end of the experiment. As can already been observed from the heavy metal concentrations (see Fig. 2c) the abundance ratio in R4 shifted from 58% at the start to 79% after 2 weeks, 63% after 6 and only 14% after 11 weeks of operation. Obviously R4 has been operated with transient heavy metal load which more and more inhibits the activity of SRB. R5 finally shows an immediate inhibition. The SRB abundance significantly decreased from 58% at the start to 30% after one week; no positive hybridized cell was detected by specific probe for SRB (SRB385) after four weeks of operation. In addition, a distinct biomass loss through the significant
decrease of the total cell counts from 1.3 108 cell mL1 at the start to 3.5 104 cell mL1 after four weeks of experiment was observed in R5. FISH results are consistent with the obtained results of sulfate reduction, dissolve sulfide production and heavy metal removal, indicated that SRB played a key role in heavy metal removal.
3.5.
Effect of heavy metals on pH value
The effects of heavy metals on the SRB consortium were also observed by the change of pH value (Fig. 4). The pH increased from an initial value of 6 in the influent to more than 7 at the start of continuous operation (after 9 day pre-incubation) and in the effluent of R1eR3 (1.5e4.5 mg L1 d1) throughout the
3868
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 6 3 e3 8 7 0
Table 2 e Dissolved metal concentrations in the effluent at the end of the experiments.
7.6
Reactor
7.2 pH value
7.4
1
1
experiment as well as in the effluent of R4 (6 mg L d ) during the first seven weeks. However, the effluent pH of R4 dropped gradually to between 6.91 and 6.58 toward the end of the experiment. The effluent pHs of R5 (7.5 mg L1 d1) were more than 7 at only during the first two weeks and dropped gradually to 6.88 after three weeks and to 6.55 after six weeks of operation.
3.6.
Qualitative EDS analysis
Qualitative EDS analysis of the precipitate of R2 (3 mg L1 d1) experiment was performed. The precipitates as CuS, ZnS, and NiS were confirmed by the EDS spectrum with strong peaks of copper, zinc, nickel and sulfur shown in Fig. 5.
4.
Discussion
The impact of heavy metals on microorganisms is classified as toxic (causing death) and/or inhibitory (causing a reduction in metabolic activity). Toxic concentration is the lowest initial dissolved metal concentrations at which no bacterial growth 1.00E+09 1.00E+08 1.00E+07
Cells mL
-1
1.00E+06
R1 R2 R3 R4 R5
6.6 6.4 6.2 6
12
280 281 265 26.3 0
11
834 917 902 1860 2200
9
0.1 0.24 0.55 1.25 135
10
0.36 0.4 0.5 3.5 138
8
0.3 0.3 0.6 1.9 141
7
<0.2 0.26 0.25 1.3 135
6.8
6
S2
5
SO42
4
Cr6þ
3
Ni2þ
2
Zn2þ
1
Cu2þ
7
In f. st ar t
R1 R2 R3 R4 R5
Dissolved heavy metal, sulfate and sulfide concentrations in the effluent of R1-R4 after 12 weeks and of R5 after 6 weeks (mg L1)
Time (weeks)
Fig. 4 e The changes of pH value in the effluent of R1-R5 (1.5e7.5 mg LL1 dL1) with time.
is observed. The obtained results of this study showed that heavy metal mixtures (Cu2þ, Zn2þ, Ni2þ and Cr6þ) were removed effectively (94e100%) without any inhibition to the growth of SRB throughout the experiment in R1eR3 with heavy metal loading rate ranging from 1.5 to 4.5 mg L1 d1. However, SRB abundance and heavy metal removal began to decrease at heavy loading rate of 6 mg L1 d1 after 8 weeks of experiment (91e97%). A toxic effect on the growth of SRB that resulted in failure of hybridized positive cell detection and heavy metal precipitation was observed at a heavy metal loading rate of 7.5 mg L1 d1 (R5). The heavy metal loading rate of 6 mg L1 d1 (R4) and 7.5 mg L1 d1 (R5) therefore are considered to be the inhibitory and toxic concentrations to the SRB consortium in the present study, respectively. The toxic concentrations of heavy metals to single SRB species or SRB consortia ranging from a few mg L1 to as much as one hundred mg L1 were reported by other authors. A Cu2þ concentration of 1.92 mg L1 causing death of organism was shown by Sani et al., 2001 who used a single SRB species (Desulfovibrio desulfuricans) and a specific medium to prevent metal ions from abiotic precipitation. Lack of growth of SRB consortium at 12 mg L1 of Cu2þ and 20 mg L1 of Zn2þ was shown by Utgikar et al., 2001. Hao et al., 1994 reported that the toxic concentrations of individual heavy metal for SRB
1.00E+05 1.00E+04 1.00E+03 1.00E+02 1.00E+01
DAPI SRB385
1.00E+00 Start R2-2 R2-6 R2-11 R4-2 R4-6 R4-11 R5-1 R5-4 Samples
Fig. 3 e Relative number of positive hybridized cells detected by specific probe for sulfate-reducing bacteria (SRB385) and total DAPI-stained cells in three representative anaerobic semi-continuous bioreactors (referred as R2, R4, and R5) with heavy metal (Cu2D, Ni2D, Zn2D, and Cr6D) loading rate of 3, 6, and 7.5 mg LL1 dL1, respectively. Samples of R2 and R4 were analyzed after L2, L6, and L11 weeks and of R5 after L1 and L4 weeks of operation. Cell counts are on a logarithmic scale.
Fig. 5 e EDS spectrum of the precipitate of R2 (3 mg LL1 dL1) experiment.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 6 3 e3 8 7 0
consortium in batch test were 20 mg L1 Cd2þ, 20 mg L1 Cu2þ, 20 mg L1 Ni2þ, 25 mg L1 Zn2þ, and 60 mg L1 Cr3þ, 75 mg L1 Pb2þ, but only 10 mg L1 heavy metal mixture. Azabou et al., 2007a who studied heavy metal inhibition and precipitation by a mixture of SRB in batch condition showed that more than 72 mg L1 of Zn2þ caused death of organisms. Goncalves et al., 2007 reported that more than 99% of 70 mg L1 Zn and 2 mg L1 Cd were removed from synthetic wastewater in a continuous bench-scale upflow anaerobic sludge blanked reactor. The results obtained from different studies were not similar, suggesting that the toxic and inhibitory effect of heavy metals on SRB cultures are influenced by many factors such as the chemical and physiochemical properties of the surrounding SRB environment, the species composition of microbial community, and configuration of bioreactor. Moreover, the simultaneous presence of heavy metals could induce synergistic or cumulative toxic effects. Utgikar et al., 2004 reported that the toxic effects of binary mixtures of Cu and Zn were significantly higher than the toxic effect of individual heavy metal. This was also demonstrated by the study of Hao et al., 1994 mentioned above. In the present study, 94e100% of w10e30 mg L1 each of heavy metal mixture (Cu2þ, Zn2þ, Ni2þ and Cr6þ) was removed effectively by SRB consortium weekly throughout 12-week operational period. The obtained results might be due to the use of an indigenous SRB consortium isolated from heavy metal contaminated sediment, resulting in an effective heavy metal removal. Possible explanations for why the use of indigenous consortia may be more advantageous than the use of single species could be: (i) indigenous consortia containing multi-species have adapted to a heavy metal polluted environment by developing a variety of resistance mechanisms (see introduction). They are less liable to mutate and to be contaminated from other microorganisms, and (ii) they contain more than one kind of organisms that facilitate the formation of reducing conditions by oxidizing completely carbon sources (Bruins et al., 2000). In addition, it is difficult to maintain culture purity due to the ubiquity of microorganisms in the environment. Therefore, using a consortium instead of single species is an optimum option and widely applied for heavy metal treatment bioreactors. An increase of heavy metal loading rate of up to 6 and 7.5 mg L1 d1 in this study resulted in a decrease or even no detection of hybridized positive cells, sulfate reduction, sulfide production, and heavy metal precipitation. Contrary to common belief that only soluble metallic ion can be toxic or inhibitory, the insoluble metallic compounds, especially metal sulfides, could affect the activity of SRB by deposition on the surface of the cells and blocking the access to the substrate and other nutrients (Utgikar et al., 2002). At low levels of sulfide precipitate, the bacteria themselves may directly accelerate metal sulfide precipitation and facilitate settling of the solids by binding the metal in their cell walls and extracellular polymeric substances (EPS) (Beech and Cheung, 1995). This was demonstrated by Jalali and Baldwin, 2000 and Azabou et al., 2007b, who evaluated the influence on copper and zinc removal, respectively. The precipitation of copper and zinc was detected more quickly in the presence of bacteria cells than without bacteria cells in these both studies. Thus, association of copper and zinc with the bacterial cells could promote the precipitation rate.
3869
Although the presence of bacterial cells may facilitate metal precipitation, the high level of sulfide precipitate can act as a barrier between the cells and their essential growth nutrients. Therefore, the influent with heavy metal concentrations below the inhibitory level to maintain a maximum rate of sulfidogenesis are required to have a successful operation for heavy metal sulfidogenic bioreactors. Metal sulfides have been attributed to the major precipitation of Cu, Zn, and Ni by biological activity of SRB in the present study. This is confirmed by a simultaneous decrease of sulfate and heavy metal concentrations in the effluent and EDS spectrum with strong peaks of copper, zinc, nickel and sulfur. This is in agreement with Azabou et al., 2007a,b who showed that zinc removal was possible only in the presence of sulfate. This suggests that sulfide produced from sulfate reduction by activity of SRB is responsible for zinc removal as ZnS. In addition to precipitation with sulfide, heavy metals may also have been removed through sorption to the biomass or by other precipitation mechanisms i.e., hydroxide and carbonate precipitation, as well as generated alkalinity (Goncalves et al., 2007; Neculita et al., 2007). However, sulfide precipitation is the dominant mechanism, whereas other mechanisms play only a minor role for the removal of heavy metals in anaerobic sulfidogenic bioreactors (Machemer and Wildeman, 1992; Radhika et al., 2006). As mentioned in the introduction, soluble Cr6þ can be reduced into much less toxic and insoluble Cr3þ by reacting with bacterially produced hydrogen sulfide or enzymatic reduction. However, precipitation of Cr as metal sulfide is not stable in aqueous medium in comparison with Cu, Zn, and Ni. Thus, the reduction to Cr3þ is likely to be followed by rapid deposition as hydroxides with the concentration of soluble Cr3þ in equilibrium with Cr(OH)3 is w6.3 1031 mol L1 (Dean, 1999).
5.
Conclusions
There was no inhibition of SRB growth and the effective removal (94e100%) of heavy metal mixture (Cu2þ, Zn2þ, Ni2þ, and Cr6þ) can be achieved in R1eR3 with heavy metal loading rates ranging from 1.5 to 4.5 mg L1 d1 throughout 12 weeks of experiment. The heavy metal removal efficiency (91e97%) was reduced from week 9 onwards and the inhibition of SRB growth was observed at the heavy metal loading rate of 6 mg L1 d1 (R4). The toxic effect of heavy metals on SRB growth was revealed in R5 (7.5 mg L1 d1), in which no SRB could survive and no heavy metal precipitation was detected after four weeks of operation. This implied that the investigated SRB consortium might have potential application for heavy metal sulfidogenic bioreactors. To have a successful heavy metal treatment process, the heavy metal loading rates below the inhibitory level to SRB activity ranging from 1.5 to 4.5 mg L1 d1 should be selected.
Acknowledgment The authors acknowledge financial support provided by Federal Ministry for Education and Research (BMBF/Germany), Institute of Water Quality Control, Faculty of Civil Engineering and Geodesy, Technische Universita¨t Mu¨nchen, Germany.
3870
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 6 3 e3 8 7 0
references
Amann, R., Ludwig, W., Schleifer, K.H., 1995. Phylogenetic identification and in situ detection of invidual microbial cells without cultivaon. Microbiol. Rev. 59, 143e169. APHA, 1998. Standard Methods for Examination of Water and Wastewater, twentieth ed.. American Public Health Association, Washington DC, USA. Azabou, S., Mechichi, T., Sayadi, S., 2007a. Zinc precipitation by heavy-metal tolerant sulfate-reducing bacteria enriched on phosphogypsum as a sulfate source. Miner. Eng. 20, 173e178. Azabou, S., Mechichi, T., Patel, B.K.C., Sayadi, S., 2007b. Isolation and characterization of a mesophilic heavy-metals-tolerant sulfate-reducing bacterium Desulfomicrobium sp. from an enrichment culture using phosphogypsum as a sulfate source. J. Hazard. Mater. 140 (1e2), 264e270. Bartzas, G., Komnitsas, K., Paspaliaris, L., 2006. Laboratory evaluation of Feo barriers to treat acidic leachates. Mineral. Eng. 19 (5), 505e514. Beech, I.B., Cheung, C.W.S., 1995. Interactions of exopolymers produced by sulfate-reducing bacteria with metal ions. Int Biodeterior. Biodegradation 35 (1e3), 59e72. Beyenal, H., Lewandowski, Z., 2004. Dynamics of lead immobilization in sulfate reducing biofilms. Water Res. 38, 2726e2736. Bruins, M.R., Kapil, S., Oehme, F.W., 2000. Microbial resistance to metals in the environment. Ecotoxicol. Environ. Saf. 45, 198e207. Chang, I.S., Kim, B.H., 2007. Effect of sulfate reduction activity on biological treatment of hexavalent chromium [Cr(VI)] contaminated electroplating wastewater under sulfate-rich condition. Chemosphere 68, 218e226. Dean, J.A., 1999. Lange’s Handbook of Chemistry, fifteenth ed. McGraw-Hill, Inc, New York, USA, Pp. 5.6 and 8.6e8.17. Gallegos-Garcia, M., Celis, L.B., Rangel-Me´ndez, R., Razo-Flores, E., 2009. Precipitation and recovery of metal sulfides from metal containing acidic wastewater in a sulfidogenic down-flow fluidized bed reactor. Biotechnol. Bioeng. 102, 91e99. Goncalves, M.M.M., da Costa, A.C.A., Leite, S.G.F., Sant’Anna, G.L., 2007. Heavy metal removal from synthetic wastewaters in an anaerobic bioreactor using stillage from ethanol distilleries as a carbon source. Chemosphere 69, 1815e1820. Goulhen, F., Gloter, A., Guyot, F., Bruschi, M., 2006. Cr(VI) detoxification by De´ulfovibrio vulgaris strain Hildenborough: microbe-metal interactions studies. Appl. Microbiol. Biotechnol. 71, 892e897. Hao, O.J., Huang, L., Chen, J.M., Bugass, R.L., 1994. Effects of metal additions on sulfate reduction activity in wastewater. Toxicol. Environ. Chem. 46 (4), 197e212. Hulshof, A.H.M., Blowes, D.W., Gould, W.D., 2006. Evaluation of in situ layers for treatment of acid min drainage: a field comparison. Water Res. 40, 1816e1826. Jalali, K., Baldwin, S.A., 2000. The role of sulfate reducing bacteria in copper removal from aqueous sulfate solutions. Water Res. 34 (3), 797e806. Kaksonen, A.H., Vanhanen, M.L.R., Puhakka, J.A., 2003. Optimization of metal sulphide precipitation in fluidized-bed treatment of acidic wastewater. Water Res. 37, 255e266. Lenz, M., Hullebusch, E.D.V., Hommes, G., Corvini, P.F.X., Lens, P. N.L., 2008. Selenate removal in methanogenic and sulfate-
reducing upflow anaerobic sludge bed reactors. Water Res. 42, 2184e2194. Lovley, D.R., Phillips, E.J., 1994. Reduction of chromate by Desulfovibrio vulgaris and its c3 cytochrome. Appl. Environ. Microbiol. 60, 726e738. Machemer, S.D., Wildeman, T.R., 1992. Adsorption compared with sulphide precipitation as metal removal processes from acid mine drainage in constructed wetland. J. Contam. Hydrol. 9, 115e131. Miller, T.L., Wolin, M.J., 1974. A serum bottle modification of the Hungate technique for cultivating obligate anaerobes. Appl. Microbiol. 27 (5), 985e987. Nagpal, S., Chuichulcherm, S., Livingston, A., Peeva, L., 2000. Ethanol untilization by sulfate-reducing bacteria: an experimental and modeling study. Biotechnol. Bioeng. 70 (5), 533e543. Neculita, C.M., Zagury, G.J., Bussie`re, B., 2007. Passive treatment of acid mine drainage in bioreactors using sulfate-reducing bacteria: critical review and research needs. J. Environ. Qual. 36, 1e16. Postgate, J.R., 1984. The Sulfate-reducing Bacteria, second ed. Cambridge University Press, Cabridge. Radhika, V., Subramanian, S., Natarajan, K.A., 2006. Bioremediation of zinc using Desulfotomaculum nigrificans: bioprecipitaion and characterization studies. Water Res. 40, 3628e3635. Sahinkaya, E., 2009. Biotreatment of zinc-containing wastewater in sulfidogenic CSTR: performance and artificial neural network (ANN) modeling studies. J. Hazard. Mater. 164, 105e113. Sani, R.K., Peyton, B.M., Brown, L.T., 2001. Copper-induced inhibition of growth of Desulfovibrio desulfuricans G20: assessement of its toxicity and correlation with those of zinc and lead. Appl. Environ. Microbiol. 67 (10), 4765e4772. Tekerlekopoulou, A.G., Tsiamis, G., Dermou, E., Siozios, S., Bourtzis, K., Vayenas, D.V., 2010. The effect of carbon sourse on microbial community structure and Cr(VI) reduction rate. Biotechnol. Bioeng. 107, 478e487. Tsukamoto, T.K., Killion, H.A., Miller, G.C., 2004. Column experiments for microbiological treatment of acid mine drainage: low-temperature, low-pH and matrix investigations. Water Res. 38, 1405e1418. Uhrie, J.L., Drever, J.I., Colberg, P.J.S., Nesbitt, C.C., 1996. In situ immobilization of heavy metals associated with uranium leach mines by bacterial sulfate reduction. Hydrometallurgy 43, 231e239. Utgikar, V.P., Chaudhary, N., Koeniger, A., Tabax, H.H., Heines, J. R., Govind, R., 2004. Toxicity of metals and metal mixtures analysis of concentration and time dependence for zinc and copper. Water Res. 38, 3651e3658. Utgikar, V.P., Chen, B.-Y., Chaudhary, N., Tabak, H.H., Haines, J.R., Govind, R., 2001. Acute toxicity of heavy metals to acetateuntilizing mixed cultures of sulfate-reducing bacteria: EC100 and EC50. Environ. Toxicol. Chem. 20, 2662e2669. Utgikar, V.P., Harmon, S.M., Chaudhary, N., Tabak, H.H., 2002. Inhibition of sulfate-reducing bacteria by metal sulfide formation in bioremediation of acid mine drainage. Environ. Toxicol. 17, 40e48. Viggi, C.C., Pagnanelli, F., Cibati, A., Uccelletti, D., Palleschi, C., Toro, L., 2010. Biotreatment and bioassessment of heavy metal removal by sulphate reducing bacteria in fixed bed reactors. Water Res. 44, 151e158.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 7 1 e3 8 8 2
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Rheological and fractal characteristics of unconditioned and conditioned water treatment residuals Y.J. Dong, Y.L. Wang*, J. Feng College of Environmental Science and Engineering, Research Center for Water Pollution Source Control and Eco-remediation, Beijing Forestry University, Beijing 100083, China
article info
abstract
Article history:
The rheological and fractal characteristics of raw (unconditioned) and conditioned water
Received 17 February 2011
treatment residuals (WTRs) were investigated in this study. Variations in morphology, size,
Received in revised form
and image fractal dimensions of the flocs/aggregates in these WTR systems with increasing
16 April 2011
polymer doses were analyzed. The results showed that when the raw WTRs were conditioned
Accepted 25 April 2011
with the polymer CZ8688, the optimum polymer dosage was observed at 24 kg/ton dry sludge.
Available online 4 May 2011
The average diameter of irregularly shaped flocs/aggregates in the WTR suspensions increased from 42.54 mm to several hundred micrometers with increasing polymer doses.
Keywords:
Furthermore, the aggregates in the conditioned WTR system displayed boundary/surface and
Water treatment residuals
mass fractals. At the optimum polymer dosage, the aggregates formed had a volumetric
Conditioning
average diameter of about 820.7 mm, with a one-dimensional fractal dimension of 1.01 and
Polymer
a mass fractal dimension of 2.74 on the basis of the image analysis. Rheological tests indicated
Rheology
that the conditioned WTRs at the optimum polymer dosage showed higher levels of shear-
Fractal
thinning behavior than the raw WTRs. Variations in the limiting viscosity (hN) of condi-
Model
tioned WTRs with sludge content could be described by a linear equation, which were different from the often-observed empirical exponential relationship for most municipal sludge. With increasing temperature, the hN of the raw WTRs decreased more rapidly than that of the raw WTRs. Good fitting results for the relationships between lghNwT using the Arrhenius equation indicate that the WTRs had a much higher activation energy for viscosity of about 17.86e26.91 J/mol compared with that of anaerobic granular sludge (2.51 J/mol) (Mu and Yu, 2006). In addition, the Bingham plastic model adequately described the rheological behavior of the conditioned WTRs, whereas the rheology of the raw WTRs fit the HerscheleBulkley model well at only certain sludge contents. Considering the good powerelaw relationships between the limiting viscosity and sludge content of the conditioned WTRs, their mass fractal dimensions were calculated through the models proposed by Shih et al. (1990), which were 2.48 for these conditioned WTR aggregates. The results demonstrate that conditioned WTRs behave like weak-link flocs/aggregates. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
Chemical coagulants, such as alum salts, either alum or polyaluminium chloride(PACl), and ferric salts, either ferric or
polyferric sulfate (PFS), are employed in water treatment to remove fine particles (e.g., clay), organic matter, and algae from water. The sludge generated can be called water treatment residuals (WTRs), which are alum or ferric flocs and their
* Corresponding author. Tel.: þ86 10 62336 528; fax: þ86 10 62336596. E-mail address:
[email protected] (Y.L. Wang). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.04.042
3872
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 7 1 e3 8 8 2
mixtures. Compared with wastewater treatment biosolids, WTRs are generally smaller, less compressible, more compact in structure, and have lower water contents (Lai and Liu, 2004). Numerous studies in biosolid treatment and disposal have proven that chemical conditioning is a key process to improve the efficiency of dewatering and reduce the quality of the final waste product, as well as the operating costs of overall dewatering and disposal operations (Langer et al., 1994; Dentel, 1997; Abu-Orf and Dentel, 1999; Dentel et al., 2005; Ayol et al., 2005). Few studies, however, have been conducted to understand the effects of conditioning on the physicochemical characteristics and dewaterability of WTRs (Wu et al., 1997, 2003; Zhao, 2002, 2003, 2004; Bache and Papavasilopoulos, 2003; Lai and Liu, 2004; Turchiuli and Fargues, 2004; Ayol et al., 2005). Rheology is a powerful tool for characterizing the material properties of non-Newtonian fluids, such as sludge suspensions, because it can scientifically describe and even predict flow behaviors in real processes (Dentel, 1997; Dentel et al., 2005). An understanding of the rheological properties of sludge is important for its management. Such properties may be used as design parameters for transport, storage, landfill, and spreading operations and as control parameters in many treatments, such as stabilization and dewatering (Lotito et al., 1997; Yen et al., 2002; Dentel et al., 2005; Chen et al., 2005). A considerable number of scientific studies for waste-activated, anaerobic digested, and anaerobic granular sludge have been conducted regarding correlations between rheological parameters and other material characteristics, such as particle size and distribution, surface physicochemical parameters (e.g., charge, hydrophibicity/hydrophobicity, tension), and fractal dimension (Forster, 2002; Seyssiecq et al., 2003; Jin et al., 2004; Mu and Yu, 2006; Wang and Dentel, 2010). However, for conditioned WTRs, only Ayol (2005) conducted experiments on the effect of dual-conditioner doses on their rheograms. Natural and engineering flocs/aggregates, such as those formed during water and wastewater treatment (Li and Ganczarczyk, 1989; Wu et al., 2002; Chu and Lee, 2004), have generally been characterized as highly porous fractallike forms of many primary particles. Especially, alum or ferric salt-coagulated flocs can be characterized as fractals (Wu et al., 1997; Zhao, 2002, 2003, 2004; Bache and Papavasilopoulos, 2003; Turchiuli and Fargues, 2004). When a WTR is conditioned with a polymer, the aggregates packed by the aforementioned flocs also present a fractal structure. Zhao (2002, 2003, 2004). Wu et al. (2003) had calculated the corresponding mass fractal dimensions of raw and conditioned WTRs from the logelog plot of their size versus density. A powerful application of fractal geometry for describing raw and conditioned sludge aggregate structures lies in its potential to relate formation processes to their structure (Li and Ganczarczyk, 1989; Chu and Lee, 2004). Furthermore, raw and conditioned WTRs combined with polymers could behave as a gel system. Bache and Papavasilopoulos (2003) indicated that the WTR formed by alum sulfate coagulation was gelatinous in character. Thus, the relationship between its rheological properties and fractal nature can be simulated using the model by Shih et al. (1990). Mu and Yu (2006) succeeded in using the Shih model to describe the relationship between the rheological
and fractal properties of anaerobic granular sludge. Thus, determining whether or not such a relationship exists in raw and conditioned WTR systems is of considerable interest. Although a few studies have been conducted on the correlations between the physical properties of WTR flocs/aggregates and specific polymer dosages during conditioning or dewatering, little information is available regarding the rheological and fractal properties of raw and conditioned WTRs. Therefore, the main objective of this study is to explore the rheological and physical properties of WTRs, such as size and fractal structure, and compare them between raw and conditioned samples. The influence of conditioning on these properties is also identified. The results obtained will be very important in providing further information on the mechanism and practical application of WTR conditioning and dewatering processes.
2.
Materials and methods
2.1.
Raw WTRs
The raw WTRs used in this study were collected from the No. 9 water plant of the Beijing Waterworks Group Ltd., China. The water treatment plant handles 1.5 106 m3 of water per day using coagulationeflocculation, sedimentation and activated carbon filtration treatments. The coagulation process is fulfilled in the high speed mixing chambers with the coagulant of aluminum or ferric salt. In this plant, aluminum and ferric saltgenerated residuals are first conditioned with a kind of polyacrylamide (PAM) polymer, after which a centrifuge is used to dewater these conditioned residuals. The raw residuals were collected upstream of the conditioning and dewatering processes, and then immediately transferred to the laboratory at the Beijing Forestry University and stored at 4 C. Prior to the experiments, the sludge sample was warmed to 25 C. All measurements were performed within 7 d from the date of sampling. The dry-solid content of the residual samples was determined from the weight loss of the sludge samples that were dried at 105 C over 24 h (APHA, 1995). The total solids concentration (TS) of the samples was determined to be 1.253%, and their pH was 7.30. During dewaterability tests, a Triton 304B instrument was used for capillary suction time (CST) measurements, and a Buchner funnel was used for specific resistance to filtration (SRF) determination, and the corresponding operation pressure during filtration and for the SRF quantification is 0.038 MPa. The CST and SRF of the raw residuals were 97.97 5.75 s and 2.67 1011 2.64 1010 m/kg, respectively. Their average size was found to be 42.54 mm using a Mastersizer 2000 instrument (Malvern, UK). To obtain zeta potential measurements, the raw WTRs were diluted 20 times, and their diluted suspension was sampled afterward. The zeta potential of the diluted suspension was 10.2 3.6 mV.
2.2.
WTR conditioning
A high molecular weight cationic organic polymer, CZ8698, was obtained from Beijing Beike Lvjie Environmental Protection Science and Technology Ltd. This polymer is a type of PAM with an average molecular weight above ten million
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 7 1 e3 8 8 2
Daltons and a charge density of 1.5650 meq/g TS. For laboratory-scale conditioning the WTRs with CZ8698, a jar test procedure was employed using a classical six-paddle stirring apparatus (JTY-6, Tangshan Dachang Chemicals Ltd., China). Several 1 L beakers were filled to 500 mL with the WTR samples under a multipaddle stirrer. Next, a stock solution of 0.5% CZ8698 was quickly injected into the WTR sample in each beaker within 5 s under a pre-imposed mixing intensity of 800 rpm. Then, the following mixing procedure, designated to yield a certain overall energy input into the suspension, was used: a rapid mixing period of 1 min at 800 rpm followed by a slow-stir phase at 60 rpm for 5 min. After mixing, the conditioned sludge aggregate samples that formed at different polymer doses were withdrawn and subjected to dewatering and geometric testing. All the rheological tests were repeated for raw and conditioned WTRs at optimum polymer CZ8688 dosage.
2.3.
Rheological testing
A rotational viscometer (NXS-11A Rotational Viscometer, Chengdu Instrument Co. China) was employed for the rheological measurements. The rotor had a radius of 38.46 mm and a height of 70 mm, and the cup had a radius of 40 mm. The gap between the rotor and cup was 0.77 mm. Fifteen different shear rates (gdot) of 15.5e996 s1 were available at this viscometer. A continuous ramp of shear rate was applied during the rheological test for WTR. Then at each shear rate, the duration time of 30 s was employed to ensure that the equilibrium point of the stationary state can be reached. The rheograms of shear stress (s)- shear rate (gdot) for raw and conditioned WTR samples were recorded and analyzed.
2.4.
Geometric characterization
The particles of raw WTRs were too small to capture by an ordinary microscope; thus, the images of the raw and conditioned aggregates at the optimum polymer dosage were obtained by scanning electron microscopy (SEM) (Quanta 200, FEI). Samples were first fixed for 1 h at 4 C with 2.5% (w/v) glutaraldehyde in phosphate buffer and then dehydrated through a graded series of acetone-water mixtures, (10%, 25%, 50%, 75%, 90%, and 100%). The samples were brought to equilibrium for 10 min, and finally dried by the critical-point drying method before sputter-coating with gold particles. The micrographs of the dried and golden-coated samples were obtained by SEM at magnifications between 500 and 150,000. The geometric characteristics of the conditioned WTR aggregates determined at different polymer doses in the present study include their morphology, size distribution, and fractal dimension. The conditioned WTR aggregate samples were carefully withdrawn from the stirred mixture and then evenly and separately introduced into deionized water contained within a glass square dish for image recording using a wide mouthed pipette. Aggregates that dissociated by flow shearing at the mouth of the pipette were picked out and discarded. After the transfer was complete, a digital microscope of GE-5 (Aigo, China) was used to obtain pictures of the aggregates in the dish. The geometric characteristics were derived from the images using Image-pro Plus 5.0. Due to the
3873
absence of spherical aggregates, they can be characterized in many ways, and the software then determines the area, perimeter, diameter, and second-order moments of the image for each aggregate. The size distribution of the aggregates was determined through statistical analysis of the above geometric parameters. The one-dimensional fractal dimension (D1) of the flocs was calculated by regression analysis of the logarithm of their perimeters versus the logarithm of their corresponding characteristic lengths (Jin and Wang, 2001; Wang et al., 2009). The two dimensional fractal dimension (D2) for the flocs was calculated by regression analysis of the logarithm of their projected areas versus the logarithm of their corresponding perimeters (Jin and Wang, 2001; Wang et al., 2009). In this study, the longest diameter of the aggregate image was considered as the characteristic length. In addition, the mass fractal dimensions of the raw and conditioned WTRs at the optimum polymer dosage were determined based on the D2 values or rheological test results according to the methods of Shih et al. (1990) and Mu and Yu (2006). In the present study, each experiment was replicated thrice to ensure consistency of the results.
3.
Results and discussion
3.1.
Determination of the optimum polymer dose
Fig. 1 shows the values of CST and SRF with varying polymer dosage. These two parameters for WTR dewaterability displayed different variation trends. Compared with raw WTRs, CST values for conditioned WTRs sharply decreased at a polymer dose lower than 12 kg/ton dry sludge. They then decreased slowly and reached a minimum value at a polymer dosage of 24 kg/ton dry sludge. Afterward, the CST values slightly increased. Therefore, on the basis of aforementioned CST evolution, the optimum dosage for WTR conditioning with polymer CZ8688 was determined at 24 kg/ton dry sludge, and the corresponding CST value for conditioned WTRs was 11.57 s. On the other hand, the SRF showed an slight decreasing tend at initial two polymer doses, and an immediately much increasing trend until the polymer dosage reached 12 kg/ton dry sludge, after which decreased at
Fig. 1 e CST and SRF as functions of polymer CZ8688 dose.
3874
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 7 1 e3 8 8 2
a polymer dosage of 24 kg/ton dry sludge. As the polymer dosage further increased, the SRF for the conditioned WTR increased. The filter paper was finally broken during the vacuum-driven filtration because of the highly viscous conditioned WTR suspension at a polymer dose of 40 kg/ton dry sludge. In general, the SRF test can simulate the vacuum filtration process for sludge dewatering, while the CST test is conducted without any pressure. The increase of SRF values implied that the pores among the biosolids became smaller and the corresponding high hindrance was met for water transportation under pressure. At the polymer doses less than 8 kg/ton dry sludge, the conditioned WTR flocs showed slight more compact structure with increasing of polymer dosage, while as the polymer doses increased to 12 kg/ton dry sludge, a much less compact structure was formed among these conditioned WTR flocs, which showed deficient strength to resist the compression of pressure under vacuum filtration, then resulted in much high resistance to water transportation in them. With the increase of polymer doses from 12 to 24 kg/ ton dry sludge, the conditioned WTR flocs became compact again, the corresponding SRF values decreased. Furthermore, as the polymer doses increased, the pores among these flocs became smaller and even collapsed at 40 kg/ton dry sludge, so the SRF value increased to very high value. On the basis of SRF value of the conditioned WTR, the optimum polymer doses should be 8 kg/ton dry sludge or 24 kg/ton dry sludge. Combined with CST and SRF variation trend for WTR, a polymer dosage of 24 kg/ton dry sludge was selected as optimum value for its conditioning, the corresponding CST value for conditioned WTR was less than 20 s and SRF value was less than 2.0 1011 m/kg.
3.2. Rheological characterization of the raw and conditioned WTRs 3.2.1.
Typical rheogram of the raw and conditioned WTRs
The typical rheograms of the raw and conditioned WTRs are shown in Fig. 2. For raw WTRs, the shear stress versus shear rate test was conducted at a TSS of 37.50 g/L and temperature (T ) of 303 K. For conditioned WTRs at the optimum polymer dosage, the test was done at a slightly higher TSS (49.94 g/L). During the rheological test, the shear rate was increased from 15.5 to 996.1 s1, and then vice versa. The corresponding shear stress was recorded at the same time. As illustrated in Fig. 2, the shear stress values were higher on the ascending path than on the descending path; this is attributed to the thixotropic properties of the WTRs (Larson, 1999; Mezger, 2000; Seyssiecq et al., 2003). The area of the hysteresis loop could be taken as a measure of the degree of thixotropy exhibited by WTRs (Mezger, 2000; Yen and Lee, 2003). Based on the calculation results for the area of the hysteresis loop in the rheograms in Fig. 2, the conditioned WTRs are about 1.21 104 mPa/s of hysteresis loop area higher than the raw WTRs, implying that the conditioned WTRs were more thixotropic. In addition, similar shapes in the ascending and descending shear stress curves were observed in Fig. 2a and b, except those at several initial shear rate points. The shearthinning (thixotropy) behaviors of both raw and conditioned WTRs were also demonstrated by their apparent viscosity
Fig. 2 e Typical rheograms at 303 K (a) raw WTR, TS [ 37.50 g/L, (b) conditioned WTR, TS [ 49.94 g/L.
(happ, a quotient of shear stress and shear rate in the ascending path of rheograms) variations, which decreased rapidly as the shear rate was increased from 15.5 to 200 s1, and then decreased gradually, next became constant at higher shear rates. The constant viscosity at the infinite shear rate was called the limiting viscosity (hN) (Midoux, 1988; Steffe, 1996; Tixier et al., 2003; Mu and Yu, 2006; Mori et al., 2006; Khongnakorn et al., 2010). This can represent the viscosity of the WTRs corresponding to the maximum dispersion of flocs under the impact of shear rate (Tixier et al., 2003), which is also correlated with the optimal opening and orientation of the WTRs in the flow direction (Yen et al., 2002). Therefore, this parameter was used to characterize the WTRs. On the basis of data matrix of shear stress-shear rate in the ascending path of rheograms, all the values of limiting viscosities given for raw and conditioned materials have been approached on the basis of an extrapolation at very high shear rate using a rheological model of Sisko including this specific parameter (Mori et al., 2006), and the data matrix for above calculations was the array of shear stress-shear rate greater than 300 s1.
3.2.2.
Effect of residual concentration on WTR rheology
Many research studies have shown that the TS content is the main factor that strongly influences the rheological properties of sludge (Monteiro, 1997; Lotito et al., 1997; Slatter, 1997; Seyssiecq et al., 2003; Mu and Yu, 2006). The limiting viscosity as a function of TS is represented in Fig. 3. For the
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 7 1 e3 8 8 2
3875
water, also have important effects on the viscosity of the sludge (Seyssiecq et al., 2003; Mu and Yu, 2006; Khongnakorn et al., 2010). WTRs are a type of sludge produced in water treatment plants; they have fewer organic materials and a smaller biomass than municipal sludge. These differences may inevitably lead to many discrepancies in the parameters described above, and even in behaviors in which particleeparticle interactions influence sludge rheology. However, according to the research results obtained by Khongnakorn et al. (2010), the soluble organic compound (SMP) concentration and the TSS content have equivalent quantitative influence on the limiting viscosity hN, which can characterize the flow of the sludge. It seems obvious that the discrepancies in the organic content between WTR and municipal sludge do not result their different flow behavior. As indicated in Fig. 3a and b, conditioning has a significant effect on the limiting viscosity of WTRs. The limiting viscosities of raw and conditioned WTR not only showed a contrary variation trend with the increase of TS content, but also the hN of conditioned WTRs was slight lower than that of raw WTRs at the same TS content between 32.1 g/L and 37.5 g/L, while at higher TS contents, the corresponding hN values of conditioned WTRs became higher. These results implied that a higher limiting viscosity will be observed at higher TS content for conditioned WTR, though it is composed of flocs of higher size.
3.2.3.
Fig. 3 e Limiting viscosity of (a) raw and (b) conditioned WTR as a function of TS at 303 K.
raw WTRs, it fluctuated as the TS contents increased from 25 g/L to 37.5 g/L and then rapidly decreased with increasing TS content at several higher TS contents. For the conditioned WTRs at optimum polymer dosage, a good linear increase in hN as TS content increased was observed, and R2 value for the corresponding linear regression equation was greater than 0.89. However, an empirical exponential relationship was often observed between the hN and TS content of activated sludge (Tixier et al., 2003) and anaerobic granular sludge (Mu and Yu, 2006). In addition, Tixier et al. (2003) indicated that the relationship between most common rheological parameters, such as yield stress or viscosity, and the TS contents of sewage sludge and digested organic fractions of municipal solid wastes could be suitably described with exponential and power laws (Battistoni, 1997; Forster, 2002). In general, aforementioned regression laws can represent the way in which particleeparticle interactions influence sludge rheology (Forster, 2002). Therefore, the way for the WTR is different from that for municipal sludges. Besides TS content, other parameters, such as sludge composition, soluble organic compound (COD), the properties of suspended solids, flocculation state, particle size, and bound
Effect of temperature on the WTR rheology
Temperature has a strong influence on the rheological behavior of solideliquid suspensions (Steffe, 1996; Larson, 1999; Mezger, 2000; Mu and Yu, 2006). Fig. 4 shows that the limiting viscosity decreased from 8.42 0.39 to 1.70 0.58 mPa s for raw WTRs, and from 12.52 1.88 to 1.83 0.24 mPa s for conditioned WTR as the test temperature increased from 283 to 343 K. Compared with the anaerobic granular sludge suspensions reported by Mu and Yu (2006), the WTRs in this study showed a consistent decreasing trend. Greater change in hN implied that higher temperatures can weaken the network strength between particles in such solideliquid suspensions due to increase in intense thermal motions of these particles. Furthermore, similar to the study by Mu and Yu (2006), an Arrhenius type equation can suitably fit the data matrix of lghN against 1/T, and the activation energy for viscosity (Ea) can also be calculated through the slope of the regression line. The Ea values for the raw and conditioned WTRs were 17.86 and 26.91 J/mol, respectively, which were much higher than the Ea of 2.51 J/mol obtained for granular sludge samples. This difference indicates that more energy is needed to reach the same level of thermal motion or limiting viscosity for WTRs than for granular sludge. Moreover, a lower Ea implies that hN undergoes a more rapid rate of change with temperature. In addition, raw WTRs show more rapid changes in hN than conditioned WTRs.
3.2.4.
Modeling of the WTR rheology
The rheograms in Fig. 2 reveal a significant yield stress and non-linear manner in the shear stress-shear rate curve. Thus, the HerscheleBulkley model may be considered to describe the behavior of this non-Newtonian fluid (Steffe, 1996; Slatter, 1997; Moeller and Torres, 1997; Seyssiecq et al., 2003).
3876
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 7 1 e3 8 8 2
Table 1 e Fitting results for the raw WTR using the HerscheleBulkley model at various sludge concentrations. TS (g/L)
25.0 28.1 32.1 37.5 45 56.3
R2
HerscheleBulkley model s0 (mPa)
K (mPa sn)
N
2698.76 4261.75 2228.49 1719.37 4400.30 17,617.18
81.00 188.88 5753.46 7541.13 12,205.56 977.69
0.66 0.56 0.153 0.137 0.058 0.270
0.9918 0.9659 0.8848 0.8773 0.7560 0.8341
Note: T ¼ 303 K.
plastic model because of the high regression coefficients under all tested WTR concentrations. Values for s0 and hp increased from 1971.62 to 7268.97 mPa and from 3.79 to 12.70 mPa s, respectively, with increasing WTR concentrations from 32.10 to 44.94 g-TS/L. The s0 values in Table 2 indicate that, below these values, conditioned WTRs behave with solid-like features. Hence, more energy can be stored at small strains in conditioned WTR samples with higher WTR concentrations (Mu and Yu, 2006).
3.3. Geometric characterization of the raw and conditioned WTR 3.3.1. Morphology and size of the raw and conditioned WTR aggregates
Fig. 4 e Log of the limiting viscosity as a function of 1/T (a) raw WTR, TS [ 32.10 g/L, (b) conditioned WTR, TS [ 44.94 g/L.
n
s ¼ s0 þ K,ðgdotÞ
(1)
where s is the shear press (mPa), s0 is the yield stress (mPa), gdot is the shear rate (1/s), K is the consistency index (mPa sn), and n is the flow index (dimensionless). When 0 < n < 1, this model is appropriate for a shear-thinning fluid. The Bingham plastic model can be derived from the HerscheleBulkley model when n is equal to 1, and K is commonly called the plastic viscosity (hp, mPa s). The best-fit results for the rheogram data matrix at various sludge concentrations are summarized in Table 1 for raw WTRs and Table 2 for conditioned WTRs. The fitting results for the raw WTR in Table 1 indicate that the HerscheleBulkley model can describe the rheological behaviors of these sludge samples under various concentrations. All the values for s0, K, and n showed a fluctuation variation trend as the raw WTR TS content increased. Even at some WTR concentrations of 32.1, 37.5, and 45 g/L, non-linear fitting results did not converge; their fitting results can only be obtained at higher regression coefficients. Table 2 reveals that the rheological behaviors of conditioned WTRs can be adequately described by the Bingham
The raw and conditioned WTR aggregates were withdrawn and their pictures were recorded at different polymer doses. Fig. 5a provides several images of these aggregates. All the digital microphotographs prove that the aggregates had an irregular boundary and their sizes usually increased with increasing polymer dose. Comparison of the SEM images of conditioned WTRs with raw samples showed that the former aggregates displayed a more compact structure and smoother surface. The change in the size of the conditioned WTR aggregates is displayed in Fig. 5b. Considering the average size of 42.54 mm for raw WTR flocs, the sizes of conditioned WTR aggregates were about one order of magnitude greater than the flocs of raw WTRs. As well, the volumetric average diameter of the conditioned WTR aggregates increased from about 200 mm to over 900 mm as the polymer dose increased. The same trend was
Table 2 e Fitting results for the conditioned WTR using the Bingham plastic model at various sludge concentrations. TS (g/L)
32.10 34.57 37.45 42.40 44.94 Note: T ¼ 303 K.
Bingham plastic model s0 (mPa)
hp (mPa s)
1971.62 2141.50 3253.17 5061.04 7268.97
3.79 7.02 8.10 11.66 12.70
R2
0.8605 0.9914 0.9834 0.9822 0.9606
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 7 1 e3 8 8 2
3877
Fig. 5 e Geometric parameters: [(a) morphology and (b) average size] of raw and conditioned WTR flocs/aggregates as a function of polymer doses.
observed in the other three kinds of average diameters. The sharpest increases in average diameter of conditioned WTR aggregates were observed at polymer dosages of 16e24 kg/ton dry sludge. At the optimum polymer dosage, the aggregates had a volumetric average diameter of about 820.7 mm.
3.3.2.
Image fractal dimensions of conditioned WTR aggregates
According to the methods described in Section 2.4, the fractal characteristics of these WTR flocs were investigated, and their
corresponding linear regression plots for conditioned WTR are shown in Fig. 6. It was observed that the determination coefficients of these linear regressions were higher than 0.91, and the characteristic length or perimeter for fractal dimension calculation were 2 decades in magnitude. The change in the image fractal dimensions of the conditioned WTR aggregates is illustrated in Fig. 7. D2 increased from 1.50 to 1.74 as the polymer doses increased from 4 to 12 kg/ton dry sludge, and a slight decrease to 1.70 in D2 was observed at
3878
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 7 1 e3 8 8 2
Fig. 5 e (continued).
16 kg/ton dry sludge. D2 continued to increase up to 1.80 as the polymer doses further increased. D2 is equal to 1.74 at the optimum polymer dose of 24 kg/ton dry sludge. D1 showed a contrasting trend, and most D1 values were slightly greater than one. However, at a polymer dosage of 12 kg/ton dry sludge, D1 was 0.98, which did not conform to the definition of D1. This may be attributed to image resolution limitations during image analysis. Nevertheless, the D1 values obtained imply that the boundaries or surfaces of conditioned WTR aggregates were irregular, which is in good agreement with the results obtained from observations of their microphotograph in Fig. 5a.
D1 values plus unity can be used to indicate the surface fractal dimensions of the aggregates (Mandelbrot, 1983). Accordingly, aforementioned D2 values plus unity, which were in the range of 2.50e2.80, also indicate the mass fractal dimensions of the aggregates if their 3D structures can be expressed by the extension of their 2D images in the direction normal to their surface (Mandelbrot, 1983). These D2 values indicate that the conditioned WTR aggregates obtained in some polymer doses formed via the diffusion limited the particle-cluster aggregation model (Jullien, 1986). Mu and Yu (2006) observed that the mass fractal dimensions of
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 7 1 e3 8 8 2
3879
3.3.3. Mass fractal dimensions of raw and conditioned WTR particles
Fig. 6 e Double-logarithmic plots of PedL, and AeP plots for conditioned ADS at polymer dosage 24 kg/ton dry sludge. (a: lgPelgdL, b: lgAelgP).
anaerobic granular sludge were larger than those of other biosolids studied in the literature, implying that anaerobic granular sludge aggregates are more compact and dense. Comparing these results with the mass fractal dimensions of the conditioned WTR aggregates obtained in this study, the latter sometimes showed values close to those of anaerobic granular sludge aggregates. This indicates that the conditioned WTR aggregates can form a compact and dense structure. However, in other works, the mass fractal dimensions of raw alum WTRs were reported to be about 1.06 (Zhao, 2002, 2003, 2004), 1.18 (Wu et al., 1997), and 1.58 (Wu et al., 2003). The corresponding average mass fractal dimension of alum salt WTRs conditioned with polymers was found to be approximately 1.72 (Zhao, 2002, 2003, 2004), and other fractal dimensions ranging from 1.04 to 1.83 were found at different polymer doses (Wu et al., 1997). Another work showed a fractal average dimension of 1.80 (Wu et al., 2003). These mass fractal dimensions of conditioned alum WTRs were lower than the values obtained in the present study. This can be ascribed to the different methods used for the determination of fractal dimensions, aside from the discrepancies in the types of WTRs and polymers employed. Some errors that occurred during operation of the image system for determining the floc size and density may have contributed to the much-lower mass fractal dimensions obtained (Zhao, 2002, 2003, 2004).
The raw WTR is a suspension with flocs formed by fine particles in natural water flocculation with an Al or Fe-salt inorganic coagulant. These flocs are fractals (Li and Ganczarczyk, 1989; Wu et al., 1997; Zhao, 2002, 2003, 2004; Bache and Papavasilopoulos, 2003; Chu and Lee, 2004; Turchiuli and Fargues, 2004). As the suspension with flocs was conditioned with CZ8698, a kind of gel network structure was produced, which is approximately considered as a closely packed bigger fractal aggregates formed by the flocculation of aforementioned flocs throughout the suspension. As proposed by Shih et al. (1990), for colloid gels well above the gelation threshold, the scaling of the rheological properties, such as the elastic parameters, are dominated by the fractal nature of flocs or aggregates collected in the gel network system. Shih et al. (1990) and Vreeker et al. (1992) also developed the powerelaw relationships between the elastic constant (storage modulus G0 ), the limit of linearity (g0) or yield stress (s0), and particle concentration in their own gel system. Mu and Yu (2006) also gave another scaling relationship between the hN and particle concentration in a granule sludge system. Furthermore, depending on the strength of the links between the flocs/ aggregates in comparison to that of flocs/aggregates, a stronglink (interfloc/interaggregate) regime and a weak-link regime can be observed. The corresponding relationship between the rheological properties for colloid gels and the particle concentration can be described as follows: In the strong-link regime, KwFð3þxÞ=ð3Df Þ
(2)
g0 wFð1xÞ=ð3Df Þ
(3)
whereas in the weak-link regime, KwFðd2Þ=ð3Df Þ
(4)
g0 wF1=ð3Df Þ
(5)
Fig. 7 e Image fractal dimensions of conditioned WTR aggregates as a function of polymer doses.
3880
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 7 1 e3 8 8 2
aggregates, as evidence by very low Df values calculated using Eq. (2), which were less than 2.0 and did not favor the 3D structure of the aggregates. In addition, the logelog plot of yield stress (s0) and TS of the conditioned WTR in Table 2 conforms well to a powerelaw, i.e., s0wTSn, with n ¼ 3.952 (R2 ¼ 0.975), close to a range of approximately four decades, which is consistent with the result of Vreeker et al. (1992) for glycerol tristearate gel. However, the relationship between n and mass fractal dimension is not yet determined. In addition, the aforementioned powerelaw relationship was not observed in the raw WTR flocs. Gmachowski (1995, 1996) developed some equations to determine the aggregate structure factor (S ) and the ratio between of the hydrodynamic radius (RH) to the radius of aggregate (RA).
S¼
Df RH RA
RH ¼ RA
Fig. 8 e Double-logarithmic plots of the limiting viscosity versus TS at 303 K (a) raw WTR, (b) conditioned WTR.
where K is the elastic constant (storage modulus G0 or limiting viscosity hN), d is the Euclidean dimension, Df is the mass fractal dimension of flocs or aggregates, and x is the backbone fractal dimension of the flocs or aggregates, which increases from 1.0 to 1.3 as the particle concentration is lowered (Shih et al., 1990). Fig. 8 displayed the double-logarithmic plots of the limiting viscosity versus TS at 303 K for raw WTR and conditioned WTR. It is obvious observed that the logarithm of the limiting viscosity for raw WTR showed non-linear variation trend with the increase of TS content, whereas a good linear relationship between lghNelgTS was observed for conditioned WTR with the determination coefficients of 0.87. Then the slope of these straight lines can be employed to calculate the mass fractal dimension of conditioned WTR. The Df values for the conditioned WTRs were 0.92e0.77 and 2.48, as calculated using Eqs. (2) and (4), respectively. For the conditioned WTR aggregates at the optimum polymer dose, the Df value calculated using Eq. (4) for the weak-link regime was lower than the sum of unity and the D2 value estimated by the image analysis method, while both of them indicated that these aggregates had a relatively compact and dense structure. The conditioned WTRs behaved like weak-link
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2 Df 0:228 1:56 1:728 2
(6)
(7)
RH/RA and S for the conditioned WTR flocs were calculated to be 0.922 and 0.817 using Eqs. (6) and (7), respectively. Gmachowski (1995) and Mu and Yu (2006) proposed that the RH/RA could be used to describe the velocity ratio between the primary particles and the aggregates, which was valid for their solideliquid separation. Thus, RH/RA could be employed to determine the WTR settleability and filterability. In the present study, conditioned WTRs had better settleability and filterability than raw WTRs. The value of S could be employed to characterize the space-filling ability or compactness of WTRs. The conditioned WTRs had a compact structure. Gmachowski (1995) and Mu and Yu (2006) indicated that the compactness of sludge has a great effect on the fluid flow through it and its dynamic behavior. Therefore, the conditioning process can change the aforementioned properties of WTRs.
4.
Conclusions
Raw WTRs were conditioned with CZ8688; the optimum polymer dosage was found to be 24 kg/ton dry sludge. The irregular floc/aggregate sizes of WTR suspensions increased to several hundred micrometers as the polymer doses increased. Both the raw and conditioned WTRs showed shear-thinning behaviors, and the latter was more thixotropic. The variations of the limiting viscosity of the conditioned WTRs with the sludge content could be described by a linear equation. The limiting viscosity of the raw WTRs decreased more rapidly compared with that of raw WTRs with increasing temperature, and the relationships between lghNwT could be suitably fitted by the Arrhenius equation. The HerscheleBulkley and Bingham plastic models described the rheological behaviors of the raw and conditioned WTRs, respectively. The aggregates in the conditioned WTR system
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 7 1 e3 8 8 2
displayed boundary/surface and mass fractals. Furthermore, a difference in mass fractal dimension values can be derived from the WTR aggregate image analysis method and the relationship between the limiting viscosity and sludge content. The conditioned WTRs behave like weak-link aggregates.
Acknowledgments This work was supported by the National Natural Science Foundation of China (Nos. 20977008, 51078035), by the Fundamental Research Funds for the Central Universities (JC2011-1, TD2010-5), the Ph.D. Programs Foundation of Ministry of Education of China (20100014110004), the High-Tech Research and Development Program (863) of China (No. 2007AA06Z301), Major projects on control and rectification of water body pollution (2008ZX07422-002-004, 2008ZX07314-006).
references
Abu-Orf, M.M., Dentel, S.K., 1999. Rheology as a tool for polymer dose assessment and control. ASCE J. Environ. Eng. 125 (12), 1133e1141. APHA, AWWA, WEF, 1995. Standard Methods for the Examination of Water and Wastewater, nineteenth ed. American Public Health Association, Washington, DC. Ayol, A., 2005. Determination of rheological properties of sludges produced at different treatment stages. PhD dissertation, University of Delaware. Ayol, A., Dentel, S.K., Filibeli, A., 2005. Dual polymer conditioning of water treatment residuals. J. Environ. Eng. ASCE 131 (8), 1132e1138. Bache, D.H., Papavasilopoulos, E.N., 2003. Dewatering of aluminohumic sludge: impacts of hydroxide. Water Res. 37, 3289e3298. Battistoni, P., 1997. Pre-treatment, measurement execution procedure and waste characteristics in the rheology of sewage sludges and the digested organic fraction of municipal solid wastes. Water Sci. Technol. 36 (11), 33e41. Chen, B.-H., Lee, S.-J., Lee, D.J., 2005. Rheological characteristics of the cationic polyelectrolyte flocculated wastewater sludge. Water Res. 39, 4429e4435. Chu, C.P., Lee, D.J., 2004. Multiscale structures of biological flocs. Chem. Eng. Sci. 59, 1875e1883. Dentel, S.K., 1997. Evaluation and role of rheological properties in sludge management. Water Sci. Technol. 36 (11), 1e8. Dentel, S.K., Ayol, A., Filibeli, A., 2005. Modern rheometric characterization of sludges. J. Residuals Sci. Technol. 2 (4), 233e240. Forster, C.F., 2002. The rheological and physico-chemical characteristics of sewage sludges. Enzyme Microb. Technol. 30, 340e345. Gmachowski, L., 1995. Mechanism of shear aggregation. Water Res. 29, 1815e1820. Gmachowski, L., 1996. Hydrodynamics of aggregated media. J. Colloid Interface Sci. 178, 80e86. Jin, B., Wile´n, B.M., Lant, P., 2004. Impacts of morphological, physical and chemical properties.of sludge flocs on dewaterability of activated sludge. Chem. Eng. J. 98, 115e126.
3881
Jin, P.K., Wang, X.C., 2001. Morphological characteristics of alhumic floc and coagulation chemistry. Acta Scientiae Circumstantiae 21 (Suppl.), 23e29 (in Chinese). Jullien, R., 1986. Aggregation phenomena and fractal aggregates. Am. Telecommun. 41, 343e372. Khongnakorn, W., Mori, M., Vachoud, L., Delalonde, M., Wisniewski, C., 2010. Rheological properties of sMBR sludge under unsteady state conditions. Desalination 250 (2,15), 824e828. Lai, J.Y., Liu, J.C., 2004. Co-conditioning and dewatering of alum sludge and waste activated sludge. Water Sci. Technol. 50 (9), 41e48. Langer, S.J., Klute, R., Hahn, H.H., 1994. Mechanisms of floc formation in sludge conditioning with polymers. Water Sci. Technol. 30 (8), 129e138. Larson, R.G., 1999. The Structure and Rheology of Complex Fluids. University of Michigan. Oxford University Press, Ann Arbor. Li, D., Ganczarczyk, J., 1989. Structure of activated sludge flocs. Biotechnol. Bioeng. 35, 57e65. Lotito, V., Spinosa, L., Mininni, G., Antonacci, R., 1997. The rheology of sewage sludge at different steps of treatment. Water Sci. Technol. 36 (11), 79e85. Mandelbrot, B.B., 1983. The Fractal Geometry of Nature. W.H. Freeman and Co., New York. Mezger, T., 2000. The Rheology Handbook, ISBN 3-87870-567-0. Midoux, N., 1988. Me´canique et rhe´ologie des fluides en genie chimique. Technique et documentation, 2e´me ed. Lavoisier. Moeller, G., Torres, L.G., 1997. Rheological characterization of primary and secondary sludges treated by both aerobic and anaerobic digestion. Bioresour. Technol. 61, 207e211. Mori, M., Seyssiecq, I., Roche, N., 2006. Rheological measurements of sewage sludge for various solids concentrations and geometry. Process Biochem. 41, 1656e1662. Monteiro, P.S., 1997. The influence of the anaerobic digestion process on the sewage sludges rheological behaviour. Water Sci. Technol. 36 (11), 61e67. Mu, Y., Yu, H.-Q., 2006. Rheological and fractal characteristics of granular sludge in an upflow anaerobic reactor. Water Res. 40, 3596e3602. Seyssiecq, I., Ferrasse, J.-H., Roche, N., 2003. State-of-the-art: rheological characterization of wastewater treatment sludge. Biochem. Eng. J. 16, 41e56. Shih, W.H., Shih, W.Y., Kim, S.I., Lin, J., Aksay, I.A., 1990. Scaling behavior of the elastic properties of colloidal. Phys. Rev. A 42, 4772e4779. Slatter, P.T., 1997. The rheological characterization of sludges. Water Sci. Technol. 36 (11), 9e18. Steffe, J.F., 1996. Rheological Methods in Food Process Engineering. Freeman Press, USA. Tixier, N., Guibaud, G., Baudu, M., 2003. Determination of some rheological parameters for the characterization of activated sludge. Bioresour. Technol. 90, 215e220. Turchiuli, C., Fargues, C., 2004. Influence of structural properties of alum and ferric flocs on sludge dewaterability. Chem. Eng. J. 103, 123e131. Vreeker, R., Hoekstra, L.L., den Boer, D.C., Agterof, W.G.M., 1992. The fractal nature of fat crystal networks. Colloids Surf. 65, 185e189. Wang, Y.L., Dentel, S.K., 2010. The effect of high speed mixing and polymer dosing rates on the geometric and rheological characteristics of conditioned anaerobic digested sludge (ADS). Water Res. 44, 6041e6052. Wang, Y.L., Lu, J., Du, B.Y., Shi, B.Y., Wang, D.S., 2009. Fractal analysis of polyferric chloride-humic acid (PFC-HA) flocs in different topological spaces. J. Environ. Sci. 21, 41e48. Wu, C.C., Huang, C., Lee, D.J., 1997. Effects of polymer dosage on alum sludge dewatering characteristics and physical properties. Colloids Surf. A: Physicochem. Eng. Aspects 122, 89e96.
3882
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 7 1 e3 8 8 2
Wu, C.C., Wu, J.J., Huang, R.Y., 2003. Floc strength and dewatering efficiency of alum sludge. Adv. Environ. Res. 7, 617e621. Wu, R.M., Lee, D.J., Waite, T.D., Guan, J., 2002. Multilevel structure of sludge flocs. J. Colloid Interface Sci. 252, 383e392. Yen, P.S., Chen, L.C., Chien, C.Y., Wu, R.M., Lee, D.J., 2002. Network strength and dewaterability of flocculated activated sludge. Water Res. 36, 539e550. Yen, P.-S., Lee, D.J., 2003. Network strength of waste activated sludge using rheological tests. Separ. Sci. Technol. 38 (2), 375e388.
Zhao, Y.Q., 2002. Enhancement of alum sludge dewatering capacity by using gypsum as skeleton builder. Colloids Surf. A: Physicochem. Eng. Aspects 211, 205e212. Zhao, Y.Q., 2003. Correlations between floc physical properties and optimum polymer dosage in alum sludge conditioning and dewatering. Chem. Eng. J. 92, 227e235. Zhao, Y.Q., 2004. Settling behaviour of polymer flocculated watertreatment sludge II: effects of floc structure and floc packing. Sep. Purif. Technol. 35, 175e183.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 8 3 e3 8 8 9
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Degradation of 2,4,5-trichlorophenoxyacetic acid by a novel Electro-Fe(II)/Oxone process using iron sheet as the sacrificial anode Y.R. Wang, W. Chu* Department of Civil and Structural Engineering, Research Centre for Urban Environmental Technology and Management, The Hong Kong Polytechnic University, Hung Hom, Kowloon, Hong Kong
article info
abstract
Article history:
A novel electrochemically enhanced advanced oxidation process for the destruction of
Received 3 December 2010
organic contaminants in aqueous solution is reported in this study. The process involves
Received in revised form
the use of an iron (Fe) sheet as sacrificial anode and a graphite bar as cathode. In the
2 April 2011
oxidation process, once an electric current is applied between the anode and the cathode,
Accepted 19 April 2011
a predetermined amount of Oxone is added to the reactor. Ferrous ions generated from the
Available online 23 April 2011
sacrificed Fe anode mediate the generation of highly powerful radicals (SO4) through the decomposition of Oxone. The coupled process of Fe(II)/Oxone and electrochemical treat-
Keywords:
ment (Electro-Fe(II)/Oxone) was evaluated in terms of 2,4,5-Trichlorophenoxyacetic acid
2,4,5-trichlorophenoxyacetic acid
degradation in aqueous solution. Various parameters were investigated to optimize the
Electro-Fe(II)/Oxone
process, including applied current, electrolyte and Oxone concentration. In addition, low
Sacrificial anode
solution pH facilitates the system performance due to the dual effects of weak Fenton
Sulfate radicals
reagent generation and persulfate ions generation, whereas the system performance was
Transition metals
inhibited at basic pH levels through non-radical self-dissociation of Oxone and the formation of ferric hydroxide precipitates. Furthermore, the active radicals involved in the Electro-Fe(II)/Oxone process were also identified. The Electro-Fe(II)/Oxone process demonstrates a very high 2,4,5-T degradation efficiency (over 90% decay within 10 min), which justifies the novel Electro-Fe(II)/Oxone a promising treatment process for herbicide removal in water. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
2,4,5-Trichlorophenoxyacetic acid (2,4,5-T) developed in the late 1940s is a chlorophenoxy acetic acid herbicide. It has been used worldwide on a large scale in agriculture to control the growth of broad-leaved weeds on cereal crops, grasslands, lawns (Brillas et al., 2004), and in post-emergence applications (Chang et al., 1998). Although the use of 2,4,5-T was banned in some developed countries due to its toxicity concerns, it’s still
widely used nowadays in most developing countries for controlling weeds and enhancing crops yield (Chaudhary et al., 2009, de Lipthay et al., 2007). 2,4,5-T is considered to be less biodegradable than its analogous herbicide 2,4-D, and has greater resistance to microbial metabolism likely stemming from the additional chlorine constituent on the aromatic ring (Chaudhary et al., 2009). Thus, 2,4,5-T can be detectable in both surface water and groundwater not only during the application of the herbicide, but also after a long period of
* Corresponding author. Tel.: þ852 2766 6075; fax: þ852 2334 6389. E-mail address:
[email protected] (W. Chu). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.04.034
3884
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 8 3 e3 8 8 9
usage due to its comparatively high resistance to biodegradation in the environment. Various methods, including adsorption, biological degradation and advanced oxidation processes (AOPs) have been investigated and developed to degrade 2,4,5-T. Among these emerging treatment technologies, AOPs are considered effective in water treatment applications (Chan and Chu, 2009). Several AOPs have been explored for the degradation of 2,4,5-T, such as Electro-Fenton process (Brillas et al., 2004), Photoelectro-Fenton process (Boye et al., 2003), photolyticelectrolytic system (Chaudhary et al., 2009) and UV/TiO2 (Hemant et al., 2007). In recent years, existing studies have shown that sulfate radicals (SO4) generated by the combination of persulfate or peroxymonosulfate with transition metals (Fe2þ, Agþ, Co2þ) are promising in the degradation of organic contaminants. Moreover, sulfate radicals demonstrate higher standard reduction potential than hydroxyl radicals at neutral pH and are more selective for oxidation than hydroxyl radicals at acidic pH (Anipsitakis and Dionysiou, 2004). As a result, sulfate radical based-advanced oxidation processes (SR-AOPs) have been a great interest for researchers in water treatment. In this study, a new approach named as “Electro-Fe(II)/ Oxone (EFO)” process by coupling electrochemical process and ferrous mediated activation of Oxone to generate sulfate radicals was investigated. In the EFO, soluble ferrous ions were continuously released to the solution from the surface of an iron electrode via anodic oxidation. Once the Oxone is added into the system, the sulfate radicals are generated through the catalysis of transition metal as follows: Fe/Fe2þ þ 2e ðAnodeÞ
(1)
3þ Fe2þ þ HSO þ SO 5 /Fe 4 þ OH
(2)
The counter reaction, which simultaneously occurs at the cathode, is the reduction of ferric ions (Isarain-Chavez et al., 2010), shown as follows: Fe3þ þ e /Fe2þ ðCathodeÞ
(3)
One major benefit of this process is that ferrous ions are electrochemically produced in the process from the sacrificial iron anode, which is reported to be a less energy-demanding reaction (Figueroa et al., 2009). Moreover, the cathodic reduction reaction enhances the regeneration of Fe2þ, whereas in a traditional Fe(II)/Oxone process, the rapid depletion as well as the slow regeneration of Fe(II) usually terminates the production of sulfate radicals. The immediate regeneration of Fe(II) in EFO ensures a smooth and continuous generation of sulfate radicals and makes the best utilization of oxidant. The EFO also has the advantage in minimizing the transportation and utilization problems of commercial ferrous salt. In this study, the performance of the EFO to degrade 2,4,5-T was investigated through the examination of various operational parameters, including applied current, electrolyte Na2SO4, contaminant concentration, dosage of oxidant, and initial solution pH, by which the optimal experimental conditions were identified. In addition, radicals quenching study was also carried out to identify the predominant active radicals involved in the EFO process.
2.
Materials and methods
2.1.
Chemicals reactor operating conditions
Chemicals used in this study, including the probe compound, 2,4,5-T at 97% purity, Oxone (2KHSO5$KHSO4$K2SO4, DuPont product, 95%, where 1 mol Oxone provides 2 mol HSO5), ferrous sulfate (FeSO4$7H2O, 99.0%), sodium sulfate (Na2SO4, 99.0%) and tert-butanol (C4H9OH, 99.5%) were purchased from Sigma Aldrich Inc. (USA). They were used as received without further purification, and all solvents used are of HPLC grade. All solutions were prepared in deionized and distilled water with a resistivity of 18.2 MU obtained from a Bamstead NANOpure water treatment system (Thermo Fisher Scientific Inc., USA). Acetonitrile of HPLC grade purchased from Tedia company, Inc. was degassed before being used in high performance liquid chromatography (HPLC). Sulfuric acid and sodium hydroxide were used to adjust the initial solution pH and methanol of HPLC grade was used as the quenching agent.
2.2.
Electrolytic cell systems
All the preparations and experiments were performed in an air-conditioned laboratory at 23 2 C. The tests were duplicated and average values were used in presenting the results. Oxone solution was freshly prepared before use. The EFO experiments were conducted under constant current electrolysis conditions in a 50 mL undivided single-compartment glass cell by using commercial iron (Fe) sheet with total surface area of ～10 cm2 as the sacrificial anode and a graphite bar as the cathode. An Agilent E3641A DC potentionstatgalvanostat power supply was employed to provide the constant current. The probe (0.10 mM 2,4,5-T) and electrolyte (0.05 M Na2SO4) were used without pH adjustment, unless otherwise stated. The reaction is initiated by adding an appropriate amount of Oxone into the reactor once a specified constant electrical current is applied between the anode and cathode. The solution was continuously agitated by a magnetic stirrer to ensure homogeneity throughout the reaction. Exact amount of an aliquot (0.5 mL) was withdrawn from the solution at predetermined time intervals and mixed with the same amount of methanol to quench the reaction, and the samples were then filtered with 0.45 mm PTFE filters (Whatman) before the LC analysis.
2.3.
Analysis procedures
For the quantification of 2,4,5-T degradation, the remaining 2,4,5-T was analyzed by a High Performance Liquid Chromatography (HPLC) consisting of a Waters 515 HPLC pump, a Waters 2489 UV/Visible Detector, and a Water 717 plus Autosampler. The chromatographic separations were performed by using a C18 reversed phase column (250 mm 4.6 mm with i.d. of 5 mm). A mixture of acetonitrile, water and phosphoric acid in the ratio of 50:50:0.05 (V/V/V) was selected as the mobile phase with a flow rate of 1.2 mL/ min and an injection volume of 10 mL. The UV/Visible detector was set at 289 nm, which was the maximum absorbance wavelength of 2,4,5-T determined by scanning its spectra from
3885
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 8 3 e3 8 8 9
Results and discussion
3.1.
Comparative study of different processes
The following tests were carried out to evaluate the efficiency of 2,4,5-T degradation using EFO: (1) sole-Oxone as oxidant, (2) Fe(II)-mediated Oxone process by adding ferrous sulfate as the source of Fe(II) catalyst, (3) sole-electrolytic process using iron as anode and graphite as cathode, (4) Electro/Oxone process using graphite and iron as anode and cathode, respectively, and (5) EFO. The results are compared in Fig. 1. It can be obviously seen that no appreciable herbicide elimination was observed when Oxone or Electrolytic process was separately applied. A slow and incomplete 2,4,5-T decay (about 50%) was achieved by the combined Electro/Oxone process within 30 min, which demonstrates that the electrolysis can also be an useful alternative to activate the formation of sulfate radicals (from peroxymonosulfate) in addition to the conventional radiolysis, photolysis, thermal activation, and transition metal catalysis. The activation reactions are shown in Eqs. (5) and (6). HSO 5 þ e /SO4 þ OH
(5)
2 HSO 5 þ e /SO4 þ OH
(6)
A better herbicide removal efficiency (around 57%) was achieved using the Fe(II)-mediated Oxone process by applying 1.0 mM [Fe(II)] (the optimal dosage determined separately). In contrast, the EFO demonstrates the fastest and nearly complete herbicide degradation within 10 min. The exceptional performance of the EFO can be rationalized as follows.
3.2.
Effect of applied current
The effect of applied current on the degradation of 2,4,5-T was investigated in the range of 5e30 mA at current-controlled conditions without pH adjustment. The results are presented in Fig. 2, where a significant jump of decay rate (inset) was observed when the applied current increased from 5 to 10 mA. However, the decay rate leveled off (at around 0.24 min1) as the current further increased over 15 mA. The initial rate increment at lower currents is likely attributed to the elevated ferrous ion concentration nearby the anode as the current increased. A faster ferrous ion generation results in a faster production of sulfate radicals and therefore the faster 2,4,5-T decay. However, as the ferrous ions accumulated to certain level in the solution, the excess ferrous ions may act as sulfate radical scavengers as shown in Eq. (7) (Rastogi et al., 2009). Thus, the decay of the probe is retarded due to the competition of sulfate radicals by excessive Fe(II) in the solution when the applied current was too high. 3þ þ SO2 Fe2þ þ SO 4 /Fe 4
From the electrical energy consumption point of view, the electrical energy required to degrade 0.1 mM 2,4,5-T by EFO at various applied currents was calculated in terms of kWh$m3 using Eq. (8) shown as follows:
1.0
I=5 mA I=10 mA I=15 mA I=20 mA I=30 mA
1.0 0.25 mM Oxone; I=0 mA 0.25 mM Oxone; 1.0 mM Fe(II) 0 mM Oxone; I=10 mA; Iron anode 0.25 mM Oxone; I=10 mA; Graphite anode 0.25 mM Oxone; I=10 mA; Iron anode
0.8
0.8 0.6
C/C0
C/C0
0.6
(7)
0.30 0.25 -1
3.
Firstly, ferrous ions were electrochemically produced in the reactor from the sacrificial iron anode. Upon the reaction between Fe (II) and Oxone, sulfate radicals are generated to degrade the probe, while Fe (II) can be regenerated by the reduction of Fe (III) at cathode, which maintains a higher level of catalyst in the solution. Secondly, the electrolysis itself (i.e. without catalyst) may initiate the activation of peroxymonosulfate, which gives rise to the generation of powerful radical species. Therefore, the synergistic effect by coupling ferrous mediated activation of Oxone with electrochemical activation of peroxymonosulfate facilitates the EFO as an effective process in the generation of highly active radicals.
k (min )
200 to 900 nm using a spectrophotometer Spectronic (R) Genesys. To measure the solution pH, a CD510 digital pH meter was used in this study. The concentration of Fe(II) was determined by spectrophotometric method at 510 nm after complexing with 1,10-phenanthroline using a UVeVis spectrophotometer.
0.20 0.15 0.10 0 5 10 15 20 25 30 35 Applied current (mA)
0.4
0.4
0.2 0.2
0.0 0.0
0
5
10 15 20 Reaction time (min)
25
30
Fig. 1 e 2,4,5-T degradation under various reaction conditions. Experimental conditions were [2,4,5T] [ 0.10 mM; [Na2SO4] [ 0.05 M as supporting electrolyte; No pH adjustment.
0
5
10 15 20 Reaction time (min)
25
30
Fig. 2 e Effect of applied current on 2,4,5-T decay by EFO. The inset shows the kinetic constants of herbicide decay at various applied currents. Experimental conditions were [2,4,5-T] [ 0.10 mM; [Oxone] [ 0.25 mM; [Na2SO4] [ 0.05 M as supporting electrolyte; I [ 10 mA; No pH adjustment.
3886
(8)
1.0
Where U is the voltage measured during the reaction (volt), I is the applied current (A), t is the electrolysis time (h), and V is the volume of reaction solution (m3) (Khataee et al., 2009). The calculated values are summarized in Table 1. As shown, optimal energy consumption was obtained when the applied current was 10 mA. Therefore, to minimize the possible quenching effect at higher current and to achieve efficient energy consumption, applied current of 10 mA was adopted in the rest of this study. Additionally, blank experiment was carried out to investigate the Fe ions (all the Fe ions have been transferred to Fe2þ) generation using 0.05 M Na2SO4 as electrolyte at applied current of 10 mA without the addition of contaminant and oxidant. The results show that the Fe ions smoothly increase from 0 mM (0 min) to 1.06 mM (10 min) and finally to 2.08 mM (30 min) in the solution. Compared with Fe2þ dosages applied in regular electro-Fenton processes or traditional Fenton process (1 mM (Brillas et al., 2004), 15 mM (Wang et al., 2008) and 25 mM (Sun et al., 2009)), the released iron ion in the solution is relatively low.
0.8
Table 1 e The electrical energy required to degrade 0.1 mM 2,4,5-T at various applied current by EFO. Applied current (mA) 5 10 15 20 30
Degradation efficiency (%) 84.5 85.9 84.2 85.5 85.5
Required time (min) 20 7 7 7 7
The concentration of Oxone was 0.25 mM.
Energy consumption (kWh$m3) 0.0155 0.0107 0.0344 0.0717 0.1544
C/C0
0.1000 M Na2SO4 0.30
0.6 0.4 0.2 0.0
0.25 0.20 0.15 0.10 0.00 0.03 0.06 0.09 0.12 [Na2SO4] (M)
0
5
10 15 20 Reaction time (min)
25
30
Fig. 3 e Effect of [Na2SO4] on herbicide decay by EFO. The inset shows the decay rate constants as a function of [Na2SO4]. Experimental conditions were [2,4,5-T] [ 0.10 mM; [Oxone] [ 0.25 mM; I [ 10 mA; No pH adjustment.
3.4.
Effect of 2,4,5-T and oxidant dosage
The efficiency of EFO as a function of the initial herbicide concentration was investigated by varying [2,4,5-T] from 0.025 to 0.200 mM. The results in terms of pseudo first-order kinetics are illustrated in Fig. 4. As shown in the inset, the decay rate constant linearly decreases as the initial [2,4,5-T] increases. In addition, the removal percentage of 2,4,5-T decreases from 100 to 69% (at 10 min) when [2,4,5-T] increases from 0.025 to 0.200 mM. To elucidate the effect of [Oxone] in EFO, the initial [Oxone] was increased from 0.0625 to 0.5000 mM and the final degradation of 2,4,5-T increased from 49 to 97% as shown in Fig. 5. In general, the higher the oxidant dosage is, the better the herbicide removal efficiency shall be. However, it is interesting to note that the initial decay rates increase with the increment of [Oxone] from 0.0625 to 0.250 mM (i.e. optimized at 0.250 mM), and then the initial rate is retarded as [Oxone] further
0 -1
0.025 mM 0.050 mM 0.100 mM 0.150 mM 0.200 mM
-2 -3
-1
Electrolyte concentration directly affects the electrical conductivity of the solution. In this study, the influence of electrolyte concentration on the degradation of 2,4,5-T was examined by varying [Na2SO4] (0.006e0.100 M) while keeping other parameters unchanged. The results are demonstrated in Fig. 3 with inset showing the pseudo first-order rate constants of 2,4,5-T decay versus [Na2SO4]. It is found that the degradation rate increases linearly with [Na2SO4] increasing from 0.006 to 0.050 M, then levels off at higher [Na2SO4]. The linear increment from low to mid [Na2SO4] is ascribed to the higher electrolyte concentration in the solution, leading to a higher electrical conductivity and therefore a higher voltage when the current is kept constant. However, if the voltage is too high, the side reactions of O2 and H2 evolution occur, reducing the current efficiency and then leading to the decreased decay rate. In addition, the decay rate may also be retarded with the increment of [Na2SO4], which was justified by another test (data not shown here). Thus, 0.05 M Na2SO4 was chosen as the optimal electrolyte concentration and used in this study.
0.0500 M Na2SO4
k (min )
Effect of electrolyte concentration
0.0125 M Na2SO4 0.0250 M Na2SO4
Ln(C/C0)
3.3.
0.0060 M Na2SO4
-1
UIt 103 V
k (min )
E¼
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 8 3 e3 8 8 9
-4 -5
0
0.5 0.4 0.3 0.2 0.1 0.0
0.0 0.1 0.2 0.3 [2,4,5-T] (mM)
2 4 6 Reaction time (min)
8
Fig. 4 e Degradation of various initial [2,4,5-T] by EFO. The inset shows the decay rate constants as a function of [2,4,5-T]. Experimental conditions were [Oxone] [ 0.25 mM; [Na2SO4] [ 0.05 M as supporting electrolyte; I [ 10 mA; No pH adjustment.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 8 3 e3 8 8 9
1.0
0.0625 mM Oxone 0.1250 mM Oxone 0.2500 mM Oxone 0.3750 mM Oxone 0.5000 mM Oxone
-1
k (min )
0.30
0.8
0.25 0.20 0.15 0.10 0.0
C/C0
0.6
0.2 0.4 0.6 [Oxone] (mM)
0.4 0.2 0.0
0
5
10 15 20 Reaction time (min)
25
30
Fig. 5 e Degradation of 2,4,5-T with different [Oxone] by EFO. Experimental conditions were [2,4,5-T] [ 0.10 mM; [Na2SO4] [ 0.05 M as supporting electrolyte; I [ 10 mA; No pH adjustment.
increased (see inset of Fig. 5). This observation is inconsistent with previous findings; for example, the decolorization rate of acid red 88 follows zero-order with respect to [Oxone] in a Co2þ/ Oxone process (Madhavan et al., 2009). The divergence is possibly caused by the different treatment processes. The slow transformation of Co3þ to Co2þ could be a rate-limiting factor in Co2þ/Oxone system (Huang et al., 2009), while it is not the case for EFO due to the fact that ferrous ions can be effectively regenerated through ferric ion reduction at the cathode. In addition, the mechanism of colour (i.e. chromophore) removal is different from that of a molecular decay. When the [Oxone] is higher than 0.250 mM (i.e. the optimal dosage in terms of initial decay rate), although the decay rate is retarded, the overall removal of 2,4,5-T at the end of the process is still better than that at the optimal dosage. This suggests that the increase of [Oxone] can still improve the overall removal of 2,4,5-T if an extended reaction time is allowed. However, the performance in the utilization of the valuable sulfate radicals is apparently not optimized. This can be explained by the unfavorable consumption of SO4 radicals by the excessive HSO5, which leads to the scavenging of sulfate radicals, generating of less reactive SO5, and thus slower decay rates, as explained by Eq. (9) (Ling et al., 2010; Madhavan et al., 2009), 2 þ HSO 5 þ SO4 /SO5 þ SO4 þ H
3887
efficiency with the change of solution pH level. Pseudo firstorder kinetic was observed with high linear regression coefficient (r2) above 0.95 at different pH levels. As shown in the inset of Fig. 6, the herbicide decay rate sharply increases as the solution pH decreases from 4.47 to 1.50, while the rate slows down and levels off for pH levels higher than 4.47. At first glance, it looks similar to the conventional Fenton’s process, which is generally regarded to have better performance at acidic condition. However, the Fenton’s process is inhibited at extremely acidic environment due to the formation of (Fe(II)(H2O))2þ (pH < 2.5) (Gogate and Pandit, 2004) and Fe(III)hydroxyl complexes, Fe(OH)2þ (pH < 3) (Pignatello et al., 1999). The former reduces the availability of free Fe(II), while the latter inhibits the regeneration of Fe(II) from Fe (III). However, these negative effects are not found in EFO, which performs well at the extreme acidic condition. To further verify the validity of the above observation at the low pH level, additional tests were conducted at the initial pH of 1.5 without Oxone addition and with/without radical scavenger. As shown in Fig. 7, the Electro-Fe(II) (i.e. without Oxone) at pH level of 1.5 can remove 40% of 2,4,5-T at 30 min, which is much higher than that at neutral pH (about 5%, see Fig. 1). This indicates that the low pH level can improve the process performance. In this study, two typical radical quenching agents, methanol and t-butyl alcohol, were employed to investigate the mechanism of Electro-Fe(II) process. As demonstrated in Fig. 7, the presence of 1 M methanol or 1 M t-butyl alcohol significantly retards the herbicide degradation, indicating that 2,4,5-T decay is dominated by the oxidation of radical species. The degradation of 2,4,5-T in the presence of t-butyl alcohol is slightly higher than that of methanol (with a ratio of 2:1 in terms of removal %), indicating the presence of both hydroxyl and sulfate radicals in Electro-Fe(II) process. This is because methanol contains ahydrogen that can rapidly quench hydroxyl and sulfate radicals. The t-butyl alcohol however, as a weaker quencher for SO4, selectively quenches hydroxyl radicals (Anipsitakis and
(9)
Because of the overdose of Oxone, the HSO5, SO4 and SO5 radicals remaining in the solution are still in good quantities, and the overall removal of 2,4,5-T is not affected as long as the reaction time is sufficient. In practice, however, an unlimited addition of Oxone in EFO is not suggested and the determination of an optimal dosage is critical for a costeffective application.
3.5.
Effect of initial pH
The effect of initial pH levels on the EFO was also explored. Fig. 6 illustrates the corresponding 2,4,5-T degradation
Fig. 6 e Effect of initial solution pH on the degradation of 2,4,5-T. The inset shows the decay rate constants as a function of solution pH. Experimental conditions were [2,4,5-T] [ 0.10 mM; [Oxone] [ 0.25 mM; [Na2SO4] [ 0.05 M as supporting electrolyte; I [ 10 mA.
3888
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 8 3 e3 8 8 9
0.8
0.8
0.6
0.6 pH=1.5_No Oxone pH=1.5_No Oxone_1 M Methanol pH=1.5_No Oxone_1 M t-butyl alcohol
0.4
0.4 0.2
0.2 0.0
0
10
20 30 40 Reaction time (min)
50
0.0
60
Fig. 7 e Degradation of herbicide by Electro-Fe(II) system in pH [ 1.5 solution without the presence of Oxone. Experimental conditions were [2,4,5-T] [ 0.10 mM; [Na2SO4] [ 0.05 M as supporting electrolyte; I [ 10 mA.
Dionysiou, 2004) and results in about 2% (see Fig. 7) higher removal of 2,4,5-T than methanol. It is therefore very likely that in the Electro-Fe(II) process, Fenton’s reagents are generated in the acidic electrolysis medium while utilizing iron as anode and graphite as cathode. Simultaneously, S2O82 can be generated from the oxidation of SO42 and HSO4 ions (pka2 of H2SO4 is 1.26 102 so [SO42]/[HSO4] ¼ 0.42 at pH 1.5) as sulfate medium is used as the electrolyte (Brillas et al., 2009). Therefore, the following reactions (Eqs. (10e14)) may be responsible for the radical species’ generation in the Electro-Fe(II) process: O2 þ 2Hþ þ 2e /H2 O2
(10)
Fe2þ þ H2 O2 þ Hþ /Fe3þ þ H2 O þ ,OH
(11)
2 2SO2 4 /S2 08 þ 2e
(12)
2 þ 2HSO 4 /S2 08 þ 2H þ 2e
(13)
2 3þ þ SO Fe2þ þ S2 O2 8 /Fe 4 þ SO4
(14)
In Fig. 6, the decay rate is gradually leveled off from pH 4.47 to 10.31. This is a result of the formation of ferric hydroxide precipitates, Fe(OH)3(s), at higher pH level. It not only leads to the decrease of dissolved Fe(II) and Fe(III), but also retards the regeneration of ferrous catalyst by partially coating on the electrode surface, which in turn significantly reduces the herbicide degradation efficiency. In addition, Oxone is unstable at basic conditions and its self-dissociation was observed mainly through non-radical pathways, which also contributes to the retardation of the reaction rate.
3.6.
Without radical scavenger 1 M Methanol 1 M Ethanol 1 M t-butyl alcohol 2 M Methanol
C/C0
1.0
C/C0
1.0
Radicals quenching study
To investigate the 2,4,5-T decay mechanism in EFO, three radical quenching agents were utilized to evaluate the contribution of various radicals or oxidizing species to the
0
5
10 15 20 Reaction time (min)
25
30
Fig. 8 e Effect of different radical quenching agents on herbicide degradation by EFO. Experimental conditions were [2,4,5-T] [ 0.10 mM; [Oxone] [ 0.25 mM; [Na2SO4] [ 0.05 M as supporting electrolyte; I [ 10 mA. No pH adjustment.
process. As indicated in Fig. 8, the addition of 1 M methanol or 1 M ethanol significantly inhibits the system performance with herbicide decay efficiency decreased from 85.9 to 12.9 and 14.3% in 7 min, respectively, whereas the addition of 1 M tbutyl alcohol only results in 53% herbicide decay drop compared with that of 72.6 and 69.1% by the addition of methanol and ethanol, respectively. The differences in the herbicide decay drop by the three radical scavengers imply the involvement of both hydroxyl and sulfate radical in the system as discussed previously. Nevertheless, the 2,4,5-T was not fully quenched suggesting the possibility of insufficient quenching agent and/or self-dissociation of oxidant through non-radical pathway (Rastogi et al., 2009) in the solution. To verify their contribution, the methanol was further increased to 2 M and the decay of 2,4,5-T was almost stopped, which suggests the former is likely the main cause. Therefore, the decay of herbicide 2,4,5-T in EFO is dominated by radical based mechanisms.
4.
Conclusions
In this study, the electrochemically enhanced transition metal-activated Oxone process, EFO, was experimentally verified to be very effective in degrading 2,4,5-T with over 90% removal in 10 min. Degradation of herbicide 2,4,5-T by EFO was observed to follow pseudo first-order kinetics. The effects of various operational parameters of EFO were also examined to optimize the process. Based on the test results, the optimal applied current was determined to be 10 mA, in which both the herbicide removal efficiency and energy consumption were optimized. For the effect of [2,4,5-T], experimental results showed that (1) the degradation efficiency for 0.025, 0.05, 0.1, 0.15 and 0.2 mM 2,4,5-T was 100, 98, 88, 78 and 69%, respectively; (2) 2,4,5-T decay rate constant linearly decreases with the increment of initial herbicide concentration. For the effect of Oxone dosage, the test results demonstrate that 0.25 mM Oxone exhibits the optimal 2,4,5-T decay rate and an
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 8 3 e3 8 8 9
unlimited increment of [Oxone] is not favorable to the EFO. Moreover, it was found that herbicide decay rate sharply increases as the solution pH decreases. A favorable dual effect was subsequently conducted to elucidate this performance enhancement at acidic conditions. Radicals quenching study revealed the presence of both hydroxyl and sulfate radicals in the process. Generally, a highly effective radicals basedadvanced oxidation process was investigated for the degradation of 2,4,5-T in this study. Compared with other AOPs, the proposed EFO has the following advantages: (1) in-situ generation of Fe(II), (2) acceleration of regenerating Fe(II) from Fe(III), (3) minimization of the transportation and utilization problems of commercial ferrous salt, and (4) a less energy-demanding reaction, indicating the proposed EFO is a promising process.
Acknowledgements The work described in this study was fully supported by a grant from the University Research Fund of the Hong Kong Polytechnic University (RPS6).
references
Anipsitakis, G.P., Dionysiou, D.D., 2004. Radical generation by the interaction of transition metals with common oxidants. Environmental Science & Technology 38 (13), 3705e3712. Boye, B., Dieng, M.M., Brillas, E., 2003. Electrochemical degradation of 2,4,5-trichlorophenoxyacetic acid in aqueous medium by peroxi-coagulation. Effect of pH and UV light. Electrochimica Acta 48 (7), 781e790. Brillas, E., Boye, B., Sires, I., Garrido, J.A., Rodriguez, R.M., Arias, C., Cabot, P.L., Comninellis, C., 2004. Electrochemical destruction of chlorophenoxy herbicides by anodic oxidation and electroFenton using a boron-doped diamond electrode. Electrochimica Acta 49 (25), 4487e4496. Brillas, E., Sires, I., Oturan, M.A., 2009. Electro-Fenton process and related electrochemical technologies based on Fenton’s reaction chemistry. Chemical Reviews 109 (12), 6570e6631. Chan, K.H., Chu, W., 2009. Degradation of atrazine by cobaltmediated activation of peroxymonosulfate: different cobalt counteranions in homogenous process and cobalt oxide catalysts in photolytic heterogeneous process. Water Research 43 (9), 2513e2521. Chang, B.V., Liu, J.Y., Yuan, S.Y., 1998. Dechlorination of 2,4dichlorophenoxyacetic acid and 2,4,5-trichlorophenoxyacetic acid in soil. Science of the Total Environment 215 (1e2), 1e8. Chaudhary, A.J., Hassan, M.U., Grimes, S.M., 2009. Simultaneous recovery of metals and degradation of organic species: copper
3889
and 2,4,5-trichlorophenoxyacetic acid (2,4,5-T). Journal of Hazardous Materials 165 (1e3), 825e831. de Lipthay, J.R., Sorensen, S.R., Aamand, J., 2007. Effect of herbicide concentration and organic and inorganic nutrient amendment on the mineralization of mecoprop, 2,4-D and 2,4,5-T in soil and aquifer samples. Environmental Pollution 148 (1), 83e93. Figueroa, S., Vazquez, L., Alvarez-Gallegos, A., 2009. Decolorizing textile wastewater with Fenton’s reagent electrogenerated with a solar photovoltaic cell. Water Research 43 (2), 283e294. Gogate, P.R., Pandit, A.B., 2004. A review of imperative technologies for wastewater treatment I: oxidation technologies at ambient conditions. Advances in Environmental Research 8 (3e4), 501e551. Hemant, K.S., Saquib, Mohd, Haque, M.M., 2007. Titanium dioxide mediated photocatalysed degradation of phenoxyacetic acid and 2,4,5-trichlorophenoxyacetic acid, in aqueous suspensions. Journal of Molecular Catalysis A-Chemical 264, 66e72. Huang, Y.H., Huang, Y.F., Huang, C.I., Chen, C.Y., 2009. Efficient decolorization of azo dye reactive black B involving aromatic fragment degradation in buffered Co2þ/PMS oxidative processes with a ppb level dosage of Co2þ-catalyst. Journal of Hazardous Materials 170 (2e3), 1110e1118. Isarain-Chavez, E., Arias, C., Cabot, P.L., Centellas, F., Rodriguez, R.M., Garrido, J.A., Brillas, E., 2010. Mineralization of the drug beta-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. Applied Catalysis B-Environmental 96 (3e4), 361e369. Khataee, A.R., Vatanpour, V., Ghadim, A.R.A., 2009. Decolorization of CI Acid Blue 9 solution by UV/Nano-TiO2, Fenton, Fenton-like, electro-Fenton and electrocoagulation processes: a comparative study. Journal of Hazardous Materials 161 (2e3), 1225e1233. Ling, S.K., Wang, S.B., Peng, Y.L., 2010. Oxidative degradation of dyes in water using Co2þ/H2O2 and Co2þ/peroxymonosulfate. Journal of Hazardous Materials 178 (1e3), 385e389. Madhavan, J., Maruthamuthu, P., Murugesan, S., Ashokkumar, M., 2009. Kinetics of degradation of acid red 88 in the presence of Co2þ-ion/peroxomonosulphate reagent. Applied Catalysis AGeneral 368 (1e2), 35e39. Pignatello, J.J., Liu, D., Huston, P., 1999. Evidence for an additional oxidant in the photoassisted Fenton reaction. Environmental Science & Technology 33 (11), 1832e1839. Rastogi, A., Ai-Abed, S.R., Dionysiou, D.D., 2009. Sulfate radicalbased ferrous-peroxymonosulfate oxidative system for PCBs degradation in aqueous and sediment systems. Applied Catalysis B-Environmental 85 (3e4), 171e179. Sun, J.H., Li, X.Y., Feng, J.L., Tian, X.K., 2009. Oxone/Co2þ oxidation as an advanced oxidation process: comparison with traditional Fenton oxidation for treatment of landfill leachate. Water Research 43 (17), 4363e4369. Wang, C.T., Hu, J.L., Chou, W.L., Kuo, Y.M., 2008. Removal of color from real dyeing wastewater by Electro-Fenton technology using a three-dimensional graphite cathode. Journal of Hazardous Materials 152 (2), 601e606.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 9 0 e3 8 9 8
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
A novel scenario for biofouling control of spiral wound membrane systems J.S. Vrouwenvelder a,*, M.C.M. Van Loosdrecht b, J.C. Kruithof a a b
Wetsus, Centre of Excellence for Sustainable Water Technology, Agora 1, P.O. Box 1113, 8900 CC Leeuwarden, The Netherlands Delft University of Technology, Department of Biotechnology, Julianalaan 67, 2628 BC Delft, The Netherlands
article info
abstract
Article history:
Current strategies to control biofouling in nanofiltration and reverse osmosis membrane
Received 5 November 2010
systems such as chemical cleaning and use of low fouling membranes are not always
Received in revised form
successful. Based on recent studies, an alternative approach is derived, combining a lower
19 April 2011
linear flow velocity in lead modules and adapted designs for feed spacer with an advanced
Accepted 26 April 2011
cleaning strategy. This approach can be realized by small adaptations in current plant
Available online 4 May 2011
design. A lower linear flow velocity in lead spiral wound membrane modules results in (i) lower energy use, (ii) lower impact of biomass on the feed channel pressure drop, and (iii)
Keywords:
more fluffy biofilm that may be easier to remove from the lead membrane modules,
RO
especially when adapted feed spacers combined with a reversed enhanced flush are
NF
applied. This rational scenario can result in effective biofouling control at low energy
Membrane filtration
requirements, minimal chemical use and minimal cost.
Environmentally friendly biofouling
ª 2011 Elsevier Ltd. All rights reserved.
control Hydrodynamics Reversed enhanced flow Biofouling removal
1.
Introduction
High quality drinking water from water sources including seawater and sewage can be produced with membrane filtration processes like nanofiltration (NF) and reverse osmosis (RO). Because the global demand for clean freshwater is growing, these membrane technologies are increasing in importance. One of the most serious problems in NF and RO applications is biofouling - excessive growth of biomass - affecting the performance of these membrane systems, influencing the (i) amount and quality of the produced fresh water and/or (ii) reliability of water production and (iii) costs (Ridgway and Flemming, 1996; Shannon et al., 2008). Numerous authors
describe biofouling problems in membrane installations (Ridgway et al., 1983; Flemming et al., 1994; Tasaka et al., 1994; Ridgway and Flemming, 1996; Baker and Dudley, 1998; Schneider et al., 2005). In the Middle East, about 70% of the seawater RO membrane installations suffer from biofouling problems (Gamal Khedr, 2000). In spiral wound membrane modules, two types of pressure drop can be distinguished: the trans-membrane pressure drop (TMP) and the feed channel pressure drop (FCP), the pressure drop between feed and concentrate lines (Flemming et al., 1994). The TMP is the differential pressure between feed and permeate lines, caused by the frictional resistance over the membrane. When the TMP is increased by biofouling, the membrane flux is declined.
* Corresponding author. Tel.: þ31(0)58 2843000; fax: þ31(0)58 2843001. E-mail addresses:
[email protected],
[email protected] (J.S. Vrouwenvelder). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.04.046
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 9 0 e3 8 9 8
Biofouling in spiral wound NF and RO membranes has been studied on monitor, test-rig, pilot and full-scale for extensively pretreated water (Vrouwenvelder et al., 2009b,c). The monitor, named membrane fouling simulator (MFS), used in our research efforts has shown to be representative for spiral wound membrane modules used in practice (Vrouwenvelder et al., 2006). Identical behavior with respect to biofouling and FCP development was observed in membrane elements in the same position in a nanofiltration installation operated with and without flux. Irrespective whether a flux was applied or not, the FCP and biofilm concentration increased. Calculations on mass transfer aspects supported the observations that the flux is not playing a significant role in substrate supply to the fouling layer. Also, test-rig and full-scale studies with different types of feed water showed that biofouling of membrane modules correlated very well with FCP-increase (Vrouwenvelder et al., 2008). Moreover, in systems suffering from biofouling cleaning cycles are governed by the pressure drop over the feed channel. Therefore, biofouling is considered an FCP problem (Vrouwenvelder et al., 2009c). Thus, the MFS was operated under cross flow conditions without permeate production. This operation gives almost identical results as fouling in a full-scale membrane module (Vrouwenvelder et al., 2006). Magnetic resonance imaging (MRI) studies showed that even restricted biofilm accumulation on the feed channel spacer influenced the velocity distribution profile strongly (Graf von der Schulenburg et al., 2008; Vrouwenvelder et al., 2009c). The feed spacer presence strongly influenced the FCP-increase caused by biofilm accumulation: in both spiral wound NF and RO systems biofouling is dominantly a feed spacer problem. Over the years, strategies to control biofouling have not always been successful. In the late 1990s, two strategies were pursued to prevent and control membrane biofouling: (i) physical removal of bacteria from the feed water of membrane systems (for example by microfiltration or ultrafiltration pretreatment), and (ii) metabolic inactivation of bacteria by applying biocide dosage or UV irradiation. At present, the focus is on nutrient removal by biological pre-treatment (e.g. sand filtration) and membrane modification (disinfectant resistant and low fouling). In addition, membrane cleanings are applied. These cleanings are not effective to control biofouling since biomass is not removed from the module (Fig. 1), resulting in rapid bacterial regrowth (Kappelhof et al., 2003; Bereschenko et al., 2011). Use of low fouling membranes do not guarantee fouling control. For prevention of biofouling without biocides it is essential to minimize the availability of nutrients (similar to conditions to minimize regrowth of bacteria in drinking water lines without chlorination). Therefore, a certain pre-treatment will always be needed. Eventually, microbial growth will always happen, therefore it is recommended to design a system in such a way that biofilm formation doesn’t lead to operational problems or that biomass can be easily removed. In the past four years, the study of membrane biofouling has intensified strongly, compared to the previous twenty two years, shown by the more than doubling of the number of scientific papers. Evidently, biofouling is still considered an important problem in practice.
3891
Fig. 1 e Effect of cleaning on the biomass parameters adenosine-tri-phosphate, ATP (A), total direct cell count, TDC (B) and heterotrophic place count, HPC on R2A media (C) in a membrane element. One lead element was taken from the plant directly prior to cleaning and another lead element from a parallel pressure tube was taken directly after the cleaning (Vrouwenvelder et al., 2008).
The goal of this study was to develop a new, more successful strategy to control biofouling. The focus of our studies was on fresh water systems. Fresh water is characterized by having low concentrations of dissolved salts. In case of high salt concentrations like seawater desalination systems, osmotic pressure and concentration polarization plays an much more important role (on the impact of accumulated material) on membrane performance (Herzberg and Elimelech, 2007; Chong et al., 2008).
2.
Impact of feed spacers on Biofouling
Feed spacers commonly used in practice in spiral wound NF and RO membrane modules were characterized on a number of aspects such as material type, structure and thickness. The feed spacers provided by 4 membrane manufacturers were made of polypropylene. The spacers had a diamond-shaped structure (Fig. 2) and thicknesses varying between 26 and 34 mil (1 mil equals 25.4 mm), according to manufacturer specifications. In some cases thicker spacer (45e50 mil) were used in practice. The feed spacers used in practice have similar structure and orientation (Fig. 2).
3. Localization of biofouling in NF/RO installations High biomass concentrations were observed in the first stage NF/RO lead modules during studies in practice (Carnahan et al., 1995; Vrouwenvelder et al., 2008). A systematic study on biofouling development in individual membrane modules and stages of an NF pilot plant showed a pressure drop increase over the lead membrane module higher than the pressure drop increase over the following membrane modules (Fig. 3). Biomass was observed on the feed spacer (Fig. 4). In
3892
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 9 0 e3 8 9 8
Fig. 2 e Inventory of feed spacers commonly used in practice in commercially available spiral wound NF and RO membrane modules by four global membrane manufacturers. The membrane module manufacturers are coded I to IV. The feed spacer thickness is commonly expressed in mil (1 mil equals 25.4 mm) (Vrouwenvelder et al., 2009c).
agreement with observations in practice, most biomass was present in the lead element. Evidently, biofouling develops at the feed side of NF/RO installations.
4.
Effect of linear flow velocity
The relationship between linear flow velocity and pressure drop over spiral wound membrane elements can be calculated
using the methods described by Schock and Miquel (1987). Mathematically, the pressure drop is expressed as r,v2 L , Dp ¼ l, 2 dh where l is the friction coefficient, r the specific liquid density, v the linear velocity, L the length of the membrane module and dh the hydraulic diameter. The friction coefficient is given by the correlation function by Schock and Miquel (1987):
Fig. 3 e Pressure drop in time over individual membrane elements in the first pressure vessel of an NF installation (A) and biomass concentration in module 1, 2 and 6 after 146 day operation (B), The number represents the position of the elements in the pressure vessel from the feed side (1 [ lead module and 6 [ element on the concentrate side of the pressure vessel; Vrouwenvelder et al., 2009c).
3893
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 9 0 e3 8 9 8
Comparative studies with membrane fouling simulators performed with varying linear flow velocities at constant substrate load showed that the amount of accumulated biomass was constant but the pressure drop increase was a function of the linear flow velocity (Fig. 5, Vrouwenvelder et al., 2010). The impact of biomass on the pressure drop is determined by the biomass amount and the linear flow velocity. An inventory of feed flows in practice per lead element is shown in Table 1. Biofilm accumulates most on the location with the highest linear flow velocity, where it has highest impact on the feed channel pressure drop. In other words, biofouling occurs on the location where the effect on performance is strongest.
5. Effect of flow regime on biofilm cohesion strength
Fig. 4 e Biofilm on feed spacer in lead membrane element in the first pressure vessel of an NF installation after 146 day operation (see Fig. 3).
l ¼ 6:23,Re0:3 Where Re is the Reynolds number. Up to eight membrane modules are placed in series in pressure vessels. A tapered configuration of pressure vessels is applied to maintain proper flow velocities along the membrane, minimizing concentration polarization (Mallevialle et al., 1996). At an NF pilot plant, already at the start of the study the pressure drop over individual modules declined with increasing distance to the feed side of the installation (Fig. 3). This pressure drop decline is caused by the reduction of linear flow velocity due to permeate production in the membrane elements. For each element, permeate production is about 10% of the element feed flow. Obviously, the lead module in the pressure vessel contributes most to the total FCP over the membrane installation because of the highest flow rate.
The impact of flow regime on pressure drop, biomass accumulation and morphology was studied. In NF and RO membrane elements, at linear flow velocities as applied in practice voluminous and filamentous biofilm structures developed in the feed spacer channel, causing a significant increase in feed channel pressure drop. The amount of accumulated biomass was depending on the substrate load but was independent of the applied shear. A higher shear force resulted in more compact and less filamentous biofilm structure, causing a lower pressure drop increase. By water flushing, a biofilm grown at low shear was easier to remove than a biofilm grown at high shear (Figs. 6 and 7). The influence of hydrodynamic conditions on biofilm structure and compactness is referenced broadly. In several biofilm studies it has been observed that at high hydrodynamic shear force the biofilm becomes more compact, stable
Table 1 e Inventory of feed flows used in practice per lead membrane element (8 inch diameter) at the membrane installation feed side and linear flow velocity at inlet side of the lead membrane element. 13 different full-scale installations treating different water types were inventoried (Vrouwenvelder et al., 2009a). NF/RO installation
Fig. 5 e Amount of accumulated biomass in the membrane fouling simulator (A) and pressure drop increase (B) after 11 day operation with the same acetate load and different linear flow velocities. Adapted from Vrouwenvelder et al. (2009a).
1 2 3 4 5 6 7 8 9 10 11a 12a 13a
Feed flow per lead element (m3 h1)
Linear flow velocity inlet side lead element (m s1)
11.2 11.1 10.0 9.4 9.4 9.2 8.8 8.6 7.9 7.7 4.9 4.2 3.6
0.20 0.20 0.18 0.17 0.17 0.17 0.16 0.16 0.14 0.14 0.09 0.08 0.07
a full-scale installations using the hydraulic optimized pressure vessel.
3894
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 9 0 e3 8 9 8
load (Peyton, 1996). Filamentous biofilm streamers and hydrodynamics were described in a review of Purevdorj-Cage and Stoodley (2004) and other studies (Stoodley et al., 1999a,b; Hille et al., 2009; Garny et al., 2009). Garny et al. (2009) reported that filamentous bacterial growth was due to the combination of limited substrate availability and high flow rates. Evidently, biofilm formation in spiral wound NF and RO systems is analogous to biofilm formation in other systems. A high shear force results in a thin compact biofilm while a low shear force results in a thick fluffy biofilm. Depending on the process conditions, a fluffy (easy to remove) or a compact (hard to remove) biofilm may be produced. The preferred process condition in view of the best long term operation is under study. Fig. 6 e Effect of flushing (at 0.42 m sL1) on biofouling developed at low and high linear flow velocity (0.06 and 0.31 m sL1) in a membrane fouling simulator containing membranes and feed spacers. Adapted from Vrouwenvelder et al. (2010).
and dense (Van Loosdrecht et al., 1995; Kwok et al., 1998; Pereira et al., 2002). The thickness, structure, stability, adhesive strength and density of the biofilm were influenced by the substrate load and hydrodynamic shear stress/flow regime (Characklis and Marshall, 1990; Van Loosdrecht et al., 1995; Peyton, 1996; Kwok et al., 1998; Pereira et al., 2002; Wijeyekoon et al., 2004; Purevdorj-Cage and Stoodley, 2004; Chen et al., 2005). Thinner biofilms were found during growth of Pseudomonas fluorescens biofilms under turbulent flow conditions compared to laminar flow conditions (Pereira et al., 2002). Thicker biofilms were observed at increasing substrate
6.
Discussion
Feed spacers used in practice in spiral wound membrane systems have similar geometry and orientation (Fig. 2). Biomass accumulates predominantly at the feed side of the membrane installation (Fig. 3), the location with the highest impact on feed channel pressure drop (Fig. 5). Biomass accumulation was observed on the feed spacer (Fig. 4). The linear flow velocity determines the impact of accumulated biomass on the pressure drop (Fig. 5) and the biomass cohesion strength (Fig. 6). These observations were the starting point for developing an integrated scenario for biofouling control. In non-fouled conditions the linear flow velocity determines the pressure drop (Fig. 8). In fouled conditions the accumulated biomass volume increases the effective water flow velocity resulting in a more than proportional pressure
Fig. 7 e Effect of flushing (at 0.42 m sL1) on biofouling developed at low and high linear flow velocity (0.06 and 0.31 m sL1). Biofilm developed at low linear flow velocity (0.06 m sL1) before (A) and after flush (B), and biofilm developed at high linear flow velocity (0.31 m sL1) before (C) and after flush (D) (Vrouwenvelder et al., 2010).
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 9 0 e3 8 9 8
3895
is reduced. This adapted pressure design was developed to reduce pressure loss and improve membrane productivity. Low linear flow velocities in the lead elements obtained with these hydraulic optimized vessels will also favor the reduction of the pressure drop increase caused by biomass accumulation. The experience in practice with low linear flow velocities suggests that membrane systems can be operated with lower linear flow velocities without operational problems like scaling caused by concentration polarization. In this system with hydraulic optimized pressure vessels, the impact of biomass on performance is reduced but biomass is not removed from the membrane installation. Fig. 8 e Pressure drop as function of the linear flow velocity, illustrating that the effect of biomass concentration on pressure drop increase is affected by linear flow velocity enabling to reduce the impact of accumulated biomass on the pressure drop. The difference between the pressure drop in non-fouled conditions and pressure drop in fouled conditions is the pressure drop increase caused by accumulated fouling.
drop increase. The impact of accumulated biomass on the feed channel pressure drop is dependent on the linear flow velocity. The relation between fouling accumulation and reduced membrane performance is complex. The linear flow velocity in membrane modules influences the substrate load, the substrate transport, biofilm growth, biofilm morphology and the effect of accumulated biofilm on the feed channel pressure drop increase (Vrouwenvelder et al., 2009a). Biofilm accumulation can cause flow channeling, reducing the water production flux (Vrouwenvelder et al., 2009a). The effect of hydrodynamics on concentration polarization and biofilms on concentration polarization and on the enhanced osmotic pressure are also important (Kim et al., 2006; Herzberg and Elimelech, 2007; Chong et al., 2008). Furthermore, the concentration polarization may affect the biofilm cells physiology and viability by inducing higher substrate levels near the RO membrane. All these aspects are considered to adopt a new successful approach for biofouling control.
7.
Approaches for biofouling control
7.1.
Hydraulic optimized pressure vessel
In membrane installations, cross flow velocities in spiral wound membrane elements vary greatly (from w0.07e0.20 m s1, Fig. 9A) depending on element position in the pressure vessel. In most current installations, the cross flow velocity in lead membrane elements ranges between 0.14 and 0.20 m s1 (Table 1). However, several full-scale installations are operated with lower linear flow velocities in the lead elements (0.07e0.09 m s1) without problems (Van Paassen et al., 2005). The low cross flow velocity in the lead elements in these installations is caused by a hydraulic optimized pressure vessel use (Van der Meer et al., 2003; Fig. 9B and C). Using the optimized pressure vessel, the number of lead elements is doubled and consequently the linear flow velocity
7.2.
Feed flow reversal
Feed flow reversal in pressure vessels with several modules results in low FCP-levels. The effect of biofouling accumulation is strongly reduced by changing the flow direction in the pressure vessels (Fig. 9D). Before flow reversal, the lead membrane module has a much higher linear flow velocity (and pressure drop) than the membrane element in the last position of the pressure vessel. By changing the flow direction, the FCPincrease over the total installation will be instantly reduced. Simulation studies with the membrane fouling simulator also show this effect (Vrouwenvelder et al., 2009a). Feed flow reversal may also be efficient for simultaneous control of other fouling types besides biofouling, i.e. particulate fouling and mineral scaling. The principles to control individual fouling types differ. For prevention of mineral scaling or periods of system downtime, feed flow reversal can be applied resulting in re-dissolving of deposited scale into solution avoiding chemical dosage (Uchymiak et al., 2008). Practical consequences of feed flow reversal would be the use of adapted module sealing and some additional piping. In this periodic flow reversal system, the impact of biomass on performance is reduced and the biomass amount will gradually decline in time.
8.
Scenario for biofouling control
Advanced cleaning strategies involve preventive cleaning of partially fouled modules. Biofouling primarily takes place in the lead membrane module of the first stage. Therefore with a timely cleaning of the lead element(s), cleaning of the total installation can be avoided. Chemical use and energy consumption (to heat and pump/recirculate the cleaning solution) can be restricted, lowering the costs. Additionally, under standard conditions after biomass is removed from the lead membrane module the detached biomass passes several less fouled membrane modules in the same pressure vessel, which may negatively affects the cleaning performance. Isolating the lead membrane modules (during cleaning) from the rest of the membrane filtration installation prevents the possibility that detached biomass deposits in the rest of the pressure vessel. Early cleaning and isolating the lead membrane modules from the installation may have several advantages (Fig. 9E): (i) reduction of cleaning chemical use and energy consumption and (ii) more accurate monitoring of biofouling. When isolated lead modules are installed vertically (Fig. 9F), other types of cleanings such as enhanced reverse
3896
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 9 0 e3 8 9 8
Fig. 9 e Scheme illustrating selected approaches for biofouling control. Pressure vessel configuration on feed side containing membrane modules as applied in practice (A). Approaches for early biofouling detection and biofouling control (BeH). (H) is the most promising scenario for biofouling control.
flushing with water containing air eventually in combination with a copper salt solution (Cornelissen et al., 2007) can be applied as well. Application of cleaning by air sparging enhanced the flux in both ultrafiltration (Cabassud et al., 1997) and flat sheet NF systems (Ducom et al., 2002). Combining chemical cleaning with reverse flushing may be more effective than chemical cleaning only. Lead membrane modules can be adapted to improve (i) resistance to high shear forces caused by enhanced flushing and (ii) cleanability by the use of (thicker/coated/adapted geometry) feed spacers.
8.1.
Most promising scenario for biofouling control
The most promising scenario to control biofouling combines a lower linear flow velocity in lead modules and adapted designs for feed spacer with an advanced cleaning strategy (Fig. 9H). A lower linear flow velocity in spiral wound membrane modules will result in a (i) lower energy use, (ii) lower impact of biomass on the feed channel pressure drop, (iii) a more fluffy biofilm that may be easier to remove from the membrane module. Steering the biofilm morphology may reduce the impact of accumulated biomass on plant performance and facilitate biomass removal from the modules using e.g. adapted spacers and a reversed enhanced flow.
The scenario could be effective for biofouling control (Fig. 9H) at low energy requirements, minimal cost and minimal chemical use.
9.
Evaluation
Evaluation of on-going studies on biofouling of spiral wound nanofiltration and reverse osmosis membrane systems used in water treatment resulted in: A scenario for biofouling control based on combining lower linear flow velocities, adapted feed spacers and enhanced cleaning methods. The approach has low energy requirements, minimal costs and minimal chemical use.
Acknowledgments This work was performed at Wetsus, centre of excellence for sustainable water technology. Wetsus is funded by the ministry of economic affairs. The authors like to thank
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 9 0 e3 8 9 8
the participants of the theme ‘Biofouling’ for the fruitful discussions and their financial support. The input of Jacques van Paassen, Simon Bakker, Arie Zwijnenburg, Wim Borgonje, Harm van der Kooi is fully acknowledged.
references
Baker, J.S., Dudley, L.Y., 1998. Biofouling in membrane systems a review. Desalination 118, 81e90. Bereschenko, L.A., Prummel, H., Euverink, G.J.W., Stams, A.J.M., Van Loosdrecht, M.C.M., 2011. Effect of conventional chemical treatment on the microbial population in a biofouling layer of reverse osmosis systems. Water 45, 405e416. Cabassud, C., Laborie, S., Laine´, J.M., 1997. How slug flow can improve ultrafiltration flux in organic hollow fibres. Journal of Membrane Science 128, 93e101. Carnahan, R.P., Bolin, L., Suratt, W., 1995. Biofouling of PVD-1 reverse osmosis elements in the water treatment plant of the city of Dunedin, Florida. Desalination 102, 235e244. Characklis, W.G., Marshall, K.C., 1990. Biofilms. John Wiley & Sons, New York. Chen, M.J., Zhang, Z., Bott, T.R., 2005. Effects of operating conditions on the adhesive strength of Pseudomonas fluorescens biofilms in tubes. Colloids and Surfaces B: Biointerfaces 43, 61e71. Chong, T.H., Wong, F.S., Fane, A.G., 2008. The effect of imposed flux on biofouling in reverse osmosis: role of concentration polarisation and biofilm enhanced osmotic pressure phenomena. Journal of Membrane Science 325, 840e850. Cornelissen, E.R., Vrouwenvelder, J.S., Heijman, S.G.J., Viallefont, X.D., Van der Kooij, D., Wessels, L.P., 2007. Periodic air/water cleaning for control of biofouling in spiral wound membrane elements. Journal of Membrane Science 287, 94e101. Ducom, G., Matamoros, H., Cabassud, C., 2002. Air sparging for flux enhancement in nanofiltration membranes: application to O/W stabilised and non-stabilised emulsions. Journal of Membrane Science 204, 221e236. Flemming, H.C., Schaule, G., McDonogh, R., Ridgway, H.F., 1994. Effects and extent of biofilm accumulation in membrane systems. In: Geesey, G.G., Lewandowski, Z., Flemming, H.C. (Eds.), Biofouling and Biocorrosion in Industrial Water Systems. CRC Press, USA, pp. 63e89. Gamal Khedr, M., 2000. Membrane fouling problems in reverse osmosis desalination applications. Desalination and Water Reuse 10, 8e17. Garny, K., Neu, T.R., Horn, H., 2009. Sloughing and limited substrate conditions trigger filamentous growth in heterotrophic biofilms-measurements in flow-through tube reactor. Chemical Engineering Science 64, 2723e2732. Herzberg, M., Elimelech, M., 2007. Biofouling of reverse osmosis membranes: role of biofilm-enhanced osmotic pressure. Journal of Membrane Science 295, 11e20. Hille, A., He, M., Ochmann, C., Neu, T.R., Horn, H., 2009. Application of two component biodegradable carriers in a particle-fixed biofilm airlift suspension reactor: development and structure of biofilms. Bioprocess and Biosystems Engineering 32, 31e39. Kappelhof, J.W.N.M., Vrouwenvelder, H.S., Schaap, M., Kruithof, J. C., Van der Kooij, D., Schippers, J.C., 2003. An in situ biofouling monitor for membrane systems. Water Science and Technology: Water Supply 3, 205e210. Kim, A.S., Chen, H., Yuan, R., 2006. EPS biofouling in membrane filtration: an analytic modeling study. Journal of Colloid and Interface Science 303, 243e249.
3897
Kwok, W.K., Picioreanu, C., Ong, S.L., Van Loosdrecht, M.C.M., Ng, W.J., Heijnen, J.J., 1998. Influence of biomass production and detachment forces on biofilm structures in a biofilm airlift suspension reactor. Biotechnology and Bioengineering 58, 400e407. Mallevialle, J., Odendaal, P.E., Wiesner, M.R., 1996. Water Treatment Membrane Processes. McGraw-Hill, New York., ISBN 0-07-001559-7 Pereira, M.O., Kuehn, M., Wuertz, S., Neu, T., Melo, L.F., 2002. Effect of flow regime on the architecture of a Pseudomonas fluorescens biofilm. Biotechnology and Bioengineering 78, 164e171. Peyton, B.M., 1996. Effects of shear stress and substrate loading rate on Pseudomonas aeruginosa biofilm thickness and density. Water Research 30, 29e36. Purevdorj-Cage, L.B., Stoodley, P., 2004. Biofilm structure, behavior, and hydrodynamics. In: Ghannoum, M., O’Toole, G. A. (Eds.), Microbial Biofilms. ASM press, Washington DC, US, pp. 160e173. Ridgway, H.F., Flemming, H.F., 1996. Membrane biofouling. In: Mallevialle, J., Odendaal, P.E., Wiesner, M.R. (Eds.), Water Treatment Membrane Processes. McGraw-Hill, New York, pp. 6.1e6.62. Ridgway, H.F., Kelly, A., Justice, C., Olson, B.H., 1983. Microbial fouling of reverse-osmosis membranes used in advanced wastewater treatment technology: chemical, bacteriological, and ultrastructural analyses. Applied and Environmental Microbiology 45, 1066e1084. Schneider, R.P., Ferreira, L.M., Binder, P., Bejarano, E.M., Go´es, K.P., Slongo, E., Machado, C.R., Rosa, G.M.Z., 2005. Dynamics of organic carbon and of bacterial populations in a conventional pretreatment train of a reverse osmosis unit experiencing severe biofouling. Journal of Membrane Science 266, 18e29. Schock, G., Miquel, A., 1987. Mass transfer and pressure loss in spiral wound modules. Desalination 64, 339e352. Graf von der Schulenburg, D.A., Vrouwenvelder, J.S., Creber, S.A., Van Loosdrecht, M.C.M., Johns, M.L., 2008. Nuclear magnetic resonance microscopy studies of membrane biofouling. Journal of Membrane Science 323, 37e44. Shannon, M.A., Bohn, P.W., Elimelech, M., Georgiadis, J.G., Marina˜s, B.J., Mayes, A.M., 2008. Science and technology for water purification in the coming decades. Nature 452, 301e310. Stoodley, P., Dodds, I., Boyle, J.D., Lappin-Scott, H.M., 1999a. Influence of hydrodynamics and nutrients on biofilm structure. Journal of Applied Microbiology Symposium Supplement 85, 19Se28S. Stoodley, P., Lewandowski, Z., Boyle, J.D., Lappin-Scott, H.M., 1999b. Structural deformation of bacterial biofilms caused by short-term fluctuations in fluid-shear: an in situ investigation of biofilm rheology. Biotechnology and Bioengineering 65, 83e92. Tasaka, K., Katsura, T., Iwahori, H., Kamiyama, Y., 1994. Analysis of RO elements operated at more than 80 plants in Japan. Desalination 96, 259e272. Uchymiak, M., Alex, B., Christofides, P., Daltrophe, N., Weissman, M., Gilron, J., Rallo, R., Cohen, Y. 2008. RO membrane desalting in a feed flow reversal mode. In: Proceedings 8th International Congress on Membranes and Membrane Processes (ICOM 2008) Held in Honolulu, Hawaii, 12e18 (July). Van Loosdrecht, M.C.M., Eikelboom, D., Gjaltema, A., Mulder, A., Tijhuis, L., Heijnen, J.J., 1995. Biofilm structures. Water Science and Technology 32, 35e43. Van Paassen, J.A.M., Van der Meer, W.G.J., Post, J., 2005. Optiflux: from innovation to realisation. Desalination 178, 325e331. Van der Meer, W.G.J., Van Paassen, J.A.M., Riemersma, M.C., Van Ekkendonk, F.H.J., 2003. Optiflux: from innovation to realization. Desalination 157, 159e165.
3898
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 9 0 e3 8 9 8
Vrouwenvelder, J.S., Van Paassen, J.A.M., Wessels, L.P., Van Dam, A.F., Bakker, S.M., 2006. The membrane fouling simulator: a practical tool for fouling prediction and control. Journal of Membrane Science 281, 316e324. Vrouwenvelder, J.S., Manolarakis, S.A., Van der Hoek, J.P., Van Paassen, J.A.M., Van der Meer, W.G.J., Van Agtmaal, J.M.C., Prummel, H.D.M., Kruithof, J.C., Van Loosdrecht, M.C.M., 2008. Quantitative biofouling diagnosis in full scale nanofiltration and reverse osmosis installations. Water Research 42, 4856e4868. Vrouwenvelder, J.S., Hinrichs, C., Van der Meer, W.G.J., Van Loosdrecht, M.C.M., Kruithof, J.C., 2009a. Pressure drop increase by biofilm accumulation in spiral wound RO and NF membrane systems: role of substrate concentration, flow velocity, substrate load and flow direction. Biofouling 25, 543e555.
Vrouwenvelder, J.S., van Paassen, J.A.M., van Agtmaal, J.M.C., van Loosdrecht, M.C.M., Kruithof, J.C., 2009b. A critical flux to avoid biofouling of spiral wound nanofiltration and reverse osmosis membranes: fact or fiction? Journal of Membrane Science 326, 36e44. Vrouwenvelder, J.S., Graf von der Schulenburg, D.A., Kruithof, J.C., Johns, M.L., van Loosdrecht, M.C.M., 2009c. Biofouling of spiral wound nanofiltration and reverse osmosis membranes: a feed spacer problem. Water Research 43, 583e594. Vrouwenvelder, J.S., Buiter, J., Riviere, M., Van der Meer, W.G.J., Van Loosdrecht, M.C.M., Kruithof, J.C., 2010. Impact of flow regime on pressure drop increase and biomass accumulation and morphology in membrane systems. Water Research 44, 689e702. Wijeyekoon, S., Mino, T., Satoh, H., Matsuo, T., 2004. Effects of substrate loading rate on biofilm structure. Water Research 38, 2479e2488.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 9 9 e3 9 0 7
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Effects of bioaugmentation on indigenous PCB dechlorinating activity in sediment microcosms Sonja K. Fagervold a,1, Joy E.M. Watts b, Harold D. May c, Kevin R. Sowers a,* a
Center of Marine Biotechnology, University of Maryland Biotechnology Institute, 701 E. Pratt St., Baltimore, MD 21202, USA Department of Biological Sciences, Portsmouth University, Portsmouth, UK c Department of Microbiology and Immunology, Marine Biomedicine & Environmental Science Center, Medical University of South Carolina, Charleston, SC, USA b
article info
abstract
Article history:
Bioaugmentation is an attractive mechanism for reducing recalcitrant pollutants in sedi-
Received 19 January 2011
ments, especially if this technology could be applied in situ. To examine the potential
Received in revised form
effectiveness of a bioaugmentation strategy for PCB contamination, PCB dehalorespiring
27 April 2011
populations were inoculated into Baltimore Harbor sediment microcosms. A culture con-
Accepted 27 April 2011
taining the two most predominant indigenous PCB dehalorespiring microorganisms and
Available online 4 May 2011
a culture containing a strain with a rare ortho dechlorination activity and a non-indigenous strain that attacks double-flanked chlorines, were inoculated into sediment microcosms
Keywords:
amended with 2,20 ,3,5,50 ,6-hexachlorobiphenyl (PCB 151) and Aroclor 1260. Although we
Aroclor
observed a similar reduction in the concentration of PCB 151 in all microcosms at day 300,
Polychlorinated biphenyls
a reduced lag time for dechlorination activity was observed only in the bioaugmented
Anaerobic dehalorespiration
microcosms and the pattern of dechlorination was altered depending on the initial
Bioaugmentation
combination of microorganisms added. Dechlorination of Aroclor 1260 was most extensive when dehalorespiring microorganisms were added to sediment. Overall numbers of dehalorespiring microorganisms in both bioaugmented and non-bioaugmented microcosms increased 100- and 1000-fold with PCB 151 and Aroclor 1260, respectively, and they were sustained for the full 300 days of the experiments. The ability of bioaugmentation to redirect dechlorination reactions in the sediment microcosms indicates that the inoculated PCB dehalorespiring microorganisms effectively competed with the indigenous microbial populations and cooperatively enhanced or altered the specific pathways of PCB dechlorination. These observations indicate that bioaugmentation with PCB dehalorespiring microorganisms is a potentially tractable approach for in situ treatment of PCB impacted sites. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
PCBs are persistent organic pollutants that are ubiquitously dispersed throughout the ecosystem as a result of cycling between air, water, and soil. Once released into the environment, PCBs can bioaccumulate throughout the food chain as
a result of absorption in the fatty tissue of animals, such as fish and marine mammals; within humans, PCBs have been detected in human adipose tissue, milk and serum (Johansen et al., 1993; Muir et al., 1992; Neff, 1984; Safe, 1993; Schecter et al., 1994). Among their toxic effects, PCBs have been reported to act as endocrine disrupters (Crisp et al., 1998) and
* Corresponding author. Tel./fax: þ1 410 234 8878. E-mail address:
[email protected] (K.R. Sowers). 1 Present address: UPMC Univ Paris 06, FRE 3350, LECOB, Observatoire Oce´anologique, F-66650, Banyuls/Mer, France. 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.04.048
3900
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 9 9 e3 9 0 7
possible carcinogens (Safe, 1993). The commercial PCB mixture Aroclor 1260, which was widely used as dielectric fluid in liquid-filled transformers and capacitors between 1930 and 1971, is especially recalcitrant to degradation, primarily due to its high chlorine content and hydrophobicity (Alder et al., 1993; Quensen et al., 1990; Versar, 1976). However, microbial reductive dechlorination, a process in which microorganisms use PCBs as terminal electron acceptors for respiration, can transform highly chlorinated congeners in commercial mixtures such as Aroclor 1260 (Brown et al., 1987). In addition to the potential for lower cost, in situ bioaugmentation with PCB dechlorinating microorganisms would provide several advantages over traditional methods such as dredging and capping, including minimal disruption to benthic habitats in sensitive rivers and wetlands and the ability to treat shallow locations or those with restricted accessibility. Although anaerobic bioaugmentation studies have been successful for in situ treatment of chlorinated ethenes (Lendvay et al., 2003; Major et al., 2002), another class of organochlorines most commonly found in contaminated groundwater, effective bioaugmentation of PCB contaminated sites has not, as yet, been demonstrated. There have been reports on biostimulation of PCB dechlorination using other chlorinated compounds (halopriming) to enhance the numbers of dechlorinators and increase rates of contaminant breakdown (Bedard et al., 1996; Deweerd and Bedard, 1999; Klasson et al., 1996; Van Dort et al., 1997) though, unfortunately, the biostimulants are often themselves persistent organic pollutants. Other studies investigated the effects of enrichment cultures on bioaugmentation (Bedard et al., 1997; Natarajan et al., 1997), but the fates of the dechlorinating microorganisms were not determined in these earlier studies because their identity was unknown. The patterns of PCB dechlorination in anaerobic sediments varies between sites and has been attributed to the activities of different indigenous dehalorespiring populations (for reviews see Bedard and Quensen (1995) and Smidt and de Vos (2004)). Since the discovery that some “Dehalococcoides” spp. and related species within the Chloroflexi are capable of dehalorespiring PCBs, there have been reports on the enrichment and isolation of strains with specific dechlorinating capabilities (Bedard et al., 2007; Cutter et al., 2001; Fagervold et al., 2005; Wu et al., 2002). Fagervold et al. (2005) showed that PCB 132 was dechlorinated in two discreet steps by 2 phylotypes enriched from Baltimore Harbor (BH) sediments, demonstrating that PCBs could be cooperatively dechlorinated by a consortium of dehalorespiring bacteria. Furthermore, Aroclor 1260 was later shown to be dechlorinated by a consortium consisting of three phylotypes that were reproducibly enriched from BH sediments (Fagervold et al., 2007). The dechlorination pattern resulting from the activity of the microbial consortia predominantly attacks flanked meta chlorines most characteristic of process N, which has been reported in sediments from Woods Pond and Silver Lake (Bedard and Quensen, 1995). In contrast, Adrian et al. (2009) reported the reductive dechlorination of Aroclor 1260 by “Dehalococcoides” sp. CBDB1 as more predominant attack of para chlorines most characteristic of Process H, reported in Hudson River sediments (Bedard and Quensen, 1995). These combined observations support the hypothesis that the
variation of microbial dechlorination patterns observed in different sites is due to the presence and activity of specific species and consortia of dehalorespiring bacteria. However, it is unclear if the combined interactions of the indigenous population of dehalorespiring bacteria and resulting products can be altered by changing the total number and ratio of individual species in the consortium. In this study sediment microcosms, spiked with PCB 151 (2,20 ,3,5,50 ,6-hexachlorobiphenyl) or Aroclor 1260, were bioaugmented with the two most predominant indigenous PCB dehalorespiring microorganisms from BH sediments, and a culture containing a species with a rare ortho-dechlorinating activity and a non-indigenous species selective for dechlorination of double-flanked chlorines. The study examines the effects of bioaugmentation on both the dehalogenating activities and fate of the indigenous dehalorespiring population in BH sediment microcosms.
2.
Materials and methods
2.1.
Bacterial cultures used for bioaugmentation
The specific activities of the microorganisms used to bioaugment the sediments have been characterized previously (Fagervold et al., 2007; May et al., 2006; Wu et al., 2000). Bacterium o-17 reductively dechlorinates flanked meta chlorines and flanked ortho chlorines in congeners containing up to 3 ortho chlorines (Fagervold et al., 2007; May et al., 2006). “Dehalobium chlorocoercia” strain DF-1 dechlorinates double-flanked chlorines in the para or meta positions, but does not dechlorinate single flanked chlorines (Wu et al., 2000). Phylotype DEH10 dechlorinates the double-flanked meta chlorines in 2,3,4substituted chlorobiphenyl rings and para-flanked meta chlorines when no double-flanked chlorines are available. Phylotype SF1 dechlorinates all 2,3,4,5-substituted chlorobiphenyl rings, preferentially in the meta position, although some para dechlorination has been observed (Fagervold et al., 2007). The inocula used for bioaugmentation were maintained by sequential transfer of 10% v/v into estuarine salts medium (E-Cl) as described below. D. chlorocoercia strain DF-1, originally enriched from Charleston Harbor sediments, was maintained in co-culture with a Desulfovibrio spp. with 10 mM sodium formate and 2,3,4,5-CB (PCB 61) (Wu et al., 2000). A co-culture containing Bacterium o-17 originally enriched from BH sediment was maintained with a Desulfovibrio spp. with 20 mM sodium acetate and 2,3,5,6-CB (PCB 65) (Cutter et al., 2001). Phylotypes DEH10 and SF1 were initially enriched from BH sediment (Fagervold et al., 2007) and maintained with a fatty acid mixture (acetate, propionate and butyrate, 2.5 mM each) and 2,20 3,50 ,6-CB (PCB 95) or 2,20 ,3,40 ,6-CB (PCB 91), respectively, and 0.1% (w/v), sterile, dry BH sediment. Both phylotypes were maintained as highly enriched cultures.
2.2.
Microcosm experiments
Defined E-Cl medium containing a low sulfate concentration (<0.3 mM) was dispensed anaerobically (10 ml) into 20 ml anaerobe tubes and sealed under an atmosphere of N2eCO2 (4:1) (Berkaw et al., 1996) with modifications described
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 9 9 e3 9 0 7
previously (Fagervold et al., 2005). A fatty acid mixture (acetate, propionate and butyrate, 2.5 mM each) was added as an electron donor and carbon source (Berkaw et al., 1996). Aroclor 1260 or PCB 151 was added in 10 ml acetone to final concentrations of 100 and 50 ppm, respectively. All treatments contained 1.5 g v/v freshly collected BH sediment that was contaminated with less than 2 ppm weathered PCB. Treatment 1 contained 1.5 g (wet) BH sediment as a source of indigenous microorganisms at a pre-enriched background level of approximately 1.5 105 total dehalorespiring cells per 10 ml culture. Treatment 2 contained 1.5 g fresh BH sediment with the addition of 0.4 ml each of SF1 and DEH10 cultures, which is equivalent to approximately 3 105 cells of each phylotype per 10 ml. Treatment 3 contained 1.5 g fresh BH sediment with the addition of 1 ml of a co-culture containing o-17 and 1 ml of a co-culture containing DF-1, which is equivalent to 1 105 and 2 105 cells, respectively, of these phylotypes per 10 ml. Treatment 4 contained 1.5 g BH sediment and all the microorganisms at the same concentrations as in treatments 2 and 3. Sterile controls were prepared as in treatment 1, followed by sterilization (autoclaved for 20 min, 121 C). All cultures were prepared in triplicate and incubated at 30 C in the dark.
2.3.
Analytical techniques
Microcosms were sampled on day 0 immediately after inoculation and subsequently every 50 days in an anaerobic glove box (Coy Laboratory Products, Grass Lake, MI). Each culture sample (0.5 ml) was extracted with 3 ml of hexane for 12 h on a wrist shaker. The organic phase was passed through a copper/Florisil (1:4) column and analyzed using a HewlettePackard 5890 series II gas chromatograph (GC) with a DB-1 capillary column (30 m by 0.25 mm by 0.25 mm; JW Scientific, Folsom, CA) and a 63Ni electron capture detector as described previously (Fagervold et al., 2005). Unpaired two sample Student’s t-tests, assuming equal variance, were used to determine the significance between the mean concentrations of 151 at different time points, in augmented versus nonaugmented microcosms, as well as the total, meta, para and ortho chlorines at day 300 in the Aroclor 1260 microcosms.
2.4.
3901
a monophyletic group within the Chloroflexi, Chl348FGC and Dehal884R (Fagervold et al., 2005). DGGE was performed using the D-Code Universal Mutation Detection System (Bio-Rad, Hercules, CA.) (Watts et al., 2001). To identify strains from DGGE profiles, bands were excised and incubated in 30 ml TE overnight at 4 C, amplified by PCR and checked for purity using DGGE and subsequently sequenced.
2.5.
DNA sequencing and analysis
PCR products from excised bands were used as templates for dye terminator cycle sequencing using the Big Dye 3.1 kit (Applied Biosystems, Foster city, CA) and an AB3100 Genetic Analyzer (Applied Biosystems). Sequences were examined for errors and assembled using the software Pregap4 and Gap4 of the Staden software package (http://sourceforge.net/projects/ staden). Sequences similarities were analyzed using the Basic Local Alignment Search Tool (BLAST) (Altschul et al., 1990).
2.6. Quantitative assessment of PCB dechlorinating populations Cell enumeration of DEH10, SF1, o-17 and DF-1 inocula for bioaugmentation experiments were determined by most probable number estimation of 16S rRNA gene copies from total DNA. Diluted DNA from these cultures was subjected to PCR (40 cycles) using primers Chl348F and Dehal884R (Fagervold et al., 2005), and one copy of the gene per genome was assumed based on the genomes of “Dehalococcoides ethenogenes” strain 195 and Dehalococcoides sp. strain CBDB1 (Kube et al., 2005; Seshadri et al., 2005). Numbers of putative dehalogenating Chloroflexi in bioaugmentation microcosms were monitored by competitive PCR, using primers Chl348F and Dehal884R (Fagervold et al., 2005; Kjellerup et al., 2008). The assays were performed on normalized DNA extracted from microcosm sub samples as described previously (Fagervold et al., 2007).
3.
Results
3.1. Effect of bioaugmentation on the patterns of PCB 151 dechlorination
Bacterial community 16S rRNA gene analyses
DNA from pooled samples (0.5 ml from each of replicates for each treatment) was extracted every 50 days. Microcosms sampled on day 0 immediately after inoculation were extracted with the DNA SPIN For Soil kit (MP Biochemicals, Solon, OH), followed by purification with the Promega Wizard PCR Prep Kit (Promega, Madison, WI.). All subsequent DNA extractions were extracted with UltraClean Soil DNA Kit (Mo Bio, Carlsbad, CA), which did not require an additional purification step. DNA concentration was determined using a DU 650 spectrophotometer (Beckman, Fullerton, CA), and DNA extracts were diluted with TE buffer to 10 mg/ml. Diluted DNA (1 ml) was used as a template in all subsequent PCR reactions. The microbial dechlorinating population within microcosms was evaluated by denaturing gradient gel electrophoresis (DGGE) of total microbial community DNA using PCR amplification with primers specific for 16S rRNA genes of
BH sediment microcosms amended with PCB 151 were bioaugmented with selected populations of dehalorespiring microorganisms to a final concentration of approximately 1e7 104 cells ml1 and the patterns of dechlorination were monitored over 300 days. PCB 151 (2,20 ,3,5,50 ,6-CB) can be dechlorinated initially through two pathways (Fig. 1). The first pathway involves a meta dechlorination to PCB 95 (2,20 ,3,50 ,6-CB), which is further dechlorinated in the meta position to PCB 53 (2,20 ,5,60 -CB). The second pathway is a dechlorination in the ortho position to PCB 92 (2,20 ,3,5,50 -CB), which can be dechlorinated either in the meta position to PCB 52 (2,20 ,5,50 -CB) or in the ortho position to PCB 72 (2,30 ,5,50 -CB). The dechlorination pathways of PCB 151 varied as a result of the different bioaugmentation treatments (Fig. 2). The PCB dehalorespiring population was monitored by DGGE analysis of the 16S rRNA gene community at time points throughout the incubation period to: 1) investigate whether the inoculated
3902
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 9 9 e3 9 0 7
Fig. 1 e Possible pathways for dechlorination of PCB 151.
microorganisms were sustained after addition, and 2) to determine the impact of bioaugmentation on the indigenous dechlorinating community in the microcosms (Fig. 3). All treatments were found to be significantly different from the non-bioaugmented control (1: BH, Fig. 2) as determined by unpaired Student’s t-tests ( p value <0.05), except for treatment 2 (BH þ SF1 þ DEH10) at day 200 and 300. Treatment 1, BH sediment with indigenous microorganisms only, did not dechlorinate before day 150, thereafter, both meta and ortho dechlorination was observed (Fig. 2). After 300 days of incubation, the major dechlorination product of PCB 151 for two replicates was PCB 72 (69.1 and 67.3 mol %); resulting from two sequential ortho dechlorinations, likely mediated by o-17 (Fig. 3). The major product for the remaining replicate was PCB 95 (81.2 mol %) indicating a single meta dechlorination, likely mediated by SF1. In a previous report, DEH10 was shown to catalyze the single meta dechlorination of PCB 151 to PCB 95 in BH microcosms (Fagervold et al., 2007). However, both DEH10 and SF1 have specificity for dechlorinating flanked meta chlorines. These differences observed between the two studies and between the replicate cultures in the current study are likely due to biotic factors, such as competition between the indigenous species enriched in each microcosm. An indigenous Dehalococcoides phylotype was initially observed in BH sediments for all treatments at day 0 (Fig. 3), but was not subsequently detected after incubation with PCB 151. This phylotype co-migrates with DEH10 and was most similar using 16S rRNA gene similarity (472 identity over 478 bp) to another Dehalococcoides species, DHC ANAS (acc. number DQ855129) detected previously in trichloroethene enrichment cultures (Holmes et al., 2006). Treatment 2, consisting of BH sediment bioaugmented with the indigenous phylotypes SF1 and DEH10, exhibited two sequential meta dechlorinations of PCB 151 to PCB 53 (46 0.1 mol % at day 300) in all three replicates in only 50 days, with less than 5 mol % accumulation of PCB 95 (Fig. 2). Both SF1 and DEH10 were detected at day 0 in these microcosms, but only DEH10 was subsequently detected throughout the
experiment, to day 300 (Fig. 3). A second less intense band was detected in treatment 2, with 97% (471/478 bp) sequence identity to an uncultured Chloroflexi (clone VHS-B3-87). This phylotype was also detected weakly in treatments 3 and 4 during active dechlorination of PCB 151, but its role is unknown. In contrast, dechlorination was observed at 100 days in treatment 3 (BH sediment with o-17 and DF-1) and the major product after 300 days was PCB 72 at 44 7 mol %. This pathway is probably the result of two sequential ortho dechlorination steps, with PCB 92 as an intermediate. Some meta dechlorination (18 5 mol %) of the intermediate PCB 92 to PCB 52 was observed on day 300. DF-1 does not dechlorinate PCB 151 because this congener does not contain double-flanked chlorines. The greatest range of dechlorination activities was observed in treatment 4, BH sediment with all four phylotypes. Meta dechlorination of PCB 151 to PCB 53 (26 3 mol %) was detected after 300 days with little accumulation of the intermediate congener PCB 95 (<2 mol %). Ortho dechlorination of PCB 151 to PCB 72 (25 3 mol %) was detected with <5 mol % accumulation of PCB 92. However, some meta dechlorination to PCB 52 (8 2 mol %) was also detected at day 300. The activity corresponds to a more diverse microbial community of putative dechlorinators with four different phylotypes detected at day 300. DEH10 and o-17 appeared as the predominant DGGE bands (Fig. 3) and the uncultured Chloroflexi and SF1 as minor bands, which although not quantitative, is consistent with the dechlorination pathways observed.
3.2. Effect of bioaugmentation on the patterns of Aroclor 1260 dechlorination BH sediment microcosms amended with Aroclor 1260 exhibited a decrease in the total chlorines per biphenyl for all four treatments (Fig. 4), in comparison to sterile controls. All treatments were found to be significantly different ( p value <0.05) from the non-biaugmented control and treatment 4, BH sediment bioaugmented with all four dechlorinators, exhibited the most extensive dechlorination, with the reduction of
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 9 9 e3 9 0 7
1.77 chlorines per biphenyl over 300 days; while treatment 2, BH sediment bioaugmented with SF1 and DEH10, was similar with an average reduction of 1.70 chlorines per biphenyl. The remaining treatments exhibited a total reductive dechlorination of approximately 1.40 chlorines per biphenyl in 300 days.
3903
The total reduction of meta chlorines was similar to the reduction of total chlorines (Fig. 4, panels A and B), however the unpaired t-test showed that treatment 3, BH sediment bioaugmented with o-17 and DF-1 was not statistically different from the non-bioaugmented microcosm. Also some differences in the reduction of ortho chorines were observed between the different treatments. Treatment 3, BH sediment bioaugmented with o-17 and DF-1 dechlorinated PCB congeners in the ortho position to a greater extent than other treatments, with an average reduction of 0.24 and 0.21 ortho chlorines per biphenyl over 300 days, respectively. In contrast, treatment 4, BH sediment bioaugmented with all four microorganisms, dechlorinated in the para position to a greater extent than other treatments, however, this might be a reflection of overall higher dechlorination by this treatment. All treatments were significantly different ( p value <0.05) from non-bioaugmented microcosms, regarding ortho and para chlorines, at day 300.
3.3. Effect of bioaugmentation on the indigenous population of PCB dehalorespiring bacteria
Fig. 2 e Dechlorination of PCB 151 in BH sediment after the following treatments: BH sediment with no bioaugmentation (1); BH sediment bioaugmented with SF1 and DEH10 (2); o-17 and DF-1 (3); SF1, DEH10, o-17 and DF-1. Congeners detected include PCB 151 (C), PCB 95 (:), PCB 92 (O), PCB 72 (,), PCB 53 (-), PCB 52 (>) and a sterile control (B). Each datum point is the mean and standard deviation for three replicate cultures.
To determine whether microorganisms used for bioaugmentation were capable of growth and sustainable dechlorination in the presence of the indigenous dahalorespiring population, and the general microbial community within the sediment; dehalorespiring microorganisms were enumerated during growth in the number of 16S rRNA gene copies, per ml normalized DNA. In all four treatments the numbers of dechlorinators increased approximately 100- and 1000-fold with PCB 151 and Aroclor 1260, respectively, regardless of whether the culture was bioaugmented. The average numbers of 16S rRNA gene copies per ml normalized DNA in most microcosms with PCB 151 increased from 1.65 103 1.68 103 at day 0 to 2.40 105 9.48 104 at day 100, then remained relatively constant to day 300. Slower initial growth was observed only for treatment 1 with PCB 151 (indigenous population without bioaugmentation), but similar numbers were observed after 200 days. The average numbers of 16S rRNA gene copies per ml normalized DNA in microcosm dechlorinating Aroclor 1260 increased from 1.06 103 380 at day 0 to 1.32 106 6.94 105 at day 100 then remained relatively constant to day 300. The results indicate that total number of dehalorespiring microorganisms in the microcosms, both indigenous and/or inoculated microorganisms, increased to a finite total or climax community within the first 100 days of active dechlorination and this steady state was maintained for the remainder of the incubation experiments. The numbers of dechlorinating microorganisms in bioaugmented microcosms was approximately 2 104 cells per ml sediment or 1 105 cells per g dry wt of sediment at day 0. By comparison, Krumins et al. (2009) used approximately 3 106 cells per ml sediment or 7 106 cells per g dry wt of sediment of D. ethenogenes strain 195 per ml in microcosm experiment to test stimulation of PCB dechlorination activity. Although lower numbers of microbes were used for bioaugmentation in the current study, they still had a considerable effect on the composition of the microbial community and on the pathways of PCB dechlorination.
3904
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 9 9 e3 9 0 7
Fig. 3 e Composite DGGE of PCR amplified 16S rRNA from PCB 151 and Aroclor 1260 microcosms after different treatments. Day 0 samples for treatments 2, 3 and 4 was subjected to 40 PCR cycles rather than the normal 26 cycles (Fagervold et al., 2005) and run on a separate gel. aPhylotype detected in day 0 samples co-migrated with DEH10 but was a different phylotype with sequence identity (99%) to uncultured Chloroflexi clone DHC ANAS (DQ855129). bUnkn [ unknown, sequence identity (97%) to uncultured Chloroflexi clone VHS-B3-87 (DQ294968). cExpected position of DF-1.
Although indigenous DEH10, SF1 and o-17 were enriched from BH sediments in treatment 1, the effect of bioaugmentation is evident from differences in both the dechlorination pattern and changes in the microbial community in treatments 2 through 4 (Fig. 3). In treatment 1 with PCB 151 o-17 and SF1 are enriched, which is consistent with two sequential ortho dechlorinations of PCB 72 and a single meta dechlorination to PCB 95. In treatment 2, the predominance of DEH10 concurrent with a two step meta dechlorination to PCB 53 is consistent with results reported previously (Fagervold et al., 2007). In treatment 3, the predominance of o-17 is consistent with a two step ortho dechlorination to PCB 72 and the presence of SF1 is consistent with smaller amounts of single step meta dechlorinations to PCB 95 and 52. Finally, adding all four microorganisms resulted in predominance of DEH10 and o-17 and concurrent two step meta and ortho dechlorinations, respectively, as the dominant dechlorination pathways. SF1 was weakly detected, which is consistent with smaller amounts of single step meta dechlorination products observed. Likewise, bioaugmentation with different microorganisms had a substantial effect on the dehalorespiring population in cultures amended with Aroclor 1260. Treatment 1 without bioaugmentation is consistent with a prior study showing that DEH10 and SF1 are the predominant microorganisms observed in BH sediments enriched with Aroclor 1260 (Fagervold et al., 2007). Bioaugmentation with DEH10 and SF1 in treatment 2 resulted in detection of both phylotypes as expected, a reduction in the lag time and more extensive dechlorination of para and meta chlorines at 300 days. Treatment 3 resulted in sustained detection of o-17 throughout the 300-day incubation period and significant increase in ortho dechlorination compared with treatments 1 and 2. DF-1 was not detected in the microcosm and there was no major change in the meta and para dechlorination patterns, which indicates that DF-1 did not compete successfully with the indigenous dehalorespirers DEH10 and SF1 for utilizing Aroclor 1260. Bioaugmenting with all four dehalorespiring bacteria resulted in the same lag time and total dechlorination pattern as observed after
bioaugmentation with only DEH10 and SF1. However, a less intense o-17 band was detected throughout the 300 days, concurrent with some ortho dechlorination. Interestingly, there was an increase in the extent of para dechlorination than with other treatments. One possible explanation is the increased removal of ortho chlorines generated more congeners with double-flanked para chlorines, which are generally more susceptible to dechlorination than single flanked meta chlorines (Bedard and Quensen, 1995).
4.
Discussion
Bioaugmentation with PCB dehalorespiring bacteria had a considerable effect on the dehalogenating activity of PCB 151 and Aroclor 1260, in sediment microcosms. The addition of PCB dechlorinators changed the specific dechlorination activities and increased the extent of dechlorination after 300 days, compared with enrichments containing only indigenous dechlorinating populations. A one-way ANOVA analysis showed no significant difference between dechlorination rates between treatments ( p value >0.05), which indicates that the more extensive dechlorination observed in bioaugmented microcosms after 300 days was the result of the reduced lag time. This is consistent with the enumeration data indicating that total numbers of dehalorespiring microorganisms in both enriched and bioaugmented cultures increased to the same steady state maximum levels during the incubation period. This observation is not surprising as nutrients, including nitrogen, phosphate, sulfur, trace metals, vitamins, PCBs and fatty acids, were not limiting in these experiments, although it is not clear why bioaugmentation in some cases repressed growth of an indigenous dechlorinating species in the presence of excess PCB. Furthermore, we observed heterogeneity of activity between non-bioaugmented replicate samples both in this study and previously (Fagervold et al., 2007), but not in the bioaugemented samples. This may be an artificial microcosm effect, and a number of factors could be involved; for example,
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 9 9 e3 9 0 7
Fig. 4 e Dechlorination of Aroclor 1260 in microcosms with different treatments (see legend) calculated as total chlorines per biphenyl (A), and meta (B), ortho (C) and para (D) chlorines per biphenyl. BH [ Baltimore Harbor sediments. Each datum point is the mean and standard deviation for three replicate cultures.
3905
bioaugmentation of microcosms may reduce the spatial patchiness of dechlorinating microorganisms within the microcosm thereby reducing heterogeneity (Jessup et al., 2004). In addition, some indigenous species detected in enrichments with PCB 151 were no longer detectable after bioaugmentation (Fig. 3). The results indicate that bioaugmentation was successful, because bioaugmentation with greater initial cell numbers enabled the inoculum to successfully outcompete the indigenous dechlorinating population (e.g., PCB 151, treatment 2). However, this does not explain why selected activities of the indigenous population could be completely repressed in the presence of excess PCB. The free energy yield is similar for products of the alternative pathways (Holmes et al., 1993), so this selection is unlikely due to thermodynamic differences. Likewise, the similarity between rates for dechlorination of parent compound PCB 151 via alternate pathways rules out kinetic differences as a reason for selection. One possible explanation is that PCBs adsorb to sediment particles due to their hydrophobic nature and as the more abundant species colonize the particles as biofilms, the PCBs are no longer available to species that are less abundant during the initial inoculation. The observation that bioaugmented microorganisms were sustainable as long as activity continued, supports the latter hypothesis that PCBs were no longer bioavailable to other species despite the fact that high levels of parent PCB 151 (>10 ppm) were detected after 300 days for all treatments. Although these results show that bioaugmentation is a potential treatment strategy for PCB dechlorination in sediments, several questions must be addressed to determine if bioremediation of weathered PCBs is feasible. One unknown is the effectiveness of bioaugmentation for dechlorinating low levels of PCBs most commonly associated with weathered PCB contaminated sites. In the presence of weathered PCBs where the concentration is more limiting both kinetic and availability issues might affect the effectiveness of in situ PCB transformation. A recent study by May et al. (2008) showed that bioaugmentation with DF-1 stimulated the reductive dechlorination of Aroclor 1260 (>5 ppm) in contaminated soil, which suggests that using bioaugmentation for treatment of low levels of weathered PCBs is feasible. Furthermore, Krumins et al. (2009) found that the addition of D. ethenogenes strain 195 and pentachloronitrobenzene to microcosms stimulated the dechlorination of weathered PCB in contaminated sediments from Anacostia River, Washington DC. Interestingly, in microcosms bioaugmented with D. ethenogenes strain 195, this strain could not be detected after 281 days, although its addition did have an initial stimulatory effect on dechlorination. The results of the current study demonstrate clearly that the dehalorespiring microorganisms added to sediment microcosms can successfully compete with indigenous populations reducing the lag time and diverting the pathways of dehalogenation. The ability to alter the synergistic activities of the indigenous dehalogenating community suggests that bioaugmentation is a potentially viable approach for enhancing reductive dehalogenation of PCB impacted sediments. Future studies will focus on the effects of bioaugmentation on indigenous populations under PCB limiting conditions to identify the kinetic and availability limitations in weathered sediments subject to in situ treatment.
3906
5.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 9 9 e3 9 0 7
Conclusions
A reduced lag time for dechlorination activity was observed in all bioaugmented microcosms and the pattern of dechlorination was altered depending on the initial combination of microorganisms added. Dechlorination pattern heterogeneity observed in sediment enrichments was not observed in bioaugmented cultures, indicating that bioaugmentation effectively redirected the dechlorination pathways. The ability of bioaugmentation to redirect dechlorination reactions in the sediment microcosms and sustainability of bioaugmented microorganisms after 300 days indicates that the inoculated PCB dehalorespiring microorganisms effectively competed with the indigenous microbial populations and cooperatively enhanced or altered the specific pathways of PCB dechlorination. Although the mechanism by which bioaugmented microorganisms could outcompete the indigenous population in the presence of excess substrate is not known, the results of this study support the feasibility of using in situ bioaugmentation with dehalorespiring bacteria as an environmentally less evasive and lower cost alternative to dredging for treatment of PCB impacted sediments.
Acknowledgments This study was funded by the Office of Naval Research, U.S. Department of Defense (Project Numbers N000014-03-1-0035 to K.R.S. and N000014-03-1-0034 to H.D.M.) and US Department of Defense, Strategic Environmental Research and Development Program (Project Number ER-1502 to K.R.S.).
references
Adrian, L., Dudkova, V., Demnerova, K., Bedard, D.L., 2009. “Dehalococcoides” sp. strain CBDB1 extensively dechlorinates the commercial polychlorinated biphenyl mixture Aroclor 1260. Applied and Environmental Microbiology 75 (13), 4516e4524. Alder, A.C., Ha¨ggblom, M.M., Oppenheimer, S.R., Young, L.Y., 1993. Reductive dechlorination of polychlorinated biphenyls in anaerobic sediments. Environmental Science & Technology 27 (3), 530e538. Altschul, S.F., Gish, W., Miller, W., Myers, E.W., Lipman, D.J., 1990. Basic local alignment search tool. Journal of Molecular Biology 215 (3), 403e410. Bedard, D.L., Bunnell, S.C., Smullen, L.A., 1996. Stimulation of microbial para-dechlorination of polychlorinated biphenyls that have persisted in Housatonic River sediment for decades. Environmental Science & Technology 30 (2), 687e694. Bedard, D.L., Quensen, J.F., 1995. Microbial reductive dechlorination of polychlorinated biphenyls. In: Young, L.Y., Cerniglia, C.E. (Eds.), Microbial Transformation and Degradation of Toxic Organic Chemicals. A. John Wiley, New York, pp. 127e216. Bedard, D.L., Ritalahti, K.M., Loffler, F.E., 2007. The Dehalococcoides population in sediment-free mixed cultures metabolically dechlorinates the commercial polychlorinated biphenyl
mixture Aroclor 1260. Applied and Environmental Microbiology 73 (8), 2513e2521. Bedard, D.L., VanDort, H.M., May, R.J., Smullen, L.A., 1997. Enrichment of microorganisms that sequentially meta, para-dechlorinate the residue of Aroclor 1260 in Housatonic River sediment. Environmental Science & Technology 31 (11), 3308e3313. Berkaw, M., Sowers, K.R., May, H.D., 1996. Anaerobic ortho dechlorination of polychlorinated biphenyls by estuarine sediments from Baltimore Harbor. Applied and Environmental Microbiology 62 (7), 2534e2539. Brown, J.J.F., Wagner, R.E., Feng, H., Bedard, D.L., Brennan, M.J., Carnahan, J.C., May, R.J., 1987. Environmental dechlorination of PCBs. Environmental Toxicology and Chemistry 6 (8), 579e593. Crisp, T.M., Clegg, E.D., Cooper, R.L., Wood, W.P., Anderson, D.G., Baetcke, K.P., Hoffmann, J.L., Morrow, M.S., Rodier, D.J., Schaeffer, J.E., Touart, L.W., Zeeman, M.G., Patel, Y.M., 1998. Environmental endocrine disruption: an effects assessment and analysis. Environmental Health Perspectives 106 (Suppl.1), 11e56. Cutter, L.A., Watts, J.E.M., Sowers, K.R., May, H.D., 2001. Identification of a microorganism that links its growth to the reductive dechlorination of 2,3,5,6-chlorobiphenyl. Environmental Microbiology 3 (11), 699e709. Deweerd, K.A., Bedard, D.L., 1999. Use of halogenated benzoates and other halogenated aromatic compounds to stimulate the microbial dechlorination of PCBs. Environmental Science & Technology 33 (12), 2057e2063. Fagervold, S.K., May, H.D., Sowers, K.R., 2007. Microbial reductive dechlorination of Aroclor 1260 in Baltimore Harbor sediment microcosms is catalyzed by three phylotypes within the phylum Chloroflexi. Applied and Environmental Microbiology 73 (9), 3009e3018. Fagervold, S.K., Watts, J.E.M., May, H.D., Sowers, K.R., 2005. Sequential reductive dechlorination of meta-chlorinated polychlorinated biphenyl congeners in sediment microcosms by two different Chloroflexi phylotypes. Applied and Environmental Microbiology 71 (12), 8085e8090. Holmes, D.A., Harrison, B.K., Dolfing, J., 1993. Estimation of gibbs free energies of formation for polychlorinated biphenyls. Environmental Science & Technology 27 (4), 725e731. Holmes, V.F., He, J., Lee, P.K.H., Alvarez-Cohen, L., 2006. Discrimination of multiple Dehalococcoides strains in a trichloroethene enrichment by quantification of their reductive dehalogenase genes. Applied and Environmental Microbiology 72 (9), 5877e5883. Jessup, C.M., Kassen, R., Forde, S., Kerr, B., Buckling, A., Rainey, P.B., Bohannan, B.J.M., 2004. Big questions, small worlds: microbial model systems in ecology. TRENDS in Ecology and Evolution 19 (4), 189e197. Johansen, H.R., Becher, G., Polder, A., Skaare, J.U., 1993. Congenerspecific determination of polychlorinated biphenyls and organochlorine pesticides in human milk from Norwegian mothers living in Oslo. Journal of Toxicology and Environmental Health 42, 157e171. Kjellerup, B.V., Sun, X., Ghosh, U., May, H.D., Sowers, K.R., 2008. Site-specific microbial communities in three PCB-impacted sediments are associated with different in situ dechlorinating activities. Environmental Microbiology 10 (5), 1296e1309. Klasson, K.T., Barton, J.W., Evans, B.S., Reeves, M.E., 1996. Reductive microbial dechlorination of indigenous polychlorinated biphenyls in soil using a sediment-free inoculum. Biotechnology Progress 12 (3), 310e315. Krumins, V., Park, J.-W., Son, E.-K., Rodenburg, L.A., Kerkhof, L.J., Ha¨ggblom, M.M., Fennell, D.E., 2009. PCB dechlorination enhancement in Anacostia River sediment microcosms. Water Research 43 (18), 4549e4558. Kube, M., Beck, A., Zinder, S.H., Kuhl, M., Reinhardt, R., Adrian, L., 2005. Genome sequence of the chlorinated compound-
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 8 9 9 e3 9 0 7
respiring bacterium Dehalococcides strain CBDB1. Nature Biotechnology 23 (10), 1269e1273. Lendvay, J.M., Lo¨ffler, F.E., Dullhopf, M., Aiello, M.R., Daniels, G., Fathepure, B.Z., Gebhard, M., Heine, R., Helton, R., Shi, J., Krajmalnik-Brown, R., Major, C.L.J., Barcelona, M.J., Petrovskis, E., Hickey, R., Tiedje, J.M., Adriaens, P., 2003. Bioreactive barriers: a comparison of bioaugmentation and biostimulation for chlorinated solvent remediation. Environmental Science & Technology 37 (7), 1422e1431. Major, D.W., McMaster, M.L., Cox, E.E., 2002. Field demonstration of successful bioaugmentation to achieve dechlorination of tetrachloroethene to ethene. Environmental Science & Technology 36 (23), 5106e5116. May, H.D., Cutter, L.A., Miller, G.S., Milliken, C.E., Watts, J.E.M., Sowers, K.R., 2006. Stimulatory and inhibitory effects of organohalides on the dehalogenating activities of PCBdechlorinating bacterium o-17. Environmental Science & Technology 40 (18), 5704e5709. May, H.D., Miller, G.S., Kjellerup, B.V., Sowers, K.R., 2008. Dehalorespiration with polychlorinated biphenyls by an anaerobic ultramicrobacterium. Applied and Environmental Microbiology 74 (7), 2089e2094. Muir, D.C., Wagemann, R., Hargrave, B.T., Thomas, D.J., Peakall, D.B., Norstrom, R.J., 1992. Arctic marine ecosystem contamination. Science of the Total Environment 122 (1e2), 75e134. Natarajan, M.R., Nye, J., Wu, W.-M., Wang, H., Jain, M.K., 1997. Reductive dechlorination of PCB-contaminated raisin river sediments by anaerobic microbial granules. Biotechnology & Bioengineering 55 (1), 182e190. Neff, J.M., 1984. Bioaccumulation of organic micropollutants from sediments and suspended particulates by aquatic animals. Fresenius’ Journal of Analytical Chemistry 319, 132e136. Quensen, J.F., Boyd, S.A., Tiedje, J.M., 1990. Dechlorination of four commercial polychlorinated biphenyl mixtures (Aroclors) by anaerobic microorganisms from sediments. Applied and Environmental Microbiology 56 (8), 2360e2369. Safe, S., 1993. Toxicology, structure-function relationship, and human and environmental health impacts of polychlorinated
3907
biphenyls: progress and problems. Environmental Health Perspectives 100, 259e268. Schecter, A., Stanley, J., Boggess, K., Masuda, Y., Mes, J., Wolff, M., Furst, P., Furst, C., Wilson-Yang, K., Chisholm, B., 1994. Polychlorinated biphenyl levels in the tissues of exposed and nonexposed humans. Environmental Health Perspectives 102 (Suppl 1), 149e158. Seshadri, R., Adrian, L., Fouts, D.E., Eisen, J.A., Phillippy, A.M., Methe, B., Ward, N.L., Nelson, W.C., Deboy, R.T., Khoudry, H. M., Kolonay, J.F., Dodson, R.J., Daugherty, S.C., Brinkac, L.M., Sullivan, S.A., Madupu, R., Nelson, K.E., Kang, K.H., Impraim, M., Tran, K., Robinson, J.M., Forberger, H.A., Fraser, C.M., Zinder, S.H., Heidelberg, J.F., 2005. Genome sequence of the PCE-dechlorinating bacterium Dehalococcoides etheneogenes. Science 307 (5705), 105e108. Smidt, H., de Vos, W.M., 2004. Anaerobic microbial dehalogenation. Annual Review of Microbiology 58 (1), 43e73. Van Dort, H.M., Smullen, L.A., May, R.J., Bedard, D.L., 1997. Priming microbial meta-dechlorination of polychlorinated biphenyls that have persisted in Housatonic River sediments for decades. Environmental Science & Technology 31 (11), 3300e3307. Versar, I., 1976. PCBs in the United States: industrial use and environmental distribution. EPA report 560/6-76-005. Watts, J.E.M., Wu, Q., Schreier, S.B., May, H.D., Sowers, K.R., 2001. Comparative analyses of PCB dechlorinating communities in enrichment cultures using three different molecular screening techniques. Environmental Microbiology 2 (11), 710e719. Wu, Q., Sowers, K.R., May, H.D., 2000. Establishment of a polychlorinated biphenyl-dechlorinating microbial consortium, specific for doubly flanked chlorines in a defined, sediment-free medium. Applied and Environmental Microbiology 66 (1), 49e53. Wu, Q., Watts, J.E.M., Sowers, K.R., May, H.D., 2002. Identification of a bacterium that specifically catalyzes the reductive dechlorination of polychlorinated biphenyls with doubly flanked chlorines. Applied and Environmental Microbiology 68 (2), 807e812.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 0 8 e3 9 1 4
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Identification of estrogen receptor agonists in sediments from Wenyu River, Beijing, China Jianping Luo a,b, Bingli Lei a, Mei Ma a, Jinmiao Zha a, Zijian Wang a,* a
State Key Laboratory of Environmental Aquatic Chemistry, Research Center for Eco-Environmental Sciences, Chinese Academy of Sciences, P.O. Box 2871, Beijing 100085, China b Beijing Municipal Research Institute of Environmental Protection, Beijing 100037, China
article info
abstract
Article history:
Assignment of ecological impacts of contamination to specific classes of contaminants is a
Received 22 November 2010
prerequisite for risk assessment and remediation. In this study, the combination of
Received in revised form
polarity-based fractionation, two-hybrid yeast bioassay, and chemical analysis were used
18 April 2011
to evaluate and identify estrogen receptor agonists (ER-agonists) in sediments from Wenyu
Accepted 27 April 2011
River, Beijing, China. By bioassay, organic raw sediment extracts could induce significant
Available online 11 May 2011
estrogenicity and the bioassay-derived 17b-estradiol equivalents (EEQs) of raw extracts (EEQraws) ranged from 0.8 to 19.8 ng/g dry weight. By polarity-based fractionation, the raw
Keywords:
extracts were separated into three fractions, i.e. non-polar, moderately polar, and polar
Estrogen receptor agonists
fractions, which were subjected to bioassay and chemical analysis. The highest estro-
Nonylphenols
genicity was observed in the polar fraction, which accounted for more than 78% of the
Sediment
total. The medium polar fraction contains PAHs and OCPs, and the estrogenic activities in
Fractionation
this fraction contributed 3%e12% of the total in raw extract. An estrogenic activity of nonpolarity fraction was negligible in compare to other two fractions. By chemical analysis and toxic equivalent calculation, major part of the estrogenicity in polar fraction could be attributed to six natural/synthetic estrogens (16%e63%), i.e. 17b-estradiol, estrone, estriol, 17a-ethynylestradiol, diethylstilbestrol, and b-estradiol-17-valerate, and to nonylphenols (26%e55%). The proposed approach has been successfully used for characterization of ER-agonists in this case study. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
In the aquatic system, exposure of organisms to substances with an estrogenic mode of action has been linked to adverse endocrine effects observed in wild populations, such as feminization and imposex (Schlenk, 2008). Furthermore, these effects can be accumulated and irreversible, possibly appear in next generations, and endanger the sustainable development of the aquatic ecosystem (Zha et al., 2008). Complex mixtures of estrogenic chemicals may be released into the
aquatic system via domestic and industrial effluents (Pothitou and Voutsa, 2008; Ma et al., 2007). Among the estrogenic chemicals, estrogen receptor agonists (ER-agonists) include natural/synthetic estrogens, such as 17b-estradiol (E2), estrone (E1), estriol (E3), 17a-ethynylestradiol (EE2), diethylstilbestrol (DES), b-estradiol-17-valerate (EV), and phytoestrogens; xenoestrogens of industrial or agricultural origins, such as nonylphenols (NPs), phthalates and organochlorine pesticides (OCPs) (Desbrow et al., 1998; Ce´spedes et al., 2004; Peng et al., 2006; Kurihara et al., 2007; Vigano` et al., 2008).
* Corresponding author. Tel./fax: þ86 010 62849140. E-mail address:
[email protected] (Z. Wang). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.04.045
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 0 8 e3 9 1 4
Many ER-agonists such as NPs, EE2, and EV have a high tendency to accumulate in sediment or in aquatic organisms due to their hydrophobicity (Huang et al., 2007; Peck et al., 2004). This suggests that sediment may adsorb considerable amounts of ER-agonists, and pose a potential threat to sediment biota (Navarro et al., 2009; Ying et al., 2003). Therefore, the sediment may be the sink for ER-agonists, as well as form an exposure source for aquatic organisms. Evaluation and identification of ER-agonists in sediments are necessary for the risk assessment and ecological remediation of the water bodies. Although chemical analyses can be used to identify and quantify priori-selected ER-agonists in sewage effluents or contaminated aquatic systems, it is unrealistic to chemically analyze numerous known and unknown ER-agonists (Brack et al., 2007). Alternatively, estrogenic activity of an environmental sample can be evaluated by using rapid, sensitive and relatively inexpensive bioassays such as reporter gene expression assay, a MCF-7 cell proliferation assay, as well as vitellogenin induction assay (Giesy et al., 2002). However the results obtained from bioassays can be used to estimate the total potency of estrogenicity in the mixtures, but hardly to tell the exact components and their contributions to the estrogenic effects. Estrogenic compounds in aquatic environment differ greatly in physicalechemical properties, chemical structures, and in composition in water bodies that have their own distinctive profile features of discharge scenarios. Therefore, it is very difficult but imperative to identify the ER-agonists in a specific field condition with an appropriate approach rather than focus on priori-selected key pollutants alone. A combination of bioassays and chemical analyses has been considered as a powerful tool to characterize the causative agents of estrogenic activity in complex mixtures (Desbrow et al., 1998; Thomas et al., 2001; Ce´spedes et al., 2004), but in a field study it is still challenging to develop and apply a specific and suitable approach adapted to local conditions. Toxic equivalency factor (or relative potencies, RPs) method with the comparison between bioassay-based and analytical concentration-based effect estimates had been developed and used for causality assessment of biological effects in wastewater effluents and sediments (Snyder et al., 2001; Furuichi et al., 2004; Qiao et al., 2006; Houtman et al., 2006). However, successful application of relative estrogenic potency method in identification of estrogen agonists in sediment was not common. Furthermore, relative estrogenic potencies (or E2 equivalent factors, EEFs) for the same chemical may vary greatly in different bioassays, which will affect the subsequent quantitative analysis in identification of the targets, and ranking of chemicals (Legler et al., 2002). In this study a two-hybrid yeast strain containing ERa and coactivator pGAD424 GRIP1, which was constructed in our laboratory and demonstrated to have high sensitivity to a range of known ER-agonists and useful for screening estrogenic activity of compounds that act through the same mode of action (Li et al., 2008), was used to evaluate the total potency of ER-agonists in the sediments. Meanwhile, polarity-based fractionation and EEFs derived by this bioassay were used to help analyze ER-agonists with different polarities. The aim of this study was to quantitatively compare the bioassay-derived activities with relative potencies of analyzed ER-agonists
3909
based on the analytical concentrations and, and to determine which specific compound or class of compounds may account for the bioassay-derived estrogenic activity.
2.
Materials and methods
2.1.
Chemicals and reagents
The mixture standard solutions of OCPs (a-, b-, g-, and d- HCHs, p,p0 -DDD, p,p0 -DDE, p,p0 -DDT) (purity 99%), NPs (a mixture of 11 isomers with a branched C9H19 side chain, purity 97%), bisphenol-A (purity 99%), E1, E2, E3, EE2, DES, and EV (purity 97% for latter six standards) were purchased from SigmaeAldrich, USA. Surrogate standards of 2,4,5,6-tetrachloro-m-xylene, PCB209, bisphenol-A-d14, E2-d12, and internal standards of pentachloronitrobenzene, pyrene-d16 were purchased from the same vender. Derivatization reagents of bis(trimethyl)-trifluoroacetamide (BSTFA) containing 1% of trimethylchlorosilane (TCMS) and trimethylsilylimidazole (TMSI) were purchased from Supelco (Bellefonte, PA). Acetone, hexane, dichloromethane (DCM), methanol (MeOH), ethyl acetate of pesticide grade were obtained from Fisher Scientific (Fair Lawn, NJ). The silica gel (60e200 mesh, Acros Oragnics USA) and neutral alumina (50e200 mesh, Acros Organics USA) were activated at 180 C and 250 C for 12 h, respectively, deactivated by 3% water for cleaning up procedure.
2.2.
Sampling and sample preparation
Wenyu River is the major river to receive treated and untreated industrial effluents and domestic sewage from Beijing municipality. Previous work was carried out on the evaluation of estrogenic activities in water bodies (Ma et al., 2007). Surface sediments (<10 cm depth) were collected from 20 sites of Wenyu River in 2006 (Fig. 1). Composite sediment sample (mixture of 3e5 samples in one sampling location and homogenized on site) was stored in clean stainless-steel containers. The samples were immediately transferred to laboratory and kept at 20 C until further analyses. The sediment samples were mixed with sodium sulfate, which was sequentially rinsed with acetone, completely dried in the fume hood, baked at 450 C for 4 h and stored in muffle oven. The mixture was ultrasonically extracted with DCM/acetone mixture (1/1, v/v) for 20 min. The slurry was then centrifuged at 3000 rpm for 5 min, and the supernatants were collected. The ultrasonic extraction was repeated three times and the extracts were combined. Activated copper was added for desulfurization. An aliquot of the raw extracts was concentrated on a rotary evaporator, solvent-exchanged to dimethyl sulfoxide (DMSO) under a gentle stream of nitrogen gas and stored at 20 C for the two-hybrid yeast bioassay. Another aliquot of raw extracts was exchanged to hexane and fractionated based on polarity using a glass column (10 mm i.d.) containing 15 g of 1/2 alumina/silica gel. The column was eluted with 50 mL hexane to yield non-polar fraction (F1), followed by 70 mL mixture of hexane/DCM (v/v ¼ 7/3) to produce a fraction with moderate polarity (F2), and finally by 40 mL acetone/methanol (v/v ¼ 1/1) to produce
3910
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 0 8 e3 9 1 4
Fig. 1 e Map of the Wenyu River, in Beijing, China and sampling locations.
a polar fraction (F3). The sub-fractions were divided into two parts, respectively. One part was concentrated and solventexchanged to DMSO for bioassay, and the other part to hexane for chemical analysis. Total organic carbon (TOC) of the freeze-dried and sieved (425-mm sieve) sediment samples after removing the inorganic carbon with 21% phosphoric acid was analyzed by TOC analyzer (Appollo 9000, TeKmar-DOHRMANN).
2.3.
Two-hybrid yeast bioassay
The yeast was cultured overnight at 30 C, 130 rpm to reach the exponentially growing period. The 0.3 mL of overnight culture was then added to 2.7 mL of fresh SD medium and the absorption at 600 nm measured. Details of the bioassay were as described elsewhere (Li et al., 2008). Briefly, overnight cultures were diluted with culture medium to an OD 600 nm of 0.75. The bioassay was conducted on 96-well plate. The standard solutions were prepared by serial dilution of E2 stock solution that could yield a full dose-response curve (0.3e272.4 ng/L in culture). Six concentration levels of the raw extracts were prepared by three fold dilution. In each well, the final concentration of DMSO was about 0.5%. Meantime, a 0.5% DMSO solution was also used as the solvent control. Each test solution was assayed in triplicate. Positive control (E2) and negative control were assayed simultaneously on each 96-well plate. After exposure of 3 h, a 50 mL test culture was transferred to a new 96-well plate and after addition of 120 mL of Z-buffer and 20 mL chloroform, the assays were carefully mixed. The enzyme reaction was initiated by addition of o-nitrophenyl-b-D-galactopyranoside and incubated at 30 C for 60 min. The reactions were terminated by the addition of 100 mL Na2CO3 (1 mol/L). The absorbance at 420 nm was determined. The bioassay-derived E2 equivalents (EEQs) were determined by comparing the induction of activity caused by environmental sample extracts with that caused by authentic E2 standards. The minimum detectable concentration was 3.0 pg EEQ/well.
2.4.
Chemical analysis
For chemical analyses, OCPs including DDTs and HCHs, NPs, BPA, and six natural/synthetic estrogens were considered as important endocrine disrupting chemicals in the sediment. The full scan of GCeMS by using the mixture of the individual standards revealed that OCPs were eluted in F2, and NPs, bisphenol-A, and the six estrogens were eluted in F3. F2 fraction was then concentrated and internal standards were added to the final extracts prior to GCeMS analysis for OCPs. The analysis was carried out with an Agilent 6890 series gas chromatograph equipped with a 63Ni electron-capture detector (Agilent, USA). Details of the OCPs analysis have been described elsewhere (Xu et al., 2005). One part of F3 fraction was cleaned up using 5 g alumina and eluted with 20 mL DCM. The eluate was then concentrated, derivatized with bis(trimethylsily1)trifluoroacetamide (BSTFA), and added internal standard pyrene-d16 prior to GCeMS analysis (Agilent 6890 with Agilent 5973 MS, Agilent Technologies, USA). One micro-liter of derivatives was injected into HP-5 MS silica fused capillary column (30 m 0.25 mm 0.25 mm) in splitless mode to analyze NPs. The GC oven temperature was programmed from 50 C (2 min) to 100 C (2min) at 20 C/min, then to 180 C (15 min) at 25 C/ min, and finally to 300 C at 20 C/min (5 min). The results were expressed as the sum of the 11 isomers (Zhou et al., 2005). Another part of F3 was subjected to silica gel/alumina column for cleanup, and were eluted successively using 10 mL of hexane, 20 mL of hexane/acetone (65:35, v/v) and 40 mL of acetone/methanol (1:1, v/v). The six estrogens were in the acetone/methanol fraction. The eluate was concentrated, derived with BSTFA/TCMS/TMSI (99:1:0.5, v/v/v). The separation and detection of the target compounds were achieved by GCeMS using selected-ion mode based on the method from Zhou et al. (2007). A DB-5 MS column (30 m 0.25 mm 0.25 mm) was selected. The GC temperature programs were as follows: initial temperature was 80 C held 1 min, increased to 200 C at a rate of 20 C/min, then increased to 300 C at a rate of 10 C/min and held 10 min, with a total run of 27 min and solvent delay
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 0 8 e3 9 1 4
6 min. The injector and detector temperatures were set at 280 C. The ion source temperature was 230 C, and the energy of the ionizing electron was 70 eV.
2.5.
Quality assurance and quality control
Before analysis, relevant standards were run to check column performance, peak height, resolution, and the limits of detection (LOD). Laboratory quality control procedures include analyses of method blanks, spiked blanks (standards spiked into solvent), matrix spikes, and sample duplicates. None of the target compounds were detected in the procedural blanks for sediment samples. LODs of OCPs and 6 estrogens ranged from 0.10 to 0.15 ng/g and 0.06e0.10 ng/g, respectively, and LODs of NPs and bisphenol-A were 7.4 ng/g, and 0.5 ng/g, respectively. The average recoveries of OCPs, NPs, bisphenol-A, and six estrogens were 65 2% to 107 4%, 71 6%, 62 5%, and 67 4 to 106 9%, respectively. Means were compared using an independent-samples T test, and linear regression (as based on Pearson correlation coefficients) was used with SPSS version 13.0 software.
3.
Results and discussion
3.1.
EEQs of raw extractions (EEQraw)
No apparent cytotoxicity was observed in sediment extracts in the yeast bioassay. Dose-response relationships could be obtained from multiple dilutions for test samples. Significant responses were induced for raw extracts, and the EEQraw ranged from 0.8 to 19.8 ng/g dw (Table 1), which were at the same level to those in sediments from Haihe River in Tianjin (China) (Song et al., 2006) and Lake Shihwa (Korean) (Koh et al.,
Table 1 e 17b-estradiol equivalents of raw extracts (EEQraws) and total organic carbon (TOC) contents of sediments from Wenyu River, Beijing, China. Sites Site 1 Site 2 Site 3 Site 4 Site 5 Site 6 Site 7 Site 8 Site 9 Site 10 Site 11 Site 12 Site 13 Site 14 Site 15 Site 16 Site 17 Site 18 Site 19 Site 20
EEQraws (ng/g) TOC contents
Location
0
00
40 15 18 N 40 080 0800 N 40 060 5300 N 40 070 5300 N 40 070 2300 N 40 070 3800 N 40 060 3900 N 40 040 3000 N 40 040 0800 N 40 030 1100 N 40 020 2000 N 40 010 3100 N 40 000 3900 N 40 000 0700 N 39 570 4300 N 39 560 1900 N 39 550 4300 N 39 550 0100 N 39 540 2700 N 39 540 0300 N
0
00
116 16 05 E 116 160 5200 E 116 160 1500 E 116 190 3500 E 116 220 1200 E 116 270 1800 E 116 290 1000 E 116 290 5500 E 116 310 1300 E 116 320 1900 E 116 330 2100 E 116 340 1400 E 116 350 5300 E 116 380 0100 E 116 380 1300 E 116 380 2300 E 116 380 4900 E 116 390 5000 E 116 400 3900 E 116 410 1600 E
0.9 12.3 19.8 1.1 2.3 2.7 3.2 0.8 14.4 9.3 1.0 5.1 4.2 15.5 5.6 3.1 15.7 6.2 12.6 14.2
0.2 2.3 1.9 0.2 0.4 0.5 0.4 0.1 2.2 0.9 0.1 0.7 0.7 1.8 1.0 0.3 2.6 1.2 2.0 0.9
2% 28% 33% 5% 1% 2% 1% 1% 15% 2% 1% 3% 2% 27% 8% 3% 18% 15% 10% 12%
3911
2005). The presence of various ER-agonists in sediment should originate from sewage effluents and industrial discharges (Ternes et al., 1999; Thomas et al., 2001). Total amount of industrial effluents and municipal sewage that were discharged into Wenyu River was up to 1.7 105 m3/d, and about half of them were not treated (Beijing Hydraulic Research Institute (BHRI), 2006). Previous work had investigated the estrogenic activities of effluents from five sewage treatment plants and thirteen different industries in Beijing, and it was found that the wastewater discharges along the river might be a major source of ER-agonists (Ma et al., 2007). The biologically derived EEQraws were significantly correlated with TOC contents of the sediment samples (r ¼ 0.82, P < 0.01). It indicated that sediments with high organic carbon content were more likely to retain ER-agonists than those with lower organic carbon levels. This observation is in agreement with previous conclusion that the higher amounts of estrogens could occur in sediments with high organic matter content (Gobas and Maclean, 2003; Samir et al., 2006).
3.2.
EEQs of different fractions
Five sediment samples (site 3, site 9, site 12, site 18, and site 20) were selected for further fractionation-based bioassay, and the results were shown in Fig. 2. From the figure, the polar fraction (F3) showed the highest estrogenic activity. The estrogenic activity in this fraction contributed about 78%e115% of the total in raw extract. It could be noted that, for some samples, higher estrogenic activities were in polar fraction than in raw extract. It was suggested that the extracts should contain some interfering substances that interfere the bioassay results for agonists (Snyder et al., 2001). The medium polar fraction (F2) contains PAHs and OCPs, and the estrogenic activities in this fraction contributed 3e12% of the total in raw extract. An estrogenic activity of non-polarity fraction was negligible in compare to other two fractions. This fraction contains PCBs, aliphatic hydrocarbons etc., which contribute mainly to aryl hydrocarbon receptor agonistic effect (Luo et al., 2009). Based on the results of fractionation-based bioassay, it can be concluded that majority of ER-agonists in the river sediment should be in polar fraction, followed by in moderate polar fraction.
3.3. Estimation of chemical analysis derived EEQs (EEQchems) Concentrations of NPs, BPA, and six natural/synthetic estrogens in polar fraction, and OCPs including DDTs and HCHs in moderately polar fraction were determined and the results were listed in Table 2. The NPs concentrations in sediment samples ranged from 1.59 to 7.94 mg/g dw, that were two to four orders of magnitude higher than those of the other analytes. BPA concentrations in sediment samples ranged from P 4.3 to 55.3 ng/g dw. Concentrations of the six estrogens ( est) ranged from 2.2 to 39.0 ng/g dw. Concentrations of OCPs ranged from 8.9 to 35.3 ng/g dw. In comparison with published results, concentrations of NPs and BPA in Wenyu River were at the same level to those in River Po in Italy (Vigano` et al., 2008); the levels of E1, E2, E3 and EE2 in sediments were lower than
3912
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 0 8 e3 9 1 4
Fig. 2 e Estrogenic activities of individual fractions separated by alumina/silica gel column based on the two-hybrid yeast.
those from 3 rivers in Tianjin area in China (Lei et al., 2009), at the same level to those from Ouse and Uck River, England (Liu et al., 2004), and higher than those from Tokyo Bay, Japan (Isobe et al., 2006) and Rivers in Germany (Ternes et al., 2002). To derive EEQchem, relative estrogenic potencies (RPs) or E2 equivalent factors (EEFs) of different analytes were obtained according to dose-response curves of OCPs, NPs, BPA, and the six estrogens (Table 2). The results showed that EEFs of the tested chemicals were widely variable covering six orders of magnitude. The EEFs for EE2 and EV were six and seven times lower than that of E2, respectively. The EEFs for E1, DES, and BPA were about 2 orders of magnitudes lower than that for E2. The EEFs for E3 and NPs were about 3 orders of magnitudes, and that for p,p0 -DDT was about 6 orders of magnitude lower
than that for E2. EEFs of other OCPs could not be obtained or their EEFs should be less than 107 in compare to unit of E2, due to no significant enzyme induction at the highest exposure dose where even cytotoxicity occurred. EEQ of individual chemical could be calculated through multiplying its chemically determined concentration by its EEF. The total calculated EEQchem was then summarized from individual EEQ (Table 2). EEQchems ranged from 3.3 to 15.3 ng/g dw, and accounted for 65%e106% of EEQraws. Moreover, a significant correlation was observed between EEQchems and EEQraws (r ¼ 0.89, P < 0.05). From Table 2, the contributions of EEQchems derived from the six estrogens (EEQests) and NPs (EEQNPs) corresponded to 16e63% and 26e55% of the EEQraws, respectively. Among the six estrogens, E2 was obviously the
Table 2 e Concentrations, calculated EEQs, and 17b-estradiol equivalent factors (EEFs) of the 6 estrogens, NPs, BPA, and OCPs in sediments from Wenyu River, Beijing, China. EEFs Chemicals DES E1 E2 EE2 E3 EV EEQests NPs EEQNPs BPA EEQBPAs p,p0 -DDT P OCPs EEQOCPs EEQchems EEQraws
0.021 0.053 1.0 0.17 0.0049 0.14 e 0.001 e 0.015 e 0.0000054 e e e e
Concentrations at selected sites (ng/g dry weight) Site 3
Site 9
Site 12
Site 18
Site 20
1.6 21.8 3.4 1.3 0.8 10.1 6.2 6755.4 6.8 55.3 0.8 13.5 27.9 0.00007 13.8 19.8
0.3 15.8 4.2 0.9 0.8 13.4 7.1 7938.5 7.9 19.3 0.3 5.9 18.7 0.00003 15.3 14.4
0.2 1.1 0.7 0.2 ND ND 0.8 2243.4 2.2 18.5 0.3 2.1 8.9 0.00001 3.3 5.1
0.1 0.5 3.6 0.8 ND 1.28 3.9 1591.6 1.6 4.3 0.1 7.5 35.3 0.00004 5.6 6.2
0.9 4.9 2.2 ND 0.3 7.4 3.5 7440.6 7.4 7.8 0.1 2.3 13.2 0.00001 11.1 14.2
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 0 8 e3 9 1 4
most potent risk factor that contributes about 14e58% to the EEQraws, and 55e92% to the EEQests. The sum of these two types of ER-agonists could account for 60e104% of EEQraws. Contribution of BPA was only 1e5%, and that of OCPs was almost negligible. Some nonylphenol ethoxylates known as precursors of NPs might also result in estrogenic activity despite of their less EEFs (Legler et al., 2002; Routledge and Sumpter, 1996). However, their contribution may not be higher than that of NPs and negligible in the EEQchem calculation. Therefore, one can conclude that the six estrogens and NPs should be the dominant contributor to the observed estrogenic activities in the river sediment. In this study, incomplete recovery of target compounds might arise from extraction, fractionation and cleaning, which would affect the comparison between chemical analysis and bioassay. For ultrasonic extraction, the recovery of target compounds was validated to be very good in our laboratory. For fractionation, the raw extracts were fractionated into three parts based on polarity, and it is impossible that one class of chemicals were all included in one fraction. However, the carry-over in other fractions could not be remarkable based on the results of full scan with standards, and fractionation had no influence on the sum effects either. Furthermore, with comparison of our previous study (Lei et al., 2009) where no fractionation was used, it was found that the recoveries of estrogens in the two studies were similar. This also indicated that the loss caused by fractionation in this study was not remarkable. For cleaning, this process could be important for the loss of target chemicals, but cleaning process was not used for bioassay. Therefore, the bioassay result could be underestimated due to incomplete recovery, but not remarkable. Natural/synthetic estrogens and NPs have been recognized to be important ER-agonists for aquatic organisms (Brack et al., 2007; Vigano` et al., 2008). In previous field work, direct association of estrogenic activity with the natural/synthetic estrogens, such as E2, E1 and EE2, has been observed by many environmental scientists (reviewed by Petrovic et al., 2004). For example, E2 and EE2 were proposed to be important ERagonists in water samples from mid-Michigan and Lake Mead, NV (Snyder et al., 2001). Plasma vitellogen levels in male carp were found to correlate with concentration levels of NPs, possibly as well as natural/synthetic estrogens on the Llobregat basin in Spain (Sole´ et al., 2000; Petrovic et al., 2002). Other studies have reported that NPs may be the major contributor to estrogenicity in sediments, and might contribute to feminization (Quiro´s et al., 2005; Kurihara et al., 2007). However, it is still inconclusive whether the ER-agonists could induce significant adverse endocrine effects for aquatic animals in Wenyu River without evidence of in vivo bioassays. Other studies were needed to assess the bioavailability and in vivo biological effects of these ER-agonists in this river.
4.
Conclusion
The combination of the chemical analysis and the bioassay demonstrated that ER-agonists were mainly distributed in the polar fraction, and natural/synthetic estrogens and NPs dominantly contributed to the estrogenic activities in the sediments.
3913
The fractionation-based bioassay, chemical analysis and EEF method used in this study could be a useful tool in quantitatively screening and characterization of ER-agonists in river sediments for monitoring and regulatory purposes.
Acknowledgments This study was supported by Chinese Academy of Science (KZCX1-YW-06-02), the National Basic Research Program of China (2007CB407304) and National Natural Science Foundation of China (40703025).
references
Beijing Hydraulic Research Institute (BHRI), 2006. Guidelines of Water Body Remediation and Water Resource Utilization for Wenyu River, Beijing, China, pp. 30e50 (in Chinese). Brack, W., Klamer, H., de Alda, M., Barcelo´, D., 2007. Effectdirected analysis of key toxicants in European river basins, a review. Environ. Sci. Pollut. Res. 14, 30e38. ´ , Pin˜a, B., Lacorte, S., Ce´spedes, R., Petrovic, M., Raldu´a, D., Saura, U Viana, P., Barcelo´, D., 2004. Integrated procedure for determination of endocrine-disrupting activity in surface waters and sediments by use of the biological technique recombinant yeast assay and chemical analysis by LCeESI-MS. Anal. Bioanal. Chem. 378, 697e708. Desbrow, C., Routledge, E., Brighty, G., Sumpter, J., Waldock, M., 1998. Identification of estrogenic chemicals in STW effluent. 1. Chemical fractionation and in vitro biological screening. Environ. Sci. Technol. 32, 1549e1558. Furuichi, T., Kannan, K., Diesy, J.P., Masunaga, S., 2004. Contribution of known endocrine disrupting substances to the estrogenic activity in Tama River water samples from Japan using instrumental analysis and in vitro reporter gene assay. Water Res. 38, 4491e4501. Giesy, J.P., Hilscherova, K., Jones, P., Kannan, K., Machala, M., 2002. Cell bioassays for detection of aryl hydrocarbon (AhR) and estrogen receptor (ER) mediated activity in environmental samples. Mar Pollut. Bull. 45, 3e16. Gobas, F.A.P.C., Maclean, L.G., 2003. Sediment-water distribution of organic contaminants in aquatic ecosystems: the role of organic carbon mineralization. Environ. Sci. Technol. 37, 735e741. Houtman, C., Booij, P., Jover, E., del Rio, D., Swart, K., van Velzen, M., Vreuls, R., Legler, J., Brouwer, A., Lamoree, M., 2006. Estrogenic and dioxin-like compounds in sediment from Zierikzee harbour identified with CALUX assay-directed fractionation combined with one and two dimensional gas chromatography analyses. Chemosphere 65, 2244e2252. Huang, G., Hou, S., Wang, L., Sun, H., 2007. Distribution and fate of nonylphenol in an aquatic microcosm. Water Res. 41, 4630e4638. Isobe, T., Serizawa, S., Horiguchi, T., Shibata, Y., Managaki, S., Takada, H., Morita, M., Shiraishi, H., 2006. Horizontal distribution of steroid estrogens in surface sediments in Tokyo Bay. Environ. Pollut. 144, 632e638. Koh, C.H., Khim, J.S., Villeneuve, D.L., Kannan, K., Johnson, B.G., Giesy, J.P., 2005. Instrumental and bioanalytical measures of dioxin-like and estrogenic compounds and activities associated with sediment from the Korean coast. Ecotoxicol Environ. Saf. 61, 366e379. Kurihara, R., Watanabe, E., Ueda, Y., Kakuno, A., Fujii, K., Shiraishi, F., Hashimoto, S., 2007. Estrogenic activity in
3914
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 0 8 e3 9 1 4
sediments contaminated by nonylphenol in Tokyo Bay (Japan) evaluated by vitellogenin induction in male mummichogs (Fundulus heteroclitus). Mar Pollut. Bull. 54, 1315e1320. Legler, J., Dennekamp, M., Vethaak, A., Brouwer, A., Koeman, J., van der Burg, B., Murk, A., 2002. Detection of estrogenic activity in sediment-associated compounds using in vitro reporter gene assays. Sci. Total Environ. 293, 69e83. Lei, B., Huang, S., Zhou, Y., Wang, D., Wang, Z., 2009. Levels of six estrogens in water and sediment from three rivers in Tianjin area, China. Chemosphere 76, 36e42. Li, J., Li, N., Ma, M., Giesy, J.P., Wang, Z., 2008. In vitro profiling of the endocrine disrupting potency of organochlorine pesticides. Toxicol. Lett. 183, 65e71. Liu, R., Zhou, J., Wilding, A., 2004. Microwave-assisted extraction followed by gas chromatography-mass spectrometry for the determination of endocrine disrupting chemicals in river sediments. J. Chromatogr. A 1038, 143e159. Luo, J., Ma, M., Zha, J., Wang, Z., 2009. Characterization of aryl hydrocarbon receptor agonists in sediments of Wenyu River, Beijing, China. Water Res. 43, 2441e2448. Ma, M., Rao, K., Wang, Z., 2007. Occurrence of estrogenic effects in sewage and industrial wastewaters in Beijing, China. Environ. Pollut. 147, 331e336. Navarro, A., Endo, S., Gocht, T., Barth, J., Lacorte, S., Barcelo´, D., Grathwohl, P., 2009. Sorption of alkylphenols on Ebro River sediments: comparing isotherms with field observations in river water and sediments. Environ. Pollut. 157, 698e703. Peck, M., Gibson, R., Kortenkamp, A., Hill, E., 2004. Sediments are major sinks of steroidal estrogens in two United Kingdom rivers. Environ. Toxicol. Chem. 23, 945e952. Peng, X., Wang, Z., Yang, C., Chen, F., Mai, B., 2006. Simultaneous determination of endocrine-disrupting phenols and steroid estrogens in sediment by gas chromatography-mass spectrometry. J. Chromatogr. A 1116, 51e56. Petrovic, M., Sole´, M., Lo´pezde Alda, M., Barcelo´, D., 2002. Endocrine disruptors in sewage treatment plants, receiving river waters, and sediments: integration of chemical analysis and biological effects on feral carp. Environ. Toxicol. Chem. 21, 2146e2156. Petrovic, M., Eljarrat, E., Lopez de Alda, M., Barcelo, D., 2004. Endocrine disrupting compounds and other emerging contaminants in the environment: a survey on new monitoring strategies and occurrence data. Anal. Bioanal. Chem. 378, 549e562. Pothitou, P., Voutsa, D., 2008. Endocrine disrupting compounds in municipal and industrial wastewater treatment plants in Northern Greece. Chemosphere 73, 1716e1723. Qiao, M., Chen, Y., Zhang, Q., Huang, S., Ma, M., Wang, C., Wang, Z., 2006. Identification of Ah receptor agonists in sediment of Meiliang Bay, Taihu Lake, China. Environ. Sci. Technol. 40, 1415e1419. Quiro´s, L., Ce´spedes, R., Lacorte, S., Viana, P., Raldu´a, D., Barcelo´, D., Pin˜a, B., 2005. Detection and evaluation of endocrine-disruption activity in water samples from Portuguese rivers. Environ. Toxicol. Chem. 24, 389e395. Routledge, E., Sumpter, J., 1996. Estrogenic activity of surfactants and some of their degradation products assessed using
a recombinant yeast screen. Environ. Toxicol. Chem. 15, 241e248. Samir, K.K., Xian, B., Thompson, M.L., Sung, S., Ong, S.-K., Leeuwen, J.V., 2006. Fate, transport, and biodegradation of natural estrogens in the environment and engineered systems. Environ. Sci. Technol. 40, 6537e6545. Schlenk, D., 2008. Are steroids really the cause for fish feminization? A mini-review of in vitro and in vivo guided TIEs. Mar Pollut. Bull. 57, 250e254. Snyder, S., Villeneuve, D., Snyder, E., Giesy, J.P., 2001. Identification and quantification of estrogen receptor agonists in wastewater effluents. Environ. Sci. Technol. 35, 3620e3625. Sole´, M., Lo´pez de Alda, M.J., Castillo, M., Porte, C., LadegaardPedersen, K., Barcelo´, D., 2000. Estrogenicity determination in sewage treatment plants and surface waters from the Catalonian area (NE Spain). Environ. Sci. Technol. 34, 5076e5083. Song, M., Xu, Y., Jiang, Q., Lam, P.K.S., O’Toole, D.K., Giesy, J.P., Jiang, G.B., 2006. Measurement of estrogenic activity in sediments from Haihe and Dagu River, China. Environ. Int. 32, 676e681. Ternes, T., Stumpf, M., Mueller, J., Haberer, K., Wiken, R., Servos, M., 1999. Behavior and occurrence of estrogens in municipal sewage treatment plantsdI. Investigations. Sci. Total Environ. 225, 81e90. Ternes, T., Andersen, H., Gilberg, D., Bonerz, M., 2002. Determination of estrogens in sludge and sediments by liquid extraction and GC/MS/MS. Anal. Chem. 74, 3498e3504. Thomas, K., Hurst, M., Matthiessen, P., Waldock, M., 2001. Characterization of estrogenic compounds in water samples collected from United Kingdom estuaries. Environ. Toxicol. Chem. 20, 2165e2170. Vigano`, L., Benfenati, E., van Cauwenberge, A., Eidem, J., Erratico, C., Goksøyr, A., Kloas, W., Maggioni, S., Mandich, A., Urbatzka, R., 2008. Estrogenicity profile and estrogenic compounds determined in river sediments by chemical analysis, ELISA and yeast assays. Chemosphere 73, 1078e1089. Xu, Y., Wang, Z., Ke, R., Khan, S., 2005. Accumulation of organochlorine pesticides from water using triolein embedded cellulose acetate membranes. Environ. Sci. Technol. 39, 1152e1157. Ying, G., Kookana, R., Dillon, P., 2003. Sorption and degradation of selected five endocrine disrupting chemicals in aquifer material. Water Res. 37, 3785e3791. Zha, J., Sun, L., Zhou, Y., Spear, P., Ma, M., Wang, Z., 2008. Assessment of 17a-ethinylestradiol effects and underlying mechanisms in a continuous, multigeneration exposure of the Chinese rare minnow (Gobiocypris rarus). Toxicol. Appl. Pharm. 226, 298e308. Zhou, Y., Ma, M., Wang, Z., 2005. Determination of nonylphenols in water by pre-column derivatization and gas chromatography coupled with mass spectrometry. Chin. J. Anal. Chem. 32, 1109e1112. Zhou, Y., Wang, Z., Jia, N., 2007. Formation of multiple trimethylsilyl derivatives in the derivatization of 17aethinylestradiol with BSTFA or MSTFA followed by gas chromatography-mass spectrometry determination. J. Environ. Sci. 19, 879e884.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 1 5 e3 9 2 4
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
River health assessment in peri-urban landscapes: An application of multivariate analysis to identify the key variables U. Pinto, B.L. Maheshwari* School of Natural Sciences, Hawkesbury Campus, Building H3, University of Western Sydney, Locked Bag 1797, Penrith South, DC, NSW 1797, Australia
article info
abstract
Article history:
An array of river health assessment approaches and water quality variables have been
Received 19 November 2010
suggested in the past for assessing the level of river health. However, the selection of
Received in revised form
suitable variables to be monitored for the assessment remains ambiguous and often it is
20 April 2011
not practical to monitor all the suggested variables. In this study, we employ a multivariate
Accepted 27 April 2011
data reduction technique, called Factor Analysis (FA), to identify the key river health
Available online 6 May 2011
variables for a peri-urban river system, viz., the Hawkesbury-Nepean River system in New South Wales, Australia. Out of 40 water quality variables included in the analysis, the FA
Keywords:
identified nine key variables, under three varifactors (VFs), explaining 50% of the variance
Water quality monitoring
in the river water quality. Variables in the first, second and third VFs revealed anaerobic
Factor analysis
conditions, microbial quality and effects of eutrophication in the Hawkesbury-Nepean
River health assessment
River. Thus, the present work shows a notable reduction in the number of variables and
Hawkesbury-Nepean river
the application of FA for identification of key variables was found promising. The finding of this study has potential application in designing a cost-effective river health monitoring program by reducing the number of variables to be monitored in a peri-urban situation. It can also assist in partitioning variables according to their unique contribution to the total variance. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
Rivers play a key role in ecosystems and society, and they provide a range of ecosystem functions such as shelter and food source for an array of biological species, aid in flood management and ecological refuge development (Hoelzl, 2007). Socially, rivers accommodate communities by providing a medium for transport, recreation, tourism, worship, ecosystem services and a place to experience the serenity of nature. Unfortunately, many peri-urban rivers draining from extensive urban and agricultural areas have become highly degraded over the past few decades and remain a sensitive issue in the agenda of river
management authorities. For the past decade, countries across the globe have been in a constant battle developing the most suitable method to assess the health of freshwater for biotic species and humans, viz., IUCN Global Freshwater Initiative (International Union for Conservation of Nature and Natural Resources), Healthy Watershed Initiative in the US (Young and Sanzone, 2002), Pressure, State, Response model in Australia (Commonwealth of Australia, 1996) and EU Water Framework Directive in Europe (Kaika, 2003). Surface freshwater streams that are severely impacted due to anthropogenic influence are said to be suffered from ‘Urban stream syndrome’ (Walsh et al., 2005). One possible way to
* Corresponding author. Tel.: þ61 4570 1235. E-mail address:
[email protected] (B.L. Maheshwari). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.04.044
3916
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 1 5 e3 9 2 4
overcome the urban stream syndrome and manage the river health effectively is by understanding the interactions of anthropogenic and natural factors, particularly through a well-designed and cost-effective monitoring program. For this reason, many government agencies collect a standard group of variables regularly for routine assessments of river condition. These indicators are used to identify a system departing from the normal values. However, such deviations need to be in the form of extreme nature to be clearly detected by the indicators of interest (Boulton, 1999). In contrast to routine monitoring, a specific set of variables are also collected (intermittently or seasonally) to be used with specific river health assessment approaches. However, routine monitoring of multiple variables is an expensive task and measuring variables for different river health assessment methodologies becomes complex when there is a radical change in underlying approaches.
2. River health assessment and water quality monitoring Bewildering varieties of river health assessment approaches have been reported in literature, and there are often specific problems and limitations with some of the most frequent approaches. The river health assessment approaches reported to date tend to be based on single perspective studies (i.e., biological, floral and faunal species), ecological function based studies (i.e., plant respiration and photosynthesis rates) and composite studies (i.e., based on water quality and macroinvertebrate indices). The single perspective studies are based on the fact that each species is capable of reflecting the environmental stress, as a response indication, which can be monitored over different time scales. The single perspective, biological assessments also involve study of distribution and abundance of a wide range of faunal and floral groups such as macroinvertebrates, fish, diatoms and algae species (Chessman, 1995; Harris, 1995; Turner and Rabalais, 2003; Whitton and Kelly, 1995). Ecological function based methods on the other hand are indirect methods of assessing river health and examining how biological communities function and interact with the environment (Boulton, 1999). Composite indices incorporate either the quality of many aspects of a chosen ecosystem (e.g. Index of Stream Condition) or sub-levels of a single environmental component (e.g. Water Quality Index) (ISC, 2009, Vugteveen et al., 2006). Many predictive models compare the quality of an impacted site with a reference site on the basis that biological species, which share a similar habitat at different sites, are more alike than species which belong to different habitats but are found at the same site (Pearson and Norris, 1996). For example, the River Invertebrate Prediction And Classification System (RIVPACS), Index of Biotic Integrity (IBI) and the Australian River Assessment System (AUSRIVAS) are three widely used predictive models that utilise macroinvertebrates and fish assemblages data for making site-specific predictions (Wright, 1995; Karr, 1991; Barmuta et al., 2002). On the other hand, ranking matrices such as visually based habitat assessments (HABSCORE and Rapid Appraisal of Riparian Condition), are
the simplest and quickest to perform and results can be obtained within 30e60 min (Barbour et al., 1999; Jansen and Robertson, 2001). The current application of ranking scores is mostly restricted for riparian habitat assessments.
2.1.
Issues in river health assessment approaches
Major problems associated with river health assessment approaches are related to the rationale for variable selection and our limited understanding of the environmental complexity. In the past, recommendations from a panel of experts who interact remotely and anonymously provide feedback, a method commonly known as ‘DELPHI’ forecasting was considered as the best way to select variables to be included in water quality indices (Dalkey and Helmer, 1963). This has historically gained much acceptance although it suffers from lack of human understanding of complex interactions of ecosystems. Another river health assessment approach considers a hierarchical model, which combines catchment, habitat and biota, assuming that the components at higher levels of hierarchy affect components at lower levels (i.e., catchment health affects river health). In other words, this methodology chooses variables on the assumption that ecological integrity is represented by all the major components of the environment that comprise an ecosystem (Norris et al., 2007). Thus, a package of representative variables on catchment disturbance, hydrological changes, water quality, soil, physical form, fringing zone and aquatic biota are included to assess the overall health of river. Similarly, some researchers proposed to measure riparian habitat conditions and argue that habitats close to the river waters have a stronger relationship between their physical and biological composition and the river health (Jansen and Robertson, 2001). However, the above argument proves the lack of coherence in variable selection procedures for the purpose of reliable river health assessment. Secondly, it is becoming extremely difficult to find sites that are ‘pristine’ for use as reference sites to establish indicators and subsequently develop composite indices. The river health assessment approaches which are heavily dependent on condition of reference sites (Observed versus Expected commonly known as O/E ratio) have a tendency to misinterpret the true state of rivers. In Australia, many composite river health approaches such as AUSRIVAS, FARWH (National Framework for the Assessment of River and Wetland Health), IRC (Tasmanian Index of River Condition) and SRA (Sustainable River Audit, systematic river health assessment for the Murray Darling Basin) consider the state of pristine condition or pre-European reference conditions to assess the health of river systems at state and national levels (Barmuta et al., 2002; Norris et al., 2007; Askey-Doran et al., 2009; Peter et al., 2008). However, if the reference sites are already impacted, the river health assessment becomes erroneous. Similarly, selecting reference sites which are ‘minimally-disturbed’ may become difficult due to large environmental variability inherited in individual sites (Underwood, 1994). Thirdly, there are complex interactions occurring in natural environment, and for this reason incorporating ecological interactions into river health assessment approaches becomes
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 1 5 e3 9 2 4
difficult. For example, we still do not know the full extent of response thresholds to a given stressor by ecological functions or biological organisms (i.e., leaf-litter decomposition is increased by nutrient enrichment but decreased by acidic pH water) (Young et al., 2004). As such, the interpretation of ecological function based methodologies has become cumbersome due to multiple factors affecting a chosen ecological function. Similarly, the distribution and abundance of species and communities have been widely used to assess river health in response to two distinct stressor levels. However, distribution and abundance of biological species are concurrently affected by predators and internal community dynamics. Thus, it is extremely difficult to predict whether the change in species distribution and abundance is due to an external impact or community dynamics. Past studies indicated that diversity remains highest at the intermediate levels of disturbance and decreases at higher levels of disturbance due to stress (Huston, 1979; Connell, 1974). Therefore, methods based on community interactions and ecological functions require an in-depth understanding of the community dynamics before a meaningful assessment of river system health is made. In order to assess river health comprehensively, there is a need to establish a clear rationale to identify key variables of river health. There are numerous problems associated with specific water quality variables monitored for river health assessment approaches as well as variables selected for routine assessment of river condition. Often, ongoing river monitoring schemes yield a large volume of data that is expensive in terms of collection and storage. The intensity of monitoring is also questioned when budget cuts occur. Clearly, there is a need to make monitoring more effective in terms of cost and effort involved. Considering that river health is analogous to human health (Fairweather, 1999; Schofield and Davies, 1996), in this study we attempt to understand the key variables of peri-urban river systems to include in river health assessment similar to a physician assessing patient health based on major symptoms. Identifying key variables of river health has many advantages. Firstly, with a small number of representative variables, government agencies could significantly reduce their monitoring cost. Secondly, key variables can be examined for a specific river management purpose such as developing a risk assessment framework for recreation activities or identifying an emerging problem in the river health. Therefore, using the Hawkesbury-Nepean River system in New South Wales, Australia as a case study, the main objective of this study is to employ a multivariate approach to identify key variables that have a significant relationship with overall changes in river water quality in peri-urban contexts.
3.
Materials and methods
3.1.
Application of multivariate techniques
In this study, we employ a multivariate approach because multivariate tools can handle interacting water quality variables on a spatio-temporal scale and guide in developing management strategies for complex water resources within
3917
a region (Pejaman et al., 2009; Vega et al., 1998). Based on matrix algebra, Factor Analysis (FA) has been selected because it can handle large sets of data points often present in routine water quality monitoring programs. The analysis can help to reduce the dimensionality of the data set to a manageable size and keeps as much of the original information as possible. In water research, FA has been employed to link marine community interaction with environmental variables to assess seasonal and temporal variation in surface water bodies and to understand groundwater interactions in multiple locations (Clarke and Ainsworth, 1993; Shrestha and Kazama, 2007; Winter et al., 2000). For the application of FA, it is a necessary requirement to remove the highly correlated parameters prior to the analysis because singularity (variables that are highly correlated) prevents the determination of unique contribution of a particular variable to a factor (Fields, 2005). This is a crucial step because in reality agencies measure a vast range of water quality parameters that are similar in nature. Avoiding the overlapping variables or variables that explain the same dimension of water quality greatly saves a monitoring budget, especially when there are limited funds available. By extracting the most useful groups of variables, FA helps to understand the structure of a data set in terms of latent factors (i.e., the factors that are not directly measured but reflect the interactions of several individual variables) and sources of pollution unique to a chosen river system (Fields, 2005). In the present study, we employ FA to considerably reduce the number of variables obtained in a routine monitoring program in the Hawkesbury-Nepean River system (HNR) and identify the latent factors relative to river health in peri-urban landscapes.
3.2.
The Hawkesbury-Nepean River system
The HNR is the main source of water supply for the Sydney Metropolitan area. The HNR system is a combination of two major rivers, the Nepean River (155 km) and the Hawkesbury River (145 km) (Markich and Brown, 1998). The Nepean River becomes the Hawkesbury River at the Grose River confluence near a rural town of Yarramundi, NSW (Fig. 1). The river system is complex in nature, the upper one-third being edged with many poorly accessible gorges, the middle one-third running through irrigated farm lands and the lower one-third having tidal slopes with alluvial soil pockets (Diamond, 2004). The catchment of the HNR covers about 22,000 km2 and the total length of the system is about 300 km. The HNR catchment consists of three major geological formations, the Narrabeen Group (conglomerate sedimentary rocks and shale), Hawkesbury sandstone (quartz-rich sandstone) and the Wianmatta Group (green, grey and black shale) (Markich and Brown, 1998). Due to a large number of urban and peri-urban activities, the HNR catchment presents some particular challenges in terms of water quality and health of the HNR system. Currently land use in the catchment includes heavily urbanised regions, industrial regions, recreational regions, agricultural regions and scenically attractive regions (Baginska et al., 2003; Pinto and Maheshwari, 2010). There are numerous point and diffuse sources of anthropogenic
3918
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 1 5 e3 9 2 4
Fig. 1 e Map with key locations on the Hawkesbury-Nepean River system (- [ sampling location).
pollution that primarily originate from peri-urban, agricultural and industrial activities. Point source pollutions are attributed to sewerage treatment plants, mining activities and discharged industrial effluent while diffuse sources of pollution are related to urban runoff and agricultural activities associated with farms and market gardens. Agricultural runoff contributes approximately 40e50% of phosphorus loads and 25% of nitrate loads into the HNR system between Wiseman’s Ferry and the head waters which are believed to have originated from improved pasture, market gardens and animal farms (Markich and Brown, 1998). Within the Hawkesbury River catchment, vegetation clearance has been continuously practiced over the last 200 years causing increased subsurface and agricultural runoff and sediment loads into the river system (Thoms et al., 2000). Unlike other natural rivers where flow is dominated by rainfall events, flow of HNR is highly regulated by impoundments and treated effluent discharge from STPs. There are about 22 dams and 15 weirs situated along the HNR system. The major dam on this river is at Warragamba, which holds about 2.057 109 m3 of water captured from a 9000 km2 catchment area (Turner and Erskine, 2005). There are four other dams, viz., Cataract, Cordeaux, Avon and Nepean which determine flow pattern of the HNR system. A number of constructed impoundments have considerably reduced flood events of the river and natural flow regimes (Turner and Erskine, 2005). Further, there are 18 sewage treatment plants along the HNR discharging significant volumes of treated municipal wastewater into the river (Howard, 2009). In general, the condition of the HNR system has changed considerably over the past few decades and careful management is required based on efficient river health assessment methodologies. Further, the influence of urban factors (particularly the discharge of effluent and stormwater) sets
peri-urban rivers, such as HNR, apart from rivers in dominantly rural catchments.
3.3.
Data analysis
We obtained 40 parameters measured on monthly basis for four adjacent sites (Wiseman’s Ferry, Lower Portland, Wilberforce and North Richmond) on the Hawkesbury River, part of the HNR. The data supplied by the Sydney Catchment Authority, were for two consecutive years viz., 2008 and 2009. The Wiseman’s Ferry and Lower Portland sites mostly contained brackish water. These sites were selected due to the continuity in the range of water quality parameters collected between 2008 and 2009. Data reported as below detection level (
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 1 5 e3 9 2 4
Table 1 e Highly correlated variables in the study. Variable 1 Al Total Al Total Mn Total Algal BioVol ASU Cyano Coli. Therm Coli TC Algal TC Cyano BioVol Cyano BioVol Toxic Cyano BioVol Cyano BioVol NTN Loren UV UV SS
Variable 2
Pearson Correlation Coefficient
NTU Fe Total Fe Total ASU algae Algal BioVol E. coli E. coli Cyano TC ASU Cyano Toxic Cyano BioVol ASU Toxic ASU Toxic NOx Chl-a NTU Colour NTU
0.827 0.912 0.724 0.920 0.800 0.923 0.794 0.996 0.968 0.923 0.991 0.909 0.817 0.996 0.841 0.930 0.712
4.
Results and discussion
4.1.
Variability in water quality data
Table 1 and Fig. 2 show the highly correlated variables observed in the initial data set. The descriptive statistics of the monitored data for 2008 and 2009 are summarised in Tables 2a and 2b. The pre-test results of FA are shown in Table 3. Tables 4 and 5 show the results of total variance explained by the varifactors and rotated factor loading on individual variables respectively. The
3919
determinant value for the variables used in this study was zero (Table 3). In FA, existence of extreme multicollinearity (many correlated variables) and singularity (perfectly correlated variables) causes problems. When there are too many variables in the original data which are correlated, it becomes difficult to determine their unique contribution to a particular factor. Past studies suggest to keep the determinant value <0.00001 to avoid such difficulties (Fields, 2005). The KMO test statistic for sampling adequacy was 0.692 and Bartlett’s test of sphericity was significant ( p < 0.05) (Table 3). The results of pre-tests indicate that correlations of the data were relatively compact to yield distinct and reliable factors and the R-matrix was not an identity matrix. Hence, FA was appropriate for the data. The input data matrix contained 16 variables for four sites and therefore the input data matrix for FA comprised of 16 variables and 104 cases for 2008 and 2009. Five principal components which had Eigenvalues >1 were retained, summarising almost 74% of the total variance in the water quality data (Table 4). The first principal component accounted for 17% of the total variance while the second and the third principal components accounted for similar variances of 16%. Altogether, first three components reflected 50% of the total variance in the entire data set. Corresponding VFs and factor loadings are described in Table 5. Only factor loading above 0.5 have been considered for interpretation as they have a moderate to high effect on the relevant factor (Zhou et al., 2007).
4.2.
Identification of water quality factors
Water, biota and river geography are the ultimate endpoints of human induced pollution. We assume that the water
Fig. 2 e A sample of plots showing the extent of correlation among the selected variables. Refer to Tables 2a and 2b for the units of the variable on X and Y-axis.
3920
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 1 5 e3 9 2 4
Table 2a e Statistical descriptive of original water quality variables (2008). Variable and code
Descriptive statistics-2008 N
Algal total count (cells/mL)-‘Algal TC’ Algal biovolume (mm3/L)-‘Algal BioVol’ Areal standard unit (Cyano)-‘ASU Cyano’ Areal standard unit (Toxic)-‘ASU-Toxic’ Areal standard unit (algae)-‘ASU Algae’ Chlorophyll-a (ug/L)-‘Chl-a’ Cyanobacteria Total Count (cells/mL)Cyano TC’ Cyanobacterial biovolume (mm3/L) Lorenzen (ug/L)-‘Loren’ Phaeophytin (ug/L)-‘Phaeo’ Toxic Cyanobacterial count (cells/mL)‘Toxic Cyano TC’ Toxic Cyanobacterial biovolume (mm3/L)‘Toxic Cyano BioVol’ Clostridium perfringens (CFU/100 mL) Coliforms Thermotolerant (cfu/100 mL)‘Coli Therm’ Coliforms Total (cfu/100 mL)-‘Coli TC’ E. coli (orgs/100 mL)-‘Ecoli’ Enterococci (cfu/100 ml)-‘Entero’ Manganese filtered (mg/L)-‘MnFilt’ Manganese total (mg/L)-‘MnTot’ Aluminium filtered (mg/L)-‘AlFilt’ Aluminium total (mg/L)-‘AlTot’ Iron filtered (mg/L)-‘FeFilt’ Iron total (mg/L)-‘FeTot’ Dissolved organic carbon (mg/L)-‘DOC’ UV absorbing constituents (organic)-‘UV’ Nitrogen ammoniacal (mg/L)-‘N-Amm’ Nitrogen oxidised (mg/L)-‘NOx’ Nitrogen TKN (mg/L)-‘NTKN’ Nitrogen total (mg/L)-‘NTN’ Phosphorus filterable (mg/L)-‘PFilt’ Phosphorus total (mg/L)-‘TP’ Silicate reactive (SiO2 mg/L)-‘Sil’ Alkalinity (mgCaCO3/L)-‘Alk’ Conductivity field (mS/cm)-‘EC’ Dissolved oxygen (%Sat)-‘DO’ Suspended solids (mg/L)-‘SS’ Temperature (Deg C)-‘Temp’ True colour at 400 nm-‘Col’ Turbidity lab/field (NTU)-‘NTU’ pH (lab/field)-‘pH’
Min
Max
Mean
STD
Statistic
Std. error
46 46 46 46 46 52 46
770 0.214 0 0 144 2.1 0
337954 7.146 3275 1248 5523 59 325699
33944.54 1.68 296.73 119.20 1552.76 15.94 27525.36
8582.82 0.21 85.76 38.80 186.05 1.73 8272.59
58211.51 1.44 581.67 263.14 1261.87 12.45 56107.44
46 52 52 46
0 0.5 0.2 0
1.812 53 9.1 35970
0.19 14.36 1.80 2385.86
0.05 1.61 0.25 948.30
0.37 11.61 1.83 6431.67
46
0
1.008
0.12
0.04
0.26
52 52
1 1
180 3000
11.31 91.96
4.17 58.58
30.05 422.41
52 52 52 52 52 52 52 52 52 52 52 52 52 52 52 52 52 52 52 52 52 52 52 52 52 52
46 1 1 0.001 0.003 0.01 0.01 0.05 0.05 2 0.08 0.01 0.01 0.1 0.3 0.005 0.01 0.1 18 0.178 51.88 1 10.2 5 1.7 6.62
98000 2400 1200 0.123 0.144 0.2 1.17 0.52 2.23 11 0.78 0.087 1.03 0.7 1.3 0.037 0.15 5.87 59 17.38 126.55 41 26.7 50 89.1 9.03
5739.37 91.48 80.31 0.02 0.06 0.02 0.26 0.13 0.62 5.06 0.19 0.02 0.33 0.28 0.61 0.01 0.02 2.69 34.02 2.06 89.12 9.58 19.60 15.65 12.44 7.46
2575.59 49.95 32.45 0.00 0.00 0.00 0.03 0.01 0.06 0.28 0.02 0.00 0.04 0.02 0.04 0.00 0.00 0.24 1.51 0.63 2.12 1.07 0.70 1.35 1.99 0.06
18572.87 360.21 234.00 0.03 0.03 0.03 0.23 0.10 0.45 1.99 0.13 0.02 0.26 0.13 0.26 0.01 0.02 1.72 10.91 4.53 15.26 7.70 5.02 9.70 14.37 0.43
N.B.: N-Number of data points, STD-Standard Deviation.
chemistry is fundamental to river health assessments as it has multiple stressors embedded into it, such as the composition and quality of catchment and atmospheric and human influence (Markich and Brown, 1998). As such, subtle changes in water quality can be identified quickly and efficiently even before they appear in a biological community. Water chemistry being the basis for river health and using the human health analogy, we attempted to understand the health of HNR system through a few key variables. For this, we employed a popular dimension reduction multivariate approach FA (Fields, 2005). This tool has previously been used for many purposes but its potential for identifying overall river condition
based on a few variables for river health assessment purposes has not been attempted in detail. The first VF, the most important of all five VFs, contained three variables with both positive and negative factor loadings (pH 0.863, NTU 0.766, DO e 0.699). The values of pH and DO variables were negatively correlated with VF1, while NTU was positively correlated. Past studies suggested that the factors containing pH and DO with negative factor loadings show the presence of anaerobic conditions in the river (Shrestha and Kazama, 2007; Vega et al., 1998). Shrestha and Kazama (2007) obtained a negative loading for pH and DO on VF2 for Fuji River Basin in Japan while Vega et al. (1998) obtained
3921
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 1 5 e3 9 2 4
Table 2b e Statistical descriptive of original water quality variables e (2009). Variable and code
Descriptive statistics 2009 N
Algal total count (cells/mL)-‘Algal TC’ Algal biovolume (mm3/L)-‘Algal BioVol’ Areal standard unit (Cyano)-‘ASU Cyano’ Areal standard unit (Toxic)-‘ASU-Toxic’ Areal standard unit (algae)-‘ASU Algae’ Chlorophyll-a (ug/L)-‘Chl-a’ Cyanobacteria total count (cells/mL)Cyano TC‘ Cyanobacterial biovolume (mm3/L) Lorenzen (ug/L)-‘Loren’ Phaeophytin (ug/L)-‘Phaeo’ Toxic Cyanobacterial count (cells/mL)‘Toxic Cyano TC’ Toxic Cyanobacterial biovolume (mm3/L)‘Toxic Cyano BioVol’ Clostridium perfringens (CFU/100 mL) Coliforms Thermotolerant (cfu/100 mL)‘Coli Therm’ Coliforms total (cfu/100 mL)-‘Coli TC’ E. coli (orgs/100 mL)-‘Ecoli’ Enterococci (cfu/100 ml)-‘Entero’ Manganese filtered (mg/L)-‘MnFilt’ Manganese total (mg/L)-‘MnTot’ Aluminium filtered (mg/L)-‘AlFilt’ Aluminium total (mg/L)-‘AlTot’ Iron Filtered (mg/L)-‘FeFilt’ Iron Total (mg/L)-‘FeTot’ Dissolved Organic Carbon (mg/L)-‘DOC’ UV absorbing constituents (organic)-‘UV’ Nitrogen ammoniacal (mg/L)-‘N-Amm’ Nitrogen oxidised (mg/L)-‘NOx’ Nitrogen TKN (mg/L)-‘NTKN’ Nitrogen total (mg/L)-‘NTN’ Phosphorus filterable (mg/L)-‘PFilt’ Phosphorus total (mg/L)-‘TP’ Silicate reactive (SiO2 mg/L)-‘Sil’ Alkalinity (mg CaCO3/L)-‘Alk’ Conductivity field (mS/cm)-‘EC’ Dissolved oxygen (%Sat)-‘DO’ Suspended solids (mg/L)e‘SS’ Temperature ( C)-‘Temp’ True colour at 400 nm-‘Col’ Turbidity lab/field (NTU)-‘NTU’ pH (lab/field)-‘pH’
Min
Max
Mean
STD
Statistic
Std. error
41 41 41 41 41 52 41
643 0.231 0 0 217 2.2 0
233506 15.07 7158 2502 10650 45.4 229567
52845.10 1.70 555.46 179.30 1619.45 14.02 47171.90
9180.42 0.36 182.66 182.66 264.95 1.44 9003.68
58783.38 2.33 1169.60 419.86 1696.50 10.41 57651.70
41 52 52 41
0 1.8 0.2 0
8.227 41.5 8.2 36222
0.48 12.58 1.75 2602.79
0.21 1.34 0.22 933.77
1.31 9.64 1.59 5979.08
41
0
2.653
0.20
0.07
0.45
52 52
1 1
340 990
15.60 48.81
6.72 20.67
48.46 149.08
52 52 52 52 52 52 52 52 52 52 52 52 52 52 52 52 52 52 52 52 52 52 52 52 52 52
70 1 1 0.001 0.011 0.01 0.01 0.05 0.05 2 0.06 0.005 0.002 0.05 0.27 0.001 0.005 0.05 7 0.016 61.9 1 11.1 4 1.23 7.12
24000 2000 1600 0.054 0.142 0.08 0.9 0.36 1.6 19 0.3 0.088 0.85 0.92 1.24 0.037 0.122 4.81 71 17.9 120.84 25 30.4 26 24.8 9.2
2230.08 68.12 81.65 0.01 0.05 0.02 0.19 0.10 0.43 4.28 0.14 0.01 0.20 0.33 0.53 0.00 0.02 1.47 40.42 1.78 90.44 8.22 21.05 10.82 9.17 7.64
491.92 38.78 32.85 0.00 0.00 0.00 0.03 0.01 0.04 0.35 0.01 0.00 0.03 0.02 0.03 0.00 0.00 0.19 1.96 0.50 1.85 0.81 0.77 0.75 0.88 0.05
3547.27 279.66 236.86 0.01 0.02 0.01 0.19 0.07 0.32 2.54 0.07 0.02 0.21 0.15 0.21 0.01 0.02 1.36 14.15 3.62 13.34 5.83 5.55 5.44 6.32 0.35
N.B.: N-Number of data points, STD-Standard Deviation.
a negative loadings for pH and DO on VF2 for Pisuerga River system in Spain. Nevertheless, they both agree that VF2 in their studies is a clear indication of organic matter entering the waterway producing ammonia and other organic acids under anaerobic fermentation, resulting in low pH and possible depletion of oxygen in water. In the present study, first factor gives an indication of anaerobic conditions in the river. These results further explain the anaerobic fermentation, once known as a common problem for many urban rivers, is now appearing in peri-urban river system. At present, these three variables are listed as the key variables to be measured in regional streams in the community water
quality monitoring networks in Australia. These networks encourage ordinary citizens to become involved and be active in the protection and management of their waterways and catchments (Waterwatch, 2004). The present analysis further confirms the importance of the three variables to obtain a quick snapshot of the river water quality. The second VF indicates the microbiological quality of the river water. It consists of Enterococci and E. coli variables with strong positive loadings (0.860 and 0.734) on the factor. The microbial species belonging to genus Enterococcus are gram positive, facultative anaerobes (ability to survive with or without the presence of Oxygen) mostly present in the human
3922
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 1 5 e3 9 2 4
Table 3 e Results of KMO and Bartlett’s tests.
Table 5 e Rotated component matrix.
KMO and Bartlett’s Test
Rotated Component Matrix
Kaiser-Meyer-Olkin Measure of Sampling Adequacy. Bartlett’s Test of Sphericity Approx. Chi-Square df Sig.
0.692 838.834 120.000 0.000
faeces with two dominant species, viz., E. faecalis and E. faecium (Murray, 1990). On the other hand, E. Coli is a gram negative bacteria found in the intestines of warm blooded animals and has been historically used as an indicator organism of faecal contamination of water. It is now accepted the ability of Enterococci spp. to act as a more stable and reliable indicator than E. coli or faecal coliforms in brackish water (Jin et al., 2004). The EU Water Framework Directive still insists using E. coli and Enterococci species as the most robust indicator species to assess bathing water (European Environment Agency (EEA), 2009) In the present study, Enterococci have a higher correlation with the VF2 while E. coli have a slightly low correlation with VF2. Two previous studies include faecal and total coliform counts in FA while analysing water quality data obtained from Suquia River (Argentina) and Fuji River (Japan) (Alberto et al., 2001; Shrestha and Kazama, 2007). The former found coliforms as important under third VF, while the latter did not obtain coliforms as an important variable for water quality variance. Peri-urban rivers are particularly prone to microbiological contamination due to activities of large scale farms operating in the riparian lands. However, microbiological contamination is less or confounded by other highly variable water quality parameters in urban river systems draining from sewered areas. Results of the present study agree with work by Alberto et al. (2001). The original data set contained five types of microbiological indicators, viz., Clostridium perfringens, Clostridium thermotolerant, Coliform total, E. coli and Enterococci spp. The multivariate analysis carried out in the present study helped to reduce the five microbiological variables to two microbiological variables that explained 16% of the total variance in peri-urban river system. This factor further indicated the importance of Enterococci spp. as a potential indicator for river health assessment in brackish water sections of this peri-urban river system. The third VF contains two chlorophyll related variables (chlorophyll-a and Phaeophytin), one algal related (algal
Table 4 e Total variance explained by the varifactors. Total Variance Explained Component
VF1 pH NTU DO DOC Entero E.coli NTKN Chla AlgalBioVol MnFilt Phaeo Temp NOx Alk EC Sil
VF2
VF3
VF4
VF5
0.863 0.766 0.699 0.860 0.734 0.843 0.824 0.642 0.536 0.867 0.820 0.878 0.692 0.504
biovolume) variable and one plant micronutrient (Manganese). This factor can be viewed as representing the eutrophication effects of the river system. Chlorophyll-a is a pigment found in all plant cells, including eukaryotic algae and prokaryotic blue-green algae cyanobacteria (Carlson, 1996). When chlorophyll is degraded due to light, it could occur through either losing the centre Magnesium ion or the phytol tail. The former produces the pigment Phaeophytin, while the latter produces a molecule called chlorophyllide (Yentsch, 1965). Eutrophication has recently become significant threat to nutrient equilibrium in peri-urban river systems due to increased agricultural activities and discharge of municipal water from treatment facilities. Based on the periurban section of Yangtze River system China, Zhang et al. (2007) reported higher levels of nitrogen and phosphorus levels in surface river waters than in surface water reported from other districts of China. On the other hand, algal bio volumes are now being used to quantitatively measure the volume of algae cells in a sample to determine the risks associated with mixed species of toxin producing algae species for recreational activities in Australia (ACT Health, 2010). Thus, the amount of Chlorophyll-a, Phaeophytin and algal bio volumes can be viewed as suitable indicators of eutrophication and nuisance algal blooms in recreational waters of peri-urban rivers. The presence of Manganese in this factor is indicative of its association with the growth of macrophytes in the river waters as a micronutrient. It does not seem related to the effects of eutrophication or algal blooms.
Rotation Sums of Squared Loadings Total
1 2 3 4 5 Extraction Method:
Factors
% of Variance
2.80 17.49 2.68 16.77 2.68 16.73 2.00 12.48 1.67 10.44 Principal Component Analysis
Cumulative % 17.49 34.27 51.00 63.47 73.91
4.3.
Key water quality variables for monitoring
During the analysis, we obtained three groups of variables explaining 50% of variance in the river quality. By measuring nine variables listed under three factors, we are able to understand 50% of the variance in the river water quality. Therefore, depending on the time and resource availability, river management authorities can first focus on variables
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 1 5 e3 9 2 4
described in the first, second and third VFs to understand the half of the variance occurring in the river system. Variables grouped under fourth and fifth VFs can then be measured to account for up to 74% of the total variance. Nevertheless, in descending order of importance, the variables identified in the first, second and third varifactors were related to both apparent and latent dimensions (i.e., anaerobic condition, microbial quality and eutrophication) of river water quality. The foregoing discussion has indicated the possibility of identifying the most appropriate water quality variables for routine monitoring and inclusion in river health assessment approaches in peri-urban situation, where the effects of urbanisation is a dominating factor. By selecting key variables for monitoring, it is possible to greatly reduce the data collection cost over the long term. The FA is also a valuable tool in isolating key variables to be used in a composite river health index. By doing this, the whole approach expects to avoid the use of reference sites, provision of equal degree of importance to each environmental component in the river system and use the view of an expert panel to assess river health. In general, the present study has indicated that river health in peri-urban landscapes is prone to higher degrees of degradation. River health is usually assessed through regular monitoring of water quality and specific river health approaches. Although water quality variables have a significant influence on river health, there are other aspects in river ecosystems that also play a role in determining the health. As a result, the latter has gained much acceptance in assessing river health compared to routine monitoring in recent years. However, there are a number of discrepancies in either method in terms of variable selection and application of river assessment to an aquatic system. For example, ecological function based river health assessment approaches suffer from a lack of human understanding on complex ecological interactions when interpreting the results while the rationale for selecting key variables remains ambiguous for other river health assessment approaches. Furthermore, we are losing sites which were once considered ‘pristine’ and as a result the accuracy and reliability of river health assessment approaches based on reference site conditions are subject to a higher degree of scrutiny. A number of restrictions of our study and areas for future research should be mentioned. We assume that the observed parameters important for majority of the variances occurring in the HNR system in this study are subjected to a certain degree of spatio-temporal variation and require a standardisation with long term data. There is also some uncertainty associated with the ecological relationships between chosen water quality parameters and biotic communities. As such, further research is warranted to unearth the biotic interactions with selected variables.
5.
Conclusions
In this study, we followed a bottom-to-top approach for understanding the key variables and factors of river health conditions assuming that river waters are fundamental to the health of an entire river system. The FA approach employed in the present study clearly identified three groups of water quality variables that explains majority of the river
3923
health conditions using a few key variables that can easily be monitored. Anaerobic fermentation, microbial pollution and eutrophication are three key environmental problems faced by peri-urban rivers. Variables in the first, second and third VFs revealed anaerobic conditions, microbial quality and effects of eutrophication in the HNR system. Thus, the present work found nine variables are sufficient to explain up to 50% of the river variance. The application of FA for identification of key variables is promising. Its major advantages are considerable cost reduction in river health monitoring and assessment programs. The FA approach can provide guidance in variable selection for river health assessment in a peri-urban context and helping in the partition of variables according to their unique contribution to the total variance.
Acknowledgements The authors acknowledge the Sydney Catchment Authority for permission to use the data in this publication. We also sincerely thank Ms. Tracey Schultz for her great support in sending us the data sets promptly.
references
ACT Health, 2010. ACT Guidelines for the Management of Blue-Green Algae in Recreational Water [Online]. Department of Health, Canberra. Available. http://www.health.act.gov.au/c/health? a¼sendfile&ft¼p&fid¼1281404492&sid¼ accessed 15.11.10. Alberto, W., Mar a Del Pilar, D., Mar a Valeria, A., Fabiana, P., ´ ngeles, B., 2001. Pattern recognition Cecilia, H., Mar a De Los A techniques for the evaluation of spatial and temporal variations in water quality. A case study: Suqu a river basin (Co´rdobaeArgentina). Water Research 35, 2881e2894. Askey-Doran, M., Hart, B., Ladson, T., Read, M., 2009. Tasmanian River Condition Index South Horbart, Tasmania. Baginska, B., Pritchard, T., Krogh, M., 2003. Roles of land use resolution and unit-area load rates in assessment of diffuse nutrient emissions. Journal of Environmental Management 69, 39e46. Barbour, M., Gerritsen, J., Snyder, B., Stribling, J., 1999. Rapid Bioassessment Protocols for Use in Streams and Wadeable Rivers: Periphyton, Benthic Macroinvertebrates and Fish, second ed. U.S. Environmental Protection Agency, Washington, D.C. (Office of Water). Barmuta, L.A., Chessman, B.C., Hart, B.T., 2002. Australian River Assessment System: Interpretation of the Outputs from AusRivAS-(Milestone Report). Land and Water Resources Research and Development Corporation. Boulton, A.J., 1999. An overview of river health assessment: philosophies, practice, problems and prognosis. Freshwater Biology 41, 469e479. Carlson, R.E.A.J.S., 1996. A Coordinator’s Guide to Volunteer Lake Monitoring Methods. Wisconsin: North American Lake Management Society, Madison. Chessman, B., 1995. Rapid assessment of rivers using macroinvertebrates: a procedure based on habitat-specific sampling, family level identification and a biotic index. Austral Ecology 20, 122e129. Clarke, K., Ainsworth, M., 1993. A method of linking multivariate community structure to environmental variables. Marine Ecology-Progress Series 92, 205e219.
3924
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 1 5 e3 9 2 4
Commonwealth of Australia, 1996. Australia. State of the Environment. CSIRO Publishing, Collingwood, Victoria. Connell, J., 1974. Ecology: field experiments in marine ecology. Experimental Marine Biology, 21e54. Dalkey, N., Helmer, O., 1963. An experimental application of the Delphi method to the use of experts. Management Science 9, 458e467. Diamond, R., 2004. Water and Sydney’s Future e Balancing the Value of Our Rivers and Economy. Final Report of the HawkesburyNepean River Management Forum. NSW: Department of Infrastructure, Planning and Natural Resources, Sydney. European Environment Agency (EEA) 2009. Annual report, 2009 bathing season. Quality of Bathing Water e 2009 Bathing Season (EEA Report, No 3/2010). Copenhagen, Denmark: Office for Official Publications of the European Union, Available: http://ec.europa.eu/environment/water/water-bathing/ report_2010.html [10 Feb 2011]. Fairweather, P.G., 1999. State of environment indicators of ’river health’: exploring the metaphor. Freshwater Biology 41, 211e220. Fields, A., 2005. Factor Analysis Using SPSS. SAGE Publications, London. Harris, J.H., 1995. The use of fish in ecological assessments. Austral Ecology 20, 65e80. Hoelzl, U., 2007. River habitat monitoring and assessment in Germany. Environmental Monitoring and Assessment 127. Howard, M., 2009. Aquatic Ecosystems Productivity ‘IS’ Reliant on Water Managers and Systainable Cities 12th International River Symposium. Brisbane, Australia, 21e24 (September). Huston, M., 1979. A general hypothesis of species diversity. American Naturalist 113, 81. International Union for Conservation of Nature and Natural Resources. IUCN Global Freshwater Initiative. IUCN Conservation Centre, Gland, Switzerland, Available: http:// www.iucn.org/about/work/programmes/species/our_work/ about_freshwater/ (access 10.02.11.). ISC, 2009. Index of Stream Condition: The Second Benchmark of Victorian River Condition. Department of Sustainability and Environment, Victoria. Jansen, A., Robertson, A., 2001. Relationships between livestock management and the ecological condition of riparian habitats along an Australian floodplain river. Journal of Applied Ecology 38, 63e75. Jin, G., Englande, A., Bradford, H., Jeng, H., 2004. Comparison of E. coli, enterococci, and fecal coliform as indicators for brackish water quality assessment. Water Environment Research 76, 245e255. Kaika, M., 2003. The water framework directive: a new directive for a changing social, political and economic European framework. European Planning Studies 11, 299e316. Karr, J., 1991. Biological integrity: a long-neglected aspect of water resource management. Ecological Applications, 66e84. Markich, S.J., Brown, P.L., 1998. Relative importance of natural and anthropogenic influences on the fresh surface water chemistry of the Hawkesbury-Nepean River, south-eastern Australia. The Science of the Total Environment 217, 201e230. Murray, B., 1990. The life and times of the Enterococcus. Clinical Microbiology Reviews 3, 46. Norris, R.H., Dyer, F., Hairsine, Kennard, Mark, Linke, Simon, Merrin, Linda, Read, Arthur, Robinson, Wayne, Ryan, Chris, Wilkinson, Scott, Williams, D., 2007. Assessment of River and Wetland Health: A Framework for Comparative Assessment of the Ecological Condition of Australian Rivers and Wetlands. National Water Commision, Canberra. Pearson, M., Norris, R.H., 1996. The effect of habitat-specific sampling on biological assessment of water quality. Freshwater Biology 36, 419e434. Pejaman, A.H., Bidhendi, N.G.R., Karbassi, A.R., Mehrdadi, N., Bidhendi, E.M., 2009. Evaluation of spatial and seasonal variation in surface water quality using multivariate statistical
techniques. International journal of Environmental Science and Technology 6, 467e476. Peter, D., Harris, J., Hillman, T., Walker, K., 2008. SRA Report 1: A Report on the Ecological Health of Rivers in the MurrayDarling Basin, 2004e2007. Canberra: Prepared by the Independent Sustainable Rivers Audit Group for the MurrayDarling Basin Ministerial Council. Pinto, U., Maheshwari, B.L., 2010. Reuse of Greywater for irrigation around homes in Australia e Understanding community views, issues and practices. Urban Water 7, 141e153. Schofield, N., Davies, P., 1996. Measuring the health of our rivers. Water (Melbourne) 23, 39e43. Shrestha, S., Kazama, F., 2007. Assessment of surface water quality using multivariate statistical techniques: a case study of the Fuji river basin, Japan. Environmental Modelling and Software 22, 464e475. Thoms, M., Parker, C., Simons, M., 2000. The Dispersal and Storage of Trace Metals in the Hawkesbury River valley. River management: The Australasian Experience. John Wiley, Chichester, UK. 97e102. Turner, L., Erskine, W., 2005. Variability in the development, persistence and breakdown of thermal, oxygen and salt stratification on regulated rivers of south-eastern Australia. River Research and Applications 21. Turner, E.R., Rabalais, N.N., 2003. Linking landscape and water quality in the Mississippi River Basin for 200 year. BioScience 53, 563e572. Underwood, A., 1994. On beyond BACI: sampling designs that might reliably detect environmental disturbances. Ecological Applications, 4e15. Vega, M., Pardo, R., Barrado, E., Deba´n, L., 1998. Assessment of seasonal and polluting effects on the quality of river water by exploratory data analysis. Water Research 32, 3581e3592. Vugteveen, P., Leuven, R., Huijbregts, M., Lenders, H., 2006. Redefinition and elaboration of river ecosystem health: perspective for river management. Hydrobiologia 565, 289e308. Walsh, C.J., Allison, H.R., Feminella, J.W., Cottingham, P.D., Groffman, P.M., Ii, R.P.M., 2005. The urban stream syndrome: current Knowledge and the Search for a Cure. Journal of the North American Benthological Society 24, 706e723. Waterwatch, 2004. Waterwatch Australia National Technical Manual. Department of the Environment and Heritage, Canberra. Whitton, B.A., Kelly, M.G., 1995. Use of algae and other plants for monitoring rivers. Austral Ecology 20, 45e56. Winter, T., Mallory, S., Allen, T., Rosenberry, D., 2000. The use of principal component analysis for interpreting ground water hydrographs. Ground Water 38, 234e246. Wright, J., 1995. Development and use of a system for predicting the macroinvertebrate fauna in flowing waters. Austral Ecology 20, 181e197. Yentsch, C., 1965. Distribution of chlorophyll and phaeophytin in the open ocean. Deep Sea Research 12, 653e666. Young, T.F., Sanzone, S., 2002. A Framework for Assessing and Reporting on Ecological Condition. US Environmental Protection Agency, EPA Science Advisory Board. Young, R., Townsend, C., Matthaei, C., 2004. Functional indicators of river ecosystem healthean interim guide for use in New Zealand. Prepared for Ministry for Environment. Cawthron Rep., 870, 54. ¨ born, I., Blomba¨ck, K., Zhang, Q., Shi, X., Huang, B., Yu, D., O Wang, H., Pagella, T., Sinclair, F., 2007. Surface water quality of factory-based and vegetable-based peri-urban areas in the Yangtze River Delta region, China. Catena 69, 57e64. Zhou, F., Liu, Y., Guo, H., 2007. Application of multivariate statistical methods to water quality assessment of the watercourses in Northwestern new territories, Hong Kong. Environmental Monitoring and Assessment 132, 1e13.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 2 5 e3 9 3 2
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Fate of Perfluorooctanesulfonate and perfluorooctanoate in drinking water treatment processes Sokichi Takagi a,*, Fumie Adachi a, Keiichi Miyano a, Yoshihiko Koizumi a, Hidetsugu Tanaka a, Isao Watanabe a, Shinsuke Tanabe b, Kurunthachalam Kannan c a
Osaka Prefectural Institute of Public Health, 1-3-69 Nakamichi, Higashinari-ku, Osaka 537-0025, Japan Center for Marine Environmental Studies (CMES), Ehime University, Bunkyo-cho 2-5, Matsuyama 790-8577, Japan c Wadsworth Center, New York State Department of Health and Department of Environmental Health Sciences, State University of New York at Albany, Empire State Plaza, P.O. Box 509, Albany, NY 12201-0509, USA b
article info
abstract
Article history:
Perfluorooctanesulfonate (PFOS) and perfluorooctanoate (PFOA) have been recognized as
Received 17 December 2010
global environmental pollutants. Although PFOS and PFOA have been detected in tap water
Received in revised form
from Japan and several other countries, very few studies have examined the fate, especially
20 April 2011
removal, of perfluorinated compounds (PFCs) in drinking water treatment processes. In
Accepted 29 April 2011
this study, we analyzed PFOS and PFOA at every stages of drinking water treatment
Available online 13 May 2011
processes in several water purification plants that employ advanced water treatment technologies. PFOS and PFOA concentrations did not vary considerably in raw water, sand
Keywords:
filtered water, settled water, and ozonated water. Sand filtration and ozonation did not
PFOS
have an effect on the removal of PFOS and PFOA in drinking water. PFOS and PFOA were
PFOA
removed effectively by activated carbon that had been used for less than one year.
Finished water
However, activated carbon that had been used for a longer period of time (>1 year) was not
Activated carbon treatment
effective in removing PFOS and PFOA from water. Variations in the removal ratios of PFOS
Tap water
and PFOA by activated carbon were found between summer and winter months. ª 2011 Elsevier Ltd. All rights reserved.
Water treatment
1.
Introduction
Perfluorinated compounds (PFCs) such as perfluorooctanesulfonate (PFOS) and perfluorooctanate (PFOA) have been recognized as global environmental pollutants. PFOS and PFOA have been found in wildlife and human blood from all over the world (Hansen et al., 2001; Giesy and Kannan, 2001; Kannan et al., 2001, 2002; 2004; Tao et al., 2006). Because PFOS and PFOA occur in tap water (Skutlarek et al., 2006; Takagi et al., 2008; Mak et al., 2009; Quinete et al., 2009), it is recognized as a source of human exposure to PFCs (Ericson et al., 2009; Holzer et al., 2009: Nolan et al., 2010). Activated
carbon treatment is used in some water purification plants in Japan. It is reported that activated carbon is effective in the removal of PFOS and PFOA from aqueous solutions (Qu et al., 2009; Yu et al., 2009; Hansen et al., 2010). The sorption isotherms of PFOS and PFOA were fitted by the Langmuir model (Qu et al., 2009; Yu et al., 2009) or the Freundlich model (Hansen et al., 2010). The affinity of activated carbon for PFOS is similar to phenol, trichloroethylene and 2-chlorophenol (Ochoa-Herrera and Sierra-Alvarez, 2008). In an earlier study, we investigated residue levels of PFOS and PFOA in several water purification plants in Osaka, Japan, and reported the occurrence of both PFOS and PFOA in raw and treated tap
* Corresponding author. Tel.: þ6 6972 1321; fax: þ81 6 6972 2393. E-mail address:
[email protected] (S. Takagi). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.04.052
3926
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 2 5 e3 9 3 2
water (Takagi et al., 2008). Comparison of concentrations of PFOS and PFOA in raw and tap water indicated that water treatment processes did not completely remove PFOS and PFOA. The removal of PFOS and PFOA in advanced water treatmentd including ozonation and activated carbon filtrationdwas found to be incomplete. Therefore, the concentrations of PFOS and PFOA present in raw water can influence the residue levels in tap water. In addition, in tap water samples from some water purification plants, the concentrations of PFOS and PFOA were higher than those found in raw water. In wastewater treatment facilities, the concentrations of PFOS and PFOA in effluent samples were higher than those in influent samples (Sinclair and Kannan, 2006; Loganathan et al., 2007). PFOA concentrations increased after biologically activated carbon filtration in water reclamation plant (Thompson et al., 2011). These results suggested that current water treatments were not effective for the removal PFOS and PFOA in drinking water. Therefore, it is important to clarify the fate, especially removal, of PFOS and PFOA in drinking water treatment for the protection of human health from PFOS and PFOA exposures. In this study, we determined concentrations of PFOS and PFOA in water samples taken at every stage in the drinking water treatment processes from several water purification plants that use advanced water treatment technologies in Osaka, Japan. We investigated the behavior and fate of PFOS and PFOA and the removal efficiencies of these compounds in water treatment processes.
2.
Materials and methods
2.1.
Reagents and chemicals
Sodium perfluoro-1-octanesulfonate (purity >99%), 13C4sodium perfluoro-1-octanesulfonate (>99%) (13C4-PFOS) and 13 C4-perfluoro-n-octanoic acid (>99%) (13C4-PFOA) were purchased from Wellington Laboratories Inc (Guelph, ON, Canada) and PFOA (purity >95%) was purchased from Wako Pure Chemical Industries, Ltd (Osaka, Japan). 13C8-PFOA (>98%) (13C8-PFOA) was purchased from Cambridge Isotope Laboratories, Inc (Andover, MA, USA). Sep-pak plus PS-2 cartridges (Waters, Tokyo, Japan) were used in the solid phase extraction (SPE) of water samples. GF/B (47 mmF, Whatman, Tokyo, Japan) was used as a glass fiber filter (GF).
2.2.
Sample collection
Water samples were collected in July 2007 and February 2008 from five water purification plants (denoted as ‘A’ to ‘E’) in Osaka, Japan. At plants A, B, and C, we collected raw water, sand filtered water, ozonated water, activated carbon filtered water, and finished water. At plant D, we collected raw water, settled water, ozonated water, activated carbon filtered water, and finished water. At plant E, raw water, sand filtered water, activated carbon filtered water, and finished water were collected. Because there were several activated carbon filters at all plants, we collected samples from multiple activated carbon treatment steps. Details regarding sources of raw water and the treatment methods employed in each of the five plants are shown in Fig. 1.
In addition, at plant E, we collected sand filtered water and activated carbon filtered water every two weeks after new activated carbon was installed in summer in order to investigate the behavior of PFOS and PFOA following installation of new activated carbon treatment filters. We targeted two activated carbon filters installed within the plant E for this investigation.
2.3.
Solid phase extraction
Sodium sulfite was added into ozonated water samples and residual ozone was removed. If sand filtered water, settled water, and finished water samples had residual chlorine, sodium sulfite was added to these samples. For the extraction, a sample of 1 L of water was used. For raw water samples, GF was used to filter the suspended solids. Each sample was spiked with 25 ng of 13C4-PFOS and 25 ng of 13 C4-PFOA as internal standards. The solid phase extraction cartridge was conditioned by the passage of 10 mL of acetone, 10 mL of methanol, and 10 mL of purified water. Each water sample was then loaded onto the SPE cartridge and passed through it at 20 mL/min. Nitrogen was passed through the cartridge for 15 min, at 2 mL/min, to dehydrate the cartridge. The cartridge was then eluted with 2 mL of methanol. The GF was transferred into a centrifugation tube. Ten mL of methanol was added and extracted by ultrasonication for 15 min. The extract from the GF was concentrated with a rotary evaporator. The eluate from the SPE cartridge and the extract from the GF were combined. The extract was then concentrated to 1 mL under a gentle stream of nitrogen. Particles in the extract were filtered by use of a cellulose filter. The eluate was spiked with 25 ng of 13C8-PFOA and injected into a high performance liquid chromatograph interfaced with a tandem mass spectrometer (LC-MS/MS).
2.4.
Analysis
Analyte separation was performed using an Agilent 1100 series HPLC system. Five mL of the extracts were injected onto a 50 2.1 mm (5 mm) Ascentis RP-Amide column (Supelco; Bellefonte, PA, USA) with a 10-mM ammonium acetate/ acetonitrile mobile phase starting at 20% acetonitrile. At a flow rate of 200 mL/min, the gradient was increased to 80% acetonitrile at 10 min, and held for 8 min. The column temperature was maintained at 40 C. For the quantitative determination, the HPLC system was interfaced to an Applied Biosystems API 3000 tandem mass spectrometer operated in electrospray negative ionization mode. Primary and product ions (SRM transitions) monitored for PFOS, PFOA, 13C4-PFOS, 13C4-PFOA, and 13C8-PFOA were 499 > 80, 413 > 369, 503 > 80, 417 > 372, and 421 > 376, respectively. Recoveries of 13C4-PFOS and 13C4PFOA ranged from 67 to 97% and from 50 to 97%, respectively. By injecting the lowest concentration of the standard mixture seven times, we found that the limit of quantification (LOQ) for PFOS and PFOA was 0.10 ng/L. However, when procedural blanks were analyzed with every set of samples, PFOS and PFOA were detected at concentrations ranging from 0.09 to 0.27 ng/L and from 0.10 to 0.34 ng/L, respectively. Therefore, we calculated the LOQ using the standard deviation of these
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 2 5 e3 9 3 2
Fig. 1 e Schematic diagram of the drinking water treatment process. The stars represent sampling points.
3927
3928
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 2 5 e3 9 3 2
procedural blanks and the LOQ was set at 0.50 ng/L for PFOS and at 0.70 ng/L for PFOA. Total organic carbon (TOC) was analyzed by hightemperature combustion method using TOC analyzer, TOCV CSH (Shimadzu, Kyoto, Japan). All samples were analyzed three times and we report the average of three analysis for every samples.
3.
Results and discussion
3.1. Behavior of PFOS and PFOA in drinking water treatment PFOS and PFOA were found in all raw water samples at concentrations ranging from 1.2 to 4.4 ng/L and from 10 to 42 ng/L, respectively (Table 1). Concentrations of PFOA were higher than those of PFOS. When a nationwide survey of PFCs in surface water sample in Japan was carried out, the thirdhighest PFOS concentration and highest PFOA concentration were found in the Kansai region including Osaka (Lai et al., 2009). Concentrations of PFOS and PFOA found in surface water from the Kansai region were <0.25e13 and 0.77e46 ng/ L, respectively. Our results were similar to this previous report. Therefore, PFOS and PFOA concentrations in raw water from Osaka were higher than those from other areas in Japan. PFOS and PFOA were found in all sand filtered water and settled water samples at concentrations ranging from 1.2 to 3.5 ng/L and from 12 to 34 ng/L, respectively (Table 1). PFOS and PFOA were found in all ozonated water samples at concentrations ranging from 1.1 to 3.4 ng/L and from 11 to 27 ng/L, respectively (Table 1). PFOS and PFOA concentrations in raw water, sand filtered water, settled water, and ozonated water were similar within each treatment plant. Therefore, sand filtration and ozonation did not have an effect on the removal of PFOS and PFOA. Because PFOS and PFOA are stable compounds, ozonation had no effect on PFOS and PFOA. Similar result was found in a water reclamation plant in South East Queensland, Australia (Thompson et al., 2011). PFOS and PFOA can be formed from some precursor compounds such as N-EtFOSE alcohol and 8:2 FTOH (Lange, 2002; Kannan et al., 2005). These precursor compounds did not appear to degrade to PFOS and PFOA by ozonation, although further studies are needed in this regard. PFOS was detected in 17 activated carbon filtered water samples collected in summer and 18 samples collected in winter. The concentration range of PFOS in activated carbon filtered water samples was from 0.51 to 7.6 ng/L (Table 1). PFOA was detected in 23 activated carbon filtered water samples collected in summer and 22 samples collected in winter. The concentration range of PFOA in activated carbon filtered water samples was from 0.78 to 72 ng/L (Table 1). The concentrations of PFOS and PFOA in activated carbon filtered water samples varied depending on the duration of use of the activated carbon. Concentrations of PFOS and PFOA in activated carbon filtered water were lower than those in sand filtered water or ozonated water. Because activated carbon had been renewed once or twice a year at plant E, PFOS and PFOA were removed effectively throughout the year at this plant. Activated carbon in two activated carbon filters, AC1 at plant A and AC4 at plant
B, were renewed between the summer and winter months of the sampling period. As a result of this renewal, PFOS and PFOA were removed effectively through these activated carbon filters in winter. Therefore, activated carbon treatment was effective in the removal of PFOS and PFOA. However, the use of activated carbon over one year was not effective in removing PFOS and PFOA. Concentrations of PFOS and PFOA in activated carbon filtered water samples were similar to or higher than those in ozonated water samples. This aspect is described in detail later. Because activated carbon at water treatment plants is commonly used for several years in Osaka, it is clear that the current water treatment processes do not completely remove PFOS and PFOA. PFOS and PFOA were found in all finished water samples at concentrations ranging from 1.3 to 3.7 ng/L and from 6.5 to 48 ng/L, respectively (Table 1). PFOS and PFOA concentrations in finished water samples were much lower than in raw water samples at plant E. However, at other plants, PFOS and PFOA concentrations in finished water samples were similar to or higher than those in raw water samples. Therefore, the removal efficiency of PFOS and PFOA by the water treatment processes is low. This result was consistent with what we found in our previous study (Takagi et al., 2008). PFOS concentration range in tap water was reported to be from 0.066 to 4.9 in Japan (Harada et al., 2003; Mak et al., 2009). PFOS concentration found in this study was similar to values reported earlier. PFOS concentration in tap water from Shanghai, Shenzhen, Macau, Hong Kong and Taipei, China were from 5.4 to 7.6 ng/L (Mak et al., 2009) and these concentrations were higher than those in Osaka, Japan. On the other hand, PFOA concentration range in tap water was reported to be from 0.18 to 5.4 ng/L in Japan excluding Osaka (Harada et al., 2003; Mak et al., 2009). Therefore, PFOA concentration detected in finished water from Osaka was higher than that from the other areas in Japan. PFOA concentration in Shanghai, China, was 78 ng/L (Mak et al., 2009) and which is higher than that in Osaka, Japan. Because several activated carbon filtered water are mixed and chlorinated to produce final finished water, the concentrations of PFOS and PFOA in finished water were higher than those in raw water. The increase of PFOS and PFOA concentrations in activated carbon filtered water samples was considerable in summer. Therefore, concentrations of PFOS and PFOA in finished water in summer were higher than those in winter.
3.2. Removal efficiency of PFOS and PFOA by fresh activated carbon At plant E, PFOS and PFOA were removed effectively by activated carbon treatment. We analyzed PFOS and PFOA in sand filtered water and activated carbon filtered water samples every two weeks for eight months since the activated carbon filter was renewed. The target activated carbon filters were AC2 and AC3. We compared the concentrations of PFOS and PFOA in sand filtered water samples with those in activated carbon filtered water samples and investigated the removal efficiency of PFOS and PFOA by activated carbon treatment. In the sand filtered water samples, the concentration range of PFOS was from 2.3 to 3.9 ng/L and the concentration range of PFOA was from 25 to 44 ng/L (Table 2). The concentration of
3929
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 2 5 e3 9 3 2
Table 1 e Concentrations (ng/L) of PFOS and PFOA in water samples collected at various stages of drinking water treatment processes. Water Purification Plant
Sample
Summer a
Time
(Day)
Concentration TOC (mg/L)
A
Raw water (river water) Sand filtered water Ozonated water Activated carbon filtered water
B
C
D
E
AC1 AC2 AC3 AC4 AC5 AC6 AC7 AC8
Finished water Raw water (river water) Sand filtered water Ozonated water Activated carbon filtered AC1 water AC2 AC3 AC4 AC5 AC6 Finished water Raw water (river water) Sand filtered water Ozonated water Activated carbon filtered AC1 water AC2 AC3 AC4 AC5 AC6 Finished water Raw water (river water) Settled water Ozonated water Activated carbon filtered AC1 water AC2 AC3 AC4 AC5 Finished water Raw water (lake water) Sand filtered water Activated carbon filtered AC1 water AC2 AC3 Finished water
e e e e 1611 924 1328 183 1258 992 e e e e e e e 524 1012 1226 1651 1255 159 e e e e 356 266 721 1618 1602 e e e e e 896 165 1975 502 e e e e 4 180 264 e
1.4 0.9 0.8 0.8 0.7 0.6 0.7 0.2 0.7 0.6 e e 0.5 1.4 0.6 0.7 0.6 0.4 0.5 0.5 0.5 0.5 0.2 0.4 1.3 0.6 0.6 0.4 0.4 0.5 0.5 0.5 e 0.5 1.3 0.8 0.8 0.6 0.2 0.6 0.5 e 0.5 2.9 1.7 <0.1 0.6 1.1 0.45
Winter Removal Ratio Time
PFOS PFOA PFOS (ng/L) 1.3 1.7 1.7 1.6 7.6 2.1 6.3 <0.50 5.7 4.5 e e 3.7 1.6 1.6 1.5 1.4 <0.50 1.7 4.1 4.9 4.1 <0.50 2.3 1.2 1.2 1.1 <0.50 <0.50 1.6 3.7 3.3 e 1.6 1.4 1.2 1.3 4.1 <0.50 2.7 0.58 e 2.2 4.4 3.5 <0.50 0.7 <0.50 <0.50
PFOA (%)
15 20 20 20 42 72 46 0.78 54 63 e e 48 33 21 20 21 24 65 70 49 63 <0.70 42 10 12 11 14 6.4 44 15 14 e 22 26 25 23 51 2.2 24 43 e 36 42 34 <0.70 2.2 8.1 6.5
e e e e 347 24 271 100 235 181 e e e e e e e 100 13 173 250 193 100 e e e e 100 100 45 236 200 e e e e e 280 100 108 55 e e e e 100 80 100 e
Concentration TOC
(Day) (mg/L) e e e e 110 260 130 96 170 215 e e e e e e e 20 225 250 133 200 100 e e e e 27 42 300 0 27 e e e e e 122 90 4.3 87 e e e e 100 94 76 e
e e e e 4 1106 1510 e 1440 e 321 1174 e e e e e 706 1194 1408 68 1437 341 e e e e 558 468 923 e 1804 34 e e e e 1078 347 e 684 1412 e e e 186 168 180 e
1.5 1 1 1 0.1 0.7 0.7 e 0.7 e 0.5 0.7 0.6 1.5 1 0.9 0.9 0.9 0.8 0.9 0.5 1 0.8 0.6 1.4 0.9 0.9 0.7 0.7 0.8 e 0.2 0.8 0.7 1.4 1.1 1.3 0.6 0.7 0.8 0.9 0.8 2.7 1.5 0.8 0.8 0.9 0.6
Removal Ratio
PFOS PFOA PFOS (ng/L) 3.3 3.1 3.1 2.8 <0.50 1.2 2.6 e 2.8 e <0.50 1.8 1.3 3.3 3 3.1 2.9 1.3 2.5 3.1 <0.50 2.8 0.51 1.7 2.8 2.6 2.7 1.9 1.5 2.7 e 3.1 <0.50 1.9 3.3 3.3 3.4 2.5 0.74 e 1.7 3.2 2 4.1 2.9 <0.50 <0.50 <0.50 <0.50
24 26 27 26 <0.70 41 29 e 30 e 8.0 33 24 26 26 27 27 34 35 33 <0.70 33 17 25 19 20 21 25 22 28 e 22 <0.70 20 26 26 27 34 22 e 35 33 31 42 33 10 5.1 7.9 9.2
PFOA (%)
e e e e 100 61 16 e 10 e 100 36 e e e e e 58 19 0 100 3.4 82 e e e e 30 44 0 e 15 100 e e e e 26 78 e 50 6 e
e e e e 100 52 7.4 e 11 e 70 27 e e e e e 26 30 22 100 22 37 e e e e 19 5 33 e 5 100 e e e e 26 19 e 30 22 e
100 100 100
69 84 76
a period of activated carbon use.
PFOS was below LOQ in all activated carbon filtered water samples. PFOA was detected in 19 activated carbon filtered water samples. In activated carbon filtered water samples, the concentration range of PFOA was 0.82e13 ng/L (Table 2). PFOS was removed by activated carbon treatment during the study period. On the other hand, the measured concentrations of
PFOA in activated carbon filtered water samples increased gradually with time. The removal ratio (R) of PFOA by activated carbon treatment was calculated as: Rð%Þ ¼ ðCSF CAC Þ=CSF 100
(1)
3930
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 2 5 e3 9 3 2
Table 2 e Temporal changes in the concentrations of PFOS and PFOA in drinking water treatment Plant E. Date
7/27/07 8/8/07 8/22/07 9/5/07 9/19/07 10/3/07 10/17/07 10/30/07 11/14/07 11/29/07 12/12/07 12/26/07 1/8/08 1/22/08 2/5/08 2/19/08 3/05/08 3/12/08 3/16/08
Time (Day)a
PFOS (ng/L)
PFOA (ng/L)
AC2
AC3
Sand filtered water
Activated carbon filtered water(AC2)
Activated carbon filtered water(AC3)
Sand filtered water
Activated carbon filtered water(AC2)
Activated carbon filtered water(AC3)
e 2 16 30 44 58 72 85 100 115 128 142 155 169 183 197 212 219 223
1 13 27 41 55 69 83 96 111 126 139 153 166 180 194 208 223 230 e
2.8 3.2 3.4 3.6 3.9 3.7 2.5 3.6 3.7 3.5 3.4 3.3 3.3 2.9 2.6 3.1 2.8 2.3 2.9
e <0.50 <0.50 <0.50 <0.50 <0.50 <0.50 <0.50 <0.50 <0.50 <0.50 <0.50 <0.50 <0.50 <0.50 <0.50 <0.50 <0.50 <0.50
<0.50 <0.50 <0.50 <0.50 <0.50 <0.50 <0.50 <0.50 <0.50 <0.50 <0.50 <0.50 <0.50 <0.50 <0.50 <0.50 <0.50 <0.50 e
31 28 33 31 36 33 25 32 32 31 31 31 31 33 29 34 35 35 44
e <0.70 <0.70 <0.70 <0.70 <0.70 <0.70 <0.70 <0.70 <0.70 1.4 1.7 3.7 5.0 7.1 8.6 11 12 13
<0.70 <0.70 <0.70 <0.70 <0.70 <0.70 1.5 <0.70 <0.70 0.82 1.7 3.7 5.2 7.8 9.6 11 13 13 e
a period of activated carbon use.
3.3. Removal efficiency of PFOS and PFOA by activated carbon in current use
the removal efficiency of PFOS and PFOA by activated carbon treatment. Then the removal ratio (R) of PFOS and PFOA by activated carbon treatment was calculated as: Rð%Þ ¼ ðC0 C=C0 Þ 100
100 90 80 70 60 50 40 30
AC2
20
AC3
10 0
Because the concentrations of PFOS and PFOA increased in activated carbon filters that have been in use for several years, we investigated the characteristics of activated carbon in removal of PFCs when used for several years. We evaluated
(2)
where C0 and C are concentrations of PFOS and PFOA in pretreated water (ozonated water and sand filtered water) and activated carbon filtered water, respectively. When the detected concentration was below LOQ, the concentration was assigned a value of zero for this calculation. The removal ratios of PFOS and PFOA were from 347 to 100% and from 300 to 100%, respectively (Table 1). The results in this study were obtained from different water purification plants which had different conditions for activated carbon treatment. However, the behavior of PFOS and PFOA at activated carbon treatment differed in summer and winter. Fig. 3 shows the average of C/C0 for 4 groups which were
Removal ratio (%)
where CAC and CSF are concentrations of PFOA in activated carbon filtered water and sand filtered water, respectively. When the detected concentration was below LOQ, the concentration was assigned a value of zero for this calculation. PFOA was detected after approximately 120 days since the activated carbon filter was renewed. Finally, the removal ratio decreased by approximately 65% (Fig. 2). PFOS and PFOA were removed effectively by use of fresh activated carbon. The activated carbon in AC2 was made up of coal and that in AC3 was a coconut shell-derived carbon. The removal efficiencies of two kinds of activated carbon for PFOS and PFOA were similar (Fig. 2). The sorption isotherms of PFOS and PFOA on the granular activated carbon were fitted by the Langmuir model (Yu et al., 2009; Hansen et al., 2010). The calculated maximum sorption capacity of the granular activated carbon for PFOA was 1.55 108 (Yu et al., 2009) and 1100 ng/g (Hansen et al., 2010). We calculated the sorption capacity of the activated carbon at the breakthrough, where the removal ratio was less than 100%. The average flow rate, the mass of the activated carbon and the average PFOA concentration in sand filtered water samples at plant E were 1500000 L/day, 12000000 g and 32 ng/L, respectively. PFOA was not detected in finished water samples until about 130 days after the activated carbon was renewed. Therefore, the sorption capacity of the activated carbon at plant E was calculated as 520 ng/g. This value is similar to that reported earlier (Hansen et al., 2010).
0
50
100 150 Time (day)
200
250
Fig. 2 e Correlation of removal ratio of perfluorochemicals and time in Plant E.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 2 5 e3 9 3 2
classified by the period of activated carbon use (error bars are standard deviation). In summer, the effluent concentrations of PFOS and PFOA were greater than the influent concentrations at many activated carbon filters. Therefore, the removal ratios were negative at these activated carbon filters. The reasons for the increase in concentrations of PFOS and PFOA after passage through the activated carbon filter were not clear. It appears that adsorption of PFOS and PFOA to activated carbon is a reversible reaction. Poorly adsorbed compounds such as chloroform are removed from activated carbon (Rook, 1976). 1,4-dioxane, which is a hydrophilic compound like PFOS and PFOA, begins to be desorbed from activated carbon after the carbon is saturated (McGuire et al., 1978). Adsorption of chemical compounds is affected by water temperature and adsorption is strong at low temperatures (Mollah and Robinson, 1996; Chern and Wu, 2001). Because water temperature is about 25 C in summer and 8 C in winter, PFOS and PFOA are less retained in activated carbon in summer than in winter.
3.4.
Risk assessment of PFOS and PFOA
PFOS and PFOA provisional health advisory levels reported by the United States Environmental Protection Agency (USEPA) are 200 and 400 ng/L, respectively (USEPA, 2009). Concentrations of PFOS and PFOA in all finished water samples were lower than these values. It seems that the detected concentrations of PFOS and PFOA in finished water have no immediate effect on human health. However, when activated carbon was used for several years, the concentrations of PFOS and PFOA in effluent water can increase because of
3931
breakthrough from activated carbon. It is necessary to renew the activated carbon 2 or 3 times per year. However, because the activated carbon is expensive, development of cheaper water treatment technologies for the removal of PFCs may be required. Therefore, it is important to monitor PFOS and PFOA in finished water regularly and to take measures against the discharge of PFOS and PFOA into drinking water sources. Further investigation into the behavior of PFOS and PFOA in activated carbon treatment is required.
4.
Conclusions
PFOS and PFOA were detected in water samples from water purification plants. Concentrations of PFOS and PFOA in water samples collected following sand filtration, and ozonation samples were similar to those in raw water samples. Therefore, sand filtration, sedimentation, and ozonation had no effect on the removal of PFOS and PFOA. PFOS and PFOA were removed by fresh activated carbon effectively. However, PFOS and PFOA were detected in activated carbon filtered water samples when activated carbon was used for a year. Because activated carbon is normally used at water treatment plants for several years in Osaka, it is clear that the current water treatment processes do not completely remove PFOS and PFOA. Therefore, it is important to monitor PFOS and PFOA in finished water regularly and to take measures against the discharge of PFOS and PFOA into drinking water sources.
Acknowledgments 4.5 4.0
PFOS (Summer)
3.5
PFOS (Winter)
This study was supported by the Osaka Prefectural Government. We appreciate the water utilities’ cooperation in the collection of raw and treated water samples.
C/C0 (-)
3.0 2.5
references
2.0
1.5 1.0 0.5 0.0
0 ~ 500
501 ~ 1000 1001 ~ 1500 Time (day)
1501 ~ 2000
4.5 4.0
PFOA (Summer)
3.5
PFOA (Winter)
C/C0 (-)
3.0 2.5
2.0 1.5 1.0
0.5 0.0 0 ~ 500
501 ~ 1000 1001 ~ 1500 Time (day)
1501 ~ 2000
Fig. 3 e Difference in the removal of perfluorochemicals by activated carbon treatment between summer and winter.
Chern, J.-M., Wu, C.-Y., 2001. Desorption of dye from activated carbon beds: effects of temperature, pH, and alcohol. Water Res. 35, 4159e4165. Ericson, I., Domingo, J.L., Nadal, M., Bigas, E., Llebaria, X., van Bavel, B., Lindstrom, G., 2009. Levels of perfluorinated chemicals in municipal drinking water from Catalonia, Spain: public Health Implications. Arch. Environ. Contam. Toxicol. 57, 631e638. Giesy, J.P., Kannan, K., 2001. Global distribution of perfluorooctane sulfornate and related perfluorinated compounds in wildlife. Environ. Sci. Technol. 35, 1339e1342. Hansen, K.J., Clemen, L.A., Ellefson, M.E., Johnson, H.O., 2001. Compound-specific, quantitative characterization of organic fluorochemicals in biological matrices. Environ. Sci. Technol. 35, 766e769. Hansen, M.C., Børresen, M.H., Schlabach, M., Cornelissen, G., 2010. Sorption of perfluorinated compounds from contaminated water to activated carbon. J. Soils Sediments 10, 179e185. Harada, K., Saito, N., Sasaki, K., Inoue, K., Koizumi, A., 2003. Perfluorooctane sulfonate contamination of drinking water in the Tama River, Japan: Estimated effects on resident serum levels. Bull. Environ. Contam. Toxicol. 71, 31e36.
3932
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 2 5 e3 9 3 2
Holzer, J., Goen, T., Rauchfuss, K., Kraft, M., Angerer, J., Kleeschulte, P., Wilhelm, M., 2009. One-year follow-up of perfluorinated compounds in plasma of German residents from Arnsberg formerly exposed to PFOA-contaminated drinking water. Int. J. Hyg. Environ. Health 212, 499e504. Kannan, K., Yun, S.H., Evans, T.J., 2005. Chlorinated, brominated and perfluorinated contaminates in livers of polar bears from Alaska. Environ. Sci. Technol. 39, 9057e9063. Kannan, K., Corsolini, S., Falandysz, J., Fillman, G., Senthil Kumar, K. , Loganathan, B.G., Mohd, M.A., Olivero, J., Van Wouwe, N., Yamg, J.H., Aldous, K.M., 2004. Perfluorooctanesulfonate and related fluorochemicals in human blood from several countries. Environ. Sci. Technol. 38, 4489e4495. Kannan, K., Corsolini, S., Falandysz, J., Oehme, G., Focardi, S., Giesy, J.P., 2002. Perfluorooctane sulfonate and related fluorochemicals in marine mammals, fish and birds from coasts of the Baltic and the Mediterranean Seas. Environ. Sci. Technol. 36, 3210e3216. Kannan, K., Hansen, S.P., Franson, C.J., Bowerman, W.W., Hansen, K.J., Jones, P.D., Giesy, J.P., 2001. Perfluorooctane sulfonate in fish-eating water birds including bald eagles and albatrosses. Environ. Sci. Technol. 35, 3065e3070. Lai, F.Y., Yeung, L.W.Y., Taniyasu, S., Li, P., Horii, Y., Kannan, K., Lam, P.K.S., Yamashita, N., 2009. A nationwide survey of perfluorinated compounds in surface water samples from 47 prefectures in Japan. Organohalogen Compounds 71, 2627e2632. Lange, C.C., 2002. Biodegradation Screen Study for Telomer-type Alcohols. U.S. EPA Public Docket AR226-1149. 3M Environmental Laboratory, St. Paul, MN. Loganathan, B.G., Sajwan, K.S., Sinclair, E., Senthil Kumar, K., Kannan, K., 2007. Perfluoroalkyl sulfonate and perfluorocarboxylates in two wastewater treatment facilities in Kentucky and Georgia. Water Res. 41, 4611e4620. Mak, Y.L., Taniyasu, S., Yeung, L.W.Y., Lu, G., Jin, L., Yang, Y., Lam, P.K.S., Kannan, K., Yamashita, N., 2009. Perfluorinated compounds in tap water from china and several other countries. Environ. Sci. Technol. 43, 4824e4829. McGuire, M.J., Suffet, I.H., Radziul, J.V., 1978. Assessment of unit processed for the removal of trace organic compounds from drinking water. Jour. AWWA 70, 565e572. Mollah, A.H., Robinson, C.W., 1996. Pentachlorophenol adsorption and desorption characteristics of granular activated carbon-1. Isotherms. Water Res. 30, 2901e2906.
Nolan, L.A., Nolan, J.M., Shofer, F.S., Rodway, N.V., Emmett, E.A., 2010. Congenital anomalies, labor/delivery complications, maternal risk factors and their relationship with perfluorooctanoic acid (PFOA)-contaminated public drinking water. Reprod. Toxicol. 29, 147e155. Ochoa-Herrera, V., Sierra-Alvarez, R., 2008. Removal of perfluorinated surfactants by sorption onto granular activated carbon, zeolite and sludge. Chemosphere 72, 1588e1593. 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, 146e152. Quinete, N., Wu, Q., Zhang, T., Yun, S.H., Moreira, I., Kannan, K., 2009. Specific profiles of perfluorinated compounds in drinking and surface waters, mussels, fish and dolphins from southeastern Brazil. Chemosphere 77, 863e869. Rook, J.J., 1976. Haloforms in Drinkingwater. Jour. AWWA 68, 168e172. Sinclair, E., Kannan, K., 2006. Mass loading and fate of Perfluoroalkyl surfactants in wastewater treatment plants. Environ. Sci. Technol. 40, 1408e1414. Skutlarek, D., Exner, M., Farber, H., 2006. Perfluorinated surfactants in surface and drinking water. Environ. Sci. Pollut. Res. 13, 299e307. Takagi, S., Adachi, F., Miyano, K., Koizumi, Y., Tanaka, H., Mimura, M., Watanabe, I., Tanabe, S., Kannan, K., 2008. Perfluorooctanesulfonate and perfluorooctanoate in raw and treated tap water from Osaka, Japan. Chemosphere 72, 1409e1412. Tao, L., Kannan, K., Kajiwara, N., Fillmann, G., Takahashi, S., Tanabe, S., 2006. Perfluorooctanesulfonate and related fluorochemicals in albatrosses, elephant seals, penguins and polar skuas from the Southern Ocean. Environ. Sci. Technol. 40, 7642e7648. Thompson, J., Eaglesham, G., Reungoat, J., Poussade, Y., Bartkow, M., Lawrence, M., Mueller, J.F., 2011. Removal of PFOS, PFOA and other Perfluoroalkyl acids at water reclamation plants in South East Queensland Australia. Chemosphere 82, 9e17. U.S.EPA, 2009. Provisional health Advisories for perfluorooctanoic acid (PFOA) and perfluorooctane sulfonate (PFOS). Yu, Q., Zhang, R., Deng, S., Huang, J., Yu, G., 2009. Sorption of perfluorooctane sulfonate and perfluorooctanoate on activated carbons and resin: Kinetic and isotherm study. Water Res. 43, 1150e1158.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 3 3 e3 9 4 0
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Characterisation of the impact of coagulation and anaerobic bio-treatment on the removal of chromophores from molasses wastewater Linhua Fan*, Thang Nguyen, Felicity A. Roddick School of Civil, Environmental and Chemical Engineering, RMIT University, 124 La Trobe Street, Melbourne, Victoria 3001, Australia
article info
abstract
Article history:
The performance of a coagulation sequence using aluminium chlorohydrate (ACH) and
Received 27 October 2010
a low MW polydiallyldimethylammonium chloride (polyDADMAC), and ferric chloride, for
Received in revised form
decolourising a high-strength industrial molasses wastewater was compared at bench
28 April 2011
scale. At their optimum dosages, ACH/polyDADMAC gave higher colour removal than
Accepted 30 April 2011
FeCl3 (45% cf. 28%), whereas COD reduction was similar (w30%), indicating preferential
Available online 10 May 2011
removal of melanoidins (a major contributor to the colour) by ACH/polyDADMAC. Size exclusion chromatography and fluorescence excitationeemission matrix spectrometry
Keywords:
suggested that chromophoric Feeorganic complexes were formed during FeCl3 treatment
Coagulation
of the molasses wastewater, which appeared to compromise decolourisation efficiency.
Decolourisation
Anaerobic bio-treatment of the wastewater enhanced the coagulation efficiency markedly,
Melanoidins
with FeCl3 achieving 94% colour and 96% COD removal, while ACH/polyDADMAC gave
Characterisation
70% and 56% removal, respectively. The improved decolourisation was attributed to the
Molasses wastewater
decrease in low MW organics (<500 Da) and biopolymers by the biological treatment, leading to reduced competition with melanoidins for interaction with coagulant/flocculant. For both the wastewater and the biologically treated wastewater, ACH/polyDADMAC treatment gave flocs with markedly better settling properties compared with FeCl3. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
Molasses, a sugar-rich by-product from sugar production processes, is commonly used as a raw material in the fermentation industries. Molasses wastewater from industrial fermentation processes is typically high in colour and organic load, consequently this limits its direct discharge to the aquatic environment, and even to wastewater treatment plants. The origin of the colour in the molasses wastewater is primarily associated with the dark brown melanoidin pigments which are the products of the non-enzymatic reaction between sugars and amino acids, peptides or proteins
(Maillard Reaction) (Kort, 1979). Melanoidins are generally regarded to be heterogeneous, high molecular weight (MW), negatively charged, acidic and highly dispersed polymers with similar chemical properties to humic substances (i.e., humic and fulvic acids) (Migo et al., 1993). However, the detailed structure and characteristics of the melanoidins are not yet fully understood, thus hindering the development of effective processes for their removal (Satyawali and Balkrishnan, 2008). Many methods have been investigated for reducing colour and COD of melanoidin-containing wastewaters, including activated carbon adsorption (Mall and Kumar, 1997); coagulation and flocculation (Migo et al., 1993); oxidation using ozone
* Corresponding author. Tel.: þ61 3 9925 3692; fax: þ61 3 9925 3746. E-mail address:
[email protected] (L. Fan). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.04.050
3934
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 3 3 e3 9 4 0
(Kim et al., 1985; Pen˜a et al., 2003), UV/H2O2 or UV/H2O2/Fe (II) (Catalkaya and Sengu¨l, 2006; Dwyer et al., 2008); bioremediation using white-rot fungi (Kahraman and Yesilada, 2003) or acetogenic bacteria (Sirianuntapiboon et al., 2003), electro izares et al., 2009) and membrane filtration oxidation (Can (Mutlu et al., 2002). Of these, coagulation/flocculation is regarded as a simple and cost-effective means of decolourising molasses wastewater (Zhou et al., 2008; Liang et al., 2009a). A range of conventional inorganic coagulants has been tested for removing colour and organics from melanoidincontaining wastewaters. Ferric chloride was reported to perform better than other coagulants including alum, aluminium chloride, ferric sulphate and polyferric sulphate (PFS) in the treatment of a biologically treated molasses wastewater, with up to 89% and 98% reduction for COD and colour, respectively (Liang et al., 2009a,b). However, poorer settleability of the flocs for the ferric chloride-treated wastewater was observed, and this combined with the corrosive/ highly acidic nature of the ferric chloride remained a critical issue limiting its industrial application. Polyelectrolytes are used in water and wastewater treatment both for coagulation and as coagulant/flocculant aids to strengthen flocs and improve their settleability through charge patching and particle bridging. Cationic polyacrylamide was reported to markedly enhance floc settleability of a molasses wastewater after primary coagulation using ferric chloride, whereas anionic polyacrylamide had little effect (Liang et al., 2009a). However, polyacrylamides did little to improve colour and organic removal from the wastewater. Nevertheless, the use of other polyelectrolyte types, such as high charge density cationic polyDADMACs for molasses wastewater treatment has not been documented. The aims of this study were to investigate the effectiveness of sequential ACH/polyDADMAC treatment of a highly coloured molasses process wastewater, and to evaluate the treatability of the wastewater by coagulation before and after anaerobic biological pre-treatment. Coagulation tests with ferric chloride were conducted in parallel for comparison. Size exclusion chromatography (SEC) analyses using liquid chromatography with organic carbon detection (LCeOCD) and UV absorbance detection (LCeUVD), and fluorescence excitationeemission matrix (EEM) spectrometry, were employed in order to gain a better insight into the impact of the coagulation and bio-treatment on the removal of chromophores from the molasses wastewaters.
2.
Materials and methods
2.1.
Wastewater samples
Wastewater samples containing molasses (of cane origin) were collected from an industrial fermentation plant in Victoria. Two types of wastewaters were generated from the production process: the first pass (i.e., the very high-strength process supernatant) and second pass wastewater (i.e., product washings). The wastewater tested in this study was the mixture of the two wastewaters at a ratio of 40% first pass to 60% second pass, which was the typical composition of the wastewater discharged to sewer as Trade Waste. Part of the
Table 1 e Characteristics of WW and ATW. Parameter pH True colour (mg PteCo L1) DOC (mg L1) COD (mg L1) UVA254 (cm1, 1/1000 dilution)
WW
ATW
5.6e6.0 30,000e35,500 8200e12,000 18,000e30,000 0.1e0.15
7.7e8.5 31,000e36,000 3200e3900 6000e14,000 0.13e0.16
Note: Data are given in ranges based on the measurements of 5 samples collected over a period of 6 months.
molasses wastewater sample was subjected to anaerobic treatment using a lab-scale unit in a commercial laboratory, and the treated effluent was supplied for the coagulation tests. The characteristics of the wastewater (WW) and the anaerobically treated wastewater (ATW) are summarised in Table 1.
2.2.
Coagulants/flocculants and test procedures
ACH (aluminium chlorohydrate), sold as Megapac 23 (40% w/w), was supplied by Omega Chemicals. Ferric chloride (FeCl3) was supplied by BDH Pty Ltd. PolyDADMACs were supplied as Magnafloc LT410 (MW w100 kDa), LT510 (MW 200e350 kDa), LT610 (MW 400e500 kDa), and Zetag 7120 (MW w1000 kDa) by Ciba Specialty Chemicals P/L. Coagulation trials were conducted using 2 L wastewater for each run with a laboratory jar tester (Phipps and Bird, PB-700) at the solution temperature of 22 C with rapid mixing for 2 min at 250 rpm followed by slow mixing for 20 min at 30 rpm. The coagulated water was settled for 2 h and the supernatant was then analysed. For sequential ACH/polyDADMAC treatment, ACH was added to the wastewater and mixed at 250 rpm for 2 min. PolyDADMAC was then added to the solution and mixed at 250 rpm for a further minute, followed by slow mixing at 30 rpm for 20 min. Doses were as noted in the text. ACH dosages are reported in terms of pure ACH. All coagulation tests and bulk analyses were conducted in duplicate on two wastewater samples collected on different dates. The overall trends in colour and COD reduction for the two wastewater samples were in good agreement. As the variation of the measured colour and COD values was generally less than 5% for each wastewater sample, only one set of colour and COD reduction data and the associated SEC and fluorescence EEMs data were presented. The settling properties of the wastewater and anaerobically treated wastewater treated with FeCl3 or ACH/LT410 were characterised by measuring the residual turbidity and floc settling rate. The settling rate was determined by transferring the coagulated wastewater sample to a 1 L measuring cylinder, and the volume of the sludge layer (floc) was recorded at 30, 90, 150 and 210 min. The sludge production for the coagulation treatments was estimated using a sludge mass balance (SMB) model. The SMB model is based on the possible sources of suspended solids contributing to the total sludge produced during the wastewater coagulation processes (Cornwell, 1999), and can be expressed by Sludge production kg ML1 ¼ TSS þ X þ Fe þ Mn þ Al þ D þ C
3935
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 3 3 e3 9 4 0
2.3.
Analytical methods
The UV absorbance at 254 nm was determined using a double beam scanning spectrophotometer (Unicam UV2, 1 cm pathlength). The true colour of the samples was measured in PteCo units at 455 nm using a Hach DR/4000U spectrophotometer. Dissolved organic carbon (DOC) was measured with a total organic carbon analyser (Sievers 820). The pH was measured with a Hach Sension 156 pH/conductivity meter. Turbidity was determined using a Hach turbidity meter (DR/4000 UVevis spectrophotometer). The COD was measured using Hach method 8000 and a Hach DR/4000U spectrophotometer. Metal ion contents in the untreated and treated wastewaters were determined using ICP-MS (HP 4500 series 300). Total solids (TS) and total dissolved solids (TDS) were determined according to the standard methods (APHA, 1998). The molecular size distribution of dissolved organics in selected raw and treated wastewater samples was determined by SEC using LCeOCD and LCeUVD techniques at the Water Research Centre of the University of New South Wales. The organic molecules were separated according to their molecular size (i.e., 100e200,000 Da) using the SEC system (LCeOCD Model 8, DOC-Labor Dr. Huber, Germany) which consists of a SEC column (Toyopearl TSK HW-50S, diameter 2 cm, length 25 cm). The SEC system was calibrated for molar masses of humics using International Humic Substance Society (IHSS) humic acid and fulvic acid (MW 300e20,000 g mol1), and for total organic carbon using potassium hydrogen phthalate. The LCeOCD chromatograms were processed using the Labview-based program FIFFIKUS (DOC-Laboratory Dr. Huber, Germany). Fluorescence EEM spectra of the raw and treated water samples were determined using a PerkinElmer Fluorescence spectrometer (LS55). All water samples were adjusted to pH 7 prior to the EEM analysis. EEM spectra were obtained for ACH/polyDADMAC and FeCl3 solutions, and indicated that the coagulants made negligible contributions to fluorescence intensities. Samples were diluted to DOC levels of approximately 1e5 mg L1 and 1e10 mg L1 prior to SEC and EEM analyses, respectively. The dilution factor was the same for all samples subjected to SEC or EEM analysis.
3.
Results and discussion
3.1. Effect of molecular weight characteristics of polyDADMAC on colour removal In order to select a polyDADMAC coagulant with the best colour removal efficiency, coagulation tests were conducted using four commercially available polyDADMAC products with various molecular weight specifications (100e1000 kDa) at the original pH of the molasses wastewater (Fig. 1). The low
40 LT410 (MW ~100k)
35
LT510 (MW 200-350k) LT610 (MW 400-500k)
Colour reductional (%)
where, TSS ¼ Total suspended solids in raw wastewater (mg L1), X ¼ Dosage of other chemicals (e.g., polymers) (mg L1), Fe ¼ Iron removal (mg L1), Mn ¼ Manganese removal (mg L1), Al ¼ Aluminium removal (mg L1), D ¼ DOC removal (mg L1), C ¼ Conversion of coagulant dosage to product precipitation (e.g., Fe(OH)3 for FeCl3; Al(OH)3 for alum) (mg L1).
30
Zetag 7120 (MW ~1,000k)
25 20 15 10 5 0 0.5
0.7
0.9
1.1
1.3
1.5
1.7
1.9
2.1
-1
Dosage (g L )
Fig. 1 e Effect of MW of polyDADMAC on colour removal for WW (initial pH 5.7).
MW polyDADMAC (LT410) gave the highest colour removal (w35%) at the dosage of 1.2 g L1. However, the colour removal decreased greatly with increased dosage of LT410. Medium (LT510) and high MW polyDADMAC (LT610) gave similar colour removal (up to 24e27%) across the dosage range tested, whereas the very high MW polyDADMAC (Zetag 7120) was the least effective (up to 17%). Charge patching is regarded as a predominant factor in determining the efficiency for removing the negatively charged pigments from molasses wastewater using polymeric coagulant/flocculent (Liang et al., 2009a). According to the supplier of the polyDADMACs (Ciba Specialty Chemicals P/L, 2010), their products are all linear polyDADMAC homopolymers and they all have the same charge density, but differ primarily in average molecular weight and active content. The better decolourisation performance of the lower MW LT410 was hence attributed to its higher active content and lower average MW compared with the other polyDADMAC products, meaning that LT410 has more molecules per unit mass of the coagulant and can interact with more of the chromophoric molecules in the molasses wastewater. LT410 was therefore chosen as the polymeric coagulant for the subsequent coagulation treatment studies.
3.2. Colour and COD removal performance of ACH/LT410 and FeCl3 Treatment using the sequence of ACH and LT410 was trialled for enhancing the colour and COD removal (Fig. 2a). The optimum dosage of 1.2 g L1 for LT410 (Fig. 1) was used as a fixed parameter. The best colour (45%) and COD removal (29%) was obtained at the ACH dosage of 0.96e1.2 g L1, which gave 10% greater colour removal than with LT410 alone. The reduction in COD correlated well with that for colour although it was lower than for colour (by 15e18%) at all dosages trialled. This indicated that the ACH/LT410 sequence treatment gave preferential removal of the coloured organic components from the wastewater. Only a slight drop in pH (from 5.7 to 5.5) was observed after coagulation at the optimum dosages of 0.96 g L1 ACH/1.2 g L1 LT410.
3936
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 3 3 e3 9 4 0
a
b
Fig. 2 e Colour and COD removal by (a) ACH/LT410 and (b) FeCl3 treatment of WW (initial pH 5.7).
To provide a comparison with ACH/LT410 treatment the wastewater was coagulated with FeCl3. The best colour (28%) and COD removal (31%) was achieved at the FeCl3 dosage of 5 g L1 (Fig. 2b). At 1.5 g L1 FeCl3 the colour of the treated wastewater increased markedly, although the COD level was reduced considerably. Both the colour and COD reduction decreased at dosages greater than 5 g L1. Over the range of 3e7 g L1, the levels of colour and COD reduction were similar, which may indicate the low selectivity of FeCl3 in removing chromophoric organic molecules from the molasses wastewater. The pH was recorded as 2.7 after treatment with the 5 g L1 FeCl3, indicating the need for substantial adjustment of pH before discharge or some further treatments.
ACH/LT410 for the anaerobically treated wastewater were determined as 0.96 g L1/1.2 g L1 (data not shown), the same as for the wastewater, and the colour and COD removals were 70% (remaining colour: 9600 mg PteCo L1) and 56% (remaining COD: 5280 mg L1), respectively. The pH of the solutions after FeCl3 and ACH/LT410 treatment was 5.1 and 6.7, respectively. Although there was no apparent colour reduction, anaerobic treatment of the wastewater reduced the DOC and COD levels by more than 60% (Table 1) and this facilitated the colour removal by coagulation. The improved coagulation performance was attributed to the physico-chemical changes in the organic components of the wastewater after the biological treatment as discussed in Sections 3.4 and 3.5.
3.3. Effect of biological treatment on coagulation and decolourisation
3.4.
The anaerobically treated molasses wastewater was subjected to coagulation using FeCl3 and ACH/LT410. The optimum dosage of FeCl3 was 3 g L1, at which 94% of colour (remaining colour: 1920 mg PteCo L1) and 96% of COD (remaining COD: 480 mg L1) were removed (Fig. 3). The optimum dosages of 100 90
Removal efficiency (%)
80 70 60 50 40 30
Colour
COD
20 10 0 0
1
2
3
4
5
-1
6
7
FeCl3 dosage (g L )
Fig. 3 e Colour and COD removal by FeCl3 treatment of ATW (initial pH 8.3).
8
Size exclusion chromatography (SEC)
SEC with organic carbon (OC) and UVA254 detection has been widely used for characterising natural organic matter such as humic substances in drinking water and wastewater (Her et al., 2002; Laabs et al., 2006; Haberkamp et al., 2007). As melanoidins and humic substances have many structural similarities (Migo et al., 1993), the chromatograms can be interpreted within a similar context. When OC detection is used (Fig. 4a), the first peak (retention time 29 min) is related to the so-called biopolymers which consist of high MW organic compounds such as polysaccharides, proteins, and organic colloids (>20,000 Da). The humics/melanoidins (500e20,000 Da) peak follows (at 45 min) and then the peak for building blocks (350e500 Da). Building blocks are the breakdown products of humics/melanoidins. The next peak (at 59 min) contains low MW organic acids. After the low MW acids peak, neutral and amphiphilic compounds such as mono- or oligosaccharides, alcohols, aldehydes, ketones (<350 Da) may be present (Her et al., 2002; Haberkamp et al., 2007). It should be noted that the retention time of the peaks may vary depending on the specifications of the SEC columns employed and the nature of the organics in the water samples. In chromatograms generated with UVA detection (Fig. 4b), biopolymers are not often detectable because they have a low proportion of bonds which absorb at 254 nm.
3937
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 3 3 e3 9 4 0
a 4.5
b
4
UVA signal response (AU)
0.3
3.5
OC signal response (AU)
0.35
WW
3
WW-ACH-LT410
2.5
WW-FeCl3
2 1.5
WW 0.25 WW-ACH-LT410 0.2
WW-FeCl3
0.15
0.1
1 0.05 0.5 0
0 20
30
40
50
60
70
80
90
20
100
30
40
50
60
70
80
90
100
Retention time (min)
Retention time (min)
Fig. 4 e SEC for WW before and after coagulation using ACH (0.96 g LL1)/LT410 (1.2 g LL1) or FeCl3 (5 g LL1): (a) LCeOCD; (b) LCeUVD.
Coagulation of the wastewater using ACH (0.96 g L1)/LT410 (1.2 g L1) or FeCl3 (5 g L1) had only a small effect on the molecular size distribution of the dissolved organics (Fig. 4a). ACH/LT410 reduced the peak sizes for the higher MW compounds (i.e., biopolymers, melanoidins) only marginally, whereas FeCl3 gave a slightly greater reduction of these compounds (Fig. 4a). Some small increases for the smaller compounds (350e500 Da) were observed after FeCl3 treatment, this was attributed to the change in the structure of the melanoidins due to depolymerisation with progressive charge neutralisation (Manka and Rebhum, 1982), and the oxidation of some large molecules to smaller molecules due to the mildly oxidising nature of FeCl3. It is generally accepted that the conjugated bonds and aromatic rings of organic molecules absorb UV light and thus can contribute to colour, and a good correlation between UVA254 and PteCo colour was observed for the molasses wastewater used in this work (R2 ¼ 0.984). The LCeUVD chromatogram was therefore used as a tool to monitor the change in character of the coloured organic molecules
a
(Fig. 4b). ACH/LT410 led to some removal of melanoidin molecules (500e20,000 Da) and higher MW compounds (>20,000 Da), whereas FeCl3 gave greater reductions for the melanoidins and low MW molecules (<350 Da). However, two peaks appeared at 32 and 40 min after FeCl3 treatment of the wastewater, indicating the formation of some higher MW colour-causing compounds, most likely Feeorganic complexes since the anionic organic matter in molasses wastewater including melanoidins, amino acids, proteins and sugars can form stable complexes with metal cations such as Fe3þ in acidic medium (Migo et al., 1997). This appeared to compromise the efficacy of colour removal, and it is therefore considered as a major cause of the marked increase in colour of the wastewater after FeCl3 treatment at 1.5 g L1 (Fig. 2b). The low MW compounds were not greatly impacted by the coagulation processes (Fig. 4a,b). To identify the influence of biological treatment on the character of the organic molecules, SEC chromatograms for the molasses wastewater before and after anaerobic biotreatment are compared in Fig. 5a and b. The biopolymer
b
4
3.5
WW
0.35
0.3 WW
3
UVA signal response (AU)
DOC signal response (AU)
ATW ATW-ACH-LT410 ATW-FeCl3
2.5
2 1.5
0.25
ATW ATW-ACH-LT410
0.2
ATW-FeCl3
0.15
0.1
1 0.05
0.5
0
0 20
30
40
50
60
70
Retention time (min)
80
90
100
20
30
40
50
60
70
80
90
Retention time (min)
Fig. 5 e SEC for WW and ATW before and after coagulation using ACH (0.96 g LL1)/LT410 (1.2 g LL1) or FeCl3 (3 g LL1): (a) LCeOCD; (b) LCeUVD.
100
3938
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 3 3 e3 9 4 0
peak (at 30 min) disappeared after anaerobic treatment of the wastewater. The anaerobic treatment also markedly reduced the low MW organic matter content (<500 Da) of the wastewater due to biodegradation. However, the increased OC and UVA signal responses at 35e40 min suggest the formation of some medium to high MW compounds which was most probably due to the re-polymerisation of the nonbiodegradable melanoidins during the biological treatment (Pen˜a et al., 2003). Some major changes in the apparent molecular weight distribution were observed after coagulation of the anaerobically treated wastewater (Fig. 5a,b). FeCl3 (3 g L1) gave much greater removal of the higher MW compounds than ACH (0.96 g L1)/LT410 (1.2 g L1), but both coagulation treatments had no apparent effect on the low MW organic components. A new peak (at 57 min) appeared in the OC chromatogram for FeCl3, indicating some low MW organic molecules were generated during the coagulation process, consistent with the previous suggestion of the depolymerisation of melanoidins and oxidation of larger organic molecules by FeCl3. The improved decolourisation after biological treatment was attributed to the changes in the physico-chemical characteristics of the organic components, e.g., the removal of low MW organic matter (<500 Da) and biopolymers which facilitated colour removal by reducing the competition with melanoidin molecules for interaction with coagulant/flocculant. The two new peaks at 32 and 40 min on the LCeUVD chromatogram for the FeCl3-treated wastewater (Fig. 4b) were not observed for the biologically treated wastewater after FeCl3 treatment (Fig. 5b), indicating that some of the biodegradable organics in the wastewater were prone to forming complexes with Fe3þ.
3.5. Fluorescence excitationeemission matrix (EEM) spectra EEM spectra are a highly sensitive method widely used for ‘fingerprinting’ the organic components such as humic substances present in natural waters and treated wastewater (Chen et al., 2003). It is only very recently that EEM spectra were employed as a tool for examining melanoidins in wastewater (Dwyer et al., 2009). According to Chen et al. (2003), EEM spectra can be divided into five regions, as shown in Fig. 6a for the raw molasses wastewater. Regions I and II correspond to aromatic proteins and region III is associated with fulvic acid-like substances. Regions IV and V are associated with soluble microbial products (SMPs, e.g., proteins and polysaccharide-like materials) and humic acidlike materials (e.g., melanoidins (Dwyer et al., 2009)), respectively. The EEM spectra indicate that the raw molasses wastewater contained significant amounts of the five organic components described above. Coagulation with ACH/LT410 led to reductions in the intensity of fluorescence in all five regions (Fig. 6b), whereas FeCl3 (5 g L1) reduced the fluorescence only marginally (Fig. 6c). This was in accord with the coagulation results (Fig. 2a,b) which demonstrated that ACH/ LT410 removed more coloured molecules from the wastewater compared with FeCl3. The fluorescence intensity for all five regions increased markedly when a lower dose of FeCl3 (1.5 g L1) was used (Fig. 6d), which was consistent with the increase in colour (Fig. 2b) and the SEC analysis suggesting the formation of chromophoric organic compounds/complexes. Anaerobic treatment of the wastewater greatly reduced the peak intensities corresponding to aromatic proteins and soluble microbial products, but not for the fulvic acid-like
Fig. 6 e EEMs of (a) WW (DOC 5.2 mg LL1), (b) WW after ACH/LT410 (DOC 3.9 mg LL1), (c) WW after 5 g LL1 FeCl3 (DOC 4.0 mg LL1), (d) WW after 1.5 g LL1 FeCl3 (DOC 4.8 mg LL1), (e) ATW (DOC 2.5 mg LL1), (f) ATW after ACH/LT410 (DOC 1.4 mg LL1) and (g) ATW after 3 g LL1 FeCl3 (DOC 1.1 mg LL1). Regions (I) and (II): aromatic proteins, (III): fulvic acid-like materials, (IV): SMPs and (V): humic acid-like materials (melanoidins).
3939
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 3 3 e3 9 4 0
Table 2 e Sludge production and residual turbidity for WW and ATW. Treatment (wastewater type)
Estimated sludge production (kg ML1)
Measured sludge production (kg ML1)
Turbidity (NTU) at 60 min
Turbidity (NTU) at 210 min
6518 4837 4171 3841
6415 4483 4574 4143
4510 1200 728 100
4320 1110 627 87
FeCl3 5 g L1 (WW) ACH 0.96 g L1/LT410 1.2 g L1 (WW) FeCl3 3 g L1 (ATW) ACH 0.96 g L1/LT410 1.2 g L1 (ATW)
and humic acid-like melanoidins (Fig. 6e). The peak intensities for the melanoidins (Region V) were increased slightly after the bio-treatment (Fig. 6e), which was related to the re-polymerisation of the non-biodegradable melanoidins as indicated by the SEC results (Fig. 5a,b). There were considerable reductions in peak intensities for the melanoidins by ACH/LT410 treatment (Fig. 6f), whereas much greater reductions were achieved by FeCl3 (Fig. 6g). These results suggest the removal of the biodegradable organic components such as proteinaceous materials, and biopolymers such as polysaccharides, from the wastewater facilitated the effective removal of melanoidins by coagulation. The poorer performance of FeCl3 in treating the raw molasses wastewater was therefore linked to the presence of these compounds, which apparently formed chromophoric organic compounds/ complexes with FeCl3.
3.6.
Sludge production and floc settling properties
The amount of sludge produced from the coagulation treatments was estimated using a sludge mass balance (SMB) model (Cornwell, 1999), and was also measured as dry mass (Table 2). The estimated sludge production from treating wastewater and anaerobically treated wastewater with ACH (0.96 g L1)/LT410 (1.2 g L1) is less than with FeCl3 (35% and 9% less, respectively). The estimations are fairly consistent with the measured values, which show that ACH/LT410 gave 43% and 10% less sludge than FeCl3 for wastewater and anaerobically treated wastewater, respectively. The measured
1000 900
Volume of sludge layer (mL)
FeCl3 5 g/L (WW)
800
ACH/LT410 (WW)
700
FeCl3 3 g/L (ATW) ACH/LT410 (ATW)
600 500 400 300 200 100 0 0
25
50
75
100
125
150
175
200
225
Settling time (min)
Fig. 7 e Plot of sludge volume versus settling time for the coagulated WW and ATW.
sludge production for FeCl3 and ACH/LT410 for anaerobically treated wastewater is higher than the corresponding estimated values, whereas the measured values for wastewater are lower than the estimated values. The different trends are likely related to the different physico-chemical properties of the organic components after biological treatment. Wastewater treated with FeCl3 had a significantly higher residual turbidity compared with that treated with ACH/ LT410 (Table 2). A similar trend was observed for the anaerobically treated wastewater, indicating the poorer settleability of the FeCl3 flocs. It is clear that anaerobic treatment of the wastewater markedly improves the settleability of the flocs for both FeCl3 and ACH/LT410 treatment, with residual turbidity after settling for 60 min reduced by 84% and 92%, respectively. It was also observed that the ACH/LT410 flocs settled faster than the FeCl3 flocs for both the wastewater and the anaerobically treated wastewater (Fig. 7). It should be noted for each individual coagulant the apparent volume of the sludge layer for ATW was greater than for WW due to the better settleability of the flocs after biological treatment as indicated by the lower residual turbidity (Table 2). For the wastewater, for ACH/LT410 a steady state was reached after 90 min, whereas for FeCl3 (5 g L1) a steady state was not reached until 210 min. For the anaerobically treated wastewater, ACH/LT410 flocs reached a steady state after 90 min, markedly less than for FeCl3.
4.
Conclusions
The coagulation performance of a treatment sequence using ACH and a low MW polyDADMAC (LT410), and FeCl3, on a high-strength molasses wastewater with and without anaerobic bio-treatment was evaluated. For the wastewater (no anaerobic bio-treatment), ACH/LT410 gave higher colour removal than FeCl3 (45% cf. 28%), whereas COD removal was similar for the two coagulant types (w30%), indicating preferential/selective removal of coloured organic molecules by ACH/polyDADMAC. SEC and EEM data suggested that chromophoric Feeorganic complexes were formed during FeCl3 treatment of the wastewater which may lead to the reduction in decolourisation efficiency. Anaerobic pre-treatment of the molasses wastewater improved the coagulation efficiency of the wastewater markedly. FeCl3 achieved 94% colour and 96% COD removal, whereas ACH/polyDADMAC gave 70% and 56%, respectively. The improved decolourisation was associated with changes in the characteristics of the organics, as indicated by SEC and EEM data which showed a significant decrease in low MW molecules (<500 Da) and biopolymers after biological
3940
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 3 3 e3 9 4 0
treatment of the wastewater, which facilitated colour removal by reducing the competition with melanoidins for interaction with coagulant/flocculant. There was only a small change in pH values for the ACH/ LT410 treated water, whereas FeCl3 lowered the solution pH significantly. Based on the current pricing of the chemicals (industrial grade) and the optimum dosages observed, the cost of chemicals for ACH/LT410 was about 35% lower compared with FeCl3 for treating the wastewater (i.e., $4.5 cf. $7 kL1); but the cost of both treatments was similar for treating the bio-treated wastewater (i.e., $4.5 cf. $4.2 kL1). The cost of FeCl3 treatment would be greater when pH adjustment of the treated wastewaters is taken into account. In terms of sludge production and floc settling properties, ACH/LT410 performed significantly better than FeCl3. These factors would make ACH/LT410 coagulation a more suitable treatment of the nonbiologically treated molasses wastewater prior to further polishing treatments such as ozonation. In this work, it was also demonstrated that SEC with LCeOCD and UVD, and fluorescence EEMs, are valid analytical tools for characterising the removal of the chromophores from molasses wastewaters. Although the chemical coagulants used in this work are generally regarded as of low toxicity and are commonly used in potable water treatment, their dosages were significantly higher for treating the high-strength molasses wastewater compared with drinking water treatment. As the higher dosages can potentially lead to greater residuals of these chemicals in the treated wastewater, it is recommended to conduct ecotoxicity tests on the coagulated wastewaters in future research. Moreover, it would also be necessary to evaluate the disinfection by-product (DBP) formation potentials of the coagulated wastewaters if an oxidation process were used as a subsequent polishing treatment.
Acknowledgements The authors would like to acknowledge Dr Zhenren Liang for helpful discussion. This work was financially supported by the Smart Water Fund, Victoria, Australia (Project No. 62M-2025).
references
APHA, 1998. Standard Methods for the Examination of Water and Wastewater, twentieth ed. American Public Health Association, Washington, DC. izares, P., Hernandez, M., Rodrigo, M.A., Saez, C., Barrera, C.E., Can Roa, G., 2009. Electrooxidation of brown-colored molasses wastewater: effect of the electrolyte salt on the process efficiency. Ind. Eng. Chem. Res. 48, 1298e1301. C¸atalkaya, E.C ¸ ., Sengu¨l, F., 2006. Application of BoxeWilson experimental design method for the photodegradation of bakery’s yeast industry with UV/H2O2 and UV/H2O2/Fe(II) process. J. Hazard. Mater. 128, 201e207. Chen, W., Westerhoff, P., Leenheer, J.A., Booksh, K., 2003. Fluorescence excitationeemission matrix regional integration to quantify spectra for dissolved organic matter. Environ. Sci. Technol. 37, 5701e5710.
Ciba Specialty Chemicals P/L, 2010. Personal communication from Anderson J.B., Senior Chemist. Cornwell, D.A., 1999. Water treatment plant residuals management. In: Letterman, R.D. (Ed.), Water Quality and Treatment (A Handbook of Community Water Supplies), fifth ed. McGraw-Hill, New York. Dwyer, J., Kavanagh, L., Lant, P., 2008. The degradation of dissolved organic nitrogen associated with melanoidin using a UV/H2O2 AOP. Chemosphere 71, 1745e1753. Dwyer, J., Griffiths, P., Lant, P., 2009. Simultaneous colour and DON removal from sewage treatment plant effluent: alum coagulation of melanoidin. Water Res. 43, 553e561. Haberkamp, J., Ruhl, A., Ernst, M., Jekel, M., 2007. Impact of coagulation and adsorption of DOC fractions of secondary effluent and resulting fouling behaviour in ultrafiltration. Water Res. 41, 3794e3802. Her, N., Amy, G., Foss, D., Cho, J., Yoon, Y., Kosenka, P., 2002. Optimization of method for detecting and characterizing NOM by HPLC e size exclusion chromatography with UV and on-line DOC detection. Environ. Sci. Technol. 36, 1069e1076. Kahraman, S., Yesilada, O., 2003. Decolorization and bioremediation of molasses wastewater by white-rot fungi in a semi-solid-state condition. Folia Microbiol. 48, 525e528. Kim, S.B., Hayase, F., Kato, H., 1985. Decolorization and degradation products of melanoidins on ozonolysis. Agric. Biol. Chem. 49, 785e792. Kort, M.J., 1979. Colour in sugar industry. In: Birch, G.G., Parker, K.J. (Eds.), Sugar: Science and Technology. Applied Science Publishers Ltd, London. Laabs, C.N., Amy, G.L., Jekel, M., 2006. Understanding the size and character of fouling-causing substances from effluent organic matter (EfOM) in low-pressure membrane filtration. Environ. Sci. Technol. 40, 4495e4499. Liang, Z., Wang, Y., Zhou, Y., Wu, Z., 2009a. Variables affecting melanoidins removal from molasses wastewater by coagulation/flocculation. Sep. Purif. Technol. 68, 382e389. Liang, Z., Wang, Y., Zhou, Y., Liu, H., Wu, Z., 2009b. Hydrolysis and coagulation behavior of polyferric sulfate and ferric sulphate. Water Sci. Technol. 59, 1129e1135. Mall, I.D., Kumar, V., 1997. Removal of organic matter from distillery effluents using low cost adsorbent. Chem. Eng. World 32, 89e96. Manka, J., Rebhum, M., 1982. Organic groups and molecular weight distribution in tertiary effluents and renovated waters. Water Res. 16, 399e403. Migo, V.P., Matsumura, M., Rosario, E.J.D., Kataoka, H., 1993. Decolorization of molasses wastewater using an inorganic flocculant. J. Ferment. Bioeng. 75, 438e442. Migo, V.P., Del Rosario, E.J., Matsumura, M., 1997. Flocculation of melanoidins induced by inorganic ions. J. Ferment. Bioeng. 83, 287e291. Mutlu, S.H., Yetis, U., Gurkan, T., Yilmaz, L., 2002. Decolorization of wastewater of a baker’s yeast plant by membrane processes. Water Res. 36, 609e616. Pen˜a, M., Coca, M., Gonza´lez, G., Rioja, R., Garcı´a, M.T., 2003. Chemical oxidation of wastewater from molasses fermentation with ozone. Chemosphere 51, 893e900. Satyawali, Y., Balkrishnan, M., 2008. Wastewater treatment in molasses-based alcohol distilleries for COD and color removal: a review. J. Environ. Manage. 86, 481e497. Sirianuntapiboon, S., Phothilangka, P., Ohmomo, S., 2003. Decolorization of molasses wastewater by a strain NO.BP103 of acetogenic bacteria. Bioresour. Technol. 92, 31e39. Zhou, Y., Liang, Z., Wang, Y., 2008. Decolorization and COD removal of secondary yeast wastewater effluents by coagulation using aluminium sulphate. Desalination 55, 301e311.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 4 1 e3 9 5 0
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Root features related to plant growth and nutrient removal of 35 wetland plants Wen-Ling Lai a,b,1, Shu-Qiang Wang a,1, Chang-Lian Peng a, Zhang-He Chen a,* a
Key Laboratory of Ecology and Environmental Science of Guangdong Higher Education, and Guangdong Provincial Key Lab of Biotechnology for Plant Development, College of Life Science, South China Normal University, 510631 Guangzhou, PR China b Chemistry and Life Science College, Gannan Normal University, 341000 Ganzhou, Jiangxi, PR China
article info
abstract
Article history:
Morphological, structural, and eco-physiological features of roots, nutrient removal, and
Received 7 January 2011
correlation between the indices were comparatively studied for 35 emergent wetland plants
Received in revised form
in small-scale wetlands for further investigation into the hypothesis of two types of wetland
22 April 2011
plant roots (Chen et al., 2004). Significant differences in root morphological, structural, and
Accepted 2 May 2011
eco-physiological features were found among the 35 species. They were divided into two
Available online 11 May 2011
types: fibrous-root plants and thick-root plants. The fibrous-root plants had most or all roots of diameter (D) 1 mm. Roots of D > 1 mm also had many fine and long lateral roots of
Keywords:
D 1 mm. The roots of these plants were long and had a thin epidermis and a low degree of
Wetland plants
lignification. The roots of the thick-root plants were almost all thicker than 1 mm, and
Root features
generally had no further fine lateral roots. The roots were short, smooth, and fleshy, and had
Radial oxygen loss
a thick epidermis. Root porosity of the fibrous-root plants was higher than that of the thick-
Photosynthesis
root plants ( p ¼ 0.001). The aerenchyma of the fibrous-root plants was composed of large
Nutrient removal
cavities which were formed from many small cavities, and distributed radially between the exodermis and vascular tissues. The aerenchyma of the thick-root plants had a large number of small cavities which were distributed in the mediopellis. The fibrous-root plants had a significantly larger root biomass of D 1 mm, of 1 mm < D < 3 mm, above-ground biomass, total biomass, and longer root system, but shorter root longevity than those of the thick-root plants ( p ¼ 0.003, 0.018, 0.020, 0.032, 0.042, 0.001). The fibrous-root plants also had significantly higher radial oxygen loss (ROL), root activity, photosynthetic rate, transpiration rate, and removal rates of total nitrogen and total phosphorus than the thick-root plants ( p ¼ 0.001, 0.008, 0.010, 0.004, 0.020, 0.002). The results indicate that significantly different root morphological and structural features existed among different wetland plants, and these features had a close relationship to nutrient removal capacity. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
As the major component of a constructed wetland, wetland plants play an important role in nutrient removal (Kivaisi, 2001; Matheson et al., 2002; Yang et al., 2007). Research has indicated
that several processes in constructed wetland nutrient removal, such as nutrient absorption and settlement of suspended substances, are strongly linked to the functional characteristics of wetland plants, especially in regards to the important functions of the roots (Tanner, 2001; Kadlec et al.,
* Corresponding author. Tel.: þ86 20 85212758, þ86 15915819878; fax: þ86 20 85215535. E-mail address:
[email protected] (Z.-H. Chen). 1 Authors of equal contribution. 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.05.002
3942
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 4 1 e3 9 5 0
2005; Stein and Hook, 2005). However, investigation is still incomplete about the function of wetland plant roots in constructed wetland nutrient removal (Stottmeister et al., 2003; Gutknecht et al., 2006). The rhizoplane and rhizosphere of wetland plant roots are the most important areas, where the plants, microorganisms, substrates of the wetland and wastewater contact directly, and physicochemical and biological processes take place (Stottmeister et al., 2003; Greenway, 2007; Weber et al., 2008). As an important component of a constructed wetland, plant roots create favorable habitats for microorganism attachment, growth, and decomposition activities (Sinha et al., 2007; Muench et al., 2007). This function of the roots may have a close relationship to root morphology, structure, growth, and distribution. Previous studies indicated that constructed wetland nutrient removal correlated with plant root numbers, and root surface area (Kyambadde et al., 2004; Cheng et al., 2009). Different wetland plants exhibited great differences in radial oxygen loss (ROL), the phenomenon of wetland plant roots releasing oxygen through the aerenchyma to the rhizosphere (Armstrong, 1979), which correlated with morphological and structural features of the roots (Pezeshki, 2001; Colmer, 2003; Stottmeister et al., 2003), with inundation adaptation or tolerance (Armstrong, 1967; Jackson and Armstrong, 1999), and with nutrient removal (Sorrell and Brix, 2003; Sasikala et al., 2009). However, results were generally only based on a few species, and could not reveal the mechanisms of the correlations. At present, it is widely understood that microorganisms play a key role in nutrient removal from constructed wetlands (Greenway, 2007; Muench et al., 2007; Weber et al., 2008). Root systems in constructed wetlands are closely correlated with the growth and decomposition activities of the microorganisms. It is difficult to understand the roles of microorganisms in constructed wetland nutrient removal, and difficult to reveal the nutrient removal mechanisms in constructed wetlands without an extensive understanding of the growth, distribution, and function of plant roots. However, the relationship between root characteristics (morphological and structural features, growth and distribution, and eco-physiological characteristics) and microorganism characteristics, and nutrient removal is still unclear, and the roles that plant roots and root zone play in the constructed wetland are still regarded as a“black box” (Stottmeister et al., 2003; Thullen et al., 2005). In their study of the morphological differences of wetland plant roots, Chen et al. (2004) distinguished two types of wetland plants: rhizomatic-root plants, and fibrous-root plants. Rhizomatic-root plants had rhizomes or a root system composed of thicker roots, and fibrous-root plants had a root system mostly composed of thinner roots (D 3 mm). They hypothesized that there were significant differences between the two types of wetland plants in root morphological and structural features, root growth and distribution characteristics, root amount and biomass, ROL, and nutrient removal. Primary results indicated that fibrous-root species had more roots and a larger root biomass, larger root surface area, faster root growth, more ROL, and higher nutrient removal rates, while having shorter root longevity, when compared with rhizomatic-root plants (Chen et al., 2007; Cheng et al., 2009). However, the current literature on plant comparison (including the above results) has been basically
limited to a few wetland species. Comparisons based on a few species may result in inconsistent results (Qiu et al., 2011). Greater research on more plant species is necessary in order to confirm the hypothesis. The aim of the present study was to investigate the supposition by comparatively studying the morphological, structural, and eco-physiological features of roots, root growth and biomass, and nutrient removal rates based on a greater selection of wetland plants.
2.
Materials and methods
2.1.
Wetlands and the plants used in the experiment
Thirty-five plant species or varieties were used in the present study (Table 1). They were planted in small-scale wetlands which were constructed in circular plastic pails (23 19 23 cm, D1 D2 H). The plants were fixed with round cystosepiments, a diameter of which was about the same as the upper inner diameter of the wetlands. All the seedlings (tillers or clones) used in the study were well established, with the age and size being almost identical (about 3 weeks old, 20 cm height, and 10 g weight). One seedling was planted in each wetland. Each species-specific wetland had five replicas. All the wetlands were arranged randomly at a distance of 30 cm from each other in a climate chamber. The average air temperature during daylight hours (7: 00e19: 00) and during the evening was 28.0 1.0 C and 18.0 1.0 C, respectively. Relative humidity was 80% 5%, and the illumination intensity was 20 K lx during daylight hours. The plants were cultured with tap water in the 1st week, with an influent of 50% nutrient solution and 50% tap water during the 2nd week, and with nutrient solution thereafter. Every wetland was provided with 5 L of the influent, which was exchanged every 5 d, with an average hydraulic loading of 0.03 m3 m2 d1. The nutrient solution was prepared according to Wiebner et al. (2005). The average concentration (mg L1) was 326.0 of chemical oxygen demand (CODcr), 61.6 of total nitrogen (TN), and 5.0 of total phosphorus (TP).
2.2.
Root length and longevity
For longevity measurement, two roots from the new growth of each plant were marked and observed every 3 days, and the average survival period of the marked roots of five plants was considered as longevity of the species. The longest root length of each plant was measured every 10 days, and the average of the five plants was considered as the root length of the species.
2.3.
Root porosity
Root porosity was measured using a modified buoyancy-based method (Visser and Bo¨gemann, 2003). The installation was made according to the density accessories of Sartorius YDK01. A glass beaker was filled with water at room temperature (25 C), and a hang sieve was immersed into the water. About 2 g of plant roots were sampled from each wetland during the 10th week. The samples were washed to remove any debris, wiped of surface water, and cut into pieces of 2e2.5 cm in length before being weighed on a salver to an accuracy level of
3943
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 4 1 e3 9 5 0
Table 1 e Plants used in the present study and their below-ground biomass characteristics (Mean ± SE, n [ 5). Below-ground biomass (g plant1)
Species D 1 mm A
B
Acorus calamus Linn. Arundo donax var. versicolor Stokes Canna indica Linn. Commelina diffusa Burm. f. Cyperus flabelliformis Rottb. Caldesia reniformis (D. Don) Makino Eclipta prostrata (Linn.) Linn. Lythrum salicaria Linn. Oenanthe javanica (Bl.) DC. Oryza sativa Linn. Panicum repens Linn. Phragmites australis Trin. ex Steud. Pontederia cordata Linn. Sagittaria trifolia Linn. Saururus chinensis (Lour.) Baill. Scirpus lacustris subsp. validus (Vahl) Koyama Scirpus triqueter Linn. Typha angustifolia Linn. Typh Typha laxmannii Lepech. Vetiveria zizanioides (Linn.) Nash
0.85 1.35 4.30 0.39 0.36 2.96 0.17 0.92 1.13 0.35 1.01 0.55 2.04 1.62 0.12 0.52 0.12 0.50 0.26 0.66
0.26 0.27 0.83 0.04 0.25 0.64 0.08 0.35 0.59 0.23 0.18 0.29 0.23 0.68 0.07 0.19 0.00 0.33 0.11 0.31
Aglaonema commutatum Schott. cv. ‘San Remo’ Aglaonema commutatum Schott. cv. ‘Silver Queen’ Alocasia cucullata (Lour.) Schott Alocasia macrorrhiza (Linn.) Schott Alpinia zerumbet (Pers.) Burtt et Smith Colocasia tonoimo Nakai Juncellus serotinus (Rottb.) C. B. Clarke Disporum cantoniense (Lour.) Merr. Hymenocallis littoralis (Jack.) Salisb. Iris tectorum Maxim. Lycoris longituba var. flava Y. Hsu et X. L. Huang Philodendron gloriosum Andre Ottelia alismoides (Linn.) Pers. Tradescantia spathacea Sw. Zizania latifolia (Griseb.) Stapf
0.20 0.02
0.15 0.05 0.18 0.41 0.31 0.08 0.02
0.03 0.18 0.10 0.17 0.03
0.03 0.03 0.23 0.06
1 mm < D 3 mm 1.41 1.45 2.36 0.13 0.81 4.26
0.49 0.35 0.61 0.03 0.29 0.94
Lateral roots
D > 3 mm
Rhizomes
0.60 0.43
2.44 0.68 5.50 2.56 17.49 2.80a
0.02 0.04 0.07 0.10 0.04 0.06 1.21 0.41 0.46 0.16 2.66 0.41
0.61 0.14 11.98 2.97 2.27 0.57
0.33 0.15 0.76 0.33 0.21 0.06 0.19 0.64 0.76 0.64 0.21 0.45 1.34 0.05 1.53 0.16 1.20 1.09 0.58 0.38 1.36
0.14 0.14 0.45 0.23 0.03 0.11 0.75 0.02 0.59 0.08 0.14 0.47 0.34 0.10 0.36
0.06 0.12
0.83 0.62 0.17 0.96 0.93
0.42 0.23 0.03a 0.53 0.17
0.96 0.22 0.16 0.17 0.08 0.05
9.25 3.39a 5.87 1.92a 0.88 0.33a
0.24 0.05 1.16 0.38 3.49 4.17a 2.38 0.56
1.29 0.34 0.04 0.07a
þ þ þ e þ þ þ þ e þ þ þ þ þ þ þ þ þ þ þ e e e þ e e e þ e þ e e e þ e
A, fibrous-root plants; B, thick-root plants; þ, having lateral roots; e, having no lateral roots. a Rhizomes which play reproductive and storage roles.
0.0001 g (W1). Then the root pieces were carefully transferred to the hang sieve, and again weighed under water (W2). The root pieces were then immersed into the room temperature water in a glass beaker, and evacuated under negative 0.1 MP until the water surface was without bubbles. The roots were then taken out, wiped of surface water, and again weighed on the salver (W3). Porosity was calculated using the formula: Porosity (%) ¼ 100 (W3W1)/(W1W2), and porosity of each species was calculated as the average of the five replicas.
2.4.
Root cross sectional structure
Mature plant roots were sampled during the 10th week, and sectioned freehand at about 5 cm from the root tip. Using a Leica DMN microscope with a Nikon E4500 digital camera, root cross sectional structures were observed and photographed.
2.5.
Biomass
At the end of the experiment, all plants were harvested. The biomasses of the shoot, roots of D 1 mm, 1 mm < D < 3 mm,
D 3 mm, and rhizomes were weighed to an accuracy level of 0.01 g. The dry weight of each part was weighed to an accuracy level of 0.0001 g after being dried to a constant weight at 80 C.
2.6.
Radical oxygen loss
ROL was measured non-destructively using the improved method of Armstrong (1979) during the 12th week. The instruments used in the experiment included a polarograph, a Ptcylinder, a Ag/AgCl anode (supplied by Professor Armstrong, the University of Hull, United Kingdom), and a list-styled recorder. Roots of 9e11 cm in length and D < 0.15 cm were selected for measurement. A solution of 0.1% agar, KCl (5 mM), and CaSO4 (0.5 mM) was sterilized at 121 C for 30 min, and deoxygenated by bubbling the solution with oxygen-free nitrogen for 24 h before use. A glass tube with a three-hole rubber bung was prepared, and the solution was poured or siphoned onto the open surface of the bung, then the Pt-cylinder and the Ag/ AgCl electrode were dipped into the medium through the rubber bung, respectively. A plant root was dipped into the solution in the same way, with a root apex of 0.5 cm inserted
3944
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 4 1 e3 9 5 0
through the Pt-cylinder. The root was wrapped with cotton to prevent air entering the solution. The ROL was measured at 5 mm from the root apex. An electrical current measurement was initiated as soon as the balanced voltage was set. ROL was calculated using the formula: ROL ¼ nFA/i, where ROL is the O2 loss rate (ng cm2 min1); n is the number of electrons for deoxygenating one molecular oxygen, set at 4; F is Faraday’s constant 96500; A is the root surface area measured in the cylindrical platinum electrode (cm2); and i is the electric current shown in the voltammeter (mA).
2.7.
Photosynthesis
When ROL was measured, the net photosynthetic rate, transpiration rate, and stomatal conductance were measured synchronously using an LI-6400 portable photosynthesis measurement system (LI-COR Company, America). For each individual plant, five normal leaves were measured, and each assay was performed in triplicate. The species-specific photosynthetic rate was the average of the measurements.
2.8.
Root activity
Root activity was determined using the modified method of Yang et al. (2004). Plant root samples of 1e2 g were taken during the 12th week, washed to remove any debris before they were placed into a triangular flask (100 mL). The roots were treated with 25 mL of a-naphthylamine solution (50 mg mL1) and 25 mL of phosphate buffer (pH 7.0) for 10 min. Then, after the flask was fully shaken, 2 mL of the solution was drawn to determine the content of a-naphthylamine using a colorimetry (Ota, 1970) (A1). The triangular flask containing the remaining 48 mL solution was airproofed using a rubber bung, and incubated for 1 h at 25 C before analysis, and colorimetry was also used for determining the a-naphthylamine content of the incubated solution (A2). A control without roots was also set with the same amount of solution as above, and corresponding values A10 and A20 , as well as A0 (A0 ¼ A10 A20 ) were obtained. The value (A1A2A0) was the amount of the a-naphthylamine oxidized by the roots (mg g1 dry weight h1).
2.9.
Nutrient removal
Nutrient removal rates were measured during the 14th week. A solution of about 300 mL was taken for measurement from each wetland after a 48-h retention period. Concentrations of TN, TP and CODcr were measured according to the standard method (State Environmental Protection Administration of China, 2002). Nutrient removal rates (%) were calculated on an influent basis.
2.10.
Statistical analysis
The mean and standard deviation of the parameters were calculated using the Statistical Product and Service Solution 13.0 and Microsoft Excel 2003. One way ANOVA with species or type as a factor was used to test the statistically significant differences between species or between types for root length, root longevity, root biomass, root porosity, ROL, photosynthetic
rate, root activity, and nutrient removal after the variances were tested for homogeneity.
3.
Results
3.1.
Morphological and anatomical features of roots
Marked differences in root size were found among the 35 plants ( p ¼ 0.001). Eclipta prostrata only had roots of D 1 mm (Table 1). Oenanthe javanica, Lythrum salicaria, and Oryza sativa had almost all roots of D 1 mm, and very few roots of D > 1 mm. Aglaonema commutatum cv. ‘Silver Queen’, Lycoris longituba var. flava, Philodendron gloriosum, Alocasia cucullata, Zizania latifolia, and Alpinia zerumbet had no roots of D 1 mm. Iris tectorum, Ottelia alismoides, and Hymenocallis littoralis all had very small parts of roots of D 1 mm. Typha laxmannii, Alocasia macrorrhiza, Disporum cantoniense, H. littoralis, A. commutatum cv. ‘San Remo’, A. commutatum cv. ‘Silver Queen’, L. longituba var. flava, and Canna indica had parts of roots of D 3 mm (not including rhizomes), among which A. commutatum cv. ‘Silver Queen’ and L. longituba var. flava had most roots being D 3 mm. Significant differences in maximum root length existed among the 35 plants ( p ¼ 0.001) (Table 2). The maximum root length of Sagittaria trifolia and Caldesia reniformis reached 36 cm and 34 cm, respectively. Five species had maximum root lengths of more than 20 cm, and 26 species had maximum root lengths between 10 and 20 cm, while the shortest root system had only 7 cm (Saururus chinensis). Roots of 1 mm < D < 3 mm which had further lateral roots generally had longer root length, such as roots of S. trifolia, and C. reniformis. Plants with thick roots had shorter maximum root lengths, and many of the thick roots had a thick epidermis and were succulent. Significant differences in root longevity were exhibited among the 35 plants ( p ¼ 0.001) (Table 2). The greatest root longevity was 78 d (T. laxmannii), and 74 d (H. littoralis), while the shortest longevity was only 23 d (Scirpus triqueter), and 24 d (O. javanica). There were 2 species (D. cantoniense and A. cucullata) having root longevity between 60 and 70 d, 7 species between 50 and 59 d, 11 species, between 40 and 49 d, 8 species, between 30 and 39 d, and 5 species, between 20 and 29 d. There were significant differences in root biomass among the 35 plants ( p ¼ 0.001) (Table 2). Pontederia cordata had the largest root biomass (16.68 g), which was 139 times that of S. triqueter (0.12 g), which had the smallest root biomass. Two species, P. cordata and Arundo donax var. versicolor, both had rhizomes, had the largest root biomass. Three species had a root biomass between 5 and 10 g, 6 species, between 2 and 5 g, 10 species, between 1 and 2 g, and 14 species, had less than 1 g. Root/shoot ratios of the 35 species were between 0.86 and 0.034. Most species with rhizomes had higher ratios, and species with fibrous roots had lower ratios. Seven species had a ratio higher than 0.40, 6 species had a ratio between 0.30 and 0.39, 9 species between 0.20 and 0.29, 7 species between 0.10 and 0.19, and 6 species, lower than 0.10. Significant differences in porosity were found among the species ( p ¼ 0.001) (Table 2). S. trifolia (52.6%), P. cordata (49.6%), and Phragmites australis (40.2%) had higher porosity rates, and
3945
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 4 1 e3 9 5 0
Table 2 e Root length, root longevity, below-ground biomass, and root porosity of 35 wetland plants (Mean ± SE, n [ 5). Type
Species
Maximum root length (cm)
Root longevity (day)
Fibrous-root plants
Acorus calamus Arundo donax var. versicolor Canna indica Commelina diffusa Cyperus flabelliformis Caldesia reniformis Eclipta prostrata Lythrum salicaria Oenanthe javanica Oryza sativa Panicum repens Phragmites australis Pontederia cordata Sagittaria trifolia Saururus chinensis Scirpus lacustris subsp. validus Scirpus triqueter Typha angustifolia Typha laxmannii Vetiveria zizanioides
23.00 13.00 13.00 10.00 15.00 34.00 15.00 12.00 15.00 16.00 20.00 15.00 23.00 36.00 6.80 12.00 10.00 18.00 15.00 16.00
2.08 1.41 1.29 1.67 1.87 2.77 0.83 2.10 1.83 0.71 1.41 1.53 2.95 5.90 1.17 2.00 2.16 3.00 1.67 1.58
32.00 54.00 42.00 28.00 40.00 46.00 31.00 34.00 24.00 40.00 30.00 54.00 36.00 39.00 37.00 35.00 23.00 26.00 78.00 45.00
Thick-root plants
Aglaonema commutatum Aglaonema commutatum Alocasia cucullata Alocasia macrorrhiza Alpinia zerumbet Colocasia tonoimo Juncellus serotinus Disporum cantoniense Hymenocallis littoralis Iris tectorum Lycoris longituba var. flava Philodendron gloriosum Ottelia alismoides Tradescantia spathacea Zizania latifolia
12.00 25.00 15.00 11.00 13.00 12.00 25.00 10.00 12.00 10.00 22.00 13.00 18.00 9.00 16.00
0.00 1.87 3.46 1.79 0.82 1.63 1.87 1.41 2.38 0.82 2.16 1.41 1.10 0.79 1.22
50.00 52.00 62.00 54.00 56.00 47.00 42.00 67.00 74.00 43.00 46.00 58.00 41.20 27.75 45.00
1.41 3.16 2.19 3.03 2.16 0.82 2.35 1.26 1.29 0.82 1.41 0.89 1.17 1.48 0.71
P. gloriosum had the lowest porosity rate (8.0%). Seven species had a porosity rate greater than 30%, 10 species, between 20% and 30%, 11 species, between 10% and 19%, and 7 species, less than 10%. All the wetland plants had a developed aerenchyma. The root transverse structure of all plants included three parts: an epidermis, cortex and a vascular cylinder. The cortex also included three parts, which were called the exodermis, mediopellis, and endodermis. Different plants had varied sizes and patterns of cavities in the mediopellis. For the 35 plants, there were basically two types of aerenchyma. In aerenchyma type I, the aerenchyma of the root was made up of large cavities which were always formed from many small pores, and distributed radially between the exodermis and the vascular cylinder. This type of roots only had differences in the aerenchyma area, regardless of whether the area of the vascular cylinder was large (Vetiveria zizanioides) or small (Acorus calamus, and C. indica), and regardless of the number of vascular bundles in the vascular cylinder (Plate IA, B). In aerenchyma type II, the aerenchyma of the root was made up of a large number of small cavities in the mediopellis (Plate IC, D). This type of roots had a small vascular cylinder and welldeveloped vascular bundles. The porosity in type II was
2.00 1.25 1.83 1.26 3.54 1.63 0.71 1.67 2.24 1.87 1.41 1.73 2.10 2.28 1.41 3.74 0.82 1.41 1.26 1.87
Below-ground biomass (g plant1) 4.70 8.30 7.26 0.52 1.17 7.22 0.17 0.95 1.20 0.38 2.23 1.62 16.68 3.89 0.95 1.14 0.12 1.80 2.01 0.87
1.37 3.15 1.51 0.06 0.48 1.51 0.08 0.38 0.69 0.27 0.57 0.42 3.34 1.21 0.49 0.36 0.01 0.86 0.64 0.33
1.36 0.08 0.81 0.28 0.76 0.45 0.87 0.29 0.21 0.03 0.43 0.13 1.74 0.57 0.60 0.11 2.77 0.74 3.67 4.28 3.58 0.67 1.09 0.47 0.61 0.37 1.90 0.43 1.36 0.36
Root porosity (%) 25.98 24.93 20.42 26.13 32.16 35.62 14.02 20.08 15.40 26.00 20.85 40.16 49.60 52.57 26.01 30.00 17.09 16.71 15.16 39.32
4.38 6.75 3.20 4.99 3.42 4.12 3.29 3.52 3.44 3.73 3.61 3.85 4.19 4.91 4.82 1.63 1.89 2.32 3.41 5.09
9.00 0.82 9.42 1.49 8.93 2.78 10.50 1.11 12.00 3.17 11.43 1.29 8.26 0.90 8.10 2.36 10.88 2.61 28.00 1.26 11.51 2.92 7.97 1.31 8.04 2.40 27.21 4.87 18.05 3.33
mainly below 20%, and was obviously lower than that in type I ( p ¼ 0.001), which was mostly above 20%.
3.2. Root types and comparison of morphological and eco-physiological features According to the root morphological and anatomical features, the 35 plants were divided into two types: fibrous-root plants and thick-root plants (Table 1). Fibrous-root plants had many roots of D 1 mm, with almost all of the roots being of D 2 mm (Plate IIA). Roots of 1 mm < D < 3 mm had many fine and long lateral roots (D 1 mm). The roots of these plants were long and had a thin epidermis and a low degree of lignification. The aerenchyma (aerenchyma type I) was radially distributed between the exodermis and the vascular tissues (Plate IA). Roots of thick-root plants were mostly of D 1 mm, some of them were of D 3 mm, and generally had no further fine (D 1 mm) lateral roots (Plate IIB). The roots were short, smooth, and fleshy, and had a thick epidermis. The aerenchyma (aerenchyma type II) had a large number of small cavities (Plate IB). A significant difference in biomass was exhibited between the types (Table 3). Fibrous-root plants had a significantly
3946
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 4 1 e3 9 5 0
Fig. 1 e Comparison of maximum root length, root longevity and root porosity between fibrous-root plants (FRP) and thick-root plants (TRP) (Error bars indicate SE, n [ 5). Different lowercase letters indicate a statistically significant difference at p < 0.05, and different capital letters indicate a statistically significant difference at p < 0.01.
higher root biomass of D 1 mm ( p ¼ 0.003) and the percentage of the biomass in the total root biomass ( p ¼ 0.001), leaf biomass ( p ¼ 0.009), above-ground biomass ( p ¼ 0.020), and total biomass ( p ¼ 0.032) compared to those of the thick-root plants. Fibrous-root plants had a higher maximum root length ( p ¼ 0.042), and porosity ( p ¼ 0.001), but shorter root longevity ( p ¼ 0.001) when compared to the thick-root plants (Fig. 1). There were significant differences in photosynthetic rate, transpiration rate, ROL, and root activity between the two types of plants. Fibrous-root plants showed an average photosynthetic rate of 8.65 0.45, and an average transpiration rate of 3.51 0.50, which was significantly higher than those (6.06 0.47, 1.65 0.31) of thick-root plants ( p ¼ 0.010, 0.004). But there was no significant difference in stomatic conductance between the types (Fig. 2). Fibrous-root plants had an average ROL of 106.63 6.24 ng cm2 min1, which was markedly higher than the average 54.00 5.21 ng cm2 min1 of
Fig. 2 e Comparisons of photosynthetic rates, transpiration rates, and stomatal conductance between fibrous-root plants (FRP) and thick-root plants (TRP) (Error bars indicate SE, n [ 5). Different lowercase letters indicate a statistically significant difference at p < 0.05, and different capital letters indicate a statistically significant difference at p < 0.01.
Fig. 3 e Comparisons of ROL and root activity between fibrous-root plants (FRP) and thick-root plants (TRP) (Error bars indicate SE, n [ 5). Different lowercase letters indicate a statistically significant difference at p < 0.05, and different capital letters indicate a statistically significant difference at p < 0.01.
thick-root plants ( p ¼ 0.000) (Fig. 3). The average root activity of fibrous-root plants (79.26 7.83 ug g1) was also higher than that of thick-root plants (50.84 5.10 ug g1) ( p ¼ 0.008) (Fig. 3). There were significant differences in TN and TP removal. Fibrous-root plants exhibited higher TN ( p ¼ 0.020) and TP ( p ¼ 0.002) removal rates than those of the thick-root plants. However there was no significant difference in CODcr removal rate between the two types (Fig. 4).
4.
Discussion
Previous research has indicated that wetland plants exhibit significant specific differences in root size, root surface area, root growth, and nutrient removal (Cheng et al., 2009). In the present study, significant differences were found in root morphological and anatomical features, root growth rate, root longevity, root activity, ROL, and nutrient removal rates among different plants. The differences in root morphological, anatomical, and eco-physiological features reflected the
Fig. 4 e Comparisons of nutrient removal rates between fibrous-root plants (FRP) and thick-root plants (TRP) (Error bars indicate SE, n [ 5). Different lowercase letters indicate a statistically significant difference at p < 0.05, and different capital letters indicate a statistically significant difference at p < 0.01.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 4 1 e3 9 5 0
3947
Plate I e The transverse sections of fibrous-root plants (A, B) and thick-root plants (C, D). A. Arundo donax var. versicolor Stokes B. Vetiveria zizanioides (Linn.) Nash C. Alocasia cucullata (Lour.) Schott D. Philodendron gloriosum Andre.
differences in adaptation of various wetland plants to their growing environments (Vartapetian and Jackson, 1997; Stottmeister et al., 2003). When plants adopt similar adaptive strategies, they may develop similar root features. This may provide the theoretical basis for dividing root types of wetland plants. In the present study, 35 wetland plants could be divided into two types according to the morphological, anatomical, and eco-physiological features of their root systems, and significant differences in these characteristics existed between the two types. However, a few species exhibited transitional characteristics between the two types, and their division was more or less subjective. In addition, wetland plants have considerable plasticity in their root morphology and anatomy, and the plasticity has obvious specific difference (Visser et al., 2000), some plants may have more obvious increase or decrease in root numbers (especially finer root number) than other species in solution culture system. Hence, the segregation of the 35 plants in the present study may be incomplete, further research is needed in more eco-physiological comparisons, and in observation of root morphology and anatomy in solid media. According to the morphological differences of roots, Chen et al. (2004) distinguished two types of wetland plants: fibrous-root plants and rhizomatic-root plants. The former included the plants with root systems made up mostly of thinner roots (D < 1 mm). And the latter included plants with thicker roots, or with rhizomes. In the present study, we found that some rhizomatic plants also had many roots of D 1 mm,
and the anatomical root features were similar to those of the fibrous-root plants. These rhizome plants also had as high root porosity as the fibrous-root plants, and their eco-physiological attributes were close to those of the fibrous-root plants, and were different from those of the thick-root plants. They also had higher ROL, photosynthetic rate, root activity, above- and below-ground growth rates, and nutrient removal rates, which were comparable to the fibrous-root plants. Therefore, we considered that it may be more reasonable to divide these rhizome plants into the fibrous-root plants category, and allow the other rhizome plants, which had no or few fine roots, to belong to the category of thick-root plants. Certainly, in order to obtain a more accurate conclusion, further research is necessary for more plant species. Previous research has indicated that wetland plants with more adventitious roots and larger root surface areas had higher growth rates and nutrient removal rates (Kyambadde et al., 2004; Cheng et al., 2009). More adventitious roots and larger root surface areas were considered to provide a greater attachment area for nitrifying bacteria (Kyambadde et al., 2004), and to have more oxygen transferring to the rhizoplane and rhizosphere to a wider extent (Brix et al., 1992; Joseph et al., 2004), therefore enhancing corresponding nitrification activities and N removal. More adventitious roots and larger root surface areas may also be favorable in the absorption of inorganic phosphorus, and to the attachment and growth of organic-phosphorus decomposing microorganisms, and hence promote P removal. In the present study,
3948
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 4 1 e3 9 5 0
Plate II e Root morphology of fibrous-root plants (A, B) and thick-root plants (C, D). A. Cyperus flabelliformis Rottb. B. Caldesia reniformis (D. Don) Makino C. Hymenocallis littoralis (Jack.) Salisb. D. Aglaonema commutatum Schott. cv. ‘Silver Queen’.
the correlations between root morphological and anatomical features, above- and below-ground growth, eco-physiological characteristics, and nutrient removal were further confirmed by comparatively observing a relatively large amount of different species of wetland plants. The fibrous-root plants had significantly higher above- and below-ground growth rates, photosynthetic rate, ROL, root activity, and nutrient removal rates. It is notable that the result reveals a positive correlation between ROL, photosynthetic rate, and root activity. ROL is an important feature for wetland plants to adapt to an anoxic habitat. However, it is still unclear whether ROL is an active or passive process of wetland plants (Sorrell and Brix, 2003), and how close a correlation exists between ROL and photosynthesis. The present study measured ROL,
photosynthesis, and root activity simultaneously for the same plant individuals, and indicated a close correlation between ROL, photosynthesis, and root activity. This suggests that ROL may correlate with the physiological activities of wetland plants. Results of the present study with regards to root types, and concerning the correlation between root morphological and anatomical features and plant growth, as well as the correlation between root morphological and anatomical features and nutrient removal may provide guidance into constructed wetland plant selection, and into the mechanisms of nutrient removal from constructed wetlands. The present study was carried out in a climate chamber. The growing conditions (average air temperatures and illumination) were simulated to those of spring and summer in
Table 3 e Biomass of fibrous-root and thick-root plants grown in microcosm wetlands (g mL2, mean ± SE, n [ 5).
a
FRP TRPa
Root (D 1 mm)
Root (1 mm < D 3 mm)
Root (D > 3 mm)
Total root
Shoot
Leaf
Total
1.01 0.30 0.11 0.04*
1.34 0.36 0.71 0.26
0.03 0.03 0.33 0.10*
3.01 0.81 1.24 0.25
12.04 2.37 5.03 0.98*
6.76 1.43 3.51 0.38*
16.07 3.14 7.56 1.23*
*p < 0.05 (statistically significant differences were tested between FRP and TRP). a FRP: Fibrous-root plants, TRP: Thick-root plants.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 4 1 e3 9 5 0
Guangzhou, China, and the shoot growth was similar to the growth of the plants in the field. From the 5th week, plants grew and tillered rapidly. During the 9th week, the shoot height of most of the plants was near their maximum height when compared to similar plants in the field, and C. indica, C. reniformis, and Tradescantia spathacea began flowering. However, the size of the wetland may have an effect on the growth and longevity of the roots. In addition, plants were grown in solution culture systems in the present study, and plant growth, root morphological and anatomical features, and root longevity may be different from those of the plants grown in solid substrates.
5.
Conclusion
Emergent wetland plants differed in root morphological and anatomical features which correlated with eco-physiological features such as photosynthesis, ROL, and nutrient removal. Two types of emergent wetland plants were distinguished according to root morphological and anatomical features: fibrous-root plants and thick-root plants. Fibrous-root plants displayed higher photosynthesis, greater ROL, and higher nutrient removal rates when compared to thick-root plants.
Acknowledgment The project was supported by the National Natural Science Foundation of China (No.30470346), and the Natural Science Foundation of Guangdong Province (No.06025056).
references
Armstrong, W., 1967. The oxidizing activity of roots in waterlogged soils. Physiol. Plant. 20, 920e926. Armstrong, W., 1979. Aeration in higher plants. In: Woolhouse, H. W.W. (Ed.), Advances in Botanical Research. Academic Press, London, pp. 226e332. Brix, H., Sorrell, B.K., Orr, P.T., 1992. International pressurization and convective gas flow in some emergent freshwater macrophytes. Limnol. Oceanogr. 37, 1420e1433. Chen, W.Y., Chen, Z.H., He, Q.F., Wang, X.Y., Wang, C.Y., Chen, D. F., Lai, Z.L., 2007. Root growth of wetland plants with different root types. Acta Ecol. Sin 27, 450e457. Chen, Z.H., Chen, F., Cheng, X.Y., Liu, Y.C., Zhou, X.Y., 2004. Researches on macrophyte roots in the constructed wetlands (A review). Curr. Trop. Plant Biol. 5, 131e142. Cheng, X.Y., Chen, W.Y., Gu, B.H., Liu, X.C., Chen, F., Chen, Z.H., Zhou, X.Y., Li, Y.X., Huang, H., Chen, Y.J., 2009. Morphology, ecology and contaminant removal efficiency of eight wetland plants with differing root systems. Hydrobiologia 623, 77e85. Colmer, T.D., 2003. Aerenchyma and an inducible barrier to radial oxygen loss facilitate root aeration in upland, paddy and deepwater rice (Oryza sativa L.). Ann. Bot. 91, 301e309. Greenway, M., 2007. The role of macrophytes in nutrient removal using constructed wetlands. In: Singh, S.N., Tripathi, R.D. (Eds.), Environmental Bioremediation Technologies. Springer, Berlin, Heidelberg, pp. 331e351. Gutknecht, J.L.M., Goodman, R.M., Balser, T.C., 2006. Linking soil process and microbial ecology in freshwater wetland ecosystems. Plant Soil 289, 17e34.
3949
Jackson, M.B., Armstrong, W., 1999. Formation of aerenchyma and the processes of plant ventilation in relation to soil flooding and submergence. Plant Biol. 1, 274e287. Joseph, K., Frank, K., Lena, G., Gunnel, D., 2004. A comparative study of Cyperus papyrus and Miscanthidium violaceum-based constructed wetlands for wastewater treatment in a tropical climate. Water Res. 38, 475e485. Kadlec, R.H., Tanner, C.C., Hally, V.M., Gibbs, M.M., 2005. Nitrogen spiraling in subsurface-flow constructed wetlands: implications for treatment response. Ecol. Eng. 25, 365e381. Kivaisi, A.K., 2001. The potential of constructed wetlands for wastewater treatment and reuse in developing countries: a review. Ecol. Eng. 16, 545e560. Kyambadde, J., Kansiimea, F., Gumaeliusb, L., Dalhammar, G., 2004. A comparative study of Cyperus papyrus and Miscanthidium violaceum-based constructed wetlands for wastewater treatment in a tropical climate. Water Res. 38, 475e485. Matheson, F.E., Nguyen, M.L., Cooper, A.B., Burt, T.P., Bull, D.C., 2002. Fate of 15N-nitrate in unplanted, planted and harvested riparian wetland soil microcosms. Ecol. Eng. 19, 249e264. Muench, C., Neu, T., Kuschk, P., Roeske, I., 2007. The root surface as the definitive detail for microbial transformation processes in constructed wetlands - a biofilm characteristic. Water Sci. Technol. 56, 271e276. Ota, Y., 1970. Diagnostic method for measurement of root activity in rice plant. Jpn. Agric. Res. Q. 5, 1e6. Pezeshki, S.R., 2001. Wetland plant responses to soil flooding. Environ. Exp. Bot. 46, 299e312. Qiu, Z.C., Wang, M., Lai, W.L., He, F.H., Chen, Z.H., 2011. Plant growth and nutrient removal in constructed monoculture and mixed wetlands related to stubble attributes. Hydrobiologia 661, 251e260. Sasikala, S., Tanaka, N., Wah Wah, H.S.Y., Jinadasa, K.B.S.N., 2009. Effects of water level fluctuation on radial oxygen loss, root porosity, and nitrogen removal in subsurface vertical flow wetland mesocosms. Ecol. Eng. 35, 410e417. Sinha, R.K., Bharambe, G., Bapat, P., 2007. Removal of high BOD and COD loadings of primary liquid waste products from dairy industry by vermi-filtration technology using earthworms. Int. J. Environ. Pollut. 27, 486e501. Sorrell, B.K., Brix, H., 2003. Effects of water vapour pressure deficit and stomatal conductance on photosynthesis, internal pressurization and convective flow in three emergent wetland plants. Plant Soil 253, 71e79. State Environmental Protection Administration of China, 2002. Methods for Water Analysis, fourth ed. Environment Science Press, Beijing, pp. 200e285 (in Chinese). Stein, O.R., Hook, P.B., 2005. Temperature, plants and oxygen: how does season affect constructed wetland performance? J. Environ. Sci. Health A 40, 1331e1342. Stottmeister, U., Wiebner, A., Kuschk, P., Kappelmeyer, U., Ka¨stner, M., Bederski, O., Mu¨ller, R.A., Moormann, H., 2003. Effects of plants and microorganisms in constructed wetlands for wastewater treatment. Biotechnol. Adv. 22, 93e117. Tanner, C.C., 2001. Plants as ecosystem engineers in subsurfaceflow treatment wetlands. Water Sci. Technol. 44, 9e17. Thullen, J.S., Sartoris, J.J., Nelson, S.M., 2005. Managing vegetation in surface-flow wastewater-treatment wetlands for optimal treatment performance. Ecol. Eng. 25, 583e593. Vartapetian, B.B., Jackson, M.B., 1997. Plant adaptations to anaerobic stress. Ann. Bot. 79 (Suppl. A), 3e20. Visser, E.J.W., Colmer, T.D., Blom, C.W.P.M., Voesenek, L.A.C.J., 2000. Changes in growth, porosity, and radial oxygen loss from adventitious roots of selected mono- and dicotyledonous wetland species with contrasting types of aerenchyma. Plant Cell Environ. 23, 1237e1245. Visser, E.J.W., Bo¨gemann, G.M., 2003. Measurement of porosity in very small samples of plant tissue. Plant Soil 253, 81e90.
3950
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 4 1 e3 9 5 0
Weber, K.P., Gehder, M., Legge, R.L., 2008. Assessment of changes in the microbial community of constructed wetland mesocosms in response to acid mine drainage exposure. Water Res. 42, 180e188. Wiebner, A., Kappelmeyer, U., Kuschk, P., Ka¨stner, M., 2005. Influence of the redox condition dynamics on the removal efficiency of a laboratory-scale constructed wetland. Water Res. 39, 248e256.
Yang, C.M., Yang, L.Z., Yang, Y.X., Ouyang, Z., 2004. Rice root growth and nutrient uptake as influenced by organic manure in continuously and alternately flooded paddy soils. Agr. Water Manage. 70, 67e81. Yang, Q., Chen, Z.H., Zhao, J.G., Gu, B.H., 2007. Contanminant removal of domestic wastewater by constructed wetlands: effects of plant species. J. Integr. Plant Biol. 49, 437e446.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 5 1 e3 9 5 9
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Characterization of natural organic matter adsorption in granular activated carbon adsorbers Silvana Velten a,*, Detlef R.U. Knappe b, Jacqueline Traber a, Hans-Peter Kaiser c, Urs von Gunten a,d, Markus Boller a, Se´bastien Meylan a,e a
Eawag, Swiss Federal Institute of Aquatic Science and Technology, 8600 Du¨bendorf, Switzerland Department of Civil, Construction, and Environmental Engineering, North Carolina State University, Raleigh, NC 27695-7908, USA c Stadt Zu¨rich, Wasserversorgung, Hardhof 9, Postfach 1179, 8021 Zu¨rich, Switzerland d Institute of Biogeochemistry and Pollutant Dynamics, ETH Zu¨rich, 8092 Zu¨rich, Switzerland e Cimo, Route de l’Ile-au-Bois, 1870 Monthey, Switzerland b
article info
abstract
Article history:
The removal of natural organic matter (NOM) from lake water was studied in two pilot-
Received 12 August 2010
scale adsorbers containing granular activated carbon (GAC) with different physical prop-
Received in revised form
erties. To study the adsorption behavior of individual NOM fractions as a function of time
31 March 2011
and adsorber depth, NOM was fractionated by size exclusion chromatography (SEC) into
Accepted 28 April 2011
biopolymers, humics, building blocks, and low molecular weight (LMW) organics, and NOM
Available online 6 May 2011
fractions were quantified by both ultraviolet and organic carbon detectors. High molecular weight biopolymers were not retained in the two adsorbers. In contrast, humic substances,
Keywords:
building blocks and LMW organics were initially well and irreversibly removed, and their
Adsorption
effluent concentrations increased gradually in the outlet of the adsorbers until a pseudo-
Granular activated carbon
steady state concentration was reached. Poor removal of biopolymers was likely a result
Natural organic matter (NOM)
of their comparatively large size that prevented access to the internal pore structure of the
NOM characterization
GACs. In both GAC adsorbers, adsorbability of the remaining NOM fractions, compared on
Size exclusion chromatography
the basis of partition coefficients, increased with decreasing molecular size, suggesting that increasingly larger portions of the internal GAC surface area could be accessed as the size of NOM decreased. Overall DOC uptake at pseudo-steady state differed between the two tested GACs (18.9 and 28.6 g-C/kg GAC), and the percent difference in DOC uptake closely matched the percent difference in the volume of pores with widths in the 1e50 nm range that was measured for the two fresh GACs. Despite the differences in NOM uptake capacity, individual NOM fractions were removed in similar proportions by the two GACs. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
Natural organic matter (NOM) is an important water constituent that affects the performance of drinking water treatment processes and the quality of drinking water. For example, NOM exerts coagulant and oxidant/disinfectant demands,
contributes to membrane fouling, and serves as precursor material for disinfection byproducts (DBPs). Apart from the NOM concentration, which is typically measured by bulk parameters such as total organic carbon (TOC) or UV absorbance at 254 nm, the character of the NOM affects such factors as reactivity with disinfectants, membrane fouling, and biological
* Corresponding author. Tel.: þ41 44 823 5067; fax: þ41 44 823 5547. E-mail address:
[email protected] (S. Velten). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.04.047
3952
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 5 1 e3 9 5 9
stability of water in distribution systems. As a result, it is important for utilities to know both the concentration and the character of the NOM in their raw and finished waters as well as in the effluents of individual treatment stages. As utilities across the world need to increasingly rely on less pristine sources for the production of drinking water, the use of granular activated carbon (GAC) is becoming more widespread. Apart from trace contaminant removal, the need for enhanced NOM removal can be a driver for the installation of GAC adsorbers. Because aquatic NOM varies in character both spatially across different water sources and temporally within a water source (Owen et al., 1993), the rate and extent of NOM adsorption in GAC filters are difficult to predict (Matsui et al., 2002). Aspects of NOM character such as molecular weight distribution (MWD), degree of hydrophobicity, charge distribution, and ability to form hydrogen bonds with the GAC surface affect NOM adsorption. In addition, physical (surface area; size, shape and volume of pores) and chemical (charge, type, and number of surface groups; ash content) GAC characteristics influence NOM adsorption (Summers and Roberts, 1988). Furthermore, solution pH, ionic strength, and hardness influence the adsorption of NOM (Newcombe, 1999). Physical effects are largely governed by the MWD of NOM relative to the pore size distribution (PSD) of GAC (Newcombe et al., 1997; Karanfil and Kilduff, 1999). NOM adsorption primarily takes place in mesopores (2e50 nm width) and large micropores (1e2 nm width) (Lee et al., 1981; Summers and Roberts, 1988; Newcombe et al., 1997; Pelekani and Snoeyink, 1999; Ebie et al., 2001; Li et al., 2003; Cheng et al., 2005) and is negligible on AC that primarily contains small micropores (<1 nm in width) (Pelekani and Snoeyink, 1999; Cheng et al., 2005). The high molecular weight (HMW) fraction (>w10,000 Da) of NOM does not readily adsorb because of size exclusion effects; however, this NOM fraction is generally well removed by coagulation and is frequently present only at low concentrations in GAC adsorber influents (Vuorio et al., 1998; Matilainen et al., 2006). In contrast, intermediate molecular weight fractions (w500e4000 Da) are well removed by GAC. Although low molecular weight (LMW) NOM constituents have access to a large percentage of the GAC pore volume and thus could be well removed based on size considerations alone, LMW compounds may also be relatively hydrophilic and, hence, less adsorbable. Also, GAC adsorber effluents may contain LMW compounds that represent metabolic products of microorganisms living on GAC surfaces (Matilainen et al., 2006). Chemical interactions are influenced by characteristics of the adsorbate, the adsorbent surface, and the water matrix. Because of the presence of carboxylic acid and phenolic groups in the NOM structure, NOM carries a negative charge at pH values typically encountered in drinking water treatment (Perdue and Lytle, 1983). Depending on the solution pH, the GAC base material and activation process, the net pore surface charge of GAC can be positive, negative, or neutral (Karanfil and Kilduff, 1999). Consequently, electrostatic interactions can affect NOM adsorption. Overall, however, the adsorbent pore size distribution appears to be the principal factor controlling NOM uptake by GAC while electrostatic effects play a secondary role (Newcombe and Drikas, 1997; Newcombe et al., 2002).
To date, most information about NOM adsorption by activated carbon was collected in batch tests. Little information is available on the dynamic breakthrough behavior of individual NOM fractions along the depth of GAC adsorbers treating natural water. Prior studies employing SEC to study NOM removal in GAC adsorbers were limited to UV detectors (Vuorio et al., 1998; Nissinen et al., 2001; Matilainen et al., 2006; Li et al., 2007), and none of these studies developed breakthrough curves for individual NOM fractions. Therefore the objectives of this study were to (1) determine the breakthrough behavior of dissolved organic carbon (DOC) and individual NOM components such as biopolymers, humic substances, building blocks and LMW organics in pilot-scale GAC adsorbers and (2) compare the adsorbability of individual NOM fractions on two GACs with different pore size distributions and grain sizes.
2.
Materials and methods
2.1.
Experimental set up and water characteristics
Two pilot-scale GAC adsorbers, GAC 1 and GAC 2, were set up at the Zurich Water Works (WVZ-Lengg, Zurich, Switzerland) treating lake Zurich water. GAC 1 was operated with ozonated water from the intermediate ozonation step of the full-scale water treatment plant (treatment scheme: pre-ozonation, rapid sand filtration, intermediate ozonation, GAC filtration and slow sand filtration; Hammes et al., 2010). GAC 2 was part of a pilot plant (treatment scheme: ozonation, GAC filtration, ultrafiltration; Hammes et al., 2008). Both reactors were operated in downflow mode. Operational and influent water quality parameters for the two GAC adsorbers are given in Table 1. Given the short empty bed contact time (EBCT) of GAC 1 (1.65 min), only influent and effluent samples were collected for NOM characterization. For GAC 2, aqueous samples were collected at the influent, effluent, and at different bed depths
Table 1 e Operational and influent water quality parameters for GAC 1 and GAC 2 adsorbers. Parameters
GAC 1
GAC 2
Carbon type
Chemviron F400
GAC filter bed (m3) Filtration velocity (m/h) Backwash rate (m/h) Backwash time (min/d) Empty bed contact time (min) GAC depth (m) Column diameter (m) GAC particle diameter (mm) Bed porosity () Packed bed density (kg/m3) Influent DOC (mg/L) Influent UV254 (1/m) Influent SUVA L/(mg*m) Influent pH ()
0.015 8 22.9 10 1.65 0.22 0.3 0.125e0.71
Chemviron SGL 8 18 1.47 5.9 n/a n/a 15.76 1.55 1.1 1.002.38
0.47 425
0.39 460
0.96 (0.03) 1.07 (0.14) 1.12 (0.12) 8.1 (0.3)
1.1 (0.04) 1.84 (0.18) 1.58 (0.18) 7.79 (0.14)
3953
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 5 1 e3 9 5 9
that corresponded to EBCTs of about 1, 4, 8, 12, and 16 min. GAC samples from GAC 1 were taken from the top of the filter. For GAC 2 sampling, a metal tube (inner and outer diameter 0.9 and 1.1 cm, respectively) was inserted 0.8 m into the GAC filter through sampling ports located on the side walls, and about 20 g GAC particles were collected from each sampling point. To not disturb the mass transfer zone in GAC 2, the adsorber was not backwashed during the course of this study.
2.2.
NOM characterization
Glass vials for DOC sampling were acid washed and baked at 450 C for 4 h. Upon in-line filtration (0.45 mm polyethersulfone filter) and acidification, NOM was characterized by size exclusion chromatography (SEC) and quantified by both ultraviolet (UV) and organic carbon (OC) detectors (SECOCD, DOC Labor Dr. Huber, Karlsruhe, Germany). Details about the SEC-OCD instrument, analytical method, and data interpretation can be found in Huber et al. (2011). The SEC column (Toyopearl TSK HW-50S, 250 20 mm) had a fractionation range of 100e10,000 Da. The UV detector (Knauer K200) provided an online UV signal and OC was detected by an infrared (IR) detector after oxidation of NOM to CO2 in a Graentzel Thin-Film Reactor. Chromatograms were interpreted with customized software (Chrom CALC, Huber et al., 2011). A phosphate buffer was used as the eluent (24 mM, pH 6.6) and the flow rate was set at 1 mL/min. Using an injection volume of 1 mL, the detection limit for NOM fractions eluting from the SEC column was 10 mg C/L.
2.3.
NOM uptake and partition coefficients
NOM uptake by GAC was calculated as follows: q¼
n Q X $ Dti $ 0:5$ cin;i1 þ cin;i 0:5$ cout;i1 þ cout;i mGAC i¼1
q Δti cin,i1 cin,i cout,i1 cout,i Q mGAC
solid-phase NOM concentration (g/kg GAC) time interval between two sampling dates (d) influent concentration at time i-1 (g/L) influent concentration at time i (g/L) effluent concentration at time i-1 (g/L) effluent concentration at time i (g/L) flow rate (L/d) mass of GAC in column (kg)
The partition coefficient Kp of individual NOM components was calculated at pseudo-steady state conditions from the ratio of solid and aqueous-phase concentrations, q and c: KP ¼ q=c
2.4.
Activated carbons
Two bituminous-coal based GACs (Chemviron SGL 8 18 and Chemviron F400, Chemviron, Feluy, Belgium) were used. GAC characteristics are summarized in Table 2.
Table 2 e Physical GAC characteristics of two GACs that were used in the GAC filters to assess NOM adsorption. BET surface area and pore volume distributions were obtained from N2 adsorption isotherm data. Parameters
GAC 1
GAC 2
Carbon Type
Chemviron F400 1060 0.175
Chemviron SGL 8 18 790 0.165
0.207
0.127
0.203 0.232
0.124 0.146
BET Surface Area (m2/g) Primary Micropore Volumea 0e1 nm (cm3/g) Secondary Micropore Volumea 1e2 nm (cm3/g) DFT Mesopore Volumeb (cm3/g) BJH Mesopore Voumec (cm3/g)
a Micropore volume calculated by density functional theory (DFT) for pores with widths less than 2 nm. b Mesopore volume calculated by density functional theory (DFT) for pores with widths ranging from 2 to 36 nm (upper limit for DFT model). c Mesopore volume calculated by Barrett, Joyner, and Halenda (BJH) method for pores with widths ranging from 2 to 50 nm.
2.5.
Adsorbent characterization
Brunauer, Emmett and Teller (BET) surface areas, micropore and mesopore volumes, and pore size distributions (PSDs) were determined from N2 adsorption isotherm data collected at 77 K (Autosorb-1-MP, Quantachrome, Boynton Beach, FL, USA). Prior to analysis, adsorbent samples were outgassed for 24 h at 423 K. BET surface areas were determined from 18-point adsorption isotherms that were completed with a 0.1-g sample in the 0.01e0.3 relative pressure range. PSDs were determined by conducting N2 adsorption experiments in two separate phases: high and low pressure phase. Adsorption isotherms from both pressure phases were combined to produce one isotherm. Micropore volume and PSD were computed from N2 adsorption isotherm data using the Density Functional Theory (DFT) with the N2_carb1.gai kernel (PC software version 1.51, Quantachrome, Boynton Beach, FL, USA). In addition, the mesopore volume was computed using the Barrett, Joyner, and Halenda (BJH) method that captures the entire mesopore range (2e50 nm). Additional details of the experimental procedure are described in Knappe et al. (2007).
3.
Results and discussion
3.1.
GAC 1
NOM adsorption by GAC 1 was studied for 112 days or 98,000 empty bed volumes (EBV). Fig. 1 shows DOC and UV absorbance data collected at the inlet and outlet of GAC 1. Following rapid initial breakthrough, effluent DOC and UV absorbance values reached a pseudo-steady state after about 40 days of operation or 36,000 EBV. At that time, the UV absorbance values measured in the adsorber inlet and outlet were almost identical whereas the DOC data differed by about 0.1 mg-C/L (Fig. 1). This difference may be explained by biodegradation of a DOC fraction with low UV absorbance. Bacteria attached to the GAC surface can metabolize the readily bioavailable
3954
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 5 1 e3 9 5 9
A
B
C
D
E
F
Fig. 1 e Influent and effluent concentrations of DOC, UV absorbance, biopolymers, humics, building blocks and LMW organics measured for GAC 1.
fraction of DOC, which typically consists of LMW organic compounds with low UV absorbance (Herzberg et al., 2003; Huber et al., 2011). For GAC 1, 50% DOC breakthrough was reached after w2400 EBV. When GAC is used to remove NOM with the goal to lower DBP formation during chlorination, full-scale experience suggests that 50% total organic carbon (TOC) breakthrough is reached after w6400 and w7900 EBV for empty bed contact times (EBCT) of 10 and 20 min, respectively (Bond and DiGiano, 2004). For the purposes of this discussion, it was assumed that DOC was the principal contributor to TOC. The earlier occurrence of 50% DOC breakthrough obtained here resulted from the short EBCT of GAC 1 (1.65 min) relative to those used in practice. Regarding the onset of the pseudosteady state region, results obtained with GAC 1 were similar to those of a prior study (Speitel et al., 1989), during which a pseudo-steady state TOC concentration was reached
after 35,000e40,000 empty bed volumes when evaluating TOC removal from ozonated water in a GAC adsorber with an EBCT of 1.2 min. In Fig. 2, size exclusion chromatograms obtained with OC (panel A) and UV detection (panel B) are compared for influent and effluent samples obtained from GAC 1 during the first 50 days of operation. The size exclusion chromatogram of the influent that was obtained with OC detection exhibits a peak for biopolymers (28 min retention time) followed by a large humic substances peak (42 min retention time) that constituted w50% of the total DOC. Building blocks (45 min retention time) are degradation products of humic substances. LMW organics (51 min retention time) are composed of LMW acids and LMW humics that elute simultaneously. LMW neutrals (>55 min retention time) are neutral or amphiphilic molecules that interact with the column and consequently have a longer retention time than the elution time of water (55 min).
3955
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 5 1 e3 9 5 9
Fig. 2 e SEC chromatograms obtained with OC detection (A) and UV detection (B) for influent and effluent samples of GAC 1.
Compared to the size exclusion chromatogram obtained with OC detection, that obtained with UV detection differs in four aspects: (1) the biopolymer peak is absent, (2) the relative signal strength for building blocks is lower, (3) the relative signal strength for LMW organics is higher, and (4) an inorganic colloids peak (e.g. clay particles or silicates) is present which is not seen in the OC-response. Fig. 2 qualitatively shows the removal of individual NOM fractions in a GAC adsorber. Biopolymers were not retained by the GAC filter as influent and effluent concentrations were practically identical after one day of operation. Concentrations of humics, building blocks and LMW organics increased gradually in the effluent during the first 50 days of operation. The three dominant NOM fractions (humics, building blocks, and LMW organics) were initially all well removed in the GAC adsorber, and this observation was consistent between chromatograms obtained with OC and UV detection. After 50 days of operation, the removal of humics became negligible while a small percentage of the two smaller NOM fractions (building blocks and LMW organics) continued to be removed across the GAC adsorber (Fig. 2). While biodegradation likely contributed to
the removal of LMW compounds, the appearance of soluble microbial products or extrapolymeric substances was not detectable by SEC-OCD analysis. During the entire study period, concentrations of all NOM fractions decreased along the length of the filter bed. This observation suggests that displacement of one adsorbed NOM fraction by another did not occur and that humics, building blocks, and LMW compounds adsorbed irreversibly. To permit a more quantitative assessment about the adsorbability of individual NOM fractions, breakthrough curves for biopolymers, humics, building blocks and LMW organics are depicted in Fig. 1. Panel C confirms that biopolymer removal was negligible by GAC 1 (Fig. 1). Biopolymers such as protein- and polysaccharide-like substances are large molecules with MW >10,000 Da. The number of adsorption sites for such large molecules on AC is limited because of size exclusion effects. Humics were initially well retained by GAC 1 but the AC rapidly lost its adsorption effectiveness (Fig. 1D). Fifty percent humics removal was reached after w4500 EBV and the inlet concentration was reached after w30,000 EBV. Compared to biopolymers, humics are smaller
Table 3 e Solid-phase concentrations and partition coefficients (Kp) of DOC and individual NOM fractions calculated for GAC 1 and GAC 2 at pseudo-steady state. Pseudo-steady state was reached after 36,000 EBV or 40 days in GAC 1 and at 120 cm after 20,000 EBV or 168 days in GAC 2. Fractions
DOC Biopolymers Humics Building Blocks LMW organics
GAC 1
GAC 2
Solid-phase concentration (g-C/kg GAC)
Kp (L/g GAC)
Solid-phase concentration (g-C/kg GAC)
Kp (L/g GAC)
28.6 1.1 11.3 7.1 6.8
34.4 14.7 24.2 49.7 53.5
18.9 0.0 7.8 3.8 1.6
22.8 0.0 16.7 26.8 41.0
3956
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 5 1 e3 9 5 9
that the adsorbability of NOM fractions increased with decreasing size; i.e. biopolymers < humics < building blocks < LMW organics. The greater adsorbability of smaller NOM constituents is likely the result of their ability to access a larger percentage of the total AC pore volume.
3.2.
Fig. 3 e DOC breakthrough curves obtained with GAC 2 at different empty bed contact times and with GAC 1: effluent DOC concentration normalized with respect to influent DOC concentration versus empty bed volumes treated.
molecules with an average molecular weight of w1000 Da and a diameter of <2 nm; they can therefore enter the mesopores and larger micropores of activated carbon (Huber and Frimmel, 1996; Wilkinson et al., 1999). Similarly the building blocks (w300e500 Da) were well retained and 50% building blocks removal was reached after w9000 EBV. Complete breakthrough of building blocks was not reached during the operation time of GAC 1 (Fig. 1E), and about 20% of the influent building blocks were removed in the GAC filter at pseudo-steady state. LMW organics (<350 Da) followed a similar pattern as building blocks with 50% LMW organics removal after w17,000 EBV (Fig. 1F) and approximately 25% removal at pseudo-steady state. The decrease of building blocks and LMW organics is most probably due to biodegradation in the biofilm that developed in GAC 1. A concurrent study focusing on the development of the microbiology on GAC 1 showed that the number of microorganisms reached a maximum after 33 days of operation (30,000 EBV) (Velten et al., 2007). Table 3 summarizes adsorption capacities and partition coefficients for DOC and individual NOM fractions after pseudosteady state was attained in GAC 1 and GAC 2. The solid-phase concentrations were determined by integrating the difference between the influent and the effluent DOC concentration over time. Overall DOC uptake in GAC 1 was 28.6 g-C/kg GAC. Humics exhibited the highest solid-phase concentration on the GAC (11.3 g-C/kg GAC) followed by building blocks (7.1 g-C/kg GAC) and LMW organics (6.8 g-C/kg GAC) and finally biopolymers (1.1 g-C/kg GAC). The high solid-phase concentration of humics can in part be explained by its dominance in the influent NOM composition (50%). To compare the adsorbability of DOC and individual NOM fractions, partition coefficients (Kp) were calculated from the pseudo-steady state solid-phase concentration and the corresponding aqueous-phase concentration in the GAC 1 effluent (Kp ¼ q/C ). The Kp values in Table 3 suggest
GAC 2
NOM removal by GAC 2 was studied as a function of bed depth (or EBCT) over a 4-month period. The evolution of DOC concentrations as a function of EBV and GAC bed depth is depicted in Fig. 3. Fifty percent DOC breakthrough in GAC 2 was reached immediately after start up of the reactor in the top 2 layers (10 and 40 cm) and was reached after 4300 EBV at 80 cm (EBCT ¼ 8.1 min); 5300 EBV at 120 cm (EBCT ¼ 12.2 min); and 6500 EBV at 155 cm (EBCT ¼ 15.8 min). EBV values to 50% DOC breakthrough that were obtained with GAC 2 are typical when compared to U.S. data that were gathered as part of the information collection rule (6449 3935 EBV at an EBCT of 10 min; Bond and DiGiano, 2004). Plotting normalized effluent DOC concentrations as a function of treated EBV, a comparison of the breakthrough curves at different filter depths (Fig. 4) shows improved DOC removal with increasing EBCT up to an EBCT of 12.2 min; an increase in EBCT from 12.2 to 15.8 min did not offer an advantage in terms of decreasing carbon usage rate, however. Breakthrough curves for biopolymers, humics, building blocks and LMW organics obtained with GAC 2 at different bed depths are shown in Fig. 4. In addition, results obtained with GAC 1 are included for comparison. As was the case with GAC 1, biopolymers were only slightly retained in GAC 2. Humics and building blocks were well retained, and concentrations decreased with increasing filter depth. At an EBCT of 15.8 min, 50% breakthrough of humics, building blocks, and LMW organics was reached after 5100 EBV (56 days), 6500 EBV (71 days), and 10,300 EBV (112 days), respectively. Adsorption capacities and partition coefficients of DOC and individual NOM fractions for GAC 2 are summarized in Table 3. The overall uptake of DOC in GAC 2 was 18.9 g-C/kg GAC. As was the case with GAC 1, humics showed the highest solid-phase concentration (7.8 g-C/kg GAC), followed by building blocks (3.8 g-C/kg GAC), LMW organics (1.6 g-C/kg GAC) and biopolymers (negligible adsorption). Partition coefficients for individual NOM fractions were calculated once the effluent DOC concentration reached a pseudo-steady state (w168 days at an EBCT of 12.21 min). The order of NOM fraction adsorbability matched that obtained with GAC 1 (i.e., biopolymers < humics < building blocks < LMW organics). These calculations were made by assuming that NOM fractions were only removed by adsorption even though biological degradation proceeded simultaneously in the GAC filter. Considering the population of bacteria in the GAC filter and a biological degradation capacity of 7.9 107 mg C/bacterium (calculated for this filter as part of another study, see Velten et al., 2007), true solid-phase concentrations could be 20e50% lower.
3.3.
Comparison of GACs
In this study two adsorbers containing GACs with different physicochemical properties were tested. The BET surface area
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 5 1 e3 9 5 9
3957
Fig. 4 e Breakthrough curves of biopolymers, humics, building blocks and LMW organics obtained with GAC 2 at different empty bed contact times and with GAC 1.
of the AC used in GAC 2 exhibited a BET surface area that was 75% of that used in GAC 1; furthermore, the AC used in GAC 2 had secondary micropore (1e2 nm pore width range) and mesopore (2e50 nm pore width range) volumes that were approximately 60% of those measured for the AC used in GAC 1 (Table 2). In contrast, primary micropore volumes (<1 nm width) were similar for the two ACs. Also, a smaller grain size was used in GAC 1 (US mesh 25 120) compared to GAC 2 (US mesh 8 18). Results of prior studies suggest that NOM adsorbs primarily in secondary micropores and mesopores. Therefore, DOC uptake was expected to be greater by GAC 1. The solid-phase concentrations and partition coefficients shown in Table 3 indeed confirm that GAC 1 was more effective for NOM removal than GAC 2. At pseudo-steady state, DOC uptake by GAC 2 was approximately 66% of that obtained by GAC 1, a result that is in reasonable agreement with the difference in mesopore and secondary micropore volumes between the two tested ACs. This result illustrates that extent of NOM uptake in GAC adsorbers is largely controlled by pores
with widths in the 1e50 nm range and confirms prior results obtained in batch tests. To assess whether GACs with different physical properties fractionate NOM differently, size exclusion chromatograms for the influent and effluent of GAC 1 and GAC 2 were compared at points of operation, at which similar overall DOC removal was measured [for GAC 1 after 8 days or 7230 EBV and for GAC 2 after 42 days or 7434 EBV of operation (EBCT 8.14 min)]. As illustrated in Fig. 5, relative peak heights of the influents to the two GAC adsorbers were similar despite the fact that there were small differences in the bulk influent water quality parameters (Table 1). Furthermore, the data in Fig. 5 show that all NOM fractions adsorbed similarly in the two GAC adsorbers even though the physical characteristics of the two GACs differed (Table 2). Consequently, the effluent NOM composition at a given DOC removal percentage did not appear to be affected by differences in GAC pore structure. Finally, it should be noted that the smaller grain size of the AC used in GAC 1 aided the rate of NOM removal (Fig. 3).
3958
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 5 1 e3 9 5 9
Acknowledgments We acknowledge financial assistance from the Eawag Wave21 project, the Zu¨rich Water Works (WVZ) and thank Heinz Bischofberger for technical and scientific assistance.
references
Fig. 5 e SEC-OCD chromatogram for influent and effluent concentrations of GAC 1 and GAC 2 after treatment of 7230 and 7434 EBV, respectively.
As illustrated in Fig. 3, the initial performance of GAC 1 (EBCT ¼ 1.65 min) was similar to that of the GAC 2 layer with an EBCT of 8.14 min.
4.
Conclusions
To achieve a high NOM removal efficiency with GAC adsorption, GAC with a high surface area and a large volume of pores with widths of 1e50 nm should be selected by utilities. The results of this study show that biopolymers do not adsorb on GAC while the adsorbability of the remaining NOM fractions increased with decreasing molecular size (i.e., humics < building blocks < LMW organics). All NOM fractions adsorbed similarly on the two GACs; i.e., the NOM composition in the effluents of GAC1 and GAC2 were similar even though the GACs exhibited different pore size distributions. Moreover, no displacement of individual NOM fractions occurred. The adsorption of individual NOM fractions can therefore be considered as irreversible. Biopolymers and humics play an important role in membrane fouling (Howe and Clark, 2002; Costa et al., 2006; Jermann et al., 2007). This study showed that biopolymers were not removed by GAC. Humics adsorbed on fresh GAC, but the adsorption capacity decreased rapidly. GAC filtration is therefore not expected to be an effective pretreatment technology for the removal of these membrane foulants. Regarding DBP precursors, prior studies showed that DBP yields (mmol DBP/mmol C) from higher MW humic and lower MW non-humic NOM fractions are similar (Hwang et al., 2001; Imai et al., 2003). The results of this research suggest that GAC adsorption is more effective for DBP precursor control in waters containing a larger percentage of LMW NOM.
Bond, R.G., DiGiano, F.A., 2004. Evaluating GAC performance using the ICR database. Journal American Water Works Association 96, 96e104. Cheng, W., Dastgheib, S.A., Karanfil, T., 2005. Adsorption of dissolved natural organic matter by modified activated carbons. Water Research 39, 2281e2290. Costa, A.R., de Pinho, M.N., Elimelech, M., 2006. Mechanisms of colloidal natural organic matter fouling in ultrafiltration. Journal of Membrane Science 281, 716e725. Ebie, K., Li, F.S., Azuma, Y., Yuasa, A., Hagishita, T., 2001. Pore distribution effect of activated carbon in adsorbing organic micropollutants from natural water. Water Research 35, 167e179. Hammes, F., Berney, M., Wang, Y., Vital, M., Ko¨ster, O., Egli, T., 2008. Flow-cytometric total bacterial cell counts as a descriptive microbial parameter for drinking water treatment processes. Water Research 42, 269e277. Hammes, F., Berger, C., Ko¨ster, O., Egli, T., 2010. Assessing biological stability of drinking water without disinfectant residuals in a full-scale water supply system. Journal of Water Supply: Research and Technology e AQUA 59 (1). Herzberg, M., Dosoretz, C.G., Tarre, S., Green, M., 2003. Patchy biofilm coverage can explain the potential advantage of BGAC reactors. Environmental Science and Technology 37, 4274e4280. Howe, K.J., Clark, M.M., 2002. Fouling of microfiltration and ultrafiltration membranes by natural waters. Environmental Science and Technology 36, 3571e3576. Huber, S.A., Frimmel, F.H., 1996. Size-exclusion chromatography with organic carbon detection (LC-OCD): a fast and reliable method for the characterization of hydrophilic organic matter in natural waters. Vom Wasser 86, 277e290. Huber, S.A., Balz, A., Albert, M., Pronk, W., 2011. Characterisation of aquatic humic and non-humic matter with size-exclusion chromatography e organic carbon detection e organic nitrogen detection (LC-OCD-OND). Water Research 45, 879e885. Hwang, C.J., Krasner, S.W., Sclimenti, M.J., Amy, G.L., Dickenson, E., le, F., Crouec , J.-P., Violleau, D., Bruchet, A., Prompsy, C., Gisec Leenheer, J., 2001. Polar NOM: Characterization, DBPs and Treatment. American Water Works Association Research Foundation, Denver, CO, USA. Imai, A., Matsushige, K., Nagai, T., 2003. Trihalomethane formation potential of dissolved organic matter in a shallow eutrophic lake. Water Research 37, 4284e4294. Jermann, D., Pronk, W., Meylan, S., Boller, M., 2007. Interplay of different NOM fouling mechanisms during ultrafiltration for drinking water production. Water Research 41, 1713e1722. Karanfil, T., Kilduff, J.E., 1999. Role of granular activated carbon surface chemistry on the adsorption of organic compounds. 1. Priority pollutants. Environmental Science & Technology 33, 3217e3224. Knappe, D.R.U., Rossner, A., Snyder, S.A., Strickland, C., 2007. Alternative Adsorbents for the Removal of Polar Organic
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 5 1 e3 9 5 9
Contaminants. American Water Works Association Research Foundation, Denver, CO, USA. Lee, M.C., Snoeyink, V.L., Crittenden, J.C., 1981. Activated carbon adsorption of humic substances. Journal American Water Works Association 73, 440e446. Li, Q., Snoeyink, V.L., Marin˜as, B.J., Campos, C., 2003. Pore blockage effect of OM on atrazine adsorption kinetics of PAC: the roles of PAC pore size distribution and NOM molecular weight. Water Research 37, 4863e4872. Li, F., Yuasa, A., Katamine, Y., Tanaka, H., 2007. Breakthrough of natural organic matter from fixed bed adsorbers: investigations based on size-exclusion HPLC. Adsorption 13, 569e577. Matilainen, A., Vieno, M., Tuhkanen, T., 2006. Efficiency of the activated carbon filtration in the natural organic matter removal. Environment International 32, 324e331. Matsui, Y., Knappe, D.R.U., Iwaki, K., Ohira, H., 2002. Pesticide adsorption by granular activated carbon adsorbers. 2. Effects of pesticide and natural organic matter characteristics on pesticide breakthrough curves. Environmental Science & Technology 36, 3432e3438. Newcombe, G., Drikas, M., Hayes, R., 1997. Influence of characterised natural organic material on activated carbon adsorption: II. Effect on pore volume distribution and adsorption of 2-methylisoborneol. Water Research 31, 1065e1073. Newcombe, G., Drikas, M., 1997. Adsorption of NOM onto activated carbon: electrostatic and non-electrostatic effects. Carbon 35, 1239e1250. Newcombe, G., 1999. Charge vs. porosity e some influences on the adsorption of natural organic matter (NOM) by activated carbon. Water Science and Technology 40, 191e198. Newcombe, G., Morrison, J., Hepplewhite, C., 2002. Simultaneous adsorption of MIB and NOM onto activated carbon. I.
3959
characterisation of the system and NOM adsorption. Carbon 40, 2135e2146. Nissinen, T.K., Miettinen, I.T., Martikainen, P.J., Vartiainen, T., 2001. Molecular size distribution of natural organic matter in raw and drinking waters. Chemosphere 45, 865e873. Owen, D.M., Amy, G.L., Chowdhury, Z.K., 1993. Characterization of Natural Organic Matter and Its Relationship to Treatability. American Water Works Association Research Foundation, Denver, CO, USA. Pelekani, C., Snoeyink, V.L., 1999. Competitive adsorption in natural water: role of activated carbon pore size. Water Research 33, 1209e1219. Perdue, E.M., Lytle, C.R., 1983. Distribution model for binding of protons and metal-ions by humic substances. Environmental Science & Technology 17, 654e660. Speitel, G.E., Turakhia, M.H., Lu, C.J., 1989. Initiation of micropollutant biodegradation in virgin gac columns. Journal American Water Works Association 81, 168e176. Summers, R.S., Roberts, P.V., 1988. Activated carbon adsorption of humic substances .2. Size exclusion and electrostatic interactions. Journal of Colloid and Interface Science 122, 382e397. Velten, S., Hammes, F., Boller, M., Egli, T., 2007. Rapid and direct estimation of active biomass on granular activated carbon through adenosine tri-phosphate (ATP) determination. Water Research 41, 1973e1983. Vuorio, E., Vahala, R., Rintala, J., Laukkanen, R., 1998. The evaluation of drinking water treatment performed with HPSEC. Environment International 24, 617e623. Wilkinson, K.J., Balnois, E., Leppard, G.G., Buffle, J., 1999. Characteristic features of the major components of freshwater colloidal organic matter revealed by transmission electron and atomic force microscopy. Colloids and Surfaces A-Physicochemical and Engineering Aspects 155, 287e310.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 6 0 e3 9 6 8
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Distribution and potential significance of a gull fecal marker in urban coastal and riverine areas of southern Ontario, Canada Jingrang Lu a,1, Hodon Ryu a,1, Stephen Hill b, Mary Schoen a, Nicholas Ashbolt a, Thomas A. Edge b, Jorge Santo Domingo a,* a b
Office of Research and Development, U.S. Environmental Protection Agency, 26 W. MLK Dr., Cincinnati, OH 45268, USA National Water Research Institute, Environment Canada, Burlington L7R 4A6, Canada
article info
abstract
Article history:
To better understand the distribution of gull fecal contamination in urban areas of southern
Received 22 December 2010
Ontario, we used gull-specific PCR and qPCR assays against 1309 water samples collected
Received in revised form
from 15 urban coastal and riverine locations during 2007. Approximately, 58% of the water
19 April 2011
samples tested positive for the gull-assay. Locations observed to have higher numbers of
Accepted 3 May 2011
gulls and their fecal droppings had a higher frequency of occurrence of the gull marker and
Available online 13 May 2011
a higher gull marker qPCR signal than areas observed to be less impacted by gulls. Lower gull marker occurrence and lower qPCR signals were associated with municipal wastewater
Keywords:
(7.4%) and urban stormwater effluents (29.5%). Overall, there were no statistically significant
Microbial source tracking
differences in gull marker occurrence at beach sites for pore water, ankle, and chest-depth
Gull-targeted fecal assay
samples, although signals were generally higher in interstitial beach sand pore water and
Fecal contamination
ankle-depth water than in chest-depth water samples. Overall, the results indicated that
Lake Ontario
gull fecal pollution is widespread in urban coastal and riverine areas in southern Ontario and that it significantly contributes to fecal indicator bacterial loads. Published by Elsevier Ltd.
1.
Introduction
A large number of Great Lakes beaches in North America (49% in Canada and 73% in the United States) had swimming advisories, postings, or closures during 1998e2007, significantly impacting local economies (Environment Canada and USEPA, 2005). Diverse fecal sources could contribute to these beach advisories, including point sources (e.g., municipal wastewater outfalls) and non-point sources (e.g., agricultural runoff and wildlife), in particular, waterfowl (Edge and Hill, 2007). From a public health perspective, prevention of waterfowl pollution may be important as several studies have shown that waterfowl excrete human waterborne pathogens (Baudart et al., 2000; Makino et al., 2000; Kullas et al., 2002; Slodkowicz-Kowalska et al., 2006; Waldenstro¨m et al., 2002; * Corresponding author. Tel.: þ1 513 569 7085; fax: þ1 513 569 7328. E-mail address:
[email protected] (J.S. Domingo). 1 These authors contributed equally to this study. 0043-1354/$ e see front matter Published by Elsevier Ltd. doi:10.1016/j.watres.2011.05.003
Zhou et al., 2004). Aquatic birds are also natural reservoirs of influenza viruses (Krauss et al., 2007) and therefore are an important link in the evolution and environmental dispersal of these viruses. Modeling of recreational waters with negligible human fecal contamination suggests one-to-two orders of magnitude lower gastrointestinal illness risk from seagullimpacted sites at current water quality criteria (Schoen and Ashbolt, 2010; Soller et al., 2010). Yet such assessments rely on fecal indicator bacteria (FIB) numbers, pathogen data for excreta and a very limited amount of waterborne pathogen data, most of which have been measured using widely criticized culture-based techniques. While gulls have been implicated as primary sources of fecal contamination in the Great Lakes (Edge and Hill, 2007), the relative abundance of their fecal inputs in environmental waters has not been accurately
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 6 0 e3 9 6 8
assessed due to the lack of quantitative host-specific techniques. In a recent study, Lu et al. (2008) developed a PCR assay targeting the 16S rRNA gene of Catellicoccus marimammalium. Thus far, this assay has shown high specificity toward gull feces and has generated positive signals from water samples collected in various locations with a history of gull fecal pollution (Lu et al., 2008; Shibata et al., 2010). However, data on the prevalence and distribution of the bacterial species targeted by this assay in environmental waters is scarce. Moreover, the relationship between this assay and FIB such as Escherichia coli and enterococci is poorly understood. The main goal of this study was to further evaluate the gull marker assay by studying the prevalence of the proposed marker within a geographic location receiving different sources of pollution, including gull feces. Additionally, the significance of gull contamination was studied by assessing the abundance of the gull marker in relation to the presence of fecal bacteria and observations of gull fecal dropping impacts.
2.
Materials and methods
2.1.
Study sites and sampling
A total of 1309 water and wastewater samples were collected between May and October of 2007 from 12 locations (50 sublocations; Table S1) around the cities of Toronto, Ottawa, and Hamilton, Canada, and challenged against the gull marker assay. Samples were collected in sterile 500 ml bottles, and returned on ice to the Burlington lab for filtration (0.45 mm) within 6 h of collection. Water samples at beaches were collected weekly over the bathing season along transects
3961
perpendicular to the shore from interstitial sand pore water (from a hole dug in foreshore sand), and by wading out to collect surface water at ankle and chest-depth. Samples were collected from Toronto locations at Ashbridges Bay Sewage Treatment Plant (AHS), Bluffers Park (BL), Don River (DON), Humber River (HUM), Kew Beach (KW), Marie Curtis Park (MC), Rouge Park (RG), and the Western (Sunnyside) Beaches (WB); from Hamilton locations at Bayfront Park (BP), Eastport (EP), Hamilton harbor (HH); and from Ottawa locations near Petrie Island (PI) (Fig. 1). Observations of the number of gulls and their fecal droppings were made at water sampling locations in an attempt to provide a qualitative assessment of low-to-high impacts from gull fecal droppings. This assessment was more rigorous at beach locations based on previous microbial source tracking studies (Edge et al., 2007a, 2007b, 2010) and weekly enumeration of the numbers of gulls and their fecal droppings each time water samples were collected. The number of gulls was counted in the immediate vicinity of the sampling location, and the number of gull fecal droppings was enumerated by walking along the shoreline near the sampling location and counting fresh droppings on the foreshore sand within 2 m of the waterline. Since beaches varied slightly in length, the gull fecal dropping results were standardized to 100 m of beach shoreline, and the mean number of droppings from weekly observations at a beach location was used to calculate an estimate of the total cumulative number of gull fecal droppings at that beach location from each day over the bathing season sampling period from May to September.
2.2.
Molecular methods
All water samples were processed (100e300 ml) as previously described (Lu et al., 2008) with the following modifications.
Fig. 1 e Sites used in this study. Labels are described in the Material and Method section.
3962
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 6 0 e3 9 6 8
DNA extracts (2 ml) were used as template in SYBR green-based gull PCR assays, which were performed using a 7900 HT Fast Real-Time Sequence Detector (Applied Biosystems). Gullspecific PCR data were analyzed using ABI’s Sequence Detector software (version 2.2.2) and a 0.2 threshold. PCR signals were recorded as presence/absence data and signal quantity values. Disassociation curves were examined to determine the presence of potential primeredimers and other non-specific reaction products. Data points with artifacts (e.g., double peaks) that resulted in signal overestimation were not used in statistical analyses. Signal intensity values were recorded for those reactions showing one corresponding amplification peak within the disassociation curves. Serial dilutions of C. marimammalium DNA (1 nge10 fg) in duplicate were used to generate a standard curve. Two no-template controls per PCR plate were used to check for crosscontamination. A ten-fold dilution of each DNA extract was used to test for PCR inhibition. Real-time PCR (qPCR) units were calculated as fg/100 ml of filtered water sample. PCR products were visualized using 2% agarose gel electrophoresis and GelStar as the nucleic acid stain (FMC BioProducts, Rockland, ME) to confirm the size of amplification products. Ten randomly selected water samples that tested positive for the presence of the gull-targeted marker were used to develop clone libraries to examine the identity and molecular diversity of environmental products (Lu et al., 2008). General Bacteroides PCR-based assay (Bac32) was tested against water DNA extracts using 32F and 708R primers (Bernhard and Field, 2000.) and the cycling conditions described by Lamendella et al. (2007). E. coli (EC) counts were determined by membrane filtration as described by Edge and Hill (2007); where two replicate samples were collected, the counts were expressed as the mean in CFU/100 ml units.
2.3.
Statistics
Statistical analyses were performed using Statistical Analysis Software v8.2 (SAS Institute Inc., Cary, NC). Logistic analysis was used to predict the probability of occurrence of gulltargeted marker (response variable) to assess the relative importance of effect variables such gull-impacted (GI) and fecal indicator data such as E. coli counts (EC) and presence of
general Bacteroides (Bac32). The Wald chi-square test was used to determine if relationships between the gull-targeted marker and effect variables were statistically significant ( p 0.05). Quantitative analysis was conducted using the PROC GLM analysis with the F-statistical significance test at a ¼ 0.05 between the response variable (Gull: gull qPCR assay). The Logistic model was conducted using the odds ratio which is the ratio of presence to absence of gull marker between two levels of a categorical variable such as Bac32 and gull impacts, or numerical variables such as EC. Point estimate (PE) obtained from the Logistic analysis is the coefficient of the prediction model and used to determine the relationship between gull marker and predictor variables. Multiple comparisons (contrast) between the gull impact (GI) levels (L, M, and H) were also conducted. Duncan’s multiple range test, Tukey’s Studentized Range (HSD) Test, Bonferroni (Dunn) t-Tests and Scheffe’s Test were used to assess if qPCR data for the gulltargeted assay was statistically different among different sites. Correspondence analysis with two dimensions was performed to determine the association between the quantity of gull-targeted assay and other independent variables (categories of sub-locations, GI and FIB). In addition, cluster analysis was used to further assess the value of the gull marker assay in classifying gull fecal contamination at study sites.
3.
Results
Of the 1309 samples tested, 58% of the samples were positive for the gull marker and the identity of PCR products was confirmed by sequencing analysis (Table 2 and Table S1). The mean frequency of occurrence of the gull marker in sublocations ranged from 0 to 77% of the total samples collected across all 12 sampling locations. The gull marker was detected at almost all sampling locations, but there were significant differences among the categories of sites ( p < 0.0001). Specifically, the gull marker was most often detected at beach and urban lake shoreline sites where relatively high numbers of gulls and gull fecal droppings were also observed (i.e., medium and high GI). For example, the Eastport (EP) location was immediately adjacent to a large colony of gulls in Hamilton Harbor, and higher numbers of gulls were observed at
Table 1 e Identity of clone sequences from Ontario beach water samples to original gull marker. Sample type
Beach water, Ontario
Sampling locationsa
Sublocations
MC
W E 1 2 4 1 2 3 53E
WB
KW
RG Total a See Fig. 1 for identification of locations.
Number of sequences a
Observed gull impacts a
Total
100% identity
99% identity
23 23 23 24 22 24 24 23 24 210
16 15 13 20 11 13 12 15 18 133
7 8 10 4 11 11 12 8 6 77
Moderate Moderate High High High High High High High
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 6 0 e3 9 6 8
Bluffers, Western and Rouge Beaches (mean ¼ 108, 90, and 88 gulls per sampling day respectively) than Petrie Island, Marie Curtis, and Kew Beaches (mean ¼ 41, 40, and 37 gulls per sampling day respectively). In addition, two of three samples from Bayfront Park Beach were positive for the gull marker and associated with high qPCR values (6.13 Log10 mean value; data not shown). These sites are known to be heavily impacted by gulls throughout the bathing season. Similarly, the number of gull fecal droppings over the sampling season at beach locations (per 100 m shoreline) was estimated to range from as high as 10,422 droppings (Rouge Beach) and 6044 droppings (Bluffers Beach) to as low as about 254 droppings (Kew Beach) and 272 droppings (Marie Curtis Beach). In contrast, a significantly lower occurrence of the gull-targeted marker ( p < 0.05) was associated with municipal wastewater effluents and stormwater outfalls. At individual beaches, the differences in gull marker occurrence between transects (i.e., sand pore water, ankle, and chest samples) were not significant ( p ¼ 0.233, n ¼ 1009) (Fig. 2). The mean gull marker qPCR signals across categories of sampling locations ranged from 0 to 1901 pg/100 ml with beaches and urban lake shorelines yielding significantly higher qPCR signals than most creek, river, stormwater or municipal wastewater samples (Table 2). The highest mean levels of the gull marker
3963
were measured at beaches where higher numbers of gulls and their fecal droppings were observed. The lowest mean values of the gull marker were measured at sites where there were very few observations of gulls or their fecal droppings in the immediate sampling vicinity such as municipal wastewater effluents, stormwater outfalls, and some creek/river sites. Across all beach transects in which samples were taken in sand pore water, and at ankle and chest-depths, gull qPCR signals were significantly different based on ANOVA results ( p ¼ 0.015, n ¼ 1009). Pairwise tests showed that qPCR signal intensity for sand pore water samples was significantly higher than for chest-depth water samples ( p ¼ 0.009), but not significantly different from ankle-depth water samples ( p ¼ 0.771). There were significant differences in signal intensity between ankle- and chest-depth water samples at beaches (BL and WB) with the highest observed gull impacts. Logistic analysis provided additional insights on the geographical distribution of the gull marker. Overall, the presence of the gull marker was significantly associated with observed gull fecal impacts and fecal indicators (EC and Bac32) (Table 3). Sites with observations of a high gull impact (GI ¼ H ) were 4.7 times more likely to be positive for the gull marker than those with observations of a low gull impact (GI ¼ L), while only 2.5 times when compared to moderately-impacted
Fig. 2 e Distribution and multiple probability test of gull-targeted marker in beach sand and different water depth zones at Ontario beaches. Light bars represent the percentage (mean values) of samples positive for gull-targeted marker (as determined by percent from total samples). Darker bars represent the qPCR results (mean values) for the samples. Samples were grouped by collection location along a beach transects: sand interstitial pore water (P); ankle-depth zone (A); chestdepth zone (C). Asterisk indicates differences are statistically significant at a 5% significance level.
3964
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 6 0 e3 9 6 8
Table 2 e Overall occurrence and quantity of gull marker for beach, creek/river and municipal wastewater sampling locations. Categories Beach
Creek/river
Municipal wastewater Stormwater Urban lake shoreline
Locationa Number of samples BL KW MC PI RG WB BL DON HUM MC PI RG AHS
125 190 137 196 168 190 16 17 88 25 17 10 22
BL HUM EP KW
31 13 4 47
Positive gull signals (%)
Log10 mean gull qPCR Average LOG gull qPCR value values (STD) of each categoryb
102 (81.6%) 101 (53.2%) 83 (60.6%) 80 (40.8%) 118 (70.2%) 145 (76.3%) 4 (25.0%) 7 (41.2%) 51 (58.0%) 11 (44.0%) 5 (29.4%) 3 (30.0%) 2 (9.1%) 9 4 4 24
(29.0%) (30.8%) (100.0%) (51.1%)
6.44 (4.6) 2.93 (3.9) 3.44 (4.3) 1.93 (3.3) 4.77 (4.4) 5.63 (4.7) 1.52 (3.2) 0.87 (2.1) 2.74 (3.4) 2.34 (3.5) 1.29 (2.7) 1.37 (2.7) 0
4.47
1.07 (2.6) 1.23 (2.8) 7.55 (0.3) 2.22 (3.4)
1.15
Observed gull impact level High Low w Moderate Low w Moderate Low w Moderate Moderate w High High Moderate Low w Moderate Low Moderate Moderate Moderate Low
1.69
BDL
Low Low High Low w High
4.89
a See Fig. 1 for identification of locations. b Average was estimated using data not shown for sites (n ¼ 4) for which 3 samples were collected.
sites (GI ¼ M ). The analysis also showed that the occurrence of the gull-targeted marker was positively associated with Bac32 detection. For example, the odds of gull marker detection at sites negative to Bac32 (Bacteroides) were 0.256 times less likely than at sites where Bac32 was detected. With respect to E. coli counts, and with all other parameters constant, a natural log increase in E. coli count corresponded to about a 1.2 times increase in the gull marker assay. To further examine potential associations among variables, correspondence analysis was performed using two-dimension Chi-square data analysis for site categories (Fig. 3). Correspondence analysis showed that Chi-square values were significantly different across site categories. Over 73% of the total Chi-square was explained by one dimension (i.e., the horizontal dimension), indicating that association of
the variables over site categories was dominated by gull contamination as determined by the detection of the gull marker. The gull marker was strongly associated with beaches and urban lake shoreline sites. In general, creek/river and stormwater sites were poorly associated with the gull marker. Quantitatively, general linear analysis showed that gull marker was positively associated significantly with observed gull impacts (GI, p < 0.0001), E. coli counts (EC, p < 0.0001) and Bac32 ( p ¼ 0.0032). Comparisons of gull marker signal between GI levels were also highly significant ( p < 0.0001 for all three pairs of comparisons). There was about one natural log scale increase of gull marker quantity from one GI level to another (L to M or M to H; calculated in Least Squares Means). Based on gull marker quantity, cluster analysis was conducted to see whether gull-assay signal levels could be used as
Table 3 e Results from statistical logistic analyses performed on presence of PCR gull marker assay signals (sample size n [ 1254). Analysis of effectsa
Effect variables Fecal source Fecal indicator
Odds Ratio Estimates
Observed gull impact E. coli count Bac32
Degree freedom (df)
Wald Chi-Squarea
Pr > ChiSq (p value)
2 1 1
62.68 12.46 35.14
<0.0001 0.0004 <0.0001
Effect variable Fecal source: Observed Gull Impact Fecal indicator
Low vs High Moderate vs High E. coli counts Bac32: absence vs presence
Point Estimate (PE)b 0.213 0.408 1.195 0.256
Wald Confidence Limitsc 0.123 0.310 1.082 0.163
0.369 0.535 1.319 0.401
a Relationship between the gull-targeted marker and effect variables was statistically significant when Wald Chi-Square is >12.71 (df ¼ 2) or 4.30 (df ¼ 1). b Lower PE values for observed gull impacts mean a greater difference between variables. The higher PE values for E. coli counts when compared to presence/absence of Bac32 data suggests a weaker positive relationship between gull marker and E. coli than gull marker and Bac32. c Lower and upper confidence limits used to determine if estimated PE was within confidence levels at 95%.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 6 0 e3 9 6 8
3965
Fig. 3 e Correspondence analysis showing associations among categories of sampling locations (Beaches, Creek/River, Municipal wastewater (M-Wastewater), Parking Lot runoff (PL-Runoff), Stormwater, Urban lake shoreline: UL-Shorline), observed gull impact (GI) and fecal indicators (gull marker qPCR signals: CM; presence/absence of general Bacteroides marker: BAC32; E. coli counts: EC). The first circled cluster indicates the higher association between gull marker, presumed GI, and UL-Shoreline, Beaches and PL-Runoff locations. The second circled cluster indicates a lesser association between gull marker fecal indicators (BAC32 and EC) and locations like Creek-Rivers and Stormwater location. M-Wastewater location showed a poor associated with the gull marker while a stronger association with fecal indicators EC and BAC32.
a classification for the extent of gull fecal contamination. Sublocations with observations of high gull fecal impacts grouped together, most of which were located at beaches and urban lake shorelines (Fig. 4). Sub-locations with observations of moderate-to-low gull impacts grouped within corresponding categories.
4.
Discussion
With over 1300 samples tested, this represents the first largescale study involving the field application of the C. marimammalium (gull-targeted) qPCR assay. The samples were collected from various urban locations around the cities of Toronto, Ottawa, and Hamilton, providing the opportunity of studying the distribution and prevalence of the gull marker in urban areas in proximity to large colonies of gulls. There are large ring-billed gull colonies on the Toronto and Hamilton waterfronts, with approximately 55,000 breeding pairs located at Tommy Thompson Park in Toronto. In general, we showed that the gull marker was detected across all sampling locations, suggesting that gull fecal contamination is widespread and highly prevalent in these urban areas. The occurrence of the gull marker in water samples from stormwater outfalls and parking lot runoff indicates the presence of gull fecal contamination in runoff from impervious surfaces in urban areas. Sequence analysis of 210 clones from nine different water types showed that the sequences generated with the gull qPCR assay were identical or nearly identical (sequence
identity 99%) to the C. marimammalium 16S rRNA gene (Table 1) confirming the identity of the PCR signals. The lower occurrence of the gull marker in municipal wastewater and other areas with observations of lower numbers of gulls and their fecal droppings provided additional evidence suggestive of the specificity of the gull-targeted assay. Specifically, the results revealed the following features for the gull-targeted assay: (1) the occurrence of the gull marker was positively associated with observed gull impacts (gull numbers or gull fecal droppings); (2) the association between the levels of the gull marker (i.e., intensity of qPCR signals) and observed gull impacts was statistically significant; (3) statistical analysis showed that there was a positive association between gull signals and fecal indicators (E. coli counts and occurrence of Bacteroidetes). Altogether, the results from this study suggest that the gull-assay is a good predictor of the presence of gull fecal contamination and, to some extent, of gull fecal pollution levels. This was particularly noticeable as the highest levels of the gull marker were measured at sampling locations where the highest numbers of gulls and their fecal droppings were observed (e.g. Eastport gull colony location in Hamilton, and Bluffers Park Beach and Western Beaches in Toronto). Lower levels of the gull marker were measured in wastewater samples and at beaches observed to be less impacted by gulls (e.g. Petrie Island and Kew Beaches). An important objective in this study was to investigate the prevalence of the gull marker across sampling locations frequented by varying numbers of gulls. We found that the gull marker was more prevalent at sites with higher observations of gulls and their fecal droppings; however, additional
3966
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 6 0 e3 9 6 8
Fig. 4 e Cluster analysis for sampling sub-locations at Lake Ontario based on gull-targeted qPCR signal intensity.
research will be required to evaluate the ability to use the gull marker or bird observations to predict E. coli or other FIB concentrations in specific water samples. It should be noted that bird numbers and indicator bacteria might not always correlate in waters where there are other fecal sources. This is due to the significant differences in the levels of FIB in different hosts. For example, as noted recent studies one fecal dog sample could represent thousands of gull fecal events when using enterococci densities as the standard (Wright et al., 2009; Wang et al., 2010). Thus a strictly linear correlation between FIB and host-specific markers may not exist for most of the currently available assays used for fecal source identification as they do not use FIB as the targeted population. There are other variables associated with the lag time between bird observations and FIB enumeration in water (i.e., wave action, wind speed/direction, and precipitation) and with bacterial fate (i.e., inactivation and/or die off due to sunlight and desiccation) that could preclude a simple correlation between bird observations and E. coli concentrations. Indeed, several studies have not shown a good quantitative relationship between numbers of birds or bird droppings and E. coli numbers at Great Lakes beaches (Kleinheinz et al., 2006; Edge and Hill, 2007). The incidence of the gull marker at locations observed to have low-to-moderate gull impacts could be influenced by the level of survival of C. marimammalium in environmental waters and secondary habitats such as sediments and sands. Little is known about the ecology of C. marimammalium. In fact, little is known about this organism besides limited
biochemical characterization data and its phylogenetic relatedness with catalase-negative genera such as Enterococcus, Melissococcus, Tetragenococcus and Vagococcus (Enterococaceae family) (Lawson et al., 2006). Other fecal members of the Enterococaceae family are presumed to survive longer in water than other fecal bacteria such as members of the Enterobacteriaceae and Bacteroidetes (Fiskal et al., 1985; Sinton et al., 1998; Noble et al., 2003; Haller et al., 2009). The presence of C. marimammalium in areas of presumed low gull fecal pollution suggest that this organism may survive in the environment to some extent, On the other hand, host-specificity data (Lu et al., 2008) suggest that the gull gastrointestinal tract is a preferred habitat of C. marimammalium and it is also possible that its occurrence in environmental waters is the product of a recent contamination event. Future studies need to be conducted to better understand the survival potential and seasonal variation of this bacterial species in environmental waters. Similarly, fecal bacteria have been isolated in secondary habitats, implicating them as potentially important reservoirs of fecal pathogens. No data on the occurrence of C. marimammalium in sediments or sand has been reported, highlighting a general area that needs to be addressed in future studies if the assay is going to be used effectively as a source tracking tool. The presence and levels of gull fecal pollution in the areas studied, as determined by the gull marker assay, can partly explain the levels of E. coli and Bacteroidetes detected. In addition to gull feces, sewage discharges, urban runoff, and other waterfowl (i.e., Canada geese) are possible pollution
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 6 0 e3 9 6 8
sources. Indeed, Canada geese were also commonly observed on many of the beaches examined in this study. Interestingly, Bacteroidetes represented 2e18% of the sequences in geese 16S rRNA gene clone libraries, while only 1% in gull counterparts (Lu et al., 2008, 2009). While avian feces tend to harbor lower densities of Bacteroidetes than mammal feces, some of the bacterial populations detected by the Bac32 assay could be of geese origin in those areas in which geese are important pollution sources. Although a Bacteroidetes-based marker was recently developed as a goose-specific assay (Fremaux et al., 2010), to this date there is no Bacteroidetes-based marker for gulls, making it difficult to understand the contribution of each avian species to this bacterial order. Additionally, differences in Bacteroidetes composition may exist, even within the same bird type. For example, Jeter et al. (2009) showed considerable differences in Porphyromonadaceae abundance in gull feces collected from different locations around Lake Michigan. Similar findings were observed for Prevotella and Bacteroides spp. in gull samples collected from other locations in or nearby the Great Lakes (Lu et al., 2009). These issues are not exclusive to efforts tracking waterfowl sources, and highlight the difficulties in developing source tracking markers that can correlate with fecal indicator bacteria, particularly when the targeted populations belong to different bacterial groups. Also, for MST tools to be more useful in quantitative microbial risk assessments, markers should also be correlated to reference pathogens of concern.
5.
Conclusions
Several lines of evidence suggest that gull feces can play an important role in contaminating urban waters in areas near where these birds establish large colonies. The pollution of southern Ontario urban coastal and riverine waters was characterized by: (1) the widespread detection of a putative gull marker at almost all sampling sites; (2) positive association between E. coli counts and presence of Bacteroidetes with the occurrence of the gull marker; (3) correspondence analysis showed that other sources might also be contributing to the overall levels of E. coli and Bacteroidetes, while cluster analysis indicated the gull-assay could be used for the identification of locations with significant impacts from gull fecal contamination. The gull-targeted assay may be useful in a ‘waterfowl pollution toolbox’. However, additional studies are needed to better understand the ecology of the targeted population and the value of this assay in quantitative risk analysis models. In this regard, when mixed sources are present in recreational waters for which high gull contamination is expected and a low percentage of the indicator bacteria are coming from human sources, it would be of practical value if a gull-assay can demonstrate that most of the other fecal indicator bacteria are coming from gulls given the significantly lower risks associated for such scenarios. Future studies should be performed to better establish the ratio between fecal indicator bacteria, pathogens and source tracking markers using a large number of fecal samples collected from different geographic locations.
3967
Acknowledgments This research was funded in part by a New Start Award from the EPA National Center for Computational Toxicology to J.S.D, and funding from the Cities of Toronto and Ottawa, and Environment Canada’s STAGE genomics program to T.A.E. The U.S. Environmental Protection Agency, through its Office of Research and Development, funded and managed, or partially funded and collaborated in, the research described herein. It has been subjected to the Agency’s administrative review and has been approved for external publication. Any opinions expressed in this paper are those of the author (s) and do not necessarily reflect the views of the Agency, therefore, no official endorsement should be inferred. Any mention of trade names or commercial products does not constitute endorsement or recommendation for use.
Appendix. Supplementary material Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.watres.2011.05.003.
references
Baudart, J., Lemarchand, K., Brisabois, A., Lebaron, P., 2000. Diversity of Salmonella strains isolated from the aquatic environment as determined by serotyping and amplification of the ribosomal DNA spacer regions. Applied and Environmental Microbiology 66 (4), 1544e1552. Bernhard, A.E., Field, K.G., 2000. A PCR assay to discriminate human and ruminant feces on the basis of host differences in Bacteroides-Prevotella genes encoding 16S rRNA. Applied and Environmental Microbiology 66 (10), 4571e4574. Edge, T.A., Hill, S., 2007. Multiple lines of evidence to identify the sources of fecal pollution at a freshwater beach in Hamilton Harbour, Lake Ontario. Water Research 41 (16), 3585e3594. Edge, T.A., Crowe, A., Hill, S., Seto, P., Marsalek, J., 2007a. Surveillance for Potential Sources of E. Coli at Toronto’s Bluffers Park Beach 2005e06. Environment Canada, Burlington, Ontario. WSTD Technical Note No. AEP-TN07-003. Edge, T.A., Hill, S., Seto, P., Marsalek, J., 2007b. Experience with the antibiotic resistance analysis and DNA fingerprinting in tracking faecal pollution at two lake beaches. Water Science Technology 56 (11), 51e58. Edge, T.A., Hill, S., Seto, P., Marsalek, J., 2010. Library-dependent and library-independent microbial source tracking to identify spatial variation in faecal contamination sources along a Lake Ontario beach (Ontario, Canada). Water Science Technology 62 (3), 719e727. Environment Canada and USEPA, 2005. State of the Great Lakes 2005. In: State of the Lakes Ecosystem Conference. Environment Canada, Quebec, Canada, and U.S. Environmental Protection Agency, Washington, DC. 15e19. Fiksdal, L., Maki, J.S., LaCroix, S.J., Staley, J.T., 1985. Survival and detection of Bacteroides spp., prospective indicator bacteria. Applied and Environmental Microbiology 49 (1), 148e150. Fremaux, B., Boa, T., Yost, C.K., 2010. Quantitative real-time PCR assays for sensitive detection of Canada goose-specific fecal pollution in water sources. Applied and Environmental Microbiology 76 (14), 4886e4889.
3968
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 6 0 e3 9 6 8
Haller, L., Amedegnato, E., Pote´, J., Wildi, W., 2009. Influence of freshwater sediment characteristics on persistence of fecal indicatorbacteria.Water,Air,andSoilPollution203(1e4),217e227. Jeter, S.N., McDermott, C.M., Bower, P.A., Kinzelman, J.L., Bootsma, M.J., Goetz, G.W., McLellan, S.L., 2009. Bacteroidales diversity in ring-billed gulls (Laurus delawarensis) residing at Lake Michigan beaches. Applied and Environmental Microbiology 75 (6), 1525e1533. Kleinheinz, G.T., McDermott, C.M., Chomeau, V., 2006. Evaluation of avian waste and bird counts as predicators of Escherichia coli contamination at Door County, Wisconsin beaches. Journal of Great Lakes Research 32 (1), 117e123. Krauss, S., Obert, C.A., Franks, J., Walker, D., Jones, K., Seiler, P., Niles, L., Pryor, S.P., Obenauer, J.C., Naeve, C.W., Widjaja, L., Webby, R.J., Webster, R.G., 2007. Influenza in migratory birds and evidence of limited intercontinental virus exchange. PLoS Pathogens 3, e167. Kullas, H., Coles, M., Rhyan, J., Clark, L., 2002. Prevalence of Escherichia coli serogroups and human virulence factors in faeces of urban Canada geese (Branta canadensis). International Journal Environmental Health Research 12 (2), 153e162. Lamendella, R., Santo Domingo, J.W., Oerther, D.B., Vogel, J.R., Stoeckel, D.M., 2007. Assessment of fecal pollution sources in a small northern-plains watershed using PCR and phylogenetic analyses of Bacteroidetes 16S rRNA gene. FEMS Microbiology Ecology 59 (3), 651e660. Lawson, P.A., Collins, M.D., Falsen, E., Foster, G., 2006. Catellicoccus marimammalium gen. nov., sp. nov., a novel Gram-positive, catalase-negative, coccus-shaped bacterium from porpoise and grey seal. International Journal of Systematics and Evolutionary Microbiology 56 (2), 429e432. Lu, J., Santo Domingo, J.W., Lamendella, R., Edge, T., Hill, S., 2008. Phylogenetic diversity and molecular detection of gull feces. Applied and Environmental Microbiology 74 (13), 3969e3976. Lu, J., Santo Domingo, J.W., Hill, S., Edge, T.A., 2009. Microbial diversity and host-specific sequences of Canadian goose feces. Applied and Environmental Microbiology 75 (18), 5919e5926. Makino, S., Kobori, H., Asakura, H., Watarai, M., Shirahata, T., Ikeda, T., Takeshi, K., Tsukamoto, T., 2000. Detection and characterization of Shiga toxin-producing Escherichia coli from seagulls. Epidemiology and Infection 125 (1), 55e61. Noble, R.T., Moore, D.F., Leecaster, M.K., McGee, C.D., Weisberg, S.B., 2003. Comparison of total coliform, fecal coliform, and
enterococcus bacterial indicator response for ocean recreational water quality testing. Water Research 37 (7), 1637e1643. Schoen, M.E., Ashbolt, N.J., 2010. Assessing pathogen risk to swimmers at non-sewage impacted recreational beaches. Environmental Science and Technology 44 (7), 2286e2291. Shibata, T., Solo-Gabriele, H.M., Sinigalliano, C.D., Gidley, M.L., Plano, L.R., Fleisher, J.M., Wang, J.D., Elmir, S.M., He, G., Wright, M.E., Abdelzaher, A.M., Ortega, C., Wanless, D., Garza, A.C., Kish, J., Scott, T., Hollenbeck, J., Backer, L.C., Fleming, L.E., 2010. Evaluation of conventional and alternative monitoring methods for a recreational marine beach with nonpoint source of fecal contamination. Environmental Science and Technology 44 (21), 8175e8181. Sinton, L.W., Finlay, R.K., Hannah, D.J., 1998. Distinguishing human from animal faecal contamination in water: a review. New Zealand Journal Marine and Freshwater Research 32 (2), 323e348. Slodkowicz-Kowalska, A., Graczyk, T.K., Tamang, L., Jedrzejewski, S., Nowosad, A., Zduniak, P., Solarczyk, P., Girouard, A.S., Majewska, A.C., 2006. Microsporidian species known to infect humans are present in aquatic birds: implications for transmission via water? Applied and Environmental Microbiology 72, 4540e4544. Soller, J.A., Schoen, M.E., Bartrand, T., Ravenscroft, J.E., Ashbolt, N.J., 2010. Estimated human health risks from exposure to recreational waters impacted by human and non-human sources of faecal contamination. Water Research 44 (7), 4674e4691. Waldenstro¨m, J., Broman, T., Carlsson, I., Hasselquist, D., Achterberg, R.P., Wagenaar, J.A., Olsen, B., 2002. Prevalence of Campylobacter jejuni, Campylobacter lari, and Campylobacter coli in different ecological guilds and taxa of migrating birds. Applied and Environmental Microbiology 68 (12), 5911e5917. Wang, J.D., Solo-Gabriele, H.M., Abdelzaher, A.M., Fleming, L.E., 2010. Estimation of enterococci input from bathers and animals on a recreational beach using camera images. Marine Pollution Bulletin 60 (8), 1270e1278. Wright, M.E., Solo-Gabriele, H.M., Elmir, S., Fleming, L.E., 2009. Microbial load from animal feces at a recreational beach. Marine Pollution Bulletin 58 (11), 1649e1656. Zhou, L., Kassa, H., Tischler, M.L., Xiao, L., 2004. Host-adapted Cryptosporidium spp. in Canada geese (Branta canadensis). Applied and Environmental Microbiology 70 (7), 4211e4215.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 6 9 e3 9 8 0
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
The effect of UV/H2O2 treatment on disinfection by-product formation potential under simulated distribution system conditions D.H. Metz a,b,*, M. Meyer a, A. Dotson c, E. Beerendonk d, D.D. Dionysiou b a
Greater Cincinnati Water Works, 5651 Kellogg Avenue, Cincinnati, OH, USA University of Cincinnati, Department of Civil and Environmental Engineering, 765 Baldwin Hall, Cincinnati, OH, USA c University of Alaska Anchorage, Civil Engineering Department, Engineering, Building Room 214, Anchorage, AK, USA1 d KWR Watercycle Research Institute, P.O. Box 1072, 3430 BB Nieuwegein, The Netherlands b
article info
abstract
Article history:
Advanced oxidation with ultraviolet light and hydrogen peroxide (UV/H2O2) produces
Received 17 February 2011
hydroxyl radicals that have the potential to degrade a wide-range of organic micro-
Received in revised form
pollutants in water. Yet, when this technology is used to reduce target contaminants,
2 May 2011
natural organic matter can be altered. This study evaluated disinfection by-product (DBP)
Accepted 3 May 2011
precursor formation for UV/H2O2 while reducing trace organic contaminants in natural
Available online 8 May 2011
water (>90% for target pharmaceuticals, pesticides and taste and odor producing compounds and 80% atrazine degradation). A year-long UV/H2O2 pilot study was conducted
Keywords:
to evaluate DBP precursor formation with varying water quality. The UV pilot reactors were
UV/H2O2
operated to consistently achieve 80% atrazine degradation, allowing comparison of low
Advanced oxidation
pressure (LP) and medium pressure (MP) lamp technologies for DBP precursor formation.
Disinfection by-product formation
Two process waters of differing quality were used as pilot influent, i.e., before and after
potential
granular activated carbon adsorption. DBP precursors increased under most of the condi-
OH radical
tions studied. Regulated trihalomethane formation potential increased through the
Sustainability
UV/H2O2 reactors from 20 to 118%, depending on temperature and water quality. When
Water treatment
Post-GAC water served as reactor influent, less DBPs were produced in comparison to conventionally treated water. Haloacetic acid (HAA5) increased when conventionally treated water served as UV/H2O2 pilot influent, but only increased slightly (MP lamp) when GAC treated water served as pilot influent. No difference in 3-day simulated distribution system DBP concentration was observed between LP and MP UV reactors when 80% atrazine degradation was targeted. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
UV/H2O2 advanced oxidation is a promising drinking water technology for the reduction of a broad-spectrum of synthetic
organic contaminants and undesirable natural organic constituents. UV/H2O2 advanced oxidation combines direct photolysis and advanced oxidation through indirect photolysis for the destruction of organic compounds in water (Pereira
* Corresponding author. Greater Cincinnati Water Works, 5651 Kellogg Avenue, Cincinnati, OH, USA. E-mail address:
[email protected] (D.H. Metz). 1 Previously from the University of Colorado. 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.05.001
3970
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 6 9 e3 9 8 0
et al., 2007). Medium pressure (MP) and low pressure (LP) UV lamps are commonly used in drinking water treatment plants. MP lamps use more electrical energy, but require less plant area. Both MP and LP lamps emit light at wavelengths that can cause hydroxyl radical formation and photolysis, however the difference in their spectrum (polychromatic versus monochromatic light respectively) may affect various chemical bonds differently, generating different degradation products. Natural organic matter (NOM) in drinking water sources is at least a magnitude of order greater in concentration than most target contaminants. As such, it is important to evaluate NOM degradates and their impact on water quality. Upon UV/H2O2 treatment, researchers have noted a reduction in NOM aromaticity (Thomson et al., 2004; Kleiser and Frimmel, 2000; Toor and Mohseni, 2007 and Sarathy and Mohseni, 2007, 2010), a shift to smaller molecular size (Sarathy and Mohseni, 2009), the creation of more biodegradable compounds (Thomson et al., 2004; Toor and Mohseni, 2007; Sarathy and Mohseni, 2009) and a decrease in hydrophobicity (Sarathy and Mohseni, 2009, 2010). The use of granular activated carbon (GAC) with UV/H2O2 advanced oxidation can be beneficial. GAC removes NOM, particularly the higher molecular weight and more aromatic components. When used after UV/H2O2, GAC can remove unintended degradation products and quench excess hydrogen peroxide. When used before UV/H2O2, GAC adsorption can provide reactor influent water with low NOM and UV absorbance. Lower NOM reduces the UV energy requirements and potentially leaves less NOM for unintended transformation. Also, GAC can promote biodegradation of NOM. Metz et al. (2011) describes the effects of biodegradation with UV/H2O2 under the conditions of this present study. Free chlorine is a commonly used disinfectant in many parts of the world. Since the 1970s it has been known that disinfection by-products (DBPs) with potential health effects were formed when drinking water was chlorinated. Aquatic NOM is complex and while much research has been performed to evaluate chlorination of NOM, all reaction mechanisms during chlorination have not been elucidated. NOM is generally characterized as hydrophobic, transphilic, hydrophilic with acid, base and neutral subdivisions (Croue´ et al., 2006). These operationally defined fractions can provide information relative to DBP precursors. Researchers have determined that the hydrophobic/non-polar NOM accounts for the majority of DBP formation (Liang and Singer, 2003). However, Hwang et al. (2001) found that even though upon chlorination non-polar NOM generally results in more DBPs on an organic carbon basis, polar NOM can produce a significant amount of DBPs, particularly haloacetic acids. Activated aromatic species such as b-dicarbonyl compounds also have been identified as reactive DBP precursors (Dickenson et al., 2008). Many nations have set maximum contaminant levels (MCLs) for the sum of four trihalomethanes (THM) and the sum of five haloacetic acids (HAA5). In 2005, the World Health Organization set acute guidelines for the individual THMs emphasizing the varying toxicity of these compounds. Numerous researchers have studied the reduction of NOM and DBP precursors through UV/H2O2. Matilainen and Sillanpaa (2010) and Bond et al. (2011) have published comprehensive review articles that summarize this NOM and DBP formation
potential work, respectively. Some researchers have found that at lower more practical UV/H2O2 conditions small increases in DBP formation potential can occur (Kleiser and Frimmel, 2000; Toor and Mohseni, 2007; Dotson et al., 2010 and Sarathy and Mohseni, 2010). Although many researchers have reported that at disinfection doses UV does not increase DBP formation, Magnuson et al. (2002) found that UV direct photolysis and UV/H2O2 can alter extracted NOM, increasing disinfection by-product precursors. Mass spectra varied with UV dose ranging from 20 to 140 mJ/cm2, indicating a change in NOM chemical structure. The change in dose also appeared to increase the reactivity of the extracted organic matter with subsequent chlorination. The magnitude of spectral changes was greater for medium pressure than low pressure lamps at equal doses. Kleiser and Frimmel (2000) found that with short irradiation times, the UV/H2O2 process increased THM precursors. Toor and Mohseni (2007) observed that UV/H2O2 was effective in reducing DBP formation potential only at UV doses >1000 mJ/cm2 and at H2O2 doses 23 mg/L. Sarathy and Mohseni (2010) reported that at UV doses between 500 and 2000 mJ/cm2 with 15 mg/L of H2O2, there was a significant reduction in TOC and SUVA that was not matched by the small reductions in DBP precursors. However, when hydrophobic humic acids were reduced, improved reductions were noted. Dotson et al. (2010) reported that at a UV dose of 1000 mJ/cm2 and 10 mg/L H2O2 THM yield was increased. The researchers found that THM yield correlated with hydroxyl radical exposure. A UV/H2O2 pilot-scale study was conducted to evaluate disinfection by-product precursor formation. The research was conducted over a full-year in order to cover seasonal variations and different levels of GAC breakthrough. DBP formation potential was evaluated while operating the pilot plant to meet practical performance goals. Other studies have compared LP and MP lamps for NOM transformation at equal UV doses (Dotson et al., 2010; Magnuson et al., 2002). The objective of this study was to evaluate the impact of UV/H2O2 on THM and HAA5 precursor formation under UV/H2O2 conditions sufficient to achieve advanced oxidation performance goals (i.e., consistently reducing pharmaceutical compounds, pesticides and odor producing compounds by more than 90% and atrazine by 80%). Two plant-process sources with different natural organic composition and concentration were evaluated for UV/H2O2 treatment with LP and MP technologies. Disinfection by-product formation potential was investigated for each unit process in the continuous flow pilot plant, including UV/H2O2 reactors, GAC columns following the UV/H2O2 reactors, and controls for each process.
2.
Methods
2.1.
Facilities
The source water for the UV/H2O2 pilot influent was drawn from two locations within Greater Cincinnati Water Works’ (GCWW) surface water treatment plant. Location one was after coagulation, settling and filtration, i.e., conventional treatment (CONV). Location two was from combined GAC
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 6 9 e3 9 8 0
adsorber effluent (Post-GAC). Plant GAC contact time averaged about 15e20 min during the study. Water entering the GAC facility had a total organic carbon (TOC) averaging 1.9 mg/L; water exiting the facility had a TOC averaging 0.9 mg/L. The GAC facility also served to significantly reduce disinfection byproduct precursors. After becoming exhausted (average combined effluent of 150 days, maximum combined effluent 200 days), the GAC was thermally reactivated. A schematic of the full-scale process train, indicating these pilot influent stream locations is shown in Supplemental Figure 1. The CONV pilot influent water had been coagulated, flocculated and filtered which decreased the TOC concentration by 30% (from 2.5 to 1.7 mg/L), and specific ultraviolet absorbance (SUVA) from 3.4 to 2.6 L/mg-M on average, while GAC adsorption reduced the TOC concentration by 65% (from 2.5 to 0.85 mg/L) of the raw water concentration and reduced SUVA from 3.4 to 1.7 L/mg-M. The Post-GAC water was lower in total organic carbon and generally followed the seasonal trend of the CONV water. However, during the summer months GCWW reactivated the GAC more frequently to produce an average run of 108 days as compared to the annual average of approximately 150 days. The increased reactivations reduced the DBP precursors, so that even during the warmest summers the utility would be able to meet DBP regulations. Likewise, the UV254 absorbance of the Post-GAC water exhibited very little increase in the summer, although the
3971
UV254 absorbance of the CONV water has an increased UV254 value in the warmer temperature months. For most of the year, the Post-GAC water had significantly lower SUVA values than the CONV water. Thus, it can be assumed that the PostGAC water exhibited a different organic character (i.e., less aromatics and conjugated double bonds) and lower organic concentration than the CONV water. During the late winter and early Spring, however, the SUVA values of the two waters were similar. This is when the GAC is most spent, averaging about 200 run days. Fig. 1 depicts the variability of natural organic matter in the two pilot influent waters.
2.2.
Pilot plant design and operation
The pilot unit was in continuous operation from October of 2007 until October 2008. It consisted of a constant head tank, the peroxide and contaminant feed systems, the UV reactors and the GAC columns. Fig. 2 depicts the layout of the pilot including the location of the chemical injection and sampling points. CONV or Post-GAC water was pumped to a constant head tank. The water flow split into two lines before entering the UV reactors. The contaminant solutions and the 8% hydrogen peroxide solution were injected through two inline injection mixers. A 10 mg/L H2O2 and a 2 mg/L atrazine concentration were targeted. The other target contaminants (See supplemental on-line information.) were spiked at
Fig. 1 e Organic constituents in conventionally (CONV) and post-GAC treated water (2007e2008).
3972
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 6 9 e3 9 8 0
CONV Inf.
F1
UV Reactors MP
F2
M4 M3
M5 C4
GAC Control 1
Discharge
Constant-Head Static mixers w/ Tank injection ports
C5
R4
GAC Control 2 R5
Contaminants
PostGAC Inf.
H 2O2
L4
LP
L3
L5
Static mixer _2
Sample point
Fig. 2 e Pilot plant-process schematic at GCWW.
similarly low levels. The peroxide dose was based on preliminary studies performed at KWR, the Netherlands. This peroxide dose insured that the target contaminant reductions were achieved with a reasonable UV dose. The LP reactor (Aquionics model ALT320 TOC reduction range) consisted of eight LP lamps oriented parallel to the central axis and placed equidistantly at about an 11 cm radius from that axis. The MP reactor (Aquionics model Photon II TOC reduction range) consisted of one MP lamp oriented parallel to the flow, and could be operated at 4 power levels ranging from 75 to 100% of the power. The effluent from both UV reactors and the control water (pilot influent water before the hydrogen peroxide injection point but after the contaminant injection point) were pumped to four GAC pilot columns. Two GAC columns were fed by the control water, control columns 1 and 2. The remaining two columns received the effluent of the LP reactor and the MP reactor, respectively. The GAC columns contained reactivated GAC with an empty bed contact time of 15 min. For more specific details of the pilot design and construction see Metz et al. (2011) and the on-line Supplemental Information. During each quarter UV/H2O2 was studied using CONV pilot influent water and Post-GAC pilot influent water. The operation of the system was based on performance. The UV/ H2O2 system was operated to degrade atrazine by 80% through both the LP and MP reactor trains (effecting a degradation of greater than 90% for spiked pharmaceutical compounds, pesticides and odor producing compounds). Atrazine was chosen to calibrate the system because it was not completely destroyed under the practical UV/H2O2 conditions that gave excellent reduction to the other contaminants. Since the UVT254 and TOC concentration of the water varied seasonally, the operational conditions of the UV reactors were adjusted each quarter and during both CONV and Post-GAC pilot
influent phases of the study. When the contaminant spiking phases were completed, the CONV water was used as pilot influent and the reactors were maintained at the 80% atrazine degradation operating conditions. Additionally, several other water quality parameters, such as TOC, bromide and alkalinity were tested at various frequencies through the pilot. The analytical methods for these tests were from the Standard Methods for the Examination of Water and Wastewater (American Public Health Services et al., 1998) (See Supplemental Table 2 in the on-line supplemental information).
2.3. Simulation of distribution system, chlorine quenching and DBP analysis Simulated distribution system (SDS) tests were used to determine DBP formation. Water samples were held at normal plant water conditions and distribution system temperatures for three days to reflect the normal maximum detention of the drinking water distribution system. The samples were held in headspace free, organic-free amber bottles in enclosed, thermally-insulated boxes with finished plant water continually passing through the boxes. The temperature of the water varied seasonally with temperatures ranging from 6 to 29 C. The amount of chlorine added depended upon whether the sample contained hydrogen peroxide. Approximately 2.09 mg of chlorine are required to quench 1 mg of hydrogen peroxide based on stoichiometry. The samples were chlorinated with 2.2e2.3 mg/L of chlorine per mg/L of hydrogen peroxide to produce a slight chlorine residual and then analyzed for chlorine residual. Additional chlorine was added to equal the actual RMTP chlorine dose. At the end of the 3-day period, chlorine residual was determined. Samples not within normal
3973
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 6 9 e3 9 8 0
plant distribution chlorine residuals (0.6e1.2 mg/L) were not considered. At the completion of the 3-day SDS hold period, samples were prepared for DBP analysis. Residual chlorine was quenched as specified by the particular analytical method. Sodium thiosulfate was used to quench the chlorine for THM samples, and ammonium chloride was used to preserve the HAA samples. All samples were stored at 4 C until they were analyzed (within two weeks). Four THMs and five HAAs were measured using USEPA method 542.2 and method 552.2, respectively. Approximately ten percent of the samples were analyzed in duplicate as a quality control check.
3.
Results and discussion
3.1.
Atrazine calibration of the pilot plant
To achieve 80% atrazine degradation throughout the study, UV doses between 800 and 2000 mJ/cm2 were required for the LP reactor and 200e500 mJ/cm2 were required for the MP reactor with 10 mg/L hydrogen peroxide. Atrazine reduction was between 75 and 85% for most of the study quarters, with the only exception being the first quarter of the study when it was measured at 62% for the LP reactor. The reason for the low value during the first quarter was iron fouling, a situation that was remedied for subsequent experiments.
3.2.
Natural organic matter changes
The TOC concentration of the pilot influent water varied over the 12 month study, fluctuating between 1.2 and 2.6 mg/L for the CONV water and 0.6e1.0 mg/L for the Post-GAC water (See Fig. 1). GAC adsorption can greatly reduce the concentration of organic compounds, including NOM. Because the quantity of NOM is reduced and the relative composition is altered, the
CONV Inf l
LP Reactor Ef f l
MP Reactor Ef f l
concentration of DBP precursors is likely reduced. Also, as was mentioned previously, the pilot process included GAC pilot columns after the reactors when CONV water was used as pilot influent. So, GAC adsorption had the potential of reducing DBP precursors, before or after the UV/H2O2 reactors depending on the pilot influent. Fig. 3 represents TOC concentration through the pilot plant, including the effluent of the GAC pilot columns when CONV water was used as the pilot influent. The top three curves depict CONV influent TOC concentration and the two UV/H2O2 reactor effluent TOC values. Westerhoff et al. (1999) found that different methods used to determine the reactivity of NOM to the hydroxyl radical tend to yield similar reactivities for isolated NOM or natural waters and the magnitude of these hydroxyl radical reactivities fell within a narrow range, 2.6 108 to 4.5 108 L M(C) 1s 1. He later refined these constants through direct measurement 1 to 5 108 L M(C) 1s 1 (Westerhoff et al., 2007). Hydroxyl radicals generated by the UV/H2O2 process reacted with the natural organic matter, but only a small percentage of NOM mineralization was achieved at 80% atrazine degradation (Metz et al., 2011). The bottom four curves represent TOC concentrations for the GAC column effluent streams from the LP and MP reactor process trains and the two control GAC columns (See Fig. 3). Typical breakthrough curves were observed for all four GAC pilot column effluent streams. TOC concentration in the GAC effluent streams ranged from 0.2 to 1.6 mg/L over the study period. At the beginning of the GAC pilot column runs, there was excellent TOC removal, and over the first 140e150 days as the GAC became loaded with organics, the effluent TOC concentration exhibited a rising trend, even though the influent TOC concentration was declining. After this point, the GAC effluent TOC concentrations reflected the increases and decreases of the TOC concentration in the GAC influent. However, some TOC removal was observed through all GAC
Control #1 GAC Ef f l
Control #2 GAC Ef f l
LP GAC Ef f l
3.0
TOC (mg/L)
2.5
2.0
1.5
1.0
0.5
0.0
GAC Run Days Fig. 3 e Total organic carbon through pilot with CONV influent.
MP GAC Ef f l
3974
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 6 9 e3 9 8 0
columns during the study period. By run day 220 there was a clear separation in the TOC concentrations of the GAC effluent streams that had received UV/H2O2 pretreatment and those that had not, as shown in Fig. 3. Overall, the GAC effluent following the UV/H2O2 reactors resulted in 8% less TOC concentration than the control GAC effluent streams. After GAC run day 220, the GAC effluent following the UV/H2O2 reactors averaged 16% less TOC concentration than the GAC effluent of the control process streams. Metz et al. (2011) describes the bioactivity of the GAC columns and the biofilm potential differences in the experimental and control process streams. The character of the NOM was altered through UV/H2O2 as indicated by the decline in SUVA through the UV/H2O2 reactors (See Fig. 4). When the CONV water was used as pilot influent, SUVA was reduced on average through the reactors from 2.6 L/mg-M to 2.3 L/mg-M (MP) and 2.2 L/mg-M (LP) and to 1.8 L/mg-M in the control GAC columns. The CONV water that was UV/H2O2 treated and subsequently GAC treated had the lowest SUVA values, averaging 1.4 L/mg-M (See Fig. 4A). During the CONV water phase, the GAC pilot columns following the reactors preferentially removed UV254 absorbable
CONV Inf l
LP Ef f l
CNTRL #1
organics (aromatics and unsaturated chromophores), thus reducing SUVA. At study conditions, UV254 absorbance and SUVA were reduced through the reactors due to the transformation of aromatic groups and conjugated double bond molecules of the NOM. Sarathy and Mohseni (2007, 2009), Thomson, et al. (2004), Kleiser and Frimmel (2000), Toor and Mohseni (2007) and (Gallard and Von Gunten, 2002a, 2002b) noted a reduction of aromaticity of NOM with UV/H2O2 treatment. Sarathy and Mohseni 2007 reported that under advanced oxidation conditions typically applied at drinking water facilities, NOM was not mineralized but was partially oxidized resulting in significant reduction of aromaticity as measured by high performance size exclusion chromatography. Their findings agree with the SUVA results from this present research. In conventional treatment processes SUVA has been shown to correlate well with “activated” aromatic structures for specific waters (aromatic sites substituted with oxygenand nitrogen-containing functional groups, i.e., phenolics and aromatic amines) that constitute the primary sites attacked by chlorine or other oxidants (Norwood et al., 1980; Reckhow et al., 1990; Westerhoff et al., 1999). Even though SUVA can
Cntrl #2
MP GAC EFFL
LP GAC Ef f l
MP Ef f l
GAC Run Days
SUVA (L/m g-M)
135
155
175
195
215
235
255
275
3.0
CONV Influent
2.5
React. Effls.
2.0
Cntrl. GAC Effls. 1.5
React. GAC Effls.
1.0 3/23/08
1.0 4/22/08
5/22/08
MP Effl
6/21/08
LP Effl
7/21/08
8/20/08
Post-GAC Infl
SUVA (L/m g-M)
3.0
2.5
LP React Effl.
2.0
MP React Effl. Post-GAC Influent
1.5
1.0 1/3/08
2/22/08
4/12/08
6/1/08
7/21/08
9/9/08
10/29/08
Fig. 4 e SUVA through pilot plant with CONV pilot influent (5A); with Post-GAC pilot influent (5B).
3975
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 6 9 e3 9 8 0
3.3.
Trihalomethane formation
3.3.1.
Preliminary trihalomethane formation studies
Liu et al. (2003) recommended the use of 0.1 mg/mL catalase to quench the hydrogen peroxide dose because it does not require multiple measurements of chlorine and peroxide residuals. However, this would not be a feasible approach for drinking water plants. It is expected that most plants would use chlorine or GAC to quench the hydrogen peroxide. For those pilot plant samples that did not pass through GAC after H2O2 addition, chlorine was used to quench the H2O2. In order to investigate the effect of quenching hydrogen peroxide by chlorine pilot tests were performed. As described previously, excess hydrogen peroxide was quenched by chlorine in the SDS samples, and additional chlorine was added to achieve plant chlorine doses. Samples taken before H2O2 addition were dosed at typical plant chlorine doses. Chlorine doses in the SDS samples were sufficient to maintain normal distribution system chlorine residual after three days following the chlorine quenching of H2O2. In order to determine the effect of H2O2 addition and chlorine quenching of H2O2 residuals, TTHM 3day SDS concentration was monitored before and after H2O2 addition (before UV exposure) in the pilot influent water. As can be seen in Fig. 5, no differences in TTHM 3-day SDS concentration were observed in pilot influent before or after H2O2 addition, suggesting that H2O2 alone does not transform NOM under the conditions studied. Even though chlorine doses of greater than 20 mg/L were used to quench the 10 mg/L H2O2 in the post H2O2 samples and less than 2 mg/L chlorine were used in the samples without H2O2, no appreciable differences in 3day SDS THM concentrations were observed.
3.3.2. Total trihalomethane formation with conventionally treated pilot influent During the time period that the CONV water served as the pilot influent, TTHM 3-day SDS samples were collected after 100 GAC run days. Large increases in TTHM 3-day SDS concentrations were observed through the UV/H2O2 reactors under study
Pilot Inf l.
Pilot Inf l. + hydrogen peroxide
160
TTHM 3-day SDS (µg/L)
be a relative indicator of DBP formation potential through conventional physical treatment processes, it cannot be assumed that chemical alteration by advanced oxidation would result in the reduction of DBP formation potential. In fact, chemical alteration by UV/H2O2 has the potential to cause the formation of new DBP precursors (Sarathy and Mohseni, 2010; Dotson et al., 2010; Toor and Mohseni, 2007; Magnuson et al., 2002 and Kleiser and Frimmel, 2000). When Post-GAC water was used as pilot influent, no reduction in SUVA was observed through the reactors (See Fig. 4B). The Post-GAC pilot influent was consistently lower (slightly) in SUVA than the reactor effluent streams. This is because much of the UV254 absorbable organic materials (aromatics and conjugated double bonds) and NOM of higher molecular weight were removed through the GAC adsorption process and was not available for transformation by UV/H2O2 (See Fig. 1). This data suggests that the Post-GAC NOM reacted differently to the UV/H2O2 treatment than the CONV NOM. Although, it is important not to conclude too much from this low organic water with TOC and UV254 values approaching the analytical detection limits.
140 120 100 80 60 40 20 0
Fig. 5 e Effect of H2O2 and chlorine quenching on 3-day simulated distribution system (SDS) TTHM.
conditions. The TTHM 3-day SDS concentrations increased through both the MP and LP reactors similarly (See Fig. 6). TTHM formation was controlled in large part by temperature. In April and September the UV254 absorbance of CONV pilot influent water was 0.03e0.04/cm, but the temperature in April was approximately 12 C versus 24 C in September. The CONV water formed 43 mg TTHM/mg of TOC after the three-day hold in April, but 77 mg TTHM/mg of TOC in September. The reactor effluents formed 73 mg TTHM/mg of TOC in April versus 95 mg TTHM/mg of TOC in September on average. The average percent increase in mg TTHM/mg of TOC through the reactors was greater in the Spring, 69% increase versus the Fall, 23%. All GAC effluents receiving CONV pilot influent, including the controls, produced similar TTHM 3-day SDS concentrations that were considerably lower in concentration. As was noted previously, an 8% improvement in TOC reduction (average) and 16% improvement in TOC reduction (during the summer months) was observed through the GAC columns receiving UV/ H2O2 treated water, but no corresponding reduction in TTHM 3day SDS was observed (See Fig. 6).
3.3.3. Total trihalomethane formation with Post-GAC pilot influent When the Post-GAC process stream served as the influent to the UV/H2O2 reactors, there was some increase in TTHM 3-day SDS concentrations, and this increase in TTHM 3-day SDS concentrations was observed through both MP and LP reactors (See Fig. 7). As can be seen in Fig. 1, less natural organic matter was present in the influent to the reactors. Not only was the TOC concentration and UV254 absorbable materials of the Post-GAC pilot influent lower than the CONV water, but SUVA was lower (See Fig. 1). Fig. 4 illustrates that the SUVA of the Post-GAC water is less significantly altered through the reactors compared to the SUVA of the CONV treated water. The GAC treatment removed much of the NOM, thus decreasing THM precursors. Westerhoff et al. (1999) noted a strong correlation between hydroxyl radical reactivity with NOM and SUVA. In April and September the UV254 absorbance of PostGAC pilot influent water was 0.01e0.02/cm. The Post-GAC water formed 29 mg TTHM/mg of TOC in April, but 63 mg TTHM/mg of TOC in September. The reactor effluents formed 54 mg TTHM/mg of TOC in April versus 96 mg TTHM/mg of TOC in September, illustrating the effect of temperature on TTHM
3976
300
30
250
25 Temperature
200
LP Reactor Ef f l
MP Reactor Ef f l
150
15 CONV Pilot Inf l
100 50All GAC
20
Effluent
U.S Max. Contam. Level 4 Qrt. Running Ave. 80 µg/L
0
Temperature °C
TTHM 3-day SDS (µg/L)
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 6 9 e3 9 8 0
10 5 0
GAC Run Days Fig. 6 e TTHM 3-day simulated distribution system (SDS) e through reactors and pilot GAC effluent with CONV pilot influent (best fit lines are meant to guide the eye).
formation, i.e., the higher formation in September was due to the higher temperatures. The formation of TTHM/mg/L of TOC in the chlorinated reactor effluent water in September was very similar to the reactor effluent water when CONV treated water was used as pilot influent, but the April reactor effluent data using the Post-GAC pilot influent was lower than the comparable data when CONV water was used as pilot influent. The percent increase in mg TTHM/mg of TOC through the reactors, however, was greater in the Spring, 86% increase versus the Fall, 53%. Four research groups have published results using UV/ H2O2 conditions similar to this present study, Kleiser and Frimmel (2000), Toor and Mohseni (2007), Dotson et al. (2010) and Sarathy and Mohseni (2010). Kleiser and Frimmel (2000) found that the maximum TTHM formation potential was 20% higher after UV/H2O2 treatment than in the control sample when held for 48 h. After extremely high levels of irradiation, that produced greater than 10% mineralization, Kleiser and Frimmel (2000) observed that THMs began to
TTHM 3-day SDS (µg/L)
100 80
U.S. Max. Contam. Level 4 Qrt. Running Ave. 80 µg/L
LP Reactor Effluent
60 MP Reactor Effluent
40
Post-GAC Pilot Influent
20 0
Fig. 7 e TTHM 3-day simulated distribution system (SDS) e through reactors with post-GAC pilot influent (best fit lines are meant to guide the eye).
decrease. The authors speculated that the reactions of peroxyl-radicals among themselves can lead to the production of ketones or aldehydes (in incomplete mineralization) with short irradiation times. In this present study 2e3% mineralization was observed through the reactors when CONV treated water served as pilot influent and 4e7% mineralization was observed through the reactors when PostGAC water served as pilot influent, Metz et al. (2011). Toor and Mohseni (2007) observed modest increases in TTHM formation potential between 0 and 1000 mJ/cm2 when 4 mg/L H2O2 was used. Sarathy and Mohseni (2010) observed that for UV/H2O2, UV doses below 1500 mJ/cm2 did not reduce TTHM formation. However, no increase in TTHM concentration was observed below this UV dose with 15 mg/L H2O2. The natural water utilized for these tests was a reservoir water of different water quality. The natural water used by Kleiser and Frimmel (2000), Dotson et al., 2010 and this present study was from a river water source. Sarathy and Mohseni (2010) also observed greater mineralization and less DBP formation when the higher molecular weight NOM was removed. This data partially agrees with the present study. More mineralization occurred in the GAC treated water and less THMs were formed. However, the yield per mg TTHM/mg was actually slightly higher when Post-GAC water served as UV/H2O2 influent. The results of Dotson et al. (2010) agree more closely with this present study. They reported that at a UV dose of 1000 mJ/cm2 and 10 mg/L H2O2, trihalomethane yield was increased by 21e29 mg/mg of TOC, using conventionally treated water (after chlorination and a 24 h hold at room temperature conditions), while increasing 18e33 mg/mg of TOC when using GAC treated water. The researchers found that THM yield correlated with hydroxyl radical exposure. Hydroxylation of aromatics or the transformation of less chlorine reactive hydrophobic fraction of NOM into the more highly reactive hydrophilic NOM was believed to occur. However, the degree to which TTHM formation potential was increased depended upon the dose of UV and H2O2 and the
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 6 9 e3 9 8 0
constituents in the water. Partial oxidation of NOM can lead to ring opening of aromatic structures, cleavage of conjugated double bonded carbon structures, and reduction in the degree of aromatic substitution. Ring cleavage can create more reactive sites. Hydroxyl radicals formed by the UV/H2O2 process may preferentially react with hydrophobic fractions of NOM yielding hydrophilic products (Sarathy and Mohseni, 2009, 2010). Hwang et al. (2001) found that polar NOM can produce a significant amount of DBPs. If the reactor effluent water from the CONV process stream were to be delivered directly to the distribution system, the system would not meet U.S. TTHM regulations. After GAC adsorption, pilot effluents (with and without UV/H2O2) produced similar THM 3-day SDS concentrations which would meet these regulations. When using Post-GAC pilot influent, the UV/H2O2 reactors increased the 3-day SDS TTHM concentrations. However, even after UV/H2O2, the reactor effluents would meet U.S. TTHM regulations.
3.3.4.
Brominated THM formation
When CONV treated water was used as pilot influent, the bromodichloromethane (CHBrCl2) 3-day SDS concentrations varied seasonally in the influent, but there was little difference in concentration of CHBrCl2 3-day SDS concentration through the pilot reactors. The CHBrCl2 3-day SDS concentration of the CONV pilot influent water was 30 mg/L and 36 and 35 mg/L for the LP and MP reactor effluent average values, respectively (See Table 1). This is interesting considering the increase in the TTHM SDS through the reactors. The CHBrCl2 3-day SDS concentration accounted for a lesser percentage of the TTHM SDS concentration in the reactor effluent water (24% for LP effluent and 25% for MP effluent) than in the CONV pilot influent (31%) on a mg/L basis. On a molar basis the percentages were 24, 20 and 21%, respectively. In the effluent of the GAC columns, the CHBrCl2 3-day SDS concentration was lower, 17 mg/L for the GAC effluents following the reactors and 16 mg/L for the GAC effluent from the control columns. CHBrCl2 SDS comprised 30% the TTHM SDS concentration for all GAC column effluent water and 31% of the CONV influent. The results for CHBrCl2 are significant because it is considered the most toxic of the TTHMs. The CHBrCl2 3-day SDS average concentration of the CONV influent and the associated reactor effluent waters were above 20 mg/L, a level of concern based on some prenatal toxicological research (Dodds and King, 2001). However, a large, well-designed study (Savitz et al., 2005) did not deem these levels of toxicological concern. Bromoform 3-day SDS concentrations were very low ranging from an average of 2.4e6.0 mg/L for all sample points through the pilot plant when CONV treated water was used as pilot influent. On average 2.9% of the TTHM 3-day SDS concentration of the CONV treated water was comprised of bromoform, while the bromoform comprised 1.6% and 9e11% of the TTHM 3-day SDS concentration for the UV/H2O2 reactor effluents and the subsequent GAC column effluent waters, respectively (See Table 1). Dibromochloromethane (CHBr2Cl) SDS concentrations ranged from an average of 18e21 mg/L for all sample points. Approximately 19% of the TTHM 3-day SDS concentration of the CONV treated water was comprised of CHBr2Cl. Thirteen to 14% of the TTHM 3-day SDS concentration of the UV/H2O2 reactor effluent water was comprised of
3977
CHBr2Cl and 36e38% in the subsequent GAC column effluent waters (See Table 1). When Post-GAC treated water was used as pilot influent, the CHBrCl2 3-day SDS concentration showed a slight increase through the reactors. However, concentrations were lower than those of the CONV pilot followed by GAC adsorbers and showed less seasonal variation, averaging 6.9 mg/L in the PostGAC pilot influent and 13 mg/L in the LP effluent and 11 mg/L in the MP effluent (See Table 1). The CHBrCl2 3-day SDS concentration accounted for a similar percentage of the TTHM SDS concentration in the Post-GAC pilot influent (24%) as in the reactor effluent water, 22% in the MP reactor effluent and 25% in LP reactor effluent. On a molar basis the percentages were 22% in the Post-GAC pilot influent and in the LP reactor effluent and 19% in the MP reactor effluent (See Table 1). When using Post-GAC water as pilot influent, bromoform 3-day SDS concentrations were low, ranging from an average of 10e13 mg/L for all sample points. Approximately 34% of the TTHM 3-day SDS concentration of the Post-GAC treated water was comprised of bromoform, while 23e26% of the TTHM 3-day SDS concentration of the UV/H2O2 reactor effluent water was comprised of bromoform (See Table 1). Dibromochloromethane (CHBr2Cl) 3-day SDS concentrations ranged from an average of 10e17 mg/L for all sample points. Approximately 33% of the TTHM 3-day SDS concentration of the Post-GAC treated water was comprised of CHBr2Cl. Twenty-eight to 33% of the TTHM 3-day SDS concentration of the UV/H2O2 reactor effluent water was comprised of CHBr2Cl (See Table 1). Dotson et al. (2010) found that bromide incorporation decreased after UV/H2O2 treatment in both CONV treated and Post-GAC waters, but it decreased more in the Post-GAC samples. This agrees with the present study. When CONV water was used as pilot influent, the percentage of TTHM as chloroform increased from 63 to 70% on a molar basis, while bromoform decreased from 1.4 to 0.9% of TTHM. When PostGAC water was used as pilot influent, the percentage of TTHM as chloroform increased from 32 to 45% on a molar basis, while bromoform decreased from 21 to 14% of TTHM. When the CONV treated UV/H2O2 reactor effluent was GAC filtered, this trend reversed (See Table 1). The concentration of some brominated DBPs can increase after GAC adsorption and subsequent chlorination (Symons et al., 1993). Bromide is passed conservatively through the GAC and results in higher bromide to NOM ratios than in the pre-GAC water (Owen et al., 1995). So it is not surprising that the Post-GAC pilot influent had higher bromoform 3-day SDS concentrations than the CONV pilot influent. Low-aromatic components of NOM have higher bromine incorporation when chlorinated (Kitis et al., 2001). As was previously mentioned, the GAC treated waters in this study had lower SUVA values than the CONV treated waters, indicating less aromatic composition. It should be noted that the final 3-day SDS bromoform concentration in the final step of both treatment processes (Post-GAC pilot influent after the reactors and CONV pilot influent after reactors and GAC treatment) was 12.5 and. 5.5 mg/L, respectively. These differences are not of practical concern. None of the individual THM species had 3-day SDS concentrations that exceeded the World Health Organization guidelines.
3978
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 6 9 e3 9 8 0
Table 1 e Speciation of TTHM SDS by mg/L and mmol/L
CHCl3
CHBrCl2
CHBr2Cl
CHBr3
TTHM
3.4.
CONV Influent MP Effluent LP Effluent MP GAC Effluent LP GAC Effluent GAC Control #1 GAC Control #2 Post-GAC Influent MP Effluent LP Effluent CONV Influent MP Effluent LP Effluent MP GAC Effluent LP GAC Effluent GAC Control #1 GAC Control #2 Post-GAC Influent MP Effluent LP Effluent CONV Influent MP Effluent LP Effluent MP GAC Effluent LP GAC Effluent GAC Control #1 GAC Control #2 Post-GAC Influent MP Effluent LP Effluent CONV Influent MP Effluent LP Effluent MP GAC Effluent LP GAC Effluent GAC Control #1 GAC Control #2 Post-GAC Influent MP Effluent LP Effluent CONV Influent MP Effluent LP Effluent MP GAC Effluent LP GAC Effluent GAC Control #1 GAC Control #2 Post-GAC Influent MP Effluent LP Effluent
Avg. conc. mg/L
% of TTHM by conc.
Avg. conc. mmol/L
Molar % of TTHM
58 87 88 14 13 11 12 7.3 19 19 30 35 36 17 17 16 16 6.9 11 13 18 19 20 20 21 21 20 9.7 14 17 2.8 2.3 2.4 5.2 5.7 6.3 6.0 10 13 12 96 144 146 56 56 54 53 29 50 52
60% 60% 60% 25% 23% 20% 23% 25% 38% 37% 31% 24% 25% 30% 30% 30% 30% 24% 22% 25% 19% 13% 14% 36% 38% 39% 38% 33% 28% 33% 2.9% 1.6% 1.6% 9.3% 10% 12% 11% 34% 26% 23% e e e e e e e e e e
0.49 0.73 0.74 0.12 0.11 0.09 0.10 0.06 0.16 0.16 0.18 0.21 0.22 0.10 0.10 0.10 0.10 0.04 0.07 0.08 0.09 0.09 0.10 0.10 0.10 0.10 0.10 0.05 0.07 0.08 0.01 0.01 0.01 0.02 0.02 0.02 0.02 0.04 0.05 0.05 0.77 1.04 1.06 0.34 0.34 0.32 0.32 0.19 0.34 0.37
63% 70% 69% 35% 32% 29% 32% 32% 46% 43% 24% 20% 21% 31% 31% 31% 31% 22% 19% 22% 11% 9% 9% 28% 30% 32% 30% 25% 19% 22% 1.4% 0.9% 0.9% 6.1% 6.7% 7.9% 7.5% 21% 15% 13% e e e e e e e e e e
Haloacetic acid formation
HAA5 3-day SDS concentration increased through the UV/ H2O2 MP and LP reactors when CONV water was used as pilot influent (See Fig. 8.). The CONV pilot influent HAA5 3-day SDS concentration averaged 34 mg/L, while the LP reactor averaged 53 mg/L and the MP reactor averaged 51 mg/L. All pilot GAC effluent streams averaged 15e20 mg/L. HAA5 3-day SDS concentration increased slightly through the MP reactor when Post-GAC was used as pilot influent (from 14 mg/L to 20 mg/L), but did not increase through the LP reactor (See Fig. 9). All
process streams (with and without UV/H2O2) would meet U.S. HAA5 regulations based on these results. This present study agreed with the findings of Toor and Mohseni (2007). An increase in HAA formation potential was observed between 0 and 1000 mJ/cm2 for 4 and 23 mg/L H2O2. The results of both studies differ from Sarathy and Mohseni (2010) who showed a decrease in HAA formation potential at a UV dose of 500 mJ/cm2 and a H2O2 dose of 15 mg/L (in contrast to 10 mg/L in this present study). The results also differ from Dotson et al. (2010) who found increases and decreases in HAA formation potential, but no discernable trend.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 6 9 e3 9 8 0
CONV Pilot Infl.
Control GAC Effl.
Regen. GAC Effl.
MP GAC Effl.
LP Reactor Effl.
LP GAC Effl.
U.S. Max. Contaminant Level 4 Qrt. Running Ave. 60 µg/L
60 HAA5 3-day SDS (µg/L)
MP Reactor Effl.
50 40 30 20 10 0 Average
Fig. 8 e Average HAA5 3-day simulated distribution system (SDS) with CONV pilot influent.
The increase in HAA5 precursors may be the result of the UV/H2O2 reaction increasing the concentration of oxygenated NOM, creating more hydrophilic compounds. Sarathy and Mohseni (2009, 2010) reported that UV/H2O2 preferentially reacted with the hydrophobic fractions of NOM leading to the formation of hydrophilic products. Hwang et al. (2001) found that the hydrophilic/polar portions of NOM readily formed HAA5s upon chlorination. Apparently, the GAC treatment in this present study removed a significant portion of the NOM with the potential of forming HAA5 precursors upon UV/H2O2 treatment. Sarathy and Mohseni (2009, 2010) found that when the very hydrophobic fraction of NOM was removed before UV/H2O2 treatment, HAA formation was reduced. In this present study GAC reduced the very hydrophobic fraction of the NOM, producing similar results.
Post-GAC Pilot Infl.
HAA5 3-day SDS (µg/L)
60
MP Reactor Effl.
LP Reactor Effl.
U.S. Max. Contaminant Level 4 Qrt. Running Ave. 60 µg/L
50
3979
No measurable differences between the monohalogenated, di-halogenated and tri-halogenated species was observed for this present study. Small increases were seen in all species. This lack of differences in speciation upon UV/H2O2 treatment was unlike the THM results, but agrees with the work of Dotson et al. (2010). He found that the yield of the di-halogenated and tri-halogenated species did not depend on the concentration of H2O2, but on MP dose. Reckhow et al. (1990) noted the varying nature of dichloroacetic acid and trichloroacetic acid precursors. Increases in dichloroacetic acid were linked to the formation of diketones and aldehydes. Unlike this present study, Toor and Mohseni (2007) found that dichloroacetic acid formation increased with increasing UV doses, while trichloroacetic acid formation decreased. However, different source waters may account for the differences in the formation of species.
4.
Conclusions
For a typical river source at practical UV/H2O2 treatment conditions (effecting a 2e7% mineralization of NOM), the following could be concluded: Three-day SDS TTHMs significantly increased for both the CONV treated water and the Post-GAC reactor influent waters, creating regulatory concerns. Three-day SDS HAAs significantly increased for the CONV treated water and slightly increased for the Post-GAC reactor influent waters (MP lamp). Although increases in 3-day SDS TTHM consistently occurred through the UV/H2O2 reactors, the degree of increase was most affected by distribution system temperature. When MP and LP pilot units were operated to produce 80 percent atrazine destruction by UV/H2O2, the 3-day simulated distribution TTHM concentrations from both reactors were very similar year-round. GAC adsorption was beneficial in reducing DBP precursors whether it was employed before or after the UV/H2O2 reactors, although typical shifts to the more brominated THM species were observed. Designers of UV/H2O2 installations need to consider doses of UV and H2O2 that will minimize DBP precursors as well as reduce target contaminants to desired levels.
40
Acknowledgments
30 20 10 0
Average
Fig. 9 e HAA5 3-day simulated distribution system (SDS) with post-GAC pilot influent.
Various aspects of this research were conducted in cooperation with KWR Watercycle Research institute, Dunea Water, the University of Colorado, the University of Cincinnati and Philips Lighting. We would like to acknowledge the support of the Dutch Ministry of Economic Affairs and the Water Research Foundation (formerly the American Water Works Research Foundation) for their support of this research. Additionally, we would like to thank Kimberley Curry, Niranjan Selar, Katherine Jamriska, Cheri Woody and the laboratory staff of the Greater Cincinnati Water Works.
3980
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 6 9 e3 9 8 0
Appendix. Supplementary material Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.watres.2011.05.001.
references
American Public Health Association, American Water Works Association and Water Environment Federation, 1998. Standard Methods for the Examination of Water and Waste Water, twentieth ed. Washington D.C. Bond, T., Goslan, E.H., Parsons, S.A., Jefferson, B., 2011. Review: treatment of disinfection by-product precursors. Environmental Technology 32 (1), 1e25. Croue´, J.P., Korshin, G.V., Benjamin, M., 2006. Characterization of Natural Organic Matter in Drinking Water. AWWA Research Foundation, Denver, CO. Report # 90780. Dickenson, E.R., Summers, R.S., Croue´, J.P., Gallard, H., 2008. Haloacetic acid and trihaolmethane formation from the chlorination and bromination of aliphatic b-dicarboxylic model compounds. Evvironmental Science and Technology 42, 3226e3233. Dodds, L., King, W.D., 2001. Relation between trihalomethane compounds and birth defects. Occupational and Environmental Medicine 58, 443e446. Dotson, A.D., Keen, V.S., Metz, D., Linden, K.G., 2010. UV/H2O2 treatment of drinking water increases post-chlorination DBP formation. Water Research 44 (12), 3703e3713. Gallard, H., Von Gunten, U., 2002a. .Chlorination of natural organic matter: kinetics of chlorination and of THM formation. Water Research 36, 65e74. Gallard, H., Von Gunten, U., 2002b. Chlorination of phenols: kinetics and formation of chloroform. Environmental Science and Technology 36 (5), 884e890. Hwang, C., Krasner, S., Schlimenti, M., Amy, G., Dickenson, E., Bruchet, A., Prompsy, C., Filippi, G., Croue´, J.P., Violleau, D., Leenheer, J.A., 2001. Polar NOM: Characterzation, DBPs, Treatment. AWWA Research Foundation, Denver, CO. Report # 90877. Kitis, M., Karanfil, T., Kilduff, J.E., Wigton, A., 2001. The reactivity of natural organic matter to disinfection by-products formation and its relation to specific ultraviolet absorbance. Water Science and Technology 43 (2), 9e16. Kleiser, G., Frimmel, F.G., 2000. Removal of precursors for disinfection by-products (DBPs)-differences between ozoneand OH-radical-induced oxidation. Science of the Total Environment 256 (1), 1e9. Liang, L., Singer, P.C., 2003. Factors influencing the formation and relative distribution of haloacetic acids and trihalomethanes in drinking water. Environmental Science & Technology 37 (13), 2920e2928. Liu, Wenjun, Andrews, Susan A., Stefan, Mihaela I., Bolton, James R., 2003. Optimal methods for quenching H2O2 residuals prior to UFC testing. Water Research 37, 3697e3703. Magnuson, M.L., Kelty, C.A., Sharpless, C.M., Linden, K.G., Fromme, W. , Metz, D.H., Kashinkunti, R., 2002. Effect of UV irradiation on organic matter extracted from treated Ohio River water wtudied
through the use of electrospray mass spectrometry. Environmental Science & Technology 36, 5252e5260. Matilainen, A., Sillanpaa, M., 2010. Removal of natural organic matter from drinking water by advanced oxidation processes. Chemosphere 80, 351e365. Metz, D.H., Reynolds, K., Meyer, M., Dionysiou, D.D., 2011. The effect of UV/H2O2 treatment on biofilm formation potential. Water Research 45 (2), 497e508. Norwood, D.L., Johnson, J.D., Christman, R.F., Haas, J.R., Bobenrieth, M.J., 1980. Reactions of chlorine with selected aromatic models of aquatic humic material. Environmental Science and Technology 14 (2), 187e190. Owen, D.M., Amy, G.L., Chowdhury, Z.K., Paode, R., McCoy, G., Viscosil, K., 1995. NOM characterization and treatability. Journal of American Water Works Association 87 (1), 46e63. Pereira, V.J., Weinberg, H.S., Linden, K.G., Singer, P.C., 2007. UV degradation kinetics and modeling of pharmaceutical compounds in laboratory grade and surface water via direct and indirect photolysis at 254 nm. Environmental Science and Technology 41 (5), 1682e1688. Reckhow, D.A., Singer, P.C., Malcolm, R.L., 1990. Chlorination of humic materials: by-product formation and chemical interpretations. Environmental Science and Technol. 24 (11), 1655e1664. Sarathy, S.R., Mohseni, M., 2007. The impact of UV/H2O2 advanced oxidation on molecular size distribution of chromophoric natural organic matter. Environmental Science and Technology 41, 8315e8320. Sarathy, S.R., Mohseni, M., 2009. The fate of natural organic matter during UV/H2O2 advanced oxidation of drinking water. Canadian Journal of Civil Engineering 36 (1), 160e169. Sarathy, S.R., Mohseni, M., 2010. Effects of UV/H2O2 advanced oxidation on chemical characteristics and chlorine reactivity of surface water natural organic matter. Water Research 44, 4087e4096. Savitz, D.A., Singer, P.C., Hartmann, K.E., Herring, A.H., Weinberg, H.S., Makarushka, C., Hoffman, C., Chan, R., Maclehose, R., 2005. Drinking Water Disinfection By-products and Pregnancy Loss. American Water Works Assoc. Research Foundation, Denver. CO. Report No. 91088F. Symons, J.M., Krasner, S.W., Simms, L.A., Sclimenti, M., 1993. Measurement of THM and precursor concentrations revisited: the effect of bromide ion. Journal of American Water Works Association 85 (1), 51e62. Thomson, J., Roddick, F.A., Drikas, M., 2004. Vacuum ultraviolet irradiation for natural organic matter removal. Journal of Water Supply: Research and Technology e Aqua 53 (4), 193e206. Toor, R., Mohseni, M., 2007. UV-H2O2 based AOP and its integration with biological activated carbon treatment for DBP reduction in drinking water. Chemosphere 66, 2087e2095. Westerhoff, P., Aiken, G., Amy, G., Debroux, J., 1999. Relationships between the structure of natural organic matter and its reactivity towards molecular ozone and hydroxyl radicals. Water Research 33 (10), 2265e2276. Westerhoff, P., Mezyk, Cooper W.J., Minakata, D., 2007. Electron pulse radiolysis determination of hydroxyl radical rate constants with Suwannee River fulvic acid and other dissolved organic matter isolates. Environmental Science and Technology 41, 4640e4646.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 8 1 e3 9 9 0
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Effects of humic acid on physical and hydrodynamic properties of kaolin flocs by particle image velocimetry Runsheng Zhong a,*, Xihui Zhang a, Feng Xiao b, Xiaoyan Li b, Zhonghua Cai c a
Research Center for Environmental Engineering & Management, Graduate School at Shenzhen, Tsinghua University, Shenzhen 518055, China b Department of Civil Engineering, The University of Hongkong, Hongkong, China c Life Sciences Division, Graduate School at Shenzhen, Tsinghua University, Shenzhen 518055, China
article info
abstract
Article history:
The physical and hydrodynamic properties of kaolin flocs including floc size, strength,
Received 31 January 2011
regrowth, fractal structure and settling velocity were investigated by in situ particle image
Received in revised form
velocimetry technique at different humic acid concentration. Jar-test experimental results
1 April 2011
showed that the adsorbed humic acid had a significant influence on the coagulation
Accepted 5 May 2011
process for alum and ferric chloride. Kaolin flocs formed with the ferric chloride were
Available online 18 May 2011
larger and stronger than those for alum at same humic acid concentration. Floc strength and regrowth were estimated by strength factor and recovery factor at different humic acid
Keywords:
concentration. It was found that the increased humic acid concentration had a slight
Floc strength
influence on the strength of kaolin flocs and resulted in much worse floc regrowth. In
Floc restructuring
addition, the floc regrowth after breakage depended on the shear history and coagulants
Fractal dimension
under investigation. The changes in fractal structure recorded continuously by in situ
Humic acid
particle image velocimetry technique during the growth-breakage-regrowth processes
Particle image velocimetry
provided a supporting information that the kaolin flocs exhibited a multilevel structure. It was proved that the increased humic acid concentration resulted in decrease in mass fractal dimension of kaolin flocs and consequently worse sedimentation performance through free-settling and microbalance techniques. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
Floc growth and breakage by fluid shear are encountered simultaneously in coagulation and sedimentation processes at drinking water treatment works (DWTWs) because most coagulation processes are carried out in agitated suspensions, nearly always under turbulent conditions assuring thorough mixing and high collision frequency between particles and therefore rapid floc growth (McCurdy et al., 2004; Spicer and Pratsinis, 1996a). After a characteristic time of shear-induced coagulation, a steady state is reached between growth and
breakage and the floc size distribution no longer changes (Spicer and Pratsinis, 1996b; Gregory, 2009; Spicer et al. 1996). The physical and hydrodynamic properties of flocs formed at a steady state play an important role in the performance of coagulation and sedimentation. The floc properties have a close relationship with the shear rate, the coagulants under investigation and the coagulation mechanisms involved (Biggs and Lant, 2000; Jarvis et al., 2005a; Li et al., 2006; Xu et al., 2010). However, particles in source water are coated readily with ubiquitous natural organic matter (NOM) (Wilkinson et al., 1997). This occurrence modifies the physicochemical
* Corresponding author. Tel.: þ86 755 26036701; fax: þ86 755 26036709. E-mail address:
[email protected] (R. Zhong). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.05.006
3982
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 8 1 e3 9 9 0
characteristics of the particle surface with the result that adsorbed NOM layer have significant influences on particle aggregation and subsequent floc structure formed in water treatment processes (Kretzschmar et al., 1998; Lee et al., 2005; Tirado-Miranda et al., 2003; Wilkinson and Reinhardt, 2005). Little thought is generally given to the influence of NOM on the physical and hydrodynamic properties of flocs formed such as floc sizes, strength, regrowth, and settling velocity in previous studies (Bache, 2004; Kretzschmar et al., 1997; Tirado-Miranda et al., 1999, 2003; Xiao et al., 2011). It has been shown that flocs formed are fractal that are invariant to a change of length scale (Meakin, 1988). An important parameter that characterizes a fractal object is the fractal dimension. The method of fractal dimension is crucial which rely mainly upon image analysis by microscopy (Chakraborti et al., 2003; Gorczyca and Ganczarczyk, 1996; Gregory and Dupont, 2001; Li et al., 2007; Wang et al., 2007), floc size distribution (Li and Leung, 2005; Li and Logan, 1995), dry mass tests (Li et al., 2003; Mu et al., 2008), free-settling tests (Gorczyca and Ganczarczyk, 1996; Wu et al., 2002) and light scattering tests (Jarvis et al., 2005b; Li et al., 2006; Tang et al., 2001; Wu et al., 2002). The light scattering method bases on the power function of the total scattered light intensity and the scattering vector is suitable only for flocs from several to several hundred micrometer (Jarvis et al., 2005b). Although the other methods can be utilized for flocs from several micrometer to several millimeter formed at DWTWs, they need to take the flocs from the containing vessel by an endcut pipette or a peristaltic pump and could easily damage and break the flocs due to their highly fragile nature (Gregory, 2004; Jarvis et al., 2005a). Thus, the nondestructive or nonintrusive technique needs to be developed in future work (Xu et al., 2008; Xiao et al., 2011). Previous research has focused on the determination of fractal dimension at a steady state. The measured fractal dimensions of flocs generated in water and wastewater treatment processes ranged from 1.4 to 2.8 (Li and Ganczarczyk, 1989). However, few studies have addressed the changes in fractal dimension indicating that flocs can exhibit multilevel structures during floc formation (Chu and Lee, 2004; Gmachowski, 2008; Wu et al., 2002). The objectives of this paper were to investigate the structural changes during flocs formation and the influences of humic acid on floc strength, structure and settling velocity, and compare floc size distribution at a steady state during growth-breakage-regrowth by in situ particle image velocimetry (PIV) technique.
2.
Materials and methods
2.1.
Humic acid and suspension preparation
Humic acid (Pahokee Peat, International humic substance society, USA) was used as model NOM which was extracted from marsh. Kaolin particle (Tianjin Fuchen Company, China) which was typical mineral particle in source water was selected to prepare suspension in this study. The particle size distribution of the kaolin particle was determined using a particle size analyzer (Multisizer 3, Beckman Coulter, USA) and found to be a relatively monodispersed distribution with
a mean diameter of 2.0 mm. Kaolin particles were soaked in 6% NaOH solution at 80 C for 24 h. This procedure aimed to remove the impurities and NOM on kaolin particle surface. The Na-kaolin particles were prepared by saturating and equilibrating five times with 1 mol/L NaCl solutions for at least 3 h, and then were washed three times free of excess salt with ultrapure water (resistance > 1.82 107 U/cm, total organic carbon < 1 mg/L, Milli-Q, Millipore, USA).
2.2.
Floc formation, breakage and regrowth
Reagent-grade alum (Al2(SO4)3$18H2O) and ferric chloride (FeCl3$6H2O) were used as coagulants. For coagulation tests, a conventional jar test apparatus (ZR-6, Zhongrun Water Ltd., China) was used to determine the optimum alum dosage at room temperature (25 1 C), with a rapid mixing at 200 rotations per minute (rpm) for 1 min, flocculation at 20 rpm for 30 min, and quiescent settling for 1 h. The zeta potentials of naked and coated kaolin particles before and after coagulant addition were measured by a laser zeta potential analyzer (Model Delsa 440SX, Beckman Coulter Inc., USA). Floc growth, breakage and regrowth experiments were performed using a variable speed jar tester with 76 mm 25 mm flat paddle impellers and rectangular jars with 80 mm 80 mm 200 mm containing 1 L samples of each suspension. Coagulation experimental procedure at optimum alum dosage was carried out as above, and followed by a breakage phase at 100 rpm after 30 min flocculation phase. The changes in floc structure were investigated by breakage under two different shear periods: (1) a long breakage period of 15 min and (2) a short breakage period of 30 s. After the breakage phase, a slow stir at 20 rpm was reintroduced for a further 30 min for floc regrowth. The dynamic floc size distribution was measured during growth, breakage and regrowth of the flocs using in situ PIV system (Xiao et al., 2011). The PIV system consists of a laser source (600 nm in wavelength, Coherent, USA), a high-speed CCD video camera (PCO. Imaging 1200 hs with a resolution of 1280 1024 pixels, PCO, Germany), a process control (PCO. Camware, PCO, Germany) and image processing software package (Scion Image, Scion, USA). As a non-intrusive laser setup, the PIV was able to record the high-quality image of particles without disturbance in a jar tester within a millisecond and had a resolution of around 9 mm for particle tracking and imaging based on calibration. A pulsed laser beam generated from the laser source was expanded to a thin laser light sheet by a combination of a cylindrical and a spherical lenses. The laser sheet illuminated a planar region of the water for visualization of the particles and flocs in the rectangular flocculation tank. The images of laserilluminated particles or flocs could be captured by a highspeed CCD camera. The images of particles or flocs were processed with an image analysis system for floc size distribution determination. Floc sizes were calculated in terms of the equivalent diameter by d ¼ (4A/p)1/2, where A is the projected area of the floc. More than 20 consecutive images were analyzed to produce a size distribution of the particles or flocs, and the result was presented as a volumebased discrete floc size distribution.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 8 1 e3 9 9 0
To investigate the influence of humic acid on coagulation process, floc strength, regrowth and structure of kaolin flocs, suspensions containing 10 mg/L kaolin (12 NTU) and different humic acid contents were made for coagulation experiments, including (1) HA0 - no humic acid addition, (2) HA1 - 1 mg C/L humic acid, (3) HA3 - 3 mg C/L humic acid, (4) HA10 - 10 mg C/L humic acid. The pH of suspensions was held constantly at 7.05 by 104 mol/L NaHCO3 buffer.
2.3.
Settling experiment
PIV settling setup consisted of PIV system as above and rectangular settling column with a floc collector at bottom. The settling column was 500 mm in height to ensure that the terminal settling velocity could be reached, and 40 mm in width to minimize the wall effect on floc settling. Prior to being placed into the top of the settling column, the kaolin flocs were transferred in series through two separate beakers filled with water identical to that placed in the upper column with quiescent water. Each floc was measured for its settling velocity by using PIV system to record the settling time and the distance simultaneously for the floc to settle through the 25 cm distance in the lower portion of the settling column. The images of floc recorded were used to determine the maximum diameter and projected area of floc (Scion Image, Scion, USA). The size of the floc was calculated as the equivalent diameter of its projected area. The floc eventually fell into the collector and was removed for subsequent analysis such as dry mass measurement by microbalance (Model AEM 5200, Labror, Shimadzu, Japan). Flocs that broke up during settling or any transfer step were discarded.
2.4.
Strength and recovery factor
Floc strength is dependent upon the inter-particle bonds between the components of the aggregate (Bache et al., 1997; Jarvis et al., 2005a). This includes the strength and number of individual bonds within the floc, and it is proved to be difficult to measure directly inter-particle forces partly due to the inherent complexity, fragility and variation in floc size, shape and composition and also due to two modes of floc rupture including surface erosion and large-scale fragmentation (Jarvis et al., 2005a). Most previous researches have been concerned with the relative floc size to quantify floc strength under the applied shear rate. The relative breakage and regrowth of different flocculated systems have been expressed by use of floc strength (S ) and recovery factor (R) which are calculated as follows (Francois, 1987; Jarvis et al., 2005a): S¼
d2 100 d1
(1)
R¼
d3 d2 100 d1 d2
(2)
where d1 is the steady-state floc size during floc formation, d2 is the floc size after the floc breakage period, and d3 is the steady-state floc size after regrowth. A higher R value indicates a higher strength of the flocs to resist breakage when exposed to an increasing shear. Likewise, an increase in the
3983
recovery factor shows higher capability for flocs to regrow after being exposed to high shear.
2.5.
Determination of fractal dimension
Besides the floc size distribution, floc images recorded by PIV can provide more information about the morphological and structural feature of kaolin flocs. The two-dimensional fractal dimension, D2, can be calculated from regression analysis of the log of the projected floc area and the log of the characteristic length during floc growth-breakage-regrowth process (Chakraborti et al., 2000). In this study the maximum diameter of floc was used as the characteristic length. This method has been used and described in detail by other researchers as follows: AflD2
(3)
where A is the projected aggregate area, l is the maximum diameter of floc, and D2 is the fractal dimension for the areato-length relation. The mass fractal dimension of kaolin floc (D3) was determined by the power relationship between dry mass of kaolin flocs and their size as follows (Gmachowski, 2003; Li and Yuan, 2002): mfdD3
(4)
where m is the floc dry mass; D3 is the mass fractal dimension; d is the equivalent diameter of its projected area, d ¼ (4A/p)1/2, where A is the projected area.
3.
Results and discussion
3.1.
Coagulation optimization and zeta potential
Through standardized jar testing procedures it was found that the optimum coagulant dosages were 0.8, 1.2, 1.6, 1.6 mg/L as Al for alum and 1.2, 4.1, 5.2, 5.2 mg/L as Fe for ferric chloride at 0, 1, 3, 10 mg/L humic acid at pH 7.05 based on TOC removal and settled turbidity. The humic acid removal efficiency as determined by TOC removal was about 70% for alum and ferric chloride. The optimum alum and ferric chloride dosages were less than reported optimum dosage as Al and Fe for natural water with 10 mg C/L NOM (Jarvis et al., 2005b). These data showed that the alum and ferric chloride dosages increased with increasing humic acid concentration. The adsorption of humic acid on kaolin particle surface resulted in increases in the zeta potential of kaolin particle from 30 mV to 50 mV with more negatively charged surface and consequently higher coagulant doses. The zeta potential of kaolin particles at the optimum dosages shifted from about þ10 mV to about 19 mV for alum and 31 mV for ferric chloride with increasing humic acid concentration. This indicates that charge neutralization is not the only mechanism for kaolin and humic acid system in this study. The coagulation condition results in different coagulation mechanisms including coagulant dosage and physicochemical condition (Gregory, 2006; Xiao et al., 2011). It is believed that a combination of charge neutralization and entrapment/
3984
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 8 1 e3 9 9 0
adsorption of NOM onto metal precipitates are the major floc formation routes for hydrolyzing coagulants such as ferric chloride and alum (Gregor et al., 1997; Shi et al. 2007; Jarvis et al. 2008; Zhan et al., 2010; Xu et al., 2010). The removal of NOM is mainly dominated by adsorption onto precipitated metal hydroxides and insoluble precipitates by the complexation of metal coagulant with humic acid at pH > 6 (Gregor et al., 1997).
3.2.
Effects of humic acid on floc size and strength
To avoid the influence of coagulation mechanism on floc structure, strength and regrowth due to different coagulant dosage, the introduced coagulant dosages of kaolin suspension with different humic acid concentration were 1.6 mg/L as Al for alum and 5.2 mg/L as Fe for ferric chloride during floc breakage and regrowth experiments. The alum and ferric flocs under investigation can be considered similar to sweep flocs in that they are mainly composed of metal hydroxides due to neutral pH (¼7.05) and higher coagulant dosages. The results of floc size distribution of kaolin suspensions with different humic acid concentration are showed in Fig. 1 for 15 min breakage and Fig. 2 for 30 s breakage during floc formation, breakage and regrowth. After the slow stir phase, alumkaolin flocs had reached a d(50) floc size of 1371 mm and ferrickaolin flocs had a floc size of 1700 mm being the largest flocs. These approximately doubled the size of the alum-kaolinhumic flocs (629 mm) and ferric-kaolin-humic flocs (836 mm) at 10 mg C/L humic acid. The steady state floc size during floc growth decreased progressively with the increasing humic acid concentration for the two coagulants. The ferric-kaolin floc sizes were larger than alum-kaolin floc under the same humic acid concentration. As a comparison to alum- and ferric-kaolin flocs, the decreased size of the alum (ferric)kaolin-humic flocs highlights the importance humic acid plays during coagulation. Therefore, the interaction of humic acid in the floc formation process through direct chemical reaction and adsorption onto metal precipitates and kaolin particles significantly reduced growth including electrostatic and steric repulsion (Abu-Lail and Camesano, 2003; Franchi and O’Melia, 2003; Gregory, 2006). As the stirring speed increased to 100 rpm, the d50 size value showed an immediate and rapid decrease in all cases, indicating rapid floc breakage as a result of increased shear in Fig. 2. The d50 size of the flocs after 15 min exposure to shear ranged from 202 to 411 mm for the ferric chloride coagulant while the alum flocs were broken into smaller floc sizes in the range of 122e249 mm at different humic acid concentration in Figs. 1 and 2. Before a high shear was applied, a large peak of alumkaolin with no humic acid was seen at around 1400 mm in Fig. 1a. With an increase in shear to 100 rpm, the large floc size peak shifted from 1400 mme200 mm commensurate with a slight increase in the flocs of size <20 mm. The shifts of floc size peak of the other flocs were similar to that of the alumkaolin flocs. Generally, there are two theoretical modes of floc rupture classified as surface erosion and large-scale fragmentation (Jarvis et al., 2005a). Surface erosion is the removal of small particles or microflocs from the floc surface resulting in an increase in the primary particle and microfloc concentration. Large-scale fragmentation is the cleavage of flocs into pieces of a similar size without an increase in
primary particle concentration which happens in the early breakage phase. The floc size distribution in Figs. 1 and 2 after breakage indicated that fragmentation and erosion occurred simultaneously for the long shear time while the breakage mode of the flocs was mainly the large-scale fragmentation for the short shear time. Floc strength is a particularly important structural characteristic in coagulation and sedimentation processes for the efficient removal of NOM and mineral particles. The strength factor values in Table 1 shows that the humic acid concentration had slight influence on the strength of kaolin flocs for both coagulants. The ferric flocs were better able to resist shear with floc strength factor values of 24 while the alum flocs had a floc strength factor of 18 when exposed to the increased shear. The extent of floc breakage was lower at the reduced shear period of 30 s for both coagulants with the floc strength factor increasing by an additional 6e16 from the longer shear period (Table 1). The ferric floc strength under investigation is similar to that of ferric precipitates reported in previous studies (Jarvis et al., 2005b; Xiao et al., 2011). The flocs formed from alum and ferric chloride are weak and fragile partly due to sweep flocculation (Bache et al., 1997) and have little increase in floc strength with increasing humic acid concentration (Table 1). This does not agree with the finding that adsorbed NOM within a floc should provide either increased attraction or increased repulsion leading to significant increase in floc strength (Jarvis et al., 2005b; Xiao et al., 2011). Further research needs to be carried out in future work.
3.3.
Effect of humic acid on floc regrowth
As the shear was reduced to 20 rpm again, the flocs began to regrow; however, the kaolin flocs could not regrow to anywhere near their previous size known as irreversible breakage regardless of the coagulants used (Table 1). The effects of humic acid concentration and coagulants on the regrowth of kaolin flocs are shown in Table 1, Figs. 1 and 2. The increased humic acid concentration resulted in the worse regrowth of kaolin flocs for both coagulants. With a long shear time, the regrowth of ferric-kaolin flocs was poorer with recovery factor values of 7e10 lower than those for alum with recovery factor values of 11e22 at low humic acid concentration while was similar to that for alum at 10 mg C/L humic acid. This agrees with our findings in previous works (Xiao et al., 2011). However, with a short shear time, the floc regrowth was seen to improve considerably. The alum-kaolin flocs had significantly higher recovery when compared to all of the flocs investigated at the short shear time with a recovery factor of 91. The ferric flocs were only able to reform to below 31% of their initial size. The floc size distribution data showed a reduction in the flocs below 20 mm after floc regrowth for all of the flocs. It indicates that the flocs after regrowth reformed by the primary particles, microflocs and large flocs. The capacity for flocs to regrow is an important operational parameter during coagulation process. The irreversible breakage of flocs gives some indication of the floc internal bonding structure that the flocs formed in these systems were held together by chemical rather than physical bonds (Yukselen and Gregory, 2004). The results of this work have
3985
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 8 1 e3 9 9 0
25
50
a
40
15
% volume
% volume
20
Alum-kaolin-humic floc (0 mg C/L) Growth Breakage Regrowth
10 5
100
b
25 20 % volume
% volume
10000
15 10
0
0
20
Alum-kaolin-humic floc (3 mg C/L) Growth Breakage Regrowth
10000
100 35
c
30 25 % volume
15 10 5
1000
Ferric-kaolin-humic floc (3 mg C/L) Growth Breakage Regrowth
10000
g
20 15 10
0 100
20
f
5
0
25
Ferric-kaolin-humic floc (1 mg C/L) Growth Breakage Regrowth
10000
10 5
1000
1000
15
5
25
% volume
1000
Alum-kaolin-humic floc (1 mg C/L) Growth Breakage Regrowth
100
1000
Alum-kaolin-humic floc (10 mg C/L) Growth breakage Regrowth
10000
100
d
25 20
15
% volume
% volume
20
0 100
20
30
10
0
25
e
Ferric-kaolin-humic floc (0mg/L) Growth Breakage Regrowth
10
0
0 1000
10000
h
10 5
Floc size (mm)
Ferric-kaolin-humic floc (10 mg C/L) Growth Breakage Regrowth
10000
15
5
100
1000
100
1000
10000
Floc size (mm)
Fig. 1 e Floc size distributions of the kaolin flocs after growth, breakage, and regrowth after 15 min high shear time. Coagulant dosages were 1.6 mg/L as Al and 5.2 mg/L as Fe respectively.
shown that the floc regrowth had a close relationship with the coagulants and humic acid concentration. Previous researches have indicated the adsorbed NOM stabilized the suspended particles (Kretzschmar et al., 1998; Tirado-
Miranda et al., 1999; Walker and Bob, 2001). The adsorbed humic acid layer modified the physicochemical characteristics of kaolin particle surface resulting in the decrease in effective contacts between particles and flocs
3986
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 8 1 e3 9 9 0
25
a
30
Growth Breakage Regrowth
25
15
% volume
% volume
20
Alum-kaolin-humic floc (0 mg C/L)
10 5
Alum-kaolin-humic floc (1 mg C/L) Growth Breakage Regrowth
100 30
b
25
10 5
Alum-kaolin-humic floc (3 mg C/L)
10
100
c
25
Growth Breakage Regrowth
20
10
0
0 1000
Alum-kaolin-humic floc (10 mg C/L)
10000
100
d
25
Growth Breakage Regrowth
20 % volume
15 10
0
0 1000
10000
1000
Ferric-kaolin-humic floc (10 mg C/L) Growth Breakage Regrowth
10000
h
10 5
Floc size (mm)
g
15
5
100
Ferric-kaolin-humic floc (3 mg C/L) Growth Breakage Regrowth
10000
10 5
100
1000
15
5
20
f
15
10000
% volume
% volume
1000
15
25
Ferric-kaolin-humic floc (1 mg C/L) Growth Breakage Regrowth
10000
0 100
% volume
20
1000
5
0
20
10
10000
% volume
% volume
1000
15
25
15
0 100
20
e
5
0
25
20
Ferric-kaolin-humic floc (0 mg C/L) Growth Breakage Regrowth
100
1000
10000
Floc size (mm)
Fig. 2 e Floc size distributions of the kaolin flocs after growth, breakage, and regrowth after 30 s high shear time. Coagulant dosages were 1.6 mg/L as Al and 5.2 mg/L as Fe respectively.
due to stronger steric repulsion and higher energy barrier among the coated kaolin particles.
3.4.
Changes in floc structure
The floc fractal dimensions D2 are shown in Fig. 3 for alum-kaolin flocs at different humic acid concentration during floc growth,
breakage and regrowth processes. This demonstrated how the degree of floc compaction changed with different shearing patterns for the flocs in the system. The fractal dimension increased with the result that the alum flocs became more densely packed or compact during floc formation process (Chakraborti et al., 2003). There was a distinct decrease from 1.85 to 1.73 in floc fractal dimension at the steady state with
3987
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 8 1 e3 9 9 0
Table 1 e Strength and recovery factors of kaolin flocs formed from different humic acid concentration after short and long shear breakage periods. Strength factor
Recovery factor
15 min
30 sec
15 min
30 sec
18
24
22
91
18
25
17
74
18
24
11
43
20
26
8
47
24
37
10
31
21
37
8
31
22
37
7
19
24
32
7
20
Alum-kaolin-humic floc (0 mg/L) Alum-kaolin-humic floc (1 mg/L) Alum-kaolin-humic floc (3 mg/L) Alum-kaolin-humic floc (10 mg/L) Ferric-kaolin-humic floc (0 mg/L) Ferric-kaolin-humic floc (1 mg/L) Ferric-kaolin-humic floc (3 mg/L) Ferric-kaolin-humic floc (10 mg/L)
the increasing humic acid concentration during floc formation process. Once exposure to high shear, the alum flocs had a considerable reduction of D2 to about 1.6 similar to that during the early growth phase after broken for either short or long shear periods, and once the original shear had been returned to, the floc fractal dimensions of all of the flocs were seen to increase to the initial values for the short shear time (Fig. 3b) while remained unchanged for the long shear time (Fig. 3a) after regrowth. The changes in fractal dimension during the growth-breakage-regrowth processes for the short shear time agrees with findings for the alum flocs and polyDADMAC flocs (Jarvis et al., 2005b). However, current understanding suggests that flocs become more compact upon exposure to increased shear as they break at their weak points and rearrange into more stable structures (Clark and Flora, 1991; Gmachowski, 2002; Selomulya et al., 2001; Spicer et al., 1998; Xiao et al., 2011). The alum flocs did not follow this expected change and the reverse was seen. A possible explanation for these inconsistencies is that the alum flocs under investigation exhibit multilevel structures. This agrees with the findings for activated sludge
1.6
1.2 0
10
20
30
40
50
Time (min)
60
70
80
Effect of humic acid on floc settling velocity
The settling velocities of ferric-kaolin and alum-kaolin flocs are shown in Fig. 4 at different humic acid concentration. The ferrickaolin flocs with no humic acid had faster settling velocities in water that varied from 2.34 to 8.37 mm/s. As a direct comparison, the settling velocities of the ferric flocs were larger than those of the alum flocs. The settling velocities of ferric-kaolin flocs were in the order HA0 > HA3 > HA10 while the settling velocities of alum-kaolin flocs were in the order HA0 z HA3 > HA10. The slopes of the settling velocity related to the size after logelog transformation were 0.94, 1.01, 1.11
a
Alum-kaolin-humic floc (0 mg C/L) Alum-kaolin-humic floc (1 mg C/L) Alum-kaolin-humic floc (3 mg C/L) Alum-kaolin-humic floc (10 mg C/L)
2.0
3.5.
2.4 Fractal dimension D2
Fractal dimension D2
2.4
flocs (Chu and Lee, 2004; Wu et al., 2002). According to the floc formation model, the flocs formed by a multilevel progression: primary particles (2 mm) combined to form dense microflocs (13 mm), which in turn combined to form highly porous flocs (100 mm) (Gorczyca and Ganczarczyk, 1999; Jorand et al., 1995). The compact flocs with size range from several hundred micrometer to several millimeter at the steady state formed possibly by restructuring resulting from further contacts of primary particles, microflocs and porous flocs. When the shear increased to 100 rpm, the compact flocs can break at weak point into primary particles, microflocs and porous flocs (100 mm) by surface erosion and large-scale fragmentation. The fractal dimensions of the flocs after breakage were determined by measuring the flocs with size larger than 100 mm in this study while the 10 mm microflocs with compact structure did not get involved. This resulted in the decrease in fractal dimension of the flocs after breakage. As the shear was reduced to 20 rpm, the primary particles, microflocs separated from large flocs and flocs after breakage had a worse regrowth and did not form the compact flocs for the long shear time while the compact flocs were formed due to a better regrowth for the short shear time. Previous studies indicated that the slight decrease in activated sludge floc size after breakage using pipette did not result in the decrease in fractal dimension (Li and Leung, 2005) that is similar to that of kaolin flocs reformed after 30 s breakage. It therefore seems highly likely that the floc restructuring could be statistically reversible for the short shear time. The difference in fractal dimension after regrowth following the long and short shear time is further evidence of multilevel structure of the flocs.
b
Alum-kaolin-humic floc (0 mg C/L) Alum-kaolin-humic floc (1 mg C/L) Alum-kaolin-humic floc (3 mg C/L) Alum-kaolin-humic floc (10 mg C/L)
2.0
1.6
1.2 0
10
20
30
40
50
60
70
Time (min)
Fig. 3 e The changes in fractal dimension of alum-kaolin flocs at different humic acid concentration during growthbreakage-regrowth process.
3988
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 8 1 e3 9 9 0
Settling velocity (mm·s-1)
10 D3=2.28 D3=2.19
1
D3=1.77
D3=2.14 D3=1.83
D3=1.75
0.1
0.01
Alum-kaolin-humic floc (0 mg C/L) Alum-kaolin-humic floc (3 mg C/L) Alum-kaolin-humic floc (10 mg C/L) Ferric-kaolin-humic floc (0 mg C/L) Ferric-kaolin-humic floc (3 mg C/L) Ferric-kaolin-humic floc (10 mg C/L)
1
2
3
Floc size (mm) Fig. 4 e The settling velocities of the alum and ferric-kaolin flocs at different humic acid concentration.
for ferric flocs and 1.00, 0.97, 1.11 for alum flocs at 0, 3, 10 mg C/L humic acid. This is a finding consistent with a number of other studies (Li et al., 2003). The floc fractal dimensions D3 derived from floc dry mass as a function of floc size were 2.28, 2.19, 2.14 for ferric flocs and 1.83, 1.77, 1.75 for alum flocs at 0, 3, 10 mg C/L humic acid. The measured fractal dimensions of alum flocs are smaller than those reported in other studies partly due to different method used. Different measuring techniques should lead to a similar estimate of the fractal dimension of the flocs if the floc structure could be characterized by a single fractal dimension (Wu et al., 2002). There are several calculating methods of fractal dimension corresponding to different Reynold number range by the relationship of settling velocities and floc sizes (Johnson et al., 1996). It was difficult in calculating the fractal dimensions of kaolin flocs according to the power function of settling velocities and floc sizes because the Reynold number was ranged from 1 to 30 in this study. The differences of the two above methods were not easily estimated. The data above indicated that the settling velocity had a close relationship with the fractal structure. The flocs with more compact structure settled faster than those with more open structure which are consistent with the theoretical hypothesis that the increasing fractal dimension results in more compact structure and consequently faster settling velocity. However, adsorbed NOM on kaolin particle was reported to result in more compact structure in inert electrolyte resulting from increased electrostatic and steric repulsion in previous studies. This does not agree with the findings in this study possibly because of net structure formed by the complexity of humic acid with metal ions or bridging resulting in loose structure.
4.
Conclusions
The particle image velocimetry as a non-intrusive tool was successfully used to record the evolution of floc size distribution and the floc structural changes during floc growth, breakage and regrowth processes. Floc strength and regrowth
of kaolin flocs were estimated by strength factor and recovery factor for two coagulants. The coagulant types had a significant influence on steady-state floc size, floc strength and regrowing capability. The kaolin flocs formed by ferric coagulant were larger and stronger than those formed by alum coagulant for their optimum dosages at same humic acid concentration. The extent of recoverability for ferric-kaolin flocs was lower than that for alum at different humic acid concentration. Kaolin floc regrowth also depended on the breakage modes resulting from different shear times at same coagulant dosage. The flocs being exposed to short time had a better regrowth in that the flocs mainly experienced fragmentation while the flocs being exposed to long shear time experienced fragmentation and surface erosion. The floc structural changes recorded by in situ particle image velocimetry technique after breakage for long and short shear times and regrowth indicated the kaolin flocs exhibited multilevel structure. The amounts of humic acid molecules absorbed on intrafloc particle surface had significant influences on the coagulation and sedimentation. The mean size of kaolin flocs at a steady state decreased with increasing humic acid concentration. The kaolin flocs formed at higher humic acid concentration had worse floc regrowth and worse sedimentation performance. The adsorbed humic acid modified the surface characteristics of kaolin particles with the result that resulted in worse regrowth. The settling test results showed that the ferric-kaolin flocs had higher mass fractal dimensions with densely packed structure with the result that had faster settling velocities than those of the alum-kaolin flocs formed. The kaolin flocs formed at higher humic acid concentration had highly branched and loosely bound structures and consequently had smaller settling velocities of the kaolin flocs for two coagulants.
Acknowledgments This work was supported by the National Natural Science Foundation of China (No. 20721140019, U0773002). The technical assistances of Mr. Keith C. H. Wong and Mr. C. H. Tong, comments and suggestions from anonymous reviewers are greatly appreciated.
references
Abu-Lail, N.I., Camesano, T.A., 2003. Role of ionic strength on the relationship of biopolymer conformation, DLVO contributions, and steric interactions to bioadhesion of Pseudomonas putida KT2442. Biomacromolecules 4 (4), 1000e1012. Bache, D.H., 2004. Floc rupture and turbulence: a framework for analysis. Chemical Engineering Science 59 (12), 2521e2534. Bache, D.H., Johnson, C., McGilligan, J.F., Rasool, E., 1997. A conceptual view of floc structure in the sweep floc domain. Water Science and Technology 36 (4), 49e56. Biggs, C.A., Lant, P.A., 2000. Activated sludge flocculation: on-line determination of floc size and the effect of shear. Water Research 34 (9), 2542e2550.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 8 1 e3 9 9 0
Chakraborti, R.K., Atkinson, J.F., Van Benschoten, J.E., 2000. Characterization of alum floc by image analysis. Environmental Science & Technology 34 (18), 3969e3976. Chakraborti, R.K., Gardner, K.H., Atkinson, J.F., Van Benschoten, J.E., 2003. Changes in fractal dimension during aggregation. Water Research 37 (4), 873e883. Chu, C.P., Lee, D.J., 2004. Multiscale structures of biological flocs. Chemical Engineering Science 59 (8e9), 1875e1883. Clark, M.M., Flora, J.R.V., 1991. Floc restructuring in varied turbulent mixing. Journal of Colloid and Interface Science 147 (2), 407e421. Franchi, A., O’Melia, C.R., 2003. Effects of natural organic matter and solution chemistry on the deposition and reentrainment of colloids in porous media. Environmental Science & Technology 37 (6), 1122e1129. Francois, R.J., 1987. Strength of aluminum hydroxide flocs. Water Research 21 (9), 1023e1030. Gmachowski, L., 2002. Aggregate restructuring and its effect on the aggregate size distribution. Colloids and Surfaces A: Physicochemical and Engineering Aspects 207 (1e3), 271e277. Gmachowski, L., 2003. Mass-radius relation for fractal aggregates of polydisperse particles. Colloids and Surfaces A: Physicochemical and Engineering Aspects 224 (1e3), 45e52. Gmachowski, L., 2008. Free settling of aggregates with mixed statistics. Colloids and Surfaces A: Physicochemical and Engineering Aspects 315 (1e3), 57e60. Gorczyca, B., Ganczarczyk, J., 1996. Image analysis of alum coagulated mineral suspensions. Environmental Technology 17 (12), 1361e1369. Gorczyca, B., Ganczarczyk, J., 1999. Structure and porosity of alum coagulation flocs. Water Quality Research Journal of Canada 34 (4), 653e666. Gregor, J.E., Nokes, C.J., Fenton, E., 1997. Optimising natural organic matter removal from low turbidity waters by controlled pH adjustment of aluminium coagulation. Water Research 31 (12), 2949e2958. Gregory, J., 2004. Monitoring floc formation and breakage. Water Science and Technology 50 (12), 163e170. Gregory, J., 2006. Particles in Water: Properties and Processes. CRC Press, Boca Raton. Gregory, J., 2009. Monitoring particle aggregation processes. Advances in Colloid and Interface Science 147e48, 109e123. Gregory, J., Dupont, V., 2001. Properties of flocs produced by water treatment coagulants. Water Science and Technology 44 (10), 231e236. Jarvis, P., Jefferson, B., Gregory, J., Parsons, S.A., 2005a. A review of floc strength and breakage. Water Research 39 (14), 3121e3137. Jarvis, P., Jefferson, B., Parsons, S.A., 2005b. Breakage, regrowth, and fractal nature of natural organic matter flocs. Environmental Science & Technology 39 (7), 2307e2314. Jarvis, P., Jefferson, B., Dixon, D., Parsons, S.A., 2008. Treatment options and their effect on NOM-coagulant floc structures. Journal American Water Works Association 100 (1), 64e73. Johnson, C.P., Li, X.Y., Logan, B.E., 1996. Settling velocities of fractal aggregates. Environmental Science & Technology 30 (6), 1911e1918. Jorand, F., Zartarian, F., Thomas, F., Block, J.C., Bottero, J.Y., Villemin, G., Urbain, V., Manem, J., 1995. Chemical and structural (2D) linkage between bacteria within activatedsludge flocs. Water Research 29 (7), 1639e1647. Kretzschmar, R., Hesterberg, D., Sticher, H., 1997. Effects of adsorbed humic acid on surface charge and flocculation of kaolinite. Soil Science Society of America Journal 61 (1), 101e108. Kretzschmar, R., Holthoff, H., Sticher, H., 1998. Influence of pH and humic acid on coagulation kinetics of kaolinite: a dynamic light scattering study. Journal of Colloid and Interface Science 202 (1), 95e103.
3989
Lee, S.Y.A., Fane, A.G., Waite, T.D., 2005. Impact of natural organic matter on floc size and structure effects in membrane filtration. Environmental Science & Technology 39 (17), 6477e6486. Li, D-h., Ganczarczyk, J., 1989. Fractal geometry of particle aggregates generated in water and wastewater treatment processes. Environmental Science & Technology 23 (11), 1385e1389. Li, X.Y., Leung, R.P.C., 2005. Determination of the fractal dimension of microbial flocs from the change in their size distribution after breakage. Environmental Science & Technology 39 (8), 2731e2735. Li, X.Y., Logan, B.E., 1995. Size distributions and fractal properties of particles during a simulated phytoplankton bloom in a mesocosm. Deep-Sea Research Part II: Topical Studies in Oceanography 42 (1), 125e138. Li, X.Y., Yuan, Y.A., 2002. Settling velocities and permeabilities of microbial aggregates. Water Research 36 (12), 3110e3120. Li, X.Y., Yuan, Y., Wang, H.W., 2003. Hydrodynamics of biological aggregates of different sludge ages: an insight into the mass transport mechanisms of bioaggregates. Environmental Science & Technology 37 (2), 292e299. Li, T., Zhu, Z., Wang, D.S., Yao, C.H., Tang, H.X., 2006. Characterization of floc size, strength and structure under various coagulation mechanisms. Powder Technology 168 (2), 104e110. Li, T., Zhu, Z., Wang, D.S., Yao, C.H., Tang, H.X., 2007. The strength and fractal dimension characteristics of alum-kaolin flocs. International Journal of Mineral Processing 82 (1), 23e29. McCurdy, K., Carlson, K., Gregory, D., 2004. Floc morphology and cyclic shearing recovery: comparison of alum and polyaluminum chloride coagulants. Water Research 38 (2), 486e494. Meakin, P., 1988. Fractal aggregates. Advances in Colloid and Interface Science 28, 249e331. Mu, Y., Ren, T.T., Yu, H.Q., 2008. Drag coefficient of porous and permeable microbial granules. Environmental Science & Technology 42 (5), 1718e1723. Selomulya, C., Amal, R., Bushell, G., Waite, T.D., 2001. Evidence of shear rate dependence on restructuring and breakup of latex aggregates. Journal of Colloid and Interface Science 236 (1), 67e77. Shi, B.Y., Wei, Q.S., Wang, D.S., Zhu, Z., Tang, H.X., 2007. Coagulation of humic acid: the performance of preformed and non-preformed Al species. Colloids and Surfaces A: Physicochemical and Engineering Aspects 296 (1e3), 141e148. Spicer, P.T., Pratsinis, S.E., 1996a. Coagulation and fragmentation: universal steady-state particle-size distribution. Aiche Journal 42 (6), 1612e1620. Spicer, P.T., Pratsinis, S.E., 1996b. Shear-induced flocculation: the evolution of floc structure and the shape of the size distribution at steady state. Water Research 30 (5), 1049e1056. Spicer, P.T., Pratsinis, S.E., Trennepohl, M.D., Meesters, G.H.M., 1996. Coagulation and fragmentation: the variation of shear rate and the time lag for attainment of steady state. Industrial & Engineering Chemistry Research 35 (9), 3074e3080. Spicer, P.T., Pratsinis, S.E., Raper, J., Amal, R., Bushell, G., Meesters, G., 1998. Effect of shear schedule on particle size, density, and structure during flocculation in stirred tanks. Powder Technology 97 (1), 26e34. Tang, S., Ma, Y., Sebastine, I.M., 2001. The fractal nature of Escherichia coli biological flocs. Colloids and Surfaces B: Biointerfaces 20 (3), 211e218. Tirado-Miranda, M., Schmitt, A., Callejas-Fernandez, J., Fernandez-Barbero, A., 1999. Colloidal clusters with finite binding energies: fractal structure and growth mechanism. Langmuir 15 (10), 3437e3444.
3990
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 8 1 e3 9 9 0
Tirado-Miranda, M., Schmitt, A., Callejas-Fernandez, J., Fernandez-Barbero, A., 2003. Aggregation of protein-coated colloidal particles: interaction energy, cluster morphology, and aggregation kinetics. Journal of Chemical Physics 119 (17), 9251e9259. Walker, H.W., Bob, M.M., 2001. Stability of particle flocs upon addition of natural organic matter under quiescent conditions. Water Research 35 (4), 875e882. Wang, Y.L., Du, B.Y., Dou, X.M., Liu, J., Shi, B.Y., Wang, D.S., Tang, H.X., 2007. Study on the pore surface fractal dimension and surface acid-base properties of natural particles around guanting reservoir. Colloids and Surfaces A: Physicochemical and Engineering Aspects 307 (1e3), 16e27. Wilkinson, K.J., Reinhardt, A., 2005. Contrasting roles of natural organic matter on colloidal stabilization and flocculation in freshwaters. In: Droppo, I.G., Leppard, G.G., Liss, S.N., Milligan, T.G. (Eds.), Flocculation in Natural and Engineered Environmental Systems. CRC Press, Boca Raton. Wilkinson, K.J., Joz-Roland, A., Buffle, J., 1997. Different roles of pedogenic fulvic acids and aquagenic biopolymers on colloid aggregation and stability in freshwaters. Limnology and Oceanography 42 (8), 1714e1724.
Wu, R.M., Lee, D.J., Waite, T.D., Guan, J., 2002. Multilevel structure of sludge flocs. Journal of Colloid and Interface Science 252 (2), 383e392. Xiao, F., Lam, K.M., Li, X.Y., Zhong, R.S., Zhang, X.H., 2011. PIV characterization of flocculation dynamics and floc structure in water treatment. Colloids and Surfaces A: Physicochemical and Engineering Aspects. doi:10.1016/j.colsurfa.2010.11.053. Xu, G.R., Fitzpatrick, C.S.B., Gregory, J., 2008. Floc formation, size distribution, and its transformation detected by online laser particle counter. Separation Science and Technology 43 (7), 1725e1736. Xu, W.Y., Gao, B.Y., Yue, Q.Y., Wang, Y., 2010. Effect of shear force and solution pH on flocs breakage and re-growth formed by nano-Al-13 polymer. Water Research 44 (6), 1893e1899. Yukselen, M.A., Gregory, J., 2004. The reversibility of floc breakage. International Journal of Mineral Processing 73 (2e4), 251e259. Zhan, X.A., Gao, B.Y., Yue, Q.Y., Wang, Y., Wang, Q.A., 2010. Coagulation efficiency of polyaluminum chloride for natural organic matter removal from low specific UV absorbance surface water and the subsequent effects on chlorine decay. Chemical Engineering Journal 161 (1e2), 60e67.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 9 1 e3 9 9 8
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Efficient electricity production and simultaneously wastewater treatment via a high-performance photocatalytic fuel cell Yanbiao Liu a, Jinhua Li a, Baoxue Zhou a,*, Xuejin Li a, Hongchong Chen a, Quanpeng Chen a, Zhongsheng Wang b, Lei Li c, Jiulin Wang c, Weimin Cai a a
School of Environmental Science and Engineering, Shanghai Jiao Tong University, 800 Dongchuan Road, Shanghai 200240, China Laboratory of Advanced Materials, Fudan University, 2005 Songhu Road, Shanghai 200438, China c School of Chemistry and Chemical Engineering, Shanghai Jiao Tong University, 800 Dongchuan Road, Shanghai 200240, China b
article info
abstract
Article history:
A great quantity of wastewater were discharged into water body, causing serious envi-
Received 1 March 2011
ronmental pollution. Meanwhile, the organic compounds in wastewater are important
Received in revised form
sources of energy. In this work, a high-performance short TiO2 nanotube array (STNA)
3 May 2011
electrode was applied as photoanode material in a novel photocatalytic fuel cell (PFC)
Accepted 5 May 2011
system for electricity production and simultaneously wastewater treatment. The results of
Available online 12 May 2011
current work demonstrate that various model compounds as well as real wastewater samples can be used as substrates for the PFC system. As a representative of model
Keywords:
compounds, the acetic acid solution produces the highest cell performance with short-
Photocatalytic fuel cell
circuit current density 1.42 mA cm2, open-circuit voltage 1.48 V and maximum power
TiO2 nanotube array
density output 0.67 mW cm2. The STNA photoanode reveals obviously enhanced cell
Electricity production
performance compared with TiO2 nanoparticulate film electrode or other long nanotubes
Wastewater treatment
electrode. Moreover, the photoanode material, electrolyte concentration, pH of the initial solution, and cathode material were found to be important factors influencing the system performance of PFC. Therefore, the proposed fuel cell system provides a novel way of energy conversion and effective disposal mode of organics and serves well as a promising technology for wastewater treatment. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
With the rapid increase in population and fast development of industries in recent years, large amounts of organic waste discharged into water bodies have caused serious environmental pollution. In 2009, the total discharge amount of wastewater in China was w58.92 billion tons and the relevant amount of chemical oxygen demand (COD) reached w12.77 million tons. Meanwhile, the organic compounds in wastewater are
important sources of energy (Feng et al., 2010; Kaneko et al., 2010; Liu et al., 2010; Strataki et al., 2010). According to the statistics (Japan Energy Society, 2002), the energy of the biowaste in the environment has reached 130 EJ y1, corresponding to a third of the global energy demand of 450 EJ y1. Therefore, finding methods for efficient recovery of the chemical energy and rapid decomposition of the organic matters in wastewater comprise the highest priority for future wastewater treatment technologies.
* Corresponding author. Tel./fax: þ86 21 5474 7351. E-mail address:
[email protected] (B. Zhou). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.05.004
3992
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 9 1 e3 9 9 8
Fig. 1 e Continuous cyclic voltammograms (CV) of STNA electrode (curve 1 and 10 ) and TiO2 nanoparticulate film electrode (curve 2 and 20 ) over five cycles in 0.1 mol LL1 Na2SO4 (a) under UV illumination and (b) in the dark, respectively. The inset is the SEM image of typical STNA (a) and conventional TiO2 nanoparticulate film (b), respectively.
Microbial fuel cell (MFC) was once considered the best proposal for wastewater treatment, which can use bacteria as the catalysts to produce electricity and oxidize organic matters (Logan et al., 2006; Wang et al., 2009). The greatest potential of MFC lies in the use of wastewater as a substrate (fuel), which breaks the traditional concept of sewage treatment and allows combining wastewater treatment with power generation. However, the electron transfer process within MFC devices involves complicated mechanism among different cells or cell systems, which directly lead to poor cell performance (Qian et al., 2010). Furthermore, MFC also possess disadvantages such as complex operation, bacteria cultivation, long start-up time, and stringent working conditions (Logan et al., 2006). TiO2-based photocatalytic oxidation is a promising and efficient process that can be used to degrade various recalcitrant organic pollutants (Fujishima et al., 2008; Kim and Choi, 2010; Zhang et al., 2008). Upon UV illumination, the electrons are excited from the valence band to the conduction band of TiO2, generating electron/hole pairs. The positive holes are powerful oxidants for degrading organic compounds into CO2 and H2O, and the negative electrons are powerful reductants (Choi et al., 2010). By substituting the slow and complex biochemical electron transfer process in traditional MFC with the fast and direct transportation process of photogenerated electrons in photocatalysis (i.e. substituting the microbial anode of MFC with the TiO2 photoanode), the TiO2 photocatalytic process may degrade organic matter and produce photogenerated electrons that pass through the conductive substrate to the cathode. In turn, this forms a TiO2-based photocatalytic fuel cell (PFC) system and the chemical energy of organics can be transformed into electricity accompanied with organic compounds degradation at the expense of incident light. Different from traditional MFC, the generation of electrons in the PFC system comes from photoexcitation, which is a much fast and direct process. It is, therefore, possible to generate electricity and simultaneous decomposition of organic compounds via the PFC system. However, the research on PFC system was in its infancy, there are still many problems need to be solved or improved and the most crucial factor was remaining the photoanode material.
The properties of functional materials are highly dependent on their microstructure. Recently, the highly ordered TiO2 nanotube array (TNA) fabricated by anodization of titanium in HF or [F]-based electrolyte has attracted much attention for its peculiar architecture and remarkable properties (Grimes, 2007; Mor et al., 2006). Within nanotubular microstructures, vertically oriented TiO2 nanotubes directly grown on the electrically conductive Ti substrate, forms a natural Schottky-type contact and provides an unidirectional channel for the rapid transport of photogenerated electrons. Lots of work can be found regarding the photocatalytic applications of TNA into organic compounds degradation (Liu et al., 2008; Xu et al., 2006). Nevertheless, most studies have mainly focused on the efficiency and the extent of mineralization. An equally important aspect of photocatalysis, energy recovery, has not received much attention. Preparation by sonoelectrochemical anodization of a short, robust, and highly ordered TiO2 nanotube array (STNA) with superior electron transfer performance and excellent mechanical stability has been reported (Liu et al., 2009). In this work, the high-performance STNA electrode was applied as a photoanode material in the novel fuel cell system for electricity yield and simultaneously wastewater treatment. Various model compounds and actual wastewater were investigated using the fuel cell system. The results of current work demonstrate that
Fig. 2 e Schematic diagram of working principle of the STNA-based PFC system. The photogenerated electrons flow from the conduction band of TiO2 nanotubes and the holes move toward the surface to generate proton. CB and VB refer to the energy levels of the conduction and valence band of TiO2.
3993
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 9 1 e3 9 9 8
Table 1 e System currentevoltage characteristics of the PFC system using various model compounds and actual wastewater samples as substrate. Organic compounds 1
Voc (V)
Jsc (mA cm2)
JVmax (mW cm2)
FF
Model compound
Na2SO4 (0.1 mol L ) Glucose (0.05 mol L1) Glutamic acid (0.05 mol L1) Nicotinic acid (0.05 mol L1) Acetic acid (0.05 mol L1) Urea (0.05 mol L1) Ammonia (0.05 mol L1)
1.13 1.28 1.34 1.39 1.48 1.41 1.24
0.35 0.83 1.08 0.61 1.42 0.91 0.72
0.12 0.38 0.51 0.30 0.67 0.51 0.37
0.31 0.36 0.35 0.35 0.32 0.40 0.41
Actual wastewatera
Pharmaceutical wastewater (COD ¼ 24,572 mg L1) Petroleum exploiting wastewater (COD ¼ 19,087 mg L1) Dying wastewater (COD ¼ 10,842 mg L1) Chemical plant wastewater (COD ¼ 11,700 mg L1) Original urine solution (COD ¼ 9642 mg L1)
0.88 1.34 1.53 1.11 0.93
1.36 0.98 1.21 0.99 0.61
0.43 0.34 0.50 0.30 0.19
0.36 0.26 0.27 0.27 0.34
a The specific information of various actual wastewater samples are given in Supplementary data.
the PFC system provides a novel way of energy conversion and a much effective disposal mode of organics.
determined via our previously reported thin-cell technology (Zhang et al., 2009).
2.
Materials and methods
3.
Results and discussion
2.1.
Materials
3.1.
Characterization of the photoanode materials
Titanium sheets (0.25 mm thick, 99.9% purity) were supplied by Sumitomo Chemical (Japan). Unless otherwise indicated, reagents were obtained from the Sinopharm Chemical Reagent Company and were used as received. The detailed information regarding the actual wastewater is presented in Supplementary data. Natural urine was collected from male students in Shanghai Jiao Tong University directly after release.
A low magnification SEM image of the titania nanotube array obtained by sonoelectrochemical anodization of titanium in
2.2. Preparation of STNA photoanode and Pt-black/Pt cathode Details of the preparation of STNA have been published in previous work (Liu et al., 2009). The preparation of Pt-black/Pt is carried out by cathodic polarization in the electrolyte solution containing 30 kg m3 H2PtCl6 and 0.2e0.5 kg m3 Pb(CH2COOH)2 at a constant current density (10e40 mA cm2) and at room temperature within 20 min. For comparison, a conventional TiO2 nanoparticulate film was prepared by screen printing technique according to the research work of our co-worker (Wang et al., 2004).
2.3.
Apparatus and methods
The currentevoltage characteristics of the PFC system were studied using a CHI electrochemical analyzer (CHI 660C, CH Instruments, Inc., USA) in a two-electrode system, with a TiO2 photoanode and a Pt-black/Pt cathode. All runs were repeated at least three times at ambient temperature, to check their reproducibility. The circuit current was calculated from the voltage across the external resistance (1 U) which was continuously recorded using the CHI electrochemical analyzer. The output voltage of the PFC was directly measured by a high-precision digital multimeter (Victor 98A, Shenzhen Victor Hi-tech Co., Ltd.). The COD value of the sample was
Fig. 3 e System currentevoltage characteristic (a) and JV product (b) of STNA-based PFC in the presence of an electrolyte solution composed of 0.05 mol LL1 acetic acid D 0.1 mol LL1 Na2SO4 solution under UV illumination.
3994
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 9 1 e3 9 9 8
reveals obviously enhanced photocurrent response and the saturated photocurrent density of STNA electrode is w1.63 times as high as that for TiO2 nanoparticulate film electrode under the same irradiation intensity. Since within the traditional TiO2 nanoparticulate film electrode, most nanoparticles are not in direct contact with glass support, resulting in easy recombination of photogenerated charges and poor mechanical stability of the electrode material (Zhu et al., 2007). While the nanotubular microstructures are perpendicular to the electrically conductive Ti substrate, forming a Schottky-type contact naturally and providing an unidirectional electric channel for the transport of photogenerated electrons (Mor et al., 2006). Furthermore, the reproducibility of the CV curves is much better for the nanotubular electrode. Here, the five continuous CV curves nearly coincided with each other, indicating excellent stability of the STNA electrode, the saturated photocurrent of five continuous CV curves remains nearly constant over the flat stage (0.5e1.8 V). In contrast, continuous CV measurements of the traditional TiO2 nanofilm electrode indicate poor reproducibility and undesirable characteristics with repeated excursions of the saturated photocurrent over time. The observed dark current for both samples may be considered negligible (Fig. 1b). Fig. 4 e Four STNA-based PFC systems in series to light a LED indicative lamp under UV illumination using 0.05 mol LL1 acetic acid D 0.1 mol LL1 Na2SO4 solution as substrate.
HFeH2O electrolyte is shown in the inset of Fig. 1a, which reveals a regularly arranged tube structure with nanotubes w65 nm in diameter and w280 nm in length (Liu et al., 2009). Under ultrasonic wave irradiation, increased mass transfer results in a layer of highly ordered and robust nanotube film with decreased tube length and enhanced mechanical stability. The TiO2 nanoparticulate film possesses a spongelike structure with nanoparticles w23 nm in diameter (Fig. 1b and Supplementary data). Fig. 1a presents the continuous cyclic voltammograms (CV) of STNA electrode and TiO2 nanoparticulate film electrode over five cycles in 0.1 mol L1 Na2SO4 as a function of applied potential under UV illumination. By comparing the CV curves, the STNA electrode
3.2.
Working principle of the PFC system
The process involving TiO2-based photocatalytic degradation of organic compounds in aqueous media concerns energy yield and energy conversion routes. Upon UV irradiation, the STNA electrode produces electron/hole pairs in the conduction band and valence band, respectively. The photogenerated holes oxidize the organic substance R to R0 . Photogenerated electrons move through an external circuit to the cathode. Hydrogen ions generated from photooxidation move toward the cathode by diffusing through the electrolyte solution. These are either reduced by externally arriving electrons producing molecular hydrogen under anaerobic conditions, or interact with oxygen and produce water under aerobic conditions. In both cases, an electric current flows between the anode and the cathode. This reaction system is very attractive, because it leads to energy yield and water cleaning. Under O2 atmosphere, the theoretical maximum voltage of the present fuel cell system can be obtained according to the calculated value based on the redox potential of organic
Fig. 5 e (a) The variation of current of the PFC system by applying 0.05 mol LL1 acetic acid as substrate under UV illumination. Inset is the COD removal performance of acetic acid over 4 h. (b) The variation of the output voltage of the PFC system.
3995
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 9 1 e3 9 9 8
Table 2 e System currentevoltage characteristics of 0.1 mol LL1 Na2SO4 and different model compound solutions as affected by the photoanode materials. Photoanode material
Voc (V)
Jsc (mA cm2)
Jsc/Jsc
JVmax (mW cm2)
JVmax/JVmax
Acetic acid
STNA TiO2 film
1.48 1.36
1.42 0.45
3.15
0.67 0.47
1.43
Glucose
STNA TiO2 film
1.28 1.32
0.83 0.052
0.38 0.06
6.23
Urea
STNA TiO2 film
1.41 1.44
0.91 0.11
0.51 0.13
3.92
Model compounds
þ
þ
(2)
At the cathode, nHþ þ
3.3.
n n O2 þ ne / H2 O 4q ¼ 1:23 V 4 2
(3)
Electricity generation from organic wastewaters
Table 1 lists the system currentevoltage (JeV) characteristics of STNA-based PFC using various model compounds and actual wastewater samples as substrate under aerobic conditions. Fill factor (FF) is the ratio of a real maximum electric power output per theoretical maximum power output and is calculated by the following equation (O’Regan and Gra¨tzel, 1991): FF ¼
JVmax Jsc Voc
(5)
However, the presence of organic materials resulted in the production of much higher photocurrents as compared with their absence. Among the group of model compounds, acetic acid (pH ¼ 2.65) produces the best performance (see Fig. 3), with Jsc, Voc, and JVmax of 1.42 mA cm2, 1.48 V and 0.67 mW cm2, respectively. This may be partly due to the simple molecular structure of acetic acids, which can be easily degraded by the photocatalytic process of TiO2. Other complex organics might necessitate more oxidative steps to achieve complete decomposition and mineralization. Furthermore, from the viewpoint of standard electrode potential, the theoretical redox potential of acetate (CH3COOH/CO2) is ca. 4q ¼ 0.4 V (see Eq. (2)), which is much lower that the theoretical redox potential of O2/H2O (4q ¼ 1.23 V). Therefore, the acetic acid molecule can be readily oxidized by photogenerated holes and produce desirable cell performance. Human wastes, such as ammonia, can also be successfully photodecomposed by the TiO2 nanotubular electrode and present JeV characteristics. Although urea is a refractory organic, it still produces comparable Jsc and JVmax, indicating that urea molecules can be further decomposed into NH3 or [NO3]. To examine the applicability of the PFC system to actual wastewater produced in everyday life, five different types of real wastewater were investigated under UV illumination and aerobic conditions. Various samples of actual wastewater generated electrical power even under low UV intensity (3.8 mW cm2), demonstrating that the chemicals within the wastewater samples can be removed by the PFC system. However, there is no specific regularity between the cell performance of actual wastewater samples and their corresponding COD (or salinity, etc.), which might be ascribed to the complex composition of the actual wastewater. An original urine sample (without the addition of electrolyte) also generated electric power, although the Jsc, Voc and JVmax were much lower than those of other samples. The experimental data in Table 1 provides compelling support for the suitability of STNA-based PFC for producing electrical energy while
(1)
R þ nh /R0 þ nHþ
7.97
2H2 O 4e /4 Hþ þ O2 [ 4q ¼ 1:23 V
compounds (R0 /R) and the redox potential for O2/H2O. The net effect of the system is degradation of organics through the STNA photocatalytic process to produce electrical energy. The main working principle of the PFC system is presented in Eqs. (1)e(3) and Fig. 2. At the photoanode, TiO2 þ hn/h þ e
16.0
(4)
where JVmax is the maximum power density (mW cm2) yield by the fuel cell system as obtained from the JV vs. V plot. Jsc and Voc are short-circuit current density and open-circuit voltage of the PFC system, respectively. The FF then gives the extent of diversion between the actual maximum power density that can be produced by the fuel cell system and the product of Jsc$Voc. The pure electrolyte with no organic additive presents the smallest Voc, Jsc, and JVmax. In this case, oxygen was produced at the photoanode (Eq. (5)), and a concentration cell of oxygen was formed in the PFC system.
Table 3 e System currentevoltage characteristics of 0.05 mol LL1 acetic acid D 0.1 mol LL1 Na2SO4 as affected by the morphological structure of TNA.
STNA Medium TNA Long TNA
L (nm)
D (nm)
W (nm)
G
Voc (V)
Jsc (mA cm2)
JVmax (mW cm2)
280 1500 19,400
65 100 220
5 15 20
26.27 75.02 500.47
1.48 1.56 1.63
1.42 1.04 0.65
0.67 0.31 0.28
3996
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 9 1 e3 9 9 8
Table 4 e System currentevoltage characteristics of STNA-based PFC in 0.05 mol LL1 acetic acid as affected by the concentration of Na2SO4. Concentration of Na2SO4 (mol L1) 0.00 0.01 0.05 0.10 0.50
Voc (V)
Jsc (mA cm2)
FF
JVmax (mW cm2)
1.16 1.37 1.42 1.48 1.46
0.18 0.51 0.67 1.42 1.25
0.31 0.32 0.43 0.32 0.33
0.064 0.22 0.41 0.67 0.60
simultaneously photodecomposition of a large variety of organic waste. To demonstrate visually the electrical energy generation, an example of 0.05 mol L1 acetic acid could light a LED indicative lamp by applying four PFC devices in series under UV illumination, as shown in Fig. 4, since the excitation voltage of the LED indicative lamp requires w4 V. It can be predicted that the present PFC system can respond to visible light and give enhanced cell performance by modifying the STNA photoanode with existing modification technologies (e.g. by depositing noble metal on its surface, sensitizing it with dyes, or doping it with transition metal elements or complexes with matching semiconductors). This work is currently under study in our group and the results will be presented in a future publication. In order to estimate the COD removal performance of the substrates in the PFC system, acetic acid (0.05 mol L1) was selected as a typical organic compound. Fig. 5a (inset) presents the COD removal curve via the novel PFC system. It is evident that the COD removal of acetic acid followed the L-H model satisfactorily and w35% was mineralized with 4 h. The Coulombic efficiency, EC, is defined as the ratio of total Coulombs actually transferred to the anode from the substrate, to maximum possible Coulombs if all substrate removal produced current. The total Coulombs obtained is determined by integrating the current over time, so that the Coulombic efficiency for the proposed PFC evaluated over a period of t can be calculated as (Logan et al., 2006):
substrate was measured to be 17.2 C by integrating the currentetime curve shown in Fig. 5a, which leading to EC equals to w15%. The EC of the employed PFC system was relatively low, and the main reason can be ascribed to the high COD load in the substrate solution. Moreover, the EC would have been better if, for instance, the distance between the two electrodes was shorter, the higher light intensity adopted, larger photoanode area exposed to illumination, and lower COD load in the substrate. All these factors affect cell performance and the organic pollutants removal. The change of output voltage of the composite system with UV illumination is shown in Fig. 5b. The output voltage increased sharply initially, and then decreased slowly. The initial output voltage increment was attributed to the application of UV illumination contributed positively to the output voltage. During the following voltage decay, this can be ascribed to the generation of photogenerated electrons and the removal of substrates in the reaction system. Moreover, with the aim to examine the stability of the PFC system, the output voltage was measured over 3 repeated PFC cycles (Supplementary data). It is evident that the reproducibility of the output voltage curves is excellent for the PFC system.
3.4.
Parameters affecting the PFC performance
3.4.1.
Effect of photoanode materials
where M ¼ 32, the molecular weight of oxygen, F is Faraday’s constant, b ¼ 4 is the number of electrons exchanged per mole of oxygen, v is the volume of liquid in the reaction system, and DCOD is the change in COD over time t. In the above case, w35% of the acetic acid was degraded over 4 h reactions and the total Coulombs actually transferred to the anode from the
The effect on PFC performance of different photoanode materials was tested under an O2 atmosphere with the same initial concentration of acetic acid, glucose, and urea (0.05 mol L1), as shown in Table 2. The results of current work demonstrate that the STNA electrode evidently enhanced cell performance compared with TiO2 nanoparticulate film. The Jsc and JVmax produced by STNA electrode are found to be 3.15e16.0 (Jsc/Jsc) and 1.43e6.23 JVmax/JVmax times higher than those of the respective TiO2 nanoparticulate electrode in various organic compounds solutions. This result is line with the enhanced photoelectrochemical performance of the nanotubular electrode given in Fig. 1 and the reason can be ascribed to the poor mechanical stability and high recombination rate of the TiO2
Table 5 e System currentevoltage characteristics of STNA-based PFC in 0.05 mol LL1 acetic acid D 0.1 mol LL1 Na2SO4 as affected by the pH of the initial solution.
Table 6 e System currentevoltage characteristics of STNA-based PFC in 0.05 mol LL1 acetic acid D 0.1 mol LL1 Na2SO4 as affected by the cathode materials.
Zt M Idt EC ¼
0
(6)
FbvDCOD
pH
Voc (V)
Jsc (mA cm2)
FF
JVmax (mW cm2)
2.65 3.96 5.12 7.13
1.48 1.38 1.27 1.22
1.42 1.24 0.92 0.90
0.32 0.30 0.34 0.33
0.67 0.51 0.40 0.37
Cathode material
Voc (V)
Jsc (mA cm2)
FF
JVmax (mW cm2)
Ti Pt Pt-black/Pt
1.19 0.96 1.48
0.22 0.41 1.42
0.13 0.14 0.32
0.035 0.065 0.67
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 9 1 e3 9 9 8
3997
nanoparticulates electrode. As for different performances of the same photoanode material in different substrates (organic solutions), which may be ascribed to different molecular structures of the organics and the relevant degradation mechanisms still need further research. The roughness factor, i.e. the physical surface area of the film per unit of projected area, measures the internal surface area of the electrode materials and is of crucial significance in photochemical applications (Shankar et al., 2009). In a typical TNA-based photocatalytic process, the amount of light harvested by the electrode film is directly related to the film roughness factor. Herein, assuming an idealized nanotubular structure of inner diameter D, wall thickness W and tube length L, the purely geometric roughness factor G of the nanotubes is calculated as:
the electrolyte solution and varying levels of concentration (0, 0.01, 0.05, and 0.1 mol L1) were used to study the influence of electrolyte concentration (Table 4). Due to the fact that acetic acid can be easily used, this substance was chosen for the rest of the experiments. The system performance of the fuel cell increased with increasing electrolyte concentration from 0.0e0.1 mol L1. The maximum power densities were 0.064, 0.22, 0.41, and 0.67 mW cm2 for the electrolyte concentrations of 0.0, 0.01, 0.05 and 0.1 mol L1, respectively. This phenomenon is explained by the increased conductivity of the solutions, which contributes positively to the photogenerated electron transport between the TiO2/solution interface. Further increase of the electrolyte solution to 0.5 mol L1 leads to a decrement of system performance. Therefore, there is an optimum electrolyte concentration applied in the PFC system.
oi h .npffiffiffi 3ðD þ 2WÞ2 þ 1 G ¼ 4pLfD þ Wg
3.4.3.
(7)
Table 3 summarizes the structural parameters of three different kinds of TNA photoanodes (Supplementary data) and compares their corresponding PFC characteristics in 0.05 mol L1 acetic acid and 0.1 mol L1 Na2SO4 solution. The medium TNA was prepared by anodization in a solution containing 0.1 mol L1 KF, 1 mol L1 NaHSO4, and 0.2 mol L1 trisodium citrate. Samples anodized at 20 V for 8 h achieved a nanotube layer of w1.5 mm in length and w100 nm in diameter. While samples anodized at 40 V for 50 h in dimethyl sulfoxide (99.8%) and 5 wt% HF mixture results much longer nanotubes w19.4 mm in length and w220 nm in diameter. It is evident that the Voc was found increased with the increment of roughness factor G (or tube length L), while the Jsc and JVmax were quite the contrary. Literatures (Liu et al., 2008; Zhang et al., 2009) have also demonstrated that the increase in length of the nanotubes may not contribute positively to the performance of the electrode materials and the short nanotubes prepared in inorganic electrolytes (e.g. HFeH2O) were reported to possess enhanced photochemical properties contrarily. The increase in tube length (or tube diameter) favors the increment in the surface area of TNA, which can improve the light harvesting capability of the electrode materials. However, this will also increase the transport resistance for photogenerated electrons and the recombination rate between photogenerated charges. Moreover, the increase in tube length will lead to decreased mechanical stability of the electrode material and broken tubes and other debris from the organic anodization bath will readily block the top surface of the nanotube film (Paulose et al., 2006), these will inevitably detrimental to the electron transfer performance as well as the photochemical reactivity of the electrode materials. Therefore, although the short tube length of STNA electrode is not favorable for incident light absorption, the superior electron transfer properties, low recombination rate and excellent mechanical stability make it an ideal electrode material for application as a photoanode material in PFC.
3.4.2.
Effect of electrolyte concentration
In a typical photocatalytic process, the selection and concentration of the electrolyte solution is of significant importance. In this study, sodium sulphate was selected as
Effect of pH of initial solution
The pH of a solution is one of the factors known to influence the rate of degradation of organic compounds in the photocatalytic process. As such, pH is also an important operational parameter in wastewater treatment. Table 5 demonstrates the influence of pH on STNA-based PFC system using different pH levels (2.65, 3.6, 5.12 and 7.13). The maximum power density of the present system decreased with the increase of the reaction solution pH. The photocatalytic process of acetic acid performed at pH values of 2.65 and 7.13 resulted in a reduction of the maximum power density from 0.67 to 0.37. In strong acid conditions, the surface of the nanotubular material was positively charged and was favorable for the degradation of negatively charged donors (acetic acid).
3.4.4.
Effect of cathode material
To estimate the effect of cathode materials on system performance, the present work compared JV values and maximum power density of the fuel cell system made with three different kinds of cathodes (Table 6). The experimental results indicate that the cathode had a major influence on system efficiency. Pure titanium foil and pure Pt both gave much poorer performance than the Pt-black coated Pt electrode. The marked increase of overall efficiency could be attributed to the large active site produced by Pt-black, which favors O2 adsorption as well as the transport of photogenerated electrons.
4.
Conclusion
In summary, various model compounds and actual wastewater samples were applied as “fuel” in a novel PFC system, consisting of high-performance STNA photoanode and Ptblack/Pt cathode, to produce electricity at the expense of UV light. The STNA photoanode evidently enhanced cell performance compared with TiO2 nanoparticulate film electrode as well as other long nanotubes electrode. Various parameters affecting the PFC performance were studied. The STNA-based PFC system serves well as a promising technology for wastewater treatment and the present fuel cell system could utilize visible light in the near future by combining with appropriate low band-gap semiconductors to generate electrical power under solar light illumination.
3998
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 9 1 e3 9 9 8
Acknowledgement The authors would like to acknowledge the Science and Technology Commission of Shanghai Municipality (0952nm01800, 08JC1411300), the State Key Development Program for Basic Research of China (Grant No. 2009CB220004), the National High Technology Research and Development Program of China (Grant No. 2009AA063003), and Shanghai Tongji Gao Tingyao Environmental Science and Technology Development Foundation for financial support. The authors would like to thank Dr. Xianwei Liu and Dr. Xin wang for their valuable technical discussions.
Appendix. Supplementary data Supplementary data associated with this article can be found in the online version, at doi:10.1016/j.watres.2011.05.004.
references
Choi, W., Yeo, J., Ryu, J., Tachikawa, T., Majima, T., 2010. Photocatalytic oxidation mechanism of As(III) on TiO2: unique role of As(III) as a charge recombinant species. Environ. Sci. Technol. 44, 9099e9104. Feng, Y.J., Lee, H., Wang, X., Liu, Y., He, W., 2010. Continuous electricity generation by a graphite granule baffled air-cathode microbial fuel cell. Bioresour. Technol. 101, 632e638. Fujishima, A., Zhang, X.T., Tryk, D.A., 2008. TiO2 photocatalysis and related surface phenomena. Surf. Sci. Rep. 63, 515e582. Grimes, C.A., 2007. Synthesis and application of highly ordered arrays of TiO2 nanotubes. J. Mater. Chem. 17, 1451e1457. Japan Energy Society, 2002. Biomass Handbook. Ohmusha, Tokyo. Kaneko, M., Suzuki, S., Ueno, H., Nemoto, J., Fujii, Y., 2010. Photoelectrochemical decomposition of bio-related compounds at a nanoporous semiconductor film photoanode and their photocurrent-photovoltage characteristics. Electrochim. Acta 55, 3068e3074. Kim, J., Choi, W., 2010. Hydrogen producing water treatment through solar photocatalysis. Energy Environ. Sci. 3, 1042e1045. Liu, X.W., Sun, X.F., Huang, Y.X., Sheng, G.P., Zhou, K., Zeng, R.J., Dong, F., Wang, S.G., Xu, A.W., Tong, Z.H., Yu, H.Q., 2010. Nano-structured manganese oxide as a cathodic catalyst for enhanced oxygen reduction in a microbial fuel cell fed with a synthetic wastewater. Water Res. 44, 5298e5305. Liu, Y.B., Zhou, B.X., Li, J.H., Gan, X.J., Bai, J., Cai, W.M., 2009. Preparation of short, robust and highly-ordered TiO2 nanotube arrays and their applications as electrode. Appl. Catal. B: Environ. 92, 326e332.
Liu, Z.Y., Zhang, X.T., Nishimoto, S., Jin, M., Tryk, D.A., Murakami, T., Fujishima, A., 2008. Highly ordered TiO2 nanotube arrays with controllable length for photoelectrocatalytic degradation of phenol. J. Phys. Chem. C 112, 253e259. Logan, B.E., Hamelers, B., Rozendal, R.A., Schroder, U., Keller, J., Freguia, S., Aelterman, P., Verstraete, W., Rabaey, K., 2006. Microbial fuel cells: methodology and technology. Environ. Sci. Technol. 40, 5181e5192. Mor, G.K., Varghese, O.K., Paulose, M., Shankar, K., Grimes, C.A., 2006. A review on highly ordered, vertically oriented TiO2 nanotube arrays: fabrication, material properties, and solar energy applications. Sol. Energy Mater. Sol. Cells 90, 2011e2075. O’Regan, B., Gra¨tzel, M., 1991. A low-cost, high-efficiency solar cell based on dye-sensitized colloidal TiO2 films. Nature 353, 737e740. Paulose, M., Shankar, K., Yoriya, S., Prakasam, H.E., Varghese, O. K., Mor, G.K., Latempa, T.A., Fitzgerald, A., Grimes, C.A., 2006. Anodic growth of highly ordered TiO2 nanotube arrays to 134 mm in length. J. Phys. Chem. B 110, 16179e16184. Qian, F., Wang, G.M., Li, Y., 2010. Solar-driven microbial photoelectrochemical cells with a nanowire photocathode. Nano Lett. 10, 4686e4691. Shankar, K., Basham, J.I., Allam, N.K., Varghese, O.K., Mor, G.K., Feng, X.J., Paulose, M., Seabold, J.A., Choi, K.S., Grimes, C.A., 2009. Recent advances in the use of TiO2 nanotube and nanowire arrays for oxidative photoelectrochemistry. J. Phys. Chem. C 113, 6327e6359. Strataki, N., Antoniadou, M., Dracopoulos, V., Lianos, P., 2010. Visible-light photocatalytic hydrogen production from ethanolewater mixtures using a PteCdSeTiO2 photocatalyst. Catal. Today 151, 53e57. Wang, Z.S., Kawauchi, H., Kashima, T., Kashima, T., Arakawa, H., 2004. Significant influence of TiO2 photoelectrode morphology on the energy conversion efficiency of N719 dye-sensitized solar cell. Coord. Chem. Rev. 248, 1381e1389. Wang, X., Cheng, S.A., Feng, Y.J., Merrill, M.D., Saito, T., Logan, B. E., 2009. Use of carbon mesh anodes and the effect of different pretreatment methods on power production in microbial fuel cells. Environ. Sci. Technol. 43, 6870e6874. Xu, C., Killmeyer, R., Gray, M.L., Khan, S.U.M., 2006. Photocatalytic effect of carbon-modified n-TiO2 nanoparticles under visible light illumination. Appl. Catal. B: Environ. 64, 312e317. Zhang, J.L., Zhou, B.X., Zheng, Q., Li, J.H., Bai, J., Liu, Y.B., Cai, W.M., 2009. Photoelectrocatalytic COD determination method using highly ordered TiO2 nanotube array. Water Res. 43, 1986e1992. Zhang, M., Chen, C.C., Ma, W.H., Zhao, J.C., 2008. Visible-lightinduced aerobic oxidation of alcohols in a coupled photocatalytic system of dye-sensitized TiO2 and TEMPO. Angew. Chem., Int. Ed. 47, 9730e9733. Zhu, K., Neale, N.R., Miedaner, A., Frank, A.J., 2007. Enhanced charge-collection efficiencies and light scattering in dyesensitized solar cells using oriented TiO2 nanotubes arrays. Nano Lett. 7, 69e74.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 9 9 e4 0 0 7
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Ozonation of secondary treated wastewater reduces ecotoxicity to Gammarus fossarum (Crustacea; Amphipoda): Are loads of (micro)pollutants responsible? Mirco Bundschuh*, Ralf Schulz Institute for Environmental Sciences, University of Koblenz-Landau, Landau Campus, Fortstrasse 7, D-76829 Landau, Germany
article info
abstract
Article history:
Ozone application is an effective tool to reduce loads of (micro)pollutants in wastewater,
Received 4 September 2010
however, its ecotoxicological implications are largely unknown. Therefore, the feeding
Received in revised form
rates of a leaf-shredding invertebrate (Gammarus fossarum) exposed to secondary (¼non-
13 April 2011
ozone) or ozone treated wastewater were investigated to assess potential ecotoxicological
Accepted 6 May 2011
effects. Two repetitive experiments resulted in significantly higher feeding rates for gam-
Available online 17 May 2011
marids exposed to ozone compared to non-ozone treated wastewater sampled from a treatment plant equipped with a full-scale ozonation. A further experiment confirmed
Keywords:
these results also for wastewater from the same treatment plant, when ozonation was
Pharmaceuticals
conducted at the lab-scale. However, the deviations in dissolved organic carbon profiles of
Ozone
ozone and non-ozone wastewater did not seem to be the driving factor for the effects
By-products
observed. Two additional experiments displayed on the one hand a higher feeding rate of
Solid phase extraction
G. fossarum if exposed to ten-fold enriched eluates from solid phase extraction cartridges
Gammarus
loaded with ozone compared to non-ozone treated wastewater. On the other hand, the
Feeding assay
mean feeding rate of gammarids exposed to non-ozone treated wastewater, which contained hardly any (micro)pollutants (i.e. pharmaceuticals), was at the same level as wastewater from the same source additionally treated with ozone. These results suggest that not an alteration in the organic matrix but a reduction in the load of micropollutants most likely triggered the effects in the bioassay applied. Hence, the feeding rate of G. fossarum appears to be a well-suited bioassay to indicate alterations in ecotoxicological properties of wastewater due to the application of advanced oxidation processes like ozonation. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
Wastewater treatment plants (WWTPs) equipped with secondary methods, i.e. mechanical and biological treatment, are not capable of removing all contaminants present. Such contaminants or (micro)pollutants are detected frequently at concentrations up to a few mg/L in surface and even
groundwater bodies (Daughton and Ternes, 1999; Fent et al., 2006). Thus, wastewater can be considered as one of the major sources of (micro)pollutant discharge into aquatic ecosystems (Schwarzenbach et al., 2006). Therefore, (micro)pollutants may pose on the one hand, potential health risks, when humans are exposed indirectly via drinking water (Webb et al., 2003). On the other hand, some classes of (micro)
* Corresponding author. Tel.: þ49 6341 280 31322; fax: þ49 6341 280 31326. E-mail address:
[email protected] (M. Bundschuh). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.05.007
4000
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 9 9 e4 0 0 7
Nomenclature WWTP TZW
wastewater treatment plant water technology centre
pollutants, i.e. antimicrobials, may lead to multiple resistances of wild bacteria in receiving waters. Since resistance genes can be transferred to pathogenic bacteria, undesirable consequences for the control of infectious diseases can ensue (Costanzo et al., 2005). Besides consequences for human health, also ecosystems may be considerably affected. Below a WWTP effluent in Southern Germany, for instance, the reproduction of a key species in leaf litter breakdown, Gammarus fossarum, was impaired (Ladewig et al., 2006). Furthermore, single substances or substance classes, like antibiotics, have the potential to affect the activity of leaf associated microbial communities (Maul et al., 2006) and finally can alter the nutritious quality of leaves (Bundschuh et al., 2009), which may indirectly influence leaf litter breakdown. The European Water Framework Directive requires a good status of surface waters in terms of quantity and quality (¼chemical and ecological) (European Commission, 2000). End of pipe technologies might be useful in the medium term to achieve these requirements by reducing the release of (micro)pollutants via point sources like WWTP effluents. The application of e.g. ozone seems to be an economically feasible and technically realistic technology (Joss et al., 2008). A full-scale ozonation at the WWTP Wu¨eri located next to Zurich, Switzerland, was recently assessed from a chemical viewpoint displaying reduced concentrations of (micro)pollutants like pharmaceuticals and pesticides (Hollender et al., 2009). This reduced load of (micro) pollutants and any alteration in dissolved organic carbon (DOC), as for instances shown by Hammes et al. (2007), is mainly caused by the oxidative nature of ozone, which reacts with certain functional groups that exhibit the ability to donate electrons (Nakada et al., 2007; von Gunten, 2003). Besides organic also inorganic (micro)pollutants may be removed from the water phase by ozone application. Soluble iron and manganese, for instance, are converted into insoluble solids namely iron hydroxide and manganese dioxide, respectively. Both can finally be removed from the water phase by filtration (El Araby et al., 2009). Hence, a sand filter would be an appropriate tool to remove these insoluble solids. Furthermore, such a sand filter, if established below the ozonation step within a WWTP, seems to remove toxic metabolites formed during ozonation, e.g. aldehydes, carboxylic acids etc., which was recently demonstrated by a fish early life stage test also conducted at the WWTP Wu¨eri (Stalter et al., 2010). However, Stalter et al. (2010) could not confirm the decreased load of (micro)pollutants due to ozone application (see Hollender et al., 2009) by a reduction in toxicity at the level of whole organisms. Hence, the main objective of the present study was on the one hand to assess the ecotoxicity of wastewater treated with ozone by using the feeding rate of the leaf shredding amphipod G. fossarum. This endpoint has recently been shown
ANOVA analysis of variance SPE solid phase extraction DOC dissolved organic matter
to display adverse effects of secondary treated wastewater (Bundschuh et al., 2011c). Therefore, Gammarus was exposed in whole effluent toxicity tests to ozone treated or nonozone treated wastewater or various mixtures of both. On the other hand, the present study aimed at assessing the hypothesis that the reduced load of micropollutants and thus not an alteration in the organic matrix triggered the observed effects. Therefore, three further experiments were set up. The first, evaluated whether or not alterations in the DOC-profile may be a driving factor. A further experiment assessed ten-fold enriched eluates obtained via solid phase extraction (SPE) from ozone treated and non-ozone treated wastewater (cp. Escher et al., 2008). This procedure was done to ensure that any effect displayed by the bioassays is likely to be caused by the fraction of micropollutants purified via SPE. Finally, potential implications of any alteration in the organic matrix in general due to application of ozone for the sublethal response of G. fossarum were assessed. Therefore, secondary treated wastewater from a sequencing batch reactor receiving wastewater from a population equivalent of 16 people using no pharmaceutical compounds, which therefore contains hardly any pharmaceuticals or plant protection products, was used.
2.
Material and methods
2.1.
Experiments I and II
24-hour wastewater composite samples were taken in the middle of April and June 2008 after secondary treatment (¼ non-ozone treated) and below the sand filter (¼ ozone treated; 0.60 and 0.67 mg O3/mg DOC, respectively) at WWTP Wu¨eri (Fig. 1). This WWTP is located next to Zurich and treats wastewater of a population equivalent of 25,000 (10,000 from industry). Its average discharge is between 70 and 120 liters per second. Since the water quality parameters of wastewater sampled below ozonation were at the same level as secondary treated wastewater, exclusively the latter are reported in Table 1. The composite sample was taken proportional to the discharge and stored in stainless steel containers. Subsequently, wastewater samples were filtered (Whatman, GF/6, pore size <1 mm) to remove particulate organic matter potentially present and aerated for another 24 h, although the aeration might modify the properties of wastewater samples. In experiment I, river water from the Hainbach (49 140 N; 8 030 E) e a near natural stream upstream devoid of any settlement, WWTP effluent or agricultural activity e served as control. In addition, the test organisms originated from this stream. Gammarids were exposed to ozone treated and non-ozone treated wastewater samples. In experiment II, ozone treated and non-ozone treated wastewater were mixed containing
4001
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 9 9 e4 0 0 7
sampling
sampling
preliminary sedimentation
activated sludge
final sedimentation
ozonation
sand filter
Fig. 1 e Schematic diagram of the treatment processes at the WWTP Wu ¨ eri. The 24-hours composite samples for experiments I, II, III and IV were sampled after secondary treatment ([ non-ozone treated) and below the sand filter ([ ozone treated) as indicated by the arrows.
0, 25, 50, 75 and 100% of ozone treated wastewater. For both feeding trials, 20 replicates per treatment were set up.
2.2.
Experiment III
A 24-hour wastewater composite sample was taken in the middle of March 2010 proportionally to the discharge below the final sedimentation (¼ non-ozone treated) at WWTP Wu¨eri (Fig. 1). The sample was filtered to remove particulate organic matter (Whatman, GF/6, pore size < 1 mm) and stored in stainless steel containers. One half of the sample volume (20 L) was treated with an effective ozone concentration of 1.23 0.06 mg O3/mg DOC (mean SD, n ¼ 2) at the Water Technology Centre (TZW) in Karlsruhe, Germany. The other half was left untreated and served as reference to evaluate the success of the applied ozone treatment. The ozone concentration was achieved by injecting air containing approximately 31 mg O3/L for 10.5 min. After a contact time of 15 min, the batches were purged for 10 min with a stream of nitrogen to remove any residual ozone and thus to stop ozone-mediated oxidation (cp. Bundschuh et al., 2011a). Success of this procedure was determined by using the indigo-blue-method (DIN, 2007). Subsequently, ten litres of both ozone treated and non-ozone treated wastewater were stored over night at 15 1 C under aeration prior to their use in the experiment. The remaining volume of 10 litres of ozone treated and non-ozone treated wastewater, respectively, were purified with SPE as described in detail by Escher et al. (2008). Briefly, both types of wastewater were acidified and extracted in batches of 600 mL each with LiChrolut SPE-cartridges (100 mg LiChrolut EN plus 250 mg LiChrolut RP-C18, Merck, Darmstadt, Germany). Subsequently, cartridges were dried under a stream of nitrogen and eluted by acetone and
methanol into silanised glass vials. The eluates were evaporated also under a gentle stream of nitrogen to approximately 500 mL and filled up with ethanol to exactly 1000 mL. 333.3 mL of the eluate were transferred to 200 mL river water from the Hainbach aiming to achieve the same concentration of the purified substances as in the original wastewater. Finally, ecotoxicity of the two SPE-extracts and whole effluent samples (ozone and non-ozone-treated wastewater) were assessed together with river water from the Hainbach, which served as control. The number of replicates used in the feeding trial was again 20 per treatment.
2.3.
Experiment IV
A 24-hour wastewater composite sample was taken at the beginning of March 2011 below the final sedimentation of WWTP Wu¨eri and processed as described in chapter 2.2. One half of the sample volume (50 L) was treated at the TZW with an effective ozone concentration of 1.17 0.06 mg O3/mg DOC (mean SD, n ¼ 5) using a similar setting as described for experiment III. The other half was used to evaluate the success of the applied ozone treatment. Both ozone and non-ozone treated wastewater were purified with SPE as described above. Both SPE-extracts obtained were transferred to river water, however, enriched by a factor of ten. Finally, these ten-fold enriched samples were assessed together with river water from the Hainbach, which served as control. The number of replicates used in the feeding trial was 20 per treatment.
2.4.
Experiment V
A 6-hour wastewater composite sample was taken following the sedimentation step from a sequencing batch reactor (for
Table 1 e Quality parameter of secondary treated wastewater for each of the experiments. Parameter BSB5 (mg/L) CSB (mg/L) Total P (mg/L) NH4eN (mg/L) NO2eN (mg/L) NO3eN (mg/L) pH DOC (mg/L)
Experiment I 2.40 16.33 0.18 0.15 0.06 5.40 7.38 4.03
(0.86) (2.73) (0.03) (0.21) (0.06) (2.31) (0.07) (0.13)
Experiment II 2.38 17.17 0.21 0.05 0.05 9.75 7.48 5.31
(0.29) (2.71) (0.02) (0.02) (0.08) (2.19) (0.12) (0.21)
Experiment III 3.72 (1.06) 17.40 (2.30) 0.30 (0.06) 0.05 (0.01) 0.02 (0.02) 9.48 (2.12) 7.45 (0.08) 5.53 (0.06)
Experiment IV 3.41 16.93 0.23 0.07 0.04 8.22 7.42 5.58
(0.97) (2.49) (0.06) (0.11) (0.06) (2.31) (0.11) (0.30)
Experiment V 6.75 (1.26) 54.0 (13.9) 17.32 (1.51) 0.00 0.03 (0.01) 20.32 (5.68) 7.82 (0.08) 13.43 (0.32)
4002
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 9 9 e4 0 0 7
general wastewater quality parameters, see Table 1) near Kaiserslautern (49 210 N, 7 440 E) that treats wastewater from a small community of 16 people that use no pharmaceuticals (apart from ethinylestradiol) and plant protection products (Stefan Du¨lk, Du¨lk Umwelttechnik, personal communication). This wastewater was filtered as described above and stored in stainless steel containers at the beginning of July 2010. One part of the sample was treated with an effective ozone concentration of 0.61 mg O3/mg DOC also at the TZW in Karlsruhe, Germany, following the method introduced in section 2.2. The other part was left untreated and used to evaluate effects caused by the ozone treated wastewater on the feeding rate of gammarids. As experiment V was conducted to assess whether or not any alteration in organic matrix mediated by wastewater ozonation is responsible for the effects observed during experiments I to III, a power analysis was conducted. This analysis suggested 40-fold replication of the feeding trial to see a significant effect at an effect size of 40% with the variability observed in the former experiments by accepting a type I and II error rate of 0.05. This would finally result in a negative predictive value above 95%. Thus, treatments were tested initially with 45 replicates in order to account for any potential loss of replicates due to mortality.
The amphipod species Gammarus fossarum Koch was chosen as test organism since it occurs at high densities in the headwater of the Furtbach, the receiving stream of the WWTP Wu¨eri. The test organisms were obtained from another near natural stream (Hainbach) close to Landau one week prior to the start of the laboratory feeding trials since the individuals had to be prepared beforehand. Specimens were checked in the laboratory visually for parasites and brooding status. Parasite-infected and breeding animals were excluded from the experiment since parasites may affect gammarids’ behaviour (Pascoe et al., 1995) and breeding females may be more sensitive than non-brooding females or males (McCahon and Pascoe, 1988). Afterwards, the remaining G. fossarum were divided into three size classes using a passive underwater separation technique (Franke, 1977). Only adults with a cephalothorax length between 1.2 and 1.6 mm were used. Subsequently, the test organisms were kept in river water from the Hainbach at 15 1 C until the start of the experiment while preconditioned black alder leaves were provided ad libitum.
2.5.
2.8.
DOC-profile
In the course of experiment III, ozone treated and non-ozone treated wastewater as well as the respective SPE-eluates were assessed for their DOC-profiles via liquid chromatography with organic carbon detection as described by Huber and Frimmel (1996) at the TZW in Dresden, Germany. Two mL of the wastewater samples were directly injected and separated using size exclusion, hydrophobic and ion interaction chromatography. The molecules eluted were then oxidised in an oxygen free environment by radiolytic decomposition of water. An organic carbon detector measured emerging carbon dioxide at an UV wavelength of 254 nm.
2.6.
Preparation of leaf discs
Leaf discs, used for the feeding trials, were prepared as described in detail in Bundschuh et al. (2011c). Briefly, black alder leaves (Alnus glutinosa L. Gaertn.) were collected shortly before leaf fall in October 2007 from a group of trees near Landau, Germany (49 110 N; 8 050 E) and stored frozen at 20 C until further use. After thawing, discs (2.0 cm diameter) were cut from each leaf with a cork borer. To establish a microbial community on the leaf discs, they were conditioned in a nutrient medium together with alder leaves previously exposed in the Rodenbach, Germany (49 330 N, 8 020 E). Following a conditioning period of 10 days the discs were dried at 60 C to constant weight (w24 h), and weighed to the nearest 0.01 mg. The determination of the leaf disc dry weight ensured an accurate measurement of the amphipods feeding rate (cp. Maltby et al., 2002). After being soaked in water from the Hainbach for 24 h, the leaf discs
were assigned randomly to the vessels of the respective treatment.
2.7.
Test organisms
Feeding trial
One specimen of G. fossarum was placed together with two preconditioned leaf discs in a 250-ml-glass beaker filled with 200 ml of river water (with or without SPE-extracts), ozone treated, non-ozone treated wastewater or a mixture of ozone and non-ozone treated wastewater. All beakers were aerated during the whole study duration. For each treatment, the respective number of replicates as mentioned in sections 2.1e2.4 were set up, while mortality did never exceed 10% in any treatment. Five additional beakers per treatment containing only two leaf discs served as a control to account for microbial decomposition and abiotic losses in leaf mass during the feeding trials. This leaf mass loss, which was used to calculate the leaf change correction factor, deviated by a maximum of 2% among treatments. The amphipods, the remaining leaf discs and any leaf tissue shredded off were removed after seven days of exposure, dried and weighed as described above. The feeding rate was expressed in consumed leaf mass (C ) and calculated as follows (Maltby et al., 2000): C¼
Lb ðkÞ Le gt
(1)
where Lb ¼ initial dry mass of the leaf discs, Le ¼ final dry mass of the leaf discs, g ¼ dry mass of G. fossarum, and t ¼ feeding time in days, k ¼ leaf change correction factor given by: P ðLob Loe Þ Lob k¼ n
(2)
where Lob ¼ initial dry mass of the leaf discs, Loe ¼ final dry mass of the leaf discs e both measured in replicates without any G. fossarum present, n ¼ number of replicates.
4003
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 9 9 e4 0 0 7
Data analysis
0.4 0,4
ANOVAs were conducted to judge significant differences between treatments in experiments I and II. This statistical procedure was followed by Tukey’s post-hoc test to identify significant differences between each of the treatment means, namely river water, ozone and non-ozone treated wastewater. A Dunnett’s tests for multiple comparisons was performed in experiment II with the mixture containing 100% ozone treated wastewater as control treatment. In experiments III, IV and V Student’s t-tests were conducted in order to detect significant differences between (i) ozone and non-ozone treated wastewater in each of the experiments and/or (ii) their SPE extracted pendants. All tests were two-sided and performed on unpaired data. The significance level was set at p < 0.05 for all tests.
feeding rate (mg/mg animal/d)
2.9.
0.3 0,3
*
*
0.2 0,2
0.1 0,1
0.0 0,0 100
75
50
25
0
ratio of ozone treated wastewater (%)
3.
Results and discussion
Experiment I showed a significantly reduced feeding rate (approximately 60%e70%) of G. fossarum in non-ozone treated wastewater compared to ozone treated wastewater (Tukey, p ¼ 0.036, n ¼ 20) and river water (Tukey, p < 0.001, n ¼ 20), respectively (Fig. 2). These findings are supported by experiment II where significantly reduced feeding rates of gammarids exposed to a mixture of non-ozone treated and ozone treated wastewater containing equal to or less than 25% of ozone treated wastewater (Dunnett, p < 0.05, n ¼ 18e20; Fig. 3) were observed. Moreover, based on individual replication data, a significant correlation between the increasing proportion of ozone treated wastewater and the mean feeding rate of G. fossarum (Pearson, r ¼ 0.311, p ¼ 0.002, n ¼ 95) was present. These results suggest a reduced ecotoxicity of ozone treated wastewater to G. fossarum. Escher et al. (2009) likewise demonstrated a positive effect of ozone treatment for specific and
feeding rate (mg/mg animal/d)
0.4 0,4
0.3 0,3
0.2 0,2 a a
0.1 0,1 b
0.0 0,0 river water
non-ozone treated
ozone treated
Fig. 2 e Mean (±95% CI) feeding rate of G. fossarum during experiment I exposed to river water ([ control), ozone and non-ozone treated wastewater (ANOVA, p [ 0.001, n [ 20). Different letters denote significant differences between treatments based on Tukey’s test for multiple comparisons.
Fig. 3 e Mean (±95% CI) feeding rate of G. fossarum during experiment II exposed to mixtures containing different proportions of non-ozone and ozone treated wastewater (ANOVA, p [ 0.037, n [ 18e20). Asterisks denote significant differences to the mixture containing 100% ozone treated wastewater based on Dunnett’s test for multiple comparisons.
non-specific toxicity of wastewater from the same WWTP. The authors exposed, inter alia, Vibrio fischeri to eluates from SPE-cartridges loaded with ozone and non-ozone treated wastewater, respectively, and observed a reduced toxicity with increasing ozone concentration for SPE-purified samples. In contrast, Stalter et al. (2010) reported nonsignificant deviations between ozone treated and non-ozone treated wastewater using a fish early life stage test conducted also at WWTP Wu¨eri. However, reduced vitellogenin concentrations in fish exposed to ozone compared to non-ozone treated wastewater were measured. This biomarker endpoint can directly be linked to reduced loads of estrogenic active compounds, however, not to the broad range of (micro)pollutants potentially oxidised by ozonation (cp. Hollender et al., 2009). Maltby et al. (2002) also demonstrated for example impairments in the feeding rate of Gammarus pulex deployed downstream of point-source discharges. This suggests that the present study displays effects, which may be mainly driven by the reduced concentrations of (micro)pollutants in ozone compared to non-ozone treated wastewater (Hollender et al., 2009). However, besides the reduced load of organic (micro) pollutants, other parameters, which may also be altered due to the ozone application, potentially affect the feeding rate of gammarids. Ozonation, for instance, partly disinfects wastewater and hence, modifies the microbial community in ozone treated compared to non-ozone treated wastewater (Joss et al., 2008). Thus, indirect effects on the feeding of G. fossarum, mediated by alterations in the leaf associated microbial community, are conceivable, too (Bundschuh et al., 2009, 2011a; Hahn and Schulz, 2007). However, in these studies the leaf discs were conditioned in presence of inoculums for a leaf associated microbial community for
4004
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 9 9 e4 0 0 7
approximately three weeks. This extended conditioning period was sufficient to affect the palatability of leaves and hence the food selection of gammarids. These effects are most likely caused by shifts in leaf associated microbial communities, while the higher biomass of aquatic hyphomycetes was hypothesised to be the driving factor (see also Arsuffi and Suberkropp, 1989). In the present study, in contrast, the experiments lasted only one week. This is too short to allow a meaningful recolonization of the dried leaf discs that were offered as food by microorganisms (cp. Hieber and Gessner, 2002), especially as aquatic hyphomycetes are not reported in wastewater and no inoculums for a leaf associated microbial community were provided. Hence, the effects observed in experiment I and II on the feeding rate of G. fossarum are unlikely to be caused by alterations in leaf palatability mediated by the microbial recolonization of the leaf discs. But also the organic matrix is altered by ozonation, e.g. larger organic molecules are broken down to smaller ones (Lehtola et al., 2001), and may hence be responsible for the observed effects. Several approaches were followed to assess potential effects of ozone-mediated shifts in organic matrix on the feeding rate of G. fossarum in the present study. The first was to assess whether an alteration in DOC-profile is the driving factor for the direct ecotoxicological effects displayed. The assimilable organic carbon was not assessed in the present study, as the bacterial recolonization on the leaf discs and any resulting indirect effect due to an alteration in the leaf palatability is highly unlikely to affect the gammarids as also discussed above. Therefore, experiment III was conducted with secondary treated wastewater sampled below the final sedimentation (¼ non-ozone treated) at the WWTP Wu¨eri (Fig. 1), which was partly treated with ozone at the lab-scale. This feeding trial confirmed the results of experiment I & II with a significantly higher feeding rate of gammarids exposed to ozone treated compared to non-ozone treated wastewater (t-test, p ¼ 0.003, n ¼ 25; Fig. 4). Despite the higher ozone concentration applied in experiment III, the relative impairment of approximately 70% in feeding rate of gammarids exposed to non-ozone treated wastewater if compared to the respective ozone treated wastewater was at the same level as in experiment I. This indicates that both experiments, irrespective whether the ozonation was conducted at full- or lab-scale, are comparable regarding their ecotoxicological effects. The DOC-profile of ozone treated wastewater exhibited elevated (approximately 30%) concentrations of building block and low molecular weight fatty acids compared to non-ozone treated wastewater (Fig. 5A). Hence, this deviation may have caused the increased feeding rate of G. fossarum in the respective treatment. However, gammarids exposed to eluates from SPE-cartridges loaded with ozone and non-ozone treated wastewater, respectively (cp. Escher et al., 2009), did not show a significant difference in feeding rate from ozone treated wastewater, although the DOC-profiles of both SPEeluates (Fig. 5B) differed remarkably from the profile of the ozone treated wastewater (Fig. 5A). As the differences in the concentrations of building blocks and low molecular weight acids are more pronounced between both SPE-eluates and ozone treated wastewater than between non-ozone treated
Fig. 4 e Mean (±95% CI) feeding rate of G. fossarum during experiment III exposed to river water ([ control), ozone treated and non-ozone treated wastewater as well as SPE-eluates gained from both types of wastewater. Asterisk denotes a significant difference between ozone treated and non-ozone treated wastewater.
and ozone treated wastewater, the deviation in the DOCprofiles displayed in Fig. 5A is most likely not causing the higher feeding rate of gammarids exposed to ozone treated compared to non-ozone treated wastewater. Yet, the results of the feeding trials conducted with SPE-eluates during experiment III differ from those by Escher et al. (2009). Escher et al. (2009) measured a reduction in non-specific toxicity with increasing ozone concentration. However, in contrast to experiment III, they enriched the (micro) pollutants purified by the SPE-method up to an enrichment factor of 85 for bioluminescence (V. fischeri) and algal growth (Pseudokirchneriella subcapitata) bioassays conducted in microtiter plates. Enrichment of purified (micro)pollutants seems to be necessary as the applied SPE-method was optimised for universality rather than specificity, hence, recovery cannot be 100% for each (micro)pollutant (Escher et al., 2009). Obviously, this SPE-method with an enrichment factor of one, as utilized in experiment III, is not able to display the same toxicity as whole effluent toxicity tests used in experiments I and II as well as in the first part of experiment III, likely because not all organic substances were held back by the SPE-cartridges. However, if ten-fold enriched ozone and non-ozone treated wastewater samples were tested during experiment IV, significant differences were displayed between both treatments (t-test, p ¼ 0.0006, n ¼ 20; Fig. 6). The feeding rate of G. fossarum was reduced by approximately 35% if exposed to eluates from SPE-cartridges loaded with non-ozone treated compared to ozone treated wastewater. Thus, it can be hypothesised, that neither the organic matrix nor other parameters like heavy metals are causing the effects on the feeding rate of G. fossarum (Escher et al., 2009). Hence, it can be concluded that the purified organic micropollutants are responsible for the effects observed during experiment IV. Moreover, this suggests that the effects displayed in experiment I to III may also be driven by the load of micropollutants.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 9 9 e4 0 0 7
4005
Fig. 5 e DOC-profiles of (A) ozone treated and non-ozone treated wastewater and (B) extracts from SPE-cartridges loaded with ozone and non-ozone treated wastewater solved in river water.
Experiment V assessed whether an ozone-mediated alterations in organic matrix was the driving factor for the effects displayed by the Gammarus-feeding assays: First, the feeding rate of G. fossarum exposed to river water was not different compared to wastewater sampled following the sedimentation step from a sequencing batch reactor (¼ non-ozone treated wastewater; Fig. 7), which contains hardly any pharmaceuticals and plant protection products (see chapter 2.3 and exemplified by concentrations of psychoactive drugs in the Supplemental Information Table SI1). Moreover, the water quality parameters of wastewater from the sequencing batch reactor, as displayed in Table 1, show that macronutrients are below concentrations causing adverse effects in gammarids (Berenzen et al., 2001). Second, the same experiment showed no deviation (approximately 1%) in the feeding rate
of gammarids if exposed to ozone treated compared to non-ozone treated wastewater (Fig. 7). This absence of deviations between both wastewater treatments suggests this experiment to be an appropriate ozone control, during which even a potential organic matrix effect is considered. Moreover, this non significant difference is also supported by a negative predictive value of above 95% confirming a very high probability that there is no significant effect among both treatments if a non significant test result is obtained. Hence, the results of experiment V indicate, although the wastewater from WWTP Wu¨eri may not be directly comparable to the wastewater from the sequencing batch reactor (Table 1), that the alteration in organic matrix potentially caused by the application of ozone, which was not quantified in experiment V, does not affect the feeding rate of G. fossarum.
4006
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 9 9 e4 0 0 7
Schulz, 2011; Bundschuh et al., 2011b,c). Thus, the increased water quality caused by the application of ozone in secondary treated wastewater has the potential to improve both the chemical and ecological status of surface waters as required by the Water Framework Directive (European Commission, 2000).
4.
Fig. 6 e Mean (±95% CI) feeding rate of G. fossarum during experiment IV exposed to river water ([ control), ten-fold enriched extracts from SPE-cartridges loaded with ozone and non-ozone treated wastewater solved in river water. Asterisk denotes a significant difference between both SPE-eluates.
In summary, the present study clearly displays increased feeding rates of gammarids exposed to ozone treated municipal wastewater from WWTP Wu¨eri. These effects are most likely not caused indirectly or by alterations in organic matrix, which suggests (micro)pollutants to be the trigger. However, more research is needed to further support this hypothesis. Moreover, from an ecological view-point it is worthwhile to mention that higher feeding rates are also accompanied with an increased level of energy reserves and this finally influences growth and reproduction. Hence, positive implications in field populations are likely to occur (Bundschuh and
Conclusion
- Although a direct link between the load of (micro)pollutants and the outcome of the feeding assays was not yet established. The feeding rate of G. fossarum seems to be a suitable ecotoxicological tool to display effects of ozone application on whole organisms even without any sample purification (e.g. SPE). Thus it may help to assess water quality in terms of ecosystem integrity and may provide information on potential benefits of wastewater treatment technologies for ecosystem services (decomposition). - Indirect effects due to alterations in leaf associated microbial communities and deviations in the organic matrix both mediated by ozone may not influence the outcome of the bioassay. - Feeding assays conducted with gammarids carried out with wastewater treated with ozone at lab- and full-scale resulted in comparable findings. Therefore, lab-scale experiments provide a low-cost alternative for an initial ecotoxicological assessment compared to full-scale testing. - Ozone application may be a useful tool to reduce loads of (micro)pollutants released into surface waters and may hence helps to ensure a good status of European water bodies in terms of chemical and biological quality. However, due to the potential formation of more toxic by-products, further research is needed for a reliable ecotoxicological risk assessment.
Acknowledgements
feeding rate (mg/mg animal/d)
0.4 0,40
The authors are grateful to T.A. Ternes and G. Fink for chemical analysis, O. Happel and S. Mertineit for wastewater ozonation at the lab-scale, S. Du¨lk for the wastewater samples used in experiment V and the staff of the WWTP Wu¨eri for their continuous support during the study. J.P. Zubrod, B. Schreiber and T. Bu¨rgi are acknowledged for their support during the laboratory work. The manuscript benefited from comments of J.P. Zubrod and R.B. Scha¨fer on an earlier draft as well as from valuable comments from both the editor and four anonymous reviewers. This research was funded by the Swiss Federal Office for the Environment (FOEN) as part of the project “Strategy MicroPoll” (project number 07.0142.PJ/G341 e 1833).
0.3 0,30
0.2 0,20
0.1 0,10
0.0 0,00
river3,00 water
2,00 treated non-ozone
1,00 ozone treated
Fig. 7 e Mean (±95% CI) feeding rate of G. fossarum during experiment V exposed to river water ([ control), ozone treated and non-ozone treated wastewater containing hardly any pharmaceuticals and pesticides.
Appendix. Supplementary data Supplementary data associated with this article can be found in the online version, at doi:10.1016/j.watres.2011.05.007.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 9 9 9 e4 0 0 7
references
Arsuffi, T.L., Suberkropp, K., 1989. Selective feeding by shredders on leaf-colonizing stream fungi: comparison of macroinvertebrate taxa. Oecologia 79 (1), 30e37. Berenzen, N., Schulz, R., Liess, M., 2001. Effects of chronic ammonium and nitrite contamination on the macroinvertebrate community in running water microcosms. Water Research 35 (14), 3478e3482. Bundschuh, M., Gessner, M.O., Fink, G., Ternes, T.A., So¨gding, C., Schulz, R., 2011a. Ecotoxicologial evaluation of wastewater ozonation based on detritus-detritivore interactions. Chemosphere 82 (3), 355e361. Bundschuh, M., Pierstorf, R., Schreiber, W.H., Schulz, R., 2011b. Positive effects of wastewater ozonation can be displayed by in situ bioassays in the receiving stream. Environmental Science & Technology. doi:10.1021/es104195h. Bundschuh, M., Zubrod, J.P., Schulz, R., 2011c. The functional and physiological status of Gammarus fossarum (Crustacea; Amphipoda) exposed to secondary treated wastewater. Environmental Pollution 159 (1), 244e249. Bundschuh, M., Hahn, T., Gessner, M.O., Schulz, R., 2009. Antibiotics as a chemical stressor affecting an aquatic decomposer-detritivore system. Environmental Toxicology and Chemistry 28 (1), 197e203. Bundschuh, M., Schulz, R., 2011. Population response to ozone application in wastewater: an on-site microcosm study with Gammarus fossarum (Crustacea: Amphipoda). Ecotoxicology 20 (2), 466e473. Costanzo, S.D., Murby, J., Bates, J., 2005. Ecosystem response to antibiotics entering the aquatic environment. Marine Pollution Bulletin 51 (1e4), 218e223. Daughton, C.G., Ternes, T.A., 1999. Pharmaceuticals and personal care products in the environment: agents of subtle change. Environmental Health Perspectives 107 (Suppl. 6), 907e938. DIN, 2007. 38 408-G3-3: Ozon - Indigo-Methode. European Commission, 2000. Directive 2000/60/EC ot the European Parliament and of the Council of 23 October 2000 establishing a framework for Community action in the field of water policy. Official Journal of the European Communities L327. El Araby, R., Hawash, S., El Diwani, G., 2009. Treatment of iron and manganese in simulated groundwater via ozone technology. Desalination 249 (3), 1345e1349. Escher, B.I., Bramaz, N., Ort, C., 2009. Monitoring the treatment efficiency of a full scale ozonation on a sewage treatment plant with a mode-of-action based test battery. Journal of Environmental Monitoring 11 (10), 1836e1846. Escher, B.I., Bramaz, N., Quayle, P., Rutishauser, S., Vermeirssen, E.L., 2008. Monitoring of the ecotoxicological hazard potential by polar organic micropollutants in sewage treatment plants and surface waters using a mode-of-action based test battery. Journal of Environmental Monitoring 10 (5), 622e631. Fent, K., Weston, A.A., Caminada, D., 2006. Ecotoxicology of human pharmaceuticals. Aquatic Toxicology 76 (2), 122e159. Franke, U., 1977. Influence of temperature, oxygen concentration and water flow on respiration of Gammarus fossarum “Experimentelle Untersuchungen zur Respiration von Gammarus fossarum in Abha¨ngigkeit von Temperatur, Sauerstoffkonzentration und Wasserbewegung”. Archiv fu¨r Hydrobiologie 3/4 (Suppl.), 369e411. Hahn, T., Schulz, R., 2007. Indirect effects of antibiotics in the aquatic environment: a laboratory study on detritivore food selection behavior. Human and Ecological Risk Assessment 13 (3), 535e542. Hammes, F., Meylan, S., Salhi, E., Koster, O., Egli, T., von Gunten, U., 2007. Formation of assimilable organic carbon (AOC) and specific natural organic matter (NOM) fractions
4007
during ozonation of phytoplankton. Water Research 41 (7), 1447e1454. Hieber, M., Gessner, M.O., 2002. Contribution of stream detrivores, fungi, and bacteria to leaf breakdown based on biomass estimation. Ecology 83 (4), 1026e1038. Hollender, J., Zimmermann, S.G., Koepke, S., Krauss, M., McArdell, C.S., Ort, C., Singer, H., von Gunten, U., Siegrist, H., 2009. Elimination of organic micropollutants in a municipal wastewater treatment plant upgraded with a full-scale post-ozonation followed by sand filtration. Environmental Science and Technology 43 (20), 7862e7869. Huber, S.A., Frimmel, F.H., 1996. Size-exclusion chromatography with organic carbon detection (LC-OCD): a fast and reliable method for the characterization of hydrophilic organic matter in natural waters. Vom Wasser 86, 277e290. Joss, A., Siegrist, H., Ternes, T.A., 2008. Are we about to upgrade wastewater treatment for removing organic micropollutants? Water Science and Technology 57 (2), 251e255. Ladewig, V., Jungmann, D., Ko¨hler, H.R., Schirling, M., Triebskorn, R., Nagel, R., 2006. Population structure and dynamics of Gammarus fossarum (Amphipoda) upstream and downstream from effluents of sewage treatment plants. Archives of Environmental Contamination and Toxicology 50 (3), 370e383. Lehtola, M.J., Miettinen, I.T., Vartiainen, T., Myllykangas, T., Martikainen, P.J., 2001. Microbially available organic carbon, phosphorus, and microbial growth in ozonated drinking water. Water Research 35 (7), 1635e1640. Maltby, L., Clayton, S.A., Wood, R.M., McLoughlin, N., 2002. Evaluation of the Gammarus pulex in situ feeding assay as a biomonitor of water quality: robustness, responsiveness and relevance. Environmental Toxicology and Chemistry 21 (2), 361e368. Maltby, L., Clayton, S.A., Yu, H., McLoughlin, N., Wood, R.M., Yin, D., 2000. Using single-species toxicity tests, communitylevel response and toxicity identification evaluations to investigate effluent impacts. Environmental Toxicology and Chemistry 19 (1), 151e157. Maul, J.D., Schuler, L.J., Belden, J.B., Whiles, M.R., Lydy, M.J., 2006. Effects of the antibiotic ciprofloxacin on stream microbial communities and detritivorous macroinvertebrates. Environmental Toxicology and Chemistry 25 (6), 1598e1606. McCahon, C.P., Pascoe, D., 1988. Increased sensitivity to cadmium of the freshwater amphipod Gammarus pulex (L.) during the reproductive period. Aquatic Toxicology 13 (3), 183e194. Nakada, N., Shinohara, H., Murata, A., Kiri, K., Managaki, S., Sato, N., Takada, H., 2007. Removal of selected pharmaceuticals and personal care products (PPCPs) and endocrine-disrupting chemicals (EDCs) during sand filtration and ozonation at the municipal sewage treatment plant. Water Research 41 (19), 4373e4382. Pascoe, D., Kedwards, T.J., Blockwell, S.J., Taylor, E.J., 1995. Gammarus pulex (L.) feeding bioassay e effects of parasitism. Bulletin of Environmental Contamination and Toxicology 55 (4), 629e632. 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 (5790), 1072e1077. Stalter, D., Magdeburg, A., Weil, M., Knacker, T., Oehlmann, J., 2010. Toxication or detoxication? In-vivo toxicity assessment of ozonation as advanced wastewater treatment with the rainbow trout. Water Research 44 (2), 439e448. von Gunten, U., 2003. Ozonation of drinking water: part I oxidation kinetics and product formation. Water Research 37 (7), 1443e1467. Webb, S., Ternes, T., Gibert, M., Olejniczak, K., 2003. Indirect human exposure to pharmaceuticals via drinking water. Toxicology Letters 142 (3), 157e167.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 0 8 e4 0 1 8
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Assessment of nitrification in groundwater filters for drinking water production by qPCR and activity measurement W.W.J.M. de Vet a,c,d,*, R. Kleerebezem c, P.W.J.J. van der Wielen b, L.C. Rietveld d, M.C.M. van Loosdrecht b,c a
Oasen Drinking Water Company, PO Box 122, 2800 AC Gouda, The Netherlands KWR Watercycle Research Institute, PO Box 1072, 3430 BB Nieuwegein, The Netherlands c Department of Biotechnology, Delft University of Technology, Julianalaan 67, 2628 BC Delft, The Netherlands d Department of Water Management, Delft University of Technology, Stevinweg 1, 2628 CN Delft, The Netherlands b
article info
abstract
Article history:
In groundwater treatment for drinking water production, the causes of nitrification prob-
Received 8 November 2010
lems and the effectiveness of process optimization in rapid sand filters are often not clear.
Received in revised form
To assess both issues, the performance of a full-scale groundwater filter with nitrification
30 January 2011
problems and another filter with complete nitrification and pretreatment by subsurface
Accepted 8 May 2011
aeration was monitored over nine months. Quantitative real-time polymerase chain
Available online 17 May 2011
reaction (qPCR) targeting the amoA gene of bacteria and archaea and activity measurements of ammonia oxidation were used to regularly evaluate water and filter sand
Keywords:
samples. Results demonstrated that subsurface aeration stimulated the growth of
amoA gene
ammonia-oxidizing prokaryotes (AOP) in the aquifer. Cell balances, using qPCR counts of
Archaea
AOP for each filter, showed that the inoculated AOP numbers from the aquifer were
Biofilter
marginal compared with AOP numbers detected in the filter. Excessive washout of AOP was
Groundwater
not observed and did not cause the nitrification problems. Ammonia-oxidizing archaea
Nitrification
grew in both filters, but only in low numbers compared to bacteria. The cell-specific
qPCR
nitrification rate in the sand and backwash water samples was high for the subsurface
Subsurface aeration
aerated filter, but systematically much lower for the filter with nitrification problems. From this, we conclude that incomplete nitrification was caused by nutrient limitation. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
In the Netherlands and Flanders, direct filtration of anaerobic groundwater over granular material, in most cases silica sand, is the general technique for removal of inorganic compounds,
such as iron, manganese and ammonium, in water treatment plants (WTP). Because no oxidizing chemicals are applied, ammonium is removed by biological oxidation in the nitrification process. Complete removal of ammonium in drinking water treatment is essential to prevent nitrite formation and
Abbreviations: AOA, ammonia-oxidizing archaea; AOB, ammonia-oxidizing bacteria; AOP, ammonia-oxidizing prokaryotes; amoA gene, gene encoding for the a-subunit of the ammonia monooxygenase enzyme; DW, dry weight; qPCR, (quantitative) real-time polymerase chain reaction; WTP, water treatment plant; subsurface aerated filter, filter treating a mixture of subsurface aerated and nonsubsurface aerated groundwater; non-subsurface aerated filter, filter treating non-subsurface aerated groundwater only. * Corresponding author: Oasen Drinking Water Company, PO Box 122, 2800 AC Gouda, the Netherlands. Tel.: þ31 610927947; fax: þ31 152782355. E-mail addresses:
[email protected] (W.W.J.M. de Vet),
[email protected] (R. Kleerebezem), paul.van.der.wielen@ kwrwater.nl (P.W.J.J. van der Wielen),
[email protected] (L.C. Rietveld),
[email protected] (M.C.M. van Loosdrecht). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.05.005
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 0 8 e4 0 1 8
unwanted growth of microorganisms in the distribution network. Nitrification is a two-step process performed by different species of bacteria (Belser, 1979) and archaea (Francis et al., 2007). As a consequence of the high oxygen demand of the nitrification process, trickling filters, also called dry biofilters, may be used to provide sufficient oxygen when ammonium concentration exceeds two mg L1. Even in the absence of oxygen limitation, incomplete nitrification in trickling filters is frequently encountered (de Vet et al., 2009a). The observation that nitrification becomes almost complete in the first period after startup with new filter material, indicates that inhibitors do not affect nitrification in the influent water for these filters. To maintain complete nitrification during sand filtration, the drinking water company Oasen in the Netherlands needs to apply the empirically based and effective but indirect technique of subsurface aeration. Both the nitrification problems and the beneficial effect of subsurface aeration on ammonium removal in full-scale trickling filters have been previously reported (Ibid.). The aim of this work was to test two possible hypotheses to explain the nitrification performance at an Oasen groundwater filtration station. The first hypothesis states that the growth of ammoniaoxidizing prokaryotes (AOP) in the subsurface aerated well might enhance nitrification in the trickling filter by continuous inoculation of AOP. Deposition of suspended nitrifying bacteria from the influent onto the filter has been suggested to be of crucial importance for filter performance (Uhl and Gimbel, 2000). In a previous study on groundwater treatment systems at Oasen, the microbial population composition in normal and subsurface aerated groundwater and active and inhibited full-scale trickling filters was determined by denaturing gradient gel electrophoresis (DGGE; de Vet et al., 2009b). That study showed a growth of ammonia-oxidizing bacteria (AOB) in the subsurface aerated well, but the method was indecisive for the quantitative importance of this inoculation for nitrification in the filter. The second hypothesis states that the excessive washout of nitrifying microorganisms might explain the observed nitrification problems in the groundwater filters and that the application of subsurface aeration reduces the washout of nitrifying microorganisms, which improves the performance of nitrification in sub surface aerated filters. Intensive filter backwashing, required for the prevention of clogging and diffusion limitation by removing inorganic deposits and dead biomass, may have a negative effect on the biological processes in the filter by excessively removing active biomass. Subsurface aeration might have an indirect effect on this by inducing subsurface iron colloid formation (Wolthoorn et al., 2004) and subsequently changing the types of iron precipitates and the surface characteristics of the ironcoated filter material. These changes could enhance the attachment of biomass and limit its losses by detachment in the trickling filter during filtration and backwash. In order to investigate these hypotheses, quantification of the abundance and activity of the ammonia-oxidizing populations in the filters over a prolonged period is essential. The effect of backwashing is usually measured by activity measurements of nitrification during filtration (Laurent et al., 2003) or in backwash
4009
water (Tra¨nckner et al., 2008). Direct quantification methods of the active biomass, such as effluent plate counts (Ahmad et al., 1998) or microscopic enumeration of attached biomass (Kasuga et al., 2007), have been used to assess the effect of backwashing for heterotrophic bacteria, but not for autotrophic bacteria like AOB. In this article, the abundance and activity of AOP, consisting of AOB and/or ammoniaoxidizing archaea (AOA), in groundwater filters were quantified by (quantitative) real-time polymerase chain reaction (qPCR) and activity measurements. By combining qPCR and activity measurements, the specific activity of AOP was evaluated. To evaluate the above-postulated hypotheses, balances for the ammonia-oxidizing populations in the trickling filters were determined.
2.
Materials and methods
2.1.
WTP Lekkerkerk filter and water samples
To compare the development of nitrifying activity and populations, two full-scale trickling filters at Oasen WTP Lekkerkerk were externally washed and the filter performance was monitored for nine months, after restarting the filters. During external washing, the filter sand was turbulently pumped out of a filter and back in to remove part of the accumulated filter coating and biomass. One filter treated natural groundwater from a separate well field and will be indicated further as the non-subsurface aerated filter. The other filter e further indicated as the subsurface aerated filter e was fed by a mixture of subsurface aerated and non-subsurface aerated groundwater. The water abstracted from the subsurface aerated well contributed to 16% (v/v) of the raw water flow for the subsurface aerated filter. The other wells abstracted water from the same aquifer, but were not influenced by subsurface aeration. The non-subsurface aerated filter e with nitrification problems e was started on December 12, 2007, and the subsurface aerated filter e with complete nitrification e on January 23, 2008. Both filters had an identical history, run time and backwash program and treated well-buffered (HCO3 w 3.8 mM) anaerobic groundwater, with an average pH of 7.35 0.06 and a temperature of 11.6 0.3 C. Due to forced ventilation in the trickling filter, the pH was raised to 7.67 0.13 and the dissolved oxygen was close to saturation in the filter effluents of both filters. Iron and manganese were removed almost completely in both filters. Schemes of the two investigated filters and their raw water composition have been reported in de Vet et al. (2009b). Filter performance was monitored weekly by analysis of ammonium, nitrite and nitrate in the filter influents and effluents. For determination of the abundance and activity of the AOP, extra samples were taken from the groundwater, filtrate, filter sand and backwash water of each filter. The groundwater and filtrate were sampled in duplicate for each filter system for qPCR. For the non-subsurface aerated filter, the mixed raw water from all wells was sampled. For the subsurface aerated filter, samples were taken from the subsurface aerated well and another representative well in the well field. Groundwater from the subsurface aerated well was sampled regularly during the 40 days of the abstraction
4010
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 0 8 e4 0 1 8
phase of the subsurface aeration cycle. The filtrate of each trickling filter was sampled regularly during the filter run time of 48 h. Three, six and nine months after external washing, samples were taken from the filter sand and backwash water. The filter sand was sampled also after one month. Because of the cone resistance of the trickling filter beds, a specially developed pneumatic sampling device with cone, sleeves and sampling bus was used to collect filter sand. In the subsurface aerated filter, sampling could take place at four depths (0e50 cm, 50e100 cm, 100e150 cm and 150e200 cm depths), covering the whole bed depth. In the non-subsurface aerated filter, the deepest layer could not be sampled because of high cone resistance. At each depth, mixed samples were formed from samples taken at three or four different places over the filter bed area. Measurements of the specific nitrification rate were performed on each mixed sample, qPCR was performed on the mixed samples taken from 0e50 cm and 100e150 cm depths. The two phases of the backwash program (first phase: lower flow of water þ auxiliary air scour; second phase: higher flow of water) were sampled separately in the supernatant water close to the backwash water spillway. Measurements of the specific nitrification rates were performed on both samples from the two backwash phases, aliquots of both phases were mixed proportionally according to their respective volumes for analysis by qPCR. All samples were transported and stored at 4 3 C. The isolation of DNA, start of the measurement of the specific nitrification rate, and N-analyses all took place within 48 h after sampling. Control measurements showed that the storage time has not affected the outcome of the kinetic assays.
Na2HPO4$12H2O in a molar ratio of 1e0.02 (Tra¨nckner et al., 2008) was added at the start of the experiment to a final concentration of 10 mg NH4eN L1, corresponding to the concentration range in the full-scale filters. Five mL samples were taken before substrate addition and afterwards at regular intervals over one to three days, until the ammonium was depleted. At each sampling time, duplicate samples were taken from all batch reactors, except for the negative controls that were sampled only once. The samples were filtered over 0.45 mm PTFE filters into tubes, covered and stored for analyses at 4 3 C. The temperature and pH were measured at the beginning and end of each experiment. The average and standard deviation of the temperature over all measurements was 23 1 C at the start and 26 2.2 C at the end. The pH ranged from 7.7 0.2 at the start to 8.2 0.2 and 8.4 0.1 for uninhibited and ATU-inhibited samples, respectively. The pH rise was caused by carbon dioxide stripping by flushing the reactor with compressed air. The method of least squares was used to fit the measured data for NH4 and NO2 þ NO3 using exponential substrate depletion and product formation curves (see Supplementary Material C for derivation). The initial volumetric nitrification rate in the backwash water samples at room temperature was determined from the initial slope of the exponential function. To exclude (bio) adsorption effects at the start of each experiment and incorrect analyses, only samples with less than a 5% deviation from the average of total inorganic N were used for this calculation, and only samples with more than 1 mg L1 of NH4eN were included to limit the effect of substrate limitation. Because of the exponential growth phase, maintenance and decay were neglected. Results were not corrected for loss of nitrogen through assimilation.
2.3. 2.2.
N measurements
Specific nitrification rates
As the groundwater did not contain nitrite and nitrate, the bulk nitrification rate for the filters (rN) was calculated by multiplying the water flow with the sum of the nitrite and nitrate concentrations in the filter effluent, both averaged over the test period. For both filter samples and backwash water samples, the specific nitrification rates were determined in batch experiments. The batch method used for the filter samples has been described in de Vet et al. (2009b) and provides a ‘sand-specific nitrification rate’ in mg (NO2 þ NO3)eN h1 kg1 of filter sand. The volumetric nitrification rate for the backwash samples was determined as mg N h1 and L1 of backwash water. For each sample date, the specific nitrification rate was determined for both backwash phases. To test for possible heterotrophic nitrification, negative control measurements were conducted for all samples by the addition of 10 mg L1 of allylthiourea (ATU; C4H8N2S) which totally inhibits the activity of autotrophic AOB, but only partly the activity of heterotrophic AOB (Robertson et al., 1989) and AOA (Taylor et al., 2010). The specific nitrification rate for backwash water samples was determined in 1 L batch tests. All batch reactors were continuously stirred and mildly flushed with compressed air during the entire experiment. A concentrated substrate containing NH4Cl and
In all samples, ammonium, nitrite and nitrate were measured for potential losses in the total inorganic N-balance, which should be conserved when only nitrification occurs. For all full-scale filter samples, except for the backwash activity measurements, both ammonium and nitrite were determined by colorimetric measurement. Nitrate was measured by conductivity in accordance with NEN-EN-ISO 10304-1 (http:// www.nen.nl/web/Normshop/Norm/NENENISO-1030412009en.htm). For samples from the backwash activity measurements, ammonium, nitrite and nitrate were determined by colorimetric measurement on a Lachat Quikchem 8500 Flow Injection Analysis System.
2.4.
Quantification AOP by qPCR
To quantify AOP by qPCR, the functional gene coding for the a-subunit of the ammonia monooxygenase enzyme (which catalyses the first steps of ammonia oxidation) was amplified, as described by Van der Wielen et al. (2009). In short, 50 mL of autoclaved tap water were added to 2e5 g filter sand samples. Subsequently, the samples were sonicated for 2 min at 20 kHz in a Sonifier II W-250 and the liquid phase was collected. DNA was isolated from the liquid samples by filtration over a 25-mm polycarbonate filter (0.22 mm pore size, type GTTP; Millipore, the Netherlands). The filter and a DNA fragment of
4011
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 0 8 e4 0 1 8
an internal control were added to the phosphate and MT buffer of the FastDNA Spin kit for soil (Qbiogene) and stored at 20 C. The use of the internal control was used to determine PCR efficiency and was in accordance with a normalized method (NEN 6254:2009), which has been used in other studies as well (Van der Wielen et al., 2009; Van der Wielen and Medema, 2010). DNA was isolated using the FastDNA Spin kit for soil according to the supplier’s protocol. The amoA genes of AOB and AOA were amplified with previously reported primers (Francis et al., 2005; Rotthauwe et al., 1997) in an iCycler IQ real-time detection system (BioRad laboratories BV, the Netherlands). Quantification was based on a comparison of the sample Ct value with the Ct values of a calibration curve based on known numbers of the amoA gene of AOB or AOA. The AOB numbers were calculated by assuming two amoA gene copy numbers per cell (Chain et al., 2003), and the AOA numbers by assuming one amoA gene copy number per cell (Mincer et al., 2007).
2.5.
Yx;N ¼
XN mx;N DXN;Dt z rN rN Dt
(1)
where m, specific growth rate (h1); r, rate (mg N h1); Δt, time (h); X, total cell number (cells); Y, yield (cells mg1 Nconverted). Indices: N, nitrification; Δt, time (h); x, biomass.
Cell-specific nitrification rate
The cell-specific nitrification rate was calculated for all samples by combining the cell number determined by qPCR and the specific nitrification rate from the batch tests of all filter sand and backwash water samples. The results of these calculations are summarized in Table 1 of the Discussion section.
2.6.
The filter bed area was 18.0 m2 per filter. The average water flow over the monitored period was 36.5 and 47 m3 h1 for the non-subsurface and the subsurface aerated filter, the bed height 2.03 0.04 and 1.81 0.02 m, and the total (dry) (with sand mass per filter 58 and 52 103 kg rDW,sand ¼ 1600 kg m3), respectively. When no accumulation (determined by measuring cell numbers on the filters samples) in time occurs, the net washout measured by qPCR balances the growth of AOP. Under the assumption of equilibrium, yield values of AOP were determined using the calculated growth of AOP and the nitrification rate, according to Equation (1). The results are summarized in Table 1 of the Discussion section.
AOB and AOA balances
From the qPCR results, separate balances of AOB and AOA for each filter were calculated. The balance calculation for one filter run is schematized in Supplementary Material A. For every water flow entering or leaving a filter, the total values for AOB or AOA cell numbers were calculated by multiplying the measured concentration by the flow and duration of the phase. In case the concentrations were measured at different times during a phase e like in the subsurface aerated well during the abstraction cycle and the filtrate over the filter run time e the weighted average was used.
3.
Results
3.1.
Overall performance of the full-scale filters
The overall ammonium removal performance of both fullscale filters is shown in Fig. 1. The nitrification in the subsurface aerated filter was nearly complete after the startup period (A). The non-subsurface aerated filter (B) showed incomplete nitrification and a gradual relapse after the startup period of about two to three months, although less pronounced than previously reported (de Vet et al., 2009a). Because of the different groundwater source, the influent ammonium concentration in the subsurface aerated filter was lower. Previous research showed, however, that without subsurface aeration nitrification became incomplete in this filter as well (Ibid.). We have shown that the complete nitrification resulted from the application of subsurface aeration and not from the lower influent ammonium concentration (Ibid.).
Table 1 e Comparison of growth parameters between Oasen’s two filters and the literature; (non)-SA [ (non)-subsurface aerated. Parameters
Unit
Maximum volumetric nitrification rate
gN h
Cell yield
1
SA filter
3
1.5e5
1.5e5
1011 cells mol1 N
2
5
Biomass yield
gDW mol1 N
0.01a
0.04a
Specific growth rate
h1
0.0006
0.0002
Cell specific nitrification rateb
102 fg N h1 cell1
0.2e5 2e20
0.02e0.08 0.1e0.2
a Calculated. b Section 4.3.
m
Non-SA filter
Sand Backwash water
References 12e14 van den Akker et al. (2008), Westerman et al. (2000) 14e67 Belser and Schmidt (1980) 0.4e1.72 Loveless and Painter (1968), Keen and Prosser (1987) 0.012e0.088 Prosser (1989) 0.1e3 Prosser (1989)
4012
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 0 8 e4 0 1 8
The profiles of the sand-specific nitrification rates over the filter bed depth differed between both filters (see Supplementary Material B). There was neither a clear difference in the sand-specific nitrification rate between both filters, nor a clear difference over time. The averages with standard deviations of the sand-specific nitrification rate over the filter bed depth for the experimental period are shown in Fig. 2A. The nitrification activity in the backwash water samples was identified by fitting ammonium uptake and nitrite and nitrate production using an exponential growth model (see Supplementary Material C for two examples). The nitrification rate in the ATU-inhibited batch experiments was close or equal to zero in all cases, indicating that AOA and heterotrophic AOB did not significantly contribute to the oxidation of ammonium.
Fig. 1 e Ammonium concentrations in the 280 days after external washing for the subsurface aerated filter (A) and non-subsurface aerated filter (B); Δ: influent filter, >A effluent filter.
Apart from the first months after startup, there was only a small loss of 5% in total nitrogen (as ammonium, nitrite and nitrate) between the influent and effluent for both filters. Stripping of ammonia (due to the high ventilation air to water ratio), biological and chemical denitrification (abiotic nitrite reduction with oxidation of ferrous iron; Tai and Dempsey, 2009) may have caused this unbalance.
3.2.
(Specific) nitrification rates
The calculated bulk nitrification rate (rN) after a startup period of one month was 101 20 and 68 15 g (NO2 þ NO3)eN h1 for the non-subsurface and the subsurface aerated filter, respectively. The bulk sand-specific nitrification rate was 1.8 0.4 and 1.2 0.3 mg (NO2 þ NO3)eN h1 kg1 of filter sand, respectively. The lower bulk nitrification rates for the subsurface aerated filter are explained by substrate limitation because all ammonium was depleted.
Fig. 2 e Sand-specific nitrification rate for filter sand samples (A) and initial volumetric nitrification rate for backwash water samples (B) taken one to nine months after external washing; subsurface aerated filter (striped bars) and non-subsurface aerated filter (solid bars); average and standard deviation for samples over the depth of the filter bed (filter sand samples) and two phases of backwash program; backwash water sample from non-subsurface aerated filter after 6 months only from the second phase.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 0 8 e4 0 1 8
The measurements of the specific nitrification rate in the backwash water showed a profound difference between subsurface and non-subsurface aerated filters (Fig. 2B). The initial volumetric nitrification rate was higher in the backwash water samples from the subsurface aerated filter than in the non-subsurface aerated filter, and this difference increased further in time (Fig. 2B).
3.3.
AOP quantification by qPCR
In Fig. 3, the average, minimum and maximum of AOP cell numbers are shown for all filter sand, influent, effluent and backwash water samples of the subsurface aerated (A) and
4013
the non-subsurface aerated filter (B) over the monitoring period. For most samples, AOB were two or more orders higher than AOA. Exceptions were both influent samples and the effluent samples of the subsurface aerated filter. When comparing both filters, AOB cell numbers were higher in all but the influent samples of the non-subsurface aerated filter. The AOA cell numbers in the filter samples and the increase in numbers from the influent to the effluent samples show that AOA grew in both filters. AOA cell numbers were higher in all water samples of the subsurface aerated filter compared to the non-subsurface aerated filter. This difference can be partly explained by the inoculation of AOA from the subsurface aerated groundwater and partly by the enhanced growth in the subsurface aerated filter. This enhanced growth is shown by comparison of AOA cell numbers in the influent and effluent of both filters; the increase is much stronger in the subsurface aerated filter than for the other filter (see also AOA balances for both filters in Supplementary Material A). A comparison of the AOA cell numbers in the effluent and backwash water of this filter suggests a weak attachment of the AOA cells to the filter sand.
3.4.
AOP balances
AOP balances in the two full-scale filters were assessed over the monitoring period by qPCR to compare their abundance and growth in both full-scale filters. For AOB, the balance terms for one filter run of 48 h was calculated from the cell numbers, water flows, filter bed volumes and time (Fig. 4). The overall balances in Fig. 4 show some similarities and also differences for the two filters. In both filters, a distinct growth of AOB was observed, as shown by the increase in cell numbers from influent to effluent water flows. Despite lower AOB cell numbers in the influent water, the growth of AOB was enhanced and the cell yield higher in the nonsubsurface aerated filter compared to the other filter. This coincided with a higher total cell number present in the non-subsurface aerated filter (Discussion section 4.2). The balances for AOA cells confirm the limited quantitative importance of AOA in these filters and are therefore shown in Supplementary Material A. The AOA balance showed negligible growth in the non-subsurface aerated filter compared to AOB. In the subsurface aerated filter, both AOA cell numbers in the influent and growth in the filters were higher, but their cell numbers still amounted to less than 10% of the total AOP in all balance terms.
Fig. 3 e Averages of AOB (solid bars) and AOA (line bars) over the total test period of nine months; subsurface aerated filter (A), non-subsurface aerated filter (B); values for filter sand in gL1 and for water in mLL1; averages over 10 sand, 11 (SA) and 2 (non-SA) influent, 8 effluent and 4 backwash water samples per filter; cross bars indicate minimal and maximal values; results in filter influent under the detection limit of the method (<15 mLL1) are depicted as 7.5 mLL1 (half of the detection limit).
4.
Discussion
The combined approach to evaluate AOP activity by qPCR and activity measurements provided a good means to evaluate the observed nitrification performances of full-scale nitrification filters and allowed a deeper understanding of AOP population dynamics in such filters. In the next subsections, some of these major issues will be discussed and related to previous research.
4014
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 0 8 e4 0 1 8
Fig. 4 e AOB balances for the subsurface aerated filter (A) and non-subsurface aerated filter (B) during nine months after external washing; bar graphs in the center show the cumulative cell numbers per filter run time of 48 h for the influent (top) and effluent (bottom). The bar graphs in the middle center and outside present the cumulative cell numbers present in the filter and washed out during backwash, respectively; the curved arrows indicate the direction of the backwash flow and point towards the columns graph with the cumulative cell numbers present in the backwash water.
4.1. Detection of AOB and AOA using molecular techniques In this study, high numbers of AOB were found in filter samples of both filters based on the amplification of functional gene markers using qPCR. These results are in contrast to previous analysis using CTO-primers targeting the 16S rRNA, which could not detect AOB in a non-subsurface aerated filter (de Vet et al., 2009b). It is likely that the AOB specific 16S primers had some mismatches with the AOB in the system studied. Maybe other 16S primer sets (such as published by Hermansson and Lindgren, 2001) might have yielded better results. As 16S rRNA is present in all bacteria, species other than AOB are likely to be targeted too, in contrast to the approach with functional markers as the amoA gene. The inaccuracy of the 16S approach for detection of AOB has been reported before (Hoefel et al., 2005). The quantification of the amoA genes of AOB and AOA on filter sand and in backwash water suggests that AOA played a minor role in these groundwater filter systems. Only in the groundwater and filtrate of the subsurface aerated filter were AOA equally present or more abundant than AOB. Subsurface aeration stimulated the growth of AOA both in groundwater and in the trickling filter. By contrast, virtually no growth of AOA occurred in the non-subsurface aerated filter. The growth of AOA in the subsurface aerated filter did not result in a strong accumulation of AOA in the filter bed. Possibly, they do not form as strong biofilms as AOB do. For the subsurface aerated filter, the relatively low numbers of AOA compared to
AOB in the backwash water and the decreasing AOA cell numbers on the filter sand in time (data not shown) demonstrate the marginal role of AOA in ammonia oxidation. In the non-subsurface aerated filter the share of AOA in ammonia oxidation was even less significant. Recent studies suggest that pH and ammonium concentration are selecting criteria for the dominance of AOB and AOA (Prosser and Nicol, 2008 and references therein). Martens-Habbena et al. (2009) found a very high affinity of a marine AOA, with a half halfsaturation constant of 133 nM total ammonium. This affinity is much higher than of known bacteria and suggests the dominance of archaea at micromolar concentrations. It might implicate that the conditions in our studied full-scale systems and lab assays with millimolar concentrations selected for AOB. A previous study on several Dutch groundwater treatment plants showed AOB as the dominant AOP population (Van der Wielen et al., 2009), which is in agreement with our results. However, for one plant in that study, a domination of AOA in the nitrification process was found, although the treated groundwater contained 1.1 mg NH4 L1 (Ibid., Supplemental Material). A comparison of the calculated cell yield (Section 4.2) and cell specific activity in the backwash water (Section 4.3) with corresponding literature values suggests that the applied qPCR method detected only a part the AOP cells in the backwash water samples. For the sand samples (Section 4.3), it seemed that the qPCR technique detected most AOP cells. This finding, together with the relatively high recovery yield of the internal control in the q-PCR reactions, suggests that
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 0 8 e4 0 1 8
the PCR method used functioned properly for the sand samples. However, the pretreatment of the backwash water samples may not have been optimal for backwash water samples that contain a high concentration of inorganic suspended matter. Still, this research shows that the application of molecular methods has added value in analyzing the biological conversions of nitrogen in full-scale drinking water filters.
4.2.
Abundance and metabolic activity of AOB
The AOB balances for the full-scale filters (Fig. 4) provide sufficient data for a decision on the two previously posed hypotheses regarding the performance of nitrification in the Oasen trickling filters. The first hypothesis, stating that enhancement of nitrification in the trickling filters by subsurface aeration works by stimulating inoculation of AOB, was rejected. Subsurface aeration increased the AOB number in the groundwater slightly, but this phenomenon did not appear crucial for the filter’s functioning. The total AOB cell numbers inoculated from the groundwater were negligible compared to those present in the filter, filtrate or backwash water, suggesting that inoculation was marginal in numbers compared to growth. The quantitative results reported here confirm the qualitative observations reported earlier (de Vet et al., 2009b). The second hypothesis, stating that the relapse of nitrification in a non-subsurface aerated filter is caused by a stronger washout of biomass, could also be rejected. On first glimpse, this does not appear obvious. Indeed, the numbers of AOB cells washed out in the filtrate and backwash water were higher than in the subsurface aerated filter (Fig. 4). However, the nitrification capacity is being determined by the AOB attached to the filter sand and not by those washed out of the filter. The AOB cell numbers present in the nonsubsurface aerated filter sand samples were higher as well (Fig. 4) and did not decrease in time. These observations imply a better growth of not very strongly attached AOB in the non-subsurface aerated filter as compared to the subsurface aerated filter, possibly related to differences in the iron precipitate formation between both filters (de Vet et al., 2009a). With respect to the total AOB cell numbers present in the filters, the difference between the two filters shows a reverse picture. For a filter run time of 48 h, the total amount of washed out AOB cell numbers was on average 3% for the subsurface aerated filter and 1% for the non-subsurface aerated filter. Thus, it appears that the elevated detachment and washout of AOB from the nonsubsurface aerated filter is not the primary cause for the nitrification problems. To analyze other possible causes for the differences between both filters, some of the characteristic growth parameters for AOB in both filters are compared to each other and literature references. Averaged over the bed depth, the sand-specific nitrification rate was between 1 and 3 mg N h1 kg1 (see Fig. B.1 of Supplementary Material B) or 1.5 to 5 g N h1 m3 of filter sand, comparable to the bulk nitrification rate for the whole filters (Section 3.2). van den Akker et al. (2008) observed almost complete nitrification and a maximum nitrification rate of 12 g N h1 m3 of filter
4015
bed in drinking water trickling filters with influent NH4-N concentrations between 4 and 5 mg L1. Westerman et al. (2000) found a maximum removal of ammonia of 14 g N h1 m3 of filter bed in the treatment of the supernatant of swine manure by upflow aerated biofilters. Despite strongly deviating process conditions, both references give comparable maximum ammonium removal capacities, which were higher than the ones found in the filters samples in this study. The yield of AOB cells on ammonium can be calculated by combining the overall metabolic activity and the AOB growth (defined as washed out cells) in the filters for one filter run, according to Equation (1). This cell yield equals 2 1011 and 5 1011 cells mol1 N for the subsurface and the nonsubsurface aerated filter, respectively. Belser and Schmidt (1980) found values between 14 and 67 1011 cells mol1 N for pure and mixed cultures in batch experiments, which is one order higher than in our research. After the recalculation of cell numbers to biomass, the yield can also be compared to other published studies. AOB cell numbers were converted to biomass using the Nitrosomonas cell dimensions of 2.4 1019 m3 (Schmidt et al., 2004) with an assumed dry weight (DW) content of 30%, leading to a DW of 7.2 1014 g per cell. The cell yield thus calculated is 0.01 and 0.04 gDW mol- N for the subsurface and the nonsubsurface aerated filter, respectively. These values are again one to two orders lower than the values found in batch (Loveless and Painter, 1968) and continuous (Keen and Prosser, 1987) lab experiments, 0.4 till 1.72 gDW mol N. From the cell numbers present in and washed out from the filter (Fig. 4), a specific growth rate m0 of 0.0006 and 0.0002 h1 (doubling times between 7 and 25 weeks) was estimated for the subsurface and the non-subsurface aerated filter, respectively. The values are much lower than maximum specific growth rates reported in the literature for batch and continuous cultures (0.012e0.088 h1; doubling times between 60 and 8 h; Prosser, 1989, and references therein). There are several possible causes for the observed low yields. It could be that the maintenance energy forms a significant part of the ammonium conversion. The general maintenance energy requirements at 12 C are equal to 1.3 kJ mol1 biomass h1 (Tijhuis et al., 1993). The change in free energy of the catabolic reaction (NH4þ þ 1½ O2 / NO2 þ 2Hþ þ H2O) is 276 kJ mol1. This gives a specific substrate conversion rate required for the maintenance of 0.005 mol N mol1 XN h1 at 12 C. This is, after temperature correction, in good agreement with values determined by Tappe et al. (1999). With the cell DW calculated above, the total cell mass present in the filter (Fig. 4) was 4 and 60 moles (or 1 102 and 16 102 g) for the subsurface and the non-subsurface aerated filter, respectively. The substrate conversion needed for maintenance is 0.02 and 0.3 mol N h1, 0.4 and 4% of the observed bulk nitrification rate. This indicates that the limited biomass yield for nitrifiers observed in both filters was not related to high maintenance requirements. Differences between the field conditions in our research and the lab conditions in the references cited may also have influenced the yields. In the full-scale filters, factors such as the competition for nutrients with other auto- and heterotrophic bacteria,
4016
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 0 8 e4 0 1 8
inhibition by deposited organic and inorganic compounds, and pH and temperature may have impaired the growth yield of AOB. The lower growth yield may also be related to the reduced growth rate. Glover (1985) found that the C yield from nitrification declined both with increasing age in the batch culture and with decreasing growth rates in chemostats. Predation by protozoa might also decrease the observed yields. Finally, limitations of the applied qPCR method or the DNA isolation steps may also have played role, as was discussed in Section 4.1. The main growth characteristics discussed above are once again summarized in Table 1.
4.3.
Cell-specific nitrification rates
The cell-specific nitrification rates, calculated by division of the sand-specific and volumetric nitrification rate by the cell number for each sand and backwash water sample, are summarized in Fig. 5. Only AOB cell numbers were taken into account since AOA cell numbers contributed less than 2 % of the total AOP numbers in all samples. The cell-specific nitrification rates were at least one order of magnitude larger for all samples from the subsurface aerated filter than those from the non-subsurface aerated filter. In the subsurface aerated filter, the cell-specific activity increased between three and six months, coinciding with the observed drop in cell numbers (data not shown). In both the sand and backwash water, the AOB cell-specific nitrification rates were constant over time for the nonsubsurface aerated filter, and stabilized at a higher level after three months for the subsurface aerated filter samples. The results from the activity measurements in the backwash water samples are remarkable because they systematically show a much higher initial volumetric nitrification rate for the subsurface aerated filter compared to the non-subsurface aerated filter, despite the lower AOP numbers present (see Fig. 4 and Table C.1 of Supplementary Material C). In the literature, a large distribution in cell-specific nitrification rates is found in batch and continuous experiments with pure or mixed cultures. Reported values range between 0.9 and 23 fmol N h1 (0.1 to 3 102 fg N h1 cell1; Prosser, 1989, and references therein). The AOB cell-specific nitrification rate in the subsurface aerated filter samples was at the higher end of this reported range (0.2e5 102 fg N h1 cell1), and at the lower end in the nonsubsurface aerated filter samples (0.02e0.08 102 fg N h1 cell1). In the backwash water, the cell-specific nitrification rate for the subsurface aerated filter samples was again at the higher end of the values reported in the literature and even higher than the maximum reported value (2e20 102 fg N h1 cell1). For the non-subsurface aerated filter samples, the measurements were again at the lower end of the values reported in the literature (0.1e0.2 102 fg N h1 cell1). Consequently, the main finding from this research is that the incomplete nitrification in the non-subsurface aerated filter coincided with a systematically and significantly lower specific nitrification activity of the AOP cells compared to the much more active AOP in the subsurface aerated filter. This suggests that most nitrifying cells are dead or that their
Fig. 5 e Averages and standard deviations of the cellspecific nitrification rate on filter sand (A) and in the backwash water (B); striped bars subsurface aerated filter, solid bars non-subsurface aerated filter.
activity is severely inhibited in the non-subsurface aerated filter. The exponential growth in all assays with backwash water clearly shows that possible inhibitors had no effect on nitrifying organisms in the backwash water. Two possible causes for this inactivation or inhibition of AOP are: (i) diffusion limitation and (ii) absolute (bulk) limitation of nutrients. Limitation of the substrate diffusion may occur as a result of inadequate backwashing and formation of a thick biofilm or a covering layer that may be formed by inorganic deposits like iron (oxy)hydroxides in the groundwater filter. These inorganic deposits may also lead to absolute nutrient limitations in the bulk water phase because of the co-precipitation of phosphorus and trace elements in the iron (oxy)hydroxides.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 0 8 e4 0 1 8
The nutrient limitation for AOB can be aggravated further by competition for these scarce nutrients by other microorganisms in the filter.
5.
Conclusions
The combination of qPCR targeting the amoA gene and activity measurements of ammonia oxidation is a good tool to evaluate the abundance and activity of ammoniaoxidizing prokaryotes in water treatment filters; the sampling of backwash water and the extraction of DNA from these samples requires caution. The inoculation of sand filters with AOP from groundwater plays a minor role in terms of numbers in the enhancement of nitrification by subsurface aeration. The relapse of nitrification in a non-subsurface aerated filter was not caused by an excessive detachment and washout of AOP. Incomplete nitrification in the non-subsurface aerated filter was not caused by a decline in and absence of AOB, contrary to the findings of de Vet et al. (2009b). In the non-subsurface aerated filter with incomplete nitrification, a larger AOB population with lower cell-specific nitrification activity persisted compared to the subsurface aerated filter with full nitrification.
Appendix. Supplementary information Supplementary Material A, balance scheme and AOA balance calculation for filters. Supplementary Material B, profiles of cell numbers and sand-specific nitrification rates throughout the filter bed depth profile. Supplementary Material C, analytical approach and results of the backwash water batch experiments. Supplementary data associated with this article can be found in the online version, at doi:10.1016/j.watres.2011.05.005.
references
Ahmad, R., Amirtharajah, A., Al-Shawwa, A., Huck, P.M., 1998. Effects of backwashing on biological filters. Journal/American Water Works Association 90 (12), 62e73. Belser, L.W., 1979. Population ecology of nitrifying bacteria. Annual Review of Microbiology 33, 309e333. Belser, L.W., Schmidt, E.L., 1980. Growth and oxidation kinetics of three genera of ammonia oxidizing nitrifiers. FEMS Microbiology Letters 7 (3), 213e216. Chain, P., Lamerdin, J., Larimer, F., Regala, W., Lao, V., Land, M., Hauser, L., Hooper, A., Klotz, M., Norton, J., Sayavedra-Soto, L., Arciero, D., Hommes, N., Whittaker, M., Arp, D., 2003. Complete genome sequence of the ammonia-oxidizing bacterium and obligate chemolithoautotroph Nitrosomonas europaea. Journal of Bacteriology 185 (9), 2759e2773. de Vet, W.W.J.M., Rietveld, L.C., van Loosdrecht, M.C.M., 2009a. Influence of iron on nitrification in full-scale drinking water
4017
filters. Journal of Water Supply: Research and TechnologyAQUA 58 (4), 247e256. de Vet, W.W.J.M., Dinkla, I.J.T., Muyzer, G., Rietveld, L.C., van Loosdrecht, M.C.M., 2009b. Molecular characterization of microbial populations in groundwater sources and sand filters for drinking water production. Water Research 43 (1), 182e194. Francis, C.A., Beman, J.M., Kuypers, M.M.M., 2007. New processes and players in the nitrogen cycle: The microbial ecology of anaerobic and archaeal ammonia oxidation. ISME Journal 1 (1), 19e27. Francis, C.A., Roberts, K.J., Beman, J.M., Santoro, A.E., Oakley, B.B., 2005. Ubiquity and diversity of ammonia-oxidizing archaea in water columns and sediments of the ocean. Proceedings of the National Academy of Sciences of the United States of America 102 (41), 14683e14688. Glover, H.E., 1985. The relationship between inorganic nitrogen oxidation and organic carbon production in batch and chemostat cultures of marine nitrifying bacteria. Archives of Microbiology 142 (1), 45e50. Hermansson, A., Lindgren, P.E., 2001. Quantification of ammoniaoxidizing bacteria in arable soil by real-time PCR. Applied and Environmental Microbiology 67 (2), 972e976. Hoefel, D., Monis, P.T., Grooby, W.L., Andrews, S., Saint, C.P., 2005. Culture-independent techniques for rapid detection of bacteria associated with loss of chloramine residual in a drinking water system. Applied and Environmental Microbiology 71 (11), 6479e6488. Kasuga, I., Shimazaki, D., Kunikane, S., 2007. Influence of backwashing on the microbial community in a biofilm developed on biological activated carbon used in a drinking water treatment plant, pp. 173e180. Keen, G.A., Prosser, J.I., 1987. Steady state and transient growth of autotrophic nitrifying bacteria. Archives of Microbiology 147 (1), 73e79. Laurent, P., Kihn, A., Andersson, A., Servais, P., 2003. Impact of backwashing on nitrification in the biological activated carbon filters used in drinking water treatment. Environmental Technology 24 (3), 277e287. Loveless, J.E., Painter, H.A., 1968. The influence of metal ion concentrations and pH value on the growth of a nitrosomonas strain isolated from activated sludge. Journal of General Microbiology 52 (1), 1e14. Martens-Habbena, W., Berube, P.M., Urakawa, H., De La Torre, J.R. , Stahl, D.A., 2009. Ammonia oxidation kinetics determine niche separation of nitrifying Archaea and Bacteria. Nature 461 (7266), 976e979. Mincer, T.J., Church, M.J., Taylor, L.T., Preston, C., Karl, D.M., DeLong, E.F., 2007. Quantitative distribution of presumptive archaeal and bacterial nitrifiers in Monterey Bay and the North Pacific Subtropical Gyre. Environmental Microbiology 9 (5), 1162e1175. Prosser, J.I., 1989. Autotrophic nitrification in bacteria. Advances in Microbial Physiology 30, 125e181. Prosser, J.I., Nicol, G.W., 2008. Relative contributions of archaea and bacteria to aerobic ammonia oxidation in the environment. Environmental Microbiology 10 (11), 2931e2941. Robertson, L.A., Cornelisse, R., Zeng, R., Kuenen, J.G., 1989. The effect of thiosulphate and other inhibitors of autotrophic nitrification on heterotrophic nitrifiers. Antonie van Leeuwenhoek 56 (4), 301e309. Rotthauwe, J.H., Witzel, K.P., Liesack, W., 1997. The ammonia monooxygenase structural gene amoa as a functional marker: molecular fine-scale analysis of natural ammonia-oxidizing populations. Applied and Environmental Microbiology 63 (12), 4704e4712. Schmidt, I., Look, C., Bock, E., Jetten, M.S.M., 2004. Ammonium and hydroxylamine uptake and accumulation in Nitrosomonas. Microbiology 150 (5), 1405e1412.
4018
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 0 8 e4 0 1 8
Tappe, W., Laverman, A., Bohland, M., Braster, M., Rittershaus, S., Groeneweg, J., Van Verseveld, H.W., 1999. Maintenance energy demand and starvation recovery dynamics of Nitrosomonas europaea and Nitrobacter winogradskyi cultivated in a retentostat with complete biomass retention. Applied and Environmental Microbiology 65 (6), 2471e2477. Tai, Y.L., Dempsey, B.A., 2009. Nitrite reduction with hydrous ferric oxide and Fe(II): stoichiometry, rate, and mechanism. Water Research 43 (2), 546e552. Taylor, A.E., Zeglin, L.H., Dooley, S., Myrold, D.D., Bottomley, P.J., 2010. Evidence for different contributions of archaea and bacteria to the ammonia-oxidizing potential of diverse oregon soils. Applied and Environmental Microbiology 76 (23), 7691e7698. Tijhuis, L., Van Loosdrecht, M.C.M., Heijnen, J.J., 1993. A thermodynamically based correlation for maintenance Gibbs energy requirements in aerobic and anaerobic chemotrophic growth. Biotechnology and Bioengineering 42 (4), 509e519. Tra¨nckner, J., Wricke, B., Krebs, P., 2008. Estimating nitrifying biomass in drinking water filters for surface water treatment. Water Research 42 (10e11), 2574e2584.
Uhl, W., Gimbel, R., 2000. Dynamic modeling of ammonia removal at low temperatures in drinking water rapid filters. Water Science and Technology 41 (4e5), 199e206. van den Akker, B., Holmes, M., Cromar, N., Fallowfield, H., 2008. Application of high rate nitrifying trickling filters for potable water treatment. Water Research 42 (17), 4514e4524. Van der Wielen, P.W.J.J., Voost, S., Van der Kooij, D., 2009. Ammonia-oxidizing bacteria and archaea in groundwater treatment and drinking water distribution systems. Applied and Environmental Microbiology 75 (14), 4687e4695. Van der Wielen, P.W.J.J., Medema, G.J., 2010. Unsuitability of quantitative Bacteroidales 16S rRNA gene assays for discerning fecal contamination of drinking water. Applied and Environmental Microbiology 76 (14), 4876e4881. Westerman, P.W., Bicudo, J.R., Kantardjieff, A., 2000. Upflow biological aerated filters for the treatment of flushed swine manure. Bioresource Technology 74 (3), 181e190. Wolthoorn, A., Temminghoff, E.J.M., Van Riemsdijk, W.H., 2004. Colloid formation in groundwater by subsurface aeration: Characterisation of the geo-colloids and their counterparts. Applied Geochemistry 19 (9), 1391e1402.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 1 9 e4 0 2 7
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Occurrence and suitability of sucralose as an indicator compound of wastewater loading to surface waters in urbanized regions Joan Oppenheimer a,*, Andrew Eaton b, Mohammad Badruzzaman a, Ali W. Haghani b, Joseph G. Jacangelo a,c a
MWH Americas Inc., 618 Michillinda Avenue, Arcadia, CA 91007, USA MWH Laboratories, 750 Royal Oaks Avenue, Monrovia, CA 91016, USA c The Johns Hopkins University, Baltimore, MD 21205, USA b
article info
abstract
Article history:
Urban watersheds are susceptible to numerous pollutant sources and the identification of
Received 23 February 2011
source-specific indicators can provide a beneficial tool in the identification and control
Received in revised form
of input loads, often times needed for a water body to achieve designated beneficial uses.
11 May 2011
Differentiation of wastewater flows from other urban wet weather flows is needed in order
Accepted 12 May 2011
to more adequately address such environmental concerns as water body nutrient
Available online 25 May 2011
impairment and potable source water contamination. Anthropogenic compounds previously suggested as potential wastewater indicators include caffeine, carbamazepine, N,N-
Keywords:
diethyl-meta-toluamide (DEET), gemfibrozil, primidone, sulfamethoxazole, and TCEP. This
Sucralose
paper compares the suitability of a variety of anthropogenic compounds to sucralose, an
Wastewater indicator
artificial sweetener, as wastewater indicators by examining occurrence data for 85 trace
Pharmaceuticals and personal care
organic compounds in samples of wastewater effluents, source waters with known
products
wastewater point source inputs, and sources without known wastewater point source
Urban watershed
inputs. The findings statistically demonstrate the superior performance of sucralose as a potential indicator of domestic wastewater input in the U.S. While several compounds were detected in all of the wastewater effluent samples, only sucralose was consistently detected in the source waters with known wastewater discharges, absent in the sources without wastewater influence, and consistently present in septic samples. All of the other compounds were prone to either false negatives or false positives in the environment. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
Surface waters and shallow unconfined aquifers in urbanized regions are vulnerable to pollutants that cause impairment from previously established beneficial uses. Pathogens have been identified as the top impairment of assessed U.S. rivers and streams with major sources attributed to municipal
discharges/sewage, unspecified non-point source discharges, and urban runoff/stormwater (USEPA, 2009). Chemical indicators have been proposed as alternatives to microbial indicators as a more definitive means of identifying fecal contamination from human sources (Glassmeyer et al., 2005). Nitrogen input levels and oxic conditions are the major variables correlated with higher observed nitrate concentrations in groundwaters
* Corresponding author. Tel.: þ1 626 568 6006; fax: þ1 626 568 6015. E-mail address:
[email protected] (J. Oppenheimer). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.05.014
4020
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 1 9 e4 0 2 7
throughout the United States (Burow et al., 2010). Quantifying water body pollutant mass loads back to contributing sources within an urbanized watershed is complex and frequently depends upon modeling strategies that incorporate estimates and uncertainties about each source’s flow and pollutant concentration levels (Nix, 1994). Mass loadings are often times further attenuated by location specific fate and transport processes that are not adequately characterized. In order to effectively mitigate impaired water bodies, it is important to identify the major contributing sources, so effective best management practices can be established. Urban water pollution originates from wastewater point source discharge of treated effluents, wastewater non-point source intrusions from septic tanks, misapplication of recycled wastewater used for irrigation, or urban runoff from mobilization of pollutants deposited to impervious surfaces through atmospheric deposition and human activities during small rain events, erosion from pervious surfaces during large storm events, and dry weather flows (Paul and Meyer, 2001). Developing an indicator specific to wastewater effluents would aid in determining the contribution of these sources toward water quality impairment in recreational waters or potable supply sources (Glassmeyer et al., 2005; Stackelberg et al., 2004). The characteristics of an ideal wastewater indicator include: (i) source specificity, (ii) sustained effluent release because the indicator is not rapidly degraded by biological treatment processes, (iii) demonstrated analytical methodology, (iv) no attenuation during transport, and (v) virtually zero background with a sufficiently large discharge to detection level ratio able to exceed receiving water dilution factors. Traditional markers such as stable isotope ratios (e.g. d15N) and inorganic ions (e.g. chloride) often lack source specificity (Gasser et al., 2010; Katz et al., 2004). Data interpretation of stable isotope ratios is also difficult due to fractionation which obscures mixing during environmental transport (Gasser et al., 2010; Fry, 2006). Reconnaissance surveys of surface water bodies in the United States (Kumar and Xagoraraki, 2010) have suggested several anthropogenic organic compounds used as pharmaceuticals, personal care products, food products, pesticides, and hospital wastes as potential chemical markers of pollutant loading due to their behavior as persistent aqueous organic pollutants (Benotti et al., 2009; Bester, 2007; Buerge et al., 2009; Focazio et al., 2008; Glassmeyer et al., 2005; Guo and Krasner, 2009; Jjemba, 2008; Standley et al., 2008; Yamamota et al., 2009). Table 1 summarizes these candidate anthropogenic organic compounds and that which is known about their characteristics to serve as a wastewater marker due to occurrence in treated wastewater effluents, resistance to secondary wastewater treatment process operations, and persistence in U.S. waters due to a sustained rate of discharge which appears to exceed environmental attenuation processes from partitioning (i.e. volatilization, sorption, biotic uptake) or transformation (i.e. biodegradation, sunlight photolysis, and abiotic hydrolysis or redox reactions). The literature presents a potential short-list of compounds that still warrant a more methodical evaluation of their performance resistance to false positives and false negatives in identifying wastewater loading to surface water bodies. Notably absent from the short-list for most studies performed in the United States are artificial sugar
substitutes which have been included in several reconnaissance surveys in Europe (Scheurer et al., 2011, 2010, 2009; Buerge et al., 2009; Brorstro¨m-Lunde´n et al., 2008; Loos et al., 2009). Acesulfame, cyclamate, saccharin, and sucralose are commonly consumed in Europe with per capita wastewater loads (mg/Cap d) in Switzerland estimated at 10 3.4 for acesulfame, 11 6.7 for cyclamate, 3.9 1.7 for saccharin, and 1.5 0.6 for sucralose (Buerge et al., 2009) and values in Germany observed at 4.6 and 6.9 for acesulfame, 0.07 and 0.5 for cyclamate, 0.5 and 0.6 for saccharin, and 0.11 and 0.18 for sucralose (Scheurer et al., 2009). Sucralose per capita wastewater loads have been estimated at only 0.14e0.23 mg/(Cap d) in Germany and at 1.7e2.1 mg/(Cap d) in Sweden (Neset et al., 2010). Cyclamate and saccharin are unsuitable as markers because high levels of reduction (>90%) through biological treatment result in low effluent concentrations and cyclamate is additionally unsuitable in the United States since it has been banned from distribution since 1970. Acesulfame has been suggested as the most suitable chemical marker of domestic wastewater in Europe because of its approximate ten-fold higher concentration than sucralose in European wastewater effluents (approximately 10e50 mg/L in influents/effluents for acesulfame and <1e10 mg/L for sucralose). No information is available on per capita wastewater loads of sucralose in the United States, but it is anticipated to be higher than Europe due to its much longer period of availability (U.S. introduction in 1988 versus 2004 in Europe) and different regional dietary habits. Sucralose degradation through wastewater treatment facilities has also been demonstrated to be minimal for measurements through full-scale facilities and laboratory-scale aerobic biodegradation reactors (Torres et al., 2011; Buerge et al., 2009; Scheurer et al., 2009, 2010; and Neset et al., 2010). Pharmaceutical usage patterns were shown to differ between Europe and the U.S. (Sedlak and Pinkston, 2001) and differences in artificial sweetener usage and consumption is also probable. Other than one publication tracing sucralose along the salinity transect of a North Carolina river estuary impacted by municipal wastewater treatment plant effluent discharged upstream, this study is the first to evaluate the potential use of the artificial sweetener sucralose as a chemical marker for domestic wastewater input to surface waters within the United States. The evaluation considers sucralose presence in municipal wastewater effluents relative to other water sources and its susceptibility to false positives and false negatives through analysis of source waters with and without municipal wastewater discharges.
2.
Materials and methods
2.1.
Study sites
Samples collected for analysis were obtained from municipal wastewater treatment facility effluents, from drinking water intake sources with and without known wastewater point source discharges upstream of the intakes, and from active septic systems. The wastewater effluent locations were obtained from facilities located in Florida, Texas, northern and southern California, Illinois, and Michigan. The source waters with known upstream municipal wastewater discharges were
4021
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 1 9 e4 0 2 7
Table 1 e Reported detection frequencies (DF) of organic chemicals in wastewater treatment plant (WWTP) effluents, percent removal through activated sludge processes, and persistence in the environment as occurrence rank for U.S. streams and source waters. Compound (usage)
Average DF (%) in WWTP effluenta
Activated sludge treatment average percent removalb
Occurrence rank of compounds in U.S. stream/source waterc
acetaminophene (anti-inflamatory) atenolol (beta blocker) atrazine (pesticide) benzophenone (fragrance) bisphenol-A (plasticizer) caffeine (stimulant) 1,4-dichlorobenzene (household cleaning product) 1,7-dimethylxanthine (caffeine metabolite) carbamazepine (mood stabilizer) codeine (cough suppresant) coprostanol (sterol) cholesterol (sterol) cotinine (nicotine metabolite) N,N-diethyl-meta-toluamide (DEET) (insect repellent) diazinon (pesticide) diclofenac (anti-inflamatory) dehydronifedipine diltiazem (channel blocker) diphenylhydramine (antihistamine) diuron (pesticide) estrone (hormone) ethyl citrate (food additive) galaxolide (HHCB) (fragrance) gemfibrozil (anti-chloesterol) ibuprofen (anti-inflamatory) meprobamate (muscle relaxant) metolachlor (herbicide) naproxen (anti-inflammatory) 4-nonylphenol (detergent) phenytoin (dilantin) (seizure control) primidone (mood stabilizer) b-sitosterol (plant steroid) sulfamethoxazole (antibiotic) tonalide (AHTN) (fragrance) tributylphosphate (flame retardant) triclosan (antimicrobial) trimethoprim (anti-bacterial) tris(2-butoxyethyl)phosphate (flame retardant) tris(2-chloroethyl)phosphate (flame retardant) tris(dichlorisopropyl)phosphate (flame retardant)
low concentrationd low concentrationd low concentrationd 100 low concentrationd 81 not on target list
97 (n ¼ 4) 61 (n ¼ 4) no data 84 (n ¼ 6) 78 (n ¼ 41) 94 (n ¼ 7) not listed
40 25 1 30 3 not listed not listed
not on target list 88 84 low concentrationd low concentrationd low concentrationd 89
77 (n ¼ 1) 22 (n ¼ 5) 29 (n ¼ 1) 97 (n ¼ 1) 85 (n ¼ 1) not listed 54 (n ¼ 7)
48 4 78 not listed not listed 5 22
not on target list 75 not on target list not on target list 91 not on target list low concentrationd not on target list 100 92 78 83 not on target list 92 100 100 100 not on target list 94 100 not on target list 98 86 not on target list low concentrationd 100
not listed 44 (n ¼ 23) not listed not listed not listed not listed 77 (n ¼ 46) not listed 56 (n ¼ 25) 77 (n ¼ 13) 90 (n ¼ 32) not listed not listed 85 (n ¼ 18) 78 (n ¼ 10) 44 (n ¼ 1) not listed not listed 58 (n ¼ 15) 67 (n ¼ 20) not listed 60 (n ¼ 10) not listed not listed 27 (n ¼ 2) not listed
not listed 70 9 75 35 not listed 73 28 29 79 93 21 42 62 67 23 not listed not listed 24 6 not listed 54 39 2 20 7
a b c d
Dickenson et al., 2010. USEPA, 2010. Kumar and Xagoraraki, 2010. <5 limit of quantitation (LOQ).
located in northern and southern California, Colorado, Ohio, and New York. These samples were taken in the vicinity of the drinking water intake and while the percentage of wastewater present during sample collection is not known, the California location has been shown to contain sewage treatment plant effluent in the range of 20e70 percent of the flow during a severe drought year (Loraine and Pettigrove, 2006). The source waters classified without municipal wastewater discharges were located in northern and southern California, New York, Michigan, and Illinois and consisted either of rivers without upstream discharges or extremely large lakes with intakes removed from
known discharges. The septic samples were collected from systems located in Leon County and Palm Beach County, Florida.
2.2.
Sample collection and preservation
Samples were collected in 40 mL amber glass vials and field preserved with either a combination of 5 mg ascorbic acid and 50 mg sodium azide (Benotti et al., 2009; Vanderford et al., 2010) or 3 mg sodium omadine and 5 mg ascorbic acid (USEPA, 2010). Samples were transported on frozen blue ice and received in the laboratory either the same day or the next
4022
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 1 9 e4 0 2 7
day after collection and then stored refrigerated at < 6 C. Both preservatives have been shown to be effective for stabilizing many pharmaceuticals and personal care products (PPCPs) for 28 days or more under these conditions. All sample containers were provided by a single laboratory and sampling was performed with chain-of-custody documentation.
2.3.
Analysis
A fully automated on-line solid phase extraction, high performance liquid chromatography, mass spectrometry-mass spectrometry (SPE-HPLC-MS/MS) system was developed for rapid analysis of various groups of compounds of emerging concern (CECs) in water matrices. A column switching technique similar to the method of Buerge et al. (2008) was used for analysis of sucralose and other PPCP’s. The mass spectrometer used was an API 5000 triple quadrupole (AB/SCIEX) equipped with electrospray ionization (ESI) (TurboIonSpray). The interface setting and collision gas pressure were manually optimized. Parameter tuning for maximum sensitivity of multi reaction mode (MRM) detection in positive and negative ion mode was carried out by means of the optimization algorithm supported by the Analyst 1.5 software using the TurboIonSpray interface and a syringe diffusion pump
continuously supplying a standard (10 ul/min, C ¼ 10 ug/L for each compound). The resulting instrument values were then cross-checked for their validity under ESI conditions by flow injection MS/MS analysis (FIA-MS/MS) of the standard mixture and manual variation of the settings. Multiple mass transitions were used for each analyte (wherever there was adequate sensitivity for both transitions) to ensure unequivocal compound identification. For sucralose, both transitions were detectable at the MRL of 100 ng/L. The second transition was approximately 1/3 of the primary transition, but both transitions were more than ten times above the signal to noise ratio. Sample extraction was carried out using an integrated Dionex UltiMate 3000 system comprised of two HPG-3200 SD binary pumps, a WPS3000 SL semi-preparative auto sampler with 2.5 mL sample loop, and a TCC-3000SD column heater equipped with a ten port switching valve (Dionex, Sunnyvale, CA). An XBridge-C18 (2.1 150 mm 3.5 mm particle size) column (Waters, Milliford, MA) was used for ESI negative mode and a Luna C18(2)-HST (3 100 mm, 2.5 mm particle size) column (Phenomenex, Torrance, CA) was used for ESI positive mode. An Oasis HLB (2.1 10 mm 25 m) cartridge (Waters, Milliford, MA) was selected as the on-line SPE sorbent which contains both hydrophilic and hydrophobic materials suitable for multiresidue analysis in environment. One of the HPG-
Fig. 1 e Sucralose concentration in wastewater effluents collected from facilities with varying levels of nitrification and denitrification (A [ oxidzation ditch with methanol feed to denitrification filter, B [ bardenpho with dual media deep bed filters, C [ conventional activated sludge with carousel aeration and denitrification basin; D [ complete mix activated sludge with biological nutrient removal, E [ activated sludge with fine bubble diffused air and filtration, F [ modified Ludzak-Ettinger activated sludge, G [ complete mix activated sludge with anoxic basin, H [ oxidation ditch). Duplicates are a second sample bottle collected from the same site.
4023
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 1 9 e4 0 2 7
Fig. 2 e Effect of dilution factor and internal standard on sucralose quantification.
3200 SD binary pumps was used for sample delivery to the online SPE at a flow rate of 1.2 mL/min. Prior to extraction the short Oasis HLB cartridge was conditioned with 5 mL acetonitrile and 5 mL HPLC grade water acidified to pH 3. The delivery eluent pH was adjusted to approximately pH 3 by adding 250 ml of formic acid per 1 L of HPLC grade water. These
conditions represent a best compromise for the target analyte list and the HPLC eluant of 0.2% ammonium hydroxide and acetonitrile and the acidic pH 3 condition used to load the online SPE were both observed to improve the peak shape for sucralose. For both ESI modes the organic eluent B was acetonitrile and Eluent A for ESI positive mode was HPLC
Table 2 e Batch QC per Analytical Run. QC type
Frequency
calibration curve
each analytical batch
method blank MRL check continuing calibration sample matrix spikes closing blank closing standard identification criteria identification criteria
each analytical batch each analytical batch every 10 samples every 20 samples end of run end of run each detect each detect
a
Spiking level
Acceptance criterion
minimum 5 point curve plus blank
quadratic fit with 1/ weighting and correlation coefficient of 0.99; back-calculated to þ/30% of true value (þ/ 50% at MRL level) < 1/3 MRL 50e150% 70e130% 60e140% < 1/3 MRL 70e130% þ/ 1 s 2 MRM if available,ion ratio based on EU criteriab
0 low cal level 5 MRL level 5 MRL level 0 10 MRL retention time mass transitions
a An analytical batch consists of no more than 20 samples analyzed in a single 24-h run. b Draft SANCO 1805/2000 Rev. 1. [Revised Commission Decision 93/256 of 14 april 1993] laying down performance criteria for the analytical methods to be used for certain substances and residues thereof in live animals and animal products according to Council Directive 96/23/EC.
4024
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 1 9 e4 0 2 7
grade water adjusted to pH 3 For the ESI negative mode, due to the list of hormones and alkylphenols, a 0.2% solution of ammonium hydroxide was used. 2.5 mL of unfiltered sample was loaded to the 2.5 mL auto sampler loop and delivered to the on-line SPE and washed with 5% acetonitrile for 2 min, and the matrix diverted to waste. The valve position was then changed and the target analytes were refocused on the analytical column and then separated and eluted into the mass spectrometer using a gradient of 10% eluent B for 4 min, 100% eluent B at 20 min, then 5 min equilibration time at 10% eluent B again, and this conditions was kept for another 5 min during the next sample upload into the sample loop. Sucralose and other PPCPs (85 total analytes) were analyzed using the above method, with the exception of the sucralose data provided in Fig. 1 which was performed by the Florida Department of Environmental Protection using the EPA Method 3535 offline pre concentration technique and EPA Method 8321B. For the on-line concentration method, all standards and QC samples were processed the same way. Results for 44 analytes were determined via isotope dilution and when isotopes were unavailable, a close eluting isotope was utilized. Sucralose data was quantified using both gemfibrozil and sucralose-d6 as the internal standard. Fig. 2 provides a comparison of sucralose results for a suite of samples that were evaluated using multiple quantitation techniques, including dilution and quantification against gemfibrozil as an internal standard at both 1 and 10 and also quantification against sucralose-d6 at both 1 and 10 dilutions. At low concentrations (which were less “dirty” matrices) there is minimal difference between quantitation
techniques. The sucralose signal suppression from background matrix could be largely corrected by dilution, although quantitation was typically 15% lower using 10 dilution and gemfibrozil as the internal standard as compared with sucralose d6 isotope dilution. The recovery for sucralose-d6 was greater than 50%. Since all samples quantified with gemfibrozil were diluted ten-fold or more before analysis to minimize matrix effects, data reported by either internal standard method are comparable.
2.4.
Quality assurance
Minimum reporting levels (MRLs) for analytes ranged from 1 to 100 ng/L. MRLs were determined as the lowest calibration point used which also corresponds to a S/N ratio of at least 10, assuring accurate and precise quantitation (91% average recovery and relative standard deviation of 39% for sucralose and 106% average recovery and relative standard deviation of 15% for carbamazepine over a 15 month period) at the MRL levels of 100 and 5 ng/L respectively. Each analytical batch included numerous quality control (QC) samples as summarized in Table 2.
3.
Results and discussion
3.1.
Database evaluation of conservative indicators
A summary of the compounds detected in the wastewater effluents and two categories of source waters are presented in Table 3. Only compounds detected in at least 35% of the
Table 3 e Compounds detected in wastewater effluents and source waters with and without municipal wastewater discharges. Compound (MRL, ng//L)
sucralose (100) diuron (5) simazine (5) DEET (5) meprobamate (5) caffeine (10) diaminochlorotriazine (5) TCEP (5) bromacil (5) sulfamethoxazole (10) primidone (5) 2,4-D (5) amoxicillin (20) iohexal (10) atenolol (5) carisoprodol (5) gemfibrozil (5) carbamezapine (5) 1,7-dimethylxanthine (5) cotinine (10) dehydronifedipine (5) lopressor (20) theobromine (5)
Wastewater (ww) effluents
Sources with ww discharges
Sources without ww discharges
mean (ng/L)
rsd (%)
detects (%)
n
detects (%)
n
detect range (ng/L)
non detects (%)
n
detect range (ng/L)
27,000 99 21 269 323 127 36 547 95 907 159 248 1230 4780 1310 119 360 416 98 29 119 3900 151
30 78 100 135 197 159 209 66 100 116 49 262 92 120 1070 156 131 21 160 86 92 149 158
100 100 100 100 100 75 67 92 50 80 100 83 71 100 100 92 83 100 75 100 92 67 42
16 12 12 12 12 12 12 12 12 10 12 12 7 16 16 12 12 16 12 8 12 12 12
100 82 73 73 70 64 64 60 55 55 50 44 45 45 45 40 40 36 36 36 36 36 36
11 11 11 11 10 11 11 10 11 11 8 9 11 11 7 10 10 11 11 11 11 11 11
120e10,000 7.5e940 24e160 2.5e67 5.5e160 13e300 13e300 7.9e66 6e270 17e990 20e54 11e23 25e2200 73e960 6.1e200 5.4e43 13e130 31e190 8.9e23 13e27 12e120 22e270 6.4e41
100 80 20 13 100 100 40 47 93 100 100 60 100 87 92 100 100 100 100 100 87 100 67
15 15 15 15 15 15 15 15 15 15 15 15 14 15 13 15 15 15 13 15 15 13 15
e 5.3e6.7 7.1e61 2.2e7.1 e e 10e100 13e64 290 e e 7.4e21 e 16e39 19 e e e e e 7.7e70 e 7.8e25
4025
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 1 9 e4 0 2 7
40,000 35,000
Concentration, ng/L
30,000 25,000 20,000 15,000 10,000 5,000 0 Wastewater Effluent
Impacted Source
Clean Source
Fig. 3 e Nonparametric probability distribution of sucralose concentrations in different matrices depicted as box and whisker plots. Top and bottom of each box [ 75th and 25th percentiles, respectively; top and bottom of each whisker [ 90th and 10th percentile respectively; line across inside of each box [ median (50th percentile); and points beyond whiskers [ outliers. Clean source samples were all below the method reporting limit (MRL) and represented as a line at the 100 ng/L MRL.
sources with known wastewater discharges were captured in the table. Good markers of wastewater source loading should exhibit a high ratio of mean concentration to method reporting limit with 100 percent detection in the wastewater effluents. Low variability, denoted as % rsd in Table 3, also demonstrates a compound’s potential to serve as an indicator of the extent of wastewater impact (i.e., a correlation between concentration in the receiving stream and the fraction of stream flow due to upstream wastewater discharges) provided that the compound is stable in the environment during transport. The table shows that of the compounds exhibiting 100 percent detection in wastewater effluents, the best performing compounds in terms of ratio of mean concentration to MRL and lack of false positives were sucralose, meprobamate, and carbamazepine. The low relative standard deviation of sucralose and carbamazepine also indicates their potential to serve as quantitative markers of wastewater input, provided their environmental stability is adequately demonstrated. To function as an indicator of wastewater input in the environment, a compound should also demonstrate a high detection frequency in sources with known wastewater discharges (no false negatives) as well as absence from sources without known wastewater discharge influence (no false positives) (Gasser et al., 2010). For the samples and analytical methodology utilized to generate the dataset presented in Table 3, only sucralose demonstrated no false positives or false negatives. False positives demonstrate lack of source specificity and false negatives demonstrate lack of adequate sensitivity. Amoxicillin, carbamazepine, caffeine,
cotinine, gemfibrozil, meprobamate, primidone, sulfamethoxazole exhibited false negatives, stressing the need for detectability below a 5e10 ng/L MRL. Diuron, simazine, DEET, iohexal, and atenolol, exhibited false positives as well as false negatives. The efficacy of sucralose as a marker of conventional biologically treated wastewater is further demonstrated by Fig. 1 which shows the stable sustained presence of sucralose in wastewater effluents for a suite of samples collected throughout the state of Florida, regardless of the facility’s nutrient removal capabilities. Data in Fig. 1 are independent of the data presented in Table 3. Although this Fig. 1 does not include membrane bioreactor (MBR) facilities, a pilot study reported to the New Mexico Environment Department showed higher average sucralose levels of 42,400 ng/L in MBR effluent (Lee et al., 2010). This report also showed sucralose removal of approximately 40% for ozone and a biologically active filter and approximately 99% for reverse osmosis with trace levels detectable in the RO effluent indicating that sucralose is probably not appropriate as an indicator of wastewater that has gone through advanced treatment processes.
3.2.
Comparison of sucralose with carbamazepine
Sucralose and carbamazepine appear to be two of the most efficacious indicators of municipal wastewater input to surface water supplies, but adequate analytical sensitivity is an important issue. Comparison of Figs. 3 and 4 demonstrates the greater spread in the sucralose data between the wastewater effluents
4026
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 1 9 e4 0 2 7
400 350
Concentration, ng/L
300 250 200 150 100 50 0
Wastewater Effluent
Impacted Source
Clean Source
Fig. 4 e Nonparametric probability distribution of carbamazepine concentrations in different matrices depicted as box and whisker plots. Top and bottom of each box [ 75th and 25th percentiles, respectively; top and bottom of each whisker [ 90th and 10th percentile respectively; line across inside of box [ median (50th percentile); and points beyond whiskers [ outliers. Clean source samples were all below the method reporting limit (MRL) and represented as a line at the 5 ng/L MRL. For the impacted sources, 64% of the samples were left censored below the MRL and therefore only the 75th and 90th percentile and maximum values appear on the plot.
4. Table 4 e Comparison of sucralose and carbamazepine concentrations in single grab samples collected from eight septic systems located in two separate counties in Florida.
septic septic septic septic septic septic septic septic
1 2 3 4 5 6 7 8
Sucralose (ng/L)
Carbamazepine (ng/L)
69,000 40,000 80,000 42,000 24,000 40,000 12,000 12,000
<5 40 <5 <5 55 <5 <5 <5
and the impacted sources compared with the carbamazepine data. Sucralose therefore has greater ability to assess levels of dilution in an impacted source for sucralose and carbamazepine reporting limits of 100 ng/L and 5 ng/L respectively. Due to the differences in wastewater effluent concentrations, the performance of carbamazepine as a unique marker of municipal wastewater should be equivalent to sucralose only when the MRL of carbamazepine is less than 1 ng/L. Many laboratories, however, are not capable of achieving an MRL in the vicinity of 1 ng/L or lower for carbamazepine (Vanderford et al., 2010). Sucralose is also a superior indicator to carbamazepine for wastewater input originating from septic system intrusion as demonstrated by the data presented in Table 4 for eight septic samples collected in Florida.
Conclusions
These findings suggest the superior performance of the artificial sweetener sucralose to serve as an indicator for the presence of conventional biologically treated municipal and domestic wastewater and septic system sources to water bodies in the United States. The consistently high concentrations present in municipal wastewater effluents and septic systems along with a demonstrated absence of false negatives in impaired water sources and absence of false positives in clean sources supports its usefulness as an appropriate indicator compound for contaminant inputs of human origin that have not undergone advanced treatment through processes such as reverse osmosis. Additional studies on the fate of sucralose during environmental transport need to be conducted in order to assess whether the concentration of sucralose in the receiving water can also be used to quantitatively assess the fractional contribution of wastewater input.
Acknowledgments The authors gratefully acknowledge the WateReuse Research Foundation’s financial, technical, and administrative assistance in funding and managing the project through which a portion of this information was discovered. We are particularly indebted to Caroline Sherony in her role as the Foundation Project Officer. The comments and views detailed herein may not necessarily reflect the views of the WateReuse Research Foundation, its officers, directors, employees,
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 1 9 e4 0 2 7
affiliates or agents. Additional support was also received from the Florida Department of Environmental Protection, the Southwest Florida Water Management District, the South Florida Water Management District, the St Johns River Water Management District, the Orange County Utilities Department, the City of Orlando, Hillsborough County, the City of North Port and in-kind support from Miami-Dade Water and Sewer Department, JEA, and the City of Pompano Beach.
references
Benotti, M.J., Trenholm, R.A., Vanderford, B.J., Holady, J.C., Stanford, B.D., Snyder, S.A., 2009. Pharmaceuticals and endocrine disrupting compounds in U.S. drinking water. Environ. Sci. Technol. 43 (3), 597e603. Bester, K., 2007. Personal Care Compounds in the Environment. Pathways, Fate, and Methods for Determination. Wiley-VCH Verlag GmbH & Co., Weinheim, Federal Republic of Germany. Brorstro¨m- Lundẻn, E., Svenson, A., Viktor, T., Woldegiorgis, A., Remberger, M., Kaj, L., Dyem, C., Bjerke, A., Schlabach, M., 2008. Measurements of Sucralose in the Swedish Screening Program. Swedish Environmental Institute (IVL). Report B1769, January. Buerge, I.J., Kahle, M., Buser, H.-R., Mu¨ller, M.K., Piger, T., 2008. Nicotine derivatives in wastewater and surface waters: application as chemical markers for domestic wastewater. Environ. Sci. Technol. 42, 6354e6360. Buerge, I.J., Buser, H.-R., Kahle, M., Muller, M.D., Poiger, T., 2009. Ubiquitous occurrence of the artificial sweetener acesulfame in the aquatic environment: an ideal chemical marker of domestic wastewater in groundwater. Environ. Sci. Technol. 43 (12), 4381e4385. Burow, K.R., Nolan, B.T., Rupert, M.G., Dubrovsky, N.M., 2010. Nitrate in groundwater of the United States, 1991e2003. Environ. Sci. Technol. 44 (13), 4988e4997. Dickenson, E.R.V., Snyder, S.A., Sedlak, D.L., Drewes, J.E., 2010. Indicator compounds for assessment of wastewater effluent contributions to flow and water quality. Wat. Res. 45, 1199e1212. Focazio, M.J., Kolpin, D.W., Barnes, K.K., Furlong, E.T., Meyer, M.T., Zaugg, S.D., Barber, L.B., Thurman, M.E., 2008. A national reconnaissance for pharmaceuticals and other organic wastewater contaminants in the United States e (II) Untreated drinking water sources. Sci. Total Environ. 402, 201e216. Fry, B., 2006. Stable Isotope Ecology. Springer ScienceþBusiness Media, LLC, New York, N.Y. Gasser, G., Rona, M., Voloshenko, A., Shelkov, R., Tal, N., Pakrativ, I., Elhanany, S., Lev, O., 2010. Quantitative evaluation of tracers for quantification of wastewater contamination of potable water sources. Environ. Sci. Technol. 44, 3919e3925. Glassmeyer, S.T., Furlong, E.T., Kolpin, D.W., Cahill, J.D., Zaugg, S. D., Werner, S.L., Meyer, M.T., Kryak, D.D., 2005. Transport of chemical and microbial compounds from known wastewater discharges: potential for use as indicators of human fecal contamination. Environ. Sci. Technol. 39, 5157e5169. Guo, Y.C., Krasner, S.W., 2009. Occurrence of primidone, carbamazepine, caffeine, and precursors for n-nitrosodimethylamine in drinking water sources impacted by wastewater. JAWRA 45 (1), 58e67. Jjemba, P.K., 2008. Pharma-ecology, the Occurrence and Fate of Pharmaceuticals and Personal Care Products in the Environment. John Wiley & Sons, Inc., Hoboken, N.J. Katz, B.G., Chelette, A.R., Pratt, T.R., 2004. Use of chemical and isotopic tracers to assess nitrate contamination and ground-water age, Woodville Karst Plain, USA. Journal of Hydrol. 289, 36e61. Kumar, A., Xagoraraki, I., 2010. Pharamacueticals, personal care products and endocrine-disrupting chemicals in U.S. surface
4027
and finished waters: a proposed ranking system. Sci. Total Environ. 408, 5972e5989. Lee, C.O., Howe, K.J., Thomson, B.M., 2010. Ozone and Biofiltraton as an Alternative to Reverse Osmosis for Removing Ppcps and Edcs from Wastewater. Report to New Mexico Environment Department. http://www.unm.edu/whowe/UNM%20Howe% 20Final%20PPCP%20Ozone-Biofiltration%20Report.pdf. Loos, R., Gawlik, B.M., Boettcher, K., Locoro, G., Contini, S., Bidoglio, G., 2009. Sucralose screening in European surface waters using a solid-phase extraction-liquid chromatographytriple quadrupole mass spectrometry method. J. Chromatogr. 1216, 1126e1131. Loraine, G.A., Pettigrove, M.E., 2006. Seasonal variations in concentrations of pharmaceuticals and personal care products in drinking water and reclaimed wastewater in southern California. Environ. Sci. Technol. 40, 687e695. Neset, T.-S.S., Singer, H., Longree, P., Bader, H.-P., Scheidegger, R., Wittmer, A., Andersson, J.C.M., 2010. Understanding consumption-related sucralose emissions e A conceptual approach combining substance-flow analysis with sampling analysis. Sci. Total Environ. 408, 3261e3269. Nix, Stephen J, 1994. Urban Stormwater Modeling and Simulatio. CRC Press, Inc., Boca Raton, FL. Paul, M.J., Meyer, J.L., 2001. Streams in the urban landscape. Annu. Rev. Ecol. Syst. 32, 333e365. Scheurer, M., Brauch, H.-J., Lange, F.T., 2009. Analysis and occurrence of seven artificial sweeteners in German waste water and surface water and in soil aquifer treatment (SAT). Anal. Bioanal. Chem. 394, 1585e1594. Scheurer, M., Storck, F.R., Brauch, J.-J., Lange, F.T., 2010. Performance of conventional multi-barrier drinking water treatment plants for the removal of four artificial sweeteners. Wat. Res. 44, 3573e3584. Scheurer, M., Ru¨diger Storck, F., Graf, C., Brauch, J.-J., Ruck, W., Lev, O., Lange, F.T., 2011. Correlation of six anthropogenic markers in wastewater, surface water, bank filtrate, and soil aquifer treatment. J. Environ. Monit. 13, 966e973. Sedlak, D.L., Pinkston, K.E., 2001. Factors affecting the concentrations of pharmaceuticals released to the aquatic environment. Wat. Res. Update 120, 56e64. Stackelberg, P.E., Furlong, E.T., Meyer, M.T., Zaugg, S.D., Henderson, A.K., Reissman, D.B., 2004. Persistence of pharmaceutical compounds and other organic wastewater contaminants in a conventional drinking-water-treatment plant. Sci. Total Environ. 32, 99e113. Standley, L.J., Rudel, R.A., Swartz, C.H., Attfield, K.R., Christian, J., Erickson, M., Brody, J.G., 2008. Wastewater-contaminated groundwater as a source of endogenous hormones and pharmaceuticals to surface water ecosystems. Environ. Toxicol. Chem. 27 (12), 2457e2468. Torres, C.I., Ramakrishna, S., Chiu, C.-A., Nelson, K.G., Westerhoff, P., Krajmalnik-Brown-, R., 2011. Fate of sucralose during wastewater treatment. Environ. Eng. Sci. 28 (5), 325e331. USEPA, January 2009. National Water Quality Inventory: Report to Congress, 2004 Reporting Cycle: Findings. www.epa.gov/ owow/305b/2004report/factsheet2004305b.pdf. USEPA, August 2010. Treating Contaminants of Emerging Concern. In: A Literature Rev. Database. www.epa.gov/ waterscience/ppcp/studies/results.html. Vanderford, B., Snyder, S., Eaton, A., Guo, C., Ternes, T., Drewes, J., Wood, C., 2010. Fourth Periodic report for evaluation of analytical methods for EDCs and PPCPs via Interlaboratory comparison. WaterRF Project 4167 April 16, 2010. Yamamota, H., Nakamura, Y., Moriguchi, S., Nakamura, Y., Honda, Y., Tamura, I., Hirata, Y., Hayashi, A., Sekizawa, J., 2009. Persistence and partitioning of eight selected pharmaceuticals in the aquatic environment. Lab. photolysis, biodegradation, Sorption Experiments Wat. Res. 43, 351e362.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 2 8 e4 0 3 4
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Adsorption removal of boron in aqueous solutions by amine-modified tannin gel Shintaro Morisada*, Tetsuzen Rin, Takeshi Ogata, Yoen-Ho Kim, Yoshio Nakano Department of Environmental Chemistry and Engineering, Tokyo Institute of Technology, 4259 Nagatsuta, Midori-ku, Yokohama, Kanagawa 226 8502, Japan
article info
abstract
Article history:
A tannin gel (TG) synthesized from condensed tannin molecules has a remarkable ability to
Received 25 December 2010
adsorb various metal ions in aqueous solutions. In the present study, the adsorption
Received in revised form
removal of boron in solutions at various pHs and temperatures has been examined using
13 April 2011
the TG and the amine-modified tannin gel (ATG) prepared with ammonia treatment of the
Accepted 12 May 2011
TG. The adsorption amounts of boron for the TG and the ATG were relatively small and
Available online 20 May 2011
almost constant below pH 7, whereas the boron adsorption amounts increased with increasing pH in the range of pH above 7. Considering that in aqueous solutions above pH 7,
Keywords:
the mole fraction of boric acid decreases while that of tetrahydroxyborate ion increases
Tannin gel
with increasing pH, the boron adsorption onto both gels takes place probably through the
Boron removal
chelate formation of tetrahydroxyborate ion with the hydroxy and the amino groups in the
Adsorption
gels. Besides, the adsorbability of the ATG for boron was higher than that of the TG due to
Amine modification
the stable coordination bond between boron and nitrogen of the amino group in the ATG.
Water treatment
The adsorption kinetics were adequately described by the pseudo-second order kinetic equation while the adsorption isotherms followed both the Langmuir and the Freundlich equations. The boron adsorbability of both the TG and the ATG at low boron concentration were comparable or fairly good compared with other adsorbents. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
Boron occurs widely in nature, especially in seawaters and hot springs, while the waste waters from several industrial plants also include boron compounds. Although a trace amount of boron is essential for plant growth and used as a fertilizer component, it can be harmful for both plants and animals above certain concentration levels. According to the World Health Organization (WHO), the guideline value of the boron concentration in drinking water is 0.5 mg L1 (WHO, 2008). Many research groups have reported the boron removal from environmental and waste waters, using various methods such as chemical precipitation, adsorption, solvent extraction, reverse osmosis, and electrodialysis (Xu and Jiang, 2008).
Among these methods, an adsorption method, including an ion-exchange method, is promising for the boron removal because it requires the simple operating conditions and can be applied for the water treatment at low concentrations. For this reason, numerous studied on the boron adsorption have been conducted so far using various adsorbents such as natural and synthesized clay minerals (Hatcher and Bower, 1958; Sims and Bingham, 1967; Su and Suarez, 1995; Goldberg et al., 1996; Ferreira et al., 2006; Jiang et al., 2007; Xu and Peak, 2007; Kavak, 2009), ion-exchange resins (Kunin and Myers, 1947; Everest and Popiel, 1956; Simonnot et al., 2000; Kabay ¨ ztu¨rk, 2008), chelate resins (Bic¸ak and et al., 2007; Ko¨se and O S‚enkal, 1998; S‚enkal and Bic¸ak, 2003), functionalized mesoporous silicas (Kaftan et al., 2005; Wang et al., 2006),
* Corresponding author. Tel./fax: þ81 45 924 5419. E-mail address:
[email protected] (S. Morisada). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.05.010
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 2 8 e4 0 3 4
activated carbons (Choi and Chen, 1979; Kluczka et al., 2007a, b), metal oxides (Choi and Chen, 1979; Okay et al., 1985; Kluczka ¨ ztu¨rk and Kavak, 2008), cellulose (Okay et al., et al., 2007b; O ¨ ztu¨rk and Kavak, 2005; Chong et al., 2009). 1985), and ashes (O Recently, we have developed a novel adsorbent prepared from condensed tannin molecules by cross-linking with formaldehyde: the tannin molecule is an inexpensive and ubiquitous natural polymer extracted from leaves and barks of plants and has many hydroxy groups as shown in Fig. 1 (Hemingway et al., 1989). The resultant tannin gel (TG) has a significant ability to adsorb toxic metal ions (Nakano et al., 2001; Zhan et al., 2001; Kim et al., 2007b) and precious metal ions (Kim and Nakano, 2005, 2008; Ogata and Nakano, 2005; Kim et al., 2007a; Ogata et al., 2007). Moreover, we succeeded in enhancing the adsorbability for precious metal ions by the amine modification of the TG (Kim et al., 2009; Morisada et al., 2011). The application of such tannin-based materials has been also investigated actively by other research groups (Pizzi, 1982, 2009, 2000, 2006; Beltra´n-Heredia et al., 2009; Thevenon et al., 2009; Tondi et al., 2009a,b; Sa´nchez-Martı´n et al., 2010; Sa´nchez-Martı´n et al., 2011). In the present study, we have investigated the adsorption behavior of boron onto the TG and the amine-modified tannin gel (ATG) at boron concentrations not higher than 200 mg L1, where boron exists in the form of boric acid or tetrahydroxyborate ion, B(OH)3 or BðOHÞ 4 (Waggott, 1969). The effect of pH on the boron adsorption has been examined to elucidate the boron adsorption mechanism onto the TG and the ATG. Using a theoretical kinetic model, the adsorption kinetics for both gels have been analyzed. Lastly, the adsorption isotherms of boron have been obtained and then compared with those of other boron adsorbents.
4029
were prepared from boric acid (Wako Pure Chemical Industries Ltd., Japan). Deionized and distilled water was used in all procedure, and all reagents were used as received in the present study.
2.2. Preparation of tannin gel and amine-modified tannin gel The tannin gel (TG) was prepared following the previous work (Kim et al., 2007a). Wattle tannin powder (28 g) was dissolved in 45 mL of 0.25 M NaOH solution, and 6 mL of 37 wt% formaldehyde solution was successively added as a crosslinker. After gelation at 353 K for 12 h, the gel obtained was grinded into small particles and sieved by the screens with mesh sizes of 125 and 250 mm. The gel particles in the fraction of 125e250 mm were washed with distilled water and 0.05 M HNO3 solution to remove the residual chemicals and finally rinsed thoroughly with distilled water again. The amine-modified tannin gel (ATG) was prepared following our previous reports (Kim et al., 2009; Morisada et al., 2011). The freeze-dried TG particles were added in 10 wt% aqueous ammonia at 50 g-gel L1, and then the solution was shaken in a water bath at 333 K. After 12 h, the ATG particles were washed with 1 M HCl solution until the gel particles contained no ammonium ion, which can be retained in the gel through the electrostatic interaction with the deprotonated hydroxy groups of the tannin. The amine modification of the TG was confirmed by the elemental analyses and the solid-state CP-MAS 13C NMR measurements (Kim et al., 2009; Morisada et al., 2011). The zeta potentials of the TG and the ATG at 313 K and pH 3, 8.8, and 11 were obtained using a zeta potential analyzer (ZEECOM, Microtec Co., Ltd.), where the dosage of the TG or the ATG was 1 g L1 (dry basis) and the pH was adjusted using HNO3 and NaOH.
2.
Material and methods
2.3.
2.1.
Chemicals
All adsorption experiments in the present study were carried out in a batch system at a specific temperature T. The TG or the ATG particles (0.25 g, dry basis) were added into 50 mL aqueous boron solutions at different pHs, where the initial boron concentration was [B]ini ¼ 200 mg L1, and the initial pHs were adjusted using HNO3 and NaOH. The solution in a sealed vial was vigorously shaken in a water bath at T ¼ 303, 313, and 333 K and sampled at different time intervals for 8 h. To obtain the adsorption isotherms, 0.1 g of the dried gel was added into 50 mL aqueous boron solutions at [B]ini ¼ 10e200 mg L1 and pH 8.8, and then the sealed vial containing the sample solution was shaken in a water bath at T ¼ 303 K for 20 h. After filtration of the gel particles, the boron concentration of the sample solution was measured by an inductively coupled plasma spectrometer (ICPS-8100, Shimadzu). The amount of boron adsorbed onto the gel was calculated by the mass balance.
Wattle tannin powder (condensed tannin molecule) was kindly supplied by Mitsubishi Nuclear Fuel Co., Ltd., and all other reagents were of analytical grade. The stock solutions of boron
R3 OH
R2 O
HO
B OH
A OH R1
Fig. 1 e Estimated chemical structure of condensed tannin molecule: R1 [ OH and R2 [ H, phloroglucinolic; R1 [ R2 [ H, resorcinolic; R1 [ H and R2 [ OH, pyrogallolic; R3 [ H, catecholic; R3 [ OH, pyrogallolic (Hemingway et al., 1989).
Adsorption experiments
3.
Results and discussion
3.1.
Adsorption mechanism
Fig. 2 shows the adsorption amounts of boron onto the TG and the ATG, q, at initial pH 8.8 and T ¼ 303 K, as a function of time
4030
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 2 8 e4 0 3 4
10
1.0
TG
B(OH) 3 0.8
ATG Mole fraction [–]
–1
q [mg g ]
8 6 4 2 0
0
2
4
6
8
0.4
0.0
10
2
t. Both tannin gels adsorbed boron immediately, and the adsorption amounts reached their equilibrium values after t z 2 h. Besides, the adsorbability of the ATG for boron was better than that of the TG, indicating that the affinity of the amino groups for boron is higher than the hydroxy groups. Such tendency was also observed in the adsorption of the palladium and the platinum ions in aqueous solutions (Kim et al., 2009; Morisada et al., 2011). In general, the pH of the aqueous solution affects the adsorption behavior considerably. The equilibrium amounts of boron adsorbed onto the TG and the ATG, qe, at T ¼ 303 K and different initial pHs were plotted in Fig. 3. From pH 3 to 7, the values of qe for both the TG and the ATG are around 2 mg g1 and almost constant, while above pH 7, qe increases with increasing initial pH. In aqueous solution at the boron concentration below 270 mg L1, boric acid acts as a very weak and monobasic acid: þ BðOHÞ3 þ2H2 O#BðOHÞ 4 þ H3 O
(1)
where the acid dissociation constant is pKa ¼ 9.25 (Greenwood, 1975; Choi and Chen, 1979). Fig. 4 shows the distributions of B(OH)3 and BðOHÞ 4 as a function of pH. Below pH 7, boron exits in the form of B(OH)3, where the
10
TG ATG
8
4
6
8
10
12
pH [–]
Fig. 2 e Adsorption amounts of boron onto the TG and the ATG, q, at [B]ini [ 200 mg LL1, initial pH 8.8, and T [ 303 K, as a function of time t.
–1
0.6
0.2
t [h]
q e [mg g ]
–
B(OH) 4
6
Fig. 4 e Mole fractions of boric acid and tetrahydroxyborate ion, B(OH)3 and BðOHÞL 4 , in aqueous solution at the boron concentration below 270 mg LL1, as a function of pH.
boron adsorption amounts qe are small and almost constant as seen in Fig. 3. In the range of pH above 7, on the other hand, the mole fraction of B(OH)3 decreases, but that of BðOHÞ 4 increases with increasing pH. This increase in the fraction synchronizes with the increase in qe, BðOHÞ 4 indicating that boron is adsorbed by the TG and the ATG mainly in the form of BðOHÞ 4 rather than B(OH)3. One may think that the boron adsorption onto the TG and the ATG is attributable to the electrostatic attraction between BðOHÞ 4 and the tannin gel. The zeta potentials of the TG and the ATG at T ¼ 313 K and pH 3, 8.8, and 11 are listed in Table 1. The zeta potentials of both gels at pH 3.3 were near zero, while those at higher pHs were negative because of the deprotonation of the hydroxy and the amino groups. These values indicate that the electrostatic interaction between BðOHÞ 4 and the tannin gels plays no role in the boron adsorption. According to Oertel’s Raman study, BðOHÞ 4 reacts with polyols such as 1,2-ethanediol and 1,3-propanediol in aqueous solutions to form the anionic complexes with a ring structure (Oertel, 1972). Such chelate formation was also confirmed by 11 B NMR spectroscopy (Dawber and Green, 1986). Likewise, the amino groups are known to form coordination bonds with boron (Greenwood, 1975; Yao and Dong, 1995). Based on these published data and the favorable adsorption of BðOHÞ 4 in the present study, the adsorption mechanisms of boron onto the TG and the ATG are presumed as illustrated in Fig. 5. Yao and Dong (1995) reported that the boron chelate complexes with the boronnitrogen coordination bond are more stable against hydrolysis than the corresponding
4 Table 1 e Zeta potentials of the TG and the ATG at pH 3, 8.8, and 11 and T [ 313 K.
2 0
2
4
6
8
10
12
Gel
pH
zeta potential [mV]
TG
3 8.8 11 3 8.8 11
7.1 40.2 38.7 5.74 36.35 34.35
pH [–] Fig. 3 e Equilibrium adsorption amounts of boron onto the TG and the ATG, qe, at [B]ini [ 200 mg LL1, T [ 303 K, and different initial pHs.
ATG
4031
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 2 8 e4 0 3 4
a
R3
R3 OH B
b
B
B
+ OH
O
OH
HO
OH
HO
R3
B
HO
OH
+
2H2O
OH
H N
B
B
+ OH
OH
HO
B O
R3 NH2
OH
OH B
O
+
Table 2 e Pseudo-second order rate constants k2, equilibrium adsorption amounts qe, and correlation coefficients R2 for boron adsorption onto the TG and the ATG at initial pH 8.8 and T [ 303e333 K. Gel
T [K]
k2 [g mg1 h1]
qe (calcd.) [mg g1]
qe (expt.) [mg g1]
R2
TG
303 313 333 303 313 333
1.88 1.99 2.87 2.33 2.93 3.02
5.04 5.17 5.36 6.26 6.37 6.92
5.04 5.08 5.17 6.17 6.21 6.69
0.9865 0.9962 0.9937 0.9937 0.9918 0.9941
2H2O
OH
ATG
Fig. 5 e Possible adsorption mechanisms of boron onto the tannin gels: (a) TG; (b) ATG.
10 TG
borate esters without nitrogen. Due to the more stable coordination bond between boron and nitrogen in the amino group, the ATG showed the higher adsorbability for boron than the TG, as seen in Figs. 2 and 3.
Adsorption kinetics
The modeling of the adsorption kinetics is fundamentally important in water treatment process design. In the present study, we employed the pseudo-second order model to evaluate the adsorption kinetic data for the TG and the ATG obtained at initial pH 8.8 and T ¼ 303e333 K, because this model
–1
q [mg g ]
a
303 K
4
ATG
] –1
q [mg g
0
0
50
100 C e [mg L
–1
150
200
]
2 dq ¼ k2 qe q dt
10
313 K
8
4
1 1 ¼ þ k2 t qe qe q
10
]
0
c
8
(2)
where k2 is the pseudo-second order rate constant. Integrating Eq. (2) with the boundary conditions, t ¼ 0 to t ¼ t and q ¼ 0 to q ¼ q, gives
6
2
–1
2
expression provides the well correlation of the experimental and Aktay, 2002). data in many cases (Ho and McKay, 1999; Sag The kinetic equation for the pseudo-second order model is described as:
TG
0
q [mg g
4
6
2
b
6
Fig. 7 e Adsorption isotherms of boron for the TG and the ATG at initial pH 8.8 and T [ 303 K.
10 8
ATG
] –1
q e [mg g
3.2.
8
(3)
Above equation can be rewritten as: 1 q ¼ qe 1 1 þ k2 qe t
333 K
(4)
We then fitted Eq. (4) to the experimental results to estimate the rate constant k2 and the equilibrium adsorption amount qe as
6 4 2 0 0
2
4
6
8
t [h] Fig. 6 e Adsorption amounts of boron onto the TG and the ATG, q, at [B]ini [ 200 mg LL1, initial pH 8.8, and T [ 303e333 K, as a function of time t. The solid and broken lines represent fits of the pseudo-second order kinetic model to the experimental data.
Table 3 e Langmuir and Freundlich isotherm constants of boron for the TG and the ATG at initial pH 8.8 and T [ 303 K. Gel
Langmuir isotherm 1
1
kL [L mg ] qm [mg g ] TG ATG
0.00434 0.00283
11.4 24.3
Freundlich isotherm 2
R
0.9911 0.9875
kF [] 1/n [] 0.113 0.115
0.734 0.827
R2 0.9901 0.9847
4032
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 2 8 e4 0 3 4
Table 4 e Adsorption isotherms of boron for various adsorbents. Adsorbent TG ATG Polyol-grafted SBA-15 Polyol-grafted MCM-41 Activated carbon (AC) AC with tartaric acid AC with tartaric acid Activated alumina Zirconium dioxide Cerium dioxide Calcined alunite Palm oil mill boiler bottom ash
T [K] 303 303 303 303 293 293 293 293 293 293 298 298
pH 8.8 8.8 e e 6 6 6 6 6 e 10 e
Isotherm equation qe qe qe qe qe qe qe qe qe qe qe qe
¼ 0:0497Ce =1 þ 0:00434Ce ¼ 0:0687Ce =1 þ 0:00283Ce ¼ 0:542Ce =1 þ 0:0795Ce ¼ 0:125Ce =1 þ 0:0462Ce ¼ 0:0282Ce =1 þ 0:065Ce ¼ 0:0505Ce =1 þ 0:023Ce ¼ 0:0568Ce =1 þ 0:028Ce ¼ 0:0236Ce =1 þ 0:012Ce ¼ 0:0594Ce =1 þ 0:139Ce ¼ 0:0000264Ce ¼ 0:033Ce0:777 ¼ 0:0002Ce3:0345
q0.5a [mg g1] 2.48 3.43 2.61 6.11 1.37 2.50 2.80 1.17 2.78 1.32 1.93 2.44
2
10 102 101 102 102 102 102 102 102 105 102 105
Ref. This work This work Wang et al., 2006 Wang et al., 2006 Kluczka et al., 2007a Kluczka et al., 2007a Kluczka et al., 2007a Kluczka et al., 2007a Kluczka et al., 2007a ¨ ztu¨rk and Kavak, 2008 O Kavak, 2009 Chong et al., 2009
a Adsorption amount of boron onto the adsorbent at Ce ¼ 0.5 mg L1.
shown in Fig. 6. The resultant values and the correlation coefficients R2 are summarized in Table 2, where the experimental values of qe are also listed. The fairly high values of the correlation coefficients and the good agreement between the calculated and the experimental values of qe demonstrate that the pseudo-second order model adequately describes all the adsorption data obtained in the present study. The rate constant k2 increases with increasing temperature, while the value of qe is relatively temperature independent. Also, the adsorption rate for the ATG is faster than that for the TG, judging from the values of k2 listed in Table 2.
guideline value of Ce ¼ 0.5 mg L1, q0.5, are also listed for comparison. The values of q0.5 for the TG and the ATG are more or less higher compared with those for the other adsorbents except for the polyol-grafted mesoporous silica materials. However, considering that the tannin is an environmentallyfriendly and cost-effective material, in addition to the high adsorbability for boron, the TG and the ATG can be regarded as reasonable adsorbents for boron removal.
3.3.
We have examined the adsorption properties of the tannin gel (TG) and the amine-modified tannin gel (ATG) for boron in aqueous solutions at different pHs and temperatures, and then reached the following conclusions.
Adsorption isotherms
The adsorption isotherms of boron onto the TG and the ATG at initial pH 8.8 and T ¼ 303 K are shown in Fig. 7, where Ce is the equilibrium concentration of boron in the aqueous solution. The values of qe for the TG and the ATG are almost the same at Ce < 20 mg L1, whereas those for the ATG at Ce > 20 mg L1 are larger than those for the TG. We analyzed the experimental data shown in Fig. 7 using the Langmuir and the Freundlich equations, which are commonly used to describe adsorption isotherms. The Langmuir equation is represented by: qe ¼
kL qm Ce 1 þ kL Ce
(5)
where kL is the Langmuir adsorption constant related to the adsorption energy and qm is the adsorption capacity of the gel. The Freundlich equation is expressed as: qe ¼ kF Ce1=n
(6)
where kF and 1/n are the Freundlich adsorption constants. The constants in Eqs. (5) and (6), as well as the correlation coefficients R2, obtained by a nonlinear fitting are listed in Table 3. The adsorption isotherms of boron for the TG and the ATG shown in Fig. 7 are faithfully represented by both the Langmuir and the Freundlich equations.
3.4.
4.
Conclusion
(1) Both the tannin gels can adsorb boron in aqueous solutions effectively in the range of pH above 7. (2) The adsorption of boron onto both the TG and the ATG takes place probably through the chelate formation of tetrahydroxyborate ion, BðOHÞ 4 , with the hydroxy and the amino groups in the gels. (3) The adsorbability of the ATG for boron is superior to that of the TG, which is attributed to the stable coordination bond between boron and nitrogen of the amino group in the ATG. (4) The adsorption rates of boron onto both gels increase with increasing temperature, and the ATG can adsorb boron more quickly than the TG. (5) The adsorption isotherms of boron for the TG and the ATG are well described by both the Langmuir and the Freundlich equations. In comparison with other boron adsorbents, the TG and the ATG have the comparable or fairly higher adsorbability for boron at the WHO guideline concentration.
references
Comparison with other boron adsorbents
The adsorption isotherms of boron for the various adsorbents reported in the literatures are summarized in Table 4, where the amounts of boron adsorbed onto each adsorbent at the WHO
Beltra´n-Heredia, J., Sa´nchez-Martı´n, J., Solera-Herna´ndez, C., 2009. Anionic surfactants removal by natural coagulant/ flocculant products. Industrial and Engineering Chemistry Research 48 (10), 5085e5092.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 2 8 e4 0 3 4
Bic¸ak, N., S‚enkal, B.F., 1998. Sorbitol-modified poly(N-glycidyl styrene sulfonamide) for removal of boron. Journal of Applied Polymer Science 68 (13), 2113e2119. Choi, W.W., Chen, K.Y., 1979. Evaluation of boron removal by adsorption on solids. Environmental Science and Technology 13 (2), 189e196. Chong, M.F., Lee, K.P., Chieng, H.J., Ramli, I., 2009. Removal of boron from ceramic industry wastewater by adsorptionflocculation mechanism using palm oil mill boiler (POMB) bottom ash and polymer. Water Research 43 (13), 3326e3334. Dawber, J.G., Green, S.I.E., 1986. An 11B nuclear magnetic resonance study of the reaction of the tetrahydroxyborate ion with polyhydroxy compounds. Journal of the Chemical Society, Faraday Transactions 1 82 (11), 3407e3413. Everest, D.A., Popiel, W.J., 1956. Ion-exchange studies of solutions of borates. Journal of the Chemical Society, 3183e3189. Ferreira, O.P., de Moraes, S.G., Dura´n, N., Cornejo, L., Alves, O.L., 2006. Evaluation of boron removal from water by hydrotalcitelike compounds. Chemosphere 62 (1), 80e88. Goldberg, S., Forster, H.S., Lesch, S.M., Heick, E.L., 1996. Influence of anion competition on boron adsorption by clays and soils. Soil Science 161 (2), 99e103. Greenwood, N.N., 1975. The Chemistry of Boron. Pergamon Press, New York. Hatcher, J.T., Bower, C.A., 1958. Equilibria and dynamics of boron adsorption by soils. Soil Science 85 (6), 319e323. Hemingway, R.W., 1989. Structural variations in proanthocyanidins and their derivatives. In: Hemingway, R.W., Karchesy, J.J., Branham, S.J. (Eds.), Chemistry and Significance of Condensed Tannins. Plenum Press, New York, pp. 83e107. Ho, Y.S., McKay, G., 1999. Pseudo-second order model for sorption processes. Process Biochemistry 34 (5), 451e465. Jiang, J.Q., Xu, Y., Quill, K., Simon, J., Shettle, K., 2007. Laboratory study of boron removal by Mg/Al double-layered hydroxides. Industrial and Engineering Chemistry Research 46 (13), 4577e4583. ¨ ztu¨rk, N., 2008. Boron removal from aqueous Ko¨se, T.E., O solutions by ion-exchange resin: column sorption-elution studies. Journal of Hazardous Materials 152 (2), 744e749. Kabay, N., Sarp, S., Yuksel, M., Arar, O., Bryjak, M., 2007. Removal of boron from seawater by selective ion exchange resins. Reactive and Functional Polymers 67 (12), 1643e1650. lu, A.E., Shahwan, T., Artok, L., Ni, C.Y., Kaftan, O., Ac¸ikel, M., Erog 2005. Synthesis, characterization and application of a novel sorbent, glucamine-modified MCM-41, for the removal/ preconcentration of boron from waters. Analytica Chimica Acta 547 (1), 31e41. Kavak, D., 2009. Removal of boron from aqueous solutions by batch adsorption on calcined alunite using experimental design. Journal of Hazardous Materials 163 (1), 308e314. Kim, Y.-H., Nakano, Y., 2005. Adsorption mechanism of palladium by redox within condensed-tannin gel. Water Research 39 (7), 1324e1330. Kim, Y.-H., Nakano, Y., 2008. Effect of Br on the adsorption rate of palladium(II) ions onto condensed-tannin gel in chloride media. Separation Science and Technology 43 (9e10), 2386e2395. Kim, Y.-H., Ogata, T., Nakano, Y., 2007a. Kinetic analysis of palladium(II) adsorption process on condensed-tannin gel based on redox reaction models. Water Research 41 (14), 3043e3050. Kim, Y.-H., Ogawara, Y., Ogata, T., Nakano, Y., 2007b. Adsorption mechanism of selenite (Se4þ) by redox within condensedtannin gel under concentrated hydrochloric acid solution. Chemistry Letters 36 (11), 1316e1317. Kim, Y.-H., Alam, M.N., Marutani, Y., Ogata, T., Morisada, S., Nakano, Y., 2009. Improvement of Pd(II) adsorption performance of condensed-tannin gel by amine modification. Chemistry Letters 38 (10), 956e957.
4033
Kluczka, J., Ciba, J., Trojanowska, J., Zolotajkin, M., Turek, M., Dydo, P., 2007a. Removal of boron dissolved in water. Environmental Progress 26 (1), 71e77. Kluczka, J., Trojanowska, J., Zolotajkin, M., Ciba, J., Turek, M., Dydo, P., 2007b. Boron removal from wastewater using adsorbents. Environmental Technology 28 (1), 105e113. Kunin, R., Myers, R.J., 1947. The anion exchange equilibria in an anion exchange resin. Journal of the American Chemical Society 69 (11), 2874e2878. Morisada, S., Kim, Y.-H., Ogata, T., Marutani, Y., Nakano, Y., 2011. Improved adsorption behaviors of amine-modified tannin gel for palladium and platinum ions in acidic chloride solutions. Industrial and Engineering Chemistry Research 50 (4), 1875e1880. Nakano, Y., Takeshita, K., Tsutsumi, T., 2001. Adsorption mechanism of hexavalent chromium by redox within condensed-tannin gel. Water Research 35 (2), 496e500. ¨ ztu¨rk, N., Kavak, D., 2005. Adsorption of boron from aqueous O solutions using fly ash: batch and column studies. Journal of Hazardous Materials 127 (1e3), 81e88. ¨ ztu¨rk, N., Kavak, D., 2008. Boron removal from aqueous O solutions by batch adsorption onto cerium oxide using full factorial design. Desalination 223 (1e3), 106e112. Oertel, R.P., 1972. Raman study of aqueous monoborate-polyol complexes e equilibria in monoborate-1,2-ethanediol system. Inorganic Chemistry 11 (3), 544e549. Ogata, T., Nakano, Y., 2005. Mechanisms of gold recovery from aqueous solutions using a novel tannin gel adsorbent synthesized from natural condensed tannin. Water Research 39 (18), 4281e4286. Ogata, T., Kim, Y.-H., Nakano, Y., 2007. Selective recovery process for gold utilizing a functional gel derived from natural condensed tannin. Journal of Chemical Engineering of Japan 40 (3), 270e274. Okay, O., Gu¨c¸lu¨, H., Soner, E., Balkas‚, T., 1985. Boron pollution in the Simav River, Turkey and various methods of boron removal. Water Research 19 (7), 857e862. Pizzi, A., 1982. Condensed tannins for adhesive. Industrial and Engineering Chemistry Product Research and Development 21 (3), 359e369. Pizzi, A., 2000. Tannery row e the story of some natural and synthetic wood adhesives. Wood Science and Technology 34 (4), 277e316. Pizzi, A., 2006. Recent developments in eco-efficient bio-based adhesives for wood bonding: opportunities and issues. Journal of Adhesion Science and Technology 20 (8), 829e846. Pizzi, A., 2009. Polyflavonoid tannins self-condensation adhesives for wood particleboard. Journal of Adhesion 85 (2e3), 57e68. Sa´nchez-Martı´n, J., Beltra´n-Heredia, J., Carmona-Murillo, C., 2011. Adsorbents from Schinopsis balansae: optimisation of significant variables. Industrial Crops and Products 33 (2), 409e417. Sa´nchez-Martı´n, J., Gonza´lez-Velasco, M., Beltra´n-Heredia, J., Gragera-Carvajal, J., Salguero-Ferna´ndez, J., 2010. Novel tannin-based adsorbent in removing cationic dye (Methylene Blue) from aqueous solution. Kinetics and equilibrium studies. Journal of Hazardous Materials 174 (1e3), 9e16. , Y., Aktay, Y., 2002. Kinetic studies on sorption of Cr(VI) and Sag Cu(II) ions by chitin, chitosan and Rhizopus arrhizus. Biochemical Engineering Journal 12 (2), 143e153. S‚enkal, B.F., Bic¸ak, N., 2003. Polymer supported iminodipropylene glycol functions for removal of boron. Reactive and Functional Polymers 55 (1), 27e33. Simonnot, M.O., Castel, C., Nicolaı¨, M., Rosin, C., Sardin, M., Jauffret, H., 2000. Boron removal from drinking water with a boron selective resin: is the treatment really selective? Water Research 34 (1), 109e116.
4034
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 2 8 e4 0 3 4
Sims, J.R., Bingham, F.T., 1967. Retention of boron by layer silicates, sesquioxides, and soil materials: I. Layer silicates. Soil Science Society of America Proceedings 31 (6), 728e732. Su, C., Suarez, D.L., 1995. Coordination of adsorbed boron: a FTIR spectroscopic study. Environmental Science and Technology 29 (2), 302e311. Thevenon, M.-F., Tondi, G., Pizzi, A., 2009. High performance tannin resin-boron wood preservatives for outdoor end-uses. European Journal of Wood and Wood Products 67 (1), 89e93. Tondi, G., Fierro, V., Pizzi, A., Celzard, A., 2009a. Tannin-based carbon foams. Carbon 47 (6), 1480e1492. Tondi, G., Oo, C.W., Pizzi, A., Trosa, A., Thevenon, M.F., 2009b. Metal adsorption of tannin based rigid foams. Industrial Crops and Products 29 (2e3), 336e340. Waggott, A., 1969. An investigation of potential problem of increasing boron concentrations in rivers and water courses. Water Research 3 (10), 749e765.
Wang, L., Qi, T., Zhang, Y., 2006. Novel organic-inorganic hybrid mesoporous materials for boron adsorption. Colloids and Surfaces A 275 (1e3), 73e78. WHO, 2008. Guidelines for Drinking-Water Quality, third ed.. World Healthe Organization, Geneva, Switzerland. Xu, Y., Jiang, J.Q., 2008. Technologies for boron removal. Industrial and Engineering Chemistry Research 47 (1), 16e24. Xu, D.N., Peak, D., 2007. Adsorption of boric acid on pure and humic acid coated am-Al(OH)3: a boron K-edge XANES study. Environmental Science and Technology 41 (3), 903e908. Yao, J.B., Dong, J.X., 1995. Improvement of hydrolytic stability of borate esters used as lubricant additives. Lubrication Engineering 51 (6), 475e479. Zhan, X.M., Miyazaki, A., Nakano, Y., 2001. Mechanisms of lead removal from aqueous solutions using a novel tannin gel adsorbent synthesized from natural condensed tannin. Journal of Chemical Engineering of Japan 34 (10), 1204e1210.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 3 5 e4 0 4 6
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Low-temperature (7 C) anaerobic treatment of a trichloroethylene-contaminated wastewater: Microbial community development Alma Siggins, Anne-Marie Enright, Vincent O’Flaherty* Microbial Ecology Laboratory, Microbiology, School of Natural Sciences and Ryan Institute, National University of Ireland, Galway (NUI, Galway), University Road, Galway, Ireland
article info
abstract
Article history:
The feasibility of low-temperature (7 C) anaerobic digestion for the treatment of a trichlo-
Received 10 March 2011
roethylene (TCE) contaminated wastewater was investigated. Two expanded granular
Received in revised form
sludge bed (EGSB) bioreactors (R1 and R2) were employed for the mineralisation of
5 May 2011
a synthetic volatile fatty acid based wastewater at an initial organic loading rate (OLR) of
Accepted 12 May 2011
3 kg COD m3 d1, and an operating temperature of 15 C. Successive reductions in OLR to
Available online 20 May 2011
0.75 kg COD m3 d1, and operational temperature to 7 C, resulted in stable bioreactor
Keywords:
bioreactors. Subsequently, the influent to R1 was supplemented with increasing concen-
EGSB
trations (10, 20, 30 mg l1) of TCE, while R2 acted as a control. At an influent TCE concen-
Low-temperature anaerobic
tration of 30 mg l1, although phase average TCE removal rates of 79% were recorded,
digestion
a sustained decrease in R1 performance was observed, with COD removal of 6%, and % biogas
operation by day 417, with COD removal efficiency and biogas CH4 content 74%, for both
Specific methanogenic activity
CH4 of 3% recorded on days 595 and 607, respectively. Specific methanogenic activity (SMA)
Toxicity
assays identified a general shift from acetate- to hydrogen-mediated methanogenesis in
TCE
both R1 and R2 biomass, while toxicity assays confirmed an increased sensitivity of the
qPCR
acetoclastic community in R1 to TCE and dichloroethylene (DCE), which contributed to acetate accumulation. Quantitative Polymerase Chain Reaction (qPCR) analysis of the methanogenic community confirmed the dominance of hydrogenotrophic methanogens in both R1 and R2, representing 71e89% of the total methanogenic population, however acetoclastic Methanosaeta were the dominant organisms, based on 16S rRNA gene clone library analysis of reactor biomass. The greatest change in the bacterial community, as demonstrated by UPGMA analysis of DGGE banding profiles, was observed in R1 biomass between days 417 and 609, although 88% similarity was retained between these sampling points. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
Trichloroethylene (TCE) is a chlorinated aliphatic compound that is widely used in industry, and is particularly associated with the vapour degreasing of metals (Hansen et al., 2004). Recently, the
treatment of TCE-contaminated wastewaters via anaerobic digestion is emerging, as complete conversion of TCE to ethylene can be accomplished under anaerobic conditions by reductive dechlorination, due to symbiotic interactions between phylogenetic groups (Gu et al., 2004; Richardson et al., 2002).
* Corresponding author. Tel.: þ353 (0) 91 493734; fax: þ353 (0) 91 494598. E-mail address:
[email protected] (V. O’Flaherty). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.05.013
4036
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 3 5 e4 0 4 6
Although both laboratory- and full-scale anaerobic digestion bioreactor trials have been traditionally implemented at mesophilic or thermophilic conditions, the potential for lowtemperature adaptation of the anaerobic digestion process provides an attractive alternative. Significantly, previous arguments that low-temperature anaerobic digestion proceeds too slowly and inefficiently to be economically viable have now been addressed, with efficient degradation comparable to that from mesophilic trials demonstrated for a range of wastewater types (Collins et al., 2003; Connaughton et al., 2006; Enright et al., 2009; Lettinga et al., 1999; Madden et al., 2010; Nozhevnikova et al., 2000; Rebac et al., 1995). The development of bioreactor designs, such as the expanded granular sludge bed (EGSB) bioreactor (Zoutberg and de Been, 1997), was unquestionably accountable for advances towards successful anaerobic digestion trials at reduced operational temperatures. In addition, reducing the process temperature allows direct treatment of industrial wastewater at ambient temperature (average Irish temperature 2010, 7.9 C; Met Eireann, 2011), removing the costly energy requirement for heating of the system. Moreover, the treatment of volatile organic chemicals by low-temperature anaerobic digestion could decrease the rate of evaporation of these solvents during the treatment process; increased solubility of gaseous lower chlorinated compounds should allow for increased solvent:biomass contact, thereby maximising the opportunity for complete dechlorination to innocuous end products. In recent years, recognition of the importance of analysis of microbial community structure and function has allowed a greater understanding of the process, and optimisation of the technology (Enright et al., 2009; Madden et al., 2010). This is of particular importance when employing bioreactor trials for the treatment of a toxicant, and can allow for the identification of factors contributing to the failure of a system (Siggins et al., 2011). To date, nucleic acid based molecular techniques have been extensively employed for the analysis of the archaeal and bacterial communities present in anaerobic bioreactors at low temperatures (Chachkhiani et al., 2004; Enright et al., 2009; Madden et al., 2010), aiding the successful implementation of this technology.
In light of the above, the aim of this study was (1) to evaluate the feasibility of treatment of TCE by low-temperature (7 C) anaerobic digestion; (2) to determine the effect of TCE on the process of anaerobic digestion, by continued monitoring of bioreactor performance and metabolic analysis of the methanogenic activity and toxicity thresholds demonstrated by the granular biomass throughout the trial; (3) to monitor the adaptation of the archaeal and bacterial communities in response to the presence of TCE.
2.
Materials and methods
2.1.
Source of biomass
A granular anaerobic sludge was obtained from a mesophilic (37 C), full-scale bioreactor as described by Siggins et al. (2011).
2.2.
Design and operation of EGSB bioreactors
Two glass, laboratory scale (3.5 l) EGSB bioreactors, R1 and R2, were utilised for this 609 day study. R1 and R2 were each inoculated with 70 g VSS of biomass and employed for the treatment of a synthetic volatile fatty acid (VFA) based wastewater as described by Siggins et al. (2011). The initial 15 C operational temperature was decreased by 1 C on days 74, 81, 88, 95, 102, 109, 143 and 161, until a final temperature of 7 C was achieved. The initial 3 kg COD m3 d1 organic loading rate (OLR) was decreased to 1.5 kg COD m3 d1 on day 172, and subsequently to 0.75 kg COD m3 d1 on day 231, in response to an accumulation of VFA in both bioreactors. R1 influent was supplemented with TCE at increasing concentrations of 10, 20 and 30 mg l1 on days 418, 500 and 522, respectively, resulting in seven operational phases (Phase 1 e Phase 7; Table 1).
2.3.
Specific methanogenic activity and toxicity testing
Seed biomass and biomass sampled from the bioreactors on days 342 and 609 were screened for metabolic capability using
Table 1 e Operational and performance characteristics of R1 and R2 EGSB bioreactors. Values are averages of phases (P1eP7). Standard deviations are given in parenthesis, where applicable. n.a. not applicable. n.d. not determined. Phases
P1
P2
P3
P4
P5
P6
P7
Days
0e73
74e171
172e230
231e417
418e499
500e521
522e609
Operational temperature R1 Influent TCE (mg l1) % TCE removal Influent COD (mg l1) % COD removal
15 C 0 n.a. 3000 57 (17) 54 (17) 64 (12) 68 (6) n.d. n.d. n.d. n.d. n.d. n.d.
14e7 C 0 n.a. 3000 68 (13) 62 (16) 66 (12) 73 (4) 19 (12) 22 (12) 142 (45) 143 (41) 3 (3) 5 (4)
7 C 0 n.a. 1500 68 (20) 72 (15) 72 (6) 69 (12) 18 (15) 18 (13) 35 (32) 53 (57) 2 (2) 2 (2)
7 C 0 n.a. 750 76 (16) 75 (14) 74 (7) 75 (6) 10 (3) 10 (5) 6 (6) 5 (5) 1 (1) 1 (1)
7 C 10 19 (60) 750 82 (10) 83 (10) 51 (15) 70 (11) 11 (4) 8 (4) 4 (2) 1 (2) 0 (0) 0 (0)
7 C 20 86 (8) 750 78 (12) 83 (11) 68 (5) 75 (2) n.d. n.d. n.d. n.d. n.d. n.d.
7 C 30 79 (37) 750 56 (15) 86 (6) 46 (23) 73 (7) 375 (167) 135 (71) 0 (0) 0 (0) 2 (5) 0 (0)
% Biogas CH4 Effluent Acetic Acid mg COD l1 Effluent Propionic Acid mg COD l1 Effluent Butyric Acid mg COD l1
R1 R2 R1 R2 R1 R2 R1 R2 R1 R2
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 3 5 e4 0 4 6
4037
specific methanogenic activity (SMA) tests, which were performed using the pressure transducer technique (Coates et al., 1996; Colleran et al., 1992). The substrates acetate (30 mM), propionate (30 mM), butyrate (15 mM), ethanol (30 mM) and H2/CO2 (80:20, v/v) were employed to determine direct and indirect methanogenic activity, using suitable controls, as detailed in Siggins et al. (2011). SMA of seed biomass was determined at 37 C and 15 C, while SMA of bioreactor biomass sampled on days 342 and 609 was determined at 37 C, 15 C and 7 C. TCE, cis-1,2 dichloroethylene (DCE), trans-1,2 DCE and 1,1 DCE induced methanogenic toxicity of day 342 and day 609 bioreactor biomass was assessed at 7 C using the SMA based toxicity assay (acetoclastic and hydrogenotrophic) as described by Colleran and Pistilli (1994) and detailed in Siggins et al. (2011).
94 C for 10 min; followed by 30 cycles of denaturation at 94 C for 30 s, primer annealing at 55 C for 1 min and extension at 72 C for 1 min; followed by a final extension step at 72 C for 10 min. Controls containing no template DNA were employed to identify amplification of contaminants. 5 ml of each PCR product was visualised after electrophoresis on 1% agarose TAE (0.5% w/v Tris, 0.11% v/v acetic acid, 0.04% w/v EDTA) gel containing 1 mg ml1 Sybr Safe (Invitrogen) with Hyperladder IV (Bioline) as a molecular weight marker. Construction of clone libraries (TOPO XL), amplified rDNA restriction analysis and sequencing were performed as described by Siggins et al. (2011). The resulting partial 16S rRNA gene sequences generated by this study were deposited in the GenBank database under the accession numbers Seed HM749805eHM749813; day 417 HQ270537e HQ270547; day 609 HQ315889eHQ351901.
2.4.
2.7.
Analytical methods
qPCR
Reactor influent, effluent and biogas from R1 and R2 were routinely sampled. Influent and effluent COD and % biogas CH4 content were determined according to Standard Methods American Public Health Association (APHA, 1998), and % COD removal efficiency was determined from calculated influent and effluent measurements. VFA concentrations of R1 and R2 bioreactor effluent samples were determined from day 74 until the conclusion of the trial, by GC/MS as described in Siggins et al. (2011). VOC concentrations of TCE, cis-1,2 DCE, trans-1,2 DCE and 1,1 DCE in the effluent of R1 were determined from day 417 until the conclusion of the trial, using a Varian Saturn 2000 GC/ MS system. Separation was carried out on a Varian Capillary column, FactorFour VF-624ms (60 m length 0.32 mm internal diameter 1.8 mm film thickness, Varian). The temperature program was as follows: 50 C (1 min) to 250 C (1 min), at a rate of 10 C min1. All other parameters were as per VFA analysis. The minimum detection limit for VOC was 10 mg l1.
Quantitative real-time PCR was performed as described in Siggins et al. (2011). Four methanogenic primer and probe sets were employed, specific to two hydrogenotrophic orders (Methanomicrobiales and Methanobacteriales) and two acetoclastic families (Methanosaetaceae and Methanosarcinaceae), covering most methanogens present in anaerobic digesters (Lee et al., 2009; Yu et al., 2005). An adaptation of a moving-window approach, as described by Possemiers et al. (2004), was employed for statistical assessment of the extent of archaeal community dynamics. This adaptation allowed for the use of undefined time points, while traditional moving-window analysis utilises fixed time intervals. Similarity of the measured archaeal community between consecutive sampling points was calculated by Sorenson’s (BrayeCurtis) distance measurement using PC-ORD v5.0 (McCune and Grace, 2002).
2.5.
DGGE analysis of bacterial 16S rRNA genes from seed, day 417 and day 609 biomass was carried out after initial PCR-amplification using forward primer 341F (50 -CCT ACG GGA GGC AGC AG-30 ) and 517R (50 -ATT ACC GCG GCT GCT GG-30 ) (Muyzer et al., 1993), with a 40-base pair GC clamp attached to the 50 terminus of the forward primer. The touchdown PCR program consisted of an initial denaturation at 94 C for 2 min, followed by denaturation at 94 C for 30 s, annealing of primers (65e55 C; 1 cycle at 1 C increments; 20 cycles at 55 C) for 30 s and extension at 72 C for 30 s, followed by a final extension at 72 C for 10 min. A 40 ml aliquot of GC-clamped PCR product was loaded onto a 10% (w/ v) polyacrylamide gel containing a denaturing gradient of 30e70% (where 100% denaturant contained 7 M urea, 40% formamide) and ran at 60 C and 70 V for 16 h in a D-Code system (BioRad, Hercules, CA). The DGGE gels were ethidium bromide stained and photographed under UV transillumination. For sequencing and phylogenetic analysis, sixteen bands were excised from the gel using a sterile scalpel blade. Excised bands were suspended in 200 ml of sterile water, and stored at room temperature for 3 h to elute DNA from the gel for use as a PCR template. PCR reactions were performed under the conditions as described above and products were
Extraction of genomic DNA
Total genomic DNA was extracted from seed biomass and biomass sampled from both R1 and R2 on days 417, 500, 522 and 609 using an automated nucleic acid extractor (Magtration 12GC, PSS Co., Chiba, Japan) as described in Siggins et al. (2011). Extracted genomic DNA was eluted in 100 ml TriseHCl buffer and stored at 20 C.
2.6.
Clone library analysis of archaeal 16S rRNA gene
Archaeal clone libraries of seed, day 417 and day 609 biomass were constructed by the amplification of archaeal 16S rRNA genes with forward primer 21F (50 -TTC CGG TTG ATC CYG CCG GA-30 ; Stackebrandt and Goodfellow, 1991) and reverse primer 958R (50 -YCC GGC GTT GAM TCC AAT T-30 ; DeLong, 1992). Reaction mixtures (50 ml) contained 1.5 mM MgCl2, 5 ml 10 NH4 buffer (16 mM (NH4)2SO4, 67 mM TriseHCl (pH 8.8 at 25 C), 0.01% Tween-20), 50 pmol dNTP (dATP, dCTP, dGTP, dTTP), 12.5 pmol of each primer, 200 ng template DNA and 0.2 U Taq DNA polymerase. The PCR reactions were carried out under the following conditions: initial denaturation at
2.8.
Denaturing gradient gel electrophoresis (DGGE)
4038
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 3 5 e4 0 4 6
cloned using TOPO TA (Invitrogen). Five resulting plasmids per cloning reaction were extracted and a 2 ml aliquot was employed as a template for PCR using the same primers and conditions as described previously. For confirmatory purpose, PCR products were electrophoresed on a DGGE gel in parallel with the original samples. Bands obtained from extracted plasmids that underwent denaturation at the same gradient concentration as the original sample, and thereby migrated the same distance through the gel, were selected and sequenced (MWG). Sequences from this study were aligned with 16S rRNA gene sequences retrieved from the BLASTn using Clustal X (Thompson et al., 1997), and the phylogenetic inference package Paup* 4.0b8 was used for all phylogenetic analyses (Swofford, 2001). The partial 16S rRNA gene sequences obtained in this study were deposited in the GenBank database under the accession numbers HM749788eHM749796, HM749798, HM749800eHM749802, HM749804. DGGE gels were statistically analysed by creating binary matrices, where-by the presence or absence of bands, throughout the five individual samples was scored with the numeric values “1” or “0”, respectively. These matrices were used to calculate unweighted pair-group methods using arithmetic averages (UPGMAs) similarily dendograms using the PC-ORD 5.0 statistical package (McCune and Grace, 2002).
3.
Results
3.1.
Bioreactor performance
3.1.1.
Phases 1e4
A lengthy start-up period (at 15 C) of 73 days was recorded for both R1 and R2, during which time COD removal efficiencies fluctuated up to 20% between consecutive sampling points (Fig. 1a). The CH4 content of biogas from both bioreactors exceeded 70% by day 32, and remained at this level for the duration of P1 (Fig. 1b). The stability of COD removal efficiencies by both bioreactors was affected by the decreases in operational temperature implemented throughout P2, which corresponded to an accumulation of propionate in the effluents of R1 and R2 to 314 mg COD l1 and 345 mg COD l1, respectively, on day 150 (data not shown). Although stable COD removal efficiencies (ca. 70%; Fig. 1a) and % biogas CH4 (ca. 70%; Fig. 1b) were obtained for both R1 and R2 by the end of P2, the continued accumulation of propionate prompted the reduction of the applied OLR on day 171 to 1.5 kg COD m3 d1 (Table 1). While effluent propionate levels of both bioreactors reached <20 mg l1 within 20 HRT cycles of OLR reduction (data not shown), COD removal efficiencies for both R1 and R2 remained unstable throughout P3 (Fig. 1a). Additionally, a gradual decline of R1 and R2 % biogas CH4 was observed, and resulted in biogas CH4 levels in R1 of 56% on day 201, and R2 of 58% on day 203, however both bioreactors returned to >70% biogas CH4 within 4 days, and generally remained at that level until the end of P3 (Fig. 1b). In response to continued unstable COD removal efficiency, the applied OLR was further reduced to 0.75 kg COD m3 d1 on day 231 (Table 1), resulting in R1 COD removal efficiencies
Fig. 1 e (A) % COD removal efficiency, R1 (A) R2 (>); (B) % Biogas CH4 R1 (A) R2 (>).
of 88% by day 252, while R2 achieved 79% COD removal on day 265 (Fig. 1a). Further fluctuations in % COD removal efficiencies during P4 were reflected in both R1 and R2, and both bioreactors demonstrated comparable phase average % biogas CH4 (74% and 75%, respectively) and % COD removal (76% and 75%, respectively; Table 1). Finally, effluent VFA levels of both R1 and R2 remained <20 mg l1 during P4, resulting in comparable phase average effluent VFA concentrations between the two bioreactors (Table 1).
3.1.2.
Phases 5e7
The introduction of TCE to R1 at a concentration of 10 mg l1 on day 418 did not impact R1 % COD removal efficiency or effluent VFA concentrations (Fig. 1a; Table 1). Although a decrease in biogas CH4 was exhibited by both bioreactors on day 433, R2 recovered quickly, while R1 levels remained low, resulting in a P5 average of 51% biogas CH4 (Table 1). Significantly, the lowest level of R1 biogas CH4 measured during P5 corresponds with a cis-1,2 DCE peak detected in R1 effluent on day 460 (Figs. 1b and 2). Erratic periods of TCE removal and accumulation were detected during this phase,
4039
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 3 5 e4 0 4 6
with effluent TCE concentrations ranging from 21 mg l1 on day 422, to below detectable limits (10 mg l1) on day 460 (Fig. 2). This resulted in P5 average TCE removal rates of only 19%, with a high standard deviation (60; Table 1). The subsequent increase of influent TCE to 20 mg l1 on day 500 (P6) induced a temporary response in the COD removal efficiency of R1 biomass, which reached 50% on day 502, although this recovered to >90% by the next sampling point (Fig. 1a). In addition, a decrease in R2 COD removal efficiency to 62% was noted on day 502 (Fig. 1a). R1 biogas CH4 levels recovered during P6, with a phase average of 68%, only 7% lower than that of R2 (Table 1). Furthermore, effluent cis-1,2 DCE concentrations remained 5 mg l1 during P6, with 86% average TCE removal efficiency recorded during this phase (Table 1; Fig. 2) Due to GC/MS technical issues, effluent VFA concentrations were not measured during P6. A marked decline in R1 performance was noted after the influent TCE concentration was increased to 30 mg l1 on day 521. R1 COD removal efficiencies decreased to 42% on day 549 (Fig. 1a) and did not recover by the end of the trial, resulting in P7 average COD removal efficiencies of 56%, 30% lower than that of the control bioreactor, R2 (Table 1). VFA analysis of R1 effluent identified an accumulation of acetate, which reached 714 mg COD l1 on day 609, although a lesser degree of acetate accumulation was also noted in R2 (105 mg COD l1; data not shown). In addition, R1% biogas CH4 decreased to 17% by day 535, corresponding to a period of elevated levels of cis-1,2 DCE and TCE in R1 effluent (Figs. 1b and 2). A decrease in effluent TCE and cis-1,2 DCE levels by day 550 correlates with a temporary recovery in R1 biogas CH4 levels, however this recovery is not sustained, and deteriorates to <10% by the conclusion of the trial (Figs. 1b and 2). During the latter 50 days of bioreactor operation, R1 recorded TCE removal rates of 86e98%, with average P7 TCE removal of 79% (Fig. 2). Although cis-1,2 DCE was sporadically detected at concentrations up to c.16 mg l1, the accumulation of cis-1,2 DCE was not evident during the trial, indicating that further dechlorination was achieved (Fig. 2).
3.2.
Metabolic assays
3.2.1.
Specific methanogenic activity
SMA assays carried out throughout the trial confirmed the continued mesophilic nature of the biomass, as lower SMA values were recorded at temperatures lower than at 37 C (Table 2). Nevertheless, the development of a psychrotolerant methanogenic community was evident, with an increased activity recorded against all substrates at 15 C between days 342 and 609 (Table 2). On day 609, the TCE-supplemented bioreactor (R1) demonstrated higher SMA against H2/CO2 than acetate, while the control bioreactor (R2) recorded higher activity against acetate than H2/CO2 at all temperatures (Table 2). Additionally, while the acetoclastic methanogenic activity of R2 outperformed that of R1 on days 342 and 609, the opposite was true of hydrogenotrophic methanogenic activity, which was higher in R1 than R2 (Table 2). The propionate degrading activity of the seed biomass was extremely low at 15 C, recording only 2 ml CH4 g VSS1 day1 (Table 2). Between days 342 and 609, the 7 C SMA of R2 biomass against propionate increased from 3 to 9 ml CH4 g VSS1 day1, while the SMA of R1 biomass against propionate decreased from 6 to 0.4 ml CH4 g VSS1 day1 at 7 C (Table 2).
3.2.2.
Toxicity assay
Although the IC50 values for almost all of the toxicants, measured against acetate, were above the range of the assay on day 342 (Table 3), it can be seen that the R1 acetoclastic community demonstrated an increased sensitivity to the presence of TCE and its degradation derivatives from day 342 to day 609 (Table 3). The hydrogenotrophic methanogens displayed a higher tolerance to cis-1,2 DCE by day 609, with IC50 values increasing from 234 to >400 mg l1 (Table 3). However, this is not
Table 2 e Specific methanogenic activity profiles of seed, day 342 and day 609 biomass against direct and indirect methanogenic substrates, expressed as mlmethane g VSSL1 dayL1. All values are the mean of triplicates (std. error; n [ 3), except * where values are the mean of duplicates (std. error; n [ 2). Biomass Temperature
Seed R1 day 342
R2 day 342
R1 day 609
Fig. 2 e R1 VOC concentration; Influent TCE (solid line), effluent TCE (A), effluent cis-1,2 DCE (-x-), effluent trans-1,2 DCE (6), effluent 1,1 DCE (B).
R2 day 609
C
37 15 37 15 7 37 15 7 37 15 7 37 15 7
Substrates
H2/CO2 Acetate
Pro Butyrate Ethanol pionate
211 42 257 72 11 250 27 3 250 92 30 282 66 21
84 (8) 2 (1) 51 (13) 12 (3) 6 (1) 32 (5) 13 (1) 3 (1) 5 (2) 2 (1) 0.4 (0.2) 77 (3) 50 (1) 9 (1)
(12) (4) (12) (8) (3) (26) (4) (1) (8) (10) (1) (14) (1) (1)
470 23 197 62 20 313 121 42 167 42 11 307 84 24
(57) (2) (12) (4) (1)* (26) (1) (1) (4) (1) (1) (10) (2) (1)
165 (16)* 13 (3) 92 (7) 18 (1) 9 (1) 23 (4) 17 (1)* 6 (1) 94 (5) 39 (5) 12 (1) 107 (2) 35 (1) 10 (1)*
560 (31)* 29 (1) 219 (10)* 56 (2) 23 (1) 186 (20)* 73 (4) 10 (1) 202 (2) 47 (3) 14 (1)* 360 (14) 100 (1) 33 (1)
4040
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 3 5 e4 0 4 6
Table 3 e Batch toxicity assays; IC50 values against TCE and DCE isomers for day 342 and day 609 reactor biomass. Values are expressed as mg toxicant lL1. R1
R2
Acetate H2/CO2 Acetate H2/CO2 Day 342
TCE Cis-1,2 DCE Trans-1,2 DCE 1,1 DCE
>240 >240 >180 >120
>240 234 228 >300
>240 >240 >180 116
225 121 101 >300
Day 609
TCE Cis-1,2 DCE Trans-1,2 DCE 1,1 DCE
103 70 66 33
>125 >400 >125 >160
>750 310 >300 38
>250 66 >400 >500
reflected in R2, which recorded a 2-fold decrease of cis-1,2 DCE IC50 values against hydrogenotrophs from day 342 to day 609 (Table 3).
3.3.
Archaeal 16S rRNA gene clone library
Thirty-two distinct operational taxonomic units (OTUs) were identified from the 291 clones analysed from all 5 clone libraries. Specifically, the seed biomass contained eight OTUs, R1 biomass (day 417) contained five, R2 biomass (day 417) contained six, R1 biomass (day 609) contained nine and R2 biomass (day 609) contained four. BLASTn search results and phylogenetic reconstruction revealed that archaeal clones related to Methanomicrobiales were dominant in the seed biomass (49%; Fig. 3), although no Methanomicrobiales-like clones were detected in either bioreactor on trial day 417, and on day 609 were only detected in R1, accounting for 5% of the archaeal community (Fig. 3). Methanobacteriales-like clones were present in all samples, and represented 2e9% of the archaeal community (Fig. 3). The emergence of Methanosarcinales-like clones was a feature of all bioreactor samples e these organisms accounted for only 2% of the seed biomass, but between 58% and 93% of bioreactor biomass (Fig. 3). Furthermore, Methanosarcinaleslike clones from day 417 and day 609 biomass grouped closely with Methanosaeta species, while clones from the seed biomass grouped with Methanomethylovorans species (Fig. 3). Non-methanogenic archaea were detected in all samples except R2 on day 609 (Fig. 3). Specifically, Thermoplasmatales-like clones represented 21% of the seed biomass archaea, but were not detected in any other samples, while Crenarchaeota-like clones represented 28% of the seed archaeal community, 23% and 9% of R1 and R2 on day 417, respectively, and 32% of R1 on day 609 (Fig. 3). Consequently, all bioreactors demonstrated a shift in methanogenic community, from a dominance of the hydrogenotrophic methanogenic community in the seed biomass (51%), to the acetoclastic Methanosaeta in all bioreactors on days 471 and 609 (Fig. 3).
3.4.
qPCR
Methanosarcina species were only detected at levels below the lower limit of quantitation for this assay (106) and therefore were not included in any further analysis. However, the
applicable real-time PCR results show clear changes in the archaeal community structure of both bioreactors over time. Methanomicrobiales species were not detected in the seed biomass, but emerged in both R1 and R2 prior to the introduction of TCE, and were present for the remainder of the trial (Fig. 4). Furthermore, the Methanomicrobiales population developed steadily in R1 until day 522 and subsequently plateaued at ca. 1 108 copies g VSS1, while R2 levels peaked on day 500 at 2.5 108 copies g VSS1 and subsequently decreased, reaching 2.4 107 copies g VSS1 by the culmination of the trial (day 609; Fig. 4). Both bioreactors displayed comparable patterns in the community dynamics of Methanosaeta and Methanobacteriales species throughout the trial, with the exception of day 500 (Fig. 4). On day 500, when R1 influent was supplemented with 10 mg l1 TCE, R1 recorded 10- and 15-fold lower concentrations of Methanosaeta and Methanobacteriales species, respectively, than R2 (Fig. 4). Finally, although acetoclastic Methanosaeta species dominated the seed sludge, accounting for 75% of the total measured methanogenic population, hydrogenotrophic methanogens dominated both R1 and R2 biomass throughout the trial, representing 71e89% of the total measured methanogenic population (Fig. 4). Moving-window based analysis of qPCR data demonstrated a comparative shift in the archaeal community structure of both R1 and R2 from day 0 to day 417 (Fig. 5). However, divergence of R1 and R2 archaeal communities was noted between days 417 and 500, with moving window similarity coefficients (MWSCs) of 0.67 and 0.24, respectively (Fig. 5). Subsequent comparisons revealed that changes in the R2 archaeal community decreased throughout the trial, with 0.73 similarity noted between days 522 and 609 (Fig. 5). Conversely, adaptation of the R1 archaeal community throughout the trial was consistent, displaying <0.6 similarity from day 500 to day 522 and from day 522 to day 609 (Fig. 5).
3.5.
DGGE
UPGMA cluster analysis using Sorensons (BrayeCurtis) distance coefficient demonstrated that changes in the bacterial community structure from the seed biomass to biomass sampled on trial day 417 were conserved in R1 and R2, which maintained 100% similarity at that sampling point (Fig. 6). Furthermore, the bacterial community of R2 retained >98% similarity between days 417 and 609 (Fig. 6). The greatest change in community structure was exhibited by R1 between days 417 (Influent TCE 0 mg l1) and 609 (Influent TCE 30 mg l1), although R1 and R2 still retained 88% similarity on day 609 (Fig. 6). A total of 16 individual bands were excised for sequence analysis by BLASTn searches and phylogenetic reconstruction to identify the organisms affected by population changes. Eight of the sequenced bands, detected in each biomass sample, were phylogenetically associated with the phylum Proteobacteria, while the phyla Planctomycetes, Bacteroidetes and Chloroflexi, were also represented in all biomass samples (Fig. 7). Band 8, which was detected in the mesophilic seed biomass, grouped within the Thermotogae, although Thermotogae-like
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 3 5 e4 0 4 6
4041
Fig. 3 e Phylogeny of archaeal sequences obtained from seed sludge, day 417 and day 609 biomass, calculated using the Kimura-2 algorithm and the neighbour-joining method (Saitou and Nei, 1987). Bootstrap replicates (total 100 replicate samplings) supporting the branching order are shown at relevant nodes. Clone accession number and coverage of clonal sequences (%) are given in parenthesis.
species were not detected in any of the low-temperature bioreactor biomass samples (Fig. 7). The emergence of Firmicutes-like species was observed throughout the trial, as two bands (B2 and B14) were detected in the biomass of both R1 and R2 on days 417 and 609, but were not detected in the seed biomass (Fig. 7). Finally, one band (B7) grouped with unclassified bacterial sequences (Fig. 7).
4.
Discussion
This study demonstrated efficient TCE removal, of up to 98%, from acidified wastewaters at 7 C, at influent TCE
concentrations of 20 mg l1 (Table 1). The ability of a mesophilic anaerobic granular biomass to successfully adapt to lowtemperature operating conditions and to the presence of a chlorinated aliphatic compound, to which it has not been previously exposed, was also observed. Although the pre-TCE start-up phase was lengthy (417 days) compared to previous trials involving the low-temperature adaptation of mesophilic biomass, of between ca. 20 days (Enright et al., 2005; Madden et al., 2010) and 64 days (Scully et al., 2006), changes in operational parameters were undertaken slowly, as a series of minor adjustments. It is possible that this regime allowed the requisite time period for the development of an active cold-adapted anaerobic
4042
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 3 5 e4 0 4 6
Fig. 4 e Absolute quantification of the 16S rRNA gene concentration of the methanogenic archaeal orders Methanomicrobiales (MMB) and Methanobacteriales (MBT) and family Methanosaeta (MSt).
community, capable of rapid recovery following exposure to an inhibitory factor. Efficient bioreactor performance was achieved by both R1 and R2 in the latter stages of phase 1, as demonstrated by >71% COD removal efficiencies and >76% biogas CH4 content by trial day 73 (Fig. 1). Sequential decreases in operational temperature elicited negative responses in bioreactor performance, particularly 10 C, although these were temporary and stabilised after an acclimatisation period of generally <7 days (Fig. 1). Furthermore, minor changes in bioreactor performance were observed throughout the trial in both bioreactors (e.g. decrease in biogas CH4 on day 433) that appeared to be unrelated to TCE, thus confirming the importance of a control bioreactor for analysis of the treatment of toxic compounds. Overall, the development of an anaerobic consortium capable of the low-temperature anaerobic digestion of a low-
Fig. 5 e Moving analysis based on pair-wise similarity coefficients of consecutive sampling points from qPCR data; R1 (A) and R2 (>).
strength wastewater was achieved. Analysis of effluent VFA concentrations revealed that reductions in COD removal efficiencies for both R1 and R2 corresponded to an accumulation of propionate in the bioreactor system (Table 1), indicating that the degradation of propionate is a potential rate limiting step in anaerobic digestion at low temperatures, a finding in agreement with previous studies (Arbeli et al., 2006; Connaughton et al., 2006; Rebac et al., 1995). The reduction of the organic loading rates of both R1 and R2, on day 172, coincided with a decrease in propionate accumulation, although the achievement of stable COD removal performance required a further reduction in OLR to 0.75 kg COD m3 d1. Previous studies investigating low-temperature anaerobic digestion routinely employed expanded granular sludge bed e anaerobic filter (EGSB-AF) bioreactors, which maximise substrate/biomass interaction by allowing increased mixing intensity, while maintaining high levels of biomass retention. This system has been accredited with the stability of bioreactor performance at temperatures as low as 5 C (Enright et al., 2009; Madden et al., 2010; McKeown et al., 2009a; Scully et al., 2006). However, in order to decrease the surface area available for adsorption and physical removal of TCE and its degradation derivatives, an anaerobic filter was not provided in this study, possibly resulting in a less stable bioreactor, with respect to performance, than previously documented for the EGSB-AF configuration at low temperatures. Interestingly, the dechlorination of TCE appeared to affect % biogas CH4 levels, albeit temporarily, and peaks in effluent cis-1,2 DCE concentration corresponded to periods of low biogas CH4 production (Figs. 1b and 2). Therefore, it appears that the reductive dechlorination of TCE to DCE had a direct impact on biogas methane concentration. This is in correlation with previous reports that analysis of biogas methane concentration is not a reliable tool to monitor bioreactor stability as methane production by an anaerobic biomass decreases as the concentration of lower chlorinated alkenes formed by reductive dechlorination increases (DiStefano et al., 1991; Zhuang and Pavlostathis, 1994). Overall, competition for hydrogen may result in continued successful COD and TCE, in conjunction with poor biogas methane production (Zhuang and Pavlostathis, 1994). However, it appears that in this study an influent TCE concentration of 30 mg l1 (phase 7) exceeded the maximum threshold for continued successful anaerobic digestion at 7 C, as the COD removal efficiency of R1 was also affected at this TCE concentration (Fig. 1a). The provision of readily degradable substrates at low temperatures has been shown to allow the production of fatty acids to proceed more rapidly than methanogenesis, thereby resulting in the potential accumulation of acetate and propionate in the effluent of low-temperature bioreactors (Kalyuzhnyi et al., 2001; McHugh et al., 2006; Rebac et al., 1999), and this appear to be the case in this study. Specifically, the accumulation of acetate was noted in R1 and, to a lesser extent, in R2 during phase 7 (Table 1). Accordingly, SMA values at the bioreactor operation temperature (7 C) revealed a decline in methanogenic activity against acetate for both R1 and R2 biomass between days 342 and 609 (Table 2), indicating that acetoclastic methanogens of both bioreactors were adversely affected by the decreased operating temperature. Further inhibition of R1 acetoclastic methanogenesis was confirmed
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 3 5 e4 0 4 6
4043
Fig. 6 e Bacterial UPGMA cluster analysis of 16S rRNA gene fragments generated from DGGE profiles.
by toxicity assays, which demonstrated a lower tolerance of this trophic group to TCE and its degradation derivatives by the conclusion of the trial (Table 3). Moreover, accumulation of acetate could occur if the levels of acetoclastic methanogens within the bioreactor system were too low to maintain successful rates of acetate degradation (Nozhevnikova et al., 2000). qPCR analysis of specific methanogenic archaeal groups identified a shift in the proportion of the measured archaeal community away from a dominance of acetoclastic and towards a hydrogenotrophic methanogen-dominated system Specifically, the abundance of the hydrogenotrophic Methanomicrobiales species increased throughout the trial, concomitantly with increased SMA against H2/CO2. Although the emergence of Methanocorpusculum-like species has also been reported at low temperature (Madden et al., 2010; McHugh et al., 2004), no clones related to Methanocorpusculum were identified in this study. This trend towards hydrogenotrophic methanogenesis has been observed during previous low-temperature bioreactor trials (Collins et al., 2005; McHugh et al., 2004; McKeown et al., 2009a), although the precise reason for the development of hydrogenotrophic methanogenic activity is largely unknown. Overall, it appears that the acetoclastic methanogens are adversely affected by the reduced operating temperature, and are disadvantaged further by the presence of the supplemented toxic compounds. Clones phylogenetically associated with the non-methanogenic archaea, Crenarchaeota, were routinely detected throughout the trial. Crenarchaeota-like clones have previously been detected in anaerobic bioreactors (McKeown et al., 2009b) and are considered to have an important role in lowtemperature operation (Høj et al., 2008), although the precise nature of that role is as yet unknown. Both 16S rRNA gene clone libraries and qPCR were employed for analysis of the archaeal community structure, and interestingly, discrepancies were noted between the data generated by the two methods. While qPCR identified the dominance of hydrogenotrophic methanogens in both bioreactors from day 417 onwards supported by the metabolic assays, clones related to the acetoclastic Methanosaeta dominated 16S rRNA clone libraries during the same period (Figs. 3 and 4). Additionally, Methanomicrobiales-like clones represented 49% of the clone library constructed from seed biomass, but were not detected above the quantifiable limit (106) of the qPCR assay (Figs. 3 and 4). The use of species-specific primers, as opposed to general
bacterial or archaeal primers, may minimise biases and improve the quantification capabilities of PCR-amplification. Traditionally, biases of PCR-based methods, including the amplification of a more abundant template DNA fragment causing the suppression of the amplification of a minor template (Becker et al., 2000), or lack of primer specificity (Meyerhans et al., 1990), have been associated with the types of discrepancies described here. Therefore, while the construction of 16S rRNA gene clone libraries is a useful tool for the molecular analysis of microbial communities, more weight must be given to newer, fully quantitative methods, such as qPCR, particularly when undertaken in conjunction with species-specific primers. Moving-window analysis is a useful quantitative method for monitoring community dynamics and stability, and its use for monitoring microbial community dynamics in response to environmental changes has been increasingly documented (Bialek et al., 2011; Falk et al., 2009). By comparison of several statistical methods for analysis of qPCR data, Bialek et al. (2011) demonstrated that moving-window analysis better reflected performance changes in anaerobic bioreactors than nonmetric multidimensional scaling (NMS). In this study, movingwindow statistical analysis of qPCR data demonstrated that changes in the methanogenic community structure were more pronounced in the control bioreactor (R2) than in the TCE-supplemented R1. This is most apparent between days 500 and 417, i.e. after initial introduction of TCE to the influent of R1, when R1 shows > 60% similarity, while only ca. 20% similarity is retained by R2 (Fig. 5). This indicates that TCE did not induce a change in the archaeal community structure, but appeared to restrict the community development detected in the control bioreactor. It is possible, therefore, that ongoing adaptations of the bioreactors to decreased temperature and influent wastewater were hampered by the introduction of TCE to the influent of R1. The rate of change of the archaeal community of R2 decreased as the trial progressed, with >70% similarity displayed between days 522 and 609 (Fig. 5), possibly due to the development of a stable methanogenic community in response to the imposed operational conditions. A different trend was observed with respect to analysis of the bacterial community by DGGE. The same level of adaptation of the bacterial community from the seed biomass to day 417 was reflected in both bioreactors, in response to decreasing temperature and OLR (Fig. 6). Between days 0 and day 417, the
4044
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 3 5 e4 0 4 6
Fig. 7 e Phylogeny of bacterial sequences obtained by DGGE analysis of day 0 and day 299 biomass, calculated using the Kimura-2 algorithm and the neighbour-joining method (Saitou and Nei, 1987). Bootstrap replicates (total 100 replicate samplings) supporting the branching order are shown at relevant nodes. Clone accession numbers are given in parenthesis.
main differences identified in the bacterial community were the disappearance of bands associated with the thermophilic Thermotogae and the emergence of Firmicutes-like species. The abundance of Firmicutes-like species in anaerobic bioreactors has previously been reported (Chouari et al., 2005), particularly at low operating temperatures (Enright et al., 2007; Xing et al., 2009, 2010). By day 609, the bacterial community of R1 displayed the greatest difference from any of the other samples. It would appear that this response was a direct result of TCE addition to R1, as the control bioreactor only exhibited ca. 2% change in bacterial community structure during the same period (Fig. 6). In spite of the bacterial community
structures of R1 and R2 displaying 88% similarity on day 609, significant variations in bioreactor performance (Table 1) and methanogenic activity (Table 2) were noted between the two bioreactors during the final phase of the trial. Based on the findings of this study, it is possible that ongoing acclimatisation to the presence of a toxicant, such as TCE, resulted in the adaptation of the already existing microbial communities, rather than a significant shift in the community structure. Therefore, it appears that small changes in the bacterial community structure in response to TCE addition may have had an important impact on bioreactor performance. To elucidate this, further investigation targeting
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 3 5 e4 0 4 6
the roles and functions of the microbial populations involved in anaerobic digestion, particularly those species involved in syntrophic interactions, is essential.
5.
Conclusions
The following conclusions can be inferred from this study: (1) TCE-contaminated wastewater can be effectively treated at influent concentrations of up to 20 mg TCE l1 in an anaerobic EGSB bioreactor operating at 7 C; (2) An unacclimated mesophilic sludge is capable of sufficient adaptation to allow efficient methanogenesis at temperatures as low as 7 C; (3) Statistical analysis of qPCR data demonstrated that changes in the methanogenic community structure were more pronounced in the control bioreactor (R2) than in the TCE-supplemented R1, while statistical analysis of DGGE data indicated a response by the bacterial community structure of R1 biomass to the presence of TCE;
Acknowledgements The financial support of Enterprise Ireland, The Irish Environmental Protection Agency and Science Foundation Ireland is acknowledged.
references
APHA, 1998. Standard methods for the examination of water and wastewater. In: Clesceri, L.S., Greenberg, A.E., Eaton, A.D. (Eds.), twentieth ed. American Public Health Association, Washington, DC, USA. Arbeli, Z., Brenner, A., Abeliovich, A., 2006. Treatment of highstrength dairy wastewater in an anaerobic deep reservoir: analysis of the methanogenic fermentation pathway and the rate-limiting step. Water Res. 40 (19), 3653e3659. Becker, S., Bo¨ger, P., Oehlmann, R., Ernst, A., 2000. PCR bias in ecological analysis: a case study for quantitative Taq nuclease assays in analysis of microbial communities. Appl. Environ. Microbiol. 66 (11), 4945e4953. Bialek, K., Kim, J., Lee, C., Collins, G., Mahony, T., O’Flaherty, V., 2011. Quantitative and qualitative analyses of methanogenic community development in high-rate anaerobic bioreactors. Water Res. 45 (3), 1298e1308. Chachkhiani, M., Dabert, P., Abzianidze, T., Partskhaladze, G., Tsiklauri, L., Dudauri, T., Godon, J.J., 2004. 16S rDNA characterisation of bacterial and archaeal communities during start-up of anaerobic thermophilic digestion of cattle manure. Bioresour. Technol. 93 (3), 227e232. Chouari, R., Le Paslier, D., Daegelen, P., Ginestet, P., Weissenbach, J., Sghir, A., 2005. Novel predominant archaeal and bacterial groups revealed by molecular analysis of an anaerobic sludge digester. Environ. Microbiol. 7 (8), 1104e1115. Coates, J.D., Coughlan, M.F., Colleran, E., 1996. Simple method for the measurement of the hydrogenotrophic methanogenic activity of anaerobic sludges. J. Microbiol. Methods 26, 237e246.
4045
Colleran, E., Pistilli, A., 1994. Activity test system for determining the toxicity of xenobiotic chemicals to the methanogenic process. Ann. Microbiol. Enzymol. 44, 1e18. Colleran, E., Concannon, F., Golden, T., Geoghan, F., Crumlish, B., Killilea, J.E., Henry, M., Coates, J., 1992. Use of methanogenic activity tests to characterize anaerobic sludges, screen for anaerobic biodegradability and determine toxicity thresholds against individual anaerobic trophic groups and species. Water Sci. Technol. 25, 31e40. Collins, G., Foy, C., McHugh, S., Mahony, T., O’Flaherty, V., 2005. Anaerobic biological treatment of phenolic wastewater at 15e18 C. Water Res. 39, 1614e1620. Collins, G., Woods, A., McHugh, S., Carton, M.W., O’Flaherty, V., 2003. Microbial community structure and methanogenic activity during start-up of psychrophilic anaerobic digesters treating synthetic industrial wastewaters. FEMS Microbiol. Ecol. 46, 159e170. Connaughton, S., Collins, G., O’Flaherty, V., 2006. Psychrophilic and mesophilic anaerobic digestion of brewery effluent: a comparative study. Water Res. 40, 2503e2510. DeLong, E.F., 1992. Archaea in coastal marine sediments. Proc. Natl. Acad. Sci. U.S.A. 89, 5685e5689. DiStefano, T.D., Gossett, J.M., Zinder, S.H., 1991. Reductive dechlorination of high concentrations of tetrachloroethene to ethane by an anaerobic enrichment culture in the absence of methanogenesis. Appl. Environ. Microbiol. 57 (8), 2287e2292. Enright, A.M., McGrath, V., Gill, D., Collins, G., O’Flaherty, V., 2009. Effect of seed sludge and operation conditions on performance and archaeal community structure of lowtemperature anaerobic solvent-degrading bioreactors. Syst. Appl. Microbiol. 32, 65e79. Enright, A.M., Collins, G., O’Flaherty, V., 2007. Temporal microbial diversity changes in solvent-degrading anaerobic granular sludge from low-temperature (15 C) wastewater treatment bioreactors. Syst. Appl. Microbiol. 30, 471e482. Enright, A.M., McHugh, S., Collins, G., O’Flaherty, V., 2005. Lowtemperature anaerobic biological treatment of solvent-containing pharmaceutical wastewater. Water Res. 39, 4587e4596. Falk, M.W., Song, K.-G., Matiasek, M.G., Wuertz, S., 2009. Microbial community dynamics in replicate membrane bioreactors e Natural reproducible fluctuations. Water Res. 43, 842e852. Gu, A.Z., Hedlund, B.P., Staley, J.T., Strand, S.E., Steinsel, H.D., 2004. Analysis and comparison of the microbial community structures of two enrichment cultures capable of reductively dechlorinating TCE and cis-DCE. Environ. Microbiol. 6 (1), 45e54. Hansen, B.G., Munn, S.J., Berthault, F., Musset, C., Pakalin, S., Luotamo, M., deBruijn, J., Vegro, S., Pellegrini, S., Allanou, R., Scheer, S., 2004. European Union Risk Assessment Report: Trichloroethylene EUR 21057. In: Environment and Quality of Life Series 31. Høj, L., Olsen, R.A., Torsvik, V.L., 2008. Effects of temperature on the diversity and community structure of known methanogenic groups and other archaea in high Arctic peat. ISME J. 2, 74e83. Kalyuzhnyi, S.V., Gladchenko, M.A., Sklyar, V.I., Kizimenko, Y.S., Shcherbakov, S.S., 2001. One- and two- stage upflow anaerobic sludge-bed reactor pretreatment of winery wastewater at 4e10 C. Appl. Biochem. Biotech. 90 (2), 107e124. Lee, C., Kim, J., Hwang, K., O’Flaherty, V., Hwang, S., 2009. Quantitative analysis of methanogenic community dynamics in three anaerobic batch digesters treating different wastewaters. Water Res. 43, 157e165. Lettinga, G., Rebac, S., Parshina, S., Nozhevnikova, A., VanLier, J.B. , Stams, A.J.M., 1999. High-rate anaerobic treatment of wastewater at low temperatures. Appl. Environ. Microbiol. 65 (4), 1696e1702. Madden, P., Chinalia, F.A., Enright, A.M., Collins, G., O’Flaherty, V., 2010. Perturbation-independent community development in
4046
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 3 5 e4 0 4 6
low-temperature anaerobic biological wastewater treatment bioreactors. Biotechnol. Bioeng. 105 (1), 79e87. McCune, B., Grace, J.B., 2002. Analysis of Ecological Communities. MjM Software, Oregon. McHugh, S., Carton, M., Collins, G., O’Flaherty, V., 2004. Reactor performance and microbial community dynamics during anaerobic biological treatment of wastewaters at 16e37 C. FEMS Microbiol. Ecol. 48, 369e378. McHugh, S., Collins, G., O’Flaherty, V., 2006. Long-term, high-rate anaerobic biological treatment of whey wastewaters at psychrophilic temperatures. Bioresour. Technol. 97, 1669e1678. McKeown, R., Scully, C., Mahony, T., Collins, G., O’Flaherty, V., 2009a. Long-term (1243 days), low-temperature (4e15 C), anaerobic biotreatment of acidified wastewaters: bioprocess performance and physiological characteristics. Water Res. 43, 1611e1620. McKeown, R., Scully, C., Enright, A.M., Chinalia, F.A., Lee, C., Mahony, T., Collins, G., O’Flaherty, V., 2009b. Psychrophilic methanogenic community development during long-term cultivation of anaerobic granular biofilms. ISME J. 3, 1231e1242. Met Eireann, 2011. Monthly Climate Data. Last update 10 January. http://www.met.ie/climate/monthly-data.asp?Num¼69 Available:[19 January 2011]. Meyerhans, A., Vartanina, J.P., Wain-Hobson, S., 1990. DNA recombination during PCR. Nucleic Acids Res. 18, 1687e1691. 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. Appl. Environ. Microbiol. 59, 695e700. Nozhevnikova, A.N., Rebac, S., Kotsyurbenko, O.R., Parshina, S.N., Hollinger, C., Lettinga, G., 2000. Anaerobic production and degradation of volatile fatty acids in low temperature environments. Water Sci. Technol. 41 (12), 39e46. Possemiers, S., Verthe, K., Uyttendaele, S., Verstraete, W., 2004. PCR-DGGE-based quantification of stability of the microbial community in a simulator of the human intestinal microbial ecosystem. FEMS Microbiol. Ecol. 49 (3), 495e507. Rebac, S., Ruskova, J., Gerbens, S., van Lier, J.B., Stams, A.J.M., Lettinga, G., 1995. High-rate anaerobic treatment of wastewater under psychrophilic conditions. J. Ferment. Bioeng. 80 (5), 499e506. Rebac, S., Van Lier, J.B., Lens, P., Stams, A.J.M., Dekkers, F.T.M., Swinkels, K., Lettinga, G., 1999. Psychrophilic anaerobic
treatment of low strength wastewaters. Water Sci. Technol. 39 (5), 203e210. Richardson, R.E., Bhupathiraju, V.K., Song, D.L., Goulet, T.A., Alvarez-Cohen, L., 2002. Phylogenetic characterization of microbial communities that reductively dechlorinate TCE based upon a combination of molecular techniques. Environ. Sci. Technol. 36, 2652e2662. Saitou, N., Nei, M., 1987. The neighbour-joining method e a new method for reconstructing phylogenetic trees. Mol. Biol. Evol. 4 (4), 406e425. Scully, C., Collins, G., O’Flaherty, V., 2006. Anaerobic biological treatment of phenol at 9.5e15 C in an expanded granular sludge bed (EGSB)-based bioreactor. Water Res. 40, 3737e3744. Siggins, A., Enright, A.M., O’Flaherty, V., 2011. Methanogenic community development in anaerobic granular bioreactors treating trichloroethylene (TCE)-contaminated wastewater at 37 C and 15 C. Water Res. 45, 2452e2462. Stackebrandt, E., Goodfellow, M., 1991. Nucleic Acid Techniques in Bacterial Systematic. Wiley, England. Swofford, D.L., 2001. PAUP*. Phylogenetic Analysis using Parsimony (*and Other Methods), Version 4. Sinauer Associates, Sunderland, MA. Thompson, J.D., Gibson, T.J., Plewniak, F., Jeanmougin, F., Higgins, D.G., 1997. The CLUSTAL_X windows interface: flexible strategies for multiple sequence alignment aided by quality analysis tools. Nucleic Acids Res. 25 (24), 4876e4882. Xing, W., Zuo, J.E., Dai, N., Cheng, J., Li, J., 2009. Reactor performance and microbial community of an EGSB reactor operated at 20 and 15 C. J. Appl. Microbiol. 107 (3), 848e857. Xing, W., Zhao, Y., Zuo, J.E., 2010. Microbial activity and community structure in a lake sediment used for psychrophilic anaerobic wastewater treatment. J. Appl. Microbiol. 109 (5), 1829e1837. Yu, Y., Lee, C., Jaai, K., Hwang, S., 2005. Group-specific primer and probe sets to detect methanogenic communities using quantitative real-time polymerase chain reactions. Biotechnol. Bioeng. 89, 670e679. Zhuang, P., Pavlostathis, S.G., 1994. Effect of chlorinated alkenes on the reductive dechlorination and methane production processes. Water Sci. Technol. 30 (7), 85e94. Zoutberg, G.R., de Been, P., 1997. The Biobed EGSB (Expanded Granular Sludge Bed) system covers shortcomings of the upflow anaerobic sludge blanket reactor in the chemical industry. Water Sci. Technol. 35 (10), 183e188.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 4 7 e4 0 5 5
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Pesticides removal from waste water by activated carbon prepared from waste rubber tire V.K. Gupta a,b,*, Bina Gupta a, Arshi Rastogi c, Shilpi Agarwal a, Arunima Nayak a a
Department of Chemistry, Indian Institute of Technology Roorkee, Roorkee 247667, India Chemistry Department, King Fahd University of Petroleum and Minerals, Dhahran, 31261, Saudi Arabia c Chemistry Department, K.L.D.A.V (P.G.) College, Roorkee, India b
article info
abstract
Article history:
Waste rubber tire has been used for the removal of pesticides from waste water by
Received 28 January 2011
adsorption phenomenon. By applying successive chemical and thermal treatment, a basi-
Received in revised form
cally cabonaceous adsorbent is prepared which has not only a higher mesopore, macropore
30 March 2011
content but also has a favorable surface chemistry. Presence of oxygen functional groups
Accepted 17 May 2011
as evidenced by FTIR spectra along with excellent porous and surface properties were the
Available online 26 May 2011
driving force for good adsorption efficiency observed for the studied pesticides: methoxychlor, methyl parathion and atrazine. Batch adsorption studies revealed
Keywords:
maximum adsorption of 112.0 mg g1, 104.9 mg g1 and 88.9 mg g1 for methoxychlor,
Methoxychlor
atrazine and methyl parathion respectively occurring at a contact time of 60 min at pH 2
Atrazine
from an initial pesticide concentration of 12 mg/L. These promising results were confirmed
Methyl parathion
by column experiments; thereby establishing the practicality of the developed system.
Waste rubber tire
Effect of various operating parameters along with equilibrium, kinetic and thermodynamic
Adsorption isotherm
studies reveal the efficacy of the adsorbent with a higher adsorption capacity than most
Kinetics
other adsorbents. The adsorption equilibrium data obey Langmuir model and the kinetic data were well described by the pseudo-first-order model. Applicability of Bangham’s equation indicates that diffusion of pesticide molecules into pores of the adsorbent mainly controls the adsorption process. Spontaneous, exothermic and random characteristics of the process are confirmed by thermodynamic studies. The developed sorbent is inexpensive in comparison to commercial carbon and has a far better efficiency for pesticide removal than most other adsorbents reported in literature. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
Adsorption process by activated carbon is the most efficient and promising fundamental approach in the waste water treatment processes (Ali and Gupta, 2007; Gupta et al., 2009). The use of the adsorption technology in waste water treatment is still an expensive process, mainly because of the high adsorbent cost. Commercial activated carbons manufactured to produce precise surface properties are expensive and
require elaborate reactivation and regeneration processes. Such processes often result in the degradation of the adsorption properties of the carbon, which subsequently affects the economic viability of the operation. As a result lower cost adsorbents for a large-scale use in water decontamination processes are needed. Such adsorbents may be waste generated in abundance which, either as such or after their transformation into more active products, may be used in the adsorption processes.
* Corresponding author. Tel.: þ91 1332 285801. E-mail addresses:
[email protected],
[email protected] (V.K. Gupta). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.05.016
4048
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 4 7 e4 0 5 5
Discarded tires are an interesting and inexpensive medium for the adsorption of toxic pollutants from aqueous solutions (Ariyadejwanich et al., 2003). An estimated 1.5 106, 2.5 106 and 0.5 106 tonnes of used tires are generated each year in European Community, North America and Japan, respectively (Williams et al., 1990; Hamadi et al., 2004). It represents near to 15,000 tires per day, which makes the problem of solid waste management even more difficult to handle. Approximately 800,000 tires are reused each year, the remainder is land filled, stockpiled or illegally dumped (Castells, 2000). In turn, stockpiled scrap tires pose potentially serious health and safety problems. Whole tires serve as breeding grounds for disease carrying mosquitoes and rodents. Uncontrolled tire piles are fire hazards and, once ignited, tire piles can burn out of control for months, producing acrid black smoke and a hazardous oil residue. Widespread illegal dumping poses the same problems associated with stockpiling. These huge numbers of waste tires represent not only an enormous environmental problem but also a cheap source for the preparation of adsorbent materials that may be useful for the removal of toxic pollutants from solutions. Tire rubber is a mixture of different elastomers such as natural rubber (NB), butadiene rubber (BR) and styrene butadiene rubber (SBR) plus other additives like carbon black, sulfur and zinc oxide. Approximately 32% by weight of the waste tire (Murillo et al., 2006) is mainly constituted of carbon black in which the carbon content is as high as 70e75 wt.% (Castells, 2000). This carbonaceous adsorbent is rather similar to activated carbon and the only apparent physical difference is that carbon black has much less internal surface area (Knocke and Hemphill, 1981). Carbon black obtained by untreated rubber tire pyrolysis may be heated in air, carbon dioxide or steam atmosphere to develop its surface area and porosity (Mechant and Petrich, 1993; Teng et al., 1995; Cunliffe and Williams, 1999; Lin and Teng, 2002; Miguel et al., 2003; Helleur et al., 2001; Ariyadejwanich et al., 2003; Mui et al., 2004; Gonzalez et al., 2006; Suuberg and Aarna, 2007; Betancur et al., 2009; Lopez et al., 2009; Zabaniotou et al., 2004) and hence to improve its adsorption behavior. Various studies have been carried out on waste rubber tire activated carbon for liquid phase applications of pollutant removal (Lin and Teng, 2002; Miguel et al., 2003; Tanthapanichakoon et al., 2005; Mui et al., 2010). But in some waste water treatment applications involving macromolecules like pesticides and dyes which cannot easily penetrate into the micropores and adsorb onto them, the activated carbons should possess not only micropores but also mesopores (Hsieh and Teng, 2000). So our previous study centered on the development of a carbonaceous, mesoporous activated carbon (RTAC) from waste tire by physical activation method and successfully used it for the removal of dyes (Gupta et al., 2011). The present study aims to prepare activated carbons from tire char for pesticide adsorption. Keeping in consideration of the fact that pesticides are bulky and have complex chemical structures, a combination of chemical and physical treatment viz. chemical treatment of tire char prior to physical activation has been adopted in an attempt to develop surface porosity as well as to incorporate desired functional groups on the surface of the resultant activated carbons. The adsorption capacity of the carbons was further evaluated for pesticides removal from waste water. Methoxychlor, atrazine, and methyl parathion are
model pesticides selected for study; they are toxic and their applications have been banned but despite this, are most widely used in the third world countries because of low cost and versatility in industry, agriculture and public health (USEPA). Factors that are likely to affect the adsorption capacity of tire rubber were studied; in addition, adsorption isotherms and kinetic models have been determined and analyzed.
2.
Materials and methods
2.1.
Reagents
Technical grade methoxychlor, atrazine and methyl parathion, were obtained from M/s Merck, India. Physico-chemical and toxicological properties of the pesticides are reported in Table 1. The ground tire granules used in the study were purchased by Wiswani Chemicals. All reagents (NaOH, H2SO4, HNO3) used in the study were of analytical grade. Deionized water was used for making synthetic samples. Before each experiment, all glassware were cleaned with dilute nitric acid and repeatedly washed with deionized water.
2.2.
Equipment
pH measurements were made using a pH meter (Model Cyberscan 510,Singapore). The concentration of pesticides was determined by a gas chromatograph (Clarus 500 GC, India) with electron capture detector (ECD). LEO 435 VP (Leo Elektronenmikroskopie GmbH, Germany) scanning electron microscopy was used for scanning the adsorbent surface. Elemental analysis was carried out on an Elementar Vario ELHI CHNS analyzer. The infrared spectra of adsorbents were recorded in KBr discs on a infrared spectrophotometer (Model Perkin Elmer-1600 Series). The BET surface area of the adsorbent was measured on micromeritics ASAP 2010 (UK). Porosity of the adsorbent was determined by mercury porosimeter (Pascal 440; M/s Spektron Instrument Inc., India). X-ray measurements were performed by using a Philips X-ray diffractometer employing Ni-filtered Cu KR radiation and Ni filters. Deionized water was prepared using a Millipore Milli-Q (Bedford, MA) water purification system.
2.3.
Preparation of the adsorbent
Initial cleaning and carbonization of the ground tire granules were done as per our previous study (Gupta et al., 2011). The dried material (approx 2g) was mixed with 8g of KOH; mixing was allowed for 10 min, the product was further activated to 900 C for 2 h. The material was then treated with 1M HCl solution to remove the ash content, washed with deionized water and then dried at 100 C. The dried product referred to as CTRTAC henceforth was sieved to desired particle sizes and stored until required.
2.4.
Adsorption and kinetic studies
Batch adsorption studies were conducted as per our previous study (Gupta et al., 2006b, 2008, 2011). The parameters considered for study were pH, adsorbent dose, particle size,
4049
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 4 7 e4 0 5 5
Table 1 e Physico-chemical & toxicology properties of pesticides (USEPA). Pesticide
Methoxychlor
Atrazine
Methyl parathion
C16H15Cl3O2 345.65 Organochlorine Insecticide 0.045 ppm 4.68e5.08
C8H14ClN5 215.69 Organochlorine Insecticide 33.0 ppm 2.34
C8H10NO5PS 263.23 Organophosphate Insecticide 60.0 ppm 1.83e3.43
Highly persistent, non-mobile, high bio-accumulation potential Class D 0.052 mg/L III (moderately toxic) Restricted & banned 20 ppb
Moderately persistent & mobile, low bio-accumulation potential Class C <0.1 mg/L I (Highly toxic) Restricted & banned 3 ppb
Moderately persistent & mobile, low bio-accumulation potential Class A 0.163 mg/L I (Highly toxic) Restricted & banned 2 ppb
Molecular Structure
Molecular Formula Molecular Weight Chemical family Pesticide type Solubility in water (25 C) Log octanolewater partition coefficient Environmental fate Carcinogen class LC50 Toxicity Category Use classification EPA limit in drinking water
initial concentration of adsorbate, contact time and temperature. For kinetic and isothermal studies, a series of 250-mL Erlenmeyer flasks were filled with 100 mL pesticide solution of varying concentrations (2 mg/L to 12 mg/L), maintained at desired temperature and pH. The pH of the solution was kept constant by adding 0.1M NaOH or 0.1M HNO3. An equal amount of adsorbent was added separately into each individual flask. The flasks were agitated in an orbital shaker. Liquid samples were taken out at a given time interval and centrifuged at 1500 rpm for 10 min. The concentration of remaining pesticide in the adsorption medium was determined by the GC method as described later. The adsorption capacity for pesticides uptake, qe (mg/g), was determined as follows: qe ¼ ðC0 CÞV=W
(1)
where, C0 and C are the initial and final pesticide concentrations (mg/L), respectively, V is the volume of solution (L) and W is the weight of adsorbent (g). All the experiments were repeated three times and average values were reported. The standard deviation was found to be 0.14%; values of correlation coefficient being in the range 0.98e0.99.
2.5.
2.6.
GC analysis
The pesticides were extracted from aqueous solutions (obtained from batch and column experiments) using 10 mL of nhexane. The n-hexane layer was separated, concentrated to 1.0 mL and used for determination of the pesticides by gas chromatography. A HewlettePackard HP-5 MS fused silica capillary column (30 m 0.25 mm I.D., 0.25 mm film) was used with Helium as carrier gas, at a flow rate of 1.0 ml/min min with 25 mL/min as make up gas flow. A 1 mL sample was injected under split mode (split ratio 10:1). All the gas chromatographic experiments were carried out with the temperatures of the column, injector and detector as 260, 230, and 310 C, respectively. The GC column temperature program was as follows: an initial temperature 100 C, held for 3 min, and then ramp 10 C/min to 280 C held for 5 min. The retention time for methoxychlor, atrazine and methyl parathion were observed 13.8, 12.7 and 10.1 min respectively.
3.
Results and discussion
3.1.
Characterization of adsorbent material
Column studies
A glass column of length 30 cm and 1 cm internal diameter, filled with weighed amount of CTRTAC having particle size 200e250 mm, was used as a fixed bed adsorber and set up was established as per our previous study (Gupta et al., 2011). Experiments were carried out with an adsorbate solution of known concentration of 12 mg/L at 25 C. Samples of the effluent were collected and the effluent concentrations were analyzed by GC. The column was shut down when the runoff concentration matched the original concentration of the pesticide.
Since adsorption is a surface phenomenon, the rate and extent of adsorption specific to a given adsorbent are influenced by the physico-chemical characteristics of the adsorbent such as surface area, pore size, surface chemistry and elemental composition (Pal et al., 2006). CHNS analysis of CTRTAC before and after treatment shows the carbon content to be 22.2% and 78.43%, thereby revealing a transformation to carbonaceous nature. The X-ray diffraction spectrum pattern of the CTRTAC sample did not show any peak, thereby, indicating the amorphous nature of the product (Fig not shown). Prior treatment, BET and porosity of the rubber tire sample was 22 m2/g and
4050
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 4 7 e4 0 5 5
Table 2 e Effect of adsorbent dosage on the uptake of pesticides onto CTRTAC (temperature: 25 C, initial concentration: 12 mg/L; pH: 2). Adsorbent dose (g/L) 0.02 0.04 0.06 0.08 0.10 0.12 0.14
Fig. 1 e SEM micrograph (a) unactivated rubber tire (b) CTRTAC before adsorption (c) CTRTAC after adsorption of methoxychlor pesticide.
CTRTAC/ Atrazine qe (mg/g)
CTRTAC/ Methylparathion qe (mg/g)
49.60 68.90 86.60 100.20 112.00 111.80 111.90
37.90 51.20 74.40 90.70 104.90 104.60 104.00
26.40 41.80 62.70 77.40 88.90 88.40 88.00
from the SEM images. Fig. 1b reveals that the thermally and chemically treated sorbent is highly porous as compared to the untreated adsorbent (Fig. 1a). Surface texture of the CTRTAC after adsorption of pesticide (methoxychlor) is depicted in Fig. 1c. FTIR spectrum of the CTRTAC after treatment indicates the presence of two prominent bands lying at 1602 and 3456.09 cm1 which may be assigned to the presence of carbonyl group and hydroxyl group respectively. Prior to treatment, the spectra revealed an extra band at 1256 cm1 indicating the presence of carboxylic group. In the activation of KOH-impregnated samples for 2 h at 800e900 C, unstable oxygen groups (e.g. carboxyl groups) decomposes and only hydroxyl and carbonyl groups remain (Park and Jung, 2002). Successive physical and chemical treatment has helped in transforming a non-porous rubber tire sample to a well developed mesoporous adsorbent with a high surface area and favorable surface chemistry.
3.2. 2.39% respectively. CTRTAC samples showed an increase of both BET surface area to 981m2/g and porosity to 89.34%. Untreated rubber tire was non-porous but after treatment, total pore volume was 1.51cc/g out of which 72.21% pores were of 2e50 nm range with average pore diameter being 3.12 nm. The well developed mesoporous surface texture is further evident
CTRTAC/ Methoxychlor qe (mg/g)
Effect of contact time and concentration
The rate of uptake of pesticides on CTRTAC as evident from Fig. 2 depicts that the sorption is quite rapid initially, gradually slows down and then reaches the equilibrium. Equilibrium was attained at 60 min for CTRTAC-pesticide system. An increase in adsorption was observed from 41.1 to 112.0 mg/g for methoxychlor, 36.9e104.9 mg/g for atrazine and
Fig. 2 e Effect of contact time on the uptake of pesticides at two different initial concentrations [average value of 3 tests, std deviation <0.14%] (temperature: 25 C, particle size: 200e250 mm, pH: 2).
4051
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 4 7 e4 0 5 5
Fig. 4 e Adsorption isotherm of pesticides on CTRTAC at 25 C (Adsorption on untreated rubber negligible; hence not shown). [average value of 3 tests, std deviation <0.14%].
Fig. 3 e Effect of solution pH on the uptake of pesticides [average value of 3 tests, std deviation <0.14%] (temperature: 25 C, particle size: 200e250 mm, initial concentration: 12 mg/L; pH: 2).
commensurate increase in adsorption resulting from lower adsorptive capacity utilization of adsorbent (Raghuvanshi et al., 2004).
28.8e88.9 mg/g for methyl parathion with 5 mg/L and 12 mg/L initial concentration of the pesticides. The time of equilibration adsorption was unaffected with initial concentration, but the amount adsorbed increased by increasing concentration of the pesticides. The uptake capacity of pesticides molecules by the adsorbent, and the time required for establishment of equilibrium suggest the effectiveness of these materials for waste water treatment.
3.3.
3.4.
Effect of pH
To observe pH effects, the adsorption of the pesticides was studied over a pH range of 2.0e11.0 and the results are plotted in Fig. 3. Over the range of experimental pH, surface functional groups of CTRTAC (especially carbonyl, hydroxyl) become deprotonated and the extent of deprotonation increases with an increase in pH. This deprotonation results in a more negatively charged carbon surface at higher pH than at the lower pH values. The negative charge developed at the surface of the adsorbent as a result of deprotonation cause strong electrostatic repulsion for the pesticides in solution, thereby retarding diffusion and adsorption. While, with decrease in alkaline conditions or pH, deprotonation of adsorbent is reduced, thereby accelerating diffusion and adsorption (Gupta and Imran, 2008). An increase in pH from 2.0 to 11.0 resulted in decrease in adsorption from 112.0 to 56.8 mg/g for methoxychlor, 104.9 to 42.5 mg/g for atrazine and 88.9 to 27.8 mg/g for methyl parathion. Maximum adsorption of all three pesticides occurs at pH 2 and hence was selected as the optimum and used throughout the study.
Effect of adsorbent dose
The effect of adsorbent dose on the removal of pesticides was studied by varying the dose of adsorbent from 0.02 to 0.14 g/L. The experiment was carried out at fixed pesticide concentration at 12 mg/L, at a fixed temperature of 25 C and pH 2. From Table 2, it is evident that adsorbent dose significantly influences the amount of pesticide adsorbed. Initially, the amount of pesticide adsorbed was found to be rapid from 0.02 to 0.10 g/L. Further increase of adsorbent dose resulted in very less increase in adsorption, and hence, 0.1 g/L was considered the optimum dose (Table 2). The initial rise in adsorption with adsorbent dose is probably due to a stronger driving force and larger surface area. With a rise in adsorbent dose, there is a less
Table 3 e Adsorption Isotherm model parameters for the pesticides at different temperatures. Pesticides
Log Kow
Methoxychlor
4.68e5.08
Atrazine
2.34
Methyl parathion
1.83e3.43
Temp ( C)
25 35 45 25 35 45 25 35 45
Langmuir isotherm
Freundlich isotherm 2
qmax (mg/g)
b
R
111.11 97.09 94.34 100.00 95.24 83.33 90.91 66.67 50.00
9.10 5.15 2.65 5.00 2.63 2.00 3.67 1.50 1.00
0.997 0.998 0.994 0.996 0.995 0.996 0.998 0.995 0.996
Kf
n
R2
140.28 130.02 99.08 92.05 75.16 44.57 65.01 57.28 47.86
1.88 1.62 1.36 1.92 1.54 1.59 2.24 2.16 1.98
0.972 0.971 0.988 0.978 0.981 0.953 0.946 0.957 0.978
4052
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 4 7 e4 0 5 5
Table 4 e Thermodynamic Parameters for pesticide adsorption. DG (KJ/mole)
Pesticide
298K 308K 318K
DS R2 DH (KJ/mole) (J/K. mole)
Methoxychlor 34.27 27.59 20.92 233.15 Atrazine 17.15 13.09 9.02 138.27 Methyl 6.93 3.83 0.74 99.022 parathion
Fig. 5 e Langmuir adsorption isotherm of methoxychlor on CTRTAC at different temperatures [average value of 3 tests, std deviation <0.14%].
3.5.
Effect of particle size
The adsorption of pesticides was studied at different particle sizes of CTRTAC of 100e150,150e200 and 200e250 mm. The batch experiments were conducted at a fixed pesticide concentration of 12 mg/L, at a fixed temperature of 25 C and fixed pH of 2. By increasing mesh size from 100 to 150 mm to 200e250 mm, the adsorption capacity of methoxychlor was found to increase from 64.90 mg/g to 111.90 mg/g. Similar results are observed for atrazine (50.20 mg/g at 100e150 mm to 104.0 mg/g at 200e250 mm) and for methyl parathion (41.80 mg/ g at 100e150 mm to 88.0 mg/g at 200e250 mm). This leads to the conclusion that the amount adsorbed increases with increasing mesh size which is indeed true as decreasing mesh size has a decreasing particle size which in turn has a higher surface area. This may be due to the fact that the smaller adsorbent particles have shortened diffusion paths, such that the ability of adsorbate to penetrate all internal pores of the adsorbent is higher (Gupta and Imran, 2001). Particle size of 200e250 mm was, hence, chosen for further studies.
3.6.
Modeling of adsorption isotherms
Adsorption isotherms of the rubber tire before and after treatment on the pesticides are plotted in Fig. 4. It is evident from the figure that CTRTAC showed remarkable adsorption capacity for the pesticides. Untreated rubber tire showed
667.37 406.44 309.05
0.989 0.999 0.989
negligible adsorption characteristics as compared to CTRTAC and hence was not shown in Fig. 4. Thus it can be concluded that textural characteristics of CTRTAC was conducive to good adsorption capacities (Aharoni and Ungarish, 1977; Gupta et al., 2002); proving beyond doubt, that the adsorption process is a surface phenomenon and the nature of interaction between the adsorbent and adsorbate is, to some extent, physical in nature. This phenomenon is proved later in kinetic study. Isotherm data was analyzed with Langmuir and Freundlich models as discussed earlier (Gupta et al., 2011) and the results of their linear regression were used to find out the fit model. Values of Langmuir constants qmax and b, Freundlich constants KF and n, as well as the regression coefficients (R2) for all the pesticides are listed in Table 3. Langmuir model was found fit due to high correlation coefficients for CTRTACpesticides system. CTRTAC has maximum affinity for methoxychlor as indicated by the ‘b’ values. Fig. 5 indicates the linear plots of 1/qe vs 1/Ce for methoxychlor adsorption showing the applicability of the Langmuir equation. It is seen that the pesticides adsorbing onto CTRTAC follow the order: methoxychlor > atrazine > methyl parathion (Table 3). Also the log Kow (octanolewater partition coefficient) of pesticides is of the order: methyl parathion > atrazine > methoxychlor. A direct relation is thus observed between the adsorptive capacity and log Kow values of pesticides (Table 3). The log Kow, or OctanoleWater partition coefficient, is simply a measure of the hydrophobicity (water repelling) of an organic compound. The more log Kow value, the more hydrophobic a compound, the less soluble it is, therefore the more likely it will adsorb to an adsorbent (Bedient et al., 1994). Thus methoxychlor (having a log Kow value of 4.68e5.08) being least soluble in water (least affinity for water) is adsorbed at the solideliquid interface to the maximum extent. Methyparathion on the other hand shows least log Kow value of 1.83e3.43 and hence is least hydrophobic thereby is the least adsorbed onto CTRTAC.
Table 5 e Kinetic parameters for pesticides on CTRTAC. Adsorbate
Methoxychlor Atrazine Methyl parathion
qe (mg/g)
C₀
Pseudo-first-order 1
Pseudo-second-order 2
(experimental)
(mg/L)
qe(mg/g) (theoretical)
k₁ (min )
R
112.0 41.1 104.9 36.8 88.9 28.8
12 5 12 5 12 5
117.5 40.93 111.9 35.48 87.9 27.8
0.0484 0.0815 0.0609 0.0932 0.0605 0.930
0.99 0.991 0.994 0.993 0.991 0.993
qe (mg/g) (theoretical)
k₂ (gmg1min1)
R2
125.0 55.56 111.1 45.45 45.45 40.00
0.0003 0.0012 0.0008 0.0007 0.0033 0.0007
0.976 0.984 0.943 0.802 0.978 0.922
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 4 7 e4 0 5 5
4053
Fig. 6 e Lagergren’s Pseudo 1st order plot of pesticides on CTRTAC at 12 mg/L initial concentration. [average value of 3 tests, std deviation <0.14%]. Fig. 8 e Effect of flow rate on removal of pesticides on CTRTAC.
3.7.
Thermodynamic study
The free energy change (DG ), enthalpy change (DH ), and entropy change (DS ) were calculated using equations described in our earlier work (Gupta et al., 2011). The values of these parameters summarized in Table 4 indicate the sensitivity of the adsorption process toward temperature. The enthalpy change, DH , is negative (exothermic) due to the decrease in adsorption with successive increase in temperature. Negative DG values dictate spontaneous process, revealing increased randomness at the solidesolution interface during the fixation of the pesticides on the active sites of the adsorbent. Since the adsorption process is exothermic, it follows that under these conditions the process becomes spontaneous because of the negative entropy change.
3.8.
Adsorption kinetics
The Lagergren’s pseudo-first-order (Ho and McKay, 1999) and pseudo-second-order (McKay and Ho, 1999) models were used to test adsorption kinetics data in order to investigate the mechanism of adsorption and were analyzed as discussed earlier (Gupta et al., 2011). The values of the kinetic parameters as well as the regression coefficients (R2) at two different concentrations of the pesticides are listed in Table 5. The value of correlation coefficient R2 for the pseudo-firstorder adsorption model is relatively high (>0.99), and the
Fig. 7 e Bangham plot of pesticides on CTRTAC at 12 mg/L initial concentration.
adsorption capacities calculated by the model are also close to those determined by experiments. However, the values of R2 for the pseudo-second-order model are not satisfactory. Therefore, it has been concluded that the Lagergren’s pseudofirst-order adsorption model is more suitable to describe the adsorption kinetics of the pesticides over CTRTAC. Fig. 6 indicates the linear plots of log (qe-qt) vs time for pesticides adsorption on CTRTAC showing the applicability of the Lagergren’s pseudo-first-order model. Lower first-order rate constants for methoxychlor may be due to its larger size, which would not facilitate its approach through pores to adsorbent surface. As the sizes of atrazine and methyl parathion are nearly same, their rate constants should also be similar, and the experimental rate constants are found nearly same as evident from Table 5. The applicability of the following Bangham’s equation (Ho and McKay, 1998) was tested to learn about the slow step occurring in the present adsorption system: C00 k0 m ¼ log þ alog t C0 qtm 2:303V
log log
(2)
where C0 is initial concentration of adsorbate (mg/L), V the volume of solution (mL), m the weight of adsorbent used per
Fig. 9 e Breakthrough curves of pesticides on CTRTAC.
4054
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 4 7 e4 0 5 5
Table 6 e Comparison of batch and column capacities and degree of column utilization. Pesticide
Initial concentration mg/L
Batch capacity (from adsorption experiment) mg/g
12 12 12
112.00 104.90 88.90
Methoxychlor Atrazine Methyl parathion
liter of solution (g L1), qt (mg g1) the amount of adsorbate retained at time t and a (<1) and k0 are constants. log log [C0/ (C0 qtm)] values were plotted against log t in Bangham’s plot (Fig. 7). The linearity of these plots confirms the applicability of Bangham’s equation and indicates that diffusion of pesticide molecules into pores of the adsorbent mainly controls the adsorption process (Serre and Silcox, 2000). The micropores provides majority of the active sites for pesticide adsorption while mesopores act as transportation routes (Hu et al., 2003). The values of Ea (energy of activation) are calculated from the following equation: log K2 =K1 ¼ Ea =2:303 RðT2 T1 Þ=T2 T1
(3)
where K is the 1st order rate constant, T is the temperature in Kelvin and R is the gas constant. The value of Ea calculated from the slope (slope ¼ Ea/ 2.303R) of the graph plotted between log K vs 1/T (plot is not given) was 2.05, 2.71 and 3.84 kJ/mol respectively. Thus, it is confirmed that the nature of interaction between the adsorbate and adsorbent is physical in nature involving weak Van der Waals forces.
3.9.
Fixed bed adsorption studies
The application of the results obtained from adsorption batch experiments were used to remove the pesticidesmethoxychlor, atrazine and methyl parathion of fixed concentration of 12 mg/L by column experiments. The flow rate was varied to achieve maximum removal of pesticides. It was observed that the maximum removal of the pesticides (91% for methoxychlor, 82% for atrazine and 71% for methyl parathion) was achieved at the flow rate of 1.5 ml/min (Fig. 8). The breakthrough curves (Fig. 9) were used to calculate the column capacity (breakthrough capacity, exhaustion capacity and column utilization) and the results are tabulated in Table 6. Relatively lower adsorption capacity than batch capacity has been observed, which is generally attributed to relatively less time of interaction between the adsorbent and the adsorbate surface with column method.
3.10.
Cost evaluation
A distinct advantage of using waste rubber tire as adsorbent is its lower cost and its economic feasibility in comparison to commercial carbon available in India (US $285.0 per ton) (Gupta et al., 2006a). The waste rubber tire granules costs US$
Column capacity Breakthrough capacity mg/g
Exhaustion capacity mg/g
Degree of column utilization (%)
95.54 81.62 59.13
104.87 99.41 82.38
91.1 82.1 71.78
10.0 per ton and considering the cost of transport, chemicals, electrical energy used in the process and labor, the finished product would cost approximately US$ 25.0 per ton.
4.
Conclusion
This study has investigated the adsorption of pesticides onto activated carbon developed from thermal and chemical treatment of waste rubber tire. Chemical treatment prior to activation facilitated in the development of not only a well developed porous adsorbent but also incorporation of carbonyl and hydroxyl functional groups onto the adsorbent surface; these ultimately helped in the adsorption process. Batch adsorption studies revealed maximum adsorption occurring at a contact time of 60 min, particle adsorbent size of 200e250 mm, low pH and at a lower temperature. CTRTAC offered admirable adsorption capacities for the removal of methoxychlor, atrazine and methyl parathion which were 112.0 mg g1, 104.9 mg g1 and 88.9 mg g1 respectively from an initial concentration of 12mg/L signifying 93.3%, 87.4% and 74.1% removal of pesticides. These promising results were confirmed with the removal of 91%, 82.1% and 71.78% methoxychlor, atrazine and methyl parathion respectively by column experiments. Both textural as well as surface chemistry of the developed adsorbent was conducive for the enhanced adsorption capacity of bulky adsorbates like the pesticides. The decreasing trend of the adsorptive capacity of the CTRTAC is indicative of the influence of the pesticide solubility. The developed system is feasible and spontaneous as revealed from thermodynamic studies. Kinetics study indicates that the rate determining step of the adsorption process is pore diffusion of pesticides onto the adsorbent. Finally, a cost analysis study of CTRTAC and commercial activated concludes that CTRTAC is a more effective, efficient, economic adsorbent and can be used for the removal of reported and other pesticides from waste water.
Acknowledgment The authors acknowledge the financial support provided by KFUPM through NSTIP Project # 10-WAT1400-04. Arunima Nayak is thankful to Ministry of Human Resource Development. (MHRD), New Delhi, India, for providing fellowship.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 4 7 e4 0 5 5
references
Aharoni, C., Ungarish, M.J., 1977. Assessment of environmental and solution parameter impact on trace-metal sorption by soils. Chem. Soc. Faraday Trans. 73, 456e464. Ali, I., Gupta, V.K., 2007. Advances in water treatment by adsorption technology. Nat. Protoc. 1, 2661e2667. Ariyadejwanich, P., Tanthapanichakoon, W., Nakagawa, K., Mukai, S.R., Tamon, H., 2003. Preparation and characterization of mesoporous activated carbon from waste tires. Carbon 41, 157e164. Bedient, P.H., Rifai, H.S., Newell, C.J., 1994. Ground Water Contamination: Transport and Remediation. Prentice Hall, Englewood Cliffs, NJ. Betancur, M., Martinez, J.D., Murillo, R., 2009. Production of activated carbon by waste tire thermochemical degradation with CO2. J. Hazard. Mater. 168, 882e887. Castells, X.E., 2000. Reciclaje de Residuos Industriales, vol. 495. Diaz Santos, Madrid. Cunliffe, A., Williams, P.T., 1999. Influence of process conditions on the rate of activation of chars derived from pyrolysis of used tires. Energy Fuel 13, 166e175. Gonzalez, J.F., Encinar, J.M., Gonzalez-Garcia, C.M., Sabio, E., Ramiro, A., Canito, J.L., Gǎnan, J., 2006. Preparation of activated carbons from used tyres by gasification with steam and carbon dioxide. Appl. Surf. Sci. 252, 5999e6004. Gupta, V.K., Imran, A., 2001. Removal of DDD and DDE from wastewater using baggasse fly ashda sugar industry waste. Water Res. 35, 33e40. Gupta, V.K., Jain, C.K., Imran, A., Chandra, S., 2002. Removal of lindane and malathion from waste water using baggasse fly ash: a sugar industry waste. Water Res. 3, 2483e2490. Gupta, V.K., Imran, A., Suhas, Saini, V.K., 2006a. Adsorption of 2,4-D and carbofuran pesticides using fertilizer and steel industry wastes. J. Colloid Interface Sci. 299, 556e563. Gupta, V.K., Mittal, A., Jain, R., Mathur, M., Sikarwar, S., 2006b. Adsorption of Safranin-T from wastewater using waste materials e activated carbon and activated rice husk. J. Colloid Interface Sci. 303, 80e86. Gupta, V.K., Imran, A., 2008. Removal of endosulfan and methoxychlor from water on carbon slurry. Environ. Sci. Technol. 42, 766e770. Gupta, V.K., Mittal, A., Gajbe, V., Mittal, J., 2008. Adsorption of basic fuschin using waste materials-bottom ash and de-oiled soya as adsorbents. J. Colloid Interface Sci. 319, 30e39. Gupta, V.K., Carrott, P.J.M., Ribeiro Carrott, M.M.L., Suhas, 2009. Low cost adsorbents: growing approach to wastewater treatment e a review. Crit. Rev. Environ. Sci. Technol. 39, 783e842. Gupta, V.K., Gupta, B., Rastogi, A., Agarwal, S., Nayak, A., 2011. A comparative investigation on adsorption performances of mesoporous activated carbon prepared from waste rubber tire and activated carbon for an Azo dye. J. Hazard. Mater. 186, 891e901. Hamadi, N.K., Swaminathan, S., Chen, X.D., 2004. Adsorption of Paraquat dichloride from aqueous solutions by activated carbon derived from used tires. J. Hazard. Mater. 112, 133e141. Helleur, R., Popovic, N., Ikura, M., Stanciulescu, M., Liu, D., 2001. Characterization and potential applications of pyrolytic char from ablative pyrolysis of used tires. J. Anal. Appl. Pyrol. 58e59, 813e824. Hsieh, C., Teng, H., 2000. Influence of mesopore volume and adsorbate size on adsorption capacities of activated carbons in aqueous solutions. Carbon 38, 863e869.
4055
Ho, Y.S., McKay, G., 1998. Sorption of dye from aquous solution by peat. J. Chem. Eng. 70, 115e124. Ho, Y.S., McKay, G., 1999. The sorption of lead (II) ions on peat response to comment. Water Res. 33, 578e584. Hu, Z., Guo, H., Srinivasan, P.M., Yaming, N., 2003. A simple method for developing mesoporosity in activated carbon. Separat. Purif. Technol. 31, 47e52. Knocke, W.R., Hemphill, L.H., 1981. Mercury(II) sorption by waste rubber. Water Res. 15, 275e282. Lin, Y.R., Teng, H., 2002. Mesoporous carbons from waste tire char and their application in wastewater discoloration. Microporous Mesoporous Mater. 54, 167e174. Lopez, G., Olazar, M., Artetxe, M., Amutio, M., Elordi, G., Bilbao, J., 2009. Steam activation of pyrolytic tyre char at different temperatures. J. Anal. Appl. Pyrol. 85, 539e543. Mechant, A.A., Petrich, M.A., 1993. Pyrolysis of scrap tire and conversion of chars to activated carbon. AIChE J. 39, 1370e1376. McKay, G., Ho, Y.S., 1999. Pseudo-second-order model for sorption processes. Process Biochem. 34, 451e465. Miguel, G.S., Fowler, G.D., Sollars, C.J., 2003. A study of the characteristics of activated carbons produced by steam and carbon dioxide activation of waste tyre rubber. Carbon 41, 1009e1016. Murillo, R., Aylon, E., Navarro, M.V., Callen, M.S., Aranda, A., Mastral, A.M., 2006. The application of thermal processes to valorize waste tyre. Fuel Process. Technol. 87, 143e147. Mui, E.L.K., Ko, D.C.K., Mckay, G., 2004. Production of active carbon from waste tyresda review. Carbon 42, 2789e2805. Mui, E.L.K., Cheung, W.H., McKay, G., 2010. Tyre char preparation from waste tyre rubber for dye removal from effluents. J. Hazard. Mater. 175, 151e158. Pal, S., Kyeong-Hee, L., Jong-Uk, K., Seung-Hee, H., Song, J.M., 2006. Adsorption of cyanuric acid on activated carbon from aqueous solution: effect of carbon surface modification and thermodynamic characteristics. J. Colloid Interface Sci. 303, 39e48. Park, Jin-Soo, Jung, Yung-Woo, 2002. Effect of KOH activation on the formation of oxygen structure in activated carbons synthesized from polymeric precursor. J. Colloid Interface Sci. 250, 93e98. Raghuvanshi, S.P., Singh, R., Kaushik, C.P., Raghav, A.K., 2004. Kinetics study of methylene blue dye bioadsorption on baggase. Int. J. Appl. Ecol. Environ. Res. 2, 35e43. Serre, S.D., Silcox, G.D., 2000. Adsorption of elemental mercury on the residual carbon in coal fly ash. Ind. Eng. Chem. Res. 39, 1723e1730. Suuberg, E.M., Aarna, I., 2007. Porosity development in carbon derived from scrap automobile tires. Carbon 4, 1719e1726. Tanthapanichakoon, W., Ariyadejwanich, P., Japthong, P., Nakagawa, K., Mukai, S.R., Tamon, H., 2005. Adsorptionedesorption characteristics of phenol and reactive dyes from aqueous solution on mesoporous activated carbon prepared from waste tires. Water Res. 39, 347e355. Teng, H., Serio, M.A., Wojtowicz, M.A., Bassilakis, B., Solomon, P. R., 1995. Reprocessing of used tires into activated carbon and other products. Ind. Eng. Chem. Res. 34, 3102e3111. USEPA, http://www.epa.gov/pesticides/factsheets/index.htm. Williams, P.T., Besler, S., Taylor, D.T., 1990. The pyrolysis of scarp automotive tyres. Fuel 69, 1474e1482. Zabaniotou, A., Madau, P., Oudenne, P.D., Jung, C.G., Delplancke, M.P., Fontana, A., 2004. Active carbon production from used tire in two-stage procedure: industrial pyrolysis and bench scale activation with H2OeCO2 mixture. J. Anal. Appl. Pyrol. 72, 289e297.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 5 6 e4 0 6 8
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Removal of bisphenol A and 17a-ethinyl estradiol from landfill leachate using single-walled carbon nanotubes Lesley Joseph a, Qammer Zaib a, Iftheker A. Khan a, Nicole D. Berge a, Yong-Gyun Park b, Navid B. Saleh a, Yeomin Yoon a,* a
Department of Civil and Environmental Engineering, University of South Carolina, Columbia, SC 29201, USA Environmental & Energy Research Team, GS E&C Research Institute, 417-1 Deokseong-ri, Idong-myeon, Cheoin-gu, Yongin-si, Gyeonggi-do, 449-831, Republic of Korea b
article info
abstract
Article history:
In this study, the adsorption of bisphenol A (BPA) and 17a-ethinyl estradiol (EE2) from
Received 16 January 2011
landfill leachate onto single-walled carbon nanotubes (SWCNTs) was investigated.
Received in revised form
Different leachate solutions were prepared by altering the pH, ionic strength, and dissolved
14 May 2011
organic carbon (DOC) in the solutions to mimic the varying water conditions that occur in
Accepted 17 May 2011
leachate during the various stages of waste decomposition. The youngest and oldest
Available online 26 May 2011
leachate solutions contained varying DOC and background chemistry and were repre-
Keywords:
[Ca2þ] ¼ 1200 mg/L; [Mg2þ] ¼ 470 mg/L) and Type E (pH ¼ 7.5; DOC ¼ 250 mg/L;
Single-walled carbon nanotubes
conductivity ¼ 3250 mS/cm; [Ca2þ] ¼ 60 mg/L; [Mg2þ] ¼ 180 mg/L). These solutions were
Landfill leachate
subsequently combined in different ratios to produce intermediate solutions, labeled BeD,
Bisphenol A
to replicate time-dependent changes in leachate composition. Overall, a larger fraction of
sented by leachate Type A (pH ¼ 5.0; DOC ¼ 2500 mg/L; conductivity ¼ 12,500 mS/cm;
17a-Ethinyl estradiol
EE2 was removed as compared to BPA, consistent with its higher log KOW value. The total
Adsorption
removal of BPA and EE2 decreased in older leachate solutions, with the adsorptive capacity of SWCNTs decreasing in the order of leachate Type A > Type B > Type C > Type D > Type E. An increase in the pH from 3.5 to 11 decreased the adsorption of BPA by 22% in young leachate and by 10% in old leachate. The changes in pH did not affect the adsorption of EE2 in the young leachate, but did reduce adsorption by 32% in the old leachate. Adjusting the ionic strength using Naþ did not significantly impact adsorption, while increasing the concentration of Ca2þ resulted in a 12% increase in the adsorption of BPA and a 19% increase in the adsorption of EE2. DOC was revealed to be the most influential parameter in this study. In the presence of hydrophilic DOC, represented by glucose in this study, adsorption of the endocrine disrupting compounds (EDCs) onto the SWCNTs was not affected. In the absence of SWCNTs, hydrophobic DOC (i.e., humic acid) adsorbed 15e20% of BPA and EE2. However, when the humic acid and SWCNTs were both present, the overall adsorptive capacity of the SWCNTs was reduced. Hydrophobic (p-p electron donoracceptor) interactions between the EDCs and the constituents in the leachate, as well as interactions between the SWCNTs and the EDCs, are proposed as potential adsorption mechanisms for BPA and EE2 onto SWCNTs. ª 2011 Elsevier Ltd. All rights reserved.
* Corresponding author. Tel.: þ1 803 777 8952; fax: þ1 803 777 0670 E-mail address:
[email protected] (Y. Yoon). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.05.015
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 5 6 e4 0 6 8
1.
Introduction
Recently, the presence of endocrine disrupting compounds (EDCs) in drinking water has emerged as a serious issue worldwide. EDCs are compounds that can block or mimic the activity of natural hormones, thus interfering with the reproductive systems of wildlife and humans (Harrison et al., 1997; Snyder et al., 2003). The reported effects of human exposure to EDCs include reduced fertility and an increase in the incidence of breast, ovarian, and testicular cancer (Colborn et al., 1993; Harrison et al., 1997; Cheek et al., 1998). The majority of EDCs are synthetic pharmaceuticals that are excreted from humans at low concentrations and discharged into waterways and wastewater streams (Daughton and Ternes, 1999; Kolpin et al., 2002; Kasprzyk-Hordern et al., 2008). While municipal wastewater has been cited as the primary source of EDC emission into the environment (DeRudder et al., 2004), researchers have identified a wide variety of EDCs in landfill leachate. The sources of these include plastics, pharmaceuticals, and solid waste incineration residue, with reported concentrations of EDCs ranging from <0.001 mg/L to 17,200 mg/L (Holm et al., 1995; Shiraishi et al., 1999; Behnisch et al., 2001; Yamamoto et al., 2001; Wintgens et al., 2003; Asakura et al., 2004). Landfill leachate is generally characterized by its high concentration of organic compounds, and mainly consists of dissolved organic matter, inorganic compounds (e.g., Ca2þ, Mg2þ, Naþ), trace metals, and xenobiotic organic compounds (XOCs). The concentrations of each component in the leachate depend on the age and composition of the waste in the landfill (Barlaz and Ham, 1993; Reinhart and Townsend, 1998; Kjeldsen et al., 2002). The treatment of landfill leachate has been investigated using a variety of techniques, including biological (e.g., nitrification/ denitrification, activated sludge), chemical (e.g., advanced oxidation), and physical (e.g., membrane filtration, adsorption) processes. These processes have been successfully used to remove significant quantities of various contaminants, including heavy metals (Linde and Jonsson, 1995; Urase et al., 1997), nitrogen species (Peters, 1998; Aziz et al., 2004), and humic substances (Rodriguez et al., 2004). Many studies have also found these treatment methods to be effective in the removal of EDCs from contaminated drinking water, river water, and wastewater effluents (Ternes et al., 2002; Yoon et al., 2003, 2006; Alum et al., 2004; Westerhoff et al., 2005); however, little research has been conducted to determine the ability of these processes to remove target contaminants, particularly EDCs, from leachate. One advantage of using physical processes, particularly adsorption, is the ability to remove a wide variety of dissolved organic and inorganic contaminants (Rivas et al., 2006; Lim et al., 2009). Activated carbon is a commonly used adsorbent and has been used to treat landfill leachate (Foo and Hameed, 2009). The advantages of using activated carbon are its large surface area, pore structure, and thermostability; these characteristics improve its ability to remove organic contaminants from various aqueous media (Chingombe et al., 2005). Although activated carbon is a well-established adsorbent that is often used for water and wastewater treatment, alternatives, including carbon nanotubes (CNTs) have been
4057
investigated to determine their relative sorptive properties. Studies indicate that CNTs are better adsorbents than activated carbon because of their higher adsorption capacities for heavy metals (Li et al., 2005; Wang et al., 2005; Lu and Chiu, 2006), phenols (Chen et al., 2008b; Lin and Xing, 2008; Yang et al., 2008), and other organic chemicals (Lu et al., 2005; Yang et al., 2006). Another study showed that SWCNTs were more effective that activated carbon at removing EDCs at varying concentrations (Pan et al., 2010). More recently, researchers have examined the adsorptive capacities and mechanisms of various EDCs onto CNTs and determined that p-p electron donor-acceptor (EDA) and hydrophobic interactions are important mechanisms for the adsorption of EDCs onto CNTs, based on the KOW (octanol-water partitioning coefficient) values and the KHW (hexadecane-water partition coefficient) values of different EDCs (Kuo, 2009; Oleszczuk et al., 2009; Pan and Xing, 2010). Of the known EDCs, bisphenol A (BPA) and 17a-ethinyl estradiol (EE2) have been the most frequently researched with respect to water treatment (Pan et al., 2008). BPA is used in the production of polycarbonate plastics and epoxy resins (Staples et al., 1998). Studies have shown that significant amounts of BPA can leach from plastics and enter the environment, creating a public health concern (Lopez-Cervantes and Paseiro-Losada, 2003; Nerin et al., 2003). Researchers have detected BPA in landfill leachates at median concentrations of 0.07e269 mg/L (Yamamoto et al., 2001; Asakura et al., 2004). EE2 is a synthetic estrogen found in birth control pills; its toxicity is 10e50 times higher than that of estrone and 17bestradiol, two other common EDCs (Segner et al., 2003). EE2 has also been detected in landfill leachates, with concentrations of up to 5 ng/L (Behnisch et al., 2001). BPA and EE2 are compounds with wide-ranging chemical properties and are often used as representatives of EDCs when conducting water treatment research (Pan et al., 2008). However, studies focused on the removal of EDCs have only been conducted in the context of drinking water and wastewater treatment. The removal of EDCs from landfill leachate presents a new challenge due to the broad range of water chemistry characteristics that are present, including differing pH levels, increased variation of types of DOC, and increased amounts of inorganic compounds (e.g., Ca2þ, Mg2þ, Naþ). These characteristics could potentially alter the physicochemical properties of CNTs (Saleh et al., 2010), thus having a significant effect on the adsorption behavior of EDCs onto CNTs. The objective of this study is to investigate the adsorption of BPA and EE2 on single-walled carbon nanotubes (SWCNTs) under landfill leachate chemistry. With the inherent variability in the composition of landfill leachates, the interactions between different components of the leachate and the SWCNTs must be examined, along with the influence of varying chemical compositions (i.e., pH, electrolyte species, DOC, and ionic strength) on the adsorption of BPA and EE2 to SWCNT surfaces. In this study, the sorption mechanism is described, along with detailed characterization of the SWCNTs. Electron microscopy, time-resolved dynamic light scattering (DLS), and angle-dependent static light scattering (SLS) are performed to obtain insight into SWCNT cluster size, morphology, and aggregate structure in the presence of specific organic matter (i.e., humic acid and glucose).
4058
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 5 6 e4 0 6 8
Table 1 e Characteristics of bisphenol A and 17a-ethinyl estradiol. Common name [Full name]
Bisphenol A [2,20 -bis-(4-hydroxyphenyl) propane]
17a-ethinylestradiol [17a-ethinyl-1,3,5(10)-estratriene-3,17a-diol].
2.
Materials and methods
2.1.
Materials
Use
Molar Mass (g/mol)
Log KOW
pKa
Plasticizer
228.1
3.3
9.6e10.2
Ovulation inhibitor.
296.2
3.7
w10.5
BPA (purity > 99%), EE2 (purity > 98%), and humic acid were purchased from Sigma-Aldrich (St. Louis, MO, USA). Table 1 shows the physicochemical properties of BPA and EE2. BPA and EE2 were dissolved separately in methanol to produce stock solutions of 1 mM each. Humic acid stock solution was prepared by adding 1 g of dry humic acid powder to 1 L of ultrapure deionized (DI) water and stirring overnight. The solution was then filtered through 0.7 mm glass microfiber filters (Whatman, Buckinghamshire, UK) to remove particulate matter greater than 0.7 mm. Glucose, CaCl2, and MgCl2 were purchased from Sigma-Aldrich. NaCl was obtained from Fisher-Scientific (Pittsburgh, PA, USA). SWCNTs (purity > 90%) were procured from Cheap Tubes, Inc. (Brattleboro, Vermont, USA) and used without further purification. The SWCNTs have a length of 5e30 mm and an outer diameter of 1e2 nm, as reported by the manufacturer (Cheap Tubes, 2009).
2.2.
Synthetic leachate
To examine the adsorption of the EDCs from landfill leachate, solutions of synthetic leachate were produced. Sufficient volumes of the BPA and EE2 stock solutions to achieve
Structure
concentrations of 10 mM each were placed in separate beakers. To remove the DOC found in the solvent, methanol was evaporated from the stock solution for 1 hour at room temperature under a fume hood. Then, the other components of the leachate were added according to the experimental matrix presented in Table 2 and stirred overnight. Two sets of leachate were made to represent the acetogenic and the methanogenic phases that occur within a landfill. The leachate compositions utilized in this study were determined according to a literature review that accounted for various landfills around the world (Ehrig, 1989; Barlaz and Ham, 1993; Tchobanoglous et al., 1993; Reinhart and Al-Yousfi, 1996; Kjeldsen and Christophersen, 2001; Scott et al., 2005). Ranges of concentrations were adopted for leachate produced in both the acetogenic and methanogenic phases of a landfill. For this study, the synthetic leachate that is described as “young” mimics the composition of leachate in the acetogenic phase, while the synthetic leachate that is described as “old” represents the composition during the methanogenic phase. Glucose (molar mass: 180 g/mol) was used to represent the DOC typically found in leachate produced in the acetogenic phase. Humic acid (molar mass: 2000e3000 g/mol) was used to represent leachate produced in the methanogenic phase. CaCl2 and MgCl2 were used as the sources of the Ca2þ and Mg2þ ions, respectively. The pH of each leachate was adjusted using 1 M HCl and 1 M NaOH and buffered with a 1 mM phosphate buffer solution. The conductivity was measured to
Table 2 e Characteristics of synthetic young and old leachates. Parameters
pH Conductivity (mS/cm) Ca2þ (mg/L) [mM] Mg2þ (mg/L) [mM] DOC (mg/L) BPA (mg/L)b,c EE2 (mg/L)b,c
Young leachate
Old leachate a
Experimental
Range
Experimental
Rangea
5.0 12,500 1200 [300] 470 [193] 2500 2.28 2.96
4.7e7.7 1600e17,100 10e2500 [2.5e624] 50e1150 [20e473] 1500e20,000 0.013e17.2 e
7.5 3250 60 [15] 180 [74] 250 2.28 2.96
7.1e8.8 1400e4500 20e600 [5e150] 40e350 [16e144] 80e160 0.013e17.2 e
a Representative range of values developed from Reinhart and Basel Reinhart and Al-Yousfi (1996); Ehrig (1989); and Tchobanoglous et al. (1993). b Ranges of BPA taken from Yamamoto et al. (2001); no ranges have been reported for EE2, although detection in leachate has occurred (Behnisch et al., 2001). Current studies have not investigated variations in EDC concentrations based on landfill age. c Experimental values of BPA and EE2 are equivalent to 10 mM.
4059
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 5 6 e4 0 6 8
account for the contributions of the Ca2þ and Mg2þ ions, and then adjusted using NaCl to achieve the desired concentration. To account for the changes in waste composition that occur over time, the “young” and “old” leachate solutions were combined in different ratios and labeled Type AeE. Type A corresponds to the youngest leachate, while Type E corresponds to the oldest leachate. The characteristics of these combined solutions are shown in Table 3. A detailed description of the considerations made when determining the composition of the synthetic leachate solutions are discussed later in the article.
2.3.
Adsorption experiments
A stock suspension of 1000 mg/L of SWCNTs was prepared by adding 100 mg of SWCNTs to 100 mL of DI water and mixing over a magnetic stirrer at 500 rpm. The SWCNTs used in this study did not undergo any additional treatment (e.g., sonication, heating, or chemical functionalization), in order to most accurately replicate its application in a commercial water treatment system. In current water treatment applications, adsorbents (e.g., powdered activated carbon) are purchased from the manufacturer and applied directly into the existing flocculation tanks to facilitate the removal of any contaminants. This procedure was adopted for the batch experiments conducted throughout this study. Adsorption kinetics experiments were prepared using 100 mL solutions of young and old leachate in amber bottles with initial concentrations of 10 mM of both BPA and EE2. These concentrations were confirmed by measuring control samples prior to experimentation. The solutions were then spiked with a small volume of SWCNT stock solution to achieve a concentration of 50 mg/L of SWCNTs. At the predetermined time intervals of 0, 5, 10, 15, 20, 25, 30, 60, 120, 180, and 240 min, aliquots of the leachate solutions were withdrawn from the bottles and passed through 0.45 mm Durapore membrane filters (Millipore, Ireland), placed in 2 mL amber vials, and analyzed using high performance liquid chromatography (HPLC); the details of the HPLC analysis are described in the following section. Isotherms were investigated using a batch adsorption technique under ambient conditions. BPA and EE2 stock solutions were diluted to initial concentrations of 10 mM. Background solutions were prepared using the predefined ratios of young and old leachate described in Table 3. Applied
doses of SWCNTs ranged from 20 to 200 mg/L. The dilutions of the SWCNT suspension were achieved by volumetric extractions from the continuously mixed stock suspension of SWCNTs. The samples containing BPA and EE2 were placed in 40 mL screw-cap glass vials, spiked with the SWCNTs, and sealed with Teflon screw-caps. The vials were then placed in a shaker for 4 h at 13.9 rpm. While adsorption experiments are generally conducted for 7 days to reach equilibrium, kinetic experiments confirmed that 4 h is sufficient time for the solution to reach a pseudo-equilibrium state of sorption. Each sample was then filtered through a 0.45 mm membrane filter and transferred to 2 mL amber vials for HPLC analysis. For the purposes of this study, 0.45 mm filters were used to remove the SWCNTs from the treated leachate samples. In a full-scale leachate treatment facility, rapid filtration can be used to remove the SWCNTs from the treated leachate. All of the experiments conducted in this study were under abiotic conditions. Duplicates of the experiments were conducted to ensure the reproducibility of the results; the error bars represent one standard deviation of the replicates. The amount of BPA and EE2 adsorbed onto the SWCNTs was calculated using the following equation: q ¼ ðC0 Ct Þ
V m
(1)
where q is the amount of BPA or EE2 adsorbed onto SWCNTs (mg/g), C0 and Ct are the concentrations at the beginning and end of a time period (mg/L), V is the volume of the initial solution (L), and m is the mass of SWCNTs (g).
2.4.
Analytical methods
The concentrations of BPA and EE2 were quantified using an HPLC 1200 Series (Agilent Technologies, Santa Clara, CA). BPA and EE2 were detected using a fluorescence detector at an excitation wavelength of 280 nm and an emission wavelength of 310 nm. A Waters 5-mm LiChrosorb RP18 analytical column (4.6 mm 100 mm) with a 100-mL sample loop was used for the reverse-phase separations. The mobile-phase solvent profile was 45% DI water, acidified with 10 mM H3PO4, and 55% MeOH at a constant flow rate of 1 mL/min for 30 min. The detection limits were 0.88 nM (201 ng/L) for BPA and 0.96 nM (283 ng/L) for EE2. BPA and EE2 eluted from the columns at 9.4 and 20.3 min, respectively. DOC was measured by a TOC-VCSN analyzer (Shimadzu Scientific Instruments, Columbia, MD, USA).
Table 3 e Characteristics of each leachate combination. Parameters
Types of leachate A
Young and old percentages (Y:O) pH Conductivity (mS/cm) Ca2þ (mg/L)a Mg2þ (mg/L)a DOC (mg/L) [Y:O]b
100:0 5.0 12,500 1200 470 2500:0 [2653]
B 75:25 6.17 11,030 915 398 1875:63 [2028]
C
D 50:50
6.68 8090 630 325 1250:125 [1364]
25:75 7.06 5960 345 253 625:188 [889]
E 0:100 7.5 3250 60 180 0:250 [293]
a Values calculated using ratios of existing young and old leachate concentrations. b Ratios represent the theoretical contribution of DOC from young and old leachate, respectively. Measured DOC values are in brackets.
4060
2.5.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 5 6 e4 0 6 8
Characterization of SWCNTs
High resolution transmission electron microscopy (HRTEM) was performed using dry SWCNT samples. SWCNTs were dispersed in ethyl alcohol (JT Baker, ACS Grade) and sonicated for 10 min using a sonic dismembrator (S-4000, Misonix). Suspensions were subsequently placed in a bath sonicator (Bransonic-12) for 5 min before the imaging was performed. A drop of the suspension was placed on a 200-mesh copper TEM grid coated with an amorphous carbon holey film (SPI Supplies) and allowed to dry for 2 min. Excess solvent was removed using filter paper. Images were collected using a JEOL JEM-2100F 200 kV Schottky field emission gun HRTEM (JEOL Ltd., Tokyo, Japan) with a point resolution of 0.19 nm. The shape and microstructure of the SWCNTs were studied via transmission electron microscopy (TEM) using a Hitachi H-8000 with an accelerating voltage of 200 kV that was fitted with a charge-coupled device camera. Samples were prepared by dispersing SWCNTs in deionized water by mild ultrasonication and depositing them on a 200mesh nickel grid coated with formvar. A ZetaPALS analyzer (Brookhaven, USA) was used to measure the zeta potential of the SWCNTs. The pH values of the SWCNT solutions were adjusted to between 3.5 and 11.0 by adding 1 M NaOH or 1 M HCl solution. A LabRam JY Horiba Raman spectrometer fitted with a confocal microscope, a thermoelectrically-cooled charge-coupled device, and a 632.8 nm HeeNe laser for excitation was used to obtain Raman spectra of powdered samples of the SWCNTs. The integration time for each scan was 15 s, with an average of 5 scans represented by each spectrum. The average hydrodynamic radii of the SWCNT clusters and the cluster size distribution were obtained using a robust, DLS/SLS instrument (ALV/CGS-3, Langen, Germany) equipped with a 22 mW HeeNe laser at 632 nm (equivalent to an 800 mW laser at 532 nm) and a highly sensitive high QE APD detector with photomultipliers. The average cluster size of a SWCNT suspension (1 mg/L) in the presence of the two organic backgrounds (i.e., humic acid and glucose) was measured at a 90 scattering angle. A SWCNT size distribution analysis was performed using a CONTIN algorithm to determine the relative differences in the sizes of the SWCNT clusters in the presence of the different organic matter. Insight into the aggregate structure of the SWCNT clusters was obtained through angle-dependent SLS measurements performed using the precision DLS/SLS ALV/CGS-3 instrument. An angular range of 12 e150 was used to obtain scattering intensity data for each SWCNT sample. The linear fractal range of scattering (30 e80 ) was established by plotting the scattering parameter, q (where q ¼ 4pr=l0 sin ðq=2Þ,) vs. intensity, I; a linear regression fit revealed the fractal dimensions (Df) of the SWCNT clusters.
2.6.
Isotherm modeling
The adsorption data obtained in the experiments was fitted to two different isotherm models: Freundlich model : qe ¼ Kf Ce1=n
(2)
abCe 1 þ bCe
(3)
Langmuir model : qe ¼
where qe is the solid-phase concentration (mg/g), Ce is the equilibrium solution phase concentration (mg/L), Kf is the Freundlich affinity coefficient [(mg/g)/(mg/L)(1/n)], n is the dimensionless number related to surface heterogeneity, a is the maximum adsorption capacity (mg/g), and b is the Langmuir fitting parameter (L/mg).
3.
Results and discussion
3.1.
Determination of various synthetic leachates
In municipal solid waste landfills, large amounts of leachate are generated by the passage of water through the solid waste. During this process, various contaminants can leach from the waste, and appear in the leachate. According to Kjeldsen et al. (2002), these contaminants are typically separated into four categories: (i) total organic matter, often quantified as DOC; (ii) inorganic components (e.g., Ca2þ, Mg2þ, Naþ); (iii) heavy metals; and (iv) XOCs. As the waste in the landfill decomposes through the five stages of stabilization, the overall characteristics of the leachate change. For instance, the DOC in leachate can include a wide variety of organic components ranging from volatile fatty acids (VFAs) to refractory humic and fulvic-like compounds, depending on the stage of degradation in the landfill (Chian and Dewalle, 1977). Several leachate composition studies have shown that the DOC found in leachate produced during the acetogenic phase is primarily made up of VFAs and simple organics, while the DOC found in the methanogenic phase consists of complex organics, which are mainly humic-like materials (Artiola-Fortuny and Fuller, 1982; Harmsen, 1983; Frimmel and Weis, 1991; Christensen et al., 1998). The inorganic components of the leachate can also change throughout the landfill stabilization process. The concentrations of many inorganics (e.g., Ca2þ and Mg2þ) can decrease over time due to changes in the pH and lower levels of total organic carbon (TOC) in the leachate (Kjeldsen et al., 2002). During later stages in the decomposition of the landfill, these cations can potentially form complexes with the humic and fulvic acids that are present in old leachate, which can also reduce their concentrations (Krug and Ham, 1991). The water chemistry of the leachate (i.e., pH and ionic strength) can also vary significantly over the life of a landfill. In the early stages of leachate production, low pH values have been observed, which is primarily due to high concentrations of easily degradable compounds such as VFAs and simple organics (Kjeldsen et al., 2002). However, as the waste stabilizes, the pH increases as the organic carbon present in the landfill becomes less degradable (Ehrig, 1988). Ionic strength is mainly determined by the concentrations of cations such as Ca2þ, Mg2þ, Fe2þ, Naþ, and Mn2þ at any given stage in its production. The ionic strength of leachate is high during the acetogenic phase and can change as the landfill ages, due to the higher pH in the methanogenic phase, which contributes to the precipitation of these cations and enhances their sorption to various components in the waste (Scott et al., 2005). The parameters for the synthetic leachate produced for this study were chosen based on their potential influence on the adsorption of the EDCs by SWCNTs. The concentration
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 5 6 e4 0 6 8
3.2.
Characterization of SWCNTs
The average hydrodynamic radii of the SWCNT clusters (obtained by averaging the radii values collected for 25 min for each condition), as shown in Fig. 1 (and Fig. S1), indicate a smaller cluster size for SWCNTs in the presence of humic acid (100e130 nm) as compared to those in the presence of glucose (150e290 nm). The average radii for both the organic matter background cases did not change significantly with an increase in the concentration of the organics (i.e., 10, 50, and 100 mg/L DOC). The difference in the hydrodynamic radii of the SWCNT clusters depending on the type of DOC is further demonstrated by analyzing the scattering data to obtain relative cluster distributions, as presented in Fig. S2. The
300 Average hydrodynamic radius (nm)
and the complexity of the organic carbon found in the leachate may possibly affect the overall adsorption of the EDCs. This was addressed in this study by using glucose to represent the simple organic carbon found in young leachate, and humic acid to represent the more complex organic carbon in old leachate. The decrease in the concentrations of Ca2þ, Mg2þ, and Naþ over time due to the changes in pH and lower TOC can alter the ionic strength of the leachate and potentially affect the behavior of the EDCs and the overall adsorption (Kjeldsen et al., 2002). Therefore, Ca2þ, Mg2þ, and Naþ were added as inorganic components in the synthetic leachate samples. Heavy metals have also been detected in leachate at low concentrations (Christensen et al., 1994; Kjeldsen and Christophersen, 2001), but they are not believed to significantly influence the adsorption of EDCs by SWCNTs. Therefore, no heavy metals were incorporated into this study. XOCs are represented by BPA and EE2, which are the target compounds in this study. The XOCs that are most prevalent in leachates are aromatic hydrocarbons and halogenated hydrocarbons (Kjeldsen et al., 2002). Many studies have shown that these hydrocarbons can be adsorbed by CNTs (Peng et al., 2003; Gotovac et al., 2007; Chen et al., 2008a; Lin and Xing, 2008), which suggests that competition may occur between the contaminants for adsorption sites. However, this study will focus solely on the adsorption of BPA and EE2. Throughout this study, the leachate being examined is classified as “young” or “old” based on its production during the acetogenic and methanogenic phases of the landfill, respectively. However, this is a simplistic characterization. In reality, landfill leachate undergoes a gradual transformation over several decades, beginning as young leachate with a large amount of organics, low pH, and high conductivity to more stable leachate with decreased chemical oxygen demand, low conductivity, and high pH (El-Fadel et al., 1997). These changes in the composition of the leachate can alter the adsorptive capabilities of the SWCNTs and potentially complicate the removal of target micropollutants, due to the competitive nature of the DOC and the behavior of the SWCNTs under various water conditions. As shown in Table 3, leachate Types AeE represent the varying combinations of young and old leachate solutions to more accurately represent the changes in the leachate during the life of the landfill. Type A corresponds to the youngest leachate, while Type E corresponds to the oldest leachate.
4061
250 200
150 100 50 0 10 mg/L 50 mg/L 100 mg/L Concentration of organic matter (mg/L)
Fig. 1 e Average hydrodynamic radii of SWCNTs in presence of glucose and humic acid for a range of DOC concentrations (10e100 mg/L). White bars represent humic acid as the organic carbon source and gray bars represent glucose. Error bars represent one standard deviation of the average radii. The measurements were performed at 22 ± 1 C.
smaller cluster sizes of the SWCNTs in the presence of humic acid direct toward a higher dispersed state of the SWCNTs due to the possible adsorption of the hydrophobic DOC on the surface of the SWCNTs. The larger sizes of the SWCNT clusters in the presence of hydrophilic glucose demonstrate a more bundled state of SWCNTs, which may have resulted from the lower surface coverage of the sorbed hydrophilic organics. This has direct implications on the sorption behavior of the EE2 and BPA to the SWCNTs, as will be described later. Further insight into the cluster morphology of the SWCNTs was obtained from the results of the angledependent SLS (Fig. 2). The fractal dimension, Df, of SWCNTs is significantly smaller in the case of humic acid (2.2 0.06) as compared to that of glucose (2.7 0.04). The decrease in the Df values in the presence of humic acid indicates more loosely bound clusters of the SWCNTs as compared to a more dense aggregate structure in the case of glucose. TEM micrographs (Fig. S3) show that the majority of the nanotubes are SWCNTs with some impurities; which confirms the manufacturer’s claim of greater than 90% SWCNT purity. Fig. S4 presents the zeta potential values of SWCNTs as a function of pH. The SWCNTs in DI water are negatively charged with a zeta potential of 21 3 mV at pH 3.5, which increases to 53 0.5 mV at pH 11.0. The net charge on the SWCNTs is due to the formation of an electrical double layer that is a few angstroms thick (Saleh et al., 2010). The electrostatic double layer repulsion prevents the aggregation of particles, resulting in a stable SWCNT suspension.
4062
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 5 6 e4 0 6 8
3.3. Adsorption of BPA and EE2 in various synthetic leachates
10
1 0
0.005
0.01 q (1/nm)
0.015
0.02
Fig. 2 e Angle-dependent (30 e85 ) static light scattering profile of SWCNT clusters in fractal regime with glucose and humic acid. The subscripts denote independent measurement of the samples in each condition. Both the glucose and humic acid concentration was maintained at 10 mg/L DOC. The measurements were performed at 22 ± 1 C. Run 1: humic acid (A); glucose (:). Run 2: humic acid (>); glucose (Δ).
An increase in the pH results in a significantly increased net negative charge on the SWCNTs, indicating the possible deprotonation of functional groups on the surfaces of the SWCNT. The physicochemical properties of SWCNTs can also be studied using the radial breathing mode (RBM) response of Raman spectroscopy. Fig. S5 shows the Raman spectrum of the SWCNTs. The primary peak position for the SWCNTs at low frequencies, i.e., near RBM (<300 1/cm), is found to occur at 229 1/cm. The diameter of the SWCNTs can be calculated by applying the following equation for bundled SWCNTs (Saleh et al., 2010): d¼
223:75 u 14
(4)
where d is the diameter of the SWCNTs and u is the peak Raman shift at RBM. Using Eq. (4), the observed peak at 229 1/cm was found to correlate to a SWCNT diameter of 1.04 nm. The Raman spectrum in Fig. S5 also shows a second group of peaks, which appear at 1321 1/cm and 1578 1/cm, corresponding to the D- and G-bands, respectively. D/G ratio was found to be 0.074 in our sample. The D-band can be attributed to minor amorphous carbon and defective structures of hollow graphite cylinders, whereas the strong intensity of the G peak is related to the graphitization of SWCNTs (Dresselhaus et al., 2002). The intensity of the D-band becomes stronger with the expense of the RBM and the intensity of the G-band becomes stronger with an increase of the degree of sidewall functionalization. The intensity ratio of the D-band to the G-band (D/G ratio) can be used to estimate the degree of functionalization of the SWCNTs. However, the SWCNTs have not been functionalized or modified and thus, the degree of functionalization was not applicable in this study.
Fig. 3 shows the removal kinetics of BPA and EE2 from young and old leachate. Equilibrium concentrations of BPA and EE2 appear to be achieved after 4 h of adsorption, which is consistent with the rapid equilibrium times observed in other studies (Peng et al., 2003; Lu et al., 2005). The adsorption of BPA and EE2 differed greatly between the young and the old leachates. In young leachate, the removal of BPA reached 69%, while the removal of EE2 reached 99%. However, in the old leachate, the removal was much lower, with the removal of BPA and EE2 reaching 33% and 61%, respectively. Overall, a significantly larger fraction of EE2 was removed from both leachate solutions, as compared to BPA. The higher adsorption of EE2 over BPA can possibly be explained by its larger log KOW values, suggesting that hydrophobic interactions of the EDCs with the SWCNTs are the dominant adsorption mechanism. This mechanism has been cited in several other adsorption studies using SWCNTs (Chen et al., 2003; Agnihotri et al., 2005; Gotovac et al., 2007). Pan and Xing (2008) also described CNT surfaces as “evenly distributed hydrophobic sites” for organic chemical adsorption. However, an additional mechanism for the adsorption of EDCs, specifically BPA and EE2, may be p-p electron donor-acceptor interactions (Pan et al., 2008). The lower sorption in the case of the more hydrophobic old leachate is likely due to the decreased availability of sorption sites caused by humic acid sorbed onto the SWCNT surfaces. The higher sorption of humic acid to SWCNTs has been demonstrated by the relatively smaller cluster sizes of SWCNTs, as presented in Fig. 1. Fig. 4 shows the adsorption of BPA and EE2 to SWCNTs under all leachate conditions. Linearized Freundlich and Langmuir isotherm models are used to fit the sorption kinetics data. Table 4 lists the fitting parameters of both isotherms for
1 0.8 0.6
C/C0
Scattering intensity, l (kHz)
100
0.4 0.2 0 0
1
2 3 Time (h)
4
5
Fig. 3 e Effect of contact time on adsorption of BPA and EE2 onto SWCNTs in old and young leachate compositions (SWCNT dose [ 50 mg/L). The characteristics of old and young leachate are described in Table 1. BPA, young (:); EE2, young (-); BPA, old (Δ); EE2, old (,).
4063
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 5 6 e4 0 6 8
5.5
5.5
5
5
log qe (µg/g)
6
log qe (µg/g)
6
4.5
4.5
4
4
3.5
3.5
3
3 0
1
2 3 log Ce (µg/L)
0
4
0.1
0.08
0.08
2 3 log Ce (µg/L)
4
1 Ce (mg/L)
2
Ce/qe (mg/g)
Ce/qe (mg/g)
0.1
1
0.06
0.06 0.04
0.04
0.02
0.02
0
0 0
1 Ce (mg/L)
2
0
Fig. 4 e Effect of leachate composition on the adsorption of BPA and EE2 onto SWCNTs (C0 [ 10 mM; contact time [ 4 h). Freundlich isotherms: BPA (a.1) and EE2 (a.2); Langmuir isotherms: BPA (b.1) and EE2 (b.2). The composition of each leachate is described in detail in Table 2. Lines denote linearized model fitting for Freundlich and Langmuir isotherms. Leachate Types: A(,); B(Δ); C(B); D(-); E(:).
the various leachates investigated in this study. Currently, the most frequently applied models to describe the adsorption of organic chemicals onto SWCNTs are Freundlich (Liu et al., 2004; Agnihotri et al., 2005) and Langmuir isotherms (Lu et al., 2005; Su and Lu, 2007). The Freundlich isotherm model describes multi-layer adsorption, where the chemicals initially interact with the SWCNT surfaces and then with each other. Langmuir isotherms, on the other hand, assume a single-layer
adsorption process, where the chemicals only interact with the SWCNT surface throughout the adsorption process. The high correlation (R2) values for both methods, as shown in Table 4, suggest that the adsorption in this study could be explained using either isotherm model. A clear trend of decreased adsorption onto SWCNTs from the youngest to oldest leachates is observed for the removal of both BPA and EE2. Decreasing log Kf values from the Freundlich modeling
Table 4 e Freundlich and Langmuir fitting parameters for adsorption of BPA and EE2 onto SWCNTs in various leachates. Chemical
Leachate type
Freundlich
Langmuir
log Kf (mg/g)/ (mg/L)(1/n)
n1
R2
a (mg/g)
b (L/mg)
R2
BPA
A B C D E
3.80 3.71 3.75 3.65 2.29
0.219 0.237 0.191 0.215 0.625
0.954 0.995 0.998 0.977 0.976
44.8 28.9 23.0 23.5 22.6
3.91 9.89 10.9 6.46 5.60
0.995 0.988 0.994 0.979 0.991
EE2
A B C D E
4.57 3.98 3.77 3.75 3.57
0.194 0.290 0.269 0.222 0.244
0.997 0.997 0.986 0.917 0.926
120 78.1 42.0 27.8 24.9
83.0 8.53 7.93 10.3 4.28
0.999 0.985 0.999 0.997 0.981
4064
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 5 6 e4 0 6 8
(BPA: 3.80e2.29; EE2: 4.57e3.57) and lowered sorption capacity, i.e., ‘a’ values, for Langmuir modeling (BPA: 44.8e22.6; EE2: 120e24.9) from leachate Type A to E further signify a reduction in the adsorptive performance of the SWCNTs with the aging of the leachate. This effect was more pronounced in the case of EE2, as compared to BPA. This phenomenon may be related to the physicochemical properties of these EDCs or the water quality conditions (e.g., pH, ionic strength) in the leachate. The influence of each parameter of the leachate on the adsorption of BPA and EE2 is discussed in the following sections.
3.4.
Effect of pH and ionic strength
3.5.
Effect of the electrolyte species
The effect of the electrolyte species on the adsorption of BPA and EE2 from young and old leachate was examined by varying the concentration of Ca2þ. As shown in Fig. 6, the adsorption of BPA and EE2 from young leachate by SWCNTs was not affected by increasing the concentration of Ca2þ. The control experiments (with no SWCNTs being present) showed that the adsorption of the EDCs from the young leachate was minimal (<10%). However, using the old leachate conditions, an increase in the Ca2þ concentration from 0 to 150 mM increased the adsorption of BPA and EE2 by 12% and 19%, respectively. Furthermore, the control experiments showed that without SWCNTs, the adsorption of BPA and EE2 occurred with removal efficiencies reaching 20% and 35%, respectively. This observation could possibly be explained by the previously mentioned “salting-out” effect, which has been found to be stronger with Ca2þ than with Naþ (Schlautman et al., 2004). However, the adsorption data shown in Fig. 4a suggests that this mechanism may only contribute minimally. Instead, the adsorption of BPA and EE2 could be improved by the interactions between the chemicals and the complexes that can form between the humic acid and the Ca2þ ions, as
1
1
0.8
0.8
0.6
0.6
C/C0
C/C0
The effect of pH on the adsorption of BPA and EE2 in young and old leachates is shown in Fig. 5a. An increase in the pH of the leachates from 3.5 to 7.5 had no effect on the adsorption of BPA, regardless of the type of leachate. Decreased adsorption of BPA was observed in both leachate samples when the pH was increased to 11. Changes in the pH resulted in minimal adsorption variation of EE2 (<2%) in young leachate. In old leachate, however, the adsorption of EE2 decreased by 32% with an increase in pH from 3.5 to 11. The contrast between the reduction in the adsorption of EE2 and BPA with varying pH levels may be attributed to the different log KOW values of each compound. The overall decrease in adsorption may be the result of the increased ionization of BPA and EE2 caused by the increase in pH, possibly leading to reduced hydrophobic interactions with the SWCNTs (Pan et al., 2008). Decreased adsorption of organic chemicals, such as BPA and EE2, can often be expected when the pH is greater than the pKa because both the chemical and the adsorbent (i.e., SWCNTs) become negatively charged and, therefore, can experience increased electrostatic repulsion (Pan and Xing, 2008). The higher log KOW value of EE2 suggests that hydrophobic interactions are predominant, particularly at lower pH values, and significantly reduce as the pH increases. Meanwhile, the adsorption of BPA due to this mechanism remains fairly constant until the pH exceeds the pKa of BPA, at which point adsorption is reduced. The effects of ionic strength on the adsorption of BPA and EE2 are shown in Fig. 5b. Increases in the NaCl concentration, from 0 to 320 mM, did not significantly change the adsorption
of either BPA or EE2 in both leachate solutions. Several studies have shown that increased ionic strength can enhance the adsorption of compounds onto carbonaceous materials due to the screening effect of the surface charge produced by adding salt (Vinu et al., 2006; Fontecha-Camara et al., 2007). On the other hand, many researchers have found that increasing the ionic strength has a negligible effect on the adsorption of various organic chemicals (Gotovac et al., 2007; Chen et al., 2008a; Zhang et al., 2010). One explanation for this observation could be the “salting-out” effect, which refers to the reduced solubility of organic compounds in aqueous salt solutions (Xie et al., 1997). This mechanism could possibly increase the accessibility of the surface of the SWCNTs to the EDCs. However, the adsorption of the BPA and EE2 from young leachate, shown in Fig. 5b, showed no measurable differences upon the addition of NaCl, suggesting that the “salting-out” effect may not enhance the removal of EDCs.
0.4
0.4
0.2
0.2
0
0 0
3
6 pH
9
12
0
200 NaCl (mM)
400
Fig. 5 e Effect of background solution chemistry on adsorption of BPA and EE2 in young and old leachate conditions: (a) Effect of pH and (b) effect of ionic strength represented by NaCl concentration. (C0 [ 10 mM; contact time [ 4 h; SWCNT dose [ 50 mg/L). BPA, young (:); EE2, young (-); BPA, old (Δ); EE2, old (,).
4065
1
1
0.8
0.8
0.6
0.6
C/C0
C/C0
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 5 6 e4 0 6 8
0.4 0.2
0.4 0.2
0
0 0
100
200 300 Ca2+ (mM)
400
0
50
100 150 Ca2+ (mM)
200
Fig. 6 e Effect of Ca2D on adsorption of BPA and EE2 using (a) young leachate conditions and (b) old leachate conditions. Open symbols (>,B) denote adsorption of BPA and EE2 without SWCNTs, respectively; filled symbols (A,C) denote adsorption of BPA and EE2 with SWCNT dose [ 50 mg/L and contact time [ 4 h.
previously mentioned. The concentration of these complexes increases with higher concentrations of Ca2þ, and with the hydrophobic nature of humic acid, these complexes may facilitate additional hydrophobic interactions among the EDCs by providing additional surfaces that can serve as adsorption sites.
3.6.
Effect of DOC
1
1
0.8
0.8
0.6
0.6
C/C0
C/C0
Fig. 7 shows the effect of increased concentrations of DOC on the adsorption of BPA and EE2 in the absence and presence of SWCNTs. An increase in the concentration of glucose did not have a significant effect on the adsorption of BPA and EE2, as shown in Fig. 7a, which may be due to the hydrophilic nature of glucose. Fig. 7b illustrates the extent of removal of BPA and EE2 that was achieved with the addition of humic acid as DOC in the leachate solution. With removal efficiencies of 15% for BPA and 20% for EE2 achieved with humic acid only, it can be inferred that hydrophobic interactions between EDCs and humic acid contribute to their removal. However, an increase in the humic acid concentration resulted in a decrease of up to 55% in the
adsorption of BPA and 40% in the adsorption of EE2 by SWCNTs. This decrease in adsorption onto carbonaceous surfaces, such as SWCNTs, due to the presence of humic acid was also found in previous studies (Kilduff and Wigton, 1999; Pignatello et al., 2006). Researchers have proposed two potential mechanisms to account for this: (i) direct competition between the EDCs and the humic acid for adsorption sites on the SWCNTs, and (ii) physical pore blockage, which reduces the amount of surface area available for adsorption (Newcombe et al., 1997; Kilduff et al., 1998; Chen et al., 2008a; Zhang et al., 2010). Based on the available sites for adsorption (i.e., interstitial channels of the SWCNTs) and the size of the humic acid used in this study, both competitive adsorption and pore blockage could occur to reduce the adsorption of BPA and EE2 by SWCNTs in old leachate. In this study, humic acid concentrations were used to represent the range of DOC that is typically found in old leachate. Therefore, the concentration of humic acid that may completely inhibit the adsorption of EDCs by SWCNTs was not investigated, which may be important when studying contaminant removal in waterways with large quantities of hydrophobic organic matter (e.g., municipal wastewater).
0.4 0.2
0.4 0.2
0
0 0
1000 2000 DOC (mg/L)
3000
0
100 200 DOC (mg/L)
300
Fig. 7 e Effect of DOC on adsorption of BPA and EE2 using (a) young leachate conditions which uses glucose as source of DOC and (b) old leachate conditions, which use humic acid as source of DOC. Open symbols (>,B) denote adsorption of BPA and EE2 without SWCNTs, respectively; filled symbols (A,C) denote adsorption of BPA and EE2 with SWCNT dose [ 50 mg/L and contact time [ 4 h.
4066
4.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 5 6 e4 0 6 8
Conclusions
Overall, this study demonstrated high removal efficiencies of BPA and EE2 from simulated landfill leachate using SWCNTs. The adsorptive capacity of the SWCNTs was consistently higher for EE2 than for BPA, possibly due to the higher log KOW value of EE2, which leads to increased hydrophobic interactions. The increases of the pH of the leachate from 3 to 11 impaired the removal of both BPA and EE2. Changes in ionic strength, when adjusted with NaCl, had a negligible effect on adsorptive characteristics. Increasing the concentration of Ca2þ also did not have an effect on adsorption in the young leachate, but effectively enhanced the adsorption of BPA and EE2 in the old leachate. This phenomenon may be caused by complexes that are formed between the Ca2þ ions and humic acid, possibly producing additional adsorption sites for the EDCs. The most influential parameter on the adsorption of EDCs onto SWCNTs was DOC. In young leachate, the organic carbon is typically hydrophilic, while the organic carbon in old leachate is hydrophobic and has a higher molar mass. In our study, glucose had no effect on the adsorption onto SWCNTs; humic acid significantly decreased the adsorption of BPA and EE2 by SWCNTs. This observation can be explained either by direct competition between the EDCs and humic acid for adsorption sites on the SWCNTs or limited access to SWCNT surfaces due to pore blockage caused by the hydrophobic DOC. The latter mechanism is supported by the physicochemical characterization of SWCNT clusters. Solutions of SWCNTs with humic acid showed a significantly smaller aggregate size, which indicates higher SWCNT surface coverage by the organics. Moreover, in the absence of SWCNTs, humic acid adsorbed BPA and EE2, which suggests that hydrophobic interactions occur between humic acid and EDCs. Carbon nanotubes are potentially useful for the treatment of landfill leachate, particularly for the removal of micropollutants. Based on this study, SWCNTs have been shown to be effective in removing low concentrations of EDCs from leachate. The use of SWCNTs appears to be more effective in young leachates than in old leachates due to its lower pH and the presence of simple, hydrophilic organic carbon. However, the reported wide ranges of pH and variations of hydrophilic and hydrophobic organic carbon found in leachates may severely inhibit the removal of target organic contaminants due to competition between the DOC and the contaminants for adsorption sites on the SWCNTs. Future landfill leachate treatment studies using SWCNTs should examine their ability to remove other EDCs and emerging contaminants in solutions with high concentrations of hydrophobic DOC. A detailed mechanistic study of the interactions between those pollutants, SWCNTs, the DOC, and the other components in the leachate must also be conducted. Future research must also be conducted to compare the removal of EDCs using SWCNTs and activated carbon. While activated carbon is often used in the treatment of landfill leachate, no published literature related to EDC removal was found. Therefore, direct comparisons to the performance of SWCNTs and activated carbon in EDC removal cannot be made. Implementation of SWCNTs into the full-scale treatment of landfill leachate for
the removal of EDCs and emerging contaminants cannot be determined until its adsorptive capacity is shown to greatly exceed that of activated carbon.
Acknowledgements This research was supported by GS E&C Research Institute. Funding was provided by a grant (07SeaHeroB01-01) from the Plant Technology Advancement Program of the Ministry of Land, Transport, and Maritime Affairs of the Korean government and by the University of South Carolina.
Appendix. Supplementary material Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.watres.2011.05.015.
references
Agnihotri, S., Rood, M.J., Rostam-Abadi, M., 2005. Adsorption equilibrium of organic vapors on single-walled carbon nanotubes. Carbon 43, 2379e2388. Alum, A., Yoon, Y., Westerhoff, P., Abbaszadegan, M., 2004. Oxidation of bisphenol A, 17b-estradiol, and 17a-ethinyl estradiol and byproduct estrogenicity. Environ. Toxicol. 19, 257e264. Artiola-Fortuny, J., Fuller, W.H., 1982. Humic substances in landfill leachates: I. Humic acid extraction and identification. J. Environ. Qual. 11, 663e669. Asakura, H., Matsuto, T., Tanaka, N., 2004. Behavior of endocrinedisrupting chemicals in leachate from MSW landfill sites in Japan. Waste Manage. 24, 613e622. Aziz, H.A., Adlan, M.N., Zahari, M.S.M., Alias, S., 2004. Removal of ammoniacal nitrogen (NeNH3) from municipal solid waste leachate by using activated carbon and limestone. Waste Manage. Res. 22, 371e375. Barlaz, M.A., Ham, R.K., 1993. Leachate and gas generation. In: Daniel, D.E. (Ed.), Geotechnical Practice for Waste Disposal. Chapman and Hall, London, pp. 114e136. Behnisch, P.A., Fujii, K., Shiozaki, K., Kawakami, I., Sakai, S., 2001. Estrogenic and dioxin-like potency in each step of a controlled landfill leachate treatment plant in Japan. Chemosphere 43, 977e984. Cheek, A.O., Vonier, P.M., Oberdorster, E., Burow, B.C., McLachlan, J.A., 1998. Environmental signaling: a biological context for endocrine disruption. Environ. Health Perspect. 106, 5e10. Chen, R.J., Bangsaruntip, S., Drouvalakis, K.A., Kam, N.W.S., Shim, M., Li, Y.M., Kim, W., Utz, P.J., Dai, H.J., 2003. Noncovalent functionalization of carbon nanotubes for highly specific electronic biosensors. Proc. Natl. Acad. Sci. 100, 4984e4989. U.S.A. Chen, J.Y., Chen, W., Zhu, D., 2008a. Adsorption of nonionic aromatic compounds to single-walled carbon nanotubes: effects of aqueous solution chemistry. Environ. Sci. Technol. 42, 7225e7230. Chen, W., Duan, L., Wang, L.L., Zhu, D.Q., 2008b. Adsorption of hydroxyl- and amino-substituted aromatics to carbon nanotubes. Environ. Sci. Technol. 42, 6862e6868. Chian, E.S.K., Dewalle, F.B., 1977. Characterization of soluble organic matter in leachate. Environ. Sci. Technol. 11, 158e163.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 5 6 e4 0 6 8
Chingombe, P., Saha, B., Wakeman, R.J., 2005. Surface modification and characterisation of a coal-based activated carbon. Carbon 43, 3132e3143. Christensen, T.H., Kjeldsen, P., Albrechtsen, H.J., Heron, G., Nielsen, P.H., Bjerg, P.L., Holm, P.E., 1994. Attenuation of landfill leachate pollutants in aquifers. Crit. Rev. Environ. Sci. Technol. 24, 119e202. Christensen, J.B., Jensen, D.L., Gron, C., Filip, Z., Christensen, T.H., 1998. Characterization of the dissolved organic carbon in landfill leachate-polluted groundwater. Water Res. 32, 125e135. Colborn, T., Saal, F.S.V., Soto, A.M., 1993. Developmental effects of endocrine-disrupting chemicals in wildlife and humans. Environ. Health Perspect. 101, 378e384. Daughton, C.G., Ternes, T.A., 1999. Pharmaceuticals and personal care products in the environment: agents of subtle change? Environ. Health Perspect. 107, 907e938. DeRudder, J., Van de Wiele, T., Dhooge, W., Comhaire, F., Verstraete, W., 2004. Advanced water treatment with manganese oxide for the removal of 17 a-ethynyl estradiol (EE2). Water Res. 38, 184e192. Dresselhaus, M.S., Dresselhaus, G., Jorio, A., Souza Filho, A.G., Saito, R., 2002. Raman spectroscopy on isolated single wall carbon nanotubes. Carbon 40, 2043e2061. Ehrig, H.J., 1988. The Landfill.in. In: Baccini, P. (Ed.), Water and element balances of landfills. Springer Verlag, Berlin, Germany. Ehrig, H.J., 1989. Leachate quality.in. In: Christensen, T.H., Cossu, R. , Stegmann, R. (Eds.), Sanitary Landfilling: Process, Technology and Environmental Impact. Academic Press Limited. El-Fadel, M., Findikakis, A.N., Leckie, J.O., 1997. Modeling leachate generation and transport in solid waste landfills. Environ. Technol. 18, 669e686. Fontecha-Camara, M.A., Lopez-Ramon, M.V., Alvarez-Merino, M. A., Moreno-Castilla, C., 2007. Effect of surface chemistry, solution pH, and ionic strength on the removal of herbicides diuron and amitrole from water by an activated carbon fiber. Langmuir 23, 1242e1247. Foo, K.Y., Hameed, B.H., 2009. An overview of landfill leachate treatment via activated carbon adsorption process. J. Hazard. Mater. 171, 54e60. Frimmel, F.H., Weis, M., 1991. Aging effects of high molecular weight organic acids which can be isolated from landfill leachates. Water Sci. Technol. 23, 419e426. Gotovac, S., Yang, C.M., Hattori, Y., Takahashi, K., Kanoh, H., Kaneko, K., 2007. Adsorption of polyaromatic hydrocarbons on single wall carbon nanotubes of different functionalities and diameters. J. Colloid Interf. Sci. 314, 18e24. Harmsen, J., 1983. Identification of organic compounds in leachate from a waste tip. Water Res. 17, 699e705. Harrison, P.T.C., Holmes, P., Humfrey, C.D.N., 1997. Reproductive health in humans and wildlife: are adverse trends associated with environmental chemical exposure? Sci. Total Environ. 205, 97e106. Holm, J.V., Rugge, K., Bjerg, P.L., Christensen, T.H., 1995. Occurrence and distribution of pharmaceutical organic compounds in the groundwater downgradient of a landfill (Grindsted, Denmark). Environ. Sci. Technol. 29, 1415e1420. Kasprzyk-Hordern, B., Dinsdale, R.M., Guwy, A.J., 2008. The occurrence of pharmaceuticals, personal care products, endocrine disruptors and illicit drugs in surface water in South Wales. Water Res. 42, 3498e3518. UK. Kilduff, J.E., Wigton, A., 1999. Sorption of TCE by humic-preloaded activated carbon: Incorporating size-exclusion and pore blockage phenomena in a competitive adsorption model. Environ. Sci. Technol. 33, 250e256. Kilduff, J.E., Karanfil, T., Weber, W.J., 1998. Competitive effects of nondisplaceable organic compounds on trichloroethylene uptake by activated carbon. II. Model verification and applicability to natural organic matter. J. Colloid Interf. Sci. 205, 280e289.
4067
Kjeldsen, P., Christophersen, M., 2001. Composition of leachate from old landfills in Denmark. Waste Manage. Res. 19, 249e256. Kjeldsen, P., Barlaz, M.A., Rooker, A.P., Baun, A., Ledin, A., Christensen, T.H., 2002. Present and long-term composition of MSW landfill leachate: a review. Crit. Rev. Environ. Sci. Technol. 32, 297e336. Kolpin, D.W., Furlong, E.T., Meyer, M.T., Thurman, E.M., Zaugg, S. D., Barber, L.B., Buxton, H.T., 2002. Pharmaceuticals, hormones, and other organic wastewater contaminants in US streams, 1999-2000: a national reconnaissance. Environ. Sci. Technol. 36, 1202e1211. Krug, M.N., Ham, R.K., 1991. Analysis of long-term leachate characteristics. 3rd International Landfill Symposium, Cagliari, Italy. Kuo, C.Y., 2009. Comparison with as-grown and microwave modified carbon nanotubes to removal aqueous bisphenol A. Desalination 249, 976e982. Li, Y.H., Di, Z.C., Ding, J., Wu, D.H., Luan, Z.K., Zhu, Y.Q., 2005. Adsorption thermodynamic, kinetic and desorption studies of Pb2þ on carbon nanotubes. Water Res. 39, 605e609. Lim, Y.N., Shaaban, M.G., Yin, C.Y., 2009. Treatment of landfill leachate using palm shell-activated carbon column: axial dispersion modeling and treatment profile. Chem. Eng. J. 146, 86e89. Lin, D.H., Xing, B.S., 2008. Adsorption of phenolic compounds by carbon nanotubes: Role of aromaticity and substitution of hydroxyl groups. Environ. Sci. Technol. 42, 7254e7259. Linde, K., Jonsson, A.S., 1995. Nanofiltration of salt solutions and landfill leachate. Desalination 103, 223e232. Liu, G.H., Wang, J.L., Zhu, Y.F., Zhang, X.R., 2004. Application of multiwalled carbon nanotubes as a solid-phase extraction sorbent for chlorobenzenes. Anal. Lett. 37, 3085e3104. Lopez-Cervantes, J., Paseiro-Losada, P., 2003. Determination of bisphenol A in, and its migration from, PVC stretch film used for food packaging. Food Addit. Contam. 20, 596e606. Lu, C.Y., Chiu, H.S., 2006. Adsorption of zinc(II) from water with purified carbon nanotubes. Chem. Eng. Sci. 61, 1138e1145. Lu, C.S., Chung, Y.L., Chang, K.F., 2005. Adsorption of trihalomethanes from water with carbon nanotubes. Water Res. 39, 1183e1189. Nerin, C., Fernandez, C., Domeno, C., Salafranca, J., 2003. Determination of potential migrants in polycarbonate containers used for microwave ovens by high-performance liquid chromatography with ultraviolet and fluorescence detection. J. Agr. Food Chem. 51, 5647e5653. Newcombe, G., Drikas, M., Hayes, R., 1997. Influence of characterised natural organic material on activated carbon adsorption .2. Effect on pore volume distribution and adsorption of 2-methylisoborneol. Water Res. 31, 1065e1073. Oleszczuk, P., Pan, B., Xing, B.S., 2009. Adsorption and desorption of oxytetracycline and carbamazepine by multiwalled carbon nanotubes. Environ. Sci. Technol. 43, 9167e9173. Pan, B., Xing, B.S., 2008. Adsorption mechanisms of organic chemicals on carbon nanotubes. Environ. Sci. Technol. 42, 9005e9013. Pan, B., Xing, B.S., 2010. Adsorption kinetics of 17 a-ethinyl estradiol and bisphenol A on carbon nanomaterials. I. Several concerns regarding pseudo-first order and pseudo-second order models. J. Soils Sediments 10, 838e844. Pan, B., Lin, D.H., Mashayekhi, H., Xing, B.S., 2008. Adsorption and hysteresis of bisphenol A and 17 a-ethinyl estradiol on carbon nanomaterials. Environ. Sci. Technol. 42, 5480e5485. Pan, B., Sun, K., Xing, B.S., 2010. Adsorption kinetics of 17 aethinyl estradiol and bisphenol A on carbon nanomaterials. II. Concentration-dependence. J. Soils Sediments 10, 845e854. Peng, X.J., Li, Y.H., Luan, Z.K., Di, Z.C., Wang, H.Y., Tian, B.H., Jia, Z.P., 2003. Adsorption of 1,2-dichlorobenzene from water to carbon nanotubes. Chem. Phys. Lett. 376, 154e158.
4068
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 0 5 6 e4 0 6 8
Peters, T.A., 1998. Purification of landfill leachate with reverse osmosis and nanofiltration. Desalination 119, 289e293. Pignatello, J.J., Kwon, S., Lu, Y.F., 2006. Effect of natural organic substances on the surface and adsorptive properties of environmental black carbon (char): attenuation of surface activity by humic and fulvic acids. Environ. Sci. Technol. 40, 7757e7763. Reinhart, D.R., Al-Yousfi, A.B., 1996. The impact of leachate recirculation on municipal solid waste landfill operating characteristics. Waste Manage. Res. 14, 337e346. Reinhart, D.R., Townsend, T.G., 1998. Landfill Bioreactor Design and Operation. CRC Press, Boca Raton, FL. Rivas, F.J., Beltran, F.J., Gimeno, O., Frades, J., Carvalho, F., 2006. Adsorption of landfill leachates onto activated carbon Equilibrium and kinetics. J. Hazard. Mater. 131, 170e178. Rodriguez, J., Castrillon, L., Maranon, E., Sastre, H., Fernandez, E., 2004. Removal of non-biodegradable organic matter from landfill leachates by adsorption. Water Res. 38, 3297e3303. Saleh, N.B., Pfefferle, L.D., Elimelech, M., 2010. Influence of biomacromolecules and humic acid on the aggregation kinetics of single-walled carbon nanotubes. Environ. Sci. Technol. 44, 2412e2418. Schlautman, M.A., Yim, S.B., Carraway, E.R., Lee, J.H., Herbert, B.E. , 2004. Testing a surface tension-based model to predict the salting out of polycyclic aromatic hydrocarbons in model environmental solutions. Water Res. 38, 3331e3339. Scott, J., Beydoun, D., Amal, R., Low, G., Cattle, J., 2005. Landfill management, leachate generation, and leach testing of solid wastes in Australia and overseas. Crit. Rev. Environ. Sci. Technol. 35, 239e332. Segner, H., Navas, J.M., Schafers, C., Wenzel, A., 2003. Potencies of estrogenic compounds in in vitro screening assays and in life cycle tests with zebrafish in vivo. Ecotox. Environ. Safe 54, 315e322. Shiraishi, H., Nakasugi, O., Hashimoto, S., Yamamoto, T., Yasuhara, A., Yasuda, K., 1999. Endocrine disrupters in the leachate from waste disposal sites. Waste Manage. Res. 10, 33e45. Cheap Tubes, I. 2009. Single walled nanotubes-SWNTs. The Source for Single Walled Nanotubes-SWNTs. Snyder, S.A., Westerhoff, P., Yoon, Y., Sedlak, D.L., 2003. Pharmaceuticals, personal care products, and endocrine disruptors in water: implications for the water industry. Environ. Eng. Sci. 20, 449e469. Staples, C.A., Dorn, P.B., Klecka, G.M., O’Block, S.T., Harris, L.R., 1998. A review of the environmental fate, effects, and exposures of bisphenol A. Chemosphere 36, 2149e2173. Su, F.S., Lu, C.S., 2007. Adsorption kinetics, thermodynamics and desorption of natural dissolved organic matter by multiwalled carbon nanotubes. J. Environ. Sci. Health Part A 42, 1543e1552.
Tchobanoglous, G., Theisen, H., Vigil, S., 1993. Integrated Solid Waste Management: Engineering Principles and Management Issues. Irwin/McGraw Hill, Boston, MA. Ternes, T.A., Meisenheimer, M., McDowell, D., Sacher, F., Brauch, H.J., Gulde, B.H., Preuss, G., Wilme, U., Seibert, N.Z., 2002. Removal of pharmaceuticals during drinking water treatment. Environ. Sci. Technol. 36, 3855e3863. Urase, T., Salequzzaman, M., Kobayashi, S., Matsuo, T., Yamamoto, K., Suzuki, N., 1997. Effect of high concentration of organic and inorganic matters in landfill leachate on the treatment of heavy metals in very low concentration level. Water Sci. Technol. 36, 349e356. Vinu, A., Hossain, K.Z., Kumar, G.S., Ariga, K., 2006. Adsorption of L-histidine over mesoporous carbon molecular sieves. Carbon 44, 530e536. Wang, X.K., Chen, C.L., Hu, W.P., Ding, A.P., Xu, D., Zhou, X., 2005. Sorption of 243Am(III) to multiwall carbon nanotubes. Environ. Sci. Technol. 39, 2856e2860. Westerhoff, P., Yoon, Y., Snyder, S., Wert, E., 2005. Fate of endocrine-disruptor, pharmaceutical, and personal care product chemicals during simulated drinking water treatment processes. Environ. Sci. Technol. 39, 6649e6663. Wintgens, T., Gallenkemper, M., Melin, T., 2003. Occurrence and removal of endocrine disrupters in landfill leachate treatment plants. Water Sci. Technol. 48, 127e134. Xie, W.H., Shiu, W.Y., Mackay, D., 1997. A review of the effect of salts on the solubility of organic compounds in seawater. Mar. Environ. Res. 44, 429e444. Yamamoto, T., Yasuhara, A., Shiraishi, H., Nakasugi, O., 2001. Bisphenol A in hazardous waste landfill leachates. Chemosphere 42, 415e418. Yang, K., Zhu, L.Z., Xing, B.S., 2006. Adsorption of polycyclic aromatic hydrocarbons by carbon nanomaterials. Environ. Sci. Technol. 40, 1855e1861. Yang, K., Wu, W.H., Jing, Q.F., Zhu, L.Z., 2008. Aqueous adsorption of aniline, phenol, and their substitutes by multi-walled carbon nanotubes. Environ. Sci. Technol. 42, 7931e7936. Yoon, Y., Westerhoff, P., Snyder, S.A., Esparza, M., 2003. HPLCfluorescence detection and adsorption of bisphenol A, 17 b-estradiol, and 17 a-ethinyl estradiol on powdered activated carbon. Water Res. 37, 3530e3537. Yoon, Y., Westerhoff, P., Snyder, S.A., Wert, E.C., 2006. Nanofiltration and ultrafiltration of endocrine disrupting compounds, pharmaceuticals and personal care products. J. Membr. Sci. 270, 88e100. Zhang, S.J., Shao, T., Bekaroglu, S.S.K., Karanfil, T., 2010. Adsorption of synthetic organic chemicals by carbon nanotubes: effects of background solution chemistry. Water Res. 44, 2067e2074.
