WATER RESEARCH A Journal of the International Water Association
Editor-in-Chief Mark van Loosdrecht Delft University of Technology Department of Biochemical Engineering Julianalaan 67 2628 BC Delft The Netherlands Tel: +31 15 27 81618 E-mail:
[email protected]
Editors J. Block Université H. Poincaré, Nancy I France David Dixon University of Melbourne Australia Hiroaki Furumai The University of Tokyo Japan Xiaodi Hao Beijing University of Civil Engineering and Architecture China Gregory Korshin University of Washington USA Anna Ledin Formas Sweden Eberhard Morgenroth Swiss Federal Institute of Aquatic Science and Technology (EAWAG) Switzerland W. Rauch University Innsbruck Austria Maria Reis Universidade Nova de Lisboa/FCT Portugal Hang-Shik Shin Korea Advanced Institute of Science and Technology Korea Thomas Ternes Bundesanstalt für Gewässerkunde Germany Stefan Wuertz Univ. of California, Davis USA
Associate Editors Andrew Baker University of New South Wales Australia
Damien Batstone The University of Queensland Australia G-H. Chen The Hong Kong University of Science & Technology Hong Kong China Tom Curtis Univ. of Newcastle upon Tyne UK Ana Deletic Monash University USA Francis de los Reyes III North Carolina State University USA Rob Eldridge The University of Melbourne Australia Rosina Girones University of Barcelona Spain Stephen Gray Victoria University Australia Kate Grudpan Chiang Mai University Thailand E.E. Herricks University of Illinois - Urbana USA Peter Hillis United Utilities Plc UK H-Y. Hu Tsinghua University China P.M. Huck University of Waterloo Canada Bruce Jefferson Cranfield University UK Ulf Jeppsson Lund University Sweden Sergey Kalyuzhnyi Moscow State University Russian Federation Jaehong Kim Georgia Institute of Technology USA Jes La Cour Jansen Lund Institute of Technology Sweden G. Langergraber BOKU/Univ. of Natural Res. and Applied Life Scs. Austria S-L. Lo National Taiwan University Taiwan Dionisis Mantzavinos Technical University of Crete Greece
Y. Matsui Hokkaido University Japan A. Maul Université Paul Verlaine-Metz France Max Maurer EAWAG Switzerland How Yong Ng National University of Singapore Singapore Satoshi Okabe Hokkaido University Japan S.L. Ong National University of Singapore Singapore Jong M. Park Pohang University of Science & Technology Korea Susan Richardson U.S. Environmental Protection Agency USA Miguel Salgot University of Barcelona Spain David Sedlak University of California, Berkeley USA Jean-Philippe Steyer LBE-INRA France M. Takahashi Hokkaido University Japan Kai Udert EAWAG Switzerland V.P. Venugopalan BARC Facilities India E. von Sperling Federal University of Minas Gerais Brazil Hanqing Yu University of Science & Technology of China China J. Zilles University of Illinois Urbana USA A.I. Zouboulis Aristotle University Greece
Editorial Office Mailing address Water Research Radarweg 29 1043 NX Amsterdam The Netherlands E-mail:
[email protected]
Publication information: Water Research (ISSN 0043-1354). For 2011, volume 45 is scheduled for publication. Subscription prices are available upon request from the publisher or from the Elsevier Customer Service Department nearest you or from this journal’s website (http://www.elsevier.com/locate/watres). Further information is available on this journal and other Elsevier products through Elsevier’s website (http://www.elsevier.com). Subscriptions are accepted on a prepaid basis only and are entered on a calendar year basis. Issues are sent by standard mail (surface within Europe, air delivery outside Europe). Priority rates are available upon request. Claims for missing issues should be made within six months of the date of dispatch. Author Enquiries: For enquiries relating to the submission of articles (including electronic submission) please visit this journal’s homepage at http://www.elsevier.com/ locate/watres. Contact details for questions arising after acceptance of an article, especially those relating to proofs, will be provided by the publisher. You can track accepted articles at http://www.elsevier.com/trackarticle. You can also check our Author FAQs at http://www.elsevier.com/authorFAQ and/or contact Customer Support via http://support.elsevier.com. Orders, claims, and journal enquiries: Please contact the Elsevier Customer Service Department nearest you: St. Louis: Elsevier Customer Service Department, 3251 Riverport Lane, Maryland Heights, MO 63043, USA; phone: (877) 8397126 [toll free within the USA]; (+1) (314) 4478878 [outside the USA]; fax: (+1) (314) 4478077; e-mail:
[email protected] Oxford: Elsevier Customer Service Department, The Boulevard, Langford Lane, Kidlington OX5 1GB, UK; phone: (+44) (1865) 843434; fax: (+44) (1865) 843970; e-mail:
[email protected] Tokyo: Elsevier Customer Service Department, 4F Higashi-Azabu, 1-Chome Bldg, 1-9-15 Higashi-Azabu, Minato-ku, Tokyo 106-0044, Japan; phone: (+81) (3) 5561 5037; fax: (+81) (3) 5561 5047; e-mail:
[email protected] Singapore: Elsevier Customer Service Department, 3 Killiney Road, #08-01 Winsland House I, Singapore 239519; phone: (+65) 63490222; fax: (+65) 67331510; e-mail:
[email protected] Application for membership of International Water Association should be made to: Executive Director, IWA, Alliance House, 12 Caxton Street, London SW1H 0QS, U.K. (Tel.: +44 207 654 5500; Fax: +44 207 654 5555; e-mail:
[email protected]; website: http://www.IWAhq.org.uk). Registered Charity (England) No. 289269. Individual membership is available from £30 upwards. For details contact IWA.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 3 9 e2 4 5 1
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Removal of trace organics by MBR treatment: The role of molecular properties Nichanan Tadkaew a, Faisal I. Hai a, James A. McDonald b, Stuart J. Khan b, Long D. Nghiem a,* a
The Strategic Water Infrastructure Laboratory, School of Civil Mining and Environmental Engineering, The University of Wollongong, NSW 2522, Australia b Water Research Centre, The University of New South Wales, NSW 2552, Australia
article info
abstract
Article history:
This study examined the relationship between specific molecular features of trace organic
Received 15 September 2010
contaminants and their removal efficiencies by a laboratory scale membrane bioreactor
Received in revised form
(MBR). Removal efficiencies of 40 trace organic compounds were assessed under stable
27 January 2011
operating conditions. The reported results demonstrate an apparent correlation between
Accepted 28 January 2011
chemical structures and the removal of trace organic contaminants by the laboratory scale MBR system. The removal of all 14 very hydrophobic (Log D > 3.2) trace organic compounds selected in this study was consistently high and was above 85%. The occurrence and types
Keywords:
of electron withdrawing or donating functional groups appear to be important factors
Membrane bioreactor (MBR)
governing their removal by MBR treatment. In this study, all hydrophilic and moderately
Trace organic contaminants
hydrophobic (Log D < 3.2) compounds possessing strong electron withdrawing functional
Sorption
groups showed removal efficiency of less than 20%. In contrast, high removal efficiencies
Biodegradation
were observed with most compounds bearing electron donating functional groups such as
Hydrophobicity
hydroxyl and primary amine groups. A qualitative framework for the assessment of trace
Molecular structure
organic removal by MBR treatment was proposed to provide further insights into the removal mechanisms. Crown Copyright ª 2011 Published by Elsevier Ltd. All rights reserved.
1.
Introduction
Major driving forces towards water recycling today are the growing demand for water from an increasing population, changing lifestyle patterns, urbanisation, and diminishing natural water resources. In addition, better public awareness about environmental protection has resulted in progressively more stringent wastewater quality discharge regulations. Despite the growing interest in water recycling, our predictive capacity regarding the ability of treatment technologies to remove specific trace organic contaminants remains very limited. This is reflected by the public reluctance to accept reclaimed water for potable reuse and the fact that most water
recycling applications are currently still restricted to nonpotable purposes. Membrane bioreactors (MBRs) have recently emerged as an important technology for water recycling, capable of transforming wastewater to high quality effluent suitable for various water recycling applications (Atkinson, 2006). Becoming commercially available only around two decades ago, MBR technology has already been well proven and can provide a superior rating for most bulk water quality indicators such as pathogens, suspended solids and nutrient removal compared to conventional activated sludge (CAS) treatment processes (Melin et al., 2006; Visvanathan et al., 2000). However, the efficiency of MBR technology as a barrier for a range of trace organic
* Corresponding author. Tel.: þ61 2 4221 4590. E-mail address:
[email protected] (L.D. Nghiem). 0043-1354/$ e see front matter Crown Copyright ª 2011 Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.01.023
2440
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 3 9 e2 4 5 1
contaminants such as endocrine disrupting chemicals (EDCs), pesticides, and pharmaceutically active compounds (PhACs), as well as the specific removal mechanisms involved remain unclear (Clara et al., 2005; De Wever et al., 2007; Kimura et al., 2005; Qu et al., 2009; Visvanathan et al., 2005; Wintgens et al., 2004). Previous studies have indicated significant variation in the removal of trace organics by MBRs, ranging from near complete removal for some compounds (e.g. ibuprofen and bezafibrate) to almost no removal for several others (e.g. carbamazepine and diclofenac) (Clara et al., 2005; Kimura et al., 2005; Tadkaew et al., 2010; Urase et al., 2005). The reasons for such variation are not yet fully understood. Physicochemical properties of trace organics have been reported to significantly govern their removal efficiency by MBR treatment. Biosorption of trace contaminants, driven primarily by hydrophobic interaction, appears to be one of the key mechanisms controlling removal efficiency during MBR treatment. For instance, apparent improvement in removal efficiency of certain acidic trace organics such as ibuprofen, ketoprofen, and diclofenac has been observed when MBRs are operated under acidic conditions rather than neutral conditions (Tadkaew et al., 2010; Urase et al., 2005). This phenomenon was explained by the speciation of the compounds from hydrophilic ionic forms to much more hydrophobic forms at pH lower than their acid dissociation (pKa) values. A limited number of studies has shed some light on the effect of chemical structures on the removal efficiency of trace chemicals during biological treatment processes. For example, Kimura et al. (2005) attributed the poor removal of clofibric acid, diclofenac, and dichloprop to the presence of chlorine in their molecular structure or their relatively complicated aromatic rings. Several studies have utilised the US-EPA-developed Biodegradation Probability Program for Windows (BIOWIN) software package commissioned by the US-EPA which is one of the most widely used computer-based quantitative structure biodegradability relationship (QSBR) programs to estimate the biodegradability of organic compounds under aerobic conditions. Lapertot and Pulgarin (2006) investigated the biodegradability of 17 priority hazardous substances and suggested that the primary and ultimate BIOWIN models were generally suitable for removal assessment of these compounds in industrial wastewater treatment processes. On the other hand, Yu et al. (2006) reported some inconsistency between the likelihood of biodegradability predicted by BIOWIN and experimental data when they investigated the removal efficiency of 18 pharmaceutical and personal care products at a conventional municipal wastewater treatment plant. Although the connection between chemical structure and removal efficiency seems highly plausible, studies to develop a capacity to predict the removal efficiency of trace organic contaminants by MBR treatment processes based on a range of molecular parameters are still limited. Because of the involvement of many diverse and complex functional groups, the connection between chemical structure and removal efficiency has not yet been thoroughly examined in the literature. In fact, several previous attempts to identify a definitive relationship between the structures of trace organic contaminants and their removal efficiencies during CAS and MBR treatment have identified significant challenges (Joss et al., 2005; Radjenovic et al., 2007).
This study aimed to elucidate the connection between specific molecular features of trace organic contaminants and their removal efficiencies by a laboratory scale MBR. The MBR system was operated under stable conditions for an extended period to allow for a systematic examination of the removal of 40 trace organic contaminants at environmentally relevant concentrations. Hydrophobicity and molecular structures of the selected trace organic compounds were carefully delineated and correlated to their removal efficiencies. Key factors governing the removal efficiencies of trace organic contaminants were identified and reported.
2.
Materials and methods
2.1.
Laboratory scale MBR system
A laboratory scale MBR system was used in this study. Detailed description of this MBR system is available elsewhere (Tadkaew et al., 2010). The system consisted of a glass reactor, a continuous mixer, two air pumps, a pressure sensor, and influent and effluent pumps. Two ZeeWeed-1 (ZW-1) submerged hollow fibre ultrafiltration membrane modules supplied by Zenon Environmental (Ontario, Canada) were used in this set-up. The membrane has a nominal pore size of 0.04 mm. Each module has an effective membrane surface area of 0.047 m2. A chiller (Neslab RTE 7) equipped with a stainless steel heat exchanging coil was used to maintain a constant temperature in the MBR. A personal computer was used to control the permeate peristaltic pump to operate on a 14 min suction and 1 min off cycle to provide relaxation time to the membrane modules. Flow rate of the influent pump was matched with that of the permeate pump to maintain a constant reactor volume. A mixer was continuously used to ensure homogeneous conditions of the mixed liquor and to prevent the settling of biomass.
2.2.
Synthetic wastewater
A synthetic wastewater simulating municipal sewage was used to ensure a stable feeding rate throughout the experiment. Concentrated stock solution was prepared and stored in a refrigerator at 4 C. It was then diluted with MilliQ water on a daily basis to make up a feed solution containing glucose (400 mg/L), peptone (75 mg/L), KH2PO4 (17.5 mg/L), MgSO4 (17.5 mg/L), FeSO4 (10 mg/L), and sodium acetate (225 mg/L). This composition was based on a previous study (Zhang et al., 2006).
2.3.
Trace organic compounds
In this study, 40 organic compounds were selected to represent four major trace organic groups of concern in water reuse applications e namely pesticides, pharmaceutically active compounds, steroid hormones, and other endocrine disrupting chemicals. The selection of these model trace organic compounds was also based on their widespread occurrence in domestic sewage and their diverse physicochemical properties (e.g. hydrophobicity and molecular weight). The effective hydrophobicity of these compounds varies significantly as reflected by their Log D values at pH 8 (see Supplementary data)
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 3 9 e2 4 5 1
which is typical of an activated sludge reactor (Wells, 2006). The most hydrophilic compound is enalapril with Log D at pH 8 of 1.21 and the most hydrophobic compound is nonylphenol with Log D at pH 8 of 6.19. All selected trace organic compounds were of analytical grade. A combined stock solution was prepared in pure acetonitrile. The trace organic stock solution was kept in a freezer and was used within less than a month.
2.4.
Analytical techniques
The analysis of the model trace organics was based on a previously reported method (Tadkaew et al., 2010; Vanderford and Snyder, 2006). Analytes were extracted using 5 mL, 500 mg solid phase extraction hydrophilic/lipophilic balance (HLB) cartridges (Waters, Millford, MA, USA). Samples were spiked with a solution containing 50 ng of an isotopically labelled version of each analyte. The sample was then loaded onto the cartridges at 15 mL/min, after which the cartridges were rinsed with 5 mL of reagent water and dried with a stream of nitrogen for 30 min. Loaded cartridges were stored at 4 C in sealed bags until elution and analysis. Analytes were separated using an Agilent (Palo Alto, CA, USA) 1200 series high performance liquid chromatography (HPLC) system equipped with a 150 4.6 mm, 5 mm particle size, Luna C18 (2) column (Phenomenex, Torrence, CA, USA). Mass spectrometry was performed using an API 4000 triple quadrupole mass spectrometer (Applied Biosystems, Foster City, CA, USA) equipped with a turbo-V ion source employed in both positive and negative electro-spray ionisation modes. Steroid hormones were analysed using an atmospheric pressure chemical ionisation method and all other compounds were analysed using an electro-spray ionisation method. For each analyte and internal standard a precursor ion and two product ions were monitored for reliable confirmation. Relative retention times of the analyte and isotopically labelled internal standard were also monitored to ensure correct identification. Standard solutions of all analytes were prepared at 1, 5, 10, 50, 100, 500 and 1000 ng/mL. A relative response ratio of analyte/ internal standard over a 1e1000 ng concentration range was generated enabling quantification with correction for losses due to ion suppression and incomplete SPE recovery. All calibration curves had a correlation coefficient of 0.99 or better. The limit of reporting was determined using an s/n ratio of greater than 10. Conductivity and pH were measured using an Orion 4-Star Plus pH/conductivity metre. Total organic carbon (TOC) and total nitrogen (TN) were analysed using a Shimadzu TOC/TNVCSH analyser. TOC analysis was conducted in non-purgeable organic carbon (NPOC) mode. Samples were kept at 4 C until analysed and calibrations were performed in the range between 0 and 1000 mg/L and 0e100 mg/L for TOC and TN, respectively. Mixed liquor suspended solid (MLSS) and mixed liquor volatile suspended solid (MLVSS) contents in the MBR were measured in accordance to the Standard Methods for the Examination of Water and Wastewater (Eaton et al., 2005).
2.5.
MBR experimental protocol
The MBR was seeded with activated sludge from the Wollongong sewage treatment plant, NSW, Australia. After the initial
2441
start-up process, which lasted about 2 months, a small amount of sludge was regularly extracted from the reactor to keep the sludge age at approximately 70 days. The hydraulic retention time was set at 24 h, corresponding to a permeate flux of 4.3 L/m2 h. The MBR temperature and dissolved oxygen content were kept constant at 20.0 0.1 C and 2 1 mg/L, respectively. Performance of the MBR system with regard to basic water quality parameters was then monitored for an extended period of more than four weeks. Once stable operation had been achieved, trace organic contaminants were continuously introduced into the feed solution to make up a concentration of approximately 2 mg/L of each selected compound. The investigation with trace organics was conducted over a period of four weeks during which no sludge was withdrawn from the reactor. The feed solution was kept in a stainless steel reservoir at controlled room temperature (20 2 C). Feed and permeate samples were taken twice a week in duplicate and solid phase extraction was conducted immediately for subsequent trace organic analysis. Removal efficiency was calculated as R ¼ 100 ð1 CEff =CInf Þ, where CEff and CInf are effluent (permeate) and influent concentrations (ng/L) of the trace organic compound, respectively. It is noted that complete degradation of an organic compound may follow different pathways and undergo several steps. Therefore, the term removal here does not necessarily indicate complete degradation of the trace organics, but rather a loss of the specific trace chemical molecule. In many cases, stable intermediates or ‘metabolites’ may be produced, but detailed consideration of these intermediates is beyond the scope of this study.
3.
Results and discussion
3.1.
Performance stability of the MBR
In this study, synthetic feed solution was used to ensure a consistent influent composition. The MBR showed stable and good performance with respect to all key water quality parameters. The stable performance continued even following the introduction of the trace organic contaminants to the feed solution. A notable exception, however, was a significant decline in the removal of total nitrogen (TN) immediately after the introduction of the trace organic contaminants from almost complete removal to as low as 60%. The decrease in TN removal can be explained by the introduction of acetonitrile, the solvent used to introduce the trace organics, to the influent. The MBR system used in this study was operated under aerobic conditions and therefore is not expected to have any biological denitrification capacity. The original synthetic feed solution was deficient in nitrogen, and therefore, the initial high TN removal observed here could be attributed to the conversion of dissolved organic nitrogen to biomass, which would then be retained by the membrane. Because acetonitrile was used as a carrying solvent for the introduction of trace organic contaminant cocktail into the feed solution, the introduction of trace organic contaminants into the feed solution resulted in a significant increase in TN in the influent from 12 mg/L to approximately 49.5 mg/L. This was assumed to be the main reason for the observed decrease
2442
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 3 9 e2 4 5 1
in TN removal. The increase in nitrogen content of the feed water did not exert any discernible impact on any other biological performance indicators of the MBR system. There was a slight increase in the MLSS content in the reactor from 8.6 g/L to 10.0 g/L over the duration of the experiment of approximately one month while the MLVSS/MLSS ratio remained constant at approximately 0.9. Other basic performance parameters including TOC removal efficiency (98%), pH of the MLSS (7.5 0.1), effluent conductivity (559 19 mS/cm) were also relatively stable during the entire experiment. In addition, no abnormal transmembrane pressure increase was observed following the introduction of trace contaminants to the feed solution (data not shown). Stable performance of the MBR system could also be observed with respect to the removal of trace organic contaminants (Fig. 1). It is noted that the error bars shown in Fig. 1 represent the standard deviations of eight influent and effluent samples, regularly collected in duplicate throughout the experiment. It is also notable that the removal efficiencies of the 40 compounds investigated in this study vary significantly ranging from negligible removal (e.g. atrazine,
carbamazepine, dilatin, and trimethoprim) to removal to below the analytical detection limit (e.g. 17b-oestradiol, testosterone, and triclocarban), indicating a removal of at least 98%. The observed significant variation in the removal efficiency of the trace organic contaminants by MBR treatment indicates that improved understanding of the key factors that govern the elimination of specific chemicals is required to enable prediction of MBR treatment performance for any particular chemical or class of chemicals.
3.2.
Removal of trace organic contaminants
A logical approach to qualitatively predict the effectiveness of MBR treatment for the removal of a wide range of trace organic contaminants is to evaluate their removal efficiency according to the intended applications or origins of these compounds. Accordingly, Table 1 summarises the removal efficiencies of the 40 compounds selected in this study. Data previously reported in other studies, whenever available, are also included for comparison purposes. With caffeine being the only noteworthy exception, results reported here are in
4000
Influent
Effluent
3500
Concentration (ng/L)
3000 2500 2000 1500 1000 500
Bisphenol-A Nonylphenol t-octylphenol
Other EDC's
Estrone 17b-estradiol Androstenedione Estriol Steroid/hormone Testosterone Etiocholanolone Androsterone 17a-ethynylestradiol
Other drug
Atenolol Verapamil Enalapril
Cardiovascular drug
Triamterene Hydroxyzine Meprobamate Caffeine Omeprazole
Simvastatin Gemfibrozil Sim-hydroxyacid
Hypolipidemic agent
Triclosan Triclocarban Sulfamethoxazole Trimethoprim
Antibiotic/ antiseptic
Antidepressant drug
2
Clozapine Risperidone Primidone Carbamazepine Dilantin Amitriptyline
Paracetamol Ketoprofen Naproxen Ibuprofen Diclofenac
Anti-inflammatory drug
Pesticide
Atrazine Linuron DEET
0
Fig. 1 e Influent and effluent concentrations of the selected trace organic contaminants. Samples were collected twice a week and in duplicate for four weeks. Error bars represent the standard deviation of 16 measurements.
2443
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 3 9 e2 4 5 1
Table 1 e Removal efficiencies of the selected trace organic contaminants (n [ 16) obtained in this investigation and corresponding values recorded in the literature. Class Pesticides
Non-steroidal anti-inflammatory
Anti-depressants & mood stabilisers
Antibiotic & antiseptic
Compound
Cardiovascular drugs
Other drugs
Steroid hormones
Other EDCs
Literature (%) (minemax)
Atrazine Linuron DEET
4.4 3.7 21.1 4.1 4.6 2.4
9e40 Not available 0e78
Paracetamol
95.1 3.4
99
Ketoprofen
70.5 0.8
43.9e95
Naproxen
40.1 2.8
36e91.6
Ibuprofen
96.7 0.7
90
Diclofenac
17.3 4.2
0e87.4
Clozapine Risperidone Primidone Carbamazepine
84.8 95.8 12.4 13.4
5.4 2.2 4.3 4.3
Not available Not available Not available 0e13
5.4 3.6 97.8 0.8 >91.8 >98.4 91.9 0.6
0e12 Not available 61e95 Not available 52e80.8
Dilantin Amitriptyline Triclosan Triclocarban Sulfamethoxazole
Trimethoprim Hypolipidemic agents
This study (%) (Average Std)
16.6 3.7
0e90
Simvastatin Gemfibrozil Sim-hydroxyacid Atenolol
97.9 98.95 59.6 96.9
0.9 0.1 2.8 0.2
Not available 32.5e90 Not available 70
Verapamil Enalapril Triamterene Hydroxyzine Meprobamate Caffeine Omeprazole Oestrone 17b-estradiol Androstenedione Oestriol Testosterone Etiocholanolone Androsterone 17a-ethynylestradiol Bisphenol A
88.4 97.1 27.9 >92.2 14.5 49.6 62.1 98.0 >99.4 >99.5 98.2 >99.4 >99.4 >99.3 93.5 90.4
6.1 0.1 6.3
Not available Not available Not available Not available Not available 98e99 Not available 96.3 100 Not available >99 Not available Not available Not available 81.9e93.6 68.9e99.0
3.3 4.1 3.5 0.2
1.9
1.2 3.1
Nonyphenol
99.3 0.2
0e88
t-octylphenol
94.5 1.1
44.9e99.0
References Bernhard et al. (2006), Bouju et al. (2008) Bernhard et al. (2006), Kim et al. (2007), Snyder et al. (2007) Kim et al. (2007), Radjenovic et al. (2007, 2009), Joss et al. (2005) Radjenovic et al. (2007, 2009), Kimura et al. (2005), Quintana et al. (2005) Kim et al. (2007), Radjenovic et al. (2007, 2009), Joss et al. (2005), Quintana et al. (2005), Urase et al. (2005) Bernhard et al. (2006), Kim et al. (2007), Radjenovic et al. (2007, 2009), Joss et al. (2005), Quintana et al. (2005), Reif et al. (2008), Kreuzinger et al. (2004), Clara et al. (2005), Smook et al. (2008) Bernhard et al. (2006), Kim et al. (2007), Radjenovic et al. (2007, 2009), Quintana et al. (2005), Kreuzinger et al. (2004), Clara et al. (2005), Gonzalez et al. (2006), Abegglen et al. (2009)
Bernhard et al. (2006), Radjenovic et al. (2007), Joss et al. (2005), Reif et al. (2008), Clara et al. (2004, 2005) Snyder et al. (2007) Kim et al. (2007), Snyder et al. (2007) Kim et al. (2007), Radjenovic et al. (2007, 2009), Reif et al. (2008), Kreuzinger et al. (2004), Clara et al. (2005), Go¨bel et al. (2007) Kim et al. (2007), Radjenovic et al. (2009), Reif et al. (2008), Go¨bel et al. (2007) Radjenovic et al. (2007, 2009), Reif et al. (2008) Radjenovic et al. (2007), Reif et al. (2008), Quintana et al. (2005)
Kim et al. (2007), Snyder et al. (2007) Clara et al. (2005), Lyko et al. (2005) Clara et al. (2005), Lyko et al. (2005) Clara et al. (2005)
Lyko et al. (2005) Urase et al. (2005), Kreuzinger et al. (2004), Clara et al. (2005), Lyko et al. (2005), Chen et al. (2008) Kreuzinger et al. (2004), Clara et al. (2005), Cirja et al. (2006), Hu et al. (2007) Clara et al. (2005)
2444
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 3 9 e2 4 5 1
good agreement with the literature data. The mean removal efficiency of caffeine observed in our study is 49.6%, which is substantially lower than the previously reported values (Kim et al., 2007; Snyder et al., 2007). In a recent study, Santos et al. (2009) examined the performance of four CAS wastewater treatment plants in Seville city (Spain). They reported a highly variable caffeine removal efficiency among these four treatment plants with the mean value ranging from as low as 44% up to 75% (Santos et al., 2009). Given the similarity between MBR and CAS treatment, it is possible that this discrepancy can be explained by the differences in operating conditions. The literature data presented in Table 1 are from a range of sources with different operating conditions and system arrangements. The reported experimental results confirm that the MBR system used in this study behaved well within the range of typical performance data from other systems. Therefore, the results presented in this study and the conclusions drawn from them would be broadly applicable and generalisable to most typical MBR systems. In fact, data presented in Table 1 suggest that some generalisation can be made about certain groups of compounds. All the three pesticides investigated in this study showed very low removal efficiencies. Atrazine, a chloro-triazine herbicide, was removed at a rate of less than 5%. It has been reported to be poorly removed both in CAS and MBR (Bernhard et al., 2006) and that a major removal mechanism was sorption onto withdrawn sludge (Bouju et al., 2008). Linuron is a dichloro-phenylurea herbicide. Despite being a widely used herbicide, no reports on the removal of Linuron in CAS or MBR could be found. However, its slow natural attenuation rate in various soils and the evolution of more toxic and persistent chloroaniline intermediates in the process have been reported (Dejonghe et al., 2003). A mean removal of 21% of linuron as achieved in our MBR, therefore, appears to be consistent with the reported recalcitrance of this compound. DEET is a toluamide compound and is the most common active ingredient in insect repellants. In this study, a mean removal of 4.6% of DEET was recorded during MBR treatment. This removal efficiency is at the lower end of range reported in other published studies. Bernhard et al. (2006) reported nil to over 50% removal of DEET by MBR treatment and suggested that DEET removal efficiency was dependent on the sludge retention time (SRT). Kim et al. (2007) reported no removal of DEET in their study; however, no information about the SRT was provided. The highest removal efficiency of DEET of 78% was reported by Snyder et al. (2007) calculated from a one off sampling event at a pilot scale treatment facility. Near complete removal or removal to below the analytical limit of all eight steroid hormones and three other EDCs selected for investigation (bisphenol A, nonylphenol, and toctylphenol) were observed in this study. These results are consistent with other published studies (Table 1). It is noteworthy that all of these compounds possess significant hydrophobicity and bear similar molecular backbone structures; which may, in part, explain the similarities of their removal efficiencies. No generalistion can be inferred for any of the six therapeutic classes of pharmaceuticals investigated in this study (Table 1). Their removal efficiencies by MBR treatment vary widely even within the same class of compounds. The
removal efficiencies of the five non-steroidal anti-inflammatory drugs (NSAIDs) differ remarkably from one another. For example, ibuprofen registers a removal efficiency of 97% whereas the removal efficiency of diclofenac is only 17%. Unlike the other NSAIDs, diclofenac is a chlorinated compound, which can possibly explain its recalcitrant behaviour during MBR treatment. Significant variation in the removal efficiency can also be observed among compounds used as anti-depressants and mood stabilisers. Dilantin, primidone and carbamazepine were poorly removed, whereas the removal efficiencies of clozapine, risperidone, and amitriptyline were 85% and higher. Given the considerable dissimilarity in the molecular structure among these antidepressants and mood stabilisers, differences in their removal efficiencies can be expected. Further analysis of the molecular structures of these compounds is presented in Section 3.3.2. Significant variation in removal efficiency was observed among the other pharmaceutical groups (cardiovascular and other drugs) and can again be attributed to their diverse molecular structures (Table 1 and Supplementary data 1). Among the hypolipidemic agents (lipid lowering drugs) investigated in this study, simvastatin is a hydrophobic compound with Log D (at pH 8) of 4.41 and the compound registers a removal efficiency of 98% (Table 1). Simvastatin hydroxyacid shares the same molecular backbone structure with that of simvastatin. However, the 3, 5-dihydroxy-heptanoic acid functional group of simvastatin hydroxyacid renders the compound much more hydrophilic (Log D at pH 8 of 0.64). Consequently, simvastatin hydroxyacid shows a much lower removal efficiency of 60% in comparison to that of the related compound simvastatin. Results reported in Table 1 suggest that the classification of trace organics according to their intended use or origin can only be used to qualitatively predict the removal efficiencies of compounds of similar molecular structure, having similar molecular features or physicochemical properties. In fact, certain molecular features and physicochemical properties of the trace organic contaminants appear to be the underlying factors governing their rate of removal during MBR treatment.
3.3.
Role of molecular features
Attempts to fit the removal efficiency data obtained in our study and the corresponding available biodegradability scores from BIOWIN model did not result in any meaningful correlations (data not shown). Although this result is somewhat surprising, it does not necessarily invalidate the model. BIOWIN is essentially a statistical model and the discrepancies may have arisen to some extent due to the fact that the BIOWIN scores were derived from batch tests, which cannot effectively replicate the biological conditions of an MBR. It is also noteworthy that only three out of 40 compounds investigated in this study were included in the database which has been used for the development of BIOWIN. Furthermore, BIOWIN would not account for the adsorption of trace organics to biosolids which can be an important removal mechanism along with biodegradation. Given the poor correlation between the removal efficiencies experimentally obtained in this study and the BIOWIN biodegradability scores, it is necessary to further examine the key
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 3 9 e2 4 5 1
physicochemical properties and molecular features that can govern the removal efficiency of trace organic compounds.
3.3.1.
Effects of hydrophobicity
The removal of trace organic contaminants by an activated sludge treatment process is a complex function of both sorption and biological degradation. In a CAS treatment process, the sludge-bound contaminants can be subsequently removed via sludge withdrawal. In addition, sorption of trace organic contaminants to biosolids results in a longer residence time in the reactor, which may lead to further removal via biodegradation. Because the MLSS content and sludge retention time of typical MBR processes are much higher than those of CAS treatment, sorption has been suggested as a major removal mechanism for the removal of trace organic contaminants by MBR treatment. In a systematic survey of the literature data, Wells suggested that the sorption of a trace organic contaminant to the activated sludge could be assessed by considering the Log D value of the compound at a given pH (Wells, 2006). Experimental results presented in Fig. 2 indicate that this finding can be extended to MBR treatment. There appears to be a ‘removal envelop’ that can be defined by the hydrophobicity of the trace organic contaminants (Fig. 2). Removal of the very hydrophobic (Log D > 3.2) compounds is probably dominated by sorption to the activated sludge facilitating enhanced biological degradation in some cases. Therefore, these compounds consistently showed high removal efficiency (above 85%). As the Log D value of the compounds decreased to below 3.2, sorption of these trace organic contaminants onto the activated sludge was no longer a dominating removal mechanism and the removal efficiency of these compounds was much more strongly influenced by their intrinsic biodegradability. As a result, the removal efficiency of trace organics with low Log D values (at pH 8) varies significantly from less than 20% to removal to below the analytical detection limit (corresponding to a removal of at least 98%). Of particular note in Fig. 2 is a cluster of five compounds that show very low removal efficiencies despite their moderately high hydrophobicity (Log D in the rage from 2 to 3.2). It is also noteworthy that these five compounds possess one or several electron withdrawing functional
2445
groups, such as a chlorine atom or amide group. Results reported here suggest that individual molecular features can also be an important factor governing the removal efficiency of trace organics during MBR treatment.
3.3.2.
Effects of molecular weight
The molecular weights of the trace organics studied here ranged from 151 g/mol (paracetamol) to 455 g/mol (verapamil). There appears to be a weak but nevertheless discernible correlation between the removal efficiency of these trace organics and their molecular weights (Fig. 3). Compounds with molecular weight of more than 300 g/mol were relatively well removed (>60%), while the removal efficiencies of those with molecular weight of less than 300 g/mol varied from almost no removal to more than 98% (removal beyond the analytical detection limit). A plausible explanation for this observation could be the relative hydrophobicity (Log D at pH 8 in the range from 2.03 to 5.74, see Supplementary data) of the compounds having molecular weight of more than 300 g/mol. In addition, in this study, removal efficiency does not necessarily represent a complete mineralisation of the compound. Compounds with higher molecular weight may have more branches, which would offer more opportunities for the microbes to selectively cleave a certain target site and initiate degradation.
3.3.3.
Effect of chemical structure
Experimental results obtained in this study confirm the possible role of molecular functional groups in governing the removal of moderately hydrophobic and hydrophilic trace organic compounds by MBR treatment. The 40 trace organic compounds investigated in this study can be systematically categorised into three groups. Group A consists of compounds with Log D at pH 8 of above 3.2. As discussed above, sorption was a dominant removal mechanism for these hydrophobic compounds and the removal efficiencies of all compounds of group A were above 85% (Fig. 2). To further elucidate the role of different molecular features, the rest of the compounds can be categorised in terms of ring structure (heterocyclic/nonheterocyclic, mono or polynuclear) and functional groups (electron withdrawing/donating moieties). Fig. 4 shows the removal efficiency as a function of ring structure, whereas
Fig. 2 e The relationship of removal of trace organic compounds with effective hydrophobicity (Log D). The MLSS pH during the experiment was 7.5 ± 0.1. Log D values were obtained from the SciFinder Scholar (ACS) database. Error bars represent the standard deviation of 16 measurements.
2446
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 3 9 e2 4 5 1
100
60
40
20
0 200
250
300
350
400
450
500
Molecular weight, g/mol
Fig. 3 e Removal efficiency of trace organic compounds as a function of their molecular weight. Error bars represent the standard deviation of 16 measurements.
Fig. 5 presents the compounds under three distinct categories (B, C and C*) based on the presence and types of electron withdrawing or donating functional groups. No clear distinction between heterocyclic or non-heterocyclic compounds removal could be observed in this study (Fig. 4). Similarly, no discernible trend in terms of mononuclear or polynuclear compounds could be observed. It is generally considered that simple aliphatic and monocyclic aromatic compounds are readily degradable, while polycyclic structures may be more persistent (Jones et al., 2005). However, irrespective of the mono or polynuclear structure, degradation may be initiated by the mere cleavage of a side chain structure and then further mineralisation may depend on the complexity of the nucleus. In this study, removal indicates the loss of the parent structure, and not complete
80
80
Removal efficiency (%)
100
60
40
40
Na pro Sim xe n - hy dr o xya ci d
il
l ol o
roz mf ib
Ge
pr o
pa
Ate n
Ibu
ce
Ve ra
Pa ra
ac
pro
en l of
Ke to
Di c
fen
0
mi l
0
tam ol
20
Non-heterocyclic compounds
Polynuclear
60
20
fen
Removal efficiency (%)
Mononuclear
Polynuclear
Mononuclear
100
zi n e DE ET Dil a Pri ntin mi d T ri me one tho pr i m Su L l fa me inuro n tho xa z Hy dro ole xyz i ne En al a pr i l Ca r ba ma ze pi Tri am ne ter en e Ca Om ffein e ep r az Ri s ol e pe ri d on e
150
Atr a
Removal efficiency (%)
80
mineralisation. Therefore, the absence of any discernible correlation between the removal efficiency and ring structure is not entirely unexpected. As shown in Fig. 7, the compounds containing strong electron withdrawing groups (B) consistently showed very low (<20%) removal efficiency. According to Knackmuss (1996), the initial electrophilic attack by oxygenases of aerobic bacteria is often a rate-limiting step and the first of a chain of reactions responsible for the biodegradation of many organic compounds. As a result, the presence of electron withdrawing functional groups generates an electron deficiency and thus renders the compounds less susceptible to oxidative catabolism. Electron donating functional groups, on the other hand, render the molecules more prone to electrophilic attack by oxygenases of aerobic bacteria. Consequently, the removal efficiencies of organic compounds bearing strong electron donating functional groups were, in most cases, much higher than those of group B in Fig. 7. The compounds containing both electron withdrawing and donating groups however showing less than 70% removal have been placed in group C*. The elucidation of the overall influence of these functional groups and particularly their opposing effects on the biodegradability of trace organic compounds is a complex task and would generally require extensive exercise involving simultaneous application of quantitative structure activity relationship and biochemical interpretation, as demonstrated for a particular compound class (N-heterocycles) by Philipp et al. (2007). Because a large number of diverse compound classes were studied here, such a rigorous approach falls beyond the scope of this paper. Nevertheless some general inferences, can be drawn from the results in the light of metabolic pathway information retrieved from the sparse literature and also from biodegradation prediction tools such as UM-BBD PPS (Wackett and Ellis, 1999). The biodegradation of amide-only compounds needs to proceed from conversion of the amide group to primary amine (Hart and Orr, 1975). As suggested by the low removal of
Heterocyclic compounds
Fig. 4 e Removal efficiency as a function of ring structure. Error bars represent the standard deviation of 16 measurements.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 3 9 e2 4 5 1
2447
Fig. 5 e Compound classification according to the presence of electron donating or withdrawing functional groups.
carbamazepine and dilantin, this pathway appears to be extraordinarily recalcitrant. All the tested compounds possessing only methyl (weak electron donor) and amide (strong electron withdrawing) groups including primidone, DEET and meprobamate were very poorly removed. The presence of methyl groups means that the degradation could initiate from conversion of the methyl group to alcohol (Shaw and Harayama, 1992), bypassing the recalcitrant amide conversion. However, methyl and other aliphatic groups have very weak electron donating capacity, and thus in presence of a strong electron withdrawing group they may have limited activating effect. All three compounds (i.e. atenolol, enalapril and caffeine) containing both the amine (strong electron donating) and the amide (strong electron withdrawing) functional groups were quite well removed (50e97%). Degradation of compounds with one or more amine groups may proceed by converting the existing amine to a less substituted form of amine and aldehyde/ketone (Hakil et al., 1998). Comparing the performance of these three compounds (containing amide and amine groups) with that of primidone, DEET and meprobamate (containing amide and methyl groups), it appears that the coexistence of the amine, and not the methyl group, with the
amide group may make these compounds more amenable to biodegradation. The excellent removal of another amidecontaining compound paracetamol can be attributed to the presence of the hydroxyl group which is also a strong electron donating functional group. In this context, it is noteworthy that the entire set of hydroxyl group-containing compounds tested in this study showed high removal. Such positive impact of hydroxyl group on biodegradation is in line with previous literature reports (Tunkel et al., 2000). Halogenated organics comprise a superset which has many antimicrobial as well as human toxic and carcinogenic industrial chemicals as members (Ha¨ggblom and Bossert, 2004). Linuron contains both halogen and amide groups and accordingly demonstrated low removal. Interestingly, of the halogenated compounds with amine groups, risperidone (containing amine, methyl and amide) and hydroxyzine (containing amine and hydroxyl) showed good removal while diclofenac (containing amine and carboxylic) and atrazine (containing amine and methyl) showed poor removal. Literature review regarding the metabolic pathways of these compounds provided further insights but could not resolve the paradox. It is suggested in the literature that the metabolism of hydroxyzine can proceed simultaneously through the
2448
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 3 9 e2 4 5 1
80
60
40
Gro
up C
Log D
0 Caffeine
Naproxen
Sim-hydroxyacid
Trimethoprim
up C
Omeprazole
Gro
2 Triamterene
up B
Atenolol
Gro
4 Enalapril
up A
Ketoprofen
Gro
8 6 Sulfamethoxazole
0
Linuron Carbamazepine Atrazine Dilantin DEET Meprobamate Diclofenac Primidone -Verapamil Risperidone Hydroxyzine Gemfibrozil Ibuprofen Paracetamol
20
Androstenedione Estriol Triclocarban 4-tert-octylphenol Triclosan 17a-ethynylestra-diol Simvastatin 17b-estradiol Androsterone Etiocholanolone Amitriptyline Estrone Testosterone Bisphenol-A Clozapine
Removal Efficien
cy (%)
100
-2
*
Fig. 6 e The combined effects of functional group and hydrophobicity on the removal of trace organic compounds by the MBR. Error bars represent the standard deviation of 16 measurements. Group A: all compounds with Log D > 3.2 (at pH 8). Groups B, C, and C* are defined in Fig. 5.
conversion of amine to aldehyde/ketone and/or through oxidation of the alcohol moiety to a carboxylic acid (CampoliRichards et al., 1990). In the case of risperidone the degradation may initiate via 9-hydroxylation and/or via N-dealkylation at the piperadine nitrogen (Mannens et al., 1993). Diclofenac has been suggested to be degraded by hydroxylation of the 1-amino-2-unsubstituted aromatic fragment (Marco-Urrea et al., 2010). The degradation of atrazine, on the other hand, has been reported to be initiated through Nmonodealkylation, hydroxylation of the isopropyl or tertbutyl moiety (Lang et al., 1996) or in the rare case via oxidation of the s-triazine ring to hydroxy-s-triazine (De Souza et al., 1995). While it is certain that the aerobic oxidation of the halogenated compounds is initiated from the co-existing electron withdrawing groups and not via dehalogentaion, it is not clear why, despite seemingly similar metabolic pathways (e.g. hydroxylation, dealkylation), the compounds exhibit different extents of recalcitrance. Hydroxylation of the vicinal unsubstituted aromatic fragment and the mono-carbon-substituted benzenoid are the predominant initial degradation pathways (Quintana et al., 2005) for the well removed compounds ibuprofen (97%) and ketoprofen (70%), respectively. It is, however, not clear why despite possessing a suitable structure for the similar
metabolic pathway as ketoprofen, triamterene registered a rather low removal of 28%. The absence of any literature data regarding triamterene removal by CAS or MBR restricts further clarification regarding this matter. The only possible distinction that can be offered at this stage is that triamterene is a heterocyclic compound. It is known that the degradation of compounds with an aromatic-aliphatic ether fragment can proceed by ether cleavage, producing a phenol derivative and an aldehyde (Bernhardt et al., 1988). Of the tested compounds that fit into this category, gemfibrozil (98%) and verapamil (87%) were well removed while omeprazol (62%) and naproxen (40%) demonstrated moderate removal, and trimethoprim was poorly removed (16%). The predominant biodegradation route of naproxen and trimethoprim appears to be via ether cleavage (Quintana et al., 2005); however, the degradation can potentially proceed via conversion of a tertiary/secondary aliphatic group to the corresponding alcohol. On the other hand, in addition to the ether cleavage verapamil may be degraded by N-demethylation (Unadkat et al., 2008). The degradation of omeprazole can also initiate from the conversion of di-[C,O]substituted sulfoxide to sulfone (Kanazawa et al., 2003), and gemfibrozil can be degraded also through conversion of an aromatic methyl to primary alcohol (Hermening et al., 2000).
2449
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 3 9 e2 4 5 1
Trace organic contaminants
Log D
Possessing only ewithdrawing groups
3.2
Possessing both ewithdrawing & donating groups
Low removal (<20%)
Log D
3.2
Possessing only edonating groups
Removal varies
High removal (>70%)
Very high removal (>85%)
Fig. 7 e A qualitative framework for the prediction of trace organic removal by MBR treatment.
The discrepancy in the removal efficiencies of these compounds may, therefore, be attributed to the distinct alternate routes of biodegradation, which may govern the overall removal. The combined effect of functional groups and hydrophobicity on the removal of trace organic compounds by the MBR is shown in Fig. 6. It is evident from the above discussion that all the aspects of chemical structure i.e. aromatic moiety, ring composition, substituent groups, side chain and associated metabolic pathway need to be taken into account in conjunction with physical parameters namely hydrophobicity and molecular weight to explain observed variabilities in trace organic removal by MBR. As noted earlier, in an MBR, adsorption and biodegradation may simultaneously play important roles. However, for the compounds with low hydrophobicity, properties such as molecular weight, ring structure and functional groups may influence the biodegradability and consequently govern the overall removal. Although some similarities can be expected, the purpose of this section is clearly not to describe the biodegradability of trace organics in biological wastewater treatment in general. The comprehensive discussion on biodegradability and metabolic pathway as furnished here serves the important purpose of explaining the removal of compounds with low hydrophobicity in the MBR.
carefully considered. For example, because MBR usually operates at a much longer sludge retention time and can offer complete retention of the biomass, hydrophobicity of the trace organic compounds would have a more profound impact on their removal efficiency by MBR than that by CAS. For the compounds with low hydrophobicity, where biodegradability is likely to govern the overall removal, the performance of CAS operated under the same loading and sludge retention time may be comparable to MBR (Clara et al., 2005). However it also needs to be noted that MBR may facilitate growth and maintenance of special degrading microbes (Hai et al., 2010) which may contribute to enhanced removal of compounds with low hydrophobicity. To derive further insight into this matter, longterm performance of CAS and MBR will need to be compared with the same set of compounds. That, however, is beyond the scope of this study. It is prudent to note that this proposed framework has been based on a limited set of data of only 40 compounds. Nevertheless, this framework has the potential to provide significant insights to the removal of trace organic contaminants by MBR treatment. With ongoing scientific and dedicated efforts in this field, the framework can be a foundation for a future quantitative model for the prediction of trace organic removal by MBR and CAS treatment.
4. 3.4.
Conclusion
A framework to predict removal efficiency
Notwithstanding a few exceptions which will be subjected to further investigation, results reported in this study indicate a clear link between molecular features and the removal of trace organic compounds by MBR treatment. Fig. 7, based on the data presented in this study, outlines a qualitative and schematic framework for the prediction of the removal efficiency of any given compound by an aerobic MBR treatment process. Given the similarities between CAS and MBR treatment, the framework proposed here may also be applicable to CAS treatment processes to some extent. However, differences in operational conditions between MBR and CAS must be
Results reported in this study indicate an apparent correlation between molecular features and the removal of trace organic contaminants by a laboratory scale MBR system. The removal efficiencies of all 14 very hydrophobic trace organic compounds (Log D at pH 8 > 3.2) selected in this study consistently showed removal efficiencies in the range between 85% to removal below the analytical detection limit, indicating a removal of at least 98%. The occurrence of electron withdrawing or electron donating functional groups appears to be another important factor governing their removal by MBR treatment. All hydrophilic and moderately hydrophobic (Log D < 3.2) compounds possessing strong
2450
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 3 9 e2 4 5 1
electron withdrawing functional groups consistently showed removal efficiency of well below 20%. In contrast, high removal efficiency was observed with most compounds bearing electron donating functional groups such as hydroxyl groups and primary amine groups. Nevertheless, further analysis also revealed several exceptions which remained unexplainable given the current lack of biochemical data about these compounds of interest. Based on the reported data, a qualitative framework for the assessment of trace organics removal by MBR treatment was presented.
Acknowledgements We acknowledge the financial support from the Royal Thai Government to Nichanan Tadkaew for doctoral studies at the University of Wollongong. Zenon Environmental Inc (Ontario, Canada) is thanked for the provision of the submerged membrane module.
Appendix. Supplementary data Supplementary data related to this article can be found online at doi:10.1016/j.watres.2011.01.023.
references
Abegglen, C., Joss, A., McArdell, C.S., Fink, G., Schlu¨sener, M.P., Ternes, T.A., Siegrist, H., 2009. The fate of selected micropollutants in a single-house MBR. Water Research 43, 2036e2046. Atkinson, S., 2006. Research studies predict strong growth for MBR markets. Membrane Technology 2006, 8. Bernhard, M., Muller, J., Knepper, T.P., 2006. Biodegradation of persistent polar pollutants in wastewater: comparison of an optimised lab-scale membrane bioreactor and activated sludge treatment. Water Research 40, 3419e3428. Bernhardt, F.H., Bill, E., Trautwein, A.X., Twilfer, H., 1988. 4Methoxybenzoate monooxygenase from Pseudomonas putida: isolation, biochemical properties, substrate specificity, and reaction mechanisms of the enzyme components. Methods in Enzymology, 281e294. Bouju, H., Buttiglieri, G., Malpei, F., 2008. Perspectives of persistent organic pollutants (POPS) removal in an MBR pilot plant. Desalination 224, 1e6. Campoli-Richards, D.M., Buckley, M.M., Fitton, A., 1990. Cetirizine. A review of its pharmacological properties and clinical potential in allergic rhinitis, pollen-induced asthma, and chronic urticaria. Drugs 40, 762e781. Chen, J., Huang, X., Lee, D., 2008. Bisphenol A removal by a membrane bioreactor. Process Biochemistry 43, 451e456. Cirja, M., Zuehlke, S., Ivashechkin, P., Schaffer, A., Corvini, L., 2006. Fate of a 14C-labeled nonylphenol isomer in a laboratory-scale membrane bioreactor. Environmental Science & Technology 40, 6131e6136. Clara, M., Strenn, B., Ausserleitner, M., Kreuzinger, N., 2004. Comparison of the behaviour of selected micropollutants in a membrane bioreactor and a conventional wastewater treatment plants. Water Science and Technology 50, 29e36.
Clara, M., Strenn, B., Gans, O., Martinez, E., Kreuzinger, N., Kroiss, H., 2005. Removal of selected pharmaceuticals, fragrances and endocrine disrupting compounds in a membrane bioreactor and conventional wastewater treatment plants. Water Research 39, 4797e4807. De Souza, M.L., Wackett, L.P., Boundy-Mills, K.L., Mandelbaum, R. T., Sadowsky, M.J., 1995. Cloning, characterization, and expression of a gene region from Pseudomonas sp. strain ADP involved in the dechlorination of atrazine. Applied and Environmental Microbiology 61, 3373e3378. De Wever, H., Weiss, S., Reemtsma, T., Vereecken, J., Muller, J., Knepper, T., Rorden, O., Gonzalez, S., Barcelo, D., Dolores Hernando, M., 2007. Comparison of sulfonated and other micropollutants removal in membrane bioreactor and conventional wastewater treatment. Water Research 41, 935. Dejonghe, W., Berteloot, E., Goris, J., Boon, N., Crul, K., Maertens, S., Hofte, M., De Vos, P., Verstraete, W., Top, E.M., 2003. Synergistic degradation of linuron by a bacterial consortium and isolation of a single linuron-degrading Variovorax strain. Applied and Environmental Microbiology 69, 1532e1541. Eaton, A.D., Clescerl, L.S., Rice, E.W., Greenberg, A.E., 2005. Standard Methods for Examination of Water & Wastewater, 21st ed. American Public Health Association. Go¨bel, A., McArdell, C.S., Joss, A., Siegrist, H., Giger, W., 2007. Fate of sulfonamides, macrolides, and trimethoprim in different wastewater treatment technologies. Science of the Total Environment 372, 361e371. Gonzalez, S., Muller, J., Petrovic, M., Barcelo, D., Knepper, T.P., 2006. Biodegradation studies of selected priority acidic pesticides and diclofenac in different bioreactors. Environmental Pollution 144, 926e932. Ha¨ggblom, M.M., Bossert, I.D., 2004. Halogenated Organic Compounds e a Global Perspective Dehalogenation. Springer, US. 3e29. Hai, F.I., Yamamoto, K., Nakajima, F., Fukushi, K., 2010. Recalcitrant industrial wastewater treatment by membrane bioreactor (MBR). In: Gorley, S.V. (Ed.), Handbook of Membrane Research: Properties, Performance and Applications. Nova Science Publishers, Inc., New York, pp. 67e97. Hakil, M., Denis, S., Viniegra-Gonzalez, G., Augur, C., 1998. Degradation and product analysis of caffeine and related dimethylxanthines by filamentous fungi. Enzyme and Microbial Technology 22, 355e359. Hart, A., Orr, D.L.J., 1975. The degradation of paracetamol (4 hydroxyacetanilide) and other substituted acetanilides by a Penicillium species. Antonie Van Leeuwenhoek, International Journal of General and Molecular Microbiology 41, 239e247. Hermening, A., Gra¨fe, A.-K., Baktir, G., Mutschler, E., SpahnLangguth, H., 2000. Gemfibrozil and its oxidative metabolites: quantification of aglycones, acyl glucuronides, and covalent adducts in samples from preclinical and clinical kinetic studies. Journal of Chromatography B: Biomedical Sciences and Applications 741, 129e144. Hu, J.Y., Chen, X., Tao, G., Kekred, K., 2007. Fate of endocrine disrupting compounds in membrane bioreactor systems. Environmental Science & Technology 41, 4097e4102. Jones, O.A.H., Voulvoulis, N., Lester, J.N., 2005. Human pharmaceuticals in wastewater treatment processes. Critical Reviews in Environmental Science and Technology 35, 401e427. Joss, A., Keller, E., Alder, A.C., Go¨bel, A., McArdell, C.S., Ternes, T., Siegrist, H., 2005. Removal of pharmaceuticals and fragrances in biological wastewater treatment. Water Research 39, 3139e3152. Kanazawa, H., Okada, A., Higaki, M., Yokota, H., Mashige, F., Nakahara, K., 2003. Stereospecific analysis of omeprazole in human plasma as a probe for CYP2C19 phenotype. Journal of Pharmaceutical and Biomedical Analysis 30, 1817e1824.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 3 9 e2 4 5 1
Kim, S.D., Cho, J., Kim, I.S., Vanderford, B.J., Snyder, S.A., 2007. Occurrence and removal of pharmaceuticals and endocrine disruptors in South Korean surface, drinking, and waste waters. Water Research 41, 1013e1021. Kimura, K., Hara, H., Watanabe, Y., 2005. Removal of pharmaceutical compounds by submerged membrane bioreactors (MBRs). Desalination 178, 135e140. Knackmuss, H.-J., 1996. Basic knowledge and perspectives of bioelimination of xenobiotic compounds. Journal of Biotechnology 51, 287e295. Kreuzinger, N., Clara, M., Strenn, B., Kroiss, H., 2004. Relevance of the sludge retention time (SRT) as design criteria for wastewater treatment plants for the removal of endocrine disruptors and pharmaceuticals from wastewater. Water Science and Technology 50, 145e156. Lang, D., Criegee, D., Grothusen, A., Saalfrank, R.W., Bocker, R.H., 1996. In vitro metabolism of atrazine, terbuthylazine, ametryne, and terbutryne in rats, pigs, and humans. Drug Metabolism and Disposition 24, 859e865. Lapertot, M.E., Pulgarin, C., 2006. Biodegradability assessment of several priority hazardous substances: choice, application and relevance regarding toxicity and bacterial activity. Chemosphere 65, 682e690. Lyko, S., Wintgens, T., Melin, T., 2005. Estrogenic trace contaminants in wastewater e possibilities of membrane bioreactor technology. Desalination 178, 95e105. Mannens, G., Huang, M.L., Meuldermans, W., Hendrickx, J., Woestenborghs, R., Heykants, J., 1993. Absorption, metabolism, and excretion of risperidone in humans. Drug Metabolism and Disposition 21, 1134e1141. Marco-Urrea, E., Pe´rez-Trujillo, M., Cruz-Morato´, C., Caminal, G., Vicent, T., 2010. Degradation of the drug sodium diclofenac by Trametes versicolor pellets and identification of some intermediates by NMR. Journal of Hazardous Materials 176, 836e842. Melin, T., Jefferson, B., Bixio, D., Thoeye, C., De Wilde, W., De Koning, J., van der Graaf, J., Wintgens, T., 2006. Membrane bioreactor technology for wastewater treatment and reuse. Desalination 187, 271e282. Philipp, B., Hoff, M., Germa, F., Schink, B., Beimborn, D., Mersch-Sundermann, V., 2007. Biochemical interpretation of quantitative structure-activity relationships (QSAR) for biodegradation of N-heterocycles: a complementary approach to predict biodegradability. Environmental Science & Technology 41, 1390e1398. Qu, Y.-Y., Zhou, J.-T., Wang, J., Xing, L.-L., Jiang, N., Gou, M., Salah Uddin, M., 2009. Population dynamics in bioaugmented membrane bioreactor for treatment of bromoamine acid wastewater. Bioresource Technology 100, 244. Quintana, J.B., Weiss, S., Reemtsma, T., 2005. Pathways and metabolites of microbial degradation of selected acidic pharmaceutical and their occurrence in municipal wastewater treated by a membrane bioreactor. Water Research 39, 2654e2664. Radjenovic, J., Petrovic, M., Barcelo´, D., 2007. Analysis of pharmaceuticals in wastewater and removal using a membrane bioreactor. Analytical and Bioanalytical Chemistry 387, 1365e1377. Radjenovic, J., Petrovic, M., Barcelo´, D., 2009. Fate and distribution of pharmaceuticals in wastewater and sewage sludge of the conventional activated sludge (CAS) and advanced membrane bioreactor (MBR) treatment. Water Research 43, 831e841.
2451
Reif, R., Sua´rez, S., Omil, F., Lema, J.M., 2008. Fate of pharmaceuticals and cosmetic ingredients during the operation of a MBR treating sewage. Desalination 221, 511e517. Santos, J.L., Aparicio, I., Callejo´n, M., Alonso, E., 2009. Occurrence of pharmaceutically active compounds during 1-year period in wastewaters from four wastewater treatment plants in Seville (Spain). Journal of Hazardous Materials 164, 1509e1516. Shaw, J.P., Harayama, S., 1992. Purification and characterisation of the NADH: acceptor reductase component of xylene monooxygenase encoded by the TOL plasmid pWW0 of Pseudomonas putida mt-2. European Journal of Biochemistry 209, 51e61. Smook, T.M., Zho, H., Zytner, R.G., 2008. Removal of ibuprofen from wastewater: comparing biodegradation in conventional, membrane bioreactor, and biological nutrient removal treatment systems. Water Science and Technology 57, 1e8. 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. Tadkaew, N., Sivakumar, M., Khan, S.J., McDonald, J.A., Nghiem, L.D., 2010. Effect of mixed liquor pH on the removal of trace organic contaminants in a membrane bioreactor. Bioresource Technology 101, 1494e1500. Tunkel, J., Howard, P.H., Boethling, R.S., Stiteler, W., Loonen, H., 2000. Predicting ready biodegradability in the Japanese Ministry of International Trade and Industry test. Environmental Toxicology and Chemistry 19, 2478e2485. Unadkat, J.D., Chung, F., Sasongko, L., Whittington, D., Eyal, S., Mankoff, D., Collier, A.C., Muzi, M., Link, J., 2008. Rapid solidphase extraction method to quantify [11C]-verapamil, and its [11C]-metabolites, in human and macaque plasma. Nuclear Medicine and Biology 35, 911e917. Urase, T., Kagawa, C., Kikuta, T., 2005. Factors affecting removal of pharmaceutical substances and estrogens in membrane separation bioreactors. Desalination 178, 107e113. Vanderford, B.J., Snyder, S.A., 2006. Analysis of pharmaceuticals in water by isotope dilution liquid chromatography/tandem mass spectrometry. Environmental Science & Technology 40, 7312e7320. Visvanathan, C., Ben Aim, R., Parameshwaran, K., 2000. Membrane separation bioreactors for wastewater treatment. Critical Reviews in Environmental Science and Technology 30, 1e48. Visvanathan, C., Thu, L.N., Jegatheesan, V., Anotai, J., 2005. Biodegradation of pentachlorophenol in a membrane bioreactor. Desalination 183, 455e464. Wackett, L.P., Ellis, L.B.M., 1999. Predicting biodegradation. Environmental Microbiology 1, 119e124. Wells, M.J.M., 2006. Log Dow: key to understanding and regulating wastewater-derived contaminants. Environmental Chemistry 3, 439e449. Wintgens, T., Gallenkernper, M., Melin, T., 2004. Removal of endocrine disrupting compounds with membrane processes in wastewater treatment and reuse. Water Science and Technology 50, 1e8. Yu, J.T., Bouwer, E.J., Coelhan, M., 2006. Occurrence and biodegradability studies of selected pharmaceuticals and personal care products in sewage effluent. Agricultural Water Management 86, 72e80. Zhang, J., Chua, H.C., Zhou, J., Fane, A.G., 2006. Factors affecting the membrane performance in submerged membrane bioreactors. Journal of Membrane Science 284, 54.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 5 2 e2 4 6 2
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Methanogenic community development in anaerobic granular bioreactors treating trichloroethylene (TCE)-contaminated wastewater at 37 C and 15 C Alma Siggins, Anne-Marie Enright, Vincent O’Flaherty* Microbial Ecology Laboratory, Department of Microbiology and Environmental Change Institute (ECI), National University of Ireland, Galway (NUI, Galway), University Road, Galway, Ireland
article info
abstract
Article history:
Four expanded granular sludge bed (EGSB) bioreactors were seeded with a mesophilically-
Received 25 November 2010
grown granular sludge and operated in duplicate for mesophilic (37 C; R1 & R2) and low- (15 ;
Received in revised form
R3 & R4) temperature treatment of a synthetic volatile fatty acid (VFA) based wastewater (3 kg
21 January 2011
COD m3 d1) with one of each pair (R1 & R3) supplemented with increasing concentrations of
Accepted 31 January 2011
trichloroethylene (TCE; 10, 20, 40, 60 mg l1) and one acting as a control. Bioreactor perfor-
Available online 19 February 2011
mance was evaluated by % COD removal efficiency and % biogas methane (CH4) content. Quantitative Polymerase Chain Reaction (qPCR) was used to investigate the methanogenic
Keywords:
community composition and dynamics in the bioreactors during the trial, while specific
EGSB
methanogenic activity (SMA) and toxicity assays were utilized to investigate the activity and
Low-temperature anaerobic diges-
TCE/dichloroethylene (DCE) toxicity thresholds of key trophic groups, respectively. At both
tion
37 C and 15 C, TCE levels of 60 mg l1 resulted in the decline of % COD removal efficiencies to
Specific methanogenic activity
29% (Day 235) and 37% (Day 238), respectively, and in % biogas CH4 to 54% (Day 235) and 5% (Day
Toxicity
238), respectively. Despite the inhibitory effect of TCE on the anaerobic digestion process, the
TCE
main drivers influencing methanogenic community development, as determined by qPCR and
qPCR
Non-metric multidimensional scaling analysis, were (i) wastewater composition and (ii) operating temperature. At the apical TCE concentration both SMA and qPCR of methanogenic archaea suggested that acetoclastic methanogens were somewhat inhibited by the presence of TCE and/or its degradation derivatives, while competition by dechlorinating organisms may have limited the availability of H2 for hydrogenotrophic methanogenesis. In addition, there appeared to be an inverse correlation between SMA levels and TCE tolerance, a finding that was supported by the analysis of the inhibitory effect of TCE on two additional biomass sources. The results indicate that low-temperature anaerobic digestion is a feasible approach for the treatment of TCE-containing wastewater. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
Trichloroethylene (TCE; C2HCl3) is a chlorinated aliphatic compound widely used in the cleaning and degreasing industries (Doherty, 2000). It is considered a potentially
carcinogenic and mutagenic compound (US EPA, 1997) exposure to which may cause liver and kidney damage, and impaired functioning of the heart, immune system and central nervous system (ATSDR, 2003). Contamination of soil and groundwater sites by TCE has become widespread, mainly
* Corresponding author. Tel.: þ353 (0) 91 493734; fax: þ353 (0) 91 494598. E-mail address:
[email protected] (V. O’Flaherty). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.01.030
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 5 2 e2 4 6 2
due high levels of use and inappropriate disposal methods (Doucette et al., 2007). In 2003, TCE was reported as present in at least 852 of the 1430 National Priorities Lists identified by the US Environmental Protection Agency (ATSDR, 2003). Rivett et al. (1990) also reported the presence of TCE in 45 out of 59 supply boreholes tested in the Birmingham (UK) Triassic Sandstone aquifer, with 30% of those observed at high concentrations (30e5500 mg l1). While physical remediation involving the extraction of contaminated groundwater and soils has been considered inefficient, costly, slow and environmentally disruptive (Russell, 1992), it has been demonstrated that TCE is biodegradable under both aerobic and anaerobic conditions (Ozdemir et al., 2007; Volcik et al., 2005; van Agteren et al., 1998). Further studies, including Wu et al. (1997), Freedman and Gossett (1989) and Ferguson and Pietari (2000), have shown that complete dechlorination of trichloroethylene to ethylene can be carried out under anaerobic conditions. The advantages of anaerobic bioremediation were discussed in detail by Speece (1983), but were primarily due to reduced cost of excess cell disposal, reduced electricity consumption and the value of methane produced during the process. This urgent demand for greater economic efficiency from these processes has paved the way for studies into the feasibility of low-temperature reactor operations. An efficient level of lowtemperature recalcitrant compound degradation, comparable with mesophilic operation has been demonstrated for a range of wastewater types (Rebac et al., 1995; Enright et al., 2007; McKeown et al., 2008). In addition, the successful degradation of tetrachloroethylene (PCE) has been demonstrated at both mesophilic (Sponza, 2003) and low-temperature conditions (Carter and Jewell, 1993). However, little or no data is available on the response of an anaerobic biomass to the presence of chlorinated aliphatic compounds, at varying temperatures. In light of the above, the aim of this study was to monitor the community structure and metabolic response of methanogenic communities within laboratory-scale bioreactors to the presence of TCE, and its degradation derivatives, at 37 C and 15 C. A polyphasic approach, combining qPCR of the 16S rRNA gene of key methanogenic groups, specific methanogenic activity (SMA) and toxicity assays was undertaken.
2.
Materials and methods
2.1.
Source of biomass
A granular, anaerobic sludge was obtained from a mesophilic (37 C), full-scale (1500 m3) internal circulation (IC) alcohol wastewater treatment bioreactor at Carbery Milk Products, Ballineen, Co. Cork, Ireland. The granules were of a regular spherical shape (Ø, c. 1e2 mm) with a volatile suspended solids (VSS) concentration of 86 g l1. Additionally, two further granular, anaerobic sludges (B & C) were obtained from lab-scale bioreactors, previously used for the treatment of phenol (McKeown et al., 2008) and brewery wastewater (Connaughton et al., 2006), respectively, at 15 C. The volatile suspended solids of sludges B and C were 93 and 34 g VSS l1, respectively.
2.2.
2453
Design and operation of EGSB bioreactors
Four glass, laboratory-scale (3.5l), expanded granular sludge bed (EGSB) bioreactors, R1-R4, as described by Collins et al. (2004) were utilized for this study. R1 and R2 were each inoculated with 40 g VSS, and operated at 37 C while, R3 and R4 were each inoculated with 70 g VSS of biomass and operated at 15 C. All four bioreactors were used for the treatment of a synthetic volatile fatty acid based wastewater consisting of acetic acid, propionic acid, butyric acid and ethanol in the chemical oxygen demand (COD) ratio of 1:1:1:1, to a total of 3 g COD l1. The synthetic influent was buffered with NaHCO3 and fortified, as described by Shelton and Tiedje (1984) with macro(10 ml l1) and micro- (1 ml l1) nutrients. The organic loading rate (OLR) applied to all bioreactors was 3 kg COD m3 d1 with a hydraulic retention time (HRT) of 24 h. Effluent was recirculated through the systems at an applied upflow velocity of 2.5 m h1. In addition, R1 and R3 were supplemented with 10 mg l1 TCE on day 149. This was increased in a step-wise manner through 20, 40 and 60 mg l1 on days 172, 191 and 226 respectively, before TCE was removed from both R1 and R3 on day 243 in response to poor bioreactor performance.
2.3.
Specific methanogenic activity and toxicity testing
Seed biomass, sludges B and C, and samples collected from the bioreactors on days 108, 235 and 343 were screened for metabolic capability using specific methanogenic activity (SMA) tests, which were performed using the pressure transducer technique (Colleran et al., 1992; Coates et al., 1996), in which, acetate (30 mM) and H2/CO2 (80:20, v/v) were employed as substrates in order to establish the activities of acetoclastic and hydrogenotrophic methanogens, respectively. Vials without any substrate, or with the addition of N2/CO2 (80:20, v/ v) in the case of hydrogenotrophic tests, served as controls. Trichloroethylene (TCE) induced methanogenic toxicity of seed biomass and sludges B and C, and TCE, 1,1 dichloroethylene (DCE), cis-1,2 DCE, and trans-1,2 DCE induced toxicity of biomass collected from the bioreactors on days 108, 235 and 343 were assessed using the SMA based toxicity assay (acetoclastic and hydrogenotrophic) as described by Colleran and Pistilli (1994) and Enright et al. (2005). Toxicity was defined in terms of the IC50 value i.e. the concentration (mg l1) of toxicant that resulted in 50% inhibition of SMA, which was calculated from the linear regression of SMA as a function of toxicant concentration. Vials without any toxicant added served as controls. All activity and toxicity assays contained 2e5 g volatile suspended solids (VSS) l1 and were performed in triplicate at 15 and 37 C for candidate biomass, and at the reactor operational temperature for bioreactor samples (R1 & R2, 37 C; R3 & R4, 15 C).
2.4.
Analytical methods
Reactor influent, effluent and biogas from R1-R4 were routinely sampled. Influent and effluent COD concentration and % biogas CH4 content were determined according to Standard Methods American Public Health Association (APHA 1998), and % COD removal efficiency was determined from calculated influent and effluent measurements.
2454
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 5 2 e2 4 6 2
Analysis of VFA concentrations of effluent samples were performed by heated (85 C) and agitated headspace, in a Varian Saturn 2000 GC/MS system, with CombiPAL autosampler (Varian Inc., Walnut Creek, CA). Separation was carried out on a Varian Capillary column, CP-WAX 58 (FFAP) CB (25 m length 0.32 mm internal diameter 0.2 mm film thickness, Varian). The injector volume was 2 ml and the injector temperature was maintained at 250 C. Helium was employed as the carrier gas, at a flow rate of 1 ml min1. The temperature program was as follows: 50 C (20 s) to 110 C (20 s) at a rate of 2 C min1; from 110 C to 200 C (20 s) at a rate of 20 C min1. The MS-detector was operated in the scan mode in the range of 40e150 m z1 at a temperature of 210 C. Identification of VFAs was achieved by matching chromatographic retention times and spectra of standard compounds (acetic-, butyric- and propionic-acids). Calibration curves of standard VFAs were constructed and used for relative concentration of VFAs in effluent headspace samples and then expressed as mg l1.
2.5.
Extraction of genomic DNA
Total genomic DNA was extracted from biomass sampled at Day 0 (mesophilic inoculum), Day 149 (TCE, 0 mg l1), Day 173 (TCE, 10 mg l1), Day 191 (TCE, 20 mg l1), Day 224 (TCE, 40 mg l1), Day 235 (TCE, 60 mg l1) and Day 343 (TCE, 0 mg l1), using an automated nucleic acid extractor (Magtration 12GC, PSS Co., Chiba, Japan). Prior to extraction, granular biomass was finely crushed using a mortar and pestle, and re-suspended in sterile double distilled water to a ratio of 1:4. A 100 ml aliquot of the biomass suspension was loaded per extraction. The VSS of each sample was also determined. Each extraction was performed in duplicate and the extracted DNA was eluted in TriseHCl buffer (pH 8.0) and stored at 20 C.
2.6.
qPCR
Quantitative real-time PCR was performed using a LightCycler 480 (Roche, Mannheim, Germany) with four methanogenic primer and probe sets, specific to two hydrogenotrophic orders (Methanomicrobiales and Methanobacteriales) and two acetoclastic families (Methanosaetaceae and Methanosarcinaceae), covering most methanogens present in anaerobic digesters (Lee et al., 2009; Yu et al., 2005). All DNA samples were analyzed with each primer and probe set in duplicate. Each reaction mixture was prepared using the LightCycler TaqMan Master Kit (Roche): 2 ml PCR-grad water, 1 ml of probe (final concentration 200 nM), 1 ml each primer (final concentration 500 nM), 10 ml of 2 reaction solution and 5 ml of DNA template. Amplification was carried out using a two-step thermal cycling protocol consisting of predenaturation for 10 min at 94 C, followed by 50 cycles of 10 s at 94 C and 30 s at 60 C. Quantitative standard curves were constructed using the standard plasmids containing the full-length 16S rRNA gene sequences from the representative strains of the target methanogenic groups as previously described (Lee et al., 2009; Yu et al., 2005). For each primer and probe set, an equimolar mixture of its corresponding standard plasmids was used as the template solution for constructing the standard curve. The mass concentration of each plasmid was measured in duplicate
using a Qubit system (Invitrogen) and converted into its copy concentration as previously described (Lee et al., 2009). A 10-fold serial dilution series (101e109 copies ml1) was generated for each standard solution and analyzed by real-time PCR in triplicate with its corresponding primer and probe set. The threshold cycle (CT) values determined were plotted against the logarithm of their input copy concentrations. The 16S rRNA gene copy concentrations of target groups were then estimated against the corresponding standard curves within the linear range (R2 > 0.995). The volume-based concentration (copies l1) were converted into the biomass-based concentration (copies g [VSS]1) using the VSS concentration of each sludge sample.
3.
Results
3.1.
Bioreactor performance: phases 1e3
During the initial 149 days of this study (P1), a rapid start-up was recorded for R1 and R2 (37 C), which displayed % COD removal efficiencies of >80% within 3 days (Fig. 1), and maintained this capability throughout P1 with phase averages of 82% and 78%, respectively (Table 1). Although R3 and R4 (15 C) had achieved the same level of % COD removal by the end of P1 (Fig. 1), it was much slower start-up period, resulting in reduced mean P1% COD removal averages of 67% and 61% respectively (Table 1). The introduction of TCE to the influent wastewater of R1 and R3, at a concentration of 10 mg l1 on day 149, had no visible effect on bioreactor performance (Fig. 1). Contrarily, all four bioreactors generally recorded increases in mean phase % COD removal efficiencies and % biogas CH4, from P1 to P2 (Table 1). On day 173, the influent TCE concentration was increased to 20 mg l1 (P3). During this phase, COD removal efficiencies remained high for all bioreactors (mean 85%; Table 1), however, during P3 biogas CH4 levels were lower in the TCE-supplemented bioreactors (R1 and R3) than in the corresponding control bioreactors (R2 and R4; Table 1). This divergence was recorded until the trial conclusion, irrespective of bioreactor operational temperature.
3.2.
Bioreactor performance: phases 4e6
On day 191, the concentration of TCE in the influent wastewater of R1 and R3 was increased to 40 mg l1 (P4). An immediate deterioration was observed in the performance of R1, with a decrease in COD removal efficiencies of approximately 25% within 2 days, which dropped further to 40% on day 201 (Table 1; Fig. 1). A temporary recovery was observed in R1 COD removal efficiency by day 222 (to 92%; Fig. 1). By the end of P4 the bioreactor had become unstable, however, fluctuating between 60 and 90%, resulting in a decrease of phase average COD removal efficiencies to 74% (Table 1). In addition, R2 also displayed some instability in COD removal efficiencies during P4, although to a lesser degree than R1 (Fig. 1) and R2 recorded P4 average COD removal efficiencies, which were, on average, 9% higher than that of the TCE-supplemented R1 bioreactor (Table 1). The increased concentration of TCE also affected a decline in the % COD removal efficiency of the 15 C bioreactor, R3, which achieved only 45% removal on day 215 (Fig. 1), and for
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 5 2 e2 4 6 2
2455
Fig. 1 e % COD removal efficiency A: 37 C R1 (-), R2 (,); B: 15 C R3 (C), R4 (B).
the first time in the trial, R3 recorded a lower phase average COD removal efficiency than the control bioreactor, R4 (Table 1). A further deterioration in R1 and R3 bioreactor performance was observed after day 226, when the concentration of TCE in the influent wastewater was increased to 60 mg l1 (P5). In particular, R1 COD removal efficiencies and biogas CH4 content had declined to 29% and 54%, respectively on day 235 and remained low for the duration of P5 (Fig. 1), while R2 COD removal efficiencies and biogas CH4 content remained unaffected (Fig. 1; Table 1). Similarly, R3 COD removal efficiencies and biogas CH4 content became more unstable with P5 lows of 37% and 5%, respectively, on day 238 (Fig. 1). By contrast, the low-temperature control bioreactor, R4, recorded a slight increase in performance during P5, with % COD removal efficiencies and biogas CH4 production means of 79% and 76%, respectively (Table 1).
TCE was withdrawn from R1 and R3 influent on day 243 (P6), resulting in an immediate improvement in the performance of both bioreactors, with COD removal efficiencies and biogas CH4 of 90% and 70% respectively for both bioreactors (Fig. 1). Mean COD removal efficiencies and biogas methane yields for all reactors (R1-R4) during P6 were 79% and 60%, respectively (Table 1). Finally, the principal volatile fatty acids (VFA) represented in effluent COD from all bioreactors during P1 were propionate and acetate, with concentrations of each ranging from 400 to 600 mg l1 in R3 and R4 effluents during the relatively long start-up period. Acetate was the principal VFA, however, in all reactor effluent samples taken from day 150 onwards and, in particular, during process perturbations due to the addition of TCE, while propionate concentrations were maintained below 50 mg l1 during these periods (data not shown).
2456
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 5 2 e2 4 6 2
Table 1 e Influent TCE concentration and performance characteristics during Phases 1e6 of operation. Values are means of phases, standard deviations are given in parenthesis. Phases Days 1
Influent TCE (mg l )
P1
P2
P3
P4
P5
P6
0e149
150e172
173e191
192e226
227e243
243e343
0
10
20
40
60
0
Mean COD Removal (%)
R1 R2 R3 R4
82 (10) 78 (13) 67 (15) 61 (12)
86 (9) 87 (10) 87 (11) 75 (10)
91 (4) 89 (7) 84 (8) 74 (11)
74 (15) 83 (10) 76 (15) 78 (9)
40 (18) 76 (9) 74 (17) 79 (10)
79 (18) 84 (10) 84 (12) 83 (9)
Mean biogas CH4 (%)
R1 R2 R3 R4
69 (6) 50 (7) 63 (11) 56 (15)
68 (5) 61 (4) 71 (5) 70 (3)
67 (2) 71 (3) 65 (3) 74 (2)
62 (4) 69 (2) 65 (3) 73 (3)
50 (12) 70 (3) 31 (25) 76 (3)
66 (16) 70 (7) 60 (13) 74 (7)
3.3. Specific methanogenic activity of seed and bioreactor biomass An initial, transient, decline in the specific methanogenic activity (SMA) against acetate was recorded in both 37 C bioreactors between days 0 and 108 (Table 2). Following this, the activity of R1 biomass against acetate increased in the presence of TCE, but not to the same extent as the control bioreactor R2, and R1 acetate SMA was c.48% lower than that of R2 reactor by day 235 (Table 2). The removal of TCE from the influent of R1 resulted in a significant recovery in acetate SMA values, and these were actually higher than those recorded for R2 by the conclusion of the trial. Hydrogenotrophic methanogenic activity in R1 was also significantly reduced by the presence of TCE, with a c.66% decrease in activity between days 108 and 235, which was not reflected in R2 (Table 2). SMA measured against hydrogen on day 343 showed a 14-fold increase in H2/CO2 SMA when TCE was removed from R1 (Table 2).
Table 2 e Specific Methanogenic Activity of Reactor Biomass All values are expressed in ml CH4 gVSS-1 d-1 and are the means of triplicates ± std. errors (std. deviation/On, n [ 3), except * where values are the mean of duplicates ± std. errors (n [ 2). Biomass
Temperature ( C)
Day
37 37
R2
37
Seed R3
15 15
R4
15
3.4.
Toxicity of TCE toward seed and bioreactor biomass
Substrate Acetate
Seed R1
Acetoclastic SMA at 15 C increased in biomass samples taken from both bioreactors during the trial in relation to the seed biomass, but the addition of TCE to R3 limited this increase, and, by day 235, the acetoclastic SMA of R3 was 42% lower than that of R4 (Table 2). Nevertheless, after TCE was removed from R3, acetoclastic activity increased 3.5-fold and was again higher than that of R4 by the conclusion of the trial (day 343; Table 2). Similar to the results for R1 at 37 C, hydrogenotrophic methanogenesis was significantly affected by the presence of TCE with an 85% reduction in R3 H2mediated activity recorded between days 108 and 235, while H2-mediated SMA was c.2.5-fold higher in R4 biomass on day 235 (Table 2). A shift in the potential routes for methanogenesis was observed during the trial for both R1 and R2, at 37 C, and also for R3 and R4 at 15 C. The SMA of acetoclasts was >2.5-fold more active in the seed biomass than that of hydrogenotrophs at 37 C. By the end of the trial, however, hydrogenotrophic SMA was 2.5 and 3.2-fold higher than acetoclastic SMA in R1 and R2, respectively (Table 2). By contrast, although the potential for hydrogenotrophic methanogenesis was higher in the seed sludge at 15 C, the maximum acetoclastic SMA was 2.5e3-fold greater than hydrogenotrophic SMA in R3 and R4 biomass by the conclusion the trial (Table 2).
Hydrogen
0 108 235 343 108 235 343
578 183 222 344 225 423 204
(69) (60) (5) (18) (6) (1) (6)
213 (10) 189 (21) 63 (8) 867 (79) 293 (5) 698 (99) 668 (23)*
0 108 235 343 108 235 343
16 54 74 255 58 128 214
(1) (1) (4) (7)* (2) (7) (50)
41 (5) 215 (13) 28 (2) 76 (3) 66 (2) 70 (6) 83 (0.3)
The IC50 levels of TCE and the three DCE isomers against the acetoclastic methanogens of R1 generally decreased from day 108e235, when the influent TCE concentration was 60 mg l1 (Tables 1 and 3). On day 235, R1 biomass recorded an IC50 value against the acetoclastic methanogens of 40 mg TCE l1 (Table 3). However, by the end of the trial, R1 acetoclastic IC50 values increased to 67 mg TCE l1 and assay values against cis-1,2 DCE and trans-1,2 DCE were above the measured range (0e120 mg l1 and 0e100 mg l1, respectively; Table 3). Furthermore, R1 hydrogenotrophic IC50 values increased during the trial, indicating an adaptation of these organisms to TCE and its degradation derivatives (Table 3). Indeed, all concentrations of toxicants assayed on day 343 displayed little or no effect on hydrogenotrophic methanogenesis (Table 3). Limitations were imposed by the low water solubility of TCE,
2457
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 5 2 e2 4 6 2
Table 3 e IC50 values of reactor biomass. All values are the mean of triplicates. aAcetoclastic methanogens b Hydrogenotrophic methanogens. n.d. not determined. Biomass
Bioreactor day
Temperature ( C)
Toxicants TCE a
1,1 DCE b
cis-1,2 DCE
trans-1,2 DCE
A
H
A
H
A
H
A
H
Seed R1
0 108 235 343
37
36 >50 40 67
53 88 143 >160
n.d. >100 111 119
n.d. >100 >150 >250
n.d. 75 6 >120
n.d. 77 >150 >200
n.d. >100 14 >100
n.d. 56 127 >100
Seed R3
0 108 235 343
15
68 >100 >150 >240
74 >100 >150 >240
n.d. 88 82 151
n.d. >100 >100 >250
n.d. >100 >150 >240
n.d. >100 >150 >400
n.d. >100 124 119
n.d. >100 >150 >300
which precluded the use of higher test concentrations in the assays. The IC50 levels of TCE against acetoclastic and hydrogenotrophic methanogens of R3 also increased throughout the trial, to the extent that all of the IC50 values on days 108, 235 and 343 exceeded the range of the toxicity assay (maximum assayed concentration 240 mg l1; Table 3). Similarly, with the exception of some 1,1 DCE and trans-1,2 DCE assays against the acetoclastic methanogens, the IC50 concentrations of all DCE isomers generally exceeded the toxicity assay range (Table 3). These results suggested that adaptation of the methanogenic community in the R1 and R3 biomass to the presence of TCE was observed at both 15 and 37 C. A hypothesis that an inverse relationship exists between SMA and TCE toxicity thresholds was suggested by this study. To further develop this idea, the SMA and IC50 values of two additional anaerobic granular biomass samples were also measured. In both cases, low SMA values corresponded to IC50 values above the measurable range of the toxicity assay (Table 4), which was essentially dictated by the solubility of TCE in an aqueous stock solution (1.28 g l1; The Merck Index, 2006), further corroborating the supposition that SMA is inversely related to the tolerance of the methanogenic community to TCE.
3.5.
Methanogenic community development
The hydrogenotrophic order Methanobacteriales (MMB) emerged strongly in all four bioreactors during the trial. MMB were not quantified above the assay detection limit (1 106 copies g [VSS]1) in the seed or day 149 biomass samples, or in biomass
taken from R1-R3 on day 173, although >8 106 copies g[VSS]1 were detectable on day 173 in R4 biomass (Fig. 2). By day 191, MMB were quantified in all bioreactors (Fig. 2) and had reached >109 copies g[VSS]1 in R4 biomass (Fig. 2). It would appear that at 37 C, population shifts of MMB were unrelated to TCE concentration, with fluctuations in R1 16S rRNA gene copy concentration also reflected in the control bioreactor (R2; Fig. 2). The rate of emergence of MMB appeared to be enhanced somewhat at 15 C, with higher levels detected in R3 on days 224 and 235 than at any sampling point at 37 C, and R4 16S rRNA gene concentration reached 5.9 109 copies g[VSS]1 on day 235, almost 200-fold higher than R2 concentrations (Fig. 2). By contrast, the hydrogentrophic Methanobacteriales (MBT) were present at all sample points, remaining at relatively steady concentrations in both bioreactors at 37 C, with the exception of two points of interest: firstly, although both R1 and R2 displayed an increase in MBT populations on day 149, this increase was greater in R1, reaching 1.13 1010 copies g[VSS]1 (Fig. 2); secondly, R2 showed a 90-fold decrease in MBT between days 224 and 235 that was not mirrored in R1. At 15 C, the MBT populations of R4 were unstable, decreasing consecutively from day 149 until day 224 (>100-fold decrease), when it recovered, recording concentrations in the 109 copies g[VSS]1 range by the conclusion of the trial (Fig. 2). Of the two acetoclastic methanogen families analyzed, the Methanosarcinaceae were not observed at concentrations above the detection limit in any samples, while in contrast, Methanosaetaceae (MSt) were abundant in all samples, particularly in the seed biomass, where they accounted for 75% of the total measured methanogenic 16S rRNA gene
Table 4 e Specific Methanogenic Activity and IC50 values of supplementary biomass sources. All SMA values are expressed in ml CH4 gVSS-1 d-1 and are the means of triplicates ± std. errors (std. deviation/On, n [ 3), except * where values are the mean of duplicates ± std. errors (n [ 2). All IC50 values are the means of triplicates. Biomass
Temperature ( C)
B
37 15
C
37 15
SMA
IC50
Acetate
Hydrogen
Acetoclastic
Hydrogenotrophic
181 (1) 7 (2)
173 (7) 29 (3)
>100 >100
>200 >200
78 (3) 17 (8)
>100 >100
>200 >200
78 (11) 24 (2)*
2458
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 5 2 e2 4 6 2
Fig. 2 e Quantitative changes in the 16S rRNA gene concentration of methanogens: MBT Methanobacteriales; MMB Methanomicrobiales; MSt Methanosaetaceae.
concentration (Fig. 2). Furthermore, MSt levels did not appear adversely affected by a reduction in bioreactor operating temperature, as both control reactors (R2 & R4) retained steady, comparable copies g[VSS]1 throughout the trial (Fig. 2). However, the MSt family did exhibit a temperature-dependent response to the presence of TCE in the bioreactor influent. At 37 C, low levels of TCE elicited a decrease in MSt copies g[VSS]1, with a 20-fold decline
observed from pre-TCE levels on day 149, to day 191, when bioreactor influent TCE concentration was 20 mg l1 (Fig. 2). Moreover, levels of MSt detected on day 191 from R1 biomass (9.38 108 copies g[VSS]1) were the lowest levels detected at any sampling point throughout the trial (Fig. 2). Conversely, biomass sampled after subsequent increases in influent TCE concentration demonstrated a recovery in MSt, reaching 7.8 109 copies g[VSS]1 on day 224, and remained higher
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 5 2 e2 4 6 2
than seed MSt levels until the conclusion of the trial (Fig. 2). In contrast, at 15 C, the MSt population of R3 displayed no response to the presence of TCE at influent concentrations of 10 or 20 mg l1 (days 173 & 191; Fig. 2), but recorded a decline in MSt population when influent TCE levels were increased further on days 224 and 235 (Table 1; Fig. 2).
4.
Discussion
The development of elevated methanogenic activity at low temperatures, after extended exposure to psychrophilic conditions in bioreactors, has been well documented (McHugh et al., 2003; Collins et al., 2003, 2005a; Lettinga et al., 1999; Kettunen and Rintala, 1997); and Collins et al. (2005b) concluded that the use of a mesophilic biomass is a satisfactorily efficient means of cultivating a consortium for psychrophilic reactor operation. In addition, it has been shown that commencing low-temperature bioreactor trials directly at the target temperature (rather than by step-wise reduction) can minimize an inefficient pre-toxicant start-up phase (Enright et al., 2005; Scully et al., 2006). The poor lowtemperature activity demonstrated by the seed biomass (Table 2) was not, therefore, of concern for this study. A rapid start-up was observed in both 37 C bioreactors (R1 & R2), and while reduced COD removal efficiencies were noted for both 15 C bioreactors (R3 & R4; Table 1), the performance of R3 and R4 improved steadily throughout P1, and displayed COD removal efficiencies on a par with their mesophilic counterparts by the end of this phase. It was not until an increase in influent TCE concentration to 40 mg l1 on day 192 that a decline in COD removal efficiency was noted in bioreactors R1 and R3, although performance had recovered within <10 HRT cycles, while biogas CH4 levels remained unaffected (Table 1). This apparent adaptation was not sufficient to allow stable anaerobic digestion processes to take place in the presence of TCE at the increased concentration of 60 mg l1 on day 226, which elicited a contrasting, apparently temperature-dependent response. Specifically, an immediate decline in COD removal efficiency (<20%) was observed in the mesophilic bioreactor (R1; Fig. 1), which continued during P5 and did not recover for 9 HRT cycles after the removal of TCE, indicating an increased sensitivity to TCE at this temperature. By contrast, although the low-temperature bioreactor (R3) displayed reduced COD removal efficiencies during this phase, its performance was considerably more stable than the mesophilic bioreactor (Fig. 1). Furthermore, the severity of the operational decline in R3 and its recovery time after the withdrawal of TCE were greatly reduced. Although neither bioreactor had demonstrated recovery by the conclusion of P5 (Fig. 1; Table 1), process efficiency improved during P6, after the withdrawal of TCE from the influents of both R1 and R3, suggesting that the toxicity/inhibitory effects of TCE were fully reversible in the EGSB bioreactor system. This result is in agreement with previously documented trials, whereby toxicity induced by chlorophenol (Droste et al., 1998), TCP (Collins et al., 2005a) and toluene (Enright et al., 2007) was successfully reversed by withdrawal from the wastewater. Consequently, the maximum TCE loading thresholds were determined to be <60 mg l1 for both test temperatures.
2459
SMA assays carried out on bioreactor biomass indicated that hydrogenotrophic methanogenesis was adversely affected by the presence of TCE at 37 C (Table 2; Day 235), a supposition supported by the decrease in recorded biogas CH4 levels from R1 during P5. Interestingly, the IC50 values of all toxicants measured against hydrogenotrophic methanogens increased during the trial (Table 3). Similarly, SMA assays confirm that the hydrogenotrophic community in R3 (15 C) was inhibited by the supplementation of TCE, despite the increased toxicity assay IC50 values, which inferred that hydrogenotrophs were more tolerant to the presence of TCE and its derivatives at 15 C than 37 C (Table 3). There are a number of possibilities to explain the relationship between SMA and toxicity thresholds for hydrogenotrophic methanogens observed during this study. The extent of TCE reduction to one or more DCE isomers, or even further to vinyl chloride, was not determined this trial and thus the concentration of one or more of these degradation derivatives may have exceeded the capabilities of the methanogenic community (IC50 values; 1,1 DCE 7.7, cis-1,2 DCE 19, trans-1,2 DCE 48; Blum and Speece, 1991). It is also possible, however, that the decreased hydrogenotrophic methanogenic activity recorded was due to competition with other organisms responsible for the dehalorespiration of TCE, for which H2 is an essential electron donor (Middeldorp et al., 1999), and with homoacetogenic bacteria (Kassenga et al., 2004), rather than by direct toxicity alone. In fact, Ballapragada et al. (1997) determined the half velocity coefficient of hydrogen utilization for the dechlorination of TCE (KsH2) to be 19 ppm, while the KsH2 for hydrogenotrophic methanogenesis was determined to be between 9 and 13 ppm (Lovley et al., 1994), thereby favouring the use of H2 for dechlorination. Furthermore, Ballapragada et al. (1997) also report successful competition for H2 by dechlorinators up to a H2 partial pressure of 100 ppm, a level seldom exceeded in methanogenic environments. In addition, Chin and Conrad (1995) reported that H2 is consumed mainly by methanogenesis at 30 C, but mainly by homoacetogenesis at 15 C, thus indicating that when methanogens, homoacetogens and dehalorespirators compete for H2, dehalorespirators are the most successful at both temperatures, with homoacetogens out competing methanogens at 15 C and vice versa at 37 C, thus supporting the hypothesis that H2 has reduced availability to hydrogenotrophic methanogens in bioreactors supplemented with TCE, particularly at lower temperatures. It should be noted that hydrogenotrophic activity in the control bioreactors remained unaffected (Table 2), reinforcing the point that changes to the hydrogenotrophic community was a TCE, and not a temperature, related phenomenon. At both 37 C and 15 C, the SMA data clearly demonstrated a direct inhibition of acetoclastic methanogens, as a result of the presence of TCE and/or its degradation derivatives. This was supported by the fact that acetate was the principal component of effluent COD from R1 and R3 and by the IC50 value for R1 biomass against TCE, of 40 mg l1 when assayed on day 235, when influent TCE concentrations had reached 60 mg l1 (Table 3). Reasons for specific inhibition of acetoclastic methanogens by TCE remain unclear, although Bae and Lee (1999) have previously shown a higher TCE toxicity effect on the acetoclastic communities of broken granules rather than intact granules, indicating that granule structure confers some essential resistance against TCE toxicity. Evidently, this is largely dependent
2460
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 5 2 e2 4 6 2
on the type and spatial arrangement of specific acetoclastic methanogenic populations in a granule, which greatly varies between seed biomass. However, notwithstanding the importance of granule structure this was apparently not to be the case in this study as granule integrity was not compromised at either temperature during the trial. Moreover, it has been shown previously that at lower temperatures hydrogenotrophic methanogens show a higher tolerance to organic toxicants than acetoclasts (Scully et al., 2006; Collins et al., 2005b). In general, this study suggested an inverse relationship between SMA and toxicity thresholds. For example, the SMA values of the seed biomass were significantly higher at 37 C than 15 C, but the IC50 values revealed a higher sensitivity of the methanogenic community in the seed biomass to TCE at 37 C than at 15 C (Tables 2 and 3). It is conceivable, therefore, that a higher specific methanogenic activity may indicate increased vulnerability of a biomass to the effects of a toxicant such as TCE and/or its degradation derivatives, possibly due to the growth phase of the biomass. This suggestion is supported by the better operational performance of the R3 bioreactor in response to elevated TCE levels during P5 and its more rapid recovery time after the withdrawal of TCE (Fig. 1); and the SMA and IC50 values of two additional anaerobic granular biomass samples assayed (Table 4). The principal driver of methanogenic community development in this study was apparently the nature of wastewater and reactor operating conditions (including temperature) and not the presence of TCE. Degradation of a volatile fatty acid based wastewater has previously been shown to drive changes in microbial community structure, for example, due to mutual inhibition phenomena (Vavilin and Lokshina, 1996). Statistical analysis of qPCR data of the gene copy concentrations of the methanogenic orders Methanobacteriales and Methanomicrobiales and the family Methanosaetaceae using NMS based on both absolute quantity and relative abundance matrices (Bialek et al., 2010; data not shown) revealed no strong relationship between methanogenic community structure and the presence of TCE (Fig. 2). There was, however, a significant change in the nature of the hydrogenotrohic community in the biomass of all four reactors during the trial, as the Methanomicrobiales emerged from relatively low numbers in the seed sludge (<106 copies g[VSS]1) to reach levels of between 107 and 109 in all digesters. Kim et al. (2010) described a statistically positive correlation between enhanced propionate methanogenesis and the emergent dominance of Methanomicrobiales in a reactor system. Our study suggests a similar finding, as propionate represented 25% of the influent to R1-R4 on a COD basis and was initially a predominant volatile fatty acid (VFA) represented in effluent COD from all bioreactors during P1, with concentrations ranging from 400 to 600 mg l1 but, from day 150 onwards, and associated with the emergence of the Methanomicrobiales, propionate concentrations were maintained at <50 mg l1 in all four reactor effluents, while acetate was the principal component of effluent COD, and in particular during process perturbations (data not shown). The reason for the putative association between the Methanomicrobiales and enhanced propionate methanogensis is unknown, but it is possible that syntrophic associations between these organisms and syntrophic propionate oxidizers are favoured under the process conditions applied in this study,
The rate of emergence of Methanomicrobiales was apparently enhanced at lower temperatures, as the gene copy concentrations of these organisms were significantly higher in the biomass of R3 and R4 during this trial on days 191, 224 and 235 and (Fig. 2). The predominance of hydrogenotrophic methangens, and specifically the emergence of members of the Methanomicrobiales as dominant methanogenic populations during low-temperature anaerobic digestion has been described previously by a number of authors (Connaughton et al., 2006; McHugh et al., 2004), who also reported on reactors treating VFA-based influents. These authors suggested that the emergence of these organisms have been due to enhanced thermodynamic conditions for hydrogenotrophic methanogenesis at lower temperatures and a greater capacity for activity at lower temperatures by the Methanomicrobiales. The potential role of influent propionate in the selection of this group, however, should now be investigated further. Previous reports of high abundance of Methanosaetaceae in stable anaerobic granular systems (Diaz et al., 2006; Satoh et al., 2007; O’Reilly et al., 2010) were supported by this study. The abundance of the acetoclastic family Methanosaetaceae demonstrated a transient, temperature-dependent response to TCE, with significant reductions in gene copy numbers observed on day 191 at 37 C, corresponding to influent TCE levels of 20 mg l1 (Fig. 2), but by day 235 Methanosaetaceae 16S rRNA gene copies gVSS1 in R1 had recovered to levels similar to that of R2. No significant response to the addition of TCE was detected with respect to the population levels of the Methanosaetaceae community at 15 C, suggesting that at lower temperatures these acetoclastic methanogens are more tolerant to the presence of TCE, as indeed also suggested by IC50 data. The gene copy numbers of Methanosaetaceae had significantly decreased in R3, however, by the conclusion of the trial. Although the alternative acetoclastic methanogens, the Methanosarcinaceae, were not detected above the lower limit of the qPCR assay applied here (106), they still may have been present at sufficient concentrations to play some role in methane production from acetate. Despite the clear impact on reactor performance of TCE addition, and the presence of high concentrations of acetate in the effluents of R1 and R3 during process perturbations, and the SMA and toxicity data, no statistical relationships (Fig. 2) or clear association using NMS (data not shown) could be made between acetoclastic methanogenic community structure and the addition of TCE to the influents of R1 and R3. This study illustrated some potential difficulties with respect to the integration of quantitative molecular, physiological and anaerobic bioprocess data, therefore, despite the findings of previously successful studies (Bialek et al., 2010; O’Reilly et al., 2010; Lee et al., 2009; McKeown et al., 2009). In addition, the value of specific activity and toxicity profiles as tools to evaluate the trophic status of anaerobic sludge exposed to toxicants was highlighted. For example, despite a significantly reduced hydrogenotrophic SMA profile in the mesophilic reactor (R1) following TCE addition on day 235, no significant relationship between SMA and the abundance of the hydrogenotrophic methanogens was recorded. In fact, on day 235 the SMA of the control reactor biomass (R2) was >10fold higher than R1, whereas the levels of Methanobacteriales were >10-fold higher in R1 than in the control R2.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 5 2 e2 4 6 2
Furthermore, despite the development of enhanced acetoclastic methangenic activity at 15 C in the R4 bioreactor during the trial, the gene copy concentrations of the Methanosaetaceae were virtually identical to that of the seed biomass and no emergence of the Methanosarcinaceae was noted. The apparent disparity in results between the molecular and physiological data noted in this study maybe due to the detection of inactive organisms by the DNA-based qPCR assays. It is plausible that, for example, the addition of TCE did not quantifiably affect microbial community structure, but instead strongly influenced the functions of the microbes present, which was evident in SMA profiles. Previous studies have also shown that the development of psychrotolerance, or enhanced methanogenic activity at 15 C, can occur in anaerobic bioreactors, without major shifts in methanogenic community structure (Collins et al., 2003; McHugh et al., 2003). To unravel the nature of the response of reactor biomass to TCE or temperature decreases, therefore, analysis of the bacterial community structure, using RNA, rather than DNA-based approaches; or functional investigations through the use of recently developed metaproteomic analyses for anaerobic bioreactor biomass (Abram et al., 2009), is recommended for future studies.
5.
Conclusions
The mesophilic biomass utilized in this study was capable of continued methanogenic activity at an influent TCE concentration of <60 mg l1, at both 37 C and 15 C. At an influent concentration of 60 mg l1, acetoclastic methanogens were directly inhibited by the presence of TCE or its degradation derivatives, while hydrogenotrophic methanogens may have been unsuccessful in the competition for H2 against other trophic groups. Methanogenesis and methanogenic community structure was strongly influenced by wastewater composition and also impacted by temperature. An inverse correlation between SMA and IC50 values against TCE was suggested, suggesting that higher activity levels render microbial communities more vulnerable to the effects of this toxicant. Further research is required to determine the nature and extent of TCE degradation during anaerobic digestion and the important microbial species involved in this process.
Acknowledgements The financial support of Enterprise Ireland, The Irish Environmental Protection Agency and Science Foundation Ireland is acknowledged. Assistance with statistical analysis by Dr. Changsoo Lee is gratefully acknowledged.
references
Abram, F., Gunnigle, E., O’Flaherty, V., 2009. Optimisation of protein extraction and 2-DE for metaproteomics of microbial
2461
communities from anaerobic wastewater treatment biofilsm. Electrophoresis 30, 4149e4151. APHA, 1998. Standard methods for the examination of water and wastewater. In: Clesceri, L.S., Greenberg, A.E., Eaton, A.D. (Eds.), twentieth ed. American Public Health Association, Washington, DC, U.S.A. ATSDR, 2003. ATSDR, Division of Toxicology toxFAQs. Agency for Toxic Substances and Disease Registry. Atlanta, GA, USA. Bae, J.W., Lee, S.T., 1999. Layered structure of UASB granules gives microbial populations resistance to toxic chemicals. Biotechnol. Lett. 21, 159e162. Ballapragada, B.S., Stensel, H.D., Puhakka, J.A., Ferguson, J.F., 1997. Effect of hydrogen on reductive dechlorination of chlorinated ethenes. Environ. Sci. Technol. 31, 1728e1734. Bialek, K., Kim, J., Lee, C., Collins, G., Mahony, T., O’Flaherty, V., 2010. Quantitative and qualitative analyses of methanogenic community development in high-rate anaerobic bioreactors. Water Res. 45 (3), 1298e1308. Blum, D.J.W., Speece, R.E., 1991. A database of chemical toxicity to environmental bacteria and its use in interspecies comparisons and correlations. Res. J. Water Pollut. C. 63 (3), 198e207. Carter, S.R., Jewell, W.J., 1993. Biotransformation of tetrachloroethylene by anaerobic attached-films at low temperature. Water Res. 27, 607e615. Chin, K.J., Conrad, R., 1995. Intermediary metabolism in methanogenic paddy soil and the influence of temperature. FEMS Microbiol. Ecol. 18, 85e102. Coates, J.D., Coughlan, M.F., Colleran, E., 1996. Simple method for the measurement of the hydrogenotrophic methanogenic activity of anaerobic sludges. J. Microbiol. Method 26, 237e246. Colleran, E., Pistilli, A., 1994. Activity test system for determining the toxicity of xenobiotic chemicals to the methanogenic process. Ann. Microbiol. Enzymol. 44, 1e18. Colleran, E., Concannon, F., Golden, T., Geoghan, F., Crumlish, B., Killilea, J.E., Henry, M., Coates, J., 1992. Use of methanogenic activity tests to characterize anaerobic sludges, screen for anaerobic biodegradability and determine toxicity thresholds against individual anaerobic trophic groups and species. Water Sci. Technol. 25, 31e40. Collins, G., Woods, A., McHugh, S., Carton, M.W., O’Flaherty, V., 2003. Microbial community structure and methanogenic activity during start-up of psychrophilic anaerobic digesters treating synthetic industrial wastewaters. FEMS Microbiol. Ecol. 46, 159e170. Collins, G., McHugh, S., Kearney, A., O’Flaherty, V., 2004. Psychrophilic anaerobic digestion of a range of wastewaters: process technology, microbial activity and population dynamics. Wat. Intell. Online 40 (6). Collins, G., Foy, C., McHugh, S., O’Flaherty, V., 2005a. Anaerobic treatment of 2,4,6-trichlorophenol in an expanded granular sludge bed-anaerobic filter (EGSB-AF) bioreactor at 15 C. FEMS Microbiol. Ecol. 53, 167e178. Collins, G., Foy, C., McHugh, S., Mahony, T., O’Flaherty, V., 2005b. Anaerobic biological treatment of phenolic waste-water at 15e18 C. Water Res. 39, 1614e1620. Connaughton, S., Collins, G., O’Flaherty, V., 2006. Psychrophilic and mesophilic anaerobic digestion of brewery effluent: a comparative study. Water Res. 40, 2503e2510. Diaz, E.E., Stams, A.J., Amils, R., Sanz, J.L., 2006. Phenotypic properties and microbial diversity of methanogenic granules from a full-scale upflow anaerobic sludge bed bioreactor treating brewery wastewater. Appl. Environ. Microb. 72, 4942e4949. Doherty, R.E., 2000. A history of the production and use of carbon tetrachloride, tetrachloroethylene, trichloroethylene and 1,1,1-trichloroethane in the United States: part 2Trichloroethylene and 1,1,1-trichloroethane. Environ. Forensics 1, 83e93.
2462
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 5 2 e2 4 6 2
Doucette, W.J., Chard, J.K., Fabrizius, H., Crouch, C., Petersen, M.R., Carlsen, T.E., Chard, K.B., Gorder, K., 2007. Trichloroethylene uptake into fruits and vegetables: three-year field monitoring study. Environ. Sci. Technol. 41, 2505e2509. Droste, R.L., Kennedy, K.J., Lu, J., Lentz, M., 1998. Removal of chlorinated phenols in upflow anaerobic sludge blanket reactors. Water Sci. Technol. 38, 359e367. Enright, A.M., McHugh, S., Collins, G., O’Flaherty, V., 2005. Lowtemperature anaerobic biological treatment of solventcontaining pharmaceutical wastewater. Water Res. 39, 4587e4596. Enright, A.M., Collins, G., O’Flaherty, V., 2007. Low-temperature anaerobic biological treatment of toluene-containing wastewater. Water Res. 41, 1465e1472. Ferguson, J.F., Pietari, J.M.H., 2000. Anaerobic transformations and bioremediation of chlorinated solvents. Environ. Pollut. 107, 209e215. Freedman, D.L., Gossett, J.M., 1989. Biological reductive dechlorination of tetrachloroethylene and trichloroethylene to ethylene under methanogenic conditions. Appl. Environ. Microbiol. 55 (9), 2144e2151. Kassenga, G., Pardue, J.H., Moe, W.M., Bowman, K.S., 2004. Hydrogen thresholds as indicators of dehalorespiration in constructed treatment wetlands. Environ. Sci. Technol. 38, 1024e1030. Kettunen, R.H., Rintala, J.A., 1997. The effect of low temperature (5e29 C) and adaptation on the methanogenic activity of biomass. Appl. Microbiol. Biotechnol. 48, 570e576. Kim, W., Lee, S., Gu Shin, S., Lee, C., Hwang, K., Hwang, S., 2010. Methanogenic community shift in anaerobic batch digesters treating swine wastewater. Water Res. 44, 4900e4907. Lee, C., Kim, J., Hwang, K., O’Flaherty, V., Hwang, S., 2009. Quantitative analysis of methanogenic community dynamics in three anaerobic batch digesters treating different wastewaters. Water Res. 43, 157e165. Lettinga, G., Rebac, S., Parshina, S., Nozhevnikova, A., van Lier, J., Stams, A., 1999. High-rate anaerobic treatment of wastewater at low temperatures. Appl. Environ. Microbiol. 65 (4), 1696e1702. Lovley, D.R., Chapelle, F.H., Woodward, J.C., 1994. Use of dissolved H2 concentrations to determine distribution of microbially catalyzed redox reactions in anoxic groundwater. Environ. Sci. Technol. 28 (7), 1205e1210. McHugh, S., O’Reilly, C., Mahony, T., Colleran, E., O’Flaherty, V., 2003. Anaerobic granular sludge bioreactor technology. Rev. Environ. Sci. Biotechnol. 2, 225e245. McHugh, S., Carton, M., Collins, G., O’Flaherty, V., 2004. Reactor performance and microbial community dynamics during anaerobic biological treatment of wastewaters at 16,037 C. FEMS Microbiol. Ecol. 48, 369e378. McKeown, R.M., Collins, G., Chinalia, F.A., Mahony, T., O’Flaherty, V., 2008. Low temperature anaerobic biotreatment of priority pollutants. Water Sci. Technol. 57 (4), 499e503. McKeown, R.M., Scully, C., Enright, A.-M., Chinalia, F.A., Lee, C., Mahony, T., Collins, G., O’Flaherty, V., 2009. Psychrophilic methanogenic community development during long-term cultivation of anaerobic granular biofilms. ISME J. 3 (11), 1231e1242. The Merck Index, 2006. Entry 9639: Trichloroethylene, fourteenth ed..
Middeldorp, P.J.M., Luijten, M.L.G.C., van de Pas, B.A., van Eekert, M.H.A., Kengen, S.W.M., Schraa, G., Stams, A.J.M., 1999. Anaerobic microbial reductive dehalogenation of chlorinated ethenes. Bioremediat J. 3 (3), 151e169. O’Reilly, J., Lee, C., Chinalia, F., Collins, G., Mahony, T., O’Flaherty, V., 2010. Microbial community dynamics associated with biomass granulation in low-temperature (15 C) anaerobic wastewater treatment bioreactors. Bioresour. Technol. 101, 6336e6344. Ozdemir, C., Dursun, S., Karatas, M., Sen, N., Sahinkaya, S., 2007. Removal of trichloroethylene (TCE) in upflow anaerobic sludge blanket reactors (UASB). Biotechnol. Biotec. Eq. 21 (1), 107e112. Rebac, S., Ruskova, J., Gerbens, S., van Lier, J.B., Stams, A.J.M., Lettinga, G., 1995. High-rate anaerobic treatment of wastewater under psychrophilic conditions. J. Ferment. Bioeng. 80 (5), 499e506. Rivett, M.O., Lerner, D.N., Lloyd, J.W., 1990. Chlorinated solvents in UK aquifers. J. Inst. Water Environ. Manag. 4, 242e250. Russell, D.L., 1992. Remediation Manual for PetroleumContaminated Sites. Technomic Publishing Co, Lancaster, PA., USA. Satoh, H., Miura, Y., Tsushima, I., Okabe, S., 2007. Layered structure of bacterial and archaeal communities and their insitu activities in anaerobic granules. Appl. Environ. Microb. 73, 7300e7307. Scully, C., Collins, G., O’Flaherty, V., 2006. Anaerobic biological treatment of phenol at 9.5e15 C in an expanded granular sludge bed (EGSB)-based bioreactor. Water Res. 40, 3737e3744. Shelton, D.R., Tiedje, J.M., 1984. General method for determining anaerobic biodegradation potential. Appl. Environ. Microbiol. 47, 850e857. Speece, R.E., 1983. Anaerobic biotechnology for industrial wastewater treatment. Environ. Sci. Technol. 17 (9), 416Ae427A. Sponza, D.T., 2003. Enhancement of granule formation and sludge retainment for tetrachloroethylene (TCE) removal in an upflow anaerobic sludge blanket (UASB) reactor. Adv. Environ. Res. 7, 453e462. US EPA, 1997. Technology Transfer Network Air Toxics Web Site [online]. Available. http://www.epa.gov/ttn/atw/hlthef/triethy.html (accessed 01.07.10). van Agteren, M.H., Keuning, S., Janssen, D.B., 1998. Handbook on Biodegradation and Biological Treatment of Hazardous Organic Compounds. Kluwer Academic Publishers, The Netherlands. Vavilin, V.A., Lokshina, L.Y., 1996. Modeling of volatile fatty acids degradation kinetics and evaluation of microorganism activity. Bioresour. Technol. 57, 69e80. ı´k, V., Hoffmann, J., R ka, J., Sergejevova´, M., 2005. ic Volc uz Trichloroethylene (TCE) removal in a single pulse suspension bioreactor. J. Environ. Manage. 74, 293e304. Wu, W.M., Nye, J., Jain, M.K., Hickey, R., 1997. Anaerobic dechlorination of trichloroethylene (TCE) to ethylene using complex organic materials. Water Res. 32 (5), 1445e1454. Yu, Y., Lee, C., Jaai, K., Hwang, S., 2005. Group-specific primer and probe sets to detect methanogenic communities using quantitative real-time polymerase chain reactions. Biotechnol. Bioeng. 89, 670e679.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 6 3 e2 4 7 2
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Performance parameter prediction for sewage sludge digesters using reflectance FT-NIR spectroscopy J.P. Reed*, D. Devlin, S.R.R. Esteves, R. Dinsdale, A.J. Guwy Sustainable Environment Research Centre, Faculty of Health, Sport and Science, University of Glamorgan, Pontypridd, Wales CF37 1DL, UK
article info
abstract
Article history:
This study investigates the use of Fourier transform near infrared (FT-NIR) spectroscopy
Received 9 December 2010
combined with principle components analysis (PCA) and partial least squares regression
Received in revised form
(PLS-R) as part of a possible process monitoring system for sewage sludge anaerobic
28 January 2011
digesters. The ability of FT-NIR with PCA to distinguish between different stages of the AD
Accepted 31 January 2011
process was investigated, it was found that waste activated sludge (WAS), primary, feed
Available online 12 February 2011
(Primary:WAS 70:30) and digested sludge were distinguishable from each other using this technique. PLS-R was used successfully to track differing proportions of primary:WAS in
Keywords:
feedstocks of 5% total solids (Coefficient of Efficiency (CE) ¼ 0.93). The study also looked at
Anaerobic digestion
the ability of reflectance mode NIR spectroscopy to track process parameters important for
Sewage sludge
stability. Temperature and organic loading rate variations were employed to stress the
Reflectance FT-NIR
digesters. Predictive models were produced for volatile fatty acids (VFA), bicarbonate
Solids
alkalinity (BA) and total and volatile solids (TS and VS) and independently validated for
Alkalinity
each digester. The models were able to track the relevant process parameters: TS
VFA
(CE ¼ 0.75), VS (CE ¼ 0.75), BA (CE ¼ 0.71), and VFA (CE ¼ 0.69). This technique could be used
Multivariate analysis
to improve the performance of sewage sludge anaerobic digesters. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
1.1.
Industrial relevance
Anaerobic digesters have been commonly employed in wastewater treatment plants (WWTPs) in Europe for treating sewage sludges. In the UK, digestion accounts currently for over 60% of sewage sludge treatment. This process not only reduces the solids content but also decreases odours of the incoming sewage sludge before eventual re-use or disposal. A beneficial by-product of this process is the production of a methane rich biogas, which can be utilised onsite for electricity and heat generation used to support the operation of the WWTP or it can be upgraded and injected into the gas grid or used as a transport fuel.
Anaerobic digestion is a multi-stage bioprocess that requires the interaction of at least 4 trophic groups of bacteria. For stable operation it is required that the process is kept at an equilibrium where intermediate by-products are able to be utilised successfully by the bacterial consortia. Failure of anaerobic digesters can occur when the acetogenic bacteria and obligate-anaerobes (methanogens) are unable to metabolise the volatile fatty acids (VFA) produced by the acidogenic bacteria fast enough, which then leads to a build up of VFAs and a corresponding drop of bicarbonate alkalinity (BA) and eventually of the pH inhibiting the methanogens further and eventually leading to a collapse of the process. To mitigate against this possibility, anaerobic digesters are operated with organic loading rates that are below the optimum for costeffectiveness (Dixon et al., 2007). Monitoring of anaerobic
* Corresponding author. Tel.: þ44 1443 654389. E-mail address:
[email protected] (J.P. Reed). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.01.027
2464
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 6 3 e2 4 7 2
digesters can take place in the solid, liquid or gas phase. Considerable efforts have been directed at providing on-line sensors, such as organic content of feedstocks as well as BA or VFA concentration in the digester (Esteves et al., 2000; Guwy et al., 1994; Spanjers and van Lier, 2006). Typical anaerobic digesters will employ a very limited number of sensors to monitor temperature and gas flows and in some cases gas quality, but all other biochemical analysis of feedstocks or digestate characteristics or reactor performance are typically carried out sporadically off-line. Often there is no closed loop control employed for any of the parameters monitored. It has been shown that on-line monitoring of key process intermediates, such as VFAs or the level of bicarbonate alkalinity in the digester allow optimisation of the process and improved performance (Steyer et al., 2002; Lardon et al., 2004). There is also a major benefit if one on-line sensor or model is able to predict more than one state and performance variable of the process. With increased investment in centralised WWTPs and centralised sewage sludge digestion plants, a tool for rapid online analysis of the quality and composition of the incoming waste streams and digestates as well as reactor performance could be valuable. These facilities import sludges from the surrounding area to treat them centrally at a single site and mixing of sludges for digestion is in many cases performed on ad hoc basis. As such the sludge quality may vary more widely than at a traditional site. Operators may also like to assess the stage of treatment of outgoing wastes in order to ensure that the wastes have been treated to a certain level before being released from the site. For digesters operating on mixed sewage sludges i.e. primary and waste activated sludges (WAS), an important consideration is in the regulation of feeding mixtures. To avoid foaming due to the transfer of filamentous bacteria to the digester from the activated sludge process, it has been suggested that WAS should not make up more than 40% of the feedstock (Massart et al., 2006). Non-complete degradation of volatile solids leads to the presence of hydrophobic substances that promote foaming (Ganidi et al., 2009). The physical state of microbial cells in WAS makes it difficult to digest, as such pretreatment of WAS may be advantageous (Muller, 2000). However the pretreatment may have no effect on digestibility of primary sludge thus incurring an energy cost when feed mixtures are pretreated together. A tool that may be able to indicate the need/value of carrying out pretreatments or a tool that can effectively identify types of sludges to be digested separately or provide an indication that a longer retention time would aid digestion could maximise treatment efficiencies, reduce costs and energy use, and increase biogas energy.
process monitoring of laboratory scale municipal solid waste (MSW) digesters (Hansson et al., 2002), the monitoring of VFA during co-digestion of manures and maize silage (Lomborg et al., 2009), and activated sludge process monitoring (Dias et al., 2008). Whilst these studies report positive outcomes, the application of FT-NIR monitoring schemes to on-line analysis of sewage sludge is not as straightforward as its implementation for the food and pharmaceutical industries owing to a number of factors. One difficulty in the use of FT-NIR analysis for AD monitoring is the low proportion of dry matter in the samples, typically 5e8%. Water content has been stated to be problematic as water has intense absorbance bands in the NIR region that may interfere with spectral characteristics of interest of the samples (Xie et al., 2009). A solution to this difficulty would be to dry the samples before scanning, but this would preclude the use of the technique for on-line analysis and provide little benefit over traditional methods of analysis. There are three techniques commonly used in FT-NIR spectroscopy for scanning samples. These are the transmittance, transflectance and reflectance methods. Transmittance methods can be excluded for use with sewage sludge due to the turbidity of the samples involved, the particle size distribution, and heterogeneous nature of bio-solids. The requirement for short optical path lengths in order to avoid saturation of the signal makes transmittance techniques impractical in an on-line situation. The Transflexive Embedded Near Infrared Sensor (TENIRS) (Holm-Nielsen et al., 2007a) is an example of the use of the transflectance technique with a flow through cell. This technique though producing good results is dependent on the type of feedstock to be investigated. This is due to the requirement of a small optical path length (3 mm). Where the system has been used in an at-line configuration, manure has been screened through a 3 mm sieve before digestion (Holm-Nielsen et al., 2007a), a 4 mm sieve (Lomborg et al., 2009),or macerated prior to being analysed using the TENIRS system (HolmNielsen et al., 2007b). Saeys et al. (2005) investigated the use of transflectance and reflectance FT-NIR spectroscopy on pig manures. Their results showed that the transflectance technique outperformed the reflectance technique, however they noted that for on-line applications the choice of principle might depend on its physical limitations, transflectance requiring the use of a small optical path length (1 mm), whilst reflectance techniques do not. This is an important consideration to make as sewage sludges by their nature will exclude the use of small optical path lengths due to the possibility of fouling.
1.3. 1.2.
Multivariate analysis techniques
Monitoring technique
Fourier transform near infrared analysis (FT-NIR) combined with multivariate statistical techniques have been used in the food industry for the past 35 years and subsequently in the pharmaceutical, agricultural and petrochemical industries (Blanco and Villarroya, 2002). In recent years there has been increasing interest in the technique for analysis of bio-reactors. Studies have been carried out on the use of FT-NIR on
Multivariate analysis methods have been used to analyse bio-reactors. These are as follows; Principle Component Analysis (PCA) (Jolliffe, 2002) and Partial Least Squares Regression (PLS-R) (Wold et al., 2001). PCA has been used to track the progress of the anaerobic digestion of MSW (Hansson et al., 2003). By plotting scores of individual principle components (PC) against each other Hansson shows that it is possible to identify periods of
2465
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 6 3 e2 4 7 2
instability of the digester due to planned and unplanned disturbances. PCA has also been used with Fourier transform infrared spectroscopy (FT-IR) to demonstrate the ability of identifying different stages of sludge treatment within a WWTP, though this study used pre-treatments on the samples that included drying and sieving thus not applicable on-line (Smidt and Parravicini, 2009). Dias et al. (2008) used transflectance NIR spectroscopy (5 mm path length) combined with PCA to monitor an activated sludge process. They showed that it was possible to identify the moment a disturbance to the process occurred and then follow the path of the process to a new equilibrium point. PLS-R has been used to produce predictive models of performance indicators such as VFAs, chemical oxygen demand (COD), total and volatile solids (TS and VS). VFAs have been modelled using PLS-R for manure, maize silage digestion, and MSW. (Jacobi et al., 2009) Hansson et al. (2002) used PLS-R to produce a model for prediction of propionate from MSW digestion and determined that the NIR detection limit for propionate was 300 mg l1. TS and VS have been investigated using PLS-R for manure, and maize silage. COD has been investigated with FT-IR for sewage sludge (Smidt and Parravicini, 2009). This type of analysis has utility for at-line monitoring of bioreactor processes or the online monitoring and potential control of bioprocesses (HolmNielsen et al., 2007a; Jacobi et al., 2009). It is clear that for the purposes of on-line monitoring of sewage sludge anaerobic digesters, that the reflectance method of gathering spectra shows the greatest promise of the three scanning techniques from a practical standpoint. This study will investigate the use of reflectance mode FT-NIR analysis for the anaerobic digestion process treating sewage sludges and will investigate the use of multivariate statistics in order to construct models that can be used for process monitoring as well as characterisation of feedstocks and digestates.
2.
Methodology
2.1.
Sludge characterisation
The TS and VS of the sewage sludge samples were determined according to American Public Health Association (APHA) standard methods (1998). BA analyses were carried out using the method as proposed by Jenkins et al. (1983). VFAs were measured according to Cruwys et al. (2002) and expressed as mg l1 acetic acid equivalent.
2.2.
Semi-continuous anaerobic digesters
prepared based on VS content by mass followed by addition of water to result in a TS content of 5% by mass for the feedstock. The digesters were continuously stirred, initially batch fed once a day and were operated at a 15 day hydraulic retention time (HRT). The reactors were situated in a temperature controlled bath and operated initially at 35 C. Three identically operated reactors were used in order to generate a large number of samples over the given time period. The use of 3 independent but identically operated reactors allowed independent validation of the statistical models developed, instead of relying on cross-validation. The reactors were operated as per conditions defined in Table 1. These conditions were defined in order to investigate reactor ‘stress’ related conditions without any external biochemical inputs that could interfere with the FT-NIR monitoring response. The reactors were run for a total of 57 days. Digestate samples were taken twice daily, generally once in the morning immediately prior to feeding and once again at the end of the afternoon. All samples were maintained at the reactor bath temperature and scanned using the FT-NIR instrument in reflectance mode before being analysed for TS, VS, BA and VFAs.
2.3. Preparation of primary and WAS mixtures for feedstock characterisation Mixtures of primary sludge and WAS were made using the same method as for the digester feed. In this case the ratio of primary to WAS was varied over the range 0e100% in increments of 10%. All mixtures were adjusted to produce a sludge with a TS content of 5%. The fresh mixtures were then scanned using the FT-NIR instrument for analysis.
2.4.
Sample spectra were acquired using a Perkin Elmer 100N FT-NIR spectrometer fitted with a Near Infrared Reflectance
Table 1 e Operating conditions for the three semicontinuous reactors. No. 1 2 3 4
Three 22 l semi-continuous anaerobic digesters were operated under identical operating conditions. The seed for the reactors and feedstocks were collected from a working full-scale anaerobic digester treating sewage sludges. Feedstock collection took place four times in total over the course of the experiment. The feedstock was mixed from primary and WAS to match a ‘typical’ ratio used at a full-scale digestion plant. As the seed and feed were collected from full-scale plants with a similar operating procedure, no start up procedure was utilised. This mixture of 70:30 primary to WAS ratio was
Spectra acquisition
5 6 7 8
Operating conditions Mixed sewage sludges, mean OLRa ¼ 2.6 g VS l1 d1, 35 C reactor, 15 days HRT e day 1e11 WAS only, mean OLR ¼ 2.8 g VS l1 d1, 35 C reactor, 15 days HRT e day 12e13 Mixed sewage sludges, mean OLR ¼ 2.8 g VS l1 d1, 35 C reactor, 15 days HRT e day 14e25 Mixed sewage sludges, mean OLR ¼ 2.6 g VS l1 d1, 3035 C reactor, 15 days HRT e day 26e33 Mixed sewage sludges, mean OLR ¼ 2.6 g VS l1 d1, 3535 C reactor, 15 days HRT e day 34e39 Mixed sewage sludges, mean OLR ¼ 2.5 g VS l1 d1, 2535 C reactor, 15 days HRT e day 40e49 Mixed sewage sludges, mean OLR ¼ 5.7 g VS l1 d1, 25 C reactor, 7.5 days HRT e day 50e54 Mixed sewage sludges, mean OLR ¼ 5.8 g VS l1 d1, 2035 C reactor, 7.5 days HRT e day 55e57
a OLR ¼ Organic Loading Rate.
2466
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 6 3 e2 4 7 2
Accessory (NIRA). 25 ml samples were transferred to the NIRA for scanning. Samples were scanned over the range of 4000 cm1 to 12,000 cm1 at a resolution of 16 cm1. Each sample was scanned 250 times and the resulting spectra averaged to increase the signal to noise ratio of the spectra.
Primary Sludge
2.5.
Total Solids 6.32e7.83 Volatile Solids 4.94e5.98
Data analysis
PCA analysis was carried out for all groups of samples scanned (Digested, Feed, Primary, and WAS) using SOLO 4.1 software (Eigenvector Research Incorporated). No pre-processing of the spectra was used other than mean centring to comply with the needs of the PCA method. PLS-R using the SIMPLS algorithm (de Jong, 1993) as implemented within the SOLO 4.1 software was carried out on digested sludge samples and models were produced for bicarbonate alkalinity, TS, VS, and VFAs. For each variable three models were produced, one for each of the reactors. This allows the models to be compared to each other for consistency. The similarity, or otherwise, of the performance of the resulting three models allows an additional measure of confidence in the model performance. In each case, data from 2 reactors (156 samples) was used to calibrate the model using a cross-validation method of random subsets. The number of latent variables (LV) retained for each model was determined by the software. The data from the third reactor (78 samples) was then used as an independent validation of the model. Where this meant that validation samples were outside the range of the calibration set, those samples were not excluded but were kept in as part of the validation. This was to mimic what would happen in reality when the model would be applied to spectra with unknown characteristics. This provides an independent validation of the models and is more likely to represent future model performance than cross-validation and artificial control of the sample set ranges. The pre-processing applied to the spectra for each model are detailed in Table 4. For each characteristic modelled the pre-processing employed was the same for each reactor. Pre-processing techniques employed were 1st derivative (1st Deriv) and Savitzky-Golay smoothing (Sav-Gol) with a window of 151 pts. In all cases where the 1st derivative was taken, this was carried out with Sav-Gol smoothing as taking the 1st derivative emphasises high frequency noise contained by the spectra. Mean centring was carried out in all cases as a requirement of the PLS-R method. In the case of the models for TS and VS it was found that removing regions of the spectra corresponding to wavenumbers > 8932 cm1 and < 4156 cm1 improved the performance of the model.
2.6.
Table 2 e Sludge characteristics. Range (g∙l1) Mean (g∙l1)
Standard Deviation (g∙l1)
7.21 5.51
0.58 0.37
Mean (g∙l1)
Standard Deviation (g∙l1) 0.63 0.45
WAS Range (g∙l1) Total Solids 4.44e6.18 Volatile Solids 3.56e4.82
5.56 4.52
Digested Sludge Mean Standard Bicarbonate Range Alkalinity (mg CaCO3∙l1) (mg CaCO3∙l1) Deviation (mg CaCO3∙l1) Reactor 1 1231.25e3495.00 2930.77 480.30 Reactor 2 1191.25e3852.50 2926.88 480.11 Reactor 3 1268.75e3722.50 2930.89 531.81 Total Solids
Range (g∙l1)
Mean (g∙l1)
Reactor 1 Reactor 2 Reactor 3
3.02e3.79 2.53e3.79 2.96e3.81
3.30 3.29 3.31
Standard Deviation (g∙l1) 0.17 0.20 0.19
Volatile Solids Range (g∙l1)
Mean (g∙l1)
Reactor 1 Reactor 2 Reactor 3
1.97e2.61 1.67e2.60 1.92e2.60
2.22 2.22 2.22
Volatile Fatty Acids Reactor 1 Reactor 2 Reactor 3
Range (mg∙l1)
Mean (mg∙l1) Standard Deviation (mg∙l1) 229.47 332.05 223.67 289.81 229.26 299.83
28.63e1469.35 24.23e1451.35 26.08e1460.41
Standard Deviation (g∙l1) 0.14 0.16 0.15
predictions match the samples but does not necessarily provide information on the models ability to predict accurately. For example, it is possible to produce a model with R2 ¼ 1 which predicts badly if the model suffers from bias (Weglarczyk, 1998). RMSEP provides a measure of the expected error of the prediction. However, without knowledge of the range of the sample set it does not give useful information on model utility, as the usefulness of the model for a given application will be determined by the ratio of the range to the RMSEP.
RMSEP ¼
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2 uP u yp ym t
(1)
n
Performance assessment
Several methods of assessing the performance of the models are reported. The squared correlation coefficient (R2), the root mean squared error of prediction (RMSEP) (Eq. (1)), the ratio of the sample standard deviation (SD) to the standard error of prediction (SEP) (Eq. (3)) called RPD (Eq. (2)), and coefficient of efficiency (CE) (Eq. (5)). R2 is a measure of the amount of variance in the sample set explained by the model. It provides a measure of how well a particular linear function of the
Table 3 e Squared correlation coefficients between digestate sample characteristics.
TS VS BA VFA
TS
VS
BA
VFA
e
0.91 e
0.16 0.13 e
0.05 0.01 0.62 e
2467
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 6 3 e2 4 7 2
Table 4 e PLS-R summary of results. Validation set
Pre-processing
No. of LVs
R2
RPD CE 1
Bicarbonate alkalinity Reactor 1 Sav-Gol smooth 151pt window, mean centre Reactor 2 Sav-Gol smooth 151pt window, mean centre Reactor 3 Sav-Gol smooth 151pt window, mean centre
3 3 3
0.71 0.72 0.71
mg CaCO3 l 257.67 258.84 288.42
1.87 0.71 1.89 0.71 1.86 0.70
0.70 0.77 0.76
g l1 0.096 0.098 0.095
1.79 0.68 2.06 0.74 2.05 0.75
3 3 3
0.60 0.70 0.76
g l1 0.087 0.089 0.073
1.56 0.58 1.87 0.67 2.04 0.75
1st Deriv Sav-Gol 151pt window, mean centre 1st Deriv Sav-Gol 151pt window, mean centre 1st Deriv Sav-Gol 151pt window, mean centre
7 7 7
0.69 0.69 0.71
mg l1 184.89 169.89 164.86
1.80 0.69 1.74 0.65 1.86 0.69
Mean centre
5
0.97
% 7.09
7.65 0.93
Total solids Reactor 1 Reactor 2 Reactor 3
1 Deriv Sav-Gol 151pt window, mean centre, wavenumbers 4156e8932 1st Deriv Sav-Gol 151pt window, mean centre, wavenumbers 4156e8932 1st Deriv Sav-Gol 151pt window, mean centre, wavenumbers 4156e8932
3 3 3
Volatile solids Reactor 1 Reactor 2 Reactor 3
1st Deriv Sav-Gol 151pt window, mean centre, wavenumbers 4156e8932 1st Deriv Sav-Gol 151pt window, mean centre, wavenumbers 4156e8932 1st Deriv Sav-Gol 151pt window, mean centre, wavenumbers 4156e8932
Total VFAs Reactor 1 Reactor 2 Reactor 3
st
Proportion of WAS
Where ym ¼ measured value and yp ¼ predicted value. The ratio of the sample standard deviation to the standard error of prediction (RPD) (Williams, 1987) is the factor by which the model improves over the simplest predictive model available, the use of the mean. RPD ¼
SD SEP
SEP ¼
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi RMSEP2 Bias2
the commonly used RPD. In combination with R2 it provides a good measure of model performance as a model with high CE will always have a high R2, but a model with high R2, as discussed above, would not necessarily have a high CE. In general, a good model will have high values for R2 and CE, and a low RMSEP.
(2)
(3)
where
Bias ¼
RMSEP
P yp ym
(4)
n
A model with an RPD equal to one is as effective as the mean at predicting values. The relationship between RPD and model performance is application specific. Where Williams suggests a scale of interpretation from 0 to 8.1þ, Saeys et al. (2005) suggest a scale of interpretation of prediction accuracy based on the RPD value for FT-NIR reflectance analysis of pig manure from 0 to 3þ. A further measure of model performance is the coefficient of efficiency (Nash and Sutcliffe, 1970). This is the skill score (SS) for the model based on using the mean of the validation set as the reference technique. P CE ¼ 1 P
yp ym
2
ym ym
2
3.
Results and discussion
3.1.
Digestate characterisation
Table 2 shows a summary of the results of digestate characteristics during the eight operating conditions. It can be seen that the reactors all had similar performance over the period of the study. The greater range of TS and VS in reactor 2 are attributable to a temporary break down of the stirring mechanism. Table 3 shows the squared correlation coefficients (R2) between each of the digestate characteristics. It can be seen that there is high correlation between TS and VS as would be expected. The next highest correlation is between BA and VFA, with correlation between TS and VS, and BA and VFA very weak. This suggests that any PLS-R model constructed for TS would be similar to a model for VS but that the development of models for BA and VFA would be more distinct.
3.2.
Raw spectra
(5)
CE ranges from -N to 1. If CE >1 the model has skill in predicting values when compared to the use of the mean. If CE ¼ 0 then the model is no better than using the mean. If CE < 1 the model performance is worse than using the mean. This provides a measure of model effectiveness with a bounded upper limit that makes interpretation simpler than that of
A plot of raw spectra for different mixtures of primary sludge and WAS can be seen in Fig. 1. The spectra are dominated by water absorption bands with wavenumbers of around 5000 cm1 and 7000 cm1. It can be seen that absorbance is increased as the proportion of primary sludge in the sample increases. Saeys et al. (2004) suggested that absorbance is inversely related to the amount of dry matter content in pig
2468
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 6 3 e2 4 7 2
Fig. 1 e Raw spectra for various primary:WAS sludge ratios (with TS adjustment) (before pre-processing)(left) and raw spectra for various primary sludge and WAS samples (without TS adjustment and before pre-processing) (right).
manure when spectra are collected in reflectance mode. When there is less dry matter there is less chance that the light is reflected back to the detector, raising the apparent absorption. In Fig. 1 each sample scanned had the same dry matter content (5%). This demonstrates a clear influence on sample absorption due to the sludge mixture and not the dry matter content of the sludge. It is theorised that this is due to the difference in optical properties of primary and WAS which arise as a result of the distribution of the dry matter throughout the sludge. The difference in absorption is also evident in primary and WAS that had not been adjusted to a TS content of 5%. In those cases primary sludges contain more dry matter (mean TS ¼ 7.21%) than WAS (mean TS ¼ 5.56%) but counter to Saeys et al. (2004) findings, higher absorbance is seen for the samples of primary than for the WAS. This is not to say that there is no influence on absorption from dry matter content but that in the case of sewage sludges it is not the most significant factor.
3.3.
PCA
A model consisting of 2 PCs was produced. The first PC accounted for 98.54% of the variance and the second PC accounted for 1.24%. The resulting scores plot can be seen in Fig. 2. From the scores plot, it can be seen that there are four distinct groupings, each corresponding to a different type of sludge. This kind of analysis enables each type of sludge to be differentiated by their spectra. The digested, feed (70:30 primary:WAS ratio), primary and WAS samples can be distinguished by their scores on PC1 with the samples of primary sludge distinguishable from feed samples by their score on
Fig. 2 e Scores plots from PCA modelling.
PC2. The two subgroupings of feed samples can be accounted for by differences in the feedstock samples collected from the WWTP and used to mix the feed. Additionally, it can be seen that when the Primary:WAS mixtures are applied to the PCA model developed using the four types of sludges, that they have scores that are descriptive of their makeup. The Primary:WAS mixtures scores moving from near the centre of the scores plot close to the existing primary samples, through the cluster of feed samples to the top left and the pure WAS samples as the proportion of WAS in the mixture increases. Fig. 3 shows the scores plot for the digested samples from reactor 1. It can be seen that the majority of samples are located near the origin of the plot. This is to be expected as the majority of samples were taken from the digester under similar conditions. As the scores for each sample move away from the centre, the performance of the digester deviated from normal. Deviations from the centre of the plot can be shown to be correlated with disturbances to the digester. The first major deviation from the centre is that associated with operating condition 2. The return of the scores to the centre then coincide with the return to a standard feeding regime. The deviation of the scores from the centre towards the top left of the plot are coincident with operating conditions 6, 7 and 8. In this case a decrease in operating temperature from 25 C to 20 C and a halving of the HRT from 15 days to 7.5 days. This behaviour can be seen in the scores plots for all three reactors and demonstrates that FT-NIR with PCA can be used as a monitoring tool to indicate the stability of a sewage sludge digester.
Fig. 3 e Scores plots for PCA modelling for reactor 1 performance (Operating Condition (OC)).
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 6 3 e2 4 7 2
3.4.
PLS-R model results
The results of the model performance as validated are summarised in Table 4 with scatter plots for BA, VS, VFAs and proportion of WAS shown in Fig. 4. All models produced have CE > 0. This demonstrates that the models have a greater ability for prediction than simply using the mean. Fig. 5 shows time series data for the reactor 1 models. Where data was not collected the endpoints of the two adjacent periods have been merged. The periods of time over which each operating condition was carried out are indicated at the top of each figure. It can be seen (Fig. 5) that the individual models constructed for TS, VS, BA and VFA capture the gross behaviour of the digesters well. The model performance indicators for BA are similar for all 3 reactors with little to choose between them. The values of R2 and CE show that the models constructed have the ability to predict values of BA. The time series for the BA model for reactor 1 (Fig. 5) indicates that although the model is not suitable for accurate quantitative predictions, the model is sufficiently accurate to track the measured values of BA over time and that it would be possible to use such a model to determine gross changes in the operation of the reactor. This is possible because the ratio of the range to the error (range/ RMSEP) is large. The performance indicators of the models developed for TS are greater than those for VS but the two sets are similar as might be expected from the high correlation between the two data sets. This can be seen in the time series plots for TS and VS for Reactor 1 in Fig. 5, the predicted TS and VS values for reactor 1 correlating with R2 ¼ 0.98. This high correlation also exists between TS and VS models for Reactors 2 and 3 suggesting that the models are constructed from the same spectral features but simply scaled to match the measured values. The performance of the modelled results for VFA is similar to that for BA, TS or VS. The performance of the models for the three reactors is similar, though with more variance than that for BA. The predicted outputs of the models for VFA and BA are inversely correlated with R2 ¼ 0.83 for reactor 1. Some correlation is to be expected due to the level of correlation of the measured data sets and is inline with the increase in correlation observed from the measured to predicted TS and VS data sets. As for the BA models, from the time series plots in Fig. 5 the VFA models are suitable for quantitative analysis of VFA content and can track the measured values of VFA over time providing an indication of the stability of the digester. To investigate further the ability to use FT-NIR to characterise digester feedstocks it was decided to use the spectra gained from the primary:WAS mixes to build a PLS-R model. As can be seen from the results in Table 4 and Fig. 4 the model produced has a high level of ability in predicting the proportion of WAS in the feed mixture. These results would need to be confirmed using sludge samples collected at different times from the initial WWTP and also with samples taken from alternative WWTPs, where for example industrial/domestic sources of wastewaters are treated and activated sludge ages are significantly different. In order to produce a robust model it is necessary to be able to take samples from the full range of operating conditions
2469
expected (Saeys et al., 2004). This suggests that it would not be possible to build a highly predictive PLS-R model based on scans from a healthy digester as the range and standard deviation of the operating conditions used to calibrate the model would not include deviations from “normal” operation. For sewage sludge digesters the range and values of the component characteristics that can be expected to be found in a stable digester are limited. Previous studies using FT-NIR and PLS-R to analyse VFAs (Holm-Nielsen et al., 2007a; Lomborg et al., 2009; Jacobi et al., 2009) show better values of RPD for the predictive models produced than found in this study (RPD ¼ 4.11, Jacobi et al., 2009; RPD ¼ 4.3, Holm-Nielsen et al., 2007a; RPD ¼ 3.1, Lomborg et al., 2009). However, the range of VFAs looked at in those studies far exceed those utilised in this study with sewage sludge and this could lead to increased values of RPD due to the larger range of the sample set for a given RMSEP. The same effect can be seen for calculations of R2, where for a given RMSEP a larger sample set range produces a higher value of R2. The RMSEP for the VFA models developed during this study (Table 4) are lower than those found and reported in Jacobi et al. (2009) (RMSEP ¼ 820 mg l1, Jacobi et al., 2009; RMSEP ¼ 200 mg l1, Holm-Nielsen et al., 2007a; RMSEP ¼ 1590 mg l1, Lomborg et al., 2009). Jacobi et al. (2009) also indicated that VFAs in the range <1000 mg l1 behave differently to those samples above that range when using FT-NIR. As previously mentioned Hansson et al. (2002) suggest a limit of detection of 300 mg l1 for propionate. The results from this study show a similar pattern of results, the model seems to behave differently below 300 mg l1. The distribution of the measured values may also play a part as there are few values in this study above 300 mg l1 with which to make a comparison. As a stable sewage sludge digester normally operates at VFA concentrations in the region of 100e200 mg l1 the behaviour of the model below 300 mg l1 is a potential stumbling block for the use of FT-NIR as a process monitoring system for VFAs. However, the time series plot for VFAs shown in Fig. 5 suggest that this behaviour at low concentrations of VFA may not be a significant impediment to the development of a process monitoring system for VFAs using FT-NIR, as the response of the digester to stress is well described by the model. That it is not possible to quantitatively predict the values of VFA concentrations below approximately 300 mg l1 is not a critical problem for a monitoring system that only needs to determine if there is significant move away from a stable operating condition such as those seen during conditions 7 and 8. BA models do not seem to exhibit the different behaviour over different concentration ranges that VFAs do. BA models also do not suffer from limitations due to the limit of detection as might a model based on VFAs. The correlation between measured BA and VFAs also suggests the use of BA as a proxy for VFAs that can be exploited for process monitoring as demonstrated by Guwy et al. (1994). The similarity in the model results presented in this study of BA and VFAs seems to bear that out. In the case where a VFA model is deemed unsuitable for process monitoring, then it is possible that a BA model could be used successfully instead.
2470
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 6 3 e2 4 7 2
Fig. 4 e Scatter plots and regression lines for the PLS-R models (Reactor 1 (R1), Reactor 2 (R2), Reactor 3 (R3)). Previous studies utilising different substrates that have used FT-NIR and PLS-R to model TS and VS (Saeys et al., 2005; Holm-Nielsen et al., 2007a; Lomborg et al., 2009) were able to make use of greater ranges of values in their sample
sets than was possible in this study. However the results gained in this study show that it is possible to produce models able to predict values of TS and VS using FT-NIR and PLS-R with a restricted calibration set range but that care
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 6 3 e2 4 7 2
2471
Fig. 5 e Time series results for Reactor 1 PLS models (The solid line indicates measured values, the dashed line indicates predicted values, operating conditions indicated at top of figure).
must be taken when using those models outside of the calibration range. There are several approaches that could be taken in order to attempt to improve the performance of the models. It is desirable that the characteristics of the samples used to calibrate and validate the models cover the span of the values that can be expected to be found in practise. Ideally there would be an even distribution of samples across the range of interest. This is important for characteristics such as VFAs where there is some question about the behaviour of the technique at low concentrations.
4.
Conclusions
Reflectance NIR spectroscopy can be used as a tool for characterisation of feedstocks and digestates, and performance parameters for sewage sludge digesters. In particular the outcomes of the study were as follows: 1. It is possible to monitor the stability parameters of sewage sludge digestion by using reflectance FT-NIR with PLS-R by building predictive models for BA, TS/VS, and VFA without the need for sample pre-treatment. 2. In the case of BA it has been shown that a PLS-R predictive model could be used in lieu of a VFA model for process monitoring, where both model outputs correlated with R2 ¼ 0.83. 3. The technique is able to effectively distinguish between undigested feed sludge and sludge treated by anaerobic digestion using PCA, as well as between different
proportions of primary and waste activated sludge (WAS) using PLS-R. 4. The study indicates that the apparent absorbance of sewage sludge is strongly affected by the distribution of dry matter throughout the sludge, and that this effect is more significant than that seen due to the variation of dry matter in the sludge.
Acknowledgements The authors would like to acknowledge Welsh water for provision of sewage sludge samples. This work was partly funded by the Welsh Assembly Government and ERDF to the Wales Centre of Excellence for Anaerobic Digestion as well as from an University internal research grant URIS 2009.
references
American Public Health Association, 1998. Standard Methods for Examination of Water and Wastewater, twentieth ed. (Washington DC, USA). Blanco, M., Villarroya, I., 2002. NIR spectroscopy: a rapid-response analytical tool. TrAC Trends in Analytical Chemistry 21 (4), 240e250. Cruwys, J.A., Dinsdale, R.M., Hawkes, F.R., Hawkes, D.L., 2002. Development of a static headspace gas chromatographic procedure for the routine analysis of volatile fatty acids in wastewaters. Journal of Chromatography A 945 (1e2), 195e209.
2472
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 6 3 e2 4 7 2
de Jong, S., 1993. SIMPLS: an alternative approach to partial least squares regression. Chemometrics and Intelligent Laboratory Systems 18 (3), 251e263. Dias, A.M.A., Moita, I., Pa´scoa, M., Alves, M.M., Lopes, J.A., Ferreira, E.C., 2008. Activated sludge process monitoring through in-situ NIR spectral analysis. Water Science and Technology 57 (10), 1643e1650. Dixon, M., Gallop, J.R., Lambert, S.C., Lardon, L., Healy, J.V., Steyer, J.-P., 2007. Data mining to support anaerobic WWTP monitoring. Control Engineering Practice 15 (8), 987e999. Esteves, S.R.R., Wilcox, S.J., O’Neill, C., Hawkes, F.R., Hawkes, D.L., 2000. On-line monitoring of anaerobic-aerobic biotreatment of a simulated textile effluent for selection of control parameters. Environmental Technology 21 (8), 927e936. Ganidi, N., Tyrrel, S., Cartmell, E., 2009. Anaerobic digestion foaming causes e a review. Bioresource Technology 100 (23), 5546e5554. Guwy, A.J., Hawkes, D.L., Hawkes, F.R., Rozzi, A.G., 1994. Characterisation of a prototype industrial on line analyser for bicarbonate/carbonate monitoring. Water Research 44, 1325e1330. Hansson, M., Nordberg, A., Sundh, I., Mathesien, B., 2002. Early Warning of disturbances in a laboratory-scale MSW biogas process. Water Science and Technology 45 (10), 55e60. Hansson, M., Nordberg, A., Mathesien, B., 2003. On-line NIR monitoring during anaerobic treatment of municipal solid waste. Water Science and Technology 48 (4), 9e13. Holm-Nielsen, J.B., Andree, H., Lindorfer, H., Esbensen, K.H., 2007a. Transflexive embedded near infrared monitoring for key process intermediates in anaerobic digestion/biogas production. Journal of Near Infrared Spectroscopy 15, 123e135. Holm-Nielsen, J.B., Lomborg, C.J., Oleskowicz-Popiel, P., Esbensen, K.H., 2007b. On-line near infrared monitoring of glycerol-boosted anaerobic digestion processes: evaluation of process analytical technologies. Biotechnology and Bioengineering 99 (2), 302e313. Jacobi, H.F., Moschner, C.R., Hartung, E., 2009. Use of near infrared spectroscopy in monitoring of volatile fatty acids in anaerobic digestion. Water Science and Technology 60 (2), 339e346. Jenkins, S.R., Morgan, J.M., Sawyer, C.L., 1983. Measuring anaerobic sludge digestion and growth by a simple alkalimetric titration. Journal Water Pollution Control Federation 55 (5), 448e453. Jolliffe, I., 2002. Principle Component Analysis, second ed. Springer, New York. Lardon, L., Punal, A., Steyer, J.P., 2004. On-line diagnosis and uncertainty management using evidence theory e experimental illustration to anaerobic digestion processes. Journal of Process Control 14 (7), 747e763.
Lomborg, C.J., Holm-Nielsen, J.B., Oleskowicz-Popiel, P., Esbensen, K.H., 2009. Near infrared and acoustic chemometrics monitoring of volatile fatty acids and dry matter during co-digestion of manure and maize silage. Bioresource Technology 100, 1711e1719. Massart, N., Bates, R., Corning, B., Neun, G., 2006. Design and operational considerations to avoid excessive anaerobic digester foaming. Proceedings of the Water Environment Federation 31e40, 225e2575. Muller, J., 2000. Pretreatment processes for the recycling and reuse of sewage sludge. Water Science and Technology 42 (9), 167e174. Nash, J.E., Sutcliffe, J.V., 1970. River flow forecasting through conceptual models part I d a discussion of principles. Journal of Hydrology 10 (3), 282e290. Saeys, W., Darius, P., Ramon, H., 2004. Rapid on site analysis of hog manure using a visual and near-infrared diode array reflectance spectrometer. Journal of Near Infrared Spectroscopy 12 (5), 299e310. Saeys, W., Xing, J., De Baerdemaeker, J., Ramon, H., 2005. Comparison of transflectance and reflectance to analyse hog manures. Journal of Near Infrared Spectroscopy 13, 99e107. Smidt, E., Parravicini, V., 2009. Effect of sewage sludge treatment and additional aerobic post-stabilization revealed by infrared spectroscopy and multivariate data analysis. Bioresource Technology 100, 1775e1780. Spanjers, H., van Lier, J.B., 2006. Instrumentation in anaerobic treatment e research and practice. Water Science and Technology 53 (4e5), 63e76. Steyer, J.P., Bouvier, J.C., Conte, T., Gras, P., Sousbie, P., 2002. Evaluation of a four year experience with a fully instrumented anaerobic digestion process. Water Science and Technology 45 (4e5), 495e502. Weglarczyk, S., 1998. The interdependence and applicability of some statistical quality measures for hydrological models. Journal of Hydrology 206 (1e2), 98e103. Williams, P.C., 1987. Variables affecting near-infrared reflectance spectroscopic analysis. In: Williams, P., Norris, K. (Eds.), Near Infrared Technology in the Agriculture and Food Industries, first ed. Am. Cereal Assoc. Cereal Chem., St. Paul, MN, pp. 143e167. Wold, S., Sjo¨stro¨m, M., Eriksson, L., 2001. PLS-regression: a basic tool of chemometrics. Chemometrics and Intelligent Laboratory Systems 58 (2), 109e130. Xie, L., Xingqian, Y., Liu, D., Ying, Y., 2009. Quantification of glucose, fructose and sucrose in bayberry juice by NIR and PLS. Food Chemistry 114, 1135e1140.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 7 3 e2 4 8 4
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Effect of biological and chemical oxidation on the removal of estrogenic compounds (NP and BPA) from wastewater: An integrated assessment procedure Giorgio Bertanza a,*, Roberta Pedrazzani b, Mario Dal Grande c, Matteo Papa a, Valerio Zambarda d, Claudia Montani f, Nathalie Steimberg e, Giovanna Mazzoleni e,1, Diego Di Lorenzo f,1 a
DICATA Department of Civil Engineering, Architecture, Land and Environment, University of Brescia, via Branze 43, I-25123 Brescia, Italy DIMI Department of Mechanical and Industrial Engineering, University of Brescia, via Branze 38, I-25123 Brescia, Italy c Acque Veronesi s.c.a.r.l., Lungadige Galtarossa 8, I-37133 Verona, Italy d TZ Engineering, via Monte Suello 119/A, I-25015 Desenzano del Garda, Italy e General Pathology & Immunology Unit, Department of Biomedical Sciences and Biotechnologies, School of Medicine, University of Brescia, viale Europa 11, I-25123 Brescia, Italy f Laboratory of Biotechnology, Department of Laboratory Medicine, Civic Hospital of Brescia, Piazzale Spedali Civili 1, I-25123 Brescia, Italy b
article info
abstract
Article history:
A major source of the wide presence of EDCs (Endocrine Disrupting Compounds) in water
Received 22 March 2010
bodies is represented by direct/indirect discharge of sewage. Recent scientific literature
Received in revised form
reports data about their trace concentration in water, sediments and aquatic organisms, as
28 January 2011
well as removal efficiencies of different wastewater treatment schemes. Despite the
Accepted 31 January 2011
availability of a huge amount of data, some doubts still persist due to the difficulty in
Available online 19 February 2011
evaluating synergistic effects of trace pollutants in complex matrices. In this paper, an integrated assessment procedure was used, based on chemical and biological analyses, in
Keywords:
order to compare the performance of two full scale biological wastewater treatment plants
EDCs (Endocrine Disrupting
(either equipped with conventional settling tanks or with an ultrafiltration membrane unit)
Compounds)
and tertiary ozonation (pilot scale).
Estrogenic activity
Nonylphenol and bisphenol A were chosen as model EDCs, together with the parent
Mass balance
compounds mono- and di-ethoxylated nonylphenol (quantified by means of GCeMS).
MBR treatment
Water estrogenic activity was evaluated by applying the human breast cancer MCF-7 based
Tertiary ozonation
reporter gene assay. Process parameters (e.g., sludge age, temperature) and conventional pollutants (e.g., COD, suspended solids) were also measured during monitoring campaigns. Conventional activated sludge achieved satisfactory removal of both analytes and estrogenicity. A further reduction of biological activity was exerted by MBR (Membrane Biological Reactor) as well as ozonation; the latter contributed also to decrease EDC concentrations. ª 2011 Elsevier Ltd. All rights reserved.
* Corresponding author. Tel.: þ39 030 3711301; fax: þ39 030 3711312. E-mail address:
[email protected] (G. Bertanza). 1 Shared responsibility for estrogenic measurements. 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.01.026
2474
1.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 7 3 e2 4 8 4
Introduction
In recent decades, concerns regarding the occurrence of Endocrine Disrupting Compounds (EDCs) in the environment have rapidly increased worldwide. Municipal sewage and Waste Water Treatment Plant (WWTP) effluents are considered to be major sources of pollution due to the documented presence of such compounds at relevant concentrations (see, inter alia: Auriol et al., 2006; Ternes and Joss, 2006; Gonza´lez et al., 2007; Stasinakis et al., 2008; Ying et al., 2009; SanchezAvila et al., 2009). EU Directive 2008/105/EC (amending and subsequently repealing Council Directives 82/176/EEC, 83/513/EEC, 84/156/ EEC, 84/4/9/EEC, 86/280/EEC and amending Directive 2000/60/ EC) sets strict quality standards for water bodies (many EDCs being included among priority substances). Therefore, in the future, efforts to adopt feasible and reliable treatment techniques for wastewater cleaning will be made. Accordingly, two important tasks should be pursued: (1) the assessment of the removal capacity of conventional biological processes and, consequently, (2) the evaluation of possible requirements for additional (tertiary) treatment. Even though data is available in the literature on both issues, some lack in knowledge still persists: (1) the removal potential of many EDCs by conventional activated sludge plants is well-established (see for instance Farre´ et al., 2002; Ternes and Joss, 2006; Gonza´lez et al., 2007; Joss et al., 2008; Pothitou and Voutsa, 2008; Press-Kristensen et al., 2008), nevertheless, data are not easily comparable due to different treatment conditions, sampling procedures and analytical methods; (2) tertiary chemical oxidation has been successfully tested (Rosenfeldt and Linden, 2004; Auriol et al., 2006; Zhang et al., 2006; Esplugas et al., 2007; Gultekin and Ince, 2007; Ning et al., 2007; Bolong et al., 2009; Racz and Goel, 2010) but technical-economic feasibility is still to be fully demonstrated (Auriol et al., 2006; Gultekin and Ince, 2007; Koh et al., 2008); (3) chemical analysis alone is not useful to investigate synergistic effects among mixtures of different pollutants and their degradation by-products (a well-known phenomenon in the case of endocrine disruptors: Hjelmborg et al., 2006; Bjorkblom et al., 2008; Mnif et al., 2010). Several authors (Svenson et al., 2003; Hashimoto et al., 2007; Fernandez et al., 2008; Mispagel et al., 2009) have pointed out that water biological activity should also be monitored in order to better evaluate treatment suitability; actually, endocrine activity assays have been proposed in the last few years (Harris et al., 1997; Ce´spedes et al., 2003; Isobe et al., 2003; Korner et al., 2004; Tan et al., 2007; Fernandez et al., 2009; Jugan et al., 2009; Creusot et al., 2010; Sousa et al., 2010). In this work, an integrated assessment procedure, based on both chemical and biological analyses, was adopted to evaluate the performance of biological and chemical oxidation in the removal of target EDCs from municipal wastewater. The following estrogen-like substances were considered: 4-nonylphenol (NP), its parent compounds 4-nonylphenol monoethoxylate (NP1EO) and 4-nonylphenol diethoxylate (NP2EO), and bisphenol A (BPA). These substances were chosen as model EDCs since they are diffusely detected in the aquatic environment (Kolpin et al., 2002; Belmont et al., 2006; Gultekin and Ince,
2007; Loos et al., 2007; Sun et al., 2008; Ying et al., 2009) and are included in the EU priority list (EU Directive 2008/105/EC). Experimental work was conducted at two full scale WWTPs located in Northern Italy equipped with either conventional settling tanks (CAS, Conventional Activated Sludge: Verona municipality) or with an ultrafiltration unit (MBR, Membrane Biological Reactor: Brescia municipality). Tertiary chemical oxidation was tested by means of an ozone pilot plant located at the Verona WWTP. The duration of the analytical campaigns was extended so as to enable the accurate calculation of mass balances of target compounds. Hormonal activity in water samples was measured by means of human breast cancer MCF-7 based reporter gene assay, using 17b-estradiol (E2) as a standard. This cell line was chosen due to its high concentration of estrogenic receptors and sensitivity (Pons et al., 1990; Urban et al., 2001; Soto et al., 2006; Higashi et al., 2007).
2.
Materials and methods
2.1.
Treatment plants
2.1.1.
Verona WWTP
This is a CAS plant (design size 370,000 p.e.) treating mainly domestic wastewater. The process scheme includes primary settling (volume ¼ 10,400 m3, 3 parallel basins); pre-denitrification (volume ¼ 7200 m3, 5 parallel basins); oxidation-nitrification (volume ¼ 16,600 m3, 5 parallel basins); secondary settling (volume ¼ 26,100 m3, 6 parallel basins). The sludge treatment line consists of: dynamic thickening, anaerobic digestion and mechanical dewatering. The following are the main operational data (typical values): influent water flow ¼ 92,000 m3/d (dry weather); dissolved oxygen concentration in aerated tanks ¼ 2.0e2.2 mg/L; total suspended solids concentration in biological reactors ¼ 4.0e4.5 gTSS/L; influent characteristics (after screens and gritoil removal): 450 mgCOD/L, 200 mgBOD5/L, 240 mgTSS/L, 50 mgTKN/L, 5 mgPTOT/L; effluent characteristics: 30 mgCOD/L, 5 mgBOD5/L, 12 mgTSS/L, 6.5 mgTKN/L; 4 mgNHþ 4 -N/L, -N/L, <0.1 mgNO -N/L, 1.3 mgP /L. 4 mgNO TOT 3 2
2.1.2.
Brescia WWTP
This consists of 2 CAS lines and 1 MBR line (design size 380,000 p.e.), treating domestic and industrial wastewater. The process scheme includes equalization/homogenization (volume ¼ 24,000 m3); pre-denitrification (volume ¼ 11,100 m3, 3 parallel basins); oxidation-nitrification (volume ¼ 20,600 m3, 3 parallel basins); secondary settling (for conventional lines, volume ¼ 7800 m3, 2 parallel basins) and ultrafiltration (for MBR line). This configuration enabled the comparison of the CAS process with the MBR technique. The sludge treatment line consists of: dynamic thickening, anaerobic digestion and mechanical dewatering. The following are the main operational data (typical values): influent water flow ¼ 71,500 m3/d (dry weather); dissolved oxygen concentration in aerated tanks ¼ 1 mg/L; total suspended solids concentration in biological reactors ¼ 2.0 and 5.2 gTSS/L in CAS and MBR lines, respectively; influent
2475
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 7 3 e2 4 8 4
VERONA WWTP sedimentation outlet
inlet pretreatments
outlet final sedimentation
denitrification / oxidation nitrification
primary sedimentation
primary sludge
excess sludge (= return sludge)
sludge treatment
dewatered sludge
BRESCIA WWTP
equalization / homogenization
CAS outlet
inlet pretreatments
final sedimentation
denitrification / oxidation nitrification
excess sludge (= return sludge)
ultrafiltration
MBR outlet
denitrification / oxidation nitrification excess sludge (= return sludge)
Fig. 1 e Sampling points for the Verona (top) and Brescia (bottom) WWTPs (bold line [ wastewater; fine line [ sludge; dotted line [ supernatant from sludge treatment; double line [ dewatered sludge).
characteristics (after screens and grit-oil removal): 310 mgCOD/ L, 140 mgBOD5/L, 140 mgTSS/L, 29 mgTKN/L, 5 mgPTOT/L; effluent characteristics: 15 (CAS line) and 8 (MBR line) mgCOD/L, <5 mgBOD5/L, <5 mgTSS/L, 2.1 mgTKN/L, 3.1 (CAS line) and 0.5 (MBR line) mgNHþ 4 -N/L, 3.5 (CAS line) and 5 (MBR line) mgNO 3 -N/L, <0.2 mgNO2 -N/L, 0.6 mgPTOT/L.
2.1.3.
Pilot scale ozonation plant
Supplied by SIAD SpA, Bergamo, Italy, this consists of a stainless steel tubular reactor (volume ¼ 1460 L) and is equipped
NP
NP1EO
NP2EO
with a pure oxygen supply system (capacity ¼ 400 gO3/h). The reactor can be fed with a flow-rate up to 6 m3/h in a continuous mode of operation.
2.2.
Monitoring campaign and treatment tests
2.2.1.
Full scale CAS and MBR WWTPs
The Verona WWTP monitoring campaign was conducted in winter (dry weather) from 5 to 20 February 2008: sampling points were located as shown in Fig. 1 (top). It is important to
BPA
COD
TSS
600
7 6
500
5
400
4
300
3 200
2 100
1 0 4-feb
COD, TSS (mg/L)
Target compounds (µg/L)
8
0
6-feb
8-feb
10-feb
12-feb
14-feb
16-feb
18-feb
20-feb
Time (date) Fig. 2 e Verona WWTP: daily average concentration of pollutants in influent wastewater.
2476
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 7 3 e2 4 8 4
Table 1 e Verona WWTP: average concentrations of target EDCs and percentage attached to 1.6 mm particulate fraction. Influent
Primary settling tank effluent
Final effluent
Influent
Final effluent (CAS)
Final effluent (MBR)
Total (mg/L)
Total Particulate (mg/L) (%)
Total (mg/L)
Total (mg/L)
0.74 0.29 0.64 0.47
0.79 0.30 0.96 0.50
Total Particulate Total Particulate (mg/L) (%) (mg/L) (%) NP NP1EO NP2EO BPA
4.15 3.90 2.18 2.19
47 49 38 41
3.65 3.96 2.15 2.43
41 33 39 30
0.85 0.52 0.70 0.31
note that sewage entering the primary settling tanks includes supernatants from the sludge treatment line. The Brescia WWTP was monitored during a dry weather summer period (23 June e 11 July 2008). Sampling points are shown in Fig. 1 (bottom); unlike the Verona plant, influent samples were not affected by supernatants from the sludge line. For both plants, wastewater was collected daily, over 24 h, by automatic refrigerated auto-samplers equipped with Teflon pipes and dark glass containers (pre-washed with hydrochloric acid and acetone); sludge was sampled instantaneously and submitted immediately to analysis. The following parameters were measured on collected samples: NP (mixture of 4-nonylphenol isomers), NP1EO (mixture of 4-nonylphenol monoethoxylates isomers), NP2EO (mixture of 4-nonylphenol diethoxylates isomers), BPA, COD, total suspended solids (TSS). Estrogenic activity was measured only at the Brescia WWTP, on three 24-h samples collected during the monitoring campaign.
2.2.2.
Ozonation plant
Two series of tests were conducted in order to assess the effect of ozone dosage (12 and 20 mgO3/L) and, for each ozone
NP
Effluent, 20.4%
Table 2 e Brescia WWTP: average concentrations of target EDCs and percentage attached to 1.6 mm particulate fraction.
NP NP1EO NP2EO BPA
4.70 7.89 5.01 1.94
64 51 45 63
concentration, three runs were performed at increasing contact times (15, 22 and 30 min, respectively). During each test, at 1, 2 and 3 HRT (Hydraulic Retention Time) time intervals, grab samples of influent and effluent wastewater were taken and immediately submitted to chemical (NP, NP1EO, NP2EO, BPA), microbiological (total coliforms and Escherichia coli) and biological (estrogenic activity) analyses. Based on instrumentally detected data (ozone production and residue in offgas), the actual ozone dissolution percentage was calculated.
2.3.
Chemical analyses
The method of Gatidou et al. (2007) was successfully adopted for the extraction of analytes from liquid phase. The following chemicals were purchased from Sigma Aldrich (Taufkirchen, Germany): (a) standard reagents: bisphenol A, NP1EO, NP2EO, 4-NP technical mixture of isomers, as proposed by ISO 18857-1 (2005); (b) derivatization reagents: MSTFA and pyridine; (c) internal standard: bisphenol A-d16.
Primary sludge, 5.5%
Effluent, 13.2%
Biological degradation, 73.6%
Biological degradation, 84.1%
Primary sludge, 5.4%
Effluent, 13.9%
Primary sludge, 6.4%
BPA
Excess sludge, 0.4%
Effluent, 32.1%
Excess sludge, 0.5%
NP1EO
Excess sludge, 0.3%
Excess sludge, 0.5%
NP2EO
Primary sludge, 2.4%
Biological degradation, 62.0%
Biological degradation, 79.3%
Fig. 3 e Verona WWTP: mass balance of trace pollutants. “Degraded” mass obtained by subtracting the sum of each effluent mass flow (final effluent, primary and excess sludge) from influent load.
2477
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 7 3 e2 4 8 4
NP
Excess sludge CAS, 1.8%
Effluent CAS, 15.7%
Effluent CAS, 3.7%
NP1EO
Excess sludge CAS, 1.6%
Biological degradation CAS, 94.5%
Biological degradation CAS, 82.7%
NP2EO
Effluent CAS, 12.3%
BPA
Effluent CAS, 24.4%
Excess sludge CAS, 1.5%
Excess sludge CAS, 5.3%
Biological degradation CAS, 86.2%
NP
Biological degradation CAS, 70.3%
Effluent MBR, 16.8%
Effluent MBR, 3.8%
Excess sludge MBR, 1.1%
NP1EO
Excess sludge MBR, 1.8%
Biological degradation MBR, 94.4%
Biological degradation MBR, 82.1%
NP2EO Effluent MBR,
Effluent MBR, 25.7%
19.4%
BPA
Excess sludge MBR, 1.6%
Excess sludge MBR, 3.9% Biological degradation MBR, 79.0%
Biological degradation MBR, 70.5%
Fig. 4 e Brescia WWTP: mass balance of trace pollutants: CAS (top) and MBR (bottom) line, respectively. “Degraded” mass obtained by subtracting the sum of each effluent mass flow (final effluent and excess sludge) from influent load.
Influent samples were filtered on glass fiber filters (Whatman GF/A, 4 ¼ 1.6 mm particle retention) in order to separate particulate matter from the liquid phase. Liquid samples were submitted to enrichment on SPE C18 (Supelco, Bellefonte, USA) and consequent elution. Filters were weighed prior to filtration; solids retained by the filter were weighed by using a thermobalance set at 60 C. Afterward, filters were placed into 50 mL vials, and 9 mL dichloromethane-hexane 4:1, 1 mL BPA-d16
(500 ppb) and 100 mL HCl 6 N were added. Vials were submitted to sonication for 30 min at 50 C. Derivatization was performed with 900 mL MSTFA (5% in isooctane) and pyridine (100 mL). Instrumental analysis was conducted using a gas-chromatograph 5975B inert XL EI/CI MSD equipped with a split/ splitless injector and autosampler (Agilent Technologies, Palo Alto, USA).
2478
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 7 3 e2 4 8 4
Fig. 5 e Biological assay: calibration curve with the reference estrogen E2.
The RDS% (Recovery Determination Standard) varied from 7.3 to 13.7, depending on the target molecule; mean recovery percentage referring to internal standard of BPA-d16 was more than 80% (water samples) and about 60% (sludge samples); the lowest concentration of the calibration curve was equal to 100 ppb for each pollutant (for further details about the analytical procedure, see Pedrazzani et al., in preparation). COD and TSS were measured as prescribed by the Italian Standard Methods (APAT IRSA CNR No. 5130 and 2090, 2003), the former after K2Cr2O7 oxidation, and the latter after 0.45 mm filtration and 105 C drying process, respectively.
2.4.
Microbiological analyses
Raw samples of ozonation plant influent and effluent wastewater were diluted in sterile NaCl 0.1% and submitted to total coliforms and E. coli determination, accordingly with the MPN (Most Probable Number) technique (Italian Standard Methods: APAT IRSA CNR No. 7010B and 7030B, 2003). DST Colilert (IDEXX Laboratories, Westbrook, USA) was employed, based on specific enzymatic reactions with ONPG (o-nitrophenyl b-D-galactopyranoside) and MUG (4-methyl-umbelliferyl b-D-glucuronide). Multiplates trays were placed in an incubator at 36 1 C for 24 h and positive results were read and interpreted as prescribed.
2.5.
Biological analyses
The pollutant extraction and clean-up procedure was the same as reported for the chemical analyses; extracts were resuspended in 1 mL DMSO (dimethyl-sulfoxide). Human breast cancer cell line MCF-7 stably transfected with the EREtK-LUC construct was maintained in DMEM (Modified Dulbecco’s Medium, Euroclone, Milan), supplemented with 5% calf serum, at 37 C and 5% CO2. 24 h before treatment with pollutants, cells were plated at a density of 6.0$105 cells/well in six-well plates containing phenol red free DMEM and 5% charcoal-stripped fetal calf serum.
Cells were treated with either reference estrogen (E2) or pollutants culture medium solutions; dishes were kept at 37 C for 24 h (Chau et al., 1998; Spink et al., 2003). Cells were then harvested in TEN buffer (10 mM Tris, 10 mM EDTA, 150 mM NaCl, pH 8.0) and pellets were lysed in luciferase assay buffer (25 mM Tris, 150 mM NaCl, 10 mM EDTA, 1 mM dithiothreitol, 5% glycerol, 0.5% Triton X-100, pH 8.0). Lysate was spun for 20 s at 13,000 g and supernatant submitted to luciferase activity quantification, which was performed in triplicate by means of a luminometer (Centro 960, Berthold Tech., Germany) over 10 s (De Wet et al., 1987), expressed as RLU (Relative Light Units) and normalized toward protein concentration. Reference estrogen E2 (dissolved in absolute ethanol) was employed for calibration curve definition, at concentrations corresponding to physiological/sub-physiological doses, i.e., from 1013 to 107 M (the lower approaching LOD e Limit of Detection).
3.
Results and discussion
3.1.
Verona WWTP: CAS process
The mass balance of target compounds was calculated based on measured concentrations and recorded flow-rates of different streams (wastewater and sludge). It should be highlighted that the daily flow-rate was quite stable during the entire period (average value: 82.500 m3/d 5%), thus yielding reliable calculations, despite an expected slight variability of influent concentrations (similar patterns were observed for EDCs and conventional pollutants COD and TSS: Fig. 2). Average weighted concentrations of pollutants in different plant sections as well as solid-liquid phase partition percentages are detailed in Table 1; the complete mass balance is shown in Fig. 3. As far as influent wastewater is concerned, the results confirm the data from the literature, even though NP1EO and NP2EO concentrations are close to the lowest values found by several authors (Di Corcia et al., 1994; Sole´ et al., 2000; Ko¨rber et al., 2000; Fuerhacker et al., 2001; Farre´ et al., 2002; Planas et al., 2002; Fauser et al., 2003; Lagana` et al., 2004; Vethaak et al., 2005; Mart’ianov et al., 2005; Fountoulakis et al., 2005; Lee et al., 2005; Jiang et al., 2005; Clara et al., 2005a, 2007; Shen et al., 2005; Komori et al., 2006; Cantero et al., 2006; Vogelsang et al., 2006; Nakada et al., 2006; Belmont et al., 2006; Levine et al., 2006; Gonza´lez et al., 2007; Loyo-Rosales et al., 2007; Stasinakis et al., 2008). Average concentrations (Table 1) indicate that primary sedimentation exerted negligible removal of trace pollutants, notwithstanding an appreciable abatement of TSS (50%: data not shown) and the relevant percentage of pollutants associated with particulate matter. As a confirmation, mass balance revealed that only 5e6% (Fig. 3) of the influent amount of these contaminants was in primary sludge, detected concentrations being in the range 3e7 mg/kgTSS. This is in agreement with published data (Gonza´lez et al., 2004; Levine et al., 2006), even though removal percentages up to 20e30% are reported as well (in particular for NPnEO, Ahel et al., 1994). However, an exhaustive comparison with the literature is not possible because primary settling performance is likely to be
2479
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 7 3 e2 4 8 4
experiment #2
7
7
6
6 RLU/mg protein (107), NP + BPA [µg/L]
W1
RLU/mg protein (107), NP + BPA [µg/L]
experiment #1
5 4 3 2 1
3 2
0 Inlet
CAS outlet
MBR outlet
Inlet
CAS outlet
MBR outlet
Inlet
CAS outlet
MBR outlet
Inlet
CAS outlet
MBR outlet
9
7
8 RLU/mg protein (107), NP + BPA [µg/L]
6 RLU/mg protein (107), NP + BPA [µg/L]
4
1
0
W2
5
5 4 3 2
7 6 5 4 3 2
1
1
0
0 CAS outlet
MBR outlet
10
10
9
9
8
8
RLU/mg protein (107), NP + BPA [µg/L]
W3
RLU/mg protein (107), NP + BPA [µg/L]
Inlet
7 6 5 4 3
7 6 5 4 3
2
2
1
1
0
0 Inlet
CAS outlet
MBR outlet
estrogenic activity
EDCs concentration
Fig. 6 e Brescia WWTP: comparison between estrogenic activity (measured in two different experiments and in three different days W1, W2, W3) and EDC concentration (NP D BPA). Error bars represent maximum and minimum values measured in 3 replicates in the case of biological data, while they show variation percentage in the case of chemical analyses.
influenced by hydraulic retention time and sewage temperature, and these data are often missing. Taking into account final effluent, it can be observed that biological process was able to reduce the concentrations of target organics to a significant extent. These results are in accordance with the data from the literature (Koh et al., 2005; Auriol et al., 2006; Huntsman et al., 2006; Levine et al., 2006; Nakada et al., 2006; Vogelsang et al., 2006; Clara et al., 2007; Loos et al., 2007; Loyo-Rosales et al., 2007; Stasinakis et al., 2008). It must be noted that the residual amount of NP, NP1EO and NP2EO in the effluent is the result of both removal (by means
of biodegradation/sorption) and generation (as metabolites of parent compounds) processes. Therefore, while in the case of BPA we focus on primary degradation, for NP, NP1EO and NP2EO we refer to an apparent degradation. Trace pollutants were also detected in excess sludge at concentrations ranging from 0.26 mg/kgTSS (BPA) to 4.08 mg/kgTSS (NP1EO); however, mass balance showed that the amount found in excess sludge accounted for less than 0.5% of the mass entering the biological system. As already noted for primary sludge, these pollutants were not removed with solid phase (sludge). Based on the comparison between TSS (data not shown) and trace pollutant
2480
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 7 3 e2 4 8 4
concentrations in final effluent, a clear correlation could not be evidenced, as already stated in the literature (see, among others, Jiang et al., 2005), who observed that tertiary filtration does not improve the removal of EDCs.
3.2.
Brescia WWTP: CAS and MBR processes
3.2.1.
Chemical analyses
3.3. Overall comparison between CAS and MBR processes performance
Weighted mean concentrations of trace pollutants are reported in Table 2 while mass balance is shown in Fig. 4 for CAS and MBR processes, respectively. Also in this case, no appreciable scattering (5%) with respect to average value was evidenced for sewage flow-rate during the monitoring campaign. Considering influent wastewater characteristics, while NP and BPA were detected in similar concentrations as in the Verona WWTP, NP1EO and NP2EO values were higher. This may be due to several factors: the origin of influent wastewater (Brescia is located in a heavily industrialized area); influent wastewater temperature (higher during the Brescia monitoring campaign), which influences NPnEO degradation pathways, hence metabolite generation by biodegradation processes; sewer pipeline features (length, hydraulic retention time, etc.). Both CAS and MBR lines yielded a noticeable reduction of trace pollutants and, like in Verona the plant, amounts detected in excess sludge were very low: from 1.1% to 5.3% of total influent mass (concentrations ranging from 0.38 mg/ kgTSS for NP to 1.51 mg/kgTSS for NP1EO).
3.2.2.
(NP þ BPA) concentrations were similar in outlet samples taken from both lines (Fig. 6), estrogenic activity exerted by CAS effluent was almost always higher.
Biological analyses
Water samples (influent and both CAS and MBR effluents), taken on three different days of consecutive weeks (W1, W2 and W3) during the monitoring period, were submitted to biological assays, which were repeated twice (experiment #1 and #2). Prior to each experiment, cell responsivity to E2 was checked and a calibration curve was plotted (an example is presented in Fig. 5). Fig. 6 shows the results of the biological analyses. It is clear that estrogenic activity was significantly reduced by both treatments, and, in five of six cases, with greater efficiency by the MBR system. This is a relevant outcome which emphasizes the importance of biological analyses: actually, while EDC
Removal efficiency and residual effluent concentration of target compounds for all studied plants and processes are compared in Fig. 7. The experimental results show that, while the Brescia CAS and MBR lines, where different sludge ages were kept (9 d for CAS and 15 d for MBR, respectively), yielded similar performances, the Verona CAS plant, having the same sludge age as the Brescia MBR line, yielded on the contrary to slightly lower removal efficiencies (apart from BPA). This phenomenon might be due to different sewage temperature (16 C and 23 C for the Verona and Brescia WWTPs, respectively). Actually, it is well known, that sludge age and temperature are crucial parameters: Clara et al. (2005b) argue that the minimum required sludge age is 10 d at 10 C, and further increases do not lead to noticeable improvements. Moreover, several authors (e.g. Auriol et al., 2006; Koh et al., 2008, 2009) conclude that EDC removal occurs only in plants equipped with nitrification stages (as in the Brescia and Verona WWTPs). In addition, Clara et al. (2004) report that possible MBR efficiency improvements might be ascribed to an increase in sludge age, rather than to filtration. Nevertheless, biological measurements carried out in this work showed that estrogenic activity was reduced to a greater extent by a MBR process with respect to CAS treatment, even if analytes were removed at a comparable level. While the reason is still under investigation; it might be attributed to metabolic pathways exhibited by different microbial consortia growing in MBR plants (Cicek et al., 1999; Clouzot et al., 2010).
3.4.
Tertiary ozonation
3.4.1.
Chemical and microbiological analyses
Actual ozone dosages (calculated based on dissolution efficiency) were 8 and 11 mg/L, respectively, during the two series of tests. Disinfection performance was very high: total coliforms and E. coli were abated from 3.2 log efficiency (8 mg/L actual ozone dosage, 15 min contact time) up to 4.2 (11 mg/L actual
CAS (Verona plant)
A
1
CAS (Brescia plant)
90%
B
MBR (Brescia plant)
80%
Concentration (µg/L)
Overall removal efficiency (%)
100%
70% 60% 50% 40% 30% 20%
0.8
0.6
0.4
0.2
10%
0%
0 NP
NP1EO
Analyte
NP2EO
BPA
NP
NP1EO
NP2EO
BPA
Analyte
Fig. 7 e Comparison among studied processes: treatment efficiency (A) and effluent residual concentrations (B).
2481
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 7 3 e2 4 8 4
1
B
1
0,9
0,9
0,8
0,8
Normalized concentration C(t)/C0
Normalized concentration C(t)/C0
A
0,7 0,6 0,5 0,4 0,3
0,7 0,6 0,5 0,4 0,3
0,2
0,2
0,1
0,1
0
0 0
10
20
30
40
0
10
20
30
40
Time (min)
Time (min)
Fig. 8 e Ozonation: target pollutants normalized concentration vs. reaction time. White marks and fine line: BPA; black marks and bold line: NP. Actual ozone dosages: (A) [ 8 mg/L; (B) [ 11 mg/L. C0 [ influent concentration.
ozone dosage, 30 min contact time), notwithstanding the initial concentration (2.4$105e1.0$106 MPN/100 mL total coliforms, 4.3$104e1.9$105 MPN/100 mL E. coli). Influent trace pollutants concentrations were in the range 0.14e0.30 mg/L and 0.20e0.43 mg/L for NP and BPA, respectively, while both NP1EO and NP2EO were below 0.20 mg/L. Time profiles of NP and BPA normalized concentration are shown in Fig. 8; NP1EO and NP2EO are omitted since they were below detection limits. Assuming first order kinetics (and under the hypothesis of plug-flow reactor), it was possible to estimate reaction rate constants, which resulted, for both pollutants, in the range 0.028e0.093 min1 depending on ozone dosage.
3.4.2.
Biological analyses
The influence of O3 dosage on estrogenic activity abatement is shown in Fig. 9 (average values). Error bars indicate results obtained during different experiments (i.e., reaction time conditions). Chemical oxidation was able to reduce estrogenicity of wastewater remarkably. Nevertheless, while a higher O3 dosage led to an appreciable improvement of EDC (NP þ BPA) removal, 8
estrogenic activity
RLU/mg protein (107), NP + BPA [µg/L]
7
EDCs concentration
6 5 4 3 2 1 0
IN
OUT 8mg(ozone)/L
OUT 11mg(ozone)/L
Fig. 9 e Ozonation: comparison between the estrogenic activity and EDC concentration (NP D BPA) as a function of O3 dosage. Error bars indicate results obtained during different experiments (i.e., reaction time conditions).
only a slight additional reduction of hormonal activity was achieved. This may be due to the persistence of endocrine disruptors (e.g., including natural hormones) or the formation of active by-products, as recently found by other authors (Huber et al., 2004; Bila et al., 2007).
4.
Conclusions
In this work, the fate of selected trace pollutants (NP, NP1EO, NP2EO and BPA) in two full scale WWTPs was investigated. Monitoring campaigns showed that the contribution of primary settling in the removal of studied pollutants was negligible, their content in primary sludge being quite low (<10 mg/kgTSS). Biodegraded fractions ranged from 62.0% (NP2EO, Verona plant) to 94.5% (NP1EO, Brescia plant); final effluent concentrations were always <1 mg/L and excess sludge concentrations 5 mg/kgTSS for all analytes. Although the WWTPs considered have different process schemes (CAS and MBR, respectively) similar performances were observed. In fact this finding was expected based on the literature, since the most influential process parameters (sludge age and temperature) were always within the optimal range for EDC biodegradation. On the contrary, biological assays showed that MBR was more efficient in estrogenicity reduction: this is a very important finding of this research, which would not have been highlighted if only chemical analysis had been performed. As far as tertiary ozonation is concerned, chemical oxidation of trace pollutants was described by first order kinetics, rate constants being dependent on reagent dosage: for instance, a 90% removal of BPA and NP could be achieved either after 80 min at 8 mgO3/L, or 27 min at 11 mgO3/L. Biological analyses confirmed the beneficial effect of ozonation on the reduction of estrogenicity of CAS effluent. However, unlike analytes, estrogenic activity abatement was not significantly affected by ozone dosage. In summary, CAS treatment enabled a satisfactory reduction of EDCs and estrogenicity, thanks to adequate process
2482
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 7 3 e2 4 8 4
conditions; a further decrease of biological activity was achieved by means of MBR and ozonation, but the latter, at the same time, yielded an additional reduction in pollutants. Finally, the efficacy of an integrated (chemical þ biological) approach in evaluating performances of wastewater treatment processes was demonstrated: bioassays account for synergistic effects of dozens of pollutants, the simultaneous determination of which might be actually unfeasible.
Acknowledgments This work was partly conducted within the Vigoni Project (2006/2007) funded by the Italian and German Governments and by the EU project EXERA, LSHB-CT-2006-037168. The authors thank Acque Veronesi s.c.a.r.l. for supporting the experimental activities and SIAD S.p.A. for supplying the chemical oxidation pilot plant. The authors are grateful to Silvia Avesani, Davide Pensieri and Valentina Salogni for their fundamental support during experimental activities conducted within their degree theses.
references
Ahel, M., Giger, W., Koch, M., 1994. Behaviour of alkylphenol polyethoxylate surfactants in the aquatic environment e I. Occurrence and transformation in sewage treatment. Water Research 28 (5), 1131e1142. Apat Irsa Cnr, 2003. Standard Methods for the Examination of Water, Wastewater and Sludge Italy. Auriol, M., Filali-Meknassi, Y., Tyagi, R.D., Adams, C.D., Surampalli, R.Y., 2006. Endocrine disrupting compounds removal from wastewater, a new challenge. Process Biochemistry 41 (3), 525e539. Belmont, M.A., Ikonomou, M., Metcalfe, C.D., 2006. Presence of nonylphenol ethoxylate surfactants in a watershed in Central Mexico and removal from domestic sewage in a treatment wetland. Environmental Toxicology and Chemistry 25 (1), 29e35. Bila, D., Montalva˜o, A.F., Azevedo, D., Dezotti, M., 2007. Estrogenic activity removal of 17b-estradiol by ozonation and identification of by-products. Chemosphere 69 (5), 736e746. Bjorkblom, C., Salste, L., Katsiadaki, I., Wiklund, T., Kronberg, L., 2008. Detection of estrogenic activity in municipal wastewater effluent using primary cell cultures from three-spined stickleback and chemical analysis. Chemosphere 73 (7), 1064e1070. Bolong, N., Ismail, A.F., Salim, M.R., Matsuura, T., 2009. A review of the effects of emerging contaminants in wastewater and options for their removal. Desalination 239 (1e3), 229e246. Cantero, M., Rubio, S., Pe´rez-Bendito, D., 2006. Determination of alkylphenols and alkylphenol carboxylates in wastewater and river samples by hemimicelle-based extraction and liquid chromatographyeion trap mass spectrometry. Journal of Chromatography A 1120 (1e2), 260e267. Ce´spedes, R., Petrovic, M., Raldu´a, D., Saura, U., Pin˜a, B., Lacorte, S., Viana, P., Barcelo´, D., 2003. Integrated procedure for determination of endocrine-disrupting activity in surface waters and sediments by use of the biological technique recombinant yeast assay and chemical analysis by LCeESIMS. Analytical and Bioanalytical Chemistry 378 (3), 697e708.
Chau, K.Y., Peter Lam, H.Y., Daniel Lee, K.L., 1998. Estrogen treatment induces elevated expression of HMG1 in MCF-7 cells. Experimental Cell Research 241 (1), 269e272. Cicek, N., Franco, J.P., Suidan, M.T., Urbain, V., Manem, J., 1999. Characterisation and comparison of a membrane bioreactor and a conventional activated-sludge system in the treatment of wastewater containing high-molecular-weight compounds. Water Environment Research 71 (1), 64e70. Clara, M., Strenn, B., Ausserleitner, M., Kreuzinger, N., 2004. Comparison of the behaviour of selected micropollutants in a membrane bioreactor and a conventional wastewater treatment plant. Water Science and Technology 50 (5), 29e36. Clara, M., Strenn, B., Gans, O., Martinez, E., Kreuzinger, N., Kroiss, H., 2005a. Removal of selected pharmaceuticals, fragrances and endocrine disrupting compounds in a membrane bioreactor and conventional wastewater treatment plants. Water Research 39 (19), 4797e4807. Clara, M., Kreuzinger, N., Strenn, B., Gans, O., Kroiss, H., 2005b. The solids retention time e a suitable design parameter to evaluate the capacity of wastewater treatment plants to remove micropollutants. Water Research 39 (1), 97e106. Clara, M., Scharf, S., Scheffknecht, C., Gans, O., 2007. Occurrence of selected surfactants in untreated and treated sewage. Water Research 41 (19), 4339e4348. Clouzot, L., Doumenq, P., Roche, N., Marrot, B., 2010. Kinetic parameters for 17 alpha-ethinylestradiol removal by nitrifying activated sludge developed in a membrane bioreactor. Bioresource Technology 101 (16), 6425e6431. Creusot, N., Kinani, S., Balaguer, P., Tapie, N., LeMenach, K., Maillot-Marechal, E., Porcher, J.M., Budzinski, H., Ait-Aissa, S., 2010. Evaluation of an hPXR reporter gene assay for the detection of aquatic emerging pollutants: screening of chemicals and application to water samples. Analytical and Bioanalytical Chemistry 396 (2), 569e583. De Wet, J.R., Wood, K.V., De Luca, M., Helinski, D.R., Subramani, S., 1987. Firefly luciferase gene: structure and expression in mammalian cells. Molecular and Cellular Biology 7 (2), 725e737. Di Corcia, A., Samperi, R., Marcomini, A., 1994. Monitoring aromatic surfactants and their biodegradation intermediates in raw and treated sewages by solid-phase extraction and liquid chromatography. Environmental Science & Technology 28 (5), 850e858. Esplugas, S., Bila, D.M., Krause, L.G.T., Dezotti, M., 2007. Ozonation and advanced oxidation technologies to remove endocrine disrupting chemicals (EDCs) and pharmaceuticals and personal care products (PPCPs) in water effluents. Journal of Hazardous Materials 149 (3), 631e642. Farre´, M., Klo¨ter, G., Petrovic, M., Alonso, M.C., de Alda, M.J.L., Barcelo´, D., 2002. Identification of toxic compounds in wastewater treatment plants during a field experiment. Analytica Chimica Acta 456 (1), 19e30. Fauser, P., Vikelsøe, J., Sørensen, P.B., Carlsen, L., 2003. Phthalates, nonylphenols and LAS in an alternately operated wastewater treatment plant e fate modelling based on measured concentrations in wastewater and sludge. Water Research 37 (6), 1288e1295. Fernandez, M.P., Buchanan, I.D., Ikonomou, M.G., 2008. Seasonal variability of the reduction in estrogenic activity at a municipal WWTP. Water Research 42 (12), 3075e3081. Fernandez, M.P., Noguerol, T.N., Lacorte, S., Buchanan, I., Pina, B., 2009. Toxicity identification fractionation of environmental estrogens in waste water and sludge using gas and liquid chromatography coupled to mass spectrometry and recombinant yeast assay. Analytical and Bioanalytical Chemistry 393 (3), 957e968. Fountoulakis, M., Drillia, P., Pakou, C., Kampioti, A., Stamatelatou, K., Lyberatos, G., 2005. Analysis of nonylphenol and nonylphenol ethoxylates in sewage sludge by high
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 7 3 e2 4 8 4
performance liquid chromatography following microwaveassisted extraction. Journal of Chromatography A 1089 (1e2), 45e51. Fuerhacker, M., Scharf, S., Pichler, W., Ertl, T., Haberl, R., 2001. Sources and behaviour of bismuth active substances (BiAS) in a municipal sewage treatment plant. The Science of the Total Environment 277 (1e3), 95e100. Gatidou, G., Thomaidis, N.S., Stasinakis, A.S., Lekkas, T.D., 2007. Simultaneous determination of the endocrine disrupting compounds nonylphenol, nonylphenol ethoxylates, triclosan and bisphenol A in wastewater and sewage sludge by gas chromatography e mass spectrometry. Journal of Chromatography A 1138 (1e2), 32e41. Gonza´lez, S., Petrovic, M., Barcelo´, D., 2004. Simultaneous extraction and fate of linear alkylbenzene sulfonates, coconut diethanol amides, nonylphenol ethoxylates and their degradation products in wastewater treatment plants, receiving coastal waters and sediments in the Catalonian area (NE Spain). Journal of Chromatography A 1052 (1e2), 111e120. Gonza´lez, S., Petrovic, M., Barcelo´, D., 2007. Removal of a broad range of surfactants from municipal wastewater e comparison between membrane bioreactor and conventional activated sludge treatment. Chemosphere 67 (2), 335e343. Gultekin, I., Ince, N.H., 2007. Synthetic endocrine disruptors in the environment and water remediation by advanced oxidation processes. Journal of Environmental Management 85 (4), 816e832. Harris, C.A., Henttu, P., Parker, M.G., Sumpter, J.P., 1997. The estrogenic activity of phthalate esters in vitro. Environmental Health Perspectives 105 (8), 802e811. Hashimoto, T., Onda, K., Nakamura, Y., Tada, K., Miya, A., Murakami, T., 2007. Comparison of natural estrogen removal efficiency in the conventional activated sludge process and the oxidation ditch process. Water Research 41 (10), 2117e2126. Higashi, E., Fukami, T., Itoh, M., Kyo, S., Inoue, M., Yokoi, T., Nakajima, M., 2007. Human CYP2A6 is induced by estrogen via estrogen receptor. Drug Metabolism and Disposition 35 (10), 1935e1941. Hjelmborg, P.S., Ghisari, M., Bonefeld-Jorgensen, E.C., 2006. SPEHPLC purification of endocrine-disrupting compounds from human serum for assessment of xenoestrogenic activity. Analytical and Bioanalytical Chemistry 385 (5), 875e887. Huber, M.M., Ternes, T., von Gunten, U., 2004. Removal of estrogenic activity and formation of oxidation products during ozonation of 17a-ethinylestradiol. Environmental Science & Technology 38 (19), 5177e5186. Huntsman, B.E., Staples, C.A., Naylor, C.G., Williams, J.B., 2006. Treatability of nonylphenol ethoxylate surfactants in on-site wastewater disposal systems. Water Environment Research 78 (12), 2397e2404. ISO 18857-1, 2005. Water Quality e Determination of Selected Alkylphenols e Part 1: Method for Non-filtered Samples Using LiquideLiquid Extraction and Gas Chromatography with Mass Selective Detection Geneva, CH. Isobe, T., Shiraishi, H., Yasuda, M., Shinoda, A., Suzuki, H., Morita, M., 2003. Determination of estrogens and their conjugates in water using solid-phase extraction followed by liquid chromatographyetandem mass spectrometry. Journal of Chromatography A 984 (2), 195e202. Jiang, J.Q., Yin, Q., Zhou, J.L., Pearce, P., 2005. Occurrence and treatment trials of endocrine disrupting chemicals (EDCs) in wastewaters. Chemosphere 61 (4), 544e550. Joss, A., Siegrist, H., Ternes, T.A., 2008. Are we about to upgrade wastewater treatment for removing organic micropollutants? Water Science and Technology 57 (2), 251e255. Jugan, M.L., Oziol, L., Bimbot, M., Huteau, V., Tamisier-Karolak, S., Blondeau, J.P., Levi, Y., 2009. In vitro assessment of thyroid
2483
and estrogenic endocrine disruptors in wastewater treatment plants, rivers and drinking water supplies in the greater Paris area (France). Science of the Total Environment 407 (11), 3579e3587. 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. Koh, Y.K.K., Chiu, T.Y., Boobis, A., Cartmell, E., Scrimshaw, M.D., Lester, J.N., 2008. Treatment and removal strategies for estrogens from wastewater. Environmental Technology 29 (3), 245e267. Koh, Y.K.K., Chiu, T.Y., Boobis, A., Scrimshaw, M.D., Bagnall, J.P., Soares, A., Pollard, S., Cartmell, E., Lester, J.N., 2009. Influence of operating parameters on the biodegradation of steroid estrogens and nonylphenolic compounds during biological wastewater treatment processes. Environmental Science & Technology 43 (17), 6646e6654. Kolpin, D.W., Furlong, E.T., Meyer, M.T., Thurman, E.M., Zaugg, S.D., Barber, L.B., Buxton, H.T., 2002. Pharmaceuticals, hormones, and other organic wastewater contaminants in U.S. streams, 1999e2000: a national reconnaissance. Environmental Science & Technology 36 (6), 1202e1211. Komori, K., Okayasu, Y., Yasojima, M., Suzuki, Y., Tanaka, H., 2006. Occurrence of nonylphenol, nonylphenol ethoxylate surfactants and nonylphenol carboxylic acids in wastewater in Japan. Water Science and Technology 53 (11), 27e33. Ko¨rber, W., Bolz, U., Su¨ssmuth, W., Hiller, G., Schuller, W., Hanf, V., Hagenmaier, H., 2000. Input/output of estrogenic active compounds in a major municipal plant in Germany. Chemosphere 40 (9e11), 1131e1142. Korner, W., Vinggaard, A.M., Te´rouanne, B., Ma, R., Wieloch, C., Schlumpf, M., Sultan, C., Soto, A.M., 2004. Interlaboratory comparison of four in vitro assays for assessing androgenic and antiandrogenic activity of environmental chemicals. Environmental Health Perspectives 112 (6), 695e702. Lagana`, A., Bacaloni, A., De Leva, I., Faberi, A., Fago, G., Marino, A., 2004. Analytical methodologies for determining the occurrence of endocrine disrupting chemicals in sewage treatment plants and natural waters. Analytica Chimica Acta 501 (1), 79e88. Lee, H.B., Peart, T.E., Svoboda, M.L., 2005. Determination of endocrine-disrupting phenols, acidic pharmaceuticals, and personal-care products in sewage by solid-phase extraction and gas chromatographyemass spectrometry. Journal of Chromatography A 1094 (1e2), 122e129. Levine, A.D., Meyer, M.T., Kish, G., 2006. Evaluation of the persistence of micropollutants through pure-oxygen activated sludge nitrification and denitrification. Water Environment Research 78 (11), 2276e2285. Loos, R., Hanke, G., Umlauf, G., Eisenreich, S.J., 2007. LC-MS-MS analysis and occurrence of octyl- and nonylphenol, their ethoxylates and their carboxylates in Belgian and Italian textile industry, waste water treatment plant effluents and surface waters. Chemosphere 66 (4), 690e699. Loyo-Rosales, J.E., Rice, C.P., Torrents, A., 2007. Fate of octyl- and nonylphenol ethoxylates and some carboxylated derivatives in three American wastewater treatment plants. Environmental Science & Technology 41 (19), 6815e6821. Mart’ianov, A.A., Dzantiev, B.B., Zherdev, A.V., Eremin, S.A., Cespedes, R., Petrovic, M., Barcelo, D., 2005. Immunoenzyme assay of nonylphenol: study of selectivity and detection of alkylphenolic non-ionic surfactants in water sample. Talanta 65 (2), 367e374. Mnif, W., Dagnino, S., Escande, A., Pillon, A., Fenet, H., Gomez, E., Casellas, C., Duchesne, M., Hernandez-Raquet, G., Cavaille´s, V., Balaguer, P., Bartegi, A., 2010. Biological analysis of endocrinedisrupting compounds in Tunisian sewage treatment plants.
2484
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 7 3 e2 4 8 4
Archives of Environmental Contamination and Toxicology 59 (1), 1e12. Mispagel, C., Allinson, G., Allinson, M., Shiraishi, F., Nishikawa, M., Moore, M.R., 2009. Observations on the estrogenic activity and concentration of 17b-estradiol in the discharges of 12 wastewater treatment plants in Southern Australia. Archives of Environmental Contamination and Toxicology 56 (4), 631e637. Nakada, N., Tanishima, T., Shinohara, H., Kiri, K., Takada, H., 2006. Pharmaceutical chemicals and endocrine disrupters in municipal wastewater in Tokyo and their removal during activated sludge treatment. Water Research 40 (17), 3297e3303. Ning, B., Graham, N.J.D., Zhang, Y., 2007. Degradation of octylphenol and nonylphenol by ozone - Part I: direct reaction. Chemosphere 68 (6), 1163e1172. Pedrazzani, R., Dal Grande, M., Gal, C., Repice, C., (in preparation). Quantification of bisphenol A, 4-n-nonylphenol, 4-n-nonylphenol-mono and di-ethoxylate, in wastewater and sludge. Planas, C., Guadayol, J.M., Droguet, M., Escalas, A., Rivera, J., Caixach, J., 2002. Degradation of polyethoxylated nonylphenols in a sewage treatment plant. Quantitative analysis by isotopic dilution e HRGC/MS. Water Research 36 (4), 982e988. Pons, M., Gagne, D., Nicolas, J.C., Mehtali, M.A., 1990. A new cellular model of response to estrogens: a bioluminescent test to characterize (anti) estrogen molecules. Biotechniques 9 (4), 450e459. Pothitou, P., Voutsa, D., 2008. Endocrine disrupting compounds in municipal and industrial wastewater treatment plants in Northern Greece. Chemosphere 73 (11), 1716e1723. Press-Kristensen, K., Lindblom, E., Schmidt, J.E., Henze, M., 2008. Examining the biodegradation of endocrine disrupting bisphenol A and nonylphenol in WWTPs. Water Science and Technology 57 (8), 1253e1256. Racz, L., Goel, R.K., 2010. Fate and removal of estrogens in municipal wastewater. Journal of Environmental Monitoring 12 (1), 58e70. Rosenfeldt, E.J., Linden, K.G., 2004. Degradation of endocrine disrupting chemicals bisphenol A, ethinyl estradiol and estradiol during UV photolysis and advanced oxidation processes. Environmental Science & Technology 38 (20), 5476e5483. Sanchez-Avila, J., Bonet, J., Velasco, G., Lacorte, S., 2009. Determination and occurrence of phthalates, alkylphenols, bisphenol A, PBDEs, PCBs and PAHs in an industrial sewage grid discharging to a Municipal Wastewater Treatment Plant. The Science of the Total Environment 407 (13), 4157e4167. Shen, G., Yu, G., Cai, Z., Zhang, Z., 2005. Development of an analytical method to determine phenolic endocrine disrupting chemicals in sewage and sludge by GC/MS. Chinese Science Bulletin 50 (23), 2681e2687. Sole´, M., Lo´pez de Alda, M.J., Castillo, M., Porte, C., LadegaardPedersen, K., Barcelo´, D., 2000. Estrogenicity determination in sewage treatment plants and surface waters from the Catalonian area (NE Spain). Environmental Science & Technology 34 (24), 5076e5083.
Soto, A.M., Maffini, M.V., Schaeberle, C.M., Sonnenschein, C., 2006. Strengths and weaknesses of in vitro assays for estrogenic and androgenic activity. Best Practice & Research Clinical Endocrinology & Metabolism 20 (1), 15e33. Sousa, A., Schonenberger, R., Jonkers, N., Suter, M.J.F., Tanabe, S., Barroso, C.M., 2010. Chemical and biological characterisation of estrogenicity in effluents from WWTPs in Ria de Aveiro (NW Portugal). Archives of Environmental Contamination and Toxicology 58 (1), 1e8. Spink, D.C., Katz, B.H., Hussain, M.M., Pentecost, B.T., Cao, Z., Spink, B.C., 2003. Estrogen regulates Ah responsiveness in MCF-7 breast cancer cells. Carcinogenesis 24 (12), 1941e1950. Stasinakis, A.S., Gatidou, G., Mamais, D., Thomaidis, N.S., Lekkas, T.D., 2008. Occurrence and fate of endocrine disrupters in Greek sewage treatment plants. Water Research 42 (6e7), 1796e1804. Sun, Q.F., Deng, S.B., Huang, J., Shen, G., Yu, G., 2008. Contributors to estrogenic activity in wastewater from a large wastewater treatment plant in Beijing, China. Environmental Toxicology and Pharmacology 25 (1), 20e26. Svenson, A., Allard, A.S., Ek, M., 2003. Removal of estrogenicity in Swedish municipal sewage treatment plants. Water Research 37 (18), 4433e4443. Tan, B.L.L., Hawker, D.W., Muller, J.F., Leusch, F.D.L., Tremblay, L.A., Chapman, H.F., 2007. Comprehensive study of endocrine disrupting compounds using grab and passive sampling at selected wastewater treatment plants in South East Queensland, Australia. Environment International 33 (5), 654e669. Ternes, T., Joss, A., 2006. Human Pharmaceuticals, Hormones and Fragrances: The Challenge of Micropollutants in Urban Water Management. IWA Publishing, London, UK. Urban, G., Golden, T., Aragon, I.V., Scammell, J.G., Dean, N.M., Honkanen, R.E., 2001. Identification of an estrogen-inducible phosphatase (PP5) that converts MCF-7 human breast carcinoma cells into an estrogen-independent phenotype when expressed constitutively. Journal of Biological Chemistry 276 (29), 27638e27646. Vethaak, A.D., Lahr, J., Lahr, S.M., Belfroid, A.C., Rijs, G.B.J., Gerritsen, A., de Boer, J., Bulder, A.S., Grinwis, G.C.M., Kuiper, R.V., Legler, J., Murk, T.A.J., Peijnenburg, W., Verhaar, H.J.M., de Voogt, P., 2005. An integrated assessment of estrogenic contamination and biological effects in the aquatic environment of The Netherlands. Chemosphere 59 (4), 511e524. Vogelsang, C., Grung, M., Jantsch, T.G., Tollefsen, K.E., Liltved, H., 2006. Occurrence and removal of selected organic micropollutants at mechanical, chemical and advanced wastewater treatment plants in Norway. Water Research 40 (19), 3559e3570. Ying, G.G., Kookana, R.S., Kumar, A., Mortimer, M., 2009. Occurrence and implications of estrogens and xenoestrogens in sewage effluents and receiving waters from South East Queensland. The Science of the Total Environment 407 (18), 5147e5155. Zhang, H., Yamada, H., Kim, S.E., Kim, H.S., Tsuno, H., 2006. Removal of endocrine-disrupting chemicals by ozonation in sewage treatment. Water Science and Technology 54 (10), 123e132.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 8 5 e2 4 9 5
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Experimental study of the supercritical water oxidation of recalcitrant compounds under hydrothermal flames using tubular reactors Pablo Cabeza, M. Dolores Bermejo*, Cristina Jime´nez, M. Jose´ Cocero High Pressure Process Group, Dept. Chemical Engineering and Environmental Technology, University of Valladolid, Prado de la Magdalena s/n, 47011 Valladolid, Spain
article info
abstract
Article history:
The hydrothermal flame is a new method of combustion that takes place in supercritical
Received 11 November 2010
water oxidation reactions when the temperature is higher than the autoignition temper-
Received in revised form
ature. In these conditions, waste can be completely mineralized in residence times of
27 January 2011
milliseconds without the formation of by-products typical of conventional combustion.
Accepted 31 January 2011
The object of this work is to study the hydrothermal flame formation in aqueous streams
Available online 19 February 2011
with high concentrations of recalcitrant compounds: an industrial waste with a high concentration of acetic acid and various concentrated solutions of ammonia. A tubular
Keywords:
reactor with a residence time of 0.7 s was used. Oxygen was used as the oxidant and
Ammonia
isopropyl alcohol (IPA) as co-fuel to reach the operation temperature required. The increase
Acetic acid
of IPA concentrations in the feeds resulted in a better TOC removal. For mixtures con-
Tubular reactor
taining acetic acid, 99% elimination of TOC was achieved at temperatures higher than
Autoignition
750 C. In the case of mixtures containing ammonia, TOC removals reached 99% while
Industrial waste
maximum total nitrogen removals were never higher than 94%, even for reaction
Wastewater treatment
temperatures higher than 710 C. Ignition was observed at concentrations as high as 6% wt NH3 with 2% wt IPA while at IPA concentrations below 2% wt IPA, the ammonia did not ignite. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
In conditions above its critical point (T ¼ 374 C and P ¼ 22.1 MPa), water is completely miscible with organic substances and gases. Oxidation reactions can be carried out in this medium in the homogeneous phase, without mass transport limitations. The process using supercritical water most developed at the industrial scale is called supercritical water oxidation (SCWO). It consists of the total oxidation of compounds in a homogenous aqueous medium using air or oxygen as the oxidant. Due to the high temperatures, kinetics are very fast and total waste mineralization can be achieved
using the appropriate combination of temperature and residence time. (Bermejo and Cocero, 2006; Brunner, 2009). At temperatures near the critical point of water (400e450 C), residence times of about 1 min are needed. When organic waste concentrations are high enough and temperatures are above the autoignition temperatures of the compounds, it is possible to form flames called hydrothermal flames. This phenomenon is due to the reduction of the autoignition temperature at high pressures. For inflammable compounds such as methane or methanol, hydrothermal flame can occur at temperatures as low as 400 C (Pohsner and Franck, 1994). Supercritical water oxidation in the presence of hydrothermal
* Corresponding author. Tel.: þ34 983184934; fax: þ34 983423013. E-mail address:
[email protected] (M.D. Bermejo). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.01.029
2486
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 8 5 e2 4 9 5
List of symbols CHAc0 CHAc CIPA0 CIPA CNH30 CNH3 Exc O2 P
Initial concentration of acetic acid, % wt Concentration of acetic acid, % wt Initial concentration of isopropyl alcohol, % wt Concentration of isopropyl alcohol, % wt Initial concentration of ammonia, % wt Concentration of ammonia, % wt Excess O2 over the stoichiometrich amount, % Pressure, MPa
flames can reduce residence times to the order of milliseconds (Augustine and Tester, 2009) without the production of subproducts typical of conventional combustion such as NOx (Bermejo et al., 2008) or dioxins (Serikawa et al., 2002). SCWO with a hydrothermal flame has a number of advantages over the flameless process. Some of these advantages permit overcoming the traditional challenges that make the successful and profitable commercialization of SCWO technology difficult. The advantages include the following (Augustine and Tester, 2009): - It allows the destruction of the pollutants in residence times of a few milliseconds, which permits the construction of smaller reactors. - It is possible to initiate the reaction with feed injection temperatures near to room temperature when using vessel reactors (Bermejo et al., 2011, 2006; Oh et al., 1996). This avoids problems such as plugging and corrosion in a preheating system, having an advantage from the operational and energy integration perspective. - Higher operation temperatures improve the energy recovery Tubular reactors are the optimal device for kinetic determination at high pressures, as well as for studying the ignition process of hydrothermal flames of different compounds, as it allows an accurate monitoring of the evolution of the temperature along different positions in the reactors, making it possible to obtain more accurate experimental data than those obtained in vessel reactors. In a previous investigation by our research group (Bermejo et al., 2009), conditions for the formation of hydrothermal flames of isopropyl alcohol in tubular reactors were studied. It was determined that feed temperature and organic concentration are the most critical variables in hydrothermal flame formation. When working with gases, alcohols and other inflammable compounds, ignition is easily produced (Augustine and Tester, 2009; Pohsner and Franck, 1994; Serikawa et al., 2002; Wellig et al., 2009), but when working with other compounds that are more difficult to oxidize, a co-fuel may be needed to obtain the hydrothermal flame (Leybros et al., 2010) Ammonia (NH3) and acetic acid (HAc) are considered the most recalcitrant compounds for the chemical oxidation processes, including, SCWO, as they are already partially oxidized compounds, frequently acting as reaction intermediates (Li et al., 1991). For this reason, they are considered to be key compounds for the study of the SCWO process.
T tR TOC TN Tmax Tinj XNH3 z
Temperature, C Residence time, s Total Organic Carbon, ppm Total Nitrogen, ppm Maximum temperature in the reactor, C Injection temperature in the reactor, C Ammonia conversion Reactor length, m
The destruction of HAc has been reported by several authors (Mateos et al., 2005; Meyer et al., 1995; Savage and Smith, 1995). In these investigations, different kinds of reactors were used at temperatures between 400 and 600 C in a flameless regime obtaining conversions over 90% in residence times of 4e30 s. The main products of acetic acid oxidation were CO2 and water. Minor products included CO and trace amounts of methane, the main radical species in the mechanism are formic acid and methanol, a detailed mechanism for the oxidation of Acetic Acid was published by Boock and Klein, (Boock and Klein, 1993). The final oxidation product of nitrogen compounds in supercritical water is N2 (Killilea et al., 1992). Even though the complete oxidation product of nitrogen is NO 3 , in the SCWO process the oxidation to N2 is favored thermodynamically. It is known that ammonia is the main reaction intermediate and that the oxidation of ammonia is the rate-limiting step in the overall oxidation of nitrogen-containing organic waste (Al-Duri et al., 2008). This makes the study of ammonia oxidation an essential step in order to improve the process design in SCWO technology. In addition, the oxidation of ammonia is one of the slowest reactions in SCWO and it can be used for testing the performance of a new reactor design. Numerous difficulties for completely destroying ammonia in SCWO have been reported in literature, with important divergences: while some authors did not eliminate NH3, others, in similar conditions, obtained significant removals (Goto et al., 1999; Ploeger et al., 2006; Segond et al., 2002). Our group’s results showed complete destruction of nitrogen compounds and ammonia using IPA as a co-fuel in a vessel cooled-wall reactor working at flame regime with residence times in the reactor of about 1 min (Bermejo et al., 2008). In this study, hydrothermal flame ignition of two different wastes was performed: - A real industrial waste composed mainly of acetic acid, and containing also crotonaldehyde, a carcinogenic compound. - Synthetic waste consisting of different aqueous solutions of the recalcitrant compound ammonia. A simple tubular reactor with residence times of less than 0.7 s was used in this study. Isopropyl alcohol (IPA) was used as a co-fuel to provide the enthalpy of combustion needed to obtain the flame ignition. The concentrations of recalcitrant compounds were increased while the concentrations of the cofuel were decreased in order to observe how the ratio of waste to
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 8 5 e2 4 9 5
co-fuel can affect flame ignition and waste destruction. Previous to this investigation, the ignition of IPA solutions at several concentrations using oxygen as the oxidant were studied.
are opposite each other. It is similar to the experimental setup denominated as “mixer 1” described in a previous work (Bermejo et al., 2009).
2.2.
2.
Experimental
2.1.
Experimental set-up
All the experiments described in this study were carried out in the demonstration plant owned by the University of Valladolid and located at the industrial site of the company CETRANSA in Santovenia de Pisuerga (Valladolid, Spain). The demonstration plant can work at temperatures up to 750 C and pressures up to 30 MPa. Feed flows of 9 kg/h were used for these experiments. Feed was electrically preheated before being introduced in the tubular reactor. The oxidant (oxygen) is injected into the feed at the inlet of the tubular reactor. Temperatures in the mixing point and in several points of the tubular reactor were monitored. After leaving the reactor, the reaction mixture was quenched. Then products were cooled in the intercoolers and finally depressurized. Samples of the liquid were taken, and the concentrations of NH3 and NOx in the gaseous stream were checked for selected experiments. A flow diagram of the pilot plant is shown in Fig. 1.The main elements of the facility are the following: - A tank where the feed is prepared, with special attention given to the concentration of organic matter, because this concentration determines the operating temperature of the reactor. - Metering diagram pump (DOSAPRO MAX ROYAL C) to drive the feed from the feed tank to the tubular reactor. Its maximum work pressure is of 23 MPa. - Two electrical preheaters in the feed line, They consist of cylinders in which four resistance of 2500 W each are embedded. A coil is surrounding the cylinder. The whole device is thermally isolated. - The oxygen supply facility. Liquid oxygen is stored in a cryogenic deposit, from where it is pumped by a cryogenic metering pump until work pressure. The pump supplies a constant flow. After pumping, oxygen is vaporized in a finned tube heat exchanger. Then, the pressurized gaseous oxygen is stored in four reservoirs before it is mixed with the aqueous waste stream. The flow of oxygen withdrawn from these reservoirs can be controlled with a control valve. - The reactor consisted of a straight and empty tube made of Ni alloy C-276 with a length of 2000 mm and a diameter of ¼” (i.d. 3.86 mm) giving an internal volume of 18.5 mL. The flow inside the tube was upwards. - Cooling systems. They consist of two coolers, in which the hot product of reaction flows inside a titanium alloy (Tie3Ale2.5V) coil, and is refrigerated by cooling water. - The back pressure regulator valve consisting of a needle valve, SENTRY VREL11, is used. For security reasons, a second valve is placed in parallel to this valve. - Flash chamber separator and sampling device that allows taking samples of the liquid and gaseous effluents. Temperature was monitored in 6 internal points of the reactor using 6 thermocouples type K. Feed and oxygen inlets
2487
Experimental procedure
At the beginning of a new experiment, the reactor was preheated by pumping an initial feed of water and IPA (4% of IPA) through it until all the thermocouples indicated temperatures higher than 400 C. Once preheated the reactor, oxygen flow was introduced at 0e10% of the stoichiometric amount in order to start the ignition. When a sharp increase in the temperature occurred, it indicates that the hydrothermal flame was ignited, and in that moment, quenching water was connected. Apart from the electrical preheating of the reagents, no more heat was supplied to the system, the reactor being auto-thermal Even when the flame could not visually be observed, it is known that hydrothermal flames are formed at these temperatures and IPA concentrations (Serikawa et al., 2002). On the other hand the sharp increase of temperature happens in a stretch of reactor shorther than 40 cm of reactor. This implies the oxidation of IPA in a residence time of 100 mS, being this residence time typical of SCWO at flame regime, while flameless SCWO requires higher residence times. This indicates that all the rises in temperature produced inside of the reactor were due to the release of the heat of reaction during the oxidation of IPA. The system was maintained in a stationary state and samples were taken at the outlet of the system. Temperature profiles along the reactor were registered automatically when a sample was taken. After sample taking, experimental conditions were altered by changing the concentration of IPA and the one of waste or ammonia (depending on the experiment carried out), the flow of feed is maintained constant at 9 kg/h for all the experiments. All the experiments were monitored through temperature profiles in the reactor and TOC concentrations in the effluent.
2.3.
Materials
Isopropanol (99% in mass) and Ammonia (25% in mass) were supplied by COFARCAS (Spain). Experiments using a real industrial waste, mainly composed of acetic acid, were performed. The composition of the waste is shown in Table 1. Experiments were performed using different dilutions of the waste in order to obtain different concentrations of acetic acid.
2.4.
Analysis
All the Total Organic Carbon (TOC) analysis and Total Nitrogen (Total N) analysis of the samples were performed with a TOC 5050 SHIMADZU Total Organic Carbon Analyzer which uses combustion and IR analysis. The detection limit is 1 ppm. NH3 and NOx in the gas effluent were analyzed with Dra¨ger tubes detectors Lab Safety Supply CH29401 and CH31001. The NOx detection limits for these tubes ranged from 0.5 to 100 ppm and the NH3 detection limits for these tubes ranged
2488
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 8 5 e2 4 9 5
Fig. 1 e Flow diagram of the facility.
from 5 to 70 ppm (standard deviation for both tubes are between 10 and 15%). Nitrates and nitrites were characterized in the liquid effluent by ionic chromatography with an IC PAK A column of Waters. The detection limit is 1 ppm.
3.
Results and discussion
3.1. Influence of IPA concentration on the hydrothermal flame formation A study was conducted examining the influence of IPA concentrations on hydrothermal flame ignition when IPA was the only fuel present in the aqueous mixture. The experiments performed are summarized in Table 2. Each experiment was repeated at least three times. Standard deviations were calculated and shown in Table 2.
Table 1 e Composition of the industrial waste. Compound Acetic Acid Acetaldehyde Crotonaldehyde Methanol Cl
Composition % 83.53 8.95 1.79 0.95 4.77
The experimental temperature profiles obtained are depicted in Fig. 2. It is observed that when decreasing fuel concentration, the ignition and the point where the highest temperature is reached are delayed. Even when most of the IPA has reacted at that point, complete conversion is not obtained. As expected, increasing the concentration of IPA, the maximum temperature registered in the reactor is higher and the obtained conversion is higher as well. The observations of Vogel et al., (Vogel et al., 2005) can explain this phenomenon. According to these authors, the SCWO reaction mechanism of methanol yields a S-shaped conversion vs. time curve. At the beginning of the reaction an induction time
Table 2 e Summary of the experiments performed with IPA as the only fuel in the mixture. Average values and standard deviations are shown. CIPA0
TOC0
(% wt) 4.5 4.0 3.5 2.5 2.0 1.5 1.0
(ppm C) 27,000 24,000 21,000 15,000 12,000 9000 6000
TInj ( C) 406 404 415 416 407 407 390
Tmax
3 4 2 8 3 4 8
( C) 708 676 631 600 566 516 440
4 1 1 1 17 10 9
TOC TOCoutlet Exc. O2 Removal (%) 99.7 0.1 99.3 0.2 98.5 0.3 96 2 97 2 94 1 62 5
(ppm C) 77 23 172 37 323 68 494 252 344 294 575 97 2300 331
(%) 38 3 76 14 86 16 70 59 55 28 85 15 145 44
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 8 5 e2 4 9 5
3.2. acid
790 4.5 %
740
2489
Oxidation of an industrial waste containing acetic
4.0 %
690
4.5% 4.0% 3.5% 2.5% 2.0% 1.5% 1.0%
3.5 %
T (ºC)
640 590 2.5 %
540 490
2.0 % 1.5 %
440 1.0 %
390 0,00
0,20
0,40
0,60
0,80
1,00
tR (s)
Fig. 2 e Temperature profile registered in experiments with different IPA concentrations. Comparison to predicted temperature profiles in flameless SCWO. Symbols linked by discontinuous line represent experimental data and solid lines represent prediction of the model for flameless oxidation.
is observed; then in a second stage, the methanol is consumed rapidly in a quasi-steady propagation phase. When most of the reagent has been converted, termination reactions between radicals, which yield stable products, start to dominate and methanol conversion is slowed down (termination phase). Thus, even though the flame is generated and most of the TOC is eliminated, the termination step is slower and the reaction needs additional residence time to be completed; this necessary residence time decreases when reaction temperature is higher. The experimental temperature profiles are compared to the flameless temperature profile predicted by a mathematical model described in a previous work by our research work group, (Bermejo et al., 2009). The model consists of a tubular reactor in which plug flow is considered. The reactor works in stationary state. Only mass and energy balances are solved. Runge Kutta method of the 4th order was used to solve the model. Densities and enthalpies were calculated as a function of composition and temperature using the PR EoS with the translated volume correction (Magoulas, 1990). The kinetic of Wightman (Wightman, 1981) published by Li et al. (Li et al., 1991) was used. This kinetics was developed for phenol oxidation but, in a previous research (Bermejo et al., 2009) several kinetic models were tested and it was the best reproducing the experimental flameless temperature profile of IPA. It is shown that experimental temperature profiles are faster than the flameless temperature profiles predicted by the model even for the experimental data with 1% in mass, in which only 62% is removed. Nevertheless, a maximum temperature is reached, and even though there is more IPA to be oxidized, it is not, and no further temperature increase is registered. Results indicate that upon reaching a concentration low enough, flame regime is no longer possible, and for low IPA concentrations the TOC removal is lower than for higher TOC removals.
Experiments for the oxidation of the industrial waste were performed using different dilutions of waste with concentrations of acetic acid from 1.4 to 4% in mass. IPA was used as the co-fuel, thereby increasing the calorific power of the aqueous solutions in order to work at temperatures between 600 and 700 C. When waste concentration was increased, IPA concentration was decreased in order to keep the temperature value in this range. If TOC removal was not complete, IPA concentration was increased in order to increase the reaction temperature. All the experiments were performed with a fixed flow of 9 kg/h, injection temperatures between 390 and 410 C, and a pressure of 23 MPa, using oxygen as oxidant in concentrations above the stoichiometric proportion to achieve complete conversion to CO2. Every experiment was repeated at least 3 times. The average values and the standard deviation of the data are showed in Table 3. Note that the uncertainties of experimental data are higher in a pilot scale plant than in a laboratory scale plant. As expected, by increasing the amount of organic material in the feed, the maximum temperature reached in the reactor is higher and, consequently, the conversion of TOC increases as well. This behavior is shown in Fig. 3(a) where the TOC removals versus the maximum temperatures registered in the reactor are shown for several waste concentrations. Fig. 3b compares the TOC removal versus maximum reaction temperature in feeds with different IPA concentrations, and in feeds with the same IPA concentrations and 1.4% acetic acid. Higher TOC removals were observed for mixtures not containing acetic acid. Almost total TOC removals (>99.8%) were obtained at temperatures around 708 C for feeds containing only IPA, while those containing IPAþ1% HAc needed a temperature higher than 750 C to obtain 99% TOC removal. Despite good results of TOC removal of mixtures of wasteIPA, higher waste concentrations could not be tested due to significant corrosion problems detected in the preheating area. It was found that samples contained Ni concentrations as high as 232 ppm, originated in the corrosion of the Ni alloy tube. Cr concentrations of 23 ppm and Fe concentrations of 42 ppm were also found. This can be explained by the high chloride concentration of the waste. Fig. 4 shows the temperature profile along the reactor. It can be observed that the maximum residence time in the reactor is around 0.7 s. The sharp temperature increase, indicating the ignition of the hydrothermal flame, occurs at the beginning of the reactor with residence times between 100 and 150 ms, meaning that most of the TOC is eliminated in that point; but if reaction temperature is not high enough, complete TOC removal is not achieved. Again, extra residence time is needed to finish the termination step of the reaction (Vogel et al., 2005).
3.3.
Oxidation of aqueous mixtures of NH3eIPA
In order to understand hydrothermal flame behavior when dealing with nitrogen-containing compounds, experiments with concentrations from 2 to 8% of ammonia in mass were
2490
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 8 5 e2 4 9 5
Table 3 e Experiments performed with a highly acetic acid concentrated industrial waste. CHAcO (% wt)
1.4 1.4 1.4 2.0 3.0 4.0 4.0
CIPAO
TOC0
(% wt)
3.5 4.0 4.5 3.3 3.5 2.5 3.0
TOCoutlet
(ppm C)
(ppm C)
HAc
IPA
5600 5600 5600 8000 12,000 16,000 16,000
21,000 24,000 27,000 19,800 21,000 15,000 18,000
1574 257 57 1082 135 1335 139
689 137 28 195 31 473 23
performed. IPA was used as a co-fuel in order to reach temperatures between 600 and 750 C. To improve the TOC removal, IPA concentrations were increased to elevate the reaction temperature. All of the experiments were performed with a fixed flow of 9 kg/h, injection temperatures between 400 and 430 C and a pressure of 23 MPa using oxygen as the oxidant for the total oxidation of the feed to CO2 and N2. Each experiment was repeated at least 3 times. The average values and the standard deviation of the data are showed in Table 4. As shown in Table 4, it is possible to obtain TOC removal over 99% for mixtures of NH3eIPA. However, in the case of the elimination of ammonia, the maximum removal was 94% at temperatures of approximately 750 C. In general, TOC removal and Total N are improved with high temperatures. It is observed that, typically, when TOC removal is higher, N removal increases as well. Higher enthalpies content, that is, higher reaction temperatures, were needed to achieve TOC removals of over 99% when ammonia concentration was higher in the mixture. The improvements in N removal for temperatures higher than 710 C were relatively small. In our previous work (Bermejo et al., 2008), it was possible to get more than 99% removal of ammonia with temperatures close to 800 C in a cooled-wall reactor with a total residence time of 1 min. Thus, it can be inferred that for the total destruction of ammonia, longer residence times or higher reaction temperatures are needed than those for which the facility was designed.
TInj
Tmax
( C)
409 409 396 398 396 387 403
8 16 8 4 8 4 4
( C)
670 709 766 690 745 703 744
21 20 10 15 12 10 4
TOC Removal
Exc O2
(%)
(%)
409 409 396 398 396 387 403
8 16 8 4 8 4 4
28 38 6 33 25 50 1 14 89 58 5 16 25 23
The temperature profiles registered along the reactor for different experiments with increasing NH3 concentration and decreasing IPA concentration are shown in Fig. 5 It is observed that with an ammonia concentration of 4% in mass and IPA concentrations of 3 and 2.5% in mass, ignition is produced at the beginning of the reactor. In this case, the IPA auto-ignites rapidly, and with the heat released, a sharp increase in temperature is produced, the autoignition temperature of ammonia is surpassed and it also reacts in flame regime, observing only one temperature increase. Even when the ignition was registered at 0.1 s, at which point most of the pollutants were consumed, the TOC and N removal were not complete. As occurred in the case of HAc, the last step of ammonia elimination is probably slower, as in the case of methanol (Vogel et al., 2005), and the reaction could not be completed in the 0.7 s residence time. For higher NH3/IPA ratios the reaction proceeded more slowly, and the temperature profiles have been compared to the flameless temperature profiles predicted by the model. These comparisons are shown in Fig. 6. When the concentration of ammonia increased to 6% and that of IPA decreased to 2%, the position of the flame moved away from the injection point, reaching the maximum temperature only at residence times of 0.2 s. The temperature profile in the first 0.05 s coincided with the flameless temperature profile predicted by the model, as shown in Fig. 6a). Because the concentration of IPA was lower, it took
Fig. 3 e a) TOC removal versus maximum reaction temperature for mixtures with 1.4% and 4% acetic acid. Experimental conditions of the experiments: feed flow 9.5 kg/h flow. Oxygen excess over the stoichiometric concentration for complete oxidation to CO2 and water P [ 23 MPa. b) TOC removal versus maximum reaction temperature for mixtures of different IPA concentrations and IPAD 1.4 wt% acid acetic concentration.
2491
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 8 5 e2 4 9 5
Fig. 4 e Profile temperature for different experiments with acetic acid and IPA. (a) 1% acetic acid in mass; (b) 4% Acetic acid in mass.
Table 4 e Experiments performed with NH3eIPA mixtures. CIPA0 (% wt) 3.2 3.8 2.5 3.0 1.7 2.0 1.1
CNH30
TOC0
TN0
TInj
Tmax
TOC Outlet
TOC Removal
TN Outlet
TN removal
Exc O2
(% wt)
(ppm C)
(ppm N)
( C)
( C)
(ppm C)
(%)
(ppm N)
(%)
(%)
2.0 2.0 4.0 4.0 6.0 6.0 8.0
19,200 22,800 15,000 18,000 10,200 12,000 6000
16,470 16,470 32,941 32,941 49,412 49,412 65,882
428 21 408 2 418 3 412 1 421 2 414 1 405 1
710 17 727 1 709 2 762 2 609 2 744 3 574 2
157 48 106 37 156 922 186 226 422 80 103 72 476 3
96 3 99.5 0.2 98.6 0.5 99.0 1.2 95.9 0.8 97 3 92.1 0.1
1033 99 1101 60 3032 555 2754 208 12,360 593 3412 261 612,714 18
93.7 0.6 93.3 0.4 90.8 1.7 91.6 0.6 75.0 1.2 93.1 0.5 7.0 0.6
30 27 57 24 37 11 24 10 11 20 6 23 31 4
longer for the mixture to be oxidized and release the heat necessary to reach the autoignition temperature, first of IPA and later of NH3, as occurred in the previous case. When NH3 concentration was 6% and IPA concentration was 1.7%, the reaction proceeded even more slowly. But a maximum temperature is reached that is not further increased. The maximum temperature was reached only at 0.3 s, and only 95.9% IPA removal and 75% Total N removal were obtained. When compared with the flameless temperature profile predicted by the model, we observe that in the first 0.1 s residence time the temperature profile is equal to the flameless temperature profile as indicated in Fig. 7b). This occurs just before reaching the adiabatic flame temperature, which is calculated by considering only the oxidation of IPA.
Fig. 5 e Influence of Ammonia concentration for the hydrothermal flame formation.
At longer residence times, the reaction proceeded faster, which indicates that ammonia begin to react at that point. Even though the temperature increase was less sharp, IPA removal was near to 96%. Thus, apparently IPA ignition took place. The residence time at which maximum temperature occurs coincides with the maximum temperature in ignition with 1 and 2% IPA in mass. Thus, ammonia is not delaying the ignition of IPA as observed in Fig. 7. Ammonia removal was only 75%, reaching a maximum temperature of 600 C. But even with the 25% of the initical NH3 content unreacted, no further temperature increase was registered, indicating that ammonia is not reacting beyond 75% conversion, at least in its propagation step at this ratio of IPA/NH3. With NH3 concentration at 8% in mass and IPA concentration at 1% in mass, there was no sharp changes in the experimental temperature profile. Nevertheless, the temperature increased faster than indicated by the predicted flameless temperature profile showed in Fig. 7c). Surprisingly, under those conditions a 93% of TOC removal was obtained while elimination in pure 1% IPA solution was only at 63%. Only 7% of the Total N was removed, but apparently the heat released by this oxidation contributed to a higher IPA elimination due to the temperature increase. When ammonia was oxidized without IPA, no temperature increase was detected even at NH3 concentrations as high of 10% NH3 in mass. Based on these results, it is believed that the ignition temperature of ammonia may be much higher than that of IPA. Thus, when high concentrations of IPA are used, ignition of IPA is produced and high temperatures lead to ammonia
2492
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 8 5 e2 4 9 5
Fig. 6 e (a) Comparison of the experimental temperature profile to the predictions of the kinetic models for a mixture of 2% of IPA and 6% of NH3, (b) for a mixture of 1.7% of IPA and 6% of NH3. (c) for a mixture of 1% of IPA and 8% of NH3.
ignition and consequently to high ammonia removals, but the termination phase of the oxidation of NH3 is too slow to reach a complete N elimination. When the concentration of IPA is too low for flame ignition, ammonia ignition is not reached, Total N removal is even slower and neither TOC nor N are completely eliminated. Experimental temperature profiles of the oxidation of mixtures of 2.5% IPA only, 2.5% of IPA with 4% NH3 and 2.5% IPA with 4% HAc, are compared with the flameless temperature profile calculated with the model in Fig. 8. Ignition is produced from the beginning of the reactor in all cases. When HAc or NH3 are present in the mixture, temperature profiles are very similar. Only one sharp temperature increase can be observed indicating that the ignition of IPA readily leads to the ignition of the most recalcitrant compounds. Moreover, when
Fig. 7 e Comparison of the ignition of mixtures 6% NH3, 1.7% IPA to the ignition of IPA mixtures with 1.5 and 2% wt IPA.
IPA is reacting with other components, the temperature increase is faster than when it is reacting alone in the same concentration. Thus, the heat released by the oxidation of other compounds is accelerating the oxidation of IPA. In the experiments with the highest concentration of IPA, for each ammonia concentration the presence of NOx and NH3 was investigated in the gas effluent and the concentration of NO 3 in liquid effluent. The average results are shown in Table 5. It is observed that for higher NH3/IPA ratios the Total N is increased in the liquid effluent. Concentrations of NH3 between 4 and 60 ppm were found in the gas effluent. This concentration is increased when NH3 concentrations are higher in the feed. Nevertheless, concentrations of NOx in the gas effluent were below of the detection limit of 5 ppm. Concentrations of nitrates and nitrites ðNO 3 Þ between 300 and 2000 ppm were found in the liquid effluents. In general, NO 3 concentration is higher in the samples with lower initial
Fig. 8 e Profile temperatures for experiment of mixtures with IPA and HAc and NH3.
2493
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 8 5 e2 4 9 5
Table 5 e Average results of the total nitrogen, ammonia and nitrates present in samples taken form the gas effluent and liquid effluent flow. CIPA0
TN (liquid)
NH3 in Gas efluent
NOx in Gas efluent
CNO3 (liquid)
XNH3
Converted to NO3
%
%
(ppm N)
(ppm NH3)
(ppm NOx)
(ppm NO3)
(%)
(%)
2.0 4.0 6.0
3.8 3.0 2.0
1101 61 2754 208 3412 262
<5 20 60
<0.5 <0.5 <0.5
2012 349 1767 115 371 67
96.1 0.7 92.9 0.6 93.3 0.5
2.7 0.5 1.2 0.1 0.2 0.1
CNH30
Fig. 9 e (a) Nitrates concentration present in the liquid effluent versus the O2 in excess used for samples in a range of temperature of 720e750 C for samples with different ammonia concentrations, (b) Nitrates concentration present in the liquid effluent versus the maximum temperature registered in the reactor for samples with a flow in excess of O2 between 30 and 50% for samples with different ammonia concentration.
NH3 concentration; the highest overall N removal was obtained in these samples. In Fig. 9, the NO 3 concentration is represented versus excess of O2 with respect to the stoichiometric value and versus maximum reaction temperatures for samples with different initial ammonia concentrations. In general, NO 3 concentration increases with higher reaction temperature and higher oxygen excesses, as was observed our group’s previous studies (Bermejo et al., 2008; Cocero et al., 2000). So far, the flameles SCWO has been commecially applied to several waste such as sludge, Chemicals weapons etc (Bermejo and Cocero, 2006), sometimes using tubular reactors as long as 1 km long (O’Regan, 2010). The results show in this work prove that the SCWO technology using hydrotermal flame as a heat source is able to destroy high amounts of recalcitrant compounds in residence times of around a seconds, using very compact reactors. The commercialization of this technology can be promising for the treatment of waste with elevated ammonia content such as sludge (with ammonia containing around 1%) or for waste derived from electronic industry, with elevated ammonia contents. Nevertheless, previous to its commercialization the reactor design should still be optimize in order to find the appropriate residence time/temperature combination to completely destroy ammonia without generating NO 3 , and to be able to inject cold waste into the flame to avoid corrosion in the preheating area.
4.
Conclusions
The destruction of waste containing high concentrations of the recalcitrant compounds acetic acid and ammonia by supercritical water oxidation in the presence of a hydrothermal flame
using isopropyl alcohol as a co-fuel and oxygen as the oxidant was experimentally studied using a tubular reactor with a residence time of 0.7 s. In addition, as a preliminary step, the ignition of IPA solutions of different concentrations was studied. - Extremely high TOC and nitrogen conversions were obtained at the present operational conditions. Conversions increased with temperature up to a certain point: - TOC removals were as high as 99.7% at temperatures higher than 700 C when working with IPA solution. - 99% mineralization at temperatures as high as 750 C in concentrations up to 4% in mass of acetic acid: TOC removal was more difficult than in the oxidation of equivalent IPA solutions. - Maximum 99% removals of TOC and 94% of N were obtained. No significant improvement in N removal was found at temperatures higher than 710 C. - Concentrations of NH3 between 4 and 60 ppm found in the gas effluent. - Concentrations of NOx in the gas effluent were below 0.5 ppm. - Concentrations of nitrates and nitrites between 300 and 2000 ppm were found. The amount of NO 3 is favored by higher temperatures and oxygen excesses. - When using IPA as a fuel and a recalcitrant compound, only one ignition point was observed, meaning that the increase of temperature produced by IPA ignition makes it possible to reach the autoignition conditions of the recalcitrant compounds as well. - Impossible to obtain ignition using exclusively ammonia as fuel. A minimum concentration of 2% IPA in the
2494
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 8 5 e2 4 9 5
ammoniaeIPAewater mixture was necessary to obtain more than 90% ammonia elimination. - Most of the TOC and Total N were eliminated at the moment of the ignition, in a rapid temperature increase for forming the flame, but a remaining fraction of the pollutant could not be eliminated using this tubular reactor. Very high temperatures or additional residence time seem are necessary. This is due to the low rate of oxidation reaction during the termination phase. This was especially remarkable in the case of ammonia, where eliminations higher than 93% were not possible in the operational conditions of our reactor. - Even though tubular reactors are entirely appropriate for the study of the ignition process of different compounds, the reactor design must be optimize to get a compact apparatus able to work with the appropriate residence time/temperature combination to completely destroy ammonia without generating NO 3 , and to be able to inject cold waste into the flame to avoid corrosion in the preheating area.
5.
Future work
- Study of the elimination of recalcitrant compounds using a hydrothermal flame in a vessel reactor with longer residence times and lower temperatures in order to avoid NO3 formation. - Avoiding corrosion problems in the preheating area by injection of the feed at room temperature into a hydrothermal flame.
Acknowledgments The authors thank CETRANSA for providing technical support and the Spanish Ministry of Science and Innovation for the Project CTQ2010-15475 (subprogram PPQ). M.D.B. thanks the program Juan de la Cierva (JCI-2008-02877). P.C. thanks Junta de Castilla y Leo´n for predoctoral Grant.
Appendix. Supplementary material Supplementary data related to this article can be found online at doi:10.1016/j.watres.2011.01.029.
references
Al-Duri, B., Pinto, L., Ashraf-Ball, N.H., Santos, R.C.D., 2008. Thermal abatement of nitrogen-containing hydrocarbons by non-catalytic supercritical water oxidation (SCWO). Journal of Materials Science 43 (4), 1421e1428. Augustine, C., Tester, J.W., 2009. Hydrothermal flames: from phenomenological experimental demonstrations to quantitative understanding. Journal of Supercritical Fluids 47 (3), 415e430. Bermejo, M.D., Cocero, M.J., 2006. Supercritical water oxidation: a technical review. Aiche Journal 52 (11), 3933e3951.
Bermejo, M.D., Fdez-Polanco, E., Cocero, M.J., 2006. Experimental study of the operational parameters of a transpiring wall reactor for supercritical water oxidation. Journal of Supercritical Fluids 39 (1), 70e79. Bermejo, M.D., Cantero, F., Cocero, M.J., 2008. Supercritical water oxidation of feeds with high ammonia concentrations: pilot plant experimental results and modeling. Chemical Engineering Journal 137 (3), 542e549. Bermejo, M.D., Cabeza, P., Bahr, M., Fernandez, R., Rios, V., Jimenez, C., Cocero, M.J., 2009. Experimental study of hydrothermal flames initiation using different static mixer configurations. Journal of Supercritical Fluids 50 (3), 240e249. Bermejo, M.D., Cabeza, P., Queiroz, J.P.S., Jime´nez, C., Cocero, M.J., 2011. Analysis of the scale up of a transpiring wall reactor with a hydrothermal flame as a heat source for the supercritical water oxidation. The Journal of Supercritical Fluids 56 (1), 21e32. Boock, L.T., Klein, M.T., 1993. Lumping strategy for modeling the oxidation of c1-c3 alcohols and acetic-acid in hightemperature water. Industrial & Engineering Chemistry Research 32 (11), 2464e2473. Brunner, G., 2009. Near and supercritical water. Part II: oxidative processes. Journal of Supercritical Fluids 47 (3), 382e390. Cocero, M.J., Alonso, E., Torio, R., Vallelado, D., Fdz-Polanco, F., 2000. Supercritical water oxidation in a pilot plant of nitrogenous compounds: 2-propanol mixtures in the temperature range 500e750 degrees C. Industrial & Engineering Chemistry Research 39 (10), 3707e3716. Goto, M., Nada, T., Kodama, A., Hirose, T., 1999. Kinetic analysis for destruction of municipal sewage sludge and alcohol distillery wastewater by supercritical water oxidation. Industrial & Engineering Chemistry Research 38 (5), 1863e1865. O’Regan, J., Preston, S., Dune, A., 2010. Supercritical Water Oxidation of Sewage Sludge-An Update, in SCFI. Killilea, W.R., Swallow, K.C., Hong, G.T., 1992. The fate of nitrogen in supercritical water oxidation. Journal of Supercritical Fluids 5 (1), 72e78. Leybros, A., Roubaud, A., Guichardon, P., Boutin, O., 2010. Ion exchange resins destruction in a stirred supercritical water oxidation reactor. The Journal of Supercritical Fluids 51, 369e375. Li, L., Chen, P., Gloyna, E.F., 1991. Generalized kinetic-model for wet oxidation of organic-compounds. Aiche Journal 37 (11), 1687e1697. Magoulas, D.T.C., 1990. Thermophysical properties of n-alkanes from C1 to C26 and their prediction for higher ones. Fluid Phase Equilibria 56, 119e140. Mateos, D., Portela, J.R., Mercadier, J., Marias, F., Marraud, C., Cansell, F., 2005. New approach for kinetic parameters determination for hydrothermal oxidation reaction. Journal of Supercritical Fluids 34 (1), 63e70. Meyer, J.C., Marrone, P.A., Tester, J.W., 1995. Acetic-acid oxidation and hydrolysis in supercritical water. Aiche Journal 41 (9), 2108e2121. Oh, C.H., Kochan, R.J., Charlton, T.R., Bourhis, A.L., 1996. Thermalhydraulic modeling of supercritical water oxidation of ethanol. Energy & Fuels 10 (2), 326e332. Ploeger, J.M., Madlinger, A.C., Tester, J.W., 2006. Revised global kinetic measurements of ammonia oxidation in supercritical water. Industrial & Engineering Chemistry Research 45 (20), 6842e6845. Pohsner, G.M., Franck, E.U., 1994. Spectra and temperatures of diffusion flames at high-pressures to 1000 bar. Berichte Der Bunsen-Gesellschaft-Physical Chemistry Chemical Physics 98 (8), 1082e1090. Savage, P.E., Smith, M.A., 1995. Kinetics of acetic-acid oxidation in supercritical water. Environmental Science & Technology 29 (1), 216e221.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 8 5 e2 4 9 5
Segond, N., Matsumura, Y., Yamamoto, K., 2002. Determination of ammonia oxidation rate in sub- and supercritical water. Industrial & Engineering Chemistry Research 41 (24), 6020e6027. Serikawa, R.M., Usui, T., Nishimura, T., Sato, H., Hamada, S., Sekino, H., 2002. Hydrothermal flames in supercritical water oxidation: investigation in a pilot scale continuous reactor. Fuel 81 (9), 1147e1159. Wightman, T.J., 1981. Studies in Supercriticalair Oxidation Berkeley, CA.
2495
Vogel, F., Blanchard, J.L.D., Marrone, P.A., Rice, S.F., Webley, P.A., Peters, W.A., Smith, K.A., Tester, J.W., 2005. Critical review of kinetic data for the oxidation of methanol in supercritical water. Journal of Supercritical Fluids 34 (3), 249e286. Wellig, B., Weber, M., Lieball, K., Prikopsky, K., von Rohr, P.R., 2009. Hydrothermal methanol diffusion flame as internal heat source in a SCWO reactor. Journal of Supercritical Fluids 49 (1), 59e70.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 9 6 e2 5 0 6
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
In situ oxidation of petroleum-hydrocarbon contaminated groundwater using passive ISCO system S.H. Liang a, C.M. Kao a,*, Y.C. Kuo a, K.F. Chen b, B.M. Yang a a b
Institute of Environmental Engineering, National Sun Yat-Sen University, Kaohsiung 804, Taiwan Department of Civil Engineering, National Chi-Nan University, Nanto County, Taiwan
article info
abstract
Article history:
Groundwater contamination by gasoline spill is a worldwide environmental problem.
Received 20 September 2010
Gasoline contains methyl tertiary-butyl ether (MTBE) (a fuel oxygenates) and benzene,
Received in revised form
which are the chemicals of concerns among the gasoline components. In this study, an in
3 February 2011
situ chemical oxidation (ISCO) barrier system was developed to evaluate the feasibility of
Accepted 4 February 2011
applying this passive system on the control of MTBE and benzene plume in aquifer. The
Available online 12 February 2011
developed ISCO barrier contained oxidant-releasing materials, which could release oxidants (e.g., persulfate) when contact with water for the contaminants’ oxidation in
Keywords:
groundwater. In this study, laboratory-scale fill-and-draw experiments were conducted to
Barrier
determine the component ratios of the oxidant-releasing materials and evaluate the per-
Groundwater contamination
sulfate release rates. Results indicate that the average persulfate-releasing rate of 7.26 mg
In situ oxidation
S2O2 8 /d/g was obtained when the mass ratio of sodium persulfate/cement/sand/water was
Persulfate
1/1.4/0.24/0.7. The column study was conducted to evaluate the efficiency of in situ
Petroleum-hydrocarbon
application of the developed ISCO barrier system on MTBE and benzene oxidation. Results from the column study indicate that approximately 86e92% of MTBE and 95e99% of benzene could be removed during the early persulfate-releasing stage (before 48 pore volumes of groundwater pumping). The removal efficiencies for MTBE and benzene dropped to approximately 40e56% and 85e93%, respectively, during the latter part of the releasing period due to the decreased persulfate-releasing rate. Results reveal that acetone, byproduct of MTBE, was observed and then further oxidized completely. Results suggest that the addition of ferrous ion would activate the persulfate oxidation. However, excess ferrous ion would compete with organic contaminants for persulfate, and thus, cause the decrease in contaminant oxidation rates. The proposed treatment scheme would be expected to provide a more cost-effective alternative to remediate MTBE, benzene, and other petroleum-hydrocarbon contaminated aquifers. Results from this study will be useful in designing a scale-up system for field application. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
Accidental releases of petroleum products (e.g., gasoline, diesel, fuel oil) from pipelines, aboveground storage tanks (ASTs), and underground storage tanks (USTs) are the most
common causes of groundwater contamination in most countries. In most cases, released petroleum products would form non-aqueous phase liquids (NAPLs), which will become the long-term source of groundwater contamination. Among those petroleum hydrocarbons contained in gasoline,
* Corresponding author. Tel.: þ886 7 5254413; fax: þ886 7 5254449. E-mail address:
[email protected] (C.M. Kao). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.02.005
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 9 6 e2 5 0 6
benzene, toluene, ethylbenzene, xylene isomers (BTEX), and methyl tert-butyl ether (MTBE) are the major components of gasoline. They are the chemicals of concerns (COCs) from the human health point of view. The slow dissolution of residual BTEX and MTBE from the spill locations or from NAPL areas usually cause the migration of petroleum-hydrocarbon plume in the subsurface. This would usually threaten the downgradient water resource and ecosystem. Given that it is usually very difficult to remove the NAPLs with a short period of time, more attention has been paid in controlling the further migration of the dissolved petroleum-hydrocarbon plume in mid-plume or downgradient areas. In contrast to other fuel components, MTBE is highly water soluble (water solubility ¼ 51,000 mg/L) and has a low Henry’s Law constant (0.024e0.12 (Cair/Cwater at 25 C)) and soil adsorption coefficient [octanol water partition coefficient (Kow) ¼ 1.20 (Coctanol/Cwater at 25 C)] (ITRC, 2005a). MTBE has been identified as a potential human carcinogen by the US Environmental Protection Agency (EPA) (US EPA, 1993). Due to the high solubility, biorecalcitrance, and poor volatility of MTBE, groundwater treatment can be very challenging (US EPA, 2004; ITRC, 2005a; Pruden et al., 2005; Chen et al., 2010; Youngster et al., 2010). Benzene is ranked #6 on 2003 CERCLA Priority List of Hazardous Substances based on frequency of occurrence at EPA National Priorities List (NPL) sites, toxicity, and potential for human exposure to these substances (ATSDR, 2001; Crimi and Taylor, 2007). Both MTBE and benzene are highly recalcitrant under anoxic conditions (Ha¨ggblom et al., 2007; Foght, 2008). Due to the limitations of conventional groundwater cleanup technologies (e.g., pump-and-treat, air sparging), in situ chemical oxidation (ISCO) has become one of the attractive remedial alternatives for petroleum-hydrocarbon contaminated groundwater in recent years (Crimi and Taylor, 2007; Chen et al., 2009; Huling and Hwang, 2010). ISCO technology is an effective and potent groundwater remedial option, which is capable of breaking down many contaminants in solid or aqueous phases. During the pilot and field ISCO applications, hydrogen peroxide (Fenton’s reagent), permanganate (KMnO4), persulfate, and ozone (O3) are commonly used ISCO oxidants in these studies (Siedlecka and Stepnowski, 2007; Garoma et al., 2008; Liang et al., 2008; Chen et al., 2009; Forsey et al., 2010). Among the ISCO oxidants, persulfate is a promising new oxidant. It has a half-life of weeks to months in the subsurface (ITRC, 2005b; Osgerby, 2006). The oxidationereduction (redox) potential (ORP) of persulfate is about 2.01 V, which is higher than those of hydroxyl peroxide and permanganate but lower than those of hydroxyl radicals ($OH) and ozone (Huang et al., 2002; ITRC, 2005b). Persulfate can be activated thermally or chemically by initiators such as heat or transition metals [e.g., ferrous ion (Fe2þ)] to produce more powerful sulfate free radicals (SO 4 $) (Block et al., 2004). Other activation methods include high or low pH and hydrogen peroxide (ITRC, 2005b; Liang et al., 2007): 2 (1) S2 O2 8 þ initiator/2SO4 , or SO4 , þ SO4 2 SO 4 , þ e /SO4
E0 ¼ 2:6 V
(2)
Sulfate free radicals are capable of degrading organic compounds to carbon dioxide and water. For example,
2497
degrading MTBE and benzene by sulfate free radicals can be illustrated as reactions (3) and (4): 5Na2 S2 O8 þC5 H12 Oþ9H2 O/5CO2 þ30Hþ þ30Naþ þ30SO2 4
(3)
16Na2 S2 O8 þ C6 H6 þ 12H2 O/6CO2 þ 30Hþ 32Naþ þ 32SO2 4
(4)
Generally, ferrous ion is a commonly used initiator for persulfate activation (Liang et al., 2004): 2 3þ þ SO2 Fe2þ þ S2 O2 8 /Fe 4 þ SO4 , 1
(5)
1
k ¼ 2:0 10 M s (Travina et al., 1999). Where k ¼ bimolecular rate constant. 1
2þ 3þ þ SO2 SO 4 , þ Fe /Fe 4 1
(6)
1
k ¼ 4:6 10 M s (Kolthoff and Miller, 1951; Buxton et al., 1997). From reactions (5) and (6), overall reaction is obtained as reaction (7): 9
3þ þ 2SO2 2Fe2þ þ S2 O2 8 /2Fe 4
(7)
k ¼ 3:1 104 M1 s1 (Buxton et al., 1997). Among the passive plume control technologies, permeable reactive barriers (PRBs) have received a great deal of attention as an innovative and cost-effective method for in situ remediation of contaminated groundwater (Kao et al., 2003; Barton et al., 2004; Liu et al., 2006; Tsai et al., 2009). A wide variety of reactive materials have been developed and successfully used in PRBs for plume control and contaminant removal (Lenka et al., 2006; Ahmad et al., 2007; He et al., 2008; Saponaro et al., 2009; Yeh et al., 2010; Zijlstra et al., 2010). However, the main objective of each site remediation project might be different. In some cases, the treatment goal is to contain the plume and prevent its off-site migration. Controlled release techniques have attracted attention in diverse fields such as pharmaceutical and agrochemical technologies (Chu et al., 2007; Hu et al., 2007; Wu et al., 2007; Wu, 2008; Chansanroj and Betz, 2010; Shibata et al., 2010). Controlled release of an oxidant into a subsurface environment is an emerging concept relevant to environmental engineering. Particularly, the controlled release of oxidants during the application of ISCO merits investigation. Recent studies have focused on kinetics and efficiency of trichloroethylene (TCE) removal by KMnO4 (Kang et al., 2004; Lee and Schwartz, 2007a,b; Lee et al., 2008a,b, 2009). However, few studies on controlled (sustained) oxidant delivery systems have been conducted, particularly for persulfate (Liang et al., 2010). This study examined aspects of using oxidant-releasing materials as the active components of a well-based PRB system. In this proposed application, persulfate is emplaced as a controlled-release solid (oxidant-releasing materials) to create a chemically active zone in the subsurface. Thus, petroleum-hydrocarbon plume can be controlled when migrates into the chemically active zone, where contaminants would be oxidized by the persulfate released from the persulfate-releasing materials. The main objectives of this study were to (1) determine the components of the oxidant-releasing materials and optimal persulfate release rate of the materials; (2) assess the feasibility and effectiveness of applying the ISCO barrier system on the control of petroleum-hydrocarbon plume. In this study, MTBE and benzene, COCs in gasoline,
2498
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 9 6 e2 5 0 6
were used as the target compounds. Batch and column studies were conducted to fulfill the objectives.
2.
Materials and methods
2.1.
Materials
In this study, MTBE (99.97%, TEDIA Company Inc., Fairfield, Ohio) and benzene (99.97%, TEDIA Company Inc., Fairfield, Ohio), major components of gasoline, were selected as the COCs and target compounds. Persulfate was used as the oxidant for contaminant oxidation. Sodium persulfate (Na2S2O8, reagent grade, min. 99%, Riedel-de Haen, Germany) and potassium persulfate (K2S2O8, reagent grade, min. 99%, J.T. Baker, U.S.A.) were selected for persulfate supplement. Ferrous sulfate (FeSO4$7H2O, min. 99.5%, Riedel-de Haen, Germany) was used for ferrous ion supplement to activate persulfate oxidation. Cement (type I) and sand were used as the components of oxidant-releasing materials, which were obtained from Taiwan Cement Corp. In this study, regular medium sand with grain size of 0.4e2 mm was used to control the porosity and permeability of the mixture.
2.2.
Design of oxidant-releasing materials
Preliminary laboratory studies were conducted to determine an appropriate composition of the mixture (Kuo, 2009). Results from our preliminary study show that the appropriate components of the oxidant-releasing material included oxidant, binding material, water, and solids with small diameters to create an appropriate porosity. In the preliminary study, bentonite powder and cement were tested for their feasibility on binding properties. Results indicate that cement was more cost-effective and performed well on component binding compared to bentonite
powder. Thus, cement was used as the binding material in the following experiments. In this study, sodium persulfate (or potassium persulfate), cement, and sand were added in a 500-mL beaker, and distilled water was also gradually introduced into the beaker. Continuously mixing was performed for 15e20 min to ensure uniform distribution of those components in the mixture. Before the material was solidified, the pasty liquid was poured into a mold (4 cm 6 cm 7 cm) and it was air-dried at room temperature (25 C) for one day to form the solid oxidantreleasing material. The persulfate release rate experiments were performed in 2-L serum bottles. Each serum bottle contained 1 L distilled water and an oxidant-releasing cube. Persulfate would be released from the cube when it contacted with water. Water samples were collected from the serum bottles periodically for persulfate concentration analyses, and the distilled water in each bottle was replaced daily. The persulfate release rate was calculated and recorded as mg S2O2 8 /d/g of oxidant-releasing material versus time.
2.3. Evaluation of the released persulfate on groundwater remediation In this study, column experiment was performed to evaluate the effectiveness of using the designed persulfate-releasing materials on contaminated groundwater remediation. A total of six continuous-flow glass columns were used to simulate the ISCO barrier system. Fig. 1 shows the layout of the laboratory-scale column experiment. The developed system consisted of the first soil column followed by the oxidantreleasing material columns (the second and third columns), and the fourth to sixth columns (the second to fourth soil columns). Each column had a length of 25 cm with an inside diameter of 5 cm. The first column (soil column) was used to simulate the contaminated area located upgradient of the persulfate barrier. It was also used to equilibrate the feed
Fig. 1 e Diagram showing the layout of column experiment.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 9 6 e2 5 0 6
solution before it reached the oxidation zone. The second and third columns (Fig. 1) were used to simulate the oxidantreleasing barrier and were filled with persulfate-releasing cubes for persulfate supplement when they contacted with water. Approximately 270 g of cube-shaped oxidant-releasing materials were placed in the second and third columns (approximately 135 g in each column). The fourth to sixth columns were used to represent the downgradient area of the oxidant-releasing barrier, and they were applied to evaluate the effectiveness of released persulfate on MTBE and benzene oxidation. The produced byproducts after MTBE and benzene oxidation were also evaluated. In this system, a peristaltic pump (Cole Parmer MasterFlex L/S) was used to delivery the solution. All tubing materials in contact with solution were made of Teflon. Five pore volumes (PVs) of the test solution (site groundwater, pH ¼ 7.6) were flushed though the system at the beginning of the experiment to equilibration of the system. Researchers have reported that a minimum of 100e250 mg/L of ferrous ion was required to activate the persulfate oxidation process (Block et al., 2004; ITRC, 2005b). Thus, 100 mg/L of ferrous ion was added in the filling solution to enhance the persulfate oxidation. The MTBE and benzene solution were continuously pumped into the columns with an upflow mode under saturated conditions by the peristaltic pump. The MTBE and benzene concentrations entering the first soil column were approximately 0.2 mg/L and 0.3 mg/L, respectively. Approximately 102.8 PVs of groundwater was pumped through the six-column system at a flow rate of 0.2 mL/min. The operational characteristics of the column study are presented in Table 1.
2.4.
Performance evaluation and sample analysis
Effluents from each column were collected and analyzed for MTBE, benzene, and degradation byproducts of MTBE [e.g., tertbutyl formate (TBF), tert-butyl alcohol (TBA), actone]. Due to the limited amount of effluent volume, only effluents from the sixth column were analyzed for indicating parameters including persulfate, sulfate, ferrous ion, ORP, and pH. The soils and site groundwater used in the column experiment were sampled from a background and uncontaminated area of a petroleum-hydrocarbon spill site located in southern Taiwan. The site groundwater contained the following components at the specified concentrations (units are in mg per liter of water): alkalinity, 95.8; chloride, 69; dissolved oxygen, 0.68; ferrous ion, 0.06; nitrate, 2.8; phosphate, 2.0; sulfate, 1.7 total iron, 0.26; total organic carbon, 4.5 (Kao et al., 2008). The analytical methods were described in Standard Methods (APHA, 2005). The
Table 1 e Specifications of the column experiments. Column No. Pore volume (mL) Particle density (g/cm3) Porosity Packed height (cm) Flow rate (mL/min)
1
2e3*
4
5
6
202 1.68 0.29 25 0.2
e e e e 0.2
203 1.68 0.29 25 0.2
203 1.68 0.29 25 0.2
201 1.68 0.29 25 0.2
*Filled with oxidant-releasing materials.
2499
collected soils were air-dried, passed through a 2-mm sieve, and kept refrigerated at 4 C until analyzed. The characteristics of the soils used in the column experiment were described in Table 2. The soil analytical methods are described in Page (1982), ASTM (1998), and NIEA (2008). Persulfate concentration was determined by the colorimetric method (Huang et al., 2002). ORP (InLab 501) and pH (InLab 415) were measured using a pH/ORP meter (Metter MP120). Anion ion analysis was performed using an ion chromatography (Dionex 600) equipped with an analytical column (IonPac AS-17, 4 250 mm, Dionex). Water samples were pretreated by a purge and trap system (Tekmer 3000), and analyzed for MTBE, TBF, TBA, acetone, and benzene in accordance with US EPA Method 502.2, using a Varian 3800 Gas Chromatograph (GC) equipped with an HP-1 capillary column (Agilent Technologies, USA) and flame ionization detector (FID) (Kuo, 2009). Cations and anions were measured using PerkineElmer Plasma II Inductively Coupled Plasma-Argon Emission Spectrometer (ICP-AES). The surface of the persulfate-releasing cube was analyzed before and after the oxidation process. The cube was dried under nitrogen purged conditions at 30 C inside an oven. Thereafter, morphology and crystal structures of persulfatereleasing material were characterized by the FEI Quanta 200 environmental scanning electron microscope (ESEM). In this study, Pearson’s correlation coefficient was calculated to determine if MTBE (or benzene) and extracted groundwater were closely related. The Pearson’s correlation coefficient is a number between 1 and 1, which measures the degree to which two variables are linearly related. It is a measure of the linear association between two variables that have been measured on interval or ratio scales. This method was performed for extraction of the variation of MTBE and benzene concentration versus the effluents in column experiment.
3.
Result and discussions
3.1.
Design of oxidant-releasing materials
In this study, the oxidant-releasing material was developed and was able to release persulfate continuously. After a series
Table 2 e Characteristics of the tested soils used in the column experiment. Parameters pH Total chromate oxidizable matter (%) Total organic carbon (%) Cation exchange capacity (meq per 100 g) Iron (Fe) (mg/kg) Manganese (Mn) (mg/kg) Copper (Cu) (mg/kg) Cadmium (Cd) (mg/kg) Nickel (Ni) (mg/kg) Lead (Pb) (mg/kg) Zinc (Zn) (mg/kg) Soil texture
ND: not detectable.
Value 7.6 0.06 0.08 3.5 26.7 0.35 0.02 N.D. 0.018 ND 0.11 Sandy loam texture
2500
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 9 6 e2 5 0 6
of tests with varied mass ratios of four components (e.g., sodium persulfate, cement, sand, water), the optimal mass ratios of sodium persulfate/cement/sand/water were found to be 1/1.4/0.24/0.7. With this mass ratio, the oxidant-releasing material would remain stable and release oxidant continuously with a relatively constant rate. Thus, the mass ratio was used for the production of oxidant-releasing material for the following column study. The cube-shaped oxidant-releasing material with a density of 1.8 g/cm3 was composed of sodium persulfate (or potassium persulfate), cement, sand, and water were mixed together at a ratio of 1:1.4:0.24:0.7 by weight. Fig. 2 shows the variations in sodium persulfate and potassium persulfate release rates versus time during the 48-day operational period. The persulfate was presented as sodium persulfate concentration equivalent. Results reveal that the released rate of persulfate dropped rapidly within the first 6 days, then declined slowly for the remainder of the experimental period. The regression lines (from day 8 to day 48) for the released persulfate concentrations versus time are also presented in Fig. 2. Sodium and potassium persulfate concentrations were initially high at 30.79 and 5.49 mg persulfate/d/g material on day 1, but rapidly decreased to 9.74 and 1.48 mg persulfate/d/g material on day 6. Sodium and potassium persulfate concentration then gradually decreased to 6.11 and 0.45 mg persulfate/d/g material on day 48 with a fluctuation range of 1e3 mg persulfate/d/g material. Based on the calculated slopes of the regression lines, the regression lines were relatively flat, and thus, the release rates after day 8 were relatively constant. The calculated slopes for sodium persulfate and potassium persulfate-releasing lines were 2.37 103 and 2.6 102 mg persulfate/g of material, respectively. Results indicate that the designed oxidantreleasing material would be able to release persulfate for a long period of time. Therefore, the stable released rates of sodium persulfate and potassium persulfate were obtained using the mean values of the two regression lines for the period from day 8 to day 48, when the release rate reached 9.74 and 1.48 to 6.11 and 0.45 mg persulfate/d/g material, respectively (Fig. 2). Thus, the calculated mean sodium persulfate and potassium persulfate release rate was 7.26 and 1.14 mg persulfate/d/g. Results from
released persulfate mass (mg of persulfate/day/g material)
100
sodium persulfate Y = 7.31 - 2.37 x 10-3 X potassium persulfate -2 X
10
1
0.1 0
10
20
30 time (day)
40
50
Fig. 2 e Variation in sodium and potassium persulfate release rate during the 48-day operation period.
60
the mass balance calculation indicate that a total of 41% and 28% of sodium persulfate and potassium persulfate had released from the persulfate-releasing materials, respectively, after 48 days of operation. This might be due to the fact that sodium persulfate has higher solubility than potassium persulfate (ITRC, 2005b). Although more persulfate remained in the potassium persulfate-releasing materials, lower release rate was observed. Because the oxidation efficiency of MTBE and benzene would be affected by the persulfate concentrations, less persulfate release rate would cause less oxidation efficiency. Thus, sodium type persulfate-releasing material was selected for the following column experiment. Technically, release rates and life-time of the oxidant-releasing material can be adjusted by changing the component mass ratio or dimensions of the pellets. Thus, the oxidant-releasing barrier scheme could potentially be developed as a practical approach for the in situ remediation of contaminated aquifers.
3.2. Evaluation of the released persulfate on groundwater remediation In the second part of this study, column experiments were performed to evaluate the in situ oxidant-releasing barrier system on benzene and MTBE contaminated groundwater remediation. Fig. 3ed presents the variations in ferrous ion, pH, ORP, persulfate and sulfate in the effluent samples from the sixth column. Results show that persulfate and ferrous ion concentrations varied from 337 to 14,545 mg/L and 0 to 0.15 mg/L when 0e8.29 PVs of groundwater were pumped through the column system, respectively. The persulfate released from the oxidant-releasing materials could be activated by the introduced ferrous ions and produce sulfate radicals subsequently as reported in Eq. (5). The produced sulfate radicals would oxidize contaminants (MTBE and benzene) in a very efficient way. The observed persulfate and ferrous ion concentrations in the sixth column effluents varied from 14,545 to 1000 mg/L and 0.15 to 77.5 mg/L, respectively, when 8.29e102.8 PVs of groundwater was pumped through the six-column system. The decreased persulfate concentrations indicate that persulfate release pattern reached the stabilized stage with a more constant release rate after approximately 94.51 PVs of groundwater pumping. Researchers reported that excessive ferrous ion might act as an intrinsic scavenger of sulfate radicals (Eq. (6)) (Kolthoff and Miller, 1951; Buxton et al., 1997). This would cause the consumption of sulfate radicals. Therefore, ferrous ion concentrations should be controlled to prevent the reaction between ferrous ion and sulfate radical (Chen et al., 2009; Oh et al., 2009). Results from Liang et al. (2004) reveal that even though ferrous ion activated persulfate can quickly generate sulfate radicals for the destruction of trichloroethylene, incomplete trichloroethylene destruction would occur almost instantaneously and then reaction stalled. Thus, the destruction of sulfate radical could be due to (1) the presence of excess Fe2þ, and (2) rapid conversion of Fe2þ to Fe3þ, which limits the oxidizing capability. According to the literatures and results from our study (Block et al., 2004; Liang et al., 2004; ITRC, 2005b; Chen et al., 2009), the appropriate ferrous ion concentration for this system is approximately 100e250 mg/L.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 9 6 e2 5 0 6
2501
Fig. 3 e Variations of in (a) ferrous ion, (b) pH, (c) ORP, (d) persulfate and sulfate in the effluents from the sixth column.
If ferrous ion concentration is higher than 250 mg/L, then ferrous ion would compete sulfate radicals with the contaminants (Liang et al., 2008; Tsai et al., 2009). Chen et al. (2009) reported that MTBE could be degraded effectively when appropriate amount of ferrous ion was added (molar ratios of 2þ between 1/0.031 and 1/0.31). Thus, ferrous ion S2O2 8 /Fe concentrations should be controlled to minimize the adverse effects of ferrous ion on sulfate radical production. Results from ORP measurements show that ORP values decreased dramatically from 718 to 604 mV after pumping with 0.39e72.7 PVs of groundwater (Fig. 3b). This could be attributed to the production of sulfate radicals (Chen et al., 2009). The observed persulfate concentrations and ORP values varied from 14,545 to 404 mg/L and 771 to 490 mV after pumping with 8.29e102.8 PVs of groundwater, respectively. Results indicate that ORP measurement can be used as a quick indicating parameter for persulfate and sulfate radical in the subsurface during the ISCO process using persulfate as the oxidant. The variations in pH measurements versus flushed groundwater for the sixth column effluents are shown in Fig. 3c. The pH values varied from 7.24 to 1.63 during the pumping period from 0.39 to 15.7 PVs. This demonstrates that the in situ persulfate oxidation process would cause the decrease in groundwater pH because of the production of sulfate and sulfuric acid shown in Eq. (5). In the future practical application, pH would not become an issue in the field due to the intrinsic groundwater buffering system (ITRC, 2005b). In the latter part of the operational period (15.7e102.8 PVs of pumping), increase in pH values from 1.63 to 5.6 was observed. This could be attributed to the decreased production of sulfuric acid in the latter operational period. The variations in sulfate concentrations versus PV in the sixth column effluents are presented in Fig. 3d. Results show that persulfate and sulfate concentration varied from 0 to 14,545 mg/L and 145 to 10,530 mg/L during the pumping period from 0.39 to 8.29 PVs, respectively. The observed persulfate and sulfate concentrations respectively varied from 14,545 to 400 mg/L and 10,530 to 569 mg/L during the pumping period from 8.29 to 102.8 PVs, respectively.
Results from Fig. 3d presents a trend of the persulfate release from the oxidant-releasing materials and the decrease in persulfate and sulfate concentrations reveals that the material reached the latter part of the release pattern. The concentrations of MTBE, benzene, and their degradation byproducts versus PV for the effluents from the first to sixth columns were shown in Fig. 4. Insignificant removal of MTBE and benzene was observed from the first column effluents, indicating that the intrinsic biodegradability of MTBE and benzene was low under anoxic conditions. This matched with the results reported by Kao et al. (2008) and Chen et al. (2010) that MTBE and benzene had lower natural biodegradation rates compared to toluene, ethylbenzene, and xylenes. Approximately 85.0% of MTBE and 98.8% of benzene were removed during the operational periods from 0.39 to 37.9 PVs, and 20.0% of MTBE and 60.0% of benzene were removed during the operational periods from 37.9 to 102.8 PVs from the third column effluents. The significant amounts of MTBE and benzene removal indicate that the released persulfate was able to oxidize MTBE and benzene. Results also show that higher contaminant removal efficiency was observed especially during the early stage of the operational period (from 0.39 to 37.9 PVs). It should be noted that although persulfate can be activated by ferrous ion, ferrous ion would also consume sulfate free radicals. Thus, ferrous ion concentrations should be controlled to minimize the adverse effects of ferrous ion on sulfate radical application. The scavenging reaction between ferrous ion and sulfate radicals is shown in Eq. (6). In addition, ferrous ion can also decompose persulfate anion according to Eq. (5). Thus, excess addition of ferrous ion may reduce the efficiency of contaminant oxidation (ITRC, 2005b; US EPA, 2006). Chen et al. (2009) reported that MTBE degradation rates were positively related to ferrous ion concentrations within the molar ratio of 2þ between 1/0.03 and 1/0.3. However, MTBE oxidation S2O28 /Fe 2þ molar of 1/1 due rate decreased significantly with a S2O28 /Fe to the competition for sulfate free radicals between ferrous ion and MTBE. The authors concluded that the molar ratio of S2O28 /
2502
1
0.25 0.20 0.15 0.10 0.05 0.00
Concentration (mg/L)
0
40 60 80 pore volume
0.30
4
0.20 0.15 0.10 0.05 0.00 20
40 60 80 pore volume
0.30
0.30
100
2-3
0.25 0.20 0.15 0.10 0.05 0.00
100
0.25
0 Concentration (mg/L)
20
Concentration (mg/L)
0.30
0
Concentration (mg/L)
Concentration (mg/L)
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 9 6 e2 5 0 6
20
40 60 80 pore volume
0.30
100
5
0.25 0.20 0.15 0.10 0.05 0.00 0
20
40 60 80 pore volume
100
6
0.25 0.20
MTBE benzene actone
0.15 0.10 0.05 0.00 0
20
40 60 80 pore volumn
100
Fig. 4 e Concentrations of MTBE, benzene, and acetone in the first to the sixth column effluents.
Fe2þ in a persulfate oxidation system is the most important controlling factor to achieve effective MTBE removal (Chen et al., 2009). Liang et al. (2004) demonstrated that TCE degradation increased with higher ferrous ion concentration with 2þ between 1/0.25 and 1/0.75. No the molar ratio of S2O28 /Fe further TCE degradation was observed with the molar ratios of 2þ beyond 1/0.75 (Liang et al., 2004). A similar result S2O28 /Fe was obtained in a polychlorinated biphenyls (PCBs) treatment study with PMS/Fe2þ molar ratio of 1/1 (Rastogim et al., 2009). Significant radical scavenging was observed at higher ferrous ion concentrations resulting in the less removal of PCBs. 2þ was not deterAlthough the optimal molar ratio of S2O28 /Fe mined in their study, the results reveal that the applied ferrous ion concentrations were able to enhance the contaminant removal (Rastogim et al., 2009). Kong et al. (1998) reported that the naturally-occurring iron minerals (e.g., goethite, magnetite), were able to catalyze oxidants and initiate oxidation reaction of silica sand contaminated with petroleum hydrocarbons in batch systems. Tsai et al. (2009) demonstrated that iron-contained basic oxygen furnace slag can be applied for iron supplement to activate the persulfate oxidation mechanism, and cause the decrease in chlorinated hydrocarbon concentration via oxidation process. For the field application, the iron-contained soils might be able to supply iron catalyst to activate the persulfate oxidation
process. To further enhance the oxidation efficiency, other substitutes (e.g., iron-contained intrinsic soils or minerals, basic oxygen furnace slag) can be applied for the supplement of ferrous ions in the practical application. Results from the fourth to sixth column effluents show that approximately 86e92% of MTBE and 95e99% of benzene were removed in effluents from the fourth to sixth columns from 0.39 to 37.9 PVs. The removal efficiencies for MTBE and benzene were approximately 40.1e55.5% and 85.4e93.0%, respectively, during the operational periods from 22.9 to 43.5 PVs. This also reveals that the released persulfate from the oxidant-releasing materials were able to oxidize MTBE and benzene in the downgradient area of the barrier (soil columns). The oxidant-releasing barrier can be used for the control of petroleum-hydrocarbon plume. After the consumption of the persulfate in the oxidant-releasing materials, the decrease in MTBE and benzene oxidation efficiency was observed, and thus, increase in MTBE and benzene concentrations was also observed. The results show that the oxidation efficiency of MTBE and benzene corresponded with the persulfate concentrations. Higher persulfate concentrations would cause higher MTBE and benzene removal efficiency. Replacement of the oxidant-releasing material is required after flushing 102.8 PVs of groundwater to maintain a desired persulfate release rate and contaminant oxidation
2503
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 9 6 e2 5 0 6
efficiency. Researchers reported that the oxidation power of persulfate could be enhanced by ferrous ion (Liang et al., 2004, 2008; Killian et al., 2007; Chen et al., 2009). Results from this study also show that the introduced ferrous ion could activate the persulfate oxidation process and cause the removal of MTBE and benzene. In this column experiment, effluents from the columns were analyzed for TBF, TBA, and acetone, degradation byproducts of MTBE (Fig. 4). Although decrease in MTBE concentrations was observed, no significant TBF and TBA concentrations were detected [less than detection limit (8 mg/L)] (data not shown). This could be due to the fact that both TBA and TBF were produced and then rapidly degraded as the reaction proceeded. Although TBF and TBA were not detected from the effluents, significant acetone concentrations were observed. This indicates that acetone is a more persistent compound during the MTBE oxidation process, and it is also an oxidation byproduct of TBF and TBA. Liang and Su (2009) also reported that acetone is produced during the TBF and TBA degradation process. Results indicate that sulfate and hydroxyl radicals are the main cause of MTBE and benzene degradation (Eq. (5) and (8)). 2 þ SO 4 , þ H2 O/SO4 þ HO, þ H
(8)
It is generally accepted that the first step of MTBE and benzene degradation is the abstraction of an alphahydrogen by a hydroxyl radical (and/or a sulfate radical in persulfate systems) (Kormann et al., 1988; Barreto et al., 1995). The first step leads to a carbon centered radical, which then reacts with oxygen to form a peroxyl radical that may undergo acidcatalyzed hydrolysis reaction to yield TBF (Kormann et al., 1988; Barreto et al., 1995). It was reported that in various aqueous systems, TBF hydrolysis yielded stoichiometric amounts of TBA (Church et al., 1999). In addition, acetone and methyl acetate in addition to aldehydes and carboxylic acids have been found as intermediate products prior to MTBE mineralization with H2O2/UV oxidation (Stefan et al., 2000). Similar degradation patterns were observed by other researchers (Huang et al., 2002; Xu et al., 2004; Chen et al., 2009; Hwang et al., 2010). Measured ferrous ion concentrations in the effluents from the third to sixth columns varied from 0 to 5.88 mg/L. This indicates that all of the ferrous ion added in the filling solution had been oxidized, and thus, the observed oxidation efficiencies for MTBE and benzene via the ferrous ion-activated persulfate oxidation were approximately 86e92% and 95e99%, respectively, in the early stage (0e37.9 PVs) of the oxidation process. Table 3 presents the correlation coefficient matrix between MTBE and PV variables in the column experiment. Table 4
Table 3 e Correlation coefficient matrix between MTBE and PV variables in the column experiment. MTBE 1 2e3 4 5 6
presents the correlation coefficient matrix between benzene and PV variables in the column experiment. The correlation matrix containing calculated Pearson’s correlation coefficients was utilized to demonstrate relationships between variables. The obtained matrixes of MTBE and benzene concentration were subjected to multivariance analytical technique (Tables 3 and 4). There was no significant correlation between the first column effluent and the other column effluents (less than 0.05). The correlation matrix between chemical variables indicate that there was the highest correlation between the fourth and fifth column effluents (with correlation coefficient higher than 1.0) in the column experiment. The strongest correlation with MTBE and benzene degradation was obtained due to the persulfate release from the oxidant-releasing materials. To observe the surface change of the sodium type oxidantreleasing materials, ESEM technique was conducted. Fig. 5 shows the ESEM images of the oxidant-releasing material surface for the original material (before experiment) [Fig. 5a-1 (400) and Fig. 5a-2 (1000)] and residual material (after experiment) [Fig. 5b-1 (400) and Fig. 5b-2 (1000)]. ESEM results show that the sodium persulfate granules appeared in the material porosity after the experiment (Fig. 5b-1 and b-2). This indicates that sodium persulfate released from the material after contacted with water. The released persulfate would cause the contaminant oxidation in the water phase. The ESEM results verified the formation of micro-scale secondary capillary permeability through which persulfate can be released by a reaction-diffusion. Based on the results from the column study, contaminant concentration plays an important role in determining the dose of ferrous ion. Because the groundwater flow affects the contaminant concentration, groundwater velocity and groundwater flow rate need to be considered for a practical application. Because reaction time (residence time) is determined by groundwater velocity, reaction time also affects the contaminant oxidation efficiency. Chen et al. (2009) shows that a complete degradation of MTBE and its degradation byproducts was not observed with low concentration of 2þ ¼ 1/50/31 and 1/10/1.55) or high oxidant (MTBE/S2O2 8 /Fe 2þ ¼ 1/100/31). concentration of ferrous ion (MTBE/S2O2 8 /Fe This indicates that higher oxidant concentration and proper ferrous ion addition are necessary to achieve complete removal of MTBE and its degradation byproducts. Moreover, 2þ molar results of MTBE degradation from MTBE/S2O2 8 /Fe ratio of 1/10/1.55 also suggest that persulfate was able to sustain the oxidation power in the system more than 43 days because drops of MTBE and its degradation byproduct concentrations after 43 days of operation. Thus, increased
Table 4 e Correlation coefficient matrix between benzene and PV variables in the column experiment.
1
2e3
4
5
6
Benzene
1 0.30 0.30 0.28 0.05
e 1 0.84 0.83 0.74
e e 1 1.00 0.79
e e e 1 0.82
e e e e 1
1 2e3 4 5 6
1
2e3
4
5
6
1 0.41 0.56 0.38 0.25
e 1 0.77 0.55 0.73
e e 1 0.79 0.68
e e e 1 0.90
e e e e 1
2504
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 9 6 e2 5 0 6
Fig. 5 e ESEM images of the oxidant-releasing material surface for the original material (before experiment) [Fig. 5a-1 (4003) and Fig. 5a-2 (10003)] and residual material (after experiment) [Fig. 5b-1 (4003) and Fig. 5b-2 (10003)].
reaction time for contaminants and persulfate corresponded with increased contaminants removal efficiencies. In most petroleum-hydrocarbon contaminated aquifers, released free or residual products formed NAPLs would cause a long-term source of subsurface pollution. Because it is very difficult to completely remove NAPLs from the subsurface, effective control of the continuously release of petroleum hydrocarbons from NAPLs and minimizing their migration to downgradient regions are top priorities. The developed slowrelease and passive oxidant-releasing barrier can be installed downgradient of the source zone (or NAPL zone) of the contamination, and thus, the petroleum hydrocarbons (e.g., MTBE, benzene) migrated from the source zone would be degraded when contact with the released oxidants through the oxidation process. When this system is applied in the field, some chemical and hydrological/site factors need to be considered so the effective contaminant oxidation efficiency can be obtained. The major chemical factors include the following: contaminant of concern, contaminant concentration, groundwater pH, and amounts of applied oxidant-releasing materials and ferrous ion. The major hydrological/site factors include the following: groundwater flow velocity, groundwater flow rate, and distribution of the contaminant plume. With appropriate design, the developed passive treatment scheme has the potential to be developed into an economically acceptable
remediation technology for the in situ treatment of contaminated aquifer.
4.
Conclusions
In this study, oxidant-releasing materials were designed to continuously release persulfate passively for MTBE and benzene oxidation. The components and persulfate release rates were determined in batch experiments, and the effectiveness of the released persulfate on MTBE and benzene removal was evaluated in the column study. In the batch experiments, oxidant-releasing materials were prepared from a mixture of sodium persulfate (or potassium persulfate), cement, sand, and water. The persulfate release rates from sodium persulfate-contained materials and potassium persulfate-contained materials were 7.26 and 1.14 mg persulfate/d/g material, respectively. Thus, sodium type persulfate-releasing materials were applied for the following column experiments. Results from the column experiments indicate that the designed oxidant-releasing materials could release persulfate passively for contaminant oxidation during the 102.8 PVs operational period. Persulfate release and oxidation of MTBE and benzene in the column study can be verified by the following findings observed in effluents after the oxidantreleasing columns: (1) decreased MTBE and benzene
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 9 6 e2 5 0 6
concentrations in the system, (2) increased ORP values, (3) increased sulfate and persulfate concentrations, (4) increased acetone (MTBE degradation byproduct) concentrations. Results indicate that the degradation of MTBE and benzene were mainly due to the persulfate oxidation reactions. Thus, the proposed ISCO barrier system containing oxidant-releasing materials has the potential to become a cost-effective remedial alternative to remediate petroleum-hydrocarbon contaminated groundwater.
Acknowledgments This study was funded by National Science Council in Taiwan. Additional thanks are extended to the personnel at Taiwan Environmental Protection Administration for their assistance and support throughout this project.
references
APHA, (American Public Health Association), 2005. Standard Methods for the Examination of Water and Wastewater, 21st ed. American Water Works Association & Water Environment Federation, Washington, D.C., USA. Ahmad, F., Schnitker, S.P., Newell, C.J., 2007. Remediation of RDXand HMX-contaminated groundwater using organic mulch permeable reactive barriers. Journal of Contaminant Hydrology 90, 1e20. ASTM-International-D422-63, 1998. Standard Test Method for Particle Size Analysis of Soils. ATSDR, 2001. Interaction Profile for Benzene, Toluene, Ethylbenzene and Xylenes (BTEX). Agency for Toxic Substances and Disease Registry (ATSDR). http://www.atsdr.cdc.gov. Barreto, R.D., Gray, K.A., Anders, K., 1995. Photocatalytic degradation of methyl tert-butyl ether in TiO2 slurries: a proposed reaction scheme. Water Research 29 (5), 1243e1248. Barton, C.S., Stewart, D.I., Morris, K., Bryant, D.E., 2004. Performance of three resin-based materials for treating uranium-contaminated groundwater within a PRB. Journal of Hazardous Materials 116 (3), 191e204. Block, P.A., Brown, R.A., Robinson, D., 2004. “Novel activation technologies for sodium persulfate in situ chemical oxidation”. In: Proceedings of the 4th International Conference on the Remediation of Chlorinated and Recalcitrant Compounds. Buxton, G.V., Malone, T.N., Salmon, G.A., 1997. Reaction of SO 4$ with Fe2þ, Mn2þ and Cu2þ in aqueous solution. Journal of the Chemical Society, Faraday Transactions 93 (16), 2893e2897. Chansanroj, K., Betz, G., 2010. Sucrose esters with various hydrophilicelipophilic properties: novel controlled release agents for oral drug delivery matrix tablets prepared by direct compaction. Acta Biomaterialia 6 (8), 3101e3109. Chen, K.F., Kao, C.M., Wu, L.C., Surampalli, R.Y., Liang, S.H., 2009. Methyl tert-butyl ether (MTBE) degradation by ferrous ionactivated persulfate oxidation: feasibility and kinetics studies. Water Environment Research 81 (7), 687e694. Chen, K.F., Kao, C.M., Chen, C.W., Surampalli, R.Y., Lee, M.S., 2010. Control of petroleum-hydrocarbon contaminated groundwater by intrinsic and enhanced bioremediation. Journal of Environmental Sciences-China 22 (6), 864e871. Chu, L.Y., Kim, J.W., Shah, R.K., Weitz, D.A., 2007. Monodisperse thermoresponsive microgels with tunable volume-phase
2505
transition kinetics. Advanced Functional Materials 17 (17), 3499e3504. Church, C.D., Pankow, J.F., Tratnyek, P.G., 1999. Hydrolysis of tertbutyl formate: kinetics, products, and implications for the environmental impact of methyl tert-butyl ether. Environmental Toxicology and Chemistry 18 (12), 2789e2796. Crimi, M.L., Taylor, J., 2007. Experimental evaluation of catalyzed hydrogen peroxide and sodium persulfate for destruction of BTEX contaminants. Soil and Sediment Contamination 16 (1), 29e45. Environmental Protection Agency, 2004. Technologies for Treating MTBE and Other Fuel Oxygenates, EPA 542-R-04e009. U.S. Environmental Protection Agency, Washington, D.C., U.S. Environmental Protection Agency, 2006. In-Situ Chemical Oxidation e Engineering Issue, EPA-600-R-06-072. US Environmental Protection Agency, Washington, DC. Foght, J., 2008. Anaerobic biodegradation of aromatic hydrocarbons: pathways and prospects. Journal of Molecular Microbiology and Biotechnology 15 (2e3), 93e120. Forsey, S.P., Thomson, N.R., Barker, J.F., 2010. Oxidation kinetics of polycyclic aromatic hydrocarbons by permanganate. Chemosphere 79 (6), 628e635. Garoma, T., Gurol, M.D., Osibodu, O., Thotakura, L., 2008. Treatment of groundwater contaminated with gasoline components by an ozone/UV process. Chemosphere 73 (0), 825e831. Ha¨ggblom, M.M., Youngster, L.K.G., Somsamak, P., Richnow, H.H., 2007. Anaerobic biodegradation of methyl tert-butyl ether (MTBE) and related fuel oxygenates. Advances in Applied Microbiology 62, 1e20. He, Y.T., Wilson, J.T., Wilkin, R.T., 2008. Transformation of reactive iron minerals in a permeable reactive barrier (biowall) used to treat TCE in groundwater. Environmental Science and Technology 42 (17), 6690e6696. Hu, L., Chu, L.Y., Yang, M., Yu, J., Wang, H.D., 2007. A composite thermo-responsive membrane system for improved controlled-release. Chemical Engineering and Technology 30 (4), 523e529. Huang, K.C., Couttenye, R.A., Hoag, G.E., 2002. Kinetics of heatassisted persulfate oxidation of methyl tert-butyl ether (MTBE). Chemosphere 49, 413e420. Huling, S.G., Hwang, S., 2010. Iron amendment and Fenton oxidation of MTBE-spent granular activated carbon. Water Research 44 (8), 2663e2671. Hwang, S., Huling, S.G., Ko, S., 2010. Fenton-like degradation of MTBE: effects of iron counter anion and radical scavengers. Chemosphere 78 (5), 563e568. ITRC (The Interstate Technology & Regulatory Council), 2005a. Overview of Groundwater Remediation Technologies for MTBE and TBA. ITRC (The Interstate Technology & Regulatory Council), 2005b. Technical and Regulatory Guidance for In Situ Chemical Oxidation, second ed. Kang, N., Hua, I., Rao, P.S.C., 2004. Production and characterization of encapsulated potassium permanganate for sustained release as an in situ oxidant. Industrial and Engineering Chemistry Research 43 (17), 5187e5193. Kao, C.M., Chen, S.C., Wang, J.Y., Chen, Y.L., Lee, S.Z., 2003. Remediation of PCE-contaminated aquifer by an in situ twolayer biobarrier: laboratory batch and column studies. Water Research 37 (1), 27e38. Kao, C.M., Chen, C.Y., Chen, S.C., Chien, H.Y., Chen, Y.L., 2008. Application of in situ biosparging to remediate a petroleumhydrocarbon spill site: field and microbial evaluation. Chemosphere 70 (8), 1492e1499. Killian, P.F., Bruell, C.J., Liang, C.J., Marley, M.C., 2007. Iron (II) activated persulfate oxidation of MGP contaminated soil. Soil and Sediment Contamination 16 (6), 523e537.
2506
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 4 9 6 e2 5 0 6
Kolthoff, I.M., Miller, I.K., 1951. The chemistry of persulfate. I. The kinetics and mechanism of the decomposition of the persulfate ion in aqueous medium. Journal of the American Chemical Society 73 (7), 3055e3059. Kong, S.H., Watts, R.J., Choi, J.H., 1998. Treatment of petroleumcontaminated soils using iron mineral catalyzed hydrogen peroxide. Chemosphere 37 (8), 1473e1482. Kormann, C., Bahnemann, D.W., Hoffman, M.R., 1988. Photocatalytic production of H2O2 and organic peroxides in aqueous suspensions of TiO2, ZnO and desert sand. Environmental Science and Technology 22 (7), 798e806. Kuo, Y.C., 2009. Remediation of petroleum-hydrocarbon contaminated groundwater using oxidant-releasing material. Thesis of Master of Science, National Sun Yat-sen University, Kaohsiung, Taiwan. Lee, E.S., Schwartz, F.W., 2007a. Characteristics and applications of controlled-release KMnO4 for groundwater remediation. Chemosphere 66 (11), 745e750. Lee, E.S., Schwartz, F.W., 2007b. Characterization and optimization of long-term controlled release system for groundwater remediation: a generalized modeling approach. Chemosphere 69 (2), 247e253. Lee, E.S., Woo, N.C., Schwartz, F.W., Lee, B.S., Lee, K.C., Woo, M.H., Kim, J.H., Kim, H.K., 2008a. Characterization of controlledrelease KMnO4 (CRP) barrier system for groundwater remediation: a pilot-scale flow-tank study. Chemosphere 71 (5), 902e910. Lee, E.S., Liu, G.M., Schwartz, F.W., Kim, Y.J., Ibaraki, M., 2008b. Model-based evaluation of controlled-release systems in the remediation of dissolved plumes in groundwater. Chemosphere 72 (2), 165e173. Lee, B.S., Kim, J.H., Lee, K.C., Bin Kim, Y., Schwartz, F.W., Lee, E.S., Woo, N.C., Lee, M.K., 2009. Efficacy of controlled-release KMnO4 (CRP) for controlling dissolved TCE plume in groundwater: a large flow-tank study. Chemosphere 74 (6), 745e750. Lenka, V., Jan, N., Martina, S., Martin, K., 2006. The biofiltration permeable reactive barrier: practical experience from synthesia. International Biodeterioration & Biodegradation 58 (3e4), 224e230. Liang, C.J., Su, H.W., 2009. Identification of sulfate and hydroxyl radicals in thermally activated persulfate. Industrial and Engineering Chemistry Research 48 (11), 5558e5562. Liang, C.J., Bruell, C.J., Marley, M.C., Sperry, K.L., 2004. Persulfate oxidation for in situ remediation of TCE. I. Activated by ferrous ion with and without a persulfateethiosulfate redox couple. Chemosphere 55, 1213e1223. Liang, C.J., Wang, Z.S., Bruell, C.J., 2007. Influence of pH on persulfate oxidation of TCE at ambient temperatures. Chemosphere 66 (1), 106e113. Liang, C.J., Huang, C.F., Chen, Y.J., 2008. Potential for activated persulfate degradation of BTEX contamination. Water Research 42 (15), 4091e4100. Liang, S.H., Kao, C.M., Kuo, Y.C., Chen, K.F., Surampalli, R.Y., 2010. Application of persulfate-releasing material to treat MTBEcontaminated aquifer. In: 2010 International Advanced Oxidation Processes Conference. Tunghai University, Taiwan. Liu, S.J., Jiang, B., Huang, G.Q., Li, X.G., 2006. Laboratory column study for remediation of MTBE-contaminated groundwater using a biological two-layer permeable barrier. Water Research 40 (18), 3401e3408. NIEA, National Institute of Environmental Analysis, 2008. Environmental Analysis Laboratory EPA. Executive Yuan, ROC. http://www.niea.gov.tw/. Oh, S.Y., Kim, H.W., Park, J.M., Park, H.S., Yoon, C., 2009. Oxidation of polyvinyl alcohol by persulfate activated with heat, Fe2þ, and zero-valent iron. Journal of Hazardous Materials 168 (1), 346e351.
Osgerby, I.T., 2006. ISCO technology overview: do you really understand the chemistry? In: Calabrese, E.J., Kostecki, P.T., Dragun, J. (Eds.), Contaminated Soils Sediments and Water, vol. 10 Springer, New York, NY, USA, pp. 287e308. Page, A.L., 1982. Methods of Soil Analysis e Part 2 Chemical and Microbiological Properties, second ed. American Society of Agronomy, Inc., Soil Science Society of America, Inc., Madison, WI. Pruden, A., Sedran, M.A., Suidan, M.T., Venosa, A.D., 2005. Anaerobic biodegradation of methyl tert-butyl ether under iron-reducing conditions in batch and continuous-flow cultures. Water Environment Research 77 (3), 297e303. Rastogim, A., Ai-Abed, S.R., Dionysiou, D.D., 2009. Sulfate radicalbased ferrous-peroxymonosulfate oxidative system for PCBs degradation in aqueous and sediment systems. Applied Catalysis B-Environmental 85 (3e4), 171e179. Saponaro, S., Negri, M., Sezenna, E., Bonomo, L., Sorlini, C., 2009. Groundwater remediation by an in situ biobarrier: a bench scale feasibility test for methyl tert-butyl ether and other gasoline compounds. Journal of Hazardous Materials 167 (1e3), 545e552. Shibata, N., Nishumura, A., Naruhashi, K., Nakao, Y., Miura, R., 2010. Preparation and pharmaceutical evaluation of new sustained-release capsule including starch-sponge matrix (SSM). Biomedicine and Pharmacotherapy 64 (5), 352e358. Siedlecka, E.M., Stepnowski, P., 2007. Effect of chlorides and sulfates on the performance of a Fe3þ/H2O2 Fenton-like system in the degradation of methyl tert-butyl ether and its byproducts. Water Environment Research 79 (11), 2318e2324. Stefan, M.I., Mack, J., Bolton, J.R., 2000. Degradation pathways during the treatment of methyl tert-butyl ether by the UV/ H2O2 process. Environmental Science and Technology 34 (4), 650e658. Travina, O.A., Kozlov, Y.N., Purmal’, A.P., Rod’ko, I.Y., 1999. Synergism of the action of the sulfite oxidation initiators, iron and peroxydisulfate ions. Russian Journal of Physical Chemistry 73 (8), 1215e1219. Tsai, T.T., Kao, C.M., Hong, A., 2009. Treatment of tetrachloroethylene-contaminated groundwater by surfactant-enhanced persulfate/BOF slag oxidation e a laboratory feasibility study. Journal of Hazardous Materials 171 (1e3), 571e576. U.S. Environmental Protection Agency, 1993. “Assessment of Potential Health Risks of Gasoline Oxygenated with Methyl Tertiary Butyl Ether (MTBE)” EPA600-R-93-206. US Environmental Protection Agency, Washington, DC. U.S. Wu, C.S., 2008. Characterizing biodegradation of PLA and PLA-gAA/starch films using a phosphate-solubilizing Bacillus species. Macromolecular Bioscience 8 (6), 560e567. Wu, K.J., Wu, C.S., Chang, J.S., 2007. Biodegradability and mechanical properties of polycaprolactone composites encapsulating phosphate-solubilizing bacterium Bacillus sp. PG01. Process Biochemistry 42 (4), 669e675. Xu, X.R., Li, H.B., Wang, W.H., Gu, J.D., 2004. Degradation of dyes in aqueous solutions by the Fenton process”. Chemosphere 57 (7), 595e600. Yeh, C.H., Lin, C.W., Wu, C.H., 2010. A permeable reactive barrier for the bioremediation of BTEX-contaminated groundwater: microbial community distribution and removal efficiencies. Journal of Hazardous Materials 178 (1e3), 74e80. Youngster, L.K.G., Kerkhof, L.J., Ha¨ggblom, M.M., 2010. Community characterization of anaerobic methyl tert-butyl ether (MTBE)-degrading enrichment cultures. FEMS Microbiology Ecology 72 (2), 279e288. Zijlstra, J.J.P., Dessi, R., Peretti, R., Zucca, A., 2010. Treatment of percolate from metal sulfide mine tailings with a permeable reactive barrier of transformed red mud. Water Environment Research 82 (4), 319e327.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 0 7 e2 5 1 6
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Simultaneous degradation of disinfection byproducts and earthy-musty odorants by the UV/H2O2 advanced oxidation process Chang Hyun Jo a,1, Andrea M. Dietrich a,*, James M. Tanko b a b
Department of Civil and Environmental Engineering, Virginia Polytechnic Institute and State University, VA 24061, USA Department of Chemistry, Virginia Polytechnic Institute and State University, VA 24061, USA
article info
abstract
Article history:
Advanced treatment technologies that control multiple contaminants are beneficial to
Received 27 September 2010
drinking water treatment. This research applied UV/H2O2 for the simultaneous degradation
Received in revised form
of geosmin, 2-methylisoborneol, four trihalomethanes and six haloacetic acids. Experi-
3 February 2011
ments were conducted in de-ionized water at 24 1.0 C with ng/L amounts of odorants
Accepted 4 February 2011
and mg/L amounts of disinfection byproducts. UV was applied with and without 6 mg/L
Available online 12 February 2011
H2O2. The results demonstrated that brominated trihalomethanes and brominated haloacetic acids were degraded to a greater extent than geosmin and 2-methylisoborneol.
Keywords:
Tribromomethane and dibromochloromethane were degraded by 99% and 80% respec-
Hydroxyl radical
tively at the UV dose of 1200 mJ/cm2 with 6 mg/L H2O2, whereas 90% of the geosmin and
Ultraviolet
60% of the 2-methylisoborneol were removed. Tribromoacetic acid and dibromoacetic acid
Geosmin
were degraded by 99% and 80% respectively under the same conditions. Concentrations of
2-Methylsioborneol
trichloromethane and chlorinated haloacetic acids were not substantially reduced under
Trihalomethanes
these conditions and were not effectively removed at doses designed to remove geosmin
Haloacetic acids
and 2-methylisoborneol. Brominated compounds were degraded primarily by direct photolysis and cleavage of the CeBr bond with pseudo first order rate constants ranging from 103 to 102 s1. Geosmin and 2-methylisoborneol were primarily degraded by reaction with hydroxyl radical with direct photolysis as a minor factor. Perchlorinated disinfection byproducts were degraded by reaction with hydroxyl radicals. These results indicate that the UV/H2O2 can be applied to effectively control both odorants and brominated disinfection byproducts. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
Advanced oxidation processes (AOP) in water treatment involve hydroxyl radicals ($OH) as intermediates. UV-based AOPs essentially mimic the photo-initiated oxidation processes in natural systems, such as sun light on surface
water or in the atmosphere (Oppenlander, 2003). Ultraviolet (UV) irradiation is well established for disinfection of water. The UV/H2O2 provides oxidation through generation of hydroxyl radicals by photolysis of H2O2 (Liao and Gurol, 1995; Rudra et al., 2005). This process degrades organic contaminants, including recalcitrant odorants such as geosmin and
* Corresponding author. E-mail address:
[email protected] (A.M. Dietrich). 1 Present address: K-water, Daejeon, Korea. 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.02.006
2508
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 0 7 e2 5 1 6
2-methylisoborneol (2-MIB), mainly by the reaction with hydroxyl radicals and partially by direct UV photolysis of the contaminant (Beltran et al., 1993; Stefan and Bolton, 1998; Rosenfeldt et al., 2005; Rudra et al., 2005; Paradis and Hoffman, 2006; Rosenfeldt and Linden, 2007). Solution-phase AOPs are advantageous for contaminant removal because residual solids are not produced such as would occur with spent activated carbon from an adsorption process. Recently, UV systems labeled as “dual purpose” were applied to fullscale water treatment plants. These dual systems combine low intensity UV for disinfection and high intensity UV for both disinfection and advanced oxidation of odorants. Their operation is limited by both UV transmittance and water quality constraints because low UV transmittance attenuates UV light available to produce hydroxyl radicals from H2O2, and carbonates scavenge hydroxyl radicals (Ho et al., 2004; Cotton and Collins, 2006). The optimum H2O2 dose in the UV/H2O2 process should be empirically determined because excess H2O2 can be a hydroxyl radical scavenger (Wang et al., 2000). Typical doses applied for taste-and-odor control in full-scale treatment plants vary from >0 to 11 mg/L H2O2 (Royce and Stefan, 2005). Geosmin and 2-MIB are earthy-musty odorants found in surface water and drinking water. They are produced by cyanobacteria or actinomycetes, commonly in the warm summer and early fall seasons (Ju¨ttner and Watson, 2007). Concentrations in drinking water can range from a few to >100 ng/L; to minimize consumer complaints and produce acceptable quality water, concentrations of these odorants should be brought to below their odor threshold levels of 1e10 ng/L (Rashash et al., 1997; Dietrich, 2006; Piriou et al., 2009). Although not regulated for health effects, esthetic guidelines of 10 ng/L geosmin and 10 ng/L 2-MIB have been established in Korea and as secondary standards in Japan. These compounds are difficult to remove by conventional water treatment processes; thus activated carbon or AOPs are frequently used for their control during full-scale treatment. 2-MIB has repeatedly proven the more difficult to remove and reacts slower with hydroxyl radicals compared to geosmin (Glaze et al., 1990; Westerhoff et al., 2006; Peter and Von Gunten, 2007). In addition to problematic taste and odor compounds, disinfection byproducts (DBPs) are another major issue in drinking water quality. Most organic DBPs form from the reaction with humic or fulvic acid and disinfectants. The major classes of regulated DBPs in drinking water are trihalomethanes (THMs), which are regulated at a total concentration of 80 mg/L, and haloacetic acids (HAAs) which are regulated at a total concentration of 60 mg/L (USEPA, 2011). Chlorinated DBPs including trichloromethane and mono, di-, and trichloroacetic acid are the most prevalent (Krasner et al., 2006), but the occurrence of brominated and iodinated DBPs is both pervasive and problematic because bromo-and iodocontaining DBPs are more toxic than their chlorinated analogs (Krasner et al., 2006; Richardson et al., 2007). Brominated DBPs can regionally constitute > 40% THMs and 10e25% HAAs (Krasner et al., 1989; Hyun et al., 2005). In addition to regulated DBPs, over 90 other brominated DBPs were identified by GCeMS (Richardson et al., 2007) and additional polar brominated DBPs were detected by LCeMS (Zhang et al., 2008).
Advanced treatment technologies that control multiple contaminants and provide disinfection in full-scale water treatment are immensely beneficial to the drinking water industry. Studies report that UV-based AOPs control DBP precursors and consequently reduce DBP level in finished water (Wang et al., 2000; Chin and Berube, 2005; Sarathy and Mohseni, 2010). However, many water utilities apply prechlorination to control taste/odor or iron/manganese or ammonia nitrogen, and to obtain required CT values. Prechlorination occurs prior to coagulation and produces a variety of DBPs while AOPs typically occur after filtration to increase UV transmission, which is also after formation of DBPs. There have been only a few studies on the removal of DBPs by UV/H2O2 and these reports are contradictory. Rudra et al. (2005) reported over 90% degradation of THMs at high UV (17 000 mJ/cm2) and 0.1% H2O2 dose. In pilot-scale and bench-scale studies with natural surface waters, THMs increased at lower UV/H2O2 doses and decreased with higher UV/H2O2 doses, and HAAs decreased for two samples and increased for one sample (Paradis and Hoffman, 2006). Second order rate constants for the reaction of DBPs and odorous contaminants with hydroxyl radical have been measured (Mezyk et al., 2006; Westerhoff et al., 2006; Peter and Von Gunten, 2007). Few studies have addressed the possibility and mechanisms of DBPs degradation by UV/H2O2 process in aqueous phase. Two mechanisms are associated with UV/H2O2 oxidation of contaminants (Scheme 1). Photolysis of H2O2 produces hydroxyl radicals that subsequently react with contaminants, generally by abstracting a hydrogen or by adding to an unsaturated site. Direct UV photolysis of the contaminant can occur and result in bond homolysis and radical generation; the resulting carbon-centered radicals subsequently are oxidized by reaction with H2O2, O2, and contaminants. Direct UV photolysis can contribute substantially to contaminant degradation. Geosmin and 2-MIB were decreased by 40% and 20%, respectively, with a UV dose of 1700 mJ/cm2 in raw water blends with 1.7e2.3 mg/L TOC, 105 mg/L alkalinity as CaCO3, and pH 8.0 to 8.3 (Rosenfeldt et al., 2005). Other source water algal-contaminants degraded by UV include microcystin (Qiao et al., 2005) and conjugated odorous aldehydes (Jo and Dietrich, 2009). In regard to UV photolysis of THMs, only the brominated THMs were photolyzed and quantum yield of the photolysis for all brominated THMs was 0.43 with a 253.7 nm lamp (Nicole et al.,
Direct Photolysis hv
R3C-X
R 3C
+
X
Hydroxyl radical generation (H2O2 photolysis; hydrogen abstraction) hv
H 2O 2
HO
+
R-H
2 HO
H2O
+
R
Scheme 1 e Mechanisms for oxidation by UV alone and with H2O2. (“X” refers to a halogen).
2509
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 0 7 e2 5 1 6
1991). In the same research, polybrominated THMs were shown to be photolyzed faster than monobrominated compounds. Concentrations of tribromomethane and chlorodibromomethane in chlorinated swimming pool water decreased significantly with medium pressure UV irradiation of 145 mJ/cm2 (Cassan et al., 2006). Polyhalomethanes such as tribromomethane (CHBr3) and carbon tetrabromide (CBr4) were photolyzed by a proposed water-catalyzed OeH insertion/HBr elimination (Kwok et al., 2004; Lin et al., 2005; Zhao et al., 2005). For the direct UV photolysis and hydroxyl radical reactions of organic compounds such as geosmin/2-MIB and pharmaceutical compounds by UV/H2O2, pseudo-first order reaction models have been proposed because under constant irradiation, the steady-state concentration of the hydroxyl radical is approximately constant with respect to pollutant concentration because of the relatively high concentration of H2O2 (Rosenfeldt et al., 2005; Pereira et al., 2007). d½C ¼ k0 ½C dt Where, k0 ¼ k0d þ k0i k0 ¼ the observed pseudo-first order rate constant (s1) k0d ¼ the measured pseudo-first order rate constant of direct photolysis (s1) k0i ¼ the measured pseudo-first order rate constant of the reaction with $OH (s1) Second order reaction rate constants for reaction of hydroxyl radical with geosmin have been reported to range from 8.2 109 to 1.4 1010 M1s1, and for 2-MIB from 5.1 109 to 8.1 109 M1s1 (Glaze et al., 1990; Westerhoff et al., 2006; Peter and Von Gunten, 2007), which are nearly an order of magnitude more than for the reaction of hydroxyl radical with THMs (0.7 107e1.5 108 M1s1) or chlorinated HAAs (6 107e1.0 108 M1s1). The rate constant for reaction of tribromomethane with HO$ is greater than trichloromethane by a factor of 10 (Maruthamuthu et al., 1995; Mezyk et al., 2006). This paper summarizes our results from the investigation of the simultaneous degradation of odorants and DBPs at concentrations and conditions representative of those used for the removal of recalcitrant odorants in drinking water using UV/H2O2 based on typical H2O2 dose used in full-scale water treatment plant. The objectives of this research were to: 1) determine the degree of degradation of THMs and HAAs with respect to degradation of geosmin and 2-MIB, and 2) evaluate the role of UV photolysis and hydroxyl radical reaction in this degradation.
2.
Methods and materials
2.1.
Apparatus
Experiments were performed with a Rayonet RPR-100 photochemical reactor equipped with 253.7 nm wavelength UV lamps of 7.2 mW/cm2 total intensity, and quartz reaction vessels. UV dose was confirmed with the iodide/iodate
actinometer (Rahn, 2004; Rahn et al., 2006). Samples were completely mixed and headspace free while being irradiated with UV. A reaction temperature of 24 1.0 C was maintained by an electric fan set under the sample, and ice.
2.2.
Reagents and sample preparation
Samples for UV/H2O2 irradiation were prepared directly in deionized water (Nanopure) using individual neat compounds for DBPs; trichloromethane (99%, Fisher Scientific), tribromomethane (99%, Acros Organics), chloroacetic acid (MCAA) (99%, Aldrich), dichloroacetic acid (DCAA) (99%, SigmaeAldrich), trichloroacetic acid (TCAA) (99%, Alfa Aesar), bromoacetic acid (MBAA) (99%, SigmaeAldrich), dibromoacetic acid (DBAA) (99%, Fluka), tribromoacetic acid (TBAA) (99%, Acros Organics), tetrachloromethane (>99%, Acros Organics) and tetrabromomethane (>99%, Acros Organics). To prepare individual and mixed samples of HAAs to react with UV/H2O2, neat HAA compounds were dissolved in water because the commercial HAA standard mixture was dissolved in methyl tert-butyl ether (MTBE) and this solvent scavenges hydroxyl radicals (second order rate constant, k ¼ 3.9 109 M1s1) (Chang and Young, 2000). For quantitative chromatographic analyses, commercial standards solutions of THMs in methanol and HAAs in methyl-tert-butyl ether were purchased from Ultra Scientific. Geosmin (200 mg/L) and 2-MIB (100 mg/L) were purchased from Supelco as solutions in methanol and diluted in water to the appropriate concentration. Typical odorant solutions used in this research (Table 1) consistently contained about 2 mL or 1.5 mg/L (50 mM) methanol from dilution of the concentrated odorant solutions. Hydrogen peroxide (30%, Fisher) was
Table 1 e Typical initial concentrations of compounds in the research. Compounds
Typical concentrations mg/L
mM
a
Odorants Geosmin 2-MIB THMsb Trichloromethane Bromodichloromethane Dibromochloromethane Tribromomethane Tetrahalomethanes Carbon tetrachloride Carbon tetrabromide HAAsb Chloroacetic acid (MCAA) Dichloroacetic acid (DCAA) Trichloroacetic acid (TCAA) Bromoacetic acid (MBAA) Dibromoacetic acid (DBAA) Tribromoacetic acid (TBAA) Hydrogen Peroxide
0.04e0.2 0.1e0.3
0.0002e0.001 0.0006e0.002
60e500 90 80 80e550
0.5e4.2 0.5 0.4 0.3e2.2
350 1000
2.3 3.0
270 190 180 200 190 160 6000
2.9 1.5 1.1 1.4 0.9 0.5 176.5
a Added with z1.5 mg/L methanol. b It should be noted that these concentrations are slightly higher than those usually found in actual drinking waters.
2510
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 0 7 e2 5 1 6
2.4.
Statistical analyses
Microsoft Excel (2007) was applied for data manipulation and calculation of apparent first order rate constant and standard errors.
3.
Results
3.1.
UV absorbance
To investigate the relative role of hydroxyl radical production versus direct UV photolysis of the organic contaminants in
3.2. Degradation of odorants and DBPs at typical concentrations found in drinking water Concentrations of ng/L geosmin/2-MIB and mg/L DBPs were prepared in mixed aqueous samples and contaminant removals were simultaneously followed under the same UV/ H2O2 and solution conditions. The solutions contained about 1.5 mg/L methanol, because it was the solvent used to dissolve geosmin and 2-MIB; methanol is known to act as a hydroxyl radical scavenger (Nickelsen et al., 1994). In practice, a utility would apply a UV/H2O2 dose for geosmin/2-MIB degradation that met their treatment need to control the odor problem. Thus, although the water being treated may contain other contaminants and radical scavengers that could absorb at UV 253.7 nm or react with hydroxyl radicals, the applied UV/H2O2 dose would be determined by treatment goals for removing geosmin and 2-MIB.
1.20 1.00 0.80 0.60 0.40 0.20 0.00
A
H 2O ge 2 os m in 2M IB C H Br 3 C H C l3 C Br 4 C C l4 M BA A D BA A TB AA M C AA D C AA TC AA
1200 1000 800 600 400 200 0
B
H 2O ge 2 os m in 2M IB C H Br 3 C H C l3 C Br 4 C Cl 4 M BA A D BA A TB AA M C AA D C A A TC AA
Geosmin and 2-MIB were measured by solid-phase microextraction (SPME, Supelco) as reported by other researchers (Watson et al., 1999, 2000). Compounds partitioned from water were sorbed onto an SPME fiber (65 mm, PDMS/DVB) for 10 min at 60 C. The SPME fiber was injected in the splitless mode into the GC/MS (Agilent 5973) at 220 C and desorbed for 2 min. An Rtx-5Sil column (30 m, 0.25 mm ID) with He carrier gas and a temperature program of 60 Ce180 C by 15 C/min was used. Approximate retention times were 5.4 min for 2-MIB and 7.9 min for geosmin; m/z values of 112, 125, 182 for geosmin and m/z values of 95, 108, 168 for 2-MIB were monitored for qualitative analysis and m/z values of 112 or 95 were used for quantitative analysis for geosmin and 2-MIB respectively in selective ion mode. Concentrations of disinfection byproducts were determined using standardized methods. THMs were measured based on Standard Method 6232.D by purge/trap (Tekmar 3000) and GC (Tremetrics 9001) with DB-624 column (J & W) and ECD detector. GC temperature was initially maintained at 45 C for 3 min, and then increased by 11 C/min up to 200 C. HAAs were determined by liquideliquid extraction method (EPA Method 552.2) and GC (HP 5890) with an ECD detector. The injector temperature was 210 C and the initial oven temperature was set to 35 C and increased up to 140 C. UV adsorption was measured on a UV/vis spectrophotometer (Beckman, DU640). H2O2 concentration was determined by triiodide (I 3 ) titration method (Klassen et al., 1994).
Molar extinction
Chemical analyses
coefficient (M -1 cm -1 )
2.3.
this UV/H2O2 process, molar extinction coefficients were measured (Fig. 1A). The brominated compounds, geosmin, and 2-MIB had at least two order of magnitude higher molar extinction coefficients than chlorinated compounds, and one order of magnitude lower than H2O2. Absolute molar extinction coefficients, however, can be misleading. At typical concentrations used in this research (Table 1), only H2O2 and the brominated DBPs would absorb an appreciable amount of UV as shown by their absorptions relative to H2O2 (Fig. 1B). The relative absorbance of geosmin, 2-MIB, and any of the chlorinated compounds are nil. The UV absorbances of the brominated DBPs are sufficiently high that a plausible mechanism for brominated DBPs elimination is by direct UV photolysis (CeBr bond cleavage). At 6 mg/L H2O2, hydrogen peroxide absorbs most of the UV at 253.7 nm, which indicates that hydroxyl radicals can be produced from UV photolysis of H2O2 under these conditions. In this research, the absorbance of 6 mg/L H2O2 at 253.7 nm was ca. 0.035 in the experimental apparatus.
Relative absorbance
diluted to desired concentrations of 6 mg/L which was selected based on typical concentration range in pilot-scale study (Paradis and Hoffman, 2006), and added into the samples immediately before UV irradiation. Typical initial concentrations of compounds used in the research are shown in Table 1; it should be noted that these concentrations are slightly higher than those usually found in actual drinking waters. Molar extinction coefficients at 253.7 nm were determined for each compound dissolved in distilled water; these concentrations were measured: 34 mg/L H2O2, 0.4 mg/L geosmin, 0.1 mg/L 2-MIB, 340 mg/L for CHBr3, CHCl3, or CBr4, 150 mg/L CCl4, and 1000 mg/L for MBAA, DBAA, TBAA, MCAA, DCAA, and TCAA. Absorptions relative to H2O2 were calculated by applying the concentrations in Table 1, and using 100 ng/L for geosmin and 2-MIB, and 300 mg/L for CHCl3 and CHBr3.
Fig. 1 e Molar extinction coefficients (A) and relative absorption (relative to H2O2 [ 1) (B) of odorant and DBP compounds and hydrogen peroxide measured at 253.7 nm and at the concentrations used in this research. Concentrations used for these measurements are provided in the methods and materials.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 0 7 e2 5 1 6
3.2.1.
Geosmin and 2-MIB
The geosmin and 2-MIB results show 90 and 65% degradation, respectively, with a UV dose of 1200 mJ/cm2 and 6 mg/L H2O2. This dose was selected as it was sufficient to reduce the initial geosmin concentration of 40e43 ng/L to 4 ng/L (Figs. 2 and 3), which is below its guideline and upper odor threshold value of 10 ng/L. Under identical conditions, but in the absence of H2O2, only about 20% was degraded with UV photolysis (Fig. 2).Using the results in Figs. 2 and 3, apparent pseudo first order rate constants and half-lives for degradation of these odorants can be calculated (Table 2). The apparent rate constants for degradation of both geosmin and 2-MIB are approximately 3e7 greater when H2O2 is present. These results suggest that geosmin and 2-MIB concentrations are mainly reduced by reaction with hydroxyl radical (formed by photolysis of H2O2) though a small amount may also degraded by direct photolysis.
3.2.2.
2511
Fig. 3 e Photolysis of geosmin, 2-MIB and THMs in the presence of 6 mg/L H2O2. Initial concentration (C0): geosmin [ 43 ng/L, 2-MIB [ 100 ng/L, trichloromethane [ 63 mg/L, bromodichloromethane [ 94 mg/L, dibromochloromethane [ 76 mg/L, tribromomethane [ 82 mg/L, methanol z1.5 mg/L.
Geosmin/MIB and THMs
3.2.2.1. THMs degradation compared to geosmin/2-MIB by UV/H2O2. Brominated THMs were degraded better than trichloromethane at the UV/H2O2 dose effective for removing geosmin/2-MIB (Fig. 3). Tribromomethane and dibromochloromethane were degraded by 99% and 80%, respectively, at the UV dose of 1200 mJ/cm2 and 6 mg/L H2O2, while 90% of the geosmin and 65% of the 2-MIB were degraded at this dose. The brominated THMs with higher numbers of bromine atoms were degraded faster than ones with lower numbers of bromine or trichloromethane, which for all practical purposes, was not degraded by UV/H2O2 (Fig. 3). Tribromomethane was degraded faster than either geosmin/2-MIB. Another approach to determining an effective UVeH2O2 dose from the data in Fig. 3 is to use the guideline of 10 ng/L 2-MIB that has been established in Japan and Korea. Accordingly, a UV dose of 2700 mJ/cm2 and 6 mg/L H2O2 was required to remove 90% of the more recalcitrant odorant, 2-MIB, to a final concentration of 10 ng/L 2-MIB. At this UV/H2O2 dose for 90% removal of 2-MIB and the initial concentrations used for this experiment, the final concentrations of the other contaminants were: 0.8 ng/L geosmin (98% degradation);
Fig. 2 e Photolysis of geosmin and 2-MIB in the presence and absence of 6 mg/L H2O2. Initial concentration (C0): geosmin (no H2O2) [ 40 ng/L, geosmin (H2O2 6 mg/ L) [ 183 ng/L, 2-MIB (no H2O2) [ 108 ng/L, 2-MIB (6 mg/L H2O2) [ 306 ng/L, methanol z1.5 mg/L.
47 mg/L trichloromethane (26% degradation); 44 mg/L monobromodichloromethane (53% degradation); 0.6 mg/L dibromochloromethane (99% degradation); and <0.1 mg/L tribromomethane (>99.9% degradation). Although the THM concentrations used in this experiment were higher than concentrations typical for drinking water, the UVeH2O2 conditions applied in this research to degrade 2-MIB below its esthetic guideline would also be effective to eliminate dibromochloromethane and tribromomethane to below their regulatory limits.
3.2.2.2. Direct UV photolysis of brominated THMs. To investigate the contribution of direct UV photolysis, an aqueous mixture of THMs was reacted with UV in the absence of H2O2. Using the data in Figs. 3 and 4, apparent pseudo first order rate constants and half-lives were calculated (Table 3). The data clearly show that: a) the apparent rate constants for the brominated THMs were substantially faster than for trichloromethane, b) the rate constant for the brominated THMs increased with the number of bromine atoms in the molecule, c) the rate constants for the brominated THMs were identical in the absence and presence of H2O2. These observations make sense because the CeBr bond serves as the active chromophore which is effectively cleaved by UV (i.e., CeBr þ hn / C$ þ Br$) and chlorine-only containing compounds do not absorb appreciably at 253.7 nm, and thus, cannot be eliminated by direct photolysis.
3.2.2.3. Degradation mechanism for THMs. To further investigate the role of hydrogen abstraction and its effects on THM degradation, the reaction of carbon tetrachloride (CCl4) and carbon tetrabromide (CBr4) were compared to each other and to their corresponding trihalomethane analogs. All halomethane solutions were prepared by dissolving neat compounds in de-ionized water then individually treating with UV/H2O2. Because they do not possess abstractable hydrogens, carbon tetrachloride and carbon tetrabromide are effectively non-reactive toward hydroxyl radical and can only be degraded by direct photolysis. Both the tri- and tetrabrominated methanes were degraded faster than their
2512
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 0 7 e2 5 1 6
Table 2 e Apparent first order rate constants for photo-induced removal of geosmin and 2-MIB in the presence and absence of 6 mg/L H2O2. hn/H2O2, UV 253.7 nm
Contaminant
kapp, s1 Geosmin
2-MIB
a b c d
1.6 1.9 1.6 7.2 6.8 6.0
(0.3) (0.3) (0.3) (0.2) (0.3) (0.5)
kapp, s1
t½, s 2 b
10 102 102 103 103 103
c d b c d
krela
hn only, UV 253.7 nm
1.2 (0.6) 10
43.3 36.5 43.3 96.3 101.9 115.5
t½, s 3 b
1.1 (0.1) 103
b
577.6
14.1
630.1
6.1
Relative rate constant: mean kapp(hn/H2O2)/kapp(hn only). Based upon data from Fig. 2; standard error of slopes provided. Based upon data from Fig. 3; standard error of slopes provided. Based upon data from Fig. 6; standard error of slopes provided.
chlorinated analogs as shown in Fig. 5. The pseudo first order rate constant for degradation of CCl4 was 5.6 (0.4) 103 s1 and for CBr4 it was 8.4 (0.8) 102 s1 (calculated from the data in Fig. 5). Rate constants for degradation of CX4 (X ¼ Cl or Br) were greater than CHX3, even though the latter possesses a hydrogen atom that can be abstracted by hydroxyl radical. These results further confirm that the different degradation rates in UV/H2O2 between chlorinated and brominated THMs results from the different UV photolysis rates and not from the hydrogen abstraction by hydroxyl radicals.
3.2.3.
Geosmin/2-MIB and HAAs
3.2.3.1. HAAs degradation compared to geosmin/2-MIB by UV/ H2O2. Treatment with UV/H2O2 degraded brominated HAAs faster than chlorinated HAAs (Fig. 6) in distilled water that also contained odorants. Applying a UV dose of 1200 mJ/cm2 and 6 mg/L H2O2 as was done for the geosmin/2-MIB/THM data in Fig. 3, the geosmin concentration was reduced to 7 ng/L (96% removal), while TBAA and DBAA were degraded by 99% and 80% respectively. Chlorinated HAAs with no bromine atoms were barely degraded by a UV doses of 0e4300 mJ/cm2 and 6 mg/L of H2O2. Brominated HAAs degradation rates increased in proportion to the number of bromine atoms in
Fig. 4 e Photolysis of brominated THMs in the presence and absence of H2O2. Initial concentration (C0): trichloromethane [ 63 mg/L, bromodichloromethane [ 94 mg/ L, dibromochloromethane [ 76 mg/L, tribromomethane [ 82 mg/L.
the molecule. Consequently, tribromoacetic acid had the highest degradation rate among all HAAs and geosmin/2-MIB.
3.2.3.2. UV photolysis of brominated HAAs. TBAA, DBAA, and MBAA were treated with UV in the absence of H2O2 to investigate the contribution of direct UV photolysis as the primary mechanism for degradation of brominated HAAs. As shown in Fig. 7, there is not a huge difference in the degradation rates for MBAA, DBAA and TBAA between UV photolysis and UV/ H2O2 process e in fact, the rates in the presence of H2O2 appear to be slightly lower. This becomes clearer when the apparent rate constants and half-lives are compared (Table 4). The fact that the rate constants did not increase in the presence of H2O2 suggests that brominated HAAs are degraded mainly by UV photolysis in UV/H2O2 process, rather than via reaction with $OH.
4.
Discussion
For ng/L odorant concentrations and mg/L THM and HAA concentrations typical of drinking water, UV/H2O2 provided substantial simultaneous degradation of geosmin, 2-MIB, brominated THMs and brominated HAAs; however, chlorinated DBPs experienced only minor degradation under the same conditions. This research confirmed the results of Rosenfeldt et al. (2005) that geosmin and 2-MIB were mainly eliminated by the reaction with hydroxyl radicals in the UV/ H2O2 process and that direct UV photolysis is minor. This research concurs with that of Rudra et al. (2005) who reported removal of THMs by UVeH2O2 and also Nicole et al. (1991) who reported direct UV as the removal mechanism for brominated THMs. Brominated HAAs were demonstrated to be degraded by direct UV photolysis and chlorinated HAAs by UV/H2O2 which is consistent with the results of Paradis and Hoffman (2006), who determined that UVeH2O2 decreased HAAs in two out of three samples. For halogenated THMs, a possible first reaction step in UV/ H2O2 advanced oxidation is either hydrogen abstraction by hydroxyl radical or carbonehalogen bond cleavage by direct UV photolysis. Bond dissociation energies of the CeH bond in trichloromethane (CHCl3) and tribromomethane (CHBr3) are very close, 100.0 and 99.9 kcal/mol, respectively (McGivern
2513
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 0 7 e2 5 1 6
Table 3 e Apparent first order rate constants for photo-induced removal of brominated and chlorinated THMs in the presence and absence of 6 mg/L H2O2. Contaminant
hn/H2O2, UV 253.7 nm kapp, s1
CHCl3 CHBrCl2 CHBr2Cl CHBr3
1.5 (0.1) 2.2 (0.1) 1.8 (0.1) 1.0 (0.1) 1.3 (0.2) 2.9 (0.3) 2.7 (0.4)
3 b
10 103 b 103 c 102 b 102 c 102 b 102 c
krela
hn only, UV 253.7 nm t½, s
kapp, s1
t½, s
462.1 315.0 385.1 69.3 53.3 23.9 25.7
e 2.1 (0.2) 103 c
330.0
e 1.0
1.0 (0.1) 102 c
69.3
1.2
2.8 (0.3) 102 c
24.8
1.0
a Relative rate constant: mean kapp(hn/H2O2)/kapp(hn only). b Based upon data from Fig. 3; standard error of slopes provided. c Based upon data from Fig. 4; standard error of slopes provided.
et al., 2000) and thus, do not explain the faster degradation of brominated DBPs compared to perchlorinated DBPs. Carbonebromine cleavage due to UV photolysis is the likely mechanism of faster degradation of brominated DBPs. This is
Fig. 5 e Photolysis of halogenated methanes in the absence of H2O2. Initial concentration (C0): trichloromethane [ 514 mg/L, tribromomethane [ 523 mg/L, carbon tetrachloride [ 343 mg/L, carbon tetrabromide [ 926 mg/L.
Fig. 6 e Photolysis of geosmin, 2-MIB and HAAs in the presence of 6 mg/L H2O2. Initial concentration (C0): geosmin [ 183 ng/L, 2-MIB [ 306 ng/L, bromoacetic acid [ 202 mg/L, dibromoacetic acid [ 190 mg/L, tribromoacetic acid [ 161 mg/L, chloroacetic acid [ 271 mg/L, dichloroacetic acid [ 191 mg/L, trichloroacetic acid [ 176 mg/L, methanol z1.5 mg/L.
supported by the higher strengths of CeCl bonds than CeBr bonds in trichloromethane and tribromomethane, 80.1 and 70.4 kcal/mol, respectively (McGivern et al., 2000). The energy of a UV photon at 253.7 nm is about 113 kcal/mol, which is sufficient to break both the CeBr bond (Kwok et al., 2004) and CeCl (Kurac and Hlatka, 1992). The higher molar absorption coefficients of brominated DBPs (Fig. 1) and the equal rates of degradation of brominated THMs by UV photolysis with and without H2O2 (Fig. 4 and Table 3) indicate that brominated compounds absorb more UV and thus are more susceptible to photolysis than chlorinated compounds. Although the CeCl is weak, the chlorinated compounds are not degraded by direct UV photolysis because the CeCl chromophore has a low extinction coefficient, i.e., the molecule does not absorb sufficient light to be degraded by this mechanism. The degradation rates of the brominated DBPs increased with the number of bromine atoms present in the molecule. Tribromomethane and tribromoacetic acid were eliminated the fastest among the THMs and HAAs (Figs. 4 and 7, and Table 4). In this research, brominated THMs and HAAs were shown to be eliminated mainly by UV photolysis in the UV/ H2O2 process. Previously, Nicole et al. (1991) demonstrated that UV photolysis of CHBr3, CHBr2Cl, and CHCl2Br at 253.7 nm in aqueous solution led to complete conversion to the corresponding halide. Regarding UV (253.7) photolysis of tribromomethane, a water-catalyzed mechanism was proposed in which isotribromomethane isomerized from
Fig. 7 e Photolysis of brominated HAAs in the presence and absence of 6 mg/L H2O2. Initial concentration (C0): bromoacetic acid [ 202 mg/L, dibromoacetic acid [ 190 mg/L, tribromoacetic acid [ 161 mg/L.
2514
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 0 7 e2 5 1 6
Table 4 e Apparent first order rate constants for photo-induced removal of chlorinated and brominated acetic acids in the presence and absence of H2O2. (explanation of acronyms: AA [ acetic acid; B [ bromo; C [ chloro; M [ mono; D [ di; T [ tri). hn/H2O2, UV 253.7 nm
Contaminant
kapp, s1 MBAA DBAA TBAA MCAA DCAA TCAA
1.1 1.3 6.4 5.8 5.0 2.9
(0.1) (0.1) (0.4) (2.6) (1.2) (2.3)
103 102 102 104 104 104
kapp, s1
t½, s b b b b b b
krela
hn only, UV 253.7 nm
2.1 (0.1) 103 1.5 (0.1) 102 7.4 (1.1) 102 e e e
630.9 53.3 10.8 1195 1390 2390
t½, s c c c
330.0 46.2 9.4
0.52 0.87 0.86
a Relative rate constant: mean kapp(hn/H2O2)/kapp(hn only). b Based upon data from Fig. 6; standard error of slopes provided. c Based upon data from Fig. 7, standard error of slopes provided.
tribromomethane reacts with water molecule resulting in OH insertion and HBr elimination. The major products from tribromomethane were carbon monoxide and three equivalents of HBr for the major route and formic acid plus three equivalents of HBr for the minor route (Kwok et al., 2004). Brominated compounds are increasingly identified as toxic contaminants in drinking water. In addition to the regulated THMs and HAAs, other bromine-containing disinfection byproducts in drinking water include halonitromethanes, 3e4 carbon haloacids, haloaldehydes, haloketones, haloamides, and halogenated furanones (Krasner et al., 2006). Non-DBP brominated compounds of health concern also occur in source and treated water including the regulated volatile organic compound methyl bromide, brominated flame retardants, and odorous bromophenols (Piriou et al., 2007). In this research the UV/H2O2 process reduced the concentrations of geosmin and 2-MIB in distilled water. This is consistent with the results of Rosenfeldt et al. (2005), who reported degradation of geosmin and 2-MIB in distilled water, and also in source waters and treated drinking waters which possessed other reactants for UV and hydroxyl radicals including natural organic matter and bicarbonate/carbonate ions. The degradation of geosmin, 2-MIB, and perchlorinated DBPs depend on reaction with hydroxyl radicals. The degradation of brominated compounds is primarily through direct UV photolysis that would consume few hydroxyl radicals and would not be affected by the presence of hydroxyl radical scavengers. This is an advantage to the use of UV/H2O2 to simultaneously remove brominated compounds while removing geosmin and 2-MIB. Furthermore, water that contains greater than 0.10 mg/L bromide ion could result in bromate formation with ozonation (Song et al., 1997), and such source waters would be unsuitable for the application of ozone to control geosmin and/or MIB due to bromate formation (Ho et al., 2004; Westerhoff et al., 2006; Peter and Von Gunten, 2007); UV/H2O2 would be the preferred AOP in this case. This research has implications for application of UV at full-scale facilities to control pathogens and contaminants. Although application of UV at doses of approximately 40 mJ/cm2 is considered viable for full-scale drinking water disinfection (Linden et al., 2004), full-scale data for contaminant removal by UV are not readily available. Pilot- and bench-
scale research by Mofidi et al. (2002) demonstrated that in trials using 50 ng/L geosmin and 2-MIB added to source water, a UV dose of 1100 mJ/cm2 with 5.5 mg/L H2O2 reduced both odorants by >90% while a high UV dose of 10 100 mJ/cm2 was required for >90% removal in the absence of H2O2. The authors calculated that at a reasonable dose of 100 mJ/cm2 for disinfection, even with the addition of 5 mg/L H2O2, only 50e60% of the added geosmin and 2-MIB would be removed and the earthy-musty odors would persist. THM and HAA degradation were not investigated. Linden et al. (2004) concur that UV alone at doses compatible with disinfection will not be effective for destruction of organic contaminants. Their bench-scale studies with geosmin, 2-MIB, and pesticides indicated that degradation of organic contaminants would likely require 500e2000 mJ/cm2 and even then, some contaminants would not be removed without addition of H2O2. Our research confirms that degradation of geosmin and 2-MIB occurs at doses of 500e2000 mJ/cm2 with addition of H2O2, and that brominated DBPs are degraded under these conditions whereas perchlorinated DPBs are not well removed.
5.
Conclusions
The results demonstrate that mg/L concentrations of brominated THMs and HAAs can be simultaneously degraded by advanced oxidation when applying 6 mg/L H2O2 and a UV dose effective for removing ng/L concentrations of geosmin and 2-methylisoborneol. Trichloromethane and perchlorinated haloacetic acids could not be effectively degraded under these same advanced oxidation conditions. Geosmin, 2-MIB, and chlorinated DBPs were primarily degraded by the reaction with hydroxyl radicals, generated via the photolysis of H2O2. While direct UV photolysis played a minor role in removing geosmin and 2-methylisoborneol, the perchlorinated DBPs were not degraded by direct photolysis. Brominated DBPs were degraded primarily by direct photolysis, presumably via photo-induced CeBr bond cleavage. The apparent pseudo-first order rate constants for photolysis of the brominated THMs and HAAs were between
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 0 7 e2 5 1 6
103 to 102 s1 and increased with the number of bromine atoms in the molecule. Tribromomethane and dibromochloromethane were degraded by 99% and 80%, respectively, at the UV dose of 1200 mJ/cm2 with 6 mg/L H2O2, while 90% of the geosmin and 65% of the 2-methylisoborneol were degraded. Tribromoacetic acid and dibromoacetic acid were degraded by 99% and 80% respectively under the same conditions. Implications of this research are that the UV/H2O2 process, when implemented for odor control, can have the additional benefit of DBP elimination, especially in regions where source water bromide concentration leads to high concentrations of brominated DBPs. UV/H2O2 may also be more desirable than ozone in these regions due to possible formation of bromate from bromide ion during ozonation.
Acknowledgments The authors thank Kwater (Korea Water Resources Corporation) which provided a research fellowship for Dr. Chang Hyun Jo and the MILES (Macromolecular Interfaces with Life Science)eIGERT program at Virginia Tech (NSF agreement # DGE0333378) for experimental support. Neither funding agency contributed to the study design; collection, analysis or interpretation of data; decision to submit the paper for publication.
references
Beltran, F.J., Ovejero, G., Acedo, B., 1993. Oxidation of atrazine in water by ultraviolet-radiation combined with hydrogenperoxide. Water Research 27 (6), 1013e1021. Cassan, D., Mercier, B., Castex, F., Rambaud, A., 2006. Effects of medium-pressure UV lamps radiation on water quality in a chlorinated indoor swimming pool. Chemosphere 62 (9), 1507e1513. Chang, P.B.L., Young, T.M., 2000. Kinetics of methyl tert-butyl ether degradation and by-product formation during UV/hydrogen peroxide water treatment. Water Research 34 (8), 2233e2240. Chin, A., Berube, P.R., 2005. Removal of disinfection by-product precursors with ozone-UV advanced oxidation process. Water Research 39 (10), 2136e2144. Cotton, C.A., Collins, J.R., 2006. Dual purpose UV light: Using UV light for disinfection and for taste and odor oxidation. Proceedings of the 2006 Water Quality Technology Conference. American Water Works Association, Denver, CO, USA. Dietrich, A.M., 2006. Aesthetic issues for drinking water. Journal of Water and Health 4 (Suppl. 1), 11e16. Glaze, W.H., Schep, R., Chauncey, W., Ruth, C., Zarnoch, J.J., Aieta, E.M., Tate, C.H., McGuire, M.J., 1990. Evaluating oxidants for the removal of model taste and odor compound from a municipal water-supply. Journal of the American Water Works Association 82 (5), 79e84. Ho, L., Croue, J.P., Newcombe, G., 2004. The effect of water quality and NOM character on the ozonation of MIB and geosmin. Water Science and Technology 49 (9), 249e255. Hyun, K.S., Kim, Y.J., Choi, J.J., 2005. Profiles of THM and HAAs in the process of water treatment plants. Journal of Korean Society Water Science and Technology 13 (1), 55e66.
2515
Jo, C.-H., Dietrich, A.M., 2009. Removal and transformation of odorous aldehydes by UV/H2O2- efficiency and effect. Journal of Water Supply: Research and Technology-AQUA 58 (8), 580e587. Ju¨ttner, F., Watson, S.B., 2007. Biochemical and ecological control of geosmin and 2-methylisoborneol in source waters. Applied and Environmental Microbiology 73 (14), 4395e4406. Klassen, N.V., Marchington, D., McGowan, H.C.E., 1994. H2O2 determination by the I 3 method and by KMnO4 titration. Analytical Chemistry 66 (18), 2921e2925. Krasner, S.W., McGuire, M.J., Jacangelo, J.G., Patania, N.L., Reagan, K.M., Aieta, E.M., 1989. The occurrence of disinfection by-products in US drinking water. Journal of the American Water Works Association 81 (8), 41e53. Krasner, S.W., Weinberg, H.S., Richardson, S.D., Pastor, S.J., Chinn, R., Sclimenti, M.J., Onstad, G.D., Thruston, A.D., 2006. Occurrence of a new generation of disinfection byproducts. Environmental Science & Technology 40 (23), 7175e7185. Kurac, J., Hlatka, A., 1992. Photolysis of waterecarbon tetrachloride and waterechlorobenzene saturated solutions and their two phase system: quantum yields of chloride ions. Journal of Radioanalytical and Nuclear Chemistry 166 (3), 239e250. Kwok, W.M., Zhao, C.Y., Li, Y.L., Guan, X.G., Wang, D.G., Phillips, D.L., 2004. Water-catalyzed dehalogenation reactions of isobromoform and its reaction products. Journal of the American Chemical Society 126 (10), 3119e3131. Liao, C.-H., Gurol, M.D., 1995. Chemical oxidation by photolytic decomposition of hydrogen peroxide. Environmental Science and Technology 29 (12), 3007e3014. Lin, X.F., Guan, X.G., Kwok, W.M., Zhao, C.Y., Du, Y., Li, Y.L., Phillips, D.L., 2005. Water-catalyzed OeH insertion/HI elimination reactions of isodihalomethanes (CH2XeI, where X ¼ Cl, Br, I) with water and the dehalogenation of dihalomethanes in water-solvated environments. The Journal of Physical Chemistry A 109 (6), 981e998. Linden, C., Sharpless, C., Andrews, S., Atasi, K., Korategere, V., Stephan, M., Suffet, I.H., 2004. Innovative UV Technologies to Oxidize Organic and Organoleptic Chemicals. AwwaRF, Denver CO USA. Maruthamuthu, P., Padmaja, S., Huie, R.E., 1995. Rate constants for some reactions of free-radicals with haloacetates in aqueous-solution. International Journal of Chemical Kinetics 27 (6), 605e612. McGivern, W.S., Derecskei-Kovacs, A., North, S.W., Francisco, J.S., 2000. Computationally efficient methodology to calculate CeH and CeX (X ¼ F, Cl, and Br) bond dissociation energies in haloalkanes. The Journal of Physical Chemistry A 104 (2), 436e442. Mezyk, S.P., Helgeson, T., Cole, S.K., Cooper, W.J., Fox, R.V., Gardinali, P.R., Mincher, B.J., 2006. Free radical chemistry of disinfection-byproducts. 1. Kinetics of hydrated electron and hydroxyl radical reactions with halonitromethanes in water. The Journal of Physical Chemistry A 110 (6), 2176e2180. Mofidi, A., Min, J., Palencia, L., Coffey, B., Liang, S., Green, J., 2002. Advanced Oxidation Processes and UV Photolysis for Treatment of Drinking Water. California Energy Commission. http://www.energy.ca.gov/reports/2004-04-02_500-02-019/50002-019_A1.PDF (accessed January 2011). Nickelsen, M.G., Cooper, W.J., Lin, K., Kurucz, C.N., Waite, T.D., 1994. High energy electron beam generation of oxidants for the treatment of benzene and toluene in the presence of radical scavengers. Water Research 28 (5), 1227e1237. Nicole, I., De Laat, J., Dore, M., Duguet, J.P., Sutyh, E., 1991. Degradation of trihalomethanes in dilute aqueous solution by UV irradiation e determination of the quantum yield of photolysis at 253.7 nm. Environmental Technology 12, 21e31. Oppenlander, T., 2003. Photochemical Purification of Water and Air. Wiley-VHC.
2516
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 0 7 e2 5 1 6
Paradis, N., Hoffman, A., 2006. Mitigation of taste and odour compounds by UV/H2O2 advanced oxidation. In: Proceedings of the 2006 Water Quality Technology Conference. American Water Works Association, Denver, CO, USA November 6e9, 2006. Pereira, V.J., Weinberg, H.S., Linden, K.G., Singer, P.C., 2007. UV degradation kinetics and modeling of pharmaceutical compounds in laboratory grade and surface water via direct and indirect photolysis at 254 nm. Environmental Science and Technology 41 (5), 1682e1688. Peter, A., Von Gunten, U., 2007. Oxidation kinetics of selected taste and odor compounds during ozonation of drinking water. Environmental Science and Technology 41 (2), 626e631. Piriou, P., Devesa, R., De Lalande, M., Glucina, K., 2009. European reassessment of MIB and geosmin perception in drinking water. Journal of Water Supply: Research and TechnologyAQUA 58 (8), 532e538. Piriou, P., Soulet, C., Acero, J.L., Bruchet, A., Von Gunten, U., Suffet, I.H., 2007. Understanding medicinal taste and odour formation in drinking waters. Water Science and Technology 55 (5), 85e94. Qiao, R.P., Li, N., Qi, X.H., Wang, Q.S., Zhuang, Y.Y., 2005. Degradation of microcystin-RR by UV radiation in the presence of hydrogen peroxide. Toxicon 45 (6), 745e752. Rahn, R.O., 2004. Spatial distribution of upper-room germicidal UV radiation as measured with tubular actinometry as compared with spherical actinometry. Photochemistry and Photobiology 80 (2), 346e350. Rahn, R.O., Bolton, J., Stefan, M.I., 2006. The iodide/iodate actinometer in UV disinfection: determination of the fluence rate distribution in UV reactors. Photochemistry and Photobiology 82 (2), 611e615. Rashash, D.M.C., Dietrich, A.M., Hoehn, R.C., 1997. Flavor profile analysis of selected odorous compounds. Journal of the American Water Works Association 89 (4), 131e142. Richardson, S.D., Plewa, M.J., Wagner, E.D., Schoeny, R., DeMarini, D.M., 2007. Occurrence, genotoxicity, and carcinogenicity of regulated and emerging disinfection byproducts in drinking water: a review and roadmap for research. Mutation Research-Reviews in Mutation Research 636, 178e242. Rosenfeldt, E.J., Linden, K.G., 2007. The R-OH, R-UV concept to characterize and the model UV/H2O2 process in natural waters. Environmental Science and Technology 41 (7), 2548e2553. Rosenfeldt, E.J., Melcher, B., Linden, K.G., 2005. UV and UV/H2O2 treatment of methylisoborneol (MIB) and geosmin in water. Journal of Water Supply: Research and Technology-AQUA 54 (7), 423e434.
Royce, A., Stefan, M., 2005. Application of UV in drinking water treatment for simultaneous disinfection and removal of taste and odor compounds. In: Proceedings of the 2005 Water Quality Technology Conference. American Water Works Association, Quebec City, Canada November 6e10, 2005. Rudra, A., Thacker, N.P., Pande, S.P., 2005. Hydrogen peroxide and ultraviolet irradiations in water treatment. Environmental Monitoring and Assessment 109 (1e3), 189e197. Sarathy, S., Mohseni, M., 2010. Effects of UV/H2O2 advanced oxidation on chemical characteristics and chlorine reactivity of surface water natural organic matter. Water Research 44 (15), 4087e4096. Song, R., Westerhoff, P., Minear, R., Amy, G., 1997. Bromate minimization during ozonation. Journal of the American Water Works Association 89 (6), 69e78. Stefan, M.I., Bolton, J.R., 1998. Mechanism of the degradation of 1,4-dioxane in dilute aqueous solution using the UV hydrogen peroxide process. Environmental Science and Technology 32 (11), 1588e1595. United Stated Environmental Protection Agency (USEPA), 2011. National Primary Drinking Water Regulations. http://water.epa. gov/drink/contaminants/index.cfm#List (accessed 01.14.11). Wang, G.S., Hsieh, S.T., Hong, C.S., 2000. Destruction of humic acid in water by UV light e catalyzed oxidation with hydrogen peroxide. Water Research 34 (15), 3882e3887. Watson, S.B., Brownlee, B., Satchwill, T., Hargesheimer, E.E., 2000. Quantitative analysis of trace levels of geosmin and MIB in source and drinking water using headspace SPME. Water Research 34 (10), 2818e2828. Watson, S.B., Brownlee, B., Satchwill, T., McCauley, E., 1999. The use of solid phase microextraction (SPME) to monitor for major organoleptic compounds produced by chrysophytes in surface waters. Water Science and Technology 40 (6), 251e256. Westerhoff, P., Nalinakumari, B., Pei, P., 2006. Kinetics of MIB and geosmin oxidation during ozonation. Ozone-Science and Engineering 28 (5), 277e286. Zhang, X., Talley, J.W., Boggess, B., Ding, G.Y., Birdsell, D., 2008. Fast selective detection of polar brominated disinfection byproducts in drinking water using precursor ion scans. Environmental Science and Technology 42 (17), 6598e6603. Zhao, C.Y., Lin, X.F., Kwok, W.M., Guan, X.G., Du, Y., Wang, D.Q., Hung, K.F., Phillips, D.L., 2005. Water-catalyzed dehalogenation reactions of the isomer of CBr4 and its reaction products and a comparison to analogous reactions of the isomers of di- and trihalomethanes. Chemistrya European Journal 11 (4), 1093e1108.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 1 7 e2 5 2 6
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Oxidation kinetics of two pesticides in natural waters by ozonation and ozone combined with hydrogen peroxide Pamela Chelme-Ayala a, Mohamed Gamal El-Din a,*, Daniel W. Smith a, Craig D. Adams b a
Department of Civil and Environmental Engineering, Markin/CNRL Natural Resources Engineering Facility, University of Alberta, Edmonton, Alberta, Canada T6G 2W2 b Department of Civil, Environmental and Architectural Engineering, University of Kansas, Lawrence, KS 66045, USA
article info
abstract
Article history:
The oxidation of bromoxynil and trifluralin was investigated using ozone (O3) and O3
Received 12 August 2010
combined with hydrogen peroxide (H2O2) in natural waters using batch reactors. The
Received in revised form
results indicated that these pesticides could not be completely degraded during ozonation,
31 January 2011
achieving degradation levels lower than 50%. An enhancement of the level of degradation
Accepted 6 February 2011
was observed using O3/H2O2 process. A biphasic behaviour of O3 was also observed.
Available online 12 February 2011
Depending on the experimental conditions, the rate constant for O3 decomposition was estimated to be between 7.4 104 s1 to 5.8 102 s1, and 3.2 103 s1 to 4.2 102 s1
Keywords:
for bromoxynil and trifluralin samples, respectively. Acute toxicity analysis performed
Ozone
using Microtox showed a decrease in the toxic effects of the samples on the luminescent
Hydrogen peroxide
bacteria during the first few minutes of treatment, followed by an increase of the toxic
Pesticides
effects at the end of the reaction for both pesticides. The quantification of oxidation by-
Natural water
products generated during treatment was also addressed. The total molar balances of the
By-product
degradation by-products versus the initial pesticide concentrations ranged from 60 to 103%
Toxicity
under different experimental conditions. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
Pesticide contamination in the environment has been well documented around the world. Runoff from urban areas, return flow waters from agricultural fields, and leaching are considered important contributors to pesticide contamination of surface and groundwater bodies (Neumann et al., 2002). Even though the use of pesticides has been reduced drastically due to more rigorous regulations and innovative pesticideapplication methods, the pesticides continue to be detected in natural waters (Mattice et al., 2010; Kaushik et al., 2010; Navarro et al., 2010). Some of the most frequently detected pesticides in surface water and irrigation flow waters in Alberta, Canada, are
atrazine, 2,4-dichlorophenoxyacetic acid (2,4-D), dicamba, 2-methyl-4-chlorophenoxyacetic acid (MCPA), bromoxynil, trifluralin, and clodinafop (Anderson, 2005; Byrtus et al., 2002). Although few samples have exceeded the Canadian water quality guidelines (Anderson, 2005), the synergistic adverse effects of the occurrence of multiple pesticides in surfaces waters poses many uncertainties (Laetz et al., 2010; Norgaard and Cedergreen, 2010). Bromoxynil and trifluralin have not been studied extensively even though these two pesticides are intensely used worldwide (Alberta Environment, 2001; Sipyagin et al., 2004). Moreover, they have been found in reservoirs supplying drinking water to small communities in Manitoba, Saskatchewan, and Alberta, Canada (Donald et al., 2007). Bromoxynil is
* Corresponding author. 3-093 Markin/CNRL Natural Resources Engineering Facility, Department of Civil and Environmental Engineering, University of Alberta, Edmonton, Alberta, Canada T6G 2W2. Tel.: þ1 780 492 5124; fax: þ1 780 492 8198. E-mail address:
[email protected] (M.G. El-Din). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.02.007
2518
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 1 7 e2 5 2 6
a phenolic benzonitrile-based herbicide used extensively for the control of broad-leaved weeds in grain crops. Trifluralin is a pre-emergence dinitroaniline herbicide used for the control of annual grasses and broadleaved weeds in beans, cotton, groundnuts, forage legumes, and vineyards. Both pesticides are classified as group C, possible human carcinogens (U.S. EPA, 1998; Greene and Pohanish, 2005), and have a relative high toxic effect on aquatic organisms. Several treatment technologies such as adsorption and oxidation processes are used to remove pesticides from drinking water (Miltner et al., 1989; Rodriguez et al., 2009). Biological treatment has also been used as a treatment option (Gisi et al., 1997). However, the biological treatment systems are susceptible to toxic compounds that could inactivate the waste-degrading microorganisms. In such cases, the use of advanced oxidation processes (AOPs) seems to be more appropriate (Chiron et al., 2000; Sanches et al., 2010). Among the various advanced oxidation technologies used in water and wastewater treatment, ozone (O3) and O3 combined with hydrogen peroxide (H2O2) have been recognized as efficient technologies for pesticide oxidation (Ikehata and Gamal El-Din, 2005a,b; Catalkaya and Kargi, 2009). In spite of the abundant literature on pesticide removal, studies dealing with their degradation are still needed. The oxidation of atrazine, MCPA, 2,4-D, and dicamba by AOPs have been well documented; however, the studies of bromoxynil, and trifluralin have been scarcely reported, especially their removal from natural waters. Kochany and Choudhry (1990) found bromoxynil oxidation levels higher than 80% after 30 min of irradiation at 313 nm. Kearney et al. (1987) reported trifluralin degradation levels lower than 60% by using UV light combined with O3 after 1 h of reaction. Ormad et al. (2008) found trifluralin oxidation efficiency of 80 and 65% by using O3 and chlorine, respectively. In this study, the degradation of two pesticides, bromoxynil and trifluralin, in natural water by O3 and O3 combined with H2O2 was investigated. The effects of the pH, alkalinity and humic acid on pesticide degradation were investigated. The reaction kinetics of bromoxynil and trifluralin in natural waters were also studied. In addition, the acute toxicity and the quantification of the major oxidation by-products were assessed.
2.
Materials and methods
2.1.
Chemicals
The two pesticides investigated in this study e bromoxynil (3,5-dibromo-4-hydroxy benzonitrile) and trifluralin (a,a,atrifluoro-2,6-dinitro-N,N-dipropyl-p-toluidine) e together with 4-hydroxy benzonitrile, 2,6-dinitro-4-trifluoromethyl aniline, 4-trifluoromethyl aniline, tert-butyl alcohol (TBA), and p-chlorobenzoic acid (pCBA) were obtained from SigmaeAldrich (Canada) at purity higher than 99%. Humic acid as sodium salt was obtained from Acros Organic (New Jersey, USA). All treatment chemicals and solvents (Fisher Scientific, Canada) were at least reagent grade and were used without further purification.
2.2.
Sample preparation
Ozone-demand-free (ODF) water was prepared using ultrapure water (resistivity > 18.2 MU $ cm) and was produced by bubbling O3 gas (ozone generator PCI-Wedeco, Model GLS-7) through the water for about 30 min. The ozonated solution was left overnight to eliminate any O3 residual and then, stored in ODF glass containers until needed. Different natural waters were selected to assess the oxidation of these pesticides in natural systems. The water samples were collected from two locations in the Province of Alberta, Canada. Water 1 (W1) was collected from the North Saskatchewan River, upstream of the City of Edmonton, and water 2 (W2) was collected from an irrigation return flow discharging into the Redwater River, receiving no inputs from the City of Edmonton. The natural waters were filtered using 0.2 mm nylon filters within 24 h after sampling and stored at 4 C until they were used, the storage period being less than one month. The water W1 had a pH around 8.1, total organic carbon (TOC) of 2.8 mg C/L, absorbance of ultraviolet light at a wavelength of 254 nm (UV254) of 0.1 cm1, and total alkalinity of 136 mg/L as calcium carbonate (CaCO3). The water W2 had a pH of 8.4, TOC of 20.6 mg C/L, UV254 of 0.6 cm1, and total alkalinity of 230 mg/L as CaCO3. To estimate the kinetics in natural water, the samples were fortified with the pesticide and pCBA.
2.3.
Ozonation experiments
All experiments were performed at room temperature (20 1 C) and at least in duplicate. The initial concentrations of bromoxynil and trifluralin were 3.6 106 M and 3.0 106 M, respectively; while the initial O3 doses ranged from 2.1 105 M to 1.2 104 M. Hydrogen peroxide levels were between 1 105 M and 5 104 M, with H2O2/O3 molar ratio between 0.3 and 2.0. A custom-designed glass reactor (400 mL) was used in the experiments and operated in batch mode. By using a floating Teflon lid, no headspace was allowed in the reactors during the experiments in order to prevent O3 volatilization. The O3 stock solutions were prepared by bubbling O3 gas into the ODF water by using a diffuser. The ozonation experiments were initiated by mixing aqueous O3 stock solution (O3 concentration of 6.3e8.3 104 M) with the appropriate amounts of H2O2 (if needed) and pesticide solutions. At different time intervals, samples were withdrawn for residual aqueous O3 measurement and for the quantification of pesticide concentrations (for that purpose, the O3 residuals in the samples were quenched with sodium thiosulfate immediately after being collected). The first sets of experiments were performed in ODF water, and the pH was adjusted using a 5 mM phosphate buffer solution. The main objective of these experiments was to determine the rate constant for direct reactions with O3 and the rate constant of the reaction between the pesticides and hydroxyl radicals (OH). The direct ozonation rate constants were determined at pH 2 and in presence of 10 103 M TBA as OH radical scavenger. At pH 9, the competition kinetics method was used to determine the rate constants for the reactions of the pesticides with OH radicals. For these
2519
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 1 7 e2 5 2 6
experiments, pCBA was used as an OH probe compound. The kinetic rate constants predicted in this stage were used to model the degradation of the pesticides in natural water. To study the pH effect and impact of alkalinity and humic acid on pesticide degradation, ODF water was fortified with humic acid and sodium bicarbonate at different concentrations.
2.4.
Analytical determination
The residual bromoxynil, trifluralin, and pCBA concentrations were separated and detected by using a Shimadzu high performance liquid chromatography (HPLC) instrument (LC10AT VP) equipped with a Phenomenex Gemini 5m C18 110A (250 4.6 mm) column. Bromoxynil detection was carried out using acetonitrile and 10 mM aqueous phosphoric acid (50%:50% v/v) in an isocratic mode. The mobile phases for trifluralin analysis were acetonitrile and ultrapure water (75%:25% v/v) while an elution of methanol and 10 mM aqueous phosphoric acid (60%:40% v/v) was used for pCBA detection. UV detection at a wavelength of 220, 237 nm, and 240 nm was used to quantify bromoxynil, trifluralin, and pCBA, respectively. A Varian 500-MS ion trap mass spectrometer (Varian Canada Inc.) was used to identify the oxidation by-products by using electrospray ionization in positive and negative modes. Bromoxynil oxidation by-products detection was carried out using the nebulizer gas pressure at 276 kPa and 138 kPa as the drying gas pressure, while trifluralin by-products were detected using the nebulizer gas pressure at 138 kPa and 345 kPa as the drying gas pressure. Aqueous O3 residuals were measured by using indigo method (APHA, 2005). The TOC levels in the water matrices were estimated using a TOC combustion analyzer Apollo 9000 (Teledyne Tekmar Co., Ohio).
2.5.
Toxicity analysis
The acute toxicity of bromoxynil and trifluralin solutions, before and after treatment, was measured using the Microtox bioassay. A model 500 Microtox analyzer (Strategic Diagnostic Inc.) was used to measure the light emitted by the Vibrio fischeri as a result of its normal metabolic processes. The Microtox 81.9% screening test protocol was used for the toxicity assessment of samples (analysis performed using 81.9% of initial sample concentrations). Three replicates of toxicity analyses were performed for each sample.
3.
Results and discussion
3.1. Rate constant for reactions with ozone and hydroxyl radical The rate constants for direct reaction with O3 were estimated at pH 2, using TBA as OH radical scavenger and in the presence of at least a 10-fold excess of the O3 residual. TBA was chosen because its reaction with O3 is very slow (kO3 TBA ¼ 0:001M1 s1 ) but it reacts quickly with OH (Yao and Haag, 1991; Hoigne´ and Bader, 1983; Buxton et al., 1988). Under these conditions, oxidation by molecular O3 was the predominant pathway for
substrate degradation. Because OH radicals cannot be measured directly during ozonation processes, competition kinetics was used to determine the second-order rate constants for the reactions of bromoxynil and trifluralin with OH radicals. These experiments were carried out at pH 9 with the presence of H2O2 (molar ratio H2O2/O3 of 0.5) and pCBA as a reference compound. pCBA was selected as a reference compound because of its low reactivity with O3, with a rate constant for direct reaction with O3 being kO3 pCBA < 0:15M1 s1 (Yao and Haag, 1991). In addition, pCBA reacts readily with OH radical, with a rate constant k$OHpCBA of 5 109 M1 s1 (Buxton et al., 1988). Before performing the O3/H2O2 experiments, the pesticides were allowed to be in contact with H2O2 (6 105 M) for 24 h at room temperature. The results indicated that the pesticides were not oxidized by H2O2 alone. Table 1 summarizes the estimated second-order rate constants for the reactions of O3 (kO3 P ) and OH radical (k$OHP ) with bromoxynil and trifluralin. As shown, bromoxynil was more reactive with O3 compared to trifluralin. It was also found that both pesticides were very reactive towards 9 1 1 OH, with k s . $OHP values above 10 M In a previous research study (Chelme-Ayala et al., 2010), the second-order rate constants (kO3 P ) of the reactions of theses pesticides and O3 were estimated at pH 2.06 and different experimental conditions. In the present study, the value of kO3 P was higher (2 times), than that of the published study, although in the same order of magnitude. The effect of pH on bromoxynil may be responsible for the difference in the values of kO3 P . Dissociating compounds are generally expected to behave differently during ozonation at different pHs (Leitner and Roshani, 2010). Small changes in the pH (of about 0.3) may cause an increase of the kO3 P values as reported by Lee et al. (2007a). The second order rate constant of O3 and trifluralin are in good agreement with the values reported previously. The k$OHP values were also estimated in a previous study (Chelme-Ayala et al., 2010), but at pH 11. The second order rate constant of OH radicals and bromoxynil was found to be 8.5 109 M1 s1, two times lower than the value reported in the present study. This difference can be attributed to the reaction with OH radical taking place at a site not affected by the deprotonation of bromoxynil as observed by other researchers (Leitner and Roshani, 2010). The pH did not influence the levels of reactivity of trifluralin with OH, showing similar k$OHP values at pHs 9 and 11.
Table 1 e Second-order rate constants for the reaction of ozone (kO3 LP ) and OH radical (k$OHLP ) with bromoxynil and trifluralin in ultrapure water. Pesticide
kO3 P (M1 s1)
k$OHP (M1 s1)
Bromoxynil Trifluralin
(5.2 0.08) 102 (1.0 0.09) 102
(2.0 0.7) 1010 (7.1 0.7) 109
Experimental conditions: [bromoxynil]0 ¼ 1.8 106 M, [trifluralin]0 ¼ 1.5 106 M, [TBA]0 ¼ 10.0 103 M, [pCBA]0 ¼ 1.8 106 M in bromoxynil experiments and 1.5 106 M in trifluralin experiments; [ozone]0 ¼ 2.8 to 3.0 105 M in bromoxynil experiments and 2.6 to 2.8 105 M in trifluralin experiments, [H2O2]0/[O3]0 ¼ 0.5, pH ¼ 2.30.
2520
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 1 7 e2 5 2 6
3.2. Effect of pH, alkalinity and humic acid on pesticide degradation The effect of pH, alkalinity and humic acid levels on the degradation of bromoxynil and trifluralin during ozonation were examined in buffered ODF water. Increasing the solution pH increased the levels of pesticide degradation. At pH 7, degradation levels of 87% and 41% were achieved for bromoxynil and trifluralin, respectively, with an averaged utilized-ozone dose of 1.7 105 M; whereas 99% and 55% of bromoxynil and trifluralin were oxidized at pH 9. This was expected as the deprotonated pesticide species have higher reactivity towards O3 compared to protonated species, as the pH increases (Hoigne´ and Bader, 1976). In addition, the high levels of OH radicals at pH 9 (Hoigne´ and Bader, 1976; Sa´nchez-Polo et al., 2002) could contribute to the increment of the pesticide oxidation at this pH. At pH 9 (pH value higher than the dissociation rate constant pKa of 4.1), the phenolate form of bromoxynil is the major species in solution, being more reactive to O3. Similarly, the deprotonated trifluralin species are dominant at pH higher than its pKa value of 5.3. The same trend was observed for in experiments performed using O3/H2O2. The highest degradation was observed at pH 9 using O3/H2O2, with removals of 99% for bromoxynil and 67% for trifluralin. To simulate the effect of humic substances, a commercial humic acid (4.4 106 M to 22.0 106 M) was used in the tests. For these sets of experiments, bicarbonate alkalinity was added as scavenger to simulate natural water matrices. It is well known that humic substances can initiate, promote or inhibit radical reactions of O3 (Staehelin and Hoigne´, 1985). In the presence of bicarbonate alkalinity (1 mM), the degradation rate of both pesticides decreased as the humic acid concentration increased. Degradation levels of 88% and 72% from bromoxynil and 32% and 21% for trifluralin were observed at humic acid concentrations of 8.8 106 M and 22.0 106 M, respectively. The effect of bicarbonate alkalinity was studied at pH 7. The pesticide decreased from 89% to 47% and from 54% to 40% for bromoxynil and trifluralin, respectively, when increasing alkalinity from 0 to 5 mM as CaCO3. It is clear that
the scavenging effect of bicarbonate ions on the O3 radical chain reactions (Staehelin and Hoigne´, 1985) and the reaction of humic acid with molecular O3 decreased the availability of OH radicals available to react with the pesticides.
3.3.
Pesticide degradation in natural waters
The degradation of bromoxynil and trifluralin in natural water was examined using O3 and O3/H2O2 (see Table 2 for experimental conditions and levels of pesticide oxidation). Fig. 1 shows the degradation levels of these two pesticides in ultrapure water (UPW), river water (W1), and irrigation return flow water (W2). As shown, the rate of degradation was faster in ultrapure water, followed by the rate obtained using water W1. The lowest removals were observed in water W2. The higher levels of organic matter and the presence of bicarbonate ions may explain the lower degradation rate in water W2. As expected, the highest removals were achieved using O3/H2O2. Removals of 93% for bromoxynil and 88% for trifluralin were observed in ultrapure water using 2.1 105 M as an initial O3 concentration. In water W1, 60% and 48% removals were achieved for bromoxynil and trifluralin with 2.0 105 M of utilized O3 doses, whereas 69% removal for bromoxynil and 54% removal for trifluralin where observed in water W1 using O3/H2O2. Experiments performed using different initial O3 and H2O2 concentrations showed that increasing O3 concentrations resulted in higher removals. Fig. 2 shows the effect of different initial O3 concentrations on ozone decomposition and trifluralin oxidation in river water (W1) as an example. By increasing the initial O3 concentration from 4 105 M to 20 105 M, an additional 19% removal of trifluralin was achieved. About 10e15% of the pesticide degradations increased by adding H2O2 at a H2O2/O3 molar ratio of 0.5. As reported in several studies (Glaze et al., 1987; Acero and von Gunten, 2001; Can and C ¸ akir, 2010), an optimum H2O2/O3 ratio of 0.4e0.6 was found. Molar ratios lower than 0.5 can result in limited OH radical generation, while the excess of H2O2 competes with the organic constituents in the water for OH radicals (Glaze et al., 1987).
Table 2 e Experimental conditions and corresponding levels of pesticide degradation. Pesticide
Water matrix
Bromoxynil
UPW UPW W1 W1 W2 W2 UPW UPW W1 W1 W2 W2
Trifluralin
Initial alkalinity (mg/L as CaCO3)
Initial TOC (mg/L)
0 0 136 136 230 230 0 0 136 136 230 230
0 0 2.8 2.8 20.6 20.6 0 0 2.8 2.8 20.6 20.6
Initial H2O2 concentration (M) e 1 e 1 e 1 e 1 e 1 e 1
105 105 105 105 105 105
Pesticide degradation (%) 94 99 60 70 39 49 88 91 48 54 34 47
UPW: ultrapure water, W1: North Saskatchewan River water, W2: irrigation return flow discharging into the Redwater River; [bromoxynil]0 ¼ 3.6 106 M, [trifluralin]0 ¼ 3.0 106 M, applied ozone dose ¼ 2 105 M, reaction time ¼ 600 s.
2521
reduction of 56% was observed in the bromoxynil experiments using O3/H2O2 after 15 min of reaction and while 52% of TOC was reduced in the experiments using trifluralin samples. The low TOC reductions compared to the levels of pesticide oxidation may be explained by the assumption that these pesticides were degraded to smaller organic molecules still contributing to the TOC of the samples.
UPW
W1 Bromoxynil - O3 Bromoxynil - O3/H2O2 Trifluralin - O3 Trifluralin - O3/H2O2
W2
0
20
40
60
80
100
3.4. Toxicity and identification of major oxidation by-products 120
Pesticide Degradation (%)
Fig. 1 e Pesticide degradation in natural waters using ozonation and O3/H2O2 after 10 min of reaction (UPW: ultrapure water, W1: North Saskatchewan River water, W2: irrigation return flow discharging into the Redwater River). Experimental conditions: [bromoxynil]0 [ 3.6 3 10L6 M, [trifluralin]0 [ 3.0 3 10L6 M, [ozone]0 [ 2 3 10L5 M, [H2O2]0 [ 1 3 10L5 M, averaged utilized-ozone doses [ 1.7 3 10L5 M.
The results also indicate that the rate of O3 decay was slower in ultrapure water than those obtained in river water. Due to the level of dissolved organic matter in water W2, a higher O3 degradation (due to O3 auto-decomposition and the reactions with the precursors and by-products generated during treatment) rate was observed; almost 96% of the initial O3 residual was consumed in the first minute of reaction. At an O3 dose of 2 105 M, the O3 residual was consumed after 3 min of reaction. Total organic carbon is a required measurement in municipal water and wastewater systems for process control. From the experimental work, it was found that O3/H2O2 was more efficient than O3 alone in terms of TOC reduction (see Table 2 for the details of experimental conditions). TOC
1.2 Triflyralin oxidation - O3 = 4 E-05 M Trifluralin oxidation - O3 = 20 E-05 M O3 decay - O3 = 4 E-05 M O3 decay - O3 = 20 E-05 M
1.0 0.8 0.6
1.0 0.8 0.6
0.4
0.4
0.2
0.2
0.0
0.0 1000
[O3]t/[O3]0
[Trifluralin]t/[Trifluralin]0
1.2
Because the oxidation of an organic compound may lead to the formation of by-products, in some cases more toxic than their precursors, the use of toxicity measurements becomes an important tool to test the treatment efficiency. The acute toxicity of untreated and treated samples was measured using the Microtox bioassay. The toxicity of water samples on V. fischeri after 15 min of exposure based on the Microtox 81.9% screening test protocol are illustrated in Fig. 3. The toxic effects of the treated samples on the photobacterium decreased during the first minutes of reaction. Following that, an increase of the toxicity was noted. This increase in the sample toxicity may be due to the production of some oxidation by-products that were toxic to the photobacterium. To confirm this assumption, a set of experiments were performed to identify and quantify the oxidation by-products, comparing their structures with authentic standard compounds and examining the retention time and mass spectral data. When authentic standards were not available, the by-product identification was assigned by interpreting the mass spectrum. The results from HPLC-UV and MS analyses showed the presence of several peaks after oxidation of the water samples using O3/H2O2. Three compounds were identified as possible bromoxynil oxidation by-products, including 3-bromo-4,5hydroxy benzonitrile, 3-bromo-4-hydroxy benzonitrile, and
15 % Inhibitoin of Vibrio fischeri
Type of Water Matrix
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 1 7 e2 5 2 6
10 5 0 -5 -10 -15 -20
Bromoxynil - W1 - O3 Bromoxynil - W2 - O3/H2O2 Trifluralin - W1 - O3 Trifluralin - W2 - O3
-25 -30 -35 0
0
200
400
600
800
Time (s)
Fig. 2 e Effect of the initial ozone concentration on the ozone decomposition and pesticide oxidation for experiment using trifluralin in river water W1. Experimental conditions: [trifluralin]0 [ 3.0 3 10L6 M, [ozone]0 [ 4 and 20 3 10L5 M, water matrix [ North Saskatchewan River water.
200
400
600
800
1000
Time (s)
Fig. 3 e Toxic effects of bromoxynil and trifluralin samples on Vibrio fischeri using 81.9% screening test Microtox protocol (W1: North Saskatchewan River water, W2: irrigation return flow discharging into the Redwater River). Experimental conditions: [bromoxynil]0 [ 3.6 3 10L6 M, [trifluralin]0 [ 3.0 3 10L6 M, [ozone]0 [ 2 3 10L5 M, [H2O2]0 [ 1 3 10L5 M.
2522
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 1 7 e2 5 2 6
[By-product]t /[Bromoxynil]0
0.8
3-bromo-4,5-hydroxy benzonitrile
0.7
3-bromo-4-hydroxy benzonitrile 4-hydroxy benzonitrile
0.6
1.0 0.8
Bromoxynil
0.5
0.6
0.4 0.3
0.4
0.2
0.2
0.1 0.0 0
200
400
600
800
0.0 1000
Time (s)
Fig. 4 e Formation of bromoxynil by-products generated during O3/H2O2 treatment. Experimental conditions: [bromoxynil]0 [ 3.6 3 10L6 M, [ozone]0 [ 2.1 3 10L5 M, water matrix [ North Saskatchewan River water.
[Bromoxynil]t/[Bromoxynil]0
1.2
0.9
0.20
1.2 2,6 dinitro-4-trifluoromethyl aniline 4-trifluoromethyl aniline α,α,α-trifluoro-2,6-dinitro-N-dipropyl-p-toluidine Trifluralin
0.15
1.0 0.8
0.10
0.6 0.4
0.05 0.2 0.00
[Trifluralin]t/[Trifluralin]0
[By-product]t /[Trifluralin]0
4-hydroxy benzonitrile. As illustrated in Fig. 4, the major byproducts in the natural water were 3-bromo-4,5-hydroxy benzonitrile, 3-bromo-4-hydroxy benzonitrile, and 4-hydroxy benzonitrile, being the latest the predominant species, with a concentration equivalent to 52% of initial bromoxynil concentration. After 10 min of O3/H2O2 treatment, the concentrations of 3-bromo-4,5-hydroxy benzonitrile, 3bromo-4-hydroxy benzonitrile and 4-hydroxy benzonitrile were 7.6 108 M, 2.6 107 M, and 1.3 106 M, respectively. It is speculated that the decomposition pathway of bromoxynil may occur via debromination, leading to the formation of 3-bromo-4,5dihydroxybenzonitrile, 3-bromo-4-hydroxy benzonitrile, and 4-hydroxy benzonitrile. In the case of trifluralin, three compounds were identified as possible ozonation by-products, including a,a,a-trifluoro2,6-dinitro-N-dipropyl-p-toluidine, 2,6-dinitro-4-trifluoromethyl aniline, and 4-trifluoromethyl aniline. As shown in Fig. 5 and after 10 min of ozonation, the concentrations of a,a,a-trifluoro-2,6-dinitro-N-dipropyl-p-toluidine, 2,6-dinitro4-trifluoromethyl aniline, and 4-trifluoromethyl aniline were 10.8, 1.8, and 1.9% of the initial trifluralin concentration (2.9 106 M), respectively. The results of this study suggest that the oxidation by O3 and O3/H2O2 may proceed by dealkylation of trifluralin. Table 3 present an example of a total molar balance of the pesticides and their degradation byproducts. The total molar balances of the degradation byproducts versus the initial pesticide concentrations ranged from 60 to 103% under different experimental conditions. The lower values of the total molar balances can be attributed to the mineralization of the pesticides and some by-products and/or the formation of unidentified by-products, not accounted for the total molar balance. In a previous publication, these by-products were also identified after the oxidation of bromoxynil and trifluralin in ultrapure water after ozonation (Chelme-Ayala et al., 2010), but at higher concentrations. The higher efficiency of O3/H2O2 to degrade the pesticides and the by-products generated during treatment and the different experimental conditions used in the present study such as lower pesticides levels in the natural water matrices and different pHs may explain the lower concentrations of by-products.
0.0 0
200
400
600
800
Time (s)
Fig. 5 e Formation of by-products generated during the oxidation of trifluralin by O3/H2O2. Experimental conditions: [trifluralin]0 [ 3.0 3 10L6 M, [ozone]0 [ 2.1 3 10L5 M, water matrix [ North Saskatchewan River water.
3.5.
Kinetic study in natural waters
The overall O3 oxidation reaction may proceed through two pathways, including direct reaction by molecular O3 and indirect reaction by OH radical (Hoigne´ and Bader, 1976). Using a batch reaction of ozonation process, the elimination of a pollutant (P) that reacts with both O3 and OH radicals is described by second-order kinetics:
d½P ¼ kO3 P ½O3 ½P þ k$OHP ½$OH½P dt
(1)
where kO3 P and k$OHP are second-order reaction rate constants of the micropollutant (P) with O3 and OH radical, respectively. To predict the degradation of bromoxynil and trifluralin, the procedure recommended by Elovitz and von Gunten (1999) was followed. In this approach, the evolution of the target compound is estimated based on the rate constants and the transient O3 and OH radical concentrations. The rate of O3 decomposition in natural water can be calculated assuming first order kinetics (Elovitz and von
Table 3 e Example of total molar balance of the pesticides and their degradation by-products. Compound Residual bromoxynil 3-bromo-4,5-hydroxy benzonitrile 3-bromo-4-hydroxy benzonitrile 4-hydroxy benzonitrile Total molar balance Residual trifluralin a,a,a-trifluoro-2,6-dinitro-N-dipropyl-p-toluidine 2,6 dinitro-4-trifluoromethyl aniline 4-trifluoromethyl aniline Total molar balance
Molar balance (%) 32.3 1.7 7.1 48.9 90.0 47.6 1.6 2.6 8.7 60.5
Experimental conditions: [Bromoxynil]0 ¼ 3.6 106 M, [trifluralin]0 ¼ 3.0 106 M, [ozone]0 ¼ 2.0 to 4.0 105 M, water matrix ¼ North Saskatchewan River water.
2523
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 1 7 e2 5 2 6
of a compound depends on the relationship between O3 expoR R sure and OH exposure (Rct ¼ ½$OHdt= ½O3 dt) and the rate constants. The O3 exposure can be estimated from the integration of the O3 concentration over time. To measure the OH concentration, a probe compound, pCBA, was used. The probe compound levels in the solution must be very low in order to avoid scavenging effects on OH radical. The Rct parameter can be estimated as follows:
0
ln [O3]/[O3]0
-2 -4 -6 Bromoxynil - W1 - O3 Bromoxynil - W1 - O3/H2O2 Trifluralin - W2 - O3 Trifluralin - W2 - O3/H2O2
-8 -10
d½R ¼ k$OHR ½$OH½R dt
-12 0
200
400
600
800
1000
Reaction Time (s)
Fig. 6 e Determination of first-order rate constant for ozone decomposition in natural waters. Experimental conditions: [bromoxynil]0 [ 3.6 3 10L6 M, [trifluralin]0 [ 3.0 3 10L6 M, [ozone]0 [ 1.0 3 10L4 M, [pCBA]0 [ 3.0 3 10L6 M.
Gunten, 1999; Benı´tez et al., 2007). Thus, the depletion of O3 in natural waters can be represented as follows: ½O3 ¼ kt ln ½O3 0
(2)
where [O3]0 and [O3]t are the concentrations of O3 residual at initial and any reaction time (t), and k is the first-order O3 decay rate constant. A plot of semi-log versus time allows the determination of k. The depletion of O3, in water W1 and water W2, is presented in Fig. 6. As shown in this figure, two periods were observed in the experiments performed using ozonation: an initial period of very fast O3 decomposition, in which O3 reacts rapidly with oxidizable compounds, followed by a period of slow O3 decrease, in which O3 reacts slowly with more recalcitrant compounds. This trend has been previously reported in other studies (Elovitz and von Gunten, 1999; Buffle et al., 2006; Benı´tez et al., 2007). Because the O3 residual in experiments using O3/H2O2 disappeared within 5 min of reaction, one period of decomposition was observed. In the reaction kinetic modelling approach proposed by Elovitz and von Gunten (1999), instead of using the elementary reactions, the prediction of the time-dependent transformation
½R ln t ¼ k$OHR Rct ½R0
(3)
Zt ½O3 dt
(4)
0
where [R]0 and [R]t represent the concentration of the probe compound at initial and any reaction time, k$OHR is the second-order rate constant for the reaction between the probe compound and OH radical, and ![O3]dt is the O3 exposure. A plot of ln [R]t/[R]0 versus O3 exposure leads to the estimation of Rct. It has been found that the Rct parameter depends on water quality characteristics, including pH, temperature, alkalinity, organic matter concentrations, and levels of scavengers promoting or inhibiting the OH radical reaction (Elovitz and von Gunten, 1999). Similar to the two stages found during the O3 decomposition in natural waters, two separate kinetics phases were observed, the first stage related to the rapid O3 decay. Table 4 summarizes the first-order rate constants for O3 depletion and Rct values for the initial period of very fast O3 consumption (k1 and Rct1) and the same values for the second stage (k2 and Rct2) under different treatment options. As shown in Table 4, the highest Rct values were recorded during the initial reaction stage. As illustrated, Rct values of 2.47 108 and 1.83 108 were estimated for the ozonation of bromoxynil in water W1 during the initial and second reaction stages, respectively. In general, the Rct parameter increases with decreasing carbonate concentration and increasing natural organic matter concentration. As well, Rct values are higher for O3/ H2O2 processes than for conventional ozonation because of the higher OH exposure. By considering the Rct parameter, the Equation (1) can be rewritten as:
Table 4 e First-order rate constants for ozone decay and Rct values for the first (fast ozone consumption) and second period (slow ozone decay) of oxidation in experiments using natural waters. Pesticide Bromoxynil
Trifluralin
Water type W1 W1 W2 W2 W1 W1 W2 W2
Treatment (dose, M) O3 O3 O3 O3 O3 O3 O3 O3
(1.0 (1.0 (1.0 (1.0 (0.8 (0.8 (0.8 (0.8
4
10 ) 104)/H2O2 (0.5 104) 104)/H2O2 (0.5 104) 104)/H2O2 (0.4 104) 104)/H2O2 (0.4
104) 104) 104) 104)
k1 (s1) 3.2 5.8 3.3 4.1 1.2 4.2 1.5 4.0
2
10 102 102 102 102 102 102 102
k2 (s1) 9.8 10
4
7.4 104 3.2 103 3.3 103
Rct1 2.5 1.3 2.4 5.0 1.6 1.1 2.7 6.5
Rct2 8
10 107 108 108 108 107 108 108
1.8 108 1.2 108 8.5 109 5.5 109
W1: North Saskatchewan River, W2: Irrigation return flow discharging into the Redwater River. Experimental conditions: [bromoxynil]0 ¼ 3.6 106 M, [trifluralin]0 ¼ 3.0 106 M, [pCBA]0 ¼ 3.6 106 M for bromoxynil experiments and 3.0 106 M for trifluralin experiments.
2524
d½P ¼ kO3 P ½O3 ½P þ k$OHP Rct ½O3 ½P dt ½P ¼ kO3 P þ k$OHP Rct ln ½P0
(5)
Zt ½O3 dt
(6)
0
By using Equation (6), and the kO3 P and k$OHP values listed in Table 1, and the Rct values given in Table 4 is possible to predict the evolution of pesticide concentrations during O3 and O3/H2O2 processes. As an example, Fig. 7 shows the experimental data and the model predictions for the degradation of bromoxynil and tifluralin in natural waters. In this figure, the symbols represent the experimental data, and the lines represent the model predictions. The results showed that the kinetic models yielded accurate predictions for the natural waters tested. Fig. 8 illustrates the fraction of the pesticides reacting with the OH radicals as function of Rct. Although the molecular O3 and OH radical pathways were important in the pesticide degradation, the contribution of OH radical was more significant in experiments using O3/H2O2. In experiments using ozonation, the reaction of molecular O3 with the pesticides accounted for about 48e53% of the pesticide degradation. In the O3/H2O2 treatment, the OH pathway was more dominant, whereas the molecular O3 contribution for the pesticide degradation was about 20%.
3.6.
Engineering applications
Kinetic models are very useful in the design of treatment plants because of their ability to predict the influence of several parameters on the oxidation processes. Models based on reaction mechanisms, mass balance and empirical selectivity have been used to predict the performance of the oxidation process during ozonation (Glaze and Kang, 1988; Gottschalk et al., 2000). The procedure proposed by Elovitz and von Gunten (1999) has been successfully used to predict the degradation of organic compounds such as N-nitrosodimethylamine (NDMA), and some pesticides including
[Pesticide]t/[Pesticide]0
1.2 Bromoxynil - W1 - O3 Bromoxynil - W2 - O3 Trifluralin - W2 - O3 Model
1.0 0.8
Fraction Reacting with •OH Radicals
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 1 7 e2 5 2 6
1.0 0.8 0.6 0.4 Bromoxynil Trifluralin
0.2 0.0 1.00E-09
5.10E-08
1.01E-07
1.51E-07
Rct
Fig. 8 e Fraction of the pesticides reacting with the hydroxyl radicals as function of Rct. Experimental conditions: [bromoxynil]0 [ 3.6 3 10L6 M, [trifluralin]0 [ 3.0 3 10L6 M.
metolachlor, diuron and isoproturon (Acero et al., 2003; Lee et al., 2007b; Benı´tez et al., 2007). This approach is easy to apply and requires reactors operating at batch mode. However, it is important to keep in mind that the models developed by this method are valid only for certain oxidation conditions. The Rct parameter that takes into account the characteristics of water matrix is very specific for a given water; any change in the composition of the water requires a new Rct determination. The results of this study demonstrate that O3 combined with H2O2 is an effective treatment process for the degradation of pesticides. However, special attention has to be paid to the formation of by-products. This requires the application of higher oxidant doses, able to degrade not only the target compounds but also the oxidation by-products generated during treatment. From this experimental work, it was found that O3 concentrations higher than 5 104 M were required to degrade the pesticides and the by-products formed during oxidation. The results of this study also indicate that the oxidation by-products may cause toxicological risk if they are not further oxidized. When using advanced oxidation processes, additional and more comprehensive tests such as fish toxicity (Kerr et al., 2008; Stalter et al., 2010) are recommended to assess the efficiency of the process in terms of toxicity reduction.
0.6
4.
0.4 0.2 0.0 0
500
1000
1500
2000
Reaction Time (s)
Fig. 7 e Experimental data and predicted pesticideoxidation level. Experimental conditions: [bromoxynil]0 [ 3.6 3 10L6 M, [trifluralin]0 [ 3.0 3 10L6 M, [ozone]0 [ 2 3 10L5 M.
Conclusion
Ozone and the combination of O3 and H2O2 were used to oxidize bromoxynil and trifluralin in natural waters. Degradation levels lower than 50% for both pesticides were found during conventional ozonation. An enhancement of the level of degradation was observed using O3/H2O2 process. Although the molecular O3 and OH radical pathways were important in the pesticide degradation, the contribution of OH radical was more significant in experiments using O3/H2O2. An increase of the acute toxicity of the treated samples towards V. fischeri was noted at the end of treatment at different experimental
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 1 7 e2 5 2 6
conditions. It was also found that O3 concentrations higher than 5 104 M were required to degrade the pesticides and the by-products formed during oxidation.
Acknowledgements The authors gratefully acknowledge the financial support provided by the Natural Sciences and Engineering Research Council of Canada (NSERC) and Alberta Ingenuity Center for Water Research (AICWR), now known as the Alberta Water Research Institute (AWRI) which is a part of Alberta Innovates. The authors also appreciate the technical cooperation provided by the technical staff of the Environmental Engineering Program at the University of Alberta and the technical support provided by the Environmental Engineering Group at the Missouri University of Science and Technology.
references
Acero, J.L., von Gunten, U., 2001. Characterization of oxidation processes: ozonation and the AOP O3/H2O2. J. Am. Water Works Assoc. 93 (10), 90e100. Acero, J.L., Benı´tez, F.J., Real, F.J., Maya, C., 2003. Oxidation of acetamide herbicides in natural waters by ozone and by the combination of ozone/hydrogen peroxide: kinetic study and process modeling. Ind. Eng. Chem. Res. 42 (23), 5762e5769. Alberta Environment, 2001. Pesticide Use in Alberta 1998. Alberta Environment, Government of Alberta, Edmonton, Alberta, Canada. Anderson, A.M., 2005. Overview of Pesticide Data in Alberta Surface Waters Since 1995. Environmental Monitoring and Evaluation Branch, Publication No. T/772. Alberta Environment, Edmonton, Alberta, Canada. American Public Health Association (APHA), 2005. Standard Methods for the Examination of Water and Wastewater, twentyfirst ed. American Water Works Association and Water Environment Federation, Washington, DC, USA. Benı´tez, F.J., Real, F.J., Acero, J.L., Garcia, C., 2007. Kinetics of the transformation of phenyl-urea herbicides during ozonation of natural waters: rate constants and model predictions. Water Res. 41 (18), 4073e4084. Buffle, M.O., Schumacher, J., Salhi, E., Jekel, M., von Gunten, U., 2006. Measurement of the initial phase of ozone decomposition in water and wastewater by means of a continuous quench-flow system: application to disinfection and pharmaceutical oxidation. Water Res. 40 (9), 1884e1894. 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. J. Phys. Chem. Ref. Data 17 (2), 513e886. Byrtus, G., Anderson, A.M., Saffran, K., Bruns, G., Checknita, L., 2002. Determination of New Pesticides in Alberta’s Surface Waters (1999e2000). The Water Research User Group, Publication. No. T/674. Alberta Environment, Edmonton, Alberta, Canada. Can, S.Z., C¸akir, E., 2010. Treatability of organic constituents in the Pas‚ako¨y wastewater treatment plant effluent by O3 and O3/H2O2. Ozone Sci. Eng. 32 (3), 209e214. Catalkaya, E.C., Kargi, F., 2009. Degradation and mineralization of simazine in aqueous solution by ozone/hydrogen peroxide advanced oxidation. J. Environ. Eng. ASCE 135 (12), 1357e1364.
2525
Chelme-Ayala, P., Gamal El-Din, M., Smith, D.W., 2010. Kinetics and mechanism of the degradation of two pesticides in aqueous solutions by ozonation. Chemosphere 78 (5), 557e562. Chiron, S., Fernandez-Alba, A., Rodriguez, A., Garcia-Calvo, E., 2000. Pesticide chemical oxidation: state-of-the-art. Water Res. 34 (2), 366e377. Donald, D.B., Cessna, A.J., Sverko, E., Glozier, N.E., 2007. Pesticides in surface drinking-water supplies of the Northern Great Plains. Environ. Health Perspect. 115 (8), 1183e1191. Elovitz, M.S., von Gunten, U., 1999. Hydroxyl radical/ozone ratios during ozonation processes. I. The Rct concept. Ozone Sci. Eng. 21 (3), 239e260. Gisi, D., Stucki, G., Hanselmann, K.W., 1997. Biodegradation of the pesticide 4,6-dinitro-ortho-cresol by microorganisms in batch cultures and in fixed-bed column reactors. Appl. Microbiol. Biotechnol. 48 (4), 441e448. Glaze, W.H., Kang, J.W., Chapin, D.H., 1987. The chemistry of water treatment processes involving ozone, hydrogen peroxide and ultraviolet radiation. Ozone Sci. Eng. 9 (4), 335e352. Glaze, W.H., Kang, J.W., 1988. Advanced oxidation processes for the treatment of groundwater with TCE and PCE: laboratory studies. J. Am. Water Works Assoc. 80 (5), 57e63. Gottschalk, C., Libra, J.A., Saupe, A., 2000. Ozonation of Water and Waste Water. A Practical Guide to Understanding Ozone and Its Application. Wiley-Vch, Weinheim, Federal Republic of Germany. Greene, S.A., Pohanish, R.P., 2005. Sittig’s Handbook of Pesticides and Agricultural Chemicals. William Andrew Publishing, NY, USA. Hoigne´, J., Bader, H., 1976. The role of hydroxyl radical reactions in ozonation processes in aqueous solutions. Water Res. 10 (5), 377e386. Hoigne´, J., Bader, H., 1983. Rate constants of reactions of ozone with organic and inorganic compounds in water. 1. Water Res. 17 (2), 173e183. Ikehata, K., Gamal El-Din, M., 2005a. Aqueous pesticide degradation by ozonation and ozone-based advanced oxidation processes: a review (part I). Ozone Sci. Eng. 27 (2), 83e114. Ikehata, K., Gamal El-Din, M., 2005b. Aqueous pesticide degradation by ozonation and ozone-based advanced oxidation processes: a review (part II). Ozone Sci. Eng. 27 (3), 173e202. Kaushik, A., Sharma, H.R., Jain, S., Dawra, J., Kaushik, C.P., 2010. Pesticide pollution of river Ghaggar in Haryana, India. Environ. Monit. Assess. 160 (1e4), 61e69. Kearney, P.C., Muldoon, M.T., Somich, C.J., 1987. UV-Ozonation of eleven major pesticides as a waste disposal pretreatment. Chemosphere 16, 2321e2330. Kerr, J.L., Guo, Z., Smith, D.W., Goss, G.G., Belosevic, M., 2008. Use of goldfish to monitor wastewater and reuse water for xenobiotics. J. Environ. Eng. Sci. 7 (4), 369e383. Kochany, J., Choudhry, G.G., 1990. Photochemistry of halogenated benzene derivatives. Part XII. Effects of sodium nitrite on the environmental phototransformation of bromoxynil (3,5Dibromo-4-Hydroxy benzonitrile) herbicide in water: the photoincorporation of nitrite ions into bromoxynil. Toxicol. Environ. Chem. 27 (4), 225e239. Laetz, C.A., Baldwin, D.H., Collier, T.K., Hebert, V., Stark, J.D., Scholz, N.L., 2010. The synergistic toxicity of pesticide mixtures: implications for risk assessment and the conservation of endangered Pacific salmon. Environ. Health Perspect. 117 (3), 348e353. Lee, C., Schmidt, C., Yoon, J., von Gunten, U., 2007a. Oxidation of N-nitrosodimethylamine (NDMA) precursors with ozone and chlorine dioxide: kinetics and effect on NDMA formation potential. Environ. Sci. Technol. 41 (6), 2056e2063.
2526
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 1 7 e2 5 2 6
Lee, C., Yoon, J., von Gunten, U., 2007b. Oxidative degradation of N-nitrosodimethylamine by conventional ozonation and the advanced oxidation process ozone/hydrogen peroxide. Water Res. 41 (3), 581e590. Leitner, N.K.V., Roshani, B., 2010. Kinetic of benzotriazole oxidation by ozone and hydroxyl radical. Water Res. 44 (6), 2058e2066. Mattice, J.D., Skulman, B.W., Norman, R.J., Gbur, E.E., 2010. Analysis of river water for rice pesticides in eastern Arkansas from 2002 to 2008. J. Soil Water Conserv. 65 (2), 130e140. Miltner, R.J., Baker, D.B., Speth, T.F., Fronk, C.A., 1989. Treatment of seasonal pesticides in surface waters. J. Am. Water Works Assoc. 81 (1), 43e52. Navarro, A., Tauler, R., Lacorte, S., Barcelo´, D., 2010. Occurrence and transport of pesticides and alkylphenols in water samples along the Ebro river basin. J. Hydrol. 383 (1e2), 18e29. Neumann, M., Schulz, R., Schafer, K., Muller, W., Mannheller, W., Liess, M., 2002. The significance of entry routes as point and non-point sources of pesticides in small streams. Water Res. 36 (4), 835e842. Norgaard, K.B., Cedergreen, N., 2010. Pesticide cocktails can interact synergistically on aquatic crustaceans. Environ. Sci. Pollut. Res. 17 (4), 957e967. Ormad, M.P., Miguel, N., Claver, A., Matesanz, J.M., Ovelleiro, J.L., 2008. Pesticides removal in the process of drinking water production. Chemosphere 71, 97e106. Rodriguez, A., Garcia, J., Sotelo, J.L., Ovejero, G., Mestanza, M., 2009. Removal of the pesticides diuron and carbofuran in
aqueous solutions by activated carbon. Fresen. Environ. Bull. 18 (11), 2093e2101. Sanches, S., Barreto, M., Pereira, V.J., 2010. Direct water treatment of priority pesticides using low pressure UV photolysis and advanced oxidation. Water Res. 44 (6), 1809e1818. Sa´nchez-Polo, M., Rivera-Utrilla, J., Zaror, C.A., 2002. Advanced oxidation with ozone of 1,3,6-naphthalenetrisulfonic acid in aqueous solution. J. Chem. Technol. Biotechnol. 77 (2), 148e154. Sipyagin, A.M., Enshov, V.S., Kashtanov, S.A., Bateman, C.P., Mullen, B.D., Tan, Y.T., Thrasher, J.S., 2004. New 4-pentafluorosulfanyl and 4-perfluoroalkylthio derivatives of 1-chloro-2-nitro- and 1-chloro-2,6-dinitrobenzenes. J. Fluor. Chem. 125 (9), 1305e1316. Staehelin, J., Hoigne´, J., 1985. Decomposition of ozone in water in the presence of organic solutes acting as promoters and inhibitors of radical chain reactions. Environ. Sci. Technol. 19 (12), 1206e1213. Stalter, D., Magdeburg, A., Weil, M., Knacker, T., Oehlmann, J., 2010. Toxication or detoxication? in vivo toxicity assessment of ozonation as advanced wastewater treatment with the rainbow trout. Water Res. 44 (2), 439e448. U.S. EPA, 1998. Registration Eligibility Decision (RED) Facts e Bromoxynil. U.S. Environmental Protection Agency. Available from: http://www.epa.gov (accessed July 2010). Yao, C.C.D., Haag, W.R., 1991. Rate constants for direct reactions of ozone with several drinking-water contaminants. Water Res. 25 (7), 761e773.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 2 7 e2 5 3 8
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Identification and phylogeny of the small eukaryote population of raw and drinking waters Eli Otterholt*, Colin Charnock Department of Health Sciences, Oslo University College, Pilestredet 46, 0167 Oslo, Norway
article info
abstract
Article history:
Culture-dependent and -independent methods were used to investigate the small
Received 9 July 2010
eukaryote composition of raw and finished waters in the Norwegian cities of Oslo, Tromsø,
Received in revised form
Fredrikstad and Oppega˚rd. Probes with general applicability to the 18S rRNA genes of the
4 February 2011
small eukaryote consortium were used for PCR-denaturing gradient gel electrophoresis
Accepted 7 February 2011
(DGGE), and in the generation of clone libraries using the TOPO cloning and sequencing
Available online 17 February 2011
system. The chosen probes invariably gave a single band in agarose gel electrophoresis, indicating amplification of an area of similar size. DGGE and cloning analyses resolved the
Keywords:
bands into components representing many unique amplicons. Diversity and composition
18S rRNA
in the collection were studied by DNA-sequencing, and visual examination of DGGE
Drinking water
patterns. The cloning approach enabled the putative identification of a total of approxi-
Eukaryote diversity
mately 100 unique small eukaryotes. The major fraction of these represented ciliated and
DGGE
flagellated protozoal species. This was in keeping with the findings from protozoal culti-
Clone libraries
vation. DNA from a number of multicellular eukaryotes was also detected. Amoebal and fungal DNA was rarely found. The latter may indicate a low incidence or a bias in the analysis technique. The population of small eukaryotes appears typical for pristine waters and no primary pathogens were detected by culture-independent techniques. However, the potentially pathogenic protozoa Acanthamoeba castellanii was grown on one occasion from Oslo’s drinking water. DGGE allowed the identification of fewer amplicons (by excision and sequencing of bands) than by the cloning-transformation approach. The DGGE analysis revealed clear similarities between the compositions of the raw and treated waters, indicating that cells or DNA in the raw water pass through the treatment trains. Protozoal culture and heterotrophic plate count analysis consistently revealed viable cells in both raw and treated waters in Oslo. This indicates that a fraction of the clone library represents eukaryotic species surviving the treatment trains. The analyses here presented represent the first published study of the general small eukaryotic fraction of the Capital’s drinking water, and those of three other Norwegian cities. We suggest that DGGE profiles may have a value in judging physical treatment efficacy (removal of cells), but that direct cloning and sequencing studies is more amenable for characterization of uncultured microbes. ª 2011 Elsevier Ltd. All rights reserved.
* Corresponding author. Tel.: þ4722452358; fax: þ47222452305. E-mail address:
[email protected] (E. Otterholt). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.02.008
2528
1.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 2 7 e2 5 3 8
Introduction
In addition to its direct use as a thirst quencher, potable water is also used in the production of food, beverages and medicines. Water quality thus impinges on our health. The European drinking water directive (CEC, 1998) includes water quality parameters based on the detection of indicator bacteria. Testing for eukaryotic microbes is not specified. Monitoring of pathogenic eukaryotic microbes is implicit in the stated requirement that drinking water should not contain biological components associated with health risks. However, strategies for meeting this requirement are not specified. In addition, the potential for a breach of hygienic barriers by protozoal cysts are only met indirectly by the testing for the endospore-forming Clostridium perfringens (CEC, 1998). Endospores produced by this species are presumably expected to meet a similar fate to parasites and viruses during water bacterial treatment. However, the indicator organism pathogen paradigm has been challenged (Harwood et al., 2005). Testing for indicator bacteria such as Escherichia coli, is primarily an investigation of the contamination of water with biological matter, particularly fecal material, which might contain primary pathogens. Such bacterial indicators might thus give information on the potential for contamination by fecal protozoa such as Giardia and Cryptosporidium. However, a number of protozoa as well as fungi associated with disease, such as the yeast species Candida and Cryptococcus, certain molds and dermatophytes, do not belong to the fecal contentdisease paradigm. Among the small eukaryotes, fungi are the most amenable to detection using culture techniques. Two recent studies of the mold content of Norwegian drinking waters based chiefly on cultivation techniques, showed a wide range of species, some of which are possible pathogens (Hageskal et al., 2006, 2007). No similar studies of yeasts or protozoa appear to have been performed. With the exception of fungi and some protozoa (eg, several amoebae), small eukaryotes are not easily grown as pure cultures. Gross structural similarities and a lack of biochemical rapid identification systems suitable for environmental strains, make this approach to identification unreliable. Numerous studies documenting a myriad of nucleotide sequences in water and other environmental samples that do not correspond to cultivated phenotypes, suggest that only a small fraction of microorganisms can be grown under normal laboratory conditions (Dorigo et al., 2005 and references therein). PCR using total or taxon-specific probes in combination with separation techniques such as DGGE and molecular cloning, has revolutionized our understanding of the composition of environmental samples. In particular, nucleotide fingerprinting approaches are enabling an identification and taxonomy of the small eukaryotic biota in water (Dorigo et al., 2005; Moon-van der Staay et al., 2001; Valster et al., 2009; Otterholt and Charnock, 2011). Cloning and sequencing of rRNA genes has shown itself to be a robust, reproducible and reliable indicator of microbial taxonomy (Diez et al., 2001 and references therein). DGGE has been widely used for the separation and visualization of PCRamplified genes (chiefly those coding for 16S rRNA) creating a community profile, and also enabling sequence analysis of
some of the individual bands (Diez et al., 2001). Direct ligation of amplified genes into cloning and sequencing vectors, does not provide an immediate visual representation of the community structure such as that obtained with DGGE. However, the problems related to the adequate separation and isolation of DGGE bands prior to sequencing such as overlapping bands, bands close together and the need for band excision is avoided with the direct cloning approach. Although DGGE and direct cloning have been used to characterize the microbial composition of several environments, few studies of drinking water have been published (Dewettinck et al., 2001; Hoefel et al., 2005). Free living protozoa are ubiquitous in natural freshwater environments, but also proliferate in engineered watersystems. These organisms can enter, pass through and even reproduce inside water treatment plants by colonizing granular activated carbon filter media (Schreiber et al., 1997). Genera of Free living protozoa commonly observed in these systems and in tap water installations include Acanthamoeba, Echinamoeba, Hartmannella, Platyamoeba, Vahlkampfia and Vannella (Valster et al., 2009, and references therein). Some Free living amoebae such as Acanthamoeba (Rowbotham, 1980) and Hartmannella (Thomas et al., 2006, and references therein) can in addition to their own pathogenicity, also act as Trojan horses for bacteria involved in human infections including Legionella (Thomas et al., 2006). Acanthamoeba spp. have also been associated with keratitis in persons wearing contact lenses and meningoencephalitis (Marciano-Cabral and Cabral, 2003). A recent culture-independent study of protist diversity in freshwater environments found that the majority of phylotypes at the 97% level affiliated within a few well established eukaryotic kingdoms or phyla, including alveolates, cryptophytes, heterokonts, Cercozoa, Centroheliozoa and haptophytes (Slapeta et al., 2005). A few sequences did not display a clear taxonomic affiliation. Studies have shown that the small eukaryotic assemblage is very diverse with many hitherto undiscovered taxa (Diez et al., 2001; Moon-van der Staay et al., 2001; Slapeta et al., 2005). However, the diversity of eukaryotes present in drinking water environments has only recently been examined (Poitelon et al., 2009; Slapeta et al., 2005; Valster et al., 2009). Thus further work may uncover new eukaryotic lineages and provide a framework for an evaluation of the true distribution and health significance of these species in potable waters. The aim of the present study was to analyze the diversity and composition of the small eukaryotic community in Norwegian drinking and source raw waters by use of culture and culture-independent techniques. Results are also considered in light of the differing water treatment trains at the installations studied. To our knowledge, no similar studies have been performed on Norwegian drinking waters.
2.
Materials and methods
2.1.
Water samples
Raw and treated waters prior to distribution from the Norwegian cities of Oslo (lake water treated by either coagulation/ flocculation/sedimentation and UV-disinfection (new plant)
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 2 7 e2 5 3 8
2529
or microsieving and chlorination (reserve plant)), Tromsø (lake water treated by sieving and chlorination e distribution net sample), Fredrikstad (lake water treated by coagulation/ flocculation/sedimentation, sand filtration and chlorination) and Oppega˚rd (lake water treated by coagulationeflotationefiltration and chlorination) were included in the study. In addition, a sample from Oslo University College, a point on the distribution net about 10 km from the plant, were included in the cloning and cultivation studies. Samples for testing were collected in sterile 3.0 l bottles and transported directly to the laboratory. Samples were maintained at 4 C during transport and were analyzed (filtration for DNA extraction, heterotrophic plate counts and protozoa culture) within 24 h of collection.
membrane filters (Gelman membranes, Pall Corporation, New York). The filters were frozen and stored at 20 C until further analysis. Extraction of total community DNA from the filters was performed by using the alternative protocol in the UltraClean Soil DNA Isolation kit (Mo BIO Laboratories, Inc, Solana Beach, California) following modifications described by others (Winter et al., 2007). According to the manufacturer the DNA extraction kit efficiently lyses spores and fungi. In brief, the filters were cut into approximately 20 pieces using sterile scissors and transferred into the supplied DNA extraction tubes. To achieve better wetting, the tubes were the shaken for 10 min at 250 rpm prior to DNA extraction as described. This gave typically 5e10 ng/ml DNA with an A260/A280 ratio of 1.5e1.7.
2.2.
2.5.
Heterotrophic plate count (HPC)
In order to supplement the DGGE and cloning data (non-culture based techniques) the HPC was determined by the surface inoculation technique using 0.1 ml samples spread in triplicate on R2A agar (Oxoid, Basingstoke, UK). The plates were incubated at 22 2 C in the dark. The HPC was read after 7 d.
2.3.
Isolation and culture of Free living protozoa
In order to supplement the DGGE and cloning data (nonculture based techniques), an attempt was made to grow, describe and where possible isolate protozoa from samples (Oslo). Amoebae species were chosen for special attention as indicator organisms because some of these are pathogens. Amoebae strains were isolated and identified based on 18S rRNA studies. Duplicate samples of 500 ml were filtered slowly (low pressure) to reduce sheer forces through a 0.45 mm poresize filter. Filters were then placed sample side down onto non-nutrient agar (NNA) having the following composition (g/L): 24.0 g NaCl, 0.8 g MgSO4$7H2O, 1.2 g CaCl2$6H2O, 28.4 g Na2HPO4, and 27.2 g KH2PO4 in 1 l of distilled water (Page’s Amoebae Saline Solution, PAS) amended with 15 g bacteriological agar (Oxoid, UK). All media were autoclaved for 15 min at 121 C. The NNA plates were seeded with a thick suspension (>McFarland 5) of E. coli ATCC 25922. The plates were sealed with tape, placed in a humidified atmosphere in the dark at 22 2 C and observed over a period of 2 weeks. Protozoa were analyzed by light microscopy and described particularly with regard to their shape, inclusions and mode of mobility. Samples containing amoebae were resolved by serial dilution in PAS to obtain theoretically w1 cell (per sample) which was subsequently transferred to E. coli-seeded NNA. For DNA isolation prior to sequencing, cultures were enriched by pelleting of amoebae cells and discarding of the E. coli-containing supernatant. This involved centrifugation at 500 g for 5 min, discarding of the supernatant and resuspension of amoebae in PAS a total of 3e4 times. DNA for PCR amplification was purified using the Blood protocol of the Blood and tissue Genomic DNA Extraction Miniprep system (Viogene, Taiwan).
2.4.
Direct DNA extraction from water samples
Water samples (350 ml raw water and 2500 ml finished water) were filtered through 47 mm diameter, 0.45 mm pore-sized
PCR amplification
Partial 18S rRNA gene sequences were amplified for DGGE analyses or directly for cloning/sequencing using PCR primer sets A and B (Table 1). Primer set A corresponds to bases 576e596 and 1147e1166 of the 18S rRNA gene of Tetrahymena setosa (AF364041.1), whereas primer set B corresponds to bases 1396e1411 and 1600e1614. Primer set C was used for sequencing of cloned fragments. All primers were obtained from Eurofins MWG-Biotech (Ebersberg, Germany) and the PCR reactions were performed in a Palm-Cycler (Corbett Life Science, Sydney). PCR reaction mixtures and thermal cycling were performed under the conditions described in the referenced articles (Table 1). Primer set A gives an expected fragment size of approximately 600 bp. In brief, for primer set A the PCR mixture (50 ml) contained: 5e10 ng template, 1x PCR Buffer, 2 mM MgCl2, 0.2 mM dNTPs (all SigmaeAldrich, St. Louis, MO), 2.5 U of Taq DNA polymerase (Sigma enzyme, or Go Taq Hot Start polymerase, Promega Corporation, Madison) and a primer concentration of 0.167 mM each. After a denaturation step of 5 min at 94 C, forty cycles of amplification were performed by using 94 C for 1 min, 55 C for 30 s and 72 C for 2 min, followed by a final extension at 72 C for 10 min. Primer set B produces a product of approximately 210 bp. For DGGE a 40 bp GC-clamp was attached to the 30 -end of primer Euk1209f. The reaction mix (50 ml) contained 2e10 ng template, 200 mM dNTPs (SigmaeAldrich), 1.5 mM MgCl2 (Promega), each primer at a concentration of 0.3 mM, 2.5 U Go Taq Hot Start polymerase (Promega) and the PCR buffer supplied with the enzyme. After an initial denaturation step of 1 min at 94 C and 10 touchdown cycles at 94 C for 1 min, annealing at 65 C for 1 min, and extension at 72 C for 3 min, 20 cycles of amplification were performed by using 94 C for 1 min, 55 C for 1 min and 72 C for 3 min followed by a final additional extension at 72 C for 4 min.
2.6.
DGGE
The PCR amplicons produced with primer set B were separated by DGGE using the D-Code Universal Mutation Detection system (Bio-Rad Laboratories Inc., Hercules, California). The DGGE procedure was based on a modification of the protocol of Diez et al. (2001). In brief, approximately 6 mg PCR product was loaded into each well (i.e. representing equal amounts of
2530
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 2 7 e2 5 3 8
Table 1 e Overview of the oligonucleotide sequences used in the study. Primer set
Primer name
Sequence (50 e30 )
Specificity
Reference
A
Ami6F1 Ami6F2 Ami 9R Euk1209f (GC)a Uni1392r M13 Forward M13 Reverse
CCAGCTCCAATAGCGTATATT CCAGCTCCAAGAGTGTATATT GTTGAGTCGAATTAAGCCGC CAGGTCTGTGATGCCC ACGGGCGGTGTGTRC GTAAAACGACGGCCAG CAGGAAACAGCTATGAC
Eukarya Eukarya Eukarya Eukarya Universal Plasmid linker region Plasmid linker region
Thomas et al., 2006 Thomas et al., 2006 Thomas et al., 2006 Diez et al., 2001 Diez et al., 2001 TOPO, Promega. TOPO, Promega.
B C
Primer set A and B was used for the generation of PCR products for cloning in TOPO system, primer set A for the generation of PCR products from pure cell cultures (protozoa) and for direct sequencing. Primer set B was used for the generation of PCR products for DGGE analysis. Excised bands were sequenced using the same primer set without clamp. Primer set C was used for sequencing of PCR products cloned in TOPO system. a The GC-clamp sequence is CGCGCGCCGCGCCCCGCGCCCGTCCCGCCGCCCCCGCCCG.
DNA) of a 6.5% polyacrylamide gel, with a parallel denaturing gradient ranging from 40 to 70%. The denaturant (100%) contained 7 M Urea and 40% formamide. The gel was polymerized for 2 h. Electrophoresis was run in 1X Tris-acetateeEDTA buffer at 60 C for 6 h at 200 V. After electrophoresis, the gel was stained for 30 min in the dark with SYBR Gold (Invitrogen) nucleic acid gel stain (1:10,000 dilution in 1X TAE) for documentation, excision and reamplification of fragments, or stained with Silver using the DNA Silver Staining Kit (GE Healthcare, Bio-Sciences AB) for purposes of documentation. The SYBR-stained gels were photographed using a UV-transilluminator and a digital document imaging system (Dock-It LS, UVP INC., California). Silver stained gels were placed on a visible light box and photographed with a digital camera by use of a gel document system (GeneSnap version 7.07, SynGene, Cambridge). Well-separated and dominant DGGE bands were excised from SYBR-stained gels using a clean, sterile scalpel. Gel slices were transferred to a sterile tube containing 20 ml sterile deionized water. After incubation at 4 C for 12e24 h 1 ml of the diffusion eluted DNA was reamplified with primer set B (Table 1). These PCR products were electrophoresed in an 0.8% agarose gel to confirm purity and sequenced either directly or after cloning using the TOPO system (Invitrogen, California).
2.7.
18S rRNA gene cloning and sequencing
Clone libraries were constructed using amplicons from primer sets A and B (Table 1). Cloning was performed using a TOPO TA Cloning Kit (Invitrogen) as described in the manufacturer’s One Shot chemical transformation approach. In brief 0.5e1.5 ml of PCR product per ml vector were ligated as specified in the kit’s complete protocol. Chemically competent E. coli cells were transformed with the ligation mixes as described and selected after liquid culture by plating onto Luria-Bertani agar plates containing 100 mg ml1 ampicillin (Sigma). Plasmid DNA was isolated using the Pure Link Quick Plasmid Miniprep Kit (Invitrogen, California) or the QIAprep Spin Miniprep Kit (Qiagen, Hilden, Germany) and checked for the presence of an insert prior to sequencing by restriction analysis using EcoR1 which cuts on both sides of the linker region. All clones with insert of the expected size were selected for sequencing by using the vector specific primers M13F and M13R (primer set C, Table 1).
2.8. Sequencing, sequence analysis and assignment of phylotypes Sequencing reactions were performed by a commercial laboratory (ABI-Lab, Blindern, Oslo). Sequences were compared with those in the BLAST database (Altschul et al., 1990) for similarity. Chimeric sequences were identified by using the SSU rDNA chimera detection tool Pintail (Ashelford et al., 2005) and by BLAST analysis of several regions of each sequence. Sequences were grouped into operational taxonomic units (OTUs) at the levels of 97% (Slapeta et al., 2005) and 99% (eg, Poitelon et al., 2009; Valster et al., 2009) by use of Geneious Pro 4.8.4 (http://www.geneious.com/) (Drummond et al., 2009).
2.9.
Phylogenetic tree
Bootstrap maximum likelihood trees (1000 bootstrap replicates) were calculated for the partial 18S rRNA sequences using Geneious Pro 4.8.4 (Drummond et al., 2009). A phylogenetic tree for the aligned sequences of protists and fungi obtained by using primer set A was constructed based on the Jukes-Cantor distance model and neighbor-joining. Additional sequences from public databases were selected to optimally illustrate the phylogenetic positions of the reported sequences.
2.10.
Nucleotide sequence accession numbers
The 18S rDNA gene sequences obtained during this study have been submitted to GenBank: accession numbers JF317688eJF317810.
3.
Results
3.1.
DGGE analysis of raw and finished drinking waters
It has been reported that separation of PCR products using DGGE is best for fragments less than 500 bp (Myers et al., 1985). Consequently, following the lead of others, clamped primers (B, Table 1) providing fragments of about 250 bp were used for DGGE (Diez et al., 2001). DGGE profiles of raw and drinking waters were generated for three water works. Fig. 1c (1e2) shows the DGGE profiles for the raw and drinking waters at the Oslo Oset plant. The profiles are
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 2 7 e2 5 3 8
2531
Only a net sample was made available from Tromsø. As for the Oslo plant, gross similarities between the profiles are seen, and bands of approximately equal intensity and migration distance in the raw and treated samples were shown to have the same clonal identity. As in the Oslo profiles a number of the identified clones were crustaceans (Thermocyclops e bands 1 and 2; Skistodiaptomus e band 3b; Daphnia e bands 4 and 5). In addition, a rotifer clone (Polyarthra remata e band 3a) and an amoeba (Vanella sp. e band 6) were identified. Cloning studies showed that in at least one instance (Fig. 1a: band 3) a single band could represent more than one amplicon. The similarity of the raw and finished water eukaryote populations was also evident in the clone libraries (see below).
3.2.
Fig. 1 e DGGE profiles of raw water (R), finished water (F) and water from the distribution net (N) from (a) Frevar (Fr), (b) Tromsø (Tr) and (c) Oslo (Os). Equal amounts of DNA were applied in each well. Tentative identification of numbered DGGE bands by sequencing and BLAST analysis: 1, 2: Thermocyclops decipiens, 3a: Polyarthra remata, 3b: Skistodiaptomus pygmaeus, 4, 5: Daphnia cf. Magna, 6: Vannella sp, 7, 8: Meseres corlissi, 9: Attheyella crassa, 10, 11: Bosmina longirostris, 12, 13: Acroperus harpae.
visually highly similar when using both SYBR and silver staining, although some bands are unique to a particular water (see below). The high degree of clustering in areas of the DGGE profiles may be indicative of assemblages of highly related organisms (Fig. 1). The treated water sample used in the study represents according to the plant staff (Aasebø and Rommetveit, 2008) an approximately 4:1 mix of waters produced by the main (coagulation/flocculation/sedimentation and UV-disinfection) and reserve plant treatment (microsieving, pH adjustment and post-chlorination). The bands lying outside the main cluster of amplicons in the gels were most amenable to comparison and also these were generally represented in both profiles. Major bands were identified based on sequencing. Table 2 gives an overview of putative eukaryote identities in the sampled waters based on both DGGE and cloning studies. A major band unique to the raw water was identified as the crustacean Attheyella crassa (Fig. 1c e band 9). Other major bands present and identified in both profiles were the crustaceans Bosmina longirostris (bands 10 and 11) and Acroperus harpae (bands 12 and 13). In addition, the cyst-producing ciliate Meseres corlissi was represented in both profiles (bands 7 and 8). That bands present in the same positions in both profiles represented the same amplicon, was confirmed by sequencing of both bands excised from the gels. Fig. 1a and b shows the DGGE profiles for the raw and treated waters from the Frevar (2 hygienic barriers including filtration) and Tromsø (sieving and chlorination) plants respectively.
18S rRNA clone libraries
Table 2 gives an overview of putative eukaryote identities in the sampled waters based on both DGGE and cloning studies. Only sequences shown to be non-chimeric are included in the table. Sequences of the protists and fungi obtained by using primer set A were subjected to phylogenetic analysis using the Geneious Pro programme (Fig. 2). Sequences were classified into operational taxonomic units (OTUs) at the 97% and 99% similarity levels. Among the 86 partial eukaryotic 18S rRNA sequences obtained using primer set A, 11 OTUs at 97% similarity level and 7 OTUs at 99% similarity level were identified (Table 3). The taxonomic assignment of clones from primer set A shows that 36% of the OTUs at a 97% similarity level consist of dinoflagellates (Strobilidium, Woloszynskia, Pfiesteriaceae, Gymnodinium), 18% were ciliates (Askenasia, Kahliella), 18% of the OTUs can be assigned to the Metazoa, 9% belong to stramenopiles (Dinobryon) and 9% to Telonema. At a 99% similarity level, 29% were assigned to dinoflagellates (Gymnodinium, Pfiesteriaceae), 29% to the Metazoa, 14% to ciliates (Kahliella), 14% to stramenopiles (Dinobryon) and 14% belong to Telonema. In summary, the major fraction of the OTUs could be affiliated to free living protozoa. Only 6 fungal species are represented in the clone collection, and each only a single time. These were putative Chytridiales (raw water), Catenomyces and Sistotrema sernanderi (net water sample), Rhizoclosmatium, Rhizophlyctis harderi and Rhizophlyctis rosea (finished water). Only one of the clones affiliated with the amoebozoa (Acanthamoeba e OpR5). However, the similarity was less than 97%, which is a value that has commonly been used as the lower limit when defining microbial genera (Slapeta et al., 2005; Caron et al., 2009). In order to provide a comparison to DGGE bands, a number of clones were produced using the same primer set (primer set B, Table 1) as that used for the DGGE analysis. Among the 27 partial eukaryotic 18S rRNA sequences obtained using primer set B, 7 OTUs at 97% similarity level (FrR13-OpF6-FrR15; OpF1-OpR2FrF13; TrF14-TrF13-OpR8-OpR6; OpR3-OpF7-FrF12; OpR2-OpF1; OpF6-FrF13; OpF4-FrF18) and 5 OTUs at 99% similarity level (TrF14-TrF13-OpR8-OpR6; OpR3-OpF7-FrF12; OpR2-OpF1; OpF6FrF13; OpF4-FrR18) were identified (Table 2). The primer set B clone library overlaps with the DGGE profiles. For example, in Fig. 1b bands 4 (Tromsø: raw water) and 5 (Tromsø: distribution net water) were identical to clone TrN13 (Tromsø: distribution net water). Furthermore, band 3b in Fig. 1a (Frevar: finished
2532
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 2 7 e2 5 3 8
Table 2 e Tentative identification of clones and DGGE bands by sequencing and BLAST analysis. Sequences were aligned with the closest relatives in the GenBank database with BLAST. Nucleotide sequences can be accessed by using the accession numbers at http://www.ncbi.nlm.nih.gov. R [ Raw water, F [ Finished water, N [ Water from the distribution net, Fr [ Frevar, Os [ Oslo, Op [ Oppega˚rd, Tr [ Tromsø. Sequence size (bp)
Closest relative (Accession no.)
Alignment (% similarity)
Max Score
FrR1 FrR2 FrR3 FrR4 FrR5
615 610 601 613 608
FrR6 FrR7 FrR8 FrR9
613 462 615 611
FrR10 FrR11 FrR12 FrR13 FrR14 FrR15 FrR16
615 615 615 217 220 217 218
FrR17
217
FrR18
217
FrF1 FrF2 FrF3 FrF4 FrF5
623 615 615 598 615
FrF6 FrF7
611 481
FrF8 FrF9 FrF10
620 618 615
FrF11
600
FrF12 FrF13 OsR1 OsR2
218 217 620 614
OsR3
568
OsR4
611
OsR5
613
OsR7 OsR6 OsR8
613 614 614
OsR9 OsR10 OsR11 OsR12 OsR13 OsR14
581 614 615 613 612 614
Glenodinium inaequale (EF058237.1) Ephydatia muelleri (EU702416.1) Stentor amethystinus( AM713191.1) Mallomonas caudate (EF469643.1) Uncultured marine alveolate clone (DQ244038.1) Rana sphenocephala pathogen (EF675616.1) Monosiga ovate (AF084230.1) Hydra sp.(DQ683368.1) Gymnodinium sp (EF492493.1) Uncultured alveolate clone (DQ504314.1) Pentapharsodinium tyrrhenicum (AF022201.1) Chrysamoeba tenera (EF165102.1) Uncultured eukaryotic picoplankton clone (AY642722.1) Pfiesteriaceae sp. Masanensis (EU048553.1) Synura uvella (U73222.1) Woloszynskia pascheri (EF058253.1) Poterioochromonas sp. (EU586184.1) Uncultured apicomplexan clone (EU143868.1) Rana sphenocephala pathogen (F675616.1) Uncultured ciliate clone (EU143862.1) Uncultured choreotrichid ciliate (AY821916.1) Uncultured alveolate clone (EU162628.1) Rana sphenocephala pathogen (EF675616.1) Cyclops sp. (AY626998.1) Woloszynskia pascheri (EF058253.1) Apicoporus parvidiaboli (EU293238.1) Woloszynskia leopoliensis (AY443025.1) Uncultured marine eukaryote clone (DQ310268.1) Pfiesteriaceae sp. Masanensis (EU048553.1) Notommata allantois (DQ297710.1) Uncultured eukaryote clone (EF659887.1) Woloszynskia halophila (EF058252.1) Eudiaptomus gracilis (AY339148.1) Trichocerca tenuior (DQ297723.1) Uncultured eukaryote clone (EU733826.1) Peridinium polonicum (AY443017.1) Uncultured eukaryote clone (DQ455742.1) Urotricha sp. (EU024981.1) Skistodiaptomus pygmaeus (AY339161.1) Mesocyclops thermocyclopoides (EF581894.1) Eudiaptomus gracilis (AY339148.1) Uncultured freshwater eukaryote (AY919792.1) Dinophyceae sp. Lucy-8 (AY251290.1) Uncultured marine eukaryote clone (AY426938.1) Heliophrya erhardi clone (AY007448.1) Uncultured stramenopile clone (EU162647.1) Bicosoeca petiolata (EU162647.1) Uncultured freshwater clone (AY919778.1) P.foraminifera (Z38025.1) Uncultured opisthokont clone (FJ976648.1) Eunotia sp. (AM501963.1) Uncultured freshwater eukaryote clone (AY919778.1) Paraphysomonas imperforate (AF109323.1) Unidentified eukaryote 18S ribosomal RNA (AJ130851.1) Fragilaria rumpens (AM497722.1) Dinobryon cylindricum (EF165140.1) Uncultured eukaryote 18S rRNA gene (AJ564770.1) Chlamydomonad sp (AY220571.1) Uncultured eukaryotic picoplankton clone (AY642736.1) Woloszynskia pascheri (EF058253.1)
97 99 95 99 93 91 93 99 99 95 95 97 99 98 99 99 98 97 97 99 99 99 93 93 97 98 98 98 97 98 99 99 99 99 98 98 94 93 100 99 99 99 93 90 87 98 89 96 94 89 98 98 94 99 99 99 99 98 98 97
1048 1122 968 1116 917 819 896 837 1109 983 979 1057 1103 1092 396 396 385 375 370 396 396 396 324 941 1059 1070 1066 1070 1064 1085 878 867 1123 1109 1098 1098 911 904 403 396 1134 1118 902 737 630 1068 776 1003 941 778 1070 1064 946 1057 1127 1120 1105 1081 1074 1062
Sample
2533
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 2 7 e2 5 3 8
Table 2 (continued) Sequence size (bp)
Closest relative (Accession no.)
Alignment (% similarity)
Max Score
OsN1 OsN2 OsN3
614 615 603
OsN4 OsN5 OsN6 OsN7 OsN8 OsN9
615 615 614 622 614 511
OsN10 OsN11
620 613
OsN12 OsN13
495 606
OsN14 OsN15
561 600
OsN16 OsN17
547 597
OsF1
613
OsF2
606
OsF3
604
OsF4
597
OsF5
622
OsF6
604
TrR1
587
TrR2 TrR3 TrR4 TrR5
592 615 620 614
TrR6
615
TrR7 TrR8
612 613
TrR9
597
TrR10 TrR11
625 588
TrR12 TrR13 TrR14
615 624 599
TrN1 TrN2 TrN3 TrN4 TrN5 TrN6
615 624 403 620 557 545
Monosiga ovate (AF084230.13) Catenomyces sp. (AY635830.14) Uncultured eukaryote clone (EF100279.1) Strobilidium caudatum (AY143573.1) Nais variabilis (DQ459978.1) Woloszynskia pascheri (EF058253.1) Sistotrema sernanderi (AY757264.1) Uncultured cercozoan clone (AY620270.1) Amphidinium herdmanii (AF274253.1) Uncultured marine alveolate clone (DQ244036.1) Kahliella sp. (EU079472.1) Eudiaptomus gracilis (AY339148.1) Uncultured eukaryote (AJ564770.1) Telonema antarcticum (AJ564773.1) Pfiesteriaceae sp. masanensis isolate (EU048553.1) Uncultured marine alveolate clone (DQ244036.1) Kahliella sp. (EU079472.1). Pirsonia verrucosa partial 18S rRNA gene (AJ561113.1) Uncultured marine alveolate clone (DQ244026.1) Urotricha sp. (EU024981.1) Gymnodinium sp. (EF492493.1) Uncultured eukaryotic picoplankton clone (AY642698.1) Askenasia sp. (EU024991.1) Uncultured eukaryotic picoplankton clone (AY642735.1) P.foraminifera (Z38025.1) Uncultured marine alveolate clone (DQ244036.1) Kahliella sp. TT2005 (EU079472.1) Uncultured eukaryote clone (EF100279.1) Strobilidium caudatum (AY143573.1) Uncultured eukaryotic picoplankton clone (AY642698.1) Askenasia sp. FU44-55 (EU024991.1) Uncultured marine eukaryote clone (DQ314809.1) Protaspis grandis (DQ303924.1) Uncultured eukaryotic picoplankton clone (AY642730.1) Strobilidium caudatum (AY143573.1) Uncultured freshwater eukaryote clone (AY919716.1) Dinophyceae sp. (EU418969.1). Tetrahymena sonneborni (EF070258.1) Gymnodinium sp (EF492493.1) Eudiaptomus gracilis (AY339148.1) Uncultured eukaryotic picoplankton (AY642736.1) Woloszynskia pascheri (EF058253.1) Uncultured freshwater eukaryote clone (AY919747.1) P.foraminifera (Z38025.1) Chlamydomonad sp. (AY220571.1) Uncultured eukaryote 18S rRNA gene (AJ564770.1) Telonema antarcticum (AJ564773.1) Unidentified eukaryote 18S rRNA, (AJ130855.1) Vorticella campanula (DQ662849.1) Pedinellales sp. (EU247836.1) Uncultured freshwater eukaryote clone (AY919716.1) Scrippsiella precaria (DQ847435.1) Uncultured eukaryote clone (EF100355.1) Thermocyclops sp. (DQ107580.1) Askenasia sp. (EU024989.1) Ochromonas tuberculata (AF123293.1) Dinobryon cylindricum (EF165140.1) Thermocyclops sp. (DQ107580.1) Woloszynskia pascheri (EF058253.1) Eudiaptomus graciloides (AY339149.1) Rhizophlyctis harderi (AF164272.2) Cyclops kolensis (GU066284.1)
96 96 94 94 99 98 99 96 97 98 98 99 98 95 99 98 97 96 99 94 99 96 94 99 96 99 98 95 95 96 94 98 97 95 94 98 97 96 99 99 97 97 98 94 97 99 95 98 98 95 96 96 95 99 99 95 99 98 99 99 92 99
1003 1014 931 926 1120 1081 1112 1027 1059 905 900 1129 1094 965 893 1081 1042 917 1081 931 983 992 922 1101 996 1092 1053 965 961 998 933 1083 1066 977 922 1027 1007 987 1109 1129 1057 1047 1094 942 1059 1105 976 1044 1038 987 904 887 981 1120 1085 981 1116 1109 728 1123 761 979
Sample
(continued on next page)
2534
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 2 7 e2 5 3 8
Table 2 (continued) Sample
Sequence size (bp)
TrN7
613
TrN8
615
TrN9 TrN10 TrN11 TrN12 TrN13 TrN14 OpF1 OpF2
615 641 613 620 220 220 217 217
OpF3 OpF4 OpF5 OpF6 OpF7 OpF8 OpF9 OpR1 OpR2 OpR3 OpR4 OpR5
220 21 220 217 218 219 219 221 217 216 220 220
OpR6 OpR7 OpR8 Dgge1,2 Dgge3a Dgge 3b Dgge 4,5 Dgge 6 Dgge 7,8 Dgge 9 Dgge 10,11 Dgge 12,13
220 219 220 217 218 162 220 241 218 217 219 220
Closest relative (Accession no.) Uncultured eukaryotic picoplankton (AY642734.1) Rhizophlyctis rosea isolate (AY635829.1) Uncultured freshwater eukaryote clone (AY919716.1). Dinophyceae sp. (EU418969.1) Dinophyceae sp. (EU418969.1) Aphanomyces sp. (FJ794897.1) Spumella oblique (AJ236860.1) Pythiopsis cymosa (AJ238657.1) Daphnia cf. magna (EU370423.1) Acroperus harpae (AM490272.1) Cyclops insignis (EF532821.2) Uncultured eukaryotic picoplankton (AY642699.1) Uncultured cryptophyte clone (EU143967.1) Stephanodiscus minutulus (DQ514916.1) Uncultured alveolate clone (EU162628.1) Aphanomyces invadans (DQ403202.1) Synura uvella (U73222.1) Mastigodiaptomus nesus (AY339156.1) Rhizoclosmatium sp. (AY601709.1) Entophlyctis luteolus (AF164326.1) Uncultured freshwater cercozoan clone (DQ243991.1) Cyclops insignis (EF532821.2) Mastigodiaptomus nesus (AY339156.1) Chytridiales sp. (AF164262.1) Uncultured cercozoan clone (EU143889.1) Acanthamoeba (EF023782.1) Acroperus harpae (AM490272.1) Thaumatomonadida (EF024801.1) Daphnia cf. magna (EU370423.1) Thermocyclops decipiens (GQ848504.1) Polyarthra remata (DQ297716.1) Skistodiaptomus pygmaeus (AY339161.1) Daphnia cf. Magna (EU370423.1) Vannella sp (AY929914.1) Meseres corlissi strain (DQ244039.1) Attheyella crassa (EU380307.1) Bosmina longirostris (AM490275.1) Acroperus harpae (AM490272.1)
water) was shown to be identical to clone FrF12 (Frevar: finished water).
3.3.
Isolation and culture of Free living protozoa
The presence of free living protozoa in raw, treated and net water samples at the Oslo plant was investigated. Particular emphasis was placed on amoebae which are known to contain human pathogens and to shield pathogenic bacteria. Culture techniques raised protozoal populations for both raw, treated and net water samples perhaps indicating that these can cross hygienic barriers. For example, 8 of 8 samples taken at 4 points in the distribution system all grew protozoal cultures. In only one instance was an amoeba present (see below). Protozoa populations were documented by visual examination of the size, shape and means of motilitity using phase contrast microscopy. The amoeba-species were isolated in pure culture and identified by genotyping (see below). Based on morphological criteria described elsewhere (Finlay et al., 1988), the following genera were tentatively identified using phase
Alignment (% similarity)
Max Score
99 93
1110 902
97 97 98 99 98 99 99 100 99 100 98 100 99 99 98 99 98 96 99 99 100 99 99 95 100 99 100 100 98 100 100 97 9 98 98 96
1057 1037 1070 1123 1083 1114 401 407 390 401 379 407 396 396 396 398 388 366 403 390 399 401 396 351 407 399 407 392 387 300 407 405 381 385 383 363
contrast microscopy: Chrysomonad flagellates (small, with two flagella of unequal length eg, Ochromonas and Spumella) were commonly grown and some of these are also represented in the clone library. Volvocid flagellates including presumptive Monsiga (with several anterior flagella and stalk) and kintetopid flagellates including presumptive Bodo (bean-shaped with trailing and leading flagella) were seen. Flagellates were in greatest abundance. Only 2 ameobozoa are represented in the clone libraries and sequences obtained from DGGE (Table 2). These were a presumptive Acanthamoeba sp (raw water from Oppega˚rd, direct cloning, OpR5) and a Vannella sp. (DGGE band 6, finished water from Tromsø). In addition, Acanthamoeba castellanii was isolated in pure culture from Oslos distribution system. Strain identity was confirmed based on sequencing (AF260724.1, 100% similarity, max score 911).
3.4.
Heterotrophic plate count (HPC)
The viable heterotrophic plate counts for raw and processed waters at the Oset water plant in Oslo were similar (about
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 2 7 e2 5 3 8
2535
Fig. 2 e Phylogenetic tree of protists and fungi 18S rDNA obtained using primer set A. Fr [ Frevar, Tr [ Tromsø, Os [ Oslo, R [ Raw water, F [ Finished drinking water, N [ Water from the distribution net. Sequences obtained from the GenBank database (http://www.ncbi.nlm.nih.gov) are included for purposes of comparison. The scale bar represents 0.05 substitutions per site.
2536
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 2 7 e2 5 3 8
Table 3 e Classification of clones produced with primer set A as operational taxonomic units (OTUs) at >97% and >99% similarity levels. Taxonomic level Alveolata Dinoflagellates
Ciliates Stramenopiles Telonema Chlamymonads Metazoa
Percentage similarity 97
99
OsF3-OsN3 OsR14-TrR5-TrN3-OsN5-FrF2-TrN8-TrN9-TrR1 FrR12-OsN12 FrR8-TrR3-FrF7-OsN16 OsF4-OsN17 OsF2-OsN13-OsN9 OsR11-TrN1 OsR12-TrR8-OsN11 OsR13-TrR7 OsN10-OsR1-TrR4-FrF8-TrN4 TrN2-TrN13-TrN6
FrR12-OsN12 FrR8-TrR3-FrF7-OsN16 OsF2-OsN13-OsN9 OsR11-TrN1 OsR12-TrR8-OsN11 OsN10-OsR1-TrR4-FrF8-TrN4 TrN2-TrN13-TrN6
Key: Fr ¼ Frevar, Tr ¼ Tromsø, Os ¼ Oslo, R ¼ Raw water, F ¼ Finished drinking water, N ¼ Water from the distribution net.
500 colony forming units (cfu)/ml). This shows that bacteria are able to survive treatment at the plant. At the sampling point on the Oslo net the count was often >1000 cfu/ml. The HPC for raw water at the Frevar water plant in Fredrikstad was about 800 cfu/ml whereas the finished water had a considerably lower HPC (about 70 cfu/ml) indicating an efficient disinfection process. No colonies were observed for Oppega˚rds finished water whereas in raw water the HPC was about 340 cfu/ml. For Tromsø only a distribution net sample was made available. The cfu/ml in the net sample was significantly higher (about 850 colony forming units (cfu)/ml) than in the raw water (about 550 colony forming units (cfu)/ml).
have a pathogenic potential for humans, and none are classified as primary pathogens. A. castellanii is a notable eye infectant (Zanetti et al., 1995) and disease has been recorded in Norway (Aasly and Bergh, 1992). It has also been demonstrated that some waterborne pathogens i.e. Legionella pneumophila and Cryptosporidium oocysts are able to maintain viability inside certain species of amoebae and ciliated protozoa (reviewed by Bichai et al., 2008). Nonetheless the species is considered usual in water and soil (Zanetti et al., 1995). A number of the potential fish pathogens Pfesteriaceae and Aphantomyces were detected and these are worthy of further studies to evaluate the significance of the findings.
4.
4.2. The eukaryote community of raw and finished waters
Discussion
4.1. The eukaryotic diversity in Norwegian drinking waters To our knowledge the present study represents the first attempt at providing a general overview over the small eukaryotic population of several Norwegian drinking water supplies, including the Capital. Both non-culture based techniques (DGGE and direct cloning) and cultivation of HPC and protozoa were used. Single-celled ciliates and flagellates, particularly dinoflagellates dominated both the culture based protozoan cultures and the non-culture based clone libraries. The composition of the small eukaryotic population (Table 2 and Fig. 2) in the samples tested, shows similarities with previous studies of small eukaryotes in aquatic environments (Diez et al., 2001; Moon-van der Staay et al., 2001; Valster et al., 2009). For example, Moon-van der Staay (2001) identified large populations of choanoflagellates (eg, Monsiga), dinoflagellates and stramenopiles (eg, Ochromonas) in pristine oceanic water. These are also amongst the most widely represented taxa in the current study. With the exception of primary pathogens, it is difficult to define good criteria to judge if named organisms represent a health threat. The clone identifications were compared against citations in the PubMed article database using key words such as “Disease” and “Pathogen”. None of the indicated strains with the exception of A. castellanii, seems to
The DGGE profiles for raw and treated waters were visually similar. Given that equal amounts of DNA were applied in each well, the figure seems to indicate low removal of eukaryote cells (Fig. 1). Furthermore, similarities with respect to OTUs were found in the raw and treated waters, suggesting similar small eukaryotic populations. In addition, culturing of protists raised similar protozoal populations from raw and drinking waters. Taken together these results may indicate limited removal of small eukaryotes during the treatment processes in Oslo. However, the Oslo samples were taken during a period when a more comprehensive treatment train was being optimalized (coagulation/flocculation/sedimentation and UV-disinfection) and the analyzed samples contained about 20% water from the reserve plant. Water from the reserve plant represents only a single hygienic barrier (chlorination). It is, therefore, planned to re-evaluate the new Oslo treatment train at a later date. An additional similarity is provided by the HPC which were comparable for raw and treated waters at the plant. Even though the DGGE profiles for raw and treated waters showed similarities, the results from HPC-analysis suggest that some treatment trains (eg, Fredrikstad) effectively disinfect the water during production. Although not directly representing the small eukaryote population, HPC measures provides a convenient indication of disinfectant efficacy. Future work will aim at
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 2 7 e2 5 3 8
finding a suitable eukaryotic indicator organism. Our group are currently working on real-time PCR assay for Cryptococcus seen in raw and treated waters.
4.3.
Considerations of the methodology used
The cloning results indicate low numbers of amoebozoa and these were found only once in cultured samples. Similarly, fungal species account for only about 5% of the clone population. Given that free living amoebae are common in soil and water (Rodrı´guez-Zaragoza, 1994) the result suggest that the methods used are not adequate to give a good representation of the amoebozoan fraction of drinking water. Biases in the method of DNA isolation (particularly with regard to fungi) and/or the PCR reaction are expected to affect the numbers and types of cells detected. The presence of a fungal cell wall requires a more stringent regime for DNA isolation prior to PCR. It has also been noted that the cells of amoeboid taxa are less easily disrupted than many others (Berney et al., 2004). According to the manufacturer the UltraClean Soil DNA Isolation kit efficiently lyses fungi (Mo BIO Laboratories, Inc, Solana Beach, California). It has also previously been shown that this kit released fungal DNA from soil samples which could then be concentrated by PCR (Gryndler et al., 2008). However, to our knowledge no data exists on the general efficiency of the test system for the isolation of fungal DNA. Similarly, the UltraClean Soil DNA isolation kit has been used for DNA isolation from free living amoebae (Declerck et al., 2007) but its efficiency in this respect has also not been documented. Another potential source of underrepresentation in the clone library is the choice of primers used for PCR amplification. However, primer sets A and B have been used in studies of free living amoebae (Weekers et al., 1994; Corsaro et al., 2009). Furthermore, our search into primer sequences revealed that primer sets A and B show 100% homology with for example Saccharomycetales, Penicillium, Acanthamoeba and Hartmannella 18S rRNA genes. The low abundance of amoebozoan clones is also reflected in cell culture where only one sample produced a free living amoeba. More studies are required to show to what extent the chosen methods are suitable to reveal the presence of fungi and amoebae. We are thus currently engaged in an extensive study of culturable yeasts in Oslo’s drinking water to complement the present study and previous studies of molds (Hageskal et al., 2006, 2007). The finding that even well-isolated DGGE bands could represent multiple clones suggests direct cloning and sequencing to be a more suitable tool for the identification of the small eukaryotic population. However, DGGE provides an easily graspable overview of the numbers and possibly abundances of organisms in tested samples (Muyzer and Smalla, 1998).
5.
Major conclusions
Cultivation-independent and dependent techniques revealed highly diverse eukaryote communities in both raw, finished and net drinking water samples. The communities were dominated by dinoflagellates, ciliates and metazoans. No
2537
primary human pathogens were detected using the cultivation-independent molecular methods. The potentially pathogenic species A. castellanii was cultivated from a net sample. DGGE profiling conveniently displays the small protist populations of raw and treated waters and may provide an indication of the physical removal of such organisms during water treatment. Although some bands were unique to either the raw or processed waters, the profiles showed gross similarities regardless of the treatment train, indicating that no treatment effectively removed cellular material. However, HPC-analysis suggests that some plants (eg, Fredrikstad) had efficient disinfection processes. Cloning, though more time consuming than DGGE, gives a more detailed overview of the species composition and provides better quality DNA for sequencing purposes. Taken together the two techniques are able to describe and visualize changes in drinking water quality during production. However, they do not distinguish between the living and dead biota. Therefore, in addition to HPC testing we are currently working on real-time PCR assays of eukaryote indicators (including Cryptococcus). With the exception of the finding of small numbers of transient Giardia in Oslo in 2007 (Robertson et al., 2009) there have been to our knowledge no reported cases of disease traceable to eukaryotes in Oslo’s drinking water, and the current study seems to support the contention that the Capital’s water appears to hold a satisfactory quality, while maintaining a normal population of small eukaryotes.
Acknowledgments We would like to thank employees at the water plants included in the study for collecting the water samples.
Appendix. Supplementary material Supplementary data related to this article can be found online at doi:10.1016/j.watres.2011.02.008.
references
Aasebø, Kari Rommetveit (2008). Personal communication. Aasly, K., Bergh, K., 1992. Acanthamoeba keratitis; report of the first Norwegian cases. Acta Opthalmologica (Copenh) 70 (5), 698e701. Altschul, S.F., Gish, W., Miller, W., Myers, E.W., Lipman, D.J., 1990. Basic local alignment search tool. Journal of Molecular Biology 215, 403e410. Ashelford, K.E., Chuzhanova, N.A., Fry, J.F., Jones, A.J., Weightman, A.J., 2005. At least one in twenty 16S rRNA sequence records currently held in public repositories estimated to contain substantial anomalies. Applied and Environmental Microbiology 71 (12), 7724e7736. Berney, C., Fahrni, J., Pawlowski, J., 2004. How many novel eukaryotic “kingdoms”? Pitfalls and limitations of environmental DNA surveys. BioMed Central Biology 2, 1e13. Bichai, F., Payment, P., Barbeau, B., 2008. Protection of waterborne pathogens by higher organisms in drinking water. Canadian Journal of Microbiology 54 (7), 509e524.
2538
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 2 7 e2 5 3 8
Caron, D.A., Countway, P.D., Savai, P., Gast, R.J., Schnettzer, A., Moorthi, S.D., Dennett, M.R., Moran, D.M., Jones, A.C., 2009. Defining DNA-based operational taxonomic units for microbial-eukaryote ecology. Applied and Environmental Microbiology 75 (18), 5797e5808. CEC (Council of the European Communities), 1998. Directive (98/83/EC) on the Quality of Water Intended for Human Consumption (accessed 27.03.07) Available from: http://eurlex.europa.eu/LexUriServ/site/en/oj/1998/l_330/l_ 33019981205en00320054.pdf. Corsaro, D., Feroldi, V., Saucedo, G., Ribas, F., Loret, J.F., Greub, G., 2009. Novel chlamydiales strains isolated from a water treatment plant. Environmental Microbiology 11, 188e200. Declerck, P., Behets, J., van Hoef, V., Ollevier, F., 2007. Detection of Legionella spp. and some of their amoeba hosts in floating biofilms from anthropogenic and natural aquatic environments. Water Research 41 (14), 3159e3167. Dewettinck, T., Hulsbosch, W., van Hege, K., Top, E.M., Verstraete, W., 2001. Molecular fingerprinting of bacterial populations in groundwater and bottled mineral water. Applied Microbiology and Biotechnology 57 (3), 412e418. Diez, B., Pedros-Alio, C., Marsh, T.L., Massana, R., 2001. Application of denaturing gradient gel electrophoresis (DGGE) to study the diversity of marine picoeukaryotic assemblages and comparison of DGGE with other molecular techniques. Applied and Environmental Microbiology 67 (7), 2942e2951. Dorigo, U., Volatier, L., Humbert, J.-F., 2005. Molecular approaches to the assessment of biodiversity in aquatic microbial communities. Water Research 39 (11), 2207e2218. Drummond, A.J., Ashton, B., Cheung, M., Heled, J., Kearse, M., Moir, R., Stones-Havas, S., Thierer, T., Wilson, A., 2009. Geneious V4.7 Available from: http://www.geneious.com/. Finlay, B.J., Rogerson, A., Cowling, A.J., 1988. A Beginners Guide to the Collection, Isolation, Cultivation and Identification of Freshwater Protozoa. Titus Wilson and Son Ltd, Kendal. Gryndler, M., Krofta, K., Gryndlerova´, H., Soukupova´, L., Hrselova´, H., Gabriel, J., 2008. Potentially dangerous fusarioid microorganisms associated with rot of hops (Humulus lupulus L.) plants in field culture. Plant, Soil and Environment 54, 149e154. Hageskal, G., Knutsen, A.K., Gaustad, P., de Hoog, G.S., Skaar, I., 2006. Diversity and significance of mold species in Norwegian drinking water. Applied and Environmental Microbiology 72 (12), 7586e7593. Hageskal, G., Gaustad, P., Heier, B.T., Skaar, I., 2007. Occurrence of moulds in drinking water. Journal of Applied Microbiology 102 (3), 774e780. 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. Hoefel, D., Monis, P.T., Grooby, W.L., Andrews, S., Saint, C.P., 2005. Profiling bacterial survival through a water treatment process and subsequent distribution system. Journal of Applied Microbiology 99 (1), 175e186. Marciano-Cabral, F., Cabral, G., 2003. Acanthamoeba spp. as agents of disease in humans. Clinical Microbiology Reviews 16 (2), 273e307.
Moon-van der Staay, S.Y., De Wachter, R., Vaulot, D., 2001. Oceanic 18S rDNA sequences from picoplankton reveal unsuspected eukaryotic diversity. Nature 409, 607e610. Muyzer, G., Smalla, K., 1998. Application of denaturing gradient gel electrophoresis (DGGE) and temperature gradient gel electrophoresis (TGGE) in microbial ecology. Antonie Van Leeuwenhoek 73 (1), 127e141. Myers, R.M., Fischer, S.G., Lerman, L.S., Maniatis, T., 1985. Nearly all single base substitutions in DNA fragments joined to a GC-clamp can be detected by denaturing gradient gel electrophoresis. Nucleic Acid Research 13 (9), 3131e3145. Otterholt, E., Charnock, C., 2011. Microbial quality and nutritional aspects of Norwegian brand waters. International Journal of Food Microbiology 144 (3), 455e463. Poitelon, J.-B., Joyeux, M., et al., 2009. Identification and phylogeny of eukaryotic 18S rDNA phylotypes detected in chlorinated finished drinking water samples from three Parisian surface water treatment plants. Letters in Applied Microbiology 49 (5), 589e595. Robertson, L., Gjerde, B., Hansen, E.F., Stachurska-Hagen, T., 2009. A water contamination incident in Oslo, Norway during October 2007; a basis for discussion of boil-water notices and the potential for post-treatment contamination of drinking water supplies. Journal of Water Health 7, 55e66. Rodrı´guez-Zaragoza, S., 1994. Ecology of Free living amoebae. Critical Reviews in Microbiology 20 (3), 225e241. Rowbotham, T.J., 1980. Preliminary report on the pathogenicity of Legionella pneumophila for freshwater and soil amoebae. Journal of Clinical Pathology 33, 1179e1183. Schreiber, H., Schoenen, D., Traunspurger, W., 1997. Invertebrate colonization of granular activated carbon filters. Water Research 31 (4), 743e748. Slapeta, J., Moreira, D., Lo´pez-Garcı´a, P., 2005. The extent of protist diversity: insights from molecular ecology of freshwater eukaryotes. Proceedings of the Royal Society: Biological Sciences 272, 2073e2081. Thomas, V., Herrere-Rimann, K., Blanc, D.S., Greub, G., 2006. Biodiversity of amoebae and amoeba-resisting bacteria in a hospital water network. Applied and Environmental Microbiology 72 (4), 2428e2438. Valster, R.M., Wullings, B.A., Bakker, G.B., Smidt, H., van der Kooij, D., 2009. Free living protozoa in two unchlorinated drinking water supplies, identified by phylogenic analysis of 18S rRNA gene sequences. Applied and Environmental Microbiology 75 (14), 4736e4746. Weekers, P.H., Gast, R.J., Fuerst, P.A., Byers, T.J., 1994. Sequence variations in small-subunit ribosomal RNAs of Hartmannella vermiformis and their phylogenetic implications. Molecular Biology and Evolution 11 (4), 684e690. Winter, C., Hein, T., Kavka, G., Mach, R.L., Farnleitner, A.H., 2007. Longitudinal changes in the bacterial community composition of the Danube River: a whole river approach. Applied and Environmental Microbiology 73 (2), 421e431. Zanetti, S., Fiori, P.L., Pinna, A., Usal, S., Carta, F., Fadda, G., 1995. Susceptibility of Acanthamoeba castellanii to contact lens disinfecting solutions. Applied and Environmental Microbiology 39 (7), 1596e1598.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 3 9 e2 5 4 9
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Common key acidogen populations in anaerobic reactors treating different wastewaters: Molecular identification and quantitative monitoring Jaai Kim a, Seung Gu Shin b, Gyuseong Han b, Vincent O’Flaherty c, Changsoo Lee a,*, Seokhwan Hwang b,** a
Division of Environmental and Water Resources Engineering (EWRE), School of Civil and Environmental Engineering, Nanyang Technological University, Singapore b School of Environmental Science and Engineering, POSTECH, San 31, Hyoja-dong, Nam-gu, Pohang, Gyungbuk 790-784, Republic of Korea c Microbial Ecology Laboratory, Department of Microbiology and Environmental Change Institute (ECI), National University of Ireland, Galway, Ireland
article info
abstract
Article history:
Bacterial population dynamics during the start-up of three lab-scale anaerobic reactors
Received 19 October 2010
treating different wastewaters, i.e., synthetic glucose wastewater, whey permeate, and
Received in revised form
liquefied sewage sludge, were assessed using a combination of denaturing gradient gel
24 January 2011
electrophoresis (DGGE) and real-time PCR techniques. The DGGE results showed that
Accepted 4 February 2011
bacterial populations related to Aeromonas spp. and Clostridium sticklandii emerged as
Available online 15 February 2011
common and prominent acidogens in all reactors. Two real-time PCR primer/probe sets
Keywords:
tatively investigate their dynamics in relation to changes in reactor performance.
targeting Aeromonas or C. sticklandii were developed, and successfully applied to quantiAcidogenesis
Quantitative analysis demonstrated that both Aeromonas- and C. sticklandii-related pop-
Aeromonas
ulations were highly abundant for acidogenic period in all reactors. Aeromonas populations
Clostridium sticklandii
accounted for up to 86.6e95.3% of total bacterial 16S rRNA genes during start-up, sug-
DGGE
gesting that, given its capability of utilizing carbohydrate, Aeromonas is likely the major
Real-time PCR
acidogen group responsible for the rapid initial fermentation of carbohydrate. C. sticklandii, able to utilize specific amino acids only, occupied up to 8.5e55.2% of total bacterial 16S rRNA genes in the reactors tested. Growth of this population is inferred to be supported, at least in part, by non-substrate amino acid sources like cell debris or extracellular excretions, particularly in the reactor fed on synthetic glucose wastewater with no amino acid source. The quantitative dynamics of the two acidogen groups of interest, together with their putative functions, suggest that Aeromonas and C. sticklandii populations were numerically as well as functionally important in all reactors tested, regardless of the differences in substrate composition. Particularly, the members of Aeromonas supposedly play vital roles in anaerobic digesters treating various substrates under acidogenic, fermentative start-up conditions. ª 2011 Elsevier Ltd. All rights reserved.
* Corresponding author. Tel.: þ65 56 6790 5316; fax: þ65 56 6791 0676. ** Corresponding author. Tel.: þ82 54 279 2282; fax: þ82 54 279 8299. E-mail addresses:
[email protected] (C. Lee),
[email protected] (S. Hwang). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.02.004
2540
1.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 3 9 e2 5 4 9
Introduction
Anaerobic digestion (AD) has been widely used for treating various wastewaters mainly due to producing energy in the form of methane simultaneously with reducing pollution load. AD process is underpinned by a series of biological reactions performed by a consortium of diverse microorganisms that can be roughly grouped into acidogenic bacteria (acidogens) and methanogenic archaea (methanogens). The former hydrolyzes and ferments organic materials to produce organic acids and alcohols, and further converts these intermediates to acetate and hydrogen. The latter utilizes the acidogenic end products to generate methane (Speece, 1996). A wide variety of populations with distinct physiological characteristics thus coexist in an AD system even when the influent contains only one organic substrate, and their concerted activity is important for the stable conversion of organic compounds to methane (Fernandez et al., 2000). Therefore, a comprehensive understanding of the microbial community structures and dynamics in AD is desired for fundamentally improving the overall process technology. Application of culture-independent molecular techniques, particularly those based on 16S rRNA gene sequences, has remarkably raised our ability to approximate real microbial world by eliminating the need for cultivating individual community members. Such advance enabled to link microbial behaviors with process performance data, and many studies have been made to investigate microbial communities in different AD processes, both qualitatively and, more recently, quantitatively (Akarsubasi et al., 2005; Hori et al., 2006; McMahon et al., 2001; Yu et al., 2006; Zumstein et al., 2000). Previous studies have mostly focused on methanogens because they are directly responsible for producing methane at the final step of AD pathway (Speece, 1996). However, it has been posed that AD is desired to be studied also in light of the behavior of acidogens playing primary and critical roles in producing major substrates for methanogens. It has been reported that an imbalance among or overgrowth of acidogens can cause an accumulation of unfavorable (or even toxic) intermediates to methanogens and stable performance of anaerobic digestion systems depends greatly on the establishment of a suitable microbial community during start-up (McMahon et al., 2001, 2004). This is the principal reason for the frequently observed digester upset during start-up, leading to poor performance or even to process failure (Leclerc et al., 2001). Despite such important influences, relatively little has focused on the community behavior of acidogens as compared with methanogens, particularly during start-up that is an important period for the long-term stability of an anaerobic digester. This study thus aimed to identify and quantitatively monitor key acidogen populations during startup of anaerobic digesters, which may play vital roles for initiating the overall process and for accumulating acidogenic products. For a comparative study, three lab-scale batch digesters treating different wastewaters, i.e., synthetic glucose wastewater, whey permeate, and liquefied sewage sludge, were operated to investigate acidogen communities under various substrate composition conditions. Contrary to methanogens,
all belonging to only five orders under one phylum, acidogens are extremely widely spread across >20 phyla, indicating that their phylogenetic positions and phenotypic functions (i.e., capability of producing organic acids) are not tightly linked. This is one of the main reasons why the application of molecular techniques, mostly based on 16S rRNA gene sequences today, in studying acidogen communities is to be limited. In this study, to tackle this limitation, bacterial community structures were first analyzed from the tested digesters by denaturing gradient gel electrophoresis (DGGE) to identify the key, common acidogens in all trials. The acidogens commonly identified in three digesters were then further examined using real-time PCR, to quantitatively analyze their dynamics and to ascertain their contributions to total bacterial communities, in relation to process performance data. For this, new real-time PCR primer/probe sets detecting the acidogens of interest were designed and evaluated in this study. To the best of the author’s knowledge, this is the first study reporting the identification, quantification, and monitoring of key acidogen populations common in anaerobic digesters treating different wastewaters. This study provides not only specific and quantitative insights into the behavior of acidogens but also new tools for investigating key acidogens, which may be ubiquitous and vital in many anaerobic processes regardless of substrate composition, although likely containing some common components.
2.
Materials and methods
2.1.
Bioreactor operation
Three anaerobic completely mixed tank reactors (6-L working volume) were run in batch mode to treat different types of wastewaters: synthetic glucose wastewater (G), whey permeate (W) and liquefied sewage sludge (S). Glucose synthetic wastewater contains (mg/L): Glucose, 4868; yeast extract, 50; NH4Cl, 955.5; KH2PO4, 63.6; K2HPO4∙3H2O, 123; NaCl, 600; KCl, 185; MgSO4∙7H2O, 123.6; Nitrilotriacetic acid, 40; CaCl2∙2H2O, 20; FeCl3∙6H2O, 0.1; MnCl2∙4H2O, 0.9; H3BO3, 0.2; CoCl2∙6H2O, 1.5; CuCl2∙2H2O, 2.2; NiCl2∙6H2O, 1.2; 45% Na2SeO3, 0.6; ZnCl2, 0.9; Citric acid∙H2O, 105. Whey permeate was prepared by dissolving whey permeate powder in distilled water (DW). No supplementary nutrient was added because whey permeate already contains most of the essential nutrients for microbial growth (Hwang and Hansen, 1992). Thickened sewage sludge was liquefied by a thermochemical treatment with 4 g/L of NaOH at 121 C for 30 min (Kim et al., 2003), and then filtered through Whatman GF/C glass-fiber filter (pore size, 1.2 mm). The initial substrate concentration was adjusted to 5 g/L as soluble chemical oxygen demand (SCOD) in all reactors. Physicochemical characteristics of the wastewaters used are summarized in Table 1. Each reactor was inoculated with mesophilic anaerobic sludge with a seeding ratio of 1% (v/v). Temperature was held at 35 C and pH was maintained over 7.0 with 3 N NaOH solution. Thickened sewage sludge and anaerobic seed sludge were collected from a full-scale municipal wastewater treatment plant (Pohang, South Korea).
2541
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 3 9 e2 5 4 9
Table 1 e Physical and chemical characteristics of the wastewaters studied. Concentrations (mg/L)a
Parameters Synthetic glucose wastewater Total COD Soluble COD Total carbohydrate Soluble carbohydrate Total protein Soluble protein Total suspended solids Volatile suspended solids Sulfate
Whey permeate wastewater
5038 (28) 4959 (42) 4701 (39) 4686 (101) ndb nd 100 (5) 140 (23) 87 (3)
5406 5044 3733 3757 544 374 163 125 117
(128) (76) (78) (27) (22) (17) (18) (35) (6)
Liquefied sewage sludge 5869 (217) 5021 (27) 980 (17) 945 (24) 2169 (10) 2275 (83) 138 (18) 62 (18) 883 (2)
a Standard deviations are in parentheses. b nd, not detected.
2.2.
Bacterial strains
Seven bacterial strains were purchased from culture collections and cultivated as described by the suppliers. Aeromonas caviae (KCTC 1653) and Aeromonas hydrophila (KCTC 2358) were obtained from Korean Collection of Type Cultures (KCTC). Escherichia coli K12 (DSM 498), Clostridium sticklandii (DSM 519), Desulfotomaculum carboxydivorans (DSM 14880), Gelria glumatica (DSM 14054) and Thermomonas fusca (DSM 15424) were obtained from Deutsche Sammlung von Mikroorganismen und Zellkulturen (DSMZ). The strains were used for testing the specificity of primer/probe sets and/or for constructing the standard curves for real-time PCR assays.
2.3.
Extraction of DNA
DNA was extracted from pure- and mixed-culture samples using an automated nucleic acid extractor (Magtration System 6 GC, PSS, Japan). Samples for extraction were prepared as previously described (Yu et al., 2006). Possible PCR inhibitors and DNA from cell debris were eliminated from a sample by repeated centrifuging, decanting and resuspending. Prepared cell suspensions were loaded on the extractor with a proper Genomic DNA Purification Kit (PSS). The purified DNA was eluted with 100 mL of Tris-HCl buffer (pH 8.0) and stored at 20 C for subsequent analyses. All DNA extraction runs were performed in duplicate.
2.4.
DGGE and phylogenetic affiliation
Bacterial 16S rRNA gene fragment was PCR-amplified using the primers, BAC338F (50 -ACTCCTACGGGAGGCAG-30 ) and BAC805R (50 -GACTACCAGGGTATCTAATCC-30 ), targeting the domain Bacteria (Yu et al., 2005). For stabile melting behavior of amplicons, a 40-bp GC-clamp (50 -CGCCCGCCGCGCGCGGC GGGCGGGGCGGGGGCACGGGGGG-30 ) was attached to the 50 end of the forward primer, BAC338F (Muyzer et al., 1993). A touch-down PCR was conducted according to the following protocol: initial denaturation at 94 C for 10 min; 20 cycles of denaturation at 94 C for 30 s, annealing at 65e55 C (reducing the temperature by 0.5 C/cycle), and extension at 72 C for 1 min; additional 15 cycles of 94 C for 30 s, 55 C for 30 s, and 72 C for 1 min; final extension at 72 C for 7 min. DGGE
analysis was then performed in a D-code system (Bio-Rad, Inc., Hercules, CA). Fifteen microliters of each PCR product were loaded onto an 8% (w/v) acrylamide gel containing a 30e55% denaturant gradient, where 100% is defined as 7 M urea with 40% (v/v) formamide. After electrophoresis run at 150 V for 7 h in 1ⅹ TAE buffer, the gel was stained with ethidium bromide and then scanned under UV transillumination. Bands of interest were cut out of the gel and eluted in 40 mL of sterile DW. Two microliters of the eluted solution were PCR-amplified using BAC338F and BAC805R, with no GC-clamp attached. The PCR amplicons were gelpurified and cloned into pGEM-T Easy vector (Promega, Madison, WI). The cloned fragments were sequenced and compared against the GenBank and RDP databases. Sequence alignment and phylogenetic analysis were carried out using MEGA4 software (Tamura et al., 2007).
2.5.
Design and evaluation of primer/probe sets
Two real-time PCR primer/probe sets targeting the genus Aeromonas (AMN-set) or the species C. sticklandii (Cst-set) were newly designed based on the 16S rRNA gene sequence, using PrimeRose software (Ashelford et al., 2002) with the RDP database. Detection specificity of was first tested for each set in silico using OligoCheck software (Ashelford et al., 2002) and ProbeMatch program of the RDP database (Cole et al., 2003). Each probe was labeled with a 6-carboxyfluorescein (FAM) and a 6-carboxytetramethylrhodamine (TAMRA) at the 50 and 30 ends, respectively. Along with the new sets, a primer/probe set targeting the domain Bacteria (BAC-set) was used to detect total bacteria in the reactors (Yu et al., 2005). Table 2 summarizes the characteristics of the primer/probe sets used in this study. New primer/probe sets were further experimentally examined using the pure-culture total DNA from the corresponding target strains (Table 2). In addition to the test with target strains, non-specific amplifications of AMN- and Cstsets were tested using the non-target strains, available from culture collections, with minimal mismatches as potential false-positive candidate: T. fusca (DSM 15424) for AMN-set; D. carboxydivoran (DSM 14880) and G. glutamica (DSM 14054) for Cst-set. Near full-length 16S rRNA gene fragment was PCRamplified from each target or non-target test strain with BAC 8F (50 -AGAGTTTGATYMTGGCTCAG) and BAC1392R
2542
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 3 9 e2 5 4 9
Table 2 e Characteristics of the real-time PCR primer and probe sets used in this study. Set name/Target group AMN-set/Aeromonas
Cst-set/Clostridium sticklandii
BAC-set/Bacteria
Sequence (50 / 30 )a
Representative target strainsb
F: GCCTTGACATGTCTGGAA P: TCGGGAATCAGAACACAGGTGCT R: ACTATCGCTAGCTTGCAG F: CTTCGGGTCGTAAAGCT P: CCATAGGAGGAAGCCCCGGCTA R: AAGTTCACCAGTTTCAGAG F: ACTCCTACGGGAGGCAG P: TGCCAGCAGCCGCGGTAATAC R: GACTACCAGGGTATCTAATCC
Amplicon size (bp)
A. caviae (KCTC 1653) A. hydrophila (KCTC 2358)
286
C. sticklandii (DSM 519)
235
E. coli K12 (DSM 498)
468
a F, P, and R indicate the forward primer, the probe, and the reverse primer, respectively. b Culture collection numbers are in parentheses.
(50 -ACGGGCGGTGTGTRC) primers (Amann et al., 1995). The PCR amplicons were cloned into pGEM-T Easy vector (Promega), and sequencing confirmed. Each cloned plasmid was diluted to 106 copies ml1 and amplified by real-time PCR, in triplicate, to check the amplification specificity of AMN- or Cstset (Yu et al., 2005). Mass concentration of the plasmid DNA solution was determined using PicoGreen dye and converted to copy concentration as previously described (Yu et al., 2006). The target plasmids, carrying the correct sequences obtained from the representative target strains (Table 2), constructed above were used to test the applicability of the new primer/probe sets to real environmental systems. A certain quantity of each plasmid (final concentration, 106 copies m/L) was added into PCR-grade water and total DNA extraction from anaerobic sludge. The prepared solutions, in parallel with the negative controls with no addition of target DNA, were realtime PCR analyzed with AMN- or Cst-set in triplicate. The quantification results from the target-only solution and the target-spiked mix were compared using t-test.
2.6.
Real-time PCR and standard curve construction
Real-time PCR amplification was performed using a LightCycler 1.2 instrument (Roche Diagnostics, Mannheim, Germany) with the corresponding primer/probe sets listed in Table 2. A reaction mixture (20 mL) was prepared using the LightCycler TaqMan Master kit (Roche Diagnostics): 11 mL of PCR-grade water, 1 mL of the TaqMan probe (final concentration, 200 nM), 1 mL of each primer (final concentration, 500 nM), 4 mL of 5 reaction mix solution, and 2 mL of template DNA. Amplification was performed in a two-step thermal cycling procedure: initial denaturation for 10 min at 94 C followed by 40 cycles of 10 s at 94 C (for denaturation) and 30 s at 60 C (for combined annealing and extension). Quantitative standard curve for each primer/probe set was generated using a 10-fold serial dilution, ranging from 102 to 109 copies m/L, of the corresponding target plasmid which carries one copy of specific sequence for the set (see Section 2.4 for target plasmid construction). The target plasmid for BACset was constructed using the 16S rRNA gene from E. coli K12 as described above for AMN- and Cst-sets. Each dilution series was analyzed in triplicate by real-time PCR with the corresponding primer/probe set. Analyzed CT values were plotted against the logarithm of the corresponding initial template copy concentrations. The dynamic range of a standard curve
was determined based on the linear regression r2 greater than 0.99. For AMN-set, the equimolar mixture of two target plasmids, carrying different target sequences from two representative target strains (Table 2), was used as the standard solution. Copy concentration of target sequence was determined against the corresponding standard curve. All reactor DNA samples were analyzed with each set in duplicate.
2.7.
Analytical methods
Solids and COD were measured according to Standard Methods for the Examination of Water & Wastewater (APHAAWWA-WEF, 2005). Carbohydrate and protein were quantified by the phenol-sulfuric acid method (Dubois et al., 1956) and the Kjeldahl method (Zapsalis and Beck, 1986), respectively. Volatile fatty acids (VFAs, C2eC6) and ethanol were measured using a gas chromatograph (6890 Plus, Agilent, Palo Alto, CA) equipped with an Innowax capillary column (Agilent) and a flame ionization detector. Biogas content was analyzed using another identical gas chromatograph equipped with a HP-5 capillary column (Agilent) and a thermal conductivity detector. An ion chromatograph (Personal 790, Metrohm, Switzerland) equipped with a PRP-X300 column (Hamilton, Reno, NV) was used for measuring formate and lactate. Anions including sulfate were quantified using another identical ion chromatograph equipped with Metroseop A Supp 5-100 column (Metrohm). A fluorometer (TD-700, Turner Designs, Inc., Sunnyvale, CA) with the PicoGreen dsDNA Quantification Reagent (Molecular Probes, Eugene, OR) was used to measure DNA mass concentration (w/v). All experimental analyses mentioned were performed in duplicate.
2.8.
Nucleic acid accession numbers
DNA sequence data have been deposited in the GenBank database under accession numbers FJ372572eFJ372616 and EU090149eEU090165, EU090167 and EU090169.
3.
Results and discussion
3.1.
Bioreactor performance
Three reactors, treating synthetic glucose wastewater, whey permeate and liquefied sewage sludge, were named as G-,
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 3 9 e2 5 4 9
W- and S-reactors, respectively (Table 1). The SCOD contributions of carbohydrate and protein in substrate were 99.2 and 0% for G-reactor, 79.5 and 11.1% for W-reactor, and 20.1 and 68.0% for S-reactor. The nutritional ingredients in the wastewaters are markedly different. Glucose is the sole carbon source in the synthetic glucose wastewater, whereas whey permeate is mainly composed of lactose and milk protein (Tejayadi and Cheryan, 1995). The liquefied sewage likely comprises an even more complex mixture of organic compounds compared to the other wastewaters tested. The concentration profiles of residual substrates and acidogenic products in all reactors are shown in Fig. 1. The acidogenic period, of our particular interest, was defined as being from time 0 to when the total COD equivalent of acidogenic products reaches its maximum (refer to the shaded areas in Fig. 1). Complete degradation (97%) of initial SCOD was observed after 39 d in both G- and W-reactors, whereas the SCOD removal in S-reactor was as low as 66% even after 60 d of operation. Correspondingly, the cumulative methane
2543
productions in G- and W-reactors (6.5 and 6.8 L, respectively) were remarkably greater than that in S-reactor (2.8 L). A small amount of hydrogen (<0.06 L) was produced in G- and W-reactors within the first 1-d period and no more production afterward, whereas no detectable hydrogen production was observed in S-reactor (data not shown). In all trials, acetate and propionate were the first and second major acidogenic products, and a biphasic production of methane along with successive consumption of the major organic acids was observed (data not shown), agreeing with previous reports (Stams et al., 1998). During the start-up of G-reactor, soluble carbohydrate was degraded fast and exhausted (>99% removal) in 0.8 d. W-reactor consumed 98.0% of soluble carbohydrate and 82.0% of soluble protein for the same 0.8d period, and the rest were completely utilized by 1.2 d. On the other hand, in S-reactor, only 56.0 and 20.0% of soluble carbohydrate and soluble protein, respectively, were degraded during the first 1-d operation. Even at the end of acidogenic period (at 21 d; Fig. 1), 23.4% of soluble carbohydrate and 30.6%
Fig. 1 e Changes in chemical profiles during the anaerobic digestion in G-, W-, and S-reactors treating synthetic glucose wastewater, whey permeate, and liquefied sewage sludge, respectively. The shaded areas in the overall plots (left) defined as acidogenic periods are shown enlarged in detail (right).
2544
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 3 9 e2 5 4 9
of soluble protein remained untreated. This inferior efficiency of S-reactor can be ascribed to the less favorable nature of liquefied sewage sludge for anaerobic decomposition than the other wastewaters tested. In both G- and W-reactors, initial fermentation products were mainly acetate, lactate, and ethanol, and the lactate and ethanol accumulations were then further converted to acetate and propionate (Ahring, 2003). A rapid and high initial accumulation of lactate up to 2.4 g/L by 0.8 d (>4 times any other product) was shown only in W-reactor. This corresponds to the fact that lactate is readily produced from lactose, the major carbonaceous component in whey permeate, by homolactic bacteria under anaerobic conditions (Lewis and Yang, 1992). On the other hand, no lactate or ethanol was detected in S-reactor. At the end of acidogenic period, acetate and propionate accumulated up to 2.0 and 1.0 g/L, respectively, in G-reactor, and 2.7 and 0.8 g/L, respectively, in W-reactor. In both reactors, the two major products accounted for >90% of the maximum total acidogenic products accumulation, in COD equivalent. By contrast, in S-reactor, the acetate accumulation after 21 d was only 1.1 g/L, but its COD contribution to total acidogenic products was 80.0% due to the low concentrations (<0.1 g/L) of the other products throughout the trial period. From an engineering standpoint, W-reactor treating whey permeate presented favorable acidogenesis performance with rapid reaction rate, high acetate content, and high acetate to propionate ratio (Speece, 1996).
3.2.
Bacterial community structures
Changes in bacterial community structures were examined based on the DGGE profiles of the reactors. In total, 10 (GA1 to 10), 19 (WA1 to 19), and 16 (SA1 to 16) bands from the gels of the G-, W-, and S-reactors, respectively (Fig. 2), were cloned and sequenced to determine their phylogenetic affiliations with the GenBank database (Table 3). The DGGE profile of
W-reactor analyzed throughout the whole anaerobic digestion was reported in a recent study from the author’s group investigating the qualitative bacterial community shifts in the reactor (Lee et al., 2008). In this study, only focusing on acidogenic period (i.e., 0e4.7 d in W-reactor) to track down common acidogens in different reactors, the bands shown in the gel picture were relabeled for convenience and the original labeling information (as in the previous paper cited) is given in the legend of Fig. 2. Twenty five of the total 45 bands were closely related (>97% sequence similarity) to known bacterial species, and 13 of them were affiliated with the genus Clostridium often involved in anaerobic hydrolysis and fermentation of organic compounds (Bitton, 1999). GA3 and SA4 were closely related to Clostridium propionicum, which reduces lactate to propionate through the acrylate pathway (Akedo et al., 1983), with 99.8 and 98.9% similarities, respectively. The microbes deduced from those bands were thus likely to participate in propionate production. WA2, 4, and 6 showed high sequence similarities of 99.1e99.5% to Clostridium acetibutylicum, while the sequence of WA13 was identical to the 16S rRNA gene of a Clostridium tertium strain. Both Clostridium species are often found in anaerobic environments and produce acetate and butyrate through saccharolytic fermentation (Sneath et al., 1986). These indicate that the corresponding microorganisms were likely involved in fermentation of lactose, the principal carbohydrate in whey permeate. WA9 and 10 were closely related to Clostridium magnum, producing acetate from H2 þ CO2, formate or methanol (Bomar et al., 1991), with 98.9 and 98.4% similarities, respectively. Although this species can also use various carbohydrates to form acetate (Schink, 1984), WA9 and 10 appeared from 1.2 d with an abrupt drop in formate level after the exhaustion of carbohydrate in substrate (Figs. 1 and 2). It was thus inferred that the corresponding microorganisms to these bands likely participated in converting formate or H2 þ CO2, derived from the decomposition of substrate or
Fig. 2 e Bacterial DGGE profiles of the 16S rRNA gene PCR fragments from (A) G-, (B) W-, and (C) S-reactors treating synthetic glucose wastewater, whey permeate, and liquefied sewage sludge, respectively. Lane labels along the top show sampling time (days) from the initiation of reactor operation and anaerobic seed sludge (seed). For W-reactor bands, the original labeling information (as reported in (Lee et al., 2008)) is given in parenthesis below: WA1 (W1), WA2 (W5), WA3 (W16), WA4 (W6), WA5 (W14), WA6 (W7), WA7 (W15), WA8 (W21), WA9 (W10), WA10 (W8), WA11 (W9), WA12 (W19), WA13 (W2), WA14 (W3), WA15 (W17), WA16 (W11), WA17 (W4), WA18 (W12), W19 (W13).
2545
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 3 9 e2 5 4 9
Table 3 e Phylogenetic affiliation of the 16S rRNA gene sequences from DGGE bands. DGGE bands (sequence similarity) Closest species and taxon (accession no.) Aeromonas hydrophila (X87271) Anaerofilum agile (X98011) Azovibrio restrictus (AF011346) Clostridium acetibutylicum (X81021)
G-reactor
W-reactor
S-reactor
GA7 (100%)
WA17 (99.8%) WA16 (94.1%)
SA13 (99.9%)
GA8 (100%) WA2 (99.5%) WA4 (99.1%) WA6 (99.5%) WA9 (98.9%) WA10 (98.4%)
Clostridium magnum (X77835) Clostridium propionicum (X77841) Clostridium sp. 13A1 (AY554421) Clostridium sticklandii (M26494) Clostridium tertium (AJ245413) Desulfotomaculum reducens (CP000612) Desulfobotulus sapovorans (M34402) Desulfovibrio vulgaris (AF418179) Frigovirgula patagoniensis (AF450134) Klebsiella oxytoca (AY786181) Pseudomonas fluorescens (AJ583090) Streptococcus bovis (AF429766) Sulfurospirillum halorespirans (AF218076) Trichococcus flocculiformis (AJ306611) Ubc 16SX-2 (U27711) Ubc AKAU4083 (DQ125852) Ubc BSA2B-13 (AB175385) Ubc D7 (AY766467) Ubc E3 (AY426468) Ubc E16 (AY426460) Ubc Galb35 (AY193168) Ubc IA-5 (AJ488076) Ubc IIB-27 (AJ488087) Ubc JN18_A107_G (DQ168658) Ubc LCFA-B10 (AB244317) Ubc M13 (DQ378233) Ubc mek63d03 (AY537432)
GA3 (99.8%) GA6 (99.8%)
SA4 (98.9%) WA12 (100%) WA15 (99.8%)
SA9 (99.8%) SA10 (99.1%)
WA13 (100%) SA11 (97.5%) SA16 (96.3%) SA15 (100%) SA6 (98.6%) GA2 (99.4%) SA8 (99.4%) WA11 (100%) SA3 (99.1%) WA14 (100%) SA12 (100%) SA7 (88.9%) GA9 (100%) SA2 (99.3%) GA5 (100%)
GA1 (100%)
WA8 WA1 WA5 WA3
(99.8%) (99.8%) (98.5%) (100%) SA5 (99.4%) SA14 (91.2%)
GA4 (99.3%) GA10 (100%)
Ubc MTG-104 (DQ307696) Ubc RB016 (AB240286)
WA7 (98.9%) WA19 (100%) SA1 (92.2%) WA18 (99.6%)
Ubc, uncultured bacterial clone.
intermediates, to acetate. SA11 and 15 were closely related to Desulfotomaculum reducens and Desulfovibrio vulgaris, with 97.5 and 100% similarities, respectively. SA16 also shared a considerable similarity of 96.3% (although <97%) with Desulfobotulus sapovorans. Such appearance of sulfatereducing bacteria (SRB)-related bands only in S-reactor is likely due to the significantly greater concentrations of protein (potential S source) and sulfate in the substrate (Table 1). Fifteen of the total 45 bands were most closely related to environmental clones, whose existence and roles are yet unknown, and five were not closely related (<97% sequence similarity) to any database sequence. These indicate that many microbial populations in the reactors tested have not been characterized by conventional culture-dependent methods. Some of the uncultured clones showed prominent band intensity. GA1 and WA3 that appeared in the later acidogenic period showed 100% similarity to an uncultured clone from a bacterial consortium removing chlorobenzenes
(AJ488087; direct submission to GenBank). WA18 was most closely related (99.6% similarity) to a cloned sequence from the rhizosphere biofilm of a lab-scale reed bed reactor (AB240286; direct submission to GenBank). SA5, of which band intensity remained fairly constant from 1 d onwards, showed a high similarity of 99.4% to an uncultured clone from an anaerobic culture dechlorinating polychlorinated biphenyls (PCBs) (Bedard et al., 2006). The functions of the uncultured clones in their source environments are yet unclear and so are those of the corresponding microbes in the reactors tested in this study. Although the retrieved band sequences were distributed across five different phyla, almost all of them (95.6%; 43/45) were assigned to Bacteriodetes (24.4%), Firmicutes (51.1%), and Proteobacteria (20.0%) that are often abundant in wastewater treatment systems. The most interesting finding is the appearance of Aeromonas- and C. sticklandii-related bands, with outstanding intensities, in all reactors (Fig. 2 and Table 3). GA7, WA17, and
2546
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 3 9 e2 5 4 9
SA13 were closely related to A. hydrophila, the type species of the genus, with high similarities of 99.8e100%, and in addition, they also displayed high similarities greater than 99% to several other Aeromonas species. This corresponds to the fact that Aeromonas species share a greatly high full-length 16S rRNA gene sequence similarity (>97.8%) (Brenner et al., 2005; Laganowska and Kazanowski, 2004), resulting in difficulty in identifying them to species level based on 16S rRNA gene sequences. Aeromonas is often found in sewage and water environments, and ferments a variety of carbohydrate to produce organic acids and alcohols, such as acetate, formate, lactate, and ethanol (Brenner et al., 2005). Its strains also frequently occur, sometimes in high numbers, in municipal wastewater treatment processes and hence in activated sludge (Nsabimana et al., 1999). These, together with their dominant band intensities (although not robustly quantitative) during start-up, suggest that the Aeromonas-related microorganisms deduced from GA7, WA17, and SA13 were responsible for the rapid initial utilization of carbohydrate with the accumulation of acidogenic products (Figs. 1 and 2). On the other hand, GA6, WA15, SA9 and SA10 showed high similarities of 99.1e99.8% to C. sticklandii that only utilize specific amino acids in pairs via the Stickland reaction to form organic acids such as acetate, propionate, and butyrate (Sneath et al., 1986). The C. sticklandii-related microorganisms corresponding to these bands are thus suggested to be involved in the degradation of protein, a rich source of amino acids. Intriguingly, a C. sticklandii-related band (GA6) was also observed even in G-reactor that was supplied with no amino acid source in the substrate. This may indicate the presence of utilizable amino acids from non-substrate sources, potentially such as cell debris and extracellular excretions.
3.3.
Specificity evaluation of primer/probe sets
Aeromonas- and C. sticklandii, commonly detected acidogens in all reactors, were selected for further quantitative investigation. Given the putative roles and DGGE banding patterns, these populations were likely to be functionally important and numerically dominant in all reactors, regardless of different substrate compositions (Table 1). For quantitative real-time PCR assay, two primer/probe sets detecting the genus Aeromonas (AMN-set) and the species C. sticklandii (Cst-set) were newly developed in this study (Table 2). Because of the high 16S rRNA gene sequence homology among Aeromonas species (>97.8%), as aforementioned, AMN-set was designed to have a genus-level resolution, whereas Cst-set has a species-level resolution The in silico screening of false-negative and false-positive candidates were achieved based on the criteria described previously (Yu et al., 2005). A false-negative candidate is defined as a target strain carrying one or more mismatches with any of three oligonucleotides composing its corresponding primer/probe set. By contrast, a non-target strain carrying two or less mismatches with each of three oligonucleotides is considered as a false-positive candidate. The in silico assay identified a few false-positive but no false-negative candidates for either of the new sets. For convenience, the mismatch combinations were shown in square brackets in the following order: number of mismatches in the probe (P),
the forward primer (F), and the reverse primer (R). For AMNset, two false-positive candidates with [P2-F0-R1] and [P1-F0R2] mismatches, both of which were uncultured clones from zebrafish intestine (Rawls et al., 2006), were screened. On the other hand, nineteen false-positive candidates were identified for Cst-set, fifteen among which had 2 mismatches with each oligonucleotide (i.e., 6 mismatches with the set). Among the rest four, two with [P1-F2-R2] mismatches belonged to the genus Gelria and the other two carrying no mismatch were uncultured clones identified from oil-rich environments (Grabowski et al., 2005). It should be noted that the probability of false-positive detection declines as the number of mismatches increases, and a total of 5 or 6 mismatches in a set will greatly reduce the chance of non-specific amplification (Yu et al., 2005). Set specificity was then empirically verified using the representative target strains listed in Table 2 and the nontarget strains, available at present from culture collections, with minimal mismatches. Thermomona fusca (DSM 15424) with [P1-F0-R4] mismatches was used as the false-positive test strain for AMN-set, whereas G. glumatica (DSM 14054) with [P1-F2-R2] mismatches and D. carboxydivorans (DSM 14880) with [P2-F2-R2] mismatches were tested for Cst-set. Each set selectively amplified its target strains without amplifying non-target strains or negative control (data not shown). This indicates that only the non-target microorganisms, carrying fewer mismatches than the tested non-target strains, may cause false-positive detection. As shown above in the in silico test results, however, only two uncultured clones satisfying this condition were identified for each set. Moreover, the possibility of false-positive detection (i.e., non-specific amplification) decreases significantly in a complex mixedculture DNA sample since a specific primer/probe set will be far more selective to its correct target sequence (Becker et al., 2000). Consequently, for both sets, the false-positive detection leading to overestimation was suggested to be insignificant, especially in acidogenic systems.
3.4.
Applicability and nesting test of primer/probe sets
Most environmental samples are mixed cultures of extremely diverse microorganisms and consequently contain a large amount of various genes. Accordingly, unexpected amplification of random non-16S rRNA genes may not be thoroughly ruled out by the design and specificity evaluation procedures based on the 16S rRNA gene databases. To this end, the applicability of AMN- and Cst-sets to environmental samples was experimentally examined by spike-recovery test using anaerobic sludge collected from a full-scale anaerobic digester. A 16S rRNA gene clone library (290 bacterial clones) of the anaerobic sludge sample showed no clone affiliated with Aeromonas or C. sticklandii (Lee et al., 2009), indicating their very low population (<0.3% clone frequency) in the sludge tested. For AMN-set, the 16S rRNA gene concentrations quantified (n ¼ 3) in the target-only solution and in the targetspiked mix were 9.25 ⅹ 105 (sd, 7.3%) and 1.12 ⅹ 106 (sd, 14.3%) copies/mL, respectively. For Cst-set, the estimated concentrations (n ¼ 3) were 8.81 ⅹ 105 (sd, 1.8%) and 9.21 ⅹ 105 (sd, 5.3%) copies/mL in the target-only and the mixed samples, respectively. According to the t-test results for each set, the
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 3 9 e2 5 4 9
quantification results in the target-only solution and in the target-spiked mix were not different at 5% significance level. This suggests that there is little possibility of non-specific amplification or detection when applying the new sets to complex environmental samples like anaerobic sludge. Consequently, AMN- and Cst-sets were verified to be selectively specific to their corresponding target groups and readily applicable to real environmental systems without producing significant false results. Total DNA samples from the pure cultures of their target strains (Table 2), extracted within log-phase growth, were analyzed using either AMN- or Cst-set in parallel with BAC-set. Ideally, given the taxonomic hierarchy, the AMN- or Cst-set quantification results should equal the results obtained using BAC-set. The determined concentrations in all trials were in the range of 108e109 copies ml1 and highly reproducible with standard deviations of 1.8e4.5% in five replicates (data not shown). Moreover, for each sample, the CT values and target gene concentrations measured using different taxonomiclevel sets were not different at 5% significance level (t-test). These results indicate that the hierarchical nesting of the newly developed primer/probe sets was successfully done, suggesting that the quantitative monitoring of total bacteria and individual target populations can be performed using the new sets together with BAC-set.
3.5.
2547
start-up, in anaerobic digestion has not been reported, making it the most intriguing point of this paper. This finding hints that Aeromonas populations may play critical roles, at least in some case, in acids accumulation and reactor souring, in response to overloading (Gerardi, 2003). C. sticklandii initially constituted only 0.01, 0.16, and 0.10% of total bacterial 16S rRNA gene concentration in G-, W-, and S-reactors, respectively. Its concentration increased up to 4.37 ⅹ 107, 1.34 ⅹ 107, and 2.89 ⅹ 107 copies/mL after 3.3, 4.7, and 5.6 d (ca. 13,000e40,000 fold rise) in G-, W-, and S-reactors, respectively. At the maximum points, the C. sticklandii to total bacteria ratio was 8.5, 55.2, and 17.1% in G-, W-, and S-reactors, respectively. These indicate that C. sticklandii, which only utilize amino acids, was abundant in all reactors, even
Quantitative dynamics of target acidogens
Bacterial community dynamics in the reactors were quantitatively analyzed using real-time PCR with AMN-, Cst-, and BAC-sets (Fig. 3). Both Aeromonas and C. sticklandii 16S rRNA gene concentrations exhibited dynamic profiles, characterized by a steep increase during the initial phase, in all reactors. Aeromonas at time 0 occupied 0.02, 0.05, and 0.08% of the total bacterial 16S rRNA gene concentration, in G-, W-, and S-reactors, respectively. However, its population increased explosively (ca. 37,000e170,000 fold rise) with the rapid degradation of carbohydrate, and reached up to 1.37 ⅹ 109, 6.60 ⅹ 108, and 2.26 ⅹ 108 copies/mL after 0.8, 0.6 and 0.6 d in G-, W-, and S-reactors, respectively (Figs. 1 and 3). At the peak points, Aeromonas accounted for 86.6, 87.6, and 95.3% of the total bacterial 16S rRNA gene concentration in G-, W-, and S-reactors, respectively. Such dominance did not last after the completion of acidogenesis (data not shown). In G-reactor, the 16S rRNA gene ratio of Aeromonas to total bacteria declined down to <10% by 20 d. The ratio dropped down to <10% only by 4 d and 3 d in W- and S-reactors, respectively, showing the tendency to decrease faster under complex substrate conditions. It continued to decrease to reach <1% level in all reactors at the end of incubation, indicating that Aeromonas population declined and was replaced by other bacteria during methanogenic period. These results suggests that Aeromonas populations (deduced from DGGE bands GA7, WA17, SA13) likely formed the most important acidogen group, at least in terms of abundance, for acidogenic initiation in all reactors. Given the physiological characteristics of Aeromonas, together with process performance data, this group is suggested to be mainly responsible for the rapid fermentation of carbohydrate, building up organic acids and alcohols. Such striking emergence and absolute dominance of Aeromonas, during
Fig. 3 e Changes in the 16S rRNA gene concentrations of bacterial populations in (A) G-, (B) W-, (C) S-reactors treating synthetic glucose wastewater, whey permeate, and liquefied sewage sludge, respectively. Each symbol shows the following: Bacteria (filled circles); Aeromonas (open triangles); and C. sticklandii (filled diamonds).
2548
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 3 9 e2 5 4 9
including G-reactor provided with no amino acid source. Noteworthy is that the C. sticklandii 16S rRNA gene concentration began to increase after a lag of 0.7 d only in G-reactor, whereas it increased promptly with no lag in W- and S-reactors treating protein-containing wastewater. In addition, only in G-reactor, the increase in C. sticklandii population occurred concurrently with the decrease in total bacteria, particularly Aeromonas (Fig. 3A). On the other hand, in W-reactor, C. sticklandii continued to increase even after running out of substrate protein at 1.2 d until the end of acidogenic period at 4.7 d, showing a 4.3-fold increase in the 16S rRNA gene concentration during that 3.5-d period (Fig. 3B). These observations circumstantially support the view that the growth of C. sticklandii in the reactors was supported, in part, by amino acid sources released from cell debris or extracellular excretions. Although, in this study, the DGGE profiles were generally in accordance with the real-time PCR results, some apparent discrepancy did exist. For example, in each reactor, the bands related to C. sticklandii were not detected on the gel during the initial period while this group was actively growing and present in considerable concentrations of >105 copies/mL (Figs. 2 and 3). In G-reactor, although C. sticklandii began to increase rapidly after 0.7 d (Fig. 3A), a C. sticklandii-related band GA6 appeared from 2.4 d (Fig. 2A). The 16S rRNA gene ratio of C. sticklandii to total bacteria was 0.1% at 1.4 d (i.e., the DGGE point right before the appearance of GA6) whereas the ratio was 7.0% at 2.4 d (Fig. 3A). In W-reactor, additionally, a C. sticklandii-related band WA15 started to show up from 2.2 d (Fig. 2B) in spite of the active increase in the C. sticklandii 16S rRNA gene concentration from time 0 (Fig. 3B). The contribution of C. sticklandii to total bacteria, in terms of 16S rRNA gene concentration, was 1.7% and 3.9% at 1.2 and 2.2 d, respectively. This discrepancy is expectable and can be ascribed to the inherent limitation of DGGE that a numerically minor population (ca. < 1% of total community) is typically missed out due to the PCR bias (after many thermal cycles; typically >30), which is a common problem to PCR-based molecular fingerprinting methods (Forney et al., 2004). This limitation readily causes the omission of a numerically minor but functionally vital community member, leading to mislinking microbial information with process performance data, especially when studying dynamic communities. Therefore, more quantitative studies, in addition to qualitative studies, are desired for a better understanding of environmental microbial communities.
4.
Conclusions
This study demonstrated that Aeromonas and C. sticklandii populations were common and abundant in all anaerobic reactors tested, irrespective of difference in wastewater composition. Their qualitative and quantitative dynamics associated with changes in process performance data, together with their putative functions, suggest that both are not only numerically but also functionally important acidogens in initiating anaerobic digestion. This observation, although one seeding source was used, indicates that these acidogen populations might appear in abundance in other
anaerobic digestion systems treating different types of wastewaters. Although a few have studied bacterial community structures and dynamics under start-up or substrate perturbation conditions and identified diverse acidogens (Chachkhiani et al., 2004; Delbes et al., 2000; Fernandez et al., 2000; Sundh et al., 2003), explosive emergence and/or overwhelming dominance of Aeromonas have not yet been quantitatively observed. To the authors’ knowledge, this is the first, or one of the first, reports on the identification and monitoring of common key acidogens initiating anaerobic digestion of different wastewaters. The design, verification, and application of new real-time PCR primer/probe sets detecting the common key acidogens (i.e., AMN- and Cst-sets) were also first carried out in this study, and the sets are ready for practical application.
Acknowledgments This work was supported by the Korea Ministry of Environment (MOE) as ‘Human resource development Project for Energy from Waste & Recycling’, by the Korea Research Foundation Grant funded by the Korean Government (MOEHRD), and by the Korea government Ministry of Knowledge Economy as the New & Renewable Energy of the Korea Institute of Energy Technology Evaluation and Planning (KETEP) grant funded by the Korea government Ministry of Knowledge Economy (Grant no. 20101T100100366). This work was also partly supported by a Science Foundation Ireland (SFI) Charles Parsons Energy Research Award (06/CP/E006) and by a Start-Up Grant (M58030010) from Nanyang Technological University.
references
Ahring, B.K., 2003. Biomethanation I. Springer, New York, USA. Akarsubasi, A.T., Ince, O., Kirdar, B., Oz, N.A., Orhon, D., Curtis, T.P., Head, I.M., Ince, B.K., 2005. Effect of wastewater composition on archaeal population diversity. Water Res. 39, 1576e1584. Akedo, M., Cooney, C.L., Sinskey, A.J., 1983. Direct demonstration of lactate-acrylate interconversion in clostridium propionicum. Biotechnology 1, 791e794. Amann, R.I., Ludwig, W., Schleifer, K., 1995. Phylogenetic identification and in situ detection of individual microbial cells without cultivation. Microbiol. Rev. 59, 143e169. APHA-AWWA-WEF, 2005. Standard Methods for the Examination of Water and Wastewater. American Public Health Association, Washington, D.C. Ashelford, K.E., Weightman, A.J., Fry, J.C., 2002. Primrose: a computer program for generating and estimating the phylogenetic range of 16S rRNA oligonucleotide probes and primers in conjunction with the RDP-II database. Nucl. Acids Res. 30, 3481e3489. Becker, S., Boeger, P., Oehlmann, R., Ernst, A., 2000. Pcr bias in ecological analysis: a case study for quantitative taq-nuclease assays of microbial communities. Appl. Environ. Microbiol. 66, 4945e4953. Bedard, D.L., Bailey, J.J., Reiss, B.L., Jerzak, G.V.S., 2006. Development and characterization of stable sediment-free anaerobic bacterial enrichment cultures that dechlorinate aroclor 1260. Appl. Environ. Microbiol. 72, 2460e2470.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 3 9 e2 5 4 9
Bitton, G., 1999. Wastewater microbiology. John Wiley & Sons, Inc., New York. Bomar, M., Hippe, H., Schink, B., 1991. Lithotrophic growth and hydrogen metabolism by clostridium magnum. FEMS Microbiol. Lett. 67, 347e349. Brenner, D.J., Krieg, N.R., Staley, J.T., 2005. Bergey’s Manual of Systematic Bacteriology: Proteobacteria. Springer, New York. Chachkhiani, M., Dabert, P., Abzianidze, T., Partskhaladze, G., Tsiklauri, L., Dudauri, T., Godon, J.J., 2004. 16S rDNA characterization of bacterial and archaeal communities during start-up of anaerobic thermophilic digestion of cattle manure. Biores. Technol. 93, 227e232. Cole, J.R., Chai, B., Marsh, T.L., Farris, R.J., Wang, Q., Kulam, S.A., Chandra, S., McGarrell, D.M., Schmidt, T.M., Garrity, G.M., Tiedje, J.M., 2003. The ribosomal database project (RDP-II): previewing a new autoaligner that allows regular updates and the new prokaryotic taxonomy. Nucl. Acids Res. 31, 442e443. Delbes, C., Moletta, R., Godon, J.-J., 2000. Monitoring of activity dynamics of an anaerobic digester bacterial community using 16S rRNA polymerase chain reaction-single-strand conformation polymorphism analysis. Environ. Microbiol. 2, 506e515. Dubois, M., Gilles, K.A., Hamilton, J.K., Rebers, P.A., Smith, F., 1956. Colorimetric methods for determination of sugars and related substances. Anal. Chem. 28, 350e356. Fernandez, A.S., Hashsham, S.A., Dollhopf, S.L., Raskin, L., Glagoleva, O., Dazzo, F.B., Hickey, R.F., Criddle, C.S., Tiedje, J.M., 2000. Flexible community structure correlates with stable community function in methanogenic bioreactor communities perturbed by glucose. Appl. Environ. Microbiol. 66, 4058e4067. Forney, L.J., Zhou, X., Brown, C.J., 2004. Molecular microbial ecology: land of the one-eyed king. Curr. Opin. Microbiol. 7, 210e220. Gerardi, M.H., 2003. Wastewater Microbiology Series. Wileyinterscience, New Jersey. Grabowski, A., Nercessian, O., Fayolle, F., Blanchet, D., Jeanthon, C., 2005. Microbial diversity in production waters of a low-temperature biodegraded oil reservoir. FEMS Microbiol. Ecol. 54, 427e443. Hori, T., Haruta, S., Ueno, Y., Ishii, M., Igarashi, Y., 2006. Dynamic transition of a methanogenic population in response to the concentration of volatile fatty acids in a thermophilic anaerobic digester. Appl. Environ. Microbiol. 72, 1623e1630. Hwang, S.H., Hansen, C.L., 1992. Performance of upflow anaerobic sludge blanket (UASB) reactor treating whey permeate. Trans. ASAE 35, 1665e1671. Kim, J., Park, C., Kim, T.H., Lee, M., Kim, S., Kim, S.W., Lee, J., 2003. Effects of various pretreatments for enhanced anaerobic digestion with waste activated sludge. J. Biosci. Bioeng. 95, 271e275. Laganowska, M., Kazanowski, A., 2004. Restriction fragment length polymorphism of 16S-23S rDNA intergenic spacer of Aeromonas spp. Syst. Appl. Microbiol. 27, 546e557. Leclerc, M., Delbes, C., Moletta, R., Godon, J., 2001. Single strand conformation polymorphism monitoring of 16S rDNA archaea during start-up of an anaerobic digester. FEMS Microbiol. Ecol. 34, 213e220. Lee, C., Kim, J., Chinalia, F., Shin, S., Hwang, S., 2009. Unusual bacterial populations observed in a full-scale municipal sludge digester affected by intermittent seawater inputs. J. Ind. Microbiol. Biotechnol. 36, 769e773. Lee, C., Kim, J., Shin, S.G., Hwang, S., 2008. Monitoring bacterial and archaeal community shifts in a mesophilic anaerobic
2549
batch reactor treating a high-strength organic wastewater. FEMS Microbiol. Ecol. 65, 544e554. Lewis, V.P., Yang, S., 1992. Propionic acid fermentation by Propionibacterium acidipropionici: effect of growth substrate. Appl. Microbiol. Biotechnol. 37, 437e442. McMahon, K.D., Stroot, P.G., Mackie, R.I., Raskin, L., 2001. Anaerobic codigestion of municipal solid waste and biosolids under various mixing conditions e II: microbial population dynamics. Water Res. 35, 1817e1827. McMahon, K.D., Zheng, D., Stams, A.J.M., Mackie, R.I., Raskin, L., 2004. Microbial population dynamics during start-up and overload conditions of anaerobic digesters treating municipal solid waste and sewage sludge. Biotechnol. Bioeng. 87, 823e834. Muyzer, G., Waal, E.C., Uitterlinden, A.G., 1993. Profiling of complex microbial populations by denaturing gradient gel electrophoresis analysis of polymerase chain reactionamplified genes coding for 16S rRNA. Appl. Environ. Microbiol. 59, 695e700. Nsabimana, E., Belan, A., Bohatier, J., 1999. Analysis at the genomospecies level of microbial populations changes in activated sludge: the case of Aeromonas. Water Res. 34, 1696e1704. Rawls, J.F., Mahowald, M.A., Ley, R.E., Gordon, J.I., 2006. Reciprocal gut microbiota transplants from zebrafish and mice to germfree recipients reveal host habitat selection. Cell 127, 423e433. Schink, B., 1984. Clostridium magnum sp. nov., a non-autotrophic homoacetogenic bacterium. Arch. Microbiol. 137, 250e255. Sneath, P.H.A., Mair, N.S., Sharpe, M.E., 1986. Bergey’s Manual of Systematic Bacteriology. Williams & Wilkins, Baltimore. Speece, R.E., 1996. Anaerobic Biotechnology for Industrial Wastewaters. Archae Press, Nashville, TN. Stams, A.J.M., Dijkema, C., Plugge, C.M., Lens, P., 1998. Contribution of 13C-NMR spectroscopy to the elucidation of pathways of propionate formation and degradation in methanogenic environments. Biodegradation 9, 463e473. Sundh, I., Carlsson, H., Nordberg, A., Hansson, M., Mathisen, B., 2003. Effects of glucose overloading on microbial community structure and biogas production in a laboratory-scale anaerobic digester. Bioresour. Technol. 89, 2003. Tamura, K., Dudley, J., Nei, M., Kumar, S., 2007. MEGA4: molecular evolutionary genetics analysis (MEGA) software version 4.0. Mol. Biol. Evol. 24, 1596e1599. Tejayadi, S., Cheryan, M., 1995. Lactic acid from cheese whey permeate. Productivity and economics of a continuous membrane bioreactor. Appl. Microbiol. Biotechnol. 43, 242e248. Yu, Y., Kim, J., Hwang, S., 2006. Use of real-time pcr for groupspecific quantification of aceticlastic methanogens in anaerobic processes: population dynamics and community structures. Biotechnol. Bioeng. 93, 424e433. Yu, Y., Lee, C., Kim, J., Hwang, S., 2005. Group-specific primer and probe sets to detect methanogenic communities using quantitative real-time polymerase chain reaction. Biotechnol. Bioeng. 89, 670e679. Zapsalis, C., Beck, R.A., 1986. Food Chemistry and Nutritional Biochemistry. Macmillan Publishing Co., NewYork. Zumstein, E., Moletta, R., Godon, J., 2000. Examination of two years of community dynamics in an anaerobic bioreactor using fluorescence polymerase chain reaction (PCR) singlestrand conformation polymorphism analysis. Environ. Microbiol. 2, 69e78.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 5 0 e2 5 6 0
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Fate of N-nitrosomorpholine in an anaerobic aquifer used for managed aquifer recharge: A column study M.M. Pitoi a,b, B.M. Patterson a,b,*, A.J. Furness a, T.P. Bastow a, A.J. McKinley b a b
CSIRO Land and Water, Private Bag 5, Wembley, WA 6913, Australia School of Biomedical, Biomolecular and Chemical Sciences, University of WA, Crawley, WA 6009, Australia
article info
abstract
Article history:
The fate of N-nitrosomorpholine (NMOR) was evaluated at microgram and nanogram per
Received 3 December 2010
litre concentrations. Experiments were undertaken to simulate the passage of groundwater
Received in revised form
contaminants through a deep anaerobic pyritic aquifer system, as part of a managed
4 February 2011
aquifer recharge (MAR) strategy. Sorption studies demonstrated the high mobility of NMOR
Accepted 12 February 2011
in the Leederville aquifer system, with retardation coefficients between 1.2 and 1.6.
Available online 19 February 2011
Degradation studies from a 351 day column experiment and a 506 day stop-flow column experiment showed an anaerobic biologically induced reductive degradation process
Keywords:
which followed first order kinetics. A biological lag-time of less than 3 months and
N-Nitrosomorpholine
a transient accumulation of morpholine (MOR) were also noted during the degradation.
Degradation
Comparable half-life degradation rates of 40e45 days were observed over three orders of
Recycled water
magnitude in concentration (200 ng L1 to 650 mg L1). An inhibitory effect on microor-
N-Nitrosamines
ganism responsible to the biodegradation of NMOR at 650 mg L1 or a threshold effect at
MAR
200 ng L1 was not observed during these experiments. Crown Copyright ª 2011 Published by Elsevier Ltd. All rights reserved.
1.
Introduction
The high demand for water of drinking quality and increasing wastewater flow, as the consequences of population growth, have resulted in investigations of Managed Aquifer Recharge (MAR) using recycled water. However, potential trace organic contaminants in recycled water, sourced from wastewater or produced during disinfection, present a risk to recycled water use for MAR. If these trace organics are not removed, they may contaminate the aquifer and subsequently be present in the groundwater extracted from the aquifer. While reverse osmosis (RO) treatment removes many of the larger hydrophobic compounds from recycled water (Steinle-Darling et al., 2007), a number of disinfection byproducts such as nitrosamines are poorly removed because these molecules are small, polar and uncharged (Mitch et al.,
2003; Steinle-Darling et al., 2007). One of nitrosamines which has become a concern in recycled water is N-nitrosomorpholine (NMOR). NMOR contamination of recycled water is of concern due to health risks if the water is to be used for drinking. NMOR has been classified as a possible human carcinogen (IARC, 1987) and a drinking water guideline value of 10 ng L1 has been recommended in Germany (Planas et al., 2008). Chemical and physical properties of NMOR are given in Table 1. NMOR has been detected in the human environment notably in association with the rubber industry (de Vocht et al., 2007; Fajen et al., 1979; Spiegelhalder and Preussmann, 1983). It has also been found in wastewater effluent (Krauss and Hollender, 2008; Krauss et al., 2009; Planas et al., 2008; Schreiber and Mitch, 2006b) with concentrations up to 12.7 mg L1 (Krasner et al., 2009). Moreover, NMOR has been reported as one of the most abundant nitrosamines in primary
* Corresponding author. CSIRO Land and Water, Private Bag 5, Wembley, WA 6913, Australia. Tel.: þ61 8 93336276; fax: þ61 8 9333 6211. E-mail address:
[email protected] (B.M. Patterson). 0043-1354/$ e see front matter Crown Copyright ª 2011 Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.02.018
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 5 0 e2 5 6 0
Table 1 e Properties of NMOR. O O
N N
2551
comparable fate data; and (iii) changes in degradation rates during longer-term experiments and relate these data to degradation lag-times and bacterial acclimation.
Chemical structure Molecular formula CAS number Molecular weight Colour/form Boiling point Melting point Octanol/water partition coefficient Density (specific gravity) Henry’s law constant Solubility
C4H8N2O2 59-89-2 116.12 Yellow crystal 139e140 C at 25 mmHg 29 C Log Kow ¼ 0.44 0.791 g cm3 2.45 108 atm m3 mol1 at 37 C Miscible in water in all prop; soluble in organic solvent
All properties are taken from Hazardous Substance Data Bank (HSDB, 2003).
and secondary wastewater effluent (Krauss and Hollender, 2008; Krauss et al., 2009). NMOR can be formed during disinfection, particularly chloramination (Schreiber and Mitch, 2006a), and ozonation followed by chlorination (Zhao et al., 2008). However, the formation of NMOR during disinfection was usually lower compare to N-nitrosodimethylamine (NDMA), another nitrosamine which was detected frequently in disinfected wastewater (Krasner et al., 2009). Moreover, NMOR is less degradable in sewage treatment plants when compared to other nitrosamines such as NDMA, N-nitrosodiethylamine, N-nitroso-di-nbutylamine, and N-nitrosopiperidine (Hollender et al., 2009; Krauss et al., 2009). Activated sludge and sand filtration treatments have been reported to reduce NMOR concentrations in wastewater. However these methods are only partially effective. From 21 wastewater treatment plants in Switzerland, the average percent removal of NMOR by activated sludge treatment was 65% and by sand filtration was only 34% (Krauss et al., 2009). The fate of NMOR at microgram per litre concentrations in aquifer have been previously assessed under anaerobic and aerobic conditions via column studies. These studies have shown limited degradation with degradation half-lives of >100 days compared to a sterile control under anaerobic condition (Patterson et al., 2010) and >50 days under aerobic condition (Patterson et al., 2011). Moreover, sorption studies of NMOR showed a low retardation coefficient of 1.0 for an aerobic soil with low organic carbon content of 0.02% (w/w) (Patterson et al., 2011) and 1.2 for an anaerobic soil with a higher organic carbon content of 0.32% (w/w) (Patterson et al., 2010). Low retardation coefficients suggest that NMOR is highly mobile in the different aquifer systems. In this paper we report further experiments designed to mimic the injection of recycled water containing NMOR into an aquifer under anaerobic conditions. These experiments examine: (i) the degradation of NMOR in an aquifer at ng L1 concentrations that are more typical of recycled water; (ii) the concentration effects on degradation rates and determine if higher concentrations (mg L1 concentrations) provide
2.
Materials and methods
2.1.
Chemicals
NMOR, morpholine (MOR) and sodium azide were sourced from SigmaeAldrich (Sydney, Australia). d8-NMOR was sourced from CDN isotopes (Honsby, Australia) and sodium bromide was sourced from Hayashi Pure Chemical Ind. Ltd. (Osaka, Japan).
2.2.
Aquifer sediment
Anaerobic Leederville aquifer sediment used in the column experiments was collected from a trial MAR site in Perth, Western Australia. The sediment was collected via rotary auger and coring from the confined Leederville aquifer on the Swan Coastal Plain of Western Australia over the proposed MAR injection depth interval (between 120 m and 220 m below ground level). The sediment in this zone consisted of discontinuous interbedded sands, silts and clays (Playford et al., 1976). To prevent sediment oxidation, collected sediment was immediately stored in either sealed air-tight 4 L tins flushed with nitrogen or sealed air-tight 10 L buckets saturated with anaerobic groundwater and kept at 4 C. The sediment mineralogy, based on X-ray defraction (XRD), was predominantly quartz (72%), and K-feldspar (24%), with minor quantities of pyrite (2%) and Na-feldspar (2%). All other minerals were below analytical detection (<1%) (Patterson et al., 2010). The sediment organic matter (SOM) content was 0.32% w/w and the chromium reducible sulphur (SCr) content of 0.71% w/w, which is consistent with the XRD pyrite content (Patterson et al., 2010). Particle size distribution of the sediment was >1 mm (10%); 1 mm to 500 mm (20%); 500e250 mm (50%); 250e125 mm (17%) and <125 mm (3%).
2.3.
Column influent water
Column influent water was collected from the Beenyup microfiltration/RO pilot plant, located at the Beenyup Wastewater Treatment Plant, Perth Western Australia. A chemical analysis of the influent water is given in Table 2.
Table 2 e Chemical analysis of influent water. Chloride Sulphate Nitrate-N Sodium Potassium Magnesium Calcium Conductivity Dissolved oxygen Dissolved organic carbon (DOC)
3.0 mg L1 0.38 mg L1 <1 mg L1 5.0 mg L1 0.55 mg L1 0.18 mg L1 1.6 mg L1 34 mS cm1 8.5 mg L1 <1 mg L1
2552
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 5 0 e2 5 6 0
2.4.
Experimental setup
2.4.1.
Low NMOR concentration column setup
Two stainless steel columns were constructed for the low NMOR concentration experiments. One column was used as a sterile control column, while the other column was used as a non-sterile column to differentiate between abiotic and biotic processes. Each column was 100 cm in length with an internal diameter of 14.5 cm. To avoid sediment migrating into the influent and effluent tubing, a stainless steel grate with holes 1.0 cm in diameter and stainless steel mesh (0.1 cm diameter) was fixed at the bottom and the top of each column. The base of each column, below the stainless steel mesh, provided a 500 mL mixing chamber to collect influent water prior to entering the column. Eleven sampling ports were placed at 0 (just below the stainless steel mesh), 4, 11, 18, 25, 32, 39, 46, 59, 75, and 92 cm from the base of each column for the collection of water samples. Each sampling port consisted of a 0.4 cm i.d. stainless steel tube that protruded 6.0 cm from the wall of the column into the centre of the column. The inner end of the tube contained a stainless steel mesh (0.1 cm diameter) to prevent sediment entering, while the outer end contained a silicon septum allowing a syringe needle to be inserted for the collection of water samples. The columns were wet packed with homogenized sediment while the column gas space was flushed with nitrogen gas to maintain anaerobic conditions. During column packing anaerobic groundwater from the Leederville aquifer was added to the column to maintain a water level above the packed sediment to prevent sediment stratification during packing. The columns were then operated in a saturated up flow mode. To confirm the permeability of the sediment-packed columns, anaerobic groundwater was pumped through the columns for two weeks at a flow rate of 5000 mL d1, then the flow was reduced to 190 mL d1. Anaerobic groundwater flow at 190 mL d1 continued for a period of two months to stabilize the groundwater chemistry in the columns. After two months the groundwater was replaced with recycled water to start the experiment. Influent water for the experimental column was stored in a 5 L SKC Flexfoil Grab Bag and spiked with NMOR and sodium bromide (conservative tracer). The final concentration of NMOR was 200 ng L1 while the concentration of sodium bromide was 50 mg L1 in all columns. Influent water for the control column was the same as the experimental column influent water, except for the inclusion of sodium azide (final concentration of 0.65 g L1) to inhibit biological activity in the influent water and on the sediment and column infrastructure. To reduce the potential for NMOR degradation prior to injection into the columns, fresh influent water solutions were prepared every two weeks. Analysis of influent water after the two week storage period showed that concentration of NMOR in sample bags remained at >95% of the initial concentration for both the sterile and non-sterile columns. The influent water solution was pumped through each column using a separate MCP Standard drive pump (ISMATEC) at a flow rate of 190 mL d1, giving a linear velocity of
approximately 2 cm d1, based on a porosity of 0.57 estimated from the bromide tracer test.
2.4.2.
High NMOR concentration column setup
Two columns were used for the high NMOR concentration experiment with a similar setup to the low NMOR concentration experiments, except that the columns were 2 m long. Seventeen sampling ports were placed at 0, 4, 11, 18, 25, 32, 39, 46, 59, 75, 92, 109, 125, 142, 158, 175 and 192 cm from the base of each column for the collection of water samples. For this experiment, an MCP Standard drive pump (ISMATEC) semi-continuously injected the influent water from a 3 L SKC Flexfoil Grab Bag into each of the column’s recycled water influent lines immediately prior to entering the column. A fresh solution of NMOR and bromide (conservative tracer) was prepared every two weeks with final concentration of 650 ng L1 and 50 mg L1 respectively. Sodium azide (final concentration of 0.65 g L1) was added into the influent water of sterile control column. The flows of recycled water through the columns were regulated at the effluent by a peristaltic pump (ISMATEC Reglo) to give a column flow of approximately 360 mL d1. Further details of the column setup have been reported in Patterson et al. (2010).
2.4.3.
High NMOR concentration stop-flow column setup
To increase column water residence time within the columns, water flow and trace organic delivery to the high NMOR concentration experiment columns were ceased after w12 months of continuous flow.
2.5.
Water sample collection
For the low NMOR concentration column experiment, water samples from each sampling port of the columns were collected using a 100 mL gas tight syringe (SGE). For NMOR analysis, 50 mL water samples were stored in Pyrex screw cap test tubes with Teflon lined septa caps (Corning) prior to extraction and analysis. For the high NMOR concentration experiment, a minimum volume was taken from each sampling port (5 mL) to minimize perturbing effects on the columns. A 50 mL sub-sample was diluted to 50 mL using MilliQ water and then processed similarly to the samples from low NMOR concentration column experiment. The remainder of the sample was used for MOR and bromide analysis. For MOR and bromide analyses, each water sample was filtered using a Bulk Aerodisc 32 mm filter with 0.45 mm Supor membrane (PAL Life Science) and stored in a 750 mL polypropylene vial with a PTFE/silicone liner cap, prior to analysis by ion chromatography.
2.6.
Analytical methods
Analysis for NMOR was carried out using a method developed by Ranwala (2009) which involved micro liquideliquid extraction followed by GC-MS analysis. This method had a limit of detection of 2 ng L1 with the recovery of 40 10%. A 50 mL water sample collected for NMOR analysis was spiked with d8-NMOR as a surrogate standard. The water sample was then extracted with dichloromethane (DCM).
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 5 0 e2 5 6 0
The extraction was carried out in triplicate where 10 mL of fresh DCM was used for each extraction. DCM containing NMOR was then dried with anhydrous sodium sulphate and evaporated to 2 mL. A clean-up step was subsequently applied to the 2 mL sample in order to remove interferences. For the clean-up step, the sample (2 mL) was loaded onto a small silica gel column and eluted with 5% of acetone in DCM (2 mL) followed by 10% of acetone in DCM (2 mL). The first 4 mL of the eluent was discharged and the last 2 mL (10% of acetone in DCM) containing NMOR was collected, concentrated to 50 mL, an internal standard added (d8naphthalene, 2.5 ng) and the sample subjected to GC-MS analysis. GC-MS analysis was carried out with an Agilent 5975 MSD, interfaced with an Agilent 6890 gas chromatograph fitted with fused-silica open tubular column coated with a AT-5 ms stationary phase (Alltech, 60 m 0.25 mm i.d., 0.25 mm film thickness). The GC oven temperature was programmed from 35 C (1 min) to 150 C at 5 C min1 and then to 280 C (5 min) at 30 C min1. Samples for analysis were injected (1.2 mL) into a vaporising injector (260 C) operated in splitlesis mode (0.4 min) using an autosampler. Helium was used as carrier gas at a constant flow of 1.2 mL min1. Typical MSD conditions were: ionisatioin energy 70 eV; source temperature 230 C; electron multiplier voltage 1700 V. The mass spectrometer was run in selected ion monitoring (SIM) mode. The ions monitored for NMOR were 116 and 94 m/z. The peaks in the chromatograms were integrated using MSD Chemstation D 2.00.275 and Microsoft Excel software. Filtered water samples (10 mL) were analysed for MOR using a Dionex ICS-3000 reagent free ion chromatography (RFIC) system equipped with a CG16 5 mm column. Methane sulfonic acid was used as an eluent with the flow of 1 mL min1. The eluent concentration was 25 mM for the first 22 min then increased at 0.28 mM per minute to 30 mM over a period of 18 min. The temperature of the column was set at 40 C. For bromide analysis an internal standard was added to the filtered water sample prior to analysis, i.e. 50 mL of 200 mg L1 of LiF was added to 500 mL of a filtered water sample. 10 mL of the mixture was then analysed using a Dionex ICS-3000 RFIC with an AG181 4 mm column. A 33 mM potassium hydroxide eluent was used at a flow rate of 1 mL min1. The column temperature was set at 40 C.
3.
Results and discussion
3.1.
Characteristics of Leederville sediment
The high potential reductive capacity of the sediment (890 mM O2 g1 sediment; Patterson et al., 2010) and the rapid oxygen consumption rate (2.0 mg L1 h1; Patterson et al., 2010) indicated that the dissolved oxygen from aerobic recycled water would be consumed rapidly and aerobic conditions would only be present at the inlet of the columns. Therefore, for the NMOR that migrates more than 2 cm (based on a water flow of 190 mL d1) along the column, the potential mechanism for removal would be through anaerobic reductive degradation or co-metabolism.
3.2.
2553
Retardation coefficient
After NMOR and bromide were introduced into the columns, a retardation coefficient (R) was determined by comparing the migration rate of NMOR along the column to the migration rate of the conservative tracer bromide (Fig. 1). Data from the sterile column were used in preference to the non-sterile column data to eliminate the potential for biodegradation confounding the interpretation of the retardation data. Porosity studies based on bromide breakthrough data of the sterile and non-sterile columns showed that column packing of both columns were similar. R values (assuming linear sorption isotherms) for NMOR were determined by (i) initially fitting the bromide data to the convectionedispersion equation (Parker and Vangenuchten, 1984) with a nonlinear least squares fitting routine based on the LevenbergeMarquardt algorithm (MICROCAL, 1995) using Origin v7 software, then (ii) fitting the data for NMOR to the convectionedispersion equation constrained using the bromide fitted parameters, except the R value which was used as the fitting parameter. The estimated R value of 1.6 (for the low NMOR concentration experiment) was only marginally greater compared to the previously reported value of 1.2 in the same sediment, but at an NMOR concentration 3 orders of magnitude higher (Patterson et al., 2010). Sorption of non-ionized chemicals in soil is correlated to soil organic content of soil (Calvet, 1989; Nicholls, 1991). However, the sorption behaviour of NMOR was low even in sediment with relatively high organic carbon content such as the Leederville sediment. This was, possibly due to low hydrophobicity of NMOR as indicated by its low octanole water partition coefficient (log Kow ¼ 0.44; HSDB, 2003). While sorption studies at only two different concentrations are not adequate to confirm the linearity of sorption behaviour of NMOR, the similar R values for the low and high NMOR concentrations would suggest that the sorption isotherm of NMOR in the Leederville aquifer was near linear, and sorption was not substantially concentration dependent over the ng L1 to mg L1 concentration range. The marginally lower R value for the high NMOR concentration column experiment would be consistent with only minor competition for sorption sites (Xing et al., 1996). The sorption behaviour of NMOR was similar to NDMA, a more widely studied nitrosamine. Sorption of NDMA (log Kow of NDMA ¼ 0.57; HSDB, 2005a) was reported to be linear for several types of landscape soils (Yang et al., 2005) and subsurface soils (Gunnison et al., 2000) with r2 > 0.95 and r2 > 0.90 respectively. Moreover, NDMA was also reported to be weakly sorbed with distribution coefficient ranged from 0.45 to 1.14 L kg1 (Gunnison et al., 2000; Yang et al., 2005). Based on this NMOR sorption data, the flow of NMOR in groundwater through the Leederville aquifer would only be marginally retarded, with a flow velocity between 65 and 85% of the groundwater flow.
3.3. Degradation e low NMOR concentration column experiment For the low NMOR concentration column experiment, relative (compared to influent concentrations) NMOR concentration
2554
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 5 0 e2 5 6 0
Fig. 1 e NMOR (dBd) and Bromide (d-d) breakthrough profiles for A) high NMOR concentration after 15 days after NMOR delivery (modified after Patterson et al., 2010) and B) low NMOR concentration after 21 days after NMOR delivery.
data for the non-sterile and sterile columns are shown in Fig. 2. These data showed that NMOR degradation commenced within 3 months of its introduction into the nonsterile column. This suggests that initial numbers of NMORdegrading bacteria in the non-sterile column were low and increased with time. The combination of approximately 3
month lag-time (see Section 3.6) and then increased degradation rate caused a “cut-off plume” with the head of the NMOR plume not being degraded (likely as a result of insufficient numbers of NMOR-degrading bacteria) and migrating past the end of the column (Fig. 2B, w100 days). This “cut-off plume” effect was also observed in similar experiments that
Fig. 2 e NMOR concentrations as a fraction of influent concentration for A) sterile and B) non-sterile low NMOR concentration column experiments.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 5 0 e2 5 6 0
investigated iohexol degradation (Patterson et al., 2011). No substantial NMOR degradation was evident in the sterile control column over the 351 day experiment. To determine degradation rate of NMOR, the concentration data of NMOR from the last sampling event (which was chosen in preference to the earlier sampling events to provide sufficient time for degradation to commence) was plotted against the column residence time (Fig. 3). The distance of the sampling ports along the columns (cm) was converted to column residence time (days) based on the linear flow velocity of NMOR (linear velocity of bromide tracer divided by the R value of NMOR). As a first order degradation profile could be fitted to the concentration data with a greater degree of confidence than a zero order profile, a half-life curve was fitted to the experimental data (non-sterile column) to give an estimated degradation half-life of 40 3 days. Minor losses of NMOR were also observed in the control column. The losses observed in the sodium azide sterilized column may be due to analytical variability or slow degradation of the NMOR in the sodium azide sterilized column as some microorganisms were reported to be persistent under azide treatment (Lichstein and Soule, 1943). Assuming slow degradation, a half-life degradation rate of 260 50 days was estimated. However, the water residence time of 51 days was insufficient for substantial NMOR removal to provide more accurate half-life estimations.
3.4. Degradation e high NMOR concentration stop-flow column experiment While some minor loss of NMOR was observed in previous (Patterson et al., 2010) high concentration column experiments (Fig. 4), there was no statistical difference (P ¼ 0.26, student t test) between the sterile control column (half-
Fig. 3 e NMOR concentrations (, sterile column; - nonsterile column), bromide tracer concentrations (dBd sterile column; dCd non-sterile column) and fitted NMOR half-life curves (••• sterile column; d non-sterile column) for the low NMOR concentration column experiments at day 351. Column residence time was converted from the sampling port distance, based on the linear velocity of NMOR and bromide.
2555
Fig. 4 e NMOR concentrations (, sterile column; - nonsterile column) and NMOR half-life fitting (••• sterile column; d non-sterile column) for the high concentration column experiments after w12 months of operation (modified after Patterson et al., 2010).
life ¼ 160 80 days) and the non-sterile column (72 8 days). Therefore, an estimated half-life degradation rate >100 days compared to the control column data was reported (Patterson et al., 2010). For these experiments, the water residence time of 39 days within the column was insufficient to provide a more accurate estimation of NMOR degradation. To increase the accuracy of the NMOR degradation rate, the water residence time in the high NMOR concentration columns were increased by ceasing the water flow and trace organic delivery within the sterile and non-sterile columns. At this time, the experiment was continued as a stop-flow column experiment. Prior to ceasing water flow through the columns, column water samples were collected from sampling ports at 4, 25, 46, 92, 125, 142, 158, 175, and 192 cm from the base of each column. Based on the column flow rate, an average column residence time for water samples collected was 21 days. Column flow was then ceased, and column water samples collected again after approximately 6, 12 and 17 months (average column residence time of 158, 361 and 507 days). NMOR concentrations for column residence of 158, 361 and 507 days in the non-sterile column were observed to decrease over time at all sampling port locations along the column (Fig. 5). However, the rate of NMOR concentration decrease was slower at 4 cm from the base of the column than at other locations. The slower NMOR concentration decrease observed at the 4 cm location compared to the other port locations could be due to (i) diffusion of NMOR into the column from the 500 mL mixing chamber at the base of the column containing influent water concentrations of NMOR, (ii) different geochemical conditions at the base of the column as the result of aerobic water injection over a period of 12 months into the reductive pyritic sediment, prior to ceasing of influent water injection (iii) biodegradation competition of NMOR with other high adsorbing trace organics present in the recycled water that were retarded in the sediment at the base of the column. The
2556
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 5 0 e2 5 6 0
column) to give an estimated degradation half-life of 45 2 days. Fitting of control data gave an estimation half-life of 530 160 days.
3.5. Comparison of degradation rates at high and low concentration
Fig. 5 e NMOR concentrations along the sterile and nonsterile columns during the stop-flow column experiment. Results are for the sterile column on day 21 (,,B,,), day 158 (,,,,,), day 361 (,,6,,) and day 506 (,,>,,) and for the non-sterile column on day 21 (-C-), day 158 (---), day 361 (-:-) and day 506 (-A-).
presence of other chemicals can act as competitors and decrease biodegradation rates (Qiu et al., 2009; Stringfellow and Aitken, 1995). Due to potential confounding additional processes at port 4, data from this location was not used to estimate the average NMOR degradation half-life for the nonsterile and sterile stop-flow column experiments. The degradation half-life for the stop-flow column experiment was estimated using the average NMOR concentration (excluding the data for the 4 cm port) for the different stopflow column residence times (Fig. 6). A first order degradation profile could be fitted to the concentration data with a greater degree of confidence than a zero order profile, therefore a halflife curve was fitted to the experimental data (non-sterile
Fig. 6 e NMOR concentrations (, sterile column; - nonsterile column) and NMOR half-life fittings (••• sterile column; d non-sterile column) for the stop-flow column experiments.
Half-life degradation rates for all experiments are given in Table 3. The substantial difference of half-lives between sterile and non-sterile anaerobic columns indicates that the degradation process is likely to be an anaerobic biologically induced reductive NMOR degradation. Electron donation for this reductive degradation was possibly from the oxidation of SOM (0.32% w/w) or oxidation of reduced minerals such as pyrite (2% w/w) present in the Leederville sediment. Rapid oxygen consumption (8.5 mg L1 to <2 mg L1) and increased sulphate concentration (from <2 mg L1 to 40 mg L1) along the sterile and non-sterile columns of the low NMOR concentration experiment provided evidence of pyrite oxidation. Electron donation from DOC was unlikely due to the low DOC concentration of the influent water used for all experiments (<1 mg L1; Table 2). As, groundwater DOC from the site of sediment collection was only marginally higher with an average DOC of 1.2 mg L1 and an average biodegradable organic carbon of 0.1 mg L1, electron donation from SOM or oxidation of reduced minerals would be likely under field conditions. Degradation of NMOR in mammalian enzyme systems has been reported previously through several reaction pathways (Fig. 7) including a-hydroxylation (Hecht and Young, 1981; Jarman and Manson, 1986), a-hydroxylation (Manson et al., 1978) and denitrosation by reduction (Appel et al., 1980). Oxidation of NMOR (pathways A, B and C in Fig. 7) is unlikely in the Leederville aquifer due to the reductive anaerobic conditions of the aquifer. In experiments with several groups of bacteria isolated from rats’ gastrointestinal tract under anaerobic conditions, Rowland and Grasso (1975) found that the degradation mechanism of nitrosamines was different from what had been observed in other mammalian enzyme systems. In their studies, they observed that the nitrosamines were converted into their corresponding parent amine and nitrite. Appel et al. (1980) suggested NMOR was reduced through denitrosation into its parent amine (morpholine; MOR) and nitric oxide which then converted into nitrite. Involvement of cytochrome P-450 as a catalyst in the reduction was demonstrated in their experiments. However, the role of cytochrome P-450 as a reductase in the denitrosation of NMOR was arguable since it has also been observed as an oxidase in a-hydroxylation of nitrosamines (Lorr et al., 1982). Keefer and Lunn (1985) concluded that NMOR could be reduced into its parent amine (MOR) through the breakdown of the NeN bond. Analysis of MOR in column water from the high NMOR concentration stop-flow column experiments at day 506 showed MOR concentrations up to 0.14 mg L1 in the nonsterile column, while concentrations were below the detection limit (<0.03 mg L1) in the sterile column. Based on these data, degradation of NMOR in the Leederville aquifer is likely to occur with the formation of MOR as a degradation product (pathway D in Fig. 7).
2557
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 5 0 e2 5 6 0
Table 3 e Comparison of degradation rates for all experiments. Experiment
Column Stop-flow column Columna
NMOR initial concentration (ng L1)
200 650,000 650,000
Water residence time (days)
51 w506 39
NMOR degradation half-life (days) Non-sterile
Sterile
Relative to control
40 3 45 2 72 8
260 50 530 160 160 80
>100a
a Patterson et al., 2010.
Bae et al. (2002) reported that MOR was persistent under anaerobic degradation. However, Dmitrenko and Gvozdyak (as cited in IPCS, 1996) reported that MOR degraded anaerobically by MOR-degrading mycobacteria. Under our experimental conditions, MOR concentrations in the non-sterile column were less than predicted assuming complete stoichiometric conversion of NMOR to MOR and no biodegradation of MOR. To our knowledge, there is no specific report on MOR sorption studies in soil but estimation of its partition coefficient based on its low octanol water coefficient (log Kow ¼ 0.86; HSDB, 2005b) suggests that MOR is highly mobile in soil and hydrophobic sorption to organic matter would be low (HSDB, 2005b). Minor hydrophilic sorption (sorption to mineral surfaces) of MOR may have been possible given its pKa ¼ 8.5 (HSDB, 2005b) and the column pH of 8.2. Lower MOR concentration than predicted might also suggest that MOR has undergone slow reductive degradation resulting in only transient accumulation of MOR.
Alternatively, other minor NMOR degradation pathways may also explain the less than predicted MOR concentrations. At environmental levels of exposure, MOR does not present a toxic risk to humans or pose a substantial risk to biota in the environment (IPCS, 1996). While biodegradation of some chemicals is limited to either high or low concentrations, these experiments showed that NMOR was able to be degraded biologically at relatively high and low concentrations. A comparable half-life degradation rate for the high and low NMOR concentration in the non-sterile experiments suggests neither an inhibitory effect of NMOR (or the degradation products) at 650 mg L1 nor a threshold effect at 200 ng L1 was apparent in these experiments. Inhibition of degradation at concentration higher than the toxic level or concentration lower than the threshold level has been observed for several compounds. Chemicals such as 2,4-dichlorophenoxyacetate (Rubin et al., 1982) and pentachlorophenol (Crawford and Mohn as cited in Providenti
Fig. 7 e Metabolism of NMOR (modified after Hecht and Young, 1981 and Jarman and Manson, 1986).
2558
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 5 0 e2 5 6 0
et al., 1993) showed persistency at certain concentrations which appeared to be higher than their toxic levels since they were degraded at lower concentrations. At these concentrations, the compounds acted as toxins to the microorganisms responsible for their degradation and inhibit biodegradation (Providenti et al., 1993). In contrast, degradation of other compounds such halobenzoate (Boethling and Alexander, 1979a; Rubin et al., 1982), nitrotriacetic acid, 2,4-dichlorophenol and isopropyl N-phenylcarbamate (Hoover et al., 1986) were inhibited at low concentrations. In these cases, a threshold concentration has to be met to support growth of corresponding microorganisms in order for biodegradation to occur (Alexander, 1985; Providenti et al., 1993). Currently, there are no reports of concentration effects on biodegradation of NMOR. However, there are reports of concentration effects for another nitrosamine, NDMA. NDMA was reported to degrade over a range of concentrations. Fournier et al. (2009) reported that NDMA degraded from mg L1 concentrations to <2 ng L1. Sharp et al. (2010) also reported NDMA degradation over the mg L1 to ng L1 concentration range. A toxic effect for natural microorganisms in lake water was not observed at NDMA concentrations as high as 1000 mg L1 (Kaplan and Kaplan, 1985). However, an increase in initial concentration of NDMA was observed to decrease the degradation rate (a longer half-life at a higher concentration) and this effect was significant for orders of magnitude differences in concentration (Gunnison et al., 2000; Kaplan and Kaplan, 1985; Yang et al., 2005). Moreover, biodegradation studies for secondary amines showed that the biodegradation rate was strongly influenced by initial concentration (Boethling and Alexander, 1979b). In this study, there was a three order of magnitude difference between the two tested concentrations (Table 3; low concentration column experiment and high concentration stop-flow column experiment) but the degradation half-life was only marginally different (40 and 45 days). These results indicate that the degradation of NMOR was not substantially affected by the initial concentration as has been reported for NDMA and secondary amines. Half-lives of NMOR at high and low concentrations indicate that in the Leederville aquifer, NMOR will degrade under anaerobic conditions if there is sufficient residence time for the biodegradation to occur. Based on an aquifer residence time of 12 months, an NMOR concentration decrease of >99.5% would be predicted. Therefore, although NMOR is highly mobile (travelling at up to 85% of the rate of groundwater), if it is introduced into the aquifer, biodegradation during aquifer passage should prevent the establishment of an NMOR contaminant groundwater plume. In addition, low sorption behaviour of NMOR may benefit biodegradation since low sorption increases the availability of the electron donor to microorganisms thus increasing the biodegradation rate (Calvet, 1989; Providenti et al., 1993).
3.6.
Degradation lag-times
For all experiments, there was no rapid increase in degradation rates during the experiment, thus a biological lag-time for NMOR biodegrading bacteria was difficult to determine. This was likely due to the relatively slow degradation rate of NMOR
observed. However, data from the low concentration experiment indicated that degradation commenced within 3 months of NMOR introduction into the non-sterile column.
4.
Conclusion
These findings suggest: 1. Degradation of NMOR in Leederville aquifer sediment is likely an anaerobic biologically induced reductive degradation process which follows first order kinetics with a biological lag-time of less than 3 months. 2. With consistent half-life degradation of 40e45 days for concentrations up to 650 mg L1, there was no inhibitory effect observed on microorganisms responsible to the biodegradation of NMOR at 650 mg L1, nor was there a threshold effect at 200 ng L1 observed during the experiments. 3. The low retardation coefficient of NMOR indicates that NMOR would only be marginally retarded in the Leederville aquifer, with a flow velocity up to 85% of the groundwater flow. 4. Any NMOR present in recycled water recharged to the aquifer should be naturally degraded during aquifer passage with sufficient aquifer residence time or travel distance between recycled water injection and groundwater extraction.
Acknowledgments This research was made possible through funding from CSIRO Water for a Healthy Country Flagship Program, and the Water Corporation of Western Australia.
references
Alexander, M., 1985. Biodegradation of organic chemicals. Environ. Sci. Technol. 19 (2), 106e111. Appel, K., Schrenk, D., Schwarz, M., Mahr, B., Kunz, W., 1980. Denitrosation of N-nitrosomorpholine by liver microsomes; possible role of cytochrome P-450. Cancer Lett. 9 (1), 13e20. Bae, H.-S., Cho, Y.-G., Oh, S.-E., Kim, I.-S., Lee, J.M., Lee, S.-T., 2002. Anaerobic degradation of pyrrolidine and piperidine coupled with nitrate reduction. Chemosphere 48 (3), 329e334. Boethling, R., Alexander, M., 1979a. Effect of concentration of organic chemicals on their biodegradation by natural microbial communities. Appl. Environ. Microbiol. 37 (6), 1211e1216. Boethling, R., Alexander, M., 1979b. Microbial degradation of organic compounds at trace levels. Environ. Sci. Technol. 13 (8), 989e991. Calvet, R., 1989. Adsorption of organic chemicals in soils. Environ. Health. Perspect. 83, 145e177. de Vocht, F., Burstyn, I., Straif, K., Vermeulen, R., Jakobsson, K., Nichols, L., Peplonska, B., Taeger, D., Kromhout, H., 2007. Occupational exposure to NDMA and NMOR in the European rubber industry. J. Environ. Monit. 9 (3), 253e259.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 5 0 e2 5 6 0
Fajen, J.M., Carson, G.A., Rounbehler, D.P., Fan, T.Y., Vita, R., Goff, U.E., Wolf, M.H., Edwards, G.S., Fine, D.H., Reinhold, V., Biemann, K., 1979. N-nitrosamines in the rubber and tire industry. Science 205 (4412), 1262e1264. Fournier, D., Hawari, J., Halasz, A., Streger, S.H., McClay, K.R., Masuda, H., Hatzinger, P.B., 2009. Aerobic biodegradation of N-nitrosodimethylamine by the propanotroph Rhodococcus ruber ENV425. Appl. Environ. Microbiol. 75 (15), 5088e5093. Gunnison, D., Zappi, M.E., Teeter, C., Pennington, J.C., Bajpai, R., 2000. Attenuation mechanisms of N-nitrosodimethylamine at an operating intercept and treat groundwater remediation system. J. Hazard. Mater. 73 (2), 179e197. Hecht, S., Young, R., 1981. Metabolic a-hydroxylation of N-nitrosomorpholine and 3,3,5,5-tetradeutero-Nnitrosomorpholine in the F344 rat. Cancer Res. 41 (12), 5039e5043. Hollender, J., Zimmermann, S., Koepke, S., Krauss, M., McArdell, C., Ort, C., 2009. Elimination of organic micropollutants in a municipal wastewater treatment plant upgraded with a full-scale post-ozonation followed by sand filtration. Environ. Sci. Technol. 43 (20), 7862e7869. Hoover, D., Borgonovi, G., Jones, S., Alexander, M., 1986. Anomalies in mineralization of low concentrations of organic compounds in lake water and sewage. Appl. Environ. Microbiol. 51 (2), 226e232. HSDB, 2003. Hazardous Substance Data Base, Nnitrosomorpholine. U.S. National Library of Medicine. HSDB, 2005a. Hazardous Substance Data Base, Nnitrosodimethylamine. U.S. National Library of Medicine. HSDB, 2005b. Hazaerdous Substance Data Base, Morpholine. U.S. National Library of Medicine. IARC, 1987. IARC Monographs on the Evaluation of the Carcinogenic Risks to Humans. World Health Organization, Lyon. IPCS, 1996. International Program on Chemical Safety, Environmental Health Criteria 179, Morpholine. World Health Organization, Geneva. Jarman, M., Manson, D., 1986. The metabolism of Nnitrosomorpholine by rat liver microsomes and its oxidation by the Fenton system. Carcinogenesis 7 (4), 559e565. Kaplan, D.L., Kaplan, A.M., 1985. Biodegradation of NNitrosodimethylamine in aqueous and soil systems. Appl. Environ. Microbiol. 50 (4), 1077e1086. Keefer, L.K., Lunn, G., 1985. Reductive Destruction of Nitrosamines, Hydrazines, Nitramines, Azo- and Azoxycompounds United States of America. Krasner, S.W., Westerhoff, P., Chen, B., Rittmann, B.E., Amy, G., 2009. Occurrence of disinfection byproducts in United States wastewater treatment plant effluents. Environ. Sci. Technol. 43 (21), 8320e8325. Krauss, M., Hollender, J., 2008. Analysis of nitrosamines in wastewater: exploring the trace level quantification capabilities of a hybrid linear ion trap/orbitrap mass spectrometer. Anal. Chem. 80 (3), 834e842. Krauss, M., Longree, P., Dorusch, F., Ort, C., Hollender, J., 2009. Occurrence and removal of N-nitrosamines in wastewater treatment plants. Water Res. 43 (17), 4381e4391. Lichstein, H., Soule, M., 1943. Studies of the effect of sodium azide on microbic growth and respiration: I. The action of sodium azide on microbic growth. J. Bacteriol. 44 (3), 221e230. Lorr, N., Tu, Y., Yang, C., 1982. The nature of nitrosamine denitrosation by rat liver microsomes. Carcinogenesis 3 (9), 1039e1043. Manson, D., Cox, P.J., Jarman, M., 1978. Metabolism of N-nitrosomorpholine by the rat in vivo and by rat liver microsomes and its oxidation by the Fenton system. Chem. Biol. Interact. 20 (3), 341e354.
2559
MICROCAL, 1995. ORIGIN User’s Manual. Microcal Software Inc., Northampton. Mitch, W.A., Sharp, J.O., Trussell, R.R., Valentine, R.L., AlvarezCohen, L., Sedlak, D.L., 2003. N-nitrosodimethylamine (NDMA) as a drinking water contaminant: a review. Environ. Eng. Sci. 20 (5), 389e404. Nicholls, P.H., 1991. In: Jones, K.J. (Ed.), Organic Contaminants in the Environment. Elsevier Science Publisher Ltd., London. Parker, J.C., Vangenuchten, M.T., 1984. Flux-averaged and volume-averaged concentrations in continuum approaches to solute transport. Water Resour. Res. 20 (7), 866e872. Patterson, B.M., Shackleton, M., Furness, A.J., Pearce, J., Descourvieres, C., Linge, K.L., Busetti, F., Spadek, T., 2010. Fate of nine recycled water trace organic contaminants and metal(loid)s during managed aquifer recharge into a anaerobic aquifer: column studies. Water Res. 44 (5), 1471e1481. Patterson, B.M., Shackleton, M., Furness, A.J., Bekele, E.B., Pearce, J., Linge, K.L., Busetti, F., Spadek, T., Toze, S., 2011. Fate of nine recycled water trace organics during managed aquifer recharge in an aerobic aquifer. J. Contam. Hydrol 122, 53e62. Planas, C., Palacios, O., Ventura, F., Rivera, J., Caixach, J., 2008. Analysis of nitrosamines in water by automated SPE and isotope dilution GC/HRMS occurrence in the different steps of a drinking water treatment plant, and in chlorinated samples from a reservoir and a sewage treatment plant effluent. Talanta 76 (4), 906e913. Playford, P.E., Cockbain, A.E., Low, G.H., 1976. Geology of the Perth basin, Western Australia. Geol. Surv. Western Aust. Bull. 124, 169. Providenti, M., Lee, H., Trevors, J., 1993. Selected factors limiting the microbial degradation of recalcitrant compounds. J. Ind. Microbiol. 12 (6), 379e395. Qiu, Y., Pang, H., Zhou, Z., Zhang, P., Feng, Y., Sheng, G.D., 2009. Competitive biodegradation of dichlobenil and atrazine coexisting in soil amended with a char and citrate. Environ. Pollut. 157 (11), 2964e2969. Ranwala, N., 2009. Nanogram per litre detection of Nnitrosodimethylamine and N-nitrosomorpholine in reverse osmosis treated wastewater using low volume samples and GC-MS(EI). University of Western Australia, Perth. Rowland, I., Grasso, P., 1975. Degradation of N-nitrosamines by intertinal bacteria. Appl. Microbiol. 29 (1), 7e12. Rubin, H., Subbarao, R., Alexander, M., 1982. Rates of mineralization of trace concentrations of aromatic compounds in lake water and sewage samples. Appl. Environ. Microbiol. 43 (5), 1133e1138. Schreiber, I.M., Mitch, W.A., 2006a. Nitrosamine formation pathway revisited: the importance of chloramine speciation and dissolved oxygen. Environ. Sci. Technol. 40 (19), 6007e6014. Schreiber, I.M., Mitch, W.A., 2006b. Occurrence and fate of nitrosamines and nitrosamine precursors in wastewaterimpacted surface waters using boron as a conservative tracer. Environ. Sci. Technol. 40 (10), 3203e3210. Sharp, J.O., Sales, C.M., Alvarez-Cohen, L., 2010. Functional characterization of propane-enhanced Nnitrosodimethylamine degradation by two actinomycetales. Biotechnol. Bioeng. 107 (6), 924e932. Spiegelhalder, B., Preussmann, R., 1983. Occupational nitrosamine exposure .1. Rubber and tyre industry. Carcinogenesis 4 (9), 1147e1152. Steinle-Darling, E., Zedda, M., Plumlee, M.H., Ridgway, H.F., Reinhard, M., 2007. Evaluating the impacts of membrane type, coating, fouling, chemical properties and water chemistry on reverse osmosis rejection of seven nitrosoalklyamines, including NDMA. Water Res. 41 (17), 3959e3967.
2560
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 5 0 e2 5 6 0
Stringfellow, W., Aitken, M., 1995. Competitive metabolism of naphthalene, methylnaphthalenes, and fluorene by phenanthrene-degrading pseudomonads. Appl. Environ. Microbiol. 61 (1), 357e362. Xing, B., Pignatello, J., Gigliotti, B., 1996. Competitive sorption between atrazine and other organic compounds in soils and model sorbents. Environ. Sci. Technol. 30 (8), 2432e2440.
Yang, W.C., Gan, J., Liu, W.P., Green, R., 2005. Degradation of N-nitrosodimethylamine (NDMA) in landscape soils. J. Environ. Qual. 34 (1), 336e341. Zhao, Y.-Y., Boyd, J.M., Woodbeck, M., Andrews, R.C., Qin, F., Hrudey, S.E., Li, X.-F., 2008. Formation of N-nitrosamines from eleven disinfection treatments of seven different surface waters. Environ. Sci. Technol. 42 (13), 4857e4862.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 6 1 e2 5 7 2
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Escherichia coli, enterococci, and Bacteroides thetaiotaomicron qPCR signals through wastewater and septage treatment Sangeetha Srinivasan a, Asli Aslan a, Irene Xagoraraki b, Evangelyn Alocilja c, Joan B. Rose a,* a
Department of Fisheries and Wildlife, 13, Natural Resources, Michigan State University, East Lansing, MI 48824, USA Department of Civil and Environmental Engineering, Michigan State University, East Lansing, MI 48824, USA c Department of Biosystems and Agricultural Engineering, Michigan State University, East Lansing, MI 48824, USA b
article info
abstract
Article history:
Fecal indicators such as Escherichia coli and enterococci are used as regulatory tools to
Received 19 October 2010
monitor water with 24 h cultivation techniques for possible input of sewage or feces and
Received in revised form
presence of potential enteric pathogens yet their source (human or animal) cannot be
7 February 2011
determined with routine methods. This critical uncertainty has furthered water pollution
Accepted 8 February 2011
science toward new molecular approaches. Members of Bacteroides genus, such as Bacter-
Available online 17 February 2011
oides thetaiotaomicron are found to have features that allow their use as alternative fecal indicators and for Microbial Source Tracking (MST). The overall aim of this study was to
Keywords:
evaluate the concentration and fate of B. thetaiotaomicron, throughout a wastewater treat-
Bacteroides thetaiotaomicron
ment facility and septage treatment facility. A large number of samples were collected and
Escherichia coli
tested for E. coli and enterococci by both cultivation and qPCR assays. B. thetaiotaomicron
Enterococci
qPCR equivalent cells (mean: 1.8 107/100 mL) were present in significantly higher
qPCR
concentrations than E. coli or enterococci in raw sewage and at the same levels in raw
Sewage
septage. The removal of B. thetaiotaomicron target qPCR signals was similar to E. coli and
Septage
enterococci DNA during the treatment of these wastes and ranged from 3 to 5 log10 for wastewater and was 7 log10 for the septage. A significant correlation was found between B. thetaiotaomicron marker and each of the conventional indicators throughout the waste treatment process for both raw sewage and septage. A greater variability was found with enterococci when compared to E. coli, and CFU and equivalent cells could be contrasted by various treatment processes to examine removal and inactivation via septage and wastewater treatment. These results are compared and contrasted with other qPCR studies and other targets in wastewater samples providing a view of DNA targets in such environments. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
Wastewater is the source of many human enteric pathogens (Nayak and Rose, 2007; Lee et al., 2006; Kamel et al., 2010; Robertson et al., 2006) and often associated with swimmingacquired illnesses in natural waters (Wade et al., 2006). Adequate wastewater treatment prior to effluent discharge
plays a critical role in minimizing public health risks. On-site wastewater disposal using septic tanks has also been an issue regarding pathogen entry into and transmission through water particularly groundwater (Fong et al., 2007) and septage treatment and application on land has received little attention in regard to microbial quality. In most states, fecal coliform bacteria are still used to address wastewater treatment using
* Corresponding author. Tel.: þ1 517 432 4412; fax: þ1 517 432 1699. E-mail address:
[email protected] (J.B. Rose). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.02.010
2562
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 6 1 e2 5 7 2
the National Pollutant Discharge Elimination System permit programs and are disconnected from ambient water quality monitoring, in which coastal states are moving toward Escherichia coli (E. coli) and enterococci. E. coli and enterococci have long been used as indicators of fecal pollution for recreational and drinking waters (USEPA, 2002, 2005a); and cultivation methods are used as the gold standard for the enumeration of these bacteria in water (Messer and Dufour, 1998). Recent advances in molecular biology such as polymerase chain reaction (PCR) and particularly quantitative PCR (qPCR) have revolutionized microbiology. Quantitative PCR has many advantages over standard cultivation methods, such as producing results rapidly (30 mine2 h), the ability to detect viable but non-cultivatable (VNBC) pathogens, and providing quantitative results with a wide detection range (100e108 copies/reaction). However, it is still recognized that both inactivated and live microbes will be detected, which is a disadvantage when evaluating disinfection processes. Never-the-less, evaluation and application of these qPCR methods for routine monitoring of fecal contamination in recreational waters is ongoing in the US (Haugland et al., 2005; Noble et al., 2006; Wade et al., 2006; Lavender and Kinzelman, 2009) as well as in wastewaters (Silkie and Nelson, 2009; Varma et al., 2009; Wery et al., 2008; Frahm and Obst, 2003). One of the other disadvantages of using routine bacterial indicators is that the source cannot be identified while the specificity of molecular methods has lead to development of a field known as Microbial source tracking (MST) that has enabled the identification of animal or human sources of fecal contamination (Scott et al., 2002; Simpson et al., 2002; USEPA, 2005b). Many human fecal specific assays to address sewage discharges impacting water quality have targeted species in the BacteroidesePrevotella group directed toward detection of 16S rRNA genes (Seurinck et al., 2005; Layton et al., 2006; Okabe et al., 2007). Bacteroides spp. are obligately anerobic, Gram negative, rod shaped, and non-endospore forming bacteria and are normally commensals that constitute the most numerous members of the intestinal flora of all warm blooded animals (Wexler, 2007). There are still concerns regarding cross reactivity of some of these genetic markers with feces from humans and other animals including cats and dogs (Sadowsky et al., 2007; Kildare et al., 2007). Some qPCR assays based on 16S rRNA genes have been reported to cross react with fish DNA (McLain et al., 2009). Moreover, the exact copy number of these 16S rRNA genes present in one cell is not known which makes conversion of qPCR copy number to cell equivalents rather difficult. Recently, Yampara-Iquise et al. (2008) examined a single copy putative mannanase 1-6 gene of Bacteroides thetaiotaomicron as a human fecal source tracking marker with good specificity. The overall aim of this study was to evaluate the concentration and fate of B. thetaiotaomicron, throughout a wastewater treatment facility and septage treatment facility in contrast to E. coli and enterococci as measure by qPCR and cultivation. In this study, a qPCR assay targeting uidA gene for E. coli was developed and used, modified from Frahm and Obst (2003). Enterococci qPCR assay focused on the use of primers and probes designed for detection of the 23S rRNA gene sequences (Haugland et al., 2005; Silkie and Nelson, 2009).
Samples were collected throughout the wastewater treatment processes as well as from septage before and after treatment. Cultivable counts of E. coli and enterococci were compared to the qPCR data generated and an automated DNA extractor was compared to a commercially available QIAmp DNA mini kit (Qiagen, Valencia, CA, USA).
2.
Materials and methods
2.1.
Samples
a) Wastewater treatment plant samples: Over 200 samples were collected from a municipal wastewater treatment plant, located in East Lansing, Michigan that serves a population of 90 000. The plant receives, on an average basis, a little less than 13.40 MGD (million gallons per day) wastewater inflow. Samples collected from this facility included: i) Raw sewage (RS) ii) Primary effluent (PE), after the solids have settled iii) Secondary effluent (SE pre-chlorination), after activated sludge process and secondary clarification iv) Secondary effluent (SE post-chlorination), after disinfection by chlorination, and v) Tertiary effluent (TE), effluent from secondary treatment post sodium bi-sulfite dechlorination and filtration through rapid sand filters. For comparison of auto and manual DNA extraction, raw sewage (n ¼ 9), and primary effluent samples (n ¼ 9), secondary post-chlorinated effluent (n ¼ 9), and tertiary effluents (n ¼ 9) were used, which is a total of 36 samples from the wastewater environment. For assessment of conventional indicators and B. thetaiotaomicron (a-mannanase gene) RS, PE, SE and TE samples were collected in triplicates during 18 sampling events from January 2009 to January 2010. During six sampling events within this time frame, secondary treated effluent prior to the chlorination step was also collected in triplicates. During each sampling event, one hundred milliliters of RS and PE, 500 mL of pre-chlorinated SE and 2 L of SE and TE were collected in triplicates. Chlorinated SE and TE samples were collected in bottles pre-loaded with sodium thiosulphate (1 mL of 10% solution) to neutralize any residual chlorine present in the effluents. All samples were transported on ice and processed within 2 h after collection. b) Septage treatment plant samples: Samples were collected from a septage treatment plant located in Charveloix, Michigan. This treatment plant utilizes an aerobic biological treatment system to treat septage wastes (solid waste from septic tanks) and discharges the treated effluent to a municipal sewer system. Samples were collected during eight sampling events between January 2009 and November 2009. During each event, triplicates of 50 mL raw septage and 500 mL of septage effluent were collected, placed on ice and shipped to Water Quality and Health Laboratory at Michigan State University, East Lansing, MI.
2563
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 6 1 e2 5 7 2
2.2.
Comparison of DNA extraction methods
Comparison between automated DNA extraction using Roche MagNaPure LC instrument (Roche Applied Sciences, Indianapolis, Ind.) and manual extraction using QIAmp DNA mini kit (Qiagen, Valencia, CA, USA) was performed. Fifty milliliters RS and PE samples were centrifuged at 8000 g for 20 min and for treated sewage, 1 L of SE and TE was filtered through 90 mm, 0.45 mm pore size nitrocellulose membrane filters (Millipore, Billerica, Mass.). The filters were folded and immersed into sterile phosphate buffered saline (PBS) in 50 mL centrifuge tubes. The tubes were then vortexed at high speed to detach the cells from membrane for 2 min. The filters were removed and the tubes were centrifuged at a speed of 8000 g for 15 min. Around 48 mL of the supernatant was discarded and the remaining sample was mixed well by vortexing. The volume was recorded and from this, 400 mL of the pellet was used for manual DNA extraction with the QIAmp DNA mini kit (Valencia, CA, USA), another 400 mL of the aliquot was used for extraction by the Roche MagNaPure LC instrument and remainder was stored at 80 C. These volumes were included in the calculation for concentrations of targets by qPCR. For DNA extraction using Roche MagNaPure automated instrument, an additional lysis step was performed by mixing 400 mL of concentrated samples with 180 mL of MagNaPure lysis buffer and 20 mL of Proteinase K (20 mg/ml), and incubated at 65 C for 30 min. The mixture was then centrifuged at 500 g for 30 s to settle down the particles. The supernatant was used for DNA extraction by the instrument. Simultaneously, a manual DNA extraction was carried out on the same
samples using the QIAmp DNA mini kit. Both extraction methods resulted in 200 mL of DNA suspended in TE buffer. Negative controls (molecular grade water) were used to check for cross contamination in both extraction methods. The concentrations of extracted DNA were determined by using Nanodrop ND-1000 Spectrophotometer. Extracted DNA was stored at 20 C until further analyses.
2.3.
Reproducibility of the auto extraction method
In order to assess the intervariability in extraction during comparison of auto and manual extraction methods, eight replicates of concentrated RS samples were subjected to DNA extraction by both instrument and manual methods following the procedures described above. For all replicates, the same volume of lysis buffer (180 mL) and proteinase K (20 mL) were added to the tubes and subjected to auto extraction. DNA concentrations in all extracts were analyzed by qPCR methods for both E. coli and enterococci. The extractions were also carried by two analysts to assess the variability in both instrument and manual methods.
2.4. DNA extraction used for assessment of septage samples DNA extraction from treated and untreated sewage samples were performed using automated method described previously. Untreated Septage (5 mL) samples were mixed thoroughly by vortexing for 1 min and 600 mL was taken for DNA extraction. For septage effluent samples, 50 mL of the sample
Table 1 e The bacteria, gene targets, primer/probe sequences, and PCR programs used for the assays. Primer/probe sequence (50 e30 )
Amplicon size (bp)
PCR programs
163
6 s at 95 C 8 s at 58 C 8 s at 72 C
This study
AGAAATTCCAAACGAACTTG CAGTGCTCTACCTCCATCATT 6FAM-TGGTTCTCTCCGAAATAGCTT TAGGGCTA-TAMRA
91
15 s at 95 C 30 s at 60 C 15 s at 72 C
Frahm and Obst (2003)
qPCR
CATCGTTCGTCAGCAGTAACA CCAAGAAAAAGGGACAGTGG 6FAM-ACCTGCTG-NFQ
63
15 s at 94 C 60 s at 60 C 5 s for 72 C
Yampara-Iquise et al. (2008)
E. coli/uidAc
Regular PCRd
GCAGTCTTACTTCCATGATTTCTTTA TAATGCGAGGTACGGTAGG
522
30 s at 95 C 30 s at 57 C 60 s at 72 C
This study
Enterococci/23S rRNAc
Regular PCRd
ATCTACCCATGTCCAGGTTGAAG CCATCTCGGGTTACCGAATTCAG
223
30 s at 95 C 30 s at 57 C 60 s at 72 C
This study
B. thetaiotaomicron/a-1-6 mannanasec
Regular PCRd
GCGGTACACATAACGGG ATCGACTTATATCTACTGGCAAC
306
30 s at 95 C 30 s at 60 C 60 s at 72 C
This study
Bacteria/gene
Type of assay
E. coli/uidAa
qPCR
CAATGGTGATGTCAGCGTT ACACTCTGTCCGGCTTTTG 6FAM-TTGCAACTGGACAAGGCA CCAGC-BBQ
Enterococci/23S rRNAa
qPCR
B. thetaiotaomicron/ a-1-6 mannanaseb
a b c d
References
PCR programs were repeated for 40 cycles, after an initial cycle of 10 min at 95 C (For E. coli and enterococci). PCR programs were repeated for 45 cycles, after an initial cycle of 15 min at 95 C (For B. thetaiotaomicron). PCR programs were repeated for 35 cycles, after an initial cycle of 10 min at 95 C and terminated by a final extension cycle at 72 C for 8 min. Regular PCR was used for producing the amplicon required for the preparation of standards.
2564
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 6 1 e2 5 7 2
was centrifuged at 8000g for 20 min. The supernatant was discarded and 1 mL of the pellet was left behind. From this, 600 mL was used for DNA extraction. For both treated and untreated sewage and septage samples, DNA extraction was carried out using Roche MagNaPure automated machine following the protocol described above.
2.5.
qPCR analyses
In order to prepare the standards, the DNA was extracted from the bacterial strains E. coli ATCC 15597, and Enterococcus faecalis ATCC 19433, and for B. thetataiomicron, genomic DNA from ATCC (number 29148D-5) was used. The uidA gene of E. coli, 23S rRNA gene of enterococcus and a-mannanase gene of B. thetataiomicron were amplified separately using primer sets that flanked the qPCR target amplicon sequences (Table 1). These primers were developed using the Roche LightCycler Primer Design Software. Polymerase chain reaction (PCR) was performed in a 25 mL total reaction mix which contained 15 mL Hotstart DNA Polymerase Mastermix, 0.4 mM of each primer, 2 mL of the template DNA and molecular graded water (QIAgen, Valencia, CA, USA) to make up a final volume of 25 mL. The amplified PCR products for all three target genes were cloned into TOPO PCR 2.1 and transformed with the TOPO10 F0 competent cells (Invitrogen Inc., Carlsbad, CA, USA), according to the protocol provided by the manufacturer. Plasmids were extracted with QIAprep Spin MiniPrep kit (Valencia, CA, USA) and were sequenced at the Research Technology Support Facility (RTSF) at Michigan State University that confirmed the insertion of the target inside the vector. The plasmids were quantified using Nano-Drop spectrophotometer and then serially diluted ten-fold to construct qPCR standard curves. Triplicates of dilutions ranging from 106 to 100 were used for the standard curve. Targeted bacterial genes, primers/probes and corresponding qPCR programs are described in Table 1. For the E. coli uidA gene, satisfactory results with no homology with other targets were noted after checking the primer/probe sequences by BLAST analysis. All isolates of E. coli were detected whereas non-E. coli bacterial strains and other coliform bacteria were not detected by this assay. The B. thetataiomicron a-mannanase assay was tested, in our laboratory, against 226 non-human fecal samples and exhibited excellent specificity (bird, cow, cat, dog, horse and pig feces) (unpublished). DNA extracted from all samples was analyzed by qPCR using Roche LightCycler 2.0 Instrument (Roche Applied Sciences, Indianapolis, IN). The reaction mixture for amplification of E. coli uidA gene consisted of 2 mL of Roche Fast Start LightCycler Mastermix, forward and reverse primers (0.5 mM each), probe (0.2 mM), 3.2 mM MgCl2 and nuclease free water to a final volume of 15 mL. Cycle threshold (Ct) was measured during each amplification and target gene concentration was analyzed automatically by absolute quantification method by the LightCycler Software 4.0. The enterococci 23S rRNA and B. thetaiotaomicron a-mannanase 1-6 (B. theta a) qPCR assays were carried out using 10 mL of LightCycler 480 Probes Mastermix (Roche, Indianapolis, IN), forward and reverse primers (0.5 mM each for enterococci and 0.2 mM each for B. theta a), probe (0.2 mM for enterococci and 0.1 mM for B. theta a {probe number 62 from Roche UPL}), and nuclease free water to make up
a final volume of 15 mL. Five microliters of the extracted DNA sample was used as the template and run in duplicates. All sample extracts were diluted (1:5 dilution) and checked for any qPCR inhibition based on the difference in corresponding threshold cycle values. Every four to five runs, a standard curve was run (using various dilutions in triplicates) and the average efficiency of qPCR assays for the uidA gene of E. coli, 23S rRNA of enterococci gene and a-mannanase 1-6 gene of B. thetaiotaomicron were 102 0.5%, 98 0.3% and 96 1.2%, respectively and r2 value was always higher than 0.99 for all three assays. A diluted plasmid standard was included in triplicates as a positive control during each qPCR run and the average threshold cycle was compared with the original standard curve. The copies of uidA gene of E. coli, 23S rRNA of enterococci gene and a-mannanase 1-6 gene of B. thetaiotaomicron present in the sample were quantified from the standard curves obtained earlier. The copies of the corresponding genes were converted to cell equivalents; in the case of E. coli and B. thetaiotaomicron, only one copy of the target gene is present in a cell; thus, one copy number corresponds to one cell. However, in case of enterococci, it has been suggested that there are four copies of 23S rRNA present in a cell; therefore, DNA copies to cell conversions were completed based on four copies of enterococci qPCR targets corresponded to one cell in this study (Viau and Peccia, 2009). All final concentrations for qPCR analyses were reported as qPCR equivalent cells/100 mL.
2.6.
Cultivation methods
U.S.EPA methods 1603 and 1600 were used for enumerating E. coli and enterococci, respectively by cultivation (USEPA, 2002, 2005a). Serial dilutions of the raw sewage and primary effluents (101 through 105) were made and one mL from these dilutions was filtered through 47 mm diameter, 0.45 mm pore size, membrane filters. For pre-chlorinated secondary effluents, secondary and tertiary treated effluents, one mL and 100 mL, respectively, were filtered. One mL of raw septage samples was serially diluted and these dilutions were used for further bacterial indicator analysis. For septage effluent samples, volumes of 0.1 mL, 1 mL and 10 mL were filtered. The filters were placed on mTEC agar and mEI agar plates. The mTEC agar plates were incubated 35 0.5 C for 2 h, followed by incubation in a water bath at 44.5 0.2 C for 22 h. The mEI plates were incubated for 24 h at 41 C. The concentrations of E. coli and enterococci were reported as colony forming units (CFU) per 100 mL.
2.7.
Statistical analysis
All statistical analyses were performed using SAS software 9.2 (SAS Inc, 2002) and significance level was set at a ¼ 0.05. The data were log10-transformed to achieve normal distribution and meet the assumptions of a parametric test. Simple t-tests were used to compare the means of concentrations of the qPCR and the cultivation method results ( p < 0.05). The coefficient of variation (CV%) was calculated to evaluate the intervariability in extraction procedures using the formula; CV% ¼ ðstandard deviation=meanÞ 100
2565
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 6 1 e2 5 7 2
Table 2 e t-Test results for comparison of means between the qPCR cell equivalent (CE) concentrations of E. coli and enterococci from DNA extracted by auto and manual methods. Average CE/100 mL Samplesa
Automated method
Manual method
Mean difference
Standard error for mean difference
t-statistic (p)
E. coli RS PE SE TE
4.62Eþ06 6.79Eþ06 3.20Eþ03 2.58Eþ03
2.18Eþ06 2.24Eþ06 1.22Eþ03 1.50Eþ03
0.31 0.55 0.61 0.42
0.07 0.16 0.23 0.38
4.09 (0.001) 3.51 (0.004) 2.60 (0.02) 1.11 (0.28)
Enterococci RS PE SE TE
2.23Eþ07 1.99Eþ07 1.34Eþ05 1.49Eþ05
2.10Eþ07 1.58Eþ07 7.16Eþ04 7.96Eþ04
0.26 0.24 0.23 0.09
0.29 0.23 0.29 0.30
0.57 (0.57) 1.02 (0.32) 0.80 (0.43) 0.25 (0.80)
a n ¼ 9 for each type of sample. RS (raw sewage) PE (primary effluent-after primary clarification), SE (secondary effluent-after biological treatment and disinfection) TE (Teritiary effluent after dechlorination and filtration).
Analysis of variance (ANOVA) was performed to determine the differences in mean concentrations of target organisms in different treatment groups and if significant, multiple pairwise comparisons were carried out using Fisher’s Least Square Difference (LSD) test. Pearson’s correlation coefficient was used to test the relationship between conventional indicators and B. theta a. Linear regression analysis was performed using scatter plots of log10 cells/100 mL of E. coli and enterococci for comparison between manual and instrument DNA extraction from the samples tested and to estimate the coefficients of the linear equation for conventional indicators that best predicted the concentrations of B. theta a after various wastewater treatment processes.
Intervariability of total DNA concentration (ng/ml) by auto extraction showed a CV of 5.07% and manual DNA extraction showed a CV (%) of 20.65. The CE concentration of E. coli per qPCR reaction for RS showed a CV (%) of 0.86 for auto DNA extraction and 1.88 for manual DNA extraction whereas cell concentration of enterococci per reaction showed a CV (%) of 1.36 for auto DNA extraction and 1.83 for manual DNA extraction by the second individual. Thus, for further examination of the DNA signals through the treatment processes, the data from the auto DNA extraction were used.
3.2. Assessment of qPCR signals through sewage and septage treatment
3.
Results
3.1. Comparison of manual and automated extraction for E. coli and enterococci DNA Statistical analyses showed that qPCR cell equivalent (CE) concentrations of E. coli/100 mL of RS, PE, and SE were statistically higher for the autoextractor as compared to manual extraction ( p < 0.05) (Table 2). In tertiary effluent samples, both methods returned equivalent cell concentrations. For enterococci, there was no statistically significant difference found between extraction methods in any of the samples (Table 2). The mean difference of qPCR CE concentrations of E. coli and enterococci between auto extraction and manual extraction ranged from 0.31 to 0.61, and 0.09 to 0.26 log10 units, respectively, in various samples. Table 3 shows that total DNA concentration (ng/mL) extracted by the DNA extraction instrument had less variation (CV: 4.50%), than manual DNA extraction (CV: 13.29%). The qPCR CE concentration of E. coli per reaction for RS showed a CV (%) of 1.31 for auto DNA extraction and 1.67 for manual DNA extraction whereas cell concentration of enterococci per reaction for RS showed a CV (%) of 1.48 for auto DNA extraction and 1.71 for manual DNA extraction.
3.2.1. Occurrence of bacterial targets in raw sewage and treated effluents The average concentrations and standard deviations for the 216 samples (54 after each treatment location) collected from the wastewater treatment plant are shown in Table 4. The
Table 3 e Coefficient of variation (CV %) for DNA concentrations and cell equivalent concentrations of E. coli and enterococci performed by two individuals for auto and manual extraction methods. Sample (raw sewage)a
DNA concentration (ng/mL) Enterococci (CE/reaction) E. coli (CE/reaction)
Individual 1
Individual 2
Instrument Manual Instrument Manual CV%
CV%
CV%
CV%
4.50
13.29
5.07
20.65
1.48
1.71
1.36
1.83
1.31
1.67
0.86
1.88
a Raw sewage (n ¼ 9) from the same sample concentrate was used for the experiment.
2566
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 6 1 e2 5 7 2
Table 4 e Average log10 concentrations for E. coli, enterococci and B. thetaiotaomicron throughout wastewater treatment by cultivation and qPCR. Samplesa RSb PEc SEd TEe
E. coli (log10 CFU/100 mL) Mean Mean Mean Mean
6.21 6.17 1.01 1.04
Enterococci (log10 CFU/100 mL)
E. coli (log10 CE/100 mL)
5.72 (0.42) 5.58 (0.27) 0.64 (0.91) 0.75 (1.01)
6.46 (0.59) 6.48 (0.72) 3.05 (0.95) 2.82 (1.19)
(0.26) (0.26) (0.84) (0.93)
Enterococci (log10 CE/100 mL) 6.63 6.75 4.13 3.59
(0.51) (0.40) (0.84) (1.12)
B. theta a (log10 CE/100 mL) 7.26 7.31 4.19 3.67
(0.24) (0.41) (0.42) (0.60)
Number in parentheses represents standard deviation. a n ¼ 54 for each treatment location. b RS-Raw sewage. c PE-Primary effluent. d SE-secondary effluent. e TE-Tertiary effluent.
qPCR concentrations of E. coli in RS ranged from 1.47 105 to 1.48 107 cell equivalent/100 mL, for enterococci, ranged from 7.08 105 to 5.75 107 CE/100 mL, and for B. theta a, ranged from 7.76 106 to 5.68 107 CE/100 mL. The TE had a range of concentrations of 2.16 101 to 4.39 104 for E. coli, 6.35 100 to 3.81 105 for enterococci and 9.79 102 to 1.59 105 for B. theta a cell equivalent/100 mL. There were statistically significant differences in concentrations of all targets either by cultivation (represents cells that are alive and are in dynamic state) or qPCR methods (measure both dead and live cells, presenting a stable data) between samples collected prior to disinfection and after disinfection ( p < 0.05). Among the different target microorganisms, Fisher’s LSD showed that concentrations of B. theta a were significantly higher than E. coli or enterococci in all samples except in SE (effuent after biological treatment and disinfection) and TE (final effluent after dechlorination and filtration) where their concentrations were not significantly different from that of enterococci ( p < 0.05). Concentrations of E. coli were significantly higher than enterococci in all samples by cultivation methods; however with qPCR enumeration, enterococci concentrations were found to be significantly higher than E. coli in SE and TE samples ( p < 0.05).
3.2.2.
(after filtration), for CFU of E. coli and enterococci were found to be 4.99 and 4.92 log10 units and for qPCR cell equivalent of E. coli, enterococci, and B. theta a were 3.13, 2.60 and 3.51 log10, respectively (Fig. 1).
3.2.3. Correlation between conventional indicators and human specific marker in raw sewage and treated effluents Table 5 presents Pearson’s correlation coefficients between both the conventional indicators and B. theta a in treated and untreated group (SE and PE were grouped in this analysis) as well as pooled data (from all locations). The analysis was done on the pooled data sets in order to observe the relationship as the qPCR targets moved through the treatment process. Correlation coefficient (r) between E. coli and B. theta a was found to be 0.93 for pooled samples. In SE/PE samples and SE/ TE samples, the coefficients were 0.33 and 0.66, respectively. Correlation coefficient between enterococci and B. theta a was 0.88 for pooled data; in untreated and treated samples, the coefficients were 0.37 and 0.53, respectively. All these correlations were found to be statistically significant ( p < 0.05). Regression analysis was used to test the strength of association between conventional indicators and B. theta a in the pooled data set. The equation for each regression is displayed on Fig. 2. Fig. 2a shows scatter-plot for E. coli and B. theta
Removal during wastewater treatment process
During secondary activated sludge treatment prior to the chlorination step, there was a significant reduction in all target organisms. There was an average log10 removal of 2.69 and 3.05 for E. coli concentrations, respectively, by cultivation method and qPCR. For enterococci, the log10 reduction was 2.65 and 1.71, by cultivation method and qPCR, respectively. The log10 removal for B. theta a in this process was found to be 3.11, as measured by qPCR. When chlorine was added to the effluent, there was a further significant reduction in the average concentrations of cultivable cells; 2.42 log10 for E. coli and 2.63 log10 for enterococci. But as expected by qPCR methods, this reduction was only 0.31 and 0.40 log10 for E. coli and enterococci, respectively, and only 0.12 log10 for B. theta a cells, none of which were found to be significant. Filtration using rapid sand gravity filters did not significantly reduce any of the CFU or qPCR cell equivalents except for enterococci qPCR cells, which were reduced slightly by 0.49 log10 units. The overall log10 removal, which is the difference in log10 concentrations between raw sewage and tertiary effluent
Fig. 1 e Average log10 reductions of E. coli, enterococci (CFU and CE) and B. thetaiotaomicron CE by wastewater treatment.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 6 1 e2 5 7 2
Table 5 e Pearson’s correlation between conventional indicators and B. thetaiotaomicron in treated, untreated groups of data and pooled data by qPCR. Pearson’s correlation coefficient (R) Indicator bacteria
Untreateda
Treatedb
Pooledc
0.33 0.37
0.66 0.53
0.93 0.88
E. coli Enterococci
a Raw sewage and primary effluent data combined (n ¼ 108 samples). b Secondary and tertiary effluents data combined (n ¼ 108 samples). c Treated and untreated data combined (n ¼ 216 samples).
a concentrations from pooled data from four treatments which displayed strong correlations with R2 ¼ 0.87. The scatter-plot for enterococci and B. theta a from four treatments, which displayed strong correlation with R2 ¼ 0.79 is in Fig. 2b.
the range of 1.19 107 to 1.17 108 cells/100 mL. The final septage effluent had a range of concentration of 9.09 102 to 5.92 104 qPCR equivalent cells/100 mL of E. coli, 4.23 104 to 4.57 105 cells/100 mL of enterococci and 3.83 103 to 3.67 105 cells/100 mL of B. theta a. There was no significant difference between B. theta a and conventional indicators as measured by qPCR in raw septage and they were present in concentrations of >107 cells/100 mL. The mean log10 difference between concentrations of E. coli by cultivation methods and by qPCR in raw septage was 0.85 units whereas for enterococci, the mean log10 difference was 0.95. The mean log10 difference between concentrations of enterococci by cultivation methods and that by qPCR in effluent was 1.25 log10 units. All these differences were statistically significant ( p < 0.05). However, there was no statistically significant difference found between cultivable levels and qPCR equivalent cells for E. coli in septage effluent (there was no disinfection used in the septage treatment process train).
3.2.5. 3.2.4. Occurrence of bacterial targets in raw septage and treated effluent
Log10 B. thetaiotaomicron concentration, (CE/100 mL)
The average concentrations and standard deviations for the 48 samples (24 each for raw septage and treated effluent) collected from the septage treatment plant are shown in Table 6. The qPCR equivalent concentrations of E. coli in raw septage ranged from 6.80 106 to 6.23 108 cells/100 mL, enterococci ranged from 3.72 106 to 6.235 107 cells/100 mL and B. theta a were in
10.00
a y = 0.8523x + 1.5687 R² = 0.8766
8.00 6.00
2.00
Log10 B. thetaiotaomicron concentration, (CE/100 mL)
Raw sewage
10.00
2.00 4.00 6.00 Log10 E. coli concentration (CE/100 mL)
Primary effluent
Secondary effluent
8.00
Tertiary effluent
y = 0.9739x + 0.4525 R² = 0.7916
b
8.00 6.00 4.00 2.00 0.00 0.00
2.00
4.00
6.00
8.00
Log10 enterococci concentration, (CE/100 mL) Raw sewage
Primary effluent
Removal during septage treatment process
Following the septage treatment process, there was a significant average log10 reduction in all target organisms, shown in Fig. 3. The highest log10 reduction during treatment was found for E. coli qPCR equivalent cells with a log10 difference of 3.82 whereas the difference in log10 concentrations for E. coli as measured by cultivation methods was 2.52. However, this difference in log10 removals between both methods was due to the higher initial qPCR equivalent concentrations in raw septage. The log10 reduction for enterococci by the cultivation method was 2.29 whereas this difference was only 1.99 by qPCR. The log10 reduction for B. theta a was found to be 3.13.
3.2.6. Correlation between conventional indicators and human specific marker in raw septage and treated septage effluents
4.00
0.00 0.00
2567
Seondary effluent
Tertiary effluent
Treated and untreated data pooled, n=216
Fig. 2 e Correlation between a) E. coli and B. thetaiotaomicron, and b) enterococci and B. thetaiotaomicron by qPCR from wastewater treatment samples.
Regression analysis was used to test the strength of association between conventional indicators and B. theta a using qPCR (Fig. 4). The equation for each regression is displayed on each chart. E. coli and B. theta a concentrations displayed strong correlations with R2 ¼ 0.91; and enterococci and B. theta a showed correlation of R2 ¼ 0.92.
4.
Discussion
Comparison of commercial kits for extraction of DNA from water samples have been undertaken (Lebuhn et al., 2005; Rose et al., 2003), however, in this study, an automated method for extracting DNA was evaluated, and showed significantly higher or equal efficiency compared to manual extraction using a commercially available kit for extracting DNA from treated and untreated sewage. The same instrument had been used for the study of genetic diversity of Legionella spp. (Wullings and van der Kooij, 2006), but the auto extraction method was not evaluated or compared with any other existing methods as method evaluation was not a goal of their study. Advantages in using the autoextractor included improved consistency and decreased variability. Based on the DNA concentrations, this appeared to be a recovery issue, and
2568
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 6 1 e2 5 7 2
Table 6 e Average log10 concentrations of bacteria in raw septage and effluent by cultivation and qPCR methods. Samplesa Raw septage Effluent
E. coli (log10 CFU/100 mL)
Enterococci (log10 CFU/100 mL)
E. coli (log10 CE/100 mL)
Enterococci (log10 CE/100 mL)
B. thetaiotaomicron (log10 CE/100 mL)
6.47 (0.45) 3.96 (0.86)
6.36 (0.82) 4.07 (0.96)
7.33 (0.68) 3.51 (0.67)
7.31 (0.36) 5.32 (0.28)
7.55 (0.34) 4.42 (0.55)
a Total number of samples, n ¼ 48 (24 samples for each treatment).
while removal of inhibitors was not evaluated, the available technical information by the manufacturers suggests similar inhibitor removal efficiency as the basic steps of extraction are the same. However, the performance of various extraction methods in removing complex inhibitors from environmental samples should be explored further. With the pre-treatment step included, the time it took for extracting DNA from eight samples was slightly less when using the automated method (1 h) as compared to the manual kit (1.5 h). Previous studies in clinical samples have also compared the cost of DNA extraction using commercially available manual kits and automated methods (Knepp et al., 2003). Roche MagNaPure LC extraction is currently more expensive, with QIAamp kit and the MagNaPure reported to cost $2.64 and $3.58 per specimen, respectively (the price excludes cost of plastics). Scientific developments such as qPCR will likely yield changes in future regulations and monitoring of drinking and recreational waters. Therefore, efficient and reliable data become important issues for these emerging techniques. These preliminary studies have demonstrated that automation can be used to improve efficiency and reproducibility of DNA extraction and that qPCR can be used to describe bacterial concentrations in wastewater. New instrument configurations for automation, which can handle more samples and a greater diversity of matrices for environmental testing, would be beneficial in analysis of wastewater and recreational waters. This study characterized human wastewater treatment environments for a novel MST target, B. thetaiotaomicron, in comparison with the conventional indicators E. coli and enterococci. To our knowledge, this study is the first to contrast raw sewage and septage, and the fate of molecular signals of all these bacterial indicators during waste treatment. The B. thetaiotaomicron qPCR equivalent cells were present in significantly higher concentrations than that of E.
E.coli (CFU/100mL) enterococci (CFU/100mL)
2.52 2.29
E.coli (CE/100mL)
3.82
B. thetaiotaomicron (CE/100mL) enterococci (CE/100mL)
coli or enterococci in raw sewage but similar in septage. Yampara-Iquise et al. (2008) found a wide range of B. thetaiotaomicron (6.88 102e1.07 109 cells/g) in human feces, when stool samples from 10 human subjects were analyzed by qPCR targeting the a-mannanase 1-6 gene. This variation was suggested to be attributed to the variation in DNA extraction efficiency, yet no supporting data were available in the publication. Other human fecal MST markers, such as the host-specific BacteroidesePrevotella 16S rRNA gene markers, have been detected in raw sewage and septage at concentrations of 108e109 cells/100 mL (Seurinck et al., 2005; Silkie and Nelson, 2009; Sercu et al., 2009) whereas in human feces, concentrations ranging from 105e1011 cells/gram (wet weight) were observed (Seurinck et al., 2005; Okabe et al., 2007; Sercu et al., 2009). Comparison between Bacteroides 16S rRNA human markers, whose exact gene copy number is not known, and B. thetaiotaomicron a-mannanase 1-6 gene, where one copy of gene represents one cell, may not be the best approach for determining the sensitivity of these markers. E. coli was found to occur at concentrations of 106 target gene copies/100 mL of raw sewage in this study, which is almost three logs lower than the concentrations reported by another study, where 109 copies/100 mL of lacZ gene targets for E. coli by qPCR were found (Wery et al., 2008). But in five other studies, in which the uidA gene was targeted, qPCR cell concentrations of E. coli in raw sewage agree with this study (Table 7). One possibility for such high numbers of E. coli found by Wery et al. (2008) could be due to the cross amplification of the E. coli lacZ qPCR assay with other bacterial strains (coliforms) that harbor this gene, tested only with four bacteria during the assay development (Foulds et al., 2002). Enterococci qPCR assays published before have mostly targeted 23S rRNA
3.13 1.99
Number of samples, n=48 (24 samples from each treatment location)
Fig. 3 e Average log reduction of E. coli, enterococci, B. thetaiotaomicron targets by cultivation and qPCR during septage treatment process.
Total number of samples, n=48 (24 samples from raw septage and 24 treated effluents) RS1= E. coli vs B. theta α in raw septage RS2= enterococci vs B. theta α in raw septage Eff1= E. coli vs B. theta α in treated effluent Eff2= enterococci vs B. theta α in treated effluent
Fig. 4 e Correlation between E. coli, enterococci and B. thetaiotaomicron by qPCR in septage treatment process.
2569
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 6 1 e2 5 7 2
Table 7 e Summary of qPCR concentrations for various DNA targets in wastewater and treated effluents. Treatment
RS PE SE (Pre-cl) SE (Post-cl) TE Septage STE
Sampling
Grab Grab Grab Grab Grab Grab Grab
Samples, n
E. colia
54 54 18 54 54 24 24
uidA* 2.88Eþ06 3.02Eþ06 1.01Eþ03 1.12Eþ03 6.61Eþ02 2.14Eþ06 3.24Eþ03
RS PE SE (Pre-cl) SE (Post-cl)
Grab
9 9 9 9
RS
Grab
12
Enterococcia
Bacteroides Human Markera
References
23S rRNA* 1.71Eþ07 2.25Eþ07 1.68Eþ05 5.40Eþ04 1.56Eþ04 8.17Eþ07 8.36Eþ05
alpha-mannanase b,* 1.82Eþ07 2.04Eþ07 1.84Eþ04 1.55Eþ04 4.68Eþ03 3.55Eþ07 2.63Eþ04
e
2.76Eþ07 3.71Eþ07 4.12Eþ05 1.58Eþ05
16S rRNAc 1.75Eþ09 1.59Eþ09 3.47Eþ07 3.17Eþ07
uidA* 2.00Eþ07
23S rRNA* 1.00Eþ07
16S rRNAc 7.94Eþ08
Silkie and Nelson 2009
16S rRNAc 7.8Eþ08 3.9Eþ08
Sercu et al., 2009
RS Septage
Grab Grab
3 3
e e
e e
RS PE TE (UV disinfected)
Grab Grab Grab
5 5 5
uidA* e e e 2.85d
e e e 2.59d
RS PE SE (Pre-cl)
composite
6 6 6
RS PE
Grab Grab
RS
RS TE (UV disinfected)
e e e
This study
Varma et al., 2009
Lavender and Kinzelman 2009
lacZ* 1.34Eþ09 1.05Eþ09 2.95Eþ05
e
e
12 12
e
e
16SrRNAc 2.10Eþ05 2.90Eþ05
Savichtcheva et al., 2007
Grab
4
e
e
16SrRNAc 1.72Eþ10
Seurinck et al., 2005
Grab Grab
3 3
uidA* 1.66Eþ07 6.12Eþ02
e
e
Lee et al., 2006
Wery et al., 2008
RS ¼ Raw sewage. PE ¼ Primary effluent after grit removal. SE(Pre-cl) ¼ Secondary effluent from secondary clarifier prior to chlorination. SE(Post-cl) ¼ Secondary effluent after disinfection. TE ¼ Teritiary effluent after filtration or UV disinfection. STE ¼ septage treated effluent. *Target gene. a qPCR target gene copies/100 mL. b B. thetaiotaomicron. c Order Bacteriodales. d Log10 difference between raw sewage and tertiary effluent (Initial raw sewage or final effluent concentrations not reported).
gene using the same primer sets as used in this study and have found similar concentrations of target gene copies in raw sewage. Log10 removal of enterococci qPCR signals across the wastewater treatment found in this study is also very similar to the study by Varma et al. (2009) (Table 7). Primary treatment in the wastewater treatment process train (primary sedimentation) does not reduce fecal bacteria and this has been shown with CFU and is similar when monitoring DNA targets as shown in this study and by others (Lucena et al., 2004; Puig et al., 2010). Significant reduction in
all targets (both cultivable and qPCR equivalent cells) were observed during the secondary treatment process comprising of biological process (activated sludge), sedimentation and disinfection (chlorination). However there was a difference between enterococci (gram positive cells) and the gram negative cells of E. coli and B. thetaiotaomicron. Enterococci were the least removed during secondary treatment in both sewage and septage but were better removed by filtration by rapid sand filters. Loss during secondary treatment before disinfection may be due to biological and sedimentation
2570
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 6 1 e2 5 7 2
processes (e.g. attachment to solids, sedimentation, predation by protozoa), thus a difference in attachment to activated sludge versus sand of various types of bacterial cells with different cell-wall properties needs further exploration. Sludge was not monitored for in this study; however, others have observed that in liquid sludge, 3.7 log10 higher concentrations of E. coli were found by qPCR compared to cultivation (Wery et al., 2008). This study observed that the rapid sand filtration used in this wastewater treatment facility did not adequately remove the cells. In the absence of coagulants, microorganisms may not be retained efficiently in the filters thus increasing their concentrations in final effluents (Koivunen et al., 2003). There is currently a disconnect between sewage treatment monitoring requirements and recreational standards or criteria. Under the Clean Water Act of 1972, wastewater treatment plant discharge into streams and the effluent has to be monitored to meet the permitted levels of pollutants specified in NPDES. Wastewater treatment plants use mostly fecal coliforms to monitor the treatment efficacy (primarily disinfection). Total and fecal coliforms were used for monitoring recreational water quality after epidemiological studies conducted by U.S. Public Health Service (PHS) found that any increase in the levels of total coliforms could possibly be used as a “warning signal” (Stevenson, 1953) for public health risk and related this to the numbers of fecal coliform bacteria in the water. Later, U. S. EPA adapted E. coli and enterococci as new water quality criteria based on other epidemiological studies conducted in marine and fresh water beaches USEPA (1986). A qPCR method is an effective tool that gives quicker results as compared to cultivation, and is being explored for same day analysis to address beach closures. One key limitation of qPCR methods is the inability to differentiate between live and dead cells and we found significant difference between qPCR and cultivable levels of E. coli and enterococci following disinfection (chlorination). However, it is not clear if this difference in number is solely due to the dead cells or if some of the cells had transformed into a viable but non-cultivable stressed state, which prevents their growth in culture media. Cell death upon chlorination is attributed to different mechanisms such as membrane damage, intracellular or extracellular enzyme inactivation, or uncoupling of the electron chain (Virto et al., 2005). Where low concentrations of chlorine may alter the E. coli cytoplasmic membrane and injure the cells without necessarily causing the cell inactivation, a higher dosage is required to cause the damage to nucleic acids (Lisle et al., 1998; Phe et al., 2007). Resistance to chlorine by different cell-wall types in the presence of organic matter has also been suggested (Virto et al., 2005). Viable qPCR methods using dyes such as Ethidium Monoazide (EMA) and Propidium Monoazide (PMA) distinguish live cells and dead cells (heat killed) but when applied to untreated and treated sewage samples in this study as well as by Varma et al. (2009) there was no observed reduction in the qPCR signals. After disinfection, the qPCR signal was, definitely not comparable to colony counts suggesting EMA was not able to penetrate some of the cells present in the disinfected effluent samples (data not shown). It is not clear if this is due to the effect of turbidity reducing the photolysis of the dyes and binding with DNA, no change in membrane permeability, or if it is due to the
presence of viable but non-culturable cells present in the disinfected effluents. Thus currently qPCR can be used a tool to monitor loading and physical removal or dilution but cannot be used to address viability. The Beaches Environmental Assessment and Coastal Health (BEACH) Act of 2000 is supporting science for qPCR to monitor recreation beaches (EPA, 2000). A recent study found qPCR gave higher values than cultivation based methods for recreational water samples from sites closer to treated effluent discharge points (Byappanahalli et al., 2010). In contrast, another study suggested underestimation bias for qPCR (Noble et al., 2010) and the sampling sites for this study were far from such outfalls. These contradictory results suggest that qPCR tool may underestimate the public health risks or cause unnecessary economic losses due to beach closures when used for regulatory purposes. The levels of qPCR signals detected from wastewater effluents (as in this current study) may be very useful in mapping the fate and transport of these signals from discharge points to exposure sites, getting a better understanding of wastewater impacts on recreational sites. These signals should also be explored in relationship to virus and parasite pathogens as well. The discharged effluents from wastewater treatment plant in this study had a log10 qPCR cell concentration of 3.59 per 100 mL of enterococci. The B. thetaiotaomicron qPCR equivalent cell levels (log10 average of 3.67 cells/100 mL) were almost at the same levels as that of enterococci in 100 mL of treated final effluents. Following a CSO or SSO event the levels of these signals would be much higher. Wade et al. (2006) found a strong correlation between enterococci qPCR daily average concentrations and swimming related gastrointestinal illnesses at two of the Great Lakes beaches they studied by applying the 23S rRNA qPCR assay for enterococci. According to their study, a log10 increase in the daily average of qPCR cell equivalents was associated with 1.30 (95% CI, 1.08e1.57) increase in the odds of gastrointestinal illness for any contact with water. The enterococci concentrations found at the beaches ranged from 1.90 to 2.04 log10 qPCR cell equivalents per 100 mL. Wastewater effluents were suspected as impacting these beaches. These correlations between qPCR signals and public health risks, and the high qPCR target signals detected in the wastewater effluents in the current study suggest that in spite of the viability issues; qPCR may be a valuable tool in monitoring water environment for both enterococci and Bacteroides to protect public health risk.
5.
Conclusions
Automated methods can be effectively used to extract DNA from water and wastewater samples, thus reducing the risks of cross contamination and human errors. B. thetaiotaomicron qPCR equivalent cells based on the a-mannanase target were present at statistically significantly higher concentrations (w7 log10) than that of E. coli or enterococci (w6 log10) in raw sewage ( p < 0.05) and at the same levels in raw septage (7.3e7.5 log10). There was a significant correlation between this MST marker and each of the conventional indicators throughout the waste treatment process for both wastewater and
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 6 1 e2 5 7 2
septage ( p < 0.05). A clear difference was observed between indicators and their removal rates in different treatment processes; enterococci were better removed by filtration and E. coli and B. thetaiotaomicron were better removed by secondary treatment. Effluents discharged from wastewater treatment plants have concentrations of 2e3 log10 cell equivalents/100 mL of both the conventional indicators and B. thetaiotaomicron human marker.
Acknowledgments This research was funded in part by the grants from the U.S. Environmental Protection Agency (RD83300501) and the National Oceanic and Atmospheric Administration (Grant # NA04OAR4600199).
references
Byappanahalli, M.N., Whitman, R.L., Shively, D.A., Nevers, M.B., 2010. Linking non-culturable (qPCR) and culturable enterococci densities with hydrometeorological conditions. Science of the Total Environment 408 (16), 3096e3101. Fong, T.T., Mansfield, L.S., Wilson, D.L., Schwab, D.J., Molloy, S.L., Rose, J.B., 2007. Massive microbiological groundwater contamination associated with a waterborne outbreak in Lake Erie, South Bass Island, Ohio. Environmental Health Perspectives 115 (6), 856e864. Foulds, I.V., Granacki, A., Xiao, C., Krull, U.J., Castle, A., Horgen, P.A., 2002. Quantification of microcystin-producing cyanobacteria and E-coli in water by 50 -nuclease PCR. Journal of Applied Microbiology 93 (5), 825e834. Frahm, E., Obst, U., 2003. Application of the fluorogenic probe technique (TaqMan PCR) to the detection of Enterococcus spp. and Escherichia coli in water samples. Journal of Microbiological Methods 52 (1), 123e131. Haugland, R.A., Siefring, S.C., Wymer, L.J., Brenner, K.P., Dufour, A.P., 2005. Comparison of Enterococcus measurements in freshwater at two recreational beaches by quantitative polymerase chain reaction and membrane filter culture analysis. Water Research 39 (4), 559e568. Kamel, A.H., Ali, M.A., El-Nady, H.G., Aho, S., Pothier, P., Belliot, G., 2010. Evidence of the co-circulation of enteric viruses in sewage and in the population of Greater Cairo. Journal of Applied Microbiology 108 (5), 1620e1629. Kildare, B.J., Leutenegger, C.M., McSwain, B.S., Bambic, D.G., Rajal, V.B., Wuertz, S., 2007. 16S rRNA-based assays for quantitative detection of universal, human-, cow-, and dogspecific fecal Bacteroidales: a Bayesian approach. Water Research 41 (16), 3701e3715. Knepp, J.H., Geahr, M.A., Forman, M.S., Valsamakis, A., 2003. Comparison of automated and manual nucleic acid extraction methods for detection of enterovirus RNA. Journal of Clinical Microbiology 41 (8), 3532e3536. Koivunen, J., Siitonen, A., Heinonen-Tanski, H., 2003. Elimination of enteric bacteria in biologicalechemical wastewater treatment and tertiary filtration units. Water Research 37 (3), 690e698. 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
2571
water quality monitoring programs. Water Research 43 (19), 4967e4979. Layton, A., McKay, L., Williams, D., Garrett, V., Gentry, R., Sayler, G., 2006. Development of Bacteroides 16S rRNA gene TaqMan-based real-time PCR assays for estimation of total, human, and bovine fecal pollution in water. Applied and Environmental Microbiology 72, 4214e4224. Lebuhn, M., Effenberger, M., Garces, G., Gronauer, A., Wilderer, P.A., 2005. Hygienization by anaerobic digestion: comparison between evaluation by cultivation and quantitative real-time PCR. Water Science and Technology 52 (1e2), 93e99. Lee, D.Y., Shannon, K., Beaudette, L.A., 2006. Detection of bacterial pathogens oligonucleotide microarray in municipal wastewater using an ind real-time quantitative PCR. Journal of Microbiological Methods 65 (3), 453e467. Lisle, J.T., Broadway, S.C., Prescott, A.M., Pyle, B.H., Fricker, C., McFeters, G.A., 1998. Effects of starvation on physiological activity and chlorine disinfection resistance in Escherichia coli O157:H7. Applied and Environmental Microbiology 64 (12), 4658e4662. Lucena, F., Duran, A.E., Moron, A., Calderon, E., Campos, C., Gantzer, C., Skraber, S., Jofre, J., 2004. Reduction of bacterial indicators and bacteriophages infecting faecal bacteria in primary and secondary wastewater treatments. Journal of Applied Microbiology 97 (5), 1069e1076. McLain, J.E.T., Ryu, H., Kabiri-Badr, L., Rock, C.M., Abbaszadegan, M., 2009. Lack of specificity for PCR assays targeting human Bacteroides 16S rRNA gene: cross-amplification with fish feces. FEMS Microbiology Letters 299 (1), 38e43. Messer, J.W., Dufour, A.P., 1998. A rapid, specific membrane filtration procedure for enumeration of enterococci in recreational water. Applied and Environmental Microbiology 64 (2), 678e680. Nayak, A.K., Rose, J.B., 2007. Detection of Helicobacter pylori in sewage and water using a new quantitative PCR method with SYBR (R) green. Journal of Applied Microbiology 103 (5), 1931e1941. Noble, R.T., Griffith, J.F., Blackwood, A.D., Fuhrman, J.A., Gregory, J.B., Hernandez, X., Liang, X.L., Bera, A.A., Schiff, K., 2006. Multitiered approach using quantitative PCR to track sources of fecal pollution affecting Santa Monica Bay, California. Applied and Environmental Microbiology 72 (2), 1604e1612. Noble, R.T., Blackwood, A.D., Griffith, J.F., McGee, C.D., Weisberg, S.B., 2010. Comparison of rapid quantitative PCR-based and conventional culture-based methods for enumeration of Enterococcus spp. and Escherichia coli in recreational waters. Applied and Environmental Microbiology 76 (22), 7437e7443. Okabe, S., Okayama, N., Savichtcheva, O., Ito, T., 2007. Quantification of host-specific BacteroidesePrevotella 16S rRNA genetic markers for assessment of fecal pollution in freshwater. Applied Microbiology and Biotechnology 74 (4), 890e901. Phe, M.H., Dossot, M., Guilloteau, H., Block, J.C., 2007. Highly chlorinated Escherichia coli cannot be stained by propidium iodide. Canadian Journal of Microbiology 53 (5), 664e670. Puig, S., van Loosdrecht, M.C.M., Flameling, A.G., Colprim, J., Meijer, S.C.F., 2010. The effect of primary sedimentation on full-scale WWTP nutrient removal performance. Water Research 44 (11), 3375e3384. Robertson, L.J., Hermansen, L., Gjerde, B.K., 2006. Occurrence of Cryptosporidium oocysts and Giardia cysts in sewage in Norway. Applied and Environmental Microbiology 72 (8), 5297e5303. Rose, P., Harkin, J.M., Hickey, W.J., 2003. Competitive touchdown PCR for estimation of Escherichia coli DNA recovery in soil DNA extraction. Journal of Microbiological Methods 52 (1), 29e38. Sadowsky, M.J., Call, D.R., Santo Domingo, J.W., 2007. Application of microbial source tracking studies. In: Santo Domingo, J.W.,
2572
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 6 1 e2 5 7 2
Sadowsky, M.J. (Eds.), Microbial Source Tracking. ASM Press, Washington, DC, pp. 235e277. Savichtcheva, O., Okayama, N., Okabe, S., 2007. Relationships between Bacteroides 16S rRNA genetic markers and presence of bacterial enteric pathogens and conventional fecal indicators. Water Research 41 (16), 3615e3628. Scott, T.M., Rose, J.B., Jenkins, T.M., Farrah, S.R., Lukasik, J., 2002. Microbial source tracking: current methodology and future directions. Applied and Environmental Microbiology 68 (12), 5796e5803. Sercu, B., Van De Werfhorst, L.C., Murray, J., Holden, P.A., 2009. Storm drains are sources of human fecal pollution during dry weather in three urban southern california watersheds. Environmental Science & Technology 43 (2), 293e298. Seurinck, S., Defoirdt, T., Verstraete, W., Siciliano, S.D., 2005. Detection and quantification of the human-specific HF183 Bacteroides 16S rRNA genetic marker with real-time PCR for assessment of human faecal pollution in freshwater. Environmental Microbiology 7 (2), 249e259. Silkie, S.S., Nelson, K.L., 2009. Concentrations of host-specific and generic fecal markers measured by quantitative PCR in raw sewage and fresh animal feces. Water Research 43 (19), 4860e4871. Simpson, J.M., Santo Domingo, J.W., Reasoner, D.J., 2002. Microbial source tracking: state of the science. Environmental Science & Technology 36 (24), 5279e5288. Stevenson, A.H., 1953. Studies of bathing water quality and health. American Journal of Public Health 43 (5), 529e538. USEPA, 1986. Bacteriological Water Quality Criteria for Marine and Fresh Recreational Waters. EPA-440/5-84-002. U.S. Environmental Protection Agency, Office of Water Regulations and Standards, Cincinnati, OH. USEPA, 2002. Enterococcus spp. in Water by Membrane Filter Using Membrane Enterococcus Indoxyl-B-D-Glucoside Agar (MEI). EPA-821-R-02e022. Office of Water, Washington D.C. USEPA, 2005a. Escherichia coli (E. coli) in Water by Membrane Filtration Using Modified Membrane-Thermotolerant
Escherichia coli Agar (Modified MTEC). EPA 821-R-04e025. Office of Water, Washington D.C. USEPA, 2005b. Microbial Source Tracking Guide Document. EPA-600/R-05/064. Office of Research and Development, Washington D. C. Varma, M., Field, R., Stinson, M., Rukovets, B., Wymer, L., Hauglanda, R., 2009. Quantitative real-time PCR analysis of total and propidium monoazide-resistant fecal indicator bacteria in wastewater. Water Research 43 (19), 4790e4801. Viau, E., Peccia, J., 2009. Evaluation of the enterococci indicator in biosolids using culture-based and quantitative PCR assays. Water Research 43 (19), 4878e4887. Virto, R., Manas, P., Alvarez, I., Condon, S., Raso, J., 2005. Membrane damage and microbial inactivation by chlorine in the absence and presence of a chlorine-demanding substrate. Applied and Environmental Microbiology 71 (9), 5022e5028. 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 gastrointestional illness. Environmental Health Perspectives 114 (1), 24e28. Wery, N., Lhoutellier, C., Ducray, F., Delgenes, J.P., Godon, J.J., 2008. Behaviour of pathogenic and indicator bacteria during urban wastewater treatment and sludge composting, as revealed by quantitative PCR. Water Research 42 (1e2), 53e62. Wexler, H.M., 2007. Bacteroides: the good, the bad, and the nittygritty. Clinical Microbiological Reviews 20, 593e621. Wullings, B.A., van der Kooij, D., 2006. Occurrence and genetic diversity of uncultured Legionella spp. in drinking water treated at temperatures below 15 C. Applied and Environmental Microbiology 72 (1), 157e166. Yampara-Iquise, H., Zheng, G., Jones, J.E., Carson, C.A., 2008. Use of a Bacteroides thetaiotaomicron-specific alpha-1-6, mannanase quantitative PCR to detect human faecal pollution in water. Journal of Applied Microbiology 105 (5), 1686e1693.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 7 3 e2 5 8 6
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Assessment of global nitrogen pollution in rivers using an integrated biogeochemical modeling framework Bin He a,b,*, Shinjiro Kanae c, Taikan Oki d, Yukiko Hirabayashi e, Yosuke Yamashiki b, Kaoru Takara b a
Center for Promotion of Interdisciplinary Education and Research, Educational Unit for Adaptation and Resilience for a Sustainable Society, Kyoto University, Kyoto, Japan b Disaster Prevention Research Institute (DPRI), Kyoto University, Kyoto, Japan c Department of Mechanical and Environmental Informatics, Tokyo Institute of Technology, Japan d Institute of Industrial Science, The University of Tokyo, Japan e Institute of Engineering Innovation, The University of Tokyo, Japan
article info
abstract
Article history:
This study has analyzed the global nitrogen loading of rivers resulting from atmospheric
Received 14 September 2010
deposition, direct discharge, and nitrogenous compounds generated by residential,
Received in revised form
industrial, and agricultural sources. Fertilizer use, population distribution, land cover, and
3 February 2011
social census data were used in this study. A terrestrial nitrogen cycle model with a 24-h
Accepted 10 February 2011
time step and 0.5 spatial resolution was developed to estimate nitrogen leaching from soil
Available online 19 February 2011
layers in farmlands, grasslands, and natural lands. The N-cycle in this model includes the major processes of nitrogen fixation, nitrification, denitrification, immobilization, miner-
Keywords:
alization, leaching, and nitrogen absorption by vegetation. The previously developed Total
Global rivers
Runoff Integrating Pathways network was used to analyze nitrogen transport from natural
Water quality
and anthropogenic sources through river channels, as well as the collecting and routing of
Nitrogen pollution load
nitrogen to river mouths by runoff. Model performance was evaluated through nutrient
Terrestrial ecosystem
data measured at 61 locations in several major world river basins. The dissolved inorganic
Anthropogenic sources
nitrogen concentrations calculated by the model agreed well with the observed data and demonstrate the reliability of the proposed model. The results indicate that nitrogen loading in most global rivers is proportional to the size of the river basin. Reduced nitrate leaching was predicted for basins with low population density, such as those at high latitudes or in arid regions. Nitrate concentration becomes especially high in tropical humid river basins, densely populated basins, and basins with extensive agricultural activity. On a global scale, agriculture has a significant impact on the distribution of nitrogenous compound pollution. The map of nitrate distribution indicates that serious nitrogen pollution (nitrate concentration: 10e50 mg N/L) has occurred in areas with significant agricultural activities and small precipitation surpluses. Analysis of the model uncertainty also suggests that the nitrate export in most rivers is sensitive to the amount of nitrogen leaching from agricultural lands. ª 2011 Elsevier Ltd. All rights reserved.
* Corresponding author. Center for Promotion of Interdisciplinary Education and Research, Educational Unit for Adaptation and Resilience for a Sustainable Society, Kyoto University, Kyoto, Japan. E-mail address:
[email protected] (B. He). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.02.011
2574
1.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 7 3 e2 5 8 6
Introduction
Increasing world population has resulted in higher food and energy demand and consumption over the past half century (United Nations, 1996). Human activities have greatly accelerated and enlarged the natural cycles of nutrients and nitrogen in the soil, water, and atmosphere. Through activities such as fertilizer application, fossil fuel consumption, and leguminous crop production, humans have more than doubled the rate at which biologically available nutrients enter the terrestrial biosphere in comparison to pre-industrial levels (Galloway et al., 2004). One of the most important nutrients in this respect is nitrogen, which is an integral component of many essential plant nutrients. However, while nitrogen is an essential nutrient that plays important roles in increasing crop yields and quality, it is also a major pollutant in terrestrial ecosystems (Baker, 2003; Oenema et al., 1998; Schepers et al., 1995). Excess nitrogen used in fertilization has disturbed the biogeochemical nitrogen cycle of natural ecosystems, resulting in stratospheric ozone depletion, soil acidification, eutrophication, and nitrate pollution of ground and surface waters (Davis and Koop, 2006; Ding et al., 2006; Hantschel and Beese, 1997; Rijtema and Kroes, 1991). Water quality degradation associated with nitrate leaching from agricultural soils is an important environmental issue worldwide (Galloway, 1998, 2000; Galloway and Cowling, 2002; Galloway et al., 1995). Losses of nitrogenous compounds in the atmosphere and aquatic systems have inverse impacts not only on human health and global warming, but also on natural and agricultural terrestrial and aquatic ecosystems. The effects of agricultural diffuse source nitrogen pollution on water quality and aquatic ecosystems have received considerable research attention in recent years (Howarth et al., 2002; Hudson et al., 2005). Nitrogen pollution is one of the major pollutants, yet it is difficult to estimate because its sources are widely spread. In addition, both natural and anthropogenic emitters are responsible for nitrogen pollution (He et al., 2009a and 2009b). For example, natural reactions of atmospheric forms of nitrogen can result in the formation of nitrate and ammonium ions. In addition, the large anthropogenic sources of septic tanks, application of nitrogen-rich fertilizers, and agricultural processes have greatly increased the nitrate concentration, particularly in groundwater. Since the characteristics of each river basin is different, the relative contribution from each emitters has to be analyzed based on the database of land use, population, agricultural fertilizer, industrial production, livestock, etc. To date, research on the nitrogen cycle has primarily focused on the river basin scale (Dumont et al., 2005; Seitzinger et al., 2005). Very few national- or global-scale studies exist (Dumont et al., 2005), and prediction of nitrogen export is still insufficient (Seitzinger et al., 2005). In addition, most large- or global-scale nitrogen studies have treated entire river basins as the basic unit and, as a result, the calculated nitrogen leaching or nitrate concentration mainly reflects the amount of nitrogen in river outlets (Bouwman et al., 2005a; Harrison et al., 2005; Howarth et al., 2002; He et al., 2009b). Detailed information on nitrogen leaching or nitrate concentration in individually distributed grids is lacking in the current literature. Furthermore, global nitrogen
cycle models have relied on calculated amounts of nitrogen fertilizer application based on yearly statistical databases for each country. Monthly nitrogen fertilizer application amounts have not been available for global-scale study. The aim of this study was to estimate the global nitrogen loading from point and nonpoint sources separately and apply an integrated biogeochemical model to nitrogen export for global rivers. The nitrogen fertilizer application amount and nitrate leaching were first calculated for each grid box, measuring 0.5 by 0.5 , using the process-based N-cycle model. In this article, we present an initial overview of the integrated modeling framework, including the structure, database, and model results. Section 2 presents the methodology, and Section 3 discusses the database for the point and nonpoint sources. The integrated simulation using the above model and database is discussed in Section 4.
2.
Method
2.1.
Integrated modeling framework
This study proposed an integrated biogeochemical modeling of global nitrogen loads from anthropogenic and natural sources. The global runoff was simulated by a land surface model driven by atmospheric forcing in an off-line mode (Fig. 1). Then, the nitrogen load (NL) from different sources such as crop, livestock, industrial plant, urban and rural population were calculated by applying datasets of fertilizer utilization, population distribution, land cover map, and social census. The number of livestock and population in each country was collected from national census database. The
Fig. 1 e Integrated modelling framework for calculating global terrestrial nitrogen load and rivers’ nitrogen concentration.
2575
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 7 3 e2 5 8 6
Land surface model
Several land surface models have been developed for use in global or regional climate models (Sellers et al., 1996; Dickinson et al., 1998). These models incorporate radiation transfer, evaporation, transpiration, snow, runoff and also take into consideration the effects of vegetation. In the result, the energy and water exchange between the land and atmosphere is illustrated as a vertical one dimensional processes. In this study, we employed the Minimal Advanced Treatment of Surface Interaction and Runoff Model (MATSIRO) which is projected to be used for long-term simulations of climate studies (Takata, 2000, 2001; Takata et al., 2003). The MATSIRO model computes vertical energy and water fluxes in a grid cell based upon specifications of soil properties and vegetation coverage for each grid (Hirabayashi et al., 2005). It was used to estimate the long-term terrestrial water fluxes by long-term atmospheric forcing data that was stochastically estimated from monthly mean time series of precipitation and temperature (Hirabayashi et al., 2005, 2008). MATSIRO model has a single-layer albedo. The bulk exchange coefficients are evaluated based on a multilayer canopy model. The fluxes are calculated from the energy balance at the ground and canopy surfaces in both snow-free and snow-covered portions that consider the subgrid snow distribution. The snow has the variable number of layers from one to three in accordance with snow water equivalent (SWE), and the snow temperature is calculated by a thermal conduction equation. The snow mass is prognosticated from snowfall, snowmelt, refreezing of rainfall and snowmelt, and sublimation. The detailed description of snow process can be found in Takata et al. (2003). Evaporation of water on the canopy and transpiration parameterized on the basis of photosynthesis (Sellers et al., 1996) are included. A simplified TOPMODEL (Beven and Kirkby, 1979) calculates baseflow runoff, in addition to surface flows. The original TOPMODEL usually requires a detailed elevation data over the domain of interest, however, it is difficult to treat such data at a global scale. Therefore, in MATSIRO, the subgrid topography in a grid cell is roughly approximated as repetition of a slope with a uniform slope angle and with the distance between ridge and valley (Takata et al., 2003). There are five soil layers in which energy and water movements are treated with physical equations that consider freezing and condensation. The model is originally designed as a land surface sub-module of an atmospheric general circulation model (AGCM). The coupled model appropriately reproduced the
2.3.
Terrestrial nitrogen cycle model
The TNCM (Fig. 2) is developed to consider the mass balance of nitrogen in vegetation and organic soil of the ecosystem. It is based on the original model by Lin et al. (2000, 2001). The ecosystem was divided into an atmospheric and a terrestrial reservoir. The terrestrial nitrogen cycle consists of biological processes that depend on a variety of the environmental factors. The model contains five variables as for nitrogen: nitrogen in vegetation (Nveg, unit: ton N km2), organic N in detritus (Ndet, unit: ton N km2), organic nitrogen in humus (Nhum, unit: ton N km2), ammonium (Namm, unit: ton N km2), and nitrate (Nnit, unit: ton N km2). The nitrogen balance for each process was shown as below (He et al., 2009a): vNveg ¼ nuptake nf ð1 hvstÞ þ nfix vt
(1)
vNdet ¼ nf ndm ndh vt
(2)
vNhum ¼ ndh nhm þ fert hum þ lst vt
(3)
vNamm Namm ¼ ndm þ nhm þ nammd nuptake nnitrif vt Namm þ Nnit nvola þ fert amm
ð4Þ
vNnit Nnit ¼ nnitrif nnitrgas þ nnitd nuptake ndenitr vt Namm þ Nnit nleach þ fert nit
ð5Þ
nfix
hvst
Nveg nf
Ndet ndh
ndm
nvola
Nhum n Namm nnitrif hm lst fert_hum
fert_amm
Internal flux
Input
nnitd ndentr nnitrgas
2.2.
observed seasonal cycles of the energy and water balance at both regional (Hawaiian Islands: Sakamoto et al., 2004; Japan: Sakimura, 2007) and global scale (Hirabayashi et al., 2008). The model is also driven by atmospheric forcing in an off-line mode. Model application results in an off-line mode are described in previous studies such as Hirabayashi et al. (2005). This study used runoff output of off-line simulation of MATSIRO as an input of a nitrogen cycle model.
nammd
fertilizer consumption for each country were derived from FAO census database (FAOSTAT). The nitrate leaching from soil layers in farmland, grassland and natural conditions was calculated by using a terrestrial nitrogen cycle model (TNCM) (He et al., 2009a). A river routing model was used to transport nitrogen from natural and anthropogenic sources through river channels, as well as collect and route nitrogen to the river mouths. Subsequently, we will discuss the land surface model in Section 2.2, the terrestrial nitrogen cycle model in Section 2.3, the river routing model in Section 2.4, biological N fixation, atmospheric N deposition, and denitrification in Section 2.5, nitrate leaching in Section 2.6, and the design of the integrated simulation in Section 2.7.
Nnit nleach fert_nit Output
Fig. 2 e Flow chart for the terrestrial nitrogen cycle model (He et al., 2009a). The variables in this figure are described in the text, Section 2.3.
2576
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 7 3 e2 5 8 6
Where, nuptake is flux of nitrogen uptake by plant (ton N km2 day1), nf is flux of litter-fall from leaf, trunk, and root as in nitrogen (ton N km2 day1), nfix is flux of nitrogen fixation as in nitrogen (ton N km2 day1), ndm is flux of detritus mineralization as in nitrogen (ton N km2 day1), ndh is flux of detritus huminification as in nitrogen (ton N km2 day1), nhm is flux of humus mineralization as in nitrogen (ton N km2 day1), nammd is flux of nitrogen deposition as in ammonium (ton N km2 day1), Namm is potential nitrogen storage as in ammonium (ton N km2), Nnit is potential nitrogen storage as in nitrate (ton N km2), nnitrif is flux of nitrification (ton N km2 day1), nvola is flux of ammonia volatilization (ton N km2 day1), nnitrgas is flux of gaseous emissions during nitrification process (ton N km2 day1), nnitd is flux of nitrogen deposition as in nitrate (ton N km2 day1), ndenitr is flux of denitrification (ton N km2 day1), nleach is flux of nitrate leaching (ton N km2 day1), fert_hum is the amount of fertilizer in humus (ton N km2 day1), fert_amm is the amount of fertilizer in ammonium (ton N km2 day1), lst is the amount of fertilizer from livestock (ton N km2 day1), hvst is the ratio of harvested crops. For natural ecosystem, all of fertilizer amount i.e., fert_hum, fert_amm, and lst equal to zero (He et al., 2009a). The mathematical formulas describing all these processes and parameters in detail can be found in Lin et al. (2000, 2001) and He et al. (2009a). Most of the parameter values required in this model were either cited from reference papers or determined by model calibrations as described in Lin et al. (2000).
2.4.
River routing model
The aim of river routing model is to give directions for lateral water and pollutant movement by creating an idealized network of river channels. In this study, Total Runoff Integrating Pathways (TRIP) (Oki et al., 1999; Oki and Sud, 1998; Ngo-Duc et al., 2007) was employed to transport water and nitrogen flow through channels. The source of global digital elevation map (DEM) is ETOPO5 (Edwards, 1986). The basin delineation data used in this study have a spatial resolution of 0.5 by 0.5 . It was used to transport nitrogen from natural and anthropogenic sources through river channels, as well as collect and route nitrogen to the river mouths. Amounts of total nitrogen (TN) in direct runoff, lateral subsurface flow and percolation are estimated as the products of the volume of water and the average concentration. Transport or retention factors are taken into account through routing of water and nitrogen in the river flow via transmission losses (He et al., 2009a). Since dissolved inorganic nitrogen (DIN) is often the most abundant and bioavailable form of N and contributes significantly to coastal eutrophication (Veuger et al., 2004), we calculated DIN input into rivers from point sources by multiplying the amount of TN with an estimated fraction of TN that is DIN in sewage effluents (Dumont et al., 2005) as below: DIN ¼ TN$½0:485 þ TN $0:255=maxðTN Þ
(6)
where, TN is a country by country fraction of TN removed by wastewater treatment compiled by Bouwman et al. (2005a), 0.485 is an estimate of the fraction of TN that is DIN in sewage effluent (Seitzinger, 1995), and 0.255 is the maximum increase
in DIN to TN ratio that can be achieved by sewage treatment (Seitzinger, 1995). A detailed explanation can be found in Bouwman et al. (2005a) and Dumont et al. (2005). The time resolution of TRIP simulation was set as 1.0 day in this study.
2.5. Biological N fixation, atmospheric N deposition, and denitrification Biological N fixation of atmospheric N in natural ecosystems was estimated by using TNCM. It was assumed to be the sum of symbiotic and nonsymbiotic fixations, which can be modeled by the function in Lin et al. (2000), as shown in the equation below. nfix ¼ nfix
sy
þ nfix
nsy
(7)
where, nfix is flux of nitrogen fixation as in total nitrogen (ton N km2 day1), nfix_sy is nitrogen symbiotic fixation (ton N km2 day1), nfix_nsy is nitrogen nonsymbiotic fixation (ton N km2 day1). Nitrogen deposition includes dry and wet deposition of ammonia gas, nitrate, and nitrogen compounds from the atmosphere to soil by rain, snow, and dust. The deposition of ammonium and nitrate was modeled by using the method in Lin et al. (2000), where wet deposition was modeled as a linear function of precipitation (Hudson et al., 1994). Nitrogen discharging from land surface to rivers was assumed to infiltrate through soil where some fraction was removed by denitrification and organic matter accumulation. The surplus nitrogen flows to the river and then to the sea, were analyzed in conjunction with precipitation surplus. In this study, nitrogen was assumed to be denitrified and accumulated in the soil by a first-order reaction expressed in Shindo et al. (2003) and He et al. (2009a). C ¼ C0 expðkT $tR Þ
(8)
kT ¼ 2ðT20Þ=10 $k20
(9)
where, C0 indicates the original nitrogen concentration (mg/L), kT and k20 are coefficients of denitrification and accumulation at T and 20 (k20 ¼ 3.0), respectively, and tR is residence time in soil (day).
2.6.
Nitrate leaching
Many different models are used for the detailed simulation of the average nitrate leaching and denitrification process (Brisson et al., 2003; Johnsson et al., 1987; Shaffer et al., 1991). However, such models are too detailed for the 0.5 by 0.5 resolution and these models require data on environmental conditions (i.e., daily condition of root growth, phenology stage, crop yield, leaf area index, etc.) and agricultural management (i.e., irrigation option, drainage option, precise planting and cultivation date, fertilizer application), which are not available on the spatial scale of our model. In the present stage of this study, the TNCM was applied to estimate nitrate leaching from natural ecosystems such as grassland and forest with fertilizer application rate as zero. Furthermore, it was used to estimate nitrate leaching from croplands with the application of fertilizer amounts. The nitrate leaching is
2577
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 7 3 e2 5 8 6
strongly related to soil water content, soil texture, and NO3 concentration. For modeling the nitrate leaching flux, the below equation was employed (He et al., 2009a): Rt Nleach ¼ Nnit $ $103 qs
(10)
Where, Nleach is the flux of nitrate leaching (ton N km2 day1), Nnit is potential nitrogen storage as in nitrate (ton N km2) which was calculated by TNCM model, Rt is runoff (tonne km2 day1) which was calculated by MATSIRO model, and qs is soil water storage (mm) which was calculated by MATSIRO model (He et al., 2009a).
2.7.
model close to a steady-state so that negligible climate drift is experienced in the control run which follows. In this study, the spin-up procedure with a time step of 1 day was used for 100-year spin-up period.
Design of the integrated simulation
The integrated modeling framework for calculating global terrestrial nitrogen load and river’s nitrogen concentration is illustrated in Fig. 1. The model operates on a daily time step and at a spatial resolution of 0.5 by 0.5 over the world. After the input datum are read from files, the three-step modeling procedure is applied. First, water discharge, nitrogen balance, and nitrate leaching are calculated for each grid (0.5 by 0.5 ) by the MATSIRO and TNCM. Then the outputs from each grid (e.g. lateral water flows, nitrate flow) are summed with point pollution load (e.g. from industrial source, sewage plant). Finally, the routing procedure TRIP is applied to transport point and nonpoint pollution along rivers, taking transmission losses into account. Among these, the hydrological module is fundamental for all the modeling systems in this study. It was tested and validated in Hirabayashi et al. (2005, 2008). It reproduces well the observed seasonal cycles of the energy and water balance. In addition, before commencing a long-term nitrogen cycle simulation, it is usually necessary to allow the land surface to adjust to a steady-state. Some groups have nevertheless apparently used spin-up techniques successfully to initialize the ocean state for long climate studies (Manabe et al., 1991,, 1992; Stouffer et al., 1994; Thornton and Rosenbloom, 2005; Lin et al., 2000). The objective of spin-up is to bring the
3.
Data
Most of the available input database, which was used in global river nutrient export models, and the available model validation all choose 1995 as the base year. Therefore, the initial database constructed in this study is also based on the year 1995 as an example. All the input datasets have a spatial resolution of 0.5 by 0.5 . The following section will describe the detailed information about the database used in this study (Table 1).
3.1.
Hydrometeorological database
Air temperature, precipitation, short wave downward radiation with the spatial resolution of 1 by 1 and the temporal resolution of one day are from the second Global Soil Wetness Project (GSWP2; Dirmeyer et al., 2006) database. The soil temperature, soil water content, and runoff are calculated and validated by MATSIRO model (Hirabayashi et al., 2005) at the spatial resolution of 1 by 1 and the temporal resolution of one day. All the input data were divided into the same spatial resolution of 0.5 by 0.5 (allocation of the same value in four 0.5 grids within 1 grid) and the dataset in 1995 was used in this study since the final validation and model evaluation are all based on this year.
3.2.
Land cover map
The land cover map was generated from the database of GLCC (USGS). Its original spatial resolution is on a 30 s grid (Loveland et al., 2000). In this study, the spatial resolution of 0.5 has been applied for TNCM, LSM and RRM. Therefore, in each 0.5 grid, the area ratio of each land cover was calculated and
Table 1 e Overview of data used in this study. Data
Spatial Resolution
Land cover map Digital elevation map Basin delineation map River routing map Soil temperature, moisture Precipitation, Air temperature, Solar radiation Runoff Ammonia fertilizer Nitrous fertilizer Crop calendar Global population map (urban, rural, total) Manure N addition Sewage point sources DIN load
0.5 0.5 0.5 0.5 1 1
Temporal resolution
Data period
Source
e e e e Daily Daily
2000 e e e 1995 1995
http://edc2.usgs.gov/glcc Oki and Sud, (1998) Oki and Sud, (1998) Oki and Sud, (1998) Hirabayashi et al. (2005) Hirabayashi et al. (2005)
1 0.5 0.5 1 0.5
Daily Yearly Yearly Yearly Yearly
1995 1995 1995 1995 1995
Dirmeyer et al. (2006) FAOSTAT FAOSTAT Hanasaki et al. (2008) Bengtsson et al. (2006)
0.5 0.5 River basin
Yearly Yearly Yearly
1995 1995 1995
Bouwman et al. (2005a) Bouwman et al. (2005a) Dumont et al. (2005)
2578
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 7 3 e2 5 8 6
accumulated from the 30 s grid. The land cover data were reclassified among cropland, grassland, forest, water wetland, tundra, and other land. The area of each land cover type in each 0.5 by 0.5 cell was then calculated.
3.3.
Nitrogen load estimation
The spatial and temporal distributions of on-ground N fertilizer use from various crops and agricultural practices were quantified in this study (Fig. 3). The assessment of on-ground N fertilizer due to multiple land use activities can be complex and the traditional method is to only use the national census data without considering the crops’ spatial distribution, fertilization and harvest patterns. In this study, the monthly fertilizer application rates in the world were calculated by using global fertilizer statistics data and each crop’s agricultural manuals (FAOSTAT). Then, the land cover map showing the spatial distribution of areal ratio of each crop in each grid with a spatial resolution of 0.5 by 0.5 was generated for the whole global land surface. Finally, the spatial distribution map of fertilizer utilization with a temporal resolution of one month was generated. The crop distribution map was generated from the crop area. Then, the fertilizer amount for different land cover was calculated in each grid (Fig. 4). The nitrogen load from livestock was calculated for various livestock species based on literature values (Bouwman et al., 2005a, 2005b) by using the animal numbers and pollutant emission load per animal. Point sources of N are primarily associated with human excreta and industrial water use (wastewater drainage). As for the calculation of nitrogen load from domestic water use for populations provided with sewage plants, the database of diffusion rate of public sewerage and population distribution was utilized. The distribution of population provided with wastewater service was calculated by using the database of population without sewage plants and population with wastewater service. Generally, the population without sewage plants is distributed in rural area and the population with sewage plants is distributed in urban area. As for the
Agricultural manual (Crop)
FAOSTAT data Land cover (Fertilization data) map
Fertilizer application rate (Crop land)
FAOSTAT data (Livestock)
Pollutant emission basic unit
Nitrogen load (Livestock)
Nitrogen load (Agriculture) Fig. 3 e Flow chart for the calculation of nitrogen load from agricultural sources.
calculation of nitrogen load from industrial water use, this was calculated by using the database of the production of pollutant emission basic unit of industrial classification. The distribution of nitrogen load from industry can then be calculated by land cover data and nitrogen load. Furthermore, the distribution of sewage diffusion rate was calculated from total population distribution and population without sewage plants. The nitrogen load from industrial water use was calculated by the distribution of nitrogen load from industry and the distribution of sewage diffusion rate (Bouwman et al., 2005a; He et al., 2009a). DIN input into rivers from point sources was estimated by the method proposed by Dumont et al. (2005). Total nitrogen (TN), was used to calculated DIN from human excreta and industrial wastewater in sewage effluents by multiplying the amount of TN with an estimated fraction of TN and DIN, which was described in Equation (6).
3.4.
Validation data collection
Nitrogen has many chemical forms and compounds, which are very mobile and dynamic both in space and time. In addition, biogeochemical modeling of nitrogen cycle at the global or national scale with large grid cells usually only consider the vertical flows. In this study, the lateral flows are included by using the routing procedure of TRIP in the modeling system so that the chemical fluxes at the global scale can be validated using the data of measurements at the river outlet. For nitrogen cycle and routing model’s validation, we collected the observed dissolved inorganic nitrogen (DIN) concentration in major global rivers from literature records, which are published values and available for open access. Table 2 shows all selected 61 rivers for model validation (Alexander et al., 1996; EEA, 1998; Dumont et al., 2005; Seitzinger et al., 2005; Van Drecht et al., 2003).
4.
Result and discussion
4.1.
Model validation
Firstly, the land surface model was run to obtain a steadystate model. The global nitrogen cycle model restarted after an annual simulation, with the output used as the new initial conditions for the next year. The experiment began with a 100-year spin-up of the model, forced by the repeated initial annual nitrogen cycle and climatological data. After spin-up was completed, a steady-state of global nitrogen storage could be obtained. Then, the modeled annual dissolved inorganic nitrogen (DIN) load (ton N y1) at each river was compared with the existing data of observed DIN load. As for the hydrological model results, the long-term terrestrial water fluxes were estimated well using the land surface model driven by long-term atmospheric forcing data (Hirabayashi et al., 2005). High correlations between predicted and observed annual runoff were obtained at many basins globally, but correlations are low in dry areas and in cooltemperate zones. Moreover, annual snow covered area in North America and northern Europe and annual summer soil moisture in Mongolia were successfully replicated by the model (Hirabayashi et al., 2005).
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 7 3 e2 5 8 6
2579
Fig. 4 e Map of global annual nitrogen fertilization use in 1995.
The predicted annual DIN yield (ton N km2 y1), calculated by dividing the annual DIN export by the river basin area, was also compared with the measured yield. Fig. 5 shows a scatter plot of calculated and observed (Dumont et al., 2005) DIN load and yield for the selected 61 rivers in 1995. The figure shows a linear relationship between the logarithms of observed and modeled annual values of DIN load and yield. The regression lines between observed and predicted DIN load and yield are very close to a 1:1 ratio. The model efficiencies (R2, the coefficient of determination) were 0.88 and 0.81 for DIN load and yield, respectively. Consequently, we can conclude that the model reproduced the DIN load in selected rivers with reasonable accuracy. Fig. 6 shows calculated and observed DIN load and yield for the selected 61 rivers in this study. The x-axis is the river ID from Table 2; rivers are ordered in size from the largest (Amazon) to the smallest (Pee Dee) basin. From the top figure in Fig. 6, we can see that the DIN load in most of the rivers is proportional to the size of the river basin. However, some river basins with larger areas have smaller DIN loads, such as the Murray River (ID ¼ 10) in Australia and Rio Grande (ID ¼ 13) in North America, which have relatively smaller river discharge. In Table 2, we can find some river basins in which the DIN loads are smaller than 1.5 103 ton/year. As shown in the bottom figure in Fig. 6, the DIN yield in these two rivers is very small; this is because the DIN yield is calculated by dividing the DIN export by the river basin area. Among the 61 rivers, the river with the largest DIN yield is the Rhine River (ID ¼ 24). From Fig. 6, we can see that the model used in this study reproduced DIN yield and load well for most of the selected 61 rivers, which have different spatial locations and basin sizes. Fig. 7 illustrates the correlation of river discharge with DIN load and DIN yield for selected rivers in the world. The logarithmelogarithm relationship in Fig. 7 shows that river
discharge has stronger correlation with DIN load than DIN yield. It means increased river discharge generally exerts a positive effect on DIN load in rivers at a long term (one year) time scale. At higher annual flows, rivers deliver more nitrogen from upstream to downstream by reducing residence times. However, as for DIN yield, its correlation with river discharge will be more complicated since DIN yield is also affected by the catchment characteristics, such as area, elevation, time and distance of transport for nitrogen in a drainage network. Because river discharge is correlated with vegetation, the relationship between DIN yield and river discharge corresponds to a relationship between DIN yield and vegetation type (Lewis et al., 1999).
4.2.
Annual nitrate leaching from terrestrial ecosystems
Fig. 8 presents a map of annual nitrate leaching from the terrestrial ecosystem in 1995. The range of nitrate leaching was very large across the world. Similar to the results from Dumont et al. (2005), higher nitrate leaching was predicted for tropical humid-climate river basins (i.e., in Indonesia, West Africa, the Amazon, and the Zaire River basin), densely populated basins with high GDP (i.e., Rhine and Thames river basins), and basins with extensive agricultural activities (i.e., Yangtze and Ganges river basins). Other areas with high values were predicted in New Zealand and Japan. The lowest nitrate was predicted for basins with low population density such as most basins at high latitudes and also for arid regions (i.e., Nile River basin and Tamanrasett River basin). In this paper, nitrate was calculated by the terrestrial nitrogen cycle model, for which the main input data included simulated runoff, fertilizer application, and precipitation. Therefore, the predicted nitrate was greatly influenced by these input data and the resulting spatial pattern was very similar to those of
2580
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 7 3 e2 5 8 6
Table 2 e Selected global rivers for model validation. ID
Name
Continent
Area (104 km2)
Average Q(km3/yr)
Pop. density (person/km2)
Agriculture land (%)
DIN load (103ton)
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61
Amazon Mississippi Ob Parana Yenisei Lena Yangtze Jiang Amur Indus Murray Yellow Yukon Rio Grande Columbia Kolyma Don Pearl Pechora Churchill Neva Yana Rufiji Wisla Rhine Elbe Brazos Balsas Colorado Odra Kuskowin Anabar Nemanus Penzhina Daugava Mezen Seine Tejo Susquehanna Bug Usumacinta Copper Kuban Paraiba do Sul Sacramento Narva Sakarya Appalachicola Saint John Stikine Kamchatka Trinity Glama Weser Hudson Altamaha Potomac Nushagak Tornionjoki Klamath Dalalven Pee Dee
S. A. N.A. Asia S. A. Asia Asia Asia Asia Asia Australia Asia N.A. N.A. N.A. Asia Europe Asia Europe N.A. Europe Asia Africa Europe Europe Europe N.A. N.A. N.A. Europe N.A. Asia Europe Asia Europe Europe Europe Europe N.A. Europe N.A. N.A. Europe S. A. N.A. Europe Asia N.A. N.A. N.A. Asia N.A. Europe Europe North N.A. N.A. N.A. Europe N.A. Europe N.A.
583.30 319.10 301.50 265.40 256.90 243.30 178.80 174.80 113.90 102.80 89.05 85.27 80.19 72.93 66.32 42.16 40.71 31.31 30.24 28.35 22.42 18.61 18.00 16.45 14.80 12.46 12.26 12.08 11.94 11.54 9.86 9.66 8.55 8.32 7.54 7.32 7.31 7.19 6.90 6.79 6.70 6.36 6.28 5.87 5.80 5.68 5.47 5.29 5.12 5.04 4.74 4.73 4.55 4.31 4.15 3.83 3.53 3.45 3.21 2.98 2.76
5025.75 538.37 289.73 595.59 423.42 284.92 436.57 261.14 38.37 20.46 18.46 42.4 6.53 198.21 67.57 33.42 142.01 78.54 46.09 71.34 17.21 62.53 22.64 58.47 16.61 5.45 31.83 3.88 11.42 14.66 7.3 16.76 18.11 16.17 18.65 9.75 8.67 25.44 5.64 100.25 31.75 19.74 29.62 26.16 11.26 4.42 27.44 32.67 22.4 25.92 9.38 18.46 7.92 19.27 13.75 11.36 7.3 7.54 12.78 8.39 8.57
4.39 22.47 10.14 27.80 3.06 0.56 244.07 36.37 169.23 3.51 158.88 0.13 17.45 9.19 0.09 48.60 207.09 1.79 0.21 28.31 0.04 23.92 135.49 300.35 166.61 27.57 230.44 16.04 121.44 0.07 0.00 46.26 0.09 30.57 1.84 210.31 101.42 53.55 74.05 36.12 0.08 66.84 68.34 16.88 19.66 105.59 79.56 8.46 0.04 0.72 150.21 28.54 196.98 172.40 38.59 91.50 0.05 1.60 7.83 11.44 72.75
8.62 74.69 37.23 59.47 13.65 0.00 69.34 26.78 37.02 53.89 81.50 0.00 73.85 17.40 0.00 98.70 69.11 0.00 4.75 2.33 0.00 54.02 50.97 45.99 53.68 85.11 37.64 81.81 63.49 0.00 0.00 37.55 0.00 20.72 0.00 81.63 57.97 23.14 84.09 30.35 0.00 61.20 58.98 15.51 30.58 66.67 25.88 7.81 0.00 0.00 81.95 0.00 28.99 5.22 5.50 29.39 0.00 6.47 6.73 0.00 27.34
1006.19 815.62 295.47 116.51 110.72 51.34 585.57 139.32 155.93 1.13 107.31 26.09 0.48 54.04 11.94 8.05 213.04 20.26 2.87 21.01 5.81 51.34 66.92 361.97 117.72 6.98 8.96 2.92 46.54 15.80 1.15 13.37 2.18 12.58 1.87 99.90 8.98 35.44 1.95 38.17 21.79 21.06 11.62 2.24 4.25 8.81 12.85 3.16 11.92 4.47 4.37 9.07 54.78 16.41 4.69 15.16 3.72 0.31 2.28 1.69 6.06
Note: Rivers are sequenced by basin area. (S.A: South America; N.A: North America).
2581
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 7 3 e2 5 8 6
10000000
7 1:1 DIN load
6
5
Model
100000 10000 1000 100
4
1
5
10 15 20 25 30 35 40 45 50 55 61 River Number
10000.00
3
1000.00
2 2
3
4
5
6
7
Logarithm of observed DIN load 4
Logarithm of estimated DIN yield
Observation
1000000
y = 0.23 + 0.83x R= 0.90
1:1
3.5
DIN yield
Logarithm of estimated DIN load
y = -0.03 + 0.98x R= 0.94
100.00 10.00 1.00 0.10
1
3
5
10 15 20 25 30 35 40 45 50 55 61 River Number
Fig. 6 e Comparison of calculated and observed DIN load (ton N yL1) and DIN yield (ton N kmL2 yL1) for selected rivers.
2.5 2 1.5 1 0.5 0 0
0.5
1
1. 5
2
2. 5
3
3. 5
4
Logarithm of observed DIN yield Fig. 5 e Scatter plot of comparison of calculated and observed DIN yield (ton N kmL2 yL1) and DIN load (ton N yL1) for selected rivers. Dark diagonal line represents the 1:1 line.
runoff and fertilizer. Harrison et al. (2005) noted similar findings for the global distribution of dissolved organic nitrogen yield. As discussed by Seitzinger and Kroeze (1998) and shown in Fig. 8, Asia exports the most DIN to its coasts. This is due to its large surface area, high population, and large cultivated land area. The World Health Organization (WHO) has recommended healthy drinking water quality standards of 10 mg/L or less for nitrateenitrogen, whereas most rivers in populated regions, according to the Global Environment Monitoring System (GEMS) database, have values about seven times this number at their mouths. The levels of dissolved nitrogen in these rivers are no longer due solely to natural processes such as weathering and soil organics, but also due to a substantial contribution by human activities, particularly in Asia (Subramanian, 2004; Jacks and Sharma, 1983).
4.3. Spatial distribution of global nitrateenitrogen concentration To identify where the nitrogen pollution is most serious, Fig. 9 shows the spatial distribution of the nitrateenitrogen
concentration in global rivers in 1995. The nitrateenitrogen concentration is generally low for areas with low temperature and little precipitation. By contrast, the nitrateenitrogen concentration becomes especially high in the eastern United States, the Rhine River, the Thames River, the lower portion of the Amazon River, the Yellow River and Yangtze River, northeast China, the east coast of the North China Plain, and some parts of the Republic of Korea and Japan. In comparing the maps of nitrogen fertilizer, nitrate leaching, and nitrate concentration, it is evident that, in specific areas, high nitrogen fertilizer use does not necessarily correspond with equally high nitrate leaching and nitrate concentration. This is apparent in the upstream region of the Yangtze River, upstream region of the Yellow River (China), upstream region of the Mississippi River (U.S.A), Murray River (Australia), Nelson River (Canada), and upstream region of the Danube River (Germany, Austria, Slovakia), midstream region of the Amur River (Russia, China). Conversely, other areas have relatively low nitrogen fertilizer use but high nitrate leaching and concentration. This is seen in places such as downstream region of the Amazon River, and the midstream region of Congo River. The processes of the nitrogen cycle are complex and nitrate leaching from soil layers is controlled by nitrogen load input, hydrometeorological conditions, and management practices. However, from a global perspective, the most serious nitrogen pollution has occurred in distinct areas exhibiting extensive agriculture and low precipitation surpluses (Shindo et al., 2003). This is visible from Fig. 7 in which large values of nitrate concentration (10e50 mg N/L) were found in the Northern plains of China, Northern India, and North-western portions of the U.S.A. In the groundwater of Shandong Province, China, for example, the average measured nitrate concentration was found to be 38.5 mg N L1 and maximum concentration could exceed 100 mg N L1 (Shindo et al., 2006).
2582
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 7 3 e2 5 8 6
10000000
DIN load
1000000 100000 10000 Observation Model
1000 100 1
10
100
1000
10000
River discharge 10000
DIN yield
1000 100 10 Observation Model
1 0.1 1
10
100
1000
10000
River discharge Fig. 7 e Correlation of river discharge (km3 yL1) with DIN load (ton N yL1) and DIN yield (ton N kmL2 yL1) for selected rivers.
4.4.
Uncertainty analysis
From the model results and analysis in the above sections, we can see that a degree of discrepancy remains between simulated and observed values of both nitrateenitrogen flux and nitrateenitrogen yield. The most likely causes originate from the uncertainties and variables inherent to the dataset, and model limitation. For example, as the main input to nitrogen loading on a global scale, nitrogen fertilizer usage has a large impact on the final calculation of nitrate leaching. The
calculated nitrogen fertilizer value was compared with that of FAO census data. However, since census datasets are inevitably prone to a level of error, this data also contains some uncertainties surrounding their reported values. Nitrogen produced from livestock was calculated from the number of animals and excretion rate per head (Bouwman et al., 1997; Shindo et al., 2003). All the animals were considered to be full grown. For nitrogen deposition data, the simple empirical relationship between nitrogen deposition and precipitation was employed. However, long-term transport of nitrogen compounds, particularly NOx, is important and thus for more precise estimation a transport model is needed (Shindo et al., 2003). For denitrification and organic matter accumulation in the soil, a simple reaction model considering temperature and residence time was applied. Nitrogen removal due to instream nitrogen retention is affected by complex conditions, which also creates uncertainty. In addition, we collected the available observed DIN concentration data in major rivers from literature record for model validation. However, the observed data itself has uncertainty related to discharge measurement, sample collection (location and frequency), sample storage, and laboratory analysis. Moreover, data processing can contribute uncertainty to measured data because of missing data, assumptions made to estimate missing values, and mistakes in data management and reporting (Harmel et al., 2006). Therefore, to reduce the effect from the uncertainty of measured data, the frequency of water quality sampling for the collected data in this study commonly ranges from quarterly to monthly, with differences occurring by network, constituent, and time period. The dataset collected in this study was restricted to include only long term (>4 years) annual averages with at least 85% of the measurements taken between 1990 and 1997 (Dumont et al., 2005). The collected datasets also include world river basins with a broad range of area, land cover, climate, and topography (Table 2).
Fig. 8 e Simulated annual NO3eN leaching from the terrestrial ecosystem in 1995.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 7 3 e2 5 8 6
2583
Fig. 9 e Map of simulated global NO3eN concentration in 1995.
To assess the model uncertainty, five scenarios were considered in which the nitrogen load was increased by 10% for nitrogen leaching from the agricultural soil layer, nitrogen load from manure, nitrogen load from nitrogen fixation in agricultural lands, nitrogen load from nitrogen deposition in agricultural land, and nitrogen from sewage, respectively. The upper part of Fig. 10 presents the resulting ratios of simulated annual nitrate export under these five scenarios. The average nitrateenitrogen export in 1995 is shown in the bottom part of Fig. 10 in comparison to the absolute nitrateenitrogen export in different rivers. The figure shows that the contribution of nitrogen leaching from the agricultural soil layer is the most sensitive source for most of the selected 61 rivers. The second largest source is the nitrogen from sewage, followed by the nitrogen contribution from manure. The contributions of nitrogen deposition and nitrogen fixation in agricultural lands are nearly identical. These results indicate that the nitrateenitrogen export in most rivers will be sensitive to the amount of nitrogen compounds leaching from agricultural soil layers and urban sewage. The nitrogen leaching from the agricultural soil layer is substantially affected by agricultural activities such as nitrogenous fertilizer application. Accordingly, the uncertainty from fertilizer application will have the largest impact on the estimation of nitrateenitrogen export in rivers.
4.5.
Discussion
The models in this study were constructed to examine dissolved nitrogen forms in particular. Through this emphasis, hydrological pressures, such as the positively correlated relationship between N export and runoff rate, contributing to the diffusion of anthropogenic and natural nitrogen sources could be incorporated (Seitzinger et al., 2005). The transport of nitrogen was only considered from the land surface into rivers
and then out to sea. Transport of nitrogen in groundwater and interaction between river water and groundwater were not investigated in this study. Furthermore, the observed data for DIN concentration were collected at the outlet of each river. However, estimates of the absolute values of nitrogen concentration in river water involved uncertainties because sufficient data was not available for the ratio of total nitrogen export to nitrate export. This ratio would vary spatially according to temperature, river discharge, catchment characteristics, etc. (Shindo et al., 2003). Consideration of these factors would likely result in improved estimates. Despite the limitations described above, we can still obtain good insight and estimates from the nitrogen-loading simulation at the global scale, which can improve our understanding of global spatial patterns and the magnitudes of nitrogen export from global river basins. In addition, the nitrogen model used in this study was specifically developed to obtain nitrogen export from river basins at the global scale. Similar to other global biogeochemical models, it differs in its degree of spatial resolution and mechanism formulation from models specifically developed for use in individual watersheds such as the Riverstrahler model (Billen et al., 1999; Seitzinger et al., 2005). Moreover, the current global nitrogen model is a process-based model considering the major process of nitrogen fixation, nitrification, denitrification, immobilization, mineralization, leaching, and nitrogen taken by vegetation. The fertilizer application rate for major crops and nitrogen load from major livestocks in the world were calculated to be inputs of the model. As results, the nitrogen leaching from soil layers was calculated through the model. This is different with most of the other nitrogen models in which the pollutant emission basic unit is applied to estimate the nitrogen leaching from soil layers (Sferratore et al., 2005). Moreover, the models in this study can provide a detailed information of calculated nitrateenitrogen leaching from the terrestrial ecosystem and
2584
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 7 3 e2 5 8 6
Increased Nitrate Nitrogen Export (%)
40.00
Increase 10% n_leach
35.00
n_manure nfx_agr
30.00
ndp_agr 25.00
n_sewage
20.00
15.00
10.00
5.00
0.00
1
5
10 11 20 25 30 35 40 45 50 55 60 River Number
1
5
10 11 20 25 30 35 40 45 50 55 60 River Number
Nitrate Nitrogen Export (ton N/yr)
1.E+07
1.E+06
1.E+05
1.E+04
1.E+03
Acknowledgements
1.E+02
Fig. 10 e Annual NO3eN export and ratios of simulated NO3eN export under five scenarios in which the nitrogen load was increased by 10% for nitrogen leaching from the soil layer (n_leach), nitrogen load from manure (n_manure), nitrogen load from nitrogen fixation in agricultural land (nfx_agr), nitrogen load from nitrogen deposition in agricultural land (ndp_agr), and nitrogen load from sewage (n_sewage).
nitrate concentration in rivers not only at each river outlet, but also at each grid with a spatial resolution of 0.5 by 0.5 . Therefore, the nitrogen pollution condition in each grid of these river basins can be assessed accordingly. The calculated nitrogen loadings from agricultural lands, livestock, industrial plants, and domestic water use can provide useful databases for global-scale nitrogen pollution evaluation.
5.
datasets of fertilizer use, maps of land cover and population, social census data, and literature records. The nitrate leaching from soil layers in the terrestrial ecosystem was calculated using a global-scale terrestrial nitrogen cycle model. The model validation results indicate that discrepancy remains between simulated and observed DIN yield and load in some places. Uncertainties associated with the dataset and model limitations are the primary reasons for these discrepancies. However, this study aimed at understanding the spatial and temporal distribution of nitrogen load and it provided an initial overview of an integrated framework with which to estimate nitrogen load and nitrogen pollution in rivers at a global scale. As for future directions of related work, it is important to improve our understanding the mechanisms and time scales involved in the terrestrial response to nitrogen deposition, taking into account long-term nitrogen fertilization. The relationship between anthropogenic river water removal and DIN export is also required in future study. In addition, the same model architecture used in this study can be applied to a region (or several ones) with good data availability over several years to perform an in-depth test of the model. The database built in this study will provide a useful foundation for further model development and improvement, considering both anthropogenic and natural scenarios. As improved temporal and spatial resolution of validation datasets and the development of hydrological models that route materials downstream through river networks become available, it will be possible for us to further examine seasonal patterns and finer spatial resolution of DIN export.
Conclusion
This paper has described integrated biogeochemical modeling of nitrogen load and its export to global rivers. The amount of nitrogen loading from various sources were calculated using
This study was supported by the Kyoto University Global COE program “Sustainability/Survivability Science for a Resilient Society Adaptable to Extreme Weather Conditions” and the JSPS Grants-in-Aid for Scientific Research. This work was also partially supported by JSPS KAKENHI, Grants-in-Aid for Scientific Research (S)(19106008). We wish to thank Dr. Lex Bouwman for providing us with global data on N input including sewage effluents, net N input via fertilizer and manure, etc., and Mr. Leif Harum for giving us good comments about this article. The authors are grateful for their supports.
references
Alexander, R.B., Slack, J.R., Ludtke, A.S., Fitzgerald, K.K., Schertz, T.L., 1996. Data from Selected U.S. Geological Survey National Stream Water Quality Monitoring Networks (WQN) [CD-ROM]. U.S. Geol. Surv., Denver, Colo. Baker, A., 2003. Land use and water quality. Hydrological Processes 17, 2499e2501. Bengtsson, M., Shen, Y.J., Oki, T., 2006. A SRES-based gridded global population dataset for 1990e2100. Population and Environment 28, 113e131. Beven, K.J., Kirkby, M.J., 1979. A physically based, variable contributing area model of basin hydrology. Hydrological Sciences Bulletin 24, 43e69.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 7 3 e2 5 8 6
Brisson, N., Gary, C., Justes, E., Mary, B., Roche, R., Ripoche, D., Zimmer, D., Sierra, J., Bertuzzi, P., Burger, P., Bussie`re, F., Cabidoche, Y.M., Cellier, P., Debaeke, P., Gaudille`re, J.P., Maraux, F., Seguin, F.B., Sinoquet, H., 2003. An overview of the crop model STICS. European Journal of Agronomy 18, 309e332. Billen, G., Garnier, J., Ficht, A., Cun, C., 1999. Nitrogen transfers through the Seine drainage network: a budget based on the application of the Riverstrahler model. Hydrobiologia 410, 139e150. Bouwman, A.F., Lee, D.S., Asman, W.A.H., Dentener, F.J., Van Der Hoek, K.W., Oliver, J.G.J., 1997. A global high-resolution emission inventory for ammonia. Global Biogeochemical Cycles 11, 561e587. Bouwman, A.F., Van Drecht, G., Knoop, J.M., Beusen, A.H.W., Meinardi, C.R., 2005a. Exploring changes in river nitrogen export to the world’s oceans. Global Biogeochemical Cycles 19 (GB1002). doi:10.1029/2004GB002314. Bouwman, A.F., Van Drecht, G., Van der Hoek, K.W., 2005b. Nitrogen surface balances in intensive agricultural production systems in different world regions for the period 1970e2030. Pedosphere 15 (2), 137e155. Davis, J.R., Koop, K., 2006. Eutrophication in Australian rivers, reservoirs and estuaries e a southern hemisphere perspective on the science and its implications. Hydrobiologia 559, 23e76. Dickinson, R.E., Shaikh, M., Bryant, R., Graumlich, L., 1998. Interactive canopies for a climate model. Journal of Climate 11, 2823e2836. Ding, C., Ohmori, H., Takamura, H., 2006. Relationship between water quality and land use in the Mamagawa river, Tokyo Metropolitan Area. Journal of Japanese Association of Hydrological Sciences 36 (4), 219e233. Dirmeyer, P.A., Gao, X.A., Zhao, M., Guo, Z.C., Oki, T., Hanasaki, N., 2006. GSWP-2: multimodel analysis and implications for our perception of the land surface. Bulletin of the American Meteorological Society 87, 1381e1397. Dumont, E., Harrison, J.A., Kroeze, C., Bakker, E.J., Seitzinger, S.P., 2005. Global distribution and sources of dissolved organic nitrogen export to the coastal zone: results from a spatial explicit, global model. Global Biochemical Cycle 19 (GB4S402), 1e13. Edwards, M.H. 1986. Digital image processing of local and global bathymetric data. M.S. thesis, Department of Earth and Planetary Sciences, Washington University, St. Louis, MO, 106pp. Europe’s Environment: Statistical Compendium for the Second Assessment, Report, 1998. European Environment Agency, Copenhagen. FAOSTAT. http://faostat.fao.org/site/291/default.aspx. Galloway, J.N., Schlesinger, W.H., Levy II, H., Michaels, A., Schnoor, J.L., 1995. Nitrogen fixation: anthropogenic enhancement e environmental response. Global Biogeochemical Cycles 9, 235e252. Galloway, J.N., 1998. The global nitrogen cycle: changes and consequences. Environmental Pollution 102, 15e26. Galloway, J.N., 2000. Nitrogen mobilization in Asia. Nutrient Cycling in Agroecosystems 57, 1e12. Galloway, J.N., Cowling, E.B., 2002. Nitrogen and the world. Ambio 31, 64e71. Galloway, J.N., Dentener, F.J., Capone, D.G., Boyer, E.W., Howarth, R.W., Seitzinger, S.P., Asner, G.P., Cleveland, C.C., Green, P.A., Holland, E.A., Karl, D.M., Michaels, A.F., Porter, J.H., Townsend, A.R., Vorosmarty, C.J., 2004. Nitrogen cycle: past, present and future. Biogeochemistry 70, 153e226. Harmel, R.D., Cooper, R.J., Slade, R.M., Haney, R.L., Arnold, J.G., 2006. Cumulative uncertainty in measured streamflow and water quality data for small watersheds. Transactions of the American Society of Agricultural and Biological Engineers (ASAE) 49 (3), 689e701. Harrison, J.A., Caraco, N., Seitzinger, S.P., 2005. Global patterns and sources of dissolved organic matter export to the coastal
2585
zone: results from a spatial explicit, global model. Global Biogeochemical Cycles 19 (GB4S04), 1e16. Hanasaki, N., Kanae, S., Oki, T., Masuda, K., Motoya, K., Shirakawa, N., Shen, Y., Tanaka, K., 2008. An integrated model for the assessment of global water resourcesdPart 2: applications and assessment. Hydrology and Earth System Sciences 12, 1027e1037. Hantschel, R.E., Beese, F., 1997. Site-oriented ecosystem management: precondition to reducing the contamination of waters and the atmosphere. In: Rosen, D., Tel-Or, E., Hadar, Y., Chen, Y. (Eds.), Modern Agriculture and the Environment. Kluwer Academic Publishers, Dordrecht, pp. 135e145. He, B., Oki, T., Kanae, S., Mouri, G., Kodama, K., Komori, D., Seto, S., 2009a. Integrated biogeochemical modelling of nitrogen load from anthropogenic and natural sources in Japan. Ecological Modelling 220, 2325e2334. He, B., Oki, K., Wang, Y., Oki, T., 2009b. Using remotely sensed imagery to estimate potential annual pollutant loads in river basins. Water Science and Technology 60 (8), 2009e2015. doi: 10.2166/wst.2009.596. Hirabayashi, Y., Kanae, S., Oki, T., 2005. A 100-year (1901e2000) global retrospective estimation of terrestrial water cycle. Journal of Geophysical Research 110, D19101. Hirabayashi, Y., Kanae, S., Emori, S., Oki, T., Kimoto, M., 2008. Global projections of changing risks of floods and droughts in a changing climate. Hydrological Sciences Journal 53 (4), 754e772. Howarth, R.W., Sharpley, A., Walker, D., 2002. Sources of nutrient pollution to coastal waters in the United States: implications for achieving coastal water quality goals. Estuaries 25 (4B), 656e676. Hudson, D., Hite, D., Haab, T., 2005. Public perception of agricultural pollution and Gulf of Mexico hypoxia. Coastal Management 33 (1), 25e36. Hudson, R.J.M., Gherini, S.A., Goldstein, R.A., 1994. Modelling the global carbon cycle: nitrogen fertilization of the terrestrial biosphere and the ‘missing’ CO2 sink. Global Biogeochemical Cycles 8 (3), 307e333. Jacks, G., Sharma, V.P., 1983. Nitrogen circulation and nitrate in groundwater in an agricultural catchment in southern India. Environment Geology 5, 61e64. Johnsson, H., Bergstro¨m, L., Jansson, P.E., Paustian, K., 1987. Simulated nitrogen dynamics and losses in a layered agricultural soil. Agriculture Ecosystems and Environment 18 (4), 333e356. Lewis, J.R.W.M., Melack, J.M., McDowell, W.H., McClain, M., Richey, J.E., 1999. Nitrogen yields from undisturbed watersheds in the America. Biogeochemistry 46, 149e162. Lin, B.L., Sakoda, A., Shibasaki, R., Goto, N., Suzuki, M., 2000. Modelling a global biogeochemical nitrogen cycle model in terrestrial ecosystems. Ecological Modelling 135-1, 89e110. Lin, B.L., Sakoda, A., Shibasaki, R., Suzuki, M., 2001. A modelling approach to global nitrate leaching caused by anthropogenic fertilisation. Water Research 35 (8), 1961e1968. Loveland, T.R., Reed, B.C., Brown, J.F., Ohlen, D.O., Zhu, J., Yang, L., Merchant, J.W., 2000. Development of a global land cover characteristics database and IGBP DISCover from 1-km AVHRR data. International Journal of Remote Sensing 21 (6/7), 1,303e1,330. Manabe, S., Stouffer, R.J., Spelman, M.J., Bryan, K., 1991. Transient responses of a coupled ocean-atmosphere model to gradual changes of atmospheric CO2. Part I: annual mean response. Journal of Climate 4, 785e818. Manabe, S., Spelman, M.J., Stouffer, R.J., 1992. Transient responses of a coupled ocean-atmosphere model to gradual changes of atmospheric CO2. Part II: seasonal response. Journal of Climate 5, 105e126.
2586
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 7 3 e2 5 8 6
Ngo-Duc, T., Oki, T., Kanae, S., 2007. A variable stream flow velocity method for global river routing model: model description and preliminary results. Hydrology and Earth System Sciences Discussions 4, 4389e4414. Oenema, O., Boers, P.C.M., Willems, W.J., 1998. Leaching of nitrate from agriculture to groundwater: the effect of policies and measures in the Netherlands. Environmental Pollution 102, 471e478. Oki, T., Nishimura, T., Dirmeyer, P., 1999. Assessment of annual runoff from land surface models using Total Runoff Integrating Pathways (TRIP). Journal of the Meteorological Society of Japan 77 (1B), 235e255. Oki, T., Sud, Y.C., 1998. Design of total runoff integrating Pathways (TRIP) e A global river channel network. Earth Interactions 2. Rijtema, P.E., Kroes, J.G., 1991. Some results of nitrogen simulations with the model ANIMO. Fertilizer Research 27, 189e198. Sakamoto, T.T., Emori, S., Nishimura, T., Hasumi, H., Suzuki, T., Komoto, M., 2004. Far-reaching effects of the Hawaiian Islands in the CCSR/NIES/FRCGC high-resolution climate model. Geophysical Research Letters 31, L17212. Sakimura, T., 2007. Development of an approach to predict flood risks based on water and energy balance simulation over Japan. Master Thesis. The University of Tokyo. Schepers, J.S., Varvel, G.E., Watts, D.G., 1995. Nitrogen and water management strategies to reduce nitrate leaching under irrigated maize. Journal of Contaminant Hydrology 20, 227e239. Seitzinger, S.P., 1995. Data Collection Program in Support of the Harborwide Eutrophication Model for New York e New Jersey Harbor Estuary Program, Report. U.S. Environ. Prot. Agency, New York. Seitzinger, S.P., Kroeze, C., 1998. Global distribution of nitrous oxide production and N inputs in freshwater and coastal marine ecosystems. Global Biochemical Cycles 12, 93e113. Seitzinger, S.P., Harrison, J.A., Dumont, E., Beusen, A.H.W., Bouwman, A.F., 2005. Sources and delivery of carbon, nitrogen, and phosphorus to the coastal zone: an overview of global Nutrient Export from Watersheds (NEWS) models and their application. Global Biogeochemical Cycles 19 (GB4S01), 1e11. Sellers, P.J., Randall, D.A., Collatz, G.J., Berry, J.A., Field, C.B., Dazlich, D.A., Zhang, C., Collelo, G.D., Bounoua, L., 1996. A revised land surface parameterization (SiB2) for atmospheric GCMs. Part I: model formulation. Journal of Climate 9, 676e705. Sferratore, A., Billen, G., Garnier, J., Thery, S., 2005. Modeling nutrient budget in the Seine watershed: application of the Riverstrachler model using data from local to global scale
resolution. Global Biogeochemical Cycles 19, 1e14. doi:10. 1029/2005GB002496. Shaffer, M.J., Halvorson, A.D., Pierce, F.J., 1991. Nitrate leaching and economic analysis package (NLEAP): model description and application. In: Follet, R.F., Keeney, D.R., Cruse, R.M. (Eds.), Managing Nitrogen for Groundwater Quality and Farm Profitability. Soil Science Society of America, Madison, Wis, pp. 285e322. Shindo, J., Okamoto, K., Kawashima, H., 2003. A model-based estimation of nitrogen flow in the food production e supply system and its environmental effects in East Asia. Ecological Modeling 169, 197e212. Shindo, J., Okamoto, K., Kawashima, H., 2006. Prediction of the environmental effects of excess nitrogen caused by increasing food demand with rapid economic growth in eastern Asian countries, 1961e2020. Ecological Modelling 193 (3e4), 703e720. Stouffer, R.J., Manabe, S., Vinnikov, K.Y., 1994. Model assessment of the role of natural variability in recent global warming. Nature 367, 634e636. Subramanian, V., 2004. Water quality in South Asia. Asian Journal of Water, Environment and Pollution 1, 41e54. Takata, K., 2000. Sensitivity Study on Snowmelt Discharge of Lena Using a Land Surface Model, “MATSIRO”. Activity Report of GAME-Siberia, 1999. No. 21, Nagoya Univ.. GAME Publication, Nagoya, Japan, pp. 73e76. Takata, K., 2001. Estimation of Summer Heat and Water Fluxes at Tiksi Using a One-dimensional Land Surface Model. Activity Report of GAME-Siberia, 2000. No. 26, Nagoya Univ.. GAME Publication, Nagoya, Japan, pp. 207e208. Takata, K., Emori, S., Watanabe, T., 2003. Development of the minimal advanced treatments of surface interaction and runoff. Global and Planetary Change 38, 209e222. Thornton, P.E., Rosenbloom, N.A., 2005. Ecosystem model spinup: estimating steady state conditions in a coupled terrestrial carbon and nitrogen cycle model. Ecological Modelling 189, 25e48. United Nations, 1996. Country Population Statistics and Projections 1950e2050, Report. Food and Agric. of the U.N., Rome. USGS. http://edc2.usgs.gov/glcc/globdoc2_0.php. Van Drecht, G., Bouwman, A.F., Knoop, J.M., Beusen, A.H.W., Meinardi, C.R., 2003. Global modeling of the fate of nitrogen from point and nonpoint sources in soils, groundwater, and surface water. Global Biogeochemical Cycles 17 (4), 1e20. doi: 10.1029/2003GB002060. Veuger, B., Middelburg, J.J., Boschker, H.T.S., Nieuwenhuize, J., van Rijswijk, P., Rochelle-Newall, E.J., Navarro, N., 2004. Microbial uptake of dissolved organic and inorganic nitrogen in Randers Fjord. Estuarine Coastal and Shelf Science 61, 507e515.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 8 7 e2 5 9 4
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Degradation of PAHs by high frequency ultrasound Ioannis D. Manariotis a,*, Hrissi K. Karapanagioti b, Constantinos V. Chrysikopoulos a a b
Department of Civil Engineering, Environmental Engineering Laboratory, University of Patras, Patras 26500, Greece Department of Chemistry, University of Patras, Patras 26500, Greece
article info
abstract
Article history:
Polycyclic aromatic hydrocarbons (PAHs) are persistent organic compounds, which have
Received 25 October 2010
been reported in the literature to efficiently degrade at low (e.g. 20 kHz) and moderate (e.g.
Received in revised form
506 kHz) ultrasound frequencies. The present study focuses on degradation of naphtha-
31 January 2011
lene, phenanthrene, and pyrene by ultrasound at three different relatively high frequencies
Accepted 10 February 2011
(i.e. 582, 862, and 1142 kHz). The experimental results indicate that for all three frequencies
Available online 17 February 2011
and power inputs 133 W phenanthrene degrades to concentrations lower than our experimental detection limit (<1 mg/L). Phenanthrene degrades significantly faster at
Keywords:
582 kHz than at 862 and 1142 kHz. For all three frequencies, the degradation rates per unit
Polycyclic aromatic hydrocarbons
mass are similar for naphthalene and phenanthrene and lower for pyrene. Furthermore,
Naphthalene
naphthalene degradation requires less energy than phenanthrene, which requires less
Phenanthrene
energy than pyrene under the same conditions. No hexane-extractable metabolites were
Pyrene
identified in the solutions. ª 2011 Elsevier Ltd. All rights reserved.
Ultrasonic treatment High frequency
1.
Introduction
Polycyclic aromatic hydrocarbons (PAHs) are widespread in the environment. PAHs often accumulate in various environmental systems including coastal estuaries and marine sediments (Daskalakis and O’Connor, 1995; Karapanagioti et al., 2009) as well as in drinking water supplies (Yang and Silverman, 1988; Kim Oahn et al., 2002). PAHs generally occur as complex mixtures and not as single compounds. Although their concentration in environmental systems is very low due to their low solubility (wmg/L), they are of great importance because they are listed as priority pollutants (Callahan et al., 1979; EEC, 1980). The European drinking water standard for the sum of four PAHs is 0.1 mg/L (EC, 1988). PAHs can originate from various sources by thermal combustion processes (i.e. coal burning, cooking and heating oils), vehicular emissions (i.e. automobiles), and biomass
burning (Simoneit, 1984; Stein et al., 2006). Combustion-derived PAHs present in the atmosphere can enter the water supply by gaseous exchange in the airewater interface, dry deposition of particulate matter, wet deposition (rainfall), and urban runoff (Chen et al., 2004). The PAH loading pathways to coastal waters in highly urbanized areas are mainly via storm water runoff, tributary inflow, wastewater treatment plant effluent, atmospheric deposition, and dredged material disposal (Chrysikopoulos et al., 1992; Oros et al., 2007). Concentrations of PAHs are generally higher in samples from urban streams and in combined overflow sewers than in effluents of wastewater treatment plants (Bergqvist et al., 2006; Phillips and Chalmers, 2009; Roswell et al., 2010). PAHs are removed in wastewater treatment plants by absorption onto the generated sludge (Jiries et al., 2000). During the past several years, ultrasound has been effectively employed as an emerging advanced oxidation process (AOP) for a wide variety of micropollutants (Virkutyte and
* Corresponding author. E-mail address:
[email protected] (I.D. Manariotis). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.02.009
2588
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 8 7 e2 5 9 4
Rokhina, 2010; Adewuyi, 2001, 2005; Kotronarou et al., 1991; Hoffmann et al., 1996; Taylor et al., 1999; Lifka et al., 2003; Matouq et al., 2008; David, 2009). Biological treatment can be inhibited by toxic or persistent pollutants, it is very sensitive to several environmental factors, and it is slow, therefore uneconomical for high concentrations of micropollutants (Adewuyi, 2005). Mahamuni and Adewuyi (2010) reported that the cost of ultrasound treatment of various contaminants is higher compared to other AOPs, and combination of ultrasound with different AOPs is economically more attractive than the use of ultrasound alone. Acoustic cavitation in aqueous solutions results in the formation of micro bubbles producing local areas of high energy. Within these localized high energy regions, chemical transformations occur including the cleavage of chemical bonds, oxidation, pyrolysis and/or combustion of organic compounds (Suslick, 1990). Temperature and pressure can reach levels of 3000e5000 K and 500e10,000 atm, respectively (Riesz et al., 1985; Suslick, 1990). In the region of cavitation, three reaction zones exist: (a) the gas phase, (b) the gaseliquid interface, and (c) the liquid surrounding the cavitation bubble (Riesz et al., 1985). Various studies suggest that the sonodegradation of PAHs occurs via: (i) oxidation by HO$ in the solution phase, (ii) pyrolytic processes within the bubble, and (iii) combination of oxidation by HO$ and pyrolytic processes (Wheat and Tumeo, 1997; Taylor et al., 1999; Laughrey et al., 2001; David and Riguier, 2002; Little et al., 2002; Psillakis et al., 2004). David (2009) concluded that because of their properties, PAHs are expected to be mainly localized in the center and/or in the surrounding shell of the bubble inhibiting the production of hydroxyl radicals and hence the oxidation pathway, and that they are mainly pyrolyzed within the bubble. Most studies so far are focused on low ultrasound frequencies (20e80 kHz) and volumes treated ranging from 50 to 200 mL. The power and frequency of ultrasound are reported to affect the degradation rate of PAHs. The mean bubble diameter increases with increasing acoustic power and decreases with increasing ultrasound frequency. At 1056 kHz the mean diameter of active bubbles have been observed to reach a limiting value of about 4.5 mm, and for acoustic power greater than 6 W the bubble size is unaffected (Brothie et al., 2009). The acoustic power enhances the degradation rate of PAHs (Psillakis et al., 2004). The increase of input power intensity from 50 to 600 W results in higher degradation efficiency of pyrene after 60 min sonication at 20 kHz and 20 C (Park et al., 2000). The effect of frequency is associated with the dynamics of bubble formation. Generally, higher ultrasound frequencies may increase the number of free radicals because there are more cavitational events that consequently lead to increased degradation of PAHs (Psillakis et al., 2004). David (2009) reported that high frequency (506 kHz) is more efficient than low frequencies. The degradation of PAHs is affected positively or negatively by the presence of matrix compounds, and the predominant mechanism for the decomposition of PAHs is known to involve oxygen derived chemicals (Laughrey et al., 2001). Furthermore, chemical alteration of PAHs can occur in aqueous solutions exposed to high intensity ultrasound (Wheat and Tumeo, 1997; Park et al., 2000; Little et al., 2002). The sonochemical degradation of hydrophobic pollutants can also be significantly reduced by the presence of additional
dissolved species, which may inhibit PAH access to the cavitation sites (Taylor et al., 1999; Laughrey et al., 2001). The objective of this work is to evaluate the effectiveness of ultrasound waves at relatively high frequencies (i.e. 582, 862 and 1142 kHz) to degrade three PAH molecules of varying sizes (naphthalene, phenanthrene and pyrene) in pure aqueous solutions. Furthermore, the effect of input power on PAH degradation, the degradation kinetics as well as the presence of hexane-extractable metabolites are examined. To our knowledge, PAHs sonodegradation results at 582, 862 and 1142 kHz have never been reported in the literature. In addition, the present study employs a system able to treat higher volume solutions (500 and 1000 mL) than those reported in the literature (50e200 mL). Finally, a correlation between the energy required to degrade a compound and its molecular weight is determined.
2.
Materials and methods
2.1.
Test system
An ultrasonic system (Meinhardt Ultraschalltechnik, Leipzig, Germany) composed of a 75-mm diameter titanium transducer operating at 582, 862, and 1142 kHz, a function generator, and an amplifier was employed in this study (see Fig. 1). The transducer was mounted at the bottom of a cylindrical 2-L glass laboratory reactor with double walls to allow water circulation for cooling. PAH aquatic solutions with volume 500 or 1000 mL were poured into the ultrasonic reactor. A sample solution was used to measure and record the initial PAH concentration. During each experiment, 4 mL samples were taken at 5, 10, 15, 30, 60, 90, and 120 min. Samples were collected with a glass pipette and stored in glass vials. All samples were analyzed at the end of each experiment and the PAH concentrations were compared to the initial concentration. The temperature of the reactor liquid was measured with a digital thermometer equipped with a thermistor (Oakton Temp 5 Acorn Series, Eutech Instruments Ltd., Singapore). The electric power of the system could be adjusted to any level up to 200 W.
Fig. 1 e Experimental setup arrangement.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 8 7 e2 5 9 4
2.2.
Analytical methods
Naphthalene, phenanthrene, and pyrene with two, three, and four ring chemical structure, respectively, were used in this study as model PAH compounds. Naphthalene (99.8% Dr. Ehrenstorfer GmbH, Germany), phenanthrene (>97% FLUKA Chemika, Switzerland), and pyrene (98% Aldrich Chemicals, Milwaukee, WI) were dissolved in methanol (>99.8% Merck) to yield a 12.5, 1.0, and 0.1 g/L stock solutions, respectively, which were kept in the dark at 4 C. Aquatic solutions of 3000, 100, and 10 mg/L were prepared for naphthalene, phenanthrene, and pyrene. The pH of each PAH solution was 6.5. Aqueous PAH concentrations were measured by a cuvette mode fluorescence spectrophotometer (Cary Eclipse, Varian Australia PTY LTD, Australia) using a quartz cuvette and excitation/emission wavelengths of 278/324 for naphthalene, 249/347 for phenanthrene, and 272/373 for pyrene. The conditions of the various experiments conducted in this study are listed in Table 1. Selected experiments were performed in duplicates or triplicates. All experiments were performed at 25 C. A temperature control system was used and the highest temperature increase was less than 5 C for all experiments. To determine whether PAH hexane-extractable metabolites were produced during sonication, selected naphthalene and pyrene solutions (500 mL) were analyzed before and after sonication. The solutions were extracted with Pestiscan grade hexane (10 mL) in separatory funnels that were shaken for 10 min (modified from USEPA, 1999). After the hexane phase was separated from the aqueous phase, it was condensed to about 300 mL using a gentle stream of nitrogen gas and then, was cleaned through microcolumns containing glass wool, copper, and anhydrous sodium sulfate. Subsequently, a 2 mL sample was injected in a gas chromatograph with mass spectrometer (Shimadzu QP2000 GCeMS, with a 30 m Quadrex column). The detection limit of this analysis is about 1 mg/L. Fluorescence spectra of the three PAHs at various sonication times with ultrasound frequency of 582 kHz and an electric power of 133 W were examined for possible formation of
Table 1 e Experimental conditions. Run number 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
fa (kHz)
Power input (W)
Solution volume (mL)
PAH
582 582 582 582 582 582 582 862 862 862 862 1142 1142 1142 1142 1142 1142
67 100 133 133 133 133 166 100 133 133 133 100 133 133 133 133 166
500 500 500 500 1000 1000 500 500 500 500 500 500 500 500 1000 1000 1000
Phenanthrene Phenanthrene Naphthalene Phenanthrene Pyrene Phenanthrene Phenanthrene Phenanthrene Naphthalene Phenanthrene Pyrene Phenanthrene Naphthalene Phenanthrene Pyrene Phenanthrene Phenanthrene
2589
aromatic byproducts. Excitation wavelengths in the UV range are generally used for the excitation of aromatic compounds (Little et al., 2002).
2.3.
Data analysis
Experimental data plotted as logarithmic normalized concentrations vs time (not shown here) suggested that the degradation process follows a first-order kinetic equation: C ¼ k1 t ln C0
(1)
where C [mg/L] is the PAH aqueous concentration measured at a given time, C0 [mg/L] is the initial PAH aqueous concentration, k1 [1/min] is the first-order reaction constant, and t [min] is time. The half-life of PAH degradation reaction (t1/2) [min] was calculated by the following equation: 0:693 k1
t1=2 ¼
(2)
The bubble radii as a function of ultrasound frequency employed were calculated from the following empirical expression (Brothie et al., 2009): log10 ½rb ¼ log10 fa þ 3:5
(3)
where rb [mm] is the bubble radius, and fa [kHz] is the acoustic frequency. The bubble radius for a given frequency does not change for acoustic power higher than 8 W (Brothie et al., 2009). In this study, the acoustic power in all experiments was higher than 8 W. Therefore, the bubble radius is assumed to be similar for all experiments performed at the same frequency regardless of the acoustic power used. The calorimetric power of the system, Pcal [W], was determined by recording the temperature fluctuation of the liquid during the initial 30 s of each experiment and using the following equation (Kimura et al., 1996): Pcal ¼
DT cp Mw Dt
(4)
where DT [K] is the temperature difference, cp ¼ 4.2 [J/g$K] is the heat capacity of the water, and Mw [g] is the water mass in the reactor. The acoustic pressure, Pa, was calculated using the following equation (Mason and Lorimer, 2002): Pa ¼
pffiffiffiffiffiffiffiffiffiffiffiffi 2rcIA
(5)
where r ¼ 1000 (kg/m3) is the density of water, c ¼ 1500 (m/s) is the speed of sound in water, IA ¼ Pcal/Ap (W/m2) is the intensity, which is defined as the amount of energy flowing per unit area, Ap [cm2] is the plate surface area, and Pa is measured in [N/m2]. The plate diameter was 75 mm and thus, the plate surface area was Ap ¼ 44 104 m2. The void fraction was calculated based on scarce experimental results found in the literature (Burdin et al., 1999) and the extrapolation method used by David (2009). The void fraction at fa ¼ 20 kHz and Pa ¼ 3.2 bar was reported as 104 and at fa ¼ 308 kHz and Pa ¼ 0.2 bar was reported as 3 104. Through linear extrapolations among frequencies and acoustic pressures, the void fraction for each experimental
2590
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 8 7 e2 5 9 4
condition was predicted. The bubble number per liter is the void fraction divided by the bubble volume calculated based on the bubble radius (David, 2009). Note that the void fraction value of 5 103 used by Servant et al. (2000) at fa ¼ 477 kHz and high acoustic pressure is in line with the results of this study, and supports the assumption of a linear relationship between acoustic frequencies and void fraction. Although the bubbles were not uniform in size and shape (David, 2009), our simplified calculations provided a means to compare the different experimental systems employed in this study (i.e. various frequencies and power inputs).
3.
Results and discussion
3.1. Effect of input power and frequency on PAHs concentration Fig. 2 presents the experimental data of the normalized phenanthrene concentration with time at fa ¼ 582 kHz and at varying power inputs. Determination of the optimum input power (resulting in faster degradation rate with the least input power) is necessary in order to operate the instrument in a cost effective mode. For power inputs P 133 W, phenanthrene degraded to concentrations lower than the detection limit (<1 mg/L) within 100 min. Similar trends for varying power inputs were observed for the other two frequencies tested (862 and 1142 kHz). In Fig. 2, a delay of 10e15 min for the phenanthrene removal from the solution is observed for P ¼ 100 W. Note that at P ¼ 167 W there was a slight increase in concentration during the initial period. This degradation delay is probably attributed to the formation of a dense cloud around the probe at high powers, which blocks the energy transmitted from the probe to the fluid (Thompson and Doraiswamy, 1999). It was reported that doubling the power from 75 to 150 W, phenanthrene degradation in a mixture of PAHs at fa ¼ 24 kHz increased from 60 to 83% after 60 min of sonication and at fa ¼ 80 kHz increased from 58 to 90% after 30 min of sonication
(Psillakis et al., 2004). For naphthalene, increasing P from 75 to 150 W resulted in slightly higher degradation of naphthalene; (83e84%) at fa ¼ 24 kHz and (92e99%) at fa ¼ 80 kHz. Park et al. (2000) reported that at fa ¼ 20 kHz and increasing P from 50 to 600 W, the degree of pyrene degradation doubled after 60 min of sonication. In the present study at fa ¼ 582 kHz, by progressively increasing the input power from 67e100 to 133e167 W, phenanthrene degradation increased from 3e54 to 84e84%, respectively, after 60 min of sonication. The above results show that increasing P improved degradation in a nonlinear fashion. Worthy to note is that the reaction rate reaches a maximum value and a subsequent power increase does not alter reaction rates. Fig. 3 presents the experimental data of the normalized PAH concentrations with time at one power input (133 W) and at three acoustic frequencies (582, 862, and 1142 kHz). Phenanthrene and pyrene degraded to concentrations lower than 10% of C0 in around 100 min for all three frequencies employed. Naphthalene degraded to concentrations lower than 15% of C0 in 100 min at 582 kHz, and lower than 25% of C0 at 862 and 1142 kHz. Note that at fa ¼ 582 kHz, the degradation of all three PAHs examined was significantly faster than at the other two frequencies (862 and 1142 kHz). Also, the degradation behavior of all three PAHs examined at fa ¼ 862 and 1142 kHz was similar. The properties of the bubbles generated at each of the three acoustic frequencies employed in this study are presented in Table 2. Bubble size decreased with increasing frequency. The void fraction and number of bubbles per volume considerably increased with increasing P. Whereas for the same P, the number of bubbles per volume increased by a factor of 4 by increasing fa from 582 to 862 kHz and another factor of 4 by increasing fa from 862 to 1142 kHz. No correlation was observed between the properties of the bubbles and the sonodegradation results. The results of the present study and those reported by David (2009) suggest that possibly at fa w500 kHz bubbles with optimum radius are formed that lead to better degradation. Although higher frequencies were expected to yield better activity because they result in more void volume (cavitation bubbles created per unit of time), at the same time higher frequencies result in smaller bubbles that might not be as active as bubbles at fa w 500 kHz. David (2009) reported that fa from 20 kHz to 506 kHz resulted in higher rate constants for five PAHs examined. It was reported that the main mechanisms of PAH degradation was their pyrolysis in the heart of the bubble, and oxidation by hydroxyl radicals was a minor pathway. David and Riguier (2002) also studied the degradation of anthracene in aqueous suspensions adsorbed on silica at low (20 kHz) and high (506 kHz) acoustic frequencies and observed that the degradation rate was higher at the high fa. The oxidation pathway by hydroxyl radicals was ruled out. As it is demonstrated in the present study, the degradation rates did not significantly increase for increasing fa in the range 582e1142 kHz. On the contrary, the degradation rates were lower for increasing fa in the range 582e1142 kHz.
3.2. Fig. 2 e Phenanthrene degradation as a function of time at four different input power levels (67, 100, 133, and 167 W) with fa [ 582 kHz.
Degradation kinetics
Table 3 presents the t1/2 of the three PAHs examined in this study for the different experiments performed. In all cases,
2591
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 8 7 e2 5 9 4
a
Table 2 e Bubble properties for different experimental conditions.
b
c
Fig. 3 e Degradation of (a) naphthalene, (b) phenanthrene, and (c) pyrene as a function of time at three different acoustic frequencies (582, 862, and 1142 kHz).
Pa (bar)
fa ¼ 582 kHz 67 100 5.1 133 167
13 21 37 48
0.92 1.2 1.6 1.8
2.6 3.4 4.5 5.1
fa ¼ 862 kHz 133 3.5
37
1.6
6.6 103
3.7 1010
fa ¼ 1142 kHz 133 2.6
37
1.6
8.8 103
1.2 1011
Void fraction ()
No of bubbles (number/L)
103 103 103 103
4.7 6.1 8.1 9.2
109 109 109 109
a Calculated from Equation (3).
treated by other investigators was 200 mL (Psillakis et al., 2004). Once aqueous solution volume was increased to 1 L, longer times were necessary for phenanthrene to degrade (t1/2 ¼ 45 min). Decreasing the volume by half also decreased the half-life time by half. For this reason, the 500 mL volume was chosen for further study. The values of the first-rate constants for the three PAHs tested are given in Table 4, together with the corresponding values obtained from literature. No obvious correlation between compound size, hydrophobicity, and degradation rates was observed. Such correlations have been reported in the literature for monocyclic aromatic compounds (Nanzai et al., 2008). However, there exists a positive correlation between the treated PAH compound size and the required energy, E ¼ P time [W h]. Naphthalene degradation required almost 1.5 orders of magnitude less energy than phenanthrene, which required one order of magnitude less energy than pyrene. The energy required for the degradation of phenanthrene at different input power levels is presented in Fig. 4. Note that less energy was required when operating at high P. The degradation of naphthalene with a molecular weight (MW) of 128 required 0.22 Wh/mg, almost 1.5 orders of magnitude less energy than phenanthrene (MW ¼ 178) that required 5.1 Wh/mg, which is one order of magnitude less energy than the 57 Wh/mg required by pyrene (MW ¼ 202) (see Fig. 5).
Table 3 e Estimated t1/2 (min). P(W)
faster degradation was observed at fa ¼ 582 kHz and P 133W. For initial concentrations of the three PAHs at 10% of their aqueous solubility, the faster degradation was observed at fa ¼ 582 kHz and P 133W for phenanthrene and pyrene. Degradation rates per unit mass were similar for naphthalene and phenanthrene and lower for pyrene. For phenanthrene, the lower t1/2 ¼ 23 2 min was observed for 500 mL at fa ¼ 582 kHz and P 133W. Similar results have been observed by other researchers at lower frequencies (e.g. at 20 kHz, Taylor et al., 1999; 24 and 80 kHz, Psillakis et al., 2004; 20 and 506 kHz, David, 2009). However, these results were obtained by treating lower solution volumes compared to the volumes treated in the present study. The highest solution volume
rba (mm)
Pcal (W)
P (W)
Sample volume (mL)
Naphthalene (C0 ¼ 3000 mg/L) 133 500 Phenanthrene (C0 ¼ 100 mg/L) 67 500 100 500 133 500 133 1000 166 500 166 1000 Pyrene (C0 ¼ 10 mg/L) 133 500 a Mean standard experiments.
fa (kHz) 862
582
1142
38a 3.3
57
57
580 51 23 2.0 45 21 e
e 47 32 4.0 e e e
e 65 31 1.4 e e 87
24 4.2
32 1.2
33 1.4
deviation
of
triplicate
or
duplicate
2592
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 8 7 e2 5 9 4
Table 4 e Estimated sonodegradation first-order rate (k1) values of various PAH. P (W)
Sonicated volume (mL)
C0 (mg/L)
fa (kHz)
k1 (1/min)
Reference
150 150 600 150e180 30 30 30 30 150 150 200 200 133 133 133 600 150e180 150 150 200 200 133 133 133 600 150e180 200 200 133 133 133
200 200 50e100 50e70 150 150 150 150 200 200 150 150 500 500 500 50e100 50e70 200 200 150 150 500 500 500 50e100 50e70 150 150 500 500 500
150 450 18 18e89 110 110 12 12 150 450 227 227 3000 3000 3000 18 18e89 150 450 97 97 100 100 100 20 20e100 118 118 10 10 10
80 80 20 20 20 506 20 506 80 80 20 506 582 862 1142 20 20 80 80 20 506 582 862 1142 20 20 20 506 582 862 1142
0.0846 0.0678 0.9 0.36 1.6 0.018 0.040 0.003 0.093 0.006 0.008 0.001 0.028 0.001 0.1217 0.0989 0.12 0.01 0.27 0.01 0.018 0.002 0.012 0.012 0.354 0.018 0.66 0.12 0.0836 0.0422 0.047 0.003 0.17 0.01 0.030 0.003 0.022 0.002 0.023 0.0001 0.36 0.18 0.624 0.054 0.029 0.002 0.067 0.003 0.030 0.005 0.021 0.0008 0.021 0.0009
Psillakis et al. (2004) Psillakis et al. (2004) Taylor et al. (1999) Laughrey et al. (2001) David (2009) David (2009) David (2009) David (2009) Psillakis et al. (2004) Psillakis et al. (2004) David (2009) David (2009) This study This study This study Taylor et al. (1999) Laughrey et al. (2001) Psillakis et al. (2004) Psillakis et al. (2004) David (2009) David (2009) This study This study This study Taylor et al. (1999) Laughrey et al. (2001) David (2009) David (2009) This study This study This study
PAH Acenaphthylene Acenaphthylene Anthracene Anthracene Anthracene Anthracene Benzo(k)fluoranthene Benzo(k)fluoranthene Naphthalene Naphthalene Naphthalene Naphthalene Naphthalene Naphthalene Naphthalene Phenanthrene Phenanthrene Phenanthrene Phenanthrene Phenanthrene Phenanthrene Phenanthrene Phenanthrene Phenanthrene Pyrene Pyrene Pyrene Pyrene Pyrene Pyrene Pyrene
3.3.
Intermediate products
C (µg/L)
100
50 67 W 100 W 133 W 167 W
0 0
200
400
600
Energy (Wh) Fig. 4 e Phenanthrene degradation as a function of energy required at different input power levels (67, 100, 133, and 167 W) with fa [ 582 kHz.
Required energy (Wh/µg)
100 Naphthalene and pyrene solutions were analyzed for hexaneextractable metabolites using liquideliquid extraction and GCeMS. Naphthalene solutions at two concentrations (3 and
Pyrene 4 2
10 Phenanthrene 4 2
1 4 2
Naphthalene
140
160
180
200
Molecular Weight Fig. 5 e Energy required for the three PAHs examined in this study as a function of the molecular weight at an input power level of 133 W with fa [ 582 kHz (experimental data are presented with solid circles and the correlation curve is presented with a line; Required energy [ 0.0318 MW L 4.784, R2 [ 0.985).
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 8 7 e2 5 9 4
0.3 mg/L) were analyzed before and after sonication. For both cases considered, naphthalene was detected in the solution before and after sonication. Also, a 10 mg/L pyrene solution was analyzed before and after sonication. Pyrene was only detected in the solution before sonication but was not detected in the solution after sonication. No additional peaks from other ions present in the hexane extract were observed in the chromatograms obtained by using GCeMS in the SCAN mode. Also, no additional peaks were observed even when the GCeMS was operated in the SIM mode, searching for specific ion masses of possible byproducts such as 1,2 dihydrohynaphthalene, salicylaldehyde, salicylate, gentisate, catechol, etc. Little et al. (2002) identified phenanthrene byproducts in solutions sonicated at fa ¼ 30 kHz. A possible explanation could be that the sonication mechanism is different at higher frequencies or the byproducts were present at concentrations lower than the detection limit (w1 mg/L). High frequencies result in shorter bubble life times and production of free radicals. Also, at high frequencies the amount of free radicals which are able to escape from the cavitation site to the bulk volume may increase and facilitate the bulk reaction (Thompson and Doraiswamy, 1999). Wheat and Tumeo (1997) found phenanthrene-diol as an intermediate product after sonication of the phenanthrene. Taylor et al. (1999) did not observe oxidized compounds after the sonication of phenanthrene and anthracene. Finally, Park et al. (2000)
800
Intensity (a.u.)
a
0 min 10 15 30 60 90 120
600 400
4.
Conclusions
High frequency ultrasounds (i.e. 582, 862, and 1142 kHz) were tested to degrade three PAH molecules of varying sizes (naphthalene, phenanthrene and pyrene) in pure aqueous solutions. Phenanthrene and pyrene degraded to concentrations lower than 10% of C0 in about 100 min for all three frequencies employed. Naphthalene degraded to concentrations lower than 15% of C0 in 100 min at 582 kHz, and lower than 25% of C0 at 862 and 1142 kHz. Degradation rates per unit mass were similar for naphthalene and phenanthrene and lower for pyrene. A positive correlation was observed between the treated PAH compound size and the required energy.
200
b Intensity (a.u.)
reported that after 1 h ultrasonic irradiation of pyrene, the major products were tetrahydro-2,5-dimethyl-furan, tetrahydro-2-methyl-2-furanol, 2,2-dimethyl-3-propyl-oxirane, 3,4dihydro-6-methyl-2H-pyran, and 1,2-benzedicarboxylic acid. The fluorescence spectra at various sonication times for naphthalene, phenanthrene, and pyrene at fa ¼ 582 kHz and P ¼ 133 W are presented in Fig. 6. A decrease in the fluorescence intensity with increasing time was observed without any significant change in the spectral shape. This observation indicated that there were no fluorescent interferences by other intermediate compounds with the excitation and emission wavelengths used. Similar results have been reported in the literature for the sonication of anthracene, phenanthrene and pyrene at fa ¼ 20 kHz (Taylor et al., 1999; Laughrey et al., 2001).
references
0 600 400 200 0
c Intensity (a.u.)
2593
600 400 200 0 300
350
400
450
Wavelength (nm) Fig. 6 e Fluorescence spectra at various times for: (a) naphthalene, (b) phenanthrene, and (c) pyrene.
500
Adewuyi, Y.G., 2001. Sonochemistry: environmental and engineering applications. Industrial and Engineering Chemistry Research 40 (22), 4681e4715. Adewuyi, Y.G., 2005. Sonochemistry in environmental remediation. 1. Combinative and hybrid sonophotochemical oxidation processes for the treatment of pollutants in water. Environmental Science and Technology 39 (10), 3409e3420. Bergqvist, P.A., Augulyte, L., Jurjoniene, V., 2006. PAH and PCB removal efficiencies in Umea (Sweden) and Siauliai (Lithuania) municipal wastewater treatment plants. Water, Air, and Soil Pollution 175 (1), 291e303. Brothie, A., Grieser, F., Ashokkumar, M., 2009. Effect of power and frequency on bubble-size distributions in acoustic cavitation. Physical Review Letters 102 (8), 084302e084304. Burdin, F., Tsochatzidis, N.A., Guiraud, P., Wilhelm, A.M., Delmas, H., 1999. Characterisation of the acoustic cavitation cloud by two laser techniques. Ultrasonics Sonochemistry 6, 43e51. Callahan, M.A., Slimak, M.W., Gabel, N.W., May, I.P., Fowler, C.F., Freed, J.R., Jennings, P., Durfee, R.L., Whitmore, F.C., Maestri, B., Mabey, W.R., Holt, B.R., Gould, C., 1979. EPA-440/ 4-79-029b. Water Related Environmental Fate of 129 Priority Pollutants, vol. II. U.S. Environmental Protection Agency, Washington, DC. Chen, B., Xuan, X., Zhu, L., Wang, J., Gao, Y., Yang, K., Shen, X., Lou, B., 2004. Distributions of polycyclic aromatic hydrocarbons in surface waters, sediments and soils of Hangzhou City, China. Water Research 38 (16), 3558e3568. Chrysikopoulos, C.V., Hildemann, L.M., Roberts, P.V., 1992. Modeling the emission and dispersion of volatile organics
2594
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 8 7 e2 5 9 4
from surface aeration wastewater treatment facilities. Water Research 26 (8), 1045e1052. Daskalakis, K.D., O’Connor, T.P., 1995. Distribution of chemical concentrations in US coastal and estuarine environment. Marine Environmental Research 40 (4), 381e398. David, B., 2009. Sonochemical degradation of PAH in aqueous solution. Part I: monocomponent PAH solution. Ultrasonics Sonochemistry 16, 260e265. David, B., Riguier, D., 2002. Ultasonication of aqueous and micellar suspensions of anthracene fixed on silica. Ultrasonics Sonochemistry 9, 45e52. EC, 1988. Council Directive 98/83/EC of 3 November 1998. On the quality of water intended for human consumption. Official Journal of the European Communities L 300 (5.12.1998), 32e54. EEC, 1980. Council Directive 80/778/EEC. Relating to the quality of water intended for human consumption. Official Journal of the European Communities L 229 (30.8.1980), 11. Hoffmann, M.R., Hua, I., Hochemer, R., 1996. Application of ultrasonic irradiation for the degradation of chemical contaminants in water. Ultrasonics Sonochemistry 3 (3), S163eS172. Jiries, A., Hussain, H., Lintelmann, J., 2000. Determination of polycyclic aromatic hydrocarbons in wastewater, sediments, sludge and plants in Karak province, Jordan. Water, Air, and Soil Pollution 121, 217e228. Karapanagioti, H.K., Siavalas, G., Kalaitzidis, S., Papatheodorou, G., Christanis, K., 2009. Distribution of polycyclic aromatic hydrocarbons (PAHs) in the Gulf of Aliveri, Central Greece. 9th Symposium on Oceanography & Fisheries, 2009-Proceedings, vol. I, pp. 251e255. Kim Oahn, N.T., Nghiem, L.H., Phyu, Y.L., 2002. Emission of polycyclic aromatic hydrocarbons, toxicity, and mutagenicity from domestic cooking using sawdust briquettes, wood, and kerosene. Environmental Science and Technology 36 (5), 833e839. Kimura, T., Sakamoto, T., Leveque, J.M., Sohmiya, H., Fujita, M., Ikeda, S., Ando, T., 1996. Standardization of ultrasonic power for sonochemical reaction. Ultrasonics Sonochemistry 3, S157eS161. Kotronarou, A., Mills, G., Hoffmann, M.R., 1991. Ultrasonic irradiation of p-nitrophenol in aqueous solution. Journal of Physical Chemistry 95 (9), 3630e3638. Laughrey, Z., Bear, E., Jones, R., Tarr, M.A., 2001. Aqueous sonolytic decomposition of polycyclic aromatic hydrocarbons in the presence of additional dissolved species. Ultrasonics Sonochemistry 8, 353e357. Lifka, J., Ondruschka, B., Hofmann, J., 2003. The use of ultrasound for the degradation of pollutants in water: aquasonolysis e a review. Engineering in Life Sciences 3 (6), 253e262. Little, C., Hepher, M.J., El-Sharif, M., 2002. The sono-degradation of phenanthrene in an aqueous environment. Ultrasonics 40, 667e674. Mahamuni, N.N., Adewuyi, Y.G., 2010. Advanced oxidation processes (AOPs) involving ultrasound for waste water treatment: a review with emphasis on cost estimation. Ultrasonics Sonochemistry 17 (6), 990e1003. Mason, T.J., Lorimer, J.P., 2002. Applied Sonochemistry: Uses of Power Ultrasound in Chemistry and Processing. WileyeVCH Verlag GmbH & Co., KGaA, Germany. Matouq, M.A., Al-Anber, Z.A., Tagawa, T., Aljbour, S., Al-Shannag, M., 2008. Degradation of dissolved diazinon pesticide in water using the high frequency of ultrasound wave. Ultrasonics Sonochemistry 15, 869e874.
Nanzai, B., Okitsu, K., Takenaka, N., Bandow, H., Maeda, Y., 2008. Sonochemical degradation of various monocyclic aromatic compounds: relation between hydrophobicities of organic compounds and the decomposition rates. Ultrasonics Sonochemistry 15 (4), 478e483. Oros, D.R., Ross, J.R.M., Spies, R.B., Mumley, T., 2007. Polycyclic aromatic hydrocarbon (PAH) contamination in San Francisco Bay: a 10-year retrospective of monitoring in an urbanized estuary. Environmental Research 105, 101e118. Park, J.K., Hong, S.W., Chang, W.S., 2000. Degradation of polycyclic aromatic hydrocarbons by ultrasonic irradiation. Environmental Technology 21 (11), 1317e1323. Phillips, P., Chalmers, A., 2009. Wastewater effluent, combined sewer overflows, and other sources of organic compounds to lake Champlain. Journal of the American Water Resources Association 45 (1), 45e57. Psillakis, E., Goula, G., Kalogerakis, N., Mantzavinos, D., 2004. Degradation of polycyclic aromatic hydrocarbons in aqueous solutions by ultrasonic radiation. Journal of Hazardous Materials B108, 95e102. Riesz, P., Berdahl, D., Christman, C.L., 1985. Free radical generation by ultrasound in aqueous and nonaqueous solutions. Environmental Health Perspectives 64, 233e252. Roswell, V.F., Tangney, P., Hunt, C., Voulvoulis, N., 2010. Estimating levels of micropollutants in municipal wastewater. Water, Air, and Soil Pollution 206 (1e4), 357e368. Servant, G., Caltagirone, J.P., Gerard, A., Laborde, J.L., Hita, A., 2000. Numerical simulations of cavitation bubble dynamics induced by ultrasound waves in a high frequency reactor. Ultrasonics Sonochemistry 7, 217e227. Simoneit, B.R.T., 1984. Organic matter of the troposphere III. Characterization and sources of petroleum and pyrogenic residues in aerosols over the western United States. Atmospheric Environment 18 (1), 51e67. Stein, E.D., Tiefenthaler, L.L., Schiff, K., 2006. Watershed-based sources of polycyclic aromatic hydrocarbons in urban storm water. Environmental Toxicology and Chemistry 25 (2), 373e385. Suslick, K.L., 1990. Sonochemistry. Science 247 (4949), 1439e1445. Taylor Jr., E., Cook, B.B., Tarr, M.A., 1999. Dissolved organic matter inhibition of sonochemical degradation of aqueous polycyclic aromatic hydrocarbons. Ultrasonics Sonochemistry, 175e183. Thompson, L.H., Doraiswamy, L.K., 1999. Sonochemistry: science and engineering. Industrial and Engineering Chemistry Research 38, 1215e1249. USEPA, 1999. Method 1664, Revision A: n-hexane Extractable Material (HEM; Oil and Grease) and Silica Gel Treated nhexane Extractable Material (SGT-HEM; Non Polar Material) by Extraction and Gravimetry. EPA-821-R-98e002. United States Environmental Protection Agency. Virkutyte, J., Rokhina, E.V., 2010. Hybrid advanced oxidation techniques based on cavitation for micropollutants dedgradation. In: Virkutyte, J., Varma, R.S., Jegatheesan, V. (Eds.), Treatment of Micropollutants in Water and Wastewater. IWA Publishing, London, UK. Wheat, P.E., Tumeo, M.A., 1997. Ultrasound induced polycyclic aromatic hydrocarbon reactivity. Ultrasonics Sonochemistry 4 (1), 55e59. Yang, S.K., Silverman, B.D., 1988. Polycyclic Aromatic Hydrocarbon Carcinogenesis: StructureeActivity Relationships, vol. I. CRC Press, Florida, USA.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 9 5 e2 6 0 6
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
A novel tracer technique for the assessment of fine sediment dynamics in urban water management systems K.L. Spencer a,*, I.G. Droppo b, C. He b, L. Grapentine b, K. Exall b a b
Queen Mary University of London, School of Geography, Mile End Road, London E1 4NS, UK Environment Canada, 867 Lakeshore Rd., P.O. Box 5050, Burlington, Ontario L7R 4A6, Canada
article info
abstract
Article history:
Urban storm water run off can reduce the quality of receiving waters due to high sediment
Received 14 October 2010
load and associated sediment-bound contaminants. Consequently, urban water manage-
Received in revised form
ment systems, such as detention ponds, that both modify water quantity through storage
9 February 2011
and improve water quality through sediment retention are frequently-used best manage-
Accepted 11 February 2011
ment practices. To manage such systems effectively and to improve their efficiency, there is
Available online 18 February 2011
a need to understand the dynamics (transport and settling) of sediment, and in particular the fine sediment fraction (<63 mm) and its associated contaminants within urban storm water
Keywords:
management systems. This can be difficult to achieve, as modelling the transport behaviour
Cohesive sediment
of fine-grained and cohesive sediment is problematic and field-based measurements can be
Sediment tracer
costly, time-consuming and unrepresentative.
Sediment transport
The aim of this study was to test the application of a novel cohesive sediment tracer
Storm water detention pond
and to determine fine sediment transport dynamics within a storm water detention pond.
Flocculation
The cohesive sediment tracer used was a holmium labelled montmorillonite clay which flocculated and had similar size and settling velocity to the natural pond sediment it was intended to mimic. The tracer demonstrated that fine sediment was deposited across the entire pond, with the presence of reed beds and water depth being important factors for maximising sediment retention. The results of the sediment tracer experiment were in good agreement with those of a mathematical sediment transport model. Here, the deposited sediment tracer was sampled by collecting and analysing surface pond sediments for holmium. However, analysis and sampling of the three dimensional suspended tracer ‘cloud’ may provide more accurate information regarding internal pond sediment dynamics. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
During high precipitation events large volumes of water flow over a range of urban surfaces, significantly altering both the quantity and quality of water entering the drainage basin. The use of wet detention ponds is a structural best management
practice (BMP) which can modify storm water quantity by providing storage and increasing residence times, and improve storm water quality through a range of physical, biological and chemical processes (e.g. Weiss et al., 2006). As many contaminants, in particular heavy metals, have an affinity for particulate matter, a primary mechanism for the removal of
* Corresponding author. Tel.: þ44 (0)20 7882 8200; fax: þ44 (0)20 7882 7032. E-mail addresses:
[email protected] (K.L. Spencer),
[email protected] (I.G. Droppo),
[email protected] (C. He), Lee.Grapentine@ ec.gc.ca (L. Grapentine),
[email protected] (K. Exall). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.02.012
2596
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 9 5 e2 6 0 6
contaminants and hence improvement in water quality is physical sedimentation (Li et al., 2007), including both the settling of sediment and the physical trapping of sediment by vegetation. Consequently, the efficiency of detention ponds for removing contaminants can depend on a range of factors which control sedimentation including pond hydrodynamics, sediment and contaminant characteristics, and the presence of vegetation (Li et al., 2007). In order to manage storm water ponds effectively and to inform better design, there is a need to understand and predict the key parameters which control contaminant removal efficiency and in particular sediment and fluid dynamics. Detention pond efficiency can be measured directly by comparing contaminant flux at the pond inlet and outlet or, by examining sediment accumulation and estimated flux of sediment/contaminant entering the pond (e.g. Starzec et al., 2005; Hossain et al., 2005). However, these approaches are not predictive and are site specific, and the dynamic nature of flow through a detention pond can make representative sampling of inflow and outflow conditions difficult. In addition, this approach treats the detention pond as a ‘black box’ with no understanding of internal sediment dynamics or sediment and contaminant reworking. An alternative approach is to model sediment dynamics within the pond, providing information on sediment transport and settling, and hence, pond efficiency. Such models rely upon the settling theory for discrete particles (Li et al., 2007) and assume that sediment particles are spheres and settle according to Stoke’s law, with settling velocity proportional to particle diameter. However, modelling the fine, cohesive sediments responsible for much contaminant transport in detention ponds may be problematic, in particular as these sediments flocculate and interact with their carrying medium, with shape and size continually changing during transport (Droppo, 2001; Krishnappan and Marsalek, 2002). As a result, few models successfully predict sediment/contaminant treatment performance for urban storm water management systems (Wong et al., 2006). Another alternative to understanding sediment dynamics within these environments may be the use of particle tracers, where natural or synthetic materials are tagged with a ‘signature’ that can be detected following release and subsequent dispersion of the tracer in the environment. Such tracers have been widely used to examine the transport dynamics of soils (e.g. Polyakov and Nearing, 2004) and sand-sized sediment (e.g. Silva et al., 2007; Tonk and Masselink, 2005) in a range of terrestrial and aquatic environments. A number of workers have also investigated the use of labelled clays as tracers for cohesive sediment in a range of aquatic environments including lakes, estuaries and Karst systems (e.g. Mahler et al., 1998a; Mahler et al., 1998b; Spencer et al., 2007). The use of labelled clays as cohesive tracers does have its difficulties (Spencer et al., 2007). However, they offer significant advantages over synthetic alternatives (e.g. labelled polymeric substances) as they have proven dynamic characteristics which are similar to the natural cohesive mud that they are intended to mimic (Spencer et al., 2010). Therefore, particle tracers may be used to understand cohesive sediment transport in the field and to improve and validate cohesive sediment transport models. However, there are few examples of the
application of cohesive sediment tracers in field-based studies and there is little available information on protocols for their deployment. Here, we have used both a cohesive sediment tracer (a holmium (Ho) labelled montmorillonite clay) and numerical modelling to determine the spatial and temporal distribution of fine sediment circulation and deposition within the East Storm Water Detention Pond (hereafter referred to as EP) of the regional Waste Management Site, Halton, Ontario, Canada following a simulated storm flow event. Concern has been expressed at this site over the periodically high metal concentrations in whole water samples collected down stream of the EP. The results of this study will provide a better understanding of how the pond contributes to the removal of fine sediments and sediment-bound contaminants and will help in the development of improved water management strategies and treatment technologies for fine sediment and their associated contaminants. In addition, this serves as a test case to demonstrate the use of cohesive sediment tracers for measuring fine sediment dynamics and suggests potential protocols for their deployment in urban storm water management systems.
2.
Materials and methods
2.1.
Site description and monitoring
The storm water detention pond assessed (the EP) is located on the Regional Municipality of Halton’s Waste Management Site in Milton, Ontario, Canada. The detention pond receives no direct surface runoff or sub-flow from the landfill, but does receive water and sediment from a small first order creek/ditch which runs adjacent to the landfill area and conveys water from a predominantly agricultural catchment and surrounding roadway. The creek enters the EP on the south east corner via two parallel culverts and exits via a concrete outlet channel at the north end (Fig. 1). In a typical year, the EP is 120 50 m and 2 m deep and discharges to the Oakville Sixteen Mile Creek. However, during the summer of 2007, when the tracer release field experiment took place, water levels in the detention pond were very low (average 0.7 m) due to low rainfall and high water extraction for dust control. Although this is representative of conditions in the pond for much of the summer, the low water level was below the level of the outlet culvert so that no outflow occurred during the tracer release experiment. A flume with a calibrated V-notch weir was installed in the south culvert of the inlet (the second culvert was dry during the simulated storm flow event e for detailed explanation see section 2.3) and a second calibrated V-notch weir was installed at the outlet concrete channel. Data logging area flow velocity metres (Sigma 950) were installed at the inlet and outlet (although no flow was measured here) to allow for accurate measurements of flow. A data logger (Campbell Scientific Canada) was used to collect meteorological data for use in the hydrodynamic modelling and recorded air temperature, relative humidity, wind speed and direction every 5 min from a portable meteorological station. The station was set up at the southwest corner of the pond (Fig. 1).
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 9 5 e2 6 0 6
2597
Fig. 1 e Location map showing the pond dimensions, inlets/outlets, meteorological station and bed sediment coring sampling strategy.
2.2.
Hydraulic and particle transport modelling
A three dimensional hydrodynamic numerical model (MIKE 3) was selected for this study based on several key features: a flexible unstructured mesh, conservation of mass of all modelled components, and flexible coordinates in the vertical direction. The model has been verified in previous studies by comparing simulations with field data (He et al., 2008). MIKE 3 was adopted to investigate the flow conditions and transport of fine sediment in the pond under measured meteorological conditions, inflow rate data and measured suspended fine sediment concentration at the inlet. The numerical simulation of the EP provides information on flow residency times, flow pattern and the fate of sediment within the pond. In the mud transport simulations, the pond was divided into 1054 small cells in the horizontal plane and 10 evenly distributed vertical layers. The model calculated mean sediment concentration at the centre of each individual 3 dimensional cell at every time step. A constant horizontal eddy viscosity (0.0002 m2 s1) was used and the widely used two equation kee turbulent model (built in MIKE 3) was adopted for calculating the vertical eddy viscosity (Svensson, 1978). A constant drag coefficient of 0.01 was specified in the quadratic drag formula for the calculation of bed resistance. These two constants
were chosen based on the best fit of modelled and measured hydraulic conditions in the pond. The wind force varied in time and was constant in the domain. The measured inflow rate was used to specify the hydraulic model inlet boundary condition. The model started at rest with velocity being set to zero everywhere in the simulation domain. The mean size and density of particles in the mixed sediment tracer slurry (see section 2.3) were used to calculate the settling velocity from Stoke’s law for the mud transport model set up. The inlet boundary condition of the sediment transport model was specified with particle concentrations measured in the inlet. In the model simulation particles were not released into the pond through the model inlet boundary until 1 h after the initial model run to allow the simulated velocity to increase from zero up to a realistic value. Due to high turbulent flow in the vicinity of the pond inlet, and the low suspended sediment (SS) concentrations, flocculation was not included in the simulation.
2.3.
Tracer release experiment
A cohesive sediment tracer was deployed in the EP in order to trace the movement and deposition of suspended fine sediment within the pond following a simulated single inflow or
2598
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 9 5 e2 6 0 6
storm event. The experimental design allowed for the identification of areas of maximum sediment deposition indicating which areas of the pond contributed most effectively to the removal of sediments and associated contaminants. The tracer used was a Ho-labelled montmorillonite clay which has previously been proposed as a potential cohesive sediment tracer for a range of aquatic environments (Spencer et al., 2007). Prior to deployment of the tracer, Ho concentrations in the EP bottom sediment were below analytical detection. Consequently, any Ho measured in the sediments following tracer release could be attributed to the distribution and settling of the tracer within the pond. The montmorillonite was labelled with Ho in batch sorption experiments. The conditions for sorption were as follows; clay mineral: Ho solution ratio (1:10), Ho solution concentration (0.025 M), ionic strength (NaNO3, 0.01 M), pH 5, temperature (20 C), and equilibration time (72 h). Details for the tracer’s preparation and physical and chemical behaviour in the environment are presented elsewhere (Grapentine et al., 2007; Spencer et al., 2007, 2010). The tracer had a Ho content of approximately 4000 mg kg1 and following preparation was dried and ground to <63 mm using a Tema ball mill. Prior to the deployment of the tracer, an extensive study was carried out to investigate its potential toxicological effects on the aquatic biota in the EP. Four benthic invertebrate taxa (Chironomus riparius, Hexagenia spp., Hyalella azteca and Tubifex tubifex) were exposed to 100% Ho-montmorillonite and mixtures of 0, 10, 25 and 50% Ho-montmorillonite and field-collected reference pond sediment for 10e28 days in standard laboratory toxicity tests (Grapentine et al., 2007). Only the 100% Ho-montmorillonite treatment resulted in significantly higher toxicity than the reference pond sediment. Exposure to a negative control treatment of 100% montmorillonite (no Ho) resulted in minor reductions in growth, but no lethal responses. At concentrations of approximately 2000 mg kg1 (i.e. mixtures of 50% pond sediment and 50% Homontmorillonite), no toxicity was observed (Grapentine et al., 2007) and consequently before the tracer was deployed it was mixed with natural pond sediment so that the final Ho concentration was no more than 2000 mg kg1. A sediment slurry (total dry mass approximately 6 kg) of 50% pond bottom sediment and 50% Ho-montmorillonite (50PS:50Ho) was thoroughly mixed in a 100 L pail using a mixer shaft rotated by a high speed drill and was used as the cohesive sediment tracer in these experiments. Prior to deployment SS concentration and Ho concentration in the slurry were determined and samples were collected for particle structure and settling velocity measurements (section 2.4). This slurry was then introduced to the storm water pond via the inlet flume during an artificial inflow or storm event. In order to simulate a storm event, a tanker truck was filled with 11000 L of pond water, which was released into the storm water detention pond via the inlet flume at a rate of approximately 6 L s1. Two artificial storms were created. The first on October 10, 2007 when a slurry of pond bottom sediment was injected into the flume at the inlet as a test run of procedures, and the second on October 12 with the injection of the 50PS:50Ho sediment tracer slurry. Two complete flushes of the truck were used in each case to represent a bimodal storm and to allow for the complete flushing of the sediment slurry into
the pond. Each complete flush of the truck took approximately 30 min. During the initial flush of the tanker truck, and prior to the slurry injection, samples were taken at the V-notch weir to analyse for background Ho concentrations in the pond water and sediment, SS concentration and floc structure. The 50PS:50Ho sediment tracer slurry was then injected into the flume of the inlet using a Cole-Palmer peristaltic pump and samples were collected for analysis of SS and Ho concentrations every 2 min beginning at 30 s after sediment tracer slurry release. Floc samples were collected for structural characterisation and sizing at the start, middle and end of the slurry injection time, while 4 L samples were collected at the middle of the injection for settling velocity determination. Following release of the sediment tracer slurry, a sediment plume within the pond and extending from the inlet was clearly visible during the artificial storm events and grab samples were collected from within the plume using a pole sampler and 500 ml sample bottles. The samples were analysed for SS concentration and served as a validation for the mathematical model described below. In addition, surface floats were deployed at the inlet and a video was recorded to help delineate the extent of plume migration over the duration of the tracer deployment. Fig. 1 illustrates the physical set up during the tracer deployment. Note that flow from the inlet culvert passed through a small dry reed bed (which is normally submerged when water levels are higher) before reaching the water level of the pond. Seven days after tracer release, surface sediment samples were taken from the reed bed and pond shoreline, while sediment cores were collected from the pond bed. All sampling locations were marked using GPS coordinates. Three composite reed bed sediment samples were collected by taking surface scrapes over approximately a 5 cm 5 cm area along a transect from the inflow culvert to the water’s edge using a stainless steel spatula. Thirty pond shore surface sediment samples were collected approximately every 15 m along the perimeter of the pond, 30 cm from the edge and from an area of approximately 5 cm 5 cm under a few cm of water. This sediment was collected using a 50 cm3 plastic syringe with 50 cm of plastic tubing attached to the end of the syringe. Surface sediments were collected by directing the tubing over the sediment surface while carefully extending the syringe plunger and ‘sucking up’ the very soft and unconsolidated surface sediment. The syringe was emptied and then flushed with pond water into the sample container. In order to collect surface sediments from within the pond forty-six core samples were collected from a boat along multiple transects T1eT9 (Fig. 1) with sample density increasing towards the inlet. Using a boat, cores were collected from below the water level using a 10 cm diameter plastic tube. Core tubes containing both collected sediment and over-lying water were then extruded on site. As the cores were extruded most of the over-lying water was discarded. However, 4 cm of water directly above the sedimentewater interface and any disturbed flocs were siphoned off and retained. Once over-lying water had been removed, the sediment was extruded and a 5 mm sub-sample was sliced from the core surface using a small core ring and stainless steel scraper. This sub-sample was added to the overlying water and disturbed flocs and retained for analysis. All sediment samples were returned to the laboratory and stored
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 9 5 e2 6 0 6
at 4 C within a few hours of collection. To avoid cross contamination in the field all equipment was washed in pond water, rinsed three times in deionised water and then rinsed in clean pond water between use.
2.4.
Laboratory analyses
2.4.1.
Ho analysis
Sediment samples were first freeze dried, then ground, followed by microwave digestion with 0.5 g of concentrated nitric acid. The digestate was then diluted to 50 mL with deionised water and Ho was analysed by ICP-OES (Perkin Elmer Optima 5300v), with an analytical detection limit of 0.5 mg g1 Ho. Suspended solids concentration was determined gravimetrically by filtering a known volume of sample through a pre-weighed 0.45 mm cellulose acetate filter followed by drying at 100 C for 1 h and re-weighing the filter to determine the mass of sediment per unit volume.
2.4.2.
Floc analysis
Floc samples, from both pond sediment and the 50PS:50Ho sediment tracer slurry, were collected within 3 mL plankton chamber cells for the characterisation of floc size distributions and structure following the method of Droppo et al. (1997). This allows for the non-destructive direct sampling and observation/measurement of flocculated material. The flocs were imaged (sized) down to a lower resolution of approximately 2 mm (10 objective) using a Zeiss Axiovert 100 microscope interfaced with an image analysis system (Open Lab). Floc settling velocity, density and porosity were determined using the settling chamber method (Droppo et al., 1997). A drop of sediment collected with a wide mouth pipette (3.74 mm) was introduced into an insulated 2.5 L capacity settling column. Settling flocs pass through the field of view of a Nikon stereoscopic microscope where they are digitally captured trough a digital Hamamatsu video camera interface. Using Open Lab, the settling velocity was measured by digitally overlaying two digitally captured frames containing an identified floc separated by a known time interval. Floc density was estimated using Stokes’ law given the measured aggregate size and settling velocity. Although Stokes’ law is not ideal for the determination of aggregate density due to the heterogeneous structure and irregular shape of aggregates, Stoke’s law or a modification thereof has often been used to determine the wet density of singular aggregates (Li and Ganczarczyk, 1987; Droppo et al., 1997) and provides an indication of how aggregate settling velocity, density, and porosity are related to aggregate size. The aggregate porosity can be expressed by a mass balance equation (Equation (1)) assuming a typical density of dried silt and clay of 1.65 g cm3. . 3 ¼ rs rf ðrs rw Þ
(1)
Where 3 ¼ aggregate porosity, rs ¼ density of the dried solid material, rf ¼ wet density of the aggregate and rw ¼ density of the water (Li and Ganczarczyk, 1987).
3.
Results
3.1.
The tracer release experiment
2599
The sediment tracer slurry used in this experiment (50PS:50Ho) had the same visual appearance as the natural pond sediment. Both sediments (pond sediment and Ho-montmorillonite) were observed to be irregular in shape with an open matrix and high water content. The 50PS:50Ho sediment tracer slurry produced flocs with a smaller size range and formed fewer large flocs compared to the pond sediment (Fig. 2a and b). The settling velocities of the pond sediment and 50PS:50Ho sediment tracer flocs were very similar with no significant difference between the slopes of the regression line (Fig. 3a and b; t-test, p ¼ 0.05) and greater than 80% of the variance explained. The 50PS:50Ho sediment tracer formed flocs that were slightly denser with lower porosity (Fig. 3d and f respectively) than the pond sediment flocs (Fig. 3c and e respectively). However, the effects of density and porosity on settling velocity are less important than floc size when floc densities are so close to that of water (floc density ranged from 1.01 to 1.1 g cm3) (Droppo, 2004). Following tracer release, a sediment plume was clearly observed by the delineated brown sediment colour and the movement of the surface floats. Within the first few hours the plume moved in a northeast direction, followed by a due east flow with the plume dispersing as the sediment dissipated due to topographical interactions with general currents initiated by flow and wind. During tracer release, water samples were collected from the inlet flume to assess the change in total Ho concentrations with changes in SS concentration. Fig. 4 shows that although the tanker truck flush took approximately 30 min, the sediment slurry was injected over a 10 min interval with the majority of the tracer released in the first few minutes of discharge. During this time the SS concentration reached 1300 mg L1. As sediment loading increased, so too did the Ho concentrations, reaching a maximum of 2200 mg L1. The plume was allowed to disperse and settle for 7 days following tracer release. All surface sediment samples collected 7 days after tracer release contained Ho concentrations above the analytical detection limit and therefore above natural background concentrations of Ho in these sediments. Therefore, any measurable Ho present in these sediments could be attributed to distribution and settling of the tracer sediment during/ following the simulated storm event. Fig. 5 provides the spatial interpolation of the Ho concentrations with the highest concentrations close to the inlet and extending out in the direction of the observed sediment plume suggesting that the tracer was distributed to all parts of the pond within the 7-day experiment with most sediment tracer deposition occurring near the inlet. The concentration of Ho in the reed bed sediments increased from 1.1 mg g1 close to the culvert to 3.7 mg g1 at the water’s edge.
3.2.
Hydrodynamic modelling
Fig. 6 provides SS concentrations for three modelled time steps for the bottom most layer of SS (0e10 cm from the pond
2600
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 9 5 e2 6 0 6
Fig. 2 e Representative images of flocs collected at the start of artificial storms a) pond sediment and b) 50:50 mix of pond sediment and Ho-montmorillonite tracer.
Fig. 3 e Example settling velocity, excess density (density e density of water) and porosity for pond sediment (a, c and e) and 50:50 mix of pond sediment and Ho-montmorillonite tracer (b, d and f).
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 9 5 e2 6 0 6
2601
Fig. 4 e Pollutograph of suspended sediment and total Ho concentrations (whole water samples) with time during injection of slurry into flume at the pond inlet.
bottom). The time steps displayed (51, 106 and 500) corresponded to 4.5, 9 and 42 h after the start of the simulated storm, respectively. The bottom most layer is assumed to represent proportional depositional areas within the pond. All three time steps indicate that SS is distributed throughout the entire pond, which agrees with the data from the tracer release experiment. The model also allows for the observation of temporal changes in the transport and likely deposition of SS throughout the pond. Over the time period there was
a gradual decrease in the concentrations of SS in the bottom layer of the water column (note the different scales for SS concentration on the three figures). This suggests that sediment was settling out with time and within 42 h, most sediment had been deposited. This also confirms that any sediment re-distribution caused by the first simulated storm event is unlikely to impact on our tracer release experiment 48 h later. The model shows that the highest concentrations of SS in the bottom layer (360e400 mg L1) and consequently, the
Fig. 5 e Spatial interpolation of Ho distribution in surface sediments (mg gL1).
2602
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 9 5 e2 6 0 6
Fig. 6 e Model outputs of suspended sediment tracer for 3 time steps over a 2-day period for the bottom most simulation layer (10 cm above the wateresediment interface) (axes indicate grid reference). greatest proportion of sediment that was removed from the pond water column, occurred close to the inlet during the initial phases of the simulation (Fig. 6a). The peak concentration of SS then migrated towards the outlet (Fig. 6b and c)
with time. The final time step (Fig. 6c), although with very low concentrations of SS, shows a secondary ‘hot spot’ of SS concentration near the inlet, indicating circulation of flow within the pond. This may be attributed to the low flow
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 9 5 e2 6 0 6
conditions and absence of outflow at the time of the experiment. Fig. 7 provides model outputs for SS concentrations for the same three time steps for the surface layer of the pond (10 cm below the surface). The SS concentrations in the surface layer
2603
are consistently lower than SS concentrations in the bottom layer for each corresponding time step. Here, we see the influence of the wind on the surface SS concentration with the highest SS concentrations always further down wind than those at the corresponding time step in the bottom simulation
Fig. 7 e Model outputs of suspended sediment tracer for 3 time steps over a 2-day period for the upper most simulation layer (10 cm below the surface of the pond) (axes indicate grid reference).
2604
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 9 5 e2 6 0 6
layer (Fig. 6) which indicates that the flow is stratified even in a shallow small pond. This is particularly evident in Fig. 7b, where there is a ‘tail’ trailing southward from the higher SS concentration zone. This ‘tail’ was also visually observed and validated by the position of float markers. Note that for the same time step the highest SS concentration for the lower layer (Fig. 6b) is slightly more north (up wind) with no tail. In this way, the three dimensional model showed the dynamic nature of SS transport due to different flow conditions in the surface and bottom layers of the pond. Fig. 8 provides model outputs for the total sediment tracer accumulation in the final time step of the simulation, and as such can be directly compared with the measured Ho distribution in surface sediments at the end of the tracer release experiment. This shows highest concentrations of sediment tracer accumulation near the inlet and along the SE shore of the pond in close proximity to the inlet with a ‘tail’ extending NW into central parts of the pond.
4.
Discussion
4.1. Assessment of fine sediment dynamics within the detention pond A key characteristic of a sediment tracer is that its physical and transport properties mimic those of natural sediment (Black et al., 2007). For the fine sediment fraction (i.e. silts and clays), which is cohesive and primarily transported as flocs (Droppo, 2001), this means that the tracer must flocculate and those flocs must have the same dynamic characteristics (structure and transport) as natural sediment flocs. The cohesive tracer slurry (50PS:50Ho) used in these experiments formed flocs with similar settling velocities to pure natural pond sediment. The similarity between the slopes of the regression curves for the natural mud and tracer also suggests that both materials have similar structures and flocculation behaviours with particles of a similar diameter
having similar settling velocities. This also indicates that diameter is probably the most important factor for controlling settling velocity for these materials. However, the tracer formed fewer large flocs and in particular did not form any flocs >600 mm in diameter, compared to the natural pond sediment. For the pond sediment it is these large flocs that will have the highest settling velocities and therefore would settle out and be trapped most effectively in the detention pond. Suspended particulate mass (SPM) concentration and mass settling flux (MSF), both important parameters for understanding transport behaviour of fine sediments, were not observed in this study. However, in studies comparing the dynamic behaviour of the Ho-labelled tracer to natural estuarine mud (Spencer et al., 2010), both the tracer and flocs comprising tracer and natural sediment were found to have different SPM distributions and higher MSF even though floc size and settling velocities were in similar ranges. Therefore, it is possible that the tracer slurry used in these experiments (50PS:50Ho) may have slightly different transport characteristics to the natural pond sediment. As flocs increase in size they have lower densities and higher porosity values as the flocs incorporate pore space, water and biological material into their structure (Droppo, 2004). Here, the tracer flocs were slightly denser and less porous than the natural pond sediment and this is likely due to lower organic matter concentrations in the 50PS:50Ho tracer slurry compared to pure pond sediment. These observations do not agree with a detailed study of the physical and dynamic characteristics of the Ho-labelled montmorillonite tracer (Spencer et al., 2010) where flocs comprising mixtures of natural sediment and Ho-montmorillonite tracer were larger, less dense, more porous and faster settling than either pure tracer or natural sediment. In Spencer et al. (2010)’s study this was attributed to the difference in tracer and natural mud microfloc shape resulting in poor packing of the larger macroflocs. Such detailed observations were not carried out for this study, however, in spite of these slight physical differences, for the purpose of this experiment, the tracer sediment
Fig. 8 e Model outputs for total sediment tracer accumulation on the pond bed in time step 500 (42 h) (axes indicate grid reference).
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 9 5 e2 6 0 6
demonstrated similar transport characteristics to natural pond sediment and is therefore likely to provide a good representation of cohesive sediment dynamics within the detention pond. However, the absence of very large tracer flocs may mean that the tracer has slightly longer transport distances (assuming no floc breakage) which would suggest that our experiment and modelling is representative of a worst case scenario for the pond operation in terms of sediment settling and retention. During the artificial storm, the tracer sediment flowed from the culvert through a dry reed bed and then into the wet pond. Seven days after the artificial storm, Ho was detected within the reed bed sediment samples with decreasing concentrations from the culvert to the water’s edge. These concentrations were considerably higher than Ho concentrations in the pond bottom sediments, suggesting that the reeds are effectively trapping fine sediment and associated contaminants as it is carried from the culvert to pond. Typically, this vegetation would be submerged with higher water levels and this would increase the sediment trapping capacity by spreading out the flow path and reducing flow velocity. There was strong agreement between the results of the hydrodynamic model and the tracer experiment. Both indicated that the majority of fine sediment and consequently, any sediment-bound contaminants were removed quickly near the inlet and only small amounts (likely finer sediments) were transported further into the pond. The model output tends to show quite a narrow distribution of sediment tracer near the inlet, whilst the measured Ho concentration in surface sediments has been distributed more widely (e.g. compare Figs. 5 and 8). This is because in the model, inflow is assumed to enter the pond directly from the inlet pipe, whilst in the field experiment, due to low water levels, the inflow first passes through an emergent reed bed which would have diffused the tracer as it entered the pond. Whilst Ho concentrations were low (<1 mg g1) in the majority of the pond sediment samples, Ho was detected throughout the pond suggesting that, under these experimental flow and climatic conditions, the entire pond contributed to the removal of sediments and associated contaminants. This was also supported by the hydrodynamic model. It should be noted that over the 7-day tracer experiment period, water was being extracted from the pond by the waste management authority for dust control. This extraction was however, infrequent, of low volume relative to the pond, and occurred as far away from the outlet as possible at the south end of the pond. As such, it is assumed that this had little effect on the distribution of Ho within the EP. We chose not to include flocculation within the model simulations. In part, this was due to the low SS concentrations and high levels of turbulence that would occur during the simulated storm event. However, this also allows comparison with the outputs of many other commonly used sediment transport models which do not have a flocculation component. Yet, even without the inclusion of a flocculation component in the model there was good agreement between the tracer (which does flocculate) and the sediment transport model. Sediment transport in the surface water layers is likely to be affected by climatic conditions such as wind direction.
2605
During the numerical simulation, after two days the peak SS concentration in the surface water was near to the outlet (Fig. 6), albeit of low SS concentration (<0.042 mg L1). At the time of this experiment the detention pond was not discharging due to low water levels. It is possible that with these low water levels, re-suspension of the settled bed sediment in low water depths was occurring, contributing to the redistribution of Ho observed with the experiments and the modelling exercise. With higher water levels, there would likely be lower re-suspension for similar wind conditions suggesting the importance of maintaining an optimal pond depth. With higher water levels, however, there would be discharge from the pond via the outlet culvert, so any sediments and associated contaminants reaching the outlet can then move down stream. Within this research, we were only able to determine the cohesive sediment tracer’s settled footprint rather than sampling the three dimensional, temporally changing suspended tracer ‘cloud’, which would have allowed better comparison with the time integrated sediment transport model. However, this is a typical approach to sampling non-cohesive sediment tracers (e.g. Kimoto et al., 2006) where sediment load tends to settle very quickly and protocols for sampling are relatively well established. In order to provide a more detailed picture of suspended sediment dynamics within the pond and consequently to predict the discharge of sediment via the outlet, it would be beneficial to sample the tracer cloud and verify with a modelling exercise, particularly when there is an outflow from the EP.
4.2. Possible implications for the Halton East storm water detention pond From this combined modelling and tracer study it can be concluded that all areas of the EP are retaining suspended solids and consequently associated contaminants (although no outflow was occurring at the time of sampling). The distribution of retention is not equal over the pond (Figs. 6 and 7), being dependent on sediment dynamics (floc structure and settling), hydraulics (flow and wind induced currents) and pond geometry. The presence of the macrophytes influenced sediment retention at the inlet of the pond; additional plantings or physical modifications to the pond could be used to further enhance sedimentation if needed. The current regime of water extraction for dust control may also indirectly be a useful management operation resulting in no outflow from the pond. If the pond can be maintained within the optimal depth of 1e2 m but with no outflow, then an added storage capacity is utilised with outflow only occurring when water levels are higher and the outlet grade is exceeded. No flow to the receiving stream could, however, have detrimental wildlife effects that must be considered.
5.
Conclusion
Following ecotoxicological assessments of the Ho-montmorillonite, a 50:50 mixture of Ho-montmorillonite and natural pond sediment was used in this study as a cohesive sediment tracer. An examination of the tracer indicated that it formed flocs with similar physical and settling characteristics to
2606
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 5 9 5 e2 6 0 6
natural pond sediment, although the tracer did not form very large flocs and was slightly denser and less porous than natural sediment. Consequently, it satisfies the basic assumptions of tracer technology. Sediment transport pathways and areas of deposition identified by the sediment tracer are in agreement with sediment transport pathways predicted by the hydrodynamic sediment transport model. This validates the use of the cohesive sediment tracer and demonstrates its suitability for the assessment of sediment dynamics in small confined areas such as the Halton East storm water detention pond which are dominated by fine, cohesive sediments. Sampling of surface sediments for settled tracer provided information on sediment distribution and sediment retention in the pond. However, to fully understand sediment dynamics, sediment re-suspension and potential discharge of suspended sediment through the outlet culvert it may be beneficial to develop protocols for the sampling of the three dimensional tracer ‘cloud’. For the conditions present at the time of this study, the entire area of the storm water detention pond contributed to the removal of sediments and consequently, any associated contaminants. The reed bed at the pond inlet provides some fine sediment removal, however further macrophyte plantings within the pond at the inlet and outlet could further improve pond performance and effluent water quality.
Acknowledgements The authors would like to thank the Regional Municipality of Halton and the National Water Research Institute for their support related to this project. In particular, the support and mobilisation of staff for this study by Mr. A. Mercer is gratefully appreciated. The students, H. Labelle (McMaster University) and S. Deignan (University of Waterloo) were also strong contributors to this project. Development of the Ho-montmorillonite sediment tracer was funded by the National Environmental Research Council (NER/D/S/2003/00706).
references
Black, K.S., Athey, S., Wilson, P., Evans, D., 2007. Coastal and shelf sediment transport. In: Balson, P.S., Collins, M.B. (Eds.). Geological Society of London, London, pp. 73e91. Droppo, I.G., Leppard, G.G., Flannigan, D.T., Liss, S.N., 1997. The freshwater floc: a functional relationship of water and organic and inorganic floc constituents affecting suspended sediment properties. Water Air and Soil Pollution 99 (1e4), 43e53. Droppo, I.G., 2001. Rethinking what constitutes suspended sediment. Hydrological Processes 15 (9), 1551e1564. Droppo, I.G., 2004. Structural controls on floc strength and transport. Canadian Journal of Civil Engineering 31 (4), 569e578. Grapentine, L., Webber, J., Thompson, S., Milani, D., Labelle, H., Droppo, I., Spencer, K., 2007. Toxicity of Holmium-labelled Clay to Four Benthic Invertebrates. Water Science and
Technology Directorate Contribution # 07-541. Environment Canada, 29 pp. He, C., Curtis, C., Yerubandi, R., Marvin, C., 2008. Randle Reef Hydrodynamic Numerical Modeling Study. National Water Research Institute Technical Note No. AEMRB-TN08-002. Environment Canada. Hossain, M.A., Alam, M., Yonge, D.R., Dutta, P., 2005. Efficiency and flow regime of a highway stormwater detention pond in Washington, USA. Water Air and Soil Pollution 164 (1e4), 79e89. Kimoto, A., Nearing, M.A., Shipitalo, M.J., Polyakov, V.O., 2006. Multi-year tracking of sediment sources in a small agricultural watershed using rare earth elements. Earth Surface Processes and Landforms 31, 1763e1774. Krishnappan, B.G., Marsalek, J., 2002. Modelling of flocculation and transport of cohesive sediment from an on-stream stormwater detention pond. Water Research 36 (15), 3849e3859. Li, D.H., Ganczarczyk, J.J., 1987. Stroboscopic determination of settling velocity, size and porosity of activated-sludge flocs. Water Research 21 (3), 257e262. Li, Y., Deletic, A., Fletcher, T.D., 2007. Modelling wet weather sediment removal by stormwater constructed wetlands: insights from a laboratory study. Journal of Hydrology 338 (3e4), 285e296. Mahler, B.J., Bennett, P.C., Zimmerman, M., 1998a. Lanthanidelabeled clay: a new method for tracing sediment transport in karst. Ground Water 36 (5), 835e843. Mahler, B.J., Winkler, M., Bennett, P., Hillis, D.M., 1998b. DNAlabeled clay: a sensitive new method for tracing particle transport. Geology 26 (9), 831e834. Polyakov, V.O., Nearing, M.A., 2004. Rare earth element oxides for tracing sediment movement. Catena 55 (3), 255e276. Silva, A., Taborda, R., Rodrigues, A., Duarte, J., Cascalho, J., 2007. Longshore drift estimation using fluorescent tracers: new insights from an experiment at Comporta Beach, Portugal. Marine Geology 240 (1e4), 137e150. Spencer, K.L., James, S.L., Taylor, J.A., Kearton-Gee, T., 2007. Coastal and Shelf Sediment Transport. In: Balson, P.S., Collins, M.B. (Eds.). The Geological Society of London, London, pp. 17e24. Spencer, K.L., Manning, A., Droppo, I., Leppard, G.G., Benson, T., 2010. Dynamic interactions of cohesive sediment tracers and natural mud. Journal of Soils and Sediments. doi:10.1007/ s11368-010-0291-6. Starzec, P., Lind, B.O.B., Lanngren, A., Lindgren, A., Svenson, T., 2005. Technical and environmental functioning of detention ponds for the treatment of highway and road runoff. Water Air and Soil Pollution 163 (1e4), 153e167. Svensson, U., 1978. A Mathematical Model of the Seasonal Thermocline. Lund Institute of Technology, University of Lund, Lund, Sweden. Tonk, A., Masselink, G., 2005. Evaluation of longshore transport equations with OBS sensors, streamer traps, and fluorescent tracer. Journal of Coastal Research 21 (5), 915e931. Weiss, J.D., Hondzo, M., Semmens, M., 2006. Storm water detention ponds: modeling heavy metal removal by plant species and sediments. Journal of Environmental EngineeringASCE 132 (9), 1034e1042. Wong, T.H.F., Fletcher, T.D., Duncan, H.P., Jenkins, G.A., 2006. Modelling urban stormwater treatment e A unified approach. Ecological Engineering 27 (1), 58e70.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 0 7 e2 6 1 5
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Investigation of factors affecting the accumulation of vinyl chloride in polyvinyl chloride piping used in drinking water distribution systems Ryan K. Walter a,1,2, Po-Hsun Lin a,2, Marc Edwards b, Ruth E. Richardson a,* a b
School of Civil and Environmental Engineering, Cornell University, Ithaca, NY, USA Department of Civil and Environmental Engineering, Virginia Polytechnic Institute and State University, Blacksburg, VA, USA
article info
abstract
Article history:
Plastic piping made of polyvinyl chloride (PVC), and chlorinated PVC (CPVC), is being
Received 4 September 2010
increasingly used for drinking water distribution lines. Given the formulation of the
Received in revised form
material from vinyl chloride (VC), there has been concern that the VC (a confirmed human
3 February 2011
carcinogen) can leach from the plastic piping into drinking water. PVC/CPVC pipe reactors
Accepted 12 February 2011
in the laboratory and tap samples collected from consumers homes (n ¼ 15) revealed vinyl
Available online 19 February 2011
chloride accumulation in the tens of ng/L range after a few days and hundreds of ng/L after two years. While these levels did not exceed the EPA’s maximum contaminant level (MCL)
Keywords:
of 2 mg/L, many readings that simulated stagnation times in homes (overnight) exceeded
Vinyl chloride
the MCL-Goal of 0 mg/L. Considerable differences in VC levels were seen across different
PVC
manufacturers, while aging and biofilm effects were generally small. Preliminary evidence
CPVC
suggests that VC may accumulate not only via chemical leaching from the plastic piping,
Drinking water
but also as a disinfection byproduct (DBP) via a chlorine-dependent reaction. This is sup-
Disinfection byproducts
ported from studies with CPVC pipe reactors where chlorinated reactors accumulated more
Biofilms
VC than dechlorinated reactors, copper pipe reactors that accumulated VC in chlorinated
Leaching
reactors and not in dechlorinated reactors, and field samples where VC levels were the same before and after flushing the lines where PVC/CPVC fittings were contributing. Free chlorine residual tests suggest that VC may be formed as a secondary, rather than primary, DBP. Further research and additional studies need to be conducted in order to elucidate reaction mechanisms and tease apart relative contributions of VC accumulation from PVC/ CPVC piping and chlorine-dependent reactions. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
1.1.
PVC pipe use in drinking water delivery
As drinking water distribution system lines are replaced in the United States at a cost of $250 billion over the next few decades, research is needed on the strengths and limitations
of the materials that will be installed ((AWWA) 2001). Polyvinyl chloride (PVC) piping is popular due to its relatively low cost, structural strength, ease of installation, and corrosion-resistant properties (Al-Malack and Sheikheldin, 2001). It is currently estimated that 69% of the piping used in the main drinking water distribution system is plastic, and the majority of the plastic pipe is PVC (Burn, 2005). During the
* Corresponding author. 317 Hollister Hall, Ithaca, NY 14853 USA. Tel.: þ1 607 255 3233; fax: þ1 607 255 9004. E-mail address:
[email protected] (R.E. Richardson). 1 Present Address: Environmental Fluid Mechanics Laboratory, Stanford University, Stanford, CA, USA. 2 Note: these authors contributed equally to this work. 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.02.016
2608
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 0 7 e2 6 1 5
manufacturing of PVC, chlorine and ethylene are combined to create ethylene dichloride, which is converted into vinyl chloride (VC) monomers through a cracking process before polymerization to PVC (Saeki and Emura, 2002). Additionally, chlorinated PVC (CPVC), which is PVC that has been chlorinated via a free radical chlorination reaction and the application of heat, is commonly found in hot water drinking systems and residential homes ((ATSDR) Sept. 2004).
1.2.
Vinyl chloride monomer in PVC pipe
There have been reports that residual VC monomer in the pipe matrix of the PVC/CPVC piping can be released into air or drinking water (Sano et al., 2001). VC is a known human carcinogen and is regulated by the Environmental Protection Agency (EPA) with a maximum contaminant level (MCL) of 2.0 mg/L and an MCL-goal (MCLG) of 0 mg/L in water for potable water (Flournoy and Monroe, 1999). Vinyl chloride concentrations well above the EPA’s MCL were reported in stagnant PVC pipe lines in Kansas, Missouri, Texas, and Arkansas ((MDNS), 2006; Flournoy and Monroe, 1999). Additional studies have detected VC levels above the EPA’s MCL at 14 mg/L in PVC pipe manufactured before 1977 (Flournoy and Monroe, 1999; The Vinyl Institute and Uni-Bell PVC Pipe Association, 1994). While modifications to the manufacturing process in 1977 drastically reduced the VC monomer residual in US-manufactured pipe, little work exists that examines the leaching of VC into water from modern US PVC/CPVC pipes e particularly as the pipe ages. Recent work in Saudi Arabia has been performed indicating that static unplasticized PVC (uPVC) accumulates 2.3 mg/L of VC over 14 days when exposed to ultraviolet radiation (Al-Malack, 2004), 2.5 mg/L of VC over 30 days when exposed to temperatures of 45 C (Al-Malack et al., 1999), 2.0e2.1 mg/L of VC in both raw groundwater and chlorinated drinking water (Al-Malack et al., 1999), and 2.5 mg/L of VC over 30 days when exposed to direct solar radiation (Al-Malack and Sheikheldin, 2001); however, all of these analyses used locally manufactured pipes from Saudi Arabia. Another study with Japanese pipe (unknown manufacturer and production date) found that static PVC pipes filled with deionized water or phosphate buffer did not accumulate detectable VC over a three day period (detection limit in mg/L range), but that segments sealed in serum vials did produce detectable VC at more than 50 mg/L (Ando and Sayato, 1984). Finally, an Italian study looking at the migration of VC into drinking water bottled in locally obtained plasticized PVC found that VC accumulated at a rate of 1 ng/L/day (Benfenati et al., 1991).
1.3.
Study objectives
The objectives of this study were to compare the accumulation of VC from new pipe reactors from different US manufacturers and different PVC formulations (PVC vs. CPVC) over periods from 4 h up to 2 years and to understand the effects of aging, biofilm coverage, and chlorine residual on VC levels. Laboratory studies were performed by creating static pipe reactors using new PVC/CPVC from different manufacturers. Likewise, in-use piping segments aged for 2, 15, and 25 years were procured and tested. Finally, copper piping typically used in water distribution systems was tested, eliminating the
contribution of VC via chemical leaching from plastic piping. Analyses of the tap water of local homes was also conducted to determine which factors led to detectable VC levels at the tap.
2.
Materials and methods
2.1.
Static pipe reactors
Various 3/4 00 diameter PVC, CPVC, and copper pipe types obtained from different manufacturers, or from in-use drinking water lines, were cut into 600 segments. Before use, all new pipes were rinsed with three changes of water in accordance with NSF 60 standards. All pipes obtained were NSF PW certified. For aged piping, pipe segments were maintained full of water and kept in the dark until the time of the experiment. To remove biofilms, a sterile scraper (Falcon) was used to scrape biofilm from the walls, and tap water was used to rinse the scraped surface. One end of the reactor had a fitted cap, of the same pipe material, glued onto it using Miller Stephenson Epoxy 907 two to three days before the actual experiment. A hole was drilled into the end of the second cap, and a 20 mm blue butyl rubber septum (Bellco Technology) was glued in place several days before the start of the experiment. The static reactors were 90% filled with cold, laboratory tap water (chlorinated or dechlorinated) e leaving 10% headspace to allow gas sampling. After sealing the caps and ensuring that no air bubbles formed in the epoxy, the reactors were allowed to dry for at least 2 h before sampling. All reactors were incubated at room temperature in the dark during the experiment. One milliliter gas samples were removed to monitor VC over time and 1 mL of room air was then injected to avoid the creation of a vacuum in the reactors. VC recovery controls were created by spiking known amounts of VC into subsets of reactors.
2.2.
Water characteristics
For all studies performed, cold, laboratory tap water, from Cornell University’s Water Filtration Plant (CU-WFP), was employed. The chlorine residual was measured using a Hach Chlorine Test Kit. Total chlorine residual was consistently between 0.55 and 0.7 mg/L, while the free chlorine residual was constantly between 0.5 and 0.6 mg/L. Other characteristics of the source water reported by the CU-WFP were as follows: pH ¼ 6.9; alkalinity ¼ 128 mg/L as CaCO3; total organic carbon ¼ 2.1 mg/L; turbidity ¼ 0.07 NTU. During all experiments, a 500 mL beaker was filled with water, tested for chlorine, and then used to fill all the reactors in order to ensure the same water in all reactors. For the dechlorinated water, sodium thiosulfate was employed to neutralize the total chlorine in the samples. A 4.03 104 M stock solution was created and a stoichiometric weight ratio of sodium thiosulfate per mg/L of residual chlorine of 0.556 mg/mg was used based on Metcalf and Eddy’s standard for dechlorination (Burton et al., 2002). Residual chlorine testing after dechlorination consistently indicated 0.0 mg/L of both free and total chlorine. Samples of all experimental runs were preserved in glass vials.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 0 7 e2 6 1 5
2.3.
Analytical techniques
2.3.1.
VC measurements
VC concentrations in pipe reactors were measured using gas chromatography with either a flame ionization detector (GCFID) or a mass spectrometer (GC-MS). Concentrations of VC in water were calculated from gas phase concentrations using Henry’s Law constant and the relative volumes of the aqueous and gas phases.
2.3.1.1. GC-FID. A gas tight syringe was used to inject 1000 mL of headspace gas into a Perkin Elmer Autosystem GC-FID with a 60/80 Carbopack B Column (Supelco custom 052300 phase, 1% SP-1000, 80 1/800 ). Vinyl chloride standards were prepared by injecting known amounts of VC (99.5%, Matheson Gas Product Inc.) into 14.6 mL serum bottles with a PTFE-coated septum and crimp cap. Headspace analysis was carried out, and the data integrator gave peak area values, which were used to create a standard curve. The temperature schedule in the column was 70 C for 4 min, ramping by 30 C per minute to reach a final temperature of 190 C, and holding at 190 C for 4 min. The FID was held at a constant temperature of 130 C, and the gas flowrates were 45 mL/min and 450 mL/min for hydrogen (H2) and air, respectively. The limits of quantification (LOQ) and detection (LOD) were taken to be concentrations whose signal was ten times and three times the standard deviation of the baseline signal, respectively. The LOQ for the GC-FID was found to be 95 ng/L (ppt) in water. Any signal between these two limits was deemed “detectable, but not quantifiable” (DNQ), while signals below the LOD were deemed “below the detection limit” (BDL). 2.3.1.2. GC-MS. In order to obtain a lower detection limit, a HewlettePackard GC, 5890 Series II, was coupled to a HewlettePackard MS, 5971 Series to analyze VC according to EPA method 524.2 (EPA, 1995). The mass selective detector was operated at an electron energy and detector voltage of 70 eV and 2000 V, respectively. The column was initially held at 80 C for 1 min, ramped by 10 C per minute to 190 C, and held at 190 C for an additional minute. The select-ion monitoring mode was used in order to obtain the highest analytical sensitivity. Primary and secondary quantification ions were used with the following m/z values: 62 (corresponding to 35Cl) and 64 (corresponding to 37Cl) (EPA, 1995). The GC-MS method limit of detection (LOD) was found to be 1.8 ng/L, with a limit of quantification (LOQ) of 6.0 ng/L. Subsamples were sent to a certified laboratory for quality assurance of the GCeMS method. VC levels were within 30% of the values reported by the certified laboratory. 2.4.
2609
samples were brought back to the lab (within 12 h), a headspace was created by injecting 10 mL of air into the bottle and then removing 10 mL of liquid. The bottle was allowed to equilibrate at room temperature for at least 1 h, followed by a headspace analysis for VC. In all of the houses sampled, a “first flush” sample was taken after 8e12 h of overnight stagnancy in the PVC/CPVC. In homes where non-PVC/CPVC fittings linked the PVC/CPVC lines and the tap, enough water was flushed to clear the line of water that sat in those nonPVC/CPVC fittings. Additionally, in a subset of sampling locations, a second sample was taken after flushing the lines with at least 1.5 times the volume of the PVC/CPVC lines. This allowed access to water that sat overnight in pipes upstream of the PVC/CPVC lengths.
2.5.
Epoxy negative controls
To test whether the Epoxy could contribute to the VC signal, approximately 2 g of epoxy was placed in a 14.6 mL tube and crimp capped immediately. Headspace samples were analyzed by GC-MS for VC using the same method as used for the pipe reactors.
2.6.
Chlorine tests
For the chlorine tests, the free chlorine test kit (HACH kit Model CN-66) with the long path viewing adapter was used to measure free chlorine concentrations. A control was set up in a 160 mL serum bottle with 145 mL of tap water. This serum bottle was crimp capped and incubated alongside the pipe reactors at room temperature in the dark.
2.7.
Statistical analysis
Two sample, Student t-tests, assuming unequal variances, were used for statistical analysis. Also, levels that were DNQ, were taken to be at half the LOQ (3.0 ng/L for the GC-MS and 95 ng/L for the GC-FID), while levels that were BDL, were taken to be zero.
3.
Results
Any samples that revealed levels of VC that were DNQ are depicted at one half the LOQ, and BDL are shown at zero (or as black triangles) in the figures that follow. Standard curves run over a time span of two years had coefficients of variation of 0.16 and 0.04 for the GC-FID (n ¼ 3) and GC-MS (n ¼ 2) methods, respectively. All epoxy controls had VC levels BDL by GC-MS on the first day, and after four months.
Field samples
Field samples from the Ithaca, NY area were collected from different residents’ tap water. Where possible, two water samples were collected in every house: the source water before the PVC/CPVC pipes (well, lake, or municipal clear wells) and also the tap water from locations downstream of runs of PVC/CPVC pipes. Samples were collected by completely filling a 160 mL serum bottle and immediately sealing with a PTFE-coated septum and crimp cap. After the
3.1.
Vinyl chloride accumulation in new pipe
3.1.1.
Long term study with new schedule 40 PVC
The long term trend of VC measurements made on triplicate PVC reactors filled with chlorinated tap water can be seen in Fig. 1. VC concentrations in the reactors were below the detection limit of the GC-FID, until the 63rd day. VC levels reached approximately 130 ng/L after one year and eventually reached
2610
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 0 7 e2 6 1 5
Fig. 1 e Time course of VC concentration found in new Schedule 40 PVC filled with chlorinated tap water. Error bars represent standard error across triplicate reactors and black triangles indicate values below the detection limit (BDL) of the GC-FID. All samples were measured with the GC-FID (LOQ [ 95 ng/L), except for the 715th day where the GC-MS was used (LOQ [ 6 ng/L).
a value of over 300 ng/L on the 581st day. On the 715th day, the improved method of the GC-MS corroborated a value of just over 300 ng/L in water. To estimate losses from the pipe reactors via leaking, control reactors were spiked with VC at 330 ng/L and 180 ng/L. These showed expected levels of VC after nine months.
3.1.2. Short term studies on PVC/CPVC pipes from different manufacturers To better understand VC accumulation on time frames more relevant to consumer usage, repeat studies were conducted with a VC method with improved detection limits (GC-MS method). In the various Schedule 40 PVC reactors, quantifiable VC did not show up until the 29th hour of the experiment (Fig. 2a). By the 101st hour of the experiment all of the samples had detectable VC with concentrations varying from 9.3 to
24 ng/L, with the highest concentration in water contained in manufacturer A’s piping (24 ng/L and 19 ng/L, for the two pipe replicates). As seen in Fig. 2, Manufacturer A’s piping accumulated more VC than the other piping manufacturers. Two types of new CPVC were tested: Schedule 80 CPVC (from manufacturers A, B, and C) and SDR11 CPVC (from manufacturer E) (Fig. 2b). While the VC levels after 101 h were similar in all cases to the Schedule 40 PVC, the CPVC reactors showed faster accumulation at early time points (SDR11 CPVC shown in Fig. 4). At 4 h, the SDR11 CPVC reactors (n ¼ 5) were already at 7 ng/L VC, and 2 of the 3 types of Schedule 80 CPVC were above the GC-MS detection limit. At the 101 h time point, all new reactors, regardless of formulation and manufacturer, showed VC levels between 11 and 25 ng/L. As with PVC, there was a slight trend in the Schedule 80 CPVC by manufacturer at the 101 h time point with manufacturer A > B > C.
3.2.
Aged pipe with and without biofilm
To investigate the effects of pipe aging and biofilm coverage on VC accumulation, three different aged pipes were procured and used to establish aged pipe reactors: 1) laboratory-aged 2-year old Schedule 40 PVC pipe (manufacturer A); 2) 15-year old schedule 80, gray PVC piping extracted from the lunchroom line of the City of Ithaca Water Treatment Plant (IWTP) (originally from manufacturer D); 3) 25-year aged, SDR11, CPVC piping extracted from a consumer’s home (originally from manufacturer E). Subsets of pipe were scraped to remove biofilms. Results from triplicate reactors filled with chlorinated tap water are shown in Fig. 3. For the 2-year old Schedule 40 PVC (Fig. 3a) and the 25-year old SDR11 CPVC (Fig. 3c), the VC accumulation slightly decreased or increased, respectively, in the presence of the biofilm at the end of the time course; however, all reactors with the biofilm intact behaved similarly to the new pipe (Fig. 2, Manufacturer A and Fig. 4, respectively) with similar ranges of VC concentrations after several days. The 15-year old Schedule 80 CPVC pipe showed the most surprising results. Within 2 h, as seen in Fig. 3b,VC in both
Fig. 2 e Time course of VC in new, Schedule 40 PVC (top panel) and new, Schedule 80 CPVC (bottom panel) with chlorinated water. Error bars show standard error across triplicate reactors. The LOQ for the GC-MS was 6.0 ng/L.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 0 7 e2 6 1 5
2611
Fig. 3 e Time course of VC in different types of aged piping with biofilm intact versus biofilm scraped. Solid black lines with triangles denote piping with the biofilm intact, dashed-dotted lines with squares indicate piping with the biofilm scraped off, and error bars represent the standard error across triplicate reactors. The LOQ for the GC-MS was 6.0 ng/L. Panel A (top): 2-year aged Schedule 40 PVC; Panel B (middle): 15-year aged Schedule 80 PVC; Panel C (bottom): 25-year aged SDR11 CPVC.
samples were quantifiable, and by the 8th hour the aged pipe with the biofilm and the aged pipe with the biofilm scraped reached VC levels of 34 and 41 ng/L in water, respectively. On the 72nd hour, the aged piping showed VC levels slightly higher than that with the biofilm scraped at 119 versus 111 ng/ L in water. However, on the 168th hour, the aged pipe with the biofilm scraped had higher levels of VC at 266 versus 230 ng/L in water. Yet, these VC levels are considerably higher than any other short term study, and are higher than the VC levels in new Schedule 40 PVC pipe on the 479th day of the long term study (227 ng/L in water). This pipe’s surface film had a notable rust color relative to other pipes investigated. Some rust color remained even after scraping. The cross section of
this piping with the biofilm, and with the biofilm scraped, is shown in Supplementary Fig. 1.
3.3.
Water samples from PVC/CPVC-utilizing homes (n ¼ 15) around the Ithaca area were collected, and the majority of the samples were below the detection limit for VC; however, several samples contained quantifiable levels of VC. Results are summarized in Table 1. Note that VC levels were measured using a GCeMS for all samples, with the exception of three early samples which were analyzed via GC-FID. These samples are noted in Table 1. The results indicate that VC is only showing up at quantifiable levels in small (<100 ), CPVC piping, where municipal water is the water source. Additionally, various pipe lengths produced quantifiable levels of VC, with pipe age having no deterministic effect. The positive detections were from three different municipal distribution systems. Values of VC that were quantifiable ranged from 11 to 23 ng/L in water, with a mean of 16 ng/L in water and a median of 13 ng/L in water. In five homes, two samples were taken: one first flush and a second after at least one-and-a-half holdup volumes of fresh water had flushed the line. Three of the five locations were VC positive, and first flush and post-flushing samples had the same levels of VC ( p ¼ 1).
3.4.
Fig. 4 e Time course of VC in new and aged SDR11 piping with chlorinated and/or dechlorinated water. Error bars show standard error across replicates (n [ 2e3) and different experiments (n [ 2) and black triangles indicate values BDL. The LOQ for the GCeMS was 6.0 ng/L.
Field samples
Effects of chlorine on VC accumulation results
Following the trends observed in the field sampling results, laboratory tests were conducted to compare results when chlorinated versus dechlorinated tap water was used. The experiment tested both new and 25-year aged SDR11 CPVC pipe. Averages from two replicate studies are plotted in Fig. 4. At all time points, the average VC levels for pipes receiving
2612
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 0 7 e2 6 1 5
Table 1 e Results from vinyl chloride tests of tap water samples from homes with PVC/CPVC pipes installed. Groupings are based on combinations of pipe type, pipe age, pipe length, and type of water source. Grouping
1 2 3 4a 5 6 7 8 9 10a 11a 12
Pipe type
Pipe age (years)
Pipe length (ft)
Pipe diameter (inch)
Water source
# of Sample locations
# Positive for VC
# Negative for VC
PVC PVC PVC PVC PVC CPVC CPVC CPVC CPVC CPVC CPVC CPVC
<3 <3 ¼3 4e6 4e6 Unknown 4e6 7e10 11e20 11e20 >20 >20
<10 >50 >50 10e50 >50 <10 10e50 10e50 <10 10e50 10e50 10e50
>¼1 >¼1 >1 >¼1 >¼1 <1 <1 <1 <1 <1 <1 <1
Municipal water Well water Municipal water Lake water Well water Municipal water Municipal water Well water Municipal water Well water Well water Municipal water
1 1 1 1 1 1 1 1 1 3 2 1
0 0 0 0 0 1 1 0 0 0 0 1
1 1 1 1 1 0 0 1 1 2 3 0
a One sample for group 4, 10, and 11 were analysed with GC-FID which had a higher limit of quantification of 95 ng/L.
chlorinated water were higher than those receiving dechlorinated water. The results for the chlorinated water repeats were similar to those in Fig. 2b (Sch. 80 CPVC) and 3c. However, when dechlorinated water was used, the new pipe did not show detectable VC until the 101st hour. An additional sample taken after 2 months (1368 h) suggested that while much more total VC accumulated by that time, the trend was maintained: reactors with dechlorinated water having lower VC levels than their chlorinated counterparts. Also notable is that the new pipes accumulated more VC after 2 months than aged pipe. To further investigate the effects of chlorine on VC accumulation, additional tests were done with new, type L copper piping (obtained from manufacturer F) in order to eliminate leaching as a source of VC. Triplicate reactors were filled with chlorinated versus dechlorinated water. As seen in Fig. 5, all samples were below the detection limit (BDL) on the 5th and 29th hours. On the 53rd hour, one of the chlorinated triplicates was detectable but not quantifiable, while the rest of the reactors were BDL. By the 101st hour, all of the chlorinated reactors were detectable with one being quantifiable, while all of the dechlorinated reactors were BDL. Likewise, at the two month time point, all of the dechlorinated reactors were BDL, while two of the triplicate chlorinated water reactors showed detectable VC (up to 10 ng/L). Overall, the levels of VC were significantly lower than in any experiments with PVC/CPVC piping at an equivalent time point.
3.5.
Chlorine residual over time
Chlorine residual was monitored in replicate reactors (CPVC, copper pipe, and serum bottle control). Supplementary Fig. 2 shows the free chlorine residual trends in the reactors. For copper and CPVC negative control pipes filled with deionized water, the free chlorine residual concentrations were all below the detection limit of the DPD (N,N-diethyl-p-phenylediamine) method (the detection limit of the HACH kit was 0.04 mg/L using the long path viewing adapter). The residual free chlorine concentrations in the copper pipes over four days were all below detection limit, suggesting rapid consumption of disinfectant. Free chlorine was present in the
CPVC pipes at the 5th hour but was not detected afterward. For the serum bottle control, the free chlorine levels were 0.3 and 0.1 mg/L at the 5th and 29th hour, respectively and then were below detection at the 53rd (2 days) and 101st hour (4 days).
4.
Discussion
4.1. New piping and concentrations relative to other studies Concentrations in the long term study reached 300 ng/L over nearly a two year time period. An Italian study found that VC accumulated in plasticized PVC drinking water bottles to levels of around 180 ng/L after 200 days (Benfenati et al., 1991). In the current study, it took between 450 and 479 days to reach 180 ng/L (225 ng/L after 479 days). Looking at the shorter time period reactors, it is clear that VC accumulates in PVC (Sch. 40) and CPVC (Sch. 80 and SDR11) in the range of 11e25 ng/L after about four days. There were no statistically significant differences between the types of piping; however, different manufacturers of the same pipe type showed variations. For instance, in PVC, manufacturer A was greater than both B ( p-value ¼ 0.080) and C ( p-value ¼ 0.049), while in Sch. 80 CPVC, manufacturer A was greater than C ( p-value ¼ 0.060). This indicates that manufacturing processes in the United States affect the vinyl chloride concentration. Studies done under various conditions on unplasticized PVC (uPVC) produced locally in Saudi Arabia showed much higher VC levels in the mg/L range after several weeks (Al-Malack, 2004; Al-Malack et al., 1999; Al-Malack and Sheikheldin, 2001). Likewise, one Japanese study (Ando and Sayato, 1984) found that static PVC pipe reactors (unknown manufacturer) did not produce VC over a three day period, while an Italian study (Benfenati et al., 1991) found VC accumulated in plasticized PVC at a rate of 1 ng/L/day. These results highlight that manufacturing processes within the United States and abroad greatly affect VC accumulation, as there may be some factor in the production of the piping that influences residual VC monomer concentrations. This effect of manufacturing
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 0 7 e2 6 1 5
2613
Biofilm effects may be important at longer times as they might provide a barrier to VC diffusion, or could promote the degradation of VC. Conversely, they can be chlorinated by residual chlorine, leading to production of chlorinated organics that could breakdown into low molecular weight compounds including VC. Further exploration is needed.
4.4. Field and lab evidence of VC as a disinfection byproduct (DBP)
Fig. 5 e Time course of VC in new Type L copper pipe filled with either chlorinated or dechlorinated tap water. Black triangles indicate values BDL. The LOQ for the GC-MS was 6.0 ng/L.
process on VC accumulation is not surprising considering the study of Fluornoy and Monroe on pre-1977 pipes. Both in lab studies and field studies, VC accumulated in drinking water exposed to the PVC at levels well above the MCL of 2 mg/L (Flournoy and Monroe, 1999).
4.2.
Aging effects
Aging had no significant effect on the accumulation of VC in the Sch. 40 PVC and SDR11 CPVC over short term accumulation experiments, as the two-year old Sch. 40 PVC and 25-year old SDR11 CPVC showed very similar VC levels to corresponding new pipe of the same type at comparable time points over short durations on the order of days (Figures 2a/3a and Figures 3c/4, respectively). However, after two months, SDR11 CPVC showed much higher VC levels in new pipe compared to the aged pipe (new SDR11 CPVC versus aged SDR11 CPVC with biofilm after two months; p-value ¼ 1.65E-05). On the other hand, the 15 year-old Sch. 80 PVC (manufacturer D obtained from the IWTP), showed levels much greater than any other pipe with levels over 200 ng/L after 168 h. This may be a manufacturer-specific result (we could not find new pipe from the same manufacturer for additional experiments). Alternately, the result could be an effect of the rust-colored (bio) film coating the pipe (see Supplementary Fig. 1). The rust color of the biofilm could indicate iron oxides, which could serve as a catalyst for reactions that promote the accumulation of VC. It is unlikely merely an effect of pipe age given that the 25-year old SDR11 CPVC accumulated a lower VC concentration compared to the new pipe after two months ( p-value ¼ 1.65E-05), as discussed above.
4.3.
Biofilm effects
The removal of biofilms from the aged pipes was shown to slightly increase the accumulation of VC in pipe reactors for the Sch. 40 PVC ( p-value ¼ 0.100) and Sch. 80 CPVC (n ¼ 1) after several days, while the SDR11 CPVC showed higher levels when the biofilm was left intact ( p-value ¼ 0.063) (Fig. 3).
The field study observed detectable VC levels only in tap water from homes on municipal (chlorinated) water supplies. In those homes, the fact that first flush water and water after 1.5 pore-volumes showed statistically indistinguishable VC levels ( p ¼ 1.0) suggested that residual chlorine may be a contributor to the VC observations. The follow up lab experiment using chlorinated and dechlorinated water on the SDR11 CPVC with new and aged pipe (biofilm intact) supported this (Fig. 4). While VC in both types of pipe using chlorinated water was detectable after 5 h, VC in the dechlorinated reactors was detectable only after 101 h. Even after two months, chlorinated reactors were consistently higher than the dechlorinated reactors for both the new ( p-value ¼ 0.088 at two months) and aged pipe ( p-value ¼ 0.087 at two months). These results suggest that the chlorine in the tap water, and not simply the piping, may be contributing to the VC accumulation. In the analysis of the copper piping (Fig. 5), the chlorinated reactors accumulated VC at quantifiable levels after 101 h ( p-value ¼ 0.054) and two months ( p-value ¼ 0.10) compared to the dechlorinated reactors that never showed detectable VC levels. Since the use of copper eliminates leaching as a source of VC, it is evident from the copper pipe reactors that there is another source of VC accumulation when residual chlorine (0.6 mg/L total chlorine) is present. The discovery of disinfection byproducts (DBPs) in drinking water has caused public health concerns and become one of the major issues for the drinking water industry. Drinking water DBPs result from the reaction of natural organic matter (NOM) and/or bromide/iodide with disinfectants, such as chlorine or chloramines present in drinking water supplies (Bougeard et al., 2010). Although more than 500 DBPs have been reported in the literature, there are numerous DBPs that have not been identified. It has been estimated that the DBPs currently quantified in drinking water account for no more than 50% of the total organic halogen (TOX) measured in chlorinated drinking water (Richardson, 2003; Hua and Reckhow, 2008). A significant fraction of TOX formed from chlorine reactions is still of unknown structure. Results from the controlled laboratory experiments with chlorinated and dechlorinated tap water with CPVC and, especially, copper pipes provided the preliminary evidence that VC may be a DBP. Additionally, field samples with VC were all from a municipal (chlorinated) source, and VC levels were the same before and after flushing the lines in the homes. This suggests that VC produced earlier in the distribution system and transported to the home (non-plastic fittings) was a more significant contribution than leaching from the CPVC piping connected to the tap. Finally, free chlorine residual tests (Supplementary Fig. 2) suggest that VC is not likely a primary DBP, but rather, forms from a secondary reaction with more
2614
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 0 7 e2 6 1 5
complex organic matter that gets chlorinated. This secondary DBP hypothesis is supported by the time lag in VC detection seen in the copper piping only after 53 h (Fig. 5) in comparison to the quick consumption of free chlorine (<5 h; Supplementary Fig. 2). More extensive lab studies (with extensive DBP monitoring) as well as field samples from more PVC/CPVC-using systems (and non plastic systems as well) on utility distribution lines (with a chlorine residual) would allow further testing of this claim.
5.
Acknowledgments This work was supported by the American Water Works Association Research Foundation (Project #2991). Additional funding to Ryan Walter was provided through Cornell University’s Engineering Learning Initiatives program. Thanks to Kelly Krohn and Mindy Sun for help with initial experiments. We also thank the three local water treatment works (Bolton Point Water Treatment Plant; Ithaca Water Treatment Plant; and Cornell Water Filtration Plant) for their assistance with sampling and providing valuable materials and information.
Conclusions
The purpose of this research was to investigate VC levels encountered in drinking water as a result of the use of PVC/ CPVC pipe. Furthermore, we sought to determine factors affecting the accumulation of VC in PVC/CPVC piping. This research has shown accumulation of VC in PVC/CPVC piping, which is being used increasingly in water distribution systems. Although all values measured in this study are below the EPA’s MCL of 2.0 mg/L, many readings, especially at longer times, were quantifiable and, therefore, above the MCL-Goal of 0 mg/L. For new pipe manufactured in the US, long term studies in the lab suggested equilibrium concentrations of about 300 ng/L. In time frames that simulate overnight stagnancy in a home (5e8 h), VC levels varied from below detection (<6 ng/L) to 20 ng/L, with pipe age and/or manufacturer having some effect. Viewing our results in the context of other studies, considerable differences in VC accumulation can occur across different manufacturers, especially pipe manufactured outside the U.S. and before 1977. Aging effects of pipe were generally negligible, except for an anomalously high level of VC (>200 ng/L in one week) found in one aged pipe with a rust-colored biofilm. Likewise, biofilm effects were mostly small, but need to be further explored at longer times. Field studies found that VC can accumulate in tap water to detectable levels mainly in small diameter (<100 ) CPVC pipe hooked up to municipal (chlorinated) water, with pipe age and length having no obvious effect. While it has been shown that PVC and CPVC pipe reactors leach VC in short (up to 7 days) and long term (up to 2 years) studies, this research provided preliminary evidence that chlorine may also contribute to VC accumulation via DBP reactions. This is supported by the fact that CPVC reactors accumulated more VC in chlorinated reactors compared to dechlorinated reactors over time, indicating that the chlorine in the water, in addition to leaching from the plastic piping, may contribute to VC accumulation. Likewise, copper pipe tests, which eliminate VC accumulation via leaching from the plastic pipe, showed detectable VC levels in the chlorinated reactors and not the dechlorinated reactors. Free chlorine residual tests indicate that VC is more likely a secondary, not primary, DBP. Further research needs to be conducted in order to validate this trend and to elucidate reaction mechanisms. Follow up studies need to be tightly controlled with different experimental variables, such as different types of disinfectants, additional studies with metal piping, different total organic carbon levels, and various levels of chlorination, all in an attempt to tease apart the relative contributions from the plastic piping and from chlorine-dependent reactions.
Appendix. Supplementary material Supplementary data related to this article can be found online at doi:10.1016/j.watres.2011.02.016.
references
(ATSDR), Agency for Toxicity Substances and Disease Registry, 2004. Toxicological Profile for Vinyl Chloride. http://www. atsdr.cdc.gov/toxprofiles/tp20.pdf (accessed 25.04.08.). (AWWA), American Water Works Association, 2001. Water Industry Technical Action Fund Sponsored Report: Dawn of the Replacement Era: Reinvesting in Drinking Water Infrastructure. (EPA), Environmental Protection Agency, 1995. Measurement of Purgeable Organic Compounds in Water by Capillary Column Gas Chromatography/Mass Spectrometry Method 524.2, Environmental Protection Agency (EPA). (MDNS), Missouri Department of Natural Resources, 2006. Controlling vinyl chloride in drinking water distribution systems. Water Protection Program Technical Bulletin June 2006. Al-Malack, M.H., 2004. Effect of UV radiation on the migration of vinyl chloride monomer from unplasticized PVC pipes. Journal of Environmental Science and Health, Part AdToxic/ Hazardous Substances & Environmental Engineering A39 (1), 145e157. Al-Malack, M.H., Sheikheldin, S.Y., 2001. Effect of solar radiation on the migration of vinyl chloride monomer from unplasticized PVC pipes. Water Research 35 (14), 3283e3290. Al-Malack, M.H., Sheikheldin, S.Y., Fayad, N.M., Khaja, N., 1999. Effect of water quality parameters on the migration of vinyl chloride monomer from unplasticized PVC pipes. Water Air and Soil Pollution 120 (1e2), 195e208. Ando, M., Sayato, S., 1984. Studies on vinyl chloride migrating into drinking water from polyvinyl chloride pipe and reaction between vinyl chloride and chlorine. Water Research 18 (3), 315e318. Benfenati, E., Natangelo, M., Davoli, E., Fanelli, R., 1991. Migration of vinyl chloride into PVC-bottled drinking-water assessed by gas chromatography-mass spectrometry. Food and Chemical Toxicology 29 (2), 131e134. Bougeard, C.M.M., Goslan, E.H., Jefferson, B., Parsons, S.A., 2010. Comparison of the disinfection by-product formation potential of treated waters exposed to chlorine and monochloramine. Water Research 44, 729e740. Burn, S., 2005. Long Term Performance Prediction for PVC Pipe; Report Order 91092F, AwwaRF.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 0 7 e2 6 1 5
Burton, F.L., Stensel, H.D., Tchobanoglous, G., 2002. Wastewater Engineering Treatment and Reuse, fourth ed.. Metcalf and Eddy, McGraw Hill, New York City, N.Y. Flournoy, R.L., Monroe, D., 1999. Health effects from vinyl chloride monomer leaching from pre-1977 PVC pipe. Annual American Water Works Association Conference Proceedings, pp. 1211e1230. Hua, G., Reckhow, D.A., 2008. DBP formation during chlorination and chloramination: effect of reaction time, pH, dosage, and temperature. Journal American Water Works Association 100 (8), 82e95. Richardson, S.D., 2003. Disinfection by-products and other emerging contaminants in drinking water. TrAC Trends in Analytical Chemistry 22 (10), 666e684.
2615
Saeki, Y., Emura, T., 2002. Technical progresses for PVC production. Progress in Polymer Science 27 (10), 2055e2131. Sano, T., Negishi, N., Kutsuna, S., Takeuchi, K., 2001. Photocatalytic mineralization of vinyl chloride on TiO2. Journal of Molecular Catalysis A: Chemical 168 (1e2), 233e240. The Vinyl Institute and Uni-Bell PVC Pipe Association, 1994. Investigation of vinyl chloride monomer (vcm) contamination in Doniphan county, Kansas rural water District (RWD) #5. A Report for U.S. Environmental Protection Agency Office of Drinking Water Washington, DC.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 1 6 e2 6 2 6
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Anaerobically digested biosolids odor generation and pathogen indicator regrowth after dewatering Yen-Chih Chen a,*, Matthew J. Higgins b, Steven M. Beightol b, Sudhir N. Murthy c, William E. Toffey d a
Environmental Engineering Program, Penn State Harrisburg, Middletown, PA 17057, USA Department of Civil and Environmental Engineering, Bucknell University, Lewisburg, PA 17837, USA c DC Water and Sewer Authority, 5000 Overlook Avenue, S.W., Washington, DC 20032, USA d Effluential Synergies LLC, 9001 Verree Road, Philadelphia, PA 19115, USA b
article info
abstract
Article history:
The objective of this research was to investigate whether a preferential stimulation of
Received 15 April 2010
microorganisms in anaerobically digested biosolids can occur after dewatering and if it can
Received in revised form
lead to pathogen indicator regrowth and odor generation upon storage. Laboratory incu-
9 February 2011
bation simulating biosolids storage indicates that both odorant generation, based on total
Accepted 12 February 2011
volatile organic sulfur compound concentrations (TVOSCs) and pathogen indicator
Available online 19 February 2011
regrowth, based on fecal coliform densities follow similar formation and reduction patterns. The formation and reduction patterns of both odor compounds and fecal coli-
Keywords:
forms imply that groups of microorganism are induced if shearing disturbance is imposed
Biosolids
during dewatering, but a secondary stabilization can be achieved soon after 1e2 weeks of
Odor
storage. The occurrence of the induction is likely the microbial response to substrate
Fecal coliforms
release and environmental changes, such as oxygen, resulting from centrifuge shearing.
Regrowth
The new conditions favor the growth of fecal coliforms and odor producing bacteria, and
Volatile organic sulfur compounds
therefore, results in the observed fecal coliforms regrowth and odor accumulation during subsequent storage. However, when both substrate and oxygen deplete, a secondary stabilization can be achieved, and both odor and fecal coliforms density will drop. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
Biosolids reuse for land application is not only economical but also beneficial to resource recycling. Biosolids are composed of mostly organic matter, with nutrient levels at roughly 55, 30, and 1.5 g/kg for N, P, and K respectively (Wallace et al., 2009). Organic matter is one of the most important attributes of soil quality, and applications of biosolids on farmlands have shown improved soil aggregation, C and N storage, water retention, nutrient availability, and overall health of pastures (National Research Council, 1993; Wallace et al., 2009).
Anaerobic digestion is widely adopted by many wastewater treatment facilities for biosolids stabilization. In addition to solids reduction, biological activity of biosolids is greatly reduced through anaerobic digestion. However, recent reports have indicated that odor generation can occur and pathogen indicators can reactivate and regrow after dewatering, especially when centrifuge dewatering is used (Adams et al., 2004; Cheung et al., 2003; Erdal et al., 2003; Hendrickson et al., 2004; Higgins et al., 2006a, 2007; Iranpour and Cox, 2006; Qi et al., 2007). Reactivation was defined as the sudden increase of culturable bacteria immediately after dewatering, while
* Corresponding author. Tel.: þ1 717 948 6695; fax: þ1 717 948 6580. E-mail address:
[email protected] (Y.-C. Chen). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.02.014
2617
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 1 6 e2 6 2 6
regrowth was the continued growth of bacteria upon further storage of dewatered biosolids (Higgins et al., 2007; Qi et al., 2008). Though many biosolids can experience both reactivation and regrowth, regrowth was found to be independent of reactivation (Higgins et al., 2008; Qi et al., 2008), and as high as 109 cells/g DS (dry solids) of E. coli regrowth was reported (Higgins et al., 2008). Offensive odor is not currently regulated and yet drew the most concern from the public; therefore, it was recommended by the US National Research Council to be included in the future biosolids disposal standards (National Research Council, 2002). Densities of coliform indicators, on the other hand, are already regulated by the US Environmental Protection Agency, and thus regrowth of fecal coliform can potentially impact the ability to meet regulations. Though pathogen indicators are not pathogenic, their regrowth may imply growth of actual pathogens and consequently an increased health risk. This is especially true for pathogenic bacteria such as Salmonella that are closely related to E. coli, which had already been reported by some for their regrowth (Higgins et al., 2008; Zaleski et al., 2005). Biosolids odor has long been one of the major issues for the wastewater treatment utilities. Odors are typically a problem for dewatered biosolids since digested liquid biosolids usually show low to no odor (Adams et al., 2004). Major biosolid odor compounds were identified to be volatile organic sulfur compounds (VOSCs) including methanethiol (MT), dimethyl sulfide (DMS), and dimethyl disulfide (DMDS), which are all microbial degradation byproducts of sulfur containing amino acids, methionines, and cysteines (Adams et al., 2004; Higgins et al., 2006a). MT was observed to be the dominant VOSCs during biosolids storage (Higgins et al., 2006a), and can be produced by several bacteria and fungi, including E. coli and other bacteria that are known to present in biosolids (Onitake, 1938; Segal and Starkey, 1969; Weimer et al., 1999). Methanogens were also shown to be the key VOSCs degraders, and recovery of methanogensis activity during biosolids storage can lead to odor removal (Sipma et al., 2002; Higgins et al., 2006a). For long-term storage, odorous volatile aromatic compounds (OVACs) were suspected to contribute to the persistent odor that remains due to their resistance to degradation under field storage conditions (Chen et al., 2006a). Since residual proteins are the precursors for odorant production (Novak et al., 2006), methods attempting to reduce bioavailability of proteins, such as metal binding, had shown success in odor control (Chen et al., 2007). Biosolids pathogen indicator regrowth, including fecal coliforms and E. coli, is a new but yet pressing issue for the wastewater treatment utilities. Recent reports showed that stabilized liquid biosolids exhibit active microbial activity after polymer conditioning and centrifuge dewatering, and pathogen indicators can regrow to as high as just before digestion (Cheung et al., 2003; Erdal et al., 2003; Hendrickson et al., 2004; Higgins et al., 2007; Iranpour and Cox, 2006; Qi et al., 2007). Based on a recent research survey conducted by a team sponsored by the Water Environment Research Foundation, the observed regrowth seems to be associated with centrifuge dewatering (Higgins et al., 2008). Moreover, in addition to fecal coliform, pathogens such as Salmonella were also found to experience regrowth in Class B biosolids (Higgins et al., 2008). Understanding the mechanisms of how and why
pathogen indicators regrow is, therefore, an important step to future developments on regrowth control and prevention. Since biosolids odor production and pathogen indicator regrowth are both signs of microbial activity during storage, it is hypothesized that similar mechanisms are behind both phenomena. The goal of this research was to investigate the mechanisms of biosolids odor production and pathogen indicator regrowth, and observe factors that contribute to the stimulation of microbial growth from the originally stable liquid biosolids.
2.
Materials and methods
2.1.
Plant description and sample collection
Five municipal wastewater treatment plants with anaerobic digestion, operating at various digestion timeetemperature combinations and configurations, were analyzed in this research. Detailed specifications of these treatment plants can be found in Table 1. For dewatered cakes, all samples were collected from the immediate cake exit point of the centrifuge or belt filter press (BFP), and liquid biosolids were collected at the closest point to the dewatering equipment prior to polymer addition. Duplicate samples were collected aseptically with sterile containers and tools, stored immediately in either wet ice (for microbial culture), dry ice (for molecular DNA analysis), or room temperature (for odor analysis), and analyzed within 24 h of collection.
2.2.
Effect of shearing
2.2.1.
Field analysis e centrifuge vs belt filter press
Two field samples, one with centrifuge dewatering (Pre-Past-1) and one with belt filter press dewatering (Meso-1), were collected and analyzed side by side for their total volatile organic sulfur compounds (TVOSCs) production and fecal coliform regrowth. Separate and extensive work on TVOSCs production and fecal coliforms regrowth were conducted previously and the two plants of choice had shown consistent and typical TVOSCs and fecal coliform regrowth profiles based on prior experiences (Adams et al., 2007; Higgins et al., 2008).
2.2.2.
Field analysis e high solids vs low solids centrifuge
A field study was conducted at the Meso-2 plant, which was equipped with two types of centrifuges. One centrifuge,
Table 1 e Operational Data for the Sampled Wastewater Treatment Plants. Plant Pre-past-1
Temperature ( C)
1st: 66 2nd: 35 Meso-1 1st: 35 2nd: unheated Meso-2 36 Meso-3 38 Thermo-1 55
SRT
Configuration
1st: 2 h 2 stages in series 2nd: 29 d 1st: 30 d 2 stages in series 2nd: 40 d 22 d 32 d 15e20 d
Parallel Parallel CSTR in Parallel
EPA Class A/B A B
B B B
2618
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 1 6 e2 6 2 6
Humboldt, operating at 2200 rpm speed and 2.3 rpm differential, dewatered biosolids to solid contents ranging between 30 and 34%, and is termed high solids (HS) centrifuge in this report. A second centrifuge, a Centrysis modified Bird, operating at 1100 rpm speed and 1.6 rpm differential, produced lower solid contents (26e30%), and is termed low solids (LS) centrifuge in this report. Both centrifuges received the same liquid biosolids feed with identical polymer dose, and both operated at flow rates between 300 and 450 gal/min. The HS centrifuge was set to operate at higher bowl differential speed, and thus resulted in a higher level of cake shear by the internal scroll conveyer.
2.2.3.
Laboratory dewatering and shearing simulation
A laboratory experiment simulating the field belt filter press and centrifuge dewatering was used to observe the effects of shearing on biosolids TVOSCs production and fecal coliforms regrowth. Anaerobically digested liquid biosolids collected from Prepast-1 plant was used for this experiment. Three key steps were included in the laboratory protocol: polymer conditioning, dewatering, and shearing. Liquid biosolids were first conditioned to their optimum polymer dose (OPD) based on the capillary suction time (Higgins et al., 2006b). Once OPD is achieved, the majority of liquid water was removed by centrifuging at 3000 g for 10 min in a 250 mL Nalgene bottle, which typically results in solids contents around 15%. To increase solids content matching field cakes and to simulate the belt filter press dewatering, centrifuged pellets were further dewatered by a Crown Press, a laboratory scale belt filter press simulator (Phipps&Bird, Richmond, VA). Finally, to simulate centrifuge shear, the cake was processed through a KitchenAid meatgrinder apparatus that pushes the cake forward using a scrollconveyor (similar to the internal part of an actual full-scale centrifuge), followed by extrusion through a small opening at the end of the conveyor. Various amounts of passes were used to provide different levels of shear to the cakes. For the purpose of this research, 90 kg, 4 min of belt filter pressing was used to generate biosolids with cake solid ranging between 28 and 33%, and 0, 5, or 10 shearing passes were used to observe shearing impacts on TOVSCs production and fecal coliforms regrowth. Though the setting does not provide identical simulation to the field scale dewatering process, it allowed us to produce cakes of similar solid contents with minimal to high shear, which mimic belt filter press and centrifuge cakes, respectively.
2.3.
Effect of substrates
2.3.1.
Substrate addition to belt filter press cakes
Belt filter pressed cakes typically do not experience regrowth in storage and therefore, this test was to see if substrate is the growth limiting factor that is responsible for fecal coliforms regrowth. To accomplish this, belt filter press dewatered cakes collected from Meso-1 plant was used and surface-applied with 1 mL of the stock food substrate solution (3 g/L of glucose/bactopeptone in 9:1, filter-sterilized) in each 30 g of cakes. Since mixing would introduce undesired shearing, the containers were rolled around for about 1 min to provide needed mixing with minimal shear. An identical amount of de-ionized (DI)
water was spiked in another 30 g of cakes to serve as the control. Both treatment and control were incubated at 35 C for 24 h, and fecal coliform densities were analyzed at time 0 and 24 h.
2.3.2. Substrate addition and shearing to stabilized liquid biosolids Stabilized digested liquid biosolids from Meso-1 plant was also collected and used to evaluate both shearing and substrate effects on fecal coliforms regrowth. Exactly 10 mL of the stock substrate solution or sterile DI water was incorporated into each 300 mL of treatment and control biosolids. Samples were immediately blended for 2 min at high speed in a Waring Blender (EPA Method 1680) and analyzed for their fecal coliform densities by multiple tubes e most probable number (MPN) analysis. The left-over blended samples were stored in the blender jar at 35 C for 24 h to test for potential impact on regrowth. After 24 h of storage, the samples were subjected to a quick 10 s blending before dilutions and MPN analysis.
2.3.3.
Methionine addition to sheared and un-sheared cakes
Methionine is a known TVOSCs precursor (Higgins et al., 2006a) and thus was used to evaluate the precursor effect on odor generation. For methionine addition tests, 0.005 mmol of methionine was introduced into each 10 g of belt filter press dewatered cakes (Meso-1) through 1 mL of pH 7.2, 50 mM phosphate buffer, with the same mixing process as the substrate tests. The same 1 mL phosphate buffer without methionine was applied to cakes to serve as the control. Both un-sheared BFP cakes and BFP cakes with 5-pass shearing were spiked with either methionine or phosphate buffer and compared side-byside for their TVOSCs generation.
2.4.
Bound protein extraction and quantification
Bound proteins are defined here as the fraction of proteins in biosolids that is easily extracted and thus are considered bioavailable to microbial degradation for odor production (Higgins et al., 2006a). Bound proteins were collected through extractions of 10 g biosolids with 100 mL of pH 8, 50 mM phosphate buffer, followed by centrifugation at 3000 g for 10 min, and filtration through a 4 mm glass membrane (Higgins et al., 2006a). The recovered filtrates were quantified for their protein concentrations using the Bio-Rad RC DC Protein Assay Kit (Hercules, CA) with a SpectraMax M2 Microplate Reader (Molecular Device, Sunnyvale, CA). Triplicate analyses were conducted and concentrations were expressed as relative to the known bovine serum albumin (BSA) standards.
2.5.
TVOSCs and methane profile analysis
Methane and volatile organic sulfur compounds (VOSCs), including MT, DMS, and DMDS, were measured and monitored daily using a standard laboratory procedure for odor profile analysis, which was described in detail previously (Higgins et al., 2006a). In brief, each 10 g of dewatered biosolids were incubated in a sealed 160 mL serum bottle at 25 C. Target compounds were monitored daily from the head-space for its methane and VOSCs concentrations using a HP GC-FID
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 1 6 e2 6 2 6
until VOSCs decreased to below the detection limits, which typically took around 20 days. Total VOSCs (TVOSCs), which is the sum of the sulfide-weighted MT, DMS, and DMDS concentrations (1MT þ 1 DMS þ 2 DMDS), were used for the purpose of this research to represent total odor production. For each treatment, duplicate bottles of samples were prepared and analyzed.
2.6.
Fecal coliforms/E. coli regrowth and enumeration
For regrowth experiments, each 30 g of biosolids were transferred to separate sterile containers and stored at 35 C to allow for bacterial growth. Sacrifice biosolids samples were monitored periodically using the USEPA Method 1680 (USEPA, 2005) to determine fecal coliforms densities. In brief, serially diluted samples were enriched in lauryl tryptose broth (LTB), and incubated at 35 C for up to 48 h. Positives tubes were transferred to EC broth and incubated in water bath at 44.5 C for 24 h, where gas production indicates presence of fecal coliforms. E. coli were enumerated by membrane filtration as described in Standard Method 9222 (Standard Methods, 1998) with m-ColiBlue24 medium (HACH Company, Loveland, Colorado) and incubated at 35 C for 24 h. In most cases, samples were analyzed at days 0, 1, 2, 4, and 7, which were found sufficient to demonstrate their regrowth profile. Determination of most probable numbers of fecal coliforms was done by using the MPN calculation software available at the USEPA website.
2.7. E. coli and total microbial DNA extraction and quantification A solvent based DNA extraction method providing high DNA recoveries was used to better represent total microbial contents in biosolids (Chen et al., 2006b). Total DNA was determined by staining extracted DNA with the Quant-iTTM PicoGreen (Invitrogen, Carlsbad, CA) and quantified by a SpectraMax M2 Fluorescent Microplate Reader (Molecular Device, Sunnyvale, CA). Triplicate samples were analyzed along with known concentrations of Lambda DNA as the reference standards. E. coli was also quantified for their total DNA copies through a previously developed real-time polymerase chained reaction (rt-PCR) protocol (Chen et al., 2006b). The detection limit for this protocol was estimated to be around 50,000 E. coli/g dry solids. Triplicate analyses were performed and log-averages of E. coli copies are reported here.
2.8.
Data and statistical analysis
For odor analyses, duplicate sample bottles were analyzed and only average data are presented here. In general, percent deviations were below 10% of the average value. For fecal coliforms regrwoth, the 95% confidence intervals were generally within 0.5-log. The t tests were used for pairedcomparisons and significance are reported at p < 0.05. For MPN analysis, a chi-squared distribution was constructed and used to testify their differences at p < 0.05 (Haas et al., 1999). An Excel spreadsheet was developed specifically for this statistical analysis and is available upon request.
3.
2619
Results and discussion
3.1. Effects of shearing on biosolids odor generation and coliform regrowth 3.1.1.
Centrifuge vs belt filter press (BFP) dewatering
In recent years, biosolids odor production has been identified to be the results of post-dewatering residual microbial activity, which generates gaseous malodor through protein degradation (Higgins et al., 2006a; Novak et al., 2006). Recently, the observed regrowth of pathogen indicators after biosolids dewatering has raised concerns by many, which is undoubtedly also a phenomenon of residual microbial activities (Higgins et al., 2007; Iranpour and Cox, 2006; Qi et al., 2007). Fig. 1A presents a composite odor and regrowth profile of a dewatered biosolids sample from a wastewater treatment plant (Pre-past-1) using pre-pasteurization/mesophilic anaerobic digestion followed by centrifuge dewatering. The results indicate that both TVOSCs production and fecal coliforms regrowth have a similar trend, in that they both increased soon after dewatering, peaked after a few days, and decreased to low levels within two weeks. A similar trend was also observed in centrifuge dewatered cakes with mesophilic, thermophilic, and pre-pasturization operations (Chen et al., 2008). A second biosolids sample obtained from another wastewater treatment plant using mesophilic anaerobic digestion (Meso-1) followed by belt filter press (BFP) dewatering, on the other hand, did not show any odor accumulation or display any fecal coliforms regrowth (Fig. 1B). Similar observations were reported from other mesophilic plants with BFP dewatering (data not published), but unfortunately, none of the thermophilic or pre-pasteurization plants that were sampled used BFP dewatering and thus cannot provide further verification. However, based on current available results, all evidence seems to indicate that centrifuge dewatering is the key factor in TVOSCs generation and fecal coliforms regrowth, which is independent from digestion types. Centrifuge dewatering produces biosolids with high solid contents, which provides savings on post-disposal handling for the utilities. However, shearing of biosolids imposed during centrifuge dewatering has been reported to result in elevated odor accumulation during storage (Erdal et al., 2008), and now a similar problem seems to occur with regrowth of pathogen indicators.
3.1.2.
Field high solids vs low solids centrifuges
To observe the effect of shearing on odor production and fecal coliforms regrowth, two cakes from the same digester (Meso-2) but dewatered by two separate centrifuges of various shearing stresses were analyzed. Meso-2 plant is equipped with two types of centrifuges, one operated at 2200 rpm with 2.3 rpm bowl differential which corresponded to more shear and higher solids content (HS). The second centrifuge operated at 1100 rpm with 1.6 rpm bowl differential and corresponded to less shear and lower solids content (LS). Solids contents for the HS and LS cakes collected were 33% and 29%, respectively. Note that, at the time of this experiment, the laboratory was unable to perform fecal coliform analysis and thus only E. coli were enumerated, although research has shown that the majority of
2620
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 1 6 e2 6 2 6
200
7.0
100
6.0
0 2
4
6
8
10
8
60
6
40
4
20
2
0
5.0 0
TVOSCs fecal coliform
0 0
12
10
8.0
Fecal Coliform Density (MPN/g DS)
300
10
B
80
10
9.0
TVOSCs (ppmv)
fecal coliform
Mean log
TVOSCs
400 TVOSCs (ppmv)
100
10
A
Mean log Fecal Coliform Density (MPN/g DS)
500
1
2
3
4
5
6
7
8
Storage Time (d)
Storage Time (d)
Fig. 1 e TVOSCs production and fecal coliforms regrowth profiles of A) centrifuge dewatered biosolids, and B) belt filter press dewatered biosolids at a storage temperature of 25 C (Plant A: pre-pasteurization/mesophilic anaerobic digestion; Plant B: mesophilic anaerobic digestion).
biosolids fecal coliforms are E. coli and regrowth patterns of the two are comparable (Higgins et al., 2006b). Fig. 2 showed that, when cakes were subjected to high shearing, more TVOSCs were produced compared to cakes that were subjected to low shearing. At peak days, approximately twice the TVOCs concentration and E. coli density were observed from the HS cake compared to the LS cake (note that E. coli is expressed in log scale).This increase is significant on TVOSCs production ( p < 0.05), but only marginal on E. coli ( p ¼ 0.07). Overall, there was not a clear separation between the regrowth curves of HS 600 E. coli (HS) E. coli (LS)
500
8 TVOSC (HS) TVOSC (LS)
400
6 300
TVOSCs (ppmv)
4
10
Mean log E. coli Density (CFU/g DS)
10
200 2
100
0
0 0
5
10
15
20
25
Incubation Time (d)
Fig. 2 e TVOSCs production and E. coli regrowth profiles of two centrifuge dewatered cakes from a mesophilic anaerobic digestion plant. The plant is equipped with two centrifuges, one produces high solids cakes (HS), and the other produces low solids cakes (LS). Reprinted with permission from Proceedings of Water Environment Federation Specialty Conference, Residuals and Biosolid 2008, March 30eApril 2 2008,Phildelphia, Pennsylvania. Copyrightª 2008 Water Environment Federation, Alexandria, Virginia.
and LS biosolids. This was likely due to the fact that the level of difference was within 1-log between the two samples. While the level of increase was sufficient for odor comparison, it was within the typical analytical variances for microbial enumeration to show any significance. On the timing wise, the odor and regrowth peaks were not right on top of each other. This is likely because 1) E. coli are just one of the many odor producers in biosolids. Therefore, their growth may indicate odor production but the peak odor is the collective effect of all odorproducing bacteria, and 2) several substrates released during biosolids dewatering can support E. coli growth, but proteins may not be the first priority substrate. Therefore, their degradation and odor production may not match the growth of E. coli. The LS cakes also produced a higher amount of methane than the HS cakes (Fig. 3), indicating that shearing negatively impacts methanogenic activities, which is consistent with prior reports (Chen et al., 2005; Higgins et al., 2006a). The experiment, again, demonstrates the negative impact of shearing on odor production but, however, was unable to see its impact on E. coli regrowth with the level shear increase. The impacts of centrifuge on odor production were investigated by several researchers and some mechanisms were proposed, which include 1) shearing release of proteins and substrates for microbial degradation, 2) shearing inhibition of methanogenic populations, and 3) inhibition of methanogenic population by the increase of solids content (Chen et al., 2005; Erdal et al., 2008; Higgins et al., 2006a; Kiene et al., 1986; Oremland et al., 1989; Qi et al., 2008). It is likely that similar mechanisms may also be behind the observed pathogen indicator regrowth.
3.1.3.
Laboratory simulation of centrifuge dewatering
A separate laboratory experiment simulating both belt filter press and centrifuge dewatering was also conducted on liquid biosolids collected from the Pre-past-1 plant. Fig. 4 shows the results of the laboratory shearing test, where 0 pass was believed to be similar to the belt filter press process. The pressed cakes were then subjected to 5 and 10 conveying
2621
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 1 6 e2 6 2 6
2 10
5
Methane (ppmv)
HS LS
1.5 10
5
1 10
5
5 10
4
0 0
2
4
6
8
10
12
14
Incubation Time (d)
Fig. 3 e Methane production profiles of two centrifuge dewatered cakes from a mesophilic anaerobic digestion plant. The plant is equipped with two centrifuges, one produces high solids cakes (HS), and the other produces low solids cakes (LS).
coliforms regrowth. This is similar to the field experiment where only less than 1-log of increase was observed, and therefore, the impact from additional 5 passes seemed minimal. The 0 pass cake also showed a low level of fecal coliforms regrowth (up to 3-log) without producing any observable TVOSCs. This is possibly due to the fact that the level of regrowth was not as high as those typically observed on odor producing cakes; however, additional work is needed in order to verify this hypothesis. Increased shear also resulted in decreased methane production (Fig. 5), indicating perturbations on the methanogenic population. Methanogens are known TVOSC degraders and damages of their population would result in increased TVOSC accumulation (Higgins et al., 2006a). Although a previous report has indicated potential impacts of solids content on fecal coliform regrowth (Qi et al., 2008), the solid content of all samples in this experiment were identical and thus was not a factor in this experiment. Oxygen, on the other hand, was not excluded in the process, and could potentially impact the experiment. Oxygen was previously found to induce fecal coliforms regrowth (Qi, 2008), and is inhibitive to methanogenic growth. Since oxygen can be potentially introduced during cake shearing, further work is needed to exclude its effect.
3.2. Mechanisms of shearing impact on odor generation and fecal coliforms regrowth 3.2.1.
passes to simulate the internal conveying scroll of a decanter centrifuge at two levels of shearing force. The results indicate that the more shear that is applied to biosolids, the greater the odor and regrowth during subsequent storage. Fecal coliform density of the 10-pass cakes grew up to as high as 5-log increase from the original (<8 MPN/g DS) after 2 days of storage, while TVOSCs accumulated past 1000 ppmv. Comparing to 0 pass cake, both 5 and 10 pass cakes showed significant increase on both TVOSCs and fecal coliforms ( p < 0.05). Between the two sheared cakes, though the 10-pass cake showed significant increase on TVOSCs compared to the 5-pass cake, there was no apparent difference on fecal
6
1200
A
B
0 pass 5 passes 10 passes
5
0 pass 5 passes 10 passes
1000
TVOSCs (ppmv)
4
3
2
800
600
400
10
Mean log Fecal Coliform Density (MPN/g DS)
Impact of substrates on fecal coliforms regrowth
It was previously reported that centrifuge shearing of biosolids releases soluble sulfur-containing proteins which become odor precursors in subsequent biosolids storage (Erdal et al., 2008; Higgins et al., 2006a; Novak et al., 2006). In addition to proteins, other organics are likely released during shearing and thus serve as substrates for microbial growth. An experiment was conducted to investigate the role of external food substrate on the observed regrowth phenomenon. A BFP cake was collected for the purpose of this experiment that, unlike centrifuged cake, have minimal substrate release during dewatering based on the shearing-substrate release hypothesis. Quantification of fecal coliforms was performed on day 1
1
200
0
0 0
1
2
3
4
5
Storage Time (d)
6
7
8
0
5
10
15
20
25
Storage Time (d)
Fig. 4 e A) Fecal coliform regrowth profile, and B) odor production profile of dewatered biosolids with pre-pasteurization/ mesophilic anaerobic digestion subjecting to various amounts of shearing passes.
2622
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 1 6 e2 6 2 6
5
2 10
0 pass 5 pass 10 pass
Methane (ppmv)
5
1.5 10
5
1 10
4
5 10
0 0
5
10
15
20
25
Storage Time (d)
Fig. 5 e Methane production profile of dewatered biosolids with pre-pasteurization/mesophilic anaerobic digestion subjecting to various amounts of shearing passes.
immediately after collection and subsequently analyzed on day 2 after it was spiked with 1 mL of sterile substrate (glucose/ bacto-peptone mixture) and water control for 24 h Fig. 6A shows that fecal coliform density increased almost 2 orders of magnitude in biosolids spiked with substrate compared to the water control, which indicates that substrate is the limiting factor for microbial growth in biosolids, and provision of substrates stimulates microbial growth. Similar stimulation of fecal coliforms growth was also observed when shearing is imposed to liquid biosolids, which is further amplified with the addition of substrates (Fig. 6B). In both experiments, both control samples showed slight increase in fecal coliforms, which was likely due to the unavoidable agitation, such as mixing to obtain homogeneous samples, during sample preparations at day 0. Overall, these 2 tests implied that substrates are stimulant for microbial activity in biosolids, and substrates release resulting from centrifuge shearing is likely the cause of the observed fecal coliforms regrowth. When the released substrates contain odor precursor, odor can be generated. Since a previous report has indicated that microbial consumption of centrifuge released soluble proteins and polysaccharides occurs during biosolids storage (Higgins et al., 2006a), the observed reduction of fecal coliforms at the later storage stage is, therefore, likely due to the depletion of available substrates.
3.2.2.
Impact of substrates on TVOSCs production
A separate test, using BFP cake plus laboratory shearing, was conducted to verify the role of proteins in biosolids shearing through protein extraction/quantification and methionine addition. Bound protein content, the measurement of the easily extracted soluble proteins suggestive of the bioavilable proteins, showed that biosolids contain more bioavailable proteins in sheared cake than the non-sheared control, which
also resulted in increased TVOSCs accumulation (Figs. 7 and 8). Addition of methionine, a TVOSCs precursor, also showed corresponding increase in TVOSCs production (Fig. 8). However, the level of TVOSCs increase was a lot higher in cake subjected to shear than those without shear, even though an identical amount of methionine was added in both treatments. Assuming 100% conversion of methionine into TVOSCs, a total of 766 ppmv increase in TVOSCs was expected; however, only 39 ppmv peak concentration increase was observed in cake without shear, compared to the 488 ppmv increase in the sheared cake. Clearly, in addition to releasing food substrates, changes in microbial activity had also occurred to result in this observation. Past research has also shown potential shearing impacts on microbial composition and activity, especially the methanogen population, and links to the observed odor accumulation (Adams et al., 2007; Erdal et al., 2008; Higgins et al., 2006a). Overall, the results showed that although available protein/amino acids were the source of odor during biosolids storage, changes of microbial activity through shearing can amplify the odor problem to a much larger extent.
3.2.3.
E. coli vs total microbial population
During the dewatering process, biosolids are exposed to oxygen when water is removed. Continuous flow decanter centrifuges contain an internal scroll conveyer that can break up dewatered cakes further, and therefore, the produced cakes can likely experience more oxygen exposure. Anaerobic digestion creates predominantly anaerobic bacteria, including both oxygen tolerant and non-tolerant anaerobes, which are susceptible to oxygen exposure. Oxygen non-tolerant anaerobic bacteria, which unable to produce oxidant scavengers, can even be killed when exposed to oxygen (Rolfe et al., 1978). On the other hand, fecal coliforms, being facultative anaerobic bacteria that prefer oxygen as the terminal electron acceptors, can thrive under aerobic condition. Oxygen exposure during dewatering process, therefore, creates a selective advantage toward fecal coliforms over other bacteria, which likely contributes to the observed increase of fecal coliforms during initial storage. This advantage quickly disappears as the storage condition turns anaerobic with the depletion of oxygen. Using molecular DNA techniques, it was observed that E. coli DNA increased more than one order of magnitude even though total microbial DNA decreased by 50% after storing biosolids for 2 days (Fig. 9). This result supports the hypothesis that after centrifuge dewatering and during the initial stage of biosolids storage, the overall environment favors establishment of E. coli over the bulk average population. The observed decrease in background DNA was likely the destruction of oxygen non-tolerant anaerobes. However, since E. coli is only a pathogen indicator, whether the environment also favors true pathogens needs to be verified. Though, recent reports did indicate Salmonella, also a facultative anaerobe, can follow a similar regrowth pattern if not completely destroyed during digestion (Higgins et al., 2008). On the odor perspective, it was reported that oxygen can potentially disturb methanogens, the key TVOSCs degraders in biosolids, which results in early accumulation of TVOSCs that later dissipates when oxygen depletes and methanogens reestablish (Chen et al., 2005; Higgins et al., 2006a). Overall, oxygen exposure creates selective advantage for E. coli regrowth, and a disadvantage for methanogenic odor removal.
2623
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 1 6 e2 6 2 6
12
B Mean log Fecal Coliform Density (MPN/g DS)
A
10
8
6
10
4
10
Mean log Fecal Coliform Density (MPN/g DS)
12
2
0
Time = 0 hr Time = 24 hr
10
8
6
4
2
0
24hr Cake Control
BFP Cake
Liquid Biosolids
24hr Cake w/ Food
Liquid Biosolids + Food
Fig. 6 e Induction of fecal coliform regrowth by substrates addition to A) belt filter press (BFP) dewatered cakes, and B) digested liquid biosolids from a mesophilic anaerobic plant. The liquid biosolids were also subjected to 2 min of blender shearing prior to 24 h incubation. Error bars represent 95% confidence interval and there is a statistical significance between substrates treatments and the controls (chi-distribution, p < 0.05).
Although results presented here demonstrate the high possibility of TVOSCs generation and fecal coliforms regrowth for centrifuge dewatered cake, both TVOSCs and fecal coliforms profiles indicate that a secondary stabilization can occur with lengthened storage time, which can potentially be used for both odor and regrowth control. Storage temperature, however, has been shown to impact the length of time required for complete odor removal, in which longer time is required at lower temperature (Chen et al., 2005). A similar
temperature impact was also observed for pathogen indicator regrowth (data not shown). Therefore, storage temperature should be carefully taken into consideration if the utility is to rely on additional storage time for both odor and fecal coliforms destruction. In summary, based on the current results, the authors believe that biosolids subjecting to high shearing during dewatering can release substrates, including proteins. The released substrates can provide fecal coliforms regrowth, while the released proteins can be degraded to form TVOSCs odor. Shearing possibly also introduced oxygen and thus impacts
1000 0 Pass Control 0 Pass + Methionine 5 Passes Control
800
5 Passes + Methionine
10
TVOSCs (ppmv)
Bound Protein (mg/g DS)
15
5
600
400
200
0 0 pass
5 passes
Number of Shearing Passes
Fig. 7 e Bound protein concentrations of BFP dewatered cakes from a mesophilic anaerobic plant, and after subjecting to 5 laboratory shearing passes. Error bars represent 1 standard deviation (n [ 3) and there is a statistical significance between the two treatments (t-test, p < 0.05).
0 0
2
4
6
8
10
12
Incubation Time (d)
Fig. 8 e Total volatile organic sulfur compounds (TVOSCs) production profiles of BFP cakes from a mesophilic anaerobic plant, BFP cake with 5 shearing passes, and cakes with 0.005 mmol methionine addition per 10 g cakes.
2624
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 1 6 e2 6 2 6
5000
10 Total DNA 8
DNA (µg/g DS)
10
Mean log E. coli Density (copies/g DS)
E. coli DNA
4000
3000
6
2000
4
1000
2
0
0 Fresh Cake
Stored Cake
Fig. 9 e Total bacterial and E. coli DNA of centrifuge dewatered biosolids with mesophilic anaerobic digestion (Meso-3) and the same biosolids subjecting to 2 days of storage at room temperature. Error bars represent 1 standard deviation (n [ 3) and there is a statistical significance between fresh and store cakes on both total DNA and E. coli DNA (t-test, p < 0.05).
methanogenic activity, which consequently increased TVOSCs accumulation. In addition, oxygen provided a selective growth advantage to fecal coliforms over other obligate anaerobes, and therefore facilitated their growth. However, given enough time for biosolids storage, both oxygen and substrate will be depleted and would result in a drop in both TVOSCs and fecal coliforms.
4.
Implications of research
The results presented in this work indicate that both biosolids odor production and fecal coliforms regrowth seem to coexist, in that when high level of fecal coliforms regrowth is observed, organic sulfur odors are likely to accumulate. The results also demonstrated that centrifuge shearing of biosolids likely released odor precursor and food substrates, which resulted in the observed odor accumulation and fecal coliforms regrowth. This finding implies that attempts in preventing release, facilitating removal, or reducing availability of odor precursors and food substrates can likely control both odors and regrowth simultaneously. However, biosolids may also expose to oxygen during dewatering which can also result in increased odor and fecal coliforms growth. Prevention of biosolids from oxygen stress, therefore, can likely also mitigate the problems. One should also understand that, though closely related, both phenomena are not mutually exclusive. In other words, they do not always coexist. For example, a biosolids sample can exhibit organic sulfur odor without fecal coliforms regrowth if odor producers are not fecal coliforms or E. coli. Literatures indicate that many microorganisms are known to produce the enzyme, methionine-g-lyase, that converts methionine into MT (Segal and Starkey, 1969; Weimer et al., 1999; Coombs and Mottram, 2001), and therefore, their presence can also produce odor despite the absence of E. coli. In
addition, TVOCs can also be accumulated if only the methanogenic population is damaged, while little or no growth advantage favors fecal coliforms. This was demonstrated previously where addition of BES (2-bromoethanesulfonic acid), a methanogen inhibitor, would result in accumulation of TVOSCs (Chen et al., 2005; Qi, 2008). Similarly, if methanogens are injured at a relatively higher extent than fecal coliform, odor can prolong even after fecal coliform has achieved a secondary stabilization. Methanogens are known to be a sensitive population and their recovery from an environmental perturbation can be much longer than other microorganisms due to their slow growth, which explains observations of odor problems with certain environmental stresses (Chen et al., 2005). On the other hand, coliform regrowth can also occur without experiencing odor if the provided food substrates are not odor precursors. In such a case, coliforms can grow without generation of any odorous metabolites. Another possibility is the production of a different group of odorants, such as organic volatile aromatic compounds (OVACs), whose production profile is different from what is observed for fecal coliforms regrowth (Chen et al., 2006a). OVACs were found to produce after TVOSCs dissipate and can persist for an extended period of time. Unlike TVOSCs, OVACs are difficult to degrade and since OVACs have low threshold concentrations to human olfactory, it is likely that humans can still experience odor from OVACs residuals even after stabilization of fecal coliforms. However, since OVACs are also known metabolic byproducts of protein degradation, the aforementioned treatment approaches should also reduce their production. The results from this work showed that microorganisms in digested biosolids can still be induced if disturbed, which results in undesirable odor and fecal coliform growth. Therefore, minimal agitation is recommended when handling biosolids in order to maintain their stability. For example, choosing a low shearing dewatering method will allow reduction of odor and coliform growth. Furthermore, conveyance, loading/unloading, spreading, and incorporation of bioslids are all operations that can potentially introduce shear to the biosolids, and therefore precaution should be taken to minimize shear. Biosolids odor has always been one of the major issues for wastewater utilities and was identified to be the principal complaints from the public based on a report released by the National Research Council (National Research Council, 2002). Although there is no current regulation on biosolids odor, it was highly recommended by the NRC based on civilian concerns. Fecal coliforms density, on the other hand, is one of the principal limits for biosolids land application. For Class A biosolids, since regulations stipulate that biosolids should contain less than 1000 MPN fecal coliform/g DS, occurrence of regrowth will likely jeopardize their reuse. However, research showed that no pathogens were recovered despite regrowth of fecal coliforms in Class A biosolids, and thus pursuit of an alternative pathogen indicator was recommended (Higgins et al., 2008). For Class B biosolids, wastewater utilities are not required to perform routine monitoring of fecal coliforms as long as a PSRP (Processes to Significantly Reduce Pathogens) treatment process is used for stabilization (USEPA, 1994). Therefore, despite potentially exceeding Class B fecal coliform
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 1 6 e2 6 2 6
limit and likely experiencing regrowth of pathogenic Salmonella (Higgins et al., 2008), biosolids can still be considered as Class B for disposal and reuse. This does not necessarily indicate any danger to the public, because it was the intention of Class B designation to allow low level of pathogens to remain in biosolids that do not pose any immediate threat to the public health (USEPA, 2003). The required site restriction after land application allows time for pathogen to die-off in soil, where native soil microorganisms are predominant.Future efforts should, therefore, be focused on continued monitoring of pathogen die-off during, and beyond site-restriction, especially when biosolids are again disturbed by farming activities.
5.
Conclusions
Overall, both odor production and fecal coliforms regrowth during biosolids storage are indicators of residual microbial activity. Microorganisms in stabilized biosolids can still be induced upon centrifuge dewatering, which releases food substrates. Oxygen stress also likely contributes to both phenomena where it increases odor through suppressing methanogenic activity, and increases fecal coliforms regrowth by providing an environment benefitting their growth. As substrates and oxygen deplete, both odor and fecal coliforms density gradually decrease until a new stabilization is reached.
Acknowledgments This research was funded by the Water Environmental Research Foundation and the Mid-Atlantic Biosolids Association. The authors also like to express their appreciations for the assistance provided from the associated treatment facilities.
references
Adams, G.M., Witherspoon, J.R., Card, T., Erdal, Z., Forbes, R., Geselbracht, J., Glindemann, D., Hargreaves, R., Hentz, L., Higgins, M.J., McEwen, D., Murthy, S.N., 2004. Identifying and Controlling the Municipal Wastewater Odor Environment Phase 2: Impacts of In-plant Operational Parameters on Biosolids Odor Quality Water Environment Research Foundation Report No. 00-HHE-5T, Alexandria, VA, USA. Adams, G.M., Witherspoon, J.R., Erdal, Z.K., Forbes, R.H., Hargreaves, J.R., Higgins, M.J., McEwen, D.W., Novak, J.T., 2007. Identifying and Controlling the Municipal Wastewater Odor Environment Phase 3: Biosolids Processing Modifications for Cake Odor Reduction Water Environment Research Foundation, Report No.03-CTS-9T, Alexandria, VA. Chen, Y., Higgins, M.J., Maas, N.A., Murthy, S.N., Toffey, W.E., Foster, D.J., 2005. Roles of methanogens on volatile organic sulfur compound production in anaerobically digested wastewater biosolids. Water Science and Technology 52 (1e2), 67e72. Chen, Y., Higgins, M.J., Murthy, S.N., Maas, N.A., Covert, K.J., Toffey, W.E., 2006a. Mechanisms for the production of odorous volatile aromatic compounds in wastewater biosolids. Journal of Residual Science & Technology 3 (4), 193e202.
2625
Chen, Y., Higgins, M.J., Maas, N.A., Murthy, S.N., 2006b. DNA extraction and E. coli quantification of anaerobically digested biosolids using the competitive touchdown PCR method. Water Research 40 (16), 3037e3044. Chen, Y., Adams, G., Erdal, Z., Forbes, R.H., Hargreaves, R., Higgins, M.J., Murthy, S.N., Novak, J.T., Witherspoon, J., Toffey, W.E., 2007. The effect of aluminum sulfate addition during condition on production of volatile organic sulfur compounds from anaerobically digested biosolids. Water Practice 1 (4) 1e13 (13). Chen, Y., Higgins, M.J., Beightol, S.M., Murthy, S.N., 2008. The link between odors and regrowth of fecal coliforms after dewatering. In: Proceedings of the WEF/AWWA Joint Residuals and Biosolids Management Conference. Philadelphia, PA. Cheung, K.H.M., Lang, N.L., Smith, S.R., 2003. The effects of centrifugation dewatering on Escherichia coli numbers in digested sewage sludge. In: Proceedings of the Joint CIWEM Aqua Environ. Technology Transfer 8th European Biosolids and Organic Residuals Conference. Wakefield, UK. Coombs, G.H., Mottram, J.C., 2001. Trifluoromethionine, a prodrug designed against methionine-g-lyase-containing pathogens, has efficacy in vitro and in vivo against Trichomonas vaginalis. Antimicrobial Agents and Chemotherapy 45 (6), 1743e1745. Erdal, Z.K., Mendenhall, T.C., Neely, S.K., Wagoner, D.L. Quigley, C., 2003. Implementing improvements in a North Carolina residuals management program. In: Proceedings of the WEF/ AWWA Joint Residuals and Biosolids Management Conference. Baltimore, MD. Erdal, Z.K., Forbes Jr., R.H., Witherspoon, J., Adams, G., Hargreaves, R., Morton, R., Novak, J., Higgins, M., 2008. Recent findings on biosolids cake odor reductioneresults of WERF phase 3 biosolids odor research. J Environ Science Health. Part A: Toxic/Hazardous Substances Environmental Engineering 43 (13), 1575e1580. Haas, C.N., Rose, J.B., Gerba, C.P., 1999. Quantitative Microbial Risk Assessment. John Wiley & Sons, Inc, New York. Hendrickson, D.A., Denard, D., Farrell, J., Higgins, M.J. Murthy, S.N., 2004. Reactivation of fecal coliforms after anaerobic digestion and dewatering. In: Proceedings of the WEF/AWWA Joint Residuals and Biosolids Management Conference. Salt Lake City, UT. Higgins, M.J., Chen, Y., Yarosz, D.P., Murthy, S.N., Maas, N.A., Glindemann, D., Novak, J.T., 2006a. Cycling of volatile organic sulfur compounds in biosolids and its implications for odors. Water Environment Research 78 (10), 243e252. Higgins, M.J., Murthy, S.N., Chen, Y., 2006b. Understanding Factors Affecting Conditioning and Dewatering Water Environment Research Foundation Report No. 01-CTS-1, Alexandria, VA. Higgins, M.J., Chen, Y., Murthy, S.N., Maas, N.A., Hendrickson, D., 2007. Reactivation and growth of non-culturable indicator bacteria in anaerobically digested biosolids after centrifuge dewatering. Water Research 41 (3), 665e673. Higgins, M.J., Chen, Y., Murthy, S.N., Hendrickson, D., 2008. Evaluation of Bacterial Pathogen and Indicator Densities after Dewatering of Anaerobically Digested Biosolids: Phase II and III Water Environment Research Foundation Report No. 04-CTS-3T, Alexandria, VA. Iranpour, R., Cox, H.H.J., 2006. Recurrence of fecal coliforms and Salmonella species in biosolids following thermophilic anaerobic digestion. Water Environment Research 78 (9), 1005e1012. Kiene, R.P., Oremland, R.S., Catena, A., Miller, L.G., Capone, D.G., 1986. Metabolism of reduced methylated sulfur compounds in anaerobic sediments and by a pure culture of an estuarine methanogen. Applied and Environmental Microbiology 52 (5), 1037e1045. National Research Council, 2002. Biosolids Applied to Land: Advancing Standards and Practices. National Academies Press, Washington DC, USA.
2626
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 1 6 e2 6 2 6
National Research Council, 1993. Soil and Water Quality: An Agenda for Agriculture. National Academies Press, Washington DC, USA. Novak, J.T., Adams, G., Chen, Y., Erdal, Z., Forbe, R.H., Glindemann, D., Hargreaves, J.R., Hentz, L., Higgins, M.J., Murthy, S.N., Witherspoon, J., 2006. Generation pattern of sulfur containing gases from anaerobically digested sludge cakes. Water Environment Research 78 (8), 821e827. Onitake, J., 1938. On the formation of methylmercaptan from L-cystine and L-methionine by bacteria. Journal of the Osaka Medical Association 37, 263e270. Oremland, R.S., Kiene, R.P., Mathrani, I.M., Whiticar, M.J., Boone, D.R., 1989. Description of an estuarine methylotrophic methanogen which grows on dimethyl sulfide. Applied and Environmental Microbiology 55 (4), 994e1002. Qi, Y., Dentel, S.K., Herson, D.S., 2007. Increases in fecal coliform bacteria resulting from centrifugal dewatering of digested biosolids. Water Research 41 (3), 571e580. Qi, Y., Dentel, S.K., Herson, D.S., 2008. Effect of total solids on fecal coliform regrowth in anaerobically digested biosolids. Water Research 42 (14), 3817e3825. Qi, Y., 2008. Effect of centrifugal dewatering on the regrowth of fecal coliforms and Salmonella in anaerobically digested biosolids. Ph.D. dissertation. University of Delaware. UMI Number: 3329790. Rolfe, R.D., Hentges, D.J., Campbell, B.J., Barrett, J.T., 1978. Factors related to the oxygen tolerance of anaerobic bacteria. Applied & Environmental Microbiology 36 (2), 306e313. Standard Methods, 1998. Standard Methods for Examination of Water and Wastewater, twentieth ed.. American Public Health
Association/American Water Works Association/Water Environment Federation, Washington, DC, USA. Segal, W., Starkey, R.L., 1969. Microbial decomposition of methionine and identity of the resulting sulfur products. Journal of Bacteriology 98 (3), 908e913. Sipma, J., van Bree, R., Janssen, A.J., Arena, B., Hulshoff, P.L., Lettinga, G., 2002. Degradation of methanethiol in a continuously operated upflow anaerobic sludge-blanket reactor. Water Environment Research 74 (3), 264e271. USEPA, 1994. A Plain English Guide to the EPA Part 503 Biosolids Rule EPA/832/R-93/003. Washington, DC. USEPA, 2003. Environmental Regulations and Technology: Control of Pathogens and Vector Attraction in Sewage Sludge EPA/625/ R-92/013. Cincinnati, OH. USEPA, 2005. Method 1680: Fecal Coliforms in Sewage Sludge (Biosolids) by Multiple-tube Fermentation Using Lauryl Tryptose Broth (LTB) and EC Medium EPA/821/R-04/026. Washington, DC. Wallace, B.M., Krzic, M., Forge, T.A., Broersma, K., Newman, R.F., 2009. Biosolids increase soil aggregation and protection of soil carbon five years after application on a crested wheatgrass pasture. Journal of Environmental Quality 38 (1), 291e298. Weimer, B., Seefeldt, K., Dias, B., 1999. Sulfur metabolism in bacteria associated with cheese. Antonie van Leeuwenhoek 76, 247e261. Zaleski, K.J., Josephson, K.L., Gerba, C.P., Pepper, I.L., 2005. Potential regrowth and recolonization of Salmonellae and indicators in biosolids and biosolid-amended soil. Applied and Environmental Microbiology 71 (7), 3701e3708.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 2 7 e2 6 3 7
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Fouling indices for low pressure hollow fiber membrane performance assessment Anh H. Nguyen a, John E. Tobiason b,*, Kerry J. Howe c a
Siemens Water Technologies, Memcor Australia, South Windsor, NSW 2756, Australia Department of Civil and Environmental Engineering, University of Massachusetts Amherst, MA 01003, USA c Department of Civil Engineering, University of New Mexico, Albuquerque, NM 87131-1351, USA b
article info
abstract
Article history:
This study evaluated the use of fouling indices to describe low pressure membrane fouling.
Received 5 November 2010
One critical aspect of this study was the use of a bench-scale hollow fiber membrane
Received in revised form
system that imitated full-scale operation (constant flux with automatic hydraulic back-
13 February 2011
wash and chemical cleaning). Fouling indices were based on a resistance-in-series model.
Accepted 14 February 2011
Two different hollow fiber membrane types (membrane A and B) were tested with water
Available online 22 February 2011
from two water utilities (A and B) and three other natural sources (oligotrophic, algal bloom impacted, and wastewater impaired). The bench-scale testing included use of the same
Keywords:
membrane as utilized at Utility B. Most fouling was reversible by hydraulic backwash and
Low pressure
chemical cleaning. Specific flux and fouling indices for the bench-scale system were higher
Constant flux
than those determined from full-scale data but fouling index ratios were comparable,
Fouling index
suggesting a similar fouling nature. At similar organic loading, fouling was specific to water
Backwash
source and membrane type, i.e., no generalization on the impact of water source was
Chemical clean
possible. Full-scale data were compared with bench-scale data to validate the use of
Irreversible fouling
fouling indices. Fouling indices based on a resistance-in-series are useful tools to describe membrane performance data for both raw and pretreated water, for different water sources, and different membrane types. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
A significant technical challenge for membrane processes is membrane fouling, manifested at full-scale by the increase in required operating pressure to maintain a targeted water production rate. Fouling includes the short term and reversible increase in transmembrane pressure (TMP) due to the accumulation of rejected materials that cannot be avoided in the dead-end mode operation of low pressure (LP) membranes. Irreversible fouling, hydraulic and chemical, is the longer-term loss of permeability not recoverable after hydraulic backwash (BW) or chemical cleaning (CC).
Although different fouling types (reversible versus irreversible, hydraulic versus chemical) are well defined in the literature, results reported from bench-scale fouling studies conducted in laboratories have generally not reported different fouling types. These studies were often conducted at conditions not typical of full-scale practice (not including hydraulic BW and/or CC), use of flat-sheet membranes at constant pressure operating mode instead of using hollow fibers (HF) at constant flux operating mode, or use of a crossflow system instead of dead-end as used at full-scale (Crozes et al. (1997), Tarabara et al. (2002), and Howe et al. (2007)). Due to the complex nature of membrane fouling, changing
* Corresponding author. Department of Civil and Environmental Engineering, University of Massachusetts Amherst, 130 Natural Resources Road, Marston Hall, Amherst, MA 01003, USA. Tel.: þ1 413 545 5397; fax: þ1 413 545 2202. E-mail address:
[email protected] (J.E. Tobiason). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.02.020
2628
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 2 7 e2 6 3 7
one component of a membrane system can drastically change membrane performance and make the bench-scale data far from applicable to realistic practice. It is therefore critical to conduct studies at controlled conditions, as well as simulating those at full-scale. Most membrane fouling by natural water sources involves natural organic matter (NOM). Inconsistent results have been reported in the literature regarding membrane fouling by NOM, possibly due to different testing conditions, with many deviations from full-scale practice. Different measures of NOM, including total organic carbon (TOC), total nitrogen (TN), dissolved organic nitrogen (DON), ultraviolet (UV) absorbance, NOM molecular weight distribution, and NOM source have been found to have certain impacts on membrane fouling but no single characteristic has been demonstrated to control fouling (e.g., Lee et al. (2004, 2006), Lozier et al. (2008)). More likely, complex interactions between characteristics of the water, membrane material, and system configuration that vary from site to site are responsible for fouling. For instance, Huang et al. (2008b) found that both particle size and stability were important in fouling. Higher TOC and DON levels have also been reported to cause more severe fouling of low-pressure membranes (Lozier et al. (2008)). Fouling indices have been developed by means of simple, short, empirical filtration tests to quantify the level or degree of membrane fouling. The silt density index (SDI) is a widely used fouling index for reverse osmosis (RO) membranes. The modified fouling index (MFI) (also used for RO membranes) has been reported to be insensitive to the presence of smaller particles and an unsatisfactory correlation with colloidal fouling has been observed for full-scale membrane installations (Boerlage et al., 1998, 2003). The MFI-UF (tested with a polyacrylonitrile 13 kDa UF membrane) was developed to account for the presence of smaller particles but at constant pressure (Boerlage et al., 1998, 2003). Many researchers have attempted to attribute membrane fouling to one of several mechanisms, including blockage of membrane pores at the membrane surface, development of a cake layer at the membrane surface, and adsorption of matter within the membrane pores, using equations summarized by Hermia (1982) and extended by numerous other researchers. This approach to fouling, known as the blocking laws, has several limitations, including that the original model was derived for constant pressure, declining flux operation, and that the model assumes that only one of these mechanisms is active at any given time in the operation cycle, whereas in full-scale operation it is likely that several fouling mechanisms might occur simultaneously. Thus, the blocking laws have not had predictive value for relating bench-scale results to full-scale operation. Recently, a group of researchers (Jacangelo et al. (2006); Lozier et al. (2008); Huang et al. (2008a,b, 2009)) proposed a unified modified fouling index for low pressure membrane performance assessment at constant flux. The original model was based on cake layer formation to determine a reversible fouling index and on intermediate pore blockage to determine an irreversible fouling index. The newer version presented in Lozier et al. (2008) and Huang et al. (2009) assumed fouling is solely caused by the formation of a cake layer. Different types of fouling indices could be calculated using different data, with or without hydraulic backwash or chemical cleaning. A
question remaining is whether bench-scale and full-scale data are comparable. This paper focuses on further development and validation of fouling indices to assess membrane performance. Fouling studies were conduced at bench-scale while simulating fullscale conditions. The fouling indices were based on a resistancein-series model. A key feature of this development is that fouling is not attributed to a specific mechanism so the model can be valid regardless of whether cake filtration, pore constriction, or some combination of fouling mechanisms is occurring. Different types of fouling were clearly defined and described. The validated fouling indices were then used to describe membrane fouling and compare performance of two different membrane types tested with three different characteristic waters normalized to similar organic loading (TOC and DON/DOC). The work is part of a recent larger study of low pressure membrane fouling (Nguyen (2010); Tobiason and Nguyen (2011)).
2.
Materials and methods
2.1.
Water sources and membrane properties
Waters from two water utilities (plant A and B) and three other natural surface water sources were used. Both raw and coagulated waters from water treatment plant A were tested. This water was colored (30e40 color units, CU) with moderate TOC (3.5 mg/L). Utility B used the same membrane (membrane A) as the one used for some of the bench-scale tests. The membrane feed water was coagulated with alum and mixed with backwash water (90%e10%, respectively). This water had moderate TOC levels (3.8 mg/L and 4.9 mg/L for the raw and membrane feed water respectively; full-scale recycle may have caused higher TOC in the membrane feed water). Three natural water sources with different NOM origins (oligotrophic, algal impacted, and wastewater impaired) but with similar NOM concentrations achieved by dilution with DI water (2 mg C/L and 1/17 DON/DOC ratio) were tested. Table 1 summarizes the quality of all water sources utilized in this study. Two commercial membranes, named membrane A and B, were utilized for the research project. While membrane A was utilized throughout this study, membrane B was used only for the experiments with 3 natural water sources. Both membranes are hollow fiber, outside-in type, and have PVdF base material. The membrane characteristics are shown in Table 2.
2.2.
Membrane module and system
Mini membrane modules were fabricated in the laboratory using the two types of hollow fiber membranes. A semi-rigid clear plastic tube 25.4 cm long and 1.3 cm diameter was used as the membrane housing. On one end of the tube the gap between the membrane fibers and the membrane housing was sealed by epoxy; this end served as the permeate collection and backwash feed side. At the other end, the fibers were potted/sealed by either injecting epoxy inside the lumen of each fiber or deadended together. This allowed one end of the membrane module to be open for better removal of solids during backwashing. This potting technique also made it
2629
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 2 7 e2 6 3 7
Table 1 e Utility A and B water quality. Water quality parameter
TOC (mg/L) UVA254 (cm1) TN (mg/L) DON (mg/L) DON/DOC pH Turbidity (NTU)
Plant A
Plant B
Oligotrophic (Quabbin Reservoir, MA)
Raw
Coag.
Raw
Mem. feed
3.5 0.07 e e e 6.15 0.24
4.5 0.077 e e e 5.58 2.17
3.8 0.13 0.265 e e 6.14 0.35
4.9 0.09 0.339 e e 6.63 1.58
2.1 0.031 0.23 0.116 0.056 6.8 0.32
Algal bloom impacted-ABI (Forge Pond, MA)
Wastewater impaired e WWI (Blackstone River, MA)
As is
Mem. feed
As is
Mem. feed
8.8 0.407 0.66 0.36 0.04 6.4 1.45
1.94 0.117 0.13 0.11 0.057 6.6 0.53
5.72 0.11 2.96 0.187 0.032 6.8 4.1
2.2 0.052 1.71 0.115 0.052 6.8 0.28
(Source: Tobiason and Nguyen. Hollow Fiber Membrane Fouling: Characterization & Role of Organic Nitrogen. ª2011 Water Research Foundation. Reprinted with permission).
easier to sample the hollow fibers for surface analyses. The fiber length varied from 18 to 25 cm (flow was adjusted accordingly to fiber length to maintain comparable fluxes). 5 fibers were used for each membrane A module and 8 fibers were used for each membrane B module to achieve the same fiber density (30%). The fiber density was determined as the ratio of the total cross sectional area of all the fibers to the cross sectional area of the module tube. The membrane system was constructed with automatic backwash and control to continuously monitor and record the transmembrane pressure (TMP). Constant suction based flux was maintained using a positive-displacement pump. Hydraulic backwashing was conducted using a separate backwashing pump washing the fibers from inside to outside. The overflow, which was less than 1% of the total feed volume, was recycled back to the system. Water quality measurements (TOC, DON, UV, pH, and turbidity) were monitored throughout the test and showed insignificant change in feed water quality.
Fouling indices can be quantified using a resistance-in-series approach to membrane fouling. In general, the pressure driven water flux (J, flow per unit area) though a low pressure membrane is described as: DP TMP ¼ mK mK
K ¼ Kmem þ Kcake þ KHIFI þ KCIFI
(1)
Table 2 e Membrane properties. Property
Membrane A
Membrane B
Membrane type Outside/inside diameter (mm) Membrane area (m2/ft2) Flow pattern Material Pore size (mm)
Hollow fiber 1.9/0.8 0.0063/0.067 Outside-in PVdF 0.02
Hollow fiber 1.0/0.56 0.0058/0.062 Outside-in PVdF 0.05
(Source: Tobiason and Nguyen. Hollow Fiber Membrane Fouling: Characterization & Role of Organic Nitrogen. ª2011 Water Research Foundation. Reprinted with permission).
(2)
If the resistance due to fouling increases linearly with the volume of permeate produced, then Ki ¼ kiV, where ki is a rate constant for the increase in resistance, and V is specific volume (permeate volume/membrane area). Thus the total resistance can be written as follows. K ¼ Kmem þ ðkHR þ kHI þ kCI ÞV ¼ Kmem þ ktotal V
2.3. Development of fouling indices using a resistancein-series approach
J¼
Where DP or TMP is the transmembrane pressure, K is a coefficient for the resistance to flow through the membrane, and m is the water viscosity. K is the total of several possible resistances that operate in series. At any time during operation, K is the sum of the resistance of the clean membrane (Kmem), the resistance due to a cake formed on the membrane (Kcake), the resistance due to irreversible fouling not recoverable by hydraulic backwashing (KHIFI), and the resistance due to irreversible fouling not recoverable by chemical cleaning (KCIFI), as summarized in equation (2):
(3)
Where kHR, kHI, kCI are rate constants for increase in resistance due to hydraulic-reversible, hydraulic-irreversible, and chemical-irreversible fouling, respectively. Combining (1) and (3) yields, Js ¼
J 1 ¼ DP mðKmem þ ktotal VÞ
(4)
J 1 . Membrane For a new membrane, V ¼ 0 so ð Þ0 ¼ DP mKmem performance can be normalized by dividing J/DP at any specific volume by the initial (clean membrane) condition as follows J0s ¼
ðJ=DPÞV ¼ ðJ=DPÞ0
1 1 ktotal V or 0 ¼ 1 þ ktotal Kmem Js V 1þ Kmem
(5)
Membrane performance data for different operational cycles and cleaning procedures can be used to observe and possibly quantify various fouling indices: the total fouling index (TFI), the hydraulic-irreversible fouling index (HIFI), and the chemical-irreversible fouling index (CIFI). The indices have units of inverse length and can be described as follows.
2630
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 2 7 e2 6 3 7
Average values for all data for a series of CC cycles can be used to determine the CIFI; i.e., the CIFI was determined based on data from one complete multiple CC cycle experiment.
The total fouling index is the sum of all the indices, TFI ¼ HRFI þ HIFI þ CIFI (HRFI is the fouling reversible by hydraulic backwash). If the rate of increase in resistance is linearly proportional to V, i.e., a plot of (1=J0s ) versus (V) data is linear, then the fouling indices can be quantified using linear regression. However, the rate of resistance increase might be a non-linear function of V, which means that fouling does not depend linearly on the specific volume through the membrane. In that case, the fouling indices can be determined based on a 2point method; i.e., instead of using all performance data, the first and the last points can be used to determine the average rate of increase in resistance. In this study, averages of the 10 first and 10 last data points (approximately 2.5 min of membrane operation at the beginning and at the end of each cycle) were used to determine the TFI (using averages of the first and last few data points reduces the impact of noisy data). For the HIFI, the average specific flux values for the first and last HBW cycles were used. For the CIFI, the average specific flux values for the first and last chemical clean cycles were used. Fig. 1 presents a schematic illustration of membrane operation and how data from each cycle are used to determine a fouling index value. Equation (5) can be presented as followed
1 ¼ 1 þ ðCIFIÞV J0s
1 ¼ 1 þ ðFIÞV J0s
For any single cycle between hydraulic backwashes, referred to as one hydraulic backwash cycle (HBW cycle), the TFI can be related to the normalized specific flux and the specific volume as: 1 ¼ 1 þ ðTFIÞV J0s
(6)
Although equation (6) has the same mathematical form as the unified membrane fouling index developed by Huang and coworkers, the development of this equation has the advantage that it was not necessary to assume that fouling was caused by a particular mechanism. For multiple HBW cycles without any chemical cleaning, which is referred to as one chemical cleaning cycle or chemical enhanced backwash cycle (e.g., a CEB cycle), HIFI can similarly be related to average values of J0s during individual filtration cycles and V as: 1 ¼ 1 þ ðHIFIÞV J0s
(7)
(8)
(9)
Fig. 1 e Operating cycle and fouling index determination: A e Operating cycle illustration; B e FI calculation using linear regression; C e FI calculation using 2-point method.
2631
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 2 7 e2 6 3 7
Table 3 e Operating conditions at bench- and full-scale. Exp.
Flux (GFD/LMH)
Backwash conditions
Chemical cleaning conditions
Utility A e bench-scale Utility B e bench-scale
40/68
Utility B e full-scale
21/36 or lower
Every 15 min for 30 s with air & permeate at 25 to 42 GFD (42e72 LMH)
Natural water sources e bench -scale
30/51
Every 30 min for 1 min at 90 GFD (153 MLH) with permeate
If the flux is constant, the left hand side of equation (9) can be written as: 1 Js0 J0 =DP0 DPs ¼ ¼ ¼ Js J=DPs DP0 J0s
(10)
where J0 is the flux for filtering DI water, J is the flux for filtering feed water, DP0 is the transmembrane pressure needed to obtain flux J0, and DPs is the transmembrane pressure needed to obtain flux Js. Combining equations (9) and (10) yields: DPs ¼ 1 þ ðFIÞV or DPs ¼ DP0 þ ðFIÞðVÞðDP0 Þ DP0
(11)
In other words, the operating transmembrane pressure is the sum of the initial transmembrane pressure for the clean membrane and the pressure needed to overcome membrane fouling after a certain specific volume, V, is filtered. Equation 11 is analogous to head loss in granular media filtration caused by initial, or clean-bed, head loss of the media plus head loss caused by deposited particles. Knowing the initial DP0 and the fouling index, the pressure needed to filter water over time can be calculated. For example, an average HIFI can be used to calculate the expected volume filtered between chemical enhanced backwashes (i.e., when the TMP after BW is too high) while an average CIFI could be used to estimate permeate volume prior to the need for chemical cleanin-place. Assessment and reporting of standard, non-proprietary, fouling index values as developed in this study should aid in increasing the body of knowledge concerning fouling of low pressure membranes in drinking water treatment.
2.4.
Membrane operating conditions
Different BW cycle times and fluxes were tested. For experiments conducted with water collected from Utility A, the flux was kept constant at 68 L/square meter/hour (LMH) or 40 gallons/square foot/day (GFD). Hydraulic backwash was conducted with permeate every 30 min for 1 min at 204 LMH (120 GFD). No chemical cleaning was conducted. For experiments conducted with water collected from Utility B, the testing
Mem used
None conducted
7h
A
Every 24 h for 20 min with 10 ppm NaOCl, and 20 min with DI waters at room temperature (20 C) CC: 20 min with 10 ppm NaOCl; CIP every 1.5 month or at TMP of 9 psi Every 24 h for 20 min with 100 ppm NaOCl, and 20 min with DI waters at room temperature (20 C)
4 days
A
Data collected for 1 year
A
3e7 days
A&B
conditions at bench-scale were kept as similar as possible to those used at the full-scale plant (36 LMH (21 GFD)) constant flux, 15 min filtration, 30 s backwashing, daily chemical maintenance wash with 10 mg/L chlorine. At full-scale, air scouring was conducted before hydraulic backwash with permeate. Flux and BW flowrate fluctuated daily and seasonally to follow water demands. Chemical clean-in-place (CIP) was conducted every 1.5 months or when the TMP reached 9 psi, referred to as one CIP cycle. For the experiments conducted with the three natural water sources, flux was kept constant at 51 LMH (30 GFD). Hydraulic backwash was conducted every 30 min with permeate at 3 times the filtration flowrate. No air scouring was conducted. Chemical cleaning (CC) was conducted periodically at room temperature. Chemical cleaning involved 20 min backflushing with 100 mg/L chlorine solution (pH 9.9) and 20 min backflushing with deionized (DI) water. Table 3 summarizes all the testing conditions.
3.
Results and discussion
3.1.
Bench-scale results for water from Utility A
Significant fouling occurred for both raw and coagulated (alum) feed waters. Over the duration of the experiment Specific Flux (LMH/kPa)
21/36
Every 30 min for 1 min at 120 GFD (204 MLH) with permeate Every 15 min for 30 s at 25 GFD (42 LMH) with permeate
Length
12
Coagulated Raw
8
4 Monitor DI
0
0
100
200 Specific Volume (L/m )
300
400
Fig. 2 e Specific flux versus specific volume for raw and coagulated water e Utility A. (Source: Tobiason and Nguyen. Hollow Fiber Membrane Fouling: Characterization & Role of Organic Nitrogen. ª2011 Water Research Foundation. Reprinted with permission).
2632
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 2 7 e2 6 3 7
Fig. 3 e HIFI of raw and coagulated water from WTP A: left e all data; right e average points for each HBW cycle. (Source: Tobiason and Nguyen. Hollow Fiber Membrane Fouling: Characterization & Role of Organic Nitrogen. ª2011 Water Research Foundation. Reprinted with permission).
(400 L/m2 or 12 hydraulic backwash cycles), the specific flux decreased from 8 to 4 LMH/kPa for the raw water and from 6 to 4 LMH/kPa for the coagulated water (Fig. 2). The data within one hydraulic backwash (HBW) cycle were used to determine the TFI using the 2-point method. There were significant variations of TFI values between each HBW cycle, presumably due to variable effectiveness of the HBW. The raw water and coagulated average TFI values were 38.8 104 m2/L and 30.0 104 m2/L respectively. Fig. 3 presents the performance data and the hydraulicirreversible fouling index results for raw and coagulated waters. Using all data (Fig. 3A) or average data points for each HBW cycle (Fig. 3B) resulted in similar HIFI values. The coagulated water had a much lower HIFI than the raw water (0.0023 m2/L for the raw versus 0.0009 m2/L for the coagulated water), showing that coagulation reduced hydraulic-irreversible fouling. For the raw water, only 41% of the total fouling could be reversed by hydraulic backwash, versus 69% for the coagulated water. Hydraulicirreversible fouling was relatively high for water samples from Utility A.
3.2.
Bench-scale results for water from Utility B
For Utility B membrane feed water, the specific flux loss for 1 day of operation (including HBW but no CC) was quite significant (approximately 3 LMH/kPa). TFIs were determined using the 2-point method. Similar to utility A, TFI value variations were quite significant between HBW cycles with an average value of 255 104 m2/L. Fig. 4 presents performance data and the hydraulic-irreversible fouling indices (HIFI). Either linear regression of all data or of average values for each HBW cycle resulted in similar HIFI values. The HIFI values for each chemical cleaning cycle varied quite significantly, presumably due to variable effectiveness of chemical cleaning even though the chemical cleaning procedure was kept constant. The same scenario was observed for the full-scale plant (presented in the next section). Fig. 5 presents the performance data used to determine the chemical-irreversible fouling index. The error bars show the 95% confidence interval of the data. Each data point is the average value of one chemical cleaning cycle. Using data from all 4 chemical cleaning cycles resulted in a poor linear
3
3
2
1/J' s
1/J' s
2
1
1 Day 1
Day2
Day3
Day 4
Day 1
0 0
500
1000
1500
2000
2500
Specific Volume (L/m 2) 2
3000
HIFI1=0.00093±0.00003 (R = 0.77) HIFI2=0.00167±0.00006 (R2 = 0.81) HIFI3=0.0016±0.00003 (R2 = 0.78) HIFI4=0.0013±0.00005 (R2 = 0.67)
3500
Day 2
Day 3
Day 4
0 0
500
1000
1500
2000
2500
3000
3500
Specific Volume (L/m2)
HIFI1=0.00092±0.00003 (R2 = 0.99) HIFI2=0.00169±0.00006 (R2 = 0.99) HIFI3=0.00159±0.00003 (R2 = 0.99) HIFI4=0.00132±0.00005 (R2 = 0.98)
Fig. 4 e Fouling data and hydraulic-irreversible fouling indices (HIFI); bench-scale for water from Utility B: left e all data; right e average data for each HBW cycle (Source: Tobiason and Nguyen. Hollow Fiber Membrane Fouling: Characterization & Role of Organic Nitrogen. ª2011 Water Research Foundation. Reprinted with permission).
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 2 7 e2 6 3 7
2.5 2.0
1/J's
1.5 1.0 0.5 CIFIall data = 2.14E-04
CIFI2-point = 2.51E-04
2
R = 0.46 0.0 0
500
1000 1500 2000 2 Specific Volume (L/m)
2500
3000
Fig. 5 e Fouling data and chemically irreversible fouling index (Utility B, bench-scale). (Source: Tobiason and Nguyen. Hollow Fiber Membrane Fouling: Characterization & Role of Organic Nitrogen. ª2011 Water Research Foundation. Reprinted with permission).
relation between 1/Js’ and specific volume (R2 ¼ 0.46). Thus, the 2-point method was used to determine the chemicalirreversible fouling index. CIFI2-point and CIFIall-data values are relatively comparable (CIFI2-point ¼ 1.17 CIFIall-data).
3.3.
Membrane performance of full-scale Utility B
To validate the use of fouling indices in describing membrane performance, full-scale data for one year of operation from Utility B that utilized the same membrane as tested at bench-scale (membrane A) were analyzed and compared to
2633
the bench-scale results. Bench-scale and full-scale data comparisons were made for the first 4 days in July 2008 when water samples were collected and tested with the bench-scale system and for the whole year from July 2008 to June 2009. Fig. 6 presents the membrane permeability (specific flux) corrected to 20 C for one membrane cassette at full-scale Utility B. Chemical clean-in-place (CIP) occurred between groups of data (every 1.5 months or at 9 psi TMP, referred to as one CIP cycle). The permeability for the full-scale plant was lower than for the bench-scale system. The differences between membrane permeability might be due to the variations in the membrane material, changes in water quality, and/or scale-up. Also note that the full-scale membrane performance data were obtained after one year of operation whereas the bench-scale tests were conducted with new membranes. Each data point represents one HBW cycle (data were recorded every 16 min). The permeability drop was quite significant even with hydraulic backwash and daily chemical enhanced cleaning (referred as chemical cleaning in the bench-scale experiments). However, chemical CIP was quite effective in recovering the permeability loss. TFI values were determined using the 2-point method. TFI values varied significantly between each HBW cycle. Hydraulic backwash could be effective, i.e., resulting in a lower resistance, or “ineffective”, resulting in higher resistance after hydraulic backwash (negative values of TFI). Excluding the “ineffective” hydraulic backwash cycles, TFI values ranged from to 0.614 104 to 88.9 104 m2/L (averaging 31 104 m2/L) for the first 4 days in July 2008, and ranged from 0.0027 104 to 640 104 m2/L, averaging 51 104 m2/L for the whole year. While the high values of the full-scale TFI are of the same magnitude as those of the bench-scale (TFIbench-scale varied from 216 104 to 398 104 m2/L), the TFI values for full-scale
3
Permeability (LMH/kPa)
2.5
2
1.5
1
0.5
0 Jun-08 Jul-08 Jul-08 Aug- Sep-08 Oct-08 Nov08 08
Dec- Jan-09 Feb-09 Mar- Apr-09 May- Jun-09 08 09 09
Fig. 6 e Permeability of one membrane cassette from Utility B; one year of full-scale data, chemical clean-in-place (CIP) between groups of data. (Source: Tobiason and Nguyen. Hollow Fiber Membrane Fouling: Characterization & Role of Organic Nitrogen. ª2011 Water Research Foundation. Reprinted with permission).
2634
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 2 7 e2 6 3 7
CIFIJuly = 0.368E-04
4.0
R2 = 0.86 CIFIAug. = 1.16E-04 R2 = 0.73 CIFISep. = 0.617E-04 R2 = 0.92 CIFIOct. = 0.254E-04
3.0
R2 = 0.87
1/J's
CIFINov. = 0.473E-04 R2 = 0.93 CIFIDec. = 1.119E-04
2.0
R2 = 0.95 CIFIJan. = 1.591E-04 R2 = 0.96 CIFIFeb. = 2.07E-04 R2 = 0.97
1.0
CIFIMar. = 2.27E-04 R2 = 0.96 CIFIApril = 2.12E-04
0.0 0
30000
60000 90000 120000 Specific Volume (L/m2)
150000
R2 = 0.98 CIFIMay = 2.21E-04 R2 =0.97
Fig. 7 e Full-scale fouling data and CIFI for 1 year of data. (Source: Tobiason and Nguyen. Hollow Fiber Membrane Fouling: Characterization & Role of Organic Nitrogen. ª2011 Water Research Foundation. Reprinted with permission).
were in general significantly lower than those measured at bench-scale. One of the main reasons is likely to be the impact of air scouring on the effectiveness of hydraulic backwash at full-scale, which was not conducted at bench-scale. Fig. 7 presents the fouling data and CIFI for all 12 CIP cycles from July 2008 to June 2009. As the fit between resistance 1/Js’ and specific volume V to a linear plot was very good, all data points were utilized to determine the CIFI instead of using average values for each chemical cleaning cycle. Similar to TFI and HIFI, the CIFI values varied significantly for the first 7 CIP cycles which lasted for 7 months. After January 2009, membrane performance seemed to reach a steady state with CIFI values averaging 2.1 104 m2/L. The fit between inverse specific flux and specific volume to a linear plot was excellent for all the CIP cycles, suggesting that fouling indices developed based on a resistance-in-series model could be used to describe chemical-irreversible fouling.
Table 4 compares the average values of permeability and fouling indices for the bench-scale and full-scale systems. Comparisons were made for the first 4 days in the month of July 2008 when water samples were collected and tested and for the whole year (July 2008 to the end of June 2009). In general, bench-scale permeability and fouling indices were higher than those for the full-scale plant. For the first 4 days in July, the TFI, HIFI and CIFI for the bench-scale system were approximately 8, 7, and 6 times higher than those at fullscale, respectively. Taking the whole year into consideration, the TFI and HIFI at bench-scale were 5 times higher than values at full-scale while the CIFI was 2 times higher than that at full-scale. The steady state CIFI values (CIFI for the last 4 months) were more comparable to the CIFI at benchscale (2.1 104 m2/L for the full-scale versus 2.5 104 m2/L for the bench-scale system). However, the fouling index ratios (HIFI/TFI, CIFI/HIFI, CIFI/TFI) for the bench-scale and
Table 4 e Utility B bench-scale and full-scale FI comparison. Bench-scale
Full-scale First 4 days of July
Permeability (LMH/kPa) TFI 104 (m2/L) HIFI 104 (m2/L) CIFI 104 (m2/L) HIFI/TFI CIFI/HIFI CIFI/TFI
a
3.78 (2.3e7.5) 255 (212e398)a 13.4 2.5 0.05 0.19 0.01
1.91 31 1.86 0.42 0.06 0.23 0.01
Ratio of bench-scale to full-scale 1 year a
1.34 (0.5e2.5) 51.5 (0.0027e640)a 2.7 1.3 0.05 0.48 0.03
First 4 days of July
1 year
2.0 8.2 7.2 6.0 0.83 0.83 1.0
2.8 5.0 5.0 1.9 1.0 0.4 0.33
a Values in parentheses are range. (Source: Tobiason and Nguyen. Hollow Fiber Membrane Fouling: Characterization & Role of Organic Nitrogen. ª2011 Water Research Foundation. Reprinted with permission).
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 2 7 e2 6 3 7
Fig. 8 e Productivity loss for membrane A for three water sources, chemical cleaning between groups of data. (Source: Tobiason and Nguyen. Hollow Fiber Membrane Fouling: Characterization & Role of Organic Nitrogen. ª2011 Water Research Foundation. Reprinted with permission).
the full-scale data for the first 4 days in July had very similar values, suggesting that the nature of the fouling was similar (the fouling fraction reversible by hydraulic backwash and chemical cleaning at bench- and full-scale were similar). When comparing the whole year of full-scale data with the bench-scale, the fouling index ratios were less similar which was expected due to changes in water quality and permeate flux fluctuation by demand. However, the bench-scale results provided an estimation of long-term membrane performance. These results suggest that testing at benchscale can be used to assess membrane performance at fullscale. Results for comparisons between membrane types or pre-treatment options are likely to be useful as well.
1.8
1/J's
1.6
1.4
High initial productivity loss
1.2
Membrane A Membrane B
1.0 0
1000
2000
3000
4000
5000
Specific Volume (L/m2)
Fig. 9 e Productivity loss for membranes A and B for the wastewater-impaired source; chemical cleaning between groups of data. (Source: Tobiason and Nguyen. Hollow Fiber Membrane Fouling: Characterization & Role of Organic Nitrogen. ª2011 Water Research Foundation. Reprinted with permission).
2635
Fig. 10 e Productivity loss for membrane B for the three water sources; chemical cleaning between groups of data. (Source: Tobiason and Nguyen. Hollow Fiber Membrane Fouling: Characterization & Role of Organic Nitrogen. ª2011 Water Research Foundation. Reprinted with permission).
3.4.
Membrane performance with natural water sources
Fouling index values through bench-scale testing were used to compare performance of two different membrane types tested with three different natural water sources (oligotrophic, algal bloom impacted (ABI), and wastewater impaired (WWI)). These natural water sources were normalized to have the same TOC (2 mg/L) and DON/DOC ratio (1/18). Fig. 8 compares performance data for membrane A for the three different water sources. While membrane A had high hydraulically irreversible fouling with the WWI source, it fouled much less with the ABI source and the oligotrophic source. In general, the fit of inverse specific flux (1/J’s) and specific volume (V) to the linear plot was good, thus linear regression was used to determine HIFI values. The HIFI values were fairly constant for the WWI source, relatively constant for the ABI source, and varied from 0.8 to 1.8 104 m2/L for the oligotrophic source. The HIFI varied between chemical cleaning cycles even though the chemical clean procedure was kept constant; the same scenario occurred for the fullscale plant as presented earlier. Fig. 9 compares the performance data for membranes A and B for the wastewater impaired source. While membrane A fouled significantly, the change in specific flux over one day was much less for membrane B, showing a significant impact of membrane type on membrane fouling. Membrane B had high initial specific flux loss but the rate of fouling decreased significantly after 2 to 3 hydraulic backwash cycles. The same scenario occurred for membrane B tested using the algal impacted water source. This suggests that a few hours of testing are not sufficient for a good assessment of longer-term membrane performance. Although membranes A and B both have PVdF based material, they behaved quite differently. Fig. 10 presents the performance data for membrane B for all three water sources, using all data to determine the HIFI values. Among the three water sources, the ABI source caused
2636
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 2 7 e2 6 3 7
Table 5 e FI summary for membranes A and B tested with natural water sources. Mem.
FI 104 (m2/L)
Water
TFI
A
B
Olig ABI WWI Olig ABI WWI
HBW Cyc. 1
Ave. first 4 HBW cycs
12 6.3 6.5 9 20 7
5.9 5.0 7.8 13.6 13.5 7.5
HIFI
CIFI
1.3 1.2 4.7 0.6 1.5 0.5
0.27 0.42 0.16 0.16 0.42 0.10
HIFI/TFIave
CIFI/HIFI
CIFI/TFIave
0.29 0.24 0.81 0.06 0.11 0.13
0.23 0.17 0.04 0.24 0.27 0.20
0.07 0.08 0.03 0.01 0.03 0.02
(Source: Tobiason and Nguyen. Hollow Fiber Membrane Fouling: Characterization & Role of Organic Nitrogen. ª2011 Water Research Foundation. Reprinted with permission).
the highest productivity loss. The HIFI values for each chemical cleaning cycle stayed relatively constant for the oligotrophic water source and the WWI source, but varied from 0.7 to 2.5 104 m2/L for the ABI source. Table 5 summarizes the fouling indices for the two membranes and the three different water sources. The TFI values were reported for the first HBW cycle and for the average of the first four HBW cycles. The HIFI and CIFI were significantly lower than the TFI indicating that most fouling was reversible by hydraulic BW and chemical cleaning. ANOVA test results showed that at the 95% confidence level, the TFI differed between membrane types while water source did not have any statistically significant impact on the TFI values. For the HIFI, ANOVA multiple pair-wise comparisons show that each membrane type and water source combination resulted in significantly different HIFI values. Chemical cleaning with chlorine was very effective in reversing the hydraulic-irreversible fouling for both membrane types. Although membrane B had higher initial total fouling than membrane A, membrane B had much less hydraulic-irreversible fouling than membrane A. For membrane A, overall the WWI source caused the most total and hydraulic-irreversible fouling; more than 80% of the fouling could not be reversed by hydraulic backwash. However, chemical cleaning with chlorine reversed more than 95% of the hydraulic-irreversible fouling and the WWI source caused the lowest chemical-irreversible fouling of the three water sources. There was no significant difference in hydraulic-irreversible fouling between the ABI source and the oligotrophic source for membrane A. More than 70% of the total fouling was reversed by hydraulic backwash and the chemical-irreversible fouling was only 8% (ABI) and 7% (oligotrophic source) of the total fouling. Although membrane B had higher initial TFI values than membrane A, the TFI values of subsequent cycles decreased significantly, resulting in much lower HIFIs than those for membrane A. Among the three water sources, the ABI source caused the most fouling to membrane B, both reversible and irreversible, followed by the oligotrophic source and then the WWI source. This suggests that the total and hydraulic-irreversible fouling are very specific to water source and membrane type. Given the same normalized TOC and DON/DOC levels, both membrane A and B shared the same trend with regards to chemical-irreversible fouling: highest for the ABI source, less for the oligotrophic source, and least for the WWI source. ANOVA test results show
that the impact of water source is significant while membrane type did not cause any statistically significant differences in the CIFI value. High initial fouling and hydraulic-irreversible fouling rates did not lead to high chemical-irreversible fouling rates, which further confirms the necessity of longer-term testing (at least a few days) to better assess longer-term membrane performance.
4.
Conclusions
Fouling indices developed from the resistance-in-series model were validated and are shown to be useful tools to describe membrane performance data for both raw and pretreated water, for different water sources, and for two different membrane types. Fouling was proportional to specific volume for both raw and pretreated water, at different fluxes, and different membrane types; different fouling indices were determined to describe different aspects of fouling. For highly variable TMP data, a common occurrence at short time scale, it is best to use average values to determine fouling index values. For total fouling index values, the two-point method (the difference in TMP from start to end of one cycle prior to hydraulic BW) should be used for highly noisy data. To reduce the impact of noisy data, average values for the first and last few data points should be used instead of the first and the last single data points (10 first and last data points were used in this study). Fouling was specific to water source and membrane type and no generalization on the impact of water source on fouling was found. Direct comparison of bench-scale and full-scale data shows that specific flux and fouling indices for the benchscale system were higher than those determined from the full-scale system data. Scale-up effects, differences in material, feed water quality, and the absence of air scouring in the bench-scale system are likely the cause of these differences. It is recommended that air scouring be included in the bench-scale system for a better comparison of bench-scale and full-scale. However, the fouling index ratios (HIFI/TFI, CIFI/HIFI, and CIFI/TFI) for the bench-scale and the full-scale data during the period when water samples were collected and tested at bench-scale were very similar, suggesting a similar fouling nature at the bench- and full-scales. This suggests that fouling index ratios should be used as one
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 2 7 e2 6 3 7
measure of membrane operation management. Comparison of the fouling index ratios could direct the focus on hydraulic backwash or chemical cleaning for optimal membrane operation. For all the trials in this study, hydraulic backwash was not as effective as chemical cleaning in reversing irreversible fouling. Investigations of the addition of air scouring, backwash frequency, backwash flowrate and intensity are logical steps to decrease hydraulic-irreversible fouling. The study results suggest that testing at the bench-scale could be potentially used for membrane selection screening. A few hours of initial testing are not a good indicator of long-term membrane performance; at least a few days of testing is recommended.
Acknowledgments We thank the Vietnam Education Foundation (VEF) and the Water Research Foundation (formerly the American Water Works Association Research Foundation, AwwaRf) for funding this project. We thank Siemens Water Corporation and GE Water & Process Technologies for providing the membrane samples. Special thanks to the Amherst Water Treatment Plant, Norwalk CT water treatment plant (Mr. Tom Villa) and Mr. William Clunie of AECOM in facilitating interactions with the Norwalk CT water utility. The views expressed in this article are not necessarily those of the funding agencies, the membrane providers, or the water utilities.
references
Boerlage, S., Kennedy, M., Meseret, A., Elhadi, A., Gilbert, G., Schippers, J., 1998. Monitoring particulate fouling in membrane systems. Desalination 118 (1), 131e142. Boerlage, S., Kennedy, M., Meseret, A., Elhadi, A., Gilbert, G., Zeyad, T., Schippers, J., 2003. The MFI-UF as a water quality test and monitor. Journal of Membrane Science 211 (2), 271e289. Crozes, G.F., Jacangelo, J.G., Anselme, C., Laine, J.M., 1997. Impact of ultrafiltration operating conditions on membrane
2637
irreversible fouling. Journal of Membrane Science 124 (1), 63e76. Hermia, J., 1982. Constant pressure blocking filtration lawsapplication to power-law non-newtonian fluids. Transactions of the Institution of Chemical Engineers 60, 183e187. Howe, K.J., Marwah, A., Chiu, K.P., Adham, S.S., 2007. Effect of membrane configuration on bench-scale MF and UF fouling experiments. Water Research 41 (17), 3842e3849. Huang, H., Thayer, Y., Jacangelo, J., 2008a. Unified membrane fouling index for low pressure membrane filtration of natural waters: principles and methodology. Environmental Science and Technology 42 (3), 714e720. Huang, H., Spinett, R., O’Melia, C.R., 2008b. Direct-flow microfiltration of aquasols e I. Impacts of particle stabilities and size. Journal of Membrane Science 314 (1e2), 90e100. Huang, H., Thayer, Y., Jacangelo, J., 2009. Novel approach for the analysis of bench-scale, low pressure membrane fouling in water treatment. Journal of Membrane Science 334 (1e2), 1e8. Jacangelo, J., Huang, H., Young, T., Amy, G., Lozier, J., Myrose, C., 2006. Factors Influencing Natural Organic Matter Fouling in Low Pressure, Hollow Fiber Membrane Filtration of Natural Waters. Proc. AWWA WQTC, Denver, Colorado. Lee, N., Amy, G., Croue, J.P., Buisson, H., 2004. Identification and understanding of fouling in low-pressure membrane (MF/UF) filtration by natural organic matter (NOM). Water Research 38 (20), 4511e4523. Lee, N., Amy, G., Croue, J.P., 2006. Low-pressure membrane (MF/ UF) fouling associated with allochthonous versus autochthonous natural organic matter. Water Research 40 (12), 2357e2368. Lozier, J., Cappucci, L., Amy, G., Lee, N., Jacangelo, J., Huang, H., Young, T., Mysore, C., Emeraux, C., Clouet, J., Croue, J., Heijmann, B., 2008. Natural Organic Matter Fouling of LowPressure Membrane Systems. Water Research Foundation Report, USA. Nguyen, A. 2010. Bench-Scale Assessment of Low Pressure Membrane Fouling: Characterization and Examination the Role of Organic Nitrogen Compounds. Ph.D. Dissertation, Department of Civil and Environmental Engineering, University of Massachusetts, Amherst, MA, USA. Tarabara, V.V., Hovinga, R.M., Wiesner, M.R., 2002. Constant transmembrane pressure vs. constant permeate flux: effect of particle size on crossflow membrane filtration. Environmental Engineering Science 19 (6), 343e355. Tobiason, J.E., Nguyen, A., 2011. Hollow Fiber Membrane Fouling: Characterization & Role of Organic Nitrogen. Water Research Foundation, Denver, Colorado.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 3 8 e2 6 4 6
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Ability of three DNA-based assays to identify presumptive Escherichia coli colonies isolated from water by the culturebased mFC agar method Andre´e F. Maheux a,b, E`ve Be´rube´ a, Dominique K. Boudreau a, Philippe Cantin c, Maurice Boissinot a,b, Luc Bissonnette a,b, Lynda Rodrigue d, Michel G. Bergeron a,b,* a
Centre de recherche en infectiologie de l’Universite´ Laval, Centre de recherche du CHUQ, Que´bec City, Que´bec, Canada De´partement de microbiologie-infectiologie et immunologie, Faculte´ de me´decine, Universite´ Laval, Que´bec City, Que´bec, Canada c Centre d’expertise en analyse environnementale du Que´bec, Ministe`re du de´veloppement durable, de l’environnement et des parcs, Que´bec City, Que´bec, Canada d Exova Canada Inc., Que´bec City, Que´bec, Canada b
article info
abstract
Article history:
We tested the ability of three PCR assays, targeting uidA and tuf genes to correctly identify
Received 8 November 2010
Escherichia coli colonies isolated from water and we compared them to two b-glucuroni-
Received in revised form
dase-based culture methods (Colilert and Readycult), in terms of specificity and sensi-
15 January 2011
tivity. E. coli isolates recovered on mFC agar were first tested for the presence of the uidA
Accepted 15 February 2011
positive colonies were presumed to be E. coli. For further characterization, uidA-negative
Available online 22 February 2011
colonies were subsequently identified using the Vitek 2 automated system. Colilert and Readycult detected 436 and 442 of 468 colonies identified as E. coli on mFC corresponding
Keywords:
to sensitivities of 93.2 and 94.4%, respectively. None of the 59 non-E. coli isolates was
Water analysis
detected by both methods for a specificity of 100%. Two (2) uidA and 1 tuf PCR assays were
Escherichia coli
also tested. The uidA PCR assays yielded positive signals for 447 (95.5%) and 434 (92.7%) of
b-Glucuronidase-based assays
468 E. coli isolates tested respectively, whereas the tuf PCR assay showed a sensitivity of
mFC agar
100%. None of the 59 non-E. coli isolates was detected by both uidA PCR assays (100%
Real-time PCR
specificity), whereas tuf PCR false-positive signals were obtained with Escherichia fergusonii and Escherichia albertii. However, since these 2 species are principally found in the feces of mammals and birds, their detection indicates a fecal contamination. Consequently, using a 1-h tuf rtPCR assay to confirm the identity of E. coli colonies on mFC agar is as specific, more sensitive, and potentially more cost-efficient than culture methods based on bglucuronidase detection. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
In the 80’s, the membrane filtration technique performed on mFC agar incubated at 44.5 C was a widely used and accepted method for the analysis of thermotolerant fecal coliforms in
water and wastewater (APHA, 1981). Since all thermotolerant coliforms are not of fecal origin, confirmatory tests are required (Geldreich et al., 1965). Thus, standard procedures recommended the confirmation of blue colonies by the EC MUG test, where the 4-methylumbelliferone glucuronide is cleaved by the
* Corresponding author. Centre de recherche en infectiologie de l’Universite´ Laval, Centre de recherche du CHUQ, 2705 Laurier Blvd., Que´bec City, Que´bec, Canada G1V 4G2. Tel.: þ1 418 656 4141x48753; fax: þ1 418 654 2715. E-mail address:
[email protected] (M.G. Bergeron). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.02.021
2639
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 3 8 e2 6 4 6
Table 1 e Ability of the membrane filtration method on mFC agar to detect E. coli strains. E. coli strains (n ¼ 53)
Strain no.
Serotype
Origin
Growth on mFC
E. coli E. coli E. coli E. coli E. coli E. coli E. coli E. coli E. coli E. coli E. coli E. coli E. coli E. coli E. coli E. coli E. coli E. coli E. coli E. coli E. coli E. coli E. coli E. coli E. coli E. coli E. coli E. coli E. coli E. coli E. coli E. coli E. coli E. coli E. coli E. coli E. coli E. coli E. coli E. coli E. coli E. coli E. coli E. coli E. coli E. coli E. coli E. coli E. coli E. coli E. coli E. coli E. coli All E. coli strains:
ATCC 11775 ATCC 43886 ATCC 23511 CCRI-474 CCRI-475 CCRI-476 CCRI-477 CCRI-481 ATCC 43890 ATCC 43894 ATCC 43895 ATCC 43896 CCRI-493 CCRI-494 CCRI-495 CCRI-496 CCRI-497 CCRI-498 CCRI-499 CCRI-500 CCRI-501 CCRI-502 CCRI-1213 CCRI-2099 CCRI-2166 CCRI-8825 CCRI-8831 CCRI-8832 CCRI-8833 CCRI-8834 CCRI-8835 CCRI-8836 CCRI-8837 CCRI-8838 CCRI-8839 CCRI-8840 CCRI-14858 CCRI-14871 CCRI-14881 CCRI-16465 CCRI-16485 CCRI-16527 CCRI-16528 CCRI-16537 CCRI-16539 CCRI-16540 CCRI-16579 CCRI-16580 CCRI-17151 CCRI-17158 CCRI-17161 CCRI-17172 CCRI-17176
O1:K1:H7 O25:K98:NM O16:K1(L):NM O12:NM O4:H5 O8:H9 O7:NM O157:H7 O157:H7 O157:H7 O157:H7 O78:K80:H12 O26:NM O111:NM O113:H21 O117:H4 O128:NM O128:H8 O157:H7 O157:H7 O157:H7 O18:NM N/A N/A N/A N/A O157:H7 O157:H7 O103:H2 O103:H2 O111:HO111:HO26:NM O26:NM O145:NM O145:NM N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A
Clinical Clinical Clinical Clinical Clinical Clinical Clinical Clinical Clinical Clinical Clinical Clinical Clinical Clinical Clinical Clinical Clinical Clinical Clinical Clinical Clinical Clinical Clinical Clinical Clinical Clinical Clinical Clinical Clinical Clinical Clinical Clinical Clinical Clinical Clinical Clinical Environmental Environmental Environmental Environmental Environmental Environmental Environmental Environmental Environmental Environmental Environmental Environmental Environmental Environmental Environmental Environmental Environmental
Blue Blue Blue Blue Blue Blue Blue Blue Blue Blue Blue Blue Blue No growth Blue Blue Blue Blue Blue Blue Blue Blue Blue Blue Blue Blue Blue Blue Blue Blue Blue Blue Blue Blue Blue Blue Blue Blue Blue Blue Blue Blue Blue Blue Blue Blue Blue Blue Blue Blue Blue Blue Blue 52/53 (98.1%)
b-glucuronidase enzyme produced by Escherichia coli cells to release a fluorogenic end product that is visible under longwave UV light (APHA, 1981; Feng and Hartman, 1982). However, the combination of a presumptive test (mFC agar) and a confirmed test (EC MUG) requires at least 72 h for completion. The use of media containing specific substrates for the E. coli b-D-glucuronidase has become widely adopted to assess
water quality (Cowburn et al., 1994; Eckner, 1998; Edberg et al., 1988; Manafi et al., 1991). Used in combination with substrates for detecting b-galactosidase activity, these tests offer a simple method for the simultaneous detection of total coliforms and E. coli in water samples. Among all recommended methods, Colilert is a defined substrate technologybased test widely used to monitor water quality. This
2640
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 3 8 e2 6 4 6
Table 2 e Specificity of the membrane filtration method on mFC agar. Strains (n ¼ 101) Acinetobacter baumanii Acinetobacter calcoaceticus Acinetobacter haemolyticus Acinetobacter johnsonii Acinetobacter lwoffii Acinetobacter radioresistens Acinetobacter ursingii Aeromonas hydrophila Burkholderia cepacia Burkholderia stabilis Burkholderia thailandensis Citrobacter amalonaticus Citrobacter braakii Citrobacter farmeri Citrobacter freundii Citrobacter gillenii Citrobacter koseri Citrobacter murliniae Citrobacter sedlakii Citrobacter werkmanii Citrobacter youngae Cronobacter sakazakii Enterobacter aerogenes Enterobacter agglomerans Enterobacter amnigenus Enterobacter asburiae Enterobacter cancerogenus Enterobacter cloacae Enterobacter dissolvens Enterobacter gergoviae Enterobacter hormaechei Enterobacter intermedius Enterobacter nimipressuralis Enterobacter pyrinus Escherichia albertii Escherichia blattae Escherichia fergusonii Escherichia fergusonii Escherichia fergusonii Escherichia hermanii Escherichia vulneris Klebsiella oxytoca Klebsiella pneumoniae Klebsiella pneumoniae Klebsiella pneumoniae Klebsiella pneumoniae subsp. ozaenae Klebsiella pneumoniae subsp. pneumoniae Klebsiella pneumoniae subsp. rhinoscleromatis Pantoea dispersa Proteus hauseri Proteus mirabilis Proteus myxofaciens Proteus penneri Proteus vulgaris Providencia alcalifaciens Providencia rettgeri Providencia rustigianii Providencia stuartii Pseudomonas aeruginosa Pseudomonas oryzihabitans Pseudomonas stutzeri Raoultella ornithinolytica Raoultella planticola Raoultella terrigena
Strain no. ATCC 19606 CCRI-10481 ATCC 17906 ATCC 17909 LSPQ 2006 CCUG 34434 ATCC BAA-617 ATCC 7966 ATCC 25416 CCUG 13348 ATCC 700388 ATCC 25416 ATCC 43162 ATCC 51112 ATCC 8090 ATCC 51117 ATCC 27156 ATCC 51641 ATCC 51115 ATCC 51114 ATCC 29935 ATCC 29544 ATCC 13048 ATCC 27989 ATCC 33072 ATCC 35953 ATCC 35317 ATCC 13047 ATCC 23373 ATCC 33028 ATCC 49162 ATCC 33110 ATCC 9912 ATCC 49851 ATCC 12032 ATCC 29907 ATCC 35469 CCUG 21144 CCUG 29925 ATCC 33650 ATCC 33821 ATCC 13182 CCRI-821 CCRI-17697 ATCC 27799 ATCC 29015 ATCC 33495 ATCC 29018 ATCC 14589 ATCC 13315 ATCC 35659 ATCC 19692 ATCC 33519 ATCC 29513 ATCC 9886 ATCC 9250 ATCC 33673 ATCC 43664 ATCC 39018 ATCC 43272 ATCC 17588 ATCC 31898 ATCC 33531 ATCC 33257
Serotype N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A Type 13 N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A
Origin Human Human Human Human Human Human Human Food Food Human Soil Food Human Human N/A Human N/A Human Human Human Food Human Human Human Soil Human Human Human Food Human Human Water Tree Plant N/A Animal Human N/A Human Human Human Human Human Human Human Human Human N/A Soil Human N/A Animal Human Animal Human Human Human Human N/A Food Human Human Food Water
Growth on mFC No growth No growth No growth No growth No growth No growth No growth No growth No growth No growth No growth No growth No growth No growth No growth No growth No growth No growth No growth No growth No growth No growth No growth No growth No growth No growth No growth No growth No growth No growth No growth No growth No growth No growth No growth No growth Purple Purple Purple No growth No growth No growth No growth No growth No growth No growth No growth No growth No growth No growth No growth No growth No growth No growth No growth No growth No growth No growth No growth No growth No growth No growth No growth No growth
2641
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 3 8 e2 6 4 6
Table 2 (continued) Strains (n ¼ 101) Salmonella bongori Salmonella enterica subsp. Salmonella enterica subsp. Salmonella enterica subsp. Salmonella enterica subsp. Salmonella enterica subsp. Salmonella enterica subsp. Salmonella enterica subsp. Salmonella enterica subsp. Salmonella enterica subsp. Salmonella enterica subsp. Salmonella enterica subsp. Salmonella enterica subsp. Salmonella enterica subsp. Salmonella enterica subsp. Salmonella enterica subsp. Salmonella enterica subsp. Salmonella enterica subsp. Salmonella enterica subsp. Salmonella enterica subsp. Serratia entomophila Serratia ficaria Serratia fonticola Serratia grimesii Serratia liquefaciens Serratia marcescens Serratia odorifera Serratia proteamaculans Serratia rubidaea Shigella boydii Shigella boydii Shigella dysenteriae Shigella flexneri Shigella flexneri Shigella sonnei Shigella sonnei Shigella sonnei
Strain no. arizonae enterica Choleraesuis enterica Enteritidis enterica Gallinarum enterica Heidelberg enterica Infantis enterica Kottbus enterica Newport enterica Panama enterica Paratyphi A enterica Paratyphi B enterica Pullorum enterica Putten enterica Thompson enterica Typhi enterica Typhimurium houtenae indica salamae
ATCC 43975 ATCC 13314 ATCC 7001 ATCC 13076 ATCC 9184 ATCC 8326 CCRI-10033 CCRI-10012 CCRI-10047 CCRI-10017 ATCC 9150 ATCC 8759 ATCC 9120 ATCC 15787 CCRI-10038 ATCC 27870 ATCC 14028 ATCC 43974 ATCC 43976 ATCC 43972 ATCC 43705 ATCC 33105 ATCC 29844 ATCC 14460 ATCC 27592 ATCC 8100 ATCC 33077 ATCC 33765 ATCC 27593 ATCC 9207 ATCC 8700 ATCC 11835 ATCC 12022 CCRI-2198 ATCC 29930 ATCC 25931 CCRI-2196
Serotype N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A Type 1 Type 2b N/A N/A N/A N/A
Origin N/A N/A N/A N/A N/A N/A Human Human Human Human N/A Human Human Animal Human N/A N/A N/A N/A N/A Animal Food Water N/A Food N/A Human Food N/A N/A N/A Clinique Clinique N/A Clinique N/A N/A
Growth on mFC No growth No growth No growth No growth No growth No growth No growth No growth No growth No growth No growth No growth No growth No growth No growth No growth No growth No growth No growth No growth No growth No growth No growth No growth No growth No growth No growth No growth No growth No growth No growth No growth No growth No growth Gray Gray No growth
N/A: non available. CCRI: culture collection of Centre de recherche en infectiologie.
technique, which uses costly media, is easy and rapid to perform, yields results after 18e24 h, and does not require confirmatory tests, according to the manufacturer. However, an unexpected level of false-negative results was observed with the Colilert method by Clark et al. (1991), Hall and Moyer (1989), Lewis and Mak (1989), and Pitkanen et al. (2007), when it was compared to the standard membrane filtration method for fecal coliforms (mFC). Indeed, Feng and Hartman (1982), Hartman (1989), Martin et al. (1993), and Moberg et al. (1988), found that 3e7% of E. coli isolates found in food and water were b-glucuronidase negative. However, blue colonies that grow on mFC agar require confirmation tests and those tests are often based on b-glucuronidase production (APHA, 1981). Thus, the result remains the same since b-glucuronidase-negative E. coli colonies that have grown on mFC agar are missed. PCR-based assays can be completed in approximately 1 h, allowing a faster turnaround time than that required to confirm the identity of a blue colony on mFC agar by recommended phenotypic characterization tests. PCR amplification
also detects b-glucuronidase-negative E. coli and is more sensitive than b-glucuronidase activity-based methods (Maheux et al., 2009). As the occurrence of potable water samples presenting fecal contamination is very low, we are proposing that it could be more advantageous for water analysis laboratories to test the microbiological quality of water by the cost-effective mFC agar and then confirm the identity of blue colonies isolated from contaminated samples with a 1-h E. coli-specific PCR test, rather than testing the microbiological quality of every water sample by using a more expensive b-glucuronidase-based method. In this study, we have compared the performance of Colilert and Readycult, and that of 3 PCR assays to identify presumptive E. coli colonies isolated from water samples on mFC agar after membrane filtration. The results of our study will provide water analysis laboratories with a better appreciation of the advantages and disadvantages of using enzymatic-based methods or molecular-based methods to recover E. coli from water or confirm the identity of a presumptive E. coli colony on mFC agar.
2642
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 3 8 e2 6 4 6
2.
Materials and methods
2.1.
Bacterial strains
The ubiquity (ability of the medium to detect all or most E. coli strains) of the membrane filtration method on mFC agar to detect E. coli strains in water was first demonstrated by testing 53 E. coli strains of different serotypes and of different geographic origins (Table 1). These strains were obtained from various sources: ATCC (n ¼ 7), Huashan Hospital (Shangai, China; n ¼ 1), Institut fu¨r Hygiene und Mikrobiologie der Universita¨t Wu¨rzburg (Wu¨rzburg, Germany, n ¼ 10), Laboratoire de Sante´ Publique du Que´bec (Sainte-Anne-de-Bellevue, Que´bec, Canada; n ¼ 15), National Institute of Public Health (Warsaw, Poland; n ¼ 1), South African Institute for Medical Research (Johannesburg, South Africa; n ¼ 1), and Wyeth-Ayerst Research (Pearl River, NY; n ¼ 1). Environmental isolates of E. coli (n ¼ 17) were obtained from various sources and isolated by different methods including Colilert (n ¼ 4), MI agar (n ¼ 2), Chromocult coliforms agar ES (n ¼ 3), modified mTEC agar (n ¼ 5), and sheep blood agar (n ¼ 3). These environmental strains were isolated from (i) drinking water samples obtained from Exova Canada Inc. (Que´bec City, Que´bec, Canada; n ¼ 9), (ii) water samples collected in Bermuda (n ¼ 3), (iii) and river water samples from across Que´bec obtained from the Centre d’expertise en analyse environnementale du Que´bec (Que´bec City, Que´bec, Canada; n ¼ 5). The sensitivity of the membrane filtration method on mFC agar was also verified by testing all 305 E. coli isolates that have grown on 10 different mFC agar plates during summer 2009 in Que´bec City (Panel 1). These environmental strains were isolated from surface water (n ¼ 297), sewage (n ¼ 3) as well as lake (recreational water; n ¼ 5). To enlarge our sampling, we also selected 222 colonies randomly isolated (1e5 colonies by plate) on 59 mFC agar plates containing between 1 and 300 colonies (Panel 2). Those plates have been chosen because colonies were not confluent. These environmental strains were isolated from surface water (n ¼ 130), river water (recreational water; n ¼ 5), well water (n ¼ 17), sewage (n ¼ 62), as well as lake water (recreational water; n ¼ 8) samples obtained from Exova Canada Inc. (Que´bec City, Que´bec, Canada) and from the Centre d’expertise en analyse environnementale du Que´bec. The uidA PCR published by Bej et al. (1991) was used to identify each colony isolated on mFC agar. The uidA-positive colonies were considered as E. coli. Subsequently, the remaining uidA-negative isolates were identified using the Vitek 2 automated system. The analytical specificity of mFC agar was evaluated against 101 reference and environmental non-E. coli strains phylogenetically related to E. coli (Table 2). The identity of these strains was confirmed using a MicroScan Autoscan-4 automated system (Siemens Healthcare Diagnostics Inc., Newark, DE, USA) or a Vitek 2 automated system (bioMe´rieux SA, Marcy l’E´toile, France). Bacterial strains were grown from frozen stocks, kept at 20 C in brain heart infusion (BHI) medium (Becton, Dickinson and Company, Mississauga, Ontario, Canada) containing 10% glycerol, and cultured on sheep’s blood agar. Three culture passages were performed prior the preparation of the bacterial suspension preparation.
2.2.
Bacterial suspension preparation
Each bacterial strain described in Tables 1 and 2 was grown to semi-logarithmic phase (0.5e0.6 OD600) in liquid BHI medium and adjusted to a 0.5 McFarland standard (Fisher Scientific Company, Ottawa, Ontario, Canada), before being serially diluted ten-fold in phosphate-buffered saline (PBS; 137 mM NaCl, 6.4 mM Na2HPO4, 2.7 mM KCl, 0.88 mM KH2PO4, pH 7.4). For each strain, an aliquot of the 105 dilution was spiked in source water (Labrador, Anjou, Que´bec, Canada) to produce a suspension having approximately 100 colony forming units (CFU) per 100 mL of water. Bacterial count was verified by filtering 100 mL of each spiked water sample through a GN-6 membrane filter (47 mm diameter, 0.45 mm pore size; Pall Life Sciences, Ville St. Laurent, Que´bec, Canada) with a standard platform manifold (Millipore Corporation, Billerica, MA, USA). Then, the filter was incubated on sheep blood agar plates for 24 2 h at 35.0 0.5 C prior to the determination of colony counts. Tests to confirm the sterility of filter membranes and buffer used for rinsing the filtration apparatus were also performed.
2.3.
Membrane filtration method
To establish the ability of the mFC agar to detect E. coli, each strain of our panel was spiked in a 100 mL volume and filtered on a GN-6 membrane filter with a standard platform manifold. Filters were then incubated on mFC agar plates (Becton, Dickinson and Company, Franklin Lakes, NJ, USA) for 24 2 h at 44.5 0.5 C. Subsequently, colony count and color were determined. To establish the false-positive and false-negative rates of mFC agar, contaminated water samples from Que´bec City area were used. First, volumes of water samples were filtered on GN-6 membrane filter with a standard platform manifold. Filters were then incubated on mFC agar plates for 24 2 h at 44.5 0.5 C. Subsequently, all colonies harvested from plates having approximately 50 non-confluent colonies were separately inoculated on tryptic soy agar (TSA) plates and incubated for 24 2 h at 35.0 0.5 C. Each pure culture obtained was then analyzed with Colilert, Readycult, and the molecular assays, as described below. Each preparation of mFC and TSA plates was tested for performance using pure cultures of target and non-target microorganisms, as recommended by the USEPA microbiology methods manual (USEPA, 1978). Tests to confirm the sterility of the filter membranes and buffer used for rinsing the filtration apparatus were also performed.
2.4.
Liquid culture methods
For liquid culture methods, preparation, validation, storage, and handling were performed according to the manufacturer’s instructions. For the Colilert method, one snap pack containing the Colilert reagent (IDEXX Laboratories Canada Corp., Toronto, Ontario, Canada) was dissolved in 100 mL of sterile water. The medium obtained was then aseptically separated in 2-mL aliquots and conserved for a maximum of 48 h at 4 C prior to inoculation. The colonies isolated on mFC agar were aseptically inoculated in a 2-mL aliquot of Colilert
2643
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 3 8 e2 6 4 6
medium and incubated at 35.0 0.5 C for 24 h. E. coli-positive aliquots present a yellow color and fluoresce under UV light (l ¼ 365 nm). Positive and negative controls of the Colilert medium, using pure cultures of target and non-target microorganisms, were performed as recommended by the manufacturer. For the Readycult method, one snap pack containing the Readycult reagent (Merck, Darmstadt, Germany) was dissolved in 100 mL of sterile water. The medium obtained was then aseptically separated in 2-mL aliquots and conserved for a maximum of 48 h at 4 C prior to inoculation. The colonies isolated on mFC agar were aseptically inoculated in a 2-mL aliquot of Readycult medium and incubated at 35.0 0.5 C for 24 h. E. coli-positive cultures present a green color and fluoresce under UV light (l ¼ 365 nm). Positive and negative controls of the Readycult medium, using pure cultures of target and non-target microorganisms, were also performed as recommended by the manufacturer.
2.5.
Molecular methods primers and probes
The sequence of PCR and real-time (rt)PCR primers and probes used in this study to detect E. coli are shown in Table 3. The TEco573-T1-B1 E. coli/Shigella-specific rtPCR dual-labeled (TaqMan) rtPCR probe was developed by building a multiple alignment of tuf gene sequences retrieved from public databases with GCG programs (version 8.0; Accelrys, Madison, WI, USA). The Oligo primer analysis softwares (version 5.0; National Biosciences, Plymouth, MN, USA) were used to select candidate primer and probe sequences from the alignment. Oligonucleotide primers and probes were synthesized by Integrated DNA Technologies (Coralville, IA, USA).
2.6.
Molecular assays
PCR and rtPCR amplifications were performed using a bacterial suspension adjusted to a 0.5 McFarland standard. Bacterial cells were lysed using the BD Diagnostics-GeneOhm Rapid Lysis kit as recommended by the manufacturer (BD Diagnostics-GeneOhm, Que´bec City, Que´bec, Canada). Brielfy, 1 mL of bacterial crude lysate was transferred directly to a 19-mL PCR (24 mL for rtPCR) mixture containing
50 mM KCl, 10 mM Tris-HCl (pH 9.0), 0.1% Triton X-100, 2.5 mM MgCl2, 0.4 mM of each primer, (0.2 mM of probe for rtPCR), 200 mM each deoxyribonucleoside triphosphate (GE Healthcare Life Sciences Inc., Baie d’Urfe´, Que´bec, Canada), 3.3 mg per mL of bovine serum albumin (BSA; SigmaeAldrich Canada Ltd., Oakville, Ontario, Canada), 0.025 enzyme unit (U) of Taq DNA polymerase (Promega, Madison, WI, USA), and TaqStart antibody (Clontech Laboratories, Mountain View, CA, USA). Molecular decontamination of the PCR mixtures prior to PCR or rtPCR was done according to Maheux et al. (2009). For each experiment, 1 mL of sterile water was added to PCR mixtures to serve as the negative control. PCR mixtures were subjected to thermal cycling (3 min at 95 C and then 30 cycles of 1 s at 95 C, 30 s at 58 C, and 30 s at 72 C, with a 5-min final extension step at 72 C) in a PTC-200 DNA Engine thermocycler (MJ Research, now Bio-Rad Laboratories, Hercules, CA, USA). The rtPCR mixtures were subjected to thermal cycling in a Rotor Gene 3000 (Corbett Life Science, Sidney, Australia, now a QIAGEN company) under the following conditions: 3 min at 95 C and then 35 cycles consisting of a denaturation step of 15 s at 95 C and an annealing step of 60 s at 60 C.
3.
Results and discussion
3.1. Sensitivity of the membrane filtration method on mFC agar The ubiquity of the membrane filtration method on mFC agar was first evaluated by testing 53 E. coli strains of different serotypes and of different geographic origins (Table 1). Blue colonies grew on mFC agar for 51 of 53 strains after an incubation time of 24 h at 44.5 C for an analytical sensitivity of 96.2%. In a previous report, Maheux et al. (2008) determined the ubiquity of the culture-based methods MI agar, Chromocult coliform ES, Colilert, and Readycult to detect E. coli by testing 74 E. coli strains of different serotypes isolated from fecal and environmental settings. The results of that study showed a ubiquity of 79.7, 79.7, 51.4, and 81.1%, respectively. Contrary to detection on mFC agar which is based on lactose enzymatic degradation and incubation
Table 3 e Primers and probe sets used in this study for (real-time) PCR amplification. Assay
Genetic target
Primers and probe
Primers and probe sequence (50 / 30 )
Reference
A
uidA uidA
C
tuf
AAAACGGCAAGAAAAAGCAG ACGCGTGGTTACAGTCTTGCG GTGTGATATCTACCCGCTTCGC GAGAACGGTTTGTGGTTAATCAGGA FAMa-TCGGCATCCGGTCAGTGGCAGT-BHQ-1b TGGGAAGCGAAAATCCTG CAGTACAGGTAGACTTCTG TETc-AACTGGCTGGCTTCCTGG-BHQ-1b
Bej et al., 1991
B
UAL754 UAR900 784F 866R EC807 TEcol553 TEcol754 TEco573-T1-B1
a FAM, 6-carboxyfluorescein, fluorescence reporter dye. b BHQ-1, Black Hole Quencher-1, fluorescence quencher dye. c TET, tetrachlorofluorescein, fluorescence reporter dye.
Frahm and Obst, 2003
Maheux et al., 2009 This study
2644
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 3 8 e2 6 4 6
at 44.5 C, the 4 above-mentioned tests are based on b-glucuronidase activity, a property specific to E. coli/Shigella among the Enterobacteriaceae (Cleuziat and Robert-Baudouy, 1990). In contrast to these 4 methods that failed to detect between 20 and 50% of E. coli strains, the culture-based method on mFC agar presents an ubiquity of 96.2% and is even able to detect O157:H7 serotype strains in which b-glucuronidase activity is mostly absent (Doyle and Schoeni, 1984; Krishnan et al., 1987; Ratnam et al., 1988). The sensitivity of the membrane filtration method on mFC agar was also demonstrated by testing all E. coli colonies isolated on 10 mFC agar plates following the filtration of 10 water samples during summer 2009 in Que´bec City (Panel 1). Of 290 colonies isolated on mFC agar that were positive for the uidAspecific PCR assay or identified as E. coli by Vitek, 289 were blue for a sensitivity of 99.7%; the only discordant colony identified as E. coli grew lilac on mFC. This result is in accordance with a study of Lavoie (1983) that identified strains isolated on mFC agar as fecal coliforms demonstrating that none of the 24 nonblue colonies isolated on mFC agar was E. coli for a sensitivity of 100%.
3.2. Specificity of the membrane filtration method on mFC agar The analytical specificity of the membrane filtration method on mFC agar was verified by testing 101 non-E. coli strains that are frequently encountered in environmental settings or phylogenetically closely related to E. coli (Table 2). None of these bacterial species grew positive (blue) colonies on mFC agar after an incubation time of 24 h at 44.5 C. However, Escherichia fergusonii and 2 of 3 Shigella sonnei strains, two species closely related to E. coli, grew on mFC agar, yielding purple and gray colonies, respectively. The specificity of mFC agar was also verified by testing all non-E. coli colonies of Panel 1. Of 305 blue colonies isolated from mFC agar, 15 were not identified as E. coli for a specificity of 95.1%. False-positive results were caused by Acinetobacter baumanii (4), Bordetella sp. (1), Cronobacter sakazakii (2), E. fergusonii (3), Escherichia hermanii (1), and Klebsiella pneumoniae (4). The selectivity of the mFC medium is mainly based on lactose utilization as well as a high incubation temperature of 44.5 C (Rose et al., 1975). Unfortunately, and in contrast with the activity of b-glucuronidase, these 2 parameters are not specific to E. coli/Shigella strains, as isolates belonging to genera Klebsiella, Enterobacter, Citrobacter, Salmonella, Serratia, Proteus, and Yersinia have been found to give false-positive results on mFC agar (Stramer, 1976; Lavoie, 1983). Methods based on b-glucuronidase activity are thus more specific than mFC agar. For this reason, the identity of the colonies isolated on mFC agar must be confirmed when this method is used to assess water quality.
3.3. Ability of the liquid-based methods Colilert and Readycult to confirm the identity of E. coli strains isolated from membrane filtration method on mFC agar The ability of the liquid-based methods Colilert and Readycult to detect E. coli was demonstrated by using the 305 strains isolated from Panel 1 as well as the 222 colonies randomly isolated on 59 mFC agar plates following the
Table 4 e Ability of the liquid-based methods Colilert and Readycult to confirm the identity of E. coli colonies isolated from membrane filtration method on mFC agar. Liquid-based Panel methods
False positive results
Sensitivity
Non-E. coli colonies E. coli colonies detected (%) detected (%) Colilert
Panel 1 Panel 2
0/15 (0%) 0/44 (0%)
267/290 (92.7%) 169/178 (94.9%)
Readycult
Panel 1 Panel 2
0/15 (0%) 0/44 (0%)
273/290 (94.1%) 169/178 (94.9%)
filtration of 59 different water samples during summer 2009 in Que´bec City (Panel 2). Results are presented in Table 4. Since, Colilert and Readycult are methods based on the detection of b-glucuronidase activity, there was no falsepositive results. In independent studies, Feng and Hartman (1982), Hartman (1989), Martin et al. (1993), and Moberg et al. (1988), found that 3e7% of E. coli isolates found in food and water are b-glucuronidase negative. These results are similar with those obtained in this study (5e8%). After they evaluated the detection of E. coli from 83 different water samples, Clark et al. (1991) concluded that the major disagreement between b-glucuronidase activity-based methods and the mFC method was due to the occurrence of false-negative results with b-glucuronidase activity-based methods. In the present study, the mFC agar method recovered 5e8% of E. coli strains that would not have been detected on the basis of b-glucuronidase activity.
3.4. Specificity and ability of molecular-based methods to detect E. coli strains isolated from membrane filtration method on mFC agar Even in the absence of b-glucuronidase activity, the uidA gene that encodes b-glucuronidase remains available for detection by molecular amplification, thereby allowing a faster turnaround time than that required to perform recommended phenotypic characterization tests while being more sensitive than Colilert and Readycult b-glucuronidase activity-based methods (Feng and Lampel, 1994). Maheux et al. (2009) showed that the uidA PCR assay of Bej et al. (1991) was 100% specific and ubiquitous (ability to detect all strains of targeted species) for the detection of E. coli when challenging a panel of strains composed of 79 E. coli strains of both clinical and environmental origins. The specificity and the ability of the uidA PCR of Bej et al. (1991) to detect E. coli strains isolated on mFC agar was verified by testing the 527 strains of Panels 1 and 2. The results are presented in Table 5. The uidA rtPCR assay designed by Frahm and Obst (2003) was also tested since an rtPCR assay would be more appropriate for water analysis laboratories. Indeed, rtPCR has the advantages of being more rapid, requires less manipulations, and the realization of amplification and concomitant detection in closed tubes avoids contamination of the laboratory environment with amplification products. By detecting 2.8% less E. coli strains
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 3 8 e2 6 4 6
Table 5 e Specificity and ability of molecular-based methods to detect E. coli strains isolated from membrane filtration method on mFC agar. Molecular assays (reference)
uidA PCR (Bej et al., 1991) uidA rtPCR (Frahm and Obst, 2003) tuf rtPCR (This study)
False positive results
Sensitivity
Non-E. coli colonies detected (%)
E. coli colonies detected (%)
0/59 (0%)
447/468 (95.5%)
0/59 (0%)
434/468 (92.7%)
8/59 (13.5%)a
468/468 (100%)
a False-positive results were observed with E. fergusonii and E. albertii.
than the Bej et al. uidA PCR, the Frahm and Obst uidA rtPCR assay is less sensitive for the detection of E. coli strains. Sensitivity and specificity are two important parameters of a PCR assay. However, these 2 parameters present contradictory and opposite characteristics; while specificity requires gene regions presenting traits unique to the target species, the sensitivity depends on conserved features present in all strains of the species (Maheux et al., 2009). Microbes are known to evolve rapidly. Hence, different strains of E. coli could present uidA genetic polymorphisms and the design of (rt)PCR primer pairs enabling to detect all E. coli strains could become more difficult. Using conserved genes to design primer sets constitutes a good alternative since these genes contain more conserved DNA regions. To verify this, we tested the rtPCR assay targeting the conserved tuf gene encoding for the translation elongation factor Tu (Tables 3 and 5). Among the 59 non-E. coli strains isolated, the tuf rtPCR assay yielded a positive signal for 8 strains. False-positive signals were obtained with E. fergusonii, E. hermanii, and 2 strains identified by Vitek 2 as low discrimination organisms. Based on the phenotypical results obtained by the Vitek 2, we believe that the low discrimination organisms isolated in this study could be Escherichia albertii, a new strain for which biochemical properties are not included in the Vitek database (Abbott et al., 2003). Furthermore, 16S rRNA analysis of the strain identified as E. hermanii by Vitek analysis did not match with 16S rRNA sequence of the E. hermanii type strain. Moreover, E. hermanii does not grow on mFC agar. Further, 16S rRNA analysis showed that this strain is not E. hermanii as it belongs to the E. coli/Shigella/E. fergusonii/E. albertii group. The species belonging to this group present similar 16S rRNA sequences making it difficult to distinguish them on the basis of this gene sequence only. In addition, on the basis of biochemical profiling, Walk et al. (2009) found that Escherichia species are highly similar in phenotype and cannot be easily distinguished from E. coli. In this study, we showed that E. fergusonii cells represent 1% of the colonies growing on mFC medium. According to Walk et al. (2009) and Oaks et al. (2010), E. fergusonii has been related to humans and birds whereas E. albertii has been mostly related to humans and animals, but also found in water and soil. Confirming the identity of a presumptive E. coli colony isolated on mFC agar by a molecular method can be done in
2645
less than 1 h, simply after touching the colony with a micropipette tip or a sterile toothpick, resuspending it in saline buffer, transferring 1 mL of this suspension to a pre-prepared rtPCR mixture, and putting it in an rtPCR instrument ready for a thermal cycling run of about 50 min. However, before implementing such an approach, an rtPCR assay must be thoroughly validated and the user must be aware of the false-positive and negative results associated with the selected rtPCR assay in order to make a good interpretation of the results. For the 3 molecular-based methods evaluated in this study, the Bej et al. uidA PCR assay presented a perfect specificity as it detected only E. coli strains and also a good sensitivity as it detected more than 95% of E. coli strains tested. However, a standard PCR assay is more laborious and more sensitive to contamination than an rtPCR assay done in closed tubes. The Frahm and Obst uidA rtPCR assay also had a perfect specificity. Nevertheless, its sensitivity was lower than the sensitivity of the Bej et al. uidA PCR assay. Finally, the tuf rtPCR assay correctly identified all E. coli strains isolated. However, false-positive results were obtained with E. fergusonii and E. albertii isolates. The uidA gene presents genetic polymorphisms that do not allow the design of an rtPCR assay able to detect all E. coli strains. Hence, using a conserved gene for this purpose can be a good alternative to achieve the detection of all strains. Since E. coli, E. fergusonii, and E. albertii are highly similar in phenotype and genotype, they cannot be easily discriminated from E. coli. However, E. fergusonii and E. albertii are also principally found in feces like E. coli. Therefore, their detection mostly indicates a fecal contamination. Consequently, using the 1-h tuf rtPCR assay to confirm the identity of E. coli colonies isolated on mFC agar constitutes a strategy which is as specific, more sensitive, and potentially more cost-effective than culture methods based on b-glucuronidase detection.
4.
Conclusion
In this study, we have shown that the culture-based method on mFC agar detects 5e8% more E. coli isolates in water samples than approved culture methods based on b-glucuronidase activity. We also showed that both PCR assays targeting uidA were 100% specific but failed to detect 4.5e7.3% of E. coli colonies isolated on mFC agar. Finally, the rtPCR assay targeting tuf correctly identified 100% of E. coli colonies isolated on mFC agar. However, it also detected E. fergusonii and E. albertii, two species that are highly similar in phenotype and genotype to E. coli, and also found in feces. Thus, their detection would indicate a fecal contamination. Consequently, as the occurrence of potable water samples presenting fecal contamination is very low, we suggest that it could be more advantageous for water analysis laboratories to test the microbiological quality of potable water by the cost-effective mFC agar that detects all b-glucuronidase-negative E. coli strains, and then eliminate all possibilities of false-positive results by confirming the identity of isolated blue colonies with a 1-h E. coli-specific rtPCR test, rather than systematically testing the microbiological quality of every water sample with a more expensive b-glucuronidase-based method.
2646
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 3 8 e2 6 4 6
Disclosure of interests The authors of this study do not have links with companies fabricating or commercializing water testing products.
Acknowledgments This research was supported by grant PA-15586 from the Canadian Institutes of Health Research (CIHR) and by grant FCI-5251 from Canada Foundation for Innovation (CFI). Andre´e Maheux was supported by a scholarship from Nasivvik (Center for Inuit Health and Changing Environment; Canadian Institutes for Health Research).
references
Abbott, S.L., O’Connor, J., Robin, T., Zimmer, B.L., Janda, J.M., 2003. Biochemical properties of a newly described Escherichia species, Escherichia albertii. J. Clin. Microbiol. 41, 4852e4854. American Public Health Association, 1981. Standard Methods for the Examination of Water and Wastewater, fifteenth ed. American Public Health Association, Washington, D.C. Bej, A.K., DiCesare, J.L., Haff, L., Atlas, R.M., 1991. Detection of Escherichia coli and Shigella spp. in water by using the polymerase chain reaction and gene probes for uid. Appl. Environ. Microbiol. 57, 1013e1017. Clark, D.L., Miller, B.B., Stewart, M.H., Wolfe, R.L., Olson, B.H., 1991. Comparative study of commercial 4-methylumbelliferylb-d-glucuronide preparations with the Standard methods membrane filtration fecal coliform test for the detection of Escherichia coli in water samples. Appl. Environ. Microbiol. 57, 1528e1534. Cleuziat, P., Robert-Baudouy, J., 1990. Specific detection of Escherichia coli and Shigella species using fragments of genes coding for betaglucuronidase. FEMS Microbiol. Lett. 60, 315e322. Cowburn, J.K., Goodall, T., Fricker, E.J., Walter, K.S., Fricker, C.R., 1994. A preliminary study of the use of Colilert for water quality monitoring. Lett. Appl. Microbiol. 19, 50e52. Doyle, M.P., Schoeni, J.L., 1984. Survival and growth characteristics of Escherichia coli associated with hemorrhagic colitis. Appl. Environ. Microbiol. 48, 855e856. Eckner, K.F., 1998. Comparison of membrane filtration and multiple tube fermentation by the Colilert and Enterolert methods for the detection of waterborne coliform bacteria, Escherichia coli and enterococci used in drinking and bathing water quality monitoring in southern Sweden. Appl. Environ. Microbiol. 64, 3079e3083. Edberg, S.C., Allen, M.J., Smith, D.B., 1988. National field evaluation of a defined substrate method for the simultaneous enumeration of total coliforms and Escherichia coli from drinking water: comparison with the standard multiple tube fermentation method. Appl. Environ. Microbiol. 54, 1595e1601. Feng, P.C.S., Hartman, P.A., 1982. Fluorogenic assays for immediate confirmation of Escherichia coli. Appl. Environ. Microbiol. 43, 1320e1329. Feng, P., Lampel, K.A., 1994. Genetic analysis of uidA expression in enterohaemorrhagic Escherichia coli serotype O157:H7. Microbiology 140, 2101e2107. Frahm, E., Obst, U., 2003. Application of the fluorogenic probe technique (TaqMan PCR) to the detection of Enterococcus spp. and Escherichia coli in water samples. J. Microbiol. Methods 52, 123e131.
Geldreich, E.E., Clark, H.F., Huff, C.B., Best, L.C., 1965. Fecalcoliform-organism medium for the membrane filter technique. J. Am. Water Works Assoc. 56, 208e214. Hall, N.H., Moyer, N.O., 1989. Evaluation of multiple tube fermentation test and the autoanalysis Colilert test for the enumeration of coliforms and Escherichia coli in private well water samples. In: AWWA Technol. Conf. Proc., pp. 479e496. Hartman, P.A., 1989. The MUG (glucuronidase) test for Escherichia coli in food and water. In: Balows, A., Tilton, R.C., Turano, A. (Eds.), Rapid Methods and Automation in Microbiology and Immunology. Brixia Academic Press, Brescia, Italy, pp. 290e308. Krishnan, C., Fitzgerald, V.A., Dakin, S.J., Behme, R.J., 1987. Laboratory investigation of outbreak of hemorrhagic colitis caused by Escherichia coli O157:H7. J. Clin. Microbiol. 25, 1043e1047. Lavoie, M.C., 1983. Identification of strains isolated as total and fecal coliforms and comparison of both groups as indicators of fecal pollution in tropical climates. Can. J. Microbiol. 29, 689e693. Lewis, C.M., Mak, J.L., 1989. Comparison of membrane filtration and autoanalysis Colilert presenceeabsence techniques for analysis of total coliforms and Escherichia coli in drinking water samples. Appl. Environ. Microbiol. 55, 3091e3094. Maheux, A.F., Huppe´, V., Boissinot, M., Picard, F.J., Bissonnette, L., Bernier, J.-L., Bergeron, M.G., 2008. Analytical limits of four beta-glucuronidase and beta-galactosidase-based commercial culture methods used to detect Escherichia coli and total coliforms. J. Microbiol. Methods 75, 506e514. Maheux, A.F., Picard, F.J., Boissinot, M., Bissonnette, L., Paradis, S., Bergeron, M.G., 2009. Analytical comparison of nine PCR primer sets designed to detect the presence of Escherichia coli/ Shigella in water samples. Water Res. 43, 3019e3028. Manafi, M., Kneifel, F., Bascon, S., 1991. Fluorogenic and chromogenic substrates used in bacterial diagnosis. Microbiol. Rev. 55, 335e348. Martins, M.T., Rivera, I.G., Clark, D.L., Stewart, M.H., Wolfe, R.L., Olson, B.H., 1993. Distribution of uidA gene sequences in Escherichia coli isolated in water sources and comparison with the expression of beta-glucuronidase activity in 4methylumbelliferyl-b-D-glucuronide media. Appl. Environ. Microbiol. 59, 2271e2276. Moberg, L.J., Wagner, M.K., Kellen, L.A., 1988. Fluorogenic assay for rapid detection of Escherichia coli in chilled and frozen foods: collaborative study. J. Assoc. Off. Anal. Chem. 71, 589e602. Oaks, J.L., Berrer, T.E., Walk, S.T., Gordon, D.M., Beckmen, K.B., Burek, K.A., Haldorson, G.J., Bradway, D.S., Ouellette, L., Rurangirwa, F.R., Davis, M.A., Dobbin, G., Whittam, T.S., 2010. Escherichia albertii in wild and domestic birds. Emerg. Infect. Dis. 16, 638e646. Pitkanen, T., Paakkari, P., Miettinen, I.T., Heinonen-Tanski, H., Paulin, L., Hanninen, M.L., 2007. Comparison of media for enumeration of coliform bacteria and Escherichia coli in nondisinfected water. J. Microbiol. Methods 68, 522e529. Ratnam, S., March, S.B., Ahmed, R., Bezanson, G.S., Kasatiya, S., 1988. Characterization of Escherichia coli serotype O157:H7. J. Clin. Microbiol. 26, 2006e2012. Rose, R.E., Geldreich, E.E., Litsk, W., 1975. Improved membrane filter method for fecal coliform analysis. Appl. Microbiol. 29, 532e536. Stramer, S.L., 1976. Presumptive identification of Klebsiella pneumoniae on m-FC medium. Can. J. Microbiol. 22, 1774e1776. U.S. Environmental Protection Agency (USEPA), 1978. Microbiological Methods for Monitoring the Environment: Water and Wastes. USEPA, Washington DC. Walk, S.T., Alm, E.W., Gordon, D.M., Ram, J.L., Toranzos, G.A., Tiedje, J.N., Whittam, T.S., 2009. Cryptic lineages of the genus Escherichia. Appl. Environ. Microbiol. 75, 6534e6544.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 4 7 e2 6 5 8
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Treatment of a sanitary landfill leachate using combined solar photo-Fenton and biological immobilized biomass reactor at a pilot scale Vı´tor J.P. Vilar a,*, Elisangela M.R. Rocha b, Francisco S. Mota b, Ame´lia Fonseca c, Isabel Saraiva c, Rui A.R. Boaventura a a
LSRE e Laboratory of Separation and Reaction Engineering, Departamento de Engenharia Quı´mica, Faculdade de Engenharia da Universidade do Porto, Rua Dr. Roberto Frias, 4200-465 Porto, Portugal b Universidade Federal do Ceara´, Campus do Pici, Centro de Tecnologia, Departamento de Engenharia Hidra´ulica e Ambiental, Laborato´rio de Saneamento (LABOSAN), Avenida da Universidade, 2853 e Benfica, 60020-181 Fortaleza, Brazil c Efacec Ambiente, SA, Rua Eng. Frederico Ulrich e Guardeiras, Apartado 3003, 4471-907 Moreira da Maia, Portugal
article info
abstract
Article history:
A solar photo-Fenton process combined with a biological nitrification and denitrification
Received 20 December 2010
system is proposed for the decontamination of a landfill leachate in a pilot plant using
Received in revised form
photocatalytic (4.16 m2 of Compound Parabolic Collectors e CPCs) and biological systems
7 February 2011
(immobilized biomass reactor). The optimum iron concentration for the photo-Fenton
Accepted 14 February 2011
reaction of the leachate is 60 mg Fe2þ L1. The organic carbon degradation follows a first-
Available online 22 February 2011
order reaction kinetics (k ¼ 0.020 L kJUV1, r0 ¼ 12.5 mg kJUV1) with a H2O2 consumption rate of 3.0 mmol H2O2 kJUV1. Complete removal of ammonium, nitrates and nitrites of the
Keywords:
photo-pre-treated leachate was achieved by biological denitrification and nitrification,
Sanitary landfill Leachate
after previous neutralization/sedimentation of iron sludge (40 mL of iron sludge per liter of
Solar-driven photo-Fenton
photo-treated leachate after 3 h of sedimentation). The optimum C/N ratio obtained for the
Anoxic and aerobic biological
denitrification reaction was 2.8 mg CH3OH per mg NeNO3, consuming 7.9 g/8.2 mL of
treatment
commercial methanol per liter of leachate. The maximum nitrification rate obtained was
Pilot plant
68 mg NeNH4þ per day, consuming 33 mmol (1.3 g) of NaOH per liter during nitrification and 27.5 mmol of H2SO4 per liter during denitrification. The optimal phototreatment energy estimated to reach a biodegradable effluent, considering ZahneWellens, respirometry and biological oxidation tests, at pilot plant scale, is 29.2 kJUV L1 (3.3 h of photo-Fenton at a constant solar UV power of 30 W m2), consuming 90 mM of H2O2 when used in excess, which means almost 57% mineralization of the leachate, 57% reduction of polyphenols concentration and 86% reduction of aromatic content. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
The overall aim of the Landfill Directive (Council Directive, 1999/31/EC of 26 April 1999) on the landfill of MSW (Municipal
Sludge Wastes) is “to prevent or reduce as far as possible negative effects on the environment, in particular the pollution of surface water, groundwater, soil and air, and on the global environment, including the greenhouse effect, as well as any
* Corresponding author. Tel.: þ351 918257824; fax: þ351 225081674. E-mail address:
[email protected] (V.J.P. Vilar). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.02.019
2648
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 4 7 e2 6 5 8
resulting risk to human health, from the landfilling of waste, during the whole life-cycle of the landfill” (1999/31/EC, 1999). Leachate production is inevitable due to rainwater percolation through the landfill and wastes decomposition, so a leachate treatment plant is required in a sanitary landfill. The high variability of leachate composition and quantity (Tatsi and Zouboulis, 2002; Kulikowska and Klimiuk, 2008), reinforced by the presence of high recalcitrant substances, such as humic and fulvic acids (Kang et al., 2002), xenobiotics and pesticides (Baun et al., 2004), heavy metals and inorganic macroconstituents (Ca2þ, Mg2þ, Kþ, Naþ, NH4þ, Cl, HCO3, SO42, ¨ man and Junestedt, 2008), etc.) (Christensen et al., 2001; O depending on the age of the landfill, climatic conditions, soil properties, waste type and composition (Christensen et al., 2001), is nowadays recognized as one of the greatest environmental problems in terms of water pollution. Normally, old landfill leachates (>10 years), present a low BOD5/COD ratio (<0.1), indicating low biodegradability due to the release of high molecular weight recalcitrant organic molecules and high concentration of NH3eN, which makes difficult being treated by conventional biological processes (Chian and DeWalle, 1976; Calace et al., 2001; Renou et al., 2008). The main source of ammoniacal nitrogen in leachate results from the slow anaerobic hydrolysis of proteins present in the solid wastes, resulting in a slow release of ammonia, which can reach high concentrations in stabilized landfill leachates (Jokela and Rintala, 2003). Normally, for concentrations higher than 100 mg L1 (Burton and Watson-Craik, 1998), untreated NH3eN is highly toxic to aquatic organisms, as confirmed by toxicity tests using Brachydanio rerio (Silva et al., 2004), Daphnia magna (Assmuth and Penttila¨, 1995) and Vibrio fischeri luminescent bacteria (Nohava et al., 1995). Due to its ability to enhance the biodegradability of recalcitrant compounds in the leachate, advanced oxidation processes (AOPs), using the combination of strong oxidants, e.g. O3, H2O2, irradiation, e.g. ultraviolet (UV) (Kurniawan et al., 2006), ultrasound (US) (Gonze et al., 2003) or electron beam (EB) (Deng and Englehardt, 2007), and catalysts, e.g. Fe2þ (Primo et al., 2008) and photocatalysts, e.g. TiO2 (Weichgrebe et al., 1993), to produce hydroxyl radicals, have been considered as one of the most promising options for leachate treatment, which can be improved through the use of renewable solar energy, as UV/Vis photon source. It has been demonstrated that the photo-Fenton reaction is more efficient for the treatment of the leachates (Rocha et al., 2011) than TiO2, since the reaction rate is much higher and very low iron concentrations is enough for promoting leachate treatment. The advantage of the photo-Fenton process is the higher light sensitivity up to 580 nm, corresponding to 35% of solar radiation spectrum, when compared with z5% for TiO2 photocatalysis. Biological nitrogen removal requires a two-step process: aerobic nitrification of ammonia to nitrite (Eq. 1) and then nitrite is converted to nitrate (Eq. 2); anoxic denitrification of nitrate to nitrogen gas (Eq. 3).
3 þ NHþ 4 þ O2 /NO2 þ H2 O þ 2H ðammonia oxidizing bacteriaÞ 2 (1)
1 NO 2 þ O2 /NO3 ðnitrite oxidizing bacteriaÞ 2
(2)
NO 3 /NO2 /NO/N2 O/N2
(3)
Denitrification occurs in the absence of oxygen, where nitrate or nitrite is the electron acceptor and requires a carbon source as electron donor. Normally, methanol is used as substrate when an external carbon source is necessary (Eqs. 4 and 5), however acetic acid, methane and others can be used (Randall et al., 1992). 6NO 3 þ 5CH3 OH þ CO2 /3N2 þ 6HCO3
þ 7H2 O ðnitrate removal processÞ
(4)
6NO 2 þ 3CH3 OH þ CO2 /3N2 þ 6HCO3
þ 3H2 O ðnitrite removal processÞ
(5)
Complete nitrification requires 2 mol of oxygen per mol of ammonia to be nitrified (Eqs. 1 and 2). If denitrification is to be considered after a nitrification process, partial nitrification to nitrite implies 25% less oxygen demand compared to complete nitrification, and this shortcut of the nitrate would mean a reduction in the total carbon source required for denitrification because carbon is needed for conversion of nitrate to nitrite, which can yield up to 40% savings in methanol consumption (Eqs. 4 and 5). Approximately 7.14 mg of alkalinity (as CaCO3) are consumed per mg NeNH4 oxidized and 3.57 mg (as CaCO3) of alkalinity production per mg of NeNO3 reduced, meaning that for nitrogen biological elimination in a nitrificationedenitrification cycle, 3.57 mg of alkalinity (as CaCO3) is consumed. Temperature, pH and dissolved oxygen are the main parameters that control the nitrification and denitrification processes efficiency (Randall et al., 1992). Alleman (1984) showed that the optimal pH values are between 7.9 and 8.2 for nitrification and between 7.2 and 7.6 for denitrification. Ruiz et al. (2003) studied the nitrification of synthetic wastewater with high ammonia concentration (10 g NeNH4þ L1) at a temperature of 30 C and concluded that for pH values lower than 6.45 and higher than 8.95 complete inhibition of nitrification takes place. Setting a DO concentration in the reactor at 0.7 mg L1, it was possible to accumulate more than 65% of the loaded ammonia nitrogen as nitrite with a 98% ammonia conversion, representing a reduction of 20% in the oxygen consumption. Below 0.5 mg L1 of DO, ammonia was accumulated and over a DO of 1.7 mg L1 complete nitrification to nitrate was achieved. Different methanol demands, as external carbon source, for denitrification process of landfill leachates have been reported: 2.43 g CH3OH/g NeNO3 (3.6 g COD/g NeNO3) (Kulikowska and Klimiuk, 2004), 2.8e3.0 g CH3OH/g NeNO3 (4.5e4.1 g COD/g NeNO3) (Christensson et al., 1994), which are more than 2 times higher than the stoichiometric mass ratio between consumed methanol-C and nitrate-N (C/N ¼ 0.46). Modin et al. (2007) studied the denitrification using methane as external carbon source, showing a C/N ratio of 7.1 g CeCH4/ g NeNO3, approximately seven times higher than the stoichiometric mass ratio, according to Eq. (6) (Randall et al., 1992).
2649
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 4 7 e2 6 5 8
8NO 3 þ 5CH4 /4N2 þ 8HO þ 6H2 O þ 5CO2
(6)
The objective of this study is to evaluate the possibility of combining a solar photo-Fenton process, as pre-oxidation step, to enhance the biodegradability of a landfill leachate, with a biological oxidation treatment (nitrification and denitrification) for nitrogen and residual DOC removal.
2.
Experimental methodology
2.1.
Sanitary landfill leachate samples
Leachate samples were collected at MSW sanitary landfill nearby Porto. Table 1 presents the main chemicalephysical characteristics of the leachate used in this work.
2.2.
Solar CPC pilot plant
The photo-Fenton experiments were carried out under sunlight in 100-L pilot plant installed at the roof of the Chemical Department, Faculty of Engineering, University of Porto (FEUP), Portugal. The pilot plant consists of a photocatalytic system, constituted by compound parabolic collectors (CPCs) (4.16 m2),
Table 1 e Landfill leachate characterization. Parameters pH T ( C) Redox Potential (mV) Conductivity (mS cm1) Volatile Suspended Solids (mg L1) Total Suspended Solids (mg L1) COD (mg O2 L1) BOD5 (mg O2 L1) BOD5/COD Total Dissolved Carbon (mg C L1) Inorganic Carbon (mg C L1) DOC (mg C L1) Polyphenols (mg caffeic acid L1) Dissolved Iron (mg (Fe2þ þ Fe3þ) L1) Absorbance at 254 nm (diluted 1:25) Nitrite (mg NeNO2 L1) Nitrate (mg NeNO3 L1) Ammoniacal nitrogen (mg NeNH4þ L1) Total Nitrogen (mg N L1) Phosphates (mg PO43 L1) Total Phosphorous (mg P L1) Sulfate (mg SO42 L1) Chloride (mg Cl L1) Sodium (g Naþ L1) Potassium (g Kþ L1) Copper (mg Cu2þ L1) Total chromium (mg Cr L1) Manganese (mg Mn2þ L1) Arsenic (mg As L1) Lead (mg Pb2þ L1) Zinc (mg Zn2þ L1) Cadmium (mg Cd2þ L1) Nickel (mg Ni2þ L1)
Values 7.6 21.5 98 20.7 235 337 4505 300 0.07 1158 60 1098 93.0 8.5 1.07 469 85.9 167 1780 3.2 10.8 374 3823 2.6 2.8 0.1 2.2 0.9 95.6 36.4 1.2 0.4 0.8
mounted on a fixed platform tilted 41 (local latitude), two polypropylene storage conic tanks (50 and 100 L), two recirculation pumps (20 L min1) (ARGAL model TMB), two flowmeters (Stu¨be, model DFM 165-350), polypropylene valves (FIP) and connecting tubing, being operated in batch mode. The pilot plant has also a sedimentation tank and a biological reactor system, consisting in a conditioner tank and an immobilized biomass reactor. The solar collectors are made-up of four CPC units (1.04 m2) with 5 borosilicate tubes each (Schott-Duran type 3.3, Germany, cut-off at 280 nm, internal diameter 46.4 mm, length 1500 mm and thickness 1.8 mm) connected by polypropylene junctions. The pilot plant can be operated in two ways: using the total CPCs area (4.16 m2) or using 2.08 m2 of CPCs area individually, giving the possibility of performing two different experiments at the same time and at the same solar radiation conditions. The intensity of solar UV radiation is measured by a global UV radiometer (ACADUS 85-PLS) mounted on the pilot plant at the same angle, which provides data in terms of incident WUV m2. Eq. (7) allows to obtain the amount of accumulated UV energy (QUV,n kJ L1) received on any surface in the same position with regard to the sun, per unit of volume of water inside the reactor, in the time interval Dt: QUV;n ¼ QUV;n1 þ Dtn UVG;n
Ar ; Vt
Dtn ¼ tn tn1
(7)
where tn is the time corresponding to n-water sample, Vt total reactor volume, Ar illuminated collector surface area and UVG;n average solar ultraviolet radiation measured during the period Dtn . All experiments were done from April to December 2009 during cloudy and sunny days.
2.3.
Biological oxidation system
The biological oxidation system is composed by a neutralization/sedimentation conic-bottom tank (75 L), a conditioner flat-bottom tank (50 L) and an immobilized biomass reactor (IBR) (50 L). The neutralization/sedimentation tank is equipped with a pH meter (CRISON) and a mechanical stirrer (TIMSA). The conditioner tank is equipped with a pH control unit (CRISON, electrode and PH27P controller) for pH adjustment using either H2SO4 or NaOH dosed by means of two metering pumps (DOSAPRO MILTON ROY, GTM series, model A) and a mechanical stirrer (TIMSA). The IBR is a flat-bottom container packed with 25e30 L of propylene rings (nominal diameter 50 mm), colonized by activated sludge from a municipal wastewater treatment plant (Freixo WWTP). The bioreactor is also equipped with a dissolved oxygen control unit (CRISON, electrode and OXI49P controller) and air is supplied by a blower (compressor-HAILEA model V-20; ceramic air diffuser) for maintaining the oxygen in the system in the selected range.
2.4.
Analytical determinations
Evaluation of H2O2 concentration during experiments was performed by the metavanadate method, based on the reaction of H2O2 with ammonium metavanadate in acidic medium, which results in the formation of a redeorange color peroxovanadium cation, with maximum absorbance at
2650
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 4 7 e2 6 5 8
450 nm (Nogueira et al., 2005). Iron concentration was determined by colorimetry with 1,10-phenantroline according to ISO 6332. The total polyphenols concentration was measured by spectrophotometry at 765 nm using the reagent Folin-Ciocalteau (Merck) (Folin and Ciocalteau, 1927). The polyphenols content is expressed as mg L1 of caffeic acid. Sulfate, chloride, nitrate and nitrite were measured by ion chromatography (Dionex DX-120), using a Dionex Ionpac AS9-HC 4 mm 250 mm column. The programme for anions determination comprises a 20 min run using 9 mM Na2CO3 as eluent, at a flow rate of 1.0 mL min1. Ammonium, phosphates, total phosphorous, total nitrogen and COD were determined by MerckSpectroquant kits. The quantification of total suspended solids and volatile suspended solids was carried out according to the Standard methods (Clesceri et al., 2005). Dissolved organic carbon (DOC) was measured in a TCeTOCeTN analyzer (Shimadzu, model TOC-VCSN) provided with an NDIR detector. UVeVis spectrum between 200 and 700 nm, absorbance at 450 nm (vanadate method), 510 nm (phenantroline method), 765 nm (Folin-Ciocalteau method) and 254 nm (aromatic content) were obtained using a UNICAM Helios a spectrophotometer. All samples were prefiltrated through 0.2 mm Nylon VWR membrane filters before analysis. pH, temperature, conductivity and ORP were measured using a pH meter HANNA HI 4522. Metal ions concentrations (Kþ, Naþ, Cu2þ, total Cr, Mn2þ, total As, Pb2þ, Zn2þ, Cd2þ and Ni2þ) were obtained, after a previous digestion of the leachate according to Standard Methods (Clesceri et al., 2005), by atomic absorption spectrometry (GBC 932 Plus Atomic Absorption Spectrometer and GBC GF 3000 graphite furnace system).
2.5.
Biodegradability assays
Before biological tests and other analysis involving chemical oxidation, excess H2O2 present in samples was removed using a small volume of 0.1 g L1 solution of catalase (2500 U mg1 bovine liver) after adjusting the sample pH to 6.5e7.5. Biochemical oxygen demand (BOD5) was determined according to OECD-301F test using an OxiTop (manometric respirometry), described in Standard Methods (Clesceri et al., 2005). Respirometric tests were performed using a YSI Model 5300 Biological oxygen monitor and a YSI Model 5301B thermostatic bath, allowing to measure the oxygen uptake rate (OUR) by the active sludge in the presence of the partially photo-treated samples. The respirometer cell was loaded with 5 mL of the sample saturated with air, and continuously magnetically stirred. 1 mL of activated sludge suspension from a WWTP of Porto (Portugal) was added to the sample, and the oxygen consumption (OUR) measured during 30 min. A 28 days biodegradability ZahneWellens test was performed according to the EC protocol, Directive 88/303/EEC (EPA, 1996). 250 mL of the pre-treated samples at different photo-Fenton times, without hydrogen peroxide, were added to an open glass vessel, magnetically stirred and kept in the dark at 25 C. Activated sludge from a WWTP in Porto, previously centrifuged, and mineral nutrients (KH2PO4, K2HPO4, Na2HPO4, NH4Cl, CaCl2, MgSO4 and FeCl3) were added to the
samples. The control and blank experiments were prepared using glucose as carbon source, which is highly biodegradable, and distilled water, respectively, and was also added the mineral nutrients and activated sludge. The percentage of biodegradation (Dt) was determinate by the following equation: Ct CB 100 Dt ¼ 1 CA CBA
(8)
where CA and CBA are the DOC (mg L1) in the sample and in the blank, measured 3 h after starting the experiment, Ct and CB are the DOC (mg L1) in the sample and in the blank, measured at the sampling time t. The photo-Fenton pretreated samples are considered biodegradable when Dt is higher than 70% (EMPA, 1992).
2.6.
Experimental procedure
A volume of 50 or 105 L of sanitary landfill leachate was added to the recirculation tank of the CPC units (2.08 or 4.16 m2) and homogenized by turbulent recirculation during 15 min in darkness (a first control sample was taken to characterize the wastewater). pH was adjusted with H2SO4 (Pancreac, 98% purity) to 2.6e2.9 (2 mL H2SO4 L1) to avoid iron hydroxide precipitation and another sample was taken after 15 min to confirm the pH. Afterward, iron salt (60 mg Fe2þ L1) was also added (FeSO4.7H2O, Panreac) and well homogenized for 15 min and a third sample was taken for iron concentration control. Finally, the first dose of hydrogen peroxide (30% w/v, Panreac) was added, the CPCs were uncovered and samples were taken at pre-defined times to evaluate the degradation process. In the kinetic study, the hydrogen peroxide concentration was maintained in excess, between 200 and 500 mg L1, by supplementing small amounts of H2O2 as consumed. For the biodegradability tests, a new photo-Fenton experiment was performed maintaining all the parameters, with the exception of H2O2 dose. In this case, a small amount of H2O2 was added to the photoreactor, and after H2O2 total consumption, a sample was taken for bioassays and a new dose of H2O2 was added. This procedure of “additionetotal consumptionesample collectioneaddition” is very important since it prevents any reaction in dark conditions after sample collection, during the storage and possible interferences in the bioassays. Considering this procedure, the experimental data must be expressed in terms of H2O2 consumption and not accumulated UV energy per liter of leachate. The pre-treated leachate by photo-Fenton reaction was pumped into the neutralization tank, where pH was neutralized with NaOH to a pH around 7 under stirring (1 mL NaOH 50% per L pre-treated leachate), leading to iron precipitation as Fe(OH)2, followed by a period of 3 h for sedimentation of iron sludge (40 mL of iron sludge per liter of leachate phototreated). A very small concentration of iron (<0.1 mg L1, analyzed by AAS after preliminary acid digestion) was detected in the supernatant after iron precipitation. Following this preliminary step, the neutralized photo-treated effluent was pumped to the conditioner tank, and afterward to the IBR previously colonized by activated biomass, which operates as an up-flow reactor at a recirculation of 4 L min1 (EHEIM
2651
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 4 7 e2 6 5 8
pump) between the conditioner tank and the IBR and pH is controlled in a range of 6.5e7.5.
Leachate characterization
3.3. Evaluation of combined biological and photo-Fenton treatment
Table 1 presents the main chemical/physical characteristics of the leachate after a pre-treatment in a reception/equalization lagoon with pure oxygen injection, an activated sludge reactor with an anoxic and aerobic zone and a secondary clarifier. The pre-treated effluent presents a strong dark-brown colour associated with a high organic charge (DOC ¼ 1098 mg C L1; COD ¼ 4505 mg O2 L1), high nitrogen content (1780 mg N L1), high polyphenols concentration (93.0 mg caffeic acid L1) and a low BOD5/COD (0.07) ratio, indicating the low biodegradability of the leachate. The leachate presents also a high aromatic content measured as absorbance at 254 nm (the wavelength at which the aromatic compounds present maximum absorption) (Mrkva, 1983). Another relevant point is the high conductivity attributed to the high concentration of chloride, sulfate, ammonium ions, potassium and sodium. The effluent contains also a high concentration of suspended, dissolved and volatile solids. Heavy metals were relatively low, except total chromium.
3.2.
Solar photo-Fenton process
Rocha et al. (2011) showed that the photo-Fenton reaction is faster than the Fenton reaction and the heterogeneous photocatalysis with TiO2, TiO2/H2O2/UV or homogeneous with H2O2/UV. The optimum iron concentration for the photoFenton treatment of this leachate is 60 mg Fe2þ L1 (Fig. 1). The photo-Fenton kinetics shows a slow initial reaction rate, followed by a first-order kinetic behavior (k ¼ 0.020 L kJUV1,
In order to assess the biocompatibility of the pre-treated effluent, different biodegradability tests, such as ZahneWellens and activated sludge respirometry, were performed at different stages of the solar photo-Fenton reaction. Fig. 2 presents the evolution of DOC, COD and two parameters, AOS (average oxidation state) and COS (carbon oxidation state), which can be used to evaluate the oxidation degree and efficiency of the oxidative process, respectively (Amat et al., 2007; Arques et al., 2007): COD AOS ¼ 4 1:5 DOC
(9)
COD COS ¼ 4 1:5 DOC0
(10)
where DOC is the dissolved organic carbon at time t (mg of C L1), DOC0 is the initial dissolved organic carbon of the solution (mg of C L1) and COD is the chemical oxygen demand at time t (mg of O2 L1). AOS takes values between þ4 for CO2, the most oxidized state of C, and 4 for CH4, the most reduced state of C. The AOS only takes into consideration the organic matter in the solution. In COS calculation, CO2 eliminated from the solution (with an oxidation state þ4) is also taken into account (Amat et al., 2007; Arques et al., 2007). The COD concentration decreases 89% (from 4348 to 477 mg O2 L1), showing a strong oxidation of the organics, which is well correlated with the COS parameter, which increased from 1.9, indicating the presence of rather reduced
350
Abs. 254 nm H2O2 cons. DOC Fe
RAD-ON 1000
300
DOC (mg/L)
800
250
200
600
150 400 100 200 50
0
0 -20
0
20
40
60
80
70
1,2
1,0
0,8
0,6
0,4
0,2
0,0
60
50
40
30
Fe (mg/L)
3.1.
Absorbance at 254 nm (dil. 1:25)
Results and discussion
H2O2 consumed (mM)
3.
r0 ¼ 12.5 mg kJUV1), with a H2O2 consumption rate of 3.0 mmol H2O2 kJUV1, and finally, a reaction period characterized by a lower DOC degradation and H2O2 consumption until the end of the experiment, presumably due to formation of lowmolecular carboxylic groups (Kavitha and Palanivelu, 2004).
20
10
0
100
QUV (kJ/L) Fig. 1 e DOC degradation of landfill leachate, H2O2 consumed, iron concentration and aromatic content (absorbance at 254 nm) during the photo-Fenton reaction as a function of amount of accumulated UV energy per liter of effluent.
2652
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 4 7 e2 6 5 8
4 3
3000
2 1
2400 0 1800
600
-1 S 1 S 2 S 3
-2 S S 4 5
S 6
S 7
S 8
-3
S 9
S 10
S 11
0 0
30
60
90
120
150
180
-4 210
H2O2 consumed (mM)
Fig. 2 e DOC, COD, AOS, and COS evolution as a function of the hydrogen peroxide consumption during the photoFenton process. organic compounds, to þ3.4, which means strong mineralization and generation of highly oxidized intermediates. Fig. 2 also shows that AOS starts at 1.9, increases rapidly to approximately 0.28 due to acidification, decreasing again to the initial value and then increases to 0.25 and remains almost constant during the phototreatment. The increase of AOS suggests that more oxidized organic intermediates are formed during the treatment and, after AOS reaches a plateau, the chemistry of the intermediates generated does not vary significantly (Sarria et al., 2002). The BOD5/COD ratio has been established as a more reliable parameter to evaluate the biodegradability (Marco et al., 1997; Esplugas et al., 2004; Metcalf and Eddy, 2005), as it is not affected either by the amount or by the oxidation state of organic matter (Amat et al., 2009). Fig. 3 shows a very
1200
1000
60
6
50
0.4 5
A
800
DOC (mg/L)
7
0.5
OUR/OUR0 BOD5/COD Fe DOC
RAD-ON
40 0.3
600 0.2 400
BOD5/COD
1200
4
3
30
Fe (mg/L)
COD DOC AOS COS
AOS, COS
DOC, COD (mg/L)
3600
important increase of BOD5/COD ratio (from 0.07 to 0.44), suggesting that photo-oxidation enhanced significantly the biodegradability. The OUR/OUR0 profile shows also the same trend, showing a high increase after 155 mM of H2O2 consumed (OUR0 is the oxygen uptake rate for the raw leachate used in this work). At the end of the phototreatment the OUR increased approximately six times when compared with the raw leachate (OURraw leachate ¼ 1.04 103 mol O2 mol1 C h1), indicating that photo-oxidation treatment leads to more biodegradable organic carbon, which can be assimilated by the activated sludge. Polyphenols concentration and aromatic content given by absorbance at 254 nm after dilution 1:25 (Fig. 4), shows a similar profile, leading to 78% reduction after 155 mM of H2O2 consumed, and only more 12% until the end of the experiment (204 mM). According to the ZahneWellens test (Fig. 5), which involves longer periods of contact (28 days) of the sample with microorganisms to allow some adaptation of the biomass, the first four samples (non-treated, after pH adjustment, after iron addition and after 22 mM H2O2 consumed) present a poor biodegradation level, between 14 and 20%. However, as expected, the biodegradability of the leachate was enhanced during the photo-Fenton treatment and a value higher than 70% biodegradation after 28 days was achieved for sample 10. Although sample 10 seems to be the best point to stop the preoxidation process, samples 7, 8 and 9 also present high biodegradability levels of 61%, 66% and 68%, respectively, which corresponds to savings in the UV energy requirements of 50%, 41% and 24% and H2O2 consumption of 54%, 47% and 22%, relatively to sample 10. The high concentration of nitrogen in landfill leachates constitutes, beyond the presence of recalcitrant carbon compounds, a big environmental concern that must be solved. The concentration profiles of nitrate, nitrite, ammonium and total nitrogen during the photo-Fenton reaction are shown in Fig. 6. The total nitrogen concentration, after a first decrease
OUR/OUR0
RAD-ON
4200
20 2
0.1
200
1
0 0
30
60
90
120
150
180
0.0 210
0
10
0
H2O2 (mM) Fig. 3 e DOC, BOD5/COD ratio, OUR/OUR0 and Fe concentration evolution as a function of the hydrogen peroxide consumption during the photo-Fenton process.
2653
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 4 7 e2 6 5 8
40
800
DOC (mg/L)
80
A
60 600 40
400
20
200
1,0
0,8
0,6
0,4
0,2
35
30
T (°C)
100
1000
Absorbance 254 nm (dil. 1:25)
Polyphenols DOC Abs. 254 nm T
Polyphenols (mg caffeic acid/L)
1200
25
RAD-ON 0 210
0 0
30
60
90
120
150
180
0,0
20
H2O2 (mM) Fig. 4 e DOC, polyphenols concentration, absorbance at 254 nm and temperature evolution as a function of the hydrogen peroxide consumption during the photo-Fenton process.
80
70
70
60
60
50
50
40
40
30
30
20
20
10
10
1800
1800
1500
1500
1200
1200
+
80
8
12
16
20
24
28
900
900 -
NO2
Fig. 5 e ZahneWellens test for samples taken during the photo-Fenton process (initial sample is also showed, as the reference): 5 e S1, DOC [ 1098 mg LL1; 9 e S2, DOC [ 865 mg LL1; > e S3, DOC [ 688 mg LL1; A e S4, DOC [ 650 mg LL1; + e S5, DOC [ 678 mg LL1; 7 e S6, DOC [ 647 mg LL1; ; e S7, DOC [ 470 mg LL1; B e S8, DOC [ 374 mg LL1; C e S9, DOC [ 315 mg LL1; 6 e S10, DOC [ 213 mg LL1; : e S11, DOC [ 158 mg LL1; - e Reference.
-
600
-
NO3
-
+
NO2+NO3+NH4
600
+
NH4
-
Time (days)
NTOTAL
-
4
-
0
-
0
0
N-NO2; N-NO3;N-NH4; NTOTAL (mg/L)
90
+
100
90
N-NO2; N-NO3;N-NH4; NTOTAL (mg/L)
100
process. However, almost all nitrite were easily oxidized to nitrate after acidification. The leachate presents initially a low concentration of nitrates (86 mg NeNO3 L1) that increased abruptly to w1178 mg NeNO3 L1 at sample 4 (22 mM of H2O2 consumed) and remained approximately constant until the end of the experiment. This behavior is explained by the nitrite conversion into nitrate and by the conversion of organic nitrogen into ammoniacal nitrogen, which remained approximately constant between 115 and 154 mg NeNH4þ L1, and then into NO3. The oxidation of ammonia to nitrate was also observed during the photo-Fenton process applied to a biorecalcitrant industrial compound, a-methylphenylglycine (Oller et al., 2007). After sample 6, the organic nitrogen remained approximately constant (360 mg N L1), which corresponds to the
Dt (%)
Dt (%)
during the acidification process, maybe due to retention of nitrogenated compounds in the foam formed, returned to the initial value (between sample 4 and 5) and remained approximately constant during the photo-Fenton reaction (z1.7 g N L1). The high concentration of nitrite (z469 mg NeNO2 L1) can be justified by the nitrification/denitrification stages in the biological system installed in the sanitary landfill, taking into consideration that leachate samples used for the experiments performed in this work were collected after the biological
300
300
0
0 1
2
3
4
5
6
7
8
9
10
11
Sample Number
Fig. 6 e Profile of nitrate, nitrite, ammoniacal nitrogen and total nitrogen concentration during the photo-Fenton process.
2654
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 4 7 e2 6 5 8
leachate (DOC ¼ 324 mg L1) and 10 L of suspended biomass resulted from the inoculation suspension (DOC ¼ 1150 mg L1, resulting from the use of methanol during the adaptation period), were added to the biological reactor. Fig. 8 presents the evolution of DOC, NH4þ, NO3 and total nitrogen concentration during the denitrification process. It can be seen an increase in DOC due to the addition of 10 L of suspended biomass and a decrease of total nitrogen and nitrates due to dilution and adsorption of nitrogenated organic compounds on the biomass. During the first three days, almost no denitrification was observed, but almost all the DOC present initially in the pre-treated leachate was consumed. After the first addition of methanol (370 mL, 0.96 g cm3) on the third day (point a), a decay ratio of 312 mg C day1 and 175 mg NeNO3 day1, which is equivalent to 4.75 mg CH3OH per mg NeNO3 (2.48 mg NO3 per mg C), was observed. Fig. 8 also shows a small decay of ammonia along the denitrification process. Between the 9th and 14th days the DOC remained almost constant, meaning that this residual organic carbon (120 mg C L1) is not biodegradable by this kind of biomass. Another addition of methanol (25 mL, point b) was enough to eliminate all the nitrates present in the phototreated leachate. The second experiment was performed with a phototreated leachate (446 mg L1, point 7), in order to study the denitrification and nitrification processes and to evaluate the biodegradability of all organic carbon. In this case, the mass ratio between methanol and nitrate obtained was to 3.6 mg CH3OH per mg NeNO3 achieving the same final residual DOC of 120 mg L1. After complete removal of nitrates (1st denitrification step), aeration was promoted in order to have sufficient oxygen to perform a nitrification cycle. This test was performed in winter, at an average temperature of 13 C. In these conditions almost no nitrification occurred during 20 days. So, it was decided to collect 4 L of the wastewater and perform the study at controlled temperature (23 C) and aeration conditions.
difference between the total nitrogen and the sum of nitrites, nitrates and ammoniacal nitrogen. It seems that between sample 1 and 6 (50.2 mM of H2O2 consumed), 700 mg of organic nitrogen per liter was converted into ammonia, and then into nitrite and nitrate. Approximately 90% of the organic nitrogen converted to ammonia was due to leachate acidification to pH 2.8. The concentration profile of Cl, SO42, PO43 ions and total phosphorous are presented in Fig. 7. The chloride concentration in solution decreased sharply from 3.8 g L1 to 2.2 g L1, after the acidification process, possible due to the retention in the foam formed, but it was gradually released into the solution, remaining almost constant during the photo-Fenton reaction at 3.2 g L1. The sulfate concentration increased drastically after the addition of sulfuric acid for the acidification of leachate to pH z 2.8 and addition of iron sulfate, and remained constant during the reaction (2.6 g L1). The concentration of total phosphorous and phosphates remained approximately constant during the photo-Fenton treatment at 1 mg PePO43 and 11.6 mg P L1, respectively.
3.4. Evaluation of biological nitrification and denitrification In order to evaluate the optimal phototreatment time to reach a biodegradable effluent and nitrogen removal by a biological nitrificationedenitrification system, another photo-Fenton experiment was carried out in order to get a photo-treated leachate with a residual organic carbon of 446 mg L1, 324 mg L1 and 220 mg L1, which corresponds to the more biodegradable samples, 7, 9 and 10 of the ZahneWellens test. After the pre-oxidation, the pre-treated leachate was neutralized to pH 7, leading to iron precipitation as Fe(OH)2, and after iron sludge removal, the wastewater was introduced into the biological system constituted by a conditioner tank and IBR previously colonized by activated sludge from a municipal WWTP. A volume of 60 L of the pre-treated
3000
4000 3500
18
18
16
16
14
14
2500
Chloride Sulphate Total phosphorous Phosphate
2000 1500
1500
1000 1000
10 8 6 4
12 10 8 6 4
500 500 0
0 1
2
3
4
5
6
7
8
9
10
2
2
0
0
Total phosphorous (mg P/L)
3-
12
2-
2000
Sulphate (mg SO4 /L)
-
Chloride (mg Cl /L)
3000
Phosphate (mg P-PO4 /L)
2500
11
Sample Number Fig. 7 e Profile of chloride, sulfate, phosphate and total phosphorous concentration during the photo-Fenton process.
2655
1000
600
800 600
400
400
a
200
200
-
b
175 mg N-NO 3 /day 0
0 0
3
6
9
12
15
Time (days) Fig. 8 e Evolution of total nitrogen, nitrate, nitrite, ammonia and DOC concentration during the biological denitrification of the photo-treated leachate (initial DOC of 324 mg LL1).
Ilies and Mavinic (2001) showed that decreasing the temperature from 20 to 10 C, nitrification and denitrification processes suffered major inhibition. Isaka et al. (2007) showed that using nitrifying bacteria entrapped in a gel carrier increased the nitrification rates at low temperatures (0.71 kg N m3 day1 at 10 C, DO > 7 mg L1, 7.5 < pH < 8.0). However, ammonium nitrogen at low concentrations was detected in the effluent, showing that cultivated ammonium-oxidizing bacteria at low temperatures had low affinity for ammonium nitrogen. Fig. 9 shows that after four days of adaption period, the ammonia removal rate was 87 mg NH4þ per day (68 mg NeNH4þ per day), converted into 90.1 mg NO3 per day (20.3 mg NeNO3 per day) and 33 mg NO2 per day (10 mg
350
250
+
DOC + NH4 NO3 NO2
68 mg N-NH4 /day 300
70
350
60
200 300
DOC (mg/L)
-
250
400
-
3,4 g NO3 /g C
200
150
150
100
100 50 50
b
a
0
0 0
4
8
12
16
20
250 200 150 100
50 -
312 mg C/day
Nitrites (mg NO2 /L)
DOC (mg/L)
1200 800
+
1400
Ammonium (mg N-NH4 /L)
1000
1600
NeNO2 per day), which leads to the ratios 3.4 mg NeNH4þ mg1 NO3 and 6.8 mg NeNH4þ mg1 NeNO2. At the end of the nitrification reaction 305 mg NeNH4þ L1 were converted into 43 mg NeNO2 L1 and 178 mg NeNO3 L1, which means that approximately 170 mg N were converted into N2, consuming 33 mmol of NaOH per liter (1.3 g NaOH per liter). After the nitrification cycle, a second denitrification cycle was performed adding 3.4 mL of methanol at point a and 0.2 mL of methanol at point b, obtaining a denitrification ratio of approximately 3 mg CH3OH per mg of (NeNO3 þ NeNO2), which indicates that less carbon is needed for reduction of nitrites instead of nitrates. Canziani et al. (2006) studied the nitrification and denitrification of old landfill leachates (0.2e4 g NeNH4þ L1, T ¼ 33 C, pH ¼ 8.0) showing that for values of DO between 0.2 and 0.5 mg L1, 90% oxidation of ammonia to nitrite was achieved. The impact of organic carbon on nitrification performance performed by Ling and Chen (2005) showed that 60e70% nitrification rate reduction was observed when COD/N ration increased from 0 to 3 (T ¼ 20 C). During nitrification, DOC removal is very slow, however, some of the organic carbon from the photo-treated leachate is eliminated during denitrification, leading to a final DOC of 86 mg L1 and final COD of 227 mg O2 L1. Denitrification of the leachate consumed 27.5 mmol H2SO4 per liter. A last denitrification/nitrification experiment was performed in a 4 L- suspended biomass reactor at temperature (22 C) and aeration controlled conditions, using a phototreated leachate with a residual DOC of 240 mg L1, which is equivalent to sample 10 of the biodegradability tests (Fig. 10). In this case the C/N ratio obtained for the denitrification reaction was 2.8 mg CH3OH per mg NeNO3 (points a, b, c and d indicate the addition of 15, 5, 5 and 2.5 mL of methanol), which is similar to the ratios reported by Kulikowska and Klimiuk (2004), 2.43 g CH3OH g1 NeNO3 (3.6 g COD g1 NeNO3), and Christensson et al. (1994), 2.8e3.0 g CH3OH g1 NeNO3 (4.5e4.1 g COD g1 NeNO3), although those values are more than 2 times higher than the stoichiometric mass ratio between consumed
Nitrates (mg N-NO3 /L)
1200
1800
Total Nitrogen + NH 4 NO 3 DOC
addition of biomass
Total Nitrogen, Ammonium, Nitrates (mg N/L)
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 4 7 e2 6 5 8
40
30
20
50
10
0
0
24
Time (days) Fig. 9 e Evolution of total nitrogen, nitrate, nitrite, ammonium and DOC concentration during the denitrification/nitrification of the photo-treated leachate (initial DOC of 446 mg LL1).
2656
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 4 7 e2 6 5 8
1250
600
500
600
b 500 400 250
300 250 200 150
c
0 20
30
40
50
125 100 75 50 25
50
a 10
350
150
100
d
0
400
-
750
200
175
Nitrites (mg NO2 /L)
450
-
800
DOC (mg/L)
200
550 1000
+
1000
225
650
Ammonium (mg N-NH4 /L)
DOC + NH4 NO3 NO2
Nitrates (mg N-NO3 /L)
1200
60
70
0 80
0
0
Time (days) Fig. 10 e Evolution of total nitrogen, nitrate, nitrite, ammonium and DOC concentration during the denitrification/ nitrification of the photo-treated leachate (initial DOC of 220 mg LL1).
methanol-C and nitrate-N (C/N ¼ 0.46) (Eqs. 4 and 5). Modin et al. (2007) showed that methane can be a potentially inexpensive carbon source for the denitrification of leachates, since methane is generated in landfills, although the use of methane for electricity generation, presence of hydrogen sulphide impurities and indirect methane oxidation when used for denitrification (Thalasso et al., 1997) can be negative points. Considering the ratio above mentioned, the total denitrification of nitrates, considering the nitrates formed during the nitrification process, requires 4.4 g of methanol per liter of leachate or 4.6 mL of commercial methanol per liter of leachate. It was observed a parallel nitrification reaction with a decay of 11.7 mg NeNH4þ per day even at very low oxygen concentrations (<0.2 mg O2 L1). The absence of biodegradable carbon between days 30 and 50, stopped the denitrification
reaction. Oxygen concentration was maintained between 0.5 and 2.0, and nitrification of the remaining ammonia is observed and consequently lead to the formation of nitrates and nitrites. After total elimination of ammonia, two more additions of methanol (points c and d), were sufficient for complete denitrification. In this experiment the lowest dissolved organic carbon concentration obtained was 83 mg L1. According to these results, complete removal of ammoniacal nitrogen, nitrates and nitrites is possible for this pretreated leachate and sample 7 of the phototreatment can be considered the best one for the combination with a biological oxidation process. So, the optimal phototreatment energy estimated to reach a biodegradable effluent is 29.2 kJUV L1 (3.3 h of photo-Fenton at a constant solar UV power of 30 W m2), consuming 90 mM of H2O2 when used in excess,
Table 2 e Landfill leachate characteristics at the best phototreatment time and discharge limits (Decree n. 236/98). Parameter pH BOD5, 20 C (mg O2 L1) COD (mg O2 L1) DOC (mg C L1) Nitrite (mg NeNO2 L1) Nitrate (mg NO3 L1) Ammoniacal nitrogen (mg NeNH4þ L1) Total Nitrogen (mg N L1) Sulfate (g SO42 L1) Total phosphorous (mg P L1) Phosphates (mg PO42 L1) Polyphenols (mg caffeic acid L1) Chloride (g Cl L1) Dissolved iron (mg L1) 3
LOPT
LPBT
DLPL
DLBL
2.9 260 1174 470 0 5300 149 1280 2.6 11.4 2.9 59 3.2 39.3
7.0 e 227 83 <0.5 <5 <1 <10 2.6 7.0 2.0 e 3.2 <0.05*
6.0e9.0 40 150 e e 50 10 15 2 10 e e e 2*
6.0e8.5 120** 360** e e e e 10 1 e e e 10
DLUL
DLCHL
6e9 220 e e e e 10 e e e e e e e
5.5e8.5 20 60 e e e 12 20 e 1.5 e e e 5
*Total iron; **Flow rates between 200 and 1000 m /day; LOPT-Leachate at the Optimum Phototreatment time; LPBT-Leachate Photo-Bio-Treated; DLPL-Discharge Limits in the Portuguese legislation (Decree no 236/98); DLUL-Discharge Limits in the Brazilian legislation (Portaria 05/89, 1989); DLUL-Leachate Discharge Limits in USA Legislation (EPA, 2000); DLCHL-Discharge Limits in Chinese Legislation (Bureau, 2002).
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 4 7 e2 6 5 8
which means almost 57% mineralization of the leachate, 57% reduction of polyphenols concentration and 86% reduction of aromatic content. Although total nitrogen removal was achieved using a nitrification/denitrification process, aeration and an external organic carbon source are needed. A promising process, called Anammox, can save up to 90% of operation costs as compared to the traditional nitrogen removal process (Jetten et al., 2001), since in this case microbial oxidation of ammonium with nitrite to nitrogen gas occurs under strict anoxic conditions, and no organic carbon source is needed (Graaf et al., 1996; Strous et al., 1998). Liu et al. (2010) used shortcut nitrification combined with Anammox for treating diluted effluent from an UASB reactor fed by landfill leachate and found an average total inorganic nitrogen removal efficiency of 87%. Table 2 compares the characteristics of the photo-biotreated leachate with the discharge limits according to the Portuguese, Chinese, Brazilian and USA Legislation. Sulfate ions concentration greatly exceeds the discharge limits imposed by Portuguese legislation. COD is also another parameter that exceeds the discharge limits imposed by Portugal and China, although accomplishes Brazilian regulations. However, considering the complex nature of leachates, higher discharge limits can be stipulated, as recommended by EPA, that allows BOD5 values of 220 mg O2 L1.
4.
Conclusions
Leachate from landfills usually has high content of pollutants, especially recalcitrant organic compounds, as humic and fulvic acids, xenobiotics, pesticides, and other harmful substances, that result in environment pollution problems. The solar photo-Fenton process was found to be very efficient in the treatment of leachates, enhancing the biodegradability of the leachate and making possible a subsequent treatment by a biological oxidation process. Biological nitrogen removal was achieved by a two-step process: aerobic nitrification of ammonia to nitrite and then to nitrate followed by anoxic denitrification of nitrate to nitrite, nitric oxide, nitrous oxide and nitrogen gas using an external carbon source.
Acknowledgments Financial support for this work was in part provided by LSRE financing by FEDER/POCI/2010 and EFACEC Ambiente SA.
references
1999/31/EC, 1999. Council directive 1999/31/EC of 26 April 1999 on the landfill of waste. Off. J. Eur. Communities (in European Union). Alleman, J.E., 1984. Elevated nitrite occurrence in biological wastewater treatment systems. Water Sci. Technol. 17, 409e419. Amat, A.M., Arques, A., Galindo, F., Miranda, M.A., SantosJuanes, L., Vercher, R.F., Vicente, R., 2007. Acridine yellow as solar photocatalyst for enhancing biodegradability and
2657
eliminating ferulic acid as model pollutant. Appl. Catal. B. 73 (3e4), 220e226. Amat, A.M., Arques, A., Garcı´a-Ripoll, A., Santos-Juanes, L., Vicente, R., Oller, I., Maldonado, M.I., Malato, S., 2009. A reliable monitoring of the biocompatibility of an effluent along an oxidative pre-treatment by sequential bioassays and chemical analyses. Water Res. 43 (3), 784e792. Arques, A., Amat, A.M., Garcia-Ripoll, A., Vicente, R., 2007. Detoxification and/or increase of the biodegradability of aqueous solutions of dimethoate by means of solar photocatalysis. J. Hazard. Mater. 146 (3), 447e452. Assmuth, T., Penttila¨, S., 1995. Characteristics, determinants and interpretations of acute lethality in daphnids exposed to complex waste leachates. Aquat. Toxicol. 31 (2), 125e141. Baun, A., Ledin, A., Reitzel, L.A., Bjerg, P.L., Christensen, T.H., 2004. Xenobiotic organic compounds in leachates from ten danish msw landfillsechemical analysis and toxicity tests. Water Res. 38 (18), 3845e3858. Bureau C. N. E. P, 2002. Discharge Standards of Municipal Wastewater Treatment Plants in China issued by Chinese national environment protection bureau, GB18918. Burton, S.Q., Watson-Craik, I.A., 1998. Ammonia and nitrogen fluxes in landfill sites: applicability to sustainable landfilling. Waste Manage. Res. 16, 41e53. Calace, N., Liberatori, A., Petronio, B.M., Pietroletti, M., 2001. Characteristics of different molecular weight fractions of organic matter in landfill leachate and their role in soil sorption of heavy metals. Environ. Pollut. 113 (3), 331e339. Canziani, R., Emondi, V., Garavaglia, M., Malpei, F., Pasinetti, E., Buttiglieri, G., 2006. Effect of oxygen concentration on biological nitrification and microbial kinetics in a cross-flow membrane bioreactor (MBR) and moving-bed biofilm reactor (MBBR) treating old landfill leachate. J. Memb. Sci. 286 (1e2), 202e212. Chian, E.S.K., DeWalle, F.B., 1976. Sanitary landfill leachates and their treatment. J. Environ. Eng. Div., 411e431. Christensen, T.H., Kjeldsen, P., Bjerg, P.L., Jensen, D.L., Christensen, J.B., Baun, A., Albrechtsen, H.-J., Heron, G., 2001. Biogeochemistry of landfill leachate plumes. Appl. Geochem. 16 (7e8), 659e718. Christensson, M., Lie, E., Welander, T., 1994. A comparison between ethanol and methanol as carbon sources for denitrification. Water Sci. Technol. 30, 83. Clesceri, L.S., Greenberg, A.E., Eaton, A.D., 2005. Standard Methods for Examination of Water & Wastewater, 21st ed. American Public Health Association (APHA), American Water Works Association (AWWA) & Water Environment Federation (WEF). Deng, Y., Englehardt, J.D., 2007. Electrochemical oxidation for landfill leachate treatment. Waste Manage. (Oxford) 27 (3), 380e388. EMPA, 1992. OCDE Guideline for Testing of Chemicals. adopted by the council on 17th july 1992, Zahn-Wellens/EMPA test. Swiss Federal Laboratories for Materials Testing and Research. EPA, 1996. U.S. Environmental Protection Agency, Prevention Pesticides and Toxic Substances (7101). Fates; Transport and Transformation Test Guidelines Oppts 835.3200 ZahnWellens/EMPA Test EPA 712-C-96e084, Washington, DC. EPA, 2000. Code of Federal Regulations (CFR): Protection of the Environment chapter I, subchapter N, effluent guidelines and standard, 445 landfill point source category. Esplugas, S., Contreras, S., Ollis, D.F., 2004. Engineering aspects of the integration of chemical and biological oxidation: simple mechanistic models for the oxidation treatment. J. Environ. Eng. 130 (9), 967e974. Folin, O., Ciocalteau, V., 1927. On tyrosine and tryptophane determinations in proteı´ns. J. Biol. Chem. 73, 627e650. Gonze, E., Commenges, N., Gonthier, Y., Bernis, A., 2003. High frequency ultrasound as a pre- or a post-oxidation for paper
2658
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 4 7 e2 6 5 8
mill wastewaters and landfill leachate treatment. Chem. Eng. J. 92 (1e3), 215e225. Graaf, A.A.V.d., Bruijn, P.d., Robertson, L.A., Jetten, M.S.M., Kuenen, J.G., 1996. Autotrophic growth of anaerobic ammonium-oxidizing microorganisms in a fluidized bed reactor. Microbiology 142 (8), 2187e2196. Ilies, P., Mavinic, D.S., 2001. The effect of decreased ambient temperature on the biological nitrification and denitrification of a high ammonia landfill leachate. Water Res. 35 (8), 2065e2072. Isaka, K., Yoshie, S., Sumino, T., Inamori, Y., Tsuneda, S., 2007. Nitrification of landfill leachate using immobilized nitrifying bacteria at low temperatures. Biochem. Eng. J. 37 (1), 49e55. Jetten, M.S.M., Wagner, M., Fuerst, J., van Loosdrecht, M., Kuenen, G., Strous, M., 2001. Microbiology and application of the anaerobic ammonium oxidation (‘anammox’) process. Curr. Opin. Biotechnol. 12 (3), 283e288. Jokela, J.P.Y., Rintala, J.A., 2003. Anaerobic solubilisation of nitrogen from municipal solid waste (MSW). Rev. Environ. Sci. Biot. 2 (1), 67e77. Kang, K.-H., Shin, H.S., Park, H., 2002. Characterization of humic substances present in landfill leachates with different landfill ages and its implications. Water Res. 36 (16), 4023e4032. Kavitha, V., Palanivelu, K., 2004. The role of ferrous ion in Fenton and photo-Fenton processes for the degradation of phenol. Chemosphere 55 (9), 1235e1243. Kulikowska, D., Klimiuk, E., 2004. Removal of organics and nitrogen from municipal landfill leachate in two-stage SBR reactors. Pol. J. Environ. Stud. 13 (4), 389e396. Kulikowska, D., Klimiuk, E., 2008. The effect of landfill age on municipal leachate composition. Bioresour. Technol. 99 (13), 5981e5985. Kurniawan, T.A., Lo, W.-h., Chan, G.Y.S., 2006. Radicals-catalyzed oxidation reactions for degradation of recalcitrant compounds from landfill leachate. Chem. Eng. J. 125 (1), 35e57. Ling, J., Chen, S., 2005. Impact of organic carbon on nitrification performance of different biofilters. Aquacult. Eng. 33 (2), 150e162. Liu, J., Zuo, J.E., Yang, Y., Zhu, S., Kuang, S., Wang, K., 2010. An autotrophic nitrogen removal process: short-cut nitrification combined with anammox for treating diluted effluent from an uasb reactor fed by landfill leachate. J. Environ. Sci. 22 (5), 777e783. Marco, A., Esplugas, S., Saum, G., 1997. How and why combine chemical and biological processes for wastewater treatment. Water Sci. Technol. 35 (4), 321e327. Metcalf and Eddy, 2005. Wastewater Engineering Treatment and Reuse, fourth ed. Metcalf & Eddy. Modin, O., Fukushi, K., Yamamoto, K., 2007. Denitrification with methane as external carbon source. Water Res. 41 (12), 2726e2738. Mrkva, M., 1983. Evaluation of correlations between absorbance at 254 nm and COD of rivers waters. Water Res. 17, 231e235. Nogueira, R.F.P., Oliveira, M.C., Paterlini, W.C., 2005. Simple and fast spectrophotometric determination of H2O2 in photoFenton reactions using metavanadate. Talanta 66 (1), 86e91.
Nohava, M., Vogel, W.R., Gaugitsch, H., 1995. Evaluation of the luminescent bacteria bioassay for the estimation of the toxicological potential of effluent water samples e comparison with data from chemical analyses. Environ. Int. 21 (1), 33e37. Oller, I., Malato, S., Sanchez-Perez, J.A., Gernjak, W., Maldonado, M.I., Perez-Estrada, L.A., Pulgarin, C., 2007. A combined solar photocatalytic-biological field system for the mineralization of an industrial pollutant at pilot scale. Catal. Today 122 (1e2), 150e159. ¨ man, C.B., Junestedt, C., 2008. Chemical characterization of O landfill leachates e 400 parameters and compounds. Waste Manage. (Oxford) 28 (10), 1876e1891. Portaria 05/89, 1989, Secretaria da sau´de e do meio ambiente, crite´rios e padro˜es de efluentes lı´quidos a serem observados por todas as fontes poluidoras que lancem seus efluentes nos corpos d’a´gua interiores do estado do Rio Grande do Sul. Primo, O., Rivero, M.J., Ortiz, I., 2008. Photo-Fenton process as an efficient alternative to the treatment of landfill leachates. J. Hazard. Mater. 153 (1e2), 834e842. Randall, C.W., Barnard, J.L., Stensel, H.D., 1992. Design and retrofit of wastewater treatment plants for biological nutrient removal. In: Eckenfelder, W.W., Malina, J.F., Patterson, J.W. (Eds.). Technomic Publishing AG, Lancaster, Pensylvania. Renou, S., Givaudan, J.G., Poulain, S., Dirassouyan, F., Moulin, P., 2008. Landfill leachate treatment: review and opportunity. J. Hazard. Mater. 150 (3), 468e493. Rocha, E.M.R., Vilar, V.J.P., Fonseca, A., Saraiva, I., Boaventura, R. A.R., 2011. Landfill leachate treatment by solar-driven AOPs. Sol. Energy 85 (1), 46e56. Ruiz, G., Jeison, D., Chamy, R., 2003. Nitrification with high nitrite accumulation for the treatment of wastewater with high ammonia concentration. Water Res. 37 (6), 1371e1377. Sarria, V., Parra, S., Adler, N., Peringer, P., Benitez, N., Pulgarin, C., 2002. Recent developments in the coupling of photoassisted and aerobic biological processes for the treatment of biorecalcitrant compounds. Catal. Today 76 (2e4), 301e315. Silva, A.C., Dezotti, M., Sant’Anna, G.L., 2004. Treatment and detoxification of a sanitary landfill leachate. Chemosphere 55 (2), 207e214. Strous, M., Heijnen, J.J., Kuenen, J.G., Jetten, M.S.M., 1998. The sequencing batch reactor as a powerful tool for the study of slowly growing anaerobic ammonium-oxidizing microorganisms. Appl. Microbiol. Biot. 50, 589e596. Tatsi, A.A., Zouboulis, A.I., 2002. A field investigation of the quantity and quality of leachate from a municipal solid waste landfill in a Mediterranean climate (Thessaloniki, Greece). Adv. Environ. Res. 6 (3), 207e219. Thalasso, F., Vallecillo, A., Garcı´a-Encina, P., Fdz-Polanco, F., 1997. The use of methane as a sole carbon source for wastewater denitrification. Water Res. 31 (1), 55e60. Weichgrebe, D., Vogelpohl, A., Bockelmann, D., Bahnemann, D., 1993. Treatment of Landfill Leachates by Photocatalytic Oxidation Using TiO2: A Comparison with Alternative Photochemical Technologies. Elsevier Science Publishers B.V., Amsterdam, pp. 579e584.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 5 9 e2 6 6 9
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Amazonian former gold mined soils as a source of methylmercury: Evidence from a small scale watershed in French Guiana Stephane Guedron a,*, Michel Grimaldi b, Catherine Grimaldi c,d, Daniel Cossa e, Delphine Tisserand a, Laurent Charlet a a
IRD, Institut des Sciences de la Terre (ISTerre), UMR5559 (IRD/UJF/CNRS), University Joseph Fourier, BP 53, F-38041 Grenoble, France IRD, UMR Bioemco-Bioge´ochimie et Ecologie des Milieux Continentaux, UMR211, Institut de Recherche pour le De´veloppement, 32 avenue Henri Varagnat, F-93143 Bondy, France c INRA, UMR1069, Soil Agro and Hydrosystem, F-35000 Rennes, France d Agrocampus Rennes, UMR1069, Soil Agro and Hydrosystem, F-35000 Rennes, France e IFREMER, Laboratoire de Bioge´ochimie des Contaminants Me´talliques, Centre de Me´diterrane´e, BP 330, Zone Portuaire de Bre´gaillon, F-83507 La Seyne-sur-mer, France b
article info
abstract
Article history:
Total mercury (HgT) and monomethylmercury (MMHg) were investigated in a tropical head
Received 22 October 2010
watershed (1 km2) of French Guiana. The watershed includes a pristine area on the hill slopes
Received in revised form
and a former gold mined flat in the bottomland. Concentrations of dissolved and particulate
16 February 2011
HgT and MMHg were measured in rain, throughfall, soil water and at three points along the
Accepted 17 February 2011
stream. Samples were taken in-between and during 14 storm events at the beginning and
Available online 24 February 2011
middle of the 2005 and 2006 rainy seasons. Dissolved and particulate HgT concentrations in the stream slightly increased downstream, while dissolved and particulate MMHg concen-
Keywords:
trations were low at the pristine sub-watershed outlet (median ¼ 0.006 ng L1 and 1.84 ng g1,
Mercury
respectively) and sharply increased at the gold mined flat outlet (median ¼ 0.056 ng L1 and
Methylmercury
6.80 ng g1, respectively). Oxisols, which are dominant in the pristine area act as a sink of HgT
Tropical watershed
and MMHg from rain and throughfall inputs. Hydromorphic soils in the flat are strongly
Gold placers
contaminated with Hg (including Hg0 droplets) and their structure has been disturbed by
Stream water
former gold-mining processes, leading to multiple stagnant water areas where biogeo-
Oxisols
chemical conditions are favorable for methylation. In the former gold mined flat high dis-
Hydromorphic soils
solved MMHg concentrations (up to 0.8 ng L1) were measured in puddles or suboxic soil pore waters, whereas high dissolved HgT concentrations were found in lower Eh conditions. Ironreducing bacteria were suggested as the main methylators since highest concentrations for dissolved MMHg were associated with high dissolved ferrous iron concentrations. The connection between saturated areas and stagnant waters with the hydrographic network during rain events leads to the export of dissolved MMHg and HgT in stream waters, especially at the beginning of the rainy season. As both legal and illegal gold-mining continues to expand in French Guiana, an increase in dissolved and particulate MMHg emissions in the hydrographic network is expected. This will enhance MMHg bio-amplification and present a threat to local populations, whose diet relies mainly on fish. ª 2011 Elsevier Ltd. All rights reserved.
* Corresponding author. Tel.: þ33 0 4 76 63 59 28; fax: þ33 0 4 76 63 52 52. E-mail address:
[email protected] (S. Guedron). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.02.022
2660
1.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 5 9 e2 6 6 9
Introduction
Mercury (Hg) contamination of Amazonian ecosystems through the use of elemental Hg for gold amalgamation has been highlighted by many scientific studies (Dorea and Barbosa, 2007; Lechler et al., 2000; Malm, 1998; Pfeiffer et al., 1993; Roulet et al., 1999b; Wasserman et al., 2003).Toxicological concerns related to high monomethylmercury (MMHg) concentrations in Amazonian fish have been evidenced for Amerindian populations whose diet relies mainly on fish (Brabo et al., 2000; Durrieu et al., 2005; Frery et al., 2001; Harper and Harris, 2008; Porvari, 1995). The main sources of MMHg have been identified in areas where oxygen content drops sharply, such as river and lake sediments, as well as in lake water columns at the oxyclines and in temperate flooded environments (Coquery et al., 2003; Hall et al., 2008). Amazonian ecosystems combine most of the surrounding geochemical conditions favorable for Hg methylation such as high temperature, high organic matter content, abundant electron acceptors (i.e., sulfate ions and ferric iron contents), and intensive microbial activities (Benoit et al., 2003; Ullrich et al., 2001). Nevertheless, most studies performed in Amazonian watersheds have focused on total Hg distribution in waters, river sediments and soils (Barbosa et al., 2003; Dorea and Barbosa, 2007; Marchand et al., 2006; Roulet et al., 1998b) and few data are available for MMHg. It is important to gain knowledge about Hg methylation in tropical watersheds since correlations found between total Hg and MMHg concentrations are commonly weak and related to external environmental factors, such as the chemical form of HgII, which have a strong influence on its bioavailability for methylation (Birkett and Lester, 2005; Lambertsson and Nilsson, 2006; Ullrich et al., 2001). The presence of elemental Hg is also an important factor which must be considered in gold-mining areas, since Dominique et al. (2007) have recently shown that, under experimental conditions, the presence of elemental Hg droplets can enhance MMHg production. In this study, we examined if MMHg was produced in flooded soils of a former gold mined area and tested in the field the experimental findings of Dominique et al. (2007) regarding Hg methylation in the presence of Hg0 droplets. We focussed on particulate and dissolved HgT and MMHg outputs from a small watershed covered by tropical-humid vegetation. This watershed was chosen (i) because it includes a pristine area and a Hg contaminated former gold mined flat, and (ii) because of its small scale (1 km2), which enables an optimal understanding of Hg distribution in and between the different pedological and hydrological compartments. Finally, we attempted to analyze the influence of internal (i.e., geochemical, hydrological, geomorphological) and external (i.e., seasonality) determinants on Hg methylation and emissions to the watershed’s hydrological network.
2.
Site, material and methods
2.1.
Environmental settings
2.1.1.
Location
The study site is the Combat Creek (CC) watershed, located in French Guiana, South America (52 230 W, 4 350 N) (Fig. 1),
covered by tropical rain forest. Its surface area is approximately 1 km2. The climate is tropical-humid with a mean annual rainfall of 4000 mm (Barret, 2004). Precipitation mainly occurs from December to July, with May and June as the wettest months.
2.1.2.
Bedrock
The CC watershed is located in the ‘Amina series’ of the Guiana Proterozoic shield consisting primarily of dark schist and thin sandstone (Mile´si et al., 1995). Vast gravel deposits from ancestral rivers within the valleys contain large quantities of placer gold, derived from the weathering of goldequartz veins.
2.1.3.
Soil cover
Soil associations of the CC watershed have been described in detail in a previous publication (Guedron et al., 2009). Oxisols are developed on the hill tops and on the steep upper- and middle-slopes, ultisols occur mostly on the foot-slopes, and hydromorphic soils are found in the valley referred to as “flats”. A large part of the watershed was considered to be pristine while in the lowland, ancient “Long Tom” sluices, gold-bearing gravel heaps and elemental Hg droplets attest to the former gold-mining activities dating from the early 1950s.
2.1.4.
Hydrology
The Combat Creek is a tributary of the Boulanger River (BR). The CC watershed outlet exhibits a permanent discharge, in contrast to intermittent flow in upstream channels. Water discharge response to rain is progressive and lasting, with a high amplitude at the outlet, contrary to upstream and midstream sections, where the response is rapid and short with low amplitude. Surface runoff is visible during heavy rainfalls on the steep hill slopes (often between 15 and 30%). In the former gold mined flat, due to the disorganized original topography, the flow is fractionated into a web of small creek tributaries and multiple stagnant water zones which are not always linked to each other or connected to the hydrographic network.
2.2.
Sampling procedure
Four points were monitored along the streams (Fig. 1): PS (pristine spring) is a spring of the CC which drains a small subwatershed and only flows during the rainy season; MS (middlestream) and CO (contaminated flat outlet) are respectively in the upstream and downstream part of the former gold mined flat; BR is located on the Boulanger River, upstream the confluence with the CC (Fig. 1). Rain and stream waters were sampled during and between fourteen rain events at the beginning (08th, 09th and 13th December 2005; 08th, 09th and 12th December 2006) and in the middle of the rainy season (24th, 25th, 27th and 30th May 2005; 18th, 20th, 21st and 25th June 2006). In addition, several superficial stagnant water areas of the former gold mined flat were sampled, as well as the hydromorphic soil’s pore waters. Metacrylate-lined rain gauges were set up close to each sampling point to measure rainfall under forest cover and to collect throughfall samples after each rain event. At the pristine spring, an additional rain
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 5 9 e2 6 6 9
2661
Fig. 1 e The watershed study site with water sampling (PS e pristine spring, MS e middlestream and CO e contaminated flat outlet, in the Combat Creek and BR e Boulanger River), soil sampling locations and soil types (O: oxisol, U: ultisol and H: hydromorphic soil).
gauge was set up outside the forest cover to collect rain samples. The creek flow was gauged using a C2 OTT current meter set, a mobile decameter and vertical ladders, as described by Herschy (1995). Mercury specific fluxes were calculated using discharge and Hg concentration data obtained during and between rain events, and considered the surface area of upstream watersheds. On the CC watershed, soil profiles next to PS, MS and CO were sampled systematically along toposequences (Fig. 1). The toposequences next to the MS and CO reached the gold mined flats. Soil samples were collected, every 10 or 20 cm depths, down to 1e2 m, using an auger. All soil samples were collected in sterile plastic bags. Water sample collection and treatment were performed using ultra-clean techniques (Cossa and Gobeil, 2000). All materials in contact with samples were acid-washed (5 days in 20% HNO3 v/v, then 3 days in HCl 10% v/v) and rinsed several times with demineralized water (Milli-Q) before use. Polyethylene gloves were used for handling operations. Clean Teflon (FEP) bottles were stored in double polyethylene bags until use. Aliquots for total dissolved mercury ((HgT)D) and
dissolved methylmercury ((MMHg)D) analysis were collected in FEP bottles and acidified (HCl 0.5% v/v, Millipore Seastar) after filtration on 22-mm-filters (0.45 mm Sterivex-HV) (Parker and Bloom, 2005). Dissolved Organic Carbon (DOC) samples were stored in Pyrex bottles (previously heated at 500 C) and acidified (HCl 1% v/v, Millipore Seastar) after filtration (0.45 mm Sterivex-HV). Total particulate mercury ((HgT)P) and methylmercury ((MMHg)P) samples were obtained by filtration on 47-mm-diameter filters (0.45 mm hydrophilic e LCR Teflon) (Cossa et al., 1996). All samples for major elements were filtered (0.45 mm PVDF). Samples for cations analysis were acidified (HNO3 2% v/v, Seastar), and samples for anion analysis were stored in polyethylene bottles and frozen until analysis. Particulate Organic Carbon (POC) samples were obtained through filtration (0.7 mm GF/F, Whatman). Interstitial soil waters were sampled using Milli-Q rinsed microporous polymer tube samplers (Rhizon SMS, Rhizosphere Research Products). The samplers were placed on the CC bank between the stream sampling points. One interstitial soil water profile was sampled using dialysis membrane techniques (metacrylate peeper) with a 0.45 mm HAWP
2662
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 5 9 e2 6 6 9
membrane. The peeper was first acid-washed as described above, then filled with Milli-Q water and degassed with Hgfree nitrogen during a 15 day period. The peeper was placed in a hydromorphic soil from the former gold mined flat for 15 days for osmotic equilibration. The Eh profile was monitored several centimeters away from the peeper on the last day of equilibration just before the removal of the peepers. The sulfide-accumulating zone (SAZ) was identified with sulfidesensitive sellotape (fixed on the peeper), through the formation of a surface darkening Ti-S complex (Jezequel et al., 2007). Aliquots for (MMHg)D, DOC, anions, sulfides and (FeII) analysis were sampled in peeper cells every 3 cm and stored as previously described.
chromatography (Dionex, model CD20). Dissolved reduced iron ([FeII]) and sulfide ([SII]) were measured in the field with a Hach (model DR/850) spectrometer (methods 8146 and 8131 for [FeII] and [SII], respectively). [DOC] was determined using a Non Dispersive Infra-Red CO2 spectrometer (NDIR, Shimatzu) after humid oxidation in a sodium persulfate solution at 100 C. [POC] was determined by dry combustion of GF/F filters, using a Fisons 1500 CHN autoanalyzer. Detection limits (3SDblk) were 0.06 and 0.05 mg L1 for [SO42] and [NO3], 0.01 mg L1 for [FeII] and [SII], and 0.2 mg L1 for [DOC]). Eh, pH and conductivity were performed in situ using a Sentron Eh probe (model Argus with probe 67597), a WTW pH meter and a conductimeter (model 340i), respectively.
2.3.
Analytical measurements
2.4.
2.3.1.
Soil analysis
Chloride (Cl) was used as a hydrological index to trace the proportion of mixing in the stream between recent water, generally surface runoff characterized by low [Cl] close to throughfall concentrations, and old water, corresponding to subsurface runoff or deep groundwater with typically higher [Cl] (Christophersen and Neal, 1990; Grimaldi et al., 2004; Peters and Ratcliffe, 1998; Soulsby et al., 2007). The [Cl] increase between both hydrological compartments is due to evapo-transpiration related to the residence time of water in the soil. Cl was thus used as a conservative element to compare the behavior of the various Hg species.
Soil samples were freeze-dried, sieved, and the fraction <2 mm was crushed to grain size smaller than 63 mm for Hg analysis. Total Hg concentrations ([HgT]) were determined by atomic absorption spectrophotometry after dry mineralization and gold amalgamation with an automatic mercury analyser (Altec, Model AMA 254) with a relative precision of 10% (Roos-Barraclough et al., 2002). Concentrations obtained for repeated analyses of certified reference materials (CRM) never exceeded the range of concentration given for standards CRM 7002 (0.09 0.012 mg g1 e Czech Metrological Institute) and MESS-3 (0.091 0.008 mg g1 e National Research Council of Canada). The analytical quality was assured by analyzing every sample twice. Typically, the measurement error is usually about 5% and always below 10%. The detection limit, defined as three times the standard deviation of the blank (SDblk), was 0.005 mg g1.
2.3.2.
Water analyses
Samples were analyzed for [(HgT)D] and [(HgT)P], [(MMHg)D] and [(MMHg)P] by cold vapor atomic fluorescence spectrometry (CVAFS) after conversion of all mercury species into Hg0 (Bloom and Fitzgerald, 1988) followed by detection using a Tekran (Model 2500) mercury detector. The principles of the methods used originate from the Bloom and Fitzgerald (1988) gold amalgamation method for (HgT)D, from the Liang et al. (1994) ethylation method for (MMHg)P and from the Tseng et al. (1998) hydruration method for (MMHg)D modified by Cossa et al. (2009). (HgT)P was performed after HCl/HNO3 digestion (10 h at 70 C) in PFA Teflon reactors (Coquery et al., 1997). The detailed procedures are described elsewhere (Cossa et al., 2003, 2002). These quantifications were performed after checking for possible interference with the internal spikes (Coquery et al., 2003). The accuracy of analyses was checked using the CRM ORMS-3 (National Research Council of Canada) for (HgT)D, CRM 7002 for (HgT)P and CRM ERM-AE670 (IRMM e European Commission) for (MMHg)D and (MMHg)P. Analytical quality was assured by triplicating several samples, and the measurement error usually was approximately 10% and always below 15% for [(HgT)D], [(HgT)P], and [(MMHg)D] and always below 30% for [(MMHg)P]. The detection limits (3SDblk) were 0.01 ng L1 for (HgT)D, 0.004 ng L1 for (MMHg)D, and 0.05 ng g1 for (MMHg)P. Dissolved chloride, sulfate and nitrate ([Cl], [SO42] and [NO3]) were determined using ionic
2.5.
Use of chloride as hydrological index
Statistical treatment
Since most geochemical data were not normally distributed, we have listed in tables: the mean, the median and the standard error of the mean (SEM), and as suggested by Webster (2001), the 6 following parameters (Supplementary Data): the mean, the standard error of the mean (SEM), the median, 25th percentiles (25th perc.), 75th percentiles (75th perc.) and the number of observations (N). We also used non-parametric tests, the ManneWhitney rank sum test (U test), or the KruskaleWallis one way analysis of variance on ranks (H test), to compare two, or more than two sets of data, respectively, and pairwise multiple comparison according to Dunn’s method to isolate the set or sets that differed from the others. Correlation coefficients (R) and P values (P) are reported. All statistical analyses were performed using Sigmastat.
3.
Results
3.1.
Total Hg in soils
On the slopes of the pristine part of the watershed, soil [HgT] globally decreased from oxisols to ultisols (Table 1), but the difference was statistically significant (t test, P < 0.001) only for the deeper layers (>0.5 m), because [HgT] rapidly decreased with depth in the ultisols. Soil surface layer [HgT] significantly increased (U Test, P ¼ 0.006) from the pristine to the former gold mine flat soils (by a factor between 2 and 4 for the median values), up to a [HgT] maxima of 5.47 mg g1 [HgT] were highly variable over short distances and with depth in the former gold mined flat.
2663
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 5 9 e2 6 6 9
Table 1 e Soil total mercury (HgT) median and mean concentrations, and standard error of the mean (SEM) for different soil types; presented as median/mean (SEM). Soil type Oxisol (pristine slopes) Ultisol (pristine slopes) Hydromorphic soil (contaminated flat)
3.2.
Soil depth (cm) 0e50 >50 0e50 >50 0e50 >50
1
HgT (mg g ) 0.35/0.37 0.42/0.38 0.20/0.25 0.07/0.10 0.82/1.31 0.75/1.01
(0.04) (0.03) (0.07) (0.04) (0.35) (0.26)
Total Hg in stream and rain waters
A global increase in total Hg concentrations (Table 2) in the CC stream waters was observed between the pristine spring (PS) and the contaminated flat outlet (CO). [(HgT)P] and [(HgT)D] significantly increased downstream (H test, P < 0.001) from PS to CO. On the contrary [(HgT)P] and [(HgT)D] were not significantly different between CO and BR. [(HgT)D] in both throughfall and rain waters were higher than in stream waters at PS and MS (U test, P < 0.001), and were in the range of [(HgT)D] measured at CO (Tables 2 and 3 and Fig. 2). Conversely, [Cl] were higher in stream waters at PS than in throughfall and rain waters (H test, P < 0.001) due to the residence time in soil. Downstream, [Cl] increased at CO, fed by older waters than the outlet of the pristine sub-catchment. Most of the total Hg measured in the stream was bound to suspended particles with log partitioning coefficients (logKD ¼ log[(HgT)P]log[(HgT)D]) ranging from 3.8 to 8.3, with a median value of 5.9. HgT was not specifically associated with suspended organic particles since no correlation was observed between [(HgT)P] and [POC]. Seasonal variations for total Hg concentrations were noticeable at MS and CO. At PS, the absence of discharge at the beginning of the rainy season precluded water sampling and seasonal comparison. [(HgT)P] at CO were lower during the rain events at the beginning of the rainy season than those in the middle (U test, P < 0.05; respective medians of 1.01 and 0.77 mg g1). [(HgT)D] at MS and CO were higher at the beginning of the rainy season than later in the season (U test, P < 0.05) (Fig. 2). At the same time, [Cl] decreased between the beginning and the middle of the rainy season (U test, P < 0.05) due to the progressive renewal of soil waters and/or to the increase of surface water runoff in relation with intense rain events and to soil water saturation.
3.3.
Methylmercury in stream and rain waters
The mean [MMHg]/[HgT] ratios in stream waters equaled 2% and 1%, for the particulate and dissolved phases, respectively. For total Hg, a large downstream increase was observed for MMHg concentrations between the pristine sub-watershed and the contaminated flat (Table 2). [(MMHg)P] were similar at PS, MS and BR but higher at CO (H test, P < 0.001), with a 3 times median increase between PS and CO. [(MMHg)D] measured at PS and MS were very low, often under the detection limit (<0.004 ng L1), but significantly increased at CO (H test, P < 0.001), with a high variability. [(MMHg)D] at BR ranged between the concentrations measured at PS, MS, and CO. In rain samples, [(MMHg)D] were not significantly different than those in throughfalls, however, both were a slightly larger than the concentrations measured in PS stream waters (Table 3). MMHg had a strong affinity for particulate organic matter since log partitioning coefficients (logKD ¼ log[(MMHg)P]log [(MMHg)D]) were high (between 7.4 and 9.5), and [(MMHg)P] measured at CO were correlated with [POC] (R ¼ 0.457, P < 0.05, N ¼ 24). [(MMHg)P] monitored at CO were lower at the beginning than in the middle of the rainy season (U test, P < 0.001; respective medians of 12.26 and 5.59 ng g1), as opposed to [(MMHg)D] which was highly variable but globally higher (U test, P < 0.001) at the beginning of the rainy season than in the middle.
3.4. Dissolved total and methylmercury in soil waters of the gold mined flat [(MMHg)D] measured in various locations of the former gold mined flat area largely exceeded those measured in the pristine area for both overlying and subsurface soil pore waters (Table 3, Fig. 3). The highest [(MMHg)D] were related to slightly negative or positive Eh values and high [FeII]. They also were associated with pH values between 5 and 6, while pH values measured in stream waters ranged from 4 to 5. High [(HgT)D] also were found in these soil waters, especially with low Eh conditions and high [FeII] (Fig. 3); however, no relation was observed between [(HgT)D] and [(MMHg)D]. Fig. 4 illustrates a complete vertical profile of [(MMHg)D] for both overlying and pore waters of a disorganized hydromorphic soil of the flat, sampled near an ancient sluice at the beginning of the rainy season. In this profile, [(MMHg)D] were low in the pore water and sharply increased in the overlying water. High [(MMHg)D] were found in the upper part of the SAZ where the Eh increased, and above the SAZ where [FeII] sharply increased.
Table 2 e Stream water total dissolved mercury (HgT)D, particulate mercury (HgT)P, dissolved monomethylmercury (MMHg)D, and particulate monomethylmercury (MMHg)P median and mean concentrations, and SEM; presented as median/mean (SEM). Sampling location: PS MS CO BR
(HgT)D (ng L1) 0.94/0.98 1.34/2.77 1.57/4.78 1.67/4.94
(0.13) (0.75) (1.46) (3.43)
(HgT)P (mg g1)
(MMHg)D (ng L1)
0.25/0.53 (0.13) 0.61/0.82 (0.17) 0.88/1.99 (0.94) 1.51/1.42 (0.33)
0.006/0.016 (0.004) 0.009/0.048 (0.033) 0.056/0.062 (0.005) 0.024/0.025 (0.004)
(MMHg)P (ng g1) 1.84/1.86 0.75/0.75 6.80/9.59 1.69/2.67
(0.37) (0.67) (1.28) (1.33)
2664
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 5 9 e2 6 6 9
Table 3 e Rain, throughfalls, stagnant (overlying) waters and pore hydromorphic soil waters. Total dissolved mercury (HgT)D and dissolved monomethylmercury (MMHg)D median and mean concentrations, and SEM; presented as median/ mean (SEM). Sampling location Rain (PS) Throughfall (PS) Soil overlying water (Pristine sub-watershed) Soil overlying watera (Contaminated flat) Soil overlying waterb (Contaminated flat) Soil pore watera (Contaminated flat) Soil pore waterb (Contaminated flat)
(HgT)D (ng L1) 4.83/4.63 3.98/3.93 1.70/3.11 e 2.01/2.01 e 2.34/5.49
(0.74) (0.68) (1.25) (0.40) (1.65)
(MMHg)D (ng L1) 0.010/0.011 0.017/0.026 0.005/0.007 0.615/0.670 0.082/0.243 0.136/0.113 0.161/0.231
(0.003) (0.008) (0.002) (0.080) (0.098) (0.034) (0.047)
a Relates to single location of the gold mined flat between 1 and 10 cm for stagnant waters and 0e15 cm for pore waters. b Relates to various locations in the gold mined flat.
Both median [(MMHg)D] and [(HgT)D] in overlying soil and pore waters were higher at the beginning of the rainy season (0.27 ng L1 and 4.27 ng L1, for (MMHg)D and (HgT)D, respectively) than in the middle of the rainy season (0.11 ng L1 and 1.60 ng L1, for (MMHg)D and (HgT)D, respectively e H test, P < 0.01).
4.
Discussion
4.1. Oxisols as a sink of total mercury and methylmercury in the pristine area In the pristine area of the CC watershed, the decreasing [(HgT)D] with increasing [Cl] found between rain or throughfall waters and the stream (Fig. 2) is due to the adsorption of Hg on oxisol components during water percolation. The strong adsorption capacity of oxisols already was observed in another study on the Hg distribution in oxisols at the same site (Guedron et al., 2009). The comparison of our data with those of other studies showed that in rain and throughfall waters, [(HgT)D] were within the range of reported concentrations for 83 rain events (2.34 0.27 ng L1) monitored in French Guiana (Muresan et al., 2007b) and that [(MMHg)D] were within the range of concentrations published for temperate regions (Lawson and Mason, 2001) and for 50 rain events (0.03 0.09 ng L1) monitored in French Guiana (Muresan et al., 2007b). In spring waters, [(HgT)D] were similar to concentrations reported for larger Amazonian rivers (0.4e2.8 ng L1 in Lechler et al., 2000; Roulet et al., 1998a) and [(MMHg)D] were within the range of concentrations reported by several authors (Bisinoti et al., 2007; Roulet et al., 1999a) for large Amazonian rivers (0.02e0.24 ng L1) and for large French Guiana rivers (0.06e0.10 ng L1 according to Muresan, 2006; Roulet et al., 1999a). These comparisons confirm that tropical soils act as a sink for Hg and regulate Hg fluxes towards small as well as large hydrosystems (Roulet et al., 2001).
4.2. Evidence of soil total mercury contamination in the former gold mined flat The [HgT] reaching up to 100 times the values reported from pristine hydromorphic soils in French Guiana (Guedron et al., 2006) demonstrates the large contamination of the former
gold mined flat by mercury. This large Hg contamination of the flat explains the downstream increase of [(HgT)P] along the stream (Table 2), which carries an increased proportion of contaminated particles from the pristine spring to the flat outlet. [(HgT)P] in suspended particles are higher than in the soil surface (Tables 1 and 2), since suspended particles consist of the soil’s finest granulometric fraction, which are enriched in Hg (Grimaldi et al., 2008). In contrast, [(HgT)P] monitored at the outlet of the pristine sub-catchment are within a similar range of reported data from larger Amazonian rivers (0.11e0.36 mg g1 in Lechler et al., 2000; Roulet et al., 1998a).
4.3. Hydromorphic soils as a source of methylmercury for the stream While the downstream increase in [(HgT)P] and [(HgT)D] is gradual until reaching the watershed outlet, the sharp increase in [(MMHg)P] and [(MMHg)D] between MS and CO suggests a particularly strong MMHg input from the contaminated flat. Both [(MMHg)P] and [(MMHg)D] monitored at the outlet of the watershed exceeded the range of concentrations given in the literature cited above for large Amazonian and French Guiana rivers, including the Boulanger river (Muresan, 2006; Roulet et al., 1999a). Puddles and pore waters of hydromorphic soils are the most probable MMHg sources in the flat. The numerous locally isolated water puddles in the flat are attributed to former mining activities which have strongly disturbed the flat’s topography. The former “Long Tom” mining process and stream tapping, which was shifted laterally by miners along the flat, have led to a web of small creek tributaries and to multiple stagnant water areas which are not always connected to the hydrographical network. The intense disturbance of soil structure leads to local drainage deficiency reflected by the presence of quasi-permanently flooded soils. Because the ancient gold-mining activities, such as gold amalgamation and burning of amalgams, were performed on site, Hg contamination in the flat is due both to the loss of Hg0 droplets, and to the rapid deposition of atmospheric Hg in the local environment. This explains the high variability of [HgT] in soils. In addition, Hg was observed to be present mainly in its elemental form in these contaminated soils (Guedron et al., 2009). The high [(MMHg)D)] monitored in various locations of the flat corroborates the laboratory simulation of gold-mining activities (Dominique et al., 2007), which indicated that the
2665
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 5 9 e2 6 6 9
0.10
5
PS
PS 0.08 [(MMHg)D ] (ng L-1 )
[(HgT)D ] (ng L-1)
4
3
2
1
0.04 0.02 0.00
0 0
1
2
3
4
5
60 58 56 54
6
0
MS [(MMHg)D ] (ng L-1 )
[(HgT)D ] (ng L-1 )
0.06
15
10
5
1
2
3
4
5
0.60 0.55 0.50 0.05
6
MS
0.04 0.03 0.02 0.01 0.00
0 1
2
3
140 120 100 80 60 25
4
0
1
2
3
0.15
CO [(MMHg)D ] (ng L-1)
[(HgT)D ] (ng L-1 )
0
20 15 10
4
CO
0.10
0.05
5 0.00 0 0
1
2 3 [Cl-] (mg L-1)
Recent Rain Beginning rainy season Middle rainy season Throughfall Beginning rainy season Middle rainy season
4
Old water
0
Recent
1
2 3 [Cl-] (mg L-1)
4
Old water
Pristine Spring (PS) Contaminated flat outlet (CO) Beginning rainy season Beginning rainy season Middle rainy season Middle rainy season Middlestream (MS) Beginning rainy season Middle rainy season
Fig. 2 e Dissolved [(HgT)D] and [(MMHg)D] versus [ClL] for rain, throughfall and stream waters at each sampling location (PS, MS and CO), sampled during and out of rain events at the beginning and the middle of the rainy season.
presence of Hg0 can enhance Hg methylation in the Amazonian environment. The local soil disturbance favors the existence of high methylation areas. In hydromorphic soils and isolated puddles, water residence time can be long, which induces anoxic conditions and high concentrations of
dissolved elements. Low Eh, pH between 5 and 6, iron oxides, high concentrations of DOC, SO42 and (HgT)D are favorable for methylation (Benoit et al., 1999, 2003; Bisogni and Lawrence, 1975; Mason et al., 1995). Under these conditions iron-reducing as well as sulfate-reducing bacteria (IRB and
2666
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 5 9 e2 6 6 9
1.0
10
-1 [(MMHg)D ] (ng L )
-1 [(MMHg)D ] (ng L )
0.8
0.6
0.4
1
0.1
r² = 0.59
0.01
0.2
0.0 -400 -300 -200 -100 0
0.001 100 200 300 400
0
1
2
3
4
5
6
7
-1 [FeII] (mg L )
Eh (mV)
Beginning rainy season Middle rainy season Pristine subwatershed - Soil overlying water Contaminated Flat - Soil overlying water Contaminated Flat - Soil pore water Contaminated Flat - Soil pore water Contaminated Flat - Soil overlying water 100
10 r² = 0.50
1
0.1 -400 -300 -200 -100 0
-1 [(HgT)D ] (ng L )
-1 [(HgT)D ] (ng L )
100
10 r² = 0.74
1
0.1 100 200 300 400
Eh (mV)
0
1
2
3 4 5 -1 [FeII] (mg L )
6
7
Fig. 3 e [(MMHg)D] and [(HgT)D] versus Eh and [FeII] in soil waters. Soil pore waters and overlying waters were sampled near the PS and in the former gold mined area between MS and CO, in the beginning and middle of the rainy season. Regression lines are plotted if significant ( p < 0.05).
SRB, respectively) are reported to be the main mercury methylators (Barkay et al., 1997; Fleming et al., 2006). The increase of [(MMHg)D] within the 10 first centimeters of overlying waters with high dissolved [FeII], and to a lesser extent in the upper SAZ in soil pore waters (Fig. 4) reinforces the hypothesis of a contribution of IRB and SRB in the methylation, as supported by archetypal chemical changes where microorganisms shift from FeOOH(s) to SO42 as an electron acceptor. The occurrence of high [(MMHg)D] with high [FeII] suggests a greater availability of (HgT)D for IRB than for SRB in the SAZ, since in absence of sulfides, the adsorption and coprecipitation of Hg(II) onto FeS(s) is restricted (Fink, 2002; Mehrotra et al., 2003; Mehrotra and Sedlak, 2005). The striking feature of high [(MMHg)D], associated with slightly negative or slightly positive Eh values measured in various overlying and soil pore waters of the flat (Figs. 3 and 4), corresponds well with the observations made in lake sediment/water interfaces (Birkett and Lester, 2005; Muresan et al., 2007a, 2007b; Ullrich et al., 2001).
4.4. Seasonal influence on MMHg emissions to the stream The favorable geochemical conditions for mercury methylation can occur during the dry season when they are favored by a long residence time of water in puddles and soils of the gold mined flat. At the beginning of the rainy season (MMHg)D is discharged by pulses during rain events which leads to high concentrations in the stream. In the middle of the rainy season the decrease of [(MMHg)D] is due both to unfavorable methylation conditions and global dilution, as indicated by [Cl] decrease at the same time. Similarly, the gold mined flat puddles and soil pore waters are more concentrated in Cl at the beginning of the rainy season than later (U test, P < 0.05). As for Cl, the renewal of stagnant overlying and pore waters of the contaminated flat leads to the dilution of (MMHg)D and (HgT)D, as well as other dissolved elements such as DOC and SO42. The dissolution of particulate OM and minerals, as well as methylation reactions are slow reactions (Bisogni and
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 5 9 e2 6 6 9
Eh (mV) -300 -200 -100
0
100
10 ng lyi er er Ov wat
0 -5 [FeII] (mg L-1)
-10
[(MMHg)D] (ng L-1)
Eh (mV)
re Po ter wa
Depth (cm)
5
-15 0.0 0.3 0.6 0.9 3.0 4.0 5.0 Concentration
Fig. 4 e [(MMHg)D] (ng LL1), [FeII] (mg LL1), and Eh (mV) vertical profiles and sulfide-accumulating zone (SAZ e gray color area) of the overlying (10 cm thick water puddle) and pore water of a flooded hydromorphic soil in the former gold mined area between MS and CO, at the beginning of the rainy season. Soil water interface (0 cm) is plotted as a horizontal line.
Lawrence, 1975; Langley, 1973). Therefore, since DOC is the most important complexing ligand in surface waters in the absence of sulphide the bioavailability of Hg for methylating bacteria may decrease in the middle of the rainy season (Benoit et al., 2003; Ullrich et al., 2001). The change in geochemical conditions such as the decrease of pH and conductivity and the increase in Eh from the beginning to the middle of the rainy season, also limits bacterial activity and Hg availability, which can lead to the decrease of methylation rates (Birkett and Lester, 2005; Ullrich et al., 2001).
4.5. Hg and MMHg export from the CC watershed to the stream The comparison of (HgT) and (MMHg) specific fluxes between the outlet of the pristine oxisol sub-watershed (PS) and the entire watershed (CO) indicates that the contribution of the
Table 4 e Stream waters at the outlet of the pristine oxisol sub-watershed and the entire watershed. Total dissolved mercury (HgT)D, particulate mercury (HgT)P, dissolved monomethylmercury (MMHg)D, and particulate monomethylmercury (MMHg)P median and mean specific fluxes, and SEM; presented as median/mean (SEM). Specific fluxes (HgT)D (ng s1 km2) (HgT)P (ng s1 km2) (MMHg)D (ng s1 km2) (MMHg)P (ng s1 km2)
PS (surface 0.12 km2)
CO (surface 1.27 km2)
113/147 (32) 517/1458 (521) 0.67/0.97 (0.30)
117/628 (276) 1104/2951 (1158) 4.20/4.17 (0.51)
3.5/4.2 (0.96)
14.0/21.6 (3.80)
2667
contaminated flat is especially substantial for (MMHg). Additionally, calculated specific fluxes for both dissolved and particulate MMHg were increased by factors of 6 and 4 between PS and CO, respectively (U tests, P 0.001 e Table 4), while those for (HgT)D and (HgT)P were within the same range and doubled (U test, P ¼ 0.043) between PS and CO, respectively. The estimation of annual specific fluxes from the CC watershed highlights its large contamination, all the more since the flat’s contribution is diluted by the fluxes originating from the pristine soils which cover the majority of the watershed and act as a sink for Hg and MMHg. The comparison of our data with calculated fluxes in other Amazonian locations corroborates this point. For example, the annual (HgT)P export at the outlet of the CC watershed (mean ¼ 93 37 mg m2 yr1) is much larger than the fluxes (2.6e8.5 mg m2 yr1) calculated for a small forested Amazonian area (1.6 km2), located far from gold-mining activities (Fostier et al., 2000), and for fluxes (30e35 mg m2 yr1) measured for the large Cururai floodplain system (Maia et al., 2009) where the large size of the basin (3800 km2) may dilute the fluxes originating from gold mined areas significantly. Our data also exceed the range of fluxes found for the Seine river (Coquery et al., 1997) and for urban-type watersheds (0.2e20 mg m2 yr1) according to the Wisconsin’s US system of rivers classification (Hurley et al., 1995). No MMHg fluxes are available for Amazonian watersheds. However, a comparison with boreal and temperate environments shows that the (MMHg)P exported from the CC watershed (mean fluxes ¼ 0.68 0.12 mg m2 yr1) is larger compared to fluxes (0.02e0.183 mg m2 yr1) measured for a selection of sixteen US streams (Balogh et al., 2005; Brigham et al., 2009).
5.
Conclusions
This study shows that, even 60 years after exploitation, former gold-mining activities largely contribute to the in stream load of MMHg, whereas their contribution for total mercury remains moderated. Hydromorphic soils, disturbed and strongly Hg contaminated (including Hg0 droplets) by former gold-mining, were identified as the main sources of MMHg. Methylation was suggested to be induced mainly by IRB during the dry season when the surface and pore waters are stagnant, whereas emissions of MMHg occurred during rainy season events when these waters are discharged into the stream. Such former goldmines can still contribute to MMHg inputs in larger hydrological systems. Numerous former and current artisanal or semi-industrial goldmines exist in French Guiana and elsewhere in Amazonia, but these areas are rarely mapped or referenced. Thus, the evaluation of the real contribution of these former activities is sorely quantifiable and suggests that it is an important contributor of MMHg emissions in Amazonian hydrosystems. The continuous expansion of legal and illegal gold-mining in French Guiana implies an increase in dissolved and particulate MMHg emissions in the hydrographic network, with an enhancement of MMHg contamination of aquatic ecosystems and the consequent increase in the toxicological threat for local human populations whose diet relies mainly on fish.
2668
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 5 9 e2 6 6 9
Acknowledgments This research was supported mainly by the CNRS, FNS, FEDER, MATE/DIREN, and ANR as a part of the Mercury in French Guiana research program and by the Boulanger Mine Company (CMB). We acknowledge Jennifer Harris-Hellal, Noureddine Bousserrhine, Genlis Gallifet and Dennis Lahondes for additional support on this project. We also, acknowledge Bernard Averty from the IFREMER Laboratory of Nantes for support in methylmercury analysis.
Appendix. Supplementary data Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.watres.2011.02.022.
references
Balogh, S.J., Nollet, Y.H., Offerman, H.J., 2005. A comparison of total mercury and methylmercury export from various Minnesota watersheds. Sci. Total Environ. 340 (1e3), 261e270. Barbosa, A.C., Souza, J.d., Dorea, J.G., Jardim, W.F., Fadini, P.S., 2003. Mercury biomagnification in a tropical black water, Rio Negro, Brazil. Arch. Environ. Contam. Toxicol. 45 (2), 235e246. Barkay, T., Gillman, M., Turner, R., 1997. Effects of dissolved organic carbon and salinity on bioavailability of mercury. Appl. Environ. Microbiol. 63 (11), 4267e4271. Barret, J., 2004. Illustrated Atlas of French Guyana (in French). French Guyana Publications, Cayenne. Benoit, J., Gilmour, C., Mason, R., Heyes, A., 1999. Sulfide controls on mercury speciation and bioavailability to methylating bacteria in sediment pore waters. Environ. Sci. Technol. 33, 951e957. Benoit, J.M., Gilmour, C.C., Heyes, A., Mason, R.P., Miller, C.L., 2003. In: Cai, Y., Braids, O.C. (Eds.), Biogeochemistry of Environmentally Important Trace Elements. Oxford University Press, pp. 262e297. Birkett, J., Lester, J., 2005. Distribution of mercury and methylmercury in the sediments of a lowland river system. Proc. R. Soc. A Math. Phys. Eng. Sci. 461 (2057), 1335e1355. Bisinoti, M.C., Sargentini Ju´nior, E., Jardim, W.F., 2007. Seasonal behavior of mercury species in waters and sediments from the Negro river basin, Amazon, Brazil. J. Braz. Chem. Soc. 18 (3), 544e553. Bisogni, J.J., Lawrence, A.W., 1975. Kinetics of mercury methylation in aerobic and anaerobic aquatic environments. J. Water Pollut. Control Fed. 47 (1), 135e152. Bloom, N.S., Fitzgerald, W.F., 1988. Determination of volatil mercury spices at the picogram level by low-temperature gas chromatography with cold-vapor atomic fluorescence detection. Anal. Chim. Acta 208, 151e161. Brabo, E.D., Santos, E.D., de Jesus, I.M., Mascarenhas, A.F.S., Faial, K.D., 2000. Mercury contamination of fish and exposures of an indigenous community in Para State, Brazil. Environ. Res. 84 (3), 197e203. Brigham, M.E., Wentz, D.A., Aiken, G.R., Krabbenhoft, D.P., 2009. Mercury cycling in stream ecosystems. 1. Water column chemistry and transport. Environ. Sci. Technol. 43 (8), 2720e2725. Christophersen, N., Neal, C., 1990. Linking hydrological, geochemical, and soil chemical processes on the catchment
scale: an interplay between modelling and field work. Water Resour. Res. 26, 3077e3086. Coquery, M., Cossa, D., Azemard, S., Peretyazhko, T., Charlet, L., 2003. Methylmercury formation in the anoxic waters of the Petit-Saut reservoir (French Guiana) and its spreading in the adjacent Sinnamary river. J. Phys. IV 107, 327e331. Coquery, M., Cossa, D., Sanjuan, J., 1997. Speciation and sorption of mercury in two macro-tidal estuaries. Mar. Chem. 58 (1e2), 213e227. Cossa, D., Averty, B., Bretaudeau, J., Senard, A.S., 2003. Dissolved Mercury Speciation in Marine Waters. Analysis Methods in Marine Environment. Ifremer and French Ecology, Durable Development Ministry (in French). Cossa, D., Averty, B., Pirrone, N., 2009. The origin of methylmercury in open Mediterranean waters. Limnol. Oceanogr. 54 (3), 837e844. Cossa, D., Coquery, M., Gobeil, C., Martin, J., 1996. In: Baeyens, W., Ebinghaus, R., Vasiliev, O. (Eds.), Regional and Global Cycles of Mercury: Sources, Fluxes, and Mass Balances. Kluwer Academic Publishers, Dordrecht, The Netherlands, pp. 229e247. Cossa, D., Coquery, M., Nakhle, K., Claisse, D., 2002. Total Mercury and Monomethylmercury Analysis in Marine Organisms and Sediments. Analysis Methods in Marine Environment. Ifremer and French Ecology, Durable Development Ministry (in French). Cossa, D., Gobeil, C., 2000. Mercury speciation in the lower St. Lawrence estuary. Can. J. Fish. Aquat. Sci. 57, 138e147. Dominique, Y., Muresan, B., Duran, R., Richard, S., Boudou, A., 2007. Simulation of the chemical fate and bioavailability of liquid elemental mercury drops from gold mining in Amazonian freshwater systems. Environ. Sci. Technol. 41 (21), 7322e7329. Dorea, J.G., Barbosa, A.C., 2007. Anthropogenic impact of mercury accumulation in fish from the Rio Madeira and Rio Negro rivers (Amazonia). Biol. Trace Elem. Res. 115 (3), 243e254. Durrieu, G., Maury-Brachet, R., Boudou, A., 2005. Goldmining and mercury contamination of the piscivorous fish, Hoplias aimara in French Guiana (Amazon basin). Ecotoxicol. Environ. Saf. 60 (3), 315e323. Fink, L., 2002. Status Report on the Effect of Water Quantity and Quality on Methylmercury Production, pp. 2B2.1e2B2.17, EPA. Fleming, E.J., Mack, E.E., Green, P.G., Douglas, C.N., 2006. Mercury methylation from unexpected sources: molybdate-inhibited freshwater sediments and iron-reducing bacterium. Appl. Environ. Microbiol. 72 (1), 457e464. Fostier, A.-H., Forti, M.c., Guimaraes, J.R.D., Melfi, A.J., Boulet, R., Espirito Santo, C.M., Krug, F.J., 2000. Mercury fluxes in a natural forested Amazonian catchment (Serra do Navio, Amapa State, Brazil). Sci. Total Environ. 260, 201e211. Frery, N., Maury-Brachet, R., Maillot, E., Deheeger, M., Merona de, B., Boudou, A., 2001. Goldmining activities and mercury contamination of native Amerindian communities in French Guiana: key role of fish in dietary uptake. Environ. Health Perspect. 109 (5), 449e456. Grimaldi, C., Grimaldi, M., Gue´dron, S., 2008. Mercury distribution in tropical soil profiles related to origin of mercury and soil processes. Sci. Total Environ. 401, 121e129. Grimaldi, C., Grimaldi, M., Millet, A., Bariac, T., Boule`gue, J., 2004. Behavior of chemical solutes during a storm in a rainforested headwater catchment. Hydrol. Process. 18, 93e106. Guedron, S., Grangeons, S., Lanson, B., Grimaldi, M., 2009. Mercury speciation in a tropical soil association; consequence of gold mining on Hg distribution in French Guiana. Geoderma 153, 331e346. Guedron, S., Grimaldi, C., Chauvel, C., Spadini, C., Grimaldi, M., 2006. Weathering versus atmospheric contributions to mercury concentrations in French Guiana soils. Appl. Geochem. 21, 2010e2022.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 5 9 e2 6 6 9
Hall, B.D., Aiken, G.R., Krabbenhoft, D.P., Marvin-DiPasquale, M., Swarzenski, C.M., 2008. Wetlands as principal zones of methylmercury production in southern Louisiana and the Gulf of Mexico region. Environ. Pollut. 154 (1), 124e134. Harper, B.L., Harris, S.G., 2008. A possible approach for setting a mercury risk-based action level based on tribal fish ingestion rates. Environ. Res. 107 (1), 60e68. Herschy, R.W., 1995. Streamflow Measurement. Elsevier Applied Science. Hurley, J.P., Benoit, J.M., Babiarz, C.L., Shafer, M.M., Andren, A.W., Sullivan, J.R., Hammond, R., Webb, D.A., 1995. Influence of watershed characteristics on mercury levels in Wisconsin rivers. Environ. Sci. Technol. 29 (7), 1867e1875. Jezequel, D., Brayner, R., Metzger, E., Viollier, E., Prevot, F., Fievet, F., 2007. Two-dimensional determination of dissolved iron and sulfur species in marine sediment pore-waters by thin-film based imaging. Thau lagoon (France). Estuar. Coast. Shelf Sci. 72 (3), 420e431. Lambertsson, L., Nilsson, M., 2006. Organic material: the primary control on mercury methylation and ambient methyl mercury concentrations in estuarine sediments. Environ. Sci. Technol. 40 (6), 1822e1829. Langley, D.G., 1973. Mercury methylation in an aquatic environment. J. Water Pollut. Control Fed. 45 (1), 44e51. Lawson, N.M., Mason, R.P., 2001. Concentration of mercury, methylmercury, cadmium, lead, arsenic, and selenium in the rain and stream water of two contrasting watersheds in Western Maryland. Water Res. 35 (17), 4039e4052. Lechler, P.J., Miller, J.R., Lacerda, L.D., Vinson, D., Bonzongo, J.-C., Lyons, W.B., Warwick, J.J., 2000. Elevated mercury concentrations in soils, sediments, water, and fish of the Madeira river basin, Brazilian Amazon: a function of natural enrichments? Sci. Total Environ. 260, 87e96. Liang, L., Bloom, N.S., Horvat, M., 1994. Simultaneous determination of mercury speciation in biological materials by GC/CVAFS after ethylation and room-temperature precollection. Clin. Chem. 40 (4), 602e607. Maia, P.D., Maurice, L., Tessier, E., Amouroux, D., Cossa, D., Pe´rez, M., Moreira-Turcq, P., Rhe´ault, I., 2009. Mercury distribution and exchanges between the Amazon river and connected floodplain lakes. Sci. Total Environ. 407, 6073e6084. Malm, O., 1998. Gold mining as a source of mercury exposure in the Brazilian Amazon. Environ. Res. 77 (2), 73e78. Marchand, C., Lallier-Verges, E., Baltzer, F., Alberic, P., Cossa, D., Baillif, P., 2006. Heavy metals distribution in mangrove sediments along the mobile coastline of French Guiana. Mar. Chem. 98 (1), 1e17. Mason, R.P., Reinfelder, J.R., Morel, F.M.M., 1995. Bioaccumulation of mercury and methylmercury. Water Air Soil Pollut. 80, 915e921. Mehrotra, A.S., Horne, A.J., Sedlak, D.L., 2003. Reduction of Net mercury methylation by iron in desulfobulbus propionicus (1pr3) cultures: Implications for engineered wetlands. Environ. Sci. Technol. 37 (13), 3018e3023. Mehrotra, A.S., Sedlak, D.L., 2005. Decrease in net mercury methylation rates following iron amendment to anoxic wetland sediment slurries. Environ. Sci. Technol. 39 (8), 2564e2570. Mile´si, J.P., Egal, E., Ledru, P., Vernhet, Y., Thie´blemont, D., Cocherie, A., Tegyey, M., Martel-Jantin, B., Lagny, P., 1995. Mineralizations of the northern French Guiana in their geological setting. Mining Res. Chron. 518, 5e58 (in French). Muresan, B., 2006. Mercury geochemistry in the continuum of Petit Saut reservoir and the Sinnamary estuary, French Guiana (in French). P.H.D, University of Bordeaux I, PhD thesis.
2669
Muresan, B., Cossa, D., Jezequel, D., Prevot, F., Kerbellec, S., 2007a. The biogeochemistry of mercury at the sediment-water interface in the Thau lagoon. 1. Partition and speciation. Estuarine Coastal Shelf Sci. 72 (3), 472e484. Muresan, B., Cossa, D., Richard, S., Burban, D., 2007b. Mercury speciation exchanges at the air-water interface of a tropical artificial reservoir, French guiana. Sci. Total Environ. 385 (1e3), 132e145. Parker, J.L., Bloom, N.S., 2005. Preservation and storage techniques for low-level aqueous mercury speciation. Sci. Total Environ. 337 (1e3), 253e263. Peters, N., Ratcliffe, E., 1998. Tracing hydrologic pathways using chloride at the Panola mountain research watershed, Georgia, USA. Water Air Soil Pollut. 105, 263e275. Pfeiffer, W.C., Lacerda, L.D., Salomon, W., Malm, O., 1993. Environmental fate of mercury from gold mining in the Brazilian Amazon. Environ. Rev. 1, 26e37. Porvari, P., 1995. Mercury levels of fish in Tucurui hydroelectric reservoir and in River Moju in Amazonia, in the state of Para, Brazil. Sci. Total Environ. 175 (2), 109e117. Roos-Barraclough, F., Givelet, N., Martinez-Cortizas, A., Goodsite, M.E., Biester, H., Shotyk, W., 2002. An analytical protocol for determination of total mercury concentration in solid peat samples. Sci. Total Environ. 292, 129e139. Roulet, M., Lucotte, M., Canuel, R., Farella, N., Freitos Goch, Y.G.D., Pacheco Peleja, J.R., Guimaraes, J.-R.D., Mergler, D., Amorim, M., 2001. Spatio-Temporal geochemistry of mercury in waters of the Tapajos and Amazon rivers, Brazil. Limnol. Oceanogr. 46 (5), 1141e1157. Roulet, M., Lucotte, M., Canuel, R., Rheault, I., Tran, S., De Freitos Gog, Y.G., Farella, N., Souza do Vale, R., Sousa Passos, C.J., De Jesus de Silva, E., Mergler, D., Amorim, M., 1998a. Distribution and partition of total mercury in waters of the Tapajos river basin, Brazilian Amazon. Sci. Total Environ. 213, 203e211. Roulet, M., Lucotte, M., Dolbec, J., Gogh, Y.F., Pelaja, J.R.P., 1999a. 5th International Conference Mercury as a Global Pollutant, Rio de Janeiro, Brazil. Roulet, M., Lucotte, M., Farella, N., Serique, G., Coelho, H., Sousa Passos, C.J., De Jesus Da Silva, E., Scavone De Andrade, P., Mergler, D., Guiaraes, J.R.D., Amorim, M., 1999b. Effects of recent human colonization on the presence of mercury in Amazonian ecosystems. Water Air Soil Pollut. 112, 297e313. Roulet, M., Lucotte, M., Saint-Aubin, A., Tran, S., Rheault, I., Farella, N., Da Silva, E.D., Dezencourt, J., Passos, C.J.S., Soares, G. S., Guimaraes, J.R.D., Mergler, D., Amorim, M., 1998b. The geochemistry of mercury in central Amazonian soils developed on the Alter-do-Chao formation of the lower Tapajos river valley, Para state, Brazil. Sci. Total Environ. 223 (1), 1e24. Soulsby, C., Tetzlaff, D., van den Bedem, N., Malcolm, I.A., Bacon, P.J., Youngson, A.F., 2007. Inferring groundwater influences on surface water in montane catchments from hydrochemical surveys of springs and streamwaters. J. Hydrol. 333, 199e213. Tseng, C.M., de Diego, A., Pinaly, H., Amouroux, D., Donard, O.F.X., 1998. Cryofocusing coupled to atomic absorption spectrometry for rapid and simple mercury speciation in environmental matrices. J. Anal. Atom Spectrom. 13, 755e764. Ullrich, S.M., Tanton, T.W., Abdrashitova, S.A., 2001. Mercury in the aquatic environment: a review of factors affecting methylation. Crit. Rev. Environ. Sci. Technol. 31 (3), 241e293. Wasserman, J.C., Hacon, S., Wasserman, M.A., 2003. Biogeochemistry of mercury in the Amazonian environment. Ambio 32, 336e342. Webster, R., 2001. Statistics to support soil research and their presentation. Eur. J. Soil Sci. 52 (2), 331e340.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 7 0 e2 6 8 0
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Evaluating the importance of faecal sources in human-impacted waters5 Mary E. Schoen a,*, Jeffrey A. Soller b, Nicholas J. Ashbolt a a b
Office of Research and Development, U.S. Environmental Protection Agency, 26 West Martin Luther King Drive, Cincinnati, OH 45268, USA Soller Environmental, LLC, 3022 King St, Berkeley CA 94703, USA
article info
abstract
Article history:
Quantitative microbial risk assessment (QMRA) was used to evaluate the relative contri-
Received 23 November 2010
bution of faecal indicators and pathogens when a mixture of human sources impacts
Received in revised form
a recreational waterbody. The waterbody was assumed to be impacted with a mixture of
20 January 2011
secondary-treated disinfected municipal wastewater and untreated (or poorly treated)
Accepted 21 February 2011
sewage, using Norovirus as the reference pathogen and enterococci as the reference faecal
Available online 2 March 2011
indicator. The contribution made by each source to the total waterbody volume, indicator density, pathogen density, and illness risk was estimated for a number of scenarios that
Keywords:
accounted for pathogen and indicator inactivation based on the age of the effluent (source-
Recreational water
to-receptor), possible sedimentation of microorganisms, and the addition of a non-path-
Quantitative microbial risk
ogenic source of faecal indicators (such as old sediments or an animal population with low
assessment
occurrence of human-infectious pathogens). The waterbody indicator density was held
Human-impact
constant at 35 CFU 100 mL1 enterococci to compare results across scenarios. For the
Norovirus
combinations evaluated, either the untreated sewage or the non-pathogenic source of
Sedimentation
faecal indicators dominated the recreational waterbody enterococci density assuming a culture method. In contrast, indicator density assayed by qPCR, pathogen density, and bather gastrointestinal illness risks were largely dominated by secondary disinfected municipal wastewater, with untreated sewage being increasingly less important as the faecal indicator load increased from a non-pathogenic source. The results support the use of a calibrated qPCR total enterococci indicator, compared to a culture-based assay, to index infectious human enteric viruses released in treated human wastewater, and illustrate that the source contributing the majority of risk in a mixture may be overlooked when only assessing faecal indicators by a culture-based method. Published by Elsevier Ltd.
1.
Introduction
Numerous epidemiology studies have generated evidence of adverse health outcomes attributed to swimming in municipal disinfected wastewater effluent-impacted waters, or what we will call here, human-impacted waters (Pru¨ss, 1998; 5
Wade et al., 2006; Zmirou et al., 2003). Other epidemiology studies have focused on bather risk from non-human sources of faecal contamination, and are summarized in reviews by Sinton et al. (1998) and updated by Soller et al. (2010b). In 2012, the U.S. Environmental Protection Agency (EPA) will issue new or revised recreational water quality criteria. Although the
The views expressed in this article are those of the authors and do not necessarily reflect the views or policies of the U.S. Environmental Protection Agency. * Corresponding author. E-mail address:
[email protected] (M.E. Schoen). 0043-1354/$ e see front matter Published by Elsevier Ltd. doi:10.1016/j.watres.2011.02.025
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 7 0 e2 6 8 0
Nomenclature S Dwp DSp Vs FSp ts kp DwI DSI DSqPCR VF FSI
source of the indicator (i.e. POTW, Raw, or Other) or the pathogen (i.e. POTW or Raw) density of pathogen ( p) in the waterbody (w) density of pathogen ( p) in a source (S ) volume of waste from a source (S ) fraction of pathogen ( p) from a source (S ) that sediments out of the waterbody age of the waste from a source (S ) when mixed into the waterbody decay coefficient of the pathogen ( p) in the waterbody density of indicator (I ) by culture method in the waterbody (w) density of indicator (I ) by culture method in a source (S ) density of indicator (I ) by qPCR method in a source (S ) volume of water in the waterbody fraction of indicator (I ) from a source (S ) that sediments out of the waterbody
existing epidemiology studies are extremely valuable in this process, there are many faecal source mixtures unstudied, and indeed, some may not be possible to study through current epidemiology approaches. In previous work, we used quantitative microbial risk assessment (QMRA) as a complement to the epidemiology evidence to further understand the potential risks from a variety of faecally contaminated waterbodies (Schoen and Ashbolt, 2010; Soller et al., 2010a,b; U.S. EPA, 2010). We presented a QMRA approach for predicting and comparing the potential probability of gastrointestinal (GI) illness from accidental ingestion of recreational water impacted by alternative sources of faecal contamination (Schoen and Ashbolt, 2010). Waterbodies impacted by seagull excreta and primary sewage effluent were compared at the same faecal indicator bacterial density with the result of a lower predicted illness risk from seagullimpacted waters. The same approach was extended to compare the relative risks from exposure to recreational waters impacted by direct contamination by gulls, chickens, pigs, and/or cattle and those associated with human-impacted waters (Soller et al., 2010b; U.S. EPA, 2010). The primary finding from this work was that the predicted illness risk associated with non-sewage impacted beaches was dependent on the source of contamination. Generally, the existing QMRA work has estimated GI risks for recreational water exposures from a single source of faecal contamination, with the exception of an illustrative example of a seagull and primary sewage impacted recreational beach presented in Schoen and Ashbolt (2010). Here, we used QMRA to model mixtures of faecal contamination and compare different ways to interpret the mixture. Our attention was first on human-impacted waters because the existing epidemiology and QMRA studies provide estimates of GI risk. In previous work, we used QMRA to evaluate the etiological agents potentially responsible for the
2671
KI
decay coefficient of the indicator (I ) by culture method in the waterbody decay coefficient of the indicator (I ) by qPCR KqPCR method in the waterbody load of indicator (I ) from a non-pathogenic, LI environmental source (S ) percent of the total waterbody volume from PSV a source (S ) percent of the total indicator (I ) load from PSI a source (S ) percent of the total pathogen ( p) load from PSp a source (S ) volume of waterbody ingested over duration of Vi one swimming event pathogen dose ingested over duration of one uwp swimming event P(illjinf) conditional probability of illness given infection Pinf(a,b,-u) dose-response function for pathogen ( p) with parameters (a,b) total probability of illness PillT probability of illness attributable to a source (S ). PillS
reported epidemiology results in human-impacted waters (Soller et al., 2010a). The results of that work indicated that human enteric viruses and in particular, Norovirus represented the vast majority of the observed swimming-associated GI illnesses in the human-impacted freshwaters studied. The epidemiology studies that supported the 1986 Ambient Water Quality Criteria (AWQC) (Cabelli et al., 1982; Dufour et al., 1984) probably resulted from a combination of both well treated, disinfected municipal wastewater and less well treated or untreated sewage contamination (directly from swimmers, poorly operating septic systems, sewage bypassing treatment etc.) (Soller et al., 2010a). Here we simulate possible mixtures of untreated (or poorly treated) sewage and secondary-treated disinfected municipal wastewater and then apply QMRA, using Norovirus as the reference pathogen, to estimate each source contribution to the total GI risk. The main objective was to synthesize the various approaches commonly used to describe a mixture of two (or more) sources of faecal contamination. The least complicated way to describe a mixture is to report and compare the mass or volume of waste from each source. A second way to describe a mixture is to estimate the portion of the total indicator load from a source, such as in microbial source tracking (Wang et al., 2010). Rarely does a study attempt to determine the actual pathogen or risk contribution from each source in a mixture of sources. Hence, four approaches were used to describe mixtures: (a) the percent of the total waterbody volume from each source, (b) the percent of total faecal indicator load by traditional culture and rapid methods from each source, (c) the percent of the waterbody index pathogen load from each source, and (d) the probability of illness attributable to each source. All four measures are synthesized in the discussion to provide context for understanding mixtures of sources in human-impacted waters.
2672
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 7 0 e2 6 8 0
2.
Methods
2.1.
Model human-impacted waterbody
To compare the four approaches commonly used to describe a mixture of faecal contamination, a model of a simplified, human-impacted waterbody was constructed. This waterbody was a completely mixed, batch reactor with a known volume of water (VF). Two sources of contamination were mixed into this waterbody, untreated (or poorly treated) sewage (Raw) and secondary-treated disinfected publicallyowned treatment works (POTW) municipal wastewater effluent. The mixture was subject to two constraints. First, the total waterbody faecal indicator level was that of the 1986 AWQC for marine waters, 35 CFU 100 mL1 enterococci. The
DIw
FSI fraction of indicator (I ) by culture method from a source (S ) that sediments out of the waterbody ts age of the waste from a source (S ) when mixed into the waterbody kI decay coefficient of the indicator by culture method (I ) in the waterbody LI load of indicator (I ) by culture method from a nonpathogenic, environmental source (S ) VF volume of fresh or saltwater in the waterbody. Equation (1a) accounts for the first-order inactivation of indicator organisms, sedimentation of the indicator out of the waterbody, and an indicator load from a non-pathogenic, environmental source such as some populations of seagulls with low occurrence of human-infectious pathogens (Schoen and Ashbolt, 2010). Equation (1a) can be written explicitly in terms of the two human sources (Raw & POTW) as:
PI I I DIPOTW VPOTW 1 FIPOTW 10tPOTW k þ DIRaw VRaw 1 FIRaw 10tRaw k þ DIw Other ðVF þ VPOTW þ VRaw Þ 100 ¼ VF þ VPOTW þ VRaw
marine geometric mean standard of 35 CFU 100 mL1 enterococci (75th percentile value of 104 CFU 100 mL1) is based on the previously reported occurrence of GI illness associated with swimming in marine recreational waters impacted by effluent from POTWs (U.S. EPA, 1986). Second, the total probability of GI illness for the waterbody was set at the corresponding risk of 0.03, based on the epidemiology studies performed at human-impacted waters (Soller et al., 2010a; Wade et al., 2008). Since there is a range of risk associated with an observed faecal indicator density in a humanimpacted waterbody (Wade et al., 2003), the second constraint was later relaxed to explore a range of GI risks (i.e. 0.001e0.3). Alternative scenarios were constructed to include the possible physical processes of indicator and pathogen removal such as sedimentation in the environment (Skraber
Dpw ¼
In Equation (1b), the environmental source load of enterococci by culture method LI was rewritten as a function of total indicator load. When there was no sedimentation, no other environmental source of indicator, and the effluents were fresh (such that FSI ¼ 0, LI ¼ 0 CFU, and ts ¼ 0 h). Equation (1b) reduced to: DIw ¼
DIPOTW VPOTW þ DIRaw VRaw VF þ VPOTW þ VRaw
P Dpw ¼
S
p p p DS VS 1 FS 10tS k P VF þ S VS
Dpw ¼
(2a)
(2b)
p
where DwI density of indicator (I) by culture method in the waterbody (w) DSI density of indicator (I) by culture method in a source (S) VS volume of waste from a source (S)
(1c)
Similarly, the total pathogen density in the waterbody was expressed in full form (Eq. (2a)), human-impacted form (Eq. (2b)), and reduced form (Eq. (2c)):
p p p p p p DPOTW VPOTW 1 FPOTW 10tPOTW k þ DRaw VRaw 1 FRaw 10tRaw k VF þ VPOTW þ VRaw
et al., 2009) and removal by sunlight inactivation. The total indicator density by traditional culture method in the waterbody (DwI) accounting for the physical processes of indicator and pathogen removal was expressed: P I I D VS 1 FIS 10tS k þ LI P (1a) DIw ¼ S S VF þ S VS
(1b)
p
DPOTW VPOTW þ DRaw VRaw VF þ VPOTW þ VRaw
(2c)
where Dwp density of pathogen (p) in the waterbody (w) DSp density of pathogen (p) in a source (S) Vs volume of waste from a source (S) FSp fraction of pathogen (p) from a source (S) that sediments out of the waterbody ts age of the waste from a source (S) when mixed into the waterbody kp decay coefficient of the pathogen (p) in the waterbody VF volume of fresh or saltwater in the waterbody.
2673
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 7 0 e2 6 8 0
The volumes of untreated (or poorly treated) sewage and secondary disinfected municipal wastewater were estimated using Equations (1b) and (2b) and the specified constraints on indicator density and risk. First, the density of the indicator by culture method in the waterbody in Equation (1b) was set so that DwI ¼ 35 CFU 100 mL1 enterococci. Then the density of the pathogen (Dwp) in Equation (2b) was set based on a target probability of GI illness PillT ¼ 0.03. Finally, Equation (1b) was rewritten to solve for the volume of the Raw source (VRaw):
VRaw
2.3.
Measures of impact
To investigate the relative importance of each source contributing pathogens to the recreational waterbody, five
PI PI I DIw VF 1 Other þ VPOTW DIw DIPOTW 1 FIPOTW 10tPOTW k DIw Other 100 100 ¼ I POther I I tRaw kI I 1 þ DRaw 1 FRaw 10 Dw 100
And substituted into Equation (2b) to solve for the volume of secondary disinfected municipal wastewater (VPOTW):
p
p
DIw DRaw Y DPOTW XDIw
measures of impact were considered based on predicted source volumes: the percent of the total waterbody volume
PIOther PI p p p p DIRaw RDPOTW X þ DIRaw RDpw DIPOTW QDpw þ DIPOTW QDRaw Y þ DPOTW XDIw þ DRaw YDIw Other 100 100
where: I Q ¼ 1 FIPOTW 10tPOTW k
p p X ¼ 1 FPOTW 10tPOTW k Y ¼ 1
p FRaw
PVS ¼ tRaw
10
kp
VS P 100 VF þ VS
(7)
S
Here, the density of the pathogen in the POTW wastewater (DPOTWp) was expressed as a function of the density in the p p p untreated source DRAWp, DPOTW ¼ DRaw =10LRPOTW where LRPOTWp is the log reduction of the pathogen in secondary disinfected wastewater effluent. Solving Equations (3) and (4) provided a mixture of effluent volumes that resulted in a waterbody enterococci concentration of 35 CFU 100 mL1 and the specified target risk (PillT). The target risk (PillT) was translated into a target density of pathogen in the waterbody (Dwp) using a dose-response relationship for the selected reference pathogen Pinf(a,b,-u): PTill ¼ 1 Pinf a; b; upw PðilljinfÞ (5) where P(illjinf) is the conditional probability of illness given infection. Using Eq. (5), the pathogen dose ingested during one swimming event (uwp) to reach the target risk was estimated. Then the pathogen dose associated with the target risk was divided by the volume ingested during one swimming event (Vi) to estimate the target density of pathogen in the waterbody: Dpw ¼ upw =Vi
(4)
from a source (PSV); the percent of the total indicator load from a source using a culture method (PSI); the percent of the total indicator load from a source using a qPCR method (PSqPCR); the percent of the total pathogen load from a source (PSp); and the probability of illness attributable to a source (PillS):
I R ¼ 1 FIRaw 10tRaw k
(3)
PI p p DIRaw RDw VF þ DRaw YDIw VF 1 Other 100
VPOTW ¼
2.2.
Norovirus dose-response relationship for GI infection was used with hypergeometric parameters (a,b) estimated by Teunis et al. (2008) for healthy adults, along with the conditional probability of illness given infection P(illjinf) (see Section 2.6.2).
(6)
Reference pathogen
The reference pathogen used in this analysis was Norovirus, as justified previously (Soller et al., 2010a,b). The non-aggregated
PIS ¼ P S
I DIS VS 1 FIS 10tS k 100 DIS VS 1 FIS 10tS kI þ LI qPCR
¼P PaPCR S S
DS
VS 10tS k
qPCR DS VS
10tS
(8)
qPCR
kqPCR
þ LI
100
(9)
p p p D VS 1 FS 10tS k p PS ¼ P S p 100 p t S DS VS 1 FS 10 kp
(10)
p PSill ¼ 1 Pinf a; b; upw PS PðilljinfÞ
(11)
S
where: uwp pathogen dose ingested over duration of one swimming event P(illjinf) conditional probability of illness given infection Pinf(a,b,-u) dose-response function for pathogen (p) with parameters (a,b) PillS probability of illness attributable to a source (S).
2.4.
Scenarios
Six scenarios were explored that captured the potential physical processes occurring in the waterbody (Table 1). Runs
2674
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 7 0 e2 6 8 0
Table 1 e Scenario and parameter inputs. Run
Constraintsa PillT
DwI
1aei 2aeb
0.001e0.3c 0.03
35 35
3aec
0.03
4aeb
Parametersb
Description of sources POtherI
tPOTW
0 a ¼ 50 b ¼ 90 0
0 0
0 0
0 0
0 0
0 0
0 0
35
Raw & POTWd Raw & POTW plus other source of enterococci Raw & aged POTW
0
0
0
0
0
0.03
35
Aged Raw & POTW
0
a¼6 b ¼ 10 c ¼ 72 0
0
0
0
0
5aec
0.03
35
Raw & POTW with Norovirus sedimentation
0
0
a¼6 b ¼ 10 0
0
0
0
6aec
0.03
35
POTW & Raw with proportional Norovirus and enterococci sedimentation
0
0
0
a ¼ 0.50 b ¼ 0.75 c ¼ 0.99 0
0
a ¼ 0.5 b ¼ 0.90 c ¼ 0.99
a ¼ 0.5 b ¼ 0.90 c ¼ 0.99
tRaw
FPOTWp
FPOTWI
FRawp
FRawI
a Probability of GI illness and density of indicator enterococci in waterbody. b P is the percent of indicator in waterbody from non-pathogenic, non-human, other source; t is the age of source (POTW or Raw) when mixed in waterbody; F is the fraction of pathogen ( p) or indicator (i) that sediments out of the source (POTW or Raw). c Run was repeated for a range of risks separately (0.001, 0.01, 0.02, 0.03, 0.04, 0.1, 0.2, 0.3). d Fresh poorly treated sewage (Raw) and fresh secondary-treated disinfected municipal wastewater (POTW).
1aei considered the simplified system described in Equations (1b) and (2b) with fresh untreated (or poorly treated) sewage (Raw) and freshly discharged secondary-treated disinfected municipal wastewater (POTW). In Runs 1aei, a range of target probability of illness was explored around 0.03 (i.e. 0.001e0.3). Runs 2aeb added a non-pathogenic source of faecal indicator bacteria, such as some populations of seagulls with low occurrence of human-infectious pathogens (Schoen and Ashbolt, 2010). Runs 3aec assumed a secondary-treated disinfected municipal wastewater (POTW) effluent aged up to 72 h. Runs 4aeb assumed an untreated sewage source aged up to 10 h. Runs 5aec incorporated the removal of the POTW Norovirus genomes by sedimentation. Finally, Runs 6aec incorporated the removal of the untreated sewage Norovirus genomes and indicator bacteria by sedimentation.
2.5.
Computation
Many of the input parameters in the equations for the waterbody pathogen density (Equations (5) and (6)), source volumes (Equations (3) and (4)), and measures of impact (Equations (7)e(11)) have natural variability. Therefore, these outputs were estimated concurrently for each scenario run in a Monte Carlo simulation of 10,000 iterations. For each iteration, a set of input variables was drawn from probability distributions used to capture the observed natural variability (Table 2). Some of the iterations resulted in a negative volume, indicating that the selected faecal indicator density and target risk were not physically possible given that set of Monte Carlo input parameters. The negative iterations were discarded when reporting results. When the number of negative volume iterations was large (i.e. >90%), then the target risk was likely not achievable given the sources and enterococci waterbody density. Simulations were implemented in R version 2.8.1 (R Development Core Team, 2009).
2.6.
Parameter inputs
2.6.1.
Characteristics of sources at human-impacted beaches
The input parameters related to the human sources of faecal indicator and pathogen contamination are presented in Table 2. The microbial and pathogenic content of untreated (or primary) sewage and secondary-treated disinfected municipal wastewater have been reported in numerous studies across geographic regions (Bae and Schwab, 2008; de Roda Husman et al., 2009; Haramoto et al., 2006; He and Jiang, 2005; Katayama et al., 2008; Lemarchand and Lebaron, 2003; Noble et al., 2004; Rose et al., 2004; Teunis et al., 2008; Tschobanoglous et al., 2003; Varma et al., 2009; Walters et al., 2009). These studies showed that primary and raw sewage effluent contains extremely high densities of enterococci faecal indicator (104.8e107.0 CFU 100 mL1 and 104.85e106.25 calibrator cell equivalents (CCE) 100 mL1) while the secondary-treated disinfected municipal wastewater contains low densities by traditional culture (100.5e101.7 CFU 100 mL1) but higher densities by quantitative PCR (qPCR) (103.56e105.14 CCE 100 mL1). Estimating the density of enterococci by qPCR in wastewater or untreated sewage was complicated by the various existing methodologies. Evaluation of total Enterococcus spp. in secondary-treated disinfected effluent from three POTWs during one sampling event provides some insight into the variability of qPCR in sewage (Varma et al., 2009). The minimum and maximum reported mean enterococci densities in one sampling event under normal operating conditions were 107.02 and 107.65 target sequences (TS) 100 mL1 in untreated sewage samples and 104.95 and 105.46 TS 100 mL1 in secondary-treated disinfected wastewater. This was roughly equivalent to 105.61 and 106.25 CCE 100 mL1 and 103.56 and 104.06 CCE 100 mL1 using a conversion factor of 0.04 TS/CCE (provided by Dr Richard Haugland via personal communication). In a separate study, He and Jiang (2005) estimated total Enterococcus spp. using qPCR for primary and secondary
2675
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 7 0 e2 6 8 0
Table 2 e Parameter inputs for Monte Carlo simulation. Parameter
Description
Units
Reference
DRawI
Density enterococci in primary sewage
Log10 CFU 100 mL1
DPOTWI DRawqPCR DPOTWqPCR kqPCR kI DRawp
Density enterococci in POTW Density qPCR enterococci in primary sewage Density qPCR enterococci in POTW Log10 decay of qPCR enterococci Log10 decay of enterococci Density Norovirus in raw sewage
Log10 CFU 100 mL1 Log10 CCE 100 mL1 Log10 CCE 100 mL1 h1 h1 Genomes 100 mL1
LRPOTWp kp
Reduction of Norovirus in POTW Log10 decay of Norovirus
Log10 removal h1
Pilljinf (a, b)c
Conditional probability of illness on infection Non-aggregated, hypergeometric Norovirus dose-response parameters
NA NA
Lemarchand and Lebaron (2003); Tschobanoglous et al. (2003) Rose et al. (2004) He and Jiang (2005); Varma et al. (2009) He and Jiang (2005); Varma et al. (2009) Walters et al. (2009) Noble et al. (2004); Walters et al. (2009) Haramoto et al. (2006); Katayama et al. (2008) Haramoto et al. (2006) Bae and Schwab, 2008; de Roda Husman et al. (2009) Teunis et al. (2008) Teunis et al. (2008)
Parameter rangea 4.8
7
0.5 4.85 3.56 0.012 0.04 2
1.7 6.25 5.14 NA 0.111 5
2.23 3 0.001 0.004 0.03 NA
1.0b NA
a Parameter range corresponds to the lower an upper bounds of a Log-uniform distribution for indicator and pathogen densities, a uniform distribution for the inactivation coefficients, and a triangle distribution for the conditional probability of illness. b Triangle distribution a ¼ 0.03, b ¼ 0.6, c ¼ 1.0 based on the range of conditional illnesses observed by Teunis et al. (2008). ^ ¼ (0.04, 0.055). c 10,000 bootstrap parameter estimates provided by Teunis (pers. comm.) with maximum likelihood estimates of ð^ a; bÞ
effluent with chlorine disinfection at a single POTW. The minimum and maximum reported mean enterococci density for three sampling events under normal operating conditions were 105.54 and 106.41 genomes 100 mL1 in primary effluent and 105.72 and 105.84 genomes 100 mL1 in secondary-treated chlorinated effluent. This was roughly equivalent to 104.85 and 105.60 CCE 100 mL1 and 105.03 and 105.14 CCE 100 mL1 using the conversion factor of 0.2 genomes/CCE (provided by Dr. Richard Haugland, US EPA via personal communication). Although the limited available data may underestimate the variability of enterococci density by qPCR in municipal wastewater, the intent of including the qPCR indicator was to account for the apparent difference between culture and qPCR enterococci densities in secondary-treated disinfected wastewater. Similar to enterococci by qPCR, Norovirus genomes, as measured by real-time RT-PCR, have been detected in both untreated sewage and secondary-treated disinfected wastewater (Haramoto et al., 2006; Katayama et al., 2008). These microbial and pathogenic densities are represented by Log-uniform distributions (Table 2) to capture the range of observed densities. No correlation between pathogens and indicators was included, consistent with studies that indicate little relationship between indicator and pathogen densities in sewage (Payment and Locas, 2010). The inactivation coefficient of enterococci kI was modeled as a uniform distribution with parameters based on the range of inactivation rates reported by Noble et al. (2004) and Walters et al. (2009) for summer seawater at ambient temperatures. The inactivation of infectious Norovirus is unknown, due to the lack of an infectivity assay, yet there is data for murine Norovirus and other human enteric viruses. Therefore, the inactivation of infectious Norovirus kN was modeled as a uniform distribution with parameters based on the range of inactivation rates described by Bae and Schwab (2008) and de Roda Husman et al. (2009). de Roda Husman et al. (2009) measured inactivation of coxsackievirus B4, poliovirus 1, and polovirus 2 at 22 C in the dark at a laboratory scale. Bae and Schwab (2008) compared the
inactivation of murine Norovirus, feline calicivirus, poliovirus 1, and Norovirus by culture (where applicable) and qPCR at 25 C in surface waters with no UV, and dropped feline calicivirus and polovirus 1, being unsuitable Norovirus surrogates due to their rapid inactivation. A few studies estimate the persistence of the ‘total enterococci’ qPCR signal over time in the environment (Bae and Wuertz, 2009; Walters et al., 2009); however this subject remains underinvestigation. The decay coefficient of 0.012 h1 was initially set based on light and dark seawater microcosm experiments (Walters et al., 2009). With this selected decay coefficient, the qPCR signal is more persistent than enterococci by culture, but less persistent than Norovirus.
2.6.2.
Dose-response parameters
The hypergeometric dose-response parameters (a,b) and the conditional probability of illness given infection P(illjinf) for non-aggregated Norovirus have uncertainty. We accounted for this uncertainty by using 10,000 pairs of dose-response parameters (provided by Dr. Teunis via personal communi^ ¼ (0.04, cation) with maximum likelihood estimates of ð^ a; bÞ 0.055). The assumed distribution of P(illjinf) presented in Table 2 is a triangle distribution with a mode of 0.6, minimum of 0.03, and maximum of 1.0. Limited data is available for estimating the P(illjinf), but work by Teunis et al. (2008) indicated that the value is variable with a maximum, conservative estimate of 1.0. The average across high-dose trials was 0.6 and hence set as the mode.
3.
Results
3.1. Fresh secondary-treated disinfected wastewater (POTW) and untreated sewage (raw) Runs 1aei estimated POTW and Raw indicator and pathogen densities in human-impacted recreational waters assuming
2676
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 7 0 e2 6 8 0
3.1.1.
Range of possible GI illness
Using Runs 1aei input parameters, the range of probabilities of illness resulting in POTW and Raw mixture solutions was 0.01e0.3. Many of the Monte Carlo iterations did not satisfy the waterbody risk and enterococci density constraints: 22e43% of the iterations resulted in solutions when the GI risk was constrained from 0.3 down to 0.01. Run 1b (probability of illness of 0.001) was discarded because <1% of the Monte Carlo iterations resulted in feasible solutions. The risk range generating feasible solutions for a waterbody enterococci density of 35 CFU 100 mL1 changed based on the input parameters for alternative scenarios described in Table 1.
3.1.2.
Source impact
Five measures of source impact, Equations (7)e(11) were estimated for a waterbody with fresh POTW and Raw inputs (Runs 1aei) and an enterococci density of 35 CFU 100 mL1. Four of these measures showed little variation across the range of possible probabilities of illness (i.e. 0.01e0.3; results not shown) e the percent of the total waterbody indicator load using a culture method from each source (PSI), the percent of the total indicator load using a qPCR method from a source (PSqPCR), the percent of the total waterbody volume from each source (PSV), and the percent of the total waterbody Norovirus load from each source (PSp). Representative results of the measures of source impact (Run 1a) are shown in Fig. 1. When impact was measured in terms of the indicator density by a culture method, the Raw source contributed 99% of the enterococci load while the POTW effluent contributed less than 1% (Fig. 1) for all Runs 1aei. In contrast, when impact was measured in terms of the indicator density by a qPCR method, the Raw source contributed less than 1% of the
Percent from Source
100 10 1
enterococci load while the POTW effluent contributed more than 99% for all Runs 1aei. When impact was assessed in terms of volume (PSV), less than 0.01% of the total waterbody volume needed to come from the Raw source while 10e22% needed to be POTW effluent (remainder being background water). Both POTW and Raw sources contributed Norovirus genomes (PSp) in Runs 1a-i with median predicted contributions between 84 and 91% and 9e16%, respectively. Constraining the enterococci density to 35 CFU 100 mL1 and GI illness between 0.01 and 0.3, both fresh POTW effluent and the Raw source contributed to the total probability of illness from accidental ingestion of recreational water (Fig. 2). The x-axis in Fig. 2 presents the range of possible probabilities of illness for the fresh scenarios (Runs 1aei) with the corresponding source-specific risks on the y-axis. For example, for a fresh secondary-treated disinfected wastewater effluent and Raw source mixture with a probability of illness of 0.03 (Run 1a), the predicted median GI illness risk attributable to fresh POTW effluent was 0.026 and the median risk attributable to the Raw source was 0.004.
3.2.
Sensitivity to input parameters
The measures of impact presented above for the fresh wastewater effluent and untreated (or poorly treated) sewage mixture scenario (Run 1a) change when alternative scenarios are considered (Runs 2e6 from Table 1). The sensitivity of each of the four measures of impact is discussed below for changes in the percent of the total indicator load from a non-pathogenic source (POI) (Runs 2aeb), the age of the POTW wastewater when mixed into the waterbody (tPOTW) (Runs 3aec), the age of the Raw source when mixed into the waterbody (traw) (Runs 4aeb), the sedimentation of POTW Norovirus out of the waterbody (FPOTWp) (5aec), and sedimentation of the Raw source Norovirus and indicator out of the waterbody (Frawp, FrawI) (Runs 6aec).
Risk from Source
no inactivation, sedimentation nor additional sources of the pathogens and indicators (Tables 1 and 2). For this, Equations (3)e(10) were solved to estimate the source mixture necessary to reach a waterbody enterococci density of 35 CFU 100 mL1 and a total probability of illness from accidental ingestion of recreational water while swimming in a range around 0.03 (i.e. 0.001e0.3).
1 0.1 0.01 0.001 0
0.1
0.1
0.2
0.3
Probability of Illness
0.01 POTW Raw
0.001 0.0001 Indicator CFU
Indicator CCE
Volume
Norovirus
Fig. 1 e Predicted median source contribution of total indicator density, volume, and pathogen density for a waterbody with a total indicator density of 35 CFU 100 mLL1 enterococci and impacted by fresh poorly treated sewage (Raw) and fresh secondary-treated disinfected municipal wastewater (POTW) (Run 1a). The 25th and 75th percentile value predictions are shown as error bars.
median POTW
median Raw
25th and 75th %tile POTW
25th and 75th %tile Raw
Fig. 2 e Predicted median probability of illness from accidental ingestion while swimming in a waterbody with a total indicator density of 35 CFU 100 mLL1 enterococci and impacted by fresh poorly treated sewage (Raw) and fresh secondary-treated disinfected municipal wastewater (POTW) (Runs 1aei). The x-axis presents a range of feasible probabilities of illness for the total indicator density and sources. The median (boldface) and 25th and 75th percentile value predictions are plotted.
2677
Percent Enterococci (cfu) from Source
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 7 0 e2 6 8 0
for a total risk of 0.03 (also presented in Fig. 1). When the nonpathogenic source increased to 90% of the indicator load, the risk attributable to the Raw source decreased so that the risk was almost exclusively from POTW effluent.
100 10 1 0.1
3.2.2. Age of the POTW wastewater when mixed into the waterbody
0.01 0.001 0
20
40
60
80
100
Percent Enterococci from Non-pathogenic Source POTW
Raw
Fig. 3 e Comparison of median predicted source contribution of total indicator density (CFU) for a waterbody with 35 CFU 100 mLL1 enterococci and impacted by fresh poorly treated sewage (Raw), fresh secondary-treated disinfected municipal wastewater (POTW), and a non-pathogenic indicator source (Runs 2aeb). The percent of the total enterococci from the nonpathogenic source is shown on the x-axis. The median (boldface) and 25th and 75th percentile value predictions are plotted.
3.2.1.
Indicator load from a non-pathogenic source
A non-pathogenic source of indicator was included in the input parameters for Runs 2aeb, which were otherwise identical to Run 1a (i.e. fresh secondary-treated disinfected municipal wastewater and untreated (or poorly treated) sewage sources with no inactivation or sedimentation). The addition of a non-pathogenic indicator source decreased the predicted volume (results not shown) and the percent of total indicator load from the Raw source (Fig. 3). The sensitivity of the risk attribution for fresh POTW or Raw (PillS) in a waterbody with enterococci density of 35 CFU 100 mL1 to changes in the percent of the total indicator load from a non-pathogenic source (POI) is shown in Fig. 4. When the percentage of the total indicator load from a non-pathogenic source (POI) ¼ 0% (Run 1a), both POTW effluent and untreated sewage contribute risk
3.2.3. Age of the untreated (or poorly treated) sewage when mixed into the waterbody The age of the untreated (or poorly treated) sewage (tRaw) was included in the input parameters for Runs 4aeb, which were otherwise identical to Run 1a (i.e. fresh secondary-treated disinfected municipal wastewater with no other sources of faecal indicator, inactivation or sedimentation). When the Raw source was aged, the indicator by culture method inactivated at a faster rate than the Norovirus genomes. As a result, the predicted Raw volume slightly increased but with minimal effects on the percent of total indicator load and the total Norovirus load (results not shown). An increase in the median risk attributable to the Raw source was observed when the age of the untreated (or poorly treated) sewage source approached 10 h (Fig. 5).
1
1
0.1 0.01 0.001 0.0001 0
20
40
60
80
100
Percent Enterococci from Non-pathogenic Source
Risk from Source
Risk from Source
The age of the POTW wastewater effluent (tPOTW) was included in the input parameters for Runs 3aec, which were otherwise identical to Run 1a (i.e. fresh untreated (or poorly treated) sewage sources with no other sources of faecal indicators, inactivation or sedimentation). The age of the POTW effluent (up to 72 h) had little effect on the contribution of POTW and Raw to the total indicator (both CFU and CCE), volume, pathogen, and risk. This lack of change in indicator by culture method, volume, pathogen, and risk is due to the minimal inactivation of Norovrius in the environment and already low density of enterococci CFU in the POTW effluent. Although there was removal of the qPCR enterococci CCE in the POTW effluent, the overall contribution of the enterococci CCE from the POTW effluent was still large compared to the contribution from the Raw source using the limited available data available for the densities and persistence of enterococci by the qPCR method.
0.1 0.01 0.001 0.0001 0
median POTW
median Raw
25th and 75th %tile POTW
25th and 75th %tile Raw
Fig. 4 e Comparison of median predicted probability of illness for a waterbody with 35 CFU 100 mLL1 enterococci and impacted by fresh poorly treated sewage (Raw), fresh secondary-treated disinfected municipal wastewater (POTW), and a non-pathogenic indicator source (Runs 2aeb). The percent of the total enterococci from the nonpathogenic source is shown on the x-axis. The median (boldface) and 25th and 75th percentile value predictions are plotted.
2
4
6
8
10
Age of Raw (h)
median POTW
median Raw
25th and 75th %tile POTW
25th and 75th %tile Raw
Fig. 5 e Comparison of median predicted probability of illness for a waterbody with 35 CFU 100 mLL1 enterococci and impacted by aged poorly treated sewage (Raw) and fresh secondary disinfected municipal wastewater (POTW) (Runs 4aeb). The age of the poorly treated sewage is shown on the x-axis. The median (boldface) and 25th and 75th percentile value predictions are plotted.
2678
3.2.4.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 7 0 e2 6 8 0
Sedimentation (removal) of POTW Norovirus genomes
Norovirus sedimentation out of the secondary-treated disinfected municipal wastewater prior to mixing (FPOTWp) was included in the input parameters for Runs 5aec, which were otherwise identical to Run 1a (i.e. fresh secondary-treated disinfected municipal wastewater and untreated sewage sources with no other sources of faecal indicator or inactivation). When less than 99% of the Norovirus was removed (Runs 5aeb), there were minimal effects on the Run 1 results (not shown). When the removal of POTW Norovirus was 99%, 17% of the Monte Carlo runs resulted in solutions, indicating that it was difficult to construct a physical scenario within the bounds of the input parameters when the removal of Norovirus genomes from the POTW effluent was 99% and the waterbody risk was set to 0.03.
3.2.5. Sedimentation (removal) of poorly treated sewage Norovirus genomes and indicator Norovirus and enterococci sedimentation from the Raw source (Frawp, FrawI) suspension was included in the input parameters for Runs 6aec, which were otherwise identical to Run 1a (i.e. fresh secondary-treated disinfected municipal wastewater and untreated (or poorly treated) sewage sources with no other sources of faecal indicator or inactivation). When 50% of the Norovirus genomes and enterococci CFU were removed (Runs 5aeb), there were minimal effects on the Run 1 results (not shown). When the removal of Norovirus genomes and enterococci was 90% or greater, less than 20% of the Monte Carlo runs resulted in solutions, indicating that it was difficult to construct a physical scenario within the bounds of the input parameters. When the risk was lowered to 0.01 and 0.001, the percentage of Monte Carlo runs resulting in solutions remained below 20%.
4.
Discussion
4.1.
Assessment of human source contribution
To address the large natural variability in the QMRA input parameters and uncertainties in using Norovirus as the reference pathogen, the results were examined for general trends across multiple scenarios with clear, large differences in the relative importance of each source. Accounting for the age of the effluent, possible sedimentation of microorganisms, and other non-pathogenic sources of indicator, predicted results for a human-impacted waterbody at 35 CFU 100 mL1 enterococci showed that: 1. Enterococci density assayed by culture was dominated by the untreated (or poorly treated) sewage (Raw) or possible other non-pathogenic sources; 2. Enterococcus density assayed by qPCR was dominated by the secondary-treated disinfected municipal wastewater (POTW) effluent or possible other non-pathogenic sources; 3. Effluent volume was dominated by the secondary-treated disinfected municipal wastewater effluent; 4. Norovirus genome density and risk from accidental ingestion of recreational waters from swimming were dominated by POTW effluent, but also influenced by Raw sources
when there were no other dominant sources of enterococci; and 5. Norovirus genome density and risk from accidental ingestion of recreational waters from swimming were dominated by POTW effluent when there were dominant contributions of enterococci from additional non-pathogenic sources. Clearly, the way in which one measures source impact is important. If looking at faecal indicator contributions only by a culture method, the predicted contribution from the Raw source dominates that from secondary-treated disinfected municipal wastewater effluent; however, the POTW wastewater effluent played an important role in the predicted GI illness risk at human-impacted beaches. Interestingly, when 90% of the waterbody enterococci as measured by culture were contributed from other non-pathogenic sources, it was possible to predict a GI illness risk of 0.03 with the risk mainly attributable to POTW effluent.
4.2.
Faecal indicator insights
Although the primary objective of this work was to assess the importance of different human sources of faecal contamination at human-impacted beaches, the results confirm existing beliefs about the use of faecal indicator bacteria as a measure of human health risk at recreational waters. First, the results confirm that a range of risks may be associated with a level of indicator in waterbodies impacted by human faecal sources. Second, the predicted contribution of enterococci assayed by culture from secondary-treated disinfected wastewater effluent is extremely small compared to untreated sewage and other potential indicator sources. This is particularly concerning since a large portion of the total risk was predicted to be attributed to the POTW effluent, even with 90% of the indicator coming from other, non-pathogenic sources. This analysis illustrates that the source contributing the majority of risk in a mixture may be overlooked when only faecal indicators assayed by a culture method are assessed. Third, the predicted contribution of enterococci by a qPCR method from secondarytreated disinfected wastewater effluent is large compared to the Raw source using the limited available data. Based on the conditions examined for this paper, there is strong evidence that the use of a calibrated qPCR total enterococci indicator, compared to a culture-based assay, could provide a better measure of potential pathogen (Norovirus) risk, since both the qPCR waterbody density and total Norovirus risk were predicted to be largely attributed to the POTW effluent.
4.3.
Limitations
The QMRA approach used to estimate the source mixtures has a number of limitations. The results of this analysis may not be applicable to every human-impacted site or during events. For example, sites with a non-human pathogenic source of indicator and pathogens, such as cattle, may have different total risks and source apportionments (Soller et al., 2010b; U.S. EPA, 2010). Also, episodic events at human-impacted beaches such as rainfall may also introduce different sources and apportionments of indicators or pathogens. Furthermore, the
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 7 0 e2 6 8 0
approach assumes distributions of indicator and pathogen densities for secondary-treated disinfected effluents and untreated sewage sources. There may be human sources with indicator and pathogen densities different from the assumed distributions. Norovirus genome densities in POTW effluent may vary with type of treatment, season (Haramoto et al., 2006; Katayama et al., 2008), and population density contributing to the effluent and geographic location. Here, wastewater Norovirus densities were restricted to the swimming season. If Norovirus densities were in actuality much larger (say during an outbreak), less predicted POTW volume would be necessary to achieve a GI risk of 0.03. Furthermore, the geographic location of the waterbody and its contributing sources may differently affect the inactivation of the indicator and pathogen. For purposes of this paper, enterococci inactivation was estimated for temperate location seawaters during summer ambient conditions. The characteristics of the Raw source may not be applicable to all poorly treated human wastes, such as human shedders. Human shedders are very likely to show more spatial and temporal heterogeneity in indicator and pathogen densities, as seen for on-site wastewater systems (Charles et al., 2003). As the number of shedders decreases, the assumptions made here of a well mixed, relatively homogeneous Raw source become less applicable. Due to these spatial and temporal variations in indicator and pathogen densities in many surface waters (Payment and Locas, 2010), the exact source contributions may vary for any single point of exposure. The use of Norovirus as the reference pathogen, although a strong selection for human faecal contamination, has a number of outstanding limitations. The available dose-response relationship for Norovirus infection is based on a population of healthy, young adults. Therefore, what is reported here does not specifically account for susceptible sub-populations. The approach assumes that everyone is susceptible to Norovirus with an uncertain dose-response relationship for infection and probability of illness conditional on infection. This seems reasonable if we believe there may be many types of infectious Norovirus present in the effluent. In addition, the infectivity of the secondary disinfected wastewater effluent is assumed to be similar to that of the inoculum used in estimating the doseresponse relationship (Teunis et al., 2008). Lastly, the predicted Norovirus doses are small (sometimes 1 genome) compared to those administered when estimating the dose-response relationship. Possible alternative candidates for a suitable reference pathogen for human-impacted waters might include other enteric viruses (Bofill-Mas et al., 2010) or Giardia (Graczyk et al., 2010) based on previous study results which estimated the relative contribution of specific bacterial, protozoan, and viral pathogens to the source-specific risks (Soller et al., 2010a). If, for example, Giardia was selected as the reference pathogen in this study, the overall results would be very similar based on the relatively low inactivation of Giardia in the environment for up to 72 h and the high density in POTW effluents (unless a membrane filtration process or UV disinfection is utilized). The results for any sewage-associated bacterial species would be very different since active enteric bacterial species are generally in low densities in disinfected POTW effluents and can be rapidly inactivated in the environment (Noble et al., 2004).
5.
2679
Conclusion
To assess the relative importance of different human sources at human-impacted beaches, QMRA was used with Norovirus as the reference pathogen to estimate possible mixtures of secondary disinfected municipal wastewater effluent and untreated (or poorly treated) sewage needed to provide a waterbody enterococci density of 35 CFU 100 mL1 and total probability of GI illness from accidental ingestion of water in the range of 0.01e0.3. The contribution to the total waterbody volume, indicator density (both culture and qPCR methods), pathogen density, and risk were estimated for each source for a number of scenarios that accounted for pathogen and indicator inactivation based on the age of the effluent, possible sedimentation of microorganisms, and other non-pathogenic sources of indicator. The major findings from a synthesis of alternative scenarios with a waterbody enterococci density of 35 CFU 100 mL1 were: 1. A broad range of probabilities of illness were possible for a waterbody impacted by untreated sewage and secondarytreated disinfected municipal wastewater effluent at the water quality limit of 35 CFU 100 mL1 enterococci; 2. By volume, secondary-treated disinfected municipal wastewater would need to be the major contaminant, but alone is unlikely to reach the water quality limit; 3. Enterococci assayed by culture in the waterbody could largely come from untreated sewage or possibly other nonpathogenic sources; 4. Enterococci estimated by qPCR in the waterbody could predominantly be contributed by secondary-treated disinfected (POTW) effluent or possibly other non-pathogenic sources; and 5. Norovirus genome density and significant GI risk from accidental ingestion of recreational waters is most likely to result from a combination of untreated sewage and secondarytreated disinfected municipal wastewater effluent.
Acknowledgements The research described in this article was funded, in part, by the U.S. EPA Office of Water, Office of Science and Technology under contract #EP-C-07-036 to Clancy Environmental Consulting, Inc. This work has been subject to formal Agency review and does not necessarily reflect the views of the Agency, and no official endorsement should be inferred. The authors gratefully acknowledge the valuable contributions of Dr Timothy Bartrand, John Ravenscroft, and Dr Richard Haugland.
references
Bae, J., Schwab, K.J., 2008. Evaluation of murine Norovirus, feline calicivirus, poliovirus, and MS2 as surrogates for human norovirus in a model of viral persistence in surface water and groundwater. Applied and Environmental Microbiology 74 (2), 477e484.
2680
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 7 0 e2 6 8 0
Bae, S., Wuertz, S., 2009. Rapid decay of host-specific fecal Bacteroidales cells in seawater as measured by quantitative PCR with propidium monoazide. Water Research 43 (19), 4850e4859. Bofill-Mas, S., Rodriguez-Manzano, J., Calgua, B., Carratala, A., Girones, R., 2010. Newly described human polyomaviruses Merkel cell, KI and WU are present in urban sewage and may represent potential environmental contaminants. Virology Journal 7, 141. Cabelli, V.J., Dufour, A.P., McCabe, L.J., Levin, M.A., 1982. Swimming-associated gastroenteritis and water quality. American Journal of Epidemiology 115, 606e616. Charles, K., Ashbolt, N., Ferguson, C., Roser, D., McGuinness, R., Deere, D., 2003. Centralised versus decentralised sewage systems: a comparison of pathogen and nutrient loads released into Sydney’s drinking water catchments. Water Science and Technology 48 (11e12), 53e60. de Roda Husman, A.M., Lodder, W.J., Rutjes, S.A., Schijven, J.F., Teunis, P.F.M., 2009. Long-term inactivation study of three enteroviruses in artificial surface and groundwaters, using PCR and cell culture. Applied and Environmental Microbiology 75 (4), 1050e1057. Dufour, A., 1984. Health Effects Criteria for Fresh Recreational Waters. US EPA, Cincinnati, OH. Graczyk, T., Sunderland, D., Awantang, G., Mashinski, Y., Lucy, F., Graczyk, Z., Chomicz, L., Breysse, P., 2010. Relationships among bather density, levels of human waterborne pathogens, and fecal coliform counts in marine recreational beach water. Parasitology Research 103 (5), 1103e1108. Haramoto, E., Katayama, H., Oguma, K., Yamashita, H., Tajima, A., Nakajima, H., Ohgaki, S., 2006. Seasonal profiles of human Noroviruses and indicator bacteria in a wastewater treatment plant in Tokyo, Japan. Water Science & Technology 54 (11e12), 301e308. He, J., Jiang, S., 2005. Quantification of enterococci and human adenoviruses in environmental samples by real-time PCR. Applied and Environmental Microbiology 71 (5), 2250e2255. Katayama, H., Haramoto, E., Oguma, K., Yamashita, H., Tajima, A., Nakajima, H., Ohgaki, S., 2008. One-year monthly quantitative survey of noroviruses, enteroviruses, and adenoviruses in wastewater collected from six plants in Japan. Water Research 42 (6e7), 1441e1448. Lemarchand, K., Lebaron, P., 2003. Occurrence of Salmonella spp. and Cryptosporidium spp. in a French coastal watershed: relationship with fecal indicators. FEMS Microbiology Letters 218 (1), 203e209. Noble, R., Lee, I., Schiff, K., 2004. Inactivation of indicator microorganisms from various sources of faecal contamination in seawater and freshwater. Journal of Applied Microbiology 96 (3), 464e472. Payment, P., Locas, A., 2010. Pathogens in water: value and limits of correlation with microbial indicators. Ground Water 49 (1), 4e11. Pru¨ss, A., 1998. Review of epidemiological studies on health effects from exposure to recreational water. International Journal of Epidemiology 27 (1), 1e9. R Development Core Team, 2009. R: a Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria. Rose, J.B., Nowlin, H., Farrah, S.R., Harwood, V., Levine, A., Lukasik, J., Mendendez, P., Scott, T.M., 2004. Reduction of Pathogens, Indicator Bacteria, and Alternative Indicators by Wastewater Treatment and Reclamation Processes. Water Environment Research Foundation. Report 00-PUM-2T.
Schoen, M.E., Ashbolt, N.J., 2010. Assessing pathogen risk to swimmers at non-sewage impacted recreational beaches. Environmental Science & Technology 44 (7), 2286e2291. Sinton, L.W., Finlay, R.K., Hnnah, D.J., 1998. Distinguishing human from animal faecal contamination in water: a review. New Zealand Journal of Marine and Freshwater Research 32, 323e348. Skraber, S., Schijven, J., Italiaander, R., de Roda Husman, A.M., 2009. Accumulation of enteric bacteriophage in fresh water sediments. Journal of Water and Health 7 (3), 372e379. Soller, J.A., Bartrand, T., Ashbolt, N.J., Ravenscroft, J., Wade, T.J., 2010a. Estimating the primary etiologic agents in recreational freshwaters impacted by human sources of faecal contamination. Water Research 4 (16), 4736e4747. Soller, J.A., Schoen, M.E., Bartrand, T., Ravenscroft, J.E., Ashbolt, N.J., 2010b. Estimated human health risks from exposure to recreational waters impacted by human and nonhuman sources of faecal contamination. Water Research 44 (16), 4674e4691. Teunis, P.F.M., Moe, C.L., Liu, P., Miller, S.E., Lindesmith, L., Baric, R.S., Pendu, J.L., Calderon, R.L., 2008. Norwalk virus: how infectious is it? Journal of Medical Virology 80 (8), 1468e1476. Tschobanoglous, G., Burton, F., Stensel, H.D., 2003. Wastewater Engineering: Treatment and Reuse. McGraw-Hill, New York. U.S. EPA, 1986. Ambient Water Quality Criteria for Bacteria e 1986 EPA 440/5-84-002. Office of Water Regulations and Standards Division, U.S. Environmental Protection Agency, Washington, DC. U.S. EPA, 2010. EPA 822-R-10-005, Quantitative Microbial Risk Assessment to Estimate Illness in Freshwater Impacted by Agricultural Animal Sources of Fecal Contamination. Varma, M., Field, R., Stinson, M., Rukovets, B., Wymer, L., Haugland, R., 2009. Quantitative real-time PCR analysis of total and propidium monoazide-resistant fecal indicator bacteria in wastewater. Water Research 43 (19), 4790e4801. Wade, T.J., Pai, N., Eisenberg, J.N.S., Colford, J.M., 2003. Do US Environmental Protection Agency water quality guidelines for recreational waters prevent gastrointestinal illness? A systematic review and meta-analysis. Environmental Health Perspectives 111 (8), 1102e1109. 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., Calderon, R.L., Brenner, K.P., Sams, E., Beach, M., Haugland, R., Wymer, L., Dufour, A.P., 2008. High sensitivity of children to swimming-associated gastrointestinal illness: results using a rapid assay of recreational water quality. Epidemiology 19 (3), 375e383. Walters, S.P., Yamahara, K.M., Boehm, A.B., 2009. Persistence of nucleic acid markers of health-relevant organisms in seawater microcosms: implications for their use in assessing risk in recreational waters. Water Research 43 (19), 4929e4939. Wang, D., Silkie, S.S., Nelson, K.L., Wuertz, S., 2010. Estimating true human and animal host source contribution in quantitative microbial source tracking using the Monte Carlo method. Water Research 44 (16), 4760e4775. Zmirou, D., Pena, L., Ledrans, M., Letertre, A., 2003. Risks associated with the microbiological quality of bodies of fresh and marine water used for recreational purposes: summary estimates based on published epidemiological studies. Archives of Environmental Health 58 (11), 703e711.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 8 1 e2 6 9 0
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Trichloramine in swimming pools e Formation and mass transfer Christina Schmalz a, Fritz H. Frimmel b, Christian Zwiener a,* a
Environmental Analytical Chemistry, Center for Applied Geoscience (ZAG), Eberhard Karls University Tuebingen, 72076 Tuebingen, Germany b Water Chemistry, Engler-Bunte-Institut, University Karlsruhe (TH), Engler-Bunte-Ring 1, 76131 Karlsruhe, Germany
article info
abstract
Article history:
Trichloramine is a volatile, irritant compound of penetrating odor, which is found as
Received 13 December 2010
a disinfection by-product in the air of chlorinated indoor swimming pools from reactions of
Received in revised form
nitrogenous compounds with chlorine. Acid amides, especially urea, ammonium ions and
9 February 2011
a-amino acids have been found as most efficient trichloramine precursors at acidic and
Accepted 19 February 2011
neutral pH. For urea a relative NCl3 formation of 96% at pH 2.5 and 76% at pH 7.1 was
Available online 1 March 2011
determined. Even under sub-stoichiometric molar ratios of Cl/N the formation of NCl3 is favored over mono and dichlorinated products. However, the reaction kinetics of urea with
Keywords:
chlorine is slow under conditions relevant for swimming pools. Also the mass transfer of
Trichloramine
NCl3 from water to the gas phase which was calculated by the Deacon’s boundary layer
Swimming pool
model could be shown as a relatively slow process. Mass transfer would take 20 h or 5.8 d
Chlorination
for a rough or a quiescent surface of the water, respectively. This is much more than
Mass transfer
a typical turnover rate of 6e8 h of a treatment cycle of a 25 m swimming pool. Therefore
Urea
processes to remove NCl3 and its precursors can help to minimize the exposure of bathers. ª 2011 Elsevier Ltd. All rights reserved.
Acid amide
1.
Introduction
For hygienic safety and protection of bathers against infectious diseases the disinfection of swimming pool water (SPW) is absolutely essential. Chlorine dosed in gaseous form or as sodium hypochlorite is the most common method for disinfection of SPW. To achieve a sufficient disinfection capacity the concentration of free chlorine should be kept in the range between 0.3 and 0.6 mg/L according to the German Pool Water Standard DIN 19643. In the United States, Australia and many other European countries concentrations up to 3 mg/L of free chlorine are recommended. Under operating conditions SPW is continuously loaded with organic carbon and particularly with nitrogen compounds. Each bather contributes with considerable amounts of body
fluids, hair, skin particles, microorganisms and cosmetics. Components of the bather load are only partly removed in the treatment and therefore cycle many times in the system “pool water e treatment e pool water”. In each passage through the treatment the water is subsequently dosed with chlorine, which can react with water constituents. Thus the compounds are partly degraded and transformed to disinfection by-products (DBPs). The reaction between components of the bathers load and the dosed chlorine to toxic DBPs cannot be completely avoided but has to be minimized. Besides trihalomethanes (THM) e the best known DBPs e many other chlorinated DBPs have been analyzed in SPW and drinking water, for example halogenated acetic acids, aldehydes or acetonitriles (Zwiener et al., 2007; Richardson et al., 2007). Chlorinated nitrogenous reaction products are generally less known and are routinely
* Corresponding author. Tel.: þ49 7071 2974702; fax: þ49 7071 295059. E-mail address:
[email protected] (C. Zwiener). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.02.024
2682
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 8 1 e2 6 9 0
monitored only by the not well defined sum parameter combined chlorine. Recently published epidemiological studies on respiratory functions and asthma give evidence for adverse health effects from attending chlorinated swimming pools. Reports show an increase in lung epithelium permeability (Bernard et al., 2003), respiratory complaints (Levesque et al., 2006) or developing asthma (Nickmilder and Bernard, 2007). Mostly trichloramine is suggested to cause eye and upper respiratory tract irritation (Massin et al., 1998), biomarker changes in the lung and development of asthma (Bernard et al., 2006). Hery et al. (1995) and Gagnaire et al. (1994) reported on irritating complaints of eye and throat among pool attendants and irritating effects on mice at trichloramine concentration levels higher than 0.5 mg/m3 in the air. The proposed INRS (French Institute for Occupational Health and Safety) guideline value of 0.5 mg/m3 for NCl3 for indoor pool air is based on these investigations. Trichloramine itself is an irritant and in higher concentrations an explosive compound of penetrating odor. It is slightly soluble in water (0.025 mol/L, 25 C, pH 1e10) and its octanol-water partition coefficient log D is 2.64 (25 C, pH 1e10) (SciFinder Scholar, 2006). It is four times more volatile than chloroform and partitions to the air of swimming pools (Henry coefficient, NCl3 10 atm L mol1 (25 C); Holzwarth et al., 1984; Sander, 1999). Air measurements of NCl3 in indoor pool settings reveal concentrations between 0.1 and 18.8 mg/m3 (Hery et al., 1995; Stottmeister and Voigt, 2006). Air concentrations of trichloramine strongly depend on air circulation and on water agitation by bathers and attractions (Zirbs, 2008). However, they are not correlated with the concentration of the indicator parameter combined chlorine in the SPW (Stottmeister and Voigt, 2006), since air ventilation is the predominant parameter. Furthermore, the contribution of NCl3 and dichloramine to the total combined chlorine is low and typically in the range between 15 and 20% (Weaver et al., 2009). Further monitoring programs for trichloramine concentrations in air of different pool settings have been started meanwhile in Germany by Bavarian and Federal Consumer and Occupational Health and Safety Agencies (LGL-Bayern, BGIA). Up to now no systematic data are available which show the relation between the occurrence of precursor substances, the reaction conditions and the resulting trichloramine exposure in water and air of swimming pool settings. N-chlorination reactions are described by an initial electrophilic attack of HOCl to the nucleophilic nitrogen. For ammonium ions decreasing chlorine reactivity with increasing numbers of substituted chlorine atoms were reported. At pH 7 second order reaction rates of ammonium ions with HOCl of 1.3 104 M1 s1 and monochloramine with HOCl of 2.7 102 M1 s1 were given by Deborde and von Gunten (2008) (calculated from data of Qiang and Adams, 2004; Morris and Isaac, 1983). The reaction kinetics of trichloramine formation is more complex. Here a general base catalyzed mechanism with third order rate constants was proposed (Hand and Margerum, 1983). In water the two species of ammonia (NH3 and NH4þ) are present, but chlorine reactivity of NH4þ species was reported to be negligible (Qiang and Adams, 2004). Under acidic conditions ammonium ions are the prevailing species, here
the first chlorination step is slowed down by a factor of 106 compared to the uncharged ammonia molecule (Saguinsin and Morris, 1975). Ammonium ions, urea and to some extent creatinine and amino acids have been described as precursors for trichloramine formation in chlorination processes (Li and Blatchley, 2007; Shang and Blatchley, 1999; Na and Olson, 2004; Blatchley and Cheng, 2010). The kinetics of chlorine degradation in the presence of urea at neutral and alkaline pH conditions have been first described by Fuchs (1962). Due to the slow reaction he also showed a continuous increase of the urea concentration in the water within a period of three months up to a concentration of 24 mg/L for a swimming pool where only water was added to balance evaporation (sand filtration, 0.1e0.3 mg/L Cl2, pH 8e10). The objective of this study was to determine the major reactions and processes involved in the formation of trichloramine in swimming pools and the resulting exposure of bathers. For that purpose different nitrogenous compounds and the major components of urine and sweat have been tested for their relative NCl3 formation. Reaction parameters, reaction kinetics and the mass transfer of NCl3 from water to air were considered. The results should set a sound basis to apply suitable measures to minimize the exposure of the bathers and bath attendants by NCl3.
2.
Materials and methods
2.1.
Experimental procedures
All chemicals, phosphate salts and nitrogen compounds used in this study were p.a. reagent grade and mostly obtained from Merck (Darmstadt, Germany), except sodium hypochlorite solution (w90 g/L Cl2, daily controlled by DPD method) was from Roth (Karlsruhe, Germany). Lysine, arginine, acetamide, glutamic acid and b-alanine were from SigmaeAldrich (Steinheim, Germany). Free and combined chlorine were analyzed with the photometric DPD method (Spectroquant chlorine cell test, according to EN ISO 7393-2:2000), urea with the enzymatic and colorimetric indophenol blue method (Microquant urea-test; both from Merck, Darmstadt, Germany). UV-spectra were recorded on a Cary 50 spectra-photometer (Varian, Darmstadt, Germany). GC-MS instruments and operating parameters are described in the supporting information (Figure S6). All chlorination experiments were performed in glassware, which was soaked before use in sodium hypochlorite solution for 24 h to remove the chlorine demand. NCl3 formation experiments were carried out in aqueous phosphate buffer solutions (pH 2.5e7.7) with 1.0 102 mol/L nitrogen compound (as N). Sodium hypochlorite solutions were added at molar chlorine to nitrogen ratios between 0.15 and 5. The phosphate buffer concentration in the reaction mixture was 0.25 mol/L. The pH varied at maximum by 0.2 at pH 5.9, pH 6.3, pH 7.1 and pH 7.7, by 0.3 at pH 2.5 and by 0.5 at pH 4.0. The reaction time was 30 min to compare the reactivity of all N-compounds under the same conditions. After 30 min reaction was assumed to be completed for all compounds studied at the concentrations applied. Reaction products were analyzed by UV-spectroscopy (l ¼ 190e500 nm) after extraction to hexane. To check the effects of the hexane
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 8 1 e2 6 9 0
extraction on the yield of NCl3 formation, comparative measurements were done with urea chlorine mixtures. NCl3 yields were measured after extraction with hexane and directly in the aqueous phase. Similar results were achieved with both methods (detailed information is given in Supporting information paragraph 3). Additionally, the hexane phases were analyzed after washing steps with deionized water by UV-spectroscopy and GC-MS. UV-spectra formed by overlaid spectra of further chlorination products were deconvoluted with Gaussian functions for each compound (paragraph 1 Supporting information; Figures S1eS4). Quantitative data were calculated from the measured and deconvoluted UVspectra with the UV-absorption coefficients given in the Supporting information in Table S1. The absorption coefficients of NCl3 solutions in hexane were obtained from measurements of the nitrogen content (N-Analyzer, chemiluminesence of NO with O3, TN 05, Abimed, Langenfeld, Germany) and of the chloride content after reduction of NCl3 with As2O3 in a sodium carbonate solution (Hery et al., 1995). The chloride concentration was analyzed after cation exchange (Dowex 50 WX 8) with ion chromatography (IC 790, Metrohm, Herisau, Switzerland; column: Metrosep Anion Dual 2, 75 mm 4.6 mm, 6 mm particles; eluent 5 mM phthalic acid with 2% acetonitrile at pH 4.5). Chlorination experiments with initial urea concentrations of 3.33 105 mol/L (nitrogen concentration c(N) ¼ 6.67 105 mol/L) and 3.33 104 mol/L were performed in stirred volumetric flasks. The products were extracted with 5 mL and 10 mL hexane to achieve concentration factors of 10e100. The hexane phases were analyzed after reaction times of 1 h, 3 h, 24 h, 48 h, 72 h and 96 h by UV-spectroscopy. Maximum total NCl3 yields for experiments with urea concentrations of 3.33 104 mol/L were achieved after 24 h and with urea concentrations of 3.33 105 mol/L after up to 72 h. The experiments for the reaction kinetics were performed in stirred 2 L-Erlenmeyer flasks, which were wrapped in aluminum foil to exclude light. The temperature was controlled by an Ikatron ETS D-4 fuzzy system (IKA-Works, Staufen, Germany). The initial chlorine concentration was 1.13 105 mol/L and the urea concentration was 3.33 105 mol/L (molar Cl/N ratio ¼ 0.17). A pH of 6.9 0.1 was adjusted with phosphate buffer (c ¼ 1.7 104 mol/L). Chlorine was measured with the DPD method and urea with the enzymatic and colorimetric indophenol blue method.
2.2.
2683
controlled by the aqueous phase (Schwarzenbach et al., 2003). So the gas phase regime could be neglected and the flux of NCl3 to the air was described by a one film model with liquid mass transfer coefficients (vNCl3 w ) according to (Eq. (3)) (Schwarzenbach et al., 1979). To calculate the liquid mass transfer coefficient for NCl3 (vNCl3 w ) the Deacon’s boundary layer model (Deacon, 1977; Schwarzenbach et al., 2003) was used. Different water surface characteristics were considered on the basis of a direct linear correlation (Mackay and Yeun, 1983) with water evaporation rates for unused swimmer pools, used swimmer pools and whirl pools given in the German Guideline VDI 2089-1. Additional information is given in the paragraph 6 supporting information. Pia Fia=w ¼ via=w ciw KHi
(1)
1 1 1 ¼ þ via=w viw via Kia=w
(2)
FNCl3 w ¼ vNCl3 w cNCl3 w
(3)
ciw ¼ concentration of the substance i in the water (mol L1) Pia ¼ partial pressure for the substance i in the air above (bar) viw ¼ liquid phase mass transfer coefficient for substance i (cm s1) via ¼ air phase mass transfer coefficient for substance i (cm s1) Kia/w ¼ dimensionless Henry coefficient;Kia=w ¼ KHi =RT; KHi ¼ Henry coefficient in L bar mol1; R ¼ 0,08314 L bar K1 mol1; T ¼ temperature in Kelvin The Henry constant for NCl3 at the temperature of 28 C was calculated according to Sander (1999) with the enthalpy of solution DsolnH/R ¼ 4100 K.
2.2.1.
Safety advice
NCl3 is an explosive, unstable compound that can explode in form of its pure liquid above its boiling point of 71 C or in concentrated solutions when catalyzed by light or other catalysts. Concentrations higher than 2 g/L in water (Savickas et al., 1989; Schlessinger, 1966) or isolation of the pure compound have to be imperatively avoided. Never isolate or concentrate NCl3. Only handle diluted gas streams or solutions of NCl3 behind a shield and in a well-ventilated hood.
Calculation of trichloramine concentrations in air
The air concentrations of NCl3 were modeled for an indoor pool with a water volume Vw ¼ 960 m3, a water temperature T ¼ 28 C, a water surface area A ¼ 312.5 m2 and an air volume of the pool hall Va ¼ 5760 m3. The general Fick’s first law to describe mass fluxes (Eq. (1)) of volatile compounds Fia/w from the water to gas phase was reduced under assumption of a low NCl3 partial pressure (PNCl3 a ) in the air above the water surface compared to the concentration in the water. The total mass transfer coefficient (via/w) (Eq. (2)) considers both phases of the boundary e air and water phase. For substances with Henry coefficients much higher than 0.025 L bar mol1 which is given for NCl3 with KH;NCl3 ¼ 11.6 L bar mol1 (T ¼ 301 K) the mass transfer is
3.
Results and discussion
3.1.
Screening for NCl3 precursor compounds
Organic precursor compounds with different functional nitrogen containing groups and different forms of bound nitrogen have been investigated for their potential to form trichloramine. The precursors were chosen with the aim to include the major components of urine and sweat and to get a systematic structural variation of relevant nitrogenous compounds (Fig. 1). Urea is considered as the most important N-compound followed by ammonia and creatinine. The functional groups selected were acid amides, amino acids and alkyl
2684
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 8 1 e2 6 9 0
Fig. 1 e Nitrogenous compounds with different functional groups and different forms of bound nitrogen.
amines as well as compounds with guanidine type structures (arginine, creatinine). The results of the relative trichloramine formation after chlorination at a molar Cl/N ratio of 5 at varied pH are shown in Table 1. The compounds were arranged according to decreasing NCl3 formation at pH 7.1. Generally, increasing NCl3 yields were found with decreasing pH values. This pH effect can be rationalized on the basis of the pH dependent equilibrium reaction of both free chlorine and the nitrogen compound. HOCl is the main reactive chlorine species in aqueous solution. As expected
considerable lower NCl3 yields were found at pH 7.7 where OCl are the main species. The other effect is that low pH values result in protonation of the bound nitrogen, which slows down the reaction of the first chlorine attack and thus favors the reaction of the second and third chlorine attack. Data for the apparent rate constants kapp of chlorination reactions can be deduced from the literature (Deborde and von Gunten, 2008). kapp for NH3 at pH 7 is 1.3 104 M1 s1 and much lower at pH 4 (1.2 102 M1 s1). Whereas for monochloramine kapp is 1.2 102 M1 s1 at pH 7 and
2685
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 8 1 e2 6 9 0
Table 1 e Relative NCl3 formation of different nitrogenous substances. Reaction of 10 mmol/L nitrogenous compound (calculated as N) with chlorine at a molar Cl/N ratio of 5 in buffered solution. Names and numbers in italics indicate that the spectra had to be deconvoluted by Gaussian functions (cf. to Supporting information Figures S1eS4) (n.d. [ not detectable <0.1%). pH value 2.5 0.3
4.0 0.5
5.9 0.2
6.3 0.1
6.7 0.1
7.1 0.1
7.7 0.2
Relative NCl3-formation Molar ratio c(NCl3)/c(N) 100 (%) ( standard deviation) Urea Ammoniumchloride Formamide Glycine Histidine Asparagine Acetamide Serine Creatinine Glutamine Arginine Glutaminc acid Alanine b-Alanine Lysine Proline Propylamine Methylamine
96.0 5.4 98.6 5.6 91.3 0.6 22.0 2.1 31.6 0.7 62.5 0.8 64.7 5.0 73.5 4.7 4.9 51.6 0.2 28.5 44.3 3.1 79.8 2.8 6.2 0.5 32.6 n.d. n.d. n.d.
e 89.3 1.3 36.2 1.8 49.3 2.0 38.8 1.6 51.3 3.3 77.4 0.3 61.0 3.4 14.2 43.8 2.5 28.5 30.0 2.3 51.0 3.1 0.8 0.1 e n.d. n.d. n.d.
94.7 3.2 64.5 3.7 52.5 1.1 48.0 1.8 34.9 0.8 42.1 1.9 39.4 0.2 15.1 0.4 34.6 13.3 1.0 15.7 19.3 0.5 5.4 0.2 0.8 0.1 n.d. n.d. n.d. n.d.
1.5 102 M1 s1 at pH 4. Also the decomposition of NCl3 is described by a base catalyzed reaction or by a reaction with NHCl2. However, NHCl2 played only a minor role in our experiments and was not considered in detail. Therefore, increasing pH values result in increasing NCl3 decay (Kumar et al., 1987). Decay products are nitrogen and at basic conditions nitrate (Jafvert and Valentine, 1992). Maximum NCl3 yields were found for glycine, histidine and acetamide at pH 4 and for creatinine at pH 5.9. Ammonia, urea, formamide and the a-amino acids a-alanine and serine revealed as the most efficient NCl3 precursors at pH 2.5. However, at pH 7.1 the ranking with decreasing NCl3 formation is urea, ammonia, formamide, glycine, and histidine. Important prerequisites for an efficient NCl3 formation seem to be both the amine or amide function, which guarantees efficient chlorination, and the polarization of the CeN bond for facilitated cleavage of the chlorinated nitrogen moieties. This can be demonstrated nicely by the drop of the NCl3 formation yield going from an a-amino acid (a-alanine) to a b-amino acid (b-alanine). Asparagine and glutamine having both a-amino acid and acid amide functions also showed high NCl3 yields. Nitrogen bound in other forms like imidazole or guanidine like structures does not seem to contribute considerably to the NCl3 formation like the data reveal for histidine, arginine or creatinine at pH 2.5. The NCl3 yield found for histidine and arginine can be attributed mostly to the a-amino acid function. However, at increased pH the guanidine structure can also contribute to the NCl3 formation as the data show for creatinine at pH 5.9 (34.6%) and pH 6.3 (26.1%). For compounds with a single amine group without further polarization of the CeN bond no NCl3 formation could be observed. Methylamine, propylamine (pH 2.5e7.7) and the side chain of lysine (pH 5.9e7.7) almost quantitatively react to
90.2 1.3 60.0 1.4 62.0 0.2 34.5 0.3 35.4 0.5 39.6 1.4 32.2 3.5 16.3 0.3 26.1 13.6 0.6 16.6 13.2 2.2 4.6 0.2 0.9 0.1 n.d. n.d. n.d. n.d.
86.1 0.2 55.0 2.4 59.3 3.6 30.8 2.9 32.8 2.0 30.4 5.0 27.8 3.2 17.7 0.7 17.3 12.2 0.7 11.4 9.2 0.4 3.7 0.4 0.9 0.2 n.d. n.d. n.d. n.d.
75.8 4.3 37.4 2.0 36.8 2.0 31.9 0.2 26.2 1.1 22.7 0.3 22.5 1.0 19.1 0.4 10.6 10.1 0.2 6.6 4.6 0.6 2.9 0.1 0.8 0.1 n.d. n.d. n.d. n.d.
23.9 0.3 20.4 1.8 18.0 2.8 18.4 2.6 13.6 0.4 10.0 1.6 12.5 2.1 11.5 0.3 3.1 10.5 0.2 3.3 0.6 0.2 2.6 0.1 0.7 0.1 n.d. n.d. n.d. n.d.
N,N-dichlorination products which was shown by the shift in UV absorption to 309 nm and confirmed by GC-MS data (Figure S6aec). Also for the amino acid proline no NCl3 formation could be observed. But again the UV absorbance band with a maximum at 309 nm was a typical hint for N,Ndichlorinated products.
3.2. ratios
NCl3 formation for different chlorine to nitrogen
Urea and ammonia as major precursors and glycine as an example for an amino acid of medium NCl3 formation potential were selected to investigate the effect of different molar chlorine to nitrogen ratios on the NCl3 formation (Cl/ N ¼ 0.5 to 5). The results in Fig. 2 are given in % of the stoichiometric theoretical yield. The theoretical yield is defined as the complete conversion of the whole initial nitrogen or chlorine to NCl3. The NCl3 formation from ammonium ions and urea at substoichiometric Cl/N ratios (e.g. 1) at pH 2.5 revealed the preference of the formation of trichloramine compared to mono and dichlorinated products (Fig. 2a). If this would be not the case no NCl3 would be formed at Cl/N ratios of 2 and below. This is in accordance with the nearly quantitative transformation reactions of monochloramine (Eq. (4)) and dichloramine (Eq. (5)) at pH < 3 (Corbett et al., 1953). 2NH2 Cl þ Hþ #NHCl2 þ NHþ 4
(4)
3NHCl2 þ Hþ #2NCl3 þ NHþ 4
(5)
Urea revealed a considerably high NCl3 formation potential at neutral pH (Fig. 2b). 38% and 27% of the theoretical NCl3 yield could be achieved at the sub-stoichiometric ratios of Cl/N of 1 and
2686
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 8 1 e2 6 9 0
Fig. 3 e NCl3 formation of urea at different initial concentrations (6.7 3 10L5 mol/L N; 6.7 3 10L4 mol/L N and 1.0 3 10L2 mol/L N) and varied molar Cl/N ratios in buffered solutions at pH 6.8 ± 0.1.
Fig. 2 e NCl3 e yield in % of the stoichiometric theoretical yield for urea, ammonium chloride and glycine after chlorination at different molar chlorine to nitrogen ratios (Cl/N) - Reaction of 1.0 3 10L2 mol/L N-compound (as nitrogen) with chlorine, in buffered solutions at pH 2.5 ± 0.2 (2a) and pH 6.8 ± 0.1 (2b).
0.5 which means that 13% and 5% of the nitrogen in urea have been transformed to NCl3 under this conditions (Cl/N ¼ 1 and Cl/ N ¼ 0.5). As expected glycine revealed as a less efficient NCl3 precursor under sub-stoichiometric conditions. At Cl/N ratio of 1 only 3.5% and 0.8% of the theoretical NCl3 yield could be measured as reaction product of glycine with chlorine at acidic and neutral pH, respectively. This can be explained by the rapid conversion of glycine to N-mono- and N,N-dichloroglycine followed by the rate limiting decay to further products (Na and Olson, 2006).
Due to the slow reaction of urea with chlorine the required reaction time had to be increased with decreasing urea concentrations. For a nitrogen concentration of 1.0 102 mol/L 30 min were sufficient, whereas for a concentration of 6.7 105 mol/L up to 72 h were required to reach the maximum NCl3 yields. Generally, decreasing initial urea concentrations resulted in decreasing NCl3 yields independent of the Cl/N ratio. The lower NCl3 yields at low urea concentrations can be explained by the increasing importance of the decomposition reaction of NCl3 or other side reactions in aqueous solutions with increasing reaction time. Data on half-lives between 6 and 25 h of aqueous NCl3 solutions at pH 7 were published by Saguinsin and Morris (1975) and Kumar et al. (1987), whereas the higher half-lives are consistent with the experiments at low concentrations. The results for a Cl/N ratio of 1 show that 13% of the urea nitrogen were transformed to NCl3 for c(N) of 1.0 102 mol/L, and 7.5% and 4.7% of the urea nitrogen were transformed to NCl3 for c(N) of 6.7 104 mol/L and c(N) of 6.7 105 mol/L. The results are in accordance to that of Blatchley and Cheng (2010) who found NCl3 yields in the same order of magnitude. At typical conditions relevant for swimming pools with urea concentrations of 2 mg/L and chlorine concentrations of 0.6 mg/L (Cl/N ¼ 0.15) up to 1% of the urea nitrogen can be transformed to NCl3 despite the sub-stoichiometric Cl/N ratio.
3.4. Trichloramine in swimming pools e modeling the indoor pool situation and practical consequences 3.3.
Trichloramine formation from urea
Since urea revealed as the most important NCl3 precursor, also at neutral pH, and is a main constituent of the bathers load, the formation of NCl3 from urea was considered in more detail. Urea concentrations up to 2 mg/L could be measured in different swimming pools (Table S3). Initial urea concentrations of 2 mg/L (6.7 105 mol/L nitrogen) and 20 mg/L (6.7 104 mol/L nitrogen) were selected for chlorination experiments, and the data were compared to the already discussed results at higher concentrations. In Fig. 3 the NCl3 yields at pH 6.8 are given for different molar Cl/N ratios.
In swimming pools the air concentration of NCl3 is governed by several, partly coupled processes: - Formation and degradation of NCl3 in water from the reaction of precursors and free chlorine. - Partitioning of NCl3 between the water and gas phase. - Advective flow (air exchange) and degradation in the gas phase. The kinetics of NCl3 formation, degradation and the mass transfer from the water to the gas phase will be considered in
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 8 1 e2 6 9 0
more detail. The understanding of the major drivers is important to take the best measures to minimize NCl3 concentration in the water and in the gas phase of swimming pools, especially of indoor pools. In the lab experiment only chlorine and urea are present as reaction partners. So a pseudo first order reaction rate constant of 4.58 106 s1 for the reduction of chlorine concentration could be determined for reaction conditions typical for real SPW with concentrations of free chlorine of 1.13 105 mol/L and urea of 3.3 105 mol/L (Cl/N ¼ 0.17; pH 6.9; T ¼ 30 C) (Fig. 4). Since urea is present in high excess compared to chlorine, only 4% of urea would be converted to NCl3, if chlorine is consumed completely. The initial urea concentration won’t be decreased considerably. NCl3 is a known instable product in aqueous solutions with first order decay rate constants between 3.4 105 s1 (20 C, Saguinsin and Morris, 1975) and 7.5 106 s1 (25 C, Kumar et al., 1987) at pH 7. With this data the reaction of chlorine and urea was described as an irreversible consecutive reaction were NCl3 is formed as an intermediate product (Eq. (7)). k1 and k2 are first order reaction rate constants. k1
k2
Urea þ HOCl/NCl3 /Products
(7)
The following time dependent modeled concentrations of free chlorine and NCl3 shown in Fig. 4 were obtained from (Eq. (8)) and (Eq. (9)). ½Chlorine ¼ ½Chlorine0 ek1t
(8)
½NCl3 ¼ k1=ðk2 k1Þ ek1t ek2t ½Chlorine0
(9)
With the low decay constant from the literature of 7.5 106 s1 a steady state level of 2.96 106 mol/L NCl3 and
2687
with the higher decay constant of 3.4 105 s1 NCl3 concentrations up to 1.03 106 mol/L were obtained. This results in a transformation between 1.5 and 4.4% of the nitrogen in urea to NCl3. In the next step the mass transfer of NCl3 from the water to the air was considered. According to the Deacon’s boundary layer model (in Schwarzenbach et al., 2003) the calculated liquid mass transfer coefficients were vNCl3 w ¼ 0.6 103 cm/s for quiescent (unused) pools, vNCl3 w ¼ 2.4 103 cm/s for rippled (normal used) pools and vNCl3 w ¼ 4.4 103 cm/s for rough surfaces (whirlpool) (paragraph 6 supporting information). The NCl3 concentration in water cNCl3 w of 0.08 mg/L is a very realistic assumption and would result from an urea concentration of 2 mg/L and a 1% conversion to NCl3 (cf. long term NCl3 formation at sub-stoichiometric Cl/N ratios in Fig. 3 and kinetic consideration as an irreversible consecutive reaction Fig. 4). The NCl3 concentration of 0.08 mg/L is also in accordance to literature data for mean NCl3 levels of 11 swimming pools between 0.02 and 0.19 mg/L (Weaver et al., 2009). Resulting fluxes for NCl3 from water to air are then calculated to 1.8 103 g h1 m2, 7.0 103 g h1 m2 and 12.6 103 g h1 m2 again for quiescent, rippled and rough water surfaces, respectively (Table 2). To transfer the NCl3 fluxes to air concentrations of NCl3 a typical indoor pool geometry (water surface 312.5 m2, air volume 5760 m3) and two scenarios for air ventilation rates have been assumed. A minimum air ventilation rate of 0.5 per hour at a fresh air supply of 30% is recommended by the German standard (VDI 2089-1, 2008; DIN 4108, 2003). A typical ventilation rate is 2 air exchanges per hour at a fresh air supply of 50%. The resulting air concentrations of NCl3 range from 0.10 mg/m3 for smooth water surface at a high ventilation rate to 4.5 mg/m3 for rough water surface at a low ventilation rate. The data clearly show that for a typical bather load NCl3 concentrations in the air can be well above the
Fig. 4 e Measured and modeled reaction progress of chlorine with urea. - c0 (free chlorine) [ 1.13 3 10L5 mol/L; c0 (urea) 3.33 3 10L5 mol/L, pH 6.9 ± 0.1; T [ 30 C. (error bars 1 s (n [ 2) for the measured values). Pseudo first order decay of free chlorine: y [ 1.05EL05eL4.58EL06x; R2 [ 0.988). Modeled NCl3 concentrations from decay rate k2a [ 7.5 3 10L6 sL1 (Kumar et al., 1987) (dashed line) and k2b [ 3.4 3 10L5 sL1 (Saguinsin and Morris, 1975) (dotted-dashed line).
2688
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 8 1 e2 6 9 0
Table 2 e Calculated trichloramine concentrations in indoor pool air. - Modeling parameters: c(NCl3) (water): 0.08 mg/L; volume of water Vw: 960 m3; surface area of water A: 312.5 m2; volume of air Va: 5760 m3, water temperature T: 28 C.
Input vNCl3 w (cm s1) cNCl3 w (mg L1) Output FNCl3 w (g h1 m2) RNCl3 w (g h1) RNCl3 w =Vair (mg m3 h1) cNCl3 w air (mg m3)
Quiescent surface (unused pool)
Rippled surface (used swimmer pool)
Rough surface (whirlpool)
0.6 103 0.08
2.4 103 0.08
4.4 103 0.08
1.8 103 0.55 0.10 high ventilationb low ventilationa 0.64 0.10
7.0 103 2.20 0.38 low ventilationa high ventilationb 2.54 0.38
12.6 103 3.93 0.68 low ventilationa high ventilationb 4.54 0.68
a air ventilation rate 0.5/h and fresh air percentage of 30%. b air ventilation rate 2/h and fresh air percentage of 50% (typical).
proposed INRS guideline value of 0.5 mg/m3 of NCl3 in air (Hery et al., 1995; Gagnaire et al., 1994) under low air exchange and high stripping conditions. The water-air exchange reveals as a slow process compared to the turnover rate of the pool water treatment and the decay rate of NCl3 in water. The transfer of NCl3 from water to air of a swimming pool would take 5.8 d or 20 h for smooth or rough water surface, respectively. Therefore a part of the dissolved NCl3 can be removed in the treatment cycle. Hence two means are possible to minimize the NCl3 exposure of bathers: the reduction of urea due to its slow reaction kinetics of NCl3 formation and the degradation of NCl3 due to its slow mass transfer. Both processes are slower than a typical treatment cycle of a 25 m swimming pool which is between 6 and 8 h. Up to now urea is predominantly degraded by chlorine only. Ozonation is not very efficient due to the short contact time of often less than 1 min and due to the slow reaction rates of ozone with urea (k ¼ 0.05 L mol1 s1) (Eichelsdoerfer and von Harpe, 1970; Hoigne´ and Bader, 1983). The decomposition of NCl3 in water can also be taken into consideration if sufficient reaction time is provided or for example UV irradiation was already used to control combined chlorine and NCl3 (Cassan et al., 2006; De Laat and Berne, 2009). A more challenging task for the future is to develop a cost efficient and quantitative decomposition process for urea and other precursors in pool water treatment.
irreversible consecutive reaction with a slow formation and a faster degradation of trichloramine. Under real swimming pool water conditions, a conversion of 1% of the urea nitrogen to trichloramine was determined. Furthermore, the mass transfer of trichloramine from the water into the gas phase was calculated with the Deacon’s boundary layer model. For typical NCl3 concentrations in the water, air concentrations of NCl3 varied over a wide range and strongly correlated with the characteristics of the air-water surface and the ventilation. However the mass transfer of NCl3 from water to air is quite slow compared to a typical treatment cycle. Therefore the removal of precursor compounds and the disinfection byproduct trichloramine in the pool water treatment would be an attractive possibility to improve the water and air quality of swimming pools.
Acknowledgment The German Ministry of Research and Education (BMBF) is acknowledged for financial support of the Joint Project “Health related optimization of swimming pool water treatment” (Project number 02WT1090).
Appendix. Supporting information 4.
Conclusions
The formation of trichloramine in swimming pool water and the mass transfer from the aqueous to the gas phase was investigated. The results demonstrated that urea, ammonium ions, a-amino acids and creatinine from the bathers load can be considered as precursors for trichloramine formation in swimming pool water. Urea revealed as the most important trichloramine precursor. At neutral pH values NCl3 yields of chlorination experiments at sub-stoichiometric chlorine to nitrogen ratios revealed the preference of the formation of trichloramine compared to mono- and dichlorinated products. The reaction kinetics of trichloramine formation and degradation was regarded in a simplified model as an
Supplementary data related to this article can be found online at doi:10.1016/j.watres.2011.02.024.
references
Bernard, A., Carbonnelle, S., de Burbure, C., Michel, O., Nickmilder, M., 2006. Chlorinated pool attendance, atopy, and the risk of asthma during childhood. Environmental Health Perspectives 114 (10), 1567e1573. Bernard, A., Carbonnelle, S., Michel, O., Higuet, S., de Burbure, C., Buchet, J.-P., Hermans, C., Dumont, X., Doyle, I., 2003. Lung hyperpermeability and asthma prevalence in schoolchildren: unexpected associations with the attendance at indoor
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 8 1 e2 6 9 0
chlorinated swimming pools. Occupational and Environmental Medicine 60 (6), 385e394. Blatchley, E.R., Cheng, M., 2010. Reaction mechanism for chlorination of urea. Environmental Science & Technology 44 (22), 8529e8534. Cassan, D., Mercier, B., Castex, F., Rambaud, A., 2006. Effects of medium-pressure UV lamps radiation on water quality in a chlorinated indoor swimming pool. Chemosphere 62 (9), 1507e1513. Corbett, R.E., Metcalf, W.S., Soper, F.G., 1953. N-Halo compounds. IV. The reaction between ammonia and chlorine in aqueous solution, and the hydrolysis constants of chloramines. Journal of the Chemical Society 1927e1929. Deacon, E.L., 1977. Gas transfer to and across an air-water interface. Tellus 29 (4), 363e374. Deborde, M., von Gunten, U., 2008. Reactions of chlorine with inorganic and organic compounds during water treatment e kinetics and mechanisms: a critical review. Water Research 42 (1e2), 13e51. De Laat, J., Berne, F., 2009. Removal of chloramines from swimming-pool waters by UV irradiation. Literature review. European Journal of Water Quality 40 (2), 129e149 (in French). Eichelsdoerfer, D., von Harpe, T., 1970. Effect of ozone on urea in swimming pool water treatment. Vom Wasser 37 (1), 73e81 (in German). European standard EN ISO 7393-2, 2000. Water Quality Determination of Free Chlorine and Total Chlorine e Part 2: Colorimetric Method Using N, N-diethyl-1,4-phenylendiamine for Routine Control Purposes. European Committee for Standardization, Beuth Verlag, Berlin. Fuchs, J., 1962. Chlorination of swimming-pool water. ChemikerZeitung 86 (3), 76e82 (in German). Gagnaire, F., Azim, S., Bonnet, P., Hecht, G., Hery, M., 1994. Comparison of the sensory irritation response in mice to chlorine and nitrogen trichloride. Journal of Applied Toxicology 14 (6), 405e409. German Guideline VDI 2089 part 1, 2008. Building Services in Swimming Baths - Indoor Pools. Technische Geba¨udeausru¨stung im Verein Deutscher Ingenieure e.V., Du¨sseldorf. Beuth Verlag, Berlin. German Standard DIN 19643-1, 1997. Treatment of Water of Swimming-Pools and Baths. Part 1: General Requirements. Normenausschuss Wasserwesen (NAW) im DIN Deutsches Institut fu¨r Normung. Beuth Verlag, Berlin (in German). German Standard DIN 4108, 2003. Thermal Protection and Energy Economy in Buildings. Part 2: Minimum Requirements to Thermal Insulation. Normenausschuss Bauwesen (NABau) im DIN Deutsches Institut fu¨r Normung. Beuth Verlag, Berlin (in German). Hand, V.C., Margerum, D.W., 1983. Kinetics and mechanisms of the decomposition of dichloramine in aqueous solution. Inorganic Chemistry 22 (10), 1449e1456. Hery, M., Hecht, G., Gerber, J.M., Gendre, J.C., Hubert, G., Rebuffaud, J., 1995. Exposure to chloramines in the atmosphere of indoor swimming pools. Annals of Occupational Hygiene 39 (4), 427e439. Hoigne´, J., Bader, H., 1983. Rate constants of reactions of ozone with organic and inorganic compounds in water-II. Dissociating organic compounds. Water Research 17 (2), 185e194. Holzwarth, G., Balmer, R.G., Soni, L., 1984. The fate of chlorine and chloramines in cooling towers. Henry’s law constants for flashoff. Water Research 18 (11), 1421e1427. Jafvert, C.T., Valentine, R.L., 1992. Reaction scheme for the chlorination of ammoniacal water. Environmental Science & Technology 26 (3), 577e586. Kumar, K., Shinnes, R.W., Margerum, D.W., 1987. Kinetics and mechanisms of the base decomposition of nitrogen trichloride in aqueous solution. Inorganic Chemistry 26 (21), 3430e3434.
2689
Levesque, B., Duchesne, J.F., Gingras, S., Lavoie, R., Prud’homme, D., Bernard, E., Boulet, L.P., Ernst, P., 2006. The determinants of prevalence of health complaints among young competitive swimmers. International Archives of Occupational and Environmental Health 80 (1), 32e39. Li, J., Blatchley III, E.R., 2007. Volatile disinfection byproduct formation resulting from chlorination of organic-nitrogen precursors in swimming pools. Environmental Science & Technology 41 (19), 6732e6739. Mackay, D., Yeun, A.T.K., 1983. Mass transfer coefficient correlations for volatilization of organic solutes from water. Environmental Science & Technology 17 (4), 211e217. Massin, N., Bohadena, A.B., Wild, P., Hery, M., Toamain, J.P., Hubert, G., 1998. Respiratory symptoms and bronchial responsiveness in lifeguards exposed to nitrogen trichloride in indoor swimming pools. Occupational and Environmental Medicine 55 (4), 258e263. Morris, J.C., Isaac, R.A., 1983. In: Jolleys, R.L., Brungs, W.A., Cotruvo, J.A., Cumming, R.B., Mattice, J.S. (Eds.), A Critical Review of Kinetic and Thermodynamic Constants for Aqueous Chlorine-ammonia System. Na, C., Olson, T.M., 2006. Mechanism and kinetics of cyanogen chloride formation from the chlorination of glycine. Environmental Science & Technology 40 (22), 1469e1477. Na, C., Olson, T.M., 2004. Stability of cyanogen chloride in the presence of free chlorine and monochloramine. Environmental Science & Technology 38 (22), 6037e6043. Nickmilder, M., Bernard, A., 2007. Ecological association between childhood asthma and availability of indoor chlorinated swimming pools in Europe. Occupational and Environmental Medicine 64 (1), 37e46. Qiang, Z., Adams, C., 2004. Determination of monochloramine formation rate constants with stopped-flow spectrophotometry. Environmental Science & Technology 38 (5), 1435e1444. Richardson, S.D., Plewa, M.J., Wagner, E.D., Schoeny, R., DeMarini, D.M., 2007. Occurrence, genotoxicity, and carcinogenicity of regulated and emerging disinfection byproducts in drinking water: a review and roadmap for research. Mutation Research, Reviews in Mutation Research 636 (1e3), 178e242. Saguinsin, L.S., Morris, C.J., 1975. The chemistry of aqueous nitrogen trichloride. In: Johnson, J.D. (Ed.), Disinfection Water and Wastewater. Ann Arbor Science, Michigan, pp. 277e299. Sander, R., 1999. Compilation of Henry’s Law Constants for Inorganic and Organic Species of Potential Importance in Environmental Chemistry. Max-Planck Institute of Chemistry, Mainz, Germany. http://www.mpch-mainz.mpg.de/wsander/ res/henry.html Version 3(28.01.10). Savickas, P.J., LaPack, M.A., Tou, J.C., 1989. Hollow fiber membrane probes for the in situ spectrometric monitoring of nitrogen trichloride formation during wastewater treatment. Analytical Chemistry 61 (20), 2332e2336. Schlessinger, G.G., 1966. Nitrogen trichloride-ether mix explodes. Chemical & Engineering News 44 (8), 46. Schwarzenbach, R.P., Molnar-Kubica, E., Giger, W., Wakeham, S.G., 1979. Distribution, residence time, and fluxes of tetrachloroethylene and 1,4-dichlorobenzene in Lake Zurich, Switzerland. Environmental Science & Technology 13 (11), 1367e1373. Schwarzenbach, R.P., Gschwend, P.M., Imboden, D., 2003. Environmental Organic Chemistry, second ed. Wiley, New York, pp. 887e939. SciFinder Scholar, Version 2006. Calculated using ACD/Labs Software V8.14 for Solaris (ª 1994-2009 ACD/Labs). Shang, C., Blatchley III, E.R., 1999. Differentiation and quantification of free chlorine and inorganic chloramines in
2690
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 8 1 e2 6 9 0
aqueous solution by MIMS. Environmental Science & Technology 33 (13), 2218e2223. Stottmeister, E., Voigt, K., 2006. Trichloramine in the air of indoor swimming pools. Archiv des Badewesens 3, 158e162 (in German). Weaver, W.A., Li, J., Wen, Y., Johnston, J., Blatchley, M.R., Blatchley III, E.R., 2009. Volatile disinfection by-product analysis from chlorinated indoor swimming pools. Water Research 43 (13), 3308e3318.
Zirbs, R., 2008. Messprojekt: trichloramine in the air of indoor pools. Pollutants in indoor pools. Unfallversicherung aktuell 3, 10e13 (in German). Zwiener, C., Richardson, S.D., De Marini, D.M., Grummt, T., Glauner, T., Frimmel, F.H., 2007. Drowning in disinfection by-products? Assessing swimming pool water. Environmental Science & Technology 41 (2), 363e372.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 9 1 e2 6 9 9
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Redistribution of wastewater alkalinity with a microbial fuel cell to support nitrification of reject water Oskar Modin a,*, Kensuke Fukushi a, Korneel Rabaey b, Rene´ A. Rozendal b, Kazuo Yamamoto c a
Integrated Research System for Sustainability Science, The University of Tokyo, 7-3-1, Hongo, Bunkyoku, Tokyo 113-8654, Japan Advanced Water Management Centre, The University of Queensland, St Lucia, Queensland 4072, Australia c Environmental Science Center, The University of Tokyo, 7-3-1 Hongo, Bunkyoku, Tokyo 113-0033, Japan b
article info
abstract
Article history:
In wastewater treatment plants, the reject water from the sludge treatment processes
Received 18 November 2010
typically contains high ammonium concentrations, which constitute a significant internal
Received in revised form
nitrogen load in the plant. Often, a separate nitrification reactor is used to treat the reject
21 February 2011
water before it is fed back into the plant. The nitrification reaction consumes alkalinity,
Accepted 23 February 2011
which has to be replenished by dosing e.g. NaOH or Ca(OH)2. In this study, we investigated
Available online 21 March 2011
the use of a two-compartment microbial fuel cell (MFC) to redistribute alkalinity from influent wastewater to support nitrification of reject water. In an MFC, alkalinity is
Keywords:
consumed in the anode compartment and produced in the cathode compartment. We use
Microbial fuel cell
this phenomenon and the fact that the influent wastewater flow is many times larger than
Nitrification
the reject water flow to transfer alkalinity from the influent wastewater to the reject water.
Wastewater
In a laboratory-scale system, ammonium oxidation of synthetic reject water passed
Reject water
through the cathode chamber of an MFC, increased from 73.8 8.9 mgN/L under opencircuit conditions to 160.1 4.8 mgN/L when a current of 1.96 0.37 mA (15.1 mA/L total MFC liquid volume) was flowing through the MFC. These results demonstrated the positive effect of an MFC on ammonium oxidation of alkalinity-limited reject water. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
Microbial fuel cells (MFCs) have been proposed as a wastewater treatment system that simultaneously removes biochemical oxygen demand (BOD) and produces electrical energy (Habermann and Pommer, 1991; Liu et al., 2004; Min et al., 2005). An MFC consists of an anode compartment in which microorganisms oxidize organic compounds and use the anode as electron acceptor. The electrons flow through an external circuit to the cathode, where e.g. oxygen is reduced. The difference in redox potential between the organic
compounds at the anode and the oxygen at the cathode drives the electrical current and makes it possible to recover energy. To balance the flow of electrical charge, an equal but opposite ionic charge migrate between the anode and cathode in the liquid. A variety of MFC configurations have been developed, including single-chamber MFCs (Liu and Logan, 2004) and MFCs with liquid recirculation between the anode and cathode chambers (Freguia et al., 2008). However, a common MFC design is a system where the anode and cathode chambers are separated by an ion exchange membrane that limits the crossover of oxygen from the cathode to the anode and
* Corresponding author. Chalmers University of Technology, Department of Civil and Environmental Engineering, Water Environment Technology section, 412 96 Go¨teborg, Sweden. Fax: þ46 31 189705. E-mail address:
[email protected] (O. Modin). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.02.031
2692
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 9 1 e2 6 9 9
organics from the anode to the cathode, but allows the exchange of ions between the two compartments (Logan et al., 2006). The application of ion exchange membranes in microbial fuel cells treating wastewater has led to operational problems. Gil et al. (2003) used a Nafion cation exchange membrane and observed that the pH decreased in the anode chamber and increased in cathode chamber, leading to lower current. The pH shifts could be counteracted by using a strong phosphate buffer (Gil et al., 2003); however, the addition of buffers is not practical in real wastewater treatment. Assuming acetate is the organic substance oxidized at the anode and oxygen is reduced at the cathode, the stoichiometric equations responsible for current production in an MFC are as follows:
Anode: CH3COO þ 3H2O / CO2 þ HCO3 þ 8Hþ þ 8e
(1)
Cathode: 2O2 þ 8Hþ þ 8e / 4H2O
(2)
As shown in Equations (1) and (2), for pH shifts to be avoided the current must be balanced by an equal transfer of protons through the ion exchange membrane. Rozendal et al. (2006) investigated transfer of ions through a Nafion cation exchange membrane used in MFCs operated on synthetic wastewater. They showed that mainly ions other than Hþ, such as Naþ, Kþ, NH4þ, Ca2þ, and Mg2þ, were responsible for the ionic charge transfer through the membrane (Rozendal et al., 2006). Since the cations mentioned above are present in wastewater in concentrations significantly higher than Hþ, the use of cation exchange membranes in MFCs treating wastewater will invariably lead to pH shifts in both compartments. This phenomenon, however, does not necessarily have to be a drawback but could be utilized for specific applications. For example, Cao et al. (2009) developed a three-compartment MFC using cation and anion exchange membranes to desalinate water (Cao et al., 2009). Rabaey et al. (2010) developed a bioelectrochemical system for the production of alkaline solutions. In line with this, a two-compartment MFC in which the anode and cathode compartments are separated by an ion exchange membrane could potentially also be used to redistribute alkalinity within a wastewater treatment plant. Reject water from the sludge dewatering processes in a treatment plant contains high ammonium concentrations (Arnold et al., 2000). To reduce the ammonium load to the plant, the reject water is often nitrified in a separate process before it is fed back into the plant. As the nitrification reactions consume alkalinity, NaOH or Ca(OH)2 has to be added to counteract pH drops that would otherwise damage the biological activity (Arnold et al., 2000). With an MFC, we could extract alkalinity from the influent wastewater and transfer it to the reject water to support nitrification. This could potentially eliminate the need for an external input of chemicals to control the pH. The goal of this study was to investigate the concept of using an MFC to redistribute alkalinity in a wastewater treatment plant to support nitrification of reject water. We used a laboratory-scale experimental setup to demonstrate the effect of alkalinity transfer accomplished by an MFC on nitrification and assess the feasibility of the concept.
2.
Materials and methods
2.1.
Experimental setup
The experimental setup is shown in Fig. 1. The MFC consisted of two acrylic frames with internal dimensions 10 4 2 cm3. The frames formed two chambers, which were separated by a Nafion 117 cation exchange membrane with a surface area of 40 cm2. The Nafion membrane rested against a perforated acylic plate that provided structural support (the perforation had an area of 15.3 cm2). An acrylic plate was used to close the anode chamber, which contained a 10-cm long, 0.615-cm diameter graphite rod (Alfa Aesar) in contact with an 8.3 3.6 0.5 cm3 carbon fiber felt (TMIL Tsukuba, Japan) acting as the anode. On the cathode side, an Avcarb P75T carbon paper sheet coated with Pt (1%) on activated carbon (Aldrich Chemicals) and 30% PTFE was used to close the cathode chamber, with the Pt-coated side facing the liquid. The air-facing side was coated with a layer of 40% PTFE mixed with graphite powder to reduce liquid loss. A stainless steel mesh was pressed against the air-facing side of the carbon sheet to provide electrical contact, and a perforated acrylic plate was pressed against the stainless steel mesh to provide structural support (the perforations had an area of 6.4 cm2). The liquid volume was 64 mL in the anode chamber and 66 mL in the cathode chamber. The synthetic wastewater was continuously fed through the anode chamber of the MFC and collected in an effluent vessel. The synthetic reject water was fed through the cathode of the MFC and then through a nitrification column. The nitrification column was made of a 25-cm high, 3.5-cm diameter acrylic pipe. An aeration stone was placed at the bottom of the pipe, which was then filled with 200 g of 1-mm diameter glass beads that served as support material for microbial growth. The glass beads reached a height of 15 cm in the column. The effluent port was placed 20 cm from the bottom. The influent port was located at the bottom of the
Fig. 1 e Experimental setup. The synthetic wastewater was continuously fed through the anode chamber of the MFC and collected in an effluent vessel. The synthetic reject water was fed through the cathode of the microbial fuel cells and then through a nitrification column. The anode and cathode compartments were separated by a cation exchange membrane (CEM).
2693
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 9 1 e2 6 9 9
column. An airflow of 150 mL/min was continuously provided to the column through the aeration stone.
2.2.
Operation
The synthetic wastewater was prepared by adding the following compounds to tap water (mg/L): 200 MgSO47H2O, 150 CaCl22H2O, 10 FeSO47H2O, 2925 NaCl, 1.6 KH2PO4, 3.0 Na2HPO4, 100 NH4Cl, and 500 CH3COONa. The synthetic reject water was prepared the same way with the following exceptions: 1000 mg/L NH4Cl, 1000 mg/L NaHCO3, and no acetate. The wastewater was fed through the anode chamber at a flow rate of approximately 0.4 L/d. The reject water was fed through the cathode chamber and nitrification column at a flow rate of approximately 0.05 L/d. The MFC anode was inoculated by connecting it in series with an operating MFC for 36 days. A 150-U resistor was connected between the anode and cathode. The MFC was then operated in a stand-alone fashion (with 50 mM NaCl being recirculated through the cathode) for 19 days before the cathode was connected with the reject water flow. The nitrification column was inoculated with activated sludge and operated stand-alone for 19 days before the reject water flow was fed through the cathode. The experiment was divided into three runs. In Run A, the system was operated as shown in Fig. 1. In Run B, the MFC was bypassed and the reject water was fed directly to the nitrification column. In Run C, the nitrification column was disconnected and the MFC was operated alone. The runs were further subdivided into a total of nine periods with varying experimental conditions. Run A consisted of periods 1e4, Run B consisted of period 5, and Run C consisted of periods 6e9 (Table 1).
2.3.
2.4.
Calculations
The current-dependent alkalinity transfer from the anode compartment to the cathode compartment was calculated assuming ions other than Hþ or OH were responsible for charge transfer through the cation exchange membrane used in the MFC (Equation (3)). DAlk ¼
I 3600 FQ
(3)
where DAlk is the alkalinity change in the cathode compartment (meq/L), 3600 represents seconds per hour, I is the current (mA), F is Faraday’s constant (96,485.3 mC/meq), and Q is the flow rate through the cathode compartment (L/h). The oxidation of ammonium to nitrite consumes alkalinity (Equation (4)).
Analytical methods
The acetate concentration was measured by HPLC (HP 1100 series) and UV absorbance at 210 nm or by ion chromatography
Table 1 e Operational conditions and duration of the experimental periods. Run
(Metrohm 761 Compact IC). Nitrate, nitrite, sulfate, and phosphate concentrations were measured with ion chromatography (Metrohm 761 Compact IC). The total carbonate alkalinity was measured by titrating 50 mL of a sample with 0.1 M or 0.5 M HCl to a pH less than 4.5. Ammonium concentrations were measured with a salicylate method (Hach High Range TNT). Total nitrogen (TN) concentration was measured with a TOC-V analyzer equipped with TNM-1 unit (Shimadzu). Calcium concentrations were measured using ICP-AES (Perkin Elmer Optima 3000DV). Voltages were recorded using an NI USB-6008 data logger (National Instruments) connected to a PC. Polarization curves of the MFC were recorded by varying the external resistances. The anode and cathode potentials were measured against Ag/AgCl reference electrodes (BAS Inc.) but are reported against the Normal Hydrogen Electrode (NHE), with an assumed offset of 0.197 V.
Period
Days
Comment
A A A A
1 2 3 4
0e40 41e53 53e91 91e106
B
5
107e121
150 U resistor Open-circuit 18 U resistor 18 U resistor, no aeration of nitrification column Synthetic reject water fed directly to the aerated nitrification column without passing the cathode 18 U resistor, no nitrification column Open-circuit, no nitrification column 0.5 V input voltage, current measured over 10 U resistor, no nitrification column Open-circuit, no nitrification column
C
6
128e142
C
7
142e152
C
8
152e162
C
9
162e177
NH4þ þ 1.5O2 / NO2 þ 2Hþ þ H2O
(4)
The alkalinity concentration consumed in the cathode compartment and nitrification column was calculated from the measured concentrations of nitrite (nitrate was generally not produced in the experiments) (Equation (5)). þ DNO 2 Nðmg=LÞ 2 meq H =mmol NO2 DAlk ¼ 14ðmg N=mmolÞ
3.
Results and discussion
3.1.
Microbial fuel cell performance
(5)
The anode of the MFC was fed with synthetic wastewater containing acetate. The resistance connected between the anode and cathode was varied (see Table 1) to achieve currents of varying magnitude (Fig. 2). In one experimental period, Run C e period 8, an input voltage of 0.5 V was supplied to the MFC to achieve a higher current. The anode influent and effluent acetate concentration and pH are shown in Fig. 2. The effluent pH was affected by the current. In periods 1, 3, 4, 6, and 8 when current was flowing in the MFC, there was typically a drop in pH. In periods 2, 7, and 9
2694
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 9 1 e2 6 9 9
Fig. 3 e Cell voltage, power curves, and electrode potentials for the MFC on day 40 and 106. Fig. 2 e Current flowing in the MFC (upper diagram) and acetate concentration and pH in the influent and effluent synthetic wastewater fed to the MFC anode (lower diagram).
when no current was flowing, there was a small increase in pH as a result of acetate being oxidized to HCO3. In period 1, the coulombic efficiency (CE) was 4.6e8.8%. In periods 3 and 4, the CE increased to 10.8e40.0%, probably as a result of the higher current in the MFC. In periods 6 and 8 the CE was 8.1e14.9%. Sulfate reduction, methanogenesis, and aerobic oxidation in the effluent collection vessel are all possible mechanisms responsible for lowering the CE. Sulfate reduction was quantified by measuring the sulfate concentration in the anode influent and effluent between day 89 and 105. During this period, sulfate reduction accounted for 20.5 1.0% of the removed acetate, whereas current production accounted for 11.9 0.9%. Varying degree of aerobic oxidation of acetate in the collection vessel probably caused the large variation of observed CE. Polarization curves of the MFC were obtained on day 40 and day 106. The results are shown in Fig. 3. The maximum power output was 0.18 mW (1.38 mW/L total MFC liquid volume) on day 40 and 0.17 mW (1.31 mW/L) on day 106. The internal resistance of the MFC was calculated from the slope of the linear portion of the cell voltage vs. current curves in Fig. 3. On day 40 it was 153 U and on day 106 it was 158 U.
3.2.
The influent and effluent NH4þeN concentrations and the effluent NO2eN and NO3eN concentrations are shown in Fig. 4. In Run A, we can see the effect of current on the effluent NH4þ concentrations. In period 1 with a current of 1.01 0.01 mA, the effluent NH4þeN concentration dropped to 56.3 8.8 mg/L. In period 2 with zero current, the concentration increased to 110.1 8.3 mg/L. In period 3 with a current of
Nitrification performance
NH4þeN, NO2eN, and NO3eN concentrations were measured in the influent and effluent from the cathode-nitrification column system. TN concentration measurements correlated well with the sum of NH4þeN, NO2eN, and NO3eN concentrations showing that other forms of nitrogen were not playing a major role in the system.
Fig. 4 e Influent and effluent ammonium concentrations (upper diagram) and effluent nitrate and nitrite concentrations (lower diagram) from the cathode-coupled nitrification column (Run A), nitrification column alone (Run B), and cathode compartment alone (Run C).
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 9 1 e2 6 9 9
1.96 0.37, effluent NH4þeN concentrations decreased to only 12.4 5.1 mg/L. In period 4 when the aeration was turned off in the nitrification column, the concentration increased to 75.7 12.1 mg/L. In Run B and C, when the nitrification column and MFC were disconnected from each other, the effluent NH4þeN concentration was higher, between 124 and 206 mg/L. Throughout the experiment, the effluent NO3eN concentration was near zero suggesting that nitritation of ammonium to nitrite occurred, but nitrite was not further oxidized to nitrate. This peculiarity of the study is further elaborated on in the supplementary data. We can see the effect of current on ammonium oxidation in periods 1 to 3. The effluent NO2eN concentration was 117.0 3.9 mg/L in period 1, decreased to 81.9 5.5 mg/L in period 2, and increased up to 176.7 3.0 mg/ L in period 3. When the aeration was shut off in period 4, the concentration dropped down to as low as 42.8 mg/L but increased gradually. Dissolved oxygen present in the influent, oxygen entering the cathode chamber through the gas-diffusion cathode, and oxygen dissolution at the air/water interface in the nitrification column could have supported ammonium oxidation even when the air pump was shut off. In Run B, the concentration stabilized at 79.0 4.2 mg/L, which is close to the concentration observed in period 2 of Run A. Even when the nitrification column was disconnected in Run C, NO2eN concentrations ranging from 10 to 80 mg/L could be observed in the cathode effluent suggesting ammonium-oxidizing biomass existed in the cathode compartment as well. The influent and effluent pH and alkalinity concentrations are shown in Fig. 5. In Run A and B, the effluent alkalinity concentration was always close to zero suggesting it limited nitrification except in period 4 when oxygen was the limiting factor. In Run C, the alkalinity concentration in the effluent ranged from 3.4 to 20.5 meq/L. In this phase of the experiment, oxygen supply and the amount of ammonium-oxidizing biomass in the cathode compartment limited nitrification. The pH was the lowest in periods 2 and 5. Since period 2 was open-circuit and period 5 took place during Run B when the nitrification column was disconnected from the MFC, no flow of current was contributing alkalinity to the reject water in these two periods, resulting in a low pH in the effluent.
3.3.
Mass balance on nitrogen
Mass balances on nitrogen for the cathode-nitrification column systems are shown in Table 2. Run A had higher NH4þ removal and NO2 production than Run B and C. The results show that the magnitude of the current flowing in the MFC had a clear effect on the NH4þ removal and NO2 production when alkalinity-limited nitrification. In Run A, the NH4þ removal and NO2 production was highest in period 3, which had the highest current. Low NH4þ removal and NO2 production was observed in period 2 when no current was flowing in the system, in period 4 when the aeration was shut off, and in Run B and C when the MFC and nitrification column were operated separately. In Run A and C, a significant amount of removed NH4þ could not be accounted for by production of NO2. Possible nitrogen sinks include (i) assimilation of nitrogen by bacteria
2695
Fig. 5 e Influent and effluent pH (upper diagram) and influent and effluent alkalinity concentrations (lower diagram) from the cathode-coupled nitrification column (Run A), nitrification column alone (Run B), and cathode compartment alone (Run C).
in the cathode-nitrification column system, (ii) denitrification of NO3 and NO2 to N2 gas, (iii) precipitation of struvite, (iv) volatilization of NH3 gas, and (v) diffusion of NH4þ from the cathode through the cation exchange membrane to the anode. (i) The assimilated nitrogen can be estimated since the respiration rate (i.e. the produced NO2) is known. Assuming ammonium-oxidizers have a yield coefficient of 0.34 gVSS/ gNH4þeN and a biomass elemental composition of C5H7O2N (Rittmann and McCarty, 2001), the N removal caused by assimilation in the experiment can be estimated to vary between 0.7 0.5 mg/L in period 9 to 6.7 0.2 mg/L in period 3. This means that at most 10% of the unaccounted N was assimilated into biomass. (ii) Denitrification was an unlikely N sink because no organic carbon source was supplied to the cathode-nitrification column system. Cathodes can also serve as electron donors for denitrification (Clauwaert et al., 2007; Virdis et al., 2008, 2010). However, in our experiments the unaccounted N concentration was similar with and without current flow current in the MFC (e.g. compare periods 1, 2, and 3). Therefore, bioelectrochemical denitrification did not seem to play an important role in the system. (iii) The chemical formula of struvite is (NH4)MgPO46H2O. Our synthetic reject water had a P concentration of 0.033 mM, thus struvite precipitation can only have been responsible for a loss of 0.46 mg/L of NH4eN, which is negligible compared to the total N loss.
2696
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 9 1 e2 6 9 9
Table 2 e Mass balance on nitrogen in the cathode-nitrification column system (Run A, periods 1e4), nitrification column alone (Run B, period 5), and cathode compartment alone (Run C, periods 6e9). Run
Period
A A A A B C C C C
1 2 3 4 5 6 7 8 9
DNH4þeN (mg/L) 184.5 141.1 218.4 149.4 58.5 85.5 83.3 130.7 64.1
11.8 10.0 6.0 15.4 18.0 9.7 5.7 1.0 4.1
DNO2eN (mg/L) 114.9 73.8 160.1 33.2 49.2 18.9 38.4 36.2 17.6
5.3 8.9 4.8 22.0 4.6 2.5 7.9 1.4 12.6
(iv) Volatilization of ammonia is possible at pH values approaching 9.3, which is the pKa of NH4þ. In all periods except 4, 6, and 8, the effluent pH was too low for volatilization to be likely. Particularly in period 4 when the pH in the effluent from the nitrification column reached 9.1 and in period 8 when the pH in the effluent from the cathode compartment reached 9.4, volatilization may have played an important role. (v) The NH4þeN concentration going into the cathode compartment was 235 23 mg/L whereas the NH4þeN concentration in the anode compartment was around 20e25 mg/L, thus there was a steep diffusion gradient across the cation exchange membrane. All experimental periods except 4, 5, and 8 had similar concentrations of unaccounted N, most of which could probably be attributed to diffusion into the anode. Periods 4 and 8 had higher unaccounted N, which could be attributed to NH3 gas volatilization, since the effluent pH in these periods was close to the pKa of ammonium. In period 5 (Run B), the concentration of unaccounted N was nearly zero. This makes sense because the two sinks assumed to have contributed most to the concentration of unaccounted N in the other periods, namely diffusion through the cation exchange membrane and volatilization at high pH, were not possible.
3.4.
Mass balance on alkalinity
Mass balances on total carbonate alkalinity in the cathodenitrification column system are shown in Table 3. The alkalinity increase caused by transfer of cations through the membrane was calculated based on the measured current assuming the charge transfer accomplished by Hþ was
DNO3eN (mg/L) 0.6 1.0 0.6 1.4 0.0 1.1 1.4 1.6 1.3
0.2 0.4 0.2 0.2 0.0 0.2 0.1 0.3 0.3
Unaccounted N (mg/L) 69.6 67.2 58.3 116.3 9.3 66.6 44.8 94.4 46.5
13.0 13.4 7.7 26.8 18.6 11.4 10.7 2.2 14.3
negligible (see Section 2.4), which is a reasonable assumption in wastewater systems (Rozendal et al., 2006). The alkalinity consumption caused by the nitritation reaction was calculated based on the nitrite concentration increase in the effluent (see Section 2.4). There was a low NO3eN concentration (<2 mg/L) in the tap water used to prepare the synthetic reject water. The change in NO3 concentration between the influent and effluent was very small, so its possible effects of on the alkalinity mass balances were not included. In periods 1, 3, 4, 6, and 8, there were alkalinity drops that cannot be accounted for by production of NO2. A possible explanation is precipitation of calcium carbonate. To check the plausibility of this explanation, the influent and effluent calcium concentrations were measured between day 31 and 121 (Fig. 6). Theoretically, there should be a slight increase in calcium concentration when current is flowing due to transfer through the cation exchange membrane. However, in periods 1 and 3 the effluent concentration was clearly lower than the influent, whereas in period 2, the effluent concentration was higher. This suggests that the alkalinity production in the cathode chamber when current was flowing led to precipitation of calcium, and some of the precipitates redissolved when the current was stopped. In period 5 (Run B), when the reject water was fed directly to the nitrification column without passing the cathode, the influent calcium concentration is almost identical to the effluent. Volatilization of NH4þ to NH3(g), which is assumed to have occurred in period 4, would have contributed to a portion of the unaccounted alkalinity (roughly around 3.7 meq/L).
Table 3 e Mass balance on total carbonate alkalinity in the cathode-nitrification column system (Run A, periods 1e4), nitrification column alone (Run B, period 5), and cathode compartment alone (Run C, periods 6e9). Run A A A A B C C C C
Period
D(Infl. Effl.) (meq/L)
1 2 3 4 5 6 7 8 9
11.4 0.5 10.1 0.8 8.8 0.2 6.4 3.2 7.2 0.3 3.5 1.0 3.3 2.0 7.6 1.1 2.0 2.4
I-dep. (meq/L) 20.5 0 39.9 33.4 0 23.0 0 47.8 0
0.9 7.6 2.6 1.2 2.5
DNO2-dep. (meq/L)
Unaccounted Alk (meq/L)
16.4 0.4 10.5 0.6 22.9 0.3 4.7 1.6 7.0 0.3 2.7 0.4 5.5 1.1 5.2 0.2 2.5 1.8
15.5 1.1 0.4 1.0 25.8 7.6 22.2 4.4 0.2 0.4 16.9 1.6 2.2 2.3 35.0 2.7 0.6 3.0
2697
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 9 1 e2 6 9 9
between the anode and cathode compartments. An alkalinity increase in the cathode compartment corresponds to an equal alkalinity decrease in the anode compartment. In a wastewater treatment plant, the influent wastewater flow rate is usually more than 100 times larger than the reject water flow rate. This means that if alkalinity was redistributed between the influent wastewater and the sludge stream, a small decrease in the alkalinity concentration in the influent wastewater would correspond to an over 100 times larger increase in the alkalinity concentration in the reject water flow stream. The relationship between anode and cathode flow rates and alkalinity concentration is described by Equation (6). Qanode DAlkanode ¼ Qcathode DAlkcathode
Fig. 6 e Influent and effluent calcium concentrations from the cathode-coupled nitrification column (Run A) and nitrification column alone (Run B).
In the periods when no current was flowing in the MFC, the alkalinity mass balances closed, which means that the difference of alkalinity concentration in the influent and effluent approximately corresponded to the alkalinity consumption by ammonium-oxidizing bacteria. However, in the periods when current was flowing in the MFC, there was always an alkalinity concentration unaccounted for in the reactor. The magnitude of the alkalinity transfer caused by current was calculated assuming ions other than Hþ was responsible for charge transfer through the cation exchange membrane. Combining the results from all experimental periods with flowing current, the average ratio of the unaccounted alkalinity to the current-dependent alkalinity transfer was 71 5%. This means that only 29 5% of the alkalinity transfer caused by current could be useful for supporting nitrification in this reactor setup. At the end of the experiment, the cathode compartment was flushed first with deionized water and then with 1% HNO3. The content of the effluent 1% HNO3 solution was analyzed to give a qualitative indication of the precipitates existing in the compartment. The three major cations were Ca2þ 24.0 mM, Naþ 2.3 mM, and Mg2þ 1.0 mM, giving a Ca:Na:Mg ratio of 1:0.09:0.04. This can be compared to the ratio in the synthetic reject water, which was 1:42.1:0.7. Thus, Ca2þ made up a much larger proportion both in relation to Naþ and Mg2þ, suggesting most precipitates in the cathode compartment were calcium. The PO43 concentration in the 1% HNO3 solution was 0.36 mM, which suggests that the precipitates were calcium carbonate rather than calcium phosphate.
3.5.
Discussion of concept
In this study we have presented a new concept for using an MFC for alkalinity redistribution within a wastewater treatment plant to support nitrification of reject water. A twocompartment MFC containing a cation exchange membrane and operating on wastewater may redistribute the alkalinity
(6)
where Q is the flow rate through the anode and the cathode compartments, DAlk is the change in alkalinity concentration, which is negative in the anode and positive in the cathode. Fig. 7 shows how an MFC could be integrated in a conventional wastewater treatment plant to transfer alkalinity from the wastewater flow stream to support nitrification of the reject water. The feasibility of this concept is shown by a simple calculation. Assume 6 mg/L of BOD is converted into current at the anode. This corresponds to a small alkalinity drop of 0.75 meq/L in the wastewater (Equation (7)).
DAlkðmeq=LÞ ¼
BODðmg=LÞ 4ðmeq=mmol O2 Þ 32ðmg=mmol O2 Þ
(7)
However, if the ratio between wastewater flow and reject water flow is 100/1, it also corresponds to an alkalinity increase of 75 meq/L in the reject water flow. This could theoretically support nitrification of an additional 525 mg/L of NH4þeN (Equation (8)). DNH 4 Nðmg=LÞ ¼
DAlkðmeq=LÞ 14ðmg N=mmolÞ 2 meq Hþ produced=mmol NHþ 4 oxidized (8)
Denitrification of the nitrite and nitrate fed into the activated sludge treatment would somewhat counteract the wastewater alkalinity drop in the anode. To reduce resistive losses it will likely be beneficial to install several small MFCs rather than one large unit. In this study, feeding synthetic reject water through the cathode of an MFC increased the produced NO2eN concentration from 73.8 8.9 mg/L in period 2 when no current was flowing to 160.1 4.8 mg/L in period 3 with a current of 1.96 0.37 mA flowing in the MFC. Apart from oxidation to NO2, diffusion of NH4þ from the cathode chamber to the cation exchange membrane to the anode chambers is believed to have been a major sink of ammonium. Assimilation of N into biomass was likely responsible for a smaller ammonium loss in the system. We estimate that in period 3, which had the highest ammonium oxidation, 73 3% of the ammonium was converted to nitrite, 3 0% was assimilated into biomass, and 24 4% diffused into the anode compartment. Other ammonium sinks were negligible. A large portion of the current-dependent alkalinity transfer in the system was not useful to support ammonium oxidation, but lost in the system. On average, only 29 5% of the
2698
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 9 1 e2 6 9 9
Fig. 7 e Schematic showing how a microbial fuel cell with anode and cathode compartments separated by a cation exchange membrane could be integrated in a conventional wastewater treatment plant to redistribute alkalinity between the wastewater and reject water flow streams.
alkalinity that should theoretically have been transferred to the cathode compartment was actually useful for ammonium oxidation. Measured calcium concentrations suggested precipitation of calcium carbonate may have been responsible for a portion of the lost alkalinity. Transfer of Hþ ions and precipitation of other species may also have contributed. An alternative reactor configuration that might address these issues is a combined cathode-nitrification compartment where the cathode is also acting as support for nitrifiers. Such a system was investigated by You et al. (2009) with the aim of improving the cathode performance of an MFC. Virdis et al. (2010) also investigated such as system for combined nitrification and denitrification at the cathode. Another alternative to utilize the alkalinity increase in the cathode would be not to focus on ammonium oxidation but rather ammonium recovery by either struvite precipitation or volatilization and capture. A third alternative would be to produce a concentrated caustic solution in a separate reactor and then inject the caustic into the reject water stream. In this study, the system was operated as an MFC with an output of electrical energy (except in period 8). It would also be possible to operate the system in short circuit or with a voltage input to improve the rate of the process. Further studies are required to improve efficiencies and ensure the long-term stability of the process.
4.
Conclusions
In this study we have presented a new concept for how MFC technology could be integrated into a wastewater treatment plant. The purpose of the MFC is to support nitrification of reject water from the sludge treatment processes by redistributing alkalinity from the influent wastewater. We operated a laboratory system on synthetic wastewater and reject water. The ammonium oxidation in the system increased from 73.8 8.9 mgN/L under open-circuit condition to 160.1 4.8 mgN/L when a current of 1.96 0.37 mA was flowing in the MFC (i.e. an increase by 117 20%).
Acknowledgments The project was partially supported by Japan Society for Promotion of Science (JSPS) KAKENHI (21360250). O.M. was supported by a post-doctoral fellowship from JSPS. K.R. is supported by the Australian Research Council (DP0879245) and a University of Queensland Research Excellence Award. RR is supported by the University of Queensland (UQ Early Career Researcher 2009; UQ Post-doctoral Research Fellowship 2010).
Appendix A. Supplementary material Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.watres.2011.02.031.
references
Arnold, E., Bo¨hm, B., Wilderer, P.A., 2000. Application of activated sludge and biofilm sequencing batch reactor technology to treat reject water from sludge dewatering systems: a comparison. Water Science and Technology 41 (1), 115e122. Cao, X., Huang, X., Liang, P., Xiao, K., Zhou, Y., Zhang, X., Logan, B. E., 2009. A new method for water desalination using microbial desalination cells. Environmental Science & Technology 43, 7148e7152. Clauwaert, P., Rabaey, K., Aelterman, P., de Schamphelaire, L., Pham, T.H., Boeckx, P., Boon, N., Verstraete, W., 2007. Biological denitrification in microbial fuel cells. Environmental Science & Technology 41 (9), 3354e3360. Freguia, S., Rabaey, K., Yuan, Z., Keller, J., 2008. Sequential anodecathode configuration improves cathodic oxygen reduction and effluent quality of microbial fuel cells. Water Research 42, 1387e1396. Gil, G.-C., Chang, I.-S., Kim, B.H., Kim, M., Jang, J.-K., Park, H.S., Kim, H.J., 2003. Operational parameters affecting the performance of a mediator-less microbial fuel cell. Biosensors and Bioelectronics 18, 327e334.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 6 9 1 e2 6 9 9
Habermann, W., Pommer, E., 1991. Biological fuel cells with sulphide storage capacity. Applied Microbiology and Biotechnology 35, 128e133. Liu, H., Ramnarayanan, R., Logan, B.E., 2004. Production of electricity during wastewater treatment using a single chamber microbial fuel cell. Environmental Science & Technology 38, 2281e2285. Liu, H., Logan, B.E., 2004. Electricity generation using an aircathode single chamber microbial fuel cell in the presence and absence of a proton exchange membrane. Environmental Science & Technology 38, 4040e4046. Logan, B.E., Hamelers, B., Rozendal, R., Schro¨der, U., Keller, J., Freguia, S., Aelterman, P., Verstraete, W., Rabaey, K., 2006. Microbial fuel cell: methodology and technology. Environmental Science & Technology 40 (17), 5181e5192. Min, B., Kim, J.R., Oh, S.E., Regan, J.M., Logan, B.E., 2005. Electricity generation from swine wastewater using microbial fuel cells. Water Research 39, 4961e4968. Rabaey, K., Butzer, S., Brown, S., Keller, J., Rozendal, R.A., 2010. High current generation coupled to caustic production using
2699
a lamellar bioelectrochemical system. Environmental Science & Technology 44 (11), 4315e4321. Rittmann, B.E., McCarty, P.L., 2001. Environmental Biotechnology: Principles and Applications. McGraw-Hill, New York. Rozendal, R.A., Hamelers, H.V.M., Buisman, C.J.N., 2006. Effects of membrane cation transport on pH and microbial fuel cell performance. Environmental Science & Technology 40, 5206e5211. Virdis, B., Rabaey, K., Yuan, Z., Keller, J., 2008. Microbial fuel cells for simultaneous carbon and nitrogen removal. Water Research 42, 3013e3024. Virdis, B., Rabaey, K., Rozendal, R.A., Yuan, Z., Keller, J., 2010. Simultaneous nitrification, denitrification and carbon removal in microbial fuel cells. Water Research 44, 2970e2980. You, S.-J., Ren, N.-Q., Zhao, Q.-L., Kiely, P.-D., Wang, J.-Y., Yang, F.-L., Fu, L., Peng, L., 2009. Improving phosphate-buffer free cathode performance of microbial fuel cell based on biological nitrification. Biosensors and Bioelectronics 24, 3698e3701.
