WATER RESEARCH A Journal of the International Water Association
Editor-in-Chief Mark van Loosdrecht Delft University of Technology Department of Biochemical Engineering Julianalaan 67 2628 BC Delft The Netherlands Tel: +31 15 27 81618 E-mail:
[email protected]
Editors J. Block Université H. Poincaré, Nancy I France David Dixon University of Melbourne Australia Hiroaki Furumai The University of Tokyo Japan Xiaodi Hao Beijing University of Civil Engineering and Architecture China Gregory Korshin University of Washington USA Anna Ledin Formas Sweden Eberhard Morgenroth Swiss Federal Institute of Aquatic Science and Technology (EAWAG) Switzerland W. Rauch University Innsbruck Austria Maria Reis Universidade Nova de Lisboa/FCT Portugal Hang-Shik Shin Korea Advanced Institute of Science and Technology Korea Thomas Ternes Bundesanstalt für Gewässerkunde Germany Stefan Wuertz Univ. of California, Davis USA
Associate Editors Andrew Baker University of New South Wales Australia
Damien Batstone The University of Queensland Australia G-H. Chen The Hong Kong University of Science & Technology Hong Kong China Tom Curtis Univ. of Newcastle upon Tyne UK Ana Deletic Monash University USA Francis de los Reyes III North Carolina State University USA Rob Eldridge The University of Melbourne Australia Rosina Girones University of Barcelona Spain Stephen Gray Victoria University Australia Kate Grudpan Chiang Mai University Thailand E.E. Herricks University of Illinois - Urbana USA Peter Hillis United Utilities Plc UK H-Y. Hu Tsinghua University China P.M. Huck University of Waterloo Canada Bruce Jefferson Cranfield University UK Ulf Jeppsson Lund University Sweden Sergey Kalyuzhnyi Moscow State University Russian Federation Jaehong Kim Georgia Institute of Technology USA Jes La Cour Jansen Lund Institute of Technology Sweden G. Langergraber BOKU/Univ. of Natural Res. and Applied Life Scs. Austria S-L. Lo National Taiwan University Taiwan Dionisis Mantzavinos Technical University of Crete Greece
Y. Matsui Hokkaido University Japan A. Maul Université Paul Verlaine-Metz France Max Maurer EAWAG Switzerland How Yong Ng National University of Singapore Singapore Satoshi Okabe Hokkaido University Japan S.L. Ong National University of Singapore Singapore Jong M. Park Pohang University of Science & Technology Korea Susan Richardson U.S. Environmental Protection Agency USA Miguel Salgot University of Barcelona Spain David Sedlak University of California, Berkeley USA Jean-Philippe Steyer LBE-INRA France M. Takahashi Hokkaido University Japan Kai Udert EAWAG Switzerland V.P. Venugopalan BARC Facilities India E. von Sperling Federal University of Minas Gerais Brazil Hanqing Yu University of Science & Technology of China China J. Zilles University of Illinois Urbana USA A.I. Zouboulis Aristotle University Greece
Editorial Office Mailing address Water Research Radarweg 29 1043 NX Amsterdam The Netherlands Tel: 31 20 485 2701 Fax: 31 20 485 2521; 31 20 485 3264 E-mail:
[email protected]
Publication information: Water Research (ISSN 0043-1354). For 2011, volume 45 is scheduled for publication. Subscription prices are available upon request from the publisher or from the Elsevier Customer Service Department nearest you or from this journal’s website (http://www.elsevier.com/locate/watres). Further information is available on this journal and other Elsevier products through Elsevier’s website (http://www.elsevier.com). Subscriptions are accepted on a prepaid basis only and are entered on a calendar year basis. Issues are sent by standard mail (surface within Europe, air delivery outside Europe). Priority rates are available upon request. Claims for missing issues should be made within six months of the date of dispatch. Author Enquiries: Please visit this journal’s homepage at http://www.elsevier.com/locate/watres. You can track accepted articles at http://www.elsevier.com/ trackarticle and set up e-mail alerts to inform you of when an article’s status has changed. Also accessible from here is information on copyright, frequently asked questions and more. Contact details for questions arising after acceptance of an article, especially those relating to proofs, will be provided by the publisher. Orders, claims, and journal enquiries: Please contact the Elsevier Customer Service Department nearest you: St. Louis: Elsevier Customer Service Department, 11830 Westline Industrial Drive, St. Louis, MO 63146, USA; phone: (877) 8397126 [toll free within the USA]; (+1) (314) 4537076 [outside the USA]; fax: (+1) (314) 5235153; e-mail:
[email protected] Oxford: Elsevier Customer Service Department, The Boulevard, Langford Lane, Kidlington OX5 1GB, UK; phone: (+44) (1865) 843434; fax: (+44) (1865) 843970; e-mail:
[email protected] Tokyo: Elsevier Customer Service Department, 4F Higashi-Azabu, 1-Chome Bldg, 1-9-15 Higashi-Azabu, Minato-ku, Tokyo 106-0044, Japan; phone: (+81) (3) 5561 5037; fax: (+81) (3) 5561 5047; e-mail:
[email protected] Singapore: Elsevier Customer Service Department, 3 Killiney Road, #08-01 Winsland House I, Singapore 239519; phone: (+65) 63490222; fax: (+65) 67331510; e-mail:
[email protected] Application for membership of International Water Association should be made to: Executive Director, IWA, Alliance House, 12 Caxton Street, London SW1H 0QS, U.K. (Tel.: +44 207 654 5500; Fax: +44 207 654 5555; e-mail:
[email protected]; website: http://www.IWAhq.org.uk). Registered Charity (England) No. 289269. Individual membership is available from £30 upwards. For details contact IWA.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 1 9 e1 5 2 8
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Feasibility of a two-stage reduction/subsequent oxidation for treating Tetrabromobisphenol A in aqueous solutions Si Luo, Shao-gui Yang*, Cheng Sun*, Xiao-dong Wang State Key Laboratory of Pollution Control and Resource Reuse, School of the Environment, Nanjing University, Nanjing 210093, PR China
article info
abstract
Article history:
A “two-stage reduction/subsequent oxidation” (T-SRO) process consists of FeeAg reduction
Received 3 July 2010
and Fenton-like oxidation under ultrasound (US) radiation. Due to the refractory oxidation
Received in revised form
of brominated flame retardant, T-SRO was employed to remove Tetrabromobisphenol
22 October 2010
A (TBBPA) by the combination of first debromination and succeeding oxidation. It indicated
Accepted 31 October 2010
that the T-SRO process resulted in a complete decrease in TBBPA concentration and
Available online 10 December 2010
a 99.2% decrease in BPA concentration. The T-SRO process for the removal of TBBPA is much effective than Fenton-like oxidation of TBBPA alone. The result showed that US
Keywords:
radiation improved the Fenton-like oxidation rate of BPA solutions. The addition of dis-
Two-stage
solved iron into the Fenton-like oxidation system could accelerate the first 2 min reaction,
Debromination
but had little effect on the following process. The main intermediate products resulting
Fenton-like oxidation
from TBBPA reduction and BPA oxidation were identified by GCeMS and LC-MS/MS. On the
Tetrabromobisphenol A
basis of this analysis, reactions with OH radical were identified as the major chemical
Bimetallic nanoparticles
pathways during BPA oxidation. ª 2010 Published by Elsevier Ltd.
Ultrasound radiation
1.
Introduction
Tetrabromobisphenol A (TBBPA) is one of the most widely used brominated flame retardant around the world. It can be covalently bound to the polymer in the manufacturing process (de Wit, 2002). TBBPA and its dimethylated derivative have been detected in various environmental matrices, and they negatively affect various aspects of mammalian and human physiology (Sellstro¨m and Jansson, 1995; Helleday ¨ berg et al., 2002). Conseet al., 1999; Meerts et al., 2000; O quently, removal of TBBPA in the contaminated environment is necessary and significant. The reported treatment mainly includes biotransformation, photochemical transformations and thermal decomposition (Mackenzie and Kopinke, 1996; Barontini et al., 2004; Eriksson et al., 2004). In addition, it also indicated that removal of the halogen substituent is a key step in the degradation of halogenated aromatic compounds.
This may occur as an initial step via reductive, hydrolytic, or oxygenolytic mechanisms or may occur after ring cleavage at a later stage of degradation (Monserrate and Haggblom, 1997). Zero valent iron (ZVI) and bimetallic particles have been used for degradation of halogen-containing organic substance (Orth and Gillham, 1996; Cwiertny et al., 2006). In our previous work (Luo et al., 2010), we reported that TBBPA was reductively debrominated to bisphenol A (BPA) over FeeAg bimetallic nanoparticles under US radiation. However, it is well known that BPA exhibits estrogenic activity, which increases the proliferation rate of breast cancer cells and induces the acute toxicity to freshwater and marine species (Pulgar et al., 1998; Kaiser, 2000). Therefore, the debromination of TBBPA in FeeAg/US system is incomplete, BPA must be further degraded. An effective method for BPA mineralizing is the application of Fenton (Fenton-like) oxidation technologies (Go¨zmen et al., 2003; Ioan et al., 2007). On the other hand,
* Corresponding authors. Tel./fax: þ86 25 89680580. E-mail addresses:
[email protected] (S.-g. Yang),
[email protected] (C. Sun). 0043-1354/$ e see front matter ª 2010 Published by Elsevier Ltd. doi:10.1016/j.watres.2010.10.039
1520
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 1 9 e1 5 2 8
ZVI can be used to substitute ferrous salts in the Fenton-like oxidation, and it seems to have similar degradation rates to homogeneous ferrous catalyst. It has confirmed that phenol (Bremner et al., 2006) and 4-chlorophenol (Zhou et al., 2008) could be rapidly degraded in ZVI/H2O2 system. Since iron can reductively transform the electron-withdrawing moieties and render recalcitrant compounds more amenable to subsequent oxidation processes, several researchers presented the ZVI reduction for the pretreatment of wastewater (Mantha et al., 2001, 2002; Oh et al., 2005). Oh et al. (2003) reported the enhanced Fenton oxidation of TNT and RDX through pretreatment with ZVI. Thus, in consideration of the complete treatment of TBBPA, ZVI-based reductive debromination followed by Fenton-like oxidation is proposed, where FeeAg bimetallic nanoparticles are used to debrominate TBBPA because of its higher catalytic activity relative to ZVI. This paper evaluates the effectiveness and feasibility of a T-SRO treatment of TBBPA. Experiments are conducted to examine separately the performance of the FeeAg nanoparticles reductive and Fenton-like oxidative systems. The effect of US radiation in Fenton-like oxidation process is discussed; the influence of dissolved iron (ferrous and ferric ions) on the oxidation kinetics of ZVI/H2O2 system is also investigated. On the basis of identifying intermediate and final products, the reaction pathways are proposed.
2.
Experimental section
2.1.
Materials
2.3.
The reduction of TBBPA (5 mg L1) was conducted by FeeAg nanoparticles (0.8 g L1) under US radiation (40 kHz and 100 W). Debromination experiments were performed in a chamber as shown in Fig. SM-1 (a) attached in Supplemental Material (SM). The detailed procedure of reduction experiment was reported in the literature (Luo et al., 2010).
2.4.
2.2. Synthesis and characterization of FeeAg bimetallic nanoparticles FeeAg bimetallic nanoparticles with core-shell structure were synthesized by reductive deposition of Ag on ZVI nanoparticles as described in the literature (Luo et al., 2010). Various analytical techniques including XRD, XPS and XRF were used to characterize the fresh and reacted (after reduction process) FeeAg bimetallic samples. X-ray diffraction (XRD) analyses of the samples were performed using ˚ ). Switzerland ARL X’TRA X-ray diffractometer (l ¼ 1.5418 A The metal oxidation states and surface atomic composition of FeeAg samples was examined via X-ray photoelectron spectroscopy (XPS, Thermo VG Scientific ESCALAB 250). X-ray fluorescence (XRF, Switzerland ARL Corporation) was used to measure the mass of Ag deposited on the surface of nanoiron.
Fenton-like oxidation experiment
To keep a constant temperature (25 1 C), the Fenton-like process was conducted in a chamber as presented in Fig. SM-1 (b). The reduction and oxidation experiments were carried out in the same vessel. In each bottle, the solution contained BPA and FeeAg nanoparticles after reduction. Its initial pH was adjusted to 3.0 0.1 with 0.1 M H2SO4 and 0.1 M NaOH solutions. The oxidation experiments were started by dropping H2O2 solutions into the mixture by a separatory funnel. The flow rate was controlled at 2 mg L1 min1 and lasted 10 min in the whole oxidation process. At the given reaction time intervals, 1 mL sample was withdrawn. 10 mL 1 M tert-butanol was immediately added into the sample as reaction inhibitor. Then the samples were filtered by a syringe filled with a little silanized glass wool. The concentrations of BPA and intermediates in the filtrate were measured by high-performance liquid chromatography (HPLC). If no specific instructions are given, initial pH of Fenton-like oxidation is 3.0 0.1, nanoparticles loading is 0.8 g L1 and Ag content in FeeAg composite material is 1 wt.%.
2.5. Tetrabromobisphenol A, bisphenol A and tert-butanol were obtained from SigmaeAldrich Company. H2O2 (30%, v/v) was purchased from Fisher Company. AgCl, FeSO4$7H2O, Fe2(SO4)3, H2SO4, NaOH, Na2SO3, 1,10-phenanthroline and ferrous ammonium sulfate were provided by Nanjing Chemical Company. HPLC-grade methanol and dichloromethane were purchased from Tedian Company and used without further purification. Milli-Q water was used throughout this study. The zero valent iron used was iron powder (Shenzhen Junye Nano Material Co., Ltd, >99.9％, <60 nm).
Reduction experiment of TBBPA
Analytical methods
The concentrations of TBBPA were analyzed via HPLC (Agilent 1200, USA), with a C18 reversed-phase column (150 mm 4.6 mm, 5 mm particles, Agilent, USA). Identification of reductive debromination products was performed by LC-ESI-MS/MS (Thermo LCQ Advantages, QuestLCQ Duo, USA) equipped with electrospray ionization with Beta Basic-C18 HPLC column (150 mm 2.1 mm id, 5 mm Thermo, USA). The specific operation conditions of HPLC and LC-ESI-MS/MS were provided in the literature (Luo et al., 2010). The HPLC operation conditions of BPA were as follows: the mobile phase was 30% water in methanol. At the detection wavelength of 226 nm and flow rate of 1.0 mL min1, the BPA retention time was 3.4 min. Since the oxidation products were so complex, a Thermo Finnigan Trace gas chromatography interfaced with a Polaris Q ion trap mass spectrometer (GC/ MS, Thermo, Finnigen, USA) equipped with DB-5 fused-silica capillary column (30 m 0.32 mm i.d, 0.25 mm film thickness) was used for analyzing the samples. Prior to GCeMS analysis, the samples were extracted with dichloromethane for three times. The extracted solution was dehydrated using anhydrous sodium sulfate and concentrated to 1 mL by rotary evaporation. After the solvent was blown away by the gentle nitrogen, trimethylsilylation was carried out at 50 C for 30 min using 0.2 mL of bis(trimethylsilyl)trifluoroacetamide (BSTFA). The initial temperature of the column oven was 40 C and following 1 min hold at this temperature, and then increased up to 300 C with a heating rate of 6 C min1.
1521
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 1 9 e1 5 2 8
Helium was used as the carrier gas. Mass spectrometric detection was operated with 70 eV electron impact (EI) mode. The degradation products were also analyzed by LC-ESI-MS/ MS. The injection volume was 20 mL, and the mobile phase was methanol to water (70:30). The MS analysis was conducted with negative-ionization mode with electrospray interface (ESI) source. Full scanning analyses were performed by scanning m/z range from 50 to 600 in profile mode. The dissolved bromide ion was determined using a Dionex ion chromatograph (IC, Dionex model ICS 1000) equipped with a dual-piston (in series) pump, a Dionex IonPac AS11-HC analytical column (4 mm 250 mm) and a Dionex DS6 conductivity detector. Suppression of the eluent was achieved with a Dionex anion ASRS 300 electrolytic suppressor (4 mm) in the auto suppression external water mode. The concentration of dissolved iron and ferrous ion was measured by the o-phenanthroline colorimetric method (l ¼ 510 nm, e ¼ 1.1 104 M1 cm1). Ferric concentration calculated by subtracting dissolved iron with ferrous concentration.
3.
Results and discussion
3.1.
Characterization
Detailed morphology and structure characterization of the FeeAg particles were presented in the literature (Luo et al., 2010). Fig. 1 is the XRD spectra of the fresh and reduced FeeAg particles. The XRD patterns of aged samples show peaks associated with iron oxides, indicating that surface of the used reductant is covered by a layer of oxide film which may form during the reduction process. The XPS survey scans
Fig. 1 e XRD of Fe-Ag bimetallic nanoparticles. (a) Fresh samples, (b) Reacted samples.
Table 1 e Composition of surface elements. Element
Ag 3d Fe 2p O 1s C 1s
XPS surface composition (atom %) Before
Reacted
0.57 40.59 36.11 22.73
0.53 24.69 53.88 20.91
of Fe and Ag over the surface of the FeeAg nanoparticles are described in Fig. SM-2. Table 1 shows elements (iron, silver and oxygen) composition of the samples before and after reaction. After reduction, iron oxide layer is adopted a larger amount of oxygen which indicated the structure of the particles with iron oxide in the outer of the sphere and ZVI in the inner.
3.2. Fenton-like oxidation of TBBPA in a heterogeneous FeeAg/H2O2 system In order to investigate the Fenton-like oxidation effect of TBBPA in a heterogeneous FeeAg/H2O2/US system, experiments were conducted in an 150 mL conical bottle with 5 mg L1 TBBPA and 0.8 g L1 FeeAg nanoparticles. The other conditions were as the same as the Fenton-like oxidation of BPA, which was mentioned in the Section 2.4. A typical reaction profile of TBBPA in FeeAg/H2O2/US system is given in Fig. 2. The result shows that more than 30% of TBBPA was removed in 2 min and then the degradation rate decreased. Up to 40% of TBBPA disappeared as the reaction was prolonged to 30 min. To gain more chemical structure information on the reaction products, reaction solution was subsequently extracted and possible polar products were silylated with BSFTA, and then subjected to GCeMS analysis. The detailed data are provided in the Supplemental Material. The results show that the main TBBPA oxidation products are brominated phenol species. In fact, there is strong
Fig. 2 e Degradation of TBBPA in aqueous solution by Fe-Ag/H2O2/US. [Fe-Ag bimetallic] [ 0.8 g LL1; flow rate of H2O2 [ 2 mg LL1 minL1, lasted 10 min; [TBBPA]0 [ 5 mg LL1; US conditions: 40 kHz, 100 W.
1522
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 1 9 e1 5 2 8
Fig. 3 e (a) Temporal disappearance of TBBPA and appearance of by-products in aqueous solution during FeeAgeUS reduction treatment. [Fe-Ag bimetallic] [ 0.8 g LL1, Ag content (wt %) [ 1%, and [TBBPA]0 [ 5 mg LL1. (b) Temporal change of BPA and its hydroxylated products concentration in the solution during Fe-Ag/H2O2/US oxidation treatment. [Fe-Ag bimetallic] [ 0.8 g LL1; flow rate of H2O2 [ 2 mg LL1 minL1, lasted 10 min; [BPA]0 [ 4.64 mg LL1; US conditions: 40 kHz, 100 W.
evidence that halogen, when placed in the organic moleculecarbon group, could increase the bio-toxicity of the organic compound. Bromine, due to its weaker electronegativity, could be considered better leaving groups and hence should be more toxic (Lag et al., 1994; DeWeese and Schultz, 2001; Huang et al., 2007). Several studies have presented the results from ecotoxicologic investigations of the brominated aromatic compounds, including brominated benzenes, brominated phenols and indoles, 2-bromo- hydroquinone and ska, 1998; Bruchajzer et al., 2002; Reineke et al., so on (Szyman 2006). Therefore, the Fenton-like oxidation of TBBPA in a heterogeneous FeeAg/H2O2 system is unsuitable and inadequate. Considering that the primary step in the degradation of halogenated compounds is the removal of halides from the
aromatic ring, such a T-SRO process mentioned above is proposed for the treatment of TBBPA.
Fig. 4 e Degradation of BPA in aqueous solution by Fe-Ag/ H2O2 with and without ultrasonic radiation. [Fe-Ag bimetallic] [ 0.8 g LL1; flow rate of H2O2 [ 2 mg LL1 minL1, lasted 10 min; [BPA]0 [ 5 mg LL1; US conditions: 40 kHz, 100 W.
Fig. 5 e Degradation of BPA in aqueous solution by Fe-Ag/ H2O2/US with and without addition of dissolved iron. [Fe-Ag bimetallic] [ 0.8 g LL1; flow rate of H2O2 [ 2 mg LL1 minL1, lasted 10 min; [BPA]0 [ 5 mg LL1; [Fe2D] [ 1.26 mg LL1, [Fe3D] [ 0.37 mg LL1; US conditions: 40 kHz, 100 W.
3.3. TBBPA debromination by FeeAg bimetallic nanoparticles Fig. 3 (a) shows the time course profiles for disappearance of TBBPA and appearance of by-products in aqueous solution over FeeAg bimetallic nanoparticles coupled with US radiation. The initial TBBPA concentrations and catalysts amount were chosen based on results of previous reduction experiments (Luo et al., 2010). In neutral conditions (pH ¼ 6.94), FeeAg bimetallic nanoparticles completely degraded TBBPA (5 mg L1) in 20 min coupled with US radiation.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 1 9 e1 5 2 8
Along with the decrease of TBBPA concentration, almost 100% of TBBPA was gradually debrominated to lowly brominated compounds. The released bromine was in hydrogen bromine form. TBBPA might be transformed into tri-BBPA and di-BBPA, then tri-BBPA and di-BBPA were dehalogenated to
1523
mono-BBPA, at last, BPA was generated by the debromination reaction of these three intermediates. The BPA concentration increased stably and continuously in the all 70 min and reached 4.64 mg L1 at the end of reduction process. The carbon mass balance at the end of the experiment (calculated
Fig. 6 e (a) Total ions chromatogram of BPA in the solution during Fe-Ag/H2O2/US oxidation treatment. (b) LC-MS/MS analysis of BPA degradation intermediates.
1524
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 1 9 e1 5 2 8
as the sum of all organic species measured) was approximately 94% of the calculated initial TBBPA concentration. We suspect that minor losses may have occurred during filtering through the silanized glass wool.
3.4. Fenton-like oxidation of BPA in a heterogeneous FeeAg/H2O2/US system Experiments of BPA control, FeeAg nanoparticles alone and H2O2 alone were also carried out in 150 mL conical bottles under US radiation, and the initial concentration of BPA was 5 mg L1. The results in Fig. 3 (b) indicate that no obvious removal of BPA was observed in the control and FeeAg nanoparticles alone experiments throughout 30 min. In the H2O2 alone degradation experiment, 15% degradation of BPA was observed after 30 min reaction. It means that the hydroxyl radicals generated by ultrasounds radiation could destroy the BPA with adding H2O2. The effect of US radiation will be discussed in detail in Section 3.4.1 As shown in Fig. 3 (b), after 2 mg L1 min1 of H2O2 (lasted 10 min) was added to the BPA solution debrominated from TBBPA, BPA concentration rapidly declined from 4.64 to 0.11 mg L1 over 30 min with removal efficiency of 99.2%. Concentration change profile of BPA derivatives is also given in Fig. 3 (b) in terms of the HPLC peak areas of the corresponding
compounds versus the charge applied to the system. The typical chromatogram of BPA during the oxidation in 30 min is depicted in Fig. SM-4. It can be seen that peak I corresponded to the parent compound BPA and two main intermediates peaks appeared (II and III). Peak II had a retention time of 2.7 min, which increased at first but decreased quickly after radiation for 6 min. Peak III had a retention time of 2.5 min, which appeared after reaction for 8 min and decreased afterward. These two intermediates were the products for the Fenton-like oxidation of BPA, whose decrease implied that further oxidation of BPA continued in solution.
3.4.1. Effect of ultrasonic radiation on BPA Fenton-like oxidation It is all well known that introduction of US can evidently improve pollutant degradation in Fenton system (Ioan et al., 2007; Namkung et al., 2008). When US were introduced into Fenton-like reaction, the mechanism would be more complex. The results of oxidation experiments in the presence and absence of US radiation are compared and presented in Fig. 4. Under US radiation, an obvious increase of BPA degradation efficiency in the solution was observed. Thus, there was a synergistic effect between US and Fenton-like reactant, which could improve the degradation of BPA in the heterogeneous FeeAg/H2O2/US system.
Fig. 7 e GC/MS analysis of BPA degradation intermediates and final products.
1525
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 1 9 e1 5 2 8
In general, introduction of US radiation into a heterogeneous system could produce both chemical and physical effect. The possible reasons for synergistic effect between US and Fenton-like reactant are as follows. Firstly, US could increase the OH radical concentration by US cavitation. Guo and Feng (2009) described that the US elimination of BPA in aqueous solution is mainly attributed to the OH radical oxidation. BPA was decomposed by OH radicals principally at the bubble-bulk solution interface in absence of FeeAg nanoparticles. In FeeAg/H2O2/US Fenton-like oxidation experiments, US cavitation increased the amount of OH radical and therefore accelerated the degradation reaction. Secondly, US shock wave could enhance dissolution of ferrous ion from iron surface. The main reason was ascribed to the removal or destruction of passivation films on the metal surface by cavitation effects (Tomlinson, 1990; Namkung et al., 2008). Thus, the increase of iron concentration in the solution might lead to enhance the oxidation of BPA. In addition, US radiation can transform BPA slightly with H2O2, as shown in Fig. 3 (b).
3.4.2. Effect of dissolved iron (Fe2þ and Fe3þ) on BPA Fenton-like oxidation To demonstrate the role of dissolved iron, ferrous ion and ferric ion were added to the FeeAg/US system with 5 mg L1 BPA. The concentrations of Fe2þ and Fe3þ in the solution after debromination were 1.26 mg L1 and 0.37 mg L1, respectively. These values were taken as the dissolved iron amount used for addition in the oxidation experiments for comparison. Fig. 5 shows that BPA decayed during the treatment by only FeeAg and FeeAg with dissolved iron Fenton-like process. In the two cases, BPA concentration was under the detection limit after 20 min, indicating that the oxidation capacity of the two systems (FeeAg and FeeAg þ dissolved iron) showed less difference throughout the reaction. However, in the first 2 min, 56% decomposition of BPA was achieved in the presence of Fe2þ and Fe3þ ion while only 22% BPA removal was observed in the absence of Fe2þ and Fe3þ ion. Zhou et al. (2008) reported that the two-stage kinetic of 4chlorophenol degradation in a heterogeneous ZVI/H2O2 system
Table 2 e Identification of the intermediates of BPA during the Fenton-like oxidation by GC/MS. Product
Rt (min)
m/z
Name
A
3.28
90
Oxalic acid
B
4.89
116
Maleic acid
C
5.67
108
p-Benzoquinone
D
6.97
94
phenol
E
7.45
90
2-hydroxypropanoic acid
Molecular structure HO
O
O
OH
O O HO
OH
O
O
OH
OH OH
F
13.01
92
propane-1,2,3-triol
G
14.72
134
4-isopropenylphenol
H
15.75
110
p-hydroquinone
I
19.20
136
4-hydroxyacetophenone
J
30.52
228
BPA
K
35.53
244
BPA-o-catechol.
OH
OH
HO
O
OH
OH
1526
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 1 9 e1 5 2 8
Scheme 1 e Main Fenton-like oxidation reaction pathway of BPA.
was composed of an initial slow degradation stage and a followed rapid degradation stage. In the first stage, the decomposition of H2O2 occurred on/near the ZVI surface, dissolving iron to ferrous, and Fenton reaction occurred concomitantly on/near surface in the presence of ferrous and H2O2. Based on this, the addition of dissolved irons species could increase the degradation rates of BPA, especially in the initial stage. As the reaction progresses, ferrous ion and ferric ion leached from iron surface and their concentrations increased gradually. Accordingly, addition of dissolved iron would not affect the removal efficiency in the entire Fenton-like process.
3.5. Identification of Fenton-like oxidation products of BPA The analyses of main BPA products formed during the Fentonlike oxidation were carried out using LC-MS/MS and GCeMS. The intermediates and products were identified by the analysis of mass spectra obtained from LC-MS/MS and/or GCeMS applications and by the comparison with the library data of NIST. Fig. 6 (a) is the total ion chromatogram of BPA solution throughout oxidation reaction, which shows three peaks with retention time at 12.69 min, 9.78 min and 8.23 min, respectively. Fig. 6 (b) is the mass spectra for them. It can be seen that the compound at 12.69 min was BPA (I) with m/z 227. The peaks labeled with II (9.78 min, m/z ¼ 243) and III (8.23 min, m/z ¼ 257) corresponded to the monohydroxylated BPA product and the dihydroxylated BPA product, respectively. The former gave base ion at 243 m/z on ESI negative mode detection, was BPA-ocatechol. This compound also showed other ion at 241 m/z, suggesting that BPA could be transformed via the following reversible equilibration between BPA-o-catechol (MW ¼ 244) and BPA-o-quinone (MW ¼ 242). These products obviously have a greater hydrophilicity than BPA, so their retention times were shorter under the HPLC condition in our experiment.
Other products in the Fenton-like oxidation, including 4isopropenylphenol and the open ring products, were detected by GCeMS (Fig. 7). This result is similar to that reported by Go¨zmen et al. (2003) in their studies on the electro-Fenton oxidation of BPA. After all the peaks in the chromatograms were carefully examined, up to 10 compounds were identified as possible intermediates. Detailed data are listed in Table 2. The intermediates with benzene ring could be further oxidized into small organic molecules, with the retention times at 3.28, 4.89, 7.47 and 13.01 min, most of which were organic acids, such as oxalic acid, and so on. However, some products such as HCOOH were not observed due to their low levels and unavoidable loss during the sample preparation course. Based on the intermediates and the results obtained by other researchers (Katsumata et al., 2004; Guo and Feng, 2009), the possible pathways of BPA oxidation were represented in Scheme 1. The expatiation of these pathways is presented in the Supplemental Material. In addition, the brominated compounds were not present in the Fenton-like degradation products. After reduction and oxidation processes, the amount of dissolved Br measured by IC was 2.5 mg L1 and 2.3 mg L1, respectively, which showed that the variation of Br concentration in the solution was negligible in Fenton-like experiments. Based on analysis described above, we referred that the debromination of TBBPA was to release HBr and the bromination of decomposition products by bromine that was not formed even in the present of OH radicals. In other words, the bromine generated from TBBPA reduction did not lead to the formation of secondary brominated products by interaction with the primary oxidation products of BPA.
4.
Conclusions
The degradation of TBBPA in aqueous solution was investigated by Fe-Ag reduction and Fenton-like oxidation under US
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 1 9 e1 5 2 8
radiation. TBBPA was completely debrominated to BPA after 70 min over FeeAg bimetallic nanoparticles. For BPA, removal efficiency of 99.2% was achieved after 30 min under the optimum conditions. The US radiation enhanced the degradation rate of BPA as compared to Fenton-like only. In addition, dissolved iron was found to increase the rate of the first 2 min oxidation reactions, however, the existence of Fe2þ and Fe3þ did not affect entire Fenton-like degradation of BPA. This work shows that it is feasible for T-SRO treatment to completely decompose TBBPA in aqueous solutions. GCeMS and LC-MS/MS techniques were used to identify the oxidation intermediates in order to gain a deeper insight of the reaction mechanism. The main intermediates include monohydroxylated BPA, 4-isopropenylphenol, p-hydroquinone, 4-(1-hydroxy-1-methyl-ethyl)-phenol, 4-hydroxyacetophenone and phenol. Besides, OH radical-mediated oxidation was found to be the major destruction pathway during BPA decomposition. Future work will need to be done to improve the potential mineralization of BPA, as well as the fate of the organic molecules to confirm mineralization.
Acknowledgements The authors greatly acknowledge the National Natural Science Foundation of China (20707009), National Major Project of Science & Technology Ministry of China (NO. 20082X07421-002) for financial support, and National Natural Science Foundation of China (50938004).
Appendix. Supplementary data Supplementary data associated with the article can be found in online version, at doi:10.1016/j.watres.2010.10.039.
references
Barontini, F., Cozzani, V., Marsanich, K., Raffa, V., Petarca, L., 2004. An experimental investigation of Tetrabromobisphenol A decomposition pathways. Journal of Analytical and Applied Pyrolysis 72 (1), 41e53. Bremner, D.H., Burgess, A.E., Houllemare, D., Namkung, K.C., 2006. Phenol degradation using hydroxyl radicals generated from zero-valent iron and hydrogen peroxide. Applied Catalysis B-Environmental 63 (1e2), 15e19. Bruchajzer, E., Szymanska, J.A., Piotrowski, J.K., 2002. Acute and subacute nephrotoxicity of 2-bromophenol in rats. Toxicology Letters 134 (1e3), 245e252. Cwiertny, D.M., Bransfield, S.J., Livi, K.J.T., Fairbrother, D.H., Roberts, A.L., 2006. Exploring the influence of granular iron additives on 1,1,1-trichloroethane reduction. Environmental Science & Technology 40 (21), 6837e6843. DeWeese, A.D., Schultz, T.W., 2001. Structure-activity relationships for aquatic toxicity to Tetrahymena: halogensubstituted aliphatic esters. Environmental Toxicology 16 (1), 54e60. de Wit, C.A., 2002. An overview of brominated flame retardants in the environment. Chemosphere 46 (5), 583e624.
1527
˚ , Jakobsson, E., Eriksson, J., Rahm, S., Green, N., Bergman, A 2004. Photochemical transformations of Tetrabromobisphenol A and related phenols in water. Chemosphere 54 (1), 117e126. Go¨zmen, B., Oturan, M.A., Oturan, N., Erbatur, O., 2003. Indirect electrochemical treatment of bisphenol A in water via electrochemically generated Fenton’s reagent. Environmental Science & Technology 37 (16), 3716e3723. Guo, Z.b., Feng, R., 2009. Ultrasonic irradiation-induced degradation of low-concentration bisphenol A in aqueous solution. Journal of Hazardous Materials 163 (2e3), 855e860. ˚ , Jenssen, D., 1999. Helleday, T., Tuominen, K.L., Bergman, A Brominated flame retardants induce intragenic recombination in mammalian cells. Mutation Research 439 (2), 137e147. Huang, C.P., Wang, Y.J., Chen, C.Y., 2007. Toxicity and quantitative structure-activity relationships of nitriles based on Pseudokirchneriella subcapitata. Ecotoxicology and Environmental Safety 67 (3), 439e446. Ioan, I., Wilson, S., Lundanes, E., Neculai, A., 2007. Comparison of Fenton and sono-Fenton bisphenol A degradation. Journal of Hazardous Materials 142 (1e2), 559e563. Kaiser, J., 2000. Endocrine disrupters: panel cautiously confirms low-dose effects. Science 290 (5492), 695e697. Katsumata, H., Kawabe, S., Kaneco, S., Suzuki, T., Ohta, K., 2004. Degradation of bisphenol A in water by the photo-Fenton reaction. Journal of Photochemistry and Photobiology A: Chemistry 162 (2e3), 297e305. Lag, M., Omichinski, J.G., Dybing, E., Nelson, S.D., Soderlund, E.J., 1994. Mutagenic activity of halogenated propanes and propenes: effect of bromine and chlorine positioning. Chemico-Biological Interactions 93 (1), 73e84. Luo, S., Yang, S.G., Wang, X.D., Sun, C., 2010. Reductive degradation of Tetrabromobisphenol A over iron-silver bimetallic nanoparticles under ultrasound radiation. Chemosphere 79 (6), 672e678. Mackenzie, K., Kopinke, F.D., 1996. Debromination of duroplastic flame-retarded polymers. Chemosphere 33 (12), 2423e2430. Mantha, R., Biswas, N., Taylor, K.E., Bewtra, J.K., 2002. Removal of nitroaromatics from synthetic wastewater using twostep zero-valent iron reduction and peroxidase-catalyzed oxidative polymerization. Water Environmental Research 74 (3), 280e287. Mantha, R., Taylor, K.E., Biswas, N., Bewtra, J.K., 2001. A continuous system for Fe0 reduction of nitrobenzene in synthetic wastewater. Environmental Science & Technology 35 (15), 3231e3236. Meerts, I.A.T.M., van Zanden, J.J., Luijks, E.A.C., van Leeuwen˚ , Brouwer, A., 2000. Bol, I., Marsh, G., Jakobsson, E., Bergman, A Potent competitive interactions of some brominated flame retardants and related compounds with human transthyretin in Vitro. Toxicological Sciences 56 (1), 95e104. Monserrate, E., Haggblom, M.M., 1997. Dehalogenation and biodegradation of brominated phenols and benzoic acids under iron-reducing, sulfidogenic, and methanogenic conditions. Applied and Environmental Microbiology 63 (10), 3911e3915. Namkung, K.C., Burgess, A.E., Bremner, D.H., Staines, H., 2008. Advanced Fenton processing of aqueous phenol solutions: a continuous system study including sonication effects. Ultrasonics Sonochemistry 15 (3), 171e176. ¨ berg, K., Warman, K., O ¨ berg, T., 2002. Distribution and levels of O brominated flame retardants in sewage sludge. Chemosphere 48 (8), 805e809. Oh, S.Y., Chiu, P.C., Kim, B.J., Cha, D.K., 2003. Enhancing Fenton oxidation of TNT and RDX through pretreatment with zerovalent iron. Water Research 37 (17), 4275e4283.
1528
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 1 9 e1 5 2 8
Oh, S.Y., Chiu, P.C., Kim, B.J., Cha, D.K., 2005. Zero-valent iron pretreatment for enhancing the biodegradability of RDX. Water Research 39 (20), 5027e5032. Orth, W.S., Gillham, R.W., 1996. Dechlorination of trichloroethene in aqueous solution using Fe0. Environmental Science & Technology 30 (1), 66e71. Pulgar, R., Olea-Serrano, F., Villalobos, M., Rivas, A., Metzler, M., Pedraza, V., Olea, N., 1998. The estrogenicity of bisphenol A-related diphenylalkanes with various substituents at the central carbon and the hydroxy groups. Environmental Health Perspectives 106 (3), 167e174. Reineke, N., Biselli, S., Franke, S., Francke, W., Heinzel, N., Hu¨hnerfuss, H., Iznaguen, H., Kammann, U., Theobald, N., Vobach, M., Wosniok, W., 2006. Brominated indoles and phenols in marine sediment and water extracts from the north and baltic seas-concentrations and effects. Archives of
Environmental Contamination and Toxicology 51 (2), 186e196. Sellstro¨m, U., Jansson, B., 1995. Analysis of Tetrabromobisphenol A in a product and environmental samples. Chemosphere 31 (4), 3085e3092. ska, J.A., 1998. Hepatotoxicity of brominated benzenes: Szyman relationship between chemical structure and hepatotoxic effects in acute intoxication of mice. Archives of Toxicology 72 (2), 97e103. Tomlinson, W.J., 1990. Effect of ultrasonically induced cavitation on corrosion. In: Mason, T.J. (Ed.), Advances in Sonochemistry, Vol. 1. Jai Press, London, pp. 173e195. Zhou, T., Li, Y.Z., Ji, J., Wong, F.S., Lua, X.H., 2008. Oxidation of 4-chlorophenol in a heterogeneous zero valent iron/H2O2 Fenton-like system: kinetic, pathway and effect factors. Separation and Purification Technology 62 (3), 551e558.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 2 9 e1 5 3 8
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Extracellular polymeric substances diversity of biofilms grown under contrasted environmental conditions Monique Ras a, Dominique Lefebvre a, Nicolas Derlon b,c,d, Etienne Paul b,c,d, Elisabeth Girbal-Neuhauser a,* a
LBAE, Laboratoire de Biologie applique´e a` l’Agro-alimentaire et a` l’Environnement, Institut Universitaire de Technologie, Universite´ Paul Sabatier Toulouse III, 24 Rue d’Embaque`s, 32000 Auch, France b Universite´ de Toulouse; INSA, UPS, INP; LISBP, 135 Avenue de Rangueil, F-31077 Toulouse, France c INRA, UMR792 Inge´nierie des Syste`mes Biologiques et des Proce´de´s, F-31400 Toulouse, France d CNRS, UMR5504, F-31400 Toulouse, France
article info
abstract
Article history:
Extracellular Polymeric Substances (EPS) analysis was undertaken on three biofilms
Received 2 September 2010
grown under different feeding conditions and offering diverging microbial activities and
Received in revised form
structural characteristics. EPS were extracted by a multi-method protocol including soni-
15 November 2010
cation, Tween and EDTA treatments and were characterized by size exclusion chroma-
Accepted 15 November 2010
tography (SEC). Tween and sonication extracts presented higher EPS size diversity
Available online 24 November 2010
compared to EDTA extracts. EPS size diversity also increased with microbial functions
Keywords:
biofilms presenting autotrophic activity. Another specific size cluster (180 kDa) occurred in
Microbial biofilm
Tween extracts provided from the mechanically stable biofilms. Such specific EPS appear
Autotrophic
as potential indicators for describing microbial and structural properties of biofilms.
Extracellular polymeric substances
This study brings new elements for designing EPS fractionation and shows that size
Size distribution
distribution analysis is an interesting tool to relate EPS diversity with macro-scale char-
Extraction strategy
acteristics of biofilms.
within the biofilms and a specific 25e50 kDa cluster was identified only in extracts from
ª 2010 Elsevier Ltd. All rights reserved.
1.
Introduction
Biofilms are described in literature as fixed micro-organisms on an interface and immobilized in a matrix of extracellular polymeric substances (or EPS) of microbial origin. The stable environment offered by the EPS matrix cradles the development of a large span of microbial communities of which several can be deleterious. The microbial heterogeneity of biofilms can also be of great interest in the environmental sector since such concentrated and diversified microbial activities can be beneficially exploited for treating organic and inorganic water pollutants. However, municipal wastewater
treatment facilities generally use suspended floc forming biomasses which are often washed out from the system (Liu et al., 2004; Matsumoto et al., 2007) and hence experience low microbial diversity functions. Fixed biomass such as biofilms can prevent such losses by retaining bacterial diversity inside the system and particularly slow-growing bacterial populations, such as nitrifiers. Such configurations can hence increase the treatment efficiency. The EPS matrix is often stated as consolidating material for the entire biofilm. Indeed, the extracellular compartment can reach 98% of the total organic carbon fraction of biofilms (Jahn and Nielsen, 1998). EPS compounds are excreted by the
* Corresponding author. Tel.: þ33 5 62 61 28 13; fax: þ33 5 62 61 63 01. E-mail address:
[email protected] (E. Girbal-Neuhauser). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.11.021
1530
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 2 9 e1 5 3 8
microbial population, but can also result from natural cell lysis or from hydrolytic activities. A wide variety of polymers are reported within the matrix, where a major proportion is attributed to proteins and polysaccharides, while lipids and nucleic acids are rather found in minor proportions (Azeredo et al., 1999; Jahn and Nielsen, 1995). The influence of environmental conditions on the composition of EPS compounds has already been suggested in literature (Branda et al., 2005). Regarding the use of carbon and nitrogen elements for EPS production, the carbon/nitrogen ratio of the influent is liable to impact the type of produced EPS, i.e. carbohydrate and protein production (Durmaz and Sanin, 2001; Li et al., 2008). In addition, the carbon/nitrogen ratio can specify the microbial ecology of the biofilm (Ohashi et al., 1995), by promoting either heterotrophic growth (high ratio) or autotrophic microorganisms (low ratio). Regarding the biochemical responses to environmental and microbial parameters, characterizing the EPS fraction of biofilms could thus be a relevant procedure for describing relations between EPS and biofilm structure. In literature, studies undertaken on molecular characterization of EPS in biofilms are few. Under axenic conditions, exopolysaccharides and more specifically uronic acids containing polymers extracted from biofilms are described as essential for providing the matrix framework through strong anionic interactions (Chen and Stewart, 2002; Davies et al., 1993). For multi-species biofilms, the conditions are even more complex since a wide range of other molecular interactions have to be considered. Proteins are characterized by ionic, hydrophobic and neutral amino-acids and a large range of chemical interactions (electrostatic, hydrophobic and low energy hydrogen bonds) are able to link proteins to the biological matrix (Mayer et al., 1999). Proteins also include functionally active enzymes which take part in the production and degradation of the matrix. Therefore, inherent chemical properties of EPS and especially proteins can offer qualitative information on both physical and dynamic properties of the biofilm. However, the structural heterogeneity and complex functional properties in environmental associated biofilms make EPS characterization somewhat difficult. Several analytical methods including physical and chemical techniques are used (Denkhaus et al., 2007) but with care depending on the aim of investigation as well as the type of studied biofilm. Microscopic and optical methods which involve EPS staining techniques are not always appropriate for visualizing these components in thick and complex biofilms due to light attenuation or probe penetration problems. Infrared spectroscopy is also widely used in biofilm analysis with similar limitations relative to the penetration capacity of the IR radiations (Boualam et al., 2002). Considering these technical restrictions, molecular characterization of complex biofilms can be achieved by extracting the EPS from the biofilm and then characterizing the soluble extract by chromatography or electrophoresis separation methods. Although widely used on activated sludge samples (Comte et al., 2007; Garnier et al., 2005), this molecular scale investigation strategy was never applied for biofilm EPS characterization. Molecular weight (MW) distributions of extracted EPS can offer global characteristics of the sample and has been suggested as a useful tool for fingerprint identification. Garnier
et al. (2005) evidenced different MW profiles depending on the origin of activated sludge. Authors showed that proteins where generally found in the high MW fractions (10e600 kDa) while sugars were rather found in the lower MW fractions (1 kDa). However, studying the size distribution of EPS in complex bacterial aggregates reveals to be tricky since such analysis implies prior extraction methods which can affect not only the proportion of extracted EPS (Ras et al., 2008a; Zhang et al., 1999) but also the qualitative aspect of these polymers (Comte et al., 2007; Simon et al., 2009). The present paper explores EPS size distributions within biofilms in order to figure out specific molecular characteristics which could explain particular biofilm biological and/or physical properties. In order to validate such an approach, the investigated biofilms were grown under contrasted environmental conditions to promote diverging microbial activities within each biofilm. A multi-method protocol previously described for extracting EPS from activated sludge (Ras et al., 2008a) was used to sample EPS compounds from the biofilms. This protocol, based on mechanical, hydrophobic and ionic extraction methods, offers a globally diversified EPS extract which can be consistent of the studied biofilms. The distribution of EPS contents as well as EPS molecular weight profiles were investigated in order to relate specific molecular EPS characteristics to biofilm growth conditions and/or microbial populations. The impact of EPS extraction procedures on this molecular fingerprint diagnosis was also considered. According results are expected to help improve knowledge on biofilm growth control which is lacking in the wastewater and water distribution sectors.
2.
Methods
2.1.
Experimental setup
Three biofilms were grown in hydrodynamic controlled Couette Taylor reactors as described by Coufort et al. (2007). For a fixed gap between the two concentric cylinders, the rotational speed of the inner cylinder was fixed in order to have a wall shear stress of 0.5 Pa during the growth period. Biofilms grew on 25 polyethylene plastic plates (100 50 5 mm) distributed around the external cylinder.
2.2.
Biofilm growth conditions
A mixed carbon source composed of ethanol, propionic acid, glucose and sodium acetate was used as organic substrate for the development of the biofilms. Reactors were inoculated with conventional activated sludge sampled from the aeration tank of a local municipal wastewater treatment plant. Two biofilms were developed under organic substrate-limiting conditions and with a constant surface loading rate of 2.5 g COD m2 d1 (COD: Chemical Oxygen Demand). In order to obtain either a heterotrophic biofilm (B1) or a mixed autotrophic/heterotrophic biofilm (B2), the feed diverged in COD/ NH4eN ratios. The feed for B1 was fixed at 73 g COD g1 NeNH4 (9.5 mg NH4eN L1) and the feed for B2 at 4 g COD g1 NH4eN (175 mg NH4eN L1). For these two cases, the oxygen concentration in the bulk liquid was kept constant at a value
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 2 9 e1 5 3 8
closed of the oxygen saturation concentration. A third biofilm (B3) was grown under a high substrate loading rate of 25 g COD m2 d1 and with a ratio of 4 g COD g1 NH4eN (175 mg NH4eN L1). In this case, the oxygen concentration was kept constant between 6 and 7 mg O2 L1, inducing oxygen-limiting growth conditions. During the overall characterization period, ammonia (NHþ 4 ), nitrite (NO2 ), nitrate (NO3 ) and COD were measured daily in the inlet and in the outlet of the Couette Taylor reactors. The ammonia concentration was measured using the Nessler method, nitrite and nitrate concentrations were determined by spectrometry and the COD was obtained with the method based on the oxidation by potassium dichromate (Standard Methods, 1995). Biofilm at steady state was defined as a biofilm characterized by stable COD removal, nitrification and denitrification rates. Stable COD removal, nitrification and denitrification and thus steady state were usually reached after 60 days of biofilm development.
2.3.
Biofilm characterization
The average biofilm thickness was measured by image analysis as described in Coufort et al. (2007). Mean accumulated mass was measured after biofilm detachment from the polyethylene plastic plates by gentle scraping and suspension in the liquid reactor. Detached biomass was then recovered by centrifugation (1500g; 15 min) and measured in terms of Suspended Solids and Volatile Suspended Solids concentration (g VSS L1) according to the standard procedures (Standard Methods, 1995). Total COD removal, nitrification efficiency and denitrification efficiency were evaluated by comparing the inlet and outlet values of COD, ammonium, nitrate and nitrite concentrations.
2.4.
EPS extraction by the multi-method protocol
Bound EPS were extracted according to the previously described multi-method protocol validated on activated sludge samples (Ras et al., 2008a). Biofilm samples were centrifuged (10 000g; 20 min) and pellets were washed twice in Phosphate Buffer Saline (PBS) pH 7. Each biofilm sample was subdivided in three 10 mL aliquots containing around 5 g VSS L1 for triplicate extractions. One protocol involved three extraction methods in sequence: sonication (3 2 min in PBS), Tween (0.25% in PBS, 1 h) and then EDTA (2% in Tris-HCl 0.3 mol L1, pH 8.5, 1 h), with intermediate centrifugation steps (10 000g; 20 min). EPS extracts were measured the same day for protein and polysaccharide contents as well as for G6P-DH activity, and then stored at 20 C for further analysis. The protocol extraction efficiency was evaluated after repeating three times the protocol sequence on the same biofilm sample. The decrease of the protein content recovered after each protocol sequence fitted an exponential curve as described in Ras et al. (2008a). The total protein content in biofilm extracts obtained by repeating the extraction protocol reached 246 mg eq. BSA g1 VSS, with 116 mg eq. BSA g1 VSS obtained by applying the protocol only once (results not shown). The extraction yield performed on the biofilm was
1531
hence 47%, which is similar to yields obtained on activated sludge samples (Ras et al., 2008a). Protein measurements were performed on all soluble extracts from B1, B2 and B3 biofilms using the Bicinchoninic Acid (Smith et al., 1985) or BCA reagent (SigmaeAldrich), according to Ras et al. (2008b) procedure. This quantification method was chosen according to its better tolerance towards chemicals used during extraction compared to modified Lowry method (Ras et al., 2008a). Bovine Serum Albumin (BSA) was used as standard. Polysaccharide concentrations were determined using the Anthrone method (Dreywood, 1946). Glucose was used as standard. Each measurement was undertaken on duplicate samples.
2.5.
Cell lysis control
The activity of the intracellular G6P-DH was measured according to Ras et al. (2008a). Enzyme substrate solution was prepared with 0.2 M Tris-HCl pH 8.5, 0.2 M 2-mercaptoethanol (Acros), 0.0005 M Nicotine Adenine Dinucleotide (NAD, Acros) and 0.01 M D-glucose-6-phosphate (Fluka). Enzyme activity was evaluated after incubating 200 mL of sample with 800 mL of the enzyme substrate solution at room temperature and measuring NADH production at 340 nm during 30 min G6P-DH activity was expressed as units (U) per mg of VSS, one unit corresponding to the number of nmol of NADH produced per min in the assay conditions. In order to correlate the G6P-DH activities measured in the extracts or in the whole biofilms with a number of lysed cells, a preliminary calibration was performed using Cupriavidus necator DSM 545 suspensions. C. necator was cultured as previously described by Ramsay et al. (1990). Briefly, the culture was first grown for 12 h in a liquid Nutrient Broth medium (Merck) under agitation (200 rpm) and at 30 C. 10 mL of the suspension was then inoculated to 150 mL of a Mineral Medium supplemented with glucose (10 g L1) and incubated for 12 h at 30 C at 100 rpm. Every 4 h, 10 mL of a culture medium sample was filtered on a cellulose 0.2 mm filter then dried and weighed in order to determine the total biomass concentration. Bacterial population was also evaluated by serial dilution of the samples and numeration on TCA agar plates: a value of 4.08 106 g of dry biomass per 106 cells was determined. After 12 h, bacteria were harvested by centrifugation (10 000g, 10 min) and resuspended in a equal volume of TES buffer (Tris-HCl 50 mM pH 8, EDTA 0.29 g L1, saccharose 25%). Cell lysis was then induced by adding 50 mL of a lysozyme solution at 47 000 U/mg (SigmaeAldrich) to 1 mL of the TES bacterial suspension. After 1 h at 37 C, numeration was performed on the suspension and the G6P-DH measured on the supernatant. Data obtained on three independent samples indicated that 0.2 U were released per 106 disrupted cells, also corresponding to 49,020 U per g of dry cells.
2.6.
Chromatography analysis
Chromatography was performed using a high-performance liquid chromatography system (AKTA Purifier, GE Healthcare) equipped with a 1 mL injection loop, a UV detector and a conductivity cell. Size exclusion chromatography (SEC) used
1532
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 2 9 e1 5 3 8
a 24 mL sepharose gel filtration column (Superose 6, GE Healthcare). Elution was carried out at room temperature using PBS at constant 1 mL min1 flow rate. According to manufacturer information, size fractionation is performed between 5 and 5000 kDa. Calibration of the column was undertaken by injecting ten different size protein standards (high and low molecular weight calibration kits GE Healthcare: aprotinin (6500 Da), ribonuclease (13 700 Da), chymotrypsin (25 000 Da), carbonic anhydrase (29 000 Da), bovine serum albumin (67 000 Da), conalbumin (75 000 Da), aldolase (158 000 Da), catalase (232 000 Da), ferritin (440 000 Da) and thyroglobulin (669 000 Da)). The calibration curve revealed the following equation: log (MW) ¼ 0.2939V þ 9.8481 with Molecular Weight (MW) expressed in Da and the elution volume V in mL. The total exclusion volume was determined after injection of Blue Dextran 2000 (GE Healthcare, 2$106 Da) and was found at 8 mL. Chromatogram profiles were recorded with UNICORN 5.1 software (GE Healthcare). Peak retention times and peak areas were directly calculated and delivered by the program.
3.
Results
3.1.
Global characteristics of the developed biofilms
Three biofilms (B1, B2 and B3) were developed under different feeding conditions in terms of COD/NH4eN ratios as well as surface organic loading rates. These experimental conditions were chosen according to previous results which reported the influence of growth conditions on biofilm structure and biological activity (Coufort et al., 2007; Derlon et al., 2008; Wijeyekoon et al., 2004). The B1 biofilm grew under a high COD/NH4eN ratio of 73 (nitrogen limitation) while B2 and B3 biofilms grew under a low COD/NH4eN ratio of 4 (excess nitrogen). This carbon/nitrogen ratio varied by modifying ammonium concentration in the feed. Both B1 and B2 biofilms were grown under a low surface loading rate of 2.5 g COD m2 per day, while B3 biofilm received a high surface loading rate of 25 g COD m2 per day. Physical measurements (Table 1) and microscopic observations (Fig. 1) of all biofilms were undertaken when steady state COD and nitrogen removal rates were reached. B1 and B3 biofilms were characterized by a particularly thick and filamentous structure in opposition to the thin and denser aspect of
Fig. 1 e Microscopic side views of B1 (A), B2 (B) and B3 (C) biofilms. S: Substratum ; B: Biofilm.
B2 biofilm (Fig. 1). B1 and B2 biofilms were fed under a low organic load and exposed a homogeneous colonization over the surface plates. B3 biofilm, on the other hand, was fed under a high organic load and experienced sloughing events which caused partial colonization of the surface plates. Thickness measurements were difficult to proceed on B3 biofilm due to the strong surface heterogeneity (values ranging from 0.5 to 4 mm). B1 accumulated more biofilm mass (8.5 g VSS m2) compared to B2 (4.2 g VSS m2), and in spite of detachment events, B3 revealed the highest accumulated mass (16.6 g VSS m2) (Table 1).
3.2.
Biofilm microbial activities
Microbial activities were investigated in B1, B2 and B3 biofilms after reaching steady state conditions. Fig. 2 reveals that heterotrophic activity was found in all biofilms. However, carbon removal efficiencies were higher for B2 and B3 biofilms (respectively 93% and 97%) compared to B1 biofilm grown under a higher COD/NH4eN ratio (84%). Nitrogen removal activities where only be measured in the B2 and B3 biofilms grown under low COD/NH4eN ratios. Indeed, B2 and B3 biofilms performed simultaneous nitrification and denitrification activities, while B1 biofilm did not express any nitrogen removing activity. However, nitrification efficiency was found to be higher in B2 biofilm compared to B3 biofilm (85% versus 66%), and denitrification efficiency on the other hand was two fold higher in B3 biofilm compared to B2 biofilm (100% versus 50%). According to these results, B1 biofilm, fed on a high carbon/ nitrogen ratio, was identified as a single heterotrophic biofilm while both B2 and B3 biofilms, fed on low carbon/nitrogen
Table 1 e Physical and structural characteristics of B1, B2 and B3 biofilms. Growth conditions, colonization aspect, biofilm thickness, accumulated biomass and natural cell lysis measured on B1, B2 and B3 biofilms. Biofilm COD/TKN Surface loading (g COD m2 d1) Aspect Surface colonization Average biofilm thickness (mm) Mean accumulated mass (g VSS m2)
B1
B2
B3
73 2.5 Homogeneous Filamentous Complete 4.4 1.1 8.5
4 2.5 Homogeneous Dense Complete 1.6 0.4 4.2
4 25 Heterogeneousa Filamentous Partial 0.5e4a 16.6
a Heterogeneous biofilm thickness due to sloughing events.
1533
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 2 9 e1 5 3 8
Proteins (mg.gVSS )
A
500 400
US Tw EDTA TOTAL
300 200 100 0
Fig. 2 e Carbon removal (-), nitrification ( ) and denitrification (,) efficiencies measured in B1, B2 and B3 biofilms, when reached steady state conditions.
ratios, were identified as mixed autotrophic and heterotrophic biofilms performing simultaneous nitrification and denitrification. The developed biofilms appeared as mature and thick structures, potentially offering a large range of microbial populations. Biochemical properties of these three diverging biofilms were then investigated in terms of EPS contents and size characteristics.
Sugars (mg.gVSS )
B
B1
B2
B3
B1
B2
B3
150
100
50
0
C
6 5
3.3.
EPS content in biofilms
A multi-method extraction protocol, previously described for quantifying EPS from activated sludge (Ras et al., 2008a), was applied on each B1, B2 and B3 biofilm. The extraction protocol involved three different extraction methods (sonication, Tween and EDTA) applied sequentially on the same sample in order to collect a consistent fraction of EPS. Soluble extracts were harvested by intermediate centrifugation steps and quantified in terms of proteins and polysaccharides. Fig. 3 shows that protein contents in all extracts were systematically higher compared to polysaccharide contents, and thus independently of the applied extraction method (sonication, Tween or EDTA) as well as the biofilm (B1, B2 or B3). Extraction yields diverged between the applied methods, but revealed similar trends between the biofilms. Indeed, both protein and polysaccharide contents were always higher in the extracts obtained by EDTA and sonication steps, while Tween steps always appeared as the least efficient extraction method. Total EPS contents in each biofilm were defined by summing the amounts of proteins and polysaccharides obtained by each extraction method (sonication þ Tween þ EDTA). Fig. 3A and B show that B1 biofilm had the lowest amount of proteins (43 mg g1 VSS) and polysaccharides (15 mg g1 VSS) whilst protein and polysaccharide contents was four fold higher in B2 and 10 fold higher in B3 biofilms. As shown in Fig. 3C, protein/polysaccharide ratios in the various extracts varied between 1.8 and 5.4 but protein to polysaccharide ratio of the total extracted EPS were similar
P/S
4 3 2 1 0 B1
B2
B3
Fig. 3 e Protein content (A), Sugar content (B) and Protein to Sugar (P/S) ratio (C) in soluble extracts obtained by the multi-method extraction protocol. Soluble extracts were harvested after each extraction method (ultrasonic, Tween, or EDTA) and both proteins and sugars were assayed. Error bars are evaluated from doubled extractions and duplicate measurements. for both B1 and B2 biofilms (2.9 0.2) and slightly higher in B3 biofilm (3.7 0.2). In order to control potential cell lysis during the extraction procedure, G6P-DH activity was systematically measured in each soluble extract. The measured units obtained in each extract were added in order to evaluate the total released G6PDH activity per biofilm. Table 2 shows that some G6P-DH activity was detected in B2 and B3 but not in B1 biofilm extracts. G6P-DH units can be related to a number of disrupted cells and hence to a mass of organic cell compounds. This conversion is possible by using experimental correlation factors established with a C. necator culture (described in the Material and Methods). G6P-DH units measured in B2 extracts
1534
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 2 9 e1 5 3 8
G6P-DH activity (U g1 VSS) Number of eq. lysed cellsb (106 g1 VSS) Released cellular compoundsc (mg g1 VSS) Total extracted proteins and sugars (mg g1 VSS) Level of extract contamination by released cellular compoundsd (%)
B3 extract
0
82
2547
0
412
12735
0
2
52
58
217
539
0
0.8
9.7
B3 900
2
4
3.4.
Global EPS size distribution in biofilms
A global EPS fingerprint investigation was undertaken by pooling each EPS extract obtained from each extraction step of the protocol (sonication, Tween and EDTA), and this for B1, B2 and B3 biofilms individually. Fig. 4 shows the size distribution of the pooled fractions from each individual biofilm. Since proteins were predominant in all extracts (Fig. 3), the absorbance signal was chosen at 280 nm. Moreover, results are expressed in mAU g1 VSS in order to standardize the signal between each biofilm sample and to compare the relative predominance of size fractions between each other, by direct evaluation of their peak area. The column was previously calibrated by injecting standard size proteins, which led to a linear semi-logarithmic relation between molecular weight of proteins and elution volume (Fig. 4A). Chromatographic profiles obtained from the pooled extracts highlight qualitative and quantitative differences between B1, B2 and B3 biofilms. Nevertheless, three fractions occurred systematically in all biofilm profiles: (i) a high molecular weight fraction eluted inside the exclusion volume of the column (8 mL) indicating size fractions above 5000 kDa, (ii) an intermediate size fraction eluted between 20 and 22 mL, represented
100000 10000 1000
13 16 17 20 22 Elution volume (mL)
6
8 10 12 14 16 18 20 22 24 26 28 30 Elution volume (mL)
B
100% < 0.5 kDa (> 24 mL)
(5)
2 - 0.5 kDa (22 -24 mL)
(4)
60%
7 - 3 kDa (20-22 mL)
(3)
40%
25 - 20kDa (17 – 18 mL)
(2)
20%
> 5000 kDa (8 mL)
(1)
80%
0%
B1
are equivalent to 412$106 C necator disrupted cells, which is liable to the release of 1.7 mg of cellular components per g of biofilm VSS. Comparing this amount with the amount of proteins and sugars measured in the soluble extracts indicates that the multi-method protocol did not induce significant cell breakage in B2 biofilm since the level of contamination of the extracted EPS by released cellular molecules was estimated to 0.8%. However, by performing similar determination for B3 biofilm extracts results indicate a higher level of intracellular compounds that was estimated as 9.7% of the total extracted sugars and proteins.
1000000
400
-100 0
a Total G6P-DH activity as the sum of the G6P-DH units measured in sonication, Tween and EDTA extracts. b Evaluated by measurement of the G6P-DH activity released after lysis of cupriavidus necator pure suspensions: 0.2 U per 106 equivalent lysed cells. c Evaluated using the correlation factor of dry biomass per number of cells: 4.08$106 g per 106 cupriavidus necator cells. d Released cellular compounds after extraction/total extracted proteins and sugars.
B1
Molecular weight (kDa)
B2 extract
1400
B2
Quantitative EPS distribution (% peak area)
a
B1 extract
A Absorbance (mAU).g VSS
Table 2 e Controls of cell lysis during the extraction performed on B1, B2 and B3 biofilm and evaluation of the related contamination level of the EPS extracts.
B2
B3
Fig. 4 e Global size distribution profiles at 280 nm of total EPS extracted from each B1, B2 and B3 biofilm (A) by size exclusion chromatography. Linear semi-logarithmic relation between molecular weight of standard proteins and elution volume (A, insert). Five different EPS size clusters (1 to 5) were identified between 0.5 kDa and 5000 kDa and their relative distribution inside each biofilm was evaluated by peak integration of the 280 nm signal (B).
by 3e7 kDa size molecules and (iii) a range of small size molecules eluted beyond 24 mL, which corresponds to the total inclusion volume of the column. These latter small fractions are not in the optimal separation range offered by the column but are expected to be under 0.5 kDa and are grouped in one single category. Fig. 4A also shows that these three recurring size fractions compose alone the B1 biofilm profile. On the other hand, additional peaks were identified in B2 and in B3 biofilm profiles. Indeed, both B2 and B3 biofilm profiles revealed a fraction eluted at 17e18 mL (i.e. 20e25 kDa), and B3 biofilm alone revealed a fraction eluted at 22e24 mL (i.e. 0.5e2 kDa). A total of five different size clusters were identified among the three studied biofilms: cluster 1 (>5000 kDa), cluster 2 (20e25 kDa), cluster 3 (3e7 kDa), cluster 4 (0.5e2 kDa) and cluster 5 (<0.5 kDa). The relative abundance of EPS size clusters between each other and between each biofilm was undertaken by peak integration of each chromatographic profile. Fig. 4B compares size clusters between each biofilm, and highlights the predominance of the three recurring EPS size clusters (1, 3 and 5). Cluster 3 (3e7 kDa) was the most represented and with 86%, 60% and 46% occurrence of the total peak areas eluted from B1, B2 and B3 chromatograms respectively. The cluster 2 was specifically found in B2 and B3 biofilms, and in the same proportions (3%). The cluster 4 appeared in B3
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 2 9 e1 5 3 8
A
B1
180
B2
B3
160
Absorbance (mAU)
140 120 100 80 60 40 20 0 -20 0
2
4
6
8 10 12 14 16 18 20 22 24 26 28 30
elution volume (mL)
B
90 80
Absorbance (mAU)
70 60 50 40 30
1535
protocol (sonication, Tween and EDTA extracts). Fig. 5 shows size distribution of EPS obtained by each extraction method individually. Fig. 5C shows that EDTA-extract profiles were generally poor in EPS size diversity and offered similar profiles between the three biofilms. Indeed, this EDTA extraction step revealed the three recurring EPS clusters alone (clusters 1, 3 and 5) with a predominance of cluster 3, i.e. EPS belonging the 3e7 kDa fraction eluted between 20 and 22 mL. Fig. 5Aand B show that sonication and Tween extract size profiles, were more diversified and diverged between biofilm samples. Indeed, only the recurring clusters (1, 3 and 5) were found in B1 biofilm, whilst all clusters (1e5) were found in B2 and B3 biofilms. This result shows that cluster 2 was found only within the heterotrophic/autotrophic B2 and B3 biofilms independently on the extraction method. This latter result confirms the global analysis performed previously. On the other hand, cluster 4 which was identified in B3 biofilm alone in the global analysis is finally identified by this specific analysis, in the sonication and Tween extracts of B1 and B2 biofilms. Interestingly, a new size cluster not yet identified in previous profiles was only visualized in Tween extract profiles provided from B1 and B2 biofilms. This latter fraction was eluted at 15.6 mL, indicating a specific size of 180 kDa (Fig. 5B).
20 10
4.
0 -10
0
2
4
6
8 10 12 14 16 18 20 22 24 26 28 30 elution volume (mL)
C 1380 Absorbance (mAU)
1180 980 780
The aim of this study was to evaluate whether molecular diversity of EPS are potential markers for biofilm macro-scale characteristics. In order to validate such an approach, molecular investigations were undertaken on three biofilms, each being differentiated by their growth conditions, i.e. different substrate loading rates or different nitrogen content in the supply.
580
4.1.
380
The COD/NH4eN ratio was first chosen as a key parameter to promote the development of carbon or nitrogen removing micro-organisms. This ratio was decreased from 73 (nitrogen limitation) for B1 biofilm, to 4 (excess nitrogen) for B2 biofilm, by increasing the NHþ 4 content in the supply. Neither nitrification, nor denitrification activity was measured in this B1 biofilm indicating that the small amount of NHþ 4 in the feed was consumed for heterotrophic growth only. On the other hand, B2 biofilm which grew under excess nitrogen conditions, showed simultaneous autotrophic and heterotrophic activities. These observations are in agreement with other findings (Matsumoto et al., 2007) which showed that in spite of carbon deficiency heterotrophic bacteria can out-compete other communities such as autotrophic ammonium-oxidizing bacteria, and this due to their higher growth rate (Elenter et al., 2007; Morgenroth and Wilderer, 2000; Okabe et al., 1995). Whilst the heterotrophic B1 biofilm exposed a filamentous structure with a high accumulated mass, B2 biofilm grew into a dense granular type biofilm with a lower accumulated mass. This type of structure is also in agreement with theories valuable for biofilm or granule formation involving slow-growing
180 -20
0
2
4
6
8 10 12 14 16 18 20 22 24 26 28 30 elution volume (mL)
Fig. 5 e Specific size distribution profiles at 280 nm of EPS extracted from each step of the multi-method protocol, sonication (A), Tween (B) and EDTA (C) from B1 biofilm ), B2 biofilm ( ) and B3 biofilm ( ). (
biofilm alone and represented 5% of the total peak areas. These data indicate that EPS diversity was higher in the mixed autotrophic/heterotrophic biofilms (B2 and B3) compared to the simple heterotrophic biofilm.
3.5.
Discussion
EPS size distribution versus extraction methods
A more specific EPS fingerprint investigation was undertaken on the three biofilms by identifying the previously described size clusters in individual extracts provided by the extraction
Relating feed to biofilm properties
1536
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 2 9 e1 5 3 8
nitrifiers which seem to affect the mass density of biological matrixes (Derlon et al., 2008; Elenter et al., 2007; Liu et al., 2004). Changing the organic loading rate by 10 fold between B2 and B3, without modifying the COD/NH4eN ratio, also affected the structural and microbial properties of biofilms. As expected, the high organic load applied for B3 biofilm promoted bacterial growth which was confirmed by the high accumulated biomass measurements. However, this B3 biofilm presented partial filaments with sloughing events which caused heterogeneous colonization. It is probable that this thick biomass developed by B3 biofilm might have been more exposed to hydrodynamics, which could have triggered localized detachment events (Ohashi et al., 1995), compared to the thinner and homogeneous structure described for B2 biofilm. In addition, such a thick structure promoted anaerobic zones inside the B3 biofilm, which was confirmed by a two fold higher denitrification activity compared to B2 biofilm. Moreover, oxygen deficiency in B3 biofilm could have promoted bacterial mortality and nitrifiers, who often lose out when competing heterotrophic bacteria for oxygen, might have been particularly affected. This hypothesis is supported by the fact that nitrification efficiency was lower in the thick B3 biofilm (65%) compared to the thin B2 biofilm (85%) and by the detection of G6P-DH activity in B3 biofilm prior to EPS extraction (results not shown). Results clearly show that controlled environmental conditions can pilot microbial activities inside growing biofilms, and also modify their macro-scale structural properties.
4.2. Influence of environmental conditions on EPS production In order to harvest a representative pool of biofilm EPS, a multi-method protocol based on both mechanical and chemical extraction steps was applied on the three biofilms. Quantitative analysis of extracted proteins and polysaccharides suggests that excess nitrogen in the feed (B2 and B3 biofilms) triggered more EPS production than nitrogen limitation (B1 biofilm). These results do not join those reported by Miqueleto et al. (2010) who related decreasing values of soluble and bound EPS to decreasing carbon/nitrogen ratios in the feed of an anaerobic sequence batch biofilm reactor. EPS were produced only when oxygen, even at very low concentration was available, suggesting that microaerophilic micro-organisms were the main secretors. Li et al. (2008) showed that different thicknesses of membrane-aerated biofilms in which counter-gradients of oxygen and substrate existed, led to different EPS distributions. These authors reported a maximum EPS content (120e140 mg EPS g1 VSS extracted by formaldehyde and NaOH) in the aerobic region of the studied biofilm where carbon limitation occurred and autotrophic ammonia oxidizing bacteria developed. This latter content can be compared to the nitrifying B2 biofilm where the extractible EPS reached 210 mg g1 VSS. The total amount of extracted EPS was 2.5 times higher in B3 biofilm (high organic load) compared to B2 biofilm (low organic load) and several theories can be quoted for this increase in EPS content. Firstly, the higher substrate load
applied on B3 biofilm might have promoted bacterial growth rates, forming a thick and less cohesive structure as confirmed by sloughing events. This is in agreement with previous data which report that heterotrophic fast growing bacteria develop lower resistance towards either mechanical or chemical disintegration methods (Denkhaus et al., 2007). Consequently, extraction of EPS might be easier in such a fragile structure leading to a higher content of proteins and polysaccharides in extracts. Secondly, the extracted molecules were contaminated by soluble intracellular compounds but the level of contamination, estimated around 10%, was not high enough to justify the 2.5 fold increase of proteins and sugars observed in B3 compared to B2 biofilm. As stated earlier in the discussion, denitrification activity evidenced anoxic areas in the B3 biofilm and Adav et al. (2009) recently located proteolitic activities in anaerobic cores of bacterial granules that might have been responsible for the occurrence of granule breakdown. Therefore, possible proteolytic activity in anaerobic zones in B3 biofilm could partly explain the associated unstable structure and the high content in released proteins. Proteins were quantified in majority in all extracts with a global protein/polysaccharide ratio of 2.9 0.2 for B1 and B2 biofilms and of 3.7 0.2 for B3 biofilm. These data are in agreement with those of Gao et al. (2008) showing protein/ polysaccharide ratios varying between 1.3 and 3.3 along vertical profiles inside heterogeneous aerobic bio-filters. According to Durmaz and Sanin (2001), the amount of substrate converted to polymers by the cell depends on the composition of the growth medium. Indeed, substrates with low nitrogen content, as found in B1 biofilm may favor polysaccharide production, and on the other hand, substrates with excess nitrogen, as found in B2 and B3 biofilms, should promote protein production. Therefore, while high protein contents of B2 and B3 biofilms (164 mg g1 VSS and 424 mg g1 VSS) are in agreement with expectations, the polysaccharide content of B1 biofilm somehow low compared to B2 (15 mg g1 VSS versus 54 mg g1 VSS). This could be explained by the low organic load applied to B1 which seems to favor primarily cell growth and hence proteins (enzymatic material) rather than carbon storage (polysaccharides).
4.3.
EPS size fingerprinting of biofilms
In order to obtain a global molecular fingerprint of each biofilm matrixes, EPS extracts obtained from the multi-method extraction protocol were pooled for global analysis of the size distribution of the extracted EPS. Fractionation of these pooled extracts by SEC revealed a total of five different EPS size clusters. Three clusters were found in common between each biofilm, of which cluster 1 (>5000 kDa) is excluded from the column due to too high molecular weight EPS. Garnier et al. (2005) have already shown the existence of associated proteins/polysaccharides/mineral compounds in fractions eluted near the size exclusion volume when characterizing EPS extracted from activated sludge by SEC. Therefore, the EPS size cluster 1 might probably be represented by polymers eluted as a colloidal structure. Cluster 5 (<0.5 kDa) is, on the other hand, eluted in the total inclusion volume of the column, where the separation efficiency is reduced. This
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 2 9 e1 5 3 8
cluster 5 may either be effectively low molecular weight organic molecules such as amino-acids and peptides, or otherwise molecules which can interact with the column and hence be partially retained during elution. Hydrophobic retention of EPS on the sepharose column beads has already been proven by Comte et al. (2007) and Garnier et al. (2005), when eluting EPS extracted from activated sludges with 5% methanol. However, performing similar experimental conditions did not allow to evidence particular hydrophobic retention in this study (results not shown). Finally, the intermediate 3e7 kDa cluster 3 was predominant overall other clusters and in all biofilms. The recurrence of this cluster 3 in the three highly diversified biofilms suggests that the associated size molecules are either associated to the heterotrophic activity confirmed in all biofilms, or to mandatory EPS involved in bacterial aggregate consolidation and/or adhesion. The three recurrent EPS size clusters 1, 3 and 5, identified in this present study, were the only components of the global EPS fingerprint from the heterotrophic biofilm B1. By introducing nitrification and denitrification activities in B2 and B3 biofilms, another EPS size fraction appeared between 20 and 25 kDa (cluster 2). This latter fraction (cluster 2) could be associated to the presence of bacteria involved in nitrogen removal processes. This cluster was represented in similar proportions within B2 and B3 biofilms (3% of the eluted EPS). Since these latter biofilms exposed different nitrification and denitrification levels, cluster 2 cannot be specifically related to nitrification or denitrification microbial activities. Concerning B3 biofilm, G6P-DH measurements indicated natural cell lysis as well as cell breakage after EPS extraction. About 9.7% of the proteins and sugars measured in the B3 extract may originate from the intracellular compartment. However, due to their low proportion and to the fact that these intracellular compounds may be natural constituents of the biofilm matrix, size fingerprint of B3 biofilm can be considered as relevant. The global EPS size profile of B3 biofilm revealed an additional size cluster between 0.5 and 2 kDa, named cluster 4, that was not identified in the global EPS size profile of B1 and B2 biofilms. Performing a more specific EPS size fractionation focused on each soluble extract indicated that cluster 4 was finally found in all three biofilms. Such a result indicates that pooling extracts from one same biofilm sample can hide under-represented size fractions and hence bias final fingerprint profiles. Interestingly, the size cluster 2 (20e25 kDa) which was identified as specific to nitrogen removal activities measured in B2 and B3 biofilms, was also highlighted in sonication and Tween extracts of B2 and B3 biofilms whilst absent in either B1 biofilm extracts. These results suggest that a 20e25 kDa EPS size fraction can effectively be related to the presence and activity of the nitrogen removing micro-organisms, evidenced within B2 and B3 biofilms in spite of their diverging growth conditions and structural properties. Still in a specific view of EPS diversity through extraction methods, chromatographic profiles pointed out the strong diversity of EPS size fractions in sonication and Tween extracts in opposition to the EDTA extraction step. EDTA extracts showed poor size diversity, although the EPS content in these extracts were the highest compared to sonication and Tween extracts. Therefore, EDTA extracts alone would not be
1537
appropriate for a size diversity fingerprint study. On the other hand, the mechanical sonication and hydrophobic Tween methods are able to extract all size clusters (1e5) identified previously. Interestingly, Tween extracts revealed an additional size fraction of 180 kDa in B1 and B2 biofilms only, which was not visualized during the global study. Tween step thus reveals the most diversified EPS size profiles although EPS content in the extracts were the lowest compared to sonication and EDTA extracts. These results suggest that extraction method-specificity could be a relevant parameter for fingerprint diagnosis. The Tween-specific 180 kDa size fraction revealed in B1 and B2 biofilms may be associated to the low organic load applied to these two biofilms. In other words, the occurrence of this size fraction may rather be related to a biochemical response towards substrate-limiting conditions than to a specific microbial function. B1 and B2 biofilms were also characterized by stable and homogeneous structures in opposition to B3 biofilm, therefore, the occurrence of this Tween-specific size fraction might also mark the mechanical stability of both biofilms in opposition to B3 where this 180 kDa size fraction was absent. The specificity of this fraction towards Tween treatments indicate that the associated EPS have hydrophobic properties. Hydrophobic properties of EPS might hence be implicated in the mechanical stability of biofilms. Such results could be of interest for the understanding of attachment and detachment processes. Authors expect that in the future, configuration of appropriate coatings could be suitable to improve biofilm adherence or on the other hand to prevent biofilm development. Indeed, integrating these hydrophobic EPS fractions inside or on top of coatings could promote molecular interactions and hence biofilm strength. On the other hand, integrating specific enzymes which are liable to digest these hydrophobic EPS can also provide an alternative to toxic biocides in order to prevent biofilm growth (e.g. for heat exchangers, drinking water distribution systems). However, further studies are required before hand, such as characterizing hydrophobic EPS in different cohesive parts of biofilms. These latter investigations are already under progress.
5.
Conclusions
Characterization of EPS extracted from multi-species biofilms was investigated using a multi-method extraction procedure coupled with a SEC analysis. Results showed that EPS size diversity was higher in the two mixed heterotrophic/autotrophic biofilms compared to the heterotrophic biofilm. The multi-method extraction strategy provided consistent quantitative and qualitative EPS fractions. However, by focusing on each extraction steps, results showed that each method offered different quantities and different size diversity profiles. Nevertheless, the occurrence of a 25e50 kDa size fraction was systematically associated to biofilms exposing nitrogen removing activities. Moreover, a 180 kDa size fraction occurred in Tween extracts only and was associated to mechanically stable biofilms. This study has put forward the importance of methodology in qualitative investigations of EPS in biofilms. Hydrophobic
1538
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 2 9 e1 5 3 8
EPS seem to provide highly diversified size profiles with a particular size category (180 kDa) which might be a print of mechanical stability. Analysis of the hydrophobic EPS of biofilms developed under different shear stress conditions is currently under investigation.
references
Adav, S.S., Lee, D.J., Lai, J.Y., 2009. Proteolytic activity in stored aerobic granular sludge and structural integrity. Bioresour. Technol. 100, 68e73. Azeredo, J., Lazarova, V., Oliveira, R., 1999. Methods to extract the exopolymeric matrix from biofilms: a comparative study. Water Sci. Technol. 39, 243e250. Boualam, M., Quiles, F., Mathieu, L., Block, J.-C., 2002. Monitoring the effect of organic matter on biofilm growth in low nutritive waters by ATR-FT-IR spectroscopy. Biofouling 18, 73e81. Branda, S., Ashild, V., Friedman, L., Kolter, R., 2005. Biofilms: the matrix revisited. Trends Microbiol. 13, 20e26. Chen, X., Stewart, P.S., 2002. Role of electrostatic interactions in cohesion of bacterial biofilms. Appl. Microbiol. Biotechnol. 59, 718e720. Comte, S., Guibaud, G., Baudu, M., 2007. Effect of extraction method on EPS from activated sludge: an HPSEC investigation. J. Hazard. Mater. 140, 129e137. Coufort, C., Derlon, N., Ochoa-Chaves, J., Line, A., Paul, E., 2007. Cohesion and detachment in biofilm systems for different electron acceptors and donors. Water Sci. Technol. 55, 421e428. Davies, D.G., Chakrabarty, A.M., Geesy, G.G., 1993. Exopolysaccharide production in biofilms: substratum activation of alginate gene expression by Pseudomonas aeruginosa. Appl. Environ. Microbiol. 59, 1181e1186. Denkhaus, E., Meisen, S., Telgheder, U., Wingender, J., 2007. Chemical and physical methods for characterisation of biofilms. Mikrochim. Acta 158, 1e27. Derlon, N., Masse, A., Escudie´, R., Bernet, N., Paul, E., 2008. Stratification in the cohesion of biofilms grown under various environmental conditions. Water Res. 42, 2102e2110. Dreywood, R., 1946. Qualitative test for carbohydrate material. Ind. Eng. Chem. 18, 499. Durmaz, B., Sanin, F.D., 2001. Effect of carbon to nitrogen ratio on the composition of microbial extracellular polymers in activated sludges. Water Sci. Technol 44, 221e229. Elenter, D., Milferstedt, K., Zhang, W., Hausner, M., Morgenroth, E., 2007. Influence of detachment on substrate removal and microbial ecology in a heterotrophic/autotrophic biofilm. Water Res. 41, 4657e4671. Gao, B., Zhu, X., Xu, C., Yue, Q., Li, W., Wey, J., 2008. Influence of extracellular polymeric substances on microbial activity and cell hydrophobicity in biofilms. J. Chem. Technol. Biotechnol. 83, 227e232. Garnier, C., Go¨rner, T., Lartigues, B.S., Abdelouhab, S., Donato, P., 2005. Characterization of activated sludge exopolymers from various origins: a combined size-exclusion chromatography and infrared microscopy study. Water Res. 39, 3044e3054. Jahn, A., Nielsen, P.H., 1995. Extraction of extracellular polymeric substances (EPS) from biofilms using cation exchange resin. Water Sci. Technol. 32, 157e164.
Jahn, A., Nielsen, P.H., 1998. Cell biomass and exopolymer composition in sewer biofilms. Water Sci. Technol. 37, 17e24. Li, T., Bai, R., Liu, J., 2008. Distribution and composition of extracellular substances in membrane-aerated biofilm. J. Biotechnol. 135, 52e57. Liu, Y., Yang, S.F., Tay, J.H., 2004. Improved stability of aerobic granules by selecting slow-growing nitrifying bacteria. J. Biotechnol. 108, 161e169. Matsumoto, S., Terada, A., Tsuneda, S., 2007. Modeling of membrane-aerated biofilm: effects of COD/TKN ratio, biofilm thickness and surface loading of oxygen on feasibility of simultaneous nitrification and denitrification. Biochem. Eng. J. 37, 98e107. Mayer, C., Moritz, R., Kirschner, C., Borchard, W., Maibaum, R., Wingender, J., Flemming, H.C., 1999. The role of intermolecular interactions: studies on model systems for bacterial biofilms. Int. J. Biol. Macromol. 26, 3e16. Miqueleto, A., Dolosic, C., Pozzi, E., Foresti, E., Zaiat, M., 2010. Influence of carbon sources and C/N ratio on EPS production in anaerobic sequencing batch reactors for wastewater treatment. Bioresour. Technol. 101, 1324e1330. Morgenroth, E., Wilderer, P.A., 2000. Influence of detachment mechanisms on competition in biofilms. Water Res. 34, 416e426. Ohashi, A., Mobarry, B., Manem, J.A., Stahl, D.A., Rittmann, B.E., 1995. Influence of substrate COD/TKN ratio on the structure of multi-species biofilms consisting of nitrifiers and heterotrophs. Water Sci. Technol. 32, 75e84. Okabe, S., Hiratia, K., Ozawa, Y., Watanabe, Y., 1995. Spatial microbial distributions of nitrifiers and heterotrophs in mixed-population biofilms. Biotechnol. Bioeng. 50, 24e35. Ramsay, B.A., Lomaliza, K., Chavarie, C., Dube´, B., Bataille, P., Ramsay, J.A., 1990. Production of poly-(beta-hydroxybutyricco-beta-hydroxyvaleric) acids. Appl. Environ. Microbiol. 56, 2093e2098. Ras, M., Girbal-Neuhauser, E., Paul, E., Lefebvre, D., 2008a. A high yield multi-method extraction protocol for protein quantification in activated sludge. Bioresour. Technol. 99, 7465e7471. Ras, M., Girbal-Neuhauser, E., Paul, E., Spe´randio, M., Lefebvre, D., 2008b. Protein extraction from activated sludge: an analytical approach. Water Res. 42, 1867e1878. Simon, S., Paı¨ro, B., Villain, M., D’Abzac, P., Van Hullebusch, E., Lens, P., Guibaud, G., 2009. Evaluation of size exclusion chromatography (SEC) for the characterization of extracellular polymeric substances (EPS) in anaerobic granular sludges. Bioresour. Technol. 100, 6258e6268. Smith, P.K., Krohn, R.I., Hermanson, G.T., Mallia, A.K., Gartner, F.H., Provenzano, M.D., Fujimoto, E.K., Goeke, N.M., Olson, B.J., Klenk, D.C., 1985. Measurement of protein using bicinchoninic acid. Anal. Biochem. 150, 76e85. Standard Methods for the Examination of Water and Wastewater, ninetienth ed., 1995 APHA, AWWA, WPCF, Washington DC, USA. Wijeyekoon, S., Mino, T., Satoh, H., Matsuo, T., 2004. Effects of substrate loading rate on biofilm structure. Water Res. 38, 2479e2488. Zhang, X.Q., Bishop, P.L., KinKle, B.K., 1999. Comparison of extraction methods for quantifying extracellular polymers in biofilms. Water Sci. Technol. 39, 211e216.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 3 9 e1 5 4 8
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Long term case study of MIEX pre-treatment in drinking water; understanding NOM removal Mary Drikas*, Mike Dixon, Jim Morran Australian Water Quality Centre, South Australian Water Corporation, GPO Box 1751, Adelaide SA 5001, Australia
article info
abstract
Article history:
Removal of natural organic matter (NOM) is a key requirement to improve drinking water
Received 25 June 2010
quality. This study compared the removal of NOM with, and without, the patented
Received in revised form
magnetic ion exchange process for removal of dissolved organic carbon (MIEX DOC) as
18 November 2010
a pre-treatment to microfiltration or conventional coagulation treatment over a 2 year
Accepted 18 November 2010
period. A range of techniques were used to characterise the NOM of the raw and treated
Available online 24 November 2010
waters. MIEX pre-treatment produced water with lower concentration of dissolved organic carbon (DOC) and lower specific UV absorbance (SUVA). The processes incorporating MIEX
Keywords:
also produced more consistent water quality and were less affected by changes in the
MIEX
concentration and character of the raw water DOC. The very hydrophobic acid fraction
Coagulation
(VHA) was the dominant NOM component in the raw water and was best removed by MIEX
Microfiltration
pre-treatment, regardless of the raw water VHA concentration. MIEX pre-treatment also
NOM
produced water with lower weight average apparent molecular weight (AMW) and with the
Fractionation
greatest reduction in complexity and range of NOM. A strong correlation was found
Molecular weight
between the VHA content and weight average AMW confirming that the VHA fraction was a major component of the NOM for both the raw water and treated waters. ª 2010 Elsevier Ltd. All rights reserved.
1.
Introduction
Natural organic matter (NOM) has a significant impact on drinking water quality either directly, by reacting with water treatment chemicals (to form disinfection by-products), or indirectly, by impacting water treatment processes (including fouling of membranes and reducing the effectiveness of activated carbon for contaminant removal). Therefore the water industry has focussed on improving current treatment and developing new processes to increase the removal of NOM. Conventional treatment comprising coagulation/flocculation/ sedimentation/filtration is one of the most widely used methods to remove NOM. Extensive research has been undertaken to increase the extent of NOM removal by conventional treatment, including the use of increased
coagulant doses and reduced pH, referred to as enhanced coagulation (Crozes et al., 1995; White et al., 1997; Bell-Ajy et al., 2000). A more recent technology developed specifically for the removal of NOM is the patented MIEX DOC Process (Morran et al., 1996; Drikas et al., 2002). This process utilises a strong base anion-exchange resin, incorporating magnetic iron oxide particles within its core, which is applied to raw water utilising a stirred contactor. The small resin beads facilitate rapid reaction whilst the magnetic component allows separation of the resin and recycling of the resin in a continuous process. This differs from other more traditional applications where ion exchange resin is applied as the final polishing step within a filter (Brattebo et al., 1987; Baker et al., 1995). Laboratory scale testing of the MIEX resin has proven the effectiveness of the process for rapid removal of NOM, to a greater extent than that
* Corresponding author. Tel.: þ61 8 7424 2110; fax: þ61 8 7003 2110. E-mail address:
[email protected] (M. Drikas). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.11.024
1540
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 3 9 e1 5 4 8
possible by coagulation, or enhanced coagulation, in a range of waters (Drikas et al., 2002, 2003a; Singer and Bilyk, 2002; Morran et al., 2004; Fearing et al., 2004; Humbert et al., 2005; Boyer and Singer, 2005). Some studies have also been conducted comparing pilot plant or full scale MIEX treatment with coagulation (Drikas et al., 2003b; Allpike et al., 2005; Boyer and Singer, 2005; Warton et al., 2007; Singer et al., 2007, 2009; Jarvis et al., 2008) although all of these studies have been conducted over a short period of time. A few studies have also assessed the effectiveness of MIEX as a means of reducing fouling of microfiltration or ultra filtration membranes (Fabris et al., 2007; Humbert et al., 2007; Dixon et al., 2010). MIEX has been shown to remove both hydrophobic and hydrophilic organic acid fractions of NOM (Singer and Bilyk, 2002; Morran et al., 2004; Fearing et al., 2004; Boyer and Singer, 2005; Mergen et al., 2008, 2009) to a greater extent than possible with coagulation alone. MIEX was also found to remove a wider range of molecular weight components than coagulation with alum (Drikas et al., 2003a, b; Morran et al., 2004; Allpike et al., 2005; Humbert et al., 2005; Singer et al., 2007). The first MIEX plant was commissioned at Mt Pleasant in South Australia in August 2001 (Drikas et al., 2003b). The Mt Pleasant Water Treatment Plant (WTP) is a small (2.5 ML/d) potable water treatment plant supplying high quality treated water to the local community. However the plant is more complex, innovative and diverse in processes than necessary to enable the MIEX DOC process to be fully evaluated. The WTP has been divided into two streams of 1.25 ML/d capacity, each incorporating the MIEX DOC process but the plant also enables comparison of two separate subsequent processes for the removal of suspended matter e conventional treatment (comprising coagulation, flocculation, sedimentation, and filtration) and submerged microfiltration (MF) (Drikas et al., 2003b). This study enabled an extended evaluation of the impact of the MIEX DOC process on NOM removal by comparing the performance of the two processes operating at the Mt Pleasant WTP with separate pilot plant installations utilising the same processes (conventional treatment and submerged MF) but without MIEX pre-treatment, over a 2 year period. A quantitative assessment of the NOM removed by all the treatment processes was undertaken together with a detailed study of the character of the remaining NOM using a rapid fractionation technique and molecular weight profiles for 16 months of this period. This study has identified novel benefits of the continuous operation of the MIEX DOC process and provided a clearer understanding of the character of the NOM removed.
2.
Materials and methods
2.1.
Treatment processes
A schematic of the treatment trains used is provided in Fig. 1. A conventional pilot plant (Conv) consisting of coagulation, flocculation, sedimentation and rapid filtration was established on site at the Mt Pleasant WTP using the same raw water that supplied the WTP. The flash mixer was a vessel of 1.5 L volume which was stirred at a rate of 200 rpm with a flat paddle agitation blade. Alum was dosed directly into the top of this vessel via a piston dosing pump incorporating a flow
dampening device. The two flocculation bays held 45 L each and were separated by a plate which the water laundered over. The first vessel was stirred at 80 rpm and the second at 40 rpm by an overhead stirrer with 25 offset flat paddles. Three 50 mm pipes delivered flocculated water into the 65 L sedimentation bay. The inverted pyramid shaped sedimentation bay allowed sludge to be collected over three days and be drained off to waste. Settled water was laundered from the top of the sedimentation bay via flexible beverage tubing which ran to a peristaltic pump. The pipe was split into three via a manifold in order to reach the desired flow rate and pumped via three peristaltic heads to the top of the filter column. The filter column consisted of a 140 mm diameter acrylic column and was filled with 600 mm of gravel (of varying grades), with 300 mm of sand of size 0.5e0.6 mm and 750 mm anthracite of size 1.0e1.1 mm. The media was of the same type and depth as the filters on Stream 1 of the Mt Pleasant WTP. The conventional pilot plant operated for three days on and four days off. The pilot plant throughput was 36 L/h which gave 2.5 h flocculation and 2 h settling time. The alum dose was 40 mg/L (as Al2(SO4)3$18H2O) over the study period. This was selected by the use of a model (van Leeuwen et al., 2005) and confirmed by regular jar tests to achieve the optimum DOC removal (defined as the point of diminishing return, where an additional 10 mg/ L alum produces <0.1 mg/L DOC reduction). The pH was not optimised but was between 6.5e6.8 throughout the study. Conventional treatment (Conv) was compared with Stream 1 at the Mt Pleasant WTP which incorporates MIEX followed by conventional treatment comprising coagulation, flocculation, sedimentation, rapid filtration (MIEX Coag) (Fig. 1). During the period July 2005 to June 2007, MIEX was applied to maintain the resin dose at or above 10 mL/L for 10 min contact followed by sedimentation and removal of the resin before entering the separate particulate removal processes. The actual resin dose varied between 8 and 16 mL/L (average 12 mL/L) over this period. The resin was recirculated in a continuous process with 10% removed for regeneration using sodium chloride. Fresh regenerated resin was returned continuously to the resin contact tank to maintain a constant resin dose while regeneration was undertaken separately on a batch process as required. Virgin makeup resin was added on an infrequent basis to compensate for resin lost due to attrition. Coagulation in Stream 1 during this period varied between 6 and 10 mg/L (average 8 mg/L) (as Al2(SO4)3$18H2O) and 0.2 mg/L poly dimethyl diallyl ammonium chloride (DADMAC) as a coagulant aid to ensure filtered water turbidity was maintained below 0.2 NTU. The throughput of Stream 1 at Mt Pleasant WTP remained steady at 0.3 ML/day which gave 3.5 h flocculation and 1.5 h settling time prior to filtration. Filter run times averaged 2 days. The second stream at the Mt Pleasant WTP incorporates MIEX followed by submerged continuous microfiltration (CMFS) with polyvinylidenefluoride (PVDF) membranes which have a nominal pore size of 0.04 mm (Memcor S10 V). However the MF pilot plant was used to provide the comparison of MF with and without MIEX pre-treatment to ensure operating conditions were identical for both operating systems. The MF pilot plant consisted of a single module CMF-S membrane, the same variety as that used in the Mt Pleasant WTP. Two separate membrane modules were used in the MF pilot plant unit in a one week on, one week off rotation. The source water for
1541
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 3 9 e1 5 4 8
Raw
Conventional Treatment Pilot plant
Stream1 WTP MIEX
Submerged Microfiltration Pilot Plant
WTP Conventional Treatment
Conv
Stream 1 WTP MIEX
Submerged Microfiltration Pilot Plant
MIEX Coag
Raw MF
MIEX MF
Fig. 1 e Schematic of Treatment Trains. Conv-conventional treatment pilot plant consisting of coagulation, flocculation, sedimentation and rapid filtration; MIEX Coag-Stream 1 at Mt Pleasant water treatment plant (WTP) consisting of MIEX followed by conventional treatment utilising coagulation, flocculation, sedimentation and rapid filtration; Raw MF-raw water followed by passage through submerged microfiltration membrane in pilot plant; MIEX MF-MIEX treated water sourced from Stream 1 at Mt Pleasant WTP followed by passage through submerged microfiltration membrane in pilot plant.
one module was raw water (Raw MF) and the other MIEX treated water sourced from Stream 1 (prior to coagulation) (MIEX MF) (Fig. 1).
2.2.
Analyses
Samples were taken from (a) the raw water (Raw), (b) after the filtration stage of both the conventional pilot plant (Conv) and the WTP Stream 1 (MIEX Coag) and (c) after passage through each of the MF membranes (Raw MF and MIEX MF). Monitoring of water quality included measuring the extent of NOM removal using dissolved organic carbon (DOC) and UV absorbance at 254 nm (UV254) three times a week. Characterization of NOM was determined from August 2005 to December 2006 by measuring molecular weight distribution and rapid resin fractionation. Characterization was undertaken monthly for the first 6 months and every 2 months for the remainder of the study. Samples for DOC and UV254 were filtered through 0.45 mm prerinsed membranes. UV254 was measured through a 1 cm quartz cell. DOC was measured using a Sievers 820 Total Organic Carbon Analyser (GE Analytical Instruments, USA). Specific UV absorbance (SUVA) was calculated as (UV254 100)/DOC in L/mg-m. Rapid fractionation analysis separates DOC into four fractions by adsorption onto different adsorbent resins in a sequential process based on hydrophobicity. After filtration through 0.45 mm pre-rinsed membranes a 500 mL sample was acidified and flowed through packed columns of Supelite DAX-8 and then Amberlite XAD-4 resin with intermediate samples taken for DOC analysis. Samples were then adjusted to pH 8 and flowed through a strong anion-exchange resin (Amberlite IRA-958). All resins were obtained from Supelco (Sigma Aldrich, USA). Fraction concentrations were obtained by calculation of DOC concentration measured before and after each resin. Fractions produced are defined as very hydrophobic acids (VHA), slightly hydrophobic acids (SHA), charged hydrophilics (CHA) and neutral hydrophilics (NEU).
Specifics of the technique and definitions have been described elsewhere (Chow et al., 2004). Apparent molecular weight (AMW) distribution profiles were determined using high performance size-exclusion chromatography (HPSEC) utilising UV detection after filtration through 0.22 mm pre-rinsed membranes. Weight and number average molecular weight (Mw and Mn respectively) were derived using the following equations (Chin et al., 1994). P ni M2i i (1) Mw ¼ P ni Mi i
and P
ni Mi Mn ¼ iP ni
(2)
i
where ni is the height of the HPSEC curve and Mi is the equivalent calculated molecular weight of an analyte eluted at volume i. The polydispersity (ratio of Mw to Mn) was also calculated (Chin et al., 1994). A polydispersity value of 1 indicates the presence of a single homogenous compound while greater values indicate a more disperse, complex mixture of compounds. HPSEC analysis was undertaken using a Waters Alliance 2690 separations module and 996 photodiode array detector at 260 nm (Waters Corporation, USA). Phosphate buffer (0.02 M) with 0.1 M sodium chloride was flowed through a Shodex KW802.5 packed silica column (Showa Denko, Japan) at 1.0 mL/min. AMW was derived by calibration with poly-styrene sulphonate molecular weight standards of 35, 18, 8 and 4.6 kDa.
3.
Results and discussion
3.1.
NOM removal
The raw water quality at the Mt Pleasant WTP varied over the study period; turbidity, 8.7e60 Nephelometric turbidity units
1542
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 3 9 e1 5 4 8
7 6
Period A
Period B
Raw
Period C
Conv MIEX Coag Raw MF
DOC (mg/L)
5
MIEX MF
4 3 2 1 0 Jul-05
Oct-05
Feb-06
May-06
Aug-06
Nov-06
Mar-07
Jun-07
Fig. 2 e DOC present after each treatment process (Period A, Auguste14 November 2005; Period B, 18 November 2005e15 June 2006; Period C, 19 June 2006e30 June 2007).
a
7 6
DOC (m g/ L)
5 4 3 2 1 0 Raw
b
Conv
MIEX Coag
Raw MF
MIEX MF
0.4 0.35
UVab s(/ cm )
0.3 0.25 0.2 0.15 0.1 0.05 0 Raw
c
Conv
MIEX Coag
Raw MF
MIEX MF
8 7 6
SUVA (L/m g-m )
(NTU), colour, 6e23 Hazen units, and DOC, 2.7e5.8 mg/L. The DOC variation is shown in detail in Fig. 2. Three distinct periods were observed; Period A, Auguste14 November 2005, (winter/spring), Period B, 18 November 2005e15 June 2006, (summer/autumn), and Period C, 19 June 2006e30 June 2007 (the full season cycle during drought conditions upstream). Period A consisted of variable DOC between 3.0 and 4.8 mg/L, generally decreasing, over the 4 month period. Period B began with an increase in DOC to 5.7 mg/L, due to rain and inflow from storages upstream, and remained reasonably stable within the band 4.5e5.8 mg/L over the 7 month period. Period C was less distinct but appeared to start mid June 2006 and showed a gradual steady decrease from 4.5 mg/L to 3 mg/L over the next year, under drought conditions and very little inflow upstream. The DOC concentration after each of the treatment processes over the periods A, B and C is also shown in Fig. 2. Pre-treatment with MIEX achieved the lowest DOC, with similar concentration obtained regardless of whether MIEX was followed by coagulation or MF. The DOC concentration after MIEX treatment remained nearly constant over the entire study period despite the variation in raw water DOC. Conventional treatment did not remove as much DOC as the MIEX pre-treatment whilst MF alone with no pre-treatment was found to consistently remove only a small amount of DOC. The greater DOC removal by MIEX alone compared with coagulation has been observed previously (Singer and Bilyk, 2002; Drikas et al., 2003a,b; Morran et al., 2004; Boyer and Singer, 2005; Warton et al., 2007; Jarvis et al., 2008). The extent of DOC remaining, as well as UV254 and SUVA, can also be compared over the period of the study using box and whisker plots as summarised in Fig. 3a. Outliers were excluded for each treatment train but the number of analyses conducted over this period and used in the calculations was similar for each process e Raw, 137; Conv, 132; MIEX Coag, 132; Raw MF, 139; MIEX MF, 134. As shown in Fig. 3a, the two processes incorporating MIEX pre-treatment achieved the lowest DOC concentrations with median of 1.7 mg/L after MIEX Coag and 1.6 mg/L after MIEX MF compared with 2.5 mg/L after Conv and 3.3 mg/L after Raw MF. This equates to greater average removal of DOC using MIEX pre-treatment (54% and 57%) compared with the Conv (33%) and the Raw MF (11%)
5 4 3 2 1 0 Raw
Conv
MIEX Coag
Raw MF
MIEX MF
Fig. 3 e Box and whisker plot of (a) DOC, (b) UVabs and (c) SUVA remaining after each treatment, for the total period of study. The bottom of the box is the 25th percentile and the top is the 75th. The whiskers represent the maximum and minimum for each treatment process.
processes. The 95 percent confidence intervals around the medians (not shown in this figure) indicate that the MIEX pretreatment trains are statistically different from those not including MIEX. The extent of removal of DOC by the processes incorporating MIEX pre-treatment was within the range previously observed for waters with similar SUVA values (Boyer and Singer, 2005; Singer et al., 2007). Fig. 3a also shows that the raw water over the period of study was skewed towards higher DOC and that this trend was continued after passage through the MF (Raw MF) and to a lesser extent after conventional treatment (Conv). However the data for the two treatments incorporating MIEX did not show any skew around the median results indicating that they produced more consistent water quality and were less affected by the raw water DOC with 50% of the remaining DOC results contained within a region of less than 0.5 mg/L DOC for the duration of the study.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 3 9 e1 5 4 8
The UV254 absorbance removal followed similar trends to the DOC removal although removals obtained were significantly higher - MIEX pre-treatment achieved 81% and 78% removal for the median data compared with the Conv (55%) and the Raw MF (25%) processes. Fig. 3b indicates that the UV254 absorbance of the raw water was also skewed towards higher values but confirms that all treatments used were more effective in removal of UV absorbing components than DOC with the whiskers after all treatments significantly reduced. This is not surprising as these types of structures are generally larger and more hydrophobic which makes them easier to remove by both coagulation and physical filtration. The unexpected removal of a small amount of DOC by the Raw MF can be attributed to removal of these larger UV absorbing compounds as evidenced by the reduced range of UV254 absorbance apparent in Fig. 3b. The removal of larger and more UV absorbing organics by MF membranes has been observed by other researchers (Fan et al., 2001; Lee et al., 2005). The reduced spread in UV254 absorbance of the Conv treatment is expected based on the known preferred removal of these compounds by coagulation (Chow et al., 1999; Archer and Singer, 2006). The MIEX treated waters had a very small spread of UV254 absorbance with 50% of data within a region of less than 0.01 Abs/cm and the total spread of data less than 0.03 Abs/cm. This data confirms previous observations that the removal of UV absorbing compounds by MIEX is very effective (Drikas et al., 2002; Singer and Bilyk, 2002; Fearing et al., 2004; Boyer and Singer, 2005; Humbert et al., 2005; Singer et al., 2007). The small spread in UV254 absorbance data also confirms the consistency in treated water quality following MIEX pre-treatment regardless of raw water changes observed with DOC removal. The change in SUVA would be expected to follow similar trends for all of the treated waters as it is a parameter derived from the DOC and UV254 absorbance data. Fig. 3c confirms that Raw MF reduced the SUVA of the raw water from a median of 2.4 to 2.0 L/mg-m, while Conv achieved a median SUVA of 1.5 L/mg-m. MIEX pre-treatment consistently produced water with the lowest SUVA because more UV absorbing organics were removed by MIEX relative to the DOC. This differed from Boyer and Singer (2006) who found that the treated water SUVA was similar to the raw water SUVA but was consistent with the findings of Singer et al. (2007) who found that SUVA at each utility studied decreased with MIEX treatment. MIEX MF had a median of 1.1 L/mg-m while MIEX Coag was lower with a median of 0.9 L/mg-m. Again the differences in SUVA observed following MIEX treatment were statistically significantly based on a 95 percent confidence interval.
3.2.
Rapid fractionation
To better understand the mechanism of NOM removal, the types of organics removed by each of the treatment processes were investigated. An example of the information obtained from rapid fractionation is shown in Fig. 4 for the sample taken in November 2005. DOC of the raw water was high in the early stages of the study, especially during Period B, and contained high concentration of VHA (1.7e2.3 mg/L). The VHA fraction, consisting predominantly of higher molecular weight humic and fulvic acids, was the dominant fraction
1543
Fig. 4 e Rapid Fractionation for all treatments for sample taken in November 2005. Fractions are defined as very hydrophobic acids (VHA), slightly hydrophobic acids (SHA), charged hydrophilics (CHA) and neutral hydrophilics (NEU). Error bars represent limit of detection.
present and was removed best by the two processes containing MIEX pre-treatment. The SHA fraction, variously described as transphilic (Boyer and Singer, 2005) or hydrophilic acid (Fearing et al., 2004), although not present to a large extent in the raw water during this study, was also preferentially removed by the MIEX pre-treatment. The high proportion of VHA in this water and its preferential removal by the MIEX processes resulted in a significant difference in the amount of DOC removed and in the resulting character of the waters after treatment, as illustrated in Fig. 4. Similar comparative removal was obtained for all samples analysed over the period of the study. Fearing et al. (2004), Sharp et al. (2006) and Mergen et al. (2008, 2009) utilised a modified fractionation process where they further separated the VHA fraction using pH adjustment into two additional fractions, which they defined as HAF (humic acid fraction) and FAF (fulvic acid fraction), and did not separate the non-adsorbed hydrophilic fraction into CHA and NEU as in this study. Fearing et al. (2004) found that MIEX removed a larger proportion of one component of the VHA fraction (the FAF fraction) than coagulation with ferric sulphate but similar amounts of the SHA and the other component of the VHA fraction (the HAF fraction) while Chow et al. (2005) showed a strong correlation of the VHA fraction with alum dose in a conventional treatment plant. The CHA fraction was removed most effectively, particularly by those processes containing coagulation (Conv and MIEX Coag) and/or ion exchange (MIEX Coag and MIEX MF). Other studies have also shown that the CHA fraction is the most amenable to removal by coagulation (Chow et al., 2004; Bolto et al., 2002; van Leeuwen et al., 2002) and ion exchange (Bolto et al., 2002; Morran et al., 2004). Fearing et al. (2004) found that MIEX performed better at removal of the nonadsorbed hydrophilic fraction (combination of CHA and NEU) than ferric coagulation. Boyer and Singer (2005) showed that the MIEX removed all fractions to a greater extent than coagulation with removal increasing with increased resin dose while Singer et al. (2007) found that MIEX removed the VHA and SHA fraction more than the non-adsorbed hydrophilic fraction. Study of the regenerant produced by MIEX treatment of four waters by Mergen et al. (2009) showed that
1544
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 3 9 e1 5 4 8
MIEX removed VHA (both HAF and FAF components), SHA and the non-adsorbed hydrophilic fraction but that the SHA fraction was the only fraction which increased in concentration in the regenerant solution for one water and after coagulation of the regenerant for the four waters studied suggesting better removal of this fraction by MIEX than by coagulation. The NEU hydrophilic fraction generally consists of lower molecular weight components such as polysaccharides and proteins and is often indicative of biologically derived material (Leenheer, 1981; Buchanan et al., 2005). Only very small amounts of the NEU were removed by any of the treatments. van Leeuwen et al. (2002) identified the NEU fraction as recalcitrant following fractionation and coagulation of two waters while Chow et al. (2004) showed that removal of all fractions, except the NEU fraction, could be improved by change in coagulation conditions. Sharp et al. (2006) also showed strong removal of the hydrophobic components by alum coagulation with poor removal of the non-adsorbed hydrophilic fraction (combination of CHA and NEU). The concentration of the dominant VHA fraction present in the raw and remaining after treatment over the period of the study is summarised in Fig. 5. Initial samples taken in Period A showed some inconsistency, particularly with the raw water sample. This was due to development and optimisation of the rapid fractionation method during this initial period. Notwithstanding the early results, it appears that the VHA content of the raw water decreased during Period A, increased at the junction of Period A and B and then steadily decreased during period B and C of the study. Fig. 5 also shows that generally the Raw MF removed little or no VHA, compared with the filtered raw water. This is despite the Raw MF achieving a measurable removal of DOC (median DOC difference of 0.4 mg/L). The lack of observation of consistent removal of VHA is attributed to the limit in sensitivity of the rapid fractionation analysis. The analysis requires measurement of the DOC of the raw sample and of each fraction after passage through the resins which results in an additive limit in detection of 0.2 mg/L for each fraction. This means that small differences in VHA concentrations are not able to be detected. It is likely that some DOC was removed by the Raw MF by adsorption onto particulates during filtration by the mechanism of cake layer formation. During the start of Period B, on the 18th November 2005, there was an increase in raw water turbidity from 30 NTU to above 50 NTU. This higher turbidity level (also associated with higher DOC) was present until late February 2006 when the
turbidity again was reduced to around 30 NTU. This higher turbidity period coincided closely with the measured removal of some VHA by the Raw MF observed at the start of Period B and provides support for this theory with enhanced cake layer formation at the higher turbidity. Conv treatment removed some VHA with the extent of removal apparently dependent on the amount of VHA in the raw water. The MIEX pre-treated waters achieved the lowest VHA concentration and this concentration remained nearly constant over the entire study period, regardless of the raw water VHA concentration. This supports the previous observation, noted with the other water quality parameters, of consistent treated water quality following MIEX pre-treatment regardless of raw water quality changes. There was some change in the NOM character of the raw water during the study. Initially the raw water consisted of 60% VHA during period A and B but this gradually decreased to 50% VHA by the end of the study period. The difference in NOM character of the waters after treatment is also apparent, particularly during period B, where the two MIEX pre-treated waters consistently had between 10 and 20% less VHA present after treatment, than either the Conv or Raw MF. This indicates that the DOC of the MIEX pre-treated waters consisted of a greater percentage of the other fractions; in particular the NEU fraction was a significant component of the remaining DOC, as this was not removed by any of the treatment processes. As stated earlier this NEU component is the most recalcitrant to all forms of treatment. The percentage of the VHA fraction removed by each of the treatment processes over the study period was also calculated (Fig. 6). This confirms that the raw MF removed little or no VHA during the study. Conv treatment removed very little VHA during Period A, a constant amount of VHA, about 35%, during Period B with decreasing removal observed during Period C. The two treatments incorporating MIEX achieved the highest percentage of VHA removal throughout the study. During Period B, when the raw water had very high VHA concentration, the two treatments incorporating MIEX pretreatment consistently removed a significantly greater proportion of the VHA fraction (w70%). During this same period, the Raw MF and the Conv treatments were not able to achieve the same removal of VHA (0% and 35% respectively), resulting in higher concentration of VHA in the treated waters (as shown in Fig. 5). The MIEX pre-treated waters continued to remove a high percentage of VHA even during Period C when the raw water VHA concentration decreased and the Conv
2.50 Raw
100 50
Oct-05
Dec-05
Feb-06
Apr-06
Jun-06
Aug-06
Oct-06
Dec-06
Fig. 5 e Concentration of VHA fraction present after each treatment process.
Fig. 6 e Percentage removal of VHA Fraction by each treatment process.
Dec-06
0
Oc t-06
Aug-05
Raw MF
150
-50
0.00
MIEX MF
200
Aug-06
0.50
MIEX Coag Conv
250
J un-06
1.00
Period C
Apr-06
1.50
Period B
Feb-06
MIEX MF
Period A
300
Aug-05
Raw MF
Dec-05
MIEX Coag
% VHA Fraction Removed
VHA Concentration (mg/L)
Conv
2.00
Nov-05
Period C
Oc t-05
Period B
Sep-05
Period A
1545
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 3 9 e1 5 4 8
0.006
Raw
0.005
Raw MF
UV Abs @ 260 nm
0.004
Conv 0.003
MIEX Coag
0.002
MIEX MF
0.001
0.000 100
1000
10000
-0.001
Apparent Molecular Weight (Daltons)
Fig. 7 e Apparent Molecular Weight profiles for all samples taken in November 2005.
3.3.
Apparent molecular weight distribution profiles
AMW distribution profiles of each of the treatment processes were also determined. It should be noted that as these profiles were obtained using UV detection only UV absorbing NOM would be detected. To illustrate the type of data that was obtained, the AMW profile of the same sample selected from November 2005 that was used in Fig. 4 is shown in Fig. 7. The Raw MF removed minimal DOC, resulting in molecular weight distribution similar to the raw water. Conventional treatment (Conv) removed predominantly higher AMW material, mainly above 1000 Da. The removal of high AMW organic matter with coagulation is well documented in the literature (Collins et al., 1986; Owen et al., 1995; White et al., 1997; Chow et al., 1999). MIEX treatment removed organics across the complete AMW range, including above and below 1000 Da, as has been observed in other studies (Drikas et al., 2003a,b; Morran et al., 2004; Allpike et al., 2005; Boyer and Singer, 2005; Humbert et al., 2005; Singer et al., 2007; Mergen et al., 2009). In the example shown in Fig. 7, MIEX Coag removed more NOM than MIEX MF. This was consistent for all samples up to June 2006 (Period A and B) after which removal in the two MIEX treated trains was virtually equivalent up to the last sample analysed in March 2007. The greater removal of organics by MIEX Coag
than MIEX MF evident in the AMW profile has been attributed to the inclusion of the coagulation step, which would be expected to remove additional organics, particularly above 1000 Da. To illustrate the impact of treatment on the AMW distribution, the weight-average MW (Mw) achieved after each treatment was calculated for the period of study and is summarised in Fig. 8. The Mw of the raw water over the period of study followed a similar, albeit less pronounced, pattern to that observed with the VHA; an initial high Mw which decreased during Period A, increased slightly at the junction of Period A and B and then steadily decreased during period B and C of the study. The variation in Mw of all the treated waters also followed similar, but reduced, trends to those observed with the VHA. Fig. 8 shows that the Raw and Raw MF had virtually identical Mw. The removal of DOC by the MF was low (of the order of 0.4 mg/L) and, as apparent from the example in Fig. 7, observed differences in AMW were small, due in part to the use of tighter membranes for filtration before HPSEC analysis (0.22 mm for filtration of samples compared with 0.45 mm for the other organic analyses). Fig. 8 confirms the trends observed previously for other parameters and shows that DOC removal across the entire AMW range by the processes incorporating MIEX pre-treatment
1600
Weight Average Molecular Weight (M ) (Daltons)
treatment had a reduced VHA removal. The consistently high VHA removal compares well with the work of Mergen et al. (2008) where waters containing more than 50% hydrophobic DOC were found to have removals around 70% DOC with MIEX. However in their work MIEX applied in batch tests to water containing over 75% VHA was found to result in reduced DOC removal with repeated resin use whereas DOC removal in water containing only 20% VHA and 60% SHA remained constant. In our study, raw water VHA varied between 50 and 65% (assuming raw water VHA for Period A approximated that obtained with Raw MF) but the DOC removal with MIEX did not show a decline with time based on repeated resin use such as would occur for a proportion of the resin in a continuous process.
Period A
Period B
Period C
Raw Conv
1400
MIEX Coag Raw MF
1200
MIEX MF 1000 800 600 400 200 0 Aug-05
Oct-05
Dec-05
Feb-06
Apr-06
Jun-06
Aug-06
Oct-06
Dec-06
Fig. 8 e Weight average AMW (Mw) for all treatment processes.
1546
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 3 9 e1 5 4 8
3 2.5
1.25
VHA (m g/L)
Polydispersity (M /M )
1.3
1.2
1.15
1.1
1.5 1 0.5 0
1.05 p Raw
p Conv
p MIEX Coag
p Raw MF
p MIEX MF
Fig. 9 e Median of polydispersity achieved after each treatment processes. Error bars denote interquartile range.
resulted in the lowest Mw, around 630 Da, nearly half that of the raw water and significantly less than that achieved by the Conv treatment. The remaining Mw in the MIEX pre-treated waters concurs with the observation of Humbert et al. (2005) that the components of NOM that are refractory to MIEX (and even more to coagulation) are the lowest MW components. The data from this study provides more detailed data on the range of recalcitrant organics than that observed in other studies. The polydispersity (ratio Mw/Mn) achieved after each treatment was also calculated (Chin et al., 1994). Polydispersity is a measure of the homogeneity of the NOM in a sample, with values above 1 indicating the presence of a diverse, complex mixture of compounds. The median of the polydispersity in the raw water and after each treatment for the twelve samples is shown in Fig. 9. The raw water had the greatest polydispersity, which was unchanged after passage through the MF membrane. It is likely that the higher AMW organics that may be removed by the MF membrane (Fan et al., 2001; Lee et al., 2005) would also be removed by filtration through 0.22 mm filters prior to HPSEC analysis of both the Raw and Raw MF samples and therefore differences between their molecular weight profiles and consequently polydispersity would not be measurable. Conventional treatment (Conv) reduced the range of organics somewhat whilst the two processes incorporating MIEX pre-treatments had the lowest polydispersity confirming that these treatments had the greatest reduction in complexity and range of AMW of the organics. The smaller interquartile range for MIEX Coag in Fig. 9 also confirms that MIEX Coag was more effective in reducing the range of organics than MIEX MF. This was attributed to the removal of additional organics above 1000 Da by the coagulation step.
3.4.
2
Correlations
The different characterization techniques employed whilst measuring different components are not mutually exclusive. For example, SUVA has been shown to provide a measure of the extent of conjugation and aromaticity of water (Chin et al., 1994; Weishaar et al., 2003) which is likely to be associated with both higher AMW and hydrophobic compounds such as those predominant in the VHA fraction. Therefore it would be expected that there would be a relationship between the observed SUVA and the Mw and VHA of the various waters
0
200
400
600
800
1000
1200
1400
1600
Weight Average Molecular Weight Mw (Daltons)
Fig. 10 e Correlation of VHA and weight average AMW (Mw) for raw and treated waters (r2 [ 0.83; n [ 53).
studied. These relationships were studied utilising a linear regression function for the raw water and all of the various waters following treatment. It was found that SUVA was not strongly correlated to the amount of VHA present (r2 ¼ 0.68; n ¼ 53) or the Mw (r2 ¼ 0.63; n ¼ 57). The SUVA of the raw water alone showed a similar poor relationship with both VHA (r2 ¼ 0.55; n ¼ 8) and Mw (r2 ¼ 0.47; n ¼ 10) most likely due to the large variation in the SUVA of the raw water (these 10 samples varied from 5.1 to 1.8). However the relationship between SUVA and Mw was significantly improved when the raw data was excluded from the data set (r2 ¼ 0.85; n ¼ 47) supporting the relationship between molar absorptivity and weightaveraged molecular weight observed by Chin et al. (1994). The improved correlation of SUVA with Mw following removal of the raw data was due to removal of the vast spread in the SUVA data for very similar Mw; attributed to a significantly higher concentration of UV absorbing compounds within the same molecular weight range in some raw water samples. The relationship between SUVA and VHA was not impacted when raw water data was excluded from this data set because the data range was not markedly affected by the removal of the raw data as the range in VHA concentrations in the raw water was not significantly different to the range observed after treatment. It also suggests that SUVA is not an accurate measure of the organic character for all the waters studied and while it could be correlated with the Mw of the treated waters, or those without high levels of UV absorbing organics, it could not be related to the Mw of all waters nor could it be attributed solely to the VHA component of waters. There was, however, a strong relationship between VHA and Mw when using data from both the raw water and the waters following treatment (r2 ¼ 0.83; n ¼ 53) as illustrated in Fig. 10. This relationship was also retained when the raw water data was excluded from the data set (r2 ¼ 0.85; n ¼ 44) and improved when only the raw water data was considered (r2 ¼ 0.90; n ¼ 9). This confirms that the VHA fraction was the major component of the molecular weight distribution for both the raw and treated waters supporting the premise that the hydrophobic fraction of NOM obtained using XAD-8 resin comprises a large component of the UV absorbing NOM in this natural water and concurs with Chin et al. (1994) that the relative amount of aromatic moieties in aquatic fulvic acid increases with increasing molecular weight. The close correlation between these two parameters also supports the
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 3 9 e1 5 4 8
relevance of these two techniques as valuable tools to characterise NOM. However it should also be noted that NOM arising from sources such as algal blooms may be more hydrophilic and that the observed relationship between VHA and Mw may not hold.
4.
Conclusion
Pre-treatment with MIEX prior to both conventional treatment and microfiltration over a 2 year period was found to result in consistently greater removal of DOC with resultant lower SUVA in the treated water than the same processes without MIEX over the entire period of study. The processes incorporating MIEX produced more consistent water quality and were less affected by changes in the raw water DOC. This would make treatment plants incorporating MIEX easier to operate and maintain good quality water even when subjected to large variations in the raw water DOC concentration and character. Characterization of the organics indicated that the VHA fraction was the dominant fraction present in the raw water and this was best removed by the processes incorporating MIEX pre-treatment. The high proportion of VHA in this water and its preferential removal by MIEX pre-treatment resulted in a significant difference in the amount of DOC removed and in the resulting character of the waters after treatment. The lowest VHA concentration was achieved after treatment with processes incorporating MIEX, regardless of the raw water VHA concentration. The processes incorporating MIEX pre-treatment also removed organics across the complete AMW range resulting in lower weight average AMW in the treated water than the other treatment processes. Greater removal of organics by MIEX Coag than MIEX MF was evident in some of the specific AMW profiles and this has been attributed to the inclusion of the coagulation step, which would be expected to remove additional organics, particularly above 1000 Da. The two processes incorporating MIEX pre-treatments had the lowest polydispersity confirming that these treatments had the greatest reduction in complexity and range of AMW of the organics. The greater reduction in concentration of NOM and range of organics following treatment incorporating MIEX results in more stable water which will be less reactive to disinfectants. A strong correlation was found between the VHA and weight average AMW suggesting that the VHA fraction was a major component of the AMW for both the raw water and treated waters. SUVA was found to correlate well with the weight average AMW of the treated waters, but it could not be related to the weight average AMW of all waters nor attributed to the VHA component of the waters.
Acknowledgements The authors would like to thank the following people: Nick Nedelkov for technical assistance at Mt Pleasant WTP and Edith Kozlik, Miriam Nedic and Leanne Biddiss for analytical support.
1547
references
Allpike, B.P., Heitz, A., Joll, C.A., Kagi, R., Abbt-Braun, G., Frimmel, F., Brinkmann, T., Her, N., Amy, G., 2005. Size exclusion chromatography to characterize DOC removal in drinking water treatment. Environ. Sci. Technol. 39 (7), 2334e2342. Archer, A.D., Singer, P., 2006. Effect of SUVA and enhanced coagulation on removal of TOX precursors. J.AWWA 98 (8), 97e107. Baker, J., Lavinder, S., Fu, P., 1995. Removal of natural organic matter with anion exchange resins. AWWA Annu. Conf. Proc., 547e564. Bell-Ajy, K., Abbaszadegan, M., Ibrahim, E., Verges, D., LeChevallier, M., 2000. Conventional and optimized coagulation for NOM removal. J.AWWA 92 (10), 44e58. Bolto, B., Dixon, D., Eldridge, R., King, S., 2002. Removal of THM precursors by coagulation or ion exchange. Water Res. 36 (20), 5066e5073. Boyer, T.H., Singer, P.C., 2005. Bench-scale testing of a magnetic ion exchange resin for removal of disinfection by-product precursors. Water Res. 39 (7), 1265e1276. Boyer, T.H., Singer, P.C., 2006. A pilot-scale evaluation of magnetic ion exchange treatment for removal of natural organic material and inorganic anions. Water Res. 40 (15), 2865e2876. Brattebo, H., Ødegaard, H., Halle, O., 1987. Ion exchange for the removal of humic acids in water treatment. Water Res. 21 (9), 1045e1052. Buchanan, W., Roddick, F., Porter, N., Drikas, M., 2005. Fractionation of UV and VUV pretreated natural organic matter from drinking water. Environ. Sci. Technol. 39 (12), 4647e4654. Chin, Y.-P., Alken, G., O’Laughlin, E., 1994. Molecular weight, polydispersity and spectroscopic properties of aquatic humic substances. Environ. Sci. Technol. 28 (11), 1853e1858. Chow, C.W.K., Fabris, R., Drikas, M., 2004. A rapid fractionation technique to characterise natural organic matter for the optimisation of water treatment processes. J. Water Supply: Res. Technol. . AQUA 53 (2), 85e92. Chow, C.W.K., Fabris, R., Drikas, M., Holmes, M., 2005. A case study of treatment performance and organic character. J. Water Supply: Res. Technol. . AQUA 54 (6), 385e395. Chow, C.W.K., van Leeuwen, J.A., Drikas, M., Fabris, R., Spark, K. M., Page, D.W., 1999. The impact of the character of natural organic matter in conventional treatment with alum. Water Sci. Technol. Water Supply 40 (9), 97e104. Collins, M.R., Amy, G.L., Steelink, C., 1986. Molecular weight distribution, carboxylic acidity, and humic substances content of aquatic organic matter: implications for removal during water treatment. Environ. Sci. Technol. 20 (10), 1028e1032. Crozes, G., White, P., Marshall, M., 1995. Enhanced coagulation: its effect on NOM removal and chemical costs. J. AWWA 87 (1), 78e89. Dixon, M.B., Morran, J.Y., Drikas, M., 2010. Extending membrane longevity by using MIEX as a pre-treatment. J. Water Supply: Res. Technol. . AQUA 59 (2), 92e99. Drikas, M., Morran, J.Y., Pelekani, C., Hepplewhite, C., Bursill, D.B., 2002. Removal of natural organic matter - a fresh approach. Water Sci. Technol. Water Supply 2 (1), 71e79. Drikas, M., Chow, C.W.K., Cook, D., 2003a. The impact of recalcitrant organic character on disinfection stability, trihalomethane formation and bacterial regrowth - an evaluation of magnetic ion exchange resin (MIEX) and alum coagulation. J. Water Supply: Res. Technol. . AQUA 52 (7), 475e487. Drikas, M., Morran, J.Y., Cook, D. and Bursill, D.B, (2003b), Operating the MIEX process with microfiltration or
1548
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 3 9 e1 5 4 8
coagulation. In: Proceedings of the AWWA Water Quality Technology Conference, Philadelphia, USA, November 2003. Fabris, R., Lee, E.K., Chow, C.W.K., Chen, V., Drikas, M., 2007. Pretreatments to reduce fouling of low pressure micro-filtration (MF) membranes. J. Memb Sci. 289 (1e2), 231e240. Fan, L., Harris, J., Roddick, F., Booker, N., 2001. Influence of the characteristics of natural organic matter on the fouling of microfiltration membranes. Water Res. 35 (18), 4455e4463. Fearing, D.A., Banks, J., Guytand, S., Eroles, C.M., Jefferson, B., Wilson, D., Hillis, P., Campbell, A.T., Parsons, S.A., 2004. Combination of ferric and MIEX for the treatment of a humic rich water. Water Res. 38 (10), 2551e2558. Humbert, H., Gallard, H., Suty, H., Croue, J.P., 2005. Performance of selected anion exchange resins for the treatment of a high DOC content surface water. Water Res. 39 (9), 1699e1708. Humbert, H., Gallard, H., Jacquemet, V., Croue, J.P., 2007. Combination of coagulation and ion exchange for the reduction of UF fouling properties of a high DOC content surface water. Water Res. 41 (17), 3803e3811. Jarvis, P., Mergen, M., Banks, J., McIntosh, B., Parson, S.A., Jefferson, B., 2008. Pilot scale comparison of enhanced coagulation with magnetic resin plus coagulation systems. Environ. Sci. Technol. 42 (4), 1276e1282. Lee, N.H., Amy, G., Lozier, J., 2005. Understanding natural organic matter fouling in low-pressure membrane filtration. Desalination 178 (1e3), 85e93. Leenheer, J.A., 1981. Comprehensive approach to preparative isolation and fractionation of dissolved organic carbon from natural waters and wastewaters. Environ. Sci. Technol. 15 (5), 578e587. Mergen, M.R.D., Jarvis, P., Jefferson, B., Parsons, S.A., 2008. Magnetic resin treatment: impact of water type and resin use. Water Res. 42 (8e9), 1977e1988. Mergen, M.R.D., Adams, B.J., Vero, G.M., Prices, T.A., Parsons, S.A., Jefferson, B., Jarvis, P., 2009. Characterisation of natural organic matter (NOM) removed by magnetic ion exchange resin (MIEX Resin). Water Sci. Technol. Water Supply 9 (2), 199e205. Morran, J.Y., Bursill, D.B., Drikas, M., Nguyen, H., (1996). A new technique for the removal of natural organic matter,
Proceedings of the AWWA WaterTECH Conference, Sydney, Australia. Morran, J.Y., Drikas, M., Cook, D., Bursill, D.B., 2004. Comparison of MIEX treatment and coagulation on NOM character. Water Sci. Technol. Water Supply 4 (4), 129e137. Owen, D.M., Amy, G.L., Chowdbury, Z.K., Paode, R., McCoy, G., Viscosil, K., 1995. NOM characterisation and treatability. J.AWWA 87 (1), 46e63. Sharp, E.L., Jarvis, P., Parsons, S.A., Jefferson, B., 2006. Impact of fractional character on the coagulation of NOM. Colloids Surf. A Physicochem Eng. Asp 286 (1e3), 104e111. Singer, P.C., Bilyk, K., 2002. Enhanced coagulation using a magnetic ion exchange resin. Water Res. 36 (16), 4009e4022. Singer, P.C., Schneider, M., Edwards-Brandt, J., Budd, G.C., 2007. MIEX for removal of DBP precursors: pilot plant findings. J. AWWA 99 (4), 128e139. Singer, P.C., Boyer, T., Holmquist, A., Morran, J., Bourke, M., 2009. Integrated analysis of NOM removal by magnetic ion exchange. J. AWWA 101 (1), 65e73. van Leeuwen, J., Chow, C., Fabris, R., Withers, N., Page, D., Drikas, M., 2002. Application of a fractionation technique for the better understanding of the removal of NOM by alum coagulation. Water Sci. Technol. Water Supply. 2 (5e6), 427e433. van Leeuwen, J., Daly, R., Holmes, M., 2005. Modelling the treatment of drinking water to maximize dissolved organic matter removal and minimize disinfection by-product formation. Desalination 176 (1e3), 81e89. Warton, B., Heitz, A., Zappia, L.R., Franzmann, P.D., Masters, D., Joll, C.A., Alessandrino, M., Allpike, B., O’Leary, B., Kagi, R.I., 2007. Magnetic ion exchange drinking water treatment in a large-scale facility. J.AWWA 99 (1), 89e101. Weishaar, J.L., Aiken, G.R., Bergamaschi, B.A., Fram, M.S., Fujii, R., Mopper, K., 2003. Evaluation of specific ultraviolet absorbance as an indicator of the chemical composition and reactivity of dissolved organic carbon. Environ. Sci. Technol. 37 (20), 4702e4708. White, M.C., Thompson, J.D., Harrington, G.W., Singer, P.C., 1997. Evaluating criteria for enhanced coagulation compliance. J. AWWA 89 (5), 64e77.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 4 9 e1 5 6 0
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
The implications of household greywater treatment and reuse for municipal wastewater flows and micropollutant loads D. Michael Revitt a,*, Eva Eriksson b, Erica Donner a,1 a b
Urban Pollution Research Centre, Middlesex University, Hendon Campus, The Burroughs, London NW4 4BT, United Kingdom Department of Environmental Engineering, Technical University of Denmark, Miljoevej B113, Kgs. Lyngby, DK-2800, Denmark
article info
abstract
Article history:
An increasing worldwide interest in water recycling technologies such as greywater
Received 10 September 2010
treatment and reuse suggests that additional research to elucidate the fate of xenobiotics
Received in revised form
during such practices would be beneficial. In this paper, scenario analyses supported by
2 November 2010
empirical data are used for highlighting the potential fate of a selection of xenobiotic
Accepted 19 November 2010
micropollutants in decentralised greywater treatment systems, and for investigation of the
Available online 24 November 2010
possible implications of greywater recycling for the wider urban water cycle. Potential potable water savings of up to 43% are predicted for greywater recycling based on Danish
Keywords:
water use statistics and priority substance monitoring at a greywater treatment plant in
Greywater treatment
Denmark. Adsorption represents an important mechanism for the removal of cadmium,
Wastewater influent
nickel, lead and nonylphenol from influent greywater and therefore the disposal route
Recycling
adopted for the generated sludge can exert a major impact on the overall efficiency and
Priority substances
environmental sustainability of greywater treatment.
Scenario analyses
ª 2010 Elsevier Ltd. All rights reserved.
Sludge disposal
1.
Introduction
With pressures on potable water supplies continuing to increase worldwide, interest in the use of alternative water sources such as recycled wastewater is also growing (Chu et al., 2004; Bixio et al., 2006). In particular, greywater treatment and reuse is receiving increasing attention (e.g. Maimon et al., 2010; Liu et al., 2010). This is because greywater generally has a lower organic pollutant and pathogen content than combined municipal wastewater which also contains toilet waste (Eriksson et al., 2002). Thus, greywater is considered particularly suitable for on-site (i.e. decentralised) treatment and
reuse. Greywater treatment and reuse schemes have already been piloted in many countries around the world and are becoming increasingly commonplace in water stressed areas such as Australia and Mediterranean countries (Friedler and Gilboa, 2010; Masi et al., 2010; Pinto and Maheswari, 2010). However, related research has largely been restricted to studies of standard water quality parameters such as total organic carbon, biological oxygen demand, chemical oxygen demand and faecal and total coliforms (e.g. Pidou et al., 2008; Paulo et al., 2009). In contrast, there has been very little greywater research investigating the loads and dynamics of micropollutants. Nevertheless, Eriksson et al. (2002, 2003) and Palmquist and
* Corresponding author. Tel.: þ44 (0)20 8411 5308; fax: þ44 (0)20 8411 6774. E-mail address:
[email protected] (D.M. Revitt). 1 Present address: Centre for Environmental Risk Assessment and Remediation (CERAR), University of South Australia, Building X, Mawson Lakes Campus, Mawson Lakes, SA-5095, Australia. 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.11.027
1550
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 4 9 e1 5 6 0
Hanaeus (2005) have collectively shown that a large number of xenobiotic substances can find their way into greywater via bathroom and laundry products. Donner et al. (2010) have reported initial investigations into the fate of a range of pollutants within greywater treatment and reuse systems. However, given the increasing implementation of greywater recycling technology, it is evident that additional research to elucidate the behaviour of xenobiotic micropollutants during greywater treatment would be beneficial. It would also be useful to understand the potential implications of more widespread greywater recycling for urban wastewater loads and dynamics. Greywater treatment and reuse is a very diverse field, encompassing a wide range of potential treatment trains and spatial scales, as well as numerous reuse options (Li et al., 2009; Misra et al., 2010). Current treatment options vary widely in sophistication from simple filter systems to constructed wetlands, multi-stage biological treatment systems, and membrane bioreactors. Nevertheless, all systems are based on a combination of chemical, physical and biological processes such as adsorption, coagulation, precipitation, filtration, aeration, biodegradation, and disinfection. Reuse options cover a wide range of non-potable applications, from those involving a higher risk of human exposure such as spray irrigation and car washing, to lower risk options such as toilet flushing and sub-surface irrigation of non-food crops. Although pathogen transfer is generally considered the most pressing concern, it is nonetheless important to ensure that the lack of information regarding the chemical pollutant dynamics of greywater does not lead to the prevalence of suboptimal treatment trains or inappropriate reuse practices. This is currently being brought into focus with the development of national standards and codes of practice for both greywater treatment and specific reuse applications (e.g. in the UK and Australia). Fatta-Kassinos et al. (2010) have recently reviewed the practice of wastewater reuse for irrigation purposes and concluded that the benefits associated with improved water balances and nutritional levels need to be assessed against the current lack of knowledge relating to possible impacts on ecosystems and human health of the applied organic xenobiotics and heavy metals. In this paper, scenario analyses are used to highlight the potential fate of a selection of xenobiotics in decentralised greywater treatment systems, and to investigate the possible implications of greywater recycling for the urban water cycle. All of the substances investigated are listed under the European Water Framework Directive (WFD) (European Commission, 2000a) as ‘Priority Substances’ (PS) or ‘Priority Hazardous Substances’ (PHS) and are known to be present in greywater. A range of different greywater treatment and reuse scenarios are compared in order to ascertain the likely benefits/shortcomings of the different scenarios in terms of micropollutant persistence and fate, including the possible impacts on municipal wastewater flow dynamics and pollutant source control. Due to the limited availability of relevant data, the presented results focus on cadmium (Cd), nickel (Ni), lead (Pb), benzene and 4-nonylphenol (4-NP). Cadmium, Ni and Pb are metal pollutants of high concern in the municipal wastewater treatment process, as their tendency to accumulate in sludge can counteract its beneficial reuse for nutrient recovery and soil conditioning. For instance, national and European regulations
specify acceptable levels of metal pollutants in sludge destined for recycling to agricultural land (e.g. European Commission, 1986) and sludge not meeting those criteria must be disposed of via alternative means such as incineration or landfilling. Particular focus is given in this paper to the potential for greywater treatment to act as an emission control barrier for Cd. Recognised as a PHS under the WFD and highlighted as a major element of concern in relation to sludge quality, Cd is toxic to humans, has no known biological function and is one of the more mobile metals in soil. It is thus of particular concern in terms of crop uptake potential as it can pose health risks to humans and animals at levels well below phytotoxic concentrations (McLaughlin et al., 2000). Some sludge regulations (including the Danish national regulations) also specify acceptable levels of key organic pollutants, such as nonylphenols which have been found to accumulate in the sludge fraction during wastewater treatment (e.g. Abad et al., 2005; Koh et al., 2005). For contrast, benzene has also been included among the selected substances because being a relatively volatile substance, it tends to partition predominantly to air rather than sludge or water, and can thus be expected to demonstrate a differing behaviour during greywater treatment. Both benzene and 4-NP are resistant to biodegradation, as is typically the case for substances identified as PS/PHS. This investigation of the fate of selected greywater micropollutants facilitates a good overview of the possible implications of more widespread implementation of greywater reuse technologies.
2.
Materials and methods
2.1.
Greywater treatment at Nordhavnsga˚rden
The Nordhavnsga˚rden treatment plant is located in the basement of an apartment block in Copenhagen, Denmark, and consists of a primary settling tank, a three-stage rotating biological contactor (RBC), a secondary settling tank, a sand filter, an ultraviolet disinfection unit, and a service-water storage tank. Eighty-four one-bedroom apartments (w117 inhabitants) are connected to this facility which treats bathroom greywater for reuse as toilet flushing water and is automatic and self-cleaning.
2.2. Chemical analysis of PS and PHS in greywater and greywater treatment sludge The selected PS (benzene, Ni and Pb) and PHS (Cd and 4-NP) were measured both in hot and cold potable water, and in the influent and effluent greywater from the ‘Nordhavnsga˚rden’ greywater treatment system. Sixteen time-proportional samples of influent and effluent greywater were collected over a one-week period (29 November to 5 December 2007) using acid washed bottles. In addition, bottles used to collect samples for organic analysis were pre-heated at high temperature (220 C for 24 h). All samples (except for benzene analysis) were filtered prior to analysis (GF/A 1.6 mm for metal analysis and GF/C 1.2 mm for organics analysis). Cadmium, Ni, and Pb were analysed by Inductively Coupled Plasma - Optical Emission Spectroscopy (Varian Vista-MPX
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 4 9 e1 5 6 0
1551
Table 1 e Greywater treatment and reuse scenarios considered for this study. Scenario A B C D E F G H I J K L
Treatment system No treatment Indoor - RBC Indoor - RBC Indoor - RBC Indoor - RBC Indoor - RBC Indoor - RBC Indoor - RBC Indoor - RBC Indoor - RBC Outdoor - reedbed Outdoor - reedbed
Scenario analyses
The twelve greywater treatment and reuse scenarios investigated during this study are documented in Table 1. They range
Table 2 e Proportion of household water used for different domestic purposes (after Kjellerup and Hansen, 1994). Location/use of household water Bathrooms Laundry activities Kitchens Toilet flushing Irrigation a Average percentages in parenthesis.
Reuse of greywater
e
e Toilet Toilet þ Irrigation Toilet Toilet þ Laundry Toilet þ Irrigation Toilet þ Laundry þ Irrigation Toilet þ Laundry Toilet þ Irrigation Toilet þ Laundry þ Irrigation Groundwater recharge Groundwater recharge
Bathroom Bathroom Bathroom Bathroom Bathroom Bathroom Bathroom Bathroom Bathroom Bathroom Bathroom
CCD Simultaneous ICP-OES). Benzene was determined by purge and trap (Tekdyn Tekmar Velocity XPT Purge and Trap Sample Concentrator) and gas chromatography (Shimadzu Gas Chromatograph GC-14B, equipped with a Flame Ionization Detector). 4-nonylphenol was isolated and concentrated by solid phase extraction prior to analysis by GCeMS (Agilent 6890N GC-system with an Agilent 5973 Mass Selective Detector). All instrumental analyses were performed in triplicate. Quality control procedures included determination of detection limit, quantification limit, linearity, and precision. The detection limits for the employed analytical procedures were benzene (1.4 mg l1), 4-NP (0.005 mg l1), Cd (0.01 mg l1), Ni (0.1 mg l1) and Pb (0.03 mg l1). Internal reference materials were also included in all analyses for quality control purposes. The total greywater sludge was collected from the primary settling tank and rotating biological contactor on three occasions (separated by 4 monthly intervals) and was initially dewatered by centrifugation (4000 rpm for 20 min). The settled material was dried at 105 C for 1 h, pulverised and weighed, then acid digested (7 M nitric acid at 125 C and 2 atm for 30 min according to Danish Standards (DS259, 2003, DE/EN15586, 2004) prior to metal analysis by ICP-OES. The sludge was not analysed for benzene and 4-NP. Total solids (TS) were determined according to APHA et al. (2005) to facilitate normalisation of the sludge metal content to the concentration per unit of dry weight (DW).
2.3.
Source of greywater
Range and average percentagesa 35e37 (36) 13e15 (14) 17e25 (21) 20e27 (23) 5e7 (6)
þ Laundry þ Laundry þ Laundry þ Laundry þ Laundry þ Kitchen þ Laundry þ Kitchen þ Laundry þ Kitchen þ Laundry
from a baseline scenario of no treatment and no reuse (Scenario A) to full household greywater treatment and recycling (Scenario J; bathroom, laundry and kitchen greywater treated and reused for toilet flushing, laundry washing and irrigation). The identified scenarios differ in terms of the type of treatment plant (e.g. an indoor system using an RBC system and outdoor land-based treatment systems using reedbeds), in terms of the source of the greywater being treated (e.g. bathroom vs. bathroom þ laundry) and in terms of the enduse of the recycled water (e.g. toilet flushing vs. toilet flushing þ laundry washing). In practice, bathroom greywater is the fraction most commonly recycled and this is the reason for the relative dominance of this fraction in the selected scenarios (Table 1).
2.4. Water use statistics and input data to scenario analyses The scenario analyses reported in this paper are based on Danish water use statistics. The potential effects of greywater recycling on wastewater flows under the different scenarios (assuming that 100% implementation of greywater recycling technology is practised) have been calculated based on an average Danish potable water consumption of 119 l person1 day1 and a 43% contribution from households to the influent of municipal wastewater treatment plants (DANVA, 2007). The other major inputs to wastewater treatment plants are from industrial and commercial wastewater, stormwater and sewer infiltration. The proportion of household water used for different domestic purposes (Kjellerup and Hansen, 1994) is identified in Table 2. Similar distributions have been reported by Memon and Butler (2006) for residential properties in the UK although with an increased proportion for toilet flushing and a reduced percentage for general bathroom use.
2.5.
Pollutant fate analysis
The fate of the selected substances during greywater treatment and reuse has also been evaluated under the different scenarios. Hypothetical pollutant removal efficiencies of 10%, 50% and 90% were used for the pollutant fate calculations in order to cover a broad range of potential treatment situations. With such a broad range of treatment systems potentially available and little attention given to optimising these
1552
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 4 9 e1 5 6 0
Table 3 e Nordhavnsga˚rden monitoring data used in the scenario calculations, and other relevant data from the literature (all values in mg lL1). Cd
Ni
Pb
Benzene
4-NP
Greywater influent concentration (Danish and Swedish greywater literature data)
Range: 0.01e0.22 Mean: 0.08 Mediana: 0.07 Rangeb: 0.06e0.66 Meanb: 0.22 2.5c < 0.1d Rangee: 0.06e0.16 Meane: 0.10
Range: 5.15e26.5 Mean: 9.32 Median: 6.76 Rangeb: 3.86e10.2 Meanb: 6.2 1.3c 1.5d Rangee: 4.45e28.1 Meane: 11.0
Range: 4.89e10.2 Mean: 6.95 Median: 6.82 Rangeb: 1.1e6.9 Meanb: 3.4 1.8c <2d Rangee: 2.14e3.14 Meane: 2.52
Range: <1.4e9.85 Meana: 3.61 Mediana: 2.51 All values 1.9b
Range: 0.35e1.63 Mean: 0.90 Median: 0.90 All values <0.5b,g
Potable water concentration (Nordhavnsga˚rden) Concentration in Copenhagen potable water abstraction wellsg
Cold water: <0.01 Hot water: <0.01 Range: 0.03e0.07 Mean: 0.04
Cold water: 0.24 Hot water: 0.35 Range: 0.46e8.9 Mean: 2.21
Cold water: 7.27 Hot water:6.21 Range: <0.03e0.11 Mean: 0.22
Cold water: <1.4 Hot water: <1.4 All values <1.4
Influent concentration (Nordhavnsga˚rden) (n ¼ 8)
0.76c,g 0.9d,g Rangee: 2.85e5.95 Meane: 3.8 Rangef: 0.56e1.1 Meanf: 0.76 No data All values <0.5
a 38% of the values for benzene were below the detection limit; for the purposes of calculating mean and median values these were assumed to be equal to half of this value (i.e. 0.7 mg l1 for benzene). b BO90 (apartment block), Copenhagen, Denmark (Ledin et al., 2006). c Gals Klint (campingsite), Denmark (Nielsen and Pettersen, 2005). d Vestbadet I/S, Denmark (Andersson and Dalsgaard, 2004). e Vibya˚sen (housing area), Sollentuna, near Stockholm, Sweden (Palmquist and Hanaeus, 2005). f Gebers (apartment block), Skarpnack, near Stockholm, Sweden (Palmquist, 2004). g Indicates that a measurement includes not only 4-NP but nonylphenols collectively.
systems for micropollutant removal it is prudent to conclude that many systems may have limited effectiveness in terms of non-standard parameters. Pollutant load data used for the pollutant fate calculations have predominantly been based on the Nordhavnsga˚rden data presented in this paper. However, only bathroom greywater is recycled at the Nordhavnsga˚rden site. Thus, in order to facilitate Cd fate calculations for the full suite of scenarios (Scenarios A-L), additional data on greywater Cd loads for kitchen and laundry greywater was taken from Wall (2002) and Bergstrom (2007) and the Cd load in blackwater (i.e. toilet wastewater including faeces and urine) was taken from Palmquist and Hanaeus (2005). These studies were conducted in Swedish households. As measured data for laundry and kitchen greywater were not available for benzene, 4-NP, Ni, and Pb only those scenarios involving bathrooms as the source of greywater (Scenarios B and C) have been investigated for these pollutants but a complete scenario analysis has been conducted for Cd. The physicochemical characteristics of the different pollutants have been taken into account in assessing their removal behaviour during the greywater treatment process. For the metals and their compounds the main removal process will be adsorption with negligible removal by biodegradation and no susceptibility to volatilisation. A precise assessment of metal adsorption capability is difficult due to the variety of compounds and complexes which can exist in wastewater samples but in a review of the potential of metals to be removed from stormwater, Revitt et al. (2008) have identified the highest adsorptive removal to be associated with Pb followed by Ni and with Cd demonstrating the lowest removal potential. The behaviours of benzene and 4-NP can be correlated with the relevant physiochemical parameters such as adsorption coefficients, biodegradation
half-lives and Henry’s Law constant for volatilisation (Scholes et al., 2007). These parameters suggest equal, but limited, susceptibilities for both pollutants to aerobic biodegradation but clear differences with regard to adsorption and volatilisation. Benzene is predicted to have a high potential to be removed by volatilisation compared to the moderate removal for 4-NP and the reverse is true for adsorption although to a less exaggerated extent.
3.
Results and discussion
3.1.
Priority substances in greywater
A summary of relevant pollutant monitoring data for greywater influent to the Nordhavnsga˚rden treatment plant is given in Table 3. All of the selected PS/PHS were detected at measurable concentrations and the results are generally comparable to existing Danish and Swedish greywater monitoring data for these substances (also given in Table 3), with some exceptions such as the high concentration of Cd (2.5 ug l1) measured at the Gals Clint camping site (Nielsen and Pettersen, 2005). However, a high level of consistency is not to be expected given that greywater flows and pollutant loads are inherently variable and highly dependent on the behaviour of individuals. In addition to the concentrations of the selected PS/PHS in greywater, measured values for these substances in the potable water at Nordhavnsga˚rden, and in the abstraction wells used to supply the potable water distribution network in Copenhagen (Copenhagen Energy, 2008a, 2008b) are also presented in Table 3. The abstraction well data clearly demonstrate the low background levels of the monitored substances.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 4 9 e1 5 6 0
Fig. 1 e a and b Diagrammatic representation of water flow for Scenarios A and J (dashed borders indicate water use options which are not relevant to that particular scenario).
3.2.
Flow calculations
Based on monitored greywater inflow rates and the Danish water use statistics specified in Section 2.4, effluent flow rates (expressed as litres per person per day; 1 p1 d1) have been calculated for each of the identified scenarios. Fig. 1a and b provide diagrammatic representations of the flow pathways associated with Scenarios A and J and serve as examples of the method by which the proportional potable water savings and the proportional reductions in wastewater treatment plant effluent in columns 2 and 3, respectively of Table 4 were derived. It can be seen that under the baseline conditions represented by Scenario A (i.e. no greywater treatment followed by reuse but direct use of greywater for irrigation purposes) a daily potable water use of 119 l p1 d1 results in 111.9 l p1 d1 of household wastewater being released to the municipal wastewater system. In contrast, under Scenario J (where bathroom, laundry and kitchen greywater are treated and reused for irrigation, laundry washing and toilet flushing), the effluent volume is reduced to 60.7 l p1 d1, representing a reduction in the effluent to the municipal wastewater treatment plant (WWTP)
1553
of 20% (when the 43% contribution of households to this wastewater stream is taken into account). This scenario also achieves a potable water saving of 51.2 l1p1 d1 due to the use of greywater for toilet flushing, the continued recycling of laundry effluents through the greywater treatment system and avoidance of using potable water for irrigation. The effective water use is 67.8 l1p1 d1 which amounts to a saving of 43% compared to the baseline situation represented by Scenario A. The calculations for Scenario J (Fig. 1b) also show that 33.3 l1p1 d1 of treated greywater will be produced for which there is no identified reuse application. This would represent an inefficient use of treatment resources and the described scenario analysis approach therefore offers a route for optimising the treated volumes according to user requirements. The flow calculation results provided in Table 4 demonstrate the implications of the different scenarios in terms of both potential potable water savings and reduced wastewater influent volumes at municipal WWTPs. Significant potable water savings (up to 43% for the described scenarios) can be achieved by recycling greywater. However, subsequent reductions in wastewater flows to large-scale municipal WWTP are predicted to be more modest (up to 27% for Scenario K) as the assumption has been made that only 43% of the total WWTP influent volume is derived from households (DANVA, 2007). The most beneficial combination of potable water savings and WWTP influent reductions are achieved when the volume of recycled water is sufficient to cover the requirements for toilet flushing, laundry washing, and outdoor irrigation uses (e.g. Scenarios G and J). It is important to note however that these impacts have been calculated on the basis of 100% uptake of the relevant greywater recycling scenario. Whilst this is feasible for new developments (or large-scale refurbishments), particularly in water stressed countries where water recycling regulations on new-builds are increasingly likely to be introduced, it should be recognised that implementation of greywater reuse in more established built environments without existing dual reticulation plumbing systems is likely to remain much lower than 100%.
3.3. Micropollutant fate during greywater treatment and reuse For each indoor treatment and reuse scenario (Scenarios A-J), the fates of the pollutants have been calculated based on hypothetical greywater treatment removal efficiencies of 10%, 50% and 90%. These hypothetical removal efficiencies span the wide range anticipated for the available treatment options of varying sophistication which can be expected to differ substantially in their ability to remove micropollutants. For example, losses due to volatilisation are likely to be greater in systems incorporating rotating biological contactors, than in simple filtration systems without additional aeration and will therefore exert the greatest influence on the removal of benzene. Treatment systems also vary widely in their ability to remove suspended solids and adsorbed pollutants from greywater (Donner et al., 2010). This is a process which has been identified as being important for the removal of Pb and 4NP. The composition and condition of the microbial
1554
Table 4 e Implications of Scenarios A-L for municipal wastewater flows and Cd loads, assuming on-site greywater treatment Cd removal efficiencies of 10%, 50% and 90%. Scenario
Reduction in WWTP influent (%)
Reduction in Cd load to WWTP based on 10% removal efficiencya Assuming sludge is discharged to WWTP
A B Cb D E Fb Gb H Ib Jb K L
e 23 29 23 37 29 43 37 29 43 0 0
e 11 13 11 17 13 20 17 13 20 27 17
e 0 0.45 (2.2%) 0 0 0.82(4.1%) 1.25(6.2%) 0 0.69(3.4%) 0.84(4.2%) N/A N/A
Assuming sludge is removed from WW stream e 0.31 0.75 0.77 1.19 1.59 2.28 1.13 1.62 1.97 N/A N/A
(1.5%) (3.8%) (3.8%) (5.9%) (7.9%) (11.3%) (5.6%) (8.1%) (9.8%)
Reduction in Cd load to WWTP based on 50% removal efficiencya Assuming sludge is discharged to WWTP
Assuming sludge is removed from WW stream
e 0 0.25 (1.2%) 0 0 0.46 (2.3%) 0.60 (3.0%) 0 0.38 (1.9%) 0.42 (2.1%) N/A N/A
e 1.53 (7.6%) 1.78 (8.9%) 3.85 (19.1%) 4.56 (22.7%) 4.31 (21.5%) 5.09 (25.3%) 5.16 (25.7%) 5.02 (25.0%) 5.58 (27.8%) N/A N/A
a Main value given is the reduction in load in mg p1 d1; values in brackets show the reduction in load as a percentage of the total household load). b Values given show the reduction in load after 5 cycles of the given scenario (i.e. laundry water recycled 5 times).
Reduction in Cd load to WWTP based on 90% removal efficiencya Assuming sludge is discharged to WWTP e 0 0.05 0 0 0.09 0.11 0 0.08 0.08 N/A N/A
(0.2%)
(0.4%) (0.5%) (0.4%) (0.4%)
Assuming sludge is removed from WW stream e 2.74 2.80 6.93 7.15 7.02 7.24 8.53 8.43 8.60 N/A N/A
(13.5%) (13.9%) (34.5%) (35.6) (35.0%) (36.1%) (42.5%) (42.0%) (42.8%)
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 4 9 e1 5 6 0
Potable H2O saving (%)
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 4 9 e1 5 6 0
Scenario B Irrigation Irrigation Bathroom
0 µg p d
[A]
3.04 µg p d
Greywater Treatment Plant
Laundry 0 µg p d
3.04 µg p d
Potable water
Laundry 4.65 µg p d
Sludge 2.74 µg p d
[B]
[D]
Toilet 11.16 µg p d
Kitchen 1.58 µg p d
[C]
Surplus 0.11 µg p d
Toilet 0 µg p d
Potable H2O saving = 27 l p d (23 %) WWTP influent reduction = 11 %
Municipal [F] Wastewater Treatment Plant
[E]
20.23 µg p d
Fig. 2 e Diagrammatic representation of Scenario B and associated Cd load calculations (based on 90% removal efficiency during treatment) as described in Box 1. Letters in square brackets can be used to match with the associated calculation in Box 1.
community or biofilm in biological systems will significantly affect the biodegradation potential for organic micropollutants (Donner et al., 2010; Giri et al., 2006) and has been identified as being equally important for the removal of both benzene and 4-NP. Biological greywater treatment systems can take some time to mature and establish reliable performance and may be inhibited by pollutant shock loadings, such as a predominance of bleach or other cleaning products. Treatment efficiencies can be expected to vary over time and the use of hypothetical removal efficiencies of varying effectiveness is thus a useful approach for providing an overview of the possible impacts of different greywater treatments and reuse scenarios on the wider urban water cycle. In Table 4 the results of the Cd fate calculations for the full range of scenarios are presented. These results also demonstrate how two different hypothetical pathways for sludge disposal will influence the influent Cd load to a WWTP. One set of calculations are based on the assumption that the greywater treatment sludge will be discharged or transferred periodically to the municipal WWTP (as is in fact most commonly the case) with the second set of calculations being designed to investigate the effect of employing a separate sludge disposal route (such as disposal to land). As an example of the manner by which pollutant pathways have been evaluated for the different scenarios, the fate of household-derived Cd pollution under Scenario B (see Fig. 2) is described in detail in Box 1. The different steps in the calculation can be matched to the scenario diagram by means of the square bracketed letters in both Fig. 2 and Box 1. According to Scenario B, bathroom greywater is treated on-site using an RBC and reused for toilet flushing, and the results show that treatment and reuse according to this scenario will have no positive effect on WWTP Cd influent loads unless the sludge is removed from the wastewater stream entering the associated WWTP (Table 4). Furthermore, even under conditions of separate sludge disposal, the greatest potential decrease in Cd
1555
loading at the treatment plant will be 2.74 mg p1 d1 (assuming 90% removal efficiency during treatment and 100% implementation of Scenario B). Compared to the baseline scenario (Scenario A) which incorporates no greywater treatment and reuse, this represents a fairly minor overall reduction (13.5%) on the influent Cd load at the WWTP, as baseline calculations indicate a total household load of 20.2 mg p1 d1. It is clear that the incorporation of Cd in the sludge is a critical pathway in controlling the fate of this and similar pollutants. In those situations where the sludge from the greywater treatment process is eventually discharged or transferred to a WWTP, there will be no overall Cd removal unless the scenarios incorporate removal of some of the treated greywater from the municipal wastewater stream by using it for irrigation purposes (i.e. Scenarios C, F, G, I and J). When irrigation is practised, it is interesting to note that the impact on the WWTP load is not consistent with the increasing treatment efficiency of the greywater plant. Thus for Scenario C, it can be seen that the overall removal of Cd from the wastewater stream in terms of the decrease in total household load arriving at the WWTP decreases from 2.2% to 1.2%e0.2% as the applied greywater treatment efficiencies increase from 10% to 50%e90% (Table 4). This can be explained by the fact that the higher treatment removal efficiencies (i.e. 50% and 90%) produce treated greywater with lower Cd concentrations, and hence the proportion of Cd removed from the total WWTP system due to losses via irrigation is reduced. If it is feasible to remove the sludge produced by the greywater treatment system from the external wastewater stream, it can be seen that all scenarios (other than A, K and L) produce overall Cd removal efficiencies which are consistent with the expected results based on the applied greywater treatment values. For 10% greywater treatment efficiency, the most efficient overall Cd removal is demonstrated by Scenario G (11.3%) whereas for the higher greywater treatment performances Scenario J proves to be most efficient (27.8% and 42.8%). Scenarios G and J both involve continuous recycling of laundry greywater and the results in Table 4 are based on predictions after the completion of 5 cycles. All scenarios incorporating laundry water recycling (Scenarios E, G, H and J) involve micropollutants being continually added to the system and the wastewater being continually circulated and treated for reuse. The calculations indicate that the Cd concentration in these systems initially increases but approaches an equilibrium situation with regard to the greywater Cd loading and an optimal removal efficiency is established within 5 cycles or less. This suggests that there should not be any detrimental impact on washing machine functioning due to micropollutant build-up although the elevated pH levels during typical laundry washing may encourage the precipitation of some constituents and corrosion may occur due to increased salinity. The annual influent loads of Cd, Ni, Pb, benzene and 4-NP to the Lynetten WWTP, which services the area of Copenhagen where the Nordhavnsga˚rden greywater treatment plant is located, are 21 kg, 386 kg, 1064 kg, 12.6 kg and 178 kg (Lynettefællesskabet I/S, 2008). Because of the differences in influent flows (5.7 m3/year to Nordhavnsga˚rden greywater treatment plant compared to 74 million m3/year to the WWTP), the contributions deriving from untreated Nordhavnsga˚rden
1556
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 4 9 e1 5 6 0
Box 1 Cadmium fate calculations for greywater treatment and reuse according to Scenario B (based on 90% removal efficiency).
[A] With an estimated bathroom greywater flow rate of 42.8 l p1 d1 (based on DANVA (2007) and Kjellerup and Hansen, 1994) and a median measured Cd concentration in the Nordhavnsga˚rden bathroom greywater of 0.071 mg l1, the median Cd load in untreated bathroom greywater is 3.04 mg p1 d1. [B] Assuming a greywater treatment removal efficiency of 90%, the maximum effluent Cd loading will be 0.30 mg p1 d1. The remaining Cd (2.74 mg p1 d1) will be entrained in the sludge produced by the greywater treatment system. The greywater treatment effluent has a Cd concentration of 0.0071 mg l1 (0.30 mg p1 d1 O 42.8 l p1 d1). [C] As with most treatment systems of this type the sludge produced at the Nordhavnsga˚rden treatment plant is periodically transferred directly to the municipal WWTP without further pre-treatment. [D] The Cd loading in the treated water used for toilet flushing is 0.19 mg p1 d1 (27.4 l p1 d1 0.0071 mg l1). Additionally, Cd could be added due to the addition of faeces and urine at this stage. Based on published measurements of Cd in blackwater (Palmquist and Hanaeus, 2005) it is estimated that the concentration of Cd in toilet wastewater would be 0.4 mg l1. Therefore, in a volume of 27.4 l, the maximum Cd loading contribution from the addition of blackwater would be 10.96 mg p1 d1. Hence, the total Cd load which would be discharged to the WWTP upon toilet flushing is 11.15 mg p1 d1 (0.19 þ 10.96 mg p1 d1). [E] Under Scenario B, surplus greywater treatment effluent (i.e. treated greywater not required for toilet flushing) will be discharged directly to the WWTP. The surplus flow rate is 15.4 l p1 d1 and the Cd concentration is 0.0071 mg l1 which equates to a Cd loading of 0.11 mg p1 d1. [F] The total Cd load discharged to the WWTP after greywater treatment and reuse is 14.00 mg p1 d1 (2.74 þ 11.15 þ 0.11). The three contributing sources to this Cd load are sludge [C], reused water after toilet flushing [D] and surplus treated water [E]. Under this scenario, additional household Cd releases will also occur due to laundry washing or kitchen activities as these waste streams are discharged directly to the WWTP. The relevant Cd loads from these sources are estimated to be 4.65 mg p1 d1from the laundry greywater and 1.58 mg p1 d1 from kitchen greywater (1.16 mg p1 d1 for dishwashing þ 0.26 mg p1 d1 from sink wiping þ 0.16 mg p1 d1 from food preparation) (Wall, 2002). Therefore the total Cd load to the wastewater treatment plant would be 20.23 mg p1 d1 (14.00 þ 4.65 þ 1.58). Impact: The total household Cd load without greywater treatment (Scenario A) is estimated to be 20.23 mg p1 d1 (comprising 3.04 mg p1 d1 from bathroom greywater, 4.65 mg p1 d1 from laundry greywater, 1.58 mg p1 d1 from kitchen greywater, and 10.96 mg p1 d1 from toilet wastewater). Therefore, as expected, under Scenario B there will be no decrease in Cd loading going to the WWTP unless the greywater sludge is removed from the system and treated separately. If this was practised, it would equate to a decrease in WWTP influent Cd loading of 2.74 mg p1 d1 and a potential overall per capita Cd removal efficiency of 13.5%.
greywater are very low, typically of the order of 0.001%. Therefore, clearly in terms of assessing the benefits which could be accrued by comprehensive application of greywater treatment, it is more realistic to compare per capita pollutant reductions. On this basis, the results reveal that full implementation of the most effective scenario (i.e. Scenario J with full greywater treatment and recycling and separate sludge disposal) could lead to a calculated reduction in the Cd load to the WWTP of 8.6 mg p1 d1 which is equivalent to a reduction of 14.1% of the overall Cd influent load at the WWTP (61 mg p1 d1). Although this is relatively low, it is apparent that in areas of low industrial activity and/or with separate stormwater treatment (i.e. where household wastewater is the major contributor to the municipal WWTP influent), the introduction of greywater treatment and reuse technologies may be beneficial in terms of pollutant emission control as well as water conservation. Clearly, the magnitude of the emission control function in relation to micropollutants will be highly dependent on the greywater sludge disposal pathway. The results presented in Tables 4 and 5 show that even when greywater treatment removes a substantial proportion of micropollutants from influent
greywater, for elemental pollutants such as Cd, Ni and Pb and for hydrophobic substances such as 4-NP the resulting impact at the WWTP is highly dependent on the fate of the greywater treatment sludge. In Table 5, the results derived for the bathroom greywater reuse scenarios are presented for two metals (Ni and Pb) and two organic micropollutants (benzene and 4-NP), respectively. Both metals follow similar trends to those described for Cd although with considerably elevated loading values. The magnitude of the differences in pollutant reductions according to the disposal route of the greywater treatment sludge are indicative of the adsorption potentials of different pollutants and are clearly less significant for benzene for which volatilisation plays an important role in controlling pollutant removal from the aqueous phase. The results for benzene and 4-NP shown in Table 5 have been informed by apportioning the contributions to the different removal processes during greywater treatment according to the distribution calculated using a pollutant fate model for an activated sludge WWTP (STPWIN, EPI Suite v3.20, US EPA, 2007). As expected from a consideration of the physicochemical properties, only 1.1%
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 4 9 e1 5 6 0
Table 5 e Implications of Scenarios A-C for Ni, Pb, benzene and 4-nonylphenol loads in bathroom greywater treatment sludge and household wastewater, assuming greywater removal efficiencies of 10%, 50% and 90%. Reduction in load to WWTP (mg p1 d1)a Scenario A Scenario B Scenario C Ni
Pb
Benzene
4-NP
10% removal
e
50% removal
e
90% removal
e
10% removal
e
50% removal
e
90% removal
e
10% removal
e
50% removal
e
90% removal
e
10% removal
e
50% removal
e
90% removal
e
0 (29.1) 0 (145.3) 0 (261.6) 0 (29.3) 0 (146.6) 23.87 (170.5) 10.58 (10.8) 52.89 (53.8) 95.20 (96.8) 0.03 (3.9) 0.17 (19.4) 0.31 (34.8)
42.59 (71.7) 23.66 (169.0) 4.73 (266.3) 42.98 (72.3) 0 (263.9) 4.76 (268.7) 26.33 (26.5) 61.64 (62.5) 96.95 (98.5) 5.70 (9.5) 3.32 (22.5) 0.94 (35.6)
a Main value given is the reduction in load in mg p1 d1 assuming the greywater treatment sludge is discharged to the WWTP; values in brackets show the reduction in load assuming the greywater treatment sludge is removed from the wastewater stream. Removal due to sorption, volatilisation and biodegradation is apportioned according to the distribution calculated using STPWIN (EPI Suite v3.20, US EPA, 2007).
of benzene is predicted to be removed by adsorption to sludge with volatilisation representing the major removal route (67.8%) in an overall removal capability of 68.9%. This raises concerns regarding the overall environmental effectiveness of greywater treatment as an emission control barrier for benzene. In contrast, 4-NP which has a low volatility (<1% removal by volatilisation) is predicted to partition predominantly to the sludge (90% removal by adsorption) and therefore behaves in a similar way to the metals placing the fate of this pollutant firmly on the adopted sludge disposal route during greywater treatment. Both benzene and 4-NP are identified as possessing low potentials for removal by biodegradation (<1%). Scenarios K and L investigate the potential implications of land-based greywater treatment systems. Under these scenarios, the greywater is treated using reedbed technology resulting in advantageous overall reductions in terms of the municipal WWTP influent pollutant load, but also raising concerns regarding the possible environmental impacts. For example, under Scenario K, the removal of bathroom greywater for treatment in a reedbed equates to a decrease in Cd
1557
WWTP influent loading of 3.04 mg p1 d1. Therefore, the reduction in Cd being directed to the WWTP due to this greywater treatment scenario is 15.0%. According to Scenario L, in which both bathroom and laundry greywater are treated, the corresponding reduction in WWTP influent load is 38.4%. In both cases, it is important to consider the environmental implications. Depending on the substrate of the treatment system, Cd may build up in the sediment/soil/solid phase over time and may also leach through to the groundwater. For the Nordhavnsga˚rden greywater treatment plant the annual release of Cd to the environment would be 130.4 mg and 329.0 mg for Scenarios K and L, respectively. A median wet weather removal efficiency of 84.7% has been measured for Cd passing through a sub-surface constructed wetland (Revitt et al., 2004). If applied to Scenario K this would indicate that a discharge loading of 3.04 mg p1 d1 could be reduced to 0.46 mg p1 d1 after passing through an appropriately designed vegetated greywater treatment plant. Given the hydraulic loading rate of 42.8 l p1 d1, this corresponds to a discharge concentration of 0.011 mg l1 which is well below the proposed AA-EQS value (European Commission, 2008) for Cd for the most sensitive inland surface waters (0.08 mg l1) before any dilution has occurred within the receiving water. By contrast for Scenario L, the discharge of 7.69 mg p1 d1 at a hydraulic loading of 59.5 l p1 d1 corresponds to a discharge concentration of 0.13 mg l1. Treated greywater with this Cd concentration would require an appropriate dilution on entering the receiving water. More critically, if discharged to ground the adsorption characteristics of the soil would need to ensure that appropriate protection existed for an underlying aquifer.
3.4.
Sludge fate and pollutant loading
One of the major drivers for further reducing micropollutant influent loads to municipal WWTPs is to facilitate the beneficial reuse of sewage sludge (i.e. biosolids) for soil conditioning of agricultural land. The European Directive most pertinent to the agricultural use of sewage sludge is Directive 86/278/EEC (European Commission, 1986) which establishes concentration limits for a number of metals that are typically present within sludge. The concentration limits are effectively ceiling limits, meaning that if sludge exceeds the metal concentration limit for any of the listed metals it should not be permitted for land application. Directive 86/278/EEC is currently under revision and the working draft for the revised Directive indicates that future limits will be more conservative (European Commission, 2000b). To enable some member states to achieve the new limits, it is probable that water companies will need to further tighten trade effluent consents for industries as well as seeking further means of reducing WWTP influent loads of key pollutants. The alternative would be an unwanted reduction in land recycling of sludge and a waste of a potentially beneficial resource. Currently, some member states, including Denmark, impose more stringent requirements than those in the EC Directive. For example, the current limit for Cd in the Danish regulations is 0.8 mg/kg DW compared to 20 mg/kg DW in the EC Directive and for nonylphenols the Danish value of 10 mg/kg DW is considerably
1558
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 4 9 e1 5 6 0
Table 6 e Measured concentrations of Cd, Ni and Pb in Nordhavnsga˚rden greywater treatment sludge and Danish wastewater treatment plant sludge, together with Danish and European sludge guideline limits for the relevant substances. All values are given in mg kgL1 DW. Substance
Measured concentration in Nordhavnsga˚rden primary settling tank sludgea
Concentration in Danish WWTP sludgeb
Cd
Range: 0.7e1.2 Mean: 1.0 Median: 1.1 Range: 22e35 Mean: 27 Median: 24 Range: 34e45 Mean: 37.7 Median: 34.0 No data
1995: 1.5 (0.8e6.0) 2002: 1.3 (0.3e3.2)
Ni
Pb
Nonylphenols
0.8
1995: 25.7 (10e141) 2002: 20 (11e42) 1995: 72 (26e155) 2002: 50 (11e96) 1995: 8 (0.3e61) 2002: 4 (1e25)
Danish sludge guideline limits (mg kg1)c
European sludge guideline limitsd, f 20e40
Proposed European sludge guideline limits in working drafte, f 10
30
300e400
300
120
750e1200
750
10
N/A
50
a n ¼ 3, 1 sample was taken from the primary settling tank and 2 samples were taken from the biological treatment module. b Values given are derived from a national survey of sludge quality in Danish WWTPs and are shown as median values, with the 5th and 95th percentiles in brackets (Jensen and Jepsen, 2005). c Cited in Jensen and Jepsen (2005). d Directive 86/278/EEC (European Commission, 1986). e Working document on sludge, 3rd draft (European Commission, 2000a,b). f Limit value applies to the substances nonylphenol and nonylphenolethoxylates with 1 or 2 ethoxy groups.
lower than a proposed European sludge guideline limit of 50 mg/kg DW (Table 6). Measured concentrations in the greywater treatment plant sludge from Nordhavnsga˚rden are provided in Table 6. The measured metal concentrations in the Nordhavnsga˚rden greywater treatment sludge confirm that adsorption to suspended solids is an important removal process for these substances during treatment. With median sludge concentrations of 1.1, 24 and 34 mg kg1 DW for Cd, Ni and Pb respectively it is evident that removal of greywater treatment sludge from the WWTP influent load could assist in the reduction of metal loadings in municipal WWTP sludge. The separate treatment and disposal of greywater sludge is an attractive prospect because it is unlikely to contain a significant nutrient content, and yet does effectively concentrate unwanted substances such as metals and nonylphenols. The separation of the greywater treatment sludge from community scale treatment and reuse systems is feasible and could effectively reduce WWTP sludge metal loads without significantly impacting on sludge nutrient value. In contrast, sludge separation from single household system designs is unlikely to be practical and currently these systems are typically designed to periodically backwash or flush particulate matter to the sewerage system.
4.
Conclusions
The results of the conducted scenario analyses are important in the face of increasing pressures on potable water supplies, showing that greywater recycling can potentially save significant volumes of potable water. Within a greywater treatment plant, the dominant removal process for a particular pollutant
is heavily dependent on the physical, chemical and biological properties of that pollutant. For example, some substances will be more readily biodegraded than others, and some substances will be more susceptible to sorption or volatilisation. The potential for the greywater treatment and reuse system to act as a pollutant emission barrier is thus highly substance dependent. In general, a system such as that installed at Nordhavnsga˚rden will only act as a significant pollutant barrier for substances which are readily biodegradable (but this is not the case for most PS/PHS and certainly not for metals). Thus, on the basis of current designs, which typically do not facilitate separate treatment and disposal of greywater treatment sludge, the results indicate that the potential for extra benefits associated with the emission control of xenobiotics are likely to be quite limited. On the other hand, if greywater treatment sludge were to be removed from the wider municipal WWTP load this could potentially improve the sludge quality and hence help meet the requirements of the various national and European sludge regulations.
Acknowledgements The presented results have been obtained within the framework of the ScorePP project - “Source Control Options for Reducing Emissions of Priority Pollutants”, contract no. 037036, a project coordinated by the Department of Environmental Engineering, Technical University of Denmark, within the Energy, Environment and Sustainable Development section of the European Community’s Sixth Framework Programme for Research, Technological Development and Demonstration. COST Action 636 ‘Xenobiotics in the Urban Water Cycle’ is also acknowledged.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 4 9 e1 5 6 0
references
Abad, E., Martinez, K., Planas, C., Palacios, O., Caixach, J., Rivera, J., 2005. Priority organic pollutant assessment of sludges for agricultural purposes. Chemosphere 61 (9), 1358e1369. Andersson, M., Dalsgaard, A., 2004. Demonstrationsprojekt med genanvendelse af gra˚t spildevand fra en større gra˚vandsproducent Økologisk byfornyelse og spildevandrensning Nr. 46 (accessed 02 09 10) at: http://www2.mst.dk/Udgiv/publikationer/2004/87-7614149-7/html/default.htm (In Danish). APHA, AWWA, WEF, 2005. Standard methods for the examination of water and wastewater. American Water Works Association, and Water Environment Federation, first ed. American Public Health Association, Washington D.C. Bergstrom, A., 2007. Cadmium sources in the area connected to a sewage treatment plant in Stockholm. Poster presentation at 5th IWA Specialised Conference on Assessment and Control of Micropollutants/Hazardous Substances in Water. Frankfurt am Main/Germany 17e20 June 2007. Bixio, D., Thoeye, C., De Koning, J., Joksimovic, D., Savic, D., Wintgens, T., Melin, T., 2006. Wastewater reuse in Europe. Desalination 187 (1e3), 89e101. Chu, J., Chen, J., Wang, C., Fu, P., 2004. Wastewater reuse potential analysis: implications for China’s water resources management. Water Research 38 (11), 2746e2756. Copenhagen Energy, 2008a. The content of your drinking water: chemical investigation of treated water (accessed 05 06 09) at: http://www.ke.dk/portal/pls/portal/docs/436003.PDF. Copenhagen Energy, 2008b. The content of your drinking water: content of xenobiotic substances in drinking water (accessed 05 06 09) at: http://www.ke.dk/portal/pls/portal/docs/436005. PDF. Danish Standard, 2003. Determination of Metals in Water, Sludge and Sediments - General Guidelines for Determination by Atomic Absorption Spectrophotometry in Flame, vol. 259. DS, pp. 2003. Danish Standard, 2004. Water quality - Determination of trace elements using atomic absorption spectrometry with graphite furnace, no. DS/EN ISO 15586:2004 DANVA(Danish Water and Wastewater Association), 2007. VAND I TAL DANVA’s Benchmarking og Vandstatistik (accessed 02 09 10) at: http://www.bessy.dk (In Danish). Donner, E., Eriksson, E., Revitt, D.M., Scholes, L., HoltenLutzhof, H.-C., Ledin, A., 2010. Presence and fate of priority substances in domestic greywater treatment and reuse systems. Science of the Total Environment 408 (12), 2444e2451. Eriksson, E., Auffarth, K.P.S., Henze, M., Ledin, A., 2002. Characteristics of grey wastewater. Urban Water 4 (1), 85e104. Eriksson, E., Auffarth, K., Eilersen, A.-M., Henze, M., Ledin, A., 2003. Household chemicals and personal care products as sources for xenobiotic organic compounds in grey wastewater. Water SA 29 (2), 135e146. European Commission, 1986. Directive 86/278/EEC of 12th June 1986 on the protection of the environment and in particular of the soil, when sewage sludge is used in agriculture (accessed 02 09 10) at: http://ec.europa.eu. European Commission, 2000a. Directive 2000/60/EC of the European Parliament and of the Council establishing a framework for the Community action in the field of water policy (accessed 02 09 10) at: http://ec.europa.eu/emvironment/ water/water-framework/index_en.html 23rd October, 2000. European Commission, 2000b. Working Document on Sludge, 3rd draft, ENV.E.3/LM (accessed 02 09 10) at:. European Commission, Brussels http://europa.eu.int/comm/ environment/sludge/report10.htm 27th April, 2000.
1559
European Commission, 2008. Directive 2008/105/EC of the European Parliament and of the Council of 16 December 2008 on environmental quality standards in the field of water policy, amending and subsequently repealing Council Directives 82/176/EEC, 83/513/EEC, 84/156/EEC, 84/491/EEC, 86/ 280/EEC and amending Directive 2000/60/EC of the European Parliament and of the Council. Fatta-Kassinos, D., Kalavrouziotis, I.K., Koukoulakis, P.H., Vasquez, M.I., 2010. The risks associated with wastewater reuse and xenobiotics in the agroecological environment. Science of the Total Environment. doi:10.1016/j.scitotenv.2010. 03.036. Friedler, E., Gilboa, Y., 2010. Performance of UV disinfection and the microbial quality of greywater effluent along a reuse system for toilet flushing. Science of the Total Environment 408 (9), 2109e2117. Giri, R.R., Takeuchi, J., Ozaki, H., 2006. Biodegradation of domestic wastewater under the simulated conditions of Thailand. Water and Environment Journal 20 (3), 169e176. Jensen, J., Jepsen, S.-E., 2005. The production, use and quality of sewage sludge in Denmark. Waste Management 25, 239e247. Kjellerup, M., Hansen, A.M., 1994. Vandbesparende foranstaltninger. ISBN 87-571-1435-9. Teknisk Forlag. Koh, Y.K.K., Lester, J.N., Scrimshaw, M.D., 2005. Fate and behaviour of alkylphenols and their polyethoxylates in an activated sludge plant. Bulletin of Environmental Contamination and Toxicology 75 (6), 1098e1106. Ledin, A., Auffarth, K., Eriksson, E., Smith, M., Eilersen, A.-M., Mikkelsen, P.S., Dalsgaard, A., Henze, M., 2006. Udvikling af metode til karakterisering af gra˚t spildevand. Økologisk byfornyelse og spildevandsrensning 58 (accessed 05 06 10) at: http://www.mst.dk/Udgivelser/Publikationer/2006/07/877052-116-6.htm. Li, F.Y., Wichmann, K., Otterpohl, R., 2009. Review of the technological approaches for grey water treatment and reuses. Science of the Total Environment 407 (11), 3439e3449. Liu, S., Butler, D., Memon, F.A., Makropoulos, C., Avery, L., Jefferson, B., 2010. Impacts of residence time during storage on potential of water saving for grey water recycling system. Water Research 44 (1), 267e277. Lynettefællesskabet I/S. 2008. (accessed 12 10 10) at: http://www. lyn-is.dk/Lynettef%C3%A6llesskabet/Gr%C3%B8nt_regnskab,_ Milj%C3%B8data_og_%C3%85rsberetning.aspx. Maimon, A., Tal, A., Friedler, E., Gross, A., 2010. Safe on-site reuse of greywater for irrigationeA critical review of current guidelines. Environmental Science and Technology 44 (9), 3213e3220. Masi, F., El Hamouri, B., Shafi, H.A., Baban, A., Ghrabi, A., Regelsberger, M., 2010. Treatment of segregated black/grey domestic wastewater using constructed wetlands in the Mediterranean basin: the zer0-m experience. Water Science and Technology 61 (1), 97e105. McLaughlin, M.J., Hamon, R.E., McLaren, R.G., Speir, T.W., Rogers, S.L., 2000. A bioavailability-based rationale for controlling metal and metalloid contamination of agricultural land in Australia and New Zealand. Australian Journal of Soil Research 38 (6), 1037e1086. Review. Memon, F.A., Butler, D., 2006. Water consumption trends and demand forecasting techniques. In: Butler, D., Memon, F.A. (Eds.), Water Demand Management, vol. 2006. IWA Publishing, pp. 1e26. Misra, R.K., Patel, J.H., Baxi, V.R., 2010. Reuse potential of laundry greywater for irrigation based growth, water and nutrient use of tomato. Journal of Hydrology 386 (1e4), 95e102. Nielsen, M., Pettersen, T., 2005. Genanvendelse af gra˚t spildevand pa˚ campingpladser - Fase 2 og 3 Økologisk byfornyelse og spildevandsrensning no. 57. Report to the Danish EPA
1560
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 4 9 e1 5 6 0
(accessed 02 09 10) at: http://www2.mst.dk/Udgiv/ publikationer/2005/87-7614-899-8/html/default.htm. Palmquist, H, 2004. Hazardous substances in wastewater management, PhD thesis. Luleaˆ University of Technology, Sweden. 2004:47, ISSN: 1402e1544. http://epubl.luth.se/14021544/2004/47/LTU-DT-0447-SE.pdf, (accessed 02 09 10) Palmquist, H., Hanaeus, J., 2005. Hazardous substances in separately collected grey- and blackwater from ordinary Swedish households. Science of the Total Environmen 348 (1e3), 151e163. Paulo, P.L., Begosso, L., Pansonato, N., Shrestha, R.R., Boncz, M.A., 2009. Design and configuration criteria for wetland systems treating greywater. Water Science and Technology 60 (8), 2001e2007. Pidou, M., Avery, L., Stephenson, T., Jeffrey, P., Parsons, S.A., Liu, S., Memon, F.A., Jefferson, B., 2008. Chemical solutions for greywater recycling. Chemosphere 71 (1), 147e155. Pinto, U., Maheshwari, B.L., 2010. Reuse of greywater for irrigation around homes in Australia: understanding community views, issues and practices. Urban Water Journal 7 (2), 141e153.
Revitt, D.M., Shutes, R.B.E., Jones, R.H., Forshaw, M., Winter, B., 2004. The performances of vegetative treatment systems for highway runoff during dry and wet conditions. Science of the Total Environment 334, 261e270. Revitt, D.M., Scholes, L., Ellis, J.B., 2008. A pollutant removal prediction tool for stormwater derived diffuse pollution. Water Science and Technology 57 (8), 1257e1264. Scholes, L., Revitt, M., Gasperi, J., Donner, E., 2007. Priority pollutant behaviour in stormwater Best management practices (BMPs). Deliverable 5.1, ScorePP project (accessed 22 10 10) at: http://www.scorepp.eu/index.php?option¼com_ content&task¼view&id¼30&Itemid¼56. US EPA(United States of America Environmental Protection Agency), 2007. EPI Suite v3.20 (February 2007) (accessed 02 09 10) at: http://www.epa.gov/opptintr/exposure/pubs/episuite.htm. Wall, E., 2002. Kadmium i husha˚llsspillvatten. Stockholm Vatten (Stockholm Water) Report No. 9, April 2002. (accessed on 12 10 10) at: http://www.stockholmvatten.se/commondata/ rapporter/avlopp/Processer/Kadmium_spillvatten.pdf (In Swedish).
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 6 1 e1 5 7 0
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Investigating the decay rates of Escherichia coli relative to Vibrio parahemolyticus and Salmonella Typhi in tropical coastal waters Choon Weng Lee a,*, Angie Yee Fang Ng a, Chui Wei Bong a, Kumaran Narayanan b, Edmund Ui Hang Sim c, Ching Ching Ng a a
Laboratory of Microbial Ecology, Institute of Biological Sciences, University of Malaya, 50603 Kuala Lumpur, Malaysia School of Science, Monash University, Sunway Campus, Selangor, Malaysia c Department of Molecular Biology, Faculty of Resource Science and Technology, University Malaysia Sarawak, Malaysia b
article info
abstract
Article history:
Using the size fractionation method, we measured the decay rates of Escherichia coli, Salmo-
Received 24 June 2010
nella Typhi and Vibrio parahaemolyticus in the coastal waters of Peninsular Malaysia. The size
Received in revised form
fractions were total or unfiltered, <250 mm, <20 mm, <2 mm, <0.7 mm, <0.2 mm and <0.02 mm.
18 November 2010
We also carried out abiotic (inorganic nutrients) and biotic (bacterial abundance, production
Accepted 19 November 2010
and protistan bacterivory) measurements at Port Dickson, Klang and Kuantan. Klang had
Available online 27 November 2010
highest nutrient concentrations whereas both bacterial production and protistan bacterivory rates were highest at Kuantan. We observed signs of protistebacteria coupling via the
Keywords:
following correlations: Protistan bacterivoryBacterial Production: r ¼ 0.773, df ¼ 11, p < 0.01;
Bacterial decay rate
ProtistBacteria: r ¼ 0.586, df ¼ 12, p < 0.05. However none of the bacterial decay rates were
Size fractionation
correlated with the biotic variables measured. E. coli and Salmonella decay rates were
Top-down control
generally higher in the larger fraction (>0.7 mm) than in the smaller fraction (<0.7 mm) sug-
Straits of Malacca
gesting the more important role played by protists. E. coli and Salmonella also decreased in the
South China sea
<0.02 mm fraction and suggested that these non-halophilic bacteria did not survive well in seawater. In contrast, Vibrio grew well in seawater. There was usually an increase in Vibrio after one day incubation. Our results confirmed that decay or loss rates of E. coli did not match that of Vibrio, and also did not correlate with Salmonella decay rates. However E. coli showed persistence where its decay rates were generally lower than Salmonella. ª 2010 Elsevier Ltd. All rights reserved.
1.
Introduction
Coastal waters account for less than 10% of the ocean area. However they are highly productive and account for 25% of primary production in the ocean (Berger et al., 1989). Coastal waters are increasingly exploited by humans for food, recreation, transport and other needs, and at present most are in various stages of degradation (Alongi, 1998). There is also an increasing public health threat from pathogens (Hazen and
Toranzos, 1990; Moe, 1997). Disposal of inadequately treated waste is considered faecal pollution and a main source of bacterial pathogens in the sea (Solo-Gabriele et al., 2000). For faecal pollution studies, the concept of bacterial indicator is standard (Wolf, 1972). A fundamental assumption to this concept is the parity in the survival of indicator and enteric pathogens over a wide range of aquatic environments (Bonde, 1977). It is however acknowledged that these indicators are inadequate to predict the presence of pathogenic
* Corresponding author. E-mail address:
[email protected] (C.W. Lee). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.11.025
1562
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 6 1 e1 5 7 0
microorganisms (Rhodes and Kator, 1988; Kay et al., 1994; Borrego and Figueras, 1997). At present, the coliform group of microorganisms or microorganisms found in the intestines of all warm blooded animals especially Escherichia coli is used as a standard indicator of faecal contamination in many countries (Hazen and Toranzos, 1990) including Malaysia (Department of Environment, 2008). The value of E. coli as an indicator microorganism is significantly enhanced by the ease with which it can be detected and cultured (Wolf, 1972) when compared with other bacterial pathogens. In order to use E. coli as an indicator bacterium in seawater, knowledge of its survival must be acquired as it is not a halophilic microorganism. Loss rates of indicator microorganisms pose an interesting problem especially in coastal waters as many factors affect the survival times of E. coli for example temperature, light, salinity, predation, nutrients and pollutants (Fujioka et al., 1981; Munro et al., 1989; Presser et al., 1998; Solo-Gabriele et al., 2000; Rozen and Belkin, 2001; Sinton et al., 2002). Since the 1950s, Carlucci and Pramer (1959; 1960a; 1960b) have studied the survival of bacteria including E. coli in seawater. Numerous studies have shown that decay rates of E. coli do not reflect Vibrio cholerae in estuarine waters and also Salmonella spp. (Colwell et al., 1981; Rhodes and Kator, 1988). However to the best of our knowledge, there is no study that compares the relative loss rates of E. coli with both halophilic and non-halophilic bacterial pathogens. In this study, we investigated whether the decay rates for E. coli were similar to a non-halophilic bacterial pathogen (i.e. Salmonella Typhi) and a halophilic bacterial pathogen (i.e. Vibrio parahaemolyticus)? Our results could help properly evaluate the risk posed by such bacteria either to the health of bathers in recreational waters or to the safety of fisheries or marine aquaculture.
2.
Materials and methods
2.1.
Sampling
We sampled at three stations located on the east (Kuantan: 03 48.40 N 103 20.60 E) and west (Klang: 03 00.10 N 101 23.40 E and Port Dickson: 02 29.50 N 101 50.30 E) of Peninsular Malaysia from April until October 2006 (Fig. 1). The stations at Kuantan and Klang were located in estuaries whereas the station at Port Dickson was sandy coast. Surface seawater samples were collected during high tide, using an acid-cleaned bucket, and in-situ measurements of temperature (0.1 C) and salinity (0.1 ppt) were carried out using a salinometer (YSI-30, US). Seawater samples were then kept in a cooler box for no more than 4 h until processing in the laboratory. In the laboratory, seawater samples for dissolved nutrient analyses were filtered through pre-combusted (450 C for 5 h) Whatman GF/F filters, and stored at 20 C until analysis.
2.2.
Environmental conditions
Dissolved inorganic nitrogen (nitrate (NO3), nitrite (NO2), ammonium (NH4)), and phosphate (PO4) concentrations were measured using a spectrophotometer (Parsons et al., 1984). All nutrient measurements above were carried out in triplicates.
Fig. 1 e Map showing the location of the sampling sites, east (Kuantan) and west (Klang and Port Dickson) of Peninsular Malaysia.
Coefficient of variation (CV) for NH4, NO2 and PO4 analyses were <5%, and <10% for NO3 analysis. Bacteria and protist were determined using the direct count method by an epifluorescence microscope (Olympus BX60, Japan) with a U-MWU filter cassette (excitor 330385 nm, dichroic mirror 400 nm, barrier 420 nm). For protist, 10 ml sample was filtered onto a black 0.8 mm pore size Isopore filter (Millipore, Ireland), and then stained with the fluorochrome primulin (40 mg ml1 final concentration) for 5 min (Bloem et al., 1986) whereas for bacteria, 2 ml sample was filtered onto a black 0.2 m pore size Isopore filter, and then stained with 40 6diamidino-2-phenylindole (DAPI, 1 mg ml1 final concentration) for 10 min (Kepner and Pratt, 1994). Slides were kept frozen for < 3 days before enumeration. A minimum of 10 microscope fields or 500 cells were counted for bacteria and for protist, at least 30 microscope fields were observed.
2.3.
Bacterial decay rates
For bacterial decay experiments, seawater samples were sizefractionated as total or unfiltered, <250 mm (through a 250 mm stainless steel mesh), <20 mm (20 mm pore size nylon mesh), <2 mm (2.0 mm polycarbonate membrane filter), <0.7 mm (GF/F filter), <0.2 mm (0.2 mm polycarbonate membrane filter) and <0.02 mm (0.02 mm Whatman Anodisc). The size fractions were inoculated separately with 1% (v/v) fresh cultures of E. coli, Salmonella Typhi (hereafter referred to as Salmonella) and V. parahaemolyticus (hereafter referred to as Vibrio), and then incubated at 30 C for about three days. Both E. coli and Salmonella inocula were prepared on nutrient broth whereas
1563
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 6 1 e1 5 7 0
for Vibrio, 3% NaCl (final concentration) was added to the nutrient broth. From the inoculated size fractions, the initial E. coli, Salmonella and Vibrio concentration was determined as colony forming unit (cfu) ml1 via spread plating. The number of cfus was observed one day after spread plating on MacConkey agar (for E. coli and Salmonella) and Thiosulfate Citrate Bile Salts Sucrose (TCBS) agar with 3% NaCl (for Vibrio). A portion of the sample was also removed every day to determine the change in E. coli, Salmonella and Vibrio concentration. Bacterial decay rate was modeled by the standard differential equation: dN/dt ¼ kN (Chick, 1908) where k ¼ decay constant, N is the cfu ml1, and t ¼ time of incubation. Integrating the differential equation gave the following ln N ¼ kt þ C. Hence for decay rate assessment, we transformed the cfu data by natural logarithm and plotted the ln cfu ml1 against incubation time. Linear regression analysis was then used to find the best-fit slope i.e. decay rate. We also measured separately, the concurrent bacterial production and protistan bacterivory rates by measuring bacterial growth rate in both <0.7 mm and <20 mm fractions after 12 h incubation. Bacterial growth rate in each fraction was calculated as the increase in natural logarithmic bacterial abundance over time. The bacterial growth rate in the <0.7 mm fraction was assumed without grazing (m0.7) whereas growth rate in the <20 mm fraction (m20) was the product of both growth and grazing. Bacterial production (BP) was then estimated by the following equation: BP ¼ Bacterial abundance m0.7 (Lee et al., 2009a) whereas bacterivory rate was estimated by Bacterial abundance (m0.7 m20) (McManus, 1993).
2.4.
Statistical analyses
Statistical tests such as coefficient of variation (CV), analysis of variance, Student’s t-test, Tukey’s test, linear regression and correlation analyses were carried out according to Zar (1999). Count data were log-transformed to meet parametric assumptions of equality of variances and normal distribution before correlation and linear regression analyses. All data, unless mentioned otherwise, were reported as mean S.D.
3.
Results
3.1.
Abiotic measurements
Table 1 shows the physico-chemical parameters measured at the three stations. Average surface seawater temperature
observed ranged from 29 to 30 C whereas average salinity measured at the estuaries (Klang and Kuantan) was lower than at Port Dickson. Salinity fluctuated over a wider range at the estuaries (CV ¼ 7% and 46%) than Port Dickson (CV ¼ 1%). Dissolved inorganic nitrogen (DIN) measured showed that concentrations at Klang were more than two-fold higher than both Kuantan and Port Dickson. Among the nitrogen species, NH4 was the dominant species at both Klang and Kuantan, accounting for > 70% of DIN. At Port Dickson, NO3 was the dominant species (about 45% of DIN). Average PO4 concentration at Klang was also higher than at both Port Dickson and Kuantan.
3.2.
Biotic measurements
Bacterial abundance ranged from 0.85 to 3.49 106 cell ml1, and was significantly higher at Kuantan than both Klang (q ¼ 4.77, df ¼ 12, p < 0.05) and Port Dickson (q ¼ 5.77, df ¼ 12, p < 0.01) (Table 2). For protist, abundance ranged from 0.58 to 6.64 103 cell ml1, and was about three orders lower than bacteria. Protist abundance at Kuantan was significantly higher than Port Dickson (q ¼ 4.34, df ¼ 12, p < 0.05) but not Klang. Table 2 also shows the bacterial production measured in this study. Bacterial production at both Klang and Port Dickson ranged from 0.72 to 1.69 105 cell ml1 h1 whereas bacterial production at Kuantan ranged from 1.99 to 4.78 105 cell ml1 h1. Similar to patterns exhibited by microbial abundance data, bacterial production at Kuantan was significantly higher than both Klang (q ¼ 4.69, df ¼ 12, p < 0.05) and Port Dickson (q ¼ 5.79, df ¼ 12, p < 0.01). Protistan bacterivory at Kuantan was also higher than both Klang (q ¼ 14.48, df ¼ 9, p < 0.001) and Port Dickson (q ¼ 13.64, df ¼ 9, p < 0.001). Bacterivory rates at Kuantan ranged from 1.60 to 2.15 105 cell ml1 h1 whereas bacterivory rates at both Klang and Port Dickson were about one order lower, and ranged from 1.11 to 4.19 104 cell ml1 h1. E. coli concentration from 1990 until 2005 were also obtained from the Department of Environment Malaysia monitoring stations located near our study area. Although E. coli concentration varied over four-order at all three locations (Fig. 2), analysis of variance showed significant differences among the different locations (F ¼ 19.2, df ¼ 1095, p < 0.001). At Port Dickson and Klang, average E. coli was 1300 3600 MPN (Most Probable Number) per 100 ml, and was significantly higher than Kuantan (640 1900 MPN per 100 ml) (KlangeKuantan: q ¼ 8.52, df ¼ 1095, p < 0.001; Port DicksoneKuantan: q ¼ 6.38, df ¼ 1095, p < 0.001).
Table 1 e Physico-chemical characteristics of the sampling stations in this study. Mean (±S.D.) of surface water temperature, salinity, ammonium (NH4), nitrite (NO2), nitrate (NO3), and phosphorus (PO4). Station Klang (n ¼ 4) Port Dickson (n ¼ 5) Kuantan (n ¼ 5)
Temperature C
Salinity ppt
NH4 mM
NO2 mM
NO3 mM
PO4 mM
30.3 0.5 30.0 1.0
27.7 1.9 28.4 0.3
17.84 24.75 0.42 0.51
2.97 0.64 0.18 0.27
1.24 0.61 0.50 0.28
1.66 1.54 0.12 0.05
29.1 0.8
22.4 10.2
2.89 2.64
0.51 0.29
0.61 0.33
0.60 0.31
1564
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 6 1 e1 5 7 0
Table 2 e Bacterial abundance, protist abundance, bacterial production and protistan bacterivory rates measured in this study. Date
Bacteria (106 cell ml1)
Protist (103 cell ml1)
Bacterial production (105 cell ml1 h1)
Protistan bacterivory (104 cell ml1 h1)
Klang
24-Apr-06 12-Jun-06 26-Jun-06 31-Jul-06 Average (S.D.)
1.13 0.93 1.29 1.53 1.22 0.26
2.89 4.08 2.57 1.08 2.65 1.23
1.69 0.97 1.66 1.64 1.49 0.35
1.55 1.11 3.70 2.77 2.28 1.18
Port Dickson
04-Jul-06 03-Oct-06 10-Oct-06 17-Oct-06 30-Oct-06 Average (S.D.)
1.10 0.85 0.91 1.11 0.98 0.99 0.12
2.43 1.16 2.01 0.58 1.73 1.58 0.73
e 0.72 1.25 1.36 1.08 1.10 0.28
e e 3.41 4.19 1.95 2.26 2.07
Kuantan
18-Apr-06 17-May-06 19-Jun-06 25-Jul-06 15-Aug-06 Average (S.D.)
2.91 1.31 2.19 2.40 3.49 2.46 0.82
3.02 1.76 5.42 6.64 5.51 4.47 2.01
4.78 1.99 3.60 2.50 2.84 3.14 0.11
21.45 e 16.05 e 16.00 17.83 3.13
Station
3.3.
Bacterial decay rates at different size fractions
Fig. 3 shows the change in E. coli concentration after inoculation in the different fractions of seawater collected from Port Dickson, Kuantan and Klang. There was an apparent decrease in E. coli in most of the larger fraction seawater (i.e. total, <250 mm, <20 mm, <2 mm) whereas in the smaller fraction (i.e. <0.7 mm, <0.2 mm and <0.02 mm), the concentration of E. coli often did not change significantly. Fig. 4 shows the statistically significant ( p < 0.05) decay rates measured in this study, and confirmed that E. coli decay rate was significantly higher in the larger fraction (>0.7 mm) (2.28 0.90 d1) than the smaller fraction (<0.7 mm) (0.58 0.22 d1) (Student’s t-test: t ¼ 12.35, df ¼ 64, p < 0.001). A similar trend was observed for Salmonella where its concentration generally decreased with time, especially in the larger fraction seawater (Fig. 5). Salmonella decay rates in the larger fraction (3.17 1.19 d1) were significantly higher than in the smaller fraction (1.51 0.84 d1) (Student’s t-test: t ¼ 6.65, df ¼ 59, p < 0.001) (Fig. 4). In contrast to E. coli and Salmonella, Vibrio exhibited a different trend (Fig. 6). There was often an increase in Vibrio concentration after one day incubation, and less discernible difference in Vibrio concentration between the larger and smaller fractions of seawater. Few of the experiments gave significant decay rates (Fig. 4), and available Vibrio decay rates ranged 0.90e1.78 d1 at Klang, and 0.54e4.64 d1 at Kuantan. At Port Dickson, only two significant decay rates were observed (1.36 0.21 d1). When we compared E. coli decay rates in the larger fractions among the stations, E. coli decay rate was highest at Klang (1.96e4.90 d1), followed by Kuantan (0.88e2.80 d1) and Port Dickson (0.36e2.93 d1) (KlangeKuantan: q ¼ 6.55, df ¼ 51, p < 0.001; KlangePort Dickson: q ¼ 9.99, df ¼ 51, p < 0.001; KuantanePort Dickson: q ¼ 3.44, df ¼ 51, p < 0.05). However
Fig. 2 e Long term E. coli counts (log MPN/100 ml) from selected Department of Environment monitoring stations near Port Dickson (PD) (n [ 578), Klang (n [ 196) and Kuantan (n [ 322).
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 6 1 e1 5 7 0
1565
Fig. 3 e E. coli decay (ln cfu mlL1) measured in different fractions at Port Dickson (PD) (n [ 5), Klang (n [ 3) and Kuantan (n [ 5).
there was no significant difference in Salmonella and Vibrio decay rates among the stations.
4.
Discussion
4.1.
Environmental conditions
Surface seawater temperature observed was typical of tropical waters. Klang had the highest nutrient concentrations, and reaffirmed earlier observations (Lee and Bong, 2006, 2008; Lee et al., 2009a). One possible reason is the rapid pace of
development and industrialization taking place upstream of Klang where the capital of Malaysia, Kuala Lumpur is located (Lee and Bong, 2006). In this study, the abundance of both bacteria and protist were within the range for coastal waters of Peninsular Malaysia (Lee et al., 2005; Lee and Bong, 2007, 2008). Protist abundance was about three orders of magnitude lower than bacteria, and this observation was consistent with the analysis by Sanders et al. (1992). Both bacterial production and protistan bacterivory rates measured in this study were also within the range previously published for tropical coastal waters (Lee et al., 2005, 2009a; Lee and Bong, 2006, 2007, 2008).
1566
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 6 1 e1 5 7 0
Fig. 4 e Summary of decay rates (P < 0.05) measured in this study at Klang, Port Dickson (PD) and Kuantan. Different symbols represent different sampling dates. Error bar denotes the standard error of decay rate and is shown except when smaller than symbol.
However, in contrast to the pattern exhibited by nutrient concentrations where Klang had the highest levels, all biotic variables were highest at Kuantan. It was not clear why biotic variables at Klang were not reflective of its nutrient concentrations as found in earlier studies (Lee and Bong, 2008; Lee et al., 2009a). However both Klang and Kuantan are estuaries, and can experience episodic nutrient inputs that can stimulate microbial biomass and productivity. Long term monitoring data showed the extent of faecal pollution via the indicator E. coli. Although coastal water quality in Malaysia has deteriorated over time (Department of Environment, 2008), there was no apparent increase in E. coli over the 15-year period for all three locations. At present, Malaysia has an interim marine water quality standard that set the criterion for E. coli at 100 MPN per 100 ml (Department of Environment, 2008). From a 15-year period of monitoring data, >60% of the samples at Klang and Port Dickson exceeded the standard whereas at Kuantan, 35% of the samples exceeded the standard. Dow (1995) had earlier reported higher E. coli concentration for stations along the Straits of Malacca. Our study showed that faecal pollution remains a problem for coastal waters here. One reason is many coastal communities in Malaysia lack proper sewage disposal systems and often discharge sewage directly into the sea (Law, 1992). The faecal pollution is accentuated for stations along the Straits of Malacca as the population density is higher along the west coast of Peninsular Malaysia. Although sewage treatment facilities have increased over time, the problem showed no sign of alleviation at both Klang and Port Dickson.
4.2.
Bacterial decay rates at different size fractions
The decay rates obtained in this study were within the range reported by Anderson et al. (2005) but relatively higher than decay rates reported for temperate waters (Lessard and Sieburth, 1983; Rhodes and Kator, 1988). Relative to temperate waters, the higher decay rates obtained in this study could be due to the fact that microbial activity (including protistan bacterivory) in tropical waters is at its optimum (Pomeroy and Wiebe, 2001). For both E. coli and Salmonella, decay rates were generally higher in the larger fraction (>0.7 mm) than in the smaller fraction (<0.7 mm). The main bacterial predators in the larger fractions are nanoflagellates and ciliates (Sanders et al., 1992) whereas the cause of bacterial mortality in the smaller fraction is mainly by viral lysis (Fuhrman, 2000). In the smaller fraction, lytic bacteria may also play a role albeit a minor one (Enzinger and Cooper, 1976). Bacterial decay rates measured in this study suggested that protistan bacterivory was more important, similar to the conclusion by Enzinger and Cooper (1976). A reason why viral lysis played a minor role in this study was because both Salmonella and E. coli are not natural seawater organisms and are not active in the sea (Carlucci and Pramer, 1960b). Bacteriophages require the host physiological activity in order to replicate (Pretorius, 1962). The decay rates in the <0.02 mm fraction is effectively not due to viral lysis or bacterivory as both viruses and protists do not pass through this pore size. The <0.02 mm fraction is therefore suitable to observe the response of halophilic and
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 6 1 e1 5 7 0
1567
Fig. 5 e Salmonella decay (ln cfu mlL1) measured in different fractions at Port Dickson (PD) (n [ 4), Klang (n [ 3) and Kuantan (n [ 4).
non-halophilic pathogens in seawater. We observed that Vibrio counts in the <0.02 mm fraction increased after day one in all the experiments, and no significant decay of Vibrio could be observed throughout this study. In contrast, the decay rates for E. coli and Salmonella in the <0.02 mm fraction were still statistically significant ( p < 0.05). Our results concurred with earlier reports that non-halophiles are usually not able to grow well in seawater (Gerba and McLeod, 1976). However relative to protistan bacterivory, the decay rates for E. coli and Salmonella in the <0.02 mm fraction were low, and accounted for <20% of total decay rates. These decay rates also did not correlate ( p > 0.15) with salinity, suggesting that bacterivory was more important than salinity for non-halophilic bacterial decay. There were obvious differences in the response of Vibrio in the different seawater fractions relative to both E. coli and Salmonella. In this study, significant Vibrio decay was seldom observed, and rates observed were usually lower than E. coli and Salmonella. One reason why Vibrio decay rates were lower was probably due to the ability of V. parahaemolyticus to grow in a salty environment (Holt et al., 1994). Increase in Vibrio was still apparent after one day incubation before Vibrio counts started decreasing. The latter reduction could be due to other stresses for example limited food availability that resulted in Vibrio growth rates falling below loss rates.
E. coli is widely used as an indicator for faecal pollution and for pathogenic microorganisms (Bonde, 1977). However the validity of E. coli as an indicator is questionable especially in coastal waters (Solo-Gabriele et al., 2000) where presence of E. coli is more likely the balance between supply and loss. This is because E. coli is non-halophilic and is not known to grow well in coastal waters (Gerba and McLeod, 1976). Our results confirmed that decay or loss rates of E. coli did not match that of the halophilic Vibrio and also did not correlate significantly with Salmonella decay rates ( p > 0.50) (Table 3) even though the survival characteristic of E. coli is presumed similar to Salmonella (Bonde, 1977). When we compared the different bacterial responses in the larger fractions, we found that at Klang, Salmonella and E. coli decay rates were significantly higher than Vibrio (SalmonellaeVibrio: q ¼ 6.58, df ¼ 33, p < 0.001; E. colieVibrio: q ¼ 6.16, df ¼ 33, p < 0.001) whereas at Kuantan, Salmonella decay rate was significantly higher than both Vibrio (q ¼ 7.70, df ¼ 46, p < 0.001) and E. coli (q ¼ 4.92, df ¼ 46, p < 0.01). Similarly at Port Dickson, Salmonella decay rate was higher than E. coli (Student’s t-test: t ¼ 4.10, df ¼ 29, p < 0.001). For Port Dickson, only decay rates from Salmonella and E. coli were compared as there were too few decay rates from Vibrio experiments. Generally, we found that Salmonella decay rates were higher
1568
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 6 1 e1 5 7 0
Fig. 6 e Vibrio decay (ln cfu mlL1) measured in different fractions at Port Dickson (PD) (n [ 3), Klang (n [ 3) and Kuantan (n [ 3).
than E. coli. The persistence of E. coli relative to Salmonella fulfilled one of the criteria for an indicator organism i.e. the indicator organism should survive longer than the pathogen itself (Cabelli, 1978; Allwood et al., 2003). Although our study provided snapshots of the environment, we showed consistently the importance of protistan bacterivory in bacterial decay. E. coli was a poor indicator for halophilic pathogens, and although E. coli decay rates did not correlate with Salmonella, E. coli persisted longer than Salmonella in coastal waters. Our results provided support to the continuous use of E. coli as an indicator organism for nonhalophilic pathogens especially Salmonella.
4.3.
Protistebacteria coupling
Multiple correlation analysis showed that E. coli, Salmonella and Vibrio decay rates did not correlate with each other or other biotic variables (Table 3). Protistan bacterivory also did not correlate with E. coli, Salmonella and Vibrio decay rates. As these pathogens form only a minor fraction of the total bacterial population in the sea (Lee et al., 2009b), their decay rates played only a minor role towards total bacterivory rates. However, we did observe evidences of significant coupling between protist
and bacteria similar to Sanders et al. (1992) and Lee et al. (2005) (Protistan bacterivoryeBacterial Production: r ¼ 0.773, df ¼ 11, p < 0.01; ProtisteBacteria: r ¼ 0.586, df ¼ 12, p < 0.05). Our study showed that protistan bacterivory accounted for 9e56% of bacterial production (Table 2), and was significantly higher at Kuantan (>44% bacterial production) than both Klang (q ¼ 9.68, df ¼ 9, p < 0.001) and Port Dickson (q ¼ 6.68, df ¼ 9, p < 0.01).
Table 3 e Pearson productemoment correlation coefficient (r) between variables measured i.e. bacterial production (BP, cell mlL1 hL1), log bacterial abundance (BA, cell mlL1), grazing (cell mlL1 hL1), log Protist (cell mlL1), decay rates of E. coli (dL1), Salmonella (dL1), and Vibrio (dL1). * is P < 0.05, ** is P < 0.01, *** is P < 0.001. BP
BA
BA 0.896*** Grazing 0.773** 0.794*** Protist 0.546* 0.586* E. coli 0.214 0.120 Salmonella 0.174 0.082 Vibrio 0.129 0.135
Grazing Protist E. coli Salmonella
0.295 0.291 0.120 0.064
0.190 0.121 0.077 0.276 0.040
0.106
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 6 1 e1 5 7 0
Bacterivory rates at both Klang and Port Dickson accounted for only 9e31% bacterial production. When bacterivory was less than bacterial production, other factors might have played a role in the removal of bacterial production as bacterial abundance is stable and does not change significantly with time (Lee and Bong, 2008). Among the factors that are important include viral lysis (Fuhrman, 2000), removal by benthic filter feeders (Strom, 2000), and sedimentation (Pedro´s-Alio´ and Mas, 1993). Sedimentation is often overlooked as a loss factor for bacteria due to the low sinking rates of microorganisms. However sinking speed can increase for bacteria attached to particles and there might be significant losses by sedimentation (Pedro´s-Alio´ and Mas, 1993). Although these loss factors were not investigated here, studies on viral ecology in tropical waters have suggested that viral lysis may not be an important loss factor (Bettarel et al., 2006; Cissoko et al., 2008). Moreover the higher suspended solids in the coastal waters of Peninsular Malaysia (Bong and Lee, 2008; Lee et al., 2009b) implied that sedimentation is probably more important. However further investigations are needed.
5.
Conclusion
1. Bacterial decay in tropical coastal waters is mainly due to protistan bacterivory. 2. Via the <0.02 mm fraction, our results showed that E. coli and Salmonella do not survive well in seawater. In contrast, Vibrio grows well in seawater. Therefore E. coli is a poor indicator for halophilic pathogens. 3. E. coli decay rates do not correlate with both Salmonella and Vibrio decay rates. However E. coli persists longer than Salmonella in coastal waters. Our results provided support to the continuous use of E. coli as an indicator organism for non-halophilic pathogens especially Salmonella. 4. There is protist‒bacteria coupling where protist counts correlated with bacterial abundance, and protistan bacterivory correlated with bacterial production.
Acknowledgements We are grateful to the Department of Environment, Malaysia for providing the monitoring data of E. coli concentration. Funding for this research was provided by University of Malaya (RG064-09SUS) and Ministry of Science, Technology & Innovation (06-01-03-SF0457). The researchers would also like to thank anonymous reviewers who helped improve the manuscript and University of Malaya for providing the research facilities.
references
Allwood, P.B., Malik, Y.S., Hedberg, C.W., Goyal, S.M., 2003. Survival of F-specific RNA coliphage, feline calicivirus and Escherichia coli in water: a comparative study. Applied and Environmental Microbiology 69 (9), 5707e5710.
1569
Alongi, D.M., 1998. Coastal Ecosystem Processes. CRC Press, Boca Raton, FL. Anderson, K.L., Whitlock, J.E., Harwood, V.J., 2005. Persistence and differential survival of fecal indicator bacteria in subtropical waters and sediments. Applied and Environmental Microbiology 71 (6), 3041e3048. Berger, W.H., Smetacek, V.S., Wefer, G., 1989. Ocean productivity and paleoproductivityean overview. In: Berger, W.H., Smetacek, V.S., Wefer, G. (Eds.), Dahlem Workshop on Productivity of the Ocean: Present and Past. John Wiley & Sons, Chichester, pp. 1e34. Bettarel, Y., Bouvy, M., Dumont, C., Sime-Ngando, T., 2006. Virusbacterium interactions in water and sediment of West African inland aquatic systems. Applied and Environmental Microbiology 72 (8), 5274e5282. Bloem, J., Ba¨r-Gilissen, M.J.B., Cappenberg, T.E., 1986. Fixation, counting, and manipulation of heterotrophic nanoflagellates. Applied and Environmental Microbiology 52, 1266e1272. Bonde, G.J., 1977. Bacterial indication of water pollution. Advances in Aquatic Microbiology 1, 273e364. Bong, C.W., Lee, C.W., 2008. Nearshore and offshore comparison of marine water quality variables measured during SESMA 1. Malaysian Journal of Science 27 (3), 25e31. Borrego, J.J., Figueras, M.J., 1997. Microbiological quality of natural waters. Microbiology SEM 1, 169e180. Cabelli, V., 1978. New standards for enteric bacteria. In: Mitchell, R. (Ed.), Water Pollution Microbiology. John Wiley & Sons, New York, pp. 233e271. Carlucci, A.F., Pramer, D., 1959. Factors affecting the survival of bacteria in sea water. Applied and Environmental Microbiology 7, 388e392. Carlucci, A.F., Pramer, D., 1960a. An evaluation of factors effecting the survival of Escherichia coli in sea water. II. Salinity, pH and nutrients. Applied and Environmental Microbiology 8, 247e250. Carlucci, A.F., Pramer, D., 1960b. An evaluation of factors effecting the survival of Escherichia coli in sea water. IV. Bacteriophages. Applied and Environmental Microbiology 8, 254e256. Chick, H., 1908. An investigation of the laws of disinfection. Journal of Hygiene 8, 92e158. Cissoko, M., Desnues, A., Bouvy, M., Sime-Ngando, T., Verling, E., Bettarel, Y., 2008. Effects of freshwater and seawater mixing on virio- and bacterioplankton in a tropical estuary. Freshwater Biology 53, 1154e1162. Colwell, R.R., Seidler, R.J., Kaper, J., Joseph, S.W., Garges, S., Lockman, H., Maneval, D., Bradford, H., Roberts, N., Remmers, E., Huq, I., Huq, A., 1981. Occurrence of Vibrio cholerae serotype O1 in Maryland and Louisiana estuaries. Applied and Environmental Microbiology 41, 555e558. Department of Environment, 2008. Malaysia Environmental Quality Report 2007. Department of Environment, Ministry of Natural Resources and Environment, Malaysia. Dow, K.M., 1995. An overview of pollution issues in the Strait of Malacca. MIMA Issue Paper 5 (95), 1e34. Enzinger, R.M., Cooper, R.C., 1976. Role of bacteria and protozoa in the removal of Escherichia coli from estuarine waters. Applied and Environmental Microbiology 31, 758e763. Fuhrman, J.A., 2000. Impact of viruses on bacterial processes. In: Kirchman, D.L. (Ed.), Microbial Ecology of the Oceans. WileyLiss, New York, pp. 327e350. Fujioka, R.S., Hashimoto, H.H., Siwak, E.B., Young, R.H.F., 1981. Effect of sunlight on survival of indicator bacteria in sea water. Applied and Environmental Microbiology 41, 690e696. Gerba, C.P., McLeod, J.S., 1976. Effects of sediments on the survival of Escherichia coli in marine waters. Applied and Environmental Microbiology 32, 114e120. Hazen, T.C., Toranzos, G.A., 1990. Tropical source water. In: McFeters, G.A. (Ed.), Drinking Water Microbiology: Progress
1570
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 6 1 e1 5 7 0
and Recent Developments. Springer-Verlag, New York, USA, pp. 32e53. Holt, J.G., Krieg, N.R., Sneath, P.H.A., Staley, J.T., Williams, S.T., 1994. Bergey’s Manual of Determinative Bacteriology, ninth ed. Williams & Wilkins, Baltimore. Kay, D., Fleischer, J.M., Salomon, R.L., Jones, F., Wyer, M.D., Goodfree, A.F., Zelenauch-Jacquotte, Z., Shore, R., 1994. Predicting likelihood of gastroenteritis from sea bathing results from randomized exposure. Lancet 344, 905e909. Kepner Jr., R.L., Pratt, J.R., 1994. Use of fluorochromes for direct enumeration of total bacteria in environmental samples: past and present. Microbiology Reviews 58, 603e615. Law, H.D., 1992. The implementation constraints in waste management in Malaysia. In: Chua, T.E., Garces, L.R. (Eds.), Waste Management in the Coastal Area of the ASEAN Region: Roles of Government, Banking Institutions, Donor Agencies, Private Sector and Communities. ICLARM Conference Proceedings 33. ICLARM, Philippines, pp. 169e172. Lee, C.W., Bong, C.W., 2006. Carbon flux through bacteria in a eutrophic tropical environment: Port Klang waters. In: Wolanski, E. (Ed.), The Environment in Asia Pacific Harbours. Springer, The Netherlands, pp. 329e345. Lee, C.W., Bong, C.W., 2007. Bacterial respiration, growth efficiency and protist grazing rates in mangrove waters in Cape Rachado, Malaysia. Asian Journal of Water Environment and Pollution 4, 11e16. Lee, C.W., Bong, C.W., 2008. Bacterial abundance and production, and their relation to primary production in tropical coastal waters of Peninsular Malaysia. Marine and Freshwater Research 59, 10e21. Lee, C.W., Bong, C.W., Hii, Y.S., 2009a. Temporal variation of bacterial respiration and growth efficiency in tropical coastal waters. Applied and Environmental Microbiology 75, 7594e7601. Lee, C.W., Bong, C.W., Mohamed Yusoff, M.A., Alias, S.A., 2005. Bacterial mediated carbon flux in mangrove waters: a Malaysian perspective. International Journal of Ecology and Environmental Sciences 31, 203e211. Lee, C.W., Ng, A.Y.F., Narayanan, K., Sim, E.U.H., Ng, C.C., 2009b. Isolation and characterization of culturable bacteria from tropical coastal waters. Ciencias Marinas 35, 153e167. Lessard, E.J., Sieburth, JMcN, 1983. Survival of natural sewage populations of enteric bacteria in diffusion and batch chambers in the marine environment. Applied and Environmental Microbiology 45 (3), 950e959. McManus, G.B., 1993. Growth rates of natural populations of heterotrophic nanoplankton. In: Kemp, P.F., Sherr, B.F., Sherr, E.B., Cole, J.J. (Eds.), Handbook of Methods in Aquatic Microbial Ecology. Lewis Publishers, Boca Raton, pp. 557e562. Moe, C.L., 1997. Waterborne transmission of infectious agents. In: Hurst, C.J., Knudsen, G.R., McInerney, M.J., Stetzenbach, L.D.,
Walter, M.V. (Eds.), Manual of Environmental Microbiology. American Society for Microbiology Press, Washington, D. C., pp. 136e152. Munro, P.M., Gauthier, M.J., Breittmayer, V.A., Bongiovanni, J., 1989. Influence of osmoregulation processes on starvation survival of Escherichia coli in seawater. Applied and Environmental Microbiology 55, 2017e2024. Parsons, T.R., Maita, Y., Lalli, C.M., 1984. A Manual of Chemical and Biological Methods for Seawater Analysis. Pergamon Press, Oxford. Pedro´s-Alio´, C., Mas, J., 1993. Bacterial sinking losses. In: Kemp, P.F., Sherr, B.F., Sherr, E.B., Cole, J.J. (Eds.), Handbook of Methods in Aquatic Microbial Ecology. Lewis Publishers, Boca Raton, pp. 677e684. Pomeroy, L.R., Wiebe, W.J., 2001. Temperature and substrates as interactive limiting factors for marine heterotrophic bacteria. Aquatic Microbial Ecology 23, 187e204. Presser, K.A., Ross, T., Ratkowsky, D.A., 1998. Modelling the growth limits (growth/no growth interface) of Escherichia coli as a function of temperature, pH, lactic acid concentration, and water activity. Applied and Environmental Microbiology 64, 1773e1779. Pretorius, W.A., 1962. Some observations on the role of coliphage on the number of Escherichia coli in oxidation ponds. Journal of Hygiene 60, 279e281. Rhodes, M.W., Kator, H., 1988. Survival of Escherichia coli and Salmonella spp. in estuarine environments. Applied and Environmental Microbiology 54, 2902e2907. Rozen, Y., Belkin, S., 2001. Survival of enteric bacteria in seawater. FEMS Microbiology Reviews 25, 513e529. Sanders, R.W., Caron, D.A., Berninger, U.G., 1992. Relationship between bacteria and heterotrophic nanoplankton in marine and freshwaters: an inter-ecosystem comparison. Marine Ecology Progress Series 86, 1e14. Sinton, L.W., Hall, C.H., Lynch, P.A., Davies-Colley, R.J., 2002. Sunlight inactivation of fecal indicator bacteria and bacteriophages from waste stabilization pond effluent in fresh and saline waters. Applied and Environmental Microbiology 68, 1122e1131. Solo-Gabriele, H.M., Wolfert, M.A., Desmarais, T.R., Palmer, C.J., 2000. Sources of Escherichia coli in a coastal subtropical environment. Applied and Environmental Microbiology 66, 230e237. Strom, S.L., 2000. Bacterivory: interactions between bacteria and their grazers. In: Kirchman, D.L. (Ed.), Microbial Ecology of the Oceans. Wiley-Liss, New York, pp. 351e386. Wolf, H.W., 1972. The coliform count as a measure of water quality. In: Mitchell, R. (Ed.), Water Pollution Microbiology. Wiley Interscience, New York, pp. 333e345. Zar, J.H., 1999. Biostatistical Analysis, fourth ed. Prentice Hall, Upper Saddle River, NJ.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 7 1 e1 5 7 8
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
The effect of real-time external resistance optimization on microbial fuel cell performance R.P. Pinto a,b, B. Srinivasan b, S.R. Guiot a, B. Tartakovsky a,b,* a b
Biotechnology Research Institute, National Research Council, 6100 Royalmount Ave., Montre´al, Que., Canada H4P 2R2 Departement de Ge´nie Chimique, E´cole Polytechnique Montre´al, C.P.6079 Succ., Centre-Ville Montre´al, Que., Canada H3C 3A7
article info
abstract
Article history:
This work evaluates the impact of the external resistance (electrical load) on the long-term
Received 23 July 2010
performance of a microbial fuel cell (MFC) and demonstrates the real-time optimization of
Received in revised form
the external resistance. For this purpose, acetate-fed MFCs were operated at external
21 November 2010
resistances, which were above, below, or equal to the internal resistance of a correspond-
Accepted 22 November 2010
ing MFC. A perturbation/observation algorithm was used for the real-time optimal selec-
Available online 30 November 2010
tion of the external resistance. MFC operation at the optimal external resistance resulted in increased power output, improved Coulombic efficiency, and low methane production.
Keywords:
Furthermore, the efficiency of the perturbation/observation algorithm for maximizing
Microbial fuel cell
long-term MFC performance was confirmed by operating an MFC fed with synthetic
External resistance
wastewater for over 40 days. In this test an average Coulombic efficiency of 29% was
Optimal control
achieved. Crown Copyright ª 2010 Published by Elsevier Ltd. All rights reserved.
1.
Introduction
The environmental impact of using fossil fuels to produce energy and their low reserves are leading to a search for renewable energy technologies. Electricity production in Microbial Fuel Cells (MFCs) from a variety of highly diluted organic matter, including wastewater, is one of such technologies (Logan and Regan, 2006; Lovley, 2008). When wastewater is used, MFCs perform waste treatment while recovering energy, thus leading to the possibility of energy-producing wastewater treatment plants. However, the low power density and the restricted output voltage of MFCs limits their industrial application (Pant et al., 2010). Therefore, intense research is now focused on improving MFC power output through the development of new anode and cathode materials (Kang et al., 2003; Logan et al., 2007; Rismani-Yazdi et al., 2008; ter Heijne et al., 2008),
better MFC design (Logan, 2010; Logan et al., 2006; Shimoyama et al., 2008), understanding of electron transfer mechanisms (Debabov, 2008; Reguera et al., 2005; Torres et al., 2010), and optimizing operational conditions (Jadhav and Ghangrekar, 2009). Furthermore, stacks of MFCs are used to increase the operating voltage (Aelterman et al., 2006; Ieropoulos et al., 2008) although challenges such as voltage reversal have been encountered (Oh and Logan, 2007), leading to significant efficiency losses. One simple alternative that is often overlooked is to enhance MFC’s power output by controlling the electrical load (i.e. external resistance) thereby always producing the maximum power output (Woodward et al., 2010). As in any electric power source, maximum power is drawn when the external resistance (Rext) equals the power source’s internal resistance (Fuel Cell Handbook, 2005). An incorrect selection of Rext, either larger or smaller than the internal resistance
* Corresponding author. Biotechnology Research Institute, National Research Council, 6100 Royalmount Ave., Montre´al, Que., Canada H4P 2R2. Tel.: þ1 514 496 2664; fax: þ1 514 496 6265. E-mail address:
[email protected] (B. Tartakovsky). 0043-1354/$ e see front matter Crown Copyright ª 2010 Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.11.033
1572
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 7 1 e1 5 7 8
(Rint), may lead to large losses in power output. The Rext control is an important requirement for industrial application of MFCs, since their Rint might vary with changes in operational parameters such as temperature, pH, influent strength, influent composition, and other factors. The problem of optimizing the external load for power sources has been addressed before by on-line control, and it is often referred to as Maximum Power Point Tracking (MPPT). Woodward et al. (2010) successfully applied and compared several non-model based real-time MPPT methods for tracking optimal Rext in MFCs fed with acetate. However, only short-term performance of the tracking algorithms was evaluated, with no attempt to study long-term consequences of MFC operation at an optimal Rext. Our recent work using a model-based simulation of microbial populations in an MFC suggests a performance decrease due to changes in microbial populations when there is a significant deviation from the optimal MFC electrical load (Pinto et al., 2010a, 2010b, 2010c). In addition, an application of the MPPT technique in the case of an MFC operated on wastewater, a more complex influent, has not been reported. Therefore, the goals of this paper are (i) to study the long-term effect of Rext on electricity and methane production in an MFC and (ii) to verify the applicability of the MPPT technique for optimizing Rext of an MFC fed with wastewater.
2.
Materials and methods
2.1.
Analytical methods
Acetate, propionate, and butyrate were analyzed on an Agilent 6890 gas chromatograph (Wilmington, DE, USA) equipped with a flame ionization detector. Method details are provided in Tartakovsky et al. (2008). Chemical oxygen demand (COD) of synthetic wastewater was estimated according to Standard Methods (APHA, 1995). Both total COD (tCOD) and soluble COD (sCOD) values were analyzed. Gas production in the MFC anodic chamber was measured on-line using glass U-tube bubble counters interfaced with a data acquisition system. Gas composition was measured using a gas chromatograph (6890 Series, Hewlett Packard, Wilmington, DE) equipped with a 11 m 3.2 mm 60/80 mesh Chromosorb 102 column (Supelco, Bellefonte, PA, USA) and a thermal conductivity detector. The carrier gas was argon. A detailed description of all analytical methods used in the study can be found in Tartakovsky et al. (2008).
2.2.
Inoculum and media composition
Each MFC was inoculated with 5 mL of anaerobic sludge with volatile suspended solids (VSS) content of approximately 40e50 g L1 (Lassonde Inc, Rougemont, QC, Canada) and 20 mL of effluent from an operating MFC. The stock solution of nutrients was composed of (in g L1): yeast extract (0.8), NH4Cl (18.7), KCl (148.1), K2HPO4 (64.0), and KH2PO4 (40.7). Concentration of acetate in the stock solution was varied in order to obtain the desired organic load by adding sodium acetate (20e80 g L1). Synthetic wastewater
stock solution had a tCOD of 48 g L1 and was composed of (in g L1): pepticase (15.0), beef extract (15.0), yeast extract (9.0), NH4HCO3 (5.1), NaCl (2.8), K2HPO4 (0.5), and KH2PO4 (0.4). A stock solution of the trace elements contained (in g L1): FeCl24H2O (2), H3BO3 (0.05), ZnCl2 (50), CuCl2 (0.03), MnCl24H2O (0.5), (NH4)6Mo7O244H2O (0.5), AlCl3 (0.5), CoCl26H2O (0.5), NiCl2 (0.5), EDTA (0.5), and concentrated HCl (1 mL). One mL of the trace elements stock solution was added to 1 L of deionized water, which was fed to the MFCs (dilution water). Deionized water was used for solution preparation, and the chemicals and reagents used were of analytical grade. All acetate solutions were sterilized by filtration (0.22 mm filtration unit) and maintained at 4 C, while and synthetic wastewater solution was frozen and maintained at 6 C until use.
2.3.
MFC design, operation, and characterization
Four single-chamber membraneless air-cathode MFCs were constructed using polycarbonate plates. The anodes were made of 5 mm thick carbon felt measuring 10 cm 5 cm (SGL Canada, Kitchener, ON, Canada). For MFC-1, MFC-2 and MFC-3 the cathodes were made of a gas diffusion electrode with a Pt load of 0.5 mg cm2 (GDE LT 120 EW, E-TEK Division, PEMEAS Fuel Cell Technologies, Somerset, NJ, USA). The MFC-4 cathode was made using ClFeTMPP (TriPorTech GmbH) on carbon Vulcan XC-72R (Cabot) as a precursor and contained 0.4% Fe with a total catalyst (Fe þ C) load of 2 mg cm2 (Birry et al., 2010).The electrodes were separated by a J-cloth with a thickness of about 0.7 mm. An external recirculation loop was installed for improved mixing of the anodic liquid. The anodic chamber temperature was maintained at 25 C by a PID temperature controller (Model JCR-33A, Shinko Technos Co., Ltd., Osaka, Japan) and a heating plate (120 V-10 W, Volton Manufacturing Ltd, Montreal, Qc, Canada). The electrical load of each MFC was controlled individually by an external resistor. MFC-1 and MFC-2 had a manually controlled external resistors, while the resistors connected to MFC-3 and MFC-4 were computer controlled digital resistors (Innoray, Montreal, QC, Canada) with a resistor variation range from 2.5 U to 1000 U. The MFCs were continuously fed with the stock solutions of carbon source (acetate or synthetic wastewater) and dilution water. The carbon source and dilution streams were combined before entering the anodic chamber. MFC-1, MFC-2, and MFC-3 were operated at several influent concentrations, while maintaining a hydraulic retention time (HRT) of 6 h. MFC-4 was operated at an HRT of 13 h. To account for process variability, each mode of operation was maintained long enough to ensure a steady state performance, which was assessed based on on-line measurements of the output voltage. Table 1 summarizes the operating conditions of each MFC. Polarization tests (PTs) were periodically performed for each MFC. In each PT Rext was disconnected for 30 min, then open circuit voltage (OCV) was measured. Subsequently, the external resistance was re-connected and progressively decreased from 1000 U to 5 U every 10 min with voltage measurements at the end of each period. The resulting voltage and current values were used to construct polarization curves, i.e. voltage vs. current plots from where the MFC’s total (ohmic
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 7 1 e1 5 7 8
Therefore, a smaller ΔR can be used to decrease the distance between Rext and the optimal external resistance, but the time of convergence will increase. A detailed description of this algorithm can be found in Woodward et al. (2010).
Table 1 e MFCs operating conditions. Organic load (g-COD/La/day)
Rext setting
Acetate
2.1, 4.3 or 8.5
Synthetic wastewater
4 or 6
1000 U (above Rint) 5 U (below Rint) Optimal (Rext wRint) Optimal (Rext wRint)
MFC 1 2 3 4
Influent
and solution) Rint was estimated by the slope of the linear region (Fan et al., 2008). Also, cathode and anode open circuit potentials were measured against an Ag/AgCl reference electrode (222 mV vs. normal hydrogen electrode). The Coulombic efficiency (CE) was estimated as: CE ¼
IMFC Dt DSnF
(1)
where I*MFC is the average current produced by the MFC during Dt [A]; Dt is the time interval (typically one day) used to calculate average current and substrate consumption [s]; DS is the amount of substrate consumed [mol-S]; n is the number of electrons transferred per mol of substrate [mol-e mol-S1]; and F is the Faraday constant [A s mol-e1]. Volumetric power output (Pout, mW L1 a ) was calculated using measurements of MFC output voltage, a corresponding value of Rext, and MFC anodic chamber volume (La). Maximum volumetric power output (Pmax) was estimated from the power curves (Pout vs current) obtained during the polarization tests. Since Rext in the PTs was changed stepwise, the accuracy of Pmax estimation was improved by using a linear interpolation of the polarization curve in the region of constant voltage drop, UMFC ¼ a0 þ a1 IMFC, where UMFC is the MFC output voltage (V), IMFC is the MFC current (A), and a0, a1 are the regression coefficients. Based on this interpolation, Pmax was calculated as described in Logan (2008), p. 47 : Pmax ¼
a20 4a1 La
(2)
Notably, for an MFC with small overpotentials a0 is close to the OCV estimation and a1 corresponds to Rint (Logan, 2008).
2.4.
Maximum power point tracking (MPPT) algorithm
The perturbation observation (P/O) method is commonly used in MPPT of solar panels (Hua and Shen, 1998). It was selected for this study because of its robustness, demonstrated in short-term MFC tests (Woodward et al., 2010), and its simplicity. The P/O algorithm applied in this work modified Rext with a predetermined amplitude (ΔR) at each iteration. The direction of resistance change was selected by comparing the value of the power output with that at the previous resistance. The method can be expressed as follows: PMFC ðk þ 1Þ PMFC ðkÞ Rext ðk þ 2Þ ¼ Rext ðk þ 1Þ þ DRsign Rext ðk þ 1Þ Rext ðkÞ
1573
(3)
where ΔR is the amplitude of change in Rext [U]; PMFC is the MFC power output [W]; and k is the iteration number. Once the algorithm converges to an optimum, the Rext will oscillate around this optimum with a maximum distance of ΔR.
3.
Results and discussion
3.1. The impact of external resistance on MFC performance The effect of Rext on long-term MFC performance was studied by simultaneously operating acetate-fed MFC-1, 2, and 3 at high, low, and optimal Rext settings, respectively, for 30e35 days (Table 1). The selected Rext values were significantly different (e.g high Rext corresponded to 1000 U and low Rext corresponded to 5 U, Table 1). This approach reduced the impact of MFC performance variability due to the microbiological nature of the process on the comparison of MFC power outputs and other performance parameters at each mode of operation. Throughout the tests, variations in influent acetate concentration were simultaneously imposed for all MFCs. The profile of acetate influent concentration is shown in Fig. 1a. Acetate concentration measurements in the effluent streams showed similar substrate removal in all MFCs, as can be seen from the values presented in Fig. 1a. At steady state, the effluent acetate concentration varied from 20 to 160 mg L1. At the same time, power outputs were quite different. Fig. 1b summarizes power production observed throughout the tests. This figure shows that in all MFCs power output began to increase after approximately 3 days of operation, reaching steady state values after 7e10 days. Power outputs at steady state strongly depended on Rext selection with power outputs around 6, 15, and 58 mW L1 a , observed for MFC-1 (high Rext), MFC-2 (low Rext), and MFC-3 (optimal Rext), respectively. Notably on day 13 the cathode of MFC-3 was punctured and replaced by a new cathode, made of the same material. Following the replacement, MFC-3 power output approached 135 mW L1 a for about two days before returning to its previous level. This period was excluded from Fig. 1b. Also, due to technical problems, MFC-2 voltage was not recorded between days 8e12, 14.5 to 16.5, and 28.5 to 30.5. Fig. 1c presents changes in Rext of MFC-3 imposed by the P/O algorithm over time. At startup, Rext value maintained by the algorithm was above 400 U, then after about 5 days of operation Rext sharply decreased to values below 40 U. This figure also shows Rint values estimated for MFC-3 during polarization tests. As expected, the P/O algorithm provided timely adjustment of Rext to maintain it close to Rint values, such that the Rext oscillated around the optimum value equal to MFC-3 internal resistance. The profile of Rint change in the MFC-2 test was similar to that observed for MFC-3 rapidly decreasing to 15e30 U after first 6 days of MFC operation, while Rint of MFC-1 remained at around 200 U for most of the test. While the growth of anodophilic microorganisms during the startup period resulted in relatively slow changes in Rint, the variations in operating conditions had an almost immediate impact on Rint and therefore on MFC performance. The P/O algorithm’s ability to track fast variations of operating conditions (e.g. influent composition) during MFC-3 operation
1574
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 7 1 e1 5 7 8
b
Influent MFC-1 MFC-2 MFC-3
-1
1
MFC-3
60 40
0
0 5
0
10 15 20 25 30 35
5
10 15 20 25 30 35
Time (d)
Time (d)
d R ext
75 50 25 0
0
5
50
65
Pout
40
R int Rext (Ω)
Resistance (Ω)
100
30
45
20
35
10 23.75
10 15 20 25 30 35
55
Rext
25 24.25
24
Time (d)
-1
0
c
MFC-2
MFC-1
20
Pout (mW L a )
-1 Acetate (g L )
2
Pout (mW La )
a
Time (d)
Fig. 1 e (a) Acetate concentration in the influent and effluent, and (b) power production for MFC-1 MFC-2, and MFC-3 against time; (c) Rext and Rint values for MFC-3 (MFC-1 and MFC-2 Rext values were always kept at 1000 U and at 5 U, respectively); (d ) Rext and Pout values during an increase in the MFC-3 influent concentration at t [ 23.9 days.
is illustrated in Fig. 1d. Here, the influent acetate concentration was increased from 0.5 to 2 g L1 on day 24. This increase in acetate concentration caused a decrease of Rint and, accordingly, the MPPT algorithm decreased the Rext value thus maximizing power output under new operating conditions. Fig. 2 presents estimations of current density and Coulombic efficiency (CE ) for MFC-1, MFC-2 and MFC-3. As expected, MFC-1, which was operated at a high Rext, always had a low current density and a low CE, while MFC-2 and MFC-3 showed larger values. MFC-2, which was operated at the lowest Rext, was expected to have the highest Coulombic efficiency. However, current densities (Fig. 2a) and CE (Fig. 2b) of MFC-3, which was operated at an optimal Rext, on average were slightly
b
3
100%
MFC-1 MFC-3
80%
2 1.5 MFC-1 MFC-2 MFC-3
1 0.5
2.5 2
-1
2.5
MFC-2 Influent
60%
1.5
40%
1
20%
0
0.5 0
0%
0
5
10
15
20
Time (d)
25
30
35
Acetate (g L )
3.5
CE (%)
-2
Current density (A m )
a
higher than that of MFC-2. CE calculations showed a more pronounced difference between MFC-2 and MFC-3, however this difference could be attributed to the variability of the effluent acetate measurements. Also, MFC-3 featured the shortest startup time, as can be seen from the comparison of current densities in Fig. 2a. Polarization tests provided additional information for the comparison of MFC performances. Fig. 3a shows the evolution of the cathode and anode open circuit potentials (OCP) over time. As expected, cathode OCP values were similar for all MFCs and remained constant throughout the experiment. Following the startup, anode OCP values decreased for all MFCs, with the fastest decrease observed for MFC-2 (low Rext),
0
5
10
15
20
25
30
35
Time (d)
Fig. 2 e Current density (a) and Coulombic efficiency (b) observed at various acetate loads (Fig. 1a).
1575
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 7 1 e1 5 7 8
a
100
Cathode
150
0
5
10
-150
15
20
25
30
35
MFC-1 MFC-2 MFC-3
-300 -450
Anode
-1
0
-600
b
80
Time (d)
Pmax (mW L a )
Potential vs. Ag/AgCl (mV)
300
60 MFC-1
40
MFC-2 20
MFC-3
0
0
5
10
15
20
25
30
35
Time (d) Fig. 3 e (a) The evolution of the cathode and anode OCP values over time, and (b) maximum power densities estimated using polarization curves according to Eq. (2).
followed by MFC-3 (optimal Rext). MFC-1, which was operated at a high Rext, was the last to reach a steady state value of the anode potential. Nevertheless, the anode OCP values for all MFCs were similar after 15 days of testing. It can be hypothesized that this pattern of anode OCP decrease over time was reflective of anode colonization by the anodophilic microorganisms. Indeed, lower values of Rext facilitate the electron transfer process thus providing growth advantages to the anodophilic microorganisms. Consequently MFC-2 operated at the lowest Rext featured the fastest rate of the anodophilic biofilm formation. However, MFC operation at an Rext below Rint results in low power output (Logan, 2008), thus requiring Rext optimization. When maximal power outputs were estimated from the polarization test results, it was observed that MFC-1 always had low Pmax, never exceeding 9 mW L1 a . Maximal power outputs estimated for MFC-2 and MFC-3 increased during the first 15e20 days of the experiment (Fig. 3b). A maximal power output of 95 mW L1 a was estimated based on the polarization test for MFC-3 on day 15, which was carried out shortly after the cathode replacement in this MFC. As mentioned above, the cathode replacement resulted in higher than usual power output between days 13 and 15. On average, MFC-3 maximal power output remained around 70 mW L1 a , which agreed well with the power densities measured during MFC-3 test (Fig. 1b). At the same time, power output of MFC-2 during normal operation was very low because of the Rext choice. The selection of Rext for MFC-1 (Rext ¼ 1000 U) and MFC-2 (Rext ¼ 5 U) was confirmed by the polarization tests. After the startup period, Rint in MFC-1 varied between 50 and 200 U, while in MFC-2 Rint varied between 15 and 25 U. Thus, MFC-1 was operated at Rext >> Rint, and MFC-2 was operated at Rext << Rint throughout the test, as intended. Besides power output comparison, methane production in the anodic compartment of each MFC was measured throughout the tests and was used to compare the long-term effects of Rext selection on methane production. Since the MFCs were inoculated with anaerobic sludge, methane production was observed at the beginning of the operation in all MFCs. The methane production in MFC-1 increased over time, while it decreased in MFC-2 and MFC-3. To compare electricity and methane production for MFC-1, MFC-2, and
MFC-3 steady state values were calculated using experimental results obtained between day 7 and 30 of MFC operation, when constant organic load was applied. The average methane production rates were 2.0 0.7, 0.2 0.1, 0.6 0.2 mL d1 in MFC-1, MFC-2, and MFC-3, respectively. The corresponding power outputs for these MFCs were 5.6 0.3, 14.6 7.7, and 60.8 17.2 mW L1 a . It should be acknowledged that methane flow measurements were complicated by the small volume (50 mL) of the anodic compartments and low methane production rates, in the range of several mL per day. Measurements of such small flow rates resulted in large standard deviations. Nevertheless, the overall trends were clear showing significantly higher methane production in MFC-1 operated at high Rext. In contrast, methane production in MFC-1 and MFC-2 was very low. These results agree with methane production measurements reported in Martin et al. (2010), where MFCs were operated at Rext above estimated values of Rint resulting in significant methane production. It was previously demonstrated using biomolecular methods, that a high Rext, which implies MFC operation at more positive anode potentials, decreases the amount of the anodophilic microorganisms (Torres et al., 2009). There is a significant body of evidence proving that the external resistance at which the MFC operates has an influence on the microbial communities and the long-term performance (Aelterman et al., 2008; Chae et al., 2010; Lyon et al., 2010; Pinto et al., 2010c; Torres et al., 2009). Aelterman et al. (2008) observed the impact of Rext on electricity and methane production in an MFC inoculated with a mixed anaerobic culture and concluded that low methane production and stable power output are only obtained if Rext is set close to the MFC internal resistance. Furthermore, Chae et al. (2010) compared methanogenic activity and methane production in MFCs subjected to several external perturbations (pH, temperature, oxygen exposure, addition of a methanogenesis inhibitor, and Rext variation). They also concluded that electricity production was increased and methane production was decreased only when a methanogenesis inhibitor (BES, 2-bromoethanesulfonate) was added to the anodic chamber or by setting a Rext close to the Rint values. Computer simulations using a two population MFC model (Pinto et al., 2010c) corroborated with the experimental
1576
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 7 1 e1 5 7 8
evidence above. The model was used to simulate the outcome of the competition between the anodophilic and methanogenic microorganisms for a common carbon source (acetate) in MFCs operated at different organic loads and Rext settings (Pinto et al., 2010c). It was predicted that, independent of the microbial composition of the inoculum, proliferation of the anodophilic microorganisms could only be achieved at Rext values that are equal to or less than the MFC internal resistance. The same conclusion can be inferred from the results of Lyon et al. (2010). In this study the microbial composition of biofilms sampled from MFCs inoculated with the same sludge and operated at different Rext values were analyzed. By using Ribosomal Intergenic Spacer Analysis (RISA) they demonstrated, based on the profiles observed, that the microbial community structure in MFC’s biofilm operated at Rext appreciably above Rint (1 kU and 10 kU) was significantly different from that observed in the MFC operated at low Rext (10 U, 100 U, and 470 U). As mentioned above, a low Rext promotes growth and metabolic activity of the anodophilic microorganisms since electron transport to the cathode is facilitated. However Rext, which is lower than the MFC Rint value leads to a low power output, i.e. an optimal Rext value (Rext w Rint) should be maintained. Notably, all of the tests mentioned above were carried out by manual adjustment of Rext without using a real-time algorithm, which would guarantee timely correction of Rext. The results from the citations above are in perfect agreement with the findings presented in this paper, where acetatefed MFCs were inoculated with the same anaerobic sludge and consumed comparable quantities of carbon source. They were operated at high, low, or optimal Rext settings, leading to Coulombic efficiencies remarkably different (Fig. 2b). High Rext led to, essentially, an anaerobic reactor with low Coulombic efficiency and significant methane production. At the same time, both MFC-2 (low Rext) and MFC-3 (optimal Rext) featured high Coulombic efficiency, and by the end of the 30 day test methane production in these MFCs declined to near zero values. Notably, the high power output with negligible methane production observed during MFC-3 operation are in agreement with the results of ClFeTMPP cathode tests, where an MFC was also operated using the P/O algorithm for the online optimization of Rext (Birry et al., 2010). Considering that MFC operation at Rext values below Rint leads to a sharp drop in power output (Fuel Cell Handbook, 2005), it is sufficient to maintain Rext at an optimal value in order to minimize methane production and maximize power production.
3.2.
MFC operation on synthetic wastewater
The efficiency of the P/O algorithm when operating an MFC on a complex feed was confirmed by feeding MFC-4 with synthetic wastewater. The power output observed during this test and Rext values selected by the P/O algorithm are shown in Fig. 4. This graph also shows Rint estimations obtained in the polarization tests. A sharp decrease in Rint values obtained from the polarization tests and Rext values selected by the P/O algorithm can be seen after the first 5 days of MFC operation (Fig. 4c). After this point, Rint remained between 20 U and 60 U with Rext matching these values. As one can see, Pout increased steadily during the first 15 days of the experiment (Fig. 4b),
which is longer than the 5 day startup period observed in the acetate-fed MFC-3 (Fig. 1b). A longer startup period could be attributed to the use of synthetic wastewater containing complex organic matter. We hypothesize that wastewater hydrolysis and fermentation steps, which were required because of the complex wastewater composition, limited the amount of volatile fatty acids available for growth and metabolism of the anodophilic microorganisms. To demonstrate the robustness of the P/O algorithm, MFC4 was subjected to two types of perturbations: First the organic load was increased by 50% between days 16 and 20 (Fig. 4a), then between days 23 and 33 the MFC temperature was increased from 25 C to 35 C (Fig. 4d). The increase in organic load did not result in a substantial increase of power production and Rint remained unchanged (Fig. 4b,c). It can be seen that the increased organic load did not result in an immediate increase in the effluent COD concentration (Fig. 4a), likely due to a short duration of the test. Effluent VFA analysis showed that concentrations of acetate, propionate, and butyrate always remained below 70e90 mg L1 during the test. Apparently, the biotransformation process was limited by the rate of hydrolysis rather than the rate of VFA consumption by the anodophilic microorganisms. Accordingly, Rint remained unchanged during the test. MFC response to temperature increase was more pronounced with an immediate drop in Rext from about 48 U to 30 U (40% decrease) and a 50% increase of Pout. Once MFC temperature was returned to 25 C, the P/O algorithm adjusted Rext close to its previous level. A similar response to a temperature perturbation was observed by Woodward et al. (2010). Notably, an additional external perturbation during MFC-4 test was caused by the fermentation of the synthetic wastewater solution kept at room temperature in a syringe. This feeding solution was replaced every two or three days. Fermentation of the synthetic wastewater resulted in a visible change in the feed solution appearance with an average of 0.15 g L1 of volatile fatty acids at the end of each feeding period, although tCOD concentration remained constant. The arrows in Fig. 4b represent the times when the synthetic wastewater was replenished. It can be seen that synthetic wastewater fermentation had a direct impact on the Pout. After each wastewater replenishment the MFC power output declined for about 24 h, while reaching its maximal values by the end of each feeding period. Also, abrupt drops in Pout observed on days 25, 28, and 33 are related to delays in replacing the syringe, which resulted in an interruption of the feed. These perturbations of the organic load were promptly followed by the P/O algorithm thus demonstrating its excellent robustness, as can be seen from the results presented in Fig. 4b. Polarization tests not only confirmed the choice of Rext values by the P/O algorithm, but also demonstrated timerelated changes in OCV and Pmax values resulting from MFC operation at an optimal Rext. OCV and Pmax values presented in Fig. 5 show that OCV slowly increased in about 20 days stabilizing around 500 mV, while Pmax more then doubled after day 7 and fluctuated between 17 and 35 mW L1 a with higher outputs observed by the end of the test. Overall, the Coulombic efficiency of MFC-4 varied between 14% and 41%, with an average of 29% throughout the experiment.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 7 1 e1 5 7 8
1577
Fig. 4 e MFC-4 performance: (a) total COD in the influent (tCOD-in) and soluble COD in the effluent (sCOD-out). The organic load was increased by 50% between days 16 and 20; (b) power output; (c) external and internal resistance; (d ) External resistance and power density response during an increase in temperature from 25 C to 35 C at 23.25 days. Arrows show syringe replacement times.
During MFC-3 and MFC-4 operation, real-time correction of Rext was instrumental in maintaining its value at or close to Rint. Indeed, because of the variations of organic load, composition, and MFC temperature during the tests, the internal resistance of the MFCs varied in time. Polarization tests performed after each change in operating conditions were used to estimate Rint variations, as can be seen in Fig. 1c and Fig. 4c, where Rint estimations are shown along with Rext values selected by the P/O algorithm. The P/O algorithm demonstrated excellent stability and fast convergence so that 50
600 500
OCV (mV)
-1
Pmax (mW La )
40
400 30 300 20 200 OCV
100 0
Rext always remained close Rint as illustrated in Fig. 1d and Fig. 4d. The real-time strategy for Rext control also allowed for avoiding a sharp decrease in MFC performance even when MFC feed was interrupted due to a technical problem (e.g. syringe pump malfunction or a delay in syringe replacement). These events led to a sharp short-term increase in the internal resistance, which was successfully tracked by the P/O algorithm (e.g. days 25, 28 and 33 in Fig. 4c). Notably, implementing an MPPT on-line technique for MFC’s Rext control could be used for individual control of electric loads of MFC stacks. This strategy might prevent MFC operation at Rext values below its Rint thus helping to avoid voltage reversal (Oh and Logan, 2007) after a feed disruption or another operating problem.
10
Pmax 0
10
20
30
40
0
Time (d) Fig. 5 e Open circuit voltage and maximum power output estimated from the polarization tests in MFC-4.
4.
Conclusion
In this study, a simple perturbation/observation algorithm was used to maximize MFC power output by matching Rext and Rint values. The real-time optimization of Rext was tested through long-term operation of MFCs fed with either acetate or synthetic wastewater. The P/O algorithm demonstrated an excellent performance when the MFC was subject to perturbations in operating conditions such as variations in organic load, influent composition, and temperature. A comparison of MFC performance at an optimal Rext value, with MFCs operated at either high (Rext >> Rint) or low (Rext << Rint) external resistances showed that real-time resistance optimization led to significantly higher power outputs with less methane
1578
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 7 1 e1 5 7 8
production. MFCs operated at an optimal resistance showed average Coulombic efficiencies of 57% and 29% on acetate and synthetic wastewater, respectively. It is concluded that the real-time optimal control of MFC electrical load might be instrumental in the development of stackable MFCs for combined wastewater treatment and power production.
Acknowledgement This research was supported by the National Research Council of Canada (NRC publication #53342).
references
Aelterman, P., Rabaey, K., Pham, H.T., Boon, N., Verstraete, W., 2006. Continuous electricity generation at high voltages and currents using stacked microbial fuel cells. Environmental Science and Technology 40 (10), 3388e3394. Aelterman, P., Versichele, M., Marzorati, M., Boon, N., Verstraete, W., 2008. Loading rate and external resistance control the electricity generation of microbial fuel cells with different three-dimensional anodes. Bioresource Technology 99 (18), 8895e8902. APHA, 1995. Standard Methods for the Examination of Water and Wastewater. American Public Health Association, Washington, DC. Birry, L., Mehta, P., Jaouen, F., Dodelet, J.P., Guiot, S.R., Tartakovsky, B., 2010. Application of iron-based cathode catalysts in a microbial fuel cell. Electrochimica Acta, in press. Chae, K.-J., Choi, M.-J., Kim, K.-Y., Ajayi, F.F., Park, W., Kim, C.-W., Kim, I.S., 2010. Methanogenesis control by employing various environmental stress conditions in two-chambered microbial fuel cells. Bioresource Technology 101 (14), 5350e5357. Debabov, V., 2008. Electricity from microorganisms. Microbiology 77 (2), 123e131. Fan, Y., Sharbrough, E., Liu, H., 2008. Quantification of the internal resistance distribution of microbial fuel cells. Environmental Science and Technology 42, 8101e8107. Fuel Cell Handbook, 2005. National Energy Technology Laboratory U.S. Department of Energy. University Press of the Pacific. Hua, C., Shen, C., 1998. Comparative Study of Peak Power Tracking Techniques for Solar Storage System. In: Anon (Ed.). IEEE, Anaheim, CA, USA, pp. 679e685. Ieropoulos, I., Greenman, J., Melhuish, C., 2008. Microbial fuel cells based on carbon veil electrodes: stack configuration and scalability. International Journal of Energy Research 32, 1228e1240. Jadhav, G.S., Ghangrekar, M.M., 2009. Performance of microbial fuel cell subjected to variation in pH, temperature, external load and substrate concentration. Bioresource Technology 100 (2), 717e723. Kang, K.H., Jang, J.K., Pham, T.H., Moon, H., Chang, I.S., Kim, B.H., 2003. A microbial fuel cell with improved cathode reaction as a low biochemical oxygen demand sensor. Biotechnology Letters 25 (16), 1357e1361. Logan, B., Cheng, S., Watson, V., Estadt, G., 2007. Graphite fiber brush anodes for increased power production in air-cathode microbial fuel cells. Environmental Science and Technology 41, 3341e3346. Logan, B.E., 2008. Microbial Fuel Cells. John Wiley & Sons, Hobeken, New Jersey.
Logan, B.E., 2010. Scaling up microbial fuel cells and other bioelectrochemical systems. Applied Microbiology and Biotechnology 85 (6), 1665e1671. Logan, B.E., Hamelers, B., Rozendal, R.A., Schroder, U., Keller, J., Freguia, S., Aelterman, P., Verstraete, W., Rabaey, K., 2006. Microbial fuel cells: methodology and Technology. Environmental Science and Technology 40 (17), 5181e5192. Logan, B.E., Regan, J.M., 2006. Microbial fuel cells - Challenges and applications. Environmental Science and Technology 40 (17), 5172e5180. Lovley, D.R., 2008. The microbe electric: conversion of organic matter to electricity. Current Opinion in Biotechnology 19 (6), 564e571. Lyon, D.Y., Buret, F., Vogel, T.M., Monier, J.-M., 2010. Is resistance futile? Changing external resistance does not improve microbial fuel cell performance. Bioelectrochemistry 78 (1), 2e7. Martin, E., Savadogo, O., Guiot, S.R., Tartakovsky, B., 2010. The influence of operational conditions on the performance of a microbial fuel cell seeded with mesophilic anaerobic sludge. Biochemical Engineering Journal 51 (3), 132e139. Oh, S.E., Logan, B.E., 2007. Voltage reversal during microbial fuel cell stack operation. Journal of Power Sources 167 (1), 11e17. Pant, D., Van Bogaert, G., Diels, L., Vanbroekhoven, K., 2010. A review of the substrates used in microbial fuel cells (MFCs) for sustainable energy production. Bioresource Technology 101 (6), 1533e1543. Pinto, R.P., Perrier, M., Tartakovsky, B. and Srinivasan, B., 2010a. Performance analyses of microbial fuel cells operated in series, Proceedings of 9th International Symposium on Dynamics and control of process systems, Leuven, Belgium. Pinto, R.P., Srinivasan, B., Manuel, M.F., Tartakovsky, B., 2010b. A two-population bio-electrochemical model of a microbial fuel cell. Bioresource Technology 101 (14), 5256e5265. Pinto, R.P., Tartakovsky, B., Perrier, M., Srinivasan, B., 2010c. Optimizing treatment performance of microbial fuel cells by reactor Staging. Industrial & Engineering Chemistry Research 49 (19), 9222e9229. Reguera, G., McCarthy, K.D., Mehta, T., Nicoll, J.S., Tuominen, M.T., Lovley, D.R., 2005. Extracellular electron transfer via microbial nanowires. Nature Biotechnology 435, 1098e1101. Rismani-Yazdi, H., Carver, S.M., Christy, A.D., Tuovinen, I.H., 2008. Cathodic limitations in microbial fuel cells: an overview. Journal of Power Sources 180, 683e694. Shimoyama, T., Komukai, S., Yamazawa, A., Ueno, Y., Logan, B.E., Watanabe, K., 2008. Electricity generation from model organic wastewater in a cassette-electrode microbial fuel cell. Applied Microbiology and Biotechnology 80 (2), 325e330. Tartakovsky, B., Manuel, M.F., Neburchilov, V., Wang, H., Guiot, S.R., 2008. Biocatalyzed hydrogen production in a continuous flow microbial fuel cell with a gas phase cathode. Journal of Power Sources 182 (1), 291e297. ter Heijne, A., Hamelers, H.V.M., Saakes, M., Buisman, C.J.N., 2008. Performance of non-porous graphite and titanium-based anodes in microbial fuel cells. Electrochemica Acta 53, 5697e5703. Torres, C.I., Krajmalnik-Brown, R., Parameswaran, P., Marcus, A.K., Wanger, G., Gorby, Y.A., Rittmann, B.E., 2009. Selecting anoderespiring bacteria based on anode potential: phylogenetic, electrochemical, and microscopic characterization. Environmental Science and Technology 43 (24), 9519e9524. Torres, C.I., Marcus, A.K., Lee, H., Parameswaran, P., KrajmalnikBrown, R., Rittmann, B.E., 2010. A kinetic perspective on extracellular electron transfer by anode-respiring bacteria. FEMS Microbiology Reviews 34 (1), 3e17. Woodward, L., Perrier, M., Srinivasan, B., Pinto, R.P., Tartakovsky, B., 2010. Comparison of real-time methods for maximizing power output in microbial fuel cells. AIChE Journal 56 (10), 2742e2750.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 7 9 e1 5 8 6
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Electrochemical oxidation of trace organic contaminants in reverse osmosis concentrate using RuO2/IrO2-coated titanium anodes Jelena Radjenovic*, Arseto Bagastyo, Rene´ A. Rozendal, Yang Mu, Ju¨rg Keller, Korneel Rabaey Advanced Water Management Centre, The University of Queensland, Brisbane, QLD 4072, Australia
article info
abstract
Article history:
During membrane treatment of secondary effluent from wastewater treatment plants,
Received 26 July 2010
a reverse osmosis concentrate (ROC) containing trace organic contaminants is generated.
Received in revised form
As the latter are of concern, effective and economic treatment methods are required. Here,
24 November 2010
we investigated electrochemical oxidation of ROC using Ti/Ru0.7Ir0.3O2 electrodes, focussing
Accepted 24 November 2010
on the removal of dissolved organic carbon (DOC), specific ultra-violet absorbance at
Available online 1 December 2010
254 nm (SUVA254), and 28 pharmaceuticals and pesticides frequently encountered in secondary treated effluents. The experiments were conducted in a continuously fed reactor
Keywords:
at current densities (J ) ranging from 1 to 250 A m2 anode, and a batch reactor at
Water recycling
J ¼ 250 A m2. Higher mineralization efficiency was observed during batch oxidation (e.g.
Reverse osmosis
25.1 2.7% DOC removal vs 0% removal in the continuous reactor after applying specific
Advanced oxidation
electrical charge, Q ¼ 437.0 A h m3 ROC), indicating that DOC removal is depending on
Pharmaceuticals
indirect oxidation by electrogenerated oxidants that accumulate in the bulk liquid. An
Pesticides
initial increase and subsequent slow decrease in SUVA254 during batch mode suggests the
LC-MS
introduction of auxochrome substituents (e.g. eCl, NH2Cl, -Br, and -OH) into the aromatic
Microtox test
compounds. Contrarily, in the continuous reactor ring-cleaving oxidation products were generated, and SUVA254 removal correlated with applied charge. Furthermore, 20 of the target pharmaceuticals and pesticides completely disappeared in both the continuous and batch experiments when applying J 150 A m2 (i.e. Q 461.5 A h m3) and 437.0 A h m3 (J ¼ 250 A m2), respectively. Compounds that were more persistent during continuous oxidation were characterized by the presence of electrophilic groups on the aromatic ring (e.g. triclopyr) or by the absence of stronger nucleophilic substituents (e.g. ibuprofen). These pollutants were oxidized when applying higher specific electrical charge in batch mode (i.e. 1.45 kA h m3 ROC). However, baseline toxicity as determined by Vibrio fischeri bioluminescence inhibition tests (Microtox) was increasing with higher applied charge during batch and continuous oxidation, indicating the formation of toxic oxidation products, possibly chlorinated and brominated organic compounds. ª 2010 Elsevier Ltd. All rights reserved.
* Corresponding author. Tel.: þ61 7 3346 3234; fax: þ61 7 3365 4726. E-mail address:
[email protected] (J. Radjenovic). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.11.035
1580
1.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 7 9 e1 5 8 6
Introduction
Due to the growing pressure on water resources, the use of treated municipal wastewater for groundwater recharge and indirect potable reuse is increasingly considered. A number of reuse facilities worldwide currently employ microfiltration (MF) followed by reverse osmosis (RO) for the treatment of secondary treated effluent prior to aquifer or reservoir recharge. High-pressure RO membranes have gained popularity due to their outstanding performance in rejecting trace organics such as endocrine disrupting compounds, pesticides, pharmaceuticals and others (Bellona and Drewes, 2007; Radjenovic et al., 2008; Snyder et al., 2007). These compounds will be concentrated four to seven times in the waste stream (reverse osmosis concentrate, ROC) that normally comprises about 15e25% of the incoming water flow. The ROC offers an opportunity for reducing human and ecotoxicological risk by implementing brine treatment prior to environmental discharge. Several advanced oxidation treatment options (e.g. TiO2 photocatalysis, sonolysis) showed moderate performance in removing the organic matter from ROC (Dialynas et al., 2008; Westerhoff et al., 2009). In the past years, electrochemical oxidation processes received renewed interest due to several perceived advantages such as efficient control of reaction conditions, no chemical requirements, simplicity and robustness of operation at ambient temperature and pressure. Van Hege and co-workers pioneered the electrochemical oxidation of ROC (Van Hege et al., 2002). More recent work by Dialynas et al. (2008) reported moderate dissolved organic carbon (DOC) removal in electrolytic oxidation of ROC on boron-doped diamond (BDD) anodes, while Perez et al. (2010) observed an excellent performance by BDD in eliminating chemical oxygen demand (COD) and 10 pharmaceuticals and stimulant drugs from ROC. Due to recent advances in electrode stability and performance, electrochemically driven processes are becoming an attractive option for the remediation of problematic waste streams. Notably, mixed metal oxide (MMO) coated electrodes have found widespread environmental applications in recent years for the treatment of pesticide contaminated water, landfill leachate, organic petroleum wastewater and other difficult to treat waste streams (Panizza, 2010). MMO anodes such as RuO2/IrO2-coated titanium with improved electrocatalytic behaviour and stability are readily available in practical mesh geometries and have extended life-time and lower costs relative to BDD electrodes. The latter have thus far been considered as the standard for electrochemical oxidation, but suffer from high product costs. Because of their low overpotential for chlorine evolution, RuO2/IrO2-coated electrodes were effectively used for the degradation of pharmaceuticals, pesticides and other organic compounds via indirect electrolysis (Malpass et al., 2006; Carlesi Jara et al., 2007; Gallard et al., 2004). However, while on one hand in-situ generated active chlorine can effectively oxidize many pollutants, on the other hand the formation of chlorinated byproducts could lead to increased toxicity levels. Different operational strategies were proposed in literature for minimizing the formation of chlorinated organic compounds, such
as application of long residence time and activated carbon polishing treatment (Rajkumar et al., 2005), continuous operation and short residence time (Bergmann and Koparal, 2005a), and over saturation of the solution by chlorine. The latter approach increases the degradation rate of chlorinated intermediates relative to their rate of formation (Gallard et al., 2004). In this study, the electrochemical oxidation using Ti/ Ru0.7Ir0.3O2 electrode was investigated for the treatment of ROC. The process efficiency was evaluated based on the removal of DOC, specific ultra-violet absorbance at 254 nm (SUVA254), and 28 pesticides and pharmaceuticals encompassing diverse molecular structures and physico-chemical properties (Table S1), selected for their ubiquity in municipal wastewater effluents and brine streams. The influence of operational mode, current density, and applied electrical charge on the oxidation efficiency was investigated. The baseline toxicity of electrochemically oxidized ROC was evaluated in the bioluminescence inhibition tests (Microtox) using the marine bacterium Vibrio fischeri.
2.
Materials and methods
2.1.
Chemicals
All standards for pharmaceuticals and pesticides used were of analytical grade (99%) (Text S1). All solvents (methanol, acetonitrile and water) were HPLC-grade and were purchased from Merck (Germany), as well as hydrochloric acid (37%), ammonium acetate, sodium hydroxide and formic acid (98%).
2.2.
Reverse osmosis concentrate
The ROC used in the experiments was sampled at an advanced water treatment plant (AWTP) in Bundamba, 30 km west of Brisbane, Australia. This AWTP receives a mixture of secondary treated effluents from four wastewater treatment plants. After the pre-treatment of secondary effluents (addition of iron coagulants, separation of solids in the clarifier), raw water is passed through MF and RO membranes. The generated ROC is further subjected to nitrification in a moving bed biofilm reactor (MBBR), coagulation and denitrification by anoxic filters, and is finally discharged into the Brisbane River. In order to investigate the removal of trace organic contaminants during electrochemical oxidation, ROC collected prior to the nitrification stage was spiked with the concentrated solutions of target compounds prepared in water and methanol. The spiking solution of compounds soluble in water was prepared from pure standards at w1 mg L1 concentration, and for each 10 L of ROC, 100 mL of spiking solution in water was added. The spiking solution of compounds poorly soluble in water was prepared in methanol at 1 g L1 concentration, and for each 10 L of ROC, 100 mL of spiking solution in methanol was added. Since methanol can act as a scavenger of the generated oxidants, the amount added was minimized. In order to ensure homogeneity, the spiked ROC sample was filtered through 0.7 mm glass fibre filters (Whatman, UK) prior to the experiments. The obtained final concentrations of
1581
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 7 9 e1 5 8 6
target analytes were in the range of 7.8 (carbamazepine) to 37.4 mg L1 (iopromide) (Table S2).
2.3.
Experimental setup
The electrochemical cell was constructed by assembling two equal rectangular polycarbonate frames with internal dimensions of 20 5 1.2 cm. The frames were bolted together between two polycarbonate square plates. The anode cell was separated from the cathode cell by a cation exchange membrane Ultrex CMI-7000 (Membranes International, U.S.A.). Sealing was ensured by a rubber o-ring inserted between the two frames. The total volume for each compartment was 114 mL. The anode used was an MMO Ti/ Ru0.7Ir0.3O2 electrode with 12 g m2 coating on Ti mesh (dimensions: 4.8 5 cm; thickness: 1 mm; specific surface area: 1.0 m2 m2), supplied by Magneto Special Anodes (The Netherlands). The cathode used was a stainless steel woven wire mesh (dimensions: 4.8 5 cm; 80 mm 0.050 mm wire diameter), and both the anode and cathode had a projected electrode surface area of 24 cm2. For electrochemical control, a Wenking potentiostat/galvanostat (KP07, Bank Elektronik, GmbH, Pohlheim, Germany) was used. The anodic half-cell potentials were measured by placing an Ag/AgCl reference electrode (assumed þ0.197 V vs SHE) in the anode compartment. The cathode medium in was a 0.1M HCl solution. The experiments were performed at constant liquid flowrate at room temperature (25 1 C), with galvanostatic control in: i) continuous mode, with step-wise increase in current density (J ) (J ¼ 1, 10, 30, 50, 100, 150, 200 and 250 A m2), and samples taken after 75 min of operation at each current density, and ii) batch mode at J ¼ 250 A m2, and samples taken after 2, 4, 7 and 23.5 h. In the continuous mode experiments, the system was operated at a hydraulic retention time (HRT) of 8.8 min (i.e. flow-rate of 13 mL min1). In order to maintain well-mixed conditions and avoid concentration gradients, both anolyte and catholyte were recirculated internally at a rate of 162 mL min1. In order to avoid gas trapping inside the anodic and cathodic compartment and enable stable potentials and ambient pressure, two degassers were installed as illustrated in Figure S1. The volume of ROC in the degassers was maintained at 100 mL, providing a ratio of active and total volume, VACT/VTOT of 0.51 in the continuous mode. Preparative continuous oxidation experiments were carried out in order to establish the time required to reach steady state. In the batch mode experiments, anolyte and catholyte were recirculated at a rate of 162 mL min1. The volume of ROC used in batch experiments was 10 L, hence the ratio of VACT/VTOT was 0.011. The results of each experimental condition are given as the means of triplicates, with their corresponding standard deviations (SDs). In the continuous
experiments two different ROC samples were used, collected at the abovementioned AWTP within a one month time span and marked as ROC-1 and ROC-2 (physico-chemical characteristics of ROC-1 and ROC-2 are given in Table 1), while in batch experiments only ROC-1 was used.
2.4.
Analytical methods
130 mL samples were collected in amber glass bottles. As a variety of generated oxidants can prevent efficient sample stabilization, no quenching agent was added to the sample. Furthermore, quenching of free chlorine may lead to errors in analytical determination of trace organics due to the formation of the original compounds from their N-chloro analogues (Wulfeck-Kleier et al., 2010). Immediately after the sampling, sample pH was adjusted to pH 7.0 by adding an appropriate amount of 0.1M NaOH or 0.1M HCl, and 100 mL samples were extracted on a Visiprep manifold system (SigmaeAldrich, U.S.A.) using Oasis HLB cartridges (200 mg, 6 mL) from Waters Corporation (U.S.A.), previously conditioned with 10 mL of methanol and 10 mL of deionised water (HPLC grade). Additionally, 30 mL samples were taken for the analyses of free and combined chlorine, non-purgeable organic carbon (NPOC) and ultra-violet absorption at 254 nm (UV254). The samples were filtered prior to all measurements using 0.45 mm filters (Millipore, Ireland), thus the determined NPOC is equivalent to DOC. NPOC was calculated as the difference between the total carbon (TC) and inorganic carbon that were determined by the standard high-temperature method (APHA Standard methods, 5310B) at a TC analyser (Tekmar Dohrmann DC-190), and UV254 absorbance was measured using a Varian Cary50 spectrophotometer. Ion chromatography (IC)Dionex 2010 i was used to determine Cl and SO24 , while concentrations of Fe2þ and Mn2þ ions were determined by ICPOES Vista Pro-CCD (Varian, Australia). Conductivity was measured using a Eutech electrical conductivity meter, while the pH was measured with a Mettler Toledo Seven easy pH meter (Mettler Toledo, Australia). Free available chlorine (FAC) and total chlorine were measured with the N,N-diethyl-pphenylenediamine (DPD) ferrous titrimetric method (APHA Standard methods, 409E). It is important to stress that oxidants other than FAC present in the solution (e.g. H2O2, ClO2) will similarly to chlorine react with DPD to form a red dye, thus interfering with the measurements. Liquid chromatography-mass spectrometry (LC-MS) analyses were performed using a Shimadzu Prominence ultra-fast liquid chromatography (UFLC) system (Shimadzu, Japan) coupled with a 4000 QTRAP hybrid triple quadrupole-linear ion trap mass spectrometer (QqLIT-MS) equipped with a Turbo Ion Spray source (Applied Biosystems-Sciex, U.S.A.). Chromatographic separation was achieved with an Alltima C18
Table 1 e Main characteristics of the two ROC samples used in the experiments.
ROC-1 ROC-2
DOC, mg L1
SUVA254, L mg1 m1
Conductivity, mS cm1
pH
[Fe2þ], mg L1
[Mn2þ], mg L1
[Cl], g L1
[SO2 4 ], mg L1
57.1 57.2
1.6 2.3
4.25 3.97
7.5 7.7
0.22 0.35
227 234
1.5 1.2
241.5 238.7
1582
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 7 9 e1 5 8 6
Column (250 4.6 mm, particle size 5 mm) run at 40 C, supplied by Alltech Associates Inc (U.S.A.). The multi-residue method used is described in the Text S2, and Tables S3 and S4. The recoveries and method quantification limits (MQLs) are summarized in Table S2.
2.5.
Microtox bioassays
For the V. fischeri bioassays, ROC-1 spiked with the mixture of target analytes was electrochemically oxidized in continuous mode with a step-wise increase in current density (i.e., J ¼ 50, 100, 150, 200 and 250 A m2), and in batch mode at J ¼ 250 A m2. Sampling times and sample pre-treatment were identical to the ones described in the sections Experimental setup and Analytical methods, respectively. The results of the acute toxicity tests are expressed in baseline-toxic equivalent concentration (TEQ) units, derived from a baseline toxicity quantitative structureeactivity relationship (QSAR) model using a virtual compound with octanol-water partition coefficient (log KOW) of 3 and molecular weight (MW) of 300 g mol1 as a reference, which equates to an EC50 of 12 mg L1 (Escher et al., 2008).
3.
Results and discussion
3.1. Electrochemical oxidation - overall organics removal in batch and continuous mode Fig. 1 illustrates the observed removal of DOC and SUVA254, and measured free and total chlorine during continuous oxidation of ROC-1 and ROC-2. There was no DOC removal for
100
0.8
80
0.6
60
0.4
40
0.2
20
0.0
0.0
2.0
100
1.8
0.6
60
0.4
40
0.2
20
0.0
1.6 SUVA254
80
Free and total Cl , mg L
0.8
DOC/DOC0
b 1.0 ROC/ROC
Free and total Cl , mg L
ROC/ROC
a 1.0
low specific electrical charges (Qsp, expressed as A h m3 ROC) applied (Qsp 307.7 and 153.8 A h m3 for ROC-1 and ROC-2, respectively), while SUVA254 was relatively constant at Qsp 153.8 A h m3 for both ROCs tested. By increasing the applied charge, a gradual decrease in DOC was observed, reaching 8.9 1.4 and 26.5 2.3% removal after 769.2 A h m3 supplied to ROC-1 and ROC-2, respectively. Considering that the initial chloride ion concentrations as well as residual chlorine concentrations measured for the ROC-1 and ROC-2 were similar, more efficient mineralization in the latter case is possibly a consequence of the higher initial specific aromaticity of ROC-2 as expressed by its SUVA254 value (Table 1). The enhanced formation of putative electron shuttles (e.g. porphyrins, quinones) from the aromatic fraction could be responsible for the higher NPOC removal observed for ROC-2, as these species accelerate the electron transfer between the organic matter and oxidants (Nurmi and Tratnyek, 2002). The removal of SUVA254 was also enhanced with increasing the supplied charge, and 28.7 1.8% and 42 2.4% removal was observed after applying 769.2 A h m3 to ROC-1 and ROC-2, respectively. The DOC removal achieved in batch oxidation of ROC-1 was 30.7 1.8% after passing 1.45 kA h m3 (Fig. 2). When comparing the oxidation of ROC-1 in the two operational modes tested, the same values of Qsp (w770 Ah m3) and J (250 A m2) in batch and continuous and oxidation of ROC-1 rendered DOC removal of w27 and 8.9 1.4%, and SUVA254 was 1 and 0.71 (i.e. 28.7 1.8% of SUVA254 removal), respectively (Figs. 1 and 2). Thus, the removal of organic carbon depended on the accumulation of long lived oxidants (e.g. HClO/ClO, HOBr and H2O2) in the bulk liquid. For the final DOC removal achieved in continuous (8.9 1.4%) and batch mode (30.7 1.8%) oxidation of ROC-1 at 250 A m2, energy consumption calculated according to Bolton et al. (2001) was 0.704 kWh g1 DOC and 0.350 kWh g1 DOC, respectively. Thus, more efficient removal of organic matter is achieved in batch mode, and the energy input per mass unit of DOC removed is lower compared to the continuously operated reactor. Nevertheless, lower throughput of batch mode compared to a continuous mode may increase significantly the total cost of the treatment.
1.4 1.2 1.0
0.0 0
1 (3.1)
10 (30.8)
30 (92.3)
50 100 150 (153.8) (307.7) (461.5)
200 (615.4)
250 (769.2)
Current density, A m (Specific electrical charge supplied, A h m )
Fig. 1 e Removal of DOC, SUVA254 and generation of free and total chlorine during anodic oxidation in continuous mode of: a) ROC-1 and b) ROC-2 spiked with the trace organic contaminants, vs J (in A mL2) and Q (in A h m-3). ADOC/DOC0, -SUVA254/(SUVA254)0, e e e free chlorine, ————total chlorine.
0.8 0.6
0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
1.6
Q, kA h m-3
Fig. 2 e DOC removal and SUVA254 vs Q in anodic oxidation in batch mode at 574 J [ 250 A mL2 of ROC-1 spiked with the SUVA254, DOC/ trace organic contaminants. DOC0.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 7 9 e1 5 8 6
The decrease in SUVA254 and the limited DOC removal observed in the continuous experiments suggested that organic intermediates were mainly generated by oxidative cleavage and opening of the aromatic moieties. On the other hand, introduction of auxochrome substituents (e.g. eCl, NH2Cl, -Br, -OH) into the aromatic rings during batch oxidation led to an initial increase in SUVA254, as DOC decay was faster than the decrease in UV254 absorbance. Much higher share of active volume (i.e. ROC oxidized inside the anodic compartment) in a continuous reactor (VACT/VTOT ¼ 0.51) compared to the batch reactor (VACT/VTOT ¼ 0.011) led to enhanced reaction of organic matter with short-lived radical oxygen species (ROS) and/or radical halogen species (RHS), which are more capable of ring-opening than more stable oxidants (e.g. FAC, O2, H2O2). However, only partial DOC removal was achieved in both continuous and batch oxidation, indicating the accumulation of oxidation intermediates.
3.2. Removal of trace organic contaminants in continuous mode Fig. 3 illustrates the removals of trace organic compounds observed in the continuous experiments conducted at higher current densities (J ¼ 100e250 A m2). The removals obtained at lower currents (J ¼ 1e50 A m2) are represented in Figure S2. The term “removal” is used here for the conversion of a target analyte to compounds other than the parent compound. Since at low current densities (J ¼ 1, 10 A m2) the likely mechanism for oxidation of pollutants is direct electrolysis, the rate of direct oxidation depends on the adsorption properties of the anode surface, and concentration and nature of trace organic
1583
compounds and their degradation intermediates (Panizza, 2010). Except for sertraline, removed at 70% efficiency, for most of the target analytes no removal or very low removal (30%) was observed under these conditions (Figure S2). Considering the adsorption properties of RuO2/IrO2 anodes, very high removal of sertraline even at J ¼ 1 A m2 can be explained by its hydrophobic nature, as sertraline has the highest log KOW value (5.29) among the selected analytes. The increase in applied charge to 92.3 and 153.8 A h m3 (i.e. J ¼ 30 and 50 A m2) exerted a notable effect on the removal of acetaminophen (40% and 90%), diclofenac (40% and 88%), sulfadiazine (44% and 88%), diazinon (40% and 70%) and norfloxacin (65% and 90%), and led to a complete removal of ranitidine (100%) and lincomycin (>90% and 100%, respectively). Bergmann and Koparal (2005a) reported an increase in RuO2/ IrO2 electrode activity anode potentials (EAN) higher than 1.3 V, related to the production of FAC. In accordance with this study, already at J ¼ 30 A m2 (i.e. EAN ¼ 1.32 0.05V, Table S5) the combined chlorine concentration of 3.0 mg L1, calculated as the difference between the total and free chlorine measured (Fig. 1) implied that chlorine had already reacted to form organic and inorganic chloramines. Considering the affinity of the abovementioned compounds towards chlorine (Bedner and MacCrehan, 2006a; Westerhoff et al., 2005; Acero et al., 2008; Dodd et al., 2005), it can be assumed that they were oxidized by free chlorine. However, under the same conditions, other compounds known to have a high affinity for FAC (Bedner and MacCrehan, 2006a, 2006b; Westerhoff et al., 2005; Acero et al., 2008; Chen and Young, 2008; Chamberlain and Adams, 2006; Gould and Richards, 1984) had low removals. Examples of these are trimethoprim (25% and 46%), gemfibrozil (11% and
Fig. 3 e Removals of target analytes during anodic oxidation in continuous mode of: a) ROC-1 and b) ROC-2 spiked with the trace organic contaminants, for the tested J in the range 100e250 A mL2. RNT-ranitidine, LNC lincomycin, ACTPacetaminophen, DCF-diclofenac, DZN diazinon, ENR-enrofloxacin, SDZ-sulfadiazine, TMP-trimethoprim, NFL-norfloxacin, SRL-sertraline, GMF-gemfibrozil, CTP-citalopram, VNF-venlafaxine, MTP-metoprolol, DIU-diuron, HCThydrochlorothiazide, CAFF-caffeine, ROX-roxithromycin, TML-tramadol, CBZ carbamazepine, ATZ-atrazine, METmetolachlor, IBU-ibuprofen, PNT-phenytoin, IPM-iopromide, 2,4-D-2,4-dichlorophenoxyacetic acid, TPR triclopyr. Values are expressed as mean with their SDs.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 7 9 e1 5 8 6
43%), caffeine (14% and 26%), diuron (14% and 33%), metoprolol (10% and 30%) at J ¼ 30 and 50 A m2, respectively. Given that their initial concentrations were very similar, it appears unlikely that this would be a due to lower mass transfer coefficients. Moreover, at J ¼ 100 A m2 (EAN ¼ 1.70 0.2 V) only 60% of caffeine was oxidized in both ROC-1 and ROC-2, despite its reported high reactivity with FAC in the pH range found in our experiments (e.g. 6.17 at 100 A m2,Table S5) (Gould and Richards, 1984). At the same current density of 100 A m2 (Qsp ¼ 307.7 A h m3), a complete disappearance of enrofloxacin was noted. Interestingly, this compound is known to react very slowly with FAC and to be recalcitrant towards NH2Cl (Dodd et al., 2005). A complete disappearance of carbamazepine from the ROC oxidized at J ¼ 150 A m2 was probably achieved by oxidants other than chlorine, since carbamazepine is recalcitrant even towards the much stronger oxidant ClO2 (Huber et al., 2005). The largely incomplete oxidation by FAC for some of the compounds and unexpectedly rapid disappearance of others known to react slowly with chlorine suggest that other oxidants in the bulk and/or surface reactions may play an important role. Besides reacting in the bulk, chlorine can react electrochemically at the anode forming adsorbed chloro- and oxychloro-radicals, which mediate the degradation of adsorbed organics (Bergmann and Koparal, 2005a; Papastefanakis et al., 2010). Moreover, electrogenerated O2 can indirectly oxidize the bulk organics and form organic radicals via the hydrogen abstraction mechanism (Carlesi Jara et al., 2007). Organic radicals can then further react with O2 to form organic hydroperoxides (ROOH) that are short-lived and tend to decompose, often leading to the formation of molecules with a lower number of carbon atoms. Furthermore, in the presence of Fe2þ and Mn2þ ions in ROC (Table 1) and electrogenerated H2O2, contribution of Fenton reaction to bulk oxidation can be expected. Compounds that were more recalcitrant during electrochemical oxidation were characterized by either the absence of nucleophilic substituents that have an activating effect on the aromatic ring (e.g. ibuprofen, phenytoin, metolachlor, N,N-diethyl-meta-toluamide (DEET)), or by the decreased electron density on the aromatic ring due to the presence of electrophilic halogen groups (e.g. 2,4-dichlorophenoxyacetic acid (2,4-D), atrazine, triclopyr and iopromide). Further increasing the applied current density up to 250 A m2 (Qsp ¼ 769.2 A h m3) led to an enhanced oxidation of atrazine (92%), metolachlor (57% and 84%), ibuprofen (46% and 67%) and phenytoin (61% and 67%) in the ROC-1 and ROC-2, respectively. Considering the EAN (i.e. 2.5 0.07 V) and the measured pH (i.e. pH 2.6, Table S5), oxidants such as ClO2, HClO/OCl, H2O2 and other radical ROS (e.g. O2 , OH, HO2) --and RHS (Br , Br2 , Cl , Cl2 ) probably had a greater participation in the indirect oxidation at this current density. Furthermore, at acidic pH (i.e. pH3) chlorination is acidcatalysed through a mechanism involving H2OClþ (Rebenne et al., 1996). This phenomenon could be responsible for the sharp increase in atrazine removal when the current density increased from 150 to 200 A m2, as in the latter conditions the pH rapidly dropped from 5.13 to 2.91 (Table S5). In a previous study Malpass et al. (2006) demonstrated a dependency of atrazine electrochemical oxidation on the presence of FAC. Moreover, the pH will influence the active surface sites of
MMO anodes as well as their redox properties, accelerating the direct electron transfer reactions (Rossi et al., 2009). On the other hand, only around 30e40% of 2,4-D, triclopyr and iopromide removal was observed under these conditions, while DEET could not be oxidized.
3.3. mode
Removal of trace organic contaminants in batch
To further investigate the effect of a prolonged exposure of trace organic contaminants in ROC to the generated oxidants, batch experiments with external circulation were performed at J ¼ 250 A m2. Triclopyr, 2,4-D, ibuprofen, iopromide, metolachlor, phenytoin, atrazine, and DEET were also in the batch tests more persistent than the other analysed compounds (Fig. 4 and Table S6). The oxidation of all target compounds was more efficient in batch mode due to the accumulation of oxidants in the bulk liquid, and consequently more intense indirect oxidation (Figure S3). With triclopyr as exception, all persistent trace compounds were completely oxidized in batch mode after applying 1.45 kA h m3. Therefore, these compounds require considerably higher electrical charge supplied per anolyte unit volume and longer residence times in order for indirect oxidation to occur.
3.4.
Bioluminescence inhibition in Vibrio fischeri
To verify the effect of electrochemical oxidation on the toxicity of ROC a number of bioluminescence assays were performed using V. fischeri. Although for specifically acting compounds the baseline toxicity is generally marginal, in a mixture of a large number of trace organic contaminants and other chemicals with a variety of specific modes of action, the baseline toxicity could dominate the overall mixture effect (Escher and Schwarzenbach, 2002). Thus, baseline toxicity provides an integrative measure of the combination of chemicals that act together in concert. The contribution of each chemical is weighted only by its hydrophobicity. In order to follow the change of baseline toxicity of the target compounds, the bioassays were performed using the samples
100 80 Removal, %
1584
60 40 20 0
0.0
0.2
0.4
0.6
0.8 1.0 Q, kA h m-3
1.2
1.4
1.6
Fig. 4 e Removal of persistent target analytes vs Q in anodic oxidation in batch mode at J [ 250 A mL2, of ROC-1 spiked TPR, with the trace organic contaminants. 2,4-D, IBU, IPM, MET, PNT, ATZ, DEET.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 7 9 e1 5 8 6
enriched by the same SPE protocol as the one previously described in the section Analytical methods. The majority of matrix components such as salts, particulates, and generated oxidants were removed by the SPE sample pre-treatment. While the sample extracts of the untreated ROC-1 spiked with target analytes exhibited a TEQ value of 3.7 0.02 mg L1, the toxicity of sample extracts of ROC-1 oxidized in continuous mode drastically increased with increased current density (209.1 14.4 mg L1 at J ¼ 250 A m2) (Fig. 5a). Likewise, the toxic response of V. fischeri increased with higher applied charge for the sample extracts of ROC-1 oxidized in batch mode (Fig. 5b). The TEQ value was increased from the initial 4.3 0.1 mg L1 to 151.0 4.9 mg L1 for ROC-1 oxidized in the batch reactor after applying 1.45 kA h m3. Although the relative increase in aromaticity (i.e. SUVA254) in batch oxidation suggested a higher accumulation of substituted aromatic intermediates than in the continuous mode, toxic oxidation products were formed in both reactors. It is important to note that the bioassay with V. fischeri cannot distinguish between the effects of the SPE enriched trace organic pollutants and the co-extracted organic matter. The hydrophobic (i.e. peptides and protein fragments) and/or aromatic fraction of the organic matter (fulvic- and humic-like substances) contributed to the
Toxic equivalent concentration (TEQ), mg L-1
a
250 200 150 100 50 0
Toxic equivalent concentration (TEQ), mg L-1
b
0
50 (153.8)
100 (307.7)
150 (461.5)
200 (615.4)
250 (769.2)
J, A m-2 (Q, A h m-3) 180
1585
observed increase in the toxicity, as this fraction was probably well retained on the SPE cartridge. Nevertheless, TEQ values determined for the ROC oxidized in batch and continuous mode are significantly higher than the values that could be expected owing to any potential by-products of the selected micropollutants.
4.
Conclusions
Electrochemical oxidation at various current densities was investigated for the treatment of a reverse osmosis concentrate spiked with a mixture of pharmaceuticals and pesticides. The removal of DOC depended on the accumulation of oxidants in the bulk liquid. Based on the changes in specific aromaticity, expressed as SUVA254, it appears that the formation of chloro-, bromo- and hydroxyl-substituted aromatic intermediates was enhanced due to the prolonged indirect oxidation. Furthermore, the employed Ti/Ru0.7Ir0.3O2 electrode was capable of oxidizing most of the selected pharmaceuticals and pesticides, while the most persistent compounds had electrophilic substituents at the aromatic ring (2,4-D, atrazine, triclopyr, iopromide), or an aromatic ring insufficiently activated towards nucleophilic attack (ibuprofen, phenytoin, metolachlor, DEET). Nevertheless, the results of the bioluminescence bioassay imply the formation of toxic by-products during both continuous and batch electrochemical oxidation of ROC. Although the contribution of by-products of the investigated pharmaceuticals and pesticides, and compounds formed by oxidation of other organic matter (e.g. fulvic- and humic-like substances) to the measured toxicity is uncertain, participation of oxidants such as active chlorine and bromine in indirect oxidation will likely cause transformation of organic compounds to their halogenated derivatives. This may represent an insurmountable barrier to the environmental applications of RuO2/IrO2-coated Ti electrodes as such byproducts could be persistent and/or require extreme treatment conditions (i.e. very high electrical charges applied). We are currently investigating downstream treatment options.
160 140 120
Acknowledgements
100 80 60 40 20 0
0
121.6
246.4
437.0
1455.7
Q, A h m-
Fig. 5 e Bioluminiscence inhibition of ROC-1 spiked with target contaminants on Vibrio fischeri at 30 min expressed as baseline e Toxic Equivalent Concentration (TEQ) in mg L-1 in: a) continuous experiment conducted at J [ 50, 100, 150, 200 and 250 A mL2, b) batch experiment conducted at J [ 250 A mL2. Each replicate of the sample was tested in duplicates, at eight different concentrations. Results are expressed as average of duplicates ± standard deviation.
This research was supported by the Australian Research Council (grants LP0989159 and DP0985317), Veolia Water Australia, Water Secure, Magneto Special Anodes and The Urban Water Security Research Alliance. The authors would like to thank to Dr Miroslava Macova and Prof Beate Escher from National Research Centre for Environmental Toxicology, The University of Queensland, for performing the Microtox bioassays. We would also like to thank Mr Pieter Hack (Magneto Special Anodes) and Mr Yvan Poussade (Veolia Water Australia) for valuable comments.
Appendix. Supplementary data Supplementary data related to this article can be found online at: doi:10.1016/j.watres.2010.11.035.
1586
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 7 9 e1 5 8 6
references
Acero, J.L., Benitez, F.J., Real, F.J., Gonzalez, M., 2008. Chlorination of organophosphorus pesticides in natural waters. J. Hazard. Mater 153, 320e328. APHA 409E, 1975. Standard Methods for the Examination of Water and Wastewater. American Water Works Association (AWWA), Washington, DC. APHA 5310B, 1998. Standard Methods for the Examination of Water and Wastewater. American Water Works Association (AWWA), Washington, DC. Bedner, M., MacCrehan, W.A., 2006a. Transformation of acetaminophen by chlorination produces the toxicants 1,4benzoquinone and N-acetyl-p-benzoquinone imine. Environ. Sci. Technol. 40, 516e522. Bedner, M., MacCrehan, W.A., 2006b. Reactions of the aminecontaining drugs fluoxetine and metoprolol during chlorination and dechlorination processes used in wastewater treatment. Chemosphere 65, 2130e2137. Bellona, C., Drewes, J.E., 2007. Viability of a low-pressure nanofilter in treating recycled water for water reuse applications: a pilot-scale study. Water Res. 41, 3948e3958. Bergmann, M.E.H., Koparal, A.S., 2005a. Studies on electrochemical disinfectant production using anodes containing RuO2. J. Appl. Electrochem. 35, 1321e1329. Bolton, J.R., Bircher, K.G., Tumas, W., Tolman, C.A., 2001. Figuresof-merit for the technical development and application of advanced oxidation technologies for both electric- and solardriven systems. Pure Appl. Chem. 73, 627e637. Carlesi Jara, C., Fino, D., Specchia, V., Saracco, G., Spinelli, P., 2007. Electrochemical removal of antibiotics from wastewaters. Appl. Catal. B. Environ. 70, 479e487. Chamberlain, E., Adams, C., 2006. Oxidation of sulfonamides, macrolides, and carbadox with free chlorine and monochloramine. Water Res. 40, 2517e2526. Chen, W.H., Young, T.M., 2008. NDMA formation during chlorination and chloramination of aqueous diuron solutions. Environ. Sci. Technol. 42, 1072e1077. Dialynas, E., Mantzavinos, D., Diamadopoulos, E., 2008. Advanced treatment of the reverse osmosis concentrate produced during reclamation of municipal wastewater. Water Res. 42, 4603e4608. Dodd, M.C., Shah, A.D., von Gunten, U., Huang, C.H., 2005. Interactions of fluoroquinolone antibacterial agents with aqueous chlorine: reaction kinetics, mechanisms, and transformation pathways. Environ. Sci. Technol. 39, 7065e7076. Escher, B.I., Schwarzenbach, R.P., 2002. Mechanistic studies on baseline toxicity and uncoupling of organic compounds as a basis for modeling effective membrane concentrations in aquatic organisms. Aquat. Sci. 64, 20e35. Escher, B.I., Bramaz, N., Mueller, J.F., Quayle, P., Rutishauser, S., Vermeirssen, E.L.M., 2008. Toxic equivalent concentrations (TEQs) for baseline toxicity and specific modes of action as a tool to improve interpretation of ecotoxicity testing of environmental samples. J. Environ. Monit. 10, 612e621. Gallard, H., Leclercq, A., Croue, J.P., 2004. Chlorination of bisphenol A: kinetics and by-products formation. Chemosphere 56, 465e473.
Gould, J.P., Richards, J.T., 1984. The kinetics and products of the chlorination of caffeine in aqueous solution. Water Res. 18, 1001e1009. Huber, M.M., Korhonen, S., Ternes, T.A., von Gunten, U., 2005. Oxidation of pharmaceuticals during water treatment with chlorine dioxide. Water Res. 39, 3607e3617. Malpass, G.R., Miwa, D.W., Machado, S.A., Olivi, P., Motheo, A.J., 2006. Oxidation of the pesticide atrazine at DSA electrodes. J. Hazard. Mater 137, 565e572. Nurmi, J.T., Tratnyek, P.G., 2002. Electrochemical properties of natural organic matter (NOM), fractions of NOM, and model biogeochemical electron shuttles. Environ. Sci. Technol. 36, 617e624. Panizza, M., 2010. In: Comninellis, C., Guohua, C. (Eds.), Importance of Electrode Material in the Electrochemical Treatment of Wastewater Containing Organic Pollutants in Electrochemistry for the Environment. Springer New York, New York. Papastefanakis, N., Mantzavinos, D., Katsaounis, A., 2010. DSA electrochemical treatment of olive mill wastewater on Ti/ RuO2 anode. J. Appl. Electrochem. 40, 729e737. Perez, G., Fernandez-Alba, A.R., Urtiaga, A.M., Ortiz, I., 2010. Electro-oxidation of reverse osmosis concentrates generated in tertiary treatment. Water Res. 44, 2763e2772. Radjenovic, J., Petrovic, M., Ventura, F., Barcelo, D., 2008. Rejection of pharmaceuticals in nanofiltration and reverse osmosis membrane drinking water treatment. Water Res. 42, 3601e3610. Rajkumar, D., Kim, J.G., Palanivelu, K., 2005. Indirect electrochemical oxidation of phenol in the presence of chloride for wastewater treatment. Chem. Eng. Technol. 28, 98e105. Rebenne, L.M., Gonzalez, A.G., Olson, T.M., 1996. Aqueous chlorination kinetics and mechanism of substituted dihydroxybenzenes. Environ. Sci. Technol. 30, 2235e2242. Rossi, A., Alves, V.A., Da Silva, L.A., Oliveira, M.A., Assis, D.O.S., Santos, F.A., De Miranda, R.R.S., 2009. Electrooxidation and inhibition of the antibacterial activity of oxytetracycline hydrochloride using a RuO2 electrode. J. Appl. Electrochem. 39, 329e337. Snyder, S.A., Adham, S., Redding, A.M., Cannon, F.S., DeCarolis, J., Oppenheimer, J., Wert, E.C., Yoon, Y., 2007. Role of membranes and activated carbon in the removal of endocrine disruptors and pharmaceuticals. Desalination 202, 156e181. Van Hege, K., Verhaege, M., Verstraete, W., 2002. Indirect electrochemical oxidation of reverse osmosis membrane concentrates at boron-doped diamond electrodes. Electrochem. Commun. 4, 296e300. Westerhoff, P., Moon, H., Minakata, D., Crittenden, J., 2009. Oxidation of organics in retentates from reverse osmosis wastewater reuse facilities. Water Res. 43, 3992e3998. Westerhoff, P., Yoon, Y., Snyder, S., Wert, E., 2005. Fate of endocrine-disruptor, pharmaceutical, and personal care product chemicals during simulated drinking water treatment processes. Environ. Sci. Technol. 39, 6649e6663. Wulfeck-Kleier, K.A., Ybarra, M.D., Speth, T.F., Magnuson, M.L., 2010. Factors affecting atrazine concentration and quantitative determination in chlorinated water. J. Chromatogr. A. 1217, 676e682.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 8 7 e1 5 9 6
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Transformation of the antiepileptic drug oxcarbazepine upon different water disinfection processes Zhi Li a,b, He´le`ne Fenet a, Elena Gomez a, Serge Chiron b,* a b
UMR 5569 ‘Hydrosciences Montpellier’ University of Montpellier I, 15 Avenue Ch. Flahault, BP 14491, 34093 Montpellier cedex 5, France Laboratoire Chimie Provence, Aix-Marseille Universite´s-CNRS (UMR 6264), 3 place Victor Hugo, 13331 Marseille cedex 3, France
article info
abstract
Article history:
Transformation of the pharmaceutical oxcarbazepine (OXC), a keto analogue of carbama-
Received 16 July 2010
zepine (CBZ) was investigated under different water disinfection processes (ozonation,
Received in revised form
chlorination and UV irradiation) to compare its persistence, toxicity and degradation
18 November 2010
pathways with those of CBZ. Analysis by LCeion trapeMSn allowed for the identification of
Accepted 25 November 2010
up to thirteen transformation products (TPs). The major abundant and persistent TPs
Available online 3 December 2010
(10,11-dihydro-10,11-trans-dihydroxy-carbamazepine (DiOH-CBZ), acridine (ACIN) and 1-(2benzaldehyde)-(1H, 3H)-quinazoline-2,4-dione (BQD)) were identical to those previously
Keywords:
reported during water treatment of CBZ. Only one new compound arising from an intra-
Oxidation processes
molecular cyclisation reaction was identified during UV irradiation. OXC reacted quickly
Ozone
with hydroxyl radical and relatively rapidly with free chlorine while slow reaction rates
Chlorine
were recorded in presence of ozone and upon UV irradiation. An increase of the acute
UV irradiation
toxicity of UV irradiated solutions, monitored by a Daphnia magna bioassay, was recorded,
Oxcarbazepine
probably due to the accumulation of ACIN. The formation of ACIN is of concern due to the carcinogenic properties of this chemical. ACIN was also generated during the direct UV photo transformation of DiOH-CBZ and 10-hydroxy-10,11-dihydro-carbamazepine (OHCBZ), two metabolites of OXC and CBZ widely detected in water resources. Analysis of tap water samples revealed the occurrence at ng/L levels of the major TPs detected under laboratory scale experiments, except ACIN. ª 2010 Elsevier Ltd. All rights reserved.
1.
Introduction
Contamination of water resources by micro-contaminant residues is one of the major current challenges for the preservation and sustainability of the environment. In this field, the focus for water pollution research has recently been extended from priority contaminants such as pesticides to the so-called emerging contaminants. An important group of these emerging contaminants are the pharmaceutical products (PPs) since many industralized countries have discovered drug products in their water resources often used for drinking water purposes. Detection of PPs in drinking water has been
up to date quite scarce (Monpelat et al., 2009; Benotti et al., 2009). From water resource to drinking water network, the removal efficiency of PPs in drinking water treatment plants (DWTPs) is rather well established (Vieno et al., 2007) reaching for instance, 98% for acetaminophen and 85% for carbamazepine in a classical DWTP including clarification, filtration and disinfection by chlorination (Stackelberg et al., 2007). However, there is currently a lot of concern regarding the fate of these compounds during chemical disinfection processes. Chlorine is the most widely used disinfectant and as a selective oxidant only allows for the removal of a limited number of PPs from water matrices (Acero et al., 2010). However, many
* Corresponding author. Tel.: þ33 4 91 10 85 25; fax: þ33 4 91 10 63 77. E-mail address:
[email protected] (S. Chiron). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.11.038
1588
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 8 7 e1 5 9 6
drinking water facilities are changing their primary disinfectant from chlorine to alternative disinfectants (e.g., ozone, ultraviolet (UV) irradiation, chlorine dioxide, and chloramine) because they generally reduce regulated trihalomethane and haloacetic acid formation levels. The use of UV irradiation, in addition to pre- or post-chlorination is gaining interest due its effectiveness for inactivation of Cryptosporidium. At the UV-C (254 nm) drinking water fluence (dose) of 400 J m2, the degree of selected pharmaceutical elimination at pH 7 strongly depends on their chemical structure and ranges for instance from 0.4% for 17-b-ethinylestradiol to 15% for sulfamethoxazole (Canonica et al., 2008), but indicates that photo transformation should be seriously taken into account when evaluating the possibility of formation of transformation products (TPs). The application of ozone in drinking water treatment is also widespread affording higher oxidation rates of micro-contaminants than free chlorine does (Westerhoff et al., 2005). Recently, it has been shown that the oxidation of selected PPs by ozone can lead to an effective transformation of many drugs during drinking (Brose´us et al., 2009) and wastewater (West et al., 2009) treatment. Upon disinfection processes, PPs can be therefore transformed into potentially toxic TPs but TPs identification has been limited to a few cases. A wide variety of (multi)-chlorinated and hydroxylated products may be formed and can be expected for conditions typical of wastewater and drinking water chlorination (Dodd and Huang, 2007). For instance, halogenated derivatives of salicylic acid occur in drinking water after chlorination (Quintana et al., 2010) while the chlorination of paracetamol yields the N-acetyl-p-benzoquinone imine and 1,4-benzoquinone toxicants (Bedner and MacCrehan, 2006). Aldehyde moieties which are known to interact with DNA are generated from the ozonation of b-blockers (Benner and Ternes, 2009) and roxithromycin TPs generated under ozonation still possess an unmodified desosamine group and might preserve some biological activity (Radjenovic et al., 2009). The carcinogen acridine is also formed during the UV irradiation of the antiepileptic drug carbamazepine (CBZ) (Vogna et al., 2004). Consequently, to assess the overall efficiency of the elimination of PPs during water treatment, the formation of TPs, their persistence, and relative toxicity compared to the parent should be considered. This will be the main contribution of this work by taking oxcarbazepine (OXC) as a probe compound. OXC is the keto analogue of CBZ and is a more recently marketed antiepileptic drug which has chemical and therapeutic similarities to CBZ but with reduced side effects. Consequently, CBZ is frequently substituted by OXC. OXC is mainly metabolized to 10-hydroxy-10,11-dihydroxy-CBZ (OH-CBZ) and this therapeutically active metabolite further yields 10,11-dihydro-10,11-trans-dihydroxy-CBZ (DiOH-CBZ) a common metabolite of CBZ and OXC (Breton et al., 2005). OXC together with its major metabolites have been detected at concentration levels as high as those of CBZ at WWTPs outlet (Leclercq et al., 2009). Consequently, the different aims of this work are the followings: 1) To investigate kinetics and transformation pathways of OXC upon different chemical oxidation processes (ozonation, chlorination and UV irradiation), 2) To compare the fate of OXC with that of CBZ (Kosjek et al., 2009) in water treatment and 3) To screen for the occurrence of major identified TPs in tap water.
2.
Materials and methods
2.1.
Chemicals
Carbamazepine (CBZ), acridine (ACIN) and acridone (ACON) were purchased from SigmaeAldrich (Saint Quentin Fallavier, France). Oxcarbazepine (OXC) was kindly supplied by Novartis Pharma (Basel, Switzerland). 10-hydroxy-10,11-dihydro-carbamazepine (OH-CBZ), 10,11-dihydro-10,11-trans-dihydroxy-carbamazepine (DiOH-CBZ), [2H2]carbamazepine were purchased from Toronto Research Chemicals (Toronto, Canada). 10,11-dihydro-10,11-cisdihydroxy-carbamazepine was easily prepared by osmiumtetroxide oxidation of CBZ as previously reported (Baker et al., 1973). Sodium hypochlorite solution NaOCl (10% available chlorine) was purchased from Fluka. Water used in all experiments and in the preparation of buffers, was purified by a Milli-Q filtration system (Millipore). Acetonitrile and methanol (LC grade) were purchased from Merck (Darmstadt, Germany). 1-(2-benzaldehyde)-(1H, 3H)-quinazoline-2,4-dione (BQD) was synthesized as previously described (McDowell et al., 2005). Briefly, 50 mg L1 (Milli-Q-water) CBZ were ozonated. An approximate ozone dose of 400 mM was added and allowed to react for 20 min. Excess ozone was then purged with helium, and the sample was freeze-dried to yield the ozonation TPs. The freeze-dried sample was then dissolved in approximately 2 mL of acetonitrile:water (50:50, v/v). To isolate BQD from the other oxidation products, fractions were collected from 300 mL injections of the dissolved material after separation on an LCeUV (l ¼ 278 nm) Merck Lichrospher semi-preparative C-18 column (250 mm 10 mm i.d.). The LC mobile phase was isocratic using acetonitrile:water (50:50, v/v) and a flow rate of 1 mL min1. Fractions were combined and the solvent was removed under a gentle nitrogen stream (T ¼ 50 C) to yield a white crystalline residue. Ozonation of 50 mg of CBZ produced up to 10 mg of BQD with this method for use as an analytical standard. 1H and 13C NMR data were compatible with the structure and are consistent with those reported by McDowell et al. (2005).
2.2.
Treatment experiments
All experiments were carried out either in distilled water or in synthetic surface water containing 25 mg L1 as CaCO3 total alkalinity, 4.5 mg (C) L1 dissolved organic carbon (DOC) as humic acids from SigmaeAldrich, 12.5 mg L1 nitrate ions and showing 75% UV transmission. pH adjustment was done with borate buffer (10 mM) for pH value above 8 and phosphate buffer (20 mM) for pH value below 8. The measured pH never varied by more than 0.1 unit during the course of the experiments. At evently spaced time intervals, 1 mL aliquots of the reaction mixture were removed for OXC, TPs and oxidant analysis. All experiments were carried out in triplicate.
2.2.1.
UV irradiation
The experiments were carried out with an immersion-type photoreactor. A low-pressure mercury (LP Hg) lamp Heraeus Noblelight model TNN 15/32 (nominal power 15 W) emitting at 254 nm was employed as UV radiation source. The 185 nm line of the low-pressure Hg-arc was absorbed by the quartz sleeve of the photoreactor. A cooling jacket system made of quartz
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 8 7 e1 5 9 6
and pure water (with negligible light absorption at the wavelength of emitted radiation) maintained a temperature of 25 0.2 C. The whole assembly was mounted on a magnetic stirrer and wrapped with aluminum foil. Initial pH (7e8) was adjusted by dropwise addition of either 0.1 M NaOH or 0.1 M H2SO4. pH was monitored and remained stable during irradiation.
2.2.2.
Chlorination
Stock solutions of chlorine (5e20 mM) were prepared by diluting a commercial solution of sodium hypochlorite (10% active chlorine). Available chlorine in the NaOCl solutions was determined by the iodometric titration method (APHA, 1998). The kinetic experiments were performed by adding an excess of chlorine (30:1 ratio of chlorine to pharmaceutical on a molar basis). The reactions were carried out in glass vials at room temperature (20 2 C) in the presence of 20 mM buffer. Aliquots of the reaction mixture were removed and quenched immediately after sampling by adding 0.1 mL of a fresh sodium sulfite solution (24 mM). The chlorine concentration in surface water experiments was analyzed by the ABTS method (Pinkernell et al., 2000) and remained nearly constant during the experiment course.
2.2.3.
Ozonation
Ozone was produced with a Fischer 502 ozone generator and its stock solutions (w1.5mM) were produced by sparging ozone-containing oxygen through Milli-Q water that was cooled in an ice bath (Bader and Hoigne´, 1981). Stock solutions of ozone were standardized spectrophotometrically based on their molar absorption coefficient 3 ¼ 3000 M1 cm1 at 260 nm (Liu et al., 2001). Experiments were conducted with ozone in excess at pH 8 (phosphate buffer, 50 mM) with or without tertbutanol (t-BuOH, 10 mM) as hydroxyl radical scavengers. The kinetic runs were started by adding an excess of ozone (30:1 ratio of ozone to pharmaceutical on a molar basis). Ozone decay was determined by the indigo/UV method (Bader and Hoigne´, 1981). Results showed that ozone decrease is <5% and it was assumed that the ozone concentration remained constant during the experiment course. The ozone residual was quenched immediately after sampling by adding 0.1 mL of a fresh sodium sulfite solution (24 mM).
2.3.
Determination of kinetic rate constants
The values of the second-order rate constants for the reaction of OXC with chlorine and ozone were calculated from the pseudo-first-rate constants (kobs) by dividing the kobs values by the total concentration of oxidants. In contrast, competition kinetics were used to determine the fast second-order rate constants for the reaction with hydroxyl (OH) radicals according to a methodology described in details elsewhere (Huber et al., 2003). For this purpose, p-chlorobenzoic acid (pCBA, kOH ¼ 5 109 M1 s1) was selected as reference compound. The experiments were carried out at 25 C with Milli-Q water solutions containing equal concentrations of OXC and reference compound (5 mM) and the pH was kept constant at 7 using 5 mM phosphate buffer. Since OXC undergoes slow direct photolysis, OH radicals were generated by photolysis of H2O2. These experiments were carried out
1589
with a 0.5 L cylindrical immersion-type photoreactor (Heraeus TQ 150 Model, radiation path length 2 cm), equipped with a water-cooled, medium-pressure mercury lamp with maximum emission wavelengths at 313, 366, 406, 436, 546, and 578 nm. The reactor is made of Pyrex glass in order to cut off the wavelengths shorter than 290 nm and to minimize direct photolysis of OXC. The whole assembly was mounted on a magnetic stirrer and wrapped with aluminum foil. The k(OH) for OXC was calculated according the following equation: Ln(Pt/P0) ¼ Ln (Rt/R0) (kP/kR), where P0 and Pt represent OXC concentration at the initial and at any reaction time, respectively, Rt and R0 those of the reference compound, and kP, kR are the second-order reaction rate constants of OXC and reference compound with OH radical, respectively.
2.4.
Samples
Five 2 L tap water samples were collected at five different private homes distributed along the city of Marseille (France) within a 2-h time period in April 2010. All samples were collected in silanized amber glass bottles, preserved with sodium azide and previously washed with acetone, methanol and Milli-Q water. Subsequently after sampling, they were filtered through 0.45-mm nitrocellulose filters (Millipore) and stored at -4 C until analysed (within 48 h). Ascorbic acid (0.6 mg/mL) was added to these samples in order to eliminate residual chlorine.
2.5.
Analytical methods
For kinetic studies, the concentration of OXC was followed by means of a VWR Hitachi HPLC coupled with a UV detector set at l ¼ 220 nm and a LiChrospher C-18 column (250 4 mm i.d., Merck). Elution was carried out with a 50:50 mixture of acetonitrile and aqueous H3PO4 (pH 3), at a flow rate of 1.0 mL min1. Injection volume was 50 mL. For TPs and parent compounds analysis in water samples, sample pH was adjusted to a value of 7 and 1 L samples were preconcentrated with SPE on Oasis HLB 200 mg cartridges (Waters). The analytes were eluted with 2 5 mL methanol. The extract was evaporated to dryness and then redissolved in 200 mL of a water/methanol mixture (50:50 v/v). The analysis of the extract was performed by LCeMS/MS in the multiple reaction monitoring mode (MRM) and in positive electrospray (ESI) ionization mode. Only the most abundant product ion in MS/MS mode was chosen because at ultratrace levels (ng/L), the abundance of the second transition ion was too low to be detected. The selected transition ions were 237 > 194, 253 > 236, 255 > 237, 271 > 253, 180 > 152, 196 > 167 for CBZ, OXC, OH-CBZ, DiOH-CBZ, ACIN, ACON, respectively. Moreover, qualitative analysis of compounds VII, VIII and IX (see Fig. 2) were carried out by using 267 > 249, 283 > 265 and 242 > 224 as specific transition ions. The HPLC system consisted of a Metachem C-18 column 150 2 mm i.d., 3 mm particle size (Varian), and an Esquire 6000 ion trap mass spectrometer (Bruker, Bremen, Germany). The mobile phase used in chromatographic separation consisted of a binary mixture of solvents A (acetonitrile) and B (0.1% formic acid solution) at a flow rate of 0.2 mL min1. The gradient was operated from 5% to 100% A for 25 min and then back to the initial conditions in 5 min. Identification of the target analytes
1590
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 8 7 e1 5 9 6
Fig. 1 e Typical Total Ion Chromatograms (TIC) obtained after a) 1 h UV irradiation in synthetic water, b) 2 min ozonation in distilled water without tert-butanol, c) 2 h chlorination of 30 mg LL1 OXC.
in unknown samples was based on the LC retention time compared to that of a standard (30 s) and the unique combination of a precursor-product ion. The percent absolute recovery was 93 5%, 77 6%, 93 5%, 96 5%, 96 5%, and 90 6% for CBZ, OXC, OH-CBZ, DiOH-CBZ, ACIN and ACON, respectively. Limits of quantification (LOQs) ranged from 1 ng L1 to 5 ng L1 depending on the compound. Quantification was carried out by means of an isotope labeled internal
standard (IS) procedure ([2H2]CBZ, transition ions 239 > 196) spiked at 100 ng L1 in the extracts, because this is the only effective way to overcome the matrix ion suppression effects in electrospray LCeMS/MS. Calibration curves were constructed by plotting the ratio peak area/internal standard peak area against concentration levels. This ratio remained constant whether the matrix was present or not (FeitosaFelizzola et al., 2007). For TPs identification, the mass
1591
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 8 7 e1 5 9 6
HO HO
N C O NH2
indirect photolysis or O3 decomposition
HO
.
OH
.
N C O NH2
O
OH
.
+ H
O
.
O3
X
HOCl / ClOHO O
O O
HO
OO O
HO H2O
O
O
Cl
N C O NH2
XII
Cl
Cl
N C O NH2
VI
V
O
OH
OH O O
IX
N H O
OH HO
-H2O2
O
O
N H
N O
HN N
VII
N C O NH2 O
direct photolysis
HN O
OH
N C O NH2
N C O NH2
N H O
N C O NH2
VI
N C O NH2
III
O
- H
HO
XI
IV
photolyis
N C O NH2
II
H
.
N C O NH2
OH
OH
OH
direct
O
I
.
OH
OO
N C O NH2
O VIII
N
OH O
O HN O VIII
N
OH O
HN O
N C O NH2
VII
N O
Fig. 2 e Proposed transformation pathways of OXC.
spectrometer was operated in the full scan mode with the same LC gradient as used for TPs analysis in water samples.
2.6.
Ecotoxicity assays
Acute toxicity immobilization tests on Daphnia magna were carried out according to the OECD Guideline 202. The test was performed using six concentrations of each standard substance (OXC, DiOH-CBZ and BDQ), working in triplicate in the 2.5e10 mg L1 range for pure standards. To investigate the potential toxicity of the TPs, two dilutions (1:3 and 1:6) of the sample corresponding at the end of treatment time (150e180 min) were used, working in triplicate. In addition, blank solutions were added for each inhibition curve.
3.
Results and discussion
3.1. Identification of transformation products and transformation pathways Thirteen TPs of OXC have been detected by LCeMS. TPs molecular weight (MW) was assigned on the basis of the pseudo-molecular ions [M þ H]þ. On the basis of the MS2/MS3 mass fragmentation patterns alone, a definitive assignment of TPs structure was not possible. The TPs structures were tentatively elucidated by coupling MS data with knowledges on the reactivity of investigated oxidants. Confirmatory methods were also used including the use of authentic standards and matching the TPs fragmentation patterns with those reported in published mass spectra. MS data together with Total Ion Chromatograms (TIC) under different conditions are reported
in Table 1 and Fig. 1, respectively. All MS2/MS3 spectra are provided in the Supplementary Material appendix. Proposed degradation pathways of OXC under different degradation conditions is also provided in Fig. 2.
Table 1 e Intermediates and products detected by ESI (D) LCeMS/MS upon OXC chlorination, ozonation, UV irradiation and biodegradation reactions. [M þ H]þ m/z
MS/MS ions m/z
Identification/ Confirmation method
I (OXC)
253
Authentic standard
II (DiOH-CBZ)
271
III IV V (ACIN) VI
210 226 180 269
VII (BQD)
267
VIII
283
IX
242
236a, 208, 210, 180 253a, 236, 210, 180 195, 182a 208a, 198, 182 152 252a, 226, 226, 224, 208, 196 249a, 239, 224, 196 265, 240, 237, 222a, 194 224a, 196
X
269
XI XII XIII (OH-CBZ)
287 321 255
Compound
251a, 224, 208, 195, 180 251 285 237, 220, 194a
a indicates the most abundant fragment ion.
Authentic standard Kosjek et al. (2009) Kosjek et al. (2009) Authentic standard This study Authentic standard This study This study, Lu et al. (2009) This study This study This study Authentic standard
1592
3.1.1.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 8 7 e1 5 9 6
UV irradiation
In distilled water (direct photolysis), none TPs could be detected while in synthetic surface water UV photolysis of OXC (I) is mainly due to indirect photolysis since the addition of a hydroxyl radical scavenger such as 2-propanol (2% in volume) significantly inhibited its transformation. In water, I is in equilibrium with an enol tautomer, which is problaby stabilized by resonance. UV photolysis of I in surface water is mainly due to indirect photolysis since the addition of a hydroxyl radical scavenger such as 2-propanol (2% in volume) significantly inhibited its transformation. Similarly to CBZ (Kosjek et al., 2009), UV photolysis of I proceeds through two main routes. Both are likely initiated by hydroxyl radical attacks on the a-carbon of the keto moiety due to the nucleophilic properties of this peculiar position. The first route involves the formation of a diol II (Fig. 2), which was identified as DiOH-CBZ by comparing its chromatographic and mass spectrometric behavior with that of an authentic standard. II eluted as two peaks because both the trans and cis stereoisomers were formed, the trans stereoisomer being the most abundant. The trans and cis stereoisomers were distinguished on the basis of the retention time of individual standard. The formation of II is somewhat surprising since it entails a reductive step. However, upon OH radical generation, the formation of an alpha-hydroxy ketone radical might be hypothezised (Fig. 2). This latter radical might evolve by a hydrogen radical transfer reaction between two molecules of this transient radical giving simultaneously rise to II and VI. II further underwent a ring contraction process probably upon direct photolysis as previously described in details (Chiron et al., 2006) to yield III, IV and V. V was identified as ACIN due to the availability of an authentic standard, III as (9H,10H)acridine-9-aldehyde and IV as (hydroxy-(9H,10H)-acridine-9aldehyde due to the consistency of MS2 fragmentation patterns of III and IV with previously published data (Kosjek et al., 2009). For compound IV, several structural isomers are probably possible and further investigation is needed to confirm with certainty its structure. ACON was never detected in our experiments indicating a slow phototransformation rate of ACIN into ACON upon UV irradiation. The second route involves the OXC heterocyclic ring oxidation and the formation of a new carbonyl group to yield VI. The MS2 spectrum of VI was characterized by losses of 17 (m/z 252) and 43 (m/z 226) mass units due to losses of NH3 and CONH2, respectively which supports the preservation of CONH2 lateral chain. In contrast, the lack of the product ion at m/z 180 which accounts for the unmodified acridine structure revealed that the heterocyclic ring has been modified. Additional fragment ions at m/z 224 (-CONH3) and at m/z 196 (-COCONH3), might reveal the formation of a new aldehyde moiety. A detailed mechanism of ring closure of VI by an intramolecular reaction to a quinazoline moiety (VII) is provided elsewhere (McDowell et al., 2005). VII was assigned the structure of BQD due to the availability of an authentic standard. VII is further oxidized to VIII. With an increase in MW of 16 mass units with respect to VII and fragmentation patterns similar to those of VII, VIII was suggested to derive from the oxidation of an aldehyde moiety into a carboxyl
one. VII and VIII both eluted as two peaks, exhibiting the same MW and fragmentation patterns. In each case the second compound is likely to be a stereoisomer of the first one resulting from the hindered rotation of the partial double bond character of the phenyl-N bond. IX probably arises from the cleavage of the CONH2 lateral chain (lack of losses of 17 and 43 mass units featuring the CONH2 lateral chain) together with the oxidation of one of the two aldehyde functions of VI into a carboxylic acid function and followed by an intramolecular cyclisation reaction as previously proposed by Hu et al. (2009). Finally, X with a MS2 spectrum depicting fragment ions at m/z 251 (OH loss) and m/z 208 (CONH2 loss) would probably derive from a hydroxyl radical attack on a C6 aromatic ring and was assigned the structure of hydroxycarbamazepine.
3.1.2.
Ozonation
Ozonation involves two different major oxidative species: ozone and OH radical. To monitor the influence of OH radical reactions in the oxidation process of OXC, experiments in the presence and absence of t-BuOH as a OH radical scavenger were performed.
3.1.3.
Direct oxidation by ozone
According to the results from the ozonation of CBZ (McDowell et al., 2005), the two non activated aromatic rings of OXC are not reactive against ozone attacks. The ozone reaction might be entirely due to a reaction with the enol and in presence of tBuOH only VIII was determined as a major TPs of OXC (see Fig. 1 in the Supporting Material). Ozonide could be formed by ozone cyclo-addition on unsaturated bond of the enol (Fig. 2). The cleavage of the unstable ozonide might lead to a ring opening and the formation of a zwitterionic specie. The hydroxy-hydroperoxide could then evolve to compound VIII by hydrogen peroxide release from the hydroxy-hydroperoxide form following a similar ozonolyis mechanism to that suggested for progesterone for instance (Barron et al., 2006).
3.1.4.
Indirect oxidation by OH radicals
Without t-BuOH, when OH radical and ozone molecule can be concomitantly present, II, VII and VIII were identified as the three major TPs of OXC (Fig. 1b) OH radical originating from ozone decay was probably responsible for the generation of II and VII, by following the same mechanism hypothesized for the indirect photochemical transformation of OXC.VIII might arise both from direct ozone attacks and from further oxidation of VII. Surprisingly, IX was not detected in our experimental conditions.
3.1.5.
Chlorination
Under chlorine treatment, the chlorination pathway of OXC was clearly evidenced by the occurrence of the chlorine isotopic pattern in the XI and XII mass spectra, with XI and XII exhibiting one and two chlorine atoms, respectively. Moreover, the MS2 spectrum of XI and XII showed a chlorine loss with an ion at m/z 250 and m/z 285, respectively. At least two isomers seem to be possible for XI and XII since both compounds were detected as two peaks. Ketone chlorination generally results from initial substitution reactions on the
1593
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 8 7 e1 5 9 6
a-carbon to the carbonyl group, inducing the successive replacement of hydrogen by chlorine (Deborde and von Gunten, 2008). However, in case of OXC, chlorine might also react with the C6 aromatic ring by electrophilic substitutions at the Meta position due to the electron-withdrawing properties of the keto group. In contrast, chlorine is not expected to react with amide. Surprisingly, the formation of VII (BQD) was highly preserved under chlorine treatment and VII was even the major TPs of the reaction at the end of the reaction time (180 min, Fig. 1c). This result implies that the haloform reaction with free chlorine probably prevailed. During this chlorination step, hydrolysis of mono- and dichlorinated TPs probably occurred to yield VI and VII as previously reported in case of the chlorination of ketones (Guthrie et al., 1991).
3.2. Kinetics for oxcarbazepine reaction with different oxidants and upon UV irradiation In Fig. 3 are illustrated the concentration evolution profiles of OXC and its major TPs plotted against time, taking into account that chlorination and ozone reactions were batch experiments while in UV irradiation the number of incident UV quanta increased with time. In Table 2 are reported kinetic rate constants for the reaction of OXC and CBZ with oxidants and upon UV irradiation.
3.2.1.
Direct photolysis
Similarly to CBZ (Pereira et al., 2007), OXC reacted very slowly upon UV irradition in distilled water at pH 7.6. The degradation of OXC was still less than 10% after 180 min reaction (results not shown). Despite its relatively high absorbance (decadic molar absorption coefficient 3 ¼ 7245 M1 cm1 at 254 nm), the quantum yield for transformation of OXC, determined from the low measured pseudo-first order constant, is quite low (f ¼ 8 0.8 104 mol E1). This quantum yield value was experimentally determined as the ratio between the pseudo-first order rate constant (kuv) and the specific rate of light absorbed by OXC at 254 nm (Ks (l ¼ 254 nm) ¼ 11.5 E mol1 s1).
3.2.2.
Indirect photolysis
In synthetic surface water, the degradation of OXC increased probably due to an indirect production of OH radical from reaction of the UV light with DOM and NO 3 ions (Fig. 3). A first apparent order reaction constant value of kuv ¼ 9.2 0.8 103 min1 was calculated. VII (BQD) and IX were the two major TPs of the reaction and the persistence of both compounds was observed. At the end of treatment time (180 min), the concentration of VII reached 56% relative to the initial concentration of OXC. The other two identified TPs
Table 2 e A comparison between apparent second-order rate constants for the reaction of OXC and CBZ with selected oxidants in surface water. Compound O
N CONH2
Oxidant
k at pH 8 (M1 s1)
HO O3a Cl2 UVb
8.4 0.24 109 550 45 6.2 0.2 102 9.2 0.8 103
HO
8.8 0.2 109
O3
3.0 0.3 103
Cl2
<0.1
UVb
2.2 1.3 103
Reference This This This This
study study study study
Oxcarbazepine
N CONH2
Fig. 3 e Product evolution against time of five transformation products of OXC: (B) OXC, (;) BDQ, (>) DiOH-CBZ, (-) ACIN, (6) IX (second axis), (A) VIII (second axis) during a) UV irradiation in synthetic water, b) ozonation and c) chlorination of 30 mg LL1 of OXC.
Carbamazepine
a in presence of tert-butanol. b kuv in min1.
Huber et al. (2003) Huber et al. (2003) Lee and von Gunten (2010) Pereira et al. (2007)
1594
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 8 7 e1 5 9 6
Fig. 4 e Transformation product time profiles under chlorination (pH 7, 50 mg LL1 Cl2). (B) monochloro-OXC, (,) dichloro-OXC; (;) BDQ. Results normalised, e.g. 1 is the maximum observed value in the set of experiments.
(II and V) were encountered in smaller quantities but the increase in V was observed at the expense of II, suggesting the transformation of the former into the latter, probably through a direct photolysis process as previously reported (Chiron et al., 2006). At the end of the experiment, V accounted for 12% of the initial concentration of OXC. The formation of V is of concern because of the carcinogenic properties of this chemical. However, as an end-product of the reaction, V is not expected to be generated under UV doses typically required for water disinfection (40e140 mJ/ cm2) because those latter values were reached in less than 5 min in our experimental set-up. In contrast, V might arise from the direct photo transformation of II (DiOH-CBZ) and
XIII (OH-CBZ), two metabolites of OXC and CBZ which have been detected at higher concentrations than their parent compounds at the outlet of WWTPs (Leclercq et al., 2009) and in surface waters (Miao and Metcalfe, 2003). V is the major TPs resulting from the UV irradiation of II and to a lesser extent XIII after 15 min treatment, as shown in Figs. 2 and 3 in the Supplementary Material. In contrast, II and XIII have turned out to be stable in presence of ozone and chlorine. When t-BuOH (10 mM) was added as OH radical scavengers to solutions used for measurement of O3 rate constant, OXC reacted relatively slowly with O3 (ko3 ¼ 550 45 M1 s1) at pH 8 in distilled water. The constant calculated in this work is far lower than the value reported for CBZ in spite of their similar chemical structure but probably due to the lack of the major site of reaction for ozone (e.g., olefinic carbon atoms). Without t-BuOH, ozonation is influenced by OH radical generated through reactions of O3 with specific functional moieties such as phenol and amine (Buffle and von Gunten, 2006) in DOM and from autocatalytic O3 decomposition. In this case, a pseudo-first-order kinetic constant kobs ¼ 7.7 0.1 103 s1 was determined. Accordingly, >99% transformation of OXC was achieved in 30 min (see Fig. 3b). OH radical oxidation prevails over O3 oxidation due to the high second-order rate constant of the reaction of OH radical with OXC (8.4 0.2 109 M1 s1, determined in this study). OXC is nearly quantitatively oxidized into VII (Fig. 3b) which is subsequently transformed into VIII as previously reported by McDowell et al. (2005). At the end of the reaction time (150 min), VII still accounted for 32% of the initial concentration of OXC while II remained stable during the course of the reaction with a concentration accounted for 9% of the initial concentration of OXC. In contrast to CBZ (kCl2 < 0.1 M1 s1, Lee and von Gunten, 2010), OXC reacted with free chlorine at significantly higher reaction rates in comparison to CBZ. As Deborde and von Gunten (2008) pointed out, the oxidation rate for most of the
Fig. 5 e Multiple Reaction Monitoring (MRM) LCeMS/MS chromatogram obtained after preconcentration of 1 L of tap water: II (12 ± 5 ng LL1), IS (100 ng LL1). VII, VIII and IX detected but not quantified with precision (<10 ng LL1).
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 8 7 e1 5 9 6
chlorine reactions with organic compounds can be described by second-order kinetics, first-order in the free active oxidant ([Cl2]t ¼ [HOCl] þ [OCl]) and first-order in the target compound. As expected, the reaction rates increased as the chlorine concentration increased and as the pH decreased from 10 to 6. Second-order rate constants k were found to be 3.8 0.1 101, 8.9 0.1 102, 6.2 0.2 102, 4.6 0.2 102, 4.1 0.3 102 M1 s1 at pH 6, 7, 8, 9 and 10, respectively. A decrease of one order of magnitude in the k values was recorded between pH 6 and 10 probably due to the higher reactivity of HOCl versus OCl. At pH 7.6, OXC was readily transformed into VII which accounted for 81% of the initial concentration of OXC, while chlorinated derivatives were only detected in small amounts (Fig. 3c). However, due to the potential toxicity of chlorinated derivatives, the influence of pH and chlorine concentrations on their quantitative formation was further investigated. The most important factor for their formation was the chlorine concentration. Mono- and dichlorinated TPs of OXC were quickly generated at high chlorine levels (50 mg L1), but their further degradation occurred also quickly and the only remaining TPs at long reaction times was VII (see Fig. 4).
3.3.
Ecotoxicity determined in D. magna tests
Ecotoxicity results did not show any acute toxic effects for OXC, BQD and DiOH-CBZ in the tested concentration range. For the test solutions containing final ozonation and chlorine treatment TPs, neither acute toxicity nor statistically significant growth inhibition was noted. In contrast, UV irradiated solutions of OXC prepared in synthetic surface water exhibited an increase in acute toxicity corresponding to a percentage of immobilization of 25 5% which was probably due to the formation of acridine. This latter chemical has demonstrated a high toxicity against D. magna (Parkhurst et al., 1981).
3.4.
Application to tap water samples
TPs of OXC and CBZ together with the parent compounds were screened for their occurrence in tap water samples by LCeMS/MS in the MRM mode after the SPE of the samples. A chromatogram of a tap water analysis is presented in Fig. 5. The parent compounds were never detected although they have been widely detected in surface and ground waters. Among the metabolites or TPs of OXC and CBZ, DiOH-CBZ was detected in all samples at concentration levels below 15 ng L1. VII, VIII and IX were detected in three out of five samples. Although a precise quantification of these chemicals was not possible either due to the lack of analytical standard (VIII and IX) or because the recoveries were not determined (VII), they may occur at an average concentration below 10 ng L1 in tap water, assuming a similar response for the three compounds in (þ) ESI ionization mode. These occurrence levels are consistent with those reported for other PPs in drinking water (Benotti et al., 2009). The persistence of TPs of OXC in water distribution system illustrates that longer contact with secondary disinfectants do not play a significant role in removing or lowering the concentrations of these refractory compounds. This is consistent with their resistance to chlorine and ozone oxidation as shown in this work.
4.
1595
Conclusions and outlook
Although several previous studies have investigated on the fate of CBZ in water treatment, kinetics and reaction pathways of OXC, a keto homologue of CBZ, upon different water disinfection treatments were largely unknown. The present work provides substantial new details related to these aspects of the OXC reaction with chlorine, ozone and upon UV irradiation. OXC reacted quickly with OH radical and relatively rapidly with free chlorine while slow reaction rates were recorded in presence of either ozone or upon UV irradiation. Product analysis indicated that abundant and persistent TPs of OXC such as DiOH-CBZ, ACIN and BQD, were identical to those previously reported during water treatment of CBZ. An increase in acute toxicity was only observed for irradiated solutions of OXC prepared in synthetic surface water probably due to the accumulation of ACIN. Analysis of tap water samples revealed the occurrence at ng/L levels of several metabolites/transformation products of OXC and CBZ except ACIN.
Acknowledgments This work was funded by the “Agence Franc¸aise de Se´curite´ Sanitaire de l’environnement et du travail (AFSSET), Programme de recherche Environnement-Sante´-Travail 2007.
Appendix A. Supplementary material Supplementary information for this manuscript can be found in the online version at, doi:10.1016/j.watres.2010.11.038.
references
Acero, J., Benitez, F., Real, F., Roldan, G., 2010. Kinetics of aqueous chlorination of some pharmaceuticals and their elimination from water matrices. Water Res. 44, 4158e4170. APHA, 1998. Standard Methods for the Examination of Water and Wastewater. APHA, AWWA, WPCF, Washington, DC. Bader, H., Hoigne´, J., 1981. Determination of ozone in water by the indigo method. Water Res. 15, 449e456. Baker, K.M., Csetenyl, J., Frigerio, A., Morselli, P.L., 1973. 10,11dihydro-10,11-dihydroxy-5H-dibenz[b, z]azepine-5carboxamide, a potential metabolite of carbamazepine isolated from human and rat urine. J. Med. Chem. 16, 703e708. Barron, E., Deborde, M., Rabouan, S., Mazellier, P., Legube, B., 2006. Kinetic and mechanistic investigations of progesterone reaction with ozone. Water Res. 2006 (40), 2181e2189. Bedner, M., MacCrehan, W.A., 2006. Transformation of acetaminophen by chlorination produces the toxicants 1,4benzoquinone and N-acetyl-p-benzoquinone imine. Environ. Sci. Technol. 40, 516e522. Benner, J., Ternes, T.A., 2009. Ozonation of propranolol: formation of oxidation products. Environ. Sci. Technol. 43, 5086e5093.
1596
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 8 7 e1 5 9 6
Benotti, M.J., Trenholm, R.H., Vanderford, B.J., Holady, J.C., Stanford, B.D., Snyder, S.A., 2009. Pharmaceuticals and endocrine disrupting compounds in US drinking water. Environ. Sci. Technol. 43, 597e603. Breton, H., Cociglio, M., Bressolle, F., Peyriere, H., Blayac, J.-P., Hillaire-Buys, D., 2005. Liquid chromatographyeelectrospray mass spectrometry determination of carbamazepine, oxcarbazepine and eight of their metabolites in human plasma. J. Chromatogr. B. 828, 80e90. Brose´us, R., Vincent, S., Aboulfadi, K., Daneshvar, A., Sauve´, S., Barbeau, B., Pre´vost, M., 2009. Ozone oxidation of pharmaceutical, endocrine disruptors and pesticides during drinking water treatment. Water Res. 43, 4707e4717. Buffle, M.O., von Gunten, U., 2006. Phenols and amine induced HO generation during the initial phase of natural water ozonation. Environ. Sci. Technol. 40, 3057e3063. Canonica, S., Meunier, L., von Gunten, U., 2008. Phototransformation of selected pharmaceuticals during UV treatment of drinking water. Water Res. 42, 121e128. Chiron, S., Minero, C., Vione, D., 2006. Photodegradation processes of the antiepileptic drug carbamazepine, relevant to estuarine waters. Environ. Sci. Technol. 40, 5977e5983. Deborde, M., von Gunten, U., 2008. Reactions of chlorine with inorganic and organic compounds during water treatmentdKinetics and mechanisms: a critical review. Water Res. 42, 13e51. Dodd, M., Huang, C.-H., 2007. Aqueous chlorination of the antibacterial agent trimethoprim. Reaction kinetics and pathways. Water Res. 41, 647e655. Feitosa-Felizzola, J., Temime, B., Chiron, S., 2007. Evaluating on line SPE LC/ion trap/MS for reliable quantification and confirmation of several classes of antibiotics in urban wastewaters. J. Chromatogr. A 1164, 95e104. Guthrie, J.P., Cossar, J., Lu, J., 1991. Dihydroxyacids from the chlorination of ketones: an unexpected process. Can. J. Chem. 69, 1904e1908. Hu, L., Martin, H.M., Arce-Bulted, O., Sugihara, M.N., Keating, K.A., Strathmann, T.J., 2009. Oxidation of carbamazepine by Mn(VII) and Fe(VI): reaction kinetics and mechanism. Environ. Sci. Technol. 43, 509e515. Huber, M.M., Canonica, S., Park, G.Y., von Gunten, U., 2003. Oxidation of pharmaceuticals during ozonation and advanced oxidation processes. Environ. Sci. Technol. 37, 1016e1024. Kosjek, T., Andersen, H., Kompare, B., Ledin, A., Heath, E., 2009. Fate of carbamazepine during water treatment. Environ. Sci. Technol. 43, 6256e6261. Leclercq, M., Mathieu, O., Gomez, E., Casellas, C., Fenet, H., Hillaire-Buys, D., 2009. Presence and fate of carbamazepine, oxcarbazepine and seven of their metabolites at wastewater treatment plants. Arch. Environ. Contam. Toxicol. 56, 408e415. Lee, Y., von Gunten, U., 2010. Oxidative transformation of micropollutants during municipal wastewater treatment: comparison of kinetic aspects of selective (chlorine, chlorine dioxide, ferrateVI, and ozone) and non selective oxidants (hydroxyl radicals). Water Res. 44, 555e566.
Liu, Q., Scurter, L.M., Muller, C.E., Aloisio, S., Francisco, J.S., Margerum, D.W., 2001. Kinetics and mechanisms of aqueous ozone reactions with bromide, sulfite, hydrogen sulfite, iodide and nitrite ions. Inorg. Chem. 40, 4436e4442. Lu, L., Martin, H.M., Arce-Bulted, O., Sugihara, M., Keating, K., Strathmann, T., 2009. Oxidation of carbamazepine by Mn(VII) and Fe(VI): reaction kinetics and mechanism. Environ. Sci. Technol. 43, 509e515. McDowell, D., Huber, M., Wagner, M., von Gunten, U., Ternes, T., 2005. Ozonation of carbamazepine in drinking water: identification and kinetic study of major transformation products. Environ. Sci. Technol. 39, 8014e8022. Miao, X.-S., Metcalfe, C.D., 2003. Determination of carbamazepine and its metabolites in aqueous samples using liquidchromatography-electrospray tandem mass spectrometry. Anal. Chem. 75, 3731e3738. Monpelat, S., Le Bot, B., Thomas, O., 2009. Occurrence and fate of pharmaceutical products and by-products, from resource to drinking water. Environ. Intern. 35, 803e814. Parkhurst, B., Bradshaw, A., Forte, J., Wright, G., 1981. The chronic toxicity to Daphnia magna of acridine, a representative azaarene present in synthetic fossil fuel products and wastewaters. Environ. Pollut. Ser. A 24, 21e30. Pereira, V.J., Weinberg, H.S., Linden, K.G., Singer, P.C., 2007. UV degradation of pharmaceutical compounds in surface water via direct and indirect photolysis at 254 nm. Environ. Sci. Technol. 41, 1682e1688. Pinkernell, U., Nowack, B., Gallard, H., von Gunten, U., 2000. Methods for the photometric determination of reactive bromine and chlorine species with ABTS. Water Res. 34, 4343e4350. Quintana, J.B., Rodil, R., Lopez-Mahia, P., Muniategui-Lorenzo, S., Prada-Rodriguez, D., 2010. Investigating the chlorination of acidic pharmaceuticals and by-product formation aided by an experimental design methodology. Water Res. 44, 243e255. Radjenovic, J., Godehardt, M., Petrovic, M., Hein, A., Farre´, M., Jekel, M., Barcelo, D., 2009. Evidencing generation of persistent ozonation products of antibiotics roxithromycin and trimethoprim. Environ. Sci. Technol. 43, 6808e6815. Stackelberg, P.E., Gibs, J., Furlong, E.T., Meyer, M.T., Zaugg, S.D., Lippincott, R.L., 2007. Efficiency of conventional drinking water treatment processes in removal of pharmaceuticals and other organic compounds. Sci. Total Environ. 377, 255e272. Vieno, N., Ha¨rkki, H., Tuhkanen, T., Kronberg, L., 2007. Occurrence of pharmaceuticals in river water and their elimination in a pilot-scale drinking water treatment plant. Environ. Sci. Technol. 41, 5077e5084. Vogna, D., Marotta, R., Andreozzi, R., Napolitano, A., D’Ischia, M., 2004. Kinetic and chemical assessment of the UV/H2O2 treatment of antiepileptic drug carbamazepine. Chemosphere 54, 497e505. West, E., Rosario-Ortiz, F.L., Snyder, S., 2009. Effect of ozone exposure on the oxidation of trace organic contaminants in wastewater. Water Res. 43, 1005e1014. Westerhoff, P., Yoon, Y., Snyder, S., West, E., 2005. Fate of endocrine disrupter, pharmaceutical, and personal care product chemicals during simulated drinking water treatment processes. Environ. Sci. Technol. 39, 6649e6663.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 9 7 e1 6 0 6
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Temperature phased anaerobic digestion increases apparent hydrolysis rate for waste activated sludge Huoqing Ge, Paul D. Jensen, Damien J. Batstone* Advanced Water Management Centre (AWMC), Environmental Biotechnology CRC, The University of Queensland, St Lucia, QLD 4072, Australia
article info
abstract
Article history:
It is well established that waste activated sludge with an extended sludge age is inherently
Received 28 September 2010
slow to degrade with a low extent of degradation. Pre-treatment methods can be used prior
Received in revised form
to anaerobic digestion to improve the efficiency of activated sludge digestion. Among these
26 November 2010
pre-treatment methods, temperature phased anaerobic digestion (TPAD) is one promising
Accepted 28 November 2010
method with a relatively low energy input and capital cost. In this study, an experimental
Available online 4 December 2010
thermophilic (50e70
C)emesophilic system was compared against a control meso-
philicemesophilic system. The thermophilicemesophilic system achieved 41% and 48% Keywords:
volatile solids (VS) destruction during pre-treatment of 60 C and 65 C (or 70 C) respec-
Temperature phased anaerobic
tively, compared to 37% in the mesophilicemesophilic TPAD system. Solubilisation in the
digestion
first stage was enhanced during thermophilic pre-treatment (15% at 50 C and 27% at 60 C,
Thermophilic pre-treatment
65 C and 70 C) over mesophilic pre-treatment (7%) according to a COD balance. This was
Mesophilic pre-treatment
supported by ammoniaenitrogen measurements. Model based analysis indicated that the
Waste activated sludge
mechanism for increased performance was due to an increase in hydrolysis coefficient
Hydrolysis rate
under thermophilic pre-treatment of 60 C (0.5 0.1 d1), 65 C (0.7 0.2 d1) and 70 C (0.8 0.2 d1) over mesophilic pre-treatment (0.2 0.1 d1), and thermophilic pre-treatment at 50 C (0.12 0.06 d1). ª 2010 Elsevier Ltd. All rights reserved.
1.
Introduction
Anaerobic digestion is a biological decomposition process used to treat, stabilise, and reduce the quantities of organic wastes prior to disposal or beneficial re-use. Anaerobic digestion has been used extensively in municipal wastewater treatment to stabilise primary sludge and activated sludge. Over the past decade, municipal wastewater treatment processes have adapted to meet reduced discharge limits on the effluent nitrogen concentration. Process adaptations include the removal of primary settlings and primary sludge streams; and increased retention times for biological nutrient removal (BNR) processes, resulting in increased sludge age. Increased sludge age results in waste activated sludge with
inherently low degradability, as inert materials in the influent, as well inert decay products accumulate in the activated sludge (Gossett and Belser, 1982). Adaptations of modern wastewater treatment processes have introduced new challenges for anaerobic digestion, as poor degradability of activated sludge requires long digester retention times, higher mixing costs, and also results in poor gas production. Incorporating a pre-treatment into anaerobic treatment may enhance the sludge digestion by accelerating hydrolysis, which is generally accepted as the rate-limiting step in anaerobic digestion. Pre-treatment can enhance overall digestion, and requires a minimal capital investment in comparison with methods such as aerobic digestion (Ros and ic , 2003). Temperature phased anaerobic digestion Zupanc
* Corresponding author. Tel.: þ61 7 3346 9051; fax: þ61 7 3365 4726. E-mail address:
[email protected] (D.J. Batstone). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.11.042
1598
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 9 7 e1 6 0 6
(TPAD), combines a short (1e3 days) thermophilic pre-treatment stage (50e70 C) applied prior to a conventional mesophilic anaerobic digestion (35 C, 10e20 days). TPAD is highly scalable as the process incorporates standard digestion vessels and low quality heat is the main energy input. Thermophilicemesophilic TPAD has been shown to be an effective treatment for increasing methane production and volatile solids (VS) destruction, compared with a single-stage mesophilic digestion. Han and Dague (1997) reported 39% VS destruction of primary sludge achieved in a TPAD system (55 C, 3 days hydraulic retention time (HRT) and 35 C), which was higher than 32% in a single-stage control system (35 C). A corresponding methane production in the TPAD system was also 16% higher over the control system. An improvement of methane production with a TPAD system treating activated sludge was also observed by Bolzonella et al. (2007). They found the highest specific methane production from the TPAD (70 C, 2e3 days HRT and 37 C) was 370 ml gVS1 added, 30e50% higher than that from a single-stage control system (37 C). Extending pre-treatment HRT to 5 days did not improve methane production further. Nges and Liu (2009) evaluated the effect of pre-treatment temperature, and reported that the digestion performance of mixed primary and activated sludge was not influenced by thermophilic temperature, as the same VS destruction of 42% was achieved at pre-treatment temperatures of both 50 and 70 C (2 days HRT). However, this result was still greater than the VS destruction achieved in the single-stage mesophilic control (39%). Watts et al. (2006) reported a substantial improvement of activated sludge digestion with increased thermophilic temperature. They found that a TPAD system (47 C, 2 days HRT and 37 C) achieved the similar VS destruction of 24% as a single-stage mesophilic digester (37 C). VS destruction was not improved with the thermophilic temperature increased to 54 C, but was significantly enhanced to 34% at 60 C. The majority of research has focused on achieving improved performance by varying pre-treatment conditions during TPAD, with performance comparisons against singlestage thermophilic or mesophilic anaerobic digestion. However, there is little analysis to determine the nature of the pre-treatment process; whether it improves the rate or extent of subsequent sludge degradation, or both properties. As a result, optimal pre-treatment conditions (temperature, pH and HRT) have not been established. Additionally, thermophilicemesophilic TPAD rarely has been evaluated in a parallel comparison of mesophilicemesophilic with the same retention times. This study is based on our previous investigation of primary sludge (Ge et al., 2010), and further investigates pre-treatment mechanisms of TPAD on waste activated sludge, with a direct comparison against a control mesophilicemesophilic process.
2.
Materials and methods
2.1.
Substrate
Substrate was waste activated sludge, collected from a biological nutrient removal (BNR) process with 10 days sludge age
and water temperature of approximately 20 C in the Elanora wastewater treatment plant, located at Gold Coast, Australia. The feed was prepared monthly by centrifuging the sludge to a total solids (TS) concentration of 2e3%, and subsequently stored below 4 C. Regular analysis was performed to determine the characteristics and consistency of the feed material. Table 1 shows the average characteristics of the activated sludge feed based on 14 feed collections over 15 months.
2.2.
Start-up and operation
Two identical two-stage systems were used throughout. These consisted of thermophilic pre-treatment (TP) and mesophilic pre-treatment (MP) pre-treatment stages (0.6 L, 2 days HRT), and mesophilic methanogenic stages (4.2 L, 14 days HRT), as shown in Fig. 1. The basic set-up and operation of thermophilic (TP1)emesophilic (TP2) TPAD and mesophilic (MP1)emesophilic (MP2) TPAD systems were described in Ge et al. (2010). Approximately 0.3 L per day of substrate was fed simultaneously by pumping 0.05 L through the pre-treatment stage and methanogenic stage at intervals of 4 h per day (6 times daily, also weighed daily). Gas production was measured daily from each reactor using tipping bucket gas meters, and continuously logged. Reactor pH was also recorded online from each reactor continuously. Each reactor was inoculated using methanogenic inoculum from the methanogenic second stage (35 1 C, 14 days HRT) of a lab-scale thermophilicemesophilic TPAD system (Ge et al., 2010). This provided a diverse microbial community and a common starting point for each reactor. The systems were operated in parallel for over 15 months. During this time the temperature of TP1 was altered to create different operating periods: Period 1: 50 C (186 days), including two periods of pH 5 by dosing 1 M HCL (Day 39e48, and Day 55e77) Period 2: 60 C (100 days) Period 3: 65 C (67 days), including a period of HRT reduced to 12 days in TP2 (Day 330e356) Period 4: 70 C (68 days). The temperature of TP2, MP1 and MP2 was held constant at 35 C during all periods. During periods where the pH of TP1 was reduced, the pH of MP1 was also reduced, and periods
Table 1 e Characteristics of the waste activated sludge used in this study. Measure 1
TS (g L ) VS (g L1) pH COD (g L1) VFA (g COD L1) TKN (g N L1) 1 NHþ 4 eN (g L )
Activated sludge 25.4 0.1 17.5 0.1 6.5e7.5 27.4 3.5 0.2 0.1 1.9 0.5 0.06 0.04
Error margins indicate standard deviation across 14 different feed collections used in the study over 15 months.
1599
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 9 7 e1 6 0 6
Gas to exhaust
Gas meter F Gas meter PLC
F
Water jacket temperature control
Heating coil
Feed reservior
Effluent drum Pretreatment 0.6L TP1 = 50-70°C MP1 = 35°C
Feed pump
Main Digester 4L TP2=35°C MP2=35°C
Effluent pump
Digester pump
Fig. 1 e Schematic diagram of thermophilic pre-treatment TPAD system and mesophilic pre-treatment TPAD system.
where the loading of TP2 was increased, the loading of MP2 was also increased, as a result of HRT shortened to 12 days. This was done in steps of 20% around the average in an attempt to provide better model-parameter identifiability. After each acid dosing period, the pH of TP1 and MP1 returned to their natural levels of 6.6 and 6.8, respectively.
2.3.
Chemical analysis
Gas production and composition (H2, CH4, CO2) were analysed by GCeTCD as described previously (Tait et al., 2009). Liquid samples were collected from each reactor three times per week. Analysis was performed for TS, VS, volatile fatty acid (VFA), chemical oxygen demand (COD), total Kjeldahl nitrogen Analytical (TKN) and ammoniumenitrogen (NHþ 4 eN). methods were based on Standard Methods (APHA, 1998). The preparation and measurement of VFA, soluble COD (COD(S)) and NHþ 4 eN were as described previously (Ge et al., 2010).
VS destruction% ¼
VSconc;in VSconc;out 100 VSconc;in
where VSconc,in ¼ VS concentration of inlet; VSconc,out ¼ VS concentrations of outlet. Results of the mass balance calculation are sensitive to systematic sampling issues, which may cause dilution, while the results of the Van Kleeck calculation are influenced by accumulation of mineral inerts within the reactor (under nonsteady state conditions).
2.4.2.
Extent of solubilisation
Extent of sludge solubilisation in each pre-treatment stage was calculated using the ratio of total solubilised products (methane production and COD(S)) and particulate COD concentration in the inlet feed (Song et al., 2005). Hydrogen was not detected in either pre-treatment stage. Extent of solubilisation can be expressed as Extent of solubilisation% ¼
2.4.
Calculation
2.4.1.
VS destruction
VSfrac;in VSfrac;out VSfrac;in VSfrac;in VSfrac;out
CODCH4 þ CODðSÞo CODðSÞi 100 CODðTÞi CODðSÞi (3)
The two calculation methods used to determine VS destruction were the Van Kleeck equation and the mass balance equation. The Van Kleeck equation (1) assumes the amount of mineral solids is conserved during digestion (Switzenbaum et al., 2003), and uses the volatile fractions (VS/TS VSfrac) in the inlet and outlet as references. VS destruction% ¼
(2)
(1)
where VSfrac,in ¼ volatile fraction (VS/TS) in the inlet solids; VSfrac,out ¼ volatile fraction (VS/TS) in the outlet solids. The mass balance equation (2) uses VS concentrations (VSconc) in the inlet and outlet, expressed as
where CODCH4 ¼ methane production as mg COD during pretreatment; COD(S)i ¼ COD(S) concentration of inlet; COD (S)o ¼ COD(S) concentration of outlet; COD(T)i ¼ total COD concentration of inlet.
2.5.
Mathematical analysis
Mathematical analysis was based on the IWA Anaerobic Digestion Model No. 1 (ADM1) (Batstone et al., 2002). Implementation of ADM1 for a TPAD process is described by Ge et al. (2010), with the input model of Nopens et al. (2009). Initial conditions were adjusted based on measurements of organic solids, organic acids, ammonia, TKN, etc. There were approx 420 input changes over 450 days used in the model. Degradability
1600
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 9 7 e1 6 0 6
extent (fd) and apparent first order hydrolysis rate coefficient (khyd) were the main parameters used to assess and compare two TPAD systems. In each system, khyd and fd were simultaneously estimated to achieve the average optimal values for the whole TPAD process, which were then set to determine the twoparameter uncertainty surface for khyd and fd based on the method of Batstone et al. (2003, 2009). In the TP system, confidence regions were estimated based on system performance at each pre-treatment temperature. For the MP system, operating conditions were constant throughout the experiment; therefore only one confidence region was estimated for comparison. A 95% confidence limit was used, with appropriate F-value (2.996) for 2 parameters and the number of degrees of freedom. Van Kleeck VS destruction was used as a measured variable, with sum of squared errors (c2) as an objective function. Mass balance VS destruction gave similar optimal parameter estimates. However, confidence regions were enlarged, and the upper limit of the region could not generally be determined. Therefore the analysis is based on the Van Kleeck data and mass balance based regions are not shown.
3.
Results
3.1.
Performance of combined TPAD systems
Fig. 2 shows VS destruction in each system calculated by mass balance (Fig. 2A) and Van Kleeck (Fig. 2B) equations. During all periods, VS destruction determined using the mass balance equation (2), was consistent with VS destruction determined using the Van Kleeck equation (1), and this applied to both TPAD systems. Consistent results from the VS calculation methods confirm systematic sampling errors and/or unexpected behaviours were minimal. During Period 1 (50 C pre-treatment), thermophilic pretreatment offered no advantage over mesophilic pre-treatment. Increasing thermophilic pre-treatment temperature to 60 C (Period 2) improved VS destruction in the TP system from 34 1% to 41 1%. A further increase of VS destruction to 48 2% was observed when thermophilic pre-treatment temperature was increased to 65 C (Period 3), but no further enhancement at 70 C (Period 4). Statistical analysis (student’s t-test, a ¼ 0.05) confirmed that VS destruction in the TP system (65 C) was significantly greater than that achieved at 60 C, which was also a significant improvement over that achieved at 50 C. VS destruction in the TP system during Periods 2e4 was also significantly better as compared to the MP system (student’s t-test, a ¼ 0.05). Additionally, pre-treatment pH was temporarily lowered to pH 5 twice during 50 C pre-treatment (Day 39e48 and Day 55e77), which did not influence VS destruction in either system compared to previously. Similarly, VS destruction in each system was maintained as previously described when HRT of methanogenic stages was shortened from 14 to 12 days during Period 3 (Day 330e356). Total methane produced from TP system was consistently higher than that from MP system except Period 1, as shown in Fig. 3 and Table 2. This was confirmed by student’s t-test (a ¼ 0.05) as a statistically significant improvement, and was consistent with enhanced VS destruction in TP system over
MP system during Periods 2e4. The methane production increase in TP system was observed when thermophilic pretreatment temperature was increased to 60 C, but not at 65 C and 70 C. This was not consistent with further enhancement of VS destruction at thermophilic pre-treatment of 65 C compared to 60 C. This is likely due to three reasons: (a) As temperature increased, different portions of VS were degraded by the microbial community resulting in different gas yields per VS destroyed, (b) the decrease in stage 1 methane production (and increase in stage 2 methane production) at higher temperatures is complicating analysis, and (c) it is possible that gas leaks (approx. 10% losses overall) were occurring despite our best efforts. Thus, gas flows have not been used in the detailed model based analysis below, but have been used in solubilisation analysis, and quantitatively compared against model outputs in Section 4.3. During Periods 1e2, approximately 50e70% of methane generated from the TP system was produced in the thermophilic pre-treatment stage, with small amounts from the subsequent mesophilic digestion stage. Increasing temperature of TP1 to 65 C and 70 C caused a substantial decrease in methane production in the first stage. A correspondingly larger amount of methane was produced from TP2, especially during Period 4. In contrary, in MP system, the methane production from MP1 was less compared to that from MP2 in all periods. Moreover, acid dosing was used to lower pH to 5 in TP1 at 50 C, in order to reduce the activity of methanogens and cause washout. As expected, methane production was severely decreased in TP1, producing a corresponding increase in production from TP2. Once the acid dosing stopped, methane production rapidly returned to previous levels.
3.2.
Model based analysis
Fig. 4 shows the 95% confidence regions for degradability (fd) (x-axis) and apparent hydrolysis rate (khyd) ( y-axis) in MP system and TP system at 50 C, 60 C, 65 C and 70 C, respectively. The MP system confidence regions overall at 35 C and TP at 50 C overlapped, indicating statistically the same apparent properties. khyd in the TP system increased significantly above both MP and TP at 50 C as temperature in TP1 was increased to 60 C, but did not increase further to 70 C. The confidence region moved to the right and upward from 60 C to 65 C, indicating a slight improvement in properties, but with full overlap between the regions at 65 C and 70 C. fd in the TP system were comparable for all tested temperatures in TP1, and statistically overlapped with the fd observed in MP. Overall, the results indicate consistent increases in hydrolysis coefficient with increased temperature, but with a relatively constant degradability of 30e60% (Table 3).
3.3.
Analysis of pre-treatment stage in TPAD systems
Extent of solubilisation of the activated sludge determined using equation (3) is shown in Fig. 5. The solubilisation in TP1 was increased from 15% to 27% with the thermophilic temperature increased from 50 C to 60 C, and was not improved further at 65 C and 70 C (Fig. 5). The main profile of solubilisation at 50 C and 60 C was methane production,
1601
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 9 7 e1 6 0 6
70 60
HRT 12 days
% VS destruction in TP system % VS destruction in TP system
pH 5
pH 5
A
50 40 30
VS destruction (%)
20 10 0
% VS destruction in TP system % VS destruction in MP system
60
B
50 40 30 20 10 Period 1
0 0
50
100
150
200
250
Period 4
Period 3
Period 2
300
350
400
450
Time in operation (days) Fig. 2 e VS destruction calculated by mass balance equation (A) and Van Kleeck equation (B) during each period in the thermophilic pre-treatment (TP) and mesophilic pre-treatment (MP) systems (% VS destruction is based on the activated sludge feed characteristics).
with a relatively lower level of VFA and other soluble products presented by COD(S). However, the majority of solubilisation profile was changed to the COD(S) with increasing thermophilic temperature to 65 C and 70 C, as the methane production dropped at higher thermophilic temperatures. For all periods, solubilisation in TP1 was higher than that in MP1, and did not appear to be affected by low pH (pH 5) at 50 C (Day 39e48 and Day 55e77). The VFA profiles were similar in TP1 and MP1 during all periods with acetate as the primary VFA produced, followed by propionate as the second major acid. Other VFAs (iso-butyrate, butyrate, iso-valerate, valerate and hexanoate) were also detected at much lower levels. The acetate concentration in TP1 was lowest during thermophilic pre-treatment of 50 C and 60 C, possibly due to the combination of poor solubilisation
and good methane production (Fig. 6). Increasing the thermophilic pre-treatment temperature to 65 C and 70 C resulted in substantial increases in acetate and propionate concentrations, which was consistent with measurements of COD(S), suggesting that most material hydrolysed was converted to organic acids. It also indicated the methanogenesis was limited at the higher thermophilic temperatures. Methane production was also suppressed under acidic conditions and resulted in increased accumulation of VFA in TP1, as hydrolysis and fermentation processes continued to producing intermediate products (VFAs). Nitrogenous organic compounds contained in the sludge (e.g. protein) is solubilised in the form of NHþ 4 eN during pretreatment process, thus NHþ 4 eN is another important indicator of solubilisation. Solubilisation according to NHþ 4 eN in TP1
1602
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 9 7 e1 6 0 6
1.2
1.0
CH4 production in MP system
-1
CH4 production (L day )
CH4 production in MP1
CH4 production in TP1 CH4 production in TP system
0.8
0.6
0.4
0.2
0.0
Period 1
Period 3
Period 2
Period 4
Fig. 3 e Average methane production during each period in the thermophilic pre-treatment (TP) and mesophilic pretreatment (MP) systems (Error bars are 95% confidence in mean methane production).
4.1.
Overall performance of TPAD systems
An increase of pre-treatment temperature from 50 C to 65 C substantially improved the overall VS destruction from 34% to 48% for activated sludge, but not for primary sludge (54% at 50e65 C) (Ge et al., 2010). It should be noted that performance on primary sludge is not diminished above 65 C, and thus mixed feed digesters should be operated at elevated temperatures. This not only decreases sludge disposal costs substantially, it also allows for much better hygenisation and pathogen removal. In addition, it provides performance of VS destruction
1.2 Degrades faster
Discussion
TP 70 C
1.0
0.8 TP 60 C
-1
4.
above 38%, which is one of the legislative levels for performance implemented by US EPA (EPA, 1994), and most Australian legislation. At higher performance levels (Period 3), VS destruction was not strongly impacted by a change from 14 to 12 days, consistent with a hydrolysis coefficient of >0.5 d1. Correspondingly, the implementation of smaller digesters or increased organic loading rates could be possible, which could substantially reduce capital costs. Another advantage of the TPAD system is the use of renewable methane to produce energy, which is used to compensate heat requirements (heat demand and heat losses) in the TPAD system, and is conventionally produced by cogeneration engines. The main heat demand from the TPAD system is to heat up the sludge to the required temperature in
khyd (d )
increased with each step increasing thermophilic pre-treatment temperature (Fig. 7). Release of NHþ 4 eN was greater from TP1 than from MP1 for all periods, which was consistent with the extent of solubilisation results. The NHþ 4 eN results indicated protein fermentation was improved under thermophilic condition, and improved with thermophilic temperature increase. Again, low pH did not have an impact on NHþ 4 eN release in TP1.
0.6 TP 65 C MP 35 C
0.4
Table 2 e A summary of methane production (L gVSfedL1 fed) in the thermophilic pre-treatment (TP) system and mesophilic pre-treatment (MP) system during each period. TP system Period Period Period Period
1 2 3 4
(50 C) (60 C) (65 C) (70 C)
0.10 0.16 0.15 0.17
0.03 0.02 0.02 0.03
MP system 0.07 0.11 0.10 0.09
0.04 0.03 0.03 0.04
Error margins indicate standard deviation across different gas measurements over each period.
TP 50 C 0.2
0.0 0.2
0.3
0.4
0.5
fd
0.6 Degrades more
Fig. 4 e 95% confidence regions for apparent hydrolysis coefficient (khyd, dL1) and degradability (fd) using Van Kleeck VS destruction as an objective function in the mesophilic pre-treatment (MP) system and thermophilic pre-treatment (TP) system at 50, 60, 65 and 70 C, respectively.
0.7
1603
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 9 7 e1 6 0 6
Table 3 e A summary of apparent hydrolysis coefficient (khyd) and degradability (fd) in the mesophilic pre-treatment (MP) system and thermophilic pre-treatment (TP) system. Parameter
TP system
1
khyd (d ) fd
MP system
50 C
60 C
65 C
70 C
0.12 0.06 0.4 0.1
0.5 0.1 0.41 0.04
0.7 0.2 0.51 0.04
0.8 0.2 0.53 0.02
0.2 0.1 0.4 0.1
Numbers after ‘’ are the linear, uncorrelated 95% confidence in parameter values.
the thermophilic stage, as well as to the mesophilic temperature in the second stage. Heat from thermophilic stage will also be used in the mesophilic stage, and minimises overall heating requirements in a TPAD process. A detailed evaluation of the heat balance, including losses and sensitivity to feed concentration is contained in the supplementary information. This analysis also considers the increased performance provided by thermophilic operation, but does not consider a smaller main digester. The heat balance for thermophilic and mesophilic systems is generally positive at a 2% feed concentration, with either supplemental heat needed, or diversion of methane from electricity to heat energy, as shown in Fig. i (Supplementary information). The heat balance became negative when increasing the feed concentration to 4% in both systems, indicating the potential heat production could fully offset heat requirements in both systems. At the feed concentration of 6%, the potential heat production was greatly in excess, especially in the TP system with thermophilic pre-treatment of 65 C and 70 C. This emphasises the need for pre-thickening, but importantly, indicates that the heat balance is very similar for standard mesophilic and TPAD systems, with TPAD processes operated above 60 C generally producing more excess energy than a mesophilic process. It should also be noted that we assume only waste heat is used, from cogeneration engines with electricity being produced as the main process.
The extent of solubilisation (%)
40
Methane VFAs Others
30
20
TP1
4.2.
Regions measurably moved upwards and to the right with increased temperature from 50 C to 75 C, reflecting the increase in performance. The regions also decreased in area as temperature increased, likely due to the increase in VS destruction. Because VS destruction is a product (or fractional) term, it can be determined with better accuracy at higher destruction levels. There are a wide range of hydrolysis rates reported in literature for mixed streams containing both primary sludge and activated sludge, between 0.1 and 1 d1 (Pfeffer, 1974; Ghosh, 1981; Bolzonella et al., 2007). However, the determination of hydrolysis rate for digestion of WAS only has been limited, and is now addressed in our study. Hydrolysis coefficient was sensitive to temperature, with no improvement at 50 C, but significant increases at 60 C and higher. This may be due to emergence of true thermophiles at the higher temperatures that do not emerge at the intermediate temperature of 50 C. However, this contrasts to our previous work on primary sludge, where hydrolysis rate was improved above 50 C, but not sensitive to temperature above this. Apparent degradability was not significantly impacted, indicating that the improved VS destruction observed in this study was due to increase in apparent hydrolysis rate, rather than an increase in degradable fraction. Analysis of degradability is consistent with our previous observations on primary sludge (Ge et al., 2010). This increase in rate rather than degradability fraction is similar to the effects of mechanical pre-treatments, e.g. sonication (Climent et al., 2007; Khanal et al., 2007). This is in contrast to high impact methods such as thermal hydrolysis, which increase rate and extent substantially (Batstone et al., 2009). In a full-scale plant, faster degradation could be utilised in either design of a smaller main digester, or intensification of an existing process.
4.3. 10
0
MP1
Period 1
Period 2
Period 3
Period 4
Fig. 5 e Extent of solubilisation during each period in the thermophilic pre-treatment stage (TP1) and the mesophilic pre-treatment stage (MP1) (% solubilisation is based on the activated sludge feed characteristics and equation (3)).
Model based analysis
Pre-treatment mechanisms
Solubilisation was enhanced in TP1 at 50 C over MP1 at 35 C (Fig. 5), even though hydrolysis coefficient remained the same (Fig. 4). Both measures were consistently enhanced at 60 C and above. The inconsistency at 50 C is likely because the apparent hydrolysis coefficient was acquired from the overall performance, and was therefore dominated by the second stage performance, whereas the information in Fig. 5 is based on the first stage (methane þ VFAs þ other products). This was further tested by simulating the first stage further, and comparing model outputs to observed results. The model simulation of
1604
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 9 7 e1 6 0 6
-1
tVFA concentration (mg L )
4000
tVFA in TP1 tVFA in MP1 pH 5
pH 5 3000
2000
1000
Period 2 (60 C)
Period 1 (50 C)
0 0
50
100
150
200
250
Period 3 (65 C) 300
350
Period 4 (70 C) 400
450
Time in operation (days) Fig. 6 e tVFA concentrations during each period in the thermophilic pre-treatment stage (TP1) and mesophilic pre-treatment stage (MP1).
valid overall), while hydrolysis coefficients estimated at increased pre-treatment temperatures were valid across both digesters. However, solubilisation results and model analysis both confirmed that the solubilisation or hydrolysis was improved in TP1 at 60e70 C compared to MP1. Therefore, improved first stage hydrolysis is a major factor contributing to enhanced performances in the TP system. The increased thermophilic temperature may improve production of extracellular enzymes to hydrolyse more complex or inert substrate materials, and have selected the specialised microbial community,
solubilisation followed the same trend of measured solubilisation in TP1, except at 50 C, where model predictions were conservative compared to solubilisation measurements. Improvements at 60 C to 65 C (or 70 C) were reflected in increased solubilisation according to both model and measurements. Additionally, fractions of solubilisation predicted by model and measurements in the TP1 were consistent, but the model could not predict the decrease in methane production at 65 C and 70 C, indicating a model limitation. This comparison suggests that the hydrolysis coefficient determined at 50 C was conservative for the first stage (but
1200 NH4+-N in TP1 NH4 -N in MP1
pH 5
pH 5
-1
NH4 -N concentration (mg L )
+
1000
800
600
+
400
200 Period 2
Period 1
0 0
50
100
150
200
250
Period 3 300
Period 4 350
400
450
Time in operation (days) Fig. 7 e NHD 4 -N concentrations during each period in the thermophilic pre-treatment stage (TP1) and mesophilic pretreatment stage (MP1).
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 9 7 e1 6 0 6
which will result in an optimized hydrolysis. All these possible improvements will increase the substrate availability for digestion in the subsequent methanogenic stage. NHþ 4 eN release in the pre-treatment stage was consistent with the extent of solubilisation data and had a significant influence on overall NHþ 4 eN release. It was exhibited as the 1 NHþ 4 eN released from TP2 (approx. 880 mg L ) and was 30% 1 higher than that from MP2 (approx. 680 mg L ). However, the enhanced NHþ 4 eN release in TP1 with increased thermophilic temperature did not influence the NHþ 4 eN release in methanogenic stage. This result was different from the similar NHþ 4 eN release observed in both methanogenic stages with individual thermophilic (50e65 C) and mesophilic pre-treatment (35 C) treating primary sludge (Ge et al., 2010). It indicated that overall conversion was improved further in the methanogenic stage.
4.4. Methanogenesis during thermophilic pre-treatment process There is some uncertainty over how best to operate TPAD systems. If the initial step is regarded purely as a hydrolytic step, methanogenesis is not required. This is not the case for our results, and no difference is seen where methanogenesis is inhibited by low pH or high temperature compared to where methanogenesis is allowed to occur. Particularly for activated sludge compared to primary sludge (Ge et al., 2010), the methane production in the pre-treatment stage is far higher, with approx 65% of total production occurring in the whole system. Methanogenesis in the pre-treatment stage is not detrimental, since it allows for increased overall methanogenic retention time, improved kinetics, and provides protection to the secondary main stage. The presence of substantial methanogenesis in this pre-treatment stage, with a very short retention time is of some interest. Methanogenesis at short HRT may be due to change in metabolic pathways from aceticlastis to acetate oxidation, under which acetate is first oxidised to hydrogen and carbon dioxide, and subsequently converted to methane. This is enhanced thermodynamically at higher temperatures 50 C to 65 C (Karakashev et al., 2006; Zinder et al., 1984), and is supported by microbial results indicating a dominance of Methanosarcinaceae, which has been found to be the dominant methanogen in acetate oxidising systems by isotopic carbon analyses (Karakashev et al., 2006). While methanogenesis in the pre-treatment stage complicates the process by the presence of two methane producing units, the action of a two-step acetate oxidation/ aceticlastic methanogenic process can provide advantages. This offers better resistance to process inhibition and toxins, since aceticlastis and acetate oxidisers are influenced by different factors. As an example, protein-rich cattle and piggery wastes, have high ammonia, to which acetate oxidisers are less susceptible than aceticlasts (Karakashev et al., 2005). In essence, the robust acetate oxidation step can act to protect the more sensitive aceticlastic methanogenic step, and provide a biological buffer. A decrease of 31% and 58% methane production in TP1 at 65 C and 70 C compared to 60 C suggests the activity of methanogenesis (presumptively acetate oxidation, or both) is decreased. Therefore, the level of methanogenesis in the pre-
1605
treatment stage can be tuned by temperature, especially at 60e70 C, without negative impact on overall VS destruction. Based on our results, it is also possible to suppress methanogenesis by decreasing the pH to 5, but doing so by changing temperature is lower cost, and easier implemented.
5.
Conclusions
The following conclusions can be drawn from this study: VS destruction in thermophilicemesophilic TPAD was increased by thermophilic pre-treatment at 50 C to 65 C (34%e48%), which was 11e30% higher than that in mesophilicemesophilic TPAD (37%), expect thermophilic pretreatment of 50 C. Model based analysis indicated the hydrolysis coefficient in the TP system was not improved under thermophilic pretreatment of 50 C (0.06e0.18 d1) compared to the MP system (0.1e0.3 d1), but significantly enhanced to 0.6 d1 at 60 C, up to 1 d1 at 65 C and 70 C. However, increasing thermophilic pre-treatment temperature had no impact on the overall degradability in the TP system relative to the MP system (0.30e0.55). Solubilisation was improved during thermophilic pretreatment relative to mesophilic pre-treatment, and reached to maximum of 27% at thermophilic pre-treatment of 60 C. Further thermophilic temperature increases had no further impacts. Higher NHþ 4 eN was released during thermophilic pre-treatment over mesophilic pre-treatment, and further increased by increasing the thermophilic pre-treatment temperature from 50 C to 70 C. A large amount of methane was produced from thermophilic pre-treatment stage between 50 C and 60 C, but started to decrease with further increase of temperature to 65 C and 70 C. Methane production from the pre-treatment stage was heavily inhibited at acidic conditions (pH 5).
Acknowledgements This work was funded by the Queensland State Government, under the Smart State Research-Industry Partnerships Program (RIPP), Meat and Livestock Australia, and Environmental Biotechnology Cooperative Research Centre (EBCRC), Australia as P23 “Small-medium scale organic solids stabilization”. Huoqing Ge and Paul Jensen are recipients of an EBCRC postgraduate scholarship and postdoctoral award, respectively. We thank Gold Coast City Council (Gold coast water) for supplying samples from their Elanora Wastewater Treatment Plant.
Appendix. Supplementary information Supplementary data related to this article can be found online at doi:10.1016/j.watres.2010.11.042.
1606
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 5 9 7 e1 6 0 6
references
APHA, 1998. Standard Methods for the Examination of Water and Wastewater, 20th ed. American Public Health Association, Washington, DC, USA. Batstone, D.J., Keller, J., Angelidaki, I., Kalyuzhnyi, S., Pavlostathis, S.G., Rozzi, A., Sanders, W.T.M., Siegrist, H., Vavilin, V., 2002. IWA Task Group for Mathematical Modelling of Anaerobic Digestion Processes. IWA Publishing, London, UK. Batstone, D.J., Pind, P.F., Angelidaki, I., 2003. Kinetics of thermophilic, anaerobic oxidation of straight and branched chain butyrate and valerate. Biotechnology and Bioengineering 84 (2), 195e204. Batstone, D.J., Tait, S., Starrenburg, D., 2009. Estimation of hydrolysis parameters in full-scale anaerobic digesters. Biotechnology and Bioengineering 102 (5), 1513e1520. Bolzonella, D., Pavan, P., Zanette, M., Cecchi, F., 2007. Two-phase anaerobic digestion of waste activated sludge: effect of an extreme thermophilic prefermentation. Industrial and Engineering Chemistry Research 46 (21), 6650e6655. Climent, M., Ferrer, I., Baeza, M.D., Artola, A., Vazquez, F., Font, X., 2007. Effects of thermal and mechanical pretreatments of secondary sludge on biogas production under thermophilic conditions. Chemical Engineering Journal 133 (1e3), 335e342. Ge, H.Q., Jensen, P.D., Batstone, D.J., 2010. Pre-treatment mechanisms during thermophilic-mesophilic temperature phased anaerobic digestion of primary sludge. Water Research 44 (1), 123e130. Ghosh, S., 1981. Kinetics of acid-phase fermentation in anaerobic digestion. Biotechnology and Bioengineering 11, 301e313. Gossett, J.M., Belser, R.L., 1982. Anaerobic-digestion of waster activated sludge. Journal of the Environmental Engineering Division-Asce 108 (6), 1101e1120. Han, Y., Dague, R.R., 1997. Laboratory studies on the temperaturephased anaerobic digestion of domestic primary sludge. Water Environment Research 69 (6), 1139e1143. Karakashev, D., Batstone, D.J., Angelidaki, I., 2005. Influence of environmental conditions on methanogenic compositions in anaerobic biogas reactors. Applied and Environmental Microbiology 71 (1), 331e338.
Karakashev, D., Batstone, D.J., Trably, E., Angelidaki, I., 2006. Acetate oxidation is the dominant methanogenic pathway from acetate in the absence of Methanosaetaceae. Applied and Environmental Microbiology 72 (7), 5138e5141. Khanal, S.K., Grewell, D., Sung, S., Van Leeuwen, J., 2007. Ultrasound applications in wastewater sludge pretreatment: a review. Critical Reviews in Environmental Science and Technology 37 (4), 277e313. Nges, I.A., Liu, J., 2009. Effects of anaerobic pre-treatment on the degradation of dewatered-sewage sludge. Renewable Energy 34 (7), 1795e1800. Nopens, I., Batstone, D.J., Copp, J.B., Jeppsson, U., Volcke, E., Alex, J., Vanrolleghem, P.A., 2009. An ASM/ADM model interface for dynamic plant-wide simulation. Water Research 43 (7), 1913e1923. Pfeffer, J.T., 1974. Temperature effects on anaerobic fermentation of domestic refuse. Biotechnology and Bioengineering 16 (6), 771e787. ic , G.D., 2003. Thermophilic anaerobic digestion of Ro s, M., Zupanc waste activated sludge. Acta Chimica Slovenica 50, 359e374. Song, H., Clarke, W.P., Blackall, L.L., 2005. Concurrent microscopic observations and activity measurements of cellulose hydrolyzing and methanogenic populations during the batch anaerobic digestion of crystalline cellulose. Biotechnology and Bioengineering 91 (3), 369e378. Switzenbaum, M.S., Farrell, J.B., Pincince, A.B., 2003. Relationship between the Van Kleeck and mass-balance calculation of volatile solids loss. Water Environment Research 75 (6), 572. Tait, S., Tamis, J., Edgerton, B., Batstone, D.J., 2009. Anaerobic digestion of spent bedding from deep litter piggery housing. Bioresource Technology 100 (7), 2210e2218. US Environmental Protection Agency, 1994. A plain English Guide to the EPA Part 503 Biosolids Rule. In: Environmental protection Agency 8322/R-93-003, September Washington, DC, USA. Watts, S., Hamilton, G., Keller, J., 2006. Two-stage thermophilicmesophilic anaerobic digestion of waste activated sludge from a biological nutrient removal plant. Water Science and Technology 53 (8), 149e157. Zinder, S.H., Anguish, T., Cardwell, S.C., 1984. Effects of temperature on methanogenesis in a thermophilic (58 C) anaerobic digester. Applied and Environmental Microbiology 47 (4), 808e813.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 0 7 e1 6 1 6
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Threshold concentrations of biomass and iron for pressure drop increase in spiral-wound membrane elements W.A.M. Hijnen*, E.R. Cornelissen, D. van der Kooij KWR Watercycle Research Institute, PO Box 1072, 3430 BB Nieuwegein, The Netherlands
article info
abstract
Article history:
In a model feed channel for spiral-wound membranes the quantitative relationship of
Received 27 October 2010
biomass and iron accumulation with pressure drop development was assessed. Biofouling
Received in revised form
was stimulated by the use of tap water enriched with acetate at a range of concentrations
26 November 2010
(1e1000 mg C l1). Autopsies were performed to quantify biomass concentrations in the
Accepted 29 November 2010
fouled feed channel at a range of Normalized Pressure Drop increase values (NPDi). Active
Available online 7 December 2010
biomass was determined with adenosinetriphosphate (ATP) and the concentration of bacterial cells with Total Direct Cell count (TDC). Carbohydrates (CH) were measured to
Keywords:
include accumulated extracellular polymeric substances (EPS). The paired ATP and CH
Spiral-wound membranes
concentrations in the biofilm samples were significantly ( p < 0.001; R2 ¼ 0.62) correlated
Biofouling
and both parameters were also significantly correlated with NPDi ( p < 0.001). TDC was not
NPD
correlated with the pressure drop in this study. The threshold concentration for an NPDi of
Biomass
100% was 3.7 ng ATP cm2 and for CH 8.1 mg CH cm2. Both parameters are recommended
ATP
for diagnostic membrane autopsy studies. Iron concentrations of 100e400 mg m2 accu-
Carbohydrates
mulated in the biofilm by adsorption were not correlated with the observed NPDi, thus
Fe
indicating a minor role of Fe particulates at these concentrations in fouling of spiral-wound
Threshold values
membrane. ª 2010 Elsevier Ltd. All rights reserved.
1.
Introduction
Microbial growth (‘biofouling’) in high pressure spiral-wound (SW) membranes for nanofiltration (NF) or reverse osmosis (RO) has been identified as a major cause of operational problems such as increased feed channel pressure drop (PD), decreased mass transfer coefficient (MTC) and product quality decline. First studies on biofouling date from the 1970s and 1980s of the twentieth century (Bailey and Jones, 1974; Potts et al., 1981; Ridgway et al., 1985) and a number of reviews on this issue have been published in the 1990s (Flemming, 1997; Flemming
et al., 1993a; Ridgway and Flemming, 1996) because of the increasing application of membrane processes in water treatment and desalination. Destructive membrane sampling (autopsies) has been used to analyze the composition and structure of accumulated biofilms in order to elucidate the fundamentals of the biofouling process in spiral-wound membranes. With different microscopic techniques membrane foulants have been detected and identified as bacterial matter (Ridgway and Flemming, 1996). Still there is a lack of information on the quantitative relationship between biomass concentrations and the resulting operational problems in
Abbreviations: AOC, assimilable organic carbon; ATP, Adenosinetriphosphate; CH, carbohydrates; EPS, extracellular polymeric substances; FS, feed spacer; HPC, heterotrophic colony plate count; IPC, ion chromatography; NF, nanofiltration; NPD, normalized pressure drop; MFS, membrane fouling simulator; MTC, mass transfer coefficient; PD, pressure drop; Rf, exponential fouling rate constant; RO, reversed osmosis; SW, spiral-wound; TDC, total direct cell count. * Corresponding author. E-mail address:
[email protected] (W.A.M. Hijnen). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.11.047
1608
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 0 7 e1 6 1 6
NF/RO membranes. Such information is not only needed for diagnostic purposes and improvement of pretreatment but also for assessing the efficacy of cleaning procedures to control biofouling. Microbial analysis such as Heterotrophic Plate Counts (HPC) and Total Direct Cell counts (TDC), and physical and (bio) chemical analysis including total wet weight of deposits, adenosinetriphosphate (ATP), extracellular polymeric substances (EPS) and proteins have been used to measure the amount of biomass on SW membranes (Flemming and Schaule, 1988; Griebe and Flemming, 1998; Ridgway et al., 1983; Schaule et al., 1993; Vrouwenvelder et al., 1998; Vrouwenvelder et al., 2008). Only few studies have tried to establish quantitative relationships between these biomass parameters and the pressure drop or flux decline. Flemming et al. (1993b) observed an MTC decline of 25% at total cell coverage of the membrane surface of 5 107e2 108 cells cm2 and suggested a ‘pain level’ of bacterial cells of 108 per cm2; no correlation with pressure drop was presented. This ‘pain level’ corresponds with the amount of bacterial cells observed in SW membranes with operational problems related to biofouling (Griebe and Flemming, 1998; Hijnen et al., 2009; Schaule et al., 1993; Vrouwenvelder et al., 1998). Measuring active bacterial biomass with ATP in cell cultures or biomass samples is attractive because the analytical method is rapid, cheap and simple to perform and has a low detection level; a concentration of 1 ng l1 ATP can be detected without concentration techniques. The proportional relationship between ATP and TDC (Magic-Knezev and Van der Kooij, 2004; Vrouwenvelder et al., 2008) indicates that ATP is a potential parameter to quantify the amount of accumulated biomass. Furthermore, autopsy results from full-scale SW membrane installations showed that the increase of the Normalized Pressure Drop (NPD) was related to ATP concentrations (Vrouwenvelder et al., 2008). However, establishment of a causal relationship between ATP and NPD requires more defined conditions to exclude effects of other deposits (dead biomass, EPS and other organic or inorganic substances). Quantification of carbohydrates (CH) with the Dubois method (Dubois et al., 1956) in autopsy studies enables to estimate biomass concentrations based on EPS which consists of polysaccharides with a large water-retention capacity resulting in voluminous deposits. The Dubois method is commonly used in membrane autopsy studies (Gabelich et al., 2004; Griebe and Flemming, 1998; Ridgway et al., 1983) and correlated with flux decline (Fonseca et al., 2007). Biofouling rarely occurs without mineral deposition (Ridgway and Flemming, 1996) and Fe was identified as a predominating foulant in SW elements (Baker and Dudley, 1998), but a causal relationship between Fe and PD increase was not reported. Hence, evaluation of the use of ATP, TDC and CH as quantitative biomass parameters in diagnostic autopsies and cleaning studies as well as elucidation of the role of Fe in pressure drop problems requires biofouling studies under well defined conditions. In a recent laboratory study using a Membrane Fouling Simulator (MFS) (Vrouwenvelder et al., 2006) a quantitative relationship between acetate as a model substrate and the pressure drop increase was demonstrated (Hijnen et al., 2009). Samples of the biofouled membranes were available for autopsy studies. The objectives of the current study were:
(i) elucidation of the quantitative relationship between biomass parameters ATP, TDC and CH and the extent of the PD increase and (ii) determination of the threshold concentrations of these parameters for a 100% increase of the normalized pressure drop (NPD) and (iii) to investigate the role of iron as the major mineral in the water under the experimental conditions. Such information enables the selection of proper biomass parameter(s) in autopsies to assess the cause of PD in membrane elements.
2.
Materials and methods
2.1.
Biofouling of an NF membrane
The Membrane Fouling Simulator (MFS) loaded with sheets (7 30 cm) of a “virgin” nanofiltration membrane sheet (Trisep 4040-TS80-TSF) was supplied with non-chlorinated tap water after filtration (10 and 1 mm poly-propyleen cartridge filtration; Van Borselen Ltd.) to exclude accumulation of suspended solids and spiked with low amounts of acetate-C to initiate biofouling. These experiments have been described in detail (Hijnen et al., 2009). Briefly, the MFS is a small scale continuous flow model of an SW feed channel (0.8x4x22.cm) filled with the matching Trisep feed spacer (0.8 4 20 cm; front 2 cm without feed spacer FS) and operated at a constant feed water flow of 16 l h1 (cross-flow velocity of 0.14 m s1) at a constant pressure of 1 bar without permeation. Fig. 1 depicts the experimental set up. The rate of clogging of the feed channel was measured by monitoring the pressure drop normalized (NPD) to a moderate environmental temperature of 12.5 C in the feed channel (Hijnen et al., 2009). The extent of biofouling, given by the relative NPD increase (NPDi), is calculated from the final NPD (NPDf) and the initial NPD (NPDo) by %NPDi ¼
NPDf NPDo $100% NPDo
(1)
The MFS units were supplied with pre-filtered tap water spiked with acetate-C at concentrations (Sac) of 1, 3, 5, 10, 25, 100, 500 and 1000 mg l1. Four blank MFS units with no acetate supply were operated with either filtered tap water or unfiltered tap water.
2.2.
Feed water quality
The feed water was non-chlorinated tap water produced from anaerobic groundwater using aeration and rapid sand filtration. The pH of the water was 7.98 0.05, dissolved organic carbon content was 2.0 0.1 mg C l1, assimilable organic carbon 3 concentration (AOC) was 3e5 mg acetate-C eq l1, NO 3 and PO4 1 content was 0.12 0.04 and 0.02 0.02 mg l respectively. The iron content (ion chromatography; ICP method with a lower detection limit of 0.005 mg l1) of the filtered tap water was 0.008 0.014 mg l1 and 0.32 0.24 mg l1 in the unfiltered tap water. Iron was the major mineral in the tap water and visually (brown deposits) accumulated in the biofilms. The ambient water temperature was daily monitored during the experiments and ranged from 13.5 to 16.8 C (average of 15.9 0.7 C) and was 19.4 2.0 C in one experiment.
1609
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 0 7 e1 6 1 6
PI
Tap water from ground water filtered by Cartridge filters 10 and 1µm
PI
Nutrient dosage
FI
FI
QC
PR
MFS unit
a
QC
Sensitive pressure drop registration dP
b QC
Chlorine dosage
QC
FC
FC
Fig. 1 e Experimental set up for the experiments with a total of five MFS units. Unit on the left with dosage equipment and unit on the right with the mobile pressure drop monitor; PI and FI [ pressure and flow indicator; FC [ flow controller; QuC [ quick connector.
2.3.
Autopsy of the membrane sheets
Two samples (1.5 2 cm) were cut from the membrane at the inlet section of the feed channel without spacer (no FS) and five samples (2 2 cm) were cut from the membrane with spacer. The samples were transferred to sterile glass tubes with 20 ml of autoclaved tap water and sonicated with High Energy Sonication (HES, #100) using a BRANSON Digital Sonifier (Model 250 D) at optimized conditions established in a previously published study (Magic-Knezev and Van der Kooij, 2004). The sonifier tip (size 6.5 mm) was inserted into the tube (1e2 cm) containing 20 ml of autoclaved water and the membrane/spacer sample. This tube was placed in melting ice and sonicated for one minute at an amplitude of 45% (15e20 watt) to separate the biomass from the membrane/spacer samples. This treatment was repeated in 20 ml fresh sterile tap water and both suspensions were mixed to obtain a total sample volume of 40 ml. The biomass samples of the experiments with Sac of 25, 3 and 1 mg l1 were collected by additional swabbing to enlarge the biomass recovery. The swab was treated with HES for 1 min in 20 ml autoclaved tap water and subsequently mixed with the 40 ml HES suspension.
the analytical procedure was described in detail before (Hijnen et al., 2009). The detection limit is 180 cells ml1, which corresponds to 720 cells cm2.
2.5.
The CH concentration in the biofilm samples was analyzed with the method described by Dubois et al. (1956) using glucose as the reference carbohydrate. The extinction/adsorption at 490 nm was measured directly in the biomass suspension after hydrolysis and complexation with sulphuric acid and phenol, respectively and expressed in glucose equivalent concentration. The detection limit of this parameter was approximately 5e10 mg cm2 depending on the sampled membrane area.
2.6.
Iron content
The iron (Fe) content of the obtained biomass suspension was assessed with Atomic Absorption Spectrometry resulting in a lower limit of detection of approximately 1 mg cm2 of membrane surface.
2.7. 2.4.
Carbohydrate analysis
Correlation analysis and statistics
Microbial parameters
ATP was measured to determine the amount of active bacterial biomass in the biofilm samples. The analysis is based on measuring the amount of light produced by an enzymatic reaction using the luciferineeluciferase assay in a luminometer (Celcis Ltd.) and has a lower limit of detection of 1 ng l1, which corresponds with 0.01 ng cm2 of membrane surface. The method has been described in detail previously (Magic-Knezev and Van der Kooij, 2004). The total direct cell count (TDC) was based on counting of fluorescing cells using epifluorescence microscopy (Hobbie et al., 1977) and
Correlation analyses was done by determining Pearson’s correlation coefficient between paired values of ATP, TDC, CH and Fe using SPSS 17.0 software with a significance level of p0.01. For the correlation of the biomass and Fe accumulated in the MFS units with the NPD increase (%), the weighted average concentrations (Cavg ) were calculated from the concentrations observed in the samples at different locations in the MFS units using Pn Cavg ¼
i¼1
ðCi þ Ciþn Þ=2 Aiþn Atot
(2)
1610
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 0 7 e1 6 1 6
where and Ci, Ciþn and Aiþn (cm2) are the concentration and the surface area at the ith part of n parts of the feed channel surface (n ¼ 3e7) and Atot is the total surface area of the MFS unit. The linear regression analysis of the correlation between the Cavg -values of the biomass parameters and Fe concentrations obtained from membrane autopsy and the NPDi was performed with Excel software. For the correlation of the biomass parameters and Fe concentrations with the NPDi the non-parametric Spearman’s rank correlation coefficient was calculated and multi-regression analysis was conducted with SPSS 17.0 software.
3.
Results
of concentrations in the subsequent parts of the channel (Fig. 3). In the units supplied with acetate the percentage of the total amount of ATP at the inlet section (no FS) was 1.5e9%, in the first 2 cm with feed spacer 8e16% and in the last part (18e20 cm) 3e10% (Fig. 3c). At acetate concentrations of 1000, 500 and 25 mg C l1 the ATP concentrations were higher than in the units supplied with the lower acetate concentrations (10, 5, 3 and 1 mg C l1). In the blank MFS units without acetate dosing a lower ATP concentration was observed (Fig. 3b). The spatial distribution of parameters TDC and CH and also of Fe in the channels was similar to the distribution of ATP; a declining concentration in the section with feed spacer (no figures presented; weighted average concentrations presented in Table 1).
3.1.
NPD increase
3.3.
In the MFS units supplied with acetate-enriched tap water biofouling was observed at each concentration (Fig. 2) and the NPD increase (NPDi) was characterized as a first order process (Hijnen et al., 2009). The blank MFS unit supplied with prefiltered tap water without added acetate showed no NPDi during 28 days of operation (Fig. 2a), whereas in units supplied with 1 mg of acetate-C/l biofouling was observed (Fig. 2b). Also no fouling was observed within 100 days of operation in the two blank units supplied with unfiltered water (Fig. 2c). After 100 days the pressure drop started to increase in these units. The accumulated biofilm in the feed channels was colourless at high biofouling rates and short operation times (<20 days). At lower biofouling rates and operational times of 25 days the feed channel showed accumulation of brown coloured deposits. These observations initiated the analysis of the Fe concentrations in the fouled membrane samples.
3.2.
Spatial distribution of biomass and Fe
The units were sampled for biomass and Fe concentrations at different fouling conditions with relative NPDi values ranging from 71 to 3390% (Table 1). The MFS units supplied with acetate showed high ATP concentrations at the inlet section without spacer (no FS), further elevated concentrations in the first part of the section with feed spacer (0e2 cm) and a decline
Norm. pressure drop (kPa)
100
a
1000 µg C/l
The operational periods with acetate dosing, the final NPDi values and the exponential fouling rate constant Rf and the weighted average values (Cavg ) of the biomass parameters and Fe for the correlation analysis are presented in Table 1. The correlation analysis of the paired biomass parameters showed that the log value of the ATP concentration in the MFS units was significantly ( p < 0.01) correlated with the log value of the TDC and the CH concentrations, respectively (Table 2). The linear regression equation for the relationship with TDC was Log [ATP] ¼ 0.79 (95% CI 0.69e0.89) Log [TDC] þ 4.1 (95% CI 3.6e4.6) with a goodness of fit (R2) of 0.75. Based on this correlation 1 ng of ATP equals 3 106 (95%CI 5.3 105e1.7 107) TDC cm2. The CH concentration ranged between 10 and 100 mg cm2 at ATP concentrations of 10e100 ng cm2, but the linear regression fit of paired ATP and CH values was poor (R2 ¼ 0.39). A better fit (R2 ¼ 0.62; p < 0.0001) was observed for the values of the units operated under acetate limitation conditions where the fouling rate Rf was below Rf,max (Sac 10 mg l1). ATP and Fe concentrations were not correlated when the results of the MFS units operated at high Sac values with relatively short operation times (20 days) were included. For the MFS units operated at Sac values 10 mg l1 with longer operational periods ATP and Fe concentrations were significantly correlated ( p < 0.001; Table 2). TDC did not correlate with CH and Fe. The latter two parameters correlated significantly ( p < 0.001) with a better 20 blank unfiltered
90
10 µg C/l
15
80
blank filtered
10
blank filtered
70
Correlation analysis of biomass parameters and iron
blank unfiltered
5
60
c
0 0
50
10 8 6 4 2 0
40 30 20 10 0 0
10
20
Operational time (days)
30
50
100
150
200
1 µg C/l 1 µg C/l
Stop
0
20
40
60
Start
b 80
100
Operational time (days)
Fig. 2 e The development of the Normalized pressure drop (NPD) in the MFS units supplied with filtered tap water enriched with different acetate concentrations (a,b) and (c) supplied with unfiltered tap water (lines in b and c are duplicates).
1611
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 0 7 e1 6 1 6
Table 1 e The fouling conditions of the MFS experiments and the weighted average concentration (Cavg ) of the biomass parameters and iron measured by autopsies. Acetate Sac (mg C/l)
Operational Final pressure % NPD Rf b time (days) drop (kPa) increase (ln NPDi d1)
0 pre-filtered 0 unfiltered 1 3 5 10 25 500 1000
20 184; 184 146; 98 34; 152 39; 46 35; 28 35; 33 20; 15 15; 14; 8
2.9 17.8; 15.3 3.3; 7.2 18.6; 24.5 15.7; 37.9 7.9c; 81.4 37.8; 54.7 61.3; 22.2 52.1; 18.4; 7.9
S1a S2a
31 53
46.1 45.4
a b c d e
6.0 156; 157 71; 119 306; 526 376; 919 234c; 2369 1352; 993 3390; 507 1820; 445; 182 725 1231
<0.001 0.015; 0.016 0.063; 0.027 0.102; 0.109 0.128; 0.245 0.205; 0.224 0.766; 0.696 0.859; 1.144 1.126; 1.475; 1.097 100: 1.123 1000: 1.160
ATP (ng cm2)
TDC CH Fe (cell 108 cm2) (mg gluc. eq cm2) (mg m2)
0.8 3; 2 6; 3 20; 18 16; 13 28; 46 175; 158 200; 91 118; 58; 184
0.02 0.3; 0.4 Ndd; 0.1 2.0; 2.3 0.3; 0.4 0.3; 0.6 1.4; 0.5 2.2; 3.1 1.0; 1.2; 1.4
9.4 9.6; 10.0 Nde Nde 11.7; 11.9 13.0; 44.6 Nde 52.4; 20.6 21.1; 14.4; 47.1
0.32 394; 298 96; 129 83; 173 149; 212 201; 426 82; 79 11; 1 1; nd; 7
37 37
0.5 0.8
21.3 37.3
334 154
Starvation experiments with variable acetate dosages and starvation periods (S1 ¼ 100-5 and S2 ¼ 1000-1000-10) (Hijnen et al., 2009). First order fouling rate Rf values from Hijnen et al. (2009) modified as submitted in an erratum (Hijnen et al., in press). Low NPDi caused by preferential flow path in the feed channel. Nd ¼ not determined. Unreliable CH data due to the use of cotton swab.
goodness of fit for the units with acetate-C concentrations of 10 mg l1 (Table 2).
3.4.
Correlation with NPDi
The study aimed at assessing the relationship between the biomass concentration and Fe with the NPDi at the time of the autopsy. The weighted average ATP and CH concentrations in the MFS units were both significantly ( p < 0.01) correlated to the NPDi (%) as evaluated with the non-parametric Spearman’s rank correlation coefficient (R2 of 0.71 and 0.91, respectively). The linear regression analysis also showed a significant ( p < 0.001) correlation with a high correlation coefficient for
ATP and CH (0.52, 0.70 and 0.82; Table 2 and Fig. 4). No significant correlation was observed for TDC with NPDi (Fig. 4c). The variability of ATP and CH concentrations in the MFS units is presented in Fig. 4 with the standard deviation (s.d.; n ¼ 3e7). The bars show an increased variability of both parameters at increased NPDi values which was caused by increased heterogeneity of the concentrations in the feed channel (Fig. 3). No correlation was found between the concentrations of Fe and the NPDi values (Fig. 4c), but the regression plots of ATP and CH with NPDi revealed that the low ATP and CH concentration at relatively high NPDi values contained Fe concentrations of >100e200 mg m2. However, a multi-regression analysis in combination with either ATP or CH again showed
Fig. 3 e Distribution of ATP concentrations (error bar is s.d.) in the MFS units supplied with filtered tap water enriched with different acetate-C concentrations (mg lL1) and with unfiltered tap water (a,b) and (c) % of the total ATP amount in the membrane feed channel without feed spacer (no FS) and after 0e2 and 18e20 cm with feed spacer.
1612
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 0 7 e1 6 1 6
Table 2 e Correlation matrix of the different biomass parameters, Fe and NPDi measured in the standard MFS experiments; presented are the square of the Pearson correlation coefficient R2, for all correlations p-values were <0.001 except for values indicated by * ( p-value < 0.01), the number of observations (n). Sac values (mg C L1)
TDC (cells cm2)
CH (mg cm2)
Fe (mg m2)
NPDic (%)
1e1000 10 1e1000 1e1000 10 1e1000
0.75; 85a 0.66; 43a 1
0.39; 49 0.62; 30 nc 1
ncb ( p ¼ 0.09) 0.56; 42 nc 0.34; 30 0.65; 16 1
0.52; 19 0.70*; 9 nc 0.82; 13
ATP (ng cm2) TDC (cells cm2) CH (mg cm2) Fe (mg m2)
nc
a Log transformed values. b Nc ¼ no correlation. c Correlation with the weighted average concentrations of the parameters.
no significant correlation between Fe and NPDi. This indicates that the contribution of Fe accumulation in the MFS units to the pressure drop increase was limited. The minor effect of the Fe concentrations on the NPDi was also demonstrated by the Fe content in the MFS units supplied with unfiltered tap water (unfiltered blanks; Fig. 4c). The Fe concentrations in these unfiltered blanks with a limited NPDi of 156% were 298 and 394 mg m2, whereas ATP and CH concentrations were low (2e3 ng cm2 and 9.6e10 mg cm2, respectively; Table 1). In the MFS units at Sac value of 10 mg L1 considerably higher NPDi values (234e2369%) were observed at comparable Fe content of 201e426 mg m2 and higher biomass concentrations of 28e46 ng ATP cm2 and 13e44.6 mg CH cm2. Similar observation was recorded for the MFS unit supplied with 100 and 5 mg l1 acetate and intermediate starvation period; Fe, ATP and CH content was 334 mg m2, 37 ng cm2 and 21.3 mg cm2, respectively at an NPDi of 725%. The pressure drop increase was due to a decrease of the open pore volume of the feed channel which in turn was a result of biomass accumulation. The relationship between the biofilm thickness and the NPDi has been described with hydraulic equations (Schock and Miquel, 1987) and is linear in the initial stage of biofouling but exponential in the subsequent stage (Hijnen et al., 2009). Assuming that ATP and CH 10000
concentrations were linearly related with the biofilm thickness the correlations with NPDi were also tested for an exponential relationship. Both exponential fits were significant ( p < 0.01), but the goodness of fit was lower compared to the linear regression (Fig. 4).
3.5.
The ATP concentration in the feed channels of the MFS units for an NPDi of 100% was 3.7 ng cm2 (95% CI ¼ 1.3e10.9), calculated from the equation presented in Fig. 4a. For CH the threshold concentration for this criterion was calculated from the equation given in Fig. 4b at 8.1 mg cm2 (95% CI of 6.1e11.7). This was around the detection limit of the analysis of 5e10 mg cm2. For TDC and Fe no threshold concentration was calculated because of the lack of correlation with NPDi.
3.6.
NPDi (%)
4000
c
Unfiltered blank
3000 2000
1000
100
Fouling and accumulation rate
The fouling rate in the feed channel of the MFS units could be described with the exponential fouling rate constant Rf (Table 1). Formation of biofilms on surfaces initially is an exponential process that is rapidly followed by a linear phase due to diffusion limitation of the substrate flux into the
a
Fe < 100 mg Fe > 100 mg Fe > 200 mg
Threshold concentrations
1000 0
Exponential R2 = 0.40 p=0.003
0
100
Proportional relation
300 -2
400
500
b
2
R = 0.48 p=0.009
3000 10
200
Fe mg.m
4000
2000 NPDi=63.7[CH]-413
logNPDi=0.79[logATP]+1.55
R2 = 0.82 p<0.0001
1000
2
R =0.73 p<0.001
0
1 0.1
1
10
100 -2
ATP (ng.cm )
1000
0
20
40
60
80
100
-2
CH (µg.cm )
Fig. 4 e The relationship between the accumulated biomass measured with ATP and CH (a,b) and (c) the accumulated mass of Fe with the NPDi (%); error bar is s.d. (n [ 4e7).
1613
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 0 7 e1 6 1 6
biofilm (Rittmann, 1995). This was clearly demonstrated for ATP and Fe accumulation on glass (Van der Kooij et al., 2003). On base of the concentrations of ATP, CH and Fe measured in the MFS units after the different dosing periods the linear accumulation rate of these parameters was calculated and correlated with the Sac in the influent and Rf (Fig. 5a). Correlations of both biomass parameters with Sac showed a similar saturation curve and relationship with Sac as described for Rf (Fig. 5a). The ATP and CH accumulation rates were strongly correlated ( p < 0.0001) with Rf with R2 of 0.91 for ATP and 0.90 for CH. This clearly demonstrates the proportional relationship of both biomass parameters with the porosity decline in the feed channel.
4.
Discussion
4.1.
ATP and TDC as biomass parameters in autopsies
In full-scale SW membrane filtration installations where operation is hampered by fouling problems, it is common practice to carry out an autopsy to verify the cause of the fouling process. Only few studies have been published on the quantitative correlation between biomass parameters and operational problems in membranes such as pressure drop increase and flux decline (Flemming et al., 1993b; Fonseca et al., 2007; Vrouwenvelder et al., 2008). The results of the present study show that ATP is a suitable parameter to elucidate the role of biofilm formation in the pressure drop increase in such membranes. This conclusion was based on the correlation with the observed NPDi and supported by the good correlation between the biofilm formation rate (ng ATP cm2 d1) in the feed channel of the MFS with the acetate concentration (Fig. 5a) and the exponential fouling rate constant Rf (Fig. 5b). Additionally, this clearly shows that the assessment of the biofilm formation rate for the feed water of SW membranes which is also based on ATP measurements (Van der Kooij et al., 2003) is an appropriate parameter to assess the biofouling potential of the feed water.
100
a
Rf ATP
b
ATP CH
CH
ATP/CH (ng.cm-2.d-1)
Rf (ln NPDi.d-1); ATP/CH (ng/µg.cm-2.d-1)
100
The choice of a 100% NPDi in the current study to assess a threshold biomass concentration was based on a commonly used NPDi cleaning criterion of 15% over one stage of a series of six successive membrane elements (Graham et al., 1989; Hickman, 1991; Speth et al., 1998). The NPDi is not evenly distributed over the elements and usually is mainly located in the first element. Consequently, the NPDi in this element is higher (Vrouwenvelder et al., 2009a) and may be close to 100%. The threshold ATP concentration for 100% NPDi in the MFS units was 3.7 ng cm2. A higher threshold ATP concentration for 100% NPDi of 30 ng cm2 was reported for SW elements operated under field conditions (Vrouwenvelder et al., 2008). However, one would expect this the other way around: lower for the same NPDi in the field elements because of differences in biofilm conditions. MFS units of the present study contained relatively young biofilms whereas biofilms in field elements were more aged with a lower ratio between active (ATP) and total biomass (including EPS and dead cell material). This difference between threshold values might be caused by the difference in 100% NPDi over SW elements and the MFS of the current study. It can also be caused by correlating on one hand the maximum ATP concentration with the NPDi in field elements with a length of 1 m (Vrouwenvelder et al., 2008) and on the other hand the weighted average ATP concentration with the NPDi in a 0.2 m feed channel of the MFS as done in the present study. Consequently, despite the positive correlations ATP results in field autopsies must be interpreted with care and additional parameters which are more related to the total amount of the accumulated biomass are needed. The present study and also the mentioned field study (Vrouwenvelder et al., 2008) revealed that in contrast to ATP, TDC was not correlated with NPDi. The range of TDC values of 1 107e3.1 108 corresponds with a biofilm thickness of 0.1e1.6 mm (assumed bacteria cell volume of 0.5 mm3; diameter of 1 mm). Theoretically for the 100% NPDi a biofilm thickness of 60 mm was estimated (Hijnen et al., 2009) thus indicating that microscopic cell count (TDC) is not an accurate parameter for total biomass and more importantly biofilms consist of more than bacterial cells.
10
1
0,1
ATP; c = 1.75 R2 = 0.913 p<0.0001
10
CH; c = 0.96 R2 = 0.903 p<0.0001
1
0,1 Rf = 1.14x(1-exp{-0.693*Sac/20.3}) (Hijnen et al., 2009;in press)
0,01 0,1
1
10
Sac (µg C.l-1)
100
1000
0,01 0,001
0,01
0,1
1
10
Rf (Ln NPDi.d-1)
Fig. 5 e The correlation of the exponential fouling rate Rf and the ATP and CH accumulation rates with the acetate concentrations Sac (a) and (b) the correlation of the biomass (ATP/CH) accumulation rates with Rf.
1614
4.2.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 0 7 e1 6 1 6
Carbohydrates as a biomass parameter
Biofilms are established by adsorption and adherence of bacteria followed by growth due the supply of nutrients. Extracellular polymeric substances (EPS) excreted by bacteria to anchor themselves to the surface and to each other play a key role in the development of biofilms (i.e. protection against environmental stress, nutrient availability; Or et al., 2007; Flemming and Wingerden, 2010). CH are important components of EPS (Sutherland, 1999). The method of Dubois et al. (1956) is commonly applied to quantify the CH concentration in field SW elements (Gabelich et al., 2004; Griebe and Flemming, 1998; Ridgway et al., 1983, 1984) and their reported CH concentrations were in the same order of magnitude as measured in the current study. The positive correlation between the CH concentration and pressure drop in the feed channel of the MFS (Fig. 4b) and the proportional correlation of the CH accumulation rate (mg cm2 d1) to the exponential fouling rate the Rf. (Fig. 5b) confirms that the CH concentration in SW membranes is a valuable parameter in diagnostic autopsies. The threshold CH concentration of 8.1 mg cm2 for the defined NPDi was around the current detection limit of the analysis, but the analysis can be optimized by sampling larger surface areas. More recently the CH parameter was correlated with flux decline in NF membranes (Fonseca et al., 2007) and they reported a decline of 30e80% when the CH concentrations increased to 50 mg cm2. Another study presented a >50% flux decline and 100% NPD increase in SWM elements at a CH concentration of 12.4 mg cm2 (Gabelich et al., 2004). Consequently, we propose the use of the parameters ATP and CH in membrane autopsy studies: the ATP method which is cheap and fast and reveals information on the accumulated active biomass in the feed channel and CH which represents the total amount of active and inactive biomass. Inclusion of the analysis of CH is especially of interest for studies on the effect of membrane cleaning with chemicals (Cornelissen et al., 2009).
4.3.
Fe accumulation and pressure drop increase
In the flat sheet MFS units without permeate production the process of particulate accumulation on the membrane surface was not influenced by vertical forces and particle settling which normally occur in SW elements with permeate production (Belfort and Nagata, 1985; Belfort, 1988). Thus, the observed accumulation of Fe particles in the MFS was a result of adsorption of these particles onto biomass produced on the membrane surface during the short cross-flow contact time. The results of the current study show that the accumulation of biomass has a far greater effect on NPDi in the feed channel than the accumulation of Fe particulates (Fig. 4c). In the units supplied with unfiltered tap water lower biofilm concentrations were observed than in the units supplied with 1 mg acetate-C L1 (Fig. 3) but the Fe content was much higher at similar NPDi values (71e157%; Table 1). The significant correlation between CH and Fe (Table 2) indicates that EPS plays a substantial role in the adsorption of Fe onto charged biopolymers which is related to the presence of negatively charged carboxylic and phosphate groups (Wuertz et al., 2001). The dominant role of biomass in the NPDi is explained by the high water-retention capacity of EPS. Water-retention curves
show that certain polysaccharides hold more than 50e70 g of water per gram while maintaining structural coherence (Or et al., 2007; Chenu, 1993). No studies on the role of Fe particulates in SW membranes on pressure drop are known to the authors. A recent study presented the correlation of the mass deposit of Fe micro- and nanoparticles in a porous sand column (208 cm3; empty space volume of 102 cm3 and porosity of 0.49) with the pressure drop increase (Vecchia et al., 2009). In this study an Fe concentration of 2.6 mg cm3 resulted in a PDi in the sand column of 1 kPa. The Fe concentration in the MFS supplied with unfiltered water was 300e400 mg m2 (Table 2) which equals a volumetric concentration of 0.41e0.54 mg cm3 (channel height of 0.0008 m and porosity of 92%). The %NPDi in this MFS was 156e157% with higher ATP and TDC values than in the pre-filtered blank (Table 1). These calculations show that Fe accumulation in the feed channel of SW elements at a level of 100 mg m2 (10 mg cm2) has no effect on the NPD. Further studies under field conditions are required, however, to collect additional data on the relationship of particulate accumulation and biofouling.
4.4.
Feed spacer enhances biofilm accumulation
The spacer in the feed channel enhanced biomass accumulation (Fig. 3) which is consistent with observations in other studies (Picioreanu et al., 2009; Vrouwenvelder et al., 2009b). Possible explanations for this observation are: an increase in attachment area or/and enhanced mass transfer of nutrients to the biofilm due to increased turbulence. Based on a specific surface area of the feed spacer of 7700 m2 m3 (Picioreanu et al., 2009) it can be estimated that the spacer contributes with around 25% to the attachment surface in the feed channel. An earlier autopsy study on SWM elements from field locations showed more accumulation of biomass (ATP) on the membrane (38e90%) than on the feed spacer (5e62% of the total amount) (Vrouwenvelder et al., 2008). Preferential flow paths shown by Computational Fluid dynamics and filamentous streamers at the spacer junctions (Picioreanu et al., 2009; Vrouwenvelder et al., 2009b) were not observed in the present study. Verification of the role of the feed spacer in biofouling of SWM elements requires further research.
5.
Conclusions
The effect of biomass accumulation in spiral-wound membranes on pressure drop increase can be elucidated by measuring concentrations of active biomass with adenosinetriphosphate (ATP) and of total biomass with carbohydrates (CH; Dubois, method) in membrane autopsies. There was a significant correlation ( p < 0.001) between these parameters in the current study. This study also showed a significant ( p < 0.001) and causal relationship between both parameters and the NPDi in a model feed channel. Furthermore, the calculated ATP and CH accumulation rates were highly correlated with the observed exponential fouling rate. Threshold concentrations for 100% NPDi were 3.7 ng ATP cm2 and 8.1 mg CH cm2. Because ATP is related to active biomass and CH to the total biomass, monitoring both parameters in autopsies will reveal further information on the metabolic
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 0 7 e1 6 1 6
state of the accumulated biofilm. Iron accumulation in the feed channel was enhanced by the biofilm growth as demonstrated by the significant correlation between CH and Fe concentrations ( p < 0.001). Iron concentrations of 100 mg m2 (10 mg cm2) of membrane surface did not contribute to pressure drop increase in spiral-wound membranes. The high impact of accumulation of low biomass concentrations on pressure drop increase is attributed to the high water-retention characteristics of polysaccharides in biofilms.
Acknowledgements The research was conducted as part of the Joint Research Program of the Dutch Water Supply Companies and in the MEDINA project co-funded by the European Commission under contract number 036997. The excellent technical support by Nanda Berg, Anke Hanzens-Brouwer and Meindert de Graaf from KWR and Amandine Balthazard and David Biraud from the Ecole Nationale Supe´rieure de Chimie de Mulhouse is greatly appreciated.
references
Bailey, D.A., Jones, K.M.C., 1974. The reclamation of water from sewage effluents by reversed osmosis. Wat. Pollut. Control 74, 353e366. Baker, J.S., Dudley, L.Y., 1998. Biofouling in membrane systems e a review. Desalination 118, 81e90. Belfort, G., 1988. Membrane modules: comparison of different configurations using fluid mechanics. J. Mem. Sci. 35, 245e270. Belfort, G., Nagata, N., 1985. Fluid mechanics and cross-flow filtration: some thoughts. Desalination 53, 57e79. Chenu, C., 1993. Clay or sand polysaccharide associations as models for the interface between micro-organisms and soil: water related properties and microstructure. Geoderma 56, 143e156. Cornelissen, E.R., Hijnen, W.A.M., van der Kooij, D., 2009. Assessment of the efficiency of hydraulic and chemical cleaning for the removal of biomass from surfaces in a laboratory test, Membrane based desalination: an Integrated Approach. EU-China workshop, Qingdao, China. Dubois, M., Gilles, K.A., Hamilton, J.K., Rebers, P.A., Smith, F., 1956. Colorimetric method for determination of sugars and related substances. Anal. Chem. 23, 350e356. Flemming, H.-C., 1997. Reverse osmosis membrane biofouling. Exp. Thermal Fluid Sci. 14, 382e391. Flemming, H.-C., Schaule, G., 1988. Biofouling of membranes, a microbiological approach. Desalination 70, 95e119. Flemming, H.-C., Schaule, G., McDonough, R., 1993a. How do performance parameters respond to initial biofilm formation on separation membranes. Vom Wasser 80, 177e186. Flemming, H.-C., Schaule, G., McDonogh, R., 1993b. Permeation properties of biofilms on separation membranes and the problem of biofouling. J. Membr. Res. 87, 199e217. Flemming, H.-C., Wingender, J., 2010. The biofilm matrix. Nat. Rev. Microbiol.. doi:10.1038/nrmicro2415. Fonseca, A.C., Summers, R.S., Greenberg, A.R., Hernandez, M.T., 2007. Extra-cellular polysaccharides, soluble microbial products, and natural organic matter impact on nanofiltration membranes flux decline. Environ. Sci. Technol. 41, 2491e2497. Gabelich, C.J., Yun, T.I., Ishida, K.P., Leddy, M.B., Safarik, J., 2004. The effect of naturally occurring biopolymers on polyamide
1615
membrane fouling during surface water treatment. Desalination 161, 263e276. Graham, S.I., Reitz, R.L., Hickman, C.E., 1989. Improving reverse osmosis performance through periodic cleaning. Desalination 74, 113e124. Griebe, T., Flemming, H.-C., 1998. Biocide-free antifouling strategy to protect RO membranes from biofouling. Desalination 118, 153e156. Hickman, C.E., 1991. Membrane cleaning techniques, AWWA Seminar Proceedings, Membrane Technologies in the Water Industry, Orlando (US, Fl), March 10e13. Hijnen, W.A.M., Biraud, D., Cornelissen, E.R., van der Kooij, D., 2009. Threshold concentration of easily assimilable organic carbon in feedwater for biofouling of spiral-wound membranes. Environ. Sci. Technol. 43, 4890e4895. Hijnen, W.A.M., Biraud, D., Cornelissen, E.R., van der Kooij, D.Erratum: threshold concentration of easily assimilable organic carbon in feedwater for biofouling of spiral-wound membranes. Environ. Sci. Technol., in press. Hobbie, J.E., Daley, R.J., Jasper, S., 1977. Use of nuclepore filters for counting bacteria by fluorescence microscopy. Appl. Environ. Microbiol. 33, 1225e1228. Magic-Knezev, A., Van der Kooij, D., 2004. Optimisation and significance of ATP analysis for measuring active biomass in granular activated carbon filters used in water treatment. Water Res. 38, 3971e3979. Or, D., Phutane, S., Dechesne, A., 2007. Extracellular polymeric substances affecting pore-scale hydrologic conditions for bacterial activity in unsaturated soils. Vadose Zone J 6, 298e305. Picioreanu, C., vrouwenvelder, J.S., van Loosdrecht, M.C.M., 2009. Three-dimensional modeling of biofouling and hydrodynamics in feed spacer channels of membrane devices. J. Mem. Sci. 345, 340e354. Potts, D., Ahlert, R.C., Wang, S.S., 1981. A critical review of fouling of reversed osmosis membrane. Desalination 36. Ridgway, H.F., Flemming, H.-C., 1996. Membrane biofouling. In: Malleviall, J., Odendal, P.E., Wiesner, M.R. (Eds.), Water Treatment Membrane Processes. McGraw-Hill, New York. Ridgway, H.F., Kelly, A., Justice, C., Olson, B.H., 1983. Microbial fouling of reverse-osmosis membranes used in advanced wastewater treatment technology: chemical, bacteriological and ultrastructural analyses. Appl. Environ. Microbiol. 45, 1066e1084. Ridgway, H.F., Rigby, M.G., Argo, D.G., 1984. Adhesion of Mycobacterium sp. to cellulose diacetate membranes used for reverse osmosis. Appl. Environ. Microbiol. 47, 61e67. Ridgway, H.F., Rigby, M.G., Argo, D.G., 1985. Bacterial adhesion and fouling of reverse osmosis membranes. J. Am. Water Works Assoc. 77, 97e106. Rittmann, B.E., 1995. Fundamentals and applications of biofilm processes in drinking water treatment. In: Hrubec, J. (Ed.), The Handbook of Environmental Chemistry. Springer Verlag, Berlin, Heidelberg, New York, pp. 61e87. Schaule, G., Kern, A., Flemming, H.-C., 1993. RO treatment of dump trickling water membrane biofouling: a case history. Desalination & Water Reuse 3, 17e23. Schock, G., Miquel, A., 1987. Mass transfer and pressure loss in spiral-wound modules. Desalination 64, 339e352. Speth, T.F., Summers, R.S., Guess, A.M., 1998. Nanofiltration foulants from a treated surface water. Environ. Sci. Technol. 32, 3612e3617. Sutherland, I.W., 1999. Polysaccharases for microbial exopolysaccharides. Carbohydr. Polym. 38, 319e328. Van der Kooij, D., Vrouwenvelder, H.R., Veenendaal, H.R., 2003. Elucidation and control of biofilm formation processes in water treatment and distribution using the unified biofilm approach. Wat. Sci. Technol 47, 83e90. Vecchia, E.D., Luna, M., Sethi, R., 2009. Transport in porous media of highly concentrated iron micro- and nanoparticles in the presence of Xanthan gum. Environ. Sci. Technol. 43, 8942e8947.
1616
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 0 7 e1 6 1 6
Vrouwenvelder, H.R., Van Paassen, J.A.M., Folmer, H.C., Hofman, J.A.M.H., Nederlof, M.M., Van der Kooij, D., 1998. Biofouling of membranes for drinking water production. Desalination 118, 157e166. Vrouwenvelder, J.S., Van Paassen, J.A.M., Wessels, L.P., Van Dam, A.F., Bakker, S.M., 2006. The membrane fouling simulator: a practical tool for fouling prediction and control. J. Mem. Sci. 281, 316e324. Vrouwenvelder, J.S., Manolarakis, S.A., van der Hoek, J.P., van Paassen, J.A.M., van der Meer, W.G.J., van Agtmaal, J.M.C., Prummel, H.D.M., Kruithof, J.C., van Loosdrecht, M.C.M., 2008. Quantitative biofouling diagnosis in full scale nanofiltration and reverse osmosis installations. Water Res. 42, 4856e4868.
Vrouwenvelder, H.R., Paassen, J.A.M., Kruithof, J.C., van Loosdrecht, M.C.M., 2009a. Sensitive pressure drop measurements of individual lead membrane elements for accurate early biofouling detection. J. Mem. Sci. 338, 92e99. Vrouwenvelder, H.R., Graf von der Schulenburg, D.A., Kruithof, J.C., Johns, M.L., van Loosdrecht, M.C.M., 2009b. Biofouling of spiralwound nanofiltration and reverse osmosis membranes: a feed spacer problem. Water Res. 43, 583e594. Wuertz, S., Spaeth, R., Hinderberger, A., Griebe, T., Flemming, H.-C., Wilderer, P.A., 2001. A new method for extraction of extracellular polymeric substances from biofilms and activated sludge suitable for direct quantification of sorbed metals. Wat. Sci. Technol. 43, 25e31.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 1 7 e1 6 2 8
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Photocatalytic degradation of benzenesulfonate on colloidal titanium dioxide Erzse´bet Szabo´-Ba´rdos a, Otı´lia Markovics a, Otto´ Horva´th a,*, Norbert To¨ro} b, Gyula Kiss c a
University of Pannonia, Institute of Chemistry, Department of General and Inorganic Chemistry, H-8200 Veszpre´m, POB. 158, Hungary University of Pannonia, Institute of Environmental Sciences, H-8200 Veszpre´m, POB. 158, Hungary c Air Chemistry Group of Hungarian Academy of Sciences at University of Pannonia, H-8201 Veszpre´m, POB. 158, Hungary b
article info
abstract
Article history:
Titanium dioxide-mediated photocatalyzed degradation of benzenesulfonate (BS) was
Received 16 August 2010
investigated by monitoring chemical oxygen demand (COD), total organic carbon (TOC)
Received in revised form
content, sulfate concentration, pH as well as the absorption and emission spectral changes
26 November 2010
in both argon-saturated and aerated systems. Liquid chromatography-mass spectrometry
Accepted 29 November 2010
analysis was utilized for the detection of intermediates formed during the irradiation in the
Available online 7 December 2010
UVA range (lmax ¼ 350 nm). The results obtained by these analytical techniques indicate that the initial step of degradation is hydroxylation of the starting surfactant, resulting in
Keywords:
the production of hydroxy- and dihydroxybenzenesulfonates. These reactions were
Benzenesulfonate
accompanied by desulfonation, which increases [Hþ] in both argon-saturated and aerated
Photocatalysis
systems. In accordance with our previous theoretical calculations, the formation of ortho-
Titanium dioxide
and meta-hydroxylated derivatives is favored in the first step. The main product of the
Oxidative degradation
further oxygenation of these derivatives was 2,5-dihydroxy-benzesulfonate. No decay of
Intermediates
the hydroxy species occurred during the 8-h irradiation in the absence of dissolved oxygen.
Oxygenation
In the aerated system much more efficient desulfonation and hydroxylation, moreover, a significant decrease of TOC took place at the initial stage. Further hydroxylation led to cleavage of the aromatic system, due to the formation of polyhydroxy derivatives, followed by ring fission, resulting in the production of aldehydes and carboxylic acids. Total mineralization was achieved by the end of the 8-h photocatalysis. It has been proved that in this photocatalytic procedure the presence of dissolved oxygen is necessary for the cleavage of the aromatic ring because hydroxyl radicals photochemically formed in the deaerated system too alone are not able to break the CeC bonds. ª 2010 Elsevier Ltd. All rights reserved.
1.
Introduction
Sulfonated aromatic compounds are used in consumer products and in many industrial processes (Tully, 1997). Linear alkylbenzenesulfonates (LASs) are utilized as surfactants in laundry and cleansing products; sulfonated azo dyes are used in diverse applications including the textile industry to color natural fibers. Fluorescent whitening agents are based on
sulfonated aromatics, benzene- and naphthalenesulfonates are used mainly as intermediates for the manufacturing of azo dyestuffs, pharmaceuticals and tanning agents. While the frequently used linear alkylbenzene sulfonate surfactants have been thoroughly investigated for their pollution effects and degradation possibilities, little work has been carried out with benzene- and naphthalenesulfonates (Lange et al., 2000). In contrast to LASs, which have been found to be readily
* Corresponding author. Tel.: þ36 88 624 159; fax: þ36 88 624 548. E-mail address:
[email protected] (O. Horva´th). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.11.045
1618
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 1 7 e1 6 2 8
biodegradable (Takada and Ishiwatari, 1990; Hashim et al., 1992), aromatic sulfonates without long alkyl side chains proved to be resistant to biodegradation (Cain, 1981). This is especially true for benzene- and naphthalenesulfonates with sulfo, nitro and amino groups (Brilon et al., 1981; Zu¨rrer et al., 1987; Wittich et al., 1988). Because of their low n-octanolewater partition coefficients (Greim et al., 1994) and high mobility within aquatic systems, polar aromatic sulfonates are regularly found in natural waters. These compounds have been detected in wastewater effluents, surface waters (Alonso et al., 1999; Zerbinati et al., 1999), and landfill leachates (Suter et al., 1999; Riediker et al., 2000a, 2000b). Low biodegradability and high mobility make them potentially hazardous with respect to contamination of ground water and drinking water supplies (Reemtsma, 1996). The aquatic toxicity of aromatic sulfonates appears to be small (Integral Consulting Inc., 2007). However, a toxicity study using Photobacterium phosphoreum indicated that among several naphthalene- and benzenesulfonates, the unsubstituted benzenesulfonate turned out to be the most toxic (Alonso et al., 2005). Several methods have been tested for the degradation of these pollutants over the past 15 years. Mineralization of benzenesulfonates was accomplished by contact glow discharge electrolysis (Amano et al., 2004; Amano and Tezuka, 2006). Ozonation proved to be efficient only in the presence of activated carbon, which ensures high local concentration of the reactants, due to their adsorption (Faria et al., 2008). Carboxylic acids (oxalic and formic acid) were detected as intermediates in both cases. Electrochemical oxidation of 1, 5-naphthalenedisulfonic acid proved to be promising by in situ generation of silver(II) or peroxydisulfate as mediators (Ravera et al., 2004). Recently, oxidative degradation of this pollutant was achieved in the presence of hydrogen peroxide activated by microwaves or UV irradiation (Ravera et al., 2009, 2010). Photocatalytic methods have been proved to be suitable for the treatment of water polluted with organic and inorganic contaminants, such as surfactants (Horva´th and Husza´nk, 2003; Horva´th et al., 2005), heavy metals (Kanki et al., 2004), chromium(VI) (Kajitvichyanukul et al., 2005; Gkika et al., 2006), and various pesticides (Konstantinou and Albanis, 2003; Devipriya and Yesodharan, 2005). In heterogeneous photocatalytic methods applied for the degradation of various organic pollutants the most widely used material is titanium dioxide, TiO2 (Szabo´-Ba´rdos et al., 2003, 2004; Fabbri et al., 2006; Patsoura et al., 2007). The most efficient oxidizing agent in TiO2-mediated photocatalysis is the HO radical, which can be formed in aqueous systems via the oxidation of adsorbed water by the positively charged hole (hþ vb ) formed in the valence band of the semiconductor upon excitation (Hoffmann et al., 1995). þ TiO2 þ hv/TiO2 e cb þ hvb þ TiO2 hvb þ H2 Oads /TiO2 þ HO þ Hþ
(2)
2.
Experimental section
2.1.
Materials
The titanium dioxide catalyst used in all experiments was Degussa P25 (70% anatase, 30% rutile; with a surface area of 50 m2 g1). Benzenesulfonic acid as well as 4-hydroxybenzenesulfonic acid and 2,5-dihydroxybenzenesulfonic acid (as standards for the analyses) of pure reagent grade were purchased from Merck. Compressed air or argon for stirring was introduced into the reaction mixtures from gas bottles. Beside stirring, air also served as an electron acceptor (i.e., oxidizer). High purity water used in these experiments was double distilled and then purified with a Milli-Q system.
(1)
In aerated systems, electrons (e cb ) photogenerated in the conduction band can react with dissolved oxygen, resulting in the formation of superoxide and peroxide ions. (3) TiO2 e cb þ O2ads /TiO2 þ O2 2 TiO2 e cb þ O2 /TiO2 þ O2
TiO2-based techniques were applied for degradation of various amino acids (Matsushita et al., 2007; Szabo´-Ba´rdos et al., 2006) and surfactants (Zhang et al., 2003; Hegyi and Horva´th, 2004). Efficient photocatalytic mineralization of 1,5naphthalenedisulfonate was achieved on colloidal titanium dioxide (Szabo´-Ba´rdos et al., 2008). In this case, HPLC/MS analysis indicated that the degradation pathway leads to the formation and subsequent decay of benzenesulfonate. Although the photocatalytic degradation of benzenesulfonate was studied in the past (Sangchakr et al., 1995), neither a detailed analysis regarding the possible degradation pathways was carried out nor was the role of dissolved oxygen investigated. Since benzenesulfonate is involved in several industrial processes, from which it can get into natural waters as a pollutant, it is important to gain more information about its photocatalytic mineralization. Elucidation of its decomposition contributes to a better understanding of the TiO2-mediated photodegradation of disulfonatonaphthalenes too, in which it is an important intermediate (Szabo´-Ba´rdos et al., 2008). The objective of this study is to investigate the TiO2-based photocatalytic treatment of benzenesulfonate (abbreviated as BS) in a laboratory-scale reactor in order to elucidate the oxidative degradation mechanisms of this surfactant. The changes in the properties of the treated solution during photocatalysis were followed by the measurement of the pH, the absorption and emission spectra, the total organic carbon (TOC) content and the chemical oxygen demand (COD) of the irradiated mixture. TOC was chosen as a mineralization index of BS, while COD is related to the average oxidation number of the carbon atoms in the system. In addition, the intermediates were identified with MS, giving vital information to determine the main steps of the degradation mechanism. Beside their theoretical importance, the results of this work can be utilized in the design of photocatalytic procedures for wastewater treatment.
(4)
2.2.
Photochemical experiments
Photochemical experiments were carried out using a laboratory-scale reactor with an effective volume of 1.6 dm3. The heterogeneous reaction mixture (TiO2 suspension) was circulated by a peristaltic pump through the reactor and the buffer vessel and by continuously bubbling air or other gases, such as Ar with a flow rate of 40 dm3 h1 within the reactor. The photon flux of the internal light source (40 W,
1619
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 1 7 e1 6 2 8
3.
Results and discussion
3.1. Change of TOC and COD in photoassisted degradation of benzenesulfonate Practically no change of TOC was observed over the entire irradiation period of 480 min in the argon-saturated system. In the aerated system, as can be seen in Fig. 1, an almost total mineralization of 7 104 M BS was achieved within a 480 min
6 5
-3
COD
150
COD/TOC
4
100
3 2
COD/TOC
For analysis, 4 mL samples were taken with a syringe from the reactors. The solid phase of samples, when necessary, was separated by filtration using Millipore Millex-LCR PTFE 0.45 mm filters. The pH of the aqueous phase of the reaction mixture was measured with SEN Tix 41 electrode. The concentration of benzenesulfonate was determined by HPLC, using calibration curves previously prepared. The absorption and emission spectra were recorded with a Specord S 100 diode array spectrophotometer and a PerkinElmer LS50B spectrofluorometer, respectively, using quartz cuvettes of various pathlengths. Mineralization was followed by measuring the total organic carbon (TOC) concentration, with a Thermo Electron Corporation TOC TN 1200 apparatus. The dichromate method was applied for the determination of the chemical oxygen demand (COD). Liquid chromatography-mass spectrometry (HPLC-MS) analyses were performed on an Agilent 1100 Series LC/MSD Trap VL System. The HPLC consisted of an Agilent 1100 binary gradient pump, an Agilent 1100 manual injector valve with a 20 ml loop and a 1100 diode array detector that recorded absorbance in the wavelength range of 190e800 nm. Isocratic separations were carried out at room temperature on a NovaPack C18 column (150 mm 3.9 mm I.D., 5 mm particle diameter) with a flow rate of 0.5 cm3/min. The eluent was composed of 1.5% methanol þ 1.5% acetonitrile þ 96.9% HPLC grade water þ 0.1% HCOOH. The chromatograms were recorded for 15 min. MS was performed with an ion-trap mass spectrometer operated in negative electrospray ionization mode with nebulizer gas pressure of 40 psi, drying gas temperature and flow rate of 305 C and 9 dm3/min, respectively. Anions were recorded from 50 m/z to 500 m/z. Sulfate and sulfite concentrations were determined with ion chromatography, using a Dionex model 2010i apparatus described elsewhere (Horva´th and Hajo´s, 2006). All samples were analyzed in triplicates with a flow rate of 1.7 cm3/min. The separator column (250 mm 4 mm) was based on a 13 mm polystyrenedivinylbenzene copolymer agglomerated with completely aminated anion exchange latex. The ion exchange capacity of the column was 20 mequiv/column. In order not to disturb the subsequent analyses, especially the HPLC-MS experiments, no buffer was used in this system. Thus, the initial pH of the reaction mixtures was the natural one without any adjustment, i.e., pH ¼ 5.0e5.2.
TOC, COD/mg dm
Analytical procedures
200 TOC
50 1 0
0
120
240
360
480
0
Irradiation time/min
B -3
2.3.
A
TOC/mg dm
lmax ¼ 350 nm, i.e., UVA range) was determined by tris(oxalato)ferrate(III) chemical actinometry (Rabek, 1982; Kirk and Namasivayam, 1983). It was estimated to be 1.45 105 E s1.
50 40 30 20 10 0
0
140
280
420
Irradiation time/min Fig. 1 e A) Change of chemical oxygen demand (COD) (B) and total organic carbon (TOC) content (C) during the photocatalytic treatment of an aerated system containing 7 3 10L4 M benzenesulfonate and 1 g dmL3 TiO2, pH [ 5.2. The ratio of COD/TOC (6) is also plotted as the function of irradiation time. B) TOC belonging to the whole system (C), the starting material (B), and the intermediates formed (6) as functions of the irradiation time in the same system.
photocatalytic treatment. The decay curves for TOC and COD are very similar, with gradually decreasing rates, but the change of COD was significantly faster (Fig. 1A). Accordingly, the COD/TOC ratio continuously decreased during the whole irradiation period. It clearly indicates that the oxidation of the original substrate and of the intermediates formed is faster than the formation of hydrogencarbonate or carbon dioxide, i.e., mineralization. Generally, in the first period of oxidation processes oxygenation is the predominant type of reaction, and subsequent cleavage steps result in mineralization. Nevertheless, the TOC in this system perceptibly decreased in the initial stage too (in the first 30 min period of time), indicating that cleavage took place rather early in this process. Then an approximately linear decrease is shown in the interval of 30e240 min. At longer irradiation times (240e480 min) the decrease of TOC gradually slows down due to the consumption of the oxidizable intermediates, approaching total mineralization. The actual concentrations of the starting material were determined from the HPLC-MS signals as shown later. Thus, the TOC values corresponding to the unreacted benzenesulfonate
1620
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 1 7 e1 6 2 8
could also be calculated. Fig. 1B displays the TOC versus time plots belonging to the overall system, the unreacted surfactant (BS), as well as the intermediates formed during the degradation process. The latter curve is the difference of the previous two. The TOC values corresponding to the intermediates show a maximum at 80e90 min where the concentration of the unreacted benzenesulfonate is still significant. At longer times (above 90 min) the TOC representing the intermediates is diminishing because the rate of the mineralization of these species exceeds that of their formation. In this period of irradiation, especially above 120e140 min, the total TOC mostly belongs to the intermediates because the predominant part of BS has already been transformed. Hence, further decrease of TOC can be practically attributed to the mineralization of the intermediates.
3.2.
Change of pH and release of sulfate ions
During the photocatalytic oxidation of BS, especially in the first 180 min, a continuous decrease of pH was observed in both the argon-saturated and the aerated systems (Fig. 2). From the initial value of pH ¼ 5.2, it decreased to pH ¼ 4.5 in the deaerated system, while in the air-saturated reaction mixture down to pH ¼ 3.4. This corresponds to an increase of [Hþ] by about 2.3 105 M in the argon-saturated and 3.9 104 M in the aerated system. Similar values can be observed for the concentration of the sulfate ions formed during the photocatalytic degradation of BS. The concentration of these species increased during the entire irradiation period, displaying a sigmoidal shape. In the case of the air-saturated system, after 8 h irradiation, the concentration of SO2 4 was close to the amount of sulfur that the initial reaction mixture contained. Similar results were observed for the formation of SO2 4 in the case of the photocatalytic degradation of 1,5-naphthalenedisulfonate (Szabo´-Ba´rdos et al., 2008), and also for benzenesulfonate although under different conditions (Sangchakr et al., 1995). This phenomenon may be attributed
5.0
0.6
4.5 0.4 4.0 0.2
0
3.5
0
120
240
360
480
3.0
Irradiation time/min Fig. 2 e Change of SO2L 4 concentration (B,C) and pH (6, :) during the photocatalytic treatment of the system the concentrations of which are given at Fig. 1. The contour symbols stand for the aerated, while the full symbols for the argon-saturated system.
2 þ RSO 3 þ HO /HSO4 þ R 4H þ SO4 þ R
(5)
Carbon oxidation, however, does not significantly contribute to the drop of pH because neither small molecular weight organic acids, which can be formed as intermediates, nor carbon dioxide release appreciable amount of hydrogen ions: eCH] þ 3HO / CO2 þ 2H2O
(6)
Hence, in the subsequent, longer period of irradiation (from 180 min to 480 min) only a very slight increase of the Hþ concentration was observed. In argon-saturated system, in accordance with the change of pH, only a very slow but continuous release of sulfate took place. Even after 8 h irradiation about 2.2 105 M of sulfate was formed indicating that the generation of hydroxyl radicals is over one order of magnitude slower in the absence of dissolved oxygen. Under anaerobic conditions the photogenerated electrons are not captured by O2. In our system BS itself is the most probable candidate for this reaction because its electronscavenging rate constant is high (4 109 M1s1 (Buxton et al., 1988)) and it is readily adsorbed on the surface of the catalyst at pH ¼ 5.0. On the basis of our measurements, at pH ¼ 5.0 the extent of the adsorption of BS is 13% (at 1 g/dm3 TiO2 and 103 mol/dm3 BS). This efficient adsorption can be attributed to the Coulombic attraction between the positively charged surface of the catalyst (pHzpc ¼ 6.5 for anatase (Sun et al., 2005)) and the anionic benzenesulfonate (pK ¼ 2.36 (Faria et al., 2008)). Besides, using ion chromatography, we have ions in the argon-saturated system. Its detected SO2 3 concentration increased during the first 180 min of irradiation. This species is the primary product of electron-scavenging by BS. Sulfite was gradually transformed to SO2 4 at longer irradiation times.
3.3.
5.5
pH
Conc. of SO42-/mM
0.8
to the following reaction involving an HO radical, the main oxidizing agent in this photocatalytic system:
Change of the absorption and emission spectra
The absorption and emission spectra of the solution (after removal of the colloidal catalyst) were recorded for the qualitative monitoring of the chemical change in the system during the photolysis. The absorption spectrum of benzenesulfonate is characterized by strong bands in the 250e270 nm range, assigned to p*)p transitions in the aromatic system. In the argon-saturated system an apparently continuous increase of the original bands can be observed (Fig. 3A). However, this change is accompanied by even stronger increases of the absorbance below 250 nm and above 270 nm with the appearance of a new band at 277 nm and a shoulder at about 305 nm. This phenomenon indicates that new aromatic derivatives were gradually formed over the whole irradiation period. Their absorbance overcompensates the simultaneous decrease of the original bands (superposing on the increasing “baseline”). A similar spectral change is shown for the first (30e50 min) period of the irradiation in the aerated system, indicating the formation of intermediates and the disappearance of the
1621
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 1 7 e1 6 2 8
0.9
0.6
0.3
0.0 225
A
0 20 45 90 180 240 360 420
275
325
Absorbance
0.9
375
0 5 10 20 30 50 90 120 150 180 240
0.3 0.0 225
275
325
320
370
420
60
0 5
50
10 15 40 50
40 30
90 120 150
20 10
375
Wavelength/nm Fig. 3 e Change of the absorption spectrum of the bulk solution during the irradiation of argon-saturated (A) and aerated (B) systems the concentrations of which are given at Fig. 1, ([ [ 2.0 cm). The samples were taken at the indicated times in min.
starting material (Fig. 3B). The increase in the longer-wavelength region suggests that the intermediates formed in this period of time kept the aromatic structure. After the increase of absorption, a gradual decrease can be observed at irradiation times longer than 50 min (thick lines in Fig. 3B). This phenomenon suggests the degradation of the aromatic intermediates formed during the initial period of the photocatalytic process. The emission spectrum of benzenesulfonate displays a strong band at 287 nm. In accordance with the absorption changes, in both argon-saturated and aerated solutions, the luminescence intensity at this band was found to disappear (Fig. 4), indicating the decay of the starting material. Simultaneously, a longer-wavelength band with a shoulder arose and continuously increased in the deoxygenated system. While the position of the stronger band gradually shifted to the red (to 312 nm), the wavelength of the shoulder remained the same (350 nm). This is the effect of the shorter-wavelength (287 nm) emission of the starting material, also causing that the relative intensity of the 312-nm band compared to that of the shoulder gradually decreased. This phenomenon indicates that this new emission may be attributed to at least one intermediate keeping the aromatic structure. HPLC-MS analysis (further explanation follows) indicated that the 312-nm
470
Irradiation time/min
B
0.6
40
0 270
Intensity
1.2
0 10 20 30 45 60 90 120 180 240 300 360
20
Wavelength/nm
B
80 60
Intensity
Absorbance
A
0 270
180
320
370
420
470
Wavelength/nm Fig. 4 e Change of the emission spectrum of the bulk solution during the irradiation of argon-saturated (A) and aerated (B) systems the concentrations of which are given at Fig. 1, (lexc [ 262 nm, [ [ 1.0 cm, slit [ 15 nm). The samples were taken at the indicated times in min.
emission band can be attributed to hydroxy derivatives, while the 350-nm band (shoulder) to dihydroxy species. We have carried out independent experiments with standard compounds and got the same type of emission, confirming the previous assignments. After the first 100 min apparently no change in the characteristic wavelengths of the new emission can be observed, indicating that no further luminescent intermediates were formed during the subsequent period of irradiation, at least not in a significant concentration. In the aerated system, during the first 50-min irradiation a similar spectral change can be observed (thin lines in Fig. 4B), but the maximum intensity reached is lower than in the absence of dissolved oxygen, and the position of the main band is more red-shifted (to 320 nm) at the maximum intensity. Besides, the intensity difference between the emission band at 320 nm and the shoulder at 350 nm is significantly smaller too. In the subsequent period of irradiation, similarly to the corresponding absorption spectral change, the emission intensity continuously decreased in the whole range of wavelength (thick blue lines in Fig. 4B). The band at 320 nm gradually merged into the shoulder at 350 nm, indicating that the latter one belongs to other, new intermediates, the ratio of which increased. This conclusion is confirmed by the emission of these latter intermediates obtained by excitation at a different wavelength, which results in a single-band spectrum deviating
1622
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 1 7 e1 6 2 8
200
100 50
360
410
2.0E+5
2,1 2,7
1.6E+5
Intensity
Intensity
150
0 310
A
0 5 10 20 30 45 60 90 120 150 180 240 300 360
3
1.2E+5 8.0E+4 4.0E+4
460
0.0E+0
Wavelength/ nm Fig. 5 e Change of the emission spectrum of the bulk solution during the irradiation of an aerated system the concentrations in which are given at Fig. 1, (lexc [ 302 nm, [ [ 1.0 cm, slit [ 15 nm). The samples were taken at the indicated times in min.
0
100
200
300
400
Irradiation time/min
B
8.0E+3 2,1 3
from the previous mixed ones (Fig. 5). At this wavelength (302 nm) only the latter intermediates (dihydroxy derivatives as identified by HPLC-MS measurements) are excited. Since no new bands appeared during the decay of the luminescence in the period of 150e480 min, the emitting intermediates were transformed into derivatives which are not aromatic any more, and, hence, not luminescent either.
3.4.
Analysis of the intermediate products
For the detection of intermediates formed during the oxidative degradation of benzenesulfonate, HPLC-MS analyses were also carried out. Identification of the intermediates provides useful hints for the determination of the degradation pathways in this photocatalytic process. The samples for HPLC-MS analysis were taken at various times during the irradiation period. Both argon-saturated and aerated systems were studied by this method. Although most of the intermediates formed during the photocatalytic treatment could not be ideally separated by HPLC, formation and decay of several species were detected by following the intensity of the significant m/z signals as functions of time. It should be mentioned that the absolute intensities of the signals corresponding to various species are generally not comparable because the sensitivity of the detector system with regard to these ions can be significantly different. Hence, these intensity values can rarely be used for quantitative comparative analysis. Nevertheless, the change of the relative concentration of each separate species can be followed by measuring the intensity of the corresponding m/z signal. In the case of the argon-saturated system the predominant species detected, besides the starting material (m/z ¼ 157), belong to the m/z signals 173 and 189. The first value corresponds to the hydroxybenzenesulfonates, while the second one to the dihydroxy derivatives. The formation of these species can be attributed to the reaction of BS with HO radical. As Fig. 6A indicates, three isomers of hydroxybenzenesulfonate were detected at different retention times. A control run with a standard solution of the 4-hydroxy ( para-)
Intensity
6.0E+3
4.0E+3
2.0E+3
0.0E+0 0
100
200
300
Irradiation time/min Fig. 6 e Changes of mass spectrometric signal intensities observed at in the samples taken during the irradiation of argon-saturated system (A) for the m/z [ 173 (hydroxybenzenesulfonates) at 2.1, 2.7, and 3.0 min retention times, and (B) for m/z [ 189 (dihydroxybenzenesulfonates) at 2.1 and 3.0 min retention times. The concentrations of the initial system are given at Fig. 1.
derivative indicated that its retention time was 2.1 min. Thus, the para-hydroxylated isomer gave the smallest signal. The sensitivities of the MS detector for the three hydroxy isomers do not significantly deviate because they are determined by the sulfonate group. Thus, the para-hydroxylated isomer was formed in the lowest concentration. Thus, the other two more intensive signals belong to the ortho- and meta-hydroxylated isomers. Since on a C18 column, at the applied conditions, the retention order is ortho > meta > para (Lake, 2010), according to Fig. 6A, the ortho-hydroxy species was formed in the highest concentration. This observation deviates from earlier results, which indicated the exclusive formation of the para-hydroxy derivative (Sangchakr et al., 1995), and it is in agreement with our previous quantum chemical calculation, which favored the formation of the ortho-hydroxy derivative (Szabo´-Ba´rdos et al., 2008). According to the general rule, however, the electron-withdrawing sulfo group decreases the electron density at the ortho- and para- positions of the aromatic ring
1623
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 1 7 e1 6 2 8
through resonance. Hence the meta- position is favored for an electrophilic attack. Nevertheless, recently, in photocatalytic degradation of benzene with electron-withdrawing groups such as eNO2 and eCN, formation of all the three hydroxy derivatives were observed with o:m:p ratio of 29:34:37 for nitrobenzene, and 45:30:25 for cyanobenzene (Palmisano et al., 2007b, 2007a). For the dihydroxy derivatives only two signals of different retention time were observed, with lower intensities than for the hydroxy isomers (Fig. 6B). The signal at 3.0 min retention time could be assigned to the 2,5-dihydroxy derivative as confirmed by control runs with the standard compound. This seems to be the predominant dihydroxy isomer formed in this process because the intensity of the other signal is more than one order of magnitude lower. The one hydroxy substituent of the 2,5-dihydroxy derivative is at the ortho-position and the other at the meta-. The predominant formation of this isomer is in accordance with the ratio of the hydroxy derivatives (ortho > meta > para) observed in this
A
9.0E+4 2.1
system and the theoretical considerations favorizing hydroxylation at ortho and meta positions. The concentration of all of these intermediates monotonously increased, although at different rates, during the irradiation of the argonsaturated reaction mixture. In the aerated system both the formation and decay of these intermediates were observed (Fig. 7). In this case the maximum concentrations of the hydroxy derivatives are significantly lower than those in the argon-saturated system (Fig. 7A) as a consequence of their fast transformation to dihydroxy and other intermediates. Hence, the formation of dihydroxy species is much more efficient in the aerated mixture, and their maximum concentrations are about one order of magnitude higher than the corresponding values in the deoxygenated system (Fig. 7B). Fig. 8A summarizes the decay of the starting material (benzenesulfonate, m/z ¼ 157), and the formation and decay of the hydroxy (m/z ¼ 173), dihydroxy (m/z ¼ 189), and trihydroxy (m/z ¼ 205) derivatives. The values of the signals belonging to m/z ¼ 173, m/z ¼ 189, and m/z ¼ 205 are cumulative intensities, i.e., the corresponding isomers are not distinguished here. In
Intensity
2.7 3
6.0E+4
A
157 173
2.0E+5
Intensity
3.0E+4
0.0E+0 0
100
200
300
189 205
1.5E+5 1.0E+5 5.0E+4
Irradiation time/min
B
2.5E+5
1.2E+5
0.0E+0
2.1
0
Intensity
3
100
200
300
Irradiation time/min
8.0E+4
B
10000 207 239
7500
Intensity
4.0E+4
0.0E+0 0
100
200
300
273 193
5000
2500
Irradiation time/min Fig. 7 e Changes of mass spectrometric signal intensities observed in the samples taken during the irradiation of aerated system (A) for the m/z [ 173 (hydroxybenzenesulfonates) at 2.1, 2.7, and 3.0 min retention times, and (B) for m/z [ 189 (dihydroxybenzenesulfonates) at 2.1 and 3.0 min retention times. The concentrations of the initial system are given at Fig. 1.
0
0
60
120
180
240
300
Irradiation time/min Fig. 8 e Changes of the mass spectrometric signal intensities at various m/z values in the samples taken during the irradiation of the aerated system the concentrations of which are given at Fig. 1.
1624
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 1 7 e1 6 2 8
accordance with the absorption and emission data, the predominant part (over 90e95%) of the original surfactant was converted during the first 120-min treatment. As shown in Fig. 8A, the maximum concentration of the hydroxy derivatives is at about 50 min, while that of the dihydroxy intermediates at about 90 min. Further oxygenation resulted in the formation of trihydroxy derivatives with m/z value of 205 (Fig. 8A). Their maximum concentrations appeared at about 110e120 min. These intermediates still retain their aromaticity. These maximum positions in time are in good accordance with the changes of the absorption and emission spectra, confirming that aromatic luminescent derivatives were formed. With maximum concentration at about 120e150 min species with m/z ¼ 207 and 239 were also detected (Fig. 8B). According to the absorption and emission spectra in this period of time, these compounds are not aromatic, moreover they can be ring-opened intermediates, such as aldehydes and carboxylic acids. The transformation of these intermediates involves gradual hydroxylation (and saturation of the C]C bond due to addition of HO radicals) and shortening of the carbon-chain as indicated by the m/z values of 273 and 193, respectively (Fig. 8B). Desulfonation takes place with the intermediate of m/z ¼ 273 as suggested by the m/z value of 209 (Fig. 9). The intermediate of shorter carbon-chain (m/z ¼ 193) undergoes gradual hydroxylation (and saturation of the C]C bond) too and oxidation of the terminal carbon atom (m/z ¼ 243). Desulfonation and saturation of the latter species (m/z ¼ 243) is suggested by the detection of the intermediate of m/z ¼ 179. However, decarboxylation of the intermediate with longer carbon-chain (m/z ¼ 209) leads also to the species of m/z ¼ 179. These results indicate that, depending on the order of chainshortening, desulfonation, saturation, and hydroxylation of the ring-opened intermediates, degradation can take place via different parallel pathways. This conclusion is confirmed by the fact that for several saturated intermediates the m/z values of the corresponding unsaturated species can also be found in the mass spectrum, and vice versa (not shown in Figs. 8 and 9).
7000
209 243 179 123 89
6000
Intensity
5000 4000 3000 2000 1000 0 0
120
240
360
480
Irradiation time/min Fig. 9 e Changes of the mass spectrometric signal intensities at various m/z values in the samples taken during the irradiation of aerated system the concentrations of which are given at Fig. 1.
Notably, besides the addition of HO radicals, the saturation can be the consequence of the reaction with H atoms, which are formed via scavenging of the photogenerated electrons by hydrogen ions. The low pH of the system after 120e150 min of irradiation (about 3.5, see Fig. 2), and the high value of the electron-scavenging rate constant of Hþ (2.8 1010 M1s1, (Buxton et al., 1988)) confirm the formation of nascent hydrogen. Further oxidation of the intermediate of m/z ¼ 179 is accompanied by shortening of the aliphatic chain (m/z ¼ 123). During the progress of mineralization, generation of intermediates of lower m/z values was observed, such as the formation of oxalate (m/z ¼ 89). This conclusion is confirmed by the observation in the case of degradation of BS by ozonation and glow discharge electrolysis, where oxalic and maleic acids were detected (Faria et al., 2008; Amano et al., 2004; Amano and Tezuka, 2006). The efficient photocatalytic oxidation of oxalic acid on colloidal titanium dioxide was studied earlier in our group (Szabo´-Ba´rdos et al., 2003, 2004). The final products of the total mineralization are water, carbon dioxide, and sulfate ions. Scheme 1 tentatively summarizes the possible degradation pathways of BS, which were compiled by taking into consideration the succession of formation and decay of the detected intermediates. Although this is not a detailed mechanism, it contains most of the key steps in the degradation of BS. Similar steps were proposed in the oxidative degradation pathways of p-toluenesulfonic acid treated with thermally activated hydrogen peroxide (Sto¨ffler and Luft, 1999).
3.5.
The role of dissolved oxygen
As it can be seen in the previous sections, the presence of oxygen significantly affects both the rate and the pathways of the degradation of BS. Oxygen increases the initial formation rate of the meta- and para-hydroxylated derivatives, and dramatically (at least with one order of magnitude) enhances the formation rate of the dihydroxy species (see Figs. 6 and 7). This phenomenon can be partly attributed to the electronscavenging effect of dissolved oxygen, by which it hinders the recombination of the electronehole pair on the surface of the photocatalyst (Schwitzgebel et al., 1995; Dionysiou et al., 2002). Thus, it can accelerate the formation of hydroxyl radicals via oxidation of H2O or HO with hþ vb (see Eq. (2)). Besides, O2 formed via electron-scavenging by dissolved oxygen (see Eq. (3)) can also take part in oxidation of both the starting material and the intermediates in this system. (Notably, O2 2 , which can also be formed in the presence of oxygen (see Eq. (4)), cannot directly oxidize benzenesulfonate in a thermal reaction, as indicated by our blind probes.) For the reaction with a conduction band electron, adsorption of oxygen on the surface of the catalyst particles is an important step. Since the holes react much faster with organic species in the solution or solvent molecules than electrons do with oxygen, the rate of oxygen reduction by the conduction band electron is usually the rate-limiting step in the photocatalytic process. Accordingly, the Langmuir-type of tendency of the photocatalytic rates on oxygen concentration, also called as Langmuir-Hinselwood (L-H) mechanism has been observed (Dionysiou et al., 2002). Although oxygen appears to
1625
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 1 7 e1 6 2 8
SO3
SO 3
SO3
SO 3
OH
OH
OH
HO
OH
173
157 SO 3 -
OH HO
COOCOOH
HO
HO
HO
209
OH
HO
HO
COOH
HO
COOH
CHO
273
HO
123
207
COOH CHO
193
OH
SO42-, CO2, H2O
O
OH
OH
SO 3 -
243 O
239
OH
COOH
O
-
COOH
179 OH
SO3
CHO
COOH
HO
-
COOH
SO 3 -
COO-
OH
SO3
205
COOH
OH
OH
OH
189
O
89
Scheme 1 e A possible degradation pathway of benzenesulfonate in photocatalytic oxidation. The numbers indicate the m/z values of the corresponding species.
be weakly adsorbed on TiO2, it does not compete with the organic contaminants for adsorption sites because it adsorbs at Ti3þ sites, whereas hydroxide ions and organic pollutants adsorb at Ti4þ-lattice oxygen sites (Mills et al., 2006). In the aerated system the decay of the hydroxy and dihydroxy derivatives is strongly promoted via further oxidation steps leading to ring cleavage. On the contrary, no ring cleavage could be achieved in argon-saturated reaction mixture. We also tried to achieve ring fission in the case of pyrogallol (benzene-1,2,3-triol) in the same type of deaerated photocatalytic system, but its aromatic ring could not be cleaved in spite of the trihydroxylated structure. In the aerated system, however, it can be easily degraded, even without photocatalysis as it is well-known (Marklund and Marklund, 1974). These observations clearly indicate that the presence of oxygen is indispensable for the cleavage of the aromatic ring. They also suggest that hydroxyl radicals alone are not able to cleave the aromatic ring. This can be achieved by the attack of other oxidizing agents formed in situ from dissolved oxygen, such as for example O2 /HO2 and/or 1O2 (singlet excited-state dioxygen). The superoxide radical anion can react via a single electron transfer mechanism, while 1O2 via the formation of the dioxetane intermediate (Wahab et al., 2008). Although transient intermediates cannot be detected by the HPLC/MS technique, it is well-known that the reaction of
hydroxyl radicals with organic substrates generally leads to the formation of carbon-centered radicals (Cooper et al., 2009). In aerated solution these species can react with dissolved oxygen to form peroxyl radicals, the fate of which in water is very complex. In several cases they undergo self-reaction to form tetroxide intermediates, which decompose to form a variety of products. In order to demonstrate the crucial role of dissolved oxygen in the photocatalytic degradation of BS in this system irradiation was accompanied by alternating bubbling with argon and air. Monitoring the change of TOC and pH as the functions of irradiation time clearly shows the importance of O2 for the mineralization process (Fig. 10). During the first (120-min) irradiation period in the argonsaturated system the pH significantly decreased, due to the reaction between the sulfonate groups and the hydroxyl radicals (see Eq. (5)), while no change of TOC was observed, indicating that no mineralization, i.e., formation of CO2 or HCO 3 occurred. Introducing air into the system resulted in a further but gradually slower decrease of pH, reaching an almost constant value at the end of this 180-min period. This can be attributed to the total desulfonation of BS. In the presence of dissolved oxygen TOC decreased relatively fast as a consequence of efficient mineralization steps involving ring cleavage. Changing back to argon, the decrease of TOC
1626
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 1 7 e1 6 2 8
60
7
detected by mass spectrometry, a tentative scheme including possible pathways of degradation was also compiled. Our results can contribute to the development of photocatalytic procedures for wastewater treatment.
6 40 30
5
pH
TOC /mg dm-3
50
20 4 10 0 0
120
240
360
3 480
Irradiation time/min Fig. 10 e Change of total organic carbon (TOC) content (C, B) and pH (:, 6) during the photocatalytic treatment of a system the concentrations of which are given at Fig. 1. Bubbling with argon and air were alternated. Full symbols stand for the argon-bubbling periods, while contour symbols for the air-bubbling periods.
became slower due to the decreasing concentration and the final absence of dissolved oxygen. The slight decrease of TOC in the second half of this 120-min period of argon-bubbling can be attributed to the mineralization of open-chain intermediates formed in the previous stage. Finally, air was led into the system again, instantly causing a further, significantly fast decrease of TOC. Thus, the crucial role of dissolved oxygen in this photocatalytic mineralization could be unambiguously observed through this experiment.
4.
Conclusions
According to our experimental results gained by various analytical methods, in the degradation of BS the initial step is the hydroxylation of the starting compound, leading to the production of hydroxy then dihydroxy derivatives, along with desulfonation in both the presence and the absence of dissolved oxygen. HPLC-MS measurements indicated that the formation of ortho- and meta-hydroxy derivatives is favored in this system, which is in accordance with our earlier quantum chemical calculations. As a consequence, and also in agreement with theoretical considerations, 2,5-dihydroxybenzenesulfonate is the predominant intermediate formed in the second step. While in the argon-saturated system no decay of the hydroxy derivatives was observed, in the presence of oxygen, gradual oxidation steps resulted in the destruction of the aromatic system, followed by ring cleavage producing aldehydes and carboxylic acids. Efficient desulfonation took place both before and after the ring opening. The importance of dissolved oxygen for the ring fission, and, thus, for the complete mineralization, has been unequivocally proved in this system, suggesting the crucial role of O2 /HO2 and/or 1O2 in this process. Considering the intermediates
Acknowledgments This work was supported by the National Development ´ MOP 4.2.2.-08/1/2008-0018, Livable environment Agency (TA and healthier people e Bioinnovation and Green Technology research at the University of Pannonia, the project is being cofinanced by the European Social Fund with the support of the European Union). The authors thank Prof. Pe´ter Hajo´s and Dr. Krisztia´n Horva´th for the determination of sulfate and sulfite with ion chromatography.
references
Alonso, M.C., Castillo, M., Barcelo, D., 1999. Solid-phase extraction procedure of polar benzene- and naphthalenesulfonates in industrial effluents followed by unequivocal determination with ion-pair chromatography/electrospray-mass spectrometry. Analytical Chemistry 71 (13), 2586e2593. Alonso, M.C., Tirapu, L., Ginebreda, A., Barcelo´, D., 2005. Monitoring and toxicity of sulfonated derivatives of benzene and naphthalene in municipal sewage treatment plants. Environmental Pollution 137 (2), 253e262. Amano, R., Tomizawa, S., Tezuka, M., 2004. Mineralization of aqueous benzenesulfonates by contact glow discharge electrolysis. Electrochemistry 72 (12), 836e838. Amano, R., Tezuka, M., 2006. Mineralization of alkylbenzenesulfonates in water by means of contact glow discharge electrolysis. Water Research 40 (9), 1857e1863. Brilon, C., Beckmann, W., Knackmuss, H.-J., 1981. Catabolism of naphthalenesulfonic acids by Pseudomonas sp. A3 and Pseudomonas sp. C22. Applied and Environmental Microbiology 42 (1), 44e55. Buxton, G.V., Greenstock, C.L., Helman, W.P., Ross, A.B., 1988. Critical review of rate constants for reactions of hydrated electrons, hydrogen atoms and hydroxyl radicals ($OH/$O ) in aqueous solution. Journal of Physical Chemistry Reference Data 17 (2), 513e886. Cain, R.B., 1981. In: Leisinger, T., Cook, A.M., Hutter, R., Nuesch, J. (Eds.), Microbial Metabolism of Xenobiotics and Recalcitrant Compounds. Academic Press, London, pp. 325e370. Cooper, W.J., Cramer, C.J., Martin, N.H., Mezyk, S.P., O’Shea, K.E., von Sonntag, C., 2009. Free radical mechanisms for the treatment of methyl tert-butyl ether (MTBE) via advanced oxidation/reductive processes in aqueous solutions. Chemical Reviews 109 (3), 1302e1345. Devipriya, S., Yesodharan, S., 2005. Photocatalytic degradation of pesticide contaminants in water. Solar Energy Matererials and Solar Cells 86 (3), 309e348. Dionysiou, D.D., Burbano, A.A., Suidan, M.T., Baudin, I., Laıˆne´, J.-M., 2002. Effect of oxygen in a thin-film rotating disk photocatalytic reactor. Environmental Science and Technology 36 (17), 3834e3843. http://pubs.acs.org/action/ showMultipleAbstracts-es0113605AF1#es0113605AF1. Fabbri, D., Bianco Prevot, A., Pramauro, E., 2006. Effect of surfactant microstructures on photocatalytic degradation of phenol and chlorophenols. Applied Catalysis B: Environmental 62 (1e2), 21e27.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 1 7 e1 6 2 8
´ rfa˜o, J.J.M., Pereira, M.F.R., 2008. Catalytic Faria, P.C.C., O ozonation of sulfonated aromatic compounds in the presence of activated carbon. Applied Catalysis B: Environmental 83 (1e2), 150e159. Gkika, E., Troupis, A., Hiskia, A., Papaconstantinou, E., 2006. Photocatalytic reduction of chromium and oxidation of organics by polyoxometalates. Applied Catalysis B: Environmental 62 (1e2), 28e34. Greim, H., Ahlers, J., Bias, R., Broecker, B., Hollander, H., Gelbke, H. P., Klimisch, H.J., Mangelsdorf, I., Paetz, A., Schong, N., Stropp, G., Vogel, R., Weber, C., Ziegler-Skylakakis, K., Bayer, E., 1994. Toxicity and ecotoxicity of sulfonic acids: structure activity relationship. Chemosphere 28 (12), 2203e2236. Hashim, M.A., Kulandai, J., Hassan, R.S., 1992. Biodegradability of branched alkybenzene sulfonates. Journal of Chemical Technology and Biotechnology 54 (3), 207e214. Hegyi, J., Horva´th, O., 2004. Photocatalytic reduction of mercury (II) and simultaneous oxidative degradation of surfactants in titanium dioxide suspension. Progress in Colloid and Polymer Science 125, 10e16. Hoffmann, M.R., Martin, S.T., Choi, W., Bahnemann, D.W., 1995. Environmental applications of semiconductor photocatalysis. Chemical Reviews 95 (1), 69e96. Horva´th, K., Hajo´s, P., 2006. Retention profiles and mechanism of anion separation on latex based pellicular ion exchanger in ion chromatography. Journal of Chromatography, A 1104 (1e2), 75e81. Horva´th, O., Husza´nk, R., 2003. Degradation of surfactants by hydroxyl radicals photogenerated from hydroxoiron(III) complexes. Photochemical and Photobiological Sciences 2 (9), 960e966. Horva´th, O., Bodna´r, E., Hegyi, J., 2005. Photoassisted oxidative degradation of surfactants and simultaneous reduction of metals in titanium dioxide dispersions. Colloids and Surfaces A: Physicochemical and Engineering Aspects 265 (1e3), 135e140. Integral Consulting Inc., 2007. Toxicological Profiles for Three Organic Acids. ICI, Annapolis, MD, pp. 3e1 e 3e4. Kajitvichyanukul, P., Ananpattarachai, J., Pongpom, S., 2005. Solegel preparation and properties study of TiO2 thin film for photocatalytic reduction of chromium(VI) in photocatalysis process. Science and Technology of Advanced Materials 6 (3e4), 352e358. Kanki, T., Yoneda, H., Sano, N., Toyoda, A., Nagai, C., 2004. Photocatalytic reduction and deposition of metallic ions in aqueous phase. Chemical Engineering Journal 97 (1), 77e81. Kirk, A.D., Namasivayam, C., 1983. Errors in ferrioxalate actinometry. Analytical Chemistry 55 (14), 2428e2429. Konstantinou, K.I., Albanis, T.A., 2003. Photocatalytic transformation of pesticides in aqueous titanium dioxide suspensions using artificial and solar light: intermediates and degradation pathways. Applied Catalysis B: Environmental 42 (4), 319e335. Lake, R., 2010. Optimized RP-HPLC Method for Hydroxybenzoic Acids. Restek Pharmaceutical Article. http://www.restek.com/ aoi_pharm_A009.asp. Lange, F.T., Furrer, R., Brauch, H.J., 2000. Polar aromatic sulfonates and their relevance to waterworks. Nieuwegein, Association of River Waterworks e RIWA, 21e60. Marklund, S., Marklund, G., 1974. Involvement of the superoxide anion radical in the autoxidation of pyrogallol and a convenient assay for superoxide dismutase. European Journal of Biochemistry 47, 469e474. Matsushita, M., Tran, T.H., Nosaka, A.Y., Nosaka, Y., 2007. Photooxidation mechanism of L-alanine in TiO2 photocatalytic systems as studied by proton NMR spectroscopy. Catalysis Today 120 (2), 240e244. Mills, A., Wang, J., Ollis, D.F., 2006. Dependence of the kinetics of liquid-phase photocatalyzed reactionson oxygen
1627
concentration and light intensity. Journal of Catalysis 243 (1), 1e6. Palmisano, G., Addamo, M., Augugliaro, V., Caronna, T., Paola, A. D., Lopez, E.G., Loddo, V., Marci, G., Palmisano, L., Schiavello, M., 2007a. Selectivity of hydroxyl radical in the partial oxidation of aromatic compounds in heterogeneous photocatalysis. Catalysis Today 122 (1e2), 118e127. Palmisano, G., Loddo, V., Augugliaro, V., Palmisano, L., 2007b. Photocatalytic oxidation of nitrobenzene and phenylamine: pathways and kinetics. AIChE 53 (4), 961e968. Patsoura, A., Kondarides, D.I., Verykios, E.X., 2007. Photocatalytic degradation of organic pollutants with simultaneous production of hydrogen. Catalysis Today 124 (3e4), 94e102. Rabek, J.F., 1982. Experimental Methods in Photochemistry and Photophysics. Wiley-Interscience Publication, John Wiley & Sons Ltd., New York, pp. 944e6. Ravera, M., Ciccarelli, C., Gianotti, V., Scorza, S., Osella, D., 2004. Electro-assisted methods for waste destruction: silver(II) and peroxydisulfate reagents in the electrochemically mediated oxidation of polyaromatic sulfonates. Chemosphere 57 (7), 587e594. Ravera, M., Buico, A., Gosetti, F., Cassino, C., Musso, D., Osella, D., 2009. Oxidative degradation of 1,5-naphthalenedisulfonic acid in aqueous solutions by microwave irradiation in the presence of H2O2. Chemosphere 74 (10), 1309e1314. Ravera, M., Musso, D., Gosetti, F., Cassino, C., Gamalero, E., Osella, D., 2010. Oxidative degradation of 1,5naphthalenedisulfonic acid in aqueous solutions: noncatalytic remediation by UV-photolysis in the absence and presence of H2O2. Chemosphere 79 (2), 144e148. Reemtsma, T., 1996. Methods of analysis of polar aromatic sulfonates from aquatic environments. Journal of Chromatography A 733 (1e2), 473e489. Riediker, S., Suter, M.J.F., Giger, W., 2000a. Benzene- and naphthalenesulfonates in leachates and plumes of landfills. Water Research 34 (7), 2069e2079. Riediker, S., Ruckstuhl, S., Suter, M.J.F., Cook, A.M., Giger, W., 2000b. Toluenesulfonate in landfill leachates: leachability from foundry sands and aerobic biodegradation. Environmental Science and Technology 34 (11), 2156e2161. Sangchakr, B., Hisanaga, T., Tanaka, K., 1995. Photocatalytic degradation of sulfonated aromatics in aqueous TiO2 suspension. Journal of Photochemistry and Photobiology A: Chemistry 85 (1e2), 187e190. Schwitzgebel, J., Ekerdt, J.G., Gerischer, H., Heller, A., 1995. Role of the oxygen molecule and of the photogenerated electron in TiO2-photocatalyzed air oxidation reactions. Journal of Physical Chemistry 99 (15), 5633e5638. Sto¨ffler, B., Luft, G., 1999. Oxidative degradation of p-toluenesulfonic acid using hydrogen peroxide. Chemosphere 38 (5), 1035e1047. Sun, B., Reddy, E.P., Smirniotis, P.G., 2005. Visible light Cr (VI) reduction and organic chemical oxidation by TiO2 photocatalysis. Environmental Science and Technology 39 (16), 6251e6259. Suter, M.J.F., Riediker, S., Giger, W., 1999. Selective determination of aromatic sulfonates in landfill leachates and groundwater using microbore liquid chromatography coupled with mass spectrometry. Analytical Chemistry 71 (4), 897e904. Szabo´-Ba´rdos, E., Czili, H., Horva´th, A., 2003. Photocatalytic oxidation of oxalic acid enhanced by silver deposition on a TiO2 surface. Journal of Photochemistry and Photobiology A: Chemistry 154 (2e3), 195e201. Szabo´-Ba´rdos, E., Czili, H., Megyery-Balog, K., Horva´th, A., 2004. Photocatalytic oxidation of oxalic acid enhanced by silver and copper deposition on TiO2 surface. Progress in Colloid and Polymer Science 125, 42e48.
1628
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 1 7 e1 6 2 8
Szabo´-Ba´rdos, E., Pe´terva´ri, E., El-Zein, V., Horva´th, A., 2006. Photocatalytic decomposition of aspartic acid over bare and silver deposited TiO2. Journal of Photochemistry and Photobiology A: Chemistry 184 (1e2), 221e227. Szabo´-Ba´rdos, E., Zsila´k, Z., Lendvay, G., Horva´th, O., Markovics, O., } , N., 2008. Photocatalytic degradation of Hoffer, A., To¨ro 1,5-naphthalenedisulfonate on colloidal titanium dioxide. Journal of Physical Chemistry B 112 (46), 14500e14508. Takada, H., Ishiwatari, R., 1990. Biodegradation experiments of linear alkylbenzenes (LABs): isomeric composition LABs as an indicator of the degree of LAB degradation in the aquatic environment. Environmental Science and Technology 24 (1), 86e91. Tully, P.S., 1997. In: Grant, M.H. (Ed.)Kirk-Othmer Encyclopedia of Chemical Technology, fourth ed., vol. 23. JohnWiley & Sons, New York, p. 194. Wahab, H.S., Bredow, T., Aliwiand, S.M., 2008. Computational investigation of the adsorption and photocleavage of chlorobenzene on anatase TiO2 surfaces. Chemical Physics
353 (1e3), 93e103. Wittich, R., Rast, H.G., Knackmuss, H.-J., 1988. Degradation of naphthalene-2,6- and naphthalene-1,6-disulfonic acid by a Moraxella sp. Applied and Environmental Microbiology 54 (7), 1842e1844. Zerbinati, O., Diana, I., Baiocchi, C., 1999. Trace naphthalenesulfonates determination in natural water sample. International Journal of Environmental Analytical Chemistry 74 (1e4), 43e54. Zhang, T., Oyama, T., Horikoshi, S., Zhao, J., Serpone, N., Hidaka, H., 2003. Photocatalytic decomposition of the sodium dodecylbenzene sulfonate surfactant in aqueous titania suspensions exposed to highly concentrated solar radiation and effects of additives. Applied Catalysis B: Environmental 42 (1), 13e24. Zu¨rrer, D., Cook, A.M., Leisinger, Th., 1987. Microbial desulfonation of substituted naphthalenesulfonic acids and benzenesulfonic acids. Applied and Environmental Microbiology 53 (7), 1459e1463.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 2 9 e1 6 3 8
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Accelerated biodegradation of pyrene and benzo[a]pyrene in the Phragmites australis rhizosphere by bacteriaeroot exudate interactions Tadashi Toyama a,*, Tetsuya Furukawa b, Noritaka Maeda c, Daisuke Inoue b, Kazunari Sei b, Kazuhiro Mori a, Shintaro Kikuchi c, Michihiko Ike b a
Department of Research, Interdisciplinary Graduate School of Medicine and Engineering, University of Yamanashi, 4-3-11 Takeda, Kofu, Yamanashi 400-8511, Japan b Division of Sustainable Energy and Environmental Engineering, Osaka University, 2-1 Yamadaoka, Suita, Osaka 565-0871, Japan c Division of Applied Sciences, Muroran Institute of Technology, 27-1 Mizumoto, Muroran 050-8585, Japan
article info
abstract
Article history:
We investigated the biodegradation of pyrene and benzo[a]pyrene in Phragmites australis
Received 12 January 2010
rhizosphere sediment. We collected P. australis plants, rhizosphere sediments, and unve-
Received in revised form
getated sediments from natural aquatic sites and conducted degradation experiments
6 September 2010
using sediments spiked with pyrene or benzo[a]pyrene. Accelerated removal of pyrene and
Accepted 30 November 2010
benzo[a]pyrene was observed in P. australis rhizosphere sediments with plants, whereas
Available online 15 December 2010
both compounds persisted in unvegetated sediments without plants and in autoclaved rhizosphere sediments with sterilized plants, suggesting that the accelerated removal
Keywords:
resulted largely from biodegradation by rhizosphere bacteria. Initial densities of pyrene-
Phragmites australis
utilizing bacteria were substantially higher in the rhizosphere than in unvegetated sedi-
Rhizosphere
ments, but benzo[a]pyrene-utilizing bacteria were not detected in rhizosphere sediments.
Pyrene
Mycobacterium gilvum strains isolated from rhizosphere sediments utilized pyrene aerobi-
Benzo[a]pyrene
cally as a sole carbon source and were able to degrade benzo[a]pyrene when induced with
Accelerated biodegradation
pyrene. Phragmites australis root exudates containing phenolic compounds supported
Root exudates
growth as a carbon source for the one Mycobacterium strain tested, and induced benzo[a] pyrene-degrading activity of the strain. The stimulatory effect on benzo[a]pyrene biodegradation and the amounts of phenolic compounds in root exudates increased when P. australis was exposed to pyrene. Our results show that Mycobacteriumeroot exudate interactions can accelerate biodegradation of pyrene and benzo[a]pyrene in P. australis rhizosphere sediments. ª 2010 Elsevier Ltd. All rights reserved.
1.
Introduction
Polycyclic aromatic hydrocarbons (PAHs) are ubiquitous pollutants in aquatic environments. PAHs are generated from anthropogenic activities such as industrial processing,
spillage of petroleum, and incomplete combustion of fossil fuels. In aquatic environments, PAHs accumulate in sediments because of their hydrophobicity (Juhasz and Naidu, 2000; Chen et al., 2006; Liang et al., 2007). In particular, high molecular weight (HMW) PAHs, such as pyrene and benzo[a]
* Corresponding author. Tel./fax: þ81 55 220 8346. E-mail address:
[email protected] (T. Toyama). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.11.044
1630
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 2 9 e1 6 3 8
pyrene, are resistant to biodegradation, persist in the environment, and show toxic, mutagenic, and carcinogenic effects in humans and in animals in the wild (Cerniglia, 1992; Menzie et al., 1992; Juhasz and Naidu, 2000; Kanaly and Harayama, 2000). Thus, there is an acute need for removal of HMWPAHs from contaminated sediments to reduce HMW-PAH risks to human health and the aquatic environment. Phytoremediationdthe use of plants to enhance biodegradation and removal of pollutantsdis a cost-effective and environmentally friendly remediation technology. Previous studies have revealed that various types of terrestrial plants promote HMW-PAH degradation in soils by stimulating bacterial metabolism (Aprill and Sims, 1990; Reilley et al., 1996; Banks et al., 1999; Liste and Alexander, 2000; Yoshitomi and Shann, 2001; Mueller and Shann, 2007; Yi and Crowley, 2007; Fan et al., 2008). However, little is known about the potential of aquatic plants to promote the biodegradation of HMW-PAHs in sediments. There is only one report (Jouanneau et al., 2005) of the accelerated biodegradation of pyrene in the rhizosphere sediment of the aquatic plant Phragmites australis (reed). Because efficient and rapid PAH biodegradation in the environment generally depends on molecular oxygen availability (Cerniglia, 1992; Chung and King, 1999; Juhasz and Naidu, 2000), the accelerated biodegradation of HMW-PAHs in the rhizosphere can be attributed to the release of oxygen to the rhizosphere by plant roots (rhizosphere oxygenation), and it has been confirmed that oxygen released by P. australis roots promotes pyrene biodegradation in the rhizosphere (Jouanneau et al., 2005). Another possible important mechanism for the accelerated biodegradation of HMW-PAHs in the rhizosphere is through stimulation of bacterial metabolism by plant root exudates. Previous studies have clearly shown that the root exudates of some terrestrial plants stimulate biodegradation of organic chemicals, including phenanthrene (Miya and Firestone, 2000, 2001), pyrene (Yoshitomi and Shann, 2001; Mueller and Shann, 2007; Yi and Crowley, 2007), benzo[a]pyrene (Rentz et al., 2005; Yi and Crowley, 2007), and polychlorinated biphenyls (PCBs) (Donnelly et al., 1994; Gilbert and Crowley, 1997; Leigh et al., 2002). Some phenolic compounds released by plants might enhance biodegradation of PCBs by serving as growth substrates for PCB-degrading bacteria (Donnelly et al., 1994; Leigh et al., 2002) and inducers of PCB-degrading enzymes (Gilbert and Crowley, 1997). We have recently found that the root exudates of another aquatic plant, Spirodela polyrrhiza (giant duckweed), contain highly concentrated and diverse phenolic compounds that contribute markedly to the accelerated biodegradation of simple aromatic compounds in the rhizosphere (Toyama et al., 2009b), suggesting that the exudate-mediated enhancement of biodegradation of chemicals is possible not only with terrestrial plants but also with aquatic plants. However, the effects of aquatic plant root exudates on HMWPAH biodegradation have not been investigated. Identification of the mechanisms underlying the accelerated biodegradation of HMW-PAHs in the rhizosphere of aquatic plantsdespecially the relationship between the accelerated biodegradation and phenolic root exudates of aquatic plantsdis essential for developing efficient use of aquatic plants for remediation of HMW-PAH-contaminated sediments.
Our objectives in this study were (i) to verify the accelerated biodegradation of pyrene and benzo[a]pyrene in natural rhizosphere sediment of P. australis and (ii) to clarify the mechanisms of this accelerated biodegradation. We followed the degradation of pyrene and benzo[a]pyrene in rhizosphere and non-rhizosphere (i.e., unvegetated) sediments spiked with pyrene or benzo[a]pyrene. We isolated pyrene-degrading Mycobacterium gilvum strains from the P. australis rhizosphere sediments, analyzed the chemical nature of P. australis root exudates, and investigated the effects of the root exudates on the biodegradation of pyrene and benzo[a]pyrene. Our results provide direct experimental evidence that P. australis root exudates stimulate biodegradation of pyrene and benzo[a] pyrene by pyrene-degrading Mycobacterium in the rhizosphere.
2.
Materials and methods
2.1.
Chemicals
Pyrene was purchased from Wako Pure Chemical Industries (Osaka, Japan). Benzo[a]pyrene was purchased from AccuStandard Inc. (New Haven, Connecticut, USA). N,O-bis(trimethylsilyl)trifluoroacetamide (BSTFA) was purchased from Tokyo Chemical Industry (Tokyo, Japan).
2.2.
Plant, sediment, and water samples
Phragmites australis plant samples and sediment and surfacewater samples were obtained from 3 freshwater sites in Japan: Inukai Pond (Yamadaoka, Suita) and Yodo River (Juso, Osaka) in Osaka prefecture, and Kamo River (Shijokawaramachi, Kyoto) in Kyoto prefecture. Young P. australis plants (4e6 leaves, 40e60 cm tall) were collected from within the P. australis vegetation zone, and rhizosphere sediment samples were collected from a depth of 0e20 cm around their roots. Sediment samples from unvegetated areas (“unvegetated sediment” samples) were collected at the same depth, but at least 1 m from the nearest P. australis plant. The unvegetated sediment samples did not have root materials. Water samples were collected from the surface around the plants. To obtain sterile (bacteria-free) plants, P. australis seeds were sterilized by a 1-min wash in 70% ethanol and a 5-min wash in sodium hypochlorite solution (5% available chorine), rinsed twice with autoclaved deionized water, and germinated on sterile Hoagland’s nutrient medium (Toyama et al., 2006) solidified with 0.25% (w:v) gellan gum. Each young plant was aseptically transferred to a flask containing 200 mL of sterile Hoagland’s nutrient solution and maintained in an incubation chamber (30 C, 10,000 lx, 16:8 h light:dark).
2.3.
Culture media
Basal salts medium (BSM) (Toyama et al., 2009a) containing pyrene (PYR-BSM) or benzo[a]pyrene (BaP-BSM) as the sole carbon source was used for bacterial culture and degradation tests. Bacterial isolates were routinely maintained on a 1/10 dilution of Luria-Bertani (LB) medium. Agar solid medium was prepared with 2.0% (w:v) agar.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 2 9 e1 6 3 8
1631
2.4. Pyrene and benzo[a]pyrene degradation experiments in sediment microcosms
2.5. Enrichment, isolation, and identification of pyrenedegrading bacteria
To check for potential HMW-PAH degradation in the natural rhizosphere sediments of P. australis, degradation experiments were performed in microcosms as follows. Sediment samples (equal to about 100 g dry-weight) were placed in 150mL vials. Pyrene or benzo[a]pyrene dissolved in n-hexane was added to each vial to a final concentration of 50 mg (kg dry sediment)1, and the n-hexane was allowed to evaporate. The vial was then shaken for 3 h at room temperature. Next, a P. australis plant was replanted in a vial containing rhizosphere sediment from the same site as the plant. Finally, 10 mL of surface-water from the same site as the sediment and plant was added to the vial. Sterile control experiments using autoclaved sediment samples (121 C, 20 min; 3 replicates), 2-month-old sterile P. australis plants (4e6 leaves, about 30 cm tall), and autoclaved surface-water samples were also performed, to evaluate adsorption and uptake of HMW-PAHs by the plants. In total, 6 treatments were performed for each of 3 freshwater sites: (i) rhizosphere sediment with plant and pyrene, (ii) rhizosphere sediment with plant and benzo[a]pyrene, (iii) unvegetated sediment without plant and with pyrene, (iv) unvegetated sediment without plant and with benzo[a]pyrene, (v) autoclaved rhizosphere sediment with sterile plant and pyrene, and (vi) autoclaved rhizosphere sediment with sterile plant and benzo[a] pyrene. Nine identical microcosms were prepared for each treatment. All microcosms were incubated statically (30 C, 10,000 lx, 16:8 h light:dark). Three vials from each treatment were sampled at the start of the 28-d experimental period and at days 14 and 28 for analysis of pyrene and benzo[a]pyrene levels and determination of the numbers of pyrene and benzo[a] pyrene-degrading bacteria. The upper parts of plants (shoots and leaves) were separated from the roots, which were left in the sediment before extraction of PAHs. Both types of sediment (i.e., rhizosphere sediment with the roots and unvegetated sediment without roots) were acidified with 5 mL salting-out solution (1 N HCl, 30% [w:v] NaCl), shaken with 20 mL of 1:1 (v:v) dichloromethane:methanol at 300 rpm for 20 min, sonicated in an ultrasonic bath (20 kHz, 200 W, 5-s interval, 4 C) for 20 min, and shaken again for 20 min, after which the organic layer was collected. This extraction was performed two more times (total ¼ 3) for each sample. The extract was dried under flowing nitrogen, dissolved in 1 mL acetonitrile, and analyzed by highperformance liquid chromatography (HPLC) (see section 2.9 below). We have confirmed that almost all of the pyrene (more than 94%) and benzo[a]pyrene (more than 92%) was recovered from the pyrene and benzo[a]pyrene-amended sediments, respectively. The numbers of pyrene and benzo[a]pyrene-degrading bacteria in the microcosms were determined by most probable number (MPN) assay, as previously described (Miya and Firestone, 2000). Bacterial densities are expressed as MPN per gram of dry sediment. For the MPN assay, pyrene (100 mg L1) or benzo[a]pyrene (100 mg L1) was provided as the sole carbon source for pyrene or benzo[a]pyrene-degrading bacteria, respectively.
For enrichment of pyrene-degrading bacteria from the initial rhizosphere sediment samples (no benzo[a]pyrene-degrading bacteria were found), about 1 g wet-weight of rhizosphere sediment from each location was added to 100 mL of PYR-BSM (pyrene, 100 mg L1), and the mixture was incubated at 28 C on a rotary shaker at 120 rpm for 14 d. After confirmation of bacterial growth, 1 mL of each culture was transferred to fresh PYR-BSM (100 mg L1) and incubated for 14 d. After a third transfer, enrichment cultures were serially diluted and spread on BSM agar plates, and the plates were sprayed with a 1:1 (v:v) solution of n-hexane and acetone containing pyrene (about 10% [w:v]) and incubated at 28 C. Morphologically different colonies that produced clear zones were screened for their ability to degrade pyrene (Kiyohara et al., 1982). Subsequently, the isolated bacterial strainsddesignated IPF, KTM-1, KTM-2, and YTMdwere characterized and identified by physiological and phylogenetic analyses, as described previously (Inoue et al., 2008). The 16S rDNA sequence data of strains IPF, KTM-1, KTM-2, and YTM have been submitted to the DDBJ/EMBL/GenBank databases under accession numbers AB491971, AB491972, AB491973, and AB491974, respectively.
2.6. Pyrene and benzo[a]pyrene degradation tests using pure cultures of isolates Each isolate was grown in PYR-BSM (100 mg L1) for 5 d. Cells were harvested by centrifugation (9600 g, 4 C, 10 min), washed twice with 50 mM potassium phosphate buffer (pH 7.5), and inoculated to a final cell density (as determined by optical density at 600 nm [OD600]) of 0.02 (i.e., OD600 ¼ 0.02) in 30-mL vials containing 10 mL of PYR-BSM (100 mg L1). The cells (whole cells) were also inoculated at an OD600 of 0.2 in 30mL vials containing 10 mL of BaP-BSM (benzo[a]pyrene, 20 mg L1). We prepared 42 identical vial cultures for each test. Culture was carried out at 28 C and 120 rpm. Triplicate vials from each test were removed over the 5-d experimental period, and the cell density (OD600) and substrate concentrations (see section 2.9 below) were monitored. We also performed sterile control tests, using autoclaved (121 C, 20 min) bacterial cells of each isolate, and control tests, using the cells of each isolate pre-grown in 1/10 LB medium.
2.7.
Characterization of root exudates of P. australis
To examine changes in P. australis root exudates due to pyrene-induced stress, 2-month-old sterile plants (4e6 leaves, about 30 cm tall) were transferred to 50 mL sterile Hoagland’s nutrient solution with 1.0 mg L1 pyrene (“pyrene-exposed” plants) or without pyrene (“unexposed” plants) and grown in an incubation chamber (30 C, 10,000 lx, 16:8 h light:dark) for 1 d. Three plants were used for each of these two treatments. Each plant was then collected and gently washed twice in 50 mL sterile water for 3 min. Each plant was then transferred to a flask containing 50 mL sterile pure water and grown in the incubation chamber for 7 d. The root exudates in the bulkwater fraction and root-surface fraction were then collected
1632
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 2 9 e1 6 3 8
from each flask and analyzed separately to examine the distribution of root exudates in the rhizosphere. Fifty milliliters of water was directly collected as the bulk-water fraction. To obtain the root-surface fraction, the roots, separated from the plant, were transferred to 50 mL of sterile pure water and sonicated in an ultrasonic bath (20 kHz, 200 W, 5-s interval, 4 C) for 10 min, and the water was then collected as the rootsurface fraction. Both fractions were subjected to total organic carbon (TOC) and total phenolic compounds (TPC) quantification (Toyama et al., 2009b). Results are presented as the mean values of triplicate samples. The ability of P. australis to release organic and phenolic compounds was calculated as milligrams of TOC or TPC per gram of wet root material per day (mg TOC [g wet root]1 d1 or mg TPC [g wet root]1 d1). Furthermore, to qualitatively examine constituents in total root exudates, 100 mL combined volume of the bulk-water and root-surface root exudate fractions (1:1 [v:v] bulk-water and root-surface fractions) was analyzed by HPLC. For this analysis, the combined root exudate mixtures were concentrated by using an Oasis HLB polymeric cartridge (500 mg/6 mL, Waters, Massachusetts, USA) (Toyama et al., 2009b) and dissolved in 200 mL acetonitrile, and the sample was analyzed by HPLC (see section 2.9 below).
2.8. Pyrene and benzo[a]pyrene degradation tests in the presence or absence of root exudates Strain IPF was selected for additional testing for degradation of pyrene and benzo[a]pyrene in the presence or absence of root exudates. Strain IPF was grown in 1/10 LB medium for 2 d. The cells were harvested, washed, and added at an OD600 of 0.01e10 mL PYR-BSM (100 mg L1) or BaP-BSM (10 mg L1). Freeze-dried concentrate of the combined root exudate mixture (1:1 [v:v] bulk-water and root-surface fractions) from pyreneexposed plants or unexposed plants was added to the BSM before incubation to investigate the effect of root exudates on strain IPF (see Results section 3.5). Cells of strain IPF were also added to 1/10 LB medium containing pyrene (100 mg L1) or benzo[a]pyrene (10 mg L1) as control experiments. We prepared 27 identical vial cultures for each test. Culture was carried out in the dark at 28 C and 120 rpm. Triplicate vials from each test were removed over the 5-d experimental period and the cell density and concentrations of pyrene and benzo[a] pyrene were monitored (see section 2.9).
2.9. Analytical procedures for PAHs and phenolic compounds Pyrene and benzo[a]pyrene concentrations in sediment and cultures were determined by HPLC, and metabolites of benzo [a]pyrene were analyzed by gas chromatographyemass spectrometry (GCeMS) after HPLC separation. For HPLC and GCeMS analyses, the entire collected culture was acidified with 1 N HCl, shaken for 3 min with an equal volume of 1:1 (v:v) n-hexane:ethyl acetate, and centrifuged (9600 g, 4 C, 10 min). The organic layer was then collected. The extract was dried under flowing nitrogen, and the dry extract was dissolved in 1 mL acetonitrile and analyzed. HPLC analysis was conducted using a Shimadzu HPLC system with a UV/vis
detector and a Shim-pack VP-ODS column (150 mm 4.6 mm, particle size 5 mm; Shimadzu, Kyoto, Japan). The mobile phase for the HPLC analysis was 8:2 acetonitrile:water (v:v), with detection at a wavelength of 254 nm. GCeMS analysis was conducted using a Shimadzu GC/MS system (GCMS-QP2010) and an Rxi-5ms capillary column (30 m 0.25 mm ID, 1.00 mm df; Restek, Bellefonte, PA USA). For GCeMS analysis of metabolites from benzo[a]pyrene degradation, the fractions containing the metabolite peaks detected by HPLC were collected, dried, and trimethylsilylated with BSTFA-acetonitrile solution at 60 C for 1 h. For GCeMS analysis, the column temperature was maintained at 90 C for 1 min, increased to 150 C by 15 C min1, increased to 300 C by 5 C min1, and maintained at 300 C for 6 min. For HPLC analysis of the concentrated and re-dissolved root exudate mixtures, we applied a gradient elution starting with 20% acetonitrile for the first 3 min, after which the acetonitrile percentage was increased linearly to 100% over 27 min; this was followed by elution with 100% acetonitrile for 10 min and then with 20% acetonitrile for the final 5 min. The flow rate was 1.0 mL min1 at 40 C, and the detection wavelength was 254 nm. The HPLC analysis was conducted using a Shim-pack VP-ODS column (250 mm 4.6 mm, particle size 5 mm; Shimadzu). Results are presented as the mean values and 95% confidence intervals of triplicate experiments.
3.
Results
3.1. Degradation of pyrene and benzo[a]pyrene in P. australis rhizosphere and unvegetated sediments Detailed physicochemical analysis of the sediment samples showed them to have a pH of 7.0e7.3, low organic carbon content (ignition loss ¼ 1.0%e2.3%), and low levels of pyrene (0.0029e0.0079 mg [kg dry sediment]1) and benzo[a]pyrene (0e0.0023 mg [kg dry sediment]1). We conducted degradation experiments in both rhizosphere and unvegetated sediments spiked with pyrene (50 mg [kg dry sediment]1) or benzo[a]pyrene (50 mg [kg dry sediment]1). Pyrene and benzo[a]pyrene levels declined slightly or did not decline at all in the unvegetated sediments without plants; 2.99%e14.4% of the pyrene and 0%e14.5% of the benzo [a]pyrene disappeared from these sediments over 28 d (Fig. 1). In contrast, in rhizosphere sediments with plants, 50.2%e 61.3% of the pyrene and 40.7%e61.5% of the benzo[a]pyrene were removed from the sediments over 28 d (Fig. 1). In autoclaved sediments with sterile plants, there was no substantial decrease in pyrene or benzo[a]pyrene content over 28 d (Fig. 1). Initial densities of pyrene-degrading bacteria in the rhizosphere sediments from all 3 sites were 100e204 times those in the unvegetated sediments (Fig. 2). After the 28-d degradation experiments with Inukai Pond and Kamo River samples, the densities of pyrene-degrading bacteria in the rhizosphere sediments were 197 and 42 times, respectively, those in the unvegetated sediments. However, in the Yodo River samples, the densities of pyrene-degrading bacteria in the rhizosphere and unvegetated sediments at the end of the experiment were almost equal (Fig. 2). Although only 14.4% of the pyrene was
1633
Pyrene, Benzo[a]pyrene (mg kg–1)
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 2 9 e1 6 3 8
A
50
B
50
40
40
40
30
30
30
20
20
20
10
10
10
0
0
10
20
30
0
0
10
20
30
C
50
0
0
10
20
30
Time (days) Fig. 1 e Removal of pyrene and benzo[a]pyrene from pyrene or benzo[a]pyrene-spiked sediment microcosms from (A) Inukai Pond, (B) Yodo River, and (C) Kamo River, Japan. Closed symbols represent microcosms with rhizosphere sediment and P. australis plants; open symbols represent microcosms with unvegetated sediment and no P. australis. Dotted lines represent sterile control experiments using autoclaved sediment samples, 2-month-old sterile P. australis plants, and autoclaved surface-water samples. Squares and circles represent pyrene and benzo[a]pyrene concentrations, respectively. Data are means of triplicate experiments, and error bars indicate 95% confidence intervals.
removed from the unvegetated sediment of the Yodo River over the experimental period, the pyrene-degrading bacterial density increased substantially for still unknown reasons. We did not detect bacteria capable of utilizing benzo[a]pyrene as the sole carbon source in any of the sediment samples.
3.2. Isolation and identification of pyrene-degrading bacteria We enriched for pyrene-degrading bacteria from the initial rhizosphere sediment samples because they had been detected in these samples by the MPN assay. In total, 4 pyrenedegrading bacterial strains were isolated from the rhizosphere sediments. Strain IPF was isolated from the Inukai Pond sediment, strains KTM-1 and KTM-2 were isolated from the Kamo River sediment, and strain YTM was isolated from the Yodo River rhizosphere sediment. The 4 bacterial isolates were rod-shaped, gram- and catalase-positive, and oxidasenegative. They utilized glucose and D-mannitol as sole carbon sources, but not L-arabinose, D-mannose, N-acetyl-D-glucosamine, maltose, gluconate, n-caprate, adipate, D,L-malate, citrate, or phenylacetate. The partial 16S rRNA gene sequences of the 4 isolates showed the highest identity (99.3%) with Mycobacterium gilvum ATCC 43909T (accession no. X55599), although there were slight differences among the isolates. Thus, the 4 isolates were identified as M. gilvum.
3.3. Degradation of pyrene and benzo[a]pyrene by isolated strains The 4 isolated M. gilvum strains that were pre-grown on pyrene completely degraded 100 mg L1 pyrene within 3 d under aerobic conditions with concomitant cell growth (Fig. 3A); however, pyrene degradation and cell growth were not observed under anaerobic conditions (data not shown). No metabolites were detectable after complete pyrene disappearance. Although the 4 isolates did not use benzo[a]pyrene
for growth, whole cells pre-grown on pyrene (i.e., whole cells induced with pyrene) did degrade approximately 40% of 20 mg L1 benzo[a]pyrene within 5 d (Fig. 3B). Benzo[a]pyrene degradation was not observed in experiments using whole cells pre-grown in 1/10 LB medium (Supplementary information: Fig. S1). Three major metabolites of benzo[a]pyrene were detected by HPLC at retention times (RTs) of 2.3, 2.8, and 3.7 min, along with the decrease in benzo[a]pyrene with an RT of 11.6 min (Supplementary information: Fig. S2); these metabolites were identified by GCeMS. The trimethylsilyl (TMS)-derivatized metabolites were uniform in mass spectra (m/z of fragment ions [% relative intensity, characterization]): m/z 430 (50, Mþ), 415 (9, Mþ e CH3), 357 (10, Mþ e TMS), 341 (54, Mþ e OTMS), 284 (12, Mþ e 2TMS), 252 (16, Mþ e 2OTMS), 147 (43, adjacent TMS groups), and 73 (100, TMS). The spectral data were consistent with those of ditrimethylsilyl-benzo[a]pyrene dihydrodiol (Moody et al., 2004); therefore, the metabolites were identified as benzo[a]pyrene dihydrodiol isomers.
3.4.
Characterization of P. australis root exudates
We compared root exudates from plants with and without pyrene exposure. The TOC and TPC (mg [g wet root]1 d1) in root exudates released from sterile P. australis plants into bulk-water and root-surface fractions of the rhizosphere are shown in Table 1. Detailed chemical analyses by HPLC and GCeMS showed that the root exudates of pyrene-exposed plants had undetectable levels of pyrene (<0.001 mg L1). The percentage TPC in the root-surface fractions of both pyreneexposed and unexposed plants were about 4.0 and 14 times, respectively, those in the bulk-water fractions. In addition, the percentage TPC in the total root exudate (the combined root exudate mixture of bulk-water and root-surface fractions) of pyrene-exposed plants was about 4.0 times that of unexposed plants.
1634
8 7 6 5 4 3 2 1 0
0.30
A
120
0.25
100
0.20
80 0.15 60 0.10
40
0.05
20 0
28
days Inukai Pond
0
28
days Yodo River
0
0
28
days Kamo River
25
3.5. Degradation of pyrene and benzo[a]pyrene by M. gilvum strain IPF in the presence of root exudates Strain IPF was selected for further degradation studies because it exhibited the highest pyrene and benzo[a]pyrenedegrading abilities among the 4 isolates. We compared pyrene (100 mg L1) and benzo[a]pyrene (10 mg L1) degradation by strain IPF in the presence and absence of P. australis root exudates. We added the root exudate mixture from the pyrene-exposed plants (final concentration: 182 mg TOC L1; 7.62 mg TPC L1) or unexposed plants (final concentration: 161 mg TOC L1; 1.38 mg TPC L1) to BSM before incubation. Both pyrene degradation and cell growth were greater in the presence of root exudates from pyrene-exposed or unexposed plants than in the absence of these exudates (Fig. 5A). The rates and extents of pyrene degradation were similar, irrespective of the type of root exudate. In addition, cell growth in 1/10 LB medium without root exudates (control experiment) was greater than that in BSM without root exudates, but the rates and extents of pyrene degradation were similar in the two experiments (Fig. 5A). In the benzo[a]pyrene degradation experiments, neither cell growth nor benzo[a]pyrene degradation was observed in the absence of root exudates (Fig. 5B). However, the cell density increased in the presence of both types of root exudates, and 51% and 22% of the benzo[a]pyrene was degraded within 5 d in the presence of root exudates from pyrene-exposed or unexposed plants, respectively (Fig. 5B). The rates and extents of benzo[a]pyrene degradation were substantially higher in the presence of root exudates from
Benzo[a]pyrene (mg L–1)
Fig. 2 e Densities of pyrene-degrading bacteria in pyrenespiked rhizosphere sediment with plants (filled bars) and unvegetated sediment without plants (open bars) at the start and end of a 28-d degradation experiment. Data are means of triplicate experiments, and error bars indicate 95% confidence intervals.
HPLC profiles of the total root exudates differed for pyreneexposed and unexposed plants (Fig. 4). Nineteen distinct peaks with RTs between 10.0 and 40.0 min were detected in the root exudates of pyrene-exposed plants, whereas only 11 were detected in those of unexposed plants, indicating that there was a markedly higher number of constituents in the root exudates from pyrene-exposed plants.
0
1
2
3
4
5
2
3
4
5
Cell density (OD600)
140
Pyrene (mg L–1)
Pyrene-degrad in g bacterial d ensity (Log 10MP N g –1 sediment)
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 2 9 e1 6 3 8
0
B
20 15 10 5 0
0
1
Time (days) Fig. 3 e Pyrene and benzo[a]pyrene degradation by the isolated strains. (A) Pyrene degradation by, and cell growth of, M. glivum strains IPF (squares), KTM-1 (diamonds), KTM-2 (triangles), and YTM (circles), grown on pyrene in basal salts medium (BSM). (B) Benzo[a]pyrene degradation by whole cells of strains IPF (squares), KTM-1 (diamonds), KTM-2 (triangles), and YTM (circles), pre-grown on pyrene in BSM. Closed symbols represent concentrations of pyrene or benzo[a]pyrene, and open symbols in A represent cell densities. Dotted line in B represents data from sterile control tests. Data are means of triplicate experiments, and error bars indicate 95% confidence intervals.
pyrene-exposed plants than in the presence of root exudates from unexposed plants. Coincident with benzo[a]pyrene degradation, we detected benzo[a]pyrene dihydrodiol isomers in the presence of both root exudates. In the benzo [a]pyrene degradation control experiment in 1/10 LB medium without root exudates, we observed substantial cell growth but no substantial benzo[a]pyrene degradation (Fig. 5B).
4.
Discussion
We observed the accelerated removal of pyrene and benzo[a] pyrene in P. australis rhizosphere sediments containing plants, whereas both compounds persisted in unvegetated sediments without plants and in autoclaved rhizosphere sediments with sterile plants. Initial densities of pyrenedegrading bacteria were substantially higher in the
1635
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 2 9 e1 6 3 8
Table 1 e Characteristics of root exudates released by sterile (bacteria-free) P. australis plants into bulk-water and rootsurface fractions of the rhizosphere. Total root exudate release rate (mg [g wet root]1 d1)
Distribution of root exudates (mg [g wet root]1 d1) Bulk-water fraction TOCa
TPCb
Root-surface fraction
TPC/TOC (%)
Unexposed plants 44.5 1.48 0.0684 0.0062 Pyrene-exposed plants 47.8 3.91 0.592 0.102
TOC
TPC
TPC/TOC (%)
TOC
TPC
TPC/TOC (%)
2.10 5.02
69.4 108
0.591 3.63
0.85 3.35
24.9 2.90 0.523 0.112 60.6 5.02 3.04 0.60
0.15 1.24
a TOC, total organic carbon. Data are means standard deviations of triplicate experiments. b TPC, total phenolic compounds. Data are means standard deviations of triplicate experiments.
rhizosphere sediments than in the unvegetated sediments. These results suggest that the accelerated pyrene and benzo [a]pyrene removal in the P. australis rhizosphere sediments largely resulted from biodegradation by rhizosphere bacteria rather than from adsorption and uptake by plants. These findings were common to the natural rhizosphere sediments collected from 3 freshwater sites that had not been substantially contaminated with PAHs, supporting the conclusion that phytoremediation using P. australis can be useful for pyrene and benzo[a]pyrene removal from sediments. In natural environments, low molecular weight (LMW)PAHs, such as naphthalene and phenanthrene, are relatively easily biodegraded, whereas HMW-PAHs are persistent (Cerniglia, 1992; Juhasz and Naidu, 2000; Kanaly and Harayama, 2000). Even within the same environment, the bacterial group that degrades HMW-PAHs is different than the group that degrades LMW-PAHs (Zhou et al., 2008). Recently, a variety of naphthalene and phenanthrene-degrading bacteria (mostly Pseudomonas and Paenibacillus) have been isolated from hydrocarbon-contaminated rhizosphere sediments of salt marsh plants (Daane et al., 2001; Launen et al., 2008). However, HMW-PAH-degrading bacteria in the aquatic plant rhizosphere have not been investigated. Most pyrenedegrading bacteria previously reported were fast-growing
mycobacteria and were commonly isolated from HMW-PAHor oil-contaminated soils and sediments (Heitkamp and Cerniglia, 1988; Heitkamp et al., 1988; Schneider et al., 1996; Cheung and Kinkle, 2001; Dean-Ross et al., 2002; Habe et al., 2004; Miller et al., 2004; Kim et al., 2005). Pyrene-degrading mycobacteria were recently shown to be prevalent in many HMW-PAH-contaminated sediments, where they could play an important role in the natural attenuation of HMW-PAHs (DeBruyn et al., 2007, 2009). Here, we isolated pyrene-degrading bacteriadthe M. gilvum strains IPF, KTM-1, KTM-2, and YTMdfrom P. australis rhizosphere sediments. To our knowledge, our isolates are the first bacteria from the rhizosphere of aquatic plants capable of utilizing pyrene as a sole carbon source. Interestingly, the 4 isolates from the 3 different P. australis rhizosphere sediments are all M. gilvum. Our isolates also are fast-growing mycobacteria and might play a major role in pyrene degradation in the P. australis rhizosphere. Although our rhizosphere sediments were not contaminated with HMW-PAHs, the rates of pyrene degradation by our isolates were comparable to, or greater than, those by pyrene-degrading mycobacteria derived from contaminated soils (Heitkamp and Cerniglia, 1988; Heitkamp et al., 1988; Dean-Ross et al., 2002; Habe et al., 2004; Miller et al., 2004).
225000
In ten sity (
V)
200000 175000
3 12
150000
8 4 5
125000
7
9
6
14 10 11 13 12 15
100000
18
Pyrene-exposed plant
17
19
1
75000
16
50000 7 8
25000 0 0.0
16
5.0
10.0
15.0
20.0
9
25.0
14 11 13 15
30.0
Unexposed plant 17
19
35.0
40.0
Retention time (min) Fig. 4 e Constituents of P. australis root exudates. HPLC UV/vis chromatograms (detection at 254 nm) of constituents in root exudate mixtures (bulk-water D root-surface fractions; 1:1 [v:v]) from pyrene-exposed and non-exposed P. australis plants.
1636
80
Benzo[a]pyrene (mg L–1)
0.3
60 0.2 40 0.1
20 0
B
0.4
0
1
2
3
4
5
0 0.20
12 10
0.15 8 0.10
6 4
0.05 2 0
Cell density (OD600)
100
0
1
2
3
4
5
Cell density (OD600)
A Pyrene (mg L–1)
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 2 9 e1 6 3 8
0
Time (days) Fig. 5 e Effects of P. australis root exudates on pyrene and benzo[a]pyrene degradation. (A) Pyrene and (B) benzo[a] pyrene degradation by M. glivum strain IPF in BSM in the presence of root exudate mixtures from non-pyreneexposed plants (circles) or pyrene-exposed plants (squares), or in the absence of root exudates (triangles). Diamonds in A and B represent data from control experiments in 1/10 LB medium. Closed symbols represent concentrations of pyrene or benzo[a]pyrene, and open symbols represent cell densities. Data are means of triplicate experiments, and error bars indicate 95% confidence intervals.
The presence of these pyrene-degrading mycobacteria in the P. australis rhizosphere, even under non-PAH-contaminated conditions, is of some interest. Also, although we did not determine the abundances of pyrene-degrading bacteria relative to those of total heterotrophic bacteria in the sediments, the densities of pyrene-utilizing bacteria in the P. australis rhizosphere sediments were clearly higher than those in the unvegetated sediments. This increase in microbial numbers in rhizosphere sediments compared with those in non-rhizosphere sediments is known as the “rhizosphere effect” (Shaw and Burns, 2003). Although we observed accelerated benzo[a]pyrene removal in P. australis rhizosphere sediments, we did not detect benzo [a]pyrene-utilizing bacteria. However, because the isolated pyrene-degrading mycobacteria were able to degrade benzo[a] pyrene when induced with pyrene, they might also play an important role in benzo[a]pyrene degradation in the rhizosphere sediments. Benzo[a]pyrene degradation by our isolates required both growth substrate and an inducer of benzo[a]
pyrene-degrading enzymes, such as pyrene, as reported previously (Juhasz and Naidu, 2000; Rentz et al., 2005). We detected benzo[a]pyrene dihydrodiols as the benzo[a]pyrene metabolites in our isolate cultures, suggesting a benzo[a]pyrene degradation pathway similar to that proposed for Mycobacterium sp. strains PYR-1 (Moody et al., 2004) and RJGII-135 (Schneider et al., 1996). Rhizosphere oxygenation by the roots of P. australis undoubtedly contributed to the accelerated biodegradation of pyrene and benzo[a]pyrene in this study, as shown by the inability of the isolated mycobacteria to degrade pyrene under anaerobic conditions. However, the accelerated biodegradation of both pyrene and benzo[a]pyrene and the high diversity of pyrene-degrading bacteria in the rhizosphere sediments cannot be fully explained only by rhizosphere oxygenation. We suspected that the root exudates of P. australis were another key factor in these phenomena. We focused especially on phenolic root exudates, because they are known to contribute to the accelerated rhizosphere biodegradation of simple aromatic compounds (Toyama et al., 2009b) and PCBs (Gilbert and Crowley, 1997; Leigh et al., 2002). To test this hypothesis, we characterized the root exudates of P. australis. Two-month-old sterile P. australis plants released root exudates into the rhizosphere at the specific release rates of 69.4 mg TOC and 0.591 mg TPC (g wet root)1 d1, confirming that the root exudates of P. australis contain phenolic compounds. The percentage of TPC in the root-surface fraction was about 14 times that in the bulkwater fraction. This suggests that the phenolic compounds released by P. australis roots probably have the greatest effect on rhizosphere bacteria on the root-surface, and that this effect would decrease with increasing distance from the roots. It is well documented that most higher plants respond to various stresses by activating secondary metabolic pathways, such as the phenylpropanoid mechanism, and that production of phenolic compounds plays an important role in resistance to stresses (Hutzler et al., 1998; Bagniewska-Zadworna et al., 2008). A pollutant-induced stress in plants causes a change in both the quantity and quality of the root exudates (Porteous et al., 2000; Mingji et al., 2009). In this study, the amount of TPC and the diversity of constituents in the root exudates substantially increased when the roots of P. australis were exposed to pyrene. The amounts of TOC and TPC in the total root exudates of pyrene-exposed plants were about 1.6 and 6.2 times, respectively, those of unexposed plants, suggesting that exposure of the roots to pyrene affected the exudation of TPC to a greater degree than the exudation of TOC. Our experiments also showed that the root exudates of P. australis supported cell growth of the pyrene-degrading Mycobacterium strain IPF (the only strain tested) and stimulated the degradation of pyrene and benzo[a]pyrene by this strain. In particular, we found that the root exudates were essential for benzo[a]pyrene degradation by strain IPF. Thus, the root exudates undoubtedly functioned as a carbon source for growth and an inducer for benzo[a]pyrene-degrading activity in strain IPF. Another interesting finding is the increase in the stimulatory effect of the root exudates on benzo[a]pyrene biodegradation, but not on pyrene biodegradation, after the roots of P. australis were exposed to pyrene. That is, the root exudates of pyrene-stressed P. australis
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 2 9 e1 6 3 8
stimulated benzo[a]pyrene degradation to a greater degree than the root exudates of an unexposed plant, free from pyrene stress. The increase in the amount of phenolic compounds in the root exudates following pyrene stress provides supporting evidence that phenolic exudates are most probably involved in the stimulatory effect on benzo[a]pyrene degradation by the pyrene-degrading Mycobacterium strain IPF. Identification of the key compounds involved in the stimulatory effects could lead to the use of P. australis as a source of natural compounds for bacterial stimulation; these compounds could then be applied directly to bioremediation.
5.
Conclusions
We investigated the biodegradation of pyrene and benzo[a] pyrene in P. australis rhizosphere sediments. Our conclusions are summarized as follows: (1) Accelerated removal of pyrene and benzo[a]pyrene was observed in P. australis rhizosphere sediments containing plants, whereas both compounds persisted in unvegetated sediments without plants and in autoclaved rhizosphere sediments with sterilized plants. Accelerated removal of both compounds largely resulted from biodegradation by rhizosphere bacteria. (2) We isolated pyrene-degrading Mycobacterium gilvum strains from P. australis rhizosphere sediments. The strains utilized pyrene aerobically as a sole carbon and energy source and were able to degrade benzo[a]pyrene when induced with pyrene. (3) The root exudates of P. australis clearly supported cell growth of the pyrene-degrading Mycobacterium strain tested and stimulated benzo[a]pyrene degradation by the straindthat is, the induction of benzo[a]pyrene-degrading activity. This stimulatory effect and the amounts of phenolic compounds in the root exudates increased when P. australis roots were exposed to pyrene. It appears that phenolics are key candidates for explaining the stimulation of benzo[a]pyrene biodegradation in the P. australis rhizosphere. This is the first study demonstrating that P. australis root exudates stimulate biodegradation of pyrene and benzo[a] pyrene. We conclude that interactions between Mycobacterium spp. and root exudates can accelerate removal of pyrene and benzo[a]pyrene in the P. australis rhizosphere sediment. The synergetic effects of oxygen and exudates released by P. australis roots in accelerating the biodegradation of HMW-PAHs increase the potential usefulness of aquatic plants for remediation of HMW-PAH-contaminated sediments.
Acknowledgments This research was supported partly by a Grant-in-Aid for Encouragement of Young Scientists (A) (no. 21681010) and Young Scientists (B) (no. 19710060) from the Ministry of Education, Culture, Sports, Science, and Technology of Japan.
1637
Appendix. Supplementary data Supplementary data associated with this article can be found in online version at doi:10.1016/j.watres.2010.11.044.
references
Aprill, W., Sims, R., 1990. Evaluation of the use of prairie grasses for stimulating polycyclic aromatic hydrocarbon treatment in soil. Chemosphere 20, 253e265. Bagniewska-Zadworna, A., Zenkteler, E., Karolewski, P., Zadworny, M., 2008. Phenolic compound localization in Polypodium vulgare L. rhizomes after mannitol-induced dehydration and controlled desiccation. Plant Cell Rep. 27, 1251e1259. Banks, M.K., Lee, E., Schwab, A.P., 1999. Evaluation of dissipation mechanisms for benzo[a]pyrene in the rhizosphere of tall fescue. J. Environ. Qual. 28, 294e298. Cerniglia, C.E., 1992. Biodegradation of polycyclic aromatic hydrocarbons. Biodegradation 3, 351e368. Chen, S.J., Luo, X.J., Mai, B.X., Sheng, G.Y., Fu, J.M., Zeng, E.Y., 2006. Distribution and mass inventories of polycyclic aromatic hydrocarbons and organochlorine pesticides in sediments of the Pearl River Estuary and the Northern South China Sea. Environ. Sci. Technol. 40, 709e714. Cheung, P.-Y., Kinkle, B.K., 2001. Mycobacterium diversity and pyrene mineralization in petroleum-contaminated soils. Appl. Environ. Microbiol. 67, 2222e2229. Chung, W.K., King, G.M., 1999. Biogeochemical transformations and potential polyaromatic hydrocarbon degradation in macrofaunal burrow sediments. Aquat. Microb. Ecol. 19, 285e295. Daane, L.L., Harjono, I., Zylstra, G.J., Ha¨ggblom, M.M., 2001. Isolation and characterization of polycyclic aromatic hydrocarbon-degrading bacteria associated with the rhizosphere of salt marsh plants. Appl. Environ. Microbiol. 67, 2683e2691. Dean-Ross, D., Moody, J., Cerniglia, C.E., 2002. Utilization of mixtures of polycyclic aromatic hydrocarbons by bacteria isolated from contaminated sediment. FEMS Microbiol. Ecol. 41, 1e7. DeBruyn, J.M., Chewning, C.S., Sayler, G.S., 2007. Comparative quantitative prevalence of Mycobacteria and functionally abundant nidA, nahAc, and nagAc diosygenase genes in coal tar contaminated sediments. Environ. Sci. Technol. 41, 5426e5432. DeBruyn, J.M., Mead, T.J., Wilhelm, S.W., Sayler, G.S., 2009. PAH biodegradative genotypes in Lake Erie sediments: evidence for broad geographical distribution of pyrene-degrading mycobacteria. Environ. Sci. Technol. 43, 3467e3473. Donnelly, P.K., Hegde, R.S., Fletcher, J.S., 1994. Growth of PCBdegrading bacteria on compounds from photosynthetic plants. Chemosphere 28, 981e988. Fan, S., Li, P., Gong, Z., Ren, W., He, N., 2008. Promotion of pyrene degradation in rhizosphere of alfalfa (Medicago sativa L.). Chemosphere 71, 1593e1598. Gilbert, E.S., Crowley, D.E., 1997. Plant compounds that induce polychlorinated biphenyl biodegradation by Arthrobacter sp. strain B1B. Appl. Environ. Microbiol. 63, 1933e1938. Habe, H., Kanemitsu, M., Nomura, M., Takenura, T., Iwata, K., Nojiri, H., Yamane, H., 2004. Isolation and characterization of an alkaliphilic bacterium utilizing pyrene as a carbon source. J. Biosci. Bioeng. 98, 306e308. Heitkamp, M.A., Cerniglia, C.E., 1988. Mineralization of polycyclic aromatic hydrocarbons by a bacterium isolated from
1638
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 2 9 e1 6 3 8
sediment below an oil field. Appl. Environ. Microbiol. 54, 1612e1614. Heitkamp, M.A., Franklin, W., Cerniglia, C.E., 1988. Microbial metabolism of polycyclic aromatic hydrocarbons: isolation and characterization of a pyrene-degrading bacterium. Appl. Environ. Microbiol. 54, 2549e2555. Hutzler, P., Fischbach, R., Heller, W., Jungblut, T.P., Reuber, S., Schmitz, R., Veit, M., Weissenbo¨ck, G., Schnitzler, J.-P., 1998. Tissue localization of phenolic compounds in plants by confocal laser scanning microscopy. J. Exp. Bot. 49, 953e965. Inoue, D., Hara, S., Kashihara, M., Murai, Y., Danzl, E., Sei, K., Tsunoi, S., Fujita, M., Ike, M., 2008. Degradation of bis(4hydroxyphenyl)methane (bisphenol F) by Sphingobium yanoikuyae strain FM-2 isolated from river water. Appl. Environ. Microbiol. 74, 352e358. Jouanneau, Y., Willison, J.C., Meyer, C., Krivobok, S., Chevron, N., Besombes, J.-L., Blake, G., 2005. Stimulation of pyrene mineralization in freshwater sediments by bacterial and plant bioaugmentation. Environ. Sci. Technol. 39, 5729e5735. Juhasz, A.L., Naidu, R., 2000. Bioremediation of higher molecular weight polycyclic aromatic hydrocarbons: a review of the microbial degradation of benzo[a]pyrene. Int. Biodeterior. Biodegradation 45, 57e88. Kanaly, R.A., Harayama, S., 2000. Biodegradation of highmolecular-weight polycyclic aromatic hydrocarbons by bacteria. J. Bacteriol. 182, 2059e2067. Kim, Y.-H., Engesser, K.-H., Cerniglia, C.E., 2005. Numerical and genetic analysis of polycyclic aromatic hydrocarbondegrading mycobacteria. Microb. Ecol. 50, 110e119. Kiyohara, H., Nagao, K., Yana, K., 1982. Rapid screen for bacteria degrading water-insoluble, solid hydrocarbons on agar plates. Appl. Environ. Microbiol. 43, 454e457. Launen, L.A., Dutta, J., Turpeinen, R., Eastep, M.E., Dorn, R., Buggs, V.H., Leonard, J.W., Ha¨ggblom, M.M., 2008. Characterization of the indigenous PAH-degrading bacteria of Spartina dominated salt marshes in the New York/New Jersey Harbor. Biodegradation 19, 347e363. Leigh, M.B., Fletcher, J.S., Fu, X., Schmitz, F.J., 2002. Root turnover: an important source of microbial substrates in rhizosphere remediation of recalcitrant contaminants. Environ. Sci. Technol. 36, 1579e1583. Liang, Y., Tse, M.F., Young, L., Wong, M.H., 2007. Distribution patterns of polycyclic aromatic hydrocarbons (PAHs) in the sediments and fish at Mai Po marshes nature Reserve, Hong Kong. Water Res. 41, 1303e1311. Liste, H.-H., Alexander, M., 2000. Plant-promoted pyrene degradation in soil. Chemosphere 40, 7e10. Menzie, C.A., Potocki, B.B., Santodonato, J., 1992. Exposure to carcinogenic PAHs in the environment. Environ. Sci. Technol. 26, 1278e1284. Miller, C.D., Hall, K., Liang, Y.N., Nieman, K., Sorensen, D., Issa, B., Anderson, A.J., Sims, R.C., 2004. Isolation and characterization of polycyclic aromatic hydrocarbon-degrading Mycobacterium isolates from soil. Microb. Ecol. 48, 230e238. Mingji, X., Chongling, Y., Jing, Y., Lily, W., 2009. Impact of phenanthrene on organic acids secretion and accumulation by
perennial ryegrass, Lolium perenne L., root. Bull. Environ. Contam. Toxicol 83, 75e80. Miya, R.K., Firestone, M.K., 2000. Phenanthrene-degrader community dynamics in rhizosphere soil from a common annual grass. J. Environ. Qual. 29, 584e592. Miya, R.K., Firestone, M.K., 2001. Enhanced phenanthrene biodegradation in soil by slender oat root exudates and root debris. J. Environ. Qual. 30, 1911e1918. Moody, J.D., Freeman, J.P., Fu, P.P., Cerniglia, C.E., 2004. Degradation of benzo[a]pyrene by Mycobacterium vanbaalenii PYR-1. Appl. Environ. Microbiol. 70, 340e345. Mueller, K.E., Shann, J.R., 2007. Effects of tree root-derived substrates and inorganic nutrients on pyrene mineralization in rhizosphere and bulk soil. J. Environ. Qual. 36, 120e127. Porteous, F., Killham, K., Meharg, A., 2000. Use of a lux-marked rhizobacterium as a biosensor to assess changes in rhizosphere C flow due to pollutant stress. Chemosphere 41, 1549e1554. Reilley, K.A., Banks, M.K., Schwab, A.P., 1996. Dissipation of polycyclic aromatic hydrocarbons in the rhizosphere. J. Environ. Qual. 25, 212e219. Rentz, J.A., Alvarez, P.J.J., Schnoor, J.L., 2005. Benzo[a]pyrene cometabolism in the presence of plant root extracts and exudates: implications for phytoremediation. Environ. Pollut. 136, 477e484. Schneider, J., Grosser, R., Jayasimhulu, K., Xue, W., Warshawsky, D., 1996. Degradation of pyrene, benzo[a] anthracene, and benzo[a]pyrene by Mycobacterium sp. strain RJGII-135, isolated from a former coal gasification site. Appl. Environ. Microbiol. 62, 13e19. Shaw, L.J., Burns, R.G., 2003. Biodegradation of organic pollutants in the rhizosphere. Adv. Appl. Microbiol. 53, 1e60. Toyama, T., Yu, N., Kumada, H., Sei, K., Ike, M., Fujita, M., 2006. Accelerated aromatic compounds degradation in aquatic environment by use of interaction between Spirodela polyrrhiza and bacteria in its rhizosphere. J. Biosci. Bioeng. 101, 346e353. Toyama, T., Sato, Y., Inoue, D., Sei, K., Chang, Y.-C., Kikuchi, S., Ike, M., 2009a. Biodegradation of bisphenol A and bisphenol F in the rhizosphere sediment of Phragmites australis. J. Biosci. Bioeng. 108, 147e150. Toyama, T., Sei, K., Yu, N., Kumada, H., Inoue, D., Hoang, H., Soda, S., Chang, Y.-C., Kikuchi, S., Fujita, M., Ike, M., 2009b. Enrichment of bacteria possessing catechol dioxygenase genes in the rhizosphere of Spirodela polyrrhiza: a mechanism of accelerated biodegradation of phenol. Water Res. 43, 3765e3776. Yi, H., Crowley, D.E., 2007. Biostimulation of PAH degradation with plants containing high concentrations of linoleic acid. Environ. Sci. Technol. 41, 4382e4388. Yoshitomi, K.J., Shann, J.R., 2001. Corn (Zea mays L.) root exudates and their impact on 14C-pyrene mineralization. Soil Biol. Biochem. 33, 1769e1776. Zhou, H.W., Luan, T.G., Zou, F., Tam, N.F.Y., 2008. Different bacteria groups for biodegradation of three- and four-ring PAHs isolated from a Hong Kong mangrove sediment. J. Hazard. Mater. 152, 1179e1185.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 3 9 e1 6 5 0
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Comparison of MFI-UF constant pressure, MFI-UF constant flux and Crossflow Sampler-Modified Fouling Index Ultrafiltration (CFS-MFIUF) Lee Nuang Sim, Yun Ye, Vicki Chen*, Anthony G. Fane UNESCO Centre for Membrane Science and Technology, University of New South Wales, 2052 Sydney, NSW, Australia
article info
abstract
Article history:
Understanding the foulant deposition mechanism during crossflow filtration is critical in
Received 13 March 2010
developing indices to predict fouling propensity of feed water for reverse osmosis (RO).
Received in revised form
Factors affecting the performance on different fouling indices such as MFI-UF constant
30 November 2010
pressure, MFI-UF constant flux and newly proposed fouling index, CFS-MFIUF were inves-
Accepted 1 December 2010
tigated. Crossflow Sampler-Modified Fouling Index Ultrafiltration (CFS-MFIUF) utilises
Available online 9 December 2010
a typical crossflow unit to simulate the hydrodynamic conditions in the actual RO units followed by a dead-end unit to measure the fouling propensity of foulants. CFS-MFIUF was
Keywords:
found sensitive to crossflow velocity. The crossflow velocity in the crossflow sampler unit
Fouling index
influences the particle concentration and the particle size distribution in its permeate.
Colloidal fouling
CFS-MFIUF was also found sensitive to the permeate flux of both CFS and the dead-end cell.
Reverse osmosis
To closely simulate the hydrodynamic conditions of a crossflow RO unit, the flux used for
Crossflow filtration
CFS-MFIUF measurement was critical. The best option is to operate both the CFS and deadend permeate flux at flux which is normally operated at industry RO units (w20 L/m2 h), but this would prolong the test duration excessively. In this study, the dead-end flux was accelerated by reducing the dead-end membrane area while maintaining the CFS permeate flux at 20 L/m2 h. By doing so, a flux correction factor was investigated and applied to correlate the CFS-MFIUF measured at dead-end flux of 120 L/m2 h to CFS-MFIUF measured at dead-end flux of 20 L/m2 h for RO fouling rate prediction. Using this flux correction factor, the test duration of CFS-MFIUF can be shortened from 15 h to 2 h. ª 2010 Elsevier Ltd. All rights reserved.
1.
Introduction
Significant effort has been focused on developing a reliable and useful fouling index to assess fouling propensity of feed water prior to the reverse osmosis (RO) units. A fouling index can be used to predict how rapidly a given feed water will foul the RO system due to colloidal fouling. With this information in hand, appropriate pre-treatment schemes can then be selected prior to the RO system which can ultimately reduce colloidal fouling in RO.
There are several fouling indices such as the Silt Density Index (SDI), and the Modified Fouling Index (MFI0.45). Due to its simplicity and the short duration of measurement, SDI is currently the most widely used index in water industry. However, SDI has been proven unreliable by many researchers (Boerlage et al., 2003a; Lipp et al., 1990; Schippers and Verdouw, 1980; Yiantsios and Karabelas, 2002). SDI was found to have no relationship with foulant concentration and is not derived from any fouling mechanism. These downsides led to the emergence of MFI0.45 which is calculated based on cake filtration theory
* Corresponding author. Tel.: þ61 2 9385 4813; fax: þ61 2 9385 5966. E-mail address:
[email protected] (V. Chen). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.12.001
1640
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 3 9 e1 6 5 0
Abbreviations 2
A filtration area (m ) solid concentration (g/L) Cb CFS-MFIUF Crossflow Sampler-Modified Fouling Index Ultrafiltration (s L2) standard reference pressure (207 kPa) DP0 DP transmembrane pressure (Pa) h circular channel diameter (m) I cake resistivity (m4) J flux (L m2 h1) cake mass per unit area (kg/m2) mc MFI-UFconst flux Modified Fouling Index-Ultrafiltration constant flux (s L2)
(Schippers and Verdouw, 1980). MFI0.45 has a linear relationship with the feed concentration, but the subsequent research carried out by Schippers et al. (1981) indicated that MFI0.45 was unable to account for small colloids (<0.45 mm). The Modified Fouling Index-Ultrafiltration (MFI-UF) and Nanofiltration-Modified Fouling Index (NF-MFI) were subsequently introduced, where ultrafiltration and nanofiltration membranes are used respectively as the test filters. MFI-UF was originally measured under constant pressure conditions. However, due to the prolonged duration of the test, Boerlage et al. (2004) proposed a MFI-UF which can be measured under constant flux mode. From their findings, the duration of the MFI-UF test could be shortened from 20 h to approximately 2 h if operated at constant flux of 75 L/m2 h using canal water. More recently, Choi et al. (2009) proposed a novel fouling index known as Combined Fouling Index (CFI) which uses a set of different membrane filters to determine the fouling potential of water. The authors suggested that no single method can be successfully used for accurate prediction of fouling potential of feed waters, but combination of different fouling indices using different types of test membranes such as hydrophobic MF, hydrophilic MF and hydrophilic UF may be possible as each of these membranes can capture different portions of foulants in a given feed. For example, hydrophobic MF is used to capture the fouling potential of hydrophobic foulants whereas hydrophilic UF is sensitive to the effects of colloidal matter and macromolecules on fouling. However, the capability of the fouling tests relied on their ability to capture the critical factor of feed water components which may contribute significantly to the fouling of RO. The above mentioned indices are carried out in a pressurized dead-end filtration cell, whereas in actual RO systems, crossflow filtration is the most widely chosen operating condition. These two operation modes have very different hydrodynamic conditions. In crossflow filtration, particles movement to and from the membrane surface is governed by the flux towards the membrane and the back transport of particles which includes Brownian diffusion, inertial lift (Green and Belfort, 1980) and shear-induced diffusion (Romero and Davis, 1988, 1991). If conditions are such that the back transport is greater than the permeate flux, then the particles are not expected to be deposited on the membrane surface. This effect is more likely for larger particles because back
MFI-UFconst.pressure Modified Fouling Index-Ultrafiltration constant pressure (s L2) Q flowrate (L h1) t filtration time (s) v crossflow velocity (m s1) V permeate volume (L) Greek symbols a specific cake resistance (m/kg) g shear force (N) 3 cake porosity m viscosity of fluid (Pa s) u compressibility factor particle density (kg m3) rp
transport velocities increase with the particle diameter (Green and Belfort, 1980; Romero and Davis, 1988, 1991). Therefore, in the crossflow process, large particles that have larger back transport velocities will tend to migrate away from the membrane surface. These hydrodynamic conditions that occur during crossflow membrane processes are neglected in the conventional dead-end MFI test. Without considering the hydrodynamics in the RO crossflow process and the mode of operation, some important issues regarding the fouling potential of the feed might be overlooked. The Crossflow Sampler-Modified Fouling Index (CFS-MFI) was introduced to incorporate the crossflow hydrodynamic behaviour during fouling index measurement (Adham and Fane, 2008). SDI and MFI0.45 constant pressure obtained after the crossflow sampler were found to be significantly lower than standard SDI and MFI0.45 for different types of feed water, emphasising the importance of crossflow hydrodynamics (Adham and Fane, 2008; Javeed et al., 2009). Our recent work has further extended this work where the crossflow sampler unit (CFS) was directly connected with the dead-end cell and the test was carried out under constant flux, known as CFSMFIUF constant flux (Sim et al., 2010). The aim of the CFS is to simulate the selective deposition of colloids during the crossflow RO process. Due to the shear, only the portion of the particles that will potentially deposit on the membrane can permeate through the CFS and enter the dead-end MFI device. These components represent the composition that is most likely to cause fouling in a RO crossflow system if the same feed was used. The fouling potential of these foulants can hence be determined through the dead-end device. In our previous study, the sensitivity of both MFI-UFconst.flux and CFSMFIUF was validated through lab scale RO experiments using synthetic silica suspension. MFI-UFconst.flux was found less sensitive when compared to CFS-MFIUF. The fouling rate prediction based on CFS-MFIUF agreed well with the actual RO fouling behaviour with the deviation of 11%, whereas MFIUFconst.flux deviated significantly from the actual trend (>30%) even with the deposition factor correction (Sim et al., 2010). However, the factors affecting the performance of this improved MFI test such as crossflow velocity on CFS-MFIUF values have not yet been presented. This paper aims to understand the particle capture and fouling mechanism in different fouling indices at which the
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 3 9 e1 6 5 0
effects of operating conditions such as flux and pressure on MFI-UFconst.pressure and MFI-UFconst.flux were investigated. In addition, a detailed study on the effects of different operating conditions such as crossflow velocity, permeate flux of CFS and dead-end flux, and the influence of spacer on CFS-MFIUF is presented.
2.
Theory
The concept for both MFI0.45 and MFI-UFconst.pressure can be derived based on cake filtration theory without cake compression (Schippers and Verdouw, 1980). Under constant pressure, both MFI0.45 and MFI-UFconst.pressure are defined as the slope of the linear portion of the t/V and V filtration curve, where t refers to the filtration time and V is the filtrate volume. They can also be represented as follows (Schippers and Verdouw, 1980): MFI UFconst:pressure ¼
mI 2DPA2
(1)
where m is the viscosity of the feed solution, I is the resistivity, DP is the transmembrane pressure (TMP) and A is the filtration area. The resistivity I, is the product of the specific cake resistance (a) of the deposit and concentration of particles in the feed water (Cb) (Boerlage et al., 1998): I ¼ aCb
(2)
The specific cake resistance for a known concentration solution can be calculated by substituting Eq. (2) into Eq. (1): a¼
2MFI UFconst:pressure DPA2 mCb
(3)
In many cases, the solids concentration is unknown. So, the exact value of the specific cake resistance is unlikely to be known. Eq. (3) simply shows the relationship between specific cake resistance and MFI-UFconst.pressure. High values of MFIUFconst.pressure indicate that the cake on the test membrane is high in resistance. MFI-UFconst.pressure not only acts as an indicator for water quality, but also indirectly reflects the cake structure formed during the test. For compressible cakes, Almy and Lewis (1912) developed an empirical relationship between specific cake resistance and pressure as follows: a ¼ a0 DPu
(4)
where u is the compressibility factor of the cake, DP is the operating pressure and a0 is a constant. For incompressible cakes, u is zero and the higher the compressibility factor, the more compressible the cake. For compressible cakes, MFI-UFconst.pressure has the following relationship (Boerlage et al., 2003b): MFI UFconst:pressure
ma0 DPu Cb ¼ 2DPA2
(5)
Simplifying Eq. (5), gives MFI UFconst:pressure ¼ Constant DPu1 where Constant refers to ma0Cb/2A2.
(6)
1641
The derivation of MFI-UFconst.flux under constant flux is also based on cake filtration theory (Boerlage et al., 2004). MFIUFconst.flux is represented as follows: MFI UFconst:flux ¼
mI 2DP0 A2
(7)
where cake resistivity, I is obtained from the slope of TMP against time filtration curve. DP0 refers to the standard pressure of 2 bar and A is active filtration membrane area of 0.0014 m2. Similarly, CFS-MFIUF measured at constant flux is defined as: CFS MFIUF ¼
mI0 2DP0 A2
(8)
where I0 is the cake resistivity modified by the Crossflow Sampler.
3.
Materials and methods
3.1.
Experimental setup
3.1.1.
Constant pressure dead-end filtration setup
The feed tank which has a volume of 1.8 L was connected to a dead-end cell. The dead-end cell with active membrane area of 0.0014 m2 (diameter ¼ 42.2 mm) is equipped with a porous support at which ultrafiltration membrane was placed on. A nitrogen gas cylinder was used to pressurize the system to the desired operating pressure. When the applied pressure reached the required level, the outlet of the dead-end cell was opened and the permeate was collected. The feed pressure was remained constant and was monitored using a pressure transducer (Labom, CB1020) located at the feed line whereas the mass of the collected permeate was measured by a balance (Mettler Toledo, PB8001-S). Both pressure and mass data from the pressure transducer and balance respectively were recorded continuously by LabVIEW 8.2.
3.1.2.
Constant flux dead-end filtration setup
In constant flux filtration, the feed tank (1.8 L) was connected to a peristaltic pump (Masterflex L/S) prior to the dead-end cell (diameter ¼ 42.2 mm, active filtration area ¼ 0.0014 m2). The pump was used to ensure constant amount of feed to be fed in to the dead-end cell and the same amount of permeate would result. Even though the setup was operated under constant flux condition, a nitrogen gas cylinder was still needed to pressurize the system as the peristaltic pump was unable to cope with high pressure difference. Two pressure transducers (Labom, CB1020) were located at the upstream and downstream of the dead-end cell. The pressures and mass data were recorded continuously by LabVIEW 8.2.
3.1.3.
CFS-MFIUF setup
The CFS-MFIUF combined a crossflow filtration module and a dead-end MFI assessment device on the permeate of the CFS cell. The crossflow cell was made from acrylic glass (also known as Perspex) which has a flow channel of 111 mm 25 mm 4 mm and effective membrane area of 0.0028 m2. An important feature of the CFS was the use of very large pore membrane (Refer Section 3.2). A feed spacer of
1642
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 3 9 e1 6 5 0
approximately 0.79 mm thick was used. A centrifugal pump was used to pump the feed to the crossflow cell and the crossflow velocity was manipulated through the valve located at the retentate line. A peristaltic pump (Masterflex L/S pump) was located in the permeate line to withdraw the permeate from the crossflow cell and pump into the dead-end filtration cell (diameter ¼ 42 mm, unless specified) as shown in Fig. 1. The purpose for this pump was to maintain constant flux for both crossflow and dead-end filtration. The membranes used in the dead-end cell were 10 kDa ultrafiltration membranes. TMP of the dead-end cell was continuously measured and recorded by computer and plotted against filtration time to obtain resistivity I0 . The CFS-MFIUF value can be determined by substituting I0 into Eq. (8). For the flux correction factor experiments, dead-end filtration cell with radius of 12 mm (membrane area ¼ 4.524 104 m2) was used to determine CFS-MFIUF (6:1).
Membranes
Ultrafiltration membranes with MWCO of 10 kDa were used in the dead-end filtration cell for the purpose of capturing fine particulates. The CFS membrane is a “non-retentive” filter with straight through pores. Four different pore sizes of membrane filters namely 5 mm, 11 mm, 40 mm and 100 mm were compared and used in this study. The specifications of all types of membranes used are given in Table 1.
3.3.
Feed water
Silica was selected as a representative of colloidal foulant. Three types of silica colloidal suspensions were used as the feed solution in most of the experiments. They were monodispersed 22 nm (LUDOX TM-50, Sigma Aldrich), mixed particles sizes of 70e100 nm silica colloidal (SNOWTEX-ZL, Nissan Chemical Industries Ltd) and 3 mm (Sigma Aldrich). Silica suspensions were prepared by adding Milli-Q to the desired concentration and underwent ultrasonic treatment for 5 min to ensure that the suspension was stable and without large aggregates. In order to ensure that the pH of the feed remained constant throughout the experiment, a buffer solution of pH 8 was used as the background solution. The
4.
Results and discussion
4.1.
MFI measured at constant pressure mode
4.1.1.
Effect of membrane pore size on MFI value
Using constant pressure of 207 kPa, the effect of different membrane pore sizes on MFI values was investigated. The membranes used for comparison ranged from 0.45 mm down to 10 kDa are presented in Table 2. Meanwhile, the feed suspension used was 50 ppm of 3 mm silica. As can be observed, MFI values increased with decreasing membrane pore size. One possible reason was due to the presence of small particles in the solution. This trend was first demonstrated by Schippers et al. (1981) who showed that MFI value of the pre-treated river Rhine water increased sharply as membrane pore size decreased. They suggested that this may be attributed to the presence of fine particles in the pretreated river Rhine water. Small particles which could not be retained by larger pore size membrane were not incorporated in the MFI measurement and hence the fouling propensity of the water was underestimated. In order to verify this possibility, the particle size used in this study was analysed using Malvern Mastersizer. The particle size distribution ranged from 1 mm to 11 mm with the average particle size of 4.12 mm. The results also indicated that the smallest particle size for this solution was 1 mm, suggesting that all particles in the solution can be retained by 0.45 mm membrane. Hence, the increase in MFI value with decreasing membrane pore size in current study cannot be due to the retention of small particles.
Crossflow Sampler (A1, J1)
Computer
PG
CFS permeate (Q1)
Centrifugal Pump
Flow meter Peristaltic Pump
PT
Valve 1
Dead-end cell (A2, J2) Valve 2
Dead-end permeate (Q2)
Cooling coil
Feed
3.2.
buffer solution was a mixture of 6.81 g of KH2PO4 and 461 ml of 0.1 M NaOH. Besides colloids, humic acid was also used as a representative of organic foulant. The concentration of humic acid samples was characterised in terms of their total organic carbon (TOC) concentration using Shimadzu TOCVCSH. Two different dynamic light scattering particle distribution detectors (NanoDLS, Brookhaven Instrument Corporation, and Malvern Mastersizer) were used to investigate the particle size distribution of the suspension under different conditions.
Water bath Balance PT
Pressure Transducer
PG
Pressure Gauge
Fig. 1 e Experimental rig for CFS-MFIUF measurement.
1643
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 3 9 e1 6 5 0
Table 1 e Specifications of different types of membrane used in this study. Membrane type
Membrane pore size
Material
Net filter
5 mm 11 mm 41 mm 100 mm
Hydrophilic Hydrophilic Hydrophilic Hydrophilic
Microfiltration membrane (MF)
0.45 mm 0.22 mm
Ultrafiltration membrane (UF)
10 kDa 30 kDa 100 kDa
% Porosity 15 6 31 44
GE Osmonics Millipore Corporation Millipore Corporation Millipore Corporation
Hydrophilic PVDF Hydrophobic PVDF
70 75
Millipore Corporation Millipore Corporation
Polyethersulfone Polyethersulfone Polyethersulfone
e e e
Omega by Pall Corporation Omega by Pall Corporation Omega by Pall Corporation
Another explanation for the dependency of MFI value on membrane pore size might be the effect of membrane surface morphology upon cake resistance. Membranes with low surface porosity and irregular distribution of pores were found to promote higher cake resistances as suggested by Fane et al. (1991). In their study, a comparison of cake resistance was made between two iso-porous membranes with the same pore size but different in porosity and pore size distribution for the filtration of Escherichia coli suspension. They suggested that irregular distribution of pores on the membrane surface could cause higher local fluxes, and hence lead to a higher cake resistance. Recent studies by Boerlage et al. (2002) found similar phenomena where MFI-UF was dependent of MWCO (1e100 kDa) varying from 2000 to 13,300 s/L2 when heterogeneously porous and low surface porosity membranes were used. Boerlage et al. stated that low surface porosity would decrease the effective filtration area; hence based on Eq. (7) artificially high MFI-UF values were expected. In the same study, when Polyacrylonitrile (PAN) membranes with different MWCO (6, 13, 50 kDa) were used, MFI-UF was found independent of MWCO varying to only a small extent from 2000 to 2800 s/L2. FESEM images of these three MWCO membranes revealed that these membranes had high surface porosity and were homogeneously porous. The results tabulated in Table 2 clearly show that the MFI values for microfiltration membrane were markedly lower than ultrafiltration membrane. MF membranes which have higher surface porosity and larger pore size than UF membrane would increase the effective filtration area, and hence lead to lower MFI value. By the use of the Carman Kozeny relationship as shown in Eq. (9), cake porosity (3) for
Polycarbonate Nylon Nylon Nylon
Suppliers
the corresponding filtration can be estimated based on specific cake resistance (a), particle density (rp) and particle diameter (dp) a¼
180ð1 3Þ rp d2p 33
(9)
The results suggest that cake formed during MF is more porous than that formed on UF membrane with the porosity value lying between 0.3 and 0.4. For rigid spherical particles, the void fraction (3) of a randomly packed cake is about 0.4 (Belfort et al., 1994). Therefore, this phenomenon can be explained by the random deposition of particle at higher filtration flux during MF filtration. The much lower cake porosity for the low MWCO UF membranes could be due to two factors. Firstly, as explained above the more sparsely surface porous membranes tend to overestimate specific cake resistance. Secondly, the lower fluxes allow formation of a more regular cake and the possible ‘infiltration’ of the smallest particles in the feed into the existing cake deposit. In summary, in this set of experiments, the dependency of MFI on membrane pore size was not due to the retention of fine particles, but membrane surface morphology. However, it is believed that fine particles are often responsible for the membrane fouling in RO, thus UF membrane is often recommended as the reference membrane in MFI measurement, known as MFI-UF.
4.1.2.
Effect of operating pressure on MFI-UFconst.pressure
Colloidal silica (22 nm, Sigma Aldrich) was employed to investigate the dependency of MFI-UFconst.pressure on applied pressure. In Fig. 2, no obvious trend can be observed between MFI-UFconst.pressure and applied pressure from 25 to 200 kPa.
Table 2 e Comparison of MFI for different membranes and their corresponding cake properties. Type of membrane MF MF UF UF UF
Membrane pore size (mm)/MWCO (kDa) 0.45 mm 0.22 mm 100 kDa 30 kDa 10 kDa
Membrane material PVDF PVDF PES PES PES
PVDF e Polyvinylidene fluoride; PES e Polyethersulphone.
Clean membrane resistance (m1)
MFI (s/L2)
Specific cake resistance, a (m/kg)
Cake porosity
1.61 1010 5.21 1010 1.45 1011 5.68 1011 7.51 1011
3.90 7.25 25.49 162.79 292.80
7.25 1010 1.34 1011 4.68 1011 2.97 1012 5.33 1012
0.42 0.35 0.24 0.14 0.11
1644
5
34
a
b
4 33
3
ln
MFI-UFconst.pressure x104 (s/L2)
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 3 9 e1 6 5 0
2
y = 1.0053x + 21.017
32
1 0
31 0
100
200
10
11
12
13
ln P
Applied Pressure (kPa)
Fig. 2 e (a) Effect of applied pressure on MFI-UFconst.pressure; (b) Compressibility check for 50 ppm of mono-size 22 nm silica suspension.
3
4.2.
MFI-UF measured at constant flux mode
4.2.1.
Effect of applied flux on MFI-UFconst.flux
As most industrial RO membrane systems are operated under constant flux mode, the MFI-UF measured under constant pressure might not be able to adequately reflect the fouling potential of the feed due to the difference in hydrodynamic behaviour. Hence, the effect of applied flux on MFI-UF was studied. The feed solution used in these experiments was 50 ppm of 22 nm colloidal silica solution. The results obtained are shown in Fig. 4. MFI-UFconst.flux was found to increase as applied flux increased. At a high applied flux, due to the great permeate drag force, more particles can easily deposit on the membrane surface. These particles that deposit on the membrane surface contributed significantly to the major resistance and hence leading to the higher values of MFI-UFconst.flux. Similar trend was observed by Boerlage et al. (2004) when investigating the effect of flux on MFI-UFconst.flux using tap and canal water. Membranes tend to foul more rapidly during high applied flux operation than in low flux operation which is why low to moderate applied flux was chosen in membrane water industry. As MFI-UFconst.flux is sensitive to operating flux, MFIUFconst.flux measured under different fluxes is not directly comparable. In order to more closely simulate the fouling behaviour in RO, the operating flux of MFI-UFconst.flux should be the same as that in the real RO filtration, which is around 35.0
a
b
2.5 34.5
2 1.5
ln
MFI-UFconst.pressure x 106 (s/L2)
This observation can be explained by using Eq. (6). When compressibility u equals to 1, MFI-UFconst.pressure becomes independent of applied pressure. In order to verify this, the compressibility factor was obtained by plotting ln a against ln DP where compressibility factor is the slope of the line. Compressibility factor for the solution used is approximately 1 as demonstrated in Fig. 2b. Hence, MFI-UFconst.pressure was proven to be independent of applied pressure for this type of suspension. Khirani et al. (2006) observed similar trend, MFI measured at 2 bar was similar to that measured at 4 bar using NOM solution (Bioiberica). The authors interpreted this phenomenon by the high compressibility factor of the cake with the value close to 1. For solutions that are less compressible (u < 1), the MFI-UFconst.pressure is expected to dependent on operating pressure. A good example is represented in Fig. 3, where a less compressible suspension (Mixture of 50 ppm 22 nm and 50 ppm 70e100 nm silica) with compressibility factor of 0.54 was used. These results indicated that MFI-UFconst.pressure depends on the applied pressure as well as the compressibility of the feed. A pressure correction factor is therefore important when MFIUFconst.pressure is not measured under the reference pressure. Boerlage et al. (2003a) incorporated the compressibility factor to correct the experimentally determined MFI-UFconst.pressure at 0.5 and 1 bar to the reference pressure of 2 bar and observed good agreement between experimentally obtained MFIUFconst.pressure and the corrected MFI-UFconst.pressure with only a slight difference of 5e10%.
34.0
1
y = 0.5435x + 28.038
0.5 0 0
50
100
150
Applied Pressure x 10 3 (kPa)
200
33.5 10.5
11.0
11.5
12.0
12.5
ln P
Fig. 3 e (a) Effect of applied pressure on MFI-UFconst.pressure; (b) Compressibility check for 50 ppm 22 nm and 50 ppm 70e100 nm mixture solution.
1645
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 3 9 e1 6 5 0
foulants formed under this case should be less than that formed under constant flux of 121 L/m2 h. Thus, the cake resistance under flux of 28 kPa is expected to be less resistive than that formed under constant flux of 121 L/m2 h. This section highlighted the fact that a direct comparison between MFI-UFconst.flux and MFI-UFconst.pressure cannot be made as this may result in misleading information on the fouling propensity of the given feed.
MFI-UFconst.flux (s/L2)
8000
6000
4000
2000
4.3.
Study of CFS-MFIUF
0 0
50
100
150
Applied Flux (L/m2.h)
Fig. 4 e Effect of applied flux on MFI-UFconst.flux; Feed [ 50 ppm of 22 nm silica colloidal in buffer.
20e30 L/m2 h. However, this prolongs the duration of the test substantially. Section 4.5.3 will further consider this issue and proposed a method to streamline the test.
4.2.2. Comparison between MFI-UF constant flux and MFI-UF constant pressure From Figs. 2 and 4, it is obvious to note that MFI-UFconst.flux showed lower values than MFI-UFconst.pressure. MFI-UFconst.flux has values ranging from 1000 s/L2 to 8000 s/L2, whilst under constant pressure, MFI-UFconst.pressure lies between 30,000 s/L2 and 40,000 s/L2. It should be noticed that the values of MFIUFconst.flux have been scaled to the standard pressure of 2 bar based on Eq. (7), whilst MFI-UFconst.pressure was determined based on its operating pressure. Thus, the values of the resistivity, I were calculated. I values for constant pressure operation ranged from 3.8 1012 to 2.52 1013 m2 while, for constant flux experiments, I lies between 1.45 1012 and 6.2 1012 m2. Two experiments with similar initial fluxes and initial pressures obtained at the start of the experiments were compared and tabulated in Table 3. MFI-UFconst.flux is lower than MFI-UFconst.pressure, but the resistivity value of constant flux is higher than the one from constant pressure. The difference between the resistivity values indicates cakes formed under these two conditions varied significantly. The cake formed under constant flux of 121 L/m2 h has higher I value and therefore higher specific cake resistance than the cake formed under constant pressure of 28 kPa. The most likely explanation for this observation is that the operating flux of 121 L/m2 h lies above the critical flux zone, severe fouling could have occurred instantly at the very early stage of the filtration. Under constant pressure mode, the flux declined from 116 L/m2 h to lower a permeate flux. The amount of
RO systems are usually operated under constant flux mode, hence, our focus is an MFI-UFconst.flux but in combination with crossflow sampler (CFS), which is known as CFS-MFIUF.
4.3.1.
Effect of CFS membrane pore size on CFS-MFIUF
As mentioned earlier a crossflow sampler (CFS) with large membrane pore size is implemented before the dead-end filtration cell in order to simulate the hydrodynamic conditions of the crossflow RO. Due to the selective deposition in the crossflow cell, large particles are unlikely to migrate near to the crossflow cell membrane surface, whereas small particles do. These particles which are able to permeate through the CFS are the foulants that represent the composition that is most likely to cause fouling in a RO crossflow system when the same feed is being used. In order to ensure that the membrane in the crossflow sampler is permeable to any foulants that approach it, the membranes must have large pore, high porosity, and ideally they would be straight through pores. This leads to the investigation of the effect of CFS membrane pore sizes on CFS-MFIUF. Four types of MF membranes with pore size ranging from 5 mm to 100 mm were investigated. The difference of CFS-MFIUF values between these four types of membrane was insignificant, with values ranging from 54,000 to 57,000 s/L2 (standard deviation of 2544 s/L2). These results imply that for the feed used in this study, membranes with pores R 5 mm can be used as the CFS membrane as the resulting CFS-MFIUF values are similar. The selection of CFS membrane could be based on the pre-treatment method used prior to RO. For example, if microfiltration membrane (5 mm) was used as the pre-treatment stage, then any membrane with higher pore size than 5 mm can be used as CFS membrane as long as no cake build up occurs to block pores.
4.3.2.
Effect of crossflow velocity on CFS-MFIUF
The mechanism of particle deposition during crossflow filtration not only depends on particle size itself, but also the hydrodynamic conditions such as crossflow velocity and filtration flux (Lu and Ju, 1989; Chellam and Wiesner, 1998;
Table 3 e Flux, TMP and MFI-UF value at their corresponding operating mode. Constant flux Operating flux (L/m2 h) 121.11
Constant pressure Initial TMP (kPa)
MFI-UFconst.flux (s/L2)
Resistivity, I (m2)
Operating TMP (kPa)
Initial flux (L/m2 h)
MFI-UFconst.pressure (s/L2)
26.7
7043
6.20 1012
28
115.71
30,730
Resistivity, I (m2) 3.81 1012
1646
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 3 9 e1 6 5 0
Fischer and Raasch, 1985). As for this reason, the potential effect of crossflow velocity on CFS-MFIUF was studied. Fig. 5 shows the trend of CFS-MFIUF as a function of crossflow velocity. Clear minimum values of CFS-MFIUF for both model feed and raw seawater were observed. The results indicated the competing effects of two phenomena: (1) Increase in crossflow velocity generates surface shear which tends to reduce the load of particles reaching the CFS surface. In this case Cb in Eq. (2) is decreased. Hence CFS-MFIUF decreases as crossflow velocity increases. (2) Increase in crossflow velocity causes the average particle size in the deposit to decrease. In this case a in Eq. (2) is increased. Consequently, this leads to a more resistant cake structure and therefore CFS-MFIUF values increase. It is suggested that in the low crossflow velocity range, an increased crossflow velocity produced a lower CFS-MFIUF values, dominating by phenomenon 1. The additional surface shear generated near the CFS membrane surface prevented particulates to flow near to the CFS surface and thus reduced the load of particles that permeate through the CFS unit. In fact, the reduction in cake mass as crossflow velocity increased has already been reported. Baker et al. (1985) observed that the cake mass during the crossflow filtration process is strongly dependent on crossflow velocity at which they correlated their cake loading data (mc, kg/m2) with the crossflow velocity (v, m/s),
a
100000
CFS-MFIUF (s/L2)
80000
60000 Phenomenon 2
Phenomenon 1
40000
20000
0 0
b
0.2
0.4
0.6
0.8
3000
CFS-MFIUF (s/L2)
2500 2000 1500 Phenomenon 2
Phenomenon 1
1000 500 0 0
0.2
0.4
0.6
0.8
1
Crossflow velocity (m/s)
Fig. 5 e CFS-MFIUF at different crossflow velocities (a) Mixture of 50 ppm 22 nm and 50 ppm 70e100 nm silica colloidal in buffer (b) raw seawater. Conditions: Crossflow flux [ 22.5 L/m2 h; Dead-end flux [ 55 L/m2 h.
mc ¼
0:17 v2
(10)
Based on this correlation, the cake mass reduces as the crossflow velocity increases. Notwithstanding this, the decrease in the cake mass is expected to reach a stable cake mass above a certain velocity. No significant cake mass was observed above crossflow velocity of 4 m/s in experimental results illustrated by Baker et al. (1985). Recently, Chong et al. (2008) who studied the crossflow RO fouling behaviour using silica colloids observed that the amount of particle that deposited on the membrane surface decreased as crossflow velocity increased. They characterised the amount of particles convected to and finally deposited on the membrane surface in term of fractional deposition constant (f), which varies between 0 and 1. A value of 1 means all particles are deposited on the membrane surface during the crossflow filtration process, while f ¼ 0 means no particle deposition. In their study, the fractional deposition constant was about 1 at low crossflow velocity (shear rate < 430 s1; v < 0.1 m/s), whereas this factor dropped to about 0.1 at high shear rate (shear rate > 940 s1; v > 0.23 m/s). They observed that the fractional deposition constant remained at a value of 0.1 when the crossflow velocity reached above 0.23 m/s. This result further confirmed that the decrease in mc may reach a stable value above a certain crossflow velocity during crossflow filtration. As shown in Fig. 5a, CFS-MFIUF appears to increase above crossflow velocity of 0.37 m/s. The rise of the CFS-MFIUF can be explained by the dominating effect of phenomenon 2 as well as the diminishing effect of phenomenon 1. The influence of the crossflow velocity on the particle size distribution of the CFS permeate was investigated using NanoDLS and is illustrated in Fig. 6. As can be seen, when the crossflow velocity increased, the particle distribution in the CFS permeate suspension tended to shift to lower average particle sizes. Such decrease in average cake particle size consequently raised the values of the specific cake resistance (i.e. CFS-MFIUF value). This phenomenon has been reported consistently in several studies (Lu and Ju, 1989; Chellam and Wiesner, 1998; Fischer and Raasch, 1985; Meier et al., 2002; Tanaka et al., 2001; Baker et al., 1985). A similar trend was observed when using raw seawater, with a minimum CFS-MFIUF occurring at crossflow velocity of 0.57 m/s, indicated in Fig. 5b. The seawater was obtained from the Sydney area and has been pre-filtered with 40 mm filter. The minimum values of CFS-MFIUF for silica suspension and seawater occurred at different crossflow velocities. This is due to the difference in particle size population as well as the composition in both feeds. In summary, crossflow velocity is an important factor that may affect the measurement of CFS-MFIUF. The selection of operating crossflow velocity in CFS-MFIUF measurement relies on the feed flowrate in the actual RO system.
4.3.3.
Influence of spacers on CFS-MFIUF
To further illustrate CFS-MFIUF can successfully capture the hydrodynamic behaviour during the crossflow filtration, the influence of spacers on CFS-MFIUF was studied. CFS-MFIUF obtained without the presence of spacers has the value of
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 3 9 e1 6 5 0
capture smaller particle sizes and the cake formed is expected to be denser, the effective thickness as well as the amount of the cake is still relatively small compared to that formed in MFI-UFconst.flux measurement. This is because MFI-UFconst.flux is measuring the effect of all components of the feed, so the cake formed on MFI-UFconst.flux membrane consists of both large and small particles.
120
Multimodal size distribution
Feed 100
0.26m/s 0.37m/s 0.47m/s
80
1647
60 40
4.5. Effect of crossflow sampler and dead-end permeate flux ratio on CFS-MFIUF
20 0 0
50
100
150
200
250
Particle size (nm)
Fig. 6 e Intensity weighted particle size distribution curves at different crossflow velocities. Conditions: Flux [ 55 L/ m2 h; Feed [ Mixture of 50 ppm 22 nm and 50 ppm 70e100 nm silica colloidal in buffer. Particle size distribution of feed and at crossflow velocity of 0.37 m/s have been previously presented in Sim et al. (2010).
One of the special features of CFS-MFIUF method is that the dead-end MFI unit is directly connected with the crossflow sampler. As such, the permeate flowrates for both devices are the same. The CFS permeate flowrate, Q1 is the product of CFS permeate flux and CFS membrane area A1. Q1 ¼ J1 A1
(11)
Similarly, the permeate flowrate for dead-end unit is Q2 ¼ J2 A2
(12)
Since, Q1 ¼ Q2, 62,500 s/L2 which was about 12% higher than that obtained with spacers. This is mainly because with the presence of spacers, greater shear forces were generated compared to the channel without spacers, thus hindering the particle migration towards the CFS membrane surface. Consequently, the load of particles in the CFS permeate (Cb) is less when compared to the situation without spacers.
4.4.
Comparison of CFS-MFIUF and MFI-UFconst.flux
In order to confirm that the crossflow sampler actually has influence on the MFI-UF values, a comparison was made between CFS-MFIUF and MFI-UFconst.flux using different types of feed. Fig. 7 shows consistent trends. The CFS-MFIs were lower than the MFI-UFconst.flux with differences ranging from 20% to 30%. This trend can be anticipated. Due to the hydrodynamic conditions in the crossflow filtration, the foulants permeate through the crossflow sampler selectively, hence the feed that enters the dead-end MFI assessment device differs from the original feed, containing small particle size range components and less foulants load. Even though CFS-MFIUF is expected to
100000 Fouling Index (s/L2)
MFI-UFconst.flux 80000
CFS-MFI-UF
60000 40000 20000
J1 A2 ¼ J2 A1
(13)
As A1 and A2 are fixed parameters, the selected CFS permeate flux will affect the operating flux in subsequent dead-end unit. In our CFS-MFIUF setup, the membrane ratio (A2:A1) is 1:2. In most our experiments presented earlier, the operating flux at the dead-end unit was 55 L/m2 h, thus the respective crossflow flux was approximately 28 L/m2 h. Since CFS and dead-end cell are interconnected, varying the operating flux in J2 subsequently leads to different J1. To single out the potential effect of each unit’s flux, a series of experiments was performed with either fixed value of J1 or J2. To achieve this, the experiments in Section 4.5.1 and 4.5.2 were conducted using the setup at which dead-end cell was disconnected from the CFS unit.
4.5.1. Variation in crossflow sampler permeate flux (J1) but fixed dead-end permeate flux (J2) In this set of experiments, permeate from the crossflow cell was first collected at different crossflow fluxes, then the permeate was used to measure fouling index at the same dead-end flux (55 2 L/m2 h). The feed solution used in these experiments was the mixture solution of 50 ppm 22 nm and 50 ppm 70e100 nm silica colloidal. Crossflow velocity was maintained at 0.37 m/s and with constant temperature of 22 C. The corresponding CFS-MFIUF values at different crossflow permeate fluxes are plotted in Fig. 8. CFS-MFIUF values increased as the permeate flux of crossflow sampler increased. This is an expected trend because at higher crossflow sampler permeate flux, more particles were captured and permeate through the CFS membrane.
0 Silica only
HA only (5ppm TOC)
Silica + HA (5ppm TOC)
Silica + NaCl(1000ppm)
Fig. 7 e Comparison of CFS-MFIUF and MFI-UFconst.flux using different type of feeds. Conditions: Flux [ 55 L/m2 h; Crossflow velocity [ 0.37 m/s.
4.5.2. Variation in dead-end permeate flux (J2) but fixed crossflow sampler permeate flux (J1) In the case of constant crossflow permeate flux at 28 L/m2 h and altering the dead-end permeate flux, the results are plotted given in Fig. 8. It is evident that, CFS-MFIUF also
1648
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 3 9 e1 6 5 0
60000
40000
20000
0 0
20
40
60
80
100
120
Applied flux (L/m2.h) Crossflow permeate flux (Section 4.5.1)
Dead-end permeate flux (Section 4.5.2)
Fig. 8 e CFS-MFIUF at different crossflow permeate fluxes (:) and at different dead-end permeate fluxes ( ). Conditions: Crossflow velocity [ 0.37 m/s; Feed [ mixture of 50 ppm 22 nm and 50 ppm 70e100 nm silica colloidal in buffer.
increased as dead-end flux increased. Similar explanations as given in previous section can be made in this case.
4.5.3.
Flux correction factor
Previous results in Sections 4.5.1 and 4.5.2 not only signify that the CFS-MFIUF is highly dependent on the permeate fluxes of both units, but also highlight the importance of the choice of membrane area ratio. Since the permeate fluxes in both units are affecting the CFS-MFIUF, one option is to have the same membrane areas, so CFS and dead-end permeate flux are the same. The purpose of the CFS is to simulate the hydrodynamic conditions in the RO unit. To be comparable to the RO flux operated in the industry which is approximately 20e30 L/ m2 h, both the dead-end and CFS permeate flux should also maintain at this range of fluxes. However, such a low operating flux would lead to a very slow fouling rate in the deadend MFI assessment unit, and thus an extended period for measurement of CFS-MFIUF values. In order to accelerate MFI measurement, we proposed to reduce the dead-end filtration area, A2 to approximately 1/6 of A1. As such, when CFS is operated at flux of 20 L/m2 h, the corresponding dead-end flux is 120 L/m2 h. However, in doing so, the CFS-MFIUF obtained at high flux (120 L/m2 h) may not be able to adequately represent the fouling behaviour at flux of 20 L/m2 h. Even though the fouling potential of different feeds can still be compared through this accelerated formed of CFSMFIUF, it can only act as a fouling propensity indicator for the feed. The RO fouling rate prediction cannot be made based on this accelerated CFS-MFIUF due to the permeate flux difference. Therefore, in this paper, we intend to investigate if a flux correction factor does exist to correlate CFS-MFIUF measured under this accelerated condition with CFS-MFIUF determined under low flux condition. Different types of feed solution such as humic acid and real water were used in a series of experiments to investigate if the flux correction factor exists for specific operation conditions regardless of the types of feed. For each type of feed, two experiments with different flux ratios (J1 ¼ 20 L/m2 h and
10000 7.5ppm Humic acid
CFS-MFIUF (s/L2) [1:1]
CFS-MFIUF (s/L2)
80000
J2 ¼ 20 L/m2 h vs J1 ¼ 20 L/m2 h and J2 ¼ 120 L/m2 h) were conducted. In determination of CFS-MFIUF 6:1 (i.e. J1 ¼ 20 L/m2 h and J2 ¼ 120 L/m2 h), a new dead-end cell with membrane area of 4.524 104 m2 was used. For CFS-MFIUF 1:1 (i.e. J1 ¼ 20 L/ m2 h and J2 ¼ 20 L/m2 h), both the CFS and dead-end fluxes were operated at 20 L/m2 h. The results were plotted in Fig. 9. Fig. 9 shows good correlation (R2 ¼ 0.92) for different feeds with a wide range of MFIs. Based on these data a flux correction of about 0.67 can be applied when converting CFS-MFIUF 6:1 to CFS-MFIUF 1:1. The flux correction factor is particularly important when conducting the RO fouling rate prediction based on CFS-MFIUF. This is because CFS-MFIUF 6:1 is measured under accelerated flux of 120 L/m2 h, whereas the RO operating flux is 20 L/m2 h. When predicting fouling rate of a RO unit, the obtained CFS-MFIUF 6:1 has to be multiplied by this correction factor before applying it to the fouling prediction model discussed in our previous work (Sim et al., 2010). One of the major advantages of using this approach is the duration of the test. The test required for TMP to increase by 3 kPa took at least 15 h for raw seawater when using the CFSMFIUF 1:1 while, CFS-MFIUF 6:1 took only 2 h for the TMP to increase to 15 kPa using the same type of feed water. Another set of experiments correlating different flux ratios were also conducted, a good correlation (R2 ¼ 0.97) between these fluxes regardless of the types of feed was also observed (figure not included). Therefore, we believed that there exists a flux correction factor that can be used to correlate different fluxes. However, further work is required to confirm the more general applicability of this correction factor. In spite of the benefits mentioned, the major drawback of this approach is that the flux correction factor is hydrodynamic specific. Both the permeate flux and crossflow velocity can affect the particle deposition during the crossflow filtration. Permeate flux can affect the cake composition in the crossflow filtration system. Several researchers have indicated that lower filtration rate induces finer deposit particles permeate flux (Baker et al., 1985; Knutsen and Davis, 2006). Furthermore, as previously shown in Section 4.3.2, CFS-MFIUF is dependent on the crossflow velocity of the system. The crossflow velocity not only determines the size of particles that being captured but also the quantity. Hence the correction
J = 20L/m .h; J = 20L/m .h
100000
8000
25ppm 70-100nm silica
6000
2.5ppm Humic acid
4000 y = 0.6738x R² = 0.9202
Seawater Batch B
2000 Seawater Batch A
Dam Water
0 0
2000
4000
6000
8000
10000
12000
14000
CFS-MFIUF (s/L2) [6:1] J = 20L/m .h; J = 120L/m .h
Fig. 9 e Relationship between CFS-MFIUF obtained at different membrane area ratios [1:1 and 6:1] using different feed solutions. Conditions: crossflow velocity [ 0.37 m/s.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 3 9 e1 6 5 0
factor might vary at different crossflow velocities or various permeate fluxes. In our study, the flux correction factor was determined under crossflow velocity of 0.37 m/s and for correlating fluxes between 120 L/m2 h and 20 L/m2 h. The changes in either crossflow velocity or the flux may lead to a different value of this factor. As a result, in real RO operation, a new calibration on flux correction factor must always be conducted in accordance to the plant operating flux and crossflow velocity.
5.
Concluding remarks
Factors affecting the performance of MFI-UFconst.pressure, MFIUFconst.flux and CFS-MFIUF were discussed. The study began with the investigation of fouling index that was obtained under constant pressure mode. The dependency of MFIUFconst.pressure on the applied pressure was sensitive to the compressibility of the feed. If the compressibility factor was less than 1, then MFI-UFconst.pressure decreased as the applied pressure increased, whereas, when compressibility factor equal to 1, MFI-UFconst.pressure became independent to the applied pressure. For the case of MFI-UF measured under constant flux mode, the dependency of MFI-UFconst.flux on the operating flux was investigated. The results indicated that MFI-UFconst.flux increased as the applied flux increased. In addition to investigating different operating modes, the effect of different membrane pore sizes in dead-end cell was studied. The results indicated that MFI-UFconst.pressure was greatly dependent on the pore size due to the difference in membrane surface properties. The effects of operating conditions such as applied flux and crossflow velocity on CFS-MFIUF were also investigated. In the case of different crossflow velocities, CFS-MFIUF was found to decrease gradually as crossflow velocity increased; however when a certain crossflow velocity was reached, CFS-MFIUF started to increase and plateau off. CFS-MFIUF was found sensitive to the applied flux and hence indicated that the importance of selecting appropriate operating flux during fouling index test. The effects of varying either crossflow sampler or dead-end permeate flux on CFS-MFIUF were studied. The results indicated that permeate fluxes of both units affect the CFS-MFIUF values significantly. In this fouling index test, CFS should simulate the hydrodynamic conditions in RO, at which the permeate flux should lie between 20 L/m2 h and 30 L/m2 h. However, this flux is too low for dead-end unit to be fouled quickly. So, in order to accelerate fouling, dead-end membrane area was reduced to have high permeate flux while maintaining CFS permeate flux to be comparable to crossflow in RO units. By accelerating the fouling test, the duration of the test can be shortened from 15 h to 2 h. To quantitatively predict RO fouling rates, the accelerated fouling rate in dead-end cell may not represent the true fouling behaviour in low flux mode, CFS-MFIUF obtained at 120 L/m2 h must be corrected to CFSMFIUF obtained at 20 L/m2 h using a correction factor of 0.67. Since the correction factor used in this study is flux and crossflow velocity specific, a different value of correction factor may be required in accordance to the industry operating conditions.
1649
This study highlighted several conditions to be considered during CFS-MFIUF measurement. With these considerations, CFS-MFIUF might be a more realistic testing protocol suitable for industry in determination of fouling potential of RO feed.
Acknowledgements This project is supported by International Science Linkages (DEST-ISL:CG110188) established under the Australian Government’s innovation statement, Backing Australia’s Ability. Furthermore, this study is in collaboration with the European Union 6th Framework Program Membrane-Based Desalination: An Integrated Approach (MEDINA).
references
Adham, S.S., Fane, A.G., 2008. Crossflow Sampler Fouling Index. Report. National Water Research Institute, California, USA. Almy, C., Lewis, W.K., 1912. Factors determining the capacity of a filter press. Ind. Eng. Chem. 4 (7), 528e532. Baker, R.J., Fane, A.G., Fell, C.J.D., Yoo, B.H., 1985. Factors affecting flux in crossflow filtration. Desalination 53 (1e3), 81e93. Belfort, G., Davis, R.H., Zydney, A.L., 1994. The behavior of suspensions and macromolecular solutions in crossflow filtration. Journal of Membrane Science 96, 1e58. Boerlage, S.F.E., Kennedy, M.D., Aniye, M.P., Abogrean, E.M., Galjaard, G., Schippers, J.C., 1998. Monitoring particulate fouling in membrane systems. Desalination 118 (1e3), 131e142. Boerlage, S.F.E., Kennedy, M.D., Dickson, M.R., El-Hodali, D.E.Y., Schippers, J.C., 2002. The modified fouling index using ultrafiltration membranes (MFI-UF): characterisation, filtration mechanisms and proposed reference membrane. Journal of Membrane Science 197 (1e2), 1e21. Boerlage, S.F.E., Kennedy, M., Aniye, M.P., Schippers, J.C., 2003a. Applications of the MFI-UF to measure and predict particulate fouling in RO systems. Journal of Membrane Science 220 (1e2), 97e116. Boerlage, S.F.E., Kennedy, M.D., Aniye, M.P., Abogrean, E., Tarawneh, Z.S., Schippers, J.C., 2003b. The MFI-UF as a water quality test and monitor. Journal of Membrane Science 211 (2), 271e289. Boerlage, S.F.E., Kennedy, M., Tarawneh, Z., De Faber, R., Schippers, J.C., 2004. Development of the MFI-UF in constant flux filtration. Desalination 161 (2), 103e113. Chellam, S., Wiesner, M.R., 1998. Evaluation of crossflow filtration models based on shear-induced diffusion and particle adhesion: complications induced by feed suspension polydispersivity. Journal of Membrane Science 138 (1), 83e97. Choi, J.-S., Hwang, T.-M., Lee, S., Hong, S., 2009. A systematic approach to determine the fouling index for a RO/NF membrane process. Desalination 238 (1e3), 117e127. Chong, T.H., Wong, F.S., Fane, A.G., 2008. Implications of critical flux and cake enhanced osmotic pressure (CEOP) on colloidal fouling in reverse osmosis: experimental observations. Journal of Membrane Science 314 (1e2), 101e111. Fane, A.G., Fell, C.J.D., Hodgson, P.H., Leslie, G., Marshall, K.C., 1991. Microfiltration of biomass and biofluids: effects of membrane morphology and operating conditions. Filtration & Separation 28 (5), 332e340. Fischer, E., Raasch, J., 1985. Cross-flow filtration. German Chemical Engineering 8, 211.
1650
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 3 9 e1 6 5 0
Green, G., Belfort, G., 1980. Fouling of ultrafiltration membranes: lateral migration and the particle trajectory model. Desalination 35, 129e147. Javeed, M.A., Chinu, K., Shon, H.K., Vigneswaran, S., 2009. Effect of pre-treatment on fouling propensity of feed as depicted by the modified fouling index (MFI) and cross-flow samplermodified fouling index (CFS-MFI). Desalination 238 (1e3), 98e108. Khirani, S., Ben Aim, R., Manero, M.-H., 2006. Improving the measurement of the Modified Fouling Index using nanofiltration membranes (NF-MFI). Desalination 191 (1e3), 1e7. Knutsen, J.S., Davis, R.H., 2006. Deposition of foulant particles during tangential flow filtration. Journal of Membrane Science 271 (1e2), 101e113. Lipp, P., Gorge, B., Gimbel, R., 1990. A comparative study of fouling-index and fouling-potential of waters to be treated by reverse osmosis. Desalination 79 (2e3), 203e216. Lu, W., Ju, S., 1989. Selective particle deposition in crossflow filtration. Separation Science and Technology 24 (7), 517e540. Meier, J., Klein, G.M., Kottke, V., 2002. Crossflow filtration as a new method of wet classification of ultrafine particles. Separation and Purification Technology 26 (1), 43e50.
Romero, C.A., Davis, R.H., 1988. Global model of crossflow microfiltration based on hydrodynamic particle diffusion. Journal of Membrane Science 39 (2), 157e185. Romero, C.A., Davis, R.H., 1991. Experimental verification of the shear-induced hydrodynamic diffusion model of crossflow microfiltration. Journal of Membrane Science 62 (3), 249e273. Schippers, J.C., Verdouw, J., 1980. The modified fouling index, a method of determining the fouling characteristics of water. Desalination 32, 137e148. Schippers, J.C., Hanemaayer, J.H., Smolders, C.A., Kostense, A., 1981. Predicting flux decline of reverse osmosis membranes. Desalination 38, 339e348. Sim, L.N., Ye, Y., Chen, V., Fane, A.G., 2010. Crossflow Sampler Modified Fouling Index Ultrafiltration (CFS-MFIUF) e an alternative Fouling Index. Journal of Membrane Science 360 (1e2), 174e184. Tanaka, T., Yamagiwa, Y., Nagano, T., Taniguchi, M., Nakanishi, K., 2001. Relationship between cake structure and membrane pore size in crossflow filtration of microbial cell suspension containing fine particles. Journal of Chemical Engineering of Japan 34 (12), 1524. Yiantsios, S.G., Karabelas, A.J., 2002. An assessment of the Silt Density Index based on RO membrane colloidal fouling experiments with iron oxide particles. Desalination 151 (3), 229e238.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 5 1 e1 6 5 8
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Exploiting a new electrochemical sensor for biofilm monitoring and water treatment optimization Giovanni Pavanello a,*, Marco Faimali a, Massimiliano Pittore b, Angelo Mollica c, Alessandro Mollica c, Alfonso Mollica a a
Istituto di Scienze Marine e Consiglio Nazionale delle Ricerche (ISMAR-CNR), via De Marini 6, 16149 Genova, Italy e-magine IT Srl, via Greto di Cornigliano 6r, 16152 Genova, Italy c Newlab Snc, via Greto di Cornigliano 6r, 16152 Genova, Italy b
article info
abstract
Article history:
Bacterial biofilm development is a serious problem in many fields, and the existing biofilm
Received 29 June 2010
monitoring sensors often turn out to be inadequate. In this perspective, a new sensor
Received in revised form
(ALVIM) has been developed, exploiting the natural marine and freshwater biofilms elec-
29 September 2010
trochemical activity, proportional to surface covering. The results presented in this work,
Accepted 3 December 2010
obtained testing the ALVIM system both in laboratory and in an industrial environment,
Available online 10 December 2010
show that the sensor gives a fast and accurate response to biofilm growth, and that this response can be used to optimize cleaning treatments inside pipelines. Compared to the
Keywords:
existing biofilm sensors, the proposed system show significant technological innovations,
Biofilm monitoring
higher sensitivity and precision. ª 2010 Elsevier Ltd. All rights reserved.
Biosensor MIC prevention Electrochemically active biofilm Cathodic depolarization
1.
Introduction
Serious wide-range technological problems (corrosion, equipment failure, energy loss, reduced performance and resistance to antimicrobial treatments) can be caused by bacterial biofilm development on any artificial apparatus exposed to natural water (fluid flow systems, water distribution lines, sensors, etc.), with subsequent highly negative economic repercussions (Parr and Hanson, 1965; Whitehouse et al., 1991; Borenstein, 1994; Geesey et al., 1994; Gilbert et al., 1997; Flemming and Shaule, 1994; Schultz and Swain, 2000). In the water lines of industrial plants, for example, large amounts of disinfectants and other chemical substances are
usually employed as a countermeasure against biofilm (Wirtanen et al., 2001; Prince et al., 2002; Maxwell, 2005). Real-time, continuous monitoring of bacterial growth is extremely useful in order to optimize these (and others) biofilm hindering treatments, making possible to apply them as soon as biofilm appears; regarding chemical treatments, biofilm monitoring allows also the optimization of biocides dosage, entailing a reduction of both costs and of biocide treatments environmental impact. This led, in the past years, to the study of different biofilm sensing techniques: measurement of (a) light scattering (Flemming et al., 1998), (b) turbidity (Klahre et al., 1998), (c) electrochemical impedance (Mun˜oz-Berbel et al., 2006; Dheilly et al., 2008), (d ) vibration response of the monitored surface
* Corresponding author. Tel.: þ39 010 6475407; fax: þ39 010 6475400. E-mail address:
[email protected] (G. Pavanello). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.12.003
1652
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 5 1 e1 6 5 8
(Pereira et al., 2008), (e) diffusion limitation (Foret et al., 2010). These techniques are affected by several limitations, since all of them: - cannot discriminate between biological and inorganic fouling; this is a major problem, since these two different kinds of fouling require different treatments; - have a low sensitivity, e.g. cannot detect biofilm initial colonization phases, but only thicker bacterial layers; on the other hand, many biofilm-related problems, such as Microbiologically Influenced Corrosion (MIC), start as soon as the first bacterial spots appears on a surface (Mollica and Trevis, 1976; Dexter and La Fontaine, 1998; Kimio et al., 2002). Moreover, some of the above mentioned studies stopped at the laboratory testing phase, and have never been implemented within a real sensor (c). With the aim of overcoming the limitations presented by the existing sensors, a new device (“ALVIM”) has been developed (see Section 2.1), exploiting the cathodic depolarization induced by biofilm growth on active-passive alloys exposed to natural aerated waters. This phenomenon contributes to explain the higher corrosiveness of a natural water, in comparison with a sterile one, toward the mentioned alloys and less noble materials coupled with them. The cathodic depolarization induced by biofilm growth has been largely studied in the last 20 years and has been observed in different parts of the world, both in seawater and in freshwater (Mollica and Trevis, 1976; Scotto et al., 1985; Dexter and Zhang, 1990; Mattila et al., 1997; Dexter and La Fontaine, 1998; Kimio et al., 2002; Wang et al., 2004; Acun˜a et al., 2006; Dulon et al., 2007; Little et al., 2008). Recently, the electrochemical activity of natural aquatic biofilms was proven to be proportional to the surface area covered by bacteria (Faimali et al., 2008, 2010), therefore measuring the biofilm electrochemical signal (BES, expressed as current density or potential, see Section 2.1) is possible to know, on-line and in real-time, which is the biofilm covering on a surface. The aim of this work was to evaluate the performances of this new biosensor, meant to be used for biofilm monitoring and anti-microfouling treatments optimization in industrial environments. The work followed three subsequent steps: (1) Preliminary sensor characterization, in laboratory, to study the response of this innovative probe to biofilm growth, in controlled conditions. (2) Verify the sensor response to biofilm growth in a pilot reverse-osmosis desalination plant, during ordinary working. (3) Optimization of pipeline chemical cleaning treatments, inside the above mentioned plant, basing on biofilm growth real-time data collected by the sensor. Biofilm, indeed, represents a major problem in this kind of industrial environment, both for microfiltration (MF) modules and for reverse-osmosis (RO) membranes, increasing management costs (chemical treatments, membranes cleaning) and contributing to cause cloggings which can bring to plant stop (Fritzmann et al., 2007; Vrouwenvelder et al., 2008).
2.
Materials and methods
2.1.
ALVIM working principle
As described in detail in recent papers (Mollica et al., 1997; Faimali et al., 2008, 2010), cathodic current density i (E,t), measured at a given time t on a stainless steel (SS) sample exposed to natural seawater and polarized at a fixed potential E, can be described by the relation: iðE; tÞ ¼ i1 ðEÞ þ ½ i2 ðEÞ i1 ðEÞ qðtÞ
(1)
where i1(E ) is the current density measured on the “clean” fraction of the SS surface and i2(E ) is the one measured on the surface fraction q (t) [0 q (t) 1 ] covered by biofilm. Fig. 1A shows, schematically, the evolution of the overall cathodic curve (i Vs. E ) during the gradual development of biofilm on the SS surface: curve 1 describes the oxygen reduction kinetics, i1(E ), measured at the beginning of the exposure to aerated seawater on a clean SS surface, whereas curve 4 shows the cathodic curve measured on an SS sample completely covered by biofilm. Curve 2 and 3 describe the trend of the cathodic current in two intermediate conditions. If, as suggested by 1), the evolution of cathodic current is only due to biofilm evolution, any technique able to signal the gradual cathodic depolarization from curve 1 to 4, in Fig. 1, can be utilized to build sensors which can provide information on biofilm growth. At least two classical techniques can be applied to this purpose: a potentiostatic technique or an intensiostatic one. Following Eq. (1), the potentiostatic technique provides information on biofilm development through the measurement of the cathodic currents on an SS sample polarized at a fixed potential E (Fig. 1B), whereas the intensiostatic technique provides similar information through the measurement of the potentials able to sustain a fixed cathodic current i during biofilm growth (Fig. 1C). The choice of the most suitable technique between the potentiostatic and the intensiostatic one, in a particular condition or environment, can depend on the specific biofilmrelated problem that has to be studied. Potentiostatic polarization was already proved to provide detailed information on the rate of biofilm development (Mollica et al., 1997; Faimali et al., 2008, 2010) from a q value less than 1% up to a complete covering of the SS surface. A possible defect of a sensor based on the potentiostatic technique is that a gradual carbonate precipitation is possible if the high cathodic current requested when biofilm is completely developed (in the order of 50 mA cm2) is sustained for a long time; it causes, in turn, a gradual decrease of the “active” surface of the sensor which must, hence, be periodically restored by acid cleaning. The intensiostatic technique, which can operate at cathodic currents lower than 1 mA cm2, avoids this inconvenient, but provides only the information that a specific biofilm covering threshold (e.g. 10% of the sensitive surface) was reached. Fig. 1C shows, in fact, that the shape of the curve potential Vs. time is similar to a sigmoidal curve which rises rapidly in a relatively narrow time range [t1 < t < t2 ; q(t1)< q < q(t2)],
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 5 1 e1 6 5 8
1653
Fig. 1 e Evolution of the overall cathodic curve (current i Vs. potential E ) during the gradual development of biofilm on a stainless steel surface [A], cathodic currents measured at a fixed potential E [B] and potentials measured at a fixed cathodic current i [C] during biofilm growth.
depending on the selected cathodic current density i. The inflection point of the curve can be used to define the threshold value of biofilm covering signaled by the sigmoidal curve obtained at a given cathodic current.
2.2.
Basing on the illustrated working principle, the ALVIM sensor can work both in potentiostatic and intensiostatic mode, but, given the previously mentioned considerations, the tests presented here have been performed using the intensiostatic mode. This is considered to be the best one for general industrial applications, since in this working mode the sensor requires less maintenance and gives a clear signal when biofilm covering exceeds the given threshold. In this case biofilm threshold was set to 1% of the working electrode, to test the device maximum sensitivity. The ALVIM probe (Fig. 2) is compact, requires little periodic maintenance, and can be adapted to fit different kinds of pipeline plug. The sensor is a three-electrodes system, in which the zinc counter-electrode (CE) plays also the role of pseudoreference (RE). Connected to the zinc and to the stainless steel working electrode (WE), on which biofilm growth is evaluated, there is an acquisition system, composed of three main parts: the first for substratum conditioning, the second for signal transduction and elaboration, the third for data transmission, over local/GSM/GPRS network. This system can be easily scaled and can manage up to several hundreds sensors at a time. The Biofilm Electrochemical Signal (BES, expressed as current density or potential), measured in real-time, at chosen time intervals, is sent to a remote database. It is possible to graphically visualize data and to raise different alarms in case of BES abrupt changes or achievement of a preset threshold value, corresponding to a defined biofilm covering percentage. The flexibility of this system allows to change electrodes dimensions and shapes, to fit different needs.
2.3.
Fig. 2 e ALVIM biosensor.
The biosensor
Biosensor preliminary characterization
Laboratory characterization has been performed at the ISMAR Marine Station (member of the European Network of Marine
1654
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 5 1 e1 6 5 8
Research Institutes and Stations e MARS), located in the Genoa harbor (Italy), in a tank of about 100 L in which seawater was constantly renewed (1.5/2 L min1) with natural water pumped directly from the sea. Three biofilm probe replications have been immersed for about 35 days, during the period SeptembereOctober (1st year), when seawater temperature ranged from 22 to 24.5 C, and BES was automatically registered every hour. After the reaching of the chosen biofilm covering threshold value (1%), probe sensitive surfaces were detached from the sensor to verify the effective biofilm covering.
2.4.
Biofilm covering evaluation
WE surface area covered by bacteria was quantified by means of epifluorescence microscopy and software analysis. After the detachment from the ALVIM probes, each sensitive surface was gently rinsed in seawater sterilized by filtration (Millipore, 0.22 mm pore size), in order to remove unattached cells, then fixed with 2% paraformaldehyde solution for 30 min, and washed in filtered phosphate buffer saline (PBS). Samples were stored at 4 C in PBS, before staining and microscopic analysis. After staining of bacterial cells with DAPI (40 -6-diamidino-2-phenylindole, Sigma) (Takata and Hirano, 1990), samples were observed at 400 magnification using an Olympus BX41 epifluorescence microscope coupled with an UV filter block for DAPI. A digital camera CAMEDIA 5060 (Olympus) was used to acquire 30 images of 67,500 mm2 each, randomly chosen on the surface of each sample. Images were converted to tiff format (RGB colour) and the surface fraction covered by bacteria was measured, on the 30 images, by means of “Image J” software (Rasband, 1997); mean standard error (SE) was then calculated.
2.5.
Biofilm monitoring system in-plant testing
The following experiments took place in a FISIA e Italimpianti pilot reverse-osmosis (RO) desalination plant, located at the ISMAR Marine Station too. The pilot plant drew feeding water (1.7 m3/h) directly from the sea, with a first 100 mm prefiltration and a second 0.1 mm microfiltration (MF). After the MF there was a water storage tank, and then the RO. Three ALVIM probes were installed, by means of threaded locks, in the plant pipelines (Fig. 3): the first in a newly installed pipeline, between prefiltration and microfiltration, where the biofilm was expected to grow sooner; the second between MF and the storage tank, where the biofilm was expected to grow later or never, since all the
Fig. 3 e Scheme of FISIA e Italimpianti reverse-osmosis desalination plant and ALVIM probes disposition.
particles larger than 0.1 mm were filtered and this section was treated every few days with strong cleaning agents (NaClO, NaOH and HCl); the third between the tank and the RO, where the biofilm was expected to grow, after an initial incubation, because the tank, positioned just before this section, represented a possible large-surface bacteria incubator, more suitable than pipeline for bacterial growth, since water flux was slower and temperature could slightly increase. These monitoring positions were therefore chosen to obtain a complete view of plant conditions, with reference to biofilm possible problems. A first 12-days trial (December of the 1st year) was conducted using all the three above mentioned probes. After some months (July of the 2nd year), a second 12-days trial was performed, employing only the first two probes, in the same plant, to better characterize the differences between biofilm growth dynamics before and after MF, verifying, at the same time, the effectiveness of this filtration against biofilm development. During the last days of this period, continuous chlorination (1 ppm) at the water intake was applied. In the course of the testing periods, pipelines pressure ranged from 0.2 to 0.8 bar, and seawater temperature from 11.6 to 16 C during the first trial, and from 23.5 to 27 C during the second trial. As for preliminary characterization, after the reaching of the chosen biofilm covering threshold value (1%), sensitive surfaces were detached from the probes to verify the effective biofilm covering.
2.6.
ALVIM system as a chlorination triggering device
The subsequent step, a few months later (December of the 2nd year), was the optimization of pipelines chemical cleaning treatments inside the above mentioned pilot plant, basing on biofilm growth real-time data collected by the ALVIM system. For this aim, biofilm growth signal from sensor no.1 was employed as a trigger to remotely start the 1 ppm chlorination at the water intake. During the 20-days testing period, pipelines pressure ranged from 0.2 to 0.8 bar, and seawater temperature from 11 to 15 C.
3.
Results and discussion
3.1.
Biosensor preliminary characterization
During preliminary testing, after just a few days of immersion in the seawater tank, biofilm probes showed an increase of the BES from around 500 mV Vs. Zn to more than 1100 mV Vs. Zn (Fig. 4), corresponding to a biofilm covering, quantified by microscopic analysis, of 3e4% of the WE surface. Considering that the sampling has been done some days after the reaching of the threshold value (marked in Fig. 4 by an asterisk), these data fit well with the chosen biofilm covering threshold (1% of the WE surface). These results are consistent both in terms of BES evolution curves (reasonably low standard error among replications and few differences among subsequent repetitions) and of biofilm covering (actual data match expected
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 5 1 e1 6 5 8
Fig. 4 e ALVIM probes BES evolution (mean ± SE, mV Vs. Zn) during preliminary testing in a tank with constantly renewed natural seawater. The asterisks mark the reaching of the chosen biofilm covering threshold (1% of the WE), the arrows mark sensitive surface sampling/ analysis (biofilm covering data are indicated in the boxes, as % ± SE, calculated on the three replications) and replacement.
values), confirming that the ALVIM sensor worked reliably over the considered time period.
3.2.
Biofilm monitoring system in-plant testing
After preliminary experiments, ALVIM testing proceeded in the pilot reverse-osmosis desalination plant (Fig. 5). After 2e5 days, BES of biosensors no.1 and no.3 started to increase, signaling that the biofilm covered more than 1% of WE surface. The BES rise occurred later in the stretch crossed by just prefiltered water (sensor no.1), but with a new and clean pipeline, than in the section after MF, between tank and RO (sensor no.3), never cleaned. This highlights the fact, suggested also by other experimental evidences (Donlan, 2002; Nikolaev and Plakunov, 2007; Sriyutha Murthy and Venkatesan, 2009), that already existing biofilm, inside
1655
pipelines, can have more influence in biofilm propagation/ development than new bacteria transported by feeding water, underlining the importance of an appropriate pipeline periodic cleaning. Sensor no.2, positioned in a section treated every few days with strong cleaning agents, did not show any biofilm growth signal, indeed. Microscopic examination showed that, after nine days of immersion, biofilm covering percentages on sensitive surfaces no.1 and no.3 were, respectively, about 1% and 2% (matching the chosen threshold of 1%). After the sampling, the plant was stopped, the section between MF and water storage tank was chemically cleaned (NaClO, NaOH); sensitive surface no.3 was replaced, while sensor no.1 was disconnected, for technical reasons related to the plant. At the end of the 12-days testing period, sensitive surfaces no.2 and no.3 (the last one replaced after the sampling on day 9) were nearly clean. The lower biofilm covering of sensitive surfaces no.1 and no.3 sampled on day 9, compared to those observed during preliminary testing (see Section 3.1), could be due to the different conditions, such as temperature, light, water filtration. The quick and precise ALVIM system response to biofilm growth, observed in the course of the preliminary characterization, was therefore confirmed during in-plant testing. Moreover, the BES showed a clear peak in correspondence to chlorination (marked in Fig. 5 by an oval), suggesting that ALVIM could also be used to monitor chlorine-based treatments. During the second testing period (Fig. 6), the signal of sensor no.1 started to increase one day earlier than in the first testing and, thereafter, grew at an higher rate. This reflects the different seasons, the different natural biological activity and, likely, possible differences of nutrients load in the water during the two tests, since the first one took place in December, while the second one in July. This observation confirms the flexibility of the employed biofilm monitoring system, underlining at the same time the impossibility of adopting a “one-for-all” cleaning treatment approach.
Fig. 5 e ALVIM probes BES evolution (mV Vs. Zn) during the first test in a pilot reverse-osmosis desalination plant. The arrow marks sensitive surfaces no.1 and no.3 sampling/analysis before plant stop and cleaning of the section between MF and water storage tank; on sensor no.3 a new sensitive surface was installed. The oval marks a chlorination in the same pipeline.
1656
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 5 1 e1 6 5 8
Fig. 6 e ALVIM probes BES evolution (mV Vs. Zn) during the second test in a pilot reverse-osmosis desalination plant. The circles mark chlorinations in the section between MF and water storage tank. The gray area marks plant stop for maintenance. The arrow on day 6 marks sensitive surface no.1 sampling, analysis and replacement. The arrow on day 9 marks the start of continuous chlorination (1 ppm) at the intake (whole pipeline treatment).
The signal of sensor no.2 did not increase over the baseline, except for the peaks observed when the signal was acquired during or nearby the chlorinations of the pipeline in which the sensor was placed. The height of those peaks was related with the proximity of the data acquisition moment with the chlorination time. The bio-electrochemical signal of sensor no.2 confirmed that the cleaning treatments carried out in this pipeline were effective against the biofilm; the bacterial covering of probe no.2 WE, at the end of the testing period, was nearly 0%, indeed. These observations sustain the effectiveness of ALVIM as monitoring device for chlorination treatments, as suggested by the data of the first testing period. During plant stop for maintenance (gray area in Fig. 6), the signal of sensor no.2 showed a drop, because the pipeline in which that sensor was placed got empty (no water). Just before plant restart, probe no.1 WE was sampled and replaced. The biofilm covering percentage was about 1%, fitting well, as the previous data, with the chosen threshold value. On day 9, a chlorination treatment (continuous, at 1 ppm) at the water intake was started, and the signal of sensor no.1, on which the biofilm had grown, nearly immediately decreased, confirming the effectiveness of that treatment. The signal remained 100e150 mV over the baseline, because of the chlorine continuous presence in the water. Chlorine was likely “consumed” before sensor no.2 by the organic matter present inside microfiltration module and pipelines, in fact this sensor did not show any signal increase. As previously mentioned, at the end of the testing period, biofilm covering on probe no.2 WE was nearly 0%.
3.3.
ALVIM system as a chlorination triggering device
During the last testing period, in line with what was observed during the above mentioned tests, after two days of incubation the BES started to grow (Fig. 7), signaling that the biofilm
surface covering, on the probe WE, reached the chosen threshold (1%). On day 4, the first ALVIM-triggered chlorination was started; the treatment time was set to 30 min. Immediately after this chemical cleaning treatment, the BES dropped to the initial value (around 700 mV Vs. Zn). In about two days, the biofilm growth signal increased again, and this time the 30-min chlorination was started in advance with respect to the previous one (the BES was around 900 mV Vs. Zn, while on day 4 it was nearly 1100 mV Vs. Zn). After the treatment, the sensor WE was immediately sampled and replaced. Biofilm covering on the sampled surface, quantified by laboratory analysis, was about 1%, matching the expected value. This evidence highlights the fact that, obviously, the applied chlorination treatment did not imply an immediate detachment of the biofilm from the surface. After WE substitution, it took four days to the BES to start growing again. Chlorination time was extended to 60 min, to verify the requested treatment frequency with a longer treatment time. Chlorination frequency continued to be about one treatment every two days, and, moreover, after the chemical cleaning performed on day 14 the BES did not return to the initial value, showing an increased after-treatment bio-electrochemical activity. This meant that the biofilm was not completely inactivated/killed by the chlorination, and implied also the need of more frequent treatments. Such information is essential to adjust timing and dosage of chemical cleaning, since even the survival of a small part of the settled bacteria is a guarantee of the fact that the biofilm will quickly grow again, thanks to the replication of the microorganisms still alive. These data suggest that, compatibly with environmental impact consideration and by law limit, higher chlorine concentrations or longer treatments had to be used, in this case, to completely remove the biofilm from the internal surfaces of plant pipelines.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 5 1 e1 6 5 8
1657
Fig. 7 e ALVIM probe BES evolution (mV Vs. Zn) during the use of ALVIM system as a chlorination triggering device. The arrows mark pipeline chemical cleaning by means of water chlorination (1 ppm) at the intake.
At the same time, the ALVIM system demonstrated its usefulness for the monitoring of biofilm growth and consequent cleaning treatments optimization inside water pipelines.
3.4. ALVIM compared to the other biofilm monitoring devices Comparing ALVIM sensor to the other biofilm monitoring devices, such as those, discussed in the introduction, based on light scattering (Flemming et al., 1998), turbidity (Klahre et al., 1998), electrochemical impedance (Mun˜oz-Berbel et al., 2006; Dheilly et al., 2008), vibration response of the monitored surface (Pereira et al., 2008) and diffusion limitation (Foret et al., 2010), it possible to see that the experimental results of this new device testing highlight significant advantages: - the possibility of monitoring just the biological fouling (microfouling), discerning it from the inorganic fouling; the above mentioned sensors, indeed, are not able to discriminate between these two different kinds of fouling; - the detection of the biofilm since its first colonization phase (i.e. the first bacterial layer); the above mentioned sensors detect only thicker (several mm) biofilms. Among the sensors based on electrochemical techniques, the most known are BIoGEORGE (Licina and Nekoska, 1993) and BIOX, born, earlier than ALVIM, from the same research activity (Cristiani et al., 1998). They usually show an higher sensitivity, with respect to those based on the previously mentioned techniques, but no clear quantitative data has been found in literature. Considering device flexibility, both BIoGEORGE and BIOX have a fixed working mode and sensitivity, while ALVIM can be set to monitor different extents of biofilm covering and can work both in intensiostatic and in potentiostatic mode. The last one, discussed only marginally in this work, will be the subject of future studies. From the technical point of view, the electronics of BIoGEORGE for the control, data acquisition and data analyses are
housed in an external box, where the readings are stored in a database (DB). In this way data are not available in real-time from remote, but has to be downloaded to a PC. The BIOX sensor needs external hardware too, moreover device control and data reading are basically analogical (Cristiani et al., 1998, 2000; Cristiani, 2005). On the other hand, ALVIM has a fully digital management, and its electronic is completely integrated within sensor housing; in industrial environments, indeed, device compactness represents a valuable advantage. About data storage, ALVIM uses a remote DB, thus the collected information can be viewed in real-time even from remote.
4.
Conclusions
Experimental results show that the ALVIM system works reliably in a real industrial environment, representing an efficient biofilm monitoring solution; it gives a fast and accurate information about the bacterial covering, even at early stages of colonization. Furthermore, the data provided by this system proved to be very useful if applied to cleaning treatments optimization, enabling to hinder biofilm growth as soon as it starts. This is a promising technology in any field affected by biofilm-related problems, prefiguring a wide application range for the ALVIM system. Next biosensor developments will concern longer trials, experiments in different conditions (e.g. freshwater, other industrial environments) and testing of new materials for biosensor components.
Acknowledgements Authors wish to thank FISIA e Italimpianti and University of Genoa e Department of Chemistry and Industrial Chemistry staff for the contribution to the ALVIM system field testing.
1658
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 5 1 e1 6 5 8
references
Acun˜a, N., Ortega Morales, B.O., Valadez-Gonzalez, A., 2006. Biofilm colonization dynamics and its influence of austenitic UNS S31603 stainless steel exposed to Gulf of Mexico seawater. Mar. Biotechnol. 8, 62e70. Borenstein, S.W., 1994. Microbiologically Influenced Corrosion Handbook. Industrial Press Inc., New York, p. 288. Cristiani, P., Mollica, A., Ventura, G., 2000. System for Monitoring Biocide Treatments. U. S. Patent 6,054,030. Cristiani, P., 2005. Solutions to fouling in power station condensers. Appl. Therm. Eng. 25, 2630e2640. Cristiani, P., Mollica, A., Ventura, G., 1998. On-line Monitoring of Biofilm and TRO by a New ENEL’s System, Proceedings of OCEANS ’98, Nice, France, pp. 1507e11. Dexter, S.C., Zhang, H.J., 1990. Effect of Biofilms on Corrosion Potential of Stainless Steel Alloys in Estuarine Waters, Proceedings of the 11th International Corrosion Congress, Florence, Italy, pp. 333e40. Dexter, S.C., La Fontaine, J.P., 1998. Effect of natural marine biofilms on galvanic corrosion. Corrosion 54, 851e861. Dheilly, A., Linossier, I., Darchen, A., Hadjiev, D., Corbel, C., Alonso, V., 2008. Monitoring of microbial adhesion and biofilm growth using electrochemical impedancemetry. Appl. Microbiol. Biotechnol. 79, 157e164. Donlan, R.M., 2002. Biofilms: microbial life on surfaces. Emerg. Infect. Dis. 8, 881e890. Dulon, S., Parot, S., Delia, M.L., Bergel, A., 2007. Electroactive biofilms: new means for electrochemistry. J. Appl. Electrochem. 37, 173e179. Faimali, M., Chelossi, E., Garaventa, F., Corra`, C., Greco, G., Mollica, A., 2008. Evolution of oxygen reduction current and biofilm on stainless steel cathodically polarized in natural aerated seawater. Electrochim. Acta 54, 148e153. Faimali, M., Chelossi, E., Pavanello, G., Benedetti, A., Vandecandelaere, I., De Vos, P., Vandamme, P., Mollica, A., 2010. Electrochemical activity and bacterial diversity of natural marine biofilm in laboratory closed-systems. Bioelectrochemistry 78, 30e38. Flemming, H.C., Shaule, G., 1994. Microbial deterioration of materials e biofilm and biofouling: biofouling (Mikrobielle Werkstoffzersto¨rung e Biofilm und Biofouling: biofouling). Mater. Corros. 45, 29e39. Flemming, H.C., Tamachkiarowa, A., Klahre, J., Schmitt, J., 1998. Monitoring of fouling and biofouling in technical systems. Water Sci. Technol. 38, 291e298. Foret, C., Merlet, N., Chaussec, G., Martemianov, S., Tribolet, B., Hater, W., 2010. Measurement of Biofilm Thickness: An Effective Legionella Risk Assessment Tool. BKG Technical Publication. http://www.bkgwater.com/clients/bkgwater/ upload/fichiers/2010-02-bkgws_publication-measurement-ofbiofilm-thickness.pdf. Fritzmann, C., Lo¨wenberg, J., Wintgens, T., Melin, T., 2007. State-ofthe-art of reverse osmosis desalination. Desalination 216, 1e76. Geesey, G.G., Lewandowski, Z., Flemming, H.C., 1994. Biofouling and Biocorrosion in Industrial Water Systems. CRC Press, Boca Raton, p. 297. Gilbert, P., Das, J., Foley, I., 1997. Biofilm susceptibility to antimicrobials. Adv. Dent. Res. 11, 160e167. Kimio, I., Matsuhashi, R., Toshiaki, K., Miki, O., Kihira, H., Watanabe, K., Baker, P., 2002. Potential Ennoblement of Stainless Steel by Marine Biofilm and Microbial Consortia Analysis Proceedings of Corrosion/2002, Denver, Colorado, Paper No. 403. Klahre, J., Lustenberger, M., Flemming, H.C., 1998. Microbial problems in paper manufacturing. Part III: monitoring (Mikrobielle Probleme bei der Papierfabrikation. Teil III: monitoring). Das Papier 52, 590e596.
Licina, G.J., Nekoska, G., 1993. An Electrochemical Method for OnLine Monitoring of Biofilm Activity Proceedings of CORROSION/93, Houston, Texas, Paper No. 403. Little, B.J., Lee, J.S., Ray, I.R., 2008. The influence of marine biofilms on corrosion: a concise review. Electrochim. Acta 54, 2e7. Mattila, K., Carpen, L., Hakkarainen, T., Salkinoja-Salonen, M.S., 1997. Biofilm development during ennoblement of stainless steel in Baltic sea water: a microscopic study. Int. Biodeterior. Biodegradation 40, 1e10. Maxwell, S., 2005. Controlling Corrosive Biofilms by the Application of Biocides SPE International Symposium on Oilfield Corrosion, Aberdeen, UK. Mollica, A., Trevis, A., 1976. Correlation Between the Formation of the Primary Film and the Change in the Cathodic Reaction on Stainless Steels in Seawater at Speeds Between 0.3 and 5.2m/s (Correlation entre la formation de la pellicule primaire et de la modification de la cathodique sur des aciers inoxydables experimente´s en eau de mer aux vitesses 0.3 a` 5.2m/s), Proceedings of the Fourth Congress on Marine Corrosion and Fouling, Antibes, France, pp. 351e365. Mollica, A., Traverso, E., Thierry, D., 1997. In: On Oxygen Reduction Depolarisation Induced by Biofilm Growth on Stainless Steels in Seawater. European Federation of Corrosion publications N 22. The Institute of Materials, London, pp. 51e65. Mun˜oz-Berbel, X., Mun˜oz, V., Vigue´s, N., Mas, J., 2006. On-chip impedance measurements to monitor biofilm formation in the drinking water distribution network. Sens. Actuators B 118, 129e134. Nikolaev, Y.A., Plakunov, V.K., 2007. Biofilms e ‘city of microbes’ or an analogue of multicellular microorganims? Microbiology 76, 125e138. Parr, J.G., Hanson, A., 1965. An introduction to stainless steel. ASM International, Metals Park, p. 147. Pereira, A., Mendes, J., Melo, L.F., 2008. Using nanovibrations to monitor biofouling. Biotechnol. Bioeng. 99, 1407e1415. Prince, E.L., Muiry, A.V.G., Thomasz, W.M., Stollardx, R.J., Sampsonx, M., Lewis, J.A., 2002. An evaluation of the efficacy of aqualox for microbiological control of industrial cooling tower systems. J. Hosp. Infect. 52, 243e249. Rasband, W.S., 1997. ImageJ. US National Institutes of Health. Bethesda, Maryland. http://rsb.info.nih.gov/ij/. Schultz, M.P., Swain, G.W., 2000. The influence of biofilms on skin friction drag. Biofouling 15, 129e139. Scotto, V., Di Cintio, R., Marcenaro, G., 1985. The influence of marine aerobic microbial film on stainless steel corrosion behaviour. Corros. Sci. 25, 185e194. Sriyutha Murthy, P., Venkatesan, R., 2009. Industrial biofilms and their control. In: Flemming, H.C., Venkatesan, R., Sriyutha Murthy, P. (Eds.), Marine and Industrial Biofouling. Springer, New York, pp. 65e101. Takata, K., Hirano, H., 1990. Use of fluorescein-phalloidin and DAPI as a counterstain for immunofluorescence microscopic studies with semithin frozen sections. Acta Histochem. Cytochem. 23, 679e683. Vrouwenvelder, J.S., Manolarakis, S.A., van der Hoek, J.P., van Paassen, J.A.M., van der Meer, W.G.J., van Agtmaal, J.M.C., Prummel, H.D.M., Kruithof, J.C., van Loosdrecht, M.C.M., 2008. Quantitative biofouling diagnosis in full scale nanofiltration and reverse osmosis installations. Water Res. 42, 4856e4868. Wang, W., Wang, J., Li, X., Xu, H., Wu, J., 2004. Influence of biofilm growth on corrosion potential of metals immersed in seawaters. Mater. Corros. 55, 30e35. Whitehouse, R.L., Peters, E., Lizotte, J., Lilge, C., 1991. Influence of biofilms on microbial contamination in dental unit water. J. Dent 19, 290e295. Wirtanen, G., Salo, S., Helander, I.M., Mattila-Sandholm, T., 2001. Microbiological methods for testing disinfectant efficiency on Pseudomonas biofilm. Colloids Surf. B 20, 37e50.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 5 9 e1 6 6 8
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Efficacy of monitoring and empirical predictive modeling at improving public health protection at Chicago beaches Meredith B. Nevers*, Richard L. Whitman U.S. Geological Survey, Great Lakes Science Center, Lake Michigan Ecological Research Station, 1100 N. Mineral Springs Road, Porter, IN 46304, USA
article info
abstract
Article history:
Efforts to improve public health protection in recreational swimming waters have focused
Received 17 August 2010
on obtaining real-time estimates of water quality. Current monitoring techniques rely on
Received in revised form
the time-intensive culturing of fecal indicator bacteria (FIB) from water samples, but
3 December 2010
rapidly changing FIB concentrations result in management errors that lead to the public
Accepted 6 December 2010
being exposed to high FIB concentrations (type II error) or beaches being closed despite
Available online 13 December 2010
acceptable water quality (type I error). Empirical predictive models may provide a rapid solution, but their effectiveness at improving health protection has not been adequately
Keywords:
assessed. We sought to determine if emerging monitoring approaches could effectively
E. coli
reduce risk of illness exposure by minimizing management errors. We examined four
Fecal indicator bacteria
monitoring approaches (inactive, current protocol, a single predictive model for all bea-
Recreational water quality
ches, and individual models for each beach) with increasing refinement at 14 Chicago
Lake Michigan
beaches using historical monitoring and hydrometeorological data and compared
Swimming
management outcomes using different standards for decision-making. Predictability (R2) of
Risk
FIB concentration improved with model refinement at all beaches but one. Predictive models did not always reduce the number of management errors and therefore the overall illness burden. Use of a Chicago-specific single-sample standarddrather than the default 235 E. coli CFU/100 ml widely useddtogether with predictive modeling resulted in the greatest number of open beach days without any increase in public health risk. These results emphasize that emerging monitoring approaches such as empirical models are not equally applicable at all beaches, and combining monitoring approaches may expand beach access. Published by Elsevier Ltd.
1.
Introduction
In recent years, efforts to improve public health protection in recreational swimming waters have focused on obtaining realtime estimates of water quality. Current monitoring techniques rely on the culturing of fecal indicator bacteria (FIB)d
such as Escherichia coli or enterococcidfrom water samples, a process that requires an incubation time often in excess of the rate of change of bacteria concentrations in the water (Boehm et al., 2002; Whitman et al., 1999). Because of the lapse in results availability, the public are often either unknowingly swimming in contaminated beach water or are prohibited from
Abbreviations: FIB, fecal indicator bacteria; CFU, colony-forming units; MPN, most probable number; IA, inactive monitoring program model; CM, current model; RM, regional predictive model for all study beaches; IM, individual beach predictive model. * Corresponding author. Tel.: þ1 219 926 8336x425; fax: þ1 219 929 5792. E-mail address:
[email protected] (M.B. Nevers). 0043-1354/$ e see front matter Published by Elsevier Ltd. doi:10.1016/j.watres.2010.12.010
1660
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 5 9 e1 6 6 8
swimming in water that meets the public health criteria. Efforts have been focused on two means of correcting this shortcoming: shorten the analytical time for the current indicator or find an alternate, faster way to assess water quality. To accomplish the latter, empirical predictive models have been attempted with various levels of success and application. Predictive models have been suggested by numerous authors as a potential means for minimizing errors in beach closings (Hou et al., 2006; Kim and Grant, 2004; Nevers and Whitman, 2005). These models range from simple models that associate weather conditions with direct bacteria loadingsesuch as rainfall and associated runoff (McPhail and Stidson, 2009) e to more advanced models that integrate multiple hydrometeorological variables (Kim and Grant, 2004). Model accuracy at predicting FIB concentration depends on beach location, instrument accuracy, wealth of available data, and level of effort, but predictive models can be successfully incorporated into beach management (Nevers and Whitman, 2005). Beaches at which models have been attempted tend to be high profile beaches with heavy visitor use (Boehm, 2007; Hou et al., 2006), directly or strongly impacted by a large point source (He and He, 2008), or having frequent swimming advisories (Olyphant and Whitman, 2004). The accuracy or success of a given modeling approach has typically been assessed by analyzing the amount of variation in the target FIB explained by the model, the error explained by the model, or the specificity (the percent of false negatives, or type II errors) and sensitivity (the percent of false positives, or type I errors) of the model. The first two calculations determine the accuracy of the model at predicting all FIB concentrations. The third, error-based calculation is used due to the use of a binary approach in beach management policies: beaches are either open or closed to swimming, depending on where the FIB concentration falls relative to a designated standard (acceptable health risk); errors occur when the predicted concentration is not equal to actual concentration. Errors result in either inadvertent exposure of the public to high concentrations of FIB (type II error) or exclusion of swimmers from water that meets the exposure standard (type I error). More type II errors result in more swimmers exposed to high concentrations of FIB and therefore a higher public health risk; decreasing the instances of type II errors is necessary to increase public health protection. Current water quality standards for freshwater were developed using epidemiological studies and based on historical acceptable illness rate (Pru¨ss, 1998, US EPA, 1986). Within the monitoring guidance, however, some measure of flexibility was provided for beach managers, including choice of application of two mathematical estimates of illness risk, based on the concentration of indicator bacteria (US EPA, 1986). Generally, beach managers have applied the single-sample maximum for an individual water sample because of its ease of use and interpretation (Nevers and Whitman, 2010, US EPA, 1986), but others use the 5-day geometric mean, both of which should theoretically provide equal levels of health protection. In this paper, we examine four potential monitoring approaches with increasing refinement at 14 Chicago beaches: inactive, current monitoring model, use of one predictive model for all beaches, and use of individual predictive models for each beach. Further, we examine alternate applications of
monitoring standards under these four approaches to assess the health and management outcome possibilities. Using historical monitoring and beach attendance data we compare the accuracy of each model with several calculations and also the relative public health protection provided by each of these models. Specifically, we sought to determine whether predictive modeling at Chicago beaches could be used as a monitoring tool to increase public health protection over traditional monitoring practices.
2.
Materials and methods
2.1.
Study site
Chicago beaches in general are not impacted by a major point source of contamination. Urban sewage is regularly discharged through the Chicago River and a series of man-made or modified canals to the Mississippi River. In events of extreme precipitation, the system override leads to sewage being directed to Lake Michigan (<1 per year); all beaches are then preemptively closed to swimming. Sources of FIB at the Chicago beaches are unknown but likely include beach sand, birds, and algae (Whitman and Nevers, 2003; Whitman et al., 2003). Beaches included in the current study were (from north to south) Loyola, Albion, Hollywood, Foster, Montrose, North Avenue, Oak, 12th Street, 31st Street, 57th Street, 63rd Street, South Shore, Rainbow, and Calumet.
2.2.
Beach monitoring data
E. coli monitoring data, measured as most probable number (MPN)/100 ml of water, were obtained from the Chicago Park District for 2000e2004. Beaches were sampled at least five days a week; replicate samples (up to three) were averaged. E. coli concentrations above or below detection limits were set at detection limits after determining that occurrences were rare (Boehm et al., 2002; Whitman and Nevers, 2008). Missing data points for individual beaches were calculated; values were estimated from the nearest 6 values (average of three previous and three subsequent readings). Beach management is a binary decision: if E. coli concentration <235 MPN/100 ml, the beach is open for swimming; if E. coli >235 a swimming advisory is issued. This model assumes that E. coli concentration today ¼ E. coli concentration yesterday. Inaccurate predictions, therefore, result in a type I or type II error (Table 1). Type I errors occur when the model predicts an E. coli concentration >235 when the actual concentration is <235, resulting in an unnecessary swimming advisory. A type II error occurs when the model predicts E. coli concentration <235 when the actual concentration is >235, resulting in swimmers being exposed to high concentrations of indicator bacteria and associated pathogens. A simple characterization is that type I errors are associated with economic losses because swimmers are denied access and type II errors are associated with greater public health risk, as swimming occurs in the presence of excessive FIB concentrations (Rabinovici et al., 2004).
1661
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 5 9 e1 6 6 8
2.3.
Predictive models
Data for developing predictive models were accumulated from several existing hydrometoerological stations, as described in Whitman and Nevers (2008). Predictors included solar insolation and precipitation (24 h); air temperature, barometric pressure, and wave height (mean for 4e10 AM); and day of year. All E. coli results were log-transformed prior to analysis. Predictor parameters were transformed to z-scores prior to regression analyses. Models were developed using multiple linear regression of the available predictors on the independent variable E. coli; Mallows Cp was used to compare resulting models. For individual models, Akaike’s Information Criterion (AIC) was used to select the best fit model for each individual beach. The regional model was developed previously (Whitman and Nevers, 2008) using barometric pressure, wave height, and day of year as the predictors; here, it was applied separately to each study beach. Model comparisons were made using SPSS (Chicago, IL) and SAS (Cary, NC) software: the adjusted coefficient of determination (R2), which describes the amount of variation in E. coli; and percent of type I and type II errors. The four models compared included inactive model (IA): beach is always open to swimming, regardless of microbiological water quality; current model (CM): E. coli concentration for day 1 is used to make a management decision for day 2; regional model (RM): a universal model used for all beaches; and an individual model (IM): a separate predictive model developed for each beach. Comparisons between models for estimating excess illnesses associated with type II errors were only made on days for which there were results for all four models.
2.4.
Calculating illness rates
Overall beach attendance rates were compared for 20 years, and an average per year estimate was calculated (Chicago Park District, unpublished data). With an annual reported range of 14e31 million visitors to all Chicago beaches, 20 million was used as an average. An estimated 91% of all visits, or 18.2 million, were associated with the 14 beaches included in this study. Number of visitors per beach was calculated based on two years for which individual beach data were available (1999 and 2007). It was assumed that 50% of beach visitors had full-body contact with the water as defined by US EPA (Dufour, 1984).
Lower estimates have been published (Rabinovici et al., 2004), but based on data from several Ontario lakes (Seyfried et al., 1985), estuaries (Lepesteur et al., 2006), and marine waters (Dwight et al., 2007; Given et al., 2006), 50% may be conservative for a large freshwater lake. Illness rates for each beach were calculated based on a 100-day swimming season, which assumes equal distribution of visits across the summer (Rabinovici et al., 2004). Illness rate (Y ) was calculated from Dufour (1984): Y ¼ 11:74 þ 9:397log10 ðECÞ
(1)
where Y ¼ the rate of swimming-associated gastrointestinal illness symptoms per 1000 swimmers and EC ¼ E. coli CFU/ 100 ml water. Monitoring standards based on the these epidemiological studies recommend a geometric mean of 126 E. coli CFU/100 ml for five samples over 30 days (Dufour, 1984): an acceptable illness rate of approximately 0.008%. A singlesample limit of 235 CFU/100 ml is also provided, which is within the confidence limits of the calculated geometric mean. These calculations were developed for beaches influenced by a point source (Dufour, 1984), but the sources affecting Chicago’s beaches have not been confidently identified. Using the derived regression equation (Dufour, 1984), we calculated our acceptable illness rate for the 235 CFU as 0.01054%, following Rabinovici et al. (2004). It should be noted that the standards were established using membrane filtration analysis. Chicago uses a defined substrate technique (Colilert; IDEXX, Westbrook, Maine); and while studies have favorably compared the two outcomes (Buckalew et al., 2006), differences in confidence intervals may influence results outcome (Gronewold et al., 2008). A daily excess illness rate was calculated following Given et al. (2006), that is, the number of beach swimmers expected to exhibit symptoms of illness beyond the acceptable 0.01054%. GI ¼ ðY Y0 Þðv=dÞf
(2)
Where Y0 ¼ acceptable illness rate within the monitoring criteria of 10.54/1000, v ¼ number of visitors in a swimming season, d ¼ number of swimming days in the season, and f ¼ percent of beach visitors estimated to have full-body contact with the beach water. The illness regression equation has a y-intercept <0, so calculated illness rates <0 were set at 0 for calculating the mean. There is conflicting research on the association of illness risk and traditional fecal indicators at beaches without a point
Table 1 e Outcome possibilities using the current beach management model: management decision based on previous day’s E. coli concentration. Management outcome
Error
Correct open Correct closed Incorrect open Incorrect closed
None None type 2 type 1
Health risk Loss of Use
Low High High Low
Low High Low High
Health protection Accurate Accurate Overly liberal Overly conservative
Range of reported outcome Range of mean illness rate frequency for for Chicago beaches Chicago beaches (swimmers/1000) 49e78% 3e14% 12e21% 9e17%
2.7e5.0 13.6e15.3 14.4e16.1 4.2e6.4
1662
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 5 9 e1 6 6 8
source (Calderon et al., 1991; Sinigalliano et al., 2010). Questions about the appropriateness of using E. coli as an indicator have also been raised (Harwood et al., 2005; Wong et al., 2009), but a review of epidemiological studies associate elevated E. coli with higher risk of gastrointestinal illness (Marion et al., 2010; Wade et al., 2003). Because alternate standards have not been established, all beaches are managed using the E. coli standard, despite its shortcomings. Chicago does not issue a swim ban until E. coli concentration exceeds 1000 MPN/100 ml. Currently, there are no estimates of the percent of swimmers who enter the water during a swimming advisory (i.e., 235 < E. coli <1000) (Hou et al., 2006). Here, we elect to use >235 to express excess illnesses and define model errors because it is the more conservative calculation. Rate of excess illness would increase by 0.0059% using the 1000 CFU standard.
2.5. Different management applications of monitoring standard In an attempt to improve model performance, the E. coli standard for the binary outcome was targeted. As described previously, monitoring standards recommended a five-day geometric mean, and a general single-sample maximum was provided (US EPA, 1986). The guidance suggested, however, that single-sample maximums be calculated for individual jurisdictions based on local log standard deviation of E. coli concentration (US EPA, 1986). The calculation provided in the ambient criteria (US EPA, 1986) is ss ¼ 10^ log10 GM þ cl log10 sd
14/day. Overall, for the 14 beaches, a mean of 35 illnesses/day could be expected (range: 0e659 66 SD).
3.2.
Amount of E. coli variation explained by models (R2)
Coefficient of determination was lowest for the CM, with a range of 0.049 (South Shore) to 0.135 (63rd Street) for individual beaches and an overall adjusted R2 of 0.141 (Fig. 2a). The higher R2 for 63rd Street results from the persistent high E. coli concentrations at this location with little variation. Generally, more variation in E. coli concentrations could be explained using the RM, resulting in a range of R2 from 0.111 (63rd Street) to 0.287 (North Avenue). Further improvement in R2 was seen with the IM (Fig. 2a). The model for 63rd Street included the lowest amount of variation explained (0.141); the highest adjusted R2 was at North Avenue (0.313). The general pattern was an increasing R2 with increasing model refinement, seen at all beaches except 63rd Street. The change in R2 between the CM and the RM was greatest for Hollywood and South Shore beaches, with notable improvements at Foster, North Avenue and South Shore. Change in R2 between RM and IM was not as great overall; most improvement was at the south side beaches. Precision of the predictions increased with model refinement as seen in a decreasing RMSE.
3.3. Binary water quality standard outcome (number of prediction errors) Using the CM, the beach was correctly left open to swimming 68% of the time (range ¼ 49e78%) (Fig. 3). This increased to 78%
(3)
where ss ¼ the single-sample maximum; GM ¼ 126, geometric mean E. coli concentration for acceptable illness rate of 8 per 1000; cl ¼ 0.675, the 75% calculated one-sided confidence level for a designated heavily used beach area, and sd ¼ is the calculated standard deviation for a given jurisdiction, 0.718 for Chicago beaches (mean log E. coli ¼ 1.776) Chicago’s singlesample maximum would therefore be 385 CFU/100 ml.
3.
Results
Visits to Chicago’s lakefront beaches are often in excess of 27,000,000 annually, with fully half of the visits associated with two beaches: Oak Street and North Avenue (Chicago Park District, unpublished data).
3.1.
Baseline water quality
Chicago beaches exceed the 235 single-sample maximum 14e35% of the time, with an overall rate of 20%. Given the current water quality and visitation of Chicago’s beaches and the hypothetical absence of any monitoring or associated beach closures (IA model), the majority of expected illnesses were associated with the high visitation beaches (Fig. 1). The highest illness rate was 138/day for North Avenue, and the lowest was 7/day for Rainbow. Despite having the highest mean E. coli concentration, 63rd Street illness rate averaged
Fig. 1 e Comparison of mean E. coli concentration at each beach and mean number of swimmers that can be estimated to develop gastrointestinal illness each day. Gradation of circles indicates the expected individual illness rate. Beaches include Loyola (LY), Albion (AL), Hollywood (HW), Foster (FO), Montrose (MO), North Avenue (NA), Oak (OK), 12th Street (12), 31st Street (31), 57th Street (57), 63rd Street (63), South Shore (SS), Rainbow (RB), and Calumet (CA).
1663
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 5 9 e1 6 6 8
Fig. 2 e Comparison of (a) coefficient of variations, (b) percent type I errors, and (c) percent type II errors by modeling approach.
(range ¼ 53e85%) with the RM and was 77% (range ¼ 54e84%) with the IM. Rate of correct closures under the CM is 6% (range ¼ 3e14%). That decreased to 0.8% (0.3e9) with the RM, and with the IM, the rate was 1.4% (0.3e11%). Type I errors, in which a swimming advisory is posted although E. coli concentration is lower than the single-sample maximum, were made 11.5% (9e17%) of the time using the CM for monitoring (Fig. 2b). With the application of an RM, that
inactive
rate decreased to 2.2% (0.4e4) of the time, and with the IM application, the rate was 3.2% (1e11%). Type II errors were common under the CM, occurring 14% (12e21%) (Fig. 2c). That number increased with both predictive models, to 19% (16e27) with the RM and 18.5% (14e24) with the IM. Comparison of the percent of type II errors showed highly variable results. The greatest percent change between models was an increase in percent type II errors at 12th Street and
current
outcome correct closed correct open type I error type II error
region
individual
Fig. 3 e Management outcome using the binary approach, i.e., an E. coli concentration >235 results in a beach closure.
1664
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 5 9 e1 6 6 8
Loyola between CM and RM. Minimizing type II errors between the RM and IM was limited.
3.4. Association of monitoring and modeling approaches with illness burden A comparison of the four models reveals, most noticeably at the high visitation beaches North Avenue, Montrose and Oak Street, that the leveling off of number of type II errors was mirrored in the rate of excess illnesses prevented with increasing model refinement. There was no significant difference in rate of excess illnesses associated with model approach at any of the individual beaches. In fact, model refinement between RM and IM showed little improvement in the number of excess illnesses prevented. At several beaches, cumulative illnesses were identical for all models except for CM. Eight of the beaches showed a difference in cumulative excess illnesses associated with the four models (Table 2). The CM was associated with lower cumulative excess illnesses, largely because there were fewer instances of type II errors; however, for all beaches but 63rd Street, the CM was associated with higher mean excess illnesses, indicating that with this model, the instances of very high E. coli concentrations tend to be missed. The IA model excluded type I errors because the beach would never be closed to swimming and therefore swimmers would not be prevented from water contact under any water quality conditions. Increasingly refined models were effective at reducing type I errors because they rarely predicted events when water quality exceeded the single-sample standard; most predictions were low E. coli concentrations. This tendency is apparent in the rapid decrease in number of type I errors with increasing model refinement.
3.5.
Chicago-specific monitoring standard
With the more liberal, jurisdiction-specific single-sample maximum (385 CFU/100 ml), the overall percentage of errors for all model types decreased (Fig. 4). Reduction in percent type II errors was more pronounced with the RM and IM (mean decrease by 35e36%) than for the CM (mean decrease by 22%). Dramatic reductions in percent of type II errors (>45%) occurred at some of the beaches with lower mean E. coli concentrations (North Avenue, Oak, Albion) with the RM and IM. The difference in number of illnesses should theoretically
remain unchanged because the illness rate is within the confidence limits of the original derived regression. Because the predictive models tend to predict low overalldthe tendency leading to type II errorsdthe reduction in type I errors was not as remarkable.
4.
Discussion
With recent findings that the time-intensive current beach monitoring models are not generally predictive of real-time FIB concentration (Boehm, 2007; Whitman et al., 1999), beach managers need a means of determining water quality rapidly and efficiently in order to protect public health while maximizing beach access. A suite of approaches for expanding monitoring activities and improving timeliness of monitoring results have been proposed and considered, but minimal research has explored their capacity to meet this goal. If a new method does not provide a significant improvement over current management outcomes, it is unlikely that managers will invest the necessary effort to alter the monitoring program. The first step for many jurisdictions is to implement a water quality monitoring program using the currently used model (CM). According to our results, this activity alone decreased the number of excess illnesses at all beaches, largely as a result of keeping swimmers out of the water more often, regardless of actual water quality. Beaches were closed when water quality was within acceptable standards 9e17% of the time, a scenario that can incur a social and economic burden (Rabinovici et al., 2004). With this model, accurate closures and health protection are strictly dependent on the endurance of an E. coli contamination event. With the rapidly changing nature of water quality (Boehm, 2007; Boehm et al., 2002; Whitman and Nevers, 2004), there is high potential to make a management error (Table 1). Multiple day contamination events would result in some correct swimming advisories (correct closed), and extended periods of low E. coli concentrations would result in accurate swimming permission (correct open). The vast majority of contamination events, however, last one day or less (Leecaster and Weisberg, 2001). In Chicago, only 6% of contamination events persisted for more than one sampling day, with the exception of 63rd Street: 17%. Monitoring at beaches with persistent contamination will inevitably decrease the number of type II errors and associated illnesses because swimmers are
Table 2 e Comparison of the cumulative (mean) excess illnesses as a result of the choice of model approach: inactive, current, regional, or individual beach model. Results based on days of type II errors, ranging from 38 to 103 days of the 247e292 days (2000e2004 seasons) considered in the analysis. Inactive Montrose North Ave 31st 57th 63rd South Shore Rainbow Calumet
2492.94 4540.82 366.22 912.68 704.59 551.00 240.73 1049.60
(42.25) (105.60) (5.39) (16.90) (6.84) (10.20) (3.70) (20.58)
Current 1695.49 (42.39) 3659.66 (107.64) 250.73 (5.57) 758.64 (17.64) 389.27 (6.38) 456.19 (11.13) 176.09 (3.91) 869.01 (22.87)
Regional 2407.17 4540.82 348.00 912.51 548.65 551.00 240.73 1049.32
(41.50) (105.60) (5.52) (17.55) (6.94) (10.20) (3.70) (20.99)
Individual 2442.56 4270.00 320.70 894.31 490.04 537.40 217.60 1023.04
(42.85) (101.69) (5.34) (17.54) (6.71) (10.14) (3.69) (20.88)
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 5 9 e1 6 6 8
1665
Fig. 4 e Impact on beach advisory decisions if jurisdiction-specific water quality standard is used. The two beach examples include 63rd Street Beach, which has a high mean E. coli, lower visitation rate, and frequent swimming advisories, and North Avenue Beach, which is the most highly visited Chicago beach and has low mean E. coli and infrequent swimming advisories. Use of a Chicago-specific water quality standard would result in fewer errors overall using all three monitoring model approaches.
accurately kept from the water for some percent of the high concentration events. Likewise, beaches with infrequent contamination events will have a limited number of type I errors. To overcome these inconsistencies in actual vs. presumed water quality, predictive models have been used at some beaches to provide a real-time estimate of water quality and overcome these inherent errors. The cornerstone of predictive modeling for beach management has been the individual beach model (Nevers and Whitman, 2005), but the cost of developing such models has been one of the disincentives to widespread use; many of the models use on-site instrumentation that requires installation and continuous maintenance (Francy, 2009). For this reason, regional models have been explored, which examine the hydrometeorological factors similarly affecting groups of beaches and describe background fluctuations in FIB concentrations (Nevers and Whitman, 2008). While more crude in estimations, regional models may provide a cost-effective solution for jurisdictions with numerous monitored beaches while providing insights into source behavior. Perhaps surprising, the RM results in this exercise were quite similar to results from IM, indicating a generally predictable fluctuation in FIB concentrations across Chicago. Overall, increasing refinement of monitoring approach with the use of predictive models was associated with improved accuracy of E. coli predictions. Use of the RM increased the amount of variation explained over the current monitoring approach, and the use of beach-specific IM somewhat further improved this result. With beach-specific refinement, predictive models for all of the study beaches had higher R2 and lower RMSE: more variation in individual E. coli concentrations was
described and there was lower error in this estimation. In this study, geographically widespread predictors were used, which may have limited each model’s ability to detect beach-specific variation, although the low R2 for 63rd Street was somewhat expected due to inherent high variation in E. coli and the complex circulation pattern at this enclosed beach (Ge et al., 2010; Whitman and Nevers, 2004). Models developed for this beach have depended on higher frequency and higher intensity local data than were available for this exercise (Boehm et al., 2007; Olyphant and Whitman, 2004). Predictability improved significantly with the use of the RM for many of the beaches, with the biggest improvement between CM and RM at the north-side beaches, but it was the further refinement to IM that resulted in the greatest improvement at southern beaches. This pattern supports the idea that there is more beach-specific variation at these southern beaches and E. coli concentrations perhaps recover to background concentrations more slowly than at the northern beaches (Whitman and Nevers, 2008). Use of individual predictive models at these beaches can take these factors into account, resulting in better predictability. The number of type II errors for each model was highly variable, lacking the pattern of improvement with increasing refinement shown in the R2. All of the models failed to predict the majority of high E. coli concentrations. Because many beaches have infrequent high E. coli concentration events, it is difficult to detect a pattern of associated hydrometeorological conditions; this is a problem for many predictive modeling attempts but is an important characteristic for eliminating type II errors. The CM generally had the lowest number of type II errors, likely simply because this monitoring approach
1666
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 5 9 e1 6 6 8
closes the beach more often, an action that also results in significantly more type I errors (Fig. 3). Calculated excess illnesses were far less variable between models, indicating that more advanced predictive models do not necessarily provide improved public health protection. IM had a lower number of excess illnesses at some beaches (Table 2), but the difference was not nearly as great as might be expected. The results imply that IM detect some instances of extreme high concentrations at these beaches but generally miss these events. The majority of beaches showed a pattern of identical number of estimated cumulative illnesses for all model approaches with the exception of the CM: it was associated with the lowest cumulative excess illnesses specifically because this approach limits overall exposure to the beach water. The CM was therefore associated with lowest excess illnesses due to more closed beaches (both type I errors and correct closed).
4.1.
Application of monitoring standards
The number of excess illnesses associated with a given model is affected by the calculation of the illness rate, and although jurisdictions have the option to calculate single-sample maximums specific to their beach waters, most Great Lakes states use the default 235 CFU/100 ml standard developed using data from the original epidemiological studies for all beaches statewide (Nevers and Whitman, 2010). Use of a higher Chicagospecific standard increases swimming access without impacting illness rates by expanding the range of allowable water contact and therefore the number of accurate predictions (correct open). Percent of type II errors was reduced under the CM and drastically reduced using RM or IM. This was particularly noticeable at beaches with lower mean E. coli concentrations: upwards of 40% reduction in percent type II errors for North Avenue, Montrose, and Oak Street with RM and IM. This significant reduction is mirrored in the potential to significantly reduce the number of illnesses at these high visitation beaches; illness rates are greatly elevated during a single high FIB event at a high use beach (Turbow et al., 2003). The use of the Chicagospecific standard did not greatly reduce the number of type II errors for 63rd Street because of the higher overall mean E. coli concentration. Within the framework of the original monitoring standards, leeway is provided for level of beach use, beach-specific variation in bacteria concentrations, and calculation of the overall water quality (US EPA, 1986), and health protection is assumed to be equally provided over a range of calculated concentrations. Considering this broader range of confidence could increase beach use without influencing health outcome under a variety of monitoring approaches, including the predictive models presented here. These results indicate that the use of a higher standard, along with a predictive model could maximize access at many of the Chicago beaches without increasing public health risk. The monitoring standards recommend use of the 5-day geometric mean, but most managers opt for the single-sample maximum, likely due to ease of use. An examination of Chicago beach monitoring data reveals that use of the 5-day geometric mean results in fewer errors than either the 235 single-sample maximum or the 385 Chicago-specific standard presented here. However, the number of days exceeding the
specified limits (i.e., swimming advisories) increases significantly with use of the geometric mean. The 5-day geometric mean was developed based on studies at beaches influenced by point sources (US EPA, 1986), areas more likely to have persistent high E. coli concentration events; river discharge or sewage releases may create periods of sustained high E. coli concentrations, warranting the extended swimming advisory that results from a running geometric mean. Chicago’s beaches, however, are not influenced by a major point source except during rare events of sewage overflows, during which beaches are preemptively closed for several days.
4.2.
Maximizing public health protection
Predictive modeling results indicate that this monitoring approach would not improve health protection at all Chicago beaches. The best approach for monitoring may differ between beaches, even within an individual jurisdiction such as Chicago. The threshold for level of effort associated with increased model refinement would have to be considered for each beach, perhaps while incorporating economic considerations. Hou et al. (2006) determined that different monitoring policies provided the optimal economic and public health outcomes for each of two beaches. The application of different monitoring strategies may include combining approaches or extending to rapid methods, alternate indicators, or diverse management plans. Recent research to characterize the sources, survival, fate, and transport of FIB and the applications of monitoring programs has perhaps complicated the applicability of different monitoring and management strategies by indicating that one monitoring approach does not fit all beach types. Novel management approaches have included predictive models (Frick et al., 2008; Nevers and Whitman, 2005), rapid tests (Bushon et al., 2009; Lavender and Kinzelman, 2009), new indicators, including hostspecific markers (Bacteroides, Methanobrevibacter, virulence markers etc), and gene-based detection techniques for human pathogens (Griffith et al., 2009). Reconsiderations of monitoring standards have also been explored (Kim and Grant, 2004; Nevers and Whitman, 2010). While epidemiological studies have linked illness rates with outcomes from some of these new analyses (Wade et al., 2006), care will have to be taken to consider whether environmental conditions and sources influence the results outcomes. The ideal method for beach management may differ among beaches and jurisdictions, and it may be desirable to have a variety of monitoring approaches available to beach managers that increase accuracy and reliability at given beaches. In deciphering the best management plans for different beaches, efforts should focus on improving public health protection, perhaps considering a wide variety of available monitoring options.
5.
Conclusions
Refinement of monitoring models generally increased predictability of E. coli but did not necessarily result in fewer errors or excess illnesses Regional models provided similar levels of accuracy as individual beach models in many locations
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 5 9 e1 6 6 8
Use of a location-specific water quality standard, combined with empirical predictive models, provided the greatest beach access without sacrificing public health protection
Acknowledgments We thank Murulee Byappanahalli (USGS) for his careful review. Research was funded in part by the US Ocean Action Plan: USGS Ocean Research Priorities Plan and by the Great Lakes Restoration Initiative. Any use of trade, product, or firm names is for descriptive purposes only and does not imply endorsement by the U.S. Government. This article is Contribution 1627 of the USGS Great Lakes Science Center.
references
Boehm, A.B., 2007. Enterococci concentrations in diverse coastal environments exhibit extreme variability. Environmental Science & Technology 41 (24), 8227e8232. Boehm, A.B., Grant, S.B., Kim, J.H., Mowbray, S.L., McGee, C.D., Clark, C.D., Foley, D.M., Wellman, D.E., 2002. Decadal and shorter period variability of surf zone water quality at Huntington Beach, California. Environmental Science & Technology 36 (18), 3885e3892. Boehm, A.B., Whitman, R.L., Nevers, M.B., Hou, D., Weisberg, S.B., 2007. Statistical Framework for Recreational Water Quality Criteria and Monitoring. In: Wymer, L.J. (Ed.). Wiley, West Sussex, UK, pp. 179e210. Buckalew, D.W., Hartman, L.J., Grimsley, G.A., Martin, A.E., Register, K.M., 2006. A long-term study comparing membrane filtration with Colilert defined substrates in detecting fecal coliforms and Escherichia coli in natural waters. Journal of Environmental Management 80 (3), 191e197. Bushon, R.N., Likirdopulos, C.A., Brady, A.M.G., 2009. Comparison of immunomagnetic separation/adenosine triphosphate rapid method to traditional culture-based method for E. coli and enterococci enumeration in wastewater. Water Research 43 (19), 4940e4946. Calderon, R.L., Mood, E.W., Dufour, A.P., 1991. Health effects of swimmers and nonpoint sources of contaminated water. International Journal of Environmental Health Research 1, 21e31. Dufour, A.P., 1984. Health Effects Criteria for Fresh Recreational Waters EPA Report EPA-600/1-84-004, Cincinnati, Ohio. Dwight, R.H., Brinks, M.V., SharavanaKumar, G., Semenza, J.C., 2007. Beach attendance and bathing rates for Southern California beaches. Ocean & Coastal Management 50 (10), 847e858. Francy, D.S., 2009. Use of predictive models and rapid methods to nowcast bacteria levels at coastal beaches. Aquatic Ecosystem Health and Management 12 (2), 177e182. Frick, W.A., Ge, Z., Zepp, R.G., 2008. Nowcasting and forecasting concentrations of biological contaminants at beaches: a feasibility and case study. Environmental Science & Technology 42, 4218e4824. Ge, Z., Nevers, M.B., Schwab, D.J., Whitman, R.L., 2010. Coastal loading and transport of Escherichia coli at an embayed beach in Lake Michigan. Environmental Science & Technology 44, 6731e6737. Given, S., Pendleton, L.H., Boehm, A.B., 2006. Regional public health cost estimates of contaminated coastal waters: a case
1667
study of gastroenteritis at southern California beaches. Environmental Science & Technology 40 (16), 4851e4858. Griffith, J.F., Cao, Y., McGee, C.D., Weisberg, S.B., 2009. Evaluation of rapid methods and novel indicators for assessing microbiological beach water quality. Water Research 43 (19), 4900e4907. Gronewold, A.D., Borsuk, M.E., Wolpert, R.L., Reckhow, K.H., 2008. An assessment of fecal indicator bacteria-based water quality standards. Environmental Science & Technology 42 (13), 4676e4682. Harwood, V.J., Levine, A.D., Scott, T.M., Chivukula, V., Lukasik, J., Farrah, S.R., Rose, J.B., 2005. Validity of the indicator organism paradigm for pathogen reduction in reclaimed water and public health protection. Applied and Environmental Microbiology 71 (6), 3163e3170. He, L.M., He, Z.L., 2008. Water quality prediction of marine recreational beaches receiving watershed baseflow and stormwater runoff in southern California, USA. Water Research 42 (10e11), 2563e2573. Hou, D., Rabinovici, S.J.M., Boehm, A.B., 2006. Enterococci predictions from partial least squares regression models in conjunction with a single-sample standard improve the efficacy of beach management advisories. Environmental Science & Technology 40 (6), 1737e1743. Kim, J.H., Grant, S.B., 2004. Public mis-notification of coastal water quality: a probabilistic evaluation of posting errors at Huntington Beach, California. Environmental Science & Technology 38 (9), 2497e2504. Lavender, J.S., Kinzelman, J.L., 2009. A cross comparison of QPCR to agar-based or defined substrate test methods for the determination of Escherichia coli and enterococci in municipal water quality monitoring programs. Water Research 43 (19), 4967e4979. Leecaster, M.K., Weisberg, S.B., 2001. Effect of sampling frequency on shoreline microbiology assessments. Marine Pollution Bulletin 42 (11), 1150e1154. Lepesteur, M., McComb, A.J., Moore, S.A., 2006. Do we all face the same risk when bathing in the estuary? Water Research 40 (14), 2787e2795. Marion, J.W., Lee, J., Lemeshow, S., Buckley, T.J., 2010. Association of illness and recreational water exposure during advisory and non-advisory conditions at an inland US beach. Water Research 44 (16), 4796e4804. McPhail, C.D., Stidson, R.T., 2009. Bathing water signage and predictive water quality models in Scotland. Aquatic Ecosystem Health and Management 12 (2), 183e186. Nevers, M.B., Whitman, R.L., 2005. Nowcast modeling of Escherichia coli concentrations at multiple urban beaches of southern Lake Michigan. Water Research 39 (20), 5250e5260. Nevers, M.B., Whitman, R.L., 2008. Coastal strategies to predict Escherichia coli concentrations for beaches along a 35 km stretch of southern Lake Michigan. Environmental Science & Technology 42 (12), 4454e4460. Nevers, M.B., Whitman, R.L., 2010. Policies and practices of beach monitoring in the Great Lakes, USA: a critical review. Journal of Environmental Monitoring 12 (3), 581e590. Olyphant, G.A., Whitman, R.L., 2004. Elements of a predictive model for determining beach closures on a real time basis: the case of 63rd Street Beach Chicago. Environmental Monitoring and Assessment 98 (1e3), 175e190. Pru¨ss, A., 1998. Review of epidemiological studies on health effects from exposure to recreational water. International Journal of Epidemiology 27 (1), 1e9. Rabinovici, S.J.M., Bernknopf, R.L., Wein, A.M., Coursey, D.L., Whitman, R.L., 2004. Economic and health risk trade-offs of swim closures at a Lake Michigan Beach. Environmental Science & Technology 38 (10), 2737e2745.
1668
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 5 9 e1 6 6 8
Seyfried, P.L., Tobin, R.S., Brown, N.E., Ness, P.F., 1985. A prospective study of swimming-related illness I. Swimmingassociated health risk. American Journal of Public Health 75 (9), 1068e1070. Sinigalliano, C.D., Fleisher, J.M., Gidley, M.L., Solo-Gabriele, H.M., Shibata, T., Plano, L.R.W., Elmir, S.M., Wanless, D., Bartkowiak, J., Boiteau, R., Withum, K., Abdelzaher, A.M., He, G., Ortega, C., Zhu, X., Wright, M.E., Kish, J., Hollenbeck, J., Scott, T., Backer, L.C., Fleming, L.E., 2010. Traditional and molecular analyses for fecal indicator bacteria in nonpoint source subtropical recreational marine waters. Water Research 44 (10), 3763e3772. Turbow, D.J., Osgood, N.D., Jiang, S.C., 2003. Evaluation of recreational health risk in coastal waters based on enterococcus densities and bathing patterns. Environmental Health Perspectives 111 (4), 598e603. US EPA, 1986. Ambient Water Quality Criteria for Bacteria 1986, p. 18. US EPA, Office of Water Regulations and Standards, Washington DC. Wade, T.J., Calderon, R.L., Sams, E., Beach, M., Brenner, K.P., Williams, A.H., Dufour, A.P., 2006. Rapidly measured indicators of recreational water quality are predictive of swimming associated gastrointestinal illness. Environmental Health Perspectives 114 (1), 24e28. Wade, T.J., Pai, N., Eisenberg, J.N.S., Colford Jr., J.M., 2003. Do U.S. Environmental Protection Agency water quality guidelines for recreational waters prevent gastrointestinal illness? A
systematic review and meta-analysis. Environmental Health Perspectives 111 (8), 1102e1109. Whitman, R.L., Nevers, M.B., 2003. Foreshore sand as a source of Escherichia coli in nearshore water of a Lake Michigan beach. Applied and Environmental Microbiology 69 (9), 5555e5562. Whitman, R.L., Nevers, M.B., 2004. Escherichia coli sampling reliability at a frequently closed Chicago beach: monitoring and management implications. Environmental Science & Technology 38 (16), 4241e4246. Whitman, R.L., Nevers, M.B., 2008. Summer E. coli patterns and responses along 23 Chicago beaches. Environmental Science & Technology 42 (24), 9217e9224. Whitman, R.L., Nevers, M.B., Gerovac, P.J., 1999. Interaction of ambient conditions and fecal coliform bacteria in southern Lake Michigan waters: monitoring program implications. Natural Areas Journal 19, 166e171. Whitman, R.L., Shively, D.A., Pawlik, H., Nevers, M.B., Byappanahalli, M.N., 2003. Occurrence of Escherichia coli and enterococci in Cladophora (Chlorophyta) in nearshore water and beach sand of Lake Michigan. Applied and Environmental Microbiology 69 (8), 4714e4719. Wong, M., Kumar, L., Jenkins, T.M., Xagoraraki, I., Phanikumar, M.S., Rose, J.B., 2009. Evaluation of public health risks at recreational beaches in Lake Michigan via detection of enteric viruses and a human-specific bacteriological marker. Water Research 43, 1137e1149.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 6 9 e1 6 8 0
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Performance and biofilm activity of nitrifying biofilters removing trihalomethanes David G. Wahman a,*, Lynn E. Katz b, Gerald E. Speitel, Jr.b a
United States Environmental Protection Agency, Office of Research and Development, 26 W. Martin Luther King Dr., Cincinnati, OH 45268, USA b University of Texas at Austin, Department of Civil, Architectural and Environmental Engineering, Austin, TX 78712, USA
article info
abstract
Article history:
Nitrifying biofilters seeded with three different mixed-culture sources removed trichloro-
Received 13 July 2010
methane (TCM) and dibromochloromethane (DBCM) with removals reaching 18% for TCM
Received in revised form
and 75% for DBCM. In addition, resuspended biofilm removed TCM, bromodichloro-
7 December 2010
methane (BDCM), DBCM, and tribromomethane (TBM) in backwash batch kinetic tests,
Accepted 8 December 2010
demonstrating that the biofilters contained organisms capable of biotransforming the four
Available online 16 December 2010
regulated trihalomethanes (THMs) commonly found in treated drinking water. Upon the initial and subsequent increased TCM addition, total ammonia nitrogen (TOTNH3) removal
Keywords:
decreased and then reestablished, indicating an adjustment by the biofilm bacteria. In
Trihalomethanes
addition, changes in DBCM removal indicated a change in activity related to DBCM. The
Cometabolism
backwash batch kinetic tests provided a useful tool to evaluate the biofilm’s bacteria. Based
Nitrification
on these experiments, the biofilters contained bacteria with similar THM removal kinetics
Disinfection by-products
to those seen in previous batch kinetic experiments. Overall, performance or selection does
Drinking water
not seem based specifically on nutrients, source water, or source cultures and most likely results from THM product toxicity, and the use of GAC media appeared to offer benefits over anthracite for biofilter stability and long-term performance, although the reasons for this advantage are not apparent based on research to date. Published by Elsevier Ltd.
1.
Introduction
During drinking water disinfection, natural organic matter (NOM) combines with the disinfectant to produce disinfection by-products (DBPs), including haloacetic acids (HAAs) and trihalomethanes (THMs). Chlorine disinfection remains quite popular in the United States (AWWA Water Quality and Technology Division Disinfection Systems Committee, 2000a,b, 2008a,b), but as a result of the Stage 1 and Stage 2 Disinfectants and Disinfection Byproduct Rules, many utilities now use combinations of chlorine and chloramines to avoid excessive THM and HAA formation. A recent survey reported
* Corresponding author. Tel.: þ1 513 569 7733; fax: þ1 513 487 2543. E-mail address:
[email protected] (D.G. Wahman). 0043-1354/$ e see front matter Published by Elsevier Ltd. doi:10.1016/j.watres.2010.12.012
that 30% of the respondents currently chloraminate to maintain distribution system residual, and other recent surveys suggest that between 8 and 12% of drinking water utilities are contemplating a future switch to chloramination (AWWA Water Quality and Technology Division Disinfection Systems Committee, 2008b; Seidel et al., 2005) with chloramination for secondary disinfection in the United States predicted to increase to 57% of all surface and 7% of all ground water treatment systems (USEPA, 2005). A typical chloramine treatment scheme consists of an initial chlorination period to help achieve disinfection goals followed by quenching with ammonia at some point in the
1670
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 6 9 e1 6 8 0
treatment train to meet DBP goals through the lower regulated DBP formation rates associated with chloramines. Substantial formation of THMs and HAAs can occur within treatment plants even during relatively short chlorination periods (Singer et al., 1999b; Symons et al., 1982). Therefore, approaches for minimizing DBP formation or for removing DBPs within treatment plants are potentially of much practical value. Once formed, one possible removal mechanism for these DBPs is biological degradation. Evidence for HAA biodegradation in drinking water environments continues to mount (Baribeau et al., 2000; Bayless and Andrews, 2008; Leach et al., 2009; McRae et al., 2004; Singer et al., 1999a; Williams et al., 1997, 1998; Xie and Zhou, 2000, 2002; Zhang et al., 2009). Of course, THMs and HAAs tend to form together, so biological DBP removal processes must be able to deal with both classes of DBPs to be of any practical value in regulatory compliance. Unfortunately, THM biodegradation is more difficult than HAA biodegradation. In contrast to HAAs, which can serve as carbon and energy sources for microbial growth, THMs require cometabolic pathways that serve neither purpose. Thus, a primary substrate is needed to sustain the bacteria. Of particular interest in drinking water treatment is the observation that ammonia-oxidizing bacteria are capable of aerobically transforming the four regulated THMs (trichloromethane (TCM) or chloroform, bromodichloromethane (BDCM), dibromochloromethane (DBCM), tribromomethane (TBM) or bromoform) commonly found in treated drinking water (Wahman et al., 2006a, 2005). Implementation of biological THM removal should involve relatively minor retrofitting of existing plants. The cometabolism would occur in granular media filters consisting of an upper layer of granular activated carbon (GAC). Utilities could carry out prechlorination followed by ammonia addition at a relatively low concentration (1e4 mg N/L) sometime before the filters. A mixture of monochloramine and ammonia will result at the typical free chlorine concentrations (e.g., 2 mg Cl2/L) used in treatment plants. When the water is applied to the upper GAC layer in the filter, monochloramine will be destroyed through a catalytic reaction with the GAC, releasing ammonia (Fairey et al., 2006, 2007; Komorita and Snoeyink, 1985). At this point, an appropriate environment for microbial growth is established (i.e., an environment devoid of a disinfectant residual). With respect to THMs, nitrifiers, specifically ammonia oxidizers, can grow using the available ammonia and cometabolize THMs. The filtered water would then be post-disinfected, presumably with chloramines, before distribution. Previous biofilter experiments provided an initial demonstration of feasibility in lab-scale biofilter experiments seeded with a single nitrifying mixed culture from Lake Austin, Texas (Wahman et al., 2006b). The current research extended process evaluation to two additional mixed-culture sources studied previously in batch experiments (Wahman et al., 2006a). In addition, further exploration of the reduced performance over time seen during biofilter operation with Lake Austin feed water (Wahman et al., 2006b) was investigated. Biofilter media was expanded to include GAC. The GAC source was from an operating drinking water treatment plant filter that was preceded by monochloramine addition (i.e., the proposed process configuration), providing insight into
cultures likely seen in practice. Furthermore, batch kinetic tests were conducted on backwash from the biofilters to provide a direct evaluation of the biofilm’s ability to remove THMs.
2.
Materials and methods
2.1.
Water collection and storage
Lake Austin water was obtained from the raw water line of the drinking water treatment facilities in Austin, Texas prior to any treatment. Water was subsequently stored in a 4 C temperature controlled room in LDPE and HDPE storage tanks until use. Lake Austin water is a typical central United States surface water with an alkaline pH (8.26e8.43), moderate alkalinity (169e190 mg CaCO3/L), and dissolved organic carbon (3.4e4.6 mg C/L) (Roalson et al., 2003).
2.2.
Nitrifier mixed cultures
Three different mixed-culture sources were used for biofilter inoculation: (1) a sample collected from the influent line of a drinking water treatment facility in Laredo, Texas (Rio Grande), (2) an enriched nitrifier culture from several distribution systems in California and Wisconsin that was dominated by Nitrosomonas oligotropha representatives (provided by Dr. D. Noguera, University of Wisconsin) (N. oligotropha enrichment), and (3) in-use Filtrasorb 400 (F400) was obtained from the granular media filters at the City of Laredo drinking water treatment plant (Laredo).
2.3.
Biofilter media
Virgin anthracite of the appropriate mesh size (30 40) was obtained by grinding in a blender and sieving with the appropriately sized sieves. To remove the fines from the ground anthracite, it was first washed on the sieves with distilled-deionized (DDI) water. Further washing was achieved by placing the ground anthracite in a glass beaker with an approximate volume ratio of 9 parts Millipore water and 1 part anthracite where the mixture was stirred and allowed to settle before decanting the water. This process was repeated approximately 30 times or until the decanted water was clear. The in-use Filtrasorb 400 (F400) GAC media was hand ground with a mortar and pestle and sieved to obtain a 30 40 mesh size media. The fines were removed as per anthracite media.
2.4.
Biofilter setup
Two feed waters (Table 1) were used in the biofilter experiments: (1) 0.2 mm-filtered Lake Austin water supplemented with defined micronutrients (i.e., iron, copper, and phosphorus) or (2) 0.2 mm-filtered DDI water feed supplemented with nutrients (i.e., calcium, magnesium, copper, and iron) based on batch Nitrosomonas europaea growth (Wahman et al., 2005) and a carbonate/phosphate buffer to simulate natural waters (approximately 200 mg CaCO3/L). THMs were added via a syringe pump, and oxygen was added to the feed water when required, raising the biofilter influent dissolved oxygen (DO) to non-limiting levels (approximately 16e20 mg/L).
1671
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 6 9 e1 6 8 0
Table 1 e Summary of nominal biofilter operating conditions. Run
1
2 3
Period
I II III IV I II I II III IV
Source water
EBCT (min.)
SLR (gpm/ft2)
LA LA LA LA LA NW NW NW NW NW
4 4 4 4 2 2 2 2 2 2
0.63 0.63 0.63 0.63 1.3 1.3 1.3 1.3 1.3 1.3
Nominal chemical additions to influent TOTNH3 (mg N/L)
TCM (mg/L)
DBCM (mg/L)
Fe (mg/L)
Cu (mg/L)
P (mg P/L)
4 4 4 4 2 2 2 2 4 2
0 75 110 110 110 110 100 100 100 100
0 0 0 0 0 0 0 25 25 25
200 200 200 200 200 5.6 5.6 5.6 5.6 5.6
15 15 15 15 15 0.41 0.41 0.41 0.41 0.41
0 0 0 1.1 1.1 1.1 1.1 1.1 1.1 1.1
EBCT ¼ Empty-bed contact time. LA ¼ Lake Austin water. NW ¼ Synthetic nutrient water. SLR ¼ Surface loading rate.
Experiments were conducted with three parallel trains (A, B, and C) with each train consisting of two biofilters in series. Trains A and B were packed wet and seeded with a continuous growth mixed-culture inoculum (See Section 2.2): Rio Grande (Train A) and N. oligotropha (Train B) (Wahman et al., 2006a). Before initiating flow, the biofilters were allowed to sit for approximately 1 h to promote nitrifier attachment to the media. Train C (Laredo) was packed wet but did not receive additional seeding beyond what was present on the media from being in operation at the water treatment plant and was placed into operation immediately upon packing. All trains received the same nominal influent total ammonia nitrogen (TOTNH3) concentration. TOTNH3 represents the sum of ammoniaenitrogen (NH3eN) and ammoniumenitrogen (NHþ 4 eN). Sampling points were at the first biofilter’s influent (sample point 0), between the two biofilters in a train corresponding to the first biofilter’s effluent and second biofilter’s influent (sample point 1), and at the second biofilter’s effluent (sample point 2). Using the biofilm scaling procedure proposed by Manem and Rittmann (1990), the experimental biofilters simulated full-scale filters operating with 4e8 min empty-bed contact times (EBCTs) and 2.5e7.0 gpm/ft2 (147e205 m/d) surface loading rates (SLRs), depending on the actual operating conditions and whether biofilm shear loss or external mass transport is chosen for scaling. These operating conditions fall into typical values reported for rapid filtration (MWH et al., 2005). A summary of additional biofilter operating conditions is provided in Table 1.
2.5.
Product toxicity
Previously, the cometabolism stability index (Csi) (Equation (1)) was derived to quantify the expected product toxicity of THMs fed during biofilter experiments (Wahman et al., 2006b): ! KsNH3 N þ a1 STOTNH3 Yk k TOTNH3 d a1 STOTNH3 r0g (1) Csi ¼ ¼ P k1THM STHM ri TcTHM
Where rg0 is the net rate of bacterial cell growth (1/day); ri is the rate of THM bacterial inactivation (1/day); Y is the bacterial cell yield (mg total suspended solids (TSS)/mg TOTNH3); kTOTNH3 is the TOTNH3 maximum substrate utilization rate constant (mg TOTNH3/mg TSS-day); kd is the endogenous decay coefficient (1/day); KsNH3 N is the ammoniaenitrogen half-saturation coefficient (mg N/L NH3eN); a1 is the ratio of NH3eN/TOTNH3; STOTNH3 is the TOTNH3 concentration (mg N/L TOTNH3); k1THM is the THM pseudo-first-order rate constant (L/mg TSS-day); STHM is the THM concentration (mg/L THM); and TcTHM is the THM transformation capacity (mg THM/mg TSS). For bacteria to provide sustained biotransformation of THMs, the net growth rate on ammonia (based on Monod kinetics) must be greater than the inactivation rate from THM biotransformation. Equation (1) indicates that for stable biofilter operation Csi must be greater than or equal to one (i.e., the net growth rate, rg0 , must be greater than the sum of the THM inactivation rates, ri). As was done for previous biofilter experiments (Wahman et al., 2006b), Csi values for all operating periods were initially calculated using kinetic parameters determined previously for N. europaea.
2.6.
Backwash batch kinetic experiments
During certain biofilter operating periods, biofilter backwash water was collected and organisms were subsequently harvested by centrifugation, washed, centrifuged again, and resuspended in fresh buffer medium (8 mM phosphate and 10 mM carbonate, pH 8) for batch kinetic studies. The fresh buffer media was aerated with pure oxygen to increase the dissolved oxygen concentration to levels (greater than 20 mg/L) that would not be fully consumed by ammonia removal and would not adversely affect the organisms during the experiment (Wahman et al., 2005). The approach of Aziz et al. (1999) was used to conduct the experiment. Briefly, batch kinetic assays were carried out head-space-free in 500 mL, glass, gas-tight syringes. Each syringe was wrapped in aluminum foil and contained a Teflon-coated stir bar for mixing. Chemicals were injected through the syringe nose to
1672
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 6 9 e1 6 8 0
start an experiment. Samples were collected over time by depressing the syringe plunger and ejecting the samples into a smaller gas-tight syringe. Thus, head-space-free conditions were maintained throughout the experiment (150e300 min), thereby virtually eliminating volatilization loss of chemicals. Starting individual THM concentrations ranged from 80 to 100 mg/L each and TOTNH3 concentrations from 4 to 8 mg N/L. From these experiments, ammonia and THM kinetic parameters were determined.
2.7.
Determination of batch kinetic parameters
Kinetic parameters were determined as described previously (Wahman et al., 2006a, 2005). For ammonia kinetics, Monod kinetic coefficients were estimated by nonlinear regression using Excel’s Solver routine (Smith et al., 1997, 1998). The model (Equation (2)) accounts for TOTNH3 being measured experimentally, while recognizing that NH3 is the actual substrate (Suzuki et al., 1974): dSTOTNH3 kTOTNH3 XSTOTNH3 a1 ¼ dt KsNH3 N þ STOTNH3 a1
(2)
where X is the biomass concentration (mg/L TSS) and other parameters were described previously. A fourth-order RungeeKutta numerical approximation of the Monod equation was fitted to data by minimizing the normalized residual sum of squares between predicted and experimental values. Normalization was achieved by dividing the residual sum of squares by the experimental value squared, resulting in a dimensionless sum of squares error (Robinson, 1985). For THMs, a reductant model (Equation (3)) was used that accounted for two limiting reactants, THMs and ammoniaenitrogen (NH3eN). The reductant model (Arcangeli and Arvin, 1997) was selected because it was superior to several other models in kinetic studies with N. europaea and several mixed-culture sources (Wahman et al., 2006a, 2005): dSTHM k1THM STHM ¼ KsNH3 N dt 1þ STOTNH3 a1
(3)
Where all parameters were described previously. The ammonia kinetic parameters were incorporated into the THM model without adjustment, and the THM parameters were determined using the same fitting method as used for the ammonia kinetics. As a result, the only adjustable parameters for the THM kinetic model were the THM rate constant and each initial THM concentration. The nonlinear regression analysis yielded estimates of each THM rate constant ðk1THM Þ, the ammonia maximum specific rate of biotransformation ðkTOTNH3 Þ, and the ammonia half-saturation constant ðKsNH3 N Þ, as well as the initial concentrations (S0) for ammonia and each THM. Further statistical analyses permitted estimates of the approximate 95% joint confidence limit (CL) for each parameter (Smith et al., 1997, 1998; Wahman et al., 2005).
2.8.
Simplified THM cometabolism model
A simplified THM cometabolism biofilter model can be obtained by ignoring mass transport resistances and assuming an
ideal plug flow reactor with a residence time equal to the contact time in the biofilter. Using the rate equations for TOTNH3 and THM detailed previously, a closed form solution for the removal of a given THM can be obtained and is shown below as Equation (4): k
1 DTOTNH3 k THM STHMn TOTNH3 ¼e STHM0
(4)
Where STHM0 is the THM influent concentration (mg/L THM); STHMn is the THM effluent concentration from nth (1 or 2) biofilter in series (mg/L THM); DTOTNH3 is the influent TOTNH3 minus effluent TOTNH3; and other parameters were described previously. Based on Equation (4), the THM normalized effluent S n Þ, and therefore THM fractional removal, concentration ðSTHM THM0 will be (1) independent of the influent THM and TOTNH3 concentrations and (2) for a given DTOTNH3 removal in a biofilter, dependent on the THM rate constant ðk1THM Þ. Thus, Equation (4) can be used to approximate the removal of each THM species as a function of ammonia removal.
2.9.
Biofilter operational data statistical analyses
Tukey’s paired comparison method was used to compare performance of different operating periods of the biofilters (Berthouex and Brown, 2002). A two-sided 95% confidence interval of the Studentized Range Statistic was used for all paired comparisons (Harter, 1960).
2.10.
Analytical methods
THM concentrations were measured using USEPA Method 551.1 with modifications. Concentrations of individual THM species were analyzed on a Hewlett Packard 5890A gas chromatograph with liquid autosampler and J&W DB-5 column
Table 2 e Selected biofilter performance summary (mean ± standard deviation). Train Run Period No. of D0e1TOTNH3 TCM % DBCM % samples (mg N/L) Removal Removal A
B
C
1 2 3 3 1 2 3 3a 1 2 3 3
IV I II IV IV I II IV IV I II IV
4 6 5 2 4 6 7 2 4 6 8 1
4.2 2.1 2.4 2.7 4.0 2.0 2.2 2.0 3.9 2.0 2.2 3.6
0.13 0.064 0.054 0.069 0.066 0.053 0.10 0.036 0.069 0.10 0.088
18 11 4.9 14 17 8.3 9.5 8.3 2.7 4.5 1.1 8.7
5.2 5.5 3.8 10 6.5 8.5 6.7 1.1 1.1 6.0 2.2
60 1.3 17 3.1
14 7.6 14 0.52
46 5.1 75
D0e1TOTNH3 ¼ TOTNH3 removed through the first biofilter in series (mg N/L). a Complete TOTNH3 removal not occurring.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 6 9 e1 6 8 0
1673
Fig. 1 e Run 1 biofilter (A [ Rio Grande, B [ N. oligotropha enrichment, and C [ Laredo) TOTNH3 concentrations (1) and percent removals (2) for first biofilter in series. Biofilters fed Lake Austin water with nominal influent TCM additions of 0 mg/L (Period I), 75 mg/L (Period II), or 110 mg/L (Periods III and IV).
with constant pressure and splitless injection. The initial oven temperature was 32 C for 3.5 min. Then, the temperature was ramped at 20 C/min to 72 C and remained at 72 C for 3.5 min for a total analysis time of 9 min. A standard curve was developed to span the range of anticipated THM concentrations for each experiment. For quality control purposes, a blank and duplicate sample was placed every 20 vials and a standards calibration check was run every 10th vial. An ion selective electrode probe, Thermo Orion 9512, connected to an Orion Model 920A pH/ISE electrode meter was used to measure ammonia. An ammonia standard curve was developed to span the anticipated range of ammonia concentrations the experiments. All ammonia concentrations reported are in mg N/L. DO was measured with a YSI 5905 oxygen probe on a YSI Model 54ARC oxygen meter calibrated per the manufacturer’s recommendations. pH was measured using an Orion ROSSä combination pH electrode on an Orion Model 920A pH/ISE meter calibrated with pH standards of 4, 7, and 10. Total suspended solids (TSS) and volatile suspended solids (VSS) were measured to determine the biomass for batch kinetic experiments using Standard Methods 2540D and 2540E, respectively (APHA et al., 1998). The solids were measured with the buffer solution/biomass mixture remaining after batch kinetic experiments were completed, and the
volume of solution usually ranged from 50 to 100 mL. This volume was vacuum filtered through a Whatman cellulose nitrate 0.2-mm filter. Because VSS was equal to TSS measurements (data not shown), only TSS is reported herein.
3.
Results and discussion
Three biofilter trains were seeded with different mixedculture sources: (1) Rio Grande on anthracite (Train A), (2) N. oligotropha enrichment on anthracite (Train B), and (3) Laredo GAC (Train C). This experimental design provided biofilter performance data for two mixed cultures (N. oligotropha enrichment and Rio Grande) studied previously in batch (Wahman et al., 2006a), and the inclusion of a biofilter with GAC from the City of Laredo drinking water treatment plant provided insight into cultures likely seen in practice (i.e., a plant using monochloramine for disinfection preceding a GAC filter). For ease of presentation, the results are separated into three runs. To compare the performance of Runs 1e3, average values were calculated for pseudo-steady-state operation based on TOTNH3 removal (i.e., complete TOTNH3 removal through the first column in series). Table 2 summarizes selected results for each train along with their associated standard deviations.
1674
3.1.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 6 9 e1 6 8 0
Run 1
Run 1 consisted of an initial operating period to establish nitrification (Period I), explored TCM product toxicity and recovery (Periods II and III), and attempted to address issues of low TCM removal by phosphorus addition (Period IV). Lake Austin water was supplemented with iron and copper (Table 1) as these additions were previously shown to improve biofilter performance on Lake Austin water (Wahman et al., 2006b). Fig. 1 details the TOTNH3 concentrations and associated percent removals for the first biofilter in series of all three trains in Run 1. During Run 1, an initial period (Period I) was provided for the biofilters to reach a pseudo-steady-state removal of TOTNH3 without the presence of THMs. Trains A and B stabilized with an approximate 70e80% TOTNH3 removal through the first biofilter in series. During this same timeframe, Train C showed complete TOTNH3 removal through the first biofilter in series, indicating the presence of a greater and/or more active biomass as compared with Trains A and B. To simulate a Csi value close to one (1.1), an initial addition of 75 mg/L TCM (Period II) was started. This addition resulted in an immediate large decrease in TOTNH3 removal for both Trains A (13%) and B (0%). After this sharp initial decrease in TOTNH3 removal and by the end of Period II, Train A’s TOTNH3 removal improved to a level similar to Period I (w80%), but Train B’s removal only slightly improved to approximately 20%. In contrast, Train C showed a minimal effluent TOTNH3 concentration (0.5 mg N/L) after the initial TCM addition, with subsequent samples showing complete TOTNH3 removal through the first column in series. Train C contained GAC media; therefore, adsorption of TCM may have protected a portion of the biomass from the toxic effects of TCM cometabolism. Because the trains showed improved TOTNH3 removal after their initial decrease during Period II, the influent
Fig. 2 e Train A (Rio Grande) and Train B (N. oligotropha enrichment) recovery from initial TCM addition during Run 1 (Periods II and III) for first biofilter in series.
TCM concentration was increased to 110 mg/L (Period III) to decrease the Csi value (0.76) and evaluate whether recovery would continue. After showing an initial decrease in TOTNH3 removal when the TCM concentration increased, all three trains moved toward complete TOTNH3 removal during Period III with only Train B having a minimal TOTNH3 effluent concentration (0.25 mg N/L) at the end of Period III. A similar effect on TOTNH3 removal was seen upon initial THM addition in previous biofilter studies, but in these studies, the influent THMs were removed to allow recovery (Wahman et al., 2006b). The current experiment shows that the biofilters do not require the removal of TCM to recover from the initial adverse effects of TCM addition. Even with eventual nearly complete TOTNH3 removal in all three trains, TCM removal was variable and less than in previous biofilter experiments with comparable TOTNH3 removals (Wahman et al., 2006b). TCM removal for Trains A and B ranged from 10 to 21% and 3 to 19%, respectively. Train C’s TCM removal declined from initial removals of 8e10%
Fig. 3 e Train A (Rio Grande) Run 2 biofilter TOTNH3 (A) and TCM (B) concentrations and percent removals (C) for first biofilter in series. Biofilters fed Lake Austin (Period I) or synthetic nutrient (Period II) water with nominal influent TCM addition of 110 mg/L (Periods IeII).
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 6 9 e1 6 8 0
to final removals of 0e5% at the end of Period III. Train C’s initial TCM removal was most likely a result of GAC TCM adsorption. Train A and B’s recovery rates from TCM addition (Periods II and III) were different. Compared with Train A, Train B took approximately 800 h longer to reach complete TOTNH3 removal. For Trains A and B, the recovery pattern was approximately linear with TOTNH3 mass removal through the first biofilter in series (Fig. 2). Train A’s TOTNH3 removal recovery rate (95% CL 0.012 0.0040 mg N/L-h) was significantly greater than Train B’s (95% CL of 0.0043 0.00046 mg N/L-h), indicating a different response to TCM addition. Based on the Csi concept, Train B likely possessed a biomass with either a lower yield, kTOTNH3 , or transformation capacity or a higher k1THM or ksNH3 N . For both Train A and B, the biofilm nitrifying community made an obvious adjustment to respond to the addition of TCM to the biofilter and was able to recover under a continuous TCM feed. Because of the decreased TCM removal seen during Periods II and III, phosphorus was added at 1.1 mg P/L to evaluate its effect on TCM removal (Period IV). Phosphorus was not previously studied and may be limiting in the Lake Austin
1675
water. Lake Austin water’s average total phosphate concentration ranged from 0.01 to 0.05 mg P/L during 2004 and 2005 (City of Austin Water Utility n.d.). During Period IV, TOTNH3 removal remained complete for each train. For Trains A and B, TCM removal was variable but similar to the periods without phosphorus addition, ranging from 10 to 25%. For Train C, the TCM removal remained minimal at 2e4%, indicating little effect from the phosphorus addition on any of the trains.
3.2.
Run 2
For Run 2 (Train A, Fig. 3), the influent TOTNH3 concentration was decreased from 4 to 2 mg N/L to evaluate whether enzyme competition was occurring between TOTNH3 and the THMs. In addition, the EBCT was decreased from 4 to 2 min, with the goal of achieving a measurable steady-state effluent TOTNH3 concentration from the first biofilter in series. In an attempt to improve TCM removal during Run 2, the feed water was changed from Lake Austin (Period I) to synthetic nutrient water (Period II) as better performance with respect to TCM removal was seen with a nutrient water feed (Wahman et al., 2006b).
Fig. 4 e Run 3 biofilter (A [ Rio Grande, B [ N. oligotropha enrichment, and C [ Laredo) TOTNH3 (1) and DBCM (2) concentrations and percent removals (3) for first biofilter in series. Biofilters fed synthetic nutrient water (Periods IeIV) with a nominal influent TCM addition of 100 mg/L (Periods IeIV) and DBCM addition of 25 mg/L (Periods IIeIV).
1676
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 6 9 e1 6 8 0
During Period I, TOTNH3 removal remained complete through the first biofilter in series for all three trains. Trains A (Fig. 3) and B (data not shown) showed similar TCM removals, but as compared with Run 1 (Period IV), these average TCM removals (Table 2) decreased from 18 to 11% and 17 to 8% for Trains A and B, respectively. For Train C, TCM removal remained minimal, averaging 5%. The switch to nutrient water (Period II) resulted in little change in TOTNH3 or TCM removal for any train. Overall, Run 2 provided evidence that the low TCM removals were not a result of source water characteristics or competition with TOTNH3.
3.3.
Run 3
Previous research has shown that compared to TCM, the bromine-substituted THMs are removed faster but result in greater product toxicity to the bacteria (Wahman et al., 2006a, 2005). To evaluate the removal of a bromine-substituted THM, DBCM was added to the influent at 25 mg/L, and the TCM influent was reduced to 100 mg/L. These THM feed concentrations decreased the Csi from Run 2 (0.75e0.48), allowing evaluation of whether stable biofilter operation would continue at this decreased Csi and whether the biofilm community would adjust to a bromine-substituted THM as with TCM during Run 1. Fig. 4 details the TOTNH3 and DBCM concentrations and associated percent removals for each train during Run 3. After an initial period (Period I) of TCM and TOTNH3 addition, DBCM was added to the influent (Period II). Even with the decreased Csi, DBCM addition did not lead to a decreased TOTNH3 removal in any train as complete TOTNH3 removal occurred after the first biofilter in series (Fig. 4). During Period II, Train A showed a substantial 44e61% DBCM removal and a low 0e11% TCM removal. Compared with Train A, Train B’s DBCM removal was lower (1e22%) and TCM removal was similar (0e18%). Train C showed no TCM removal upon the DBCM addition with TCM effluent concentrations increasing through the first biofilter in series, indicating that competitive adsorption was occurring between TCM and DBCM. During this time, Train C’s DBCM removal was 39e56%. To evaluate if a higher TOTNH3 concentration would stimulate increased THM removal, the influent TOTNH3 was increased from 2 to 4 mg N/L TOTNH3 (Period III). Train A’s TOTNH3 removal remained complete with Train C moving toward complete removal. Train B approached process failure as TOTNH3 removal decreased during this period, which is predicted from the Csi (0.49). This provided further evidence that Train B’s slower recovery during Run 1 (Periods II and III) resulted from differences in the bacterial kinetics leading to a lower Csi than Train A.
Table 3 e Biofilter average k1THM/kTOTNH3 ratio summary. k1THM kTOTNH3
Train Fig. 5 e Train C (Laredo) PSDM simulations showing (A) TCM and DBCM breakthrough curves assuming virgin GAC and TCM (B) and DBCM (C) adjusted breakthrough curves and experimental data assuming GAC prior loading.
A B C
(L/mg TOTNH3)
TCM
DBCM initial
DBCM final
0.045 0.016 0.045 0.0019 0.017 0.0085
0.38 0.067 0.28
0.071 0.074 0.38
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 6 9 e1 6 8 0
Because of Train B’s performance, the influent was reduced to 2 mg N/L TOTNH3 for the remainder of Run 3 (Period IV) for all biofilters. In addition, the trains were operated to generate biomass during Period IV for backwash batch kinetic tests to determine the kinetic parameters of the biofilter biomass. During this time, samples were taken only before backwashing the biofilter for the backwash batch kinetic tests. All trains showed similar TCM removals (approximately 10%) with various DBCM removals (Table 2). During Period IV, Train A’s DBCM removal declined to 17%, which was similar
Fig. 6 e Backwash batch kinetic test parameter estimation and 95% joint CL summary for (A) kTOTNH3 , (B) KsNH3 LN , and (C) k1THM (A1 and A1_3.5 at 4986 h, A2 at 5658 h, and B at 5708 h).
1677
to Train B’s 14% DBCM removal. In contrast, Train C’s DBCM removal improved to 75% at the end of its operation.
3.4.
Run comparisons
TCM and DBCM cometabolism was accomplished for Train A in a biofilter seeded with a mixed culture from the Rio Grande fed both Lake Austin and synthetic nutrient water. For this train, an increase in TOTNH3 removal did not lead to a significant increase in TCM or DBCM removal. This result is in contrast to previous biofilter experiments in which an increase in TOTNH3 removal coincided with an increase in THM removal (Wahman et al., 2006b). Excluding the period immediately after DBCM addition (Run 3, Period II), TCM removal remained relatively consistent over time (13% average), and no statistically significant difference existed in TCM removal for any operating period for Train A. Average DBCM removal significantly declined over time from an initial 60% removal to a final 17% removal, which was similar to the TCM removal for this train (Table 2). Train B demonstrated that TCM and DBCM cometabolism was accomplished in a biofilter seeded with an N. oligotropha enrichment culture fed Lake Austin and synthetic nutrient water. As with Train A, TCM removal was not significantly different for any of the operating periods and did not significantly decrease upon the addition of DBCM. In addition, DBCM removal started and remained at 14% which was similar to the TCM removal seen during this same period (9%) and not significantly different from Train A’s final DBCM removal. Because Train C was packed with GAC, adsorption may have occurred in the biofilter in addition to or in lieu of biological THM removal. To provide a baseline on THM adsorption, simulated breakthrough curves for TCM and DBCM were generated from the pore surface diffusion model (PSDM) implemented into AdDesignS (Hokanson et al., 1998). A simulation was run at the operating conditions seen in this research for the first biofilter in series of Train C with TCM and DBCM additions. Adsorption isotherm parameters present in the software library were used to generate the breakthrough curves shown in Fig. 5A. TCM was predicted to completely breakthrough the biofilter at 865 h (42% through Run 1). By contrast, DBCM was predicted to show no breakthrough during the operating time of the biofilters (5700 h). Because the GAC from the City of Laredo would be at some point of exhaustion at the time the samples were collected from the top of the full-scale filters, the predicted breakthrough curves for TCM and DBCM might be shifted in time based on the extent of exhaustion. To address this, the TCM and DBCM breakthrough curves were shifted in time so that the first data point for TCM and DBCM removal matched that respective point on their breakthrough curve. These adjusted breakthrough curves are shown for TCM (Fig. 5B) and DBCM (Fig. 5C) overlaid with the experimental normalized effluent data for each THM. If GAC adsorption was only occurring in the filter, the normalized effluent concentrations would be expected to increase as shown by the predicted breakthrough curves. For the DBCM normalized effluent concentration (Fig. 5C), a short period of increasing effluent values was followed by a decline
1678
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 6 9 e1 6 8 0
Table 4 e Parameters for cometabolism stability index calculations. Variable
a1 Y KsNH3 N kTOTNH3 kd k1TCM k1DBCM TcTCM TcDBCM
Initial N. europaeaa
Definition
[NH3-N]/[TOTNH3] at pH 8.0 Bacterial cell yield Ammoniaenitrogen half-saturation coefficient TOTNH3 maximum substrate utilization rate constant Endogenous decay coefficient TCM pseudo-first-order rate constant DBCM pseudo-first-order rate constant TCM transformation capacity DBCM transformation capacity
0.049 0.33c 0.16 2.9 0.020 0.10 0.20 9.2 6.5
Revisedb Train A
Train B
0.049 0.33c 0.029 1.5 0.015d 0.28 0.28 66d 22d
0.049 0.33c 0.21 0.83 0.015d 0.064 0.078 66d 22d
Units
Dimensionless mg TSS/mg TOTNH3 mg N/L NH3-N mg TOTNH3/mg TSS-day 1/day L/mg TSS-day L/mg TSS-day mg TCM/mg TSS mg DBCM/mg TSS
DBCM e dibromochloromethane, NH3eN e ammoniaenitrogen, TCM e trichloromethane, TOTNH3 e total ammoniaenitrogen, TSS e total suspended solids. a (Wahman et al., 2006b) unless otherwise noted. b This work unless otherwise noted. c (Rittmann and McCarty, 2001) e value also assumes volatile suspended solids equals TSS in cells. d (Bayer, 2007).
that is in opposition to the proposed breakthrough curve, suggesting that the removal of DBCM was biological in nature and improving over time as TOTNH3 removal improved. Furthermore, the lack of DBCM removal in the second biofilter in series (data not shown) in which no TOTNH3 removal was occurring suggests that the removal of DBCM in the first biofilter in series was biological in nature. For TCM, the data were unclear, but because of the predicted short time to breakthrough for TCM on virgin GAC (approximately 900 h), the GAC should have reached exhaustion early in the biofilter run, suggesting removal after 900 h (Run 1, Period II) can be attributed to biotransformation and not adsorption. Ignoring the period in which adsorption may have occurred (i.e., the first 900 h), TCM removal did not exceed 9%. DBCM removal increased over time, reaching its maximum level (75%) at the end of the run, providing evidence of substantially better DBCM removal with GAC versus anthracite. Using the average performance data and Equation (4), k1THM =kTOTNH3 ratios were determined for all three trains (Table 3). Because of the significant changes in DBCM removal seen for Trains A and C during their operation, Table 3 provides two ratios for DBCM, corresponding to their initial and final removals. For this same reason, standard deviations are only provided for the TCM ratios. The kinetic rate constant ratios for Trains A and B were the same for TCM and approached each other for DBCM through time. The final DBCM ratio for Train C approached that of Train A’s initial value.
Table 5 e Cometabolism stability index (Csi) values for biofilter runs. Parameter basis (Table 4) Initial (N. europaea) Revised (Train A) Revised (Train B)
Calculated Csi for run (Period) 1 (II) 1 (IIIeIV) 2 (IeII) 3 (I) 3 (II, IV) 3 (III) 1.1 1.5 3.3
0.76 1.0 2.3
0.75 1.0 2.1
0.82 1.1 2.3
0.48 0.65 1.2
0.49 0.65 1.3
3.5.
Backwash batch kinetic tests
To evaluate the ability of the bacteria present in the biofilm to biotransform THMs, batch kinetic tests on the biofilter backwash from the first biofilter in series were performed (Run 3, Period IV). For Train C, no kinetic parameters could be determined because of interfering effects with residual GAC carried over from biofilter backwashing and the competitive adsorption seen with all four THMs present. Fig. 6(A, B, and C) details the ammonia and THM kinetic parameters determined from these experiments along with their 95% joint CLs. Two experiments were conducted on Train A at 4986 (A1, TSS ¼ 44 mg/L) and 5658 (A2, TSS ¼ 33 mg/ L) hours and one on Train B at 5708 h (B, TSS ¼ 65 mg/L). Because of the large THM mass removed during the experiments with Train A, the A1 batch kinetic test was analyzed in two ways to see if transformation capacity affected the results. A complete analysis was conducted of the data set (A1) as well as a subset of the data starting when the TOTNH3 concentration was 3.5 mg N/L (A1_3.5). No significant difference was seen between these two analyses, indicating that transformation capacity was not an issue. In addition, the results from the two different experiments on Train A (A1 and A2) showed similar kinetic parameters and were conducted approximately 700 h apart, indicating the stability of the biofilm present in the biofilter and reproducibility of the results. For the backwash kinetic experiment conducted, the ammonia kinetic parameters for Trains A and B appear different although this cannot be statistically justified, but the THM kinetic parameters were markedly different. The THM kinetic parameters for Train A were significantly larger than for Train B. For both Trains A and B, the only significant difference among THMs was for A1 and A2 where the TBM rate constant was significantly less than for the other three THMs. This result differs from previous batch kinetic tests where the TBM rate constant was significantly greater than those of the other three THMs (Wahman et al., 2006a, 2005). For Trains A and B, the similar THM kinetic parameters for TCM and DBCM coincide with the similar TCM and DBCM biofilter removals at the time of the backwash tests.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 6 9 e1 6 8 0
The calculated Csi values used initially to select biofilter operating conditions were based on previous research with the pure culture N. europaea (Wahman et al., 2005). To investigate the initial decrease and subsequent recovery of the biofilters upon initial THM addition and stable operation with Csi values less than one, the determined ammonia and THM kinetic parameters and recently determined transformation capacities for a mixed culture (Bayer, 2007) were used to recalculate Csi values. The initial and revised parameters used to calculate Csi are provided in Table 4 with the resulting calculated Csi values for each operating condition summarized in Table 5. Based on this reanalysis, the recovery of the biofilters from initial TCM exposure is predicted because the Csi values for Runs 1 and 2 were above 1.0 (Table 5) for Trains A and B. For Run 3, Train B’s Csi values remain slightly above one, but for Train A, only Phase 1 is above 1.0. These lower values correspond with the instability seen in removals during Run 3.
4.
Conclusions
Nitrifying biofilters seeded with three different mixed-culture sources removed TCM (up to 18%) and DBCM (up to 75%). In addition, biofilm material backwashed from the biofilters biotransformed TCM, BDCM, DBCM, TBM in backwash batch kinetic tests, demonstrating that the biofilters contained organisms capable of biotransforming the four regulated THMs commonly found in treated drinking water. Upon the initial and subsequently increased TCM addition to the biofilters, TOTNH3 removal decreased and then reestablished itself without the need to remove TCM from the influent, indicating an adjustment by the biofilm bacteria after presumably experiencing some initial TCM product toxicity. These results indicate that sustained removals (approximately 10e15% for 2 mg N/L TOTNH3 removal and depending on THM speciation) should be attainable in systems with persistent THM concentrations (i.e., drinking water treatment plants). In addition, temporal changes in DBCM removal (Trains A and C) indicated a change in activity related to DBCM. Interestingly, the results from Trains A and C trended in opposite directions with Train A decreasing in DBCM removal while Train C increased in DBCM removal, indicating a possible benefit of using GAC. Furthermore, Train C’s GAC media originated from a drinking water treatment plant filter, implying that the organisms present from a drinking water treatment plant using chloramination were capable of substantial DBCM removal (up to 75%) and modest TCM removal (up to 9%). These removals could benefit utilities requiring modest removals (e.g., 10e15%) to maintain compliance with existing and future THM regulations. For treatment plants requiring ammonia removals above approximately 2 mg N/L (depending on influent dissolved oxygen levels) to achieve THM removal targets, supplemental oxygen would be required. The backwash batch kinetic tests provided a useful tool to evaluate the biofilm bacteria and provided kinetic parameters to evaluate product toxicity (i.e., Csi). Based on revised Csi calculations and biofilter performance, future backwash kinetic experiments should include determination of relevant
1679
transformation capacities to evaluate fully the biofilter operation. For GAC filters, batch experiments should minimize GAC carryover when harvesting bacteria so that adsorption to residual GAC can be eliminated. Based on the batch kinetic experiments, the biofilters contained bacteria with similar THM kinetics and greater transformation capacities than previous batch grown cultures. Overall, biofilter performance was not based specifically on nutrients, source water, or source cultures, and changes in performance most likely resulted from THM product toxicity. Use of GAC as media appeared to offer benefits over anthracite for biofilter stability and long-term performance, although the reasons for this advantage are not apparent based on research to date. Further investigations should explore the increased removals associated with using GAC filter media and further characterization of biofilm transformation capacities from these same systems.
Acknowledgements This research was funded by the American Water Works Association Research Foundation (AwwaRF) and Texas Advanced Technology Research Program (ATP), which the authors thank for their financial, technical, and administrative assistance. The comments and views detailed herein may not necessarily reflect the views of AwwaRF, its officers, directors, affiliates, or agents. Any opinions expressed are those of the authors and do not necessarily reflect the views of the U.S. Environmental Protection Agency; therefore, no official endorsement should be inferred. Any mention of trade names or commercial products does not constitute endorsement or recommendation for use.
references
APHA, AWWA, WEF, 1998. Standard Methods for the Examination of Water and Wastewater. American Public Health Association: American Water Works Association: Water Environment Federation, Washington, D.C. Arcangeli, J.P., Arvin, E., 1997. Modeling of the cometabolic biodegradation of trichloroethylene by toluene-oxidizing bacteria in a biofilm system. Environmental Science & Technology 31 (11), 3044e3052. AWWA Water Quality and Technology Division Disinfection Systems Committee, 2000a. Committee report: disinfection at large and medium-size systems. Journal of American Water Works Association 92 (5), 32e43. AWWA Water Quality and Technology Division Disinfection Systems Committee, 2000b. Committee report: disinfection at small systems. Journal of American Water Works Association 92 (5), 24e31. AWWA Water Quality and Technology Division Disinfection Systems Committee, 2008a. Committee report: disinfection survey, part 1-recent changes, current practices, and water quality. Journal of American Water Works Association 100 (10), 76e90. AWWA Water Quality and Technology Division Disinfection Systems Committee, 2008b. Committee report: disinfection survey, part 2-alternatives, experiences, and future plans. Journal of American Water Works Association 100 (11), 110e124.
1680
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 6 9 e1 6 8 0
Aziz, C.E., Georgiou, G., Speitel, G.E., 1999. Cometabolism of chlorinated solvents and binary chlorinated solvent mixtures using M. trichosporium OB3b PP358. Biotechnology and Bioengineering 65 (1), 100e107. Baribeau, H., Krasner, S.W., Chinn, R., Singer, P.C., 2000. Impact of Biomass on the Stability of Haloacetic Acids and Trihalomethanes in a Simulated Distribution System. American Water Works Association. Bayer, B.M., 2007. Significance of trihalomethanes in preventing distribution system nitrification. Master’s thesis, The University of Texas at Austin, Austin. Bayless, W., Andrews, R.C., 2008. Biodegradation of six haloacetic acids in drinking water. Journal of Water and Health 6 (1), 15e22. Berthouex, P.M., Brown, L.C., 2002. Statistics for Environmental Engineers. Lewis Publishers, Boca Raton. City of Austin Water Utility (n.d.). http://www.ci.austin.tx.us/water/waterreports. htm (accessed 19.10.06). Fairey, J.L., Speitel, G.E., Katz, L.E., 2006. Impact of natural organic matter on monochloramine reduction by granular activated carbon: the role of porosity and electrostatic surface properties. Environmental Science & Technology 40 (13), 4268e4273. Fairey, J.L., Speitel, G.E., Katz, L.E., 2007. Monochloramine destruction by GAC: effect of activated carbon type and source water characteristics. Journal of American Water Works Association 99 (7), 110e120. Harter, H.L., 1960. Tables of range and studentized range. The Annals of Mathematical Statistics 31 (4), 1122e1147. Hokanson, D.R., Hand, D.W., Crittenden, J.C., Oman, E.J., 1998. Adsorption Design Software (AdDesignS). Center for Clean Industrial and Treatment Technologies (CenCITT) Michigan Technological University, Houghton, Michigan. Komorita, J.D., Snoeyink, V.L., 1985. Monochloramine removal from water by activated carbon. Journal of American Water Works Association 77 (1), 62e64. Leach, L.H., Zhang, P., LaPara, T.M., Hozalski, R.M., Camper, A.K., 2009. Detection and enumeration of haloacetic acid-degrading bacteria in drinking water distribution systems using dehalogenase genes. Journal of Applied Microbiology 107 (3), 978e988. Manem, J.A., Rittmann, B.E., 1990. Scaling procedure for biofilm processes. Water Science and Technology 22 (1/2), 329e346. McRae, B.M., LaPara, T.M., Hozalski, R.M., 2004. Biodegradation of haloacetic acids by bacterial enrichment cultures. Chemosphere 55 (6), 915e925. MWH, Crittenden, J.C., Trussell, R.R., Hand, D.W., Howe, K.J., Tchobanoglous, G., 2005. Water Treatment: Principles and Design. John Wiley & Sons, Inc., Hoboken, New Jersey. Rittmann, B.E., McCarty, P.L., 2001. Environmental Biotechnology: Principles and Applications. McGraw-Hill Companies, Inc., New York. Roalson, S.R., Kweon, J., Lawler, D.F., Speitel, G.E.J., 2003. Enhanced softening: effects of lime dose and chemical additions. Journal of American Water Works Association 95 (11), 97e109.
Robinson, J.A., 1985. Determining microbial kinetic parameters using nonlinear regression analysis: advantages and limitations in microbial ecology. Advances in Microbial Ecology 8, 61e114. Seidel, C.J., McGuire, M.J., Summers, R.S., Via, S., 2005. Have utilities switched to chloramines? Journal of American Water Works Association 97 (10), 87e97. Singer, P.C., Arora, H., Dundore, E., Brophy, K. , Weinberg, H.S. 1999a. Control of haloacetic acid concentrations by biofiltration: a case study, pp. 1771e1779. Singer, P.C., Harrington, G.W., Cowman, G.A., Smith, M.E., Schecter, D.S., Harrington, L.J., 1999b. Impacts of Ozonation on the Formation of Chlorination and Chloramination ByProducts. Awwa Research Foundation, Denver, CO. Smith, L.H., Kitandis, P.K., McCarty, P.L., 1997. Numerical modeling and uncertainties in rate coefficients for methane utilization and TCE cometabolism by a methane-oxidizing mixed culture. Biotechnology and Bioengineering 53 (3), 320e331. Smith, L.H., McCarty, P.L., Kitandis, P.K., 1998. Spreadsheet method for evaluation of biochemical reaction rate coefficients and their uncertainties by weighted nonlinear least-squares analysis of the integrated Monod equation. Applied and Environmental Microbiology 64 (6), 2044e2050. Suzuki, I., Dular, U., Kwok, S.C., 1974. Ammonia or ammonium ion as substrate for oxidation by Nitrosomonas europaea cells and extracts. Journal of Bacteriology 120 (1), 556e558. Symons, J.M., Awwa, Us Epa, 1982. Treatment Techniques for Controlling Trihalomethanes in Drinking Water. American Water Works Association, Denver, CO. USEPA, 2005. Economic Analysis for the Final Stage 2 Disinfectants and Disinfection Byproducts Rule. USEPA. Wahman, D.G., Henry, A.E., Katz, L.E., Speitel, G.E.J., 2006a. Cometabolism of trihalomethanes by mixed culture nitrifiers. Water Research 40, 3349e3358. Wahman, D.G., Katz, L.E., Speitel, G.E.J., 2005. Cometabolism of trihalomethanes by Nitrosomonas europaea. Applied and Environmental Microbiology 71 (12), 7980e7986. Wahman, D.G., Katz, L.E., Speitel, G.E.J., 2006b. Trihalomethane cometabolism by a mixed-culture nitrifying biofilter. Journal of American Water Works Association 98 (12), 2e20. Williams, S.L., Williams, R.L., Gordon, A.S., 1997. The impact of bacterial degradation of haloacetic acids (HAA) in the distribution system, pp. 461e465. Williams, S.L., Williams, R.L., Yuan, J., 1998. Bacterial degradation of haloacetic acids in the distribution system, pp. 632e637. Xie, Y., Zhou, H., 2000. Biologically active carbon for haloacetic acid control, pp. 1437e1443. Xie, Y.F.F., Zhou, H.J., 2002. Use of BAC for HAA removal: part 2, column study. Journal of American Water Works Association 94 (5), 126e134. Zhang, P., Lapara, T.M., Goslan, E.H., Xie, Y.F., Parsons, S.A., Hozalski, R.M., 2009. Biodegradation of haloacetic acids by bacterial isolates and enrichment cultures from drinking water systems. Environmental Science & Technology 43 (9), 3169e3175.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 8 1 e1 6 9 0
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Influence of tetracycline resistance on the transport of manure-derived Escherichia coli in saturated porous media Jacob J. Walczak a, Sonia L. Bardy b, Lucia Feriancikova a, Shangping Xu a,* a b
Department of Geosciences, 3209 N Maryland Ave, University of WisconsineMilwaukee, Milwaukee, WI 53211, USA Department of Biological Sciences, 3209 N Maryland Ave, University of WisconsineMilwaukee, Milwaukee, WI 53211, USA
article info
abstract
Article history:
In this research, tetracycline resistant (tetR) and tetracycline susceptible (tetS) Escherichia
Received 24 August 2010
coli isolates were retrieved from dairy manure and the influence of tetracycline resistance
Received in revised form
on the transport of E. coli in saturated porous media was investigated through laboratory
9 December 2010
column transport experiments. Experimental results showed that tetR E. coli strains had
Accepted 10 December 2010
higher mobility than the tetS strains in saturated porous media. Measurements of cell
Available online 21 December 2010
surface properties suggested that tetR E. coli strains exhibited lower zeta potentials than the tetS strains. Because the surface of clean quartz sands is negatively charged, the repulsive
Keywords:
electrostatic double layer (EDL) interaction between the tetR cells and the surface of sands
Manure
was stronger and thus facilitated the transport of the tetR cells. Although no difference was
Antibiotic resistant bacteria
observed in surface acidity, cell size, lipopolysaccharides (LPS) sugar content and cell-
Escherichia coli
bound protein levels between the tetR and tetS strains, they displayed distinct outer
Bacteria transport
membrane protein (OMP) profiles. It was likely that the difference in OMPs, some poten-
Groundwater contamination
tially related to drug efflux pumps, between the tetR and tetS strains led to alteration in cell surface properties which in turn affected cell transport in saturated porous media. Findings from this research suggested that manure-derived tetR E. coli could spread more widely in the groundwater system and pose serious public health risks. ª 2010 Elsevier Ltd. All rights reserved.
1.
Introduction
The occurrence and spread of antibiotic resistant bacteria in the environment undermines our ability to prevent and control microbial infections and is becoming a growing public health challenge both within the United States and across the world. It was estimated that antibiotic resistant pathogens are responsible for more than 2 million illnesses and 14 000 deaths each year in the United States (Pruden et al., 2006; World Health Organization, 2000). The Centers for Disease Control and Prevention (CDC) reported that antibiotic resistance cost the United States more than 4.5 billion dollars in 1990 (Institute of Medicine, 1998).
For several decades, antibiotics have been commonly used in animal farms at therapeutic levels to treat diseases and at sub-therapeutic levels for growth promotion and prophylactic purposes (Institute of Medicine, 1988; Kumar et al., 2005; Mellon et al., 2001; Teuber, 2001). The widespread use of antibiotics in animal farm environments has resulted in high levels of antibiotic resistant bacteria in animal waste (Halbert et al., 2006; Hofacre et al., 2000; Parveen et al., 2006; Ray et al., 2006, 2007; Sato et al., 2004, 2005; Varela et al., 2008; Varga et al., 2008a, 2008b, 2009). Parveen et al. (2006) reported that 85%, 81%, 91% and 80% of Escherichia coli isolates retrieved from manures produced in swine, dairy, poultry, and beef farms respectively, were resistant to at least one antibiotic drug.
* Corresponding author. Tel.: þ1 414 229 6148. E-mail address:
[email protected] (S. Xu). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.12.014
1682
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 8 1 e1 6 9 0
Manure produced in animal farms is usually stored in deep pits or outdoor lagoons before being applied to agricultural fields as a source of fertilizer (Burkholder et al., 2007; Gollehon et al., 2001; Sapkota et al., 2007). Leakage from deep pits and lagoons and downward infiltration of water through manureladen soil can lead to the pollution of groundwater by antibiotic resistant bacteria (Anderson and Sobsey, 2006; Koike et al., 2007; Mackie et al., 2006; Mckeon et al., 1995; Sapkota et al., 2007; Storteboom et al., 2007). Mckeon et al. (1995) found that 100% of the non-coliforms and 87% of the coliforms isolated from rural groundwater samples were resistant to at least one of 16 antibiotics, with resistance most commonly directed toward novobiocin, cephalothin, and ampicillin. Approximately 60% of the coliforms were resistant to multiple drugs. Sapkota et al. (2007) reported that a concentrated swine feeding operation resulted in groundwater pollution by Enterococci that were resistant to erythromycin, tetracycline and clindamycin. Anderson and Sobsey (2006) found that more than 80% of the E. coli isolates from the groundwater influenced by swine farms were resistant to tetracycline and chlortetracycline. Additionally, many of the E. coli isolates were also resistant to streptomycin, trimethoprim and ampicillin. Because groundwater is the primary source of drinking water, particularly in areas where animal farms are located, contamination of groundwater by antibiotic resistant bacteria poses a direct public health threat (Ellefson et al., 2002; Solley et al., 1998). Despite the growing concerns of groundwater contamination by manure-derived antibiotic resistant bacteria, our knowledge of the transport of antibiotic resistant bacteria in the groundwater systems remains very limited. Rysz and Alvarez (2006) investigated the transport of tetracycline resistant Burkholderia cepacia and found that >46% of the bacterial cells were able to travel through a 15-cm column packed with sand and higher breakthrough concentrations were observed when the concentration of the bacterial cells was increased. The influence of tetracycline resistance on the transport of B. cepacia, however, was not specifically examined. The main goal of this research is to evaluate the impact of tetracycline resistance on the transport of manure-derived E. coli through column transport experiments. Such information is needed to assess the health risks associated with groundwater contamination by antibiotic resistant bacteria and to improve manure management practices that aim at mitigating this problem.
2.
Materials and methods
2.1. Isolation and antimicrobial susceptibility test of E. coli E. coli used in this research was isolated from manure collected from a family dairy farm (w50 milking cows) located in Ozaukee County, WI using standard protocols established by US EPA (2000). Briefly, the collected manure samples were suspended in sterile phosphate buffered saline (PBS) solution and filtered through sterile PVDF 0.45 mm membranes (Millipore). The membranes were flipped and placed onto modified mTEC agar plates (Becton Dickinson). The plates were incubated at 35 C for 2 h and then 44.5 C for 22 h. Tentative E. coli isolates retrieved from the mTEC agar plates were confirmed with
Enterotube II (Becton Dickinson) and MacConkey II agar plates containing 4-methylumbelliferyl-D-glucuronide (MUG) (Becton Dickinson). The isolated E. coli were tested for their susceptibility to 7 representative antibiotics using MuellereHinton agar plates amended with various antibiotics (Clinical and Laboratory Stardards Institute, 2006; Walczak and Xu, 2011). For each antibiotic reagent, two different concentrations were tested. No E. coli isolates were resistant to gentamicin and ciprofloxacin, while resistance to cephalothin, ampicillin, erythromycin and tetracycline was prevalent (Walczak and Xu, 2011). E. coli isolates that differed in tetracycline resistance but had otherwise similar antibiotic susceptibility patterns were selected for the column transport study (Table 1). To minimize variations that may be caused by different growth conditions and nutrient status among different cows, the E. coli isolates selected for this research were all from the same cow. The tetR strains were referred to as RES1 and RES2 and the tetS strains were denoted by SUS1 and SUS2. Polymerase chain reaction (PCR) assays were performed to determine a total of 20 tetR genes (12 efflux genes and 8 ribosome protection genes) (Hu et al., 2008). The primers used for the PCR assays and the primer-dependent annealing temperatures can be found elsewhere (Aminov et al., 2002, 2001; Hu et al., 2008; Miranda et al., 2003). Plasmid DNA was isolated from each strain by alkaline lysis (Sambrook and Russell, 2001), and served as the template for the PCR reactions. The PCR mixture also contained 0.3 mM of each primer (Invitrogen), 0.4 units of Vent polymerase (New England Biolabs), 200 mM dNTPs, 3% dimethyl sulfoxide and 1 PCR buffer. The PCR amplification was performed using a Mastercycler thermocycler (Eppendorf). The temperature program consisted of an initial denaturing step of 94 C for 5 min, followed by 30 cycles of 94 C for 30 s, annealing for 30 s, extension at 72 C for 30 s, and a final extension of 10 min at 72 C. Negative control reactions were included for each set of primers. PCR products were analyzed on 1% agarose gel that was stained with ethidium bromide.
2.2.
Column transport studies
Duplicate chromatography columns measuring 2.5 cm in diameter and 15 cm in length were used for the column transport experiments. The columns were packed with silica sands (US Silica) (size range: 0.707e0.841 mm) that were alternately cleaned with hot, concentrated nitric acids and diluted NaOH. The porosity of the sand was 0.356. Peristaltic pumps (ColeeParmer) were used to maintain a constant specific discharge value of 0.31 cm/min. Packed sand columns were equilibrated with >40 pore volumes of background electrolyte solution (1e100 mM KCl) before the bacteria transport experiment. E. coli preserved in 20% glycerol under 80 C was streaked onto Muller-Hinton (MH) agar plates. After overnight incubation at 37 C, cells from the freshly formed colonies were transferred to culture tubes containing 15 mL Luria-Bertani (LB) broth. The culture tubes were incubated at 37 C for 6 h. The starter culture was used to inoculate LB broth (1:500 dilution ratio), which was then incubated at 37 C for 18 h. The bacterial cells were harvested using centrifuge (4000 g, 10 min, 4 C). To
1683
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 8 1 e1 6 9 0
Table 1 e Antibiotic resistance pattern of the E. coli isolates used for the transport experiments. E. coli isolate
Tetracycline a
RES1 RES2 SUS1 SUS2
Ampicillin
Cephalothin
Gentamicin
Ciprofloxacin
Erythromycin
Sulfomethoxazole
4
16
8
32
8
32
4
16
1
4
3
15
38
152
þ þ e e
þ þ e e
þ þ þ e
þ þ e e
þ þ þ þ
þ þ þ e
e e e e
e e e e
e e e e
e e e e
þ þ þ þ
þ þ þ þ
e e e e
e e e e
a Unit is mg/L.
remove the growth medium, the bacterial pellet was rinsed 4 times with the appropriate electrolyte solution. The concentration of cells was then adjusted to ∼4 107 cell/ml for the column transport experiments. The pH of the background electrolyte solutions and the cell suspension was around 5.7. The effluents of the columns were connected to flow-through quartz cuvettes (NSG Precision) and the cell concentration was determined at a wavelength of 220 nm at 30-s intervals using a spectrophotometer (Shimadzu UV-1700). After 60 min of injection (w3.5 pore volumes), the columns were flushed with background electrolyte solution until the absorbance of effluent returned to the background values. The deposition kinetics of bacterial cells in saturated porous media under clean-bed conditions is commonly quantified by the deposition rate coefficient (K ), which is determined using the following equation (Kretzschmar et al., 1999) n C K ¼ ln qL C0
(1)
where v is approach velocity (cm/min), L is the length of the packed bed (cm), q is the porosity of the porous medium (cm3/ cm3), C is concentration of bacteria cell in the effluent under clean-bed conditions (cell/mL) and C0 is the bacteria concentration in the influent (cell/mL). The values of C/C0 were determined for each column experiment by calculating the average normalized breakthrough concentrations (i.e., C/C0) measured between 1.8 and 2 pore volumes (Walker et al., 2005).
2.3.
Determination of cell size and surface properties
The transport of bacteria cells in saturated porous media is influenced by a variety of cell surface properties such as zeta potential, surface charge and hydrophobicity which in turn were related to factors such as cell-bound proteins and LPS (Foppen and Schijven, 2006). In this research, a range of cell surface properties were determined and related to the transport of tetR and tetS E. coli cells. Freshly harvested bacterial cells were suspended in appropriate KCl solutions (1e100 mM) at a concentration of ∼107 cells/mL. The zeta potential of the bacterial cells was then measured with a Brookhaven ZetaPALS analyzer utilizing phase analysis light scattering. The hydrophobicity of the cells was determined through microbial adhesion to hydrocarbon (MATH) test (Pembrey et al., 1999). The MATH test involved the mixing of 1 mL of n-dodecane and 4 mL of cell suspension prepared using KCl solutions. The mixture was vortexed for 2 min and then allowed to stand still for 15 min. Cell concentration in the aqueous phase was determined at a wavelength
of 546 nm. The fraction of bacterial cells that partitioned into the hydrocarbon phase was calculated based on mass balance and expressed cell hydrophobicity. To quantify EDTA-extractable cell-bound proteins, fresh cell suspensions (∼3 108 cell/ml) were prepared and mixed with 2.5% EDTA solution on a 2:3 (v/v) basis (Zhang et al., 1999). Following a 30-min incubation period at 4 C, the mixture was centrifuged at 10,400 g (4 C) for 50 min. The supernatant was then decanted and filtered through 0.45 mm filters. Protein contents in the filtrates were quantified using the Coomassie brilliant blue method developed by Bradford (1976). Standard solutions prepared with human serum albumin were used to calibrate this method. Additionally, the outer membrane proteins (OMPs) of the E. coli cells were extracted and profiled using sodium dodecyl-sulfate polyacrylamide gel (SDS-PAGE) (Ben Abdallah et al., 2009; Gatewood et al., 1994; Xu et al., 2006). The bacterial cells were harvested at 4000 g for 10 min at 4 C and rinsed twice with sterile 0.15 M NaCl. The cells were suspended in 5 mL sterile 0.15 M NaCl and subsequently disrupted by intermittent sonic oscillation (50 W, 8 cycles of 15 s of sonication, VirTis Virsonic). Following centrifugation at 5000 g for 40 min to remove cellular debris, the supernatant was transferred and centrifuged at 100 000 g for 40 min at 4 C. The pellets were resuspended in 2% sodium lauryl sarcosinate (Sigma), incubated at room temperature for 1 h and then centrifuged at 100,000 g (40 min, 4 C). The pelleted OMPs were suspended in 0.25 mL sterile 0.15 M NaCl solution and resolved on 10% sodium dodecyl-sulfate polyacrylamide gel (Laemmli). The gel was fixed in 4% perchloric acid and strained using 0.1% Coomassie G250 in hot 4% perchloric acid (Faguy et al., 1996). To extract cell-bound LPS, 5 mL cell suspension was placed into 50 mL centrifuge tubes and subjected to sonic disruption (50 W, 20 s) (Liu et al., 2007). The resulting suspension was centrifuged (10 000 g, 40 min, 4 C) and filtered through 0.2 mm cellulose acetate filters. The contents of LPS in the filtrates were measured using the phenol-sulfuric acid method and xanthan gum was used as the calibration standard (Du Bois et al., 1956). Acid-base titration of cell suspensions was performed to determine the surface acidity of bacterial cells. Cell suspensions were prepared in the same fashion as those used in the column transport experiments. Two hundred milliliter of the suspension was added to a medium bottle and purged with high purity nitrogen gas for >60 min to remove CO2. After the purging step, running nitrogen gas flow was maintained right above the surface of the suspension to maintain a CO2 free environment. Sulfuric acid solution (0.1600 N, Hach Company) was then introduced into the suspension using Hach digital
1684
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 8 1 e1 6 9 0
titrator. Once the pH of the suspension dropped to w4, NaOH (0.1600 N, Hach Company) was added at small volume increments to raise the pH to ∼10. The pH of the suspension was continuously monitored and recorded using a pH probe (AccupH, Fisher Scientific). The acidity of bacterial cells was calculated based on the amount of NaOH that was consumed to raise the pH of the cell suspension from 4 to 10. The size of the bacterial cells suspended in the KCl solutions were measured by taking photos using a Nikon Eclipse 50i microscope, which was equipped with a Photometric coolsnap ES digital camera and MetaMorph software. The length and width of a minimum of 50 cells were determined using the ImageJ software and the equivalent radii of the cells were then calculated.
3.
Results and discussion
3.1.
Characterization of tetR genes
PCR assays showed that none of the tetR genes were present in the tetS strains while the efflux gene, tetB, was present in both tetR strains (Fig. 1). This observation is consistent to previously reported results which suggested that the tetB gene was among the most commonly detected tetR genes in E. coli and coliforms (Hu et al., 2008; Marshall et al., 1983). In Hu et al. (2008), the tetB gene was found in 41% of tetracycline resistant E. coli strains isolated from a river basin impaired by both human and animal farm waste.
3.2.
Transport of the bacterial cells in the sand packs
The two tetS E. coli strains differ in ampicillin and cephalothin susceptibility but the close match in their breakthrough concentrations suggested that the difference in ampicillin and
cephalothin resistance had minimal impact on their mobility (Figs. 2 and 3). The breakthrough concentrations of the tetR E. coli strains, however, were significantly higher than those of the tetS strains under all ionic strength conditions, suggesting that tetracycline resistance can enhance the mobility of manure-derived E. coli (Figs. 2 and 3). In 1 mM KCl, for instance, the breakthrough concentrations of the tetS were ∼15% lower than the breakthrough concentrations of the tetR strains. Accordingly, the values of the deposition rate coefficients (i.e., K ) for the tetS and tetR strains were 0.0325(0.0012) (SUS1), 0.0306(0.0018) (SUS2), 0.0172(0.000005) (RES1) and 0.0169 (0.0009) (RES2) min1, respectively. When the ionic strength was increased to 3 mM, there was a significant drop in the breakthrough concentrations for all 4 E. coli isolates, suggesting that within this range, higher ionic strength facilitated their deposition at the surface of quartz sand (Figs. 2 and 3). While a further increase in ionic strength led to slightly lower breakthrough concentrations for the tetS strains, the transport of the tetR strains remained virtually unchanged (Figs. 2 and 3). Overall, the tetR E. coli isolates displayed significantly higher mobility under the ionic strength conditions tested in this research (Fig. 3). Our results suggested that environmental E. coli isolates could display marked variability in mobility, which was reported in several recent publications (Bolster et al., 2010, 2009; Foppen et al., 2010; Lutterodt et al., 2009). Bolster et al. (2009) and Bolster et al. (2010) compared the transport of 12 and 8 E. coli strains isolated from different animal sources (poultry, horse, beef and dairy cattle, human and wildlife) in saturated sands, respectively, and observed large variability in their mobility. Lutterodt et al. (2009) investigated the movement of 6 E. coli strains obtained from a soil used for cattle grazing in columns packed with sands and reported that the sticking efficiencies varied by a factor of 4e10. Foppen et al. (2010) examined the transport behavior of 54 E. coli strains and found that the attachment efficiency varied by a factor of ∼6. The observed variability in E. coli transport behavior was found to be related to a range of factors such as cell surface autotransporter proteins (e.g., Ag43 protein) (Lutterodt et al., 2009), cell width (Bolster et al., 2010; Bolster et al., 2009), cell surface/zeta potential (Bolster et al., 2010), cell LPS structure (Foppen et al., 2010) as well as cell fimbriae (Foppen et al., 2010).
3.3.
Fig. 1 e PCR detection of tetB in the E. coli isolates. Lane M: 100 bp ladder (New England Biolabs). Lanes 1e4: SUS1, SUS2, RES1, and RES2. The size of the amplicon for the tetB gene was 206 bp.
Size and surface properties of bacterial cells
The transport behavior of bacterial cells in the porous media is governed by the energy interactions between the cells and the surface of the solid matrix, which depends on a range of factors such as the Lifshitzevan der Waals force, electrostatic double layer (EDL) interactions, acid-base forces, hydrophobicity interactions and steric effects, which in turn are affected by cell size as well as cell-bound LPS and proteins and so on (Lindqvist and Bengtsson, 1991; Ong et al., 1999). In this research, various cell surface properties were measured and related to the observed difference in the transport behavior of the tetR and tetS E. coli strains. The EDL interactions between bacterial cells and sand surface were closely related to their zeta potentials. The measured zeta potential values of the bacterial cells and the
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 8 1 e1 6 9 0
1685
Fig. 2 e Breakthrough concentrations of the tetracycline susceptible (SUS1 and SUS2) and tetracycline resistant (RES1 and RES2) E. coli strains in saturated porous media under ionic strength conditions of 1, 3, 10, 30 and 100 mM KCl.
sands were negative (Fig. 4A), suggesting repulsive EDL interactions. In general, the zeta potentials of the tetR strains were more negative than those of the tetS strains. As a result, the repulsive interaction between the surface of quartz sands and the tetR E. coli cells were stronger and the deposition rates
Fig. 3 e Clean-bed deposition rate coefficients (K, minL1) for the 4 E. coli strains under ionic strength conditions of 1, 3, 10, 30 and 100 mM KCl. The values of K were calculated using Eq. (1). Error bars represent standard deviation of duplicate experiments.
should thus be lower. It is noteworthy that the zeta potentials of the tetR strains remained virtually unchanged with ionic strength, while the zeta potentials of the tetS strains increased slightly with ionic strength. This is consistent with the observation that the transport of the tetR strains was less sensitive to changes in ionic strength. The tetR strains were more hydrophobic than the tetS ones (Fig. 4B). On average, slightly over 90% of the tetR E. coli cells partitioned into the hydrocarbon phase in the MATH tests, while less than 85% of the tetS E. coli cells migrated from the aqueous phase into the hydrocarbon phase. In this research we observed higher mobility for the more hydrophobic tetR strains. It was previously suggested, however, that cell hydrophobicity could enhance the attachment of bacterial cells to clean quartz sands (Bolster et al., 2006; Mccaulou et al., 1994). It thus seemed that the EDL interactions between E. coli cells and quartz sands were more significant than the hydrophobic interactions. This is consistent to recent findings which suggested that while cell zeta potential was significantly related to the transport of manure-derived E. coli strains, the relationship between cell hydrophobicity and E. coli mobility was statistically insignificant (Bolster et al., 2010). Size is another factor that could influence the transport of colloid-sized particles in saturated porous media (Bolster et al., 2010; Yao et al., 1971). In this research, the size of the
1686
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 8 1 e1 6 9 0
Fig. 5 e Size (equivalent radius) of the bacterial cells suspended in 1, 3, 10, 30 and 100q mM of KCl. The ffiffiffiffiffiffiffiffiffiffiffiffi C equivalent size was calculated as LC 3W p , where LC and WC represent the length and width of the cell, respectively (Haznedaroglu et al., 2008). Error bars represent the standard deviation of a minimum of 50 measurements.
Fig. 4 e Zeta potential (A) and MATH test results (B). The MATH test results were expressed as the fraction of bacterial cells partitioned into the hydrocarbon phase. Error bars represent the standard deviation of triplicate measurements. For zeta potential, one measurement contained a minimum of 5 runs.
bacterial cells suspended in 5 different types of electrolyte solutions was measured. Our results showed that the size of the bacterial cells was not sensitive to ionic strength and the tetR and tetS cells have practically similar sizes (Fig. 5). Through titration of cells suspended in 3 mM KCl, the acidity of the bacterial cells were measured as 6.10 (0.16), 6.54 (0.60), 7.07 (0.83) and 6.28 (0.24) 104 meq/108 cells for SUS1, SUS2, RES1 and RES2, respectively, suggesting no significant difference between the susceptible and resistant isolates. The potentiometric titration curves showed that the pKa values of the acid-base functional groups on the surfaces of both tetracycline resistant and tetracycline susceptible cells were between 4e5 as well as 9e10. These pKa values correspond to the carboxylic/phosphoric and hydroxyl/amine groups, all of which are important components of LPS, proteins and phospholipids located on the surface of bacterial cells (Hong and Brown, 2006). Under the experimental pH conditions (∼5.7), the carboxylic/phosphoric groups were deprotonated and contributed to cell surface charges. Cell-bound LPS and protein contents varied within the range of 17.4e33.1 and 4.7e13 mg/108 cells, respectively (Fig. 6).
Fig. 6 e Comparison of the LPS sugar (A) and protein contents (B) of the tetracycline resistant and susceptible cells suspended in 1, 3, 10, 30 and 100 mM KCl. The error bars represent standard deviation of triplicate extraction attempts.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 8 1 e1 6 9 0
These values were consistent with previously reported measurements using manure-derived E. coli (Haznedaroglu et al., 2008). There were significant variations among the extraction attempts of LPS and protein, and among the electrolyte solutions used to prepare cell suspensions. Overall, we did not observe a clear pattern between the tetR and tetS strains with regard to LPS sugar and protein contents. Although no difference was identified in the cell-bound protein contents, results of SDS-PAGE analysis of OMPs suggested that different proteins existed on the outer membrane of tetR and tetS strains (Fig. 7). Specifically, there were at least four proteins present in the outer membrane of the tetR strains that were absent in the tetS strains (indicated by arrows in Fig. 7). These proteins had approximate molecular masses of 54, 47, 44 and 40 kDa. Additionally, there were three proteins (indicated with arrow heads in Fig. 7) that were present in the tetS strains but were absent in the tetR strains. As an antimicrobial agent, tetracycline has an intracellular target, inhibiting bacterial protein synthesis by disrupting the interaction of aminoacyl-tRNA with the ribosome (Walsh, 2003). Mechanisms specific to tetracycline resistance include efflux pump, ribosomal protection and modification of the antibiotic (Wax, 2008). The ribosomal protection mechanisms involve soluble structural homologues (e.g., TetM proteins) of elongation factors, which can bind to the ribosome and destabilize the interaction between tetracycline and their cellular target (Burdett, 1996; Dantley et al., 1998). Resistance to tetracycline through drug destruction was relatively rare and it was recently reported that TetX, a flavin-dependent monooxygenase, could hydroxylate the tetracycline substrate into an unstable compound which subsequently underwent non-enzymatic decomposition (Yang et al., 2004). The efflux pumps that are encoded by the tetR genes (e.g., tetB) involve the transport of tetracycline from the cytoplasm to the
periplasm through proteins inserted in the cytoplasmic membrane (Walsh, 2003). Out of the 20 tetR genes examined in this research, only the tetB gene was detected in the tetR strains (Fig. 1). Because the protein involved in the efflux pump is not exposed to the outside of the bacterial cells, it was unlikely that the presence of TetB could impact cell mobility. The tetR gene family, however, does not represent all the mechanisms responsible for tetracycline resistance in E. coli. Additional efflux pumps that involve multi-protein assemblies that often traverse both the inner and outer membranes of E. coli (e.g., the AcrAB-TolC pump) could lead to tetracycline resistance (Alekshun and Levy, 2007; de Cristobal et al., 2006; Xu et al., 2006). Increased expression of proteins such as TolC, OmpC, OmpW, along with decreased amounts of LamB and NlpB proteins have been observed in tetR E. coli (Xu et al., 2006; Zhang et al., 2008). Additionally, it was observed that deletion of TolC led to increased sensitivity of E. coli to tetracycline (Zhang et al., 2008). Because processed TolC has a molecular mass of 52 kDa, it is likely that the protein extracted from the tetR strains that migrated approximately at 54 kDa was TolC (Fig. 7). The crystal structure of the TolC protein was recently resolved and it was shown that it is a trimeric, 471-residue protein that contains an a-helical barrel and a b-barrel (Koronakis et al., 2000). The a-helical domain, which forms a tunnel through the periplasm and measure 10 nm in length, is anchored to bacterial wall by the contiguous b-barrel, which has a length of 4 nm and extends to the outside of the outer membrane. In addition to the approximately 54 kDa protein enriched in the tetR cells, there were at least three other proteins that are differentially enriched in the tetR cells. It is likely that the presence of these proteins, in combination with the absence of other proteins (indicated by arrow heads, Fig. 7), contributed to alterations in cell surface properties (e.g., zeta potential), which in turn impacted the mobility of the E. coli strains. Recently, it was reported that Ag43, an outer membrane protein of E. coli, could enhance the attachment of E. coli cells to the surface of quartz sands (Lutterodt et al., 2009). It was proposed that the positive charges of the a-domain of Ag43, which extends from the cell surface, facilitated the attachment of E. coli cells to the negatively-charged quartz surfaces (Lutterodt et al., 2009). While the Ag43 and TolC proteins are structurally different and have different impacts on E. coli transport in saturated porous media (Koronakis et al., 2000; van der Woude and Henderson, 2008), the results of Lutterodt et al. (2009) and this research suggest that cell surface proteins can have strong influences on E. coli transport and more studies will be needed to elucidate the relationship between the abundance, structure and properties of cell surface proteins and bacterial transport in porous media.
3.4. Fig. 7 e Outer membrane protein profiles of SUS1, SUS2, RES1 and RES2 using SDS-PAGE. Molecular masses (kDa) are indicated on the left. Proteins that are present in the tetracycline resistant strains, but absent in the tetracycline susceptible strains, are indicated with an arrow. Proteins that are present in the tetracycline susceptible strains, but absent in the tetracycline resistant strains are indicated with arrow heads.
1687
Environmental implications
As a result of the widespread use of antibiotics in the animal farm environment, high frequencies of antibiotic resistant bacteria have been detected in animal waste (Halbert et al., 2006; Parveen et al., 2006; Ray et al., 2006; Sapkota et al., 2007; Sato et al., 2004; Varga et al., 2008a; Walczak and Xu, 2011). On the dairy farm from which the E. coli strains used in this research were isolated, 13.1%, 72.7%, 80.5% of the E. coli isolates were resistant to tetracycline, cephalothin and
1688
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 8 1 e1 6 9 0
erythromycin, respectively (Walczak and Xu, 2011). Additionally, 100% of the tetR E. coli isolates were multi-drug resistant. The observed high mobility of the tetR E. coli strains indicated that leakage from manure storage structures and application of manure as fertilizers in agricultural fields could potentially lead to the contamination of groundwater by antibiotic resistant bacteria, which in turn could pose serious public health risks when the groundwater, often untreated, is used as a source of drinking water. Drug efflux pumps that involve outer membrane proteins like TolC are also seen in pathogenic bacteria such as Salmonella to gain antibiotic resistance (Ricci et al., 2006; Virlogeux-Payant et al., 2008). It is likely that the surface properties of these pathogens can also be altered in a similar fashion. As a result, the mobility of these pathogens in the subsurface system can be enhanced. Furthermore, it has long been observed that antibiotic resistance genes, which confer bacterial antibiotic resistance, can be transferred among a diverse group of microorganisms through conjugation, transduction and transformation (Levy et al., 1976; Lorenz et al., 1992; Mckeon et al., 1995; Nikolich et al., 1994). The antibiotic resistant genes harbored by antibiotic resistant E. coli, therefore, could potentially be horizontally transferred to bacterial pathogens such as Samonella in the subsurface environment and cause additional public health risks (Hunter et al., 1992; van Essen-Zandbergen et al., 2007).
4.
Conclusion
In this research, we observed that manure-derived, tetR E. coli strains had higher mobility than tetS E. coli strains within saturated porous media. The tetR E. coli strains had more negative zeta potentials than the tetS strains. This led to increased repulsive EDL interaction between the tetR E. coli cells and the surface of quartz sands and could explain the observed higher mobility of the tetR strains. The tetR and tetS E. coli strains had distinct outer membrane proteins profiles. It is likely that such difference led to alterations in cell surface properties (such as zeta potential), which in turn affected the transport of the tetR and tetS E. coli strains.
Acknowledgement SX was supported by University of Wisconsin Milwaukee and University of Wisconsin Groundwater Research Program (WR10R007). SLB is the recipient of an NIH grant (5R00GM08314704). We thank Dr. Douglas A. Steeber, Dr. Heather A. Owen and Steven E. Hardcastle for their assistance. We thank two reviewers whose comments led to improvements of our manuscript.
references
Alekshun, M.N., Levy, S.B., 2007. Molecular mechanisms of antibacterial multidrug resistance. Cell 128, 1037e1050.
Aminov, R.I., Chee-Sanford, J.C., Garrigues, N., Teferedegne, B., Krapac, I.J., White, B.A., Mackie, R.I., 2002. Development, validation, and application of PCR primers for detection of tetracycline efflux genes of gram-negative bacteria. Applied and Environmental Microbiology 68, 1786e1793. Aminov, R.I., Garrigues-Jeanjean, N., Mackie, R.I., 2001. Molecular ecology of tetracycline resistance: development and validation of primers for detection of tetracycline resistance genes encoding ribosomal protection proteins. Applied and Environmental Microbiology 67, 22e32. Anderson, M.E., Sobsey, M.D., 2006. Detection and occurrence of antimicrobially resistant E-coli in groundwater on or near swine farms in eastern North Carolina. Water Science and Technology 54, 211e218. Ben Abdallah, F., Kallel, H., Bakhrouf, A., 2009. Enzymatic, outer membrane proteins and plasmid alterations of starved Vibrio parahaemolyticus and Vibrio alginolyticus cells in seawater. Archives of Microbiology 191, 493e500. Bolster, C.H., Cook, K.L., Marcus, I.M., Haznedaroglu, B.Z., Walker, S.L., 2010. Correlating transport behavior with cell properties for Eight Porcine Escherichia coli isolates. Environmental Science & Technology 44, 5008e5014. Bolster, C.H., Haznedaroglu, B.Z., Walker, S.L., 2009. Diversity in cell properties and transport behavior among 12 different environmental Escherichia coli isolates. Journal of Environmental Quality 38, 465e472. Bolster, C.H., Walker, S.L., Cook, K.L., 2006. Comparison of Escherichia coli and Campylobacter jejuni transport in saturated porous media. Journal of Environmental Quality 35, 1018e1025. Bradford, M.M., 1976. A rapid and sensitive method for the quantitation of microgram quantities of protein utilizing the principle of protein-dye binding. Analytical Biochemistry 72, 248e254. Burdett, V., 1996. Tet(M)-promoted release of tetracycline from ribosomes is GTP dependent. Journal of Bacteriology 178, 3246e3251. Burkholder, J., Libra, B., Weyer, P., Heathcote, S., Kolpin, D., Thorne, P.S., Wichman, M., 2007. Impacts of waste from concentrated animal feeding operations on water quality. Environmental Health Perspectives 115, 308e312. Clinical and Laboratory Stardards Institute, 2006. Performance Standards for Antimicrobial Disk Susceptibility Tests; Approved Standard - Ninth Edition. CLSI, Wayne, PA. Dantley, K.A., Dannelly, H.K., Burdett, V., 1998. Binding interaction between Tet(M) and the ribosome: requirements for binding. Journal of Bacteriology 180, 4089e4092. de Cristobal, R.E., Vincent, P.A., Salomon, R.A., 2006. Multidrug resistance pump AcrAB-TolC is required for high-level, Tet(A)mediated tetracycline resistance in Escherichia coli. Journal of Antimicrobial Chemotherapy 58, 31e36. Du Bois, M., Gilles, K.A., Hamilton, J.K., Rebers, P.A., Smith, F., 1956. Colorimetric method for determination of sugars and related substances. Analytical Chemistry 28, 350e356. Ellefson, B.R., Mueller, G.D., Buchwald, C.A., 2002. Water Use in Wisconsin, 2000. U. S. Geological Survey. open-file report 02e356. Faguy, D.M., Bayley, D.P., Kostyukova, A.S., Thomas, N.A., Jarrell, K.F., 1996. Isolation and characterization of flagella and flagellin proteins from the thermoacidophilic archaea Thermoplasma volcanium and Sulfolobus shibatae. Journal of Bacteriology 178, 902e905. Foppen, J.W., Lutterodt, G., Roling, W.F.M., Uhlenbrook, S., 2010. Towards understanding inter-strain attachment variations of Escherichia coli during transport in saturated quartz sand. Water Research 44, 1202e1212. Foppen, J.W.A., Schijven, J.F., 2006. Evaluation of data from the literature on the transport and survival of Escherichia coli and
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 8 1 e1 6 9 0
thermotolerant coliforms in aquifers under saturated conditions. Water Research 40, 401e426. Gatewood, D.M., Fenwick, B.W., Chengappa, M.M., 1994. GrowthCondition dependent expression of Pasteurella-Haemolytica a1 outer-membrane proteins, Capsule, and Leukotoxin. Veterinary Microbiology 41, 221e233. Gollehon, N., Caswell, M., Ribaudo, M., Kellogg, R., Lander, C., Letson, D., 2001. Confined Animal Production and Manure Nutrients. Resource Economics Division, Economic Research Service. U.S. Department of Agriculture, Washington DC. 39. Halbert, L.W., Kaneene, J.B., Ruegg, P.L., Warnick, L.D., Wells, S.J., Mansfield, L.S., Fossler, C.R., Campbell, A.M., Geiger-Zwald, A.M., 2006. Evaluation of antimicrobial susceptibility patterns in Campylobacter spp isolated from dairy cattle and farms managed organically and conventionally in the midwestern and northeastern United States. Journal of the American Veterinary Medical Association 228, 1074e1081. Haznedaroglu, B.Z., Bolster, C.H., Walker, S.L., 2008. The role of starvation on Escherichia coli adhesion and transport in saturated porous media. Water Research 42, 1547e1554. Hofacre, C.L., ’de Cotret, A.R., Maurer, J.J., Garritty, A., Thayer, S.G. , 2000. Presence of fluoroquinolone-resistant coliforms in poultry litter. Avian Diseases 44, 963e967. Hong, Y., Brown, D.G., 2006. Cell surface acid-base properties of Escherichia coli and Bacillus brevis and variation as a function of growth phase, nitrogen source and C: N ratio. Colloids and Surfaces B-Biointerfaces 50, 112e119. Hu, J.Y., Shi, J.C., Chang, H., Li, D., Yang, M., Kamagata, Y.C., 2008. Phenotyping and genotyping of antihiotic-resistant Escherichia coli isolated from a natural river basin. Environmental Science & Technology 42, 3415e3420. Hunter, J.E.B., Shelley, J.C., Walton, J.R., Hart, C.A., Bennett, M., 1992. Apramycin resistance plasmids in Escherichia coli: possible transfer to salmonella typhimurium in calves. Epidemiology and Infection 108, 271e278. Institute of Medicine, 1988. Human Health Risks with the Subtherapeutic Use of Penicillin or Tetracyclines in Animal Feed. Institute of Medicine, 1998. Antimicrobial Resistance: Issues and Options. National Academy Press, Washington D.C. Koike, S., Krapac, I.G., Oliver, H.D., Yannarell, A.C., CheeSanford, J.C., Aminov, R.I., Mackie, R.I., 2007. Monitoring and source tracking of tetracycline resistance genes in lagoons and groundwater adjacent to swine production facilities over a 3year period. Applied and Environmental Microbiology 73, 4813e4823. Koronakis, V., Sharff, A., Koronakis, E., Luisi, B., Hughes, C., 2000. Crystal structure of the bacterial membrane protein TolC central to multidrug efflux and protein export. Nature 405, 914e919. Kretzschmar, R., Borkovec, M., Grolimund, D., Elimelech, M., 1999. Mobile subsurface colloids and their role in contaminant transport. Advances in Agronomy 66, 121e193. Kumar, K., Gupta, S.C., Chander, Y., Singh, A.K., 2005. Antibiotic use in agriculture and its impact on the terrestrial environment. Advances in Agronomy 87, 1e54. Levy, S.B., Fitzgerald, G.B., Macone, A.B., 1976. Spread of antibiotic-resistant plasmids from chicken to chicken and from chicken to man. Nature 260, 40e42. Lindqvist, R., Bengtsson, G., 1991. Dispersal Dynamics of groundwater bacteria. Microbial Ecology 21, 49e72. Liu, Y., Yang, C.H., Li, J., 2007. Influence of extracellular polymeric substances on Pseudomonas aeruginosa transport and deposition profiles in porous media. Environmental Science & Technology 41, 198e205. Lorenz, M.G., Reipschlager, K., Wackernagel, W., 1992. Plasmid transformation of naturally competent Acinetobacter calcoaceticus in nonsterile soil extract and groundwater. Archives of Microbiology 157, 355e360.
1689
Lutterodt, G., Basnet, M., Foppen, J.W.A., Uhlenbrook, S., 2009. The effect of surface characteristics on the transport of multiple Escherichia coli isolates in large scale columns of quartz sand. Water Research 43, 595e604. Mackie, R.I., Koike, S., Krapac, I., Chee-Sanford, J., Maxwell, S., Aminov, R.I., 2006. Tetracycline residues and tetracycline resistance genes in groundwater impacted by swine production facilities. Animal Biotechnology 17, 157e176. Marshall, B., Tachibana, C., Levy, S.B., 1983. Frequency of tetracycline resistance determinant classes among lactosefermenting coliforms. Antimicrobial Agents and Chemotherapy 24, 835e840. Mccaulou, D.R., Bales, R.C., Mccarthy, J.F., 1994. Use of Short-Pulse experiments to study bacteria transport through porousmedia. Journal of Contaminant Hydrology 15, 1e14. Mckeon, D.M., Calabrese, J.P., Bissonnette, G.K., 1995. Antibioticresistant Gram-negative bacteria in rural groundwater supplies. Water Research 29, 1902e1908. Mellon, M., Benbrook, C., Benbrook, K.L., 2001. Hogging It!: Estimates of Antimicrobial Abuse in Livestock. Miranda, C.D., Kehrenberg, C., Ulep, C., Schwarz, S., Roberts, M.C., 2003. Diversity of tetracycline resistance genes in bacteria from Chilean salmon farms. Antimicrobial Agents and Chemotherapy 47, 883e888. Nikolich, M.P., Hong, G., Shoemaker, N.B., Salyers, A.A., 1994. Evidence for natural horizontal transfer of Tetq between bacteria that normally colonize humans and bacteria that normally colonize livestock. Applied and Environmental Microbiology 60, 3255e3260. Ong, Y.L., Razatos, A., Georgiou, G., Sharma, M.M., 1999. Adhesion forces between E-coli bacteria and biomaterial surfaces. Langmuir 15, 2719e2725. Parveen, S., Lukasik, J., Scott, T.M., Tamplin, M.L., Portier, K.M., Sheperd, S., Braun, K., Farrah, S.R., 2006. Geographical variation in antibiotic resistance profiles of Escherichia coli isolated from swine, poultry, beef and dairy cattle farm water retention ponds in Florida. Journal of Applied Microbiology 100, 50e57. Pembrey, R.S., Marshall, K.C., Schneider, R.P., 1999. Cell surface analysis techniques: what do cell preparation protocols do to cell surface properties? Applied and Environmental Microbiology 65, 2877e2894. Pruden, A., Pei, R.T., Storteboom, H., Carlson, K.H., 2006. Antibiotic resistance genes as emerging contaminants: studies in northern Colorado. Environmental Science & Technology 40, 7445e7450. Ray, K.A., Warnick, L.D., Mitchell, R.M., Kaneene, J.B., Ruegg, P.L., Wells, S.J., Fossler, C.P., Halbert, L.W., May, K., 2006. Antimicrobial susceptibility of Salmonella from organic and conventional dairy farms. Journal of Dairy Science 89, 2038e2050. Ray, K.A., Warnick, L.D., Mitchell, R.M., Kaneene, J.B., Ruegg, P.L., Wells, S.J., Fossler, C.P., Halbert, L.W., May, K., 2007. Prevalence of antimicrobial resistance among Salmonella on midwest and northeast USA dairy farms. Preventive Veterinary Medicine 79, 204e223. Ricci, V., Tzakas, P., Buckley, A., Coldham, N.C., Piddock, L.J.V., 2006. Ciprofloxacin-resistant Salmonella enterica serovar Typhimurium strains are difficult to select in the absence of AcrB and TolC. Antimicrobial Agents and Chemotherapy 50, 38e42. Rysz, M., Alvarez, P.J.J., 2006. Transport of antibiotic-resistant bacteria and resistance-carrying plasmids through porous media. Water Science and Technology 54, 363e370. Sambrook, J., Russell, D.W., 2001. Molecular Cloning: A Laboratory Manual, third ed. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, N.Y.
1690
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 8 1 e1 6 9 0
Sapkota, A.R., Curriero, F.C., Gibson, K.E., Schwab, K.J., 2007. Antibiotic-resistant Enterococci and Fecal Indicators in Surface Water and Groundwater Impacted by a Concentrated Swine Feeding Operation. In: Environmental Health Perspectives, 115 1040e1045. Sato, K., Bartlett, P.C., Kaneene, J.B., Downes, F.P., 2004. Comparison of prevalence and antimicrobial susceptibilities of Campylobacter spp. isolates from organic and conventional dairy herds in Wisconsin. Applied and Environmental Microbiology 70, 1442e1447. Sato, K., Bartlett, P.C., Saeed, M.A., 2005. Antimicrobial susceptibility of Escherichia coli isolates from dairy farms using organic versus conventional production methods. Journal of the American Veterinary Medical Association 226, 589e594. Solley, W.B., Pierce, R.R., Perlman, H.A., 1998. Estimated Use of Water in the United States in 1995. U.S. Geological Survey Circular, 1200, p. 79. Storteboom, H.N., Kim, S.C., Doesken, K.C., Carlson, K.H., Davis, J.G., Pruden, A., 2007. Response of antibiotics and resistance genes to high-intensity and low-intensity manure management. Journal of Environmental Quality 36, 1695e1703. Teuber, M., 2001. Veterinary use and antibiotic resistance. Current Opinion in Microbiology 4, 493e499. US EPA, 2000. Improved Enumeration Methods for the Recreational Water Quality Indicators: Enterococci and Escherichia Coli. http:// www.epa.gov/nerlcwww/RecManv.pdf, p. 52. van der Woude, M.W., Henderson, I.R., 2008. Regulation and function of Ag43 (Flu). Annual Review of Microbiology 62, 153e169. van Essen-Zandbergen, A., Smith, H., Veldman, K., Mevius, D., 2007. Occurrence and characteristics of class 1, 2 and 3 integrons in Escherichia coli, Salmonella and Campylobacter spp. in the Netherlands. Journal of Antimicrobial Chemotherapy 59, 746e750. Varela, N.P., Friendship, R., Dewey, C., Valdivieso, A., 2008. Comparison of Agar Dilution and E-test for antimicrobial susceptibility testing of Campylobacter coli isolates recovered form 80 Ontario swine farms. Canadian Journal of Veterinary Research-Revue Canadienne De Recherche Veterinaire 72, 168e174. Varga, C., Rajic, A., McFall, M.E., Avery, B.P., Reid-Smith, R.J., Deckert, A., Checkley, S.L., McEwen, S.A., 2008a. Antimicrobial resistance in generic Escherichia coli isolated from swine fecal samples in 90 Alberta finishing farms. Canadian Journal of Veterinary Research-Revue Canadienne De Recherche Veterinaire 72, 175e180. Varga, C., Rajic, A., McFall, M.E., Reid-Smith, R.J., Deckert, A.E., Checkley, S.L., McEwen, S.A., 2009. Associations between reported on-farm antimicrobial use practices and observed
antimicrobial resistance in generic fecal Escherichia coli isolated from Alberta finishing swine farms. Preventive Veterinary Medicine 88, 185e192. Varga, C., Rajic, A., McFall, M.E., Reid-Smith, R.J., Deckert, A.E., Pearl, D.L., Avery, B.P., Checkley, S.L., McEwen, S.A., 2008b. Comparison of antimicrobial resistance in generic Escherichia coli and Salmonella spp. cultured from identical fecal samples in finishing swine. Canadian Journal of Veterinary Research-Revue Canadienne De Recherche Veterinaire 72, 181e187. Virlogeux-Payant, I., Baucheron, S., Pelet, J., Trotereau, J., Bottreau, E., Velge, P., Cloeckaert, A., 2008. TolC, but not AcrB, is involved in the invasiveness of multidrug-resistant Salmonella enterica serovar Typhimurium by increasing type III secretion system-1 expression. International Journal of Medical Microbiology 298, 561e569. Walczak, J., Xu, S., 2011. Manure as a source of antibiotic resistant Escherichia coli and Enterococci: a case study of a Wisconsin, USA family dairy farm. Water Air and Soil Pollution in Review. doi:10.1007/s11270-010-0729-x. Walker, S.L., Redman, J.A., Elimelech, M., 2005. Influence of growth phase on bacterial deposition: interaction mechanisms in packed-bed column and radial stagnation point flow systems. Environmental Science & Technology 39, 6405e6411. Walsh, C., 2003. Antibiotics: Actions, Origins, Resistance. ASM Press, Washington, D.C. Wax, R.G., 2008. Bacterial Resistance to Antimicrobials, second ed.. CRC Press, Boca Raton. World Health Organization, 2000. Annual Report on Infectious Disease: Overcoming Antimicrobial Resistance. Xu, C.X., Lin, X.M., Ren, H.X., Zhang, Y.L., Wang, S.Y., Peng, X.X., 2006. Analysis of outer membrane proteome of Escherichia coli related to resistance to ampicillin and tetracycline. Proteomics 6, 462e473. Yang, W.R., Moore, I.F., Koteva, K.P., Bareich, D.C., Hughes, D.W., Wright, G.D., 2004. TetX is a flavin-dependent monooxygenase conferring resistance to tetracycline antibiotics. Journal of Biological Chemistry 279, 52346e52352. Yao, K.M., Habibian, M.M., Omelia, C.R., 1971. Water and waste water filtration - concepts and applications. Environmental Science & Technology 5, 1105e1112. Zhang, D.F., Jiang, B., Xiang, Z.M., Wang, S.Y., 2008. Functional characterisation of altered outer membrane proteins for tetracycline resistance in Escherichia coli. International Journal of Antimicrobial Agents 32, 315e319. Zhang, X.Q., Bishop, P.L., Kinkle, B.K., 1999. Comparison of extraction methods for quantifying extracellular polymers in biofilms. Water Science and Technology 39, 211e218.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 9 1 e1 7 0 1
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Impact of dissolved organic matter on colloid transport in the vadose zone: Deterministic approximation of transport deposition coefficients from polymeric coating characteristics Vero´nica L. Morales a, Wei Zhang a, Bin Gao b, Leonard W. Lion c, James J. Bisogni, Jr.c, Brendan A. McDonough a, Tammo S. Steenhuis a,* a
Department of Biological and Environmental Engineering, Riley-Robb Hall, Cornell University, Ithaca, NY 14853-5701, USA Department of Agricultural and Biological Engineering, 285 Frazier Rogers Hall, University of Florida, Gainesville, FL 32611-0570, USA c School of Civil and Environmental Engineering, Hollister Hall, Cornell University, Ithaca, NY 14853-3501, USA b
article info
abstract
Article history:
Although numerous studies have been conducted to discern colloid transport and stability
Received 30 July 2010
processes, the mechanistic understanding of how dissolved organic matter (DOM) affects
Received in revised form
colloid fate in unsaturated soils (i.e., the vadose zone) remains unclear. This study aims to
25 October 2010
bridge the gap between the physicochemical responses of colloid complexes and porous
Accepted 25 October 2010
media interfaces to solution chemistry, and the effect these changes have on colloid
Available online 31 October 2010
transport and fate. Measurements of adsorbed layer thickness, density, and charge of DOM-colloid complexes and transport experiments with tandem internal process visuali-
Keywords:
zation were conducted for key constituents of DOM, humic (HA) and fulvic acids (FA), at
Humic acid
acidic, neutral and basic pH and two CaCl2 concentrations. Polymeric characteristics reveal
Fulvic acid
that, of the two tested DOM constituents, only HA electrosterically stabilizes colloids. This
Steric stabilization
stabilization is highly dependent on solution pH which controls DOM polymer adsorption
Air-water interface
affinity, and on the presence of Caþ2 which promotes charge neutralization and inter-
Hydrophobic expulsion
particle bridging. Transport experiments indicate that HA improved colloid transport significantly, while FA only marginally affected transport despite having a large effect on particle charge. A transport model with deposition and pore-exclusion parameters fit experimental breakthrough curves well. Trends in deposition coefficients are correlated to the changes in colloid surface potential for bare colloids, but must include adsorbed layer thickness and density for sterically stabilized colloids. Additionally, internal process observations with bright field microscopy reveal that, under optimal conditions for retention, experiments with FA or no DOM promoted colloid retention at solid-water interfaces, while experiments with HA enhanced colloid retention at air-water interfaces, presumably due to partitioning of HA at the air-water interface and/or increased hydrophobic characteristics of HA-colloid complexes. ª 2010 Elsevier Ltd. All rights reserved.
* Corresponding author. Tel.: þ1 607 255 2489; fax: þ1 607 255 4080. E-mail addresses:
[email protected] (V.L. Morales),
[email protected] (W. Zhang),
[email protected] (B. Gao),
[email protected] (L.W. Lion),
[email protected] (J.J. Bisogni),
[email protected] (B.A. McDonough),
[email protected] (T.S. Steenhuis). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.10.030
1692
1.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 9 1 e1 7 0 1
Introduction
Dissolved organic matter (DOM) plays a prominent role in many soil processes and is ubiquitous in soils; high concentrations are found in manure or wastewater sludge amended lands. Humic acid (HA) and fulvic acid (FA) are principal constituents of soil, aquatic, sewage sludge, and manure DOM (Schnitzer, 1972; Thurman, 1985). The structural properties of the moieties that make up DOM in soil environments have been explored by several investigators; reporting (as general consensus) that DOM molecules behave as flexible entities that can swell and shrink in response to changes in pH and ionic strength (Avena et al., 1999; Benedetti et al., 1996; Duval et al., 2005; Hosse and Wilkinson, 2001). The amphiphilic character (i.e., presence of hydrophobic and hydrophilic moieties) of DOM has been reported by a number of studies (Guetzloff, 1994; Lenhart and Saiers, 2004; Ma et al., 2007; von Wandruszka, 2000) and used to explain its high surface reactivity and adsorptive fractionation to solid-water and airwater interfaces (Chi and Amy, 2004; Lenhart and Saiers, 2004; Ma et al., 2007). Greater affinity of larger, more hydrophobic DOM components for mineral surfaces and air-water interfaces is of primary importance for contaminant flux through the vadose zone (Lenhart and Saiers, 2004; Meier et al., 1999), as this unsaturated soil region is the critical connection that buffers deep groundwater from surface and shallow contaminants. Physicochemical interactions between DOM and contaminants have received considerable attention in recent years. Numerous investigations have demonstrated that even small amounts of DOM greatly increase the mobility of colloidassociated contaminants (e.g., radionuclide plutonium, americium, thorium, and radium; phosphorus; hydrophobic organic compounds; uranium(IV)/(VI); carbon nanotubes; and lead) (Flury and Qiu, 2008; Granger et al., 2007; Jaisi et al., 2008; Marley et al., 1993; Mibus et al., 2007; Sen and Khilar, 2006; Tang and Weisbrod, 2009) and colloid-sized pathogens (e.g., Escherichia coli, Cryptosporidium parvum oocysts, Giardia, and bacteriophage PRD1) (Abudalo et al., 2005, 2010; Bradford et al., 2006; Foppen et al., 2008) through hydrologic pathways. Laboratory batch kinetic and isotherm experiments have also been conducted to explore the interactions between DOM and colloidal particles. Results from these experiments indicate that DOM increases the stability of colloid and nanoparticle suspensions in the presence of electrolytes through electrostatic and/or steric stabilization by way of adsorption onto colloid surfaces (Akbour et al., 2002; Chen and Elimelech, 2007; Heidmann et al., 2005; Kretzschmar et al., 1998; Pefferkorn, 2006). Moreover, complexation of surface functional groups with non-indifferent ions in solution is a widely recognized process (Amirbahman and Olson, 1995; Chen and Elimelech, 2007; Chen et al., 2006) that could significantly affect the stability and therefore the transport of colloids suspended in DOM rich solutions. A number of investigations systematically examined the impact of DOM on colloid mobility in saturated porous media in terms of pore water velocity and deposition kinetics (Akbour et al., 2002; Jaisi et al., 2008; Kretzschmar et al., 1997). The effects of mono- vs. divalent cation concentrations (Jaisi
et al., 2008), and ionic strength on attachment efficiency have been evaluated (Franchi and O’Melia, 2003; Kretzschmar and Sticher, 1997), as well as charge reversal by organic matter adsorption (Kretzschmar and Sticher, 1997). Both natural and well characterized porous media have been used in research spanning acidic and neutral pore water pH ranges (Akbour et al., 2002; Franchi and O’Melia, 2003; Jaisi et al., 2008; Kretzschmar et al., 1997; Kretzschmar and Sticher, 1997). However, only limited studies have explored the effect of DOM and pH on colloid transport in unsaturated porous media (Tang and Weisbrod, 2009) and the effect of colloid transport in alkaline DOM rich conditions (Harvey et al., 2010). This range of solution chemistry is of high relevance for soils amended with lime-treated manure as a commonly used pathogen inactivation treatment. Two general consensuses about the presence of air phases are that interfaces with air in unsaturated porous media promote colloid retention, and that organic matter of hydrophobic character preferentially fractionates to the air-water interface. Thus, it is of critical importance to understand the effect that DOM has on colloid transport in unsaturated soils; particularly if adsorbed amphiphilic DOM may provide hydrophobic characteristics to the interfaces and colloid surfaces it adsorbs onto. This study aims to experimentally bridge the gap between the physicochemical changes of DOM-colloid complexes and porous media interfaces and the effect that these systematic changes have on the transport of colloids in unsaturated soils. These objectives will be achieved by: (i) directly measuring the changes in surface charge, adsorbed layer thickness and density of organic matter-colloid complexes under acidic, neutral, and basic solution pH in the presence and absence of CaCl2, (ii) assessing with column experiments the effects that HA and FA have on colloid transport at the solution chemistries listed above, (iii) simultaneous internal observation of the dominant pore scale retention sites for each set of conditions, and (iv) mathematical modeling of the transport behavior of organic matter-colloid complexes to relate deposition coefficients with changes in solution composition.
2.
Materials and methods
2.1.
Preparation of materials
In order to meet the objective of discerning the specific steric characteristics that increase the stability of organic mattercolloid complexes (e.g., thickness of adsorbed organic matter layer and uniform adsorption of DOM onto colloid surfaces), the use of uniform and spherical colloids was essential. As such, calibration grade polystyrene and carboxylated spheres of 24 nm diameter (Bangs Laboratories Inc.; Fishers, IN) were used to measure the adsorbed layer thickness. These spheres were selected because of their exceptional size uniformity that allowed the measurement of changes in size at the nanometer scale. Similar surfactant-free, red-dyed, polystyrene and carboxylated spheres of 2.6 mm diameter (Magsphere, Pasadena, CA) were used for all other measurements. The larger red colloids were chosen because they permitted excellent visualization of individual colloids against the porous medium with Bright Field Microscopy (BFM).
1693
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 9 1 e1 7 0 1
Elliott Soil HA and FA standards purchased from the International Humic Substances Society were used for this study. The individual DOM solutions were prepared by dissolving 200 mg L1 of HA and FA in deionized water, and adjusting the stock solution pH to 7 with NaOH. HA and FA stock solutions were further diluted to create solutions of 20 mg of total dissolved organic carbon L1, as measured by persulfate oxidation with an O-I-Analytical Total Organic Carbon Analyzer model 1010 (College Station, TX). Deionized water (DI) was used as the control for no DOM. Solution pH was adjusted to pre-established experimental values (e.g., 4, 6 and 9) with NaOH and HCl immediately prior to starting the experiments. For simplification, the changes in ionic strength (IS) by addition of acid and base to adjust for pH are considered small (e.g., IS of 102 mM for pH 9 and IS of 101 mM for pH 4) relative to the change from addition of CaCl2, so the solution’s IS is referred to as 0 mM and 1 mM for solutions in the absence and presence of CaCl2, respectively. Batch measurements of surface tension of DOM solutions were significantly similar, with a mean surface tension for all tested solution chemistries of 71.7 0.11 mN m1. Translucent quartz sand (Unimin Corp., Vineland, NJ) of 0.4e0.59 mm diameter was used for the porous medium. Before use, the sand was washed according to the procedure in Morales et al. (2009) in order to remove soluble organic compounds from the surface, dissolve metal oxide coatings, and obtain a constant baseline for optical density measurements during breakthrough measurements.
2.2.
Adsorbed layer characteristics
The amount of HA and FA adsorbed onto the sand, Gs (M M1), and colloid surfaces, Gc (M L2), was measured by solution depletion for each solution composition. Briefly, for Gs, the column was wet packed with 60 g of quartz sand, and 60 mL of solution were recycled through the cell with a peristaltic pump for 24 h at a rate of 0.32 mL min1. Afterward, the equilibrium concentration of the solution was measured by spectrophotometry (l ¼ 350 nm) and the difference between the initial and the equilibrium DOM concentration was used to determine the mass of adsorbed organic matter mass per gram of sand medium. For Gc, two sets of 20 mL of each solution composition were prepared and 1x108 colloids were added to the first. The suspensions and colloid-free solutions were agitated for 24 h, then centrifuged for 2 h at 8000 rpm to settle HA-colloid complexes. This centrifugation step was verified to not significantly alter the concentration of nonadsorbed HA and FA components. The supernatant equilibrium concentration was measured by spectrophotometry, and the difference between the colloid-free and the colloid equilibrated concentration was used to determine the mass of adsorbed organic matter per m2 of colloid surface. The thickness of the organic matter layer adsorbed onto the spherical colloidal particles, d (L), was directly measured (rather than fit with existing soft particle theory that is valid only for systems with symmetric and indifferent electrolytes) as the difference between the hydrodynamic radius (RH) of particles exposed to the DOM solution and that of bare particles (i.e., suspended in DI), holding pH and ionic strength constant. For the measurement of this parameter exclusively,
24 nm microspheres were used as the core particles suspended in 18 different solution compositions of varying DOM type (DI, FA, and HA), at three different pHs (4, 6, and 9), and two different ionic strengths (0 and 1 mM). This down scaling of particle size was necessary to measure significant differences in RH at the nanometer scale for colloids with and without a brush layer of adsorbed organic matter. The RH of the colloids was determined by dynamic light scattering (DLS) at each characteristic solution composition using a BIC 200 SM DLS (Long Island, NY). The density of adsorbed DOM layer, F, was estimated with Equation (1) for colloids of radius a (L), using the DOM density value (rDOM) of 1.45x1018 mg nm3 reported by Relan et al. (1984). F¼
2.3.
h
3Gc a2 3
rDOM ðd þ aÞ a3
i
(1)
Electrophoretic mobility
Electrophoretic measurements (EM) measurements were collected with a Malvern Zetasizer nano (Worcestershire, UK) for colloid suspensions in each solution composition. Eight measurements were collected for each treatment and the mean EM values were used to calculate the zeta (z) -potential of the particles via Smoluchowski’s formula (using a measured viscosity value at 25 C of 0.91 104 0.01 104 Pa s and the dielectric constant of the water medium of 78.54). Zeta potentials were converted to surface potential, jo, in mV according to van Oss (2003): z jo ¼ z 1 þ expðkZÞ a
(2)
where z is the distance from the particle’s surface to the slipping plane (here taken to be 0.5 nm), and 1/k is the thickness of the diffuse double layer.
2.4.
Column setup
The column apparatus consisted of a transparent vertical acrylic flow cell (inside dimensions: 10-cm length, 2-cm width, 2-cm depth) with inlet and outlet tubing connected to a twochamber micropurge peristaltic pump (Masterflex, Barnat, Barrington, IL) to induce perfectly matched steady-state flow. The column apparatus is the same as that used by Morales et al. (2009) and is built with a thin acrylic front side (1.5mm thickness) to allow tandem collection of visual data of internal processes during column breakthrough experiments. Colloid transport data were collected in duplicate from 18 column experiments consisting of three background combinations of DOM (DI, 20 mg L1 FA, and 20 mg L1 HA) at three solution pHs (4, 6, 9) and two background ionic strengths. The column was wet packed with clean sand using gentle vibration to ensure uniform packing (porosity of 0.41). Then, the saturated column was conditioned by circulating colloid-free DOM background solution through the column for 24 h to allow the sand medium to equilibrate with adsorbable DOM prior to injection of the colloid pulse. Subsequently, the column was allowed to become unsaturated by increasing the outflow rate above the inflow rate until a volumetric moisture
1694
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 9 1 e1 7 0 1
content, q, of 0.24 0.03 (equivalent to 0.61 0.08 water saturation) was attained (measured gravimetrically). Upon reaching the target moisture content, the outflow and inflow rates were restored to identical and uninterrupted rates of 0.30 mL min1 to ensure steady state flow. The colloid suspension was prepared in a solution of the same composition as that of the background. It is important to note that the optical density of dissolved HA and FA was accounted for when establishing the standard curves for colloid concentration measurements with spectrophotometry. To inject a 10 mL colloid pulse the column influent was switched to the colloid suspension, after which the inflow was switched back to the colloid-free background solution to flush un-retained colloids out of the medium. The BTCs were constructed from effluent colloid concentrations assayed by spectrophotometry (l ¼ 350 nm) with a linear correlation for absorbance and colloid concentration over the concentration range tested. For mass balance comparisons, the fraction of colloids recovered (MR) from the pulse of injected colloids was calculated as: P MR ¼
JDti ðCi þ Ciþ1 Þ 2tc JCo
Transport model
A deterministic model including terms for first-order attachment kinetics and colloid-excluded volume (i.e., colloid accessible pore space responsible for early elution of colloids with respect to the conservative tracer) in homogeneous soil was utilized to determine each solution composition’s effect on colloid deposition and pore-exclusion with the following governing equation: vðqc CÞ v vC ¼ qc D JC qc kd C vt vx vx
2.6.
Internal process visualization
A horizontally mounted bright field microscope was used to visualize colloid transport inside the column from a lateral view of the cell in a fashion similar to that used by Morales et al. (2009). Briefly, observation of colloid transport and retention at the pore scale was achieved by selecting random pores at heights of 9 and 5 cm from the bottom of the column, and capturing still images at 250 magnification (resolution 0.8 mm pixel1). Wherever the pores remained in clear view for the duration of the experiment video recordings were also collected to observe the progressive retention of colloids. Supplementary video related to this article can be found at doi:10.1016/j.watres.2010.10.030.
3.
Results
3.1.
Adsorbed layer characteristics
(3)
where J is the steady state flux (L3 T1), Dti is the time difference between collected effluent samples (T), Ci is the aqueous colloid concentration of the ith effluent sample (M L3), tc is the duration of the colloid or tracer pulse (T), and Co is the initial concentration of the injected pulse (M L3). In addition, colloid BTCs were compared to an independently run nonsorbing bromide tracer (Br), analyzed with a Dionex Ion Chromatography System-2000 (Sunnyvale, California) to determine the pore-water velocity, v (L T1) and dispersion coefficient, D (L2 T1).
2.5.
complexes and polymer characteristics (as discussed in section 3.4).
(4)
where kd is the colloid deposition coefficient (T1), qc is the colloid accessible volumetric moisture content (L3 L3) (obtained from the product of E and q), E is the pore-exclusion factor (unit less), C is the concentration of the colloids (M L3), x is the distance (L), and J is the specific water flux (L3 T1). The authors acknowledge the complexities of the system and the need to construct a more detailed model to discern the contribution from specific retention mechanisms involved in DOM-colloid complexes transported through unsaturated porous media. Nonetheless, the kd term accounts for the loss of particles as the sum of all participating sinks, acceptably captures the BTC shape, and allows relationships to be interpreted between deposition of sterically stabilized colloid
A summary of the adsorbed layer characteristics determined by solution depletion, DLS, and EM is listed in Table 1 for all the conditions tested in this study. Data on the amount of DOM adsorbed onto surfaces, Gs and Gc, suggest that HA has a superior affinity for solid surface sorption than FA because of its higher molecular weight, which is consistent with the literature (Ko et al., 2005). The amount of HA adsorbed onto colloid surfaces was greatest at pH of 4 and decreased with increasing pH (see Gc values in Table 1), which is in agreement with previous reports (Jada et al., 2006; Lenhart and Saiers, 2004). Moreover, the presence of CaCl2 drastically increased Gc and Gs from 25 to 107%; particularly for adsorption of FA, which was undetectable in the absence of CaCl2. The adsorption of organic matter onto sand generally followed the same trend as that for colloids (see Gs values in Table 1), with variations here attributed to the heterogeneity of the medium. The DLS data indicate that a ‘stabilizing shell’ (Fritz et al., 2002) is developed on the colloids when they are exposed to HA. The adsorbed layer thickness, d, for HA-coated colloids d was intermediate at pH 4 (32.0e24.7 nm), greatest at pH 6 (52.8e55.1 nm), and lowest at pH 9 (1.5e11.9 nm); while d for FA-coated colloids was undetectable for all conditions except for pH 6 (see Table 1). Because the diffusivity of the colloids increased greatly with the addition of salts for suspensions of FA and DI, d could not be quantified accurately in the presence of CaCl2. Consequently, already small d values for these samples (0 < d < 2.9 nm) were assumed to not be affected significantly by electrolyte addition and are reported as the same value as those measured at IS of 0 mM. Generally, d increased with IS for experiments with equal pH levels; particularly for HA suspensions. Moreover, the thickness of d for HA at pH 4 when IS is 0 mM (32.0 nm) and at pH 9 when IS is 1 mM (11.9 nm) were almost three fold different even though the mass of adsorbed HA is nearly the same (34.6 0.7 mg m2 and 33.1 0.5 mg m2, respectively in Table 1). These results suggest that the HA adsorption conformation for pH 9 is in
1695
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 9 1 e1 7 0 1
Table 1 e Characteristics of adsorbed organic layer and transport parameters. Data is organized by pH level, dissolved organic matter type (DOM), and ionic strength by addition of CaCl2 electrolyte (IS). Characteristics of adsorbed layer include: mass of adsorbed organic matter onto colloid (Gc) and sand (Gs) surfaces, thickness of adsorbed organic matter layer (d ), density of adsorbed organic matter layer (F), electrophoretic mobility of colloid suspension (EM), and colloid surface potential (jo). Column transport parameters include the fraction of colloids recovered (MR), pore-exclusion factor (E ), and deposition rate (kd). DOM/IS (/mM) pH 4
pH 6
pH 9
HA/0 HA/1 FA/0 FA/1 DI/0 DI/1 HA/0 HA/1 FA/0 FA/1 DI/0 DI/1 HA/0 HA/1 FA/0 FA/1 DI/0 DI/1
Gca (mg m2) 34.6 43.3 0.0 31.4 0.0 0.0 27.2 38.4 0.0 13.0 0.0 0.0 16.0 33.1 0.0 35.7 0.0 0.0
0.7 0.2 0.4
1 1 2
0.2 0.5 1
Gs (mg g1)
db (nm)
F ()
EMa (mm cm Vs1)
joa (mV)
0.0024 0.0031 0.0 0.0031 0.0 0.0 0.0008 0.0006 0.0 0.0004 0.0 0.0 0.0022 0.0025 0.0 0.004 0.0 0.0
32.0 24.7 0.0 0.0 0.0 0.0 52.8 55.1 2.90 2.90 0.0 0.0 1.50 11.9 0.0 0.0 0.0 0.0
0.73 1.2 0.0 0.0 0.0 0.0 0.34 0.46 0.0 2.5 0.0 0.0 7.2 1.9 0.0 0.0 0.0 0.0
3.43 0.20 1.60 0.24 2.04 0.43 1.49 0.19 0.229 0.083 0.273 0.080 3.58 0.11 1.62 0.14 3.33 0.14 1.57 0.065 1.66 0.076 0.427 0.013 3.95 0.19 1.78 0.16 2.47 0.66 1.73 0.029 3.03 0.031 0.856 0.054
43.9 20.3 26.0 19.0 2.92 3.48 45.7 20.7 42.5 20.0 21.2 5.59 50.5 22.7 31.5 22.1 38.8 11.0
MRa () 60.7 41.4 45.7 5.20 20.2 3.00 85.7 24.9 71.1 5.00 75.6 2.60 48.1 20.6 70.2 4.50 78.9 4.00
0.35 0.29 1.8 1.3 2.8 1.1 0.34 1.3 3.1 0.28 1.3 2.0 3.2 2.1 1.8 1.1 4.2 1.0
E (cm3 cm3)
kd (hr1)
0.87 0.79 0.91 0.86 0.71 0.84 0.84 0.71 0.82 0.77 0.81 0.90 0.82 0.79 0.82 0.88 0.85 1.00
0.26 0.43 0.40 1.5 0.77 1.8 0.11 0.70 0.19 1.5 0.15 1.9 0.40 0.77 0.22 1.6 0.14 1.6
a Measurements collected using 2.6 mm colloids. b Measurement collected using 24 nm colloids.
extended and tight multilayers, while that at pH 4 is in loose folds. Likewise, the d for HA at pH 6 for both IS levels (52.8 nm and 55.1 nm for IS at 0 mM and 1 mM, respectively) was very similar although the mass adsorbed experienced a 40% increase in the presence of CaCl2 (Gc ¼ 27.2 mg m2 and Gc ¼ 38.4 mg m2 for IS at 0 mM and 1 mM, respectively) (see Table 1). This result indicates that the conformation of HA strands becomes more compact and therefore the adsorbed layer more dense with increasing IS. Consistently, the density of adsorbed DOM layer, F, increased with IS, but also varied with pH such that the lightest layers formed in HA solutions at pH 6 (F ¼ 0.34e0.46) and the most dense layers formed in HA at pH 9 (F ¼ 1.9e7.2) (see Table 1). Because F is an encompassing parameter that accounts for the presence of a steric layer on the surface of colloids, only those colloid complexes with F values greater than 0 were classified as being sterically stabilized. Measurements of the particle charge are listed as raw electrophoretic mobility, EM, values along with their corresponding surface potential, jo, value in Table 1. In general, our data indicate that for the conditions tested, colloids suspended in HA solutions have the most negative jo, followed by those suspended in FA, and lastly by the bare suspensions in DI. As expected, addition of CaCl2 screened the charge for all treatments, reflected in less negative EM and jo values at higher IS treatments.
3.2.
Colloid retention sites
Four distinct colloid retention sites were identified by visual observation during the column experiments (Figure S1): (a) straining at grainegrain contact regions, (b) attachment to the
solid-water interface (SWI), (c) straining at the air-water meniscusesolid interface (AWmSI), and (d) the accumulation/ attachment of a continuous layer of enmeshed colloids along the air-water interface (AWI) (i.e., bridge flocculation at the AWI). Here, the distinction between straining and attachment processes is based on the definitions provided by Bradford and Torkzaban (2008). Although the occurrence of colloid retention at the AWmSI was frequently observed in many of the treatments tested here, it appeared to be unrelated to any particular set of solution conditions; especially to the type of DOM in solution. Retention at the SWI and grainegrain straining (i.e., multiple SWIs) (Figure S1a and b) were most common for experiments when the solution chemistry created optimal conditions for retention (i.e., higher ionic strength and/or low pH) in experiments of DI and FA. In contrast, retention at the AWI by bridge flocculation (Figure S1d) under similar conditions was the dominant retention site for experiments containing HA as is evident in Video 1(http://soilandwater.bee.cornell.edu/colloids.html) at solution pH of 4 and IS of 1 mM. Here, red colloids exhibit ripening-like behavior (defined as an increasing rate of colloid attachment with time due to colloidecolloid interactions on the collector surface (Bradford and Torkzaban, 2008)) at the AWI and become immobilized at close proximity to other previously retained colloids. It is evident that retained colloids are enmeshed in organic matter partitioned at the AWI, adsorbed on the surface of individual colloids, or both, given that the colloids in the aggregates are not in direct contact with each other but move as a floc unit with the pulsing flow. This behavior was consistently observed for experiments in HA, with denser flocs formed as pH and ionic strength levels increased (Fig. 1).
1696
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 9 1 e1 7 0 1
Fig. 1 e Colloid retention in bridge flocs when suspended in solutions of dissolved Humic Acid at different ionic strengths (IS) and pH levels. Left column 0 mM IS, right column 1 mM IS by CaCl2 addition. Top row pH 4, middle row pH 6, bottom row pH 9. Scale bar length is 250 mm.
3.3.
Colloid transport in unsaturated sand
The various solution treatments used to conduct the column studies are presented and compared in terms of DOM type, changes in solution pH, and differences in CaCl2 concentration.
3.3.1.
Effect of HA and FA
The type of DOM in the solution of the column studies significantly affected the amount of colloids retained in the unsaturated quartz sand medium, as is evident in the HA (diamonds) vs. FA (triangles) vs. DI (circles) breakthrough curves in Fig. 2aec. In all but one solution (pH 9 and IS of
0 mM), experiments with HA displayed significantly greater elution (i.e., colloid mass recovery) (see MR values in Table 1) from the column than those treated with FA or DI with equal pH and ionic strength conditions. Fig. 2aec illustrate that experiments with FA behaved very similar to those treated with DI in all but one case (pH 4 and IS of 0 mM, open triangles in Fig. 2a), in which colloid mobility was enhanced, although not as much in the presence of HA.
3.3.2.
Effect of pH
The transport of colloids in the sandy medium was considerably different at solution pH values of 4, 6 and 9 (compare like symbols from Fig. 2aec). For treatments of both DI and FA
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 9 1 e1 7 0 1
1697
retention at pH 9 (48%) (see Table 1). When the solution IS was raised to 1 mM, increases in pH resulted in a decrease in MR from 41 to 25 to 21% for pH changes from 4 to 6 to 9, respectively. The reduced mobility of HA-colloid complexes at high pH can be explained by the lack of steric hindrance evident in the thinnest adsorbed layer (smallest d values) and most densely packed adsorbed organic matter (largest F values from Table 1).
3.3.3.
Effect of CaCl2 concentration
Increases in ionic strength through addition of CaCl2 to the column study solution produced the expected reduction in colloid breakthrough (open vs. solid symbols in Fig. 2aec). For example, the MR of DI at lower IS ranged from 20 to 79%, but was reduced to less than 4% in the presence of CaCl2 at IS of 1 mM (see MR values of DI in Table 1). The effect of CaCl2 addition was similar in magnitude for experiments with FA, where the MR at lower IS ranged from 46 to 71%, and was reduced to less than 5% in the presence of CaCl2 at IS of 1 mM (see MR values for FA in Table 1). While the addition of CaCl2 to the system drastically reduced the mobility of colloids through the porous medium for all DOM types, the presence of dissolved HA resisted the destabilizing effect of CaCl2, and permitted colloids to be eluted from the column in otherwise optimal conditions for retention. The addition of HA increased the amount of recovered colloids 5e13 fold over that observed in the same conditions but without the DOM addition (i.e., DI) (see MR values in Table 1).
3.4.
Fig. 2 e Breakthrough curves of column experiments and respective model fits for solutions containing dissolved HA, FA, and DI under ionic strength levels of 0 mM and 1 mM at: a) pH 4, b) pH 6, and c) pH 9.
at lower IS, colloid transport varied directly with pH as the MR of colloids increased from 46e20% at pH 4, 71e76% at pH 6, and 70e79% at pH 9 (MR values for DI and FA in Table 1). For experiments where IS was raised to 1 mM, changes in colloid elution with solution pH for DI and FA treatments were insignificant, as the fraction of colloids recovered was low (solid circular and triangular symbols in Fig. 2aec), with MR values ranging from 3 to 5% (see Table 1). Conversely, experiments with HA exhibited distinct and significant changes in colloid mobility for the three pH values tested. For HA experiments conducted at lower IS, MR was 61% at pH 4, reached a maximum at pH 6 (86%), and experienced greatest
Mathematical model
The transport model was run in inverse mode to quantify the influence of solution composition on colloid transport and retention. Values for the colloid pore-exclusion, E, and deposition rate, kd, were estimated from the BTC data using a leastsquares algorithm and the best fit values corresponding to each solution composition presented in Table 1. Although the authors recognize that retention is likely due to more than one mechanism, discerning quantitatively the various components of retention from the breakthrough data is not feasible. We therefore approach retention with the nonspecific parameter, kd, and shed light to the retention mechanism it represents with visual data. D and v were determined from separate conservative tracer experiments where the terms for E and kd equaled 0. Model fits are presented as solid and dashed lines in Fig. 2 with R2 values > 0.9 for experiments where MR was above 5%. This good fit shows that the simple transport model employed is capable of describing the conditions here tested. The average fitted value E was 0.83 0.07 (see Table 1), and accounted for the residence time reduction of 17% of colloids from that of the Br pulse. The values for kd varied inversely with MR. kd increased with IS, decreased with increasing pH for experiments in FA and DI, and was generally much lower for experiments in HA than for other DOM solutions under the same pH and IS conditions. A positive linear correlation was found between kd for nonsterically stabilized suspensions (i.e., suspensions with measured F ¼ 0) and their respective jo (see Fig. 3a) with a Pearson’s correlation coefficient and significance level (from two-tail test) of r ¼ 0.7 and P ¼ 0.01, respectively. As shown in
1698
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 9 1 e1 7 0 1
Fig. 3 e Correlation between deposition rate, kd, of colloids traveling through an unsaturated porous medium and the particle’s surface characteristics. a) Surface characteristic for non-sterically stabilized colloids is surface potential, jo. Data included are for experiments in FA (triangles) and DI (circles) at ionic strengths of 0 mM (open symbols) and 1 mM (solid symbols). b) Surface characteristics for sterically stabilized colloids are represented by a value that includes ratio of adsorbed layer thickness to colloid radius, d/a, density of polymeric layer, F, and surface potential, jo. Data included are for experiments in HA (diamonds) and FA (triangles) at ionic strengths of 0 mM (open symbols) and 1 mM (solid symbols).
Fig. 3b, sterically stabilized suspensions (i.e., suspensions with measured F > 0) have a Pearson’s correlation coefficient and significance level of r ¼ 0.9 and P < 0.01 with the product of the measured polymeric characteristics of sterically stabilized suspensions d/a, F, and jo.
4.
Discussion
Although in general, the partitioning of DOM-colloid complexes at the AWI generally increases with the presence of HA, the overall breakthrough is still greater when HA is present than when it is not. The physicochemical characteristics of the colloids analyzed in this research indicate that, although the charge of colloids suspended in DOM was more negative than those suspended in DI (i.e., absence of DOM)
(see EM and jo values in Table 1), only HA had the capacity to electrosterically stabilize colloids. This was a result of the formation of a highly charged brush layer (of measured polymeric characteristics d, Gc, and F > 0) on the colloid’s surface that promoted colloid mobility in solutions with higher ionic strength. FA adsorption affinity onto surfaces was in most cases minor (see Gc values for FA in Table 1) and thus did not significantly improve the mobility of colloids despite its effect on colloid charge (see changes in EM and jo for FA in Table 1). As has been recognized previously, the single effect of increasing a particle’s negative charge is not sufficient to justify the increased colloid stability observed in certain treatments (Elimelech et al., 2000). In this study disparate percentages of mass recovered colloids were observed for solution compositions that yielded comparable colloid surface charges. Hence, an additional physicochemical characteristic (to electrostatic charge) responsible for improving the suspension’s stability was suspected to play a prominent role in enhancing the transport of DOM-colloid complexes. Various studies have suggested that the development of steric surface structures by adsorbed DOM may be implicated in the unsubstantiated repulsive forces of colloid suspensions, as these have been observed to remain stable even under conditions of high ionic strength (Chen and Elimelech, 2007, 2008; Franchi and O’Melia, 2003; Phenrat et al., 2010). The adsorbed layer characteristics data demonstrate that the magnitude and range of steric repulsion depend on three factors: (i) density of adsorbed DOM layer, F, (ii) the extension of the adsorbed layer, d, and (iii) the particle charge here presented as surface potential, jo. Evidently, the eletrosterically stabilized particles with most negative surface potential, large adsorbed mass, and thickest brush layer experience the best transport enhancement, and the presence of all three polymeric characteristics is required to provide a suspension with steric stability. When either d is absent or F is very large, the colloids behave like hard colloidal particles and are sensitive to aggregation with changes in solution chemistry. These results reveal that structural hindrance, in addition to electrostatic repulsion, is the mechanisms by which HA, and in one extreme case FA, increases the stability of colloids in the bulk fluid. Clearly, the development of the organic matter brush layer grants soft particle functionalities to the otherwise chemically sensitive suspension, which, to the authors’ knowledge, has only been visualized by Chen and Elimelech (2007) and physically characterized with indirect measurements by Phenrat et al. (2010). This information explains the reduced particleeparticle interaction that results in greater mobility of HA-colloid complexes through porous media, and can be assumed to be analogous to the interactions between the particles and the porous medium through which they move. Internal observations of colloid transport demonstrate that for experiments in DI and FA with solution conditions conducive for high retention (e.g., higher ionic strength and/or low pH), colloid retention occurs at sites involving single or multiple SWI interfaces (Figure S1aed) and have high deposition rate values ranging from 0.4 to 2 h1 (see kd for FA and DI at IS of 1 mM and both IS values at pH 4 in Table 1). This relation suggests that retention of hard particles, such as the colloid suspensions in DI and FA, experience retention with
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 9 1 e1 7 0 1
higher mass transfer rates toward the SWI when the medium’s solution composition hinders transport. For experiments in HA with solution conditions that favor attachment, the dominant retention sites were along the AWI in bridge flocs (Fig. 1), which were characterized with lower deposition rates ranging from 0.3 to 0.8 h1 (see kd for HA at ionic strength solutions of 1 mM and both ionic strength solutions at pH 4 in Table 1). This relation suggests that irreversible retention of soft particles, like HA-colloid complexes, at the AWI is slower than the retention of hard particles, like FA-colloid complexes or colloids in the absence of DOM, at the SWI. Moreover, colloids retained at the AWI may be more susceptible for being remobilized with transient flow conditions than the colloids predominantly retained at the SWI. The dominant retention at AWIs suggests that: (i) HA partitioning to the AWI (as has been demonstrated previously (Lenhart and Saiers, 2004; Ma et al., 2007)) creates a rough surface or gelatinous layer that enhances colloid retention; (ii) HAcolloid complexation occurs in the bulk solution, where the amphiphilic character of HA augments the hydrophobic properties of the colloids resulting in hydrophobic expulsion toward the AWI (i.e., strong attractive non-electrostatic interactions); or (iii) both processes are occurring. Although the visual data in Fig. 1 and Video 1 suggest that hydrophobic interactions significantly influence retention mechanisms in HA-colloid systems, these types of interactions do not occur in isolation from other interaction effects, warranting future work on hydrophobic alterations of surfaces by DOM. Changes in pH affect colloid transport for suspensions in DI and FA by altering the level of functional group deprotonation on the surfaces of both colloids and the porous medium, which consequently affects the thickness of their respective electric double layers and resulting electrostatic interactions. For conditions where the system’s IS was raised to 1 mM, it is evident that neutralization of the colloid’s charge (from less negative jo values in Table 1) dominated over any effect that pH had for experiments with DI and FA, as the colloid mass recovery (MR in Table 1) did not exceed 5% at any pH value. Alternatively, experiments conducted with HA demonstrate that pH is a critical factor for determining colloid transport, as the thickness of the polymeric layer and adsorption affinity of HA onto colloid and medium surfaces was dictated by the solution’s pH (see d and Gc values for HA in Table 1). The addition of CaCl2 affected the colloidal system by: (i) neutralizing the charge of deprotonated HA functional groups, (ii) forming multidentate complexes between available functional groups of the HA macromolecules and Ca cations (as demonstrated by the increase in Gc with CaCl2 addition), and (iii) compressing the electric double layer of all charged surfaces (as is evident from jo data in Table 1). Although chemical bonding of HA with Ca2þ promoted greater adsorption of organic matter onto the colloid surfaces, the chemical reaction at the Ca2þ and HA concentration tested weakened the ability of HA to mobilize colloids (observed in decreases in MR with increasing IS in Table 1) by neutralizing charged functional groups of adsorbed HA strands (apparent in the less negative jo with increasing ionic strength in Table 1), and by bridging together HA-colloid complexes forming large flocs that can experience straining in the porous medium (apparent in visual data of internal colloid processes in Fig. 1, Figure S1, and Video 1). The effects of charge
1699
neutralization and EDL compression are well known to stimulate colloid retention by reducing electrostatic energy barriers that otherwise prevent colloid aggregation and colloid attachment to immobile sites within the porous medium (Kretzschmar and Sticher, 1997). The environmental relevance of this research is most apparent in the solution compositions selected to represent natural subsurface environments rich in DOM as well as agricultural settings that practice land application of alkaline-treated manure. This waste management practice is often preferred for its practicality at the farm level to inactivate zoonotic pathogens (Gerba and Smith, 2005) and control odor (Zhu, 2000). The findings of this study advance our current understanding of the governing mechanisms of colloid transport in a broad range of organic rich subsurface environments; particularly those associated with agriculture. Moreover, our current ability to predict the transport and fate of colloid-adsorbed contaminants and pathogenic microorganisms in the natural subsurface is limited by the incomplete understanding of the physicochemical processes responsible for abiotic porous media filtration. Thus, the use of microspheres as colloidal pathogen surrogates makes elucidation of porous-media transport processes possible without the complications of even less understood biotic processes (e.g., biologically induced attachment, inactivation, die off, growth, motility, chemotaxis, bioclogging, etc.).
5.
Conclusion
In summary, the effect of DOM on colloid characteristics can be directly measured through changes in surface potential, adsorbed layer thickness, and mass of adsorbed organic matter. Solution pH and CaCl2 presence strongly affected the polymeric characteristics by varying the adsorption affinity of DOM onto surfaces and protonation of functional groups, and by neutralizing surface charge and chemically bridging DOM strands together, respectively. The presence of all three parameters (surface potential, adsorbed layer thickness, and mass of adsorbed organic matter by way of density of adsorbed DOM) was established to be a requirement for particles to be endowed with electrostatic and structural stability. Failure to possess all three polymeric characteristics resulted in one of three scenarios: (i) weakened electrostatic repulsion if surface potential were neutralized, (ii) lack of steric hindrance if adsorbed layer thickness were smaller than the separation distance where van der Waals forces are in effect, or (iii) polymerecolloid complexes approach hard particles if adsorbed layer densities were too high. Thus, of the two types of DOM examined, it was established that only HA can consistently electrosterically stabilize colloids. HA was determined to enhance colloid mobility better than FA, even under the most chemically conducive conditions for retention. Retention of HA-colloid complexes along the AWI indicates that partitioning of DOM at the AWI may create a rough or gelatinous surface for colloids to become immobilized and/or that the amphiphilic character of adsorbed HA may encourage colloidal hydrophobic expulsion. Mathematical simulations of column transport data indicate that a simple convective dispersive model with a deposition term and colloid pore-exclusion correction is suitable to describe the fate and transport of
1700
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 9 1 e1 7 0 1
organic matter-colloid complexes in the conditions here tested. A positive and significant correlation was found between deposition coefficients of electrosterically stabilized suspensions and the product of three polymeric characteristics: adsorbed layer thickness, density, and surface potential. As such, measurements of these specific polymeric characteristics can effectively be used to improve deterministic predictions of colloid transport in a wide range of DOM rich solution compositions and assess the filtering function that the vadose zone serves to protect groundwater resources.
Acknowledgments This study was financed by the National Science Foundation, Project No. 0635954; the Binational Agricultural Research and Development Fund (BARD), Project No. IS-3962-07; and the Teresa Heinz Foundation for Environmental Research. The authors thank Dr. John F. McCarthy for helpful discussions and Dr. Yuanming Zhang and Dr. Claude Cohen for assistance with light scattering techniques. We also thank Mr. Doug Caveney for constructing the flow cell used for this study.
Appendix. Supplementary Data Supplementary data related to this article can be found online at doi:10.1016/j.watres.2010.10.030.
references
Abudalo, R.A., Bogatsu, Y.G., Ryan, J.N., Harvey, R.W., Metge, D.W., Elimelech, M., 2005. Effect of ferric oxyhydroxide grain coatings on the transport of bacteriophage PRD1 and Cryptosporidium parvum oocysts in saturated porous media. Environmental Science & Technology 39, 6412e6419. Abudalo, R.A., Ryan, J.N., Harvey, R.W., Metge, D.W., Landkamer, L., 2010. Influence of organic matter on the transport of Cryptosporidium parvum oocysts in a ferric oxyhydroxide-coated quartz sand saturated porous medium. Water Research 44, 1104e1113. Akbour, R.A., Douch, J., Hamdani, M., Schmitz, P., 2002. Transport of kaolinite colloids through quartz sand: influence of humic acid, Ca2þ, and trace metals. Journal of Colloid and Interface Science 253, 1e8. Amirbahman, A., Olson, T.M., 1995. Deposition kinetics of humic matter-coated hematite in porous-media in the presence of Ca2þ. Colloids and Surfaces A-Physicochemical and Engineering Aspects 99, 1e10. Avena, M.J., Vermeer, A.W.P., Koopal, L.K., 1999. Volume and structure of humic acids studied by viscometry pH and electrolyte concentration effects. Colloids and Surfaces A-Physicochemical and Engineering Aspects 151, 213e224. Benedetti, M.F., vanRiemsdik, W.H., Koopal, L.K., 1996. Humic substances considered as a heterogeneous donnan gel phase. Environmental Science & Technology 30, 1805e1813. Bradford, S.A., Tadassa, Y.E., Pachepsky, Y., 2006. Transport of Giardia and manure suspensions in saturated porous media. Journal of Environmental Quality 35, 749e757. Bradford, S.A., Torkzaban, S., 2008. Colloid transport and retention in unsaturated porous media: a review of interface-,
collector-, and pore-scale processes and models. Vadose Zone Journal 7, 667e681. Chen, K.L., Elimelech, M., 2007. Influence of humic acid on the aggregation kinetics of fullerene (C-60) nanoparticles in monovalent and divalent electrolyte solutions. Journal of Colloid and Interface Science 309, 126e134. Chen, K.L., Elimelech, M., 2008. Interaction of fullerene (C-60) nanoparticles with humic acid and alginate coated silica surfaces: measurements, mechanisms, and environmental implications. Environmental Science & Technology 42, 7607e7614. Chen, K.L., Mylon, S.E., Elimelech, M., 2006. Aggregation kinetics of alginate-coated hematite nanoparticles in monovalent and divalent electrolytes. Environmental Science & Technology 40, 1516e1523. Chi, F.H., Amy, G.L., 2004. Kinetic study on the sorption of dissolved natural organic matter onto different aquifer materials: the effects of hydrophobicity and functional groups. Journal of Colloid and Interface Science 274, 380e391. Duval, J.F.L., Wilkinson, K.J., Van Leeuwen, H.P., Buffle, J., 2005. Humic substances are soft and permeable: Evidence from their electrophoretic mobilities. Environmental Science & Technology 39, 6435e6445. Elimelech, M., Nagai, M., Ko, C.H., Ryan, J.N., 2000. Relative insignificance of mineral grain zeta potential to colloid transport in geochemically heterogeneous porous media. Environmental Science & Technology 34, 2143e2148. Flury, M., Qiu, H.X., 2008. Modeling colloid-facilitated contaminant transport in the vadose zone. Vadose Zone Journal 7, 682e697. Foppen, J.W., Liem, Y., Schijven, J., 2008. Effect of humic acid on the attachment of Escherichia coli in columns of goethitecoated sand. Water Research 42, 211e219. Franchi, A., O’Melia, C.R., 2003. Effects of natural organic matter and solution chemistry on the deposition and reentrainment of colloids in porous media. Environmental Science & Technology 37, 1122e1129. Fritz, G., Schdler, V., Willenbacher, N., Wagner, N.J., 2002. Electrosteric stabilization of colloidal dispersions. Langmuir 18, 6381e6390. Gerba, C.P., Smith, J.E., 2005. Sources of pathogenic microorganisms and their fate during land application of wastes. Journal of Environmental Quality 34, 42e48. Granger, S.J., Bol, R., Butler, P., Haygarth, P.M., Naden, P., Old, G., Owens, P.N., Smith, B.P.G., 2007. Processes affecting transfer of sediment and colloids, with associated phosphorus, from intensively farmed grasslands: tracing sediment and organic matter. Hydrological Processes 21, 417e422. Guetzloff, T.F., 1994. Does Humic Acid Form a Micelle?, vol. 152 31e35. Harvey, R.W., Metge, D.W., Barber, L.B., Aiken, G.R., 2010. Effects of altered groundwater chemistry upon the pH-dependency and magnitude of bacterial attachment during transport within an organically contaminated sandy aquifer. Water Research 44, 1062e1071. Heidmann, I., Christl, I., Kretzschmar, R., 2005. Aggregation kinetics of kaolinite-fulvic acid colloids as affected by the sorption of Cu and Pb. Environmental Science & Technology 39, 807e813. Hosse, M., Wilkinson, K.J., 2001. Determination of electrophoretic mobilities and hydrodynamic radii of three humic substances as a function of pH and ionic strength. Environmental Science & Technology 35, 4301e4306. Jada, A., Akbour, R.A., Douch, J., 2006. Surface charge and adsorption from water onto quartz sand of humic acid. Chemosphere 64, 1287e1295. Jaisi, D.P., Saleh, N.B., Blake, R.E., Elimelech, M., 2008. Transport of single-walled carbon nanotubes in porous media: filtration
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 6 9 1 e1 7 0 1
mechanisms and reversibility. Environmental Science & Technology 42, 8317e8323. Ko, I., Kim, J.Y., Kim, K.W., 2005. Adsorption properties of soil humic and fulvic acids by hematite. Chemical Speciation and Bioavailability 17, 41e48. Kretzschmar, R., Barmettler, K., Grolimund, D., Yan, Y.D., Borkovec, M., Sticher, H., 1997. Experimental determination of colloid deposition rates and collision efficiencies in natural porous media. Water Resources Research 33, 1129e1137. Kretzschmar, R., Holthoff, H., Sticher, H., 1998. Influence of pH and humic acid on coagulation kinetics of kaolinite: a dynamic light scattering study. Journal of Colloid and Interface Science 202, 95e103. Kretzschmar, R., Sticher, H., 1997. Transport of humic-coated iron oxide colloids in a sandy soil: influence of Ca2þ and trace metals. Environmental Science & Technology 31, 3497e3504. Lenhart, J.J., Saiers, J.E., 2004. Adsorption of natural organic matter to air-water interfaces during transport through unsaturated porous media. Environmental Science & Technology 38, 120e126. Ma, J., Jiang, J., Pang, S., Guo, J., 2007. Adsorptive fractionation of humic acid at air-water interfaces. Environmental Science & Technology 41, 4959e4964. Marley, N.A., Gaffney, J.S., Orlandini, K.A., Cunningham, M.M., 1993. Evidence for radionuclide transport and mobilization in a shallow, sandy aquifer. Environmental Science & Technology 27, 2456e2461. Meier, M., Namjesnik-Dejanovic, K., Maurice, P.A., Chin, Y.P., Aiken, G.R., 1999. Fractionation of aquatic natural organic matter upon sorption to goethite and kaolinite. Chemical Geology 157, 275e284. Mibus, J., Sachs, S., Pfingsten, W., Nebelung, C., Bernhard, G., 2007. Migration of uranium (IV)/(VI)in the presence of humic
1701
acids in quartz sand: a laboratory column study. Journal of Contaminant Hydrology 89, 199e217. Morales, V.L., Gao, B., Steenhuis, T.S., 2009. Grain surfaceroughness effects on colloidal retention in the vadose zone. Vadose Zone Journal 8, 11e20. Pefferkorn, E., 2006. Clay and oxide destabilization induced by mixed alum/macromolecular flocculation aids. Advances in Colloid and Interface Science 120, 33e45. Phenrat, T., Song, J.E., Cisneros, C.M., Schoenfelder, D.P., Tilton, R.D., Lowry, G.V., 2010. Estimating Attachment of Nano- and Submicrometer-particles Coated with Organic macromolecules in porous media: development of an empirical model. Environmental Science & Technology 44, 4531e4538. Relan, P.S., Girdhar, K.K., Khanna, S.S., 1984. Molecularconfiguration of composts humic-acid by viscometric studies. Plant and Soil 81, 203e208. Schnitzer, M., 1972. Humic Substances in the Environment. Dekker, New York. Sen, T.K., Khilar, K.C., 2006. Review on subsurface colloids and colloid-associated contaminant transport in saturated porous media. Advances in Colloid and Interface Science 119, 71e96. Tang, X.Y., Weisbrod, N., 2009. Colloid-facilitated transport of lead in natural discrete fractures. Environmental Pollution 157, 2266e2274. Thurman, E.M., 1985. Organic Geochemistry of Natural Waters. Kluwer Academic, Boston. van Oss, C.J., 2003. Long-range and short-range mechanisms of hydrophobic attraction and hydrophilic repulsion in specific and aspecific interactions. Journal of Molecular Recognition 16, 177e190. von Wandruszka, R., 2000. Humic Acids: Their Detergent Qualities And Potential Uses In Pollution Remediation, vol. 1 10e15. Zhu, J., 2000. A review of microbiology in swine manure odor control. Agriculture Ecosystems & Environment 78, 93e106.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 0 2 e1 7 1 0
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Short-term bacterial community composition dynamics in response to accumulation and breakdown of Microcystis blooms Huabing Li a,b, Peng Xing a, Meijun Chen a, Yuanqi Bian a,b, Qinglong L. Wu a,* a
State Key Laboratory of Lake Science and Environment, Nanjing Institute of Geography & Limnology, Chinese Academy of Sciences, Nanjing 210008, PR China b Graduate School of Chinese Academy of Sciences, Beijing 100049, PR China
article info
abstract
Article history:
Short-term bacterial community composition (BCC) dynamics in response to accumulation
Received 12 May 2010
and breakdown of Microcystis blooms were examined by conducting in situ mesocosm
Received in revised form
experiments with varying levels of Microcystis sp. biomass, ranging from 15 to 3217 mg/L as
27 October 2010
measured by chlorophyll-a concentration in the freshwater water column. The BCC was
Accepted 9 November 2010
assessed by means of terminal restriction fragment length polymorphism (T-RFLP) of 16S
Available online 16 November 2010
ribosomal RNA genes followed by cloning and sequencing of selected samples. The results showed that the composition of both free-living and particle-attached bacterial commu-
Keywords:
nities changed during the accumulation and breakdown phases of a Microcystis bloom, and
Microcystis
differences were also evident with different levels of Microcystis biomass. The relative
Bacterial community composition
abundance of bacteria affiliated with Micrococcineae and Legionellales increased in general
Lake
after amendment with Microcystis. Significant correlation between the relative abundance of Micrococcineae and breakdown of Microcystis biomass was also observed. Canonical correspondence analysis (CCA) showed that the changes in the free-living and particleattached bacterial community were mostly related to the changes in the concentrations of chlorophyll-a, dissolved organic carbon (DOC), dissolved oxygen (DO) and pH, which were mainly induced by the breakdown of Microcystis biomass. Overall, our study revealed the following: i) accumulation of Microcystis blooms and their breakdown have strong impacts on bacterial community composition; ii) there might be saprophytic association between Micrococcineae and decomposition of Microcystis biomass; iii) it is necessary to reveal potential associations between Legionellales organisms and Microcystis blooms in eutrophic freshwater lakes. ª 2010 Elsevier Ltd. All rights reserved.
1.
Introduction
Due to pollution and eutrophication, cyanobacterial blooms, especially Microcystis blooms, are becoming a widespread problem in the aquatic environment (e.g. Lehman, 2007; Paerl et al., 2001). In the post-bloom stage, large quantities of
Microcystis will accumulate and decompose, producing dissolved organic matter (DOC) (Cole et al., 1982), resulting in drops in pH and dissolved oxygen (DO) (Chen et al., 2010), and these physiochemical changes may have strong impacts on other aquatic organisms (e.g. Zhang et al., 2009). Recent studies indicated a strong influence of Microcystis blooms on
* Corresponding author. Tel.: þ86 25 86882107; fax: þ86 25 57714759. E-mail address:
[email protected] (Q.L. Wu). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.11.011
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 0 2 e1 7 1 0
eukaryotic microorganisms (e.g. Chen et al., 2010), while little is known about their effects on bacterial community composition (BCC) (Eiler and Bertilsson, 2004). Because heterotrophic bacteria have crucial roles in biogeochemical cycling and energy flux (e.g. Azam et al., 1983), understanding the response of the BCC to accumulation and breakdown of Microcystis blooms is very important for a better understanding of the metabolic processes in eutrophic aquatic ecosystems. Meanwhile, many algae are associated with a diverse bacterial community (e.g. Maruyama et al., 2003; Grossart et al., 2005, 2006) including some potential pathogens like Legionellale pneumophila (Tison et al., 1980), and Vibrio cholerae (Ferdous, 2009). Changes of BCC may also decrease the water quality (Paerl et al., 2003; Masango et al., 2008). Thorough understanding of the BCC’s response to Microcystis blooms may provide information necessary for searching bacterial indicator, and help to develop efficient strategies for protecting the aquatic environment and human health (Paerl et al., 2003). Lake Taihu, located in eastern China (30 55 0 40" e 31 32 0 58" N and 119 520 32" e 120 360 10" E), is a large shallow eutrophic lake. The cyanobacterial blooming, most of which were Microcystis, occurred in general from March to November with the highest biomass of Microcystis observed from Jun to September in the northern part of the lake. During the last three decades, the Microcystis blooming has become a major environmental issue for the lake (Qin et al., 2007). Due to its large surface area, the Microcystis blooms drift from place to place and even form heavy scum in some bays of the lake (Qin et al., 2007). Previous investigations hinted that the seasonal dynamics of BCC might be related to Microcystis blooms (Xing and Kong, 2007; Wu et al., 2007a, b). However, this has never been confirmed directly by experimental data, and whether Microcystis blooms may be associated with some pathogens in freshwater lakes is still unclear. Therefore, to understand the response of BCC to the accumulation and breakdown of a Microcystis bloom, we conducted in situ mesocosm experiments in Lake Taihu using different levels of Microcystis biomass that encompass the natural range of Microcystis biomass in Lake Taihu.
2.
Materials and methods
2.1.
Experimental design
In LakeTaihu, an in situ mesocosm experiment with 12 polyethylene enclosures (50 L each) was carried out between 26 August and 5 September in 2008. The enclosures were enclosed with the upside open to air. All enclosures were placed in Meiliang Bay of Lake Taihu, which was about 200 m away from the shoreline. Each of the 12 enclosures was filled with lake water from Gonghu Bay, which is a part of Lake Taihu and characterized by high water transparency, dominant submersed macrophyte communities (mainly Potamogeton sp.) and very low algae biomass consisting primarily of diatoms, green algae and cyanobacteria (chlorophyll-a concentration was about 15 mg/L). Microcystis sp. was collected with a 64-mm pore net from Meiliang Bay (31 28 0 4.5" N e 120 130 34.2" E) during a Microcystis bloom period one day before the experiment, and the samples were rinsed three times with distilled
1703
water. To reflect the natural chlorophyll-a (Chl a) concentrations found in different parts of Lake Taihu in the summer, there were four treatments: A (control) enclosures were not supplemented with Microcystis, B (low biomass) enclosures were supplemented with 113 mg/L Microcystis, C (moderate biomass) enclosures were supplemented with 303 mg/L Microcystis, and D enclosures (very high biomass) were supplemented with 3217 mg/L Microcystis. Every treatment was performed in triplicate. The measurement of physical and chemical parameters and their dynamics (Table S1) have been demonstrated by Chen et al. (2010).
2.2. Bacterial counts, biomass collection, and DNA extraction Physiochemical data were used to select appropriate dates for the BCC analysis. At day 4, significant changes in the chemical parameters occurred: the concentrations of Chl a, DO and pH decreased drastically while the DOC concentration increased strongly after 4 days of incubation in all enclosures (Table S1, for details see Chen et al., 2010). This suggests that Day 4 represented the breakdown period. In order to monitor the short-term dynamic of BCC in response to accumulation and breakdown of Microcystis blooms and to decrease the complexity for analyzing the BCC, we therefore chose the samples from days 0, 1 and 4 for an in-depth analysis of short-term BCC dynamics. Because the physiochemical parameters among enclosures A, B and C were very similar throughout the experiment, we did not analyze the BCC of enclosure C. Bacterial DNA samples were taken at day 0, 1, 4 in enclosures A, B, and D. Within 2 h after sampling, about 100 ml was filtered first through a 5.0-mm polycarbonate filter membrane (47-mm diameter, Millipore) and subsequently a 0.2-mm filter membrane (47-mm diameter, Millipore) at a pressure of <20 mbar to collect particleattached and free-living bacteria, respectively. The filters (Millipore) were stored at 80 C until nucleic acids were extracted. Total nucleic acids from particle-attached and free-living bacteria on the filters were extracted and purified using proteinase K and sodium dodecyl sulfate concomitant with chloroform extraction and isopropanol precipitation. Total bacterial numbers were determined by epifluorescence microscopy after DAPI (40 ,6-diamidino-2-phenylindole) staining (Porter and Feig, 1980) as described previously (Wu et al., 2006).
2.3. Bacterial community analysis by 16S rRNA-based T-RFLP analysis Bacterial 16S rRNA genes were amplified using the universal eubacterial primers 8f (50 - AGAGTTTGATCCTGGCTCAG -30 ; labeled with 6-carboxyfluorescein on the 50 end) and 926r (50 CCGTCAATTCCTTTGAGTTT-30 ) (Liu et al., 1997). Mixed DNA from triplicate extractions were used as PCR templates. The PCR amplification was performed in an automated thermocycler (PTC 200-cycler, MJ Research) as follows: one cycle at 94 C for 3 min, 30 cycles at 94 C for 30 s, 55 C for 30 s, 72 C for 1 min, and a final extension at 72 C for 10 min. Three to four replicate PCR samples were cleaned and concentrated into a single 50-ml aliquot with an E.Z.N.A.TM Cycle-Pure Kit (Omega, USA). Purified and concentrated T-RFLP products were digested, purified, separated and analyzed sequentially
1704
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 0 2 e1 7 1 0
(Chen et al., 2010). To account for small differences in the running time among samples, we considered fragments from different profiles with less than 1 base pair difference to be the same length. The results were then expressed as the relative area compared to the total area. Peaks of less than 60 bp, longer than 600 bp, or representing less than 1% of the total peak area were discarded.
constructed from different enclosures (Schloss et al., 2009). Regression analysis was used to measure the relationship between Chl a concentration and chemical parameters as well as the relative abundance of dominant T-RFs during the decomposition of Microcystis sp. The T-test was used to assess changes in chemical parameters during the experiments, and one-way ANOVA was used to compare chemical parameters among treatments.
2.4. Cloning, sequencing, phylogenetic analysis and assignment of T-RFs
3. To identify bacteria, three clone libraries were generated with mixed bacterial 16S DNA templates in triplicate extractions retrieved from the A enclosures at Day 0 (free-living sample) and D enclosures at day 4 (free-living and particle-attached sample respectively). Bacterial amplifications for sequence analysis were generated from 16S rRNA with the primers 8f(50 -AGAGTTTGATCCTGGCTCAG-30 ) (Lane, 1991) and 926r (50 CCGTCAATTCCTTTGAGTTT-30 ) (Muyzer et al., 1993). The PCR protocols were the same as described above. The PCR products were purified with a QIAquick gel extraction kit (QIAGEN) and ligated into the pGEM-T Easy Vector (Promega) according to the instructions of the manufacturer. The presence of the 16S rRNA gene in positive colonies was checked by PCR amplification using vector primers (M13F and M13R). Positive clones were randomly selected for sequencing, which was carried out by Invitrogen Company (Shanghai, China). Raw sequence data were processed and checked with the Lasergene software package DNASTAR (Madison, USA). All sequences were screened for potential chimeric sequences using CHIMERA_CHECK from RDP (http://rdp8.cme.msu.edu/docs/chimera_ doc.html). Correct sequences were compared to sequences in public databases by using NCBI BLAST (http://www.ncbi. nlm.nih.gov) for an initial phylogenetic affiliation. Phylogenetic analyses of retrieved 16S rRNA sequences were conducted by a neighbor-joining tree using MEGA 4 (Tamura et al., 2007). To test the phylogenetic assignments based on in silico T-RF analysis, randomly selected clones were analyzed by in vitro T-RF by finding the first Hha I enzymatic digestion site downstream from 8f. The obtained 16S rRNA gene sequences were deposited in GenBank under the accession numbers HM153606eHM153700.
3.1. Impact of Microcystis addition on bacterial abundance At Day 4, significant increase in abundance of heterotrophic bacteria was observed in Enclosure D (Fig. 1) with addition of high Microcystis biomass, while no strong variations were found in Enclosure A and B in which no or less Microcystis biomass had been added.
3.2.
The T-RFLP profiles of free-living bacterial DNA extracted from the no-addition enclosures (A enclosures) showed similar patterns after 0, 1 and 4 d of incubation (Fig. 2). In Microcystisamended enclosures, by contrast, the free-living bacterial TRFLP profiles changed after 1 and 4 d (Fig. 2). Immediately before the start of the incubation (day 0), a total of 24 distinct bacterial T-RFs were identified. These T-RFs all showed a relative abundance of >1%, and those of 65, 204, 205, 208, 369, 569, and 577 base pairs (bp) in length even exhibited a relative abundance of >5%. After 1 d, in the medium-addition (B) enclosures, the relative abundance of most of these dominant T-RFs had changed very little, so that these T-RFs were still among the most abundant ones within the bacterial T-RFLP profile. However, a new T-RF at 85/86 bp appeared and became the most abundant one (relative abundance >10%; Fig. 2), and other characteristic T-RFs of 60, 92, 513, 93, 201, 203, 365 and 565 bp (the relative abundance of the former three were all >4%) were detected for the first time. At day 4, a 202-bp and the 45
106 cells/ml)
Statistical analysis
To reveal the relationships between BBC and environmental parameters, canonical correspondence analysis (CCA) was used because the length of the first DCA (detrended correspondence analysis) axis run on species data was >2. The tested environmental variables were as follows: Chl a, TP, TN, COD, NH4eN, NOx-N, PO4eP and DO. All data were log (xþ1) transformed except for pH. In the case of the T-RFLP results, a binary matrix was constructed by scoring presence (percentage of area) and absence (0) of particular T-RFs. Environmental factors best describing the most influential gradients in community composition were identified by forward selection with 499 unrestricted Monte Carlo permutations. CCA was performed with the software CANOCO 4.5 (ter Braak and Smilauer, 2002). The LIBSHUFF program was used to statistically assess the differences between clone libraries
T-RFLP analysis of the free-living bacterial community
40
Bacterial abundance (
2.5.
Results
25
35
Day0
Day1
Day4
30
20 15 10 5 0 A
B
D
Fig. 1 e Heterotrophic bacterial abundance in different enclosures at different day.
1705
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 0 2 e1 7 1 0
A
B
D
Day 0 Day 1 Day 4
Day 1 Day 4
Day 1 Day 4
100
90
80
Relative abundance (%)
70
60
50
40
30
20
10
0
Fig. 2 e Relative abundance of free-living bacterial 16S rRNA amplicons recovered from enclosures with different biomass of Microcystis sp. Those T-RFs of less than 1% of the total were grouped together as a single group donated by others <1%.
A
B
D
100
T-R F(bp)
90 80
Relative abundance (%)
565-bp T-RFs dominated the free-living bacterial populations in these enclosures with relative abundances of 12.5% and 13.7%, respectively; also, 3 new T-RFs (213, 509, and 533 bp) were detected for the first time (Fig. 2). In the high-addition (D) enclosures at day 1, the 65-bp T-RF became the most abundant one (relative abundance >9%; Fig. 2), and the other most abundant ones were those of 60, 92, 226, 371, 513, 82 and 369 bp (relative abundances were all >4%), the first five of which were detected for the first time. Furthermore, another 3 new T-RFs (176, 203, and 412 bp) were detected for the first time. At day 4, the structure of the active bacterial community had changed again. The 359-bp T-RF dominated the free-living bacterial population with a relative abundance of 21.3% (Fig. 2). In addition, 9 (287, 291, 298, 302, 415, 421, 444, 533, and 587 bp) out of the 12 distinct bacterial T-RFs with a relative abundance of >1% were detected for the first time.
70 60 50 40 30 20 10
3.3. T-RFLP analysis of the particle-attached bacterial community Unlike those of the free-living bacteria, the T-RFLP profiles of the particle-attached bacteria in the enclosures all had changed after 1 and 4 d of incubation (Fig. 3). Immediately before the start of the incubation (day 0), a total of 14 distinct
0 D ay 0 Day 1 D ay 4
D ay 1 Day 4
Day 1 Day 4
Fig. 3 e Relative abundance of particle-attached bacterial 16S rRNA amplicons recovered from enclosures with different biomass Microcystis sp. Those T-RFs of less than 1% of the total were grouped together as a single group donated by others <1%.
1706
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 0 2 e1 7 1 0
bacterial T-RFs were identified. These T-RFs all showed a relative abundance of >1%, and those of 65, 80, 295 and 314 bp even exhibited a relative abundance of >4%. The latter two dominated bacterial populations with relative abundances of about 43.8% and 16.3%, respectively. At day 1 in the no-addition enclosures, 237-bp and 369-bp T-RFs dominated the bacterial populations with relative abundances of 25.7% and 36.7%, respectively. The latter was detected for the first time, and most of the T-RFs showing a relative abundance of >1% were newly detected (11 out of 14) (Fig. 3). At day 4, the structure of the bacterial community had changed yet again: the 85/86 bp and 134/135 bp T-RFs dominated the bacterial populations with relative abundances of 14.2% and 12.5%, respectively. Meanwhile 4 new T-RFs (111,138,151 and 168 bp) were detected for the first time (Fig. 3). In the medium-addition enclosures, a total of 19 distinct bacterial T-RFs with relative abundances of >1% were identified, of which only 5 had been previously detected on day 1. Furthermore, the bacterial populations were dominated by the newly-detected T-RFs of 85/86 and 92 bp (with relative abundances of 17.1% and 10.3%, respectively). At day 4, the relative abundance of the 134/135 bp T-RF increased from 6.5% to 13.8%, and it dominated the bacterial populations; but the 85/86 bp T-RF decreased to 10.3%. In addition, 7 new T-RFs (92,
111, 129, 138, 163, 194, and 222 bp) were detected for the first time. In the high-addition enclosures at day 1, most of the T-RFs identified also existed in the medium-addition enclosures at the same time. However, two new T-RFs of 60 and 339 bp appeared and became the most abundant (with relative abundances of 7.0% and 9.6%, respectively) in the high-addition enclosures, and other characteristic T-RFs of 94, 106, 208, 213, 325, 374, 511, 513 and 515 bp were detected for the first time. At day 4, the bacterial populations of these enclosures were dominated by T-RFs of 60, 222, 339 and 513 bp (with relative abundances of 12.8%, 13.5%, 10.7% and 9.8%, respectively, and the second one was newly detected). Another two new T-RFs (201 and 204 bp) were detected for the first time (Fig. 3).
3.4.
Phylogenetic analysis and assignment of T-RFs
To identify bacteria, three clone libraries were generated from bacterial 16S DNA templates retrieved from the A enclosures (WA0.2-0d [n ¼ 52]) and D enclosures (WD0.2-4d [n ¼ 33]; WD54d [n ¼ 50]). The phylogenetic affiliations of the various clone sequences were determined using the neighbor-joining tree method (Fig. S1) in the program MEGA 4. In general, only about half of the detected T-RFs could be assigned to defined
Table 1 e Phylogenetic affiliations of free-living bacterial 16S rRNA sequences retrieved in clone libraries generated from the samples. Characteristic T-RFs for different clone groups are given; T-RFs with relative abundance of more than 4% are indicated in bold; T-RFs detected in more than one phylogenetic group are marked with an asterisk (*); T-RFs characteristic for one time point and T-RFs with enhanced increase in their relative abundance are underlined, respectively. Phylogenetic group
T-RFs of the free-living bacteria (bp) W0.2-0d
Actinobacteria Actinomycetales Micrococcineae Bacteroidetes Flavobacteriales Sphingobacteriales Deinococcales Alphaproteobacteria Rhizobiales Rhodobacterales Sphingomonadales Unclassified Betaproteobacteria Burkholderiales Methylophilales Neisseriales Deltaproteobacteria Desulfovibrionales Unclassified Gammaproteobacteria Legionellales Unclassified Gemmatimonadales Acidobacteriales Verrucomicrobiales Planctomycetales Unidentified
WA0.2-1d
WA0.2-4d
WB0.2-1d
WB0.2-4d
65*,359 363*,369
65*,363*,369
65*,363* 365*,369
65,363*,365*,369
65,363*,369
65*,369
94* 94*, 96 80*
94* 94*,96 80*
219 96
96
226
515
60*, 515
515, 511* 511*
60*, 513 533
60*, 513,515 512
80*
80*
92
82,92
65*, 203,363*
65*, 203
155, 570
570
65*,204, 205*, 363* 155, 570
65*, 203, 205*,363* 155, 570
WD0.2-4d 359
80* 60*, 513,515
92
80* 92
65*,365, 203,363* 570
65*,203,205*, 365, 363* 570
93 374 85/86
WD0.2-1d
208* 208*
208* 208*
85/86 572 208* 208*
577 101,148,249, 344,371,541, 569,581,587
101,148,202, 344,569
201,565, 569,
93 374
93
85/86
85/86*
208* 208*
208* 208* 215
208* 208*
101,344,371, 565,569,201
202,344,509, 565, 569
176,249,344, 371,569
533
287,291,298,302, 344,415,421, 443,448,587
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 0 2 e1 7 1 0
phylogenetic groups, and many T-RFs were detected in more than one phylogenetic group (Table 1). Phylogenetic analysis of the cloned sequences revealed that both the free-living and particle-attached bacterial populations involved in breakdown process of the added Microcystis were highly diverse. The 16S rRNA sequences of free-living bacteria in the noaddition enclosures were dominated by Alphaproteobacteria (Rhizobiales and Rhodobacterales), Betaproteobacteria (Burkholderiales), Actinobacteria (Micrococcineae) and Methylophilales during the experiment (Table 1). In the medium-addition enclosures, at day 1 the bacterial populations were dominated by Legionellales (T-RF: 85/86 bp; Fig. 2; Table 1) facultative anaerobic bacteria closely related to L. birminghamiensis (91% similarity), and the 8 new T-RFs were affiliated with the Rhizobiales and Desulfovibrionales. At day 4, Micrococcineae and Betaproteobacteria (Neisseriales and another unclassified group) dominated the free-living bacterial populations in the medium-addition enclosures. Furthermore, the T-RFs of 213 bp and 533 bp that were detected for the first time were assigned to the Verrucomicrobiales and Rhodobacterales, respectively, and the 509-bp T-RF was not represented by any of the clone sequences and therefore could not be assigned to any phylogenetic group. In the high-addition enclosures, at day 1 Micrococcineae and Rhizobiales were the most abundant bacterial populations. The number of T-RFs assigned to the former group decreased from 4 to 2, and the new T-RFs were affiliated with the Rhodobacterales, Sphingobacteriales, Burkholderiales, Acidobacteriales and Gemmatimonadales. At day 4, none of the previously observed T-RFs related to Actinobacteria were observed, except for one of 359 bp that dominated the free-living bacterial populations (21.3%). Furthermore, the T-RF of 533 bp that was detected for the first time at day 4 was assigned to Rhodobacterales; while the other eight new T-RFs were not represented by any of the clone sequences and therefore could not be assigned to any phylogenetic group. In agreement with the T-RFLP results, the particleattached bacterial populations in the enclosures all had changed after 1 and 4 d of incubation (Table S3). At day 0, 16S rRNA sequences of particle-attached were dominated by Micrococcineae, Deinococcales, Sphingomonadales and Burkholderiales. At day 1, the bacterial community of the noaddition enclosure was dominated by a newly identified TRF of 369 bp that was related to Micrococcineae (36.7%). Furthermore, another two new T-RFs (96 and 226 bp) were affiliated with Sphingomonadales, the T-RF of 85/86 bp was assigned to Legionellales and the T-RF of 565 bp could not be assigned to any phylogenetic group. At day 4, Legionellales became the most abundant bacterial population (14.2%). In the medium-addition enclosures at day 1, the T-RFs of 85/86 and 92 bp that were detected for the first time and that dominated this bacterial population (their relative abundances were 13.0% and 9.7%, respectively) were assigned to L. birminghamiensis and unclassified Alphaproteobacteria; other new T-RFs were related to Flavobacteriale, Rhizobiales, Rhodobacterales, Burkholderiales, Neisseriales, Desulfovibrionales and unclassified Gammaproteobacteria. At day 4, Legionellales decreased from 17.1% to 10.3%. The 134/135 bp T-RF, which increased from 6.5% to 13.8%, was not represented by any of the clone sequences and therefore could not be assigned to
1707
any phylogenetic group. Unlike the medium-addition enclosures, the bacterial communities in the high-addition ones were dominated by L. birminghamiensis, Micrococcineae and Rhodobacterales at day 1. At day 4, the particle-attached bacterial populations were dominated by Rhizobiales, Flavobacteriales and Rhodobacterales, which were most closely related to the T-RFs of 60, 222 and 513 bp (relative abundances of 12.8%, 13.5% and 9.8%, respectively, and the second one was newly detected). Another new T-RF (201 bp) was assigned to Burkholderiales and a T-RF of 204 bp could not be assigned to any phylogenetic group (Table 2).
3.5.
Statistical analysis
The CCA model illustrated that the concentrations of Chl a, DO, DOC and pH contributed most to the variance in the freeliving bacterial community (for all canonical axes, p ¼ 0.08, Fig. 4A). The eigenvalues of the first and second axes were 0.92 and 0.45, respectively, indicating that both axes were important. The first two axes explained 54.0% of the observed variation in the composition of the free-living bacterial community, and 76.2% was explained by the full four canonical axes. The first axis showed a high canonical correlation with the concentrations of DO and DOC, while the second axis correlated with the concentrations of Chl a and DOC. Similarly, the variance in the particle-attached bacterial community composition was also attributed mostly to Chl a, DOC, pH and DO (for all canonical axes, p ¼ 0.83, Fig. 4B). The eigenvalues of the first and second axes were 0.55 and 0.40, respectively. The first two axes explained 52.4% of the observed variation in the composition of the free-living bacterial community, and 70.1% was explained by all four canonical axes. The first axis showed a high canonical correlation with the concentrations of Chl a and DOC, while the second axis correlated with DOC and pH. Regression analysis revealed that Chl a concentration was negatively correlated with pH and DO (R2 ¼ 0.42, 0.36, respectively, p < 0.05) and positively correlated with DOC (R2 ¼ 0.45, p < 0.05). A strong positive correlation between the abundance of heterotrophic bacteria and DOC was observed as well (R2 ¼ 0.77, p < 0.01). Lineal regression analysis revealed that there was a negative correlation between Chl a concentration and the relative abundance of Micrococcineae (T-RFs 365 and 565 bp) in enclosure D (R2 ¼ 0.93, p < 0.05). However, this relationship was not very significant in enclosure A (R2 ¼ 0.17, p > 0.05) and enclosure B (R2 ¼ 0.27, p > 0.05).
4.
Discussion
The objective of this study was to characterize the BCC in water columns with differing Microcystis sp. biomass, to document BCC changes following the breakdown of Microcystis and to identify physiochemical parameters in relation to BCC variability. Two culture-independent methods, T-RFLP analysis and cloning/sequencing, were used to analyze the BCC. The changes in the physiochemical parameters in the enclosures at different sampling times (Table S1) clearly indicated the breakdown of Microcystis sp. (Chen et al., 2010).
1708
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 0 2 e1 7 1 0
Table 2 e Phylogenetic affiliations of particle-attached bacterial 16S rRNA sequences retrieved in clone libraries generated from the samples. Characteristic T-RFs for different clone groups are given; T-RFs with relative abundance of more than 4% are indicated in bold; T-RFs detected in more than one phylogenetic group are marked with an asterisk (*); T-RFs characteristic for one time point and T-RFs with enhanced increase in their relative abundance are underlined, respectively. Phylogenetic group
T-RFs of the particle-attached bacteria (bp) W5-0d
Actinobacteria Actinomycetales Bacteroidetes Flavobacteriales Sphingobacteriales Deinococcales Alphaproteobacteria Rhizobiales Rhodobacterales Sphingomonadales Unclassified Betaproteobacteria Burkholderiales Methylophilales Neisseriales Deltaproteobacteria Desulfovibrionales Unclassified Gammaproteobacteria Legionellales Unclassified Gemmatimonadales Verrucomicrobiales Unidentified
65*,97
WA5-1d
WA5-4d
97,369
219
60*
WB5-1d
65*, 97,369
65*, 369
60*, 222 96
60*, 94* 94* 80*
60*, 222
60*, 155*
511*,513, 515
92
339, 511* 80* 92
60*, 513,511*, 515 339,511*
80*
155*
60*
60*, 511*, 513
80
80*
511* 80* 92
65*,151
205* 205*
85/86
74,116,134, 172,179,237, 295,314,320
74, ,90,146, 160,182,190, 237,371, 565
151, 205* 155* 205*
65*
93 374
93
93 374
93
85/86
85/86 572
85/86
103,111,122, 134,138,146, 168,172,182,190
4.1. Change of BCC and bacterial numbers induced by breakdown of Microcystis blooms Our T-RFLP analysis clearly indicate that the composition of both the free-living and particle-attached bacterial communities in Gonghu Bay, which has little Microcystis, changed after the accumulation of a Microcystis bloom, and these changes varied with the biomass of the bloom. LIBSHUFF analysis of clone libraries confirmed these changes after addition of Microcystis biomass in enclosures (Table S3). These shifts were not only detected as the disappearance of bacterial populations and the appearance of new ones but also as the replacement of the most dominant bacterial populations (Fig. 2; Tables 1 and 2). CCA analysis revealed that changes in pH and in the concentrations of DO, DOC and Chl a had potentially important impacts on BCC. This is in accordance with some previous studies (e.g. Mauricea and Leff, 2002; Lindstro¨m et al., 2005; Grossart et al., 2005, 2006; Kolehmainena et al., 2007) underlining these regulating factors of BCC. Regression analysis revealed that the former three parameters were significantly correlated with the concentration of Chl a, suggesting that these chemical parameter changes were mainly induced by Microcystis breakdown (Chen et al., 2010). Furthermore, positive correlation between the abundance of heterotrophic bacteria and DOC suggests that the breakdown of Microcystis biomass promote the growth of heterotrophic bacteria.
WD5-4d
97,65*
80
65* 155*
WD5-1d
97,65* 60* 96 80*
96, 226
WB5-4d
88,90, 103, 134,146,179, 182, 509
103,106,111,122, 129,134,138,146, 160,163,168,172, 179,182,190,194
208 213 90,103,106,116, 129,160, 325
92 204, 205* 339 205*
213 201,172,
It has been suggested that the mucilage of Microcystis contains bacteria which degrade Microcystis cellular material (Maruyama et al., 2003). So the addition of Microcystis biomass into enclosures might have introduced additional bacteria that were not present in the control mesocosm. This addition might not influence the free-living BCC immediately because the bacteria inhabited in the mucilage of Microcystis cells are in general difficult to be dispatched (Wu et al., 2007a). This is confirmed by the fact that there was no significant change of bacterial abundance between the control and enclosures amended with Microcystis (Fig. 1) at the beginning of experiment. The introduction of Microcystis blooms into enclosures may influence the particle-attached BCC. However, the significant shift of particle-attached BCC in enclosures from Day 1 to Day 4 amended with Microcystis blooms suggests that the breakdown of Microcystis blooms did impact the structure of particle-attached BCC. Thus we conclude that the BCC changes in our study were mainly induced by the breakdown of Microcystis blooms.
4.2. Dominance of Actinomycetales during the decomposition of Microcystis blooms Among the shifts in bacterial community composition that we observed, one that should have been paid more attention to is the changes in Actinomycetales in the addition enclosures, most
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 0 2 e1 7 1 0
1709
4.3. Occurrence of abundant Legionellales organisms during accumulation and crash of Microcystis blooms Another important change is the occurrence of high percentage of bacteria affiliated with Legionellales that are mostly related to Legionella birminghamiensis. Legionellales were found in freshwater environments worldwide (Fliermans et al., 1981), but not as abundant in the present study. Some studies found that anthropogenic factors (Ortiz-Roque and Hazen, 1987) such as wastewater discharge favor a diversity of Legionella species (Carvalho et al., 2007). In addition, symbiotic interactions between Legionella and free-living protozoa may be good for the colonization of aquatic habitats by these bacteria (Ortiz-Roque and Hazen, 1987). Tison et al. (1980) even found that over a pH range of 6.9e7.6., L. pneumophila was apparently using cyanobacteria extracellular products as its carbon and energy sources. We were not able to isolate these organisms for further physiological and ecological studies in the current study. Given the fact that many species belong to the Legionellales are potentially pathogenic (Wilkinson et al., 1987; Fields et al., 2002), our investigation suggests that further studies are necessary to reveal the potential associations between Legionellales and Microcystis, and risk assessment of their occurrence to human health, and their application as indicator of water quality in eutrophic freshwater lakes.
Fig. 4 e Canonical correspondence analysis (CCA) biplots showed variable composition of free-living (A) and particle-attached (B) bacteria in relation to the important environmental factors in the enclosures with different biomass of Microcystis sp.
of which were belong to Micrococcineae. In both enclosures amended with Microcystis blooms, the relative abundance of free-living Micrococcineae was higher than that in the control and increased in general from day 1 to day 4 following the decrease of Chl a. Sequences affiliated with Micrococcineae have previously been described in various freshwater habitats (Zwart et al., 2002; Allgaier and Grossart, 2006a, b; Newton et al., 2007; Wu et al., 2007a, b; Holmfeldt et al., 2009). Actinomycetales possess a wide variety of physiological and metabolic properties (Goodfellow and Williams, 1983). The negative correlation between Chl a concentration and the relative abundance of Micrococcineae in our experiment suggests that there might be saprophytic association between the breakdown of Microcystis biomass and Micrococcineae. Yamamoto and his colleagues have isolated 83 Actinomycete strains from a eutrophic lake among which half were found able to lyse cyanobacteria including Microcystis sp. (Yamamoto et al., 1998). Interestingly, some of these isolates are closely affiliated with Micrococcineae that were found in our study. Further isolation of these microbes and their eco-physiological characterizations will help to reveal their exact role in decomposition of Microcystis blooms and nutrient cycling in freshwater lakes.
Acknowledgements This work was supported by Knowledge Innovation Project of Chinese Academy of Sciences (KZCX1-YW-14-1;KZCX2-YWJC302), National Basic Research Program of China (973 program) (No. 2008CB418104), and Jiangsu Provincial Science Foundation (BK2009024). We thank Ray Zhang for the statistical analysis and Feizhou Chen for assistance in the field.
Appendix. Supplementary material Supplementary data related to this article can be found online, at doi:10.1016/j.watres.2010.11.011.
references
Allgaier, M., Grossart, H.P., 2006a. Diversity and seasonal dynamics of actinobacteria populations in four lakes in northeastern Germany. Appl. Environ. Microbiol. 72, 3489e3497. Allgaier, M., Grossart, H.P., 2006b. Seasonal dynamics and phylogenetic diversity of free-living and particle-associated bacterial communities in four lakes in northeastern Germany. Aquat. Microb. Ecol. 45, 115e128. Azam, F.T., Fenchel, J.G., Field, F.S., Gray, J.S., MeyereReil, L.A., Thingstad, F., 1983. The ecological role of water-column microbes in the sea. Mar. Ecol. Prog. Se. 10, 257e263. Carvalho, F.R.S., Vazoller, R.F., Foronda, A.S., Pellizari, V.H., 2007. Phylogenetic study of Legionella species in pristine and
1710
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 0 2 e1 7 1 0
polluted aquatic samples from a tropical Atlantic forest ecosystem. Curr. Microbiol. 55, 288e293. Chen, M., Chen, F., Xing, P., Li, H., Wu, Q.L., 2010. Microbial eukaryotic community in response to Microcystis spp. bloom, as assessed by an enclosure experiment in Lake Taihu, China. FEMS Microbiol. Ecol. 74, 19e31. Cole, J.J., Likens, G.E., Strayer, D.L., 1982. Photosynthetically produced dissolved organic carbon: an important carbon source for planktonic bacteria. Limnol. Oceanogr. 27, 1080e1090. Eiler, A., Bertilsson, S., 2004. Composition of freshwater bacterial communities associated with cyanobacterial blooms in four Swedish lakes. Environ. Microbiol. 6, 1228e1243. Ferdous, Z., 2009. Survival of Vibrio cholerae O139 in association with Anabaena variabilis in four different microcosoms. Am. J. Appl. Sci. 6, 439e444. Fields, B.S., Benson, R.F., Besser, R.E., 2002. Legionella and Legionnaires’ Disease: 25 Years of investigation. Clin. Microbiol. Rev. 15, 506e526. Fliermans, C.B., Cherry, W.B., Orrison, L.H., Smith, S.J., Tison, D.L., Pope, D.H., 1981. Ecological distribution of Legionella pneumophila. Appl. Environ. Microbiol. 41, 9e16. Goodfellow, M., Williams, S.T., 1983. Ecology of actinomycetes. Annu. Rev. Microbiol. 37, 189e216. Grossart, H.P., Czub, G., Simon, M., 2006. Algae-bacteria interactions and their effects on aggregation and organic matter flux in the sea. Environ. Microbiol. 8, 1074e1084. Grossart, H.P., Levold, F., Allgaier, M., Simon, M., Brinkhoff, T., 2005. Marine diatom species harbour distinct bacterial communities. Environ. Microbiol. 7, 860e873. Holmfeldt, K., Dziallas, C., Titelman, J., Pohlmann, K., Grossart, H.P., Riemann, L., 2009. Diversity and abundance of freshwater Actinobacteria along environmental gradients in the brackish northern Baltic Sea. Environ. Microbiol. 11, 2042e2054. Kolehmainena, R.E., Langwaldtb, J.H., Puhakka, J.A., 2007. Natural organic matter (NOM) removal and structural changes in the bacterial community during artificial groundwater recharge with humic lake water. Water Res. 41, 2715e2752. Lane, D.J., 1991. 16S/23S rRNA sequencing. In: Stackebrandt, E., Goodfellow, M. (Eds.), Nucleic Acid Techniques in Bacterial Systematics. John Wiley and Sons, Chichester, pp. 115e175. Lehman, E.M., 2007. Seasonal occurrence and toxicity of Microcystis in impoundments of the Huron River, Michigan, USA. Water Res. 41, 795e802. Lindstro¨m, E.S., Agterveld, M.P.K., Zwart, G., 2005. Distribution of typical freshwater bacterial groups is associated with pH, temperature, and lake water retention time. Appl. Environ. Microbiol. 71, 8201e8206. Liu, W.T., Marsh, T.L., Cheng, H., Forney, L.J., 1997. Characterisation of microbial diversity by determining terminal restriction length polymorphisms of genes encoding 16S rDNA. Appl. Environ. Microbiol. 63, 4516e4522. Maruyama, T., Kato, K., Yokoyama, A., Tanaka, T., Hiraishi, A., Park, H.D., 2003. Dynamics of microcystin-degrading bacteria in mucilage of Microcystis. Microb. Ecol. 46, 279e288. Masango, M., Myburgh, J., Botha, C., Labuschagne, L., Naicker, D., 2008. A comparison of in vivo and in vitro assays to assess the toxicity of algal bloom. Water Res. 42, 3241e3248. Mauricea, P.A., Leff, L.G., 2002. Hydrogeochemical controls on the organic matter and bacterial ecology of a small freshwater wetland in the New Jersey Pine Barrens. Water Res. 36, 2561e2570. Muyzer, G., Dewaal, E.C., Uitterlinden, A.G., 1993. Profiling of complex microbial populations by denaturig gradient gel electrophoresis analysis of polymerase chain reactionamplified genes coding for 16S rRNA. Appl. Environ. Microbiol. 59, 695e700. Newton, R.J., Jones, S.E., Helmus, M.R., McMahon, K.D., 2007. Phylogenetic ecology of the freshwater Actinobacteria acI Lineage. Appl. Environ. Microbiol 73, 7169e7176.
Ortiz-Roque, C.M., Hazen, T.C., 1987. Abundance and distribution of Legionellaceae in Puerto Rican waters. Appl. Environ. Microbiol. 53, 2231e2236. Paerl, H.W., Fulton, R.S., Moisander, P.H., Dyble, J., 2001. Harmful freshwater algal blooms, with an emphasis on cyanobacteria. The Scientific World 1, 76e113. Paerl, H.W., Dyble, J., Moisander, P.H., Noble, R.T., Piehler, M.F., Pinckney, J.L., Steppe, T.F., Twomey, L., Valdes, L.M., 2003. Microbial indicators of aquatic ecosystem change: current applications to eutrophication studies. FEMS Microbiol. Ecol. 46, 233e246. Porter, K.G., Feig, Y.S., 1980. The use of DAPI for identifying and counting aquatic microflora. Limnol. Oceanogr. 25, 943e948. Qin, B., Xu, P., Wu, Q., Luo, L., Zhang, Y., 2007. Environmental issues of Lake Taihu, China. Hydrobiol. 581, 3e14. Schloss, P.D., Westcott, S.L., Ryabin, T., Hall, J.R., Hartmann, M., Hollister, E.B., Lesniewski, R.A., Oakley, B.B., Parks, D.H., Robinson, C.J., Sahl, J.W., Stres, B., Thallinger, G.G., Van Horn, D.J., Weber, C.F., 2009. Introducing mothur: open-source, Platform-independent, CommunityeSupported software for describing and comparing microbial communities. Appl. Environ. Microbiol. 75, 7537e7541. Tamura, K., Dudley, J., Nei, M., Kumar, S., 2007. MEGA4: Molecular Evolutionary Genetics Analysis (MEGA) Software Version 4.0. Mol. Biol. Evol. 24, 1596e1599. ter Braak, C.J.F., Smilauer, P., 2002. CANOCO Reference Manual and CanoDraw for Windows User’s Guide: Software for Canonical Community Ordination (Version 4.5). Microcomputer Power, Ithaca, N.Y., USA. Tison, D.L., Pope, D.H., Cherry, W.B., Fliermans, C.B., 1980. Growth of Legionella pneumophila in association with blue-green algae (cyanobacteria). Appl. Environ. Microbiol. 39, 456e459. Wilkinson, H.W., Thacker, W.L., Benson, R.F., Polt, S.S., Brookings, E., Mayberry, W.R., Brenner, D.J., Gilley, R.G., Kirklin, J.K., 1987. Legionella birminghamensis sp. nov. isolated from a cardiac transplant recipient. J. Appl. Microbiol. 25, 2120e2122. Wu, Q.L., Chen, Y., Xu, K., Liu, Z., Hahn, M.W., 2007a. Intrahabitat heterogeneity of the microbial food web structure under the regime of eutrophication and sediment resuspension in the large subtropical shallow LakeTaihu. Hydrobiol. 581, 241e254. Wu, Q.L., Zwart, G., Wu, J.F., Kamst-Van Agterveldand, M.P., Liu, S.J., Hahn, M.W., 2007b. Submersed macrophytes play an important role in structuring the bacterioplankton diversity in a large subtropical shallow Taihu Lake, China. Environ. Microbiol. 9, 2765e2774. Wu, Q.L., Zwart, G., Schauer, M., Kamst-van Agterveld, M.P., Hahn, M.W., 2006. Bacterioplankton community composition along a salinity gradient of sixteen high-mountain lakes located on the Tibetan Plateau, China. Appl. Environ. Microbiol. 72, 5478e5485. Xing, P., Kong, F., 2007. Intra-habitat heterogeneity of environmental factors regulating bacterioplankton community composition in Lake Taihu, China. Aquat. Microb. Ecol. 46, 113e122. Yamamoto, Y., Kouchiwa, T., Hodoki, Y., Hotta, K., Uchida, H., Harada, K., 1998. Distribution and identification of actinomycetes lysing cyanobacteria in a eutrophic lake. J. Appl. Phycol. 10, 391e397. Zhang, X., Warming, T.P., Hu, H., Christoffersen, K.S., 2009. Life history responses of Daphnia magna feeding on toxic Microcystis aeruginosa alone and mixed with a mixotrophic Poterioochromonas species. Water Res. 43, 5053e5062. Zwart, G., Crump, B.C., Agterveld, M.P.K., Hagen, F., Han, S., 2002. Typical freshwater bacteria: an analysis of available 16S rRNA gene sequences from plankton of lakes and rivers. Aquat. Microb. Ecol. 28, 141e155.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 1 1 e1 7 1 9
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Insight into the heavy metal binding potential of dissolved organic matter in MSW leachate using EEM quenching combined with PARAFAC analysis Jun Wu, Hua Zhang, Pin-Jing He*, Li-Ming Shao State Key Laboratory of Pollution Control and Resources Reuse, Key Laboratory of Yangtze River Water Environment, College of Environmental Science and Engineering, Tongji University, 1239 Siping Road, Shanghai 200092, PR China
article info
abstract
Article history:
Dissolved organic matter (DOM) plays an important role in heavy metal migration from
Received 4 September 2010
municipal solid waste (MSW) to aquatic environments via the leachate pathway. In this
Received in revised form
study, fluorescence excitation-emission matrix (EEM) quenching combined with parallel
13 November 2010
factor (PARAFAC) analysis was adopted to characterize the binding properties of four heavy
Accepted 15 November 2010
metals (Cu, Pb, Zn and Cd) and DOM in MSW leachate. Nine leachate samples were
Available online 24 November 2010
collected from various stages of MSW management, including collection, transportation, incineration, landfill and subsequent leachate treatment. Three humic-like components
Keywords:
and one protein-like component were identified in the MSW-derived DOM by PARAFAC.
Municipal solid waste
Significant differences in quenching effects were observed between components and metal
(MSW) leachate
ions, and a relatively consistent trend in metal quenching curves was observed among
Dissolved organic matter (DOM)
various leachate samples. Among the four heavy metals, Cu(II) titration led to fluorescence
Heavy metal
quenching of all four PARAFAC-derived components. Additionally, strong quenching effects were only observed in protein-like and fulvic acid (FA)-like components with the
Fluorescence excitation-emission
matrix
(EEM)
addition of Pb(II), which suggested that these fractions are mainly responsible for Pb(II)
quenching
binding in MSW-derived DOM. Moreover, the significant quenching effects of the FA-like
Parallel factor (PARAFAC) analysis
component by the four heavy metals revealed that the FA-like fraction in MSW-derived DOM plays an important role in heavy metal speciation; therefore, it may be useful as an indicator to assess the potential ability of heavy metal binding and migration. ª 2010 Elsevier Ltd. All rights reserved.
1.
Introduction
Leachate from municipal solid waste (MSW) has raised concerns as a potential pathway of contaminant emission (Huang et al., 2009; Michalzik et al., 2007). Heavy metals are significant pollutants in MSW leachate, and are often present in concentrations ranging from micrograms to milligrams per liter (Christensen et al., 2001). Primary investigations (Baumann et al., 2006; Baun and Christensen, 2004; Li et al., 2009) of heavy metal distribution in landfill leachate have shown that
significant fractions of heavy metals were associated with MSWderived dissolved organic matter (DOM), suggesting that DOM plays an important role in heavy metal speciation and migration. Christensen et al. (1999, 1996) and Christensen and Christensen (1999) found that the DOM derived from MSW landfill leachate had a high affinity for Cu, Pb, Cd, Zn, and Ni (especially for Cu and Pb); thus, enhancing their mobility in leachate-polluted waters. By using biological assay methods, it was observed that the heavy metal binding capacities fluctuated by more than one order of magnitude among various leachates,
* Corresponding author. Tel./fax: þ86 21 6598 6104. E-mail address:
[email protected] (P.-J. He). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.11.022
1712
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 1 1 e1 7 1 9
which was attributed to their variable compositions (Ward et al., 2005). Isolation of humic substances such as humic acid (HA) or fulvic acid (FA) from natural DOM was preferable to study the relationship between DOM and heavy metals (da Silva and Oliveira, 2002; Yang and van den Berg, 2009). Humic substances have been reported to exhibit strong affinities toward metal ions due to the large number of ionizable functional groups, which are mainly carboxylic and phenolic groups (Hernandez et al., 2006; Martensson et al., 1999; Terbouche et al., 2010). However, the heterogeneity and operationally-defined extraction and purification process of humic substances via the addition of chemicals enhance the complexity of the binding potential of DOM with metals. Furthermore, metal binding properties of DOM in the leachate from MSW landfill or other stages of MSW management have seldom been investigated to the best of our knowledge. Therefore, new insight into the heavy metal binding potential of MSW-derived DOM is needed. Fluorescence excitation-emission matrix (EEM) spectroscopy is a simple, sensitive and non-destructive technique that can provide valuable information regarding the molecular structure of DOM. Recent studies have demonstrated that EEM spectroscopy combined with a quenching method can be applied as a reliable technique to enable a better understanding of the binding properties of metal ions and isolated fluorescence substances from soil and sewage sludge (Plaza et al., 2006a, b). However, the EEM spectra of in situ DOM cannot be easily identified owing to the various types of overlapping fluorophores (Henderson et al., 2009). Recent advances in parallel factor (PARAFAC) analysis, a multivariate chemometric method, have greatly improved the quantitative interpretation of EEMs (Engelen et al., 2009; Fellman et al., 2009). PARAFAC can be applied to decompose fluorescence EEMs into different independent groups of fluorescent components, which can then reduce the interference among fluorescent compounds and allow a more accurate quantification (Andersen and Bro, 2003). Ohno et al. (2008) demonstrated that the binding properties between water soluble soil organic matter and trivalent metals (Al and Fe) could be well characterized by EEM quenching combined with PARAFAC. Additionally, Yamashita and Jaffe (2008) successfully evaluated the binding properties of surface water-borne DOM and divalent metals (Cu and Hg) using the same method. The structures and compositions of MSW-derived DOM are much more complicated than the aforementioned DOMs (Christensen et al., 1998; He et al., 2006). Furthermore, a series of biochemical reactions may occur, resulting in changes in the DOM properties during integrated MSW management (Huo et al., 2009). As a result, simple quantification of the heavy metal binding potential of MSW-derived DOM has remained elusive. In this study, the feasibility of EEM quenching-PARAFAC for use in characterization of the binding properties of heavy metals and DOM in MSW leachate was investigated. Nine leachate samples from various stages in MSW management were collected, and four heavy metals (Cu, Pb, Zn and Cd) that are common in MSW leachate were used as fluorescent quenching agents for titration. Specific fractions in MSWderived DOM responsible for heavy metals binding were identified and the differences in the binding characteristics among
various PARAFAC-derived components, heavy metals and DOMs were analyzed.
2.
Materials and methods
2.1.
Sample collection and preparation
Due to the high content of moisture and putrescible organic waste, MSW leachate is produced during collection, transportation, and storage. Nine leachate samples were collected from different stages of MSW management in Shanghai, China (S1eS9 as shown in Fig. 1). S1 and S2 were collected from an MSW collection vehicle and transfer station, respectively, with waste generated within one day. S3 was collected from the refuse bunk of an incineration plant in which MSW was stored for 1e3 days before incineration. S4 was gathered from the effluent of S3 treated by a sequential batch reactor (SBR) process combined with ultrafiltration (UF). Fresh leachate (S5) and old leachate (S7) were obtained from sanitary landfills comprising refuse aged 1e2 years and 5e10 years, respectively. S6 was the effluent of S5 treated by an upflow anaerobic sludge bed (UASB) combined with a membrane bioreactor (MBR). Leachate S7 was treated by an anaerobic-anoxic-oxic (A2O) process and an SBR, and the effluent of these processes comprised S8 and S9, respectively. The leachate samples were collected in pre-cleaned (by nitric acid and Milli-Q water) brown sampling bottles, after which they were filtered through a 0.45 mm membrane filter. The filtrates were then stored at 4 C until use. The physiochemical characteristics of the filtered samples are summarized in Table 1.
2.2.
Fluorescence titration
Prior to fluorescence titration, the leachate samples were diluted with Milli-Q water to TOC < 10 mg/L to ensure that the maximum fluorescence signal was below the upper detection limit of the spectrometer. Aliquots of 25 mL of the diluted solution of DOM were titrated in 40-mL brown sealed vials with 0.01 or 0.1 mol/L Pb(NO3)2, Cu(NO3)2, Zn(NO3)2, and Cd(NO3)2 using an automatic syringe. The metal concentrations in the final solutions ranged from 0 to 100 mmol/L. To maintain a constant pH condition before and after titration, the metal titrants were adjusted to pH 4.0 for Cu(NO3)2, pH 4.5 for Pb(NO3)2, pH 6.0 for Zn(NO3)2 and pH 7.0 for Cd(NO3)2, based on the Visual MINTEQ calculation that no precipitate existed in the solution and no more than 0.025 mL of the metal titrant was added during titration. All titrated solutions were shaken for 24 h at 25 0.1 C to ensure complexation equilibrium. The fluorescence EEM spectra of the titrated samples were measured using a Cary Eclipse fluorescence spectrophotometer (Varian Inc., Palo Alto, CA, USA), and collected by subsequent scanning emission from 250 to 500 nm at 2 nm increments by varying the excitation wavelength from 200 to 450 nm at 10 nm increments. The spectra were recorded at a scan rate of 1200 nm/min, using excitation and emission slit bandwidths of 5 nm. The voltage of the photomultiplier tube was set to 800 V to enable low level light detection.
1713
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 1 1 e1 7 1 9
Incinerator S3 MBR+UF Leachate
Transfer station
Collector
S4 Effluent
landfill S1 Leachate
S5 Fresh leachate UASB+MBR
S2 Leachate
S7 Old leachate
AO SBR
S6 Effluent S8 Effluent S9 Effluent
Fig. 1 e Leachate samples collected from MSW management facilities. MBR: membrane bioreactor, UF: ultrafiltration, UASB: upflow anaerobic sludge bed, A2O: anaerobic-anoxic-oxic process, SBR: sequential batch reactor.
2.3.
PARAFAC modeling
split half analysis, residual analysis, and visual inspection (Stedmon and Bro, 2008).
The approach of PARAFAC analysis of EEMs has been described in detail elsewhere (Bro, 1997; Stedmon and Bro, 2008) and is briefly described here. PARAFAC is a generalization of bilinear principal component analysis (PCA) to higher order arrays. In other words, PARAFAC decomposes N-way arrays into N loading matrices. Therefore, if fluorescence EEMs are arranged in a three-way array X of dimensions I J K, where I is the number of samples, J the number of emission wavelengths, and K the number of excitation wavelengths, PARAFAC decomposes them into three matrices A (the score matrix), B and C (loading matrices) with elements aif, bjf, and ckf. In this study, PARAFAC analysis was conducted using MATLAB 7.0 (Mathworks, Natick, MA) with the DOMFluor toolbox (www.models.life.ku.dk). A non-negativity constraint was applied to the parameters allowing only chemically relevant results. Some preprocessing steps were adopted to minimize the influence of scatter lines and other attributes of the EEM landscape. The EEM of a control (Milli-Q water) was subtracted from each sample EEM and the other Rayleigh and Raman scatters were removed according to the protocol described by Bahram et al. (2006). After the removal of the Rayleigh and Raman scatters, EEMs were normalized by dividing the spectrum by the corresponding TOC concentration. This resulted in reduction of the impact of varying DOM concentrations in different leachate samples on the component score matrix. The PARAFAC models with two to eight components were computed for the EEMs. Determination of the correct number of components was primarily based on
2.4.
Complexation modeling
The complexation parameters between heavy metals and PARAFAC-derived components were determined using the single-site fluorescence quenching model proposed by Ryan and Weber (1982). This model is based on the assumption that metal ion binding occurs at identical and independent binding sites or ligands and only 1:1 metal/ligand complexes are formed. Because PARAFAC can decompose the complex mixture of DOM fluorophores into several independent fluorescence components, the application of this model may provide more appropriate information than the fluorescence intensity from the peak maxima of bulk samples. The complexation parameters were determined using nonlinear fitting of Eq. (1) 1 1 þ KM CL þ KM CM I ¼ I0 þ ðIML I0 Þ 2KM CL qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð1 þ KM CL þ KM CM Þ2 4K2M CL CM
(1)
where, I and I0 are the fluorescence intensity at the metal concentration, CM, and at the beginning of the titration (without adding metals), respectively. IML is the limiting value below which the fluorescence intensity does not change due to the addition of metal. KM and CL are the conditional stability constant and total ligand concentration, respectively. Nonlinear fitting using the QuasieNewton algorithm was applied to estimate the IML, KM and CL.
Table 1 e Physiochemical characteristics of the leachate samples. Sample S1 S2 S3 S4 S5 S6 S7 S8 S9
TC (mg/L)
IC (mg/L)
TOC (mg/L)
TN (mg/L)
pH
Ca (mg/L)
Mg (mg/L)
Al (mg/L)
Fe (mg/L)
Cu (mg/L)
Pb (mg/L)
Zn (mg/L)
Cd (mg/L)
Mn (mg/L)
11300 19000 13300 1230 8860 1050 2350 220 1240
780 71 1090 560 753 790 1390 93 916
10520 18929 12210 670 8107 260 960 127 324
1140 1210 2030 ND 3820 61 2380 784 1050
4.03 3.92 6.06 7.99 7.17 8.45 7.48 7.76 8.36
907 805 1400 51.7 696 33.4 253 43.5 58.1
115 190 273 210 311 133 123 154 105
8.95 6.53 <0.05 0.75 0.275 ND 0.108 ND 0.715
55.7 73.1 5.53 0.795 ND ND 1.16 ND 0.205
0.033 ND ND ND ND ND 0.045 ND <0.005
ND 0.648 ND ND ND ND ND ND ND
3.74 3.35 ND ND ND ND ND ND ND
ND 0.015 ND ND ND ND ND ND ND
0.543 2.72 7.56 0.233 1.05 0.095 0.090 0.035 0.133
ND: not detected; TC: total carbon; IC: inorganic carbon; TOC: total organic carbon.
Fig. 2 e Fluorescence excitation-emission matrix spectra of the nine leachate samples (bulk sample) with or without titration of Cu(II), Pb(II), Zn(II) and Cd(II) at a total concentration of 50 mmol/L. Blank: no added heavy metals, FI/TOC: fluorescence intensity per unit total organic carbon (arbitrary units/(mg/L)).
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 1 1 e1 7 1 9
3.
Results and discussion
3.1.
EEM contours of the leachate samples
The EEM spectra of the leachate samples measured in the absence and presence of Cu(II), Pb(II), Zn(II) or Cd(II) at a total concentration of 50 mmol/L are shown in Fig. 2. The results revealed that the EEM contours of the DOM fraction in S1 and S2 were similar, being characterized by two peaks at excitation/emission (Ex/Em) values of 230/336 and 280/334 for S1, and 230/350 and 280/352 for S2, which are commonly labeled as protein-like substances (Chen et al., 2003). As the MSW storage time increased, the fluorescence intensity of DOM in leachate at longer Ex/Em wavelengths increased gradually (S3eS5eS7) which is consistent with the humification process of MSW. Comparison of the EEMs of the DOMs before (S3, S5 and S7) and after (S4, S6, S8 and S9) treatment revealed that the protein-like substances in leachate could be removed effectively whereas humic-like substances were more concentrated per unit of total organic carbon (TOC). Addition of Cu(II), Pb(II), Zn(II) or Cd(II) to DOM solutions induced various changes in the EEM spectra depending on DOM characteristics and metal species. In particular, a marked decrease in fluorescence intensity in all nine leachate samples was observed in response to the addition of Cu(II), indicating its large quenching effect on DOM in MSW leachate. Conversely, the wavelengths of Ex/Em peaks of S1, S4, S7, S8 and S9 remained nearly constant when Zn(II) or Cd(II) was added. Additionally, Pb(II) titration induced an obvious quenching
1715
effect on the DOM in fresh leachate (S1, S2, S3, S4, S5 and S6), whereas it had only a slight effect on the DOM in old leachate (S7, S8 and S9). These results supported the suggestion that the binding characteristics between Pb(II) and various fractions in DOM might be different. As expected, further analysis based on visual peaking was quite difficult owing to serious overlaps in the peaks of MSWderived DOM, particularly for S3, S5 and S7.
3.2.
PARAFAC analysis of EEM spectra
The EEM spectra of nine leachate samples titrated with four heavy metals (Cu, Pb, Zn and Cd) at ten different concentrations were analyzed by PARAFAC. Split half analysis, residual analysis and visual inspection identified that four components were appropriate (Fig. 3). All the fluorescence EEMs can be successfully decomposed by PARAFAC analysis into a fourcomponent model, despite the dissimilar fluorescence characteristics of the nine leachate samples and the different quenching effect of different metals at various concentrations. As shown in Fig. 4, the PARAFAC model identified three humic-like substances (Component 1, Component 2 and Component 4) and one protein-like substance (Component 3). Component 1 was comprised of two peaks that were similar to the hydrophobic fraction (Ex/Em 230e245/400e430) and hydrophilic fraction (Ex/Em 300e350/400e430) of surface water-borne DOM identified by Chen et al. (2003). The fluorescence characteristics of component 2 have rarely been identified. Lu¨ et al. (2009) observed similar fluorophores
Fig. 3 e EEM analysis by DOMFluor-PARAFAC model a) Split half analysis, b) Excitation and emission loadings of four PARAFAC-derived components, c) Residual analysis of components 3e5.
1716
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 1 1 e1 7 1 9
200 Component 1 Component 2 Component 3 Component 4
150 100 50 0 100 80
Percentage (%)
(Ex/Em 230,260,340/466 and 250,310,360/464) and suggested that this may have been related to the pyrene family. The peak of component 4 was similar to that of standard Suwannee River fulvic acid (214e220,440e450) reported by Chen et al. (2003). The fluorophore of component 3, which was characterized by two peaks of Ex/Em 225e237/340e381 as well as Ex/Em 275/340, exhibited EEM peaks resembling tryptophan substances derived from sewage DOM (Henderson et al., 2009). Baker and Curry (2004) also identified one similar peak (Ex/Em 220e230/340e370) in three landfill leachate samples. PARAFAC analysis provided additional quantitative information describing the distribution of the four components in the nine leachate DOMs (Fig. 5). The protein-like substances represented by component 3, which occupied a dominant proportion in the fresh leachate samples dropped gradually from 76.3% (S1), 67.9% (S2), 58.1% (S3), and 40.5% (S5) to 29.2% (S7) as the MSW storage time increased, and further decreased by 2.43 (S4), 2.95 (S6), 1.79 (S8) and 1.69 (S9) times after various treatments, whereas the humic-like substances increased accordingly.
Fluorescence intensity per unit TOC arbitrary unit/(mg/L)
Fig. 4 e Fluorescence excitation-emission matrix contours of the four components identified by the DOMFluor-PARAFAC model.
60 40 20 0
3.3. Interactions between PARAFAC-derived components and heavy metals Fig. 6 shows the fluorescence quenching curves of each component with the addition of Cu(II), Pb(II), Zn(II) and Cd(II). Although marked differences in fluorescent fluorophores were observed between component 1 and component 2
S1
S2
S3
S4
S5
S6
S7
S8
S9
Fig. 5 e Distribution of four PARAFAC-derived components in nine MSW leachate samples (no added heavy metals). The percentage of each component was calculated by dividing its fluorescence intensity per unit of TOC by that of the corresponding bulk sample.
1717
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 1 1 e1 7 1 9
C o m po nent 1
75 60 30 15 0
C o m po nent 2
160 120 80 30 15 0
C o m p o n en t 3
75 60 45 30 15 0 90 75 60 45 30 15 0
C o m p o n en t 4
Fluorescence intensity per unit TOC arbitrary units/(mg/L)
45
0 20 40 60 80 100
0 20 40 60 80 100
0 20 40 60 80 100
0 20 40 60 80 100
Cu (µ mol/L)
Pb (µ mol/L)
Zn (µ mol/L)
Cd (µ mol/L)
S1
S2
S3
S4
S6
S7
S8
S9
S5
Fig. 6 e Changes in the fluorescence intensity of four PARAFAC-derived components with the addition of Cu(II), Pb(II), Zn(II) and Cd(II).
(Fig. 4), the fluorescence quenching curves of these two components with four heavy metals were quite similar. A larger quenching effect was observed for Cu(II), whereas negligible quenching effects were found for Pb(II), Zn(II) and Cd(II) by component 1 and 2. The quenching characteristics of these two components with the addition of Cu(II), Zn(II) and
Cd(II) were quite similar to that of commercial HA (Divya et al., 2009), soil-borne HA and compost-derived HA (Hernandez et al., 2006; Plaza et al., 2006a; Provenzano et al., 2004). However, the negligible effects of Pb(II) on those two components in this study were different from those of HA derived from various sources in previous studies.
Table 2 e The log KM values for the humic-like components in MSW-derived DOM with four heavy metals (Cu, Pb, Zn and Cd) determined by Ryan and Weber model.a Symbol
S1 S2 S3 S4 S5 S6 S7 S8 S9
Cu
Pb
Comp1
Comp2
Comp4
e e e 5.94(0.96) e 3.77(0.98) 4.61(0.94) 4.27(0.99) 3.77(0.96)
e e e 4.66(0.99) FM FM FM FM FM
5.62(0.98) 5.48(0.94) 4.71(0.96) 5.58(0.98) 4.42(0.99) 5.34(0.99) 5.03(0.95) e 5.36(0.96)
Zn
Cd
Comp4 FM 4.30(0.93) 4.78(0.94) 4.31(0.98) 4.08(0.99) 3.84(0.99) FM e 5.01(0.99)
4.44(0.96) 4.81(0.95) 3.75(0.96) FM 3.99(1.00) 4.24(0.99) 4.61(0.98) e 4.08(0.99)
FM FM FM 4.64(0.94) FM FM FM e 5.10(0.99)
a The values in parentheses are R2, FM: failed to be modeled, “e”: the fluorescence intensity per unit of TOC was too low (Fig. 4) to be modeled.
1718
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 1 1 e1 7 1 9
FA-like component 4 was quenched significantly by each of the four heavy metals, which is consistent with previous reports describing the quenching effects of the heavy metals on commercial FA (Zhao and Nelson, 2005) and soil-borne FA (da Silva and Oliveira, 2002). These results strongly suggest that FA-like substances played a key role in the complexation between heavy metals and DOM in MSW leachate, especially for Zn(II) and Cd(II). In previous studies, investigators focused on the effect that elevated organic matter had on facilitating the transport of heavy metals when MSW leachate enters surface waters (Christensen et al., 1999, 1996). In this study, component 4 was found to account for more than 20% (Fig. 5) of all leachate samples except for S8 and S9. The relatively higher FA-like content compared to surface water-borne DOM may be another reason for the improved mobility of heavy metals in leachate-polluted waters. Recognizing the important role that FA-like substances in MSW-derived DOM have played in heavy metal binding, some strategies could be adopted to improve the MSW leachate treatment and management to control the migration of heavy metals. It is well-known that fluorescent protein-like substances are quenched or enhanced by complexation of metal ions. However, the fluorescence quenching method, which is widely used to determine the binding parameters between humic fluorescence substances and metal ions, has seldom been utilized to characterize the heavy metal binding potential of specific protein-like substances. In this study, larger fluctuations in the quenching curves of protein-like component 3 were observed, compared to those of humic-like components, especially for Zn(II) and Cu(II). Similar phenomena were also observed in a study conducted by Yamashita and Jaffe (2008), who employed PARAFAC-EEM quenching to explore the interactions between surface water-borne DOM and Cu(II). These results suggest that this method may not be appropriate for evaluation of the binding characteristics of protein-like substances and metal ions. The exact reason for these findings is not yet clear, but it may be related to the stability of the complex of protein-like substances and metals. Cu(II) quenched the fluorescence intensity of all four PARAFAC-derived components. It is interesting to note that the quenching effects with the addition of Pb(II) were strong for component 3 and 4, whereas they were weak for component 1 and 2. In particular, FA-like and protein-like fractions played a dominant role in Pb(II) binding, which explains the different binding characteristics between Pb(II) with fresh and aged leachate. By contrast, the quenching effects in the presence of Zn(II) and Cd(II), which could not be observed based on visual peaking of the bulk sample fluorophore, were observed in component 3 as well as component 4. These findings clearly demonstrated that PARAFAC-based results may provide additional information that may be neglected during visual peaking analysis due to overlapping fluorophores. The stability constants (log KM) calculated using the Ryan and Weber Model for PARAFAC-derived humic-like components and four heavy metals are listed in Table 2. The log KM values ranged from 3.77 to 5.94, 3.84 to 5.01 and 3.75 to 4.81 for Cu(II), Pb(II) and Zn(II), respectively, which were in the same ranges as those found for bulk DOM samples (Luster et al., 1996), physical fractionate DOM samples (de Zarruk et al., 2007) and humic substances (Plaza et al., 2006b; Provenzano et al., 2004).
There was no systematic trend of log KM values observed among the leachate samples or PARAFAC-derived components. The binding parameters between component 1 and Cu(II) could be modeled well, whereas those between component 2 and Cu (II) could not be modeled for most samples, even though similar quenching curves were observed (Fig. 6). These findings suggested that the binding mechanisms between Cu(II) and component 1 and component 2 may be different. Additionally, the binding parameters between component 4 and Cd(II) could not be modeled well by the 1:1 complexation model, which was in agreement with the binding mechanisms of Cd(II) and commercial FA (Grassi and Daquino, 2005).
4.
Conclusions
Four components with characteristic peaks at Ex/Em of (240, 330)/412, (250, 300, 360)/458, (230, 280)/340 and 220/432, were identified by the DOMFluor-PARAFAC model. The results suggested that all the fluorescence EEMs could be successfully decomposed by PARAFAC analysis into a four-component model, despite the dissimilar fluorescence characteristics of the nine leachate samples and the different quenching effects of different metals at various concentrations. The combination of EEM quenching and PARAFAC could provide additional valuable information regarding the binding properties between heavy metals and specific humic-like fluorescence components in DOM, which may be neglected during visual peaking analysis due to overlapping fluorophores. Therefore, it is a useful tool for evaluation of the interactions between DOM and heavy metals given its nondestructive nature, high sensitivity and selectivity. PARAFACbased results revealed that the FA-like fraction in MSW-derived DOM played an important role in heavy metal speciation; therefore, it may be useful as an indicator to assess the potential ability of heavy metal binding and migration.
Acknowledgement This study was financially supported by the National Basic Research Program of China (973 Program No. 2011CB201504), the National Natural Science Foundation of China (No. 20807031) and Ministry of Education (No. 20090072120068).
references
Andersen, C.M., Bro, R., 2003. Practical aspects of PARAFAC modeling of fluorescence excitation-emission data. J. Chemom. 17 (4), 200e215. Bahram, M., Bro, R., Stedmon, C., Afkhami, A., 2006. Handling of Rayleigh and Raman scatter for PARAFAC modeling of fluorescence data using interpolation. J. Chemom. 20 (3e4), 99e105. Baker, A., Curry, M., 2004. Fluorescence of leachates from three contrasting landfills. Water Res. 38 (10), 2605e2613. Baumann, T., Fruhstorfer, P., Klein, T., Niessner, R., 2006. Colloid and heavy metal transport at landfill sites in direct contact with groundwater. Water Res. 40 (14), 2776e2786.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 1 1 e1 7 1 9
Baun, D.L., Christensen, T.H., 2004. Speciation of heavy metals in landfill leachate: a review. Waste Manage. Res. 22 (1), 3e23. Bro, R., 1997. PARAFAC. Tutorial and applications. Chemometr. Intell. Lab. 38 (2), 149e171. Chen, W., Westerhoff, P., Leenheer, J.A., Booksh, K., 2003. Fluorescence excitation-emission matrix regional integration to quantify spectra for dissolved organic matter. Environ. Sci. Technol. 37 (24), 5701e5710. Christensen, J.B., Christensen, T.H., 1999. Complexation of Cd, Ni, and Zn by DOC in polluted groundwater: a comparison of approaches using resin exchange, aquifer material sorption, and computer speciation models (WHAM and MINTEQA2). Environ. Sci. Technol. 33 (21), 3857e3863. Christensen, J.B., Jensen, D.L., Christensen, T.H., 1996. Effect of dissolved organic carbon on the mobility of cadmium, nickel and zinc in leachate polluted groundwater. Water Res. 30 (12), 3037e3049. Christensen, J.B., Jensen, D.L., Gron, C., Filip, Z., Christensen, T.H., 1998. Characterization of the dissolved organic carbon in landfill leachate-polluted groundwater. Water Res. 32 (1), 125e135. Christensen, J.B., Botma, J.J., Christensen, T.H., 1999. Complexation of Cu and Pb by DOC in polluted groundwater: a comparison of experimental data and predictions by computer speciation models (WHAM and MINTEQA2). Water Res. 33 (15), 3231e3238. Christensen, T.H., Kjeldsen, P., Bjerg, P.L., Jensen, D.L., Christensen, J.B., Baun, A., Albrechtsen, H.J., Heron, C., 2001. Biogeochemistry of landfill leachate plumes. Appl. Geochem. 16 (7e8), 659e718. da Silva, J.C.G.E., Oliveira, C.J.S., 2002. Metal ion complexation properties of fulvic acids extracted from composted sewage sludge as compared to a soil fulvic acid. Water Res. 36 (13), 3404e3409. de Zarruk, K.K., Scholer, G., Dudal, Y., 2007. Fluorescence fingerprints and Cu2þ-complexing ability of individual molecular size fractions in soil- and waste-borne DOM. Chemosphere 69 (4), 540e548. Divya, O., Venkataraman, V., Mishra, A.K., 2009. Analysis of metal ion concentration in humic acid by excitationeemission matrix fluorescence and chemometric methods. J. Appl. Spectrosc. 76 (6), 864e875. Engelen, S., Frosch, S., Jorgensen, B.M., 2009. A fully robust PARAFAC method for analyzing fluorescence data. J. Chemom. 23 (3e4), 124e131. Fellman, J.B., Miller, M.P., Cory, R.M., D’Amore, D.V., White, D., 2009. Characterizing dissolved organic matter using PARAFAC modeling of fluorescence spectroscopy: a comparison of two models. Environ. Sci. Technol. 43 (16), 6228e6234. Grassi, M., Daquino, V., 2005. 113Cd NMR and fluorescence studies of multiple binding mechanisms of Cd(II) by the Suwannee river fulvic acid. Anal. Chim. 95 (7e8), 579e591. He, P.J., Xue, J.F., Shao, L.M., Li, G.J., Lee, D.J., 2006. Dissolved organic matter (DOM) in recycled leachate of bioreactor landfill. Water Res. 40 (7), 1465e1473. Henderson, R.K., Baker, A., Murphy, K.R., Hamblya, A., Stuetz, R. M., Khan, S.J., 2009. Fluorescence as a potential monitoring tool for recycled water systems: a review. Water Res. 43 (4), 863e881. Hernandez, D., Plaza, C., Senesi, N., Polo, A., 2006. Detection of copper(II) and zinc(II) binding to humic acids from pig slurry and amended soils by fluorescence spectroscopy. Environ. Pollut. 143 (2), 212e220. Huang, J.H., Ilgen, G., Vogel, D., Michalzik, B., Hantsch, S., Tennhardt, L., Bilitewski, B., 2009. Emissions of inorganic and organic arsenic compounds via the leachate pathway from
1719
pretreated municipal waste materials: a landfill reactor study. Environ. Sci. Technol. 43 (18), 7092e7097. Huo, S.L., Xi, B.D., Yu, H.C., Liu, H.L., 2009. Dissolved organic matter in leachate from different treatment processes. Water Environ. J. 23 (1), 15e22. Li, R., Yue, D.B., Liu, J.G., Nie, Y.F., 2009. Size fractionation of organic matter and heavy metals in raw and treated leachate. Waste Manage. 29 (9), 2527e2533. Lu¨, F., Chang, C.H., Lee, D.J., He, P.J., Shao, L.M., Su, A., 2009. Dissolved organic matter with multi-peak fluorophores in landfill leachate. Chemosphere 74 (4), 575e582. Luster, J., Lloyd, T., Sposito, G., Fry, I.V., 1996. Multi-wavelength molecular fluorescence spectrometry for quantitative characterization of copper(II) and aluminum(III) complexation by dissolved organic matter. Environ. Sci. Technol. 30 (5), 1565e1574. Martensson, A.M., Aulin, C., Wahlberg, O., Agren, S., 1999. Effect of humic substances on the mobility of toxic metals in a mature landfill. Waste Manage. Res. 17 (4), 296e304. Michalzik, B., Ilgen, G., Hertel, F., Hantsch, S., Bilitewski, B., 2007. Emissions of organo-metal compounds via the leachate and gas pathway from two differently pre-treated municipal waste materials e a landfill reactor study. Waste Manage. 27 (4), 497e509. Ohno, T., Amirbahman, A., Bro, R., 2008. Parallel factor analysis of excitation-emission matrix fluorescence spectra of water soluble soil organic matter as basis for the determination of conditional metal binding parameters. Environ. Sci. Technol. 42 (1), 186e192. Plaza, C., Brunetti, G., Senesi, N., Polo, A., 2006a. Fluorescence characterization of metal ion-humic acid interactions in soils amended with composted municipal solid wastes. Anal. Bioanal. Chem. 386 (7e8), 2133e2140. Plaza, C., Brunetti, G., Senesi, N., Polo, A., 2006b. Molecular and quantitative analysis of metal ion binding to humic acids from sewage sludge and sludge-amended soils by fluorescence spectroscopy. Environ. Sci. Technol. 40 (3), 917e923. Provenzano, M.R., D’Orazio, V., Jerzykiewicz, M., Senesi, N., 2004. Fluorescence behaviour of Zn and Ni complexes of humic acids from different sources. Chemosphere 55 (6), 885e892. Ryan, D.K., Weber, J.H., 1982. Fluorescence quenching titration for determination of complexing capacities and stability constants of fulvic acid. Anal. Chem. 54, 986e990. Stedmon, C.A., Bro, R., 2008. Characterizing dissolved organic matter fluorescence with parallel factor analysis: a tutorial. Limnol. Oceanogr. Methods 6, 572e579. Terbouche, A., Djebbar, S., Benali-Baitich, O., Bouet, G., 2010. Characterization and complexing capacity of humic acid extracted from Yakouren soil with heavy metals by conductimetry and quenching of fluorescence. Soil Sediment Contam. 19 (1), 21e41. Ward, M.L., Bitton, G., Townsend, T., 2005. Heavy metal binding capacity (HMBC) of municipal solid waste landfill leachates. Chemosphere 60 (2), 206e215. Yamashita, Y., Jaffe, R., 2008. Characterizing the interactions between trace metals and dissolved organic matter using excitation-emission matrix and parallel factor analysis. Environ. Sci. Technol. 42 (19), 7374e7379. Yang, R.J., van den Berg, C.M.G., 2009. Metal complexation by humic substances in seawater. Environ. Sci. Technol. 43 (19), 7192e7197. Zhao, J., Nelson, D.J., 2005. Fluorescence study of the interaction of Suwannee River fulvic acid with metal ions and Al3þ-metal ion competition. J. Inorg. Biochem. 99 (2), 383e396.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 2 0 e1 7 2 8
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Modeling high adsorption capacity and kinetics of organic macromolecules on super-powdered activated carbon Yoshihiko Matsui*, Naoya Ando, Tomoaki Yoshida, Ryuji Kurotobi, Taku Matsushita, Koichi Ohno Graduate School of Engineering, Hokkaido University, N13W8, Sapporo 060-8628, Japan
article info
abstract
Article history:
The capacity to adsorb natural organic matter (NOM) and polystyrene sulfonates (PSSs) on
Received 3 March 2010
small particle-size activated carbon (super-powdered activated carbon, SPAC) is higher
Received in revised form
than that on larger particle-size activated carbon (powdered-activated carbon, PAC).
15 November 2010
Increased adsorption capacity is likely attributable to the larger external surface area
Accepted 15 November 2010
because the NOM and PSS molecules do not completely penetrate the adsorbent particle;
Available online 24 November 2010
they preferentially adsorb near the outer surface of the particle. In this study, we propose a new isotherm equation, the Shell Adsorption Model (SAM), to explain the higher
Keywords:
adsorption capacity on smaller adsorbent particles and to describe quantitatively adsorp-
Diffusion
tion isotherms of activated carbons of different particle sizes: PAC and SPAC. The SAM was
Isotherm
verified with the experimental data of PSS adsorption kinetics as well as equilibrium. SAM
Shell
successfully characterized PSS adsorption isotherm data for SPACs and PAC simulta-
Equilibrium
neously with the same model parameters. When SAM was incorporated into an adsorption
Homogeneous
kinetic model, kinetic decay curves for PSSs adsorbing onto activated carbons of different
surface diffusion
particle sizes could be simultaneously described with a single kinetics parameter value. On
model (HSDM)
the other hand, when SAM was not incorporated into such an adsorption kinetic model and instead isotherms were described by the Freundlich model, the kinetic decay curves were not well described. The success of the SAM further supports the adsorption mechanism of PSSs preferentially adsorbing near the outer surface of activated carbon particles. ª 2010 Elsevier Ltd. All rights reserved.
1.
Introduction
It has been thought that adsorption capacity of activated carbon does not depend on particle size because adsorption occurs in internal pores of activated carbon particles (Letterman et al., 1974; Najm et al., 1990; Peel and Benedek, 1980a and Leenheer, 2007); however, the effect of adsorbent particle size on adsorption capacity has not been examined sufficiently. With decreasing activated carbon particle size, the capacity to adsorb natural organic matter (NOM) has been reported both to not change (Randtke and Snoeyink, 1983) and
to increase (Weber et al., 1983). Possible reasons for these contradictory results have been discussed, but a clear mechanism with supporting experimental evidence has not yet been presented. Recently, very fine (median particle diameters of 0.7 mm) activated carbon particles (super-powdered activated carbon, SPAC) became available through advances in pulverization technology (Matsui et al., 2004, 2005, 2007, 2009a). Thus, it has become possible to reduce adsorbent particle diameter to the submicron range, which is ten times as small as the size of powdered-activated carbon (PAC). SPAC adsorbs NOM
* Corresponding author. Tel./fax: þ81 11 706 7280. E-mail address:
[email protected] (Y. Matsui). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.11.020
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 2 0 e1 7 2 8
efficiently because of its high adsorption capacity as well as its kinetic properties. The high adsorption capacity of SPAC raises the issue of adsorption capacity dependence on adsorbent particle size. In the early stages of SPAC research, increased capacity for NOM adsorption on SPAC was thought to be attributable to the mesopore volume increase caused by the fracture of ink-bottle pore structures during pulverization (Matsui et al., 2004). However, it later became apparent that structural changes in pore size distribution attributable to pulverization were not large, and the capacity increase for NOM adsorption could not be explained adequately by an increase in mesopore volume. Instead, it has been proposed that the capacity increase for NOM adsorption on SPAC is due to adsorbate not penetrating completely into the adsorbent particle and preferentially adsorbing in the particle outer region close to the surface of the particle; we also proposed a conceptualization of this concept (Ando et al., 2010). The objective of our current research is to verify and confirm this conceptualization through model analysis. To do so, we have developed adsorption isotherm and kinetic models based on the shell adsorption mechanism, and verified our findings with experimental data by using polystyrene sulfonates (PSSs) as model substances (Karanfil et al., 1996a,b; Li et al., 2003a,b; and Ando et al., 2010).
2.
Materials and methods
2.1.
Activated carbons
Commercially available PAC (Taikou-W, Futamura Chemical Industries Co., Ltd., Gifu, Japan) was used as received (PAC-T) or pulverized in a bead mill (Metawater Co., Ltd., Tokyo, Japan) to achieve four degrees of pulverization; we designated these super-powdered activated carbons as SPACa-T, SPACb-T, SPACc-T, and SPACd-T, in increasing order of particle size. The pore size distributions and the scanning electron micrographs of SPACa-T and PAC-T are presented elsewhere (Ando et al., 2010). Slurries of each activated carbon were prepared in pure water and stored at 4 C and used after dilution and placement under vacuum. Particle size distributions of the five activated carbons were determined by using a laser-light scattering instrument (LA-700, Horiba, Ltd., Kyoto, Japan).
2.2.
1721
We dissolved the PSSs in ultrapure water after the addition of inorganic ions to adjust the ionic strength, and we adjusted the constituent inorganic ions and their concentrations to match those of natural water and the PSS waters used in previous experiments (Table 1S in the supplementary information, Matsui et al., 2004; Ando et al., 2010). All water samples were adjusted to pH 7.0 0.1 by the addition of HCl or NaOH, as required; they were then filtered through 0.2-mm membrane filters (DISMIC-25HP, Toyo Roshi Kaisha, Ltd., Tokyo) before use in experiments. We determined PSS concentrations by UV absorption at a wavelength of 262 nm (UV-1700, Shimadzu Co., Kyoto, Japan).
2.3.
Batch adsorption tests
We conducted PSS-1800 and PSS-1000 adsorption equilibrium tests with all five activated carbons, but we conducted PSS4600 adsorption equilibrium tests only with SPACa-T, SPACdT, and PAC-T. The experimental procedure, described in detail elsewhere (Ando et al., 2010), briefly is as follows. SPAC and PAC slurries were diluted, placed under vacuum, and added to 300-mL solutions containing adsorbate with mixing (Table 2S in the supplementary information). Aliquots (100 mL) were transferred from the 300-mL solutions to 125-mL vials, which were agitated on a shaker for 3 weeks at a constant temperature of 20 C. Control tests were also conducted that did not contain carbon to confirm that concentration changes during long-term mixing were negligible. After filtering the water samples through a 0.2-mm membrane filter, we measured the liquid-phase adsorbate concentrations. We investigated adsorption kinetics of SPACa-T, SPACd-T, and PAC-T by means of batch tests with efficient mixing. Sample water (3 L) containing PSSs was placed in a beaker, and an aliquot (50 mL) was withdrawn from the beaker to determine the initial PSS concentration. After the addition of a specified amount of an activated carbon suspension (Table 3S in the supplementary information), aliquots (50 mL) were withdrawn at intervals and filtered immediately through a 0.2mm membrane filter for determination of PSS concentration.
3.
Results and discussion
3.1.
Shell adsorption model
Water samples
PSSs with various molecular weights (MWs) were selected as model substances instead of NOM because PSSs are chemically homogeneous compounds with known MWs and narrow MW ranges, while NOM is a complex mixture of compounds with unknown composition. Thus, our selection of PSSs makes model analysis of adsorption equilibrium and kinetics clear and unambiguous. We refer to our first PSS formulation as PSS-4600 (Polysciences, Inc., Warrington PA, USA), with a weight-average MW (Mw) of 5180 Da and a number-average MW (Mn) of 4600 Da. Our second PSS formulation, referred to as PSS-1800 (Polysciences, Inc.), had an Mw of 1430 Da and an Mn of 1200 Da. Our final PSS formulation, referred to as PSS1000 (Polymer Standard Service GmbH., Mainz, Germany), had an Mw of 1100 Da and an Mw/Mn of <1.2.
Clearly adsorption capacity for PSS-4600, PSS-1800, and PSS1000 on activated carbon (SPACa-T, SPACb-T, SPACc-T, SPACd-T, and PAC-T) increased with decreasing adsorbent particle size (Fig. 1 and Fig. 1S in the supplementary information). Adsorption sharply increased with increasing equilibrium concentration close to the initial concentration, in particular for PSS-1000. This could be due to the heterogeneity of PSS compounds (Karanfil et al., 1996a; Matsui et al., 1998), despite our assumption of homogeneity for the PSS compounds because of their small-MW ranges. Therefore, the data points for concentrations close to the initial concentration, indicated by red color in Fig. 1, were omitted from our mathematical analysis. Adsorption capacity for all three PSS formulations, as represented by q50, increased with decreasing adsorbent particle size (Fig. 2). PSS adsorption
1722
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 2 0 e1 7 2 8
100
10
100
10 0.1
1 10 Liquid-phase concentration, mg/L
1000
0.1
1 Liquid-phase concentration, mg/L
10
SPACa-T
PSS-1000
SPACb-T SPACc-T SPACd-T
100
PAC-T
Initial conc.
Solid-phase concentration, mg/g
PSS-1800
Initial conc.
Solid-phase concentration, mg/g
1000
PSS-4600
Initial conc.
Solid-phase concentration, mg/g
1000
by SAM equation by Freundlich equation
10 0.1
1 Liquid-phase concentration, mg/L
10
Fig. 1 e Adsorption isotherms of PSS-4600 (upper left panel), PSS-1800 (upper right panel), and PSS-1000 (lower panel). Lines are SAM and Freundlich fits to data (plots close to the initial concentration are red-colored) (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article).
capacity on SPACa-T (d50 ¼ 0.7 mm), which had the smallest particle size, was highest, followed by SPACb-T (d50 ¼ 1.1 mm), SPACc-T (1.9 mm), SPACd-T (3.0 mm), and PAC-T (11.8 mm), in increasing order of particle size: d50 is a volumetric median particle diameter, and q50 is defined as the amount adsorbed on activated carbon in equilibrium with 2.5 mg/L liquid-phase concentration equal to half the initial concentration (5 mg/L) in the adsorption experiment. Ando et al. (2010) hypothesize that the increase in adsorption capacity with decreasing adsorbent particle size is attributable to molecules adsorbing principally in the exterior region close to the external particle surface. The specific external surface area (surface area per unit mass) available for adsorption would be greater for
smaller adsorbent particles, and hence adsorption capacity could be larger on SPAC, which had a much smaller particle size than PAC. In adsorption isotherm model equations, such as the Freundlich equation, amount adsorbed is expressed as mass of adsorbate per unit mass of adsorbent (e.g., Sontheimer et al., 1988). This relationship implicitly assumes that adsorption surface area is proportional to mass of adsorbent and that adsorption capacity is independent of adsorbent particle size. In a previous study (Karanfil et al., 1996a), the Freundlich equation has been employed successfully to describe adsorption isotherms of PSSs, but the effect of adsorbent size was not studied. In our current research, we
Fig. 2 e PSS adsorption capacities represented by q50 against volumetric median diameters of adsorbents. q50 is defined as the amount adsorbed on activated carbon in equilibrium with 2.5 mg/L liquid-phase concentration equal to half of initial concentration (5 mg/L) in the adsorption experiment.
1723
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 2 0 e1 7 2 8
have modified the Freundlich equation, as per Eq. (1), in order to describe adsorption capacity dependence on adsorbent particle size, as follows: qE ¼ KCE1=n
where qE is the amount adsorbed in solid-phase in equilibrium with liquid-phase concentration (mg/g), CE is the liquid-phase concentration (mg/L), K is the Freundlich adsorption capacity parameter (mg/g)/(mg/L)1/n, and n is the Freundlich exponent. Beginning with the Freundlich approach, we have modeled the mechanism of Ando et al. (2010) such that the adsorption capacity parameter K is assumed to decrease with increasing distance from the adsorbent particle surface. Using radial coordinates, the Freundlich adsorption capacity parameter is a function of radial distance and particle radius; adsorption capacity of an adsorbent with radius R at radial distance r is then given by Eq. (2), as follows: 1=n
qS ðr; RÞ ¼ KS ðr; RÞCE
(2)
where r is the radial distance from the center of a PAC particle (mm), R is the adsorbent particle radius (mm), qS(r, R) is the local solid-phase concentration (mg/g) at radial distance r in an adsorbent with radius R, and KS(r, R) is the radially changing Freundlich adsorption capacity parameter (mg/g)/(mg/L)1/n as a function of radial distance r and adsorbent radius R. Spherical particles were assumed for the PAC and the SPACs, which is the conventional practice for adsorption kinetic models (e.g., Sontheimer et al., 1988). Therefore, adsorption capacity of an adsorbent with particle radius R in equilibrium with liquid-phase concentration CE is given by Eq. (3), as shown below: ZR
3r2 1=n 3 qS ðr; RÞ 3 dr ¼ CE 3 R R
0
ZR KS ðr; RÞr2 dr
(3)
0
Accordingly, when the adsorbent size is not uniform, the overall adsorption capacity of the adsorbent is given by Eq. (4), as follows qE ¼
1=n CE
3 R3
Adsorbent particle
(1)
3 2 ZN ZR 4 KS ðr; RÞr2 dr5fR ðRÞdR 0
(4)
0
where qE is the overall adsorption capacity of adsorbent (mg/ g), and fR(R) is the normalized particle size distribution function of adsorbent (mm1). As a model for KS(r, R), we adopted Eq. (5), in which adsorption capacity linearly decreases with distance from the external surface to a depth, d, but thereafter some of the adsorption capacity remains at a level, p, inward from that depth, as depicted in Fig. 3: rRþd ; 0 ð1 pÞ þ p KS ðr; RÞ ¼ K0 max d
(5)
where K0 is the Freundlich parameter of adsorption at the external particle surface (K0 means solid-phase concentration at r ¼ R at unity equilibrium concentration, (mg/g)/(mg/L)1/n), d is the penetration depth (or thickness of the penetration shell, mm), and p is a dimensionless parameter that defines availability of internal porous structures for adsorption.
1
p Radial distance
Fig. 3 e Schematic diagram of SAM. Molecules adsorb principally in the exterior region (black region in the figure) close to the particle surface, but to some extent do diffuse into the inner region (light gray region in the figure) of an adsorbent particle. KS(r, R)/K0, normalized adsorption capacity relative to the adsorption capacity at the external surface linearly decreases with distance from the external surface to a depth d, (from black to dark gray region) in the figure and thereafter (light gray region) it remains constant as p.
Eq. (5) evolved from the following reasoning: If adsorption occurs only at external particle surface, then adsorption capacity increase is inversely proportional to adsorbent particle size (slope of log q50 vs. log d50 ¼ 1). However, slopes for data points (Fig. 2) range only from 0.34 to 0.58 (less steep than 1), thereby indicating that some of the interior region of the adsorbent particles is available for adsorption. Some adsorbate molecules probably diffuse into and adsorb onto the interior region, while other molecules adsorb onto the exterior region close to the particle outer surface (shell region). The final form of the isotherm equation, referred to hereinafter as the Shell Adsorption Model (SAM) equation, is expressed as Eq. (6) qE ¼
8 9 ZN< ZR = rRþd max ; 0 ð1 pÞ þ p r2 dr fR ðRÞdR : ; R d
3K0 CE1=n 3
0
0
(6) We have applied this SAM equation to describe isotherm data shown in Fig. 1; in doing so, we sought a single set of isotherm parameter values for K0, n, d, and p in order to obtain the best model fit to data for PSS adsorption isotherms of SPACa-T, SPACb-T, SPACc-T, SPACd-T, and PAC-T. SAM satisfactorily described the experimental data, as shown in Figs. 1 and 2. Our SAM equation is a modified version of the Freundlich equation that is extended so that the slope in the logelog plot of solid-phase concentration vs. liquid-phase concentration is identical for each of the activated carbon preparations. However, experimentally measured slopes for SPACa-T and SPACb-T were actually slightly less steep than those for SPACd-T and PAC-T when applying the Freundlich
1724
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 2 0 e1 7 2 8
equation to data for each activated carbon (see dashed lines in Fig. 1 and Table 1). Because the change in slope after pulverization was not very marked, however, we feel that the SAM approach was successful in providing a first estimate of the dependence of adsorption capacity on particle size.
3.2.
Adsorption kinetics in the shell adsorption model
We analyzed adsorption kinetics data to determine whether incorporation of the SAM equation into an adsorption kinetic model adequately describes the kinetics data. In combining the kinetic model with SAM, we used the pore diffusion model (PDM, e.g., Sontheimer et al., 1988). Although the homogeneous surface diffusion model (HSDM) is more widely used than PDM (Sontheimer et al., 1988), we felt that it could not be applied because it assumes homogeneity inside activated carbon particles. Such homogeneity implies that adsorbed molecules have migrated into adsorbent particles by Fick’s first law of diffusion according to a local solid-phase concentration gradient, and that adsorbate molecules are ultimately distributed evenly along an adsorbent gradient such that local solid-phase concentrations become equal. Such a scenario is inconsistent with SAM. Therefore, instead of HSDM, we used PDM in which migration of molecules in the liquid-filled pores
contributes to transport of adsorbates into particles, while local solid-phase and liquid-phase concentrations in pores remain at equilibrium during the entire period of adsorption (instantaneous adsorption). At an adsorption equilibrium condition in PDM, local liquid-phase concentrations become equal, while local solid-phase concentrations do not necessarily become equal; that condition does not violate SAM. Local adsorption equilibrium is expressed by Eq. (7), as follows: cðt; r; RÞ ¼
n qðt; r; RÞ KS ðr; RÞ
(7)
where t is adsorption time in the batch system (s); c(t, r, R) is the liquid-phase concentration in an adsorbent particle having radius R at radial distance r and time t (mg/L); and q(t, r, R) is the solid-phase concentration in an adsorbent particle having radius R at radial distance r and time t (mg/g). Diffusion of adsorbate molecules in an adsorbent particle is expressed by Eq. (8), as follows vqðt; r; RÞ DP 1 v 2 vcðt; r; RÞ ¼ r 2 r r vr vt vr
(8)
where DP is the pore diffusion coefficient (cm2/s); and r is adsorbent particle density (g/L).
Table 1 e Equilibrium and kinetic parameters and ENS values. PSS-4600
Simulation 1
Adsorption equilibrium
SAM K0 ¼ 1.8 102 (mg/g)/(mg/L)1/n 1/n ¼ 0.10 d ¼ 0.22 mm p ¼ 0.038
Adsorption kinetics
PDM DP ¼ 2.9 1010 cm2/s 0.25
ENS
PSS-1800
Simulation 1
Adsorption equilibrium
SAM K0 ¼ 3.2 102 (mg/g)/(mg/L)1/n 1/n ¼ 0.15 d ¼ 0.20 mm p ¼ 0.095
Adsorption kinetics
PDM DP ¼ 7.6 1010 cm2/s 0.85
ENS
PSS-1000
Simulation 1
Adsorption equilibrium
SAM K0 ¼ 2.8 102 (mg/g)/(mg/L)1/n 1/n ¼ 0.21 d ¼ 0.21 mm p ¼ 0.18
Adsorption kinetics
PDM DP ¼ 11.0 1010 cm2/s 0.40
ENS
Simulation 2
Simulation 3
Freundlich K (SPACa-T) ¼ 110 (mg/g)/(mg/L)1/n 1/n (SPACa-T) ¼ 0.064 K (SPACd-T) ¼ 39 (mg/g)/(mg/L)1/n 1/n (SPACd-T) ¼ 0.26 K (PAC-T) ¼ 18 (mg/g)/(mg/L)1/n 1/n (PAC-T) ¼ 0.27 HSDM PDM DS ¼ 3.3 1013 cm2/s DP ¼ 1.7 109 cm2/s 2.1 0.49
Simulation 2
Simulation 3
Freundlich K (SPACa-T) ¼ 190 (mg/g)/(mg/L)1/n 1/n (SPACa-T) ¼ 0.11 K (SPACd-T) ¼ 85 (mg/g)/(mg/L)1/n 1/n (SPACd-T) ¼ 0.28 K (PAC-T) ¼ 45 (mg/g)/(mg/L)1/n 1/n (PAC-T) ¼ 0.28 HSDM PDM DS ¼ 2.1 1013 cm2/s DP ¼ 3.2 109 cm2/s 0.11 0.71
Simulation 2
Simulation 3
Freundlich K (SPACa-T) ¼ 190 (mg/g)/(mg/L)1/n 1/n (SPACa-T) ¼ 0.16 K (SPACd-T) ¼ 97 (mg/g)/(mg/L)1/n 1/n (SPACd-T) ¼ 0.27 K (PAC-T) ¼ 67 (mg/g)/(mg/L)1/n 1/n (PAC-T) ¼ 0.28 HSDM PDM DS ¼ 1.5 1013 cm2/s DP ¼ 3.3 109 0.84 0.083
cm2
/s
1725
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 7 2 0 e1 7 2 8
Cummurative percentage undersize
By particle size measurement
Discrete approximation
100 80 SPACa-T 60 SPACd-T
40
PAC-T
20 0
0.1
1
10
100
Particle diameter, µm
Fig. 4 e Particle size distributions of SPACa-T, SPACd-T, and PAC-T.
External film balance is described by equating mass balance and mass transfer from the external particle surface to inside the particle, as shown in Eq. (9): 3 2 R Z d4 1 kf 2 r qðt; r; RÞdr5 2 ¼ ½CðtÞ cðt; R; RÞ r dt R
(9)
0
where kf is the liquid film mass transfer coefficient (cm/s), r is the adsorbent particle density (g/L), and C(t) is the adsorbate concentration in the bulk water phase as a function of time, t (mg/L). When considering adsorbent particle size distribution (Matsui et al., 2003), the mass balance equation for an adsorbate in a batch reactor is given in Eq. (10), as follows: dCðtÞ 3CC kf ¼ r dt
ZN
fR ðRÞ ½CðtÞ cðt; R; RÞdR R
(10)
0
We approximated particle size distribution of adsorbent by a discrete density function consisting of M size classes, where M is 13, as shown in Fig. 4. We converted the set of model Eqs (5) and (7)e(10) for adsorption in a batch reactor into a set of ordinary differential equations with respect to time, t, using the method of orthogonal collocation. We took many collocation points in an attempt to describe precisely the change of solid-phase concentration in the vicinity of the particle surface (shell region in Fig. 3). When the number of collocation points was 40, the shell region of a PAC particle 11.8 mm in
2 PN j¼1 Cobs;j Ccal;j ENS ¼ 1 PN 2 j¼1 Cobs;j Cave
1.0
SPACd-T Remaining ratio
0.8 0.6 0.4
0.8
