WATER RESEARCH A Journal of the International Water Association
Editor-in-Chief Mark van Loosdrecht Delft University of Technology and KWR Watercycle Research The Netherlands E-mail:
[email protected]
Editors Jean-Claude Block Université H. Poincaré, Nancy I France David Dixon University of Melbourne Australia Hiroaki Furumai University of Tokyo Japan Xiaodi Hao Beijing University of Civil Engineering and Architecture China Gregory Korshin University of Washington USA Anna Ledin Formas Sweden Eberhard Morgenroth ETH Zürich and Eawag, Swiss Federal Institute of Aquatic Science and Technology Switzerland Wolfgang Rauch University of Innsbruck Austria Maria Reis Universidade Nova de Lisboa/FCT Portugal Hang-Shik Shin Korea Advanced Institute of Science and Technology Korea Thomas Ternes Bundesanstalt für Gewässerkunde Germany Stefan Wuertz University of California, Davis USA
Associate Editors Andrew Baker University of New South Wales Australia Damien Batstone University of Queensland Australia
Guang-Hao Chen Hong Kong University of Science & Technology China Tom Curtis University of Newcastle upon Tyne UK Francis de los Reyes III North Carolina State University USA Ana Deletic Monash University USA Rob Eldridge Glen Waverley Australia Monica B. Emelko University of Waterloo Canada Tim Fletcher University of Melbourne Australia Rosina Girones University of Barcelona Spain Stephen Gray Victoria University Australia Kate Grudpan Chiang Mai University Thailand Edwin Herricks University of Illinois at Urbana-Champaign USA Peter Hillis AECOM Australia Hong-Ying Hu Tsinghua University China Bruce Jefferson Cranfield University UK Ulf Jeppsson Lund University Sweden Sergey Kalyuzhnyi RUSNANO Russian Federation Jae-Hong Kim Georgia Institute of Technology USA Günter Langergraber University of Natural Resources and Applied Life Sciences, Vienna (BOKU) Austria Shang-Lien Lo National Taiwan University Taiwan Dionisis Mantzavinos Technical University of Crete Greece
Yoshihiko Matsui Hokkaido University Japan Armand Maul Université de Metz France Max Maurer Eawag Switzerland How Yong Ng National University of Singapore Singapore Satoshi Okabe Hokkaido University Japan Jong Moon Park Pohang University of Science & Technology Korea Susan Richardson US Environmental Protection Agency USA Miguel Salgot University of Barcelona Spain David Sedlak University of California, Berkeley USA Jean-Philippe Steyer LBE-INRA France Masahiro Takahashi Hokkaido University Japan Nathalie Tufenkji McGill University Canada Kai Udert Eawag Switzerland Vayalam Venugopalan BARC Facilities India Eduardo von Sperling Federal University of Minas Gerais Brazil Julie Zilles University of Illinois at Urbana-Champaign USA Anastasios Zouboulis Aristotle University of Thessaloniki Greece
Editorial Office E-mail:
[email protected]
Publication information: Water Research (ISSN 0043-1354). For 2011, volume 45 is scheduled for publication. Subscription prices are available upon request from the publisher or from the Elsevier Customer Service Department nearest you or from this journal’s website (http://www.elsevier.com/locate/watres). Further information is available on this journal and other Elsevier products through Elsevier’s website (http://www.elsevier.com). Subscriptions are accepted on a prepaid basis only and are entered on a calendar year basis. Issues are sent by standard mail (surface within Europe, air delivery outside Europe). Priority rates are available upon request. Claims for missing issues should be made within six months of the date of dispatch. Author Enquiries: For enquiries relating to the submission of articles (including electronic submission) please visit this journal’s homepage at http://www.elsevier.com/ locate/watres. Contact details for questions arising after acceptance of an article, especially those relating to proofs, will be provided by the publisher. You can track accepted articles at http://www.elsevier.com/trackarticle. You can also check our Author FAQs at http://www.elsevier.com/authorFAQ and/or contact Customer Support via http://support.elsevier.com. Orders, claims, and journal enquiries: Please contact the Elsevier Customer Service Department nearest you: St. Louis: Elsevier Customer Service Department, 3251 Riverport Lane, Maryland Heights, MO 63043, USA; phone: (877) 8397126 [toll free within the USA]; (+1) (314) 4478878 [outside the USA]; fax: (+1) (314) 4478077; e-mail:
[email protected] Oxford: Elsevier Customer Service Department, The Boulevard, Langford Lane, Kidlington OX5 1GB, UK; phone: (+44) (1865) 843434; fax: (+44) (1865) 843970; e-mail:
[email protected] Tokyo: Elsevier Customer Service Department, 4F Higashi-Azabu, 1-Chome Bldg, 1-9-15 Higashi-Azabu, Minato-ku, Tokyo 106-0044, Japan; phone: (+81) (3) 5561 5037; fax: (+81) (3) 5561 5047; e-mail:
[email protected] Singapore: Elsevier Customer Service Department, 3 Killiney Road, #08-01 Winsland House I, Singapore 239519; phone: (+65) 63490222; fax: (+65) 67331510; e-mail:
[email protected] Application for membership of International Water Association should be made to: Executive Director, IWA, Alliance House, 12 Caxton Street, London SW1H 0QS, U.K. (Tel.: +44 207 654 5500; Fax: +44 207 654 5555; e-mail:
[email protected]; website: http://www.IWAhq.org.uk). Registered Charity (England) No. 289269. Individual membership is available from £30 upwards. For details contact IWA.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 0 7 e5 8 2 5
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Associations among pathogenic bacteria, parasites, and environmental and land use factors in multiple mixed-use watersheds G. Wilkes a, T.A. Edge b, V.P.J. Gannon c, C. Jokinen c, E. Lyautey d, N.F. Neumann e,f, N. Ruecker e,g, A. Scott d, M. Sunohara a, E. Topp d, D.R. Lapen a,* a
Eastern Cereal and Oilseed Research Centre, Agriculture and Agri - Food Canada, 960 Carling Ave., Ottawa, Ontario, Canada K1A 0C6 Aquatic Ecosystem Protection Research Division, Water Science and Technology Directorate, National Water Research Institute (NWRI), Environment Canada, Burlington, Ontario, Canada c Laboratory for Foodborne Zoonoses, Public Health Agency of Canada, Lethbridge, Alberta, Canada d Southern Crop Protection and Food Research Centre, Agriculture and Agri - Food Canada, London, Ontario, Canada e Alberta Provincial Laboratory for Public Health (Microbiology), Edmonton, Alberta, Canada f School of Public Health, University of Alberta, Edmonton, Alberta, Canada g Department of Microbiology and Infectious Diseases, University of Calgary, Calgary, Alberta, Canada b
article info
abstract
Article history:
Over a five year period (2004e08), 1171 surface water samples were collected from up to 24
Received 13 September 2010
sampling locations representing a wide range of stream orders, in a river basin in eastern
Received in revised form
Ontario, Canada. Water was analyzed for Cryptosporidium oocysts and Giardia cyst densi-
15 June 2011
ties, the presence of Salmonella enterica subspecies enterica, Campylobacter spp., Listeria
Accepted 20 June 2011
monocytogenes, and Escherichia coli O157:H7. The study objective was to explore associations
Available online 26 June 2011
among pathogen densities/occurrence and objectively defined land use, weather, hydrologic, and water quality variables using CART (Classification and Regression Tree) and
Keywords:
binary logistical regression techniques. E. coli O157:H7 detections were infrequent, but
Water quality
detections were related to upstream livestock pasture density; 20% of the detections were
CART
located where cattle have access to the watercourses. The ratio of detections:non-
Watershed
detections for Campylobacter spp. was relatively higher (>1) when mean air temperatures
Land use
were 6% below mean study period temperature values (relatively cooler periods). Cooler
Agriculture
water temperatures, which can promote bacteria survival and represent times when land
Bacteria pathogens
applications of manure typically occur (spring and fall), may have promoted increased
Cryptosporidium
frequency of Campylobacter spp. Fifty-nine percent of all Salmonella spp. detections occurred
Giardia
when river discharge on a branch of the river system of Shreve stream order ¼ 9550 was
Weather
>83 percentile. Hydrological events that promote off farm/off field/in stream transport
Season
must manifest themselves in order for detection of Salmonella spp. to occur in surface water
Hydrology
in this region. Fifty seven percent of L. monocytogenes detections occurred in spring, relative to other seasons. It was speculated that a combination of winter livestock housing, silage feeding during winter, and spring application of manure that accrued during winter, contributed to elevated occurrences of this pathogen in spring. Cryptosporidium and Giardia oocyst and cyst densities were, overall, positively associated with surface water discharge, and negatively associated with air/water temperature during spring-summer-fall. Yet, some of the highest Cryptosporidium oocyst densities were associated with low discharge
* Corresponding author. Tel.: þ1 613 759 1537; fax: þ1 613 759 1701. E-mail address:
[email protected] (D.R. Lapen). 0043-1354/$ e see front matter Crown Copyright ª 2011 Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.06.021
5808
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 0 7 e5 8 2 5
conditions on smaller order streams, suggesting wildlife as a contributing fecal source. Fifty six percent of all detections of 2 bacteria pathogens (including Campylobacter spp., Salmonella spp., and E. coli O157:H7) in water was associated with lower water temperatures (<w14 C; primarily spring and fall) and when total rainfall the week prior to sampling was >w27 mm (62 percentile). During higher water temperatures (>w14 C), a higher amount of weekly rainfall was necessary to promote detection of 2 pathogens (primarily summer; weekly rainfall w>42 mm (>77 percentile); 15% of all 2 detections). Less rainfall may have been necessary to mobilize pathogens from adjacent land, and/or in stream sediments, during cooler water conditions; as these are times when manures are applied to fields in the area, and soil water contents and water table depths are relatively higher. Season, stream order, turbidity, mean daily temperature, surface water discharge, cropland coverage, and nearest upstream distance to a barn and pasture were variables that were relatively strong and recurrent with regard to discriminating pathogen presence and absence, and parasite densities in surface water in the region. Crown Copyright ª 2011 Published by Elsevier Ltd. All rights reserved.
1.
Introduction
An estimated w3% of all deaths worldwide are attributed to unsafe water caused by poor sanitation and hygiene, a problem particularly acute in the ‘developing’ world (World Health Organization, 2009). In the ‘developed’ world, treatment of sewage and drinking water, and the implementation of laws and regulations that protect water supplies, have dramatically reduced the incidence of waterborne disease (NRC, 2004). Nevertheless, periodic waterborne disease outbreaks and endemic disease due to compromised drinking water or contaminated recreational water continue to be public health issues (O’Connor, 2002). In Canada, nearly 300 outbreaks of undefined waterborne disease associated with drinking water were reported from 1974 to 2001 (Shuster et al., 2005). About 47% of these outbreaks were of unknown etiology, about 22% were of bacterial origin, 22% of parasitic origin, 8% of viral origin, and 1% from multiple origins. Several lines of evidence suggest that disease outbreaks are more frequent after certain weather conditions, such as critical precipitation events. For instance, over half of the waterborne disease outbreaks in the United States in the past 50 years were preceded by heavy rainfall, and outbreaks due to surface water contamination show the strongest association with extreme precipitation (Curriero et al., 2001). Precipitation can drive the transport of fecal contaminants from land into watercourses, and increased water flow within watercourses can (re)-mobilize fecal contaminants within the flow zone proper. Moreover, large volumes of water can result in sewage outflows (Kistemann et al., 2002). Broadly, in agriculturally dominated landscapes, sources of fecal pollution in surface waters can result from livestock production activities (e.g., pasturing livestock, land application of biosolids/manure), wildlife, and human sources (e.g., such as leaky septic systems and sewage treatment effluent) (Quy et al., 1999; Sischo et al., 2000; BodleyTickell et al., 2002; Ahmed et al., 2005; Jellison et al., 2007; Ruecker et al., 2007; Keeley and Faulkner, 2008; Lyautey et al., 2010). The capacity to predict the occurrence and the densities of pathogens in surface waters, combined with the capacity to specify associated point and non-point sources of fecal pollution, will enhance our ability to meet water quality targets and
standards, as well as properly target mitigation practices (Ice, 2004; Benham et al., 2006; USEPA, 2005; Rao et al., 2009). However, to date, predictive efforts have been applied with mixed success, and results have been generally poor and/or too site specific to be of widespread use. Limited data sets, difficulties in identifying the source of pathogens and indicator organisms, unknown background concentrations of pathogens/fecal indicator organisms, and limited understanding of the processes associated with transport and fate of microorganisms in the soil-water environment have contributed to mixed findings and marginal modeling success (Jamieson et al., 2004; Pachepsky et al., 2006; Gassman et al., 2007). Nevertheless, while empirical relationships among pathogen densities/ occurrence and pollution sources and driving factors have been documented, further study of those linkages in natural watershed settings will be required to; enhance our capacity to determine the precise nature of fecal pollution sources and drivers, improve predictions of the fate and transport of pathogens in surface water systems, and help identify practices and policies that will reduce exposure risks (LeChevallier et al., 1991; Sadeghi and Arnold, 2002; Benham et al., 2006; Peterson et al., 2006; Yang et al., 2008; Ivanek et al., 2009; USEPA, 2009). Wilkes et al. (2009) examined associations between fecal indicators, pathogenic bacteria, and parasites, the utility of using fecal indicators to understand the occurrence of pathogens, and relationships between these microorganisms and a small set of hydrological and weather variables for three years of data collected in the same study region examined in this paper. In this study, we measured over a five year period (2004e2008), a suite of bacterial and parasitic pathogens at 24 distinct water sampling sites representing river stem to municipal drainage systems in a mixed use, but agriculturally dominated landscape in eastern Ontario Canada. We identified environmental associations among pathogens and an extensive suite of independent factors (>89 environmental variables), which express: water physical/chemical properties, land use, weather, season, and hydrological information. Quantification of pathogen associations with environmental variables will help promote a better understanding of pathogen fate/ transport in surface water systems, elucidate strengths and limitations of mechanistic and stochastic pathogen-surface
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 0 7 e5 8 2 5
water risk assessment tools, and will help identify practices that can be employed to reduce pathogen exposure risks to animals and humans.
2.
Methods
2.1.
Site description
The study area (Fig. 1) is in a cool temperate humid continental climate in eastern Ontario Canada, with mean yearly air temperatures of 6.2 C, total yearly precipitation of 963 mm, and total yearly rainfalls of 771 mm (Environment Canada, 2010). The South Nation River basin is approximately 3900 km2. Slopes are generally less than 3% on agricultural fields, and most are tile drained. Soils of the region are mainly Orthic Humic Gleysols and Gleyed Melanic Brunisols (Soil Classification Working Group, 1998; CANSIS, 2007). Approximately 47% of the greater basin was under some form of agricultural activity in 2006 (AAFC, 2010). A majority of the farming activities revolve around dairy production (Statistics Canada, 2006). Manure application to land in spring and fall are common practices in the region (Lyautey et al., 2007; Ruecker et al., 2007; Cicek et al., 2010). Detailed descriptions of water sampling sites (sites 1e22) can be found in Ruecker et al. (2007) and Lyautey et al. (2007), and generalized descriptions of the sites with stream orders can be found in Table S1 of the appendix. Sample sites 23 and 24 were not included in the aforementioned studies and represent, respectively, a stream within a pasturing area where cattle have direct access to the water course and, a wetland and forested catchment where agricultural activity does not occur.
2.2.
Water sample collection
Water samples were collected bi-weekly from fall 2004 to fall 2008 (See Table S2 of the appendix). On a year by year basis, surface water sampling began post ice break-up (March to
5809
April) and ceased in late November to early December. Surface water was collected at sites 2e24, while water from site 1 was collected from a drinking water treatment plant, with the river water intake being located at a 6 m depth in the South Nation River proper. In 2004, site 1 was sampled weekly throughout the year and bi-weekly from 2005e2008. Water sampling at bridge sites 2e13 consisted of fixing sterile 1 L bottles (Systems Plus, Woodstock, Ontario, Canada) to a sampling pole (Nasco Swing Sampler, Fort Atkinson, WI) and submerging them to approximately 0.5e1 m below the surface. At non-bridge sampling sites 16-17 located on larger watercourses, the sampling pole was extended from the water course bank into the water surface for collection at depths of approximately 0.5e1 m. For sites 14e15 and 18e24, which were on smaller tributaries and drainage ditches, ‘grab samples’ were collected gently by hand using sterile gloves in a manner so as to not disturb bottom sediments. At each sampling site, 1 L of water was collected for Listeria monocytogenes detection, 2 L of water was collected for detection of Escherichia coli O157:H7, Salmonella enterica subspecies enterica, and Campylobacter spp., 20 L of water was collected for enumeration of Cryptosporidium oocysts and Giardia cyst densities, and 1 L of water was collected for nutrient analysis. Samples were stored on ice, shipped and processed as described in Lyautey et al. (2007), Ruecker et al. (2007), Wilkes et al. (2009) and Lyautey et al. (2011) for the aforementioned parasites, bacteria and nutrients. See Table S2 of the appendix regarding the temporal sampling regimes for the suite of microorganisms examined.
2.3.
Water analyses
At time of sampling, additional water was collected at each site and measured for temperature, pH, electrical conductivity, dissolved oxygen, and oxidation reduction potential using a YSI 556 Multi Probe System fitted with a YSI 5563 Probe (Yellow Springs Instruments Inc., Yellow Springs, OH). Turbidity measurements were made in the laboratory on these
Fig. 1 e Map of surface water sample sites, and hydrometric and meteorological stations for which data was utilized in this study.
5810
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 0 7 e5 8 2 5
Table 1 e Description of physical and chemical attributes measured from sampled water, and meteorological, hydrological, and land use variables. These variables were used as independent criteria in the data mining analyses. Variable Name
Variable Description (unit)
AMIA_AMN
NH3 (ammonia) þ NH4 (ammonium) concentration in sample water (mg L1)
NITRITE
1 NO 2 (nitrite) concn in sample water (mg L )
NITRATE
1 NO 3 (nitrate) concn in sample water (mg L )
REA_PHOS
Dissolved reactive phosphorus concn in sample water (mg L1)
TOTKN
Total Kjeldahl nitrogen (TKN) concn in sample water (mg L1)
TOTPHO
Total phosphorus concn in sample water (mg L1)
TEMP
Temperature of sample water ( C)
pH
pH of sample water
CONDUCTIVITY
Electrical conductivity of sample water (mS cm1)
DISS_ OXYGEN _P, DISS_ OXYGEN _MGL
Dissolved oxygen in sample water (% saturation and mg L1)
ORP
Oxidation reduction potential of sample water (mV)
TURBIDITY
Cloudiness of sample water as measured with a nephelometer sensor (NTU; nephelometric turbidity units)
DIS_SPC, DIS_RUS, DIS_PAY, DIS_WEBS
Mean daily river discharge at Spencerville (Shreve order ¼ 4805), Russell (Shreve order ¼ 9550), and Payne (Shreve order ¼ 3243) hydrometric stations and mean daily stream discharge at sample site 18 (DIS_WEBS; Shreve order ¼ 36) (m3 s1). See Fig. 1 for locations.
RUS_TOTALRAIN, RUS_TOTALRAINxD, WEBSRAIN_MM
Total Russell met. rainfall for day of sampling; total Russell met. rainfall for day of sampling and ¼ 1, 2, 3, and 7 days in advance of sampling day; total WEBs met. daily rainfall (mm)
RUS_MAXTEMP, RUS_MINTEMP, RUS_MEANTEMP, WEBSTEMP_C
Daily maximum, minimum, and mean air temperature at Russell met.; mean daily air temperature at WEBs met. ( C)
BASIN_(land use)
Proportion of land use (determined via RS) in total sample site catchment area: Cropland, developed, forage, pasture, shrubland, and forest (km2 km2)
CROPP_yK
Proportion of cropland, identified via RS, in catchment areas upstream of site with MUFL of 2, 5, and 10 km (km2 km2)
DEVELP_yK
Proportion of developed land, identified via RS, in catchment areas upstream of site with MUFL of 2, 5, and 10 km (km2 km2)
FORAGEP_yK
Proportion of forage land, identified via RS, in catchment areas upstream of site with MUFL of 2, 5, and 10 km (km2 km2)
PASTP_yK
Proportion of pasture land, identified via RS, in catchment areas upstream of site with MUFL of 2, 5, and 10 km (km2 km2)
SHRUBP_yK
Proportion of shrubland, identified via RS, in catchment areas upstream of site with MUFL of 2, 5, and 10 km (km2 km2)
VEGP_yK
Proportion of forest, identified via RS, in catchment areas upstream of site with MUFL of 2, 5, and 10 km (km2 km2)
NUD_(land use)
Upstream distance from sample location to nearest land use observations identified by ReSS of type barn (BARN), developed high (DEVHI), developed low (DEVLO), forest (FOREST), pasture (PASTURE), dairy operation (DAIRY_OP), cattle or dairy barn (CAT_DAI_BA), pasture with no restriction to water course (PAST_ACC), horse barn (HORSEBA), hog barn (HOGBA) and poultry barn (POULBA) (km)
DBARN_yK
Density of barns, identified via R-SS, in catchment areas upstream of site with MUFL of 2, 5, and 10 km (obs. km2)
DDEVHI_yK
Density of developed high, identified via R-SS, in catchment areas upstream of site with MUFL of 2, 5, and 10 km (obs. km2)
DDEVLO_yK
Density of developed low, identified via R-SS, in catchment areas upstream of site with MUFL of 2, 5, and 10 km (obs. km2)
DFORE_yK
Density of forest, identified via R-SS, in catchment areas upstream of site with MUFL of 2, 5, and 10 km (obs. km2)
DPAST_yK
Density of pasture, identified via R-SS, in catchment areas upstream of site with MUFL of 2, 5, and 10 km (obs. km2)
DDAIRYOP_yK
Density of dairy operations, identified via R-SS, in catchment areas upstream of site with MUFL of 2, 5, and 10 km (obs. km2) (continued on next page)
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 0 7 e5 8 2 5
5811
Table 1 (continued ) Variable Name
Variable Description (unit)
DCAT_OR_DA_yK
Density of cattle or dairy operations, identified via R-SS, in catchment areas upstream of site with MUFL of 2, 5, and 10 km (obs. km2)
DPAST_ACC_yK
Density of pasture with no watercourse restrictions, identified via R-SS, in catchment areas upstream of site with MUFL of 2, 5, and 10 km (obs. km2)
x ¼ number of days in advance of sample; y ¼ maximum upstream flow length (MUFL) from the sample site at which sample site catchment was defined; K ¼ km; obs., number of observations; RS (remote sensing); ReSS (road-side survey).
samples after 30 s of vigorous shaking using Global Water’s WQ770 Turbidimeter (Gold River, CA). After overnight refrigeration at 4 C, 300 mL of water from each site were delivered to the City of Ottawa, Robert O. Pickard Environmental Center (ROPEC) Laboratory for analysis of ammonia þ ammonium (automated phenate method, following Standard Method 4500-NH3), dissolved reactive phosphorus (automated ascorbic acid method, following Standard Method 4500-P), nitrite and nitrate (ion chromatography, following Standard Method 4110B), total Kjeldahl nitrogen (automated phenate method preceded by sulfuric acid digestion, following Standard Methods 4500-Norg), and total phosphorus (automated ascorbic acid method preceded by persulfate digestion, following Standard Method 4500-P F) (Clesceri et al., 1998). The microbiological methods used to enumerate and detect pathogens in the sampled water are described in depth in Wilkes et al. (2009), and the methods used to subtype Campylobacter spp. and Salmonella spp. are described in detail in Jokinen et al. (2011).
2.4.
Land use and stream order
Land use variables in Table 1 were defined to express; i) nearest upstream distance to a particular land use, ii) percent land use coverage in catchment areas upstream of a sampling site, and iii) land use densities in catchment areas upstream of a sample site. Continuous land use was characterized as described in Lyautey et al. (2007) from an unsupervised classification of Landsat 5 Thematic Mapper satellite imagery producing classes of cropland, developed land, forage land, pasture, shrubland and vegetation. Locations and types of point land uses (e.g., barns, developments) were defined by road-side surveys as described in Ruecker et al. (2007). To identify the total catchment area of a sampling site, as well as the catchment area for a sampling site with a maximum upstream distance of 2, 5, and 10 km from the site, the following Geographic Information System (GIS) method was employed. Flow direction was calculated in ArcMap 9.2 (Environmental Systems Research Institute, Redlands, CA). Next ‘catchment areas’ were identified to determine the areas contributing flow to the sampling locations using the ‘watershed tool/command’ and a digital elevation model in ArcMap. After identification of the total upstream catchment areas for each sample site, upstream flow lengths were calculated for each sample catchment and segmented to 2, 5 and 10 km (upstream maximums) using Boolean selection in GIS. Catchment areas for a sample site were then determined by constraining the area so that there was a maximum upstream
distance of 2, 5, and 10 km from the sample site. The point land use observations were spatially joined to the 2, 5 and 10 km upstream catchment area spatial layers for each of the 24 sites and summarized as the number of observations per km2 of the defined catchment area. Further, the point land use observations were ‘spatially joined’ to the upstream distance spatial layers, where pixel values of the upstream distance layer were assigned to coincident land use points, and the shortest distance per land use type for each sample location were defined. For continuous land use data, percentage coverage of cropland, developed land, forage land, pasture, shrubland and vegetation for the upstream segmented (2, 5 and 10 km max. upstream) and total upstream contributing areas were calculated using the tabulate area routine of ArcMap for all 24 sampling locations. Stream order (Shreve, 1966; Strahler, 1952) was determined via methods already discussed in Lyautey et al. (2010, 2011) (variables: SHREVE and STRAHLER).
2.5.
Discharge, weather and season data
Mean daily discharge data (m3 s1) was acquired from the Water Survey of Canada (2008) for Spencerville, Russell, and Payne gauging stations within the study region (Fig. 1). Discharge (m3 s1) at site 18 was monitored every 15 min using an area velocity sensor (Model 4150 ISCO, Lincoln, NE) fitted to a fixed culvert of known geometry and averaged to a mean daily value. Daily rainfall and (min., max., mean) temperature data were obtained from Environment Canada’s (2010) Meteorological Service for Russell, Ontario, upon which cumulative rainfall estimates were calculated on day of sample and 1, 2, 3 and 7 days in advance of the sampling day (Lyautey et al., 2010). Mean daily temperature and total daily rainfall monitored at a HOBO (Onset Computer Corp., Bourne MA) instrumented weather station (see Fig. 1 for WEBs meteorology site) in the study region were also included in the study (Sunohara et al., 2010). All samples were assigned a seasonal category relative to solstice and equinox dates (SEASON).
2.6.
Statistical analyses
Correlations amongst water physical/chemical, hydrological, weather, stream order, and parasite density data were examined using non-parametric Spearman rank correlation analyses. An Alpha (a) value of 0.05 was used as a significance threshold. Descriptive statistical summaries of the data were calculated in Statistica 9.0 (Statsoft Inc., Tulsa, OK). Mean
5812
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 0 7 e5 8 2 5
Fig. 2 e Example of how CART splits cases of a target variable (e.g., pathogen data) into four terminal nodes (maximum number of terminal nodes for two tree levels). Target data are optimally split into nodal groups on the basis of specific independent variable splitting criteria. Surrogate and competitor variable splitting criteria are associated with each optimal split. A minimum of two terminal nodes can be produced as an optimal tree in CART; these terminal nodes would consist of level 1 child nodes.
physical and chemical attributes of water were calculated by Shreve stream order categories of <95 (streams and ditches), Shreve order 95 and Shreve order <15499 (middle reaches), and Shreve order 15499 (river proper). Classification and regression tree analyses (CART Pro 6.0 Salford Systems, San Diego, CA) and binary logistic regression (Statistica 9.0 Statsoft Inc., Tulsa, OK) were statistical tools applied to identify environmental correlations and interactions among target pathogens and land use, weather, season, and water physical-chemical information. CART was used previously by Wilkes et al. (2009) in classification tree mode to examine thresholds of pathogen detection based on fecal indicator bacteria densities, and by Lyautey et al. (2010) in regression tree mode to examine E. coli density relationships with land use, climate, season and environmental factors. CART is a well-accepted automated, test and learn, nonparametric recursive partitioning methodology that attempts to split dependant variables into homogeneous nodes or data groupings based on independent variable data splitting criteria (Fig. 2) (Breiman et al., 1984; Steinberg and Colla, 1995). CART tests the best discriminating split of the dependent variable using all possible independent variable splitting possibilities. A root node in a tree consists of all available data for analysis and CART splits this data into two groups consisting of child nodes, which can be further partitioned into subsequent child nodes. Nodes that cannot be split further, as a result of user and/or CART defined stopping criteria, are designated as terminal nodes. A tree with a complete set of terminal nodes represents the predictive outcome of the tree model. The optimal CART tree produced in this manner results from a combination of optimizing predictive capacity as well as minimizing model complexity. CART regression trees were generated for Cryptosporidium oocyst and Giardia cyst densities generated using the independent environmental variables listed in Table 1, with two additional variables, season, and stream order. For CART regression tree analyses, both least absolute deviation (LAD) and least square (LS) methods were used as a basis to test data split conditions that best reduced nodal heterogeneity
(translated to a best ‘improvement score’ for each split condition). An improvement score is a measure of the quality of a split in terms of homogeneity and is used to compare the power of various splitting criteria (Steinberg and Colla, 1995). LAD employs mean absolute deviation to measure the homogeneity of the predictor in a nodal grouping, whereas LS uses mean square error in this capacity. For LAD, medians and least absolute deviations are calculated for the dependent variable of each node, whereas LS uses means and standard deviations. LAD was evaluated in this study because it is less sensitive to extreme values than LS. For categorical target variables such as the presence/ absence of Cryptosporidium, Giardia, L. monocytogenes, E. coli O157:H7, Salmonella spp., and Campylobacter spp., and the total number of detections at a particular site and time of E. coli O157:H7, Salmonella spp., and Campylobacter spp. (when these three bacterial pathogens were tested at the same time at a site), CART classification trees were generated using the default Gini data splitting criteria; a non-parametric approach for determining optimal dependent variable splitting in the manner already described above. Gini is a data splitting methodology that puts the largest variable class into one pure node, and all others into the opposite node (Breiman, 1996). Nodal improvement scores, an index of group homogeneity or ‘purity’, are presented, where higher improvement score values representing better variable performance in terms of differentiating groups where pathogens were detected and not detected. CART classification trees were generated from pathogen presence/absence data using the same independent environmental variables listed in Table 1, including season, and stream order. We constrained the CART tree analysis to target pathogens that had greater than 15 detections to eliminate data sparse targets. Given the exploratory nature of the study, and for purposes of tree model interpretation brevity, unless otherwise specified, we only interpret CART trees to a maximum of two tree levels which could produce a minimum of 2 (child nodes in level 1 with no further splitting) and a maximum of 4 terminal nodes (child nodes in level 2 with no further splitting)
Table 2 e Prediction of detections/non-detections of pathogenic bacteria and parasites resulting from use of the independent variables listed in Table 1 plus the variables SHREVE, STRAHLER, and SEASON in CART classification tree analyses. The results are restricted to two tree levels with a maximum of 4 terminal nodes (Fig. 2). We imposed a CART modeling constraint of having ‡10 observations per node. Microorganism Target (Total No. Samples)
Root Node Split Criteria
Root Node Split Level 1 Child Node Result of Root Node Split Split Criteria Root and Child Surrogate Variables Competitor Variables [Child node split code] Node Splits (Improvement Score)c (Improvement Score)c
Cryptosporidium (664) RUS_MAXTEMP 15.3 C RUS_MAXTEMP > 15.3 C RUS_MAXTEMP > 15.3 C
NS
220/73 (72)a
DDEVLO_5K 5.75 obs. km2 [1] DDEVLO_5K > 5.75 obs. km2 [1]
81/159 (26)a
Giardia (664)
RUS_MAXTEMP 14.8 C RUS_MAXTEMP > 14.8 C RUS_MAXTEMP > 14.8 C
NS
123/148 (72)a
BASIN_ SHRUBLAND 0.6% [2] BASIN_ SHRUBLAND > 0.6% [2]
27/62 (16)a
RUS_ MEANTEMP 7.7 C RUS_ MEANTEMP > 7.7 C RUS_ MEANTEMP > 7.7 C
NS
103/92 (35)a
DDEVLO_5K 1.75 obs. km2 [3] DDEVLO_5K > 1.75 obs. km2 [3]
94/191 (32)a
L. monocytogenes spp. (395)
SEASON ¼ SPRING NS SEASON ¼ SUMMER, SHREVE 1895.5 [4] FALL, WINTER SEASON ¼ SUMMER, SHREVE > 1895.5 [4] FALL, WINTER
RUS_MEANTEMP (93) RUS_MINTEMP (80) SEASON (61) DIS_WEBS r (55)
RUS_MEANTEMP (97) TEMP (83) RUS_MINTEMP (80) WEBSTEMP_C (70)
DDEVHI_10K (100) [1] DEVELP_2K (96) [1] BASIN_DEVELOPED (79) [1] DDEVLO_2K (78) [1]
DDEVHI_10K (100) [1] DEVELP_2K (96) [1] BASIN_DEVELOPED (79) [1] DDEVLO_2K (79) [1]
RUS_MEANTEMP (81) RUS_MINTEMP (58) SEASON (52) DIS_WEBS r (35)
RUS_MEANTEMP (81) TEMP (69) RUS_MINTEMP (66) SEASON (52)
DDAIRYOP_5K (96) [2] NUD_CAT_DAI_BA r (96) [2] NUD_DAIRY_OP r (96) [2] NUD_BARN r (96) [2] DBARN_10K (96) [2]
NUD_BARN (96) [2] NUD_DAIRY_OP (96) [2] NUD_CAT_DAI_BA (96) [2] DDAIRYOP_5K (96) [2] DCAT_OR_DA_5K (96) [2]
RUS_MAXTEMP (84) RUS_MINTEMP (83) WEBSTEMP_C (72) DIS_WEBS r (4)
RUS_MAXTEMP (84) RUS_MINTEMP (83) WEBSTEMP_C (72) pH (45)
DDEVLO_10K (76) [3] NUD_DEVHI r (76) [3] DDEVHI_10K (76) [3] BASIN_DEVELOPED (76) [3] DEVELP_5K (34) [3]
BASIN_DEVELOPED (97) [3] DEVELP_5K (97) [3] RUS_TOTALRAIN7D (97) [3] ORP (96) [3] TURBIDITY (96) [3]
DIS_PAY (45) DIS_RUS (43) RUS_TOTALRAIN7D (20) RUS_TOTALRAIN3D (18)
RUS_TOTALRAIN2D (54) DIS_SPC (51) DIS_PAY (51) TOTPHO (45)
STRAHLER (88) [4] TURBIDITY (91) [4] BASIN_FOREST (88) [4] STRAHLER (88) [4] BASIN_CROPLAND r (88) [4] BASIN_CROPLAND (88) [4] NUD_POULBA r (82) [4] BASIN_FOREST (88) [4] DDEVLO_5K (81) [4] DDEVLO_5K (86) [4]
DEVELP_2K r (47) DPAST_2K (38) NUD_PASTURE r (29) DEVELP_5K r (26)
VEGP_2K (85) NUD_PAST_ACC (80) DPAST_ACC_10K (80) DPAST_ACC_5K (69) TURBIDITY (68)
DIS_RUS (32) [5] TOTKN (22) [5] TOTPHO (20) [5] WEBSTEMP_C r (2) [5]
DIS_RUS (89) [5] DIS_OXYGEN_MGL (83) [5] RUS_TOTALRAIN7D (79) [5] DIS_OXYGEN_P (76) [5]
DIS_WEBS (85) DIS_PAY (77) WEBSTEMP_C r (37) DIS_SPC (36)
DIS_WEBS (86) RUS_TOTALRAIN7D (85) DIS_PAY (77) NITRATE (67)
DIS_SPC (26) [6] DIS_RUS (14) [6] RUS_MEANTEMP r (7) [6] RUS_MAXTEMP r (5) [6]
DFORE_10K (93) [6] RUS_TOTALRAIN7D (93) [6] DIS_OXYGEN_MGL (80) [6] CONDUCTIVITY (66) [6]
5/126 (2)a
20/284 (12)a
97/594 (33)a 42/30 (57)a 25/138 (34)a 7/153 (9)a
E. coli O157:H7 (1186) DPAST_5K 0.80 obs. km2 DPAST_5K > 0.80 obs. km2 DPAST_5K > 0.80 obs. km2
NS
1/614 (7)a
TURBIDITY 4.8 NTU [5]
0/152 (0)a
TURBIDITY > 4.8 NTU [5]
14/405 (93)a
Salmonella spp. (1186)
NS
53/116 (59)a
DIS_PAY 1.22 m3 s1 [6]
24/846 (27)a
DIS_PAY > 1.22 m3 s1 [6]
13/134 (14)
DIS_RUS > 11.04 m3 s1 DIS_RUS 11.04 m3 s1 DIS_RUS 11.04 m3 s1
Level 1 Child Node Split Competitor Variables (Improvement Score)c [Child node split code]
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 0 7 e5 8 2 5
Campylobacter spp. (1171)
Level 1 Child Node Split Surrogate Variables (Improvement Score)c [Child node split code]
(continued on next page)
5813
NS ¼ no split. a Target pathogen detections/non-detections (percentage of positive detections). b E. coli O157:H7, Salmonella spp. and Campylobacter spp. number of detections (0 path. detections/1 path. detections/2 path. detections) at a site on a particular date when all three pathogens were sampled together, and percentage of class (0, 1, 2) detections. c Percentages in brackets represent split improvement scores as percentages of optimal nodal split improvement scores. The top five surrogate and competitor variables are presented. In some cases CART provided less than four variables scores. r denotes a variable split in reverse direction.
b
177/107/9 (22%/33%/27%) 82/67/19 (10%/20%/56%) RUS_TOTALRAIN7D 26.7 mm [8] RUS_TOTALRAIN7D > 26.7 mm [8] TEMP 14.1 C
343/58/1 (43%/18%/3%)b 206/97/5 (26%/30%/15%)b RUS_TOTALRAIN7D 42.0 mm [7] RUS_TOTALRAIN7D > 42.0 mm [7]
b
DIS_WEBS r (87) DIS_SPC r (59) WEBSTEMP_C (54) SEASON (31)
DIS_PAY (97) DIS_WEBS (95) RUS_MEANTEMP (94) WEBSTEMP_C (91)
RUS_TOTALRAIN3D (28) DIS_WEBS (25) [7] pH r (23) [7] RUS_TOTALRAIN2D (20) RUS_TOTALRAIN1D (16) DIS_PAY (36) [8] RUS_TOTALRAIN3D (35) RUS_TOTALRAIN1D (25) RUS_TOTALRAIN2D (14)
[7] DDEVLO_5K (54) [7] DDEVHI_10K (51) [7] RUS_TOTALRAIN1D (45) [7] [7] NUD_DEVHI (45) [7] [7] BASIN_DEVELOPED (45) [7] DIS_WEBS (80) [8] [8] RUS_TOTALRAIN3D (64) [8] [8] DIS_PAY (62) [8] [8] DIS_RUS (54) [8]
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 0 7 e5 8 2 5
Site detections TEMP > 14.1 C of E. coli O157:H7, Salmonella spp., & TEMP > 14.1 C Campylobacter spp. on a single day (1171) TEMP 14.1 C
Microorganism Target (Total No. Samples)
Table 2 (continued )
Root Node Split Criteria
Level 1 Child Node Result of Root Node Split Root Node Split Split Criteria Root and Child Surrogate Variables Competitor Variables [Child node split code] Node Splits (Improvement Score)c (Improvement Score)c
Level 1 Child Node Split Surrogate Variables (Improvement Score)c [Child node split code]
Level 1 Child Node Split Competitor Variables (Improvement Score)c [Child node split code]
5814
and limited tree development to a minimum of 10 observations per node (Fig. 2). Main data structural elements are often captured at these tree levels (Lapen et al., 2002; Ouellet et al., 2008). For all CART analyses, a 10 V-fold cross-validation was used for assessing relative cost and relative errors for CART trees grown in sequence. This consisted of a 10% test and 90% learning approach (Breiman et al., 1984). We complimented our heuristic analyses using CART by examining competing and surrogate variable splits for each data split for each tree model. Here, the top five surrogate and competitor variables for each node where splitting occurred were identified and presented with associated split improvement scores represented as a percentage of the optimal nodal split improvement score. Powerful surrogate data split criteria are those that mimic optimal split criteria by sending a similar number of cases into child nodes and are judged on how well they copy the optimal split criteria and not on their ability to reduce nodal impurity, unlike competitor variables (Steinberg and Colla, 1995). Competitor variables are those that produce the next highest improvement scores, relative to optimal splitting criteria improvement score. Surrogate variable splitting analyses help reveal structure and intercorrelation between predictor variables as well as the most robust predictors in the CART generated trees (Lapen et al., 2002). Overall our data mining approach was composed of two main elements: i) the objective characterization of land use attributes for each site (quantitative variables not descriptive variables), and, ii) use of those and other quantitative environmental information in an automated data mining engine to define, ‘objectively’, variable conditions, correlations, and interactions that best discriminate higher and lower pathogen occurrence/density groupings. After the CART tree models were developed, as per above, the most frequently occurring optimal splitting variables, the top four surrogate splitting variables, and the top four competitor variables with improvements scores greater than 50% of the optimal splits were identified. The top eight of these variables were deemed the most important variables for identifying structure in the pathogen data. These ‘important’ or ‘recurrent’ independent variables identified in the data mining exercise were used in binary logistic regression (Statistica 9.0, Statsoft Inc., Tulsa, OK) and CART again to develop pathogen predictive models with more parsimonious input data; models that might be more tractable from an operational standpoint than those generated via exploratory data mining as per the initial CART analyses described. For stepwise binary logistic regression, the top eight variables were submitted to forward stepwise regression using model building significance threshold criteria of a ¼ 0.05 and only variables that were selected meeting this significance level were included in a final predictive model.
3.
Results
3.1.
Hydrology and weather conditions
Water temperature (TEMP) at time of sampling ranged from 0.06 C to just over 29 C and daily air temperature values ranged from 24 C to 31.5 C. Temperature data were highly
Table 3 e Prediction of Cryptosporidium oocyst and Giardia cyst densities from independent variables listed in Table 1 plus the variables SHREVE, STRAHLER, and SEASON in CART regression tree analyses. The results are restricted to two tree levels with a maximum of 4 terminal nodes (Fig. 2). We imposed a CART modeling constraint of having a ‡ 10 observations per node. LS (least squares), LAD (least absolute deviation). Microorganism Target (Total No. Samples)
Regression Tree Split Method
Cryptosporidium
LS
oocysts 100 L1 (664)
LAD
Root Node Split Criteria
Level 1 Child Node Split Criteria [Child node split code]
Result of Root and Child Node Splitsa
Root Node Split Surrogate Variables (Improvement Score)b
Root Node Split Competitor Variables (Improvement Score)b
Level 1 Child Node Split Surrogate Variables (Improvement Score)b [Child node split code]
Level 1 Child Node Split Competitor Variables (Improvement Score)b [Child node split code]
DIS_PAY > 0.03 m3 s1
TURBIDITY 42 NTU [1]
x ¼ 11 28 (595)
DIS_WEBS (68)
DIS_WEBS (76)
ORP (100) [1]
TOTKN(72) [1]
DIS_PAY > 0.03 m3 s1
TURBIDITY > 42 NTU [1]
x ¼ 66 128 (45)
DIS_OXYGEN_P (0)
NUD_PAST_ACC (76)
TOTKN (67 [1])
DIS_PAY (60)
DIS_OXYGEN_MGL (0)
DPAST_10K (76)
DIS_PAY (60) [1]
ORP (60) [1]
DPAST_ACC_10K (76)
CROPLANDP_10K (1) [1]
DIS_RUS (60) [1]
DPAST_5K (76)
DIS_OXYGEN_MGL (0) [1]
FORAGEP_2K (54) [1]
DIS_PAY 0.03 m3 s1
SHREVE 1767 [2]
x ¼ 733 1829 (10)
FORAGEP_5K (100) [2]
BASIN_CROPLAND (100) [2]
DIS_PAY 0.03 m3 s1
SHREVE > 1767 [2]
x ¼ 0 0 (14)
VEGP_5K (100) [2]
BASIN_FORAGE (100) [2]
FORAGEP_10K (100) [2]
BASIN_SHRUBLAND (100) [2]
PASTP_10K (100) [2]
BASIN_FOREST (100) [2]
NUD_PASTURE (71) [2]
DDEVHI_10K(100) [2]
TEMP > 8.55 C
NS
TEMP 8.55 C
DIS_PAY 7.89 m3 s1 [3]
TEMP 8.55 C
3
1
DIS_PAY > 7.89 m s
[3]
m ¼ 0 25 (446)
RUS_MAXTEMP (96)
RUS_MAXTEMP (96)
DIS_RUS (100) [3]
DIS_RUS (100) [3]
m ¼ 14 13 (192)
RUS_MEANTEMP (87)
RUS_MEANTEMP (90)
pH (1) [3]
TURBIDITY (56) [3]
m ¼ 54 39 (26)
RUS_MINTEMP (63)
SEASON (78)
CONDUCTIVITY (1) [3]
RUS_TOTALRAIN1D (52) [3]
DIS_WEBS (34)
DIS_PAY (71)
BASIN_CROPLAND (90)
BASIN_SHRUBLAND (94)
TOTKN (100) [4]
REA_PHOS (92) [4]
BASIN_DEVELOPED (90)
TOTPHO (93)
REA_PHOS (92) [4]
NUD_PASTURE (73) [4]
BASIN_FORAGE (90)
NUD_CAT_OR_DAI_BA (91) BASIN_CROPLAND (1) [4]
NUD_PAST_ACC (73) [4]
pH (34)
DCAT_OR_DA_10K (91)
DPAST_ACC_2K (73) [4]
RUS_TOTALRAIN2D (52) [3] DIS_SPC (51) [3]
Giardia cysts 1
100 L
LS
CONDUCTIVITY
NS
x ¼ 32 39 (18)
1
(664)
0.12 mS cm
CONDUCTIVITY >
TOTPHO 0.45 mg L1 [4]
x ¼ 4 18 (631)
0.12 mS cm1 CONDUCTIVITY >
TOTPHO > 0.45 mg L1 [4]
x ¼ 33 49 (15)
NS
m ¼ 9 10 (49)
ORP(0) [4]
DDAIRYOP_10K (91)
DPAST_2K (73) [4]
0.12 mS cm1 LAD
RUS_MAXTEMP 2.8 C RUS_MAXTEMP >
CONDUCTIVITY
2.8 C
0.13 mS cm1 [5]
RUS_MAXTEMP > 2.8 C CONDUCTIVITY >
m ¼ 14 24 (19)
RUS_MEANTEMP (47)
RUS_MEANTEMP (90)
BASIN_FOREST (34) [5]
BASIN_CROPLAND (34) [5]
RUS_MINTEMP (36)
CONDUCTIVITY (81)
BASIN_PASTURE (34) [5]
BASIN_DEVELOPED (34) [5]
DIS_PAY (26)
TEMP (74)
BASIN_CROPLAND (34) [5]
BASIN_FORAGE (34) [5]
WEBSTEMP_C (11)
RUS_MINTEMP (55)
BASIN_DEVELOPED (34) [5] BASIN_PASTURE (34) [5]
m ¼ 0 4 (596)
0.13 mS cm1 [5]
NS ¼ no split. a x ¼ mean value standard deviation; m ¼ median value mean absolute deviation (total no. of samples in group), all units are oocysts or cysts per 100 L1. b Percentages in brackets represent split improvement scores as percentages of optimal nodal split improvement scores. The top five surrogate and competitor variables are presented. In some cases CART provided less than four variables scores.
5816
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 0 7 e5 8 2 5
Table 4 e Stepwise binary logistic regression analysis predicting pathogen presence and absence from selected independent criteria given in Section 3.5. Independent Variable Coefficientsa and Odds Ratiosb
Microorganism Target
Intercept
SEASON Spring
Summer
SHREVE Fall
TURBIDITY
DIS_PAY
Winter
Cryptosporidium
b0 ¼ 2.11722g
b1 ¼ 1.55169g
b2 ¼ 0.14931
b3 ¼ 0
b4 ¼ 0.00005g
b5 ¼ 0.37144g
detectionc
OR ¼ 8.30800
OR ¼ 0.21189
OR ¼ 1.16103
OR ¼ 1
OR ¼ 0.99995
OR ¼ 1.44982
b0 ¼ 0.27645
b1 ¼ 1.16671g
b2 ¼ 0.08847
b3 ¼ 0
b4 ¼ 0.00002g
b5 ¼ 0.09242g
OR ¼ 1.31844
OR ¼ 0.31139
OR ¼ 0.91533
OR ¼ 1
OR ¼ 0.99998
OR ¼ 1.09683
b0 ¼ 1.84602g
b1 ¼ 1.56598g
b2 ¼ 2.27284g
b3 ¼ 2.04321g
b4 ¼ 0
b5 ¼ 0.00169g
OR ¼ 0.15786
OR ¼ 4.7874
OR ¼ 9.70693
OR ¼ 7.71534
OR ¼ 1
OR ¼ 1.00169
Giardia detectionc Campylobacter detectiond L. monocytogenes
b
Detectiond Salmonella spp. detection
0
¼ 7.41858
OR ¼ 0.0006 b
0
b1 ¼ 5.27750
b2 ¼ 0.69751
b3 ¼ 2.71018
b4 ¼ 13.00575
b5 ¼ 0.02069
OR ¼ 195.88019
OR ¼ 2.00874
OR ¼ 15.03202
OR ¼ 0.000002
OR ¼ 1.02090
¼ 4.31495g
OR ¼ 0.01337
b1 ¼ 0.42540g OR ¼ 1.53020
a Logistic regression probability of presence equation: y ¼ exp(b0 þ b1 X1. bn Xn)/(1 þ exp(b 0 þ b1 X1. bn Xn)). A positive coefficient indicates an increase in the predicted logged odds of pathogen presence and a negative coefficient indicates a decrease in the predicted logged odds of pathogen presence. b Independent variable odds ratios represent the factor by which the odds change for the dependent variable given a unit increase in the independent variable, where values >1 indicate an increase in odds of pathogen presence, and values < 1 indicate a decrease in the odds of pathogen presence when all other independent variables are held constant. c Cryptosporidium and Giardia oocysts and cysts were collected in three seasons (spring, summer and fall), represented by three Beta coefficients, b1, b2, and b3, where the three betas were multiplied against factors Spring X1 ¼ 1, X2 ¼ 0, X3 ¼ 0; Summer X1 ¼ 0, X2 ¼ 1, X3 ¼ 0; Fall X1 ¼ 0, X2 ¼ 0, X3 ¼ 1. d Campylobacter and L. monocytogenes samples were collected in four seasons (spring, summer, fall and winter), represented by four Beta coefficients, b1, b2, b3, and b4, where the four betas were multiplied against factors Spring X1 ¼ 1, X2 ¼ 0, X3 ¼ 0, X4 ¼ 0; Summer X1 ¼ 0, X2 ¼ 1, X3 ¼ 0, X4 ¼ 0; Fall X1 ¼ 0, X2 ¼ 0, X3 ¼ 1, X4 ¼ 0; Winter X1 ¼ 0, X2 ¼ 0, X3 ¼ 0, X4 ¼ 1. e Kappa statistics range from 0 to 1. f The model OR’s presented are the product of making a correct prediction over the product of making an incorrect prediction. g a < 0.05.
correlated (See Table S3 and S4 of the appendix). Mean water temperature, pH, electrical conductivity, dissolved oxygen, and oxidation reduction potential were all higher (positive) at sites situated on river systems (Shreve 15499), whereas mean turbidity and nutrient concentrations were higher in ditch and small stream networks (Shreve < 95), with the exception of nitrate, which was higher in intermediate tributaries (Shreve 95 and <15499). Discharges from the different gauging stations were correlated amongst each other, with Spearman correlation coefficients ranging from 0.76 to 0.90 (Table S4 of the appendix). Correlations among the rainfall and discharge variables were generally positive but small or insignificant (Table S4).
3.2. Pathogen detection rates and subtypes found in surface water samples Between 395 and 1171 surface water samples were analyzed between 2004 and 2008 for a variety of bacterial pathogens and parasites (Table S5 and S6 of the appendix). At sample filter volumes of 500 mL, Campylobacter spp. were detected more often (25% of tested water samples) than Salmonella spp. (8%) and E. coli O157:H7 (1%) (Table S5). Eighty four percent (n ¼ 127) of the 151 Campylobacter spp. isolated from water were C. jejuni, and 14% (n ¼ 21) were C. coli. Salmonella spp. serotypes
identified in the water samples included Salmonella ser. Muenster, Salmonella ser. Kentucky, Salmonella ser. Bovismorbificans, Salmonella ser. Heidelberg, Salmonella ser. Typhimurium, Salmonella ser. Give, and Salmonella ser. Thompson. The order of occurrence of Campylobacter spp., Salmonella spp., and E. coli O157:H7 found in this study is similar to results found by Walters et al. (2007) in river water in Alberta, Canada. At sample volumes collected that were lower than those for the aforementioned bacterial pathogens, L. monocytogenes was detected in 19% of the 100 mL sample volumes. Mean densities of Cryptosporidium oocysts were 25.0 oocysts 100 L1 and mean densities of Giardia cysts were 5.5 cysts 100 L1 (Table S6).
3.3. Variables associated with pathogen presence and absence in surface water Table 2 presents CART classification tree results associated with predicting pathogen presence and absence in water from a suite of independent environmental variables. Air temperature, water temperature, and season dominated the root node split criteria for 5 out of 7 pathogen targets modeled (Table 2). CART data mining found that air temperature splits partitioned greater detections of Cryptosporidium and Giardia when RUS_MAXTEMP values were 15.3 C and 14.8 C, respectively (72% of detections occurred at, or below, the 37
5817
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 0 7 e5 8 2 5
Table 4 e Extended. Independent Variable Coefficientsa and Odds Ratiosb RUS_ MEANTEMP
b
6
BASIN_ CROPLAND
NUD_ BARN
¼ 0.13762g
Observed Detections: Correct Predictions: Incorrect Predictions
Observed Non-Detections: Correct Predictions: Incorrect Predictions
Percentage of Detections Correct: Percentage of Non-Detections Correct
Kappa Statistic (K)e
Model Odds Ratio (OR)f
b 7 ¼ 0.00001g
306: 201:105
358: 307: 51
66%: 86%
0.52
11.52
170: 50: 120
494:465: 29
29%: 94%
0.29
6.68
294: 32: 262
877: 850: 27
11%: 97%
0.11
3.85
74: 41: 33
321: 315: 6
55%: 98%
0.62
65.23
90: 12: 78
1096: 1084: 12
13%: 99%
0.19
13.90
OR ¼ 0.87143
OR ¼ 0.99999
b6 ¼ 0.10013g
b 7 ¼ 0.00002g
OR ¼ 0.90472
OR ¼ 1.00002
b6 ¼ 0.09855g
b7 ¼ 0.00001g
OR ¼ 0.90615
OR ¼ 1.00001
b6 ¼ 0.18830g
Model Results
NUD_ PASTURE
OR ¼ 1.20720 b2 ¼ 0.06007g OR ¼ 1.06191
and 35 temperature percentiles) (Table 2). Thirty-five percent of Campylobacter spp. detections were found when RUS_MEANTEMP was 7.7 C (29 percentile for temp.). For Cryptosporidium oocyst, Giardia cyst and Campylobacter spp. detection, a vast majority of surrogate and competitor split criteria associated with the root node temperature variables were other air temperature “type” variables suggesting that air temperature was a robust and or unique classifier of pathogen presence and absence at that initial data split level. Split definitions of the level 1 child nodes for the parasite detections appeared to be less unique, as demonstrated in surrogate and competitor variables having relatively strong improvement scores. But unlike root node split criteria, land use variables were more important for level 1 child node splits, where typically greater detections of parasites occurred where non-agricultural practices were less predominant upstream of the sampling sites. For Campylobacter spp. level 1 child node splits, there were generally greater percentages of non-detections at sites where upstream non-agricultural land use was higher. The root node split for L. monocytogenes indicated that 57% of detections occurred in spring, relative to 43% detections in summer, fall and winter combined. The surrogate and competitor variable splits for the root node were related to discharge and rainfall; however, they had modest improvement scores relative to SEASON. The level 1 child node split for
L. monocytogenes was based on SHREVE stream order: 9% of the detections occurred in the summer, fall, and winter for streams with SHREVE orders >1895.5, versus detection rates of 34% at orders 1895.5, respectively. Competitor and surrogate variables for the level 1 child node splits were modestly high in terms of improvement scores, but variable types were mixed (e.g., water physical and chemical, hydrological, land use). Fourteen of the 15 detections of E. coli O157:H7 were associated with sites with DPAST_5K > 0.8 observations per km2 (root node split) and with TURBIDITY > 4.8 NTU (level 1 child node split). The mean turbidity for all sites was 26.9 NTU and the mean DPAST_5K was 0.95 observations per km2. Root and level 1 child node surrogate variables were weak in terms of improvement scores, but root node competitors (primarily pasture related variables) and child node competitors (discharge, rainfall, and dissolved oxygen) had high scores in terms of optimal split improvement scores. Surface water discharge was found to be associated with Salmonella spp. detections, with 59% of the detections occurring when DIS_RUS > 11.04 m3 s1 (i.e. discharge above the 83 percentile), versus 41% of the detections when DIS_RUS 11.04 m3 s1. The discharge split appeared to be a unique root node splitting criteria since surrogate and competitor split variables were largely associated with discharge variables. A vast majority of Salmonella spp. non-detections occurred
5818
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 0 7 e5 8 2 5
Table 5 e Prediction of Cryptosporidium oocyst and Giardia cyst densities resulting from the use of the select independent variables listed in Section 3.5 in CART regression tree analyses. Trees presented here are least square regression trees with the lowest cross validated relative error in the CART selection process, and are not limited to CART tree size constraints associated with results in Tables 2 and 3. A CART modeling constraint of having ‡ 10 observations per node was imposed however. Microorganism Target (Total No. Samples)
Root Node Split Criteria
Cryptosporidium oocysts 100 L1 (664)
DIS_PAY DIS_PAY DIS_PAY DIS_PAY
Giardia cysts 100 L1 (664)
NUD_BARN 5 km NUD_BARN > 5 km NUD_BARN > 5 km
> >
0.03 m3 0.03 m3 0.03 m3 0.03 m3
s1 s1 s1 s1
Level 1 Child Node Split Criteria
Result of Root and Child Node Splits Mean Value Standard Deviation (Total No. of Samples in Group).
SHREVE 1767 SHREVE > 1767 TURBIDITY 42 NTU TURBIDITY > 42 NTU
733 1829 (10) 0 0 (14) 11 28 (595) 67 128 (45)
NS SHREVE 43.5 SHREVE > 43.5
3 9 (528) 4 10 (52) 20 51 (84)
NS ¼ no split.
during situations when both DIS_RUS and DIS_PAY (level 1 child node split variable) were equal to or below the respective discharge thresholds of 11.04 and 1.22 m3 s1 (Table 2). The percentile of the DIS_PAY split value of 1.22 m3 s1 was 72. Non-detections of any of the three pathogens Salmonella spp., Campylobacter spp., and E. coli O157:H7 together were higher when TEMP > 14.1 C relative to TEMP 14.1 C (68% of total non-detections above, 32% of total non-detections were equal to or below this root node split variable). Average water temperature during the study was approximately 15 C (Table S3). Moreover, 71% of the detection of 2 or more of these pathogens together occurred when RUS_TOTALRAIN7D >42 mm and RUS_TOTALRAIN7D >26.7 mm, above and below this TEMP threshold (14.1 C), respectively (percentiles of >62 and >77 for this 7 day cumulative rainfall variable). In fact, 50% of detections of one pathogen, and 71% of detections of 2 pathogens occurred for the relatively higher rainfall conditions for the (water) TEMP split classes. Dominant root node surrogates and competitors were associated most strongly with variables expressing discharge and air temperature (high relative improvement scores). Dominant level 1 child node surrogate (weak relative improvement scores) and competitor variables (moderately high relative improvement scores) were linked most strongly to rainfall and discharge. Hence, hydrological/weather variables appeared to be most robust and unique in terms of independent variable classifiers.
3.4. Variables associated with parasite densities in surface water The CART regression tree based results for Cryptosporidium oocyst and Giardia cyst densities are given in Table 3. The LAD results for Cryptosporidium indicated that nodal median densities 14 oocysts per 100 L1 were found when TEMP was 8.55 C (root node split), compared to a nodal median value of 0 oocysts per 100 L1 when TEMP was >8.55 C (TEMP of 8.55 C ¼ 24 percentile). The highest median density nodal group for LAD results (54 oocysts per 100 L1) was associated with TEMP 8.55 C and when DIS_PAY was >7.89 m3 s1 (Level 1 child node split; > 93 percentile for DIS_PAY). The LS
method identified DIS_PAY (root node) and SHREVE (Level 1 child node) as data split criteria for Cryptosporidium, with highest nodal mean oocyst densities (733 oocysts per 100 L1) associated with DIS_PAY <0.03 m3 s1 and SHREVE 1767. Yet mean oocyst densities of 11 oocysts per 100 L1 of water were associated with DIS_PAY >0.03 m3 s1 (>4 percentile for DIS_PAY). Temperature related variables (with high relative improvement scores) dominated root node split surrogate and competitor variables for Cryptosporidium LAD results; and DIS_RUS was similar to the level 1 child node splitting variable DIS_PAY for Cryptosporidium LAD results in terms of child node surrogate and competitor improvement scores. For Cryptosporidium LS results, several land use based level 1 child node surrogate (primarily forage and pasture cover variables) and competitor (primarily basin scale coverage variables) variables had improvement scores equal to that for the stream order based level 1 child node split criteria. Median Giardia cyst densities for the LAD based results were highest when RUS_MAXTEMP > 2.8 C and CONDUCTIVITY 0.13 mS cm1 (the study means were 15.4 C and 0.781 mS cm1, respectively). A significant proportion of observations where median densities were 0 cysts 100 L1 were found when RUS_MAXTEMP > 2.8 C and when CONDUCTIVITY > 0.13 mS cm1. Temperature dominated surrogate and competitor split variables for root node splits suggesting robustness of the temperature variables as splitters. Level 1 child node surrogate and competitor variable splits had weak improvement scores, relative to CONDUCTIVITY. For the Giardia LS results, CONDUCTIVITY (0.12 mS cm1) stratified root node Giardia density data, but at the level 1 child node level, TOTPHO parsed out higher and lower mean Giardia cyst densities. Root node surrogate and competitors were of mixed type, but were overall strong in terms of relative improvement scores. Level 1 child node surrogate and competitor variable splits had strong affinities with water chemical status (e.g., phosphorus and nitrogen concentrations), and to a lesser degree, land use. Spearman correlation coefficients between RUS_MAXTEMP and Cryptosporidium oocyst and Giardia cyst densities were 0.54 and 0.39 respectively, complimentary to these CART findings
NS ¼ no split. a E. coli O157:H7, Salmonella spp. and Campylobacter spp. number of detections (0 path. detections/1 path. detections/2 path. detections) at a site on a particular date when all three pathogens were sampled together, and percentage of class (0, 1, 2) detections. b Percentages in brackets represent split improvement scores as percentages of optimal nodal split improvement scores. The top five surrogate and competitor variables are presented. In some cases CART provided less than four variables scores.
SHREVE (75) [1] TURBIDITY (71) [1] DIS_PAY (53) [1] SEASON (53) [1] NUD_BARN (45) [1] SEASON (53) [1] DIS_PAY (20) [1] e e e RUS_MEANTEMP (97) TURBIDITY (39) SEASON (35) SHREVE (29) PER_CROPLAND (19) 130/98/24 (16%/30%/71%)a RUS_MEANTEMP (60) SEASON (35) RUS_MEANTEMP 8.2 C [1] 78/54/3 (10%/16%/9%)a TURBIDITY (35) NUD_BARN (19) RUS_MEANTEMP > 8.2 C [1] 600/177/7 (74%/54%/21%)a SHREVE (9) NS
Number of Observed DIS_PAY > Pathogens in Water 1.41 m3 s1 Sample (1171) DIS_PAY 1.41 m3 s1 DIS_PAY 1.41 m3 s1
Level 1 Child Node Split Competitor Variables (Improvement Score)b [Child node split code] Level 1 Child Node Split Surrogate Variables (Improvement Score)b [Child node split code] Root Node Split Competitor Variables (Improvement Score)b Root Node Split Surrogate Variablesb Result of root and child node splits Root Node Level 1 Child Node Split Split Criteria Criteria [Child node split code] Microorganism Target (Total No. Samples)
Table 6 e Prediction of the number of observed pathogenic bacteria (E. coli O157:H7, Salmonella spp., and Campylobacter spp.) in surface water using the independent variables selected in Section 3.5, in CART classification tree analysis. CART tree size constraints associated with results in Tables 2 and 3 were not imposed. A CART modeling constraint of having ‡ 10 observations per node was imposed however.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 0 7 e5 8 2 5
5819
(a ¼ 0.05; Table S4 in appendix). Yet, global correlations among parasite densities and TOTPHO and CONDUCTIVITY were, for the most part, insignificant statistically.
3.5. Recurrent variables related to pathogen density and presence and absence in surface water For purposes of parsimony, a few of the most important or recurrent independent variables from the CART analyses conducted (from Tables 2 and 3), were selected for further predictive analyses. We examined a subset of variables for the purposes of application in a more operational environment, with a less exploratory purpose and a more parsimonious and applied context. Selection was based on popularity and improvement score strength as primary, surrogate, and competitor variables in the data mining analyses for all pathogens. However, we constrained variable selection in terms of broad variable classes: i) physical and chemical attributes of water, ii) stream order, iii) hydrology-weatherclimate, iv) land use coverage and proximity and, v) season, for the CART results given in Tables 2 and 3. This subset of selected variables consisted of, respectively in terms of these classes, i) TURBIDITY, ii) SHREVE, iii) DIS_PAY (a gauging station of intermediate stream order of the 4 gauging stations examined) and RUS_MEANTEMP, iv) BASIN_CROPLAND, NU_BARN, NU_PASTURE, and v) SEASON. The other variables listed in Table 1 were not considered here as they were either, insignificant, redundant, or not important within the CART analyses summarized in Tables 2 and 3.
3.6. Using select independent variables to predict pathogen density and presence-absence in surface water The independent variables identified in Section 3.5 were submitted to a forward stepwise binary logistic regression to predict pathogen presence and absence in water. Logistic regression is another more commonly used statistical tool for globally predicting binary data from continuous and categorical independent criteria. Further, CART, within the least squares regression tree framework, was used to predict Cryptosporidium oocyst and Giardia cyst densities in water from the select independent variables in Section 3.5. Due to their potential use as a broad indicator of pathogen exposure, predictions of the detections of multiple pathogens together (Salmonella spp., Campylobacter spp., E. coli O157:H7), were carried out using CART in classification tree mode since the target dependent variable was multinomial. Tree constraining to two levels, as presented in Tables 2 and 3, was not employed in the parasite and multiple pathogen CART analyses described in this Section 3.6: rather, optimal CART trees were selected using unconstrained tree sizes with the lowest cross validated relative error, and limiting terminal node sizes to a minimum of 10 observations. Due to redundancy, we did not conduct individual CART analyses using the select independent variables, in this manner, for Campylobacter spp., L. monocytogenes, E. coli O157:H7, and Salmonella spp. For logistic regression results, Kappa statistics ranged from 0.11 to 0.62 (Table 4). Model odds ratios were greater
5820
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 0 7 e5 8 2 5
than approximately 4. Positive coefficients in Table 4 indicate an increase in predicted logged odds of pathogen presence, whereas a negative coefficient indicates a decrease in the predicted logged odds of pathogen presence. Where selected and included in the model, increases in TURBIDITY and DIS_PAY were associated with an increase in predicted logged odds of pathogen presence, while an increase in SHREVE was associated with a decrease in predicted logged odds of pathogen presence. Overall, SEASON, RUS_MEANTEMP, and land use variables had mixed associations among the different pathogens. In all logistic regression equations for target pathogens, RUS_MEANTEMP was a significant independent variable (a < 0.05), with an increase in temperature associated with a decrease in predicted logged odds of parasites and Campylobacter spp., while reverse trends for Salmonella spp. and L. monocytogenes were observed. The land use variables in Table 4 were not strong predictors. The application of CART LS regression trees on Cryptosporidium and Giardia densities resulted in models with r2 of 0.13 and 0.07, respectively (Table 5). For Cryptosporidium, the model presented in Table 3 was the optimal model defined within the context of the selected independent variables. For Giardia, tree model independent variables were NUD_BARN (positively correlated root node split) and SHREVE (positively correlated level 1 child node split). Prediction of the occurrence of multiple pathogens (Campylobacter spp., Salmonella spp., E. coli O157:H7) using these selected independent variables, is given in Table 6. A vast majority of 2 incidences of these pathogens occurs when discharge is relatively higher (DIS_PAY > 1.41 m3 s1, >75 percentile for this discharge variable). Only 16% of the nondetections of these pathogens occur during these discharge conditions. Greatest occurrence of non-detections and detections of only one pathogen, occur during lower relative DIS_PAY discharge, and when RUS_MEANTEMP 8.2 C (31 percentile for this temperature variable).
4.
Discussion
The presence of enteric bacteria in the aquatic environment depends on a variety of parameters, which include nutrients and temperature (Leclerc et al., 2002). The nutrient variables were not, overall, strong or recurrent predictors of individual pathogen occurrence in surface water, however, air temperature variables were more important in this capacity. Overall, there were generally greater detections of Cryptosporidium and Giardia when maximum daily air temperatures were below approximately 15 C (35 percentile of maximum temperature). This corresponds to mean daily air temperatures less than 9.5 C (35 percentile) which occurs predominately during fall sampling, and to a lesser extent, in the spring, hence SEASON was a modest root node split surrogate variable (Table 2). Thus, the temperature relationships could be related, in part, to seasonality associated with off field/farm transport of agricultural contaminants (and/or stream/riparian zone sediment mobilization) during the cooler and wetter fall periods (Bodley-Tickell et al., 2002; Ho¨rman et al., 2004; Ruecker et al., 2007). In support of such findings,
Cryptosporidium andersoni, sourced to mature cattle, occurred ubiquitously in the study area during the fall time frame (Ruecker et al., 2007). Fall and spring are seasons when frequent land application of manure occurs in the area. But high densities of Cryptosporidium oocysts were also associated with low relative discharge and relatively smaller stream orders (Table 3). In low discharge or stagnant water conditions, localized parasite inputs on smaller tributaries could concentrate if flow conditions are not sufficient enough to dilute, or move resident or sequestered parasites out of the watercourses. Both Giardia and Cryptosporidium can be derived from wildlife in the study area (Ruecker et al., 2010). Wildlife (e.g., muskrats, voles et cetera) has been observed to proliferate in very modest riparian corridors along agricultural drainage ditches of smaller stream order, and these sources may have contributed to some exceptionally high oocyst densities for a small number of samples (Table 3). Interestingly, Giardia densities in both the CART LS and LAD models (Table 3) were stratified by physical and chemical water attributes; discharge variables were effectively irrelevant statistically. Inputs of agricultural nutrients to surface water can result in higher surface water conductivity (Hunsaker and Levine, 1995). For Campylobacter spp., the ratio of detections:nondetections when mean air temperatures were respectively below and above w8 C, were 1.1 and 0.24 (Table 2), and logistic regression indicated significant season and temperature effects. Yet physical effects of temperature on survival and recovery of Campylobacter spp. likely interplays with the time of year when surface water exposure risks may be elevated (fall and spring are seasons when manure is often applied to fields and surface water systems are flowing). This is supported to some degree by occurrences of Campylobacter spp. during higher threshold air temperatures (>8 C approximately) where density of developed lands were relatively smaller (detection:non-detection ratio ¼ 0.49); thereby supporting contentions of possible agricultural sources of Campylobacter spp. Miller and Mandrell (2005) indicate that Campylobacter spp. surface water survival rates decreased significantly between 10 and 20 C water temperature, and Altekruse et al. (1999) indicate that the recovery of Campylobacter is greatest in the cooler seasons, and survival in cold water is important for their life cycle. Greater occurrence of these pathogens at “cooler” temperatures has been demonstrated for other pathogens (Rollins and Colwell, 1986; DeRegnier et al., 1989; Payment et al., 2001; King and Monis, 2007). Overall, lower pathogen presence when maximum air temperature thresholds were below 14e15 C occurred in this study, which compliment water temperature vs. die-off/inactivation rates in water given in Auer and Niehaus (1993). For instance, the >14e15 C RUS_MAXTEMP threshold (percentiles of > 33 and 35%) equates to mean water temperatures of 18 C, whereas below 14e15 C RUS_MAXTEMP values, mean water temperatures were 6 C. Inactivation equations are used in practice to model microorganism concentrations/loads (and therefore die-off) to, and within, surface water systems (Bowie et al., 1985; Sadeghi and Arnold, 2002). Surface water discharge was a dominant variable in Salmonella spp. presence and absence data mining results (globally, Salmonella spp. presence was positively correlated with discharge). Yet it is unclear as to why discharge was
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 0 7 e5 8 2 5
a more robust predictor of Salmonella spp., relative to the other independent variables at the primary root node split level. Salmonella spp. was not as commonly detected relative to all the other pathogens outside of E. coli O157:H7. Hence, it may be that in order for this pathogen to occur (or be detected) in surface water, threshold hydrological events that promote surface water loading must occur (e.g., to promote off farm/ field transport and sediment mobilization in stream and riparian zone). In fact the DIS_RUS root node split threshold of approximately 11 m3 s1 (Table 2), is above mean study period discharge at that hydrological monitoring site (Table S3). Baudart et al. (2000) indicate in a watershed characterized by agriculture and forest that Salmonella spp. loads increased with mean discharge rates, and they suggest that sediment trapped bacteria can be resuspended during turbulent water events. Winfield and Groisman (2003) indicate that Salmonella spp. frequently occurs in soil samples from agricultural areas and that it can withstand a wider variety of stresses relative to, for example, E. coli. Thus runoff, agricultural drainage, sediment re-suspension and other processes associated with high surface water discharge events seem to promote Salmonella spp. occurrence in surface waters in this study environment. Lyautey et al. (2007) found an association among the occurrence of L. monocytogenes with water sampling proximity to dairy farming, and degree of cropped land upstream of a sampling site where manures are frequently applied. The seasonal disposition of L. monocytogenes detection (higher numbers in spring, a period not extensively monitored by Lyautey et al.) identified in this study could be related to dairy cattle feeding on silage indoors in winter, and associated stored manure being applied to fields in the spring. Silage has been reported as a risk for L. monocytogenes infection (Low and Donachie, 1997; Esteban et al., 2009) and Guerini et al. (2007) found that Listeria spp. are more prevalent on cow hides during the winter-spring, relative to summer-fall. For seasons other than spring, L. monocytogenes was detected more frequently in smaller tributaries relative to larger ones, with over a three-fold increase in detections at SHREVE 1895.5. This trend might be explained by smaller order systems in this study being water bodies that typically receive, more directly, agricultural drainage and runoff from adjacent dairy farming operations where livestock crop fields typically receive application of manure (Lyautey et al., 2010). For E. coli O157:H7 in this study, we found that 93% of the detections were observed when densities of pasturing observations in catchment areas 5 km upstream of the sample site were relatively higher, and the sampled water had a turbidity >4.8 NTU. In fact 3 out of the 15 detections of E. coli O157:H7 were at sample sites located in a pasture system where cattle have direct access to the water course. Direct fecal release in streams, and cattle in streams concurrently augmenting water turbidity, was observed on many occasions for pasture systems where animal access to the watercourses were unrestricted. In addition, the turbidity associations could be linked to runoff/drainage/sediment re-suspension events that promote surface water contamination (Effler et al., 2001; Walters et al., 2007). Recently, effort has been placed on reviewing and improving ambient water quality criteria for inland waters in
5821
North America (Edge et al., 2009; WERF, 2009). An identified knowledge gap is the role of non-point sources of enteric pathogens in surface water systems. In the context of this work, rainfall variables were not found to be, relative to discharge variables, as systematically recurrent or strong in relation to pathogen occurrence when predicted on a pathogen by pathogen basis. Yet waterborne outbreaks have been empirically associated with rainfall events in the United States (Curriero et al., 2001). In an attempt to characterize conditions/ places that are associated with multiple occurrences of different bacterial pathogens, we found that the presence of 2 pathogens in water (including only Campylobacter spp., Salmonella spp. and E. coli O157:H7 in analysis (Table 2)) was predominantly higher (82% of 2 pathogens detected) when water temperatures were 14 C. More succinctly, a majority of 2 pathogen detections were associated with conditions when cumulative rainfall over a week exceeded 27 mm during relatively cooler water temperature conditions (<14 C in primarily winter, spring, and fall water temperatures), and when cumulative weekly rainfall exceeded 42 mm during relatively warmer water temperatures (>14 C, primarily in the summer). And moreover, land use variables were not found to be important predictors in this capacity; which underscores the impact of hydrological drivers and temperature/seasonal elements with regards to pathogen detection risk in these agriculturally dominated surface water systems. Perhaps in less mixed land use systems with more complex topography, land use elements would be more prominent as predictors of pathogen presence and absence. Smith et al. (2001) identified associations between urban land cover, agricultural land on steep slopes, and water quality impairment. Discharge, perhaps a more direct measure of surface water response to contaminant loading (or lack of it) from adjacent land impacted by fecal pollution and/or suspension of contaminated sediment in the stream zone proper, was found to be more consistently linked to parasite densities as well as occurrence of pathogenic bacteria in this study (for individual pathogen models). This finding may be partially due to the disposition of the study site examined here, in that it is relatively flat and under heavy influence of field tile-drainage. Hence, for certain specific pathogenic targets in water, rainfall indices may not accurately reflect the occurrence of significant polluting events associated with tile-drainage networks in spring and fall (Lapen et al., 2008). Moreover, antecedent conditions such as soil water contents (on cultivated land) will impact directly surface runoff-subsurface flow potential and groundwater inputs to surface water systems. All of this evidence is suggestive that there is a critical requirement, within a risk assessment framework, to couple mechanistic watershed scale hydrologic models (Sadeghi and Arnold, 2002; Ferguson et al., 2007) that predict explicitly discharge and ‘antecedent’ hydrological conditions (i.e., a warning system regarding fecal pollution risk) in the landscape, with transport and fate of microorganisms.
5.
Summary and conclusion
This research identified, using data mining approaches, environmental and land use variables associated with
5822
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 0 7 e5 8 2 5
pathogen occurrence and densities in surface waters within multiple mixed-use, but agriculturally dominated, watersheds in eastern Ontario Canada. This research provided, within an exploratory framework, definition of some spatial and temporal pathogen ‘hotspots’ that could be used to help inform risk management strategies and identify dominant fecal pollution driving mechanisms. Specific findings of this work follow: Restricting cattle access to streams could reduce exposure to E. coli O157:H7 in surface water since a relatively high percentage of observations of this pathogen occurred in a stream open to pasturing livestock. One of the better surface water discharge indicators of pathogen densities or occurrence was derived from a gauging station on a river of intermediate stream order (Shreve order ¼ 3243), in the context of the suite of stream orders examined in this study (e.g., watercourses ranging from river stem (Shreve order ¼ 50,855) to drainage ditches (Shreve order ¼ 17)). Thus, for other studies where stream discharge may not be measured at all sample sites within a common climate-physiographic region, discharge from surface waters of intermediate order may be most appropriate in terms of representing discharge for these similar types of studies or modeling efforts. The ratio of detections:non-detections of Campylobacter spp., was relatively higher (>1) when mean air temperatures in the region were roughly below study period average values, relative to when air temperatures were roughly above mean air temperature values (ratios < 1). Cooler temperatures can augment out of host pathogen survival, and also represent seasons coherent with land applications of manure and heightened contaminant transport (e.g., spring and fall). Surface water discharge was the most robust and unique variable predicting Salmonella spp. Fifty nine percent of the detections occurred when river discharge (stream order ¼ 9550) was >83 percentile. It appears threshold hydrological events that promote off farm/field transport (and/or stream and riparian sediment transport) must occur for detection of Salmonella spp. in surface water in this region. The highest detection:non-detection ratio for L. monocytogenes was in spring, relative to other seasons (57% detection rate in spring). Seasonality was a robust and unique prediction criterion for L. monocytogenes. It was speculated that a combination of winter livestock housing, silage feeding during winter, and spring application of manure that accrued during winter, likely contributed to elevated occurrences of this pathogen in spring. Cryptosporidium and Giardia oocyst and cyst densities were globally positively associated with surface water discharge variables, and globally negatively associated with air/water temperature variables during spring-summer-fall. Yet, some of the highest Cryptosporidium oocyst densities were found to be associated with low discharge conditions (4 percentile) on small stream orders. Cursory evidence from other ongoing parasite studies suggests that the source of Cryptosporidium parasites during these low discharge conditions may be predominately wildlife, relative to livestock.
A total of 82% of the 2 detections of bacteria pathogens (Campylobacter spp., Salmonella spp., and E. coli O157:H7) in water was associated with lower water temperatures (w14 C) (primarily the spring and fall). Fifty six percent of the 2 detections occurred during these cooler water conditions when total rainfall the week prior to sampling was >w27 mm (>62 percentile). During higher water temperatures (>w14 C) (primarily summer) and after a higher amount of weekly rainfall (w>42 mm of rainfall (>77 percentile)), was linked to 15% of the detections of 2 pathogens. Less rainfall may have been necessary to mobilize pathogens from adjacent land (or in stream/riparian zone) during cooler water conditions; as these are times when soil water contents and water table depths are relatively higher (quicker off field transport response times to rainfall), and when land application of manures typically occur in the region. Moreover, cooler water temperatures can promote survival of pathogens in water. Overall, season, stream order, turbidity, mean daily temperature, surface water discharge, cropland coverage, nearest upstream distance to a barn and pasture were variables that were found, within respective variable classes, to be relatively strong and recurrent variables structurally associated with pathogen presence/absence and parasite densities in surface water. Yet while relationship directions and linkages among these environmental/land use variables and pathogen occurrence/densities were not always consistent among the individual pathogens, there were general tendencies toward higher detections/densities during relatively cooler air/water temperatures during primarily spring-fall seasons, relatively higher surface water discharge conditions, and higher water turbidity. Land use variables were not, overall, as strong as the weather-climate, seasonal, and hydrological predictor variables in terms of predicting pathogens, but tendencies were in the direction of greater pathogen occurrence/densities where pasture densities were relatively higher and dairy operations were relatively closer to the sample site. Environmental information such as discharge, antecedent soil water contents, and temperature will enhance risk assessment efforts considerably, as surface runoff and associated discharge events are dependent on these factors and can be adequately predicted via hydrological models. However, coupling of pathogen fate and transport in such hydrological models, as well as the integration of fecal pollution sources, are currently limited.
Acknowledgments This study was funded by the Agriculture Policy-Framework’s National Water Quality Surveillance Research Initiative through an agreement between Health Canada and Agriculture and Agri-Food Canada, Environment Canada, and the Public Health Agency of Canada; and Agriculture and AgriFood Canada’s Watershed Evaluation of Beneficial Management Practice (WEBs) program. We would like to thank Diane Medeiros for helping to facilitate this research. We also thank the South Nation Conservation Authority for assistance in obtaining samples, as well as M. Edwards, P. Bolton-Russell,
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 0 7 e5 8 2 5
and N. Gottschall for their excellent technical assistance and support. We would also like to thank Steven Mutschall, Allison Kindt, Anita James, Susan Ross, Rommy Rodriguez, and Garrett Kennedy of the Public Health Agency of Canada in Lethbridge, Alberta for their excellent assistance with laboratory analyses; Linda Cole, Betty Wilkie, Ketna Mistry, and Ann Perets of the OIE Salmonella Reference Laboratory of the Public Health Agency of Canada in Guelph, Ontario for providing serotyping results; and Irene Yong and Nina Enriquez of the E. coli O157:H7 Reference Laboratory of the Public Health Agency of Canada in Guelph, Ontario for providing serotyping results.
Appendix. Supplementary data Supplementary data associated with this article can be found in online version at doi:10.1016/j.watres.2011.06.021.
references
AAFC (Agriculture and Agri-Food Canada), 2010. Interpolated Census of Agriculture to Soil Landscapes, Ecological Frameworks and Drainage Areas of Canada (WSCSSDA Id 02LB for 2006). Agriculture and Agri-Food Canada, Ottawa. Ahmed, W., Neller, R., Katouli, M., 2005. Evidence of septic system failure determined by a bacterial biochemical fingerprinting method. J. Appl. Microbiol. 98, 910e920. Altekruse, S.F., Stern, N.J., Fields, P.I., Swerdlow, D.L., 1999. Campylobacter jejuni- an emerging foodborne pathogen. Emerging Infect. Dis. 1 (5), 28e35. Auer, M.T., Niehaus, S.L., 1993. Modeling fecal coliform bacteria I. Field and laboratory determination of loss kinetics. Wat. Res. 27 (4), 693e701. Baudart, J., Grabulos, J., Barusseau, J.P., Lebaron, P., 2000. Salmonella spp. and fecal coliform loads in coastal waters from a point vs. nonpoint source of pollution. J. Environ. Qual. 29, 241e250. Benham, B.L., Baffaut, C., Zeckoski, R.W., Mankin, K.R., Pachepsky, Y.A., Sadeghi, A.M., Brannan, K.M., Soupir, M.L., Habersack, M.J., 2006. Modeling bacteria fate and transport in watersheds to support TMDLs. Trans. ASABE 49 (4), 987e1002. Bodley-Tickell, A.T., Kitchen, S.E., Sturdee, A.P., 2002. Occurrence of Cryptosporidium in agricultural surface waters during annual cycle in lowland UK. Water Res. 36, 1880e1886. Bowie, G.L., William, W.B., Porcella, D.B., Campbell, C.L., Pagenkopf, J.R., Rupp, G.L., Johnson, K.M., Chan, P.W.H., Gherini, S.A., 1985. Rates, Constants, and Kinetics Formulations in Surface Water Quality Modeling, second edition. U.S. Environmental Protection Agency, Athens Georgia. EPA/600/3e85/040. Breiman, L., Freidman, J., Olshen, R., Stone, C., 1984. Classification and Regression Trees. Wadsworth International, Pacific Grove, CA. Breiman, L., 1996. Technical Note: some properties of splitting criteria. Machine Learn. 24, 41e47. CANSIS (Canadian Soil Information System), 2007. Soil Landscapes of Canada Version 3.1.1. Available online. Agriculture and Agri-Food Canada, Ottawa, Ontario. http://sis. agr.gc.ca/cansis/nsdb/slc/v3.1.1/index.html. Cicek, H., Sunohara, M., Wilkes, G., McNairn, H., Pick, F., Topp, E., Lapen, D.R., 2010. Using vegetation indices from satellite remote sensing to assess corn and soybean response to controlled tile drainage. Agric. Water Manage. 98, 261e270.
5823
Clesceri, L., Greenberg, A., Eaton, A. (Eds.), 1998. Standard Methods for the Examination of Water and Wastewater, twentieth ed. American Public Health Association, United Book Press, Baltimore, MD. Curriero, F.C., Patz, J.A., Rose, J.B., Subhash, L., 2001. The association between extreme precipitation and waterborne disease outbreaks in the United States, 1948e1994. Am. J. Public Health 91 (8), 1194e1199. DeRegnier, D.P., Cole, L., Schupp, D.G., Erlandsen, S.L., 1989. Viability of giardia cysts suspended in lake, river, and tap water. Appl. Environ. Microbiol. 55 (5), 1223e1229. Edge, T.A., Droppo, I., El-Shaarawi, A., Gannon, V., Hewitt, M., Kent, R., Khan, I., Koning, W., Lapen, D., Marcogliese, D., Marvin, C., Miller, J., Neumann, N., Phillips, R., Robertson, W., Schreier, H., Shtepani, I., Topp, E., van Bochove, E., 2009. An Evaluation of Escherichia Coli as a Potential Agri-Environmental Waterborne Pathogen Standard. Environment Canada, Gatineau, Quebec, Synthesis Report No. 14. National AgriEnvironmental Standards Initiative, p. 48. Effler, E., Isaacson, M., Arntzen, L., Heenan, R., Canter, P., Barrett, T., Lee, L., Mambo, C., Levine, W., Zaidi, A., Griffin, P.M., 2001. Factors contributing to the emergence of Escherichia coli O157 in Africa. Emerg. Infect. Dis. 7 (5), 812e819. Environment Canada, 2010. National Climate Data and Information Archive. Environment Canada, Ottawa. (1) Climate Normals from 1971-2000 Product & Service. (2) Daily Data from the Canadian Daily Climate Data Product & Service. Environment Canada, Ottawa. Esteban, J.I., Oporto, B., Aduriz, G., Juste, R.A., Hurtado, A., 2009. Faecal shedding and strain diversity of Listeria monocytogenes in healthy ruminants and swine in Northern Spain. BMC Vet. Res. 5 (2), 1e10. Ferguson, C.M., Croke, B.F.W., Beatson, P.J., Ashbolt, N.J., Deere, D. A., 2007. Development of a process-based model to predict pathogen budgets for the Sydney drinking water catchment. J. Water Health 5 (2), 187e208. Gassman, P.W., Reyes, M.R., Green, C.H., Arnold, J.G., 2007. The soil and water assessment tool: historical development, applications, and future research directions. Trans. ASABE 50 (4), 1211e1250. Guerini, M.N., Brichta-Harhay, D.M., Shackelford, T.S., Arthur, T. M., Bosilevac, J.M., Kalchayanand, N., Wheeler, T.L., Koohmaraie, M., 2007. Listeria prevalence and Listeria monocytogenes serovar diversity at cull cow and bull processing plants in the United States. J. Food Prot. 70, 2578e2582. Ho¨rman, A., Rimhanen-Finne, R., Maunula, L., von Bonsdorff, C.H., Torvela, N., Heikinheimo, A., Ha¨nninen, M.L., 2004. Campylobacter spp., Giardia spp., Cryptosporidium spp., Noroviruses, and indicator organisms in surface water in Southwestern Finland, 2000e2001. Appl. Environ. Microbiol. 70 (1), 87e95. Hunsaker, C.T., Levine, D.A., 1995. Hierarchical approaches to the study of water quality in rivers. BioScience 45 (3), 193e203. Ice, G., 2004. History of Innovative Best Management Practice Development and Its Role in Addressing of Environmental Engineering, vol. 130, pp. 684e689. Ivanek, R., Grohn, Y.T., Wells, M.T., Lembro Jr., A.J., Sauders, B.D., Wiedmann, M., 2009. Modeling of spatially referenced environmental and meteorological factors influencing the probability of Listeria species isolation from natural environments. Appl. Environ. Microbiol. 75 (18), 5893e5909. Jamieson, R., Gordon, R., Joy, D., Lee, H., 2004. Assessing microbial pollution of rural surface waters: a review of current watershed scale modeling approaches. Agric. Water Manage. 70 (1), 1e17. Jellison, K.L., Distel, D.L., Hemond, H.F., Schauer, D.B., 2007. Phylogenetic analysis implicates birds as a source of
5824
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 0 7 e5 8 2 5
Cryptosporidium spp. oocysts in agricultural watersheds. Environ. Sci. Technol. 41, 3620e3625. Jokinen, C., Edge, T., Ho, S., Koning, W., Laing, C., Mauro, W., Medeiros, D., Miller, J., Robertson, W., Taboada, E., Thomas, J.E., Topp, E., Ziebell, K., Gannon, V., 2011. Molecular subtypes of Campylobacter spp., Salmonella enterica, and Escherichia coli O157: H7 isolated from faecal and surface water samples in the Oldman River watershed, Alberta, Canada. Water Res. 45 (3), 1247e1257. Keeley, A., Faulkner, B.R., 2008. Influence of land use and watershed characteristics on protozoa contamination in a potential drinking water resource reservoir. Water Res. 42, 2803e2813. King, B.J., Monis, P.T., 2007. Critical processes affecting Cryptosporidium oocyst survival in the environment. Parasitology 134, 309e332. Kistemann, T., Classen, T., Koch, C., Dangendorf, F., Fischeder, R., Gebel, J., Vacata, V., Exner, M., 2002. Microbial load of drinking water reservoir tributaries during extreme rainfall and runoff. Appl. Environ. Microbiol. 68 (5), 2188e2197. Lapen, D.R., Hayhoe, H.N., Topp, G.C., McLaughlin, N.B., Gregorich, E.G., Curnoe, W.E., 2002. Measurements of mouldboard plow draft: II. Draft-soil-crop and yield-draft associations. Precision Agric. 3 (3), 237e257. Lapen, D.R., Topp, E., Edwards, M., Sabourin, L., Curnoe, W., Gottschall, N., Bolton, P., Rahman, S., Ball-Coelho, B., Payne, M., Kleywegt, S., McLaughlin, N., 2008. Effect of liquid municipal biosolid application method on tile and groundwater quality. J. Environ. Qual. 37, 925e936. LeChevallier, M.W., Norton, W.D., Lee, R.G., 1991. Occurrence of Giardia and Cryptosporidium spp. in surface water supplies. Appl. Environ. Microbiol. 57 (9), 2610e2616. Leclerc, H., Schwartzbrod, L., Dei-Cas, E., 2002. Microbial agents associated with waterborne diseases. Crit. Rev. Microbiol. 28 (4), 371e409. Low, J.C., Donachie, W., 1997. A review of Listeria monocytogenes and listeriosis. Vet. J. 153 (1), 9e29. Lyautey, E., Lapen, D.R., Wilkes, G., McCleary, K., Pagotto, F., Tyler, K., Hartmann, A., Piveteau, P., Rieu, A., Robertson, W.J., Medeiros, D.T., Edge, T.A., Gannon, V., Topp, E., 2007. Distribution and characteristics of Listeria monocytogenes isolated from surface waters of the south nation river watershed. Appl. Environ. Microbiol. 73 (17), 5401e5410. Lyautey, E., Lu, Z., Lapen, D.R., Wilkes, G., Scott, A., Berkers, T., Edge, T., Topp, E., 2010. Distribution and diversity of Escherichia coli in the south nation river drainage basin, eastern Ontario, Canada. Appl. Environ. Microbiol. 76 (5). doi:10.1128/AEM. 02288-09. Lyautey, E., Wilkes, G., Miller, J., Van Bochove, E., Schreier, H., Koning, W., Edge, T., Lapen, D.R., Topp, E., 2011. Variation of an indicator of Escherichia coli persistence from surface waters of mixed-use watersheds, and relationships with population density and other factors. Ann. Limnol. e Int. J. Lim. 47, 1e9. Miller, W.G., Mandrell, R.E., 2005. Prevalence of campylobacter in food and water supply: incidence, outbreaks, isolation and detection. In: Ketley, J.M., Konkel, M.E. (Eds.), Campylobacter Molecular and Cellular Biology. Horizon Bioscience, Norfolk U.K. NRC (National Research Council), 2004. Committee on Indicators for Waterborne Pathogens. Indicators for Waterborne Pathogens. National Academies Press, Washington DC. O’Connor, D.R., 2002. Part One: A Summary Report of the Walkerton Inquiry: The Events of May 2000 and Related Issues. Ontario Ministry of the Attorney General, Queens Printer for Ontario, Toronto, Ontario, ISBN 0-7794-2558-8. Ouellet, G., Lapen, D.R., Topp, E., Sawada, M., Edwards, E., 2008. A heuristic model to predict earthworm biomass in
agroecosystems based on selected management and soil properties. Appl. Soil Ecol. 39 (1), 35e45. Pachepsky, Y.A., Sadeghi, A.M., Bradford, S.A., Shelton, D.R., Guber, A.K., Dao, T., 2006. Transport and fate of manure-borne pathogens: modeling perspective. Agric. Water Manage. 86, 81e92. Payment, P., Plante, R., Cejka, P., 2001. Removal of indicator bacteria, human enteric viruses, Giardia cysts, and Cryptosporidium oocysts at a large wastewater primary treatment facility. Can. J. Microbiol. 47, 188e193. Peterson, S., Signor, R., Ashbolt, N., Roser, D., 2006. Microrisk, Microbiological risk assessment: a scientific basis for managing drinking water safety from source to tap. QMRA methodology Study report performed as part of the MicroRisk project, co-funded by the European Commission under the Fifth Framework Programme, Theme 4: "Energy, environment and sustainable development" (contract EVK1-CT2002e00123). Quy, R.J., Cowan, D.P., Haynes, P.J., Sturdee, A.P., Chalmers, R.M., Bodley-Tickell, A.T., Bull, S.A., 1999. The Norway Rat as a reservoir host of Cryptosporidium Parvum. J Wildlife Diseases 35 (4), 660e670. Rao, N.S., Easton, Z.M., Schneiderman, E.M., Zion, M.S., Lee, D.R., 2009. Modeling watershed-scale effectiveness of agricultural best management practices to reduce phosphorus loading. J. Environ. Eng. 90, 1385e1395. Rollins, D.M., Colwell, R.R., 1986. Viable but nonculturable stage of Campylobacter jejuni and its role in survival in the natural aquatic environment. Appl. Environ. Microbiol. 52 (3), 531e538. Ruecker, N.J., Braithwaite, S.L., Topp, E., Edge, T., Lapen, D.R., Wilkes, G., Robertson, W., Medeiros, D., Sensen, C.W., Neumann, N.F., 2007. Tracking host sources of Cryptosporidium spp. in raw water for improved health risk assessment. Appl. Environ. Microbiol. 73 (12), 3945e3957. Ruecker, N.J., Wilkes, G., Topp, E., Edge, T., Lapen, D.R., Neumann, N.F., 2010. Investigating the association between land use and Cryptosporidium Occurrence in multiple watersheds in Eastern Ontario, Canada. In: Proceedings and Presented: International Symposium on Waterborne Pathogens. American Water Works Association, Manhattan Beach, CA. Sadeghi, A., Arnold, J.G., 2002. A SWAT/microbial submodel for predicting pathogen loadings in surface and groundwater at watershed and basin scales. In: Total Maximum Daily Load (TMDL) Environmental Regulations, Proceedings of the March 11-13, 2002 Conference, Fort Worth, TX, 56-63. Shreve, R.L., 1966. Statistical law of stream numbers. J. Geolog. 74, 17e37. Shuster, C.J., Ellis, A.G., Robertson, W.J., Charron, D.F., Aramini, J. J., Medeiros, D.T., 2005. Infectious disease outbreaks related to drinking water in Canada, 1974e2001. Can. J. Public Heath 96 (4), 254e258. Sischo, W.M., Atwill, E.R., Lanyon, L.E., George, J., 2000. Cryptosporidia on dairy farms and the role these farms may have in contaminating surface water supplies in the northeastern United States. Prev. Vet. Med. 43, 253e267. Smith, J.H., Wickham, J.D., Norton, D., Wade, T.G., Bruce Jones, K., 2001. Utilization of landscape indicators to model potential pathogen impaired waters. J. Am. Water Resour. Assoc. (JAWRA) 37 (4), 805e814. Soil Classification Working Group, 1998. The Canadian system of soil classification. Agric. Agri-Food Can., 187. Publ. 1646 (Revised). Statistics Canada, 2006. 2006 Census of Agriculture. 2006 Agriculture Community Profiles (Russell, Ontario CCS). Statistics Canada, Ottawa.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 0 7 e5 8 2 5
Steinberg, D., Colla, P., 1995. CART: Tree-Structured NonParametric Data Analysis. Salford Systems, San Diego, CA. Strahler, A.N., 1952. Hypsometric (area altitude) analysis of erosional topology. Geol. Soc. Am. Bull. 63, 1117e1142. Sunohara, M., Youssef, M., Topp, E., Lapen, D.R., 2010. Measured Effect of Agricultural Drainage Water Management on Hydrology, Water Quality, and Crop Yield Paper presented at the XVIIth World Congress of the International Commission of Agricultural Engineering (CIGR), Que´bec City, Canada June 13e17, 2010. USEPA U.S. Environmental Protection Agency, 2005. Total Maximum Daily Loads: National Section 303(d) List Fact Sheet: Top 100 Impairments. USEPA Office of Water, Washington, DC. USEPA U.S. Environmental Protection Agency, 2009. Review of Zoonotic Pathogens in Ambient Waters. USEPA Office of Water, Health and Ecological Criteria Division. EPA 822-R09e002. Walters, S.P., Gannon, V.P., Field, K.G., 2007. Detection of Bacteroidales fecal indicators and the zoonotic pathogens E. coli O157:H7, salmonella, and campylobacter in river water. Environ. Sci. Technol. 41, 1856e1862. Water Survey of Canada, 2008. Water Survey Branch, Data Products and Services. Environment Canada, Burlington.
5825
WERF (Water Environment Research Foundation), 2009. Report on the Expert Scientific Workshop on Critical Research and Science Needs for the Development of Recreational Water Quality Criteria for Inland Waters (PATH4W09). Dallas/Ft. Worth Texas, February 18e20, 2009. Water Environment Research Foundation, Alexandria, VA. Wilkes, G., Edge, T., Gannon, V., Jokinen, C., Lyautey, E., Medeiros, D., Neumann, N., Ruecker, N., Topp, E., Lapen, D.R., 2009. Seasonal relationships among indicator bacteria, pathogenic bacteria, Cryptosporidium oocysts, Giardia cysts, and hydrological indices for surface waters within an agricultural landscape. Water Res. 43 (8), 3615e3628. Winfield, M.D., Groisman, E.A., 2003. Role of Nonhost environments in the lifestyles of Salmonella and Escherichia coli. Appl. Environ. Microbiol. 69 (7), 3687e3694. World Health Organization, 2009. Global Health Risks: Mortality and Burden of Disease Attributable to Selected Major Risks. WHO Press, Geneva, Switzerland. Yang, W., Chen, P., Villegas, E.N., Landy, R.B., Kanetsky, C., Cama, V., Dearen, T., Schultz, C.L., Orndorff, K.G., Prelewicz, G. J., Brown, M.H., Young, K.R., Xiao, L., 2008. Cryptosporidium source tracking in the Potomac River watershed. Appl. Environ. Microbiol. 74 (21), 6495e6504.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 2 6 e5 8 3 6
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
An in-premise model for Legionella exposure during showering events Mary E. Schoen*, Nicholas J. Ashbolt Office of Research and Development, U.S. Environmental Protection Agency, 26 West Martin Luther King Drive, Cincinnati, OH 45268, USA
article info
abstract
Article history:
An exposure model was constructed to predict the critical Legionella densities in an engi-
Received 10 February 2011
neered water system that result in infection from inhalation of aerosols containing the
Received in revised form
pathogen while showering. The model predicted the Legionella densities in the shower air,
6 July 2011
water and in-premise plumbing biofilm that might result in a deposited dose of Legionella in
Accepted 18 August 2011
the alveolar region of the lungs associated with infection for a routine showering event.
Available online 5 September 2011
Processes modeled included the detachment of biofilm-associated Legionella from the inpremise plumbing biofilm during a showering event, the partitioning of the pathogen
Keywords:
from the shower water to the air, and the inhalation and deposition of particles in the
Risk
lungs. The range of predicted critical Legionella densities in the air and water was compared
Hot water system
to the available literature. The predictions were generally within the limited set of obser-
Plumbing
vations for air and water, with the exception of Legionella density within in-premise
Biofilm
plumbing biofilms, for which there remains a lack of observations for comparison. Sensi-
Legionella
tivity analysis of the predicted results to possible changes in the uncertain input param-
Protozoa
eters identified the target deposited dose associated with infections, the pathogen
Exposure
airewater partitioning coefficient, and the quantity of detached biofilm from in-premise pluming surfaces as important parameters for additional data collection. In addition, the critical density of free-living protozoan hosts in the biofilm required to propagate the infectious Legionella was estimated. Together, this evidence can help to identify critical conditions that might lead to infection derived from pathogens within the biofilms of any plumbing system from which humans may be exposed to aerosols. Published by Elsevier Ltd.
1.
Introduction
A review of drinking water-associated disease in the United States reported that Legionella accounted for 29% of outbreaks over the period it has been recorded (2001e2006) (Craun et al., 2010). Many of these outbreaks coincided with growth of Legionella in the in-premise plumbing (Craun et al., 2010). Additional evidence confirmed the presence of Legionella within in-premise plumbing of institutions, such as hospitals (Chen et al., 2008; Moore et al., 2006; Vo¨lker et al., 2010; Zeybeck and Cotuk, 2002), but there is a paucity of data for * Corresponding author. Tel.: þ1 513 569 7647. E-mail address:
[email protected] (M.E. Schoen). 0043-1354/$ e see front matter Published by Elsevier Ltd. doi:10.1016/j.watres.2011.08.031
individual households (Mathys et al., 2008). Infection with Legionella is only known to be associated with the inhalation of aerosols containing the bacteria (Casini et al., 2008). For example, shower aerosols have been identified as a potential pathway for exposure (Bollin et al., 1985; Cowen and Ollison, 2006). Here we investigated the conditions within the inpremise plumbing of homes that could result in the inhalation of human-infectious Legionella from shower aerosols. The biofilm on surfaces within engineered water systems, such as the in-premise plumbing, can create a biological niche for Legionella growth and persistence (Declerck et al., 2009).
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 2 6 e5 8 3 6
Multiple studies have recently reviewed Legionellaebiofilm interactions (Declerck, 2010; Lau and Ashbolt, 2009). Lau and Ashbolt (2009) proposed a model for the propagation and dissemination of pathogenic Legionella spp. in engineered water systems. In their model, the Legionella interact with freeliving protozoa present within the drinking water biofilm on pipe surfaces. First, pathogenic Legionella spp. colonize the biofilm and/or are ingested by biofilm grazing protozoa. If ingested, Legionella spp. can parasitize and eventually kill the protozoan host, or the protozoa can encyst while containing intracellular legionellae. These interactions result in four alternative scenarios of release from the biofilm: (1) Legionella spp. within the biofilm can be released as this material is detached, Legionella within the (2) trophozite or (3) cyst form of certain protozoa can be transported from the biofilm, or (4) Legionella spp. can be transported within vacuoles (vesicles) released from the protozoan host (Lau and Ashbolt, 2009). Here we present a processed-based, mathematical model based on the proposed theoretical model of Lau and Ashbolt (2009) to predict biofilm-associated Legionella propagation, detachment, and exposure (i.e. release scenario 1). The main objectives were to (1) use the best available data to estimate the critical densities of infectious Legionella in the shower air, water, and in-premise plumbing biofilm required to achieve a specific, target dose associated with infection in the lower respiratory tract, and to (2) identify the key pieces of information that require additional data collection for risk assessment. The secondary objective was to approximate the density of protozoan hosts in the in-premise plumbing biofilm necessary to achieve the target dose. Important data gaps in the available literature are identified in the discussion to advance risk assessment for biofilm-forming plumbing systems.
2.
Method
2.1.
Conceptual model
A mathematical model was constructed to simulate exposure to Legionella from inhalation of shower aerosols containing biofilm-associated Legionella detached from in-premise plumbing. The processes included in the mathematical model are depicted in Fig. 1. First, the biofilm-associated Legionella multiply within the biofilm (or sediments) or within a population of infected protozoan hosts within the biofilm of the in-premise plumbing (Fig. 1A). During a showering event, the biofilm-associated Legionella detaches from the biofilm (or sediments) into the bulk water (Fig. 1B). The Legionella is then aerosolized after flowing through the shower head with a portion of the aerosolized bacteria in aerosols of a respirable size (Fig. 1C). Once inhaled (Fig. 1D), processes within the lungs result in only a fraction of the inhaled bacteria reaching the alveolar region, the deposited dose (Fig. 1E). Although the conceptual model depicted in Fig. 1 starts with the growth of Legionella within the biofilm, the accompanying mathematical model is formulated in the reverse manner. The exposure model starts with an assumed target deposited dose and predicts the corresponding Legionella density in the shower air, water, biofilm, and host density
5827
within the biofilm. The predicted Legionella and host densities are referred to herein as critical densities.
2.2.
Deposited dose
Identifying a target deposited dose that is of concern for human-infectious Legionella is not straightforward. Armstrong (2005) summarized the differences in sensitivity to Legionella pneumophila (Lp) and the median lethal dose among alternative animal models. The guinea pig model was used in subsequent risk analysis based on the comparability of in vitro survival and replication of Lp in human and guinea pig alveolar macrophages (Armstrong, 2005; Armstrong and Haas, 2008). The exponential doseeresponse relationship for Lp was based on data collected by Muller et al. (1983) using an aerosol infection chamber and expressed in terms of a “retained dose.” Muller et al. (1983) estimated the “retention” of Lp Philadelphia 1 strain in the guinea pig lungs using Guyton’s formula, assuming that 50% of the inhaled dose was retained in the lungs. Based on these assumptions, reproducible infection was observed for an estimated retained dose of 5 CFU Lp (or more). The uncertain nature of the actual inhaled and deposited dose in the available animal model data led to an assessment of multiple possible target deposited doses of Lp in the alveolar region of the lungs, roughly 1e100 CFU. For each target deposited dose, the exposure model predicted the corresponding, critical Legionella densities in the shower air, water, and biofilm accounting for the physical processes described in Fig. 1.
2.3. Deposition of the target dose from inhalation of shower aerosols For each target deposited dose, the exposure model predicted the critical density of biofilm-associated Legionella in the air (CFU m3) during a shower exposure. The critical density of biofilm-associated Legionella in the air ðCLair Þ was expressed as a function of the target deposited dose (DD) and volume of inhaled air (Vair) as well as factors that account for the fraction of biofilm-associated Legionella that partition into the respirable aerosols of size range i ðF1i Þ and the fraction of the bacterial aerosols of size range i that are deposited in the lower respiratory tract ðF2i Þ (Eq.(1)). CLair ¼
Vair
DD P 1 Fi F2i
(1)
i
The volume of air that is inhaled was the product of the inhalation rate (IR) and the time of exposure (T ). Eq. (1) assumes that the Legionella within aerosols of size range i are equally distributed. Generally, aerosols of 1e10 mm in diameter may reach the alveolar region of the lungs (U.S. EPA, 2004).
2.4.
Partitioning of Legionella into shower aerosols
The critical density of Legionella in the water ðCLwater CFU L1 Þ necessary to achieve the target deposited dose was expressed as the critical density of Legionella in the air (from Eq. (1)) divided by a partitioning coefficient (PC ) based on the measured ratio of the density of bacteria in the air to the water (CFU m3/CFU L1) (Eq. (2)).
5828
CLwater ¼
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 2 6 e5 8 3 6
CLair PC
(2)
The total load of Legionella (CFU) delivered by the shower water volume (Vwater) to achieve the target deposited dose was L Þ (Equation (3)). also estimated ðWwater L Wwater ¼ CLwater Vwater
(3)
The volume of shower water was the product of the shower flow rate (FR) and the time of exposure (T ). Input parameters used in Eqs. (1)e(3) are described in Table 1.
2.5.
Density of Legionella in the biofilm
The exposure model was also used to predict the critical density of Legionella in the biofilm of the in-premise plumbing
Fig. 1 e Conceptual model for Legionella exposure from inhalation of shower aerosols containing Legionella derived from the in-premise plumbing of buildings. First, Legionella multiply within the premise plumbing biofilm, potentially by colonizing the biofilm or within a protozoan host (A). The biofilm-associated Legionella detaches from the biofilm during a showering event (B), is transported to the shower head, and is aerosolized (C). Finally, the aerosolized Legionella is inhaled (D) and a fraction of that inhaled dose is deposited in the alveolar region of the lungs (E).
5829
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 2 6 e5 8 3 6
assuming that the biofilm was the sole source of the Legionella in the bulk liquid. The critical density of Legionella in the biofilm necessary to achieve the target deposited dose was expressed as a function of the critical load of Legionella delivered in the shower water and the quantity of biofilm sloughed into the bulk water and of respirable size (Eqs. (4)e(5)). The density of Legionella in the biofilm was expressed two ways: the number of organisms per unit surface area of biofilm ðDLSA Þ (Eq. (4)), or the number of organisms per mass (dry weight) of biofilm ðDLM Þ (Eq. (5)). L Wwater Fbiofilm SA
DLSA ¼ DLM ¼
(4)
L Wwater DR SA T
(5)
where: Fbiofilm is the fraction of biofilm surface area that is sloughed off and of respirable size during a showering event (here assumed to be 0.25 h in duration); DR is the biofilm detachment rate (g cm2 0.25 h1); T is exposure time (h). Alternatively, Eq. (4) can be used to assess the density of Legionella associated with sediments by replacing the denominator with the total volume of sediments transported during a showering event. Eq. (5) could be expressed using other commonly used measures for biofilm mass or quantity if a detachment rate is available.
2.6.
Density of host cells in the biofilm
When the source of the Legionella in the biofilm was infected hosts, the pre-infection host density in the biofilm necessary to achieve the target deposited dose of Legionella was expressed as the critical density of Legionella in the biofilm ðDLSA or DLM Þ divided by the number of Legionella released by each infected host ðLChost Þ and the fraction of potential hosts infected ðF3host Þ for a set of specific conditions (C ) (Eqs. (6)e(7)). The density of host in the biofilm was expressed two ways: the number of organisms per unit surface area of biofilm ðDhost SA Þ (Eq. (6)), or the number of organisms per mass (dry weight) of biofilm ðDhost M Þ (Eq. (7)). Dhost SA ¼
DLSA 3 Fhost LChost
(6)
Dhost ¼ M
DLM F3host LChost
(7)
It is assumed in Eqs. (6)e(7) that only a fraction of the potential hosts are infected (i.e. their Legionella infection rate) and lyse/release an average number of Legionella (i.e. the infection intensity) for the set of conditions C. In addition to host strain differences, the conditions that affect the infection rate and intensity may include temperature, available
Table 1 e Best estimate, low and high values of input parameters used in the prediction of Legionella densities in air, water and in-premise plumbing biofilm and host density within in-premise plumbing biofilm during shower exposures. Parameter DD IR T F11e5
Description
SA DR
Deposited dose Inhalation rate Exposure time Fraction of total aerosolized organisms in aerosols of size range 1e5 mm Fraction of total aerosolized organisms in aerosols of size range 5e6 mm Fraction of total aerosolized organisms in aerosols of size range 6e10 mm Fraction of aerosols of size range 1e5 mm deposited at the alveoli Fraction of aerosols of size range 5e6 mm deposited at the alveoli Fraction of aerosols of size range 6e10 mm deposited at the alveoli Partitioning coefficient Shower flow rate Fraction of biofilm surface area that is sloughed off and of respirable size Total biofilm surface area Biofilm detachment rate
F3host LChost
Trophozite infection rate Legionella infection intensity
F15e6
F16e10
F21e5 F256 F2610 PC FR Fbiofilm
a Equivalent to 6 L min-1.
Units CFU m3 h1 h
CFU m3/CFU L1 L h1
cm2 g cm-2 0.25 h-1 CFU/host
Low Value
Best Estimate Value
High Value
1
10 0.72 0.25 0.75
100 1.5 1
(U.S. EPA, 2004) (Perkins et al., 2011) (Perkins et al., 2011)
0.09
0
(Perkins et al., 2011)
0.14
0
(Perkins et al., 2011)
0.2
0.54
(Schlesinger, 1989)
0.1
0.65
(Schlesinger, 1989)
0.01
0.1
(Schlesinger, 1989)
106
105 360a 1
Best Estimate Source
(Perkins et al., 2011) (Perkins et al., 2011)
40 1560
4000
520 0.01 10
0.25 100
1 1000
(Garny et al., 2009) (Kuiper et al., 2004) (Kuiper et al., 2004)
5830
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 2 6 e5 8 3 6
nutrients, microbial community, and span of time over which the biofilm community is active before detachment (Declerck et al., 2005; Kuiper et al., 2004; Ohno et al., 2008). Input parameters used in Eqs. (4)e(7) are further described in Table 1.
2.7.
Computation and sensitivity analysis
The density of Legionella in the air ðCLair Þ, the density of Legionella in the water ðCLwater Þ, and the total load of Legionella L Þ in Eqs. (1)e(3) were first delivered by the shower water ðWwater solved using best estimate input parameters (Table 1). A number of the input parameters used in Eqs. (1)e(3) are uncertain. To assess the sensitivity of the predicted output to the uncertain input parameters, Eqs. (1)e(3) were resolved using either low or high input values for each uncertain parameter separately. Similarly, the density of Legionella in the biofilm ðDLSA and DLM Þ and the density of host in the biofilm host ðDhost SA and DM Þ in Eqs. (4)e(7) were solved using best estimate input parameters as well as either low or high input values (Table 1). The exposure time (0.25 h) and shower flow rate (360 L h1) were not included in the sensitivity analysis since these parameters are not uncertain, although they can vary. All calculations were performed in Microsoft Excelä 2007.
3.
Input parameter values
3.1.
Density of Legionella in air and water
A summary of the parameter values used to estimate the density of Legionella in the air and water are presented in Table 1 and justified below. A target deposited dose of one infectious cell was assumed to be the minimal dose that could result in infection. Unfortunately, data was generally only available as cells estimated by colony forming units (CFU). While CFUs may underestimate the number of infectious cells present (Lee et al., 2011), we were limited to a CFU proxy and therefore set 1 CFU Lp as our low value target deposited dose. Based on Armstrong’s (2005) exponential doseeresponse relationship for L. pneumophila with best estimate exponential doseeresponse parameter r ¼ 0.06, the probably of infection for a retained dose of 1 CFU Lp is 0.06. Two other target deposited doses were evaluated. The best estimate target deposited dose of 10 CFU Lp was based on Armstrong’s (2005) median infectious dose of Lp, approximately 12 CFU. The high value was set at 100 CFU Lp, which is estimated to provide a probability of infection of 1.0. The data used to estimate the fraction of total aerosolized organisms in aerosols of respirable size ðF1i Þ and the partitioning coefficient (PC ) were available from a controlled shower environment experiment where high numbers of the bacterial species Brevundimonas diminuta (Bd ) were spiked into shower source water (Perkins et al., 2011). Over the course of a 15 min shower, the bacteria contained within aerosols from the breathing zone of an enclosed shower environment (at three locations) were collected to estimate the fraction of the culturable aerosolized bacteria that were retained in the aerosols of respirable size. The partitioning coefficient was calculated as the average density of bacteria in the air (CFU m3) divided by the density in the source water (CFU L1).
The PC was between 5.18 106 and 1.64 105 CFU m3/ CFU L1 for water temperatures between 25 and 40 C. The best estimate was set at 105 CFU m3/CFU L1 and 106 CFU m3/CFU L1 as the low estimate (Table 1). The fraction of aerosolized bacteria retained within aerosols of a certain aerodynamic diameter for Bd was set as the best estimate: 0.75 for aerosols of 1e<5 mm, 0.09 for aerosols of 5e<6 mm, and 0.14 for aerosols of 6e<10 mm. This represented 98% of the total aerosolized bacteria over the entire 15 min shower. The input values related to aerosolization may be sensitive to the selected shower head design (Plumb Shop/PS2 Novi, MI, USA), flow rate, pressure, and water temperature. Therefore, the fraction of bacteria retained within aerosols of 1e<5 mm diameter was increased to 1.0 as a maximum high value (with 0 set for the other size ranges). No lower limit was run based on the evidence for preferential partitioning of bacteria of respirable size into small aerosols (Perkins et al., 2011; Blanchard and Syzdek, 1982; Bollin et al., 1985). Literature on microbial partitioning is limited. Armstrong and Haas (2008) in conducting a risk assessment for Legionella exposure from hot spring aerosols assumed a PC based on endotoxin data of 2.3 105 CFU m3/CFU L1. Blanchard and Syzdek (1982) studied the water-to-air transfer of Serratia marcescens from bursting bubbles in distilled water at 22 C. The resulting film drops (1e10 mm) were collected on nutrient agar plates. An enrichment factor was calculated as the density of bacteria in the drops divided by the density of bacteria in the source water. Enrichment factors (EFs) were between 16 and 18 (Blanchard and Syzdek, 1982). The EFs were roughly translated into a PC of 106 CFU m3/CFU L1 assuming an aerosol density in the shower air of 107 L m3. The aerosol density is variable based on water temperature and shower head usage and was estimated here from a 10 min shower with a 7e10 L min1 flow rate and 36e38 C water temperature (Xu and Weisel, 2002). The fraction of inhaled bacterial aerosols deposited in the human alveoli (F2) was estimated for healthy adults based on the deposition of radioactive labeled particles (Schlesinger, 1989). Aerosols <1 mm in diameter (in the range of 0.2e1 mm) were not considered due to the larger size of Legionella cells. The particle deposition rate for nasal breathing was set as the best estimate and the high value was based on oral breathing (Schlesinger, 1989).
3.2.
Density of Legionella and host in biofilm
A summary of the parameter values used to estimate the Legionella and host densities in the biofilm are presented in Table 1 and justified below. The total biofilm surface area for in-premise plumbing within the right conditions for Legionella persistence or amplification likely varies from system to system. For the best estimate value of available biofilm surface area we assumed that 10 cm of 1.27 cm diameter (½ inch) pipe contributed 40 cm2 of biofilm coverage (i.e. 100% of the total surface area). The fraction of the biofilm surface area that sloughs off of respirable size was set to 1.0. Changes in the total biofilm surface area that sloughs off were considered in the sensitivity analysis with the high value representing 1000 cm of in-premise plumbing hot water service pipe.
5831
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 2 6 e5 8 3 6
When expressing the density of organisms in the biofilm in terms of biofilm mass (instead of surface area), the detachment rate is necessary. The detachment rate of biofilm within inpremise plumbing during a shower event appears not to have been reported. Garny et al. (2009) studied the effects of nutrients and flow on the structure of biofilms grown in a flow through tube reactor for a period of 7e8 weeks. The biofilms were cultivated in laminar (Re 1500) or turbulent (Re 3000) flow with either glucose limited or non-limited conditions. Detachment rates for particles 0.8e25 mm were estimated over the course of an entire day. Here, we derived the best estimate detachment rate from the maximum detachment for the turbulent condition and the low value detachment rate from the maximum detachment for the laminar condition. Large sloughing events that resulted in loss of most of the biofilm were ignored. The growth of Lp in co-culture with free-living protozoa has been studied for specific host and temperature combinations (Buse and Ashbolt, 2011; Newsome et al., 2001; Ohno et al., 2008; Panikov et al., 1993; States et al., 1993). Although these studies provide an idea of the potential variability in growth of Lp in coculture with protozoa, the dynamics of the biofilm system were not captured nor were important pieces of data such as the fraction of hosts that are infected and the number of Legionella per infected host. Kuiper et al. (2004) grew Lp in biofilms spiked with a mixed microbial community originating from a plumbing system with the Legionella host Hartmannella vermiformis on pieces of polyvinyl chloride with added nutrients (nitrate and phosphate) at 37 C. The infection rate, the fraction of the total trophozoites that were infected, was estimated for different time points, i.e. roughly 0.1 at day 10 and >0.9 at day 14. The number of Lp within each infected host (the infection intensity) was reported to be in the range of one when first infected to 100 before lysis (Kuiper et al., 2004). In a co-culture study of Lp with hosts Acanthamoeba castellanii or Naegleria lovaniensis, Declerck et al. (2005) reported a similar infection intensity after 24e48 h (however, the maximum detection limit was 100 CFU and likely underestimated the infection intensity). Importantly, the infection rate in co-culture was host specific. After 72 h of co-culture, 100% of A. castellanii were infected compared with only 2% of N. lovaniensis (Declerck et al., 2005). Here, our best estimates assumed that the residence time of the biofilm in the in-premise plumbing after host infection was sufficient for a maximum infection intensity of 100 CFU and that the potential host population was a mix of different species so that 25% were infected and lysed over the residence time. Since the best estimate parameter values were inferred from limited observational data, we also explored the
infection rate and intensity parameters for perceived limits of 0.01 and 1 and 10 and 1000 CFU (Table 1).
4.
Results
4.1.
Interpretation of results
The shower exposure model predicted the density of infectious Legionella in the air, shower water, and biofilm required to achieve a target deposited dose in the alveolar region of the lungs after a 15 min exposure to shower aerosols containing detached, biofilm-associated Legionella. The reported critical densities within air, water, and biofilm include a best estimate value as well as high and low predictions based on the absolute highest and lowest outcomes of the sensitivity analysis (Table 2). The predicted densities in the air and water must be interpreted as the density of infectious biofilm-associated Legionella of respirable size, not the total Legionella density. This clarification is due to the small size (<10 mm) of the test organisms used to assess the partitioning coefficient. The density of protozoan hosts within the biofilm before infection with Legionella required to release the critical Legionella biofilm density was also predicted (Table 2).
4.2. Critical densities of Legionella of respirable size in air and water The predicted ranges of the critical densities of biofilmassociated Legionella of respirable size in shower air and water were respectively 3.5 101e3.5 103 CFU m3 and 3.5 106e3.5 108 CFU L1 for a 15 min exposure event with the conditions explored (Table 2). Based on a 15 min shower with a flow rate of 6 LPM, the range of infectious Legionella contained in particles of respirable size in shower water to achieve the target deposited dose was 3.1 108e3.1 1010 CFU. To investigate which parameters were most influential on the predicted density of Legionella in water, a sensitivity analysis was undertaken. Changes in the partitioning coefficient and target deposited dose corresponding to the range of values reported in the literature resulted in the greatest changes in the best estimate critical density of Legionella in water (Fig. 2).
4.3.
Critical density of Legionella in the biofilm
The ranges of predicted critical density of Legionella in the biofilm necessary to detach and achieve the critical water
Table 2 e Absolute low, high and best estimate predictions of Legionella of respirable size and host densities in air ðCLair Þ, host water ðCLwater Þ and biofilm ðDLSA ; DLM ; Dhost SA ; DM Þ necessary to result in infection from inhalation of shower aerosols during a 15 min shower. Organism
Medium
Legionella Legionella Legionella
Air Water Biofilm
Host
Biofilm
Units CFU m3 CFU L1 CFU cm-2 CFU g1 host cm2 host g1
Low Prediction 3.5 3.5 7.8 5.0 3.1 2.0
101 106 105 102 104 101
Best Estimate Prediction 3.5 3.5 7.8 5.0 3.1 2.0
102 107 107 104 106 103
High Prediction 3.5 3.5 7.8 5.0 7.8 5.0
103 108 108 105 107 104
5832
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 2 6 e5 8 3 6
density for a 15 min shower event were 7.8 105e7.8 108 CFU cm2 or 5.0 102e5.0 105 CFU g1 (Table 2) for the conditions explored. The sensitivity of the best estimate prediction expressed as the number of organisms per mass of biofilm to changes in the input parameters in Eq. (5) is shown in Fig. 3. Possible changes in the total surface area contributing to a sloughing event (Fig. 3) and the surface area that is detached and of respirable size (Fbiofilm*SA, results not shown) resulted in the largest changes to the critical density of Legionella in the biofilm. The detachment rate of respirablesized particles may be more influential than conveyed in Fig. 3 since the possible range of values for detachment rate in inpremise plumbing was not available from the literature.
4.4.
Critical density of protozoan host in the biofilm
The predicted critical density of protozoan host in the biofilm was 3.1 104e7.8 107 host cm2 or 2.0 101e5.0 104 host g1 (Table 2) for the conditions explored. The sensitivity of the best estimate density to changes in the input parameters in Eq. (7) is shown in Fig. 4. In addition to the previously identified important input parameters, possible changes in the host infection rate and intensity also resulted in large changes to the predicted critical density of protozoan host in the biofilm. The low value infection rate, representative of less vulnerable amoebae species and/or a short timeframe for biofilm activity (<48 h) before detachment (i.e. F3 ¼ 0.01), resulted in the second largest change in the critical protozoan density prediction.
5.
Discussion
We estimated the critical density of infectious Legionella in the shower water and air in particles of a respiratory size
Fig. 2 e Sensitivity analysis of the critical density of biofilm-associated Legionella contained within particles of respirable size in the bulk shower water ðCLwater Þ required to cause infection from inhalation of shower aerosols during a 15 min shower exposure. Deviations from the best estimate of 107.5 CFU LL1 are shown for low and high parameter value changes (Table 1) in inhalation rate (IR), fraction of the total aerosolized bacteria that are retained in the aerosols of respirable size (F1), fraction of the bacterial aerosols deposited in the human alveoli (F2), bacterial partitioning coefficient (PC ), and deposited dose (DD).
necessary to achieve a target deposited dose in the lower respiratory tract. The critical densities were based on a 15 min shower exposure event. The predicted critical densities of Legionella are reported here as colony forming units (CFU) in air, water and biofilm. This approach is flawed given that at least a sub-population of infectious Legionella may be nonculturable (Storey et al., 2004). If infectious, but nonculturable organisms partition similarly and are equally infectious to the culturable, then the predicted critical densities of Legionella in the air and water represent both potential forms of Legionella cells. Using a range of possible input values, the predicted critical range of respirable-sized Legionella in the shower air and water necessary to achieve the target deposited dose were respec 103 CFU m3 and tively 3.5 101e3.5 6 8 1 3.5 10 e3.5 10 CFU L for the conditions explored. The predicted critical Legionella densities may not be representative of all systems. For example, the partitioning of Legionella from bulk water to air may be sensitive to shower head design and age, the flow rate, and the relative humidity. Also, the predicted critical densities were based on a 15 min shower exposure event; total exposure duration may vary and/or continue long after the shower is turned off due to continued exposure to bathroom aerosols. Finally, the model does not account for inactivation of Legionella by temperature or residual water disinfectant while within the biofilm or while transported through the system. Noting that the critical densities of Legionella in the air, water, and biofilm should not change whether or not there is inactivation by temperature or disinfectant within the biofilm. However, the critical densities may change if there is inactivation of Legionella while transported through the system. Over short periods of time, the inactivation of Lp from low levels of disinfectant is most likely minimal (Storey et al., 2004). Likewise, when Lp previously co-cultured with A. castellanii were exposed to a thermal treatment of 50 C over 10 min, less than one log reduction was measured (Storey et al., 2004). Although these findings may not apply to all Legionella and host combinations, the possible inactivation from thermal or water disinfectant over a short time period of transport appears small based on available data (Storey et al., 2004). To assess whether the predicted critical densities of Legionella are possible, we compiled data on the observed densities of Legionella in the air and water from hospital and family residences (Table 3) (Bauer et al., 2008; Bollin et al., 1985; Borella et al., 2004; Breiman et al., 1990; Mathys et al., 2008; Vo¨lker et al., 2010; Zacheus and Martikainen, 1994). There are few studies, however, that report the density of Legionella in air after aerosolization from shower or sink faucets. A maximum air density of 190 CFU m3 Legionella was reported for shower aerosols in a hospital where Legionellosis cases had occurred (Breiman et al., 1990). This observed maximum air density falls between our low and best estimate predictions of critical Legionella density in the air (3.5 101e3.5 102 CFU m3), confirming that the predicted range is physically possible, but not necessarily representative of densities responsible for Legionellosis. Other studies report the density of Legionella in air by qPCR, but do not distinguish between infectious and non-infectious Legionella (Bauer et al., 2008).
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 2 6 e5 8 3 6
5833
Fig. 3 e Sensitivity analysis of the critical density of Legionella in the biofilm ðDLM Þ required to cause infection from inhalation of shower aerosols containing detached organisms during a 15 min showering event. Density is expressed as the number of organisms per mass (dry weight) of biofilm. The best estimate critical density of Legionella in the biofilm was 104.7 CFU gL1. Deviations from the best estimates are shown for low and high parameter values changes (Table 1) in bacterial partitioning coefficient (PC ), deposited dose (DD), surface area (SA), and detachment rate (DR). Deviations omitted for changes in the inhalation rate (IR), fraction of infectious aerosolized bacteria that are retained in the aerosols of respirable size (F1), and fraction of the bacterial aerosols deposited in the human alveoli (F2).
Fig. 4 e Sensitivity analysis of the critical, pre-infection density of the protozoan host in the biofilm ðDhost M Þ required to cause infection from inhalation of shower aerosols containing Legionella expressed as the number of organisms per mass (dry weight) of biofilm. The best estimate density of host in the biofilm was 103.3 CFU gL1. Deviations from the best estimate are shown for low and high parameter value changes (Table 1) in bacterial partitioning coefficient (PC ), deposited dose (DD), surface area (SA), fraction of potential hosts that are infected (F3), and number of Legionella released by each infected host (L). Deviations omitted for changes in the inhalation rate (IR), fraction of infectious aerosolized bacteria that are retained in the aerosols of respirable size (F1), fraction of the bacterial aerosols deposited in the human alveoli (F2), and detachment rate (DR).
5834
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 2 6 e5 8 3 6
Table 3 e Observed culturable Legionella densities in air and water. Data Source (Borella et al., 2004) (Zacheus and Martikainen, 1994) (Vo¨lker et al., 2010) (Mathys et al., 2008) (Bauer et al., 2008) (Bollin et al., 1985) (Breiman et al., 1990)
Location
Air (CFU m-3)
Water (CFU L-1)
Comments
Hot water sink outlet Hot water taps and showers
9.6 102 2.7 103
Geometric mean of positive samples Mean of positive samples
In-building distribution systems Hot water systems of single family residences Shower hospital Shower hospital Shower hospital
106a 106a
Maximum of positive samples Maximum of positive samples
105a
Maximum of positive samples Maximum of reported positive samples After cases diagnosed
5.88 190
a Estimated from graph.
There are more observational data available on the density of Legionella in water within in-premise plumbing (Table 3). Here, the best estimate infectious Legionella density in shower water necessary to achieve a target deposited dose of 10 CFU Legionella was 3.5 107 CFU L1. This best estimate is greater than the Legionella densities observed in studies of in-premise plumbing (Bauer et al., 2008; Borella et al., 2004; Mathys et al., 2008; Vo¨lker et al., 2010; Zacheus and Martikainen, 1994). However, the observational studies typically do not quantify infectious but non-cultivable Legionella. Bauer et al. (2008) measured the Legionella water density in hospital showers associated with Pontiac fever (PF) occurrences in elderly patients. Legionella spp. densities above 104 CFU L1 (of which 2/3 were identified as L. pneumophila) were associated with an increased PF risk. Higher densities of Legionella at 106 CFU L1 were reported in the in-premise plumbing in other studies (Mathys et al., 2008; Vo¨lker et al., 2010). Taken together, the limited evidence suggests that the range of values predicted for critical Legionella densities of concern are possible for inpremise plumbing and that best estimate predictions are not to be interpreted as conclusive for all systems, but certainly plausible. We also estimated the Legionella density in the biofilm necessary to achieve a target deposited dose after biofilm detachment during a showering event, as well as the host density required to propagate the Legionella in the biofilm. There is little evidence on the Legionella density within biofilms of in-premise plumbing for comparison. Many studies reported the density of Legionella in model hot water systems spiked with Legionella over time (Liu et al., 2006; Moritz et al., 2010; Saby et al., 2005; Van der Kooij et al., 2005). The results from these spiking studies show that it is possible to observe the predicted critical Legionella densities in biofilms in controlled experiments. Other studies assessed the presence of Legionella in-premise plumbing biofilm using molecular approaches (Ditommaso et al., 2010; Feazel et al., 2009; Perkins et al., 2009). Feazel et al. (2009) collected swabs of shower biofilms from 12 sites and looked for the presence of legionellae by partial 16S rRNA gene sequencing but found little evidence of Legionella (Feazel et al., 2009). A recent review summarizes the detection/densities of free-living amoebae in drinking water and biofilm (Thomas and Ashbolt, 2010). However, we were unable to identify studies of in-premise plumbing that specifically look for infectious Legionella densities in biofilm and associated host
densities. Method selection is an important consideration when seeking to fill this data gap. The use of qPCR to estimate pathogen densities is useful for capturing non-cultivable Legionella. However, the existing doseeresponse studies for Legionella are presented using a culture-based method and are not compatible with qPCR data. To be compatible, quantification of the organisms detected by qPCR that are potentially infectious should also be assessed and reported to ultimately assess risk. The sensitivity analysis identified a subset of uncertain model parameters with the greatest influence on the predicted critical Legionella densities in air, water, and biofilm. The partitioning coefficient and the target deposited dose associated with infection for Legionella remain uncertain and were identified as the most important parameters in predicting the critical Legionella density in shower water. Armstrong and Haas (2008) also identified the partitioning coefficient as an important data gap in a risk assessment of Legionella aerosolized from spas. Partitioning may vary with shower head or Legionella strain. For example, Parker et al. (1983) studied the water-to-air transfer of multiple strains of Mycobacterium with an approach similar to Blanchard and Syzdek (1982) but focused on the larger jet droplets (size 50e100 mm) using river water (Parker et al., 1983). The EFs for Mycobacterium intracellulare ranged between 1389 and 9755 for various strains, respectively (Parker et al., 1983). Ultimately for more precise estimates of risk, additional evidence on partitioning for different shower head designs and pathogen strains as well as evidence on the doseeresponse for both culturable and non-culturable environmental Legionella strains are required. If the Legionella originates from the biofilm, what is still unknown is the quantity of biofilm that sloughs off within the in-premise plumbing or possibly from upstream in the municipal distribution system during (or previous to) a showering event that results in material of respirable size. Detached biofilm quantities were estimated here by biofilm surface area and dry weight mass because data was available to populate the model parameters. An alternative detachment rate might be expressed as biofilm volume or as mass of the biofilm protein or carbohydrate. The quantity of detached biofilm during a showering event likely varies with flow velocity, system usage and design. The authors are not aware of studies that specify the amount of biofilm detached during a showering event or the size of those detached biofilm particles when aerosols are generated.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 2 6 e5 8 3 6
If infectious Legionella originate from protozoan hosts associated with in-premise plumbing, both the host infection rate and intensity are important in the prediction of critical host density. Here, we estimated critical host density assuming that the host infection rate and intensity were homogeneous in the biofilm where detachment occurs. In real systems, the host infection rate and intensity within the system will not be homogeneous, but will likely vary with temperature, disinfection, host and Legionella strains, and time available for Legionella infection before detachment (Declerck et al., 2005). Additional research both on the actual temperature and biofilm conditions within in-premise plumbing as well as host-Legionella interactions under those conditions are important for further improvements in estimating the critical density of host required to propagate an infectious dose of Legionella. In addition to biofilm-associated or free Legionella cells, Lau and Ashbolt (2009) identified Legionella dissemination scenarios that included, Legionella contained within host cysts, trophozoites, and/or vesicles. Cysts and trophozoites are on the upper end of the size range that could potentially be aerosolized (i.e. 10 mm). It is difficult to postulate about these scenarios given the lack of information on partitioning for particles of this size and form. Whereas, vacuoles (vesicles) released from the protozoa containing Legionella spp. are of a smaller size. Berk et al. (1998) conducted feeding experiments with L. pneumophila to observe the vesicle production and number expelled from A. castellanii and Acanthamoeba polyphaga at 25, 30, and 35 C. After 24 h, over 90% of vesicles were within the respirable size range (1e5 mm). The theoretical number of Legionella contained within a vesicle was calculated as 20-200 based on the observed vesicle size. If the vesicles stay intact and partition like B. diminuta then the critical density of vesicles in the air and water to achieve a target deposited dose of Legionella is analogous to the calculation of biofilm-bound Legionella density. If the vesicles break open through the exposure pathway, the critical density of broken vesicles is smaller and depends on the surviving number of Legionella from each vesicle. For example, if the number of Legionella per vesicle were high, say 1000 CFU, then the critical number of broken vesicles delivered during a 15 min shower event to achieve the target deposited dose is roughly 106 vesicles.
6.
Conclusion
The results from the shower exposure model provide a rough estimate of the critical number of Legionella and host associated with in-premise plumbing of concern for possible infection from inhalation of shower-generated aerosols containing Legionella. Preliminary results, although plausible, will likely vary for individual system conditions. Predictions of risk and critical host and pathogen densities in the in-premise plumbing would improve with additional information on the pathogen partitioning, biofilm detachment quantities for inpremise plumbing, and host-Legionella interactions within conditions present for in-premise plumbing. Together, this evidence can help to identify critical conditions that might lead to potential infection derived from pathogens within the biofilms of any plumbing system from which humans may be
5835
exposed to aerosols, such as from pathogenic mycobacteria (Falkinham et al., 2008).
Acknowledgments This work has been subject to formal U.S. Environmental Protection Agency review and does not necessarily reflect the views of the Agency. No official endorsement should be inferred. The authors gratefully acknowledge the valuable contributions and discussions provided by Dr. Helen Buse, Dr. Sarah Perkins, Dr. Jorge Santo Domingo, Dr. Troy Hawkins, Dr. Marc Edwards, Jacquie Thomas, Dr. Tonya Nichols, and Randi Lieberman.
references
Armstrong, T.W., 2005. A Quantitative Microbial Risk Assessment Model for Human Inhalation Exposure to Legionella. Drexel University, Philadephia, PA. Armstrong, T.W., Haas, C.N., 2008. Legionnaires’ disease: evaluation of a quantitative microbial risk assessment model. Journal of Water and Health 6 (2), 149e166. Bauer, M., Mathieu, L., Deloge-Abarkan, M., Remen, T., Tossa, P., Hartemann, P., Zmirou-Navier, D., 2008. Legionella bacteria in shower aerosols increase the risk of Pontiac fever among older people in retirement homes. Journal of Epidemiology and Community Health 62 (10), 913e920. Berk, S.G., Ting, R.S., Turner, G.W., Ashburn, R.J., 1998. Production of respirable vesicles containing live Legionella pneumophila cells by two Acanthamoeba spp. Applied and Environmental Microbiology 64 (1), 279e286. Blanchard, D.C., Syzdek, L.D., 1982. Water-to-air transfer and enrichment of bacteria in drops from bursting bubbles. Applied and Environmental Microbiology 43 (5), 1001e1005. Bollin, G.E., Plouffe, J.F., Para, M.F., Hackman, B., 1985. Aerosols containing Legionella pneumophila generated by shower heads and hot-water faucets. Applied and Environmental Microbiology 50 (5), 1128e1131. Borella, P., Montagna, M., Romano-Spica, V., Stampi, S., Stancanelli, G., Triassi, M., Neglia, R., Marchesi, I., Fantuzzi, G., Tato`, D., Napoli, C., Quaranta, G., Laurenti, P., Leoni, E., De Luca, G., Ossi, C., Moro, M., Ribera D’Alcala`, G., 2004. Legionella infection risk from domestic hot water. Emerging Infectious Diseases 10 (3), 457e464. Breiman, R., Fields, B., Sanden, G., Volmer, L., Meier, A., Spika, J., 1990. Association of shower use with Legionnaires’ disease: possible role of amoebae. Journal of the American Medical Association 263 (21), 2924. Buse, H.Y., Ashbolt, N.J., 2011. Differential growth of Legionella pneumophila strains within a range of amoebae at various temperatures associated with in-premise plumbing. Letters in Applied Microbiology. Casini, B., Valentini, P., Baggiani, A., Torracca, F., Frateschi, S., Nelli, L.C., Privitera, G., 2008. Molecular epidemiology of Legionella pneumophila serogroup 1 isolates following long-term chlorine dioxide treatment in a university hospital water system. Journal of Hospital Infection 69 (2), 141e147. Chen, Y.S., Lin, Y.E., Liu, Y.C., Huang, W.K., Shih, H.Y., Wann, S.R., Lee, S.S., Tsai, H.C., Li, C.H., Chao, H.L., Ke, C.M., Lu, H.H., Chang, C.L., 2008. Efficacy of point-of-entry copper-silver ionisation system in eradicating Legionella pneumophila in a tropical tertiary care hospital: implications for hospitals
5836
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 2 6 e5 8 3 6
contaminated with Legionella in both hot and cold water. Journal of Hospital Infection 68 (2), 152e158. Cowen, K.A., Ollison, W.M., 2006. Continuous monitoring of particle emissions during showering. Journal of the Air & Waste Management Association 56 (12), 1662e1668. Craun, G.F., Brunkard, J.M., Yoder, J.S., Roberts, V.A., Carpenter, J., Wade, T., Calderon, R.L., Roberts, J.M., Beach, M.J., Roy, S.L., 2010. Causes of outbreaks associated with drinking water in the United States from 1971 to 2006. Clinical Microbiology Reviews 23 (3), 507e528. Declerck, P., Behets, J., Delaedt, Y., Margineanu, A., Lammertyn, E., Ollevier, F., 2005. Impact of non-Legionella bacteria on the uptake and intracellular replication of Legionella pneumophila in Acanthamoeba castellanii and Naegleria lovaniensis. Microbial Ecology 50 (4), 536e549. Declerck, P., Behets, J., Margineanu, A., van Hoef, V., De Keersmaecker, B., Ollevier, F., 2009. Replication of Legionella pneumophila in biofilms of water distribution pipes. Microbiological Research 164 (6), 593e603. Declerck, P., 2010. Biofilms: the environmental playground of Legionella pneumophila. Environmental Microbiology 12 (3), 557e566. Ditommaso, S., Giacomuzzi, M., Gentile, M., Moiraghi, A.R., Zotti, C.M., 2010. Effective environmental sampling strategies for monitoring Legionella spp contamination in hot water systems. American Journal of Infection Control 38 (5), 344e349. Falkinham, J.O., Iseman, M.D., de Haas, P., van Soolingen, D., 2008. Mycobacterium avium in a shower linked to pulmonary disease. Journal of Water and Health 6 (2), 209e213. Feazel, L., Baumgartner, L., Peterson, K., Frank, D., Harris, J., Pace, N., 2009. Opportunistic pathogens enriched in showerhead biofilms. Proceedings of the National Academy of Sciences 106 (38), 16393. Garny, K., Neu, T.R., Horn, H., 2009. Sloughing and limited substrate conditions trigger filamentous growth in heterotrophic biofilmsemeasurements in flow-through tube reactor. Chemical Engineering Science 64 (11), 2723e2732. Kuiper, M.W., Wullings, B.A., Akkermans, A.D.L., Beumer, R.R., van der Kooij, D., 2004. Intracellular proliferation of Legionella pneumophila in Hartmannella vermiformis in aquatic biofilms grown on plasticized polyvinyl chloride. Applied and Environmental Microbiology 70 (11), 6826e6833. Lau, H.Y., Ashbolt, N.J., 2009. The role of biofilms and protozoa in Legionella pathogenesis: implications for drinking water. Journal of Applied Microbiology 107 (2), 268e278. Lee, J.V., Lai, S., Exner, M., Lenz, J., Gaia, V., Hasati, S., Hartemann, P., Lu¨ck, C., Pangon, B., Ricci, M.L., Scaturro, M., Fontana, S., Sabria, M., Sa´nchez, I., Assaf, S., Surman-Lee, S., 2011. An international trial of quantitative PCR for monitoring Legionella in artificial water systems. Journal of Applied Microbiology 110, 1032e1044. Liu, Z., Lin, Y., Stout, J., Hwang, C., Vidic, R., Yu, V., 2006. Effect of flow regimes on the presence of Legionella within the biofilm of a model plumbing system. Journal of Applied Microbiology 101 (2), 437e442. Mathys, W., Stanke, J., Harmuth, M., Junge-Mathys, E., 2008. Occurrence of Legionella in hot water systems of single-family residences in suburbs of two German cities with special reference to solar and district heating. International Journal of Hygiene and Environmental Health 211 (1e2), 179e185. Moore, M.R., Flannery, B., Gelling, L.B., Couroy, M., Vugia, D., Salerno, J., Weintraub, J., Stevens, V., Fields, B.S., Besser, R., 2006. In: Cianciotto, N.P., Kwaik, Y.A., Edelstein, P.H., Fields, B. S., Geary, D.F., Harrison, T.G., Joseph, C.A., Ratclif, R.M., Stout, J. E., Swanson, M.S. (Eds.), Legionella: state of the art 30 years after its recognition. ASM Press, Washington, D.C, pp. 526e528.
Moritz, M.M., Flemming, H., Wingender, J., 2010. Integration of Pseudomonas aeruginosa and Legionella pneumophila in drinking water biofilms grown on domestic plumbing materials. International Journal of Hygiene and Environmental Health 213 (3), 190e197. Muller, D., Edwards, M.L., Smith, D.W., 1983. Changes in iron and transferrin levels and body temperature in experimental airborne legionellosis. Journal of Infectious Diseases 147, 302e307. Newsome, A.L., Farone, M.B., Berk, S.G., Gunderson, J.H., 2001. Free living amoebae as opportunistic hosts for intracellular bacterial parasites. Journal of Eukaryotic Microbiology, 13Se14S. Ohno, A., Kato, N., Sakamoto, R., Kimura, S., Yamaguchi, K., 2008. Temperature-dependent parasitic relationship between Legionella pneumophila and a free-living amoeba (Acanthamoeba castellanii). Applied and Environmental Microbiology 74 (14), 4585e4588. Panikov, N.S., Merkurov, A.E., Tartakovskii, I.S., 1993. In: Barbaree, J. M., Breiman, R.F., Dufour, A.P. (Eds.), Legionella: current status and emerging perspectives. ASM, Washington, D.C. Parker, B.C., Ford, M.A., Gruft, H., Falkinham, J.O., 1983. Preferential aerosolization of Mycobacterium intracellulare from natural waters. American Review of Respiratory Disease 128. Perkins, S.D., Byerns, R., Cowen, K., Lorch, D., Shaw, M., Taylor, M. , Nichols, T., 2011. Assessing exposures to shower aerosolized Brevundimonas diminuta and Pseudomonas aeruginosa. Applied and Environment Microbiology in Review. Perkins, S.D., Mayfield, J., Fraser, V., Angenent, L.T., 2009. Potentially pathogenic bacteria in shower water and air of a stem cell transplant unit. Applied and Environmental Microbiology 75 (16), 5363. Saby, S., Vidal, A., Suty, H., 2005. Resistance of Legionella to disinfection in hot water distribution systems. Water Science and Technology 52 (8), 15e28. Schlesinger, R.B., 1989. In: McClellan, R.O., Henderson, R.F. (Eds.), Concepts in inhalation toxicology. Hemisphere Publishing Corp, New York, pp. 163e192. States, S.J., Podorski, J.A., Conley, L.F., Young, W.D., Wadowsky, R. M., Dowling, J.N., Kuchta, J.M., Navratil, J.S., Yee, R.B., 1993. In: Barbaree, J.M., Breiman, R.F., Dufour, A.P. (Eds.), Legionella: current status and emerging perspectives. ASM, Washington, D.C. Storey, M.V., Winiecka-Krusnell, J., Ashbolt, N.J., Stenstrom, T.A., 2004. The efficacy of heat and chlorine treatment against thermotolerant Acanthamoebae and Legionellae. Scandinavian Journal of Infectious Diseases 36 (9), 656e662. Thomas, J.M., Ashbolt, N.J., 2010. Do free-living amoebae in treated drinking water systems present an emerging health risk? Environmental Science & Technology 45 (3), 860e869. U.S. EPA, 2004. In: Air Quality Criteria for Particulate Matter, vol. II. U.S. EPA, Washington D.C. Van der Kooij, D., Veenendaal, H.R., Scheffer, W.J.H., 2005. Biofilm formation and multiplication of Legionella in a model warm water system with pipes of copper, stainless steel and crosslinked polyethylene. Water Research 39 (13), 2789e2798. Vo¨lker, S., Schreiber, C., Kistemann, T., 2010. Drinking water quality in household supply infrastructure e a survey of the current situation in Germany. International Journal of Hygiene and Environmental Health 213 (3), 204e209. Xu, X., Weisel, C.P., 2002. Inhalation exposure to haloacetic acids and haloketones during showering. Environmental Science & Technology 37 (3), 569e576. Zacheus, O.M., Martikainen, P.J., 1994. Occurrence of legionellae in hot-water distribution-systems of Finnish apartment buildings. Canadian Journal of Microbiology 40 (12), 993e999. Zeybeck, Z., Cotuk, A., 2002. In: Marre, Reinhard, et al. (Eds.), Legionella. ASM Press, Washington, D.C, pp. 305e308.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 3 7 e5 8 4 8
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Effective detection of human noroviruses in Hawaiian waters using enhanced RT-PCR methods Hsin-I Tong a, Christina Connell a, Alexandria B. Boehm b, Yuanan Lu a,* a b
Departments of Public Health and Sciences and Microbiology, University of Hawaii at Manoa, Honolulu, Hawaii 96822, USA Stanford University, Department of Civil & Environmental Engineering, Stanford, CA 94305, USA
article info
abstract
Article history:
The current recreational water quality criteria using growth-based measurements of fecal
Received 25 February 2011
indicator bacteria (FIB) concentration have their limitations for swimmer protection. To
Received in revised form
evaluate the possible use of enteric viruses as an improved indicator of human sewage
12 July 2011
contamination in recreational waters for enhanced health risk assessment, human nor-
Accepted 18 August 2011
ovirus (huNoV) was tested as a model in this study. To establish a highly sensitive protocol
Available online 7 September 2011
for effective huNoV detection in waters, 16 published and newly designed reverse transcription polymerase chain reaction (RT-PCR) primer pairs specific for huNoV genogroup I
Keywords:
(GI) and genogroup II (GII) were comparatively evaluated side-by-side using single sources of
Human noroviruses
huNoV RNA stock extracted from local clinical isolates. Under optimized conditions, these
RT-PCR
RT-PCR protocols shared a very different pattern of detection sensitivity for huNoV. The
Norovirus detection
primer sets COG2F/COG2R and QNIF4/NV1LCR were determined to be the most sensitive
Urban wastewater
ones for huNoV GII and GI, respectively, with up to 105- and 106-fold more sensitive as
Recreational water
compared to other sets tested. These two sensitive protocols were validated by positive detection of huNoV in untreated and treated urban wastewater samples. In addition, these RT-PCR protocols enabled detection of the prevalence of huNoV in 5 (GI) and 10 (GII) of 16 recreational water samples collected around the island of O’ahu, which was confirmed by DNA sequencing and sequence analysis. Findings from this study support the possible use of enteric viral pathogens for environmental monitoring and argue the importance and essentiality for such monitoring activity to ensure a safe use of recreational waters. Published by Elsevier Ltd.
1.
Introduction
The current recreational water quality criteria requires growth-based measurement of the traditional fecal indicator bacteria (FIB) Escherichia coli and enterococci to assess the degree of sewage and sewage-borne pathogen contamination with respect to the risk of exposure. Despite being widely used as the standard procedure since the establishment of ambient
* Corresponding author. Tel.: +1 (808) 956 2702; fax: +1 (808) 956 5818. E-mail address:
[email protected] (Y. Lu). 0043-1354/$ e see front matter Published by Elsevier Ltd. doi:10.1016/j.watres.2011.08.030
water quality criteria (AWQC; U.S. EPA, 1986), this monitoring system has its limitations in protecting the public from recreational waterborne illness. The traditional culture techniques require at least 18e24 h to perform. Thus beach managers can only issue swimming advisories based on information that is at best 1 day old (Boehm et al., 2002; Dorevitch et al., 2010; Kim and Grant, 2004). Furthermore, FIB have been reported to persist and grow in the
5838
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 3 7 e5 8 4 8
environments, which bias assessment, especially in tropical regions like Hawaii (Byappanahalli and Fukioka, 2004; Davies et al., 1995; Desmarais et al., 2002; Hardina and Fujioka, 1991; Yamahara et al., 2009). Human enteric viruses such as noroviruses (huNoV), enteroviruses and adenoviruses have been suggested as alternative indicators of fecal contamination in aquatic environments (Fong and Lipp, 2005; Jiang et al., 2001). They were found to survive in the environment for long period of time and can tolerate changing environmental conditions (Jiang et al., 2001; Kocwa-Haluch, 2001; Melnick and Gerba, 1989). Human viral pathogens are free of environmental multiplication due to the lack of permissive host organisms in water. Human enteric viruses are directly associated with human disease and can be detected using molecular-based methods such as PCR and RT-PCR. These consumptive methods are known to be more rapid as compared to traditional cultivationbased detections. Therefore, human enteric viruses could be potentially employed as bio-indicators to monitor fecal contamination in recreational waters, and to facilitate the development of a simplified health risk assessment for beach contamination management. In the present study, huNoV is tested for its potential use as a model enteric virus for recreational water quality monitoring. huNoV infection is reported to be the leading cause of nonbacterial gastroenteritis (Fankhauser et al., 2002; Patel et al., 2009), and has become an increasing public health problem worldwide. Previously known as Norwalk-like virus, this nonenveloped, positive sense single stranded RNA virus is classified as a member belonging to the genus Norovirus in Caliciviridae family, and includes 5 genogroups, GI to GV. huNoV genogroups I, II, and IV are reported to infect humans of all ages (Patel et al., 2009). Despite the lack of a cell culture system for direct infectivity studies, huNoV, similar to most enteric viruses, is believed to be relatively stable in the environments (Cannon et al., 2006). huNoV surrogates Murine Norovirus (MNV-1) and feline calicivirus (FeCV) survive in both high and low pH environments and a wide range of temperature (Cannon et al., 2006; Cox et al., 2009; Duizer et al., 2004). It was also documented in literature that inactivation of these huNoV surrogates by 70% ethanol and sodium hypochlorite solution <300 ppm could be inefficient depending on the treatment time (Doultree et al., 1999; Duizer et al., 2004). The ability of longterm survival in suspension in the environment, its prolonged duration of viral shedding into feces of infected patients even after symptoms resolve, and the low infectious dose (<10 viral particle) (Teunis et al., 2008) can all facilitate easy transmission of huNoV through contaminated food or water. Cases of gastroenteritis associated with huNoV contaminated drinking water have been reported worldwide, including in the United States (Carrique-Mas et al., 2003; Kukkula et al., 1999; Parshionikar et al., 2003). huNoV is also identified as the leading cause of food-borne diseases outbreaks directly associated with the consumption of contaminated bivalve shellfish including oysters (Jothikumar et al., 2005; Noda et al., 2008; Nenonen et al., 2009). Recent studies using molecular-based detection methods, mainly RT-PCR, have revealed the prevalence of huNoV in aquatic environments, including treated and untreated urban sewage water (da Silva et al., 2007), marine water (Katayama
et al., 2002), freshwater from rivers and lakes (Ho¨hne and Schreier, 2004; Ho¨rman et al., 2004), and tap water (Haramoto et al., 2004) in different regions of the world. However, the occurrence of huNoV in Hawaiian aquatic environments has not yet been reported. This study was aimed to establish a rapid and highly sensitive method for huNoV detection in environmental waters, and to use such protocols for day-to-day screening and monitoring of huNoV contamination in Hawaiian coastal and recreational waters.
2.
Materials and methods
2.1.
Viral nucleic acids
NoV GI.4 and Nov GII.4 RNA extracted from clinical isolates were kindly provided by the Hawaii State Department of Health. cDNA was prepared from these viral RNA using SuperScript II Reverse Transcriptase (Invitrogen, CA) with random hexamers (Invitrogen, CA) according to the manufacturer’s instructions. The final concentration of viral cDNA was determined by using a DU 800 spectrophotometer (Beckman Coulter, CA). The viral cDNA was stored at 20 C and used for the optimization of PCR conditions of the selected huNoV primer sets.
2.2.
Urban wastewater samples
Treated and untreated urban wastewater samples were collected from the Sand Island Wastewater Treatment Plant (SIWWTP, Hawaii) during May 2010. This water treatment plant processes 60 million gallons of wastewater daily, which is approximately 85% of Oahu’s wastewater. SIWWTP adopts an advanced primary treatment using ultraviolet (UV) radiation for disinfection before disposing the effluent via a deep (240 feet) ocean outfall 1.7 miles offshore into Mamala Bay. Samples were collected from 3 different treatment stages, including untreated influents, post-primary clarifying treatment/pre-disinfection, and post-disinfection/effluents. Each sample was collected in a sterile 2-L polypropylene container, transported on ice to a BSL-2 certified laboratory, and processed for sample concentration and nucleic acid extraction within 2 h.
2.3.
Environmental water samples
Surface water samples were collected from 16 different recreational water bodies around the island of O’ahu (Fig. 1). Sampling sites include seawaters from Sand Island State Park, Ala Moana Park near Magic Island, Diamond Head Beach Park, West Loch Community Shoreline Park, Maili Beach Park, Pokai’i Bay (near Waianae side), Wai’alae Beach Park, Manunalua Bay Beach Park, Bellows Field Beach Park, Kailua Bay, Waikiki Beach, and Kalaka Bay. Sites for freshwater include Kaelepulu Stream, Ala Wai Canal, Wahiawa freshwater (near the reservoir), and Palolo stream. All samples except for Palolo stream were collected in June 2010. The sample from Palolo stream was collected in February 2011. One field blank using 2 L of double distilled water and one spike sample were included as the negative and positive controls for the field
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 3 7 e5 8 4 8
5839
Fig. 1 e Geographic locations of environmental water sampling sites in O’ahu Island.
study, respectively. The spike sample was prepared by adding 100 ml huNoV GI/GII positive untreated wastewater samples into 2 L of seawater sample collected from Diamond Head Beach Park. Water samples were collected in 4-L sterile polypropylene containers, transported on ice to a BSL-2 certified laboratory and processed within 8 h.
2.4.
Sample concentration and nucleic acids extraction
Water samples were concentrated using a filtration-based method described previously (Katayama et al., 2002). Briefly, negatively-charged type HA filter membranes (Millipore Corporation, MA) with a 0.45 mm pore size and 47 mm in diameter were used with a vacuum pump system. MgCl2 was added into sewage and freshwater samples at a final concentration of 25 mM before filtration. One hundred milliliters of sewage water and 2.0 L of fresh and marine water samples were filtered through membranes for viral absorption. Mixtures of sample DNA and RNA were extracted from the recovered membranes using the PowerWater RNA isolation kit (Mo Bio laboratories, CA) according to the manufacturer’s instruction with modification. In brief, the on-column DNase I digestion step in the original protocol (steps 21e23) was skipped, and a final volume of 60 mL eluents containing mixtures of sample DNA and RNA was obtained. 15 mL of the eluent mixture was aliquot and stored at 80 C to later served as the DNA template for PCR reaction. The remaining 45 mL eluent was combined with 5 mL of 10 DNase I buffer (MoBio laboratories, CA) and 3 mL of DNase I (MoBio laboratories, CA), incubated at room temperature for 20 min. DNase I was then inactivated by incubating the reaction at 75 C for 5 min to obtain sample
RNA. The RNA samples were then stored at 80 C until RT-PCR was performed.
2.5. Comparison of different PCR protocols for huNoV detection Fourteen sets of published primer pairs targeting various huNoV genomic regions along with 2 newly designed sets of primers were comparatively tested in this study. As shown in Table 1, 6 primer sets were huNoV GI specific, 7 sets were huNoV GII specific, and 3 primer sets MJV12/RegA, ORF JA and ORF JB are both huNoV GI and GII specific. Primer sets ORF JA and ORF JB were new ones designed specifically for this study based on huNoV GI and GII consensus sequence located at the junction of ORF1 and ORF2 for the purpose of detecting these two noroviral genogroups through a single PCR reaction. All primer sets were optimized (details as described below) to identify the PCR conditions that provided for the sensitive detection of huNoV. PCR was performed with a MasterCycler Gradient (Eppendorf, Germany) to determine the optimal amplification conditions. All primer sets were first tested at different annealing temperatures using 1 mL of huNoV cDNA at a final concentration of 100 ng/25 mL combining with 24 mL of a mixture containing 1 Taq (Mg2þ free) reaction buffer (New England Biolabs, NEB, MA), 2 mM MgCl2 solution (NEB, MA), 200 nM of each dNTPs (SigmaeAldrich, MO), 400 nM of each primer (Integrated DNA technologies, IA), and 2 units of Taq polymerase (provided by Dr. Huang, University of Hawaii at Manoa). The Amplification started with an initial denaturation at 94 C for 5 min, followed by 40 cycles of denaturation at 94 C for 30 s, a gradient annealing temperature ranging
5840
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 3 7 e5 8 4 8
Table 1 e RT-PCR primers used in this study G
Primer
Sequence (50 /30 )a
þ/-b
I
MON 432 MON 434 QNIF4 NV1LCR COGIF COGIR CapA CapB2 NV65a NV120 NV1LCF NV1LCR MON 431 MON 433 QNIF2d COG2R COG2F COG2R CapC CapD3 NV107a NV117 NV2LCF NV2LCR Ni E3 MJV12 RegA ORF JAFd ORF JARd ORF JBFd ORF JBRd
TGGACICGYGGICCYAAYCA GAASCGCATCCARCGGAACAT CGCTGGATGCGNTTCCAT CCTTAGACGCCATCATCATTTAC CGYTGGATGCGNTTYCATGA CTTAGACGCCATCATCATTYAC GGCWGTTCCCACAGGCTT TATGTIGAYCCWGACAC TGGACAGGRGATCGCRATCT AYATCACCGGGGGTATTRTT CARGCCATGTTYCGYTGGATG CCTTAGACGCCATCATCATTTC TGGACIAGRGGICCYAAYCA GGAYCTCATCCAYCTGAACAT ATGTTCAGRTGGATGAGRTTCTCWGA TCGACGCCATCTTCATTCACA CARGARBCNATGTTYAGRTGGATGAG TCGACGCCATCTTCATTCACA CCTTYCCAKWTCCCAYGG TGYCTYITICCHCARGAATGG AGCCAATGTTCAGATGGATG TCGACGCCATCTTCATTCAC GARYCIATGTTYAGRTGGATG TCGACGCCATCTTCATTCAC GAATTCCATCGCCCACTGGCT ATCTCATCATCACCATA TAYCAYTATGATGCHGAYTA CTCRTCATCICCATARAAIGA ATCTGGCTCCCAGTTTGTG GCTCCAAAGCCATAACCTCA GGCTCCCAGTTTTGTGAATG GCTCCAAAGCCATAACCTCA
þ e þ e þ e þ e þ e þ e þ e þ e þ e þ e þ e þ e þ e þ e þ e þ e
I I I I I II II II II II II II I/IIc I/IIc I/IIc
a b c d
Target
References
RdRP
Beuret et al., 2002
Capsid
da Silva et al., 2007
ORF1/2
Kageyama et al., 2003
ORF2
Vinje´ et al., 2004
ORF2
Ho¨hne and Schreier, 2004
Capsid
Svraka et al., 2007
RdRP
Beuret et al., 2002
ORF1/2
Loisy et al., 2005
ORF1/2
Kageyama et al., 2003
ORF2
Vinje´ et al., 2004
ORF2
Ho¨hne and Schreier, 2004
Capsid
Svraka et al., 2007
RdRP
Green et al., 1995
RdRP
Vinje´ et al., 2004
ORF1/2
This study
ORF1/2
This study
R ¼ A þ G, Y]C þ T, S]C þ G, W ¼ A þ T, H ¼ A þ C þ T, B ¼ C þ G þ T, V ¼ A þ C þ G, D ¼ A þ G þ T, N ¼ A þ T þ C þ G, I ¼ inosine. Polarity. þ Forward primer, - reverse primer. Designed to detect both huNoV GI and GII. Primer designed for this study.
from 50 to 60 C, with approximately 2 C increment, for 30 s, extension at 72 C for 30 s, and a final extension at 72 C for 5 min. PCR products were subjected to 2% agarose gel electrophoresis, alongside a 50 bp DNA marker (NEB, MA), stained with ethidium bromide (EtBr) and viewed with the Molecular Imager Gel Doc XR þ system (BioRad Laboratories, Inc., CA). Using the optimal annealing temperature, each primer set was tested at different PCR buffer (1 and 1.6), MgCl2 (1.5, 2.0, 3.0, and 4.0 mM) and primer concentrations (0.2, 0.4, 0.6, 0.8, and 1.0 mM) in the presence or absence of 0.1 mg/mL of BSA. A 10-fold serial dilution of purified huNoV GI and GII cDNA was then tested under optimized conditions to evaluate the detection sensitivity of each PCR assay. The detection limits were based on the highest dilution that gave a clear positive signal after PCR amplification.
2.6.
RT-PCR conditions
huNoV GI specific primer set QNIF4/NV1LCR and huNoV GII specific primer set COG2F/COG2R were selected to detect huNoV in urban wastewater and environmental water samples. Seven microliter of extracted sample RNA was
reversely transcribed into cDNA using the DyNAmo cDNA synthesis kit (NEB, MA) following the manufacturer’s instructions. Five microliters of the sample cDNA was combined with 20 mL of PCR mixture of 1 Taq (Mg free) reaction buffer, 2 mM MgCl2, 200 nM dNTPs mix, 0.1 mg/mL BSA, 400 nM of each primer, and 2 units of Taq polymerase. The amplification cycle started with an initial denaturation at 94 C for 5 min, followed by 40 cycles of denaturing at 94 C for 30 s, annealing at 58 C for 30 s, extension at 72 C for 30 s, and a final extension at 72 C for 5 min. To enhance molecular cloning efficiency for DNA sequencing studies, an enhanced PCR amplification was carried out for those huNoV positive samples using the same primers under the PCR conditions as described above. E. coli specific primers URL301(TGTTACGTCCTGTAGAAAGCCC)/URR-432 (AAAACTGCCT GGCACAGCAATT) (Bej et al., 1991) targeting the E. coli uidR gene were used to detect E. coli DNA in urban wastewaters and environmental waters as amplification controls. Three microliters of sample DNA were combined with 22 mL of PCR mixture of 1 Taq (Mg2þ free) reaction buffer, 2.5 mM MgCl2, 200 nM dNTPs mix, 0.1 mg/mL BSA, 400 nM of each primer, and 2 units of Taq polymerase. The amplification cycle started with
5841
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 3 7 e5 8 4 8
an initial denaturation at 94 C for 5 min, followed by 35 cycles of denaturing at 94 C for 30 s, annealing at 60 C for 30 s, extension at 72 C for 30 s, and a final extension at 72 C for 5 min. PCR and enhanced PCR products were subjected to 2% agarose gel electrophoresis staining with EtBr.
2.7.
PCR products sequencing and analysis
To ensure the PCR amplification to be specific for huNoV, the amplicons from all the positive samples were subjected to DNA sequencing and analysis. In brief, huNoV positive PCR or enhanced PCR products from clinical, sewage and environmental isolates were recovered directly from the 2% agarose gel using the QIAquick Gel Extraction kit (Qiagen, CA) following the manufacturer’s instructions and eluted in 30 mL EB buffer. The recovered DNA was cloned into pCRÒ2.1TOPOÒ vector using TOPO TA CloningÒ kit (Invitrogen, CA) following the manufacturer’s instructions, and randomly selected positive clones were submitted along with the M13 forward primer provided in the commercial kit to the Advanced Studies in Genomics, Proteomics and Bioinformatics (ASGPB, University of Hawaii at Manoa) for sequencing. To further confirm the specificity of huNoV amplification and detection, the resulting sequences were analyzed by comparing these DNA fragments with the GeneBank sequences and EMBL databases using the BLAST program of the National Center for Biotechnolgoy Information (NCBI). An additional analysis of the complete amplicon sequences corresponding to Camberwell virus (GenBank no. AF145896) for huNoV GII and Norovirus sewage isolate virus (GenBank no. AB504701.1) was also performed with BLAST to search for sequence homology derived from organisms other than huNoV.
3.
Results
3.1. Optimized RT-PCR condition and their sensitivity for huNoV detection Because most published primer sets selected in this study were originally reported in real-time PCR assays, it is important to test them under general PCR condition first and to optimize each primer set individually by adjusting salt concentration, primer concentration, annealing temperature, and the amount of BSA enhancer addition. Positive results were obtained from 15 of the 16 primer sets tested using huNoV cDNAs, which were reversely transcribed from single sources of viral RNAs extracted from the local clinical huNoV GI and GII isolates. However, the 3 primer sets designed to detect both huNoV GI and GII, including the 2 newly designed primer sets ORF JA and ORF JA, and the published MJV12/RegA, were only able to detect huNoV GII but not GI. The optimal amplification conditions and comparative detection limits of each primer set are summarized in Table 2. In general, the addition of 0.1 mg/mL BSA into the PCR reaction mixture greatly enhanced amplification signal. The sensitivities of these PCR protocols were determined by using a 10-fold serial dilution of huNoV cDNA down to 0.1 pg (107-fold dilution). Detection limits for huNoV among these protocols varied significantly as expected; the amount of cDNA needed for a positive result ranged from 100 ng to 0.1 pg, indicating a 106-fold difference in detection sensitivity. Among all primers tested, sets GNIF2d/COG2R and COG2F/COG2R were the most sensitive primers for huNoV GII detection, and sets QNIF4/NV1LCR and CapA/CapB2 were identified to be the most sensitive PCR assays for huNoV GI. Under the optimized amplification conditions, these methods require only 0.1e10 pg of viral cDNA for positive detection of
Table 2 e Optimized PCR conditions and detection limits. Type
Primer set
Detection limitc
Optimized PCR condition [Taq rxn buffer]
[MgCl2]
[primer]
GI I I I I I
MON432/Mon434 QNIF4/NV1LCR COG1F/COG1R CapA/CapB2 NV65a/NV120 NV1LCF/NV1LCR
1 1 1 1 1.6 1
1.5 2.0 2.0 1.5 2.0 2.0
mM mM mM mM mM mM
400 800 400 400 400 400
nM nM nM nM nM nM
GII II II II II II II II II
MON431/MON433 GNIF2d/COG2R COG2F/COG2R NV107a/NV117 NV2LCF/NV2LCR E3/Ni MJV12/RegA ORF JA ORF JB
1 1 1 1.6 1.6 1 1 1 1
1.5 2.0 2.0 2.0 2.0 2.0 1.5 2.0 2.0
mM mM mM mM mM mM mM mM mM
400 nM 600 nM 800 nM 400 nM 1000 nM 1000 nM 400 nM 1000 nM 1000 nM
BSAa
TA ( C)b
e þ e e e þ
50e54 54e60 50e52 50 52e56 50e52
101 104105 103 104105 103 104
e þ þ þ þ e e e e
52e54 56e58 52e60 50e54 50e54 50 50e52 50 50
103 106107 106107 104105 104 101102 101 101 101102
a At final concentration of 0.1 mg/mL þ: Addition of BSA improved amplification signal. : addition of BSA did not influence resulting signal. b TA ¼ Annealing temperature. c Tested under optimized PCR condition with a 10X serial dilution of huNoV GI and GII cDNA. The detection limits were based on the highest dilution giving a clear positive signal after RT-PCR.
5842
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 3 7 e5 8 4 8
huNoV GII and GI, respectively. Furthermore, primer set QNIF4/NV1LCR and set COG2F/COG2R showed stronger detection signals and better product specificity compared to set CapA/CapB2 and GNIF2d/COG2R when the same amount of cDNA was used. Therefore, primer set QNIF4/NV1LCR (huNoV GI) and COG2F/COG2R (huNoV GII) were chosen for huNoV detection in sewage and environmental water samples. While establishing the optimal annealing temperatures for the above primer sets, a wide range of temperatures ranging from 48 C to 60 C was tested using pure huNoV RNA extracted from clinical samples. Results indicated that all the tested temperatures worked almost equally well for the targeted amplification. However, when testing these two protocols on total RNA extracted from a complex microbe community (such as from urban wastewater samples and seawater samples), annealing temperatures lower than 56 C for huNoV GI and 54 C for huNoV GII resulted in amplification of non-specific products (data not shown). Thus, an annealing temperature of 58 C was adopted in further studies when using the two protocols to detect huNoV in environmental water samples.
3.2. Detection of huNoV GI and GII in urban wastewaters Urban wastewater samples obtained from the Sand Island Wastewater Treatment Station were used to validate the established detection protocols. As shown in Fig. 2, huNoV GI and GII were detected in all three stages, including samples collected from the untreated influents tank, the postclarification/pre-disinfection tank, and the post-disinfection/ effluents tank. These specific PCR detections of huNoV GI and GII were confirmed through DNA sequencing. Positive detections of huNoV in post-disinfection/effluent samples could not be tested for viral infectivity due to the lack of any permissive in vitro cell culture for huNoV infection.
3.3. Detection of huNoV GI and GII in Hawaiian recreational waters Although human pathogenic viruses usually exist in low concentration in environmental waters, PCR amplification using the primer set of COG2F/COG2R under the optimized detection conditions revealed that huNoV GII was detected in 10 out of 16 samples. huNoV GII positive sites included seawaters collected from Ala Moana Park, Diamond Head Beach Park, Maili Beach Park, Poka’i Bay, Wai’alae Beach Park, Kalaka Bay, and freshwaters from Kaelepulu stream (in Kailua Bay), Ala Wai Canal, Wahiawa, and Palolo stream (Fig. 3). For huNoV GI detection, 5 of 16 samples were positive, including Sand Island Beach Park, Maili Beach Park, Bellows Field Beach Park, Waikiki Beach, and Palolo stream. Maili Beach Park and Palolo stream were the only sites to be positive for both huNoV GI and huNoV GII under the described detection conditions. The results of huNoV GI and GII detection in sewage and environmental samples are summarized in Table 3. All positive samples for huNoV detection were confirmed by DNA sequencing.
3.4. Detection of E. coli DNA in urban wastewaters and environmental waters To test whether PCR was inhibited in the environmental and wastewater templates, these samples were also tested for E. coli DNA as an internal control. E. coli was detected in wastewater samples from all treatment stages (Fig. S1A), as well as in all environmental samples (Fig. S1B). This finding confirmed that inhibitory substances were not present in templates enough to completely inhibit the PCR. This suggests that negative detection of huNoV in several samples was not due to the presence of PCR inhibitors from the environmental waters or an unsuccessful preparation of sample nucleic acids.
3.5.
Fig. 2 e Agarose gel electrophoresis of RT-PCR products from urban wastewaters. (A) huNoV GI and (B) huNoV GII were detected from 100 mL of untreated raw influents (lane 1), pre-disinfection/post-clarification treatment (lane 2), and post-disinfection/Effluents (lane 3). Lane M: 50 bp DNA marker, lane CD: positive control using cDNA from clinical huNoV isolates, and Lane C-: no template negative control.
Sequence analysis of huNoV RT-PCR amplicons
To further confirm the positive PCR results to be specific amplification products of huNoV, the amplicons of all positive samples for both huNoV GI and GII were cloned into pCRÒ2.1TOPOÒ vector (Invitrogen, CA) and sequenced. BLAST analysis revealed that resulting amplicon sequences from sewage, recreational waters and clinical samples all presented high sequence homology to huNoV strains/variants from the NCBI database (Fig. 4). Sequencing multiple clones from sewage samples showed the detection of several different huNoV variants. In particular, majority of the huNoV GI sewage clones (>10 clones) were identical to the clinical isolate (Fig. 4A). Meanwhile, huNoV GI sewage clones 5 and 6 share very high sequence homology with the huNoV GI variants detected from recreational waters (Fig. 4A, Env water 3&4) in Hawaii, which match perfectly with the huNoV GI strain NV/ Saitama T83GI/02/JP reported from Saitama Prefecture, Japan (Kageyama et al., 2004). A total of 10 clones of huNoV GII from clinical sample, urban wastewater influents and recreational waters were sequenced (Fig. 4B). BLAST results indicated while there is no any centralized match of these clones to any particular huNoV GII strain as observed for huNoV GI, all
5843
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 3 7 e5 8 4 8
Fig. 3 e Agarose gel electrophoresis of RT-PCR products of (A) initial amplification and (B) enhanced amplification from environmental surface water samples. huNoV GII was detected in Ala Moana Park, Maili Beach Park, Kaelepulu Stream, Kalaka Bay, Ala Wai Canal, and Wahiawa freshwater (lanes 2, 5, 11, 13, 14, 15) by RT-PCR and subsequent enhanced amplification. huNoV GII was only detected after enhanced amplification in Diamond Head Beach Park, Pokai Bay, and Wai’alae Beach park (lanes 3, 6, 7). huNoV GII was under detection in Sand Island State recreational area, West Loch Community Shoreline park, Maunalua Bay Beach Park, Bellows Field Beach Park, Kailua Bay, and Waikiki Beach (lanes 1, 4, 8, 9, 10, 12). Lane CD: Positive control with clinical isolated huNoV GII cDNA, Lane C-: no template control, Lane FCD: Field positive control with spike samples, and Lane FC-: Field Blank control.
detected sequences matched highly with various huNoV GII strains/variants reported previously in literature (NCBI recession numbers EF028232.1, HM624049.1, GU980585.1, DQ379713.1, GU017738.1, AB122162.1, AF414408.1, HM635126.1, AB542918.1, HM635121.1) (Phan et al., 2007; Yun
et al., 2010; Symes et al., 2007; Kageyama et al., 2004; Ando et al., 1995; Han et al., 2011; Iritani et al., 2010). The results also show that despite of no perfect match with the clinical isolates and sewage clones, the 3 environmental isolates shared a high sequence homology with sewage clone 6
Table 3 e Detection of huNoV GI and GII in O’ahu urban wastewaters and recreational waters. Sample Site SIWWTP influents tank SIWWTP clarifying tank SIWWTP effluents tank 1. Sand Island State Recreational Area 2. Ala Moan Park/Magic Island 3. Diamond Head Beach Park 4. West Loch Community Shoreline Park 5. Maili Beach Park 6. Waianae/Pokai Bay 7. Waialae Beach Park 8. Maunalua Bay Beach Park 9. Bellows Field Beach Park 10. Kailua Bay 11. Waikiki Beach 12. Kalaka Bay Beach Park 13. Kaelepulu Stream 14. Ala Wai Canal 15. Wahiawa freshwater 16. Manoa/Palolo stream Fb. Field Blank S. Spike control
Condition Sewage- Untreated Sewage- Clarified Sewage- Post disinfection Sea water- open ocean Sea water- cove Sea water- open ocean Sea water- salt water lake Sea water- open ocean Sea water- open ocean Sea water- open ocean Sea water- marina entry Sea water- estuary Sea water- open ocean Sea water- open ocean Sea water- estuary Fresh water-stream Fresh water- canal Fresh water- lake Fresh water-stream Double distilled water Sewage þ sea water
Sample volume
GI
GII
100 mL 100 mL 100 mL 2.00 L 2.00 L 2.00 L 0.80 La 2.00 L 2.00 L 2.00 L 2.00 L 2.00 L 2.00 L 2.00 L 2.00 L 2.00 L 2.00 L 2.00 L 2.00 L 2.00 L 100 mL þ 2.00 L
D D D D e e e D e e e D e D e e e e þ e D
D D D e D D e D D D e e e e D D D D D e D
a The maximum volume could pass through the filter membrane. Cannot reject possibility that negative detections come from insufficient sample volumes.
5844
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 3 7 e5 8 4 8
Fig. 4 e Nucleotide sequence alignment for (A) huNoV GI and (B) huNoV GII isolates. Dots indicate to be identical to clinical sample isolate. Arrows show locations of the primers. Sewage clones 1 to 6 were isolated from the same untreated urban wastewater sample, env water 1 was isolated from Ala Moana Beach park/Magic island, env water 2 was isolated from Kalaka Bay Beach park, env water 3 was isolated from Sand Island State Recreational Area, and env water 4 was isolated from Palolo stream.
(Fig. 4B, EW1, 2, 4 & sewage clone 6), which matches perfectly with a reported huNoV GII isolated from Baltimore, USA (Ando et al., 1995). In addition, PCR products of different sizes (visible as the lower bands of smaller molecular weights in the gel in Fig. 3B) were also sequenced using the described method in this study. However, comparative sequence analysis revealed that these sequences had no homologous match with huNoV, suggesting these smaller sized amplicoms are non-specific
PCR products. Therefore, only the PCR products of the reported sizes (86 bp for huNoV GI and 97 bp for huNoV GII) were the true positive detection.
4.
Discussion
Fecal indicator bacteria E. coli and enterococci are used to assess ambient water quality throughout the United States
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 3 7 e5 8 4 8
(U.S. EPA, 2007). These organisms are not etiological agents of disease, but their concentrations correlate to beachgoer illness rates in recreational water epidemiology studies. The use of these indicators has been criticized for a number of reasons (Boehm et al., 2002) including (1) they are not pathogens and do not affect health directly, and (2) they are bacteria while most waterborne recreational illnesses are thought to be viral. Therefore, it could be extremely informative to monitor recreational waters for human viral pathogens rather than bacterial indicators. Compared to fecal indicator organisms, pathogens usually exist in the environment in much lower and more variable numbers, making monitoring these in recreational waters challenging (U.S. EPA, 2007). To overcome this limitation, highly sensitive protocols for effective viral detection must be established. There are numerous RT-PCR protocols available for specific detection of human huNoV to date, however, little is known regarding their detection efficiency and sensitivity. An international collaborative study comparing different RTPCR assays was conducted by Vinje´ and co-workers (Vinje´ et al., 2003), but newly developed protocols using more broadly reactive huNoV detecting primers were not included. Therefore, this study was directed to address the need for identifying and establishing the most sensitive protocol for detecting human enteric viruses in water environments through the use of huNoV as a model by comparing protocols presently available in a side-by-side fashion. The PCR conditions for the 16 published and newly designed primer sets targeting different regions in huNoV genome were individually optimized prior to the determination of their detection sensitivities. As expected, the detection limits among these PCR primer sets varied significantly. The most sensitive primer sets for huNoV GII are GNIF2d/COG2R and COG2F/ COG2R, presenting a 500,000 fold higher detection sensitivity than the least sensitive published primers MJV12/RegA. As for huNoV GI, QNIF4/NV1LCR and CapA/CapB2 showed over 5 to 5000 times higher sensitivity than other published protocols. Both QNIF4/NV1LCR and CapA/CapB2 were then validated for their application in detecting huNoV GI in environmental waters. Results from this study demonstrated that the primer set of QNIF4/NV1LCR is more suitable for huNoV GI detection in environmental waters compared to the other primer sets since this set of primer pair is able to generate more specific amplification products. When the two sets of primers COG2F/COG2R and GNIF2d/COG2R are compared for huNoV GII detection, the former set appears to be superior to the latter as evidenced by the stronger bands detected under a similar amplification conditions. In summary, our experimental data strongly suggest COG2F/ COG2R and QNIF4/NV1LCR to be the most sensitive primer sets for huNoV GII and huNoV GI detection, respectively, and they should be particularly considered in testing environmental waters for huNoV contamination in the future. It should be noted that this comparative study of different PCR primer sets for their sensitivity in huNoV detection is based on the limited viral strains detectable in Hawaii, there is a need to validate our findings by testing more huNoV GI and GII strains in order to the full establishment of the suitable and wide use of these protocols for the detection of huNoV in water environment.
5845
Validation of these effective detection methods was conducted by assaying urban wastewater samples using a very simple filtration-based method with a negatively-charged filter membrane and a vacuum pump, followed by extracting viral RNA with the MoBio PowerWater RNA isolation kit. Both huNoV GI and huNoV GII RNA were detected in all stages tested, regardless of the disinfecting treatment. As a comparison, we also included polyethylene glycol (PEG) precipitation method in this study because of its wide use in viral concentration from urban wastewater (Aw and Gin, 2010; da Silva et al., 2007; Lewis and Metcalf, 1988; Monpoeho et al., 2001; Shieh et al., 1995). As a result, both huNoV GI and GII RNA were detected from the untreated urban wastewater sample, and End-point PCR assay suggested that these two sample concentration methods (PEG precipitation and membrane filtration) works equally well in concentrating HuNoV from wastewater samples (data not shown). For more turbid samples such as sludge or sewage, PEG precipitation could be the method of choice due to its relatively easy-to-perform nature and suitability to smaller sample volume. On the other hand, the membrane filtration-based method was adapted in this study since it is a relatively more rapid and practical technique for concentrating viruses from sample with large volume, such as recreational water samples. With the establishment of sensitive detection protocols, we carried out a microbiological water quality surveillance to examine the prevalence of huNoV GI and GII in recreational surface waters around the island of O’ahu. Among the 16 sites tested, 10 sites were found to be positive for GII and 5 sites for GI. It should be noted that this represents the first report of detecting huNoV in Hawaiian environmental waters. PCR amplification of nucleic acids extracted from wastewater and environmental water samples was also conducted using primers targeting to E. coli, a microorganism commonly found in environmental waters. This test showed the positive detection of E. coli in all samples, suggesting the experimental procedure employed for processing water samples efficiently removed/reduced PCR inhibitors possibly present in the samples. In addition, this validation test confirmed the prevalence of E. coli detected in the environments as reported previously, as well as the concerns of FIB environmental multiplicity in Hawaiian recreational waters (Byappanahalli and Fukioka, 2004). Due to the lack of biological system such as in vitro cell culture or a small animal model system, it is not possible to determine if the detected huNoV from sewage effluents and surface waters were infectious. Nevertheless, positive detections of huNoV represent warning signs of possible fecal and sewage-borne pathogen contaminations of these recreational waters. Thus, beach goers at these sites are potentially exposed not only to detectable amounts of huNoV, but also possibly to other enteric pathogens, putting them at risk of infection. It should be mentioned that several primer sets designed for real-time PCR assays were tested for their possible use as a sensitive tool for human huNoV detection under regular PCR condition in this present study, which is part of an on-going research project directed to the development of a PCR-based array for simultaneous detection of multiple waterborne pathogens and pathogen indicators, including huHuNoV. The main objective of this PCR array is to provide a simple and
5846
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 3 7 e5 8 4 8
easy-to-used platform for a quick screening of different water bodies for possible contamination by human sewage. Because such an array is designed for field use in a routinely manner, conventional PCR represents a needed technique which is undoubtedly more practical and accessible to many environmental monitor stations or laboratories around the world as compared to real-time PCR. In summary, comparative analysis of different RT-PCR protocols currently available for huNoV detection has led to the identification of highly sensitive RT-PCR protocols for effective detection of huNoV GI and GII in recreational waters. Identification of these highly sensitive and specific huNoV detection protocols from this study will be extremely useful for all researchers interested in huNoV monitoring and detection in the future. Findings from this study also suggest that huNoV contamination is more serious than previously expected in Hawaiian recreational waters. Since the source of huNoV is restrictively from human fecal contamination, all positive huNoV sites may also contain other human fecal borne pathogens, which raise the concerns of swimmer safety. Further monitoring those sites positive for the microbial water quality including detection of huNoV RNA and possible human fecal contamination are highly recommended. In addition, this study demonstrated that it is not impossible to detect enteric viruses from aquatic environments despite the highly diluted nature of these human pathogens in waters. Findings from this study warrant more in-depth test for the possible use of huNoV as a valid indicator for enteric virus prevalence when monitoring sewage and sewage-borne pathogen contaminations in recreational waters. Evidence showed the occurrence of huNoV contamination in Hawaiian waters and argues the necessity to monitor recreational waters for huNoV using these established sensitive huNoV detection protocols, therefore to protect the public from recreational waterborne illness associated with enteric viruses.
5.
Conclusions
1. Comparative analysis and the establishment of the most sensitive RT-PCR protocols for enhanced detection of huNoV GI and GII in environmental waters. 2. Recommendation of primer sets of QNIF4/NV1LCR (CGCTGGATGCGNTTCCAT/CCTTAGACGCCATCATCATTTAC) and COG2F/COG2R (CARGARBCNATGTTYAGRTGGATGAG/TCGACGCCATCTTCATTCACA) for specifically optimal detection of huNoV GI and GII, respectively, for aquatic environmental monitoring in future. 3. huNoV GI and GII RNA were detected in all the stages of urban wastewater treatment including its treated effluents. 4. First report of detection of huNoV in O’ahu recreational and coastal waters, suggesting a possible occurrence of human contamination. 5. Concept-demonstration of possible use of human enteric viruses as a possible indicator for monitoring coastal water quality. 6. Future surveillance of enteric virus contamination on the huNoV positive sites using the established huNoV detection protocol is highly recommended.
Acknowledgment The authors would like to thank Dr. Christian A. Whelen for the cDNA of clinical noroviral strains and Courtney Cox for her assistance in preparing the new huNoV primers for this study. We also thank Dr. Emily Viau and Mo Bio Laboratories, Inc. Technical team for technical consultation, Rey-Ian Transfiguracion for his assistance on field sampling, and Mary Margaret Byron and June Lee for reviewing the manuscript. This work was supported in part by grants from the Centers for Oceans and Human Health (COHH) program, the National Institutes of Environmental Health Sciences (P50ES012740) and the National Science Foundation (OCE04-32479 and OCE09-11000).
Appendix. Supplementary material Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.watres.2011.08.030
references
Ando, T., Jin, Q., Gentsch, J.R., Monroe, S.S., Noel, J.S., Dowell, S.F., Cicirello, H.G., Kohn, M.A., Glass, R.I., 1995. Epidemiologic applications of huNoVel molecular methods to detect and differentiate small round structured viruses (Norwalk-like viruses). Journal of Medical Virology 47 (2), 145e152. Aw, T.G., Gin, K.Y.H., 2010. Environmental surveillance and molecular characterization of human enteric viruses in tropical urban wastewaters. Journal of Applied Microbiology 109, 716e730. 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. Applied and Environmental Microbiology 57 (4), 1013e1017. Beuret, C., Kohler, D., Baumgartner, A., Lu¨thi, M., 2002. Norwalklike virus sequences in mineral waters: one-year monitoring of three brands. Applied and Environmental Microbiology 68 (4), 1925e1931. Boehm, A.B., Grant, S.B., Kim, J.-H., Mowbray, S.L., McGee, C.D., Clark, C.D., Foley, D.M., Wellman, D.E., 2002. Decadal and shorter period variability of surf zoen water quality at Huntington Beach, California. Environmental Science & Technology 36, 3885e3892. Byappanahalli, M., Fukioka, R., 2004. Indigenous soil bacteria and low moisture may limit but allow faecal bacteria to multiply and become a minor population in tropical soils. Water Science and Technology 50, 27e32. Cannon, J.L., Papafragkou, E., Park, G.W., Osborne, J., Jaykus, L., Vinje´, J., 2006. Surrogates for the study of Norovirus stability and inactivation in the environment: a comparison of murine norovirus and feline calicivirus. Journal of Food Protection 69, 2761e2765. Carrique-Mas, J., Andersson, Y., Petersen, B., Hedlund, K.O., Sjogren, N., Giesecke, J., 2003. A Norwalk-like virus waterborne community outbreak in a Swedish village during peak holiday season. Epidemiology and Infection 131, 737e744. Cox, C., Cao, S.B., Lu, Y., 2009. Enhanced detection and study of murine norovirus-1 using a more efficient microglial cell line. Virology Journal 6, 196e202.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 3 7 e5 8 4 8
da Silva, A.K., Le Saux, J.C., Parnaudeau, S., Pommepuy, M., Elimelech, M., Le Guyader, F.S., 2007. Evaluation of removal of noroviruses during wastewater treatment, using real-time reverse transcription-PCR: different behaviors of genogroups I and II. Applied and Environmental Microbiology 73, 7891e7897. Davies, C.M., Long, J.A., Donald, M., Ashbolt, N.J., 1995. Survival of fecal microorganisms in marine and freshwater sediments. Applied and Environmental Microbiology 61, 1888e1896. Desmarais, T.R., Solo-Gabriele, H.M., Palmer, C.J., 2002. Influence of soil on fecal indicator organisms in a tidally influenced subtropical environment. Applied and Environmental Microbiology 68, 1165e1172. Dorevitch, S., Ashbolt, N.J., Ferguson, C.M., Fukioka, R., McGee, C. D., Soller, J.A., Whitman, R.L., 2010. Meeting report: knowledge and gaps in developing microbial criteria for inland recreational waters. Environmental Health Perspectives 118, 871e876. Doultree, J.C., Druce, J.D., Brich, C.J., Bowden, D.S., Marshall, J.A., 1999. Inactivation of feline calicivirus, a Norwalk virus surrogate. Journal of Hospital Infection 41, 51e57. Duizer, E., Kijkerk, P., Rockx, B., De Groot, A., Twisk, F., Koopmans, M., 2004. Inactivation of caliciviruses. Applied and Environmental Microbiology 70, 4538e4543. Fankhauser, R.L., Monroe, S.S., Noel, J.S., Humphrey, C.D., Bresee, J.S., Parashar, U.D., Ando, T., Glass, R.I., 2002. Epidemiologic and molecular trends of "Norwalk-like viruses" associated with outbreaks of gastroenteritis in the United States. Journal of Infectious Diseases 186, 1e7. Fong, T.T., Lipp, E.K., 2005. Enteric viruses of humans and animals in aquatic environments: health risks, detection, and potential water quality assessment tools. Microbiology and Molecular Biology Reviews 69, 357e371. Green, S.M., Lambden, P.R., Deng, Y., Lowes, J.A., Lineham, S., Bushell, J., Rogers, J., Caul, E.O., Ashley, C.R., Clarke, I.N., 1995. Polymerase chain reaction detection of small roundstructured viruses from two related hospital outbreaks of gastroenteritis using inosine-containing primers. Journal of Medical Virology 45, 197e202. Han, T.H., Kim, C.H., Chung, J.Y., Park, S.H., Hwang, E.S., 2011. Emergence of norovirus GII-4/2008 variant and recombinant strains in Seoul, Korea. Archives of Virology 156 (2), 239e323. Haramoto, E., Katayama, H., Ohgaki, S., 2004. Detection of noroviruses in tap water in Japan by means of a new method for concentrating enteric viruses in large volumes of freshwater. Applied and Environmental Microbiology 70, 2154e2160. Hardina, C.M., Fujioka, R.S., 1991. Soil: the environmental source of E. coli and enterococci in Hawaii’s streams. Environmental Toxicology 6, 185e195. Ho¨hne, M., Schreier, E., 2004. Detection and characterization of Norovirus outbreaks in Germany: application of a one-tube RT-PCR using a fluorogenic real-time detection system. Journal of Medical Virology 72, 312e319. Ho¨rman, A., Rimhanen-Finne, R., Maunula, L., von Bonsdorff, C.H., Torvela, N., Heikinheimo, A., Ha¨nninen, M.L., 2004. Campylobacter spp., Giardia spp., Cryptosporidium spp., noroviruses, and indicator organisms in surface water in southwestern Finland, 2000-2001. Applied and Environmental Microbiology 70, 87e95. Iritani, N., Kaida, A., Kubo, H., Abe, N., Goto, K., Ogura, H., Seto, Y., 2010. Molecular epidemiology of noroviruses detected in seasonal outbreaks of acute nonbacterial gastroenteritis in Osaka City, Japan, from 1996-1997 to 2008-2009. Journal of Medical Virology 82 (12), 2097e2105. Jiang, S., Noble, R., Chu, W., 2001. Human adehuNoViruses and coliphages in urban runoff-impacted coastal waters of
5847
Southern California. Applied and Environmental Microbiology 67, 179e184. Jothikumar, N., Lowther, J.A., Henshilwood, K., Lees, D.N., Hill, V. R., Vinje´, J., 2005. Rapid and sensitive detection of Noroviruses by using TaqMan-based one-step reverse transcription-PCR assays and application to naturally contaminated shellfish samples. Applied and Environmental Microbiology 71, 1870e1875. Kageyama, T., Kojima, S., Shinohara, M., Uchida, K., Fukushi, S., Hoshino, F.B., Takeda, N., Katayama, K., 2003. Broadly reactive and highly sensitive assay for norwalk-like viruses based on real-time quantitative reverse transcription-PCR. Journal of Clinical Microbiology 41, 1548e1557. Kageyama, T., Shinohara, M., Uchida, K., Fukushi, S., Hoshino, F. B., Kojima, S., Takai, R., Oka, T., Takeda, N., Katayama, K., 2004. Coexistence of multiple genotypes, including newly identified genotypes, in outbreaks of gastroenteritis due to Norovirus in Japan. Journal of Clinical Microbiology 42 (7), 2988e2995. Katayama, H., Shimasaki, A., Ohgaki, S., 2002. Development of a virus concentration method and its application to detection of enterovirus and norwalk virus from coastal seawater. Applied and Environmental Microbiology 68, 1033e1039. Kim, J.H., Grant, S.B., 2004. Public mis-notification of coastal water quality: a probabilistic evaluation of posting errors at Huntington Beach, California. Environmental Science and Technology 38, 2497e2504. Kocwa-Haluch, R., 2001. Waterborne enteroviruses as a Hazard for human health. Polish Journal of Environmental Studies 10, 485e487. Kukkula, M., Maunula, L., Silvennoinen, E., von Bonsdorff, C.H., 1999. Outbreak of viral gastroenteritis due to drinking water contaminated by Norwalk-like viruses. Journal of Infectious Diseases 180, 1771e1776. Lewis, G.D., Metcalf, T.G., 1988. Polyethylene glycol precipitation for recovery of pathogenic viruses, including hepatitis A virus and human Rotavirus, from oyster, water, and sediment samples. Applied and Environmental Microbiology 54, 1983e1988. Loisy, F., Atmar, R.L., Guillon, P., Le Cann, P., Pommepuy, M., Le Guyader, F.S., 2005. Real-time RT-PCR for norovirus screening in shellfish. Journal of Virological Methods 123, 1e7. Melnick, J.L., Gerba, C.P., 1980. Ecology of enteroviruses in natural waters. Critical Reviews in Environmental Control 10, 65e93. Monpoeho, S., Maul, A., Mignotte-Cadiergues, B., Schwartzbrod, L. , Billaudel, S., Ferre´, V., 2001. Best viral elution method available for quantification of enteroviruses in sludge by both cell culture and reverse transcription-PCR. Applied and Environmental Microbiology 67, 2484e2488. Nenonen, N.P., Hannoun, C., Olsson, M.B., Bergstrom, T., 2009. Molecular analysis of an oyster-related norovirus outbreak. Journal of Clinical Virology 45 (2), 105e108. Noda, M., Fukuda, S., Nishio, O., 2008. Statistical analysis of attach rate in norovirus foodborne outbreaks. International Journal of Food Micrology 122, 216e220. Parshionikar, S.U., Wiilian-Ture, S., Fout, G.S., Robbins, D.E., Seys, S.A., Cassady, J.D., Harris, R., 2003. Waterborne outbreak of gastroenteritis associated with a norovirus. Applied and Environmental Microbiology 69, 5263e5268. Patel, M.M., Hall, A.J., Vinje´, J., Parashar, U.D., 2009. Noroviruses: a comprehensive review. Journal of Clinical Virology 44, 1e8. Phan, T.G., Kaneshi, K., Ueda, Y., Nakaya, S., Nishimura, S., Yamamoto, A., Sugita, K., Kakanashi, S., Okitsu, S., Ushijima, H., 2007. Genetic heterogeneity, evolution, and recombination in noroviruses. Jouranl of Medical Virology 79 (9), 1388e1400. Svraka, S., Duizer, E., Vennema, H., de Bruin, E., van der Veer, B., Dorresteijn, B., Koopmans, M., 2007. Etiological role of viruses
5848
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 3 7 e5 8 4 8
in outbreaks of acute gastroenteritis in the Netherlands from 1994 through 2005. Journal of Clinical Microbiology 45, 1389e1394. Shieh, Y.S.C., Wait, D., Tai, L., Sobsey, M.D., 1995. Methods to remove inhibitors in sewage and other fecal wastes for enterovirus detection by the polymerase chain reaction. Journal of Virological Methods 54, 51e66. Symes, S.J., Gunesekere, I.C., Marshall, J.A., Wright, P.J., 2007. Norovirus mixed infection in an oyster-associated outbreak: an opportunity for recombination. Archives of Virology 152 (6), 1075e1086. Teunis, P.F.M., Moe, C.L., Liu, P., Miller, S.E., Lindesmith, L., Baric, R.S., Le Pendu, J., Calderon, R.L., 2008. Norwalk Virus: How infectious is it? Journal of Medical Virology 80, 1468e1476. U.S. EPA, 1986. Ambient water quality criteria for bacteria. EPA440/5-84-002. US EPA, Washington, DC. U.S. EPA (2007) Report of the experts scientific workshop on critical research needs for the development of new or revised recreational water quality criteria. EPA 823-R-07e006, United States Environmental Protection Agency, Washington D.C.
Vinje´, J., Vennema, H., Maunula, L., von Bonsdorff, C., Hoehne, M., Schreier, E., Richards, A., Green, J., Brown, D., Beard, S.S., Koopmans, M.P.G., 2003. International collaborative study to compare reverse transcriptase PCR assyas for detection and genotyping of Noroviruses. Journal of Clinical Microbiology 41, 1423e1433. Vinje´, J., Hamidjaja, R.A., Sobsey, M.D., 2004. Development and application of a capsid VP1 (region D) based reverse transcription PCR assay for genotyping of genogroup I and II noroviruses. Journal of Virological Methods 116, 109e117. Yamahara, K.M., Walters, S.P., Boehm, A.B., 2009. Growth of enterococci in unaltered, unseeded beach sands subjected to tidal wetting. Applied and Environmental Microbiology 75, 1517e1524. Yun, S.I., Kim, J.K., Song, B.H., Jeong, A.Y., Jee, Y.M., Lee, C.H., Paik, S.Y., Koo, Y., Jeon, I., Byun, S.J., Lee, Y.M., 2010. Complete genome sequence and phylogenetic analysis of a recombinant Korea norovirus, CBNU1, recovered from a 2006 outbreak. Virus Research 152 (1e2), 137e152.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 4 9 e5 8 5 7
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Ozone treatment ameliorates oil sands process water toxicity to the mammalian immune system Erick Garcia-Garcia a,1, Jun Qing Ge a,d,1, Ayoola Oladiran a, Benjamin Montgomery a, Mohamed Gamal El-Din b, Leonidas C. Perez-Estrada b, James L. Stafford a, Jonathan W. Martin c, Miodrag Belosevic a,* a
Department of Biological Sciences, University of Alberta, Canada Department of Civil and Environmental Engineering, University of Alberta, Canada c Department of Laboratory Medicine and Pathology, University of Alberta, Canada d Institute of Biotechnology, Fujian Academy of Agricultural Sciences, Fuzhou 350003, China b
article info
abstract
Article history:
We evaluated whether ozonation ameliorated the effects of the organic fraction of oil
Received 29 June 2011
sands process water (OSPW) on immune functions of mice. Ozonation of OSPW eliminated
Received in revised form
the capacity of its organic fraction to affect various mouse bone marrow-derived macro-
15 August 2011
phage (BMDM) functions in vitro. These included the production of nitric oxide and the
Accepted 19 August 2011
expression of inducible nitric oxide synthase, the production of reactive oxygen interme-
Available online 7 September 2011
diates and the expression of NADPH oxidase subunits, phagocytosis, and the expression of pro-inflammatory cytokine genes. Ozone treatment also eliminated the ability of OSPW
Keywords:
organic fraction to down-regulate the expression of various pro-inflammatory cytokine
Advanced oxidation
and chemokine genes in the liver of mice, one week after oral exposure. We conclude that
Ozone
ozone treatment may be a valuable process for the remediation of large volumes of OSPW.
Immunotoxicity
ª 2011 Elsevier Ltd. All rights reserved.
Water remediation Water purification Tailing ponds
1.
Introduction
The remediation of the oil sands tailings ponds in Northern Alberta, Canada, constitutes a formidable environmental challenge for the oil sands surface mining industry. There is
great interest in developing efficient technologies that help speed up the transformation of tailing ponds into reclaimed aquatic systems (Gosselin et al., 2010; Quagraine et al., 2005). If tailings ponds are to be reclaimed (Gosselin et al., 2010), an extensive set of toxicological tests and bioassays should be
Abbreviations: BMDM, bone marrow-derived macrophages; CCL, Chemokine (C-C motif) ligand; IFNg, gamma interferon; IL, interleukin; LPS, lipopolysaccharide; NAs, naphthenic acids; NADPH, nicotin-amide adenine dinucleotide phosphate; NBT, Nitroblue Tetrazolium; NK, natural killer; NO, nitric oxide; OSPW, oil sands process water; OF, organic fraction; OSPW þ O3, ozonated OSPW; PBS, phosphate-buffered saline; PMA, phorbol myristate acetate; ROI, reactive oxygen intermediates; RQ, relative quantitation; TNF-a, tumor necrosis factor alpha. * Corresponding author. Department of Biological Sciences, CW-405 Biological Sciences Building, University of Alberta, Edmonton, AB T6G 2E9, Canada. Tel.: þ1 780 492 1266; fax: þ1 780 492 9234. E-mail address:
[email protected] (M. Belosevic). 1 Contributed equally to the work. 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.08.032
5850
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 4 9 e5 8 5 7
devised, ideally encompassing a variety of organisms that would likely be exposed to remediated OSPW. In this regard, mice represent the classical mammalian model organism, and their biology is considered representative of terrestrial mammals, including humans. Naphthenic acids (NAs) are believed to be the major toxicants in the OSPW generated during bitumen extraction (Madill et al., 2001). NAs are natural water soluble constituents of petroleum or bitumen, and these become highly concentrated in tailing ponds (Clemente and Fedorak, 2005). NAs exist as complex mixtures of cyclic and acyclic alkylsubstituted aliphatic carboxylic acids with the general formula CmH2mþZO2; where m is the number of carbon atoms, and Z is zero or an even negative integer indicating the number of hydrogen atoms lost during the formation of cyclic structures (Headley and McMartin, 2004). We recently reported that the organic fraction of oil sands process water (OSPW-OF) has immunotoxic properties in vitro and in vivo, affecting various macrophage microbicidal functions, and immune gene expression in different organs (Garcia-Garcia et al., 2011a, 2011b). NAs account for less than 50% of all the compounds in the OSPW-OF (Grewer et al., 2010; Headley et al., 2009). Our previous comparison of the effects of commercial NAs and OSPW-OF on mammalian immune responses raised the possibility that, in addition to NAs, other organic compounds in OSPW contribute to its toxicity (GarciaGarcia et al., 2011a, 2011b). The complexity of OSPW dissolved organics (including NAs), constitutes a major analytical challenge (Grewer et al., 2010). Recently, the Royal Society of Canada issued a report reviewing the environmental and health impacts of the oil sands industry, indicating the need for improved methods for the analysis of OSPW composition and toxic properties (Gosselin et al., 2010). In addition to NAs, other groups of organic contaminants in OSPW are BTEX (benzene, toluene, ethylbenzene, and xylene), phenols, and polycyclic aromatic hydrocarbons (Gosselin et al., 2010). Ultra high resolution mass spectrometry revealed many other sulfur or oxygen containing acids (Headley et al., 2011), but the structure of most OSPW constituents has only begun to be investigated (Rowland et al., 2011), and it is unclear how these additional chemicals contribute to OSPW toxicity in any organism. When the identification of all the compounds responsible for the toxicity of a given water sample is a technical and analytical obstacle, more integrative procedures combining chemical analysis and a battery of suitable bioassays is recommended (Petrovic et al., 2004). The necessity for a set of in vivo and in vitro bioassays is important, because in vitro assays do not always reliably predict the in vivo effects of the contaminants (Baker, 2001; Schlenk, 2008), due to tissue-specific differences in mechanisms of action, biotransformation, or tissue-specific bioaccumulation. OSPW ozonation seems to be an efficient method for NAs degradation, with reported complete (Scott et al., 2008) or partial (He et al., 2010; Martin et al., 2010) reduction of toxicity according to the Microtox bacterial toxicity assay (Martin et al., 2010; Scott et al., 2008) or the cell line-based H295R steroidogenesis assays (He et al., 2010). Since ozone can decay to hydroxyl radical, a strong non-selective oxidant, it is probable that ozonation degrades most organic chemicals in
OSPW, in addition to NAs. Despite the proven advantages of using ozone treatment for OSPW remediation (He et al., 2010; Martin et al., 2010; Scott et al., 2008), ozonation of wastewaters containing complex mixtures of chemicals can result in the formation of by-products which, in some cases, can be more hazardous than the substances they were derived from (Paraskeva and Graham 2002; Petala et al., 2008). In particular, NAs are known to form oxidized NAs (i.e. NA þ O) in ozonation systems (Martin et al., 2010). It is thus imperative that the toxicological properties of ozonated wastewaters are thoroughly evaluated, especially when they contain large numbers of chemicals, as is the case of OSPW. This is the first report evaluating the efficiency of ozonation in reducing OSPW toxicity in mammals, using both in vitro and in vivo bioassays.
2.
Materials and methods
2.1.
Oil sands process water (OSPW)
The Oil sands process water sample used in this study was collected from Syncrude West In-Pit [6325693N, 461372E (UTM, zone 12)] at various times during this study. A detailed composition of this water is shown in Supplementary Table 2.
2.2.
Ozonation of OSPW
OSPW was ozonated with an ozone generator (WEDECO, GSO40, Herford, Germany) utilizing extra dry high purity oxygen. The feed gas was delivered into the liquid phase through a gas diffuser. The ozone concentration in feed gas and effluent gas (off-gas) was monitored using two potassium iodidecalibrated ozone monitors (WEDECO, model HC-500, Herford, Germany). Residual ozone in the reactor was measured using the indigo method. The gas flow rate was measured by two calibrated flow meters (4e20 L/min and 0.5 to 2 L/min). The utilized ozone dose for this system was calculated as described previously (Martin et al., 2010). After ozonation, the OSPW was purged with pure nitrogen for 10 min to strip the ozone residual and oxygen from the reactor.
2.3. Extraction of the organic fractions of OSPW and ozonated OSPW The organic fractions of non-treated OSPW, or ozonated OSPW (OSPW þ O3), were extracted from 35 L clarified and unfiltered water batches, using a liquideliquid organic extraction protocol exactly as described (Garcia-Garcia et al., 2011a, 2011b). The total mass of the organic fraction of nontreated OSPW was 3.07 g, and the total mass of the organic fraction of ozonated OSPW was 1.92 g. Once the amount of NAs in the organic fractions was determined by UPLC/HRMS (0.89 g in OSPW-OF and 0.22 g in OSPW þ O3eOF), they were dissolved in 17.8 mL distilled water (pH 10) containing final NAs concentrations of 50 mg/mL for OSPW-OF NAs, and 12.5 mg/mL for OSPW þ O3eOF NAs. The volume of distilled water used to dissolve OSPW-OF was the same for OSPW þ O3eOF, to reflect the total loss of extractable organic contaminants (including NAs) resulting from ozonation. The
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 4 9 e5 8 5 7
5851
same liquideliquid organic extraction protocol was performed on distilled water, which was used as a control solution. The material obtained after performing the organic extraction on distilled water was also dissolved in the 17.8 mL of distilled water (pH 10). These stock solutions were used for oral administration to mice, or serially diluted in phosphatebuffered saline (PBS) and filter-sterilized for in vitro exposure of bone marrow-derived macrophages or (BMDM) and natural killer (NK) cells (see below, sections 2.5 to 2.11).
as phagocytosis scores, which integrate the percentage of positive cellsles internalized (cells that engulfed zymosan particles) and mean fluorescence intensity values (proportional to the amount of particles internalized), as described previously (Garcia-Garcia et al., 2011b). Phagocytosis scores varied considerably between individual BMDM, consequently the phagocytosis scores of OSPW-OF- or OSPW þ O3eOF-exposed BMDM were normalized against the scores of non-exposed BMDM from the same individual culture.
2.4.
2.9.
NAs analysis
Analysis of NAs content in OSPW-OF and OSPW þ O3eOF was performed by UPLC/HRMS exactly as described previously (Garcia-Garcia et al., 2011a, 2011b).
2.5. Exposure of cells in vitro to OSPWeOF or OSPW þ O3eOF BMDM were generated from 9 to 10 week old female C57BL/6 mice (Charles River Laboratories) using standard techniques (Hume and Gordon, 1983). Lymphokine-activated NK cells were generated from the spleens of C57BL/6 mice as described (Kirwan and Burshtyn, 2005). BMDM or NK cells were exposed in vitro to OSPW-OF or OSPW þ O3eOF, at the indicated NAs concentrations, for 18 h as described previously (Garcia-Garcia et al., 2011b).
2.6.
Nitric oxide assay
Nitric oxide production, in response to lipopolysaccharide (LPS) and gamma interferon (IFNg) stimulation, was measured using the Griess reaction, as described previously (GarciaGarcia et al., 2011b). All experiments were performed in triplicate using BMDM cultures established from 6 mice, processed individually. Data are expressed as nitrite concentration in the culture medium, which was determined using the A540 values and nitrite standard curve.
2.7.
Respiratory burst assay
The respiratory burst, stimulated by phorbol myristate acetate (PMA), was measured using the Nitroblue Tetrazolium (NBT) reduction assay, as described previously (Garcia-Garcia et al., 2011b). All experiments were performed in triplicate using BMDM cultures established from 6 mice, processed individually. Data are expressed as optical density at A630 (NBT reduction), after subtracting the background A630 readings.
2.8.
Phagocytosis assay
Zymosan (yeast cell wall) (SigmaeAldrich, catalog No. Z4250) was FITC-labeled as described (Garcia-Garcia et al., 2011b) and used as phagocytic targets. Phagocytosis assays were performed using zymosan at a ratio of 20:1 particles per cell exactly as described (Garcia-Garcia et al., 2011a). Particles adhered to the cell surface, but not internalized, accounted for less than 1% of cells that have phagocytosed zymosan (data not shown). All samples were run in triplicate using BMDM cultures established from 8 mice, processed individually. Data are expressed
Gene expression analysis of BMDMs
Constitutive (i.e. not stimulated by LPS and IFNg) or LPS and IFNg-stimulated gene expression of BMDMs was analyzed by real-time PCR, as described previously (Garcia-Garcia et al., 2011b) (see below, section 2.11). All experiments were performed in triplicate using BMDM cultures established from 6 mice, processed individually. Data are expressed as relative quantitation (RQ) values.
2.10.
Natural killer (NK) cell-mediated cytolysis assay
Target cells (721.221 cell line) were labeled with 51Cr sodium chromate, surface-labeled with the antibody L243, and washed. Target cells (2500) were mixed with 25,000 NK cells, and incubated at 37 C for 4 h. Target cell killing is the result of the recognition of the 721.221-bound L243 antibody by NK cells. Cell killing was estimated by 51Cr release into the culture supernatants, and analyzed using a 1450 Microbeta Trilux (Wallace) scintillation counter. All experiments were performed in triplicate using BMDM cultures established from 8 mice, processed individually. Data are expressed as percent lysis, which is calculated as follows: (mean sample release mean spontaneous release)/(mean total release - mean spontaneous release) 100.
2.11. Exposure of mice to OSPW and OSPW þ O3 organic fractions Six-to-eight week old female C57BL/6 mice were purchased from Charles River Laboratories. At 10 weeks of age, mice were gavaged with OSPW-OF at 100 mg NAs per kg of body weight, with OSPWeO3 þ OF at 25 mg NAs per kg body weight, or the control solution. Gavaging was selected over dissolving the contaminants in the drinking water, so that each mouse would receive specific NAs doses based on their weight. Each experimental group consisted of 4 mice. One week after exposure animals were euthanized and the liver, mesenteric lymph node, and spleen were surgically removed for gene expression analysis (section 2.11).
2.12.
Gene expression analysis
Gene expression analysis was performed exactly as described (Garcia-Garcia et al., 2011a, 2011b). Total RNA was extracted from BMDM cultures or mice organs, using the Ribopure kit (Ambion, catalogue no. AM1924). RNA was reverse-transcribed into cDNA using the High Capacity RNA-to-cDNA Kit (Applied Biosystems, catalogue no. 4387406) according to the manufacturer’s directions. Real-time PCR analysis of gene expression
5852
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 4 9 e5 8 5 7
was performed using commercial TaqMan primer/probes (Applied Biosystems), with reported 100% amplification efficiencies. The genes that were examined were TNF-a (catalogue No. Mm99999068_m1), IFNg (catalogue No. Mm00801778_m1), IL-1b (catalogue No. Mm00434228_m1), IL-6 (catalogue No. Mm00446191_m1), IL-2 (catalogue No. Mm00434256_m1), IL-12b (catalogue No. Mm01288992_m1), CSF-1 (catalogue No. Mm00432688_m1), CCL2 (catalogue No. Mm00441242_m1), CCL3 (catalogue No. Mm00441258_m1), CCL4 (catalogue No. Mm00443111_m1), CCL5 (catalogue No. Mm01302427_m1), and HPRT-1 (catalogue No. Mm03024075_m1). Real-time PCR reactions were performed in triplicate for each gene, and HPRT-1 was included as an endogenous control in each run. Gene expression analysis was performed using the DDCT method. Data are expressed as relative quantitation (RQ) values.
2.13.
Statistical analysis
Statistical analysis was performed using the Statistical Package for the Social Sciences software v. 15.0. Data were analyzed using the Levene equality of variance test, followed by independent samples Student’s T-test; or one way ANOVA, followed by Tukey post hoc tests for multiple comparisons. Differences between treatment groups were considered statistically significant when P < 0.05.
3.
Results
3.1. Ozone treatment abolished the ability of OSPW to reduce BMDM nitric oxide (NO) production We recently reported the development of in vivo and in vitro bioassays to evaluate the effects of the organic fraction of oil sands process water (OSPW-OF) on mammalian immune mechanisms (Garcia-Garcia et al., 2011a, 2011b). Using these bioassays we evaluated the efficiency of ozone treatment in ameliorating the immunotoxic effects of OSPW-OF. Ozonation of OSPW caused the degradation of parent (i.e. non-oxidized) NAs, which are believed to be the major toxicants in OSPW (Madill et al., 2001). NAs were degraded evenly, regardless of their size or complexity (i.e. m and Z values, respectively), to approximately 25% of the initial amount (Suppl. Fig. 1). LPS and IFNg-stimulated production of nitric oxide (NO) by BMDM is one of the microbicidal functions affected by OSPWOF (Garcia-Garcia et al., 2011b). To evaluate the effect of ozonation on the immunotoxic properties of OSPW-OF, BMDM were exposed to OSPW-OF (containing 50, 25, 12.5 and 6.25 mg/ mL NAs) or OSPW þ O3eOF (containing 12.5, 6.25, 3.12, and 1.56 mg/mL NAs) for 18 h and then stimulated with LPS and IFNg. Stimulation of BMDM with LPS and IFNg resulted in significantly higher NO production (Suppl. Fig. 2A). This stimulated production of NO was reduced upon BMDM exposure to the OSPW-OF dilutions containing the highest NAs concentrations (25 and 50 mg/mL) (Fig. 1A). In contrast, BMDM showed normal NO production when exposed to any of the OSPW þ O3eOF dilutions (Fig. 1A). BMDM stimulation results in the transcriptional activation of the enzyme called inducible nitric oxide synthase (iNOS) (Suppl. Fig. 2B), which is required for elevated NO production.
Exposing BMDM to OSPW-OF containing 50 mg/mL NAs resulted in approximately 80% reduction in LPS and IFNg-stimulated expression of the iNOS gene (Fig. 1B). In contrast, the OSPW þ O3eOF dilution containing the highest NAs concentration (12.5 mg/mL) caused only a small reduction of iNOS gene expression (Fig. 1B). Similarly, OSPW-OF reduced the constitutive (i.e. not stimulated by LPS and IFNg) iNOS gene expression, while OSPW þ O3eOF did not (Suppl. Fig. 3).
3.2. Ozone treatment abolished the ability of OSPW to reduce BMDM production of reactive oxygen intermediates (ROI) PMA-stimulated production of ROI is another microbicidal function affected by OSPW-OF (Garcia-Garcia et al., 2011b). To further assess the effect of ozonation on the immunotoxic properties of OSPW, BMDM were exposed to OSPW-OF (containing 50, 25, 12.5 and 6.25 mg/mL NAs) or OSPW þ O3eOF (containing 12.5, 6.25, 3.12, and 1.56 mg/mL NAs) for 18 h and then stimulated with PMA. Stimulation of BMDM with PMA significantly increased ROI production (Suppl. Fig. 4). This PMA-stimulated production of ROI was reduced by the OSPWOF dilution containing the highest NAs concentration (50 mg/ mL) (Fig. 1C). In contrast, none of the OSPW þ O3eOF dilutions affected this microbicidal response (Fig. 1C). ROI production results from the activation of a multiprotein enzyme complex known as the reduced nicotineamide adenine dinucleotide phosphate (NADPH) oxidase. The NADPH oxidase is highly expressed in mature phagocytic cells and its activation is tightly regulated. Coordinated expressions of the genes that encode p91Phox, p47Phox, and p67Phox are induced upon differentiation of myeloid cells beyond the pro-myelocyte stage (Skalnik, 2002). As shown in Fig. 1D, BMDM exposure to OSPW-OF at 50 mg/mL NAs caused an approximately 40% reduction in the gene expression of two components of the NADPH oxidase, p47Phox and p67Phox (Fig. 1D). In contrast, the OSPW þ O3eOF dilution containing the highest NAs concentration (12.5 mg/mL) had no effect on the expression of NADPH oxidase subunits (Fig. 1D).
3.3. Ozone treatment abolished the ability of OSPW to reduce BMDM phagocytic function Phagocytosis is other BMDM function previously shown to be affected by in vitro exposure to OSPW-OF (Garcia-Garcia et al., 2011b). We now used this bioassay to further evaluate the efficiency of ozone treatment in reducing the immunotoxic properties of OSPW. As shown in Fig. 2, exposure of BMDM to OSPW-OF at 50 mg/mL NAs decreased the phagocytosis of zymosan by approximately 20% (Fig. 2), while the exposure of BMDM to the OSPW þ O3eOF dilution containing the highest NAs concentration (12.5 mg/mL) had no effect on phagocytosis (Fig. 2).
3.4. Ozone treatment abolished the ability of OSPW to alter BMDM cytokine gene expression In vitro exposure of BMDM to OSPW-OF results in altered expression of various cytokine genes (Garcia-Garcia et al., 2011b). We now evaluated the efficiency of ozonation in
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 4 9 e5 8 5 7
5853
Fig. 1 e Ozonation abolished the ability of OSPW-OF to decrease the BMDM production of nitric oxide and reactive oxygen intermediates. A) Bone marrow-derived macrophage cultures (BMDM) were exposed in vitro for 18 h to serial dilutions of the organic fraction (OF) of OSPW or ozonated OSPW (OSPW D O3), containing the indicated naphthenic acids (NAs) concentrations. Nitric oxide production was then stimulated by lypopolysaccharide and gamma interferon (LPSDIFNg) and analyzed 18 h later using the Griess reaction. Results are mean ± SEM nitrite concentration in the supernatant of 6 independent BMDM cultures. B) BMDM were exposed in vitro for 18 h to dilutions of OSPW-OF or OSPW D O3eOF, containing 50 or 12.5 mg/mL NAs respectively, and then stimulated with LPSDIFNg for 18 h, iNOS expression was determined by realtime PCR using the DDCT method. The relative quantitation (RQ) values for OSPWeOFe or OSPW D O3-exposed macrophages were normalized against the RQ values of non-exposed cells. Results are mean ± SEM of macrophage cultures established from 6 mice, processed individually. C) BMDM were exposed in vitro for 18 h to OSPW-OF or OSPW D O3 as in (A). ROI production was then stimulated by PMA and analyzed 25 min later using the NBT reduction assay. Results are mean ± SEM optical density (A630) of 6 independent BMDM cultures. D) BMDM were exposed to OSPW-OF or OSPW D O3 as in (B). The expression of the NADPH subunits p47Phox, p67Phox, and p91Phox was analyzed as in (B). The relative quantitation (RQ) values for OSPW-OF or OSPW D O3eOFeexposed macrophages were normalized against the RQ values of non-exposed cells. Results are mean ± SEM of macrophage cultures established from 6 mice, processed individually. Asterisks (*) indicate statistically significant differences from controls at P < 0.05.
preventing these OSPWeOFeinduced alterations on in vitro gene expression. Supplementary Table 1 describes the functions of the immune genes analyzed in this study. Fig. 3A shows that exposure of BMDM in vitro to OSPW-OF at 50 mg/mL NAs reduced constitutive (i.e. non-stimulated by immune challenge) expression of IL-1, IL-6, and IL-12, but enhanced the expression of TNF-a. In contrast, the exposure of BMDM to the OSPW þ O3eOF dilution containing the highest NAs concentration (12.5 mg/mL) had no effect on constitutive cytokine gene expression (Fig. 3A). Upon immune stimulation, macrophages produce several cytokines that are required for the activation of proinflammatory mechanisms, and also for the initiation of adaptive immunity mechanisms (Suppl. Table 1). Transcriptional activation of cytokine genes is a major regulatory step in the production of cytokines by macrophages, and the combination of LPS and IFNg is a potent stimulus for the up-regulation of
various cytokine genes (Adcock, 1997; Silvennoinen et al., 1997). BMDM stimulation with LPS and IFNg causes up-regulation of IL-1, IL6, IL-12 and TNF-a gene expression (Suppl. Fig. 5). The exposure of BMDM to OSPW-OF decreased LPS and IFNg-stimulated expression of IL-1, but enhanced stimulated expression of IL-12 and TNF-a (Fig. 3B). In contrast, the treatment of BMDM with OSPW þ O3eOF did not affect LPS and IFNg-stimulated cytokine gene expression (Fig. 3B).
3.5. Ozone treatment abolished the ability of OSPW to alter cytokine gene expression in vivo We previously showed that exposing mice orally to OSPW-OF resulted in down-regulation of several immune genes in the liver, the mesenteric lymph nodes, and the spleen in a timedependent manner (Garcia-Garcia et al., 2011a). The liver was selected for gene expression analysis because it is the
5854
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 4 9 e5 8 5 7
Fig. 2 e Ozonation abolished the ability of OSPW-OF to decrease BMDM phagocytosis. Bone marrow-derived macrophages were exposed in vitro for 18 h to dilutions of the organic fractions of OSPW or ozonated OSPW (OSPW D O3), containing 50 or 12.5 mg/mL NAs respectively. Fluorescent zymosan particles were then added at a 20:1 target:cell ratio, cells were incubated for 45 min at 37 C, and phagocytosis analyzed by flow cytometry. Data are mean ± SEM of macrophage cultures established from 8 mice, processed individually. Asterisks (*) denote statistically significant differences from controls at P < 0.05.
major detoxifying and immune organ in the body. The mesenteric lymph node was selected for analysis because its location along the gastrointestinal tract would make it an early target for possible immune alterations induced by oral administration of OSPW. The spleen was selected for immune gene expression analysis because it is the primary immune organ in mammals. Supplementary Table 1 describes the functions of the immune genes analyzed in this study. As shown in Fig. 4A, oral exposure of mice to OSPW-OF containing 100 mg/kg NAs caused a down-regulation of TNF-a, IFNg, IL-1, CFS-1, CCL3 and CCL4 in the liver. The expression of none of these genes was altered in the liver of mice exposed to OSPW þ O3eOF (Fig. 4A). Similarly, Fig. 4B shows that the expression of IL-1 in the mesenteric lymph node was downregulated in mice exposed to OSPW-OF, but not in animals exposed to OSPW þ O3eOF. Interestingly, mice exposed to OSPW þ O3eOF, but not those exposed to OSPW-OF, exhibited elevated expression of TNF-a, IFNg, CCL2, CCL3 and CCL4. Animals exposed to either OSPW-OF or OSPW þ O3eOF exhibited no significant alterations in cytokine gene expression in the spleen (Fig. 4C).
4.
Discussion
The aim of this study was to evaluate the efficiency of ozone treatment in reducing the immunotoxicity of OSPW. Following the Oil Sands Water Release Technical Working Group recommendation (OSWRTG 1996) for OSPW toxicological studies, we used an experimental design analogous to the
Fig. 3 e Ozonation abolished the ability of OSPW-OF to alter BMDM cytokine expression. Bone marrow-derived macrophages were exposed in vitro for 18 h to dilutions of the organic fractions (OF) of OSPW or ozonated OSPW (OSPW D O3), containing 50 or 12.5 mg/mL NAs respectively. A) Non-stimulated gene expression analysis was performed by real-time PCR using the DDCT method. B) Cells were stimulated with lipopolysaccharide and gamma interferon (LPSDIFNg) for 18 h, after OSPW-OF or OSPW D O3eOF exposure. Gene expression analysis was then performed by real-time PCR as in (A). Results are expressed as relative quantification (RQ) values. RQ values for OSPW-OF or OSPW D O3eOFeexposed macrophages were normalized against the RQ values of non-exposed macrophages (control). Data are mean ± SEM of macrophage cultures established from 6 mice, processed individually. Asterisks (*) denote statistically significant differences between OSPW-OF or OSPW D O3 exposures and controls, crosses (y) denote statistically significant differences between OSPWeOF and OSPW D O3 exposures, at P < 0.05.
whole effluent experimental approach and extracted the entire organic fraction of OSPW and ozonated OSPW. NAs are believed to be the major toxicants in OSPW, and ozonation reduced the amount of parent (non-oxidized) NAs in the OSPW organic fraction by approximately 75%. In addition, even though the organic fractions of OSPW and ozonated OSPW were extracted from the same volume of tailings water, the organic mass extracted from OSPW þ O3eOF was reduced by almost 40%. The observed reduction in immunotoxicity after OSPW ozonation thus correlates with NAs degradation, but it is possible that the loss of other organic contaminants (e.g. benzene, toluene, ethylbenzene, xylene, phenols and
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 4 9 e5 8 5 7
5855
Fig. 4 e Ozonation abolished the ability of OSPW-OF to alter cytokine gene expression in vivo. Mice were exposed to a single dose of the organic fractions of OSPW or ozonated OSPW (OSPW D O3). One week later gene expression analysis in the liver (A), the mesenteric lymph node (MLN) (B) or the spleen (C) was performed by real-time PCR, using the DDCt method. Results are expressed as relative quantification (RQ) values. RQ values for the OSPW-OF or OSPW D O3eOF experimental groups were normalized against the RQ values of non-exposed (control) mice for each gene. Data are mean ± SEM of 4 mice. Asterisks (*) denote statistically significant differences between OSPW-OF or OSPW D O3 exposures and controls, crosses (y) denote statistically significant differences between OSPW-OF and OSPW D O3 exposures, at P < 0.05.
polycyclic aromatic hydrocarbons; see Suppl. Table 2) is also involved in reduced toxicity. We used a combination of in vitro and in vivo bioassays to carefully evaluate the efficiency of ozonation in reducing OSPW immunotoxicity. Such a combination is necessary because in vitro assays sometimes fail to predict the in vivo effects of complex mixtures of contaminants (Baker, 2001; Petrovic et al., 2004; Schlenk, 2008). Ozonation prevented all
the toxic effects of OSPW on BMDM functions in vitro, as well as the down-regulation of various immune genes in the liver and the mesenteric lymph node in vivo. Interestingly, mice exposure to OSPW þ O3eOF, but not to OSPW-OF, caused the over expression of TNF-a, IFNg, CCL2, CCL3, and CCL4 in the mesenteric lymph node. Elevated expression of these cytokines and chemokines is suggestive of an on-going inflammatory process in the mesenteric lymph nodes, because
5856
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 4 9 e5 8 5 7
TNF-a and IFNg are key mediators of inflammation; and CCL2, CCL3, and CCL4 are chemotactic agents for cells involved in innate immunity. This observation raises the possibility that the oxidation products of NAs, or other organic contaminants present in OSPW, have toxic effects in vivo. As illustrated in Supplementary Table 2, there are many other organic contaminants in our OSPW sample. From these additional contaminants, at least three polycyclic aromatic hydrocarbons (pyrene, fluoranthene, and benzo[a]Pyrene) have been reported to produce by-products of increased toxicity in vitro, after the complete degradation of the parent compounds by ozonation (Upham et al., 1994). We used several in vitro immunological assays based on macrophage functions to demonstrate that ozone treatment could prevent the immunotoxic effects of OSPW. In addition to BMDM, we have devised an in vitro toxicity test using natural killer (NK) cells (cells providing protection against tumors and virus-infected cells), for a more comprehensive evaluation of OSPW remediation through ozonation. As shown in Suppl. Fig. 6, exposing in vitro-generated NK cells to OSPW-OF (50 mg/mL NAs) resulted in a 30% reduction in specific NK-cell target killing, whereas the exposure of NKcells to the OSPW þ O3eOF dilution with the highest NAs content (12.5 mg/mL) had no effect on their function. These results indicate that ozonation abrogates OSPW immunotoxic effects, not only towards macrophages but also towards NK cells. We believe that the toxicological bioassays depicted in this work can be efficiently used to evaluate various OSPW remediation techniques, as well as to evaluate the toxicological properties of OSPW samples with different NAs compositions. This study is the first report indicating that ozonation can help to reduce the immunotoxic properties of OSPW in vivo and in vitro.
5.
Conclussions
Ozonation of OSPW abolished its ability to impair various in vitro functions of mammalian macrophages. Ozonation of OSPW abolished its ability to down-regulate the expression of multiple cytokines and chemokines in vivo.
Acknowledgements This work was supported by a grant from Alberta Water Research Institute (AWRI) to MB, MGED, and JWM, and Natural Sciences and Engineering Research Council of Canada (NSERC) Discovery Grant to MB. We thank Dr. Deborah N. Burshtyn for her support in the setup of NK killing assays.
Appendix. Supplementary material Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.watres.2011.08.032.
references
Adcock, I.M., 1997. Transcription factors as activators of gene transcription: aP-1 and NF-kappa B. Monaldi Archives in Chest Disease 52 (2), 178e186. Baker, V.A., 2001. Endocrine disruptersetesting strategies to assess human hazard. Toxicology in Vitro 15 (4e5), 413e419. Clemente, J.S., Fedorak, P.M., 2005. A review of the occurrence, analyses, toxicity, and biodegradation of naphthenic acids. Chemosphere 60 (5), 585e600. Garcia-Garcia, E., Pun, J., Hodgkinson, J., Perez-Estrada, L.A., Din, M.G., Smith, D.W., Martin, J.W., Belosevic, M., 2011a. Commercial naphthenic acids and the organic fraction of oil sands process water induce different effects on proinflammatory gene expression and macrophage phagocytosis in mice. Journal of Applied Toxicology DOI 10.1002/jat.1687. Garcia-Garcia, E., Pun, J., Perez-Estrada, L.A., Din, M.G., Smith, D. W., Martin, J.W., Belosevic, M., 2011b. Commercial naphthenic acids and the organic fraction of oil sands process water down-regulate pro-inflammatory gene expression and macrophage antimicrobial responses. Toxicology Letters 203 (1), 62e73. Gosselin, P., Hrudey, S.E., Naeth, M.A., Plourde, A., Therrien, R., Van Der Kraak, G., Xu, Z., 2010. Environmental and Health Impacts of Canada’s oil Sands Industry. Available at. Royal Society of Canada, Ottawa. http://www.rsc.ca/expertpanels_ reports.php. Grewer, D.M., Young, R.F., Whittal, R.M., Fedorak, P.M., 2010. Naphthenic acids and other acid-extractables in water samples from Alberta: what is being measured? Science of the Total Environment 408 (23), 5997e6010. He, Y., Wiseman, S.B., Zhang, X., Hecker, M., Jones, P.D., El-Din, M. G., Martin, J.W., Giesy, J.P., 2010. Ozonation attenuates the steroidogenic disruptive effects of sediment free oil sands process water in the H295R cell line. Chemosphere 80 (5), 578e584. Headley, J.V., Barrow, M.P., Peru, K.M., Fahlman, B., Frank, R.A., Bickerton, G., McMaster, M.E., Parrott, J., Hewitt, L.M., 2011. Preliminary fingerprinting of Athabasca oil sands polar organics in environmental samples using electrospray ionization Fourier transform ion cyclotron resonance mass spectrometry. Rapid Communications in Mass Spectrometry 25 (13), 1899e1909. Headley, J.V., McMartin, D.W., 2004. A review of the occurrence and fate of naphthenic acids in aquatic environments. Journal of Environmental Science and Health, Part A. Toxic / Hazardous Substances and Environmental Engineering 39 (8), 1989e2010. Headley, J.V., Peru, K.M., Armstrong, S.A., Han, X., Martin, J.W., Mapolelo, M.M., Smith, D.F., Rogers, R.P., Marshall, A.G., 2009. Aquatic plant derived changes in oil sands naphthenic acid signatures determined by low, high and ultra-high resolution mass spectrometry. Rapid Communications in Mass Spectrometry 23, 515e522. Hume, D.A., Gordon, S., 1983. Optimal conditions for proliferation of bone marrow-derived mouse macrophages in culture: the roles of CSF-1, serum, Ca2þ, and adherence. Journal of Cell Physiology 117 (2), 189e194. Kirwan, S.E., Burshtyn, D.N., 2005. Killer cell Ig-Like receptordependent Signaling by Ig-Like Transcript 2 (ILT2/CD85j/ LILRB1/LIR-1)1. Journal of Immunology 175, 5006e5015. Madill, R.E., Orzechowski, M.T., Chen, G., Brownlee, B.G., Bunce, N.J., 2001. Preliminary risk assessment of the wet landscape option for reclamation of oil sands mine tailings: bioassays with mature fine tailings pore water. Environmental Toxicology 16 (3), 197e208.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 4 9 e5 8 5 7
Martin, J.W., Barri, T., Han, X., Fedorak, P.M., El-Din, M.Gamal, Perez, L., Scott, A.C., Tiange, J.J., 2010. Ozonation of oil sands process-affected water Accelerates Microbial bioremediation. Environmental Science and Technology 44 (21), 8350e8356. Oil Sands Water Release Technical Working Group (O.S.W.R.T.W. G), 1996. Approaches to oil sands water release, p. 34. Paraskeva, P., Graham, N.J., 2002. Ozonation of municipal wastewater effluents. Water Environment Research 74 (6), 569e581. Petala, M., Samaras, P., Zouboulis, A., Kungolos, A., Sakellaropoulos, G.P., 2008. Influence of ozonation on the in vitro mutagenic and toxic potential of secondary effluents. Water Research 42 (20), 4929e4940. Petrovic, M., Eljarrat, E., Lopez De Alda, M.J., Barcelo, D., 2004. Endocrine disrupting compounds and other emerging contaminants in the environment: a survey on new monitoring strategies and occurrence data. Analytical and Bioanalytical Chemistry 378 (3), 549e562. Quagraine, E.K., Peterson, H.G., Headley, J.V., 2005. In situ bioremediation of naphthenic acids contaminated tailing pond waters in the athabasca oil sands regionedemonstrated field studies and plausible options: a review. Journal of
5857
Environmental Science and Health, Part A. Toxic / Hazardous Substances and Environmental Engineering 40 (3), 685e722. Rowland, S.J., Scarlett, A.G., Jones, D., West, C.E., Frank, R.A., 2011. Diamonds in the rough: identification of individual naphthenic acids in oil sands process water. Environmental Science and Technology 45 (7), 3154e3159. Scott, A.C., Zubot, W., MacKinnon, M.D., Smith, D.W., Fedorak, P. M., 2008. Ozonation of oil sands process water removes naphthenic acids and toxicity. Chemosphere 71 (1), 156e160. Schlenk, D., 2008. Are steroids really the cause for fish feminization? A mini-review of in vitro and in vivo guided TIEs. Marine Pollution Bulletin 57 (6e12), 250e254. Silvennoinen, O., Saharinen, P., Paukku, K., Takaluoma, K., Kovanen, P., 1997. Cytokine receptor signal transduction through Jak tyrosine kinases and Stat transcription factors. Apmis 105 (7), 497e509. Skalnik, D.G., 2002. Transcriptional mechanisms regulating myeloid-specific genes. Gene 284 (1e2), 1e21. Upham, B.L., Masten, S.J., Lockwood, B.R., Trosko, J.E., 1994. Nongenotoxic effects of polycyclic aromatic hydrocarbons and their oxygenation by-products on the intercellular communication of rat liver epithelial cells. Fundamental and Applied Toxicology 23 (3), 470e475.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 5 8 e5 8 7 2
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Effects of soluble and particulate substrate on the carbon and energy footprint of wastewater treatment processes Riccardo Gori a, Lu-Man Jiang b, Reza Sobhani b, Diego Rosso b,* a b
Department of Civil and Environmental Engineering - DICEA, University of Florence, Via S. Marta 3, 50139 Florence, Italy Department of Civil and Environmental Engineering, University of California, Irvine, CA 92697-2175, USA
article info
abstract
Article history:
Most wastewater treatment plants monitor routinely carbonaceous and nitrogenous load
Received 29 March 2011
parameters in influent and effluent streams, and often in the intermediate steps. COD frac-
Received in revised form
tionation discriminates the selective removal of VSS components in different operations,
20 August 2011
allowing accurate quantification of the energy requirements and mass flows for secondary
Accepted 22 August 2011
treatment, sludge digestion, and sedimentation. We analysed the different effects of COD
Available online 27 August 2011
fractions on carbon and energy footprint in a wastewater treatment plant with activated sludge in nutrient removal mode and anaerobic digestion of the sludge with biogas energy
Keywords:
recovery. After presenting a simple rational procedure for COD and solids fractions quantifi-
COD fractionation
cation, we use our carbon and energy footprint models to quantify the effects of varying
Soluble COD
fractions on carbon equivalent flows, process energy demand and recovery. A full-scale real
Hydrolysis
process was modelled with this procedure and the results are reported in terms of energy and
Carbon footprint
carbon footprint. For a given process, the increase of the ratio sCOD/COD increases the energy
Energy footprint
demand on the aeration reactors, the associated CO2 direct emission from respiration, and the
Wastewater treatment
indirect emission for power generation. Even though it appears as if enhanced primary sedimentation is a carbon and energy footprint mitigation practice, care must be used since the nutrient removal process downstream may suffer from an excessive bCOD removal and an increased mean cell retention time for nutrient removal may be required. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
The need for a rigorous scientific approach to research and model wastewater treatment processes has been paramount to water research for nearly a century: “American sewerage practice is noteworthy among the branches of engineering for the preponderating influence of experience rather than experiment upon the development of many of its features, apart from those concerned with the treatment of sewage” (Metcalf and Eddy, 1914). The qualification of the oxygen demand and suspended solids components is a key step in transcending this practice-based engineering approach.
Modelling biological wastewater treatment processes has been the major driving force for the development of protocols aimed at splitting macroscopic parameters (such as total and volatile suspended solids, TSS and VSS, or biochemical and chemical oxygen demand, BOD5 and COD) into fractions with different behaviour in the treatment processes. In particular, COD fractionation was introduced as a tool for the evaluation and modelling of biological treatment processes performance (Henze, 1992). Modelling of activated sludge processes (ASP) is usually carried out with structured models, such as the ASM family (Henze et al., 2000). The ASM1 represents the basic model used
* Corresponding author. Tel.: þ1 949 824 8661; fax: þ1 949 824 3672. E-mail address:
[email protected] (D. Rosso). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.08.036
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 5 8 e5 8 7 2
Notation biodegradable COD in flow line i (massbCOD/d) biodegradable volatile suspended solids (mg/l) biodegradable VSS in primary sludge (massVSS/d) biodegradable VSS in secondary sludge (massVSS/d) energy demand (kWh/d) eD energy demand for anaerobic digestion (kWh/d) eD,AD energy demand for activated sludge aeration eD,ASP (kWh/d) energy demand for primary sedimentation eD,PS (kWh/d) energy demand for secondary sedimentation eD,SS (kWh/d) energy demand for other equipment (kWh/d) eD,O energy recovery (kWh/d) eR fraction of biomass contributing to biomass debris fd () calorific value of the biogas (kJ/kgbiogas) hBG iTSS inert total suspended solids (massTSS/l) k maximum substrate utilisation rate (massCOD/massVSS d) decay rate of autotrophic biomass (1/d) kd,A decay rate of heterotrophic biomass (1/d) kd,H decay rate of biomass under anaerobic conditions kd,ANA (1/d) half-saturation constant (mg/l) Ks mass flow of biogas (massBG/d) mBG mBG,VSS mass flow of biogas calculated from VSS destruction (massBG/d) mBS,COD mass flow of biosolids as COD (massCOD/d) mBS,bCOD mass flow of biosolids as bCOD (massCOD/d) mBS,VSS mass flow of biosolids calculated from VSS destruction (massVSS/d) mass flow of methane (massCH4 =d) mCH4 mCH4 ;fugitive mass flow of methane fugitive emission (massCH4 =d) mCH4 ;dewatering mass flow of methane released during biosolids dewatering (massCH4 =d) mCO2 ;ASP mass flow of CO2 in the activated sludge off-gas (massCO2 =d) mCO2 ;AD mass flow of CO2 in the biogas (massCO2 =d) mCO2 eq total mass flow of CO2eq (massCO2 =d) mCO2 ;CH4 comb mass flow of CO2 due to methane combustion (massCO2 =d) mCO2 eq;PG mass flow of CO2 due to off-site power generation (massCO2 =d) mCO2 eq;offset mass flow of CO2,eq offset due to energy recovery (massCO2 =d) mCO2 eq;fugitive mass flow of CO2,eq due to fugitive emission (massCO2 =d) mDIG,in mass flow entering the digester (massbCOD/d) mPS,bCOD bCOD mass flow of waste primary sludge (massbCOD/d) mPS,COD COD mass flow of waste primary sludge (massCOD/d) mPS,VSS COD mass flow of waste primary sludge (massVSS/d) mPS,TSS COD mass flow of waste primary sludge (massTSS/d) bCODi bVSS bVSSPS bVSSSS
5859
mass flow of waste secondary sludge (massbCOD/d) mSS mbCOD-CH4 bCOD converted in methane in the anaerobic digester (massbCOD/d) mbCOD-ANA biomass bCOD converted in biomass under anaerobic conditions (massbCOD/d) mbCOD-SN bCOD in the supernatant line (massbCOD/d) mbCOD-sol bCOD solubilised in the anaerobic digester (massbCOD/d) MCRT mean cell retention time (d) concentration of oxidised nitrogen (mgN/l) NOx pbCODi particulate biodegradable COD in flow line i (mg/l) pnbCODi particulate non-biodegradable COD in flow line i (mg/l) Px,active bio daily active biomass production (massVSS/d) daily excess sludge production in terms of VSS Px,VSS (massVSS/d) daily total excess sludge production (massTSS/d) Px,TSS daily total excess sludge production in terms of Px,COD COD (massCOD/d) biomass production under anaerobic conditions Px,ANA (massVSS/d) Q influent flow rate (m3/d) flow rate of biogas leaving the anaerobic digester QBG (Nm3/d) sbCODi soluble biodegradable COD in flow line i (mg/l) snbCODi soluble non-biodegradable COD in flow line i (mg/l) SRTDIG digester solids retention time (d) VSS volatile suspended solids (mg/l) TSS total suspended solids (mg/l) autotrophic biomass yield (massVSS/massN,oxidised) YA heterotrophic biomass yield (massVSS/massCOD) YH biomass yield in anaerobic conditions YANA (massVSS/massCOD) DbCOD difference between solubilised bCOD in the digester and bCOD in the supernatant line (massCOD/d) Greek letters specific methane production (Nm3 =kgbCOD ) bCH4 bCOD removal in the activated sludge process (%) hASP bCOD solubilised in the digester (%) hDIG hDIG,VSS percent VSS destruction in the digester (%) hDIG,TSS percent TSS destruction in the digester (%) efficiency of the energy recovery unit () hER particulate removal in primary sedimentation (%) hPS fugitive emission of methane (%) hFE kBIOMASS carbon emission intensity of biomass (kgCO2 eq =kgCOD ) carbon emission intensity of COD (kgCO2 eq =kgCOD ) kCOD specific CO2 emission for off-site energy kPG generation (kgCO2 eq =kWh) methane density (kgCH4 =Nm3 ) rCH4 biogas density (kgBG/Nm3) rBG CO2 density (kgCO2 =Nm3 ) rCO2 Subscripts i primary influent (PI); primary effluent (PE); secondary effluent (SE)
5860
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 5 8 e5 8 7 2
to describe removal of organic carbon and nitrogen, consumption of nitrate and oxygen as electron acceptors, and sludge production (Henze et al., 1987). Nowadays, the ASM1 model is still often the state-of-the-art for modelling activated sludge systems (Roeleveld and van Loosdrecht, 2002). Regardless of the model used, COD and nitrogen have to be fractionated in order to differentiate compounds with different mechanisms and rates of biodegradation, which in turn determine their different fate in sedimentation and biological processes. The basic approach for fractionation relies on considerations over the biodegradability and solubility of wastewaters components. Several protocols, from simplified and practical to those of increased complexity, were developed for wastewater characterisation: two of the major were developed by STOWA (Hulsbeek et al., 2002) and Biomath (Vanrolleghem et al., 2003). Both of them rely on the combination of biological (respirometric) and physical-chemical measurements which are not widely performed in normal full-scale plant operations. On a routine basis, most treatment plants monitor COD, 3BOD5, TKN, NHþ 4 -N and PO4 -P in influent and effluent streams, and oftentimes in the intermediate process steps. The typical assumption for municipal wastewater quality is that it has minor or no industrial wastewater contribution. In reality, the composition of municipal wastewater can vary widely from one site to another and in the same location over time. Influent quality is affected by several site-specific conditions, such as the nature of discharged compounds, insewer microbial transformation, sewer length and its residence time, gas-transfers in sewers, water and ambient temperatures (Vollertsen et al., 1999). For this reason, the extension of models from one site to another may result in inaccuracies, and the preliminary collection of site-specific data for fractionation is always beneficial. The introduction of COD fractionation in a mass-balance based model discriminates the selective removal of VSS components, which is key for the accurate quantification of the energy requirements and mass flows for secondary treatment, sludge digestion, and sedimentation. For example, two limit cases could be a treatment plant operating chemically-enhanced primary treatment followed by no biological process, the other operating an activated sludge at low sludge age with inadequate or no primary treatment. A simple rational procedure for the calculation of COD fractions hence pCOD/VSS (such as that reported in the following section), would provide precious information to distinguish the two cases. Even though both plants could be hypothesised to have matching VSS mass flow rates to their digesters, the quality of each VSS stream in terms of pCOD/VSS would be substantially different. Therefore, the two plants would yield different VSS removal and biogas production in their respective digesters. Currently, most of the design procedures widely taught in textbooks rely on few empirical data (inter alia, Metcalf and Eddy, 2003). Due to the lack of published experimental data and to the range of possible COD fractions, there is a need for the quantification of the effects of pCOD/VSS variations. Presently, the reduction of carbon and energy footprint (CFP and eFP, respectively) is a worldwide concern associated with global warming mitigation and adaptation strategies. Treatment utilities can play a significantly beneficial role here
because of the large amounts of carbonaceous organic matter processed daily. This COD may be converted to biomass and CO2, and ultimately utilised for energy recovery and/or carbon sequestration (Rosso and Stenstrom, 2008). However, the removal of COD from the water stream is energy intensive and is associated with direct and indirect release of greenhouse gases (Monteith et al., 2005; de Haas et al., 2008; Park et al., 2010; Ahn et al., 2010; Foley et al., 2010), which may have a relevant weight on process CFP (Rosso et al., 2009). Regardless of the anthropogenic or biogenic nature of the COD in the influent wastewater, the treatment process can add the mitigation of greenhouse gas emissions to its already evident public and environmental health benefits. The goal of this paper is to show the effect of varying COD and solids fractions on process carbon and energy footprints. After presenting a simple rational procedure for COD and solids fractions quantification, we use our carbon and energy footprint models to quantify the effects of varying fractions on carbon equivalent flows, process energy demand and recovery. This model does not aim at replacing the detailed structured models already available (e.g. ASM, ADM), but rather at introducing a more rigorous approach to the massbalance based carbon and energy footprint analysis based on a simplified set of process input data.
2.
Materials and methods
2.1.
Plant data selection
We selected a municipal water reclamation plant (Q w 60000 m3 d-1) located in the United States in a warm area (Tww,avg ¼ 20 C) with process schematic matching that of Fig. 1, with the addition of headworks and disinfection. The main process parameters were collected and are reported in Table 1. Data sets were collected as daily measurements for 2010, and were processed with the COD fractionation procedure described below to quantify yearly averages of COD fractions for variable pCOD/VSS ratio. Also, the energy consumption per unit operation was collected from plant data logs, confirming that aeration energy was the dominant process energy component (64.0e74.2% of the total process energy), with the exclusion of effluent pumping. We must specify here that the sludge processing (i.e., thickening, mesophilic anaerobic digestion, dewatering and disposal) and the biogas energy recovery are performed off-site at a larger centralised facility. Therefore, the energy recovery data for this plant cannot be retrieved directly and can only be estimated using this model.
2.2.
Procedure to derive COD and TSS fractions
Starting from the typical data available for any plant (COD, BOD5, TSS and VSS), COD fractions, and TSS fractions in primary influent (PI) and primary effluent (PE) can be calculated from a mass balance on the primary settling tank. The estimation of pCOD/VSS for biodegradable and non-biodegradable particulate organics requires advanced measurements never performed during plant operations, and typically belonging to the research domain. One assumption is
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 5 8 e5 8 7 2
5861
Fig. 1 e Wastewater treatment plant layout and depiction of COD and energy flows. All symbols are defined in the paper notation.
considering the non-biodegradable fraction of the sCOD (i.e., the snbCOD) as conservative and equal to the filtered secondary effluent COD. This is acceptable when the plant is operated at an MCRT sufficiently high to remove the sbCOD to
Table 1 e Summary of process characteristics for the plant considered here. Parameter Flow rate (AVG, m3/d) TW (AVG, C) Activated sludge mean cell retention time (AVG, d) Digester sludge retention time (AVG, d) Influent CODin (mg/l) BOD5,in (mg/l) TSSin (mg/l) VSSin (mg/l) NH4eNin (mg/l) Primary effluent CODPI (mg/l) BOD5PI (mg/l) TSSPI (mg/l) VSSPI (mg/l) NH4-NPI (mg/l) Secondary effluent CODSE (mg/l) BOD5SE (mg/l) TSSSE (mg/l) VSSSE (mg/l) NH4eNSE (mg/l) Filtered secondary effluent CODfe (mg/l) BOD5,fe (mg/l) TSSfe (mg/l)
Value 60000 20 8.5 30 541 243 308 263 N/A
the highest extent. This approach has been used in previous works (e.g., Ekama, 2009; Takacs and Vanrolleghem, 2006) and is widely accepted in modelling. Also, the concentration of pCOD and VSS varies over the diurnal cycle. With our approach, we assume that all particulate materials, regardless of their biodegradability, are characterised by the same value of pCOD/VSS. This implies that the wastewater is uniform, i.e. the differential ratio of its components (lipids, carbohydrates, proteins, etc.) does not change with time. The calculations used to derive COD and TSS fractions are summarised in Table 2.
2.3. Carbon and energy footprint models as function of COD fractions 100 48 56 50
284 54 144 36 104 31 87 27 28 3.8 27 7.1 7.1 2.9 13 5.3 10 4.2 0.13 0.2 18 4.1 1,8 0.9 1.0 0.5
The CFP and eFP models are an evolution of previously published methodologies (Monteith et al., 2005; Rosso and Stenstrom, 2008). In this paper the model was applied to the process illustrated in Fig. 1. This process was selected since it is the most common scenario for a wastewater treatment plant layout worldwide. Fig. 1 illustrates the units included in this model, and summarises the mass flows amongst the biological and settling units. The expression of all mass in COD units allows us to express all the mass flows throughout the plant in terms of bCOD. In our approach, there is no difference in the transformation of primary and secondary sludge in terms of bCOD. That would not be true in terms of VSS, i.e. the primary sludge VSS destruction is expected to be far higher than secondary sludge for municipal wastewater. As usual with CFP and eFP models, several assumptions must be taken. Caution must be used when applying the model to other plants. The main assumptions taken in our work are summarised in Table 3.
5862
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 5 8 e5 8 7 2
mPS;TSS ¼ Q$TSSPI $hPS
Table 2 e Procedure for the estimation of COD and TSS fractions as calculated from commonly measured parameters.
where Q ¼ plant influent flow rate (m3/d) and
Parameter
Symbol
ASM symbol
Formula
Particulate COD Soluble COD Biodegradable COD Soluble non-biodegradable COD Soluble biodegradable COD Particulate biodegradable COD Particulate non-biodegradable COD Non-biodegradable VSS Biodegradable VSS Inert TSS
pCOD sCOD bCOD snbCOD
e e e SI a
sbCOD
SS
pCOD/VSS - VSS COD - pCOD 1,6 - BOD5 soluble COD of filtered SE sCOD - SI
pbCOD
XS
bCOD - SS
pnbCOD
XIa
pCOD - XS
nbVSS bVSS iTSS
e e e
pCOD/VSS - XI pCOD/VSS - XS TSS - VSS
pCODPI ¼ pbCODPI þ pnbCODPI
Mass flow modelling for primary treatment
For the following mass flow analysis we express all mass flows with units of massj/d, where j is any of the pertinent subscripts. For each mass flow equation, appropriate conversion factors are required, depending of the actual units desired (SI, USCS, technical metric units, etc.). The mass flow in and out of primary sedimentation is the product of primary influent and effluent concentrations and the flow rate. For each of the COD fractions a balance can be written between primary influent (PI) and effluent (PE):
(10)
is the total particulate COD in the primary influent (mg/l). The value of hPS is a key parameter defining the energy performance of the overall process, and it will be discussed below.
2.3.2.
a Conservative fractions.
2.3.1.
(9)
Mass flow modelling for secondary treatment
These flows derived from mass balance analysis analogous to the ASM1 approach with mean cell retention time (MCRT) sufficiently long to ensure complete hydrolysis of pbCOD (Henze et al., 2000). This assumption implies that the bCOD in the secondary effluent is all soluble (i.e., sbCOD), with exception of a negligible contribution of particulate from microorganisms being washed out of the secondary clarifier. Based on the mass balance analysis of a well-mixed activated sludge reactor (Metcalf and Eddy, 2003), the concentrations of COD fractions (mg/l) in the secondary effluent are: bCODSE x sbCODSE ¼
KS ½1 þ ðkd;H ÞMCRT MCRTðYH k kd;H Þ 1
(11)
snbCODSE ¼ snbCODPE
(12)
CODSE ¼ snbCODPE þ bCODSE
(13)
sbCODPE ¼ sbCODPI
(1)
snbCODPE ¼ snbCODPI
(2)
pbCODPE ¼ pbCODPI ð1 hPS Þ
(3)
where Ks ¼ half-saturation constant (mg/l) kd,H ¼ decay rate of heterotrophic biomass (1/d) YH ¼ heterotrophic biomass yield (massVSS/massCOD) k ¼ maximum substrate utilisation rate (massCOD/massVSS d) MCRT ¼ mean cell retention time (d). The oxidation of biodegradable COD and ammonia results in a daily active biomass production Px,active bio (massVSS/d):
pnbCODPE ¼ pnbCODPI ð1 hPS Þ
(4)
Px;active bio ¼
where, sbCODi ¼ soluble biodegradable COD (mg/l) snbCODi ¼ soluble non-biodegradable COD (mg/l) pbCOD i ¼ particulate biodegradable COD (mg/l) pnbCODi ¼ particulate non-biodegradable COD (mg/l) i ¼ primary influent (PI) or primary effluent (PE) and the sedimentation efficiency of the primary settler (hPS; %) is pbCODPE þ pnbCODPE pCODPE ¼ 100$ 1 hPS ¼ 100$ 1 pCODPI pbCODPI þ pnbCODPI
(5)
Q$YH $hASP $bCODPE Q$YA $NOx þ 1 þ ðkd;H Þ$MCRT 1 þ ðkd;A Þ$MCRT
(14)
where Q ¼ flow rate through the activated sludge process (m3/d) kd,A ¼ decay rate of autotrophic biomass (1/d) YA ¼ autotrophic biomass yield (massVSS/massN,oxidized) NOx ¼ concentration of oxidised nitrogen (mgN/l) and with the bCOD removal in the activated sludge reactor hASP (%) calculated from Eqs. (1), (2), and (11): bCODSE hASP ¼ 100$ 1 bCODPE KS ½1þðkd;H ÞMCRT ½MCRTðYH kkd;H Þ1 ¼ 100$ 1 sbCODPE þ pbCODPE
(15)
The mass flow of primary waste sludge in terms of COD and bCOD as well as VSS and TSS (mPS; massCOD/d, massbCOD/d, massVSS/d, massTSS/d respectively), leaving the sedimentation basin are:
Consequently, the mass flow of secondary sludge in terms of bCOD from the sludge wasting line mSS (massbCOD/d) is:
mPS;bCOD ¼ Q$pbCODPI $hPS
(6)
mSS ¼ 1:42$Px $ 1 fd
mPS;COD ¼ Q$pCODPI $hPS
(7)
mPS;VSS ¼ Q$VSSPI $hPS
(8)
(16)
where fd ¼ fraction of biomass contributing to biomass debris (). Similarly to the approach used in ASM1 and according to Grady et al. (1998), we assumed that fd represents the fraction
5863
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 5 8 e5 8 7 2
Table 3 e Main assumptions made for calculation of carbon and energy footprint. Process section Influent quality Primary sedimentation
Activated sludge
Anaerobic digestion
Dewatering section (belt-press) Energy consumption and recovery
Emission for off-site electricity generation
Assumptions Selected according to data from WWTP (Table 1) VSS and TSS removal efficiency ¼ 66% COD removal efficiency ¼ 35e54% (calculated on the basis of VSS efficiency removal and pCOD/VSS ratio between 1.07 and 1.87) Kinetic and stoichiometric parameters for heterotrophic bacteria @20 C: maximum growth rate (m): 6 d-1 Half-saturation constant: 20 g bCOD/m3 Yield (YH): 0.67 gCOD/gCOD Net decay rate kd: 0.12 d-1 Fraction of biomass contributing to biomass debris ( fd): 0.15 Kinetic and stoichiometric parameters for autotrophic bacteria @20 C: maximum growth rate (mN): 0.5 d-1 3 Half-saturation constant: 0.5 g NHþ 4 eN/m Yield (YN): 0.12 gVSS/g NHþ 4 -N Net decay rate kdN: 0.12 d-1 Arrhenius constants for correction of kinetic parameters with temperature: 1.07 for maximum growth rate, 1.04 for net decay rate MCRT ¼ 10 d Dissolved oxygen ¼ 2 mg/l Kinetic and stoichiometric parameters were assumed from Metcalf and Eddy (2003) Fine pore aeration with aSOTEa ¼ 15% bCOD solubilised ¼ 85% Biogas: 65% methane and 35% CO2 Methane production: 0.35 Nm3/kg bCOD,rem Methane fugitive emission: 2% VSS reduction calculated according to pCOD/VSS values Biomass production calculated according to Metcalf and Eddy (2003) bCOD in supernatant line returned to the headworks assumed as 3% of solubilised bCOD YANA ¼ 0.08 gVSS/gCOD kd,ANA ¼ 0.03 d-1 SRTdig ¼ 30 d Dry matter in influent biosolidsa: 2.1% Dry matter in dewatered sludgea: 19.2% Specific energy consumptiona (kWh/m3treated in the section): 0.026 for influent pumping, 0.007 for preliminary treatments, 0.032 for primary settling tank, 0.024 for secondary settling tank, 2.116 for sludge dewatering, 0.16 kWh/kg dry solids fed to the anaerobic digesters Efficiency of energy recovery from methane combustion: 50% Methane specific energy: 35.80 MJ/m3 Specific CO2eq emission for electricity generation and for calculation of CO2eq offset from biogas combustion: 0.245 kg CO2eq/kWhb
a Actual data at the WWTP modelled. b From EIA (2009).
of biomass that can be considered not biodegradable. The conversion factor 1.42 is required to express the mass of microbial VSS in COD units, assuming the generally acknowledged empirical formula C5H7NO2 for bacterial biomass (Metcalf and Eddy, 2003). In order to complete the analysis of energy and carbon footprint for the whole process, the total excess sludge production from ASP both in terms of VSS mass and COD must be calculated. The calculation of total excess sludge production is based on the mass balance analysis of both endogenous and exogenous biodegradable and non-biodegradable fractions as well as the inert suspended solids in a well-mixed activated sludge reactor. The typical assumption (Metcalf and Eddy, 2003) is that both the pnbCOD and the inert total suspended solids (iTSS, mg/l) not removed in primary sedimentation are embedded in the activated sludge flocs. Using this hypothesis, the daily excess sludge production Px,VSS
(massVSS/d), Px,TSS (massTSS/d), and Px,COD (massCOD/d) can be calculated as follows: Px;VSS ¼
Px;active bio 1 fd $kd;H $Q$YH $MCRT þ þ Q$nbVSSPE 0:85 0:85 1 þ kd;H $MCRT
Px;TSS ¼
Px;active bio 1 fd $kd;H $Q$YH $MCRT þ þ Q$nbVSSPE 0:85 0:85 1 þ kd;H $MCRT þ Q$iTSSPE
Px;COD ¼
Px;active bio þ
(18)
fd $kd;H $Q$YH $MCRT $1:42 1 þ kd;H $MCRT
þ Q$nbVSSPE $pCOD=VSS
2.3.3.
(17)
(19)
Mass flow modelling for sludge digestion
In our approach we assumed a value for bCOD transformation in the digester (hDIG) and CH4 and CO2 production were
5864
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 5 8 e5 8 7 2
estimated performing a bCOD mass balance analysis on the digester (summarised in Fig. 1). The influent mass flow of bCOD to the digester mDIG,in (massbCOD/d) can be calculated from the sum of Eqs. (6) and (16):
Considering that there is no COD destruction in an anaerobic reactor but only COD transformation, and assuming no nitrate or sulphate reduction, we can write:
mDIG;in ¼ mPS;bCOD þ mSS ¼ Q$ pCODPI $hPS þ 1:42Px;active bio $ 1 fd
(
(20) Assuming that a fraction of digester’s influent bCOD solubilises and that of the solubilised bCOD a fraction leaves as sbCOD in the supernatant line and a fraction is converted to anaerobic biomass, the bCOD mass flow in biogas can be calculated as:
Total COD mass
9 =
flow in digester inlet ; (
9 COD mass flow = in supernatant ;
¼
þ
9 8 < COD mass flow = :
in biogas
;
8 9 < COD mass flow = : in biosolids ;
8 9 8 9 mbCODCH4 > > mbCODsol 9 8 9 8 > > > > > > > > > > mbCODANA biomass > mbCODSN > > > > > > > = < = < = < = < bCOD bCOD bCOD bCOD of anaerobic ¼ > > > > > > > > transformed > ; : ; > > : > > > > solubilised > > > > in supernatant biomass produced : ; > ; : to biogas in the digester ¼ hDIG $ Q$pCODPI $hPS þ 1:42,Px; active bio , 1 fd mbCODSN mbCODANA biomass where hDIG ¼ bCOD solubilised in the digester (%; see Table 3).
mBS;bCOD
Px;ANA ¼
YANA ,DbCOD 1 þ ðkd;ANA Þ,SRTDIG
(22)
where YANA ¼ biomass yield in anaerobic conditions (massVSS/ massCOD) SRTdig ¼ digester solid retention time (d) Px,ANA ¼ biomass production under anaerobic conditions (massVSS/d) DbCOD ¼ difference between solubilised bCOD and bCOD in the supernatant line (massCOD/d). The methane mCH4 (massCH4/d) and biogas production mBG (massBG/d) can be calculated using Eq. (21): mCH4 ¼ mbCODCH4 ,bCH4 , rCH4 mBG ¼
mCH4 rBG 0:65 rCH4
(23) (24)
where QBG ¼ flow rate of biogas leaving the anaerobic digester (m3/d) bCH4 ¼ specific methane production (Nm3 =kgbCOD ) rCH4 ¼ methane density (kgCH4 =Nm3 ) rBG ¼ biogas density (kgCH4 =m3 ).
(21)
From Eq. (25), the biosolids mass flow in terms of both COD (mBS,COD as massCOD/d) and bCOD (mBS,bCOD as massbCOD/d) are:
9 8 9 8 9 8 < COD of = < COD of = < Total bCOD = COD in þ in the digester ¼ biomass biogas supernatant ; : ; : ; : influent produced produced ¼ mPS;COD þ Px;COD þ 1:42$Px;ANA ¼ Q$pCODPI $hPS þ 1:42$Px; active bio $ 1 f d mbCODCH4 mbCODSN þ 1:42$Px;ANA
In this paper we assumed that 85% of the bCOD in the digester influent is solubilised within the digester (i.e., hDIG ¼ 0.85), and that 3% of the solubilised bCOD leaves the digester in the supernatant line (mbCOD,SN). The bCOD converted in biomass under anaerobic conditions is calculated as:
(25)
(26)
COD is conservative and therefore only transformed from influent solids to effluent solids and gas, while VSS is not conservative and is destroyed producing different biogas yield for different types of VSS. Only the VSS portion of the TSS is destroyed whilst the non-volatile remainder is conservative, hence not all TSS are destroyed but are in fact reduced. Therefore, throughout the paper we refer to COD transformation, VSS destruction, and TSS reduction. The percentage reduction in terms of VSS and TSS in the digester can be calculated as follows:
Q$pCODPI $hPS þ Px; active bio $ 1 fd hDIG;VSS ¼ hDIG $ pCOD=VSS 100 1:42$Px;ANA $ (27) Px;VSS þ mPS;VSS hDIG;TSS ¼
Q$pCODPI $hPS þ Px; active bio $ 1 fd hDIG $ pCOD=VSS 100 1:42$Px;ANA $ Px;TSS þ mPS;VSS
(28)
where hDIG,VSS ¼ VSS destruction in the digester (%) hDIG,TSS ¼ TSS reduction in the digester (%) The balance in terms of COD units is a more detailed approach than the textbook-based VSS analysis: mBS;VSS ¼ ðQ$bVSSPS þ Q$bVSSSS Þ, 1 hDIG;VSS mBG;VSS where
¼ ðQ$bVSSPS þ Q$bVSSSS Þ, hDIG;VSS
(29) (30)
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 5 8 e5 8 7 2
mBS,VSS ¼ mass flow of biosolids calculated from VSS destruction (massVSS/d) mBG,VSS ¼ mass flow of biogas calculated from VSS destruction (massBG/d) bVSSPS ¼ biodegradable VSS in primary sludge (massVSS/d) bVSSSS ¼ biodegradable VSS in secondary sludge (massVSS/d) hDIG,VSS ¼ percent VSS destruction in the digester (%) The percent VSS destruction in the digester hDIG,VSS is solely a function of the digester solid retention time SRTdig (d; Metcalf and Eddy, 2003), which makes no distinction of the nature of the influent VSS (e.g., primary sludge vs. secondary sludge and above all biodegradable vs. non-biodegradable). Hence, the specific biogas production (expressed in Nm3 =kgVSSdestroyed ) is much more uncertain. The relative performance in terms of bCOD transformation, bVSS and VSS destruction, TSS reduction, and COD of the supernatant for variable pCOD/VSS values are reported in Table 4.
2.3.4.
Energy flow modelling
In Fig. 1 we also report the relationship between energy demand, energy requirements and the COD fractions. Total energy demand eD (kWh/d) is calculated as: eD ¼ eD;PS þ eD;ASP þ eD;SS þ eD;AD þ eD;O
(31)
where eD,PS ¼ energy demand for primary sedimentation (kWh/d) eD,ASP ¼ energy demand for activated sludge aeration (kWh/d) eD,SS ¼ energy demand for secondary sedimentation (kWh/d) eD,AD ¼ energy demand for anaerobic digestion (kWh/d). eD,O ¼ energy demand for other equipment (kWh/d) The energy demand eD (kWh/d) of the process illustrated in Fig. 1 is dominated by the aeration energy for biological oxidation in the activated sludge process and details on the calculation of eD,ASP can be found in Monteith et al. (2005) and in Rosso and Stenstrom (2005) and according to values indicated in Table 3. The dominance of aeration energy on the overall process energy consumption was confirmed for the location where this model was tested. The other significant energy demanding process (digestion) mainly relies on waste heat and the values used in this paper are specific to the plant modelled here, and are again summarised in Table 3. We excluded here influent and effluent pumping, since pumping is site-specific and coastal plants discharging against tidal cycles or plants injecting their effluent in deep aquifers or plants with deep sewers would be unfairly disadvantaged in the overall COD-based calculations. This is because pumping energy is calculated only on the basis of
Table 4 e Performance variation of the anaerobic digester in terms of bCOD, VSS, and TSS. 1.07 / 1.87
pCOD/VSS (gCOD/gVSS) bVSS destruction VSS destruction TSS reduction COD transformation COD in supernatant
% % % % mg/l
84.9 / 86.9 47.9 / 55.9 37.1 / 45.8 44.1 / 36.6 441 / 948
5865
hydraulic flow, regardless of the COD concentration. Also, the energy demand for effluent disinfection heavily depends on the disinfection modality (chemical with on-site chemical generation, chemical with purchase of disinfectant from an off-site location, or by irradiation) and is therefore sitespecific. The actual calculation of energy demand adsorbs in eO all other minor contributions assumed from literature and site-specific data (details on methodology published in Rosso and Stenstrom, 2005, 2008). In this paper, we assume a fixed concentration of ammonia and TKN entering the plant, so that their effect on oxygen therefore energy demand is constant throughout our pCOD/ VSS domain. The energy recovery eR (kWh/d) from digester biogas is proportional to the biogas produced (Eq. (24)) which is in turn proportional to the removal of bCOD in the digester (Eqs. (22) and (23)): eR ¼ hER $hBG $mBG
(32)
where hER ¼ efficiency of the energy recovery unit () hBG ¼ calorific value of the biogas (kJ/kgbiogas)
2.3.5.
Carbon footprint calculation
The following contributions to CO2,eq emission were considered: direct CO2 emission from biological processes (ASP and AD); direct CO2 emission from biogas combustion; indirect CO2 emission from off-site power generation; CO2,eq offset from biogas energy recovery; CO2,eq emission due to CH4 fugitive emission; CO2,eq emission due to CH4 released during biosolids dewatering. The global warming potential for CH4 was calculated at the 100 year horizon, i.e. 25 kgCO2 eq =kgCH4 (IPCC, 2007). We quantify here only the carbonaceous emissions or emission equivalents from treatment, since the focus of this paper is the effects of COD fractional variations on carbon and energy footprint. N2O has no effect on energy footprint and at the present moment there is no evidence linking COD fractional variations on varying N2O emissions. In the calculation of the CO2 direct emission due to microbial respiration in ASP, it is important to highlight the carbon emission intensity of COD (i.e. the amount of CO2 emitted per unit of COD oxidised, or kCOD). In this study we hypothesised that wastewater organic compounds could be represented through the formula C10H19O3N which is widely used for the case of domestic wastewater (inter alia, Monteith et al., 2005; Shahabadi et al., 2010). Considering the following oxidation reaction of the compound: C10 H19 O3 N þ 25=2O2 /9CO2 þNHþ 4 þHCO3 þ 7H2 O
(33)
we obtain that kCOD ¼ 0.99 kgCO2 eq =kgCOD . Analogously, in case of activated sludge biomass, considering the following stoichiometric relationships for biomass decay in aerobic environment: C5 H7 O2 N þ 5O2 /4CO2 þNHþ 4 þHCO3 þ H2 O
(34)
we obtain that kBIOMASS ¼ 1.03 kgCO2 eq =kgCOD . Both Eqs. (33) and (34) assume that the HCO 3 produced is in equilibrium in the liquid environment, i.e. it is not converted to H2CO3. In this way, we are able to evaluate the CO2 emitted due to bacterial respiration as a function of MCRT, whose variations are in
5866
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 5 8 e5 8 7 2
turn reflected on the amount of biomass that undergoes decay. With the exception of CH4 traces released in the activated sludge off-gas, the remaining bCOD not used for synthesis in the activated sludge reactor is oxidised to CO2: mCO2 ;ASP ¼ Q$kCOD $ð1 YH Þ$hASP $bCOD þ Q$kBIOMASS $YH $hASP $bCOD
kd;H $MCRT 1 fd 1 þ kd;H $MCRT
NOx 1 þ kd;A $MCRT kd;A $MCRT 1 fd þ Q$kBIOMASS $YA $NOx $ 1 þ kd;A $MCRT x Q$kCOD $ð1 YH Þ$hASP $bCOD kd;H $MCRT þ Q$kBIOMASS $YH $hASP $bCOD 1 fd 1 þ kd;H $MCRT
Q$kBIOMASS $YA $
performing sludge dewatering in this case, and we assumed that half of the methane contained in the biosolids is released during dewatering, and the other half is returned to biological oxidation where it ultimately becomes CO2 with negligible contribution to the total CO2 from influent bCOD oxidation emitted in the off-gas. Hence, the CO2,eq emission due to CH4 fugitive emission (mCH4 eq;fugitive ) and to CH4 released during biosolids dewatering (mCH4 ;dewatering ) is calculated as: mCO2 eq;fugitive ¼ 25$ mCH4 ;fugitive $hFE þ mCH4 ;dewatering
Finally, total CO2,eq emission (kgCO2 eq =d) is calculated as the sum of all contributions: mCO2 eq ¼ mCO2 ;ASP þ mCO2 ;AD þ mCO2 ;CH4 comb þ mCO2 eq;PG mCO2 eq;offset þ mCO2 eq;fugitive
(35) where kCOD ¼ carbon emission intensity of COD (kgCO2 eq =kgCOD ) kBIOMASS ¼ carbon emission intensity of biomass (kgCO2 eq =kgCOD ). The first term summed in Eq. (35) is the direct oxidation of influent bCOD, while the second term is the endogenous respiration of microbial biomass. Direct CO2 emission from AD is calculated as: mCO2 ;AD ¼
mBG $r $0:35 rBG CO2
(36)
Direct CO2 emission from biogas combustion is calculated according to the combustion reaction of methane: mCO2 ;CH4 comb ¼ mBG $ð1 hFE Þ$
44 16
(37)
where hFE represents the methane fugitive emission (% of total methane produced). Here we assume that 100% of the methane entering the combustion chamber is fully combusted to CO2, although if site-specific data were available an efficiency of methane oxidation during combustion (<100%) could be defined. The indirect CO2 emission from off-site power generation (mCO2 ;PG ), is calculated on the basis of specific CO2 emission per unit of energy generated in the region where the treatment plant is located (EIA, 2009) (see Table 3 for details) as: mCO2 eq;PG ¼ kPG $eD
(38)
The CO2,eq offset from biogas energy recovery (mCO2 eq;offset ) is calculated as: mCO2 eq;offset ¼ eR $kPG
(39)
An assumption must be made on the methane released through fugitive emission. This is a non-point source due to the extensive length of the spatially distributed network of pipes, valves, safety units, and appurtenances for biogas conveyance. Currently, most large scale biogas flow metres cannot measure flow with an error below 5%, and therefore a biogas mass in the percent range is always unaccounted for. We assumed here a 2% fugitive emission. Furthermore, depending on the sludge dewatering technology employed, some of the methane contained in the biogas-saturated biosolids to dewater may be released directly to the atmosphere. This is the case of belt-press facilities, such as the one
(40)
(41)
Both CFP and eFP are normalised per unit bCOD removed from the wastewater (kgCO2 eq =kgbCOD;rem and kWh/kgbCOD,rem, respectively) and per unit volume of treated wastewater (kgCO2 eq =m3ww;treated and kWh/m3ww,treated, respectively).
2.4.
Model domain
Previously, several publications reported ranges of pCOD/VSS ratios (e.g. Takacs and Vanrolleghem, 2006). These ranges span over what we refer here as a range of possible pCOD/VSS values (from 1.07 to 2.87 gCOD/gVSS, for pure carbohydrates and pure lipids, respectively). By utilising our procedure for COD fractions calculation, the range of possible pCOD/VSS reported by Takacs and Vanrolloghem produces a number of unacceptable COD fractions (defined as sCOD < 0). The procedure fails (100% of fractions with sCOD < 0) with pCOD/ VSS higher than 2.59 gCOD/gVSS. This is because very high values of pCOD/VSS, more representative of pure lipids, are unrealistic for a municipal wastewater. In IWA (2008), Henze and Comeau report the typical composition of raw municipal wastewater with minor contribution of industrial wastewaters. Their values, used here as the most probable range, lead to pCOD/VSS in the range 1.22e1.50 gCOD/gVSS (corresponding to sCOD/COD of 0.40e0.26, respectively). On the other hand, in the literature review discussed by Ekama (2009), when the same value of pCOD/VSS is assumed for biodegradable and non-biodegradable particulate organics, pCOD/VSS spans over 1.40e1.62 gCOD/gVSS for primary sludge and over 1.32e1.57 and 1.19e2.17 for biodegradable and non-biodegradable particulate organics, respectively. Based on these sources we selected a pCOD/VSS domain of 1.07e1.87. This corresponds to 26.2% unacceptable COD fractions (i.e., sCOD < 0), which are excluded. These correspond to the influent conditions not usual for regular operations (e.g., during and immediately following storm events, during and immediately following holidays, sampling errors, etc.). The use of input data smoothing or averaging would reduce short-term fluctuations and highlight longer-term trends or cycles, thus reducing the amount of unacceptable COD fractions. On the other hand, this would protract the irregularities of short-term phenomena over the time-averaged period. In this research we do not aim at modelling COD fractions dynamically but
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 5 8 e5 8 7 2
rather at developing a simplified approach for modelling the process energy in terms of COD fractions. Thus, our aim in defining the model domain is to reduce the effects of the inevitable plant data variation.
3.
Results and discussion
We analysed process data sets and produced yearly averages, reported in Fig. 2. The average coefficient of variation for the COD fractions in primary influent and effluent was 0.37 and 0.40, respectively. When calculating the average coefficient of variation for secondary effluent fractions, its value exceed 1.0 due to the substantial decrease of SS and XI. The COD, VSS and TSS removal during primary settling, calculated using plant data, equals 47.5%, 67.1% and 66.3%, respectively. Assuming a 66.7% VSS and TSS removal, and using our procedure, within the pCOD/VSS domain range, the COD removal in primary settling would span between 35% and 54%. The actual (i.e., based on plant data) and estimated (i.e., based on our procedure) COD removal in primary settling match when pCOD/VSS equals 1.57 gCOD/gVSS. In Fig. 2, one can notice the expected relative increase in SS (i.e., sbCOD) after primary settling, due to the decreased relative amount of particulate COD after a successful primary sedimentation. Even though the fractions reduce in absolute terms, when pCOD/VSS increases (Fig. 2, right panel) the relative increase in SS (i.e., sCOD) is maximum. Due to its conservative nature, SI always dominates the secondary effluent fraction, followed by XS which is associated with the fraction of activated sludge biomass that does not settle well and is released from the secondary clarifiers. Since the process energy demand is largely dominated by the ASP energy (Reardon, 1995; Rosso and Stenstrom, 2005), in all cases within the pCOD/VSS range the reduction of SS from primary effluent to secondary effluent is largely associated with the aeration energy demand for secondary oxidation. Conversely, since the energy recovery is associated with the solids flow to
5867
the digester, even though the relative ratios of XS along the process train appear to be similar for all the pCOD/VSS cases, the absolute value of digestible solids sent to the digesters increases with increasing pCOD/VSS, thereby increasing the energy recovery of the process. In sum, with higher pCOD/VSS (hence, with lower sCOD/COD), the process energy demand is decreased and its energy recovery increased, with compounding beneficial effect (highlighted in Figs. 3 and 4). We report our process modelling results in terms of both mass of CO2 equivalent emitted per unit bCOD removed (kgCO2 eq =kgbCODremoved ) and per unit volume of wastewater treated (kgCO2 eq =m3ww;treated ). Both normalisations are valuable and have different significance, although in our results they overlap since we do not vary the average wastewater concentration. Reporting carbon equivalent emission per unit bCOD is the theoretical calculation of the global warming potential of the wastewater, since the alternative of no treatment at all would still yield a carbon equivalent emission (untreated bCOD is typically assumed to eventually become 50% CO2 and 50% CH4), beside the obvious public and environmental health threats associated with the untreated sewage discharge (Rosso and Stenstrom, 2008). The other output normalisation (i.e. kgCO2 eq =m3ww;treated ) is a measure of the site-specific carbon footprint of a process, as it implicitly includes information on the wastewater concentration. The two normalised measures are linearly scalable once the concentration of the influent wastewater is not varied, such as in our case. When dynamic modelling is practiced, the conversion of one normalisation to the other would follow the dynamic wastewater concentration curves. Finally, we choose here to omit the results in terms of absolute values of CO2 mass flows (i.e., kgCO2 eq =d), for these are highly site- and assumption- specific and have little or no value for modelling research, but are useful for emission reporting and process performance benchmarking, both outside the scope of this research. Fig. 3 shows the carbon equivalent emission for the CFP contributions, expressed as kgCO2 eq =kgbCODremoved and
Fig. 2 e COD fractions from the case plant considered for increasing pCOD/VSS ratios (average of 365 days for each panel; cases with negative sCOD excluded).
5868
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 5 8 e5 8 7 2
sCOD/COD 0.47
0.44
0.40
0.36
0.33
0.29
0.25
0.22
0.18
0.20
0.16
0.40 0.12
0.08 0.20
0.04
0.00
Carbon-equivalent emission (kg CO2/m3ww,treated)
Carbon-equivalent emission (kgCO2/kgCOD,removed )
0.60
0.00 1.07
1.27
1.47
1.67
1.87
pCOD/VSS (gCOD/gVSS) Fig. 3 e Carbon footprint components (expressed as kgCO2 eq =kgbCODremoved and as kgCO2 eq =m3ww;treated ) against varying pCOD/ VSS and sCOD/COD ratios. sCOD/COD 0.47
0.40
0.36
0.33
0.29
0.25
0.22
0.18
40%
0.35
20% 0.12
0.30
0.11
0.25
(kgCO2 / kgCOD,removed)
0%
(kgCO2 / m3ww,treated)
Carbon-equivalent emission
0.13
0.44
-20%
-40% 1.07
1.27
1.47
1.67
1.87
pCOD/VSS (gCOD/g VSS) Fig. 4 e Energy footprint components, and energy deficit for varying pCOD/VSS and sCOD/COD ratios.
5869
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 5 8 e5 8 7 2
sCOD/COD 0.47
0.44
20%
Parameter variation
0.40
0.36
0.33
0.29
0.25
0.22
0.18
CO2 from activated sludge respiration CO2 from combustion of CO2 in biogas CO2eq from biogas fugitive emission CO2eq from biosolids dewatering CO2eq from combustion of CH4 in biogas
30%
10%
0%
-10%
-20%
-30% 1.07
1.27
1.47
1.67
1.87
pCOD/VSS (g COD/gVSS)
sCOD/COD 0.47
0.44
0.40
0.36
0.33
0.29
0.25
0.22
0.18
30%
20%
Parameter variation
kgCO2 eq =m3ww;treated against varying pCOD/VSS and sCOD/COD ratios. Higher particulate in the influent implies higher removal of bCOD in the primary settler (affecting all but the blue line in Fig. 3), which results in lower CO2 emitted by bacterial respiration (blue line). This also implies lower activated sludge biomass production, which increases the fraction of sludge from primaries in the digester’s influent, yielding more methane and proportionally lower biosolids mass for disposal. The same discussion is applicable to Fig. 4, where the energy demand, recovery, and deficit are plotted for varying pCOD/VSS and sCOD/COD values. In this graph, as the fraction of sCOD decreases, the energy burden to the aeration system decreases, in turn driving down the overall process energy demand. Concurrently, the particulate fraction of the COD increases, thereby increasing the solids load to the digester and its biogas production. The energy recovery then increases, and the process energy balance shifts from a net energy deficit at low pCOD/VSS (i.e., high sCOD/COD) to a net energy recovery at high pCOD/VSS (i.e., low sCOD/COD). The energy demand and recovery trends in Fig. 4 may appear discrepant, but it must be remembered that when the bCOD is treated in the secondary rather than the primary, its fate is both respiration corresponding to a net energy demand, and synthesis corresponding to a partial energy recovery through biomass digestion. But since not all VSS are equally digestible (in fact pCOD/VSS for primary sludge span over 1.07e1.87 while is constant at 1.42 gbCOD/gVSS for the activated sludge biomass), by treating more bCOD in the secondary treatment the process shifts its energy balance from high recovery to much higher demand, in proportion. Hence, the different slopes of energy demand and recovery in Fig. 4. The energy deficit shifts at a certain pCOD/VSS value, showing that process energy selfsufficiency may be possible, as demonstrated by Wett et al. (2007), although unlike the case of Wett et al. (2007) most processes are fully aerobic (i.e., with higher energy requirements) and wastewaters could be in the lower range of pCOD/ VSS and therefore seldom able to yield sufficient energy to fully offset the energy demand, as discussed by Ekama (2009). Also, the energy recovery is calculated assuming an efficiency for the energy recovery process. This varies amongst energy recovery processes (e.g., combined heat and power generation, fuel cells, mechanical engines, etc.) and for large installations could be as efficient as 33e40% as electrical energy recovery and 48e50% as thermal energy recovery (Po¨schl et al., 2010). Combined energy recovery (i.e., electrical plus thermal) up to 60% is possible (IWA, 2008). In case of warm climates the waste heat may be excessive and only partially recovered, therefore the combined energy recovery may be lower than 60%, hence our assumption of 50% energy recovery for this process. Nevertheless, in cold weather facilities the waste heat from co-generation may not be sufficient and energy importation (e.g., as supplemental natural gas purchase) may be required. We present the results of the model sensitivity analysis in the graphs in Fig. 5 and in Table 5. The relative behaviour of the parameter variation in the pCOD/VSS range confirms the results reported above. CO2 emission from respiration decreases with pCOD/VSS increasing. This is due to a decreased fraction of soluble substrate to be biologically
10%
0%
-10%
-20%
-30% 1.07
1.27
1.47
1.67
1.87
pCOD/VSS (g COD/gVSS )
Fig. 5 e Sensitivity charts.
oxidised in the aeration tank. At the same time, the energy demand increases with decreasing pCOD/VSS ratio due to the higher fraction of soluble material to be biologically oxidised in the aeration tank. Being aeration the largest contributor to power consumption, the overall power consumption is driven upwards. The energy recovery increases with increasing pCOD/VSS ratio, due to the higher solids fraction sent to digestion directly from primary sedimentation. Coupled with the associated decrease in soluble matter to be biologically oxidised, an increased pCOD/VSS ratio leads to overall decreased net energy usage and lowering energy deficit.
5870
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 5 8 e5 8 7 2
Table 5 e Summary of sensitivity analysis. pCOD/VSS (gCOD/gVSS) / Variation of: CO2 from respiration CO2 in biogas CO2eq from biogas CH4 combustion CO2eq of biogas fugitive emissions CO2eq of CH4 emission during sludge dewatering CO2eq of N2O emission from activated sludge CO2eq of energy demand CO2eq of energy recovery
% % % % % % % %
1.07
1.17
1.27
1.37
1.47
1.57
1.67
1.77
1.87
25.23 24.18 24.18 24.18 4.08 0.00 10.16 24.18
18.40 18.06 18.06 18.06 2.93 0.00 7.35 18.06
12.04 11.89 11.89 11.89 1.88 0.00 4.78 11.89
5.89 5.90 5.90 5.90 0.91 0.00 2.32 5.90
0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
5.45 6.18 6.18 6.18 0.83 0.00 2.08 6.18
10.03 11.64 11.64 11.64 1.53 0.00 3.81 11.64
14.57 17.25 17.25 17.25 2.21 0.00 5.49 17.25
19.26 23.18 23.18 23.18 2.89 0.00 7.22 23.18
Biogas production decreases with decreasing pCOD/VSS ratio, due to the decreased particulate fraction sent to digestion from primary sedimentation and to slightly increased soluble substrate available to activated sludge synthesis. Along with biogas production, its fugitive emission, here quantified as a percent of biogas production, decreases with the same trend. Biogas emission during dewatering, however, decreases by a lesser extent because of the varying biogas yield (expressed as Nm3biogas/kgVSS,destroyed) due to varying composition of primary vs. secondary sludge in the digester influent. Our analysis cannot distinguish, at this moment, the different weight that regional practices, such as disposing of food waste in sewers, or site-specific operations, such as the addition of coagulants to primary sedimentation, could have on the modelling results. Food waste disposal in wastewater, which could be presented as a carbon footprint mitigation factor (Evans, 2009), is typical in the United States. Other areas of the world with moderate or no disposal of food waste in sewers would be expected to have not only lower TSS and VSS values in their influent, but also different pCOD/VSS ratios. These would not necessarily be lower, as food waste is largely composed of cellulosic matter, which is in the lower range of pCOD/VSS, according to Takacs and Vanrolleghem (2006). Also, the practice of source-separation of fats, oil and grease (FOG), especially for commercial establishments such as restaurants or communal kitchens, and its diversion to digestion via collection programs, offers the opportunity for increased biogas generation and for sheltering the activated sludge from the elevated energy requirements associated with FOG oxidation. Since FOG is a “rich” VSS component (i.e., very high pCOD/VSS ratio according to Takacs and Vanrolleghem, 2006), the expectable biogas yield in the digester could be higher, in proportion. In the case of industrial wastewater treatment plants, the pCOD/VSS ratios would be skewed by sector, but the model structure would not need modification. For example, a ratio of pCOD/VSS close to 1.07 is characteristic of wastewater rich in carbohydrates such as winery or brewery wastewater, and previous CFP modelling (e.g. Rosso and Bolzonella, 2009) could be further refined by introducing the current approach on COD fractions. This model offers a rigorous tool to evaluate the effect of variations in primary treatment on the process eFP and CFP. For example, this model can be used to quantify the eFP and CFP effects and the associated costs vs. the benefits of enhancing the removal of sludge in the primary treatment via
the aid of coagulants. Certainly, the addition of a coagulant to enhance primary sedimentation, despite its operating cost, would be expected to relieve the activated sludge from part of the load which in turn would be sent to digestion, with higher energy recovery. The energy deficit would be expected to decrease, but more often than not the choice of adding coagulants is driven by either land constraints for sedimentation expansion or need to sequester sulphides to curb odorous release. In this case, the ratio of pCOD/VSS in the primary effluent would be different than in the case of no coagulant added, and the model should be adapted accordingly. The opportunity to effectively enhance primary sedimentation must be also analysed under the point of view of incidental enhanced removal of the fraction of bCOD (i.e., rbCOD) that favours denitrification, hence BNR. Nutrient removal could suffer in case of wastewater with low C/N and C/P ratios (Puig et al., 2010) and extended MCRT to achieve nutrient removal in activated sludge may be required. Further research should address the effects of differential removal of COD fractions during enhanced primary sedimentation on the effectiveness of nutrient removal. The temporal variability in data input and its reflection on the COD fractions and on the model domain discussed in the methodology should be the object of future research. Using a moving average of the influent data to smoothen the effects of short-term irregularities would highlight longer-term trends and seasonal cycles. For example, one here should argue that the temperature of the influent wastewater and the hydraulic residence time in the sewer network have an effect on the influent sCOD/COD ratio, thereby on pCOD/VSS. That is expected to be true, due to higher rate of substrate hydrolysis and consequently methanogenesis at warmer temperature and longer residence time (Guisasola et al., 2008). The plant we modelled here is located in an area with warm wastewater throughout the year where major seasonal variations could not be detected. Another plant with four seasons and a substantial difference in wastewater temperature over the yearly cycle would be expected to exhibit different behaviour. Also, in our modelling we do not currently compare plants fed by sewers of different length or by combined sanitary and storm sewers. Sewer length would have an effect on the sCOD/COD ratio, especially at higher wastewater temperatures where sewers act in effect as plug flow reactors for hydrolysis (Guisasola et al., 2008). In any case, our model begins calculating the COD fractions at the plant influent well, and is hence independent of the in-sewer processes. However, due to different sewer geometry and sewage characteristics,
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 5 8 e5 8 7 2
a plant could have different influent fractions for same sewage or similar fractions for different sewage, respectively. Future research efforts should address these two sensitive components. Also, future research should investigate the effects of varying NOD/TOD for a given pCOD/VSS ratio, as temperature and sewer length would affect the release of ammonia from organic nitrogen. The ammonia released in the sewer is treated in the secondary process, thereby increasing even further the energy burden on the aeration system. Colder wastewaters and shorter sewer lines could be expected to have an advantage, as the organic nitrogen still embedded in the solids would be sent to digestion instead of secondary nitrification/denitrification. This would be of no consequence in the (usual) case of supernatant return to the head of the primary treatment. Therefore, in our current modelling work we did not consider variable ammonia values, since the supernatant here is fully returned to the headworks. However, as alternative low energy footprint options for side-stream treatment of highly rich nitrogen flows (such as digesters’ supernatant) become available [e.g., the simultaneous chemical precipitation of nitrogen with magnesium ammonium salts (Schultze-Rettmer, 1991; Zdybiewska and Kula, 1991), or biological treatment with anaerobic or microaerophilic processes (e.g., Siegrist, 1996; van Dongen et al., 2001; Fux et al., 2003; Wett, 2007)], they should also be considered for reducing the overall process energy and carbon footprint by relieving the main biological process form loading shocks and energy peaks. This model does not aim at replacing the full-featured process models (e.g., ASM, ADM) but introduces a simplified methodology that relies on the restricted data sets typically available for most treatment plants. The next logical step should be the plant-wide integration of the full-featured process models and their extension to include the energy and carbon footprint components.
4.
Conclusions
We analysed in this paper the different effects of COD fractions on carbon and energy footprint of a wastewater treatment process using activated sludge with nutrient removal and anaerobic sludge digestion with biogas energy recovery. Also, we introduce a simplified procedure to evaluate COD fractions using process parameters typically collected in plants (e.g., VSS, BOD5, etc.) and assuming a pCOD/VSS value. A real municipal wastewater treatment process was modelled with this procedure and the results for energy and carbon footprint, as direct and indirect emissions, are reported here. Our results show that the ratio sCOD/COD may alter the carbon and energy footprint of a process, as it dictates where the COD is processed: either recovering energy after biogas production in a digester, or demanding energy for aerobic oxidation in the activated sludge reactor. For a given process, the increase of the ratio sCOD/COD increases the process carbon and energy footprint. Conversely, an increase in particulate (e.g., pCOD/COD) removed in the primary sedimentation would reduce the energy demand on the aeration
5871
reactors and the associated CO2 direct emission from respiration and indirect emission for power generation. Even though the apparent conclusion is to promote primary sedimentation, care must be used during process analysis since a fraction of the COD necessary for proper nutrient removal downstream may be incidentally removed.
references
Ahn, J.H., Kim, S., Park, H., Rahm, B., Pagilla, K., Chandran, K., 2010. N2O emissions from activated sludge processes, 2008e2009: results of a national monitoring survey in the United States. Environ. Sci. Technol. 44, 4505e4511. de Haas, D., Foley, J. and Barr, K., 2008. Greenhouse gas inventories from WWTPs e The trade-off with nutrient removal. Proc. Water Environment Federation Sustainability Conference, National Harbour, MD. EIA - United States Energy Information administration, 2009. Annual Energy Outlook 2009 e With projections to 2030 DOE/ EIA-0383(2009), Washington, DC. Ekama, G.A., 2009. Using bioprocess stoichiometry to build a plant-wide mass balance based steady-state WWTP model. Wat. Res. 43 (8), 2101e2120. Evans, T.D., 2009. Climate change impacts of food waste diversion to anaerobic digesters. Proc. 23rd Water Environment Federation, Annual Residuals & Biosolids Conference, 1056e1076, Portland OR. Foley, J., de Haas, D., Hartley, K., Lant, P., 2010. Comprehensive life cycle inventories of alternative wastewater treatment systems. Wat. Res. 44 (5), 1654e1666. Fux, C., Lange, K., Huber, P., Grueniger, B., Siegrist, H., 2003. Nitrogen removal from digester supernatant via nitrite e SBR or SHARON? Wat. Sci. Technol. 48 (8), 9e18. Grady, L.C.P., Daigger, G.T., Lim, H.C., 1998. Biological Wastewater Treatment, second ed. CRC Press, New York. Guisasola, A., de Haas, D., Keller, J., Yuan, Z., 2008. Methane formation in sewer systems. Wat. Res. 42 (6e7), 1421e1430. Henze, M., 1992. Characterization of wastewater for modelling of activated sludge processes. Wat. Sci. Technol. 35 (9), 113e120. Henze, M., Grady, C.P.L., Gujer, W., Marais, G.v.R., Matsuo, T., 1987. Activated Sludge Model No. 1. Scientific and Technical Report No. 1. IAWPRC, London. Henze, M., Gujer, W., Mino, T., van Loosdrecht, M.C.M., 2000. Activated Sludge. Hulsbeek, J.J.W., Kruit, J., Roeleveld, P.J., van Loosdrecht, M.C.M., 2002. A practical protocol for dynamic modelling of activated sludge systems. Wat. Sci. Technol. 45 (6), 127e136. IPCC, 2007. Climate change 2007: the physical science basis. In: Solomon, S., Qin, D., Manning, M., Chen, Z., Marquis, M., Averyt, K.B., Tignor, M., Miller, H.L. (Eds.), Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge University Press, Cambridge, NY, p. 996. IWA, 2008. Biological Wastewater Treatment e Principles, Modelling and Design. IWA Publishing. Metcalf, & Eddy, Inc, 2003. Wastewater engineering: treatment and reuse. In: Tchobanoglous, G., Burton, F.L., Stensel, H.D. (Eds.), McGraw-Hill Series in Civil and Environmental Engineering, fourth ed. (New York). Metcalf, L., Eddy, H.P., 1914. American Sewerage Practice, first ed. McGraw-Hill, New York. Monteith, H.D., Sahely, H.R., MacLean, H.L., Bagley, D.M., 2005. A rational procedure for estimation of greenhouse-gas emissions from municipal wastewater treatment plants. Wat. Environ. Res. 77, 390e403.
5872
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 5 8 e5 8 7 2
Models ASM1, ASM2, ASM2d, and ASM3. Scientific and Technical Report n. 9. IWA Publ., London. Po¨schl, M., Ward, S., Owende, P., 2010. Evaluation of energy efficiency of various biogas production and utilization pathways. Appl. Energy 87 (11), 3305e3321. Park, H.D., Lee, Y.H., Kim, H.B., Moon, J., Ahn, C.H., Kim, K.T., Kang, M.S., 2010. Reduction of membrane fouling by simultaneous upward and downward air sparging in a pilotscale submerged membrane bioreactor treating municipal wastewater. Desalination 251 (1e3), 75e82. 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. Wat. Res. 44 (11), 3375e3384. Reardon, D.J., 1995. Turning down the power. Civ. Eng. 65 (8), 54e56. Roeleveld, P.J., van Loosdrecht, M.C.M., 2002. Experience with guidelines for wastewater characterisation in the Netherlands. Wat. Sci. Technol. 45 (6), 77e87. Rosso, D., Bolzonella, D., 2009. Carbon footprint of aerobic winery wastewater treatment. Wat. Sci. Technol. 60 (5), 1185e1189. Rosso, D., Stenstrom, M.K., 2005. Comparative economic analysis of the impacts of mean cell retention time and denitrification on aeration systems. Wat. Res. 39 (16), 3773e3780. Rosso, D., Stenstrom, M.K., 2008. The carbon-sequestration potential of municipal wastewater treatment. Chemosphere 70, 1468e1475. Rosso, D., Desai, A.S., Tseng, L.Y., 2009 Effects of nitrous oxide emissions on process carbon footprint of wastewater treatment plants, Proc. 2nd IWA BNR Conf., Krakow, Poland. Schultze-Rettmer, R., 1991. The simultaneous chemical precipitation of ammonium and phosphate in the form of
magnesium-ammonium-phosphate. Wat. Sci. Technol. 23, 659e667. Shahabadi, B.M., Yerushalmi, L., Haghighat, F., 2010. Estimation of greenhouse gas generation in wastewater treatment plants e Model development and application. Chemosphere 78, 1085e1092. Siegrist, H., 1996. Nitrogen removal from digester supernatant e Comparison of chemical and biological Methods. Wat. Sci. Technol. 34 (1e2), 399e406. Takacs, I. and Vanrolleghem P.A., 2006. Elemental balances in activated sludge modelling. Proc. of the IWA World Water Congress, 10e14 September 2006, China. van Dongen, U., Jetten, M.S.M., van Loosdrecht, M.C.M., 2001. The SHARON-Anammox process for treatment of ammonium rich wastewater. Wat. Sci. Technol. 44 (1), 153e160. Vanrolleghem, P.A., Insel, G., Petersen, B., Sin, G., De Pauw, D., Nopens, I., Weijers, S. and Gernaey, K., 2003. A comprehensive model calibration procedure for activated sludge models. Proc. of WEFTEC 76th Annual Technical Exhibition and Conference, October 11e15, Los Angeles, California. Vollertsen, J., Almeida, M.d.C., Hvitved-Jacobsen, T., 1999. Effects of temperature and dissolved oxygen on hydrolysis of sewer solids. Wat. Res. 33 (14), 3119e3126. Wett, B., 2007. Development and implementation of a robust deammonification process. Wat. Sci. Technol. 56 (7), 81e88. Wett, B., Buchauer, K. and Fimml, C., 2007. Energy self-sufficiency as a feasible concept for wastewater treatment systems. Proc. IWA LET Conf., Singapore. Asian Water, 09/07, 21e24. Zdybiewska, M.W., Kula, B., 1991. Removal of ammonia nitrogen by the precipitation method, on the example of some selected waste waters. Wat. Sci. Technol. 24 (7), 229e234.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 7 3 e5 8 8 5
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
The impact of transfers on water quality and the disturbance regime in a reservoir Roberta Fornarelli*, Jason P. Antenucci 1 Centre for Water Research, University of Western Australia, M023, 35 Stirling Hwy, Crawley 6009, Western Australia, Australia
article info
abstract
Article history:
The effect of water transfers between two reservoirs on the water quality of the receiving
Received 7 June 2011
reservoir was investigated over a 9-year period (2000e2008). Different management strat-
Received in revised form
egies were implemented in term of the magnitude and timing of water transfers, i.e. the
1 August 2011
amount of transferred volume and the frequency at which transfers occurred. These
Accepted 24 August 2011
different operational modes were analysed to determine changes in nutrient and metal
Available online 1 September 2011
concentrations, chlorophyll a, algal genera and biovolume. During high water transfers,
Keywords:
selected due to the high silica content of the pumped inflow and a significant shift in
Inter-basin water transfer
cyanobacteria genera occurring from Microcystis to nitrogen-fixing genera. The magnitude
Multi-objective management
and timing of water transfers exerted a strong control on phytoplankton competition and
Magnitude
disturbed the typical seasonal succession during low pumping years of a spring diatom
Timing
bloom followed by summer cyanobacteria dominance: intensive and frequent water
Internal production
transfers resulted in dominance by diatoms for the whole year and effectively limited
Cells transport
cyanobacteria summer growth. From this analysis, we identified iron concentration and
chlorophyll a and total algal biovolume increased, with larger diatoms preferentially
diatom biovolume as the key water quality indicators to be included in any optimal management, able to control the transfer regime from both a water quantity and water quality prospective. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
In reservoir systems, water is transferred from one reservoir to another, causing hydraulic connections of different basins that are often otherwise disconnected. Any inter-basin water transfer system, driven purely by economic purposes (e.g. water supply and hydropower production), can have complex social and economic impacts (Gupta and van der Zaag, 2008; Karamouz et al., 2010) as well as physical, chemical, hydrological and biological implications both in the sending and receiving system (Davies et al., 1992). Inter-basin water transfers are well studied from a quantity management point
of view (Westphal et al., 2003; Castelletti and Soncini-Sessa, 2007; Moraga et al., 2007) but the increasing pressure on water resources worldwide has encouraged decision-makers to move towards a more holistic management strategy, coupling both quality and quantity issues (Jager and Smith, 2008; Karamouz et al., 2010). Within this framework of Integrated Water Resource Management (Gupta and van der Zaag, 2008), understanding of the effect of transfers on water quality has become essential for sustainable management of reservoir systems (Castelletti and Soncini-Sessa, 2006). The effects of inter-basin water transfers on the donor and receiving ecosystem have been studied both on schemes
* Corresponding author. Tel.: þ61 8 6488 1691; fax: þ61 8 6488 3053. E-mail address:
[email protected] (R. Fornarelli). 1 Present address: Hatch Associates, 144 Stirling St, Perth, Western Australia, Australia. 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.08.048
5874
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 7 3 e5 8 8 5
conveying large volumes of water over large distances and on smaller transfer schemes connecting closer basins. Matthews et al. (1996) and Gibbins et al. (2001) studied the effects of transfer schemes on the fish population of the receiving river, while Snaddon and Davies (1998) studied the implications of a large transfer scheme on the invertebrate community, showing the possibility of transferring alien species and toxic cyanophytes through the pipelines. Several factors contributed to the observed effects of water transfers in these studies: physical and biological differences between the connected systems (Soulsby et al., 1999); the nature of the connection (presence of pipelines, canals, tunnels, Soulsby et al., 1999, and Gibbins et al., 2000); and the magnitude, frequency and duration of transfers (Gibbins et al., 2000). Few studies, however, have concentrated on the effects of water transfers on nutrient concentrations and biological variables such as chlorophyll a and phytoplankton communities. Davies et al. (1992) observed how pumping upstream nutrient-rich water could trigger cyanobacteria blooms in the receiving reservoir. Barbosa et al. (1999) found that water releases from upstream reservoirs caused more diverse and oligotrophic phytoplankton assemblages in downstream reservoirs and Padisak et al. (2000) showed how improvement of downstream reservoirs water quality could be achieved by using the upstream ones as storing agents of nutrient loads. Hu et al. (2008, 2010) and Zhai et al. (2010) showed the positive effect of water transfers from a river to a reservoir in decreasing phytoplankton concentrations but argued that intensive water transfers for long periods could increase nitrogen and phosphorus concentrations and thus accelerate eutrophication. The effects of water transfers on phytoplankton community and organisation can be interpreted by considering it as a disturbance: diverting water from one reservoir to another represents an external disturbance to the receiving reservoir water quality and, as such, it can strongly influence the nutrient concentrations, the natural seasonal succession of phytoplankton and the internal composition of each phytoplankton group (Reynolds, 1993, and Reynolds et al., 1993). Many ecological studies investigated the effects of disturbances on the phytoplankton community (Padisak, 1993; Sommer, 1993, and Lindenschmidt and Chorus, 1998) but this principle has rarely been linked to management practices such as inter-basin water transfers. In this study we analyse how water transfers between two reservoirs affected the water quality of the receiving reservoir. The studied scheme is the Shoalhaven System, Australia, which was built in the 1970s as a water supply and hydropower generation system. Our analysis is supported by a 9year dataset of weekly phytoplankton biovolume and monthly nutrient and metal concentrations that allowed for a long-term and in-depth analysis of the system’s response to water transfers, using methods widely applicable to any kind of reservoir system. The main objectives of this study were therefore: (i) to quantify the effects of water transfers on the reservoir water quality, in term of chemical and biological variables; (ii) to quantify how the reservoir behaviour, in term of internal production/consumption, and the pumping activity contributed to the measured concentrations according to different
water transfer management strategies; (iii) to identify those decision variables and water quality indicators to be used in the multi-objective management problem which couples not only water quantity and pure economic aspects but also water quality and ecological issues.
2.
Materials and methods
2.1.
Study site
The Shoalhaven System is a network of dams, pumps and canals, located 100 km south-west of Sydney (Australia). The system provides water locally, generates hydropower, provides recreation space and also acts as a back-up source for Sydney’s water supply. The system is formed by three major reservoirs (Fig. 1): Lake Yarrunga (volume ¼ 85.5 106 m3, also known as Tallowa Dam), Fitzroy Falls (23.5 106 m3) and Wingecarribee (25.9 106 m3). A small pondage (Bendeela, 1.2 106 m3) is sited between Lake Yarrunga and Fitzroy Falls to aid hydropower generation. Water is pumped 127 m uphill from Lake Yarrunga to Bendeela Pondage and a further 480 m uphill from Bendeela Pondage to Fitzroy Falls during the night. For hydroelectric power generation, water is released back from Fitzroy Falls to Lake Yarrunga during the peak energy demand period. During droughts, water is also pumped upstream from Fitzroy Falls to Wingecarribee, where it exits the Shoalhaven System, and released to Lake Burragorang, the major water supply reservoir of the Sydney region. Fitzroy Falls was selected as the target (receiving) reservoir for this study. The reservoir (Fig. 1) has a storage volume of 23.5 106 m3 and a surface area of 5.2 km2. The full storage level is 663.5 m ASL, and the maximum and mean depths are 9.89 m and 6 m, respectively. It drains a catchment area of 31 km2, mainly used for agricultural and rural development. The pumped inflow from Lake Yarrunga is the biggest inflow: when water transfers are active, the natural catchment inflow is less than 10% of the pumped transfers. All the inflows and outflows are highlighted in Fig. 1. The variability of inflows and outflows is high and it is strongly regulated by the pumping and release system between Fitzroy Falls and Lake Yarrunga. Retention times range from 28 to 40 days during high transfers and exceed 150 days during low transfer rates.
2.2.
Water balance
Flow rate data were available for the following periods: (1) pumped and released daily flow rates between Lake Yarrunga and Fitzroy Falls from December 2003eMarch 2010; (2) natural catchment inflow and dam outflow daily flow rates from January 2000eMarch 2010; (3) pumped daily flow rates from Fitzroy Falls to Wingecarribee from January 2000eMarch 2010; (4) daily water level from January 2000eMarch 2010. A daily water balance was computed to determine the magnitude and timing of the pumped/released flow rates from/to Lake Yarrunga prior to December 2003 from January 2000eMarch 2010. Note that the evaporation volume was not considered in the water balance as it was negligible comparing to the inflows. Weekly averages of daily inflows and daily outflows were used for the pumping activity characterisation. Water retention
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 7 3 e5 8 8 5
5875
Fig. 1 e Shoalhaven system, Lake Yarrunga and Fitzroy Falls Reservoir. Contours at 0, 4, 8 m depth, inflows and outflows for Fitzroy Falls Reservoir (figure not to scale). Pumped Inflow [ from Lake Yarrunga to Fitzroy Falls; Released Outflow [ from Fitzroy Falls to Lake Yarrunga; Pumped Outflow: water supply [ from Fitzroy Falls to Wingecarribee. DFF6 and DTA8 are the water quality monitoring stations for Fitzroy Falls and Lake Yarrunga, respectively. Elevation difference: 610 m; distance between the two reservoirs: 10 km. The presence of Bendeela pondage between Lake Yarrunga and Fitzroy Falls was neglected.
time was calculated as the ratio between the storage volume and the total inflow. The effect of Bendeela pondage was neglected as the retention time was less than 2 days when water transfers were active.
2.3.
Water quality data sampling and analysis
Nutrients (total nitrogen, ammonia, oxidised nitrogen, total and dissolved phosphorus and silica), metals (total iron and total manganese) and chlorophyll a concentrations at 0, 3 and 6 m depths were available monthly at one location in Fitzroy Falls reservoir close to the dam wall (Fig. 1). Water temperature data were available at the same location at depth intervals of 1 m. Depth average values were used as the reservoir is shallow and no persistent stratification was evident. Algal genera and biovolume data were available at weekly intervals from January 2000 onwards from the surface (0e1 m depth) at the same location. Similar data was available at a monitoring station approximately 1 km from the pumping station in Lake Yarrunga (Fig. 1), and were used to represent the inflow water quality data in Fitzroy Falls from water transfer. Meteorological data were available hourly at a station located 15 km west of Fitzroy Falls. The entire available dataset was sampled and analysed by the Sydney Catchment Authority as the organisation in charge of managing the Shoalhaven System. All measurements were conducted using APHA standard protocols (American Public Health Association, 2005).
2.4. Broad-scale overall influences of pumping activity on water quality The effect of water transfers (also referred to as pumping activity) on Fitzroy Falls water quality was evaluated through a direct comparison between water quality measurements made during low (from June 2000 to May 2003) and high pumping periods (from June 2004 to May 2008). Using all data available in the two periods, for each variable, the percentage increase (or decrease) with respect to the mean low pumping concentration (from June 2000 to May 2003) was calculated. To determine the statistical significance of the difference in the mean concentrations, randomized intervention analysis (RIA) was used (Carpenter et al., 1989): the actual difference between mean concentrations during high and low pumping years was ranked along with its probability distribution to produce a p-value. The probability distribution was obtained by random permutations of high and low pumping concentrations time series.
2.5.
Mass balance calculation
Annual mass balances were calculated for total iron, cyanobacteria and diatom biovolume. Outflow concentrations were those measured in Fitzroy Falls at the only monitoring station available close to the dam wall (Fig. 1). We considered that the water quality of Fitzroy Falls was uniform in depth and space as the reservoir is small
5876
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 7 3 e5 8 8 5
and not stratified. Moreover the absence of spatial gradients of iron, cyanobacteria and diatom concentrations was corroborated by a field work conducted by the authors (data not published). Inflow concentrations were taken from the Lake Yarrunga measurements at the closest monitoring station to the pumping station. The water quality at the monitoring station was representative of the water quality extracted at the pumping station as the two sites had the same characteristics (shallow, not stratified and both in the river-like part of the reservoir, Fig. 1). Each measurement was multiplied by the weekly mean of daily flow rate (pumped flow rate from Lake Yarrunga for incoming fluxes and total outflow for outgoing fluxes). One flux per month (one measurement every four weeks) was obtained for iron while four fluxes per month (one measurement every week) were obtained for algae biovolume. Fluxes were averaged over a year and thus, for each year, a mean value of incoming and outgoing flux (Min and Mout) was obtained. The difference between outgoing and incoming flux (Mout Min) was used to quantify the reservoir’s response in terms of internal production, or consumption if negative. The propagated standard error of the internal production/consumption, calculated as a function of two means (outgoing and incoming fluxes), was computed using the error propagation equation as presented in Lehrter and Cebrain (2010). Note that, to compare the results of different years, internal production and consumption were normalised relative to the inflow, so that a value of unity indicates internal production was equal to the pumped flux, and a value of 2 indicates internal production was twice the pumped flux. Based on the austral summer, we considered annual periods from June to May and six years were selected for comparison (Table 1): three are characterised by low pumping (2000e2001, 2001e2002 and 2002e2003) while three are subjected to very high water transfers but with different timing (2004e2005, 2006e2007 and 2007e2008).
3.
Results
3.1.
Water balance and water transfers management
The water balance was validated by comparing the weekly average of measured and calculated pumped flow rate from Lake Yarrunga to Fitzroy Falls from January 2004 to March 2010. A good match between the two time series was found: mean absolute error and root mean square error were 81 and 132, x103 m3 d1 respectively, while coefficient of determination and index of agreement were 0.82 and 0.96, respectively (Willmott, 1982). Therefore, the calculated flow rates were used for the whole period 2000e2010. The results of the water balance allowed characterisation of the pumping system between Lake Yarrunga and Fitzroy Falls from 2000 to 2010 (Fig. 2). The amount of transferred water varied on a daily basis and, from May 2003 until the end of 2008, the system was mostly used as an emergency water supply source for Sydney. Water was pumped from Lake Yarrunga through Fitzroy Falls to reach Wingecarribee and exit the Shoalhaven System towards Lake Burragorang. This caused high fluxes from Lake Yarrunga to the upstream reservoirs. During these years a small amount of water was released back from Fitzroy Falls to Lake Yarrunga for hydropower generation. For all the other years, the connection between Lake Yarrunga and Fitzroy Falls was mainly used for hydropower generation as the amount of water pumped from Lake Yarrunga to Fitzroy Falls was released back within the same week. A number of annual pumping regimes exist in the data record (Table 1). For each year, annual mass balances were developed. Low pumping years were 2000e2001, 2001e2002 and 2002e2003 and from May 2003 much higher volumes were pumped upstream. Based on the timing of pumping, 2004e2005 and 2007e2008 were similar as high volume of water was pumped for almost the whole year. Mostly important, the same amount of water was pumped during spring and summer months. In 2006e2007, very high volumes of
Table 1 e Pumped flow rate: mean ± standard deviation (thousands of m3 dL1) and retention time in brackets (days) for different year periods. Timing of water transfers is highlighted in grey. NP [ no pumping. Low pumping years from 2000 to 2003. High pumping years from 2004 to 2008.
Jun
Jul
Aug
2000 - 2001
96±33 (206)
2001 - 2002
NP
2002 - 2003
NP
2004 -2005
NP
2006 - 2007 2007 - 2008
Sep
Oct
Nov
Dec
Feb
Mar
Apr
154±55 (128)
NP
510±243 (39) NP 539±159 (38)
May NP
134±65 (153) 115±64 (185)
706±214 (29)
Jan
NP NP
298±115 (72)
Annual Mean 103±91 (187) 81±99 (236) 119±146 (174)
387±130 (54)
349±317 (56)
682±169 (30)
546±394 (37) 539±252 (37)
5877
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 7 3 e5 8 8 5
Fig. 2 e Water balance results. Weekly averages of daily flow rate from June 2000 to February 2010. Positive black solid line: pumped inflow to Fitzroy Falls from Lake Yarrunga. Negative grey dotted line: released outflow from Fitzroy Falls to Lake Yarrunga for hydroelectric power generation. Negative black dashed line: pumped outflow from Fitzroy Falls to Wingecarribee for water supply purposes.
water were pumped in winter, spring and fall but a threemonth halt occurred during summer.
3.2. Broad-scale overall influences of pumping activity on water quality Two periods were chosen for direct comparison. The years from June 2000 to May 2003 were referred as the “low pumping years” and, because of the very small transferred volume, pumping activity was expected to have minimal influence on water quality of Fitzroy Falls reservoir during this period. The years from June 2004 to May 2005 and from June 2006 to May 2008 were characterised by high pumped volumes and were referred as the “high pumping years”. Note that the period from June 2005 to May 2006 was neglected because of lack of measurements during winter and spring. Low pumping years
were characterised by a mean pumped flow rate equal to 101 115 103 m3 d1 (196 days retention time) with a maximum pumped flow rate equal to 558 103 m3 d1. High pumping years were characterised by a mean pumped flow rate equal to 480 337 103 m3 d1 (42 days retention time) with a maximum pumped flow rate equal to 1276 103 m3 d1. Note that the effects of water temperature and meteorological variables, i.e. wind speed and direction, were not considered in the explanation of the observed changes in the water quality. Non-significant changes between low and high pumping years occurred for these variables ( p-values from randomized intervention analysis higher than 0.05).
3.2.1.
Nutrients, metals and chlorophyll a
High pumping caused significant increases in silica, total iron and chlorophyll a (Table 2). The large increase in mean silica
Table 2 e Nutrients, metals and algal groups comparison between low and high pumping years: mean ± standard deviation (sample size). Percentage of increase or decrease (L) respect to low pumping concentrations. P-values as results of randomised intervention analysis (significant if p < 0.05, NS [ not significant). Water Quality Variable
Low Pumping Years
High Pumping Years
Percentage
P-value
1
[mg L ] Silica Tot. Iron Chlorophyll a Oxidised Nitrogen Tot. Manganese Tot. Phosphorus Tot. Nitrogen Ammonia
271 270 (41) 183 59 (39) 8 4 (39) 46 42 (38) 25 11 (39) 16 4 (38) 494 123 (38) 44 41 (40)
[mm3 L1] Total Algae Diatom Dinoflagellate Cryptophyte Chrysophyte Chlorophyte Cyanobacteria Euglenophyte
1.41 0.24 0.03 0.06 0.04 0.80 0.20 0.05
0.86 (131) 0.22 (131) 0.10 (131) 0.05 (131) 0.04 (131) 0.68 (131) 0.16 (131) 0.23 (131)
1065 1395 (42) 440 119 (40) 12 5 (42) 60 67 (42) 27 14 (40) 15 6 (42) 386 101 (42) 25 19 (42) 2.98 1.64 0.17 0.12 0.05 0.83 0.14 0.02
1.75 (127) 1.42 (127) 0.36 (127) 0.09 (127) 0.06 (127) 1.23 (127) 0.13 (127) 0.04 (127)
293 140 49 31 10 7 22 44
0 0 0 NS NS NS 0 0
111 571 447 108 34 5 30 46
0 0 0 0 0.0181 NS 0.0002 0
5878
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 7 3 e5 8 8 5
Fig. 3 e Time series of total algae biovolume during low (a) and high (b) pumping years. Percentage contribution of each group relative to the total algae biovolume. ‘Others’ group is the sum of cryptophytes, dinoflagellates, chrysophytes and euglenophytes.
(293%) and iron (140%) concentrations implied a very strong effect of pumping of silica and iron rich water from Lake Yarrunga. Chlorophyll a showed a significant increase of 49% during high pumping years, indicating that, with intensive pumping activity, phytoplankton occurrence was likely to increase. Oxidised nitrogen, total phosphorus and manganese showed no change while significant decreasing trends were observed for total nitrogen and ammonia.
(Fig. 3a) whilst diatoms became the only dominant group during high pumping years (Fig. 3b). Note that the peak of chlorophytes occurring in winter 2004 (Fig. 3b) corresponded to a three-month halt of water transfers (Table 1 and Fig. 2). Note that this analysis accumulates at the group level, and shifts between genera within a group could be possible and will be discussed in the next section.
3.2.2.
3.2.3.
Total algae biovolume
For the characterisation of algal growth during low and high pumping years, each algal group was analysed (Table 2). A significant, large (111%) increase was observed in total algae biovolume during high pumping (Table 2 and Fig. 3). The majority of this large increase was explained by a large significant increase in diatoms and dinoflagellates biovolume (571% and 447%, respectively). Significant increases were also observed for cryptophytes and chrysophytes while a significant decrease was observed for cyanobacteria. Chlorophytes dominated during low pumping years
Diatoms and dinoflagellates
Diatom mean biovolume increased from 0.24 to 1.64 mm3 L1 during high pumping years: they constituted, on average, the 19% and 53% of the total algae biovolume, during low and high pumping years respectively. The most dominant genera were Urosolenia, Cyclotella and Aulacoseira (Table 3). In Fig. 4 a and b, the concentration of each genus, relative to the total diatom biovolume, is presented. A clear shift from Urosolenia and Cyclotella to Aulacoseira occurred during high pumping years. Urosolenia and Cyclotella dominated up to the middle of 2003, while, as soon as water transfers increased, Aulacoseira
Table 3 e Diatom, dinoflagellate and cyanobacteria genera comparison between low and high pumping years: mean ± standard deviation (sample size). Percentage of increase or decrease (L) respect to low pumping concentrations. Pvalues as results of randomised intervention analysis (significant if p < 0.05, NS [ not significant). Functional group classification is specified for each genus (some genera can belong to more than one group according to different species). Genus [mm3 L1]
Low pumping years
High pumping years
Percentage
P-value
Diatom Aulacoseira Urosolenia Cyclotella
0.06 0.09 (130) 0.09 0.15 (130) 0.08 0.10 (130)
1.28 1.05 (127) 0.07 0.09 (127) 0.06 0.08 (127)
1871 27 27
0 NS 0.028
C, P A A, B
Dinoflagellate Peridinium Ceratium Gymnodinium
0.03 0.10 (65) 0.02 0.06 (65) 0.01 0.02 (65)
0.23 0.38 (77) 0.04 0.17 (77) 0.01 0.01 (77)
608 166 70
0 NS 0
LO LM LO, Y
Cyanobacteria Anabaena/Aphanizomenon Potential toxic genera Aphanocapsa/Aphanothece Microcystis
0.01 0.09 0.09 0.09
0.05 0.10 0.05 0.02
400 11 44 77
0 NS 0.0002 0
H1 M, H1 K M, LM
0.01 (146) 0.08 (146) 0.08 (146) 0.23 (146)
0.08 0.12 0.10 0.03
(127) (127) (127) (127)
Functional group
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 7 3 e5 8 8 5
5879
Fig. 4 e Time trend and contribution of each genus relative to the total diatoms (a and b) and cyanobacteria (c and d) biovolume; comparison between low (a and c) and high pumping years (b and d).
became dominant. Cyclotella and Aulacoseira non-random changes were also verified by RIA p-values (Table 3). For dinoflagellates, a significant increase in biovolume was observed during high pumping years: genera characterised by larger but less numerous cells dominated the dinoflagellate pool. From genus data, a significant shift from Gymnodinium to Peridinium was found (Table 3).
3.2.4.
Cyanobacteria
Significant changes were observed within cyanobacteria genera. Aphanocapsa and Microcystis dominated the cyanobacteria group during low pumping years (Fig. 4c). After the starting of high pumping activity, together with these genera, Aphanothece, Anabaena and Aphanizomenon were also present (Fig. 4d). To simplify the discussion, we grouped these genera according to the functional classification proposed by Reynolds et al. (2002). Aphanocapsa and Aphanothece belong to the same functional group (K) and so were summed together, as were Anabaena and Aphanizomenon (H1). Microcystis belongs to group M and all these three functional groups typically occur in shallow, nutrient-rich, small-to-medium size lakes,
which is an adequate qualitative description of Fitzroy Falls reservoir. The high pumping activity caused a significant decrease in Microcystis and Aphanocapsa/Aphanothece cyanobacteria (Table 3). On the other hand, a significant increase in nitrogen-fixing species occurred (Table 3) during high pumping years, possibly in response to the decrease of total nitrogen and ammonia concentrations. Grouping together the potentially toxic species (Microcystis and Anabaena/Aphanizomenon), no significant increase was observed: high water transfers caused an internal shift from Microcystis to Anabaena/Aphanizomenon without increasing the biovolume of the potentially toxic species as a whole.
3.2.5.
Seasonal succession
Cyanobacteria and diatom seasonal succession changed with the pumping activity (Fig. 5). Identifying each genus with its functional group (Reynolds et al., 2002), during low pumping years (Fig. 5a), Cyclotella (group A/B) dominated in spring, Aphanothece and Aphanocapsa (group K) dominated in summer while in late-summer and fall Microcystis (M) and Urosolenia (A) started to grow. This well known pattern of spring diatoms
Fig. 5 e Seasonal succession of diatoms and cyanobacteria during low (a) and high (b) pumping years. Letters correspond to functional group classification: A [ Urosolenia; A, B [ Cyclotella; C, P [ Aulacoseira; K [ Aphanothece/Aphanocapsa; M [ Microcystis; H1 [ Aphanizomenon/Anabaena. Missing data in late winter during high pumping years (b).
5880
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 7 3 e5 8 8 5
and summer cyanobacteria dominance changed during high pumping years. Aulacoseira (C/P) strongly dominated during the whole year (note the different scale of diatoms biovolume axis in Fig. 5a and b) while the cyanobacteria summer peak was replaced by early-summer nitrogen-fixing cyanobacteria (H1). Aphanothece-Aphanocapsa (K) and late-summer Microcystis (M) were still present, as during low pumping years, but at much lower biovolume.
3.3.
Impact of timing and magnitude on water quality
The analysis presented above gives a broad indication of the impact of pumping on water quality in Fitzroy Falls as it considers only the bulk in-reservoir concentrations. We therefore investigated the effects of magnitude and timing of the pumping activity on different water quality variables, and determined how internal production, internal consumption and cells pumping contributed to the measured concentrations. The magnitude effect was analysed by comparison of results of mass balances in low (2000e2001, 2001e2002 and 2002e2003) and high (2004e2005 and 2007e2008) pumping years (Table 1). The timing effect was studied by investigating the set of high pumping years, 2004e2005, 2006e2007 and 2007e2008 as although similar amounts of water were transferred, the timing of the transfers differed substantially (Table 1). We investigated the mass balances of total iron and diatoms as these were the variables mostly affected by the pumping activity with the largest percentage increase (Table 2). Additionally, diatoms became the dominant genus during the pumping activity (Fig. 3b). Cyanobacteria were considered in the following analysis as we found a decreasing trend during high pumping years and a clear effect of the timing of pumping on the prevalence of this group.
3.3.1.
Fig. 6 e Annual mass balance results for diatoms (a), cyanobacteria (b) and total iron (c) during low (00e01, 01e02, 02e03) and high (04e05, 06e07, 07e08) pumping years. Internal production/consumption (±SE) is normalised relative to the inflow.
Effect of magnitude
Pumping low or high volumes of water had a strong impact on the internal production of diatoms, cyanobacteria and the consumption of iron. The internal production of diatoms (Fig. 6a) was between 60% and 90% of the pumped flux during 2004e2005 and 2007e2008, whilst internal consumption occurred during low pumping years. Note that, during high pumping years, the internal production of diatoms was a fraction of the pumped flux, indicating that the main source of diatom biovolume was cells transported from Lake Yarrunga rather than internal growth. The internal production of cyanobacteria (Fig. 6b) always occurred, and similar high internal production (about 4 times higher than the pumped flux) was observed during low pumping years and in 2006e2007 (a more detailed analysis of the reservoir’s response during 2006e2007 will be discussed in Section 3.3.2). This suggests that during low pumping years the source of the observed biovolume was mainly internal production while, during high pumping years (2004e2005 and 2007e2008), either internal production or pumped flux could dominate the in-reservoir concentration. Nutrient loads set an overall control on the internal production of diatoms and cyanobacteria, where growth in 2002e2003 increased relative to 2001e2002 in line with the
increase in nutrient load as pumping increased (particularly ammonia increase, Fig. 7). Importantly, the same internal production of cyanobacteria was observed in 2002e2003 and 2006e2007 despite the largest nutrient load occurring in 2006e2007, indicating that a threshold level of loading was reached in 2002e2003 beyond which other factors began to control overall production. Competitive interaction between diatoms and cyanobacteria was an important factor affecting the internal production of cyanobacteria: during high pumping years (2004e2005 and 2007e2008), when silica loads increased (Fig. 7d), diatoms started to be internally produced and, at the same time, the internal production of cyanobacteria decreased, becoming less than the pumped flux from Lake Yarrunga. The lower internal production of diatoms in 2006e2007, despite the high nutrient loads, is a function of timing of the pumping and will be discussed below. Iron was always internally consumed by the reservoir, but the rate of consumption dropped from 70% of the pumped flux to less than 40% of the pumped flux during high pumping years (Fig. 6c). The effect of pumping activity was therefore to decrease the ability of Fitzroy Falls to consume iron through internal processes. The similar rates of iron consumption
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 7 3 e5 8 8 5
5881
Fig. 7 e Annual average of the daily nutrient load entering Fitzroy Falls from Lake Yarrunga. Dissolved phosphorus, data not available in 00e01.
during high pumping years indicate that the timing of pumping does not exert a strong control on iron consumption.
3.3.2.
Effect of timing
Despite the high and similar annual average pumping rates during 2004e2005, 2006e2007, and 2007e2008, the previous section demonstrated a different response for diatoms and cyanobacteria in 2006e2007 relative to the other high pumping years. The major difference with 2006e2007 relative to the other high pumping years was the absence of pumping during the austral summer months of NovembereJanuary (Table 1). Very high volumes were pumped during winter and spring 2006 (retention time of 29 days), causing high nutrient loads (Fig. 7) and a spring bloom of diatoms (SeptembereOctober 2006, Fig. 8a). The cessation of pumping in November 2006 resulted in a drop in silica concentrations (Fig. 8b) and a substantial increase in cyanobacteria biovolume as the diatom bloom declined to values lower than 1 mm3 L1 (Fig. 8a). On the other hand, when pumping occurred for the whole year (for example in 2007e2008, Fig. 9b), diatom biovolume during summer months (from December to February) was twice the value of summer 2006e2007 whilst the
cyanobacteria summer bloom did not occur (Fig. 9a). Note that the only cyanobacteria peak occurred during 2007e2008 matched exactly with the decrease in pumped flow rate and diatom biovolume in November 2007 (Fig. 9a). It is therefore apparent that the timing of transfers can control competition between diatom and cyanobacteria growth. When high water transfers occurred for the whole year, the internal production of cyanobacteria was limited in favour of diatom production (Figs. 6a, b and 9) whereas ceasing the transfers during summer caused a dominance of cyanobacteria over diatom growth (Figs. 6a, b and 8). The relationship between summer pumping and the cyanobacteria summer bloom, dominated by Aphanothece-Aphanocapsa genera (K group, Reynolds et al., 2002), is clearly shown in Fig. 10. When water transfers occurred intensively in summer 2004e2005 and 2007e2008, no K-species peak was detected (Fig. 10b). On the contrary, the cessation of pumping activity in summer 2006e2007 not only caused a cyanobacteria bloom (Fig. 10b), as in the low pumping years (Fig. 10a), but also resulted in the largest bloom on record, possibly due to the higher nutrient load from the high winter and spring water transfers.
Fig. 8 e Diatoms and cyanobacteria biovolume (a), pumped flow rate and monthly silica concentrations (b) from June 2006 to May 2007.
5882
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 7 3 e5 8 8 5
Fig. 9 e Diatoms and cyanobacteria biovolume (a), pumped flow rate and monthly silica concentrations (b) from June 2007 to June 2008.
4.
Discussion
4.1. Water transfers act as a disturbance for phytoplankton succession The effect of the pumping activity on algal growth in the receiving reservoir was a complex process resulting in an internal change in the growth and organisation of the phytoplankton community. Following Reynolds (1993) and Hoyer et al. (2009) definition and application of disturbance, it is apparent from the data that the pumping activity acted as an external disturbance to the organisation of the algal community of the receiving reservoir and, as such, its severity and frequency directly influenced the successional state of the phytoplankton community and the dominating species. During low pumping years, the seasonal succession of algal groups followed the well known seasonal pattern of spring diatoms replaced by cyanobacteria during summer, following the functional group sequence A/B / K / M,A (Fig. 5). This pattern is known to occur in systems affected by low frequency disturbances (Reynolds, 1993). During high pumping years, the community shifted towards a seasonal succession indicative of more frequent disturbances (8e50 Yr1,
Reynolds, 1993): the diatom Aulacoseira dominated for the whole year and co-existed with cyanobacteria. Aulacoseira is known to be a fast growing R-strategist, able to tolerate disturbance and short duration habitat (Reynolds, 2006), thus more likely to occur during the high pumping activity. Similarly, Hoyer et al. (2009) demonstrated how management actions (in their case hypolimnetic withdrawal) constituted an important external forcing on phytoplankton dynamics, disrupting the seasonal succession governed by a sequence of diatoms, cyanobacteria and dinoflagellate towards a more diverse, rich community with diatoms being favoured. The disturbance regime characterised by intensive summer transfers contributed to reduce the internal production and summer peak of cyanobacteria. On the other hand, stopping the transfers in summer reversed this condition to the dominance of cyanobacteria over diatoms: cyanobacteria still persisted during high disturbance periods, but they started to grow only when the calm phase occurred in summer 2006e2007 (Reynolds, 1993). The strong effects of external disturbances on cyanobacteria summer blooms has already been demonstrated in literature (Ferris and Lehman, 2007; Lehman et al., 2009, and Mitrovic et al., 2011) but has not previously been linked to management practices such as
Fig. 10 e Monthly average of K-species cyanobacteria biovolume from June to May of low (a) and high pumping years (b). The timing of transfers within high pumping years is highlighted.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 7 3 e5 8 8 5
inter-basin water transfers. Following this line, in this paper, we extended the use of the disturbance principle as proposed by Reynolds (1993) from an explanatory concept that supports the understanding of changes in phytoplankton communities to a potential tool supporting more comprehensive water transfer management: we demonstrated the strong link between magnitude and timing of water transfers with phytoplankton organisation and now use this result to design appropriate management practices (Section 4.3).
4.2. Chemical and physical connection between two reservoirs Chemical and biological differences between the two basins and reservoirs contribute to the effects of transfers on water quality (Soulsby et al., 1999, and Gibbins et al., 2000). Transferring iron and silica rich water from one reservoir to the other caused an obvious increase of these variables in the receiving reservoir. On the other hand, dilution (Welch et al., 1992) contributed to lower concentrations of ammonia and total nitrogen. The increase in silica supported significant additional growth of diatoms and a shift from Cyclotella dominance to Aulacoseira. Cyclotella is known to be more tolerant to nutrient deficiency than Aulacoseira and thus is typical of lower trophic habitats (Reynolds et al., 2002; Padisak et al., 2009). It is known to have very low silica requirements (0.08 mgSiO2 L1, Reynolds, 1998) thus it was not silica limited during low pumping years when the mean concentration was 0.27 mgSiO2 L1. On the other hand, Aulacoseira grows in more eutrophic habitats and is sensitive to silica depletion (Reynolds et al., 2002). During the pumping activity, silica concentrations increased to 1 mgSiO2 L1, well-above the Aulacoseira half saturation constant (0.5 mgSiO2 L1, Bormans and Webster, 1999), and thus silica was no longer limiting the presence of this genera. The genera that dominated diatoms (Aulacoseira, group C/P) and dinoflagellates (Ceratium and Peridinium, groups LM and LO) under the high pumping regime are known to generally exist in eutrophic systems (Reynolds et al., 2002). This is clearly indicative of an eutrophication trend imposed on Fitzroy Falls reservoir during high water transfers, the same conclusion drawn from the higher observed nutrient loading and higher chlorophyll a measurements. The physical connection through pipelines and canals between Lake Yarrunga and Fitzroy Falls favoured the pumping of cells from the first to the second and this represented an important contribution to cyanobacteria and diatom biovolume during intensive water transfers (Davies et al., 1992; Snaddon and Davies, 1998). Indeed, mass balances results showed that, when high volumes of water were transferred for the whole year, more than half of the measured cyanobacteria and diatom biovolume was directly transported from the source reservoir. Moreover, the genera that dominated the diatom and dinoflagellate groups during high pumping years (Aulacoseira, Ceratium and Peridinium) were the same genera that dominated in the source waters of Lake Yarrunga and therefore it is likely that they were transported from one reservoir to the other. This is an essential result from the management perspective and stresses the importance of the timing of transfers on
5883
Fitzroy Falls water quality: pumping water from the source reservoir during periods of no or low bloom can significantly decrease the biovolume in the receiving reservoir. Thus future investigations on methods to minimise algal growth in the source reservoir will also have a follow on benefit of decreasing the impact of water transfers on the receiving reservoir.
4.3.
Multi-objective management of water quality
The analysis presented so far evaluated the effects of different management strategies driven by water-quantity related services (water supply and hydroelectric power generation) on the reservoir water quality. Importantly, practical management constraints on the pumping regime are yet to be considered. Therefore, the present study represents the first step towards the implementation of an integrated reservoir system management combining quantity and quality objectives in the same decision framework. The effects of two decisions, magnitude and timing of water transfers, were quantified and, relating these findings with the guidelines ranges, appropriate water quality indicators can be identified. In this specific case, regulatory guideline ranges (ANZECC, 2000) were applied to the reservoir water quality and two appropriate water quality indicators in the receiving reservoir were identified as chlorophyll a and iron concentrations. Note that historical maximum cyanobacteria biovolume was lower than 1 mm3 L1, well below the guideline of 4 mm3 L1. Furthermore, the summer blooms were formed by non harmful K-species (Kankaanpaa et al., 2005; Dasey et al., 2005), thus cyanobacteria biovolume can be considered as a secondary issue in Fitzroy Falls and were not included. However, if decision-makers decide to reduce the internal growth of cyanobacteria to the minimum possible level, pumping high water volumes during summer has been demonstrated herein as being a very effective strategy to achieve this target. Iron concentration was a straightforward function of magnitude of water transfers and thus the timing of water transfers didn’t impact on this variable. In this specific case, to minimise iron concentration transferred volumes should be less than 300 103 m3 d1, keeping the retention time above 80 days. On the other hand, the observed increase in chlorophyll a during high pumping years was mainly due to an increase in diatoms and thus minimizing chlorophyll a concentration is equivalent to minimising diatom biovolume. Two sources can be targeted to reach this objective: minimising cell transport from Lake Yarrunga (which counts for more than half of the measured concentration in Fitzroy Falls, Fig. 6a) or minimizing the internal growth of diatoms in Fitzroy Falls. In the first case both magnitude and timing of pumping becomes fundamental: transferring more water during non-bloom periods in the source water of Lake Yarrunga will obviously prevent high biovolume being transported in Fitzroy Falls. On the contrary, to minimise the internal growth of diatoms in Fitzroy Falls, the timing of pumping should be organised to promote competition and dominance of cyanobacteria over diatoms. It thus appears that, based on the results of this study and the indicators selected, management of water transfers to
5884
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 7 3 e5 8 8 5
maximise water quality involves keeping the retention time above 80 days and, whenever more intensive transfers are required, pumping should take place during winter with minimum pumping during summer in order to decrease chlorophyll a concentrations by keeping diatom biomass to a minimum. Further exploration of the decision space associated with this system and these indicators requires a combination of a simulation model and an appropriate multi-objective decision making process (Castelletti et al., 2010). Moreover, future investigations towards an holistic integration of water quantity and water quality objectives in an optimisation framework require the consideration of additional decision variables, such as the amount of released water for hydropower generation (from Fitzroy Falls to Lake Yarrunga) and its implication on Lake Yarrunga water quality.
5.
Conclusion
Two operational decision variables, the magnitude and timing of water transfers, influenced the reservoir water quality and the following key findings should be considered for integrated management of the reservoir system. High water transfers caused: (i) an increase in silica and iron concentrations; (ii) an increase in diatom biovolume, thus an increase in chlorophyll a concentrations; (iii) shifting of dominant genera within the dinoflagellate and diatom groups, which acted as indicator of a higher trophic state; (iv) no cyanobacteria growth if transfers occurred during summer. The mass balances also indicated the importance of cells pumped from the source reservoir, thus the timing of transfer is also extremely important in controlling phytoplankton biovolume: avoiding pumping during algal bloom periods in the source reservoir could half diatom and cyanobacteria biovolume in the receiving reservoir in this case. The analysis presented here links simple methods applicable to any reservoir with routine monitoring and understanding of water transfers as disturbances to the receiving system. This paper represents a novel contribution to the integration between the management of reservoirs and an indepth analysis and quantification of the effects on water quality. This provides a bridge between ecological knowledge of reservoir algae populations and the common practice of inter-basin water transfers, and shows how targeting a few significant water quality indicators can result in the basis for an integrated multi-objective reservoir system management.
Acknowledgements The first author was the recipient of a Scholarship for International Research Fees from the University of Western Australia and of a University International Stipend from the Centre for Water Research. We gratefully acknowledge the support of the Sydney Catchment Authority for providing access to the data, and Jessica Harris for useful discussion regarding the Shoalhaven System. The findings, opinions and conclusions expressed herein are those of the authors and do not represent the views or opinions of the Sydney Catchment
Authority or any person employed in the Sydney Catchment Authority Division. This article represents Centre for Water Research reference 2397-RF.
references
American Public Health Association, 2005. Standard Methods for the Examination of Water and Wastewater, 21st ed. American Public Health Association (APHA), American Water Works Association (AWWA) & Water Environment Federation (WEF). ANZECC, 2000. National water quality management strategy. Australian and New Zealand Guidelines for Fresh and Marine Water Quality. Barbosa, F.A.R., Padisak, J., Espindola, E.L.G., Borics, G., Rocha, O., 1999. The cascading reservoir continuum concept (CRCC) and its application to the river Tiete-basin, Sao Paulo State, Brazil. In: Tundisi, J.G., Straskraba, M. (Eds.), Theoretical Reservoir Ecology and its Applications. International Institute of Ecology. Brazilian Academy of Sciences and Backhuys Publishers, pp. 425e437. Bormans, M., Webster, I.T., 1999. Modelling the spatial and temporal variability of diatoms in the River Murray. Journal of Plankton Research 21, 581e598. Carpenter, S.R., Frost, T.M., Heisey, D., Kratz, T.K., 1989. Randomized intervention analysis and the interpretation of whole-ecosystem experiments. Ecology 70 (4), 1142e1152. Castelletti, A., Soncini-Sessa, R., 2006. A procedural approach to strengthening integration and participation in water resources planning. Environmental Modelling & Software 21 (10), 1455e1470. Castelletti, A., Soncini-Sessa, R., 2007. Coupling real-time control and socio-economic issues in participatory river basin planning. Environmental Modelling & Software 22, 1114e1128. Castelletti, A., Pianosi, F., Soncini-Sessa, R., Antenucci, J.P., 2010. A multiobjective response surface approach for improved water quality planning in lakes and reservoirs. Water Resources Research 46 (W06502), 16. Dasey, M., Ryan, N., Wilson, J., McGregor, G., Fabbro, L., Neilan, B. A., Burns, B.P., Kankaanpaa, H., Morrison, L.F., Codd, G.A., Rissik, D., Bowling, L., 2005. Investigations into the taxonomy, toxicity and ecology of benthic cyanobacterial accumulations in Myall Lake, Australia. Marine and Freshwater Research 56, 45e55. Davies, B.R., Thoms, M., Meador, M., 1992. An assessment of the ecological impacts of inter-basin water transfers and their threats to river basin integrity and conservation. Aquatic Conservation: Marine and Freshwater Ecosystems 2, 325e349. Ferris, J.A., Lehman, J.T., 2007. Interannual variation in diatoms bloom dynamics: roles of hydrology, nutrient limitation, sinking and whole lake manipulation. Water Research 41, 2551e2562. Gibbins, C.N., Jeffries, M.J., Soulsby, C., 2000. Impacts of an interbasin water transfer: distribution and abundance of Micronecta poweri (Insecta: Corixidae) in the River Wear, north-east England. Aquatic Conservation: Marine and Freshwater Ecosystem 10, 103e115. Gibbins, C.N., Soulsby, C., Jeffries, M.J., Acornley, R., 2001. Developing ecological acceptable river flow regimes: a case study of Kielder reservoir and the Kielder water transfer system. Fisheries Management and Ecology 8, 463e485. Gupta, J., van der Zaag, P., 2008. Interbasin water transfers and integrated water resources management: where engineering, science and politics interlock. Physics and Chemistry of the Earth 33, 28e40. Hoyer, A.B., Moreno-Ostos, E., Vidal, J., Blanco, J.M., PalominoTorres, R.L., Basanta, A., Escot, C., Rueda, F.G., 2009. The
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 7 3 e5 8 8 5
influence of external perturbations on the functional composition of phytoplankton in a Mediterranean reservoir. Hydrobiologia 636, 49e64. Hu, W., Zhai, S., Zhu, Z., Han, H., 2008. Impacts of the Yangtze River water transfer on the restoration of Lake Taihu. Ecological Engineering 34, 30e49. Hu, L., Hu, W., Zhai, S., Wu, H., 2010. Effects on water quality following water transfers in Lake Taihu, China. Ecological Engineering 36, 471e481. Jager, H.I., Smith, B.T., 2008. Sustainable reservoir operation: can we generate hydropower and preserve ecosystem values? River Research and Application 24, 340e352. Kankaanpaa, H.T., Holliday, J., Schroder, H., Goddard, T.J., Von Fister, R., Carmichael, W.W., 2005. Cyanobacteria and prawn farming in northern New South Wales, Australia e a case study on cyanobacteria diversity and hepatotoxin bioaccumulation. Toxicology and Applied Pharmacology 203, 243e256. Karamouz, M., Mojahedi, S.A., Ahmadi, A., 2010. Interbasin water transfer: economic water quality-based model. Journal of Irrigation and Drainage Engineering 136, 90e98. Lehman, E.M., McDonald, K.E., Lehman, J.T., 2009. Whole lake selective withdrawal experiment to control harmful cyanobacteria in an urban impoundment. Water Research 43, 1187e1198. Lehrter, J.C., Cebrain, J., 2010. Uncertainty propagation in an ecosystem nutrient budget. Ecological Applications 20 (2), 508e524. Lindenschmidt, K.E., Chorus, I., 1998. The effect of water column mixing on phytoplankton succession, diversity and similarity. Journal of Plankton Research 20, 1927e1951. Matthews, W.J., Schorr, M.S., Meador, M.R., 1996. Effects of experimentally enhanced flows on fishes of a small Texas (U.S.A.) stream: assessing the impact of interbasin transfer. Freshwater Biology 35, 349e362. Mitrovic, S.M., Hardwick, L., Dorani, F., 2011. Use of flow management to mitigate cyanobacterial blooms in the Lower Darling River, Australia. Journal of Plankton Research 33 (2), 229e241. Moraga, R., Garcia-Gonzalez, J., Parrilla, E., Nogales, S., 2007. Modeling a nonlinear water transfer between two reservoirs in a midterm hydroelectric scheduling tool. Water Resources Research 43 (4). Padisak, J., 1993. The influence of different disturbance frequencies on the species richness, diversity and equitability of phytoplankton in shallow lakes. Hydrobiologia 249, 135e156. Padisak, J., Barbosa, F.A.R., Borbely, G., Borics, G., Chorus, I., Espindola, E.L.G., Heinze, R., Rocha, O., Torokne, A.K.,
5885
Vasas, G., 2000. Phytoplankton composition, biodiversity and a pilot survey of toxic cyanoprokaryotes in a large cascading reservoir system (Tiete basin, Brazil). International Association of Theoretical and Applied Limnology 27, 2734e2742. Padisak, J., Crossetti, L.O., Naselli-Flores, L., 2009. Use and misuse in the application of the phytoplankton functional classification: a critical review with updates. Hydrobiologia 621, 1e19. Reynolds, C.S., 1993. Scales of disturbance and their role in plankton ecology. Hydrobiologia 249, 157e171. Reynolds, C.S., Padisak, J., Sommer, U., 1993. Intermediate disturbance in the ecology of phytoplankton and the maintenance of species diversity: a synthesis. Hydrobiologia 249, 183e188. Reynolds, C.S., 1998. What factors influence the species composition of phytoplankton in lakes of different trophic status? Hydrobiologia 369/370, 11e26. Reynolds, C.S., Huszar, V., Kruk, C., Naselli-Flores, L., Melo, S., 2002. Towards a functional classification of the freshwater phytoplankton. Journal of Plankton Research 24, 417e428. Reynolds, C.S., 2006. The Ecology of Phytoplankton. Cambridge University Press. Snaddon, C.D., Davies, B.R., 1998. A preliminary assessment of the effects of a small South African inter-basin water transfer on discharge and invertebrate community structure. Regulated Rivers: Research and Management 14, 421e441. Sommer, U., 1993. Disturbance-diversity relationships in two lakes of similar nutrient chemistry but contrasting disturbance regimes. Hydrobiologia 249, 59e65. Soulsby, C., Gibbins, C.N., Robins, T., 1999. Inter-basin water transfers and drought management in the Kielder/Derwent system. Journal of the Chartered Institution of Water and Environmental Management 13, 213e223. Welch, E.B., Barbiero, R.P., Bouchard, D., Jones, C.A., 1992. Lake trophic state change and constant algal composition following dilution and diversion. Ecological Engineering 1, 173e197. Westphal, K.S., Vogel, R.M., Kirshen, P., Chapra, S.C., 2003. Decision support system for adaptive water supply management. Journal of Water Resources Planning and Management 129, 165e177. Willmott, C.J., 1982. Some comments on the evaluation of model performance. Bulletin of American Meteorological Society 63, 1309e1313. Zhai, S., Hu, W., Zhu, Z., 2010. Ecological impacts of water transfers on Lake Taihu from the Yangtze River, China. Ecological Engineering 36, 406e420.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 8 6 e5 8 9 4
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Sonolytic degradation of dimethoate: Kinetics, mechanisms and toxic intermediates controlling Juan-Juan Yao a,b,c,*, Michael R. Hoffmann b, Nai-Yun Gao c, Zhi Zhang a, Lei Li c,d a
Key Laboratory of the Three Gorges Reservoir Regions Eco-Environment, Ministry of Education, Chongqing University, Chongqing 400045, China b W.M. Keck Laboratories, California Institute of Technology, Pasadena, CA 91125, USA c State Key Laboratory of Pollution Control and Resources Reuse, Tongji University, Shanghai 200092, China d Department of Civil and Environmental Engineering, University of California at Berkeley, Berkeley, CA 94720, USA
article info
abstract
Article history:
The sonolytic degradation of aqueous solutions of dimethoate, O,O-dimethyl S-[2-(meth-
Received 4 January 2011
ylamino)-2-oxoethyl]dithiophosphate, was examined. Optimal degradation rates were
Received in revised form
obtained at 619 kHz for continuous sonolysis and 406 kHz for pulse sonolysis. The primary
8 May 2011
pathways for degradation include hydroxyl radical oxidation, hydrolysis and pyrolysis on
Accepted 24 August 2011
collapsing cavitation bubble interfaces. Reaction mechanisms coupled with the corre-
Available online 5 September 2011
sponding kinetic models are proposed to reproduce the observed concentration versus time profiles for dimethoate, omethoate and N-(methyl) mercaptoacetamide during
Keywords:
sonolysis. The oxidation and hydrolysis of dimethoate and omethoate occurred at the
Dimethoate
water-bubble interface was the rate-determining step for sonolytic overall degradation of
Sonolytic degradation
dimethoate. More than 90% toxicity of dimethoate was reduced within 45 min ultrasonic
Kinetic model
irradiation. Ferrous ion at micro molar level can significantly enhance the sonolytic
Pathways
degradation of dimethoate and effectively reduce the yields of toxic intermediate
Molecular orbital theory
omethoate. ª 2011 Elsevier Ltd. All rights reserved.
Toxicity
1.
Introduction
Organophosphates pesticides (OPs) function by inhibiting the enzyme acetylcholinesterase (AChE) (Uchida and Obrien, 1967; Tarbah et al., 2006). Limiting the use of hypertoxic organophosphate pesticides has become a widespread goal (Ragnarsdottir, 2000). However, as one of the representative moderate thio-OPs, dimethoate is legally used for agriculture and, as a consequence, it is frequently detected in surface waters and ground waters at concentration levels ranging from ng L1 to mg L1 in China (Zhang et al., 2002; Gao et al., 2009). In addition to the inherent toxicity of dimethoate,
there is a growing public health concern over its O-analog, omethoate, which is known to have a higher level of toxicity due to its higher AChE inhibitory effect (Buratti and Testai, 2007). Omethoate is formed during the oxidation of dimethoate (Evgenidou et al., 2006, 2007; Echavia et al., 2009). Moreover, O-analogs are frequently reported to be resistant or accumulate in chlorination processes of thio-Ops (Acero et al., 2008; Duirk et al., 2009, 2010). Given these concerns, we are motivated to explore the kinetics and mechanism of dimethoate degradation in a variety of engineered systems with a particular focus paid on the formation of the intermediate omethoate.
* Corresponding author. Key Laboratory of the Three Gorges Reservoir Regions Eco-Environment, Ministry of Education, Chongqing University, Chongqing 400045, China. Tel./fax: þ86 23 65120811. E-mail address:
[email protected] (J.-J. Yao). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.08.042
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 8 6 e5 8 9 4
Acoustic cavitation as driven by ultrasonic irradiation has been utilized in the past to eliminate ester-derived pesticides, such as carbofuran (Dust and Warren, 2001; Hua and PfalzerThompson, 2001), parathion (Kotronarou et al., 1992; Wang et al., 2006a, b; Yao et al., 2010), methy-parathion (Wang et al., 2006a, b, 2007), omethoate (Farooq et al., 2003), dimethoate (Liu et al., 2008a,b), diazinon (Matouq et al., 2008), p-nitrophenyl acetate (Hua et al., 1995) and dichlorvos (Schramm and Hua, 2001). Ultrasound irradiation induces the formation of cavitation bubbles in water that undergo transient collapse events. Quasi-adiabatic compression during transient collapse yields temperatures approaching 5000 K and pressures up to 10,000 bar (Hoffmann et al., 1996) for periods of time in the range of microseconds for each individual bubble event. In water during ultrasonic irradiation, three different reaction zones have been identified. They are (i) the vapor phase of collapsing cavitation bubbles where water is pyrolytically cleaved to form ( OH) and hydrogen atoms ( H), as shown in Eq. (1); in addition, volatile substrates can also undergo pyrolytic degradation in the hot spots of the collapsing bubbles;
ÞÞÞ
H2 O / H$ þ $ OH
(1)
(ii) the interfaces of the collapsing cavitation bubbles where there temperatures have been estimated to range from 600 to 1000 K in addition, a thin shell of transiently supercritical water may also exist in this region; and finally (iii) reactions also take place in the bulk aqueous solution where reactions with OH or H may occur (Adewuyi, 2005a,b). However, little is known about the mechanism and kinetics of dimethoate degradation during ultrasonic irradiation. In earlier studies, Liu et al. (2008a,b) investigated a combination of ozonolysis and sonocatalytic degradation of dimethoate at low ultrasonic frequencies. However they did not carry out a detailed kinetic study of the formation of reaction by-products nor did they propose a detailed reaction mechanism. In light of the environmental problems associated with the formation of omethoate, we are motivated to explore the fundamental reaction kinetics and reaction pathways leading to the degradation of dimethoate and formation of its principal intermediates under ultrasonic irradiation. Our experimental goals included the determination of effect of ultrasonic frequency for degradation at different modes and reaction pathways taking place during dimethoate sonolysis. A reaction mechanism and a corresponding kinetic model for dimethoate degradation with the formation the major reaction intermediates are presented. Finally, a simple but effective method for toxic intermediate omethoate controlling is proposed.
2.
Experimental details
2.1.
Chemicals
All the chemical reagents were analytical grade and used without further treatment. Dimethoate (98%, purity) and omethoate (97%, purity) were purchased from Ehrenstorfer
5887
GmbH (Germany). N-(methyl)mercaptoacetamide (97%, purity) and dichloromethane (PESTANAL), HCl, NaOH, methanol (PESTANAL) and acetic acid were obtained from SigmaeAldrich (USA). The luminescent bacterium Q67 was supplied by East China Normal University. All the solutions were prepared with water purified with a Milli-Q Gradient water purification system (18.2 MU).
2.2.
Experiments
Ultrasonic transducers at the fixed frequencies of 58, 226, 406, 619 and 800 kHz were used to evaluate the role of applied frequency (SF2460A, SF200, SF400, SF600, SF800, Shanghai Acoustics Laboratory, Institute of Acoustics, Chinese Academy of Sciences, China). A 600 mL cylindrical stainless steel reaction vessel was directly connected to the chosen ultrasonic transducer with flanges and a flexible Teflon O-ring for sealing. All reactions were performed under atmospheric pressure at a constant temperature of 20.0 1.0 C. In each kinetic run, 300 mL of a previously prepared dimethoate solution was introduced into the reaction vessel. Reactions were initiated by turning on the ultrasonic generator. The initial pH of the dimethoate solutions was adjusted to 7.00 0.05 with 1.0 M HCl and/or 1.0 M NaOH; however, the reaction solutions were not buffered and thus the pH was allowed to vary during the course of ultrasonic irradiation. The initial concentrations of dimethoate employed in Sections 3.1e3.5 were 2.1, 6.3, 6.3, 15.0, 6.3 mM, respectively. At pre-set time intervals, sample aliquots were collected for analysis. All the experiments were carried out in triplicate with standard deviations of less than 10%. Each data point represents an average value.
2.3.
Analysis
Dimethoate degradation intermediates were identified by solid phase extraction-gas chromatography/electron impact mass spectrometry (SPE-GC/EI-MS) (Yao et al., 2010). Identification of the reaction intermediates and by-products was confirmed by comparing retention times as well as mass spectra of standard samples and the interpretation of mass spectra of unknowns through NIST 02 mass spectral library searches, standard solutions or previously reported spectra (Evgenidou et al., 2006). Simultaneous quantification of dimethoate and omethoate was achieved via Agilent 1100 series LCeESI-MS equipped with a Bonus-RP column (250 mm 4.6 mm i.d., 3 mM particle sizes). Isocratic elution was performed with a mobile phase composed of methanol/ water/acetic acid (50:49.5:0.5, v/v/v) at a fixed flow rate of 0.6 mL min1. MS was operated in the positive ion mode for dimethoate (m/z ¼ 230), omethoate molecular ion (m/z ¼ 214) and N-(methyl)mercaptoacetamide molecular ion (m/z ¼ 106). The nebulizer gas pressure was 50 psi. The drying gas flow rate and temperature were 10 L min1 and 300 C, respectively. The capillary and fragmentor voltage were set at þ2500 V and þ60 V, respectively. The solution toxicities were determined using the Microtox procedure (Zhang et al., 2008, 2009). Calorimetry measurements were made to determine the acoustic power transferred to solution (Kimura et al., 1996).
5888
3.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 8 6 e5 8 9 4
ultrasonic frequency from 58 kHz to 619 kHz. The first-order constant decreased slightly to 0.083 min1 at 800 kHz. Therefore, 619 kHz was identified as the optimal frequency for dimethoate sonolysis under the conditions employed in subsequent experiments. Sonochemical kinetic frequency dependences are mainly due to effects on the mass transfer and cavitation dynamics such as the number of bubble events taking place per unit time. In this case, higher frequencies produce smaller cavitation bubbles with higher surface areato-volume ratios (Hua et al., 1997) and the smaller cavitation bubbles oscillate more frequently leading to greater effects of rectified diffusion (Campbell et al., 2009). As a consequence, more dimethoate is transported to the vaporeliquid interfacial region by rectified diffusion, where it undergoes decomposition. Moreover, the decreasing sizes of bubbles with the increasing frequency shortens the characteristic time for bubble collapse such that more bubble events take place per unit time resulting in faster overall decomposition kinetics. Petrier et al. (1994) suggested that, at higher frequencies, hydroxyl radicals are ejected out of the bubble before they recombined in the gas phase because the collapse times at higher frequencies were shortened. However, at much higher frequencies, the number of cavitational events per unit time is reduced either because the rarefaction cycle of the sound
Results and discussion
3.1. Effect of ultrasonic frequency, ultrasonic intensity and ultrasonic mode on dimethoate degradation The effect of ultrasonic frequency and intensity on dimethoate is shown in Fig. 1(a) and (b). In Fig. 1(a), the ultrasonic intensity as determined by calorimetry was measured as 0.57 W cm2 and held constant as frequency was varied. In Fig. 1(b), the frequency was held constant at 619 kHz while the ultrasonic intensity was varied. As illustrated in Fig. 1(a) and (b), the degradation of dimethoate followed pseudo 1st-order kinetics at each ultrasonic frequency (i.e., from 58 to 800 kHz) and each applied intensity over the range of 0.15e0.69 W cm2. Concentration vs. time profiles are therefore analyzed with a simple first-order equation as follows: ln
½dimethoatet ¼ kapp t ½dimethoate0
(2)
where, kapp, is the apparent first-order rate constant, t is the irradiation time, and [dimethoate]0 and [dimethoate]t are the concentrations at time ¼ 0 and time ¼ t, respectively. kapp increased from 0.031 min1 to 0.112 min1 with an increase in
b 0
0
-1
-1
Ln(ct/c0)
Ln(ct/c0)
a
-2 58 kHz 226 kHz 406 kHz 619 kHz 800 kHz
-3
-4
-2 2
0.69 W/cm 2 0.57 W/cm 2 0.48 W/cm 2 0.24 W/cm 2 0.15 W/cm
-3
-4 0
10
20
30
40
50
0
10
20
Time (min)
c
0.12
30
40
50
60
time (min) continuous mode pulse mode
0.10
-1
kapp(min )
0.08 0.06 0.04 0.02 0.00 100
200
300
400
500
600
700
800
900
Frequency (kHz)
Fig. 1 e The relative effects of ultrasonic frequency, ultrasonic intensity and mode on the sonolytic degradation of dimethoate where [dimethoate]0 [ 2.1 mM. (a) Applied ultrasonic intensity [ 0.57 W cmL2, continuous mode. (b) Ultrasonic frequency [ 619 kHz, continuous mode. (c) Applied ultrasonic intensity [ 0.57 W cmL2, pulse mode (pulse length [ 100 ms; pulse interval [ 100 ms). Note: the error bars are not added in panels (a) and (b) due to the limited space. The experimental errors for the data points in panels (a) and (b) are below 10%.
5889
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 8 6 e5 8 9 4
wave produces a negative pressure which is insufficient in its duration and/or intensity to initiate cavitation or the compression cycle occurs faster than the collapse times of these bubbles (Hung et al., 1998). Previous research found that the OH radical yield at different frequencies followed the order 354 > 620 > 803 > 206 > 1062 kHz (Yang et al., 2008). Therefore, the total number of cavitational bubble events per unit time and the mass transfer of the dimethoate to the bubble surface are two frequency-dependent factors that act concomitantly on dimethoate degradation kinetics. The effect of ultrasonic intensity on dimethoate degradation with time is shown in Fig. 1b. kapp proportionally increases from 0.027 min1 to 0.130 min1 with an increase in ultrasonic intensity from 0.15 to 0.69 W/cm2 (kapp ¼ 0.188 [ultrasonic intensity], R2 ¼ 0.990). The positive effect was primarily due to the enhanced cavitational activity at higher ultrasonic intensity that increased the number and violence of bubble collapse. In addition, the mixing intensity was also enhanced with the increasing ultrasonic intensity due to the acoustic streaming generated by the cavitational processes. The effect of ultrasonic mode (i.e. continuous mode and pulse mode) on dimethoate degradation with time at different frequencies is shown in Fig. 1(c). We set the ultrasonic pulse length at 100 ms and pulse interval at 100 ms for pulse mode because w104 acoustic cycles are required for bubbles become active (Henglein 1993). It is clear that the optimal frequency for dimethoate degradation changed to 406 kHz under pulse mode, compared that under continuous mode at the same ultrasonic intensity. Pulsing had a negative effect on sonolytic degradation of dimethoate over the discussed ultrasonic frequency range, especially at the higher frequency. It is indicated that the pulsing significantly affects the cavitational process. Under pulse mode, the lifetimes of the bubble-water interfaces increase and therefore it can provide more absorption time for dimethoate toward the bubbles. However, bubbles present in solution during the pulse interval are inherently unstable and either tend to dissolve away due to high Laplace pressure in the bubble or alternatively coalesce and become too large to undergo collapse before the next pulsing cycle coming. Both the two phenomena can considerably attenuate the cavitational process. In addition, at
higher frequency, the bubbles are more unstable due to higher Laplace pressure caused by smaller size (Yang et al., 2005). Therefore, the pulsing cannot enhance the sonolytic degradation of dimethoate.
3.2. Identification of reaction intermediates and reaction mechanism Twelve reaction intermediates were identified using GCeMS detection (as shown in Table 1). A possible set of reaction pathways consistent with the observed intermediates is graphically illustrated in Fig. 2. Dimethoate can be degraded through four parallel reaction pathways as follows: Pathway I proceeds via OH attack on the P]S bond, which results in the formation of omethoate (compound 11) and with the production of cyclic hexaatomic sulfur (compound 10) or acyclic octaatomic sulfur (compound 13) (Echavia et al., 2009); Pathway II in which OH attack the PeS bond leads to the formation of a N-(methyl) mercaptoacetamide radical and O,O-dimethyl phosphorothioate; Pathway III the hydrolytic cleavage of thioester bond leading to the formation of N-(methyl) mercaptoacetamide (compound 4) and O,O-dimethyl phosphorothioate; and Pathway IV which proceeds via the pyrolytic decomposition of dimethoate and daughter products either at the bubble interfaces or within the vapor phase to form. N-(methyl) mercaptoacetamide radicals react with either H, CH3 and SCH3 to yield N-(methyl) mercaptoacetamide (compound 4), 2-S-methyl-(N-methyl) acetamide (compound 6) and 1-methyl-2-(acetyl-N-methyl)methane disulfide (compound 9), respectively. O,O,O-trimethyl thiophosphate (compound 3) and O,O,S-trimethyl phosphorothioate (compound 5) are likely formed through secondary methylation of O,O-dimethyl phosphorothioate. O,O,S-trimethyl phosphorodithioate (compound 7) and O,S,S-trimethyl phosphorodithioate (compound 8) are formed by the attack of SCH3 on O,O-dimethyl phosphorothioate. The CH3 and SCH3 could derive from decomposition of some small molecular organic acids or intermediates containing methyl or S-methyl functional groups under ultrasonic irradiation. These by-products were also observed during the photocatalytic degradation of dimethoate (Evgenidou et al., 2006).
Table 1 e GCeMS-EI retention times and characteristic m/z numbers of primary reaction intermediates. No. 1 2 3 4 5 6 7 8 9 10 11 12 13
Product identity
Retention time (min)
Similarity (%)
Primary fragments (m/z ratio)
Dimethyl phosphite O,O,O-trimethyl phosphate O,O,O-trimethyl thiophosphate N-(methyl) mercaptoacetamide O,O,S-trimethyl phosphorothioate 2-S-methyl-(N-methyl) acetamide O,O,S-trimethyl phosphorodithioate O,S,S-Trimethyl phosphorodithioate 1-Methyl-2-(acetyl-N-methyl-)methane disulfide Cyclic hexaatomic sulfur Omethoate Dimethoate Cyclic octaatomic sulfur
4.333 6.583 7.417 9.275 9.783 10.108 11.125 13.217 14.550 16.858 18.142 20.150 25.858
97 95 90 80 93 80 94 86 85 82 100 100 93
109,95,80,77,65,51,47,45 140,110.109,95,79,65,47 156,126,93,79,63,47 105,78,72,58,48 156,141,126,110,95,93,79,62,47,45 119,73,62,58,47 172,157,141,125,109,93,79,77,63,47 172,157,141,125,94,79,62,47, 45 151,105,93,79,73,58,45,42 206,174,160,142,128,110,96,78,64,45 156,141,126,110,95, 79,58,47,44 229,171,157,143,125,109,104,93,87,63,58,47 256,224,192,160,128,96,64
5890
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 8 6 e5 8 9 4
Fig. 2 e Proposed reaction mechanism for dimethoate degradation during ultrasonic irradiation.
O,O,O-trimethyl phosphate (compound 2) most likely results from the oxidation of O,O,O-trimethyl thiophosphate (compound 3) or via the methylation of O,O-dimethyl phosphonic ester. If hydrogen atom is inefficiently scavenged by dissolved oxygen, O,O-dimethyl phosphonic ester may be reduced to diethyl phosphite (compound 1)(Hoffmann et al., 1996). The attack of CH3 on N-(methyl)mercaptoacetamide (compound 4) most likely leads to the formation of 2-S-methyl(N-methyl)acetamide (compound 6), whereas reaction with SCH3 may lead to 1-methyl-2-(acetyl-N-methyl-)methane disulfide (compound 9). In order to provide additional support for our hypothesis that the primary sonolytic degradation pathway of dimethoate involves an initial cleavage of P]S and the PeS bonds, MO calculations were carried out applied to dimethoate. Structures were optimized with the 6-31G (d) basis-set at the level of the RHF, and then the frontier electron densities (FEDs) of the highest occupied molecular orbital (HOMO) and the lowest unoccupied molecular orbital (LUMO) were calculated. The evaluated FEDs are listed in Table 2, while the atom serial number of the dimethoate molecule is shown in Fig. 2. On the basis of FED2HOMO ; FED2LUMO ; and FED2HOMO þ FED2LUMO values, the HOMO is located in the positions of S1 and S5, while the LUMO is in P2 and S1. The S1, S5 and P2 locations with higher ðFED2HOMO þ FED2LUMO Þ values are the most likely attack locations on the dimethoate molecule for OH (Fleming, 1976; Ohko et al., 2002). The attack of OH which lead to the cleavage of P]S and PeS, especially the PeS with lower bond energy. P2 position with highest positive charge location is most likely to be attacked by nucleophile chemicals, such as water or hydroxide ion at the water-bubble
interface. The electron transfer from P2 to S1 leads to the PeS bond cleavage.
3.3. A kinetic model for the sonolytic degradation of dimethoate, the formation of omethoate and N-(methyl) mercaptoacetamide The sonolytic degradation of dimethoate proceeds via 4 parallel pseudo first-order pathways. The overall dimethoate degradation can be expressed as follows:
Table 2 e Frontier electron densities (FEDs) and point charge on heavy atoms of dimethoate. No.
Atom
FED2HOMO
FED2LUMO
FED2HOMO þFED2LUMO
1 2 3 4 5 6 7 8 9 10 11 12
S P O O S C C C C N O C
0.182 0.029 0.018 0.002 0.543 0.002 0.002 0.000 0.002 0.001 0.002 0.000
1.031 1.770 0.079 0.082 0.589 0.614 0.023 0.029 0.033 0.020 0.014 0.010
1.213 1.799 0.098 0.084 1.132 0.616 0.025 0.029 0.035 0.021 0.016 0.011
Point charge
0.040 1.144 0.668 0.705 0.507 0.108 0.405 0.410 0.756 0.362 0.605 0.280
5891
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 8 6 e5 8 9 4
dimethoate, omethoate and N-(methyl) mercaptoacetamide as a function of time:
Table 3 e Non-linear least-square optimization of key rate constants and reaction pathway fractions. Constant
Value (s1)
Fractional coefficient
Value
0.115 0.0168 0.3278
h1 h2 h3
0.2423 0.3878 0.6901
k0 k1 k2
d½dimethoate ¼ ðkI þ kII þ kIII þ kIV Þ½dimethoate dt ¼ k0 ½dimethoate
(3)
where, kI, kII, kIII and kIV are the pseudo first-order rate constants for each pathway of dimethoate degradation in Fig. 2, and k0 is the overall apparent first-order rate constant. In addition, the formation of omethoate and N-(methyl) mercaptoacetamide can be expressed as follows: d½omethoate ¼ h1 k0 ½dimethoate ðkII þ kIII dt þ kIIII Þ½omethoate ¼ h1 k0 ½dimethoate k1 ½omethoate
(4)
where, kI-I, kI-II and kI-III are the pseudo first-order rate constants for each pathway for omethoate degradation as shown in Table 3, and k1 is the overall first-order rate constant for sonolytic degradation of omethoate. d½N ðmethylÞmercaptoacetamide dt ¼ h2 k0 ½dimethoate þ h3 k1 ½omethoate (5)
where, h1 and h2 are the fractional coefficients (h1 ¼ kI/k0, h2 ¼ (kII þ kIII)/k0) of dimethoate degradation via omethoate and N-(methyl)mercaptoacetamide, respectively. h3 is the fractional coefficient (h3 ¼ (kI-I þ kI-II)/k1) of omethoate degraded via yielding N-(methyl)mercaptoacetamide. k2 is the overall rate pseudo first-order rate constant for the sonolytic degradation of N-(methyl)mercaptoacetamide. Integration and rearrangement of Eqs. (3)e(5) give Eqs. (6)e(8), which can be used to predict the concentrations of
Dimethoate (micro mole)
6 5 4
R2=0.9976
2 1 0 20
h1 k0 ½dimethoate0 eðk1 k0 Þt 1 k t 1 k1 k0 e
40
Time (min)
60
80
(7)
½N ðmethylÞmercaptoacetamidet k0 ½dimethoate0 h2 h3 k1 þ h1 h3 k1 h2 k0 eðk2 k0 Þt 1 ¼ k t 2 h3 k1 k0 k2 k0 e ðk2 h k1 Þt 3 h1 h3 k1 e 1 k2 h3 k1 ek2 t (8) The rate constants k0, k1, k2 and the fraction coefficients h1, h2, h3 were determined by the least-square method using Matlab 2010a (MathWorks, Inc.), as shown in Fig. 3, which gives a comparison of the observed and modeled concentrations of dimethoate, omethoate, and N-(methyl) mercaptoacetamide versus time during dimethoate sonolysis. 99.90% of 6.3 mM dimethoate is decomposed within 80 min. In addition, 24.23% of the dimethoate is degraded into omethoate, and 38.78% is transformed into N-(methyl) mercaptoacetamide. Furthermore, 69.01% of the omethoate produced during sonolysis is degraded into N-(methyl) mercaptoacetamide. In the case of the photolytic or photocatalytic degradation of dimethoate, formation of omethoate appeared to be the primary pathway (Evgenidou et al., 2006; Echavia et al., 2009). However, with sonolytic degradation of dimethoate, both the OH oxidation and the hydrolytic pathways lead to the formation of N-(methyl) mercaptoacetamide. The subsequent degradation of N-(methyl) mercaptoacetamide is 2.9 and 19.5 times higher than that observed for dimethoate and omethoate, respectively, because N-(methyl) mercaptoacetamide, a volatile intermediate, can partition into the vapor phase of the cavitation bubbles leading to its pyrolytic degradation. Therefore, the oxidation and hydrolysis of dimethoate and omethoate occurred at the water-bubble interface was the rate-determining step for sonolytic overall degradation of dimethoate.
Omethoate & N-(methyl) mercaptoacetamide (micro mole)
7
0
½omethoatet ¼
(6)
k2 ½N ðmethylÞmercaptoacetamide
3
½dimethoatet ¼ ½dimethoate0 ek0 t
1.4 omethoate
1.2 N-(methyl) mercaptoacetamide
1 0.8
R2=0.9834
0.6 0.4 R2=0.9523
0.2 0 0
20
40
60
80
100
120
Time (min)
Fig. 3 e Mathematic model fits of the sonolytic degradation of dimethoate degradation and the formation of N-(methyl) omethoate and N-(methyl)mercaptoacetamide at 619 kHz with an applied ultrasonic intensity of 0.69 W cmL2.
5892
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 8 6 e5 8 9 4
40
Table 4 e The effect of different Fe2D dosages on the sonolytical degradation of dimethoate (ultrasonic frequency: 619 kHz; ultrasonic intensity: 0.69 W/cm2; initial pH [ 7.0).
35 30 25
Inhibition (%)
20 15 10 5
Fe2þ/mM
kapp/min1
R2
0 50 100 250 500
0.115 0.010 0.177 0.016 0.205 0.014 0.197 0.012 0.189 0.013
0.992 0.998 0.993 0.993 0.990
0
3.5.
-5
Toxic intermediate product controlling
-10 0
20
40
60
80
100
120
Time (min)
Fig. 4 e Relative toxicity as a function of irradiation time for an initial dimethoate concentration of 15 mM at a frequency of 619 kHz with an applied ultrasonic intensity of 0.69 W cmL2, continuous mode.
3.4.
Toxicity
In Fig. 4, we present the observed variation in the percentage inhibition of bioluminescence versus time. The relative initial toxicity was 19% and increased to 22% after 10 min of reaction. This increase can be attributed to the formation of the omethoate. Subsequently, the toxicity is decreased to 1.4% after 45 min of reaction time. This trend appears to correlate with the degradation profile of dimethoate. After 100 min, a quasi steady-state value of 3% is attained. The slight secondary increase in toxicity may be due to the accumulation of H2O2 and an increase in acidity during the course of the sonolytic reaction (Inoue et al., 2006; Pe´rier et al., 2007).
2+
Fe =0 μΜ 2+ Fe =50 μΜ 2+ Fe =100 μΜ 2+ Fe =250 μΜ 2+ Fe =500 μΜ
0.20
Ct,OMT /C0,DMT
0.15
0.10
According to the conclusions drawn above, how to effectively control the production of toxic intermediate omethoate is critical for sonolytic degradation of dimethoate. Omethoate has a high solubility and naturally trends to migrate to bulk solution after formation, while the real activated reaction zone for sonochemistry was internal the bubble or to less extent, at the water-bubble interface. Hence, increasing the $ OH radicals in bulk solution is a feasible method for secondary degradation of omethoate during sonolytic degradation of dimethoate. The accumulation of H2O2 and an increase in acidity caused by sonolysis provide an ideal condition for Fenton-like reaction. The combination of ultrasonic and Fe2þ was adopted to create a sonochemistryeFenton reaction condition for enhancement of omethoate degradation. The effect of different Fe2þ dosages on the sonolytic degradation of dimethoate was shown in Table 4. The maximum degradation rate was obtained with concentrations of 100 mM Fe2þ. The degradation rate constants were then about 1.9 times greater than that in the presence of ultrasound alone. The extra fraction of degradation rate should be attributed to OH radical oxidation in bulk solution. The slight decrease of the degradation rates were observed at Fe2þ concentrations over 100 mM due to the direct scavenging effect of the hydroxyl radical by Fe2þ, the excessive quantity of Fe2þ would result in a decrease in the degradation rate. The molar concentration ratios of omethoate to initial dimethoate during the course of sonolytic degradation of dimethoate were shown in Fig. 5. Compared with ultrasound alone, yields of omethoate after 60 min of reaction time can be reduced by 66e87% with the additions of Fe2þ at micro molar level and meanwhile, the peak values of the ratios curves can be reduced by 47e60%. It is found that the optimal addition dosage of Fe2þ for both dimethoate degradation and omethoate controlling was 100 mM.
0.05
4.
Conclusions
0.00 0
10
20
30
40
50
60
Time (min) Fig. 5 e The molar concentration ratios of omethoate formed to initial dimethoate versus time where [dimethoate]0 [ 6.3 mM, applied ultrasonic frequency [ 619 kHz, ultrasonic intensity [ 0.69 W cmL2, continuous mode.
This paper provides a full picture of dimethoate degradation under ultrasonic irradiation. The optimal frequency for dimethoate degradation was found to be 619 kHz under continuous mode and 406 kHz under pulse mode, respectively. The total number of cavitation bubble events per unit time and the mass transfer of the dimethoate to the bubble surface are two frequency-dependent factors that act concomitantly on dimethoate degradation kinetics. The
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 8 6 e5 8 9 4
primary pathways for degradation include hydroxyl radical oxidation, hydrolysis and pyrolysis on collapsing cavitation bubble interfaces. Dimethoate is decomposed into omethoate and N-(methyl) mercaptoacetamide in the first step, which is in agreement with the theoretical molecular orbital (MO) calculations. The kinetics models were established to well predict the concentrations of dimethoate, omethoate and N-(methyl) mercaptoacetamide as a function of time. In addition, 24.23% of the dimethoate is degraded into omethoate, and 38.78% is transformed into N-(methyl) mercaptoacetamide. Furthermore, 69.01% of omethoate produced was degraded into N-(methyl) mercaptoacetamide. The oxidation and hydrolysis of omethoate occurred at the waterbubble interface was the rate-determining step for sonolytic overall degradation of dimethoate. This observation indicates that the primary mechanism of dimethoate sonolysis is different from photo-degradation of dimethoate, in which the formation of omethoate is the dominant one. Moreover, more than 90% toxicity of dimethoate was reduced within 45 min ultrasonic irradiation. Fe2þ at micro molar level can significantly enhance the sonolytic degradation of dimethoate and effectively reduce the yields of toxic intermediate omethoate. To sum up, our results demonstrate that ultrasonic irradiation is an effective treatment for control of aqueous dimethoate and its toxic intermediates.
Acknowledgments Support for this research was provided by the National Major Project of Science & Technology Ministry of China (Grant No. 2008ZX07421-002, 2009ZX07424-004), Development Project of Ministry of Housing and Urban-Rural Development of China (Grant No. 2009-K7-4) and the Fundamental Research Funds for Central Universities (Project No. CDJRC 11210002) by the Ministry of Education, China.
references
Acero, J.L., Benitez, F.J., et al., 2008. Chlorination of organophosphorus pesticides in natural waters. Journal of Hazardous Materials 153 (1e2), 320e328. Adewuyi, G.Y., 2005a. Sonochemistry in environmental remediation. 2. Heterogeneous sonophotocatalytic oxidation processes for the treatment of pollutants in water. Environmental Science & Technology 39 (22), 8557e8570. Adewuyi, Y.G., 2005b. Sonochemistry in environmental remediation. 1. Combinative and hybrid sonophotochemical oxidation processes for the treatment of pollutants in water. Environmental Science & Technology 39 (10), 3409e3420. Buratti, F.A., Testai, E., 2007. Evidences for CYP3A4 autoactivation in the desulfuration of dimethoate by the human liver. Toxicology 241 (1e2), 33e46. Campbell, T.Y., Vecitis, C.D., et al., 2009. Perfluorinated surfactant chain-length effects on sonochemical kinetics. Journal of Physical Chemistry A 113 (36), 9834e9842. Duirk, S.E., Desetto, L.M., et al., 2009. Transformation of organophosphorus pesticides in the presence of aqueous chlorine: kinetics, pathways, and structureeactivity
5893
relationships. Environmental Science & Technology 43 (7), 2335e2340. Duirk, S.E., Desetto, L.M., et al., 2010. Chloramination of organophosphorus pesticides found in drinking water sources. Water Research 44 (3), 761e768. Dust, J.M., Warren, C.S., 2001. Kinetics and activation parameters for the alkaline aqueous rearrangement of trichorfon O,Odimethyl-(2,2,2-trichloro-1-hydroxyethyl)phosphonate to dichlorvos O,O-dimethyl-O-(2,2-dichloroethenyl)phosphate. Water Quality Research Journal of Canada 36 (3), 589e604. Echavia, G.R.M., Matzusawa, F., et al., 2009. Photocatalytic degradation of organophosphate and phosphonoglycine pesticides using TiO2 immobilized on silica gel. Chemosphere 76 (5), 595e600. Evgenidou, E., Konstantinou, I., et al., 2006. Study of the removal of dichlorvos and dimethoate in a titanium dioxide mediated photocatalytic process through the examination of intermediates and the reaction mechanism. Journal of Hazardous Materials 137 (2), 1056e1064. Evgenidou, E., Konstantinou, I., et al., 2007. Oxidation of two organophosphorous insecticides by the photo-assisted Fenton reaction. Water Research 41 (9), 2015e2027. Farooq, R., Lin, F.K., et al., 2003. Sonochemical degradation of organophosphorus pesticide in dilute aqueous solutions. Journal of Environmental Sciences-China 15 (5), 710e714. Fleming, I., 1976. Frontier Orbitals and Organic Chemical Reactions. John Wiley & Sons., Ltd, New York. Gao, J.J., Liu, L.H., et al., 2009. The occurrence and spatial distribution of organophosphorous pesticides in Chinese surface water. Bulletin of Environmental Contamination and Toxicology 82 (2), 223e229. Henglein, A., 1993. Contributions to Various Aspects of Cavitation Chemistry. JAI Press Ltd, London. Hoffmann, M.R., Hua, I., et al., 1996. Application of ultrasonic irradiation for the degradation of chemical contaminants in water. Ultrasonics Sonochemistry 3 (3), S163eS172. Hua, Inez, et al., 1997. Optimization of ultrasonic irradiation as an advanced oxidation technology. Environmental Science & Technology 31 (8), 2237e2243. Hua, I., Hochemer, R.H., et al., 1995. Sonochemical degradation of p-nitrophenol in a parallel-plate near-field acoustical processor. Environmental Science & Technology 29 (11), 2790e2796. Hua, I., Pfalzer-Thompson, U., 2001. Ultrasonic irradiation of carbofuran: decomposition kinetics and reactor characterization. Water Research 35 (6), 1445e1452. Hung, M.Hui, et al., 1998. Kinetics and mechanism of the enhanced reductive degradation of CCl4 by elemental iron in the presence of ultrasound. Environmental Science & Technology 32 (19), 3011e3016. Inoue, M., Okada, F., et al., 2006. A new development of dyestuffs degradation system using ultrasound. Ultrasonics Sonochemistry 13 (4), 313e320. Kimura, T., Sakamoto, T., et al., 1996. Standardization of ultrasonic power for sonochemical reaction. Ultrasonics Sonochemistry 3 (3), S157eS161. Kotronarou, A., Mills, G., et al., 1992. Decomposition of parathion in aqueous solution by ultrasonic irradiation. Environmental Science & Technology 26 (7), 1460e1462. Liu, Y.N., Jin, D., et al., 2008a. Study on degradation of dimethoate solution in ultrasonic airlift loop reactor. Ultrasonics Sonochemistry 15 (5), 755e760. Liu, Y.N., Jin, D., et al., 2008b. Treatment of dimethoate aqueous solution by using ultrasonic airlift loop reactor. Chinese Journal of Chemical Engineering 16 (3), 361e364. Matouq, M.A., Al-Anber, Z.A., et al., 2008. Degradation of dissolved diazinon pesticide in water using the high frequency of ultrasound wave. Ultrasonics Sonochemistry 15 (5), 869e874.
5894
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 8 6 e5 8 9 4
Ohko, Y., Iuchi, K.-i., et al., 2002. 17b-Estradiol degradation by TiO2 photocatalysis as a means of reducing estrogenic activity. Environmental Science & Technology 36 (19), 4175e4181. Pe´rier, C., Combet, E., et al., 2007. Oxygen-induced concurrent ultrasonic degradation of volatile and non-volatile aromatic compounds. Ultrasonics Sonochemistry 14 (2), 117e121. Petrier, C., Lamy, M.-F., et al., 1994. Sonochemical degradation of phenol in dilute aqueous solutions: comparison of the reaction rates at 20 and 487 kHz. Journal of Physical Chemistry 98 (41), 10514e10520. Ragnarsdottir, K.V., 2000. Environmental fate and toxicology of organophosphate pesticides. Journal of the Geological Society 157 (4), 859e876. Schramm, J.D., Hua, I., 2001. Ultrasonic irradiation of dichlorvos: decomposition mechanism. Water Research 35 (3), 665e674. Tarbah, F.A., Shaheen, A.M., et al., 2006. Distribution of dimethoate in the body after a fatal organphosphate intoxication. In: 44th Annual Meeting of the International-Association-of-ForensicToxicologists (TIAFT), Ljubljana, Slovenia. Uchida, T., Obrien, R.D., 1967. Dimethoate degradation by humin liver and its significance for acute toxicity. Toxicology and Applied Pharmacology 10 (1), 89. Wang, J., Ma, T., et al., 2006a. Investigation on the sonocatalytic degradation of parathion in the presence of nanometer rutile titanium dioxide (TiO2) catalyst. Journal of Hazardous Materials 137 (2), 972e980. Wang, J., Pan, Z., et al., 2006b. Sonocatalytic degradation of methyl parathion in the presence of nanometer and ordinary
anatase titanium dioxide catalysts and comparison of their sonocatalytic abilities. Ultrasonics Sonochemistry 13 (6), 493e500. Wang, J., Sun, W., et al., 2007. Sonocatalytic degradation of methyl parathion in the presence of micron-sized and nanosized rutile titanium dioxide catalysts and comparison of their sonocatalytic abilities. Journal of Molecular Catalysis A: Chemical 272 (1e2), 84e90. Yang, Limei, et al., 2008. Effect of ultrasound frequency on pulsed sonolytic degradation of octylbenzene sulfonic acid. Journal of Physical Chemistry B 112 (3), 852e858. Yang, L., Rathman, J.F., et al., 2005. Degradation of alkylbenzene sulfonate surfactants by pulsed ultrasound. Journal of Physical Chemistry B 109 (33), 16203e16209. Yao, J.J., Gao, N.Y., et al., 2010. Mechanism and kinetics of parathion degradation under ultrasonic irradiation. Journal of Hazardous Materials 175 (1e3), 138e145. Zhang, J., Liu, S.-S., et al., 2009. Effect of ionic liquid on the toxicity of pesticide to Vibrio-qinghaiensis sp.-Q67. Journal of Hazardous Materials 170 (2e3), 920e927. Zhang, Y.-H., Liu, S.-S., et al., 2008. Prediction for the mixture toxicity of six organophosphorus pesticides to the luminescent bacterium Q67. Ecotoxicology and Environmental Safety 71 (3), 880e888. Zhang, Z., Dai, M., et al., 2002. Dissolved insecticides and polychlorinated biphenyls in the Pearl River Estuary and South China Sea. Journal of Environmental Monitoring 4 (6), 922e928.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 9 5 e5 9 0 4
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
SAFL Baffle retrofit for suspended sediment removal in storm sewer sumps Adam Howard a,*, Omid Mohseni a, John Gulliver b, Heinz Stefan b a b
Barr Engineering Company, 4700 West 77th St., Suite 200, Minneapolis, MN 55435, USA Department of Civil Engineering, University of Minnesota, 500 Pillsbury Dr. SE, Minneapolis, MN 55455, USA
article info
abstract
Article history:
Standard sumps (manholes) provide a location for pipe junctions and maintenance access
Received 4 April 2011
in stormwater drainage systems. Standard sumps can also remove sand and silt particles
Received in revised form
from stormwater, but have a high propensity for washout of the collected sediment. With
8 July 2011
appropriate maintenance these sumps may qualify as a stormwater best management
Accepted 24 August 2011
practice (BMP) device for the removal of suspended sediment from stormwater runoff. To
Available online 12 September 2011
decrease the maintenance frequency and prevent standard sumps from becoming a source of suspended sediment under high flow conditions, a porous baffle, named the SAFL Baffle,
Keywords:
has been designed and tested as a retrofit to the sump. Multiple configurations with
Best management practices
varying percent open area and different angles of attack were evaluated in scale models.
Resuspension
An optimum configuration was then constructed at the prototype scale and evaluated for
Sediment
both removal efficiency and washout. Results obtained with the retrofit indicate that with
Settling
the right baffle dimensions and porosity, sediment washout from the sump at high flow
Stormwater
rates can be almost eliminated, and removal efficiency can be significantly increased at low
Sumps
flow rates. Removal efficiency and washout functions have been developed for standard
Testing
sumps retrofitted with the SAFL Baffle. The results of this research provide a new, versatile stormwater treatment device and implemented new washout and removal efficiency testing procedures that will improve research and development of stormwater treatment devices. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
The implementation of the 1987 Amendments to the Clean Water Act (Smith, 2001) resulted in major changes in the management of stormwater. Detention ponds, infiltration basins, and underground proprietary devices became popular stormwater best management practices (BMPs) to meet the National Pollutant Discharge Elimination System (NPDES) regulations in urban areas. The European countries have also established regulatory frameworks known as “water framework directives (WFD)” to protect the discharge of pollutants
into surface and groundwater. Detention ponds and infiltration basins often require large surface areas, i.e. land, which may be expensive or not even available in urban environments. Reaching the stormwater pollution prevention goals can therefore be challenging in urban environments. Within the last 25 years, many cities, counties and watershed districts have turned to underground stormwater treatment devices to reach their goals. These devices create flow patterns suitable for capturing suspended sediment and floatables. Understanding how well these devices remove particulates from stormwater has been the subject of many
* Corresponding author. Tel.: þ1 712 898 0838. E-mail address:
[email protected] (A. Howard). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.08.043
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 9 5 e5 9 0 4
Since the standard sump has already been proven effective at capturing sediment, a simple retrofit to increase sediment retention during large storms would be useful. If a simple, effective retrofit could be found to decrease sediment washout from standard sumps, they may qualify as a stormwater BMP. Therefore, the objectives of this study were to identify the causes of sediment resuspension and washout in standard sumps and to design a retrofit for significant reduction in washout of previously captured sediment that can be easily implemented in the field.
2.
Methods
2.1.
Measurement of flow pattern in a sump
To better understand scour and deposition in the sump, Howard et al. (2011) did extensive flow mapping using a 3dimensional Acoustic Doppler Velocimeter (ADV). The vector plot developed from velocity measurements indicates that after entering the sump the flow plunges and creates a downward velocity at the downstream end, a lateral velocity near the sediment bed, and an upward velocity at the upstream end. This circulation pattern can result in scouring of the sediment deposit at the downstream end and more deposition at the upstream end of the sump. It was determined that an effective retrofit should break the circulation pattern and dissipate the energy of the plunging flow.
2.2.
Scale model testing of SAFL baffle design
Utilizing information on the flow patterns, testing of various sump retrofit designs was initiated in a 1:4.17 Froude scale model. The goal of the retrofit was to reduce the power of the strong circulation pattern. Solid baffles have already been installed in many sumps across the country, therefore, two types of solid baffles were tested by placing them perpendicular to the flow in the center of the sump with a space between the bottom of the baffles and the top of the sediment deposit. The bottom of the solid baffle 1 and solid baffle 2 were located 0.14 and 0.36 m (5.5 and 6.8 in), respectively, above the top of the sediment layer at the bottom of the sump. As shown in Fig. 1, the standard sump with a solid baffle had higher 10000.0
1000.0
100.0
10.0
1.0
Fig. 1 e Standard sump scale model washout results compared to several porous baffle designs.
Solid Baffle 2
Solid Baffle 1
Without Baffle
Porous Baffle 7
Porous Baffle 6
Porous Baffle 5
Porous Baffle 4
Porous Baffle 3
Porous Baffle 2
0.1 Porous Baffle 1
studies (Carlson et al., 2006; Mohseni et al., 2007; Wilson et al., 2009; Kim et al., 2007; Howard et al., 2011). As noted by Saddoris et al. (2010), the overall performance of underground proprietary devices depends on two processes. First, sediment laden water enters the device at low flow rates and the sediment will settle to the bottom. This process is often called the device’s ‘particulate removal efficiency’ or ‘capture efficiency’. The second process occurs when the previously captured sediment is resuspended and moved from the bottom of the device to the outlet pipe at high water flow rates. This process is called ‘washout’ (Howard et al., 2011). Carlson et al. (2006) and Mohseni and Fyten (2007) evaluated two proprietary underground settling devices (hydrodynamic separators) in the laboratory for particulate removal efficiency. A mass balance approach was used starting with the introduction of a known dry quantity of particulates with a known particle size into a measured water flow rate for a set duration. The captured sediment quantity was then collected and measured following the test. The captured sediment was dried, sieved, and weighed in order to quantify the removal efficiency of the device. This testing procedure was also implemented in the field by Wilson et al. (2009) on four hydrodynamic separators. Mohseni and Fyten (2007) also evaluated the washout of sediment from a swirl flow device in the laboratory using a mass balance approach. The measurement of washout required the loading of a known sediment quantity prior to the introduction of water flow rates higher than the maximum design treatment rate specified by the manufacturer. Following the test the amount of sediment remaining in the sump was determined by taking sediment depth measurements throughout the device using a hand held ruler. The volume of sediment and the effluent concentration due to washout could then be determined. Saddoris et al. (2010) also implemented this testing procedure in the field for one device, using a laser distance measurement and a hand held ruler and added load cells in the laboratory for three other devices. The results of these studies indicate that hydrodynamic separators can effectively remove sediment larger than silt from stormwater runoff. The ability of a device to retain this captured sediment is largely dependent on the particular design and the maintenance plan designed by the owner. Some devices may have near zero washout at high flows, while other devices can exhibit sediment effluent concentrations exceeding 1000 mg/L (Saddoris et al., 2010). Howard et al. (2011) evaluated standard sumps of different sizes for both removal efficiency and washout in the laboratory using the mass balance approach described above. It was found that standard sumps also remove suspended sediment from stormwater runoff, although not quite as well as hydrodynamic separators. However, the main detriment of standard sumps is their propensity for washout. Standard sumps may exhibit washout effluent concentrations up to 1000 mg/L. Nevertheless they do have two main advantages over proprietary devices. First, since the standard sump does not require any internal components, it is inexpensive in comparison to hydrodynamic separators. Second, many standard sumps have already been installed in urban stormwater collection systems, at pipe junctures and for maintenance purposes.
Average Effluent Concentration (mg/L)
5896
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 9 5 e5 9 0 4
washout rates than a sump without the solid baffle, i.e. the washout rate with solid baffles 1 and 2 increased by a factor of approximately 25 and 10, respectively. A solid baffle increases washout rates because the same flow rate is required to pass through a smaller cross section between the bottom of the baffle and the sediment surface. The reduction in cross sectional area increases flow velocities on the sediment surface, which results in more sediment entrainment and washout. Solid baffle 1 had more washout than baffle 2 because of the smaller opening between the bottom of the baffle and the sediment surface. To avoid this reduction in cross-sectional area of the flow through the sump, porous baffles were tested. The angle of attack and the porosity were varied in these tests. Fig. 1 gives the washout results for a variety of porous baffles in the 1:4.17 scale model. Porous baffles 1 and 2 in Fig. 1 were tested with angles of attack of þ20 striking face pointing down) and 20 (striking face pointing up), respectively, and porosities of 33%. Baffles 3 through 7 were vertical plates with porosities of 33%, 36%, 40%, 46% and 51%, respectively. The flow rate for the tests shown in Fig. 1 was 5.7 L/s (0.2 cfs) for the scale model (199 L/s or 7 cfs in the prototype), and the preloaded sediment was U.S. Silica F-110 (with a median size of 110 mm). The duration of all tests shown in Fig. 1 was 40 min. All porous baffles tested had reduced sediment washout compared to standard sumps. The tests in the scale model showed that with a porosity of 46%e 51% and vertical orientation (baffles 6 and 7), the baffles were very effective: the washout effluent concentration can become negligible. As a reference in Fig. 1, the standard sump without baffles had effluent concentrations above 70 mg/L, almost two orders of magnitude more than standard sumps with porous baffles 6 and 7. Indications from the scale model are that, for the same flow rate and particle size a baffle with an appropriate orientation and porosity will produce effluent concentrations near zero. With an appropriate porosity, the baffle will dissipate the energy of the plunging flow by redistributing the flow across the entire sump width. The improvements to the sump during washout tests were also examined visually. The sediment scour downstream and the deposition upstream of the sump were clearly evident in the 1:4.17 scale model. The surface of the sediment deposit remained flat with the porous baffle 6 in place. The porous baffle 6 was determined to be an acceptable configuration and was given the name, SAFL Baffle. The SAFL Baffle could now be tested at the full scale for both sediment washout and removal efficiency.
5897
walls of the sump. For good access to the sump bottom during maintenance, the baffle was installed in a vertical position perpendicular to the flow at the center of the sump. Fig. 2 show an oblique view from above of the inlet pipe and the SAFL Baffle installed in the 1.2 m (4 ft) sump. The second straight flow-through sump evaluated with the SAFL Baffle was 1.8 m (6 ft) in diameter and 0.9 m (3 ft) deep with 0.61 cm (24 in) inlet and outlet pipes. This baffle had 7.7 cm (3 in) holes and 46% porosity. The baffle installation in the 1.8 m (6 ft) sump was similar to the 1.2 m (4 ft) sump. Each sump was connected to the SAFL plumbing system which provides approximately 13.7 m (45 ft) of head of Mississippi River water. The free surface flow into each sump through the 0.31 m (12 in) supply pipe was measured using two Pitot cylinders (Silberman, 1947). The free surface flow into both sump configurations was controlled using a hydraulic gate valve in the supply pipe. For removal efficiency testing, sediment was fed at a calibrated rate as a slurry from a Schenk AccuRate sediment feeder into the inlet pipe approximately one foot upstream of the sump. For washout testing, the sump, mounted on precision strain gauge load cells, was weighed before, during and after tests for accurate determination of the sediment loss during the test (Saddoris et al., 2010). For the 1.2 m (4 ft) sump removal efficiency testing, washout rates were measured at flow rates of 17.0, 34.0, 51.1 and 64.1 L/s (0.6, 1.2, 1.8, and 2.4 cfs) in triplicate. The 1.8 m (6 ft) sump was tested at flow rates of 51.1, 99.3 and 150.4, and 198.2 L/s (1.8, 3.5, 5.3, and 7 cfs). Additional removal efficiency tests were conducted at flow rates of 8.5, 11.3 and 14.2 L/s (0.3, 0.4, and 0.5 cfs) to obtain removal efficiencies of approximately 100%. Carlson et al. (2006) showed that the results of removal efficiency tests are independent of the influent concentration when sediment particles do not impact each other’s flow path. Therefore, the influent concentrations of the removal efficiency tests were varied between relatively low values of 100e200 mg/L. According to Wilson et al. (2007), a total sediment input of 10e15 kg minimized the errors associated with sediment collection at the bottom of the sump. To meet the requirements of this test series, the duration of each test was
2.3. Experimental setup and procedure for full scale testing Two full scale fiberglass sumps were constructed for testing and placed on a test stand in the St. Anthony Falls Laboratory (SAFL) of the University of Minnesota in Minneapolis, Minnesota. The first straight flow-through sump evaluated with the SAFL Baffle was 1.2 m (4 ft) in diameter and 1.2 m (4 ft) deep (the height of the inlet pipe from the sump bed), and had 38.1 cm (15 in) inlet and outlet pipes. There was a one percent drop between the inlet and outlet pipes of the sump. The baffle for this setup had circular holes of 2.5 cm (1 in) diameter and 46% porosity. The frame of the baffle was bolted to the
Fig. 2 e SAFL Baffle in the 1.2 3 1.2 m (4 ft 3 4 ft) sump.
5898
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 9 5 e5 9 0 4
3.
Results of full scale testing
The effects of the SAFL Baffle on the washout rate and removal efficiency of standard sumps were investigated in 1.2 1.2 m (4 4 ft) and 1.8 0.9 m (6 3 ft) sumps (The first number is the sump diameter and the second number is the height of the sump below the outlet pipe invert). The results are presented and compared with those obtained in standard sumps with no baffle.
3.1.
Removal efficiency results
3.1.1.
1.2 1.2 m (4 4 ft) sump with the SAFL Baffle
Removal efficiencies measured in the 1.2 1.2 m (4 4 ft) sump are plotted against water flow rate in Fig. 3. With the
Fig. 3 e Sediment removal efficiencies in the 1.2 3 1.2 m (4 3 4 ft) sump with and without the SAFL Baffle. The first values in the legend are the diameter 3 depth, and the second value is the particle diameter in microns.
SAFL Baffle installed, there is a significant increase in removal efficiency (approximately 15%) of small size particles (i.e. 107 mm) at low flow rates, e.g. below 14.2 L/s (0.5 cfs). The SAFL Baffle could not improve the performance for large particles (i.e. 545 mm) at 14.2 L/s (0.5 cfs) and below, because the sump without the SAFL baffle is already achieving near 100% removal. At medium flow rates, around 28.4 L/s (1 cfs), however, standard sumps with the SAFL Baffle exhibited between 10 and 20% increase in removal efficiency for all particle sizes tested. At higher flow rates, above 42.6 L/s (1.5 cfs), smaller sediment particles (i.e. 107 mm) were not removed more efficiently with the SAFL Baffle in place, but larger sediment sizes (i.e. 545 mm) were removed approximately 10% more efficiently.
3.1.2.
1.8 0.9 m (6 3 ft) sump with the SAFL Baffle
The results of the removal efficiency tests conducted in the 1.8 0.9 m (6 3 ft) sump are shown in Fig. 4. At very high and 120% 1.8x0.9, 110
100%
Removal Efficiency, η
selected a priory so that 10e20 kg of sediment would be fed. The test duration was 1 h for the lowest flow and 18 min for the higher flow rates. Three distinct particle sizes were used for the tests. The median sediment sizes were 545 mm (500 mme589 mm), 303 mm (250 mme355 mm), and 107 mm (88 mme125 mm). Following the test, the sediment captured in the sump was removed, dried and post-sieved into the original distinct particle sizes. Each particle size was then weighed and compared to its initial weight to determine its respective removal efficiency. Washout tests required preloading of the sump with sediment and then applying high flow rates to the sump. Prior to each washout test, 0.31 m (12 in) of U.S. Silica sand F-110 with a median particle size of 110 mm was placed in the sump. Great care was taken to ensure that the initial conditions were identical for all tests. Two methods were used to measure the weight of sediment in the sump before and after each test. In the first method, stick measurements were taken to measure the depth of wet sediment at 24 locations, which were then used to estimate an average depth of the wet sediment. The difference between the average depths of sediment before and after each test was used to determine the volume of the washed out sediment. Tests were conducted on several sediment samples to determine the bulk specific weight under water. The bulk specific weight typically varied from 1650 kg/ m3 to 1730 kg/m3 (spec. gravity from 1.65 to 1.73) for the 8 to 10 samples taken from different locations and depths within the sump. In the second method, precision strain gauge load cells were used to weigh the sediment deposited in the sump before and after each test. The 5,000 lb precision strain gauge load cells from Tovey Engineering, Inc. provided accurate measurements of weight throughout each test. In order to take accurate weight measurements it was necessary to record both the initial and final weight measurements with a precise measurement of water elevation in the sump. The water level in the sump was measured before and after each test using a precision manometer. The flow rate in the washout tests on the 1.2 m and 1.8 m (4 ft and 6 ft) sumps varied from 78 to 156 L/s (2.75e5.5 cfs) and 142 to 539 L/s (5e19 cfs), respectively. The flow rate in each test was kept constant.
1.8x0.9, 335 1.8x0.9, 545
80% 1.8x0.9, 110, SAFL Baffle 1.8x0.9, 335, SAFL Baffle
60%
1.8x0.9, 545, SAFL Baffle
40%
20%
0% 0.00
50.00
100.00
150.00
200.00
250.00
Flow Rate (L/s)
Fig. 4 e Sediment removal efficiencies in the 1.8 3 0.9 m (6 3 3 ft) sump with and without the SAFL Baffle. The first values in the legend are the diameter 3 depth, and the second value is the particle diameter in microns.
5899
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 9 5 e5 9 0 4
very low flow rates the removal efficiencies were more or less the same for all particle sizes tested. However, at median flow rates and median particle sizes (i.e. 303 mm), the sump exhibited a 15% on average increase in sediment capture with the SAFL Baffle.
3.2.
Washout results
3.2.1.
1.2 1.2 m (4 4 ft) sump with the SAFL Baffle
Fig. 5 gives the results of the washout tests both with and without the SAFL Baffle, where the sediment concentration in the sump effluent is plotted against the flow rate through the sump. The SAFL Baffle reduced effluent concentrations from 500 mg/l to nearly 0 mg/l at 142 L/s (5 cfs) flow. The uncertainty of the load cells used in the tests was þ/ 10 lbs to the 95% confidence interval. In the majority of tests conducted with the SAFL Baffle the measured washout was less than þ/ 10 lbs after 2 h of testing. To obtain results outside the uncertainty range of the load cells, it was decided to use a finer sediment which would be expected to give a higher washout rate. The top three inches of the F110 Silica sand (median size ¼ 110 mm) in the sump were therefore replaced with SCS250 (median size ¼ 45 mm), and the tests were repeated. The results obtained with the SCS250 were similar to those obtained with the F110 silica sand, i.e. effluent concentrations obtained with the SCS250 stayed near zero after the sump was subjected to high flow rates for 2 h (Fig. 5).
3.2.2.
1.8 0.9 m (6 3 ft) sump with the SAFL Baffle
The SAFL Baffle was tested in a shallow 0.9 m (3 ft) deep sump of 1.8 m (6 ft) diameter which imposes more challenging conditions for washout, i.e. increased the flow rates and decreased sump depth. The baffle used in this test series had the same porosity, but had 7.6 cm (3 in) holes, compared to the 2.5 cm (1 in) holes used in the 1.2 m (4 ft) sump tests with the SAFL Baffle. In practice, 7.6 cm (3 in) holes allow more trash to pass through the sump reducing the clogging potential of the baffle. A SAFL Baffle with 7.6 cm (3 in) holes may require less maintenance and therefore be more desirable. From a hydrodynamics point of view, the difference between the two is
small. Fig. 6 provides a comparison of the results of the tests conducted on the 1.8 0.9 m (6 3 ft) sump with and without the SAFL Baffle. The baffle improved the ability of the sump to retain sediment. At 454 L/s (16 cfs) flow, the SAFL Baffle decreased the sediment washout concentration from 800 mg/ L to 50 mg/L. A review of the removal efficiency and washout test data shows that with the SAFL Baffle retrofit, even small particle sizes which have collected in the sump will remain in the sump at high flow conditions. For instance, a 1.2 1.2 m (4 4 ft) sump can capture 18% of a 107 mm particle size during a flow event with a magnitude of 28 L/s (0.6 cfs). When followed by larger storm event, such as a 156 L/s (5.5 cfs) event which will capture nearly 0% of this sediment size, all of what was captured previously will remain in the sump. Without the SAFL Baffle, most of this sediment will be washed out of the sump during the larger storm event.
3.3.
Head loss induced by the SAFL Baffle
A possible concern with the SAFL Baffle is the increase in head loss. If the head loss is too high at high flow rates, the sump may overtop and/or cause flooding upstream of the sump. With the baffle in the sump, the maximum increase in the head loss was determined to be 0.061 m (0.2 ft) at a flow rate of 51 L/s (5.5 cfs). This increase in head loss does not cause a problem for most storm sewer designs. Fig. 7 is a plot of head loss vs. flow rate with and without the SAFL Baffle in place. This head loss does not include clogging, which is a topic of current investigations.
4.
Interpretation of the results
4.1.
Scaling
Standard sumps in storm sewer systems are designed and built with a variety of sump diameters and depths, and inlet and outlet pipe diameters. They also have to handle a wide range of flow rates, water temperatures, sediment sizes and densities. Since standard sumps can not be tested for all
600 1000
400
Original Sump with F-110
900 Effluent Concentration (mg/L)
Effluent Concentration (mg/L)
500
SAFL Baffle with F-110
300 SAFL Baffle with SCS 250
200
100
800
1.8x0.9
700 600 500 400
1.8x0.9 SAFL Baffle
300 200 100
0 0.00
50.00
100.00
150.00
200.00
Flow Rate (L/s)
Fig. 5 e Effluent concentrations of the washout tests in the 1.2 3 1.2 m (4 3 4 ft) sump with and without SAFL Baffle for a mean particle size of 110 mm.
0 0.0
100.0
200.0
300.0
400.0
500.0
Flowrate (L/s)
Fig. 6 e Effluent concentrations of the washout tests in the 1.8 3 0.9 m (6 3 3 ft) sump with and without SAFL Baffle for a mean particle size of 110 mm.
5900
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 9 5 e5 9 0 4
Pe Vs D2 hg ¼ 2 Frj U2j Q
0.14 Standard Sumps
0.12
Since the same parameters also affect the removal efficiency of standard sumps, the removal efficiency was replotted versus the Pe´clet-Froude number ratio. It was shown that Pe=Fr2j can predict particle capture under low flow conditions for all Froude-scaled conditions, especially with a change in length scales. For removal efficiency, a three parameter, nonlinear, increasing function was fitted to the data to create a removal efficiency function that gives correct results at the low and high Pe=Fr2j limits. The function is described by the relationship shown in Eq. (3).
SAFL Baffle
Headloss (m)
0.10
0.08
0.06
0.04
0.02
0.00 0
20
40
60
80
100
120
140
160
180
Flow Rate (L/s)
Fig. 7 e Head loss vs. flow rate in a 1.2 3 1.2 m (4 3 4 ft) sump with and without the SAFL Baffle.
possible combinations, it is helpful to find parameters and functions which can effectively scale the test results. The Pe´clet number, given in Eq. (1), is the ratio of settling to mixing by turbulent diffusion (Dhamotharan et al., 1981) and has been used by Carlson et al. (2006), Mohseni et al. (2007) and Wilson et al. (2009) to scale removal efficiency in underground settling devices. Pe ¼
Us h D Q
(2)
(1)
In Eq. (1), Us is the settling velocity of the particles, h is the sump depth, D is the sump diameter, and Q is the flow rate. As shown by Wilson et al. (2009), when removal efficiency is plotted versus the Pe´clet number, a performance function in the form of removal efficiency vs. Pe´clet number can be developed for a stormwater treatment device. The Pe´clet number works well to describe the performance of a single device under a wide range of flow rates, particle sizes and densities, and water temperatures. However, by changing the length scales, h and D, in Eq. (1), the function may not be capable of predicting the removal efficiency for all sizes, i.e. a different fitted model may be required for each device size. Howard et al. (2011) measured removal efficiencies and washout rates in five different standard sumps: 1.8 1.8 m (6 6 ft), 1.8 0.9 m (6 3 ft), 1.2 1.2 m (4 4 ft), 1.2 0.6 m (4 2 ft) and 0.3 0.3 m (1 1 ft). Howard et al. (2011) were able to derive washout and removal efficiency functions of standard sumps for this wide range of length scales using the ratio of the Pe´clet number to the square of the inflow jet Froude number (Pe=Fr2j ) as the scaling parameter. The inflow jet Froude number was defined as Frj ¼ Uj/(gD)1/2, where Uj is the mean velocity of the inflow (jet) and g is the acceleration of gravity. In the derivation, the power supplied by the inflow was compared to the power required for the incipient motion of the particles, entrainment into the water column and lifting to the height of the outflow pipe. The Pe´clet-Froude number ratio, Eq. (2), was obtained from the ratio of the settling power of particles to the power supplied.
31=b 2 1 1 h¼4 bþ b 5 R a Pe=Fr2j
(3)
In Eq. (3), R, a, and b are fitted parameters and h is predicted removal efficiency. For washout, a three parameter, nonlinear, decreasing function was fitted to the data to develop a washout function. The fitted sump washout function is given in Eq. (4). 2 ^ ¼ CðSG 1Þ ¼ a þ belPe=Frj C rw SG Pe=Fr2j
(4)
In Eq. (4), a, b and l are fitted parameters, C is the effluent ^ is the non-dimensional effluent concentraconcentration, C tion, SG is the particle specific gravity and rw is the density of water. The root mean square error (RMSE), Eq. (5), is calculated for Eqs. (3) and (4). The RMSE measures the goodness of fit between the data and estimated value from the fitted equation. sP ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ðXi Xmi Þ2 RMSE ¼ np
(5)
In Eq. (5), Xi is the measured value (removal efficiency for Eq. (3) or non-dimensional concentration for Eq. (4)), Xmi is the computed (modeled) variable using either Eq. (3) or 4, n is the number of data points, and p is the number of fitted parameters. Another measure of the goodness of fit for Eqs. (3) and (4) is the Nash-Sutcliffe Coefficient (NSC), which is described by Eq. (6). An NSC of one represents a perfect fit, i.e. all measured data lie on the curve. P ðXi Xmi Þ2 NSC ¼ 1 P ðX Xi Þ2
(6)
In Eq. (6), X is the mean of measured values. NSC is a measure of goodness of fit for nonlinear functions with a maximum of unity. The closer the NSC value is to one, the higher the correlation between the fitted model and the measured data.
4.2.
Increase of removal efficiency by the SAFL Baffle
Fig. 8 is a plot of the measured removal efficiency versus the dimensionless parameter Pe=Fr2j . The data plotted have been obtained in tests conducted on the 1.2 1.2 m (4 4 ft) and
5901
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 9 5 e5 9 0 4
100% 90%
Removal Efficiency, η
80% 70% 60% 50%
Standard Sump Model
40% Standard Sump Data
30%
SAFL Baffle
20% SAFL Baffle Model
10%
0% 0.01
0.1
1
10
100
1000
10000
Pe/Fr j2
Fig. 8 e Comparison of removal efficiency for standard sumps with and without SAFL Baffle.
1.8 0.9 m (6 3 ft) sumps with and without the SAFL Baffle in place. The fitted function for removal efficiency h for the data including both the 1.2 1.2 m (4 4 ft) and 1.8 0.9 m (6 3 ft) standard sumps is given by Eq. (7). 123
0 B B B h¼B B1 þ B @
C C C !32 C C Pe C 0:028 2 A Frj 1
(7)
Eq. (7) has a RMSE of 8.2% and a NSC of 0.944. The fitted function for the data including both the 1.2 1.2 m (4 4 ft) and 1.8 0.9 m (6 3 ft) standard sumps with the SAFL Baffle is given by Eq. (8). 1178
0 B B B h¼B B1 þ B @
C C C !178 C C C Pe 0:021 2 A Frj 1
(8)
Eq. 8 has a RMSE of 4.9% and a NSC of 0.98. Using the fitted functions (Eqs. (7) and (8)), it is possible to predict the removal efficiency for a particular standard sump, influent particle size, water temperature, and flow rate, both with and without the SAFL Baffle. A standard sump can also be designed to meet a particular removal efficiency goal for a given influent particle size and flow rate. The results presented in Fig. 8 suggest that standard sumps retrofitted with the SAFL Baffle exhibit only a slight increase in removal efficiency at Pe=Fr2j values above approximately 50. However, it is important to note that at a given flow rate and for a given particle size and density, the value of Pe=Fr2j is not the same for standard sumps with and without the SAFL Baffle. The inflow jet velocity is an important parameter in Pe=Fr2j and it is squared in the denominator. Any slight increase in head loss due to the presence of the SAFL Baffle causes a decrease in the jet velocity and therefore
Fig. 9 e Comparison of the inflow Fr2j in sumps with and without the SAFL Baffle for a variety of flow rates.
a larger increase in Pe=Fr2j . Fig. 9 is a plot of estimated Fr2j evalues in the experiments versus flow rate. The figure gives the data for standard sumps with and without the SAFL Baffle. For the same flow rate, the standard sump with a SAFL Baffle has a lower Fr2j -value. Thus, for the same flow rate, a sump with the SAFL Baffle will have a higher Pe=Fr2j , i.e. for the same flow rate and particle size, a sump with the SAFL Baffle will be more efficient in removing suspended sediments as was already shown in Figs. 3 and 4. For example, consider that it is desired to design a standard sump that removes 80% of the suspended sediment with a 110 mm particle size at a flow rate of 17 L/s (0.6 cfs). If the inflow pipe has a diameter of 0.38 m (15 inches), the typical inflow jet velocity will be about 0.49 m/s (1.6 ft/s). An 80% removal efficiency for the standard sump with SAFL Baffle occurs at a Pe=Fr2j value of 61. If the water temperature is 2.2 C (35 F), the settling velocity will be 0.45 cm/s (0.0129 ft/s). At 17 L/s (0.6 cfs), the value of h*D2 will become 7.4 m3 (225 ft3). The sump would need to be 1.95 m (6 ft) deep and 1.95 m (6 ft) in diameter to satisfy these requirements. For complete sizing, continuous hydrologic modeling of the drainage basin is required (NCHRP 2006; Mohseni et al., 2009). By incorporating the removal efficiency function into a hydrologic model, the annual sediment removal and removal efficiency can be determined for a known particle size distribution. Through trial and error minimum length scales can be determined based on a total suspended sediment removal goal.
4.3.
Reduction in washout by the SAFL Baffle
Fig. 10 gives the results of washout tests for the standard sump with and without the SAFL Baffle. The washout functions of the standard sumps with and without the SAFL Baffle are provided in Eqs. (9) and (10), respectively. With SAFL Baffle: 0:69 2 CðSG 1Þ 1:67 106 F j ¼ þ 5:16 107 e Pe pw SG 2 Fj
Pe
(9)
5902
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 9 5 e5 9 0 4
Table 1 e Average annual removal efficiency of standard sumps without incorporating washout for a 2 ha urban drainage basin in Minneapolis, MN and assuming OK110 as the gradation of suspended sediments in stormwater. The numbers in the first column are the diameter 3 depth in meters. Standard sumps model 1.2 1.2 1.8 1.8
0.6 1.2 0.9 1.8
Total Load (kg)
Total Load Removed (kg)
Removal Efficiency (%)
27323 27323 27323 27323
4253 7196 11261 14795
15.6 26.3 41.2 54.1
Fig. 10 e Comparison of washout functions of standard sumps with and without the SAFL Baffle.
Without SAFL Baffle: 3:18 2 CðSG 1Þ 8:3 106 F j ¼ þ 4:7 104 e Pe pw SG F2j
Pe
(10)
The Nash-Sutcliffe Coefficient (NSC) for Eq. (9) is 0.55 and the NSC value for Eq. 10 is 0.67. The RMSE for Eqs. (9) and (10) is nearly zero. These two fitted models can be used to predict sediment effluent concentrations from a standard sump for any combination of particle size, particle specific weight, water temperature, sump depth, sump diameter, sump inlet pipe size, and flow rate. In addition, the models can also be used to select the required design of a standard sump if a desired effluent concentration at high flow rate has been selected. When a sump without the SAFL Baffle is designed for high sediment removal efficiency and low washout, the controlling process will be washout. This is due to the high propensity of standard sumps for washout. In other words, to meet the washout requirements a larger sump size may be required than to achieve the required removal efficiency. The opposite is true if the sump is designed with the SAFL Baffle. The baffle creates a substantial reduction in washout, and the size of the sump will be determined by the required removal efficiency. The use of standard sumps for pipe junctions and maintenance access has made them plentiful in many stormwater management systems. The implementation of the SAFL Baffle as a retrofit for standard sumps makes it an inexpensive and readily applicable option for the treatment of sediment laden stormwater. In addition, the SAFL Baffle can be installed as a new installation for current and future development projects. This installation flexibility makes it an easy option for many stormwater mangers.
4.4.
Maintenance
Maintenance plays an important role in the cost and overall performance of stormwater BMPs and should be considered in all municipal stormwater plans. To evaluate the maintenance requirements of standard sumps a free software download from Barr Engineering Company, named SHSAM (Mohseni et al., 2009) was used. This software utilizes the fitted removal efficiency and washout functions described above in
a continuous runoff model from small urban drainage basins. The output provides a prediction of the amount of sediment that would be captured in the standard sump over a period of time, e.g. 15 years. To quantify the significance of the SAFL Baffle on the required maintenance of standard sumps, a relatively typical residential drainage basin in the Midwest USA was used in the analysis. The input data, therefore, included a 2 ha (5 acres) watershed with 25% imperviousness, a 366 m (1200 ft) hydraulic length, 2% average watershed slope, and a curve number of 78. The climatic data came from Northfield, MN and included both 15-min precipitation and daily air temperature readings for 17 years. The example also utilized a constant influent concentration of 400 mg/L and with the assumption that the suspended sediments in stormwater runoff resembled the OK110 particle size distribution (D50 ¼ 110 mm). The model was run under four scenarios for the tested sizes of standard sumps: (1) Removal efficiency of standard sumps ignoring washout and including one cleaning of the sump at the end of each year (Table 1), (2) removal efficiency of standard sumps incorporating washout and one cleaning of the sump at the end of each year (Table 2), (3) removal efficiency of standard sumps with washout and increasing the number of sump cleaning per year to approach the values of the first scenario (Table 3), and (4) removal efficiency of standard sumps with the SAFL Baffle and one cleaning of the sump at the end of each year (Table 4). For scenarios 1, 2 and 4 the program assumed additional cleaning to occur whenever the depth of the deposit exceeded one foot. The data provided in Table 1 are the maximum possible removal efficiencies of standard sumps for the given drainage
Table 2 e Average annual removal efficiency of standard sumps with washout for a 2 ha urban drainage basin in Minneapolis, MN and assuming OK110 as the gradation of suspended sediments in stormwater. The numbers in the first column are the diameter 3 depth in meters. Standard sumps model 1.2 1.2 1.8 1.8
0.6 1.2 0.9 1.8
Total load (kg)
Total load removed (kg)
Removal efficiency (%)
27323 27323 27323 27323
2398 4283 7279 12278
8.8 15.7 26.6 44.9
5903
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 9 5 e5 9 0 4
Table 3 e Average annual removal efficiency of standard sumps with washout for a 2 ha urban drainage basin in Minneapolis, MN and assuming OK110 as the gradation of the suspended sediments in stormwater. The frequency of sump cleaning has increased to achieve the removal efficiencies summarized in Table 1. The numbers in the first column are the diameter 3 depth in meters. Standard sumps model 1.2 1.2 1.8 1.8
0.6 1.2 0.9 1.8
Total load (kg)
Total load removed (kg)
Removal efficiency (%)
Average number of cleanings per year
Maximum number of cleanings per year
Minimum number of cleanings per year
27323 27323 27323 27323
3733 6569 11073 14428
13.7 24.0 40.5 52.8
2.6 5.8 4.6 5.2
5 9 7 7
1 2 2 2
Table 4 e Average annual removal efficiency of standard sumps equipped with the SAFL Baffle with washout for a 2 ha urban drainage basin in Minneapolis, MN and assuming OK110 as the gradation of suspended sediments in stormwater. The numbers in the first column are the diameter 3 depth in meters. Standard sumps model 1.2 1.2 1.8 1.8
0.6 1.2 0.9 1.8
Total load (kg)
Total load removed (kg)
Removal efficiency (%)
Average number of cleanings per year
Maximum number of cleanings per year
Minimum number of cleanings per year
27323 27323 27323 27323
4875 9247 14647 17714
17.8 33.8 53.6 64.8
0.4 0.9 0.7 0.8
1 1 1 1
0 0 0 0
basin, climate condition and particle size distribution. Table 3 was derived with the goal of achieving the maximum removal efficiencies shown in Table 1 through implementation of a maintenance program. By incorporating washout, the overall removal efficiency of standard sumps drops by 7e15 percent (see Tables 1 and 2). These results indicate that any analysis that does not include washout in predicting removal efficiency is over-predicting sediment removal from stormwater. For example, a 1.8 m (6 ft) diameter by 1.8 m (6 ft) deep sump must be cleaned an average of 6.2 times per year (5.2 times on average per year plus once at the end of the year) over a 17-year period to nearly achieve its maximum removal efficiency. In comparison, the results with a SAFL Baffle in place, provided in Table 4, show that the cleaning of the sump is only required 1.8 times on average per year (0.8 on average times per year plus once at the end of the year) to achieve an efficiency 10% higher than the maximum value shown in Table 1. The results also show that a 1.2 0.6 m standard sump, with an overall removal of between 8.8% and 17.8% is likely too small for the assumed drainage basin, and the SAFL Baffle will not provide much benefit.
5.
Summary and conclusions
Standard sumps effectively remove suspended sediment from stormwater runoff under low flow conditions. A major deficiency of standard sumps is their inability to retain the captured sediment under high flow conditions. This drawback can be overcome by either frequent maintenance (sediment removal by maintenance crew) of a sump or by retrofitting a sump such that the flow through the sump results in reduced washout. Only then can the standard sump be considered an effective device for removal of suspended sediment from stormwater runoff.
The design of this retrofit was possible through the implementation of the washout testing method originally developed by Saddoris et al., (2010). The application of load cells to quantify the amount of material removed from the sump during high flow events provides a level of accuracy and repeatability that the conventional sampling methods cannot achieve. Another unique characteristic of this research is the utilization of a mass balance approach to quantify sediment capture initially proposed by Carlson et al. (2006) and a nondimensional scaling function proposed by Howard et al. (2011) to evaluate the SAFL Baffle design. This approach provided a confidence in the results that allowed for further design refinement. This testing and scaling approach is a powerful tool for research and development of hydrodynamic separators. Innovative technologies such as the SAFL Baffle and the methods used in its design will play an important role as regulations continue to support stormwater management and budget shortfalls remain prevalent for stormwater management entities. The first objective of the research was accomplished by observing that during the sediment washout tests with standard sumps sediment was picked up, moved upstream, and deposited just below the inlet pipe at the bottom of the sump. This flow pattern was verified by measuring velocities in the sump. The circulation pattern consisting of a downward flow due to the plunging of the incoming jet, a lateral flow near the sediment bed, and an upward flow at the upstream end of the standard sump enhanced the washout of previously deposited sediment at the bottom of the sump. To break the adverse circulation pattern and to dissipate the energy of the incoming and plunging flow, a porous baffle, called SAFL Baffle, was designed to meet the goals of the second objective of the study. The SAFL Baffle with 46% porosity and a vertical orientation was first evaluated in a 1:4.17 Froude scale model. After the baffle was evaluated in
5904
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 8 9 5 e5 9 0 4
the scale model, it was installed in the 1.2 1.2 m (4 4 ft) and 1.8 0.9 m (6 3 ft) sumps for prototype-size testing of both sediment removal efficiency and washout. The results of the full scale testing showed that the SAFL Baffle can improve sediment capture by 10e15% and decrease effluent concentrations from 800 mg/L to a maximum of 50 mg/ L for sediment washout. Removal efficiency and washout functions were developed from the measured values using the dimensionless parameter Pe=Fr2j , i.e. the ratio of Pe´clet number to the square of the inflow jet Froude number. The performance functions can be used for the prediction of removal efficiencies of standard sumps retrofitted with the SAFL Baffle and aid in the design of sumps including a SAFL Baffle.
Acknowledgments This project was funded by the Minnesota Department of Transportation (MnDOT). Barbara Loida was the Technical Liaison for MnDOT. We would like to thank Benjamin Plante, Patrick Brokamp, Teigan Gulliver, Kurt McIntire, Andrew Sander, Mike Plante, Andrew Fyten, Matthew Lueker, and Benjamin Erickson for assisting with the construction of the experimental setup, data collection and laboratory analysis of sediment samples.
references
Carlson, L., Mohseni, O., Stefan, H., Lueker, M., 2006. Performance Evaluation of the BaySaver Stormwater Separation System. St. Anthony Falls Laboratory Project Report No. 472. University of Minnesota, Minneapolis, MN. Dhamotharan, S., Gulliver, J.S., Stefan, H.G., 1981. Unsteady onedimensional settling of suspended sediment. Water Resources Research 17 (4), 1125e1132.
Howard, A.K., Mohseni, O., Gulliver, J.S., Stefan, H.G., 2011. Mn/ DOT Research Services Interim Report No. 2011-08. Assessment and Recommendations for Operation of Standard Sumps as Best Management Practices for Stormwater Treatment, vol. 1. Minnesota Department of Transportation (Mn/DOT), St. Paul, MN. Kim, J., Pathapati, S., Liu B., Sansalone J., 2007. Treatment and Maintenance of Stormwater Hydrodynamic Separators: A Case Study. Proceedings of the 9th Biennial Conference on Stormwater Research and Watershed Management, Orlando, FL. Mohseni, O., Fyten, A., 2007. Performance Assessment of Modified EcoStorm Hydrodynamic Separator. St. Anthony Falls Laboratory Project Report No. 495B. University of Minnesota, Minneapolis, MN. Mohseni, O., Kieffer, J.M., Koehler, J.A., 2009. A Tool for the Performance Assessment of Hydrodynamic Separators. Proceedings of World Environmental and Water Resources Congress. EWRI, ASCE, Kansas City, Missouri. National Cooperative Highway Research Program (NCHRP), 2006. Evaluation of Best Management Practices for Highway Runoff Control. Report No. 565. Transportation Research Board, Washington, D.C. 89e96. Saddoris, D.A., McIntire, K.D., Mohseni, O., Gulliver, J.S., 2010. Hydrodynamic Sediment Retention Testing. Mn/DOT Research Services Report No. 2010-10. Minnesota Department of Transportation (Mn/DOT), St. Paul, MN. Silberman, E., 1947. The Pitot Cylinder. St. Anthony Falls Hydraulic Laboratory, Circular No. 2. University of Minnesota, Minneapolis, MN. Smith, E., 2001. Pollutant concentrations of stormwater and captured sediment in flood control sumps draining an urban watershed. Water Research 35 (13), 3117e3126. Wilson, M.A., Gulliver, J.S., Mohseni, O., Hozalski, R.M., 2007. Performance Assessment of Underground Stormwater Devices. SAFL Project Report No. 494. St. Anthony Falls Laboratory, University of Minnesota, Minneapolis, MN. Wilson, M.A., Mohseni, O., Gulliver, J.S., Hozalski, R.M., Stefan, H., 2009. Assessment of hydrodynamic separators for stormwater treatment. Journal of Hydraulic Engineering 131 (5), 383e392.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 0 5 e5 9 1 5
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Facilitated transport of Cu with hydroxyapatite nanoparticles in saturated sand: Effects of solution ionic strength and composition Dengjun Wang a,g, Marcos Paradelo b, Scott A. Bradford c, Willie J.G.M. Peijnenburg d,e, Lingyang Chu a,f, Dongmei Zhou a,* a
Key Laboratory of Soil Environment and Pollution Remediation, Institute of Soil Science, Chinese Academy of Sciences, Nanjing 210008, PR China b Soil Science Group, Department of Plant Biology and Soil Science, Faculty of Sciences, University of Vigo, Ourense E-32004, Spain c US Salinity Laboratory, Agricultural Research Service, United States Department of Agriculture, 450 W. Big Springs Road, Riverside, CA 92507, USA d Laboratory of Ecological Risk Assessment, National Institute of Public Health and the Environment, P.O. Box 1, 3720 BA Bilthoven, The Netherlands e Institute of Environmental Sciences (CML), Leiden University, P.O. Box 9518, 2300 RA Leiden, The Netherlands f College of Resources and Environmental Sciences, Anhui Agricultural University, Hefei, 230036, PR China g Graduate School of the Chinese Academy of Sciences, Beijing 100049, PR China
article info
abstract
Article history:
Column experiments were conducted to investigate the facilitated transport of Cu in
Received 2 May 2011
association with hydroxyapatite nanoparticles (nHAP) in water-saturated quartz sand at
Received in revised form
different solution concentrations of NaCl (0e100 mM) or CaCl2 (0.1e1.0 mM). The experi-
23 August 2011
mental breakthrough curves and retention profiles of nHAP were well described using
Accepted 24 August 2011
a mathematical model that accounted for two kinetic retention sites. The retention coef-
Available online 16 September 2011
ficients for both sites increased with the ionic strength (IS) of a particular salt. However, the amount of nHAP retention was more sensitive to increases in the concentration of divalent
Keywords:
Ca2þ than monovalent Naþ. The effluent concentration of Cu that was associated with
Hydroxyapatite nanoparticles (nHAP)
nHAP decreased significantly from 2.62 to 0.17 mg L1 when NaCl increased from 0 to
Cu
100 mM, and from 1.58 to 0.16 mg L1 when CaCl2 increased from 0.1 to 1.0 mM. These
nHAP-facilitated Cu (nHAP-F Cu)
trends were due to enhanced retention of nHAP with changes in IS and ionic composition
Co-transport
(IC) due to compression of the double layer thickness and reduction of the magnitude of the
Ionic strength (IS)
zeta potentials. Results indicate that the IS and IC had a strong influence on the co-
Ionic composition (IC)
transport behavior of contaminants with nHAP nanoparticles. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
Nanotechnology focuses on the investigation and application of materials with at least one characteristic dimension less than
100 nm. Properties of nanomaterials such as small size, high surface area per unit volume and great reactivity make them a highly promising class of materials for a variety of potential applications. For example, hydroxyapatite nanoparticles
* Corresponding author. Tel./fax: þ86 25 86881180. E-mail address:
[email protected] (D. Zhou). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.08.041
5906
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 0 5 e5 9 1 5
(Ca10(PO4)6(OH)2, nHAP), which are the main component of hard tissues of vertebrates such as bones and teeth, have been widely applied for the remediation of contaminated soil and purification of wastewaters polluted by metal ions and actinides (Cu2þ, Pb2þ, Cd2þ, Co2þ, and Sr2þ) because of their strong ability to fix them (Ma et al., 1994; Smiciklas et al., 2006; Handley-Sidhu et al., 2011). The fixation of metal ions on nHAP may take place through one or more mechanisms, including: ion exchange, surface complexation, and dissolution of nHAP to form new metal phosphates (Ma et al., 1994; Smiciklas et al., 2006; Handley-Sidhu et al., 2011). To date, little attention has been paid to potential environmental risks of using nHAP to remediate soils contaminated by heavy metals. Specifically, nHAP may alter the transport and fate of common environmental metal contaminants, such as Cu, by dramatically affecting their distribution among mobile and immobile phases (Wang et al., 2011). The transport potential of many chemicals is known to be greatly enhanced when they are associated with mobile colloids (McCarthy and Zachara, 1989; de Jonge et al., 2004; Simunek et al., 2006; Bradford and Kim, 2010). This process is often referred to as “colloid-facilitated contaminant transport” (Grolimund et al., 1996; Roy and Dzombak, 1997). Colloidfacilitated contaminant transport in subsurface environments has attracted considerable attention in recent years, especially for nanoscale colloids such as buckminsterfullerene (Zhang et al., 2011) and TiO2 (Fang et al., 2011). However, essentially no published data are available on how nHAP might mediate the transport and fate of Cu in subsurface environments. The solution ionic strength (IS) and composition (IC) are known to have a large influence on the transport behavior of colloid-associated contaminants. For example, Cheng and Saiers (2010) reported that the capacity of sediment-colloids to bind 137Cs decreased with increasing IS (Naþ cation), leading to a decrease of the mass of 137Cs eluted from columns packed with Hanford coarse sand. Walshe et al. (2010) found that increasing the IS (Ca2þ cation) of the bulk solution reduced peak concentrations for both kaolinite and kaolinitefacilitated MS2 coliphage from columns composed of gravel aquifer media. The influence of IS and IC on colloid and nanoparticles interactions with solid surfaces is typically explained using theory developed by DerjaguineLandaue VerweyeOverbeek (Derjaguin and Landau, 1941; Verwey and Overbeek, 1948). This theory predicts that increasing the solution IS tends to decrease the double layer thickness and magnitude of the surface charge, and thereby increase colloid retention. Divalent ions have a greater effect on these properties than monovalent ions (Israelachvili, 1992; Elimelech et al., 1995). However, up till now there has been no systemic investigation concerning the effects of IS on the co-transport behavior of Cu with nHAP. Furthermore, divalent Ca2þ can compete with Cu for ion-exchange sites of nHAP, leading to enhance dissociation of Cu2þ from nHAP, and thus altering the co-transport behavior of Cu with nHAP. It is therefore crucial to investigate the impact of the divalent Ca2þ on the co-transport behavior of Cu with nHAP in saturated packed columns. The overall objective of this study was to systemically investigate the effects of solution IS and IC on the co-transport of Cu with nHAP.
2.
Materials and methods
2.1.
Quartz sand
Quartz sand (Sinopharm Chemical Reagent Co., Ltd., China) was used as column packing material. The grain size distribution of the sand was determined by sieve analysis. The median grain size (d50) of the sand was 600 mm, and the coefficient of uniformity (Ui ¼ d60/d10 where x% of the mass was finer than dx) was 1.3. Prior to use, the sand was cleaned thoroughly by the procedure described elsewhere (Zhou et al., 2011) to remove any metal oxide and absorbed clay on the sand surface. The z-potential of the quartz colloid was measured by the method described in our previous work (Wang et al., 2011; Zhou et al., 2011).
2.2.
nHAP
The nHAP used in this study (purity >99.9%) was purchased from Aipurui Nanomaterial Company (Nanjing, China). The physicochemical properties of the nHAP were determined in our previous work (Wang et al., 2011). Briefly, the nHAP particles are rod-shaped and their mean size is 20 nm in width and 100 nm in length, the Ca/P molar ratio is 1.65 and the specific surface area of the nHAP is 154 m2 g1.
2.3.
Co-transport tests
Cleaned quartz sand was dry-packed into a glass chromatographic column (20 cm 2.6 cm, Shanghai, China) with 80 mm stainless-steel screens on both ends. Each column contained approximately 150 g of quartz sand and had an average length of 20 cm. To achieve uniform packing, the sand was carefully added to the column using a spatula and then gently vibrated. The column was slowly saturated by pumping ultrapure water (18.2 MU, Millipore, Inc., USA) in the upward direction at an approach velocity of 0.1 cm min1 for 1 h. Following this saturated step, the water content of each column was determined gravimetrically. The porosity of the packed columns varied between 0.39 and 0.41. Table 1 shows the most salient properties of the packed columns used in the co-transport tests of nHAP. The longitudinal dispersivity of the packed column was estimated from transport tests using bromide as tracer.
Table 1 e Properties and parameters of the nHAP and packed quartz sand columns used in this study. nHAP length (L) nm nHAP width (L) nm nHAP density (rn) g cm3 Collector diameter (dc) mm Fluid density (rf) kg m3 Fluid viscosity (m) kg m1 s1 Temperature (T ) K Hamaker constant (A) J Porosity ( f ) cm3 cm3 Column length (L) cm Column diameter (L) cm Happel model parameter (As)
100 20 3.2 0.6 103 8.9 104 293 1.0 1020 0.39e0.41 20.0 2.6 36.1e41.2
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 0 5 e5 9 1 5
To determine the effects of IS and IC on the co-transport behavior of Cu with nHAP, Cu-bearing nHAP suspensions at varying ISs and ICs were prepared as follows. First, 0.10 g of nHAP powder and 50 mL ultrapure water were added to a 500 mL volumetric flask and shaken vigorously for 1 min to disperse the nHAP. Then, 50 mL of 1.0 mM Cu(NO3)2 (analytical grade) and 400 mL of different concentrations of NaCl or CaCl2 solution were added. The final volume was 500 mL, and the concentrations of nHAP and Cu in suspension were 200 mg L1 and 0.1 mM, respectively. The electrolyte concentrations added to the suspension were 0, 1, 10, 50, and 100 mM of NaCl or 0.1, 0.3, 0.5, and 1.0 mM of CaCl2. The nHAP suspensions were sonicated for 30 min before use. The z-potentials of the Cu-bearing nHAP suspensions were measured using a microelectrophoresis instrument (JS94 G, Zhongcheng Digital Technology Co., Ltd., Shanghai, China).
2.4.
Experimental design
Packed columns were initially equilibrated by flushing several pore volumes (PVs) of ultrapure water and at least 5 PVs of the colloid-free background electrolyte solution in order to establish steady state flow and to standardize the chemical conditions. Experiments was conducted in the following steps: (1) phase 1, the Cu-bearing nHAP suspensions in solutions of varying IS and IC (as described in Section 2.3) were gently stirred while applied to the bottom end of the column via a peristaltic pump (YZP-15, Baoding Longer Precision Pump Co., Ltd., Hebei, China) at a constant approach velocity (0.44 cm min1) for about 3.75 PVs, and (2) phase 2, several PVs of nHAP-free background electrolyte solution with the same pH and IS were pumped into the column to ensure that almost no colloidal particles were detected in the effluent. Column outflow was collected in 15-mL glass tubes at regular time intervals using a fraction collector (BS-100A, Huxi Analytical Instrument Factory Co., Ltd., Shanghai, China). All column experiments were conducted in duplicate. Following the completion of each co-transport test, the spatial distribution of nHAP retained in the column was determined. The fitting at the column end was removed, and the quartz sands were carefully excavated in 2 cm increments and transferred into ten 50 mL vials. Excess ultrapure water was added to fill the vials. In this low IS solution, highly negatively charged surfaces of both nHAP and quartz sand grains caused the release of retained nHAP from the sand surface. After 1 h, the vials containing the sand-nHAP solution were gently shaken to obtain a homogeneous concentration of nHAP in the supernatant. The concentration of nHAP in the excess aqueous solution was measured with a UV/vis spectrophotometer (see Section 2.5.1). The sand samples were then oven dried at 100 C overnight to obtain the dry weight of the solid. A mass balance was calculated for nHAP in the effluent and retained in the sand after normalizing by the total amount of nHAP injected into the column. After completion of the co-transport experiment with 1.0 mM of CaCl2, some of the quartz sand grains that were excavated from the inlet of the packed column were examined using a SEM-EDX (Scanning Electron Microscope-Energy Dispersive X-ray, SSX-550, Shimadzu) to determine mechanisms of nHAP retention in the sand columns.
2.5.
Analytical procedures
2.5.1.
Concentration of nHAP and Cu in the effluent
5907
The concentrations of nHAP in the outflow were determined with a UV/vis spectrophotometer (721-100, Jinghua Science and Technology Instrument Co., Ltd., Shanghai, China), at a wavelength of 300 nm (Wang et al., 2011). A calibration curve was constructed by diluting a 200 mg L1 suspension of nHAP. Spectrometer response versus nHAP concentration was linear in the range of 0e200 mg L1 with a coefficient of determination of R2 ¼ 0.999. The calculated molar extinction coefficient of nHAP was 4.0 103 M1 cm1 which was consistent with the result observed by Boussiba and Richmond (1979). The detection lower limit was 1.0 mg L1. The calibration curve was verified to be independent of the solution chemistry for our experimental conditions. Aliquots (5 mL) of all effluents were centrifuged at 170,000 g for 1 h (Optima TM L-80XP Ultracentrifuge, Beckman) and then filtered through a 0.22 mm membrane filter to determine the dissolved Cu. Five mL of 16 M HNO3 was added to other aliquots (5 mL) of all effluents to determine the total Cu concentration. The nHAP-facilitated (nHAP-F) Cu was calculated as the difference between total and dissolved Cu. The Cu concentration was determined using an Atomic Absorption Spectrophotometer (AAS, Hitachi Z-2000, Japan).
2.5.2.
Size of Cu-bearing nHAP colloid
The average Cu-bearing nHAP aggregate size and the intrinsic size distributions in various suspensions were measured using dynamic light scattering (DLS) (BI-200 SM, Brookhaven Instrument Corp., USA) at 25 C. The CONTIN algorithm was used to convert intensity autocorrelation functions to an intensity-weighted, Cu-bearing nHAP aggregate, hydrodynamic diameter distribution based on the StokeseEinstein equation (Stankovich et al., 2006). The intensity-weighted distributions were further converted to number-weighted size distribution, using 1.56 as the refractive index for nHAP (Onuma et al., 2000).
2.5.3.
Data analysis
The transport of nHAP through the packed columns was described using a one-dimensional form of the convectionedispersion equation with two kinetic retention sites as (Schijven and Simunek, 2002; Bradford et al., 2003): vqc vðs1 Þ vðs2 Þ v vc vqc þ rb þ rb ¼ qD vt vt vt vx vx vx
(1)
where q is the volumetric water content [e], c is the Cu-bearing nHAP colloid concentration in the aqueous phase [N L3, where N and L denote number and length, respectively], rb is the bulk density of the porous matrix [M L3, where M denotes mass], t is the time [T], x is the vertical spatial coordinate [L], D is the hydrodynamic dispersion coefficient [L2 T1], q is the flow rate [L T1], and s1 [N M1] and s2 [N M1] are the solid phase concentrations associated with retention sites 1 and 2, respectively. The two kinetic retention sites described mass transfer of nHAP between the aqueous and solid phase. The first kinetic
5908
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 0 5 e5 9 1 5
site (site 1, Eq. (2)) assumes reversible retention, whereas the second kinetic site (site 2, Eq. (3)) assumes irreversible, depthdependent retention as: rb
vðs1 Þ ¼ qk1 c rb k1d s1 vt
(2)
rb
vðs2 Þ ¼ qk2 jx c vt
(3)
Breakthrough curves (BTCs) and retention profiles, obtained from the experiments above, were analyzed using the HYDRUS-1D code (Simunek et al., 2008), which allowed us to fit the nHAP transport parameters (k1, k1d, and k2) using a nonlinear least square optimization routine based on the LevenbergeMarquardt algorithm (Marquardt, 1963). The approach velocity and dispersivity that were used in the Hydrus-1D code were obtained by fitting the solution of the convective dispersion equation for the tracer (bromide) breakthrough curves.
where k1 [T1] and k2 [T1] are first-order retention coefficients on site 1 and 2, respectively, k1d [T1] is the first-order detachment coefficient, and jx [e] is a dimensionless function to account for depth-dependent retention. The value of jx is given as (Bradford et al., 2003):
jx ¼
b dc þ x x0 dc
3.
Results and discussion
3.1.
Properties of Cu-bearing nHAP and collector surface
(4) Electrokinetic potential (z-potential) will affect the retention and transport behavior of Cu-bearing nHAP. The measured zpotentials of the Cu-bearing nHAP and quartz sand (colloidal fragments) in solutions of varying IS and IC became less negative as the electrolyte concentration of the bulk solution increased (Table 2). This is due to the compression of the electrostatic double layer (Elimelech et al., 1995). With the monovalent Naþ as the background electrolyte, the z-potential of Cu-bearing nHAP and quartz colloids decreased from 79.6 to 24.7 mV and from 51.2 to 20.0 mV, respectively, when the IS was increased from 0 to 100 mM. In the case of the divalent Ca2þ, the z-potential of Cu-bearing nHAP declined from 34.9 to 22.1 mV when the Ca2þ concentration in the bulk solution was increased from 0.10 to 1.0 mM. It is obvious that the divalent Ca2þ is significantly more effective at screening the surface charge of nHAP than the monovalent Naþ. Similar results have been observed elsewhere (Kim et al., 2010; Wang et al., 2011). Table 2 presents the average hydrodynamic sizes of 200 mg L1 Cu-bearing nHAP suspensions in the various IS and
where dc is the median diameter of the sand grains [L], x0 is the coordinate [L] of the location where the depth-dependent retention starts and b is an empirical factor controlling the shape of the spatial distribution. Bradford et al. (2003) found that a value of 0.432 provides an optimum for experiments in which significant depth-dependency occurred. The above approach assumes that retention near the column inlet is dominated by site 2, and away from the inlet by site 1. It should be mentioned that some researchers have attributed attachment/detachment and straining mechanisms of colloid retention to sites 1 and 2, respectively (e.g., Bradford et al., 2003; Gargiulo et al., 2007, 2008). However, additional microscopic information is frequently needed to substantiate this assumption. In this work we do not attempt to attribute specific nHAP retention mechanisms to a given site without other experimental evidence. Rather, Eqs. (1)e(4) are viewed as a simple and flexible approach to describe nHAP breakthrough curves and retention profiles that are not exponential with depth.
Table 2 e Properties of Cu-bearing nHAP suspensions at varying ISs and ICs used in transport tests. NaCl (mM) 0 1.0 10 50 100 0 0 0 0
CaCl2 (mM)
pH
0 0 0 0 0 0.10 0.30 0.50 1.0
5.7 5.8 5.9 5.8 5.8 5.9 5.8 5.7 5.7
Sizea (nm) 100 102 108 119 133 103 109 114 135
2 3 5 4 4 4 5 3 5
z-potential (mV) Quartz sand 79.6 71.0 50.2 35.8 24.7 47.0 36.2 31.7 26.0
1.2 1.0 0.8 0.9 1.2 1.4 0.7 0.9 1.1
b
nHAP colloid 53.0 35.2 31.4 25.8 20.0 34.9 31.5 28.4 22.1
2.8 0.7 0.8 1.2 1.3 1.2 0.9 0.5 1.2
a The average particle hydrodynamic size, determined by DLS measurement. b z-potential of the Cu-bearing nHAP colloid. c The maximum energy barrier, Fmax, calculated by DLVO theory. d The secondary minimum, Fmin2, calculated by DLVO theory. e Separation distance from colloid surface to secondary minimum. f Total Cu concentration of the Cu-bearing nHAP suspension. g Dissolved Cu concentration in the bulk solution. h Not applicable.
Fmaxc (kT )
Fmin2d (kT )
he (nm)
Total Cuf (mg L1)
Dissolved Cug (mg L1)
NAh 160 121 82 44 138 108 54 48
NA 1.6 4.1 7.7 11 1.3 2.1 2.7 4.4
NA 48 16 7 4 58 37 28 19
6.65 6.56 6.56 6.65 6.60 6.61 6.65 6.62 6.59
2.42 2.44 2.43 2.46 2.45 2.38 2.57 2.73 2.96
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 0 5 e5 9 1 5
5909
IC solutions. The nHAP size increased from 100 to 133 nm as the concentration of NaCl increased from 0 to 100 mM, and from 103 to 135 nm as the concentration of CaCl2 increased from 0.1 to 1.0 mM of CaCl2. Similar results have been observed by Chen and Elimelech (2006, 2007). These nHAP sizes were used for the interaction energy calculations between nHAP and quartz sand provided in Table 2. Both Cubearing nHAP and quartz were negatively charged for all the considered IS and IC. Consequently, electrostatic repulsion occurred between Cu-bearing nHAP and quartz (energy barrier ranged from 44 to 160), even for the highest electrolyte concentration used in the column transport experiments. However, the theoretical prediction of the DLVO interaction energy for nHAP upon their approach to the collector surface shows the presence of a secondary minimum (Table 2) that ranged from 1.3 to 11 depending on the IS and IC. The height of the primary energy barrier decreases and the depth of the secondary minimum increases with IS. This behavior is due to compression of the diffuse double layer thickness with increasing IS according to standard PoissoneBoltzmann
Results of Cu-bearing nHAP transport studies with various concentrations of NaCl (0, 1, 10, 50, and 100 mM) and CaCl2 (0.1, 0.3, 0.5, and 1.0 mM) are shown in Figs. 1 and 2, respectively. Breakthrough curves are plotted in Figs. 1 and 2a as the relative effluent concentration (Ci/C0, where Ci and C0 are the effluent and influent concentration of Cu-bearing nHAP, respectively) as a function of PVs. Retention profiles are plotted in Figs. 1 and 2b as normalized concentration (quantity of the Cu-bearing nHAP recovered in the sand, Nt, divided by the quantity in a unit volume of the input nHAP suspension, N0) per gram of dry sand as a function of distance from the column inlet. The corresponding mass recovery of nHAP in the effluent and sand is shown in Table 3. Very good total mass balance was obtained for nHAP and Cu (90e96% and 87e95%, respectively), which provides a high degree of
Fig. 1 e Measured and fitted breakthrough curves (a) and retention profiles (b) for nHAP under the IS of 0, 1, 10, 50 and 100 mM NaCl, respectively. Fitted curves were obtained using the two-site kinetic retention model. In (a) the relative effluent concentration is plotted as a function of pore volume. In (b) the normalized concentration (quality of the nHAP recovered in the sand, Nt, is divided by the quality in a unit volume of the input nHAP suspension, N0) per gram of dry sand is plotted as a function of the distance from the column inlet.
Fig. 2 e Measured and fitted breakthrough curves (a) and retention profiles (b) for nHAP under the IC of 0.1, 0.3, 0.5 and 1.0 mM CaCl2, respectively. Fitted curves were obtained using the two-site kinetic retention. In (a) the relative effluent concentration is plotted as a function of pore volume. In (b) the normalized concentration (quality of the nHAP recovered in the sand, Nt, is divided by the quality in a unit volume of the input nHAP suspension, N0) per gram of dry sand is plotted as a function of the distance from the column inlet.
theory. Therefore, nHAP are likely to interact with the quartz surface in the secondary minimum, especially at higher IS.
3.2.
Co-transport behavior of Cu-bearing nHAP
5910
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 0 5 e5 9 1 5
Table 3 e Mass balance percentages for nHAP and Cu in the saturated packed column experiments at varying ISs and ICs used in this study. NaCl (mM)
CaCl2 (mM)
nHAP a
Effluent
Cu b
Retained
c
Total
Effluent nHAP-Fd
Retainedf
Totalg
15 20 36 44 47 29 36 42 45
93 91 89 94 93 87 92 95 94
Dise
% 0 1.0 10 50 100 0.10 0.30 0.50 1.0 a b c d e f g
89 80 68 44 11 74 68 64 26
7 15 28 50 79 20 27 32 65
96 95 96 94 90 94 95 96 91
39 33 15 12 8 21 17 10 3
39 38 38 38 38 37 39 43 46
refers to the effluent percentage of nHAP recovered from column experiments. refers to the retained percentage of nHAP recovered from column experiments. refers to the total percentage of nHAP recovered from column experiments. refers to the percentage of nHAP-facilitated (nHAP-F) Cu in the effluent. refers to the percentage of dissolved Cu recovered in the effluent. refers to the retained percentage of Cu recovered from column experiments. refers to the total percentage of Cu recovered from column experiments.
confidence in our experimental procedures. Each co-transport experiment was duplicated and the replicated experiments exhibited similar transport and retention behavior for given conditions (data not shown). Significant nHAP retention occurred despite the unfavorable electrochemical conditions for attachments in the primary minimum predicted by DLVO theory (Table 2). In general, an increase in the IS of the bulk solution resulted in increasing retention of the Cu-bearing nHAP onto the quartz grains for both monovalent and divalent cations. In particular, the percentage of nHAP mass that was recovered in the effluent decreased from 89 to 11% when the IS was increased from 0 to 100 mM of NaCl, and from 74 to 26% as the electrolyte concentration was raised from 0.1 to 1.0 mM of CaCl2, respectively. Greater retention of Cu-bearing nHAP at higher electrolyte concentration can be explained in part by a reduction in the thickness of the double layer and a corresponding increase in the depth of the secondary minimum (Table 2). This finding is in accordance with previously reported literature (Elimelech et al., 1995; Walshe et al., 2010; Wang et al., 2011). The retention profiles of Cu-bearing nHAP typically exhibited a hyperexponential shape with greater retention in the section adjacent to the column inlet (0e4 cm) and rapidly decreasing retention with depth (Figs. 1 and 2b). The shape of the retention profiles was not consistent with classical filtration theory (Yao et al., 1971) which predicts an exponential shape with depth. Hyperexponential profiles were more pronounced at higher electrolyte concentrations. For example, at IS ¼ 0 about 40% of the total retained Cu-bearing nHAP were near the column inlet (<4 cm), whereas the value was as high as 65% at an IS of 100 mM NaCl (the same trend was observed for the Ca2þ). A number of potential explanations for hyperexponential retention profiles have been proposed in the literature including: chemical heterogeneity of colloids (Li
et al., 2004; Tufenkji and Elimelech, 2005; Johnson and Li, 2005), straining of colloids (Bradford et al., 2002, 2003), secondary minimum and surface charge heterogeneity on the sand grain (Redman et al., 2004; Tufenkji and Elimelech, 2005; Johnson et al., 2010), surface roughness (Shellenberger and Logan, 2002; Yoon et al., 2006), colloid aggregation (Chen and Elimelech, 2006, 2007), and enhanced colloid retention in low velocity regions (Torkzaban et al., 2007, 2008; Bradford et al., 2009). These potential explanations will be discussed in greater detail in Section 3.4. Figs. 1 and 2 present the simulated nHAP breakthrough curves and retention profiles for the various IS and IC conditions. Values of optimized model parameters are summarized in Table 4. The two-site kinetic retention model provides a good description of both the breakthrough curves and the retention profiles (see R2 in Table 4). Values of k1 and k2 both increased with the electrolyte concentration of NaCl and CaCl2, suggesting that the greater retention of nHAP was related to the depth of the secondary minimum (Tufenkji and Elimelech, 2005; Bradford et al., 2009). The percentage of retained nHAP in sites 1 and 2 were relatively constant with IS and IC, ranging from 30 to 37% on site 1 and 63e70% on site 2. The value of k1d also increased with electrolyte concentration and was greater than or of a similar magnitude as k1. Similar results have been observed elsewhere (Gargiulo et al., 2007, 2008).
3.3.
Effects of IS and IC on nHAP-F Cu transport
The total and dissolved Cu concentrations of suspensions were measured before conducting the co-transport experiments. Table 2 indicates that the suspension pH and total Cu concentration varied from 5.7 to 5.9 and from 6.56 to 6.65 mg L1, respectively. The dissolved Cu concentration was little changed (2.42e2.46 mg L1) with IS when NaCl was the background electrolyte. However, the dissolved Cu
5911
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 0 5 e5 9 1 5
Table 4 e Fitted parameters of the two-kinetic attachment model as estimated from the breakthrough data for saturated packed quartz sand at varying ISs and ICs used in transport tests.a Parab
NaCl (mM)
1
k1 (min ) k1d (min1) k2 (min1) Mass site 1 (%) Mass site 2 (%) b R2
CaCl2 (mM)
0
1
10
50
100
0.1
0.3
0.5
1.0
6.92e-03 (1.34e-03) 2.98e-02 (6.17e-03) 4.82e-02 (1.78e-03) 2.4 4.6 0.432 0.994
9.05e-03 (1.79e-03) 3.88e-02 (8.06e-03) 5.52e-02 (1.90e-03) 5.6 9.4 0.432 0.991
9.85e-03 (3.32e-03) 4.15e-02 (1.44e-02) 7.87e-02 (2.99e-03) 9.9 18.1 0.432 0.979
1.60e-02 (6.52e-03) 4.05e-02 (2.23e-02) 1.52e-01 (7.78e-03) 17.7 32.3 0.432 0.911
5.57e-02 (2.81e-01) 2.42e-01 (1.32eþ00) 4.91e-01 (3.27e-02) 24.1 54.9 0.432 0.942
7.40e-03 (1.65e-03) 3.68e-02 (8.82e-03) 7.60e-02 (2.12e-03) 6.4 13.6 0.432 0.990
8.21e-03 (5.70e-03) 4.14e-02 (2.99e-02) 8.7e-02 (5.07e-03) 8.5 18.5 0.432 0.949
1.04e-02 (5.67e-03) 4.47e-02 (2.53e-02) 9.70e-02 (4.93e-03) 10.9 21.1 0.432 0.951
4.87e-02 (6.57e-02) 3.47e-01 (5.32e-01) 2.55e-01 (9.90e-03) 19.8 45.2 0.432 0.951
a 95% confident intervals on fitted parameters are shown in parentheses. b Parameters: k1, the first-order retention coefficient on site 1; k1d, the first-order detachment coefficient on site 1; k2, the first-order detachment coefficient on site 2; Mass sites 1 and 2 are percentages of mass of Cu-bearing nHAP deposited on site 1 and 2, respectively; b, empirical factor controlling the shape of the spatial distribution; R2, Person’s squared correlation coefficient.
concentration changed from 2.38 to 2.96 mg L1 when the electrolyte concentration was increased from 0.1 to 1.0 mM of CaCl2. This observation suggests that the divalent Ca2þ competes with Cu2þ for ion-exchange sites on nHAP. Even through the ion selectivity coefficients favor the adsorption of Cu2þ over Ca2þ (Bakker, 1997), the larger concentration (4e40 mg L1) of Ca2þ displaces some of the Cu2þ on nHAP exchange sites. Fig. 3 shows representative breakthrough curves for nHAP-F Cu under the various IS and IC conditions. Corresponding mass balance information for the dissolved and nHAP-F Cu in the column effluents is provided in Table 3. The solution chemistry had a marked effect on the co-transport of Cu with nHAP. Increasing the IS of the solution reduced the peak effluent concentration and total mass of nHAP-F Cu. The peak effluent concentrations of nHAP-F Cu diminished from 2.62 to 0.17 mg L1 as the concentration of NaCl increased from 0 to 100 mM (Fig. 3a) and from 1.58 to 0.16 mg L1 as the concentration of CaCl2 increased from 0.1 to 1.0 mM (Fig. 3b). The nHAP-F Cu accounted for about 39 and 8% of the total Cu in the effluent at an IS of 0 and 100 mM, respectively. Similarly, nHAP-F Cu decreased from 21 to 3% of the total Cu in
a 3.0
Phase 1
-1
nHAP-F Cu (mg L )
b 1.8
Phase 2 0 mM NaCl 1 mM NaCl 10 mM NaCl 50 mM NaCl 100 mM NaCl
2.5 2.0
the effluent when the concentration of CaCl2 increased from 0.1 to 1.0 mM (Table 3). Qualitatively similar relationships of colloid-facilitated contaminant transport with IS and IC have been reported elsewhere (Cheng and Saiers, 2010; Walshe et al., 2010). The decline in Cu transport with increasing IS reflects the influences of IS on Cu-bearing nHAP mobility discussed in the previous section and on the partitioning of Cu between nHAP and dissolved phases. As mentioned above, divalent Ca2þ competes with Cu2þ for ion-exchange sites on nHAP and this leads to enhanced dissociation of Cu into bulk solution, whereas dissolved Cu was relatively unaffected with increasing concentration of NaCl (Table 3). Thus, decreases in the suspended load (increased retention of Cu-bearing nHAP) combined with decreases in the affinity of Cu for nHAP (increased ion exchange) accounted for the observed decline in nHAP-F Cu mobilization with increasing IS. It should be noted that dissolved Cu actually accounted for a greater percentage of the total Cu in the effluent than nHAP-F Cu, especially as the concentration of NaCl and CaCl2 increased in solution (Table 3). This observation indicates that nHAP tended to inhibit Cu transport because of retention of nHAP-F Cu.
Phase 2
Phase 1
0.1 mM CaCl2 0.3 mM CaCl2 0.5 mM CaCl2 1.0 mM CaCl2
1.5 1.2
1.5
0.9
1.0
0.6
0.5
0.3 0.0
0.0 0
1
2
3
4 PV
5
6
7
0
1
2
3
4 PV
5
6
7
8
Fig. 3 e Representative breakthrough curves for nHAP-F Cu as a function of PV under different concentrations of NaCl (a) and CaCl2 (b) of bulk solution, respectively. Note that the concentration of nHAP-F Cu was calculated as the difference between dissolved and total Cu of Cu-bearing nHAP suspension.
5912
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 0 5 e5 9 1 5
3.4. Mechanisms influencing nHAP retention and nHAPF Cu transport The above information clearly indicates that factors that influence the retention of nHAP will have a large impact on the facilitated transport of Cu. Additional experiments were therefore conducted to better understand and quantify the mechanisms of nHAP retention. The nHAP transport experiments shown in Figs. 1 and 2 indicate that increasing retention occurred with increasing NaCl and CaCl2 concentrations. Under chemically unfavorable attachment conditions (electrostatic repulsion), nHAP may interact with the solid phase by virtue of the presence of the reversible secondary minimum at a small separation distance (Franchi and O’Melia, 2003; Redman et al., 2004; Tufenkji and Elimelech, 2005). In order to further test this hypothesis, we ran several three phase transport experiments using 200 mg L1 of nHAP and 0.1 mM of Cu. Phases 1 and 2 were the same as for other experiments described above, whereas phase 3 consisted of flushing the column with several pore volumes of ultrapure water (IS ¼ 0) to eliminate the secondary minimum (Franchi and O’Melia, 2003). Release of previously deposited colloids through lowering the bulk solution IS has been used as supportive evidence for particle deposition in the secondary minimum (Franchi and O’Melia, 2003; Redman et al., 2004; Tufenkji and Elimelech, 2005). The breakthrough curves for the three phase experiments at the electrolyte concentrations of 100 mM of NaCl and 1.0 mM of CaCl2 are shown in Fig. 4a and b, respectively. The transport and retention of nHAP during phases 1 and 2 was similar to that previously discussed for Figs. 1 and 2 (Table 3). In Fig. 4a (100 mM NaCl during phases 1 and 2) rinsing the column with ultrapure water during phase 3 resulted in a sharp peak (Ci/C0 ¼ 1.7) of released nHAP that accounted for approximately 35% of the previously retained mass. In contrast, in Fig. 4b (1.0 mM CaCl2 during phases 1 and 2) rinsing the column with ultrapure water during phase 3 resulted in a minor peak (Ci/C0 ¼ 0.04) and release of retained nHAP in the effluent. These observations suggest that the secondary minimum was likely involved in the nHAP retention for both 100 mM NaCl and 1.0 mM CaCl2 systems, but that it cannot account for most of the retained nHAP mass.
Ci/C0
a
2.0 1.8 1.6 1.4
Phase 1
Phase 2
The z-potential of nHAP suspensions at IS ¼ 0 was measured before and after filtering them through a 1 mm pore diameter glass fiber filter in order to test the hypothesis of nHAP surface charge heterogeneity. The value of the nHAP z-potential increased from 53.0 to 60.4 mV before and after filtering at IS ¼ 0. Li et al. (2004) reported that such a variation in the z-potential was sufficient to cause hyperexponential retention profiles. Specifically, the fraction of the colloid population with higher (less negative) z-potentials is assumed to preferentially attach at the column inlet, whereas the remaining particles experience a slower attachment rate and greater transport potential. Straining of colloids at grainegrain contacts has been considered to be insignificant when the diameter ratio of colloid to grain is less than 0.005 (Bradford et al., 2003; Johnson et al., 2007). In this study, the diameter ratio of Cu-bearing nHAP aggregates to sand was more than an order of magnitude lower than this threshold. Straining of nHAP aggregates at grainegrain contacts was therefore not considered to be important. Surface roughness has been demonstrated to play a critical role in colloid deposition (Shellenberger and Logan, 2002; Yoon et al., 2006; Torkzaban et al., 2010). Similar to graine grain contacts, surface roughness locations are low velocity regions that are associated with lower hydrodynamic forces and torques and flow vortices (Vaidyanathan and Tien, 1988; Taneda, 1979), and therefore enhanced colloid retention under unfavorable attachment conditions. In particular, when the colloid radius is less than the roughness height, the applied hydrodynamic torque will be zero and create locations that are hydrodynamic favorable for colloid deposition even for an IS of zero (Vaidyanathan and Tien, 1988). In addition, Torkzaban et al. (2010) demonstrated that nanoscale surface roughness can create locally favorable conditions for nanoscale colloid deposition. Fig. 5a shows a SEM image of a cleaned quartz sand grain. This grain possesses significant surface roughness of different sizes, ranging from several nanometers to dozens of microns. Considering that the length of nHAP is 100 nm, these rough patches can likely create local regions that are hydrodynamically favorable for nHAP retention. Fig. 5b shows an SEM image of quartz sand excavated from the column inlet after
b 1.0
Phase 3
100 mM NaCl
Phase 3
0.8
0.6
0.6
0.4
0.4
0.2
0.2
0.0
Phase 2
Phase 1
1.0 mM CaCl2
0.0 0
1
2
3
4
5 6 PV
7
8
9 10 11
0
1
2
3
4
5 6 PV
7
8
9 10 11
Fig. 4 e The three phase breakthrough curves for Cu-bearing nHAP under electrolyte concentration of (a) 100 mM of NaCl, and (b) 1.0 mM of CaCl2, respectively. Phase 1: Cu-bearing nHAP suspension, phase 2: eluted with background electrolyte solution, and phase 3: further eluted with ultrapure water.
5913
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 0 5 e5 9 1 5
a
c 35
Si
b
Counts (cps)
30 25 20
P
15
O
10 5
Ca Cu
C
Au
Au
0 0
1
2
3
4
5
6
7
8
9
10
11
12
Energy (keV)
Fig. 5 e SEM micrographs of cleaned quartz sand (a) and the quartz sand excavated from the inlet of the packed column (b) after completion of the Cu-bearing nHAP transport test using 1.0 mM of CaCl2. On the right (c) is the corresponding EDX spectrum of the retained nHAP on the quartz sand marked in the box.
completion of the transport test at bulk solution IS of 1.0 mM of CaCl2. Fig. 5b demonstrates that large spherical-like aggregates were retained at surface roughness locations. Furthermore, the aggregates were not uniformly distributed on the grain surface as would be expected for secondary and/ or primary minimum attachment upon collision with the sand, even though a large secondary minimum of 4.4 occurred for this system. The result of an EDX chemical analysis of these aggregates is presented in Fig. 5c. The presence of Ca, P and O elements are consistent with the basic components of nHAP, and the Cu peak confirms that the large aggregates were composed of Cu-bearing nHAP. Neither Fe nor Al oxides were present on the surfaces of the quartz grain, suggesting low amounts of chemical heterogeneity. To further test the hypothesis that nHAP aggregation influenced transport and retention, the hydrodynamic diameter distribution of nHAP in the effluent was determined by DLS measurement after completion of the co-transport experiment using 1.0 mM of CaCl2. The hydrodynamic diameter distribution ranged from 150 to 250 nm, with a mean diameter of 210 5 nm. This size distribution is much larger than that for the influent suspension of nHAP in 1.0 mM of NaCl solution (135 5 nm). The nHAP tended to form larger aggregates in the presence of divalent Ca2þ than monovalent Naþ likely because Ca2þ acted as a bridging agent (Wang et al., 2011). The above information indicates that the secondary minimum, nHAP surface charge heterogeneity, surface roughness, and aggregation are involved in the retention of nHAP. An explanation of the interrelation of these factors on nHAP retention is briefly discussed below. The probability that
colloids colliding with the grain will remain associated with the solid has been estimated from the depth of the secondary minimum (which is a function of IS and IC in Table 2) under unfavorable attachment conditions (Ryan and Elimelech, 1996; Simoni et al., 1998; Shen et al., 2007). However, this approach neglects the effects of hydrodynamics and pore space geometry on colloid retention as revealed by Torkzaban et al. (2008). Colloids that are weakly associated with the solid surface via the secondary minimum can be translated and/or funneled by fluid drag forces to such low velocity regions and “eddy zones” at locations associated with surface roughness or grainegrain contact points where they are immobilized (Kuznar and Elimelech, 2007; Torkzaban et al., 2008, 2010). A larger number of colloids may enter these low velocity regions at higher electrolyte concentrations due to the greater depth of the secondary minimum. In addition, aggregated nHAP experience a greater depth of the secondary minimum and hydrodynamic forces than smaller colloids (e.g., Bradford et al., 2011). Hence, the amount of nHAP retention (see k1 and k2 in Table 4) is expected to be greater with increasing electrolyte concentration and nHAP size (aggregation). Hyperexponential profiles will be influenced by nHAP surface charge heterogeneity (Li et al., 2004; Tufenkji and Elimelech, 2005) as well as the hydrodynamics to low velocity regions (Bradford et al., 2009) such as surface roughness locations.
4.
Conclusions
Facilitated transport of Cu by nHAP in water-saturated quartz sand was found to be a strong function of the solution IS and
5914
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 0 5 e5 9 1 5
IC. In particular, the concentration of nHAP-F Cu declined significantly from 2.62 to 0.17 mg L1 when NaCl increased from 0 to 100 mM, and from 1.58 to 0.16 mg L1 when CaCl2 increased from 0.1 to 1.0 mM. The total mass recovery of Cu in the effluent solution exhibited the same trend. This facilitated transport behavior was due to difference in the capacity of nHAP to bind Cu and the retention of nHAP with changes in IS and IC. The capacity of nHAP to bind Cu2þ decreased with increasing concentrations of Ca2þ as a result of ion exchange. More importantly, the amount of Cu-bearing nHAP that was retained in the sand increased with solution concentration of NaCl and especially CaCl2. Most of the Cu-bearing nHAP was retained close to the column inlet (0e4 cm), and the rate of retention rapidly decreased with depth. The experimental breakthrough curves and retention profiles of nHAP were well described using a mathematical model based on the CDE and two kinetic retention sites. Site 1 considered reversible retention, whereas site 2 was for irreversible, depth-dependent retention. The first-order retention coefficients on both site 1 (k1) and 2 (k2) increased with the NaCl and CaCl2 concentration due to increases in the depth of the secondary minimum. However, elimination of the secondary minimum after recovery of the breakthrough curves did not release the majority of the retained nHAP. SEM images in association with EDX measurements and DLS information demonstrated that nHAP aggregated occurred and that retention was strongly influenced by the grain surface roughness.
Acknowledgments The authors grateful acknowledge the support of the National Basic Research and Development Program (2007CB936604), the Knowledge Innovative project of Chinese Academy of Sciences (KZCX2-YW-Q02-02) and the open fund of the State Key Laboratory of Soil and Sustainable Agriculture (Y052010027). M. Paradelo is granted by the FPU program from the Spanish Ministry of Education. We are grateful to the critical review of Dr. J. Eugenio Lopez (University of Vigo) that led to revision and substantial improvement of this manuscript.
references
Bakker, E., 1997. Determination of unbiased selectivity coefficients of neutral-carrier based cation-selective electrodes. Analytical Chemistry 69 (6), 1061e1069. Boussiba, S., Richmond, A.E., 1979. Isolation and characterization of phycocyanins from the blue-green alga Spirulina platensis. Archives of Microbiology 120 (2), 155e159. Bradford, S.A., Kim, H., 2010. Implications of cation exchange on clay release and colloid-facilitated transport in porous media. Journal of Environmental Quality 39 (6), 2040e2046. Bradford, S.A., Yates, S.R., Bettahar, M., Simunek, J., 2002. Physical factors affecting the transport and fate of colloids in saturated porous media. Water Resource Research 38. doi:10.1029/ 2002WR001340. Bradford, S.A., Simunek, J., Bettahar, M., van Genuchten, M.T., Yates, S.R., 2003. Modeling colloid attachment, straining, and exclusion in saturated porous media. Environmental Science & Technology 37 (10), 2242e2250.
Bradford, S.A., Torkzaban, S., Leij, F., Simunek, J., van Genuchten, M.T., 2009. Modeling the coupled effects of pore space geometry and velocity on colloid transport and retention. Water Resources Research 45. doi:10.1029/ 2008WR007096. Bradford, S.A., Torkzaban, S., Wiegmann, A., 2011. Pore-scale simulations to determine the applied hydrodynamic torque and colloid immobilization. Vadose Zone Journal 10 (1), 252e261. Chen, K.L., Elimelech, M., 2006. Aggregation and deposition kinetics of fullerene (C-60) nanoparticles. Langmuir 22 (26), 10994e11001. Chen, K.L., Elimelech, M., 2007. Influence of humic acid on the aggregation kinetics of fullerene (C-60) nanoparticles in monovalent and divalent electrolyte solutions. Journal of Colloid and Interface Science 309 (1), 126e134. Cheng, T., Saiers, J.E., 2010. Colloid-facilitated transport of cesium in vadose-zone sediments: the importance of flow transients. Environmental Science & Technology 44 (19), 7443e7449. de Jonge, L.W., Kjaergaard, C., Moldrup, P., 2004. Colloids and colloid-facilitated transport of contaminants in soils: an introduction. Vadose Zone Journal 3 (2), 321e325. Derjaguin, B., Landau, L., 1941. Theory of the stability of strongly charged lyophobic sols and of the adhesion of strongly charged particles in solutions of electrolytes. Acta Physicochima USSR 14, 633e662. Elimelech, M., Gregory, J., Jia, X., Williams, R.A., 1995. Particle Deposition and Aggregation: Measurement, Modeling and Simulation. Butterworth-Heinemann Ltd., Oxford. Fang, J., Shan, X.Q., Wen, B., Lin, J.M., Owens, G., Zhou, S.R., 2011. Transport of copper as affected by titania nanoparticles in soil columns. Environmental Pollution 159 (5), 1248e1256. Franchi, A., O’Melia, C.R., 2003. Effects of natural organic matter and solution chemistry on the deposition and reentrainment of colloids in porous media. Environmental Science & Technology 37 (6), 1122e1129. Gargiulo, G., Bradford, S.A., Simunek, J., Ustohal, P., Vereecken, H., Klumpp, E., 2007. Transport and deposition of metabolically active and stationary phase Deinococcus radiodurans in unsaturated porous media. Environmental Science & Technology 41 (4), 1265e1271. Gargiulo, G., Bradford, S.A., Simunek, J., Ustohal, P., Vereecken, H., Klumpp, E., 2008. Bacteria transport and deposition under unsaturated flow conditions: the role of water content and bacteria surface hydrophobicity. Vadose Zone Journal 7 (2), 406e419. Grolimund, D., Borkovec, M., Barmettler, K., Sticher, H., 1996. Colloid-facilitated transport of strongly sorbing contaminants in natural porous media: a laboratory column study. Environmental Science & Technology 30 (10), 3118e3123. Handley-Sidhu, S., Renshaw, J.C., Moriyama, S., Stolpe, B., Mennan, C., Bagheriasl, S., Yong, P., Stamboulis, A., PatersonBeedle, M., Sasaki, K., Pattrick, R.A.D., Lead, J.R., Macaskie, L.E., 2011. Uptake of Sr2þ and Co2þ into biogenic hydroxyapatite: implications for biomineral ion exchange synthesis. Environmental Science & Technology 45 (16), 6985e6990. Israelachvili, J.N., 1992. Intermolecular and Surface Forces. Academic, New York. Johnson, W.P., Li, X.Q., 2005. Comment on breakdown of colloid filtration theory: role of the secondary energy minimum and surface charge heterogeneities. Langmuir 21 (23), 10895. Johnson, W.P., Li, X.Q., Assemi, S., 2007. Deposition and reentrainment dynamics of microbes and non-biological colloids during non-perturbed transport in porous media in the presence of an energy barrier to deposition. Advances in Water Resources 30 (6e7), 1432e1454. Johnson, W.P., Pazmino, E., Ma, H.L., 2010. Direct observations of colloid retention in granular media in the presence of energy barriers, and implications for inferred mechanisms from indirect observations. Water Research 44 (4), 1158e1169.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 0 5 e5 9 1 5
Kim, H.N., Walker, S.L., Bradford, S.A., 2010. Coupled factors influencing the transport and retention of Cryptosporidium parvum oocysts in saturated porous media. Water Research 44 (4), 1213e1223. Kuznar, Z.A., Elimelech, M., 2007. Direct microscopic observation of particle deposition in porous media: role of the secondary energy minimum. Colloids and Surfaces A: Physicochemical and Engineering Aspects 294 (1e3), 156e162. Li, X.Q., Scheibe, T.D., Johnson, W.P., 2004. Apparent decreases in colloid deposition rate coefficients with distance of transport under unfavorable deposition conditions: a general phenomenon. Environmental Science & Technology 38 (21), 5616e5625. Ma, Q.Y., Traina, S.J., Logan, T.J., Ryan, J.A., 1994. Effects of aqueous Al, Cd, Cu, Fe(II), Ni, and Zn on Pb immobilization by hydroxyapatite. Environmental Science & Technology 28 (7), 1219e1228. Marquardt, D.W., 1963. An algorithm for least-squares estimation of nonlinear parameters. Journal of the Society for Industrial and Applied Mathematics 11 (2), 431e441. McCarthy, J.F., Zachara, J.M., 1989. Subsurface transport of contaminants. Environmental Science & Technology 23 (5), 496e502. Onuma, K., Oyane, A., Tsutsui, K., Tanaka, K., Treboux, G., Kanzaki, N., Ito, A., 2000. Precipitation kinetics of hydroxyapatite revealed by the continuous-angle laser lightscattering technique. The Journal of Physical Chemistry B 104 (45), 10563e10568. Redman, J.A., Walker, S.L., Elimelech, M., 2004. Bacterial adhesion and transport in porous media: role of the secondary energy minimum. Environmental Science & Technology 38 (6), 1777e1785. Roy, S.B., Dzombak, D.A., 1997. Chemical factors influencing colloid-facilitated transport of contaminants in porous media. Environmental Science & Technology 31 (3), 656e664. Ryan, J.N., Elimelech, M., 1996. Colloid mobilization and transport in groundwater. Colloids and Surfaces A: Physicochemical and Engineering Aspects 107, 1e56. Schijven, J.F., Simunek, J., 2002. Kinetic modeling of virus transport at the field scale. Journal of Contaminant Hydrology 55 (1e2), 113e135. Shellenberger, K., Logan, B.E., 2002. Effect of molecular scale roughness of glass beads on colloidal and bacterial deposition. Environmental Science & Technology 36 (2), 184e189. Shen, C.Y., Li, B.G., Huang, Y.F., Jin, Y., 2007. Kinetics of coupled primary- and secondary-minimum deposition of colloids under unfavorable chemical conditions. Environmental Science & Technology 41 (20), 6976e6982. Simoni, S.F., Harms, H., Bosma, T.N.P., Zehnder, A.J.B., 1998. Population heterogeneity affects transport of bacteria through sand columns at low flow rates. Environmental Science & Technology 32 (14), 2100e2105. Simunek, J., He, C.M., Pang, L.P., Bradford, S.A., 2006. Colloidfacilitated solute transport in variably saturated porous media: numerical model and experimental verification. Vadose Zone Journal 5 (3), 1035e1047. Simunek, J., Sejna, M., van Genuchten, M.T., 2008. The HYDRUS1D Software Package for Simulating the One-dimensional
5915
Movement of Water, Heat, and Multiple Solutes in Variablysaturated Media. Version 4. Beta 1. ICGWMC-TPS-70. International Ground Water Modeling Center, Colorado School of Mines, Golden, Colorado. Smiciklas, I., Dimovic, S., Plecas, I., Mitric, M., 2006. Removal of Co2þ from aqueous solutions by hydroxyapatite. Water Research 40 (12), 2267e2274. Stankovich, S., Piner, R.D., Nguyen, S.T., Ruoff, R.S., 2006. Synthesis and exfoliation of isocyanate-treated graphene oxide nanoplatelets. Carbon 44 (15), 3342e3347. Taneda, S., 1979. Visualization of separating Stokes flows. Journal of the Physical Society of Japan 46 (6), 1935e1942. Torkzaban, S., Bradford, S.A., Walker, S.L., 2007. Resolving the coupled effects of hydrodynamics and DLVO forces on colloid attachment in porous media. Langmuir 23 (19), 9652e9660. Torkzaban, S., Tazehkand, S.S., Walker, S.L., Bradford, S.A., 2008. Transport and fate of bacteria in porous media: coupled effects of chemical conditions and pore space geometry. Water Resources Research 44 (4). doi:10.1029/2007WR006541. Torkzaban, S., Kim, H.N., Simunek, J., Bradford, S.A., 2010. Hysteresis of colloid retention and release in saturated porous media during transients in solution chemistry. Environmental Science & Technology 44 (5), 1662e1669. Tufenkji, N., Elimelech, M., 2005. Breakdown of colloid filtration theory: role of the secondary energy minimum and surface charge heterogeneities. Langmuir 21 (3), 841e852. Vaidyanathan, R., Tien, C., 1988. Hydrosol deposition in granular beds. Chemical Engineering Science 43 (2), 289e302. Verwey, E.J.M., Overbeek, J.T.G., 1948. Theory of the Stability of Lyophobic Colloids. Elsevier, Amsterdam. Walshe, G.E., Pang, L.P., Flury, M., Close, M.E., Flintoft, M., 2010. Effects of pH, ionic strength, dissolved organic matter, and flow rate on the co-transport of MS2 bacteriophages with kaolinite in gravel aquifer media. Water Research 44 (4), 1255e1269. Wang, D.J., Chu, L.Y., Paradelo, M., Peijnenburg, J.M.G.W., Wang, Y.J., Zhou, D.M., 2011. Transport behavior of humic acid-mediated nano-hydroxyapatite in saturated packed column: effects of Cu, ionic strength, and ionic composition. Journal of Colloid and Interface Science 360 (2), 398e407. Yao, K.M., Habibian, M.T., O’melia, C.R., 1971. Water and waste water filtration: concepts and applications. Environmental Science & Technology 5 (11), 1105e1112. Yoon, J.S., Germaine, J.T., Culligan, P.J., 2006. Visualization of particle behavior within a porous medium: mechanisms for particle filtration and retardation during downward transport. Water Resources Research 42 (6). doi:10.1029/2004WR003660. Zhang, L.L., Wang, L.L., Zhang, P., Kan, A.T., Chen, W., Tomson, M.B., 2011. Facilitated transport of 2, 20 , 5, 50 polychlorinated biphenyl and phenanthrene by fullerene nanoparticles through sandy soil columns. Environmental Science & Technology 45 (4), 1341e1348. Zhou, D.M., Wang, D.J., Cang, L., Hao, X.Z., Chu, L.Y., 2011. Transport and re-entrainment of soil colloids in saturated packed column: effects of pH and ionic strength. Journal of Soils and Sediments 11 (3), 491e503.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 1 6 e5 9 2 4
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
A full scale worm reactor for efficient sludge reduction by predation in a wastewater treatment plant J. Tamis a,b,*, G. van Schouwenburg b, R. Kleerebezem a, M.C.M. van Loosdrecht a a b
Department of Biotechnology, Delft University of Technology, Julianalaan 67, 2628 BC, Delft, The Netherlands SR Technologie BV, Postbus 236, 8100 AE, Raalte, The Netherlands
article info
abstract
Article history:
Sludge predation can be an effective solution to reduce sludge production at a wastewater
Received 15 November 2010
treatment plant. Oligochaete worms are the natural consumers of biomass in benthic
Received in revised form
layers in ecosystems. In this study the results of secondary sludge degradation by the
4 May 2011
aquatic Oligochaete worm Aulophorus furcatus in a 125 m3 reactor and further sludge
Accepted 24 August 2011
conversion in an anaerobic tank are presented. The system was operated over a period of 4
Available online 6 September 2011
years at WWTP Wolvega, the Netherlands and was fed with secondary sludge from a low loaded activated sludge process. It was possible to maintain a stable and active population
Keywords:
of the aquatic worm species A. furcatus during the full period. Under optimal conditions
Worms
a sludge conversion of 150e200 kg TSS/d or 1.2e1.6 kg TSS/m3/d was established in the
Sludge reduction
worm reactor. The worms grew as a biofilm on carrier material in the reactor. The surface
Aquatic Oligochaetes
specific conversion rate reached 140e180 g TSS/m2d and the worm biomass specific
Full scale implementation
conversion rate was 0.5e1 g TSS sludge/g dry weight worms per day. The sludge reduction
Biological disintegration
under optimal conditions in the worm reactor was 30e40%. The degradation by worms was an order of magnitude larger than the endogenous conversion rate of the secondary sludge. Effluent sludge from the worm reactor was stored in an anaerobic tank where methanogenic processes became apparent. It appeared that besides reducing the sludge amount, the worms’ activity increased anaerobic digestibility, allowing for future optimisation of the total system by maximising sludge reduction and methane formation. In the whole system it was possible to reduce the amount of sludge by at least 65% on TSS basis. This is a much better total conversion than reported for anaerobic biodegradability of secondary sludge of 20e30% efficiency in terms of TSS reduction. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
The activated sludge process is worldwide the most abundantly used process to treat wastewater. The principle of organic carbon removal with this process is based on partial
aerobic respiration and partial conversion of organic matter to biomass that can be separated from the treated wastewater in a settling tank. Taken into account that 40e50% of all organic carbon removed is converted to biomass (i.e. sludge) this can be regarded as an unwanted side-product of the process.
Abbreviations: COD, chemical oxygen demand; HRT, hydraulic retention time; p.e., people equivalent; SRT, solid retention time; ST, sludge tank; TSS, total suspended solids; VVM, volume of gas per volume of liquid per minute; WR, worm reactor; WWTP, waste water treatment plant. * Corresponding author. Department of Biotechnology, Delft University of Technology, Julianalaan 67, 2628 BC, Delft, The Netherlands. Tel.: þ31 152781482. E-mail address:
[email protected] (J. Tamis). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.08.046
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 1 6 e5 9 2 4
Nomenclature CO2G;IN [kg/m3] Oxygen concentration of inflow gas 3 COG 2 [kg/m ] Oxygen concentration of off-gas fN [gN/gTSS] Nitrogen content of organic sludge biodegradability fd [-] k [d1] degradation rate constant 3 NHIN 4 [gN/m ] Ammonium concentration of inflow 3 NH4 [gN/m ] Ammonium concentration of outflow 3 NOIN 3 [gN/m ] Nitrate concentration of inflow NO3 [gN/m3] Nitrate concentration of outflow rx [kgTSS/m3d] rate of sludge reduction rsCOD [gCOD/m3d] rate of soluble COD formation rNO3 [gN/m3d] rate of nitrate formation rN2 [gN/m3d] rate of nitrogen gas formation t [d] time
Produced sludge in many places has to be disposed of by incineration or other costly techniques as reviewed by Fytilli and Zabaniotou (2008). Therefore sludge minimisation techniques have become a major research topic in recent years (Wei et al., 2003; Pe´rez-Elvira et al., 2006; Carre`re et al., 2010). These techniques can be applied directly in the activated sludge process or in combination with anaerobic sludge digestion. Some proposed schemes seem to rely on decreasing sludge production by increasing the total sludge age (e.g. Cannibal, OSA). Other techniques are based on the increase of the solubilisation/hydrolysis rate and or increase the biodegradability. Used methods include mechanical, chemical, thermal and biological processes (Fig. 1) and may be placed inline or as pre-treatment step. The drawback of most of these methods is that they are energy and cost intensive. Most of the proposed strategies are not based on the basic cycling of organic carbon in natural systems. Complex organic carbon in nature is generally converted by higher organisms including benthic worms and it seems a logical route to incorporate worms into a system for sludge minimisation. Implementation of this ecological principle for enhanced sludge degradation in WWTP was demonstrated on laboratory scale by various researchers: Ratsak et al. (1993) described the growth of Naididae worm species in a wastewater treatment plant; Rensink et al. (1997) measured sludge reduction by Tubificidae worms in a trickling filter to be roughly 50% on TSS basis although Luxmy et al. (2001) did not measure any
5917
TSSIN [kg/m3] solids concentration sample point 0 TSSOUT [kg/m3] solids concentration sample point 1 DTSS [kg] amount of solids degraded TSSinit [kg] initial amount of solids VR [m3] reactor volume gx [gCOD/gTSS] conversion factor for oxidation of organic carbon to CO2 gNO3 [gCOD/gN] conversion factor for oxidation of ammonia to nitrate gN2 [gCOD/gN] conversion factor for oxidation of ammonia to nitrogen gas d [-] relative difference between gas measurements and dry weight based conversion rates FG [m3/h] gas flow FL [m3/d] liquid flow over worm reactor
significant breakdown of sludge by worms. Wei et al. (2006) described a system that could be used for cultivation of Tubificidae. At about the same time Elissen et al., 2006 described a reactor concept in which Lumbriculus variegatus could be cultivated. Hendrickx et al. (2009) investigated this system set-up further and explored the influence of operational parameters. While the abovementioned research established evidence that sludge conversion by aquatic worm species was possible and explored the influence of various operational parameters, a proper design for a full scale system has not been reported nor has the type of most effective benthic worm for practical conditions been investigated. In order to get insight in the relevant factors that play a role in the full scale operation of worm predation on a wastewater treatment plant a pilot system was developed. The main goal was to show the feasibility of adding a benthic worm reactor in a sludge reduction strategy and to demonstrate that stable sludge reduction rates can be established at a wastewater treatment plant. In this study the results of conversion of secondary sludge by the aquatic Oligochaete worm Aulophorus furcatus in a 125 m3 reactor and further degradation in an anaerobic tank are presented. This system was operated over a period of more than 4 years at WWTP Wolvega, the Netherlands (coordinates 52.8867N; 5.9898W, http://maps.google.nl/maps? q¼52.8867,þ5.9898).
Fig. 1 e Overview of methods that aim for sludge disintegration for enhanced biodegradability.
5918
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 1 6 e5 9 2 4
2.
Material and methods
2.1.
WWTP and sludge characteristics
The sludge was produced in a low loaded conventional activated sludge system, using nitrification/denitrification for N removal and with phosphate precipitation with iron for Premoval. The sludge loading rate was 0.07 kg COD/kg TSS/ d and the COD/N ratio of the influent was around 10:1. There was no primary settling tank present and the sludge age was approximately 20 days. The VSS/TSS ratio of the secondary sludge varied over the year between 65% and 75%. The capacity of the WWTP was 70.000 p.e. and approximately 25% of the produced sludge was processed via the worm reactor system. The sludge was obtained from the settling tank and mixed with effluent to obtain the appropriate concentration of 1e3 g/l of TSS. Large particles were removed by a drum filter with a mesh size of 500 mm. Although large particles were selectively removed it was observed that the filter unit operation had no significant influence on the VSS/TSS ratio of the sludge. Ammonium and soluble COD concentrations in the obtained secondary were generally low with ammonium concentrations of around 1e3 mg/l and soluble COD concentrations at around 50e100 mg/l.
2.2.
Process configuration and operational conditions
The important unit operations are the worm reactor and the anaerobic sludge sedimentation and holding tank (further referred to as sludge tank). A schematic overview of the process set-up is presented in Fig. 2. Part of the sludge was
directed via the worm reactor (stream 0) to the sludge tank (stream 1). Another part was directly fed to the anaerobic tank sludge tank via a bypass stream designated 0b. Operational parameters varied over time and 3 operational phases could be distinguished. 1. Focus on maximum degradation efficiency in the worm reactor in 2007. The worm reactor loading rate was around 2e3 kg TSS/m3d while the sludge tank loading rate was around 0.3 kg TSS/m3d. Bypass to the sludge tank was not active. 2. Focus on operational parameters for maximum degradation in total system (i.e. worm reactor in combination with sludge tank) in 2008. The worm reactor loading rate was around 3 kg TSS/m3d while the sludge tank loading rate was around 0.3 kg TSS/m3d. Around 25% of the total processed sludge was added via the bypass to the sludge tank. 3. High throughput conditions with the goal of maximizing the total degradation in terms of mass and not percentages in 2009 and 2010. The worm reactor loading rate was between 4 and 8 kg TSS/m3d while the sludge tank loading rate was around 0.5 kg TSS/m3d. Around 40% of the total processed sludge was added via the bypass to the sludge tank. In the sludge tank the sludge was thickened to a concentration of 25e35 g/l of TSS and periodically transported to a central treatment facility. The water released from the sedimentation process was recirculated to the aeration tank (stream 2). The basic of the design of the worm reactor included a carrier for immobilisation of the worms in the system, and
Fig. 2 e Schematic diagram of worm reactor set-up, system boundaries for this study are indicated with the dashed box. Sample points have been numbered 0e3. Unit operations of interest are the worm reactor (WR) and the sludge tank (ST).
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 1 6 e5 9 2 4
an appropriate sludge concentration in the feed. The latter is important for maintaining an adequate dissolved oxygen concentration and preventing toxic ammonium levels. The worm reactor was designed to be able to retain A. furcatus worm biomass using a specific carrier material consisting of a plastic mesh of 5 mm. The reactor temperature was maintained at 25 C using a heat exchanger and heat pump. The reactor was aerated in an airlift mode to provide appropriate mixing. Perspectives of the worm reactor are presented in Fig. 3. It was not necessary to inoculate the reactor since A. furcatus appeared to be naturally present in the secondary sludge and was able to colonize the reactor with a doubling time of approximately 1 week; however at some points in time, mainly after technical maintenance of the reactor inoculations were performed to achieve quicker start-up. The reactor was operated for 1434 days (and counting); data on sludge conversion rates was available for a period from 7th of November 2006 until 11th of October 2010. Average operational parameters (including the full period) for the worm reactor and the sludge tank are presented in Table 1. The solid retention time (SRT) in the worm reactor was estimated to be 1e3 times higher than the hydraulic retention time (HRT) because of sludge retention by the worms on the carrier material. The SRT in the sludge tank was approximately 30 times higher than the HRT depending on the final concentration of the sludge during sedimentation. SRT was established by calculating the amount of solids present divided by the solid mass flow.
2.3.
Analysis e measurements
Reactor performance and operational conditions were monitored according to the scheme presented in Table 2.
2.4.
Analysis e kinetic analysis
Degradation rates were calculated from online TSS concentration and volumetric flow measurements according to Eq. (1). FL TSSIN TSSOUT rX ¼ VR
(1)
To investigate the contribution of sludge degradation by worms to the total degradation, the degradation rate of secondary sludge from WWTP Wolvega was determined in
5919
aerated batch experiments with no worms present. Like in the worm reactor large particles were removed from the sludge by a drum filter with a mesh size of 500 mm. Temperature was maintained at 25 1 C to ensure operational conditions comparable with the worm reactor. The obtained data was fitted to a first order model according to Eq. (2). DTSS TSSinit
¼ fd 1 ek$t
(2)
Fitted parameters included the biodegradable fraction ( fd) and degradation rate constant (k). These parameters were subsequently used to estimate what the extent of degradation would be in a continuous system identical to the worm reactor using Eq. (3). DTSS TSSinit
2.5.
¼
fd $SRT SRT þ 1=k
(3)
Analysis e mass balancing
Results were evaluated by on site off-gas analyses (Hellinga et al., 1996). Assuming steady-state conditions, oxygen transfer rates could be related to sludge degradation using a COD balance (Eq. (4)) while assuming nitrite concentration were negligible. FG G;IN CO2 CGO2 ¼ gX rX rsCOD þ gNO3 rNO3 þ gN2 rN2 VR
(4)
Rates on sludge degradation, soluble COD and nitrate formation were readily available from direct measurements. In order to quantify the rate of nitrogen gas formation that was required for evaluation of the COD balance, it was assumed that the nitrogen content of the organic fraction of the degraded sludge was 10% on mass basis (Gujer et al., 1999) and that this was the same inflow and outflow. Subsequently the nitrogen gas production rate was calculated as follows: rN2 ¼ fN rX þ
FL FL NOIN þ NHIN 4 NH4 3 NO3 VR VR
(5)
To evaluate the COD balance the relative difference between gas measurements and dry weight based conversion rates was calculated according to Eq. (6). FG G;IN CO2 CGO2 VR d¼1 gX rX rsCOD þ gNO3 rNO3 þ gN rN2
(6)
Fig. 3 e From left to right: operational worm reactor; carrier material with worm biomass; zoom in on carrier material with mesh size of 5 mm; Aulophorus furcatus under microscope.
5920
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 1 6 e5 9 2 4
Table 1 e Average worm reactor and sludge tank operational parameters.
Volume Carrier material Volumetric flow Hydraulic retention time Solid retention time Solids concentration Temperature Dissolved oxygen
3.
Worm reactor
Sludge tank
125 1100 250 12 [12e36] [1e3] [20e29] >4
900 n/a 250 86 [1500e2500] [25e35] [4e18] <0,1
m3 m2 m3/d h h g/l C mg/l
Results and discussion
Over a period of 1434 days, the reactor was operated with a fluctuating but relatively stable worm activity. A relatively stable worm population was maintained with densities of on average 200 90 g dry weight worms per m2 of carrier material; the dominant species was identified as A. furcatus by microscopic inspection (a movie of worm biomass from the reactor is included as supplementary data available at http://homepage. tudelft.nl/8n9f7/aulophorus.avi and will be embedded in final pdf version). The worms formed a kind of biofilm that almost fully covered the carrier material. Apparently, reactor conditions favoured growth of this species since over the years A. furcatus remained the dominant species as was observed by microscopic inspection. Supplementary video related to this article can be found at doi:10.1016/j.watres.2011.08.046.
3.1.
Conversion in the worm reactor
Sludge conversion rates based on dry weight and volumetric flow measurements were determined on a daily basis and this reflected overall reactor performance in terms of sludge conversion (Fig. 4).
The sludge loading to the reactor was varying during the operational period due to changes in experimental goals. It should be noted that this included periods with lower conversion due to deliberate reactor operations that were not optimal. Under optimal conditions a sludge conversion efficiency of 150e200 kg TSS/d or 1.2e1.6 kg TSS/m3/d or 30e40% on TSS basis could be achieved. Assuming an average VSS content of 70% the sludge reduction in the worm reactor was therewith estimated to be around 50% on VSS basis. The worm surface specific conversion rate under these conditions was 140e180 g TSS/m2d and the worm biomass specific conversion rates was 0.5e1 g TSS sludge/g dry weight worms per day. This biomass specific consumption rate was in the same order of magnitude as was measured for L. variegatus in a continuous reactor by Hendrickx et al. (2009). In general only a weak correlation between the sludge loading and conversion rates was observed. Apparently many other factors were influencing the conversion rates, with adequate supply of sludge being just one of the boundary conditions for adequate functioning of the system. Furthermore, reactor performance was sometimes influenced by technical failures of the full scale installation: the low conversion in the last months of 2009 was due to technical problems with pump machinery. Variations in sludge conversions could however be explained mainly by taking into account that it was not well feasible to maintain a constant high growth rate of the Aulophorus biomass. The worm biomass retention time was controlled by regularly manually removing a fraction of the biomass from the carrier material (this biomass leaves the reactor then via the normal effluent flow and is included in the solids mass balance). In periods when the worm biomass was controlled at a low biomass amount of around 50e100 g dry weight/m2 (i.e. high growth rate) conversions were higher than when large amounts (300e400 g dry weight/m2) of biomass accumulated. Considering the generally and comparatively high surface specific conversion rates of 140e180 g TSS/m2d it is to be expected that at high biomass
Table 2 e Overview of the sampling and analysis methods. Measurement Volumetric flow TSS concentration Temperature Dissolved oxygen Volumetric flow TS concentration Ammonium concentration Nitrite concentration Nitrate concentration Phosphate concentration Soluble COD concentration Worm biomass on carrier O2 gas consumption CO2 gas consumption
Sample pointa
Frequency (week1)
Method
Accuracy
0,1 0,1 WR,ST WR 3 3 WR,0,1,2 WR,0,1,2 WR,0,1,2 WR,0,1,2 0,1,2 WR WR WR
Online Online Online Online [0,05e1] [0,05e1] 1 1 1 1 1 2 online online
Magnetic field Spectrophotometric Thermometer LDO Manualb Dry weight Spectrophotometric Spectrophotometric Spectrophotometric Spectrophotometric Spectrophotometric Manualc Paramagnetic IR spectrometry
5% 10% 1% 5% 5% 10% 10% 10% 10% 10% 10% 20% 5% 5%
a See process overview (Fig. 2). WR ¼ worm reactor, ST ¼ sludge tank. b Transport: 36 m3 per truck. c Removal of biomass with a water vacuum cleaner and subsequent dry weight determination.
5921
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 1 6 e5 9 2 4
Fig. 4 e Loading and conversion rates in the worm reactor. Values are average dry weight based measurements over 15 days.
densities the inside of the biofilm becomes anaerobic causing the worm biomass to function sub-optimal. Although ammonium and soluble COD were released during the conversion of the sludge by the worms concentration remained relatively low with ammonium concentrations around 5e20 mg/l and soluble COD concentrations at around 100e200 mg/l.
3.2.
Conversion in combination with the sludge tank
The sludge tank was originally intended for sedimentation and storage. However, significant anaerobic degradation was observed in this stage of the process despite a low temperature (Table 3). Results varied over the 3 distinct operational phases: in 2007 the loading rate of the system was relatively low resulting in a good conversion efficiency in the worm reactor. In this phase it was found that after predation in the worm reactor, an additional 50% of the sludge disappeared upon storage in the sludge tank. Off-gas analysis demonstrated that TSS removal was associated with conversion to methane at ambient temperature (i.e. 4e20 C). In the second phase in 2008 the conversion efficiency in the sludge tank was even higher while conversion in the worm
reactor was lower. Results from the high throughput experiment in 2009 and 2010 indicated that conversion efficiency decreased at high loading rates while volumetric conversion rates increased. In conclusion it was observed that it was possible to obtain 65% sludge reduction on TSS basis by application of the worm reactor in combination with the sludge tank. This is a much better conversion than reported for anaerobic digestion which typically has a degradation efficiency of no more than 20e30% on TSS basis (Li and Noike, 1992; Shimizu et al., 1993; Lin et al., 1997; Lafiitte-Trouque´ et al., 2002; Valo et al., 2004; Bolzonella et al., 2005; Bougrier et al., 2006a, 2006b).
3.3.
Control experiment results
To compare the decay of biomass by worm grazing with endogenous decay in the sludge we performed batch control experiments with the secondary sludge (Fig. 5). The aerobic endogenous decay could be described with a first order decay kinetics expression with a decay coefficient of 0.12 [d1] and a maximal degradability of 14%. This decay coefficient is in accordance with the aerobic endogenous respiration rates typically used in the Activated Sludge Model No. 3 (Gujer et al., 1999; Koch et al., 2001) which
Table 3 e Overview of the conversion in the worm reactor and the sludge tank. Worm reactor loading rate 3
2007 2008 2009e2010
Sedimentation tank
conversion 3
loading rate 3
Combined conversion
conversion
kg/m /d
kg/m /d
%
kg/m /d
kg/m3/d
%
%
2.66 3.03 4.37
0.86 0.61 0.65
32% 20% 15%
0.34 0.39 0.53
0.16 0.21 0.26
49% 55% 40%
66% 64% 49%
5922
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 1 6 e5 9 2 4
Fig. 5 e Results of batch aerobic degradation of the activated sludge from WWTP Wolvega with no worms present. Errors bars represent standard deviation of triplicate tests. Dashed line represent fitted first order model.
were 0.3e0.4 [d1] at 25 C for the active biomass. The observed value of 0.12 d1 refers to the total sludge which is a mixture of active biomass, residual particalute COD and inorganic solids. Preliminary experiments on anaerobic decay of the sludge indicated a decay coefficient of 0.04 d1. Comparison with the worm reactor using a SRT of 2 days indicates that the degradation efficiency in a continuous system without worms would be only 2% on TSS basis. Apparently the observed sludge degradation in the worm reactor was primarily caused by predation of sludge by worms leading to order of magnitude increase in the degradation rate over standard endogenous processes.
3.4.
Evaluation of nitrogen and phosphorus release
Sludge degradation leads to nitrogen and phosphate release. In the aerated worm reactor this leads to nitrification. Denitrification occurs in the worm biofilm on the carrier material and in the sludge tank. The nitrogen and phosphorus release during sludge degradation in the experimental period is reported in Fig. 6. Measurements were only conducted on the worm reactor during an 872 days period (18-01-2008e6-8-2010) and sludge tank data were taken from a 414 days period (20-42009e6-8-2010). Total nitrogen (NH4 þ NO2 þ NO3) release from the worm reactor was 70 g N per kg TSS conversion. Assuming 10% (w/w) nitrogen content of the sludge this leads to the conclusion that an additional 30 g N per kg TSS conversion was potentially processed by denitrifying bacteria to produce nitrogen gas. Phosphate release was 8 g PO4eP per kg TSS conversion in the worm reactor. Assuming a 1e2% (w/w) phosphate content of the sludge leads to indications of occurrence of precipitation reactions (e.g. CaPO4, NH4MgPO4 or FePO4); this has a very limited influence of the solids balance, but could be further investigated in future optimisation of the process.
Fig. 6 e Soluble nitrogen and phosphate production per amount of TSS degraded in the worm reactor and sludge tank. Note that negative net values for nitrite and nitrate production are due to denitrification in the sludge tank.
In the sludge tank the total nitrogen release was 16 g N per kg TSS conversion, which was significantly lower than in the worm reactor. Note that not all nitrate produced in the worm reactor was subsequently converted by denitrification processes in the sludge tank: at 2 days HRT on average inflow contained 10e20 mg/l nitrate-N while outflow contained 3e10 mg/l nitrate-N, since mixing was only done sporadically for short periods this can partly be resulting from short circuiting flows in the tank. Phosphate release in the sludge tank was negligible possibly due to the high concentration of iron that was set free from the sludge under anaerobic conditions. Relatively large quantities of iron were present in the sludge since WWTP Wolvega applies chemical phosphate removal. Over the total process the net release of nutrients was approximately 30 g nitrogen and 2 g phosphorus per kg TSS conversion. The cause of the low net formation of nitrogen and the underlying processes in the sludge tank are interesting topics for further investigation, but are for practical purposes outside the scope of this report.
3.5.
Mass balance
Since measurement of sludge amounts at wastewater treatment plants is always cumbersome we evaluated the sludge removal also by measuring the oxygen conversion in the worm reactor. Worm reactor inflow and off-gas oxygen and CO2 concentrations were measured over a period of 23 days and oxygen and carbon dioxide conversion was compared to the dry weight balance over the worm reactor (Fig. 7). Average inflow oxygen gas concentration was 20.91% while average outflow gas concentration was 20.16%. The oxygen uptake rate was 1.70 kg/m3d at a gas flow rate of 0.2 VVM (Dissolved oxygen was around 5 mg/l). For short periods dry weight measurements indicated strong deviations from the gas measurements. It appears that the gas measurements are giving a more reliable picture of the conversions in the worm reactor while the dry weight
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 1 6 e5 9 2 4
5923
Fig. 7 e Worm reactor oxygen and CO2 transfer rates and dry weight based conversion estimation (Eq. (1), material and methods).
balance only gives an accurate measure for the amount of sludge reduction over longer periods. Variance in the dry weight based conversion estimation was primarily caused by accumulation effects of sludge in the worm biofilm in the reactor. This trend was evaluated by comparing oxygen gas transfer to dry weight based conversion estimation, nitrification and denitrification rates (using the Eq. (2), material and methods). The used nitrate production rate was 35 mg/g sludge conversion and the denitrification rate (obtained from the overall N-balance was) 30 mg/g sludge conversion while nitrite production was negligible; based on evaluation of the nitrogen release (Fig. 6). The relative error in this was subsequently calculated (using Eq. (4), material and methods) for periods of different lengths and plotted in Fig. 8. These results indicate that measurements of reduction based on dry weight were only representative over longer periods. For establishing the amount of sludge degradation,
Fig. 8 e Relative difference between the cumulative sludge conversion estimated from dry weight measurements and gas transfer measurements. X-axis represents the length of the evaluated period.
periods of at least 2e3 week need to be evaluated if based on dry weight measurements.
3.6.
General evaluation
This work shows that use of benthic worms can greatly reduce sludge production. Compared to conventional reported values for secondary sludge digestion the sludge reduction was roughly doubled. Learning from nature and adapting this in engineered systems is a well known approach in wastewater treatment that can be extended to sludge treatment. It was possible to maintain a relatively stable worm population in a simple reactor design. Regarding economic potential, the proposed system with a self-organizing worm biofilm in a well mixed reactor will be favoured over more complex designs proposed in literature (i.e. Elissen et al., 2006; Hendrickx et al., 2009). There is still room for improvement of this system. We observed that worms have a high biomass specific consumption rate at low worm biomass densities. Having a good worm biomass control in the process will therefore improve the efficiency of the process. Conversion in the worm reactor appears to be limited by the maximum oxygen transfer rate from the bulk liquid to the worm biofilm on the carrier material. An interesting finding was that while in the worm reactor the sludge removal was in the same order as for digestion systems on secondary sludge, there was still a sludge degradation in the subsequent anaerobic sludge holding tank. This anaerobic sludge digestion occurred under psychrophilic conditions. This was not anticipated beforehand and therefore not optimised. It appeared as if the worms formed a kind of pre-treatment system not only oxidising sludge but also converting it in a better biodegradable form similar as reported for e.g. ozone treatment or physical processes. Possibly, the mechanism by which worms degrade sludge inside their body works also outside of their body. Further investigation is needed to clarify the working of this mechanism.
5924
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 1 6 e5 9 2 4
Be that as it may, the observed phenomenon opens the option to optimise the total system by minimising the conversion in the worm reactor and maximising the anaerobic digestion. Another option might be reversal of the order of the unit operations (i.e. the worm reactor and the anaerobic digestion) in order to obtain a maximal gas production. This way all easily degradable material can be converted into biogas, while relatively slow degradable material is fed to worms.
4.
Conclusion
1. It was possible to maintain a stable and active population of the aquatic worm species A. furcatus in a full scale reactor for a period of 4 years. 2. A reduction of 65% on TSS basis was realised during this period resulting from worm predation and subsequent digestion, whereby the predated sludge digested spontaneously under psychrophilic anaerobic conditions. Overall biodegradability increased and was much higher than total conversion as reported in literature for anaerobic digestion of secondary sludge 3. Reported observations reflect the results of a large scale demonstration plant that was not optimized. It is believed that this is the first step towards a completely new and successful approach to decrease waste activated sludge combined with anaerobic digestion of the sludge.
Acknowledgements This research was funded by the Dutch government (InnoWATOR) and the Dutch board for water innovation (STOWA).
references
Bolzonella, D., Pavan, P., Battistoni, P., Cecchi, F., 2005. Mesophilic anaerobic digestion of waste activated sludge: influence of the solid retention time in the wastewater treatment process. Process Biochemistry 40 (3e4), 1453e1460. Bougrier, C., Albasi, C., Delgene`s, J.P., Carre`re, H., 2006a. Effect of ultrasonic, thermal and ozone pre-treatments on waste activated sludge solubilisation and anaerobic biodegradability. Chemical Engineering and Processing 45 (8), 711e718. Bougrier, C., Delgene`s, J.P., Carre`re, H., 2006b. Combination of thermal treatments and anaerobic digestion to reduce sewage sludge quantity and improve biogas yield. Process Safety and Environmental Protection 84 (4), 280e284. Carre`re, H., Dumas, C., Battimelli, A., Batstone, D.J., Delgene`s, J.P., Steyer, J.P., Ferrer, I., 2010. Pretreatment methods to improve
sludge anaerobic degradability: a review. Journal of Hazardous Materials 183 (1e3), 1e15. Elissen, H.J.H., Hendrickx, T.L.G., Temmink, H., Buisman, C.J.N., 2006. A new reactor concept for sludge reduction using aquatic worms. Water Research 40 (20), 3713e3718. Fytili, D., Zabaniotou, A., 2008. Utilization of sewage sludge in EU application of old and new methods e A review. Renewable & Sustainable Energy Reviews 12 (1), 116e140. Gujer, W., Henze, M., Mino, T., van Loosdrecht, M., 1999. Activated sludge model no. 3. Water Science and Technology 39 (1), 183e193. Hellinga, C., Vanrolleghem, P., van Loosdrecht, M.C.M., Heijnen, J.J., 1996. The potential of off-gas analyses for monitoring wastewater treatment plants. Water Science and Technology 33 (1), 13e23. Hendrickx, T.L.G., Temmink, H., Elissen, H.J.H., Buisman, C.J.N., 2009. Aquatic worms eating waste sludge in a continuous system. Bioresource Technology 100 (20), 4642e4648. Koch, G., Kuhni, M., Siegrist, H., 2001. Calibration and validation of an ASM3-based steady-state model for activated sludge systems e part I: prediction of nitrogen removal and sludge production. Water Research 35 (9), 2235e2245. Lafitte-Trouque´, S., Forster, C.F., 2002. The use of ultrasound and [gamma]-irradiation as pre-treatments for the anaerobic digestion of waste activated sludge at mesophilic and thermophilic temperatures. Bioresource Technology 84 (2), 113e118. Li, Y.Y., Noike, T., 1992. Upgrading of anaerobic-digestion of waste activated-sludge by thermal pretreatment. Water Science and Technology 26 (3e4), 857e866. Lin, J.-G., Chang, C.-N., Chang, S.-C., 1997. Enhancement of anaerobic digestion of waste activated sludge by alkaline solubilization. Bioresource Technology 62 (3), 85e90. Luxmy, B.S., Kubo, T., Yamamoto, K., 2001. Sludge reduction potential of metazoa in membrane bioreactors. Water Science and Technology 44 (10), 197e202. Pe´rez-Elvira, S., Nieto Diez, P., Fdz-Polanco, F., 2006. Sludge minimisation technologies. Reviews in Environmental Science and Biotechnology 5 (4), 375e398. Ratsak, C.H., Kooijman, S.A.L.M., Kooi, B.W., 1993. Modeling the growth of an Oligochaete on activated-sludge. Water Research 27 (5), 739e747. Rensink, J.H., Rulkens, W.H., 1997. Using metazoa to reduce sludge production. Water Science and Technology 36 (11), 171e179. Shimizu, T., Kudo, K., Nasu, Y., 1993. Anaerobic waste-activated sludge digestionea bioconversion mechanism and kinetic model. Biotechnology and Bioengineering 41 (11), 1082e1091. Valo, A., Carre`re, H., Delgene`s, J.P., 2004. Thermal, chemical and thermo-chemical pre-treatment of waste activated sludge for anaerobic digestion. Journal of Chemical Technology & Biotechnology 79 (11), 1197e1203. Wei, Y., Liu, J., 2006. Sludge reduction with a novel combined worm-reactor. Hydrobiologia 564 (1), 213e222. Wei, Y.S., Van Houten, R.T., Borger, A.R., Eikelboom, D.H., Fan, Y.B., 2003. Minimization of excess sludge production for biological wastewater treatment. Water Research 37 (18), 4453e4467.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 2 5 e5 9 3 3
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Nitrogen and phosphorus removal from municipal wastewater effluent using microalgal biofilms N.C. Boelee a,b,c,*, H. Temmink a,c, M. Janssen a,b, C.J.N. Buisman a,c, R.H. Wijffels b a
Wetsus, Centre of Excellence for Sustainable Water Technology, P.O. Box 1113, 8900 CC Leeuwarden, The Netherlands Bioprocess Engineering, Wageningen University, P.O. Box 8129, 6700 EV Wageningen, The Netherlands c Sub-Department of Environmental Technology, Wageningen University, P.O. Box 8129, 6700 EV Wageningen, The Netherlands b
article info
abstract
Article history:
Microalgal biofilms have so far received little attention as post-treatment for municipal
Received 22 April 2011
wastewater treatment plants, with the result that the removal capacity of microalgal
Received in revised form
biofilms in post-treatment systems is unknown. This study investigates the capacity of
24 August 2011
microalgal biofilms as a post-treatment step for the effluent of municipal wastewater
Accepted 25 August 2011
treatment plants. Microalgal biofilms were grown in flow cells with different nutrient loads
Available online 1 September 2011
under continuous lighting of 230 mmol/m2/s (PAR photons, 400e700 nm). It was found that the maximum uptake capacity of the microalgal biofilm was reached at loading rates of
Keywords:
1.0 g/m2/day nitrogen and 0.13 g/m2/day phosphorus. These maximum uptake capacities
Algal biofilm
were the highest loads at which the target effluent values of 2.2 mg/L nitrogen and 0.15 mg/
Nitrogen removal
L phosphorus were still achieved. Microalgal biomass analysis revealed an increasing
Phosphorus removal
nitrogen and phosphorus content with increasing loading rates until the maximum uptake
Effluent polishing
capacities. The internal nitrogen to phosphorus ratio decreased from 23:1 to 11:1 when
Wastewater treatment
increasing the loading rate. This combination of findings demonstrates that microalgal biofilms can be used for removing both nitrogen and phosphorus from municipal wastewater effluent. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
Microalgae have been used to treat wastewater in large ponds for many years. Interest in microalgae and wastewater treatment has been renewed by recent findings suggesting that microalgal biofuel production could be made economically viable and sustainable when using wastewater as a nutrient supply (Clarens et al., 2010; Pittman et al., 2011; Wijffels et al., 2010). The costs of harvesting microalgae from diluted suspensions has led to the investigation of alternative microalgal systems, including biofilms (Shi et al., 2007). Microalgal
biofilm systems have the advantage that they are able to retain the biomass, while operating at a short hydraulic retention time. It is also expected that little or no separation of microalgae and water is required before discharging the effluent (Roeselers et al., 2008; Schumacher et al., 2003), presumably making harvesting much easier than in suspended systems. Moreover, no stirring is needed in the system, resulting in a lower energy requirement compared to suspended microalgal systems. Nevertheless, the performance of algal biofilm systems could be limited by photoinhibition and diffusion limitation of nutrients or carbon dioxide (CO2) (Liehr et al., 1988; Murata et al., 2007).
* Corresponding author. Wetsus, Centre of Excellence for Sustainable Water Technology, P.O. Box 1113, 8900 CC Leeuwarden, The Netherlands. Tel.: þ31 0 58 2843000; fax: þ31 0 58 2843001. E-mail address:
[email protected] (N.C. Boelee). URL: http://www.algae.wur.nl 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.08.044
5926
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 2 5 e5 9 3 3
So far, little attention has been given to the possibility of using microalgal biofilms as a post-treatment system for municipal wastewater. However, the EU Water Framework Directive’s objective to obtain good chemical and ecological status for all surface waters by 2015, leads to the need for wastewater treatment plants to further reduce their nitrogen (N) and phosphorus (P) emissions. Microalgae have the ability to assimilate N and P down to very low concentrations (Collos et al., 2005; Hwang et al., 1998). With this ability and the potential to recover N and P from the algal biomass, microalgae present an interesting and more sustainable alternative to existing post-treatment systems such as denitrifying filters, which require an organic carbon source and emit CO2. The aim of this study was to investigate whether microalgal biofilms can be used as a post-treatment system for municipal wastewater treatment plants. The main issues addressed in this study are the uptake capacity and the final effluent concentrations obtained by the microalgal biofilms. The values 2.2 mg/L total N and 0.15 mg/L total P were used as target values for the effluent processed by the microalgal biofilms. These values are currently in use by the Dutch water boards as discharge guidelines for sensitive water bodies, and were used in this study because the classification of a good chemical and ecological status of surface water of the Water Framework Directive is yet unknown. Furthermore, the biomass growth, composition and washout, as well as the photosynthetic efficiency of the biofilm were studied in order to evaluate the potential of the microalgae biofilm process as post-treatment system.
2.
Material & methods
2.1.
Experimental setup
All experiments were performed in a system consisting of a flow cell (STT products b.v.) with an inflow of (synthetic) wastewater effluent, and a recycle vessel with both an outflow of effluent and a recycle flow back to the flow cell. This system is shown in Fig. 1. In the flow cell, a water layer of 2 cm flowed over a 1 mm plastic sheet (PVC 0.018 m2), on which the microalgal biofilm grew. The microalgal biofilm was continuously illuminated by a bank of fluorescent lamps (Sylvania, CF-LE 55W/840) at a light intensity of 230 mmol/m2/s (PAR photons, 400e700 nm). Light intensity was measured with an LI-COR SA190 2p PAR quantum sensor at the level of the biofilm surface. The transparent top of the flow cell contained an outlet covered by a septum, through which gas formed during the experiment was sampled and removed. In the 400 mL recycle vessel, the pH was measured (Endress þ Hauser CPS11D-7AA21) and controlled at pH 7 by pulse-wise addition of CO2 gas. The temperature was controlled at 22 C with the water jacket of the recycle vessel. The dissolved oxygen concentration was measured continuously (Mettler Toledo InPro 6050/120) at the inlet and outlet of the flow cell. Using the oxygen measurement at the inlet of the flow cell, the dissolved oxygen concentration in the inflowing synthetic wastewater was controlled at 35% air saturation by pulse-wise addition of N2 gas to the recycle vessel.
The inflow of synthetic wastewater effluent was adjusted between 0.3 mL/min and 5 mL/min, depending on the desired nitrogen and phosphorus load. The recycle flow was 40 mL/ min, giving laminar flow velocities of about 0.6 mm/s and a retention time inside the flow cell of around 9 min. Based on the large recycle flow which was efficiently mixed with the influent before entering the flow cell, and on visual observations, it can be safely assumed that all of the biofilm was exposed to the same loading rate. The effluent flow was collected and stored for a maximum of 24 h at 2 C, in order to measure the microalgal dry weight in the effluent. To prevent microalgal growth in the system outside the flow cell, all tubing was black, glassware was brown and all glassware and connections were covered in aluminum foil. Table 1 shows the 18 experiments with their loading rates, duration, and corresponding hydraulic retention times. The 3 NO 3 -N and PO4 -P concentrations were measured daily in the influent and effluent and the suspended solids were measured daily in the effluent. At the end of the experiment the biomass was harvested by scraping the biofilm from the plastic sheet. From 11 experiments this wet microalgal biomass was frozen at 80 C until analyses were performed to determine the total amount of biomass and its C, N, P content.
2.2.
Microalgal biofilm cultivation
Microalgae were scraped off the surface of a settling tank of the effluent of the municipal wastewater treatment plant in Leeuwarden, the Netherlands. These microalgae were grown on four pieces of PVC sheet in 250 mL Erlenmeyer flasks containing 100 mL synthetic wastewater effluent. The Erlenmeyers were kept in a growth chamber (New Brunswick Scientific Innova 44) on an orbital shaker (100 rpm) at a temperature of 25 C. The growth chamber was continuously illuminated with 40 mmol photons/m2/s, and a concentration of 2% CO2 was maintained in the gas phase. Every two weeks, synthetic wastewater effluent was replaced and most of the microalgal biofilm was scraped from the plastic sheet to allow the microalgal biofilms to re-grow and keep the culture viable. The plastic sheet of the flow cell was scratched with sandpaper before the inoculation procedure prior to the experiment. Starting this procedure, the pieces of the inoculum from the growth chamber were rubbed over the flow cell sheet. Afterwards, the flow cell sheet was left in synthetic wastewater effluent for a minimum of 2 h before being put into the flow cell. The synthetic wastewater effluent contained N and P in the typical species and concentrations of municipal wastewater 3 effluent, being 10 mg/L NO 3 -N and 1.1 mg/L PO4 -P. In addition, the synthetic effluent contained (micro) nutrients based on the algae WC medium (Andersen, 2005), to rule out limitations of any nutrients other than N or P. The synthetic effluent lacked an organic carbon source in order to obtain a microalgal biofilm with as little heterotrophic bacteria as possible. The synthetic wastewater effluent composition was as follows: 60.67 mg/L NaNO3, 36.76 mg/L CaCl2.2H2O, 36.97 mg/L MgSO4.7H2O, 420.04 mg/L NaHCO3, 28.42 mg/L Na2SiO3.9H2O, 6.19 mg/L K2HPO4. Trace elements and vitamins: 3.82 mg/L EDTA.2H2O, 1.90 mg/L FeCl3, 1.00 102 mg/L CuSO4.5H2O, 2.20 102 mg/L ZnSO4.7H2O, 9.99 103 mg/L CoCl2.6H2O, 0.147 mg/L
5927
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 2 5 e5 9 3 3
Fig. 1 e Scheme of the experimental setup.
MnCl2.2H2O, 6.00 103 mg/L Na2MoO4.2H2O, 1.00 mg/L H3NO3, 0.10 mg/L vitamin B1, 5.00 104 mg/L vitamin H, 5.00 104 mg/L vitamin B12. Apart from the experiments with synthetic wastewater two experiments were performed with real wastewater effluent. The effluent was collected from the wastewater treatment plant in Leeuwarden, the Netherlands, and was enriched with NaNO3 and used as inflow in the experiment. 3 The concentrations of NO 3 -N and PO4 -P were respectively 5.57 mg N/L and 0.97 mg P/L.
2.3.
Analytical procedures
Samples were taken from the influent and effluent flow during operation. After filtering through a 0.45 mm filter, samples 3 were analyzed for NO 3 -N and PO4 -P with ion chromatography (Metrohm Compact IC 761 equipped with a conductivity detector, using the pre-column Metrohm Metrosep A Supp 4/5 Guard and the column Metrohm Metrosep A Supp 5, 150/ 4.0 mm). Gas samples were analyzed with gas chromatography to determine the amount of oxygen gas formed in addition to the dissolved oxygen (Varian, CP-4900 equipped
Table 1 e Settings of the experiments performed in the flow cells with loading rates, duration time and hydraulic retention time (HRT). Experiment 1a 2a 3 4a,b 5a 6 7a 8a,b 9a 10 11a 12a 13 14 15a 16a 17a 18a
Load NO 3 -N (gN/m2/d)
Load PO3 4 -P (g P/m2/d)
Duration experiment (d)
HRT system (d)
0.11 0.17 0.18 0.22 0.31 0.34 0.36 0.45 0.52 0.64 0.78 1.01 1.23 1.23 1.49 1.97 3.60 4.53
0.011 0.018 0.022 0.039 0.033 0.033 0.034 0.078 0.076 0.079 0.072 0.094 0.126 0.126 0.152 0.200 0.399 0.502
9 11 15 10 11 21 9 10 14 13 15 15 22 25 13 13 12 12
1.9 1.4 1.5 1.4 0.8 2.3 0.6 0.7 3.5 0.4 0.7 0.5 0.6 0.6 0.4 0.3 0.2 0.1
a Biomass collected at the end of the experiment. b Experiments performed with real wastewater effluent, all other experiments were performed with synthetic wastewater effluent.
5928
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 2 5 e5 9 3 3
˚ with a thermal conductivity detector using a Mol Sieve 5 A PLOT 10 m column at 80 C and a PoraPlot U 10 m column at 65 C, and argon as carrier gas at 1.47 mL/min). Microalgal dry weight in the effluent was determined by filtration of the effluent through pre-weighed glass fiber filters (Whatman GF/ F) and oven drying at 105 C for at least 24 h. The stored biomass from the flow cells was also oven dried at 105 C for 24 h, ground and dried for at least another 24 h. The C and N content of this microalgal biomass was measured in duplicate with an elemental analyzer (EA 1110, ThermoQuest CE Instruments using a 1000 C vertical quartz tube with a constant flow of helium at 120 ml/min, an oxidation catalyst (WO3) zone, a copper zone followed by a Porapack PQS column at 60 C and a TCD detector). To determine the P content, duplicates of the biomass were digested using 8 mL HNO3 (68%) per 0.4 g microalgal biomass, first heating to 180 C at up to 1000 W for 15 min, followed by 15 min at 180 C at up to 1000 W in a microwave (Milestone ETHOS 1). After this digestion, the total P concentration was measured with inductive coupled plasma (ICP) (Perkin Elmer Optima 5300 DV equipped with an optical emission spectrometer).
2.4.
Scanning electron microscopy
At the end of the experiment, a small piece of plastic sheet with the microalgal biofilm was cut from the plastic sheet in the flow cell. This piece was gently washed three times in phosphate buffer solution (137 mM NaCl, 2.7 mM KCl, 10 mM Na2HPO4, 2 mM KH2PO4, pH 7.4). The biofilm was then fixed using 2.5% glutaraldehyde at room temperature for 2 h. The fixed sample was dehydrated sequentially with 30%, 50%, 70%, 90%, 100% (v/v) ethanol, each step taking 20 min. The sample was then dried at 35 C for a minimum of 15 min. The samples were sputter coated with a thin 5 nm gold layer, and observed with a scanning electron microscope (SEM) (JEOL JSM-6480LV) in high vacuum mode (acceleration voltage 10e15 kV, working distance 10 mm).
2.5.
Calculations
The measured percentage of oxygen (%) was converted to oxygen concentration (mM) using the maximum oxygen solubility in the air-saturated synthetic wastewater. This
solubility was determined by sequentially adding 20 mL of sodium sulfite to 8 mL of air-saturated synthetic wastewater. Each step displayed a 15% oxygen decrease, resulting in an oxygen solubility of 0.24 mM or 7.61 mg/L. As a measure of photosynthetic efficiency, the quantum yield of oxygen evolution (QYO2) was calculated: QY02 ¼
P O2 ðgÞ$Vg þ O2 ðIÞ$QI $t ½mol O2 =mol photons PFD$A$t
where O2(g) is the oxygen concentration in the gas phase (mol/L), Vg is the volume of the gas formed inside the flow cell (L), O2 (l) is the oxygen concentration in the liquid (mol/L), Ql is the flow rate of the liquid (L/d), t is the time (d), PFD is the PAR photon flux density (mol photons/m2/d) and A is the area (m2). 3 The final effluent concentrations of NO 3 -N and PO4 -P of each experiment were calculated as the average concentration of the effluent samples taken during the quasi-steady state at the end of the experiment.
3.
Results
3.1.
Biofilm growth
During the experiments, the microalgal biofilm covered the plastic sheet with a thin layer after about four days, although the plastic could still be seen in some places. After this coverage, filamentous green microalgae started to grow from the biofilm into the overlaying water layer, forming streamers. After about 12 days, these streamers started to detach. This pattern of growth was similar at all loading rates. However, higher loading rates resulted in a visually greener and more loosely attached biofilm. SEM pictures revealed that the top layer of the biofilm grown on synthetic wastewater effluent consisted mainly of pennate diatoms (Nitzschia sp.) and that the green filaments were lying on top of this biofilm, as presented in Fig. 2A. It can be seen from Fig. 2B that the experiments with real effluent showed a larger diversity of microalgae, although these experiments were started with the same inoculum originating from a municipal wastewater treatment plant. Presumably, the wastewater effluent further inoculated the biofilm with new species.
Fig. 2 e SEM pictures of the microalgal biofilm of experiments with synthetic wastewater effluent (A) and real wastewater effluent (B).
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 2 5 e5 9 3 3
3.2. Removal of nitrate and phosphate from synthetic and real wastewater effluent Fig. 3A and B show the removal of nitrate (NO 3 ) and phosphate ) by the microalgal biofilm at representative low, inter(PO3 4 mediate and high nutrient loads, using synthetic and real wastewater effluent. Fig. 3A and B show that a similar removal pattern was observed during the experiments. The first two to four days of the experiment were considered a start-up phase. During this phase the biomass began to populate the carrier 3 material and little uptake of NO 3 -N and PO4 -P was observed. 3 As the algae grew, the NO3 -N and PO4 -P concentrations decreased further, and finally remained stable during four to eight days. This phase was described as quasi-steady state, 3 were the uptake of NO 3 -N and PO4 -P was stable. The effluent concentrations reached during this quasi-steady state phase were dependant on the applied loading rate, where the higher loading rates gave higher final effluent concentrations. Moreover, at the higher loading rates an increase of both 3 NO 3 -N and PO4 -P concentrations was observed at the end of the experiment. Finally, Fig. 3A and B show that the removal pattern of the real wastewater effluent was similar to the removal pattern of the synthetic wastewater effluent.
3.3.
Uptake capacity and microalgae washout
Fig. 4A presents the uptake rate of NO 3 -N during the quasisteady state at the end of the experiments performed at different loading rates. All rates are expressed relative to the biofilm surface in m2. In addition, Fig. 4A shows the removal of NO 3 -N during the two experiments with real effluent. As was expected, the uptake first increased approximately linearly with increasing loading rates. The uptake was no longer linear 2 above a loading rate of 1.0 g NO 3 -N/m /d. The uptake of -P showed a similar behavior, as can be seen in Fig. 4B. PO3 4 Fig. 4A and B also show that at low loading rates, the measured uptake rates of the experiments applying the real effluent were comparable to the uptake rates of the experiments applying the synthetic effluent. It can be seen from Fig. 5A that the final NO 3 -N concentrations in the effluent increased with increasing loading
A
5929
rates, from 0.03 mg NO 3 -N/L at a low loading rate of 0.18 g 2 NO 3 -N/m d to 7.3 mg NO3 -N/L at the highest load of 4.5 g 2 2 -N/m /d. Up to a loading rate of 1.0 g NO NO 3 3 -N/m /d the NO3 -N effluent concentrations remained below the target of 2 2.2 mg N/L. Therefore, 1.0 g NO 3 -N/m /d was referred to as the maximum uptake capacity of NO3 -N, being the highest load at which the target effluent value was still reached. Fig. 5B shows the effluent concentrations of PO3 4 -P. With one exception, the final effluent concentrations of PO3 4 -P also remained below the target value of 0.15 mg P/L up to a loading rate of 0.13 g 2 3 PO3 4 -P/m /d. Thus, the maximum uptake capacity of PO4 -P 2 3 was found to be 0.13 g PO4 -P/m /d. Some washout of microalgae occurred during the experiments. The amount of suspended solids that washed out remained stable until about 12 days when the quasi-steady state period ended and chunks of biofilm were released. Fig. 5A and B show the average microalgae washout, as associated N and P concentration, until the end of the quasisteady state period. This N and P concentration was calculated with an internal N and P content of the microalgae of 0.039 g N/g biomass and 0.0055 g P/g biomass, the average values of the measurements described in the following paragraph (3.4). Interestingly, the average washout of microalgae from the different experiments was always around 3.2 mg/L suspended solids. This corresponds to an average washout of 0.13 mg N/L and 0.018 mg P/L.
3.4.
Microalgal biomass
The carbon, nitrogen and phosphorus content of the microalgal biomass were determined. The carbon content remained nearly constant during the experiments and was on average 0.45 g C/g biomass. Fig. 6A and B show the internal N and P content of the microalgal biomass at different loading rates of 3 NO 3 -N and PO4 -P. Both the internal N and P content increased with increasing loading rates until the maximum 2 2 3 loading rates of 1.0 g NO 3 -N/m /d and 0.13 g PO4 -P/m /d. At higher loading rates, the N content was stable with 0.048 g N/g biomass and the P content only increased marginally from 0.0072 until 0.0099 g P/g biomass. Fig. 6B also shows the molar N:P ratio of the different experiments with increasing loading
B
3L Fig. 3 e Effluent concentrations of NOL 3 -N (A) and PO4 -P (B) during four flow cell experiments using synthetic wastewater 2 L effluent at loading rates of 0.18, 1.01 and 1.49 g NO3 -N/m2/d and 0.022, 0.094 and 0.152 g PO3L 4 -P/m /d, and using real 2 2 L 3L wastewater effluent at a loading rate of 0.45 g NO3 -N/m /d and 0.078 g PO4 -P/m /d.
5930
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 2 5 e5 9 3 3
A
A
B
B
Fig. 4 e Uptake of nitrogen (A) and uptake of phosphorus (B) by the microalgal biofilms at different nitrogen loads using synthetic and real wastewater effluent. The straight dotted line indicates a hypothetical linear relationship.
rates. The N:P ratio decreased from approximately 23:1 at low loading rates to approximately 14:1 at the maximum uptake capacity. At higher loads the ratio only decreased slightly as a result of the marginally increasing P content while the N content remained stable. The N:P ratio had decreased to 11:1 2 at the highest loading rate of 4.5 g NO 3 -N/m /d. Fig. 7 presents the average biomass production, calculated from all biomass produced during the entire experiment. Although this average biomass production showed variations, the production appeared to increase with the increasing nutrient load. The highest biomass production of 7.7 g/m2/ 2 d was found at a loading rate of 1.97 g NO 3 -N/m /d, and the 2 lowest production was 2.1 g/m /d and was found at a loading 2 rate of 0.11 g NO 3 -N/m /d.
3.5.
Photosynthetic efficiency
Fig. 8 shows the calculated quantum yield of oxygen production during the quasi-steady state period of the 16 experiments. It is apparent from the figure that this yield varied with the experiments, similar to the average biomass production, and no direct relationship was observed between the yield and the
3L Fig. 5 e Effluent concentrations of NOL 3 -N (A) and PO4 -P (B) and the calculated average washout of microalgal biomass (assuming 0.039 g N/g biomass and 0.0055 g P/g biomass) at different nitrogen and phosphorus loads. The dotted lines indicate the target values for the final effluent of 2.2 mg N/L and 0.15 mg/L P.
loading rate. The highest quantum yield of 0.043 mol O2/mol 2 photons was found at a loading rate of 2.0 mg NO 3 -N/m /d, while the lowest yield of 0.012 mol O2/mol photons was found 2 at the loading rate 0.34 g NO 3 -N/m /d.
4.
Discussion
The results of this study show that it is possible to simulta3 neously decrease NO 3 -N and PO4 -P concentrations in wastewater effluent to the target values 2.2 mg N/L and 0.15 mg P/L using a microalgal biofilm. These results are in contrast to initial expectations that N and P could not be removed simultaneously from the effluent due to the supply ratio of around 20 N:1P in municipal wastewater effluent. This N:P ratio is not optimal for freshwater microalgae, which have an average N:P ratio of 12:1 (Healey, 1973; Duboc et al., 1999; Ahlgren et al., 1992). However, varying internal compositions are reported in literature for different algal species (Ahlgren et al., 1992; Ho et al., 2003) and under different growth
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 2 5 e5 9 3 3
5931
A
B Fig. 8 e Calculated quantum yield of oxygen production at different nitrogen loading rates for the experiments with synthetic wastewater effluent.
Fig. 6 e Measured nitrogen content (A) and phosphorus content (B) and the corresponding molar N:P ratio (B) of the microalgal biomass grown under different nitrogen and phosphorus loads using synthetic and real wastewater effluent. The synthetic wastewater effluent contained nitrogen and phosphorus at a molar N:P ratio of 20:1.
Fig. 7 e The average microalgal biomass production rate (dry weight) during the experiments at different nitrogen loads for both the experiments with synthetic wastewater effluent and the experiments with real wastewater effluent.
conditions, the latter especially through luxury uptake of P (Klausmeier et al., 2004; Powell et al., 2008). The possibility of varying internal ratios was confirmed during this study, where the internal N:P ratio was found to decrease with increasing loading rates. Despite the variable data at low loading rates, the results suggest that the internal N:P ratio decreased only until the maximum uptake capacities 2 2 3 of 1.0 g NO 3 -N/m /d and 0.13 g PO4 -P/m /d. At higher loading rates the decrease in N:P ratio was only marginal. At the highest loading rates an N:P ratio of about 11:1 was found, equal to the average N:P ratio in freshwater microalgae. In addition, nutrients remained in the effluent at the loading rates above the maximum uptake capacity, which indeed shows that there were enough nutrients and the microalgae assimilated the nutrients in preferred ratios. At the low loading rates the N:P ratio of the microalgal biomass was close to the 20:1 ratio of the supplied synthetic wastewater effluent. It can thus be suggested that at, or below, the maximum uptake capacity it will be possible to simultaneously obtain the desired low effluent values for both NO 3 -N -P. Due to the varying internal N:P ratio in the and PO3 4 microalgae, this will also be possible when the ratio in the supplied wastewater is suboptimal for microalgae and varying. The final effluent concentrations that can be achieved using microalgal biofilms in practice are determined by the 3 uptake capacity of NO 3 -N and PO4 -P as measured in this study, but also by the molecular form of the nutrients and by the biomass washout. The N and P in wastewater effluent will 3 not only consist of NO 3 -N and PO4 -P, as was the case in our synthetic wastewater effluent. It is probable that only the inorganic forms of the nutrients will be taken up by the microalgae, and that organic and particulate forms of N and P will either remain in the effluent or be entrapped into the biofilm and mineralized. Although the N in the wastewater effluent typically comprises of less than 10% dissolved organic N, this dissolved organic N can become dominant, up to 85% of the total N, when the total N concentration reaches very low levels (Pehlivanoglu and Sedlak, 2004; Pagilla et al., 2006). A bioavailability assay of phosphorus in municipal wastewater
5932
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 2 5 e5 9 3 3
effluent (Ekholm and Krogerus, 1998) showed that 36% of the total phosphorus was available to algae. Therefore, in practice, the final total N and P concentrations could be higher than the concentrations measured during this study. In this study, the washout of algae was measured as suspended solids and calculated with the measured content of N and P in the microalgal biomass of the biofilm. As the washout consisted of detached microalgae from the biofilm, it was assumed that the internal nutrient content was the same. Contrary to expectations, the washout of microalgae remained stable throughout each experiment with 3.2 mg/L suspended solids, giving an average washout of 0.13 mg N/L and 0.07 mg P/L. Taking these concentrations into account, the N effluent concentrations were still below target up to the 2 loading rate of 1.0 g NO 3 -N/m /d, while the P concentrations were often slightly above target up to this loading rate. It is expected that the washout would increase during longer experiments, when the biofilm is no longer in a quasi-steady state (Horn et al., 2003). This biomass loss and also unwanted mobilization of fixed nutrients are the result of biofilms becoming too thick. Too thick biofilms also lead to a decreased uptake rate and increasing effluent concentrations, as was seen at the end of the flow cell experiments (Fig. 3B). However, this will not be pertinent for the application in practice where the biofilm will be harvested while at steady state, in order to maintain low effluent concentrations. With the maximum uptake capacity obtained in this study, a first rough estimation can be made of the size of a full scale post-treatment system. The required area of microalgal biofilms for 100 000 inhabitants would be around 10 ha (100 000 inhabitants * 130L wastewater/inhabitant/day * (10e2.2 mg N/L) 2 O 1.0 g NO 3 -N/m /d). However, this calculated area is dependent on the uptake capacity, which in turn depends on various factors including the actual supply of sunlight in an outdoor system and the efficiency at which the algae convert this light into biomass. This study found maximum uptake capacities of 1.0 g 2 2 3 NO 3 -N/m /d and 0.13 g PO4 -P/m /d. These values are higher than the removal rates of 0.1e0.6 gN/m2/d and 0.006e0.09 gP/m2/d measured in tubular biofilm photoreactors treating swine slurry (de Godos et al., 2009; 2 Gonza´lez et al., 2008). However, 0.13 g PO3 4 -P/m /d is lower 2 than 0.73 gP/m /d measured in an algal turf scrubber treating wastewater (Craggs et al., 1996). The continuous illumination of 230 mmol/m2/s used in this study could be considered modestly low for summer and very high for winter. This irradiation corresponds to a daily irradiation of 20 mol/m2/d, while an average summer day in the Netherlands gives about 34 mol/m2/d and a winter day gives 5 mol/ m2/d (Huld and Suri, 2007). An irradiation of 34 mol/m2/ d would decrease the area requirement to about 6 ha. In the Netherlands the irradiation is 20 mol/m2/d or higher from April until September. As this half year period corresponds to the time when eutrophication of surface waters can take place, it is possible to use the algal biofilm as post-treatment of municipal wastewater effluent removing the residual N and P to prevent eutrophication. Regarding the efficiency at which algae convert light into biomass, the oxygen quantum yield was assessed. Under low light intensities minimally 10 photons (PAR) are required for
the liberation of one molecule of O2 (Ley and Mauzerall, 1982; Bjorkman and Demmig, 1987) corresponding to an oxygen quantum yield of 0.1. This study found a large range of oxygen quantum yields of 0.012e0.043 mol O2/mol photons. The reason for this large range is unknown, although it is likely that the yield increases with loading rate. A higher efficiency leads to a higher biomass production and thus a higher removal capacity. The previously mentioned 10 ha could decrease to 6 ha when the yield is improved from 0.03 mol/mol to 0.05 mol/mol (100 000 inhabitants 130 L wastewater/ inhabitant/day (10e2.2 mg N/L)/(0.05/0.03 1.0 gN/m2/d)). In addition to area requirement, there are three other aspects to consider when applying microalgal biofilms as post-treatment systems. Firstly, CO2 availability to the algae in the biofilm will need to be considered. Although a modeling study (Liehr et al., 1988) has shown that it is probable that CO2 will become a limiting factor when the biofilm is exposed to ambient air, wastewater effluent can contain sufficient inorganic carbon to sustain algal growth (Van Vooren et al., 1999). Therefore, further investigations can determine if limitation imposed by CO2 will be an important restraint to the biofilm system. Secondly, the effect of the diurnal light cycle on the uptake of nutrients needs to be considered. Most researchers have studied microalgae and microalgal nutrient uptake under continuous illumination, as was also done in this study. However, for application in practice it will need to be known if there is any uptake at night, as this is one of the factors determining the system design. There are indications that uptake may continue at night at a reduced rate, for instance uptake of NHþ 4 was found to be more than 50% of the daylight value in N-limited microalgae (Vona et al., 1999). This reduced uptake rate may be compensated by the expected lower loading rate of N and P at night. A final consideration for the future application of microalgal biofilm systems is the harvesting of the microalgal biomass. The microalgal biomass production measured during this study indicates a significant biomass production in a microalgal biofilm system operating on municipal wastewater effluent. The production for the post-treatment system for 100 000 inhabitants would be around 2 ton/d (1.0 gN/m2/d/ 0.0512 g N/g biomass 10 ha). The harvesting of this biomass can either take place passively or actively, the former being the collection of the biomass that naturally detaches from the biofilm, and the latter involving techniques like backwashing, pH shock or scraping.
5.
Conclusions
The present study has shown that microalgal biofilms can be used to treat municipal wastewater effluent and remove 3 residual NO 3 -N and PO4 -P to the lower discharge demands of 2.2 mg N/L and 0.15 mg P/L. A maximum uptake capacity was 2 2 3 found at a load of 1.0 g NO 3 -N/m /d and 0.13 g PO4 -P/m /d 2 under a light intensity of 230 mmol/m /s. Up to this maximum uptake capacity the internal N and P content of the microalgae was dependent on the loading rate. This implies that microalgae can assimilate both nitrogen and phosphorus at N:P ratios present in wastewater effluent. Furthermore, it was
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 2 5 e5 9 3 3
estimated that a full scale microalgal biofilms post-treatment system for 100 000 inhabitants would be around 10 ha, producing 2 ton of biomass per day.
Acknowledgements This work was performed in the TTIW-cooperation framework of Wetsus, Centre of Excellence for Sustainable Water Technology (www.wetsus.nl). Wetsus is funded by the Dutch Ministry of Economic Affairs, the European Union Regional Development Fund, the Province of Fryslaˆn, the City of Leeuwarden and the EZ/Kompas program of the "Samenwerkingsverband Noord-Nederland”. The authors like to thank the participants of the research theme “Advanced waste water treatment” and the steering committee of STOWA for the discussions and their financial support. The authors also thank M. Balouet, O. Galama and D. Kunteng, for their help with the experiments, A. H. Paulitsch-Fuchs for her help with the SEM and P. Kuntke and J. Racyte for critically reading the manuscript.
references
Ahlgren, G., Gustafsson, I.B., Boberg, M., 1992. Fatty-acid content and chemical-composition of fresh-water microalgae. Journal of Phycology 28 (1), 37e50. Andersen, R.A., 2005. Algal Culturing Techniques. Elsevier Academic Press,. Bjorkman, O., Demmig, B., 1987. Photon yield of O2 evolution and chlorophyll fluorescence characteristics at 77 K among vascular plants of diverse origins. Planta 170, 489e504. Clarens, A.F., Resurreccion, E.P., White, M.A., Colosi, L.M., 2010. Environmental life cycle comparison of algae to other bioenergy feedstocks. Environmental Science & Technology 44 (5), 1813e1819. Collos, Y., Vaquer, A., Souchu, P., 2005. Acclimation of nitrate uptake by phytoplankton to high substrate levels. Journal of Phycology 41 (3), 466e478. Craggs, R.J., Adey, W.H., Jenson, K.R., St. John, M.S., Green, F.B., Oswald, W.J., 1996. Phosphorous removal from wastewater using algal turf scrubber. Water Science and Technology 33 (7), 191e198. de Godos, I., Gonza´lez, C., Becares, E., Garcı´a-Encina, P., Mun˜oz, R. , 2009. Simultaneous nutrients and carbon removal during pretreated swine slurry degradation in a tubular biofilm photobioreactor. Applied Microbiology and Biotechnology 82 (1), 187e194. Duboc, P., Marison, I., von Stockar, U., 1999. Quantitative calorimetry and biochemical engineering. In: Kemp, R.B. (Ed.), Handbook of Thermal Analysis and Calorimetry, vol. 4. Ekholm, P., Krogerus, K., 1998. Bioavailability of phosphorus in purified municipal wastewaters. Water Research 32 (2), 343e351. Gonza´lez, C., Marciniak, J., Villaverde, S., Leo´n, C., Garcı´a, P.A., Munoz, R., 2008. Efficient nutrient removal from swine manure in a tubular biofilm photo-bioreactor using algae-bacteria consortia. Water Science and Technology 58 (1), 95e102.
5933
Healey, F.P., 1973. Inorganic nutrient uptake and deficiency in algae. Critical Reviews in Microbiology 3, 69e113. Ho, T.Y., Quigg, A., Finkel, Z.V., Milligan, A.J., Wyman, K., Falkowski, P.G., Morel, F.M.M., 2003. The elemental composition of some marine phytoplankton. Journal of Phycology 39 (6), 1145e1159. Horn, H., Reiff, H., Morgenroth, E., 2003. Simulation of growth and detachment in biofilm systems under defined hydrodynamic conditions. Biotechnology and Bioengineering 81 (5), 607e617. Huld T, Suri M (2007) Photovoltaic Geographical Information System. http://re.jrc.ec.europa.eu/pvgis/apps4/pvest.php. Accessed 1.10.10. Hwang, S.-J., Havens, K.E., Steinman, A.D., 1998. Phosphorus kinetics of planktonic and benthic assemblages in a shallow subtropical lake. Freshwater Biology 40 (4), 729e745. Klausmeier, C.A., Litchman, E., Simon, A.L., 2004. Phytoplankton growth and stoichiometry under multiple nutrient limitation. Limnology and Oceanography 49 (4), 1463e1470. Ley, A.C., Mauzerall, D.C., 1982. Absolute absorption crosssections for photosystem II and the minimum quantum requirement for photosynthesis in Chlorella vulgaris. Biochimica et Biophysica Acta (BBA) - Bioenergetics 680 (1), 95e106. Liehr, S.K., Eheart, J.W., Suidan, M.T., 1988. A modeling study of the effect of pH on carbon limited algal biofilms. Water Research 22 (8), 1033e1041. Murata, N., Takahashi, S., Nishiyama, Y., Allakhverdiev, S.I., 2007. Photoinhibition of photosystem II under environmental stress. Biochimica et Biophysica Acta (BBA) - Bioenergetics 1767 (6), 414e421. Pagilla, K.R., Urgun-Demirtas, M., Ramani, R., 2006. Low effluent nutrient technologies for wastewater treatment. Water Science & Technology 53 (3), 165e172. Pehlivanoglu, E., Sedlak, D.L., 2004. Bioavailability of wastewaterderived organic nitrogen to the alga Selenastrum Capricornutum. Water Research 38 (14e15), 3189e3196. Pittman, J.K., Dean, A.P., Osundeko, O., 2011. The potential of sustainable algal biofuel production using wastewater resources. Bioresource Technology 102 (1), 17e25. Powell, N., Shilton, A.N., Pratt, S., Chisti, Y., 2008. Factors Influencing luxury uptake of phosphorus by microalgae in waste stabilization ponds. Environmental Science & Technology 42 (16), 5958e5962. Roeselers, G., Loosdrecht, M., Muyzer, G., 2008. Phototrophic biofilms and their potential applications. Journal of Applied Phycology 20 (3), 227e235. Schumacher, G., Blume, T., Sekoulov, I., 2003. Bacteria reduction and nutrient removal in small wastewater treatment plants by an algal biofilm. Water Science and Technology 47 (1), 195e202. Shi, J., Podola, B., Melkonian, M., 2007. Removal of nitrogen and phosphorus from wastewater using microalgae immobilized on twin layers: an experimental study. Journal of Applied Phycology 19, 417e423. Van Vooren, L., Lessard, P., Ottoy, J.P., Vanrolleghem, P.A., 1999. pH buffer capacity based monitoring of algal wastewater treatment. Environmental Technology 20 (6), 547e561. Vona, V., Rigano, V.D.M., Esposito, S., Carillo, P., Carfagna, S., Rigano, C., 1999. Growth, photosynthesis, and respiration of Chlorella sorokiniana after N-starvation. Interactions between light, CO2 and NHþ 4 supply. Physiologia Plantarum 105 (2), 288e293. Wijffels, R.H., Barbosa, M.J., Eppink, M.H.M., 2010. Microalgae for the production of bulk chemicals and biofuels. Biofuels, Bioproducts and Biorefining 4 (3), 287e295.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 3 4 e5 9 4 4
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
The effect of pH on N2O production under aerobic conditions in a partial nitritation system Yingyu Law, Paul Lant, Zhiguo Yuan* Advanced Water Management Centre, The University of Queensland, St Lucia 4072, Australia
article info
abstract
Article history:
Ammonia-oxidising bacteria (AOB) are a major contributor to nitrous oxide (N2O) emissions
Received 13 April 2011
during nitrogen transformation. N2O production was observed under both anoxic and
Received in revised form
aerobic conditions in a lab-scale partial nitritation system operated as a sequencing batch
4 August 2011
reactor (SBR). The system achieved 55 5% conversion of the 1 g NHþ 4 -N/L contained in
Accepted 27 August 2011
a synthetic anaerobic digester liquor to nitrite. The N2O emission factor was 1.0 0.1% of
Available online 7 September 2011
the ammonium converted. pH was shown to have a major impact on the N2O production rate of the AOB enriched culture. In the investigated pH range of 6.0e8.5, the specific N2O
Keywords:
production was the lowest between pH 6.0 and 7.0 at a rate of 0.15 0.01 mg N2O-N/h/g
Nitrous oxide
VSS, but increased with pH to a maximum of 0.53 0.04 mg N2O-N/h/g VSS at pH 8.0. The
pH
same trend was also observed for the specific ammonium oxidation rate (AOR) with the
Anaerobic digester liquor
maximum AOR reached at pH 8.0. A linear relationship between the N2O production rate
Ammonia-oxidising bacteria
and AOR was observed suggesting that increased ammonium oxidation activity may have
Partial nitritation
promoted N2O production. The N2O production rate was constant across free ammonia (FA)
Wastewater treatment
and free nitrous acid (FNA) concentrations of 5e78 mg NH3-N/L and 0.15e4.6 mg HNO2-N/L,
Sequencing batch reactor
respectively, indicating that the observed pH effect was not due to changes in FA or FNA
Greenhouse gas emissions
concentrations. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
Nitrogen removal from anaerobic digester liquor is a sidestream process in municipal wastewater treatment plants (WWTPs). Digester liquor has a high ammonium content (500e1500 mg N/L) and an unfavourable carbon to nitrogen ratio for the conventional nitrification and denitrification treatment. One treatment option is partial nitritation (ammonium oxidation to nitrite) followed by anammox (van Dongen et al., 2001). The molar ratio of bicarbonate to ammonium in the sludge liquor is typically around 1.1:1 (Khin and Annachhatre, 2004), which allows approximately 50% of the ammonium being converted to nitrite through partial nitritation without pH control. This produces a mixture of
ammonium and nitrite at a molar ratio of approximately 1:1, as a suitable feed for the anammox reactor. Ammonium oxidation in such a system is achieved by ammonia-oxidising bacteria (AOB), which are known to produce N2O as a side product (Lipschultz et al., 1981; Kim et al., 2010). Nitrifier denitrification has been reported to be the predominant N2O production pathway by AOB especially under suboptimal conditions such as low dissolved oxygen (DO) concentrations and also when ammonium and nitrite concentrations are high (Kampschreur et al., 2008a; Tallec et al., 2006). These conditions are typical in a partial nitritation reactor treating digester liquor. The N2O emission factor reported for such systems ranged from 0.4 to 6.6% of the nitrogen load in full-scale systems (Kampschreur et al., 2008b,
* Corresponding author. Tel.: þ61 7 3365 4374; fax: þ61 7 3365 4726. E-mail address:
[email protected] (Z. Yuan). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.08.055
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 3 4 e5 9 4 4
2009a; Joss et al., 2009; Desloover et al., 2011) and from below 0.001 to 2.6% of the nitrogen load in lab-scale systems (de Graaff et al., 2010; Pellicer-Na`cher et al., 2010). In comparison to continuous stirred tank reactors (CSTR) (Kampschreur et al., 2008b, 2009a,b), higher ammonium to nitrite conversion rates can be achieved in sequencing batch reactor (SBR) systems with biomass retention (Fux et al., 2003; Lai et al., 2004). However, SBR systems may result in higher N2O emissions due to the more dynamic conditions. pH has been widely studied as a controlling factor for optimising the nitritation process in SBR systems (Gustavsson et al., 2008; Ganigue´ et al., 2007; Fux et al., 2006; Peng et al., 2004). Apart from having a direct effect on the AOB activity (Van Hulle et al., 2010), pH also affects the concentrations of free ammonia (FA) and free nitrous acid (FNA) (Eqs. (1) and (2)). An increasing pH shifts the equilibrium to FA, which is the true substrate of AOB (Suzuki et al., 1974), and is inhibitory to nitrite oxidising bacteria (NOB) (Anthonisen et al., 1976; Vadivelu et al., 2007). Conversely, a decreasing pH increases the FNA concentration, which inhibits both AOB and NOB (Hellinga et al., 1998; Vadivelu et al., 2006). þ NHþ 4 4NH3 þ H
pka ¼ 9:25 at 25 C
NO 2 þ H2 O4HNO2 þ OH
pka ¼ 3:35 at 25 C
(1) (2)
It is generally acknowledged that transient changes in operating conditions can increase N2O production by AOB. In partial nitritation systems operated in SBR mode, pH variations between pH 6.5 and 8.0 have been observed under fully aerobic conditions (Fux et al., 2003; Ganigue´ et al., 2009; de Graaff et al., 2010). The effect of such changes in pH and the subsequent changes in FA and FNA concentrations could have a significant impact on N2O production by AOB. However, such effects have not been reported to date. The objective of this study is to investigate the effect of pH on N2O production by an enriched AOB culture cultivated in an SBR achieving partial ammonium conversion to nitrite under fluctuating pH conditions. N2O and pH were monitored during normal reactor operation to establish the emission factor and to reveal an association between N2O emission and varying pH. Batch tests were carried out to reveal the effects of pH, FA and FNA on N2O production. A slow-feeding strategy was trialed on the SBR to mitigate the pH effect on N2O production.
2.
Material and methods
2.1.
Operation of an SBR achieving partial nitritation
An SBR was operated to select for AOB performing partial nitritation. The return activated sludge from a domestic wastewater treatment plant in Brisbane, Australia, was used as the inoculum. The reactor had a working volume of 8 L. The mixed liquor temperature was controlled at 33 1 C using a water jacket, mimicking that typical for a reactor treating anaerobic digester liquor. The SBR was operated in identical cycles of 6 h. The cycle consisted of the following 10 phases: 25 min settling, 5 min decanting, 10 min idle I, 2.5 min feeding I (aeration on), 67.5 min aerobic reaction I, 86.9 min idle II,
5935
2.5 min feeding II (aeration on), 67.5 min aerobic reaction II, 86.9 min idle III and 1.2 min wasting (aeration on). In each feeding period, synthetic wastewater (composition described below) of 1 L was added. This resulted in a hydraulic retention time (HRT) of 1 day. In all but the settling and decanting phases, the reactor was mixed using a magnetic stirrer at 200 rpm. During the feeding and aerobic reaction phases, aeration was supplied with a constant air flow rate of 2.5 L/ min, giving DO concentrations varying between 0.5 and 0.8 mg O2/L. 100 mL of mixed liquor was wasted per cycle giving rise to a theoretical sludge retention time (SRT) of 20 days. During the settling phase, biomass settling caused a N2O concentration gradient across the SBR column. Idle phase I was thus introduced after decanting to equilibrate the N2O concentration in the reactor through mixing for measurement of total N2O formed during the settling phase. Idle phases II and III were introduced based on the observation that nitritation ceased (due to alkalinity limitation) long before the next feed phase was due. The SBR was fed synthetic wastewater with characteristics of anaerobic digester liquor. The daily nitrogen load was 8 kg N/m3/day. The composition of the wastewater (modified from Kuai and Verstraete, 1998) was: 5.63 g/L of NH4HCO3 (1 g/L NHþ 4 -N), 0.064 g/L of each of KH2PO4 and K2HPO4 and 2 mL of a trace element stock solution. The trace element stock solution contained: 1.25 g/L EDTA, 0.55 g/L ZnSO4$7H2O, 0.40 g/L CoCl2$6H2O, 1.275 g/L MnCl2$4H2O, 0.40 g/L CuSO4$5H2O, 0.05 g/L Na2MoO4$2H2O, 1.375 g/L CaCl2$2H2O, 1.25 g/L FeCl3$6H2O and 44.4 g/L MgSO4$7H2O. The feed had a pH of 8 0.1 and a molar ratio of ammonium to bicarbonate of 1:1. As such, the pH of the reactor mixed liquor increased to pH 7.4 0.1 after feeding and dropped to approximately 6.4 when ammonium oxidation ceased. 1 M NaHCO3 was then used to automatically adjust pH at 6.4 0.05 using a programmable logic controller (PLC) system to ensure that pH did not decrease further. DO and pH in the reactor were continuously monitored online using a miniCHEM-DO2 and a pH metre, respectively. The gas and liquid phase N2O were periodically measured using methods to be further described in Section 2.4. In addition, cycle studies were performed regularly by measuring the ammonium, nitrate and nitrite concentration throughout the 6 h cycle with a sampling interval of 15e30 min. The mixed liquor suspended solids (MLSS) concentration and its volatile fraction (MLVSS) were monitored once a week. The system achieved stable operation at the time of the batch experiments reported below. The sampling and measurement procedures are as described in Section 2.4.
2.2.
Batch experiments
2.2.1.
Batch reactor design and operation
All batch tests were carried out in a batch reactor with a volume of 1.3 L. The reactor was completely sealed with a headspace of 200 mL and a gas outlet to allow off-gas measurements. For each batch experiment, 1.1 L of mixed liquor was collected from the SBR 30 min into an aerobic phase. Temperature was controlled at 33 1 C for all experiments using a water jacket. The batch reactor was mixed with a magnetic stirrer at a stirring speed of 200 rpm. DO and pH
5936
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 3 4 e5 9 4 4
were continuously monitored and controlled. DO concentration was manually controlled in all batch tests at a level similar to the average concentration in the parent reactor (0.55 0.05 mg O2/L). This was achieved by using a gas mixture of N2 and air. The N2 flow and air flow were adjusted using two mass flow controllers (Smart-Trak 50 series e 1 L/ min and 5 L/min, Sierra). The total gas flow rate of the N2 and air mixture entering the batch reactor was maintained at a fixed rate of 0.65 0.05 L/min. These arrangements allowed DO to be controlled at a set-point, while ensuring a constant gas flow. NaHCO3 (0.5 M) and HCl (0.5 M) were used to control pH automatically at 0.05 of the pre-designated set-point using a PLC system. NaHCO3 was chosen to maintain consistency with the buffering in the feed, which was made out of bicarbonate (HCO 3 ). In all batch experiments tested below pH 7, HCl (0.5 M) was used to lower the initial pH of the culture. As ammonium oxidation proceeded, pH tended to decrease in all cases and consequently NaHCO3 (0.5 M) was added to maintain the pH at the desired set-point during all experiments. In all batch tests, the liquid phase and gas phase N2O concentrations were continuously monitored as described in Section 2.4. For each tested pH level and FA or FNA concentration (further described in Sections 2.2.2e2.2.4), the liquid phase and gas phase N2O concentrations were monitored for 30 min after pseudo steady state was reached.
2.2.2.
pH batch tests
To investigate the effect of pH on the N2O production rate of the AOB enriched culture, the batch tests were conducted at seven different pH values: 6.0, 6.5, 7.0, 7.5, 8.0 and 8.5. Each test was performed in triplicates. In all tests, a control period was included whereby pH was controlled at pH 7.0 prior to the test period during which the test pH was applied. This allowed the comparison of the AOB activity between batch tests. pH 7.0 was chosen for the control period as it was the average pH value in the parent reactor (pH ranged between 6.4 and 7.4). To investigate the recovery of the culture from the tested pH, pH was adjusted back to 7.0 after the test period. Mixed liquor samples were taken every 15 min for ammonium, nitrate and nitrite analysis and mixed liquor volatile suspended solids (MLVSS) was measured at the end of each test. The sampling and measurement procedures are as described in Section 2.4. In addition, a single batch test was conducted to investigate the effect of progressive pH increase on the N2O production by the culture. pH was stepwise increased from 7.0 to 8.5 with a 0.5 increment for each step.
2.2.3.
FA and FNA batch tests
Additional tests were performed to investigate the possible effect of FA and FNA on N2O production. Stock solutions of nitrite (NaNO2), at a concentration of 110 g NO 2 -N/L and ammonium ((NH4)2SO4), at a concentration of 110 g NHþ 4 -N/L were used to adjust the FA and FNA concentrations in individual batch tests. Each of the FA and FNA tests involved two stages. In the first stage, different volumes of the stock solution were added into the batch reactor, providing an initial total nitrite or total ammonium concentrations of 0.5, 1.0, 1.5, 2.0 and 2.5 g N/L. The total nitrite concentration was 500 50 mg NO 2 -N/L in all FA batch tests and the total ammonium concentration was 500 50 mg NHþ 4 -N/L in all
FNA batch tests. The pH was controlled at 7.0 0.05 for 30 min in each test. In the second stage, pH was increased to and controlled at pH 7.5 0.05 for the FA test by adding 0.5 M NaHCO3 whereas pH was reduced to pH 6.0 0.05 with 0.5 M HCl for the FNA test. The resulting concentrations of FNA and FA in both stages were calculated according to Eqs. (1) and (2).
2.2.4.
Negative control experiments
Two negative control experiments were conducted to investigate potential N2O production in the absence of ammonium or nitrite. 1.1 L of mixed liquor was harvested from the reactor and allowed to settle. The settled sludge was then washed twice with the synthetic wastewater (without ammonium or nitrite). The washed sludge was re-suspended into 1.1 L of synthetic wastewater supplemented with NaNO2 to give 500 mg NO 2 -N/L in the first negative control. In the second negative control, (NH4)2SO4 was supplemented instead of NaNO2 to give 500 mg NHþ 4 -N/L. The N2O production from the washed sludge was monitored at pH 7.0 (30 min) followed by pH 8.0 (30 min) in both control experiments. DO was maintained at 0.55 0.05 mg O2/L in all cases. The control of DO and pH was as described previously.
2.3.
Varying feeding rate in the parent SBR
Based on the results obtained in batch tests, experiments were designed and carried out on the SBR to investigate if aerobic N2O emission could be reduced by providing the same amount of feed over a longer period. As alkalinity is provided in the feed, a slower feeding rate would avoid sudden rise of pH. The experiment was done in three consecutive cycles. In the first cycle, the feeding flow rate was maintained as per normal operation at 1 L/2.5 min for both aerobic phase I and aerobic phase II. The feeding flow rate in the aerobic phases was reduced to 1 L/25 min and 1 L/50 min in the following two cycles. Both the gas phase and liquid phase N2O concentrations were monitored throughout the three consecutive cycles.
2.4. Online N2O monitoring and offline chemical analysis The gas phase and liquid phase N2O were measured continuously in all batch tests and periodically in the parent reactor. The gaseous N2O concentration was analysed with an infrared analyser (URAS 14 Advance Optima, ABB) and data was logged every 30 s. A moisture filter was installed at the gas inlet of the analyser to capture all moisture entering the analyser. A t-shaped tubing joint was fitted on to the gas sampling tube connecting the gas outlet of the reactor and the gas analyser. This allowed the excess gas flow to escape from the system during aerated phases and gas influx into the system during non-aerated phases, maintaining atmospheric pressure in the reactor at all times. During aerated phases, the sampling pump of the analyser was always adjusted to be lower than the total gas flow rate in the reactor. Liquid phase N2O was measured online using a N2O microsensor (N2O25, Unisense A/S. Aarhus, Denmark). Mixed liquor samples were taken using a syringe and immediately filtered through disposable Milipore filters (0.22 mm pore size) for ammonium, nitrate and nitrite analyses. The ammonium, nitrite and nitrate concentrations were
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 3 4 e5 9 4 4
analysed using a Lachat QuikChem8000 Flow Injection Analyser (Lachat Instrument, Milwaukee). MLSS and MLVSS were analysed according to the standard methods (APHA, 1998).
2.5.
Calculations
The net N2O produced (mg N2O-N) in the parent reactor during each phase in a cycle was calculated using Eq. (3): Net N2 O Produced ¼MN2 ON;liq;end MN2 ON;liq;begin þ N2 O Emitted N2 O Emitted ¼
X
CN2 ON;gas Qair Dt
ð3Þ
(4)
where MN2 ON;liq;end ¼ the mass of dissolved N2O at the end of the phase (mg N2O-N); MN2 ON;liq;begin ¼ the mass of dissolved N2O at the beginning of the phase (mg N2O-N); The N2O emitted was calculated based on Eq. (4), where CN2 ON;gas ¼ the N2O concentration in the off-gas (mg N2O-N/L); Qair ¼ the flow rate of the aeration (L/h) during an aerated phase or gas flow rate through the analyser during a non-aerated phase (L/h); Δt ¼ time interval by which the off-gas N2O concentration was recorded. the N2O concentration in the off-gas in mg N2O-N/L was converted from ppmv based on the volume occupied by 1 mole of ideal gas at standard temperature and pressure (0 C and 101.3 kPa), 22.4 m3/kg-mole and corrected for temperature of the gas sample (25 C). The emission factor of the parent reactor was calculated based on the total amount of N2O emitted in the entire 6 h cycle relative to the total ammonium conversion in the particular cycle (mg N2O-N/mg NHþ 4 -N). The emission factor was reported as average value with the standard deviation. The net biomass-specific N2O production rate (mg N2O-N/ h/g VSS) in each batch test was calculated from the sum of total N2O emitted (Eq. (4)) and total N2O accumulated in the liquid phase (mg N2O-N) over the experimental period (h), normalised to the MLVSS concentration (g VSS). The ammonium oxidation rate (AOR) in each batch test was determined based on the change in ammonium concentration (mg NHþ 4 -N) overtime (h), normalised to the MLVSS concentration (g VSS). In all the plotted graphs, error bars show the standard error calculated from triplicate tests.
2.6.
Microbial characterisation
Biomass samples were prepared according to Daims et al. (2001) for Fluorescence in situ Hybridisation (FISH). The following probes were used: NEU, specific for Nitrosomonas sp.; NSO190 and NSO1225, specific for Beta-proteobacterial AOB, Nsv443, specific for Nitrosospira sp. (Mobarry et al., 1996); NIT3, specific for Nitrobacter sp. (Wagner et al., 1996); Ntspa662, specific for the Nitrospira genera (Daims et al., 2001) and EUB-mix (EUB338, EUB338-II, and EUB338-III), covering most bacteria. All probes were either labelled with 50 FITC (fluoroscein isothiocyanate), or one of the sulfoindocyanine dyes, indocarbocyanine (Cy3) or indodicarbocyanine (Cy5). Hybridisation was performed at 46 C for 1.5 h followed by washing and 15 min incubation with pre-warmed (48 C) washing buffer. FISH-probed samples were visualised using a Zeiss LSM 510 Meta confocal laser scanning microscope (Carl Zeiss,
5937
Jena, Germany) and images collected using a Zeiss Neofluar 40/1.3 oil objective. FISH images (25e30 images) were analysed using DAIME version 1.3, to determine the biovolume fraction of the bacteria of interest (Daims et al., 2001). Reported values are mean percentages with the standard deviations.
2.7.
Detection of intracellular nitric oxide (NO)
The presence of intracellular NO was determined by staining with DAF-FM diacetate (4-amino-5-methylamino- 20 ,70 difluorofluorescein diacetate) (Molecular Probes, Eugene, OR). 1 mL of mixed liquor was harvested from the SBR during the end of the aerobic phase I and the end of idle phase I of a typical cycle. Working solution of DAF-FM diacetate in anhydrous DMSO (99.99%) was added into the cell suspension to obtain a final concentration of 5 mM. Cells were incubated in the dark at room temperature for 30 min. Stained cells were then washed twice with sterile phosphate buffer and visualised using the confocal laser scanning microscope described in Section 2.6.
3.
Results
3.1.
Reactor performance and N2O emissions
Stable operation of the SBR was established 3 months after the reactor start-up. 55 5% of the 1 g NHþ 4 -N/L in the feed was converted to nitrite at the end of a cycle. Ammonium and nitrite concentrations were maintained at 450 50 mg NHþ 4N/L and 550 50 mg NO 2 -N/L, respectively at the end of each cycle. Nitrate concentration was at all times lower than 10 mg NO 3 -N/L, indicating a minimal NOB activity. Fig. 1a shows the typical ammonium, nitrite and nitrate concentration profiles over a cycle. Characterisation of the biomass composition using FISH revealed that 81 3% of the EUBMix probe targeted cells bound to the NSO probe, suggesting that the bacterial population was dominated by ammonia-oxidising beta-proteobacteria comprising the Nitrosospira, Nitrosococcus and Nitrosomonas species (covered by the NSO190 probe). 67 7% of the EUBMix probe targeted cells bound to the NEU probe, suggesting that Nitrosomonas sp. was the dominating AOB. All the NOB probes applied did not give any signal. This is supported by the minimal nitrite oxidation activity in the SBR. The relatively low DO concentration (Blackburne et al., 2008) and high nitrite concentration were likely the selection pressure against the growth of NOB despite the long SRT in the SBR (Vadivelu et al., 2006; Ma et al., 2009). Fig. 1b shows that N2O was produced during both aerated and non-aerated phases (idle and settling phases). Intracellular NO was also detected in cells collected from both phases. DO was completely depleted at the start of the non-aerated phases with immediate onset of N2O accumulation in the liquid phase. However, the transfer of N2O from liquid to gas was minimal due to the lack of aeration. The dissolved N2O was subsequently stripped during the following aerobic phase resulting in N2O spikes (up to 250 ppmv) in the gas phase at the start of each aerobic phase. Mass balance analysis showed
5938
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 3 4 e5 9 4 4
Fig. 1 e (a) Concentration profiles of ammonium e , nitrite e and nitrate e , (b) liquid phase e and gas N2O profiles, (d) DO e and pH e profiles in a typical six hour SBR cycle; (c) Net nitrous oxide phase e and emitted during each phase of an SBR cycle (error bars are standard error, n [ 3). Cycle phases in produced sequence: settling e 25 min, decanting e 5 min, idle phase 1 e 10 min, feeding phase 1 e 2.5 min, aerobic phase 1 e 67.5 min, idle phase 2 e 86.9 min, feeding phase 2 e 2.5 min, aerobic phase 2 e 67.5 min, idle phase 3 e 86.9 min, wasting e 1.2 min.
that N2O accumulated during the non-aerated phases contributed 94 4% to the total N2O emitted during the initial 15 min of the aerated phases. In contrast to the non-aerated phases, N2O produced during aerobic phases was simultaneously stripped, resulting in a continuously measurable N2O level (4e20 ppmv) in the gas phase (Fig. 1b). Fig. 1c shows the amount of N2O emitted during each of the six phases (feed phases were merged with the corresponding aerobic phases). While most of the transfer occurred during the aerobic phases, the settling and idle phases contributed 70% of the net emission with the remaining contributed by aerobic N2O production. Fig. 1d shows the DO and pH profiles over a cycle. DO in the aerobic phases varied between 0.5 and 0.8 mg O2/L. pH rose sharply from 6.4 to 7.4 following a feed period, and decreased to 6.4 0.05 as a result of ammonium oxidation to nitrite and the subsequent pH control (Fig. 1a). Such a trend of pH change (Fig. 1d) coincided with an increase in N2O production at the beginning of the aerobic phase which gradually decreased towards the end of the aerobic phase (Fig. 1b). In an interrupted cycle where the idle phases were bypassed, the dependency between the N2O and pH profiles was still observed (Fig. S1 in Supporting Information), suggesting
a potential role of pH in N2O production in the aerobic phases. The pH effect was therefore further investigated in batch tests.
3.2.
pH effect on N2O production
A pH range of 6.0e8.5 was selected to investigate the effect of pH on the N2O production rate of the AOB culture through batch tests. A control period (pH 7.0) was carried out prior to the test period in each batch test as shown in Fig. 2a. The N2O production rate during the control periods in all tests was consistent with an average of 0.16 0.02 mg N2O-N/h/g VSS. The change in pH during the test period triggered an immediate response in the N2O production rate, achieving a new steady state gas phase N2O concentration within a short period of time (3 min) (Fig. 2a). In comparison to the control period, an increase in N2O production rate was observed when pH was increased above 7.0. The N2O production rate was the highest at pH 8.0 with a 3.5 fold increase (0.53 0.04 mg N2ON/h/g VSS) (Fig. 2b). However, there was a decrease in the N2O production rate when pH was further increased to pH 8.5. In contrast, a decrease of pH from 7.0 to 6.5 or 6.0 did not change the N2O production rate (Fig. 2b).
5939
10
0.6
a
b
8
Net N 2O Production Rate (mg N2O-N/hr/g VSS)
Gas Phase N2O (ppmv)/ DO (mg O2/L)/ pH
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 3 4 e5 9 4 4
6
4
2
0
0.5
0.4
0.3
0.2
0.1
0.0
0.2
0.4
0.6
0.8
1.0
1.2
5.5
6.0
6.5
7.0
Gas Phase N2O (ppmv)/ DO (mg O2 / L)/ pH
10
c
8
6
4
2
0 0.0
0.5
1.0
1.5
7.5
8.0
8.5
9.0
pH
2.0
2.5
3.0
Time (hour)
10
0.6
d 8
0.5
6
0.4
4
0.3
2
0.2
0
Net N 2 O Production Rate (mg N 2 O-N/ hr/ g VSS)
Gas Phase N2O (ppmv) / DO (mg O2/ L) / pH
Time (hour)
0.1
0.0
0.5
1.0
1.5
2.0
2.5
3.0
Time (hour)
Fig. 2 e (a) The DO, gas phase N2O and pH profiles in a typical pH batch test; (b) The effect of pH on the net N2O production rate e ; (c) Reversibility of pH effect on N2O production e an example; (d) The effect of progressive pH increase from pH 7 to 8.5 on , DO e , pH e profiles and N2O production rate e . N2O production. In (a), (c) & (d), gas phase N2O e
To investigate the reversibility of the pH effect, batch tests were performed where the pH was adjusted back to pH 7.0 after being exposed to the tested pH. As demonstrated in the case of pH 8.0, the gas phase N2O increased instantly when pH was increased from 7.0, eventually achieving a steady state (Fig. 2c). Restoring the pH to 7.0 caused an immediate drop in the N2O level back to the baseline (Fig. 2c). Similar trends were observed when pH was adjusted to 8.5, 7.5, 6.5 and 6.0 (Fig. S2 in Supporting Information). In addition, the effect of progressive pH increase on N2O production was also tested. When pH was increased from pH 7.0 to 7.5 and then to 8.0, the gas phase N2O concentration increased accordingly from 1.4 0.2 to 4.9 0.2 and then to 7.0 0.1 ppmv (Fig. 2d). However the gas phase N2O decreased to 2.3 0.1 ppmv when pH was further increased to 8.5. The N2O production rate determined at each tested pH (Fig. 2d) was similar to that in the individual pH batch test (Fig. 2a). This further demonstrated the strong dependency of N2O production by the AOB enriched culture on pH.
3.3.
FA and FNA effects on N2O production
Although pH was the only parameter that was changed during the pH test, pH directly affected the concentration of FNA and FA in the culture medium, given that FNA and FA concentrations have an exponential relationship with pH (Fig. S3 in Supporting Information). In addition, the FA and FNA concentrations varied between 1.1e13.4 mg NH3-N/L and
0.04e0.4 mg HNO2-N/L in the SBR. To distinguish the effect of pH from the potential effects of FA and FNA, the AOB culture was exposed to a range of FA and FNA concentrations under pH controlled conditions. At pH 7.0, the specific N2O production rate was relatively constant (0.17 0.01 mg N2O-N/h/g VSS) despite the increase in FA concentration from 4.5 to 39.5 mg NH3-N/L (Fig. 3a). Similarly, the N2O production rate was also approximately constant (0.42 0.02 mg N2O-N/h/g VSS) at pH 7.5 when FA varied between 15 and 77 mg NH3-N/L (Fig. 3a). However, the N2O production rates were consistently higher at all tested FA concentrations at pH 7.5 compared to pH 7.0, with a magnitude similar to that observed in previous pH batch tests. These results verified that the changes in N2O production in response to pH variation were not caused by changes in FA concentration. Similar observations were made with varying FNA concentrations. At pH 7.0, the N2O production rate was approximately constant at 0.16 0.02 mg N2O-N/h/g VSS while FNA varied between 0.09 and 0.46 mg HNO2-N/L (Fig. 3b). Analogous to that, at pH 6.0, the N2O production rate was about 0.25 0.03 mg N2O-N/h/g VSS, while the FNA concentration varied between 0.9 and 4.6 mg HNO2-N/L. However, the N2O production rate measured at pH 6.0 was slightly higher than those measured at pH 7.0 (Fig. 3b), with reasons unclear. Nevertheless, the results presented in Fig. 3b showed that FNA unlikely had an effect on N2O production by the AOB culture. Two negative controls were conducted showing that there was no N2O production in the absence of ammonium or nitrite
5940
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 3 4 e5 9 4 4
Net N O Production Rate (m g N O-N/hr/g VSS)
a
b
0.5
0.5
0.4
0.4
0.3
0.3
0.2
0.2
0.1
pH 7.0 pH 7.5
0.1
0.0 0
10
20
30
40
50
60
70
pH 7.0 pH 6.0
0.0
80
0
1 2 3 4 Free Nitrous Acid (mg HNO -N/ L)
Free Ammonia (mg NH -N /L)
5
Fig. 3 e The effect of (a) FA concentration and (b) FNA concentration on N2O production.
when pH was adjusted to pH 8.0 (data not shown). This confirmed that the increase in N2O production within the tested pH range of 6.0e8.5 was from bacterial activity and both ammonium and nitrite were required for the production under the tested conditions.
N2O emission factor to 0.8% and 0.7% of ammonium converted at feeding rate of 1 L/25 min and 1 L/50 min, respectively. The pH and the N2O profiles during aerobic phases 1 and 2 in all three cycles are presented in Figs. S4e6.
3.4. AOR
4.
Discussion
4.1.
N2O emission factor
In addition to the N2O production rate, the AOR of the AOB enriched culture also varied with pH. The AOR in all the control runs was consistently at 53.8 4 mg NH3-N/h/g VSS. The AOR across pH 6.0 to pH 7.0 was comparable to the control. When pH was increased to 7.5 and 8.0 AOR increased substantially to 68.1 1 and 74.2 4 mg NH3-N/h/g VSS, respectively (Fig. 4a). The trend was very similar to that observed with N2O production. Indeed, a linear correlation between these two parameters within the pH range of 6.0e8.5 is observed, as shown in Fig. 4b. This suggests that the pH likely induced a change in the AOR, which may have in turn affected the N2O production rate.
3.5.
Effect of feeding rate on N2O produced in SBR
The average N2O emission factor was determined to be 1.0 0.1% of the total ammonium converted. This is lower
a Specific Ammonium Oxidation Rate (mg NH -N/hr/g VSS)
Correlation between the N2O production rate and
100
80
60
40
20
0 5.5
b Average Specific N O Production Rate (mg N O-N/hr/g VSS)
The pH batch study showed that the AOB enriched culture in this study had the lowest N2O production within the pH window of 6.0e7.0. It was hypothesised that the N2O emissions from the reactor would be reduced by operating the reactor in the pH range of 6.0e7.0. The large pH variation in the reactor was caused by the fast feeding (Fig. 1a). To achieve a more uniform pH in the reactor, the feeding rate was reduced gradually over three consecutive cycles. The total N2O produced during aerobic phases 1 and 2 were the highest (1.5e2 mg N2O-N) at the feeding rate of 1 L/2.5 min as per normal operation, corresponding to a pH range of 6.4e7.4 (Fig. 5). When the feeding rate was reduced to 1 L/25 min (pH range 6.4e7.2) and 1 L/50 min (pH range 6.4e6.9) in the following two cycles, the total N2O produced was reduced by 2 folds and 4 folds, respectively (Fig. 5). The proportion of ammonium converted to nitrite in all three cycles was approximately 55%. This resulted in an overall reduction in
6.0
6.5
7.0
7.5
8.0
8.5
9.0
pH
0.6 0.5
r = 0.92 y = 0.014x - 0.53
0.4 0.3 0.2
Test Regression Line
0.1 0.0 40
50
60
70
80
Average Specific Ammonium Oxidation Rate (mg NH -N/hr/g VSS)
Fig. 4 e (a) The effect of pH on the specific AOR; (b) Correlation between the N2O production rate and the AOR from pH 6.0 to 8.5.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 3 4 e5 9 4 4
Net N2O Produced (mg N2O-N)
2.5
Aerobic Phase I Aerobic Phase II
2.0
1.5
1.0
0.5
0.0 1 L/ 2.5 min
1 L/ 25 min
1 L/ 50 min
Feeding Rate
Fig. 5 e The effect of feeding flow rate on aerobic N2O production.
5941
N2O emission from the SBR was predominantly due to N2O accumulation during non-aerated phases upon oxygen depletion. Such an observation has previously been reported for full-scale reject water treatment systems and other enriched nitrifying culture (Kampschreur et al., 2008a,b). It has been demonstrated that N2O production will be increased if partial nitritation systems are momentarily left idle during full-scale application. In this system, the overall N2O emission factor can be reduced to 0.5% of ammonium converted without the idle phases, comparable to the full-scale system reported by Joss et al. (2009). In addition, the N2O emission factor was also lower (0.7% of ammonium converted) when the SBR was fed slowly as described in Section 4.3. While aerobic N2O production was less than the anoxic production, it is still significant. In addition, N2O produced during aerobic phase is immediately emitted due to vigorous stripping by aeration.
4.2. Effect of pH on N2O production under aerobic conditions than emission factors previously reported for two full-scale two-step partial nitritation-anammox processes, namely 3.4% (Kampschreur et al., 2008b) and 8.1e10.5% (Desloover et al., 2011) of ammonium converted. It is also lower than the emissions factors reported for a full-scale single-step nitritation-anammox system (1.7% of N removed) (Kampschreur et al., 2009a) and a lab-scale partial nitritation reactor (1.2e5.2% of ammonium converted) (de Graaff et al., 2010). However, it is higher than the full-scale combined nitritation-anammox process studied in Joss et al. (2009) with an emission factor of 0.4e0.6% of N removed, and a lab-scale one-stage nitritation-anammox system with an emission factor of 0.02% of N removed (Pellicer-Na`cher et al., 2010). In addition, the emission factor of this system is comparable to other full-scale biological nutrient removal plants. For example, the average N2O emission factor of 12 full-scale plants studied by Ahn et al. (2010) was 0.8% of N removed. Unlike other partial nitritation-anammox systems, which were operated with a suspended biomass, the lower N2O emissions observed by Pellicer-Na`cher et al. (2010) were achieved in a membrane-aerated biofilm reactor. Spatially redoxstratified biofilms were formed by applying sequential aeration. This allowed N2O produced by AOB in the inner layer of the biofilm to be consumed by heterotrophic bacteria residing in the outer layer of the biofilm (Pellicer-Na`cher et al., 2010). For suspended biomass systems, the lowest N2O emission factor was reported in a combined partial nitritation-anammox SBR, postulated to be due to the much lower concentration of nitrogen species in the reactor (Joss et al., 2009). The significantly higher N2O emission reported by Desloover et al. (2011) was proposed to be due to (1) high nitrite þ (114e117 mg NO 2 -N/L) and high ammonium (72e82 mg NH4 N/L) concentrations in the system; (2) relatively low DO of approximately 1 mg O2/L and (3) more variable process conditions due to the batch feeding operation mode. However, in this study, the concentrations of nitrite (500e600 mg NO 2N/L) and ammonium (400e500 mg NHþ 4 -N/L) in the SBR were significantly higher and the DO (0.5e0.8 mg O2/L) concentration was significantly lower. These conditions did not result in higher N2O production.
The effect of pH on N2O production in wastewater environment has not been widely studied and has been assumed to play a minor role (Kampschreur et al., 2009b). However, pH has been shown to be a key factor in determining the microbial source of N2O from soil environment (Baggs et al., 2010). In this study, increasing pH within the range of 6.0e8.0 stimulated the N2O production rates of the enriched AOB culture. Similar trends have been observed in a pure culture study using Nitrosomonas europeae where N2O production was lowest at pH 6.0 and highest at pH 8.5 beyond which further pH increase caused a decrease in N2O production (Hynes and Knowles, 1984). A higher operating pH range (7.5e7.8) in the partial nitritation system reported by Desloover et al. (2011) could also be a possible reason for the significantly higher N2O emissions, compared to the other systems that were operated at approximately pH 7.0 (Kampschreur et al., 2008b; Joss et al., 2009). Although chemodenitrification (the chemical breakdown of nitrite, Wrage et al., 2001) was postulated by Desloover et al. (2011) to contribute to the relatively high N2O emission, it is unlikely that this reaction was responsible for the N2O production observed in our system. The chemical decomposition of nitrite depends on the FNA concentration and mainly occurs under acidic conditions (pH < 5.5) (van Cleemput and Baert, 1984; Udert et al., 2005). However, the pH applied in this study was significantly higher: 6.4e7.4 in the parent reactor, and 6.0e8.5 in the batch tests. The pH induced N2O production predicted by the chemodenitrification pathway would be the opposite to what was experimentally observed in this study. Also, in the negative control experiments, it was demonstrated that the presence of nitrite (500 mg NO 2 -N/L) at pH 7.0 and 8.0 in the absence of ammonium did not result in N2O production. The above led to the conclusion that N2O was produced biologically in our system. Two metabolic pathways have been proposed to explain N2O production by AOB. AOB can reduce nitrite and NO to N2O through the so-called nitrifier denitrification pathway, utilising ammonia, molecular hydrogen or hydroxylamine as electron donors (Wrage et al., 2001). This pathway is believed to be used by AOB mainly under oxygen stress conditions (Poth and Focht, 1985; Kampschreur et al.,
5942
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 3 4 e5 9 4 4
2008a), or under high nitrite conditions (Yu and Chandran, 2010). N2O produced in a partial nitritation system has indeed been previously suggested to be from the nitrifier denitrification pathway (Kampschreur et al., 2008b). Alternatively, N2O and NO can also be produced as a by-product by AOB during hydroxylamine oxidation to nitrite either through enzymatic or chemical degradation of reaction intermediates such as nitrosyl radical (NOH) (Poughon et al., 2001). However, this has only been demonstrated using cell free extract (Anderson, 1964; Ritchie and Nicholas, 1972) and is thus far not clearly understood. Unfortunately, our experimental data do not discriminate any of the above two metabolic pathways. While a direct effect of pH on various reactions involved in ammonia oxidation or nitrite reduction could not be ruled out, the linear relationship between the AOR and the N2O production rate suggests that the pH effect could be indirect, through increasing the ammonium oxidation activity. Our experimental data clearly showed that AOR increased when pH increased from 6.0e7.0 to 7.5e8.0 but decreased when pH further increased to 8.5. The pH dependency of AOR on pH observed in this study is in agreement with previous studies, which reported maximum nitrifying activity at pH 7.5e8.2 (Loveless and Painter, 1968; Antoniou et al., 1990). Electrons are made available at a higher rate under a higher AOR. It is possible that more electrons were diverted to the nitrifier denitrification pathway, especially considering the fact that DO applied in our studies was relatively low (0.55 0.05 mg O2/L). This would lead to increased N2O production as AOB likely do not have the ability to reduce N2O to N2 (Poth and Focht, 1985). An alternative explanation is that a higher AOR would lead to higher accumulation, possibly intracellularly, of the reaction intermediates during ammonium oxidation to nitrite, including hydroxylamine (NH2OH) and NOH. This could subsequently result in their faster breakdown to form N2O. Hynes and Knowles (1984) suggested the chemical decomposition of NOH during hydroxylamine oxidation is responsible for N2O production by a pure culture of N. europeae under optimal pH conditions (8.5). The dependency of N2O production on AOR observed in this study is in general agreement with several other studies. Yu et al. (2010) reported that the rapid increase in N2O production during recovery from anoxic to aerobic condition is due to a shift in the ammonium oxidation metabolism from a low specific activity (q < qmax) towards the maximum specific activity (qmax). Su¨mer et al. (1995) found that increased N2O production coincides with increased oxygen concentration in the activated sludge process of a domestic WWTP, which presumably increased the ammonium oxidation rate. Similarly, Kampschreur et al. (2009b) observed that N2O production by AOB in a partial nitration-anammox process decreases with decreasing DO concentrations. Contrasting to other studies that observed increased N2O production with high nitrite concentrations in partial nitritation systems (Chuang et al., 2007; Kampschreur et al., 2008b, 2009a; Desloover et al., 2011), the FNA batch tests showed that an increase in nitrite and FNA concentration did not affect the N2O production rate. FNA is shown to be the main electron acceptor for the denitrification activity of AOB (Shiskowski and Mavinic, 2006) and is carried out as a detoxification mechanism under high nitrite concentrations (Yu and Chandran, 2010). It is
likely that the high background nitrite concentration (450e550 mg NO 2 -N/L) in the SBR was sufficient to support the low nitrite or FNA concentration requirement for nitrifier denitrification, as reflected by the relatively low N2O production rate (0.16e0.53 mg N2O-N/h/g VSS). Since nitrite was not limiting, any further increase in nitrite would unlikely cause an increase in the N2O production rate. Given that the rate of N2O production is two orders of magnitude lower (approximately 100e200 times in this study) compared to the nitrite production rate, it is also unlikely that nitrifier denitrification was carried out for detoxification purpose. Such a low nitrite reduction rate would be in vain to prevent nitrite or FNA toxicity compared to the high nitrite production rate. However, in agreement with other studies (Kampschreur et al., 2008a, Schreiber et al., 2009), the presence of nitrite was important for the observed N2O production in this study as demonstrated in the negative control experiment. Although intracellular NO was detected in this study, the role of NO in N2O production cannot be fully explained due to the lack of liquid and gas phase NO monitoring. Recent study with N. europeae has shown that the regulation mechanism for NO and N2O production is different (Yu et al., 2010). NO is produced mainly during anoxic condition whereas N2O is produced exclusively during the transition from anoxic to aerobic condition. In addition, co-monitoring of NO and N2O in full-scale partial nitritation-anammox systems showed that NO and N2O production by AOB does not necessarily coincide under changing operating conditions (e.g. DO and nitrite concentrations) (Kampschreur et al., 2008b, 2009a). While the gas phase NO remained fairly constant, a 10-fold decrease in N2O concentration from the beginning until the end of the aerated period was observed (Kampschreur et al., 2008b). Although N2O production by AOB has been suggested to be closely coupled to NO production (Schreiber et al., 2009), this has not been consistent in all studies reporting N2O and NO production from AOB cultures. It should be noted that experiments conducted in this study were carried out only in short term batch tests with transient changes in pH. It is not known whether operating the SBR at a constant high pH (e.g. pH 8.0) would lead to adaptation of AOB to cope with high AOR. This may potentially result in the dampening of the N2O production overtime. In fact, pure culture of Alcaligenes faecalis was reported to develop adaptation to changing conditions whereby the N2O emission decreased from 86% to 28% of NO 2 converted after 10 cycles of starvation followed by acetate pulse feeding (Schalk-Otte et al., 2000).
4.3.
A mitigation strategy
Most studies on N2O production from wastewater treatment systems to date have focused on quantifying N2O emissions and on the mechanisms involved and responsible. Few studies have investigated the mitigation of N2O emissions. In this study, reducing the feeding rate during the aerobic phase was shown to substantially reduce N2O production without affecting the overall ammonium to nitrite conversion in this system. The results showed that reducing the feeding rate could be a potential operational strategy for minimising the N2O production during aerobic phases in the nitritation system.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 3 4 e5 9 4 4
The relatively lower pH achieved with the slow-feeding rate was likely responsible for the reduced N2O. Under normal operating conditions, the ammonium oxidation activity increased immediately to 7.4 after feeding, leading to higher N2O production. When the feeding rate was reduced, the AOR was lower due to the non-optimal pH (6.0e7.0) and also controlled supply of alkalinity. The lowered AOR would lead to lowered N2O production as revealed by the experimental data. It is recognised that the lowered feeding rate also lowered the ammonium loading rate. However, the changes in the ammonium loading rate unlikely played a role in reducing the N2O production. The total ammonium concentration in the SBR was maintained at a relatively high concentration of 450 50 mg NHþ 4 -N/L. It is unlikely that an increase in the ammonium concentration in the SBR to 515 20 mg NHþ 4 -N/L during pulse feeding could have affected the AOR and subsequently the N2O production rate. It was indeed shown in the FA batch tests that an increase of the total NHþ 4 concentration to above 500 mg N/L did not have an effect on the N2O production rate under controlled pH and DO conditions.
5.
Conclusions
pH was shown to be an important factor affecting the N2O production by the AOB culture under aerobic conditions. N2O production was shown to vary significantly as pH varied between 6.0 and 8.5, with minimum production between pH 6.0 and 7.0 and maximum at pH 8.0. Such an effect was not related to the corresponding changes in the FA and FNA concentrations. A linear relationship was observed between the N2O production rate and the ammonium oxidation rate within the pH range of 6.0e8.5. This indicated that the tested pH caused an increase in the N2O production rate by promoting the ammonium oxidation rate of the AOB culture. Despite the high ammonium to nitrite conversion rates and constant exposure to relatively high concentrations of ammonium and nitrite, the N2O emission factor of the SBR system (1 0.1% of ammonium converted) is comparable to a normal biological nitrogen removal wastewater treatment plant. Adopting a slow-feeding strategy in the parent reactor to control the pH within the window of 6.0e7.0 was shown to reduce the N2O production during aerobic phases by a factor of 4.
Acknowledgements The authors would like to thank the Australian Research Council (ARC) and the Western Australian Water Corporation for funding this study through project LP0991765 and DPO0987204. Y.L. is an Australian Postgraduate Award recipient.
Appendix. Supplementary material Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.watres.2011.08.055.
5943
references
Ahn, J.H., Kim, S.P., Park, H.K., Rahm, B., Pagilla, K., Chandran, K., 2010. N2O emissions from activated sludge processes, 2008e2009: results of a national monitoring survey in the United States. Environmental Science & Technology 44 (12), 4505e4511. Anderson, J., 1964. The metabolisms of hydroxylamine to nitrite by Nitrosomonas europaea. Biochemical Journal 91, 8e17. Anthonisen, A.C., Loehr, R.C., Prakasam, T.B.S., Srinath, E.G., 1976. Inhibition of nitrification by ammonia and nitrous acid. Journal (Water Pollution Control Federation) 48 (5), 835e852. Antoniou, P., Hamilton, J., Koopman, B., Jain, R., Holloway, B., Lyberatos, G., Svoronos, S.A., 1990. Effect of temperature and ph on the effective maximum specific growth rate of nitrifying bacteria. Water Research 24 (1), 97e101. APHA, 1998. Standard Methods for Examination of Water and Wastewater, 20th ed. American Public Health Association, Washington, DC. Baggs, E.M., Smales, C.L., Bateman, E.J., 2010. Changing pH shifts the microbial sourceas well as the magnitude of N2O emission from soil. Biology and Fertility of Soils 46 (8), 793e805. Blackburne, R., Yuan, Z., Keller, J., 2008. Partial nitrification to nitrite using low dissolved oxygen concentration as the main selection factor. Biodegradation 19 (2), 303e312. Chuang, H.P., Ohashi, A., Imachi, H., Tandukar, M., Harada, H., 2007. Effective partial nitrification to nitrite by down-flow hanging sponge reactor under limited oxygen condition. Water Research 41 (2), 295e302. de Graaff, M.S., Zeeman, G., Temmink, H., van Loosdrecht, M.C.M., Buisman, C.J.N., 2010. Long term partial nitritation of anaerobically treated black water and the emission of nitrous oxide. Water Research 44 (7), 2171e2178. Daims, H., Nielsen, J.L., Nielsen, P.H., Schleifer, K.H., Wagner, M., 2001. In situ characterization of Nitrospira-like nitriteoxidizing bacteria active in wastewater treatment plants. Applied and Environmental Microbiology 67 (11), 5273e5284. Desloover, J., De Clippeleir, H., Boeckx, P., Du Laing, G., Colsen, J., Verstraete, W., Vlaeminck, S.E., 2011. Floc-based sequential partial nitritation and anammox at full scale with contrasting N2O emissions. Water Research 45 (9), 2811e2821. Fux, C., Lange, K., Faessler, A., Huber, P., Grueniger, B., Siegrist, H., 2003. Nitrogen removal from digester supernatant via nitrite e SBR or SHARON? Water Science & Technology 48 (8), 9e18. Fux, C., Velten, S., Carozzi, V., Solley, D., Keller, J., 2006. Efficient and stable nitritation and denitritation of ammonium-rich sludge dewatering liquor using an SBR with continuous loading. Water Research 40 (14), 2765e2775. Ganigue´, R., Lo´pez, H., Balaguer, M.D., Colprim, J., 2007. Partial ammonium oxidation to nitrite of high ammonium content urban landfill leachates. Water Research 41 (15), 3317e3326. Ganigue´, R., Gabarro´, J., Sa`nchez-Melsio´, A., Ruscalleda, M., Lo´pez, H. , Vila, X., Colprim, J., Balaguer, M.D., 2009. Long-term operation of a partial nitritation pilot plant treating leachate with extremely high ammonium concentration prior to an anammox process. Bioresource Technology 100 (23), 5624e5632. Gustavsson, D.J.I., Nyberg, U., la Cour Jansen, J., 2008. Operation for nitritation of sludge liquor in a full-scale SBR. Water Science & Technology 58 (2), 429e444. Hellinga, C., Schellen, A.A.J.C., Mulder, J.W., van Loosdrecht, M.C. M., Heijnen, J.J., 1998. The SHARON process: an innovative method for nitrogen removal from ammonium-rich waste water. Water Science & Technology 37 (9), 135e142. Hynes, R.K., Knowles, R., 1984. Production of nitrous oxide by Nitrosomonas europaea: effects of acetylene, pH, and oxygen. Canadian Journal of Microbiology 30, 1397e1404.
5944
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 3 4 e5 9 4 4
Joss, A., Salzgeber, D., Eugster, J., Kŏnig, R., Rottermann, K., Burger, S., Fabijan, P., Leumann, S., Mohn, J., Siegrist, H., 2009. Full-scale nitrogen removal from digester liquid with partial nitritation and anammox in one SBR. Environmental Science & Technology 43 (14), 5301e5306. Kampschreur, M.J., Poldermans, R., Kleerebezem, R., Van der star, W.R.L., Van Loosdrecht, M.C.M., 2009a. Emission of nitrous oxide and nitric oxide from a full-scale single stage nitritation anammox reactor. Water Science & Technology 60 (12), 3211e3217. Kampschreur, M.J., Temmink, H., Kleerebezem, R., Jetten, M.S.M., van Loosdrecht, M.C.M., 2009b. Nitrous oxide emission during wastewater treatment. Water Research 43 (17), 4093e4103. Kampschreur, M.J., Tan, N.C.G., Kleerebezem, R., Picioreanu, C., Jetten, M.S.M., Van Loosdrecht, M.C.M., 2008a. Effect of dynamic process conditions on nitrogen oxide emission from a nitrifying culture. Environmental Science & Technology 42 (2), 429e435. Kampschreur, M.J., van der star, W.R.L., Wielders, H.A., Mulder, J. W., Jetten, M.S.M., van Loosdretch, M.C.M., 2008b. Dynamics of nitric oxide and nitrous oxide emission during full-scale reject water treatment. Water Research 42, 812e826. Khin, T., Annachhatre, A.P., 2004. Novel microbial nitrogen removal processes. Biotechnology Advances 22 (7), 519e532. Kim, S.W., Miyahara, M., Fushinobu, S., Wakagi, T., Shoun, H., 2010. Nitrous oxide emission from nitrifying activated sludge dependent on denitrification by ammonia-oxidizing bacteria. Bioresource Technology 101 (11), 3958e3963. Kuai, L., Verstraete, W., 1998. Ammonium removal by the oxygenlimited autotrophic nitrificationedenitrification system. Applied and Environmental Microbiology 64 (11), 4500e4506. Lai, E., Senkpiel, S., Solley, D., Keller, J., 2004. Nitrogen removal of high strength wastewater via nitritation/denitritation using a sequencing batch reactor. Water Science & Technology 50 (10), 27e33. Lipschultz, F., Zafiriou, O.C., Wofsy, S.C., McElroy, M.B., Valois, F. W., Watson, S.W., 1981. Production of NO and N2O in soil nitrifying bacteria. Nature (London) 294, 641e643. Loveless, J.E., Painter, H.A., 1968. The influence of metal ion concentrations and pH value on the growth of a Nitrosomonas strain isolated from activated sludge. Journal of General Microbiology 52 (1), 1e14. Ma, Y., Penga, Y., Wanga, S., Yuan, Z., Wanga, X., 2009. Achieving nitrogen removal via nitrite in a pilot-scale continuous predenitrification plant. Water Research 43 (3), 563e572. Mobarry, B., Wagner, M., Urbain, V., Rittmann, B., Stahl, D., 1996. Phylogenetic probes for analyzing abundance and spatial organization of nitrifying bacteria [published erratum appears in Appl Environ Microbiol 1997 Feb;63(2):815]. Applied and Environmental Microbiology 62 (6), 2156e2162. Pellicer-Na`cher, C., Sun, S., Lackner, S., Terada, A., Schreiber, F., Zhou, Q., Smets, B.F., 2010. Sequential aeration of membraneaerated biofilm reactors for high-rate autotrophic nitrogen removal: experimental demonstration. Environmental Science & Technology 44 (19), 7628e7634. Peng, Y.Z., Li, Y.Z., Peng, C.Y., Wang, S.Y., 2004. Nitrogen removal from pharmaceutical manufacturing wastewater with high concentration of ammonia and free ammonia via partial nitrification and denitrification. Water Science & Technology 50 (6), 31e36. Poth, M., Focht, D.D., 1985. 15N Kinetic analysis N2O production by Nitrosomonas europaea: an examination of nitrifier denitrification. Applied and Environmental Microbiology 49 (5), 1134e1141.
Poughon, L., Dussap, C.G., Gros, J.B., 2001. Energy model and metabolic flux analysis for autotrophic nitrifiers. Biotechnology & Bioengineering 72 (4), 416e433. Ritchie, G.A.F., Nicholas, D.J.D., 1972. Identification of the sources of nitrous oxide produced by oxidative and reductive processes in Nitrosomonas europaea. Biochemical Journal 126, 1181e1191. Schalk-Otte, S., Seviour, R.J., Kuenen, J.G., Jetten, M.S.M., 2000. Nitrous oxide (N2O) production by Alcaligenes faecalis during feast and famine regimes. Water Research 34 (7), 2080e2088. Schreiber, F., Loeffler, B., Polerecky, L., Kuypers, M.M.M., de Beer, D., 2009. Mechanisms of transient nitric oxide and nitrous oxide production in a complex biofilm. ISME Journal 3 (11), 1301e1313. Shiskowski, D.M., Mavinic, D.S., 2006. The influence of nitrite and pH (nitrous acid) on aerobic-phase, autotrophic N2O generation in a wastewater treatment bioreactor. Journal of Environmental Engineering and Science 5, 273e283. Su¨mer, E., Weiske, A., Benckiser, G., Ottow, J.C.G., 1995. Influence of environmental conditions on the amount of N2O released from activated sludge in a domestic waste water treatment plant. Cellular and Molecular Life Sciences 51 (4), 419e422. Suzuki, I., Dular, U., Kwok, S.C., 1974. Ammonia or ammonium ion as substrate for oxidation by Nitrosomonas europaea cells and extracts. Journal of Bacteriology 120 (1), 556e558. Tallec, G., Garnier, J., Billen, G., Gousailles, M., 2006. Nitrous oxide emissions from secondary activated sludge in nitrifying conditions of urban wastewater treatment plants: effect of oxygenation level. Water Research 40 (15), 2972e2980. Udert, K.M., Larsen, T.A., Gujer, W., 2005. Chemical nitrite oxidation in acid solutions as a consequence of microbial ammonium oxidation. Environmental Science & Technology 39 (11), 4066e4075. Vadivelu, V.M., Keller, J., Yuan, Z., 2007. Effect of free ammonia on the respiration and growth processes of an enriched Nitrobacter culture. Water Research 41 (4), 826e834. Vadivelu, V.M., Yuan, Z., Fux, C., Keller, J., 2006. The inhibitory effects of free nitrous acid on the energy generation and growth processes of an enriched Nitrobacter culture. Environmental Science & Technology 40 (14), 4442e4448. van Cleemput, O., Baert, L., 1984. Nitrite: a key compound in N loss processes under acid conditions? Plant and Soil 76 (1), 233e241. van Dongen, U., Jetten, M.S.M., van Loosdrecht, M.C.M., 2001. The SHARONeAnammox process for treatment of ammonium rich wastewater. Water Science & Technology 44 (1), 153e160. Van Hulle, S.W.H., Vandeweyer, H.J.P., Meesschaert, B.D., Vanrolleghem, P.A., Dejans, P., Dumoulin, A., 2010. Engineering aspects and practical application of autotrophic nitrogen removal from nitrogen rich streams. Chemical Engineering Journal 162 (1), 1e20. Wagner, M., Rath, G., Koops, H.P., Flood, J., Amann, R.I., 1996. In situ analysis of nitrifying bacteria in sewage treatment plants. Water Science & Technology 34 (1e2), 237e244. Wrage, N., Velthof, G.L., van Beusichem, M.L., Oenema, O., 2001. Role of nitrifier denitrification in production of nitrous oxide. Soil Biology and Biochemistry 33, 1723e1732. Yu, R., Chandran, K., 2010. Strategies of Nitrosomonas europaea 19718 to counter low dissolved oxygen and high nitrite concentrations. BMC Microbiology 10 (1), 70. Yu, R., Kampschreur, M.J., Loosdrecht, M.C.M., Chandran, K., 2010. Molecular mechanisms and specific directionality of autotrophic nitrous oxide and nitric oxide generation during transient anoxia. Environmental Science & Technology 44 (4), 1313e1319.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 4 5 e5 9 5 2
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Reduced iron induced nitric oxide and nitrous oxide emission Marlies J. Kampschreur a,1, Robbert Kleerebezem a, Weren W.J.M. de Vet b, Mark C.M. van Loosdrecht b,c,* a
Delft University of Technology, Department of Biotechnology, Delft, The Netherlands Department of Water Management, Delft University of Technology, Delft, The Netherlands c KWR Watercycle Research, Groningenhaven 7, Nieuwegein, The Netherlands b
article info
abstract
Article history:
Formation of the greenhouse gas nitrous oxide in water treatment systems is predomi-
Received 9 April 2011
nantly studied as a biological phenomenon. There are indications that also chemical
Received in revised form
processes contribute to these emissions. Here we studied the formation of nitric oxide (NO)
5 July 2011
and nitrous oxide (N2O) due to chemical nitrite reduction by ferrous iron (Fe(II)). Reduction
Accepted 27 August 2011
of nitrite and NO coupled to Fe(II) oxidation was studied in laboratory-scale chemical
Available online 12 September 2011
experiments at different pH, nitrite and iron concentrations. The continuous measurement of both NO and N2O emission showed that nitrite reduction and NO reduction have
Keywords:
different kinetics. Nitrite reduction shows a linear dependency on the nitrite concentra-
Nitrous oxide
tion, implying first order kinetics in nitrite. The nitrite reduction seems to be an equilib-
Nitric oxide
rium based reaction, leading to a constant NO concentration in the liquid. The NO
Iron
reduction rate is suggested to be most dependent on reactive surface availability and the
Greenhouse gas emission
sorption of Fe(II) to the reactive surface. The importance of emission of NO and N2O
Anammox
coupled to iron oxidation is exemplified by iron reduction experiments and several
Wastewater
examples of environments where this pathway can play a role.
Drinking water
ª 2011 Elsevier Ltd. All rights reserved.
Sediments
1.
Introduction
Rising atmospheric concentrations of N2O are contributing to global warming and stratospheric ozone destruction, a topic which is recently gaining increased attention (e.g.Wuebbles, 2009). NO and N2O emission from soil, sediments and also wastewater treatment plants are generally attributed to biological nitrification and denitrification processes (Kampschreur et al., 2007; Kester et al., 1997; Tallec et al., 2006). In a recent study on N2O emissions form a full scale anammox reactor (Kampschreur et al., 2008b) we noticed N2O emission that could not be easily explained by biological processes and therefore we
investigated the potential role of iron based chemical nitrite reduction (Tai and Dempsey, 2009; Van Cleemput and Samater, 1996) as potential source of N2O emissions. Ferric iron (Fe(III)) is an important oxidant in anoxic environments both for biological and chemical reactions (Weber et al., 2006). The nitrogen cycle and iron cycle are potentially coupled in anoxic environments due to biological reduction of Fe(III) and chemical or biological reduction of nitrate, nitrite and NO coupled to Fe(II) oxidation. The chemical reduction of nitrate and nitrite is often believed to be of little significance in natural environments, however under certain conditions like increased nitrite concentrations it can be very significant (Van
* Corresponding author. Department of Biotechnology, Delft University of Technology, Delft, The Netherlands. Tel.: þ31 15 278 1618; fax: þ31 15 278 2355. E-mail addresses:
[email protected] (M.J. Kampschreur),
[email protected] (M.C.M. van Loosdrecht). 1 Present address: Waterboard Aa en Maas, Pettelaarpark 70, ’s-Hertogenbosch, The Netherlands. 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.08.056
5946
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 4 5 e5 9 5 2
Cleemput and Samater, 1996). Recently, evidence was generated that nitrous oxide fluxes from a hypersaline lake can be caused by the abiotic reaction of minerals containing Fe(II) with nitrite (Samarkin et al., 2010). In this respect it is important to realize that the end product of chemical nitrite reduction with Fe(II) at near neutral pH is nitrous oxide (N2O) with NO as intermediate (Van Cleemput, 1998). The chemical conversions are: 2þ þ 2Hþ /Fe3þ þ NO þ H2 O NO 2 þ Fe
DG01 ¼ 35:8kJ reaction
1
(1)
NO þ Fe2þ þ 1Hþ /Fe3þ þ 0:5N2 O þ 0:5H2 O DG01 ¼ 38:9kJ reaction
1
ð2Þ
Under standard conditions, the first reaction is thermodynamically unfavorable, but in a natural aqueous system the Fe(III) precipitates (typically as FeOOH) and the reaction is pulled to the product side. Also under anoxic conditions the Fe(III) will be kept very low by immediate biological reduction to Fe(II) (Nealson, 1994). In many studies nitrite and NO reduction coupled to iron oxidation are not separately considered, but only N2O emission is measured. In this study the emission of NO and N2O upon nitrite reduction by Fe(II) was investigated, this with special attention to the dynamic behavior of both nitrite and NO reduction. The influence of environmental parameters like pH, nitrite and iron concentrations on the emissions were considered. The importance of emission of NO and N2O coupled to iron oxidation was here exemplified by iron reduction experiments and several examples of environments where this pathway can play a role.
2.
Materials and methods
The study was conducted in a laboratory-scale reactor with a working volume of 2 L at 35 C (see Fig. 1). Gas flow (either nitrogen gas or air for respectively anaerobic and aerobic conditions) was provided at 0.8 L min1, stirring speed was 500 rpm.
acid
pH On-line analysis of NO, NO2, N2O
base
DO Gas (N2 or air)
MFC Eh Off-line analysis nitrite, ammonium, nitrate, Fe2+
2+
Nitrite or Fe during 30 minutes
V = 1.5 L at end
Start with 1L KNO2 or FeSO4 solution at pH 6.6
Fig. 1 e Experimental set-up.
Two types of experimental approaches were used: (1) an anaerobic reactor containing 1 L 10 mM FeSO4 at pH 2.0 was flushed with N2-gas, and base (NaOH) was supplied with the pH-controller to establish a pH of 6.6. Subsequently, 50 mL of 0.86 M KNO2 solution was added during 30 min using a peristaltic pump. In the following reaction time NO and N2O emissions were monitored until Fe(II) was depleted. Alternatively, (2) the experiment started with an anaerobic reactor containing 1 L KNO2 solution at pH 6.6 (initial nitrite concentrations were between 4 and 43 mM), and a 40 mM FeSO4 solution of pH 2 was dosed during 30 min while maintaining the reactor pH constant at pH 6.6 using base dosage with the pH controller. Practically, the first approach was easier due to the need of less pH control during the course of the experiment when the two reactants nitrite and reduced iron are mixed. As will be exemplified later, the rate of reactions in both experimental approaches was very similar in the two types of experiments. NO and N2O emissions were monitored during the subsequent reaction phase. The pH was controlled during the reaction phase at pH 6.6 by base addition. The reactor was continuously monitored by on-line measurement of dissolved oxygen (DO), pH, Redox potential (Eh), base dosage, and NO and NO2 (by Rosemount Chemiluminescence NOx analyzer) and N2O concentration in the offgas (by GC, see Kampschreur et al., 2008a,b for details). During the experiment ammonium, nitrite and nitrate were analyzed off-line, by standard spectrophotometric methods (Dr. Lange spectrophotometry kit). To check the mass balance of electrons accepted during nitrite and NO reduction and electrons donated by iron oxidation, also dissolved Fe(II) was analyzed in several experiments using the 1,10-phenanthroline method for Fe(II) according to NEN-6482 which is based on (American Public Health Association (APHA), 1975). Samples where taken under full exclusion of oxygen by using air thight sampling tubing and by flushing with nitrogen gas. The DO and Eh potential (350 mV at start and gradually increasing to 20 mV when Fe(II) was depleted) during the experiments also showed that oxygen intrusion into the reactor system was adequately prevented. The iron samples were immediately analyzed as the oxidation reaction with nitrite could not be stopped by acidification as in the case for iron oxidation with oxygen. Stock Fe(II) solution was prepared with FeSO4 acidified to pH < 2 and flushed with nitrogen gas to ensure anaerobic conditions. Fe(III) was calculated from the electron balance, assuming that every N2O that was emitted was created by the oxidation of four Fe(II). The chemical iron oxidation coupled to nitrite reduction was tested at several nitrite and iron concentrations and pH. Additionally, a chemical control at pH 6.6 was performed in absence of iron, which showed no formation of NO and N2O.
3.
Results
3.1.
NO and N2O emission coupled to iron oxidation
Addition of potassium nitrite to a Fe(II) solution immediately led to NO emission and subsequently also N2O emission (see Fig. 2). The products of nitrite reduction with Fe(II) at pH 6.5 are NO and N2O. No ammonium production was observed
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 4 5 e5 9 5 2
5947
Fig. 2 e Example of the nitrite reduction/iron oxidation experiment. Nitrite addition from t [ 0.5e1.0 h (indicated by grey box) to an anaerobic Fe(II) solution at 10 mM Fe(II) and 43 mM Nitrite. A. NO and N2O concentration in off-gas B. Cumulative NO and N2O emitted during the experiment C. Iron and nitrite in the liquid (total mmoles so volume changes are corrected, Fe(III) is calculated based on mass balance), D. cumulative donated electrons (calculated from Fe(II) measurements and accepted electrons (calculated from NO and N2O emission).
probably due to the relatively low pH-values applied, as opposed to ammonium production in experiments at pH higher than 8 (Buresh and Moraghan, 1976; Huang and Zhang, 2006). Electron transfer balances confirmed that iron oxidation (based on Fe(II) measurements) was in good agreement with nitrite reduction to NO and N2O (see Fig. 2D). This observation also confirmed that no significant reduction of nitrite to dinitrogen gas occurred. The NO and N2O gas phase concentrations showed different dynamics. The NO emission started as soon as nitrite was added to the reactor system, whereas N2O emission was only observed after 0.4 h. The N2O formation also stopped much sooner, 3 h before the NO formation terminated due to the depletion of Fe(II). The experiment at pH 6.5, with initially 43 mM nitrite and 10 mM Fe(II), was performed in quadruplicates to analyse the reproducibility of the experiments. The NO emission rates were very reproducible (Fig. 3A see data points at 43 mM nitrite), while the N2O emission rates showed a larger variability (Fig. 3B see data points at 660 ppm NO), which is further discussed in section 4.1. During the experiments precipitation occurred, which led to a color change of the reactor medium from transparent at the start of the experiment, to green during the first hour from the start of iron addition toward dark green and brown when the reaction proceeded and orange when the reactions were terminated (see Fig. 4). The color changes were obviously related to the predominant iron species, which changed during the course of the experiment because the ratio of Fe(II) and Fe(III) changed (see Fig. 2C). The green color can be correlated to green rust formation (e.g. Trolard et al., 1997), a Fe(II)Fe(III)-species, but analysis of the exact mineral composition and crystallinity
went beyond the scope of this study. These experiments clearly showed the coupling between iron oxidation and nitrite reduction leading to NO and N2O as end products at slightly acidic pH’s.
3.2.
Effect of pH, nitrite, NO and Fe(II) concentration
The influence of the concentration of the reactants on the reaction kinetics was tested by several experiments with different initial concentrations. In Fig. 3A the effect of the nitrite concentration on NO formation rate is shown. The dependence of the NO emission rate on the nitrite concentration was linear. The N2O emission rate (and thus the NO reduction rate) was found to be dependent on the NO concentration (Fig. 3B), however the large variability in rates shows that other factors also greatly affected the rate, which we will be discussed in section 4.1. Second order kinetics in NO can be expected based on the reaction mechanism but the reaction order could not be distinguished based on these data. The relation between the Fe(II) concentration and the kinetics of the nitrite and NO reduction is more complicated. Doubling the initial concentration of iron did not greatly affect the reaction rates during the first phases of the experiment, which indicates that nitrite has a stronger effect on the kinetics. The NO emission very steeply decreased when Fe(II) became limiting, which indicates a high affinity of the nitrite reduction for Fe(II). The NO reduction rate is probably strongly dependent on reactive surface availability and the sorption of Fe(II) to the reactive surface (Tai and Dempsey, 2009) as will be discussed in section 4.1.
5948
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 4 5 e5 9 5 2
that there is no real difference in rates when feeding nitrite to a reduced iron solution or viceversa. These experiments showed that the iron reduction rates are greatly affected by the nitrite and NO concentration and that the pH clearly influenced the NO and N2O formation rates.
NO emission rate (mmol e/h)
A 1,6 1,4 y = 0,0328x R2 = 0,9791
1,2 1 0,8 0,6 0,4 0,2 0 0
N2O emission rate (mmol e/h)
B
10
20
30
40
50
nitrite(mmol) 7 6 5
y = 0,0041x R2 = 0,9134
4 3 2 1 0 0
100
200
300
400
500
600
700
NO (ppm)
Fig. 3 e Relation between (A) NO emission rate and the initial nitrite concentration and (B) the N2O emission rate and the average NO concentration in the off-gas (during the calculated period for the N2O emission rate). The experiments were performed at pH 6.6. The open symbols show rates originating from experiments with KNO2solution dosage (type 1, see experimental section), the closed symbols from experiments with FeSO4-solution addition (type 2). The NO and N2O emission rates were calculated for the linear phase of the cumulative NO and N2O emission curve (see Fig. 2B).
The pH clearly affected the rates of nitrite and NO reduction coupled to iron oxidation. At pH 6.6 both NO and N2O production rates were higher than at pH 5.6 and 7.6 (See Fig. 5). At pH 5.6 N2O production is insignificant and all electrons from iron oxidation will go to nitrite reduction leading to NO formation (see supplementary material A for the emission profiles during the experiments at different pH). At pH 7.6 the product of nitrite reduction is predominantly N2O, but the rate of N2O production is lower than at pH 6.6 although the difference might be insignificant due to the large variability of the rate (note the standard deviation of the four measurements at pH 6.6) and the absence of replicates at pH 7.6. The NO emission rate at pH 7.6 is lower than at pH 5.6 and 6.6, however, as all N2O production goes via NO, the nitrite reduction rate at pH 7.6 is higher than at pH 5.6. As pH changes both the speciation of nitrite (by equilibrium with free nitrous acid) and iron (via the formation of different precipitates and iron oxyhydroxide species) the mechanistic basis for the effect of pH cannot be determined based on these experiments. Comparing the open and closed symbols in Fig. 3 indicates
4.
Discussion
4.1.
Reaction kinetics
In the experimental section we showed that a chemical denitrification of nitrite coupled to Fe(II) oxidation can result in the formation of N2O. In the pH range tested NO and N2O were the only products of the chemical denitrification. The emission of NO appeared to be first order in the nitrite concentration whereas for production of N2O a more complicated reaction pattern was observed. There was also no clear direct relation with pH in the range studied (6.6e7.6). The observed reaction kinetics for nitrite and NO reduction showed clearly different kinetics, as can be seen from the difference in NO and N2O emissions in Fig. 2A. The NO emission showed a linear dependency on the nitrite concentration, implying first order kinetics in nitrite (see Fig. 3). There seems to be a constant NO concentration at a certain nitrite concentration independent of the NO reduction rate. Even though the N2O emission rate (and thus the NO consumption by NO reduction) greatly varied at 43 mM nitrite, the NO concentration in solution and NO emission rate were constant (see Fig. 3B for NO concentrations and Fig. 6 for a scheme of the processes involved). The standard Gibbs energy for this reaction calculated at pH 7 is positive for nitrite reduction with Fe(II). The actual Gibbs energy change is likely slightly negative under the reactor conditions, implying equilibrium-based kinetics. Whether nitrite or free nitrous acid is the actual reactive specie cannot be concluded based on our experiments: the nitrite reduction rate was higher at pH 6.6 than at pH 5.6 and 7.6. However, the pH does not only affect the free nitrous acid concentration but also the iron speciation which strongly complicated the comparison of the experiments at different pH. The evaluation of our experiments clearly indicates that the nitrite reduction process is an equilibrium reaction. NO emission and production due to chemical denitrification will therefore be directly related to the stripping conditions in the actual process. The occurrence of linear dependence of NO emission on the stripping rate was observed in a full-scale nitritation reactor (Kampschreur et al., 2008b) and a nitritation-anammox reactor (Kampschreur et al., 2009b). Due to presence of biomass in these systems the role of iron reduction in nitrite reduction could not be separately determined. The NO reduction process is generally assumed to proceed mostly via a surface dependent reaction, in which solid-bound Fe(II) is the primary reactant (Tai and Dempsey, 2009). An important aspect is the sorption density of Fe(II) on the Fe(III) mineral, which can be interpreted as the density of solidbound Fe(II) per unit of Fe(III)-surface. This means that NO reduction will only proceed significantly when sufficient iron
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 4 5 e5 9 5 2
5949
Fig. 4 e Pictures of the reactor as an indication of the colour change. Reactor conditions were the same as for the experiment of which the gaseous en liquid concentration profiles are presented in Fig. 2 (initial concentrations are 10 mM Fe(II) and 43 mM Nitrite). A. 1 min after nitrite addition, B. 10 min after nitrite addition, C. 60 min after nitrite addition, D. 270 min after nitrite addition. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
mineral surface area is available (Sørensen and Thorling, 1991; Tai and Dempsey, 2009). This can also be observed in our experiments; N2O emission only started after a significant amount of iron oxidation by nitrite had occurred, leading to the formation of Fe(III) precipitates. Based on the colour change to green during our experiments (see Fig. 4), the Fe(III) surface could be a Fe(II)Fe(III) mineral like green rust. The decrease in sorption density of Fe(II) can be the reason why NO reduction ceased much earlier than nitrite reduction: when the reaction proceeded both dissolved and surface-bound Fe(II) decreased while surface-bound Fe(III) increased. In Fig. 7, the capacity for green rust formation (a Fe(II)Fe(III)2 mineral was assumed) is plotted, which shows that N2O emission indeed decreases when the capacity for green rust precipitation decreases. The relatively large variability in terms of N2O emissions of the quadruplicate experiments may also be explained by mineral
formation kinetics, which is generally more variable than liquid phase reactions. This may result in a different reactant surface available for NO reduction to N2O.
4.2. Relevance for chemical iron oxidation with nitrite in the environment The chemical iron oxidation coupled to nitrite reduction described here, may play a role in specific natural and manmade environments. As Fe(II) is absent in many environments due to its reactivity, the iron oxidation reaction should be preceded by microbial iron reduction. The process would then be a chemical nitrite and NO reduction with FeII coupled to a biological iron reduction process using organic carbon (or sulfide) as electron donor (Fig. 8). In that case the iron cycle acts as electron mediator. The net conversion is then very
5950
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 4 5 e5 9 5 2
7
9
N2O
5.6 6.6 7.6
5 4
Fe(III)
Fe(II)
400
8
350
7
300
6
250
5
200
4
150
3
100
2
3
N2O (ppm)
mmol e/h
6
450
2 1 0 NO emission rate
N2O emission rate
total emission rate
Fig. 5 e Effect of pH on NO and N2O emission rates in the batch tests (for comparison expressed in moles electrons per hour) at pH 5.6 (n [ 1), 6.6 (average; n [ 4) and 7.6 (n [ 1).
similar to a biological nitrite reduction process which also uses organic carbon as electron donor. Ammonium could be an alternative electron donor for biological iron reduction in certain environments, however pure-culture evidence is still required to support this observation (Shrestha et al., 2009). The iron turnover rate is highly nitrite concentration dependent, but is fast enough to be responsible for significant NO and N2O formation in certain conditions, especially where higher nitrite concentrations are available.
4.2.1.
Wastewater treatment plants
Nitrite accumulation is known to greatly increase N2O emission from wastewater treatment plants (Su¨mer et al., 1995; Kampschreur et al., 2009a). Nitrite accumulation can be caused by a variety of circumstances, like high-nitrogen loading events, low oxygen concentrations during nitrification, substrate limitation during denitrification or presence of toxic substances. Also systems with higher temperatures will be more prone to nitrite accumulation since above 20 C the growth rate of ammonium oxidizing bacteria is faster than for nitrite oxidizing bacteria (Hellinga et al., 1998). Modern wastewater treatment plants include anoxic periods to establish biological nitrogen and phosphate removal. During these anoxic periods Fe(II) from the influent or obtained by biological Fe(III) reduction may induce biological and chemical nitrate and nitrite reduction to NO and N2O. This can also occur in anoxic zones within sludge flocs. The biological iron oxidation with nitrate yields only small amounts of N2O accumulation and predominantly N2 as end-product (Nielsen and Nielsen, 1998). The chemical denitrification yields predominantly N2O as final product as found in this study and earlier reports (Van Cleemput, 1998). Activated sludge has been reported as having a considerable iron reducing activity
NO (g)
N2O (g) 4.
2.
NO2-
1.
Fe(II)
NO (l) Fe(III)
3.
Fe(II)
½ N2O (l) Fe(III)
Fig. 6 e Scheme of the chemical denitrification process 1) nitrite reduction with Fe(II), 2) NO G/L mass transfer, 3) NO reduction with Fe(II), 4) N2O G/L mass transfer.
Fe(II)Fe(III)2
50
1
0
Fe(II), Fe(III), Fe(II)Fe(III)2 potential (mmol)
8
0 0
1
2
3 Time (h)
4
5
6
Fig. 7 e Chemical denitrification batch test. Fe(II) concentration, N2O in off-gas, Fe(III) concentration and capacity for Fe(II)Fe(III)2 formation (calculated based on rough estimation that all Fe(II) and Fe(III) precipitate as green rust when sufficient Fe(II) or Fe(III) is available) during the same experiment as in Fig. 2. Nitrite addition from t [ 0.5e1.0 h, initial concentrations are 10 mM Fe(II) and 43 mM nitrite.
(8e57 mmol Fe.gVSS1 h1) depending on local conditions (Nielsen and Nielsen, 1998) and activated sludge can contain iron in significant amounts (Rasmussen and Nielsen, 1996). Our observation that NO emission is strongly dependent on the NO liquid concentration and the implication of equilibrium-based nitrite reduction kinetics was observed in a full-scale nitritation reactor (Kampschreur et al., 2008b) and a nitritation-anammox reactor (Kampschreur et al., 2009b) for treatment of sludge digestor effluents. In these systems relatively high nitrite concentrations were observed (up to 50 mM in the former and 1 mM in the latter). The observations hint to a role for chemical iron oxidation coupled to nitrite reduction in these wastewater treatment systems.
4.2.2.
Anammox reactor
Recently, it was unexpectedly observed that N2O emission occurs from a full-scale anammox reactor fed with a partially nitrified digester effluent (Kampschreur et al., 2008b). Literature reports indicate that N2O does not play a role in the metabolism of anammox bacteria (Kartal et al., 2007). Since the influent came from a well working aerated nitritation process there was no significant amount of BOD in the influent. Therefore heterotrophic denitrification (partially yielding also N2 as a product) is implausible due to the unavailability of sufficient electron donor. Nitrifier denitrification was mentioned as an alternative route, but the unavailability of traces of oxygen (necessary for the aerobic ammonia oxidation step) would strongly limit the capacity of this pathway (Kampschreur et al., 2008a). Chemical Fe(II) oxidation coupled to nitrite reduction however, can prove an alternative explanation for the N2O emission from the anammox reactor. Iron is present in the reactor influent as Fe(III)-precipitate due to dosage of FeCl3 in earlier stages of the wastewater treatment plant. Iron reduction can be performed by a wide range of
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 4 5 e5 9 5 2
Fig. 8 e Scheme of NO and N2O emission by organic carbon or ammonium oxidation with iron as electron mediator.
organisms including anammox bacteria (Strous et al., 2006; Weber et al., 2006). The iron reduction capacity of full-scale anammox sludge from the Rotterdam treatment plant was tested in this study by incubating the sludge with Fe(III) and formate in a laboratory-scale reactor and measuring the Fe(II) production. This yielded an iron reducing activity of 19 mmol Fe2þ/g. DW/h, which is in the same range as found for activated sludge (Nielsen and Nielsen, 1998). The sum of the specific NO and N2O emission (to the air and the effluent water) from the full-scale anammox reactor was 60 mmol e/g DW/h (based on nitrite reduction to NO and N2O) and depended on the nitrite accumulation in the reactor (approximately around 3 mM nitrite). Hence it can be concluded that the iron reduction capacity measured in the lab-scale reactor is in the same order of magnitude as the electron accepting capacity by chemical denitrification. This indicates that iron reduction can be an important factor to the overall N2O production in the anammox reactor. The NO emission from the anammox reactor is orders of magnitude smaller than the N2O emission, while in the preceding (aerated) nitrititation reactor the production of both gasses is in the same order of magnitude. In the anammox reactor the gas is intensively recycled to ensure mixing of the reactor compartment and only the produced amount of nitrogen gas is discharged. Consequently there is very limited effective stripping of NO and NO can effectively react on with Fe(II) to form N2O. The same phenomenon was also seen in an experiment where infinite gas residence time was applied (batch for both gas and liquid) and subsequently only N2O was formed as final product (Tai and Dempsey, 2009) while in the system with a standard once-through aeration, as applied in this study, large amounts of NO where formed. The rate of chemical iron oxidation with nitrite is strongly dependent on the nitrite concentration, which means that controlling the nitrite concentration at low levels can be an effective measure to prevent (NO and) N2O emission. Additionally, minimising FeCl3 intrusion into a digester effluent treatment system can prevent NO and N2O emission.
4.2.3.
Drinking water trickling filters
Reduction of nitrite by Fe(II) oxidation and subsequent NO and N2O emission might also explain the unexpected nitrogen loss found in a full-scale groundwater trickling filter for drinking water production at the Water Treatment Plant Lekkerkerk of the Drinking Water Company Oasen in the Netherlands. The aim of the trickling filter is iron and manganese removal by oxidation, nitrification of ammonium and methane stripping (de Vet et al., 2009). The trickling filter showed a gap in the
5951
nitrogen balance after the filter material was externally washed to remove accumulated iron oxyhydroxide deposits and iron-oxidizing and nitrifying biomass (see Fig. 2 in supplementary material B). In the weeks after external washing of the filter material, partial ammonium oxidation to nitrite occurred and both Fe(II) and nitrite oxidation rates were significantly decreased. In this period Fe(II) and nitrite were both present in the filter bed. Based on these observations, we hypothesize that the chemical oxidation of Fe(II) by nitrite occurred during the start-up phase of this filter. After the start-up period, an abundant population of iron-oxidizing bacteria developed in the filter bed top, which out-competed the chemical iron oxidation and forced nitrifying bacteria to deeper layers in the filter. Consequently, Fe(II) and nitrite presence did not occur simultaneously anymore and chemical denitrification stopped.
4.2.4.
Soils and sediments
The described abiotic denitrification with Fe(II) can also occur in soils and sediments, especially during nitrite accumulation. High nitrite concentrations are found at sites of increased nitrogen availability, especially when alkaline fertilizers are applied (like urea, ammonium carbonate, diammonium phosphate, ammonium phosphate, and anhydrous NH3 are used) or at urine spots and sites of animal waste application (Van Cleemput, 1998). The capacity of microbial denitrification and dissimilatory nitrate reduction to ammonium (DNRA) is limited in sediments poor in organic matter and consequently abiotic denitrification can be the main process for nitrate and nitrite removal (Ernstsen and Mørup, 1992). Also in agricultural soil 31e75% of N2O emission was estimated to originate from abiotic processes (Venterea, 2007). Fe(II) can originate from activity of iron reducing bacteria, leading to Fe(III) reduction associated with various types of minerals. Alternatively Fe(II) minerals like Pyrite and FeII-III hydroxyl-salts like Fougerite can be present in the soil and sediments. Laboratory experiments indicated that the N2O emission that was recently observed from a hypersaline lake could be caused by the abiotic reaction of nitrite-rich brine with minerals containing Fe(II) (Samarkin et al., 2010). The aspect of environmentally unwanted NO and N2O production by this pathway should be realised when nitrogen removal from nitrogen-rich soils and waters by addition of iron minerals is considered, as proposed in (Ruby et al., 2006). Also addition of iron salts to wetland for phosphate removal as a method for prevention of eutrophication (Rentz et al., 2009) has a possible risk of increased NO and N2O emission.
5.
Conclusion
Chemical denitrification is an often neglected process when considering nitrogen conversion processes in (waste)water treatment processes. For the denitrification as such this is likely correct, but our results point out that when considering formation of the greenhouse gas N2O, chemical iron oxidation should be considered as a potential cause. Especially, in systems where iron is available and where nitrite is present at elevated concentrations, chemical iron oxidation can be a significant cause of N2O formation.
5952
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 4 5 e5 9 5 2
Acknowledgments The authors gratefully acknowledge the contribution of Vikash Anroedh and Mattheus Mimpen in performing part of the presented experiments and Rolf Poldermans and Dana Vejmelkova for their contribution at various stages of the project. The project was supported by SenterNovem (Innowator IWA06005) and the Drinking Water Company Oasen.
Appendix. Supplementary material Supplementary material associated with this article can be found, in the online version, at doi:10.1016/j.watres.2011.08. 056.
references
American Public Health Association (APHA), 1975. Standard Methods for the Examination of Water and Wastewater. Washington, DC. Buresh, R.J., Moraghan, J.T., 1976. Chemical reduction of nitrate by ferrous iron. Journal of Environmental Quality 5, 320e325. de Vet, W.W.J.M., van Genuchten, C.C.A., van Loosdrecht, M.C.M., van Dijk, J.C., 2009. Water quality and treatment of river bank filtrate. Drinking Water Engineering and Science Discussions 2, 127e159. Ernstsen, V., Mørup, S., 1992. Nitrate reduction in clayey till by iron(II) in clay minerals. Hyperfine Interactions 70, 1001e1004. Hellinga, C., Schellen, A.A.J.C., Mulder, J.W., Van Loosdrecht, M.C. M., Heijnen, J.J., 1998. The SHARON process: an innovative method for nitrogen removal from ammonium-rich waste water. Water Science and Technology 37 (9), 135e142. Huang, Y.H., Zhang, T.C., 2006. Nitrite reduction and formation of corrosion coatings in zerovalent iron systems. Chemosphere 64, 937e943. Kampschreur, M.J., Temmink, H., Kleerebezem, R., Jetten, M.S.M., van Loosdrecht, M.C.M., 2009a. Nitrous oxide emission during wastewater treatment. Water Research 43, 4093e4103. Kampschreur, M.J., Picioreanu, C., Tan, N.C.G., Kleerebezem, R., Jetten, M.S.M., van Loosdrecht, M.C.M., 2007. Unraveling the source of nitric oxide emission during nitrification. Water Environment Research 79, 2499e2509. Kampschreur, M.J., Tan, N.C.G., Kleerebezem, R., Picioreanu, C., Jetten, M.S.M., van Loosdrecht, M.C.M., 2008a. Effect of dynamic process conditions on nitrogen oxides emission from a nitrifying culture. Environmental Science & Technology 42, 429e435. Kampschreur, M.J., van der Star, W.R.L., Wielders, H.A., Mulder, J. W., Jetten, M.S.M., van Loosdrecht, M.C.M., 2008b. Dynamics of nitric oxide and nitrous oxide emission during full-scale reject water treatment. Water Research 42, 812e826. Kampschreur, M.J., Poldermans, R., Kleerebezem, R., van der Star, W. R.L., Haarhuis, R., Abma, W.R., et al., 2009b. Emission of nitrous oxide and nitric oxide from a full-scale single-stage nitritationanammox reactor. Water Science & Technology 60, 3211e3217. Kartal, B., Kuypers, M.M.M., Lavik, G., Schalk, J., Op den Camp, H.J. M., Jetten, M.S.M., Strous, M., 2007. Anammox bacteria disguised as denitrifiers: nitrate reduction to dinitrogen gas via nitrite and ammonium. Environmental Microbiology 9, 635e642.
Kester, R.A., Meijer, M.E., Libochant, J.A., De Boer, W., Laanbroek, H.J., 1997. Contribution of nitrification and denitrification to the NO and N2O emissions of an acid forest soil, a river sediment and a fertilized grassland soil. Soil Biology & Biochemistry 29, 1655e1664. Nealson, K.H.a.S.D., 1994. Iron and manganese in anaerobic respiration: environmental significance, physiology, and regulation. Annual Reviews Microbiology 48, 311e343. Nielsen, J.L., Nielsen, P.H., 1998. Microbial nitrate-dependent oxidation of ferrous iron in activated sludge. Environmental Science and Technology 32, 3556e3561. Rasmussen, H., Nielsen, P.H., 1996. Iron reduction in activated sludge measured with different extraction techniques. Water Research 30, 551e558. Rentz, J.A., Turner, I.P., Ullman, J.L., 2009. Removal of phosphorus from solution using biogenic iron oxides. Water Research 43, 2029e2035. Ruby, C., Upadhyay, C., Gehin, A., Ona-Nguema, G., Genin, J.-M.R., 2006. Situ Redox Flexibility of FeII-III Oxyhydroxycarbonate green rust and Fougerite. Environmental Science & Technology 40, 4696e4702. Samarkin, V.A., Madigan, M.T., Bowles, M.W., Casciotti, K.L., Priscu4, J.C., McKay, C.P., Joye, S.B., 2010. Abiotic nitrous oxide emission from the hypersaline Don Juan Pond in Antarctica. Nature Geoscience 3, 341e344. Shrestha, J., Rich, J.J., Ehrenfeld, J.G., Jaffe, P.R., 2009. Oxidation of ammonium to nitrite under iron-reducing conditions in wetland soils: laboratory, field demonstrations, and push-pull rate determination. Soil Science 174, 156e164. Sørensen, J., Thorling, L., 1991. Stimulation by lepidocrocite (gamma -FeOOH) of iron(II)-dependent nitrite reduction. Geochimica et Cosmochimica Acta 55, 1289e1294. Strous, M., Pelletier, E., Mangenot, S., Rattei, T., Lehner, A., Taylor, M.W., et al., 2006. Deciphering the evolution and metabolism of an Anammox bacterium from a community genome. Nature 440, 790e794. Su¨mer, E., Weiske, A., Benckiser, G., Ottow, J.C.G., 1995. Influence of environmental conditions on the amount of N2O released from activated sludge in a domestic wastewater treatment plant. Experientia 51, 419e422. Tai, Y.-L., Dempsey, B.A., 2009. Nitrite reduction with hydrous ferric oxide and Fe(II): stoichiometry, rate, and mechanism. Water Research 43, 546e552. Tallec, G., Garnier, J., Billen, G., Gousailles, M., 2006. Nitrous oxide emissions from secondary activated sludge in nitrifying conditions of urban wastewater treatment plants: effect of oxygenation level. Water Research 40, 2972e2980. Trolard, F., Genin, J.M.R., Abdelmoula, M., Bourrie, G., Humbert, B., H, A., 1977. Identification of a green rust mineral in a reductomorphic soil by M o¨ ssbauer and Raman spectroscopy. Geochimica Cosmochimica Acta 61, 1107e1111. Van Cleemput, O., 1998. Subsoils: chemo- and biological denitrification, N2O and N2 emissions. Nutrient Cycling in Agroecosystems 52, 187e194. Van Cleemput, O., Samater, A.H., 1996. Nitrite in soils: accumulation and role in the formation of gaseous N compounds. Fertilizer Research 45, 81e89. Venterea, R.T., 2007. Nitrite-driven nitrous oxide production under aerobic soil conditions: kinetics and biochemical controls. Global Change Biology 13 (13), 1798e1809. Weber, K.A., Achenbach, L.A., Coates, J.D., 2006. Microorganisms pumping iron: anaerobic microbial iron oxidation and reduction. Nature Reviews Microbiology 4, 752e764. Wuebbles, D.J., 2009. Nitrous oxide: no laughing matter. Science (Washington, DC, United States) 326, 56e57.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 5 3 e5 9 5 8
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Dissolution of D2EHPA in liquideliquid extraction process: Implication on metal removal and organic content of the treated water Po-Ching Lee a, Chi-Wang Li a,*, Jie-Yuan Chen a, Ying-Sheng Li a, Shiao-Shing Chen b a
Department of Water Resources and Environmental Engineering, Tamkang University, 151 Ying-Chuan Road, Tamsui district 25137, New Taipei city, Taiwan b Institute of Environmental Engineering and Management, National Taipei University of Technology, Taipei, Taiwan
article info
abstract
Article history:
Effects of pH, extractant/diluent ratios, and metal concentrations on the extent of
Received 21 April 2011
extractant dissolution during liquideliquid extraction were investigated. Experimental
Received in revised form
result shows that D2EHPA dissolution increases dramatically at pH above 4, leveling off at
11 August 2011
pH 6e7. The phenomenon is consistent with deprotonation of D2EHPA and the domination
Accepted 27 August 2011
of negatively charged D2EHPA species at pH of higher than 4. Concentration of D2EHPA in
Available online 7 September 2011
the aqueous phase, i.e., the extent of extractant dissolution, drops after addition of metal
Keywords:
the organic phase is calculated to be 2.04 mol per mol of Cd added, which is quite closed to
Dissolution
the stoichiometric molar ratio of 2 between D2EHPA and Cd via ion exchange reaction. The
Extraction
effect of metal species on the extent of extractant/metal complexes re-entering is in the
Solvent
order of CdzZn > Ag, which might be coincident to the complexation stability of these
and decreases with increasing metal concentration. The amount of D2EHPA ‘re-entering’
metals with D2EHPA. The extent of extractant dissolution in liquideliquid extraction
Metals
process depends on the type and concentration of metal to be removed, pH of aqueous phase, and extractant/diluent ratios. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
Liquideliquid extraction is widely applied for recovery of precious metals and removal of heavy metals (Chang et al., 2010, 2011, da Silva et al., 2008; Marinho et al., 2010; Sun and Lee, 2011; Xie and Dreisinger, 2009). Extractants generally have low solubility in water, but greatly dissolve under some conditions. During extraction, extractant (i.e., metal carriers) may depart the organic solvent phase (i.e., diluent) entering aqueous phase, where the phenomenon is called ‘extractant dissolution’. Extent of extractant dissolution might be dependent of solution pH, concentration and types of metals,
extractant/diluent weight ratio, and solubility of extractants. Previous studies of liquideliquid extraction mainly emphasized on metal removal efficiency but rarely focused on the phenomenon of extractant dissolution. While extractant dissolution might reduce extraction efficiency of metal, it would definitely increase organic content of treated water. Investigating Cu removal by liquideliquid extraction, Kocherginsky and his coworks (2007) proposed an extraction mechanism called ‘Big carrousel’ model which involves diffusion of extractant into aqueous phase, followed by formation of metal/extractant complexes, and then transportation of metal/extractant complexes back into solvent
* Corresponding author. Tel.: þ886 2 26239343; fax: þ886 2 26209651. E-mail address:
[email protected] (C.-W. Li). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.08.054
5954
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 5 3 e5 9 5 8
phase as shown in Fig. 1 (Ritcey and Ashbrook, 1979). It is hypothesized that increasing concentration of metal/extractant complexes through increasing metal concentration in aqueous phase would mitigate extractant dissolution. The possibility has not been explored before and will be tested in this study. Meanwhile, complexation stability of metal with extractant will also affect extractant dissolution. As shown in Fig. 1, the higher complexation stability constant of metal with extractant, the higher concentration of metal/extractant complex in the aqueous phase (MAn (aq)). As the result, more metal/extractant complexes will be transported back to the organic solvent phase, resulting in reducing extractant dissolution. In this study, three metals, namely Cd (II), Zn (II), and Ag (I) which might have different complexation stability with extractant, are chosen to test this effect. Also shown in Fig. 1, distribution of extractant speciation in the aqueous phase, which is dependent of pH and acidity constant (Ka), would impact the extractant dissolution. Dependent of the type of diluent and content of the aqueous phase, pKa of between 0.47 and 3.01 has been reported for D2EHPA (Biswas et al., 2000; Biswas and Hayat, 2002; Zhang et al., 1995). Since the concentrations of the neutral species of extractant, i.e., protonated species, in the solvent and aqueous phases are governed by distribution coefficient (Kd), and concentration of the negatively charged extractant species increases with increasing pH, the dissolved extractant, i.e., sum of protonated and deprotonated D2EHPA in the aqueous phase, increases with increasing pH. For example, Gum et al. (2000) investigated Pb and Cd extraction by D2EHPA, finding that D2EHPA dissolution increases with increasing pH between 2.5 and 4.5 due to acidic property of D2EHPA. However, the abovementioned phenomena might not be the case for neutral extractant, such as TBP (tributyle phosphate) which is a widely used neutral extractant (Wasewar et al., 2010; Wright and Paviet-Hartmann, 2010), and for anionic extractant, Aliquat 336 (methyl-trioctylammonium chloride). Aliquat 336 is a quaternary ammonium salt, and is always presented as
Solvent Phase
MAn (org)
HAaq Ka
Aqueous Phase
Kd,MA
Kd
HAorg
_
Mn+ + nAaq
Km
MAn (aq)
+ nH+
Fig. 1 e Distribution of extractant in organic and aqueous phases. Adapted from Ritcey and Ashbrook (1979).
a negatively charged species when extract is dissolved in aqueous phase. Therefore, Aliquat 336 dissolution might be also pH-independent. Although dissolution of the above-mentioned extractants, TBP and Aliquat 336, is less pH-dependent, the extent of extractant dissolution depends on types of extractants and their solubility. For example, solubility of D2EHPA in water is between 0.1 and 1% and is around 0.1% for TBP. On the other hand, solubility of Aliquat 336 in water, >10%, is much higher than those of TBP and D2EHPA. The objectives of this study are to investigate the dissolution of D2EHPA under various pH conditions, extractant/ diluent ratios, and metal concentrations. Implication of extractant dissolution on metal removal efficiency is reported.
2.
Experimental section
2.1.
Chemical and materials
All chemicals used were of reagent grade. Ag (I), Cd (II), and Zn (II) stock solutions of 5000 mg/L were prepared using AgNO3 (Merck), Cd(NO3)2.4H2O (Showa Chemical Industry Co. Ltd., Japan) and ZnSO4.7H2O (Merck), respectively. Metalcontaining solution was diluted from stock solution before each experiment. D2EHPA (Fluka) was selected as the extractant for metal removal, and kerosene (CPC corporation, Taiwan) purchased from local gas station was chosen as the organic diluent. Various weight ratios of extractant/diluent (w/w) were prepared. The effects of different D2EHPA/kerosene weight ratios (5:1, 1:1, 1:5, and 1:10) and pHs (3e7) on extractant dissolution were investigated.
2.2.
Experimental methods
Triplicate experiments were conducted with average value from these tests reported. Effect of pH on the extractant dissolution was explored under various extractant/kerosene ratios at fixed extractant dosage of 4 g per L of aqueous phase. In this phase of experiments, only DI was used, i.e., no metal added. For 1:10 D2EHPA/kerosene ratio tested, to achieve D2EHPA of 4 g, 44 of solvent, i.e., mixture of D2EHPA and kerosene, will be added per L of aqueous phase, resulting in oil-to-aqueous ratio (O/A ratio) of 4.4%. The amount of solvent to be added for 5:1, 1:1, and 1:5 D2EHPA/kerosene ratios are 4.8, 8, and 24 g/L, respectively, corresponding to O/A ratios of 0.48, 0.8, and 2.4%. After solvent was added, mixture was circulated and mixed with a centrifugal pump and aeration with pH fixed at pH 3.0. After a 30-min mixing, one sample (w30 ml) was collected, and pH was increased to the next level using 0.1 N NaOH by a pH titrator (AT 420, KEM, Japan). Collected samples were filtered immediately under vacuum through a 1.2 mm GF/C glass microfiber filter paper (Whatman, Middlesex, UK) and then through a 0.45 mm cellulose acetate filter paper (Advantec MFS, Pleasanton, USA). As indicated in our previous study (Lee et al., 2009; Li et al., 2008, 2010), membrane filtration is quite effective for separating emulsified solvent from aqueous solution. Since
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 5 3 e5 9 5 8
extractant dissolution is the main focus of the current study and to make sure no emulsified solvent droplets interfering the result, the filtrate is further polished/filtered with a hydrophilic membranes having MWCO of 10 kDa (YM 10, Millipore, USA) using a 62-mm diameter stirred cell (Model 8200, Millipore, USA) pressurized at 1.38 bar. The filtrate was then analyzed for COD and metal concentration. According to our data, dissolution of kerosene is very low with COD of less than 40 mg/L for an emulsified solution containing 40 g kerosene/L of DI. Therefore, COD detected in the filtrate (from several hundreds to thousands mg/L) is mainly attributed to the dissolution of D2EHPA into aqueous phase. Impact of extractant dissolution on the metal removal efficiency was investigated with D2EHPA dosage of 4 g/L at pHs ranging from 3 to 7 under different D2EHPA/kerosene ratios using different metal ions (concentration of 100 mg/L). Solution was circulated and mixed as stated above. After a 30-min reaction, one sample (10e15 ml) was collected and pre-treated with method shown above before analyzing COD and metal concentration. To investigate the effect of metal concentration on the ‘reentering’ of extractant/metal complexes into diluent phase, concentrated Cd(II) solution (20000 mg/L) was added stepwisely 5-ml each time to an solution with volume of 1 L initially containing DI and D2EHPA of 4 g using D2EHPA/kerosene ratio of 1:10. After each addition, Cd(II) increases about 100 mg. The solution was then mixed for 30 min with pH being kept at 7.0. Solution was circulated and mixed with a centrifugal pump and aeration with pH fixed using 0.1 N NaOH by pH titrator. After a 30-min reaction, one sample (40 ml) was collected and filtered before COD and Cd(II) analysis.
2.3.
Analytical methods
Metal concentration was determined by Flame atomic Absorption (AA) Spectrophotometer (GBC 932 plus, Australia) according to the standard method. The detection limits for Ag (I), Cd (II), and Zn (II) are 0.01, 0.002, and 0.005 mg/L, respectively. Before analysis, five-point calibration curve for each metal was obtained using the corresponding AA standard (J.T. Baker). The concentration of metal in the organic solvent phase was calculated from the difference between its concentration in the aqueous phase before and after extraction. COD was analyzed followed the standard method 5520C (APHA et al., 1998). According to Eq. (1), 1 mol of D2EHPA will consume 16.75 mol of oxygen. In the current test, 4 g/L of D2EHPA was prepared which corresponds to 6.45 g/L as COD if all of D2EHPA was dissolved into aqueous phase. C16 H35 O4 P þ 16:75 O2 /16 CO2 þ 17:5 H2 O þ PO4
3.
Results and discussion
3.1.
Effect of pH and extractant/diluent ratio
(1)
Fig. 2(a) shows dissolution of D2EHPA as functions of pH and extractant/kerosene ratio. As indicated in the figure, dissolution of D2EHPA shows classical S shape distribution of acid/ base protonation/deprotonation speciation in which COD
5955
a
b
Fig. 2 e (a) Dissolution of D2EHPA as a function of pH for various D2EHPA/kerosene weight ratios. Reaction time of 30 min. 4 g of D2EHPA per L of DI water. (b) Dissolution of D2EHPA measured as COD for experimental data and data predicted using Eq. (6) with ka of 10L3.01. Error bars represents one standard deviation from the mean for triplicate experiments.
increases with increasing pH. Meanwhile, it increases with increasing D2EHPA/kerosene weight ratio. As indicated in Fig. 1, neutral species of extractant is distributed between aqueous and organic phases, and concentration of D2EHPA in the organic and aqueous phases can be related using partition coefficient, kd, which is independent of D2EHPA/kerosene weight ratio, as shown in Eq. (2). Meanwhile, concentration (assuming ideal solution) of protonated/deprotonated D2EHPA species in the aqueous phase can be related using acidity constant, ka, as indicated in Reaction (3). kd ¼
½HAo ½HAaq
(2)
ka ¼
þ A H ½HAaq
(3)
The mass balance of D2EHPA is expressed as in Eq. (4) where TOTAm is D2EHPA added in terms of COD mass, i.e., 6.45 g in this study, and volume of aqueous and solvent phases (Vaq and Vo, respectively). Combining Eq. (4) with Eqs. (2) and (3),
5956
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 5 3 e5 9 5 8
one can express TOTAm using COD measured in aqueous phase (OM), Vaq, Vo, kd, and a0 as shown in Eq. (5). a0 is the fraction of D2EHPA neutral species in aqueous phase. Rearranging Eq. (5), OM can be written as a function of TOTAm, kd, ka, and proton concentration shown in Eq. (6) which is then used to fit dissolution of D2EHPA under various pH and extractant/kerosene ratios with kd being the only fitting parameter. To fit the data, TOTAm is fixed at the value of 6450, and two ka values of 100.47 and 103.01 are used, respectively. As shown in Fig. 2b, the model predicts the experimental value reasonably well with the only fitting parameter, kd. The partition coefficient, kd, are in the range of 102.96w105.49 for ka values ranging from 100.47 to 103.01. It is quite closed to the range of 103w105 for the partition coefficient of D2EHPA reported by others (Biswas and Hayat, 2002). TOTAm ¼ ½HAaq þ A aq Vaq þ ½HAo Vo
(4)
TOTAm ¼ ½HAaq þ A aq Vaq þ kd ½HAaq Vo ¼ OM Vaq þ kd a0 OM Vo
(5)
TOTAm
OM ¼
Vaq þ kd Vo
3.2.
þ H ka þ Hþ
(6)
Implication on metal removal
Effect of extractant dissolution on the metal removal was tested at pH of 3e7 using three kids of metal ions, namely, Ag(I), Cd(II), and Zn(II). As indicated in Fig. 3, extent of extractant dissolution is quite different for system with or without metals, and is also dependent of the type of metal. Degree of extractant dissolution is in the order of system without metal > Ag-system > Zn-system > Cd-system. At pH of 7.0, COD in the systems are 6431, 6036, 5461, and 4937 mg/L, respectively, for systems without metal and systems with Ag, Zn, and Cd. According to the ‘Big carrousel’ model (Kocherginsky and Yang, 2007), extractant in the aqueous 100
phase will form complexes with metal ions and transport back into solvent phase (Ritcey and Ashbrook, 1979). Based on the COD data, complexation stability of metal with D2EHPA should be in the order of Cd > Zn > Ag, which is also coincident to the removal efficiency of these metals shown in Fig. 3. The removal efficiency of metal is in the order of Cd > Zn > Ag for the most of pHs except at pH 3 where Cd removal efficiency is less than that for Zn. It has been shown that Zn is readily removed by D2EHPA at low pH with 50% of removal efficiency at pH of 0.9, i.e., pH1/2 of 0.9 (Pereira et al., 2007). On the other hand, pH1/2 for Cd is around 3.0. It is interested to note that removal efficiency of Ag (I) increases with increasing pHs from 3 to 5 but decreases thereafter. The observed phenomena might relate to the dissolution of D2EHPA at higher pHs as indicated by the increasing D2EHPA concentration in the aqueous phases and inefficacy of forming Ag/D2EHPA complexes. To validate this argument, Ag(I) removal was studied with fixed amount of D2EHPA at pH 5.0 but under different D2EHPA/kerosene ratios where D2EHPA dissolution is function of D2EHPA/kerosene ratio (see Fig. 2a). As indicated in Fig. 4, Ag(I) removal efficiency decreases linearly (R2 of 0.76) with increasing D2EHPA dissolution as represented as D2EHPA distribution between aqueous and kerosene phases. Based on the finding above, the impact of extractant dissolution on metal removal efficiency is marginal if the complexation stability of the metal with extractant is strong such as in the cases of Cd and Zn but will have profound impact if the complexation stability of the metal with extractant is weak as in the case of Ag.
3.3. Effect of metal concentration on extractant dissolution As shown in previous section, extractant dissolution is suppressed with presence of metal ions especially Cd(II). It is hypothesized that increasing concentration of metal/extractant complexes through increasing metal concentration in
7000
90
6000 5000
70 60
4000
50
Ag-R%
40
Cd-R%
30
Zn-R%
3000 2000
Ag-COD
20
Cd-COD
10
Zn-COD
COD (mg/L)
Removal Efficiency (%)
80
1000
w/o metal-COD
0
0 3
4
5 pH
6
7
Fig. 3 e Metal removal efficiency and COD dissolution as a function of pHs. Initial metal concentration [ 100 mg/L. D2EHPA/kerosene ratio [ 1:10. 4 g of D2EHPA added per L of aqueous solution. Reaction time of 30 min. Error bars represents one standard deviation from the mean for triplicate experiments.
Fig. 4 e Removal efficiency of Ag (I) as a function of D2EHPA dissolution. Initial Ag concentration [ 100 mg/L 4 g of D2EHPA added per L of aqueous solution. Reaction time of 30 min.
1.0
100
8000
0.9
90
7000
0.8
80
0.6 0.5 0.4 0.3 0.2
6000
70
5000
60 50
Cd/D2EHPA molar ratio
40 30
R (%)
20
COD
0.1
10
0.0
0
4000 3000
COD (mg/L)
0.7
Removal efficiency (%)
Cd / D 2EHPA molar ratio
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 5 3 e5 9 5 8
2000 1000
0
400
800 Cd added (mg)
1200
0 1600
Fig. 5 e Cd (II) removal and dissolution of D2EHPA as function of initial Cd (II) concentration. D2EHPA/kerosene ratio [ 1:10. pH [ 7.0. Error bars represents one standard deviation from the mean for triplicate experiments.
aqueous phase would mitigate extractant dissolution. To investigate the effect of metal concentration on the ‘reentering’ of extractant/metal complexes into diluent phase, concentrated Cd(II) solution (20000 mg/L) was added stepwisely 10-ml each time, i.e., 200 mg Cd(II) added each time, to n solution with initial DI volume of 1 L containing 4 g of D2EHPA using D2EHPA/kerosene ratio of 1:10. As indicated in Fig. 5, COD in the aqueous phase decreases linearly with the amount of Cd added up to 800 mg. The amount of D2EHPA ‘reentering’ the organic phase is calculated to be 2.04 mol per mol of Cd added with R2 of 0.95. The value is quite closed to the stoichiometric molar ratio of 2 between D2EHPA and Cd via ion exchange reaction (Pereira et al., 2007). Also shown in Fig. 5, Cd removal efficiency drops below 100% after more than 600 mg of Cd added, and COD in the aqueous phase are all less than 300 mg/L. Calculation of the Cd and D2EHPA molar ratio in the solvent phase shows that the saturated Cd and D2EHPA molar ratio is around 0.63, i.e., 1.6 mol of D2EHPA per mol of Cd. This value is less than the stoichiometric molar ratio of 2 between D2EHPA and Cd in the solvent phase (Pereira et al., 2007). The reason for the discrepancy might be arose when aqueous D2EHPA is in very low concentration and precipitation of Cd(OH)2 at pH 7 might be prevalence at low ligand (D2EHPA) concentration. As the consequence, portion of Cd removed in the form of Cd(OH)2(s) solid is counted as if they are removed by extraction into solvent phase. As the result, molar ratio of D2EHPA and Cd in the solvent phase is less than the stoichiometric molar ratio of 2.
4.
Conclusions
D2EHPA dissolution increases dramatically at pH above 4, levels off at pH 6e7, and is consistent with deprotonation of D2EHPA at pH of higher than 4, showing classical S shape distribution of acid/base protonation/deprotonation speciation. It also increases with increasing D2EHPA/kerosene weight ratio. COD in the aqueous phase was modeled under various pH and extractant/kerosene ratios with kd of the only fitting parameter. The developed model fitted the
5957
experimental value reasonably well with kd of 102.96w105.49 depending on the value of ka chosen. Extent of extractant dissolution is in the order of system without metal > Ag-system > Zn-system > Cd-system. The result might be related to the complexation stability of metal with D2EHPA. The impact of extractant dissolution on metal removal efficiency is marginal for Cd and Zn which have strong complexation stability with D2EHPA. However, the effect is very profound for Ag which has lower complexation stability with D2EHPA. As the result, Ag(I) removal efficiency decreases linearly with increasing D2EHPA dissolution. Increasing metal concentration in aqueous phase would mitigate extractant dissolution. By adding Cd in the aqueous phase sequentially, COD in the aqueous phase decreases accordingly. The amount of D2EHPA ‘re-entering’ the organic phase is 2.04 mol per mol of Cd added, which is quite closed to the stoichiometric molar ratio of 2 between D2EHPA and Cd via ion exchange reaction. COD in the aqueous phase decreases from around 7000 mg/L to less than 300 mg/L after 800 mg of Cd added.
Acknowledgment This work is funded by the National Science Council (Taiwan) under Grant No. NSC 95-2221-E-032-024 and 96-2221-E-032010-MY3.
references
APHA, AWWA, WEF (Eds.), 1998. Standard Methods for the Examination of Water and Wastewater. American Public Health Association: American Water Works Association: Water Environment Federation, Washington, DC. Biswas, R.K., Habib, M.A., Islam, M.N., 2000. Some physicochemical properties of (D2EHPA). 1. Distribution, dimerization, and acid dissociation constants of D2EHPA in a Kerosene/0.10 kmol m3 (Naþ, Hþ)Cl system and the extraction of Mn(II). Ind. Eng. Chem. Res. 39 (1), 155e160. Biswas, R.K., Hayat, M.A., 2002. Kinetics of solvent extraction of zirconium(IV) from chloride medium by D2EHPA in kerosene using the single drop technique. Hydrometallurgy 65 (2e3), 205e216. Chang, S.H., Teng, T.T., Ismail, N., 2010. Extraction of Cu(II) from aqueous solutions by vegetable oil-based organic solvents. J. Hazard. Mater. 181 (1e3), 868e872. Chang, S.H., Teng, T.T., Ismail, N., Alkarkhi, A.F.M., 2011. Selection of design parameters and optimization of operating parameters of soybean oil-based bulk liquid membrane for Cu(II) removal and recovery from aqueous solutions. J. Hazard. Mater. 190 (1e3), 197e204. da Silva, G.C., Cunha, J.W.S.D.d., Dweck, J., Afonso, J.C., 2008. Liquideliquid extraction (LLE) of iron and titanium by bis-(2ethyl-hexyl) phosphoric acid (D2EHPA). Miner. Eng. 21 (5), 416e419. Gum, T., Oleinikova, M., Palet, C., Valiente, M., Mun˜oz, M., 2000. Facilitated transport of lead(II) and cadmium(II) through novel activated composite membranes containing di-(2-ethyl-hexyl) phosphoric acid as carrier. Analytica Chim. Acta 408 (1e2), 65e74. Kocherginsky, N.M., Yang, Q., 2007. Big Carrousel mechanism of copper removal from ammoniacal wastewater through supported liquid membrane. Sep. Purif. Technol. 54 (1), 104e116.
5958
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 5 3 e5 9 5 8
Lee, P.-C., Li, C.-W., Chen, S.-S., Chiu, C.-H., 2009. Compressed airassisted solvent extraction (CASX) for Chromate removal: regeneration and recovery. Separ. Sci. Technol. 44 (16), 3911e3922. Li, C.-W., Chen, Y.-M., Hsiao, S.-T., 2008. Compressed air-assisted solvent extraction (CASX) for metal removal. Chemosphere 71 (1), 51e58. Li, C.-W., Chiu, C.-H., Lee, Y.-C., Chang, C.-H., Lee, Y.-H., Chen, Y.M., 2010. Integration of ceramic membrane and compressed air-assisted solvent extraction (CASX) for metal recovery. Water Sci. Tech. 62 (6), 1274e1280. Marinho, R.S., Afonso, J.C., da Cunha, J.W.S.D., 2010. Recovery of platinum from spent catalysts by liquid-liquid extraction in chloride medium. J. Hazard. Mater. 179 (1e3), 488e494. Pereira, D.D., Rocha, S.D.F., Mansur, M.B., 2007. Recovery of zinc sulphate from industrial effluents by liquid-liquid extraction using D2EHPA (di-2-ethylhexyl phosphoric acid). Sep. Purif. Technol. 53 (1), 89e96.
Ritcey, G.M., Ashbrook, A.W., 1979. Solvent Extraction: Principles and Applications to Process Metallurgy. Elsevier Scientific Pub. Co, Amsterdam. Sun, P.P. Lee, M.S. 2011. Separation of Pt(IV) and Pd(II) from the loaded Alamine 336 by stripping. Hydrometallurgy doi:10. 1016/j.hydromet.2011.06.003 in press. Wasewar, K.L., Shende, D., Keshav, A., 2010. Reactive extraction of itaconic acid using tri-n-butyl phosphate and aliquat 336 in sunflower oil as a non-toxic diluent. J. Chem. Technol. Biot. 86 (2), 319e323. Wright, A., Paviet-Hartmann, P., 2010. Review of physical and chemical Properties of Tributyl Phosphate/Diluent/Nitric acid systems. Separ. Sci. Technol. 45 (12), 1753e1762. Xie, F., Dreisinger, D., 2009. Copper solvent extraction from Waste Cyanide solution with LIX 7820. Solvent Extr. Ion Exch. 27 (4), 459e473. Zhang, P., Inoue, K., Tsuyama, H., 1995. Recovery of metal values from spent hydrodesulfurization catalysts by liquid-liquid extraction. Energy & Fuels 9 (2), 231e239.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 5 9 e5 9 6 8
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Isolation, identification and utilization of thermophilic strains in aerobic digestion of sewage sludge Shugen Liu a,b, Nanwen Zhu a,*, Loretta Y. Li b, Haiping Yuan a a b
School of Environmental Science and Engineering, Shanghai Jiao Tong University, No. 800, Dongchuan Rd., Shanghai 200240, China Department of Civil Engineering, University of British Columbia, Vancouver, Canada V6T 1Z4
article info
abstract
Article history:
Two representative thermophilic bacterial strains (T1 and T2) were isolated from a one-
Received 22 May 2011
stage autothermal thermophilic aerobic digestion pilot-scale reactor. 16S rRNA gene
Received in revised form
analysis indicated that they were Hydrogenophilaceae and Xanthomonodaceae. These
18 August 2011
isolated strains were inoculated separately and/or jointly in sewage sludge, to investigate
Accepted 29 August 2011
their effects on sludge stabilization under thermophilic aerobic digestion condition. Four
Available online 3 September 2011
digestion conditions were tested for 480 h. Digestion without inoculation and inoculation with strain T2, as well as joint- inoculation with strains T1 and T2, achieved 32.6%, 43.0%,
Keywords:
and 38.2% volatile solids (VS) removal, respectively. Removal in a digester inoculated with
Thermophilic strains
stain T1 only reached 27.2%. For the first 144 h, the three inoculated digesters all experi-
Sewage sludge
enced higher VS removal than the digester without inoculations. Both specific thermophilic
Thermophilic aerobic digestion
strains and micro-environment significantly affected the VS removal. DGGE profiles
Removal of volatile solids
revealed that the isolated strains T1 and T2 can successfully establish in the thermophilic
Microbial community
digesters. Other viable bacteria (including anaerobic or facultative microbes) also appeared in the digestion system, enhancing the microbial activity. ª 2011 Published by Elsevier Ltd.
1.
Introduction
The treatment and disposal of sewage sludge represents a rising challenge for wastewater treatment plants (WWTPs) with increased consideration for economic, environmental and regulatory factors. Thermophilic aerobic digestion processes, especially autothermal thermophilic aerobic digestion (ATAD), are promising for sewage sludge stabilization in small- and medium-sized WWTPs due to efficient pathogen inactivation, rapid removal of volatile solids (VS), high energy consumption and simple control requirements (USEPA, 1990). Since ATAD process was first introduced in the early 1970s, it has been adopted by numerous small and medium-sized WWTPs in Europe and North America (Kelly and Mavinic, 2003).
* Corresponding author. Tel./fax: þ86 21 34203732. E-mail address:
[email protected] (N. Zhu). 0043-1354/$ e see front matter ª 2011 Published by Elsevier Ltd. doi:10.1016/j.watres.2011.08.052
Significant progresses (Layden et al., 2007) have been made in optimizing and adapting ATAD process for sewage sludge treatment in order to achieve Class-A biosolids. Most research pertaining to two-stage or multiple-stage ATAD system has focused on laboratory-scale or pilot-scale process performance (Kelly et al., 1993; Mavinic et al., 2001), mathematical modeling (Gomez et al., 2007; Rojas et al., 2010), and the dewatering characteristics of biosolids (Zhou, 2003; Saurabh, 2004). With the continuous evolution of ATAD technology, a one-stage ATAD process, requiring only a simple and small unit, has been developed and patented in recent years (Zhu et al., 2005; Cheng et al., 2009). Experimental results (Cheng, 2006) indicated 40% VS removal in 12e16 d in the one-stage ATAD system, achieving the same stabilization effects as earlier two-stage and multi-stage ATAD processes. Although
5960
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 5 9 e5 9 6 8
this one-stage ATAD process has potential for application to sewage sludge treatment and has a number of advantages over conventional aerobic and anaerobic digestion processes, no one-stage ATAD system are currently implemented due to lack of clarity regarding the stabilization mechanism and insufficient guidelines for practical operation (Liu et al., 2010). Much research has focused on the microbial analysis for thermophilic aerobic digestion process to improve understanding of the sludge stabilization mechanism. Thermophilic digestion temperature plays a primary role in selecting both the individual species and in specifying the overall bacterial diversity (LaPara et al., 2000). Thermophilic digestion systems generally show low levels of biodiversity (Liu et al., 2010; Hayes et al., 2011). The predominant thermophilic populations in digestion processes are Hydrogenophilaceae, Thermotogaceae, Clostridiaceae, and the genus Bacillus such as Bacillus stearothermophilus (Kim et al., 2002), Brevibacillus (Li et al., 2009), Ureibacillus (Liu et al., 2010), and Geobacillus (Hayes et al., 2011). Some of these thermophilic stains could produce thermostable protease (Kim et al., 2002; Li et al., 2009) enhancing the efficiency of sludge thermophilic aerobic digestion. Liu et al. (2009, 2010) analyzed the microbial diversity in a simulated one-stage ATAD system and found that the bacterial consortia experienced a gradual shift as digestion conditions (such as temperature) varied. This finding is supported by Hayes et al. (2011); however, the impact of these variations on the treatment efficiency and on metabolic properties remains unclear. The physiological state of thermophilic microbes in ATAD system has not been fully established. In all likelihood, the bacterial consortia are comparable to that found in composting biomass. Air is routinely supplied for the digestion system, it is still not fully understood whether the viable thermophilic microbes in ATAD reactor are aerobic or anaerobic. However, molecular analysis of bacterial communities (Liu et al., 2010; Hayes et al., 2011) has revealed the presence of both aerobic and anaerobic thermophiles. Though an increasing number of investigations have been conducted to characterize and comprehend the microbial communities of ATAD systems, there is limited evidence on whether most of thermophilic microbes can be isolated from the ATAD reactor and whether those thermophiles have a similar effect on VS reduction. To better elucidate the effects of thermophilic microbes on sludge stabilization process, this study successfully isolated two thermophilic strains from a one-stage ATAD pilot-scale reactor, and categorized them using 16S ribosomal RNA gene sequence analysis. The isolated strains were then separately and/or jointly inoculated into sewage sludge to test for their effects on digestion under thermophilic aerobic digestion condition, followed by analysis of the bacterial community to gain insight into the biological aspects of the digestion process.
treatment. After the sampling sludge was diluted 10-fold with distilled water, the isolates were cultured by streaking onto a Luria-Bertani (LB) solid medium containing (per a liter of distilled water) 10 g tryptone, 5 g yeast extract, 5 g sodium chloride, and 30 g agar (Manaia et al., 2003; Li et al., 2009). Plates were incubated at 55 C for 48 h, then the representative strains of all colony types that could be distinguished on plates were isolated by sub-culturing onto the same LB agar plates and at the same temperature until eventually a single colony formed, differing in appearances. The isolated typical strains were inoculated onto LB agar slants and incubated at 37 C for 48 h, then preserved in refrigerator at 4 C. A loopful of the bacterial culture from the LB slant was inoculated into the 500-mL Erlenmeyer flask containing 200 mL of LB liquid medium and incubated on a rotary shaker at 100 g, 55 C for 24 h. The culture broth was concentrated at 4000 g for 5 min, and then the genomic DNA of thermophilic strains was extracted with 3S DNA Isolation kit V2.2 (Shanghai Biocolor BioScience and Technology Company, Shanghai) following the manufacturer’s instructions (Liu et al., 2010). The extracted genome was used as the template for 16S rRNA amplification with universal primers, BSF8/20 and BSR1541/20 (BSF8/20: 50 AGAGT TTGATCCTGG CTCAG-30 , BSR1541/20: 50 -AAGGA GGTGATCCAGCCGCA-30 ) (Wuyts et al., 2002; Liu et al., 2010). The PCR products were purified and sequenced by Shanghai Invitrogen Biotechnology Co. Ltd., China. Sequence comparisons were conducted with the BLAST search option in the NCBI nucleotide sequence database (http://blast.ncbi.nlm.nih.gov/ Blast.cgi). After the sequences were aligned with Clustal X 1.83 software (Chenna et al., 2003), an evolutionary distance tree was created based on the neighbor-joining method using MEGA 4.1 (Kumar et al., 2001), and the reference sequences used in tree construction were acquired from GenBank. The nucleotide sequences of the two thermophilic strains, T1 and T2, have been deposited in the NCBI database under accession number HQ436531 and HQ436532, respectively.
2.2. Inoculation and startup of thermophilic aerobic digestion process Sewage sludge was sampled from the sludge thickening tank of Minhang municipal wastewater treatment plant in Shanghai, China in which anaerobiceanoxiceoxic process was applied to treat domestic wastewater. Prior to use, the sludge was sieved to remove matter coarser than 0.5 mm, then centrifuged at 2200 g for 3 min yielding total solids (TS) levels between 5 and 6% (Staton et al., 2001). The main characteristics of the sludge sample are presented in Table 1.
Table 1 e Physico-chemical properties of the sludge employed in thermophilic aerobic digestion.a
2.
Experimental methods
2.1.
Isolation and identification of thermophilic strains
Parameter pH
Value
Thermophilic microorganisms were isolated from a one-stage ATAD pilot-scale reactor of effective volume 10 m3, operated continuously over a 6-month period for sewage sludge
6.5
VS SCOD TN Water TS content (g L1) (g L1) (mg L1) (mg L1) (%) 94.2
57.6
42.2
1615
173
a SCOD, soluble chemical oxidation demand; TN, supernatant total nitrogen.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 5 9 e5 9 6 8
The main body of the thermophilic aerobic digestion reactor (Fig. 1) was a cylindrical stainless steel digester submerged in a water bath to maintain a certain digestion temperature. The diameter and height of the digester were 180 and 250 mm, respectively, and the effective volume was 5 L due to potential foaming. Before the startup of the thermophilic aerobic digestion process, two representative isolated thermophilic stains, T1 and T2 were each required to proliferate for 48 h. A loopful of the preserved strain from the LB slant was inoculated into the 500-mL Erlenmeyer flask containing 200 mL of LB liquid medium and incubated on a rotary shaker at 100 g and 37 C for 12 h, followed by 55 C for 36 h. The culture broth was next concentrated at 4000 g for 15 min, and then the pellet was diluted with distilled water to achieve an OD600 of 0.8. After 5 L sludge sample was transferred to each of four digesters, 150 mL diluted T1 and T2 cultures were inoculated into digester R1 and R2, respectively. Digester R3 was inoculated with 75 mL each of diluted T1 and T2 cultures. Digester R0 received no inoculation. Each digester was aerated with a flow rate of 0.10e0.13 L min1 (Cheng, 2006; Liu et al., 2010) by means of a microporous diffuser. All digesters maintained a temperature of 55 C and a constant stirring speed of 30 revolutions per minute during the entire digestion process. The pH was not regulated. Evaporation losses were made up by adding distilled water before each sampling. The pH and oxidation reduction potential (ORP) of the digestion system were monitored every 12 h, and sludge samples were collected every 48 h to determine VS, as well as SCOD and TN of supernatant. The samples collected at 0, 96, 240 and 480 h were chosen to analyze the microbial communities.
2.3.
Chemical analysis
TS and VS were determined in accordance with Standard Methods (APHA et al., 2005). The pH and ORP of the sludge was
5961
measured by a pH meter (pHs-3C, Leici Co., Ltd., Shanghai) and an ORP meter (ORP-502, Ruosull Technology Co., Ltd., Shanghai). The samples were centrifuged for 15 min at 12,000 g, then filtered through 0.45 mm mixed cellulose ester membranes. The filtrate was collected to measure SCOD and TN. SCOD was measured by the standard reflux titrimetric method (SEPAC, 2002a), whereas TN was determined by a TOC/TN analyzer (Analytik jena multi N/C 3000, Germany).
2.4.
Molecular biological analysis
2.4.1.
Extraction of genomic DNA
Genomic DNA of the digestion sludge samples was extracted using a modification of a previously described method (Elisabeth and Patrick, 2000). Prior to adding lysozyme, samples were pre-washed twice with TE buffer (10 mM TriseHCl, 1 mM EDTA, pH 8.0) (Hayes et al., 2011; Piterina et al., 2010). Pellets were re-suspended in 600 ml lysozyme buffer (50 mM TriseHCl, 5 mM EDTA, pH 8.0), and 0.3 g of acidwashed beads were added to the mixture. Next, 20 min of high speed vortexing was performed in a 5-mL tube to physically disintegrate sludge granules. 15 ml of 200 mg mL1 lysozyme was added, and then the mixture was incubated at 37 C for 2 h, with shaking at intervals. Following this step, 10 ml of 20% SDS and 3 ml of 20 mg mL1 protease K were added, and the sample was incubated at 55 C for 30 min. Next, 100 ml of 5 M NaCl was added and mixed in gently, and then 100 ml of 3 M potassium acetate was added into the mixture. After incubation on ice for 30 min to precipitate the proteins, the tube was centrifuged at 12,000 g for 5 min. The supernatant was transferred to a new sterile tube and extracted with an equal volume of 25:24:1 phenol:chloroform:isoamyl alcohol. The phases were mixed gently for 5 min at room temperature and then separated by centrifugation at 13,000 g for 10 min. The aqueous phase was then re-extracted with an equal
Fig. 1 e Schematic diagram of thermophilic aerobic digestion for sewage sludge treatment.
5962
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 5 9 e5 9 6 8
volume of 24:1chloroform:isoamyl alcohol. The supernatant was transferred to a sterile 1.5-mL tube, with an equal volume of isopropanol added again to precipitate the total nucleic acid. After incubation at 20 C for 20 min, the resulting precipitate was collected by centrifugation at 12,000 g for 15 min and washed twice with 70% ethanol to remove traces of isopropanol. After total evaporation of the ethanol at 37 C, the genomic DNA pellets were re-suspended in 40 ml of 10 mM TE buffer for further analysis.
2.4.2.
for 24 h. PCR amplification was performed using 1.0 ml mixtures of extracted DNA solutions as templates with primers 518R and 338F (without GC clamp). The PCR products were purified and sequenced by Shanghai Invitrogen Biotechnology Co. Ltd. The closest phylogenetic affiliations of the sequences recovered from DGGE gels were analyzed with the BLAST search option in the NCBI nucleotide sequence database. The accession numbers here for sequences submitted to NCBI database were JF682323e682340.
PCR-DGGE
PCR amplification of bacterial 16S rRNA V3 region gene fragments (Muyzer et al., 1993; Hayes et al., 2011) was carried out in an MJ Mini Thermal Cycler (Bio-Rad Corp., USA). Firstly, the 1e10 ng extracted genomic DNA was used as the template for 16S rRNA amplifications with universal primers BSF8/20 and BSR1541/20. PCR amplification was performed in a 50-ml reaction mixture containing 1.5 U TaKaRa Taq DNA polymerase (TaKaRa Code: DR001A, TaKaRa Biotechnology Dalian Co. Ltd., China), 5 ml 10 PCR buffer (Mg2+plus), 4 ml dNTP (each 2.5 mM), and 1.5 ml each primer (20 mM) (Shanghai Invitrogen Biotechnology Co. Ltd., China). PCR protocol was: 94 C (for 5 min), followed by 33 cycles of 94 C (for 60 s), 54 C (for 60 s), 72 C (for 90 s) and a final extension step of 72 C for 10 min (Liu et al., 2010). After the 1.6-kb gene fragment of 16S rRNA was successfully amplified, the product was diluted 10-fold and used as the template for the second amplification with primers 338F GC and 518R (338F GC: 50 -CGC CCG CCG CGC GCG GCG GGC GGG GCG GGG GCA CGG GGG GAC TCC TACGGG AGGCAGAG30 ; 518R: 50 -ATT ACC GCG GCT GCT GG-30 ) (Liu et al., 2010). PCR amplifications of bacterial 16S rRNA V3 region were performed in a 50-ml reaction mixture containing 1 ml of template, 1 ml of each primer (20 mM) (Shanghai Invitrogen Biotechnology Co. Ltd., China), 4 ml of dNTP (each 2.5 mM), and 1.25 U TaKaRa Taq polymerase. The PCR cycle for amplification included 3 min of initial denaturation at 94 C, 33 cycles of 94 C for 30 s, 56 C for 30 s, and 72 C for 25 s, and a final elongation step of 72 C for 7 min. The PCR products were electrophoresed in a 1.5% (w/v) agarose gel, stained with ethidium bromide and quantified by comparison with a standard marker (DL 2000, TaKaRa Biotechnology Dalian Co. Ltd., China). DGGE profiling was performed with a Bio-Rad D-Code Universal Mutation Detection System (Bio-Rad Corp., USA). Approximately 150 ng of each PCR product was loaded into each well of a 10% (w/v) acrylamide gel (acrylamide/bis acrylamide solution, 37.5:1) containing a linear denaturing gradient from 35 to 60% (A 100% denaturant contains 7 M urea and 40% (v/v) deionized formamide). Electrophoresis was for 450 min at 150 V in 1 TAE buffer at 60 C. The gel was visualized by silver staining (Bassam et al., 1991). Then the developed gel was scanned using the Gel Doc XRþ Imaging System (Bio-Rad, USA) and saved in uncompressed TIFF format for further analysis.
2.4.3. Sequencing of the DGGE bands the analysis of nucleotide sequence Differential or dominant bands containing specific DNA were aseptically excised from DGGE gels and washed with 50 ml sterile water for 10 min, then eluted in 40 ml TE buffer at 4 C
3.
Results and discussion
3.1.
Strain’s isolation and identification
After four repetitions by streaking colonies onto LB plates, two representative thermophilic strains were isolated in accordance with the typical characters of bacterial colonies. Strain T1 had the following character: irregular colony, light yellow, smooth surface, and translucent. On the other hand, the T2 colony presented a yellow, round, wet surface, with little bulge in the center. The sequence of T1 (Accession no. HQ436531) has 98% similarity with both Tepidiphilus margaritifer (Accession no. NR025556) and Petrobacter sp. (DQ539621), whereas T2 (Accession no. HQ436532) has 100% and 99% similarity with Pseudoxanthomonas taiwanensis (FR774559) and Pseudoxanthomonas sp. (AB039336), respectively. Fig. 2 presents phylogenetic dendrogram based on 16S rRNA gene sequences between the isolated strains and representatives of related genera. Strains, T1 and T2 clearly belong to Hydrogenophilaceae and Xanthomonodaceae, respectively. As a result, the two thermophilic stains isolated from the onestage ATAD pilot-scale reactor exhibited considerable differences with respect to the characteristics of bacterial colonies and their phylogenetic relationship.
3.2.
Effects of thermophilic stains on sludge stabilization
3.2.1.
Removal efficiency of VS
The removal efficiency of VS is a significant indicator for sludge digestion systems (USEPA, 1993; Layden et al., 2007). Fig. 3 presents the variation of VS removal in the thermophilic aerobic digestion process. Digester R1, R2, and R3 all had higher VS removal than R0 up to 192 h. However, digester R1 only achieved moderate increase in VS removal after 144 h, and its removal began to be lower than that of digester R0 after 240 h. During the whole digestion process, digester R2 and R3 always kept higher VS removal than digester R0, which had no inoculation. After the sludge was digested for 432 h, VS removal in digesters R0, R2, and R3 were 30.0, 41.0, and 35.6%, respectively; at the end of digestion (i.e., 480 h), the removal in these three digesters reached up to 32.6, 43.0, and 38.2%, respectively. In the initial digestion period, abundant organic substrate was released into the supernatant due to cell lysis (Yan et al., 2008) under thermophilic temperature of 55 C, resulting in rapid degradation of VS. On the other hand, the inoculated thermophilic strains can rapidly develop into predominant microbes in the digestion system, favoring degradation of the organic substrate (Liu et al., 2009, 2010). For a digestion time of
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 5 9 e5 9 6 8
5963
Fig. 2 e Phylogenetic dendrogram based on 16S rRNA gene sequences between the isolated strains and representatives of related genera. Bootstrap values (expressed as a percentage of 1000 replications) were shown at branch points. Bar 0.01 nucleotide substitutions per position.
96 h, VS removal in digester R1, R2, and R3 was 15.5, 18.3, and 18.3%, respectively; whereas the removal in digester R0 only reached 10.9%. It is clear that inoculation with thermophilic strains can speed up the fast degradation of VS during the early stages of thermophilic aerobic digestion of sewage sludge. As the digestion process continued, the viable thermophilic bacteria increased substantially in digester R0, contributing to the degradation of VS, so that digester R0 steadily increased its VS removal after 144 h. However, digester R1 inoculated with the T1 strain achieved a moderate increase of VS removal in a later digestion period, indicating that the T1 strain adversely affected the thermophilic digestion under the changed digestion conditions. Digester R2 and R3, wholly or partially inoculated with the T2 strain, removed more VS than R0. The results revealed that the T2 strain can completely acclimatize itself to thermophilic digestion conditions, facilitating rapid degradation of VS. They also demonstrated that the factors affecting digestion depended not only on the dominant thermophilic strains, but also on the digestion conditions. At digestion time of 432 h, the VS removal in digester R0 was only 30.0%; however, the removal of VS in digester R2 reached 41.0%, meeting the GB18918-2002 guideline (that the
45% R0 R2
40%
R1 R3
Removal of VS
35% 30% 25% 20% 15% 10% 5% 0% 0
48
96
144 192 240 288 336 384 432 480 528 Digestion time ( h )
Fig. 3 e Variation of VS removal in thermophilic aerobic digestion process.
degradation of organic matter must exceed 40% when aerobic digestion is applied to stabilize the sewage sludge from municipal wastewater treatment plants) (SEPAC, 2002b). The removal efficiency was also in line with the Class-A biosolids definitions in Part 503 of the USEPA 40 CFR (USEPA, 1993).
3.2.2.
Changes of SCOD and TN in the supernatant
The concentration of SCOD in the thermophilic aerobic digestion system firstly experienced a sharp increase during the initial digestion period (up to 48 h), then a moderate variation from 48 to 144 h, followed by steady decline after 144 h (Fig. 4A). Digester R0, R1, R2, and R3 reached the highest SCOD concentrations at 48, 144, 96, and 48 h, respectively. The maximum SCOD was 18,000 mg L1 in digester R0. Though digester R1 achieved the lowest VS removal among the four digesters after 288 h (Fig. 3), it had a relatively higher concentration of SCOD from then onward. Investigations (van Loosdrecht and Henze, 1999; Liu et al., 2010) have revealed that a large amount of organic matter, including lipids, polysaccharides, proteins, and nucleic acid, could be released from dead microorganisms into the supernatant during thermophilic aerobic digestion. This finding is supported by the rapid removal of VS (Fig. 3) and the sharp increases of SCOD during the first 48 h (Fig. 4A). Li et al. (2009) found that thermostable enzymes can be released by cell lysis as less-temperature-tolerant microorganisms die, and the released enzymes can enhance sludge degradation. van Loosdrecht and Henze, (1999) also reported that thermophilic microorganisms can utilize the released compounds to sustain their own activities and achieve rapid growth. These investigations revealed that the variation of SCOD is a composite result of degradation of organic substrates and the metabolic process. As digestion continued, the digesters achieved moderate VS removal; however, the digestion system approached a stable bacterial consortium due to the synergetic effects of the inoculated thermophilic strains and the other viable thermophiles, resulting in a sound and steady metabolic process. Therefore, SCOD underwent a moderate decline after 144 h. The change of total nitrogen in the supernatant is shown in Fig. 4B. In the initial 48 h, the TN concentration in all four digesters increased rapidly, reaching w2000 mg L1. TN
5964
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 5 9 e5 9 6 8
Concentration of SCOD ( mg L-1 )
A
20000
R0 R2
18000
R1 R3
16000 14000 12000 10000 8000 6000 2000 1500 1000 0
48
96
144 192 240 288 336 384 432 480 528
Digestion time ( h )
B
R0 R2
-1 Concentration of TN ( mg L )
2400
R1 R3
2000 1600 1200 800
3.2.3.
400 0
0
48
96
144
192
240
288
336
384
432
480
528
Digestion time ( h )
C 7.5 R0
R1
R2
R3
7.0
pH
6.5
6.0
5.5 0
D
48
96
144 192 240 288 336 384 432 480 528 Digestion time ( h )
50 R0
R1
R2
R3
0 -50 -100 ORP (mV)
-150 -200 -250 -300 -350 -400 -450
0
reached its highest concentration at 144 h, with its concentrations in digesters R0, R1, R2, and R3 being 2207, 2422, 2292, and 2191 mg L1, respectively. TN began to decrease after 144 h, and continued to decline moderately from 384 h to the end of the digestion process (480 h). During the thermophilic digestion process, nitrogen was released due to the protein degradation in the extracellular polymeric substance and decay of less-temperature-tolerant cells (Lee et al., 2004; Li et al., 2009), resulting in a rapid increase of TN in the first 144 h. In these batch experiments, the concentration of NO 3 1 and NO 2 were always less than 3 and 1 mg L , respectively, indicating that nitrification and denitrification were indeed inhibited during thermophilic aerobic digestion. Since biological nitrification and denitrification effectively stop, ammonia stripping represents the only way to decrease TN because ammonia can be stripped from the digestion system during continuous aeration. This speculation is supported by the finding that ammonia nitrogen was the main form of TN in a simulated ATAD digester (Liu et al., 2010). After 144 h, the concentration of degradable organic substrates decreased, and the reduction of VS was no longer as significant as before; as a result, the concentration of TN declined continuously due to the ammonia stripping.
48
96
144
192
240
288
336
384
432
480
528
Digestion time ( h )
Fig. 4 e Digestion system time variations: (A) SCOD, (B) TN, (C) pH, and (D) ORP.
Variations of pH and ORP of the digestion system
The pH in the digester was not regulated during the overall digestion process. It declined slightly within the first 12 h due to hydrolytic-acidification and then increased gradually from 12 to 48 h (Fig. 4C). At a digestion time of 12 h, the pH in digester R0, R1, R2, and R3 reached the lowest value of 5.8, 5.8, 5.9, and 5.9, respectively. The pH fluctuated between 6.2 and 7.2 from 48 h to the end of digestion process. The pH has previously been found to depend on the rate of biodegradation and hydrolysis of protein-based material, followed by further deamination of the peptide and amino acid products with subsequent release of ammonia (Staton et al., 2001; Piterina et al., 2009). However, the pH levels in digester R0, R1, and R3 did not exceed 7.0, a finding which differed from previous investigations where the pH in ATAD systems corresponded to alkalescence (Mavinic et al., 2001). It is generally believed that the released ammonia becomes solubilized and accumulates in the bulk sludge water of the digester, contributing to an increase in pH (Grady et al., 1998). However, the pH in digester R1 did not achieve a higher level, even though digester R1 generally maintained a higher TN concentration than the other three digesters after 96 h (Fig. 4B). Poggi-Varaldo et al. (1997) pointed out that the efficiency and biogas productivity decreased, whereas propionic, butyric, and valeric acid increased significantly when the anaerobic digestion system was inhibited by high ammonia. In this study, 6 short-chain fatty acids (SCFAs), including acetic, propionic, n-butyric, isobutyric, n-valeric, and iso-valeric acid, were determined using gas chromatography, and the results indicated that digester R1 had a higher concentration of SCFAs than the other three digesters. The main constituent of SCFAs was propionic, n-butyric, and iso-valeric acid, in addition to acetic acid. This finding indicates that the pH level is determined not by the concentration of total nitrogen or ammonia nitrogen, but by the integrated results of SCFAs and ammonia nitrogen, i.e., acidebase equilibrium.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 5 9 e5 9 6 8
5965
Fig. 4D shows the variation of ORP during the thermophilic aerobic digestion process. The ORP slightly increased during the initial startup period and then increasingly declined to a minimum at 48 h. The ORP levels in digesters R0, R2, and R3 rapidly increased from 48 h to 84 h, followed by moderate fluctuation after 84 h. Similarly, ORP in digester R1 varied in a similar manner, the only difference being that ORP in digester R1 increased from 48 to 108 h and then declined. After 144 h, ORP in digester R0 and R1 ranged between 200 and 0 mV, whereas that in digesters R2 and R3 fluctuated from 270 to 20 mV. As the digestion continued, the ORP continued to increase in general after 240 h, despite moderate fluctuation. During the whole digestion process, continuous aeration was supplied; however, the ORP remained below 0 mV most of the time, indicating that the digestion system was not entirely aerobic. This finding is also consistent with the finding that the ORP often varied between 250 and 50 mV during stable operation in an ATAD system (Kelly et al., 1993).
3.3. Microbial community structure during thermophilic aerobic digestion 3.3.1.
Analysis of DGGE profiles
PCReDGGE is perhaps the most popular technique for biodiversity assessment in bioreactor samples. It has also been popular for fast comparative analysis of communities from different reactors (Boon et al., 2002; Stamper et al., 2003; Piterina et al., 2010). Fig. 5 shows DGGE profiles for the four digesters at different digestion times. The abundant bands in untreated sludge reveal that the original sludge had more abundant bacterial population than the thermophilic aerobic digestion system. As the system experienced thermophilic digestion, some bands (e.g., bands 3, 4, and 5) gradually disappeared, while others (e.g., bands 10 and 11) became detectable, consistent with the fact that the predominant microbes changed from less-temperature-tolerant microorganisms to thermophiles. The intense bands were bands 8 and 11 in addition to bands 7 and 12 in digester R0 at 240 h, implying that some thermophilic bacteria were un-culturable. The thermophilic digestion system had fewer bands, however it presented relatively steady DGGE patterns after 240 h, indicating that the digestion system achieved a steady bacterial community.
3.3.2.
Species identification
Dominant DGGE bands were excised from DGGE gels for DNA sequencing. The closest relatives matched in the GenBank database are shown in Table 2. All of the sequences from bands 1, 12, 14, 16, and 18 presented no less than 95%, and above 96%, similarity with thermophilic stain T1 (Accession no. HQ436531) and T. margaritifer (Accession no. NR025556), respectively. On the other hand, the sequences for bands 2, 7, 13, 15, and 17 all showed 99% similarity with thermophilic stain T2 (Accession no. HQ436532). The Phylogenesis analysis and the DGGE patterns (Fig. 5) revealed that the thermophilic stains, T1 and T2, appeared in the digestion system of digester R1, R2, and R3, as well as in digester R0 at 96, 240, and 480 h. This shows that the thermophilic stains T1 and T2 can not only successfully establish in the thermophilic digestion system, but also achieve viable growth.
Fig. 5 e DGGE profiles of the V3 region of 16S rRNA gene fragments extracted from biomass samples from the thermophilic aerobic digestion system. Samples were taken from 2 thermophilic stains (T1 and T2) and the untreated sludge S0 (0 h) as well as the digested sludge in 4 digesters at 96, 240, and 480 h. The numbers 1e18 represent the bands excised from corresponding positions of DGGE gels. The same band numbers is used in Table 2.
As shown in Table 2, bands 3, 4, 5, and 6 were derived from bacteria belonging to Betaproteobacteria, Comamonadaceae, Nocardioidaceae, and Pseudomonadaceae, respectively. These bands were only observed in samples from untreated sewage sludge (Fig. 5). Table 2 also showed that the sequences from bands 8 and 10 were very similar to those of Thermoanaerobacteriaceae, whereas the bacteria represented by bands 9 and 11 belonged to Bacillaceae and Sphaerobacteraceae, respectively. As a result, molecular analysis of bacterial communities revealed that anaerobic or facultative microbes, in addition to aerobic bacteria, can be present in a thermophilic aerobic digestion process. This result is supported by previous reports (Staton et al., 2001; Liu et al., 2010; Hayes et al., 2011), and it is also consistent with the fact that the digestion system always kept below 0 mV most of the digestion time (Fig. 4C) in spite of continuous aeration. As the thermophilic aerobic digestion process continued, some high viable bacteria (presented by bands 8e11) were observed in digesters (Fig. 5), and these thermophilic or thermoduric species increased substantially after a period of adaptive phase. Some of these viable bacteria belonged to anaerobic or facultative microbes (Table 2). For most of the digestion time, these anaerobic or facultative microbes can co-exist in a digestion process like thermophilic stains T1 and T2 (Fig. 5). This finding implied that those anaerobic or facultative microbes may also play an important role in VS degradation. It has been pointed out that “the more complex
5966
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 5 9 e5 9 6 8
Table 2 e Closest phylogenetic affiliations of sequences recovered from DGGE gels. Band
Accession no.
Band 1
JF682323
Band 2
JF682324
Band 3
JF682325
Band 4
JF682326
Band 5
JF682327
Band 6
JF682328
Band 7
JF682329
Band 8
JF682330
Band 9
JF682331
Band 10
JF682332
Band 11
JF682333
Band 12
JF682334
Band 13
JF682335
Band 14
JF682336
Band 15
JF682337
Band 16
JF682338
Band 17
JF682339
Band 18
JF682340
*1
HQ436531
*2
HQ436532
Closest relatives in GenBank (Accession)
Similarity
Bacterium thermus-lsg2 (HQ436531) Tepidiphilus margaritifer (NR025556) Pseudoxanthomonas taiwanensis (FR774559) Bacterium thermus-lsg3 (HQ436532) Uncultured Nitrosospira sp. (GU097368) Uncultured beta proteobacterium (EF073293) Comamonas sp. TBA14 (FR745416) Uncultured Comamonas sp. (EU639125) Nocardioides sp. 02SU6 (HQ424128) Uncultured Nocardioidaceae bacterium (GU202371) Uncultured Flavimonas sp. (FM175401) Uncultured Pseudomonas sp. (GQ183242) Pseudoxanthomonas taiwanensis (FR774559) Bacterium thermus-lsg3 (HQ436532) Coprothermobacter sp. (AB537980) Thermoanaerobacteriaceae bacterium (GU129121) Ureibacillus thermosphaericus (AB210996) Ureibacillus composti strain (DQ348071) Coprothermobacter sp. (AB537980) Thermoanaerobacteriaceae bacterium (GU129121) Uncultured compost bacterium (FN667457) Sphaerobacter thermophilus (AJ420142) Tepidiphilus margaritifer (NR025556) Bacterium thermus-lsg2 (HQ436531) Pseudoxanthomonas taiwanensis (FR774559) Bacterium thermus-lsg3 (HQ436532) Tepidiphilus margaritifer (NR025556) Bacterium thermus-lsg2 (HQ436531) Pseudoxanthomonas taiwanensis (FR774559) Bacterium thermus-lsg3 (HQ436532) Tepidiphilus margaritifer (NR025556) Bacterium thermus-lsg2 (HQ436531) Pseudoxanthomonas taiwanensis (FR774559) Bacterium thermus-lsg3 (HQ436532) Tepidiphilus margaritifer (NR025556) Bacterium thermus-lsg2 (HQ436531) Tepidiphilus margaritifer (NR025556) Petrobacter sp. (DQ539621) Pseudoxanthomonas taiwanensis (FR774559) Pseudoxanthomonas sp. (AB039336)
99% 96% 100% 99% 87% 87% 92% 92% 90% 91% 94% 93% 99% 99% 99% 99% 92% 92% 99% 99% 99% 99% 99% 96% 100% 99% 99% 97% 99% 99% 99% 95% 99% 99% 99% 95% 98% 98% 100% 99%
a reactor’s ecosystem is, the more stable and resilient it will be” (Zein et al., 2004). In this study, Samples at 240 h in digester R1 had fewer bands than digesters R0, R2, and R3 and showed low bacterial diversity (Fig. 5). This corresponded with the moderate increase of VS removal in digester R1 from 144 to 336 h (Fig. 3). How the micro-environments in thermophilic aerobic digestion process affected the removal of VS is unknown. However, this analysis of bacterial diversity provides some insight.
4.
Conclusions
It is hard to culture most of the thermophilic bacteria from ATAD system using a specific culture method. Two representative thermophilic stains, T1 and T2, were successfully isolated from a one-stage ATAD digester in this study, belonging to Hydrogenophilaceae and Xanthomonodaceae, respectively.
Phylogenesis Hydrogenophilaceae Xanthomonadaceae Nitrosomonadaceae Betaproteobacteria Comamonadaceae Nocardioidaceae Pseudomonadaceae Xanthomonadaceae Thermoanaerobacteriaceae Bacillaceae Thermoanaerobacteriaceae (unknown) Sphaerobacteraceae Hydrogenophilaceae Xanthomonadaceae Hydrogenophilaceae Xanthomonadaceae Hydrogenophilaceae Xanthomonadaceae Hydrogenophilaceae Hydrogenophilaceae Xanthomonadaceae
Inoculation with thermophilic strains can speed up the degradation of organic matter during the early stage of thermophilic aerobic digestion of sewage sludge. As those viable thermophilic bacteria increased substantially, digester R0 (with no inoculation) began to maintain a moderate increase in VS removal. Both specific thermophic strains and microenvironment significantly affected the thermophilic digestion process. Thermophilic strain T1 had adverse effects on digestion after 240 h due to the changed microenvironment, and digester R1 (inoculated with strain T1) obtained lower VS removal than digester R0. Strains T2 can completely acclimatize itself to thermophilic digestion and favor the digestion process. Digester R2 (inoculated with strain T2) achieved higher VS removal than other three digesters, and the removal reached 41.0% at 432 h, meeting the GB18918-2002 guideline. So it is very helpful to inoculate some thermophilic strains as the startup of ATAD system. The isolated thermophilic strains T1 and T2 can successfully establish in thermophilic digesters. Though continuous aeration was supplied, those anaerobic or facultative
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 5 9 e5 9 6 8
microbes could be viable and appear in thermophilic aerobic digester, which jointly constituted the steady microbial community and contributed to sludge stabilization.
Acknowledgments This study was financially supported by the National Hi-Tech Research and Development Program of China (863) (No. 2011AA060906) and the Natural Science Foundation of China (No. 50878127).
references
APHA, AWWA, WEF, 2005. Standard Methods for the Examination of Water and Wastewater, twentyfirst ed. Americal Public Health Association/American Water Works Association/Water Environment Federation, Washington, DC, USA. Bassam, B.J., Caetano, A.G., Gresshoff, P.M., 1991. Fast and sensitive silver staining of DNA in polyacrylamide gels. Analytical Biochemistry 196, 80e83. Boon, N., de Windt, W., Verstraete, W., Top, E.M., 2002. Evaluation of nested PCReDGGE (denaturing gradient gel electrophoresis) with group specific 16S rRNA primers for the analysis of bacterial communities from different wastewater treatment plant. FEMS Microbiology Ecology 39 (2), 101e112. Cheng, J.H., 2006. The Study of Sewage Sludge Treatment by Autothermal Thermophilic (micro) Aerobic Digestion Process. Tongji University, Shanghai. Cheng, J.H., Zhou, Q.F., Zhu, N.W., 2009. Study of oxidationreduction potential (ORP) on autothermal thermophilic aerobic digestion process. Research Journal of Environmental Sciences 22 (4), 484e489. Chenna, R., Sugawara, H., Koike, T., Lopez, R., Gibson, T.J., Higgins, D.G., Thompson, J.D., 2003. Multiple sequence alignment with the Clustal series of programs. Nucleic Acids Research 31 (13), 3497e3500. Elisabeth, C., Patrick, S., 2000. The Nucleic Acid Protocols Handbook. Humana Press Inc., NJ. Gomez, J., Gracia, M., Ayesa, E., Garcia-Heras, J.L., 2007. Mathematical modelling of autothermal thermophilic aerobic digesters. Water Research 41, 959e968. Grady, P.L., Daigger, G.T., Lim, H.C., 1998. In: Biological Wastewater Treatment, second ed. CRS Press Taylor & Francis, New York, USA, pp. 572e573. Hayes, D., Izzard, L., Seviour, R., 2011. Microbial ecology of autothermal thermophilic aerobic digester (ATAD) systems for treating waste activated sludge. Systematic and Applied Microbiology 34, 127e138. Kelly, H.G., Mavinic, D.S., 2003. Autothermal thermophilic aerobic digestion research application and operational experience. In: Los Angeles, WEFTEC 2003 Workshop W104 Thermophilic Digestion. Kelly, H.G., Melcer, H., Mavinic, D.S., 1993. Autothermal thermophilic aerobic digestion of municipal sludges: A one-year, full-scale demonstration project. Water Environment Research 65 (7), 849e861. Kim, Y.K., Bae, J.H., Oh, B.K., Choi, J.W., 2002. Enhancement of proteolytic enzyme activity excreted from Bacillus stearothermophilus for a thermophilic aerobic digestion process. Bioresource Technology 82, 157e164.
5967
Kumar, S., Tamura, K., Jakobsen, I.B., Masatoshi, N., 2001. MEGA2: molecular evolutionary genetics analysis software. Bioinformatics 17 (12), 1244e1245. LaPara, T.M., Konopka, A., Nakatsu, C.H., Alleman, J.E., 2000. Effects of elevated temperature on bacterial community structure and function in bioreactors treating a synthetic wastewater. Journal of Industrial Microbiology and Biotechnology 24, 140e145. Layden, M.N., Mavinic, D.C., Kelly, H.G., Moles, R., 2007. Autothermol thermophilic aerobic digestion (ATAD) e Part I: review of origins, design, and process operation. Journal of Environmental Engineering and Science 6, 665e678. Lee, J.W., Lee, H.W., Kim, S.W., 2004. Nitrogen removal characteristics analyzed with gas and microbial community in thermophilic aerobic digestion for piggery waste treatment. Water Science and Technology 49, 349e357. Li, X.S., Ma, H.Z., Wang, Q.H., Matsumoto, S., Maeda, T., Ogawa, H.I., 2009. Isolation, identification of sludge-lysing strain and its utilization in thermophilic aerobic digestion for waste activated sludge. Bioresource Technology 100, 2475e2481. Liu, S.G., Zhu, N.W., Lou, Z.Y., 2009. Analysis of microbial diversity in thermophilic aerobic digestion of sewage sludge. Acta Scientiae Circumstantiae 30, 990e995. Liu, S.G., Song, F.Y., Zhu, N.W., Yuan, H.P., Cheng, J.H., 2010. Chemical and microbial changes during autothermal thermophilic aerobic digestion (ATAD) of sewage sludge. Bioresource Technology 101, 9438e9444. Manaia, C.M., Nogales, B., Nunes, O.C., 2003. Tepidiphilus margaritifer gen. nov., sp. nov., isolated from a thermophilic aerobic digester. International Journal of Systematic and Evolutionary Microbiology 53, 1405e1410. Mavinic, D.S., Mahendraker, V., Sharma, A., Kelly, H.G., 2001. Effect of microaerophilic conditions on autothermal thermophilic aerobic digestion process. Journal of Environmental Engineering 127 (4), 311e316. Muyzer, G., Dewaal, E.C., Uitterlinden, A.G., 1993. Profiling of complex microbial populations by denaturing gradient gel electrophoresis of polymerase chain reaction-amplified genes coding for 16S rRNA. Applied and Environmental Microbiology 59, 69e5700. Piterina, A.V., Bartlett, J., Pembroke, J.T., 2009. 13C-NMR assessment of the pattern of organic matter transformation during domestic wastewater treatment by Autothermal Aerobic Digestion (ATAD). International Journal of Environmental Research and Public Health 6, 2288e2306. Piterina, A.V., Bartlett, J., Pembroke, J.T., 2010. Molecular analysis of bacterial community DNA in sludge undergoing autothermal thermophilic aerobic digestion (ATAD): pitfalls and improved methodology to enhance diversity recovery. Diversity 2, 505e526. Poggi-Varaldo, H.M., Rodriguez-Vazquez, R., FernandezVillagomez, G., 1997. Inhibition of mesophilic solid-substrate anaerobic digestion by ammonia nitrogen. Applied Microbiology and Biotechnology 47, 284e291. Rojas, J., Zhelev, T., Bojarski, A.D., 2010. Modelling and sensitivity analysis of ATAD. Computers and Chemical Engineering 34, 802e811. SEPAC, 2002a. In: Analytical and Monitoring Methods of Water and Wastewater, fourth ed. China Environmental Science Press, Beijing, pp. 210e213. SEPAC, 2002b. GB18918e2002, Discharge Standards for Pollutants from Municipal Wastewater Treatment Plants. China Environmental Science Press, Beijing, China. Saurabh, A., 2004. Conditioning and Dewatering Behavior of ATAD Sludges. Virginia Polytechnic Institute and State University, USA, Virginia. Stamper, D.M., Walch, M., Jacobs, R.N., 2003. Bacterial population changes in a membrane bioreactor for gray water treatment
5968
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 5 9 e5 9 6 8
monitored by denaturing gradient gel electrophoretic analysis of 16S rRNA gene fragments. Applied and Environmental Microbiology 69 (2), 852e860. Staton, K.L., Alleman, J.E., Pressley, R.L., Eloff, J., 2001. 2nd generation autothermal thermophilic aerobic digestion: conceptual issues and process advancements. In: WEF/AWWA/CWEA Joint Residuals and Biosolids Management Conference, San Diego. USEPA, 1990. Autothermal Thermophilic Aerobic Digestion of Municipal Wastewater Sludge. Environmental Protection Agency, Washington, DC. Report EPA/625/10-90/007. USEPA, 1993. 40 CFR PART 503, Standards for the Use or Disposal of Sewage Sludge. United States Environmental Protection Agency, Washington, D.C. van Loosdrecht, M.C.M., Henze, M., 1999. Maintenance, endogeneous respiration, lysis, decay and predation. Water Science and Technology 39 (1), 11e20.
Wuyts, J., Van de Peer, Y., Winkelmans, T., De Wachter, R., 2002. The European database on small subunit ribosomal RNA. Nucleic Acids Research 30, 183e185. Yan, S.T., Miyanaga, K., Xing, X.H., Tanji, Y., 2008. Succession of bacterial community and enzymatic activities of activated sludge by heat-treatment for reduction of excess sludge. Biochemical Engineering Journal 39, 598e603. Zein, M.M., Suidan, M.T., Venosa, A.D., 2004. MtBE biodegradation in a gravity flow, high-biomass retaining bioreactor. Environmental Science & Technology 38 (12), 3449e3456. Zhou, J.P., 2003. Factors Influencing Dewaterability of Thermophilic Aerobically Digested Biosolids. University of British Columbia, Canada, Vancouver. Zhu, N.W., Lin, J.M., Cheng H., Zhang, S.F., Jia, J.P., 2005. The device of thermophilic aerobic digestion of sludge. China Patent CN1587110, 2005-03-02.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 6 9 e5 9 7 6
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Conditioning of wastewater sludge using freezing and thawing: Role of curing Kai Hu a, Jun-Qiu Jiang a, Qing-Liang Zhao a,b,*, Duu-Jong Lee b,c,d, Kun Wang a, Wei Qiu a a
School of Municipal & Environmental Engineering, Harbin Institute of Technology, Harbin 150090, China State Key Laboratory of Urban Water Resources and Environment (SKLUWRE), Harbin Institute of Technology, Harbin 150090, China c Department of Environmental Science and Engineering, Fudan University, Shanghai 200344, China d Department of Chemical Engineering, College of Engineering, National Taiwan University of Science and Technology, Taipei, Taiwan 10617, China b
article info
abstract
Article history:
Freeze/thaw (F/T) treatment is an efficient pre-treatment process for biological sludges.
Received 14 April 2011
When bulk sludge was frozen, tiny unfrozen regimes in the ice matrix were continuously
Received in revised form
dehydrated by surrounding ice fronts, termed as the “curing stage”. This work demon-
18 August 2011
strated that the F/T treatment could not only enhance sludge dewaterability, but also
Accepted 29 August 2011
solubilize organic matters from sludge matrix. Most enhancement of sludge dewaterability
Available online 3 September 2011
was achieved during bulk freezing stage, with the waste activated sludge more readily dewatered than the mixed sludges after treatment. Conversely, the freezing stage released
Keywords:
only limited quantities of organic matters to liquid. Conversely, the curing contributed
Freeze/thaw treatment
mostly on chemical oxygen demand (COD) solubilization and NH3eN release. The crys-
Curing
tallization of intra-aggregate moisture was claimed to damage cell membranes so to
Waste activated sludge
release intracellular substances to surroundings. The F/T treatment with sufficient curing
Mixed sludge
is advised to effectively condition biological sludge as the feedstock of the following
Solubilization
anaerobic digestion process.
Dewaterability
1.
Introduction
Organic matters hydrolysis presents the rate-limiting step in sludge anaerobic digestion process (Elliott and Mahmood, 2007). Sludge pre-treatment techniques, including mechanical (like sonication), chemical (such as alkali treatment), thermal (heat treatment or freeze/thaw (F/T)) and biological (enzymatic treatment), were studied in detail (Chu et al., 2002a, 2002b; Whiteley and Lee, 2006). The F/T process presents a cost-effective sludge conditioning unit in case natural freezing on sludge is feasible in field (Vesilind et al., 1991a; Hedstrom and Hanaeus, 1999).
ª 2011 Elsevier Ltd. All rights reserved.
Studies on F/T for biological sludge and metal hydroxide precipitates considered the associated changes in sludge dewaterability (Martel and Diener, 1991; Parker et al., 1998a; Wang et al., 2001; Kawasaki et al., 2004) and floc structure (Vesilind et al., 1991a; Chang et al., 2004). Freezing temperature and time were revealed as two of the major factors that influenced performance of F/T treatment. Wang et al. (2001) demonstrated that improvement of sludge dewaterability and degree of elution of intracellular water were more favorable at slow-frozen (20 C) than at fast-frozen (80 C) tests. Gao (2011) conducted bench scale experiments to examine the effect of freezing temperature and freezeethaw cycles on the
* Corresponding author. School of Municipal & Environmental Engineering, Harbin Institute of Technology, Harbin 150090, China. Tel.: þ86 451 86283017; fax: þ86 451 86282100. E-mail address:
[email protected] (Q.-L. Zhao). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.08.064
5970
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 6 9 e5 9 7 6
yielded sludge properties. During freezing, developing ice front would partly engulf the floc and the force thereby built up would compress the unfrozen part and pull apart the network of the frozen part of sludge. By doing so, the sludge was converted into a matrix of ice crystals and compacted solid particles (Tao et al., 2006) with cell membrane integrity damaged by the intracellular ice crystals (Thomashow, 1998). The extracellular polymers (ECPs) were noted to release to the sludge supernatant following F/T treatment (Hung et al., 1996; Ormeci and Vesilind, 2001). Literature results demonstrated that the enriched supernatant is favorable to enhance anaerobic digestion of sludge (Montusiewicz et al., 2010). When bulk sludge was completely frozen, there are tiny unfrozen regimes in the ice matrix. Extended freezing, or curing of the sludge, can significantly improve sludge dewaterability of drinking water residues (Vesilind and Martel, 1990; Parker et al., 1998b; Jean and Lee, 2000). In a city such as Harbin, China, the average temperatures during November to March are below freezing temperature (Wang et al., 2010), making natural freezing a promising pre-treatment option to sludge management. The sludge dumped into a freezing pool during winter time will be kept frozen until it is thawed since April and onward. Restated, during the whole year cycle with complete freezing and complete thawing, most of the freezing stage sludge will be in the curing state. Little attention was paid to unveil the effects of curing on the solubilization of organic matters from biological sludges. This study applied F/T treatment at 18 C on wastewater sludges and investigated the changes in physical and chemical characteristics of sludge after F/T treatment. In particular, the role of curing stage on sludge characteristics was clearly demonstrated. Mechanisms corresponding to the noted changes were discussed.
2.
Materials and methods
2.1.
Sludge samples
Sludge samples were collected from the primary sedimentation tank (termed as primary sludge) and from the secondary sedimentation tank (termed as the waste activated sludge (WAS)) in a municipal wastewater treatment plant at Harbin City, China. All collected sludge samples were first gravity thickened to around 97% w/w moisture content. Then the mixed sludge samples were prepared by mixing thickened primary sludges and thickened waste
activated sludge samples at 1:4 v/v to simulate the field practice. The characteristics of sludge samples were shown in Table 1.
2.2.
F/T treatment
The thickened WAS and mixed sludge were placed in 550 ml polyethyleneterephthalate bottles sealed with polyethylene lids and frozen at 18 C at different time periods. Following freezing (and curing), the sludges were thawed for another 3 h at 29 C and at 47e56% relative humidity. Preliminary tests revealed that complete freezing of sludge samples could be reached in 3 h. Hence, the freezing tests at <3 h were at freeing stage; while those at >3 h were at curing stage.
2.3.
Analytical methods
Total chemical oxygen demand (TCOD) and soluble chemical oxygen demand (SCOD), total solids (TS), suspended solids (SS), volatile solids (VS), volatile suspended solids (VSS) and pH for the sludge samples before and after F/T treatment were measured based on the Standard Analysis Methods (China EPA, 2002). The sludge samples were centrifuged at 2770 g for 30 min prior to SCOD, alkalinity, NH3eN, SS and VSS measurements. The VS and VSS contents were determined after calcination at 600 C for 1 h. The COD solubilization was defined as the ratio of the SCOD of treated sludge (SCOD) minus the initial SCOD (SCOD0) divided by the initial particulate fraction of COD (CODp0) as follows (Bougrier et al., 2008): COD solubilizationð%Þ ¼
SCOD SCOD0 SCOD SCOD0 ¼ TCOD0 SCOD0 CODp0
where TCOD0 is the initial sludge TCOD. The particle size distribution was measured by dilution of sludge supernatant using a Liquid Particle Counting System (HIAC 9703, USA). The detected particles ranged from 2 to 300 mm. A drop of sludge sample was spread via pasteur pipette onto a microscope slide, at which point the floc structure was observed and photographed using an Olympus BX051 at 200 magnification. 100 ml of sludge sample was settled in a graduated cylinder and the settled volume of sedimentation was recorded. A vacuum filtration system equipped with Buchner funnel was installed and adopted for a 100 ml sludge sample at a pressure difference of 0.7 bar. The filtrate volume was
Table 1 e Characterization of thickened WAS and mixed sludge.a Mixed sludge 1
TCOD/mgl SCOD/mgl1 pH NH3eN/mgl1 Alkalinity/mgl1
33,200 920 6.49 148 720
Thickened WAS 28,210 948 6.45 101 580
1
TS/mgl SS/mgl1 VS/mgl1 VSS/mgl1
Mixed sludge
Thickened WAS
37,870 35,870 19,530 18,620
24,280 22,400 15,760 14,720
a TCOD: total chemical oxygen demand; SCOD: soluble chemical oxygen demand; TS; total solids; SS: suspended solids; VS: volatile solids; VSS: volatile suspended solids.
5971
TS (VS) /mg•L-1
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 6 9 e5 9 7 6
50000
Table 2 e Settled sludge volumes after 24 h (initial sludge volume: 100 ml).
40000
Mixed sludge
30000
Raw 1 h freezing 3 h freezing 72 h freezing
20000
Settled sludge volume/ml
WAS
Raw 1 h freezing 3 h freezing 72 h freezing
67 59 55 46
4 4 5 3
Settled sludge volume/ml 78 65 59 52
6 5 5 4
10000 T S of mixed sludge VS of mixed sludge
T S of WAS VS of WAS
0 0
20
40 freezing time /h
60
Table 3 e Centrifugation of sludges (initial sludge volume: 100 ml).
80
recorded to determine the average specific resistance of filter cake. A digitally controlled centrifuge (TDL-40B, ANKE Shanghai, China) with a rotational speed of 4000 rpm for 30 min was used in the centrifugal settling tests. Four tubes of sludge samples with an initial volume of 100 ml were centrifuged with the sediment volumes recorded over time.
3.
Results and discussion
3.1.
Physical characteristics
The F/T treatment did not yield significant changes in the sludge TS or VS (Fig. 1), a self-evident result since the freezing and curing did not remove organic matters or induce evaporation loss from sludge. This observation correlates with that by El-Hadj et al. (2007). The capillary suction time (CST) tests were conducted but would not be discussed based on the comments by Ormeci et al. (2001) that CST is not an appropriate indicator for
b 100
80
80
40 original sludge 1h freezing 3h freezing 72h freezing
20
Freezing time for WAS/h
2 2 1 1
47 22 30 26
0 1 3 72
60
40 original sludge 1h freezing
20
3h freezing 72h freezing
0
0 0
20
40 60 settling time /h
80
100
Sediment volume/ml
3 2 2 2
sludge dewaterability. Fig. 2a and b shows the sediment volume versus settling time data for the original and the treated sludges. The F/T treatment enhanced the settleability of sludge samples as noted by the higher settling speed and the less sediment volumes. It is also noticeable that within 1 h and 3 h freezing (and 3 h thawing), the improvement of sludge settleability was noted marginal for both WAS and mixed sludge. However, with a long curing time of about 69 h (72 he3 h), the sludge settleability was significantly improved. Curing has minimal effects on sediment volumes (Table 2). The F/T treatment enhances centrifugal settling of sludge, correlating with the report by Vesilind et al. (1991b). Similar to the gravity tests, the curing has minimal effects on sediment volumes of the centrifugated sludge (Table 3). The 1e3 h freezing of bulk sludge could not effectively improve filterability of either mixed sludges or WAS. The 69-h curing had
100
60
32 25 25 22
0 1 3 72
sedimentation volume /ml
sedimentation volume /ml
a
Sediment volume/ml
Freezing time for mixed sludge/h
Fig. 1 e TS and VS of mixed sludges and WAS versus freezing time.
0
20
40 60 settling time /h
80
Fig. 2 e Settling tests for original and treated sludges. (a) Mixed sludge, (b) WAS.
100
5972
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 6 9 e5 9 7 6
69-h curing produced excess quantity of fine particles for WAS, but had minimal effects on those for mixed sludge (Fig. 3). This result correlates with Vesilind and Martel (1990) which concluded that F/T treatment was most effective with small particles.
Table 4 e Filtration tests for original and treated sludges (initial volume: 100 ml). Filtrate volume/ml
68 63 65 70
0 1 3 72
Freezing time for WAS/h
2 2 2 2
Filtrate volume/ml
53 55 57 80
0 1 3 72
3.2.
3 2 3 2
negligible enhancement of mixed sludge filterability; conversely, curing markedly reduced cake resistance for WAS (Table 4). The result of curing on WAS was in agreement with that by Vesilind and Martel (1990). Lee and Hsu (1994) also noted that the F/T treated biological sludges could be almost completely dewatered via gravitational filtration. The F/T treatment did not alter particle size distributions for mixed sludge or WAS at 1e3 h freezing. In particular, the
a
Chemical characteristics
The COD solubilization reached 0.6% for mixed sludge and 1.6% for WAS after 3 h freezing þ 3 h thawing F/T treatment (Fig. 4). The presence of primary sludge increases resistance to solubilization action by the F/T treatment of WAS. In the subsequent curing stage (3e72 h), the COD solubilization was increased with curing time in a linearly manner, reaching 7.5% for mixed sludge and 10.5% for WAS at the end of the 72-h test. This level of solubilization is comparable to that from WAS sample treated at 100 C for 30 min (Bougrier et al., 2008) and with 0.8 W/ml ultrasound for 5 min (Zhao et al., 2010). The 1e3 h freezing þ 3 h thawing could release a limited quantity of NH3eN from sludge (Fig. 5). Conversely, curing effectively solubilizes NH3eN from sludge matrix into
60
percent of particles /%
50 40 30 original mixed sludge 1 h freezing
20
3 h freezing 10
72 h freezing
.3 3~ 12 .6 12 .6 ~1 7. 2 17 .2 ~2 3. 3 23 .3 ~3 1. 7 31 .7 ~4 3. 1 43 .1 ~5 8. 6 58 .6 ~7 9. 79 6 .6 ~1 08 10 .2 8. 2~ 14 7. 1 14 7. 1~ 20 0
6.
9.
8~ 9
.8
.0
0~ 6 5.
3.
7~ 3
2.
2~
b
7~ 5
.7
2. 7
0
90 80
percent of particles /%
70 60 50 40
original WAS
30
1 h freezing 3 h freezing
20
72 h freezing
10
1. 7 31 .7 ~4 3. 1 43 .1 ~5 8. 6 58 .6 ~7 9. 79 6 .6 ~1 08 10 .2 8. 2~ 14 7. 1 14 7. 1~ 20 0
.3 ~3
3. 3
2
~2
~1
7.
.2 17
12 .6
.6
9.
12
3 9.
3~
8 6. 8~
5. 0~
6.
0 5.
3. 7~
7 3.
2. 7~
2~
2. 7
0
23
Freezing time for mixed sludge/h
Fig. 3 e Particle size distribution for original and treated sludges. (a) Mixed sludge, (b) WAS.
5973
12
7.2
10
7.1
8
7
6
6.9
mixed sludge WAS
pH
COD solubilization /%
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 6 9 e5 9 7 6
6.8
4
6.7
mixed sludge
2
WAS
6.6
0 0
20
40 freezing time /h
60
80 6.5 0
Fig. 4 e COD solubilization with freezing time.
10
20
30
40
50
60
70
80
freezing time /h
Fig. 6 e Suspension pH with freezing time. supernatant. Using ultrasonic treatment a dramatic quantity of NH3eN was released (Feng et al., 2009). Fig. 6 shows the pH versus freezing time data for both treated sludges. The suspension pH slightly decreased following F/T treatment (from 6.89 to 6.57 for WAS and from 7.16 to 6.78 for mixed sludge). The decrease in suspension pH is attributable to the release of fatty acids from solid phase (Montusiewicz et al., 2010), which was also noted by Stabnikova et al. (2008) for food ¨ rmeci and Vesilind (2001) for activated sludge, and waste, by O by Liu et al. (2009) on marine intertidal sludge. We conducted Fourier transform infrared spectroscopy (FT-IR) tests for the original and treated sludge (Fig. 7). The intensity of the 1546 cm1 peak was decreased following F/T treatment, corresponding to the solubilization of proteins into supernatants.
3.3.
Effects of curing on sludge conditioning
Curing is a storage process of frozen sludge under subfreezing temperatures (Parker and Collins, 1997a). The unfrozen zones presented in the frozen bulk sludge can be further dewatered by the surrounding ice (Vesilind and Martel, 1990). As Jean
250
NH3-N /mg•L-1
200
150
100 mixed sludge
50
WAS
0 0
10
20
30 40 50 freezing time /h
60
70
Fig. 5 e NH3eN in suspension for treated sludge.
80
et al. (2000) mentioned, completeness of curing can be rationalized by the accomplishment of dehydration of moisture that can be frozen. Jean et al. (2000) also claimed that the mass transfer rate of intra-aggregate water to diffuse to the growing ice front determines the time needed for sludge curing. Table 5 lists the percentage of changes of sludge properties after F/T treatments for the present biological sludges. Most enhancement of sludge dewaterability was achieved in the freezing stage. The WAS was more readily conditioned by 1e3 h freezing þ 3 h thawing, as noted by the 73.1e81% reduction of sediment volumes for WAS compared with the 57.1e70% for mixed sludge. Based on the conceptual models by Vesilind and Martel (1990) and by Parker and Collins (1999), the freezing process involved rejection and entrapment of sludge flocs. Smaller solid particles associated with WAS would be more readily moved by advancing ice front and be dehydrated into solids pockets. Restated, the bulk freezing of sludge is sufficient to transform puffy sludge structures into compact aggregates to facilitate settleability and filterability. According to the postulated mechanism for sludge freezing, Vesilind and Martel (1990) pointed out that the freezing rate was possibly the governing variable that determined the dewaterability of freezeethaw sludge compared with curing temperature and time, inasmuch as freezing temperature affected the movement and aggregation of solids. If a proper freezing temperature (freezing rate) was applied, the free water surrounding the flocs and surface water surrounding the particles would be frozen in sequence, causing the water molecules being extracted from flocs interior to build the crystalline structures, and forcing the particles migrated into tightly compacted solids pockets. This conclusion on sludge dewaterability was also proposed by Parker and Collins (1997b). Conversely, the freezing stage released limited quantities of COD (15.5% for WAS and 8.1% for mixed sludge). Curing contributed mostly on COD solubilization and NH3eN release (Table 5). The interactions between unfrozen zones in bulk sludge and the surrounding front not only dehydrate the
5974
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 6 9 e5 9 7 6
Fig. 7 e FT-IR spectra for original and treated sludges. (a)Mixed sludge, (b) WAS.
Table 5 e Relative contribution of freezing (1e3 h of freezing test) and curing (3e72 h of freezing test) for mixed sludge and WAS. Sample Mixed sludge WAS
Stage
Settled sludge volume
Centrifugal settling volume
COD solubilization
NH3eN release
Freezing Curing Freezing Curing
57.1% 42.9% 73.1% 26.9%
70% 30% 81% 19%
8.1% 91.9% 15.5% 84.5%
15.2% 84.8% 4.1% 95.9%
aggregates as commented by Jean et al. (2000), but also solubilize the organic matters into supernatants. Freezing causes cell disrupt through intracellular and extracellular ice crystals formed during freezing (Thomashow, 1998; Ormeci and Vesilind, 2001). As temperature decreases below 0 C, ice forms and accumulates in the intercellular spaces, which results in the physical cell disruption (Thomashow, 1998). However, most cellular damage results from the freezeinduced dehydration. At temperatures below 10 C (such as 18 C adopted in this study), with the consequent low water potentials and severe dehydration, membrane damage can occur in the form of “fracture-jump lesions” (Thomashow, 1998). In addition, the water in sludge cells expands during freezing process, and cells vulnerable to the pressure of the expanding ice may burst. It is also possible that the compression and suction on cells exerted by the advancing ice front may cause the cells to disrupt (Ormeci and Vesilind, 2001). Gao (2011) summarizes that intracellular ice formation that usually occurs during rapid freezing may cause the mechanical disruption of cells while slow freezing (such as the one used in this study) often results in the release of more outer-membrane materials. The release of ECPs and intracellular materials to the surroundings contributed to the significant increase of SCOD and NH3eN concentrations in freezeethaw sludge. Ice formation is generally initiated in the intercellular spaces, as opposed to intracellularly, as a result of high freezing point and homogeneous ice-nucleation sites for the former. So the surface water (difficult-to-freeze water) takes a longer time to freeze (Thomashow, 1998; Vesilind and
Martel, 1990). The cell freezing process is governed by the competition between the mass transfer (intracellular water movement) and the heat transfer, which is distinguished by a freezing rate of 3000 K/min (Silvares et al., 1975). The average temperature-decreasing rate in this study was approx. 0.183 K/min, a much lower value water transport would not be dominating in the freezing process. Based on the calculation by Jean et al. (2000), our curing time should be 3380 s, of the same order of that noted in experiments. Natural freezing and thawing can be a promising pretreatment stage of biological sludge to enhance dewaterability. In case the F/T treated sludge is to be used as feedstock of the following anaerobic digestion process, sufficient curing time is needed to allow development of intra-aggregate ice to solubilize organic matters to supernatants.
4.
Conclusions
The following conclusions are drawn based on the presented experimental results: (1) Freeze/thaw (F/T) treatment could enhance biological sludge dewaterability. A 72-h treatment at 18 C decreased the sedimentation volumes by 31.2e31.3% for mixed sludge and by 33.3e44.7% for waste activated sludge, respectively. (2) F/T treatment could facilitate mass transfer from the solid phase into the aqueous phase. The maximum chemical oxygen demand (COD) solubilization obtained in the study were 7.5% for mixed sludge and 10.5% for waste activated
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 6 9 e5 9 7 6
sludge. And maximum increments in NH3eN concentration reached 45.3% for mixed sludge and 74.5% for waste activated sludge. (3) Most enhancement of sludge dewaterability was achieved during bulk freezing stage, with the waste activated sludges more readily dewatered than the mixed sludges after F/T treatment. (4) Unlike sludge dewaterability, the freezing stage released only limited quantities of organic matters to liquid. Thus, COD solubilization and NH3eN release relied mostly on the curing stage.
Acknowledgments The authors gratefully acknowledge funding from Project 50821002 (National Creative Research Groups) supported by National Nature Science Foundation of China, National Water Pollution Control and Management Key Project (2009ZX07317008), partial supports by State Key Laboratory of Urban Water Resource and Environment, Harbin Institute of Technology (No. 2010DX17), and funding from Heilongjiang Province Science Foundation for Youths (QC2009C113).
references
Bougrier, C., Delgene`s, J.P., Carre`re, H., 2008. Effects of thermal treatments on five different waste activated sludge samples solubilisation, physical properties and anaerobic digestion. Chemical Engineering Journal 139 (2), 236e244. Chang, M.R., Chiang, L.I., Lee, D.J., Liu, J.C., Wu, N.M., Chen, W.C., Hsu, B.M., 2004. Conditioning of wastewater sludge from science-based industrial park using freezing and thawing. Journal of Environmental Engineering-ASCE 130 (12), 1552e1555. China EPA, 2002. Water and Wastewater Monitoring and Analyzing, fourth ed. Chinese Environmental Science Press, Beijing, China. Chu, C.P., Lin, W.W., Lee, D.J., Chang, B.V., Liao, G.S., Peng, X.F., 2002a. Thermal treatment of activated sludge using liquid boiling. Journal of Environmental Engineering-ASCE 128 (11), 1100e1103. Chu, C.P., Lee, D.J., Chang, B.V., Liao, G.S., Tay, J.H., 2002b. Weak ultrasonic pretreatment on anaerobic digestion of flocculated waste activated sludge. Water Research 36 (11), 2681e2688. El-Hadj, T.B., Dosta, J., Marquez-Serrano, R., Mata-Alvarez, J., 2007. Effect of ultrasound pretreatment in mesophilic and thermophilic anaerobic digestion with emphasis on naphthalene and pyrene removal. Water Research 41 (1), 87e94. Elliott, A., Mahmood, T., 2007. Pretreatment technologies for advancing anaerobic digestion of pulp and paper biotreatment residues. Water Research 41 (19), 4273e4286. Feng, X., Lei, H.Y., Deng, J.C., Yu, Q., Li, H.L., 2009. Physical and chemical characteristics of waste activated sludge treated ultrasonically. Chemical Engineering and Processing 48 (1), 187e194. Gao, W., 2011. Freezing as a combined wastewater sludge pretreatment and conditioning method. Desalination 268 (1e3), 170e173. Hedstrom, A., Hanaeus, J., 1999. Natural freezing, drying, and composting for treatment of septic sludge. Journal of Cold Regions Engineering 13 (4), 167e179.
5975
Hung, W.T., Chang, I.L., Lee, D.J., Hong, S.G., 1996. Sludge chemical composition changes under uni-directional freezing. Water Science and Technology 34 (3e4), 525e531. Jean, D.S., Lee, D.J., 2000. Effects of electrolytes and curing on freezing of ferric hydroxide sludge. Colloids and Surfaces APhysicochemical and Engineering Aspects 162 (1e3), 285e288. Jean, D.S., Chu, C.P., Lee, D.J., 2000. Effects of electrolyte and curing on freeze/thaw treatment of sludge. Water Research 34 (5), 1577e1583. Kawasaki, K., Matsuda, A., Yamashita, H., 2004. The effect of freezing and thawing treatment on the solid liquid separation characteristics of bulking activated sludge. Kagaku Kogaku Ronbunshu 30 (5), 587e591. Lee, D.J., Hsu, Y.H., 1994. Fast freezeethaw treatment on excess activated sludges e floc structure and sludge dewaterability. Environmental Science and Technology 28 (8), 1444e1449. Liu, H.Y., Wang, G.C., Zhu, D.L., Pan, G.H., 2009. Enrichment of the hydrogen-producing microbial community from marine intertidal sludge by different pretreatment methods. International Journal of Hydrogen Energy 34 (24), 9696e9701. Martel, C.J., Diener, C.J., 1991. A pilot scale study of alum sludge dewaterability in a freezing bed. Journal of the American Water Works Association 83 (12), 51e55. Montusiewicz, A., Lebiocka, M., Rozej, A., Zacharska, E., Pawłowski, L., 2010. Freezing/thawing effects on anaerobic digestion of mixed sewage sludge. Bioresource Technology 101 (10), 3466e3473. Ormeci, B., Vesilind, P.A., 2001. Effect of dissolved organic material and cations on freezeethaw conditioning of activated and alum sludges. Water Research 35 (18), 4299e4306. Parker, P.J., Collins, A.G., 1999. Dehydration of flocs by freezing. Environmental Science & Technology 33 (3), 482e488. Parker, P.J., Collins, A.G., Dempsey, J.P., 1998a. Alum residual floc interactions with an advancing ice/water interface. Journal of Environmental Engineering 124 (3), 249e253. Parker, P.J., Collins, A.G., Dempsey, J.P., 1998b. Effects of freezing rate, solids content, and curing time on freeze/thaw conditioning of water treatment residues. Environmental Science and Technology 32 (3), 383e387. Parker, P.J., Collins, A.G., 1997a. Unidirectional freezing of wasteactivated sludge: the presence of sodium chloride e Comment. Environmental Science and Technology 31 (12), 3740. Parker, P.J., Collins, A.G., 1997b. Feasibility study on freeze/thaw conditioning of pulp mill waste activated sludge. Journal of Cold Regions Engineering 11 (3), 245e250. Stabnikova, O., Liu, X.Y., Wang, J.Y., 2008. Digestion of frozen/ thawed food waste in the hybrid anaerobic solideliquid system. Waste Management 28 (9), 1654e1659. Silvares, O.M., Cravalho, E.G., Toscano, W.M., Huggins, C.E., 1975. The thermodynamics of water transport from biological cells during freezing. Journal of Heat Transfer 97 (4), 582e590. Tao, T., Peng, X.F., Lee, D.J., 2006. Interaction between wastewater-sludge floc and moving ice front. Chemical Engineering Science 61 (16), 5369e5376. Thomashow, M.F., 1998. Role of cold-responsive genes in plant freezing tolerance. Plant Physiology 118 (1), 1e7. Vesilind, P.A., Wallinmaa, S., Martel, C.J., 1991a. Freezeethaw sludge conditioning and double-layer compression. Canadian Journal of Civil Engineering 18 (6), 1078e1083. Vesilind, P.A., Martel, C.J., 1990. Freezing of water and wastewater sludges. Journal of Environmental Engineering 116 (5), 854e862. Vesilind, P.A., Hung, W.Y., Martel, C.J., 1991b. Agitation and filterability of freeze/thawed sludge. Journal of Cold Regions Engineering 5 (2), 77e82. Wang, Q.H., Fujisaki, K., Ohsumi, Y., Ogawa, H.I., 2001. Enhancement of dewaterability of thickened waste activated
5976
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 6 9 e5 9 7 6
sludge by freezing and thawing treatment. Journal of Environmental Science and Health Part A-toxic/hazardous Substances & Environmental Engineering 36 (7), 1361e1371. Wang, X.A., Zheng, M.Y., Zhang, W.Y., Zhang, S., Yang, T., 2010. Experimental study of a solar-assisted ground-coupled heat pump system with solar seasonal thermal storage in severe cold areas. Energy and Buildings 42 (11), 2104e2110.
Whiteley, C.G., Lee, D.J., 2006. Enzyme technology and biological remediation: a review. Enzyme Microbial Technology 38 (3e4), 291e316. Zhao, Q.L., Hu, K., Miao, L.J., Wang, K., 2010. Effect of Ultrasonic Treatment on Characteristics of Waste Activated Sludge Environmental Pollution and Public Health (EPPH 2010) within iCBBE, June 21e23, 2010, Chengdu, China.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 7 7 e5 9 8 6
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Leaching techniques to remove metals and potentially hazardous nutrients from trout farm sludge I.S. Jung a,b,1, R.W. Lovitt a,b,* a
Centre for Complex Fluid Processing (CCFP), School of Engineering, Swansea University, Talbot building, Singleton Park, Swansea, Wales SA2 8PP, UK b Multidisciplinary Nanotechnology Centre (MNC), School of Engineering, Swansea University, Talbot building, Singleton Park, Swansea, Wales SA2 8PP, UK
article info
abstract
Article history:
A fish farm sludge high in P (2e6% w/w as dry matter), Fe (5e7%), C (40e50%) and N
Received 22 February 2011
(0.8e4%) was subjected to a series of acid leaching treatments using HCl, organic acids, and
Received in revised form
biologically mediated acid production. Additions of biodegradable organic acid solubilized
24 August 2011
heavy metals better than HCl, while additions of 1.5% w/v glucose followed by 7 day
Accepted 29 August 2011
incubation stabilized the sludge releasing 92% P, 100% Fe. The use of homo-lactic Lacto-
Available online 7 September 2011
bacillus plantarum starter cultures were more effective than hetero-lactic Lactobacillus buchneri, solubilizing 81.9% P, 92.2% Fe, 93.0% Zn and 96.4% Ca in the sludge. The anaerobic
Keywords:
sludge-glucose fermentation using L. plantarum produced a leached sludge that has low
Metal leaching
heavy metal and nutrient content while affording the recovery of nutrients. The potential
Nutrient leaching
of these methods for practical application are briefly discussed.
Phosphate
ª 2011 Elsevier Ltd. All rights reserved.
Sludge Lactobacillus buchneri Lactobacillus plantarum Recirculating aquaculture system (RAS)
1.
Introduction
The effluents and the sludge from recirculation aquaculture systems RAS are nutrient rich materials high in carbon, nitrogen and phosphorous and accumulated heavy metals. These materials can contribute to eutrophication when released to inland and coastal waters (Barak and Rijn, 2000; Cripps and Bergheim 2000; McIntosh and Fitzsimmons 2003; Lovitt et al., 2008). Furthermore an accumulation of P and Fe in recirculation water can interfere with pumping and bio-
filters resulting in poor hygiene and disease (Barak and Rijn, 2000; Bayat and Sari, 2010; Lekang et al., 2000; Hunter et al., 2001; Pathak et al., 2009; Patterson et al., 2003). Therefore adequate treatment of these wastes prior to discharge or reuse of wastewater is highly desirable. In RAS most nutrients (60e70% w/v) are in the form of solid particles of residual feeds and fish feces rather than liquid, and so can be easily reduced by removal of solid particles with coarse filters, drum filters, bio-filters and by gravitational sedimentation many of the potential harmful substances can
* Corresponding author. Centre for Complex Fluids Processing and Centre for Water Advanced Technologies and Environmental Research, College of Engineering, Singleton Park, Swansea, SA2 8PP Wales, UK. Tel.: þ44 1792295709. E-mail addresses:
[email protected] (I.S. Jung),
[email protected] (R.W. Lovitt). 1 Tel.: þ44 1792295198. 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.08.062
5978
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 7 7 e5 9 8 6
be removed as, or absorbed onto, solids or destroyed in biological oxidation (Thomas et al., 2000; Van Rijn et al., 2006) prior to discharge (Hussenot et al., 1998; Klas et al., 2006). In addition, other processes typically involve insoluble salt precipitation, use of a microbial mass and plant biomass in constructed wetlands (Doyle and Parsons 2002; de-Bashan and Bashan 2004). The separated sludge from recirculating stream of rearing water also contains high concentrations of heavy metals and is a natural sink for such materials. There are many methods of sludge disposal, however with growing environmental concerns, incineration can be a cause of the air pollution (Babel and del Mundo Dacera, 2006), land application and landfill can be restricted due to land contamination by heavy metals and the nutrient load in the sludge. In many cases heavy metals have to be removed prior to land application and landfill of the sludge other wise it can be considered a toxic waste (Sreekrishnan and Tyagi 1996). The dewatered sludge can be potentially recycled to composts, feeds and high value products if contaminating heavy metals, dioxins, poly-chlorinated biphenyls (PCBs) and human pathogens in the sludge are appropriately reduced by chemical or biological treatments. The removal of P and heavy metals in the sludge can be carried out using acid-treatment, chelating agents, biological treatment, electroreclamation and supercritical fluid extraction (SFE). In metal leaching, the pH is very important and can be achieved by the use of mineral or organic acids or it can be mediated via microbial processes that form organic acids (Chen and Lin 2001). It has been shown that microbial mediated leaching, or bioleaching, can be superior to other treatments (Sreekrishnan and Tyagi 1996; Babel and del Mundo Dacera, 2006). Also, organic acids and biological treatments are better because of their biodegradability and the relative mildness of the processes involved. Bioleaching is also more beneficial than chemical leaching alone due to an improved bioavailability of materials and its detoxification during the microbe-mediated processing of sludge (Chen et al., 2005). The efficiency of bioleaching to reduce metal content and toxicity of the sludge was verified using terrestrial and liquid-phase bioassay, rendering the sewage sludge useable for land application (Renoux et al., 2001). One drawback to biological treatment is that it is more difficult to carry out than chemical treatments as it is typically dependent on environmental control of pH and temperature and the provision of nutrient and energy sources. Energy sources like glucose, molasses and non toxic metal salts with high redox potential can accelerate leaching of the heavy metals. Acidophilic obligate chemolithoautotropics like Leptospirillum ferooxidans (Bosecker 1997), Acidothiobacillus ferrooxidans and Acidothiobacillus thiooxidans at
that are capable of degradating of organic material to release bound phosphates and poly-phosphates (Zamudio et al., 2001). Thus for example, they may be also be able modify sludge making it more easy to dewater. The leached materials are then recovered as insoluble materials if treated appropriately. For example, phosphate can be captured in a form of hydrated Amorphous Calcium phosphate (ACP, Ca3(PO4)2 X(H2O)) or hydroxyapatite (Ca10(PO4)6 (OH)2), vivianite (Fe3(PO4)2 8H2O), struvite (NH4MgPO4 6H2O) in presence of appropriate concentrations of Calcium, Ferrous and Magnesium ions in neutral or alkaline condition. The resultant capture phosphate salts have considerable value for industrial applications (Doyle and Parsons, 2002; de-Bashan and Bashan, 2004; Suzuki et al., 2005). Struvite is a route for phosphate recovery used for land application as it is precipitated with nitrogen in wastewater. ACP and vivianite can be subjected to incineration. Hydroxyapatite has potential as this a major component of bone. Solubilized metals can also be recovered from the leachates by chemical precipitation at neutral pH as hydroxide, carbonates or sulphides. In this investigation, sludges have been treated with organic and mineral acids together with inoculation of heterolactic bacterium, Lactobacillus buchneri or homo-lactic bacterium, L. plantarum. These processes were then compared for their ability to leach of metals and nutrients from sludge derived from a RAS trout farm. In this case we have investigated the leaching P, Zn, Ca and Fe from the sludge. In addition we have also studied the changes in sludge characteristics caused by the leaching process.
2.
Materials and methods
2.1. farms
The sludge and waste effluents from Danish trout
The aquaculture sludge used for stabilization and bioleaching studies were obtained from a RAS Trout Farm in Hoghoj, Denmark. The sludge from the farm was stored in an open sedimentation tank and is derived from two streams, one from a drum filter (75 micron mesh) backwash and was a continuous flow, the other stream sludge was from biofilter washing and intermittent stream from daily washing of the filter. Samples were collected and stored at 4 C for up to 24 h prior to analysis. Wastewater and the sludge were collected and centrifuged (10,000 g at 4 C) for concentration prior to treatments.
2.2. solids
Measurement of total suspended solids and volatile
Wastewater from trout farm was filtered with GF/C (Whatman) filter which was soaked in deionized water for one day before filtration and then dried in an oven at 105 C. GF/C filters with filtrates were dried at 105 C dry oven for 4 h. Total suspended solids (TSS) (g L1) is a weight difference between the blank GF/C filter and the weighed according to the standard water and wastewater assay method (Standard methods for the examination of water and waste water 20th (ed.), 1998).
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 7 7 e5 9 8 6
Volatile solids were measured by filtering sludge through GF/C filter and then determining the weight loss after burning samples in a furnace (550 C) for 1 h (Standard methods for the examination of water and waste water 20th (ed.), 1998).
2.3. Measurement of the heavy metals, total carbon and total nitrogen in permeates and sludge Permeates obtained by filtration of trout farm wastewater were used for heavy metal analysis using Inductive Coupled Plasma (ICP) e OES (Spectro Instruments). In order to measure heavy metals in the sludge (mg kg1 of sludge dry matter (DM)), the dried filtrates (105 C) were ashed in a furnace at 550 C. The resultant ash (20e30 mg) was then dissolved in 5 ml of 15.0 mol L1 nitric acid and heated at 190 C for 1 h. The solution obtained was then diluted by at least 10 times with deionized water and this solution was for ICP analysis (Standard methods for the examination of water and waste water 20th (ed.), 1998). Total nitrogen was measured by Kjeldahl method (Standard methods for the examination of water and waste water 20th (ed.), 1998). Total carbon was measured by phenol-sulfuric acid method after digestion of the sample with sulfuric acid. A standard curve to measure carbon concentration was established using glucose concentration from 0 to 10 mg L1 9 ml the sludge sample was incubated with 1 mL of 10 mol L1 sulfuric acid at 100 C for 1 h. The digests were then centrifuged in micro-centrifuge at 12000 rpm for 3 min to measure free carbon in the supernatant at 490 nm (Khodes et al., 2008).
2.4. pH titration and characterization of the sludge; particle sizes and zeta-potential The pH titration of sludge suspension was carried to measure buffering capacity of insoluble and soluble materials of aquaculture wastewater. Using a 6.6 g dry matter (DM) L1 stirred suspension, quantities of HCl 0.1 mol L1 were added to 100 ml sludge suspensions and the pH of samples was measured. Electrophoretic mobility and zeta-potential of solid particles in the sludge was measured with Malvern ZetaMastersizer 2000 as function of pH. The pH was changed to the required value by additions of HCl 0.1 mol L1. The size distribution of the sludge was measured with a Malvern Particle Mastersizer 2000 in deionized water.
2.5. acid
P and Fe extraction by hydrochloric, acetic and lactic
A sample containing 1 g sludge DM L1 of deionized water was prepared for extraction of P and Fe. Additions of lactic acid 1.0 mol L1, or acetic acid 1.0 mol L1 or HCl 0.1 mol L1 were used to reduce pH to 4.0, the samples were then kept at 5 C for 24 h, after which the filtered (GFC) soluble forms of P and Fe were measured by ICP-OES.
5979
2.6. The influence of glucose addition on stabilization the sludge For leaching of metal and phosphorus from the aquaculture sludge, 0e15 g L1 glucose was added to the sludge suspensions containing 6e8 g L1 of suspended solids. The unsterilized aquaculture suspensions with glucose were then incubated at 30 C for 7 days. The soluble forms of P, Fe, Zn and Ca together with other selected elements were measured each day for 7 days.
2.7. The influence of lactic acid bacteria in anaerobic digestion of the aquaculture sludge Cultures of hetero-lactic Lactobaccilus. buchneri (obtained from Interprise Ltd (U.K)), and homo-lactic L. plantarum (ATTC 8014) were added to the sludge to assess their ability to aid leaching. Stock cultures were prepared as follows: cultures were grown for 24 h in shaken anaerobic 20 mL pressure tubes (Bellco) containing 10 mL MRS medium (pH 6.0) (Difco Lab., Detroit, MI, USA) at 30 C. Liquid cultures using MRS broth were carried out nitrogen filled pressure tubes that were sealed with rubber seals and aluminum caps and sterilized prior to inoculation. The grown cultures were then harvested by micro-centrifuge (13000 rpm for 3 min) and then washed aseptically in sterile distilled water, three times. These stock suspensions of L. buchneri or L. plantarum were preserved in e 70 C ultra-freezer until the use. The starter cultures were prepared just before the experiment. The e 70 C ultra-freezer stock suspensions were then thawed at 30 C prior to use and resuscitated in MRS medium in pressure tubes. The pre-cultivation was performed in 10 ml pressure tube under the same growth condition as above for 24 h. The starter cultures were used for inoculation of the main cultures with 10% v/v. The main cultures were performed in 50 ml serum vials with 35 ml medium comprising of aquaculture sludge suspensions (6e8 g L1 suspended solids) and glucose 15 g L1 as a carbon source. During the fermentation of the sludge with L. buchneri or L. plantarum the amounts of P, Fe, Ca, Zn and other selected elements leached from the sludge were analyzed by ICP-OES daily, during the 7 day incubation.
3.
Results and discussion
3.1. Characterization of the aquaculture sludge from Danish trout aquaculture farm In order to improve the efficiency of water use in land-based recirculating aquaculture farms (RAS), rearing water is typically reused via inline filtering, biological treatments and disinfection. The sludge is therefore collected as a product of water recycling processing in RAS and potential problems associated with it include, phytotoxicity, hazardous organic nutrients and disease (Barak and Rijn, 2000, Cripps and Bergheim 2000; Hussenot et al., 1998; Klas et al., 2006; Lovitt et al., 2008; McIntosh and Fitzsimmons 2003; Thomas et al., 2000; Van Rijn et al., 2006).
5980
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 7 7 e5 9 8 6
Samples of sludge from the primary sedimentation tank were taken over a period of three months (annual production was over 30 t DM) and Table 1 shows the averaged composition of the sludge of 4 samples over this period. The particle size of the solids in the sludge is ranged from 10 to 500 mm. The sludge particles contain high content of total carbon (40e50% as DM) total nitrogen (0.8e4% as DM). In particularly the trout farm sludge used in this investigation was high in P (1.5%), Fe (2.7%), Ca (2.7%) and Zn (0.03%) as dry matter (DM) and were mainly associated with sludge particles which originated from fish manures, residual feeds and feed water from biofilter washes and drum filter backwashing as shown at Table 2. Most nutrients and the heavy metals are in a particulate form rather than dissolved in the liquid, therefore most components can be removed by separation of liquids from the solids. The metal distribution in the trout farm sludge showed a similar metal distribution pattern to that shown in shrimp aquaculture sludge by Nemati et al., (2009). These authors also showed high concentration of Zn, Fe and Ca in Malaysian shrimp aquaculture sludge.
3.2. pH titration of trout farm sludge and zeta-potential of sludge particles The buffering capacity of the aquaculture wastewater was measured in a titration with 0.1 mol L1 HCl. The sludge contains charged organic and inorganic materials that consequently contribute to the buffering capacity. Fig. 1 shows the pH titration curves of for 100 ml waste effluent samples containing the sludge suspension (6.6 g DM L1), water and sludge permeate using additions 0.1 mol L1 HCl. The sludge samples showed much higher buffering capacity than water and filtered waste effluents. The pH was gradually reduced from neutrality to about pH 4 using about 35 ml of 0.1 mol L1 HCl after which buffering capacity increased considerable requiring a further 50 ml to reduced the pH to 3. A calculation of the amounts of acid required to reduce pH of the sludge can be made. Therefore to reduce the pH 7 to pH 4, 5.3 mmol of HCl per g DW sludge would be required. Also, Fig. 2 shows variations in zeta-potential and mobility of solid particles in the aquaculture sludge suspension with pH using
Table 2 e The element analyses in the sludge from the primary sludge tank. Sludge tank
S P Si Co Mg K Mn Ca Fe Cr Na Cl Al Zn
GFC permeate mg L1
GFC retentate g kg1 sludge DM
7.1 0.01 <0.1 5.36 0.01 <0.01 4.72 0.03 4.11 0.02 <0.00 26.3 0.3 <0.01 <0.01 16.6 0.2 25.5 0.02 <0.01 0.05 0.005
1.76 0.07 14.72 0.1 <0.01 <0.01 0.76 0.004 0.203 0.003 1.3 0.006 27.29 0.8 26.97 1.42 <0.01 <0.01 <0.01 <0.01 0.280 003
0.1 mol L1 HCl. Electrophorietic mobility and zeta-potentials of the solid particles were reduced, with zeta-potential changing from e 23 mV at pH 7 to - 15 mV at pH 4. The reduced net negative surface charge is most probably due to changes in charge on phosphate and carboxylic acid groups as they become more protonated at low pH.
3.3. acid
Leaching by additions of HCl, lactic acid and acetic
The leaching of P, Fe, Ca, and Zn was investigated with inorganic and organic acids. HCl was compared with lactic acid and acetic acid in leaching P, Fe, Zn, and Ca out of the farm sludge and the results are shown in Table 3. When pH was lowered to pH 4.0 with 1.0 mol L1 lactic acid, soluble forms of Fe was increased to 88.21% in permeate of the sludge water and P increased 32.43% as a soluble form in permeate. Also, Zn
Table 1 e The characteristics of the sludge obtained from primary sedimentation tank of the trout farm. Characteristic pH TSS in 105 C (mg L1) Volatile Solids in 550 C (mg L1) Ash (mg L1) Particle Size distribution (mm) Total Carbon (mg L1) Total Nitrogen (mg L1) Total Phosphorus (mg L1) Total Iron (mg/L1) Total Silicon (mg L1) Total Calcium (mg L1) a mean particle size range.
Value 6.8e7.3 16,580e31,010 11,470e17,350 5110e13,660 10e500 (meana: 150e250) 5160e7800 365.7e488.5 579.5e926.9 1120e2300 3.7e6.3 313.6e614.0
Fig. 1 e The pH titration curves of water (6), permeate (B) and the sludge suspension (C) with 0.1 mol LL1 HCl. The Trout farm sludge suspension was filtered through GF/C filter (Whatman). The suspension contains 6.6 g suspended solid LL1.
5981
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 7 7 e5 9 8 6
Fig. 2 e The effect of pH on (A) the zeta-potential and (B) electrophoretic mobility of the trout farm sludge suspension. The sample contains 6.6 g suspended solid LL1. The pH was adjusted using 0.1 mol LL1 HCl.
initial pH was 6.0 and pH stayed in pH 6.0 without glucose. With an addition of 3 g L1 glucose to the sludge suspension and after 7 days incubation the pH was reduced to 5.0. Similar trends were also observed with the addition of 6 and 9 g L1 glucose to the sludge suspension. As a result the pH of the system changes to 4.5 and 4.2 respectively (Fig. 1). The average zeta-potential was also reduced in these samples (Fig. 2). The suspended solids in the sludge were slightly reduced to 8e9 g DM L1 from 10.5 to 11.5 g DM L1 in anaerobic incubation with glucose concentration over 3 g glucose L1 for 168 h, but a large reduction of the suspended sludge was not detected at higher glucose concentration (data not shown). This indicated that the sludge dry matter was stabilized under high acid conditions. Fig. 3 shows that P, Fe, Ca and Zn (Fig. 3A, B, C, and D respectively) were released from the aquaculture sludge after incubation with glucose for 168 h. The amounts of P, Fe, Ca, and Zn released from the sludge increased when more glucose was added. P was leached out by 91.34% of total P (110.23 mg L1) in the sample during anaerobic incubation of 168 h with 15 g glucose L1. Fe was leached out by 100% of total Fe (259.5 mg L1) in the samples with glucose concentration over 6.0 g/L in 168 h. Ca and Zn were leached out by 84.8% of total Ca (224.2 mg L1) and 76.9% of total Zn (4.1 mg L1) in the samples with 15 g glucose L1, respectively, in 168 h. These experiments show that substantial release of materials is possible by the addition of rapidly fermentable carbohydrate such as glucose.
3.5. A comparison of L. buchneri and L. plantarum fermentation treatment on sludge leaching and Ca solubilized in permeates increased using organic acids rather than HCl. Organic acids were show to have higher efficiency in leaching out the heavy metals than HCl. This may be attributed to the organic acid offering a chelating capacity when complexed with the metals.
3.4. Effect of glucose addition on the sludge suspension leaching The effect of adding glucose to the sludge suspension was then tested. In these experiments, carried out under anaerobic conditions, it was anticipated that the glucose would be fermented by the natural microbial flora of the sludge. This was in fact the case as indicated by the pH decline. The pH decline was dependent on the amount of glucose added (Fig. 3A). The
As well as making glucose addition to the sludge and incubating under anaerobic conditions, a further experiment was carried out to observe the effects of starter cultures of lactic acid bacteria commonly used in silage production. The effects of using L. buchneri, a hetero-lactic fermenter and L. plantarum homo-lactic fermenter were compared for their ability to leach of metals and nutrients and changing the characteristics of the sludge. Fig. 4 shows a comparison of anaerobic fermentation of trout farm sludge using glucose addition alone or with addition of either lactic acid bacteria. During the anaerobic fermentation the pH declined to pH 3.61 from pH 6.16 in all samples after 168 h (Fig. 4A). Dry matter was reduced by 20.2% from 54.4 g L1 to 43.4 g L1 using L. buchneri, and by 10.0% from 53.9 g L1 to 48.5 g L1 using only glucose, but using
Table 3 e The effect of the type of acid used on leaching of Ca, Fe, P, Zn from the trout farm sludge at pH 4.0. The elements were measured with ICP-OES as a soluble form in permeate after filtering through GF/C filter (Whatman). Elements
Ca Fe P Zn
% of total released from sludge 0.05 mol L1 HCl treatment
0.1 mol L1 HCl treatment
0.5 mol L1 HCl treatment
1.0 mol L1HCl treatment
1.0 mol L1 Lactic acid treatment
1.0 mol L1 Acetic acid treatment
52.3 74.9 25.1 15.0
52.2 80.0 26.4 12.8
52.1 81.2 27.2 12.6
51.2 78.8 24.3 12.1
53.1 88.2 32.4 24.2
54.8 77.9 27.1 18.9
5982
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 7 7 e5 9 8 6
Fig. 3 e The effect of glucose addition on trout farm sludge. The release of (A), P; (B), Ca; (C), Fe; (D), Zn from the trout farm sludge were predicted with incubation time after adding glucose. Glucose was added in a series of concentrations from 0 (B) 0 g LL1, 3 g LL1 (C), 9 g LL1 (:), 15 g LL1 (-).
Fig. 4 e The effect of incubation time on (A), the pH; (B) dry matter; (C), P; and (D), Fe- release from the trout farm sludge using 15 glucose g LL1 only (B), L. buchneri and glucose 15 g LL1 (6), and L. plantarum and glucose 15 g LL1 (,).
5983
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 7 7 e5 9 8 6
Table 4 e The effect of anaerobic fermentation treatments on element retention after 7 days incubation in water only, with glucose addition (15 g LL1) and glucose (15 g LL1) D Lactobaclilus plantarum. Element
S P Si Co Mg K Mn Ca Fe Cr Na Al Zn
Initial sludge
Sludge in water
Sludge with Glucose
Glucose þ L. plantarum
g/kg of sludge
g/kg of sludge
Change*
g/kg of sludge
Change*
g/kg of sludge
Change*
10.23 0.04 69.77 1.00 13.98 0.14 0.028 1.78 0.01 1.16 0.02 2.47 0.01 109.84 1.80 53.91 2.50 0.015 1.01 0.01 2.72 0.04 1.42 0.01
7.29 0.09 35.4 0.63 14.20 0.5 0.024 1.65 0.02 0.89 0.04 2.06 0.01 37.49 0.5 56.25 1.15 0.012 0.07 0.02 2.94 0.02 0.92 0.02
28.7% 49.3% þ1.6% 14.3% 7.3% 23.3% 16.6% 65.9% þ4.3% 20.0% 93.1% þ7.3% 35.2%
7.27 0.05 9.976 0.1 10.7 0.04 0.003 0.25 0.01 0.26 0.01 0.06 0.02 2.58 0.03 5.79 0.1 0.008 0.06 0.01 1.47 0.02 0.043 0.001
28.9% 85.7% 23.5% 89.3% 86.0% 77.6% 97.6% 97.7% 89.3% 46.7% 94.1% 46.0% 97.0%
8.27 0.11 12.64 0.05 8.2 0.34 0.003 0.3 0.04 0.43 0.01 0.18 0.01 3.93 0.05 4.21 0.03 0.006 0.12 1.82 0.01 0.1 0.002
19.2% 81.9% 41.3% 89.3% 83.2% 62.9% 92.7% 96.4% 92.2% 60.0% 88.1% 33.1% 93.0%
Change*: change (%) of the initial value. The dry matter (DM), using glucose with L. plantarum, increased by 2.67 times in 168 h. Cd. Cu, Ni and Pb in the trout farm sludge were undetectable.
L. plantarum lead to a slight increase by 4.14% in dry matter (Fig. 4B) presumably because it grew very well producing additional dry mass to that in the sludge. Using L. buchneri for anaerobic fermentation can reduce the amount of DM in the sludge because hetero-lactic acid bacterium can produce CO2, neutral solvents (such as ethanol and propandiol), acetate, lactate and biomass. On the other hand L. plantarum, a homolactic fermenting bacterium does not produce CO2 and yields only lactic acid and thus more acidity, as compared to L buchneri. Using L. plantarum can therefore be effective in preventing a loss of dry matter as CO2 production in anaerobic fermentation (Ranjit and Kung 2000; Kung and Ranjit 2001).
When L. buchneri was used, P and Fe were initially solubilized, however, after 120 h no further leaching was observed. When using only glucose or L. plantarum plus glucose, a steady increase after 120 h was observed. Table 4 shows a comparison of the predominate elements in the sludge leachates after 168 h anaerobic fermentation using sludge alone, 15 g L1 glucose or L. plantarum plus 15 g L1 glucose. During the leaching with L. plantarum plus glucose, DM (g L1) increased by 267% from 1.69 0.14 g L1 DM to 4.51 g L1 DM and with glucose alone by 5.4% from 2.23 0.15 g L1 DM to 2.25 g L1 DM. The major elements P, Fe, Zn and Ca were reduced in the sludge by 81.9%, 92.2% 97.0% and 96.4% respectively, with L.
Fig. 5 e The effect of incubation time on (A), the pH; (B), dry matter; (C), Zeta-potential; (D), mobility of solid particles during anaerobic fermentation using only water (B), 15 g glucose LL1 only (6), and L. plantarum plus 15 g glucose LL1 (,) Anaerobic fermentations were performed in 30 C for 168 h. The distributions of the heavy metals in the sludge after anaerobic fermentation were shown at Table 4.
5984
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 7 7 e5 9 8 6
Table 5 e The heavy metal components found in composts, sewage sludge, shrimp aquaculture pond sludge after natural anaerobic fermentation (g kgL1 sludge DM). Type Composts Composts Composts Sewage sludge anaerobic digests Shrimp Aqua-sludge anaerobic digests Trout Aqua-sludge anaerobic digests UK PAS 100 limits Limits/Quebec
Al 9.1
1.82 25.0
Cd
Cr
Cu
Fe
0.0035 0.0024 0.0007 0.0183 0.0009 0.001 0.0015 0.003
0.12 0.094 0.026 3.809 0.03 0.006 0.1 0.21
0.22 0.08 0.159 0.337 0.005 0.11 0.2 0.1
12.25 7.47 25.208 2.15 4.21
Mg
Mn
Na
0.22 0.042
0.3
0.3 0.18
Ni
P
0.04 0.012 0.021 0.029 0.117
25.0
0.015 0.05 0.062
12.64
Pb 0.18 0.18 0.47 0.167 0.021 0.013 0.2 0.15
Si
8.2
Zn 0.82 0.32 0.399 0.871 0.053 0.1 0.4 0.5
Ref a b c d e f g h
a Walter et al., 2006. ´ lvarez et al., 2002. b A c Smith 2009. d Fuentes et al., 2004. e Nemati et al., 2009. f This investigation. g Waller 2004. h Beauchesne et al., 2007.
plantarum plus glucose, with glucose alone by 85.7%, 89.2% 93.0% and 97.7% respectively. Fig. 5 also shows a large increase in Zeta-potential and mobility of the solid particles with rapid decline of pH in anaerobic fermentation of glucose or with glucose plus L. plantarum. If these measurements are compared with the fresh sludge using HCl additions to acidify the sludge the resultant zeta-potential is more positive when the bacterial inoculum was added as compared with the glucose addition or the control fermentation (cf. Figs. 2 and 4) indicating possible chemical modification of the sludge materials that reduce its net charge. In addition, lactic acid bacteria also can inhibit growth of other bacteria such as competing animal pathogens so potentially reducing their population during the fermentation. Organic acids produced by lactic acid bacteria are also very good not only in reducing pH to acid conditions and inhibiting growth of other bacteria, but also in improving chelation of the sludge particles bound heavy metals. The use of L. plantarum not only better preserved the dry matter, but also improved leaching P and Fe as compared with using L. buchneri. Therefore sludge treatment by fermentation using homo-lactic bacteria was very effective in removing many metals and nutrients from the sludge. This process therefore has potential use for aquaculture sludge treatment so allowing the treated sludge to be applied in agricultural land. Also, sludge treatment with L. plantarum was better for retaining dry matter than hetero-lactic acid bacteria. Additional benefits may also be found in improved dewatering characteristics of the treated sludge, the glucose fermentation with L. plantarum changed the zeta-potential significantly. Although not studied here, it would be expected that the dewatering characteristics would change which should allow greater compressibility of the sludge. Using lactic acid bacterial may also effectively preserve the sludge until it can spread on the land. In a broader context, the treated solid particles of the sludge which are low in the metals and nutrient content can be separated by filtration, centrifugation and sedimentation.
The fluids released from treated sludge that are enriched with metals can be easily treated using simple chemical methods such as neutralization with appropriate alkalis e.g. calcium hydroxide, that will cause the metals and nutrients to precipitate as salts. Alternatively the metals or nutrient can be captured by passage through constructed wetlands. Although the results of these laboratory studies are encouraging there is considerable effort required to carry out these types of process on a large scale. In such situations, the seeding and treatment would have to be optimized carefully and be correlated with the much higher sludge concentrations i.e. >80 g DW L1. For example, rapidly reducing pH may require higher concentrations of glucose to overcome the buffering capacity of the sludge. Table 5 compares the heavy metals in fermented sludge attained in this investigation with them in other composts, sludge composts and UK PAS 100 limits as a standard of the organic composts for land application. The only critical value here was that of Zn and this was reduced substantially allowing the treated sludge to pass BSI PAS 100 guideline (Waller, 2004). Cd, Cr, Cu, Ni and Pb were much less than their limits shown at BSI PAS 100 guidelines.
4.
Conclusions
The experiments on acid leaching treatments of fish farm sludge show that there is considerable potential to remove metals and nutrients. Although simple treatment with mineral acid released metals and nutrients, the use of organic acids and in particular, acids generated biogenically, enhanced this process considerably. Further improvement to a fermentation approach may be made by using rapidly utilizable waste carbon sources such as cellulose hydrolysates from straw or fermentable vegetable wastes. Although the results of these laboratory studies are encouraging there is considerable effort required before these types of processes could be on a large scale.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 7 7 e5 9 8 6
Acknowledgment The authors wish to thank Dr Robin Shields and Dr G. Profitt of CSAR Swansea and Prof V. Zonno of the University of Lecce for valuable discussions concerning the manuscript and the staff of Hoghoy Fish Farm and Aquaprie Denmark for access to sludge and water samples. We would also like to thank the EU for funding some of this work under the Aqua E-Treat project (www.aquaetreat.org) as part of the Horizontal Research Activities involving SMEs (collective research) scheme Contract no COLL-CT-2003-5003 05.
references
´ lvarez, E.A., Mocho´n, C., Sa´nchez, J.C.J., Rodrı´guez, M.T., 2002. A Heavy metal extractable forms in sludge from wastewater treatment plants. Chemosphere 47, 765e775. Babel, S., del Mundo Dacera, D., 2006. Heavy metal removal from contaminated sludge for land application: a review. Waste Management 26 (9), 988e1004. Barak, Y., Rijn, J., 2000. Biological phosphate removal in a prototype recirculating aquaculture treatment system. Aquacultural Engineering 22, 121e136. Bayat, B., Sari, B., 2010. Comparative evaluation of microbial and chemical leaching processes for heavy metal removal from dewatered metal plating sludge. Journal of Hazardous Materials 174, 763e769. Beauchesne, I., Cheikh, R.B., Mercier, G., Blais, J.-F., Ouarda, T., 2007. Chemical treatment of sludge: in-depth study on toxic metal removal efficiency, dewatering ability and fertilizing property preservation. Water Research 41, 2028e2038. Bosecker, K., 1997. Bioleaching: metal solubilization by microorganisms. FEMS Microbiology Reviews 20, 591e604. Chen, S.Y., Lin, J.G., 2001. Bioleaching of heavy metals from sediment: significance of pH. Chemosphere 44, 1093e1102. Chen, Y.X., Hua, Y.-M., Zhang, S.-H., Tian, G.-M., 2005. Transformation of heavy metal forms during sewage sludge bioleaching. Journal of Hazardous Materials B123, 196e202. Cripps, S.J., Bergheim, A., 2000. Solids management and removal for intensive land-based aquaculture production systems. Aquacultural Engineering 22, 33e56. de-Bashan, L.E., Bashan, Y., 2004. Recent advances in removing phosphorus from wastewater and its future use as fertilizer (1997e2003). Water Research 38, 4222e4246. Doyle, J.D., Parsons, S.A., 2002. Struvite formation, control and recovery. Water Research 36, 3925e3940. Fuentes, A., Liore´ns, M., Sa´ez, J., Soler, A., Aguilar, M.I., Ortun˜o, J. F., Meseguer, V.F., 2004. Simple and sequential extractions of heavy metals from different sewage sludges. Chemosphere 54, 1039e1047. Hunter, R.G., Combs, D.L., George, D.B., 2001. Nitrogen, phosphorous, and organic carbon removal in simulated wetland treatment systems. Archive Environental Contamination and Toxicology 41, 274e281. Hussenot, J., Lefebvre, S., Brossard, N., 1998. Open-air treatment of wastewater from landbased marine fish farms in extensive and intensive systems: current technology and future perspectives. Aquatic Living Resources 11 (4), 297e304. Khodes, V.B., Fernandes, L., Bhosle, N.B., Sardessai, S., 2008. Carbohydrates, uronic acids and alkali extractable carbohydrates in contrasting marine and estuarine
5985
sediments: distribution, size fractionation and partial chemical characterization. Organic Geochemistry 39, 265e283. Klas, S., Mozes, N., Lahav, O., 2006. A conceptual, stoichiometrybased model for single sludge denitrification in recirculating aquaculture systems. Aquaculture 259, 328e341. Kung Jr., L., Ranjit, N.K., 2001. The effect of Lactobacillus buchneri and other additives on the fermentation and aerobic stability of barley silage. Journal of Dairy Science 84, 1149e1155. Lekang, O.-I., Bergheim, A., Dalen, H., 2000. An integrated wastewater treatment system for land-based fish-farming. Aquacultural Engineering 22, 199e211. Lovitt, R.W., Jung, I., Proffit, G., Priess, G., Zonno, V., 2008. The sludge concentration system and environmental performance of Hoghoj fish farm; a Danish recirculation fish farm producing Trout. In: Resources Management: Natural Human and Materials Resources for the Sustainable Development of Aquaculture, ISBN 978-83-60111-30-7, pp. 389e390. Compiled by Kamlor E and Dabrowski K. Conference Short Communications Aquaculture Europe 08, Krakow Poland. McIntosh, D., Fitzsimmons, K., 2003. Characterization of effluent from an inland, low-salinity shrimp farm: what contribution could this water make if used for irrigation? Aquacultural Engineering 27, 147e156. Nemati, K., Bakar, N.K.A., Abas, M.R., 2009. Investigation of heavy metals mobility in shrimp aquaculture sludge-comparison of two sequential extraction procedures. Microchemical Journal 91, 227e231. Pathak, A., Dastidar, M.G., Sreekrishnan, T.R., 2009. Bioleaching of heavy metals from sewage sludge: a review. Journal of Environmental Management 90, 2343e2353. Patterson, R.N., Watts, K.C., Gill, T.A., 2003. Micro-particles in recirculating aquaculture systems: determination of particle density by density gradient centrifugation. Aquacultural Engineering 27, 105e115. Ranjit, N.K., Kung Jr., L., 2000. The effect of Lactobacillus buchneri, Lactobacillus plantarum, or a chemical preservative on the fermentation and aerobic stability of corn silage. Journal of Dairy Science 83, 526e535. Renoux, A.Y., Tyagi, R.D., Samson, R., 2001. Assessment of toxicity reduction after metal removal in bioleached sewage sludge. Water Research 35 (6), 1415e1424. Smith, S.R., 2009. A critical review of the bioavailability and impacts of heavy metals in municipal solid waste composts compared to sewage sludge. Environment International 35, 142e156. Sreekrishnan, T.R., Tyagi, R.D., 1996. A comparative study of the cost of leaching out heavy metals from sewage sludge. Process Biochemistry 31 (1), 31e41. Standard methods for the examination of water and waste water 20th, 1998. The Water Environment Federation, the American water Works Association, and the American Public Health Association Part 4000. Suzuki, K., Tanaka, Y., Kuroda, K., Hanajima, D., Fukumoto, Y., 2005. Recovery of phosphate from swine wastewater through crystallization. Bioresource Technology 96, 1544e1550. Thomas, M.L., Alexander, O.H., Denis, P.D., 2000. The design and operational characteristics of the CP&L EPRI fish barn: a demonstration of recirculating aquaculture technology. Aquacultural Engineering 22, 3e16. Van Rijn, J., Tal, Y., Schreier, H.J., 2006. Denitrification in recirculating systems: theory and applications. Aquacultural Engineering 34, 364e376. Waller, P., 2004. Guidelines for the Specification of Composted Green Materials Used as a Growing Medium Component BSI
5986
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 7 7 e5 9 8 6
(British Standard Institution). www.wrap.org.uk the Waste & Resource Action Programme. Walter, I., Martı´nez, F., Cala, V., 2006. Heavy metal speciation and phytotoxic effects of three representative sewage sludges for agricultural uses. Environmental Pollution 139, 507e514.
Weinberg, Z.G., Muck, R.E., 1996. New trends and opportunities in the development and use of inoculants for silage. FEMS Microbiology Reviews 19, 53e68. Zamudio, M., Gonza´lez, A., Medina, J.A., 2001. Lactobacillus plantarum phytase activity is due to non-specific acid phosphatase. Letters in Applied Microbiology 32, 181e184.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 8 7 e5 9 9 6
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Oxygen transfer and uptake, nutrient removal, and energy footprint of parallel full-scale IFAS and activated sludge processes Diego Rosso a,*, Sarah E. Lothman b,c, Matthew K. Jeung a, Paul Pitt d, W. James Gellner e, Alan L. Stone b, Don Howard f a
Department of Civil and Environmental Engineering, University of California, Irvine, CA 92697-2175, USA Hazen and Sawyer, P.C., 4011 WestChase Blvd, Raleigh, NC 27607, USA c Department of Environmental Sciences and Engineering, University of North Carolina, Chapel Hill, NC 27599 7431, USA d Hazen and Sawyer, P.C., 498, 7th Avenue, NY 10018, USA e Hazen and Sawyer, P.C., 11311 Cornell Park Drive, Cincinnati, OH 45242, USA f Department of Water Reclamation, Greensboro Water Resources, 2350, Huffine Mill Road, McLeansville, NC 27301, USA b
article info
abstract
Article history:
Integrated fixed-film activated sludge (IFAS) processes are becoming more popular for both
Received 3 April 2011
secondary and sidestream treatment in wastewater facilities. These processes are
Received in revised form
a combination of biofilm reactors and activated sludge processes, achieved by introducing
26 August 2011
and retaining biofilm carrier media in activated sludge reactors. A full-scale train of three
Accepted 29 August 2011
IFAS reactors equipped with AnoxKaldnes media and coarse-bubble aeration was tested
Available online 3 September 2011
using off-gas analysis. This was operated independently in parallel to an existing full-scale activated sludge process. Both processes achieved the same percent removal of COD and ammonia, despite the double oxygen demand on the IFAS reactors. In order to prevent
Keywords:
kinetic limitations associated with DO diffusional gradients through the IFAS biofilm, this
Activated sludge Integrated
fixed-film
activated
systems was operated at an elevated dissolved oxygen concentration, in line with the
sludge
manufacturer’s recommendation. Also, to avoid media coalescence on the reactor surface
Oxygen transfer
and promote biofilm contact with the substrate, high mixing requirements are specified.
Nutrient removal
Therefore, the air flux in the IFAS reactors was much higher than that of the parallel
Energy footprint
activated sludge reactors. However, the standardized oxygen transfer efficiency in process
Aeration
water was almost same for both processes. In theory, when the oxygen transfer efficiency is the same, the air used per unit load removed should be the same. However, due to the high DO and mixing requirements, the IFAS reactors were characterized by elevated air flux and air use per unit load treated. This directly reflected in the relative energy footprint for aeration, which in this case was much higher for the IFAS system than activated sludge. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
Biofilm processes exist in nature and are responsible for a significant contribution to the carbon and nitrogen cycles in
aquatic environments. Engineered biofilm processes, such as trickling filters preceded the development of the activated sludge a century ago, and were already included as common practice in the first edition of Metcalf and Eddy (1914). Even
* Corresponding author. Tel.: þ1 949 824 8661; fax: þ1 949 824 3672. E-mail address:
[email protected] (D. Rosso). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.08.060
5988
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 8 7 e5 9 9 6
Nomenclature AFR ATANK BHP c COD DWP hL Jair MCRT MLSS MLVSS OD OTE OTR OUR Pd Pi
air flow rate (m3 s1) tank bottom area power drawn by the blower (kW) molecular weight of air (kg kmol1) chemical oxygen demand ðmgO2 l1 Þ dynamic wet pressure (Pa) head loss in the air distribution line (Pa) Air flux (m s1) mean cell retention time (d) mixed liquor suspended solids (mgTSS l1) mixed liquor suspended solids (mgVSS l1) oxygen demand ðkgO2 s1 Þ oxygen transfer efficiency (%) oxygen transfer efficiency ðkgO2 s1 Þ oxygen uptake rate ðmgO2 l1 h1 Þ discharge pressure (Pa) inlet pressure (Pa)
though the activated sludge process (ASP) is currently the most used biological treatment for wastewater and water reclamation, several biofilm processes have been used over time (Alleman, 1983) and are being considered for both secondary treatment and sidestream applications (e.g., sludge centrate treatment). The integrated fixed-film activated sludge (IFAS) process was introduced fifteen years ago (Randall and Sen, 1996) as an evolution of the moving bed biofilm reactor (Rusten et al., 1992), and has since been demonstrated in pilot- and fullscale installations. This technology integrates the traditional biofilm characteristics and the completely mixed conditions of ASP by introducing floating plastic carriers onto which biofilm can establish. By preferentially retaining these carriers inside the reactors using screens and allowing the transit of the activated sludge media through the reactor, the bacterial colony is dual in nature and is composed of well-mixed flocs and adhered biofilm. The introduction of carrier media increases the process biomass inventory, thus allowing an increase in the process loading rates without requiring expansion in tankage. Based on these considerations, the IFAS process poses as an ideal candidate for plant retrofits and process upgrades, especially when the process site is land constrained. The enhanced removal of chemical oxygen demand (COD) and biological nutrient removal have been well-demonstrated in the application of IFAS to CAS. With IFAS, sufficient COD removal can be achieved at higher organic loads, lower hydraulic retention times and with smaller basin volumes (Andreottola et al., 2003; Sriwiriyarat et al., 2008). Enhanced nitrogen removal was accomplished up to full-scale (Randall and Sen, 1996) and enhanced phosphorous removal was accomplished up to pilot-scale (Sriwiriyarat and Randall, 2005). Besides improving the level of treatment in CAS, additional advantages have been reported in IFAS processes including greater process stability (Sriwiriyarat et al., 2008), reduced sludge production and reduced solids loadings on the secondary clarifiers (Stricker et al., 2009), and enhanced sludge settleability (Kim et al., 2010).
Q R SOTE SOTR T Wair Z
influent flow rate (m3 d1) universal gas constant (8.314 J mol1 K1) standard OTE in clean water (%) standard oxygen transfer efficiency ðkgO2 s1 Þ ambient temperature (K) ponderal air flow (kg s1) hydrostatic pressure corresponding to diffuser submergence (Pa)
Greek letters a alpha factor ¼ aSOTE/SOTE () aSOTE Standard OTE in process water (%) ε energy footprint per unit oxygen demand oxidized ¼ (kW h kg1 O2 ) g ratio of specific heats at constant pressure and volume (0.283 for air) h combined motor and blower efficiencies () air density (kgair m3air) rair
In recent years, the removal of endocrine disruptor compounds (EDC) and other recalcitrant organic compounds in wastewater treatment has been the object of research. There is potential for enhanced EDC removal in the IFAS process due to its higher biomass concentration than a suspended process with equal mean cell retention time (MCRT). The MCRTs typically associated with elevated mixed liquor suspended solids (MLSS) concentrations (i.e., larger than 10e15 days) may enhance the adsorption and/or biodegradation of these micropollutants (Soliman et al., 2006) and, although results may vary highly between activated sludge processes and membrane bioreactors, in general the enhanced removal of EDC from the water phase appears independent of MCRT when exceeding 10e15 days (Clara et al., 2005). Previous tests showed that almost all steroidal estrogens can be removed by biofilm processes such as the biological aerated filter equipped with floating biofilm carriers (Joss et al., 2004). However, Ort et al. (2010) remarked recently the need for improved sampling procedures. The differential removal of organics in the IFAS and ASP processes are currently subject to investigation, and early results show no evidence of qualitative differential removal between the two processes (Gonsior et al., 2011). Oxygen transfer tests and analyses are key components for improved process design and optimization (Plano et al., 2010). In activated sludge processes, oxygen transfer analyses using the off-gas technique have been extensively practiced for the past 30 years (inter alia, Redmon et al., 1983). In moving biofilm reactor technology, fewer published research works on oxygen transfer analyses are available: independent studies on biological aerated filters (BAF) using off-gas analysis were conducted in small scale (Harris et al., 1996) and full depth reactors (Stenstrom et al., 2008), but not for the IFAS process. The oxygen uptake rate (OUR), a key parameter that can be obtained from off-gas testing, is typically used as indicator of microbial metabolism. The OUR in IFAS systems was previously evaluated (Maas et al., 2008). Those studies suggested that COD and nutrient removal rates can be controlled by determining OUR. Maas et al. (2008) found from OUR tests that
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 8 7 e5 9 9 6
the biofilm on the media perform the majority of nitrification in an IFAS system. That study, however, did not measure oxygen transfer efficiency (OTE) and oxygen transfer rates (OTR), two performance parameters that can be measured via off-gas analysis. Off-gas analysis is a direct measurement of oxygen transfer and is based on the mass balance of the water column under a hood floating onto the aerated wastewater surface (Redmon et al., 1983). Off-gas tests were previously performed on an IFAS system by a manufacturer (Viswanathan et al., 2008). To our knowledge, our paper presents the first independent off-gas investigation on a fullscale IFAS system to date. To compare aeration performance between different aeration systems and for the same system in different process conditions and times, OTE is typically corrected to standard conditions [zero dissolved oxygen (DO), zero salinity, 101,325 Pa, 298.15 K] and SOTE is typically used to characterize performance in clean water (ASCE, 2007). Due to bubble geometry, SOTE decreases rapidly with increasing air flow rate per diffuser (Zlokarnik, 1980) and are in general lower for coarse-bubble diffusers than for fine-bubble diffusers (US. EPA, 1989). Oxygen transfer efficiency can be affected dramatically by the process conditions: for activated sludge plants, the transfer efficiency is reduced at high F/M or low MCRT operation (Rosso et al., 2005). The parameter employed to quantify the depression of oxygen transfer in process water is the a factor, calculated as (Stenstrom and Gilbert, 1981): a¼
aSOTE kL aprocess water ¼ kL aclean water SOTE
(1)
In the aerobic zones, the soluble COD and its readily biodegradable fraction (rbCOD) can suppress the a factor for fine-pore diffusers, and therefore depress the SOTE in process condition, or aSOTE (Eckenfelder and Barnhart, 1961; Mancy and Okun, 1960; Hwang and Stenstrom, 1979; Wagner and Po¨pel, 1996; Rosso and Stenstrom, 2006). When coarsebubble diffusers are installed, a is virtually unaffected by the wastewater contaminants (Eckenfelder and Ford, 1968; Rosso et al., 2005). Hence, the coarse-bubble diffusers typically employed in IFAS reactors to promote mixing and to meet high oxygen uptake rates are expected to be associated with elevated a factors but low SOTE values (IWA, 2008). Also, it is known that higher MCRT processes using fine bubble diffusers (e.g.: processes performing biological nutrient removal or BNR) have higher a factors throughout the plug flow aeration tank (Rosso et al., 2008). Hence, the oxygen transfer depression in the reactors with fine-pore diffusers is expected to be mitigated by the increased MCRT specified to perform BNR. In IFAS systems, coarse-bubble diffusers for aeration are typically installed. These diffusers release air through a macroscopic orifice (several millimeters in diameter), and bubble detach with equivalent diameter of approximately 50 mm, corresponding to the maximum diameter of stability for air bubbles in water (Clift et al., 1979). In coarse-bubble diffusers, the bubble release is dominated by the velocity of the air flow through the orifice. Coarse-bubble diffusers tend to regurgitate air at low flow rate and stabilize the release of bubbles at higher air flow rates, exhibiting in general a bubble size distribution independent of air flow (Wang and Hung,
5989
2007). Oftentimes, in IFAS systems, the bubbles may be referred to as “medium-bubbles”, because on the IFAS reactor surface may be evident that the majority of bursting bubbles are of the same size of the carrier media or smaller (i.e., below 50 mm). It can be inferred that although the diffuser itself releases coarse-bubbles, the transit of bubbles through a wellmixed reactor with solid media must be associated with bubble shearing into smaller bubbles that burst at the tank surface in the “medium” range (with size distribution between 5 and 50 mm in diameter). For a typical IFAS reactor, the air diffuser system consists of stainless steel headers with small holes in the bottom of the header or air spargers mounted to the header top. The design of the system is intended to minimize the need for in-tank maintenance. Because of the presence of the IFAS media, the coarse-bubble oxygen transfer efficiency may be expected to increase due to the bubble hold-up in the tank and to the splitting of the bubbles when passing through the areas filled with IFAS media. In this study we confirmed this phenomenon only visually and no quantitative analysis of bubble size distribution in IFAS reactors is yet published. Also, IFAS systems must be operated with increased DO concentrations relative to ASP reactors to facilitate bulk liquid DO diffusion through the biofilm. Both the coarse-bubble system and the increased target DO increase the air requirements for IFAS systems. Energy use is a growing concern at wastewater treatment facilities and since aeration has been identified as one of the most energy intensive unit operations in wastewater treatment (Reardon, 1995), aeration efficiency monitoring poses as a key candidate for the reduction of energy consumption and operating costs (Leu et al., 2009). It is therefore critical to comparatively quantify the ASP and IFAS aeration efficiencies and energy costs. The goal of this research is to quantify oxygen transfer and uptake in IFAS systems and to present the results of our comparative energy footprint analysis for the two processes. Due to the unique opportunity afforded by this plant’s configuration, we tested the hypothesis that the IFAS process can provide a much higher capacity in the same volume, but partially at the cost of a less efficient aeration system.
2.
Methodology
2.1.
Process operation
The T.Z. Osborne Water Reclamation Facility is a 15,000 m3/d plant located in Greensboro, NC (USA). Major unit processes at the plant include influent screening, primary clarifiers, conventional plug flow activated sludge, secondary clarifiers, effluent sand filters, and disinfection. The plant aeration basins are arranged to operate in parallel. Several alternatives are being evaluated as strategies to meet stringent forthcoming regulatory nutrient limits, and consequently a fullscale IFAS unit equipped with AnoxKaldnes media was installed in side-by-side configuration with the existing ASP. The plant performs BNR with a Ludzack-Ettinger configuration (sludge return rate w 100% of influent flow). For the IFAS reactors, the previously existing fine bubble disc diffusers
5990
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 8 7 e5 9 9 6
Table 1 e Plant Operating Conditions During Testing. January 2010 Parameter 3
1
Influent Flow m d COD, mg l1 Primary effluent Final effluent MLSS (mgTSS l1) MLVSS (mgVSS l1) Temp, MLSS ( C) Mixed liquor MCRT (d) NH4eN, mg l1 Primary Effluent Final Effluent
June 2010
Tank 11 ASP
Tank 12 IFAS
Tank 11 ASP
Tank 12 IFAS
11,000
22,700
11,400
22,700
182 37.8 3110a 2375a 14.2 7.0
182 36.8 1215a 928a 14.5 4.0
337 31.7 2780 N/Ab 26.8 6.0
337 35.4 1655 N/Ab 26.8 3.0
12.9 <0.1
12.9 <0.1
23.8 <0.1
23.8 0.2
a Average of values collected Tuesday-Friday of testing week since testing day was a holiday. b MLVSS not measured for June 16, 2010. June 17, 2010 MLSS values were 2840 and 1815 for Tanks 11 and 12, respectively. MLVSS values were 2220 and 1390 for Tanks 11 and 12, respectively.
(230 mm in diameter) in the three IFAS cells were removed and replaced with the coarse-bubble diffuser system recommended by the manufacturer to promote mixing and maintain elevated DO levels in the IFAS reactors. Throughout the ASP and in the portion of Tank 12 that followed the IFAS reactors, the fine-pore disc diffusers previously installed were operated. Throughout this paper we refer to as ASP for the process in Tank 11 and as IFAS for the process in Tank 12. At the time of both tests, there were no unusual plant conditions. Table 1 details the key operating parameters for the plant. At the time of testing, the two tanks were operating using the same wastewater influent, but with roughly double hydraulic load (Q) and oxygen demand OD (kgO2 s1 ) OD ¼ Q$ DCOD þ 4:33$DNHþ 4
(2)
hydraulic load on the IFAS implies that the hydraulic retention time there was half than that of the ASP. The plant was able to isolate one aeration basin and one clarifier completely from the rest of the system to allow retrofit with the IFAS process. The aeration tank 12 and a clarifier (No. 7) were isolated and used during the process pilot and this study for the IFAS installation. The ASP and IFAS reactors were hence operated independently, i.e. with separate clarifiers and sludge lines. Aside from the presence of IFAS media and different diffusers in the IFAS reactors, Tanks 11 and 12 were identical (volume ¼ 6246 m3 each), therefore acting as two separate and parallel wastewater treatment processes treating the same wastewater.
2.2.
Off-gas analysis
where DCOD ¼ COD oxidized in the aerobic reactors (mg l1) 1 DNHþ 4 ¼ ammonia oxidized in the aerobic reactors (mg l ).
Notwithstanding this, the IFAS had comparable effluent concentrations for COD and ammonia (Table 1). The double
Off-gas analysis was performed in the same fashion as described by the US EPA standard protocol for testing in process water (US. EPA, 1989). The off-gas technique is based upon the original method developed by Redmon et al. (1983). Fig. 1 illustrates the testing layout. The technique consists on
Fig. 1 e Off-gas testing layout. This non-invasive testing method allows the concurrent measurement of the actual oxygen transferred to the water (oxygen transfer efficiency [OTE, %], oxygen transfer rate ½OTR; kgO2 hL1 , and oxygen uptake rate ½OUR; mgO2 lL1 hL1 ).
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 8 7 e5 9 9 6
Fig. 2 e Testing Hood Positions.
5991
5992
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 8 7 e5 9 9 6
an oxygen mass balance on the water column underneath a floating off-gas collection hood. We constructed an analyzer with the same configuration as described in detail by Leu et al. (2009). At each measurement the apparatus self-calibrates by zeroing with an initial sample of atmospheric air. The off-gas was collected in two floating hoods providing a capture area of 3 m2 each. In the tests performed at this plant, the carbon dioxide and water vapor were removed from the off-gas prior to analysis by using a desiccating column filled with silica gel granules and sodium hydroxide pellets. The OTE is calculated from the off-gas oxygen mole fraction and the known ambient air mole fraction (20.95%). The gas flow rate is not needed to perform this calculation, although it is useful to calculate flow-weighted averages over an aeration tank or across several aeration tanks. Two testing locations were chosen in each reactor cell in order to gather a representative section of each cell. Only the last cell of each process was tested in the middle. For each of the testing locations, duplicate measurements are taken within 5 min, to limit the effect of process loading variability on the measurement itself. For each of the two tanks, 18 off-gas samples were taken for each of the Summer and Winter measurement campaigns, for a total of 72 samples taken in this research overall. Fig. 2 shows the hood locations. The total sampled area exceeded the protocol requirement of 2% tank surface coverage. The test was first performed in the Winter season (January 2010). Winter testing allows the measurements of the aeration efficiency parameters at peak loading in cold water, which is the worst case scenario for aeration efficiency and is the design target for the aeration system. The test was repeated in the Summer (June 2010) to verify the aeration efficiency parameters in warm water while the metabolic rates, hence the OUR, are expected to be highest. Also, the coldest weather is the design criteria for blower power and the warmest weather is the design criteria for blower capacity (i.e., AFR), and it is therefore valuable to repeat the test for the two temperatures for a comparison of the normalized performance data, i.e. the energy footprint per unit oxygen demand removed εðkWh kg1 O2 Þ.
3.
Results and discussion
3.1.
Process performance
Table 1 shows the secondary effluent concentrations for COD and ammonia. Both processes [ASP (Tank 11), IFAS (Tank 12)] remove the carbonaceous and nitrogenous loads to the same extent, meeting discharge requirements, notwithstanding the double oxygen demand (expressed as kgO2 s1 ) on the IFAS reactor. Concurrent investigations (Gonsior et al., 2011) showed no evidence of differential qualitative removal of organics in the two processes, but only a quantitative difference (i.e., IFAS removed a double load than ASP). Fig. 3 shows the profiles of nitrogen species along both reactors for the tests in winter and summer. The ammonia profiles show that the ammonia concentration in the first denitrification zone (Cell A in Fig. 3) during the winter are roughly half than those in the primary effluent, a dilution
associated with the mixing between denitrification influent and return activated sludge. In fact, the portion of ammonia assimilated by denitrifiers for synthesis although is usually embedded in the mass balance discrepancy and when not neglected is calculated by difference between the ammonia in the reactor influent and its oxidation products (Silverman and Schroeder, 1983; Argaman, 1986). The summer test shows less decrease in ammonia from primary effluent to the first denitrification reactor (Cell A in Fig. 3) despite the higher ammonia assimilation expected with warmer temperature, may be caused by an increased transformation of TKN to ammonia. In both tests and for both reactors, the bulk of the ammonia is removed in the first two aerated reactors (Cells D and E in Fig. 3) for both the ASP and the IFAS. For each of the tests, the two processes have consistent profiles for all nitrogen species. In the IFAS reactor (Tank 12), the reactor zones with IFAS media (Cells D, E, and F in Fig. 3) show no different behavior than their parallel ASP zones in Tank 11. The overall effect is secondary effluent values for ammonia, nitrite, and nitrate for IFAS consistently lower or equal to those of ASP. One nitrate measurement, the primary effluent concentrations for the summer test in Fig. 3 appear unusually high for an LeE process (i.e., without internal recycling) and can be considered an outlier. Although no samples upstream of the primary effluent were collected, this outlier is suspected to be due to a non-representative variation of the primary effluent at the time of sampling. The nitrate profiles in the winter increase regularly along the process train, and appear consistent with the expectations for a plug-flow L-E process.
3.2.
Off-gas test results
The off-gas testing results are summarized in Table 2. The differences between the IFAS tank and the ASP are apparent. The results from the both tests confirmed that the IFAS is characterized by elevated air flux due to mixing requirements
Fig. 3 e Comparative nutrient profiles during both tests.
5993
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 8 7 e5 9 9 6
Table 2 e Comparative summary of results for the off-gas tests performed in January and June 2010. The total number of data points in this flow-weighted averages is 18 per tank and per test (total number of data points [ 72). All values are airflow-weighted except for DO. The coefficient of variation for each of the datasets summarized below are reported in their respective figures. OTE (%)
aSOTE (%)
Air Flux (m s1)
DO (mg l1)
OUR (mg l1 h1)
JANUARY 2010 11 (ASP) 12 (IFAS)
11.7 6.3
15.7 11.0
1.53 103 5.11 103
1.0 3.6
36.5 80.0
JUNE 2010 11 (ASP) 12 (IFAS)
11.3 7.8
15.3 13.5
2.15 103 4.74 103
2.2 3.8
53.4 105.4
Tank No.
specified by the process manufacturer, with associated lower OTE. It is critical to remark that, for Tank 12, only the IFAS reactor cells (Cells D, E, F as shown in Fig. 2) contain coarsebubble diffusers. The remaining aerobic cells after the IFAS process in Tank 12 (cells G, H, I) contain fine bubble diffusers. In order to provide a comparison that accounts for different diffuser geometry, the results are also normalized as air flux (air flow per unit tank bottom area) and air use (air volume per unit mass load removed). For the latter, the total oxygen demand (OD) to be oxidized in the aerobic reactors was considered. By using the actual COD and ammonia profiles to calculate the total OD to the aerobic reactor we accounted for the denitrification credit, i.e. the fraction of rbCOD removed during denitrification. In Table 2, the results for each tank are reported as airflow-weighted averages (except for DO). Flowweighted averaging is important to account for the variability of the process conditions throughout the tank, and is performed as (ASCE, 1997):
Cell D
Cell E
Cell F
Cell G
x¼
(3)
where x ¼ parameter to average (e.g., OTE, aSOTE, etc.) xi ¼ parameter measured at position i AFRi ¼ air flow at position i i ¼ sample location. Flow-weighing is necessary to calculate an appropriate average for a process where the air flow may be variable for the different sampling locations. This is the case of the process tested here, where for example the IFAS has for some of the sampling points an air flux several times higher than that of the ASP. Due to the elevated DO values specified for the IFAS design, 1.7e3.6 times higher than the average DO in the ASP, the air flux (air flow per unit tank bottom area) and air use (air volume per unit oxygen demand removed) of IFAS vs. ASP are 1.4e3.1 times and 1.6e2.0 times higher, respectively. The OUR is consistently double for the IFAS, corresponding with the double oxygen demand being removed in the IFAS process.
Cell D
Cell H
Cell E
Cell F
Cell G
Cell H
53- 6
7 4- 8
59
20.0
150.0
6.0 14.0
OTE (%)
100.0 75.0
IFAS January ASP January IFAS June ASP June
18.0
IFAS January ASP January IFAS June ASP June
125.0
OUR (mg l-1 h-1)
Pn i¼1 ðxi $AFRi Þ P n i¼1 AFRi
12.0 10.0 8.0
50.0
6.0 25.0
4.0
IFAS Cells
IFAS Cells
2.0
0.0 1 1- 2
32- 4
53- 6
74- 8
59
Position along tank
Fig. 4 e Comparative OUR profiles for ASP and IFAS during the January and June tests. IFAS cells are installed only at testing positions 1e6, and positions 7e9 are ASP for both tanks. Total number of data points [ 70. The coefficient of variation did not exceed 19.9% and 13.8% for the ASP and IFAS reactor datasets, respectively.
1 1- 2
3 2- 4
Position along tank
Fig. 5 e Comparative OTE profiles for ASP and IFAS during the January and June tests. IFAS cells are installed only at testing positions 1e6, and positions 7e9 are ASP for both tanks. Total number of data points [ 72. The coefficient of variation did not exceed 9.2% and 13.8% for the ASP and IFAS reactor datasets, respectively.
5994
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 8 7 e5 9 9 6
Cell D
28.0
Cell F
Cell G
Cell H
53- 6
7 4- 8
59
IFAS January ASP January IFAS June ASP June
24.0
αSOTE (%)
Cell E
20.0 16.0 12.0 8.0 IFAS Cells
4.0 1 1- 2
3 2- 4
Position along tank
Fig. 6 e Comparative aSOTE profiles for ASP and IFAS during the January and June tests. IFAS cells are installed only at testing positions 1e6, and positions 7e9 are ASP for both tanks. Total number of data points [ 72. The coefficient of variation did not exceed 28.6% and 13.8% for the ASP and IFAS reactor datasets, respectively.
Figs. 4e7 show the OTE, aSOTE, OUR, and Air Flux for both the IFAS and the ASP for both tests, with their respective coefficients of variation. Note that in all figures, IFAS cells are installed only at testing positions 1e6 (cells DeF), and positions 7e9 (cells G-H) are ASP for both tanks. In Fig. 4, the OUR profiles are plotted. The IFAS zones in Tank 12, marked in the graph, are characterized by very high OUR values, atypical for municipal wastewater treatment. This is a result of the ability by the increased IFAS biomass inventory to resolve a double oxygen demand in the same reactor volume. Outside the IFAS zone in Tank 12 (positions 7e9), the OUR values for both tanks are virtually indistinguishable. In Fig. 4 we also show images of
sample biofilm carriers for each of the three IFAS reactors. The visual inspection of the biofilm shows the different thickness and suggests the potentially different ability for bubbles and liquid to reach the inner portion of the carrier which leads to the hypothesis that different microbial diversity may exists in the cavities of the IFAS media, if compared to the outer media surface. This may be associated with differences in nitrification in different points of the IFAS media. In Fig. 5, the OTE profiles illustrate the actual transfer efficiency at the existing temperature and DO concentration. The OTE for the ASP zones corresponding with the IFAS reactors (testing positions 1e6 in Fig. 5) are higher for both tests, but this is dictated by the high DO concentrations. In order to reach such elevated DO, the air flow rate required is high, therefore lowering the OTE for IFAS. Moreover, OTE at high DO is disadvantaged, since the DO is closer to oxygen saturation. Due to the lower driving force for oxygen transfer, while the IFAS process is disadvantaged on the basis of OTE, aSOTE is compensated with a zero DO correction and is in the same range for both processes. The DO correction takes into consideration the incremental difficulty of transferring oxygen when operating closer to the DO saturation value. The standardized process efficiency (aSOTE, Fig. 6) appears therefore similar for both processes.
3.3.
Energy footprint implications
Energy footprint considerations can be concluded directly from the datasets plotted in Fig. 7. For both IFAS and ASP, the energy footprint is dominated by the power requirements of the air blowers. The relationship between blower power (BHP), air flow rate (AFR), and air flux (Jair) can be described using the adiabatic formula (Metcalf and Eddy Inc, 2003): g Pd 1 Pi g Wair RT Z þ hL þ DWP þ Pi 1 $ ¼ Pi cgh g r $AFR$RT Z þ hL þ DWP þ Pi 1 $ ¼ air Pi cgh g r $ATANK $Jair $RT Z þ hL þ DWP þ Pi 1 $ ¼ air Pi cgh
BHP ¼
Wair RT $ cgh
(4)
where BHP ¼ power drawn by the blower (kW)
Fig. 7 e Comparative Air Flux profiles for ASP and IFAS during the January and June tests. IFAS cells are installed only at testing positions 1e6, and positions 7e9 are ASP for both tanks. Total number of data points [ 70. The coefficient of variation did not exceed 30.8% and 68.8% for the ASP and IFAS reactor datasets, respectively.
Wair ¼ ponderal air flow (kg s1) rair ¼ air density (kgair m3air) AFR ¼ volumetric air flow (m3 s1) ATANK ¼ tank bottom area. Jair ¼ Air flux (m s1) R ¼ universal gas constant (8.314 J mol1 K1) T ¼ ambient temperature (K) c ¼ molecular weight of air (kg kmol1) g ¼ ratio of specific heats (0.283 for air) h ¼ combined motor and blower efficiencies () Pd ¼ discharge pressure (Pa) Pi ¼ inlet pressure (Pa) Z ¼ hydrostatic pressure corresponding to diffuser submergence (Pa) hL ¼ head loss in the air distribution line (Pa)
5995
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 8 7 e5 9 9 6
Table 3 e Summary of relative process performance in terms of air use, blower power requirements, and energy footprint. Process Influent Flow Q (IFAS/ASP)winter (IFAS/ASP)summer
2.06 1.99
Oxygen Demand OD ¼ DCOD þ 1.97 1.34
DWP ¼ dynamic wet pressure (pressure drop across the diffusers, Pa) Since in this case the tank bottom area for both the ASP and the IFAS is known and identical, the air flow rate can be calculated from the air flux (plotted in Fig. 7 and averaged in Table 2). By assuming an appropriate dynamic wet pressure for the air diffusers (5475 Pa for the fine-pore diffusers and 1490 Pa for the coarse-bubble diffusers), the blower power requirements BHP can be calculated. In this case, the OD is very different between the two processes, and therefore the process with higher load (the IFAS) would be unfairly disadvantaged if only the BHP were calculated. Therefore, we also introduce the energy footprint per unit oxygen demand oxidized ε: ε ¼ 3600
BHP OD
4:33DNHþ 4
(5)
where ε ¼ energy footprint per unit oxygen demand (kWh=kgO2 ) OD ¼ oxygen demand (kgO2 s1 ) From Fig. 7 it is evident that the IFAS reactors (Cells DeF), due to their design specifications to promote mixing and high DO, are characterized by elevated air flux, which due to its tumultuous nature is associated with larger coefficients of variation. For cells G-H where both basins have fine bubble diffused aeration, the air flux for both basins is similar in the summer test and is slightly lower for the IFAS during winter. This may be due to more rapid removal of the oxygen demand in the IFAS zones upstream from positions G-H. The performance of the activated sludge reactors following the IFAS zones in Tank 12 suggests that this residual reactor volume may be unnecessary, further corroborated by the ammonia profiles in Fig. 3. The relative ratios of the flow-weighted averages of air flux, air use, and energy footprint are reported in Table 3. In theory, when OTE is the same, the air required per unit mass of oxygen demand removed should be the same, regardless of the process. Nevertheless, the specifications for elevated mixing and DO requirements impact air use significantly. Throughout the coarse bubble aeration zones [positions 1e6 in the IFAS process (Fig. 7)], the IFAS has elevated air flux which reflects directly in its energy footprint. The energy footprint per unit load oxidized ε of the IFAS exceeds 2.0 times that of the ASP (Table 3). Due to its double hydraulic load, the IFAS had a hydraulic residence time approximately half that of the ASP, since in this case both thanks were identical. This poses the IFAS as a viable alternative for land-constrained sites where the hydraulic flow rate, hence the total tankage, may be the deciding factor during process selection. This enhancement in oxygen demand
Air Flow
Air Use
Energy Footprint
AFR
AFR/OL
ε ¼ BHP/OL
3.33 2.20
2.08 2.11
2.00 2.09
removal was associated with an elevated air use (air flow per unit oxygen demand removed). The elevated air use associated with the IFAS directly relates to the energy footprint increase, due to the absence of mechanical mixers in the IFAS reactors. This trend is consistent between the January and the June tests. The elevated energy use to promote mixing may suggest that the testing of mechanical mixers combined with reduced air flow or with fine-pore diffusers for oxygen transfer should be the subject of future research.
4.
Summary and conclusions
Off-gas testing was conducted in summer and winter on parallel and independent activated sludge (ASP) and the integrated fixed-film activated sludge (IFAS) processes. The test was performed to determine the comparative oxygen transfer efficiency, oxygen uptake rate, and air use for the two processes, and to quantify temperature effects on the aeration performance parameters. The results show that in both tests the ASP had lower DO, OUR, and air flux than the IFAS, which in turn was characterized by lower oxygen transfer efficiency (OTE). Due to the elevated DO requirements for the IFAS, the standard oxygen transfer efficiency in process water (aSOTE) was comparable between the two processes after correction to zero DO. The hydraulic load and the oxygen demand applied to the IFAS process was approximately twice that of the parallel ASP, and the percent removal of the COD and ammonia was almost same for both processes. This suggests that the IFAS is a viable process for process expansion, especially for land-constrained sites. The increased oxygen demand removal for the IFAS is associated with an elevated air used (i.e., air flow per unit oxygen demand removed), which exceeds that of the ASP by a factor of 2, corresponding to the same excess for energy footprint.
Acknowledgments The authors thank the staff at the T.Z. Osborne Water Reclamation Plant in Greensboro, NC for the help during field testing. This is contribution No. 69 of the UCI Urban Water Research Center.
references
Alleman, J.E., 1983. The History of Fixed-Film Wastewater Treatment Systems. Available on:
.
5996
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 8 7 e5 9 9 6
Andreottola, G., Foladori, P., Gatti, G., Nardelli, P., Pettena, M., Ragazzi, M., 2003. Upgrading of a small Overloaded activated sludge plant using a MBBR system. J. Environ. Sci. Health Part A 38 (10), 2317e2328. Argaman, Y., 1986. Nitrogen removal in a semi-continuous process. Water Res. 20 (2), 173e183. ASCE, 1984, 1991, 2007. ASCE Standard: Measurement of Oxygen Transfer in Clean Water, ISBN 0-87262-430-7 New York, NY. ASCE, 1997. ASCE Standard: Standard Guidelines for In-Process Oxygen Transfer Testing, ISBN 0-78440-114-4 New York, NY. Clara, M., Kreuzinger, N., Strenn, B., Gans, O., Kroiss, H., 2005. The Soilds retention time e A Suitable design parameter to Evaluate the capacity of wastewater treatment plant to remove micropollutants. Water Res. 39, 97e106. Clift, R., Grace, J.R., Weber, M.E., 1979. Bubbles Drops and Particles. Academic Press, New York. Eckenfelder, W.W., Barnhart, E.L., 1961. The effect of organic Substances on the transfer of oxygen from air bubbles in water. AIChE J. 7 (4), 631e634. Eckenfelder, W.W., Ford, D.L., 1968. In: Gloyna, E.F., Eckenfelder Jr., W.W. (Eds.), New Concepts in Oxygen Transfer and Aeration, Advances in Water Quality Improvement. University of Texas Press, Austin, Tex, pp. 215e236. Gonsior, M., Tseng, L.Y., Jeung, M.K., Cooper, W.J., Hertkorn, N., Schmitt-Kopplin, P., Pitt, P., Rosso, D., 2011. Differential Analysis of the Removal of organics in parallel IFAS and activated sludge reactors using ultrahigh resolution mass spectrometry. In: Proc. IWA NOM/DOM/EfOM Conf., Costa Mesa, CA, USA. Harris, S.L., Stephenson, T., Pearce, P., 1996. Aeration investigation of biological aerated filters using off-gas analysis. Water Sci. Technol. 34 (3e4), 307e314. Hwang, H.J., Stenstrom, M.K., 1979. The Effect of Surface Active Agents on Oxygen Transfer. UCLA-ENG-79-30 Report. University of California, Los Angeles. IWA, 2008. Biological Wastewater Treatment e Principles, Modelling and Design. IWA Publishing, London, UK. Joss, A., Anderson, H., Ternes, T., Richle, P.R., Siegrist, H., 2004. Removal of estrogens in municipal wastewater treatment under aerobic and Anaerobic conditions: Consequences for plant optimization. Environ. Sci. Technol. 38, 3047e3055. Kim, H., Gellner, J.W., Boltz, J.P., Freudenberg, R.G., Gunsch, C.K., Schuler, A.J., 2010. Effects of integrated fixed film activated sludge media on activated sludge settling in biological nutrient removal systems. Water Res. 44, 1553e1561. Leu, S.-Y., Rosso, D., Larson, L.E., Stenstrom, M.K., 2009. RealTime aeration efficiency monitoring in the Acvivated sludge process and methods to Reduce energy consumption and operating costs. Water Environ. Res. 81, 2471e2481. Maas, C.L.A., Parker, W.J., Legge, R.L., 2008. Oxygen uptake rate tests to Evaluate integrated fixed film activated sludge processes. Water Environ. Res. 80 (12), 2276e2283. Mancy, K.H., Okun, D.A., 1960. Effects of surface active agents on bubble aeration. J. Water Pollut. Control Fed. 32 (4), 351e364. Metcalf, L., Eddy, H.P., 1914. American Sewerage Practice, first ed.. McGraw-Hill, New York. Metcalf and Eddy Inc, 2003. Wastewater Engineering: Treatment and Reuse, fourth ed. McGraw-Hill, New York, ISBN 0-07041-878-0. Ort, C., Lawrence, M.G., Rieckermann, J., Joss, A., 2010. Sampling for Pharmaceuticals and Personal Care products (PPCPs) and Illicit Drugs in wastewater systems: are Your Conclusions Valid? A critical Review. Environ. Sci. Technol. 44, 6024e6035. Plano, S., Rosso, D., Benedetti, L., Weijers, S., De Jonge, J., Nopens, I., 2010. Towards Dynamic Activated Sludge
Modelling without Need for Calibration. Proc. IWA World Water Congress, Montreal, QC, Canada. Randall, C.W., Sen, D., 1996. Full-Scale Evaluation of an integrated fixed-film activated sludge (IFAS) process for enhanced nitrogen removal. Water Sci. Technol. 33 (12), 155e162. Reardon, D.J., 1995. Turning down the power. Civ. Eng. 65 (8), 54e56. Redmon, D.T., Boyle, W.C., Ewing, L., 1983. Oxygen transfer efficiency measurements in mixed liquor using off-gas Techniques. J. Water Pollut. Control Fed. 55, 1338. Rosso, D., Stenstrom, M.K., 2006. Surfactant effects on alphafactors in aeration systems. Water Res. 40 (7), 1397e1404. Rosso, D., Iranpour, R., Stenstrom, M.K., 2005. Fifteen years of offgas transfer efficiency measurements on fine pore Aerators: key Role of sludge Age and normalized air flux. Water Environ. Res. 77, 266e273. Rosso, D., Larson, L.E., Stenstrom, M.K., 2008. Aeration of largescale municipal wastewater treatment plants: state of the art. Water Sci. Technol. 57 (7), 973e978. Rusten, B., Ødegaard, H., Lundar, A., 1992. Treatment of dairy wastewater in a novel moving bed biofilm reactor. Water Sci. Technol. 26 (3e4), 703e711. Silverman, J., Schroeder, E.D., 1983. Performance of SBR activated sludge process with nitrification/denitrification. J. Water Pollut. Control Fed. 55 (4), 377e384. Soliman, M.A., Pedersen, J.A., Park, H., Castaneda-Jimenez, A., Stenstrom, M.K., Suffet, I.H., 2006. Human pharamaceuticals, antioxidants, and plasticizers in wastewater treatment plant and water reclamation plant effluents. Water Environ. Res. 79 (2), 156e167. Sriwiriyarat, T., Randall, C.W., 2005. Performance of IFAS wastewater treatment processes for biological phosphorous removal. Water Res. 39, 3873e3884. Sriwiriyarat, T., Pittayakool, K., Fongsatitkul, P., Chinwetkitvanich, S., 2008. Stability and capacity enhancements of activated sludge process by IFAS technology. J. Environ. Sci. Health Part A 43, 1318e1324. Stenstrom, M.K., Gilbert, R.G., 1981. Effects of alpha, Beta and Theta factors in design, specification and operations of aeration systems. Water Res. 15, 643e654. Stenstrom, M.K., Rosso, D., Melcer, H., Appleton, R., Langworthy, A., Wong, P., 2008. Oxygen transfer in a fulldepth biological aerated filter. Water Environ. Res. 80 (7), 663e671. Stricker, A.E., Barrie, A., Maas, C.L.A., Fernandes, W., Lishman, L., 2009. Comparison of performance and operation of side-byside integrated fixed-film and conventional activated sludge processes at Demonstration scale. Water Environ. Res. 81, 219e232. US. EPA, 1989. Design Manual e Fine Pore Aeration Systems. Risk Reduction Laboratory, Cincinnati, Ohio. EPA/625/1e89/ 023. Viswanathan, S., Pham, H., Kelly, R.F., Redmon, D.T., Fernandes, W., 2008. Evaluation of Oxygen Transfer Efficiency via Off-gas Testing at Full Scale Integrated Fixed film Activated Sludge Installation, In: Proc. WEFTEC 2008 Conference. Wagner, M., Po¨pel, H.J., 1996. Surface active agents and their influence on oxygen transfer. Water Sci. Technol. 34 (3e4), 249e256. Wang, L.K., Hung, Y.-T., 2007. Advanced Physicochemical Treatment Technologies, vol. 5. Humana Press, Totowa, NJ, USA. Zlokarnik, M., 1980. Eignung und Leistungsfa¨higkeit von Volumenbelu¨ftern fu¨r biologische Abwasserreinigungsanlagen. Korrespondenz Abwasser 27 (3), 194e209.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 9 7 e6 0 1 0
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Environmental and economic profile of six typologies of wastewater treatment plants G. Rodriguez-Garcia a,*, M. Molinos-Senante b, A. Hospido a, F. Herna´ndez-Sancho b, M.T. Moreira a, G. Feijoo a a b
Department of Chemical Engineering, School of Engineering, University of Santiago de Compostela, 15782 Santiago de Compostela, Spain Department of Applied Economics II, Faculty of Economics, University of Valencia, Campus dels Tarongers, 46022 Valencia, Spain
article info
abstract
Article history:
The objective of wastewater treatment plants (WWTPs) is to prevent pollution. However, it
Received 10 May 2011
is necessary to assess their sustainability in order to ensure that pollution is being
Received in revised form
removed, not displaced. In this research, the performance of 24 WWTPs has been evaluated
17 August 2011
using a streamlined Life Cycle Assessment (LCA) with Eutrophication Potential (EP) and
Accepted 29 August 2011
Global Warming Potential (GWP) as environmental indicators, and operational costs as
Available online 5 September 2011
economic indicators. WWTPs were further classified in six typologies by their quality requirements according to their final discharge point or water reuse. Moreover, two
Keywords:
different functional units (FU), one based on volume (m3) and the other on eutrophication
Eutrophication
reduction (kg PO3 4 removed) were used to further determine sustainability. A correlation
Functional unit
between legal requirements and technologies used to achieve them was found: Organic
Global warming
matter removal plants were found to be less costly both in environmental and economic
Life cycle thinking
terms if volume was used as the functional unit, while more demanding typologies such as
Wastewater treatment costs
reuse plants showed a trade-off between lower EP and higher cost and GWP; however, this is overcome if the second FU is used instead, proving the sustainability of these options and that this FU better reflects the objectives of a WWTP. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
Over the past fifty years, public attitude toward the environment has changed. Adapting itself to the demands of an evolving society, engineering has added sustainability to its general objectives (Davidson et al., 2007). As presented in the report of the United Nations 2005 World Summit, the pillars for sustainability are environmental protection as well as economic and social development (United Nations General Assembly, 2005). This has produced a substantial change in how technology is designed and operated. In this sense, the application of sustainability criteria for the provision of goods
and services is now the main focus, rather than environmental protection based on an end-of-pipe approach (Davidson et al., 2007). Wastewater treatment, an end-of-pipe technology, must comply with environmental, social and economic requirements in order to be considered sustainable (Balkema et al., 2002). The aim of this research paper is to environmentally and economically assess the operation of 24 wastewater treatment plants (WWTPs), classified according to six different typologies. The criteria for comparison among the WWTPs include the selection of the most appropriate functional unit of the system, which assures that the conclusions
* Corresponding author. Tel.: þ34 881 816 020; fax: þ34 881 816 015. E-mail addresses: [email protected] (G. Rodriguez-Garcia), [email protected] (M. Molinos-Senante), almudena. [email protected] (A. Hospido), [email protected] (F. Herna´ndez-Sancho), [email protected] (M.T. Moreira), [email protected] (G. Feijoo). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.08.053
5998
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 9 7 e6 0 1 0
derived from the analysis are consistent. In this sense, two functional units were considered and a sensitivity analysis was carried out.
1.1.
Environmental sustainability and WWTPs
In the last two decades a number of methodologies have been developed for evaluating the environmental sustainability of a product or process. Among them, Life Cycle Assessment (LCA) is a well-established procedure quantifying inputs and outputs as well as the potential environmental impacts associated with a product throughout its whole life cycle (Bauman and Tillman, 2004; ISO, 2006; Finnveden et al., 2009). LCA has been applied to water treatment systems (water treatment plants, sewer systems, and WWTPs) from the earliest stages of the development of the methodology (a review of a majority of published papers on LCA of WWTPs can be found in Larsen et al. (2007)). More recently, literature has been focused on pharmaceutical and personal care products (PPCPs; Wenzel et al., 2008; Mun˜oz et al., 2008a), nutrient removal (Foley, 2009), tertiary treatments (Mun˜oz et al., 2009; Høibye et al., 2008; Larsen et al., 2010), sludge treatment and disposal (Lundin et al., 2004; Murray et al., 2008; Hospido et al., 2010), or global warming impact associated with wastewater treatment (Stokes and Horvath, 2010). Even when the removal of organic matter (OM) and nutrients is a key objective in the operation of a WWTP, the eutrophication produced by the treated effluent is the main environmental impact made by most plants (Roeleveld et al., 1997; Hospido et al., 2004). The magnitude of the other impact categories varies. While WWTPs with tertiary or advanced treatments seem to significantly impact on global warming and acidification (Beavis and Lundie, 2003; Clauson Kaas et al., 2006), the toxicity-related impacts, caused by the presence of heavy metals and PPCPs, both in water (Roeleveld et al., 1997; Larsen et al., 2010) and in the sludge applied to land (Hospido et al., 2005, 2010), are present in all plants. The significance of other impact categories, such as ozone layer depletion and photochemical oxidants formation, has been found negligible in most cases.
1.2.
Economical sustainability and WWTP
The established regulation mechanisms for environmental protection may not be enough to assure the objectives for environmental quality and efficient natural resource use (Zhang and Wen, 2008). A number of authors argue that economic instruments are also very important for the implementation of policies and selection of measures to assist in environmental protection; the economy provides tools, information and instruments for streamlining the decision-making process (Ashley, 2009; Wissel and Wa¨tzold, 2010; Hepburn, 2010). An example of this growing interest is the new role of economic analysis in the Water Framework Directive (WFD, EU, 2000). This directive represents a new advance in water resource planning by integrating a number of economic principles into the management of water in EU member states (Herna´ndez-Sancho et al., 2010). These principles include polluter pays, as well as additional approaches such as the
analysis of cost-effectiveness of pollution mitigation measures and the consideration of economic instruments such as water pricing (Moran and Dann, 2008). Despite the significance of economic analysis in the field of wastewater treatment processes, it has received less attention than environmentally or technologically focused assessments. Available, however, in the field of water reuse are studies with detailed cost analysis, useful in the assessment of different treatment schemes (Herna´ndez et al., 2006; Asano et al., 2007). Other researchers use cost modeling for a better understanding of the cost structure as well as for planning new investments. Several studies (Gonzalez-Serrano et al., 2005; HernandezSancho et al., 2011) have validated and defined cost functions for different wastewater treatment technologies. These studies have considered the correlation of investment, operation, and maintenance costs of the WWTPs with some representative variables. Taking into account the WFD requirements, especially those related to the cost recovery for water services, several cost benefit analyses (CBA) (Godfrey et al., 2009; Chen and Wang, 2009; Molinos-Senante et al., 2010) have been carried out with the aim of identifying cases in which the adoption of measures to achieve a good ecological status for water bodies was required. All the benefits and costs including those which qualify as “nonmarket” must be integrated into the CBA. In recent years, LCA of WWTPs has been combined with different economic indicators. Mun˜oz et al., (2008b) developed an Environmental Economic Score (EES) and applied it to different advanced oxidation technologies. In Larsen et al. (2010) the cost-efficiency of ozonation and pulverized carbon addition is calculated both as stand-alone processes or combined with sand filtration. Another group (Nogueira et al., 2007) compared the technology of activated sludge (AS) with constructed wetlands and slow rate infiltration for small communities. Lin (2009) combined one LCA study with an inputeoutput model for the wastewater system of the metropolis of Tokyo, highlighting that the implementation of the secondary treatment turned into larger GHG emissions. The present study aims to combine LCA and economic assessment based on real process data.
2.
Materials and methods
2.1.
Objectives
The main objective of this paper is to compare the environmental and economic performance of 24 Spanish WWTPs. After a preliminary analysis, the criteria for the classification of the WWTPs in different typologies was based on the quality requirements set in the European urban wastewater directive (EEC, 1991) and Spanish legislation concerning water reuse (MP, 2007) according to the final destination of the treated wastewater (Table S1, Supplementary information). Plants that discharge the treated wastewater to non-sensitive and sensitive areas (as defined by the European Directive) Type 1: Plants designed and operated for the removal of organic matter discharging the treated wastewater to nonsensitive areas.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 9 7 e6 0 1 0
Type 2: Plants designed and operated for the removal of organic matter and nutrients discharging the treated wastewater to non-sensitive areas. Type 3: Plants designed and operated for the removal of organic matter and nutrients discharging the treated wastewater to sensitive areas. Plants that reuse the treated wastewater Type 4: Plants reusing the treated wastewater for irrigation in agricultural land. Type 5: Plants reusing the treated wastewater for industrial purposes. Type 6: Plants reusing the treated wastewater for aquifer recharge. A streamlined methodology based on environmental and economic indicators is presented. Special attention has been given to the technological differences among the six typologies. To a lesser extent, the effect of influent quality, very different due to climatic conditions, such as rainfall, and water use management, has been assessed.
2.2.
Case study
A brief description of the different WWTPs studied can be found in Table 1. The plants under study were selected according to the completeness of the data offered. A representative sample of 24 plants designed for populations larger than 50,000 inhabitants was selected from two different areas: WWTP 1e5 from an Atlantic region, Galicia (NW Spain), with an average rainfall of 1289 mm/year and WWTP 6e24 from a Mediterranean region, Valencia (E Spain), with an average rainfall of 405 mm/year, as described by Rodriguez-Garcia et al. (2011). This group represents around two-thirds of the total number of large plants in the regions of interest.
2.3.
Functional unit
The definition of the functional unit (FU) comprises a physical measurement of the function provided by the system, usually expressed as a certain amount of product (i.e. 1 kg of detergent) or as the service provided by it (i.e. washing of 100 shirts). The potential impacts that can be associated with the product or service will then be totalized and further referred to the FU. When defining the FU of a WWTP, different choices are possible. On one hand, there are studies that considered the volume (m3) of treated water for a certain period of time as the FU (Suh and Rousseaux, 2001; Hospido et al., 2004). Conversely, there are those based on the environmental load associated with a one person equivalent (PE) (Tillman et al., 1998; Hospido et al., 2007). The former has the advantage of being based on physical data while the latter tends to be used for comparative purposes, since it minimizes the differences associated with the influent composition and flow. Here, a functional unit based on the volume of treated wastewater (m3) was used as the first choice for the comparison of the WWTPs (from now, FU1). This approach may give insight into the effect of differences between facilities, both in regard to the influent flow and quality, as well as the removal
5999
yields. However, speaking in a stricter sense, the function of a WWTP is to meet the restrictions imposed by the existing regulations in terms of pollutant concentrations in the discharge of the treated effluent. The existing framework legislation developed in EEC (1991) and EC (1998) directives only requires the reduction of nitrogen and phosphorous for the treated effluents returned to a sensitive area (Table S1). In fact, three of the plants addressed in this study return the treated water to a sensitive body (Type 3) and a fourth one, WWTP 24 must fulfill similar nitrogen requirements for discharging to an aquifer (Type 6, Table S1). Therefore, another choice for the definition of the FU should comprise the removal of both nutrients and organic matter, expressed in eq. removed (from now FU2).1 For that terms of kg PO3 4 reason, the use of this second FU will highlight the differences between the environmental and economic costs of reducing the potential eutrophication associated with the effluent for all the WWTPs.
2.4.
System boundaries
The selection of the system boundaries in LCA studies on wastewater treatments was studied by Lundin et al. (2000). Although the sewer systems have proved to be environmentally relevant (Doka, 2009; Lassaux et al., 2006), the treatment plant remains the main impact contributor. All the plants under study are preceded by combined sewers systems, i.e. a system that jointly collects wastewater and rainwater resulting in the dilution of the influent in all WWTPs. For this reason and for the lack of data concerning the sewer systems, they were not included in this study. However, the effect of dilution will be more relevant in the Atlantic WWTPs due to climatic reasons (Rodriguez-Garcia et al., 2011). Regarding the treatment plant, although the construction stage has been found to be responsible for 25e35% of the GWP associated with a WWTP (Tangsubkul et al., 2005; Doka, 2009), the operation of the facility is considered far more relevant for the rest of the categories (Lundie et al., 2004; Lassaux et al., 2006). On the other hand, the impact demolishing/disposal stage has been found to be negligible (Corominas et al., 2011). Therefore, this assessment considered the environmental impact associated with the operation of primary, secondary, and tertiary treatments (when present); final discharge of the treated effluent; as well as the sludge treatment and its final disposal. The latter is mainly as fertilizer in agricultural soil in the case of most plants, except WWTP 3 where an important fraction of the sludge produced is landfilled. Some plants reuse a fraction of the treated effluent for agriculture irrigation (Type 4: WWTP 17 to 22). Despite the unavailability of specific data for the fraction of reused water for each plant, most WWTPs are located in the Jucar watershed, and since this watershed displays the highest rate of reused wastewater in Spain (MARM, 2008) it was assumed that 1 According to the CML environmental impact assessment method (Guine´e et al., 2002), all substances that could potentially cause eutrophication are related to the reference or equivalence substance (PO3 4 eq) in order to establish the potential impact of a product/process referred to this impact category.
6000
Table 1 e Wastewater treatment plants under study. Size
Secondary treatment
m3/day m3/day pe (r) Activated AS þ N AS þ P AS þ N and P Extended EA þ N and P Oxidation treated treated sludge removal removal removal aireation removal ditch (d) (r) (AS) (EA)
(d) design (r) real.
26,480 54,560 24,640 8080 12,000 40,000 20,664 24,000 60,000 45,000 32,000 8400 18,000 25,000 22,486 8000 30,000 60,000 24,000 60,000 60,000 38,000 15,000 42,000
53,935 51,111 45,227 6300 14,722 38,634 13,681 20,825 37,735 42,029 12,707 10,699 7359 21,290 10,735 7945 28,870 35,613 14,048 30,584 17,676 23,695 12,517 8474
125,452 130,929 191,762 40,770 49,393 193,046 66,787 127,271 243,144 264,744 62,340 83,890 58,693 54,162 60,752 65,422 117,816 229,154 149,575 200,908 141,609 213,676 60,701 66,025
✔ ✔ e e ✔ ✔ ✔ e e e e e e e e e e e ✔ e e e e e
e e e e e e e e e e e ✔ ✔ e e e ✔ e e e e ✔ e e
e e e e e e e ✔ ✔ e e e e e e e e e e ✔ e e e e
e e e e e e e e e ✔ ✔ e e ✔ ✔ ✔ e ✔ e e ✔ e e
e e e ✔ e e e e e e e e e e e e e e e e e e e e
e e e e e e e e e e e e e e e e e e e e e e ✔ ✔
e e ✔ e e e e e e e e e e e e e e e e e e e e e
e e e ✔ e ✔ e e ✔ e ✔ e e ✔ ✔ ✔ ✔ ✔ ✔ ✔ ✔ ✔ ✔ ✔
Removal efficiency COD NT
83% 89% 93% 96% 93% 96% 94% 92% 93% 94% 96% 97% 96% 96% 95% 97% 92% 94% 94% 95% 96% 93% 93% 97%
15% 27% 50% 85% 92% 70% 39% 52% 68% 47% 73% 89% 88% 84% 63% 92% 16% 66% 29% 47% 70% 54% 70% 95%
PT Average
39% 44% 37% 63% 60% 85% 95% 87% 83% 90% 74% 84% 83% 96% 82% 92% 65% 65% 76% 71% 74% 77% 89% 93%
46% 53% 60% 82% 81% 83% 76% 77% 82% 77% 81% 90% 89% 92% 80% 94% 58% 75% 66% 71% 80% 75% 84% 95%
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 9 7 e6 0 1 0
Discharge T1. non WWTP 1 -sensitive WWTP 2 OM removal WWTP 3 WWTP 4 WWTP 5 WWTP 6 WWTP 7 T2. non sen. WWTP 8 OM þ nut WWTP 9 Rem. WWTP 10 WWTP 11 WWTP 12 WWTP 13 T3. sen. area WWTP 14 WWTP 15 WWTP 16 Reuse T4. WWTP 17 agricultural WWTP 18 WWTP 19 WWTP 20 WWTP 21 WWTP 22 T5. Ind. WWTP 23 T6. Aq. WWTP 24
Tertiary treatment
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 9 7 e6 0 1 0
all the treated effluent is reused. The presence of nutrients in reused wastewater also prevents the production of fertilizers in the same way as sludge does (see below). Since in this case wastewater reuse does not displace any marginal technology due to the high cost of desalination, no other avoided products have been included.
2.5.
Life cycle inventory
Inventory data presented in Table 2 is the annual averages for year 2008 except for WWTP 1 (average data corresponds to the period 1998e2003) and WWTP 4 (average date corresponds to 2009). Regarding the sources, data for WWTP 1 was obtained from a previous study (Hospido et al., 2007); data for WWTPs 2e5 was provided by the company in charge of their management and operation; and data for WWTPs 6e24 is from the regional wastewater treatment authority. In addition, background data was obtained from SimaPro databases as follows: - Medium voltage electricity (Dones et al., 2007): The process selected includes electricity production and import/export (data from 2004), transformation from high voltage, direct SF6 emissions to air and electricity losses from medium voltage transmission system. Electricity production and import/export data were updated for 2008 based in MITYC (2009), ONE (2008) and REE (2009). - Chemical products (Althaus et al., 2007): Acrylonitrile manufacture was used for the production of polyelectrolyte (polyacrilamide). The remaining chemicals were directly selected from the Ecoinvent database (Iron III 40%, sodium hypochlorite 15%, sodium hydroxide 50%, liquid sulfuric acid and quicklime, milled, loose and phosphoric acid 85%). - Transport (Spielmann et al., 2007): Trucks 7.5e16 t EURO 3 (2000) were selected as standard transport for chemicals, waste and sludge. The vehicles under this regulation represented 34% of total Spanish trucks in 2008 (DGT, 2009). - Waste (Doka, 2009). Waste from the primary treatment was regarded as municipal waste and treated in an incineration plant with energy recovery. Typically, grit is disposed at an inert landfill and fats are stabilized with lime and cement before disposal in a secure deposit. For those plants where sludge is landfilled (3, 9 and 11), this output was modeled as municipal waste. - Fertilizers avoided (Nemecek and Ka¨gi, 2007): After a sensitivity analysis of 10 N-based and 6 P-based fertilizers from the Ecoinvent database, ammonium sulphate and diammonium phosphate were selected as generic N and P2O5 sources, and a substitutability of 50% and 70% was assumed for the N and P present either in the sludge or in the effluent (in Type 4 plants), respectively (Bengtsson et al., 1997). - Direct emissions from sludge disposal. N2O emissions from sludge application were calculated according to Hobson (2003) and NH3 emissions as defined by Lundin et al. leakage was considered according to Doka (2000). PO3 4 (2009). - CO2 emissions from the water line as well as those derived from the biogas combustion from the anaerobic digestion of sludge were not taken into account since it was considered to be biogenic according to IPCC guidelines (Doorn et al., 2006).
2.6.
6001
Environmental sustainability indicators
As previously stated in section 1.1., the eutrophication potential (EP) has been considered the most relevant impact category in the majority of published LCAs on WWTPs (Larsen et al., 2007). For this reason, eutrophication was selected as the main environmental sustainability indicator and quantified by means of the CML method, which converts all eutrophying substances (Table S1 Supplementary information) to phosphate equivalent (Guine´e et al., 2002). According to Larsen et al. (2007), the global warming potential (GWP) is not among the most relevant impact categories for WWTPs. However, it is usually regarded as a significant environmental problem, at least from a political and social point of view, and it can also be indicative of other energy dependent impacts such as acidification. As a consequence, it was chosen as the second environmental indicator and quantified in accordance with the IPCC guidelines (Table S2, Supplementary information).
2.7.
Economic sustainability indicators
Operational costs, subdivided by categories (energy, staff and others), were chosen as economic indicators, due to their relationship with overall plant management, and presented for both FU1 and FU2. The data shown is the annual average value for 2008, except for WWTP 1 (for which no economic data was available) and WWTP 4 (whose data was from 2009).
3.
Results and discussion
Due to the large amount of data generated, results are presented in figures and tables grouped according to the 6 different typologies. However, similar information for all the individual plants can be found in the supplementary information.
3.1.
Typologies and inventory
Taking into account their legal objectives, both Type 1 and 2 plants constitute a homogenous group since they are only required to remove OM. The difference is, despite not legally being required to do so, Type 2 plants also remove nutrients (N or/and P). A reason for that might be that, on average, the Type 2 influents are noticeably more loaded than the Type 1 influents (Fig. 1, Table 2). The extent of nutrient removal is generally associated with their presence in relatively high concentrations, as occurs in WWTPs 8 and 9 in order to attain high levels of P removal (Tables 1 and 2). Another possible explanation might be that since the areas considered sensitive might vary through time, the managers of Type 2 plants built or upgraded during the last decade, and may have taken into account that their receiving body could be considered sensitive in the future; thus nutrient removal could become a legal requirement. This would also explain why some Type 2 plants have lower removal efficiency than the Type 3 plants, which discharge in sensitive areas and therefore nutrient removal must be accomplished (Table 1).
6002
Table 2 e Life cycle inventory of WWTP, all data is presented for FU1 (m3). Influent
Electricity
Chemicals consumption
COD (g)
NT (g)
PT (g)
COD (g)
NT (g)
PT (g)
From the grid (kWh)
To the grid (kWh)a
Polyelectrolyte (g)
FeCl3 (g)
CaO (g)
NaClO (g)
NaOH (g)
H2SO4 (g)
Transport (kg km)
327 340 422 220 685 648 585 623 763 673 609 918 787 419 599 846 617 623 1107 763 752 886 750 954
19.45 19.84 22.37 47.28 53.36 58.94 54.98 47.61 54.98 59.42 64.15 65.68 73.86 30.91 57.15 59.41 41.59 55.30 70.65 71.98 75.62 76.10 30.20 65.94
0.70 1.22 4.12 6.70 5.06 8.76 10.65 6.98 8.62 9.55 10.53 8.77 10.21 4.49 7.14 8.83 6.37 7.52 6.78 9.86 9.57 12.84 5.78 15.73
55 38 29 8 51 27 37 50 53 43 26 31 31 17 32 23 49 39 67 39 28 65 49 24
16.45 14.53 11.20 7.08 4.23 17.95 33.48 22.94 17.53 31.37 17.22 7.02 8.59 5.03 20.90 5.01 34.98 18.81 50.19 38.00 22.32 34.74 9.16 3.43
0.43 0.68 2.60 2.46 2.04 1.34 0.54 0.91 1.43 1.00 2.69 1.44 1.70 0.20 1.28 0.73 2.24 2.63 1.66 2.90 2.49 2.99 0.65 1.06
0.13 0.14 0.20 0.54 0.29 0.27 0.33 0.13 0.48 0.36 0.63 0.52 0.85 0.31 0.51 0.59 0.13 0.69 0.56 0.56 1.37 0.66 0.50 0.80
e e e e e 0.10 0.24 e 0.04 e 0.06 e e e e e 0.13 0.07 e e e 0.27
0.27 e 0.41 3.91 0.54 1.79 2.09 2.23 1.67 2.86 2.41 2.16 2.70 0.88 3.26 6.67 2.67 1.59 3.65 5.39 3.03 3.84 0.00 11.37
e 22.33 e e e 65.25 e e 54.48 21.75 0.74 22.54 e e 20.54 21.10 53.30 44.37 119.95 81.90 e 20.23 89.93 47.81
e 51.38 e e e e e e e e e e e e 80.63 e e e e e e e e e
e e e e 0.00 5.19 e e 2.92 5.53 e 2.58 1.16 2.86 2.93 e 2.24 23.19 6.77 21.92 3.57 4.29 e 17.93
e e e e e 0.97 e e 0.13 0.89 e 1.00 1.20 3.61 0.81 e 1.50 0.12 0.15 6.74 7.44 1.34 e 3.40
e e e e e 0.17 e e 0.03 e e 0.00 1.08 e 0.56 e 1.25 e 0.29 e 0.08b 0.26 e 0.70
0.01 1.47 0.01 0.08 0.01 1.47 0.04 0.04 1.18 0.62 0.06 0.57 0.12 0.15 2.17 0.56 1.22 1.39 2.62 2.32 0.28 0.60 1.80 1.62
e
Sludge
Waste
Operational costs (V)
To agriculture To landfill Transport Application N (g)d P2O5 (g)d N2O (g) NH3 (g) PO4 (g) Grit (g) MSW (g) Fats (g) Transport Energy (kg WW)c (kg km) (kg WW)c (kg km) as slurry (l) to air to air to water WWTP WWTP WWTP WWTP WWTP WWTP WWTP WWTP WWTP WWTP WWTP WWTP WWTP WWTP
1 2 3 4 5 6 7 8 9 10 11 12 13 14
0.25 0.34 0.17 0.87 0.70 0.54 0.72 0.78 1.06 0.76 0.90 1.44 1.75 0.53
e e 0.22 e e e e e 0.05 e 0.01 e e e
5.13 6.82 7.88 43.49 13.95 10.99 14.70 16.00 21.76 9.18 18.41 29.58 35.80 6.56
0.24 0.27 0.17 0.86 0.67 0.53 0.71 0.77 1.05 0.76 0.89 1.43 1.73 0.52
2.19 4.63 0.23 12.48 4.99 1.73 8.57 9.70 10.18 3.13 e 16.73 3.25 5.41
2.87 4.88 0.13 0.02 1.00 5.46 12.03 7.40 12.35 8.87 e 15.89 0.78 5.33
0.03 0.07 0.00 0.23 0.08 0.03 0.13 0.15 0.16 0.05 e 0.26 0.05 0.09
0.66 1.41 0.07 3.79 1.52 0.53 2.60 2.94 3.09 0.95 e 5.08 0.99 1.64
0.10 0.17 0.00 0.00 0.03 0.19 0.41 0.25 0.43 0.31 e 0.55 0.03 0.18
e 10.13 19.28 32.62 3.70 10.12 3.70 4.78 17.34 25.69 25.43 0.00 55.78 10.14
54.70 12.17 13.18 31.31 3.63 16.67 56.31 17.83 22.00 115.45 20.40 25.53 27.48 5.76
e 1.45 4.24 15.66 1.67 0.85 0.70 1.05 6.55 9.95 0.97 e 1.23 e
6.02 1.94 4.80 15.04 1.69 0.55 1.23 0.46 0.81 1.69 0.94 0.52 1.71 0.20
e 0.013 0.027 0.039 0.028 0.041 0.084 0.069 0.032 0.054 0.078 0.072 0.061 0.042
Staff
Others
e 0.024 0.033 0.053 0.015 0.060 0.108 0.074 0.032 0.082 0.063 0.065 0.153 0.103
e 0.032 0.018 0.025 0.048 0.077 0.031 0.068 0.100 0.100 0.068 0.075 0.090 0.060
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 9 7 e6 0 1 0
WWTP 1 WWTP 2 WWTP 3 WWTP 4 WWTP 5 WWTP 6 WWTP 7 WWTP 8 WWTP 9 WWTP 10 WWTP 11 WWTP 12 WWTP 13 WWTP 14 WWTP 15 WWTP 16 WWTP 17 WWTP 18 WWTP 19 WWTP 20 WWTP 21 WWTP 22 WWTP 23 WWTP 24
Effluent
Some plants produce a certain amount of electricity by combustion of the CH4 produced by the anaerobic digestion of their sludge that is sold to the net rather than used inside the WWTP. WWTP 21 consumes H3PO4 rather than H2SO4. WW: Wet weight. Nutrients values presented reflect their total amount present in the sludge, not the amount that is used by plants according to Bengtsson et al. (1997). a b c d
35.44 24.11 29.95
WWTP WWTP WWTP WWTP WWTP WWTP WWTP WWTP WWTP WWTP
15 16 17 18 19 20 21 22 23 24
1.27 1.72 0.61 0.93 2.10 1.07 1.89 1.73 1.18 1.30
e e e e e e e e e e
26.09 35.30 19.00 18.98 42.99 21.84
1.02 1.65 0.61 0.91 1.69 1.05 1.87 1.71 1.16 1.25
13.22 3.46 6.93 9.28 4.61 11.70 3.84 e 10.10 2.84
12.45 0.07 8.00 11.67 0.09 19.08 0.85 e 9.66 0.11
0.21 0.05 0.11 0.15 0.07 0.18 0.06 e 0.16 0.04
4.01 1.05 2.10 2.82 1.40 3.55 1.17 e 3.06 0.86
0.43 0.00 0.28 0.40 0.00 0.66 0.03 e 0.33 0.00
19.81 22.92 14.70 5.26 13.91 4.92 100.44 20.35 13.75 16.38
22.83 28.90 10.31 23.70 20.27 56.77 67.98 146.84 0.88 35.08
0.35 1.82 0.36 2.04 e 0.80 8.58 0.90 1.29 3.84
0.87 1.06 0.78 0.59 0.70 1.26 4.04 3.43 0.30 1.18
0.055 0.107 0.052 0.045 0.038 0.059 0.101 0.064 0.058 0.125
0.058 0.061 0.078 0.081 0.105 0.082 0.137 0.094 0.133 0.103
0.190 0.115 0.107 0.050 0.122 0.135 0.185 0.131 0.069 0.083
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 9 7 e6 0 1 0
6003
Tertiary treatment, although not required, is present in some facilities included in Type 1 and Type 2. In the Type 1 group, UV disinfection is partially responsible for the higher electricity use of WWTP 4, much in the same way as coagulation/ flocculation for the consumption of FeCl3 in WWTP 6 (Table 2). In the Type 2 group, the effect of tertiary treatment is not so obvious: both WWTP 9 and WWTP 11 present coagulation/ flocculation stages as primary and tertiary treatment, but the consumption of FeCl3 for the latter is almost irrelevant (Table 2). Even more, WWTP 11 also has UV disinfection, but the effect on the energy use is not as evident for WWTP 11 compared to other Type 2 plants as it is for WWTP 4 compared to the rest of Type 1 plants (Table 2). The Type 3 plants do not present such a high influent load as the Type 2 ones; nevertheless they are probably large enough to require nutrient removal in order to fulfill their correspondent legal requirements found in EC (1998) (Table 1, Fig. 1 and Table S1). Although tertiary treatment, and specifically disinfection, is not required to discharge in a sensitive area, this stage is present in all plants. In any case, its effect is not noticeable since all have a chlorination process, although no consumption of chlorinating agents was reported for 2008. This is because the tertiary treatment is not in use at present, and was built in case that reuse of the treated water was required instead of being discharged. All Type 4 plants present tertiary treatments since they require some kind of disinfection process in order to fulfill sanitary requirements. UV disinfection, present in all except WWTP 18 and 19, might be a reason for higher electricity use compared with Type 1. WWTP 18 presents ultrafiltration and reverse osmosis, which might be less energy efficient than UV (Beavis and Lundie, 2003; Clauson Kaas et al., 2006). On the other hand, WWTP 19, with an average consumption (0.56 kWh/m3) uses a filtration process, using electricity at a similar rate to the UV process (Clauson Kaas et al., 2006). WWTP 21 uses by far the most electricity, most likely because of the use of aerobic sludge digestion rather than anaerobic or lime stabilization like the other WWTPs. All Type 4 plants except one utilize some nutrient removal technology, which might seem contradictory since, as indicated in Section 2.4, nutrients present in the reused water partially avoid the use of N and P2O5 -based fertilizers. However, all plants were built and upgraded years before the legislation for water reuse was published, suggesting, as in the case of Type 2 plants, that they can target a specific nutrient or remove both in case the effluent has to be discharged to a sensitive area. Since the data is from 2008, the first year with water reuse regulations, it is likely the high nutrient removal efficiencies correspond to a transition process between water discharge and water reuse (Table 1). Required to supply part of the treated water for industrial purposes, WWTP 23 (Type 5) must fulfill similar requirements to Type 4 plants and thus uses similar technology. Still, their electricity use is not particularly high (Table 2), especially considering this plant makes use of filtration, ultrafiltration, and UV stages, indicating the fraction of water reused in industry might be small. WWTP 24 (Type 6) shows two important characteristics which will notably affect their environmental profile: on the one hand, this plant deals with the highest average load (Fig. 1,
6004
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 9 7 e6 0 1 0
Fig. 1 e Eutrophication Potential associated to the influent (per m3 of influent, FU1).
Table 2); but on the other, it is the facility with the highest requirements, since it not only must achieve a substantial nutrient removal (Tables 1 and 2) but also disinfect the treated waterdwater recharging an aquifer. This explains its high energy use (Table 2), associated not only with the extended aeration process, but also with the UV disinfection and the filtration of the tertiary treatment.
3.2. Environmental profile according to the volumerelated functional unit (FU1) 3.2.1.
Eutrophication potential calculated for FU1
Average results of the EP based on the FU1 (m3), grouped by typology, are presented in Fig. 2, while individual data for all the plants is detailed in Figs. S1 and S2 in the supplementary information. In addition to the impact associated with the effluent of the plants, the eutrophication associated with the treatment process, as well as the beneficial consequences of the different avoided products, are also included here. This indirect eutrophication is never higher than 10% of the effluent EP and is approximately 4% for most plants. Avoided products show a small contribution to the whole picture, although they are not insignificant for Type 4. Despite the obvious differences in the influent composition (Fig. 1) of types 1 and 2, the effluents of both groups present
a relatively similar effluent quality (Fig. 2) since the WWTPs operate to accomplish identical legal requirements (maximum concentration of 125 mg COD/m3 in the discharged effluent). As seen in Table 1 and already discussed in Section 3.1, Type 2 plants have implemented technologies for nutrient removal for their highly loaded wastewater, which probably justifies the large difference between both Figs. 1 and 2 regarding Type 1 and Type 2. For Type 3, they are legally required to remove both N and P, which explains their high efficiencies (Table 1). However, their lower EP is not only due to their efficiency, but also because of their lower influent loads (Fig. 1). The Type 4 WWTPs present a noticeably high impact (Fig. 2). Despite most of them using some kind of nutrient removal process, their removal efficiency is relatively low compared to that of Type 3, the reason being nutrients present in the water are a valuable resource for agriculture and there would be no point in totally eliminating them. It is worthwhile to note the impact of the effluent is still high, even considering more than half of the nutrients are expected to be absorbed by plants (Bengtsson et al., 1997). Type 5 presents a behavior much like the Type 3 plants as both of them remove nutrients with distinctive efficiency (Table 1) and the influent loads are not particularly different (Fig. 1. and Table 2.). Finally, although Type 6 presents the
Fig. 2 e Eutrophication Potential of the WWTP typologies considered based on FU1 (kg PO4 eq./m3).
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 9 7 e6 0 1 0
6005
Fig. 3 e Global Warming Potential of the WWTP typologies considered based on FU1 (kg CO2 eq./m3).
highest influent load of all WWTPs, its extraordinary high removal yields (Table 1) make this plant one of the facilities with the lowest effluent loads.
3.2.2.
and 5 present a similar emission rate to previous groups, although greater emissions for Type 4 plants are caused by WWTP 21, which displays by far the peak emission rate (Fig. S3) due to its high electricity use.
Global warming potential calculated for FU1
Average values of GWP impacts for the different typologies are presented in Fig. 3, while individual results are included in the Supplementary Information (Fig. S3). Not surprisingly, the increasing impacts in GWP are due to the increasing complexity of technology applied, and it is mainly associated with larger consumptions of electricity and chemicals (Table 2). Type 1 plants can be considered as a baseline scenario since they only fulfill the basic function of a WWTP: OM removal. The higher impact of the WWTPs belonging to Types 2, 3 and 4 can be partially attributed to their higher average COD concentration in the influent (Table 2), which requires larger aeration periods. Another factor would be the nitrification process (Table 1), which also demands more oxygen and thus, more electricity (Table 2). In some cases, it is also associated with tertiary treatments (Beavis and Lundie, 2003; Clauson Kaas et al., 2006), which are necessary to fulfill the requirements of the reuse water legislation but which are also employed by WWTPs discharging in sensitive areas. Types 4
3.3. Operational efficiency: environmental profile according to the eutrophication-related functional unit (FU2) 3.3.1.
Eutrophication potential calculated for FU2
Based on the FU2 (kg of PO3 4 eq. removed), average results are presented in Fig. 4, while detailed information for individual facilities is found in the supplementary information (Fig. S4). On average, this approach establishes clear differences between simple and increasingly complex technologies, penalizing the scheme used for the removal of the organic matter only (Type 1). This might suggest there is a margin for improvement for Type 1 plants, particularly WWTP 1, 2 and 3, either by optimizing the OM removal or by including the removal of nutrients. However, it is also necessary to indicate that the treatment of low load water (a feature of Type 1) has particular difficulties such as low sludge decantability (Seijas et al., 2003), and high OM removal rates could not always be guaranteed.
Fig. 4 e Eutrophication Potential of the WWTP typologies considered based on FU2 (kg PO4 eq./kg PO4 eq. removed).
6006
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 9 7 e6 0 1 0
Fig. 5 e Global Warming Potential of the WWTP typologies considered based on FU2 (kg CO2 eq./kg PO4 eq. removed).
According to the results based on FU2, the implementation of combined treatment for organic matter and nutrient removal clearly benefits the environmental performance of non-Type 1 plants, especially in the case of Type 6, which has the highest removal efficiency (Table 1). FU2 also shows, despite being less efficient in absolute terms (Table 1, Fig. 2), Type 2 presents a profile relatively similar to Type 3, reinforcing the idea that even when Type 2 plants do not need to remove nutrients, they put the same effort into doing so as plants legally required to do it (Type 3). It also balances the higher values presented in Fig. 1 for Type 4 due to its high removal in absolute terms.
3.3.2.
Global warming potential calculated for FU2
The contributions of the different wastewater treatment typologies to global warming expressed by the avoided eutrophication (FU2) are presented in Fig. 5 (individual results in Fig. S5). The differences between Types 1 and 2 are also evident here, further emphasizing that the former are on average less efficient, requiring more electricity for the same level of eutrophication reduction. To a lesser extent, this is also the case for Type 5, which presents a relatively higher profile than in Fig. 3 Type 3 and 4 share a similar profile despite the higher electricity use of the latter (Table 2, Fig. 3) due to the higher absolute removal by Type 3 plants. Aquifer recharge
(Type 6) is also revealed as a fairly good environmental option (Fig. 5), even in spite of extensive use of electricity (Table 2) and GWP emissions (Fig. 3).
3.4.
Economic profile
3.4.1. Economic profile according to the volume-related functional unit (FU1) Operational costs (OC) per FU1 (m3) are presented in Fig. 6 (as well as Fig. S6). As shown in Fig. 6, the operational costs of the six WWTP typologies are highly variable, since the minimum value is 0.127 V/m3 for Type 1 WWTPs, while the maximum is 0.311 V/ m3 for Type 6. For plants that discharge the treated wastewater to non-sensitive areas, Type 2 increase in cost of 75.6% compared Type 1 due to their nutrient removal. In regards to the two typologies of WWTPs that remove organic matter and nutrients (Types 2 and 3), the cost difference is quantified by 18%. This is because plants discharging regenerated water to sensitive areas display higher removal efficiencies for both N and P (Table 1). In relation to the three types of WWTPs that reuse the treated water, cost differences between them are very small, although the Type 6 plant presents slightly higher costs due to two factors: first, as shown in Table 1, the pollutants removal efficiency is higher than for the other
Fig. 6 e Operational costs of the WWTP typologies considered based on FU1 (V/m3).
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 9 7 e6 0 1 0
6007
Fig. 7 e Operational costs of the WWTP typologies considered based on FU2 (V/kg PO4 eq. removed).
WWTPs; second, the real wastewater flow treated by this plant (Table 1) is the lowest and therefore, is less affected by scale economies. Regarding the cost distribution: on average 26% of the total cost is associated with energy consumption, 35% with the staff, and the remaining 39% with others. That about a quarter of the total cost is linked to electricity highlights the importance of efficiency in the use of energy, both from an environmental and economic point of view. Fig. 6 shows that Types 4 and 5 are those spending the smallest percentage of total cost on the item energy. In contrast, WWTP 24 (Type 6) is the one with the highest dependence on electricity. In looking at electricity use data, expressed in kWh/m3 (Table 2), it is verified there is no direct relationship between electricity use and costs. This is because currently WWTP operators can negotiate their fee with the electric companies. Therefore, to reduce energy costs in WWTPs, a double strategy may be adopted: operators may negotiate lower electricity fees, while they may also reduce the use of electricity by increasing their efficiency. The second strategy is more beneficial since it would effect not only a reduction in costs but also reduce the carbon footprint of these facilities, but both may be implemented. In relation to staff costs, they are similar to those presented by MolinosSenante et al. (2010), who quantified them at 32% of the total operation costs. It is worth noting the percentage of the Type 5 WWTP in which staff costs represent a significantly higher percentage than the average (51% vs. 35%). That regenerated water is used for industrial purposes supposes a high monitoring of effluent parameters and ergo, high staff costs. Finally, the item “Others” is the least value-consistent parameter among WWTP types. This item consists of multiple elements (i.e. reagents, waste management and maintenance) and thus large variations would be expected.
3.4.2. Economic profile according to the eutrophication-related functional unit (FU2) Economic efficiency results based on FU2 are presented in Fig. 7, while individual results are displayed in Fig. S7. It is noted that the differences among the WWTP’s types are even more significant when using the FU2 than FU1. Type 1 and 2 experienced slightly lower costs than those discharged to sensitive areas. In the case of the non-sensitive areas, unlike
FU1 results, Type 2 plants show lower costs than Type 1. Thus, when using FU2 as a basis for comparison, it is revealed that nutrient removal is not necessarily more expensive than OM removal. Also, it is noted that plants discharging treated water to sensitive areas and those reusing the regenerated water for agricultural purposes have very similar operating costs. In regards to other plants, Type 5 and 6 are those experiencing the maximum and minimum costs, respectively. Nevertheless, it is necessary to state that these are typologies represented only by one plant and they may not be an accurate reflection of Industrial Reuse and Aquifer Recharge WWTPs.
4.
Conclusions
According to the results of this study, grouping WWTPs based on their legal requirements has exposed the link between these legalities and the technology used to achieve them. Non-Sensitive Discharge WWTPs tend to be associated with OM removal technology or with nutrient removal for specific problems and with relatively low efficiency. Sensitive Discharge and Environmental Reuse (aquifer refill) plants always present combined N and P removal as well as tertiary treatment. Agricultural and Industrial Reuse plants demand tertiary treatment due to microbiological requirements and although nutrients can be considered a valuable resource, several plants implement N and P removal, although with lower efficiency. The selection of the functional unit has proved to be a key aspect in defining both economic and environmental profiles. The first FU defined here (m3 of treated water) showed that Type 1 (non-sensitive discharge, OM removal only WWTPs), resulted in lower impacts for EP and GWP as well as lower costs, suggesting that the other typologies were less efficient. Conversely, the definition of a second FU based on EP reduction acknowledged the higher efficiency of Types 2 and 6 (nonsensitive areas discharge with OM and nutrient removal and aquifer recharge plants), resulting, on average, in a better environmental and economic performance. Although FU1 (m3) presents more intuitive results, FU2 (kg PO3 4 removed) has proved to better reflect the function of a WWTP when
6008
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 9 7 e6 0 1 0
focus is on eutrophication and thus is considered a more useful FU for comparative studies. All in all, results show obtaining an effluent of higher quality, meaning disinfected and with lower eutrophication potential, increases both with GWP and overall expense. It also revealed this higher cost is well-balanced, and is even beneficial for advanced typologies. Finally, for a wastewater treatment technology to be judged sustainable, it must comply with environmental, sociocultural and economic needs. Therefore, the on-going research is focused on incorporating social variables with the already-established approach in order to obtain a complete set of indicators of sustainability for each WWTP under consideration.
Acknowledgments The authors would like to thank the water management entities of the Generalitat Valenciana (EPSAR) and the Xunta de Galicia (Augas de Galicia-EPOSH) and the water management companies (Aquagest, IDOM and Geseco) that have supplied the data presented here. This study has been partially financed by the Spanish Ministry of Education and Science (Consolider Project-NOVEDAR) (CSD2007-00055), Xunta de Galicia (Project 09MDS010262PR) and Generalitat Valenciana (Project ACOMP/2010/138). M. Molinos and A. Hospido acknowledge the FPU program (AP2007-03483) and Isidro Parga Pondal program (IPP-06-57), respectively, for financial support.
Appendix. Supplementary material Supplementary data related to this article can be found online at doi:10.1016/j.watres.2011.08.053.
references
Althaus, H.J., Chudacoff, M., Hischier, R., Jungbluth, N., Osses, M., Primas, A., 2007. Life Cycle Inventories of Chemicals. EMPA Du¨bendorf, Swiss Centre for Life Cycle Inventories, Du¨bendorf (Switzerland). Asano, T., Burton, F., Leverenz, H., Tsuchihashi, R., Tchobanoglous, G., 2007. Water Reuse: Issues, Technologies, and Applications. McGraw-Hill Professional, New York. Ashley, R., 2009. Briefing: N. G. Semple’s ‘Principles of environmental protection’. Proceedings of the ICE-Engineering Sustainability 162 (1), 3e11. Balkema, A.J., Preisig, H.A., Otterpohl, R., Lambert, F.J.D., 2002. Indicators for the sustainability assessment of wastewater treatment systems. Urban Water 4 (2), 153e161. Bauman, H., Tillman, A., 2004. The Hitch Hiker’s Guide to LCA. Studentlitteratur AB, Lund, (Sweden). Beavis, P.C., Lundie, S., 2003. Integrated environmental assessment of tertiary and residual treatment-LCA in the wastewater industry. Water Science and Technology 47 (7e8), 109e116. Bengtsson, M., Lundin, M., Molander, S., 1997. Life Cycle Assessment of Wastewater Systems: Case Studies of Conventional Treatment, Urine Sorting and Liquid
Composting in Three Swedish Municipalities. Chalmers University of Technology. Chen, R., Wang, X.C., 2009. Cost-benefit evaluation of a decentralized water system for wastewater reuse and environmental protection. Water Science and Technology 59 (8), 1515. Clauson Kaas, J., Dahl, F., Dalgaard, O., Høibye, L., Kjølholt, J., Wenzel, H., et al., 2006. Viderega˚ende renseteknologier for kommunalt spildevand (Advanced Treatment Technologies for Municipal Wastewater). DANVA, Miljøstyrelsen, Lynettefællesskabet og Spildevandscenter. Corominas, L., Foley, J., Guest, J. S., Hospido, A., Larsen, H. F., and Shaw, A., 2011. Towards a Standard Method for Life Cycle Assessments (LCA) of Wastewater Treatment. In: Watermatex 2011, 8th IWA Symposium on System Analysis and Integrated Assessment. San Sebastian, Spain. Davidson, C.I., Matthews, H.S., Hendrickson, C.T., Bridges, M.W., Allenby, B.R., Crittenden, J.C., et al., 2007. Adding sustainability to the engineer’s toolbox: a challenge for engineering educators. Environmental Science and Technology 41 (14), 4847e4850. DGT, 2009. Anuario Estadistico General 2008 (General Statistics Yearbook 2008). Direccion General de Trafico. http://www.dgt. es/was6/portal/contenidos/documentos/seguridad_vial/ estadistica/parque_vehiculos/series_historicas_parque/ series_historicas_parque_2008.pdf (accessed 09.05.11). Doka, G., 2009. Wastewater Treatment. Life Cycle Inventories on Waste Treatment. Swiss Centre for Life Cycle Inventories, Du¨bendorf (Switzerland). Dones, R., Bauer, C., Burger, B., Emmenegger Faist, M., Frischknecht, R., Heck, T., et al., 2007. Life Cycle Inventories of Energy Systems. Paul Scherrer Institut Villigen, Swiss Centre for Life Cycle Inventories, Du¨bendorf (Switzerland). Doorn, M.R.J., Towprayoon, S., Manso-Vieira, S.M., Irving, W., Palmer, C., Pipatti, R., et al., 2006. Wastewater treatment and discharge. In: Eggleston, S., Buendia, L., Miwa, K., Tanabe, K. (Eds.), Waste. 2006 IPCC Guidelines for National Greenhouse Gas Inventories. IGES, Japan, pp. 6.1e6.28. EC, 1998. Directive 98/15/EC of 27 February 1998 mending council directive 91/271/EEC with respect to certain requirements established in annex I thereof. Official Journal of the European Communities L 67, 29e30. EEC, 1991. Directive 91/271/1991 EEC of 21 May 1991 concerning urban waste water treatment. Official Journal of the European Communities L 135, 40e52. EU, 2000. Directive 2000/60/EC of 23 October 2000 establishing a framework for community action in the field of water policy. Official Journal of the European Communities L 327, 1e72. Finnveden, G., Hauschild, M.Z., Ekvall, T., Guine´e, J., Heijungs, R., Hellweg, S., et al., 2009. Recent developments in life cycle assessment. Journal of Environmental Management 91 (1), 1e21. Foley, J., 2009. Life cycle assessment of wastewater treatment systems. PhD Thesis. University of Queensland, Australia. Forster, P., Ramaswamy, V., Artaxo, P., Berntsen, T., Betts, R., Fahey, D., et al., 2007. Changes in atmospheric constituents and in radiative forcing. In: Climate Change 2007: The Physical Science Basis. Cambridge University Press, Cambridge, UK and New York, NY, USA. Godfrey, S., Labhasetwar, P., Wate, S., 2009. Greywater reuse in residential schools in Madhya Pradesh, Indiaea case study of cost-benefit analysis. Resources, Conservation and Recycling 53 (5), 287e293. Gonzalez-Serrano, E., Rodriguez-Mirasol, J., Cordero, T., Koussis, A., Rodriguez, J., 2005. Cost of reclaimed municipal wastewater for applications in seasonally stressed semi-arid regions. Journal of Water Supply: Research and TechnologyAQUA 54 (6). http://www.iwaponline.com/jws/054/ jws0540355.htm (accessed 09.04.11.).
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 9 7 e6 0 1 0
Guine´e, J.B., Heijungs, R., Huppes, G., Keijn, R., de Koning, A., van Oers, L., et al., 2002. Handbook on Life Cycle Assessment: Operational Guide to the ISO Standards. Kluwer Academic Publishers, Dordrecht (The Netherlands). Heijungs, R., Guine´e, J.B., Huppes, G., Lankreijer, R., Udo de Haes, H., Sleeswijk, A., 1992. Environmental Life Cycle Assessment of Products. Centre of Environmental Science, Leiden (The Netherlands). Hepburn, C., 2010. Environmental policy, government, and the market. Oxford Review of Economic Policy 26 (2), 117e136. Hernandez-Sancho, F., Molinos-Senante, M., Sala-Garrido, R., 2011. Cost modelling for wastewater treatment processes. Desalination 268 (1e3), 1e5. Herna´ndez-Sancho, F., Molinos-Senante, M., Sala-Garrido, R., 2010. Economic valuation of environmental benefits from wastewater treatment processes: an empirical approach for Spain. The Science of the Total Environment 408 (4), 953e957. Herna´ndez, F., Urkiaga, A., De las Fuentes, L., Bis, B., Chiru, E., Balazs, B., et al., 2006. Feasibility studies for water reuse projects: an economical approach. Desalination 187 (1e3), 253e261. Hobson, J., 2003. CH4 and N2O Emissions from Waste Water Handling. International Panel on Climate Change. http://www. ipcc-nggip.iges.or.jp/public/gp/english (accessed 09.05.11.). Høibye, L., Clauson-Kaas, J., Wenzel, H., Larsen, H.F., Jacobsen, B. N., Dalgaard, O., 2008. Sustainability assessment of advanced wastewater treatment technologies. Water Science and Technology 58 (5), 963. Hospido, A., Carballa, M., Moreira, M., Omil, F., Lema, J.M., Feijoo, G., 2010. Environmental assessment of anaerobically digested sludge reuse in agriculture: potential impacts of emerging micropollutants. Water Research 44 (10), 3225e3233. Hospido, A., Moreira, M.T., Feijoo, G., 2007. A comparison of municipal wastewater treatment plants for big centres of population in Galicia (Spain). The International Journal of Life Cycle Assessment 13 (1), 57e64. Hospido, A., Moreira, M.T., Ferna´ndez-Couto, M., Feijoo, G., 2004. Environmental performance of a municipal wastewater treatment plant. The International Journal of Life Cycle Assessment 9 (4), 261e271. Hospido, A., Moreira, T., Martı´n, M., Rigola, M., Feijoo, G., 2005. Environmental evaluation of different treatment processes for sludge from urban wastewater treatments: anaerobic digestion versus thermal processes (10 pp). The International Journal of Life Cycle Assessment 10 (5), 336e345. ISO, 2006. ISO 14040:2006-Environmental Management e Life Cycle Assessment e Principles and Framework. http://www. iso.org/iso/catalogue_detail?csnumber¼37456. Larsen, H.F., Hauschild, M.Z., Wenzel, H., Almemark, M., 2007. Homogeneous LCA Methodology Agreed by NEPTUNE and INNOWATECH. Deliverable 4.1. EC Project “NEPTUNE”, contract No.:036845. www.eu-neptune.org. Larsen, H.F., Hansen, P.A., Boyer-Souchet, F., 2010. Decision Support Guideline Based on LCA and Cost/Efficiency Assessment. Deliverable 4.3. EC Project “NEPTUNE”, contract No.: 036845. www.eu-neptune.org. Lassaux, S., Renzoni, R., Germain, A., 2006. Life cycle assessment of water: from the pumping station to the wastewater treatment plant (9 pp). The International Journal of Life Cycle Assessment 12 (2), 118e126. Lin, C., 2009. Hybrid inputeoutput analysis of wastewater treatment and environmental impacts: a case study for the Tokyo Metropolis. Ecological Economics 68 (7), 2096e2105. Lundie, S., Peters, G.M., Beavis, P.C., 2004. Life cycle assessment for sustainable metropolitan water systems planning. Environmental Science and Technology 38 (13), 3465e3473. Lundin, M., Olofsson, M., Pettersson, G.J., Zetterlund, H., 2004. Environmental and economic assessment of sewage sludge
6009
handling options. Resources, Conservation and Recycling 41 (4), 255e278. Lundin, M., Bengtsson, M., Molander, S., 2000. Life cycle assessment of wastewater systems: influence of system boundaries and scale on calculated environmental loads. Environmental Science and Technology 34 (1), 180e186. MARM, 2008. Reutilizacion, Libro Digital del Agua (Water Reuse, Digital Water Book). Ministerio de Medio Ambiente, Rural y Marino. http://servicios2.marm.es/sia/visualizacion/lda/ recursos/noconvencionales_reutilizacion.jsp (accessed December 19.12.10.). MITYC, 2009. La Energı´a en Espan˜a 2008 (Energy in Spain 2008). Ministerio de Industria, Turismo y Comercio, Madrid (Spain). MP, 2007. Real decreto 1620/2007, de 7 de diciembre, por el que se establece el re´gimen jurı´dico de la reutilizacio´n de las aguas depuradas (Royal decree 1620/2007 of 7 December 2007 establishing the legal system for treated water reuse). BOE 294/2007, 50639e50661. Molinos-Senante, M., Herna´ndez-Sancho, F., Sala-Garrido, R., 2010. Economic feasibility study for wastewater treatment: a cost-benefit analysis. The Science of the Total Environment 408 (20), 4396e4402. Moran, D., Dann, S., 2008. The economic value of water use: implications for implementing the water framework directive in Scotland. Journal of Environmental Management 87 (3), 484e496. Mun˜oz, I., Jose´ Go´mez, M., Molina-Dı´az, A., Huijbregts, M.A., Ferna´ndez-Alba, A.R., Garcı´a-Calvo, E., 2008a. Ranking potential impacts of priority and emerging pollutants in urban wastewater through life cycle impact assessment. Chemosphere 74 (1), 37e44. Mun˜oz, I., Malato, S., Rodriguez, A., Dome`nech, X., 2008b. Integration of environmental and economic performance of processes. case study on advanced oxidation processes for wastewater treatment. Journal of Advanced Oxidation Technologies 11 (2), 270e275. Mun˜oz, I., Rodrı´guez, A., Rosal, R., Ferna´ndez-Alba, A.R., 2009. Life cycle assessment of urban wastewater reuse with ozonation as tertiary treatment: a focus on toxicity-related impacts. Science of the Total Environment 407 (4), 1245e1256. Murray, A., Horvath, A., Nelson, K.L., 2008. Hybrid life-cycle environmental and cost inventory of sewage sludge treatment and end-use scenarios: a case study from china. Environmental Science and Technology 42 (9), 3163e3169. Nemecek, T., Ka¨gi, T., 2007. Life Cycle Inventories of Swiss and European Agricultural Production Systems. AgroscopeReckenholz-Taenikon Research Station ART, Swiss Centre for Life Cycle Inventories, Zurich and Du¨bendorf (Switzerland). Nogueira, R., Brito, A., Machado, A., Salas, J., Martel, G., Vera, L., 2007. “Economic and Environmental Assessment of Small and Decentralized Wastewater Treatment Systems” in Seville (Spain). s 2007 (Activities Report 2007). ONE, 2008. Rapport d’activitec Office National de l’Electricite´, Casablanca, Morocco. REE, 2009. El Sistema Ele´ctrico Espan˜ol 08 (Spanish Electric System 08). Red Ele´ctrica de Espan˜a, Alcobendas, Madrid (Spain). Rodriguez-Garcia, G., Molinos-Senante, Hospido, A., Herna´ndez, F., Moreira, M.T., Feijoo, G., 2011. “Environmental and Economic Performance of Wastewater Treatment Plants: Comparison between Atlantic and Mediterranean Scenarios” in IWA Specialist Conference, Water and Industry 2011, Valladolid (Spain). http://iwawaterwiki.org/xwiki/bin/view/ EventsExtra/1Environmentalandeconomicperformance ofwastewatertreatmentplantscomparisonbetween AtlanticandMediterraneanscenarios. Roeleveld, P., Klapwijk, A., Eggels, P., Rulkens, W., van Starkenburg, W., 1997. Sustainability of Municipal Wastewater Treatment. http://www.iwaponline.com/wst/03510/ wst035100221.htm (accessed 09.04.11.).
6010
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 9 7 e6 0 1 0
Seijas, M., Lago, D., Omil, F., 2003. Identificacio´n de problemas durante la operacio´n de una planta de tratamiento de aguas municipal y propuesta de medidas correctoras (Identifying problems during the operation of a municipal waste water treatment plant and proposing corrective actions). Tecnologı´a del Agua 234, 33e40. Spielmann, M., Bauer, C., Dones, E., Tuchschmid, M., 2007. Transport Services. Swiss Centre for Life Cycle Inventories, Du¨bendorf (Switzerland). Stokes, J.R., Horvath, A., 2010. Supply-chain environmental effects of wastewater utilities. Environmental Research Letters 5 (1), 014015. Suh, Y.J., Rousseaux, P., 2001. Considerations in Life Cycle Inventory Analysis of Municipal Wastewater Treatment Systems Oral presentation at COST, 624. Tangsubkul, N., Beavis, P., Moore, S.J., Lundie, S., Waite, T.D., 2005. Life cycle assessment of water recycling technology. Water Resources Management 19 (5), 521e537.
Tillman, A., Svingby, M., Lundstro¨m, H., 1998. Life cycle assessment of municipal waste water systems. The International Journal of Life Cycle Assessment 3 (3), 145e157. United Nations General Assembly, 2005. 2005 World Summit Outcome. http://www.un.org/summit2005/documents.html (accessed 13.12.10.). Wenzel, H., Larsen, H.F., Clauson-Kaas, J., Høibye, L., Jacobsen, B. N., 2008. Weighing environmental advantages and disadvantages of advanced wastewater treatment of micropollutants using environmental life cycle assessment. Water Science and Technology 57 (1), 27. Wissel, S., Wa¨tzold, F., 2010. A conceptual analysis of the application of tradable permits to biodiversity conservation. Conservation Biology 24 (2), 404e411. Zhang, K., Wen, Z., 2008. Review and challenges of policies of environmental protection and sustainable development in China. Journal of Environmental Management 88 (4), 1249e1261.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 9 7 e6 0 1 0
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Environmental and economic profile of six typologies of wastewater treatment plants G. Rodriguez-Garcia a,*, M. Molinos-Senante b, A. Hospido a, F. Herna´ndez-Sancho b, M.T. Moreira a, G. Feijoo a a b
Department of Chemical Engineering, School of Engineering, University of Santiago de Compostela, 15782 Santiago de Compostela, Spain Department of Applied Economics II, Faculty of Economics, University of Valencia, Campus dels Tarongers, 46022 Valencia, Spain
article info
abstract
Article history:
The objective of wastewater treatment plants (WWTPs) is to prevent pollution. However, it
Received 10 May 2011
is necessary to assess their sustainability in order to ensure that pollution is being
Received in revised form
removed, not displaced. In this research, the performance of 24 WWTPs has been evaluated
17 August 2011
using a streamlined Life Cycle Assessment (LCA) with Eutrophication Potential (EP) and
Accepted 29 August 2011
Global Warming Potential (GWP) as environmental indicators, and operational costs as
Available online 5 September 2011
economic indicators. WWTPs were further classified in six typologies by their quality requirements according to their final discharge point or water reuse. Moreover, two
Keywords:
different functional units (FU), one based on volume (m3) and the other on eutrophication
Eutrophication
reduction (kg PO3 4 removed) were used to further determine sustainability. A correlation
Functional unit
between legal requirements and technologies used to achieve them was found: Organic
Global warming
matter removal plants were found to be less costly both in environmental and economic
Life cycle thinking
terms if volume was used as the functional unit, while more demanding typologies such as
Wastewater treatment costs
reuse plants showed a trade-off between lower EP and higher cost and GWP; however, this is overcome if the second FU is used instead, proving the sustainability of these options and that this FU better reflects the objectives of a WWTP. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
Over the past fifty years, public attitude toward the environment has changed. Adapting itself to the demands of an evolving society, engineering has added sustainability to its general objectives (Davidson et al., 2007). As presented in the report of the United Nations 2005 World Summit, the pillars for sustainability are environmental protection as well as economic and social development (United Nations General Assembly, 2005). This has produced a substantial change in how technology is designed and operated. In this sense, the application of sustainability criteria for the provision of goods
and services is now the main focus, rather than environmental protection based on an end-of-pipe approach (Davidson et al., 2007). Wastewater treatment, an end-of-pipe technology, must comply with environmental, social and economic requirements in order to be considered sustainable (Balkema et al., 2002). The aim of this research paper is to environmentally and economically assess the operation of 24 wastewater treatment plants (WWTPs), classified according to six different typologies. The criteria for comparison among the WWTPs include the selection of the most appropriate functional unit of the system, which assures that the conclusions
* Corresponding author. Tel.: þ34 881 816 020; fax: þ34 881 816 015. E-mail addresses: [email protected] (G. Rodriguez-Garcia), [email protected] (M. Molinos-Senante), almudena. [email protected] (A. Hospido), [email protected] (F. Herna´ndez-Sancho), [email protected] (M.T. Moreira), [email protected] (G. Feijoo). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.08.053
5998
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 9 7 e6 0 1 0
derived from the analysis are consistent. In this sense, two functional units were considered and a sensitivity analysis was carried out.
1.1.
Environmental sustainability and WWTPs
In the last two decades a number of methodologies have been developed for evaluating the environmental sustainability of a product or process. Among them, Life Cycle Assessment (LCA) is a well-established procedure quantifying inputs and outputs as well as the potential environmental impacts associated with a product throughout its whole life cycle (Bauman and Tillman, 2004; ISO, 2006; Finnveden et al., 2009). LCA has been applied to water treatment systems (water treatment plants, sewer systems, and WWTPs) from the earliest stages of the development of the methodology (a review of a majority of published papers on LCA of WWTPs can be found in Larsen et al. (2007)). More recently, literature has been focused on pharmaceutical and personal care products (PPCPs; Wenzel et al., 2008; Mun˜oz et al., 2008a), nutrient removal (Foley, 2009), tertiary treatments (Mun˜oz et al., 2009; Høibye et al., 2008; Larsen et al., 2010), sludge treatment and disposal (Lundin et al., 2004; Murray et al., 2008; Hospido et al., 2010), or global warming impact associated with wastewater treatment (Stokes and Horvath, 2010). Even when the removal of organic matter (OM) and nutrients is a key objective in the operation of a WWTP, the eutrophication produced by the treated effluent is the main environmental impact made by most plants (Roeleveld et al., 1997; Hospido et al., 2004). The magnitude of the other impact categories varies. While WWTPs with tertiary or advanced treatments seem to significantly impact on global warming and acidification (Beavis and Lundie, 2003; Clauson Kaas et al., 2006), the toxicity-related impacts, caused by the presence of heavy metals and PPCPs, both in water (Roeleveld et al., 1997; Larsen et al., 2010) and in the sludge applied to land (Hospido et al., 2005, 2010), are present in all plants. The significance of other impact categories, such as ozone layer depletion and photochemical oxidants formation, has been found negligible in most cases.
1.2.
Economical sustainability and WWTP
The established regulation mechanisms for environmental protection may not be enough to assure the objectives for environmental quality and efficient natural resource use (Zhang and Wen, 2008). A number of authors argue that economic instruments are also very important for the implementation of policies and selection of measures to assist in environmental protection; the economy provides tools, information and instruments for streamlining the decision-making process (Ashley, 2009; Wissel and Wa¨tzold, 2010; Hepburn, 2010). An example of this growing interest is the new role of economic analysis in the Water Framework Directive (WFD, EU, 2000). This directive represents a new advance in water resource planning by integrating a number of economic principles into the management of water in EU member states (Herna´ndez-Sancho et al., 2010). These principles include polluter pays, as well as additional approaches such as the
analysis of cost-effectiveness of pollution mitigation measures and the consideration of economic instruments such as water pricing (Moran and Dann, 2008). Despite the significance of economic analysis in the field of wastewater treatment processes, it has received less attention than environmentally or technologically focused assessments. Available, however, in the field of water reuse are studies with detailed cost analysis, useful in the assessment of different treatment schemes (Herna´ndez et al., 2006; Asano et al., 2007). Other researchers use cost modeling for a better understanding of the cost structure as well as for planning new investments. Several studies (Gonzalez-Serrano et al., 2005; HernandezSancho et al., 2011) have validated and defined cost functions for different wastewater treatment technologies. These studies have considered the correlation of investment, operation, and maintenance costs of the WWTPs with some representative variables. Taking into account the WFD requirements, especially those related to the cost recovery for water services, several cost benefit analyses (CBA) (Godfrey et al., 2009; Chen and Wang, 2009; Molinos-Senante et al., 2010) have been carried out with the aim of identifying cases in which the adoption of measures to achieve a good ecological status for water bodies was required. All the benefits and costs including those which qualify as “nonmarket” must be integrated into the CBA. In recent years, LCA of WWTPs has been combined with different economic indicators. Mun˜oz et al., (2008b) developed an Environmental Economic Score (EES) and applied it to different advanced oxidation technologies. In Larsen et al. (2010) the cost-efficiency of ozonation and pulverized carbon addition is calculated both as stand-alone processes or combined with sand filtration. Another group (Nogueira et al., 2007) compared the technology of activated sludge (AS) with constructed wetlands and slow rate infiltration for small communities. Lin (2009) combined one LCA study with an inputeoutput model for the wastewater system of the metropolis of Tokyo, highlighting that the implementation of the secondary treatment turned into larger GHG emissions. The present study aims to combine LCA and economic assessment based on real process data.
2.
Materials and methods
2.1.
Objectives
The main objective of this paper is to compare the environmental and economic performance of 24 Spanish WWTPs. After a preliminary analysis, the criteria for the classification of the WWTPs in different typologies was based on the quality requirements set in the European urban wastewater directive (EEC, 1991) and Spanish legislation concerning water reuse (MP, 2007) according to the final destination of the treated wastewater (Table S1, Supplementary information). Plants that discharge the treated wastewater to non-sensitive and sensitive areas (as defined by the European Directive) Type 1: Plants designed and operated for the removal of organic matter discharging the treated wastewater to nonsensitive areas.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 9 7 e6 0 1 0
Type 2: Plants designed and operated for the removal of organic matter and nutrients discharging the treated wastewater to non-sensitive areas. Type 3: Plants designed and operated for the removal of organic matter and nutrients discharging the treated wastewater to sensitive areas. Plants that reuse the treated wastewater Type 4: Plants reusing the treated wastewater for irrigation in agricultural land. Type 5: Plants reusing the treated wastewater for industrial purposes. Type 6: Plants reusing the treated wastewater for aquifer recharge. A streamlined methodology based on environmental and economic indicators is presented. Special attention has been given to the technological differences among the six typologies. To a lesser extent, the effect of influent quality, very different due to climatic conditions, such as rainfall, and water use management, has been assessed.
2.2.
Case study
A brief description of the different WWTPs studied can be found in Table 1. The plants under study were selected according to the completeness of the data offered. A representative sample of 24 plants designed for populations larger than 50,000 inhabitants was selected from two different areas: WWTP 1e5 from an Atlantic region, Galicia (NW Spain), with an average rainfall of 1289 mm/year and WWTP 6e24 from a Mediterranean region, Valencia (E Spain), with an average rainfall of 405 mm/year, as described by Rodriguez-Garcia et al. (2011). This group represents around two-thirds of the total number of large plants in the regions of interest.
2.3.
Functional unit
The definition of the functional unit (FU) comprises a physical measurement of the function provided by the system, usually expressed as a certain amount of product (i.e. 1 kg of detergent) or as the service provided by it (i.e. washing of 100 shirts). The potential impacts that can be associated with the product or service will then be totalized and further referred to the FU. When defining the FU of a WWTP, different choices are possible. On one hand, there are studies that considered the volume (m3) of treated water for a certain period of time as the FU (Suh and Rousseaux, 2001; Hospido et al., 2004). Conversely, there are those based on the environmental load associated with a one person equivalent (PE) (Tillman et al., 1998; Hospido et al., 2007). The former has the advantage of being based on physical data while the latter tends to be used for comparative purposes, since it minimizes the differences associated with the influent composition and flow. Here, a functional unit based on the volume of treated wastewater (m3) was used as the first choice for the comparison of the WWTPs (from now, FU1). This approach may give insight into the effect of differences between facilities, both in regard to the influent flow and quality, as well as the removal
5999
yields. However, speaking in a stricter sense, the function of a WWTP is to meet the restrictions imposed by the existing regulations in terms of pollutant concentrations in the discharge of the treated effluent. The existing framework legislation developed in EEC (1991) and EC (1998) directives only requires the reduction of nitrogen and phosphorous for the treated effluents returned to a sensitive area (Table S1). In fact, three of the plants addressed in this study return the treated water to a sensitive body (Type 3) and a fourth one, WWTP 24 must fulfill similar nitrogen requirements for discharging to an aquifer (Type 6, Table S1). Therefore, another choice for the definition of the FU should comprise the removal of both nutrients and organic matter, expressed in eq. removed (from now FU2).1 For that terms of kg PO3 4 reason, the use of this second FU will highlight the differences between the environmental and economic costs of reducing the potential eutrophication associated with the effluent for all the WWTPs.
2.4.
System boundaries
The selection of the system boundaries in LCA studies on wastewater treatments was studied by Lundin et al. (2000). Although the sewer systems have proved to be environmentally relevant (Doka, 2009; Lassaux et al., 2006), the treatment plant remains the main impact contributor. All the plants under study are preceded by combined sewers systems, i.e. a system that jointly collects wastewater and rainwater resulting in the dilution of the influent in all WWTPs. For this reason and for the lack of data concerning the sewer systems, they were not included in this study. However, the effect of dilution will be more relevant in the Atlantic WWTPs due to climatic reasons (Rodriguez-Garcia et al., 2011). Regarding the treatment plant, although the construction stage has been found to be responsible for 25e35% of the GWP associated with a WWTP (Tangsubkul et al., 2005; Doka, 2009), the operation of the facility is considered far more relevant for the rest of the categories (Lundie et al., 2004; Lassaux et al., 2006). On the other hand, the impact demolishing/disposal stage has been found to be negligible (Corominas et al., 2011). Therefore, this assessment considered the environmental impact associated with the operation of primary, secondary, and tertiary treatments (when present); final discharge of the treated effluent; as well as the sludge treatment and its final disposal. The latter is mainly as fertilizer in agricultural soil in the case of most plants, except WWTP 3 where an important fraction of the sludge produced is landfilled. Some plants reuse a fraction of the treated effluent for agriculture irrigation (Type 4: WWTP 17 to 22). Despite the unavailability of specific data for the fraction of reused water for each plant, most WWTPs are located in the Jucar watershed, and since this watershed displays the highest rate of reused wastewater in Spain (MARM, 2008) it was assumed that 1 According to the CML environmental impact assessment method (Guine´e et al., 2002), all substances that could potentially cause eutrophication are related to the reference or equivalence substance (PO3 4 eq) in order to establish the potential impact of a product/process referred to this impact category.
6000
Table 1 e Wastewater treatment plants under study. Size
Secondary treatment
m3/day m3/day pe (r) Activated AS þ N AS þ P AS þ N and P Extended EA þ N and P Oxidation treated treated sludge removal removal removal aireation removal ditch (d) (r) (AS) (EA)
(d) design (r) real.
26,480 54,560 24,640 8080 12,000 40,000 20,664 24,000 60,000 45,000 32,000 8400 18,000 25,000 22,486 8000 30,000 60,000 24,000 60,000 60,000 38,000 15,000 42,000
53,935 51,111 45,227 6300 14,722 38,634 13,681 20,825 37,735 42,029 12,707 10,699 7359 21,290 10,735 7945 28,870 35,613 14,048 30,584 17,676 23,695 12,517 8474
125,452 130,929 191,762 40,770 49,393 193,046 66,787 127,271 243,144 264,744 62,340 83,890 58,693 54,162 60,752 65,422 117,816 229,154 149,575 200,908 141,609 213,676 60,701 66,025
✔ ✔ e e ✔ ✔ ✔ e e e e e e e e e e e ✔ e e e e e
e e e e e e e e e e e ✔ ✔ e e e ✔ e e e e ✔ e e
e e e e e e e ✔ ✔ e e e e e e e e e e ✔ e e e e
e e e e e e e e e ✔ ✔ e e ✔ ✔ ✔ e ✔ e e ✔ e e
e e e ✔ e e e e e e e e e e e e e e e e e e e e
e e e e e e e e e e e e e e e e e e e e e e ✔ ✔
e e ✔ e e e e e e e e e e e e e e e e e e e e e
e e e ✔ e ✔ e e ✔ e ✔ e e ✔ ✔ ✔ ✔ ✔ ✔ ✔ ✔ ✔ ✔ ✔
Removal efficiency COD NT
83% 89% 93% 96% 93% 96% 94% 92% 93% 94% 96% 97% 96% 96% 95% 97% 92% 94% 94% 95% 96% 93% 93% 97%
15% 27% 50% 85% 92% 70% 39% 52% 68% 47% 73% 89% 88% 84% 63% 92% 16% 66% 29% 47% 70% 54% 70% 95%
PT Average
39% 44% 37% 63% 60% 85% 95% 87% 83% 90% 74% 84% 83% 96% 82% 92% 65% 65% 76% 71% 74% 77% 89% 93%
46% 53% 60% 82% 81% 83% 76% 77% 82% 77% 81% 90% 89% 92% 80% 94% 58% 75% 66% 71% 80% 75% 84% 95%
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 9 7 e6 0 1 0
Discharge T1. non WWTP 1 -sensitive WWTP 2 OM removal WWTP 3 WWTP 4 WWTP 5 WWTP 6 WWTP 7 T2. non sen. WWTP 8 OM þ nut WWTP 9 Rem. WWTP 10 WWTP 11 WWTP 12 WWTP 13 T3. sen. area WWTP 14 WWTP 15 WWTP 16 Reuse T4. WWTP 17 agricultural WWTP 18 WWTP 19 WWTP 20 WWTP 21 WWTP 22 T5. Ind. WWTP 23 T6. Aq. WWTP 24
Tertiary treatment
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 9 7 e6 0 1 0
all the treated effluent is reused. The presence of nutrients in reused wastewater also prevents the production of fertilizers in the same way as sludge does (see below). Since in this case wastewater reuse does not displace any marginal technology due to the high cost of desalination, no other avoided products have been included.
2.5.
Life cycle inventory
Inventory data presented in Table 2 is the annual averages for year 2008 except for WWTP 1 (average data corresponds to the period 1998e2003) and WWTP 4 (average date corresponds to 2009). Regarding the sources, data for WWTP 1 was obtained from a previous study (Hospido et al., 2007); data for WWTPs 2e5 was provided by the company in charge of their management and operation; and data for WWTPs 6e24 is from the regional wastewater treatment authority. In addition, background data was obtained from SimaPro databases as follows: - Medium voltage electricity (Dones et al., 2007): The process selected includes electricity production and import/export (data from 2004), transformation from high voltage, direct SF6 emissions to air and electricity losses from medium voltage transmission system. Electricity production and import/export data were updated for 2008 based in MITYC (2009), ONE (2008) and REE (2009). - Chemical products (Althaus et al., 2007): Acrylonitrile manufacture was used for the production of polyelectrolyte (polyacrilamide). The remaining chemicals were directly selected from the Ecoinvent database (Iron III 40%, sodium hypochlorite 15%, sodium hydroxide 50%, liquid sulfuric acid and quicklime, milled, loose and phosphoric acid 85%). - Transport (Spielmann et al., 2007): Trucks 7.5e16 t EURO 3 (2000) were selected as standard transport for chemicals, waste and sludge. The vehicles under this regulation represented 34% of total Spanish trucks in 2008 (DGT, 2009). - Waste (Doka, 2009). Waste from the primary treatment was regarded as municipal waste and treated in an incineration plant with energy recovery. Typically, grit is disposed at an inert landfill and fats are stabilized with lime and cement before disposal in a secure deposit. For those plants where sludge is landfilled (3, 9 and 11), this output was modeled as municipal waste. - Fertilizers avoided (Nemecek and Ka¨gi, 2007): After a sensitivity analysis of 10 N-based and 6 P-based fertilizers from the Ecoinvent database, ammonium sulphate and diammonium phosphate were selected as generic N and P2O5 sources, and a substitutability of 50% and 70% was assumed for the N and P present either in the sludge or in the effluent (in Type 4 plants), respectively (Bengtsson et al., 1997). - Direct emissions from sludge disposal. N2O emissions from sludge application were calculated according to Hobson (2003) and NH3 emissions as defined by Lundin et al. leakage was considered according to Doka (2000). PO3 4 (2009). - CO2 emissions from the water line as well as those derived from the biogas combustion from the anaerobic digestion of sludge were not taken into account since it was considered to be biogenic according to IPCC guidelines (Doorn et al., 2006).
2.6.
6001
Environmental sustainability indicators
As previously stated in section 1.1., the eutrophication potential (EP) has been considered the most relevant impact category in the majority of published LCAs on WWTPs (Larsen et al., 2007). For this reason, eutrophication was selected as the main environmental sustainability indicator and quantified by means of the CML method, which converts all eutrophying substances (Table S1 Supplementary information) to phosphate equivalent (Guine´e et al., 2002). According to Larsen et al. (2007), the global warming potential (GWP) is not among the most relevant impact categories for WWTPs. However, it is usually regarded as a significant environmental problem, at least from a political and social point of view, and it can also be indicative of other energy dependent impacts such as acidification. As a consequence, it was chosen as the second environmental indicator and quantified in accordance with the IPCC guidelines (Table S2, Supplementary information).
2.7.
Economic sustainability indicators
Operational costs, subdivided by categories (energy, staff and others), were chosen as economic indicators, due to their relationship with overall plant management, and presented for both FU1 and FU2. The data shown is the annual average value for 2008, except for WWTP 1 (for which no economic data was available) and WWTP 4 (whose data was from 2009).
3.
Results and discussion
Due to the large amount of data generated, results are presented in figures and tables grouped according to the 6 different typologies. However, similar information for all the individual plants can be found in the supplementary information.
3.1.
Typologies and inventory
Taking into account their legal objectives, both Type 1 and 2 plants constitute a homogenous group since they are only required to remove OM. The difference is, despite not legally being required to do so, Type 2 plants also remove nutrients (N or/and P). A reason for that might be that, on average, the Type 2 influents are noticeably more loaded than the Type 1 influents (Fig. 1, Table 2). The extent of nutrient removal is generally associated with their presence in relatively high concentrations, as occurs in WWTPs 8 and 9 in order to attain high levels of P removal (Tables 1 and 2). Another possible explanation might be that since the areas considered sensitive might vary through time, the managers of Type 2 plants built or upgraded during the last decade, and may have taken into account that their receiving body could be considered sensitive in the future; thus nutrient removal could become a legal requirement. This would also explain why some Type 2 plants have lower removal efficiency than the Type 3 plants, which discharge in sensitive areas and therefore nutrient removal must be accomplished (Table 1).
6002
Table 2 e Life cycle inventory of WWTP, all data is presented for FU1 (m3). Influent
Electricity
Chemicals consumption
COD (g)
NT (g)
PT (g)
COD (g)
NT (g)
PT (g)
From the grid (kWh)
To the grid (kWh)a
Polyelectrolyte (g)
FeCl3 (g)
CaO (g)
NaClO (g)
NaOH (g)
H2SO4 (g)
Transport (kg km)
327 340 422 220 685 648 585 623 763 673 609 918 787 419 599 846 617 623 1107 763 752 886 750 954
19.45 19.84 22.37 47.28 53.36 58.94 54.98 47.61 54.98 59.42 64.15 65.68 73.86 30.91 57.15 59.41 41.59 55.30 70.65 71.98 75.62 76.10 30.20 65.94
0.70 1.22 4.12 6.70 5.06 8.76 10.65 6.98 8.62 9.55 10.53 8.77 10.21 4.49 7.14 8.83 6.37 7.52 6.78 9.86 9.57 12.84 5.78 15.73
55 38 29 8 51 27 37 50 53 43 26 31 31 17 32 23 49 39 67 39 28 65 49 24
16.45 14.53 11.20 7.08 4.23 17.95 33.48 22.94 17.53 31.37 17.22 7.02 8.59 5.03 20.90 5.01 34.98 18.81 50.19 38.00 22.32 34.74 9.16 3.43
0.43 0.68 2.60 2.46 2.04 1.34 0.54 0.91 1.43 1.00 2.69 1.44 1.70 0.20 1.28 0.73 2.24 2.63 1.66 2.90 2.49 2.99 0.65 1.06
0.13 0.14 0.20 0.54 0.29 0.27 0.33 0.13 0.48 0.36 0.63 0.52 0.85 0.31 0.51 0.59 0.13 0.69 0.56 0.56 1.37 0.66 0.50 0.80
e e e e e 0.10 0.24 e 0.04 e 0.06 e e e e e 0.13 0.07 e e e 0.27
0.27 e 0.41 3.91 0.54 1.79 2.09 2.23 1.67 2.86 2.41 2.16 2.70 0.88 3.26 6.67 2.67 1.59 3.65 5.39 3.03 3.84 0.00 11.37
e 22.33 e e e 65.25 e e 54.48 21.75 0.74 22.54 e e 20.54 21.10 53.30 44.37 119.95 81.90 e 20.23 89.93 47.81
e 51.38 e e e e e e e e e e e e 80.63 e e e e e e e e e
e e e e 0.00 5.19 e e 2.92 5.53 e 2.58 1.16 2.86 2.93 e 2.24 23.19 6.77 21.92 3.57 4.29 e 17.93
e e e e e 0.97 e e 0.13 0.89 e 1.00 1.20 3.61 0.81 e 1.50 0.12 0.15 6.74 7.44 1.34 e 3.40
e e e e e 0.17 e e 0.03 e e 0.00 1.08 e 0.56 e 1.25 e 0.29 e 0.08b 0.26 e 0.70
0.01 1.47 0.01 0.08 0.01 1.47 0.04 0.04 1.18 0.62 0.06 0.57 0.12 0.15 2.17 0.56 1.22 1.39 2.62 2.32 0.28 0.60 1.80 1.62
e
Sludge
Waste
Operational costs (V)
To agriculture To landfill Transport Application N (g)d P2O5 (g)d N2O (g) NH3 (g) PO4 (g) Grit (g) MSW (g) Fats (g) Transport Energy (kg WW)c (kg km) (kg WW)c (kg km) as slurry (l) to air to air to water WWTP WWTP WWTP WWTP WWTP WWTP WWTP WWTP WWTP WWTP WWTP WWTP WWTP WWTP
1 2 3 4 5 6 7 8 9 10 11 12 13 14
0.25 0.34 0.17 0.87 0.70 0.54 0.72 0.78 1.06 0.76 0.90 1.44 1.75 0.53
e e 0.22 e e e e e 0.05 e 0.01 e e e
5.13 6.82 7.88 43.49 13.95 10.99 14.70 16.00 21.76 9.18 18.41 29.58 35.80 6.56
0.24 0.27 0.17 0.86 0.67 0.53 0.71 0.77 1.05 0.76 0.89 1.43 1.73 0.52
2.19 4.63 0.23 12.48 4.99 1.73 8.57 9.70 10.18 3.13 e 16.73 3.25 5.41
2.87 4.88 0.13 0.02 1.00 5.46 12.03 7.40 12.35 8.87 e 15.89 0.78 5.33
0.03 0.07 0.00 0.23 0.08 0.03 0.13 0.15 0.16 0.05 e 0.26 0.05 0.09
0.66 1.41 0.07 3.79 1.52 0.53 2.60 2.94 3.09 0.95 e 5.08 0.99 1.64
0.10 0.17 0.00 0.00 0.03 0.19 0.41 0.25 0.43 0.31 e 0.55 0.03 0.18
e 10.13 19.28 32.62 3.70 10.12 3.70 4.78 17.34 25.69 25.43 0.00 55.78 10.14
54.70 12.17 13.18 31.31 3.63 16.67 56.31 17.83 22.00 115.45 20.40 25.53 27.48 5.76
e 1.45 4.24 15.66 1.67 0.85 0.70 1.05 6.55 9.95 0.97 e 1.23 e
6.02 1.94 4.80 15.04 1.69 0.55 1.23 0.46 0.81 1.69 0.94 0.52 1.71 0.20
e 0.013 0.027 0.039 0.028 0.041 0.084 0.069 0.032 0.054 0.078 0.072 0.061 0.042
Staff
Others
e 0.024 0.033 0.053 0.015 0.060 0.108 0.074 0.032 0.082 0.063 0.065 0.153 0.103
e 0.032 0.018 0.025 0.048 0.077 0.031 0.068 0.100 0.100 0.068 0.075 0.090 0.060
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 9 7 e6 0 1 0
WWTP 1 WWTP 2 WWTP 3 WWTP 4 WWTP 5 WWTP 6 WWTP 7 WWTP 8 WWTP 9 WWTP 10 WWTP 11 WWTP 12 WWTP 13 WWTP 14 WWTP 15 WWTP 16 WWTP 17 WWTP 18 WWTP 19 WWTP 20 WWTP 21 WWTP 22 WWTP 23 WWTP 24
Effluent
Some plants produce a certain amount of electricity by combustion of the CH4 produced by the anaerobic digestion of their sludge that is sold to the net rather than used inside the WWTP. WWTP 21 consumes H3PO4 rather than H2SO4. WW: Wet weight. Nutrients values presented reflect their total amount present in the sludge, not the amount that is used by plants according to Bengtsson et al. (1997). a b c d
35.44 24.11 29.95
WWTP WWTP WWTP WWTP WWTP WWTP WWTP WWTP WWTP WWTP
15 16 17 18 19 20 21 22 23 24
1.27 1.72 0.61 0.93 2.10 1.07 1.89 1.73 1.18 1.30
e e e e e e e e e e
26.09 35.30 19.00 18.98 42.99 21.84
1.02 1.65 0.61 0.91 1.69 1.05 1.87 1.71 1.16 1.25
13.22 3.46 6.93 9.28 4.61 11.70 3.84 e 10.10 2.84
12.45 0.07 8.00 11.67 0.09 19.08 0.85 e 9.66 0.11
0.21 0.05 0.11 0.15 0.07 0.18 0.06 e 0.16 0.04
4.01 1.05 2.10 2.82 1.40 3.55 1.17 e 3.06 0.86
0.43 0.00 0.28 0.40 0.00 0.66 0.03 e 0.33 0.00
19.81 22.92 14.70 5.26 13.91 4.92 100.44 20.35 13.75 16.38
22.83 28.90 10.31 23.70 20.27 56.77 67.98 146.84 0.88 35.08
0.35 1.82 0.36 2.04 e 0.80 8.58 0.90 1.29 3.84
0.87 1.06 0.78 0.59 0.70 1.26 4.04 3.43 0.30 1.18
0.055 0.107 0.052 0.045 0.038 0.059 0.101 0.064 0.058 0.125
0.058 0.061 0.078 0.081 0.105 0.082 0.137 0.094 0.133 0.103
0.190 0.115 0.107 0.050 0.122 0.135 0.185 0.131 0.069 0.083
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 9 7 e6 0 1 0
6003
Tertiary treatment, although not required, is present in some facilities included in Type 1 and Type 2. In the Type 1 group, UV disinfection is partially responsible for the higher electricity use of WWTP 4, much in the same way as coagulation/ flocculation for the consumption of FeCl3 in WWTP 6 (Table 2). In the Type 2 group, the effect of tertiary treatment is not so obvious: both WWTP 9 and WWTP 11 present coagulation/ flocculation stages as primary and tertiary treatment, but the consumption of FeCl3 for the latter is almost irrelevant (Table 2). Even more, WWTP 11 also has UV disinfection, but the effect on the energy use is not as evident for WWTP 11 compared to other Type 2 plants as it is for WWTP 4 compared to the rest of Type 1 plants (Table 2). The Type 3 plants do not present such a high influent load as the Type 2 ones; nevertheless they are probably large enough to require nutrient removal in order to fulfill their correspondent legal requirements found in EC (1998) (Table 1, Fig. 1 and Table S1). Although tertiary treatment, and specifically disinfection, is not required to discharge in a sensitive area, this stage is present in all plants. In any case, its effect is not noticeable since all have a chlorination process, although no consumption of chlorinating agents was reported for 2008. This is because the tertiary treatment is not in use at present, and was built in case that reuse of the treated water was required instead of being discharged. All Type 4 plants present tertiary treatments since they require some kind of disinfection process in order to fulfill sanitary requirements. UV disinfection, present in all except WWTP 18 and 19, might be a reason for higher electricity use compared with Type 1. WWTP 18 presents ultrafiltration and reverse osmosis, which might be less energy efficient than UV (Beavis and Lundie, 2003; Clauson Kaas et al., 2006). On the other hand, WWTP 19, with an average consumption (0.56 kWh/m3) uses a filtration process, using electricity at a similar rate to the UV process (Clauson Kaas et al., 2006). WWTP 21 uses by far the most electricity, most likely because of the use of aerobic sludge digestion rather than anaerobic or lime stabilization like the other WWTPs. All Type 4 plants except one utilize some nutrient removal technology, which might seem contradictory since, as indicated in Section 2.4, nutrients present in the reused water partially avoid the use of N and P2O5 -based fertilizers. However, all plants were built and upgraded years before the legislation for water reuse was published, suggesting, as in the case of Type 2 plants, that they can target a specific nutrient or remove both in case the effluent has to be discharged to a sensitive area. Since the data is from 2008, the first year with water reuse regulations, it is likely the high nutrient removal efficiencies correspond to a transition process between water discharge and water reuse (Table 1). Required to supply part of the treated water for industrial purposes, WWTP 23 (Type 5) must fulfill similar requirements to Type 4 plants and thus uses similar technology. Still, their electricity use is not particularly high (Table 2), especially considering this plant makes use of filtration, ultrafiltration, and UV stages, indicating the fraction of water reused in industry might be small. WWTP 24 (Type 6) shows two important characteristics which will notably affect their environmental profile: on the one hand, this plant deals with the highest average load (Fig. 1,
6004
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 9 7 e6 0 1 0
Fig. 1 e Eutrophication Potential associated to the influent (per m3 of influent, FU1).
Table 2); but on the other, it is the facility with the highest requirements, since it not only must achieve a substantial nutrient removal (Tables 1 and 2) but also disinfect the treated waterdwater recharging an aquifer. This explains its high energy use (Table 2), associated not only with the extended aeration process, but also with the UV disinfection and the filtration of the tertiary treatment.
3.2. Environmental profile according to the volumerelated functional unit (FU1) 3.2.1.
Eutrophication potential calculated for FU1
Average results of the EP based on the FU1 (m3), grouped by typology, are presented in Fig. 2, while individual data for all the plants is detailed in Figs. S1 and S2 in the supplementary information. In addition to the impact associated with the effluent of the plants, the eutrophication associated with the treatment process, as well as the beneficial consequences of the different avoided products, are also included here. This indirect eutrophication is never higher than 10% of the effluent EP and is approximately 4% for most plants. Avoided products show a small contribution to the whole picture, although they are not insignificant for Type 4. Despite the obvious differences in the influent composition (Fig. 1) of types 1 and 2, the effluents of both groups present
a relatively similar effluent quality (Fig. 2) since the WWTPs operate to accomplish identical legal requirements (maximum concentration of 125 mg COD/m3 in the discharged effluent). As seen in Table 1 and already discussed in Section 3.1, Type 2 plants have implemented technologies for nutrient removal for their highly loaded wastewater, which probably justifies the large difference between both Figs. 1 and 2 regarding Type 1 and Type 2. For Type 3, they are legally required to remove both N and P, which explains their high efficiencies (Table 1). However, their lower EP is not only due to their efficiency, but also because of their lower influent loads (Fig. 1). The Type 4 WWTPs present a noticeably high impact (Fig. 2). Despite most of them using some kind of nutrient removal process, their removal efficiency is relatively low compared to that of Type 3, the reason being nutrients present in the water are a valuable resource for agriculture and there would be no point in totally eliminating them. It is worthwhile to note the impact of the effluent is still high, even considering more than half of the nutrients are expected to be absorbed by plants (Bengtsson et al., 1997). Type 5 presents a behavior much like the Type 3 plants as both of them remove nutrients with distinctive efficiency (Table 1) and the influent loads are not particularly different (Fig. 1. and Table 2.). Finally, although Type 6 presents the
Fig. 2 e Eutrophication Potential of the WWTP typologies considered based on FU1 (kg PO4 eq./m3).
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 9 7 e6 0 1 0
6005
Fig. 3 e Global Warming Potential of the WWTP typologies considered based on FU1 (kg CO2 eq./m3).
highest influent load of all WWTPs, its extraordinary high removal yields (Table 1) make this plant one of the facilities with the lowest effluent loads.
3.2.2.
and 5 present a similar emission rate to previous groups, although greater emissions for Type 4 plants are caused by WWTP 21, which displays by far the peak emission rate (Fig. S3) due to its high electricity use.
Global warming potential calculated for FU1
Average values of GWP impacts for the different typologies are presented in Fig. 3, while individual results are included in the Supplementary Information (Fig. S3). Not surprisingly, the increasing impacts in GWP are due to the increasing complexity of technology applied, and it is mainly associated with larger consumptions of electricity and chemicals (Table 2). Type 1 plants can be considered as a baseline scenario since they only fulfill the basic function of a WWTP: OM removal. The higher impact of the WWTPs belonging to Types 2, 3 and 4 can be partially attributed to their higher average COD concentration in the influent (Table 2), which requires larger aeration periods. Another factor would be the nitrification process (Table 1), which also demands more oxygen and thus, more electricity (Table 2). In some cases, it is also associated with tertiary treatments (Beavis and Lundie, 2003; Clauson Kaas et al., 2006), which are necessary to fulfill the requirements of the reuse water legislation but which are also employed by WWTPs discharging in sensitive areas. Types 4
3.3. Operational efficiency: environmental profile according to the eutrophication-related functional unit (FU2) 3.3.1.
Eutrophication potential calculated for FU2
Based on the FU2 (kg of PO3 4 eq. removed), average results are presented in Fig. 4, while detailed information for individual facilities is found in the supplementary information (Fig. S4). On average, this approach establishes clear differences between simple and increasingly complex technologies, penalizing the scheme used for the removal of the organic matter only (Type 1). This might suggest there is a margin for improvement for Type 1 plants, particularly WWTP 1, 2 and 3, either by optimizing the OM removal or by including the removal of nutrients. However, it is also necessary to indicate that the treatment of low load water (a feature of Type 1) has particular difficulties such as low sludge decantability (Seijas et al., 2003), and high OM removal rates could not always be guaranteed.
Fig. 4 e Eutrophication Potential of the WWTP typologies considered based on FU2 (kg PO4 eq./kg PO4 eq. removed).
6006
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 9 7 e6 0 1 0
Fig. 5 e Global Warming Potential of the WWTP typologies considered based on FU2 (kg CO2 eq./kg PO4 eq. removed).
According to the results based on FU2, the implementation of combined treatment for organic matter and nutrient removal clearly benefits the environmental performance of non-Type 1 plants, especially in the case of Type 6, which has the highest removal efficiency (Table 1). FU2 also shows, despite being less efficient in absolute terms (Table 1, Fig. 2), Type 2 presents a profile relatively similar to Type 3, reinforcing the idea that even when Type 2 plants do not need to remove nutrients, they put the same effort into doing so as plants legally required to do it (Type 3). It also balances the higher values presented in Fig. 1 for Type 4 due to its high removal in absolute terms.
3.3.2.
Global warming potential calculated for FU2
The contributions of the different wastewater treatment typologies to global warming expressed by the avoided eutrophication (FU2) are presented in Fig. 5 (individual results in Fig. S5). The differences between Types 1 and 2 are also evident here, further emphasizing that the former are on average less efficient, requiring more electricity for the same level of eutrophication reduction. To a lesser extent, this is also the case for Type 5, which presents a relatively higher profile than in Fig. 3 Type 3 and 4 share a similar profile despite the higher electricity use of the latter (Table 2, Fig. 3) due to the higher absolute removal by Type 3 plants. Aquifer recharge
(Type 6) is also revealed as a fairly good environmental option (Fig. 5), even in spite of extensive use of electricity (Table 2) and GWP emissions (Fig. 3).
3.4.
Economic profile
3.4.1. Economic profile according to the volume-related functional unit (FU1) Operational costs (OC) per FU1 (m3) are presented in Fig. 6 (as well as Fig. S6). As shown in Fig. 6, the operational costs of the six WWTP typologies are highly variable, since the minimum value is 0.127 V/m3 for Type 1 WWTPs, while the maximum is 0.311 V/ m3 for Type 6. For plants that discharge the treated wastewater to non-sensitive areas, Type 2 increase in cost of 75.6% compared Type 1 due to their nutrient removal. In regards to the two typologies of WWTPs that remove organic matter and nutrients (Types 2 and 3), the cost difference is quantified by 18%. This is because plants discharging regenerated water to sensitive areas display higher removal efficiencies for both N and P (Table 1). In relation to the three types of WWTPs that reuse the treated water, cost differences between them are very small, although the Type 6 plant presents slightly higher costs due to two factors: first, as shown in Table 1, the pollutants removal efficiency is higher than for the other
Fig. 6 e Operational costs of the WWTP typologies considered based on FU1 (V/m3).
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 9 7 e6 0 1 0
6007
Fig. 7 e Operational costs of the WWTP typologies considered based on FU2 (V/kg PO4 eq. removed).
WWTPs; second, the real wastewater flow treated by this plant (Table 1) is the lowest and therefore, is less affected by scale economies. Regarding the cost distribution: on average 26% of the total cost is associated with energy consumption, 35% with the staff, and the remaining 39% with others. That about a quarter of the total cost is linked to electricity highlights the importance of efficiency in the use of energy, both from an environmental and economic point of view. Fig. 6 shows that Types 4 and 5 are those spending the smallest percentage of total cost on the item energy. In contrast, WWTP 24 (Type 6) is the one with the highest dependence on electricity. In looking at electricity use data, expressed in kWh/m3 (Table 2), it is verified there is no direct relationship between electricity use and costs. This is because currently WWTP operators can negotiate their fee with the electric companies. Therefore, to reduce energy costs in WWTPs, a double strategy may be adopted: operators may negotiate lower electricity fees, while they may also reduce the use of electricity by increasing their efficiency. The second strategy is more beneficial since it would effect not only a reduction in costs but also reduce the carbon footprint of these facilities, but both may be implemented. In relation to staff costs, they are similar to those presented by MolinosSenante et al. (2010), who quantified them at 32% of the total operation costs. It is worth noting the percentage of the Type 5 WWTP in which staff costs represent a significantly higher percentage than the average (51% vs. 35%). That regenerated water is used for industrial purposes supposes a high monitoring of effluent parameters and ergo, high staff costs. Finally, the item “Others” is the least value-consistent parameter among WWTP types. This item consists of multiple elements (i.e. reagents, waste management and maintenance) and thus large variations would be expected.
3.4.2. Economic profile according to the eutrophication-related functional unit (FU2) Economic efficiency results based on FU2 are presented in Fig. 7, while individual results are displayed in Fig. S7. It is noted that the differences among the WWTP’s types are even more significant when using the FU2 than FU1. Type 1 and 2 experienced slightly lower costs than those discharged to sensitive areas. In the case of the non-sensitive areas, unlike
FU1 results, Type 2 plants show lower costs than Type 1. Thus, when using FU2 as a basis for comparison, it is revealed that nutrient removal is not necessarily more expensive than OM removal. Also, it is noted that plants discharging treated water to sensitive areas and those reusing the regenerated water for agricultural purposes have very similar operating costs. In regards to other plants, Type 5 and 6 are those experiencing the maximum and minimum costs, respectively. Nevertheless, it is necessary to state that these are typologies represented only by one plant and they may not be an accurate reflection of Industrial Reuse and Aquifer Recharge WWTPs.
4.
Conclusions
According to the results of this study, grouping WWTPs based on their legal requirements has exposed the link between these legalities and the technology used to achieve them. Non-Sensitive Discharge WWTPs tend to be associated with OM removal technology or with nutrient removal for specific problems and with relatively low efficiency. Sensitive Discharge and Environmental Reuse (aquifer refill) plants always present combined N and P removal as well as tertiary treatment. Agricultural and Industrial Reuse plants demand tertiary treatment due to microbiological requirements and although nutrients can be considered a valuable resource, several plants implement N and P removal, although with lower efficiency. The selection of the functional unit has proved to be a key aspect in defining both economic and environmental profiles. The first FU defined here (m3 of treated water) showed that Type 1 (non-sensitive discharge, OM removal only WWTPs), resulted in lower impacts for EP and GWP as well as lower costs, suggesting that the other typologies were less efficient. Conversely, the definition of a second FU based on EP reduction acknowledged the higher efficiency of Types 2 and 6 (nonsensitive areas discharge with OM and nutrient removal and aquifer recharge plants), resulting, on average, in a better environmental and economic performance. Although FU1 (m3) presents more intuitive results, FU2 (kg PO3 4 removed) has proved to better reflect the function of a WWTP when
6008
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 9 7 e6 0 1 0
focus is on eutrophication and thus is considered a more useful FU for comparative studies. All in all, results show obtaining an effluent of higher quality, meaning disinfected and with lower eutrophication potential, increases both with GWP and overall expense. It also revealed this higher cost is well-balanced, and is even beneficial for advanced typologies. Finally, for a wastewater treatment technology to be judged sustainable, it must comply with environmental, sociocultural and economic needs. Therefore, the on-going research is focused on incorporating social variables with the already-established approach in order to obtain a complete set of indicators of sustainability for each WWTP under consideration.
Acknowledgments The authors would like to thank the water management entities of the Generalitat Valenciana (EPSAR) and the Xunta de Galicia (Augas de Galicia-EPOSH) and the water management companies (Aquagest, IDOM and Geseco) that have supplied the data presented here. This study has been partially financed by the Spanish Ministry of Education and Science (Consolider Project-NOVEDAR) (CSD2007-00055), Xunta de Galicia (Project 09MDS010262PR) and Generalitat Valenciana (Project ACOMP/2010/138). M. Molinos and A. Hospido acknowledge the FPU program (AP2007-03483) and Isidro Parga Pondal program (IPP-06-57), respectively, for financial support.
Appendix. Supplementary material Supplementary data related to this article can be found online at doi:10.1016/j.watres.2011.08.053.
references
Althaus, H.J., Chudacoff, M., Hischier, R., Jungbluth, N., Osses, M., Primas, A., 2007. Life Cycle Inventories of Chemicals. EMPA Du¨bendorf, Swiss Centre for Life Cycle Inventories, Du¨bendorf (Switzerland). Asano, T., Burton, F., Leverenz, H., Tsuchihashi, R., Tchobanoglous, G., 2007. Water Reuse: Issues, Technologies, and Applications. McGraw-Hill Professional, New York. Ashley, R., 2009. Briefing: N. G. Semple’s ‘Principles of environmental protection’. Proceedings of the ICE-Engineering Sustainability 162 (1), 3e11. Balkema, A.J., Preisig, H.A., Otterpohl, R., Lambert, F.J.D., 2002. Indicators for the sustainability assessment of wastewater treatment systems. Urban Water 4 (2), 153e161. Bauman, H., Tillman, A., 2004. The Hitch Hiker’s Guide to LCA. Studentlitteratur AB, Lund, (Sweden). Beavis, P.C., Lundie, S., 2003. Integrated environmental assessment of tertiary and residual treatment-LCA in the wastewater industry. Water Science and Technology 47 (7e8), 109e116. Bengtsson, M., Lundin, M., Molander, S., 1997. Life Cycle Assessment of Wastewater Systems: Case Studies of Conventional Treatment, Urine Sorting and Liquid
Composting in Three Swedish Municipalities. Chalmers University of Technology. Chen, R., Wang, X.C., 2009. Cost-benefit evaluation of a decentralized water system for wastewater reuse and environmental protection. Water Science and Technology 59 (8), 1515. Clauson Kaas, J., Dahl, F., Dalgaard, O., Høibye, L., Kjølholt, J., Wenzel, H., et al., 2006. Viderega˚ende renseteknologier for kommunalt spildevand (Advanced Treatment Technologies for Municipal Wastewater). DANVA, Miljøstyrelsen, Lynettefællesskabet og Spildevandscenter. Corominas, L., Foley, J., Guest, J. S., Hospido, A., Larsen, H. F., and Shaw, A., 2011. Towards a Standard Method for Life Cycle Assessments (LCA) of Wastewater Treatment. In: Watermatex 2011, 8th IWA Symposium on System Analysis and Integrated Assessment. San Sebastian, Spain. Davidson, C.I., Matthews, H.S., Hendrickson, C.T., Bridges, M.W., Allenby, B.R., Crittenden, J.C., et al., 2007. Adding sustainability to the engineer’s toolbox: a challenge for engineering educators. Environmental Science and Technology 41 (14), 4847e4850. DGT, 2009. Anuario Estadistico General 2008 (General Statistics Yearbook 2008). Direccion General de Trafico. http://www.dgt. es/was6/portal/contenidos/documentos/seguridad_vial/ estadistica/parque_vehiculos/series_historicas_parque/ series_historicas_parque_2008.pdf (accessed 09.05.11). Doka, G., 2009. Wastewater Treatment. Life Cycle Inventories on Waste Treatment. Swiss Centre for Life Cycle Inventories, Du¨bendorf (Switzerland). Dones, R., Bauer, C., Burger, B., Emmenegger Faist, M., Frischknecht, R., Heck, T., et al., 2007. Life Cycle Inventories of Energy Systems. Paul Scherrer Institut Villigen, Swiss Centre for Life Cycle Inventories, Du¨bendorf (Switzerland). Doorn, M.R.J., Towprayoon, S., Manso-Vieira, S.M., Irving, W., Palmer, C., Pipatti, R., et al., 2006. Wastewater treatment and discharge. In: Eggleston, S., Buendia, L., Miwa, K., Tanabe, K. (Eds.), Waste. 2006 IPCC Guidelines for National Greenhouse Gas Inventories. IGES, Japan, pp. 6.1e6.28. EC, 1998. Directive 98/15/EC of 27 February 1998 mending council directive 91/271/EEC with respect to certain requirements established in annex I thereof. Official Journal of the European Communities L 67, 29e30. EEC, 1991. Directive 91/271/1991 EEC of 21 May 1991 concerning urban waste water treatment. Official Journal of the European Communities L 135, 40e52. EU, 2000. Directive 2000/60/EC of 23 October 2000 establishing a framework for community action in the field of water policy. Official Journal of the European Communities L 327, 1e72. Finnveden, G., Hauschild, M.Z., Ekvall, T., Guine´e, J., Heijungs, R., Hellweg, S., et al., 2009. Recent developments in life cycle assessment. Journal of Environmental Management 91 (1), 1e21. Foley, J., 2009. Life cycle assessment of wastewater treatment systems. PhD Thesis. University of Queensland, Australia. Forster, P., Ramaswamy, V., Artaxo, P., Berntsen, T., Betts, R., Fahey, D., et al., 2007. Changes in atmospheric constituents and in radiative forcing. In: Climate Change 2007: The Physical Science Basis. Cambridge University Press, Cambridge, UK and New York, NY, USA. Godfrey, S., Labhasetwar, P., Wate, S., 2009. Greywater reuse in residential schools in Madhya Pradesh, Indiaea case study of cost-benefit analysis. Resources, Conservation and Recycling 53 (5), 287e293. Gonzalez-Serrano, E., Rodriguez-Mirasol, J., Cordero, T., Koussis, A., Rodriguez, J., 2005. Cost of reclaimed municipal wastewater for applications in seasonally stressed semi-arid regions. Journal of Water Supply: Research and TechnologyAQUA 54 (6). http://www.iwaponline.com/jws/054/ jws0540355.htm (accessed 09.04.11.).
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 9 7 e6 0 1 0
Guine´e, J.B., Heijungs, R., Huppes, G., Keijn, R., de Koning, A., van Oers, L., et al., 2002. Handbook on Life Cycle Assessment: Operational Guide to the ISO Standards. Kluwer Academic Publishers, Dordrecht (The Netherlands). Heijungs, R., Guine´e, J.B., Huppes, G., Lankreijer, R., Udo de Haes, H., Sleeswijk, A., 1992. Environmental Life Cycle Assessment of Products. Centre of Environmental Science, Leiden (The Netherlands). Hepburn, C., 2010. Environmental policy, government, and the market. Oxford Review of Economic Policy 26 (2), 117e136. Hernandez-Sancho, F., Molinos-Senante, M., Sala-Garrido, R., 2011. Cost modelling for wastewater treatment processes. Desalination 268 (1e3), 1e5. Herna´ndez-Sancho, F., Molinos-Senante, M., Sala-Garrido, R., 2010. Economic valuation of environmental benefits from wastewater treatment processes: an empirical approach for Spain. The Science of the Total Environment 408 (4), 953e957. Herna´ndez, F., Urkiaga, A., De las Fuentes, L., Bis, B., Chiru, E., Balazs, B., et al., 2006. Feasibility studies for water reuse projects: an economical approach. Desalination 187 (1e3), 253e261. Hobson, J., 2003. CH4 and N2O Emissions from Waste Water Handling. International Panel on Climate Change. http://www. ipcc-nggip.iges.or.jp/public/gp/english (accessed 09.05.11.). Høibye, L., Clauson-Kaas, J., Wenzel, H., Larsen, H.F., Jacobsen, B. N., Dalgaard, O., 2008. Sustainability assessment of advanced wastewater treatment technologies. Water Science and Technology 58 (5), 963. Hospido, A., Carballa, M., Moreira, M., Omil, F., Lema, J.M., Feijoo, G., 2010. Environmental assessment of anaerobically digested sludge reuse in agriculture: potential impacts of emerging micropollutants. Water Research 44 (10), 3225e3233. Hospido, A., Moreira, M.T., Feijoo, G., 2007. A comparison of municipal wastewater treatment plants for big centres of population in Galicia (Spain). The International Journal of Life Cycle Assessment 13 (1), 57e64. Hospido, A., Moreira, M.T., Ferna´ndez-Couto, M., Feijoo, G., 2004. Environmental performance of a municipal wastewater treatment plant. The International Journal of Life Cycle Assessment 9 (4), 261e271. Hospido, A., Moreira, T., Martı´n, M., Rigola, M., Feijoo, G., 2005. Environmental evaluation of different treatment processes for sludge from urban wastewater treatments: anaerobic digestion versus thermal processes (10 pp). The International Journal of Life Cycle Assessment 10 (5), 336e345. ISO, 2006. ISO 14040:2006-Environmental Management e Life Cycle Assessment e Principles and Framework. http://www. iso.org/iso/catalogue_detail?csnumber¼37456. Larsen, H.F., Hauschild, M.Z., Wenzel, H., Almemark, M., 2007. Homogeneous LCA Methodology Agreed by NEPTUNE and INNOWATECH. Deliverable 4.1. EC Project “NEPTUNE”, contract No.:036845. www.eu-neptune.org. Larsen, H.F., Hansen, P.A., Boyer-Souchet, F., 2010. Decision Support Guideline Based on LCA and Cost/Efficiency Assessment. Deliverable 4.3. EC Project “NEPTUNE”, contract No.: 036845. www.eu-neptune.org. Lassaux, S., Renzoni, R., Germain, A., 2006. Life cycle assessment of water: from the pumping station to the wastewater treatment plant (9 pp). The International Journal of Life Cycle Assessment 12 (2), 118e126. Lin, C., 2009. Hybrid inputeoutput analysis of wastewater treatment and environmental impacts: a case study for the Tokyo Metropolis. Ecological Economics 68 (7), 2096e2105. Lundie, S., Peters, G.M., Beavis, P.C., 2004. Life cycle assessment for sustainable metropolitan water systems planning. Environmental Science and Technology 38 (13), 3465e3473. Lundin, M., Olofsson, M., Pettersson, G.J., Zetterlund, H., 2004. Environmental and economic assessment of sewage sludge
6009
handling options. Resources, Conservation and Recycling 41 (4), 255e278. Lundin, M., Bengtsson, M., Molander, S., 2000. Life cycle assessment of wastewater systems: influence of system boundaries and scale on calculated environmental loads. Environmental Science and Technology 34 (1), 180e186. MARM, 2008. Reutilizacion, Libro Digital del Agua (Water Reuse, Digital Water Book). Ministerio de Medio Ambiente, Rural y Marino. http://servicios2.marm.es/sia/visualizacion/lda/ recursos/noconvencionales_reutilizacion.jsp (accessed December 19.12.10.). MITYC, 2009. La Energı´a en Espan˜a 2008 (Energy in Spain 2008). Ministerio de Industria, Turismo y Comercio, Madrid (Spain). MP, 2007. Real decreto 1620/2007, de 7 de diciembre, por el que se establece el re´gimen jurı´dico de la reutilizacio´n de las aguas depuradas (Royal decree 1620/2007 of 7 December 2007 establishing the legal system for treated water reuse). BOE 294/2007, 50639e50661. Molinos-Senante, M., Herna´ndez-Sancho, F., Sala-Garrido, R., 2010. Economic feasibility study for wastewater treatment: a cost-benefit analysis. The Science of the Total Environment 408 (20), 4396e4402. Moran, D., Dann, S., 2008. The economic value of water use: implications for implementing the water framework directive in Scotland. Journal of Environmental Management 87 (3), 484e496. Mun˜oz, I., Jose´ Go´mez, M., Molina-Dı´az, A., Huijbregts, M.A., Ferna´ndez-Alba, A.R., Garcı´a-Calvo, E., 2008a. Ranking potential impacts of priority and emerging pollutants in urban wastewater through life cycle impact assessment. Chemosphere 74 (1), 37e44. Mun˜oz, I., Malato, S., Rodriguez, A., Dome`nech, X., 2008b. Integration of environmental and economic performance of processes. case study on advanced oxidation processes for wastewater treatment. Journal of Advanced Oxidation Technologies 11 (2), 270e275. Mun˜oz, I., Rodrı´guez, A., Rosal, R., Ferna´ndez-Alba, A.R., 2009. Life cycle assessment of urban wastewater reuse with ozonation as tertiary treatment: a focus on toxicity-related impacts. Science of the Total Environment 407 (4), 1245e1256. Murray, A., Horvath, A., Nelson, K.L., 2008. Hybrid life-cycle environmental and cost inventory of sewage sludge treatment and end-use scenarios: a case study from china. Environmental Science and Technology 42 (9), 3163e3169. Nemecek, T., Ka¨gi, T., 2007. Life Cycle Inventories of Swiss and European Agricultural Production Systems. AgroscopeReckenholz-Taenikon Research Station ART, Swiss Centre for Life Cycle Inventories, Zurich and Du¨bendorf (Switzerland). Nogueira, R., Brito, A., Machado, A., Salas, J., Martel, G., Vera, L., 2007. “Economic and Environmental Assessment of Small and Decentralized Wastewater Treatment Systems” in Seville (Spain). s 2007 (Activities Report 2007). ONE, 2008. Rapport d’activitec Office National de l’Electricite´, Casablanca, Morocco. REE, 2009. El Sistema Ele´ctrico Espan˜ol 08 (Spanish Electric System 08). Red Ele´ctrica de Espan˜a, Alcobendas, Madrid (Spain). Rodriguez-Garcia, G., Molinos-Senante, Hospido, A., Herna´ndez, F., Moreira, M.T., Feijoo, G., 2011. “Environmental and Economic Performance of Wastewater Treatment Plants: Comparison between Atlantic and Mediterranean Scenarios” in IWA Specialist Conference, Water and Industry 2011, Valladolid (Spain). http://iwawaterwiki.org/xwiki/bin/view/ EventsExtra/1Environmentalandeconomicperformance ofwastewatertreatmentplantscomparisonbetween AtlanticandMediterraneanscenarios. Roeleveld, P., Klapwijk, A., Eggels, P., Rulkens, W., van Starkenburg, W., 1997. Sustainability of Municipal Wastewater Treatment. http://www.iwaponline.com/wst/03510/ wst035100221.htm (accessed 09.04.11.).
6010
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 9 9 7 e6 0 1 0
Seijas, M., Lago, D., Omil, F., 2003. Identificacio´n de problemas durante la operacio´n de una planta de tratamiento de aguas municipal y propuesta de medidas correctoras (Identifying problems during the operation of a municipal waste water treatment plant and proposing corrective actions). Tecnologı´a del Agua 234, 33e40. Spielmann, M., Bauer, C., Dones, E., Tuchschmid, M., 2007. Transport Services. Swiss Centre for Life Cycle Inventories, Du¨bendorf (Switzerland). Stokes, J.R., Horvath, A., 2010. Supply-chain environmental effects of wastewater utilities. Environmental Research Letters 5 (1), 014015. Suh, Y.J., Rousseaux, P., 2001. Considerations in Life Cycle Inventory Analysis of Municipal Wastewater Treatment Systems Oral presentation at COST, 624. Tangsubkul, N., Beavis, P., Moore, S.J., Lundie, S., Waite, T.D., 2005. Life cycle assessment of water recycling technology. Water Resources Management 19 (5), 521e537.
Tillman, A., Svingby, M., Lundstro¨m, H., 1998. Life cycle assessment of municipal waste water systems. The International Journal of Life Cycle Assessment 3 (3), 145e157. United Nations General Assembly, 2005. 2005 World Summit Outcome. http://www.un.org/summit2005/documents.html (accessed 13.12.10.). Wenzel, H., Larsen, H.F., Clauson-Kaas, J., Høibye, L., Jacobsen, B. N., 2008. Weighing environmental advantages and disadvantages of advanced wastewater treatment of micropollutants using environmental life cycle assessment. Water Science and Technology 57 (1), 27. Wissel, S., Wa¨tzold, F., 2010. A conceptual analysis of the application of tradable permits to biodiversity conservation. Conservation Biology 24 (2), 404e411. Zhang, K., Wen, Z., 2008. Review and challenges of policies of environmental protection and sustainable development in China. Journal of Environmental Management 88 (4), 1249e1261.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 1 1 e6 0 2 0
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Thermal pre-treatment of aerobic granular sludge: Impact on anaerobic biodegradability A. Val del Rı´o a,b,*, N. Morales a, E. Isanta c, A. Mosquera-Corral a, J.L. Campos a, J.P. Steyer b, H. Carre`re b a
Department of Chemical Engineering, School of Engineering, University of Santiago de Compostela, E-15782 Santiago de Compostela, Spain INRA, UR50, Laboratoire de Biotechnologie de l’Environnement, Avenue des Etangs, F-11100 Narbonne, France c Departament d’Enginyeria Quı´mica, Edifici Q-Escola d’Enginyeria, Universitat Auto´noma de Barcelona, E-08193 Bellaterra (Barcelona), Spain b
article info
abstract
Article history:
The aerobic granular systems are a good alternative to the conventional activated sludge
Received 23 June 2011
(AS) ones to reduce the production of sludge generated in wastewater treatment plants
Received in revised form
(WWTP). Although the quantity of produced sludge is low its post-treatment is still
22 August 2011
necessary. In the present work the application of the anaerobic digestion combined with
Accepted 29 August 2011
a thermal pre-treatment was studied to treat two different aerobic granular biomasses: one
Available online 3 September 2011
from a reactor fed with pig manure (G1) and another from a reactor fed with a synthetic medium to simulate an urban wastewater (G2). The results obtained with the untreated
Keywords:
aerobic granular biomasses showed that their anaerobic biodegradability (BD) (33% for G1
Aerobic granules
and 49% for G2) was similar to that obtained for an activated sludge (30e50%) and
Anaerobic digestion
demonstrate the feasibility of their anaerobic digestion. The thermal pre-treatment before
Biochemical methane potential
the anaerobic digestion was proposed as a good option to enhance the BD when this was
(BMP)
initially low (33% G1) with an enhancement between 20% at 60 C and 88% at 170 C with
Biodegradability
respect to the untreated sludge. However when the initial BD was higher (49% G2) the
Sequencing batch reactor (SBR)
thermal pre-treatment produced a slight improvement in the methane production (14%
Thermal pre-treatment
and 18%) and at high temperatures (190 and 210 C) which did not justify the application of such a treatment. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
The sewage sludge production has increased in the European Union (EU) from 5.5 million tons of dry matter in 1992 to 10.1 million tons in 2008, and it is estimated that it will reach 13.0 million tons in 2020 (http, 2011). According to the European Commission (EC) this increase is mainly due to the implementation of the Directive 91/271/EEC for Urban Waste Water
Treatment (CEC, 1991) as well as the rise up in the number of households connected to sewers and in the level of treatment. The disposal of this excess sludge represents up to 50% of the total operational costs of a waste water treatment plant (WWTP) (Appels et al., 2008). The quantity of this waste which is spread on land for agricultural use is near 40% of the total produced amount in the EU and it is regulated by the Directive 86/278/EEC (CEC, 1986). In this context the sewage sludge
* Corresponding author. Department of Chemical Engineering, School of Engineering, University of Santiago de Compostela, E-15782 Santiago de Compostela, Spain. Tel.: þ34 881816739; fax: þ34 881816702. E-mail addresses: [email protected] (A. Val del Rı´o), [email protected] (N. Morales), [email protected] (E. Isanta), [email protected] (A. Mosquera-Corral), [email protected] (J.L. Campos), [email protected] (J.P. Steyer), [email protected] (H. Carre`re). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.08.050
6012
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 1 1 e6 0 2 0
production represents an important environmental and economic point to be considered in the design of the WWTPs and new technologies have to be developed in order to, firstly, reduce its production in the process of origin and, also, to improve its subsequent treatment. The reduction of the sludge production in the origin is an interesting alternative. In this sense, the application of the aerobic granular technology in the WWTPs instead of the conventional activated sludge process could decrease the amount of sludge generated during the secondary treatment (Campos et al., 2009). This technology, where the biomass is forced to grow in compact aggregates in sequencing batch reactors (SBRs), presents various advantages in comparison with the conventional activated sludge process: better settling properties and biomass retention, possibility to carry out simultaneous biological removal processes, capacity to treat higher loads, less surface requirements for its implantation and less sludge production. Liu et al. (2005) estimated the theoretical growth yield of aerobic granules as 0.2 g VSS/g CODremoved, which is in accordance with the results obtained by Mosquera-Corral et al. (2005) operating an aerobic granular reactor fed with a synthetic medium containing acetate. This value involves a reduction of sludge production around 30% with respect to the conventional activated sludge characterized by a sludge growth yield of around 0.3 g VSS/g CODremoved (Heijnen and van Dijken, 1992). However the development of the aerobic granular biomass is still recent and the research has mainly been focused on the establishment of the different optimum parameters for the reactor operation and formation of aerobic aggregates and nowadays on the scale up from laboratory to pilot reactors and to full scale plants. Up to now, no study has been focused on the treatment of this type of sludge before its disposal. Although granular sludge is expected to have anaerobic degradation potential similar to the activated sludge due to their similar origin, specific studies are necessary to prove it. The type of treatment applied to the sludge depends on its composition and its final application but normally the first step consists of a thickening to remove the major quantity of water from the solids and reduce its volume. In this sense the dewatering cost of the aerobic granular sludge could be lower than that of activated sludge due to its higher hydrophobicity (Wang et al., 2005) and better settling properties (Beun et al., 2000). After thickening, the biological digestion (anaerobic digestion, aerobic digestion and composting) is commonly used to transform the amount of highly degradable organic matter into a stable residue and to reduce the number of disease-causing micro-organisms present in the solids before their disposal, for example as fertilizer in agriculture. Among the biological sludge treatments the anaerobic digestion is the most suitable option due to the fact that it allows the stabilization of the sludge and also the production of energy as biogas. From previous research works it has been observed that the anaerobic biodegradability of the sewage sludge ranges from 30 to 50% depending on the type of degraded sludge and its initial organic composition (Mottet et al., 2010). To improve this conversion yield many studies have been performed applying different kinds of pre-treatment (thermal, mechanical and chemical) before the sludge anaerobic
digestion (Appels et al., 2008; Carre`re et al., 2010). The main objective of these pre-treatments is to improve the solids hydrolysis rate since it is the limiting step in the anaerobic digestion and also allow reducing the final amount of sludge to be disposed. Carre`re et al. (2010) compared different pre-treatment methods used to favor the biodegradability of the sludge. Although extracting a simplified conclusion is difficult, these authors observed that the low energy consuming methods, such as sonication and mechanical pre-treatment, increase the hydrolysis rate but with a limited improvement on VS reduction, while the high impact methods, such as thermal hydrolysis and oxidation, have a significant improvement on both aspects, but with higher operational costs. Although the thermal pre-treatment presents high energy consumption, the main part of this energy to heat can be recovered from the biogas produced in the anaerobic process. The literature shows that the thermal treatment can be applied in different ranges of temperature and with different times of treatment. Zheng et al. (1998) used a rapid thermal conditioning (30 s) at high temperature (220 C) and obtained a VS reduction of 55% and a total increase in gas production of 80%, while Gavala et al. (2003) applied the pre-treatment of sludge at 70 C during 7 days to obtain an increase of 26% in the methane production. Furthermore to know the impact that each pre-treatment has, on each sludge biodegradability, an anaerobic test under batch or continuous conditions is normally performed, which implies long operational periods (between 20 and 30 days for batch tests). In this context Mottet et al. (2010) proposed an estimating model to predict the anaerobic biodegradability of waste activated sludge based on the link between the initial composition of the sludge and its biochemical methane potential (BMP). These authors used the partial least square (PLS) regression technique to obtain a model where both macroscopic (soluble organic carbon and COD/TOC ratio) and biochemical (carbohydrates, proteins and lipids concentrations) parameters were used to predict the anaerobic biodegradability of waste activated sludge. The aim of this study was to test the effect of the thermal pre-treatment on the macroscopic and biochemical characteristics of the aerobic granular sludge and also to determine the anaerobic biodegradability enhancement when this pretreatment is applied. The obtained results were also used to validate the model proposed by Mottet et al. (2010) to estimate the anaerobic biodegradability of the aerobic granular biomass.
2.
Material and methods
2.1.
Aerobic granular sludge samples
The aerobic granular sludge samples were taken from two sequencing batch reactors (SBRs) at pilot scale (useful volume of 100 L each). The first tested sludge (G1) was collected from a reactor located at Santiago de Compostela (Spain) fed with the liquid fraction of pig slurry. In this reactor the removal of organic matter and nitrogen occurred in a SBR operated in cycles of 3 h distributed according to the following periods: 3 min of feeding, 171 min of aeration, 4 min of settling and
6013
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 1 1 e6 0 2 0
2 min of effluent withdrawal. The system operated at a solid retention time (SRT) between 4 and 14 days. The second tested sludge (G2) was collected from a reactor located at Barcelona (Spain) fed with a synthetic medium which simulated an urban wastewater. In this case the removal of organic matter, nitrogen and phosphorus took place in a SBR operated with a cycle comprising anoxic and aerobic reaction phases with a cycle length of 3 h distributed in: 60 min of feeding, 111 min of aeration, 6 min of settling and 3 min of effluent withdrawal. In this case the achieved SRT was of 20e40 days. In both cases the sludge samples were taken from the solids purge stream along the reactor operational time and stored at 4 C. In order to remove the major quantity of water and to concentrate the sludge, the samples were settled and the supernatant was discarded. The compositions of the aerobic granular sludge samples used for the experiments are presented in Table 1.
2.2.
Thermal pre-treatment
A thermal pre-treatment was applied to the sludge samples in the range of temperatures from 60 to 210 C for G1 and from 170 to 210 C for G2. When temperatures lower than 100 C were applied to the sample a reactor equipped with a heating/ cooling system was used and the sample volume was of 0.6 L. A Zipperclave (Autoclave France) was used when the assayed temperature values were higher than 100 C and the sample volume was of 0.9 L. For each experiment, once the desired temperature was reached, after 30e60 min from the beginning of the experiment, the pre-treatment was maintained during 20 min. Sludge samples without treatment were arbitrarily associated to 20 C.
2.3.
Analytical methods
In order to determine the composition of the aerobic granular sludge samples several measurements were performed on the original sludge and on the sludge after thermal treatment according to the Standard Methods for the Examination of Water and Wastewater (APHA-AWWA-WPCF, 2005). These measurements were carried out on the total, on the particulate and on the soluble fraction of the sample. To separate and obtain the particulate and soluble fractions each sample was centrifuged at 15,000 rpm for 15 min at 4 C (Beckman JA-20). The measurement of the total and volatile solids concentration was carried out on the total sludge (TS and VS) and on the solids of centrifugation (TSS and VSS). The supernatant solids concentration was calculated as the difference between the total and the suspended ones.
Chemical oxygen demand (COD), total organic carbon (TOC), proteins and carbohydrates concentrations were determined on total sludge and on supernatant (soluble fraction), the particulate fraction was deduced from the difference between both values. The total COD (CODT) was determined according to the open reflux method and the soluble COD (CODS) according to the closed reflux colorimetric method (APHA-AWWA-WPCF, 2005). The TOC was measured with a Shimadzu analyzer (TOC-VCSN) where the total sample was injected with the module SSM-5000A Shimadzu and the soluble fraction with the module ASI-V Shimadzu. The proteins concentration was measured with the Lowry method (Lowry et al., 1951) and expressed in equivalent bovine serum albumin (BSA). The carbohydrates concentration was measured with the anthrone method (Dreywood, 1946) and expressed in equivalent glucose (Glu). Lipids concentration was measured on total sludge by accelerated solvent extraction (ASE 200, Dionex) using petroleum ether and volatile fatty acids (VFA) concentration was measured in the soluble fraction by gas chromatography (GC 3900, VARIAN). The average diameter of the granules in the samples was determined by using an Image Analysis procedure (Tijhuis et al., 1994) with a stereomicroscope (Stemi 2000-C, Zeiss) for particles with a size higher than 1 mm and by using a laser radiation technique (Beckman Coulter LS200 equipped with a LS Variable Speed Fluid Module Plus) for particles with a size lower than 1 mm. The poly-hydroxialkanoates (PHA) concentration was measured according to a modification of the methodology described by Pijuan et al. (2005). An amount around 30 mg of lyophilized sludge samples was digested and methylated with 4 mL of acidulated methanol (10% H2SO4) and 4 mL of chloroform during 20 h at 100 C. Benzoic acid was used as internal standard and the analyses were performed in a GC system (Agilent 6850).
2.4.
Anaerobic biodegradability tests
The tests were carried out in glass flaks of 570 mL (useful volume of 400 mL) with coiled butyl rubber stoppers under the following operational conditions: 35 C, 120 rpm, 4 g VS/L of inoculum, 1 g VSsubstrate/g VS of inoculum, a growth medium containing macro and micro nutrients (1.8 g/L NH4Cl, 0.7 g/L KH2PO4, 0.4 g/L MgCl2$6H2O, 0.2 g/L CaCl2$2H2O, 20 mg/L FeCl2$4H2O, 5 mg/L CoCl2$6H2O, 1 mg/L MnCl2$4H2O, 1 mg/L NiCl2$6H2O, 0.5 mg/L ZnCl2, 0.5 mg/L H3BO3, 0.5 mg/L Na2SeO3, 0.4 mg/L CuCl2$2H2O and 0.1 mg/L Na2MoO4$2H2O) and 2.6 g/L NaHCO3. Two control tests were carried out under the same conditions: a blank without substrate (water) to determine the
Table 1 e Characteristics of aerobic granular sludge samples. Sample G1TOTAL G1SOLUBLE G2TOTAL G2SOLUBLE
TS (g/L) 29.6 1.3 106.1 21.0
NA: Not analyzed.
0.2 0.5 2.9 0.6
VS (g/L)
COD (g/L)
27.3 0.9 60.1 13.8
39.7 1.6 85.7 18.8
0.2 0.5 1.2 0.2
0.2 0.1 3.3 0.9
Proteins (g/L) 16.6 0.4 26.9 1.3
2.7 0.3 3.5 0.2
Carbohydrates (g/L) 3.6 0.2 6.9 0.5
0.2 0.1 1.5 0.1
Lipids (g/L)
PHA (g/L)
VFA(g/L)
0.050 0.007 NA 0.013 0.001 NA
0.8 0.1 NA 5.5 0.2 NA
NA 1.4 0.2 NA 7.5 0.7
6014
a
50
COD & Proteins (g/L)
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 1 1 e6 0 2 0
40 30 20 10
b
5
Lipids & Carb.(g/L)
0
4
0
50
100 150 Temperature (°C)
200
250
0
50
100 150 Temperature (°C)
200
250
3 2 1 0
Fig. 1 e Concentrations of the different compounds on the sample G1: (a) COD (B) and proteins (:); (b) lipids (A) and carbohydrates (>). endogenous activity of the inoculum and a test with an easily biodegradable compound (ethanol) to check the activity of the inoculum in a concentration of 4 g COD/L in the vials. The volume of biogas produced was determined by the variation of pressure inside the glass flask by means of a pressure transducer (Mano 2000 Leo2 Keller) and its composition by gas chromatography (Micro GC CP-4900 VARIAN). From each sample three different fractions were analyzed: (a) the liquid fraction of sludge; (b) the solid fraction of sludge and (c) the whole sludge. The first two determinations were performed to determine the contribution of each phase to the biodegradability of the sample while the third one was done to check the mass balance.
2.5.
Calculations
The solubilization yield due to the thermal treatment was calculated as the ratio between the soluble fraction after the treatment (CODS) minus the initial soluble fraction (CODS0) and the initial particulate fraction (CODP0). An example for the calculation of COD solubilization is presented on Eq. (1).
SCOD ¼
CODS CODS0 CODS CODS0 ¼ COD0 CODS0 CODP0
(1)
The biomethane potential (BMP) was expressed as the volume of methane produced per gram of COD of substrate (mLCH4/g CODsubstrate), and the biodegradability (BD) was determined according to Mottet et al. (2010) by dividing the BMP calculated by the theoretical value in standard conditions (350 mLCH4/g COD at 1 atm and 0 C).
3.
Results and discussion
3.1. Effects of thermal treatment on aerobic granular sludge characteristics The variation of the total COD, proteins, lipids and carbohydrates contents of sample G1 after being submitted to the different tested temperatures was evaluated (Fig. 1). The total COD, proteins and lipids content remained almost constant for all the assayed temperatures, which mean that they were not degraded with the heat. The decrease of carbohydrates content was attributed to the measurement method. As Bougrier et al. (2008) explained the carbohydrates may react with other carbohydrates (“burnt sugar” reactions), which provokes the disappearance of the carbonyl function (C ¼ 0), that is used in their quantification and this is the reason of their underestimation. The results for sample G2 are in agreement with the conclusions exposed for G1. The measured concentration of lipids was low, between 0.05 and 0.18 g/L for G1 and 0.01e0.10 g/L for G2, in comparison with the concentrations observed in activated sludge samples, between 0.24 and 3.4 g/L (Mottet et al., 2010). This is probably due to the low fats content in the feeding of the reactor of origin (pig manure and synthetic urban wastewater) compared with the fats present in an urban wastewater. The VFA concentration in the soluble phase remained constant around 1.4 g/L and 7.5 g/L for G1 and G2, respectively. This result indicates that the accumulation of VFA with the thermal treatment was not significant probably due to the low lipids content in the sludge samples.
6015
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 1 1 e6 0 2 0
a 100
Percentage (%)
80 60 40 20 0 20
b
60
90 115 140 170 Temperature (°C)
190
210
100
Percentage (%)
80 60 40 20 0 20
170 190 Temperature (°C)
210
Fig. 2 e Solids repartition between the particulate and soluble fractions: (a) G1 and (b) G2. Particulate mineral fraction ( ), soluble mineral fraction (-), particulate organic fraction ( ) and soluble organic fraction ( ).
The solids repartition (Fig. 2) indicated that the main effect of the thermal treatment was the solubilization of the organic compounds from the particulate to the soluble phase, while the mineral fraction varied slightly between both phases. This difference might be caused by the error in the measurement and associated with the heterogeneity of the aerobic granules. Differences in the solids composition between sample G1 and sample G2 can also be observed (Fig. 2). In sample G1 (from the reactor treating pig manure to remove C and N) the organic fraction represented 92% of the total solids, while in sample G2 (from the reactor treating synthetic urban wastewater to remove C, N and P) this fraction only represented 55e60%. The
high proportion of the mineral fraction in the sludge G2 can be associated to the higher SRT in the aerobic reactor and above all to the removal of phosphorus and the presence of a precipitate material in the core of some granules (Fig. 3). de Kreuk et al. (2005) operated a SBR with aerobic granular biomass and observed that the mineral content of the granules increased from 6% to 30e41% when the removal of phosphorus was promoted. These authors explained that part of the phosphate removal might be caused by the precipitation of apatite inside de granules. For the experiments performed with temperatures below 115 C the percentages of solubilization were similar and they
Fig. 3 e (a) Image of the core of an aerobic granule of sludge G2; (b) picture of the precipitate with rests of biomass after the thermal treatment at 170 C. The size bar is 0.5 mm.
6016
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 1 1 e6 0 2 0
Solubilization (%)
100 80 60 40 20 0 0
50
100 150 Temperature (°C)
200
250
Fig. 4 e Solubilization percentage of COD (A), VS ( ), proteins ( ) and carbohydrates (B) for G1 with the thermal treatment.
seemed to be independent from the temperature (Fig. 4). However in the case of temperatures above this value the percentages of solubilization increased quickly with the rise of the temperature. To explain this behavior it has to be taken into account that in the experiments of thermal treatment at 60 C and 90 C the aerobic granular sludge G1 presented a viscous aspect similar to that of a gel, reaching the maximum gel compactness structure at 115 C. After removing the sample from the interior of the thermal reactor a “block” of sludge was obtained (Fig. 5). This behavior is contrary to the tendency observed by Bougrier et al. (2008). These authors measured the apparent viscosity of a waste activated sludge after the thermal treatment (temperatures between 20 C and 190 C) and they observed that it decreased with the temperature for all the temperatures tested. An explanation for this gelatinous aspect at temperatures below or equal to 115 C could be the high EPS content, with gelforming properties, of the aerobic granular biomass with respect to the activated sludge (Seviour et al., 2009). Indeed, at
these moderate temperatures the EPS were slightly released from the surface of the granules to the media and they acted as a bond to maintain the gel structure. However, at high temperatures the EPS solubilization was higher, as it can be observed on Fig. 4 (high percentage of proteins and carbohydrates solubilization) and EPS probably lost their gel-forming properties. The maximum solubilization values were obtained at temperature of 190 C for all the parameters measured. For higher temperatures the solubilization percentage of proteins remained almost constant while a decrease of 30% for carbohydrates, 14% for VS and 4% for COD was measured. In case of sample G2 the results obtained with the studied temperatures (170, 190 and 210 C) were in agreement with the results from sample G1: maximum solubilization at 190 C (except for the carbohydrates with a maximum solubilization value of 42% at 170 C) and then a decrease at 210 C. Bougrier et al. (2008) also observed the maximum solubilization ratios around 170e190 C for activated sludge and then a dispersion of the values with higher temperatures, which could be due to the occurrence of reactions between the compounds like the Maillard reaction, which is a chemical reaction between an amino acid and a carbohydrate, usually requiring heat, to form complex molecules. The results of the solubilization percentages obtained with the aerobic granular biomass were compared to those from activated sludge at 190 C (Table 2). For all the parameters measured, the percentage of solubilization was higher in sample G1 than in sample G2 (Table 2). The solubilization values obtained for aerobic granules were in the same range as those reported for the activated sludge samples, except for the proteins, which had a higher solubilization in the case of granular sludge. The thermal treatment also provoked changes on the pH of the liquid media and average diameter of the granules (Table 3). The decrease of pH with the temperature could be associated with the formation of acid compounds due to the degradation of macromolecules, however production of VFA was not observed. The average diameter of the aerobic granules moderately decreased with the temperatures between 60 and 115 C and it strongly decreases with temperatures higher than these (Laurent et al., 2009). However for temperatures between 140 and 210 C it was impossible to obtain a tendency on the evolution of the average diameter, due to the error associated to the measurement. Then it is possible to conclude that, as it was mentioned before, at low temperatures the gelforming structure is responsible of the low diameter reduction, while when the temperature is above 115 C the reduction on
Table 2 e Comparison of the solubilization percentages at 190 C of samples G1, G2 and an activated sludge sample. Sample G1 G2 ASa
Fig. 5 e Image of the aerobic granular biomass G1 after the thermal treatment at 115 C.
SVS
SCOD
SProteins
SCarbohydrate
64 35 40e80
70 57 50e82
83 76 32e60
44 34 16e45
a AS: values corresponding to five activated sludge (Bougrier et al., 2008).
6017
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 1 1 e6 0 2 0
Table 3 e Values of pH and particle average diameter after treatment at different temperatures. Temperature ( C) PH Diameter (mm)
G1 G2 G1 G2
20
60
90
115
140
170
190
210
7.11 6.50 1630 2610
7.29 e 1380 e
7.17 e 1080 e
7.07 e e e
6.97 e 175 e
6.4 5.48 181 118
6.29 5.31 106 155
6.24 5.16 140 120
which corresponds to an enhancement of the BD of 88% with respect to the original sample. Then, with higher temperatures the BD decreased, but it remained higher than the untreated sludge. The initial BD of the sludge G2 without treatment (20 C) was of 49% and there was no significant difference after the thermal pre-treatment at 170 C. Only when the tested temperatures were as high as 190 and 210 C the BD increased slightly up to 56 and 58%, respectively, which supposed an enhancement between 14 and 18% with respect to the untreated sample G2. These results indicated that the high temperatures necessary to obtain a little improvement in the BD do not justify the application of the thermal treatment for this sludge. But it is still interesting to note that for this sample G2 at temperatures over 190 C the tendency of the BD was to increase, whereas the contrary was observed with sample G1 and with sewage sludge (Bougrier et al., 2008; Pinnekamp, 1989). Although sample G1 had a smaller granule average diameter and a lower sludge age (SRT on the reactor of origin) than G2, which could favor a higher BD, the opposite was observed. The
the median diameter is higher. The determination of the average diameter at 115 C was not possible due to the gel structure of the sample.
3.2.
Batch anaerobic digestion
a
250
N-mL CH4/g-COD
To evaluate the effect of the thermal pre-treatment on the anaerobic biodegradability of the aerobic granular sludge batch anaerobic tests were performed after submitting the biomass to the assayed temperature. The cumulative methane production per gram of CODsubstrate for the different pre-treatment conditions was followed in duplicate assays (Fig. 6). The methane production from the control experiments without substrate addition was subtracted. The percentage of biodegradability (BD) for each sample is indicated in Table 4. The initial BD of the sludge G1 without treatment (20 C) was around 33% and although it is in the range of the BD for waste activated sludge (30e50%), this value was near the low limit and indicated a poor BD of G1. The thermal treatment caused an improvement of the anaerobic BD for all the temperatures tested with the sample G1. The BD increased until a maximum value of 62% at 170 C,
200 150 100 50
b
250
N-mL CH 4/g-COD
0
200
0
4
8
12 16 Time (d)
20
24
28
0
4
8
12
20
24
28
150 100 50 0 16
Time (d) Fig. 6 e Cumulative methane production during the batch anaerobic digestion tests: (a) sample G1 and (b) sample G2. Temperature of treatment: 20 C (,), 60 C (:), 90 C ( ), 115 C (- - -), 140 C ( ), 170 C (6), 190 C (A) and 210 C (>).
6018
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 1 1 e6 0 2 0
Table 4 e Results of the anaerobic biodegradability tests with samples G1 and G2 after thermal pre-treatment. Sample G1
G2
Temperature ( C) 20 60 90 115 140 170 190 210 20 170 190 210
mL-CH4/g-VSfed 169 207 236 280 308 337 311 314 243 346 370 404
BMP mL-CH4/g-CODfed
7 10 6 12 14 5 5 18 1 7 15 23
different results obtained with both aerobic granular samples could be explained by the different concentration of PHA and VFA (easily biodegradable compounds). The concentration of these compounds represented the 3.4% and 4.7% of the CODT for PHA and VFA, respectively, on sample G1. However on sample G2 the percentages of PHA and VFA represented a higher contribution with 11.5% and 12.8% of the CODT, respectively. If it is considered that the PHAs and VFAs were degraded completely during the anaerobic digestion, the contribution of these compounds to the biodegradability represented the 8.1% for G1 and the 24.3% for G2, being the difference of 16.2%, the same that was observed in the BMP tests. If the values of BD ( Table 4) are compared, the maximum difference between both aerobic granular samples was observed for the treatments with temperatures higher than 170 C. In the case of G1 the temperature of 170 C was the optimal one for the BD enhancement, while it did not induce any change for G2. At 190 and 210 C the values of BD obtained with both samples were similar; however the enhancement with respect to the original BD was higher in G1. This is in accordance with Bougrier et al. (2008) who observed that the BD enhancement was larger with lower initial BD and higher COD solubilization after the thermal treatment. In this study similar results were observed indicating that sample G1 which had the lower initial BD presented also the higher solubilization percentage after the thermal pre-treatment in comparison with sample G2 (Table 2). The different substrate used in the reactor of origin of the samples (pig manure for G1 and synthetic medium for G2) could be the responsible for the different characteristics of the aerobic granules, being their composition determinant in the potential anaerobic biodegradability. In this sense Mottet et al. (2010) developed a model that correlates the BD and the sample composition for activated sludge. The values of BD obtained in this work were used to validate this model proposed by Mottet et al. (2010) Eq. (2) to predict the anaerobic BD of a sludge based on the following parameters: proteins (Prot), carbohydrates (Carb), lipids (Lpd), COD/TOC ratio (Ox) and soluble organic carbon (SolOc). The results with the model and the error associated with respect to the experimental data are presented in Table 4. From the twelve samples tested the better estimation was for G1 at 20 C (untreated) with an error of 2% and the worse one for G1 at
116 140 166 190 209 219 198 181 170 161 198 204
5 7 4 8 10 4 3 10 1 3 8 11
BD measured 33 40 47 54 60 62 56 52 49 46 56 58
1 2 1 2 3 1 1 3 0 1 2 3
BD model (error) 32 44 41 55 52 55 46 44 41 47 53 53
(2%) (11%) (13%) (1%) (12%) (12%) (18%) (15%) (16%) (3%) (6%) (8%)
190 C with an error of 18%. The prediction error obtained by the authors was of 11% with this model. It has to be underlined that this model has been developed for waste activated sludge samples and that it allows a good prediction of anaerobic biodegradability of aerobic granular sludge. So the model parameters seem to be adequate to predict the anaerobic biodegradability of sewage sludge (granular or not) being quicker to obtain than the results of an anaerobic test. BD ¼ 0:0430:106$Protþ0:661$Carbþ0:836$Lpdþ0:074$Ox þ0:349$SolOc
(2)
It is also interesting to know the way each fraction contributes to the total methane production in the batch anaerobic tests (Fig. 7). The contribution of the soluble fraction increased with the temperature for both samples; however the contribution of the particulate fraction was different. For sample G1 the particulate fraction at temperatures between 60 and 170 C contributed more than the particulate fraction without treatment. However the amount of organic compounds in this phase was lower due to their solubilization, indicating that for G1 the particulate phase became more biodegradable with the thermal treatment in the range of 60e170 C than without treatment. For this sample the maximum contribution of the particulate fraction was at 140 C, being at 170 C a little bit higher than without treatment and quite lower than at 140 C. Then at temperatures higher than 170 C the contribution of the particulate phase was lower than without treatment, surely due to the solubilization of the organic compounds with a consequent higher contribution of the soluble fraction. For sample G2 the contribution of the particulate fraction at 170 C was the lowest one from all the tested temperatures and also quite lower than the sample without treatment, which could explain the low BD at 170 C of G2 in comparison with sample G1 and conventional waste activated sludge. The contribution of the particulate fraction at 190 and 210 C was in the same range than G1 (70e75 N-ml CH4/g-COD). The thermal pre-treatment of the sludge provided different results depending on the origin and the composition of the sludge. In this case BD values after the thermal pre-treatment
6019
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 1 1 e6 0 2 0
N-ml CH4/g-COD
a
200 150 100 50 0
b
200
N-ml CH 4/g-COD
20
150
60
90 115 140 170 Temperature (°C)
190
210
100 50 0 20
170 190 Temperature (°C)
210
Fig. 7 e Methane volume production per gram of COD at the end of the batch anaerobic digestion for particulate (-) and soluble ( ) phase: (a) G1 and (b) G2. for sludge G1 were similar to those obtained with activated sludge. However sludge G2 presented similar tendency but the increase of BD was very low at the temperatures normally applied to activated sludge.
4.
Conclusions
The fact that the origin and composition of the sludge can affect its post-treatment makes it interesting to evaluate each kind of biomass by experimental studies. Experiences performed with two aerobic granular biomasses indicated that their BD (33% and 49%) was similar to that obtained for an activated sludge (30e50%) and demonstrated the feasibility of their anaerobic digestion. The thermal pre-treatment, before the anaerobic digestion, was proposed to enhance the BD when this is initially low and similar results to those from previous works, with conventional waste activated sludge, were obtained. The thermal treatment had a little effect on the total composition of the aerobic granular samples, but an important effect on the solubilization of the organic compounds. The maximum solubilization yield for both aerobic granular biomasses was observed at 190 C. The different characteristics of the aerobic granular samples influenced by the feeding media of the reactor, the involved removal processes, PHA and VFA concentrations, etc., could be responsible for the differences in the initial obtained BD and the optimal temperature for the thermal pre-treatment. In the case of the aerobic granules from a reactor treating pig manure with removal of C and N the initial BD was of 33% and
for all the temperatures tested (60e210 C) the pre-treatment led to an enhancement being between 20% (60 C) and 88% (170 C). However for the aerobic granules from a reactor treating a synthetic wastewater with removal of C, N and P the initial BD was higher (49%) but the thermal pre-treatment only enhanced a little (14% and 18%) the methane production and at high temperatures (190 and 210 C). This slight enhancement definitively does not justify the application of such a treatment. The results obtained in this work allowed to validate the model developed on waste activated sludge by Mottet et al. (2010) for its application to the case of aerobic granular sludge. This calculation allows the estimation of the BD of a sludge based on its chemical characteristics.
Acknowledgments This work was funded by the Spanish Government (TOGRANSYS CTQ2008-06792-C02-01, NOVEDAR_Consolider CSD 2007-00055) and Ministry of Education of Spain (FPU AP200601478).
references
APHA-AWWA-WPCF, 2005. Standard Methods for the Examination of Water and Wastewater, 21th ed. American Public Health Association/American Water Works Association/Water Environment Federation, Washington DC.
6020
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 1 1 e6 0 2 0
Appels, L., Baeyens, J., Degre`ve, J., Dewil, R., 2008. Principles and potential of the anaerobic digestion of waste-activated sludge. Progress in Energy and Combustion Science 34 (6), 755e781. Beun, J.J., van Loosdrecht, M.C.M., Heijnen, J.J., 2000. Aerobic granulation. Water Science and Technology 41 (4e5), 41e48. Bougrier, C., Delgene`s, J.P., Carre`re, H., 2008. Effects of thermal treatments on five different waste activated sludge samples solubilisation, physical properties and anaerobic digestion. Chemical Engineering Journal 139 (2), 236e244. Campos, J.L., Figueroa, M., Va´zquez, J.R., Mosquera-Corral, A., Me´ndez Pampı´n, R.J., Roca, E., 2009. Evaluation of in-situ sludge reduction technologies for wastewater treatment plants. In: Baily, R.E. (Ed.), Sludge: Types, Treatment Processes and Disposal. Nova Science Publishers Inc., New York, pp. 161e186. Carre`re, H., Dumas, C., Battimelli, A., Batstone, D.J., Delgene`s, J.P., Steyer, J.P., Ferrer, I., 2010. Pretreatment methods to improve sludge anaerobic degradability: a review. Journal of Hazardous Materials 183 (1e3), 1e15. CEC, 1986. Council Directive of 12 June 1986 on the Protection of the Environment, and in Particular of the Soil, When Sewage Sludge is Used in Agriculture. Council of the European Communities (Directive 86/278/EEC). CEC, 1991. Council Directive of 21 May 1991 concerning urban waste water treatment. Council of the European Communities (Directive 91/271/EEC). de Kreuk, M.K., Heijnen, J.J., van Loosdrecht, M.C.M., 2005. Simultaneous COD, nitrogen and phosphate removal by aerobic granular sludge. Biotechnology Bioengineering 90 (6), 761e769. Dreywood, R., 1946. Qualitative test for carbohydrate material. Industrial and Engineering Chemical Analytical 18 (8), 499. Gavala, H.N., Yenal, U., Skiadas, I.V., Westermann, P., Ahring, B.K., 2003. Mesophilic and thermophilic anaerobic digestion of primary and secondary sludge. Effect of pre-treatment at elevated temperature. Water Research 37 (19), 4561e4572. Heijnen, J.J., van Dijken, J.P., 1992. In search of a thermodynamic description of biomass yields for the chemotrophic growth of microorganism. Biotechnology and Bioengineering 39 (8), 833e858. http://ec.europa.eu/environment/waste/sludge/pdf/part_i_report. pdf. 2011 Environmental, economic and social impacts of the
use of sewage sludge on land, Final Report, Part I: Overview Report. Laurent, J., Pierra, M., Casellas, M., Dagot, C., 2009. Fate of cadmium in activated sludge after changing its physico-chemical properties by thermal treatment. Chesmosphere 77 (6), 771e777. Liu, Y., Liu, Y.Q., Wang, Z.W., Yang, S.F., Tay, J.H., 2005. Influence of substrate surface loading on the kinetic behaviour of aerobic granules. Applied Microbiology and Biotechnology 67 (4), 484e488. Lowry, O.H., Rosebrough, N.J., Fau, A.L., Randall, R.J., 1951. Protein measurement with the Folin reagent. Journal of Biological Chemistry 193, 265e275. Mosquera-Corral, A., de Kreuk, M.K., Heijnen, J.J., van Loosdrecht, M.C.M., 2005. Effects of oxygen concentration on N-removal in an aerobic granular sludge reactor. Water Research 39 (12), 2676e2686. Mottet, A., Franc¸ois, E., Latrille, E., Steyer, J.P., De´le´ris, S., Vedrenne, F., Carre`re, H., 2010. Estimating anaerobic biodegradability indicators for waste activated sludge. Chemical Engineering Journal 160 (2), 488e496. Pijuan, M., Guisasola, A., Baeza, J.A., Carrera, J., Casas, C., Lafuente, J., 2005. Aerobic phosphorus release linked to acetate uptake: influence of PAO intracellular storage compounds. Biochemical Engineering Journal 26 (2e3), 184e190. Pinnekamp, J., 1989. Effects of thermal pre-treatment of sewage sludge on anaerobic digestion. Water Science and Technology 21 (4e5), 97e108. Seviour, T., Pijuan, M., Nicholson, T., Keller, J., Yuan, Z., 2009. Gelforming exopolysaccharides explain basic differences between structures of aerobic sludge granules and floccular sludges. Water Research 43 (18), 4469e4478. Tijhuis, L., van Loosdrecht, M.C.M., Heijnen, J.J., 1994. Formation and growth of heterotrophic aerobic biofilms on small suspended particles in airlift reactors. Biotechnology Bioengineering 44 (5), 595e608. Wang, Z.W., Liu, Y., Tay, J.H., 2005. Distribution of EPS and cell surface hydrophobicity in aerobic granules. Applied Microbiology and Biotechnology 69 (4), 469e473. Zheng, J., Graff, R.A., Fillos, J., Rinard, I., 1998. Incorporation of rapid thermal conditioning into a wastewater treatment plant. Fuel Process Technology 56 (3), 183e200.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 2 1 e6 0 2 9
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Investigation of the sludge reduction mechanism in the anaerobic side-stream reactor process using several control biological wastewater treatment processes Dong-Hyun Chon a, McNamara Rome a, Young Mo Kim a, Ki Young Park b, Chul Park a,* a b
Department of Civil and Environmental Engineering, University of Massachusetts, Amherst, MA 01003, USA Department of Civil and Environmental System Engineering, Konkuk University, Seoul 143-701, South Korea
article info
abstract
Article history:
To investigate the mechanism of sludge reduction in the anaerobic side-stream reactor
Received 7 May 2011
(SSR) process, activated sludge with five different sludge reduction schemes were studied
Received in revised form
side-by-side in the laboratory. These are activated sludge with: 1) aerobic SSR, 2) anaerobic
17 August 2011
SSR, 3) aerobic digester, 4) anaerobic digester, and 5) no sludge wastage. The system with
Accepted 29 August 2011
anaerobic SSR (system #2) was the focus of this study and four other systems served as
Available online 6 September 2011
control processes with different functions and purposes. Both mathematical and experimental approaches were made to determine solids retention time (SRT) and sludge yield
Keywords:
for the anaerobic SSR process. The results showed that the anaerobic SSR process produced
Activated sludge
the lowest solids generation, indicating that sludge organic fractions degraded in this
Aerobic digestion
system are larger than other systems that possess only aerobic or anaerobic mode. Among
Anaerobic digestion
three systems that involved long SRT (system #1, #2, and #5), it was only system #2 that
Anaerobic side-stream reactor (SSR)
showed stable sludge settling and effluent quality, indicating that efficient sludge reduc-
Sludge reduction
tion in this process occurred along with continuous generation of normal sludge flocs. This
Yield
observation was further supported by batch anaerobic and aerobic digestion data. Batch digestion on sludges collected after 109 days of operation clearly demonstrated that both anaerobically and aerobically digestible materials were removed in activated sludge with anaerobic SSR. In contrast, sludge reduction in the aerobic SSR process or no wastage system was achieved by removal of mainly aerobically digestible materials. All these results led us to conclude that repeating sludge under both feast/fasting and anaerobic/ aerobic conditions (i.e., activated sludge with anaerobic SSR) is necessary to achieve the highest biological solids reduction with normal wastewater treatment performance. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
The activated sludge process is a primary method for treating municipal and industrial wastewater. Despite its high efficiency in removal of organic matter, it generates large amount of excess sludge as a byproduct. The production of excess
sludge has increased in recent years due to more stringent effluent regulatory requirements and higher number of wastewater treatment plants (WWTP) in operation (USEPA, 1999). On the other hand, sludge treatment has become more challenging and more costly with increased restrictions in reuse and disposal of sludge (USEPA, 1992). Finding more
* Corresponding author. Tel.: þ1 413 545 9456; fax: þ1 413 545 2202. E-mail address: [email protected] (C. Park). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.08.051
6022
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 2 1 e6 0 2 9
vigorous ways to remove sludge or to minimize sludge production itself, therefore, has become much more important issue in recent years. Over the last two decades, many studies have attempted to minimize excess sludge generation within the activated sludge process (Yasui and Shibata, 1994; Low and Chase, 1999; Wei et al., 2003; Øegaard, 2004; Park et al., 2004). Among several sludge reduction strategies, the anaerobic side-stream reactor (SSR) process is of particular interest because it has shown significant sludge reduction without causing negative effects on sludge settling and effluent properties (Novak et al., 2007). Although there are some configuration variations in the anaerobic SSR activated sludge process, it fundamentally consists of an aeration basin, a settling tank, and an anaerobic SSR. For the operation of the system, a portion of return sludge, or excess sludge, is recycled through the anaerobic SSR with intentionally minimized sludge wasting. It has been reported that the solids retention time (SRT) of this anaerobic SSR is usually maintained at 10 days under ambient temperature conditions (Novak et al., 2007; Johnson et al., 2008). A few studies have analyzed the reduced sludge yield in the anaerobic SSR process by performing overall solids and COD measurements. For example, Goel and Noguera (2006) showed from their laboratory enhanced biological phosphorous removal (EBPR) process that the incorporation of an anaerobic SSR into EBPR led to 63% sludge reduction (sludge yield at 0.16 mg VSS/mg COD). Novak et al. (2007) also showed that activated sludge with an anaerobic SSR resulted in about 60% less sludge generation (sludge yield from 0.11 to 0.15 mg VSS/ mg COD depending on the sludge interchange rate) than the control activated sludge system. The oxic-settling-anoxic (OSA) process shares some system configuration except for that all of the settled (return) sludge undergoes short anaerobic treatment before reaching the main aeration basin. Saby et al. (2003) reported that 58% of sludge reduction was achieved by controlling the oxidation-reduction potential in their OSA system. It is worth noting that all anaerobic SSR and OSA studies introduced above were carried out using soluble synthetic wastewater. For the possible mechanism of sludge reduction in the anaerobic SSR process, Novak et al. (2007) proposed that reduction of iron in the anaerobic SSR cause release of ironbound organic matter, primarily proteins (Park et al., 2006), that are then rapidly degraded under aerobic conditions when anaerobic sludge is recycled back to the aeration basin. Recently, Chon et al. (2011) demonstrated that about half of overall sludge reduction occurred in the aeration reactor through a long SRT condition while the other half was directly achieved by the anaerobic SSR. In spite of these earlier studies, evaluation of the anaerobic SSR process from a traditional SRT basis has been challenging and has caused controversies. Traditionally, long SRT systems have contributed to upset conditions (i.e., settling properties and effluent quality). However, the anaerobic SSR process, which should show extremely long SRT due to minimal sludge wasting, has been reported to operate well without any evidence of these upset conditions. Furthermore, determining SRT itself is also rather complicated for this process due to continuous recirculation of sludge between the main-stream and the side-stream reactor
and gradual accumulation of some truly inert solids in the system. In addition, limited understanding of the current process could be associated with the lack of knowledge regarding its comparative performance against other sludge reduction systems, for examples, activated sludge with separate aerobic digester, anaerobic digester, or even possibly with aerobic SSR. To gain a better insight of the anaerobic SSR activated sludge process and to further identify the mechanism of sludge reduction in this process, the current study directly compared five activated sludge systems with different sidestream or conventional sludge treatment schemes under control, parallel reactor operation. They are the activated sludge systems with: 1) aerobic SSR, 2) anaerobic SSR, 3) aerobic digester, 4) anaerobic digester, and 5) no sludge wastage. We also used a combination of synthetic wastewater and real wastewater as a feed substrate to better mimic real operation system. Since solids are the one most important parameter to assess the anaerobic SSR system, all solids concentrations were tracked and accounted during the entire reactor operation. Furthermore, both mathematical and experimental methods were used in this study to estimate SRT and observed sludge yield of the systems with anaerobic SSR. Finally, sludges from each system at the end of operation were subjected to both batch anaerobic and aerobic digestion to investigate the remaining pools of digestible materials within activated sludge flocs even after vigorous sludge reduction via side-stream treatment.
2.
Materials and methods
2.1.
System operation
The seed mixed liquor for the five activated sludge systems was taken from the aeration basin in Amherst WWTP, MA, USA. The same influent was fed to the five aeration reactors and consisted of 50/50 (v/v) mixture of real primary effluent from Amherst WWTP and synthetic wastewater. The synthetic wastewater was prepared in the laboratory and included for the purpose of increasing soluble chemical oxygen demand (sCOD) in the influent by adding BactoPeptone (330 mg/L COD) and CH3COONa (110 mg/L COD). Other inorganic constituents included in the synthetic feed followed the composition used in Novak et al. (2007). The characteristics of primary effluent showed variation in sCOD and solids depending on the collection date. Table 1 shows the average
Table 1 e Average values of various parameters of wastewater influent (50/50 (v/v) mixture of synthetic wastewater and primary effluent from a local WWTP). Analyses COD Ammonium nitrogen Phosphorous TSS VSS pH
Value 351 mg sCOD/L 37 mg NHþ 4 N=L 7 mg PO3 4 P=L 22 mg/L 18 mg/L 7.1
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 2 1 e6 0 2 9
values of various characteristics of the final influent (mixed) used in this study. Schematics of five laboratory activated sludge systems are shown in Fig. 1. Hydraulic retention time (HRT) of the main aeration reactor was maintained at 1 day. Sequencing batch reactor (SBR) was chosen for the activated sludge reactor and the working volume was 5 L. A circular type reactor was used for SBR (Vollrath Stainless-Steel Beakers, Fisher Scientific) and its diameter and height were 20.3 and 15.9 cm, respectively. At the bottom of SBR, 8 cm disc diffuser was placed for the aeration. The aeration provided effective mixing of activated sludge during the aeration period. Activated sludge SBR had four cycles a day and each cycle (6 h) consisted of feeding (10 min), aeration (5 h), settling (1 h), and decanting (10 min). Dissolved oxygen concentrations and temperature in five aeration reactors were maintained above 6 mg O2/L and 19 1.5 C, respectively. Systems #1 and #2 were the activated sludge systems that included an aerobic SSR and anaerobic SSR, respectively. For these systems, sludge wasting was minimized from both main aeration reactors and SSRs while 10% mass of activated sludge was recycled between the aeration reactor and SSRs. Sludge cycling happened once a day during a mixing (aeration) period of activated sludge SBR. For this, 500 mL of well mixed sludge was taken from each SBR during the aeration period in the dayetime cycle and thickened to 150 mL of sludge by 1hr settling. The decanted water was discharged to the effluent tank. Prior to feeding of this thickened sludge to SSRs, 150 mL of digested sludge was removed from SSRs to receive the same
6023
volume of fresh sludge. Systems #3 and #4 were operated as a control activated sludge for typical wastewater treatments with a separate anaerobic or aerobic digester. For these systems, both aeration reactors and digesters had a regular wastage to maintain SRT of 10 days of activated sludge and 10 days of digester, respectively. The operation was similar with system #1 and #2 except that sludge taken from digesters were permanently wasted instead of recirculation. System #5 was operated as no sludge wastage system so there was no intentional sludge wasting except for sampling for measurements. The seed sludge for SSRs and digesters was the same sludge that was prepared by digesting the field activated sludge from Amherst WWTP aerobically or anaerobically. The SSRs and digesters had a 1.5 L working volume and were maintained at 10 day HRT which was equal to SRT. Sufficient mixing using a magnetic stirrer (300 rpm) was applied to anaerobic SSR and anaerobic digester and this prohibited the formation of scum on top of the anaerobic reactors. These reactors were operated next to the activated sludge SBRs under same temperature conditions. Mixing of aerobic SSR and digester was maintained by aeration also under the same temperature conditions.
2.2.
Batch digestion study
To investigate the remaining organic pool in sludge, mixed liquor was collected from each aeration reactor at the end of system operation (day 109) and subjected to both aerobic and
Fig. 1 e Overall schematic of five activated sludge (AS) systems. (#1) AS with aerobic SSR; (#2) AS with anaerobic SSR; (#3) AS and separate sludge treatment with aerobic digester; (#4) AS and separate sludge treatment with anaerobic digester; and (#5) No wastage system.
6024
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 2 1 e6 0 2 9
organic materials. Thus, new approaches for the calculation of SRT and overall observed sludge yield were necessary. To determine SRT and sludge yield of the anaerobic SSR process, AS and SSR were considered together as one control volume (Fig. 2). There are three kinds of solids leaving out of the control volume within specific given period of time: solids leaving via effluent, wastage from AS, and wastage from SSR. Total solids staying inside the control volume can be estimated by the sum of solids in AS and SSR. Thus, SRT of whole system could be defined as the total mass of sludge in the system divided by the mass rate of sludge leaving out of the system (Eq. (1)). Fig. 2 e Conceptual schematic for the calculation of SRT and the observed sludge yield (sludge line in Bold) in the overall SSR systems.
anaerobic digestion for 20 days. The working volumes for aerobic and anaerobic batch digestion were 600 mL and 1800 mL, respectively. Aerobic digestion was conducted at room temperature and anaerobic digestion was performed at mesophilic conditions (37 C). Distilled water was added to the aerobic batch digestion to make up water loss by evaporation. A mixing speed of 300 rpm was applied to the anaerobic batch reactors using a magnetic stirrer. Volatile solids concentrations at day 0 and day 20 of batch digestion were measured and compared to estimate volatile solids reduction (VSR) efficiency.
2.3.
Analysis
Total solids (TS), total suspended solids (TSS), total volatile solids (VS), volatile suspended solids (VSS), soluble chemical oxygen demand (sCOD), and sludge volume index (SVI) were measured according to the appropriate methods shown in Standard Methods (APHA, 2005).
SRT ¼
XAS VAS þ XSSR VSSR XAS QAS;waste þ XSSR QSSR;waste þ Xeff Qeff
(1)
Where XAS and XSSR are the sludge concentration in the AS and SSR (g TSS L1), respectively. VAS and VSSR are the volume of AS and SSR (L), respectively. Qin, Qeff, QAS,waste and QSSR,waste are the flow rate of influent, effluent, wastage from AS, and wastage from SSR (L d1), respectively. Table 2 shows SRTs in the five systems that were determined using Eq. (1). The SRTs of system #1, #2, #3, #4, and #5 were found to be 63, 74, 21,16, and 81 days, respectively. These values appeared to be reasonable as systems #1, #2, and #5 were operated with minimized solids wastage except for sampling while system #3 and #4 had a regular wastage from digesters. Among several ways to express sludge yield, the observed sludge yield (Yobs) was used for this study because it can account for the effect of inert solids and slowly biodegradable substrates on sludge generation. Since solids concentrations in both AS and SSR changed in the side-stream processes, cumulative terms should be used to quantify changes in both solids and substrates and to determine Yobs. Then, as the fundamental definition of yield is “the amount of sludge formed per the amount of substrate removed” (Grady et al., 1999), Eq. (2) can be used to determine Yobs.
3. Estimation of solids retention time and observed sludge yield
Yobs ¼ ðCumulative Generated SludgeÞ=
Minimized sludge wasting and continuous recirculation of sludge in the anaerobic SSR process (system #2) led to changes in sludge concentrations in the main activated sludge (AS) reactor and SSR along with continuous operation of the system. This was also the case for system #1 that included aerobic SSR. In these systems, total solids could consistently increase due to the accumulation of inorganic and truly inert
Where
ðCumulative Consumed SubstrateÞ
(2)
Yobs is the overall observed sludge yield (g TSS g1 sCOD). The Cumulative Generated Sludge consists of the increase in sludge in the control volume (DXASVAS þ DXSSR VSSR) itself and cumulative sludge wastage from AS, SSR, and effluent. Thus,
Table 2 e The SRT values of the five systems determined using Eq. (1).
SRT (day)
System #1
System #2
System #3
System #4
System #5
63
74
21
16
81
See Fig. 1 for a full description of each system.
6025
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 2 1 e6 0 2 9
the Cumulative Generated Sludge in the specific given time can be described as Eq. (3). Cumulative Generated Sludge ¼ DXAS VAS þ DXSSR VSSR X XAS QAS;waste þ XSSR QSSR;waste þ Xeff Qeff $Dt þ
ð3Þ
Where DXAS, DXSSR, and Dt are the change of sludge (g TSS L1) in AS and SSR, and time (day), respectively. The Cumulative Consumed Substrate in the specific given time is then Cumulative Consumed Substrate ¼
X Sin Qin Seff Qeff $Dt
(4)
Where Sin and Seff are the substrate concentration of influent and effluent (g sCOD L1), respectively. Therefore, incorporating Eqs. (3) and (4) into Eq. (2) and the rearrangement of Eq. (2) leads to Eq. (5) in which the slope of linear regression line should be the observed sludge yield (Yobs). X XAS QAS;waste þXSSR QSSR;waste þXeff Qeff $Dt DXAS VAS þDXSSR VSSR þ X ¼Yobs Sin Qin Seff Qeff $Dt ð5Þ This indicates that Yobs can be determined using a regression method for the obtained experimental data. This will be further discussed in the later part of this paper.
a
120
4.
Results and discussion
4.1.
Effluent quality and sludge settling properties
Fig. 3a shows effluent TSS data for the five systems. While the anaerobic SSR process (system #2) and two control activated sludge SBRs (system #3 and #4) showed stable effluent TSS, which were mostly well below 30 mg/L, no wastage system (#5) showed significant deterioration in effluent TSS after 40 days of operation. Effluent TSS in system #5 got improved after several days of upsets but deteriorated again at around day 65 and similar trends happened throughout the whole operation. Microscopic analysis of mixed liquor from #5 did not show high number of filamentous organisms. Hence, several washout events that occurred in system #5 were most likely due to a long SRT (Table 2) employed in the given system (Grady et al., 1999). Although system #1 (aerobic SSR process) did not experience substantial increase in effluent TSS as system #5, some deterioration in effluent TSS was also observed from around day 70. Meanwhile, effluent soluble COD from all systems were well below 30 mg/L (data not shown), indicating that removal of soluble COD was not an issue for systems involving long SRT (#1, 2, and 5) and control activated sludge systems (# 3 and #4) as well. Sludge volume index (SVI) data are shown in Fig. 3b. For control activated sludge systems (#3 and #4), stable and effective sludge settling was maintained as indicated by low SVI values. However, SVIs of activated sludge #1, #2 and #5, which had no intentional wastage except for sampling, increased gradually up to day 50. The settling property of
System #1 System #2
Effluent TSS (mg/L)
100
System #3 System #4
80
System #5
60 40 20 0 0
10
20
30
40
50
60
70
80
90
Operation Time (day)
b SVI (Sludge Volume Index)
400
System System System System System
300
#1 #2 #3 #4 #5
200
100
0 0
10
20
30
40
50
60
70
80
90
Operation Time (day)
Fig. 3 e Effluent TSS (a) and sludge volume index (b) data from the five systems: (#1) AS with aerobic SSR; (#2) AS with anaerobic SSR; (#3) AS and separate sludge treatment with aerobic digester; (#4) AS and separate sludge treatment with anaerobic digester; and (#5) No wastage system.
6026
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 2 1 e6 0 2 9
sludge from the anaerobic SSR process got improved and became stable since then, which indicates that this system required some acclimation period. In contrast, sludge from the no wastage system continued to show high SVI, which simultaneously happened with substantial washout events seen in Fig. 3a. The aerobic SSR process showed some improvement in SVI but there was another period of high SVI showing at around day 70. All these are important results because all three systems #1, #2, and #5 had extremely minimized sludge wasting, resulting in very long SRT (Table 2), but showed significantly different characteristics of effluent quality and settling performances.
4.2.
Solids concentrations in the system
The MLSS concentrations in AS #1 and #2 increased initially but reached stable values from day 40 (Fig. 4a). The MLSS concentrations in AS #3 and #4 were maintained at around 1000 mg/L during the entire operation period. For system #5, MLSS gradually increased till day 45 then decreased till day 70 due to the solids loss to the effluent. After that period, MLSS in #5 increased again toward the end of operation. Fig. 4b shows solids concentrations in either SSR or digesters for systems #1 to #4. The solids concentrations in both aerobic digester (system #3) and anaerobic digester (system #4) were maintained at around 3000 to 4000 mg/L. Meanwhile, the solids concentrations in the aerobic SSR (system #1) and anaerobic SSR (system #2) increased till day 30 and then reached relatively stable values during the
a
4.3.
Comparison of observed sludge yield
According to Eq. (5), the cumulative sludge generation data were plotted against the cumulative substrate consumption data and the slope of each linear regression curve was determined for the system observed sludge yield (Yobs). As Fig. 5 shows, Yobs for #1 (AS with aerobic SSR), #2 (AS with anaerobic SSR), #3 (AS with aerobic digestion), #4 (AS with anaerobic digestion), and #5 (no wastage) were found to be 0.163, 0.159, 0.263, 0.309, 0.186 g TSS/g COD, respectively. System #2 had the lowest sludge yield among five systems investigated in this study. The Yobs of system #2 showed 49% less sludge generation than system #4 which also employed an anaerobic digestion but in a separate line of the reactor. This indicates that much lower sludge yield in system #2 resulted from further degradation of organic materials that cannot be degraded via separate anaerobic digestion. The Yobs of system #2 was also lower than that of #5. While systems #2 and #5 both revealed a similar, long SRT (Table 2), the system that incorporated an anaerobic SSR resulted in lower sludge
AS #1 AS #2 AS #3 AS #4 AS #5
5000
4000
MLSS in AS (mg/L)
remaining operation. It is also worth noting that while solids concentrations in both anaerobic and aerobic SSRs slightly increased from day 70 to day 90, it did not lead to an increase in MLSSs in the aeration reactors which had continuously received sludge from its own SSR. These results imply that AS with SSRs (system #1 and #2) achieved further sludge degradation than AS simply maintained with no wastage (system #5) and that their degradation mechanism are also different.
3000
2000
1000
0 0
10
20
30
40
50
60
70
80
90
Operation Time (day)
Solids in SSR and Digester (mg/L)
b 8000
6000
4000
2000 #1 Aerobic SSR
#2 Anaerobic SSR
#3 Aerobic digester
#4 Anaerobic digester
0 0
10
20
30
40
50
60
70
80
90
Operation Time (day)
Fig. 4 e Solids concentrations for five activated sludge systems: (a) MLSS in the aeration reactors and (b) MLSS in SSRs or separate digesters.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 2 1 e6 0 2 9
Cumulative Generated Sludge (g TSS)
50
40
30
System #1: y = 0.163x R² = 0.943
System #2: y = 0.159x R² = 0.913
System #3: y = 0.263x R² = 0.961
System #4: y = 0.309x R² = 0.984
System #1 System #2
System #5: y = 0.186x R² = 0.936
System #3 System #4
20
System #5 10
0 0
50
100
150
Cumulative Consumed Substrate (g COD)
Fig. 5 e The observed sludge yields from five systems.
production. Moreover, sludge settling properties and effluent quality were much better for system #2 as discussed above (Fig. 3). These results strongly indicate that the sludge reduction mechanism in the anaerobic SSR activated sludge process is substantially different than that from the no wastage system in which endogenous decay growth under aerobic condition is predominant. Considering schematic differences in system #2 and #5 and their subsequent effects on operational performances, recirculation of sludge between the aeration reactor (along with the entrance of fresh feed) and anaerobic SSR (starvation) does induce a favorable sludge reduction mechanism and continuously “refresh” the sludge, leading to effective flocculation and sound settling performance of sludge. Since systems #1, #3, and #5 had only aerobic treatment, comparison of their yield values is also of interest. As the data shows, Yobs of system #1 was 38% less than that of system #3 with a separate aerobic digester. The Yobs of system #1 also showed 12% less yield than system #5. These data indicate that materials that cannot be degraded by simple aerobic digestion or even under extended SRT conditions could still be degraded through feast or fasting conditions even under the same aerobic environment. Nonetheless, it needs to be noted that the observed yield of the aerobic SSR process is slightly higher than the anaerobic SSR process. Moreover, much better operational performances were observed for system #2, suggesting that combination of aerobic conditions in AS and anaerobic conditions in SSR provides the best condition for minimization of excess sludge generation and operation performance for the activated sludge process.
4.4.
6027
activated sludge flocs may contain specific organic pools that can degrade only under either aerobic or anaerobic conditions. The batch anaerobic and aerobic digestion test was performed to investigate if the sludges from long SRT systems (#1, #2, and #5) still possessed organic materials that could further degrade in one of two digestion conditions. To achieve this goal, mixed liquors were collected from each activated sludge reactor at the end of operation (day 109) and were subjected to 20 day batch digestion under both aerobic and anaerobic conditions. Volatile solids reduction (VSR) data from this batch digestion study are shown in Fig. 6. The VSR of activated sludge from system #4, which served as a control activated sludge and had a typical activated sludge SRT (10 days), were 42% and 45% in aerobic and anaerobic batch digestion, respectively. It indicates that substantial amount of both “anaerobically” and/or “aerobically” digestible materials remained in this typical activated sludge. The VSR of activated sludge from system #5 (no wastage system) showed 26% reduction by anaerobic digestion but only 5% by aerobic digestion. This data strongly indicates that most of the aerobically digestible materials were nearly exhausted when sludge grew under extremely long SRT but the same sludge still contained a large fraction of materials that can be anaerobically digested. Another recent study performed in our laboratory supported this VSR data. Field activated sludges collected from the facility that was operated in long SRT (above 25 days) showed only 10% VSR by aerobic digestion but 30% VSR by anaerobic digestion (Teague, 2011). All these observations indicate that sludge decay occurring in long SRT activated sludge systems consumes mostly aerobically digestible materials but not anaerobically digestible materials which are presumed to be iron-bound floc organics (Park et al., 2006). Batch digestion of activated sludge from system #2, the focus of this study, led to more balanced VSRs between anaerobic and aerobic digestion. The data showed 17% VSR from anaerobic digestion and 21% VSR from aerobic digestion. These VSRs were also less than half of control activated sludge. These indicate that both aerobically and anaerobically digestible sludge pools were degraded in this system and provide another line of evidence that different sludge reduction mechanism are involved for the anaerobic SSR process and the process simply maintained with no wastage.
Batch digestion test
It was previously suggested that the cell debris, or traditionally referred to as endogenous decay products, are not easily degradable under the conditions cells are grown (Novak et al., 2003). Later Park et al. (2008a,b) conducted molecular examination on activated sludge extracellular proteins and investigated their fate in aerobic and anaerobic digestion. The authors found that the fraction of activated sludge proteins that was degraded by anaerobic digestion and aerobic digestion was different. It can be inferred from these earlier studies that
Fig. 6 e Results of batch anaerobic and aerobic digestion for activated sludges collected after 109 days of operation.
6028
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 2 1 e6 0 2 9
The VSRs of activated sludge #1 were 5% and 20% from aerobic and anaerobic digestion, respectively. The VSR of this aerobic batch digestion was similar with the aerobic VSR for AS #5. The anaerobic VSR of AS #1 was also similar to that of anaerobic digestion for AS #2. This indicates that aerobically digestible materials were fully degraded like the no wastage system and a part of anaerobically digestible materials was also degraded through feast (in activated sludge) and fasting (in aerobic SSR) conditions even under the same aerobic environment. Interestingly, it can be observed that activated sludges from systems #2 and #4 led to the balanced VSR between aerobic digestion and anaerobic digestion but sludges from #1 and #5 did not. Both aerobically and anaerobically degradable organic materials are presumed to play a key role in flocculation and be related with sludge settling. It is therefore postulated that the balanced composition of aerobically and anaerobically digestible materials within floc leads to positive operational performances, as seen with better sludge settling and effluent quality data for system #2 and #4. On the other hand, a large difference in floc composition with respect to the nature of the digestibility could result in poor floc formation, thus, deteriorated operational performances, as seen with systems #1 and #5 (Fig. 3). Overall, considering both operational performance and batch digestion data, repeating of sludge under feast aerobic and fasting anaerobic conditions provides not only the necessary setting for a favorable sludge reduction but also continuously refreshes the sludge which overall lead to better flocculation and effluent quality along with effectively reduced sludge yield.
5.
Conclusions
We conducted intensive side-by-side reactor study to investigate the mechanism of sludge reduction in the anaerobic side-stream reactor (SSR) process for the activated sludge system. The specific conclusions drawn from this study are as follows: Among the five different processes investigated in this study, the anaerobic SSR process produced the lowest sludge generation. Among the three processes that involved long SRT (63 w 81 days), only the anaerobic SSR process led to sound operational performances seen with good effluent TSS and SVI data. The anaerobic SSR process degraded both aerobically and anaerobically digestible materials and continuously refreshed the floc composition leading to effective flocculation and good effluent quality. Reduced sludge yield obtained from an extended SRT operation (no wastage, #5) was mainly achieved by degrading aerobically digestible materials. A significant amount of anaerobically digestible materials still remained in that sludge. Another SSR system (activated sludge with aerobic SSR) still produced better sludge reduction than the control activated sludge with conventional digestion and the
extended SRT system, suggesting that cyclic feast/fasting conditions even under the same aerobic environment could bring further sludge reduction. The settling property, however, became challenging as the system was operated for a longer period. Therefore, recirculation of sludge under aerobic feast and anaerobic fasting condition was the necessary setting to achieve the most effective sludge reduction with sound operational performances.
references
APHA, AWWA, WEF, 2005. Standard Methods for the Examination of Water and Wastewater, 21st ed. American Public Health Association, Washington. Chon, D.-H., Rome, M., Kim, H.-S., Park, C., 2011. Investigating the mechanism of sludge reduction in activated sludge with an anaerobic side-stream reactor. Water Science and Technology 63 (1), 93e99. Goel, R.K., Noguera, D.R., 2006. Evaluation of sludge yield and phosphorus removal in a cannibal solids reduction process. Journal of Environmental Engineering ASCE 132, 1331e1337. Grady, C.P.L., Daigger, G.T., Lim, H.C., 1999. Biological Wastewater Treatment, second ed. Marcel Dekker, New York, N.Y. Johnson, B.R., Daigger, G.T., Novak, J.T., 2008. The use of ASM based models for the simulation of biological sludge reduction processes. Water Practice and Technology 3, 3e11. Low, E.W., Chase, H.A., 1999. Reducing production of excess biomass during wastewater treatment. Water Research 33, 1119e1132. Novak, J.T., Sadler, M.E., Murthy, S.N., 2003. Mechanisms of floc destruction during anaerobic and aerobic digestion and the effect on conditioning and dewatering of biosolids. Water Environment Research 37, 3136. Novak, J.T., Chon, D.H., Curtis, B., Doyle, M., 2007. Biological solids reduction using the cannibal process. Water Environment Research 79 (12), 2380e2386. Øegaard, H., 2004. Sludge minimization technologies e an overview. Water Science and Technology 49 (10), 31e40. Park, K.Y., Lee, J.W., Ahn, K.H., Maeng, S.K., Hwang, J.H., Song, K. G., 2004. Ozone disintegration of excess biomass and application to nitrogen removal. Water Environment Research 76 (2), 162e167. Park, C., Abu-Orf, M.M., Novak, J.T., 2006. The digestibility of waste activated sludges. Water Environment Research 78, 59e68. Park, C., Helm, R.F., Novak, J.T., 2008a. Investigating the fate of activated sludge exocelluar proteins in sludge digestion using sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE). Water Environment Research 80, 2219e2227. Park, C., Novak, J.T., Helm, R.F., Ahn, Y., Esen, A., 2008b. Evaluation of the extracellular proteins in full-scale activated sludges. Water Research 42, 3879e3889. Saby, S., Djafer, M., Chen, G.-H., 2003. Effect of low ORP in anoxic sludge zone on excess sludge production in oxicsettling-anoxic activated sludge process. Water Research 37, 11e20. Teague, P. 2011. The role of substrate gradient in determining EPS generation, sludge properties, and the anaerobic digestibility of activated sludge. Master’s Thesis. University of Massachusetts, Amherst, MA.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 2 1 e6 0 2 9
US EPA, 1992. Environmental Regulations and Technology: Control of Pathogens and Vector Attraction in Sewage Sludge Under 40 Cfr Part 503 EPA/625/R-92/013, Washington, DC. US EPA, 1999. Report On: BiosoLids generation, use, Disposal in the United States. Office of Solid Waste, US EPA, Washington, DC. EPA530-R-99e09.
6029
Wei, Y., Van Houten, R.T., Borger, A.R., Eikelboom, D.H., Fana, Y., 2003. Minimization of excess sludge production for biological wastewater treatment. Water Research 37, 4453e4467. Yasui, H., Shibata, M., 1994. An innovative approach to reduce excess sludge production in the activated sludge process. Water Science and Technology 30 (9), 11e20.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 3 0 e6 0 3 8
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
A potential approach for monitoring drinking water quality from groundwater systems using organic matter fluorescence as an early warning for contamination events _ ska-Sobecka b, Rasmus Boe-Hansen c, Colin A. Stedmon a,*, Bozena Seredyn b Nicolas Le Tallec , Christopher K. Waul b, Erik Arvin b a
Department of Marine Ecology, National Environmental Research Institute, Aarhus University, Frederiksborgvej 399, 4000 Roskilde, Denmark b Department of Environmental Engineering, Technical University of Denmark, Miljoevej, Building 113, 2800 Kgs. Lyngby, Denmark c Kru¨ger A/S, Gladsaxevej 363, 2860 Søborg, Denmark
article info
abstract
Article history:
The fluorescence characteristics of natural organic matter in a groundwater based drinking
Received 3 February 2011
water supply plant were studied with the aim of applying it as a technique to identify
Received in revised form
contamination of the water supply. Excitationeemission matrices were measured and
29 August 2011
modeled using parallel factor analysis (PARAFAC) and used to identify which wavelengths
Accepted 30 August 2011
provide the optimal signal for monitoring contamination events. The fluorescence was
Available online 10 September 2011
characterized by four components: three humic-like and one amino acid-like. The results revealed that the relative amounts of two of the humic-like components were very stable
Keywords:
within the supply plant and distribution net and changed in a predictable fashion
Organic
depending on which wells were supplying the water. A third humic-like component and an
Fluorescence
amino acid-like component did not differ between wells. Laboratory contamination
Drinking
experiments with wastewater revealed that combined they could be used as an indicator of
Groundwater
microbial contamination. Their fluorescence spectra did not overlap with the other
Contamination
components and therefore the raw broadband fluorescence at the wavelengths specific to their fluorescence could be used to detect contamination. Contamination could be detected at levels equivalent to the addition of 60 mg C/L in drinking water with a TOC concentration of 3.3 mg C/L. The results of this study suggest that these types of drinking water systems, which are vulnerable to microbial contamination due to the lack of disinfectant treatment, can be easily monitored using online organic matter fluorescence as an early warning system to prompt further intensive sampling and appropriate corrective measures. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
Drinking water should be thoroughly controlled to ensure the quality of water and to protect the health of the consumers. Contaminants may appear in drinking water as a result of
failures in the well field, water treatment system or the distribution network (Nyga˚rd et al., 2007; Henderson et al., 2009; Richarson et al., 2009). The “typical” contamination incident is often to be linked to one or several system malfunctions in some cases combined with environmental
* Corresponding author. Present address: National Institute of Aquatic Resources, Technical University of Denmark, Kavalerga˚rden 6, 2020 Charlottenlund, Denmark. Tel.: þ45 35883410. E-mail address: [email protected] (C.A. Stedmon). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.08.066
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 3 0 e6 0 3 8
factors such as flooding. Once pathogenic organisms have entered the drinking water system they are generally (with few exceptions) unable to proliferate, and thus will be subdued due to rapid washout. Contamination incidents are therefore characterized as being stochastic, short-termed and difficult to detect. Fecal material is generally considered to be among the most severe types of contamination, especially when originating from human wastewater. Faeces contain a vast number of pathogens that can have adverse effect on human health even at very low concentrations. The common methods used to measure the microbial quality of drinking water, are based on grab sampling followed by measurement of indicator organisms typically associated with fecal material such as Escherichia coli. (Tallon et al., 2005). The microbial techniques applied (typically heterotrophic plate counts) are slow and expensive even if automated. Additionally the grab sampling approach only allows for a marginal amount of distributed water to be controlled. Today, the chance of actually catching a contamination incident before it reaches the consumers is zero. This calls for new approaches to the challenge of monitoring the hygienic quality of drinking water, allowing for early-warning and rapid response. Monitoring organic matter fluorescence offers low maintenance and rapid measurement with vast information regarding the composition of the organic carbon pool and thus the quality of water. Although the measurement does not offer a direct measurement of the presence or concentration of pathogens in a water supply, it does offer a potentially sensitive approach to detect sudden changes in supply brought about by the intrusion of contaminated water potentially bearing pathogens. Regardless of source, drinking water contains dissolved organic matter (DOM). DOM consists of a complex mixture of organic compounds ultimately originating from the degradation of terrestrial and aquatic organisms (Thurman, 1986; Amy et al., 1987). Little is known about the actual chemical composition of this material as it consists of a vast number of compounds at very low concentrations, presenting a considerable analytical challenge. A range of different techniques have been developed and applied to characterize and quantify DOM in and across different systems according to bulk characteristics such as age, aromaticity, isotopic composition, molecular size distribution and UVevisible optical properties (Thurman, 1986; Chin et al., 1994; McKnight et al., 2001). A fraction of the organic compounds present in DOM fluoresce. This fluorescence offers a rapid and sensitive method for characterizing and tracing organic material in aquatic environments (Coble, 1996). A major advantage with fluorescence spectroscopy is that the instrumentation can be easily adapted for online measurement presenting a method for monitoring quantitative and qualitative changes in organic matter. In this study the suitability of using natural organic matter fluorescence to monitor the quality of drinking water is assessed. The application of fluorescence technique in monitoring sewage contamination in drinking water has been indicated earlier (Henderson et al., 2009). This was based on previous studies, where both wastewater biological oxygen demand, and sewage content of river waters have been found to be correlated to amino acid-like fluorescence (Reynolds and
6031
Ahmad, 1997; Baker, 2001). Organic matter fluorescence has also been used to monitor water treatment in drinking water facilities using surface waters (Baghoth et al., 2008; Bieroza et al., 2009). These studies concern drinking water produced from surface water, which is the most common global potable water source. However, some countries, for example Denmark, rely almost entirely on groundwater as sources for drinking water production. Compared to surface water, groundwater generally requires fewer treatment steps. For example, a typical Danish water treatment plant consists of only two processes: aeration and sand filtration, and does not include post disinfection. Additionally it has been estimated that 10% of the European population receive drinking water from small private supplies often in rural areas (Hulsmann, 2005), which have a higher risk of contamination (Richarson et al., 2009) compared to municipal systems. With a lack of any disinfectant used at the water works and in the distribution system, the water supply is highly vulnerable to microbial contamination. The aim of this study was to characterize the natural variability in organic matter fluorescence in a normal functional groundwater based drinking water facility. The effects of simulated contamination with wastewater on the fluorescence characteristics were then assessed experimentally as a proof of concept of using organic matter fluorescence as an indicator for groundwater drinking water contamination. The hypothesis was that the major part of the normal operational variability in the fluorescence was caused by operational changes in the production scheme, such as changes in the groundwater well configurations, and filter backwashing. Different wells may contain water with slightly different organic matter characteristics due to local conditions. In contrast to surface water supply systems where considerable seasonal variability of organic matter fluorescence is observed, these groundwater based systems are expected to have a more constant concentration and quality of fluorescent organic matter. The characteristics of the organic matter in wastewater are expected to be considerably different and therefore its presence should be easily detected above the normal natural variability. As fluorescence can be measured by online sensors (e.g. Carstea et al., 2010), this approach would be very well suited to remote monitoring of water quality, and can be developed to be used as an early warning alarm for the sampling of additional quality parameters.
2.
Methods
2.1.
Sampling
Samples originated from two sampling campaigns at the Lillevang drinking water works (Farum, Denmark). The well field consists of three wells each with a production capacity of 30 m3/h. The treatment train consists of aeration and pre- and post sand filtration followed by a 120 m3 storage tank (Fig. 1). The first sampling campaign was carried out in 2008 on the 8th, 24th and 30th of April. Samples were taken before and after the filters and at the outlet of the storage tank. In addition sampling was also carried out immediately after a filter backwash event. This was done to capture the effects that this
6032
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 3 0 e6 0 3 8
87 samples from the drinking water plant and its distribution net was modeled and a four component model was split half validated. In addition examination of the unexplained fluorescence revealed that much of the systematic fluorescence in the samples was modeled by these four components and what was left was largely instrument noise (Fig. 2c). TOC concentrations were determined on a Shimadzu TOCV WP analyzer with ASI-V autosampler. The principal of the analyzer’s operation is the use of sodium persulfate solution, UV radiation and temperature of 80 C to oxidize organic carbon. During the analysis, samples are acidified with 17% phosphoric acid to pH 2e3, sparged with nitrogen gas, and oxidized. Carbon dioxide formed in the oxidation process is subsequently quantified in an infrared detector. Fig. 1 e Schematic of the Lillevang drinking water works.
operation might have on the organic matter fluorescence. The second sampling campaign was in 2009 over the period from 27 to 30th April. Samples were taken from the storage tank and the distribution net. During the measurement period drinking water was pumped from each of the three wells individually for approximately 24 h so that the characteristics of the water supplied from each well could be assessed. At 7 am on the 26th water was pumped from well 1 alone. By the start of sampling on the 27th the water in the system at the facility was 100% well 1 water. At 7 am each day the well was switched between 1, 2 and 3 so that the gradual change in water characteristics could be followed at the storage tank and in the distribution net.
2.2.
Analyses
Water samples were collected in acid washed and precombusted (550 C) 40-ml glass vials with Teflon-lined caps. The samples were stored refrigerated (4 C) and analyzed for fluorescence and total organic carbon (TOC) within the following week. Organic matter fluorescence was measured on a Varian Eclipse fluorescence spectrophotometer in a 1 cm quartz cuvette. A series of emission scans from 300 to 600 nm (every 2 nm) were recorded whilst exciting with light from 240 to 450 nm in 5 nm steps. The excitation and emission slit widths were 5 nm. The fluorescence measurements resulted in an excitationeemission matrix for each sample, representing a map of the fluorescence characteristics of the organic matter in the sample (e.g. Fig. 2a). The data was corrected for instrument spectral biases using spectra derived from Rhodamine 101 and a silica diffuser as recommended by the manufacturer. To correct for inner filter effects the approach described by Lakowicz (2006) was applied using absorbance measurements made in a 1 cm quartz cuvette on a Shimadzu UV-2401PC spectrophotometer with Milli Q water as a reference. Finally the fluorescence signal of a Milli Q blank was subtracted and the fluorescence was calibrated to the water Raman signal from excitation at 350 nm as described in Lawaetz and Stedmon (2009), hence the results are given in Raman Units (R.U.). Organic matter fluorescence data was characterized with parallel factor analysis (PARAFAC) using the DOMFluor Toolbox (Stedmon and Bro, 2008). A data set of
2.3.
Wastewater addition experiment
Wastewater additions were performed in the laboratory to assess to what degree wastewater contamination could be detected above the background drinking water fluorescence signal. For this a raw wastewater sample from a large municipal wastewater treatment plant with a capacity of 135,000 person equivalents (Lundtofte, Kgs. Lyngby, Denmark; 80% domestic waste) was taken. The sample was filtered through a GF/C (approximate pore size of 1.6-mm) glass fiber filter. The organic carbon concentration of the filtered wastewater sample was 24 mg C/L. Wastewater additions were made at 0, 0.1, 0.25, 0.5, 0.75, 1 and 2% levels (v/v) in 100 mL volumetric flasks. These corresponded to 0e480 mg C/L wastewater organic matter additions. The fluorescence data from this experiment were characterized by fitting the four-component model derived above for the drinking water DOM fluorescence to the contaminated samples. Examination of the residuals revealed that the fourcomponent model was adequate at describing the additional fluorescence originating from the wastewater at these dilute concentration levels (e.g. Fig. 2c and f). If we had included higher concentrations of wastewater the model would be clearly inadequate as the fluorescence spectra of wastewater organic matter differs notably from that of natural waters (Baker, 2001; Stedmon and Markager, 2005). The PARAFAC model in this study was developed on the pure drinking water samples only, so that the natural variability in organic matter fluorescence could be characterized as best as possible. This would then make it easier to identify contamination events as samples where the fluorescence was not modeled adequately. So not only could one use changes in the relative fluorescence of different components as an indicator of contamination, but also the unexplained fluorescence (residuals). In this study, however, the former was found to be sufficient and therefore focused on in the following section.
3.
Results & discussion
3.1. Fluorescence characterization of natural organic matter in drinking water Fluorescence characteristics of the organic matter in the drinking water were characterized by a broad humic-like
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 3 0 e6 0 3 8
6033
Fig. 2 e Example of the fluorescence properties of dissolved organic matter in two samples. a) Measured, b) modeled and c) residual fluorescence for a sample from the storage tank at Lillevang water works taken on the 29th April. d) Measured, e) modeled and f) residual fluorescence for a drinking water sample with an addition of 240 mg C/L wastewater.
fluorescence similar to that observed in other natural waters (Fig. 2a). The four-component model explained the majority of the fluorescence signal with the unexplained fluorescence (residuals) largely consisting of instrument noise and remaining water scattering effects (e.g. Fig. 2c and f) with an order of magnitude lower signal. The fluorescence characteristics of the four components identified are shown in Fig. 3. Component 1 had an emission maximum at 428 nm and an excitation spectrum with a maximum below 240 nm and a shoulder at 320 nm. Fluorescence in this region is often referred to as the humic-like C-peak and is found in almost all natural waters (Coble, 1996, 2007; Baker, 2001; Jørgensen et al., 2011). Component 2 had a similar shape to component 1 but with excitation and emission maxima at lower wavelengths (315, 384 nm respectively). This overlaps with the region of
humic fluorescence referred to as the M-peak (Coble, 2007). This type of fluorescence was originally thought to only persist in productive oceanic environments but now has also been observed in freshwaters impacted by agriculture (Stedmon and Markager, 2005; Coble, 2007; Jørgensen et al., 2011). Component 3 has a very broad excitation spectrum spanning the whole region measured with two maxima at 270 and 400 nm. The emission spectrum has a maximum at 492 nm. Fluorescence similar to this has been found in a broad range of environments and it is thought to represent terrestrial organic matter from soils (Stedmon et al., 2003). Component 4 had spectral characteristics similar to that of the amino acid tryptophan, which has an excitation maximum at 278 nm and an emission maximum at 348 nm. Fluorescence in this region is often referred to as the T peak (Coble, 1996). It is
6034
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 3 0 e6 0 3 8
Fig. 3 e The spectral characteristics of the four fluorescent components identified by the PARAFAC model.
thought to represent the fluorescence of tryptophan present in protein structures (Yamashita and Tanoue, 2003).
3.2. Daily variability in organic matter fluorescence within the plant and distribution net The organic matter fluorescence characteristics were remarkably stable during the normal operations at the drinking water plant. During the filter backwash, water from the storage tank containing 2e4 h production is flushed back through the sand filters whereby particulate matter (iron- and manganese oxides, bacteria and organic detritus) is removed from the pore space. Only very slight changes were observed in the quantity and quality of fluorescent organic matter during the process. Data for the 24th April 2008 are shown in Fig. 4 as an example and represent samples collected directly after flowing through Filter 2 (Fig. 1). Organic matter fluorescence in the storage tank water used to carry out the backwash was slightly lower and as a consequence levels in the water passing through the filters once normal flow was reestablished were lower during the first 50 min after the flush. A similar drop was also seen in conductivity. After approximately 100 min, organic matter humic-like fluorescence (components 1e3) and conductivity had returned to original levels. The fluorescence of component 4 returned first to original levels after 200 min. TOC concentrations remained constant throughout (3.29e3.35 mg C/L on the 24th April 2008) with no systematic change. During its passage from aeration to the storage tank the organic matter levels showed very little change. Data from the 30th April are shown as an example in Fig. 5 and it is clear that the sand filters did not have any substantial effect on the
organic matter fluorescence characteristics. Similarly TOC concentrations did not vary during passage through the filters. For the samples in Fig. 5 TOC was 3.25, 3.36 and 3.26 mg C/L for after Filter 1, Filter 2 and in the storage tank, respectively.
Fig. 4 e Effects of a filter backwash event carried out on the 24th April 2008 on the organic matter fluorescence properties. The fluorescence intensities of the four PARAFAC components (Fig. 3) are plotted together with the conductivity.
6035
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 3 0 e6 0 3 8
Fig. 5 e Example of measured fluorescence properties and the fluorescence intensity of the four components from the PARAFAC model in three samples from the 30th April 2008. Samples are from after Filter 1, Filter 2 and from the Storage tank (see Fig. 1).
relationships (r2 ¼ 0.83 for both components). This agrees well with the findings of earlier studies on drinking water organic matter. For example Bieroza et al. (2010) found a strong correlation between the removal of organic carbon and fluorescence across a range of drinking water works in the U.K.
3.3. Effects of wastewater contamination on organic matter fluorescence The laboratory additions of waste water at levels <2% v/v increased the fluorescence of components 1, 2 and 4, but had 1.2
F (R.U.)
1 0.8
C1 C2 C3 C4
0.6 0.4 0.2
09 4/
20
09 20 4/
/0 30
09 20 4/
/0 29
09
/0
20 4/ /0 27
28
08 20
08 /0
4/
20 30
4/ /0 24
/0
4/
20
08
0
08
The average fluorescence intensities from different sampling dates in April 2008 and 2009 were compared in order to evaluate the stability of the fluorescence characteristics of drinking water organic matter (Fig. 6). Sampling in 2008 was a combination of samples from the storage tank and samples taken from immediately after the sand filters. During this sampling wells 1 and 2 were in use. The data from 2009 in Fig. 6 represent all samples from the storage tank and sampled approximately 24 h after a well change, averaged. After this period the tanks contain 100% water from the specific well. In general the results show that the fluorescence characteristics of the water from these wells were remarkably constant between the sampling trips despite them ranging from days to a year apart. The sampling in 2009 was focused on assessing the variability associated with well changes. The results for fluorescence and TOC are shown in Fig. 7 and one can see that when the system is flushed completely with water originating from one well, the fluorescence signal in the storage tank and distribution net is constant. Much of the variability measured during the sampling period was attributed to changes in the dominant groundwater well from which water was being pumped and the subsequent flushing of the storage tank and distribution net (Fig. 7). The results show that there is a slight delay before the well change can be detected in the distribution net. It is clear that wells 1 and 2 had very similar fluorescence intensities, whereas the water from well 3 had slightly lower organic matter fluorescence. This difference was only observed for component 1 and 2. The intensities of component 3 and 4 varied very little between well changes. The TOC concentrations mirrored the trends seen for component 1 and 2 (Fig. 7b). Although concentrations varied within a relatively limited range (2.9e3.3 mg C/L), the water from well 3 had lower TOC concentrations than well 1 and 2. Linear regressions between the fluorescence of component 1 and 2 and TOC revealed that there were strong linear
Date
Fig. 6 e Fluorescence intensities of the four components in water from the Lillevang drinking water plant from seven different days in 2008 and 2009. Data in 2008 are daily averages for samples from the storage tank and before and after the sand filters. Values in 2009 are averages of samples taken where the plant has been drawing from one groundwater well only for at least 20 h. Error bars represent standard deviations.
6036
Fl (R.U.)
a 1.2
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 3 0 e6 0 3 8
(1)
(2)
(3)
(1)
C1 ST
1.0
C2 ST
0.8
C3 ST C4 ST
0.6
C1 D
0.4
C2 D
0.2
C3 D C4 D
0.0 00:00
b
4.0
12:00
(1)
00:00
12:00 00:00 Time
(2)
12:00
(3)
00:00
12:00
(1)
TOC (mg C/L)
ST D
3.0 2.0 1.0 0.0 00:00
12:00
00:00
12:00 00:00 Time
12:00
00:00
12:00
Fig. 7 e a) Fluorescence intensities of all four components (C1eC4) during the 2009 sampling. Sampling started on the 27th April and the last sample was taken on the 30th. ST [ storage tank, D [ distribution system. Gray dashed lines indicate groundwater well change at the facility. Numbers in brackets represent the well supplying the water. b) Total organic carbon concentrations for the same samples.
no effect on component 3. The increases observed for components 1 and 2 were very slight with a 2% addition resulting in a 4% and 10% increase respectively. In contrast the fluorescence of component 4 increased by 50% after a 2% addition. As result there was a strong linear relationship between the fluorescence of component 4 and the amount of wastewater organic matter added (Fig. 8). This agrees well with results from surface waters (Baker, 2001) and recycled water systems (Hambly et al., 2010) where tryptophan-like fluorescence has been found to be a good indicator of contamination. In addition, the fact that the fluorescence of
0.16 y = 0.0003x R2 = 0.97
Change in Fl. (R.U.)
0.12
C3
0.08
C4
0.04
0.00 0
100
200
300
400
500
600
-0.04 WW added ( g C /L)
Fig. 8 e The change in fluorescence of component 3 and 4 due to wastewater additions. Additions ranged between 0 and 2% volume and equated to 0e480 mg C/L wastewater added.
component 3 does not change supports the finding that this material represents primarily soil derived fluorescent organic matter. These results show that there is clearly great potential for using organic matter fluorescence to monitor changes in the quality and quantity of organic matter in a drinking water supply. However, as is, the data treatment involved is considerable and far from ideal for simple online systems. These detailed spectral measurements combined with PARAFAC analysis, provide a complete indicator. For this drinking water plant the fluorescence of component 1 and 2 reveals the “normal” fluctuations in organic matter quality expected due to actions such as well changes. The fluorescence of component 4 offers an indicator for waste contamination and component 3 offers a stable baseline to which to relate the changes to. Background levels of component 4 are also relatively low, which is good with regards to sensitivity of contamination detection. This is also supported in part by the results of the triplicate samples taken during the 2009 sampling. In general, the replication was very good for all fluorescent components (replicates are plotted in Fig. 7 and largely lie on top of each other). However, the fluorescence of component 4 appeared to be contaminated in two samples, where one of the triplicates deviated slightly from the other two. These samples originate from the storage tank sampling at the start on the 27th (gray circle, fluorescence of 0.2 R.U.) and on the 29th at 13:10 (gray circle, fluorescence >0.2 R.U.) (Fig. 7). As only one of the triplicates deviate for these samples, sampling contamination is thought to have occurred rather than this being an indicator for contamination at the storage tank.
3.4.
Defining wavelength regions for online detection
The spectral characteristics of component 3 and 4 do not overlap considerably with each other or the other components and their fluorescence can therefore be approximated from broadband fluorescence measurements. This implies that a simpler instrumental set up could be designed, where just two specific fluorescence wavebands are monitored. One could start with just monitoring the fluorescence in the region of component 4, as this is where the correlation to contamination was present. However, there is a benefit of using two channels and normalizing the signals to each other. Normalizing the fluorescence to a relatively stable background fluorescence signal avoids problems with variations in light source intensity. The output of a light source in an in situ fluorometer can be difficult to maintain constant. With use the signal can gradually decrease and temperature fluctuations also influence lamp output. To test this approach the spectral data in two regions were averaged to simulate a fluorometer with a single UV-B excitation channel (270e290 nm) and two emission channels measuring UV-A fluorescence (325e350 nm) and visible fluorescence (480e505 nm). The UV-A fluorescence was normalized to the visible fluorescence signal. Fig. 9 shows the ratio for the data from the storage tank and distribution net sampling in 2009. Besides the two outlier samples discussed earlier, the ratio is very constant and exhibits little variability irrespective of
6037
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 3 0 e6 0 3 8
0.6
ST D
0.5
Ratio
0.4 0.3 0.2 0.1 0.0 00:00
12:00
00:00
12:00
00:00
12:00
00:00
12:00
Time Fig. 9 e The ratio of the average UV-A fluorescence (325e350 nm) to the visible fluorescence (480e505 nm) from UVB excitation (270e290 nm) for the same samples shown in Fig. 7.
well changes and sampling site (storage tank vs. distribution net). Average ratio values are 0.35 (St. dev. 0.02). The ability of the ratio to predict the wastewater contamination is shown in Fig. 10. The estimate for the lowest standard (24 mg C/L addition) is within two standard deviations of the mean ratio for the 2009 sampling and therefore not detectable during normal operating conditions. From this data it is clear that contamination can be detected down to levels equivalent to 60 mg C/L which is approximately 2% of the average TOC concentration in the analyzed drinking water samples (3.3 mg C/L). This content of wastewater TOC in drinking water is far below that can be detected by the TOC technique (Henderson et al., 2009), and it can be done rapidly, ideally within seconds, by an online fluorescence sensor. These findings support the arising consensus that organic matter fluorescence has great potential as an indicator of water quality in drinking water systems (Bieroza et al., 2009; Henderson et al., 2009; Hambly et al., 2010).
4.
Conclusions
The concept of monitoring organic matter fluorescence proved to be sensitive and specific enough for wastewater contamination of groundwater based drinking water. This work paves the way for further development of in situ sensors for early warning of contamination incidents in drinking water systems. The detailed PARAFAC analysis provided a suitable method to reveal which wavelengths are the most responsive and considerably reduce the data required. Based on the obtained results, only one excitation and two emission wavelengths are sufficient to develop a fluorescence sensor detecting wastewater contamination in water. The approach has great potential as an online indicator parameter at several locations on the distribution net, which can be used as an early warning system for a contamination event and prompt more intensive grab sampling.
600
Acknowledgments
y = 1341.7x - 490.31
WW added ( g C /L)
500
2
R = 0.9808
The authors thank Lillevang water works for allowing access to the plant and helping with the sampling. This work was funded by the Danish Agency for Spatial and Environmental Planning, Ministry of the Environment (Ref. BLS-403-00043 AquaFingerprint).
400 300 200 100 0 0.20
references 0.40 0.60 Fluorescence ratio
0.80
Fig. 10 e The ability of the broadband fluorescence ratio to predict wastewater contamination. The gray solid line represents the mean ratio for samples taken from the storage tank and in the distribution net in 2009 (data in Fig. 6) and the dashed lines represent limits of ±two standard deviations.
Amy, G.L., Collins, M.R., Kuo, C.J., King, P.H., 1987. Comparing gelpermeation chromatography and ultrafiltration for the molecular-weight characterization of aquatic organic-matter. Journal of the American Water Works Association 79 (1), 43e49. Baghoth, S.A., Maeng, S.K., Salinas Rodriguez, S.G., Ronteltap, M., Sharma, S., Kennedy, M., Amy, G.L., 2008. An urban water cycle perspective of natural organic matter (NOM): NOM in drinking water, wastewater effluent, storm water, and
6038
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 3 0 e6 0 3 8
seawater. Water Science and Technology: Water Supply 8 (6), 701e707. Baker, A., 2001. Fluorescence excitationeemission matrix characterization of some sewage impacted rivers. Environmental Science and Technology 35, 948e953. Bieroza, M.Z., Baker, A., Bridgeman, J., 2009. Relating freshwater organic matter fluorescence to organic carbon removal efficiency in drinking water treatment. Science of the Total Environment 47, 1765e1774. Bieroza, M.Z., Bridgeman, J., Baker, A., 2010. Fluorescence spectroscopy as a tool for determination of organic matter removal efficiency at water treatment works. Drinking Water Engineering and Science 3, 63e70. Carstea, E.M., Baker, A., Bieroza, M., Reynolds, D., 2010. Continuous fluorescence excitationeemission matrix monitoring of river organic matter. Water Research 44, 5356e5366. Chin, Y.P., Aiken, G., O’Loughlin, E., 1994. Molecular weight, polydispersity, and spectroscopic properties of aquatic humic substances. Environmental Science and Technology 28, 1853e1858. Coble, P.G., 1996. Characterisation of marine and terrestrial DOM in seawater using excitationeemission matrix spectroscopy. Marine Chemistry 51, 325e346. Coble, P.G., 2007. Marine optical biogeochemistry: the chemistry of ocean color. Chemical Reviews 107, 402e418. Hambly, A.C., Henderson, R.K., Storey, M.V., Baker, A., Stuetz, R.M., Khan, S.J., 2010. Fluorescence monitoring at a recycled water treatment plant and associated dual distribution system e implications for cross-connection detection. Water Research 44, 5323e5333. Henderson, R.K., Baker, A., Murphy, K.R., Hambly, A., Stuetz, R.M., Khan, S.J., 2009. Fluorescence as a potential monitoring tool for recycled water systems: a review. Water Research 43, 863e881. Hulsmann, A., 2005. Small Systems Large Problems: A European Inventory of Small Water Systems and Associated Problems. Web-based European Knowledge Network on Water. http://www.nccph.ca/docs/05_small_water_systems_ ver_june2005.pdf. Jørgensen, L., Stedmon, C.A., Kragh, T., Markager, S., Middelboe, M., Søndergaard, M., 2011. Global trends in the fluorescence characteristics and distribution of marine
dissolved organic matter. Marine Chemistry. doi:10.1016/j. marchem.2011.05.002. Lakowicz, J.R., 2006. Principles of Fluorescence Spectroscopy, third ed. Springer, Baltimore, Maryland. Lawaetz, A.J., Stedmon, C.A., 2009. Fluorescence intensity calibration using the Raman Scatter peak of water. Applied Spectroscopy 63, 936e940. McKnight, D.M., Boyer, E.W., Westerhoff, P.K., Doran, P.T., Kulbe, T., Andersen, D.T., 2001. Spectrofluorometric characterisation of dissolved organic matter for indication of precursor organic material and aromaticity. Limnology and Oceanography 46, 38e48. Nyga˚rd, K., Wahl, E., Krogh, T., Tveit, O.A., Bøhleng, E., Tverdal, A., Aavitsland, P., 2007. Breaks and maintenance work in the water distribution systems and gastrointestinal illness: a cohort study. International Journal of Epidemiology 36, 873e880. Reynolds, D.M., Ahmad, S.R.A., 1997. Rapid and direct determination of wastewater BOD values using a fluorescence technique. Water Research 31 (8), 2012e2018. Richarson, H.Y., Nichols, G., Lane, C., Lake, I.R., Hunter, P.R., 2009. Microbiological surveillance of private water supplies in England e the impact of environmental and climate factors on water quality. Water Research 43, 2159e2168. Stedmon, C.A., Bro, R., 2008. Characterizing dissolved organic matter fluorescence with parallel factor analysis: a tutorial. Limnology and Oceanography: Methods 6, 572e579. Stedmon, C.A., Markager, S.S., 2005. Resolving the variability in dissolved organic matter fluorescence in a temperate estuary and its catchment using PARAFAC analysis. Limnology and Oceanography 50 (2), 686e697. Stedmon, C.A., Markager, S., Bro, R., 2003. Tracing dissolved organic matter in aquatic environments using a new approach to fluorescence spectroscopy. Marine Chemistry 82, 239e254. Tallon, P., Magajna, B., Lofranco, C., Leung, K.T., 2005. Microbial indicators of faecal contamination in water: a current perspective. Water, Air and Soil Pollution 166 (1), 139e166. Thurman, E.M., 1986. Organic Geochemistry of Natural Waters. Martinus Nijhoff/Dr. W. Junk Publishers, The Netherlands. Yamashita, Y., Tanoue, E., 2003. Chemical characterization of protein-like fluorophores in DOM in relation to aromatic amino acids. Marine Chemistry 82, 255e271.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 3 9 e6 0 5 0
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Effects of silver nanoparticles on wastewater biofilms Zhiya Sheng, Yang Liu* Department of Civil and Environmental Engineering, University of Alberta, Edmonton, AB T6G 2W2, Canada
article info
abstract
Article history:
The goal of this research is to understand the potential antibacterial effect of silver
Received 13 May 2011
nanoparticles (Ag-NPs) on biological wastewater treatment processes. It was found that
Received in revised form
original wastewater biofilms are highly tolerant to the Ag-NP treatment. With an
5 August 2011
application of 200 mg Ag/L Ag-NPs, the reduction of biofilm bacteria measured by
Accepted 30 August 2011
heterotrophic plate counts was insignificant after 24 h. After the removal of loosely bound
Available online 10 September 2011
extracellular polymeric substances (EPS), the viability of wastewater biofilms was reduced when treated under the same conditions. By contrast, when treated as planktonic pure
Keywords:
culture, bacteria isolated from the wastewater biofilms were highly vulnerable to Ag-NPs.
Silver nanoparticles (Ag-NPs)
With a similar initial cell density, most bacteria died within 1 h with the application of
Antibacterial effects
1 mg Ag/L Ag-NPs. The results obtained here indicate that EPS and microbial community
Wastewater biofilms
interactions in the biofilms play important roles in controlling the antimicrobial effects of
Planktonic bacteria
Ag-NPs. In addition, slow growth rates may enhance the tolerance of certain bacteria to
Extracellular polymeric substances
Ag-NPs. The effects of Ag-NPs on the entire microbial community in wastewater biofilms
(EPS)
were analyzed using polymerase chain reactionedenaturing gradient gel electrophoresis,
PCReDGGE
PCReDGGE. The studies showed that the microbial susceptibility to Ag-NPs is different for each microorganism. For instance, Thiotrichales is more sensitive to Ag-NPs than other biofilm bacteria. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
Because of their antimicrobial property, silver nanoparticles (Ag-NPs) have become the most frequently used nanoparticles in consumer products. By August 2009, there were 259 consumer products containing nano-silver, which ranked first in the “Woodrow Wilson International Centre for Scholars” study on emerging nanotechnologies (2009). Silver coatings have been widely used to treat chronically infected wounds (Mcinroy et al., 2009), and to prevent biofilm formation on home appliances such as washing machines. Ag-NP coatings have also been introduced as antimicrobial agents in fabrics (Benn and Westerhoff, 2008). Ag-NP applications have been extensively studied as disinfectants in medical institutions, and an increasing amount of research has been carried out on
Ag-NP applications in the food industry and for drinking water treatment and distribution systems (Konopka et al., 2009; Kumar and Raza, 2009; Silvestry-Rodriguez et al., 2008; Zhao et al., 2010). The explosion of nanotechnology applications makes it inevitable that Ag-NPs will be released into domestic and industrial waste streams (Benn and Westerhoff, 2008; Blaser et al., 2008; Geranio et al., 2009; Hagendorfer et al., 2010; Mueller and Nowack, 2008). Because Ag-NPs are meant to exert toxic effects on microorganisms, their release into wastewater systems may adversely affect the microbial communities found in biological treatment processes. Released Ag-NPs could decrease the effectiveness of contaminant removal in biological treatment processes and cause noncompliance with effluent discharge limits. The
* Corresponding author. Tel.: þ1 780 492 5115; fax: þ1 780 492 0249. E-mail address: [email protected] (Y. Liu). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.08.065
6040
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 3 9 e6 0 5 0
antimicrobial effects of Ag-NPs can be attributed to their capacity to disturb cell membrane functions, to bind with intracellular material such as protein and DNA, and to release Agþ ions (Morones et al., 2005). The released Agþ ions can attack the thiol groups in enzymes, stop DNA replication and cause the cells to reach a nonculturable state and then cell death (Morones et al., 2005). These mechanisms are closely related to the small size and high surface/volume ratio of AgNPs (Marambio-Jones and Hoek, 2010), which equips Ag-NPs with highly active facets and increased catalytic activity. While there are increasing concerns regarding human and animal exposure to nanoparticles as emerging contaminants, little is known about the impact of nanoparticles on biological treatment processes in wastewater treatment plants. Most research has focused on the impact of Ag-NPs on individual or certain types of bacteria cultivated under laboratory conditions. However, the impact of Ag-NPs on wastewater microorganisms is not well understood. The few reported studies on nanoparticle toxicity in wastewater biological processes suggest that Ag-NPs could significantly inhibit both heterotrophic and autotrophic wastewater microorganisms (Choi et al., 2009; Choi and Hu, 2008). Recent studies showed that 1 mg Ag/L Ag-NPs inhibited the growth of laboratorycultivated autotrophic nitrifying bacteria by approximately 80% in laboratory-controlled reactors (Choi et al., 2008). Nevertheless, these reported studies focused only on the detrimental effects of Ag-NPs on the total number of planktonic (free-floating) bacteria. To reveal the impact of Ag-NPs in wastewater treatment plants, detailed information regarding nanoparticle toxicity on individual species and microbial communities is needed since different groups of microorganisms are associated with different biological treatment properties and functionalities. For instance, the nitrification process requires both ammonium-oxidizing nitrifiers (e.g., Nitrosomonas) and nitrite oxidizers (e.g., Nitrospira and Nitrobacter). Elimination of these microorganisms will lead to reduced nitrogen removal (Choi et al., 2008). To date, little research has been carried out to evaluate the impact of Ag-NPs on microbial communities found in wastewater treatment plants. In addition, microorganisms in biological wastewater treatment processes are usually in the form of microbial aggregates, such as biofilms in rotating biological contactors (RBCs) and trickling filters. For instance, RBCs are widely used in treatment plants of different scales because of their compact design (high biomass concentration per volume), high pollutant removal efficiency (especially under aerobic conditions), and simple operation (no sedimentation, thus no need for sludge recirculation) (Cortez et al., 2008). Microbial biofilms are highly stratified microbial communities embedded in a matrix of extracellular polymeric substances (EPS) on solid substrata. Previous studies have shown that microbial biofilms are more tolerant to antimicrobial agents than are planktonic bacteria (Davies, 2003; Liu et al., 2007; Rittmann and Mccarty, 2001). The physicochemical microenvironments within biofilms also play important roles in shaping the microbial community structure (Hall-Stoodley et al., 2004). Therefore, the antibacterial effect of Ag-NPs on wastewater biofilms may be significantly different from the effects of Ag-NPs on planktonic cells. Unfortunately,
knowledge regarding the fate and reactivity of Ag-NPs in complex systems such as biofilms is still lagging. Our hypotheses are (i) Ag-NPs can impact wastewater biofilm microbial community structures, depending on the characteristics of each strain, e.g., its ability to produce EPS and growth rate, and the community interactions among these strains; and (ii) the effects of Ag-NPs on planktonic cells are different than on wastewater biofilms. To assess these hypotheses, both original wastewater biofilms and isolated planktonic pure culture bacteria from the biofilms were tested under Ag-NP treatment. Possible protective mechanisms in the biofilm were investigated, such as physical exclusion due to the effects of EPS. The role of community interactions was also studied, and an artificially mixed community was tested to verify the effects of the community interaction. 16s rRNA gene based polymerase chain reactionedenaturing gradient gel electrophoresis (PCReDGGE) was used to analyze the microbial community shift after Ag-NP treatment. Three terms have been used to describe the responses of the samples to Ag-NPs: ‘tolerance’ is defined as the ability of the samples to survive under the treatment of Ag-NPs, while ‘susceptibility’ (or ‘sensitivity’) is defined as their ability to react to Ag-NPs.
2.
Materials and methods
2.1.
Wastewater biofilm samples
Wastewater biofilms were collected from the first-stage RBC unit in the Devon Wastewater Treatment Plant located in Devon, Alberta, Canada. The average biofilm thickness was 1.5 mm. Biofilm samples were cut with the plastic substratum just before each experiment, kept in a Petri dish on ice during transport, and processed within 30 min of arrival at the laboratory.
2.2. Isolation and cultivation of heterotrophic bacteria from wastewater biofilms To isolate cultivable heterotrophic bacteria from wastewater biofilms, single colonies on Reasoner’s 2A (R2A) agar plates were isolated and transferred to new plates based on their appearance and growth rate (indicated by the time needed to form visible colonies on R2A agar plates). Single colonies on R2A agar plates were then inoculated into R2A broth and shaken at 100 rpm at room temperature (25.5 C) for 30 h before further studies on planktonic pure culture bacteria. Two types of mixed planktonic wastewater bacterial cultures (i.e., biofilm mixture and artificial mixture) were prepared. The ‘biofilm mixture’ was cultured by directly inoculating wastewater biofilm into R2A broth. The ‘artificial mixture’ was generated by isolating one single colony of each isolated strain from R2A plates and inoculating them into one R2A broth. These liquid R2A cultures of planktonic wastewater bacteria were shaken at 100 rpm at room temperature (25.5 C) for 30 h before the toxicity tests. R2A media are low-nutrient media often used to recover bacteria from environmental samples. Difco R2A agar powder was purchased from Voigt Global
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 3 9 e6 0 5 0
Distribution Inc., KS, and 2 R2A broth was obtained from Teknova, CA.
2.3.
Removal of loosely bound EPS from biofilms
Among the biofilm EPS components, those that can be readily removed are defined as ‘loosely bound EPS’, while those that need vigorous removal processes are defined as ‘tightly bound EPS’. Extraction reagents such as ethanol, which can damage bacterial cells, are often applied to remove tightly bound EPS. To eliminate loss of microbial diversity and viability, only loosely bound EPS were removed in this study following the procedure described by Gong et al. (2009) as briefly summarized here. Biofilms were scraped off the plastic RBC substratum and suspended in 1% phosphate buffered saline (PBS, 1.65 mM ionic strength). A 30-s vortex was performed to mix biofilm fragments with PBS. The biofilm suspension was vortexed at the maximum speed for 1 min, then centrifuged at 4 C, 4000 g, for 20 min. The pellets were resuspended in 10 mL of 1% PBS, vortexed, and centrifuged again. Pellet resuspension, vortexing, and centrifugation were repeated three times. 1% PBS was used to provide a pH (7.0) and chloride concentration (51.4 mg/L) comparable to the first-stage wastewater (neutral pH and 58.0 mg/L chloride) at the Devon Wastewater Treatment Plant. 1% PBS was prepared by dissolving 10.93 mg/L Na2HPO4, 3.175 mg/L NaH2PO4$H2O, and 84.75 mg/L NaCl in ultra-pure water (PURELAB Maxima system, ELGA LabWater, Mississauga, Canada).
2.4.
Preparation of Ag-NP suspensions
Self-dispersing silver nanopowder was purchased from SkySpring Nanomaterials, Inc. (Houston, USA). According to the Ag-NP product description, the particle size is less than 15 nm, and the particle composition is 10% silver (99.99% purity) and 90% polyvinylpyrrolidone (PVP), similar to Ag-NPs commonly used in commercial products. Ag-NP suspensions at concentrations of 1, 50, or 200 mg Ag/L were prepared by dispersing Ag-NPs in 1% PBS and mixed by vortex at the maximum speed.
2.5.
Characterization of Ag-NPs
The particle size distribution of Ag-NPs was characterized using a Malvern Zetasizer Nano-ZS (Model: ZEN3600, Malvern Instruments Ltd, Worcestershire, UK). Since PVP dissolves in water completely, parameters of silver were adopted for the analysis: the refractive index was 2.0 and the absorption coefficient was 0.320 (Sur et al., 2010). A suspension of 1 mg Ag/L Ag-NPs in ultra-pure water was also prepared for particle size analysis. All the 24-h-old suspensions were prepared by shaking the sample at 100 rpm for 24 h in the dark at room temperature (25.5 C), which is the same as that what was done in the toxicity tests.
2.6.
Ag-NP toxicity experiments
Antimicrobial effects of Ag-NPs were tested on original wastewater biofilms, wastewater biofilms with loosely bound EPS removed, and planktonic bacteria isolated from wastewater biofilms. Biofilms with and without loosely bound EPS
6041
were suspended in freshly prepared Ag-NP suspensions (1 g wet original biofilm/5 mL Ag-NP suspension) at 0, 1, 50, and 200 mg Ag/L in sterile glass test tubes. The impact of Ag-NPs on pure and mixed cultures of planktonic bacteria isolated from the wastewater biofilms was tested by adding the appropriate bacterial culture into 0 (no-treatment control) and 1 mg Ag/L Ag-NP suspensions (initial cell density: about 1 107 CFU/mL). Samples and control were shaken at 100 rpm for 24 h in the dark at room temperature (25.5 C). Viability of cultivable bacteria was examined using a heterotrophic plate count (HPC) every 2 h during the first 12 h and every 4 h after 12 h. All toxicity tests were carried out in triplicate. As a comparison, toxicity of Agþ ions has also been tested at 200 mg Ag/L using AgNO3.
2.7.
Sorption of Ag-NPs to wastewater biofilms
1 g original biofilm samples were added into 5 mL freshly prepared Ag-NP suspensions at different concentrations (0, 1, 20 or 50 mg Ag/L) and incubated in the dark at 100 rpm and room temperature (25.5 C) for 520 min. 520 min was chosen because saturation was reached in 520 min of incubation and the concentration of free Ag-NPs in the suspension did not change significantly after that. The sample with 0 mg Ag/L was tested as the control. Samples without biofilms were also tested as abiotic controls. To evaluate the sorption of Ag-NPs to wastewater biofilms during the experimental period, aliquots of the Ag-NP suspensions were scanned periodically to obtain absorption spectra from 250 to 700 nm using a Cary 50 Bio UVevis spectrophotometer (Varian, USA). Concentration of Ag-NPs in the suspension is directly proportional to the peak absorption at 400 nm on the spectrum (Petit et al., 1993). A decrease of Ag-NPs in the residual suspension can be attributed to their sorption into the biofilm matrix. Concentrations of total silver in biofilm and in liquid suspension were measured by inductively coupled plasma mass spectrometry (ICP-MS) using the ELAN 9000 ICP mass spectrometer (PerKinElmer, Canada). Microwave digestion was performed as described by Wu et al. (1997) and briefly summarized here. 10 mL concentrated nitric acid and 2 mL ultra-pure water were added to 1 g biofilm (wet weight) or 1 mL suspension and kept at room temperature overnight for pre-digestion. Microwave digestion was then carried out using ETHOS EZ Microwave Solvent Extraction Labstation (Milestone Inc., USA) with the following heating program: heat to 190 C within 15 min and then hold at 190 C for 10 min.
2.8.
Bacterial enumeration using HPC
Bacterial enumeration was performed by HPC using the drop plate method (Liu et al., 2007; Zelver et al., 1999). A series of 10fold dilutions were performed and 10 mL of each dilution was plated on R2A agar in triplicate. Plates were incubated at 31 C for 24 h and held at room temperature for another three days. Counting was performed after 24 h (for fast-growing bacteria) and again after the four-day period (for total number of bacteria). The lower detection limit is 102 CFU/mL. For biofilm samples, the result was converted into CFU/cm2 based on the area of each biofilm sample. T-tests were performed in Microsoft Excel 2007 to examine the statistical significance of
6042
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 3 9 e6 0 5 0
the results, and corresponding p-values were calculated using Type 3 two-tailed T-test (unequal standard deviations). A pvalue less than 0.05 indicates a statistically significant difference.
2.9.
Biofilm community analysis using PCReDGGE
A fragment of 16s rRNA gene was analyzed to identify the microbial communities in the original wastewater biofilm, the wastewater biofilm with loosely bound EPS removed, as well as pure and mixed culture planktonic bacteria isolated from wastewater biofilms. Details on the PCReDGGE experiment are provided in supplementary material and briefly described here. For each sample, genomic DNA was extracted and a w550 bps fragment of 16s rRNA gene from each DNA sample was amplified. The primers were chosen according to Yu and Morrison (2004). The same amount of PCR products (600 ng) was loaded on the DGGE gel for each sample. Selected bands were retrieved from the DGGE gel and sequenced. Each sequence was matched against the NCBI nr nucleotide database using the nucleotide BLAST program. Multiple sequence alignment was built using CLUSTAL W and a neighbor-joining phylogenetic tree was calculated and constructed using TREECON (van de Peer and de Wachter, 1994). Known strains were also included in the tree for reference. Details about RBC in the Devon Wastewater Treatment Plant, sample processing and biofilm community analysis using PCReDGGE are provided in supplementary material. Supplemental experiments on biofilms with the plastic substratum are also described in the supplementary material.
3.
Results
3.1.
Characterization of Ag-NPs
As shown in Fig. 1, most Ag-NPs were smaller than 5 nm in freshly prepared water suspensions at a low concentration of 1 mg Ag/L. However, the peak of the size distribution curve moved to 10e15 nm in freshly prepared 1% PBS suspensions at
the same concentration, which indicates Ag aggregate formation under increased ionic strength and the presence of chloride in the 1% PBS suspension. After the 1 mg Ag/L suspension was stored at room temperature for 24 h, even more aggregation was observed. The mode size was about 50 nm, and 84% of the particles were in the range of 33e59 nm. In the 1% PBS suspension at 200 mg Ag/L, no significant difference in particle size was detected between newly prepared and 24 h suspensions. Most Ag-NPs were larger than 20 nm when the concentration was as high as 200 mg Ag/L in 1% PBS suspension. In freshly prepared suspensions, 76% of the particles were in the range of 26e106 nm, and a noticeable proportion of the particles were larger than 200 nm. After 24 h incubation at room temperature, Ag-NP sizes became more uniform. 88% of the particles were in the range of 26e106 nm, while less than 2% were over 200 nm. Similar size distributions under environmental conditions have been reported (Fabrega et al., 2009).
3.2.
Based on the DGGE bands from the original biofilm and isolated single strains, a total of 14 strains were identified in the wastewater biofilm, which fell into three phyla as shown in Fig. 2A. Six of the eight bands on Fig. 2B (WWBF-BeWWBF-G, corresponding to Bands BeG respectively) from the original biofilm were sequenced successfully, as well as eight bands (WWBF-1eWWBF-8) from isolated bacterial strains, which were also loaded onto the same DGGE gel and each formed a single band, confirming the purity (data not shown). Identities between each sequence and the corresponding closest homolog from the database are also shown on Fig. 2A. The closest homologs to WWBF-C, WWBF-D, WWBF-F are Thiothrix fructosivorans, Acidovorax defluvii and Rhodoferax antarcticus respectively. However, the identities are not high enough to confidently identify these three strains due to the presence of too many ‘N’s in their sequences. Bands A and H were mixtures of DNA from different strains and were not sequenced successfully. According to the DGGE bands of pure culture bacteria, Band A appeared to be a mixture of WWBF1eWWBF-4, all of which fell into the same phylum (pairwise identities 82%). In fact, WWBF-2 and WWBF-4 share a sequence identity as high as 97% and fell into the same phylum of Bacteroidetes. Only WWBF-G was clearly missing in the R2A media enriched biofilm culture, although the intensity of certain bands was slightly lower. This indicates that the majority of the microbial community could be successfully maintained using R2A media. WWBF-C maybe belongs to the same order of Thiotrichales as WWBF-G based on the information currently available here.
3.3.
Fig. 1 e Size distribution of freshly prepared and 24-h-old Ag-NPs in 1% PBS and water suspensions.
Microbial communities in wastewater biofilms
Physical protections in wastewater biofilms
Biofilm EPS are comprised of polysaccharides, proteins, phospholipids, humic substances, and nucleic acids (Kumar et al., 2011). To some extent, EPS perform as a physical barrier keeping Ag-NPs from reaching the cells. Fig. 3A and B shows the scanning electron microscopy (SEM) images of wastewater biofilms with and without loosely bound EPS, respectively. In Fig. 3A, it can be seen that bacteria in the original wastewater
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 3 9 e6 0 5 0
6043
Fig. 2 e Community profile of wastewater biofilms: (A) phylogenetic tree based on DGGE bands and isolated single strains (Pairwise identities between neighbor branches are indicated on the tree); (B) DGGE profiles of the original biofilm sample (Lane 1) and the R2A media enriched biofilm culture (Lane 2).
biofilm were immersed in compact extracellular slimes. In contrast, biofilm cells are much more exposed after the removal of loosely bound EPS (as shown in Fig. 3B), although tightly bound EPS were still present. SEM sample preparation process may produce some artifacts on the samples. However, both samples went through the same procedure to make sure that they are comparable with each other. As described in the supplemental material, a method that can effectively preserve the macrostructure for hydrated biological samples was used to prepare the SEM samples in this study. During the incubation of biofilms in Ag-NP suspensions, an evident decrease of free Ag-NPs in the suspension was observed within the first 45 min, but did not change much in further incubation. At an initial concentration of 20 mg Ag/L, a sharp drop from 1.66 to 0.78 in the absorbance of the suspension was detected within 45 min as shown in Fig. 3C. This result shows that Ag-NPs can be sorbed onto or into the biofilm. However, this does not necessarily mean that these sorbed Ag-NPs can reach the cell. Further, after 520 min
incubation, the concentration of free Ag-NPs in the suspension did not change significantly, indicating a saturation of Ag-NPs in the biofilm after 520 min. Mass balance at 45 min with an initial concentration of 20 mg Ag/L was shown in Table 1. According to the ICP results, only about 10% of Ag-NPs were sorbed to the biofilm within 45 min when the initial concentration was 20 mg Ag/L.
3.4.
Effects of Ag-NPs on original wastewater biofilms
Fig. 4A and C shows the viability of the cultivable heterotrophic bacteria in the wastewater biofilm. Colony-forming units (CFU)/mL were counted after allowing the bacteria to grow on R2A agar at 31 C for 24 h (Fig. 4A), indicating the viability of species that grow rapidly. The number of CFU/mL counted 4 days after bacteria were plated (Fig. 4C) indicates the total number of viable bacteria. Without Ag-NPs, the HPC was maintained at about 2 108 CFU/cm2 during the 24-h treatment for fast-growing bacteria, and at about 4 108 CFU/cm2
6044
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 3 9 e6 0 5 0
Fig. 3 e Physical protections in wastewater biofilms: (A) SEM image of original wastewater biofilms; (B) SEM image of wastewater biofilms with loosely bound EPS removed. Bar size: 5 mm. (C) peak absorption (400 nm) of remaining Ag-NPs in suspensions during incubation with wastewater biofilms at various initial concentrations (0, 1, 20, 50 mg Ag/L).
for the total cultivable heterotrophic bacteria. No significant growth was observed, since no nutrients were provided during the experiments.
After a 24-h treatment with Ag-NPs at 200 mg Ag/L, no significant change ( p ¼ 0.68) was detected in the viability of cultivable heterotrophic bacteria. This is consistent with the
Table 1 e Sorption of Ag-NPs to wastewater biofilms. Biotic samplea
Sample Biofilm Total silver (mg) Sum (mg) Mass balance
Abiotic controlb
Residual suspension
7.38 68.48 75.86 Biofilm þ Residual ¼ 105.7% (Abiotic control)
71.76 71.76
a Biotic sample was prepared by adding 1 g wastewater biofilm into 5 mL of Ag-NP suspension (initial concentration: 20 mg Ag/L), and the total silver in the biofilm and in the residual suspension was measured separately after 45 min of incubation at 100 rpm and room temperature (25.5 C). b Abiotic control was the 5 mL of Ag-NP suspension without wastewater biofilms (initial concentration: 20 mg Ag/L). The total silver in the suspension was measured after 45 min of incubation at 100 rpm and room temperature (25.5 C).
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 3 9 e6 0 5 0
6045
Fig. 4 e Effects of Ag-NPs (200 mg Ag/L) on wastewater biofilms with (A, C, and E) and without (B, D, and F) loosely bound EPS, compared with the no-treatment control (error bars represent one standard deviation): (A), (B) growth of heterotrophic bacteria 24 h after plating; (C), (D) growth of heterotrophic bacteria 4 days after plating; (E), (F) DGGE profile (C represents the no-treatment control, while T represents a biofilm community under 200 mg Ag/L Ag-NP treatment. Missing bands are marked with arrows).
results where the biofilms were not scraped off the substratum (Table S1), indicating that scraping off the biofilms does not have significant impact on the tolerance of biofilms to Ag-NPs under the conditions tested. Treated samples were not as stable as controls, but the difference in HPC between treated samples and no-treatment controls almost never exceeded 0.5 log units. Ag-NP treatment of 1 and 50 mg Ag/L had no pronounced impact on the survival of biofilm microorganisms (Fig. S1). Neither did bulk Agþ at 200 mg Ag/L (Fig. S2). In addition, as shown in Fig. 4E, no significant difference in community profile was detected between samples with and
without Ag-NP (200 mg Ag/L) treatment. The slightly higher intensities of several bands in untreated samples may indicate a slight decrease in the viability of these genera under Ag-NP treatment. These results indicate that wastewater biofilms with original EPS are highly tolerant to Ag-NPs.
3.5. Effects of Ag-NPs on wastewater biofilms with loosely bound EPS removed As shown in Fig. 4B and D, the density of fast-growing bacteria in the no-treatment control stabilized at about 1.7 108 CFU/
6046
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 3 9 e6 0 5 0
cm2 during the 24-h period, and the total HPC stabilized at about 2.4 108 CFU/cm2. Removal of loosely bound EPS caused only a slight loss of biofilm bacteria (less than 0.2 log units in the HPC). After the removal of loosely bound EPS, bacteria were more vulnerable to the treatment of Ag-NPs, which is consistent with reported protective effects of EPS on biofilm bacteria (Gong et al., 2009). A decrease in the HPC was detected when biofilms with loosely bound EPS removed were treated with Ag-NPs at 200 mg Ag/L. A comparison of Fig. 4B and D indicates that Ag-NPs were more toxic to fast-growing bacteria. The maximum HPC reduction for fast-growing bacteria was 1.6 log units ( p ¼ 0.02), observed 4 h after Ag-NP treatment, while loss in the total HPC never exceeded 1.0 log unit. However, after 4 h treatment, the concentration of heterotrophic bacteria in the treated samples started to recover, and the discrepancy was less than 1.0 log unit after 24 h. No significant decrease was detected for Ag-NP treatments of 1 and 50 mg Ag/L (Fig. S1). Further, our study showed that 200 mg Ag/L Agþ was more toxic than Ag-NP (Fig. S2), indicating that the release of Agþ might play an important role in the toxicity of Ag-NP to wastewater biofilms and it may be easier for Agþ ions to reach and enter the cells compared to Ag-NPs. However, in our Ag-NP treatment studies, the toxicity of Agþ cannot be differentiated from the overall toxicity of Ag-NPs. Removal of loosely bound EPS did not result in any missing bands, according to the no-treatment controls shown in Fig. 4E and F, but the intensity of several bands decreased slightly. However, without protection from the intact EPS matrix, Ag-NPs exhibited bactericidal effects on certain bacterial species, although the reduction in HPC was not very pronounced. This may be explained by the fact that these bacteria are not cultivable. WWBF-C and WWBF-G were significantly reduced in the DGGE profile in Fig. 4F. Since the higher intensity indicates a relatively higher concentration of bacteria (Liu et al., 2010), this may indicate that Thiotrichales is more sensitive to Ag-NPs than other strains. The genus of Thiothrix in the order of Thiotrichales is a group of filamentous bacteria commonly found in wastewater treatment plants (Howarth et al., 1999; Kim et al., 2002; Nielsen et al., 2000; Rossetti et al., 2003). Bacteria in this genus can oxidize reduced sulfur compounds and accumulate sulfur globules inside cells (Howarth et al., 1999). It has been reported that
Thiothrix is able to sorb heavy metals (Shuttleworth and Unz, 1993). In this case, sorption of Ag-NPs may have led to a reduction in viability.
3.6. Toxic effects of Ag-NPs on planktonic wastewater bacteria Eight bacterial strains were isolated from the wastewater biofilm, distinguished by the appearance and growth rate of their colonies on R2A agar plates, as shown in Table 2. Planktonic pure and mixed cultures were tested for viability in the presence of Ag-NPs. As shown in Fig. 5A, the initial cell density of each strain was about 1 107 CFU/mL. With the treatment of 1 mg Ag/L Ag-NPs, only three strains (i.e., WWBF3, WWBF-5 and WWBF-6) were still viable after 1 h. After 24 h, only the WWBF-5 was detected, with a 3 log unit reduction. As shown in Fig. 5B, however, when treated as biofilm mixture, other strains besides WWBF-5 (WWBF-1, WWBF-4, WWBF-6, WWBF-7 and WWBF-8) survived after 24 h. WWBF2 and WWBF-3 did not survive after the treatment, and there was a 1e2 log unit reduction in the HPC for WWBF-4, WWBF-6, WWBF-7 and WWBF-8. These results suggest that when treated individually, most of the isolated bacteria are highly sensitive to Ag-NPs, which is consistent with previous research. However, culturing these bacteria as a mixture helps to increase their viability under Ag-NP treatment, although they may be less tolerant to Ag-NPs than they are in the biofilm. This also indicates that the symbiotic effects among bacteria in wastewater biofilms can contribute significantly to their tolerance to Ag-NPs. Similar results were obtained for the artificial mixture, which also support this hypothesis.
4.
Discussion
4.1.
Mechanisms of biofilm tolerance
The high tolerance to Ag-NPs of wastewater biofilms observed in the current study can be explained by several mechanisms: (i) physical protections in the biofilms; (ii) interactions among biofilm microorganisms; (iii) the slow growth rate of certain biofilm microorganisms. More than one mechanism may occur simultaneously.
Table 2 e Colonial morphology and growth rate of isolated biofilm bacteria. Strains
Closest homologs
Color
Shape & surface
Diameter (mm)
Time to form visible colonies on R2A agar
WWBF-1 WWBF-2 WWBF-3 WWBF-4 WWBF-5 WWBF-6 WWBF-7
Flectobacillus sp. CFB group bacterium CFB group bacterium Cloacibacterium normanense Microbacterium oxydans Klebsiella sp. Aeromonas sp.
Pink Yellow Light yellow Dark yellow Greenish yellow White White
2e3 1e2 1e2 2e3 1 5e7 4e6
3d 3d 3d 3d 3d 24 h 24 h
WWBF-8
Enterobacter sp.
White
Smooth and shiny Smooth and shiny Smooth Mucoid Smooth and shiny Round, convex, smooth and shiny Semi-transparent and a little bit wrinkled, with filamentous edges Smoother, less transparent, with curled edges
3e5
24 h
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 3 9 e6 0 5 0
6047
Fig. 5 e Viability of planktonic bacteria isolated from wastewater biofilms: (A) single culture after different time periods of Ag-NP treatment (1 mg Ag/L); (B) individual strains in mixed cultures (biofilm mixture and artificial mixture) after 24 h of AgNP treatment (1 mg Ag/L) compared with the no-treatment control.
4.1.1.
Physical protections
Physical exclusion of Ag-NPs may be provided by the biofilm. 50e90% of the total organic carbon in a biofilm is usually from EPS, and the EPS matrices in wastewater biofilms are relatively complex (Flemming et al., 2000). The removal of loosely bound EPS (over 80% w/w of the original biofilm) resulted in a significant increase in the vulnerability of bacteria to Ag-NPs. Furthermore, genera such as Klebsiella that produce large amounts of EPS were relatively more tolerant to Ag-NPs when treated individually in the current study. This is consistent with previous research indicating that several species from the genus Klebsiella, often present in wastewater treatment processes, have been shown to survive in toxic industrial wastewaters or under antibiotic treatment (Chen et al., 2008;
Zahller and Stewart, 2002). Even after the removal of loosely bound EPS, the cells are still much more tolerant to Ag-NPs compared with planktonic cells. This may be associated with physical protections from tightly bound EPS. Similar results were observed in a study by Liu et al. (2007), where biofilm bacteria, in the presence of EPS, were found to be much more tolerant to TiO2 nanoparticles, compared with planktonic cells. In addition, since mass transfer into a biofilm is driven by diffusion, bacteria inside biofilms may be exposed to substantially lower concentrations of both toxins and nutrients. Furthermore, it is possible that Ag-NPs are trapped in the EPS matrix and cannot reach the microbial cells, due to significant aggregation near the cells (Holbrook et al., 2006). It has also been reported that environmentally relevant pH, ionic
6048
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 3 9 e6 0 5 0
strength, and the presence of natural organic matter, had obvious impacts on the size distribution of Ag-NPs and consequently on their toxic effects (Cumberland and Lead, 2009; Fabrega et al., 2009). In this study, Ag-NPs aggregated under the simulated environment in wastewater biofilms. This may have affected their bactericidal activity since smaller AgNPs are more toxic (Choi and Hu, 2008; Morones et al., 2005). After extended exposure times, more Ag-NPs may penetrate into the biofilm EPS matrix and bacterial cell membranes. It will be interesting to test these effects after 24 h Ag-NP exposure.
4.1.2.
Biofilm community interaction
The results in this study indicate that strains sensitive to AgNPs when tested individually survived when treated with AgNPs in a mixed community. This is consistent with studies showing that bacteria growing as a biofilm are more tolerant to antimicrobial agents due to interactions such as mutualism, commensalism, protocooperation and even competition in the symbiotic microbial community (Allison, 2000). In a mature biofilm community, vulnerable bacteria are often protected deep within the spatially organized consortium as a result of these interactions (Allison, 2000).
4.1.3.
Slow growth rate
Slow-growing bacteria are often more resistant than fastgrowing bacteria to antibiotics (Mah and O’Toole, 2001). This may explain why slow-growing bacteria WWBF-3 and WWBF5 were more tolerant to Ag-NPs. A slow growth rate may also explain the increased tolerance to nanoparticles of bacteria in mature biofilms, where nutrients are limited (Brown et al., 1988; Mah and O’Toole, 2001). The protective effects from a slow growth rate may be associated with the expression of stress response genes (Lu et al., 2009; Stewart, 2002). However, growth rate is not the only reason for the tolerance. Other factors, such as certain EPS components, may play an important role in the tolerance of each strain. It is also possible that starvation may contribute to the results obtained in this study, since there is no carbon source in PBS. No significant change of heterotrophic bacterial number was observed in no-treatment control throughout the 24 h, indicating the reliability of the results. Further studies are needed to differentiate the potential impact of starvation. It is worth noting that a 1 log unit increase in WWBF-1 was observed in the biofilm mixture. The intensity of WWBF-E from the treated sample was also even higher than the intensity of the corresponding band from the no-treatment control. This suggests that certain bacteria started to grow using cell debris of bacteria killed by Ag-NPs, causing recovery of the HPC during treatment.
4.2.
Environmental implications
Conditions in wastewater treatment plants are generally more complex than in the laboratory. Microorganisms usually congregate in suspended flocs (e.g., in the activated sludge process) or attached biofilms (e.g., in RBCs). The ionic strength in wastewater is usually high because of the presence of charged organic and inorganic ligands. Therefore, in this study, physical exclusion provided by the biofilms and the wastewater conditions were taken into account. Different from previous reports, microorganisms tested here are highly
tolerant to Ag-NP treatment, indicating that results from treatment of planktonic bacteria cannot be used to estimate the impact of nanoparticles in environmental systems. These results largely complement previous research which only focuses on planktonic cells. In addition, results here underlined the importance to understand the impact of nanoparticles on the microbial community structure in addition to counting the total number of bacteria. In this study, one group of filamentous microorganisms was observed to be highly susceptible to Ag-NPs. In the biological treatment process, filamentous microbes account for only 1% of the microorganisms. However, a reduction in filamentous microbes can significantly impact treatment efficiency, especially in activated sludge systems since filamentous microbes constitute the backbone of activated sludge flocs (Sezgin et al., 1978). Furthermore, it was found here that Ag-NPs can be sorbed to biofilm matrix, indicating a potential role for biological removal of nanoparticles from wastewater. Sorption and accumulation in microbial aggregates may increase the concentration of engineered nanoparticles in biological treatment systems and thus pose a significant threat. The concentration of engineered nanoparticles in biological aggregates has not yet been determined, and is an important area for future study, particularly in sludge treatment. Some research has been carried out based on laboratory scale reactors in this direction (Hu, 2010).
4.3.
Future research
It should be noted that this study only focused on one kind of nanoparticle under aerobic conditions. Research on other kinds of nanoparticles in conditions relevant to various engineered wastewater purification systems would be further helpful. In addition, this study focused only on the protective mechanisms for wastewater biofilms. Further studies on the mechanisms controlling Ag-NPs toxicity will provide valuable information for understanding the potential adverse impact of Ag-NPs. Further, bacterial tolerance to silver may potentially develop during the course of the treatment (Chopra, 2007), which requires further studies to understand its impacts on the bacterial susceptibility to Ag-NPs. It should also be noted that the HPC result is highly dependent on bacterial cultivability. It is possible that some bacteria are injured but still viable. These injured cells may temporarily lose their colony forming capability but are able to recover from the damage, which may also explain our observation that for biofilm samples with loosely bound EPS removed, HPC reduced then increased after 200 mg Ag/L Ag-NP treatment. Similar recovery phenomena have been reported for other nanoparticles (Hardman, 2006). Therefore, for future studies, culture-independent methods should be applied to verify bacterial viability. Longer exposure times may also provide the opportunity to observe the return from dormancy.
5.
Conclusions
Original wastewater biofilms are highly tolerant to Ag-NPs. However, accumulated Ag-NPs in wastewater biofilms may impact their microbial activity.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 3 9 e6 0 5 0
Biofilm can provide physical protections for bacteria under Ag-NP treatment, and EPS may play an important role in this protection. Biofilm bacteria with loosely bound EPS removed are more sensitive to Ag-NPs. The effects of Ag-NPs on planktonic cells are different than on wastewater biofilms. Biofilm bacteria treated as isolated pure culture are much more sensitive to Ag-NPs, compared with mixture of bacteria in the biofilm. Even artificially mixed bacteria community can survive better under Ag-NP treatment. Susceptibility to Ag-NPs is different for each microorganism in the biofilm microbial community. Thiotrichales, in this study, is more sensitive than other biofilm bacteria.
Acknowledgments This work was supported by the Natural Sciences and Engineering Research Council of Canada (NSERC) Discovery Grant Number 386457, NSERC Research Tools and Instruments, Canadian School of Energy and the Environment, and the Alberta Ingenuity Graduate Scholarship in Nanotechnology.
Appendix. Supplementary material Supplementary data related to this article can be found online at doi:10.1016/j.watres.2011.08.065.
references
Allison, D.G., 2000. Community Structure and Co-operation in Biofilms. Cambridge University Press, New York. Benn, T.M., Westerhoff, P., 2008. Nanoparticle silver released into water from commercially available sock fabrics. Environmental Science & Technology 42 (11), 4133e4139. Blaser, S.A., Scheringer, M., Macleod, M., Hungerbuhler, K., 2008. Estimation of cumulative aquatic exposure and risk due to silver: contribution of nano-functionalized plastics and textiles. Science of the Total Environment 390 (2e3), 396e409. Brown, M., Allison, D.G., Gilbert, P., 1988. Resistance of bacterial biofilms to antibiotics e a growth-rate related effect. Journal of Antimicrobial Chemotherapy 22 (6), 777e780. Chen, C.Y., Kao, C.M., Chen, S.C., 2008. Application of Klebsiella oxytoca immobilized cells on the treatment of cyanide wastewater. Chemosphere 71 (1), 133e139. Choi, O., Hu, Z., 2008. Size dependent and reactive oxygen species related nanosilver toxicity to nitrifying bacteria. Environmental Science & Technology 42 (12), 4583e4588. Choi, O., Deng, K.K., Kim Jr., N.J., Ross, L., Surampalli, R.Y., Hu, Z., 2008. The inhibitory effects of silver nanoparticles, silver ions, and silver chloride colloids on microbial growth. Water Research 42 (12), 3066e3074. Choi, O., Cleuenger, T.E., Deng, B.L., Surampalli, R.Y., Ross, L., Hu, Z.Q., 2009. Role of sulfide and ligand strength in controlling nanosilver toxicity. Water Research 43 (7), 1879e1886. Chopra, I., 2007. The increasing use of silver-based products as antimicrobial agents: a useful development or a cause for
6049
concern? Journal of Antimicrobial Chemotherapy 59 (4), 587e590. Cortez, S., Teixeira, P., Oliveira, R., Mota, M., 2008. Rotating biological contactors: a review on main factors affecting performance. Reviews in Environmental Science and Biotechnology 7 (2), 155e172. Cumberland, S.A., Lead, J.R., 2009. Particle size distributions of silver nanoparticles at environmentally relevant conditions. Journal of Chromatography A 1216 (52Sp. Iss. SI), 9099e9105. Davies, D., 2003. Understanding biofilm resistance to antibacterial agents. Nature Reviews Drug Discovery 2 (2), 114e122. Fabrega, J., Fawcett, S.R., Renshaw, J.C., Lead, J.R., 2009. Silver nanoparticle impact on bacterial growth: effect of pH, concentration, and organic matter. Environmental Science & Technology 43 (19), 7285e7290. Flemming, H.C., Wingender, J., Griegbe, T., Mayer, C., 2000. In: Evans, L.V. (Ed.), Biofilms: Recent Advances in their Study and Control. Harwood Academic Publishers, Amsterdam, pp. 19e34. Geranio, L., Heuberger, M., Nowack, B., 2009. The behavior of silver nanotextiles during washing. Environmental Science & Technology 43 (21), 8113e8118. Gong, A.S., Bolster, C.H., Benavides, M., Walker, S.L., 2009. Extraction and analysis of extracellular polymeric substances: comparison of methods and extracellular polymeric substance levels in Salmonella pullorum SA 1685. Environmental Engineering Science 26 (10), 1523e1532. Hagendorfer, H., Lorenz, C., Kaegi, R., Sinnet, B., Gehrig, R., Goetz, N.V., Scheringer, M., Ludwig, C., Ulrich, A., 2010. Sizefractionated characterization and quantification of nanoparticle release rates from a consumer spray product containing engineered nanoparticles. Journal of Nanoparticle Research 12 (7), 2481e2494. Hall-Stoodley, L., Costerton, J.W., Stoodley, P., 2004. Bacterial biofilms: from the natural environment to infectious diseases. Nature Reviews Microbiology 2 (2), 95e108. Hardman, R.C., 2006. A toxicologic review of quantum dots: toxicity depends on physicochemical and environmental factors. Environmental Health Perspectives 114 (2), 165e172. Holbrook, R.D., Morrow, J., Maugel, T., 2006. Fate of nanoparticles in biological systems e quantum dot behavior in bacterial cells and biofilms. Microscopy and Microanalysis 12 (Suppl. S02), 618e619. Howarth, R., Unz, R.F., Seviour, E.M., Seviour, R.J., Blackall, L.L., Pickup, R.W., Jones, J.G., Yaguchi, J., Head, I.M., 1999. Phylogenetic relationships of filamentous sulfur bacteria (Thiothrix spp. and Eikelboom type 021n bacteria) isolated from wastewater-treatment plants and description of Thiothrix eikelboomii sp. nov., Thiothrix unzii sp. nov., Thiothrix fructosivorans sp. nov. and Thiothrix defluvii sp. nov. International Journal of Systematic Bacteriology 49 (Part 4), 1817e1827. Hu, Z., 2010. Impact of Silver Nanoparticles on Wastewater Treatment. Water Environment Research Foundation, Alexandria. Kim, S.B., Goodfellow, M., Kelly, J., Saddler, G.S., Ward, A.C., 2002. Application of oligonucleotide probes for the detection of Thiothrix spp. in activated sludge plants treating paper and board mill wastes. Water Science and Technology 46 (1e2), 559e564. Konopka, M., Kowalski, Z., Wzorek, Z., 2009. Disinfection of meat industry equipment and production rooms with the use of liquids containing silver nano-particles. Archives of Environmental Protection 35 (1), 107e115. Kumar, R.V., Raza, G., 2009. Photocatalytic disinfection of water with Ag-TiO2 nanocrystalline composite. Ionics 15 (5), 579e587. Kumar, M.A., Anandapandian, K.T.K., Parthiban, K., 2011. Production and characterization of exopolysaccharides (EPS)
6050
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 3 9 e6 0 5 0
from biofilm forming marine bacterium. Brazilian Archives of Biology and Technology 54, 259e265. Liu, Y., Li, J., Qiu, X.F., Burda, C., 2007. Bactericidal activity of nitrogen-doped metal oxide nanocatalysts and the influence of bacterial extracellular polymeric substances (EPS). Journal of Photochemistry and Photobiology A e Chemistry 190 (1), 94e100. Liu, C., Yang, J.L., Wu, G., Zhang, S., Li, Z.X., Guo, J.B., 2010. Estimation of dominant microbial population sizes in the anaerobic granular sludge of a full-scale UASB treating streptomycin wastewater by PCR-DGGE. World Journal of Microbiology & Biotechnology 26 (2), 375e379. Lu, C., Brauer, M.J., Botstein, D., 2009. Slow growth induces heatshock resistance in normal and respiratory-deficient yeast. Molecular Biology of the Cell 20 (3), 891e903. Mah, T.C., O’Toole, G.A., 2001. Mechanisms of biofilm resistance to antimicrobial agents. Trends in Microbiology 9 (1), 34e39. Marambio-Jones, C., Hoek, E., 2010. A review of the antibacterial effects of silver nanomaterials and potential implications for human health and the environment. Journal of Nanoparticle Research 12 (5), 1531e1551. Mcinroy, L., Gibson, M., Clark, R., Cullen, B., 2009. Are silver dressings effective against biofilm formation? Wound Repair and Regeneration 17 (4), A73. Morones, J.R., Elechiguerra, J.L., Camacho, A., Holt, K., Kouri, J.B., Ramirez, J.T., Yacaman, M.J., 2005. The bactericidal effect of silver nanoparticles. Nanotechnology 16 (10), 2346e2353. Mueller, N.C., Nowack, B., 2008. Exposure modeling of engineered nanoparticles in the environment. Environmental Science & Technology 42 (12), 4447e4453. Nielsen, P.H., de Muro, M.A., Nielsen, J.L., 2000. Studies on the in situ physiology of Thiothrix spp. present in activated sludge. Environmental Microbiology 2 (4), 389e398. Petit, C., Lixon, P., Pileni, M.P., 1993. In situ synthesis of silver nanocluster in AOT reverse micelles. The Journal of Physical Chemistry 97 (49), 12974e12983. Rittmann, B.E., Mccarty, P.L., 2001. Environmental Biotechnology: Principles and Applications. McGraw-Hill, New York. Rossetti, S., Blackall, L., Levantesi, C., Uccelletti, D., Tandoi, V., 2003. Phylogenetic and physiological characterization of a heterotrophic, chemolithoautotrophic Thiothrix strain isolated from activated sludge. International Journal of Systematic and Evolutionary Microbiology 53 (Part 5), 1271e1276.
Sezgin, M., Jenkins, D., Parker, D.S., 1978. A unified theory of filamentous activated sludge bulking. Journal of the Water Pollution Control Federation 50 (2), 362e381. Shuttleworth, K.L., Unz, R.F., 1993. Sorption of heavy-metals to the filamentous bacterium Thiothrix strain A1. Applied and Environmental Microbiology 59 (5), 1274e1282. Silvestry-Rodriguez, N., Bright, K.R., Slack, D.C., Uhlmann, D.R., Gerba, C.P., 2008. Silver as a residual disinfectant to prevent biofilm formation in water distribution systems. Applied and Environmental Microbiology 74 (5), 1639e1641. Stewart, P.S., 2002. Mechanisms of antibiotic resistance in bacterial biofilms. International Journal of Medical Microbiology 292 (2), 107e113. Sur, I., Cam, D., Kahraman, M., Baysal, A., Culha, M., 2010. Interaction of multi-functional silver nanoparticles with living cells. Nanotechnology 21 (17), 175104. The Project on Emerging Nanotechnologies. http://www. nanotechproject.org/ (accessed 11.07.10). van de Peer, Y., de Wachter, R., 1994. TREECON for Windows: a software package for the construction and drawing of evolutionary trees for the Microsoft Windows environment. Computer Applications in the Biosciences 10 (5), 569e570. Wu, S., Feng, X., Wittmeier, A., 1997. Microwave digestion of plant and grain reference materials in nitric acid or a mixture of nitric acid or a mixture of nitric acid and hydrogen peroxide for the determination of multi-elements by inductively coupled plasma mass spectrometry. Journal of Analytical Atomic Spectrometry 12 (8), 797e806. Yu, Z., Morrison, M., 2004. Comparisons of different hypervariable regions of rrs genes for use in fingerprinting of microbial communities by PCR-denaturing gradient gel electrophoresis. Applied and Environmental Microbiology 70 (8), 4800e4806. Zahller, J., Stewart, P.S., 2002. Transmission electron microscopic study of antibiotic action on Klebsiella pneumoniae biofilm. Antimicrobial Agents and Chemotherapy 46 (8), 2679e2683. Zelver, N., Hamilton, M., Pitts, B., Goeres, D., Walker, D., Sturman, P., Heersink, J., 1999. Measuring antimicrobial effects on biofilm bacteria: from laboratory to field. Methods Enzymology 310, 608e628. Zhao, Y.K., Sung, W.P., Tsai, T.T., Wang, H.J., 2010. Application of nanoscale silver-doped titanium dioxide as photocatalyst for indoor airborne bacteria control: a feasibility study in medical nursing institutions. Journal of the Air & Waste Management Association 60 (3), 337e345.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 5 1 e6 0 6 2
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Characterization of bromate-reducing bacterial isolates and their potential for drinking water treatment Andrew N. Davidson a,1, Joanne Chee-Sanford b, Hoi Yi (Mandy) Lai a,2, Chi-hua Ho a,3, J. Brandon Klenzendorf c, Mary Jo Kirisits a,* a
Department of Civil, Architectural, and Environmental Engineering, The University of Texas at Austin, 1 University Station C1786, Austin, TX 78712, USA b Department of Natural Resources and Environmental Sciences, The University of Illinois at Urbana-Champaign, 1102 South Goodwin Mailcode 047, Urbana, IL 61801, USA c Geosyntec Consultants, 3600 Bee Caves Road, Suite 101, Austin, TX 78746, USA
article info
abstract
Article history:
The objective of the current study was to isolate and characterize several bromate-reducing
Received 24 May 2011
bacteria and to examine their potential for bioaugmentation to a drinking water treatment
Received in revised form
process. Fifteen bromate-reducing bacteria were isolated from three sources. According to
29 August 2011
16S rRNA gene sequencing, the bromate-reducing bacteria are phylogenetically diverse,
Accepted 1 September 2011
representing the Actinobacteria, Bacteroidetes, Firmicutes, and a-, b-, and g-Proteobacteria. The
Available online 10 September 2011
broad diversity of bromate-reducing bacteria suggests the widespread capability for microbial bromate reduction. While the cometabolism of bromate via nitrate reductase and
Keywords:
(per)chlorate reductase has been postulated, five of our bromate-reducing isolates were
Bioaugmentation
unable to reduce nitrate or perchlorate. This suggests that a bromate-specific reduction
Biologically active carbon
pathway might exist in some microorganisms. Bioaugmentation of activated carbon filters
Bromate
with eight of the bromate-reducing isolates did not significantly decrease start-up time or
Bromide
increase bromate removal as compared to control filters. To optimize bromate reduction
Drinking water
in a biological drinking water treatment process, the predominant mechanism of
Reduction
bromate reduction (i.e., cometabolic or respiratory) needs to be assessed so that appropriate measures can be taken to improve bromate removal. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
Concerns over chlorine-resistant protozoan cysts, chlorinated disinfection by-products, and taste and odor problems have led some drinking water utilities to switch from chlorine to ozone as their primary disinfectant. However, when a water contains bromide, ozonation can result in the formation of the
disinfection by-product bromate (BrO 3 ) (reviewed by von Gunten, 2003). Typical conditions in a drinking water treatment plant produce less than 60 mg/L bromate (Krasner et al., 1993), but higher bromate concentrations have been observed in some groundwaters (w1 mg/L) due to industrial contamination (Butler et al., 2006). Bromate causes cancer in rats (DeAngelo et al., 1998) and is classified as a probable
* Corresponding author. Tel.: þ1 512 232 7120; fax: þ1 512 471 0592. E-mail address: [email protected] (M.J. Kirisits). 1 Present address: GSI Water Solutions Inc., 55 SW Yamhill Street, Suite 300, Portland, OR 97204, USA. 2 Present address: CDM, 100 Pringle Avenue, Suite 300, Walnut Creek, CA 94596, USA. 3 Present address: Shine & Pretty, 456 Constitution Avenue, Camarillo, CA 93012, USA. 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.09.001
6052
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 5 1 e6 0 6 2
Nomenclature ARDRA bp BLAST BAC cm CFU C DNA DO EBCT g GAC h L LB MCL mg mL mm
amplified ribosomal DNA restriction analysis base pairs Basic Local Alignment Search Tool biologically active carbon centimeter colony forming unit degrees Celsius deoxyribonucleic acid dissolved oxygen empty bed contact time gram granular activated carbon hour liter LuriaeBertani maximum contaminant level microgram microliter micrometer
human carcinogen by the United States Environmental Protection Agency (U.S. EPA). The maximum contaminant level (MCL) for bromate in drinking water in the U.S. is 10 mg/L, although recent work demonstrating bromate reduction in simulated human gastric juice suggests that this MCL is overprotective of human health (Cotruvo et al., 2010). Similar to the microbial reduction of perchlorate to innocuous chloride (Kim and Logan, 2001; Brown et al., 2002), microbial reduction of bromate to innocuous bromide offers promise for sustainable removal in drinking water. Microbial bromate reduction is inhibited by oxygen (Hijnen et al., 1995); thus, anaerobic conditions are needed to ensure microbial bromate reduction, which can be accomplished by removing dissolved oxygen (DO) from the bulk water in a suspendedgrowth reactor or by creating anaerobic zones in a biofilm reactor. If a biological bromate-reduction process were to follow ozonation, an exogenous electron donor would need to be provided to encourage DO consumption, as is done in biological perchlorate-reduction processes (Brown et al., 2008; Webster et al., 2009). Microbial bromate reduction has been demonstrated in a variety of reactor types including biologically active carbon (BAC) filters (Kirisits et al., 2002), a denitrifying fixed-bed bioreactor supplemented with ethanol (Hijnen et al., 1999), a hydrogen gas-lift bioreactor (van Ginkel et al., 2005a), and a hydrogen-based, denitrifying membrane biofilm reactor (Downing and Nerenberg, 2007). A few studies have examined the microbial communities capable of bromate reduction. Hijnen et al. (1995) observed near-stoichiometric bromate reduction to bromide by a mixed microbial community; some of the bromate-reducing microorganisms isolated from that mixed culture were denitrifying organisms, which are defined as organisms that reduce nitrate to gaseous end products such as N2O or N2. Bromate has been shown to be a substrate for purified nitrate reductase (Morpeth and Boxer, 1985; Yamamoto et al., 1986), and it has been suggested that nitrate reductase might be involved in microbial
mg milligram mL milliliter mm millimeter mM millimolar min minute ng nanogram N medium nitrate medium NN medium no nitrate medium OTU operational taxonomic unit % percent PBS phosphate-buffered saline pmol picomole PCR polymerase chain reaction rRNA ribosomal ribonucleic acid s seconds a significance level T-RFLP terminal restriction fragment length polymorphism U.S. EPA United States Environmental Protection Agency U units
bromate reduction (Hijnen et al., 1995). However, Downing and Nerenberg (2007) demonstrated that the denitrifying bacterium Ralstonia eutropha was incapable of bromate reduction, indicating that bromate reduction is not a functionally linked trait shared by all denitrifiers. Interestingly, bromate has a higher reduction potential than does nitrate under concentration conditions representative of drinking water (Kirisits and Snoeyink, 1999). Furthermore, bromate reduction is not limited to a subset of denitrifying microorganisms. Hijnen et al. (1995) demonstrated bromate reduction by nitrate-respiring but nondenitrifying isolates (i.e., growth coupled to nitrate reduction without N2 production). In addition to the potential role of nitrate reductase in bromate reduction, some studies suggest cometabolism of bromate via other reductases. For instance, Kengen et al. (1999) demonstrated bromate-reduction activity by purified (per)chlorate reductase, and Martin et al. (2009) observed bromate reduction but no growth by the perchlorate-reducing Dechloromonas sp. PC1. van Ginkel et al. (2005a) noted cometabolic bromate reduction in a chlorate-reducing bioreactor. Additionally, the aerobically expressed selenate reductase of Enterobacter cloacae is capable of low rates of bromate reduction (Ridley et al., 2006). However, other studies suggest the existence of a specific bromate-reduction pathway. van Ginkel et al. (2005b) observed bromate reduction by an anaerobic, mixed microbial community where bromate was the sole electron acceptor. Since that mixed culture was unable to reduce nitrate, perchlorate, or chlorate, a specific bromate-reduction pathway might exist. Taken together, the literature suggests diversity in the mechanism of microbial bromate reduction (i.e., cometabolism of bromate via nitrate or (per)chlorate reductase and respiration of bromate via a specific bromatereduction pathway). The objective of the current study was to isolate and characterize several bromate-reducing bacteria and to
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 5 1 e6 0 6 2
examine their potential for bioaugmentation to a drinking water treatment process. We isolated bromate-reducing bacteria from bench-scale BAC filters treating tapwater and from natural water, determined their phylogeny via 16S ribosomal ribonucleic acid (rRNA) gene sequencing, and examined their ability to reduce nitrate. We used some of these bromate-reducing isolates for bioaugmentation to a BAC filter to determine if start-up time could be decreased or if bromate removal could be increased.
2.
Materials and methods
2.1.
Isolation of bromate-reducing bacteria
A selective agar medium (M-R2A-bromate) was prepared based on Fries et al. (1994), with some modifications. Each liter of medium (adjusted to pH 7.0) contained the following components: agar, 15 g; KH2PO4, 0.25 g; K2HPO4, 0.4 g; NaBrO3, 118 mg; KNO3, 0.101 g; CaCl2$2H2O, 0.015 g; MgCl2$6H2O, 0.02 g; FeSO4$7H2O, 0.007 g; Na2SO4, 0.005 g; NH4Cl, 0.8 g; trace metals (MnCl2$4H2O, 5 mg; H3BO3, 0.5 mg; ZnCl2, 0.5 mg; CoCl2$6H2O, 0.5 mg; NiSO4$6H2O, 0.5 mg; CuCl2$2H2O, 0.3 mg; and NaMoO2$2H2O, 0.01 mg); vitamins according to Staley (1968); and carbon sources (yeast extract, 0.5 g; peptone, 0.5 g; casamino acids, 0.5 g; and sodium pyruvate, 0.3 g). Agar plates were degassed in an anaerobic chamber for several days before use. Biomass samples were collected from the BAC filters operated in the current study (Section 2.5), from the water column of Waller Creek (Austin, TX), and from a previously operated BAC filter treating tapwater derived from groundwater (Kirisits et al., 2002). The diluted samples were plated onto M-R2A-bromate agar and incubated in an anaerobic chamber at room temperature (w21 C). Distinct colony morphologies were struck to purity. The number of isolates to be retained for further study was reduced through amplified ribosomal deoxyribonucleic acid (DNA) restriction analysis (ARDRA). Each isolate was grown aerobically in LuriaeBertani (LB) medium, and DNA was extracted (UltraClean Soil DNA Isolation kit, MoBio Laboratories, Inc., Carlsbad, CA). The 16S rRNA gene was amplified by polymerase chain reaction (PCR) using primers 8f (50 -AGAGTTTGATCCTGGCTCAG-30 ) and 1492r (50 -GGTTACCTTGTTACGACTT-30 ). PCR reactions contained 1.5 mM MgCl2, 0.2 mM of each deoxynucleoside triphosphate, 1 PCR buffer, 1.25 U Taq DNA polymerase, 50 ng template DNA, and 20 pmol of each primer in a 50-mL reaction. Each PCR tube was placed in a PTC-200 Peltier Thermal Cycler (MJ Research, Inc., Waltham, MA) and run under the following conditions: initial denaturation at 94 C for 5 min; 35 cycles at 94 C for 1 min, 55 C for 1 min, and 72 C for 1.5 min; final elongation at 72 C for 7 min. Amplicon size was verified on a 1% agarose gel. Amplicon was digested in a 20-mL reaction containing 10 mL of amplicon, 100 mg/mL bovine serum albumin, 1 NEBuffer 4, and 20 U HhaI (New England Biolabs, Ipswich, MA) for 2 h at 37 C, followed by enzyme inactivation for 20 min at 65 C. Digested products were run on a 1% agarose gel. Duplicate freezer stocks were prepared for each unique isolate and stored at 80 C.
2.2.
6053
Batch bromate reduction studies
All retained isolates were subjected to batch studies to verify their bromate-reducing ability. Liquid media with nitrate (N medium) and with no nitrate (NN medium) were prepared for batch bromate-reduction studies (pH 7.5). Acetate, lactate, and pyruvate were the electron donors and were dosed at w200% of the stoichiometric amount required for the reduction of bromate, nitrate, and oxygen (neglecting cell growth). Each liter of NN medium contained the following components: NaHCO3, 50 mg; NH4Cl, 34 mg; KH2PO4, 23 mg; K2HPO4, 30 mg; NaBrO3, 590 mg; carbon sources (sodium acetate, 2.8 mg; sodium lactate, 2.5 mg; sodium pyruvate, 2.5 mg); and the same final concentrations of trace metals and vitamins as specified in Section 2.1. N medium was the same as NN medium, with the exception that it contained (per liter) NaNO3, 6.9 mg; sodium acetate, 13.9 mg; sodium lactate, 12.6 mg; and sodium pyruvate, 12.6 mg. Balch tubes (Bellco Glass, Inc., Vineland, NJ) containing 15 mL of N medium or NN medium were prepared (four tubes for each medium per isolate). Triplicate uninoculated control tubes were prepared for each medium. The tubes were sealed with butyl rubber stoppers, crimp-capped, and purged with filtered oxygen-free nitrogen gas for 10 min. The isolates (all facultative anaerobes) were plated on LB agar and incubated aerobically at 30 C. One colony of each isolate was inoculated to three tubes of N medium and three tubes of NN medium. To safeguard against the possibility of slow bromate-reduction kinetics in the tubes inoculated with single colonies, the remaining tubes were inoculated with 10e20 colonies. The tubes were incubated with shaking at 30 C. The tubes were aseptically sampled once per month.
2.3.
Phylogenetic analysis
The 16S rRNA genes of the bromate-reducing isolates (amplified as described in Section 2.1) identified from the BAC filters operated in this study and from Waller Creek were sent to the Institute for Cellular and Molecular Biology Core Facilities (The University of Texas at Austin, Austin, TX); they were sequenced bidirectionally using primers 8f and 1492r. For the bromate-reducing isolates from a previously operated BAC filter treating tapwater derived from groundwater (Kirisits et al., 2002), the 16S rRNA genes were cloned using a TOPO TA cloning kit (Invitrogen Corp., San Diego, CA). The clone inserts were sent to the W. M. Keck Center for Comparative and Functional Genomics (University of Illinois at UrbanaChampaign Biotechnology Center, Urbana, IL); they were sequenced bidirectionally with primers M13f and M13r. The chromatograms of the nucleotide bases were edited using EditView 1.0.1 (PerkineElmer Biosystems, Foster City, CA), and the sequences were assembled using Auto Assembler (PerkineElmer Biosystems, Foster City, CA) or the Assembler tool in MacVector (v.10.6.1) software (MacVector, Inc., Cary, NC). The 16S rRNA sequences (>1200 base pairs [bp]) of the bromate-reducing isolates were aligned with sequences from reference strains obtained from the GenBank database, and a BLAST (Basic Local Alignment Search Tool) search was used to identify the closest cultivar to each isolate. A phylogenetic tree was constructed using the neighbor-joining method
6054
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 5 1 e6 0 6 2
based on JukeseCantor distance calculations in MacVector. The sequences were deposited to GenBank under accession numbers AF442522eAF442524, JF951733eJF951743, and JN002283.
2.4. Terminal restriction fragment length polymorphism (T-RFLP) T-RFLP was used to examine the persistence of the bioaugmented isolates in the early-stage bioaugmented BAC filters (Section 2.6). Each BAC core sample was sonicated for 10 min (Branson Ultrasonics Corp., Danbury, CT) to detach cells; the filter media particles were settled out, the supernatant was decanted, and cells were collected from the supernatant by centrifugation. The bioaugmented isolates were grown aerobically in LB broth. DNA was isolated from all samples with an UltraClean Soil DNA Isolation kit (MoBio Laboratories, Inc., Carlsbad, CA). The 16S rRNA genes were amplified using primers 8f-FAM-labeled and 926r (50 CCGTCAATTCCTTTGAGTTT-30 ). PCR reactions were set up as in Section 2.1, except that the template DNA consisted of 25 ng (isolates) or 100 ng (BAC samples). The 16S rRNA genes were amplified under the following conditions: initial denaturation at 94 C for 5 min; 16 cycles at 94 C for 30 s, 52 C for 30 s, and 72 C for 1 min; final elongation at 72 C for 7 min. Amplicon size was verified on a 1% agarose gel. Following the procedure described by Egert and Friedrich (2005), amplicon was subjected to Klenow treatment. Klenow-treated amplicon was cleaned using the Ultraclean PCR cleanup kit (MoBio Laboratories, Inc., Carlsbad, CA) and digested with 40 U of HhaI (New England Biolabs, Ipswich, MA) in a 20-mL reaction for 3 h at 37 C followed by 20 min at 65 C for enzyme inactivation. The digested amplicon was desalted with a Microcon spin filter (Millipore, Billerica, MA) and sent for fragment sizing at the Institute for Cellular and Molecular Biology Core Facilities (The University of Texas at Austin, Austin, TX). Fragments were sized using Genemarker 1.5 software (Softgenetics LLC, State College, PA).
2.5.
BAC filter e control experiments
Calgon F-400 granular activated carbon (GAC) (Calgon Corp., Pittsburgh, PA) was sieved, and the 30 40 mesh fraction
(425e600 mm) was retained. Two filters (A and B) were constructed in 25-mm inner-diameter borosilicate glass columns, which were fitted with Teflon endcaps (Chemglass, Vineland, NJ). GAC (25.7 g) was placed on top of a support layer (7.6-cm depth) of 4-mm diameter glass beads (Fig. 1). Influent was pumped through each filter with a Watson-Marlow 205S peristaltic pump (Watson-Marlow Inc., Wilmington, MA) for a target empty bed contact time (EBCT) of 20 min. The filters were operated at room temperature (w21 C). Influent and effluent samples were collected and stored at 4 C until analysis. Filter A was operated under control conditions (no bioaugmentation or acetate supplementation) for its entire lifetime, and filter B was operated initially under control conditions to demonstrate the reproducibility of bromate reduction in this system (Table 1). Then, filter B was used for additional experiments (Table 1, Section 2.7). Influent was prepared from local tapwater (Austin, TX), which is supplied from Lady Bird Johnson Lake. The water was dechlorinated with sodium sulfite (resulting in the release of 0.3e0.6 mg/L ammonia-N), buffered with 1 mM potassium phosphate monobasic, spiked with NaBrO3 to a final target concentration of 20 mg/L as BrO 3 , and adjusted to pH 7.5. The influent was transferred to a headspace-free influent tank and sparged with nitrogen gas to decrease the average DO concentration to 2.1 mg/L (0.2 mg/L standard deviation).
2.6. BAC filter e early-stage bioaugmentation experiments Three new filters (C, D, and E; Table 1) were prepared as described in Section 2.5. Just after construction, these filters were bioaugmented with eight bromate-reducing strains (B2, B6eB11, B15) that were isolated from a previously operated BAC filter treating tapwater derived from groundwater (Kirisits et al., 2002). The isolates were grown aerobically in liquid R2A medium. Approximately 106 colony forming units (CFU) of each isolate were added to a master inoculum that was injected to each filter via sterile syringe. Once the inoculum filled the bed, flow was stopped for 2 h to promote bacterial attachment to the GAC. Then, the filters were connected to the influent tank (same influent composition as described in Section 2.5), and flow was resumed for a target EBCT of 20 min.
Table 1 e Experimental history of each filter. Filter A B
C D E
Fig. 1 e Schematic of the BAC filter setup.
Experiment description
Bed volumes
Control conditions Control conditions Late-stage bioaugmentation Acetate addition Verification of biotic bromate reduction (autoclaving) Early-stage bioaugmentation Early-stage bioaugmentation Early-stage bioaugmentation (used to replenish cores from filters C and D)
0e23,400 0e21,800 21,800e23,400 49,300e50,700 50,700e50,950 0e12,000 0e12,000 0e12,000
6055
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 5 1 e6 0 6 2
Filters C and D were operated under the same conditions to show the reproducibility of bromate reduction in this system; influent and effluent samples were collected from filters C and D and stored at 4 C until analysis. Every two weeks, a ¼-inch inner-diameter glass tube was used to take a core of filters C and D for T-RFLP analysis (Section 2.4); the removed material was replaced with a core from filter E to maintain a consistent bed volume for the duration of the experiment.
eluent, and a 1.0 mL/min flow rate. For batch and BAC filter samples, 25- and 500-mL sample loops were used, respectively. A YSI Model 54A DO meter (Yellow Springs, OH) was used to measure DO. An Orion ion-selective electrode probe and Model 920A meter (Thermo Fischer Scientific, Inc., Waltham, MA) were used to measure pH.
2.7. BAC filter e late-stage bioaugmentation and other experiments
Parametric and nonparametric statistical analyses were conducted to make comparisons among the means of the BAC filter data sets. The parametric Student’s t-test was used on data sets with more than 15 collected samples. For smaller data sets, the nonparametric ManneWhitney test was used. A p-value was determined for each statistical test and compared to a significance level of a ¼ 0.1. In addition to comparing the means of various breakthrough curves, regression analyses were conducted to compare the shapes of the breakthrough curves. Linear regression was used to compare the rate at which breakthrough occurred (i.e., the slope of the curve) using 95% confidence intervals.
At 21,800 bed volumes, filter B was bioaugmented with the eight bromate-reducing strains (B2, B6eB11, B15) that were isolated from a previously operated BAC filter treating tapwater derived from groundwater (Kirisits et al., 2002). Each isolate was grown aerobically in LB medium and washed in 1 phosphate-buffered saline (PBS). Approximately 108 CFU of each isolate were added to a master inoculum that was injected to the filter via sterile syringe. Bacterial attachment and flow initiation proceeded as described in Section 2.6. At 49,300 bed volumes, 5 mg/L acetate was added to the influent of filter B to examine the impact of electron donor supplementation (Table 1). This is w200% of the stoichiometric amount required for the reduction of the 2.1 mg/L DO, 0.2 mg/L nitrate-N, and 20 mg/L bromate present in the influent. At 50,700 bed volumes, filter B was autoclaved to verify that the observed bromate removal in the BAC filters was biological (Table 1). When filter B was placed back on-line, the influent was supersaturated with DO (>15 mg/L) because biotic bromate reduction is inhibited by DO. The collection of influent and effluent samples continued during the different experiments conducted on filter B.
2.8.
Analytical methods
A customized Dionex ion chromatography system (Sunnyvale, CA) was used to measure the concentration of bromate in the batch bromate reduction studies (Section 2.2) and in the BAC filter samples (Sections 2.5e2.7). Additionally, bromide, nitrate, nitrite, acetate, lactate, and pyruvate were measured in the batch bromate reduction studies. System components include the LC-25 Chromatography Oven, the AS-40 Automated Sampler, the GS-50 Gradient Pump, the EG-50 Eluent Generation Unit, the ED50 Electrochemical Detector, and Chromeleon Software. The analytical method includes an anion self-regenerating suppressor (ASRS-4 mm), an AG-19 guard column, an AS-19 analytical column, a 20 mM KOH
2.9.
Statistical analysis
3.
Results and discussion
3.1.
Isolation of bromate-reducing bacteria
Selective plating followed by ARDRA and bromate-reduction batch studies identified 15 bromate-reducing isolates (four from the BAC filter operated in the current study [AUS11, AUS18, AUS23, and AUS24], three from Waller Creek [WC19eWC21], and eight from the previously operated BAC filter [Kirisits et al., 2002; B2, B6eB11, and B15]). Based on batch studies, the isolates were placed into one of four groups (Table 2) with respect to their abilities to reduce bromate in the absence or presence of nitrate (NN medium and N medium, respectively), to reduce nitrate, and to accumulate nitrite. Results from one representative isolate of each group are plotted in Fig. 2. All isolates demonstrated bromate removal and bromide production (Fig. 2), and the uninoculated controls did not show bromate removal or bromide production (data not shown). All isolates reduced bromate in the absence of nitrate, and 14 of the 15 isolates reduced bromate in the presence of nitrate (Table 2). The bromate-reducing bacteria isolated by Hijnen et al. (1995) could only reduce bromate in the absence of nitrate; their results may have been predisposed to this outcome because their bromate-reducing bacteria were
Table 2 e Grouping of isolates based on reduction of bromate or nitrate and accumulation of nitrite. Group 1 2 3 4
Isolate
Reduces bromate in absence of nitrate
Reduces bromate in presence of nitrate
Reduces nitrate
Accumulates nitrite
AUS18, AUS23, WC19, WC20, B2, B6, B8, B9, B15 AUS24, B7 WC21, B10, B11 AUS11
þ
þ
þ þ þ
þ þ
þ þ
þ
6056
400
4
300
3
200
2
100
1
0
0 0
20
40
60
80
100
Time (days) 400
6 5
300
4 3
200
2
100
1 0
0 0
20
40 60 Time (days)
80
d
100
600
6
500
5
400
4
300
3
200
2
100
1 0
0 0
20
40
60
80
100
Time (days) 500
6
400
5 4
300
3 200
2
100
1
0
0 0
20
40 60 Time (days)
80
Nitrate or nitrite (mg/L)
5
Bromate or bromide (μg/L)
500
b Nitrate or nitrite (mg/L)
6
Bromate or bromide (μg/L)
Bromate or bromide (μg/L)
c
600
Nitrate or nitrite (mg/L)
Bromate or bromide (μg/L)
a
Nitrate or nitrite (mg/L)
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 5 1 e6 0 6 2
100
Fig. 2 e Batch studies with bromate-reducing isolates in which bromate, bromide, nitrate, and nitrite were monitored. Representative isolates from each group in Table 2 are shown: (a) Isolate B2; (b) Isolate B7; (c) Isolate WC21; (d) Isolate AUS11. (-) Bromate in the absence of a nitrate spike; (,) bromate in the presence of a nitrate spike; (C) bromide in the absence of a nitrate spike; (6) nitrate; (3) nitrite.
isolated from a denitrifying inoculum. Observation of complete reduction of bromate to bromide (closure of the Br mass balance) in the current study was precluded by the presence of an anion with a similar retention time to bromide in the ion chromatographic analysis. However, others have observed the complete microbial reduction of bromate to bromide (Hijnen et al., 1995; van Ginkel et al., 2005b). Five of the 15 isolates reduced nitrate, and three of these isolates accumulated nitrite (Table 2). Some of the nitratereducing isolates (i.e., B7, B10, and B11) were tested for the ability to denitrify, but none showed evidence of gas production (e.g., N2O, N2) under nitrate-reducing conditions. Aforementioned, cometabolism of bromate via nitrate reductase and (per)chlorate reductase has been postulated in the literature. Our data do not refute that possibility because the nitrate-reducing isolates also were able to reduce bromate (Table 2, Fig. 2b,c). However, our data suggest that a separate bromate-reduction pathway also might exist because five of our isolates (B2, B6, B8, B9, B15) were unable to reduce nitrate (Table 2) or perchlorate (data not shown) but were able to reduce bromate (Table 2, Fig. 2a); thus, it appears that these isolates reduced bromate in the absence of nitrate and (per) chlorate reductases.
3.2.
Phylogeny of bromate-reducing bacteria
The phylogenetic relationships of the 15 bromate-reducing isolates to known bacterial strains, including other known bromate-reducers, denitrifiers, and (per)chlorate-reducers in the GenBank database, are represented in a tree reconstruction based on alignments of their 16S rRNA gene sequences (>1200 bp) (Fig. 3). The phylogenetic affiliations of the isolates
vary widely, with four phyla represented by the 15 isolates (i.e., Actinobacteria, Bacteroidetes, Firmicutes, and Proteobacteria). Phylogenetic diversity was evident within a single inoculum; for instance, the isolates from the previously operated BAC filter (Kirisits et al., 2002) spanned three phyla. Of the 15 bromate-reducing isolates investigated here, the Proteobacteria was the most common phylum represented, with the bulk of them representing the Gammaproteobacteria. Relatedly, Heylen et al. (2006) found that most denitrifiers in activated sludge are from the Alphaproteobacteria and Betaproteobacteria, but they also found denitrifiers representing the Gammaproteobacteria, Epsilonproteobacteria, Firmicutes, and Bacteroidetes. Thus, both bromate-reducing bacteria and denitrifying bacteria show broad phylogenetic diversity, containing Gram negative and Gram positive organisms from multiple phyla. For each bromate-reducing isolate, Table 3 summarizes the closest match and percent identity (16S rRNA gene sequence) to a cultivated organism in GenBank. In this section, each bromate-reducing isolate is discussed relative to its closest cultivated match. Seven of the bromate-reducing isolates obtained in this study were phylogenetically related to Gram negative members of the Gammaproteobacteria (Table 3). Consistent with the findings for two different strains of Pseudoxanthomonas mexicana (Thierry et al., 2004), our data show that B2 and B9 are unable to reduce nitrate (Table 2). Although the 16S rRNA gene sequences of isolates B10 and AUS11 were very similar to one another (Fig. 3), these isolates (from BAC filters treating different tapwaters) did show physiological differences. B10 reduced nitrate, but AUS11 did not (Table 2); this variability in nitrate reduction is consistent with the Stenotrophomonas literature (Heylen et al., 2007; Guzik et al., 2009). The three confirmed bromate-reducing Pseudomonas
6057
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 5 1 e6 0 6 2
(Sub-)Phylum Cyanobacteria Bactero oidetes
Microcystis aeruginosa (NC 010296)
• AUS23 (JF951735) *
Chryseobacterium sp. YJ1 (DQ521273.1)
AUS18 (JF951734)*
• WC21 (JN002283)*
Pseudomonas syringae isolate 21 (EU438852.1)
P. rhodesiae str. NO5 (FJ462694.1)
WC19 (JF951737) (JF951737)* P. fluorescens str. P17 (EF552157.1)* P. fluorescens str. Pf-5 (NC 004129) P. chloritidismutans str. ASK-1 (AY277620)**
• •WC20 (JF951738)*
P. chloritidismutans str. AW-1 (AY017341)**
γ−
Aeromonas hydrophila y p str. 45-90 (AF468055.1) ( )
B10 (JF951741)* bacteria Proteob
•
Stenotrophomonas maltophilia str. 1.22 (EF426435)
AUS11 (JF951733)* B9 (JF951740) (JF951740)* Pseudoxanthomonas mexicana (AF27308.1)
B2 (JF951739)* B11 (JF951742)*
•
•
Acidovorax sp. R24667 (AM084010)
B7 (AF442523) (AF442523)* Ralstonia eutropha str. H16 (NC008314)
β−
Azospira oryzae str. GR-1 (AY277622)** Dechloromonas hortensis (AY277621)** Dechloromonas sp. PC-1 (AY126452.1)*
• AUS24 (JF951736) * • B6 (AF442522)*
Bacillus cereus str. G8639 (AY138271)
•
Rhodococcus erythropolis (AJ717371)
B8 (AF442524) (AF442524)* Rhodococcus sp. YIM C683 (EU135645.1)
B15 (JF951743)*
α− Actinobac cteria Firmicutes
Rhizobium loti ((U50164.1))
0.05 Fig. 3 e Phylogenetic tree showing the relationship of 15 new bromate-reducing bacterial isolates (bold-face) with reference bacterial strains. The tree was constructed using the neighbor-joining distance method based on the alignment of >1200 bp of 16S rRNA gene fragments. Isolates (intact cells) demonstrating bromate-reducing activity are indicated with one asterisk (*). Cell extracts or purified enzymes demonstrating bromate-reducing activity are indicated with two asterisks (**). Phylum groups are indicated by brackets, and accession numbers are provided. Scale bar represents 5% sequence dissimilarity, and nodes with >50% bootstrap support are indicated (C, n [ 1000). The outgroup is Microcystis aeruginosa, a member of the Cyanobacteria.
6058
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 5 1 e6 0 6 2
Table 3 e Summary of bromate-reducing isolates and corresponding matches to known cultivated species in GenBank. Isolate source
Isolate
Sequence length (bp)
Closest BLAST match (% identity)
Previously operated BAC filters treating tapwater derived from groundwater (Kirisits et al., 2002)
B2 B6 B7 B8 B9 B10 B11 B15
1503 1486 1509 1481 1502 1415 1358 1305
Pseudoxanthomonas mexicana (AF273082.1) (100%) Bacillus cereus str. G8639 (AY138271) (99%) Acidovorax sp. R-24667 (AM084010) (99%) Rhodococcus erythropolis (AJ717371) (99%) Pseudoxanthomonas mexicana (AF273082.1) (100%) Stenotrophomonas sp. JRL-2 (AF181569.1) (99%) Acidovorax sp. B-206 (EF596912.1) (99%) Rhodococcus sp. YIM C683 (EU135645.1) (99%)
Gammaproteobacteria Firmicutes Betaproteobacteria Actinobacteria Gammaproteobacteria Gammaproteobacteria Betaproteobacteria Actinobacteria
BAC filters operated in current study treating tapwater derived from surface water
AUS11 AUS18 AUS23 AUS24
1441 1382 1403 1347
Stenotrophomonas maltophilia str. 1.22 (EF426435) (99%) Chryseobacterium sp. YJ1 (DQ521273.1) (98%) Chryseobacterium sp. YJ1 (DQ521273.1) (98%) Rhizobium loti (U50164.1) (97%)
Gammaproteobacteria Bacteroidetes Bacteroidetes Alphaproteobacteria
Waller Creek water
WC19 WC20 WC21
1437 1217 1303
Pseudomonas rhodesiae str. NO5 (FJ462694.1) (99%) Aeromonas hydrophila str. 45/90 (AF468055.1) (99%) Pseudomonas syringae isolate 21 (EU438852.1) (97%)
Gammaproteobacteria Gammaproteobacteria Gammaproteobacteria
microorganisms are reducing bromate in a variety of ways: cometabolically or through a specific bromate-reduction pathway. Additional studies are needed to carefully address the mechanism of microbial bromate reduction.
3.3.
Control filter experiments
Abiotic and limited biotic reduction of bromate by GAC has been observed at DO concentrations of w8 mg/L (Kirisits et al., 2000, 2001). Thus, two filters (A and B) were operated with a reduced influent DO concentration (2.1 mg/L) to promote simultaneous abiotic and biotic reduction of bromate. The duplicate filters showed similar bromate breakthrough curves (Fig. 4); the parametric t-test was conducted on these data to compare the mean of the data from each control filter with a resulting p-value of 0.368, suggesting that the two breakthrough curves are statistically indistinguishable from each other. Furthermore, the linear regression analysis conducted on the initial linear portion of these breakthrough curves (between 0 and 11,300 bed volumes) showed that the slope
100 Percent Bromate Remaining (C/Co x 100%)
strains isolated by Hijnen et al. (1995) and bromate-reducing Pseudomonas fluorescens P17 are denitrifiers; while WC21 reduced nitrate, WC19 did not (Table 2). Although our data did not show nitrate reduction by isolate WC20 (Table 2), a strain of Aeromonas hydrophila has been reported to grow anaerobically with nitrate as the terminal electron acceptor (Knight and Blakemore, 1998). Three of the bromate-reducers obtained in this study were phylogenetically related to Gram negative members of the Alphaproteobacteria and Betaproteobacteria (Table 3). AUS24 reduced nitrate (Table 2), as has been demonstrated previously for Rhizobium loti (Monza et al., 1992). Isolates B7 and B11 reduced nitrate, but B11 accumulated nitrite (Table 2), demonstrating strain-level physiological distinction for these two closely related strains derived from the same BAC filter. This is consistent with the literature, which shows that many Acidovorax strains can reduce nitrate, and some can reduce nitrite (Willems et al., 1990). The Betaproteobacteria also houses perchlorate-reducing bacteria, some of which have been shown to reduce bromate (e.g., Dechloromonas sp. PC-1 in Fig. 3). Isolates AUS18 and AUS23 were closest in identity to that of Chryseobacterium, which is a Gram negative genus of the phylum Bacteroidetes. Although Chryseobacterium species exhibit varied nitrate reduction capabilities, including denitrification (Kim et al., 2005), AUS18 and AUS23 did not reduce nitrate (Table 2). AUS18 and AUS23 were isolated from the same BAC filter and showed identical 16S rRNA sequences; thus, these isolates might be the same strain but additional physiological testing is needed for verification. Three of the bromate-reducing isolates obtained in this study were phylogenetically related to Gram positive bacteria belonging to the phyla Firmicutes and Actinobacteria. Although some Bacillus species are denitrifiers (Jones et al., 2011), isolate B6 did not show nitrate reduction in this study (Table 2). No evidence for nitrate reduction by Rhodococcus exists in the literature, which is consistent with our data for isolates B8 and B15 (Table 2). Overall, the bromate-reducing strains isolated in this study were phylogenetically diverse. It is possible that these diverse
Affiliated bacterial (sub-)division
80 60 40 20 0 0
5000
10000
15000
20000
25000
Cumulative Bed Volumes Fig. 4 e Bromate breakthrough curves for filters A and B under control conditions (nominal influent conditions: dechlorinated Austin tapwater, pH 7.5, 2 mg/L DO, 20 mg/L bromate; EBCT [ 20 min). C and Co represent the effluent and influent concentrations of bromate, respectively. (,) filter A; (-) filter B.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 5 1 e6 0 6 2
values are statistically significant ( p-value <0.001 and R2 > 0.9). The slopes for filters A and B were 0.0053 and 0.0052, respectively; the 95% confidence intervals for the slope values overlap (data not shown), suggesting that the breakthrough curves are very similar, thereby confirming the results of the ttest. These operating conditions allowed the 10 mg/L U.S. EPA drinking water standard to be met for approximately 10,000 bed volumes. After 15,000 bed volumes, bromate breakthrough began to plateau; the nonparametric ManneWhitney test was conducted due to the limited number of samples taken and demonstrated that the bromate breakthrough between 15,000 and 18,000 bed volumes was statistically indistinguishable from the bromate breakthrough between 18,000 and 21,800 bed volumes ( p-value ¼ 0.33). Thus, the plateau suggests that microorganisms account for approximately 33% bromate removal under the tested conditions, and we have previously demonstrated that bromate removal coincides with bromide production in BAC filters operated similarly to the BAC filters in the current study (Kirisits and Snoeyink, 1999; Kirisits et al., 2002). However, since bromate also can be abiotically reduced to bromide by GAC (Bao et al., 1999), it is possible that a fraction of the 33% bromate removal is abiotic in nature. To determine if bromate reduction could be increased through the addition of known bromate-reducing bacteria to the filters, two bioaugmentation experiments were undertaken. The eight bromate-reducing bacteria that had been isolated from a previously operated BAC filter (Kirisits et al., 2002; B2, B6eB11, B15) were used to bioaugment new filters (early-stage bioaugmentation of filters C, D, and E, Section 3.4) and filter B after 21,800 bed volumes of operation (late-stage bioaugmentation, Section 3.5).
3.4.
Early-stage bioaugmentation
The duplicate filters used for the early-stage bioaugmentation (filters C and D) showed similar bromate breakthrough curves (data not shown); the parametric t-test was conducted on these data with a resulting p-value of 0.314, suggesting that
Percent Bromate Remaining (C/Co x 100%)
100 80 60 40 20 0 0
3000
6000
9000
12000
Cumulative Bed Volumes Fig. 5 e Bromate breakthrough curves in filter B under control conditions as compared to early-stage bioaugmented filter C (nominal influent conditions: dechlorinated Austin tapwater, pH 7.5, 2 mg/L DO, 20 mg/L bromate; EBCT [ 20 min). C and Co represent the effluent and influent concentrations of bromate, respectively. (-) filter B; (B) filter C.
6059
the two breakthrough curves are statistically indistinguishable from each other. The early-stage bioaugmented filters C and D showed slightly higher bromate breakthrough as compared to the control filters A and B for the first 6000 bed volumes, with an average of 24% bromate remaining in the effluent for the early-stage bioaugmented filter C and 12% bromate remaining in the effluent for control filter B (Fig. 5). The parametric t-test conducted on these data resulted in a pvalue <0.001, which suggests that control filter B showed better bromate removal performance as compared to bioaugmented filter C during the first 6000 bed volumes of operation. It is possible that the R2A medium used to inoculate the isolates to the filters in the early-stage bioaugmentation experiments negatively impacted the abiotic reduction of bromate because organic matter in the medium might have blocked or occupied key functional sites on the GAC for bromate reduction. Other studies have shown decreased bromate reduction by GAC due to natural organic matter loading (Bao et al., 1999; Kirisits et al., 2000). After the initial 6000 bed volumes, the breakthrough curves for bioaugmented filter C and control filter B were similar (Fig. 5). The parametric t-test resulted in a one-sided p-value of 0.102, which suggests that the two breakthrough curves are statistically indistinguishable from each other after 6000 bed volumes. At that point, much of the labile organic matter associated with the bioaugmentation procedure might have desorbed from the GAC or been consumed by bacteria, allowing the convergence of the data sets from the two filters. This experiment indicates that the early-stage bioaugmentation of the filter with non-native bromate-reducing bacteria did not significantly improve bromate removal under the tested conditions. T-RFLP was used to examine changes in the microbial community of the early-stage bioaugmented filters C and D during the first three months of operation. After three months of operation, operational taxonomic units (OTUs) associated with bioaugmented isolates B6, B7, B10, and B11 were observed in the T-RFLP profiles and accounted for 27% of the total peak area in the electropherograms. The OTUs associated with the other four bioaugmented microorganisms were not observed. Since OTUs are not unique (i.e., multiple microorganisms can share the same terminal restriction fragment length), the presence of OTUs consistent with B6, B7, B10, and B11 does not guarantee the presence of those microorganisms. During operation, the bioaugmented filters also developed an indigenous microbial community, and the OTUs consistent with B6, B7, B10, and B11 also might have been due to native bacteria. Hence, the persistence of bioaugmented microorganisms in a complex indigenous community needs to be tracked in a more specific way, such as the chromosomal labeling of the microorganisms with a constitutive reporter (e.g., the green fluorescent protein). While some of the bioaugmented isolates might have persisted in the filter, this did not improve bromate removal (Fig. 5).
3.5.
Late-stage bioaugmentation
Subsequently, we examined the effect of late-stage bioaugmentation in an established BAC filter. For this, filter B was bioaugmented at 21,800 bed volumes, and bromate
6060
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 5 1 e6 0 6 2
breakthrough was compared to that from control filter A (Fig. 6). Prior to bioaugmentation, the GAC in filters A and B was agitated to release excess biomass. After filter cleaning, bromate breakthrough decreased slightly in both filters (Fig. 6), which might have been due to the rectification of short-circuiting in the bed. The bioaugmented bromatereducing isolates were inoculated to filter B in 1 PBS, to avoid the issues associated with the organic content of the inoculum used in the early-stage bioaugmentation. Sterile 1 PBS also was injected to filter A. Directly after bioaugmentation, bromate breakthrough increased in both filters to between 75 and 80% of the influent bromate concentration before returning to levels between 65 and 70% over the subsequent weeks (Fig. 6). It is possible that chloride in the 1 PBS solution (0.14 M chloride) could have been responsible for the temporary increase in bromate breakthrough due to ion exchange of chloride to the GAC (Kirisits et al., 2000) or to bacterial salinity shock. During the first 500 bed volumes following bioaugmentation, bromate breakthrough in bioaugmented filter B was 5% less than that from control filter A (Fig. 6). However, following this period, the bromate breakthrough curves were similar between filters A and B; the results of the nonparametric ManneWhitney test suggest that the bromate breakthrough curves from the two filters are statistically indistinguishable ( p-value >0.139). Thus, similar to the earlystage bioaugmentation results, the late-stage bioaugmentation did not significantly improve bromate removal under the tested conditions. Other bioaugmentation strategies might be tested to improve retention of the bioaugmented isolates in the BAC filter (e.g., increased concentration of bioaugmented strains, greater recirculation time prior to initiating normal column operation). On the other hand, the improvement of biological bromate reduction might need to be focused on the optimization of
other water quality parameters. For instance, if microbial bromate reduction proceeds primarily through a cometabolic route (i.e., nitrate/(per)chlorate reductase), then sufficient quantities of nitrate or (per)chlorate need to be present in the water and removed biologically during treatment to generate sufficient enzyme quantities. If the water contained low concentrations of nitrate or (per)chlorate, it might not be possible to achieve significant cometabolic reduction of bromate. If microbial bromate reduction proceeds primarily through a bromate-specific respiratory pathway, then electron donor supplementation (e.g., acetic acid) might be considered. This approach has been very successful for the biological removal of perchlorate (Brown et al., 2008). We suspect that cometabolic bromate reduction is important in the BAC filters described here. When we added 5 mg/L acetate to the influent of filter B (Table 1), which is w200% of the stoichiometric amount required for the reduction of the 2.1 mg/L DO, 0.2 mg/L nitrate-N, and 20 mg/L bromate present in the influent, bromate removal improved only mildly over three weeks of acetate dosing (from 20% to 40%, data not shown). Thus, if a portion of the bromate reduction occurring in the BAC filters is due to cometabolism by nitrate-reducing bacteria, then the low nitrate concentrations in the influent (0.2 mg/L nitrate-N) could be limiting bromate reduction. Additional work to elucidate the mechanism of bromate reduction (i.e., respiratory versus cometabolic) is necessary to effect an increase in the relatively low bromate removals observed in these BAC filters. To verify that the bromate removal observed in the BAC filters was biological in nature, filter B was sacrificed after 50,700 bed volumes of operation. The BAC was autoclaved, and the column was operated with a high DO concentration to discourage microbial bromate reduction. Under those conditions, bromate breakthrough in the effluent reached 100% (data not shown), verifying that the steady state bromate removal obtained in the BAC filters in this study was microbial.
Percent Bromate Remaining (C/Co x 100%)
Filters Bioaugment cleaned 90
4.
80 70 60 50 16000
18000 20000 22000 Cumulative Bed Volumes
24000
Fig. 6 e Bromate breakthrough curves in filter B before and after late-stage bioaugmentation as compared to control filter A (nominal influent conditions: dechlorinated Austin tapwater, pH 7.5, 2 mg/L DO, 20 mg/L bromate; EBCT [ 20 min). C and Co represent the effluent and influent concentrations of bromate, respectively. Dashed lines indicate filter cleaning (at 20,500 bed volumes) and bioaugmentation of filter B (at 21,800 bed volumes). (-) filter B before bioaugmentation; (3) filter B after latestage bioaugmentation; (,) filter A.
Conclusions
Our research shows that BAC filtration can provide sustainable bromate reduction in drinking water, but the process needs to be optimized to ensure that the U.S. EPA standard for bromate is met. Bioaugmentation of filters with bacteria known to reduce bromate to bromide did not significantly decrease start-up time or increase bromate removal as compared to control filters. However, bioaugmentation might be unnecessary because bromate-reducing bacteria are phylogenetically diverse (including Actinobacteria, Bacteroidetes, Firmicutes, and Proteobacteria), which suggests that the capability for bromate reduction is widespread in the environment. To optimize bromate reduction in a biological drinking water treatment process, the predominant mechanism of bromate reduction (i.e., cometabolic or respiratory) needs to be assessed so that appropriate measures can be taken to improve bromate removal. Given the current attention by the drinking water industry to anaerobic biological processes, the use of BAC filtration for bromate removal has potential in drinking water treatment.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 5 1 e6 0 6 2
Acknowledgments We thank Sherry Cheng, Vinh Do, and Dr. Susan De Long for laboratory assistance on this project and Bryant Chambers and Eric Solis for administrative assistance with the manuscript. Financial support for this work was provided by the University of Texas at Austin and the American Water Works Association (Larson Aquatic Research Support Scholarship to Andrew Davidson).
references
Bao, M.L., Griffini, O., Santianni, D., Barbieri, K., Burrini, D., Pantani, F., 1999. Removal of bromate ion from water using granular activated carbon. Water Research 33 (13), 2959e2970. Brown, J.C., Snoeyink, V.L., Kirisits, M.J., 2002. Abiotic and biotic perchlorate removal in an activated carbon filter. Journal of the American Water Works Association 94 (2), 70e79. Brown, J., Lauderdale, C., Estavo, G., Ettori, A., Shih, W., Poust, S., Walker, S., Raskin, L., Upadhyaya, G., Li, X., Morgenroth, E., 2008. Direct Fixed-bed Biological Perchlorate Destruction Demonstration. Final report ER-0544. Arlington, Virginia: Environmental Security Technology Certification Program. Butler, R., Ehrenberg, S., Godley, A.R., Lake, R., Lytton, L., Cartmell, E., 2006. Remediation of bromate-contaminated groundwater in an ex situ fixed-film bioreactor. Science of the Total Environment 366 (1), 12e20. Cotruvo, J.A., Keith, J.D., Bull, R.J., Pacey, G.E., Gordon, G., 2010. Bromate reduction in simulated gastric juice. Journal of the American Water Works Association 102 (11), 77e86. DeAngelo, A.B., George, M.H., Kilburn, S.R., Moore, T.M., Wolf, D. C., 1998. Carcinogenicity of potassium bromate administered in the drinking water to male B6C3F1 mice and F344/N rats. Toxicologic Pathology 26 (5), 587e594. Downing, L.S., Nerenberg, R., 2007. Kinetics of microbial bromate reduction in a hydrogen-oxidizing, denitrifying biofilm reactor. Biotechnology and Bioengineering 98 (3), 543e550. Egert, M., Friedrich, M., 2005. Post-amplification Klenow fragment treatment alleviates PCR bias caused by partially single-stranded amplicons. Journal of Microbiological Methods 61 (1), 69e75. Fries, M., Zhou, J., Chee-Sanford, J., Tiedje, J., 1994. Isolation, characterization, and distribution of denitrifying toluene degraders from a variety of habitats. Applied and Environmental Microbiology 60 (8), 2802e2810. , I., Wojcieszyn ska, D., Łabuzek, _ Guzik, U., Gren S., 2009. Isolation and characterization of a novel strain of Stenotrophomonas maltophilia possessing various dioxygenases for monocyclic hydrocarbon degradation. Brazilian Journal of Microbiology 40, 285e291. Heylen, K., Vanparys, B., Peirsegaele, F., Lebbe, L., De Vos, P., 2007. Stenotrophomonas terrae sp. nov. and Stenotrophomonas humi sp. nov., two nitrate-reducing bacteria isolated from soil. International Journal of Systematic and Evolutionary Microbiology 57 (9), 2056e2061. Heylen, K., Vanparys, B., Wittebolle, L., Verstraete, W., Boon, N., De Vos, P., 2006. Cultivation of denitrifying bacteria: optimization of isolation conditions and diversity study. Applied and Environmental Microbiology 72 (4), 2637e2643. Hijnen, W.A.M., Jong, R., van der Kooij, D., 1999. Bromate removal in a denitrifying bioreactor used in water treatment. Water Research 33 (4), 1049e1053. Hijnen, W.A.M., Vooght, R., Veenendaal, H.R., van der Jagt, H., van der Kooij, D., 1995. Bromate reduction by denitrifying bacteria. Applied and Environmental Microbiology 61 (1), 239e244.
6061
Jones, C.M., Welsh, A., Throba¨ck, I.N., Do¨rsch, P., Bakken, L.R., Hallin, S., 2011. Phenotypic and genotypic heterogeneity among closely related soil-borne N2- and N2O-producing Bacillus isolates harboring the nosZ gene. FEMS Microbiology Ecology 76 (3), 541e552. Kengen, S., Rikken, G., Hagen, W., van Ginkel, C.G., Stams, A., 1999. Purification and characterization of (per)chlorate reductase from the chlorate-respiring strain GR-1. Journal of Bacteriology 181 (21), 6706e6711. Kim, K.K., Bae, H.-S., Schumann, P., Lee, S.-T., 2005. Chryseobacterium daecheongense sp. nov., isolated from freshwater lake sediment. International Journal of Systematic and Evolutionary Microbiology 55 (1), 133e138. Kim, K., Logan, B.E., 2001. Microbial reduction of perchlorate in pure and mixed culture packed-bed bioreactors. Water Research 35 (13), 3071e3076. Kirisits, M.J., Snoeyink, V.L., 1999. Reduction of bromate in a BAC filter. Journal of the American Water Works Association 91 (8), 74e84. Kirisits, M.J., Snoeyink, V., Chee-Sanford, J., Daugherty, B., Brown, J., Raskin, L., 2002. Effect of operating conditions on bromate removal efficiency in BAC filters. Journal of the American Water Works Association 94 (4), 182e193. Kirisits, M.J., Snoeyink, V.L., Inan, H., Chee-Sanford, J.C., Raskin, L., Brown, J.C., 2001. Water quality factors affecting bromate reduction in biologically active carbon filters. Water Research 35 (4), 891e900. Kirisits, M.J., Snoeyink, V.L., Kruithof, J.C., 2000. The reduction of bromate by granular activated carbon. Water Research 34 (17), 4250e4260. Knight, V., Blakemore, R., 1998. Reduction of diverse electron acceptors by Aeromonas hydrophila. Archives of Microbiology 169 (3), 239e248. Krasner, S.W., Glaze, W.H., Weinberg, H.S., Daniel, P.A., Najm, I., 1993. Formation and control of bromate during ozonation of waters containing bromide. Journal of the American Water Works Association 85 (1), 73e81. Martin, K.J., Downing, L.S., Nerenberg, R., 2009. Evidence of specialized bromate-reducing bacteria in a hollow fiber membrane biofilm reactor. Water Science and Technology 59 (10), 1969e1974. Monza, J., Delgado, M.J., Bedmar, E.J., 1992. Nitrate reductase and nitrite reductase activity in free-living cells and bacteroids of Rhizobium loti. Plant and Soil 139 (2), 203e207. Morpeth, F.F., Boxer, D.H., 1985. Kinetic analysis of respiratory nitrate reductase from Escherichia coli K12. Biochemistry 24 (1), 40e46. Ridley, H., Watts, C.A., Richardson, D.J., Butler, C.S., 2006. Resolution of distinct membrane-bound enzymes from Enterobacter cloacae SLD1a-1 that are responsible for selective reduction of nitrate and selenate oxyanions. Applied and Environmental Microbiology 72 (8), 5173e5180. Staley, J., 1968. Prosthecomicrobium and Ancalomicrobium: new prosthecate freshwater bacteria. Journal of Bacteriology 95 (5), 1921e1942. Thierry, S., Macarie, H., Iizuka, T., Geißdo¨rfer, W., Assih, E.A., Spanevello, M., Verhe, F., Thomas, P., Fudou, R., Monroy, O., Labat, M., Ouattara, A.S., 2004. Pseudoxanthomonas mexicana sp. nov. and Pseudoxanthomonas japonensis sp. nov., isolated from diverse environments, and emended descriptions of the genus Pseudoxanthomonas Finkmann et al. 2000 and of its type species. International Journal of Systematic and Evolutionary Microbiology 54 (6), 2245e2255. van Ginkel, C.G., Middelhuis, B.J., Spijk, F., Abma, W.R., 2005a. Cometabolic reduction of bromate by a mixed culture of microorganisms using hydrogen gas in a gas-lift bioreactor. Journal of Industrial Microbiology and Biotechnology 32 (1), 1e6.
6062
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 5 1 e6 0 6 2
van Ginkel, C.G., van Haperen, A.M., van der Togt, B., 2005b. Reduction of bromate to bromide coupled to acetate oxidation by anaerobic mixed microbial cultures. Water Research 39 (1), 59e64. von Gunten, U., 2003. Ozonation of drinking water: part II. Disinfection and by-product formation in presence of bromide, iodide, or chlorine. Water Research 37 (7), 1469e1487. Webster, T.S., Guarini, W.J., Wong, H.S., 2009. Fluidized bed bioreactor treatment of perchlorate-laden groundwater to potable standards. Journal of the American Water Works Association 101 (5), 137e151.
Willems, A., Falsen, E., Pot, B., Jantzen, E., Hoste, B., Vandamme, P., Gillis, M., Kersters, K., De Ley, J., 1990. Acidovorax, a new genus for Pseudomonas facilis, Pseudomonas delafieldii, E. Falsen (EF) group 13, EF group 16, and several clinical isolates, with the species Acidovorax facilis comb. nov., Acidovorax delafieldii comb. nov., and Acidovorax temperans sp. nov. International Journal of Systematic Bacteriology 40 (4), 384e398. Yamamoto, I., Shimizu, H., Tsuji, T., Ishimoto, M., 1986. Purification and properties of nitrate reductase from Mitsuokella multiacidus. Journal of Biochemistry 99 (3), 961e969.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 6 3 e6 0 7 3
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Bacterial community characteristics under long-term antibiotic selection pressures Dong Li, Rong Qi, Min Yang*, Yu Zhang, Tao Yu State Key Lab of Environmental Aquatic Chemistry, Research Center for Eco-Environmental Sciences, Chinese Academy of Sciences, Beijing 100085, China
article info
abstract
Article history:
To investigate bacterial community characteristics under long-term antibiotic selection
Received 25 February 2011
pressures, water samples from the upstream and the downstream sections of two rivers
Received in revised form
individually receiving the treated penicillin G and oxytetracycline production wastewater,
30 August 2011
as well as the anaerobic and the aerobic effluent of the penicillin G production wastewater
Accepted 1 September 2011
treatment plant, were taken and analyzed. Antibiotic resistance ratios of bacterial
Available online 10 September 2011
communities in water samples were estimated by culture-based analysis. The majority of bacterial colonies (approximately 55%e70%) in both downstream rivers and the aerobic
Keywords:
effluent showed resistance to 80 mg/ml of antibiotics tested, while the resistance ratios
Antibiotic resistance
were less than 10% and 5% respectively for both upstream rivers. Six 16S rRNA gene clone
Bacterial community
libraries were constructed with 355 sequences and 215 OTUs totally obtained representing 465 clones. The antibiotic stresses seemed not reduce the diversities of bacterial communities in antibiotic containing water samples compared to those in the two reference upstream rivers. Bacterial groups present in the two reference upstream rivers were common residents in freshwater ecosystems, with the dominant groups as the phyla Proteobacteria including Alphaproteobacteria, Betaproteobacteria and Gammaproteobacteria, as well as Actinobacteria and Bacteroidetes. The phyla Proteobacteria and Firmicutes were dominant in all antibiotic containing water samples, with the clones belonged to Deltaproteobacteria and Epsilonproteobacteria significantly abundant, as well as Gram-positive low GC bacteria in the classes Clostridia and Bacilli. It thus seemed that Deltaproteobacteria, Epsilonproteobacteria, Clostridia and Bacilli might be specifically associated with antibiotic containing environments. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
Environments containing high concentrations of antibiotics are normally anthropogenic, such as municipal, hospital, stock raising and pharmaceutical producing wastewater, as well as polluted surface water and fishery ponds. Antibiotics in these aquatic environments could directly act on bacterial strains and exert selective pressure, resulting in the change of
bacterial community structure inevitably. However, little is known about the taxonomic composition of the whole bacterial community in antibiotic polluted environments. Although numerous researches have been performed to investigate antibiotic resistance characteristics of bacterial isolates from different environmental sources including stock farms, poultry farms, fisheries, surface water, and lakes (Wittwer et al., 2005; Huddleston et al., 2006; Jindal et al., 2006;
* Corresponding author. Tel./fax: þ86 10 62923475. E-mail address: [email protected] (M. Yang). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.09.002
6064
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 6 3 e6 0 7 3
Hamelin et al., 2007), most of these researches were focused on human related pathogens (Huddleston et al., 2006; Hamelin et al., 2007), and the remaining generally isolated environmental bacteria using non-selective rich nutrient media at aerobic conditions (Miranda and Zemelman, 2002; Messi et al., 2005). Gammaproteobacteria like Pseudomonas spp. was generally the dominant bacterial group in these culture-based studies, mainly due to the favored growth of Gammaproteobacteria in nutrient rich culture media. Several other bacterial groups such as Bacteroidetes, Actinobacteria, Betaproteobacteria and Alphaproteobacteria were also recovered with comparably lower colony numbers. As the majority of bacterial species in environments still could not be isolated and cultured, the above culture-based researches could not reflect the actual composition of the whole community. Some particular bacterial groups which are difficult to be cultured at normal conditions might be dominant in environments containing antibiotics. Furthermore, under the selection pressure of antibiotics, most of bacterial strains might become resistant to antibiotics. Some metagenomic researches have demonstrated that the diversity of antibiotic resistance genes in environments is greater than previously accounted for basing on cultured bacteria, indicating that many resistance genes are actually carried by uncultured bacteria (Riesenfeld et al., 2004; D’Costa et al., 2006; Sommer et al., 2009). The whole bacterial community including both cultured and uncultured bacteria constitutes an important reservoir of antibiotic resistance genes which could furthermore move into pathogens via horizontal gene transfer facilitated with mobile gene elements such as plasmids, transposons and so on (Thomas and Nielsen, 2005), as confirmed by numerous investigations about antibiotic resistant pathogenic bacteria (Martı´nez et al., 2007). The elucidation of bacterial community composition in antibiotic containing environments would thus help to suggest the potential antibiotic resistant bacterial groups, which might be important carriers of resistance genes and sometimes the source of clinically important resistance genes (Martı´nez, 2008). To our knowledge, few researches have been performed to elucidate the whole bacterial community structures in antibiotic containing environments until now. Co´rdova-Kreylos and Scow have elucidated the effects of ciprofloxacin on salt marsh sediment microbial communities by using phospholipid fatty acid (PLFA) analysis (Co´rdova-Kreylos and Scow, 2007). However, their conclusions were acquired through lab-experiment, which might not be able to reflect the situations in actual environments. Thus in this study, two rivers individually receiving the treated penicillin G and oxytetracycline production wastewater, which were discharged from two antibiotic producing facilities and both contained significantly higher concentrations of antibiotics than normal aquatic environments (Li et al., 2008a,b), were selected to investigate bacterial community characteristics under longterm antibiotic stresses. Considering that many environmental factors might influence bacterial community structures, several water samples from the penicillin G production wastewater treatment facility were also obtained and analyzed to compare with river samples. The antibiotic resistance ratios of bacterial communities in water samples
were first estimated using culture-based analysis. Clone libraries of 16S rRNA gene which could provide the detailed and reliable information were then constructed for water samples from each site. The results would help to complement existing knowledge of bacterial community composition in antibiotic containing environments and suggest the possible environmental antibiotic resistant groups still unknown until now.
2.
Materials and methods
2.1.
Study site and sampling
The two rivers and wastewater treatment plants (WWTPs) were all located in Hebei Province, China. Penicillin G production wastewater is treated in the WWTP which included an anaerobic digestion, a hydrolyzation and two aerobic reactors successively. The treated effluent is discharged into the receiving river, Wangyang River. Meanwhile, the production wastewater from another oxytetracycline producing facility is treated in the WWTP including a sequence batch reactor and a continuous-flow activated sludge reactor, and then discharged into the Xiao River. In December 2004, April and August 2005, surface water samples from upstream (longitude 114 420 1300 E and latitude 37 590 800 N for Wangyang River; longitude 114 270 1900 E and latitude 38 20 400 N for Xiao River) and downstream sections (longitude 114 530 2000 E and latitude 37 520 3800 N for Wangyang River; longitude 114 340 3900 E and latitude 37 510 5000 N for Xiao River) of wastewater discharging points, as well as the effluent samples from the anaerobic and aerobic apartments of the penicillin G production wastewater treatment plant were all taken in 4-L brown glass bottles. Water samples were kept at 4 C in the darkness for at most two days. Upstream and downstream sampling sites were individually approximately 5 km and 30 km away from the discharging point at the Wangyang River, and 5 km and 20 km away at the Xiao River. Penicillin G and oxytetracycline residues in water samples were all determined using LC-ESI-MS. The detailed analysis methods and the characteristics of water samples could be found elsewhere (Li et al., 2008a,b).
2.2.
Culture-based analysis
Water samples from the upstream and downstream sections of the two rivers, as well as the aerobic effluent of the penicillin G production wastewater treatment plant were applied for culture-based analysis. Two sets of plates were simultaneously incubated for each water sample at proper dilution using non-selective Tryptic soy agar (TSA) at 30 C for 24 h aerobically. Control plate was not added with any antibiotics, then 80 mg/ml ampicillin was added in agar for the upstream and downstream water samples of penicillin G containing river, as well as the aerobic effluent, and 80 mg/ml oxytetracycline was added for the upstream and downstream water samples of oxytetracycline containing river. Ampicillin was used instead of penicillin G due to its broad-spectrum ability. The percentages of the colony numbers of plates added with antibiotics accounting for those of the corresponding control
6065
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 6 3 e6 0 7 3
Table 1 e Distribution of phylogenetic groups among bacterial 16S rRNA gene clone libraries for the upstream and downstream of the two rivers individually receiving treated penicillin G and oxytetracycline production wastewater, as well as the anaerobic and the aerobic effluent of penicillin G production wastewater. Phylum
Proteobacteria
Firmicutes
Actinobacteria Planctomycetes Acidobacteria Bacteroidetes
Chloroflexi Gemmatimonadetes Verrucomicrobia Lentisphaerae Unclassified Total no.
Class
AlphaBetaGammaDeltaEpsilonUnclassified Clostridia Bacilli Unclassified Actinobacteria Planctomycetacia Acidobacteria Bacteroidetes Flavobacteria Sphingobacteria Unclassified Anaerolineae Unclassified Gemmatimonadetes Verrucomicrobiae Lentisphaerae
No. of OTUs (no. of clones) Upstream river1a
Upstream river2b
Downstream river1a
Downstream river2b
Anaerobic effluent
Aerobic effluent
5 (12) 3 (13) 2 (9) e e e 1 (1) 1 (5) e 7 (24) 1 (3) 1 (6) e e 1 (1) e e e e e e 1 (1) 23 (75)
9 (13) 10 (32) 2 (3) e e 1 (1) e e e 1 (3) 2 (3) 3 (3) 1 (1) 2 (5) 5 (5) 1 (1) e e e 1 (1) e 1 (2) 39 (73)
5 (6) 9 (15) 1 (2) 5 (10) 1 (3) e 8 (20) 3 (10) 2 (2) 2 (3) e 2 (2) 1 (1) 1 (2) 1 (1) 1 (2) e e 1 (1) e e 2 (2) 45 (82)
3 (3) 3 (6) 1 (2) 6 (12) 3 (19) 1 (1) 9 (12) 1 (5) e 1 (2) e e 3 (6) e e 3 (3) e e e 1 (1) 1 (1) 3 (4) 39 (77)
1 (1) 1 (2) e 2 (8) 5 (11) e 12 (43) 3 (7) 1 (4) e e e e e e e e e e e e 1 (1) 26 (77)
1 (1) 13 (18) e 5 (9) 5 (13) 1 (3) 2 (3) 5 (18) 1 (2) e 2 (2) e 2 (3) e 1 (1) 1 (3) 1 (1) 1 (1) e e e 3 (3) 43 (81)
a The river1 received treated penicillin G production wastewater. b The river2 received treated oxytetracycline production wastewater.
Fig. 1 e Multiple correspondence analysis of bacterial groups and water samples including upstream and downstream of the two rivers 1 and 2 individually receiving treated penicillin G and oxytetracycline production wastewater, as well as the anaerobic and the aerobic effluent of penicillin G production wastewater. Only bacterial groups with sufficient sample sizes were shown.
6066
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 6 3 e6 0 7 3
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 6 3 e6 0 7 3
plates were used to reflect antibiotic resistance ratios of bacterial community in water samples.
2.3. DNA extraction, PCR, cloning and sequencing of 16S rRNA genes 500-ml water sample was vacuum filtered through a 0.2-mmpore-size polyethersulfone membrane filter. Then samples were washed by 10 ml of phosphate buffer (pH 8.0), and DNA was extracted using the cetyltrimethylammonium bromide (CTAB) method as described previously (Crump et al., 1999). The yield and quality of DNA was estimated visually after electrophoresis on 1% (wt/vol) agarose gel through comparison with a molecular mass ladder. The 16S rRNA gene was amplified using bacteria universal primers 27F (50 -AGAGTTTGATCCTGGCTCAG) and 1492R (50 TACGGYTACCTTGTTACGACTT) (Lane, 1991). The standard 50 ml PCR mixture (Takara, Dalian, China) included 1 PCR buffer containing 1.5 mM MgCl2, 200 mM of each deoxynucleoside triphosphate, 10 pmol of each primer, 1.25 U of TaKaRa rTaq polymerase, and approximately 50 ng of template DNA. PCR conditions were as follows: 95 C for 10 min, followed by 30 cycles of 95 C for 1 min, 55 C for 1 min, and 72 C for 1 min 30 s, and a final extension at 72 C for 15 min. After confirmed by electrophoresis in 1.2% (wt/vol) agarose gel, amplification products were purified with the Qiaquick PCR cleanup kit (Qiagen, Inc., Chatsworth, Calif.). In order to minimize PCR bias in subsequent cloning steps, three separate reactions were run for each sample and pooled together, PCR products of the samples from the same sampling site was also pooled together. The amplified 16S rRNA gene products were further cloned into the TOPO TA cloning vector pCR2.1, and TOP10 E. coli transformants were further selected according to manufacture’s instructions (Invitrogen). Cloned inserts were amplified from lysed colonies by PCR with plasmid-vector specific primers M13F and M13R under the same conditions with above. PCR products were digested (3 h, 37 C) with HaeIII (Takara, Dalian, China) and separated by electrophoresis through 2% agarose gels. Clones were grouped based on RFLP patterns, and representative clones were sequenced with an ABI 3730 automated sequencer (Invitrogen, Shanghai, China).
2.4.
Phylogenetic and statistical analysis
DNA sequences were edited manually using BioEdit (Hall, 1999), and then searched against RDP II and the GenBank database (Altschul et al., 1997; Cole et al., 2007). The most similar reference sequences were retrieved and aligned with clone sequences using ClustalX (Thompson et al., 1997). Phylogenetic trees were constructed using MEGA, version 3.1 by the neighbor-joining algorithm and the Jukes-Cantor
6067
distance estimation method (Kumar et al., 2004). Possible chimeras were checked using CHIMERA_CHECK in RDP II. The sequences sharing 97% or greater similarity were grouped into the same operational taxonomic unit (OTU) and the OTU number was determined using DOTUR (Schloss and Handelsman, 2005). OTU richness SChao1 and SACE as well as Shannon diversity (H ) were calculated using EstimateS version 8.0 with 100 random sample repetitions (Colwell, 2005). Evenness (E ) indices were calculated as follows: E ¼ H/lnn, where n is OTU number. Coverage (C ) was calculated as follows: C ¼ 1 e (n1/N ), where n1 is the number of OTUs that occurred once and N is the total number of clones. Rarefaction curves were constructed using aRarefactWin (http://www.uga.edu/wstrata/ software.html). UniFrac computational analysis was performed to compare clone libraries from different sampling sites (Lozupone and Knight, 2005). Clone libraries from different sites were clustered by the application of the UPGMA method to the UniFrac metric matrix, and principal coordinate analysis (PCoA) was also performed with UniFrac metric matrix. Multiple correspondence analysis of specific bacterial groups with a sufficient sample size (5 or more clones) and water samples was also performed by using the SPSS version 16.0 release.
2.5.
Nucleotide sequence accession numbers
The 16S rRNA gene sequences could be accessed in the GenBank database under the accession numbers EU234086 to EU234324, EU864431 to EU864494 and FJ230892-FJ230941.
3.
Results
3.1.
Antibiotics and resistance ratios in water samples
The concentrations of penicillin G decreased from 72.6 3.7 mg/L in the anaerobic effluent to 1.68 0.48 mg/L in the aerobic effluent of the WWTP, and ranged from 0.35 mg/L to under the detection limit (0.031 mg/L) in the receiving river water samples (Li et al., 2008b). Meanwhile, oxytetracycline was determined in the downstream of Xiao River at 376.7 141.7 mg/L (Li et al., 2008a). No penicillin G or oxytetracycline could be detected in the reference upstream rivers. The antibiotic residual levels in this study were comparably much higher than those reported in normal aquatic environments previously (Kolpin et al., 2002). Antibiotic resistance ratios were roughly estimated for all the river water and the aerobic effluent samples using culturebased resistance assay. Approximately 65% of colonies from penicillin G containing downstream water and 70% of the aerobic effluent colonies showed resistance to 80 mg/ml of ampicillin, with the resistance percentages less than 10% for the reference upstream river. Then approximately 55% of colonies from the oxytetracycline containing downstream
Fig. 2 e Phylogenetic relationships of representative bacterial 16S rRNA gene sequences within the phylum Proteobacteria from clone libraries of this study determined by the neighbor-joining method. Bootstrap values of >50% (obtained with 1000 resamplings) are shown at nodes. The scale bar indicates 0.05 nucleotide substitution per site. The reference sequences are obtained from the Ribosomal Database Project II or GenBank. Methanobrevibacter ruminantium is used as an outgroup. GenBank accession numbers are in parentheses.
6068
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 6 3 e6 0 7 3
river showed resistance to 80 mg/ml of oxytetracycline, while the resistance ratio was less than 5% in the reference upstream river. The significant difference of resistance ratios between the antibiotic containing water samples and the reference upstream river samples indicated that the concentrations of penicillin G and oxytetracycline residues in the polluted rivers and wastewater were high enough for leading most of the bacteria population in water samples resistant to antibiotics, in addition with the fact that the concentrations of 80 mg/ml of antibiotics tested in this study were much higher than the relevant breakpoints recommended by CLSI Standards guidelines (Clinical and Laboratory Standards Institute, 2003). The CFU number of ampicillin resistant bacteria was 3.4 104 cfu/ml in penicillin G containing downstream water, 4.8 104 cfu/ml in the aerobic effluent, and less than 8.9 102 cfu/ml in the reference upstream river water. Meanwhile the CFU number of oxytetracycline resistant bacteria in the oxytetracycline containing downstream river was 7.7 102 cfu/ml, and less than 1 102 cfu/ml in the reference upstream river.
3.2.
16S rRNA clone libraries and statistical analysis
Six clone libraries were individually constructed for the upstream and downstream water samples of both rivers, as well as the anaerobic and aerobic effluent of the WWTP for treating penicillin G production wastewater. Total of 355 sequences were obtained and grouped into 215 OTUs, representing 465 clones derived from the six clone libraries of this study (Table 1). Possible chimeras were discarded. Several unique characteristics of the bacterial communities were observed for all antibiotics containing water samples, including both downstream rivers as well as the anaerobic and aerobic effluent, as illustrated in Fig. 1, of which dimension 1 explained 52.9% of the observed variation, and dimension 2 explained 21.3% of the variation. The most notable characteristics were the abundance of clones belonging to the deeply rooting classes Deltaproteobacteria and Epsilonproteobacteria, the total of which individually representing 15.9% and 40.3% of the clone numbers of the libraries for the penicillin G and oxytetracycline polluted downstream rivers, and accounting for 24.7% and 27.2% of the clones in the anaerobic and aerobic effluent libraries, respectively (Table 1). The clones grouped into Deltaproteobacteria in all antibiotic containing samples were mainly classified as Desulfovibrio spp., Desulfovibrio mexicanus, Desulfovibrio desulfuricans, Desulfobacter spp., Desulfobacter postgatei, Desulfomicrobium norvegicum, Desulfomicrobium escambiense, and Desulfomicrobium apsheronum (Fig. 2), all of which were sulfate- or sulfurreducing bacteria. These species are generally strictly anaerobic and gain energy by coupling the complete or partial oxidation of organic compounds or molecular hydrogen to the reduction of sulfur or sulfate generating hydrogen sulfide. Then all clones in the class Epsilonproteobacteria were classified into the genera Sulfurovum, Sulfurospirillum and Arcobacter, all of which are sulfur and hydrogen sulfide- or thiosulfateoxidizing bacteria, with nitrate or oxygen as electron acceptors. Additionally, the phylum Firmicutes including the classes Clostridia and Bacilli, both of which were Gram-positive low GC
bacteria, had become the second abundant bacterial group comprising of 39.0% and 22.1% of the clones in the libraries for penicillin G and oxytetracycline polluted downstream rivers, respectively (Table 1). Meanwhile, the clones belonged to Clostridia and Bacilli were dominant in the anaerobic effluent library, and more or less as abundant as those belonging to Proteobacteria in the aerobic effluent library. The Bacilli clones in all these antibiotic containing water samples were further classified into the genera Trichococcus and Streptococcus, of which Trichococcus spp., especially Trichococcus flocculiformis were the dominant species among the Bacilli clones and appeared in almost all antibiotic containing samples (Fig. 3). T. flocculiformis is an aerotolerant, fermentative organism, and originally isolated from bulking sludge in Germany (Scheff et al., 1984). Furthermore, the majority of the clones (46.2%) in the class Clostridia have been classified into Mitsuokella spp. including Mitsuokella multacida. M. multacida was generally anaerobic rumen bacteria with phytase activity and was belonged to the Sporomusa subbranch of low GC Gram-positive bacteria. The remaining clones in the Clostridia were mainly affiliated with the anaerobic genera Aminobacterium, Desulfosporosinus, Anaerovorax, Acetobacterium, Mogibacterium, Dialister, Clostridium, Thermoanaerobacterium and Ruminococcus. Several genera were spore forming bacteria, such as Desulfosporosinus, Clostridium and Thermoanaerobacterium. The similarities of OTUs among all the antibiotic containing environments were determined using software DOTUR. Totally, 13 OTUs were shared by the bacterial community libraries of the penicillin G polluted downstream river and the two penicillin wastewater samples, and 7 OTUs were shared by those of the oxytetracycline polluted downstream river and the two penicillin wastewater samples. Then 5 OTUs were shared by those of the two antibiotic containing downstream rivers, and 4 OTUs were shared by the two wastewater libraries. Two OTUs were furthermore shared across all antibiotic containing aquatic environments. Most of these OTUs fell into Deltaproteobacteria, Epsilonproteobacteria, Clostridia and Bacilli, also suggesting that these bacterial groups might be specifically associated with antibiotic containing aquatic environments at both species and classes levels. Several other bacterial groups seemed also to be related to the antibiotic containing aquatic environments, such as the class Bacteroidetes, as well as the phyla Chloroflexi, Gemmatimonadetes, and Lentisphaerae (Fig. 1 and Table 1). However, as few clones were obtained for these bacterial groups and distributed dispersedly in wastewater and downstream river samples, no more confirmative relationships could be drawn between these bacterial groups and antibiotic containing environments. The bacterial community compositions in the two reference upstream river samples were also analyzed using clone libraries. Bacterial groups present were common residents in freshwater ecosystems, with the dominant groups as the phyla Proteobacteria including Alphaproteobacteria, Betaproteobacteria and Gammaproteobacteria, as well as Actinobacteria and Bacteroidetes (Table 1). The remaining clones were classified into the phyla Planctomycetes, Acidobacteria, Firmicutes and Verrucomicrobia, present at much lower abundance. Only 1 OTU was shared by the bacterial communities of both upstream rivers. UniFrac metric analysis has demonstrated
Fig. 3 e Phylogenetic relationships of representative bacterial 16S rRNA gene sequences within the other phyla from clone libraries of this study determined by the neighbor-joining method. Bootstrap values of >50% (obtained with 1000 resamplings) are shown at nodes. The scale bar indicates 0.05 nucleotide substitution per site.
6070
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 6 3 e6 0 7 3
showed that the diversity normally decreased significantly with the treatment of antibiotics (Antonopoulos et al., 2009; Rea et al., 2011). The comparably short exposure time of human gut microbiome should be the major reason for the significant reduce of bacterial diversity. It should be noted that as rarefaction curves of all clone libraries in this study did not reach saturation (data not shown), the clone number for each sample was still not sufficient and may affect the indices values.
4. Fig. 4 e Bacterial clone libraries clustering using UPGMA method to UniFrac metric based on all 16S rRNA gene sequences from the upstream and downstream clone libraries of the two rivers 1 and 2 individually receiving treated penicillin G and oxytetracycline production wastewater, as well as the anaerobic and the aerobic effluent clone libraries of penicillin G production wastewater.
that both bacterial community compositions in the upstream rivers were distinctly different from the remaining of the antibiotic containing aquatic environments (Fig. 4). PCoA analysis also revealed the similar results (data not shown). Comparably higher richness and evenness indexes of the clone libraries derived from the aerobic effluent of the WWTP and the downstream river receiving penicillin G production wastewater were observed than those from the anaerobic effluent and the reference upstream river (Table 2), indicating that the bacterial communities in the aerobic effluent and the penicillin G polluted downstream river were more diverse. Meanwhile, the diversity of the clone library for the oxytetracycline containing downstream river was similar with that of the upstream river. It thus seems that the antibiotic stresses had not obviously reduced the diversity of bacterial community in aquatic environments. This result was accordant with that obtained before, in which phospholipid fatty acid numbers even increased in ciprofloxacin-treated microcosms (Co´rdova-Kreylos and Scow, 2007), while different from several recent studies on human gut microbiome which
Discussions
In this study, Deltaproteobacteria, Epsilonproteobacteria, Clostridia and Bacilli have been found abundant in all antibiotic containing environments. It has been reported that Deltaproteobacteria and Firmicutes were commonly abundant in the anaerobic wastewater treatment systems, with the clone percentages varied among different researches (Godon et al., 1997). Penicillin G production wastewater of this study was treated with a combination of anaerobic and aerobic processes. It is thus possible that the presence of Deltaproteobacteria, Clostridia and Bacilli in the aerobic effluent as well as the corresponding downstream river was related with the release from the anaerobic reactor. However, Epsilonproteobacteria, which is comparably uncommon in the anaerobic treatment systems reported previously (Godon et al., 1997), was the dominant Proteobacteria and accounted for 14.3% of the total clone number of the bacterial library for the anaerobic effluent. Meanwhile, many previous investigations have demonstrated that the phylum Proteobacteria was generally dominant in the aerobic reactors of WWTPs with Betaproteobacteria being the most frequently observed, and the phyla Bacteroidetes and Actinobacteria were frequently retrieved (Wagner and Loy, 2002). Then Alphaproteobacteria, Betaproteobacteria, Actinobacteria, Acidobacterium and Bacteroidetes were reported to account for the major proportion of the bacterial community in freshwater (Hugenholtz et al., 1998). Furthermore, the clones grouped in Deltaproteobacteria, Clostridia and Bacilli have totally accounted for 37.7% of the clone number of the library for the oxytetracycline containing downstream river, while no anaerobic treatment was adopted for the oxytetracycline wastewater of this study. Thus, the
Table 2 e Coverage and diversity indexes of bacterial 16S rRNA gene clone libraries for the upstream and downstream of the two rivers individually receiving treated penicillin G and oxytetracycline production wastewater, as well as the anaerobic and the aerobic effluent of penicillin G production wastewater. Sample a
Upstream river1 Upstream river2b Downstream river1a Downstream river2b Anaerobic effluent Aerobic effluent
No. of clones
No. of OTUs
SChao1
SACE
Shannon index
Evenness index
% Coverage
75 72 82 77 77 81
23 39 45 39 26 43
38 81.8 85.6 76.5 32.88 116.5
33.87 75.3 118.82 87.3 38.88 118.84
2.82 3.36 3.52 3.32 2.83 3.5
0.899 0.917 0.925 0.906 0.869 0.931
86.7 65.3 64.6 67.5 85.7 63
a The river1 received treated penicillin G production wastewater. b The river2 received treated oxytetracycline production wastewater.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 6 3 e6 0 7 3
unexpectedly high abundance of the bacterial groups including Deltaproteobacteria, Epsilonproteobacteria, Clostridia and Bacilli appeared in the aerobic effluent and both downstream rivers (the sampling sites generally 20 km away from the discharge points of wastewater) could not be only attributed to the release from the anaerobic treatment process. Furthermore, the abundance of these bacterial groups might be resulted from both the antibiotics and the co-existing pollutants in wastewater. However, the impacts of antibiotics should be much larger due to their strong bacteriostatic effects. By using molecular methods, several bacterial genera such as Clostridium spp., Eubacterium spp., Streptococcus spp. and Lactobacillus spp. belonging to the classes Clostridia and Bacilli in Firmicutes have been found sometimes still dominant in the human and poultry gut after feeding with antibiotics, together with Bacteroides spp. in Bacteroidetes (Knarreborg et al., 2002; Young and Schmidt, 2004; Jernberg et al., 2005). Antonopoulos et al. and Rea et al. have found that Proteobacteria is particularly enriched in human gut microbiome with the treatment of antibiotics (Antonopoulos et al., 2009; Rea et al., 2011). Lawrence et al. have observed a significant reduction in the abundance of Betaproteobacteria and Gammaproteobacteria in river biofilm communities exposed to a broadspectrum antimicrobial chlorhexidine using fluorescent in situ hybridization (Lawrence et al., 2008), whereas no more information has been provided about the other bacteria groups in their research. Then in the study of Co´rdova-Kreylos and Scow (2007), sulfate-reducing bacteria belonged to Deltaproteobacteria including Desulfovibrio, Desulfobacter and Desulfobulbus were obviously favored by ciprofloxacin in the sedimental microbial communities. These previous reports have partially confirmed the specific association between the bacterial groups Deltaproteobacteria, Epsilonproteobacteria, Clostridia and Bacilli and antibiotic containing environments. It still need be noted that the penicillin G is grouped into blactam antibiotics and mainly active against Gram-positive bacteria by acting on bacterial cell wall, while oxytetracycline is a broad-spectrum antibiotic belonging to tetracyclines and inhibits bacterial protein synthesis by binding to the 30S ribosomal subunit (Fluit et al., 2001). Several human infection cases caused by sulfate-reducing bacteria belonging to Deltaproteobacteria have been reported, and all of these sulfate-reducing strains had shown co-resistance to multiple antibiotics belonged to different classes (Pitcher et al., 1994; McDougall et al., 1997). Sulfate-reducing bacteria were also not affected by the administration of several different antibiotics in rats (Ohge et al., 2003). Several Arcobacter spp. belonging to Epsilonproteobacteria related to human infections had often been described to confer multi-drug resistance (Fera et al., 2003). In the classes Clostridia and Bacilli, tetracycline resistance gene tet(W) was first identified in one M. multacida isolate (Scott et al., 2000). Some isolates of T. flocculiformis in these classes had shown multiple antibiotic resistance abilities in our previous research (Li et al., 2009), and Streptococcus bovis has displayed multiple antibiotic resistance to several antibiotics (Teng et al., 2001), together with Clostridium spp. strains (Rood et al., 1978). The other bacterial genera in Clostridia and Bacilli including Aminobacterium, Desulfosporosinus, Anaerovorax, Acetobacterium, Thermoanaerobacterium, Ruminococcus, Dialister and Mogibacterium of this study are phylogenetically closely related to
6071
intestinal bacterial groups including Clostridium, Eubacterium, Streptococcus and Lactobacillus, which were sometimes still dominant in the gut after feeding the antibiotics (Knarreborg et al., 2002; Young and Schmidt, 2004; Jernberg et al., 2005), suggesting that these bacterial genera of this study possibly shared similar antibiotic resistance mechanisms with those intestinal bacterial groups. Several antibiotic resistance mechanisms have been described for bacteria, including antibiotic resistance genes which encode antibiotic modifying or inactivating enzymes and usually target one antibiotic class specifically, mutations of antibiotic target sites in bacterial cells, and efflux pump systems locating in bacterial cell membranes and mainly accounting for multi-drug resistance. The occurrence of the specific bacterial groups including Deltaproteobacteria, Epsilonproteobacteria, Clostridia and Bacilli in different antibiotics containing environments suggested that some kind of intrinsic resistance mechanisms might account for the widespread of these groups.
5.
Conclusions
Several bacterial groups including Deltaproteobacteria, Epsilonproteobacteria, Clostridia and Bacilli have been suggested to be specifically associated with antibiotic containing aquatic environments, many of which have not been recovered in previous antibiotic resistance researches using culture-based analysis.
Acknowledgments This work was financially supported by the Ministry of Science and Technology of China (2006DFA91870) and the National Natural Science Foundation of China (20877085 and 50525824).
references
Altschul, F.S., Madden, T.L., Scha¨ffer, A.A., Zhang, J., Zhang, Z., Miller, W., Lipman, D.J., 1997. Gapped BLAST and PSI-BLAST: a new generation of protein database search programs. Nucleic Acids Research 25, 3389e3402. Antonopoulos, D.A., Huse, S.M., Morrison, H.G., Schmidt, T.M., Sogin, M.L., Young, V.B., 2009. Reproducible community dynamics of the gastrointestinal microbiota following antibiotic perturbation. Infection and Immunity 77, 2367e2375. Clinical and Laboratory Standards Institute, 2003. Methods for Dilution Antimicrobial Susceptibility Tests for Bacteria that Grow Aerobically. Approved standard M7eA6. Clinical and Laboratory Standards Institute, Wayne, PA, USA. Cole, J.R., Chai, B., Farris, R.J., Wang, Q., Kulam-SyedMohideen, A.S., McGarrell, D.M., Bandela, A.M., 2007. The ribosomal database project (RDP-II): introducing my RDP space and quality controlled public data. Nucleic Acids Research 35, D169eD172. Colwell, R.K., 2005. EstimateS: Statistical Estimation of Species Richness and Shared Species from Samples. Version 8.0.
6072
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 6 3 e6 0 7 3
User’s Guide and application published at: http://purl.oclc.org/ estimates. Co´rdova-Kreylos, A.L., Scow, K.M., 2007. Effects of ciprofloxacin on salt marsh sediment microbial communities. The ISME Journal 1, 585e595. Crump, B.C., Armbrust, E.V., Baross, J.A., 1999. Phylogenetic analysis of particle-attached and free-living bacterial communities in the Columbia River, its estuary, and the adjacent coastal ocean. Applied and Environmental Microbiology 65, 3192e3204. D’Costa, V.M., McGram, K.M., Hughes, D.W., Wright, G.D., 2006. Sampling the antibiotic resistome. Science 311, 374e377. Fera, M.T., Maugeri, T.L., Giannone, M., Gugliandolo, C., La Camera, E., Blandino, G., Carbone, M., 2003. In vitro susceptibility of Arcobacter butzleri and Arcobacter cryaerophilus to different antimicrobial agents. International Journal of Antimicrobial Agents 21, 488e491. Fluit, A.D.C., Visser, M.R., Schmitz, F.-J., 2001. Molecular detection of antimicrobial resistance. Clinical Microbiology Reviews 14, 836e871. Godon, J.-J., Zumstein, E., Dabert, P., Habouzit, F., Moletta, R., 1997. Molecular microbial diversity of an anaerobic digestor as determined by small-subunit rDNA sequence analysis. Applied and Environmental Microbiology 63, 2802e2813. Hall, T.A., 1999. BioEdit: a user-friendly biological sequence alignment editor and analysis program for Windows 95/98/NT. Nucleic Acids Symposium Series 41, 95e98. Hamelin, K., Bruant, G., El-Shaarawi, A., Hill, S., Edge, T.A., Fairbrother, J., et al., 2007. Occurrence of virulence and antimicrobial resistance genes in Escherichia coli isolates from different aquatic ecosystems within the St. Clair river and Detroit river areas. Applied and Environmental Microbiology 73, 477e484. Huddleston, J.R., Zak, J.C., Jeter, R.M., 2006. Antimicrobial susceptibilities of Aeromonas spp. isolated from environmental sources. Applied and Environmental Microbiology 72, 7036e7042. Hugenholtz, P., Goebel, B.M., Pace, N.R., 1998. Impact of cultureindependent studies on the emerging phylogenetic view of bacterial diversity. The Journal of Bacteriology 180, 4765e4774. ˚ , Edlund, C., Jansson, J.K., 2005. Monitoring Jernberg, C., Sullivan, A of antibiotic-induced alterations in the human intestinal microflora and detection of probiotic strains by use of terminal restriction fragment length polymorphism. Applied and Environmental Microbiology 71, 501e506. Jindal, A., Kocherginskaya, S., Mehboob, A., Robert, M., Mackie, R. I., Raskin, L., Zilles, J.L., 2006. Antimicrobial use and resistance in swine waste treatment systems. Applied and Environmental Microbiology 72, 7813e7820. Knarreborg, A., Simon, M.A., Engberg, R.M., Jensen, B.B., Tannock, G.W., 2002. Effects of dietary fat source and subtherapeutic levels of antibiotic on the bacterial community in the ileum of broiler chickens at various ages. Applied and Environmental Microbiology 68, 5918e5924. Kolpin, D.W., Furlong, E.T., Meyer, M.T., Thurman, E.M., Zaugg, S. D., Barber, L.B., et al., 2002. Pharmaceuticals, hormones, and other organic wastewater contaminants in US streams, 1999e2000: a national reconnaissance. Environmental Science and Technology 36, 1202e1211. Kumar, S., Tamura, K., Nei, M., 2004. MEGA3: Integrated software for molecular evolutionary genetics analysis and sequence alignment. Briefings in Bioinformatics 5, 150e163. Lane, D.J., 1991. 16S/23S rRNA sequencing. In: Stackebrandt, E., Goodfellow, M. (Eds.), Nucleic Acid Techniques in Bacterial Systematics. John Wiley, Chichester, United Kingdom, pp. 115e175.
Lawrence, J.R., Zhu, B., Swerhone, G.D.W., Topp, E., Roy, J., Wassenaar, L.I., Rema, T., Korber, D.R., 2008. Community-level assessment of the effects of the broad-spectrum antimicrobial chlorhexidine on the outcome of river microbial biofilm development. Applied and Environmental Microbiology 74, 3541e3550. Li, D., Yang, M., Hu, J., Zhang, J., Liu, R., Gu, X., Zhang, Y., Wang, Z. , 2009. Antibiotic resistance profile in environmental bacteria isolated from penicillin production wastewater treatment plant and the receiving river. Environmental Microbiology 11, 1506e1517. Li, D., Yang, M., Hu, J., Ren, L., Zhang, Y., Li, K., 2008a. Determination and fate of oxytetracycline and related compounds in oxytetracycline production wastewater and the receiving river. Environmental Toxicology and Chemistry 27, 80e86. Li, D., Yang, M., Hu, J., Zhang, Y., Chang, H., Jin, F., 2008b. Determination of penicillin G and its degradation products in a penicillin production wastewater treatment plant and the receiving river. Water Research 42, 307e317. Lozupone, C., Knight, R., 2005. UniFrac: a new phylogenetic method for comparing microbial communities. Applied and Environmental Microbiology 71, 8228e8235. Martı´nez, J.L., 2008. Antibiotics and antibiotic resistance genes in natural environments. Science 321, 365e367. Martı´nez, J.L., Baquero, F., Andersson, D.I., 2007. Predicting antibiotic resistance. Nature Reviews Microbiology 5, 958e965. McDougall, R., Robson, J., Paterson, D., Tee, W., 1997. Bacteremia caused by a recently described novel Desulfovibrio species. Journal of Clinical Microbiology 35, 1805e1808. Messi, P., Guerrieri, E., Bondi, M., 2005. Antibiotic resistance and antibacterial activity in heterotrophic bacteria of mineral water origin. Science of the Total Environment 346, 213e219. Miranda, C.D., Zemelman, R., 2002. Antimicrobial multiresistance in bacteria isolated from freshwater Chilean salmon farms. Science of the Total Environment 293, 207e218. Ohge, H., Furne, J.K., Springfield, J., Sueda, T., Madoff, R.D., Levitt, M.D., 2003. The effect of antibiotics and bismuth on fecal hydrogen sulfide and sulfatereducing bacteria in the rat. FEMS Microbiology Letters 228, 137e142. Pitcher, M.C.L., Gibson, G.R., Neale, G., Cummings, J.H., 1994. Gentamicin kills multiple drug-resistant sulfate-reducing bacteria in patients with ulcerative colitis. Gastroenterology 106, A753. Rea, M.C., Dobson, A., O’Sullivan, O., Crispie, F., Fouhy, F., Cotter, P.D., Shanahan, F., Kiely, B., Hill, C., Ross, R.P., 2011. Effect of broad- and narrow-spectrum antimicrobials on Clostridium difficile and microbial diversity in a model of the distal colon. Proceedings of the National Academy of Sciences of the United States of America 108 (Suppl. 1), 4639e4644. Riesenfeld, C.S., Goodman, R.M., Handelsman, J., 2004. Uncultured soil bacteria are a reservoir of new antibiotic resistance genes. Environmental Microbiology 6, 981e989. Rood, J.I., Maher, E.A., Somers, E.B., Campos, E., Duncan, C.L., 1978. Isolation and characterization of multiply antibioticresistant Clostridium perfringens strains from porcine feces. Antimicrobial Agents and Chemotherapy 13, 871e880. Scheff, G., Salcher, O., Lingens, F., 1984. Trichococcus flocculiformis gen. nov. sp. nov. A new gram-positive filamentous bacterium isolated from bulking sludge. Applied Microbiology and Biotechnology 19, 114e119. Schloss, P.D., Handelsman, J., 2005. Introducing DOTUR, a computer program for defining operational taxonomic units and estimating species richness. Applied and Environmental Microbiology 71, 1501e1506. Scott, K.P., Melville, C.M., Barbosa, T.M., Flint, H.J., 2000. Occurrence of the new tetracycline resistance gene tet(W) in
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 6 3 e6 0 7 3
bacteria from the human gut. Antimicrobial Agents and Chemotherapy 44, 775e777. Sommer, M.O.A., Dantas, G., Church, G.M., 2009. Functional characterization of the antibiotic resistance reservoir in the human microflora. Science 325, 1128e1131. Teng, L.-J., Hsueh, P.-R., Ho, S.-W., Luh, K.-T., 2001. High prevalence of inducible erythromycin resistance among Streptococcus bovis isolates in Taiwan. Antimicrobial Agents and Chemotherapy 45, 3362e3365. Thomas, C.M., Nielsen, K.M., 2005. Mechanisms of, and barriers to, horizontal gene transfer between bacteria. Nature Reviews Microbiology 3, 711e721. Thompson, J.D., Gibson, T.J., Plewniak, F., Jeanmougin, F., Higgins, D.G., 1997. The ClustalX windows interface: flexible
6073
strategies for multiple sequence alignment aided by quality analysis tools. Nucleic Acids Research 25, 4876e4882. Wagner, M., Loy, A., 2002. Bacterial community composition and function in sewage treatment systems. Current Opinion in Biotechnology 13, 218e227. Wittwer, M., Keller, J., Wassenaar, T.M., Stephan, R., Howald, D., Regula, G., Bissig-Choisat, B., 2005. Genetic diversity and antibiotic resistance patterns in a Campylobacter population isolated from poultry farms in Switzerland. Applied and Environmental Microbiology 71, 2840e2847. Young, V.B., Schmidt, T.M., 2004. Antibiotic-associated diarrhea accompanied by large-scale alterations in the composition of the fecal microbiota. Journal of Clinical Microbiology 42, 1203e1206.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 7 4 e6 0 8 6
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Non-linear effects on solute transfer between flowing water and a sediment bed Makoto Higashino a, Heinz G. Stefan b,* a b
Department of Civil Engineering, Oita National College of Technology, 1666 Maki, Oita 870-0152, Japan Department of Civil Engineering, St. Anthony Falls Laboratory, Univ. of Minnesota, Minneapolis, MN 55414, USA
article info
abstract
Article history:
A previously developed model of periodic pore water flow in space and time, and associ-
Received 2 April 2011
ated solute transport in a stream bed of fine sand is extended to coarse sand and fine
Received in revised form
gravel. The pore water flow immediately below the sediment/water interface becomes
6 August 2011
intermittently a non-Darcy flow. The periodic pressure and velocity fluctuations consid-
Accepted 3 September 2011
ered are induced by near-bed coherent turbulent motions in the stream flow; they pene-
Available online 16 September 2011
trate from the sediment/water interface into the sediment pore system and are described by a wave number (c) and a period (T ) that are given as functions of the shear velocity (U*)
Keywords:
between the flowing water and the sediment bed. The stream bed has a flat surface without
Mass transport
bed forms. The flow field in the sediment pore system is described by the continuity
Porewater
equation and a resistance law that includes both viscous (Darcy) and non-linear (inertial)
Sediment
effects. Simulation results show that non-linear (inertial) effects near the sediment/water
Sediment-water exchange
interface increase flow resistance and reduce mean flow velocities. Compared to pure
Solute
Darcy flow, non-linear (inertial) effects reduce solute exchange rates between overlying
Stream
water and the sediment bed but only by a moderate amount (less than 50%). Turbulent
Streambed
coherent flow structures in the stream flow enhance solute transfer in the pore system of
Turbulent flow
a stream bed compared to pure molecular diffusion, but by much less than standing
Water quality
surface waves or bed forms. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
The exchange of dissolved materials (solutes) such as oxygen, sulfate or phosphate, between flowing water and the underlying sediment bed in a stream, river, or estuary is an important environmental process. It can affect the quality of the flowing water or the growth of organisms, e.g. microbes, in the sediment bed. It is well known from experimental studies (e.g. Steinberger and Hondzo, 1999; Mackenthun and Stefan, 1997) that sedimentary oxygen demand (SOD) depends strongly on the flow velocity of the water above the sediment/water interface in addition to biogeochemical processes in the
sediment. These processes, and their spatial distribution in the sediment can be complex. We will focus on the physics of (conservative) solute transport and not address microbiology or geochemistry in this paper. The exchange of solutes between overlying water and sediment has been studied in laboratory and field experiments (see review by O’Connor and Harvey, 2008), and models have been developed to quantify the solute flux across the sediment/water interface (e.g. Elliott and Brooks, 1997; Packman et al., 2000; Qian et al., 2009; O’Connor and Harvey, 2008). In sediments of low permeability, i.e. fine sediments, molecular diffusion is the major mechanism for mass (solute)
* Corresponding author. Fax: þ1 612 627 4609. E-mail addresses: [email protected] (M. Higashino), [email protected] (H.G. Stefan). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.09.004
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 7 4 e6 0 8 6
transport in the pore system because there is no interstitial water motion. In coarser sediment beds, an advective transport component becomes important or dominant. Depending on the sediment particle size, advection by interstitial flow through the sediment pore space can be a major mechanism for mass transport of solutes (Nagaoka and Ohgaki, 1990; Shimizu et al., 1990; Mendoza and Zhou, 1992; Zhou and Mendoza, 1993). Porewater (interstitial) flow in a stream bed can be driven by gravity due to the slope of a stream, or by pressure differentials along the sediment/water interface; these can be induced by bed forms (Elliott and Brooks, 1997; Packman et al., 2000, 2004; Packman and Brooks, 2001; Marion and Zaramella, 2005; Cardenas and Wilson, 2004, 2006, 2007), by surface waves (Huettel and Webster, 2001; Qian et al., 2009), or by turbulence in the stream flow (Basu and Khalili, 1999; Zhou and Mendoza, 1993). While the interstitial flow induced by bed forms and surface waves has been studied well, both theoretically and experimentally, fewer studies have addressed the propagation of turbulence from the flowing water in a stream or river into the sediment bed. A simulation that links pressure fluctuations at the sediment/water interface caused by near-bed coherent turbulent motions to interstitial flow and solute transport in the sediment (Fig. 1) was made by Higashino et al. (2009). In that study the assumption was made that the flow in the sediment bed obeys Darcy’s law. This assumption is justified in many cases, especially if the sediment particles are fine. In coarser sediment beds, turbulent eddies created by near-bed coherent motions may penetrate into the porous sediment. In that case intermittent turbulence exists in the interstitial flow below
6075
the sediment/water interface. In coarser sediment beds an underflow parallel to the surface flow can also develop, and the interaction between surface and sub-surface flow will cause a slip velocity at the sediment/water interface (Fig. 1). A large slip velocity can be an indicator of significant inertial (non-linear) resistance in the subsurface flow; in that case Darcy’s law of resistance in the sediment bed is no longer valid. Equations by Brinkman (1947) and Forchheimer (1930) have been used to describe the shear stress associated with the slip velocity, and the non-linear (inertial) resistance, respectively. The coupling of surface and sub-surface flow has also been studied by derivations starting with the full NaviereStokes equations. The significance of inertial effects for mass transport of solutes in the sediment pore system, especially near the sediment/water interface, has to the authors’ knowledge not been assessed. This is a precursor to highly turbulent flows in very coarse sediments which will not be analyzed in this paper. There are several mechanisms by which dispersion can occur in pore water flow. “Variations in local velocity, both in magnitude and direction, along the tortuous flow paths and between adjacent flow paths as a result of the velocity distribution within each pore, cause any initial tracer mass within the flow domain to spread and occupy an ever-increasing volume of the porous medium” (Bear, 1972). Advective transport and molecular diffusion are the basic physical processes that lead to dispersion. Hydrodynamic dispersion includes both processes. In Darcy flow, an oscillatory pressure gradient causes liquid elements to move back and forth, and dispersion only occurs if some of the substance is left behind along the path of
Fig. 1 e Near-bed turbulent flow over a coarse sediment bed and associated pore water flow (schematic).
6076
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 7 4 e6 0 8 6
the excursion. For shear dispersion to be effective, the solute must diffuse over a significant lateral distance within the pore over one period of oscillation. Periodic Darcy flow in a sediment bed driven by progressive surface waves, has been found to enhance vertical dispersion (Qian et al., 2009), but not nearly as much as bed forms or standing waves (Elliott and Brooks, 1997; Qian et al., 2008). A recent study on wavedriven pore water flow and associate solute transport through rippled elastic sediment suggests that the pressure perturbations along the sediment/water interface propagate within a few milliseconds to meter depths within the sediment leading eventually to a strongly transient pore water velocity field, which leads to enhanced dispersion of solutes and larger time-averaged solute fluxes (Cardenas and Jiang, 2011). Bursts in fluid flushing occur due to high-frequency pressure fluctuations, leading to larger long-term average fluid fluxes compared to a steady flow field driven by a timeaveraged pressure profile. Flows within pores are not likely to be turbulent except for large pores, and then only during a portion of the oscillatory period, but if inertial effects are significant then these will result in flow elements not returning along exactly the same paths when the flow reverses. In this case, lateral dispersion may become stronger, but longitudinal dispersion may be reduced or enhanced when inertial effects are significant. The objective of this study is to extend the previous analysis (Higashino et al., 2009) to coarser beds of streams where the subsurface flow becomes locally and intermittently turbulent due to turbulence penetration from above. The added flow resistance will be accounted for, but slip velocities will be ignored, indicating that a fully turbulent flow in the sediment bed has not developed. Specifically, we want to build a pore water flow model that includes non-linear (inertial) effects in addition to viscous resistance depending on sediment grain size. The penetration of low frequency turbulent eddies from coherent turbulent structures in the flow over the sediment/water interface, and the non-linear inertial effects will be simulated by an ‘effective hydraulic conductivity’ in the Darcy equation. Low frequency turbulence due to coherent turbulent structures in the overlying water has the most significant effect on pore water flow and will be characterized, as previously, by shear velocity between the flowing water and the sediments bed. Higher frequencies in the turbulence spectrum of stream flow will not be considered. We will determine under which conditions non-linear effects are significant, and how much they alter the pore water velocities and solute transport.
2.
Model development
2.1.
Flow model concept
We previously developed a model (pressure pulse model) for the pore water flow in a sediment bed induced by periodic pressure fluctuations due to coherent turbulent motions at the sediment/water interface (Higashino et al., 2009). In that model the sediment is a fine sand with hydraulic conductivity K ¼ 0.01~1 cm/s and particle diameter ds ¼ 0.006~0.060 cm,
and Darcy’s equation is used to describe the entire flow field. There is no underflow in the sediment bed, and no slip velocity at the sediment/water interface. When the sediment bed is coarser, inertial resistance in the pore water flow can become significant, especially near the sediment/water interface. In still coarser sediment beds the slip velocity is significant, and the pressure pulse model is no longer sufficient to describe the interstitial flow correctly. The Brinkman term (Brinkman, 1947) then needs to be taken into account. In this paper a pressure pulse model will be developed for sediments with hydraulic conductivity K ¼ 1~100 cm/s. Assuming a sediment porosity ɸ ¼ 0.4, the sediment grain diameter (ds) can be calculated as ds ¼ 0.10 cm for K ¼ 1 cm/s, ds ¼ 0.320 cm for K ¼ 10 cm/s, and ds ¼ 1.012 cm for K ¼ 100 cm/s. The shear velocity will be kept to U* < 1.6 cm/s; 1.6 cm/s is smaller than the critical shear velocities from Shields’ curve (Vanoni, 1975). Uc ¼ 2.36 cm/s for ds ¼ 0.101 cm (K ¼ 1 cm/s), and Uc ¼ 9.05 cm/s for ds ¼ 1 cm (K ¼ 100 cm/s) indicating that the sediment bed will be stationary and flat. The slip velocity can be a good parameter to characterize the interaction of surface and subsurface flows, and to show where the applicability of the pressure pulse model ends. Assuming that the slip velocity is small enough to be neglected when U* < 2.0 cm/s we will present simulation results for shear velocities U* 1.6 cm/s, and hydraulic conductivities K 100 cm/s. We will consider these two values as the upper bounds for the applicability of the extended pressure pulse model to be presented. Non-linear (inertial) effects will be considered in the analysis, but slip velocities will be assumed negligible (Table 1). The extended pressure pulse model to be presented will include (1) a model of the pressure fluctuations induced by near-bed coherent motions at the sediment/water interface, (2) a model of non-Darcy pore water flow, and (3) a derivation of effective hydraulic conductivity for non-Darcy flow. The flow model results will be applied to a mass (solute) transport model which will be compared to experimental results summarized by O’Connor and Harvey (2008).
2.2. bed
Model of near-bed turbulence above the sediment
In boundary layer flow, turbulent energy contributions occur at a continuum of time and space scales. In the proposed
Table 1 e Flow and sediment parameter range investigated for pressure pulse model with non-linear (inertial) resistance effects on the pore water flow. Parameter (units)
Units
Low value
High value
Hydraulic conductivity (Ko) Sediment particle size (ds) Porosity (ɸ) Sediment bed thickness (d) Shear velocity (U*) Slip velocity (Us =U )
cm s1 cm e cm cm s1
1 0.1 0.3 10.0 0.1 0
100 1.0 0.6 20.0 1.6 1.0
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 7 4 e6 0 8 6
model the spectra of velocity and pressure fluctuations in the free surface flow of a stream are simplified to the turbulent fluctuations produced by the coherent turbulent structures in the stream (Chien and Wan, 1998). They are associated with the largest turbulent eddies, and represent low frequencies and high amplitudes. This characterization of turbulence by coherent structures is not a serious limitation, because previous analyses of velocity pulses and/or pressure pulses at the sediment/water interface (Higashino and Stefan, 2008; Higashino et al., 2009), as well as other studies, have shown that the lowest frequencies, have by far the greatest potential to propagate into the pore space of the sediment bed of a stream. By comparison, high-frequency pulses are attenuated much more strongly with depth into the sediment bed. The spatial and the temporal scales of near-bed coherent motions and associated turbulent pressure pulses in the fully developed turbulent boundary layer are given by Eqs. (1) and (2) (Dade et al., 2001). n0 L ¼ 1000 U T ¼ 100
n0 U2
(1)
c¼
2p L
(4)
s¼
2p T
(5)
The amplitude ( p0) in Eq.(3) has been related to the turbulent kinetic energy balance near the sediment/water interface, and given as a function of the shear velocity (U*) (Higashino et al., 2009). p0 pffiffiffi ¼ 2$ð4:0w5:0ÞU2 r
2.3.
(6)
Model of pore water flow in the sediment bed
The pore water flow is governed by the continuity equation (Higashino et al., 2009). vh v vh v vh ¼ a þ a vt vx vx vz vz
(7a)
simplified to (2)
where L is the space scale or wavelength in stream-wise direction, and T is the time scale or wave period of near-bed coherent motions, n0 is the kinematic viscosity for water pffiffiffiffiffiffiffiffiffi (¼ 0.01 cm2/s at 20 C for pure water) and U ¼ s0 =r is the bed shear velocity (s0 is the bed shear stress in the stream, and r is the density of water). Eqs. (1) and (2) were derived for a solid, flat and smooth surface. Although the surface of a stream bed can be hydraulically rough, we assume that Eqs (1) and (2) are applicable. The dependence of space and time scales of near bed coherent motions on bed roughness conditions has hardly been investigated. The period of the turbulent fluctuations given by Eq. (2) is on the order of T z 1 s (with kinematic viscosity n0 ¼ 0.01 cm2/s and bed shear velocity U* z 1.0 cm/s), the penetration velocity at the sediment/water interface is on the order of 1 cm/s or less, i.e. relatively small and reversible, compared to the velocity in the overlying stream flow. The mass of water (in the eddies of the overlying flow) from which the pressure pulses at the sediment/water interface are derived, is much larger than the mass of water that flows into and out of the pore system of the sediment bed. Because of the large difference in mass, the feedback from the pore water flow onto the overlying flow is expected to be small, and will be neglected. The turbulent pore water pressure fluctuations, i.e. the pressure wave, at the sediment/water interface z ¼ 0 is given by Eq. (3). pðt; x; 0Þ ¼ p0 cos ðcx stÞ
6077
(3)
where t is time, x and z are the coordinates in longitudinal and vertical (positive upward) directions, respectively, and p0 is the amplitude of the pore water pressure wave at the sediment/water interface. The parameters c and s are the wave number and angular frequency, respectively. Parameters (c) and (s) are related to (L) and (T ) of near-bed coherent motions, from Eqs. (1) and (2), respectively.
vh v2 h v vh ¼a 2þ a vt vx vz vz
(7b)
because (a) in Eq.(7a) can be variable in the z e direction, but not in the x e direction, as will be explained later. In Eqs. (7a) and (7b) the piezometric pore pressure head h(x,z,t) ¼ p/rg þ z, where p(x, z, t) is the pore pressure, g is gravitational acceleration, and z is elevation. The coefficient (a) in Eqs. (7a) or (7b) is the effective hydraulic diffusivity and given by Eq. (8). a¼
ke rgmv
(8)
In Eq. (8) (ke) is the effective hydraulic conductivity; mv is a parameter that depends on sediment particle size and porosity; the value of mv has been given based on measured values for densely and loosely packed sand (Higashino et al., 2009). The velocity components of pore water flow in a vertical x-z- plane are expressed following Darcy’s law as qx ¼ Ke
vh vx
(9a)
qz ¼ Ke
vh vz
(9b)
in which qx and qz are velocity components in longitudinal and vertical direction, respectively, and Ke is the effective hydraulic conductivity which in our study includes locally and temporally variable non-linear (inertial effects) induced by, pore water velocities above the Darcy limit. In Darcy flow through a porous medium, non-linear (inertial) effects are absent, and Ke ¼ K, where (K ) is the hydraulic conductivity or coefficient of permeability of the porous medium; it can be anisotropic and thus a tensor quantity due to sediment stratigraphy. In this study the sediment bed is assumed to have the same hydraulic conductivity (K ) in all directions. Hydraulic conductivity (K ) and permeability (k) are related to each other by Eq. (10).
6078
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 7 4 e6 0 8 6
K ¼ kg=v0
(10)
2.4. Effective hydraulic conductivity for non-Darcy flow in the sediment bed The pressure driven flow through a porous medium, can either be laminar or turbulent (Bear, 1972; DePinto et al., 1994). When the sediment consists mainly of fine particles, the flow in the pore spaces of the sediment bed is dominated by viscous resistance. In that case Darcy’s law can be applied to the entire flow field. In Darcy flow the permeability (k) can be expressed as a function of the sediment porosity (ɸ) and the sediment grain diameter (ds) (Bear, 1972; Boudreau 1997) as k¼
5:6 103 f3 d2s ð1 fÞ2
" 1 ¼ Ke
Ke ¼1 K
,"
pffiffiffi # K 2qj 1þ 2 f g3m
(12b)
This flow resistance law including inertial effects was derived theoretically in the reference cited, and was matched to extensive experimental data. The first term on the right hand side of Eq. (12a) represents viscous flow resistance (Darcy flow) and the second term represents non-linear (inertial) effects on total flow resistance. The hydraulic radius (m) used in Eqns. (12a) and (12b), is given by Eq. (13) (Barr, 2001) m¼
f s
(13)
where s is the surface area of a sediment particle relative to its volume (units of L2/L3); s is related to sediment porosity (ɸ) and particle size represented by the radius (r) of a sphere, i.e. r ¼ ds/ 2, as s¼
3ð1 fÞ r
(14)
The effective hydraulic conductivity (Ke) defined by Eq. (12a) is likely to be smallest at the sediment/water interface (z ¼ 0) where pore velocities and inertial (non-linear) effects are likely to be the strongest; (Ke) approaches the hydraulic conductivity (K ) with vertical distance (z) from the sediment/ water interface as pore water velocities diminish, and nonlinear (inertial effects) disappear, as shown schematically in Fig. 1. When non-linear effects are absent, the effective hydraulic conductivity given by Eq. (12a) is the hydraulic conductivity (K ). Eq. (12a) was validated by Barr (2001) against data for a wide-range of flow conditions. We consider that Eq. (12a) is valid for the laminar-turbulent flow transition, and can be used to describe interstitial flow with intermittent turbulence due to penetrating turbulent eddies. The pore water velocities change periodically, but with low enough frequency to justify the quasi-steady state assumption. Non-linear (inertial) effects depend on the magnitude of the flow velocity qj in Eqs. (12a) and (12b). At and near the sediment/water interface (z ¼ 0) the vertical velocity component qz is dominant, over the horizontal velocity component qx (Higashino et al., 2009), because there is no slip velocity at the sediment/water interface. We will therefore make the non-linear effects dependent on the vertical velocity component only. This is an approximation which allows the simplification of Eq. (7a) to Eq. (7b).
2.5. bed
Fig. 2 e Relationship between sediment particle diameter (ds) and hydraulic conductivity (K ) for sediment porosity (ɸ) [ 0.4, 0.5 and 0.6.
(12a)
which can be rewritten in normalized form as
(11)
Fig. 2 illustrates the relationship between the hydraulic conductivity (K ) and the sediment grain diameter (ds) for different sediment porosities as given by Eq. (11). The hydraulic conductivity (K ) increases when the sediment particle diameter (ds) is larger. Roughly, Ky0:01 cm=s if the sediment is silt, 0.01 cm/s < K < 1 cm/s for fine sand, 1 cm/ s < K < 20 cm/s for coarse sand, and K > 20 cm/s for gravel. Although Darcy’s law is an empirical relationship, it has been shown to describe the viscous resistance in completely laminar flow without non-linear effects. When pore water flow velocities are high, added flow resistance is brought on by kinetic energy dissipation in addition to viscous flow resistance (shear) in laminar flow. Inertial effects make the law of flow resistance in a porous medium deviate from Darcy’s law. We use the term “nonlinear” effect to describe it. In the sediment bed of a stream, non-linear effects are most likely to occur in the pore water flow near the sediment/water interface. A review of turbulent flow in porous media was given by Barr (2001). He incorporated non-linear (inertial) effects into Darcy’s law by introducing the effective hydraulic conductivity (Ke) given in Eq. (12a).
pffiffiffi # 2qj 1 þ 2 K f g3m
Model of mass (solute) transport in the sediment
To analyze or model solute transport in a stream bed, the pore water flow model described above can be used. To evaluate the rate of transport of a conservative (non-reactive) solute in the sediment bed we consider a 1-D (in vertical direction) dispersion Eq. (15)
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 7 4 e6 0 8 6
vC v vC ¼ De vt vz vz
(15)
where C is the concentration of the solute, and the parameter De is an effective vertical dispersion or mass transfer coefficient; De is a bulk coefficient and accounts for the interstitial flow velocity, hydrodynamic dispersion and molecular diffusion in the porous media flow. Both C and De are averages over horizontal distance. Eq. (15) does not have an advection term because the periodic vertical pore water velocities average out to zero, both over horizontal distance and over time. Effects of periodic advection, advective dispersion and molecular diffusion in the pore system of the sediment bed are lumped into the dispersion coefficient De(z), and an explicit description of the pore water velocity field in the mass transport equation is no longer required. Eq. (15) is for a conservative solute, but a source or sink term can be added. The boundary conditions are C (0,t) ¼ C0 at the sediment/water interface, and C(N,t) ¼ 0 at great depth into the sediment layer. The initial condition is C(z,0) ¼ 0. The mass or material flux across the bed/water interface can be calculated from Eq. (15) as J ¼ De vC/vzjz ¼ 0. The vertical mass transfer coefficient De in the sediment bed of a stream is similar to the longitudinal hydrodynamic dispersion coefficient DL of 1-D porous media flow as described by Bear (1972). Dispersion in 1-D flow through a porous and permeable medium is the results of two basic processes: advection and molecular diffusion. Both are described in detail by Bear (1972). Advective dispersion is due to flow, i.e. without flow there is only molecular diffusion. The complicated system of interconnected pores of the sediment bed causes a continuous division and rejoining of the flow and the solute in it. Since the travel times through different passages are not identical, some fluid with the solute in it will travel faster than other fluid, and the net result will be a gradual longitudinal distribution of the solute over an ever longer distance. This net effect is dispersion by advection. In our case the velocities will be reversible, and hence the dispersion “front” will not be carried away because there is no net advective velocity, only a ‘pumping effect’ by the back and forth flow into and out of the sediment pore system from above. Molecular diffusion is the second process and will interact with the advective dispersion in a variety of ways, sometimes enhancing advective dispersion, and sometimes undoing it. The longitudinal hydrodynamic dispersion coefficient (DL) depends on pore flow velocity and grain size. It can be several orders of magnitude larger than the molecular diffusion coefficient (D) which for many solutes in water varies from 106 to 105 cm2/s (Lide, 1999). 1-D longitudinal dispersion in flow through columns of porous media has been studied theoretically and experimentally. These are highlighted by Bear (1972). A relationship was established between the normalized hydrodynamic dispersion coefficient (DL/D) and a Peclet number of molecular diffusion, defined as Pe ¼ V ds/D (Bear, 1972, p.607), where V is the water velocity, ds is the diameter of the sediment particles and D is the molecular diffusion coefficient of the solute. The Peclet number Pe ¼ V ds/D is closely related to the Reynolds number of the porous media flow Re ¼ V ds/n0, because Pe ¼ Re Sc, where Sc ¼ n0/D is the Schmidt number, which is e.g. on the order of 500 for dissolved oxygen in water.
6079
Eq. (16) gives the relationship for longitudinal dispersion (Bear, 1972). The relationship has been confirmed by numerical simulations in Fig. 6.9 and 6.10, and extended to transverse dispersion in Fig. 6.10 and 6.11 by de Lemos (2006). DL =DzbðPeÞm
(16)
Bear (1972) distinguishes five ranges (zones) of Pe in which different processes control the 1-D dispersive transport, and hence the values of b and m differ, although in four of the five zones m z b z 1.0. In Zone 1, roughly when Pe < 0.5, molecular diffusion dominates and DL/D ¼ constant, i.e. m ¼ 0 and b z 2/3. In Zone 2 which corresponds roughly to 0.5 < Pe < 5 molecular diffusion and mechanical dispersion are of about the same magnitude, i.e. velocity effects are still small, and the sum of the two processes is effective. Mechanical dispersion means the effect of the separation and junction of pore channels. In Zone 3, which covers roughly the range 5 < Pe < 1000, the main spreading is caused by mechanical dispersion and molecular diffusion interferes with it. The two processes are no longer additive. Experimental results yield b z 0.5 and 1 < m < 1.2. In Zone 4 which covers roughly 1000 < Pe < 100,000, experimental data can be fitted to b z 1.8 and m z 1. Mechanical (convective) dispersion is dominant as long as Darcy’s law is still valid. Effects of molecular diffusion are negligible. Zone 5, roughly beginning at Pe > 100,000 is the range where inertial effects and turbulence may no longer be neglected. According to Bear (1972) the role of inertial effects and turbulence in Zone 5 is equivalent to the role of transverse molecular diffusion in Zone 3. The slope in zone 5 is m < 1. According to the analyses by Bear (1972) and de Lemos (2006) dispersion in flow direction is roughly proportional to pore flow velocity. Inertial effects in the pore water flow decrease hydraulic conductivity from K for Darcy flow to Ke when inertial effects are present (see Eqs. (12a) and (12b)). When Ke goes down, so does velocity, and with it the dispersion coefficient De. Inertial effects would be expected to reduce the dispersion coefficient De according to the relationship De;inertial =De;Darcy ¼ Ke =K
(17)
In our streambed analysis the flow field has vertical velocity components that are periodic and dominant, and inertial or turbulence effects can be present intermittently at least near the sediment/water interface. The limits of our analysis are approximately V < 10 cm/s, ds < 1 cm and n0 z 0.01 cm2/s. For these conditions and with Sc ¼ 500, one obtains Re < 1000 and Pe < 500,000. Inertial and turbulence effects would be expected to reduce longitudinal dispersion in a water column or vertical dispersion in a stream sediment bed for two reasons: (1) they reduce mean pore water velocity due to increased flow resistance, and (2) they increase the effect of transverse dispersion and thereby reduce longitudinal dispersion (reduced value of m in Zone 5). Thus the ratio of effective dispersion coefficients in the sediment bed De,inertial/De,Darcy is expected to be less than 1.0. In previous studies of hydrodynamic longitudinal dispersion by Bear (1972) the dispersion coefficient was found to be roughly linearly related to particle size and to pore flow
6080
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 7 4 e6 0 8 6
velocity, and the same parameterization was successful in applications, e.g. by Qian et al. (2009). The effective pore water dispersion coefficients (De) in Eq. (15) were calculated from Eq. (18). qffiffiffiffiffi De ¼ const$ q2z $f2 $ds
(18)
The root-mean-square of the vertical interstitial velocity qffiffiffiffiffi component q2z was chosen as the velocity scale. The pffiffiffiffiffi magnitude of the root-mean-squares statistic qz was calculated over a period of normalized time (t) such that U t ¼ 10 000. The instantaneous interstitial flow velocity d components qz(t) were from the pore water flow model described in the previous section. Eq.(12b) was used to account for flow resistance. The grain diameter (ds) was used as the length sale after it had been reduced by sediment porosity (f) and tortuosity (q) according to Eq. (19) to account for pore size. f d s ¼ f2 d s q2
(19)
An empirical relationship 1/q2 ~ ɸ for ɸ < 0.7 (Boudreau and Joergensen, 2001) is used in Eq. (19). The constant in Eq. (18) was set equal to 1.0. Eq. (18) is in formal agreement with the experimental relationship DL/D z b (Pem)m by Bear (1972), when both b and m are equal to 1.0. The effective vertical dispersion coefficient in the sediment bed (De) given by Eq. (18) was normalized to (De/Ds) by the reference diffusion coefficient (Ds) given by Eq. (20). Ds ¼
f D ¼ f2 D q2
(20)
The effective vertical dispersion coefficient (De,inertial) including non-linear effects, and the Darcy dispersion coeffifrom Eq. (18) with vertical cient (De,Darcy) are both ffiffiffiffiffi qcalculated velocity components ð q2z Þ that are either determined with Ke from Eq. (12b) for the non-Darcy flow, or with Ke ¼ K for the Darcy flow, respectively. De,inertial is compared to De,Darcy using the ratio (De,inertial/De, Darcy).
3.
Normalizations and numerical solutions
Eqs. (7b), (9a), (9b) and (12b) were normalized and solved numerically. To normalize, the shear velocity (U*) and the sediment depth (d) were used as velocity and length scales, respectively. The size of the computational domain was L ¼ 2pd in stream-wise direction, and d ¼ L/2p in vertical direction, respectively. The depth (d) of the sediment bed was chosen so that the fluctuating pressure would dampen to near zero at the bottom of the sediment bed. The number of grid points in the x and the z-directions was chosen as 100 100 with uniform spacing in the x direction. The mesh size in the zdirection was set to be ten times finer near the sediment/ water interface, where penetrating turbulent eddies were more active, than at the bottom of the domain. The boundary condition at the sediment/water interface (z ¼ 0) was given by Eq.(3). The boundary conditions at x ¼ 0
and x ¼ L were identical periodic fluctuating pressure head distributions. A no-flow boundary, i.e. vh=vz ¼ 0, was imposed at the bottom of the sediment domain (z ¼ d). As initial condition it was assumed that no excess pore pressure head existed inside the sediment, i.e. h ¼ 0, everywhere in the computational domain. Partial derivatives in the x-direction in Eq. (7b) were evaluated spectrally, whereas partial derivatives in the z-direction were approximated by a second order finite difference formula. For time advancing, the Crank-Nicolson method was used. Eq. (12a) gives the effective hydraulic conductivity (Ke) implicitly because the velocity component (qz) which is a function of the effective hydraulic conductivity (Ke) is included in the equation. To obtain the effective hydraulic conductivity (Ke) an iterative procedure was used; as initial condition, a constant hydraulic conductivity was imposed in the z-direction; then the equations were solved numerically and iteratively until the flow field was developed enough to calculate statistics, e.g. time-averaged pressure head (h), pffiffiffiffiffi fluctuating pressure head ðhz h2 Þ and so on; third, an effective hydraulic conductivity (Ke) as a function of the vertical coordinate (z) was obtained from Eq. (12b) using the previously computed qffiffiffiffiffivertical velocity component qz,i.e. the root-mean square ð q2z Þ, as input. This process was continued until the error (E ) given by Eq. (21) was small enough at each grid point. E¼
Knþ1 Kne e Knþ1 e
! (21)
where superscript n denotes the step. A value of E ¼ 0.0001 was imposed.
4.
Model results
4.1. Pressure head amplitudes due to near-bed coherent motions The amplitude of the pressure head fluctuation at the sediment/water interface due to near-bed coherent turbulent motions in the stream flow above the sediment bed was reported by Higashino et al. (2009) as h0 z1:7 104 m (1.7 Pa) for a shear velocity U* ¼ 1.6 cm/s. By comparison, maximum pressure head variations due to bed-forms are much larger (Elliott and Brooks, 1997). For a ratio H/d ¼ 0.34, where H is the water depth, and d is the bed form height, and a mean flow velocity ¼ 0.5 m/s over the sediment surface, the maximum pressure head variation for the bed form was estimated to be hm z3:6 103 m (35 Pa), which is 20 times larger than that due to near-bed coherent motions. Why do bed forms force solutes into and out of the sediment bed more effectively than pressure pulses from coherent turbulent structures on a flat stream bed? One reason is that bed forms produce standing pressure waves, whereas coherent turbulent flow structures produce pressure fluctuations at the sediment/water interface. Qian et al. (2008, 2009) demonstrated that standing surface waves above a flat sediment bed move solutes into and out of the sediment pore system at much higher rates
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 7 4 e6 0 8 6
6081
than progressive waves. The reason is that standing surface waves and bed forms produce a stationary flow field with a pressure distribution along the sediment surface that does not change much with time, whereas progressive waves cause a pumping action by the periodic reversal of flow direction over time at a given location on the sediment bed.
4.2.
Effective hydraulic conductivity in the sediment bed
Simulated effective hydraulic conductivity (Ke), normalized by the Darcy permeability coefficient (K ) of the sediment bed, is plotted against depth below the sediment/water interface in Figs. 3e5. As expected, non-linear effects are most significant immediately below the sediment/water interface. With increasing distance from the sediment/water interface, Ke approaches K, as was schematically shown in Fig. 1. Ke ¼ K, when the sediment bed has a low hydraulic conductivity, regardless of depth in the sediment bed (z) and regardless of shear velocity (U*) at the sediment/water interface. Non-linear (inertial) effects are negligible, if K < 1 cm/s (fine sand). For K ¼ 10 cm/s (sand) and K ¼ 100 cm/s (gravel) the simulated Ke/K < 1.0 below the sediment/water interface (z ¼ 0) as shown in Figs. 3e5. Ke depends on K as shown by a comparison of Figs. 3 and 4. Ke/K becomes smaller as K becomes larger (Fig. 5). Ke also becomes smaller as U* increases. Ke/K drops to a minimum of 0.5 at U* ¼ 1.6 cm/s (Fig. 3), but Ke/K z 1.0 for U* ¼ 0.32 cm/s. It can be concluded that non-linear effects are negligible when the sediment bed has a low Darcy permeability coefficients K, or when the shear velocity U* and hence the pressure fluctuations on the sediment bed is low. In the plots of Figs. 3, 4 and 5 the fraction to the left of each curve represents the viscous resistance and the fraction to the right the non-linear (inertial) resistance. For hydraulic
Fig. 3 e Dependence of normalized effective hydraulic conductivity (Ke/K ) on depth below the sediment/water interface (z), and on shear velocity (U*) for hydraulic conductivity (K ) [ 100 cm/s.
Fig. 4 e Dependence of normalized effective hydraulic conductivity (Ke/K ) on depth below the sediment/water interface (z), and on shear velocity (U*) for hydraulic conductivity (K ) [ 10 cm/s.
conductivity K ¼ 100 cm/s and shear velocity U* ¼ 1.6 cm/s in Fig. 3, the pore water flow experiences 45% viscous resistance and 55% non-linear (inertial) resistance just below the sediment/water interface (z ¼ 0). The non-linear resistance fraction becomes smaller as the distance (z) from the sediment/ water interface increases, and finally disappears completely.
Fig. 5 e Dependence of normalized effective hydraulic conductivity (Ke/K ) on depth below the sediment/water interface (z), and on hydraulic conductivity (K ) for shear velocity (U*) [ 1.28 cm/s.
6082
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 7 4 e6 0 8 6
Barr (2001) gave a critical head gradient vh=vzz0:1, at which the pore water flow begins to become turbulent. If vh=vz < 0:1 the pore water flow in the sediment was identified as laminar or in transition. The average pressure q head ffiffiffiffiffi gradient vh=vz estimated from the statistical value jqz jz q2z zKe $vh=vz can be smaller than 0.1; however, the instantaneous head gradient due to intermittent turbulence near the sediment/water interface can exceed 0.1. Intermittent turbulence can cause the non-linear resistance that is described by the effective hydraulic conductivity (Ke) in Figs. 3 and 4.
4.3. bed
Interstitial flow velocity in the permeable sediment
The instantaneous longitudinal and vertical velocity components of pore water flow, qx(t) and qz(t), respectively, were calculated by Eqs. (9a) and (9b), respectively, from the effective hydraulic conductivity (Ke) and the pressure head gradients vh/vx and vh/vz, respectively, after the pressure head distribution h (x,z,t)had been determined from Eq. (7b). Then the qffiffiffiffiffi qffiffiffiffiffi magnitude of the root-mean-squares statistics q2x and q2z were calculated. Since the sedimentq bed isotropic, the simulated qffiffiffiffiffi ffiffiffiffiffi is assumed flow field in terms of q2x and q2z is also isotropic as long as the pore water flow obeys Darcy’s law. The necessary condition for this to occur was earlier indicated to be a low hydraulic conductivity (K < 1 cm/s). The flow field inside the sediment becomes anisotropic for K ¼ 10 and K ¼ 100 cm/s, because the non-linear term in Eq. (12a) has a significant effect on the pore water flow field in the sediment. Fig. 6 illustrates the simulated normalized vertical pore qffiffiffiffiffi q2z =U for hydraulic conducwater velocity component tivities from K ¼ 0.01 cm/s to K ¼ 100 cm/s. The profiles are for
shear velocity U* ¼ 0.8 cm/s. The vertical pore water flow velocity is large at the sediment/water interface (z ¼ 0), and diminishes with depth (z). As can be expected, the velocity becomes larger as the hydraulic conductivity (K ) increases. The simulated vertical velocity component is on the order of 0.1~1 cm/s near the sediment/water interface (z ¼ 0) for K ¼ 100 cm/s.
4.4.
Effective solute dispersion coefficient
The dependence of normalized effective pore water solute dispersion coefficient in the sediment bed (De/Ds) on hydraulic conductivity (K), and shear velocity (U*) is shown in Fig. 7. The computed values of the ratio (De,inertial/De, Darcy) are shown in Figs. 8 and 9. As K increases, the ratio (De,inertial/De,Darcy) decreases to 0.85 and 0.66 for U* ¼ 0.80 and U* ¼ 1.28 cm/s, respectively (Fig. 8). In other words, non-linear (inertial) flow resistance decreases the dispersion rate by 15%e34% compared to what it would be if Darcy flow prevailed. When U* < 0.32 cm/ s, and/or K < 30 cm/s, non-linear effects are negligible and De,inertial/De, Darcyz1.0. Non-linear (inertial) effects become significant, i.e. De,inertial/De, Darcy < 1.0, as the hydraulic conductivity and/or the shear velocity become larger. When K > 80 cm/s and U* > 1.28 cm/s, the effective diffusion coefficient (De,inertial) is less than 70% of that obtained by Darcy flow (De, Darcy) because of non-linear flow effects. K > 80 cm/s, is representative of gravel. Fig. 9 summarizes the limitation of Darcy’s law for simulating the mass transport in the sediment bed based on the hydraulic conductivity (K ) which is related to the sediment particle diameter, and based on the shear velocity (U*) which is related to the near-bed turbulence.
5.
Discussion
The hyporheic flow and associated solute transport induced by a periodic pressure field on the surface of a stream bed has been investigated in this and other studies. A comparison of the results indicates that progressive pressure waves that
Fig. 6 e Normalized vertical pore water velocity component qffiffiffiffiffiffi q2z =U vs. depth (z) for hydraulic conductivity (K) [ 0.01e100 cm/s and shear velocity (U*) [ 0.80 cm/s.
Fig. 7 e Effective solute dispersion coefficients (De/Ds) predicted by the model with and without non-linear (inertial) effects vs. hydraulic conductivity of the sediment bed (K ) for shear velocity (U*) [ 0.32, 0.80, and 1.28 cm/s.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 7 4 e6 0 8 6
Fig. 8 e Dependence of the effective solute dispersion coefficient ratio (De,inertial/De, Darcy) at the sediment/water interface on the hydraulic conductivity of the sediment bed (K ) for shear velocity (U*) [ 0.32, 0.80, and 1.28 cm/s.
cause a reversal of pore water flow velocities in time are much less effective in solute transport than stationary pressure waves. As previously shown by Qian et al. (2008, 2009) the solute flux in a sediment bed induced by a stationary wave can be twenty times larger than that induced by a progressive wave. This important result, reported in Section 4.1, may at first be surprising. An explanation of this finding may be as follows: An oscillatory flow into and out of the sediment pore system moves fluid elements and dissolved material back and forth. Laminar (Darcy) flow experiments show that periodic advection and solute transport are nearly fully reversible. Dispersion only occurs if some of the substance is left behind along
6083
the path of the excursion; if little material is left behind, dispersion is very small. A progressive pressure wave associated with laminar pore water motion (Darcy flow) in the sediment bed can produce significantly more solute dispersion than molecular diffusion because it is not a simple uni-directional up-and-down motion. Under a pressure wave hyporheic flow is moving into the pore system at a location along the stream bed where the vertical pressure is highest, and out of the pore system where the pressure is lowest. These two locations are presumably separated horizontally by about half of the wavelength L in Eq. (1). When the wave is stationary, the locations of high and low pressure are stationary, and a continuous flushing action without flow reversal is produced. When the wave is progressive, the points of highest and lowest pressure are traveling in downstream direction, a vertical pumping action associated with flow reversal and a reversible horizontal pore water flow are produced. Another important finding, stated in Section 4.4 it is that non-linear (inertial) flow resistance decreases the dispersion rate by 15%e34% compared to what it would be if Darcy flow prevailed. This comparison is made with the assumption that the same mean flow velocities occur, but one flow produces inertial effects, and the other does not. Physically this is not possible. For inertial effects to occur, higher mean flow velocities would be required. The comparison illustrates only the inertial effects on mean flow velocity, and the associated effective dispersion coefficient (De); it does not consider the inertial effect or turbulence on mixing in the pore system. This effect can only be guessed based on the 1-D dispersion studies of Bear (1972). Zone 5, roughly beginning at Pe > 100,000, is the range where inertial effects and turbulence occur. According to Bear (1972) the role of inertial effects and turbulence is equivalent to the role of transverse molecular diffusion in Zone 3. The slope m < 1 in Zone 5, is an indication that the effective hydrodynamic dispersion coefficient increases less than linearly with pore water velocity. Compared to the other zones inertial effects and turbulence appear to reduce the impact of flow velocity on longitudinal dispersion. This same effect is observed for longitudinal dispersion of a tracer in 1-D river models (Fischer et al., 1979). Although not proven here, we would expect that the inertial or turbulence effect on mixing reduces vertical dispersion near the surface of the sediment bed. So the important result is that inertial or turbulence effects reduce vertical dispersion by both reducing mean flow velocities in the sediment pore system and by increased (lateral) mixing.
6. Comparison of model results with experimental data
Fig. 9 e Iso-lines of the effective solute dispersion coefficient ratio (De,inertial/De, Darcy) at the sediment/water interface of a stream including non-linear (inertial) effects in the pore water flow plotted against the shear velocity on the stream bed (U*) and the hydraulic conductivity of the sediment bed (K ).
O’Connor and Harvey (2008) assembled a significant number of data sets from experiments in which solute transfer across a sediment/water interface was measured. O’Connor and Harvey’s (2008) data range from smooth flat glass bead beds (Richardson and Parr, 1988; Lai et al., 1994) to coarse gravel beds (Tonina and Buffington, 2007); some include bed forms (e.g. Elliott and Brooks, 1997). The data were either collected in
6084
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 7 4 e6 0 8 6
the laboratory or in the field, and bed and flow parameters were measured as well. Based on dimensional analysis, O’Connor and Harvey (2008) determined three dimensionless numbers on which the surface transfer coefficient depended, and matched the experimental data collection to the Eq. (22). De ð0Þ 6=5 ¼ 5:0 104 Re Pek Ds
(22)
in which De(0) (cm2/s) is the mass transfer coefficient at the sediment/water interface (z ¼ 0); the reference diffusion coefficient Ds is calculated by Eq. (20). The independent dimensionless numbers (variables) in Eq. (22) are the permeability (k) based Peclet number (Pek)defined by Eq. (23) and the roughness Reynolds number (Re*) defined by Eq. (24), where ks is the roughness height that is taken to be three times the particle diameter, i.e. ks ¼ 3ds. Pek ¼
pffiffiffi U k Ds
Re ¼ U ks=v0
(23)
(24)
We determined numerical values for the three dimensionless parameters in Eq. (22) from the numerical results of the pore water flow model and compared the new non-linear flow and mass transport model results with O’Connor and Harvey’s (2008) empirical formula (Eq. (22)). As shown in Fig. 10, the mass transfer coefficient predicted by the proposed nonlinear flow model is somewhat smaller than that obtained by O’Connor and Harvey’s empirical formula when De/Ds < 10.
Fig. 10 e Comparison of O’Connor and Harvey’s (2008) empirical relationship (bold dashed line) with model simulation results (thinner grey line) for normalized effective solute (mass) diffusivity across the sediment/ water interface De(0)/Ds as a function of the Reynolds number-Peclet number parameter.
That condition requires a hydraulic conductivity K < 10 (Fig. 7), and a sediment particle size ds < 0.1 cm (Fig. 2). The pore water flow in that range is most likely laminar, follows Darcy’s law and has no inertial effects. For 105 > De/Ds > 10 the mass transfer coefficient predicted by the proposed model matches O’Connor and Harvey’s (2008) result more closely (Fig. 10). In that range of (De/Ds)-values, hydraulic conductivity is roughly 100 > K > 1 cm/s (Fig. 7), and sediment particle size is roughly 0.05 < ds < 1 cm (Fig. 2). Nonlinear (inertial) resistance becomes significant for most of these conditions (see Figs. 3 and 4). However, the expression developed by O’Connor and Harvey is derived from studies some of which include bed forms which produce an exchange mechanism that is different from the pressure pulses by coherent structures (Elliott and Brooks, 1997). Also, in very coarse (gravel) sediment beds, Brinkman flows may play an important role in the exchange mechanisms that are lumped in the O’Connor Harvey relationship. That the solute transfer coefficient (De/Ds) predicted by the new model, deviates from O’Connor and Harvey’s empirical formula when De/Ds is small, and that the match gets better as De/Ds becomes larger, is comforting, but not a validation, because the proposed model includes the non-linear resistance of pore water flow in response to penetrating turbulence but not the bed form or the Brinkman effects.
7.
Summary and conclusions
The penetration of low frequency turbulent pressure fluctuations into the sediment bed of a stream has been simulated by a numerical model, and the associated mass (solute) transport between the flowing water and the pore water has been estimated. The sediment bed has a hydraulic conductivity in the range 1 < K < 100 cm/s, and both viscous and non-linear (inertial) flow resistance were considered in the pore water flow model. The pressure fluctuations are associated with the near-bed coherent turbulent motions in the flow over the sediment bed, and are linked to the shear velocity on the sediment bed. An effective hydraulic conductivity (Ke) that is smaller than Darcy’s hydraulic conductivity (K ) was used when non-linear effects were significant. The following findings stand out: 1) The pressure fluctuations on the sediment surface induced by coherent turbulent structures in the stream flow are much less than those created by bed forms or stationary surface waves. 2) Non-linear (inertial or turbulent) non-Darcy effects on the pore water flow are negligible for sediment beds that have a hydraulic conductivity K < 1 cm/s, e.g. fine sand. Nonlinear effects are also insignificant when turbulence at the sediment/water interface (parameterized by the shear velocity) is small, i.e. U* < 0.32 cm/s (Figs. 3 and 4). Nonlinear effects become significant with higher hydraulic conductivity (K > 10 cm/s in Figs. 3e5). 3) Non-linear (inertial) effects are most significant immediately below the sediment/water interface. When inertial or turbulent effects in the pore water flow are present, effective hydraulic conductivity (Ke) is smaller than the Darcy
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 7 4 e6 0 8 6
4)
5)
6)
7)
hydraulic conductivity (K ). With increasing distance from the sediment/water interface into the sediment bed, Ke approaches K. When inertial effects are present, effective hydraulic conductivity (Ke) becomes smaller as the shear velocity (U*) increases. The simulated normalized effective hydraulic conductivity (Ke/K ) drops to 0.5 at the sediment/water interface when the shear velocity U* ¼ 1.6 cm/s, but is almost uniformly equal to 1.0 when U* ¼ 0.32 cm/s. Non-linear (inertial) effects are negligible at low shear velocity (Fig. 3). The mass (solute) transfer coefficient at the sediment/ water interface predicted by the proposed mechanistic non-linear flow and mass transport model closely follows O’Connor and Harvey’s (2008) results. The effective solute diffusion coefficient (De/Ds) appears to experience only a moderate change relative to Darcy flow conditions, when non-linear inertial effects increase flow resistance and reduce flow velocities for the pore water flow. The change is a reduction of less than 50% in the flow range investigated (U* < 1.6 cm/s and K < 100 cm/s). The degree to which non-linear effects change the effective diffusion coefficient depends on hydraulic conductivity (K ) and shear velocities (U*), as shown in Figs. 8 and 9.
Acknowledgments This work was supported by the Japan Society for the Promotion of Science (Young Researcher Overseas Visit Program, Number 21-5018), and by JSPS Grant-in-Aid for Scientific Research (No.22560522). Two reviewers provided helpful comments and suggestions on the manuscript. The authors are grateful to these individuals and organizations for their support.
references
Barr, D.W., 2001. Turbulent flow through porous media. Ground Water 39 (5), 646e650. Basu, A.J., Khalili, A., 1999. Computation of flow through fluidsediment interfaces in a benthic chamber. Physics of Fluids 12, 3074e3077. Bear, J., 1972. Dynamics of Fluids in Porous Media. American Elsevier Publishing Company, New York. Boudreau, B.P., Joergensen, B.B. (Eds.), 2001. The Benthic Boundary Layer: Transport Processes and Biogeochemistry. Oxford University Press, UK. Boudreau, B.P., 1997. A mathematical model for sedimentsuspended particle exchange. Journal of Marine Systems 11, 279e303. Brinkman, H.C., 1947. A calculation of the viscous force exerted by a flowing fluid on a dense swarm of particles. Applied Science Research Section A 2, 27. Cardenas, M.B., Wilson, J.L., 2004. Impact of heterogeneity, bedforms and stream curvature on sub-channel hyporheic exchange. Water Resources Research 40, 1e13. W0830740. Cardenas, M.B., Wilson, J.L., 2006. The influence of ambient groundwater discharge on exchange zones induced by currentbedform interactions. Journal of Hydrology 331, 103e109.
6085
Cardenas, M.B., Wilson, J.L., 2007. Hydrodynamics of coupled flow above and below a sediment-water interface with triangular bedforms. Advances in Water Resources 30, 301e313. Cardenas, M.B., Jiang, H.S., 2011. Wave-driven porewater and solute circulation through rippled elastic sediment under highly transient forcing, Limnology and Oceanography. Fluids and Environments 1, 23e37. doi:10.1215/21573698-1151658. Chien, Ning, Wan, Zhaohui, 1998. Mechanics of Sediment Transport. translated into English. ASCE Press, Reston VA. de Lemos, M.J.S., 2006. Turbulence in Porous Media: Modeling and Applications. Elsevier, Amsterdam, The Netherlands, p. 335. Dade, W.B., Hogg, A.J., Boudreau, B.P., 2001. Physics of flow above the sediment-water interface. In: Boudreau, B.P., Jørgensen, B.B. (Eds.), The Benthic Boundary Layer: Transport Processes and Biogeochemistry. Oxford Univ. Press, Oxford, U. K, pp. 4e37. DePinto, J.V., Lick, W., Paul, J., 1994. Transport and Transformation of Contaminants near the Sediment-Water Interface. CRC , Florida, Boca Raton. Elliott, A.H., Brooks, N.H., 1997. Transfer of non-sorbing solutes to a streambed with bed forms: theory. Water Resources Research 33 (1), 123e136. Fischer, H.B., List, E.J., Koh, R.C.Y., Imberger, J., Brooks, N.H., 1979. Mixing in Inland and Coastal Waters. Academic Press, New York (Chapter 5). Forchheimer, P., 1930. Hydraulik, third ed.. Teubner Publ., Leipzig, Berlin, 75 pp. Higashino, M., Stefan, H.G., 2008. A ‘velocity pulse-model’ for turbulent diffusion from flowing water into a sediment bed. Journal of Environmental Engineering 134 (7), 550e560. Higashino, M., Clark, J.J., Stefan, H.G., 2009. Porewater flow due to near-bed turbulence and associate solute transfer in a stream or lake sediment bed. Water Resources Research 45. doi:10. 1029/2008WR007374 W12414. Huettel, M., Webster, I.T., 2001. Porewater flow in permeable sediment. In: Boudreau, B.P., Joergensen, B.B. (Eds.), The Benthic Boundary Layer: Transport Processes and Biogeochemistry. Oxford University Press, UK, pp. 144e179. Lai, J.L., Lo, S.L., Lin, C.F., 1994. Effects of hydraulic and medium characteristics on solute transfer to surface runoff. Water Science Technology 30, 145e155. Lide, R.D., 1999. CRC Handbook of Chemistry and Physics. CRC Press, pp. 6e190. Mackenthun, A.A., Stefan, H.G., 1997. Effect of flow velocity on sediment oxygen demand: experiments. Journal of Environmental Engineering 124 (3), 222e230. Marion, A., Zaramella, M., 2005. Diffusive behavior of bedform induced hyporheic exchange in rivers. Journal of Environmental Engineering 131 (9), 1260e1266. Mendoza, C., Zhou, D., 1992. Effects of porous bed on turbulent stream flow above the bed. Journal of Hydraulic Engineering 118 (9), 1222e1240. Nagaoka, H., Ohgaki, S., 1990. Mass transfer mechanism in porous riverbed. Water Research 24 (4), 417e425. O’Connor, B.L., Harvey, J.W., 2008. Scaling hyporheic exchange and its influence on biogeochemical reactions in aquatic ecosystems. Water Resources Research 44. doi:10.1029/ 2008WR007160 W12423. Packman, A.I., Brooks, N.H., Morgan, J.J., 2000. A physicochemical model for colloid exchange between a stream and a sand streambed with bed forms. Water Resources Research 36 (8), 2351e2361. Packman, A.I., Brooks, N.H., 2001. Hyporheic exchange of solutes and colloids with moving bed forms. Water Resources Research 37 (10), 2591e2605. Packman, A.I., Salehin, M., Zaramella, M., 2004. Hyporheic exchange with gravel beds: basic hydrodynamic interactions and bedform-induced advective flows. Journal of Hydraulic Engineering 130 (7), 647e656.
6086
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 7 4 e6 0 8 6
Qian, Q., Voller, V.R., Stefan, H.G., 2008. A vertical dispersion model for solute exchange induced by underflow and periodic hyporheic flow in a stream gravel bed. Water Resources Research 44. doi:10.1029/2007WR006366, 2008, p. 11. Qian, Q., Clark, J.J., Voller, V.R., Stefan, H.G., 2009. Depthdependent dispersion coefficient for modeling of vertical solute exchange in a lake bed under surface waves. Journal of Hydraulic Engineering 134 (3), 11. Richardson, C.P., Parr, A.D., 1988. Modified friction model for solute uptake by runoff. Journal of Environmental Engineering 114, 792e809. Shimizu, Y., Tsujimoto, T., Nakagawa, H., 1990. Experiment and macroscopic modeling of flow in highly permeable porous medium under free-surface flow. Journal of Hydroscience and Hydraulic Engineering, Japan 8 (1), 69e78. Steinberger, N., Hondzo, M., 1999. Diffusional mass transfer at the sediment-water interface. Journal of Environmental Engineering 125 (2), 192e200. Tonina, D., Buffington, J.M., 2007. Hyporheic exchange in gravel bed rivers with pool-riffle morphology: laboratory experiments and threedimensional modeling. Water Resource Research 43. doi:10.1029/2005WR004328 W01421. Vanoni, V. A. (Ed.), (1975). Sedimentation Engineering, ASCE – Manuals and Reports on Engineering Practice – No. 54, 745 pp. Zhou, D., Mendoza, C., 1993. Flow through porous bed of turbulent stream. Journal of Engineering Mechanics, ASCE 119 (2), 365e383.
Glossary a: coefficient in the continuity equation (cm2 s1) C: concentration of solute (g cm3) d: bedform height (cm) ds: sediment particle diameter (cm) D: molecular diffusion coefficient of solute (cm2 s1) De: effective dispersion coefficient of solute (cm2 s1) De(0): effective mass transfer coefficient of solute at the sediment surface (cm2 s1) Ds: reference (Darcy) diffusion coefficient of solute (cm2 s-1) DL: hydrodynamic longitudinal dispersion coefficient (cm2 s1) g: gravitational acceleration (¼ 9.8 m s2)
H: water depth (cm) h: excess pore pressure head (cm) or piezometric head (cm) K: hydraulic conductivity (cm s-1) K0: initial estimate of hydraulic conductivity (cm s1) z K Ke: effective hydraulic conductivity (cm s1) k: permeability (cm2) L: pressure wavelength mv: sediment compressibility (cm2 N1) p: pressure (N cm2) p0: amplitude of pressure fluctuation at the sediment/water interface (N cm2) Pe: Peclet number of molecular diffusion Pek: permeability-based Peclet number q: velocity in sediment bed (cm/s) qx: velocity component in stream-wise direction (cm s1) qffiffiffiffiffi q2x : root-mean-squares of velocity component in stream-wise direction (cm s1) qz: velocity component in vertical direction (cms1) qffiffiffiffiffi q2z : root-mean-squares of velocity component in vertical direction (cms1) Rek: roughness Reynolds number t: time (s) T: period of near-bed coherent motions (s) U*: bed shear velocity (cm s1) Uc : critical bed shear velocity (cm s1) Us: slip velocity (cm s1) x: horizontal coordinate (cm) z: vertical coordinate from the sediment/water interface (cm) aL: longitudinal dispersivity (cm) aT: transverse dispersivity (cm) d: thickness of sediment bed (cm) din: penetration depth of inertial effect (cm) dp: penetration depth of pore water velocity n: kinematic viscosity of water (¼ 0.01 cm2 s-1) r: water density (g cm-3) s: angular frequency (s1) s0: bed shear stress (N cm2) ɸ: sediment porosity q: tortuosity c: wave number (cm1)
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 8 7 e6 0 9 6
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Tertiary amines enhance reactions of organic contaminants with aqueous chlorine Amisha D. Shah 1, Jae-Hong Kim, Ching-Hua Huang* School of Civil and Environmental Engineering, Georgia Institute of Technology, 200 Bobby Dodd Way, Atlanta, GA 30332, USA
article info
abstract
Article history:
Through various anthropogenic inputs, tertiary amines can readily contaminate waste-
Received 21 June 2011
water and drinking water sources and can form chlorammonium species (R3Nþ-Cl) during
Received in revised form
aqueous chlorine disinfection. This study investigated the less understood concept that
30 August 2011
these chlorammonium species can potentially enhance organic contaminant loss and
Accepted 3 September 2011
increase disinfection byproduct formation to a greater extent than aqueous chlorine.
Available online 10 September 2011
Tertiary amines’ effectiveness was highly dependent on amine structure as trimethylamine (TMA) and 4-morpholineethanesulfonic acid (MES) enhanced organic contaminant
Keywords:
loss, while others (nitrilotriacetic acid (NTA) and creatinine (CRE)) were ineffective. MES
Tertiary amine
addition up to 25 mM led to increased organic contaminant chlorination by up to three
Chlorination
orders of magnitude while observing pseudo-first order kinetic behavior and a linear amine
Trimethylamine
dose response. TMA addition up to 0.5 mM accelerated organic contaminant chlorination by
4-Morpholineethanesulfonic acid
almost two orders of magnitude, but occasionally deviated from pseudo-first order kinetics
Chlorammonium
with incomplete organic contaminant degradation and a non-linear amine dose response -
Disinfection byproducts
a result linked to TMA’s rapid auto-decomposition over time. Byproduct formation was identical with and without amine addition, and thus the chlorination mechanisms are likely similar to aqueous chlorine. Results from this study improve the mechanistic understanding behind tertiary amine-enhanced chlorination. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
Tertiary amines are a class of compounds often present in natural and treated waters. They originate from a variety of sources (Hwang et al., 1995; McArdell et al., 2003; Miao et al., 2004; Park et al., 2009) and exhibit a wide range of structural and chemical properties. Tertiary amine functionalities are present in compounds such as trimethylamine (TMA), where
concentrations ranging between 16 and 78 mg/L (0.3e1.3 mM) have been detected at various sampling points in a Tokyo wastewater treatment facility (Hwang et al., 1995), as substituents in numerous pharmaceuticals often found in wastewater effluent at concentrations up to 1.0 mg/L (McArdell et al., 2003; Miao et al., 2004), and in several polymeric compounds commonly used as coagulants and flocculants in water and wastewater treatment (Kohut and Andrews, 2003;
Abbreviations: TMA, trimethylamine; NADH, nicotinamide adenine dinucleotide; NOM, natural organic matter; SA, salicylic acid; FLU, flumequine; TMP, trimethoprim; MES, morpholineethanesulfonic acid; NTA, nitrilotriacetic acid; CRE, creatinine; ACC, acetylcholine; 3Cl-SA, 3-Chlorosalicylic acid; 5-Cl-SA, 5-chlorosalicylic acid; 3,5-diCl-SA, 3,5-dichlorosalicylic acid; THAM, tris(hydroxymethyl)aminomethane; UV/vis, ultraviolet/visible; HPLC, high performance liquid chromatography; FLD, fluorescence detector; DAD, UV/vis detector; MSD, mass spectrometer detector; FAC, free available chlorine; DMA, dimethylamine; MMA, monomethylamine. * Corresponding author. Tel.: þ1 404 894 7694; fax: þ1 404 385 7087. E-mail address: [email protected] (C.-H. Huang). 1 Current address: Yale University, Mason Laboratories, 9 Hillhouse Ave., Room 308, New Haven, CT 06511, USA. 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.09.010
6088
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 8 7 e6 0 9 6
Park et al., 2009). But while individual tertiary amines can be measured, no known studies have assessed total tertiary amine concentrations in real water matrices as it is analytically challenging. Past studies suggest that the reaction of tertiary amines with aqueous chlorine (HOCl/OCl) that is added during water treatment is unique when compared to ammonia and other aliphatic amines (e.g., primary/secondary amines). First, the reaction rate constants (kHOCl) of the neutral tertiary amine species with HOCl range between 103 and 104 M1 s1 (Antelo et al., 1985) and are typically orders of magnitude lower than those of ammonia or primary/secondary amines which range from 2 106 to 5 106 M1 s1 (Margerum et al., 1978; Morris and Isaac, 1983) and 107e108 M1 s1 (Abia et al., 1998; Antelo et al., 1992, 1995; Margerum et al., 1978; Morris, 1967), respectively. Second, the effect of substituent polarity effects on the reactivity of basic tertiary amines to HOCl/OCl is opposite from primary/secondary amines. Primary/ secondary amines’ rate constants increase as substituents are more electron-withdrawing, and thus plotting them logarithmically against the Taft’s constant (s*) leads to a positive slope (r ¼ 1.14 0.26). In contrast, plotting the tertiary amines’ rate constants logarithmically against the s*, i.e. with more electron-withdrawing substituents, leads to a negative slope (r ¼ 2.24 0.82) (Abia et al., 1998; Deborde and von Gunten, 2008). This suggests that chlorination occurs through different mechanisms depending on amine type. Basic primary/ secondary amines are hypothesized to form a highly-ordered, negatively-charged transition state where water molecules are hydrogen-bonded to both HOCl and the amine N to form a ten-membered ring (Abia et al., 1998). The negative charge is formed due to an asynchronous process where the proton transfer from the amine N to water occurs prior to the chlorine transfer from HOCl/OCl to the amine N (Abia et al., 1998). Electron-withdrawing substituents thus increase the amine N’s reactivity as they stabilize the negative charge in the intermediate. In contrast, tertiary amines contain lone-pair electrons on the amine N which can directly react with the chlorine atom in HOCl/OCl to form a positively charged chlorammonium species (R3Nþ-Cl) (Abia et al., 1998), and electron-withdrawing substituents make donating such electrons less likely. An even greater discrepancy is found once the respective chlorinated amine species are formed and oxidize other organic substrates in the solution matrix. For many of these chloramines and chlorinated primary/secondary amines, their oxidizing power is relatively low. Monochloramine (NH2Cl) generally reacts about four orders of magnitude slower in comparison to HOCl (Morris, 1967). In addition, the apparent second-order rate constants (kapp) for oxidation of certain biological substrates such as NADH and ascorbate by chlorinated primary amines range from 2.8 to 76 M1 s1, similarly about 4e5 orders of magnitude lower than those by HOCl (Peskin and Winterbourn, 2001; Prutz et al., 2001). This is in contrast to chlorinated tertiary amines (chlorammonium ions) which appear to have a substantially greater chlorination potential. A number of studies indicate that chlorinated tertiary amines, either generated in-situ (e.g., formed upon addition of tertiary amines and HOCl) or pre-formed as
a chlorinated salt, may greatly enhance chlorination rates in comparison to HOCl (Dodd et al., 2005; Prutz, 1998, 2001). The objective of this study was to systematically evaluate the impact that tertiary amines and their chlorinated counterparts (i.e., chlorammonium species) have on the reaction of various organic compounds with aqueous chlorine. While studies have previously assessed the effect of tertiary amines and their chlorammonium species in biological matrices, no known studies to date have addressed their importance during wastewater and water treatment. Model compounds representing natural organic matter (NOM) constituents and organic contaminants of varying structural characteristics were selected and included salicylic acid (SA), flumequine (FLU) and trimethoprim (TMP) (Fig. 1). SA is found in aquatic humic material (Larson and Rockwell, 1979), FLU is from the group of popular fluoroquinolone antibacterial agents (Walsh, 2003), and TMP is an antibacterial agent commonly detected in wastewater effluent and surface waters (Kolpin et al., 2002). The model tertiary amines investigated were TMA, 4morpholineethanesulfonic acid (MES, common organic buffer), nitrilotriacetic acid (NTA, common chelating agent), and creatinine (CRE, found in urine (Murray et al., 2000)). One quaternary amine, acetylcholine (ACC, biological neurotransmitter (Sletten et al., 2005)), was also included in this study as quaternary amines may transform into tertiary amines under chlorination reaction conditions (see later discussions).
2.
Experimental section
2.1.
Chemicals and reagent preparation
Sources of chemicals and preparation of reagents are described in the Supplementary information Text S1.
2.2.
Reaction setup
To assess organic contaminant loss over time, batch reactions were conducted in capped 25-mL amber glass vials at circumneutral pH (pH 7.0e7.3 adjusted by 10 mM NaH2PO4/Na2HPO4 buffer) under continuous stirring at 23 C. The initial organic contaminant concentration was 5.0 mM (10.0 mM for TMP byproduct analysis). One set of experiments was performed by spiking 10 or 40 molar amounts of HOCl/OCl relative to initial organic contaminant concentrations. Another set of experiments was performed with amine addition, in which varying concentrations of tertiary or quaternary amine ([tertiary or quaternary amine]i ¼ 0.25 to 25 mM (0.05e5 [model compound]i)) were first added into the reactor, and then HOCl/OCl was added at 10 [model compound]i. Sample aliquots were periodically taken and instantaneously quenched with ascorbic acid. Ascorbic acid was verified to have negligible effect on changing either the organic contaminant (SA, FLU, and TMP) concentrations or SA and FLU byproduct concentrations over time. For TMP byproduct analysis, ascorbic acid significantly altered the byproducts formed, consistent with previous observation made with Na2S2O3 as a quenching agent (Dodd and Huang, 2007). Thus, a “softer” quenching technique using NH4Cl/tris(hydroxymethyl)aminomethane (THAM)/CH3COOH
6089
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 8 7 e6 0 9 6
Organic Contaminants: Salicylic acid (SA) O
Flumequine (FLU)
Trimethoprim (TMP) O
OH1 OH
1
OH
F O
pKa1 = 2.9a
2 NH2
N
O
N
N
O
NH2 1
O
p Ka1 = 6.5b
pKa1 = 3.2c, pKa2 = 7.1d
Tertiary/Quaternary Amines: Trimethylamine (TMA)
4-Morpholineethanesulfonic acid (MES)
Nitrilotriacetic acid (NTA) O
OH
O
1
1
N
O
S
N
1
O
HO
N
OH
O O
pKa1 = 9.8
e
pKa1= 6.1 f Creatinine (CRE) N O
pKa1 = 4.8
pKa1= 0.7-1.3 g-i Acetylcholine (ACC) O
1
3 N
HO
N+
NH2 2
O
j,k
Note: Number associated with ionization site is equal to pKa number (e.g. site corresponding to pKa1 is labeled “1”).
Fig. 1 e Structures and speciation of the model compounds and tertiary/quaternary amines investigated in this study. Structures and speciation of the model compounds and tertiary/quaternary amines investigated in this study. a(Geiser et al., 2005), b(Barbosa et al., 2001), c(Qiang and Adams, 2004a), d(Roth and Strelitz, 1969), e(Jones, 1997), f(Good et al., 1966), g(Abia et al., 1998; Martell and Smith, 1974), h(Mederos et al., 1987), i(Sanchiz et al., 1999), j(Grzybowski and Datta, 1964), k (Kotsyubynskyy et al., 2004; Qiang and Adams, 2004b).
was adopted from (Dodd and Huang, 2007) for TMP byproduct analysis. The apparent reaction rate constant (kapp) for MES with HOCl/OCl (Table 1) at pH 7.0 (10 mM phosphate buffer) and the forward reaction rate constant (kforward) for TMA and MES with HOCl/OCl at pH 13 and 10.5 (200 mM phosphate buffer), respectively, were determined by adding HOCl/OCl and varying excess amine doses ([TMA/MES] ¼ 10 80 [HOCl/ OCl]) to a capped cuvette with zero head-space. Excess amine doses were added in order to establish pseudo-first order conditions. The cuvette was placed in a UV/vis spectrophotometer where the decay of HOCl/OCl at 292 nm or 310 nm was immediately monitored.
2.3.
Analysis of organic contaminants
Loss of the organic contaminant over time was monitored using an Agilent 1100 HPLC system with a Zorbax RX-C18 column (4.6 250 mm, 5 mm) at a flow rate of 1.0 mL/min.
Table 1 e Apparent second-order rate constants for various amines and organic contaminants with free chlorine. Compound
Amines: TMA MES Organic Contaminants: SA FLU TMP a (Abia et al., 1998; Antelo et al., 1985). b Experimentally determined. c (Dodd and Huang, 2007).
Apparent Rate Constant (M1 s1) kFAC app at pH 7.0 69a 11.3b 7.8 102b 0.26b 20.7c
6090
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 8 7 e6 0 9 6
SA and FLU were detected using a fluorescence detector (FLD) with excitation and emission wavelengths of lEX ¼ 235 nm, lEM ¼ 410 nm and lEX ¼ 312 nm, lEM ¼ 366 nm, respectively. TMP was detected using a diode-array UV/vis detector (DAD) at absorption wavelength of 205 nm. The mobile phase for SA consisted of 0.1 M acetic acid þ 5%vol methanol solution (eluent A) and methanol (eluent B) in gradient mode. The mobile phase for FLU and TMP consisted of 40 mM phosphate buffer (H3PO4/NaH2PO4) solution (eluent A) and acetonitrile (eluent B) and eluted in isocratic (60% A, 40% B) and gradient mode, respectively.
2.4. Analysis of organic contaminant reaction byproducts Reaction byproducts were analyzed by an Agilent 1100 HPLC/ DAD/MSD system with a Zorbax SB-C18 column (2.1 150 mm, 5 mm) at a flow rate of 0.2 mL/min. Mobile phases similar to HPLC analysis were used for SA, FLU, and TMP except that eluents A comprised of 0.2% formic acid for SA and TMP and 0.1 M acetic acid þ 5% MeOH for FLU. MS analysis was conducted by electrospray ionization in negative (SA) and positive (FLU and TMP) modes at both low and high fragmentation voltages (80 and 220 eV) to yield optimal low and high fragmentation patterns with a mass scan range of m/ z 50 to 1000. The drying gas was at 10 L/min at 350 C, the nebulizer pressure 25 psig, and the capillary voltage 4000 V.
3.
Results and discussion
3.1. Reactivity of organic contaminants and amines with free chlorine Reactions of SA, FLU, and TMP with free available chlorine (FAC ¼ [HOCl] þ [OCl]) in the absence of tertiary amines were characterized by the following second-order rate expression: d½org:cont:T FAC ¼ kFAC app ½org:cont:T ½FACT ¼ kobs ½org:cont:T dt
(1)
Subscript T represents the total concentration of all protonated/deprotonated species for each compound. Under excess FAC conditions, the logarithm of SA, FLU, and TMP concentrations linearly decreased over time (R2 > 0.97 for all replicates), verifying first-order kinetics with respect to the organic contaminant. First-order kinetics with respect to FAC was confirmed for SA and FLU by adding varying amounts of excess FAC (0.2e0.5 mM), where a linear trend (R2 ¼ 0.97e0.99) was found between FAC concentration and the pseudo-first order rate constant, kFAC obs . For TMP, first-order kinetics with respect to FAC was confirmed in a previous study (Dodd and Huang, 2007). The apparent second-order rate constants 1 1 1 s )), obtained by dividing kFAC ) by [FAC]T at pH (kFAC app (M obs (s 7.0, for SA, FLU, and TMP, are listed in Table 1. Reactions of TMA and MES with FAC in the absence of organic contaminants were also evaluated and could be characterized by the rate expression in Eq. (1). The literature (Abia et al., 1998; Antelo et al., 1985) and experimental values 1 1 s ) for TMA and MES at pH 7.0, respectively, are of kFAC app (M listed in Table 1. The kFAC app of the other amines were not
determined as they did not enhance organic contaminant chlorination (see later discussions). By comparing the apparent rate constants in Table 1, it is evident that the selected acid-base organic compounds (except for the neutral TMP species) are unlikely to compete effectively with the tertiary amines for free chlorine when present at concentrations equimolar to or less than those of the tertiary amines.
3.2. Effects of MES, TMA and other amines on chlorination Experiments individually assessed five different amines, TMA, MES, NTA, CRE, and ACC, for their influence on SA, FLU, and TMP reactivity in the presence of excess (10) FAC. In all cases, the loss of organic contaminant was compared with and without amine addition. The amine was considered to enhance reactivity when [organic contaminant]w/amine < [organic contaminant]w/o amine at any t > 0. MES and TMA were found to enhance SA, FLU and TMP degradation when 0.05e25 mM of either amine was dosed into the reaction matrix which will be discussed in detail later. For NTA, no enhancement of SA reactivity was observed ([NTA]i ¼ 10e25 mM, pH 7.2), and only a slight increase was found for TMP when 25 mM NTA was added, in which the observed rate constant (kTMP obs ) increased to 2.6 103 s1 from the 1.5 103 s1 value observed without NTA (pH 7.2e7.3). CRE and ACC also did not influence SA, FLU, or TMP degradation ([CRE]i or [ACC]i ¼ 2.5e10 mM, pH 7.2e7.3). We herein hypothesize based on the literature (Prutz, 1998) and the present observations for TMA and MES, that certain tertiary amines, even at low concentrations, could enhance the reactivity of organic contaminants with FAC via the following catalytic mechanism: k1app
R3 N þ HOCl/R3 Nþ Cl þ OH
kchlorammonium app
R3 Nþ Cl þ org:cont: / org:cont:product þ R3 N
(2)
(3)
The enhanced transformation of an organic contaminant by the amine is therefore dependent on favorable formation of the intermediate chlorammonium species, R3Nþ-Cl. R3Nþ-Cl can be formed from the initial reaction of the amine with FAC, the degree of which could be correlated to the tertiary amine’s Lewis basicity, i.e., the overall reactivity of the amine nitrogen’s electron pair and its protonation potential in forming R3Nþ-H (quantitatively assessed using its pKa). A similar correlation has been made for other amines forming N-chloro compounds by Morris (1967). As listed in Fig. 1, the pKa values for TMA (pKa ¼ 9.8) and MES (pKa ¼ 6.1) are greater than those for NTA (pKa ¼ 0.7e1.3) and CRE, whose tertiary amine has no pKa under relevant pH conditions, but rather only a pKa at the N3 nitrogen (pKa ¼ 4.8). Thus, TMA and MES are considered to be more favorable toward forming R3Nþ-H, and consequently R3Nþ-Cl, in comparison to NTA and CRE. ACC was not found to enhance the reaction of organic contaminants over time, due to lack of reactive nitrogen lone-pair electrons. Although certain quaternary amines (R4Nþ) might dealkylate to form
6091
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 8 7 e6 0 9 6
with [TMA]i ¼ 10 mM). In contrast to the reactions enhanced by MES, reactions enhanced by TMA strongly deviated from pseudo-first order kinetics where the target organic contaminants decayed rapidly up to w45 min and then decayed at a significantly slower rate up to 360 min. This result is most likely due to auto-decomposition of TMA after its reaction with FAC. Previous studies (Ellis and Soper, 1954; Mitch and Schreiber, 2008) indicate that after being formed, the chlorammonium ion alternatively undergoes an elimination reaction to lose HCl and create an iminium ion which is then followed by hydrolysis to form dimethylamine (DMA) and formaldehyde, as follows:
tertiary amines (R3N) and subsequently react with free chlorine to form R3Nþ-Cl (Kemper et al., 2010; Park et al., 2009), such a degradation seemed negligible with ACC. Since TMA and MES were found to be the most effective as chlorination catalysts of the amines investigated here, they were the focus of subsequent analyses to better understand the mechanisms of tertiary amine’s catalytic activity toward chlorination of organic compounds. As shown in Fig. 2, complete loss of parent SA and FLU occurred well within 6 h when 2.5e25 mM MES was added, but complete conversion was never achieved with equivalent doses of TMA (only 69% and 51% conversion for SA and FLU, respectively, was reached
a
d
2.0
1.0
ln [SA] (µM)
ln [SA] (µM)
1.0
0.0
0.0
-1.0
SA + MES
SA + TMA -1.0
0
10
20
30
40
50
0
200 300
10
20
time (min)
b
100 200 300 400
time (min)
e
2.0
2.0 1.8 1.6
1.0 ln [FLU] (µM)
ln [FLU] (µM)
1.4
0.0
-1.0
1.2 1.0 0.8 0.6 0.4
-2.0
FLU + TMA
0.2
FLU + MES
0.0
0
10
20
30
40
50
200
0
300
20
40
time (min)
f
2
200
300
400
2.0
1
1.0
0
0.0
ln [TMP] (µM)
ln [TMP] (µM)
c
100
time (min)
-1 -2
-1.0 -2.0 -3.0
-3
TMP + MES
TMP + TMA -4.0
-4 0
2
4
6
time (min)
8
10
12
0
2
4
6
8
10
time (min)
Fig. 2 e SA (a,d), FLU (b,e), and TMP (c,f) degradation expressed as the ln[SA, FLU, or TMP] over time by addition of (aec) MES or (def) TMA and excess free chlorine (103[org.cont.]i). For SA and FLU, amine dosages range from no amine addition (free chlorine only) (;), 0.25 mM (,),2.5 mM (6),10 mM (>), to 25 mM (B). For TMP, amine dosages range from no amine addition (free chlorine only) (;), 0.05 mM (B), 0.1 mM (6), 0.25 mM (,), to 0.5 mM (>). Error bars represent the standard error of ‡3 replicates. (pH 7.0e7.3, 23 C).
6092
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 8 7 e6 0 9 6
As reported by Mitch and Schreiber (2008), the half-life of TMA at pH 7.0 was <2.5 min when FAC was added at 1e4 times the TMA molar concentration, and the products DMA and formaldehyde were found at near stoichiometric amounts. MES, on the other hand, is considered to be less susceptible toward the above reaction mechanism since its tertiary amine nitrogen is bound within a stable ring structure and thus less prone to the CeN bond cleavage. For TMP degradation (Fig. 2c,f), >93% conversion was achieved within just 20 min for all MES and TMA doses tested. In these cases, autodecomposition of chlorinated TMA ((CH3)3Nþ-Cl) was probably not significant enough as the observed reaction rate of TMP loss did not change over time.
3.3. Effects of amine dose, oxidation efficiency, and functional group reactivity The effect of amine dosage was examined at 0.25e25 mM for SA and FLU but at 0.05e0.5 mM for TMP due to the faster reaction of TMP with FAC alone. When higher amine doses were used, TMP degradation was too rapid (<10 s) to monitor in batch reactors. Plotting the ln[SA, FLU, or FLU] versus time showed that most experimental results (Fig. 2aec,f) fit well to a pseudo(in s1), the rate constant for first order decay, where kFACþAmine obs the decay of organic contaminant in the presence of FAC and amine, could be obtained from the linear regression of such plots ( R2 ¼ 0.95e0.99) except for the TMA-enhanced reactions FACþAmine values with SA and FLU (Fig. 2d,e). The obtained kobs were normalized against the reaction rate constant without amine present to quantify the overall rate enhancement effect, as listed in Supplementary information Table S1, and were also found to linearly increase with increasing MES dosage for all three organic contaminants and with increasing TMA dosage for TMP (Fig. 3). This linear increase suggests that the chlorammonium concentration formed in solution, given an excess FAC concentration, is linearly proportional to the amine dosage added, as expected on the basis of Eq. (2). On the other hand, the TMA-enhanced reactions of SA and FLU with FAC, while not abiding by pseudo-first order kinetics, also did not necessarily increase at a faster rate in accordance to an increased TMA dose. As seen in Fig. 4a,c, the % conversion of SA and FLU at 3 min was enhanced when [TMA]i increased from 0.25 to 10 mM but no further change was observed upon 25 mM addition. Similar trends were also observed at 360 min (Fig. 4b,d) except that the % conversion of SA and FLU decreased significantly (22e25% drop) when 25 mM TMA was added. Such an enhancement loss could be due to two factors. One possibility is related to the additional chlorine demand by TMA’s autodecomposition byproducts (DMA and monomethylamine (MMA)) which react at a faster rate with FAC at pH 7 (kapp (for DMA) ¼ 8.9 103 M1 s1 (Abia et al., 1998), kapp (for MMA) ¼ 3.2 104 M1 s1 (Margerum et al., 1978)) than TMA. This consequently limits the chlorine available to react with TMA or SA/FLU directly. In addition, an increased TMA concentration
Fig. 3 e (a) Effect of MES and TMA dosage on the observed Þ for chlorination of SA, FLU and rate constant ðkFACDamine obs TMP. Error bars represent the standard error of ‡3 values. ([FAC]0 [ 50 mM, pH [ 7.0e7.3, 23 C). kFACDamine obs
could increase its own auto-decomposition. Previous evidence has shown that the rate of chlorammonium decay was proportional to TMA at high TMA concentrations (Ellis and Soper, 1954). It was proposed that TMA could act as a nucleophile and enhance proton removal from the chlorammonium ion’s a-carbon during the initial elimination step of the autodecomposition process (Ellis and Soper, 1954), as follows:
Results shown in Fig. 3 suggest that TMA (up to 0.5 mM) values (1.4e3.3) than equivalent exhibited higher kFACþAmine obs amounts of MES. For SA and FLU, a comparison can only be made at earlier reaction times and low amine doses where R3Nþ-Cl decomposition is not yet significant (see above discussions). In this case, TMA addition was slightly more efficient than MES where % conversion of SA and FLU was 5e14% higher. These collectively suggest that a stable TMA-Clþ appears to be a more effective oxidant than MES-Clþ, but autodecomposition limits its overall oxidation potential. The experimental results in Fig. 3 also permit comparison toward how effective certain functional groups are when reacting with chlorammonium oxidants. The slopes in Fig. 3 represent susceptibility of the target organics toward MESþ-Cl and follow the order of TMP > FLU > SA. Similarly, the enhancement on a compound’s % conversion with TMA also followed the order of TMP > FLU > SA at 3 min and low amine dosages (<0.25 mM). This order of reactivity is similar to that for the reaction with FAC alone (Table 1), suggesting that the reaction mechanisms involved with the chlorammonium oxidant and with free chlorine may be similar.
3.4.
Evaluation of amine-enhanced reaction kinetics
As discussed in the previous section, the amine-enhanced reaction kinetics (i.e., MES-enhanced reactions and TMA-
6093
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 8 7 e6 0 9 6
Organic Contaminant Conversion (%)
100
a
c
SA-3 min
FLU-3 min
TMA MES
80 60 40 20 0 100
d FLU
b SA-360 min
360 min
80 60 40 20 0 0.25
2.5
10
25
0.25
Amine ( M)
2.5
10
25
Amine ( M)
Fig. 4 e SA (a,b) and FLU (c,d) % conversion at 3 or 360 min when TMA or MES was added. Error bars represent the standard error of ‡3 replicates. ([FAC]0 [ 50 mM, pH 7.0e7.3, 23 C).
enhanced TMP degradation) followed pseudo-first order decay FACþAmine ), which is then assumed to follow the rate (kobs expression of Eqs. (4) and (5) that included the additional oxidant (R)3Nþ-Cl and its apparent second-order reaction rate : constant, kchlorammonium app d½org:cont:T ¼ kFACþAmine ½org:cont:T obs dt FAC ¼ kobs ½org:cont:T ½FACT kchlorammonium app ½org:cont:T R3 Nþ Cl
(4)
chlorammonium ¼ kFAC R3 Nþ Cl kFACþAmine app ½FACT þkapp obs
(5)
To evaluate whether the right side of Eq. (4) (¼ kFAC app $ þ $½org:cont: $½R N Cl) could ½org:cont:T $½FAC kchlorammonium 3 app T accurately predict the experimentally observed kinetics, a numerical solution was established to simulate the kinetics of organic contaminant loss in the presence of FAC and the amine. The numerical solution was programmed using Matlab 7.4.0 (R2007a) (detailed code provided in Supplementary information Text S2) in which a system of five first-order ordinary differential equations (Supplementary information Text S3) derived from Eqs. (2) and (3) was solved simultaneously to determine the concentrations of the five unknown species which included the organic contaminant, FAC, R3N, R3Nþ-Cl, and organic contaminant product concentrations. In 1 this case, the kinetic rate constants, kFAC app and kobs Eq. ((2)) were was unknown but was known whereas the kchlorammonium app guessed in order to obtain the best model fit to the experimental data. Results indicated that for all organic contaminant/amine combinations that followed a pseudo-first order decay, the above model predictions significantly underestimated experimentally observed kinetics (an example for each amine provided in Supplementary information Fig. S1)
values increased up to diffusioneven when kchlorammonium app controlled rates (i.e. 1010 M1 s1). Further consideration suggested that perhaps the [R3Nþ-Cl] value determined in Eq. (4) was not properly estimated by Eq. (2) through the use of the rate constant k1app . Rather, Eq. (2) should be represented by both forward and reverse reaction rates of the deprotonated amine reacting with HOCl (Eq. (6)), as this is the dominant reaction pathway (Abia et al., 1998; Antelo et al., 1995):
d½R3 N d½HOCl ¼ ¼ k1app ½AmineT ½FACT dt dt ¼ kforward ½HOCl½R3 N þ kreverse R3 Nþ Cl OH
(6)
where [Amine]T ¼ [R3N] þ [R3Nþ]. Given a significant value for kreverse, this revised rate expression could then account for the fact that, as the organic contaminant begins to quench [R3NþCl] (Eq. (3)), the kinetics of Eq. (2) could be driven to the right as the kreverse$[R3Nþ-Cl]$[OH] value is minimized. To evaluate such a phenomenon, kforward was first determined experimentally by estimating and plotting the loss of free chlorine with time represented by kobs ( y-axis) in the presence of varying excess doses of amine [Amine]T (x-axis), as seen by Eq. (7) which was derived by rearranging Eq. (6) (Supplementary information Text S3). kforward could then be calculated from the slope of the line (a): kobs ¼
kforward Hþ Kamine ½AmineT Hþ þ Kamine Hþ þ Kw
! kforward Hþ Kamine kreverse Kw þ R3 Nþ Cl þ þ þ H þ Kamine H þ Kw H $½FACT
6094
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 8 7 e6 0 9 6
kforward Hþ Kamine and a ¼ þ H þ Kamine Hþ þ Kw
(7)
where Kamine and Kw are the acid dissociation constants for the amine and water, respectively. For both TMA and MES, kforward was determined to be equal to kapp1 when amine and free chlorine speciation effects were accounted for, such that: k1app ½AmineT ½FACT ¼ kforward ½HOCl½R3 N
(8)
and thus rendering kreverse to be negligible. Such evidence indicates that deviations present between the model predicted kinetics of enhanced organic contaminant loss in the presence of FAC and amine cannot in fact be directly attributed to the impact of kreverse but may be due to other factors that involve a more complex set of elementary reactions. Therefore, further research is needed to appropriately assess what these factors may be so that the observed kinetics can be more accurately predicted.
3.5.
Byproduct formation and distribution
The major byproducts for SA, FLU, and TMP were found to be identical whether amine was added ([MES or TMA] ¼ 10 mM for SA and FLU reactions, and 0.50 mM for TMP reactions) or not. The major byproducts for SA were 3-chlorosalicylic acid (3-Cl-SA, m/z 171), 5-chlorosalicylic acid (5-Cl-SA, m/z 171) and 3,5-dichlorosalicylic acid (3,5-diCl-SA, m/z 205). These chlorinated byproducts were identified by comparing retention times, mass/charge ratios, and mass spectral fragmentation patterns to authentic standards. The only byproduct for FLU was m/z 252 (FLU þ Cl COOH), and the major byproducts for
TMP were m/z 325 (TMP þ Cl), 359 (TMP þ 2Cl H), and 377 (TMP þ 2Cl þ OH) by MS analysis as well as comparison to similar byproducts reported in the literature (Dodd and Huang, 2007; Dodd et al., 2005). Currently though, the exact placement of the eCl and eOH substituents for these TMP byproducts remains tentative until authentic standards can be obtained. Structures of these byproducts are shown in Table S2 of the Supplementary information. Relative abundances of SA byproducts were measured as the reaction progressed following amine addition and represented in Fig. 5aec by their mole fraction, which was defined as the moles of product formed divided by the initial SA concentration. These values were then plotted against the reaction extent, represented as % SA loss. The addition of 10 mM TMA or MES did not affect formation of 3-Cl-SA (Fig. 5a) or 5-Cl-SA (Fig. 5b) and only a slight difference was observed for 3,5-diCl-SA (Fig. 5c), as its mole fraction increased from 0.04 with only FAC to 0.08 with 10 mM TMA while no difference was observed with MES. The relative abundance of TMP’s major byproducts (m/z 325, 359, and 377) were also presented in Fig. 5def in terms of the total MS ion count of each species (A) normalized by the total ion count of parent TMP at t ¼ 0 s (AMAX), assuming that all product and parent compounds exhibit comparable MS ionization potential on the basis of their similar structures. One major difference observed in TMP byproduct formation was the higher abundance of the dichlorinated byproducts (m/z 377 and 359) formed in the presence of TMA versus the higher abundance of the monochlorinated byproduct (m/z 325) formed in the presence of MES over the entire course of the reaction. This might have resulted since the presence of TMA has stronger oxidizing power than the presence of MES. In addition, Fig. 5 indicates that upon MES addition, m/z 377 formation (two
Product Mole Fraction
0.6
3-Cl SA
FAC only 10 M TMA 10 M MES
0.5 0.4
3,5-diCl SA
5-Cl SA
a
c
b
0.3 0.2 0.1 0.0 0.6
A/AMAX
m/z 377
FAC Only 0.5 M TMA 0.5 M MES
0.5 0.4
m/z 325
m/z 359
e
d
f
0.3 0.2 0.1 0.0 0
20
40
60
80
% loss of Organic Contaminant
100
0
20
40
60
80
% loss of Organic Contaminant
100 0
20
40
60
80
100
% loss of Organic Contaminant
Fig. 5 e Mole fraction of the SA byproducts (a) 3-Cl SA, (b) 5-Cl SA, and (c) 3,5-diCl SA or relative TMP byproduct abundance of (d) m/z 377, (e) m/z 359, and (f) m/z 325 that are formed as a function of % loss of SA or TMP with FAC only (403[org.cont.]i), FAC (103[org.cont.]i) D 10/0.5 mM TMA (SA/TMP reactions), or FAC (103[org.cont.]i) D 10/0.5 mM MES (SA/TMP reactions) (pH 7.0e7.3, 23 C).
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 8 7 e6 0 9 6
Cl addition) is slightly less favorable than one Cl addition (m/z 325) even in comparison to FAC alone when % TMP conversion is w38e55%. This result is possibly related to steric hindrance effects as the MES-related oxidant is likely bulkier than FAC and may not be favorably positioned to add an additional Cl to TMP as HOCl would.
4.
Conclusions
In wastewaters and drinking waters, the role of tertiary amines in serving not only as compounds reactive to free chlorine but also as precursors of strong oxidant formation has largely been overlooked in previous literature. The results of this study suggest that tertiary amines can enhance transformation of organic contaminants such as SA, FLU, and TMP by free chlorine via formation of chlorammonium species. Rate enhancement occurred by up to three orders of magnitude when the tertiary amine dose was five times the organic contaminant concentration, but also occurred even at very low tertiary amine concentrations where TMA and MES increased the rate of contaminant loss by up to 600% at levels as low as 1% compared to the organic contaminant and 0.1% compared to the free chlorine. The catalytic capability of tertiary amines to enhance organic contaminant decomposition during chlorination was further found to be highly dependent on the basicity and structure of tertiary amines as TMAþ-Cl appears to have stronger oxidizing potential than MESþ-Cl, while NTA and CRE exhibit no effect. Further numerical modeling of the reaction kinetics when both TMA and MES were added suggested that the pathways in Eqs. (2) and (3) could not simply represent the fast enhancement kinetics that were experimentally observed but that a more detailed group of elementary reactions may be involved. However, reactions of organic contaminants with these chlorammonium species did generate similar byproducts as those from free chlorine reactions with only slight differences in byproduct distribution. Thus, the mechanisms for organic contaminant chlorination in the presence of TMA or MES may be closely related to the mechanisms with FAC. Overall, these findings serve as a platform to further assess tertiary amines’ overall impact on organic contaminant transformation in real water matrices during the chlorination process in water and wastewater treatment.
Acknowledgments A.D.S. thanks partial financial support from Saehan Industries, Inc. in Korea, Georgia Water Resources Institute, and NSF IGERT Fellowship. The authors thank Michael C. Dodd for his initial work on this novel topic as well as offering helpful comments while drafting the paper.
Appendix. Supplementary information Supplementary information associated with this article can be found, in the online version, at doi:10.1016/j.watres.2011.09.010.
6095
references
Abia, L., Armesto, X.L., Canle, M., Garcia, M.V., Santaballa, J.A., 1998. Oxidation of aliphatic amines by aqueous chlorine. Tetrahedron 54, 521e530. Antelo, J.M., Arce, F., Armesto, J.L., Garciaverdugo, A., Penedo, F.J., Varela, A., 1985. A kinetic-study of the chlorination of tertiary alcoholamines in alkaline media. International Journal of Chemical Kinetics 17, 1231e1245. Antelo, J.M., Arce, F., Parajo, M., 1995. Kinetic-study of the formation of N-chloramines. International Journal of Chemical Kinetics 27, 637e647. Antelo, J.M., Arce, F., Perezmoure, J.C., 1992. Kinetics of the Nchlorination of 2-aminobutyric, 3-aminobutyric, 3aminoisobutyric and 4-aminobutyric and 4-aminobutyric acids in aqueous-solution. International Journal of Chemical Kinetics 24, 1093e1101. Barbosa, J., Barron, D., Cano, J., Jimenez-Lozano, E., SanzNebot, V., Toro, I., 2001. Evaluation of electrophoretic method versus chromatographic, potentiometric and absorptiometric methodologies for determing pKa values of quinolones in hydroorganic mixtures. Journal of Pharmaceutical and Biomedical Analysis 24, 1087e1098. 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, 13e51. Dodd, M.C., Huang, C.H., 2007. Aqueous chlorination of the antibacterial agent trimethoprim: reaction kinetics and pathways. Water Research 41, 647e655. Dodd, M.C., Shah, A.D., Von Gunten, U., Huang, C.-H., 2005. Interactions of fluoroquinolone antibacterial agents with aqueous chlorine: reaction kinetics, mechanisms, and transformation pathways. Environmental Science and Technology 39, 7065e7076. Ellis, A.J., Soper, F.G., 1954. Studies of N-Halogeno-Compounds .6. The kinetics of chlorination of tertiary amines. Journal of the Chemical Society, 1750e1755. Geiser, L., Henchoz, Y., Galland, A., Carrupt, P.A., Veuthey, J.L., 2005. Determination of pKa values by capillary zone electrophoresis with a dynamic coating procedure. Journal of Separation Science 28, 2374e2380. Good, N.E., Winget, G.D., Winter, W., Connolly, T.N., Izawa, S., Singh, R.M.M., 1966. Hydrogen ion buffers for biological research. Biochemistry 5, 467e477. Grzybowski, A.K., Datta, S.P., 1964. Ionisation constant of protonated form creatinine. Journal of the Chemical Society, 187e196. Hwang, Y., Matsuo, T., Hanaki, K., Suzuki, N., 1995. Identification and quantification of sulfur and nitrogen containing odorous compounds in wastewater. Water Research 29, 711e718. Jones, M., 1997. Organic Chemistry. W.W. Norton & Company, New York. Kemper, J.M., Walse, S.S., Mitch, W.A., 2010. Quaternary amines as Nitrosamine precursors: a role for Consumer products? Environmental Science and Technology 44, 1224e1231. Kohut, K.D., Andrews, S.A., 2003. Polyelectrolytes as amine precursors for N-nitrodimethylamine in drinking water. Proceedings of Annual Conference of American Water Works Association, 405e418. Kolpin, D.W., Furlong, E.T., Meyer, M.T., Thurman, E.M., Zaugg, S.D., Barber, L.B., Buxton, H.T., 2002. Pharmaceuticals, hormones, and other organic wastewater contaminants in US streams, 1999e2000: a national reconnaissance. Environmental Science and Technology 36, 1202e1211. Kotsyubynskyy, D., Molchanov, S., Gryff-Keller, A., 2004. Creatinine and creatininium cation in water solution.
6096
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 8 7 e6 0 9 6
Tautomerism and quantitative interpretation of the solution acidity effect on H-1, C-13 and N-14 NMR chemical shifts. Polish Journal of Chemistry 78, 239e248. Larson, R.A., Rockwell, A.L., 1979. Chloroform and Chlorophenol Production by Decarboxylation of natural acids during aqueous chlorination. Environmental Science and Technology 13, 325e329. Margerum, D.W., Gray, E.T., Huffman, R.P., 1978. Chlorination and the formation of N-chloro-compounds in water treatment. In: Brinckman, F.E., Bellama, J.M. (Eds.), Organometals and Organometalloids: Occurrence and Fate in the Environment. American Chemical Society Books, Washington, D.C, pp. 279e291. Martell, A.E., Smith, R.M., 1974. Critical Stability Constants. Plenum Press, New York. McArdell, C.S., Molnar, E., Suter, M.J.F., Giger, W., 2003. Occurrence and fate of macrolide antibiotics in wastewater treatment plants and in the Glatt Valley Watershed, Switzerland. Environmental Science and Technology 37, 5479e5486. Mederos, A., Dominguez, S., Medina, A.M., Brito, F., Chinea, E., Bazdikian, K., 1987. Equilibria in aqueous-solution between Be(II) and nitrotriacetic, methyl-c-nitrilotriacetic, nitrilodiaceticpropionic, nitriloaceticdipropionic and nitrilotripropionic acids. Polyhedron 6, 1365e1373. Miao, X.S., Bishay, F., Chen, M., Metcalfe, C.D., 2004. Occurrence of antimicrobials in the final effluents of wastewater treatment plants in Canada. Environmental Science and Technology 38, 3533e3541. Mitch, W.A., Schreiber, I.M., 2008. Degradation of tertiary alkylamines during chlorination/chloramination: implications for formation of aldehydes, nitriles, halonitroalkanes, and nitrosamines. Environmental Science and Technology 42, 4811e4817. Morris, J.C., 1967. Kinetics of reactions between aqueous chlorine and nitrogen compounds. In: Faust, S.D., Hunter, J.V. (Eds.), Principles and Applications of Water Chemistry. John Wiley & Sons, New York, pp. 23e53. Morris, J.C., Isaac, R.A., 1983. A critical review of kinetic and thermodynamic constants for aqueous chlorine-ammonia system. In: Jolleys, R.L., Brungs, W.A., Cotruvo, J.A., Cumming, R.B., Mattice, J.S., Jacobs, V.A. (Eds.), Water
Chlorination: Environmental Impact and Health Effects. Ann Arbor Science Publishers, Ann Arbor, Michigan, pp. 49e62. Murray, R.K., Granner, D.K., Mayes, P.A., Rodwell, V.W., 2000. Harper’s Biochemistry, 25th ed. Appleton & Lange, Stamford, CT. Park, S.-H., Wei, S., Mizaikoff, B., Taylor, A.E., Favero, C.d., Huang, C.-H., 2009. Degradation of amine-based water treatment Polymers during chloramination as N-Nitrosodimethylamine (NDMA) precursors. Environmental Science and Technology 43, 1360e1366. Peskin, A.V., Winterbourn, C.C., 2001. Kinetics of the reactions of hypochlorous acid and amino acid chloramines with thiols, methionine, and ascorbate. Free Radical Biology and Medicine 30, 572e579. Prutz, W.A., 1998. Reactions of hypochlorous acid with biological substrates are activated catalytically by tertiary amines. Archives of Biochemistry and Biophysics 357, 265e273. Prutz, W.A., Kissner, R., Koppenol, W.H., 2001. Oxidation of NADH by chloramines and chloramides and its activation by iodide and by tertiary amines. Archives of Biochemistry and Biophysics 393, 297e307. Qiang, Z.M., Adams, C., 2004a. Potentiometric determination of acid dissociation constants (pK(a)) for human and veterinary antibiotics. Water Research 38, 2874e2890. Qiang, Z.M., Adams, C.D., 2004b. Determination of monochloramine formation rate constants with stopped-flow spectrophotometry. Environmental Science and Technology 38, 1435e1444. Roth, B., Strelitz, J.Z., 1969. Protonation of 2,4Diaminopyrimidines I. Dissociation constants and substituent effects. Journal of Organic Chemistry 34, 821e836. Sanchiz, J., Esparza, P., Dominguez, S., Brito, F., Mederos, A., 1999. Solution studies of complexes of iron(III) with iminodiacetic, alkyl-substituted iminodiacetic and nitrilotriacetic acids by potentiometry and cyclic voltammetry. Inorganica Chimica Acta 291, 158e165. Sletten, D.M., Nickander, K.K., Low, P.A., 2005. Stability of acetylcholine chloride solution in autonomic testing. Journal of Neurological Sciences 234, 1e3. Walsh, C., 2003. Antibiotics: Actions, Origins, Resistance. ASM Press, Washington, D.C.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 9 7 e6 1 0 6
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Photodegradation of psychiatric pharmaceuticals in aquatic environments e Kinetics and photodegradation products Vaˆnia Calisto a, M. Rosa´rio M. Domingues b, Valdemar I. Esteves a,* a
Department of Chemistry and CESAM (Centre for Environmental and Marine Studies), University of Aveiro, Campus de Santiago, 3810-193 Aveiro, Portugal b Department of Chemistry, Mass Spectrometry Centre, University of Aveiro, Campus de Santiago, 3810-193 Aveiro, Portugal
article info
abstract
Article history:
Benzodiazepines are widely consumed psychiatric pharmaceuticals which are frequently
Received 8 April 2011
detected in the environment. The environmental persistence and fate of these pharma-
Received in revised form
ceuticals as well as their degradation products is of high relevance and it is, yet, scarcely
30 August 2011
elucidated. In this study, the relevance of photodegradation processes on the environ-
Accepted 3 September 2011
mental persistence of four benzodiazepines (oxazepam, diazepam, lorazepam and alpra-
Available online 10 September 2011
zolam) was investigated. Benzodiazepines were irradiated under simulated solar irradiation and direct and indirect (together with three different fractions of humic
Keywords:
substances) photodegradation kinetics were determined. Lorazepam was shown to be
Benzodiazepines
quickly photodegradated by direct solar radiation, with a half-life time lower than one
Diazepam
summer sunny day. On the contrary, oxazepam, diazepam and alprazolam showed to be
Oxazepam
highly resistant to photodegradation with half-life times of 4, 7 and 228 summer sunny
Lorazepam
days, respectively. Apparent indirect and direct photodegradation rates are of the same
Alprazolam
order of magnitude. However, humic acids were consistently responsible for a decrease in
Dissolved organic matter
the photodegradation rates while fulvic acids and XAD4 fraction caused an enhancement of the photodegradation. Overall, the results highlight that photodegradation might not be an efficient pathway to prevent the aquatic environmental accumulation of oxazepam, diazepam and alprazolam. Also, nineteen direct photodegradation products were identified by electrospray mass spectrometry, the majority of which are newly identified photoproducts. This identification is crucial to a more complete understanding of the environmental impact of benzodiazepines in aquatic systems. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
Pharmaceuticals are a vast and diverse group of chemicals which are consumed worldwide in very high quantities (Calisto and Esteves, 2009; Petrovic and Barcelo´, 2007). These compounds are being introduced into the environment on a continuous basis, mainly through sewage treatment plants (STPs), as a result of the inadequacy of the treatment processes applied in these facilities (Gros et al., 2010; Jelic
et al., 2011). Due to the growing number of studies focused on the occurrence of pharmaceuticals and their persistence in the environment, as well as possible toxicological effects on non-target organisms, pharmaceuticals are now unanimously considered as an important group of environmental contaminants. Benzodiazepines (BDZ) belong to the group of psychiatric substances which act on the central nervous system, having anxiolytic, sedative and hypnotic effects (Kar, 2007) and is one
* Corresponding author. Tel.: þ351 234 401408. E-mail address: [email protected] (V.I. Esteves). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.09.008
6098
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 9 7 e6 1 0 6
of the most prescribed groups of pharmaceuticals throughout the world (van der Ven et al., 2004). The most recent data account for 2008, when a total of 30 billion S-DDD (defined daily doses for statistical purposes) of BDZ were manufactured, the highest amount registered until now. Alprazolam (ALP), diazepam (DZP), lorazepam (LRZ) and oxazepam (OXZ) are the most relevant substances of this group, representing 46, 24, 13 and 2% of the total manufactured amount (International Narcotics Control Board, 2010), respectively. Over the last decade, there has been a significant number of studies reporting the occurrence of BDZ in environmental matrices, namely in STPs influents and effluents (Heberer, 2002; Huerta-Fontela et al., 2010), surface waters (HuertaFontela et al., 2010; Calamari et al., 2003; Gonza´lez Alonso et al., 2010) and drinking waters (Jones et al., 2005). DZP is the most commonly found BDZ with maximum concentrations ranging from 0.88 mg L1 in surface waters, Germany (Ternes et al., 2001) to 1.18 mg L1 in a sewage treatment plant, Belgium (van der Ven et al., 2004). The presence of these compounds in the ecosystems might have a significant effect on non-target organisms. Apart from their intrinsic biological activity, BDZ (and psychiatric pharmaceuticals in general) were produced with the aim of acting on the central nervous system, directly affecting the regulation of behavior and reproduction patterns (Calisto and Esteves, 2009). However, to correctly evaluate the real ecological impact of these pollutants, it is important to take into consideration their fate and persistence in aquatic environments (Boreen et al., 2003; Glassmeyer et al., 2008). In the particular case of BDZ, and when compared to other groups of pharmaceuticals, there is a significant lack of knowledge in this field. Photodegradation is one of the main factors affecting the environmental persistence of pollutants, especially in surface waters and STPs exposed to sunlight (Boreen et al., 2003; Doll and Frimmel, 2004). Organic pollutants can undergo direct photolysis by absorbing photons which are then able to induce a chemical transformation, or indirect photolysis, when the phototransformation is indirectly caused by excitation of chromophores present in natural waters (Boule et al., 2005). Indirect photolysis can play an important role on the degradation of pollutants that poorly absorb solar radiation or that resist to direct photolysis (Boule et al., 2005; Zafiriou et al., 1984). Natural dissolved organic matter (DOM), whose main fraction is referred to as humic substances (HS), is ubiquitously present in aquatic environments and is able to initiate a significant number of photochemical reactions (Boule et al., 2005). After absorbing light, DOM can produce a high number of reactive intermediates such as reactive oxygen species (singlet oxygen, superoxide, hydrogen peroxide, hydroxyl radicals), DOM excited triplet states, carbon centered radicals and hydrated electrons that are then responsible for promoting indirect photodegradation (Zafiriou et al., 1984; Zepp et al., 1985; Dalrymple et al., 2010; Guerard et al., 2009). However, the effect of DOM on the photodegradation is not trivial: they can also act as inhibitors by scavenging the reactive intermediates mentioned and by screening the radiation (Boule et al., 2005; Xia et al., 2009). In this work, the photodegradation processes of four BDZ (OXZ, DZP, LRZ and ALP) were studied. Direct and indirect photodegradation (in the presence of different fractions of HS:
humic acids (HA), fulvic acids (FA) and XAD4 fraction) were evaluated under simulated solar irradiation. Also, direct photodegradation products were identified by mass spectrometry. The available literature studies about the photodegradation products of BDZ are not environmentally relevant, focusing on drug-development, storage and handling of these pharmaceuticals (Cabrera et al., 2005; Castaneda et al., 2009). This manuscript aims to present the first data regarding photodegradation kinetics and photodegradation products of BDZ, under environmentally relevant conditions.
2.
Experimental section
2.1.
Chemicals
All chemicals used were of analytical grade: ALP, LRZ, DZP e OXZ (SigmaeAldrich), sodium dodecylsulphate (SDS, 99%, for electrophoresis, SigmaeAldrich), hexadimethrine bromide (polybrene, SigmaeAldrich), sodium chloride, ethylvanillin (99%, Aldrich), sodium tetraborate (Riedel-de Hae¨n), sodium hydroxide (Fluka), and acetonitrile (HPLC gradient grade, VWR, Prolabo). All solutions were prepared using ultra-pure water, obtained from a Milli-Q Millipore system (Milli-Q plus 185).
2.2.
Irradiation experiments
2.2.1.
Irradiation Apparatus
Samples were irradiated under simulated solar radiation using Solarbox 1500 (Co.fo.me.gra, Italy). The irradiation device was equipped with a 1500 W arc xenon lamp and outdoor UV filters that restrict the transmission of light with wavelengths below 290 nm. The irradiance of the lamp was kept constant during all the experiments at 55 W m2 (290e400 nm). Lamp spectrum, at this irradiance level, is presented in the Supporting Information (SI). To monitor the irradiance level and temperature, a multimeter (Co.fo.me.gra, Italy), equipped with a UV 290e400 nm large band sensor and a black standard temperature sensor, was used. The device was refrigerated by an air cooled system and the uniformity of the irradiation inside the chamber was guaranteed by a parabolic reflection system.
2.2.2.
Sample preparation and sampling
Irradiated samples consisted on BDZ aqueous solutions with a concentration of 10 mg L1. Due to the low solubility of the studied compounds in water, acetonitrile was used as an auxiliary solubilising agent. The concentration of acetonitrile did not exceed 1% (v/v) of the irradiated solution, in order to guarantee that acetonitrile does not influence the photodegradation rates, as recommended by the OCDE guideline TG316 (OECD, 2008). The pH of the irradiated solutions was not adjusted and samples were not buffered in order to avoid influences of the buffering agent in the photodegradation process. Solutions were irradiated in triplicate using 25 mL quartz tubes with an internal diameter of 1.5 cm. For each set of experiments, dark controls were also irradiated, in triplicate; for this purpose, quartz tubes were covered with several layers of aluminum foils. A home-made
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 9 7 e6 1 0 6
metallic holder was used to maintain the quartz tubes suspended inside the irradiation chamber, allowing for homogeneous irradiation. The extent of the irradiation experiments depended on the phototransformation rates of each pharmaceutical, varying between 12 and 304 h. 2 mL aliquots of the irradiated samples (and dark controls) were collected at specific time intervals, stored at 4 C and subsequently analyzed by micellar electrokinetic chromatography (MEKC), within 2 days, as described below.
2.2.3. Influence of dissolved organic matter on photodegradation To study the influence of different fractions of DOM on the photodegradation rates of BDZ, solutions containing 1 mg L1 of HA, FA or XAD4 fraction were irradiated in the conditions described above. For this purpose, FA and XAD4 solutions (20 mg L1) were prepared in ultra pure water. In the case of HA, due to the low solubility in water, an aqueous solution (50 mg L1) 5% (v/v) NH4OH 1 M was prepared and set to pH 6.0 with formic acid 1 M. The stock solutions were further diluted (20 and 50 fold) in order to attain a final concentration of 1 mg L1, as referred above. Similarly to direct photodegradation studies, irradiated solutions were not buffered and the pH was not adjusted. The HS used in this study were extracted and isolated from a riverine water sample, collected in a freshwater stream that flows into the Aveiro lagoon, Portugal (40 39’N, 8 44’W). HS of this riverine aquatic system are mainly derived from the decomposition of herbaceous plants. The extraction and isolation of the different fractions of HS were performed by using a system of XAD resins (XAD8 and XAD4) connected in series. This procedure is described in detail in Santos et al. (1994) and Esteves et al. (1995). The purified fractions (HA, FA and XAD4) were subsequently characterized by elemental analysis and solid-state 13C-CPMAS NMR, using the procedure described in Esteves et al. (2009).
2.3.
Capillary electrophoresis
The kinetics of photodegradation was followed by MEKC. The analyses were carried out using a commercial instrument Beckman P/ACE MDQ (Fullerton, CA, USA), equipped with a photodiode array UVeVis detector. The MEKC methodology was adapted from a previous work, where it was developed to follow the kinetics and emergence of photodegradation products of carbamazepine (Calisto et al., 2011). This method is based on the use of a dynamically coated capillary column with the aim of improving reproducibility and separation efficiency (Calisto et al., 2011; Erny et al., 2009). Detailed information about the MEKC methodology is presented in the SI.
2.4. Identification of photodegradation products by ESI(þ)MS and MS2 The direct photodegradation products of BDZ were identified by electrospray mass spectrometry (ESI-MS). Irradiated samples of each pharmaceutical were collected after 6, 24, 24 and 304 h of irradiation for LRZ, OXZ, DZP and ALP,
6099
respectively. The irradiation time was selected according to the half-life time of each compound and considering the time at which a larger number of photodegradation products could be clearly identified in the electrochromatogram. Positive-ion ESI-MS and ESI-MS2 were performed on a Micromass Q-TOF2 hybrid tandem mass spectrometer (Manchester, UK). For ESI analysis, samples were prepared by diluting the irradiated solutions in methanol (0.1% formic acid v/v). Samples were introduced into the ESI source using a syringe pump at a flow rate of 10 mL min1. The time-of-flight (TOF) mass resolution was set to approximately 9000, the cone voltage was 35 V and the capillary voltage was 3 kV. The source and the desolvation temperature were 80 and 150 C, respectively. The MS2 spectra were acquired using argon as the collision gas and collision energy was set between 10 and 32 eV. To calibrate the MS spectra, the lock mass was the calculated monoisotopic mass/charge of the ion [M þ H]þ of each pharmaceutical. The data was processed using MassLynx software, version 4.0.
3.
Results and discussion
3.1.
Performance of the MEKC methodology
The adopted MEKC method proved to be efficient in the separation of the each BDZ from its photodegradation products. The use of a dynamically coated capillary column highly improved the repeatability of both migration times and peak areas since the coating avoids fluctuations due to the modification of the chemical structure of the inner surface of the bare silica capillary between successive runs. The capillary coating was verified to be stable during several weeks. Also, the addition of sodium tetraborate to the samples contributed to a significant tightening of the peaks, thus increasing peak resolution. Under the optimized conditions, the migration times (from 10 repeated injections) of studied BDZ are presented in Table 1. Relative standard deviations did not exceed 0.55%, showing a good repeatability of successive runs. Moreover, relative standard deviations of the ratio between the analyte and the IS peak areas were always below 3% (results not shown). A linear calibration curve was obtained for each benzodiazepine using seven standard solutions with concentrations ranging from 0.5 to 10 mg L1, analyzed in quadruplicate. The results were fitted to a least-squares linear regression by plotting the ratio between the peak area of the analyte and the peak area of the IS as a function of the analyte concentration. The statistical parameters of the obtained calibration curves are presented in Table 1. Correlation coefficients range from 0.9997 to 0.9999, corroborating the excellent linear response in the studied range of concentrations. The limits of detection (LOD) and limits of quantification (LOQ) were also determined, defined as 3sx/y/b and 10sx/y/b, respectively. LOQs between 0.598 mg L1 for ALP and 0.318 mg L1 for DZP allow following the kinetics of photodegradation up to, approximately, 95% of degradation of BDZ, confirming the adequacy of the methodology to this purpose.
6100
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 9 7 e6 1 0 6
Table 1 e Linear regression equation, correlation coefficient (r), limit of detection (LOD) and limit of quantification (LOQ) of the MEKC methodology for OXZ, DZP, LRZ and ALP. Migration time (±standard deviation) of each pharmaceutical is also presented. Linear regression equation OXZ DZP LRZ ALP
y¼ y¼ y¼ y¼
(0.861 (0.917 (0.886 (0.830
0.002) 0.003) 0.004) 0.004)
x x x x
0.005 0.007 þ 0.003 0.007 þ 0.00 0.01 0.01 0.01
r
LOD/(mg L1)
LOQ/(mg L1)
mt/(min), n ¼ 10
0.9999 0.9999 0.9998 0.9997
0.103 0.0953 0.154 0.179
0.343 0.318 0.512 0.598
10.94 0.06 11.12 0.02 10.74 0.02 11.34 0.03
3.2. Photodegradation of OXZ, DZP, LRZ and ALP in absence and in the presence of DOM Direct and indirect photodegradation of OXZ, DZP, LRZ and ALP were individually evaluated in aqueous solutions of 10 mg L1, under simulated solar irradiation. Indirect photodegradation was studied in the presence of HA, FA and XAD4 fraction of HS from a freshwater sample. Dark controls, obtained under the exact same conditions, were performed for all the irradiation experiments. Degradation in the dark, due to thermal or hydrolytic processes, was not observed for OXZ, LRZ and ALP; on the other hand, DZP suffered a decrease in concentration up to 10% during 24 h. The observed degradation of DZP in the dark might be attributed to the temperature rise inside the irradiation chamber (35e40 C) as the same phenomenon was not observed in refrigerated or room temperature solutions. DZP photodegradation data were corrected in order to account for non-photolytic processes occurring in samples. The determination of kinetic parameters was performed by fitting a pseudo-first order kinetic model to each set of results. Accordingly, the natural logarithm of the ratio between the BDZ concentration, at a given irradiation time, and its initial concentration was plotted as a function of the irradiation time and a linear regression was obtained. Correlation coefficients (r), apparent pseudo-first order rate constants (k) and half-life
times (t1/2) are presented in Table 2. For both direct and indirect photodegradation, r ranges from 0.961 to 0.999, confirming the adequacy of the pseudo-first order model to describe the kinetics of BDZ photodegradation. Direct photodegradation t1/2 of the studied BDZ greatly differ. LRZ was the least photoresistant, with t1/2 of 2.6 0.1 h; OXZ and DZP had intermediate t1/2 of 15.1 0.2 and 28 2 h, respectively. Note that these compounds are structurally very similar (especially LRZ and OXZ) and, yet, exhibit a distinct resistance to photodegradation. ALP showed to be highly resistant to direct photodegradation with a t1/2 of 865 41 h. However, we must emphasize that this value was determined by an extrapolation based on results of a 204 h irradiation, due to the extremely high number of irradiation hours needed (more than 1000 h). Notwithstanding, and in our understanding, the conducted experiment is clearly conclusive about the behavior of ALP when exposed to sunlight. In order to determine the t1/2 of LRZ, DZP and OXP they were irradiated until, at least, 75% of degradation was achieved. In what concerns indirect photodegradation, the t1/2 obtained in the presence of different fractions of HS are of the same order of magnitude of those in pure water. For ALP, t1/2 were not determined seeing that no photodegradation was observed after a 48 h irradiation. Considering the conclusions taken concerning ALP direct photodegradation, irradiations under indirect degradation conditions were not extended for
Table 2 e Kinetic parameters for the photodegradation of OXZ, DZP, LRZ and ALP under different experimental conditions: correlation coefficient (r), half-life time (t1/2) and apparent photodegradation rate (k) for a pseudo-first order kinetics. n represents the number of experimental points used in the fitting and s represents the standard deviation. Half-life time in units equivalent to summer sunny days (SSD) and photolysis average quantum yields (4ave) are also shown.
OXZ
DZP
LRZ
ALP
DOM
r
n
k s/h1
t1/2 s/h
t1/2 s/SSD
4ave
e FA XAD4 HA e FA XAD4 HA e FA XAD4 HA e FA XAD4 HA
0.999 0.989 0.995 0.998 0.961 0.984 0.974 0.982 0.983 0.996 0.999 0.991 0.980 n.d. n.d. n.d.
24 21 21 24 21 18 18 21 18 21 21 18 20 n.d. n.d. n.d.
0.0458 0.0005 0.056 0.002 0.053 0.001 0.0380 0.0004 0.025 0.002 0.044 0.002 0.035 0.002 0.0154 0.0007 0.27 0.01 0.302 0.007 0.330 0.004 0.113 0.004 0.00080 0.00004 n.d. n.d. n.d.
15.1 0.2 12.4 0.4 13.2 0.3 18.2 0.2 28 2 15.6 0.7 20 1 45 2 2.6 0.1 2.29 0.05 2.10 0.02 6.2 0.2 865 41 n.d. n.d. n.d.
3.98 0.04 3.2 0.1 3.47 0.08 4.80 0.06 7.3 0.5 4.1 0.2 5.2 0.3 11.8 0.5 0.68 0.03 0.60 0.01 0.552 0.006 1.62 0.06 228 11 n.d. n.d. n.d.
4.45 106 e e e 4.3 106 e e e 7.8 105 e e e 3.4 106 e e e
6101
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 9 7 e6 1 0 6
a longer period. However, for OXZ, DZP and LRZ, the addition of HS caused a clear and consistent influence on the photodegradation rates: HA (the most hydrophobic fraction) decreased the degradation rates while FA and XAD4 fraction (the most hydrophilic fractions) enhanced the photo-induced transformations. These results suggest the existence of a correlation between the hydrophobicity of the HS and the effect on photodegradation rates of BDZ. As it was stated before, the effects of HS on photodegradation processes are of extreme complexity as they can exert, simultaneously, enhancing and reducing influences (production and/or scavenging of reactive intermediates and screening of reactive light wavelengths). In addition, the function of HS on photochemical processes is highly dependent on its concentration and chemical composition; however, a direct correlation between chemical composition and photochemical function is very often difficult to establish (Boule et al., 2005; Guerard et al., 2009; Garbin et al., 2007). The structural characterization of the used HS (13C-CPMAS NMR and Elemental Analysis data displayed in Table 3), revealed that differences in the 13CCPMAS NMR spectra of HS might suggest a higher prevalence of aromatic moieties in HA than in FA or XAD4 (spectra range between 108e145 and 145e160 ppm). Moreover, FA and XAD4 fractions seem to have a higher number of carboxylic and ester groups (spectra fraction between 160 and 190 ppm) and also a higher carbon/nitrogen ratio. The influence of these compositional differences on the distinct effect of HA and FA/ XAD4 on the photodegradation of BDZ cannot be clearly defined without further research. Also, it must be noted that apart from several evidences of an apparent correlation between the 13C NMR signal strength and the abundance of the respective chemical groups, these data should always be cautiously interpreted, especially when looking for small differences (Newman, 2007) since CPMAS experiments are known to be not quantitative. As it was referred in the experimental section, acetonitrile was used as an auxiliary solubilising agent with a concentration of 1% (v/v). This co-solvent, as well as its maximum concentration, was chosen according to the recommendations of the OECD guideline (OECD, 2008). However, it should be taken into account that possible degradation processes induced by the photogeneration of hydroxyl radicals might be masked in the presence of acetonitrile. This effect is not
expected to significantly affect the observed photodegradation rates due to the low concentration of solvent but it is, still, a source of uncertainty. In order to evaluate the influence of light absorption by the different DOM fractions in the photodegradation rates, a wavelength specific screening factor (Sl) was determined. A good estimate of the effect of light attenuation in the experiments can be obtained by calculating Sl for the wavelength corresponding to the maximum rate of light absorption by each of the studied BDZ (Schwarzenbach et al., 2003). For this purpose, the rate of light absorption of DZP, OXZ and LRZ, for each wavelength, (Iabs li ), was determined according to the following equation: 0 εli lC0 Iabs li ¼ Ili 1 10
(1)
where, I0li is the lamp emission intensity at the wavelength li (Ein L1 s1), ε is the molar absorptivity of each BDZ at li (L mol1 cm1), l is the path length inside the photoreactor (cm - diameter of the cylindrical quartz tubes, 1.5 cm) and C0 is the initial concentration of pharmaceutical in the irradiated solutions (mol L1). The selection of the wavelength corresponding to the maximum rate of light absorption, for each BDZ, was performed by considering a wavelength range in which the studied pharmaceuticals absorb significantly (290e350 nm). These wavelengths were found to be 323, 323 and 327 nm for DZP, OXZ and LRZ, respectively. Subsequently, S323 and S327 were determined, for each DOM fraction, by means of the following equation (Guerard et al., 2009; Schwarzenbach et al., 2003):
Sl ¼
1 10al l
(2)
2:303 al l
where al (cm1) is the wavelength specific attenuation coefficient (considered as the absorbance of each DOM fraction, for the selected wavelength, when the path length is 1 cm e spectra shown in Fig. 1) and l (cm) is the path length of the irradiated quartz tubes (1.5 cm). Accordingly, the values of S323 and S327 are in the range of 0.973, 0.986e0.987 and 0.990e0.991 due to the presence of HA, FA and XAD4 fraction, respectively (note that S equals 1 when no attenuation of light occurs). The determined screening factors allowed to conclude that attenuation effects due to the absorption of
Table 3 e Elemental analysisa and solid-state 13C-CPMAS NMR datab of HA, FA and XAD4 fraction. 13
C-CPMAS NMR spectra fraction (%) in the specified chemical shift range (ppm)
0e60 60e90 90e108 108e145 145e160 160e190 190e220 (alkyl and (O-alkyl (Anomeric (Aromatic (O-substituted (Carboxylic (Carbonyl methoxyl carbons) carbons; carbons) aromatic and ester carbons; carbons) carbon carbons) carbons) ketones hydrates) and quinones) HA FA XAD4
51.3 61.1 55.9
14.4 14.0 21.5
3.2 2.4 3.6
18.6 10.3 7.4
3.4 1.7 1.4
7.3 8.4 9.1
1.8 2.1 1.1
Elemental analysis (%) C
H
N
S
O
C:N
51.4 54.0 49.2
4.3 4.8 4.4
4.2 1.9 2.9
2.1 1.5 1.5
32.1 35.2 40.0
12.2 28.4 17.0
a Elemental analysis results are corrected for humidity at 60 C and ashes at 750 C. b NMR data represents the area (in percentage) of the 13C-CPMAS NMR spectra fraction due to the carbons in the specified chemical shift range (ppm). Functional groups whose carbons resonate in the specified range are given in parenthesis.
6102
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 9 7 e6 1 0 6
reactive light by DOM (1 mg L1) are of minor significance to justify the observed effects of DOM in the photodegradation rates. The higher incidence of aromatic moieties in HA, in comparison to FA and XAD4 fraction, which results in a higher absorption of UV reactive light (Fig. 1), is not sufficient to explain the decrease in the photodegradation rates observed in the presence of HA. In general, the distinct effects caused by the different fractions of HS (both enhancing and inhibitory) might be exclusively related to the production and/or scavenging of reactive intermediates differently favored by each fraction. The determined screening factors allow drawing some conclusions about the effect of DOM in the photodegradation of BDZ; nevertheless, note that these results are valid only for the upper layer of water bodies exposed to solar light. The attenuation of light is much more relevant when considering higher depths (up to a few meters). This fact highlights that the half-lives of BDZ in real water bodies are expectedly much higher than the presented values, leading to the conclusion that these pharmaceuticals persist in aqueous environments for even longer periods of time. Overall, as it was stated before, t1/2 obtained under indirect photodegradation conditions were comparable to those of direct photodegradation, showing that the presence of naturally occurring DOM is not responsible for a drastic change on the environmental t1/2 of these compounds.
3.3. Determination of environmentally relevant parameters Half-life times and apparent pseudo-first order rate constants (Table 2) were determined under simulated solar irradiation and are, consequently, strictly related to the specific experimental conditions adopted (results are mainly dependent on the total irradiance of the lamp, 55 W m2 (290e400 nm)). Assuming that the adopted lamp properly simulates sunlight, results can be converted into outdoor half-life times, in units
Fig. 1 e UVeVis spectra of the different dissolved organic matter fractions used in indirect photodegradation studies. The spectra were obtained with a UVeVis Shimadzu spectrophotometer and a cell with an optical path length of 1 cm was used.
equivalents to summer sunny days (SSD). Considering that the total energy reaching the ground on a cloudless summer day (45 N latitude) is 7.5 105 J m2 (Vione et al., 2006; Minero et al., 2007), one summer sunny day (a 24 h day/night cycle) correspond to 3.8 h of irradiation. This conversion allows the determination of the t1/2 of BDZ in environmental conditions. Results converted to SSD are presented in Table 2. Under direct photodegradation conditions, t1/2 range between 0.68 0.03 and 228 11 SSD for LRZ and ALP, respectively. While LRZ might not be considered a persistent compound in the environment, it is expected that the elimination of ALP from surface waters by photodegradation processes would take several months. Moreover, DZP and OXZ exhibit t1/2 of several days (7.3 0.5 and 3.98 0.04 SSD, respectively), which is also considered environmentally relevant, especially if we take into consideration the continuous introduction of these compounds in the environment by STPs. Note that DZP has a higher environmental t1/2 than carbamazepine (4.5 SSD) (Calisto et al., 2011), a widely studied anti-epileptic pharmaceutical considered to be significantly resistant to photodegradation. The environmental t1/2 obtained under indirect photodegradation conditions are consistent with the conclusions taken for direct photodegradation. In general, these results highlight that photodegradation processes might not be efficient elimination pathways to prevent from accumulation of OXZ, DZP and ALP in aquatic environments. On the contrary, LRZ is quickly eliminated by photodegradation. In what concerns the environmental relevance of the results, it is important to consider a few sources of uncertainty. Firstly, to convert t1/2 to SSD it was assumed that photodegradation rates at environmentally relevant temperatures are not significantly different from the ones determined inside the irradiation chamber, where the temperature rose up to 35e40 C. Moreover, possible optical phenomena caused by the quartz tubes’ surface were neglected. Finally, the t1/2 (presented in SSD) are still dependent on the initial concentration of BDZ and quantitative differences in real environmental t1/2 might be possible. However, these facts do not invalidate the general conclusions taken concerning the observed differences in the photoresistance of the studied BDZ. To allow a better understanding of the persistence of BDZ in aquatic environments, further studies involving the role of other constituents of natural waters in the photodegradation processes are mandatory. Furthermore, the environmental relevance of these results is strictly related to the fate of these drugs in the water/sediment or soil interface. Literature data regarding sorption of psychiatric pharmaceuticals onto sediments evidenced their complex behavior (Stein et al., 2008), emphasizing that sorption and degradation phenomenons should be faced as complementary; only this approach would allow to attain valid conclusions. The determination of the photolysis apparent quantum yield (4) is also of high relevance. This parameter can be defined as the ratio between the compound photodegradation rate and the rate of light absorption which is directly related to the efficiency of the photolysis process (Petrovic and Barcelo´, 2007). Moreover, the determination of 4 allows a valid comparison with other literature studies. The average quantum yield (4ave) of direct photolysis of the studied BDZ was determined based on the overall average of lamp
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 9 7 e6 1 0 6
emission intensities in the wavelength range between 290 and 800 nm, by means of the following equation: fave ¼ P
I0li
C k 0 1 10εli lC0
(3)
where k is the apparent first order rate constant (s1), C0 is the initial concentration of each BDZ in solution (mol L1), I0li is the lamp emission intensity at the wavelength li (Ein L1 s1), ε is the molar absorptivity of each BDZ at li (L mol1 cm1) and l is the path length inside the photoreactor (1.5 cm). Absorption spectra of each benzodiazepine are presented in Fig. 2 and emission lamp spectrum is presented in Figure S1 of the SI. Results are displayed in Table 2. Apparent quantum yields of OXZ, DZP and ALP are in the same order of magnitude (106); the significantly higher t1/2 of ALP is explained by poor absorption of solar light. On the other hand, absorption of LRZ (for wavelengths above 290 nm) is similar to DZP and OXZ; nevertheless, LRZ has an apparent quantum yield approximately 20 times higher (in the order of 105), justifying the increase on the photodegradation rates of this compound. To the best of our knowledge, other data concerning BDZ photolysis quantum yields are not presented in the literature.
3.4. Identification of photodegradation products by mass spectrometry The identification of the photodegradation products were performed by comparing ESI(þ)MS spectra of each irradiated sample with spectra of the respective BDZ prior to irradiation (ESI-MS spectra shown in Figures S2eS5 of the SI). Subsequently, the molecular ions [M þ H]þ and/or [M þ Na]þ of possible photodegradation products were selected. The molecular formula and structure of the products were then tentatively assigned based on the fragmentation patterns observed in the ESI-MS2 spectra (detailed data in Table S1, SI). Structures of BDZ and proposed structures for photoproducts are presented in Fig. 3. A total of 19 BDZ photodegradation products were identified. All photoproducts, with only one
Fig. 2 e UVeVis spectra of the studied benzodiazepines in aqueous solution. The spectra were obtained with a UVeVis Shimadzu spectrophotometer and a cell with an optical path length of 1 cm was used.
6103
exception, have lower molecular weight than the parent compound and photo-induced dimerization was not observed. In general, the majority of the identified compounds is monochlorinated and clearly exhibited (in the ESI-MS spectra) the typical pattern of compounds containing one chlorine atom. Also, the loss of a chlorine radical (35 Da) or of a HCl molecule (36 Da) were frequently observed in the respective ESI-MS2 spectra. Photodegradation products of OXZ, DZP and LRZ mainly resulted from the opening of the diazepinone sevenmembered ring followed by a rearrangement into a highly stabilized six-membered ring (as it is the case of OXZ IVeVI, DZP I and LRZ VI), or even into an aromatic system of 3 adjacent 6(or 7)-membered rings (DZP III, LRZ IIeIV). Differentiation of compounds containing the phenyl ring was possible due to the commonly observed loss of the phenyl group (78 Da). Oxidation to alcohols and ketones was also frequently observed; the existence of carbonyl and hydroxyl groups was confirmed by the presence of ESI-MS2 base peaks which correspond to the loss of a CO group (28 Da) or a H2O molecule (18 Da). Interestingly, and despite from only differing on the presence of one chlorine atom in the phenyl ring, photodegradation of OXZ and LRZ resulted in the appearance of distinct photoproducts, showing that the extra ortho-chlorine atom in LZP plays a decisive role on photo-induced transformations. The major photodegradation product of LRZ was LRZ V (Mw 265, [M þ H]þ ¼ 266); the emergence of the compound LRZ II (Mw 213, [M þ H]þ ¼ 214) should also be highlighted due to its structural similarity with acridine, which is known for its carcinogenic activity (Calisto et al., 2011; Chiron et al., 2006) (LRZ II corresponds to chloroacridine). In addition, it is important to emphasize that it was possible to differentiate photoproducts with the same molecular weight (OXZ V, [M þ H]þ ¼ 257 and LRZ IV, [M þ H]þ ¼ 257) based on their different ESI-MS2 fragmentation patterns. In the case of ALP, the aromatic triazole ring seems to have a protective effect in the seven-membered ring opening and only 2 photodegradation products were identified (ALP I, Mw 297, [M þ H]þ ¼ 298 and ALP II, Mw 299, [M þ H]þ ¼ 300), also explaining the extremely low photodegradation rate of this compound. These 2 photoproducts also resulted from the opening of the seven-membered ring followed by oxidation to a ketone or alcohol. In order to further validate the suggested identification of the photoproducts, proposed molecular formulas were confirmed by exact mass measurement and elemental composition determination of the new ions, observed in the ESI-MS spectra after photodegradation. For the determination of the exact mass measurement of the new ions, lock mass was the calculated monoisotopic mass/charge of the nonmodified BDZ. The exact calculated monoisotopic masses were compared with observed masses in the ESI(þ)MS spectra (detailed results for each photoproduct are displayed in Table S2, SI). For a large number of predicted formulas, the errors between the observed and calculated masses are below 9 mDa (OXZ III-V, VII; DZP I, III and LRZ IeVI), confirming with good accuracy the elemental composition of the photoproducts. In the remaining cases, errors between observed and calculated masses did not overcome 28.4 mDa; these higher errors might be, in some cases, explained by the overlap of the
6104
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 9 7 e6 1 0 6
Fig. 3 e Structures of the studied benzodiazepines and proposed structures for photodegradation products. OZX I e VII, DZP I e IV, LRZ I e VI and ALP I e II correspond to photodegradation products of oxazepam, diazepam, lorazepam and alprazolam, respectively. Molecular weight (Mw) of each compound is also shown.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 9 7 e6 1 0 6
isotopic contribution of 13C or 37Cl of other identified products (OXZ VI, DZP IV). The identification of the photodegradation products of the selected BDZ is crucial to a more complete understanding of the environmental impact of their presence in aquatic systems. The photoproducts originated in the environment should also be considered as relevant environmental pollutants and, optimally, their impact in the environment should also be evaluated. The persistence and/or toxicity of these photoproducts might, in some cases, cause a more significant concern than the parent compounds. Thus, the proposed identification of the main photodegradation products, under environmentally relevant conditions, constituted the first step to allow more comprehensive approaches when assessing scarcely studied pharmaceuticals.
4.
Conclusions
In this study, the relevance of photodegradation processes on the aquatic environmental persistence of BDZ was evaluated for the first time. This work allowed to conclude that photodegradation might not prevent the environmental accumulation of OXZ, DZP and ALP which were shown to be considerably resistant to direct photodegradation (with halflife times of 4, 7 and 228 SSD). On the other hand, LRZ is susceptible to direct photodegradation, showing a half-life time of less than 1 SSD. The presence of DOM resulted in observed half-life times of the same order of magnitude from those obtained under direct photodegradation conditions. Nevertheless, the presence of HA was consistent with a decrease in the observed photodegradation rates, while FA and XAD-4 fraction caused an enhancement of the photodegradation. Moreover, one of the main focuses of this work consisted on the identification of the direct photodegradation products of the studied BDZ, by mass spectrometry. Overall, it was possible to identify a total of 19 photodegradation products, the majority of which were reported here for the first time. This work constitutes a first approach to the study of environmentally relevant photodegradation processes of BDZ. The results presented in this manuscript are, in our opinion, a valuable tool for the development of further research concerning this topic.
Acknowledgments Vaˆnia Calisto thanks FCT (Fundac¸a˜o para a Cieˆncia e Tecnologia e Portugal) for her PhD grant (SFRH/BD/38075/2007). The authors also acknowledge FCT for financially supporting QOPNA and RNEM.
Appendix. Supplementary material Supplementary material associated with this article can be found, in the online version, at doi:10.1016/j.watres.2011. 09.008.
6105
references
Boreen, A.L., Arnold, W.A., McNeill, K., 2003. Photodegradation of pharmaceuticals in the aquatic environment: a review. Aquatic Sciences 65 (4), 320e341. Boule, P., Bahnemann, D.W., Robertson, P.K.J., 2005. Environmental Photochemistry - Part II. Springer-Verlag, Berlin. Cabrera, C.G., de Waisbaum, R.G., Nudelman, N.S., 2005. Kinetic and mechanistic studies on the hydrolysis and photodegradation of diazepam and alprazolam. Journal of Physical Organic Chemistry 18 (2), 156e161. Calamari, D., Zuccato, E., Castiglioni, S., Bagnati, R., Fanelli, R., 2003. Strategic survey of therapeutic drugs in the rivers Po and Lambro in northern Italy. Environmental Science and Technology 37 (7), 1241e1248. Calisto, V., Domingues, M.R., Erny, G.L., Esteves, V.I., 2011. Direct photodegradation of carbamazepine followed by micellar electrokinetic chromatography and mass spectrometry. Water Research 45 (3), 1095e1104. Calisto, V., Esteves, V.I., 2009. Psychiatric pharmaceuticals in the environment - a review. Chemosphere 77 (10), 1257e1274. Castaneda, B., Ortiz-Cala, W., Gallardo-Cabrera, C., Nudelman, N.S., 2009. Stability studies of alprazolam tablets: effects of chemical interactions with some excipients in pharmaceutical solid preparations. Journal of Physical Organic Chemistry 22 (9), 807e814. Chiron, S., Minero, C., Vione, D., 2006. Photodegradation processes of the antiepileptic drug carbamazepine, relevant to estuarine waters. Environmental Science and Technology 40 (19), 5977e5983. Dalrymple, R.e.M., Carfagno, A.K., Sharpless, C.M., 2010. Correlations between dissolved organic matter optical properties and quantum yields of singlet oxygen and hydrogen peroxide. Environmental Science and Technology 44 (15), 5824e5829. Doll, T.E., Frimmel, F.H., 2004. Kinetic study of photocatalytic degradation of carbamazepine, clofibric acid, iomeprol and iopromide assisted by different TiO2 materials - determination of intermediates and reaction pathways. Water Research 38 (4), 955e964. Erny, G.L., Rodrigues, J.E.A., Gil, A.M., Barros, A.S., Esteves, V.I., 2009. Analysis of non-aromatic organic acids in beer by CE and direct detection mode with diode array detection. Chromatographia 70 (11e12), 1737e1742. Esteves, V.I., Cordeiro, N.M.A., da Costa Duarte, A., 1995. Variation on the adsorption efficiency of humic substances from estuarine waters using XAD resins. Marine Chemistry 51 (1), 61e66. Esteves, V.I., Otero, M., Duarte, A.C., 2009. Comparative characterization of humic substances from the open ocean, estuarine water and fresh water. Organic Geochemistry 40 (9), 942e950. Garbin, J.R., Milori, D.M.B.P., Simo˜es, M.L., da Silva, W.T.L., Neto, L.M., 2007. Influence of humic substances on the photolysis of aqueous pesticide residues. Chemosphere 66 (9), 1692e1698. Glassmeyer, S., Kolpin, D.W., Furlong, E.T., Focazio, M.T., 2008. In: Aga, D.S. (Ed.), Fate of Pharmaceuticals in the Environment and in Water Treatment Systems. CRC Press, Taylor & Francis Books, Boca Raton, pp. 3e51. Gonza´lez Alonso, S., Catala´, M., Maroto, R.R., Gil, J.L.R., de ´ G., Valca´rcel, Y., 2010. Pollution by psychoactive Miguel, A pharmaceuticals in the rivers of Madrid metropolitan area (Spain). Environment International 36 (2), 195e201. Gros, M., Petrovic, M., Ginebreda, A., Barcelo´, D., 2010. Removal of pharmaceuticals during wastewater treatment and
6106
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 0 9 7 e6 1 0 6
environmental risk assessment using hazard indexes. Environment International 36 (1), 15e26. Guerard, J.J., Miller, P.L., Trouts, T.D., Chin, Y.P., 2009. The role of fulvic acid composition in the photosensitized degradation of aquatic contaminants. Aquatic Sciences 71 (2), 160e169. Heberer, T., 2002. Tracking persistent pharmaceutical residues from municipal sewage to drinking water. Journal of Hydrology 266 (3e4), 175e189. Huerta-Fontela, M., Galceran, M.T., Ventura, F., 2010. Fast liquid chromatography-quadrupole-linear ion trap mass spectrometry for the analysis of pharmaceuticals and hormones in water resources. Journal of Chromatography A 1217 (25), 4212e4222. International Narcotics Control Board, U.N., 2010. Psychotropic Substances - Report 2009: Statistics for 2008, (New York). Jelic, A., Gros, M., Ginebreda, A., Cespedes-Sa´nchez, R., Ventura, F., Petrovic, M., Barcelo, D., 2011. Occurrence, partition and removal of pharmaceuticals in sewage water and sludge during wastewater treatment. Water Research 45 (3), 1165e1176. Jones, O.A.H., Lester, J.N., Voulvoulis, N., 2005. Pharmaceuticals: a threat to drinking water? Trends in Biotechnology 23 (4), 163e167. Kar, A., 2007. Medicinal Chemistry. New Age International Publishers, New Delhi, India. Minero, C., Chiron, S., Falletti, G., Maurino, V., Pelizzetti, E., Ajassa, R., Carlotti, M., Vione, D., 2007. Photochemincal processes involving nitrite in surface water samples. Aquatic Sciences - Research Across Boundaries 69 (1), 71e85. Newman, R.H., 2007. Agriculture and Soils. John Wiley & Sons, Ltd. OECD, 2008. Phototransformation of Chemicals in Water- Direct Photolysis, Paris. Petrovic, M., Barcelo´, D., 2007. Analysis, Fate and Removal of Pharmaceuticals in the Water Cycle. Elsevier, Amsterdam, The Netherlands.
Santos, M.E., Esteves, V.I., Amado, F.L., Duarte, A.C., 1994. In: Senesi, N., Miano, T.M. (Eds.), Humic Substances in the Global Environment and Implications on Human Health. Elsevier, Amsterdam, pp. 877e882. Schwarzenbach, R.P., Gschwend, P.M., Imboden, D.M., 2003. Environmental Organic Chemistry. John Wiley & Sons, New Jersey. Stein, K., Ramil, M., Fink, G., Sander, M., Ternes, T.A., 2008. Analysis and sorption of psychoactive drugs onto sediment. Environmental Science and Technology 42 (17), 6415e6423. Ternes, T., Bonerz, M., Schmidt, T., 2001. Determination of neutral pharmaceuticals in wastewater and rivers by liquid chromatography-electrospray tandem mass spectrometry. Journal of Chromatography A 938 (1e2), 175e185. van der Ven, K., Van Dongen, W., Maes, B.U., Esmans, E.L., Blust, R., De Coen, W.M., 2004. Determination of diazepam in aquatic samples by capillary liquid chromatography-electrospray tandem mass spectrometry. Chemosphere 57 (8), 967e973. Vione, D., Falletti, G., Maurino, V., Minero, C., Pelizzetti, E., Malandrino, M., Ajassa, R., Olariu, R.-I., Arsene, C., 2006. Sources and sinks of hydroxyl radicals upon irradiation of natural water samples. Environmental Science and Technology 40 (12), 3775e3781. Xia, X., Li, G., Yang, Z., Chen, Y., Huang, G.H., 2009. Effects of fulvic acid concentration and origin on photodegradation of polycyclic aromatic hydrocarbons in aqueous solution: importance of active oxygen. Environmental Pollution 157 (4), 1352e1359. Zafiriou, O.C., Joussot-Dubien, J., Zepp, R.G., Zika, R.G., 1984. Photochemistry of natural waters. Environmental Science and Technology 18 (12), 358Ae371A. Zepp, R.G., Schlotzhauer, P.F., Sink, R.M., 1985. Photosensitized transformations involving electronic energy transfer in natural waters: role of humic substances. Environmental Science and Technology 19 (1), 74e81.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 0 7 e6 1 1 8
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
A photosynthetic rotating annular bioreactor (TayloreCouette type flow) for phototrophic biofilm cultures A. Paule a,b, B. Lauga c, L. Ten-Hage a,b, J. Morchain d,e,f, R. Duran c, E. Paul d,e,f, J.L. Rols a,b,* a
Universite´ de Toulouse, UPS, INP, EcoLab (Laboratoire d’e´cologie fonctionnelle et environnement), 118 route de Narbonne, F-31062 Toulouse, France b CNRS, EcoLab, F-31062 Toulouse, France c Equipe Environnement et Microbiologie, Institut Pluridisciplinaire de Recherche sur l’Environnement et les Mate´riaux - IPREM, UMR 5254 CNRS/UPPA, IBEAS, Universite´ de Pau et des Pays de l’Adour, BP1155, F-64013 Pau, France d Universite´ de Toulouse, INSA, LISBP, 135 Avenue de Rangueil, F-31077 Toulouse, France e INRA, UMR792, Inge´nierie des Syste`mes Biologiques et des Proce´de´s, F-31400 Toulouse, France f CNRS, UMR5504, F-31400 Toulouse, France
article info
abstract
Article history:
In their natural environment, the structure and functioning of microbial communities from
Received 11 April 2011
river phototrophic biofilms are driven by biotic and abiotic factors. An understanding of the
Received in revised form
mechanisms that mediate the community structure, its dynamics and the biological
1 September 2011
succession processes during phototrophic biofilm development can be gained using
Accepted 3 September 2011
laboratory-scale systems operating with controlled parameters. For this purpose, we
Available online 14 September 2011
present the design and description of a new prototype of a rotating annular bioreactor (RAB) (TayloreCouette type flow, liquid working volume of 5.04 L) specifically adapted for
Keywords:
the cultivation and investigation of phototrophic biofilms. The innovation lies in the
Rotating annular bioreactor
presence of a modular source of light inside of the system, with the biofilm colonization
TayloreCouette type flow
and development taking place on the stationary outer cylinder (onto 32 removable poly-
T-RFLP
ethylene plates). The biofilm cultures were investigated under controlled turbulent flowing
Phototrophic biofilm
conditions and nutrients were provided using a synthetic medium (tap water supple-
Microbial community
mented with nitrate, phosphate and silica) to favour the biofilm growth. The hydrodynamic
Photobioreactor
features of the water flow were characterized using a tracer method, showing behaviour corresponding to a completely mixed reactor. Shear stress forces on the surface of plates were also quantified by computer simulations and correlated with the rotational speed of the inner cylinder. Two phototrophic biofilm development experiments were performed for periods of 6.7 and 7 weeks with different inoculation procedures and illumination intensities. For both experiments, biofilm biomasses exhibited linear growth kinetics and produced 4.2 and 2.4 mg cm2 of ash-free dry matter. Algal and bacterial community structures were assessed by microscopy and T-RFLP, respectively, and the two experiments were different but revealed similar temporal dynamics. Our study confirmed the performance and multipurpose nature of such an innovative photosynthetic bioreactor for phototrophic biofilm investigations. ª 2011 Elsevier Ltd. All rights reserved.
* Corresponding author. Universite´ de Toulouse, UPS, EcoLab (Laboratoire d’e´cologie fonctionnelle et environnement), 118 route de Narbonne, F-31062 Toulouse, France. Tel.: þ33 0 6 24 38 19 04; fax: þ33 0 5 61 55 60 96. E-mail address: [email protected] (J.L. Rols). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.09.007
6108
1.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 0 7 e6 1 1 8
Introduction
Environmental phototrophic biofilms are microbial aggregates occurring on solid substrates and consisting of heterotrophic micro- and meio-organisms and phototrophic micro-organisms embedded in an extracellular polymeric substance matrix. The structure and functioning of microbial communities from phototrophic biofilms are mediated by abiotic factors such as nutrient availability (Hillebrand and Sommer, 2000a), light (Boston and Hill, 1991), substrate types (Murdock and Dodds, 2007), hydrodynamics (Battin et al., 2003), and by biotic interactions such as competition (Jackson et al., 2001) or predation (Bourassa and Cattaneo, 1998). Biofilm development has been demonstrated to be associated with population succession processes over biofilm maturation, both for the algal (McCormick and Stevenson, 1991) and the bacterial (Jackson et al., 2001; Lyautey et al., 2005) compartments. To understand how the abiotic and biotic factors (alone or combined) influence the microbial community structure, its dynamics and the biological succession processes during phototrophic biofilm development, the best approach is to use laboratory-scale systems simulating environmental conditions under different levels of experimental control. Various large and small-scale laboratory systems designed to investigate phototrophic biofilms are described in the literature (e.g. Battin et al., 2003; Singer et al., 2006). Among them, rotating annular bioreactor (RAB) designs have been suggested as a powerful tool to study the effects of environmental change on biofilm development (Neu and Lawrence, 1997). It has been shown that the hydrodynamic conditions at local level influence the composition and the structure of biofilms (Besemer et al., 2007). The geometry of RABs allows to provide a constant shear stress distribution and cultivation of biofilm under turbulent flow environments (Characklis, 1990). While RABs are described as completely mixed reactors for the liquid phase, a previous study showed heterogeneity in the growth of biofilm related to reactor geometry (Gjaltema et al., 1994). In the last decade, Lawrence et al. (2000) have developed a RAB (liquid working volume of 0.5 L) for the cultivation of phototrophic biofilms, used to investigate the various effects of environmental change occurring in a river (Che´nier et al., 2003; Lawrence et al., 2004). The main shortcomings of this RAB are its small size which limits the number of possible analyses and replicates, the external illumination, and the biofilm growing on the rotating inner cylinder. The objectives of our study were (i) to design and describe a new prototype of RAB (TayloreCouette type flow) specifically intended for the cultivation and investigation of phototrophic biofilms adapted from an RAB design for biological waste water treatment (Coufort et al., 2005), (ii) to assess the applicability of this prototype in phototrophic biofilm production and (iii) to analyze the phototrophic biofilm dynamics. Innovations of our modified RAB were the presence of a modular source of light inside the system and the biofilm colonization on the stationary outer cylinder. Two cultivation experiments were performed for periods of 6.7 and 7 weeks with different inoculation procedures and illumination intensities.
2.
Material and methods
2.1.
Experimental setup
Phototrophic biofilm culture experiments were conducted in a new prototype of a photosynthetic rotating annular bioreactor (RAB) with TayloreCouette type flow (Arias, Toulouse, France).
2.1.1.
RAB characteristics
The RAB consisted of two concentric cylinders, a stationary outer cylinder made of polyvinyl chloride and a rotating inner cylinder made of poly(methyl methacrylate) (PMMA) (Fig. 1A and B). A schematic diagram and the geometric characteristics of the RAB are given in Fig. 1C and D, respectively. This prototype presents the specificity of having (i) a modular source of light inside the system, protected by an internal water-tight cylinder made of PMMA and adjusted by changing the quality and number of the fluorescent tubes (1e8) and the frequency of light/dark cycles, and (ii) a flow generated in the annular gap (width 18.5 mm) through the rotation of the inner cylinder modulated by different motor speeds. The inside of the external cylinder supports 2 rows of 16 removable polyethylene plates or sampling units (l h ¼ 50 100 mm; 5 mm wide) for biofilm sampling. The total surface available for the biofilm colonization of plates in the RAB is 0.16 m2. To limit the occurrence of edge effects on the development of biofilm, the rows of plates were positioned at half height in the bioreactor. The plates were curved to avoid perturbation of the flow. To prevent biofilm growth on the back, upper part and leading edge of the plates, these surfaces were covered by adhesive bands during the experiments which were removed before the biofilm analyses. All bioreactor components were cleaned, with diluted detergent (Decon, 10%) for the plates, the outer cylinder and the port, or with hydrogen peroxide (30%) for the inner cylinders, and then rinsed with demineralized water. To prevent unwanted biofilm formation that could attenuate the light intensity and modify its spectrum, the surfaces of the rotating inner and internal water-tight cylinders were cleaned manually once a week. This step of 15 min required to collect the liquid contained in the RAB before opening, and allowed, if necessary, to collect some plates for biofilm analyses. Once finished, the bioreactor was closed and refilled with the collected liquid.
2.1.2.
RAB hydrodynamic behaviour
The RAB was operated at 80 rpm, which corresponds to a Reynolds number Re ¼ ri.U.(reri)/y ¼ 17,040 and Taylor number Ta ¼ Re.[(reri)/ri]1/2 ¼ 6970 where ri is the inner cylinder radius (m), re is the outer cylinder radius (m), U is the angular speed (rad s1) of the inner cylinder, and y is the cinematic viscosity of the fluid (m2 s1) (tap water). According to the literature, this value of Taylor number indicates a turbulent vortex flow with stacked axisymmetric toroidal vortices (Desmet et al., 1996). Bioreactor with TayloreCouette type flow exhibits different flow regimes (e.g. Couette, vortex flow, turbulent vortex flow, turbulent flow.) depending on the rotational speed of the inner cylinder. In the RAB designed
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 0 7 e6 1 1 8
6109
Fig. 1 e Setup of the photosynthetic rotating annular reactor (RAB) with TayloreCouette type flow. (A) 3D representation of different parts composing the RAB, a : the stationary outer cylinder, b : the rotating inner cylinder and c : the light source protected by an internal water-tight cylinder, (B) photograph of RAB, (C) schematic diagram of RAB and (D) geometric characteristics of RAB.
for the present work, the objective was to work with turbulent vortex flow with spatial periodicity and rotational speed of the inner cylinder high enough to avoid the settling of microorganisms in the annular gap. In this context, the rotational speed of the inner cylinder was set to at least 80 rpm. The RAB hydrodynamic was studied experimentally at the reactor scale by the tracer method and local flow properties were obtained through computational fluid dynamics (CFD) simulation.
2.1.2.1. Residence time distribution. The general mixing behaviour in the RAB was investigated experimentally using the pulse tracer method (10 mL of NaCl solution at 0.16 g mL1)
to determine the residence time distribution (RTD). The experiment was conducted for two different rotational speeds, 80 and 170 rpm, and the inlet throughput (tap water at 20 C) was supplied at Q ¼ 26 mL min1 for a working volume in the RAB of V ¼ 5.04 L. The conductivity of the fluid was recorded at the outlet for 15 h (corresponding to 5 times the average residence time) with a specific probe (conductivity meter 524, CRISON, SELI, probe response time of 2 s) located in an agitated cell (30 mL) positioned at the outlet valve, in absence of biofilm and without illumination in the RAB. RTD curves, defined as dimensionless concentration (E (q)) versus dimensionless time (q), were obtained from the outlet conductivity concentration data:
6110
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 0 7 e6 1 1 8
EðqÞ ¼ cðtÞ=cð0Þ
(1)
where q is given by the ratio t/s with s ¼ V/Q, c(t) and c(0) are respectively the tracer concentrations at time t and time 0 for which c(0) results from an instantaneous mixing of the injected tracer. The experimental RTD curves were compared with the RTD curve obtained from a mathematical model of reactor as described by Sugiharto et al. (2009).
2.1.2.2. Computational study of hydrodynamics. Numerical simulation was performed to evaluate the flow pattern within the annular space and the characteristic turbulent scales. The mean wall shear stress on the external cylinder and the axial average velocity profiles in the annular gap were extracted from the simulations. The computational study was performed using the CFD software Fluent (6.2) at the rotational speeds of 80 and 170 rpm. The first step was to draw the grid and mesh the two-dimensional domain using the Fluent preprocessor Gambit. The simulations were run as described by Coufort et al. (2005), i.e. Reynolds Averaged NaviereStokes equations combined with the k-ε Reynolds Stress Model, 2Daxisymmetric model in the steady state. 2.2.
Experimental design
Initially, the bioreactor was run in batch culture mode for a seeding period to allow the micro-organisms to become attached before the continuous culture mode started.
2.2.1.
Seeding procedures
The two biofilm cultures were achieved using two different seeding procedures. Seeding was conducted for 48 h, once for culture 1 and twice for culture 2. For culture 2, the two seeding phases were separated by a 24-hour period where the RAB operated in continuous culture mode. During the seeding phases, the bioreactor ran in closed recirculation, connected to an aquarium (10 L) where the inoculum was incubated. The aquarium was illuminated by fluorescent lamps including one cool daylight (F18W/GRO, Sylvania, Germany) and one fluora (F18W/54, Gt Britain) tubes, supplying average illumination values of 32 3 mmol s1 m2 with light/dark periods of 16 h/8 h. The inoculum was obtained by removing epilithic biofilms by scraping with a toothbrush, previously treated with NaOH 1N, from (i) glass slides as previously described (Paule et al., 2009) placed in the experimental channel of our laboratory for culture 1 or (ii) various river stones for culture 2. Biofilm suspensions were homogenized (tissue homogenizer at 13,500 rpm, Ultra Turrax, T25) and filtered through a 250 mm and then 100 mm pore size filter (VWR) to reduce the part of the macro fauna and coarse sediments from the natural biofilms. The end of seeding phase was defined as the start of the experiment (day 0).
2.2.2.
Experimental conditions
The experiment design used a thermostated reservoir (150 L, model CV 150, Japy) at 4 C, equipped with a peristaltic pump (520S/R2 220 T/MN pump with silicon tubes ID OD ¼ 1.6 2.4 mm) which fed the RAB continuously with a synthetic culture medium. The inlet throughput was 26 mL min1, which corresponded to a hydraulic residence time in the RAB of 3.23 h. The synthetic culture medium consisted of tap water supplemented with nutrients (SiO2, PO3 4 and NO3 ) to favour the growth of biofilm and avoid nutrient limitation. Nutrient concentrations were measured as described by Paule et al. (2009). The physical-chemical parameters (temperature, pH and dissolved oxygen concentration) were recorder using probes located in the agitated cell (30 mL) positioned at the outlet valve of the reactor. Temperature and pH were measured with a pH meter 296 WTW (electrodes sentix H 8481 HD, SCHOTT). Dissolved oxygen concentrations were measured with an oxy 296 oxymeter WTW (trioxmatic 701 sensor, WTW). Dissolved organic carbon (DOC) concentrations were measured on acidified water samples (4 mL of HCl 6N) and analyzed using a carbon analyser at 680 C (Shimadzu, Model TOC 5000H). Table 1 summarizes the chemistry of the feed waters for both cultures. For culture 2, the pH of the culture medium was adjusted to 7.0 using sulphuric acid (95%). The inside of the RAB was illuminated by fluorescent lamps including cool daylight (Osram L15W/865 Luminux, Germany) and fluora (Osram L15W/77, Germany) tubes in equal proportions, with light/dark periods of 16 h/8 h. Fluora tubes emit in the visible red, which enhances photosynthesis. At the center of the cylinder containing neon tubes, a cylinder of PMMA is positioned to improve the distribution of the light. Two neon tubes were used for culture 1 and 4 for culture 2. The illumination was measured as air photosynthetically active radiation (PAR) irradiance level by using a flat quantum sensor (model LI-189, LI-COR, Inc - Lincoln - Nebraska) and average recorded values were 130 20 and 180 10 mmol s1 m2 for cultures 1 and 2 respectively. The PAR irradiance level was measured in the air because of the small size of the annular gap, and at a distance from the rotating inner cylinder equivalent to the annular gap. The values of illumination were chosen in this study in response to two constraints. The number of neon tubes (2 and 4) is a good compromise to maintain illumination homogeneity (the fewer lamps are used, the less uniform the light field is) and to prevent an increase of temperature generated by the presence of neon tubes (the RAB is not thermostated).
2.2.3.
Biofilm characterization
The development of biofilm was monitored for 6.7 (culture 1) and 7 (culture 2) weeks. Biofilm cultures were carried out between June 11 and July 30, 2008 for culture 1 and between
Table 1 e Physical-chemical characteristics of the synthetic water used to feed the rotating annular bioreactor during cultures 1 (C1) and 2 (C2). DOC [ dissolved organic carbon concentration.
C1 C2
1 PO3 4 eP (mg L )
1 NO 3 eN (mg L )
SiO2 (mg L1)
Conductivity (mS cm1)
DOC (mg L1)
pH
0.357 0.03 0.356 0.02
6.3 0.1 4.2 0.2
13.1 0.7 10.9 2.9
311 29 368 5
0.6 0.2 1.1 0.3
8.0 0.5 7.1 0.2
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 0 7 e6 1 1 8
July 23 and September 16, 2009 for culture 2. At each sampling date, 3 plates were randomly sampled to follow the biofilm development. Access to the plates required opening the bioreactor and removing working fluid. Biofilms were removed from plates by scraping with a microscope slide previously treated with alcohol. Each plate represented one replicate. Biofilms were suspended in 50 mL (culture 1) or 90 mL (culture 2) of tap water previously filtered through a 0.2 mm pore size filter (cellulose acetate membrane, Whatman) and homogenized (tissue homogenizer at 13,500 rpm, Ultra Turrax, T25). Biofilm suspension was aliquoted for the analyses of biomass descriptors, algal diversity and bacterial community structure by T-RFLP. Sampled plates were substituted by clean plates in the RAB and the newly placed plates were excluded from the following samplings.
2.2.3.1. Biomass descriptors. From an aliquot of initial biofilm suspension, the dry mass (DM) (aliquot of 30 mL), the ash-free dry mass (AFDM) and the chlorophyll a (aliquot of 10 mL) were measured as described by Paule et al. (2009). 2.2.3.2. Algal diversity. Algal diversity was estimated from a pool of 3 aliquots of 5 mL of homogenized biofilm suspension that was preserved in formalin solution (3%) and kept in darkness at 4 C until counting and identification. The total density and abundance percentages were determined with an inverted microscope (Axiovert 10, Zeiss, West Germany) (Utermo¨hl, 1958).
6111
previously described by Smith et al. (2005) with the confidence interval of 0.5. Peaks, defined as Terminal Restriction Fragments (T-RFs), were scored as present or absent from T-RFLP profiles. The difference in physical-chemical characteristics and the difference in AFDM, chlorophyll a and the number of T-RFs between biofilm samples were assessed with the Mann Whitney test using SPSS software 13.0. Differences were considered statistically significant at p 0.05. To assess changes over time in the bacterial community structure from each culture, a Principal Component Analysis (PCA) was performed from the T-RF binary data for each biofilm culture using Primer v6 software (PrimerE, Ltd, Lutton, United Kingdom). Peaks < 0.5% of the total area were excluded from the analysis and T-RFs that differed in size by 0.5 bp or less were considered to be identical. This baseline of 0.5% was defined in accordance with the approaches of Osborne et al. (2006). Statistical analyses of PCA were run using an analysis of similarity (ANOSIM) via Past software 2.06 (Hammer et al., 2001) on Bray Curtis similarity matrices generated from binary data. This analysis generates a global R value in the range from 0 (completely random pattern) to 1 (completely separated groups) (Clarke, 1993). The global R value was considered statistically significant at p < 0.05 uncorrected.
3.
Results
3.1.
Biofilm culture conditions
2.2.3.3. Microbial community structure. After centrifugation (12,000 g at 4 C for 20 min, Heraeus Multifuge) of an aliquot of 20e50 mg dry mass of the initial biofilm suspension (Lyautey et al., 2005), the pellet was stored at 80 C until further analysis. Genomic DNA extraction was performed on the pellet using a DNeasy Plant Mini Kit according to the manufacturer’s protocol (Qiagen Laboratories). The integrity of the extracted DNA was checked as described by Paule et al. (2009). The 16S rRNA genes were amplified by PCR and the bacterial community structure was studied by T-RFLP as described by Bruneel et al. (2006) with slight modifications. The fluorescent labelled primers FAM 8F (50 -6-carboxy-fluorescein-phosphoramidite-AGA GTT TGA TCC TGG CTC AG-30 ) (Eurogentec, 295 Liege, Belgium) (Lane, 1991) and HEX 1489R (50 -hexa-chloro-fluorescein-phosphoramidite- TAC CTT GTT ACG ACT TCA-30 ) (Invitrogen, Carlsbad, USA) (Weisburg et al., 1991), described as universal within the bacterial domain, were used. The reaction mixture for PCR was made in a 50 mL volume containing 30 ng of template DNA, 25 mL AmpliTaq Gold 360 Master Mix (Applied Biosystems) and 0.5 mL of each primer. Amplification was carried out using an Applied Biosystems thermocycler with the following sequence: a 5 min hot start at 95 C, followed by 35 cycles consisting of denaturation (45 s at 95 C), annealing (45 s at 55 C) and extension (1 min at 72 C), and a final extension at 72 C for 10 min. Restriction digestion was performed with HinfI.
2.3.
Throughout the experiments and associated with daily variations and photosynthetic processes, the temperature, pH and dissolved oxygen concentration values ranged from 19 to 30 C, from 7.5 to 10, and from 4 to 18 mg L1, respectively. According to residual nutrient concentration values measured at the outlet of the RAB (data not shown), nutrients added in the synthetic culture medium were sufficient to support biofilm growth.
3.2.
RAB hydrodynamic behaviour
This section describes the characterization of the hydrodynamic behaviour of the flow in the annular space between the inner and outer cylinders through both experimental and numerical studies, performed at two rotational speeds: 80 and 170 rpm. Similar results were observed for both rotational speeds and only the data corresponding to 80 rpm are presented here.
3.2.1.
RTD experiment
Fig. 2 compares the RTD curves obtained with experimental data and predicted model simulation for one completely mixed reactor. The experimental and predicted model curves are similar. The experimental mean residence times were 209.9 and 220.8 min for rotational speeds of 80 and 170 rpm respectively.
Data analysis 3.2.2.
T-RFLP profiles from the two cultures were compared by a web-based tool, T-Align (http://inismor.ucd.ie/wtalign/) as
Computational study of hydrodynamics
For the rotational speed of 80 rpm, Fig. 3 shows a contour plot of local velocity in the annular space computed at the scale of
6112
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 0 7 e6 1 1 8
Fig. 2 e Comparison between experimental (black dots) and predicted (white dots) Residence Time Distribution (RTD) curves at 80 rpm rotational speed and 26 mL minL1 inlet flow. Predicted curve corresponds to model with a completely mixed reactor.
two vortices, together with the outer cylinder wall shear stress for the same conditions (the dotted curve in Fig. 3). The presence of vortices produced a gradient of velocity at local scale on the walls of the cylinders. Consequently, the wall shear stress, which is directly related to the velocity gradient, was clearly non-uniform along the plate. In the zone of convergence of two vortices near the plate, the shear stress was maximal and a radial flow formed from plate to inner cylinder. The two vortices separated near the plate and the shear stress was minimal when the radial flow reached the plate. The diameter of an individual vortex is approximately equal to the annular gap and thus the number of stacked vortices was 5.4 across the height of one plate. The magnitude of the wall shear stress along plates increased with the rotational speed of the inner cylinder (r2 ¼ 0.99). The mean values of shear stress were calculated as described by Coufort et al. (2005) and were found to be 1.11 and 4 Pa at rotational speeds of 80 and 170 rpm respectively. Fig. 4 presents the average tangential velocity profile at a rotational speed of 80 rpm. The profile shows a decrease in tangential velocity across the inner and outer cylinders, which is characteristic of a turbulent vortex flow (Coufort et al., 2005). As a result, tangential velocity is about 0.3 m s1 at the plate wall when rotational speed is 80 rpm and 0.7 m s1 for 170 rpm (data not shown).
3.3.
Biofilm analyses
3.3.1.
Biomass descriptors
Fig. 5 illustrates the biofilm colonization of plates over both experiments. The first colonization states occurred on the ridges of plates. Biofilm biomass as expressed by AFDM and chlorophyll a presented similar linear growth patterns, giving a biomass peak of 4.2 and 2.4 mg AFDM cm2 and 0.05 and
Fig. 3 e Contour plot of velocity magnitude field (m sL1) in the annular space between inner (left) and outer (right) cylinders computed by CFD at rotational speed of 80 rpm, and the corresponding wall shear stress along plates placed inside the outer cylinder (e is the width of the annular gap).
Fig. 4 e Profile of tangential velocity along a line of constant height in the annular space between the inner (left) and outer (right) cylinders computed by CFD for a rotational speed of 80 rpm.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 0 7 e6 1 1 8
6113
Fig. 5 e Photographic images of colonized plates on the internal surface of the external cylinder of the RAB during culture 1 (A) and culture 2 (B). Each plate has dimensions 50 3 100 mm2. Numbers indicate biofilm age in weeks. White and partially colonized plates correspond to newly placed plates after sampling.
0.03 mg chlorophyll a cm2 after 6 and 4.4 weeks of incubation for cultures 1 and 2 respectively (Figs. 6 and 7). This growth phase was followed by a plateau (Mann Whitney p > 0.05), then, for culture 2, by a slight loss of biomass (visible on the illustration of Fig. 5) (Mann Whitney, p < 0.05). Both variables (AFDM and chlorophyll a) were significantly correlated for the two cultures (C pearson ¼ 0.95, p < 0.01 and C pearson ¼ 0.93, p < 0.01, for cultures 1 and 2 respectively). The AFDM/DM ratio ranged from 39.5 to 62.1%, indicating biofilms poor in detritus and sedimentary particles (data not shown).
3.3.2.
Algal diversity
Two seeding procedures were tested using inocula of different origins. The inoculum from artificial biofilm (experimental channel) used for culture 1 presented lower species richness (7 species) than the inoculum from natural biofilm (river) used for culture 2 (27 species). Moreover, the two inoculum types had different algal community compositions (Fig. 7). Inoculum from artificial biofilm was dominated by Cyanobacteria (95.4%) composed essentially of Leptolyngbya spp. (88.5%), and inoculum from natural biofilm was dominated by Diatoms
6114
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 0 7 e6 1 1 8
Fig. 6 e Temporal evolution of photosynthetic biofilm biomass expressed as mg cmL2 of ash-free dry mass (AFDM) during the growth period in bioreactor for culture 1 (black dots) and culture 2 (white dots).
(86.3%), composed essentially of Navicula tripunctata (O.F. Mu¨ller) Bory (23.2%), Nitzschia spp. (19.4%) and Achnanthes spp. (12.3%). The species richness was relatively low and constant during culture 1 (from 5 to 7) although the species richness decreased over time during culture 2 (from 27 to 8). The 6.7week mature biofilm from culture 1 was mainly composed of Diatoms (98.6%), especially Nitzschia palea (Kutz.) W. Smith (92.6%) and the mature biofilm from culture 2 was essentially composed of green algae (95.5%) especially Scenedesmus (74.9%) and Ankistrodesmus/Monoraphidium (9.9%). Six of the ten Chlorophyceae that composed the biofilm of culture 2 are known to present planktonic ecotypes.
3.3.3.
Bacterial community structure
The dynamics of the bacterial community structure were determined by T-RFLP throughout the experiments. A total of
35 different T-RFs per culture for all sample times were identified with an average number per sample ranging from 12 to 25 and 12 to 23 T-RFs, for cultures 1 and 2 respectively. Principal component analysis (PCA) was performed on the TRF binary data for each biofilm culture (Fig. 8). The two axes accounted for 21.2 and 18.6% of the total variance for culture 1, and 30.6 and 18.1% for culture 2. Good homogeneity was observed among replicates, particularly for culture 2, suggesting little spatial variability during the culture course in the bioreactor. This analysis was strengthened by the similar trends observed with PCAs built with the first and third axes. The first three axes accounted for 65.5 and 50.2% of the variation of T-RFLP patterns for cultures 1 and 2 (data not shown) respectively. During both cultures, the bacterial community structure changed according to colonization time, based on sample clustering corresponding to a similarity of 55% (circle from Fig. 8) (culture 1 : global R ¼ 0.776 and pairwise R ranged from 0.63 to 0.8, p < 0.05; culture 2 : global R ¼ 0.998 and pairwise R ranged from 0.997 to 1, p < 0.05), followed by a stable phase after 3 weeks for culture 1 (global R ¼ 0.435, p < 0.05) and 4.4 weeks for culture 2 (global R ¼ 0.51, p < 0.05). Bacterial community composition rapidly diverged from the initial bacterial community (global R ¼ 0.895 and 0.754, for cultures 1 and 2 respectively, p < 0.05). A PCA including T-RF of cultures 1 and 2 showed that the profiles were distributed along the first axis (28.9%) according to the origin of the inoculum (data not shown). PCA showed similar temporal variations of bacterial community structures during the biofilm development irrespective of the inoculum type.
4.
Discussion
4.1.
Growth dynamics of phototrophic biofilms
In environmental phototrophic biofilms, growth basically occurs through, firstly, an accretion phase related to colonization and growth processes (increase of AFDM resulting in a biomass peak) and, secondly, an ageing phase (Biggs, 1996).
Fig. 7 e Temporal evolution of algal taxa number and their percentage abundance, and chlorophyll a as mg cmL2 of chlorophyll a during the biofilm growth period in RAB for cultures 1 (A) and 2 (B). The species richness is given for each sample at the bottom of each bar.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 0 7 e6 1 1 8
6115
Fig. 8 e Changes in bacterial community structure over time assessed by Principal Component Analysis (PCA) based on the T-RFLP data for (A) culture 1 and (B) culture 2. Circles correspond to a similarity of 55%.
In the present experiments, linear growth phases were observed to reach AFDM peaks of 4.2 and 2.4 mg cm2 after 6 and 4.4 weeks for cultures 1 and 2 respectively, followed by an ageing phase for culture 2. During the first step of each experiment, the colonization by suspended biomasses preferentially occurs on the substrate ridges. At the junction between two plates, the presence of gap was responsible for a recirculating flow near the solid surface. This stationary zone can act as a trap where biomass can accumulate, which could favour biofilm growth. Typically, the first species to colonize the substrate are heterotrophic bacteria and algal with fast growth rate and small cells, followed by the settlement and colonization of slow growth and large cells species (Biggs et al., 1998; Sekar et al., 2002; Roeselers et al., 2007). Thus, Cyanobacteria are considered as late colonizers, with a slow growth rate (Sekar et al., 2002). This could explain their fast disappearance during both cultivations and their recurrence in the 6.7-weekold biofilm during culture 1. The use of a more diversified inoculum (natural) and longer seeding phase for culture 2 did not seem to enrich the algal community. As observed in most experiments performed at laboratory scale (Bouleˆtreau et al., 2010), biofilms at the end of the experiment exhibited poor algal specific richness, of 6 and 7 species for cultures 1 and 2 respectively. As a result, the minimal number of algal species to preserve an integral biofilm in this RAB seemed to be a final number of 6 or 7 species. The use of a single short seeding phase in the present work (48 h or twice 48 h) may have limited the adhesion of micro-organisms or only selected pioneer algal species and could thus be inherent to this poor diversity. The choice of constant experimental conditions during biofilm development did not favour environmental changes as observed in a natural environment (Biggs, 1996). It is known that, in an undisturbed environment (e.g. constant hydrodynamic conditions), autogenic processes appear and the more competitive species dominate (competitive exclusion), which can explain the poor algal diversity and lead to self detachment (Bouleˆtreau et al., 2006) as observed during culture 2 at 4.4 weeks of colonization. It has been suggested
that initial high diversity is caused by the arrival of new microorganisms, while ensuing competition decreases diversity in late successional stages (Sekar et al., 2002). In our RAB, the absence of arrival of new micro-organisms throughout experiment may have caused competition processes even during the first steps of colonization. In microcosm studies, the continuous seeding processes enable the natural conditions to be reproduced but can interact with the disturbance under study (Tlili et al., 2008). The settings of the variables (temperature, light intensity, nutrient content, and flow rate), chosen to favour the growth of biofilms can be very selective for some species. For instance temperatures between 0 and 25 C increased species richness and diversity and temperatures above 30 C decreased species richness (DeNicola, 1996). Moreover, previous studies have shown that nutrient ratios (N, P and Si) greatly influence the composition of algal communities (Hillebrand and Sommer, 2000b) and that the enrichment of the medium favours the dominance of single species (Hillebrand and Sommer, 2000a). End-of-experiment biofilms were strongly dominated either by N. palea (Kutz.) W. Smith. or Scenedesmus genus for culture 1 and culture 2, respectively. N. palea (Kutz.) W. Smith. and the Scenedesmus genus are eutrophic and polysaprobic species, which reveal nutrient-rich waters with strong conductivity (Tison et al., 2004; Pen˜a-Castro et al., 2004). This is consistent with the physical-chemical characteristics of the culture medium used. Scenedesmus genus is a planktonic species and the seeding phase conducted in suspension could have induced its selection. The time taken to reach the AFDM peak was shorter for culture 2 than for culture 1 and was followed by a slight biomass removal leading to the ageing phase. Zippel and Neu (2005) concluded that green-algaldominated biofilms presented a less stable and compact structure caused by a faster growth rate. The authors observed that fast development induced the formation of poorly diversified biofilms, probably explained by an economy and partition of resources (Zippel and Neu, 2005). One possible reason why algal diversity is small can be attributed to taxonomical analyses based on morphotypes. Many different
6116
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 0 7 e6 1 1 8
species can fall into the same category and can induce an under-estimation of specific richness. Bacterial communities from both cultures changed markedly over the development of the biofilms. Despite differences in the inoculum communities, the succession was similar for both cultures associated with different trajectories. Since there was no addition of micro-organisms after the seeding phase, the temporal changes for algal and bacterial communities observed in the present study do not correspond to ecological succession processes occurring over natural biofilm maturation for either the algal (McCormick and Stevenson, 1991) or bacterial (Lyautey et al., 2005) compartments but are related to different algal species and T-RF dominance variations. The primers used in this study have been designed to target specifically the Bacteria domain. Our in silico searches using the RDP database (Cole et al., 2009) indicated that both primers (8F and 1489R) can potentially target Cyanobacteria. Among the in silico targeted organisms, around 5% corresponded to cyanobacteria. Hence, the use of these primers in the current study could over-estimate non phototrophic bacteria richness.
4.2.
Rotating annular bioreactor
4.2.1.
Improvements and advantages compared to other RAB
Previous studies have suggested that the rotating annular bioreactor (RAB) can be an appropriate system to study the effects of various environmental factors on biofilm development (e.g. Neu and Lawrence, 1997; Che´nier et al., 2003). Considering all critical points associated with RAB presented in the literature, the objective of this work was first to design an innovative bioreactor having a modular light source inside the system and second to have a good knowledge of hydrodynamic conditions as assessed using the numerical approach. Computer simulations allow us to confirm a turbulent vortex flow inside the annular gap with the presence of stacked vortices. We observed shear stress and velocity gradients at the scale of the vortices, and the distribution of shear stress described a periodic variation along the height of the bioreactor. As observed by Desmet et al. (1996), the presence of these vortices allows a faster real axial dispersion process than the plug-flow hydrodynamic type, which leads to well mixed liquid phase without of nutriment concentration gradients inside the annular gap, as verified by the tracer method. Variability between plates (n ¼ 3) for the same sampling time was relatively low (30%) for biomass analysis, and their percentage of homology was 60% for T-RFLP analysis, indicating low spatial variability of biofilm colonization and growth inside the RAB. In ecological research, it is necessary to use controlled experiments with large replication to correct for the well known heterogeneity within biofilms (Wimpenny et al., 2000). Our prototype was therefore designed with numerous, large supports associated with a large liquid working volume (5.04 L). In fact, the 32 plates provided a total colonization surface of 0.16 m2 in the present study as against the 12 plates providing 0.0132 m2 of colonization surface in Lawrence et al. (2000), or the 20 plates with 0.00187 m2 in Declerck et al. (2009). We used plates made of polyethylene, suggested to be
applicable to the growth of bacterial (Yu et al., 2010) or phototrophic (Szlauer-Lukaszewska, 2007) biofilms. Their plastic nature and flexibility made them easy to curve so as to fit the external cylinder geometry, thus limiting the disturbance on the flow. The plate fixation design allowed quick and easy sampling without destruction of the sampled biofilms. The geometry, current velocity and continuous culture mode of our prototype made it possible (i) to limit the development of phytoplankton and thus the competition processes between phototrophic biofilm and planktonic biomass, (ii) to limit potential erosion from recirculation of particle or sloughed biofilm fragments, and (iii) to avoid the settling of larger biofilm grazers in such an environment with fast rotation of the water column.
4.2.2.
Shortcomings and potential improvements
The design of our RAB prototype leads to particular operating conditions for the phototrophic biofilm development. First, flow on the plates is produced by the rotation of the inner cylinder, and not directly by the circulation of the water through the system. The consequence is the RAB functioning as a partial closed flow through system without water contact with atmosphere, and with a water residence time of a few hours. The uncoupling between flow velocity on plates and medium flow rate gives unnatural operating conditions. For example, increases of pH (up to 10) or dissolved oxygen concentration (up to oversaturation of 200%) were obtained with daily variations and photosynthetic processes. These conditions may cause temporary inorganic carbon limitation, reactive oxygen damage, and selection of algal and bacterial species. To circumvent these shortcomings, a new version of the RAB must integrate pH control and oxygen stripping, for example with an external loop to prevent the modification of the flow pattern in the bioreactor. Second, the temperature is not controlled inside the bioreactor, and values up to 30 C were obtained at the end of a diurnal period, or when the number of neon tubes was increased. Our RAB contains 3 cylinders, so the best way to control the temperature would be to thermoregulate the atmosphere inside the cylinder containing the neon tubes. Third, it can be suspected that a heterogenous distribution of light inside the RAB occurred, the total number of neon tubes being small and including two types of fluorescent lamps. This technical flaw can be circumvented by using opaque material placed inside the internal water-tight cylinder and in front of the lamps.
4.2.3.
Towards a promising tool
In spite of some improvements needed on this prototype, our study presents the applicability and the performance of a new prototype of rotating annular bioreactor (TayloreCouette flow type) which can be considered as a highly suitable tool for the cultivation, investigation and understanding of a variety of ecological concepts, including the specific richnesseresistance relationship, or the coupling between hydrodynamic level/chemical compounds and structure/function of phototrophic biofilms. As recorded in a previous study in microcosm (Bouleˆtreau et al., 2010), despite a poorly diversified algal community, the phototrophic biofilm exhibited high biomass production. This leads us to wonder about the effect of poor algal diversity
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 0 7 e6 1 1 8
on the biomass production of phototrophic biofilm. A major challenge of the last decade has been to understand the relationship between diversity loss and ecosystem processes (Loreau et al., 2001). Numerous studies have shown that species-rich communities produce more biomass than species-poor communities (Zhang and Zhang, 2006). In future experiments, it could be interesting to assess the stability and resistance of these poorly diversified algal communities obtained in the prototype when a disturbance (e.g. toxic pollutants) is imposed on them. Various works have observed greater sensitivity to disturbance for poorly diversified communities (Zhang and Zhang, 2006). Generally, it is difficult to individualize the main factors influencing epilithic biofilm development and several sources of stresses can have synergistic effects or the inverse. This prototype can bring new perspectives for characterizing the effect of a single factor (e.g. hydrodynamic). Through the ability to modulate the experimental conditions, and by the choice of a particular parameter adapted to algal ecology, the prototype can permit future investigations for the formation and cultivation of artificial biofilms as has recently been reported in the literature (Hayashi et al., 2010).
5.
Conclusion
We propose an improved RAB featuring an embedded modular source of light and the possibility to accurately control the hydrodynamic conditions. These characteristics ensure better control of the operating conditions in comparison with other RABs. Additionally, the larger size of the bioreactor permits numerous samples of biomasses to be taken along the course of experiments to ensure replicates and long term cultures. Further improvements of our RAB version would be beneficial however, including technical solutions for temperature control, homogenous distribution of light inside the system, pH control and oxygen stripping, and operating conditions with a less selective culture medium and a continuous supply of biomass inoculum. Still, our RAB may be useful for the cultivation and experimental study of phototrophic biofilms. This approach is complementary to experimental and observational studies carried out at more complex and realistic scales such as ‘open’ channel and in situ investigations. Hence, RAB-based experiments can make a significant contribution to our understanding of the mechanisms which mediate the structure and functions of phototrophic biofilm communities.
Acknowledgements This work was funded by the French National Programme EC2CO e Environmental Microbiology - and by the Midi-Pyre´ne´es Council Programme of the Pyrenean working community. We are grateful to the ARIAS (Toulouse) company, especially J.-J.Bertrand, for manufacturing the rotating annular bioreactor. We thank J.-L. Druilhe for the electrical device for continuous physical-chemical measurement, S. Karama for assistance with the T-RFLP method, S. Mastrorillo
6117
for field assistance, and E. Mazeau for the computational study of hydrodynamics. We also thank D. Dalger and T. Louis for bioreactor handling assistance, and E. Lyautey for revising the correcting English of the manuscript.
references
Battin, T.J., Kaplan, L.A., Newbold, J.D., Cheng, X., Hansen, C., 2003. Effects of current velocity on the nascent architecture of stream microbial biofilms. Applied and Environmental Microbiology 69 (9), 5443e5452. Besemer, K., Singer, G., Limberger, R., Chlup, A.-K., Hochedlinger, G., Ho¨dl, I., Baranyi, C., Battin, T.J., 2007. Biophysical controls on community succession in stream biofilms. Applied and Environmental Microbiology 73 (15), 4966e4974. Biggs, B.J.F., 1996. Patterns in benthic algae of stream. In: Stevenson, R.J., Bothwell, M.L., Lowe, R.L. (Eds.), Algal Ecology. Freshwater Benthic Ecosystem. Academic Press, San Diego, pp. 31e56. Biggs, B.J.F., Stevenson, R.J., Lowe, R.L., 1998. A habitat matrix conceptual model for stream periphyton. Archiv Fu¨r Hydrobiologie 143 (1), 21e56. Boston, H.L., Hill, W.R., 1991. Photosynthesis-light relations of stream periphyton communities. Limnology and Oceanography 36 (4), 644e656. Bouleˆtreau, S., Garabe´tian, F., Sauvage, S., Sa´nchez-Pe´rez, J.-M., 2006. Assessing the importance of a self-generated detachment process in river biofilm models. Freshwater Biology 51 (5), 901e912. Bouleˆtreau, S., Sellali, M., Elosegi, A., Nicaise, Y., Bercovitz, Y., Moulin, F., Eiff, O., Sauvage, S., Sa´nchez-Pe´rez, J.-M., Garabe´tian, F., 2010. Temporal dynamics of river biofilm in constant flows: a case study in a riverside laboratory flume. International Review of Hydrobiology 95 (2), 156e170. Bourassa, N., Cattaneo, A., 1998. Control of periphyton biomass in Laurentian streams (Que´bec). Journal of the North American Benthological Society 17 (4), 420e429. Bruneel, O., Duran, R., Casiot, C., Elbaz-Poulichet, F., Personne´, J.-C., 2006. Diversity of microorganisms in FeeAsrich acid mine drainage waters of Carnoule`s, France. Applied and Environmental Microbiology 72 (1), 551e556. Characklis, W.G., 1990. Laboratory biofilm reactor. In: Characklis, W.G., Marshall, K.C. (Eds.), Biofilms. John Wiley and Sons, New York, pp. 55e89. Che´nier, M.R., Beaumier, D., Roy, R., Driscoll, B.T., Lawrence, J.R., Greer, C.W., 2003. Impact of seasonal variations and nutrient inputs on nitrogen cycling and degradation of hexadecane by replicated river biofilms. Applied and Environmental Microbiology 69 (9), 5170e5177. Clarke, K.R., 1993. Non-parametric multivariate analyses of changes in community structure. Australian Journal of Ecology 18 (1), 117e143. Cole, J.R., Wang, Q., Cardenas, E., Fish, J., Chai, B., Farris, R.J., Kulam-Syed-Mohideen, A.S., McGarrell, D.M., Marsh, T., Garrity, G.M., Tiedje, J.M., 2009. The ribosomal database project: improved alignments and new tools for rRNA analysis. Nucleic Acids Research 37 (suppl 1), D141eD145. Coufort, C., Bouyer, D., Line´, A., 2005. Flocculation related to local hydrodynamics in a TayloreCouette reactor and in a jar. Chemical Engineering Science 60 (8e9), 2179e2192. Declerck, P., Behets, J., Margineanu, A., van Hoef, V., De Keersmaecker, B., Ollevier, F., 2009. Replication of Legionella pneumophila in biofilms of water distribution pipes. Microbiological Research 164 (6), 593e603.
6118
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 0 7 e6 1 1 8
DeNicola, D.M., 1996. Periphyton responses to temperature at different ecological levels. In: Stevenson, R.J., Bothwell, M.L., Lowe, R.L. (Eds.), Algal Ecology - Freshwater Benthic Ecosystems. Academic Press, San Diego, pp. 31e56. Desmet, G., Verelst, H., Baron, G.V., 1996. Local and global dispersion effects in CouetteeTaylor flow-II. Quantitative measurements and discussion of the reactor performance. Chemical Engineering Science 51 (8), 1299e1309. Gjaltema, A., Arts, P.A.M., van Loosdrecht, M.C.M., Kuenen, J.G., Heijnen, J.J., 1994. Heterogeneity of biofilms in rotating annular reactors: occurrence, structure, and consequences. Biotechnology and Bioengineering 44 (2), 194e204. Hammer, Y., Harper, D.A.T., Ryan, P.D., 2001. Past: paleontological statistics software package for education and data analysis. Palaeontologia Electronica 4 (1), 4e9. Hayashi, S., Jang, J.E., Itoh, K., Suyama, K., Yamamoto, H., 2010. Construction of river model biofilm for assessing pesticide effects. Archives of Environmental Contamination and Toxicology 60 (1), 44e56. Hillebrand, H., Sommer, U., 2000a. Diversity of benthic microalgae in response to colonization time and eutrophication. Aquatic Botany 67 (3), 221e236. Hillebrand, H., Sommer, U., 2000b. Effect of continuous nutrient enrichment on microalgae colonizing hard substrates. Hydrobiologia 426 (1), 185e192. Jackson, C.R., Churchill, P.F., Roden, E.E., 2001. Successional changes in bacterial assemblage structure during epilithic biofilm development. Ecology 82 (2), 555e566. Lane, D.J., 1991. rRNA sequencing. In: Stachenbradt, E. (Ed.), Nucleic Acid Techniques in Bacterial Systematics. Wiley, Chichester, pp. 115e175. Lawrence, J.R., Che´nier, M.R., Roy, R., Beaumier, D., Fortin, N., Swerhone, G.D.W., Neu, T.R., Greer, C.W., 2004. Microscale and molecular assessment of impacts of nickel, nutrients, and oxygen level on structure and function of river biofilm communities. Applied and Environmental Microbiology 70 (7), 4326e4339. Lawrence, J.R., Swerhone, G.D.W., Neu, T.R., 2000. A simple rotating annular reactor for replicated biofilm studies. Journal of Microbiological Methods 42 (3), 215e224. Loreau, M., Naeem, S., Inchausti, P., Bengtsson, J., Grime, J.P., Hector, A., Hooper, D.U., Huston, M.A., Raffaelli, D., Schmid, B., Tilman, D., Wardle, D.A., 2001. Biodiversity and ecosystem functioning: current knowledge and future challenges. Science 294 (5543), 804e808. Lyautey, E., Jackson, C.R., Cayrou, J., Rols, J.-L., Garabe´tian, F., 2005. Bacterial community succession in natural river biofilm assemblages. Microbial Ecology 50 (4), 589e601. McCormick, P.V., Stevenson, R.J., 1991. Mechanisms of benthic algal succession in lotic environments. Ecology 72 (5), 1835e1848. Murdock, J.N., Dodds, W.K., 2007. Linking benthic algal biomass to stream substratum topography. Journal of Phycology 43 (3), 449e460. Neu, T.R., Lawrence, J.R., 1997. Development and structure of microbial biofilms in river water studied by confocal laser scanning microscopy. FEMS Microbiology Ecology 24 (1), 11e25. Osborne, C.A., Rees, G.N., Bernstein, Y., Janssen, P.H., 2006. New threshold and confidence estimates for terminal restriction fragment length polymorphism analysis of complex bacterial
communities. Applied and Environmental Microbiology 72 (2), 1270e1278. Paule, A., Lyautey, E., Garabe´tian, F., Rols, J.-L., 2009. Autogenic versus environmental control during development of river biofilm. Annales de Limnologie-International Journal of Limnology 45 (1), 1e10. Pen˜a-Castro, J.M., Matı´nez-Jero´nimo, F., Esparza-Garcı´a, F., Can˜izares-Villanueva, R.O., 2004. Phenotypic plasticity in Scenedesmus incrassatulus (Chlorophyceae) in response to heavy metals stress. Chemosphere 57 (11), 1629e1636. Roeselers, G., van Loosdrecht, M.C.M., Muyzer, G., 2007. Heterotrophic pioneers facilitate phototrophic biofilm development. Microbial Ecology 54 (3), 578e585. Sekar, R., Nair, K.V.K., Rao, V.N.R., Venugopalan, V.P., 2002. Nutrient dynamics and successional changes in a lentic freshwater biofilm. Freshwater Biology 47 (10), 1893e1907. Singer, G., Besemer, K., Ho¨dl, I., Chlup, A., Hochedlinger, G., Stadler, P., Battin, T.J., 2006. Microcosm design and evaluation to study stream microbial biofilms. Limnology and Oceanography: Methods 4, 436e447. Smith, C.J., Danilowicz, B.S., Clear, A.K., Costello, F.J., Wilson, B., Meijer, W.G., 2005. T-Align, a web-based tool for comparison of multiple terminal restriction fragment length polymorphism profiles. FEMS Microbiology Ecology 54 (3), 375e380. Sugiharto, S., Su’ud, Z., Kurniadi, R., Wibisono, W., Abidin, Z., 2009. Radiotracer method for residence time distribution study in multiphase flow system. Applied Radiation and Isotopes 67 (7e8), 1445e1448. Szlauer-Lukaszewska, A., 2007. Succession of periphyton developing on artificial substrate immersed in polysaprobic wastewater reservoir. Polish Journal of Environmental Studies 16 (5), 753e762. Tison, J., Giraudel, J.L., Coste, M., Park, Y.-S., Delmas, F., 2004. Use of unsupervised neural networks for ecoregional zoning of hydrosystems through diatom communities: case study of Adour-Garonne watershed (France). Archiv Fu¨r Hydrobiologie 159 (3), 409e422. Tlili, A., Dorigo, U., Montuelle, B., Margoum, C., Carluer, N., Gouy, V., Bouchez, A., Be´rard, A., 2008. Responses of chronically contaminated biofilms to short pulses of diuron: an experimental study simulating flooding events in a small river. Aquatic Toxicology 87 (4), 252e263. Utermo¨hl, H., 1958. Zur vervollkommnung der quantitativen phytoplankton methodik. Mitteilungen Internationale Vereinigung fu¨r Theoretische und Angewandte Limnologie 9, 1e38. Weisburg, W.G., Barns, S.M., Pelletier, D.A., Lane, D.J., 1991. 16S ribosomal DNA amplification for phylogenetic study. The Journal of Bacteriology 173 (2), 697e703. Wimpenny, J., Manz, W., Szewzyk, U., 2000. Heterogeneity in biofilms. FEMS Microbiology Reviews 24 (5), 661e671. Yu, J., Kim, D., Lee, T., 2010. Microbial diversity in biofilms on water distribution pipes of different materials. Water Science and Technology 61 (1), 163e171. Zhang, Q.-G., Zhang, D.-Y., 2006. Species richness destabilizes ecosystem functioning in experimental aquatic microcosms. Oikos 112 (1), 218e226. Zippel, B., Neu, T.R., 2005. Growth and structure of phototrophic biofilms under controlled light conditions. Water Science and Technology 52 (7), 203e209.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 1 9 e6 1 3 0
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Advancing post-anoxic denitrification for biological nutrient removal Matt Winkler, Erik R. Coats*, Cynthia K. Brinkman Department of Civil Engineering, University of Idaho, PO Box 441022, Moscow, ID 83844-1022, USA
article info
abstract
Article history:
The objective of this research was to advance a fundamental understanding of a unique
Received 21 April 2011
post-anoxic denitrification process for achieving biological nutrient removal (BNR), with an
Received in revised form
emphasis on elucidating the impacts of surface oxygen transfer (SOT), variable process
2 August 2011
loadings, and bioreactor operational conditions on nitrogen and phosphorus removal. Two
Accepted 3 September 2011
sequencing batch reactors (SBRs) were operated in an anaerobic/aerobic/anoxic mode for
Available online 14 September 2011
over 250 days and fed real municipal wastewater. One SBR was operated with a headspace open to the atmosphere, while the other had a covered liquid surface to prevent surface
Keywords:
oxygen transfer. Process performance was assessed for mixed volatile fatty acid (VFA) and
Post-anoxic denitrification
acetate-dominated substrate, as well as daily/seasonal variance in influent phosphorus
Biological nutrient removal (BNR)
and ammonia loadings. Results demonstrated that post-anoxic BNR can achieve near-
Enhanced biological phosphorus
complete (>99%) inorganic nitrogen and phosphorus removal, with soluble effluent
removal (EBPR)
concentrations less than 1.0 mgN L1 and 0.14 mgP L1. Observed specific denitrification
Surface oxygen transfer
rates were in excess of typical endogenous values and exhibited a linear dependence on
qPCR
the glycogen concentration in the biomass. Preventing SOT improved nitrogen removal but
Polyphosphate accumulating
had little impact on phosphorus removal under normal loading conditions. However,
organisms (PAOs)
during periods of low influent ammonia, the covered reactor maintained phosphorus
Glycogen accumulating organisms
removal performance and showed a greater relative abundance of polyphosphate accu-
(GAOs)
mulating organisms (PAOs) as evidenced by quantitative real-time PCR (qPCR). While GAOs
Secondary phosphorus release
were detected in both reactors under all operational conditions, BNR performance was not adversely impacted. Finally, secondary phosphorus release during the post-anoxic period was minimal and only occurred if nitrate/nitrite were depleted post-anoxically. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
Phosphorus (P) and nitrogen (N) are nutrients of primary concern in regard to accelerated surface water eutrophication, and many wastewater treatment plants (WWTPs) are facing increasingly stringent effluent limitations for both nutrients. P and N can be readily removed biologically, with P removal achieved using an engineered process known as enhanced biological P removal (EBPR). EBPR is accomplished by exposing
microbes to cyclical anaerobic/aerobic and/or anoxic conditions, with influent wastewater first directed to the anaerobic zone. The prescriptive EBPR configuration provides a selective advantage to organisms capable of storing volatile fatty acids (VFAs) anaerobically as polyhydroxyalkanoates (PHAs), such as polyphosphate accumulating organisms (PAOs), which remove and store excess P as intracellular polyphosphate (poly-P) and are the putative organisms responsible for EBPR. EBPR can also enrich for glycogen accumulating organisms
* Corresponding author. Tel.: þ1 208 885 7559; fax: þ1 208 885 6608. E-mail address: [email protected] (E.R. Coats). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.09.006
6120
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 1 9 e6 1 3 0
(GAOs), which do not appear to contribute to EBPR and are therefore considered undesirable. PAOs generate energy for VFA uptake through hydrolysis of intracellular poly-P and glycogen degradation through glycolysis (Smolders et al., 1994), with glycogen degradation considered the main source of reducing power (NADH2) for PHA storage (Zhou et al., 2010). Under aerobic and/or anoxic conditions, PAOs oxidize PHA via the TCA cycle to provide energy for growth, glycogen replenishment, P uptake, and poly-P storage (Smolders et al., 1995). As will be discussed later, Candidatus “Accumulibacter phosphatis” (henceforth referred to as Accumulibacter) has been suggested to be a dominant PAO, based on lab-scale and full-scale studies (He et al., 2007). GAOs exhibit similar metabolisms, with the exclusion of P cycling. Extensive research on factors affecting the PAO-GAO competition can be found elsewhere (Lopez-Vazquez et al., 2009; Oehmen et al., 2010a, 2007). The combination of P and N removal is referred to as biological nutrient removal (BNR). Most BNR WWTPs accomplish denitrification using a pre-anoxic configuration, where the anoxic zone is located upstream of the aerobic zone. Since denitrification relies on ammonia oxidation in an aerobic zone, high mixed liquor recycle (MLR) rates are needed to provide a nitrate source in the anoxic zone. Although high specific denitrification rates (SDNRs) can be obtained with this configuration, there are several disadvantages associated with MLR pumping: higher energy costs, dissolved oxygen (DO) return from the aerobic, and dilution of influent carbon. Most importantly, the removal of oxidized nitrogen (NOx; nitrate þ nitrite) is ultimately limited by the MLR rate, and complete NOx removable is unattainable (estimated 3e5 mg L1 effluent total N) (Tchobanoglous et al., 2003). Post-anoxic denitrification eliminates the need for MLR pumping, since the anoxic tank is located downstream of the aerobic nitrifying tank, and can produce effluent less than 3 mg L1 total N (Tchobanoglous et al., 2003). In a non-EBPR system, an exogenous carbon source is typically supplied to drive denitrification. However, this approach cannot be applied to an EBPR system because the addition of carbon could promote phosphorus release (Kuba et al., 1994) and/or lead to proliferation of ordinary heterotrophic organisms which are incapable of excess P removal. Instead, a dualsludge system is employed to separate the PAO and nitrifying sludges (Bortone et al., 1996; Kuba et al., 1996b). The PAO sludge bypasses nitrification, and intracellular PHA is thus conserved for post-anoxic denitrification. While the dualsludge configuration eliminates the need for MLR pumping, it requires more underflow pumping and a larger footprint due to additional settlers.
Reactor Startup (Da y 0)
Effects of Aeration Rate Low Aeration – Fig. 3a,b,e (Day 58)
An alternative post-anoxic EBPR-based configuration would leverage residual PAO carbon reserves (PHA and/or glycogen) to drive denitrification. In this operating scenario, use of the influent organic carbon and associated electrons is maximized for efficient nutrient removal. Further, this process configuration could produce lower effluent N and P loads as compared to traditional BNR configurations. Promising results have been obtained with lab- and pilot-scale continuous flow membrane bioreactors (Bracklow et al., 2010; Vocks et al., 2005) and lab-scale sequencing batch reactors (SBRs) (Coats et al., 2011b). These systems achieved SDNRs in excess of endogenous rates, and exhibited high N and P removal efficiencies. Recognizing the potential of this novel post-anoxic BNR process, and considering prior work, the research presented and discussed herein focused on understanding the effects of process operation and wastewater loading on N and P removal. New insight on relevant post-anoxic maintenance metabolisms is provided, and the issue of secondary P release is examined. This research also considered the long-term effects of surface oxygen transfer (SOT) on the anaerobic and anoxic aspects of the process, which has not been studied in relation to EBPR. Others have observed impaired SDNRs as result of SOT/microaerophilic conditions in open anoxic basins (Martins et al., 2004; Oh and Silverstein, 1999; Plo´sz et al., 2003), and therefore the issue could be especially relevant for a carbon-limited post-anoxic environment. The research also interrogated respective microbial consortia on PAO and GAO fraction. The research presents results based on the use of real municipal wastewater rather than the much more common approach of using synthetic wastewater.
2.
Materials and methods
2.1.
Experimental setup
Two independent SBRs (reactors O and C) were operated for over 250 days, with monitoring events as shown (Fig. 1). The reactors were operated identically except for the headspace condition; reactor O had a headspace open to the atmosphere, while reactor C had zero headspace due to a liquid surface covered with a polyethylene disk. Thus, SOT could occur during the anaerobic and anoxic periods in reactor O but not in reactor C. Note that for all of the figures in this manuscript, reactor O is represented by open symbols, while reactor C is denoted by filled symbols. Each SBR (0.9 L operating volume) was inoculated with activated sludge obtained from the Moscow, ID WWTP, which operates a hybrid A2/O-oxidation
Effects of Aeration Rate High Aeration – Fig. 3c,d,e Extended Anoxic – Fig. 6 (Day 88)
DNA 1 Day 0
Day 50 Air at 0.3 L/min 90:10 Feed
Day 100
Wastewater Composition Acetate Study – Fig. 4a,b,c (Day 142)
Wastewater Composition Low NH3 Study – Fig. 4d,e (Day 207)
DNA 2
DNA 3
DNA 4 DNA 5 DNA 6
Day 150
Day 200
Day 250
Air at 1.0 L/min Raw WW + HAc
90:10 Feed
Fig. 1 e Research timeline showing time points for all sampling investigations in this study (aeration rate and substrate are provided below the timeline).
6121
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 1 9 e6 1 3 0
Wastewater source and composition
Screened and de-gritted raw wastewater was obtained weekly from the Moscow, ID WWTP and stored at 4 C in polyethylene jugs until use. Wastewater was filtered through cheesecloth prior to daily batching of feed. To supply additional VFAs, the raw wastewater was mixed with either VFA-rich fermenter liquor (90:10; 90% raw wastewater and 10% fermenter liquor by volume) or 10 mL per L of a concentrated sodium acetate solution which supplied an additional 200 mgCOD L1 to the feed (raw WW þ HAc), as illustrated in Fig. 1. VFA-rich fermenter liquor was recovered from a laboratory fermenter fed with primary solids from the Pullman, WA WWTP (Coats et al., 2011b). The VFA distribution in the 90:10 feed averaged 52 5% acetate (HAc), 34 6% propionate (HPr), 12 2% butyrate (HBu), and 5 1% valerate (HVa) on a Cmol basis; nominal amounts of iso-HBu and iso-HVa were observed as well. Influent P varied from 3.5 to 6.0 mgP L1, and influent ammonia concentrations ranged from 14 to 50 mgN L1. The reactors received the 90:10 substrate for a majority of the study (Fig. 1), and an example of the daily variations in the 90:10 characteristics are shown in Fig. 2 for about 30 days prior to the first sampling run. During this time period, total VFAs ranged from 200 to 275 mgCOD L1 which yielded an influent VFA:P ratio of 35e50 mgCOD mgP1, which was theoretically sufficient to be favorable for PAOs (Oehmen et al., 2007). Reactors were allowed to stabilize for about 3 SRTs between operational changes before performance was assessed. Table 1 provides additional information about the influent wastewater characteristics for each sampling event.
2.3.
Stoichiometric calculations
Due to the SBR configuration and post-anoxic operational mode, residual NOx carryover from the anoxic period to the subsequent anaerobic cycle occurred. While the amount of carryover was insufficient to upset EBPR performance, some of the influent VFAs would have been consumed to reduce the residual NOx (mainly nitrate). Much debate exists regarding how much influent COD would be consumed for nitrate reduction, and others have commonly assumed a ratio of 8.6 mgCOD mgNO3eN1 (Henze et al., 2008). However, this
15
60 Feed P
Feed NH3-N
PO4-P (mg L-1)
40 9 30 6 20 3
10
0 7/26
b
NH3-N (mg L-1)
50
12
0 7/31
8/5
8/10 8/15 Date
8/20
8/25
8/30
400
60 Feed VFAs
VFA:P 50
300 40 30
200
20 100 10 0 7/26
VFA:P (mgCOD mgP-1)
2.2.
a
VFAs (mgCOD L-1)
ditch process (Tchobanoglous et al., 2003). Each 6 h operational cycle consisted of 1 h anaerobic, 2 h aerobic, 2.5 h anoxic, and 0.5 h for settling/decanting. At the beginning of each cycle, 0.3 L of wastewater was fed to each reactor, resulting in a hydraulic retention time (HRT) of 18 h. The solids retention time (SRT) was maintained at 20 days by wasting 45 mL of mixed liquor at the end of one anoxic period each day. pH was not controlled, and varied between 7.0 and 7.3, 7.3e7.6, and 7.3e7.6 during the anaerobic, aerobic, and anoxic periods, respectively. The air flow rate was controlled at either 0.3 or 1.0 L min1 during the aerobic period (Fig. 1; referred to as low and high aeration studies). Mixing was accomplished using magnetic stirrers and feeding/decanting was performed with peristaltic pumps. All operations were controlled by a bench-top programmable logic controller (PLC). The reactors were operated in a temperature controlled room which resulted in average reactor temperatures of 23 3 C.
0 7/31
8/5
8/10
8/15
8/20
8/25
8/30
Date
Fig. 2 e Typical influent characteristics for (a) P and ammonia, and (b) VFAs and VFA:P ratio at the beginning of this study. The dashed lines show each time that a new batch of wastewater was collected.
ratio likely overestimates COD utilization because it assumes an anoxic yield equal to the typical aerobic heterotrophic yield value of 0.666 mgBiomassCOD (mgCODutilized)1. Others have suggested that the anoxic yield is 60e70% of the aerobic yield and also dependent on the substrate provided (Copp and Dold, 1998). In this study, residual nitrate at the beginning of the anaerobic period was low and available for less than 5 min, which would also limit the anoxic yield. Assuming that acetate was the preferred VFA for denitrification (Elefsiniotis and Wareham, 2007) and that the anoxic yield on acetate was 0.192 mgBiomassCOD (mgCODutilized)1 (Copp and Dold, 1998), a ratio of 3.54 mgCOD mgNO3eN1 was obtained. Accordingly, all EBPR stoichiometric VFA ratios were calculated assuming that 3.54 mgCOD mgNO3eN1 and 1.71 mgCOD mgNO2eN1 were utilized for nitrate and nitrite reduction, respectively (Tchobanoglous et al., 2003). Given that the NOx residual was small in comparison to the influent VFAs, this assumption caused minimal variation in the ratios.
2.4.
Analysis
2.4.1.
Chemical analyses
All soluble constituents were filtered through a 0.22 mm Millex GP syringe-driven filter (Millipore, MA, USA). Phosphate (PO4eP) and nitrate (NO3eN) were determined colorimetrically
6122
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 1 9 e6 1 3 0
Table 1 e Influent characteristics for all sampling investigations in this study. Influent characteristicsa
Study type
PO4eP (mg L1) NH3eN (mg L1) HAc HPr HBu HVa Total VFAs Influent VFA:P Influent VFA:NH3eN Low Aeration High Aeration Acetate Low NH3 - Cycle 1 Low NH3 - Cycle 2
5.97 5.48 3.87 4.42 4.99
42.4 34.2 35.9 14.6 16.0
3.27 3.72 5.62 2.95 3.45
2.62 3.14 0.57 1.53 1.79
0.91 0.72 0.00 0.55 0.00
0.39 0.41 0.00 0.21 0.00
7.28 8.24 6.19 5.24 5.24
43 54 52 41 36
6.1 8.6 5.6 12.5 11.1
a VFA units reported as Cmmol L1; VFA:P and VFA:N are mgCOD mgP1 and mgCOD mgN1, respectively.
(Coats et al., 2011b). Ammonia (NH3eN) and nitrite (NO2eN) were measured in accordance with Hach (Loveland, CO, USA) methods 10031 and 10019, respectively. MLSS and MLVSS were analyzed according to the standard methods (APHA, 1998). pH was monitored using American Marine (Ridgefield, CT, USA) Pinpoint pH controllers. VFAs were measured using a gas chromatograph (GC) equipped with a flame ionization detector (Coats et al., 2011a). Glycogen was determined with dried biomass samples as described by Parrou and Francois (1997); biomass samples were washed with a 1% NaCl solution prior to analysis to minimize potential interference with exopolysaccharide (EPS), the latter being a source of slowly biodegradable carbon for bacteria. Intracellular PHA was quantified using a GC equipped with a mass spectrometer (Coats et al., 2011b). Surface oxygen mass transfer coefficients (KLaSUR) were determined using room temperature tap water according to the dynamic gassing out method of Van’t Riet (1979), where saturation DO concentrations were estimated using temperature and pressure correction equations (APHA, 1998). KLaSUR values were calculated by minimizing the total sum of square errors between the predicted DO from the mass transfer model and the actual DO measurements. Both DO and temperature were recorded simultaneously using a Hach HQ30d Meter and LDO101 DO Probe. The volumetric oxygen mass transfer coefficient, KLaSUR, in reactor O was measured at 0.54 h1.
2.4.2.
Microbial population analyses
Genomic DNA was extracted from each reactor on the dates shown (Fig. 1) according to the procedure outlined in Coats et al. (2011c). Quantitative real-time PCR (qPCR) was used to quantify 16S rDNA genes from total bacteria, Accumulibacter
(the model PAO), and GAOs to provide an estimation of relative abundance. qPCR was conducted on a StepOne Plus Real-Time PCR system (Applied Biosystems, Foster City, CA) using iTaq SYBR Green Supermix w/ROX (Bio-Rad Laboratories, Inc., Hercules, CA, USA) with a total reaction volume of 25 ml. Total bacterial and total Accumulibacter 16S rDNA genes were quantified with primer sets 341f/534r and 518f/846r, respectively (He et al., 2007). GAOs were quantified using primer set GAOQ431f/GAOQ989r (specifically designed to target Candidatus Competibacter phosphatis, which is a putative model GAO (Crocetti et al., 2002)) and the total bacteria primer set. In addition, a primer set targeting the GB lineage (specifically GB612f/GAOQ989r (Kong et al., 2002), coupled with the total bacteria primer set) was employed to quantify Gammaproteobacteria. The GB lineage, also referred to as the Competibacter lineage, is proposed to capture the predominant GAOs within the class Gammaproteobacteria that would be present in EBPR WWTPs (Kong et al., 2006; Oehmen et al., 2007). qPCR conditions were as follows: 3 min at 95 C, 45 cycles of 30 s at 95 C, 45 s at 60 C, and 30 s at 72 C. All unknown samples were assessed in triplicate with 5 ng of total genomic DNA per reaction. Amplification efficiency was estimated for each primer set using baseline-corrected fluorescence data (from StepOne Software v2.0) with LinRegPCR (Ramakers et al., 2003). For PAO quantification, mean amplification efficiencies for the total bacterial and PAO primer sets were 96.8 1.8% (n ¼ 37) and 93.9 1.5% (n ¼ 36), respectively. For GAO and GB lineage quantification, mean amplification efficiencies for the respective primer sets were 85.2 3.1% (GAOs; n ¼ 76), 93.6 3.4% (GB lineage; n ¼ 76), and 93.9 3.0% (n ¼ 76). The cycle threshold was set at a constant value across all samples based on location within the log-linear region for
Table 2 e Anaerobic biochemical transformations and P removal performance. Study type
Feed
Low Aeration
90:10
High Aeration
90:10 HAc
Reactor name O C O C O C
MLVSS (mg L1) 3560 3770 5060 5470 1520 1996
EBPR anaerobic stoichiometry (normalized to VFAs)a Prel
Gly
PHB
PHV
PH2MV
PHA
0.29 0.29 0.20 0.21 0.15 0.16
0.18 0.40 0.68 0.50 0.71 0.61
0.36 0.36 0.28 0.30 0.84 0.93
0.53 0.46 0.73 0.81 0.27 0.33
0.12 0.12 0.16 0.14 0.00 0.00
1.02 0.94 1.17 1.24 1.11 1.18
a All ratios reported as Cmol Cmol1 except for Prel/VFA (Pmol Cmol1). b Aerobic P uptake rate reported as mgP gVSS1 h1; Effluent PO4eP reported as mg L1.
P uptake rateb
Effluent PO4ePb
13.6 8.8 25.7 22.0 7.4 9.1
0.08 0.10 0.09 0.14 0.02 0.03
6123
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 1 9 e6 1 3 0
determination of Cq values (cycle number at which the measured fluorescence exceeds the cycle threshold). Relative PAO abundance (Table 3) was estimated using the mean efficiencies for each primer set and the Cq values for the individual samples, assuming that the average 16S rDNA gene copy number is 4.1 per bacterial cell and 2 per Accumulibacter cell (He et al., 2007). For relative and comparative GAO and GB lineage abundance between sampling events, lacking a mapped genome, an average 16S rDNA gene copy number of 4.1 per bacterial cell was assumed for the targeted populations. Gel electrophoresis of qPCR products confirmed the presence of a single band for all GAO and PAO samples.
3.
Results and discussion
Two SBRs were operated continuously for over 250 days (Fig. 1). Numerous investigations were conducted on each reactor over the course of this time period in order to understand core operating fundamentals associated with this BNR process. Results from these investigations are presented and discussed below.
3.1.
Effects of aeration rate on process performance
Considering that aeration accounts for upwards of 70% of the total energy cost at biological WWTPs (Cornel et al., 2003), minimizing air requirements represents a significant cost savings opportunity. Thus, performance was assessed at both low and high aeration rates during the normal reactor cycle on the dates shown (Fig. 1). Our process configuration demands oxygen principally for nitrification, because incomplete ammonia oxidation would potentially impair the post-anoxic based process configuration. For the low aeration study, the DO remained below 1 mg L1 for the first 1.5 h before increasing to 6.25 mg L1 near the end of the aerobic period (Fig. 3a). In contrast, the high aeration study showed a rapid increase in the DO profile, surpassing 2 mg L1 within the first 7 min of aeration (Fig. 3c). DO profiles were only measured for the open reactor. The aeration, influent, and effluent piping were hard-connected through the closed reactor cover, and we elected not to remove the cover to measure DO in order to maintain the integrity of our experimental design. Although the closed reactor exhibited higher MLVSS (Table 2), which would have imposed a higher oxygen
demand, considering that the applied aeration rate was the same for both reactors the DO profiles in the closed reactor were assumed to be comparable to that measured in the open reactor. The observed removal of ammonia, VFAs, and phosphorus during the aerobic period (Fig. 3) supports this assumption.
3.1.1.
Phosphorus removal
Excellent P removal was achieved in both reactors regardless of the aeration rate (Fig. 3a, c; Table 2), and P cycling was consistent with current EBPR theory (Smolders et al., 1994). At each aeration rate, the anaerobic stoichiometric P release (Prel/ VFA; Pmol Cmol1) was identical between reactors (Table 2), suggesting negligible impact of SOT on anaerobic EBPR metabolisms. However, comparatively higher Prel/VFA ratios were observed at the low aeration rate. Zhang et al. (2008) reported Prel/VFA ratios in the range of 0.43e0.51 for their PAO-enriched SBRs fed a mixture of acetic and propionic acids; similarly, PAO-enriched cultures exhibited Prel/VFA ratios of 0.48 and 0.42 for acetate (Smolders et al., 1995) or propionate (Oehmen et al., 2005), respectively. Comparatively, the ratios observed in this study were markedly lower. Current theory would suggest that the decrease in Prel/VFA ratios and the corresponding increase in anaerobic glycogen degradation between aeration rates (Table 2) indicates a population shift toward GAOs at the higher aeration rate (Schuler and Jenkins, 2003). However, in both reactors under high aeration conditions Accumulibacter abundance was estimated to be relatively high at 24e25%, while the GAO fraction was lower (2.8e5.4%; DNA1, Table 3); DNA was not available to comparatively characterize the population at the low aeration rate. Since both configurations achieved excellent P removal, clearly the respective consortia were sufficiently enriched with PAOs. The lower Prel/VFA ratios, and the associated decrease in this ratio, would appear to have been influenced more by the use of real wastewater, which is inherently more complex than synthetic wastewater used in most EBPR studies. At the low aeration rate, biomass in both reactors accumulated comparable amounts of PHA anaerobically, producing PHB, PHV, and PH2MV (Table 2). A similar PHA composition was observed by Zhang et al. (2008) for their PAOenriched SBR fed an equal concentration of acetate and propionate. At the high aeration rate, PHV synthesis increased moderately despite nearly identical influent VFA fractions
Table 3 e Relative fraction of PAOs (n [ 3) and GAOs (n [ 6; except DNA5 n [ 8) within the bacterial community as estimated by qPCR for the DNA extraction dates shown in the timeline (Fig. 1). Sample ID
PAOs
GAOs GAO primer set
DNA1 DNA2 DNA3 DNA4 DNA5 DNA6
Reactor O
Reactor C
1.4 0.06 0.09 0.11 0.51 0.24
24.1 1.8 1.5 0.11 1.5 0.12 12.4 0.11 12.5 0.54 17.0 1.53
24.5 1.0 0.9 4.1 8.7 4.7
Reactor O 2.8 5.3 2.4 0.3 2.6 0.2
0.7 1.3 0.3 0.04 0.6 0.1
Reactor C 3.5 0.2 6.0 0.5 11.5 1.0 3.5 0.3 3.6 0.5 3.4 0.2
GB lineage Reactor O 3.6 10.4 12.9 3.4 14.9 3.5
0.2 0.8 1.0 0.6 1.8 0.9
Reactor C 5.4 11.4 46.6 8.9 15.2 10.7
0.8 1.7 3.6 0.4 2.4 0.7
6124
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 1 9 e6 1 3 0
C-P
12
O-VFA C-VFA
8
3
O-PHA
4
C-PHA O-DO
0 1
2
3 4 Time (hr)
5
24 AN
9
PO4-P (mg L -1)
16
6 O-P C-P O-VFA C-VFA
8
3
O-PHA C-PHA
4
O-DO
0
0 0
1
2
3
4
10
5
Glycogen (Cmmol L -1)
3 2
6 4
1
2 0
0
d
16
1
2
AN
3 Time (hr)
4
5
6
5
AX
AE
14
O-NH3 C-NH3 O-NO3
12 10
4 3
C-NO3 O-NO2 C-NO2
8
2
6 4
1
2 0
6
0 0
1
2
Time (hr)
e
4
8
0
20
12
12
6
AX
AE
O-NH3 C-NH3 O-NO3 C-NO3 O-NO2 C-NO2
14
0 0
5
AX
NO2-N (mg L-1)
O-P
AE
AN
NO2-N (mg L-1)
6
16
16
NH3-N, NO3-N (mg L-1)
PO4-P (mg L -1)
20
c
b
9
AX
AE
NH3-N, NO3-N (mg L-1)
AN
VFAs, PHA (Cmmol L -1) Dissolved Oxygen (mg L -1)
24
VFAs, PHA (Cmmol L -1) Dissolved Oxygen (mg L -1)
a
3
4
5
6
Time (hr)
12
AN
AX
AE
O (high DO) C (high DO) O (low DO) C (low DO)
10
8
6 0
1
2
3
4
5
6
Time (hr)
Fig. 3 e Effects of aeration rate on process performance. Low aeration rate cycle profiles for (a) P, PHA, and DO, and (b) ammonia, nitrate, and nitrite. High aeration rate cycle profiles for (c) P, PHA, and DO, and (d) ammonia, nitrate, and nitrite. Glycogen profiles (e) for both low and high aeration studies.
(Table 1); as noted, consortia in both reactors also exhibited increased Gly/VFA ratios, which likely contributed to the greater PHV fraction (Table 2). While both PAOs (Lopez et al., 2006; Lu et al., 2007) and GAOs (Liu et al., 1994) have been suggested to produce PHV from glycolysis end products and the propionate-succinate pathway, this is more commonly considered a GAO metabolism. As shown (Table 3), GAOs were present in both respective consortia.
3.1.2.
Nitrogen removal
Total nitrogen removal at the low aeration rate (ammonia, nitrate, and nitrite) was 85% for reactor O and 90% for reactor C, while at the high aeration rate reactors O and C achieved 97% removal and 100% removal, respectively (Table 4; relative to influent ammonia, Table 1). Although reactor C did not achieve complete nitrification during the low aeration rate study (Fig. 3b; Table 4), preventing SOT moderately improved overall N removal at both aeration states. Despite the
theoretical transfer of approximately 10.8 mgO2 L1 during the denitrification reaction period, reactor O maintained comparable SDNRs to reactor C. The most likely explanation for higher effluent NOx in reactor O (low aeration) was reduced intracellular glycogen (Fig. 3e). Recognizing that PHA is the preferred aerobic carbon source of PAOs, glycogen would most commonly be the only intracellular carbon source remaining post-anoxically that could potentially drive denitrification. This, in fact, was observed for both the low and high aeration studies (Fig. 3e) and is consistent with similar studies (Coats et al., 2011b; Vocks et al., 2005). All SDNRs (Table 4) were in excess of typical values for endogenous decay, which have been shown to range from 0.2 to 0.6 mgNO3eN gVSS1 h1 (Kujawa and Klapwijk, 1999), and the results suggest that PAO maintenance metabolisms were driving denitrification (see sections 3.2.2 and 3.3). Similar observations supporting PAO maintenance energy production via glycogen oxidation with nitrate
6125
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 1 9 e6 1 3 0
Table 4 e N removal performance and denitrification rates during post-anoxic period. Study type
Feed
Low Aeration
90:10
High Aeration
90:10 HAc
Effluent nitrogen (mgN L1)
Reactor name
O C O C O C
Post-AX characteristicsa
NH3eN
NO3eN
NO2eN
NOx
T ( C)
SDNR
SDNR20
0.20 3.59 0.03 0.01 0.07 0.05
6.04 0.29 0.56 0.02 8.08 5.73
0.23 0.20 0.31 0.01 0.00 0.43
6.27 0.49 0.87 0.03 8.09 6.16
24.4 24.4 25.5 25.5 22.7 22.7
0.71 0.90 1.09 1.09 0.33 0.62
0.63 0.80 0.95 0.95 0.31 0.58
a SDNR values reported as mgNO3eN gVSS1 h1.
have been reported by others (Coats et al., 2011b; Lu et al., 2007; Vocks et al., 2005).
3.2.
Wastewater composition and process performance
3.2.1.
Acetate effects
Propionate has been postulated to be a more favorable EBPR carbon source than acetate (Oehmen et al., 2007). In this regard, the wastewater used in our study contained significant propionate, as indicated by the HPr-to-total VFA ratio of approximately 0.36e0.38 Cmol Cmol1, and excellent P removal was observed. However, since the majority of EBPR studies use acetate as the sole carbon source, it was of interest to determine if process performance could be maintained with acetate-dominated feed and how a VFA switch might affect the consortia. The reactors were switched from the 90:10 substrate to the “raw WW þ HAc” feed (Fig. 1), where acetate generally comprised greater than 90% of the influent VFA fraction (Table 1; Cmol basis). Although Prel/VFA ratios decreased moderately, both reactors achieved excellent EBPR, with the lowest effluent P observed in this study (Fig. 4a, b; Table 2). However, the approximately 71% reduction for reactor O and 59% reduction for reactor C in aerobic P uptake rate (Table 2) does suggest a potential population shift associated with the dominant acetate substrate. Competibacter, for example, has been shown to prefer acetate as a carbon source and could compete with PAOs at the anaerobic pH range realized in this study (Lopez-Vazquez et al., 2009). Interestingly, as shown (Table 3; DNA2) the Accumulibacter abundance did decrease more than 10-fold in both reactors after switching to the acetate feed. Further, note that the percentages remained relatively low even 3 SRTs after transitioning back to the 90:10 feed (i.e., DNA3). Conversely, the relative GAO abundance, as measured by both the GAO primer set and the GB lineage, increased substantially in both reactors (Table 3; DNA2). These observations generally align with current EBPR theory, which would suggest that the decrease in Prel/ VFA ratios and corresponding increase in Gly/VFA ratios associated with the shift in substrate (Table 2) is indicative of a shift toward GAOs (Schuler and Jenkins, 2003). Nevertheless, recognizing that P removal remained consistently high in both reactors, it would appear that the Accumulibacter population was sufficient for EBPR success. Alternately, other PAOs not targeted by the Accumulibacter primers were present, or perhaps some of the GAOs performed EBPR.
Nitrogen cycling in the acetate-augmented reactors was generally similar to that observed in the 90:10 wastewater reactors (Fig. 3d vs. Fig. 4b; Table 4), although total inorganicN removal was 77% for reactor O and 83% for reactor C, the lowest values observed in this study. MLVSS concentrations also decreased from over 5000 mg L1 to less than 2000 mg L1. As contrasted with the 90:10 wastewater-fed reactors (Fig. 3), the synthetic acetate-augmented substrate generated lower intracellular glycogen reserves, and less glycogen was used anoxically (Fig. 4c). Notably, the SDNR in reactor O approached endogenous levels, suggesting that SOT may have been detrimental to process performance (Fig. 4b). It is also possible that the combined effects of SOT and an acetaterich substrate enriched for a consortium less efficient at, or less capable of, denitrification. Regarding PAOs, as noted the population did decrease quite substantially (Table 3). In this regard, Carvalho et al. (2007) suggested that acetate fed SBRs are more likely to select for an Accumulibacter strain that can use nitrite but not nitrate as electron acceptor (nitriteDPAOs). Further, Martin et al. (2006) determined that Accumulibacter may not harbor the metabolic capability to reduce nitrate (but could reduce nitrite), and suggested that other members of the microbial consortium provided nitrite. Finally, Flowers et al. (2009) determined that certain clades within the model PAO quantified herein could not readily use nitrate. In addition to potential metabolic limitations with PAOs, it could alternately be suggested that the increase in GAOs could have contributed to reduced denitrification capacity, since not all of the subgroups in the GB lineage can denitrify (Kong et al., 2006; Oehmen et al., 2007).
3.2.2. Post-anoxic denitrification using slowly biodegradable carbon Real wastewater contains a complex mixture of slowly- and readily-biodegradable carbon sources, whereas synthetic wastewater commonly used in most EBPR and BNR research contains 100% readily biodegradable carbon (i.e., pure VFAs). In this regard, the presence of residual slowly-biodegradable carbon could have been partially responsible for increasing post-anoxic denitrification rates beyond typical endogenous values (although the aerobic period could theoretically induce measurable oxidation of this substrate). To understand if the raw wastewater or the fermenter liquor was providing slowly-biodegradable carbon (in bulk solution; not as EPS) for post-anoxic denitrification, reactor performance was observed over two consecutive cycles, where the feed for
6126
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 1 9 e6 1 3 0
O-P C-P O-VFA C-VFA O-PHA C-PHA O-DO
6
3
16
6
1
AN
2
3 4 Time (hr)
5
O-NH3 O-NO3 O-NO2
12
C-NH3 C-NO3 C-NO2
4
10
3
8 2
6 4
1 0
6
0 0
d
1
2
3 4 Time (hr)
90:10 WW 12
Synthetic WW
AE
AN
5
AN
AX
6
NO3 Addition
AE
AX
10 PO4-P (mg L-1)
Glycogen (Cmmol L-1)
5
AX
2
AX
AE
AE
AN
14
0 0
c
b
9
AX
AE
NO2-N (mg L-1)
AN
NH3-N, NO3-N (mg L-1)
22 20 18 16 14 12 10 8 6 4 2 0
VFAs, PHA (Cmmol L-1) Dissolved Oxygen (mg L-1)
PO4-P (mg L-1)
a
4
O-Gly
8 O-P
6
C-P
4 2
C-Gly
0
2 0
1
2
3 4 Time (hr)
e
5
90:10 WW 16
NO3-N, DO (mg L-1)
0
6
AN
2
Synthetic WW
AE
AX
AN
4
6 Time (hr)
8
10
12
NO3 Addition
AE
AX
12
8 ONO3 CNO3
4
0 0
2
4
6 Time (hr)
8
10
12
Fig. 4 e Effects of wastewater composition on process performance. Acetate study cycle profiles for (a) P, PHA, and DO, and (b) ammonia, nitrate, and nitrite, and (c) glycogen. Low ammonia study cycle profiles for (d) P, and (e) nitrate and DO over two consecutive cycles, where the feed for the first cycle (0e6 h) was 90:10 and the feed for the second cycle (6e12 h) was a synthetic WW with similar composition.
the first cycle was the 90:10 wastewater mixture (5.24 Cmmol L1 VFAs) and the feed for the second cycle was entirely a synthetic wastewater (i.e., no bulk solution slowly biodegradable carbon provided). The synthetic wastewater was designed to contain the same influent P, N, and VFAs (3.45 Cmmol L1 HAc and 1.79 Cmmol L1 HPr) as the 90:10 substrate, with other nutrient concentrations provided as described by Kuba et al. (1996a). Of note, this experiment coincided with a period of low ammonia in the real wastewater (further discussed in section 3.2.3). Due to the low influent ammonia (14.6 mgN L1), minimal nitrate was produced during the first aerobic cycle. To ensure a sufficient anoxic period in the second cycle, each reactor was spiked with an additional 11.1 mgNO3eN L1 at the end of the second aerobic period. As shown (Fig. 4d, e), there was no apparent difference in performance when the consortia were supplied real vs.
synthetic wastewater. The SDNRs for the first cycle were 0.47 and 0.80 mgNO3-N gVSS1 h1 for reactors O and C, respectively, and 0.39 and 0.82 mgNO3eN gVSS1 h1 for reactors O and C in the second cycle, respectively. These comparable results suggest that residual slowly-biodegradable carbon present in the substrate was not a significant source of electrons in post-anoxic denitrification. However, we cannot rule out the contributions of EPS and/or soluble microbial products to the increased SDNRs.
3.2.3.
Effects of influent ammonia on post-anoxic BNR
Nutrient concentrations in real wastewater will vary over time, particularly with excess precipitation. The influent ammonia concentration is particularly important to this postanoxic configuration because it determines anoxic nitrate availability and associated process metabolisms and stability. As evidenced by the range of ammonia concentrations
6127
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 1 9 e6 1 3 0
observed in this study (Fig. 2a), influent ammonia loads to a WWTP will vary. Coats et al. (2011b) suggested that the influent VFA:NH3eN ratio (mgCOD mgN1) could be an important predictor of process success, where a VFA:NH3eN ratio greater than 4.1 was desirable. A higher VFA:NH3eN ratio typically indicates that sufficient VFAs are available to drive EBPR and that the influent ammonia is not so high as to create excessive nitrate carryover. However, the guideline does not address cases where the ratio is inflated not by high VFAs but by low influent ammonia concentrations (<20 mgN L1), which can lead to nitrate depletion before the end of the anoxic period and potentially induce secondary P release (Barnard and Fothergill, 1998). As discussed in section 3.2.2, reactor performance was observed over consecutive cycles during a period of low ammonia loading; the VFA:NH3eN ratio was 12.5. While both systems maintained excellent N removal, EBPR efficiency was impaired in the open air reactor (Fig. 4d, e). Prel/VFA ratios for reactors O and C were 0.12 and 0.17 for the first cycle and 0.10 and 0.14 for the second cycle, comparable to those observed when the reactors were receiving acetateaugmented substrate. However, incomplete aerobic P uptake was observed in reactor O (Fig. 4d) and the associated aerobic P uptake rate was only 1.4 mgP gVSS1 h1, markedly less than observed with the acetate and 90:10 substrates (Table 2). In contrast, the consortium in reactor C removed all P from solution aerobically for both cycles (Fig. 4d), although the consortium exhibited a slow secondary release (approximately 0.22 mgP gVSS1 h1) if nitrate was depleted anoxically. Similar trends were observed over the remainder of this study. Accumulibacter and GAO abundance was evaluated on four dates while the influent ammonia remained below 20 mgN L1 (Fig. 1; DNA 3, 4, 5, and 6). The end-aerobic P concentrations on these dates averaged 2.4 and 0.20 mgP L1 for reactors O and C, respectively. The fractional PAO population in both reactors increased substantially from DNA3 to DNA4, and then remained relatively steady for the rest of this time period (Table 3). Of note, the qPCR and effluent data indicated that reactor C was consistently enriched for more PAOs than in reactor O. Further, the GAO population also remained relatively high in the closed reactor (although decreasing substantially from the estimated peak when DNA3 was collected), as compared with the open reactor. Overall, these results suggest that limiting the extent of SOT could improve process stability, and also that the relative significant presence of GAOs will not necessarily impair phosphorus removal.
3.3.
Glycogen and post-anoxic denitrification
Glycogen, and to a lesser extent PHA, have been hypothesized by others to be important carbon sources driving post-anoxic denitrification (Vocks et al., 2005), as contrasted with endogenous decay. As discussed herein, further evidence supporting the use of glycogen for denitrification has been observations of glycogen utilization during the anoxic period and SDNRs in excess of typical endogenous rates. However, to better quantify the potential involvement of glycogen, minimum denitrification carbon requirements were estimated (assuming 2.86 mgCOD mgNO3eN1 and no growth) and compared with the measured carbon utilization (Table 5). For the quantity of nitrate reduced, measured glycogen utilization accounted for 62e76% and 55e62% of the minimum carbon requirements for the low and high aeration studies with the 90:10 wastewater, respectively. Conversely, glycogen utilization in excess of the carbon requirements was measured for both reactors during the acetate study. The excess glycogen utilization observed in these latter configurations (in particular for reactor O, which also reduced very little nitrate) could have been associated with the shift in population toward GAOs (Table 3). As noted, not all subgroups of the GB lineage can denitrify (Kong et al., 2006). Thus, when anoxic conditions were imposed, certain GAOs would have used their glycogen reserves not for nitrate reduction but for maintenance and survival, resulting in excess glycogen consumption. As further support for the involvement of glycogen in the proposed post-anoxic process, we compared SDNRs vs. glycogen for data from this study and values from Coats et al. (2011b) (Fig. 5). The SDNRs from this study were corrected to a temperature of 20 C (SDNR20) using the average anoxic temperature and an Arrhenius temperature correction coefficient, q ¼ 1.026 (Tchobanoglous et al., 2003). The SDNR values from Coats et al. (2011b) were not corrected for temperature because no temperature values were reported. As shown, the data shows that SDNRs increase with glycogen content, especially when considering each of the studies separately.
3.4.
Secondary P release considerations
One concern with our process configuration is the potential for secondary P release during the anoxic period, which could lead to elevated effluent P concentrations (Barnard and Fothergill, 1998). Specifically, process success hinges on
Table 5 e Theoretical carbon source requirements for post-anoxic denitrification. Study type
Feed
Low Aeration
90:10
High Aeration
90:10 HAc
Reactor name O C O C O C
NO3eN reduced
COD required
5.61 6.88 10.18 9.49 0.48 2.94
16.0 19.7 29.1 27.1 1.4 8.4
All units are mgCOD L1 except for NO3eN (mgN L1).
Internal carbon sources used Gly
PHB
PHV
PH2MV
Total
12.2 12.3 17.9 15.0 8.2 10.1
0.4 0.6 0.3 0.4 0.0 0.0
0.4 1.0 0.6 0.7 0.0 0.0
0.0 0.0 0.0 0.0 0.0 0.0
12.9 13.9 18.8 16.1 8.2 10.1
6128
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 1 9 e6 1 3 0
1.4
SDNR20, SDNR (mgN gVSS-1 hr -1)
1.2 1.0 0.8 0.6 0.4
This study (20°C)
0.2
Coats et al. (2011b)
0.0 0
5
10
15
Glycogen (Cmmol L-1)
Fig. 5 e Dependence of SDNR on glycogen concentration at start of anoxic period. Values for this study are corrected to a temperature of 20 C.
PAOs deriving maintenance energy during the anoxic period via oxidative and substrate-level phosphorylation rather than poly-P hydrolysis. The results presented herein indicated that secondary P release did not occur during the anoxic period as long as nitrate/nitrite remained available (Fig. 3a, c and Fig. 4a, c). However, to better understand the maintenance energy dynamics in the absence of NOx (principal electron acceptors driving post-anoxic oxidative phosphorylation), the reactors were monitored for 3 h following the end of the high aeration study (Fig. 3d shows that NOx was mostly depleted at the start of this test). As shown (Fig. 6), glycogen utilization continued in both reactors, suggesting that maintenance energy was mainly supplied via glycolysis. While secondary P release was not observed in reactor O, a nominal release occurred in reactor C (w0.15 mgP L1). Notably, the rate of P release in reactor C was markedly lower than reported for PAO anaerobic maintenance. Secondary P release rates in the range of 2e5 mgP gVSS1 h1 have been reported for PAOs (Oehmen et al., 2005; Smolders et al., 1995; Wentzel et al., 1989), but the release rate observed during this extended test (reactor C) was just
12.0
1.0
AX
Extended AX
0.9 0.8
11.0
0.7
10.5
0.6
O-Gly C-Gly
10.0
0.5
O-P C-P O-DO
9.5 9.0
0.4 0.3 0.2
8.5
PO4-P, DO (mg L-1)
Glycogen (Cmmol L-1)
11.5
0.1
8.0
0.0 3
4
5
6 7 Time (hr)
8
9
Fig. 6 e Glycogen and P transformations during the normal post-anoxic period and an extended post-anoxic period after nitrate/nitrite depletion. The DO concentration is plotted for reactor O.
0.016 mgP gVSS1 h1, more than 100 times slower. This difference cannot be attributed solely to the relative differences in Accumulibacter abundance between this study and the aforementioned studies. Rather, the microbial consortia enriched for in this study appears to have preferentially metabolized glycogen for maintenance energy before poly-P, which is also consistent with the findings of Lu et al. (2007) for their PAO anaerobic starvation experiment. It is also possible that the lower poly-P reserves of the PAOs in this study (VSS/TSS w0.78e0.84) compared to synthetic-fed EBPR cultures (VSS/TSS w0.5e0.7) made glycolysis the more favorable pathway for maintenance ATP production. Erdal et al. (2008) similarly concluded that PAOs are capable of a metabolic shift from poly-P to glycogen when poly-P pools are lowered. They postulated that the preference for the glycolytic pathway could be regulated by temperature, where glycolysis is favorable at temperatures above 20 C. Considering that the temperatures in this study were 23e25 C, the PAOs may have been relying more on glycogen to meet their ATP demands. Regardless, it would appear that secondary P release is not a significant concern in this post-anoxic BNR process.
3.5. Potential organisms involved in post-anoxic denitrification Two types of PAOs have been indentified in EBPR systems, referred to as Accumulibacter Type I (PAOI) and Accumulibacter Type II (PAOII). PAOI is postulated to be capable of full denitrification, whereas PAOII is only able to denitrify from nitrite onwards (Oehmen et al., 2010b). Both PAO types would be amplified with the primer set employed in this study. Regarding GAOs, Competibacter subgroup 6 (amplified within the GAO primer set and GB lineage) is capable of full denitrification (Kong et al., 2006), while Competibacter subgroups 1, 4, and 5 (all collectively amplified within the GAO primer set and GB lineage (Table 2)) and Defluvicoccus Cluster I (not quantified in this study) are able to reduce nitrate only (Oehmen et al., 2010b). Based on the Prel/VFA and Gly/VFA ratios, fractional PAO and GAO abundances, and overall reported N and P removal reported herein, it would appear that the imposed environmental pressures in both reactor configurations selected for a sufficiently enriched mixture of PAOs and GAOs capable of both phosphorus removal and denitrification. Since significant nitrite accumulation was never observed in any of the sampling runs, it is probable that the dominant denitrifying organisms were capable of using both nitrate and nitrite as electron acceptors (Carvalho et al., 2007; Wang et al., 2008). This suggests that most of the denitrification took place within the same microorganism rather than by multiple flanking species (with the exception of reactor O when supplied synthetic acetate, as discussed previously). Further, the PAOs in the system only hydrolyzed poly-P (i.e., secondary P release) in the absence of nitrate/nitrite, and therefore likely were capable of full denitrification (Barnard and Fothergill, 1998). Although some nitrite accumulation was observed in reactor C (Fig. 3d), nitrite was rapidly removed from solution after the depletion of nitrate. The organisms that reduced the nitrite at the end of the anoxic period seemingly relied on nitrate until it was no longer available, at which point they switched to nitrite (a less favorable electron acceptor).
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 1 9 e6 1 3 0
3.6.
Considerations for full-scale application
To the best of our knowledge, a post-anoxic denitrification EBPR process that is designed and/or operated to significantly rely on intracellular glycogen as a carbon source has yet to be applied at full-scale. A membrane-based process was investigated and patented (WO03057632) by a research group out of Berlin in 2004 (Vocks et al., 2005). The same group has operated continuous flow laboratory- and pilot-scale membrane bioreactors to demonstrate the potential viability of this process at full-scale (Bracklow et al., 2010). They observed total N and total P removal efficiencies of 86e94% and 92e99%, with SDNRs ranging from 0.5 to 1.5 mgNO3eN gVSS1 h1. However, Bracklow et al., (2010) reported a limited understanding of the microbes and carbon source driving postanoxic denitrification. Complementing our prior work (Coats et al., 2011b), the results of this study suggest that the proposed post-anoxic BNR process is also suitable for SBR configurations, and further that the process may be appropriate for continuous flow WWTPs. Specifically, our results demonstrated that the process is capable of achieving >99% soluble P and inorganic N removal, and that limiting SOT in the anaerobic and anoxic basins can improve EBPR stability under low ammonia loading. At full-scale, variable speed mixers could be installed in the anaerobic and anoxic basins to minimize mixing and reduce SOT. Further, the design of a full-scale facility would have to consider the implications of variable ammonia loadings and secondary P release, both of which were investigated in this study.
4.
Conclusions
The research presented and discussed herein focused on advancing the understanding of a relatively new post-anoxic BNR process. The major findings from this study can be summarized as follows. Post-anoxic denitrification can accomplish near-complete soluble inorganic N and P removal (>99%). Process success is enhanced at elevated aeration rates, but significant removal can be achieved at reduced aeration. Intracellular glycogen, synthesized associated with EBPR, is an important carbon source used by the mixed microbial consortium to achieve denitrification. A positive correlation between the SDNR and intracellular glycogen concentration was observed. Furthermore, glycogen oxidization for denitrification does not compromise subsequent anaerobic VFA uptake and PHA storage, which is critical to EBPR. A mixed VFA substrate (HAc, HPr, HBu, and HVa) appears to be more beneficial to process performance and supports a higher percentage of PAOs than an acetate-dominated substrate. Post-anoxic secondary P release can occur with NOx depletion. However, P release was only observed when SOT was prevented, and the rate of release was such that effluent P was only moderately increased.
6129
The proposed process configuration is potentially sensitive to low influent ammonia (<20 mgN/L), but stable performance can be maintained by minimizing SOT. All tested reactor configurations achieved significant P removal despite variability over time in the relative PAO fraction, and also considering a relatively significant GAO population.
references
APHA, 1998. Standard Methods for the Examination of Water and Wastewater. American Public Health Association. Barnard, J.L., Fothergill, S., 1998. Secondary phosphorus release in biological phosphorus removal systems. WEFTEC conference proceedings, vol. 1 Orlando, FL, USA475e485. Bortone, G., Saltarelli, R., Alonso, V., Sorm, R., Wanner, J., Tilche, A., 1996. Biological anoxic phosphorus removal - the dephanox process. Water Sci. Technol. 34, 119e128. Bracklow, U., Drews, A., Gnirss, R., Klamm, S., Lesjean, B., Stu¨ber, J., Barjenbruch, M., Kraume, M., 2010. Influence of sludge loadings and types of substrates on nutrients removal in MBRs. Desalination 250, 734e739. Carvalho, G., Lemos, P.C., Oehmen, A., Reis, M.A.M., 2007. Denitrifying phosphorus removal: linking the process performance with the microbial community structure. Water Res. 41, 4383e4396. Coats, E., Gregg, M., Crawford, R., 2011a. Effect of organic loading and retention time on dairy manure fermentation. Bioresour. Technol. 102, 2572e2577. Coats, E.R., Mockos, A., Loge, F.J., 2011b. Post-anoxic denitrification driven by PHA and glycogen within enhanced biological phosphorus removal. Bioresour. Technol. 102, 1019e1027. Coats, E.R., Watkins, D.L., Brinkman, C.K., Loge, F.J., 2011c. Effect of anaerobic HRT on biological phosphorus removal and the enrichment of phosphorus accumulating organisms. Water Environ. Res. 83, 461e469. Copp, J.B., Dold, P.L., 1998. Comparing sludge production under aerobic and anoxic conditions. Water Sci. Technol. 38, 285e294. Cornel, P., Wagner, M., Krause, S., 2003. Investigation of oxygen transfer rates in full scale membrane bioreactors. Water Sci. Technol. 47, 313e319. Crocetti, G.R., Banfield, J.F., Keller, J., Bond, P.L., Blackall, L.L., 2002. Glycogen-accumulating organisms in laboratory-scale and full-scale wastewater treatment processes. Microbiol 148, 3353e3364. Elefsiniotis, P., Wareham, D.G., 2007. Utilization patterns of volatile fatty acids in the denitrification reaction. Enzym. Microb. Technol. 41, 92e97. Erdal, U.G., Erdal, Z.K., Daigger, G.T., Randall, C.W., 2008. It is PAO-GAO competition or metabolic shift in EBPR system? Evidence from an experimental study. Water Sci. Technol. 58, 1329e1334. Flowers, J.J., He, S., Yilmaz, S., Noguera, D.R., McMahon, K.D., 2009. Denitrification capabilities of two biological phophorus removal sludges dominated by different ’Candidatus Accumulibacter’ clades. Environ. Microbiol. Rep. 1, 583e588. He, S., Gall, D.L., McMahon, K.D., 2007. Candidatus Accumulibacter" population structure in enhanced biological phosphorus removal sludges as revealed by polyphosphate kinase genes. Appl. Environ. Microbiol. 73, 5865e5874.
6130
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 1 9 e6 1 3 0
Henze, M., van Loosdrecht, M.C.M., Ekama, G.A., Brdjanovic, D., 2008. Biological Wastewater Treatment: Principles, Modelling, and Design, First ed. IWA Publishing. Kong, Y., Ong, S.L., Ng, W.J., Liu, W.-T., 2002. Diversity and distribution of a deeply branched novel proteobacterial group found in anaerobic-aerobic activated sludge processes. Environ. Microbiol. 4, 753e757. Kong, Y., Xia, Y., Nielsen, J.L., Nielsen, P.H., 2006. Ecophysiology of a group of uncultured Gammaproteobacterial glycogenaccumulating organisms in full-scale enhanced biological phosphorus removal wastewater treatment plants. Environ. Microbiol. 8, 479e489. Kuba, T., Murnleitner, E., Van Loosdrecht, M.C.M., Heijnen, J.J., 1996a. A metabolic model for the biological phosphorus removal by denitrifying organisms. Biotechnol. Bioeng. 52, 685e695. Kuba, T., van Loosdrecht, M.C.M., Heijnen, J.J., 1996b. Phosphorus and nitrogen removal with minimal COD requirement by integration of denitrifying dephosphatation and nitrification in a two-sludge system. Water Res. 30, 1702e1710. Kuba, T., van Loosdrecht, M.C.M., Murnleitner, E., Heijnen, J.J., 1994. Effect of nitrate on phosphorus release in biological phosphorus removal systems. Water Sci. Technol. 30, 263e269. Kujawa, K., Klapwijk, B., 1999. A method to estimate denitrification potential for predenitrification systems using NUR batch test. Water Res. 33, 2291e2300. Liu, W.T., Mino, T., Nakamura, K., Matsuo, T., 1994. Role of glycogen in acetate uptake and polyhydroxyalkanoate synthesis in anaerobic-aerobic activated sludge with a minimized polyphosphate content. J. Fermentation Bioeng. 77, 535e540. Lopez-Vazquez, C.M., Oehmen, A., Hooijmans, C.M., Brdjanovic, D., Gijzen, H.J., Yuan, Z., van Loosdrecht, M.C.M., 2009. Modeling the PAO-GAO competition: effects of carbon source, pH and temperature. Water Res. 43, 450e462. Lopez, C., Pons, M.N., Morgenroth, E., 2006. Endogenous processes during long-term starvation in activated sludge performing enhanced biological phosphorus removal. Water Res. 40, 1519e1530. Lu, H., Keller, J., Yuan, Z., 2007. Endogenous metabolism of Candidatus Accumulibacter phosphatis under various starvation conditions. Water Res. 41, 4646e4656. Martin, H.G., Ivanova, N., Kunin, V., Warnecke, F., Barry, K.W., McHardy, A.C., Yeates, C., He, S.M., Salamov, A.A., Szeto, E., Dalin, E., Putnam, N.H., Shapiro, H.J., Pangilinan, J.L., Rigoutsos, I., Kyrpides, N.C., Blackall, L.L., McMahon, K.D., Hugenholtz, P., 2006. Metagenomic analysis of two enhanced biological phosphorus removal (EBPR) sludge communities. Nat. Biotechnol. 24, 1263e1269. Martins, A., Heijnen, J.J., van Loosdrecht, M.C.M., 2004. Bulking sludge in biological nutrient removal systems. Water Res. 86, 125e135. Oehmen, A., Carvalho, G., Lopez-Vazquez, C.M., van Loosdrecht, M.C.M., Reis, M., 2010a. Incorporating microbial ecology into the metabolic modelling of polyphosphate accumulating organisms and glycogen accumulating organisms. Water Res. 44, 4992e5004. Oehmen, A., Lemos, P.C., Carvalho, G., Yuan, Z.G., Keller, J., Blackall, L.L., Reis, M.A.M., 2007. Advances in enhanced
biological phosphorus removal: from micro to macro scale. Water Res. 41, 2271e2300. Oehmen, A., Lopez-Vazquez, C.M., Carvalho, G., Reis, M., van Loosdrecht, M.C.M., 2010b. Modelling the population dynamics and metabolic diversity of organisms relevant in anaerobic/anoxic/aerobic enhanced biological phosphorus removal processes. Water Res. 44, 4473e4486. Oehmen, A., Zeng, R.J., Yuan, Z., Keller, J., 2005. Anaerobic metabolism of propionate by polyphosphate-accumulating organisms in enhanced biological phosphorus removal systems. Biotechnol. Bioeng. 91, 43e53. Oh, J., Silverstein, J., 1999. Oxygen inhibition of activated sludge denitrification. Water Res. 33, 1925e1937. Parrou, J.L., Francois, J., 1997. A simplified procedure for a rapid and reliable assay of both glycogen and trehalose in whole yeast cells. Anal. Biochem. 248, 186e188. Plo´sz, B., Jobba´gy, A., Grady Jr., C.P.L., 2003. Factors influencing deterioration of denitrification by oxygen entering an anoxic reactor through the surface. Water Res. 37, 853e863. Ramakers, C., Ruijter, J.M., Deprez, R.H.L., Moorman, A.F.M., 2003. Assumption-free analysis of quantitative real-time polymerase chain reaction (PCR) data. Neurosci. Lett. 339, 62e66. Schuler, A., Jenkins, D., 2003. Enhanced biological phosphorus removal from wastewater by biomass with different phosphorus contents, Part I: experimental results and comparison with metabolic models. Water Environ. Res. 75, 485e498. Smolders, G.J.F., van der Meij, J., van Loosdrecht, M.C.M., Heijnen, J.J., 1995. A structured metabolic model for anaerobic and aerobic stoichiometry and kinetics of the biological phosphorus removal process. Biotechnol. Bioeng. 47, 277e287. Smolders, G.J.F., van Dermeij, J., van Loosdrecht, M.C.M., Heijnen, J.J., 1994. Model of the anaerobic metabolism of the biological phosphorus removal process: stoichiometry and pH influence. Biotechnol. Bioeng. 43, 461e470. Tchobanoglous, G., Burton, F.L., Stensel, H.D., 2003. Wastewater Engineering: Treatment and Reuse, Fourth ed. McGraw-Hill. Van’t Riet, K., 1979. Review of measuring methods and results in nonviscous gas-liquid mass transfer in stirred vessels. Ind. Eng. Chem. Process Dev. 18, 357e364. Vocks, M., Adam, C., Lesjean, B., Gnirss, R., Kraume, M., 2005. Enhanced post-denitrification without addition of an external carbon source in membrane bioreactors. Water Res. 39, 3360e3368. Wang, X.L., Zeng, R.J., Dai, Y., Peng, Y.Z., Yuan, Z.G., 2008. The denitrification capacity of cluster 1 Defluviicoccus vanusrelated glycogen accumulating organisms. Biotechnol. Bioeng. 99, 1329e1336. Wentzel, M.C., Dold, P.L., Ekama, G.A., Marais, G.V.R., 1989. Enhanced polyphosphate organism cultures in activated sludge. Part III: kinetic model. Water SA 15, 89e102. Zhang, C., Chen, Y., Randall, A.A., Gu, G., 2008. Anaerobic metabolic models for phosphorus- and glycogen accumulating organisms with mixed acetic and propionic acids as carbon sources. Water Res. 42, 3745e3756. Zhou, Y., Pijuan, M., Oehmen, A., Yuan, Z., 2010. The source of reducing power in the anaerobic metabolism of polyphosphate accumulating organisms (PAOs) - a minireview. Water Sci. Technol. 61, 1653e1662.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 3 1 e6 1 4 0
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Characteristic transformation of humic acid during photoelectrocatalysis process and its subsequent disinfection byproduct formation potential Angzhen Li a,b, Xu Zhao a, Huijuan Liu a, Jiuhui Qu a,* a
State key Laboratory of Environmental Aquatic Chemistry, Research Center for Eco-Environmental Sciences, Chinese Academy of Sciences, Beijing 100085, China b Graduate School, Chinese Academy of Sciences, Beijing 100039, China
article info
abstract
Article history:
In this study, degradation of humic acid (HA) via photoelectrocatalysis (PEC) process and
Received 11 May 2011
corresponding disinfection byproduct formation potential (DBPFP) were investigated.
Received in revised form
Particularly, structure variation and subsequent DBPFP of HA during PEC treatment were
4 September 2011
correlated. The PEC process was found to be effective in reducing dissolved organic carbon
Accepted 5 September 2011
concentration by 75.0% and UV absorbance at 254 nm by 92.0%. Furthermore, 90.3% of
Available online 14 September 2011
haloacetic acids formation potential and 89.8% of trihalomethanes formation potential were reduced within 180 min. Based on molecular weight and resin fraction results, it was
Keywords:
demonstrated that HA with large aromatic, hydrophobic and long aliphatic chain organics
Photoelectrocatalytic process
were transformed into small and hydrophilic organics during PEC process. Combined with
Humic acid
the fourier transform infrared spectra and 13C nuclear magnetic resonance spectra analysis
Disinfection byproduct formation
of HA fractions, it was concluded that phenolic hydroxyl and conjugated double bonds
potential
tended to be attacked by hydroxyl radicals during PEC process; these groups were reactive
Structure variation
with chlorine to produce disinfection byproducts (DBP), especially trihalomethane and
Functional groups
trichloroacetic acid. By contrast, amino, carboxyl and alcoholic hydroxyl groups were relatively difficult to be oxidized during PEC process; these groups had the potential to form dichloroacetic acid during chlorination. Results of these studies confirmed that PEC process was a safe and effective technique to decrease DBP formation significantly in water treatment plant. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
Humic acid (HA) is the major fraction of natural organic matter in water, which poses significant concerns to water utility and can interfere with several water treatment processes (Chen et al., 1977). It is known that HA can react with oxidants to produce small molecular organic substances, which contribute to producing halogenated disinfection byproducts (DBP) during chlorination. In particular, HA is the
principal precursor of trihalomethanes (THMs) and haloacetic acids (HAAs), which are potentially carcinogenic (Nie et al., 2010; Xue et al., 2011). Therefore, it is important to understand the characteristic transformation of HA during the oxidation process and establish potential alternative water treatment processes to reduce DBP precursors. Coagulation and activated carbon adsorption are commonly used to remove HA from water (Bond et al., 2010). It is frequently found that these treatment processes are limited
* Corresponding author. Tel.: þ86 10 62849151; fax: þ86 10 62923558. E-mail addresses: [email protected], [email protected] (J. Qu). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.09.012
6132
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 3 1 e6 1 4 0
in the removal of HA with low molecular weight fraction. Recently, application of advanced oxidation processes (AOPs) in HA removal, such as UV/H2O2 and TiO2/UV has gained momentum (Sanly et al., 2008a, b; Siva and Madjid, 2010). The AOPs are attractive due to the potential of free radicals to degrade the organic macromolecules (Sanly et al., 2008a, b). By contrast with other AOPs, it is easier to remove DBP precursors during photocatalysis (PC) process before disinfection. However, the problems of TiO2 separation from aqueous phase and fast recombination rate of electron-hole pair become the limitation of PC treatment (Sanly et al., 2008a, b). One way to improve the performance of immobilized TiO2 is to fix the catalyst on a conductive substrate and apply external voltage to a photoelectrochemical cell, thus driving the photo-generated electrons to the cathode and, consequently, minimizing the rate of electron/hole recombination. The process, which is also referred to as photoelectrocatalysis (PEC), has successfully demonstrated the degradation of HA (Kim and Anderson, 1994; Li et al., 2002; Pinhedo et al., 2005; Huseyin and Miray, 2008; Huseyin, 2010). Previous studies on the PEC oxidation of HA focused on the effect of treatment parameters, such as pH, external potential, and HA dose on the removal efficiency (Pinhedo et al., 2005). In addition, these studies used bulk parameters, such as DOC analysis and UV spectrophotometry to assess treatment performance (Kim and Anderson, 1994; Li et al., 2002; Huseyin and Miray, 2008; Huseyin, 2010). However, these parameters may not be adequate to fully elucidate the chemical process. Relatively little information was available on how PEC process changed the composition and characteristics of HA. Furthermore, in consideration of drinking water security it is important to know the potential of the treated water to form DBP. In this study, the evolution of HA and decrease of disinfection byproduct formation potential (DBPFP) during the PEC treatment were fully investigated. HA was fractionated by ultrafiltration and resin fractionation according to the molecular weight and physicochemical properties (Aiken et al., 1992; Leenheer et al., 2000; Kitis et al., 2002; Hua and Reckhow, 2007). Each organic fraction and its speciation variation during PEC process were examined for their associated functional groups by three dimensional excitationemission matrix (3DEEM) fluorescence, fourier transform infrared (FT-IR) and 13C nuclear magnetic resonance (13C NMR) spectra analysis. Correlation of HA structure variation with corresponding DBPFP of HA during PEC process was also investigated.
2.
Materials and methods
2.1.
HA and fractionation
Commercially sourced HA from Aldrich was used. HA stock solution was prepared by mixing 4 g of HA in 1 L of 0.1 M sodium hydroxide (Univar) over a period of 1 d and then filtered through 0.45 mm glass fiber membrane filters and stored in the dark at 4 C. Diluted HA was fractionated into five fractions using a stirred ultrafiltration cell device(Model 8200, Amicon, Millipore) with nominal molecular weight cutoffs of 3, 10, 30, and
100 kDa regenerated cellulose membranes (PL, 63.5 mm, Millipore). Experimental details follow the procedure done by Kitis et al. (2002). Meanwhile, diluted HA was also fractionated by resin fractionation. The filtered HA was acidified to pH 2 using 6 M sulfuric acid and then passed through DAX-8 resin followed by XAD-4 resin, in accordance with the method of Aiken et al. (1992). Effluent from the XAD-4 resin was collected, which was referred to as the hydrophilic (Hi) fraction. The hydrophobic (Ho) and transphilic (Hs) fractions were retained by DAX-8 and XAD-4 resin respectively, and these were eluted with 0.1 M sodium hydroxide in the reverse direction. The Ho and Hs fractions were re-concentrated on the appropriate resin and hydrogen-saturated using Dowex Marathon MSC-H cation exchange resin obtained from J﹠K. The dissolved organic carbon (DOC) concentration and the UV absorbance at 254 nm (UV254) of each HA fraction were measured.
2.2.
Reactor setup and degradation experiments of HA
Electrochemical oxidation (EO), PC, and PEC processes for HA degradation were carried out in a 600 mL, singlecompartment, air-tight glass cell with a 3.5 cm-diameter quartz tube placed in the center and used as the UV bulb housing. UV irradiation was provided by a 10 W-pressure mercury lamp (253.7 nm) and the light intensity at the center of the compartment was 1.36 mW cm2 as measured with a UV radiometer (Light and Electric instruments Factory of Beijing Normal University). The experimental setup was described by Xiao et al. (2009). The reactor was controlled by a DC power supply source AMERLLPS302A (Dahua instrument corporation of Beijing). The TiO2/Ti film electrode (100 mm 60 mm 2 mm, Hengli Ti Corporation of Beijing) was selected as anode with an apparent surface area of 60 cm2, and TiO2 films were deposited onto the Ti plate via a dip-coating method as described by Shang et al. (2003). RuO2/Ti electrode with the same solid surface area was selected as cathode.
2.3.
DBP formation potential
A 24 h chlorination DBPs test of raw HA and treated water samples were carried out according to Standard Method 5710 with modifies (Greenberg et al., 1995). The sodium hypochlorite (NaOCl) stock solution with the concentration of 20 mg Cl2/mL was stored in an aluminum foil-covered glass stopped flask. Chlorine dosing solution was prepared from the dilution of NaOCl stock solution (about 5 mg Cl2/mL). Sodium hydroxide/potassium dihydrogen phosphate (NaOH/KH2PO4) buffer solutions (pH 7.0) and chlorine dosing solution were injected into each sample. The chlorine dose was determined by 4 h preliminary demand tests on each sample according to Standard Method 5710B (Greenberg et al., 1995). After being dosed with chlorine, samples were stored headspace-free at 25 2 C in the dark for 24 h. All samples were found to have measurable free chlorine residuals by an N,N-diethyl-pphenylenediamine (DPD) titrimetric method (Clesceri et al., 1998). According to the chlorine dose, water samples were then added with sodium sulfite to quench the residual free chlorine and analyzed to determine the concentrations of
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 3 1 e6 1 4 0
trihalomethane formation potential (THMFP) and haloacetic acid formation potential (HAAFP) (Li et al., 2008). Four THMs (CHCl3, CHBrCl2, CHBr2Cl, CHBr3) samples were extracted with hexane (HPLC Grade, Fisher, USA) and measured following the U.S EPA Method 551 (U.S.EPA, 1990). Nine HAAs (monochloro-, monobromo-, dichloro-, bromochloro-, dibromo-, bromodichloro-, bromodichloro-, dibromochloro-, trichloro- and tribromoacetic acid) samples were extracted with methyl-tert-butyl ether (MTBE) (HPLC Grade, J.T. Baker, USA) followed by derivation with acidic methanol according to the U.S. EPA Method 552.3 (U.S.EPA, 1990). 1,2dibromopropane (98.0%, GC, Fluka, USA) was used as the internal standard. Conditions for the analyses were as follows: (1) THMs, injector temperature 200 C, column temperature 35 C (holding 4 min) to 260 C (10 C/min), detector temperature 290 C; (2) HAAs, injector temperature 200 C, column temperature 35 C (holding 4 min) to 65 C (2 C/min), detector temperature 290 C.
2.4.
Characterization of HA
HA and each fraction powder obtained through freeze-drying were analyzed for their structural and chemical characteristics. KBr (FT-IR Grade, Aldrich Co., USA) was mixed with the HA powder at the ratio of 100 to 1 and the FT-IR spectra of the mixture was obtained with IR spectrometer (Thermo Nicolet 5700, USA). 3DEEM fluorescence spectra were recorded on a fluorescence spectrophotometer (model F-4500, Hitachi, Japan). 3DEEM spectra were obtained by measuring the emission spectra in the range from 280 to 550 nm repeatedly at the excitation wavelengths from 220 to 440 nm. Blanks of ultrapure water were included in the correction of inner filtering and Raman scattering of the fluorescence spectra. The solid-state 13C NMR spectra were acquired using a cross polarization magic angle spinning (CPMAS) on a Bruker instrument (AVANCE III, 400 MHz, Bruker, Germany) with a 4 mm H/X/Y probe. CPMAS 13C NMR was performed on 100e200 mg of samples. The main experimental parameters included contact time of 3 ms, pulse delay of 1 s, spinning rate of 5000 Hz and 20480 scans per sample. The PEC oxidation intermediates of HA were detected by gas chromatography/mass spectrometry (GC/MS) (Agilent 7890 GC-5975MSD) equipped with an HP-5 capillary column (30 m 0.25 mm, 0.25 mm film thickness). Details of GC/MS tests were included in the Supplementary Data.
Fig. 1 e Performance of photocatalysis, electrochemical oxidation and photoelectrocatalysis process during humic acid degradation. (initial pH [ 7.0, initial DOC [ 10.0 mg/L as C, Na2SO4 [ 10.0 mmol/L).
reduction of DOC were achieved during EO process. It was clear that significant synergetic effects occurred during PEC process. When a potential greater than the flat-band potential of the TiO2/Ti electrode was applied, the charge recombination process will be inhibited (Li et al., 2002; Pinhedo et al., 2005; Huseyin, 2010). Thus, the HA was efficiently degraded during PEC process. Furthermore, it is clear seen from Fig. 1 that the removal of UV254 was preferred to the removal of DOC. Within 180 min, approximately 2.5 mg/L DOC still remained in the HA solution during PEC process. Even with prolonged reaction time, removal of DOC was limited. Higher percentage of UV254 removal compared to DOC removal indicated an incomplete mineralization of the organics after the breakup of large aromatic and conjugated structures. Similar results were also reported by Huang et al. (2008), who observed that a fraction of the DOC could not be removed completely, implying the presence of refractory compounds, either present in the beginning or generated during PC process.
3.2. Characterization and variation of DBP precursors during PEC process 3.2.1.
3.
Results and discussion
3.1.
PEC degradation of DBP precursors
Degradation of HA, namely DBP precursors, was performed in the PC, EO and PEC process. Fig. 1 shows the DOC and UV254 removal efficiency of HA during PC, EC and PEC process at pH 7. With an applied current density of 5 mA/cm2, the removal efficiency of UV254 and DOC during PEC process was determined to be 92% and 75% within 180 min. By contrast, the removal efficiency of UV254 and DOC during PC process was 25% and 13%, respectively; only 17% reduction of UV254 and 8%
6133
Variation of DBP precursors fractionation
Raw HA and treated HA during PEC process were taken at given time and fractionated by ultrafiltration and resin fractionation according to molecular weight and physicochemical properties. The molecular weight distribution of samples taken during the PEC process are presented in Fig. 2(a). The fraction with molecular weight >100 kDa fraction dominated 78.4% of the DOC of raw HA at pH 7. Each of the other four fractions (<3 kDa, 3e10 kDa, 10e30 kDa and 30e100 kDa) represented less than 6% of the DOC. However, the >100 and <3 kDa fraction comprised 2.6% and 67.2% of total DOC at 180 min during PEC process, respectively. These results revealed that the PEC oxidation preferentially degraded the large molecular weight fractions of HA, forming small
6134
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 3 1 e6 1 4 0
a
b
respectively, which suggested that the Ho fraction was prone to PEC degradation. Thus, when the Ho fraction was almost completely degraded, the hydroxyl radicals (HO$) continued to attack the Hs and Hi fraction. This was also agreeable with the molecular weight results since the Ho fraction was composed of large molecular weight components, which were expected to be more readily degraded (Patrica et al., 2002). Tran et al. reported that the Ho fraction generally contained a large proportion of aromatic structures (Tran et al., 2006). It is known that aromatic compounds tend to have a high electron density and therefore an increased probability of being attacked by HO$. The Hi fraction contains more aliphatic carbon and nitrogenous compounds (such as carbohydrates, sugars, and amino acids), which are more difficult to be degraded than the Ho fraction (Hua and Reckhow, 2007). The results from this study indicated that HA degradation proceeded via breakup of large molecular weight and hydrophobic nonpolar fractions to form small molecular weight and hydrophilic charged fractions during PEC process.
3.2.2.
Fig. 2 e DOC concentration variation of the humic acid fractionation during photoelectrocatalysis process. (a) based on molecular size; (b) based on hydrophobicity. (initial DOC [ 10.0 mg/L as C, initial pH [ 7.0, Na2SO4 [ 10.0 mmol/L, current density [ 5 mA/cm2).
molecular weight fractions, which were subsequently degraded after the large molecular weight organics were nearly diminished. Fig. 2(b) illustrates DOC variations of the three resin fractions of samples taken during PEC process at pH 7. The resin fractionation result indicated that the Ho fraction predominated the DOC composition of HA: 8.3 mg/L of Ho fraction (82.7% of DOC), with the Hs and Hi fractions making up 1.1 mg/ L (11.4% of DOC), and 0.6 mg/L (5.9% of DOC), respectively. Within 180 min of PEC process, the overall DOC concentration of the Ho fraction and HA decreased by 8.1 mg/L and 7.5 mg/L, respectively. The DOC concentration of the Hs fraction remained stable with 60 min, and then decreased during PEC process. While Ho and Hs fractions continuously decreased following treatment, the DOC concentration of Hi fraction increased from 0.59 mg/L to 3.20 mg/L within 120 min, and reduced to 2.18 mg/L at 180 min. From calculations, it was found that the rate of Ho being converted to Hs/Hi was 2.3, 1.6, and 1.3 times faster than the rate of Hs/Hi being degraded to carbon dioxide (CO2) at 60 min, 120 min, and 180 min,
3DEEM fluorescence spectroscopy analysis
Detailed 3DEEM illumination for the HA fractions of Ho, Hs, and Hi is shown in Fig. 3. Excitation and emission boundaries were operationally defined into five regions based on the results by Wen et al. (2003). Additionally, f450/500 corresponded to the ratio of emission intensity at 450 nm over 500 nm at 370 nm excitation and lower f450/500 may refer to the compounds with more aromatic structures. The f450/500 for the Ho, Hs and Hi fraction was 1.13, 1.50 and 1.34, respectively. It is shown in Fig. 3 that the peak of Hi appeared in Region Ⅱ; the peaks of Ho and Hs appeared in Region Ⅴ. These peaks suggested that the Hi fraction contained protein-like substances, whereas the Ho and Hs fractions contained humic acid-like substances. Furthermore, the Ho fraction may have more aromatic sites than the other two fractions. Fig. 4 shows the variation of HA fluorescence spectroscopy with the reaction time during PEC process. The fluorescence intensity of peak at 324/440 (excitation (Ex)/emission (Em) (nm)) in Region V decreased from 50 (a.u.) to 30 (a.u.) within 60 min, then the peak in Region V almost disappeared at 180 min during PEC process. While a peak in the Region Ⅳ appeared at 180 min, which indicated that the small molecular weight organics were formed during the reaction. Furthermore, the blue shift of the peak from 325/445 (excitation (Ex)/emission (Em) (nm)) to 324/440 (Ex/Em (nm)) within 60 min was observed. It is recognized that blue shift of Ex/Em reflects the transformation of organic material into small molecular weight composition and a reduction in the degree of the p-electron system (Swietlika and Sikorska, 2004). The result suggested HA with the large molecular weight fraction was decomposed to small molecular weight fraction during PEC process. Moreover, because the breakup of aromatic structures were easily achieved, the Ho degradation was easier than the Hs and Ho fractions.
3.2.3.
FT-IR analysis
The FT-IR spectra for HA and its fractions of Ho, Hi, and Hs are shown in Fig. 5 (a). As seen in Fig. 5(a), Ho was found to have a peak at 1645 cm1, which indicates that Ho had more aromatic structures (Hyun and Myong, 2005; Elizabeth and Brandon,
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 3 1 e6 1 4 0
6135
Fig. 3 e Fluorescene spectroscopy of humic acid fractions based on hydrophobicity. (a) Hydrophobic fraction; (b) transphilic fraction; (c) Hydrophilic fraction.
Fig. 4 e Fluorescene spectroscopy of humic acid during photoelectrocatalysis process. (a) 0 min; (b) 60 min; (c) 180 min (initial DOC [ 10.0 mg/L as C, Na2SO4 [ 10.0 mmol/L, current density [ 5 mA/cm2).
6136
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 3 1 e6 1 4 0
a
b
substances (Elizabeth and Brandon, 2008). Summarily, the Ho fraction mainly contained phenolic hydroxyl and conjugated double bonds while the Hs fraction contained more carbonyl. However, the Hi fraction may contain more amino, carboxyl and alcoholic hydroxyl groups. The FT-IR spectra of HA during the PEC process is shown in Fig. 5 (b). The intense absorbance of SO24 vibration in the 12001000 cm1 region covered up the vibration of some organic groups. Thus, the magnifying 2800-1200 cm1 region was chosen for FT-IR result. As shown in Fig. 5 (b), the reaction samples taken at 60 min and 180 min showed peaks at 1385 cm1 and 1637 cm1, which are attributed to CeH deformation in methyl, methylene groups and C]C stretching vibration, respectively (Hyun and Myong, 2005; Elizabeth and Brandon, 2008). The peak intensity at 1637 cm1 was stronger for the sample at 60 min than that at 180 min; on the contrary, the peak at 1385 cm1 was weaker for the sample at 60 min than that for 180 min. While the reaction samples taken at 180 min also showed peaks at 2338 cm1 and 2366 cm1, which are attributed to CO2 stretching vibration (Valdemar et al., 2009). These peaks indicated that during PEC process aromatic sites were prone to attack by HO$ to form ring-opened products, such as carboxylic acid, and oxidation to CO2. Combined with resin fractionation results, it was concluded that PEC process primarily destroyed the aromatic and conjugated structure of HA and transformed large aromatic and long aliphatic chain organic structures to small and hydrophilic organics. Furthermore, the preferred removal of the Ho fraction revealed that phenolic hydroxyl and conjugated double bonds tended to be attacked by HO$ during PEC process. However, the slow degradation of the Hi fraction indicated that amino, carboxyl and alcoholic hydroxyl groups were relatively difficult to be oxidized during PEC process.
3.2.4.
Fig. 5 e (a) FT-IR spectra of humic acid and fractions based on hydrophobicity; (b) FT-IR spectra of humic acid during photoelectrocatalysis process.
2008). Whereas Hs showed a peak at 1714 cm1, which is attributed to carboxylic and carbonyl-C]O stretching (Hyun and Myong, 2005). Ho and Hs were similar in spectra with a peak at around 3440 cm1, 1440 cm1 and 880 cm1, which refer to OeH stretching from the presence of phenols, CeH deformation of aliphatic groups and CeH bending of the substituted aromatic groups, respectively (Valdemar et al., 2009). Compared with Ho and Hs, Hi showed a peak at around 1628 cm1, which corresponds to NeH deformation and C]N stretching of primary amides (Hyun and Myong, 2005). Additionally, Hi showed a peak at 1067 cm1, which is assigned to CeO stretching of polysaccharide or polysaccharide-like
CPMAS
13
C NMR analysis
Raw HA and HA fractions of Ho and Hs were analyzed using the CPMAS 13C NMR spectra. As shown in Fig. 6(a), HA and Ho fraction showed higher content of carbons with resonances in the range of 110e160 ppm than Hs fraction. Furthermore, HA samples taken during a given time were also presented in Fig. 6(b). It could be seen that the carbon resonance in the region 110e160 ppm decreased notably within 180 min during the PEC process. Previous studies indicated that the spectra were integrated into the following chemical shift regions: alkyl-C (0e50 ppm), O-alkyl-C (predominantly carbohydrates, 50e110 ppm), aromatic-C (110e160 ppm), carboxyl-C (160e190 ppm), and carbonyl-C (190e220 ppm) (Zhao et al., 2009). The relative intensity of these regions was determined by the integration of the corresponding peak areas. In addition, the aromaticity and aliphaticity of the samples were calculated according to Eq. (1) and Eq. (2): Aromaticity ¼ Cd110160 =Cd 100%
(1)
Aliphaticity ¼ Cd0110 =Cd 100%
(2)
Table 1 showed the relative intensity of functional groups for the HA fractions and samples taken during PEC process by CPMAS 13C NMR spectra. HA and Ho fractions presented more intense resonances in aromatic carbon than Hs fraction.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 3 1 e6 1 4 0
a
6137
accompanied by the increase of carbohydrate carbon contents from 19.01% to 35.69%. These results indicated that the aromatic structure of HA was destroyed whereas the oxygenated aliphatic structures increased during PEC process. These results indicated that the possible reactions between HA and HO involved the addition of HO to aromatic and conjugated structures, then further oxidation could yield alcohol, carboxylic acids, carbohydrates, which may be mineralized during PEC process. These findings were in agreement with the FT-IR analysis.
3.2.5.
GC/MS analysis
Typical GC/MS chromatograms of HA during PEC process are illustrated in detail in the Supplementary Data (Table S1 and Figure S1). As shown in Figure S1, at 60 min, intermediate of dibutyl phthalate was detected. At 180 min, glyoxylic acid, dibutyl phthalate, hexadecanoic acid and octadecanoic acid were identified. It was reported that the oxidation of HA was accomplished by attacking the unsaturated moieties C]C (Li et al., 2008). As a result, such organics as ketones, alcohol, ester, ether, and carboxylic acids could be formed via HO$ addition reactions to the unsaturated groups of HA (such as vinyl and aromatic groups) during PEC oxidation.
3.3. DBPFP reduction and the correlations between the functional groups of DBP precursors
b
Fig. 6 e (a) 13C NMR spectra of humic acid and fractions; (b) 13 C NMR spectra of humic acid during photoelectrocatalysis process. (initial pH [ 7.0, Na2SO4 [ 10.0 mmol/L, current density [ 5 mA/cm2).
However, the Hs fraction indicated higher aliphatic, carboxylic and carbonyl carbon intensity, as compared to the Ho fraction. Furthermore, Table 1 showed that the aromaticity for HA decreased from 40.13% to 19.47% within 180 min,
The DBPFP of HA fractions were measured to assess the reactivity of intermediates with chlorine during PEC treatment. Fig. 7 (a) and (b) indicate that the distribution of four THMs, nine HAAs and the DBPFP/DOC value from three fractions (Ho, Hs and Hi). However, the formation of CHCl3, dichloroacetic acid (DCAA) and trichloroacetic acid (TCAA) contributed to more than 90% of total DBPFP. It could be seen that Ho produced the highest value of THMFP/DOC (38.72 mg/ mg C) and HAAFP/DOC (55.13 mg/mg C) yields; Meanwhile, Ho exhibited the highest densities of CHCl3 (99.87%) and trichloroacetic acid (TCAA) (64.09%) in THMFP and HAAFP, whereas Hs produced almost equal densities of dichloroacetic acid (DCAA) (46.47%) and TCAA (47.71%). The Hi fraction produced more brominated trihalomethane (THM-Br) (19.61%) and DCAA (63.36%) in contrast to the other two fractions. However, THM-Br contributed less to the THMFP since the original HA samples had a low level of bromide (<20 mg/L). This suggests that THM and TCAA precursors were more hydrophobic, while hydrophilic fractions were important as DCAA precursors (Lin and Singer, 2003). Considering the FT-IR result of Ho and Hi fractions of HA, it could be deduced that phenolic hydroxyl and conjugated double bonds had high potential to produce THM and TCAA during chlorination. Whereas this FT-IR result also suggested that amino, carboxyl and alcoholic hydroxyl groups were prone to produce DCAA during chlorination. Fig. 7 (c) and (d) revealed the variation of DBPFP and DBPFP/ DOC of HA fractions during PEC process. It was shown that the THMFP/DOC value decreased drastically from 35.14 to 17.14 mg/mg C during PEC treatment. In contrast to THMFP/ DOC value, HAAFP/DOC value didn’t decrease all the time. The little increase in the HAAFP/DOC value from 120 min to 180 min may be attributed to the increase in hydrophilic
6138
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 3 1 e6 1 4 0
Table 1 e Characterization for HA fractions and intermediates during PEC process analyzed by Relative intensity of functional group (%) Alkyl 0e50 Carbohydrate 50e110 Aromatic 110e160 Carboxyl Carbonyl (ppm) (ppm) (ppm) 160e190 190e220 (ppm) (ppm) Humic acid Hydrophobic Transphilic Humic acid 60 min Humic acid 180 min
26.98 21.58 28.32 20.99 32.35
19.01 11.79 22.62 26.41 35.69
30.84 47.53 21.04 28.61 16.45
19.11 18.35 22.80 19.33 14.31
13
C NMR spectra.
Aromaticity Aliphaticity (%) (%)
4.06 0.75 5.22 4.66 1.2
40.13 58.75 29.23 37.64 19.47
59.87 41.25 70.77 62.36 80.53
Initial pH ¼ 7.0, Na2SO4 ¼ 10.0 mmol/L, current density ¼ 5 mA/cm2.
CHCl2Br
DBCAA DCAA
BDCAA MBAA
TCAA MCAA
DBAA
60
90
55
80
35
80
50
70
30
60
25
50 20
40
15
HAAFP (%)
100
40
THMFP/DOC (µg/mg C)
45
90
70
45
60
40
50 35
40
30
30
20
10
20
25
10
5
10
20
0
0
0 Ho
Hs
350
d
35 Hydrophobic Transphilic Hydrophilic
300 250 200
20
100 15
50 0 60
120
Hs
Hi
500
50
450
Hydrophobic Transphilic Hydrophilic
350
45 40
300 35
250 200
30
150 25
100 50
20
0
10 0
Ho
400
30 25
150
15 HA
Hi
HAAFP (µg/L)
HA
THMFP (µg/L)
TBAA BCAA
100
30
c
b
CHCl3
HAAFP/DOC (µg/mg C)
CHClBr2
HAAFP/DOC (µg/mg C)
CHBr3
THMFP/DOC (µg/mg C)
THMFP (%)
a
0
180
60
120
180
Time (min)
Time (min)
Fig. 7 e DBPFP and DBPFP/DOC value of humic acid based on resin fractionation during photoelectrocatalysis process. (a) Distribution of four THMs in chlorinated humic acid fractions; (b) distribution of nine HAAs in chlorinated humic acid fractions; (c) THMFP and THMFP/DOC of humic acid fractions during photoelectrocatalysis process; (d) HAAFP and HAAFP/ DOC of humic acid fractions during photoelectrocatalysis process. (initial DOC [ 10.0 mg/L as C, Na2SO4 [ 10.0 mmol/L, applied potential [ 8.0 V vs SCE).
fraction during PEC process. PEC oxidation preferentially degraded DBPFP of Ho fraction; whereas, at the initial stage of the PEC oxidation, DBPFP of Hi fraction increased then proceeded to decrease at a slow rate. This was in accordance with the fractionation result that the PEC process broke up large molecular weight, hydrophobic nonpolar fractions to form small molecular weight, hydrophilic charged fractions. Accordingly, THM and TCAA precursors were significantly
oxidized during PEC process; whereas, DCAA precursors were relatively stable during PEC process.
4.
Conclusions
In summary, the photoelectrocatalysis process showed higher efficiency in removal of humic acid and capacity to decrease
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 3 1 e6 1 4 0
disinfection byproducts formation of humic acid than photocatalysis and electrochemical oxidation process. Humic acid was fractionated into three fractions, and the hydrophobic fraction mainly contained phenolic hydroxyl and conjugated double bonds; the transphilic fraction contained more carbonyl in contrast to the hydrophobic fraction. The hydrophilic fraction may contain more amino, carboxyl and alcoholic hydroxyl groups. It was demonstrated that hydrophobic fractions had the highest potential to produce trihalomethanes and haloacetic acids in contrast to the transphilic and hydrophilic fractions. Loss of aromaticity and conjugation was easier to achieve than mineralization of humic acid. Large aromatic and long aliphatic chain organic structures were transformed into small and hydrophilic organics during the photoelectrocatalysis process. These results indicated that phenolic hydroxyl and conjugated double bonds were primarily responsible for disinfection byproducts formation during chlorination, especially trihalomethane and trichloroacetic acid formation; amino and alcoholic hydroxyl groups were important precursors for dichloroacetic acis. It was concluded that the superiority of photoelectrocatalysis process for reducing trihalomethane and trichloroacetic acid precursors compared to dichloroacetic acis precursors.
Acknowledgments This project is supported by Funds for the Creative Research Group of China (Grant 50921064), National Natural Science Foundation of China (Grant 50938004) and National Basic Research Program of China (2010CB933604). The kind suggestions from the editors and reviewers are deeply appreciated.
Appendix. Supplementary material Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.watres.2011.09.012.
references
Aiken, G.R., McKnight, D.M., Thorn, K.A., Thurman, E.M., 1992. Isolation of hydrophilic organic acids from water using nonionic macroporous resins. Organic Geochemistry 18 (4), 567e573. Bond, T., Goslan, E.H., Parsons, S.A., Jefferson, B., 2010. Disinfection by-product formation of natural organic matter surrogates and treatment by coagulation, MIEX and nanofiltration. Water Research 44 (5), 1645e1653. Chen, Y., Senesi, N., Schnitzer, M., 1977. Information provided on humic substances by E4/E6 ratios. Soil Science Society of America Journal 41 (2), 352e358. Clesceri, L.S., Greenberg, A.E., Eaton, A.D., 1998. Standard Methods for the Examination of Water and Wastewater, twentieth ed.. American Public Health Association, American Water Works Association. 461e463. Elizabeth, M., Brandon, S., 2008. Dissolved organic matter characteristics within the Lake Superior watershed. Organic Geochemistry 39 (11), 1489e1501.
6139
Greenberg, A.E., Clesceri, L.S., Eaton, A.D., 1995. Standard Methods for the Examination of Water and Wastewater, nineteenth ed. American Public Health Association, Washington, DC. Hua, G.H., Reckhow, D.A., 2007. Characterization of disinfection byproduct precursors based on hydrophobicity and molecular size. Environmental Science and Technology 41 (9), 3309e3315. Huang, X., Leal, M., Li, Q., 2008. Degradation of natural organic matter by TiO2 photocatalytic oxidation and its effect on fouling of low-pressure membranes. Water Research 42 (4e5), 1142e1150. Huseyin, S., 2010. Disinfection and formation of disinfection byproducts in a photoelectrocatalytic system. Water Research 44 (13), 3966e3972. Huseyin, S., Miray, B., 2008. Photocatalytic and photoelectrocatalytic humic acid removal and selectivity of TiO2 coated photoanode. Chemosphere 73 (5), 854e858. Hyun, C.K., Myong, J.Y., 2005. Characterization of natural organic matter in conventional water treatment processes for selection of treatment processes focused on DBPs control. Water Research 39 (19), 4779e4789. Kim, D.H., Anderson, M.A., 1994. Photoelectrocatalytic degradation of formic acid using a porous TiO2 thin-film electrode. Environmental Science and Technology 28 (3), 479e483. Kitis, M., Karanfil, T., Wigton, A., Kilduff, J.E., 2002. Probing reactivity of dissolved organic matter for disinfection byproduct formation using XAD-8 resin adsorption and ultrafiltration fractionation. Water Research 36 (15), 3834e3848. Leenheer, J.A., Croue, J.P., Benjamin, M., Korshin, G.V., Hwang, C.J., Bruchet, A., Aiken, G.R., 2000. Comprehensive isolation of natural organic matter from water for special characterization and reactivity testing. ACS Symposium Series 761, 68e83. Li, J., Liu, H.J., Zhao, X., Qu, J.H., Liu, R.P., Ru, J., 2008. Effect of preozonation on the characteristic transformation of fulvic acid and its subsequent trichloromethane formation potential: presence or absence of bicarbonate. Chemosphere 71 (9), 1639e1645. Li, X.Z., Li, F.B., Fan, C.M., Sun, Y.P., 2002. Photoelectrocatalytic degradation of humic acid in aqueous solution using a Ti/TiO2 mesh photoelectrode. Water Research 36 (9), 2215e2224. Lin, L., Singer, P.C., 2003. Factors influencing the formation and relative distribution of haloacetic acids and trihalomethanes in drinking water. Environmental Science and Technology 37 (13), 2920e2928. Nie, Y.L., Hu, C., Zhou, L., Qu, J.H., Wei, Q.S., Wang, D.S., 2010. Degradation characteristics of humic acid over iron oxides/Fe0 core-shell nanoparticles with UVA/H2O2. Journal of Hazardous Materials 173 (1e3), 474e479. Patrica, A.M., Michael, J.P., Stephen, E.C., Qunhui, Z., Ksenija, N.D., George, R.A., 2002. A comparison of surface water natural organic matter in raw filtered water samples, XAD, and reverse osmosis isolates. Water Research 36 (9), 2357e2371. Pinhedo, L., Pelegrini, R., Bertazzoli, R., Motheo, A.J., 2005. Photoelectrochemical degradation of humic acid on a (TiO2)0. 7(RuO2)0.3 dimensionally stable anode. Applied Catalysis B: Environmental 57 (2), 75e81. Sanly, L., May, L., Rolando, F., Christopher, C., Ken, C., Mary, D., Rose, A., 2008a. Removal of humic acid using TiO2 photocatalytic process - Fractionation and molecular weight characterization studies. Chemosphere 72 (2), 263e271. Sanly, L., May, L., Rolando, F., Christopher, C., Mary, D., Rose, A., 2008b. TiO2 photocatalysis of natural organic matter in surface water: impact on trihalomethane and haloacetic acid formation potential. Environmental Science and Technology 42 (16), 6218e6223.
6140
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 3 1 e6 1 4 0
Shang, J., Li, W., Zhu, Y.F., 2003. Structure and photocatalytic characteristics of TiO2 film photocatalyst coated on stainless steel webnet. Journal of Molecular Catalysis A: Chemicstry 202 (1e2), 187e195. Siva, S., Madjid, M., 2010. Effects of UV/H2O2 advanced oxidation on chemical characteristics and chlorine reactivity of surface water natural organic matter. Water Research 44 (14), 4087e4096. Swietlika, J., Sikorska, E., 2004. Application of fluorescence spectroscopy in the studies of natural organic matter fractions reactivity with chlorine dioxide and ozone. Water Research 38 (17), 3791e3799. Tran, H., Scott, J., Chiang, K., Amal, R., 2006. Clarifying the role of silver deposits on titania for the photocatalytic mineralization of organic compounds. Journal of Photochemistry and Photobiology A: Chemistry 183 (1e2), 41e52. U.S.EPA, 1990. Methods for the Determination of Organic Compounds in Drinking Water Supplement I. U.S.EPA, Cincinnati, OH. Valdemar, I., Marta, O., Armando, C., 2009. Comparative characterization of humic substances from the open ocean,
estuarine water and fresh water. Organic Geochemistry 40 (9), 942e950. Wen, C., Paul, W., Jerry, A.L., Karl, B., 2003. Fluorescence excitation-emission matrix regional integration to quantify spectra for dissolved organic matter. Environmental Science and Technology 37 (24), 5701e5710. Xiao, S.H., Qu, J.H., Zhao, X., Liu, H.J., 2009. Electrochemical process combined with UV light irradiation for synergistic degradation of ammonium in chlorine-containing solutions. Water Research 43 (5), 1432e1440. Xue, G., Liu, H.H., Chen, Q.Y., Hills, C., Tyrer, M., Innocent, F., 2011. Synergy between surface adsorption and photocatalysis during degradation of humic acid on TiO2/ activated carbon composites. Journal of Hazardous Materials 186 (1), 765e772. Zhao, H., Liu, H.J., Hu, C.Z., Qu, J.H., 2009. Effect of aluminum speciation and structure characterization on preferential removal of disinfection byproduct precursors by aluminum hydroxide coagulation. Environmental Science and Technology 43 (13), 5067e5072.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 4 1 e6 1 5 1
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Water reuse: >90% water yield in MBR/RO through concentrate recycling and CO2 addition as scaling control Adriano Joss a,*, Claudia Baenninger a, Paolo Foa b, Stephan Koepke a, Martin Krauss a,e, Christa S. McArdell a, Karin Rottermann a, Yuansong Wei c, Ana Zapata d, Hansruedi Siegrist a a
Eawag, Swiss Federal Institute of Aquatic Science and Technology, Ueberlandstr. 133, 8600 Duebendorf, Switzerland DIIAR-Sezione ambientale, Politecnico di Milano, Piazza Leonardo da Vinci 32, 20133 Milan, Italy c Research Centre for Eco-Environmental Sciences, Chinese Academy of Sciences, 100085 Beijing, People’s Republic of China d Plataforma Solar de Almeria, Carretera Senes, 04200 Tabernas, Spain e UFZ e Helmholtz Centre for Environmental Research, Department Effect-Directed Analysis, Permoserstr. 15, 04318 Leipzig, Germany b
article info
abstract
Article history:
Over 1.5 years continuous piloting of a municipal wastewater plant upgraded with a double
Received 27 April 2011
membrane system (ca. 0.6 m3 d1 of product water produced) have demonstrated the
Received in revised form
feasibility of achieving high water quality with a water yield of 90% by combining
4 September 2011
a membrane bioreactor (MBR) with a submerged ultrafiltration membrane followed by
Accepted 5 September 2011
a reverse osmosis membrane (RO). The novelty of the proposed treatment scheme consists
Available online 14 September 2011
of the appropriate conditioning of MBR effluent prior to the RO and in recycling the RO concentrates back to the biological unit.
Keywords:
All the 15 pharmaceuticals measured in the influent municipal sewage were retained
Water reuse
below 100 ng L1, a proposed quality parameter, and mostly below detection limits of
Municipal wastewater
10 ng L1. The mass balance of the micropollutants shows that these are either degraded or
Reverse osmosis
discharged with the excess concentrate, while only minor quantities were found in the
Micropollutants
excess sludge. The micropollutant load in the concentrate can be significantly reduced by
Inorganic precipitation
ozonation. A low treated water salinity (<10 mM inorganic salts; 280 70 mS cm1) also confirms that the resulting product has a high water quality. Solids precipitation and inorganic scaling are effectively mitigated by lowering the pH in the RO feed water with CO2 conditioning, while the concentrate from the RO is recycled to the biological unit where CO2 is stripped by aeration. This causes precipitation to occur in the bioreactor bulk, where it is much less of a process issue. SiO2 is the sole exception. Equilibrium modeling of precipitation reactions confirms the effectiveness of this scalingmitigation approach for CaCO3 precipitation, calcium phosphate and sulfate minerals. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
Water of high quality and in sufficient quantities is an increasingly crucial resource for urban and decentralized
human settlements. While seawater desalination represents a solution for coastal regions, water reuse is a cheaper solution in most cases due to its lower energy requirement (Bartels et al., 2005; Van Houtte and Verbauwhede, 2008) and is also
* Corresponding author. Tel.: þ41 44 823 5408; fax: þ41 44 823 5389 E-mail address: [email protected] (A. Joss). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.09.011
6142
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 4 1 e6 1 5 1
2.
Materials and methods
2.1.
Treatment scheme
Fig. 1 and Table 1 give an overview of the treatment scheme and its principal operating parameters. The reactor was fed with municipal wastewater taken directly from the sewers of the town of Du¨bendorf (Switzerland) after primary treatment
Influent
Conditioning CO2, chloramine
100% 5400%
feasible at landlocked locations. The achieved quality, the acceptability of the product water and the overall treatment costs are paramount for the applicability of reuse schemes. The acceptability will not be further discussed here, since non-potable reuse is mostly dominated by cost competitiveness and potable reuse is mostly given if no alternative water resources are available (Anonymous, 2009; Lahnsteiner and Lempert, 2007). In terms of quality, a multi-barrier approach is typically required, and the pathogen, micropollutant and salinity rejection capacity is relevant. It is not a simple matter to discuss the overall cost competitiveness of treatment alternatives, due to the complexity of the processes involved, the multitude of (potential) options for optimization and the diverse influence of local conditions. Investment, maintenance, chemical and energy requirements and the disposal of residuals such as sludge and concentrate are considered to determine the costs of membrane-based systems. The present article discusses the water quality achievable with dual membrane systems. This treatment scheme is applied today at full scale for potable or high quality industrial reuse e.g. in Singapore and in Belgium (Bartels et al., 2005; Van Houtte and Verbauwhede, 2008). In the present study the RO concentrate discharge is significantly reduced thanks to its recycle to the biological unit. Reliable pathogen removal by dual membrane systems has been shown and is not discussed further here (Comerton et al., 2005). The quality of the product water is discussed by presenting mass balances for organic micropollutants and inorganic ions. Pharmaceuticals, bisphenol A and the anticorrosive benzotriazoles were chosen as representatives of relatively small and often charged contaminants that are quite resistant to biodegradation and are most likely to pass through membranes. Nitrosamines (such as NDMA) were also included in the study since these are formed during chloramination of surface water and wastewater (Schreiber and Mitch, 2006a; Zhao et al., 2008). Due to scaling on the RO membrane, the fate of inorganic ions is crucial for the feasible water yield, i.e. the amount of product water. The proposed treatment scheme allows controlled precipitation in the biological reactor upstream of the reverse osmosis membrane (RO) by lowering the pH of the feed in the RO with CO2,aq and then recycling the RO concentrate back to the MBR where supersaturation is induced by stripping of CO2 with aeration, leading to salt precipitates in the MBR. It is shown that this allows significant water yields with minimal use of chemicals. The consequent reduction of discarded concentrates represents an advantage, especially at locations where concentrate disposal to the environment is not permitted (e.g. landlocked sites where the concentrate needs to be treated).
Anox
Aerob
Permeate
RO
90%
Excess sludge
1% 10% Concentrate
400%
Ozonation 90%
Fig. 1 e Schematic outline of the experimental setup for upgrading municipal wastewater for high quality water reuse. Percent indications show flowrates relative to the influent. Anox: anoxic compartment; Aerob: aerated compartment housing the UF membrane; RO: reverse osmosis unit (see S1 for details); : sampling points.
including a screen, a grid chamber and a primary settler. The feeding flow rate of 0.64 0.24 m3 d1 was controlled by the performance of the RO. The unit had been operated from March 2008 to November 2009, a total of 1.7 years (614 days). Stable operation of the entire setup was reached after 80 days. The two-lane MBR was subdivided into an anoxic part of 440 L and an aerobic compartment of 310 L. The sludge was recycled at a flow rate four times the influent flow from the aerobic to the anoxic compartment. The submerged ultrafiltration (UF) hollow-fiber membranes (eight Zenon ZW10 modules totaling 8 m2) were housed in the aerobic compartment. The cross-flow aeration of the UF membrane sufficed to cover the biological oxygen demand, resulting in a soluble oxygen concentration in the aerobic compartment >5 mgO2 L1. The hydraulic residence time was 15 3.5 h and the sludge age ca. 95 d. The sludge concentration in the MBR was 5.8 2.3 gTSS L1. The MBR effluent was conditioned in a 20 L buffer tank with chloramine to minimize biofouling of the RO (10e15 gNH2Cl m3) and CO2 to reduce inorganic scaling (0.8 kgCO2 m3 permeate ; due to the unknown losses to the headspace, the actual CO2 requirement should be estimated on the basis of the buffering capacity of the MBR effluent). The RO unit (Figure S1) consisted of a single multistage centrifugal pump (dp-Pumps model DPVF2-180) equipped with a frequency modulator for pressure generation and internal recycling, two spiral-wound RO modules (Dow Filmtec NF902540)1 operated in parallel with a 5.2 m2 surface area and an EWS OS3050 controller (EWS International -Hertogenbosch, Netherlands). The internal recycle was controlled by a Bu¨rkert flow controller type 8650 (Bu¨rkert GmbH, Ingelfingen, Germany). The rather low specific permeate flux of 4.5 1.8 Lm2 h1 was probably due to the RO being operated at only 5 bar and the irreversible fouling caused by inappropriate chemical cleaning strategies employed during the first year of operation. Until day 522, the RO was operated at a high
1
The NF90 membrane is designated as nanofiltration by the supplier. According to the very high retention of multivalent as well as monovalent ions (Fig. 6), the term reverse osmosis is deemed more appropriate in the present publication.
6143
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 4 1 e6 1 5 1
Table 1 e Operating conditions of the dual membrane setup (n.a. [ not available). Conductivity [mS cm1] Influent MBR effluent RO permeate RO concentrate
1.3 3.2 0.3 6.0
0.3 0.6 0.1 0.9
Organic content TOC [mgC L1] 140 40 10 5 <0.5 19 9
cross-flow to permeate flux ratio of 80, while this was reduced to 15 afterward to test the impact on permeability. The RO was operated at a transmembrane pressure (TMP) of 5 to 6 bar with a pressure drop of 1 bar along the module at a cross-flow of 1.9 0.1 m3 h1, dropping proportionally with reduced cross-flow. The averaged net permeate flow was 23.4 9.2 L h1. The concentrate recycle rate to the biological unit was 29.4 6.4 L h1 and the concentrate discharged about 10 4% of the recycle rate and controlled proportional to the rate of permeate production. The chemical maintenance is outlined in Table S1. Each cleaning event was composed of two consecutive steps each lasting 20e30 min. The UF was cleaned in situ by intermittent backwashing (30 s every 2 min, 8 consecutively) while submerged in sludge. The cleaning solution for the UF was disposed into the MBR. The RO was cleaned at 0.5 bar TMP by flushing the pressure vessels with cleaning solution, followed by soaking for 20 min at an exchange rate of the chemical solution of 1 L min1 and maximum cross-flow. The spent chemical cleaning solution was discarded. The permeability was calculated according to Shirazi et al. (2010): kW ¼
JW DP Dp
(1)
where kW is the permeability in L m2 h1 bar1, JW the measured specific flux in L m2 h1, DP the transmembrane pressure in bar and Dp the osmotic pressure difference over the membrane in bar. The osmotic pressure of the concentrate was calculated from the conductivity measurement according to the correlation shown in Figure S2, obtained by direct measurement during the experimental phase. The ozonation reactor consisted of a bubble column of 2 m (diameter 63 mm, volume 6 L) operated at an HRT of 0.5e1.5 h. The concentrate of the RO was fed from the top (countercurrent bubble column). The effluent of the ozonation column was recycled to the anoxic compartment of the biological unit. The concentrate discharge occurred prior to the ozonation column. An ozone generator of type Chemodata 1.0 g/Hmax (Chemonorm AG, Altendorf, Switzerland) fed with pure oxygen provided an ozone dose of 5 corresponding to 0.25 and and 16 gO3 m3 permeate 0.85 gO3 gDOC, and hence within the typical range for micropollutant removal of secondary effluent. The off-gas was confirmed to be ozone free by online measurement provided by a BMT 964 ozone analyzer (BMT Messtechnik GmbH, Stahnsdorf, Germany). The ozonation column was operated from day 259 to 381. The reactor was controlled by a Simatic S7 programmable logical controller (Siemens, Germany) and operated via a PC-
pH
Temperature [ C]
Oxygen [mgO2 L1]
19 2.3 20 2.3 26 2.7 26.5 2.8
0.0 >5 n.a. n.a.
7.7 8.1 5.8 6.8
0.4 0.3 0.5 0.4
based SCADA user interface (CitectSCADA, Schneider Electric, Australia) storing process data every 10 s. The data was evaluated with Matlab (Mathworks, USA).
2.2.
Analysis and sampling of micropollutants
The investigated micropollutants were present in the original municipal wastewater and no spiking was performed. Fig. 1 illustrates the locations of the sampling points. The micropollutant sampling was performed with flowproportional samplers cooled at 4 C with a maximum sampling interval of 30 min. Each sampling campaign was composed of 5 consecutive 24 h composite samples analyzed separately. The sampling campaigns were performed from October 20 to 25, 2008 (without ozonation; operation day 234e239) and from March 9 to 14, 2009 (with ozonation; operation day 374e379). For nitrosamines, three consecutive 24 h composite samples were analyzed (without ozonation November 5e7 2008; with ozonation December 9e11 2008). The sample analysis for micropollutants was performed by online solid phase extraction (Oasis HLB, Strata XCW/XAW, ENVþ) and LC-MS/MS on a Thermo Electron TSQ Qantum Ultra (Hollender et al., 2010). Nitrosamines were analyzed by offline SPE (Oasis HLB above Bakerbond Carbon) and LC-high resolution MS/MS (LTQ orbitrap; Krauss and Hollender, 2008). The quantification limits for micropollutants were around 10 ng L1 in general and between 0.8 and 3 ng L1 for the nitrosamines. The variation ranges of the influent and product water concentrations over the five days (Figs. 2 and 3) were used as a measure of the typical variation in the measured values. Contaminant mass flows were calculated by multiplying the measured compound concentration of each sample with the respective daily flow volume. The daily volumetric flows of permeate, excess sludge, concentrate discharge and concentrate recycle were obtained by evaluating the online data measurement (logged at 10 s intervals by the SCADA), with the influent volume calculated as the sum of all outputs (permeate, excess sludge and concentrate). The MBR volume was constant, so changes in the soluble compound mass stored in the reactor volume were calculated from measured concentrations. The accuracy of the mass balances (Figs. 4 and 5) was estimated by Monte Carlo analysis according to Ternes and Joss (2006), assuming a 10% standard error for flow measurements and 20% for the analytical measurement of the compound concentration. Both errors were assumed to be normally distributed. The represented accuracy corresponded to the standard deviation of the results from 5000 Monte Carlo simulation runs.
6144
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 4 1 e6 1 5 1
5
10
Influent Concentrate Permeate 4
Concentration [ng.L-1]
10
3
10
2
10
1
B At 5- enz e no M e t otri lol h y az l-B ol Bi en e s Ca p z o rb h en t. a o Cl ma l- A a r ze ith p i ro n e Di myc cl of in en I a b M ef upr c am o in fen M sä u et op re Na r ol Pa pr ol ra oxe ce n t Pr am im ol Su Pro ido lfa p r ne N4 me ano -A to x lo l ce az ty ol l- S e ul fa Su Sot . a lf l Tr a py ol im r i et din ho pr im
10
Fig. 2 e Comparison of micropollutant concentration in the influent municipal wastewater (no spiking) and in the product water without ozonation of the recycled concentrate. The error bars indicate the standard deviation of the five sampled days (days 234e239).
2.3. Analysis and sampling of inorganic ions and precipitates Inorganic carbon (TIC) was measured with an analyzer model TOC/TIC Analyzer IL 550 (Hach-Lange, Germany).
3 Anions (Cl, SO2 4 , PO4 and NO3 ) were measured with an ion chromatograph model Metrohm 881 Compact IC with a separation column Metrosep A Supp 4 (Metrohm AG, Switzerland). Cations (Naþ, Kþ, Ca2þ and Mg2þ) with an inductively coupled plasma atomic emission spectroscope
5
10
Influent Concentrate Permeate 4
Concentration [ng.L-1]
10
3
10
2
10
1
A Be te n o z 5- Be a fib lol M nz r a et o t Ca h yl tria e rb -Be z ol Cl am nz o a r a z t. ith ep Cl rom in e in y da ci D m n Er ic l ycin yt ofe hr n om ac M Ib yci ef up n am r o i fe M nsä n et ur o e N pr Pa a p olol ra rox ce e Pr tam n i P m ol Ro rop idon xit ra e h no Su S rom lo l u lfa lfa yc Ac me py in et th r id y l- ox in Su a z lfa ol. m Tr S e t. im ot et alo ho l pr im
10
Fig. 3 e Comparison of micropollutant concentration in the influent municipal wastewater (no spiking) and in the product water with ozonation of the recycled concentrate (0.85 gO3 gL1 DOC ). The error bars indicate the standard deviation of the five sampled days (days 374e379).
6145
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 4 1 e6 1 5 1
Mas s relativ e to influent [% ]
150
Degraded Excess sludge Concentrate Permeate
100
50
B At 5- enz e no M e t otri lol h y az l-B ol Bi en e Ca sp zo rb h en t. a o Cl ma l- A a r ze ith p i ro n e Di myc cl of in en I a b M ef upr c am o in fen M sä u et op re Na r ol Pa pr ol ra oxe ce n t Pr am im ol Su Pro ido lfa p r ne N4 me ano -A to x lo l ce az ty ole l- S ul fa Su Sot . a lf l Tr a py ol im r i et din ho pr im
0
Fig. 4 e Micropollutant mass balance over the entire sampling campaign of five days without ozonation. Error bars indicate the accuracy for each fraction as estimated by Monte Carlo analysis (sampling days 234e239).
model IC OES-Ciros (Spectro Analytical Instruments GmbH, Germany). The mass balance of the major ion species (Cl, Naþ, Kþ, 2þ Ca , Mg2þ, SO2 4 and Si) was determined from 15 consecutive 24 h composite samples starting on May 6, 2009 (day 432e446).
160
Mas s relativ e to influent [% ]
140
Elemental analysis was performed on 2e3 cm2 membrane using X-ray fluorescence (Spectro Xeposþ, Kleve, Germany). For that purpose the spiral-wound RO module was dissected and 5 g of the membrane were pressed into 32 mm pellets. Background concentrations were obtained from
Degraded Ozonation Excess sludge Concentrate Permeate
120 100 80 60 40 20
A Be ten o z 5- Be afib lol M nz r a et o t Ca h yl tria e rb -Be zol Cl am nzo a r a z t. ith ep Cl rom ine in y d a ci Di my n Er c l c i yt ofe n hr n om ac M Ib yci ef up n am r o i fe M nsä n et ur o e N pr Pa a p olol ra rox ce e Pr tam n im o P l Ro rop idon xit ra e h no Su S rom lo l lfa ulfa yc Ac me py in et th r id yl- ox in Su a z lfa ol. m Tr S e t. im ot et alo ho l pr im
0
Fig. 5 e Micropollutant mass balance over the entire sampling campaign of five days with ozonation (0.85 gO3 gL1 DOC ). Error bars indicate the accuracy for each fraction as estimated by Monte Carlo analysis (sampling days 374e379).
6146
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 4 1 e6 1 5 1
measurements of blank membranes and the measurements of the used membranes were corrected accordingly.
2.4.
Saturation indices
The PhreeqC simulation software package (Parkhurst and Appelo, 1999) with an associated standard thermodynamic database ( phreeqc.dat) was used to model the precipitation reactions. Saturation indices (SI) were calculated by the simulation package as follows: SI ¼ logIAP logK
(2)
where the ionic activity product IAP is the product of ion activities involved in the dissociation equation, and K is the temperature-dependent equilibrium constant. Positive SI values were taken to indicate oversaturation (precipitants will form and scaling can occur), zero indicated equilibrium (no precipitation) and negative values showed under-saturation (precipitants will dissolve). The solubility constant for vaterite (CaCO3) of log K ¼ 7.91 given by Sawada (1997) was added to the model database. The error in the charge balance of the modeled mixtures (Table 3) was always <7% of the total ionic strength, thus in the range of the expected analytical accuracy.
3.
Results and discussion
3.1.
Removal of micropollutants
Most organic micropollutants are degraded and/or retained to below the detection limit (10 ng L1) by RO, except for small and polar compounds such as the anticorrosive benzotriazoles, as well as some rather persistent pharmaceuticals, namely propranolol, diclofenac and carbamazepine (Figs. 2 and 3). This is in good agreement with the available literature (Hollender et al., 2009; Joss et al., 2005; Radjenovic et al., 2008). A comparison of the concentrations in the concentrate and permeate shows that high retention was achieved by RO. All measurable concentrations are further reduced by inserting an ozonation step in the concentrate recycle loop (Fig. 1).
Table 2 e Elemental analysis (XRF) of the precipitations found on the RO membrane at the end of the experiments (21 samples). Element Ca Si P Mg Al Fe Zn Sr Cu Pb Cr Br
Amount mg m2 membrane 16,500 850 430 360 224 144 106 33 14 11 6 2
5,400 440 180 180 143 47 31 10 4 3 2 1
The mass balance shows that without ozonation the retained compounds are mostly degraded but are also discharged in the concentrate, while the excess sludge contains only a minor part of the load (Fig. 4). According to the accuracy estimation performed with Monte Carlo simulations, the degradation of carbamazepine and primidone is not significant, while some degradation is probable for diclofenac (Joss et al., 2005; Kasprzyk-Hordern et al., 2009). This is speculated to be due either to the very high sludge age (ca. 100 days) or to the exposure to anaerobic conditions in the anoxic tank (enhanced biological phosphorus removal was observed; data not shown). The reason why for some compounds the sum of all effluents resulted significantly higher than 100% (i.e. benzotriazoles, carbamazepine, primidone and propranolol) is suspected to be due to underestimation of the influent load due to matrix effects (i.e. systematic errors not accounted for in the accuracy estimation). Fig. 5 shows that an ozone dose of 0.85 gO3 gDOC effectively eliminates the micropollutants and reduces the discharged amount in the concentrate. Since the DOC concentrations are high in the concentrate, a relatively high ozone dose is required to achieve good elimination of micropollutants. However, the elimination rates obtained were comparable to the results from ozonation experiments on treated municipal wastewater for similar concentrations of ozone relative to DOC (Hollender et al., 2009). Table S2 shows the elimination rates obtained for two ozone doses, and confirms that 0.25 gO3 g1 DOC is insufficient for substantial removal. A similar setup combining nanofiltration with ozonation had previously been proposed but not tested (Ernst and Jekel, 1999) Since nitrosamines can be formed from nitrogencontaining precursors during chloramination and ozonation, their concentrations were also measured (Hollender et al., 2009; Krauss et al., 2010). Nitrosamine concentrations in the MBR influent were 15e32 ng L1 for NDMA, and 4e22 ng L1 for NMOR, and generally below 10 ng L1 for the other nitrosamines. These levels indicate a rather low level of nitrosamine load typical for Swiss wastewater. The NDMA elimination rates in the MBR were 36 and 64% for treatment without respectively with ozonation, those of the other nitrosamines between 20 and 100% and thus in the range of elimination rates reported for lower ng L1 levels in the studies cited above. The addition of 10e15 mg L1 chloramine did not result in a significant formation of NDMA, while up to 23 ng L1 of Nnitrosomorpholine and up to 12 ng L1 of other nitrosamines (N-nitrosopiperidne, N-nitrosodiethylamine and N-nitrososdibutylamine) were formed. The average rejection of NDMA by the RO membrane was about 80% and those of the other nitrosamines between 70 and close to 100%, reducing permeate concentrations below 7 ng L1 for NDMA, below 24 ng L1 for NMOR, and below or close to the limit of quantification (<3 ng L1) levels for the other nitrosamines. These values indicate that nitrosamines are rejected to some extent, but not completely by reverse osmosis as reported by SteinleDarling et al. (2007). Nevertheless, NDMA concentrations in the permeate were already lowered below the interim maximum acceptable concentration of 9 ng L1 for drinking water established by the Ontario Ministry of Environment (OMot, 2003) and were close to the public health goal of
6147
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 4 1 e6 1 5 1
Table 3 e Water composition as measured in 15 samples taken between May 6 and 20, 2009 (days 432e446) is indicated as average ± standard deviation. The saturation index is given for all compounds that may potentially precipitate in the process based on the modeling of the ion composition. *: in the model pH and total inorganic carbon (TIC) concentrations have been changed for the concentrate to represent typical operating conditions rather than the specific conditions measured during the observation period when the pH was inappropriately high (7.4 ± 0.3). Influent wastewater
Wastewater þ recycle
MBR effluent
Permeate
Concentrate
Flow Temperature pH
L h1 C
26.1 3.8 18 0.6 7.4 0.5
45 6.5 20.6 1 8.1 0.1
44.6 6.5 20.6 0.8 8.2 0.1
23 1.7 26.2 1.0 6.2 0.2
2.7 0.7 24.6 1.0 6.8*
TIC Cl Naþ Kþ Ca2þ Mg2þ SO2 4 Silica species PO3 4 NO 3
mgC L1 mgCl L1 mgNa L1 mgK L1 mgCa L1 mgMg L1 mgS L1 mgSi L1 mgP L1 mgN L1
107 15 138 25 125 20 14.6 1.4 118 16 20.4 1.6 37 7 6.7 0.6 2.3 0.5 <1
233 19 472 13 365 16 36 1 277 7 66 3 133 4 17.6 0.6 2.7 0.1 2.2 0.2
191 22 465 34 352 35 35.5 3.1 248 18 63 6 132 11 17 2 1.7 0.7 2.6 1.9
16 2.5 27 4 36 4 3.6 0.3 1.5 0.3 0.4 0.1 1.6 0.2 0.8 0.2 0.1 0.5 0.3
360* 930 170 693 60 65 7 494 52 127 15 263 35 32 3.5 3.1 1.1 4.1 1
Saturation indices Calcite Aragonite Vaterite Dolomite Hydroxyapatite Chalcedony Quartz Chrysotile Sepiolite Talc
CaCO3 CaCO3 CaCO3 CaMg(CO3)2 Ca5(PO4) 3OH SiO2 SiO2 Mg3Si2O5(OH)4 Mg2Si3O7.5OH$3H2O Mg3Si4O10(OH) 2
0.5 0.3 0 0.5 4.3 0 0.5 -6.0 -3.9 -2.3
1.8 1.6 1.2 3.2 8.2 0.4 0.8 0.5 0.9 4.9
1.7 1.6 1.2 3.2 7.8 0.4 0.8 1 1.3 5.4
3.4 3.6 4.0 7.0 14 1 0.6 19 14 17
0.8 0.6 0.2 1.3 3.6 0.6 1.1 5.6 2.9 0.7
3 ng L1 of California (COoEHH, 2006). Ozonation of the reverse osmosis concentrate resulted in a slight, but not statistically significant increase of NDMA concentrations, while other nitrosamines were not formed. Overall, the low micropollutant content in the product water confirms the suitability of the treatment scheme for many reuse purposes (e.g. potable or industrial with high quality requirements), with or without concentrate ozonation.
3.2.
Salt retention and precipitation
The average RO permeate conductivity of 280 70 mS cm1 is significantly below that of the raw wastewater at 1300 300 mS cm1 (Table 1). Monovalent ions are retained at >70% relative to the influent concentration while all multivalent ions are retained and reduced to 2 mg L1 in the permeate (Table 3). Thus, the product water would also be considered as suitable for unrestricted reuse in terms of its salinity (e.g. potable reuse). The RO concentrate has a conductivity of 6000 900 mS cm1 þ þ 2þ 2þ 2 with HCO 3 ; Na ; Cl ; NO3 ; K ; Ca ; Mg ; SO4 and silicates are the dominant ions. Table 2 lists elemental compositions of the bulk precipitations found on the RO membrane after termination of the experiments (i.e. after 1.7 years of operation) determined with XRF. Calcium is clearly the prevalent element. Si, P and Mg were also present in significant amounts. The other metals were considered of minor importance ð< 10 mMol m2 membrane Þ.
During the third sampling campaign (from May 6 to 20, day 432 to 446), the concentration of the most significant ions was monitored at the main points of the integrated process, and the average values were used to predict the possible precipitations with the PhreeqC software. Table 3 shows the water composition and the major scalants to be expected according to the saturation indices modeled with highly positive values, i.e. CaCO3, SiO2 and magnesium silicates (zero indicates equilibrium, negative saturation indices indicate dissolution of precipitates). CaCO3 scaling is a major concern in membrane desalination (AlShammiri and AlDawas, 1997; Pena et al., 2010). It is mostly countered by adding sulfuric acids, hydrochloric acids or anti-scalants prior to RO (Baker et al., 1997; Ning and Netwig, 2002). In the present study, CO2 addition is used instead of sulfuric acid and the lowering of pH results in a reduction of CO2 3 concentration (in spite of the increase of total inorganic carbon species), and thus a lower saturation index for the carbonate precipitates. This effect was reversed in the biological unit, where CO2 was removed by stripping with air, the pH increased and significant CaCO3 precipitation occurred (Figure S3). According to the modeling, a pH of around 6.1 would be required to reach negative saturation indices in the concentrate for all the considered compounds except SiO2. Nevertheless, according to Lee and Lee (2005), cross-flow allows avoidance of precipitation in spite of a slight oversaturation, so that stable operation may also be achieved at a pH slightly higher than 6.1.
6148
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 4 1 e6 1 5 1
Several studies indicate that calcium phosphate precipitation may be a limiting factor for membrane performance and RO water yield (Bartels et al., 2005; Greenberg et al., 2005; Katz and Dosoretz, 2008; Steiner et al., 2010). Table 3 shows that the treatment scheme also allows the operator to maintain a significantly higher saturation index for hydroxyapatite in the biological unit compared to the membrane compartment, thus driving the precipitation of hydroxyapatite to occur preferentially in the bioreactor. XRF analysis performed on the RO membrane shows P-containing precipitates to be of minor importance (Table 2). According to the modeling of saturation indices a further decrease in pH on the membrane as required for impeding CaCO3 precipitation should also effectively mitigate hydroxyapatite precipitation. SiO2 precipitation cannot be controlled by acid conditioning nor removed via an acid clean. Thus the calculated saturation indices are slightly higher in the RO unit due to the higher Si concentration. Accordingly, it is expected that these precipitates will be a major factor limiting water yield, even if operating at a low pH in the RO feed to avoid significant CaCO3 precipitation. Magnesium silicate precipitations are expected to occur only at pH values typical for the membrane bioreactor (i.e. pH 8); i.e. dosing Mg into an appropriately configured compartment of the biological unit, so this can potentially result in a sink for SiO2. Contrary to the literature findings (Lee and Lee, 2005), CaSO4 (anhydrite and gypsum) were not found to be of concern in the present study and were significantly undersaturated in the RO as well as in the bioreactor (SI <0.5). Since the sulfate content of the influent wastewater was in a typical range (7e17 mgS L1; Tchobanoglous et al., 2003), this study indicates that CaSO4 precipitation is expected to be
a relevant scalant only if influent concentrations are significantly above that of typical municipal wastewater, e.g. due to intrusion by industrial wastewater or seawater. The saturation indices on the RO were calculated with the ion concentrations measured in the bulk medium, thus not accounting for the concentration polarization on the membrane. Nevertheless, according to Shirazi et al. (2010) it is estimated that this effect would not change the conclusions drawn under the experimental conditions: the ion concentration on the membrane is estimated to be in the range of 1.2e1.5 times higher than in the bulk solution, thus increasing the saturation indices on the membrane by only 0.2e0.5. The mass balance of specific ions confirms significant precipitation of CaCO3 in the MBR sludge or walls (Fig. 6): ca. 30% of the influent Ca2þ is accumulated in the MBR as precipitate (Figure S3). The accuracy of the mass balance does not allow recognition of silica precipitation in the RO: the influent and effluent loads correspond well (i.e. no significant silica accumulation in the reactor).
3.3.
RO membrane performance
Over the entire experimental period, the unit was operated with an overall water yield of 86.4 9.3%, which corresponds to producing only 12 7.3% concentrate for disposal. This is about half of what is normally discarded (Bartels et al., 2005; Van Houtte and Verbauwhede, 2008). This confirms that the proposed scheme solves some major scaling issues typically experienced with other installations. Nevertheless, Fig. 7 shows that permeability declines slowly during the experimental period, resulting in a rather low long-term permeability. According to the modeling of the saturation indices as
Fig. 6 e Mass balance of the principal ions. Significant accumulation in the reactor is seen only for Ca2D (effluent load smaller than the influent). All values are expressed as amounts relative to the volume of influent (sampling days 432e446).
6149
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 4 1 e6 1 5 1
5 Single clean Low CO2 dosing Double clean Low cross flow
Permeability [L· m-2· h-1· bar-1]
4
3
2
1
0 0
100
200
300 400 Days of operation
500
600
Fig. 7 e Permeability of the RO membrane over the entire experimental period calculated according to Eq. (1). From day 391 to day 430 CO2 dosing was either switched off or significantly lowered. From day 430 onwards, chemical cleaning was performed twice at weekly intervals (Table S1). From day 525 onwards the cross-flow was reduced from 1.7 to 0.5 m3 hL1. Lines show linear regressions of data subsets.
well as the elemental analysis of the deposits on the membrane (Tables 2 and 3), it is assumed that this was mainly due to insufficient CO2 dosing, since the pH on the RO averaged 7.0 0.4 during the entire study, resulting in significant CaCO3 precipitation. The observation that increased chemical cleaning of the membrane increased the permeate flux (Fig. 7, from day 430e515) confirms that acid-soluble deposits on the membrane contributed to lowering the permeate flux. It is thus argued that operating the RO at a lower pH value would significantly reduce the decline shown in Fig. 7. Several studies identify biofouling and microbial growth on the membrane as a major problem (Al-Ahmad et al., 2000; Herzberg et al., 2010). Nevertheless, based on the deposits found on the membrane in this study, it is argued that the decline in permeability shown in Fig. 7 is due primarily to inorganic precipitants. Contrary to reports in the literature (Vrouwenvelder et al., 2010), in this study the permeability loss did not correlate with a pressure drop along the feed channel of the membrane (constantly 1.0 0.1 bar at a crossflow of 960 60 L h1 and module). This is taken as a further indication that the permeability loss was primarily due to membrane scaling, rather than by biofouling occurring on both membrane and spacer within the feed channel. It is considered possible that the relatively high dosage of chloramine (10e15 gNH2Cl m3) as well as the frequent chemical cleaning of the membrane (weekly) helps to keep biofouling at a low level. The primary scope of the present work was to focus on inorganic scaling under conditions of high water yield. Biofouling was not the focus. Chloramine is typically dosed in the range of 2e5 gNH2Cl m3 and chemical cleaning is performed monthly (Bartels et al., 2005; Van Houtte and Verbauwhede, 2008; Xu et al., 2010). It is acknowledged that an overall optimization of the scheme may allow these biofouling control techniques to be reduced.
3.4.
Technical feasibility
The present work shows that conditioning with CO2 combined with concentrate recycling allows all relevant scalants except SiO2 to be controlled (the relevance of the latter for the overall performance could not be confirmed, but indirect evidence pointed to the greater significance of other scalants), and thus increasing the water yield. Assuming that a pH in the range of 6.4e6.5 on the RO is sufficient to avoid significant precipitation (not considering SiO2), and according to the buffering capacity given by the ion composition in Table 3, conditioning with ca. 2 kgCO2 m3 permeate is estimated to be required. This quantity of CO2 coming from a stripping unit must be purified to 50% purity and recycled to the RO unit. The anoxic compartment joining the influent wastewater and concentrate recycles (possibly also the internal sludge recycle) should thus be considered for upgrading to a combined stripping and precipitation unit. The degree of pH lowering needed in the RO to achieve sustainable operation is regarded as crucial for its economic feasibility. The present study is based purely on the calculation of solubility equilibria. However, several studies have shown that kinetic aspects are actually process-limiting, thus explaining why stable operation may also be feasible for some precipitates with slight oversaturation (Lee and Lee, 2005; Udert et al., 2003). Steiner et al. (2010) showed that this degree of oversaturation depends strongly on the composition of the mixture. The chemical cleaning and cross-flow regime are also considered to be relevant parameters here, besides the pH level in the RO. Finally, the concentrate recycling requires appropriate dimensioning of the UF membrane as well as a reactor configuration amenable to scalant precipitation. It is thus concluded that the economic feasibility of the present treatment scheme can be
6150
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 4 1 e6 1 5 1
assessed only by considering all involved treatment steps and their interdependencies.
4.
Conclusion
Double-membrane treatment schemes allow municipal wastewater to be upgraded to a quality suitable for many reuse purposes, with micropollutant concentrations mostly below detection limits of 10 ng L1 and salinity of <300 mS cm1. The formation of nitrosamines as byproducts of chloramination and ozonation was counterbalanced by biological degradation and reverse osmosis rejection resulting in levels below reference limits for NDMA. The continuous piloting at a water yield of 90% showed that CaCO3, hydroxyapatite, CaMg(CO3)2, SiO2 and Mg-silicates may exceed their solubility concentrations, thus potentially leading to scaling of the RO membrane. Calcium sulfate was always found to be significantly below its solubility limit. Lowering the pH with CO2 prior to membrane filtration was shown to be highly effective at mitigating inorganic scale formation, but the pH of 6.8 achieved in this study was still shown to be too high to avoid a slow permeability loss due to longer-term inorganic scaling. According to the simulation modeling results, lowering the pH down to between 6.1 and 6.5 should suffice to avoid all relevant precipitations except SiO2. The presented treatment scheme was shown to reduce the concentrate discharge by at least a factor of two from the current state of the art, while running with reduced requirements for anti-scalant chemicals. The present work does not allow the authors to draw conclusions regarding the economic competitiveness of the proposed set-ups compared to other ones, since several optimization options have not yet been thoroughly tested. Among these are the exact pH adjustment required for the RO, the impact of cross-flow, the chemical cleaning strategy and the permeate-to-recycle ratio. Further suitable solutions could be found for properly mixing influent and concentrate in a way allowing the controlled precipitation of salts (i.e. within a controlled crystallization reactor) while stripping CO2. Since around 2 kgCO2 m3permeate is estimated to be required, the recycling of the stripped CO2 back to the RO unit is expected to represent an economic solution at full-scale operation. Nevertheless, it is argued that at some locations (e.g. continental ones) the reduction of the concentrate volume for disposal by a factor two or more (i.e. from typically 20%e<10% of the treated water) as well as the fact that the scaling control does not produce any salinity increase (i.e. by not involving the addition of sulfuric or hydrochloric acid or other antiscalants) can be an advantage, thus justifying further research on the applicability of the treatment scheme presented here.
Acknowledgments Norit Process Technology NV (Enschede, The Netherlands) is kindly acknowledged for providing the nanofiltration unit. The European Commission is acknowledged for co-funding
the RECLAIM WATER Project under contract number 018309 in the Global Change and Eco-system sub-priority of the 6th Framework Programme for Research and Technological Development. The support from the Chinese Scholarship Council is acknowledged. Chemonorm AG (Altendorf, Switzerland) is kindly acknowledged for providing the ozone generator. MMS AG (Urdorf, Switzerland) supplied the osmotic pressure measurement equipment. The support of Dr. Ralf Ka¨gi and Brian Sinnet for XRF analysis is acknowledged. Stephen Tait (AWMC, Queensland University, Australia) is acknowledged for valuable discussions.
Appendix. Supplementary data Supplementary data related to this article can be found online at doi:10.1016/j.watres.2011.09.011.
references
Al-Ahmad, M., Aleem, F.A.A., Mutiri, A., Ubaisy, A., 2000. Biofuoling in RO membrane systems part 1: fundamentals and control. Desalination 132 (1e3), 173e179. AlShammiri, M., AlDawas, M., 1997. Maximum recovery from seawater reverse osmosis plants in Kuwait. Desalination 110 (1e2), 37e48. Anonymous, 2009. In: Wikipedia (Ed.), Toowoomba Water Futures Referendum 2006. Wikipedia. Baker, J.S., Judd, S.J., Parsons, S.A., 1997. Antiscale magnetic pretreatment of reverse osmosis feedwater. Desalination 110 (1e2), 151e165. Bartels, C.R., Wilf, M., Andes, K., Iong, J., 2005. Design considerations for wastewater treatment by reverse osmosis. Water Science and Technology 51 (6e7), 473e482. Comerton, A.M., Andrews, R.C., Bagley, D.M., 2005. Evaluation of an MBR-RO system to produce high quality reuse water: microbial control, DBP formation and nitrate. Water Research 39 (16), 3982e3990. COoEHH, 2006. Public Health Goal for N-Nitrosodimethylamine and Cadmium in Drinking Water. Ernst, M., Jekel, M., 1999. Advanced treatment combination for groundwater recharge of municipal wastewater by nanofiltration and ozonation. Water Science and Technology 40 (4e5), 277e284. Greenberg, G., Hasson, D., Semiat, R., 2005. Limits of RO recovery imposed by calcium phosphate precipitation. Desalination 183 (1e3), 273e288. Herzberg, M., Berry, D., Raskin, L., 2010. Impact of microfiltration treatment of secondary wastewater effluent on biofouling of reverse osmosis membranes. Water Research 44 (1), 167e176. Hollender, J., Singer, H., Hernando, D., Kosjek, T., Heath, E., 2010. In: Fatta-Kassinos, D., Ku¨mmerer, K., Bester, K. (Eds.), Xenobiotics in the Urban Water Cycle. Springer, Netherlands, pp. 195e211. Hollender, J., Zimmermann, S.G., Koepke, S., Krauss, M., McArdell, C.S., Ort, C., Singer, H., von Gunten, U., Siegrist, H., 2009. Elimination of organic micropollutants in a municipal wastewater treatment plant upgraded with a full-scale postozonation followed by sand filtration. Environmental Science & Technology 43 (20), 7862e7869. ˜ bel, A., McArdell, C.S., Ternes, T.A., Joss, A., Keller, E., Alder, A.C., GA Siegrist, H., 2005. Removal of pharmaceuticals and fragrances in biological wastewater treatment. Water Research 39 (14), 3139e3152.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 4 1 e6 1 5 1
Kasprzyk-Hordern, B., Dinsdale, R.M., Guwy, A.J., 2009. The removal of pharmaceuticals, personal care products, endocrine disruptors and illicit drugs during wastewater treatment and its impact on the quality of receiving waters. Water Research 43 (2), 363e380. Katz, I., Dosoretz, C.G., 2008. Desalination of domestic wastewater effluents: phosphate removal as pretreatment. Desalination 222 (1-3), 230e242. 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. Analytical Chemistry 80 (3), 834e842. Krauss, M., Longree, P., Van Houtte, E., Cauwenberghs, J., Hollender, J., 2010. Assessing the fate of nitrosamine precursors in wastewater treatment by physicochemical fractionation. Environmental Science & Technology 44 (20), 7871e7877. Lahnsteiner, J., Lempert, G., 2007. Water management in Windhoek, Namibia. Water Science and Technology 55 (1e2), 441e448. Lee, S., Lee, C.H., 2005. Scale formation in NF/RO: mechanism and control. Water Science and Technology 51 (6e7), 267e275. Ning, R.Y., Netwig, J.P., 2002. Complete elimination of acid injection in reverse osmosis plants. Desalination 143 (1), 29e34. OMot, 2003. Ontario Regulation 268/03 made under the Safe Drinking Water Act, 2002. Parkhurst, D.L., Appelo, C.A.J., 1999. In: Survey U.S.G (Ed.), User’s Guide to PHREEQC (Version 2) e A Computer Program for Speciation, Batchreaction, One-dimensional Transport, and Inverse Geochemical Calculations. U.S. Geological Survey. Pena, J., Buil, B., Garralon, A., Gomez, P., Turrero, M.J., Escribano, A., Garralon, G., Gomez, M.A., 2010. The vaterite saturation index can be used as a proxy of the S&DSI in sea water desalination by reverse osmosis process. Desalination 254 (1e3), 75e79. Radjenovic, J., Petrovic, M., Ventura, F., Barcelo, D., 2008. Rejection of pharmaceuticals in nanofiltration and reverse osmosis membrane drinking water treatment. Water Research 42 (14), 3601e3610. Sawada, K., 1997. The mechanisms of crystallization and transformation of calcium carbonates. Pure and Applied Chemistry 69 (5), 921e928.
6151
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. Shirazi, S., Lin, C.J., Chen, D., 2010. Inorganic fouling of pressuredriven membrane processes - a critical review. Desalination 250 (1), 236e248. Steiner, Z., Rapaport, H., Oren, Y., Kasher, R., 2010. Effect of surface-exposed chemical groups on calcium-phosphate mineralization in water-treatment systems. Environmental Science & Technology 44 (20), 7937e7943. 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 Research 41 (17), 3959e3967. Tchobanoglous, G., Metcalf and Eddy, 2003. Wastewater Engineering. McGraw-Hill. Ternes, T.A., Joss, A., 2006. Human Pharmaceuticals, Hormones and Fragrances: The challenge of micropollutants in urban water management. IWA Publishing, London, UK. Udert, K.M., Larsen, T.A., Biebow, M., Gujer, W., 2003. Urea hydrolysis and precipitation dynamics in a urine-collecting system. Water Research 37 (11), 2571e2582. Van Houtte, E., Verbauwhede, J., 2008. Operational experience with indirect potable reuse at the Flemish Coast. Desalination 218 (1e3), 198e207. Vrouwenvelder, J.S., Kruithof, J.C., Van Loosdrecht, M.C.M., 2010. Integrated approach for biofouling control. Water Science and Technology 62 (11), 2477e2490. Xu, P., Bellona, C., Drewes, J.E., 2010. Fouling of nanofiltration and reverse osmosis membranes during municipal wastewater reclamation: membrane autopsy results from pilot-scale investigations. Journal of Membrane Science 353 (1-2), 111e121. 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. Environmental Science & Technology 42 (13), 4857e4862.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 5 2 e6 1 6 2
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Xenobiotic removal efficiencies in wastewater treatment plants: Residence time distributions as a guiding principle for sampling strategies Marius Majewsky a,*, Tom Galle´ a, Michael Bayerle a, Rajeev Goel b, Klaus Fischer c, Peter A. Vanrolleghem d a
Resource Center for Environmental Technologies (CRTE), CRP Henri Tudor, 66, rue de Luxembourg, 4221 Esch-sur-Alzette, Luxembourg Hydromantis, Environmental Software Solutions, Inc., 1 James Street South, Suite #1601, Hamilton, ON, Canada L8P 4R5 c Department of Analytical and Ecological Chemistry, University of Trier, Behringstr. 21, 54296 Trier, Germany d modelEAU, De´partement de ge´nie civil et de ge´nie des eaux, Universite´ Laval, Que´bec, QC, Canada G1V 0A6 b
article info
abstract
Article history:
The effect of mixing regimes and residence time distribution (RTD) on solute transport in
Received 15 December 2010
wastewater treatment plants (WWTPs) is well understood in environmental engineering.
Received in revised form
Nevertheless, it is frequently neglected in sampling design and data analysis for the
1 September 2011
investigation of polar xenobiotic removal efficiencies in WWTPs. Most studies on the latter
Accepted 5 September 2011
use 24-h composite samples in influent and effluent. The effluent sampling period is often
Available online 22 September 2011
shifted by the mean hydraulic retention time assuming that this allows a total coverage of the influent load. However, this assumption disregards mixing regime characteristics as
Keywords:
well as flow and concentration variability in evaluating xenobiotic removal performances
Hydraulic residence time
and may consequently lead to biased estimates or even negative elimination efficiencies.
Sampling
The present study aims at developing a modeling approach to estimate xenobiotic
Xenobiotics
removal efficiencies from monitoring data taking the hydraulic RTD in WWTPs into
Conductivity
consideration. For this purpose, completely mixed tanks-in-series were applied to address
Removal efficiency
hydraulic mixing regimes in a Luxembourg WWTP. Hydraulic calibration for this WWTP
Optimal experimental design
was performed using wastewater conductivity as a tracer. The RTD mixing approach was coupled with first-order biodegradation kinetics for xenobiotics covering three classes of biodegradability during aerobic treatment. Model simulations showed that a daily influent load is distributed over more than one day in the effluent. A 24-h sampling period with an optimal time offset between influent and effluent covers less than the half of the influent load in a dry weather scenario. According to RTD calculations, an optimized sampling strategy covering four consecutive measuring days in the influent would be necessary to estimate the full-scale elimination efficiencies with sufficient accuracy. Daily variations of influent flow and concentrations can substantially affect the reliability of these sampling results. Commonly reported negative removal efficiencies for xenobiotics might therefore be a consequence of biased sampling schemes. In this regard, the present study aims at contributing to bridge the gap between environmental chemistry and engineering practices. ª 2011 Elsevier Ltd. All rights reserved.
* Corresponding author. Tel.: þ352 42 59 91 4670; fax: þ352 42 59 91 555. E-mail addresses: [email protected] (M. Majewsky), [email protected] (T. Galle´), [email protected] (M. Bayerle), [email protected] (R. Goel), [email protected] (K. Fischer), [email protected] (P.A. Vanrolleghem). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.09.005
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 5 2 e6 1 6 2
1.
Introduction
The elimination of micropollutants in wastewater treatment plants (WWTPs) became a major concern during the last decade. A variety of polar micropollutants such as pharmaceuticals or personal care products pass biological wastewater treatment without being fully degraded (Bernhard et al., 2006; Reemtsma et al., 2006). In order to estimate micropollutant emissions to receiving waters, the removal performance of WWTPs is usually assessed by either full-scale balancing or by the determination of biodegradation rates at lab-scale (Vieno et al., 2007; Wick et al., 2009). Both estimation approaches rely essentially on the time that the water remains in the plant, normally referred to as the hydraulic retention time (HRT). The latter is an easily accessible parameter since it can be calculated from flow through and tank volumes. Most work carried out uses 24-h composite samples assuming quantitative coverage of influent loads in the effluent with a temporal shift proportional to the HRT or very stable influent concentrations over relevant periods. However, taking into consideration variable influent conditions and that residence time distributions (RTD) of perfect plug-flow tanks do not apply to conventional WWTPs reactors, mass balancing based on influenteeffluent comparison may lead to biased or even negative removal efficiencies. Hence, an adequate description of the hydraulic characteristics is critical for designing sampling campaigns and predicting dynamic xenobiotic emission. The characterization of mixing regimes in wastewater treatment plants with RTDs and pulse-response techniques is well explored and common practice (De Clercq et al., 1999; Gujer, 2008; Levenspiel, 1999). It was successfully applied to describe mixing regimes in a variety of tracer test studies (Fall and Loaiza-Navı´a, 2007; Capela et al., 2009). The RTD is hereby fitted by the number and size of tanks-in-series, the type of the mixing regime (completely mixed, plug-flow etc.) as well as the flow conditions. Artificial tracer tests with appropriate substances like e.g. lithium and bromide salts or fluorescent dyes are commonly used for hydraulic characterization of mixing regimes (Olivet et al., 2005). However, recent studies showed that the latter can also be realized with data from routine measurements of WWTPs such as temperature or conductivity (Ahnert et al., 2010). The fact that the effluent concentration dynamics of hydrophilic micropollutants are largely governed by hydraulic mixing is often poorly considered and can lead to increased uncertainty and misinterpretation of the sampling results. Generally, WWTP performances are routinely evaluated by comparison of long-term influenteeffluent data, e.g. for chemical oxygen demand (COD) or NH3-N, and is therefore believed to be applicable to xenobiotics as well. However, the measurements and analyses of xenobiotics are cost- and work-intensive which is why often only a short sampling period (mostly 24 h in influent and effluent) is used as a tradeoff between cost and data density. In such a case, the effect of influent variations on sampling results is naturally potentially much larger. In this context, the present study aims to bridge the gap between hydrodynamic behavior and biodegradation in
6153
municipal WWTP to assess xenobiotic removal efficiencies and derive adequate sampling strategies. So far, the RTD concept has not been applied in combination with short-term xenobiotic mass balance calculations. For this purpose, hydraulic mixing regimes were characterized by use of an RTD approach and coupled to first-order biodegradation kinetics. Biodegradation was modeled for three different levels of xenobiotic biodegradability which are representative for persistent as well as moderately and easily biodegradable compounds. Simulations were applied to derive optimal sampling strategies and to minimize sampling errors. Moreover, an RTD-based method is proposed for an adequate estimation of overall removal efficiencies as well as a guidance tool for designing measurement campaigns at fullscale WWTPs.
2.
Material and methods
2.1.
Sampling & WWTP data
Wastewater conductivity was measured in the influent (after sand trap) and the effluent (after secondary clarification) of the Luxembourg WWTP Mamer with YSI 600 OMS probes over a period of three weeks (sampling interval Dt ¼ 10 min). Hourly inflow data and tank volumes were obtained from the plant operators (Water Syndicate SIDERO) (Fig. 1).
2.2.
Modeling
2.2.1.
Plant layout & calibration
The plant layout of WWTP Mamer was reproduced in the wastewater modeling software GPS-X from Hydromantis (Hamilton, Canada) (Fig. 1). It is equipped with standard activated sludge models (Gujer et al., 1999; Henze et al., 1987) allowing dynamic simulation of WWTPs. Completely mixed tanks-in-series with rectangular primary clarifiers and circular secondary clarifiers were selected. For the latter, a 1-D model of settler mass balance equations is used for ten horizontal layers of equal depth. Volumes, sequence and tank operation were adjusted according to the data supplied by the plant managers (Table 1). Measured wastewater conductivity was used as a tracer for model calibration at the given flow conditions of a three week period. Inlet conductivity was fed to the model as input and the predicted effluent conductivity was iteratively fit to the measured effluent data to determine the number of completely mixed tanks-in-series and to estimate the sludge recirculation flow used in the model. The difference between measured and predicted values was minimized by the chi square (Eq. (1)) within GPS-X. x2 ¼
N X 2 1 yi y^i 2 s i¼1 i
(1)
where x2 ¼ chi square, N ¼ number of observations, si ¼ standard deviation of the measurements, yi ¼ measured values, y^i ¼ predicted values.
6154
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 5 2 e6 1 6 2
Fig. 1 e Layout, tank volumes and tanks-in-series used in the model to describe the mixing regime of WWTP Mamer; arrows indicate point of sampling for conductivity.
2.2.2.
Leff ¼
Residence time distribution (RTD)
The distribution of residence times of a xenobiotic within the plant was determined by model simulations. To this purpose, concentrations pulses were created in the influent (duration: 24 h, following typical sampling periods of composite samples). The fraction of soluble inert COD Si served as model substance for xenobiotics in GPS-X. The COD Si fraction is neither degraded nor produced in the model and can therefore be used to determine RTDs. Residence time distributions were obtained for various flow conditions including dry weather conditions and a storm event. They were exported to MatLab (MathWorks), fitted with a linear interpolant function (r2 ¼ 1) and integrated using the curve fitting toolbox. Stepwise integrals were determined for equal time steps which allow the calculation of the released fraction of Si on the total Si per step. Mass balances of Si were checked to assure that 100% of the Si influent signal has been released.
2.2.3.
X
Si $ekbiol $ts $Q
(3)
where Leff ¼ total xenobiotic effluent load [g d1], ts ¼ residence time step of the RTD and Q ¼ flow [L h1]. The total elimination efficiency is then calculated by mass balancing the xenobiotic influent and effluent load: E¼
Linf Leff $100 Linf
(4)
where E ¼ elimination efficiency [%] and Linf ¼ xenobiotic influent load [g d1].
2.2.4.
Sampling scenarios
Two model scenarios were set up to derive optimized sampling strategies taking WWTP Mamer as an example (Section 3.4). In scenario 1, a perfect steady-state xenobiotic influent loading was assumed on the basis of 8-h composite samples (total load: 2.96 0.8 g d1; corresponding to
Biodegradation
Organic micropollutant biodegradation in activated sludge is typically described with pseudo first-order or first-order reaction kinetics (Joss et al., 2006; Schwarzenbach et al., 2003). Here, first-order biodegradation rate constants kbiol of 0.05, 0.5 and 5 h1 were chosen as representative values for three classes of biodegradability (persistent, moderately and easily biodegradable polar xenobiotics) to account for biological removal during the course of aerobic wastewater treatment. It was assumed that no significant degradation occurs during denitrification and in the clarifiers. To simulate xenobiotic biodegradation during aerobic conditions, first-order reaction kinetics was implemented into GPS-X: rsi ¼ kbiol $Si
(2) 1
1
where rsi ¼ reaction rate [ng L h ], kbiol ¼ biodegradation rate constant [h1], Si ¼ soluble xenobiotic concentration [ng L1]. To calculate the xenobiotic effluent loads as a function of the RTD, Eq. (2) is solved analytically for each residence time step ts of the RTD (temporal resolution ¼ 1 min) and multiplied by the flow. The degraded effluent load for a given degradation rate constant is then the sum of all partial loads over the selected time span e.g. 24 h:
Table 1 e Operational data of the investigated WWTP. WWTP Mamer Population equivalents Capacity utilization [%] Average flow during dry weather [m3 h1]a Average flow during rainfall event [m3 h1]a Hydraulic retention timeb [h] Mean HRT during dry weather Mean HRT on measured storm event Mean HRT in aerated tanks only; during dry weather Mean HRT in aerated tanks only; during measured rainfall event Recycled fraction of activated sludge [%]c (flow proportional)
20,300 100 136 54 503 44 16.7 3.7 4.6 1.4 7.3 3.5 1.9 0.17 0.8
a Flow conditions during the measurement campaign (3 weeks), daily mean during rainfall event. b Calculated by the quotient of tank volume and flow through (single pass). c Estimated from calibration.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 5 2 e6 1 6 2
measured loads of the pharmaceutical diclofenac; Table A.1; constant flow ¼ 200 m3 h1). In scenario 2, an example dataset for realistic influent variability was created from measured influent concentrations of two days on the basis of 2-h composite samples (see Appendix A). Since inlet concentration data was only available for two days, but a time series of four days was required, the scenario was completed by using generated concentration data for two additional days. Concentrations (n ¼ 12 per day) were randomly generated from a normal distribution with mean and standard deviation of the measured concentrations (703 35 ng L1, n ¼ 24). Corresponding measured hourly flow values of one week were used. The resulting average loads for these two days (day zero and three) were 2.1 0.6 g d1 and 2.2 0.7 g d1, respectively.
2.2.5.
Uncertainty analysis
Monte Carlo simulations were performed to assess the uncertainty introduced by discrete sampling on the load estimation. Following Ort and Gujer (2006), the error of a 2-h composite sample was assumed to be 20% (minimum error for sampling intervals > 5 min). The flow error was estimated to be 10%. The corresponding 2-h composite sample measurement values of day one and two were averaged for both flow and concentration in order to approximate a representative diurnal variation pattern (n ¼ 12). Each concentration and flow value was varied by an error composed of the standard deviation of the 2-h composite sample measurement value as given before and a random error taken from a normal distribution assuming non-systematic error variability: m ¼ mi þ si $ε
Model calibration
Calibration results show that modeled values matched measured effluent conductivity within i) the range of the effluent concentration and ii) the variation patterns of the effluent (Fig. 2). Artifacts in the effluent conductivity caused by measurement interferences were deleted resulting in gaps in the consecutive time series. Fig. 2 reveals that influent variations become dampened in the effluent but could be adequately reproduced by the model. The correlation coefficient was found to be R ¼ 0.76 suggesting good tracking of the conductivity variation. Nonetheless, small differences between modeled and measured values are observable that may be caused by not considering short circuits, stagnant zones and non-ideal mixing in the model. Their influence might change with variable hydraulic loading. The number of completely mixed tanks-in-series was determined to be n ¼ 4 (2 denitrification/2 aerobic treatment tanks as well as 2 clarifiers per lane, see Fig. 1) by minimizing chi square between modeled and measured effluent conductivity values. WWTP Mamer operates with FeCl3 addition for phosphate precipitation before activated sludge treatment which may influence the conductivity. Moreover, the latter can be affected by ionic compounds being produced or removed during biological treatment or changes in the pH. However, this is apparently of minor importance to the effluent conductivity since measured outlet patterns correlated well with modeled values using the measured influent conductivity as input. Also, the pH was found to be stable during the measurement period with 7.9 0.2 (n ¼ 36).
(5)
where m* ¼ varied value for flow and concentration, respectively, i ¼ number of the 2-h composite sample (1e12), mi ¼ measured value of 2-h composite sample i, s ¼ absolute standard deviation of 2-h composite sample i and 3 ¼ error taken from a Gaussian normal distribution (mean ¼ 0, standard deviation ¼ 1). An array of 10,000 simulation runs assured to asymptotically approximate normal distributions. The resulting error associated with the determination of a load was evaluated by using the relative standard deviation and the 5 and 95% percentiles of the output distribution. Error propagation was calculated according to standard equations (Refsgaard et al., 2007).
3.
3.1.
6155
3.2.
Residence time distribution
After having calibrated the model, the RTD of an inert soluble xenobiotic was determined by use of a Si injection pulse in the influent (duration: 24 h). Here, it should be kept in mind that the RTD is flow dependent. On that account, Fig. 3 shows modeled distributions at various given flow conditions in
Results and discussion
The investigated WWTP runs at full capacity with 20,300 population equivalents (PE). It operates with primary clarifiers, denitrification and two lanes with aerobic treatment followed by secondary clarifiers. The mean HRT, calculated by the quotient of average flow (hourly values over three weeks) and tank volumes was found to be 16.7 3.7 h over the whole plant and 7.3 3.5 h (single pass; one standard deviation) in the aerated tanks during dry weather conditions (Table 1). It decreased during a storm event (flow ¼ 503 44 m3 h1) to 4.6 1.4 h and 1.9 0.17 h, respectively.
Fig. 2 e Hydrodynamic calibration of WWTP Mamer using conductivity. Input data: influent conductivity (gray) and flow (hourly values, not shown).
6156
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 5 2 e6 1 6 2
originate from periods preceding the influent pulse during dry weather conditions. This aspect is addressed in detail in Section 3.4.
3.3.
Fig. 3 e Residence time distribution of xenobiotics at flows varying from 50 to 250 m3 hL1 as a result of an influent pulse in WWTP Mamer (duration: 24 h); areas of all curves are constant.
order to illustrate the retention of different fractions of Si in the tanks. From this, percentiles can be determined giving the percentage of the released fraction of the influent water volume at a certain time t (Fig. 4). The RTD becomes more leftskewed with increased flow resulting in smaller percentiles. For instance, during dry weather conditions around 20% of the influent water volume have been released within 24 h while during a rainfall event already 60% have been emitted during the same time. When comparing those percentiles to the HRT from Table 1, the mismatch becomes apparent: at a mean HRT (16.7 h) only around 10% of the water volume that entered the WWTP 16.7 h ago has been released. A 24-h effluent (composite) sample shifted by a temporal offset of the HRT would contain only 30e40% of the influent pulse. Consequently, a large proportion of sampled wastewater would
For prediction of the overall removal efficiency, first-order kinetics is usually solved for the HRT. As can be seen from Fig. 3, a single mean HRT is not suited to describe the residence of wastewater volumes in reactor tanks, in particular for strongly skewed RTDs. The use of RTDs for removal calculations is therefore much more adequate. It allows depicting decreased removal efficiencies during rainfall events (Table 2). Under high flow conditions the RTD shifts toward a left-skewed distribution, i.e. that the RTD fractions of short retention in the plant increase. Hence, high flow leads to decreased retention times in the tanks and to a decreased removal assuming that degradation rates remain constant. A visualization of flow influence on elimination efficiency is shown in Fig. 5, where the decrease of the elimination of a moderately biodegradable xenobiotic (kbiol ¼ 0.5 h1) with increasing flow through has been plotted. Using the model, elimination efficiencies can be described as a function of the flow. The dashed line indicates removal efficiencies calculated by use of the HRT. Compared to the solid line, which show the efficiencies based on the RTD, a clear mismatch of up to 45% is evident. There, the mixing regime has a significant impact causing lower degradation. As a consequence, the use of HRT and laboratory determined kbiol may lead to an overestimation of the actual removal performance (Table 2). The fact that a daily influent load is discharged in the effluent over a period longer than one day is a major concern for the determination of full-scale elimination efficiencies by influenteeffluent mass balancing. To account for this hydraulic behavior and to derive more adequate sampling strategies, influenteeffluent correspondence was investigated in the following scenarios.
3.4.
Fig. 4 e Exit-age distributions as a result of Si influent pulse injections (duration: 24 h) showing the cumulative released water volume fractions in WWTP Mamer; flow was selected according to Table 1.
RTD effects on biodegradation
Sampling scenarios
The adequate setting of sampling intervals to address mixing regimes is crucial for the determination of elimination rates at full-scale. For example, it is remarkable that negative removal efficiencies for several (biodegradable) pharmaceuticals have been reported in a variety of studies (Onesios et al., 2009). Although knowing that certain parent compounds can be formed by the cleavage of conjugates (Ternes, 1998), also inadequate sampling strategies can yield erroneous mass balances when the water volumes sampled in influent and effluent do not correspond. The importance of adapting sampling mode and frequency to influent variability and catchment structure as well as the errors being associated with discrete sampling have been shown before (Minkkinen, 2007; Minkkinen and Esbensen, 2009; Ort et al., 2010a). Modeling simulations showed that a daily water volume is distributed over more than one day when discharged in the effluent. Consequently, a daily influent load cannot be completely covered by (composite) samples taken over a period of only 24 h at the outlet. However, an optimum temporal offset can be identified, by which a 24-h effluent
6157
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 5 2 e6 1 6 2
Table 2 e Elimination efficiencies calculated on the basis of the RTD and the HRT in the aerated tanks for three different first-order degradation rate constants; dry weather: constant flow [ 136 m3 hL1; storm event: constant flow [ 503 m3 hL1. Xenobiotic
Degradation rate constant, kbiol [h1]
Total elimination efficiency Based on RTD [%]
Persistent Moderately biodegradable Easily biodegradable
0.05 0.5 5
sampling period is shifted from the beginning of the influent period to cover the maximum percentage of the released load. Given these findings, the following model scenarios were set up to derive an optimized sampling strategy for reliable xenobiotic mass balances at full-scale: Scenario 1: In this scenario, the influent concentrations are assumed to be sampled on three consecutive days on a basis of 8-h composite samples, while the effluent is sampled on one day only. The effluent sampling period (24 h) was shifted by the optimum offset (here: 18 h) from the beginning of the second measurement day and is indicated as vertical dashed lines in Fig. 6. A perfect steady-state variation pattern was assumed with a constant flow (200 m3 h1) and biodegradation was set to zero (kbiol ¼ 0 h1). The water volume sampled on day two explains only 55.6% of the sampled effluent water volume. Consequently, the remaining 44.4% originate from preceding days and one following day, as can be seen from Fig. 6. Based on the RTD approach, the fractions of each (daily) influent water volume released at the time of the effluent sampling can be calculated (Table 3). Subsequently, to obtain a reliable mass balance, it is preferable to explain the origin of 80e90% of the sampled effluent water volume. The number of influent sampling days needed to achieve this, can be derived from the cumulative
Fig. 5 e Comparison of modeled elimination efficiencies for a moderately biodegradable xenobiotic (kbiol [ 0.5 hL1) on the basis of HRT and RTD as residence times in aerated tanks at various constant flow conditions.
Based on HRT [%]
Dry weather
Rain event
Dry weather
Rain event
10 55 98
4 21 78
31 98 100
10 63 100
proportions of each inlet measurement day on the effluent load. In this case, sampling influent days 1e3 would allow explaining 91.0% of the effluent sampled during a 24-h period. An additional fourth day would result in 99.2%. When sampling influent day 1e3, 9% of the effluent sample originates from unknown water volumes (day zero, non-covered period). The proportions of the influent load on the effluent sample match exactly the captured fractions, since an ideal steady-state was assumed (Table 3). A perfect diurnal concentration pattern at constant flow in the influent would result in periodic effluent concentrations (one effluent period may extend over more than 24 h). Under these simplifying conditions it would be sufficient to sample one in- and effluent periodical pattern with any time shift. Mass balance calculations would result in the same elimination efficiency. However, such stable conditions and constant periodic loads do not reflect reality. The concentrations of xenobiotics can vary largely during a diurnal cycle depending on their usage. Nelson et al. (2011) reported intense pulses that exceeded relative standard deviations of 100% of their daily means for selected pharmaceuticals in WWTP effluents. Further, strong diurnal variation was shown e.g. for benzotriazoles in influents (Ort et al., 2005). Hence, it can be expected that high variations in influent and effluent concentrations are very likely to occur. As a consequence, the variability of empirical data was used in Scenario 2 to adapt the sampling scheme to realistic conditions. Scenario 2: Here, realistic influent concentration patterns (2-h composite samples) during dry weather conditions were introduced as well as the corresponding measured flow (hourly values) during that period (Fig. 7). A diurnal variation can be observed in the flow as well as one rainfall event on day five. Again, biodegradation was set to zero (kbiol ¼ 0 h1). The concentration variability of the influent load is propagated through the plant and can visibly be tracked in the released effluent concentrations. Applying the same sampling scheme as in scenario 1 (day 1e3), the origin of 71.1% of the effluent load could be explained with three consecutive inlet measurement days and added up to 84.9% when including day 0 (Table 4). Hence, 15.1% stem from loads of days preceding the influent sampling period and are therefore unknown. Compared to scenario 1, scenario 2 shows a lower coverage during the effluent sampling time span. This is mainly due to the lower flow conditions that decelerate the xenobiotic release. The actual elimination efficiency can now be determined by estimating the reference load that actually corresponds to the effluent sample. This reference load is composed of load
6158
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 5 2 e6 1 6 2
Fig. 6 e Sampling scenario 1: xenobiotic concentrations assumed to be sampled on three consecutive days in the influent (8h composite samples) and one sampling day in the effluent shifted by the optimum offset (18 h from the beginning of influent day 2); flow [ 200 m3 hL1 (not shown); load fractions captured during the sampling period: 8.2% of day 0, 24.2% of day 1, 55.6% of day 2 and 11.2% of day 3.
fractions of, in this case, four days (cf. Tables 3 and 4) and can thus be calculated as the sum of the latter (see Appendix B for a mathematical description). It is then used in the mass balance calculations and compared to the measured and potentially degraded effluent load. In the case that less than 100% of the relevant influent loads have been covered, the uncertainty of this non-sampled loading needs to be considered. This uncertainty decreases with decreasing contribution to the sampled effluent load. Details of this aspect are addressed in Section 3.4.1. Expanding the influent measurement period to 3 or 4 days may be no financial issue with four 24-h composite samples but can become decisive when a higher temporal resolution is required e.g. with 8-h or 2-h composite samples. There, the costebenefit ratio with regard to the reliability of the results should be assessed in advance.
3.4.1.
Biodegradation scenarios
In both scenarios, the elimination was set to 0% to calculate the fractions captured by effluent sampling. However, at full-scale, biodegradation takes place during biological treatment which reduces the influent load. Assuming that biodegradation is proportional to the concentration with a first-order rate constant kbiol, the relative fractions of the influent loads (and their ratios to each other) during the sampled effluent period remain constant, while only the absolute load changes. In this way, the calculation of the reference load is also valid for every biodegradation scenario.
In the following example, biodegradation was simulated in scenario 2 for the three xenobiotics of different persistence and a surrogate with kbiol ¼ 0 h1 (Table 5). The resulting elimination efficiencies are associated with an error because not the full 100% of the sampled load could be related to a sampled influent load. Therefore, it was assumed that the loads of the non-covered period preceding the influent sampling days had the same daily average load and varied with the same standard deviation (27.2%) as the measured loads (day one and two). This would result in an uncertainty on the total elimination efficiency of 4% for the surrogate, 3% for a persistent, 2% for a moderately biodegradable and 0% for a readily biodegradable xenobiotic. The error decreases as the variation of the non-covered period has a comparatively lower impact on the mass balance for easily and moderately biodegradable xenobiotics. Erroneous elimination efficiencies are obtained ranging from 14 to 16% for both the surrogate and the persistent xenobiotic when using the conventional 1e1 day influenteeffluent mass balancing approach for each sampling day (day 0e3) with average loads. The uncertainty of these values can be expected to increase considerably under more variable influent conditions. However, the apparent elimination efficiencies of moderately and easily biodegradable substances are well approaching the true efficiency. This is due to the fact that, relative to the influent load, variations of largely degraded effluent loads affect the mass balance to a lesser extent, as it was the case before for the variations of the noncovered period. The conventional approach is thus more
6159
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 5 2 e6 1 6 2
Table 3 e Load fractions of the consecutive influent measurement days captured during the effluent sampling period (24 h, from Fig. 6) and their proportion of the effluent load; the optimum outlet sampling period was calculated to start by an 18 h time shift from the beginning of day two. Influent measurement day Day Day Day Day
0a 1 2 (optimum offset) 3
Non-covered period
Influent load fraction captured by effluent samplingb [%]
Proportion explained of the effluent loadc [%]
8.2 24.2 55.6 11.2
8.2 24.2 55.6 11.2
0.8
0.8
91.0
99.2
a Day 0 ¼ day before the measuring period. b Referred to the influent load of each measurement day. c Referred to the calculated effluent load (2.96 g d1); in this case, the proportions of the effluent load are identical to the captured influent fractions since perfect steady-state conditions were assumed.
robust to influent variations of readily and biodegradable xenobiotics. Further, there was only a low variation of the loads during the four days in the influent. Nonetheless, disregarding the variation preceding and during a sampling campaign would make it virtually impossible to estimate how reliable the obtained elimination efficiency value actually is.
3.4.2.
Uncertainty analysis
Besides the uncertainty of the non-covered period, elimination efficiencies estimated with the proposed approach are additionally associated with the error introduced by discrete (24-h) composite samples, depending on the mode and frequency. Influent short-term variations are usually not captured by most sampling schemes and are therefore an error source leading to non-representative results for average
loads (Ort et al., 2010b). Flow measurement errors (here assumed to be 10%) affect also the accuracy of the reference load determined by the RTD approach. Ort and Gujer (2006) showed for a middle-sized catchment that a sampling interval of at least five minutes (time-proportional) was required to obtain a representative influent composite sample (2-h) with standard deviations lower than 20%. These errors must be considered in order to reliably estimate the reference load. To this purpose, Monte Carlo simulations were used to investigate the propagated error of flow and concentration sampling on the estimated reference load. Simulation results showed that fraction load estimates approximated a normal distribution with a relative standard deviation of 6.4% and a range from 8.3 to 8.3% (5 and 95% percentiles). As the total reference load is composed of multiple load fractions (in this
Fig. 7 e Sampling scenario 2: xenobiotic concentrations assumed to be sampled on four consecutive days in the influent (2-h composite samples; realistic influent variability of xenobiotic concentrations (diclofenac)) and one sampling day in the effluent shifted by the optimum offset (18 h from the beginning of influent day 2); measured flow values were used (hourly values); load fractions captured during the sampling period: 14.4% of day 0, 22.4% of day 1, 30.7% of day 2 and 16.0% of day 3.
6160
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 5 2 e6 1 6 2
Table 4 e Load fractions of the consecutive influent measurement days captured during the effluent sampling period (24 h, from Fig. 7) and their proportion of the effluent load; the optimum outlet sampling period was calculated to start by an 18 h time shift from the beginning of day two. Influent measurement day
Proportion explained of the effluent loadb [%]
Influent load fraction captured by effluent samplinga [%]
Day 0 Day 1 Day 2 (optimum offset) Day 3 Non-covered periodc
14.4 22.4 30.7 16.0 14.8
13.8 18.5 36.5 16.1 15.1
84.9
71.1
a Referred to the influent load of each measurement day. b Referred to the calculated effluent load (2.21 g d1). c Comprises two days before day 0; loads were estimated as the average load of day 0e3.
Table 5 e Elimination efficiencies estimated from four biodegradation scenarios based on mass balance calculations by i) using the fractionated reference load and ii) comparing daily average loads of each influent sampling day (0e3) to the sampled effluent load; average loads are given in Table A.1 and Section 2.2.4. Xenobiotic
Degradation rate constant, kbiol [h1]
Reference load [g d1]
Degraded load sampled in the effluent [g d1]
Elimination efficiencya [%]
Apparent elimination efficiency by conventional sampling schemeb [%]
0 0.05 0.5 5
2.21 2.21 2.21 2.21
2.21 2.01 1.09 0.08
0 4c 9 3c 51 2c 96 0c
3; 14; 10; 1 4; 6; 16; 9 48; 43; 55; 50 96; 96; 97; 96
Surrogate Persistent Moderately biodegradable Easily biodegradable
a Calculated using the reference load. b Calculated by comparing the daily average load of each day (0e3) to the sampled effluent load. c Error is caused by the assumed variation of the non-covered period preceding the influent sampling days (RSD ¼ 27.2%).
case: the fractions of four days plus non-covered period), the total propagated error would be 14.3% based on the standard deviation. The length of the sampling time span (12 h, 24 h, 36 h) in the effluent has no effect on this error assuming the same sampling mode and frequency. These sampling errors as well as the conditions of the unknown days preceding the sampling campaign, which may be highly variable, consequently affect mass balances. The latter can only be closed here for the WWTP Mamer since the non-sampled period is assumed to have similar average loads and variability as the measured days. It nonetheless
Table A.1 e Measured average concentrations and loads of diclofenac in the influent of WWTP Mamer on two independent measurement days for scenarios 1 and 2 (Figs. 6 and 7); loads calculated based on hourly flow and two hourly concentration data; stdev [ one standard deviation. Diclofenac Concentration [ng L1] (n ¼ 12) Flow [m3 h1] (n ¼ 24) Loads [g d1]
Day 1 [average stdev]
Day 2 [average stdev]
728 39
678 31
111 14
160 19
1.9 0.4
2.4 0.8
demonstrates clearly how daily influent variation can lead to misinterpretation when the sampling intervals are not accurately set and short-term data is used, in particular based on 24-h composite samples.
4.
Conclusions
This study demonstrates that hydrodynamic characteristics are crucial for elimination, emission prediction and sampling of xenobiotics in municipal WWTPs. The hydraulic retention time is only of limited use since it does not reveal any information about mixing and distribution behavior. In order to tackle this issue, a residence time distribution approach linked with biodegradation kinetics was applied. This approach illustrated the problems encountered when trying to match influent loads with effluent loads. Depending on the flow regime, a 24-h xenobiotic influent load can expand significantly over more than one day when released in the effluent. It shows that a 24-h sampling period can cover only a small percentage of the corresponding influent load. The optimal sampling setup for full-scale mass balancing at the WWTP Mamer was determined to be a coverage of four consecutive days in the inlet and a single sampling day at the outlet with an offset of 66 h to the beginning of the inlet monitoring. The presented set-up would allow explaining the origin of 84.9% of the sampled effluent load under realistic
6161
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 5 2 e6 1 6 2
conditions. This coverage can be calculated for every WWTP that should be monitored and it is advisable to use simulations for the planning and evaluation of a monitoring campaign with regard to calculation of the total elimination efficiencies. However, the number of consecutive influent sampling days must be selected plant-specifically. It is related to the prevailing mixing regime and thus requires calibration via tracer tests. We demonstrated that calibrating hydraulic models by wastewater conductivity can offer a cost-effective option compared to artificial tracers and should therefore be implemented in full-scale measurement campaigns. The uncertainty caused by non-covered periods preceding the sampling days can be estimated from the average loads and standard deviation of the measurement days assuming them to be representative for dry weather conditions. In the Mamer plant, an accurate full-scale mass balance is only possible by high inlet coverage with monitoring. The sampling mode and frequency but also analytical errors can cause additional uncertainty. Sewer network and catchment structure as well as rainfall events greatly determine the variability of flow and xenobiotic concentrations and can lead to shortterm variations in the range of minutes (Ort et al., 2010a,b). With regard to these aspects and that the origin of a sampled effluent water volume could not be explained to 100%, this study reveals that elimination effciencies of less than 15e20% are probably impossible to track in full-scale investigations. The present paper raises the issue of mass balancing influent and effluent loads on the basis of short-term WWTP measurement campaigns. We showed that apparent negative elimination efficiencies can be caused by inadequate sampling strategies. Results illustrate the need to cover influent loads over several days and to consider the hydraulic characteristics in treatment plants. Hence, the accuracy of reported full-scale elimination efficiencies should be revised under these aspects.
A.1.
Samples were collected directly after a sampling cycle, stored at 4 C and analyzed within 24 h. Diclofenac was analyzed by use of a LCeMS/MS system consisting of a Finnigan TSQ Quantum Discovery MAX from Thermo with a Surveyor MS Pump Plus (flow rate of 200 ml min1), a Surveyor LC-Pump Plus (flow rate of 2 ml min1) and an autosampler HTC PAL from CTC Analytics. A 1 ml online enrichment method was used with an extraction column Hypersil Gold (20 2.1 mm, particle size 12 mm) from Thermo. A polar endcapped C18 column Gold aQ (100 2.1 mm, particle size 3 mm) served as chromatography column. The eluent was increased from 70:30% H2O/MeOH to 0:100% within 22 min. Limits of quantification (LOQ) were found at 125 ng L1 in influent samples.
Appendix B. Estimating the total elimination efficiency The influent concentration and flow data of n consecutive measurement days can be used for model simulations in order to calculate the actual inlet reference load that corresponds to the load proportions fn [e] captured by an effluent sampling period. This reference load can be determined as: Lref ¼
Appendix A. Example data Example data introduced to scenario 2 consisted of i) the pharmaceutical diclofenac (time-proportional, 12 glass bottles, 2-h composite samples with 24 min aliquot sampling frequency in influents using ISCO 6700 autosampler units) analyzed on two independent days (28th May and 6th June 2009) as well as ii) hourly inflow data and tank volumes that were obtained from the plant operators (Water Syndicate SIDERO). The measurement error associated with the flow data is unknown and therefore assumed to be 10%.
fn $Linf;meas;n
(6)
where Lref ¼ reference load [ng d ], fn ¼ fraction of the influent load of day n [e] on an effluent sampling period, Linf,meas,n ¼ measured influent load of day n [ng d1]. Subsequently, the measured (potentially degraded) effluent load can be related to Lref in order to calculate the actual elimination efficiency E in [%]:
This study was financed by the Luxembourg Research Fund (FNR) within AFR 07/017 and the DomesticPest project funded by the Ministry of the Interior and the Greater Region, Luxembourg. We would like to thank Julien Farlin for proofreading and Ulrich Leopold for his support in the uncertainty analysis. We thank the Water Syndicate SIDERO for their cooperation. Peter Vanrolleghem holds the Canada Research Chair in Water Quality Modelling.
X
1
E¼
Acknowledgments
Xenobiotic analysis
Lref Leff;meas $100 Lref
(7)
where Leff,meas is the measured effluent load on a chosen sampling period.
references
Ahnert, M., Kuehn, V., Krebs, P., 2010. Temperature as an alternative tracer for the determination of the mixing characteristics in wastewater treatment plants. Water Research 44 (6), 1765e1776. Bernhard, M., Mu¨ller, J., Knepper, T.P., 2006. Biodegradation of persistent polar pollutants in wastewater: comparison of an optimized lab-scale membrane bioreactor and activated sludge treatment. Water Research 40 (18), 3419e3428. Capela, I., Bile´, M.J., Silva, F., Nadais, H., Prates, A., Arroja, L., 2009. Hydrodynamic behaviour of a full-scale anaerobic contact reactor using residence time distribution technique. Journal of Chemical Technology & Biotechnology 84 (5), 716e724. De Clercq, B., Coen, F., Vanderhaegen, B., Vanrolleghem, P.A., 1999. Calibrating simple models for mixing and flow propagation in waste water treatment plants. Water Science & Technology 39 (4), 61e69.
6162
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 5 2 e6 1 6 2
Fall, C., Loaiza-Navı´a, J.L., 2007. Design of a tracer test experience and dynamic calibration of the hydraulic model for a full-scale wastewater treatment plant by use of AQUASIM. Water Environment Research 79 (8), 893e900. Gujer, W., 2008. Systems Analysis for Water Technology. Springer, Berlin, Heidelberg. Gujer, W., Henze, M., Mino, T., van Loosdrecht, M., 1999. Activated sludge model no. 3. Water Science & Technology 39 (1), 183e193. Henze, M., Grady Jr., C.P.L., Gujer, W., Marais, G.v.R., Matsuo, T., 1987. Activated sludge model, no. 1. IAWPRC Scientific and Technical Report No. 1, London, UK. Joss, A., Zabczynski, S., Go¨bel, A., Hoffmann, B., Lo¨ffler, D., McArdell, C.S., Ternes, T.A., Thomsen, A., Siegrist, H., 2006. Biological degradation of pharmaceuticals in municipal wastewater treatment: proposing a classification scheme. Water Research 40 (8), 1686e1696. Levenspiel, O., 1999. Chemical Reaction Engineering, third ed. John Wiley & Sons, Hoboken, USA. Minkkinen, P., 2007. Weighting error e is it significant in process analysis? In: Costa, Joa˜o Felipe, Koppe, Jair (Eds.), Proceedings WCBS3 Third World Conference on Sampling and Blending, Porto Alegre, Brazil Publication Series Fundaca˜o Luis Englert No. 1/2007, pp. 59e68. Minkkinen, P.O., Esbensen, K.H., 2009. Grab vs. composite sampling of particulate materials with significant spatial heterogeneity e a simulation study of “correct sampling errors”. Analytica Chimica Acta 653 (1), 59e70. Nelson, E.D., Do, H., Lewis, R.S., Carr, S.A., 2011. Diurnal variability of pharmaceutical, personal care product, estrogen and alkylphenol concentrations in effluent from a tertiary wastewater treatment facility. Environmental Science & Technology 45 (4), 1228e1234. Olivet, D., Valls, J., Gordillo, M.A., Freixo´, A., Sa´nchez, A., 2005. Application of residence time distribution technique to the study of the hydrodynamic behaviour of a full-scale wastewater treatment plant plug-flow bioreactor. Journal of Chemical Technology & Biotechnology 80 (4), 425e432.
Onesios, K., Yu, J., Bouwer, E., 2009. Biodegradation and removal of pharmaceuticals and personal care products in treatment systems: a review. Biodegradation 20 (4), 441e466. Ort, C., Gujer, W., 2006. Sampling for representative micropollutant loads in sewer systems. Water Science & Technology 54 (6e7), 169e176. Ort, C., Schaffner, C., Giger, W., Gujer, W., 2005. Modeling stochastic load variations in sewer systems. Water Science & Technology 52 (5), 112e122. Ort, C., Lawrence, M.G., Reungoat, J., Mueller, J.F., 2010a. Sampling for PPCPs in wastewater systems: comparison of different sampling modes and optimization strategies. Environmental Science & Technology 44 (16), 6289e6296. Ort, C., Lawrence, M.G., Rieckermann, J., Joss, A., 2010b. Sampling for pharmaceuticals and personal care products (PPCPs) and illicit drugs in wastewater systems: are your conclusions valid? A critical review. Environmental Science & Technology 44 (16), 6024e6035. Reemtsma, T., Weiss, S., Mueller, J., Petrovic, M., Gonzalez, S., Barcelo, D., 2006. Polar pollutants entry into the water cycle by municipal wastewater: a European perspective. Environmental Science & Technology 40 (17), 5451e5458. Refsgaard, J.C., van der Sluijs, J.P., Højberg, A.L., Vanrolleghem, P.A., 2007. Uncertainty in the environmental modelling process e a framework and guidance. Environmental Modelling and Software 22, 1543e1556. Schwarzenbach, R.P., Gschwend, P.M., Imboden, D.M., 2003. Environmental Organic Chemistry, second ed. WileyInterscience, Hoboken. Ternes, T.A., 1998. Occurrence of drugs in German sewage treatment plants and rivers. Water Research 32 (11), 3245e3260. Vieno, N., Tuhkanen, T., Kronberg, L., 2007. Elimination of pharmaceuticals in sewage treatment plants in Finland. Water Research 41 (5), 1001e1012. Wick, A., Fink, G., Joss, A., Siegrist, H., Ternes, T.A., 2009. Fate of beta-blockers and psycho-active drugs in conventional wastewater treatment. Water Research 43 (4), 1060e1074.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 6 3 e6 1 7 2
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Optimization of intermittent, simultaneous dosage of nitrite and hydrochloric acid to control sulfide and methane productions in sewers Guangming Jiang, Oriol Gutierrez 1, Keshab Raj Sharma, Jurg Keller, Zhiguo Yuan* Advanced Water Management Centre, The University of Queensland, St. Lucia, Queensland 4072, Australia
article info
abstract
Article history:
Free nitrous acid (FNA) was previously demonstrated to be biocidal to anaerobic sewer
Received 5 May 2011
biofilms. The intermittent dosing of FNA as a measure for controlling sulfide and methane
Received in revised form
productions in sewers is investigated. The impact of three key operational parameters
31 August 2011
namely the dosing concentration, dosing duration and dosing interval on the suppression
Accepted 5 September 2011
and subsequent recovery of sulfide and methane production was examined experimentally
Available online 10 September 2011
using lab-scale sewer reactors. FNA as low as 0.26 mg-N/L was able to suppress sulfide production after an exposure of 12 h. In comparison, 0.09 mg-N/L of FNA with 6-h exposure
Keywords:
was adequate to restrain methanogenesis effectively. The recovery of sulfide production
Sulfide
was well described by an exponential recovery equation. Model-based analysis revealed
Methane
that 12-h dosage at an FNA concentration of 0.26 mg-N/L every 5 days can reduce the
Sewer
average sulfide production by >80%. Economic analysis showed that intermittent FNA
Nitrite
dosage is potentially a cost-effective strategy for sulfide and methane control in sewers.
Free nitrous acid
ª 2011 Elsevier Ltd. All rights reserved.
Intermittent dosing Biocidal effect Recovery Modeling
1.
Introduction
The production of hydrogen sulfide in sewers is a well-known problem. Hydrogen sulfide is the primary source of sewer odors (WERF, 2007). Additionally, it can be oxidized into sulfuric acid on concrete surface exposed to sewer air (Hvitved-Jacobsen, 2002). The low pH causes severe concrete corrosion in sewer pipes, manholes and pumping stations. Hydrogen sulfide is also toxic to human and animals (WHO, 2003).
Unlike hydrogen sulfide, it was not until very recently that methane production and emission in sewers became a serious research topic. Several reports found that significant quantities of methane forms in sewers (Foley et al., 2009; Guisasola et al., 2008, 2009). Methane is a persistent and potent greenhouse gas and its emission contributes to the global warming (IPCC, 2006). Uncontrolled release of methane can create possibilities of explosion due to its low explosion limit (approximately 5%) (Spencer et al., 2006). The consumption of
Abbreviations: COD, chemical oxygen demand; FNA, free nitrous acid; HRT, hydraulic retention time; N2O, nitrous oxide; SRB, sulfatereducing bacteria; VFA, volatile fatty acid. * Corresponding author. Tel.: þ61 7 3365 4374; fax: þ61 7 3365 4726. E-mail address: [email protected] (Z. Yuan). 1 Present address: Catalan Institute for Water Research, ICRA, Scientific and Technological Park of the University of Girona, Spain. 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.09.009
6164
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 6 3 e6 1 7 2
soluble COD by methanogenesis in sewers imposes adverse impacts on the biological nutrient removal processes. Various chemicals have been used to control hydrogen sulfide production/emission in sewer systems. Among these, oxygen, nitrate, magnesium hydroxide, and iron salts are most commonly used (Pages et al., submitted for publication; Zhang et al., 2008). Oxygen and nitrate can oxidize sulfide chemically and/or biologically (Gutierrez et al., 2008; Mohanakrishnan et al., 2009). Ferric/ferrous salts precipitate with sulfide forming insoluble metal sulfide; and ferric salts also inhibit sulfide and methane production by sewer biofilms (Zhang et al., 2009). Mg(OH)2 elevates wastewater pH, typically to 8.5e9.0, and thus reduces the transfer of hydrogen sulfide from water to the sewer air as well as sulfide and methane productions (Gutierrez et al., 2009). All these chemicals need to be added continuously in order to be effective, thus incurring large chemical consumption and operational costs. The addition of nitrite for sulfide and methane control in sewers has been tested recently in laboratory sewer reactors (Jiang et al., 2010; Mohanakrishnan et al., 2008). The lab studies showed that the continuous addition of nitrite at 40e120 mg-N/L for three weeks could completely suppress SRB and methanogenic activities. It then took 4e6 weeks for the sulfate reduction capability of sewer biofilms to fully recover. The recovery of methanogenic activity took even longer (50% recovery in 2 months). It was suggested that nitrite causes specific inhibition on dissimilatory sulfate reduction (Einsiedl, 2009; Greene et al., 2006) and therefore an exposure of sulfate-reducing bacteria (SRB) to a high-level of nitrite caused a gradual decrease of the SRB population and hence the loss of biofilm activity (Jiang et al., 2010; Mohanakrishnan et al., 2008). However, this hypothesis was not supported by the field test results reported in Jiang et al. (2010). The dosage of nitrite at a concentration of 100 mg-N/L for only 33 h over a period of three days achieved the complete suppression of both the sulfate-reducing and methanogenic activities, and the recovery of these activities resembled the re-growth pattern (Jiang et al., 2010). Such effects implied a toxic rather than an inhibitory effect. Jiang et al. (2011) showed that the toxic effect was caused by free nitrous acid (FNA or HNO2) formed from nitrite rather than by nitrite itself. It was shown that FNA has a strong biocidal effect on anaerobic sewer biofilms. It was found that the viable microbial cells in biofilms decreased from approximately 80% prior to FNA dosage to 5e15% after the biofilm was exposed to FNA at 0.2e0.3 mg-N/L for 6e24 h. The strong biocidal effect of FNA on sewer biofilms implies that the simultaneous dosage of nitrite and acid could achieve rapid inactivation of the SRB and methanogens in sewer biofilms, making it possible to achieve sulfide and methane control through short, intermittent dosages (Jiang et al., 2011). There are three key operational parameters for an intermittent dosing strategy. They are the concentration of the chemicals (dosing strength), the exposure time (dosing duration), and the dosing interval (repetition frequency). A costeffective FNA dosing strategy will entail the lowest consumption of chemicals with satisfactory controlling efficiency by optimizing these three parameters. This study aims to develop a cost-effective dosing strategy for the intermittent, simultaneous addition of nitrite and acid
into sewer systems to control sulfide and methane productions. The effects of the above identified parameters, i.e. FNA concentration, exposure time, and dosing intervals, on sulfide and methane productions were investigated experimentally using laboratory-scale sewer reactor systems. Different combinations of these parameters were tested in either single- or multiple-dosage tests in search of the optimal FNA concentration and exposure time. An exponential growth equation was used to describe the recovery of the sulfatereducing and methanogenic activities following the biofilm exposure to FNA, which supported an economic analysis of the intermittent FNA dosing strategy and its comparison with other sulfide control chemicals.
2.
Materials and methods
2.1.
Anaerobic sewer biofilm reactors
The laboratory experimental system consisted of four airtight reactors, namely R1eR4, made of Perspex (Fig. 1). Each reactor had a volume of 0.75 L, with a diameter of 80 mm and a height of 149 mm. Plastic carriers (Anox Kaldnes, Norway) of 1 cm diameter were clustered on four stainless-steel rods inside each reactor as biofilm supporters. The total volume of carriers was about 15 mL (2% of the reactor volume). The biofilm area, including both reactor wall and carriers, was 0.05 m2. The area to volume ratio (A/V) was 70.9 m2/m3. The biological activities (i.e. sulfate reduction and methanogenesis) in sewers are primarily occurring in the biofilms (Gutierrez et al., 2009). Domestic wastewater, collected weekly from a local pumping station, Brisbane (Australia), was used as the feed. The wastewater composition had large variations in sulfate, volatile fatty acids (VFAs), and chemical oxygen demand (COD) concentrations. The sewage typically contained sulfide at concentrations of <3 mg-S/L, sulfate at concentrations between 10 and 25 mg-S/L, and VFA at 50e120 mg-COD/L. Nitrite was below the detection limit. Sulfite and thiosulfate were present in negligible amounts (<1 mg-S/L). The sewage was stored at 4 C and heated up to 20 C before being pumped into the reactors. The reactors were fed with sewage through a peristaltic pump (Masterflex 7520-47) every 6 h, a typical sewage hydraulic retention time in sewers (Hvitved-Jacobsen, 2002). Every feed pumping event lasted for 2 min, delivering one reactor volume (0.75 L) of sewage into each reactor. To ensure homogeneous distribution in reactors, mixing was provided with magnetic stirrers at 200 rpm (Heidolph MR3000).
2.2.
FNA dosing schemes
The dosingerecovery experiments were conducted in three consecutive phases, namely the baseline, dosing, and recovery phases. Reactors were operated without FNA dosing to achieve similar sulfide and methane production activities during the baseline phase. During the dosing phase, R2eR4 received nitrite and hydrochloric acid (to achieve the specified FNA concentrations), while R1 was used as the control reactor (no nitrite/acid dosage). Three sets of dosingerecovery tests
6165
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 6 3 e6 1 7 2
Fig. 1 e The laboratory-scale anaerobic sewer biofilm reactors.
were conducted to investigate the effects of different parameters. The first set of tests (Tests 1e5) was designed to investigate the effects of FNA concentration and exposure time. The previous study about the biocidal effect of free nitrous acid has indicated that 90% of cell inactivation could be achieved by dosing FNA at 0.26 mg-N/L for 12 and 24 h (Jiang et al., 2011). Thus, the FNA concentrations tested in this study were 0.18, 0.26, and 0.36 mg-N/L (Table 1). The wastewater pH was adjusted to 6e6.2 by adding hydrochloric acid, and was also dosed with 100e230 mg-N/L of nitrite to achieve the selected FNA concentrations (see Table 1). For the FNA concentration of 0.26 mg-N/L, three different exposure times, namely 12, 24, and 96 h, were used in three separate tests to investigate the impact of exposure time on the sewer biofilm activities. For other FNA concentrations (i.e. 0.18, and 0.36 mg-N/L), an exposure time of 24 h was applied. Most of these tests were repeated (Table 1). The second set of dosingerecovery test was designed to investigate the effects of different dosing intervals (Table 2). Four tests (Tests 6e9) were conducted with multiple dosages, with a fixed FNA concentration at 0.26 mg-N/L, and three dosing intervals, i.e. 4, 8, and 12 days, to investigate the potential impact of dosing intervals on the suppression and recovery of sewer biofilm activities. The exposure times were designed to incorporate those tested in the first stage. An
exposure time of 12 h was chosen in most of these tests because it was found to effectively suppress both sulfide and methane productions at the applied FNA concentration. Besides, a multiple-phase dosage of 96 h with a 12-day interval was also trialed to explore the scenarios of excessive exposure. All the above tests were designed with FNA concentrations and exposure times that were expected to achieve complete suppression of sulfide and methane productions. The third set of tests was designed to explore the possibility of partial sulfide control but potentially complete methane control at low FNA concentrations combined with short exposures. Three dosingerecovery tests (Tests 10e12) were done with an exposure time of 6 h only, at FNA concentration of 0.045, 0.09, and 0.18 mg-N/L (Table 3).
2.3.
Batch tests to determine H2S and CH4 production
For the dosingerecovery experiments described in Section 2.2, batch tests were conducted prior to, and after the FNA dosing to determine biofilm activities, i.e. sulfate reduction and methanogenesis. The batch tests were performed at intervals from 2 days to 2 weeks. The sulfate-reducing activity was measured under anaerobic conditions as sulfide production rate, and the methanogenic activity was measured as the methane production rate.
Table 1 e The dosing schemes for the single-dosage tests (Test set I) to investigate effects of FNA concentration and exposure time.
Table 2 e The dosing schemes for the multiple-dosage tests (Test set II) to investigate effects of dosing interval.
Test No.
Test No.
Test Test Test Test Test
1 2 3 4 5
FNA (mg-N/L)
Exposure (h)
Replications
0.18 0.26 0.26 0.26 0.36
24 12 24 96 24
1 2 2 2 2
Test Test Test Test
6 7 8 9
FNA (mg-N/L)
Exposure (h)
Interval (day)
Dosage
0.26 0.26 0.26 0.26
12 / 12 / 12 12 / 12 12 / 12 96 / 96
4 4 8 12
3 2 2 2
6166
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 6 3 e6 1 7 2
Table 3 e The dosing schemes for the single-dosage tests (Test set III) to investigate the sulfide and methane control at low dosing concentrations and durations. Test No. Test 10 Test 11 Test 12
FNA (mg-N/L)
Exposure (h)
Replications
0.045 0.09 0.18
6 6 6
1 1 1
The batch tests were started with pumping fresh sewage into reactors, which lasted for 10 min to ensure a thorough replacement of liquid in reactors with fresh sewage. Wastewater samples were taken at 0, 30, 60, 90, and 120 min after feeding, for the analysis of dissolved inorganic sulfur and nitrogen compounds, as well as methane. Sulfide and methane production rates were calculated using linear regression of sulfide, and methane concentrations, respectively.
2.4.
Nitrous oxide measurement by microsensor
Nitrous oxide (N2O) could be produced when nitrite was added with acid into the anaerobic sewer reactors (Betlach and Tiedje, 1981; Sutka et al., 2006). The monitoring of N2O production is important because it is a potent greenhouse gas (IPCC, 2006). Wastewater samples were drawn hourly from the reactor, filtered with 0.22 mm membrane into 5 mL capped plastic tubes. Air bubble was completely avoided using parafilm before sealing the tubes. Nitrous oxide microsensor N2O100 (Unisense A/S, Denmark), with a tip diameter of 70e120 mm, attached to a PA2000 picoammeter (Unisense) was connected to a computer for data logging. It was polarized at a voltage of 0.8 V and calibrated according to the manual. Measurements in the wastewater samples were then carried out with sensor tips dipped into the sample tubes. The sample temperature was also recorded to calculate the dissolved nitrous oxide.
2.5.
Acid titration of wastewater
To determine the amount of hydrochloric acid needed to acidify the wastewater to a specified pH, three batches of titration tests were carried out with fresh wastewater (pH ¼ 7.59 0.02). One liter of wastewater in a beaker was gently mixed with a magnetic mixer at 100 rpm, with a pH probe (Metrohm Swiss, 827 pH Lab) placed in the middle to record the pH. Once the pH readings were stable, a small volume (0.1e0.5 mL) of hydrochloric acid (1 M) was added. The new pH value was recorded when getting stable after adding acid. The acid addition and pH recording were repeated till the wastewater pH reached 5.0.
2.6.
chromatograph (IC) with an UV and conductivity detector (Dionex ICS-2000). For the analysis of nitrogen species (nitrite and ammonia), 1 mL of sewage was filtered similarly, diluted 10 times and analyzed using a Lachat QuikChem 8000 (Milwaukee) flow-injection analyzer (FIA). FNA concentration is jointly determined by the wastewater pH and the dosed nitrite concentration (Anthonisen et al., 1976; Weon et al., 2002). FNA concentration was calculated as pH FNA ¼ ð46=14Þ ðNO 2 N=ðKa 10 ÞÞ, where Ka is the ionization constant of the nitrous acid equilibrium equation. The value of Ka is determined by Ka ¼ e2300=ð273þ CÞ . For methane analysis, 5 mL sewage was filtered (0.22 mm membrane) and injected into vacuumed BD vacutainer tubes using a hypodermic needle attached to a plastic syringe. The tubes were allowed to reach gaseliquid equilibrium overnight. Methane in the gas phase was measured by gas chromatography (Shimadzu GC-9A) equipped with a flame ionization detector (FID). Concentrations of methane in sewage were calculated using mass balance and Henry’s law (Guisasola et al., 2008).
3.
Results
3.1. Sulfide, methane, nitrite, and nitrous oxide in dosing periods As an example, Fig. 2 shows the profiles of sulfide, methane, and nitrite during a 24-h dosing period (Test 3). The reactor was replaced with fresh sewage followed by nitrite and hydrochloric acid addition to the pre-designed levels every 6 h. It is clear that no sulfide was accumulating when FNA was present during the whole dosing period. The slight decrease of sulfide (<0.1 mgS/L) during the exposure was likely due to the oxidation by nitrite, either chemically or biologically (Jiang et al., 2010; Okabe et al., 2003a). In contrast to sulfide, the dissolved CH4 concentration still increased when FNA was present for the 24-h dosing period. Clearly, methane production persisted although its production was reduced gradually. After the 24 h of FNA dosages, methane production rate was reduced to a negligible
Chemical analysis
For the analyses of dissolved inorganic sulfur species (sulfide, sulfite, thiosulfate, and sulfate), 1.5 mL wastewater was filtered (0.22 mm membrane) into 0.5 mL preserving solution of sulfide anti-oxidant buffer (SAOB) (Keller-Lehmann et al., 2006). Samples were then analyzed within 24 h on an ion
Fig. 2 e Dissolved sulfide, CH4, and nitrite concentrations in the sewer reactor during a 24 h. FNA dosing period. The reactor was replaced with fresh sewage followed by nitrite and hydrochloric acid addition every 6 h.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 6 3 e6 1 7 2
level, i.e. <0.3 mg-COD/L-h. Methanogens, expected to reside in the deep layers of sewer biofilms compared to sulfatereducing bacteria (Guisasola et al., 2008), were thus likely exposed to a lower level of FNA than the organisms in the outer layers due to the consumption of nitrite (Fig. 2). The biofilm arrangement and structure likely provided a shield against the initial biocidal action of FNA. The nitrite concentration decreased over all four cycles, suggesting the occurrence of nitrite reduction. However, the reduction rate was about 5 mg-N/L-h in all cycles. Nitrite concentration at the end of each pumping cycle was still above 70 mg-N/L. This implies that a relatively high nitrite concentration can be sustained in the reactor over a long time. While not shown in the figure, pH increased by less than 0.2 in all cycles. As a result of the changes in the nitrite concentration and pH level, the FNA concentration decreased from 0.25 to 0.3 mg-N/L at the beginning of the cycles to 0.1e0.13 at the end of the cycles, which is still strongly biocidal according to Jiang et al. (2010). The addition of nitrite caused the production of nitrous oxide, which was measured using a microsensor in some of the tests. The highest N2O was about 1.4 mg-N/L measured at the end of a pumping cycle.
3.2.
Effects of FNA concentration
Three FNA concentrations, namely 0.18, 0.26, and 0.36 mg-N/L were compared by the dosingerecovery experiments with the same exposure time of 24 h. Fig. 3 shows that sulfide and methane productions reached negligible levels immediately after the FNA dosing event on day 0 (measured in the absence of nitrite and acid). Even at the lowest FNA concentration (i.e. 0.18 mg-N/L), the complete suppression of sulfide and methane productions of sewer biofilm was achieved. The sulfide and methane-producing activities in reactors gradually resumed after the FNA dosing event. Sulfide production recovered to 50% of the level in control reactor in
Fig. 3 e The sulfide (A) and methane (B) production rates in the experimental reactors dosed with FNA for 24 h, relative to those of the control reactor (given as %). The vertical arrow indicates the start of the 24 h FNA dosing on day 0.
6167
about 10 days, and then stayed at approximately 70% for the rest of the study. In contrast, methanogenesis recovered much more slowly. At the end of the test (35 days after dosing), it only recovered to 20% even with the lowest FNA dosage. The results show that the loss and the subsequently recovery of the sulfide- and methane-producing activities of sewer biofilms were similar in all cases. It appears that FNA dosage at 0.18 mg-N/L would be a preferred option among the three due to its effectiveness in suppressing biofilm activities. However, an exposure time of 24 h must be maintained for this FNA dosage.
3.3.
Effects of exposure time
Three exposure times, i.e. 12, 24, and 96 h, were experimentally compared at the same FNA concentration of 0.26 mg-N/ L. Fig. 4 shows that a complete suppression of sulfide and methane productions was achieved with all these exposure times. No significant difference was found between these three cases. The exposure time required by FNA dosage (the simultaneous addition of nitrite and acid) to suppress sulfide production is much shorter than that required by the addition of nitrite alone, which was found to be 2e3 weeks (Jiang et al., 2010). Both sulfate reduction and methanogenesis started to recover once the FNA dosing event was finished. In two weeks, the sulfate-reducing activity recovered to 70% while methanogenesis reached around 20% only, consistent with the findings reported in Fig. 3. The recovery processes are very similar with all the three different exposure times. Longer exposure time (24 and 96 h) did not further slow down the recovery, in comparison to the 12-h dosage. The exposure time was found to have no prominent impacts on the recovery of sewer biofilm activities.
Fig. 4 e The sulfide (A) and methane (B) production rates in the sewer reactors after being exposed to FNA at 0.26 mgN/L for three different exposure times. The data are presented as percentages of the rates measured on the control reactor. The arrow indicates the start of the FNA dosing on day 0.
6168
3.4.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 6 3 e6 1 7 2
Effects of dosing interval
In addition to FNA concentration and exposure time, dosing interval is another important parameter of an intermittent dosing strategy. Fig. 5 shows the results obtained in Tests 6e9, which were aimed to reveal the potential impact of consecutive dosages with different intervals on the suppression and recovery of the sewer biofilm activities. It is clear that FNA dosing for a short exposure time of 12 h at an interval of 4 days was very effective in controlling the sulfide and methane production. Sulfide production was reduced by 80% and methane by 90%, averaged over the entire period of the study (Fig. 5A and B). The recovery of the sulfide production activity was similar after all five doses. When the dosing interval was increased to 8 days (at the same dosing level and exposure time, Fig. 5C), the recovery displayed a similar trend to those in Fig. 5A and B. As a result of the extended dosing interval, the average sulfide production rate was reduced by approximately 70%, slightly lower than in the cases of Test 6 and Test 7. Methane production could still be controlled below 10%. For the extreme scenario with two 96 h dosages at a 12day interval (Test 9), the sulfide and methane was reduced by 70% and 90%, respectively, averaged over the study period of 33 days (Fig. 5D), similar to the cases of Tests 6e8. However, the chemical consumption was significantly higher (1.5e3 times).
0.09, and 0.18 mg-N/L. Also, the exposure time was reduced to 6 h. Fig. 6A shows that the 6-h exposure of sewer reactors to FNA at the above levels decreased the sulfide production rates by 50e80% (varied with the dosing concentration) of that in the control reactor. Complete suppression of methane production was achieved at FNA concentrations of 0.09 and 0.18 mg-N/L, while FNA at 0.045 mg-N/L did not achieve the same level of inhibition. The results again demonstrated that methanogens are more sensitive to FNA than sulfate-reducing bacteria. After FNA dosing, sulfide production recovered almost linearly for two weeks in all cases, reaching 70e90% of the control level. The recovery rates were similar for the three FNA concentrations. Thus, lower FNA dosage reached higher recovery due to the higher initial sulfide production rate. Eventually, sulfide production recovered to about 80e90% after 30 days. However, it did not fully recover even after 60 days. The recovery of methane production rates was comparably much slower than that of sulfide production rates. After 60 days, methane production for the lowest dosage of 0.045 mg-N/L reached 70% while the other two dosages only recovered to about 40e50%. It is clear that FNA dosage at 0.09 mg-N/L and 6-h exposure time is adequate if methane control in sewers is the primary goal.
4. 3.5.
Discussion
Efficacy of low and short dosages
Test sets I and II employed FNA concentrations and exposure times that were able to suppress sulfide and methane productions completely. Partial control may be achieved with lower FNA concentrations and/or shorter exposure times. Test set III investigated three FNA concentrations, i.e. 0.045,
4.1. Intermittent FNA dosing as a sulfide and methane control strategy The impact of three key parameters involved in the intermittent dosage of FNA for sulfide and methane control in sewers, namely the dosing concentration, dosing duration
Fig. 5 e The sulfide and methane production rates for multiple dosage experiments. The data are presented as percentages of the rates measured in the control reactor. The arrows indicate the time of the FNA dosing events.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 6 3 e6 1 7 2
Fig. 6 e The sulfide (A) and methane (B) production rates in sewer reactors before and after being dosed with different concentrations of FNA for 6 h on day 0. The data are presented as percentages of the rates measured in the control reactor.
and dosing interval, was investigated in this study. The experimental data clearly showed that complete suppression of sulfide and methane productions can be achieved when the dosing concentration and dosing duration are above certain levels. While the precise threshold values were not revealed, the experimental results showed that a dosage at an FNA level of 0.18 mg-N/L for a period of 12 h would be adequate for completely suppressing sulfate-reducing bacteria and methanogens in sewer biofilms. The speed of recovery appeared to be independent of the dosage level or duration as long as complete suppression is achieved during FNA dosage. Also, multiple dosages did not affect the rate of recovery. Because FNA kills biofilm bacteria (Jiang et al., 2011), the recovery of activities reflects the regrowth process of SRB and methanogens. Thus, the recovery rate is mainly determined by the doubling time of sulfatereducing bacteria and methanogens. The recovery data of sulfide production obtained in all single-dosage experiments are well described by an exponential recovery equation r ¼ r0þ(100 r0)(1 eat) (R2 ¼ 0.88, Fig. 7A), where r0 ¼ 3.3% is the residual sulfide production rate immediately after FNA dosing, a ¼ 0.09 d1 is the recovery constant. The exponential rising pattern of the recovery resembles the bacterial re-growth. This implies that SRB and methanogens redeveloped in the sewer biofilm after stopping the dosage. The above model supports the design of the dosing frequency required for achieving certain levels of sulfide (and methane) control. The dependency of control efficiency on the dosing interval calculated using the above model is shown in Fig. 7B. For example, a dosing interval of approximately 4 days is required to maintain a sulfide control efficiency at >80%. In contrast, less frequent dosage of FNA (one dose every 20 days) would be needed for methane control.
6169
Fig. 7 e (A) The estimation of recovery parameters r0 and a by dynamic fitting to the recovery data from all the FNA dosage; and (B) the predicted decreases in sulfide and methane control efficiency (%) with time, based on the model obtained in (A).
4.2. Economic analysis and comparison with other strategies The FNA dosing involves the addition of two chemicals, nitrite and acid, to sewers simultaneously. The FNA concentration is highly sensitive to the wastewater pH. The amount of nitrite required to make up a certain FNA concentration can be reduced by 10 times when pH is decreased by one unit. Fig. S2 compares the nitrite and acid costs (prices/costs are given in US dollars) as a function of pH. The chemical costs for nitrite decrease dramatically from $0.85/m3 at pH 7.5 to $0.03/m3 at pH 6, to achieve an FNA concentration of 0.26 mg-N/L. Such a saving in nitrite is achieved with only a slight increase ($0.01/m3) in the acid costs. The acid consumption to adjust the pH was calculated from the wastewater titration results shown in Fig. S1. Thus, a lower pH is preferred from an economic point of view. However, a too low pH can cause potential corrosion of sewer pipe and installations. A pH level of 6.0 appears to be an appropriate choice as it is not harmful for sewer concrete (communication with industry partners). The experimental results show that an FNA concentration of above 0.18 mg-N/L is adequately biocidal for sewer biofilm activities. To be on the conservative side, we use 0.26 mg-N/L in the economic analysis. The experimental and modeling results show that an exposure time of 12 h and a dosing interval of 4.5 days would be adequate for achieving an overall sulfide control efficiency of 80%. In the analysis, it was assumed that the average sulfide concentration was 15 mg-S/L in the absence of a control measure. Table 4 compares chemical costs of intermittent FNA dosing with other chemicals reported in literature.
6170
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 6 3 e6 1 7 2
Table 4 e Comparison of chemical addition costs to control sulfide in sewers. Chemicals a
FNA Ca(NO3)2 b Fe(NO3)3 b FeCl2 b FeCl3 b NaNO3 c NaNO3 c NaNO3 c NaNO3 c Ca(NO3)2 c FeCl2$4H2O c FeSO4$7H2O c FeCl3 c H2O2 c H2O2 c Pure O2 Air
Control efficiency (%) >80 100 100 100 100 100 65 100 90e95 100 90 95e97 100 90e95 87e100 80e90 90
Cost ($/m3)
Cost ($/kg-S)
0.03 0.07e0.12 0.6 0.1 0.07 e e e e e e e e e e 0.04 0.11
2.2 11e19 24 11.5 8 16.5 0.6 3.4 3.4e11.2 5.9 30.2e35.2 6.5 5.0 5.4e5.7 4.7 4.8e10.8 e
References This study Bertra´n de Lis et al. (2007) Bertra´n de Lis et al. (2007) Bertra´n de Lis et al. (2007) Saracevic et al. (2006) Jenneman et al. (1986) Okabe et al. (2003a) Okabe et al. (2003b) Yang et al. (2005) Rodriguez-Gomez et al. (2005) US EPA (1992) Tomar and Abdullah (1994) Nielsen et al. (2005) US EPA (1991) Tomar and Abdullah (1994) MMBW (1989) Saracevic et al. (2006)
a Assuming average sulfide concentration of 15 mg-S/L in sewers. Costs are estimated based on chemical prices in US $: $150.00/ton for 32% hydrochloric acid, and $550.00/ton for NaNO2, http://www.alibaba.com. b Estimated data from the respective references. c Excerpt from Zhang et al. (2008).
The estimated cost of FNA dosing to achieve 80% control is $0.03/m3 or $2.10/kg-S. The chemical cost of nitrate, ferric/ ferrous and H2O2 dosing was much higher than FNA (Table 4). It is clear that the intermittent FNA dosing is more costeffective compared to the addition of other widely used chemicals. The intermittent dosing strategy greatly reduces the chemical consumption in comparison to the continuous addition of chemicals, such as iron, oxygen, nitrate, and Mg(OH)2. Further investigation is required to obtain the actual cost (based on local chemical prices) in real sewers. The additional advantage of FNA dosage is the suppression of methane production. This will subsequently reserve more soluble COD for the nutrient removal in downstream wastewater treatment plants, supported by higher VFA in effluent from reactors dosed with FNA (Fig. S3). While oxygen and nitrate addition are also expected to suppress methane formation, both oxidize organic carbon to CO2. One adverse aspect of FNA dosing is the production of N2O. The average N2O production rate measured during dosing periods was 0.23 mg-N/L-h. N2O is an electron acceptor for denitrifiers, and therefore its further reduction to N2 by denitrifiers is expected when nitrite is consumed to low levels in downstream sections. When N2O containing wastewater is discharged to gravity sewer, N2O stripping to the sewer gas phase will likely occur particularly under turbulent conditions. Precautions should be taken in planning FNA dosage, including timing and location, to prevent the unwanted N2O release. A further potential problem is that the lower pH during dosing period can enhance the H2S transfer from liquid to the sewer gas phase. However, there will be negligible H2S in the liquid when FNA is present. In real application, FNA should be added section by section in a sewer network rather than at all places simultaneously. Nitrite will thus be diluted before arriving at the wastewater treatment plants. This can avoid the adverse shock to biological treatment processes by high nitrite concentrations.
5.
Conclusions
This study investigated the intermittent dosing of free nitrous acid to control sulfide and methane production through doserecovery experiments in sewer reactors. The three key parameters, i.e. FNA concentration, exposure time, and dosing interval, of the intermittent dosing strategy were optimized by single and multiple-dosage tests. The main conclusions are as follows. FNA concentration at 0.26 mg-N/L or above, with an exposure time of 12 h or longer, could suppress sulfide and methane production by sewer biofilms. However, the sulfide and methane production rates recover gradually following the FNA dosage, with a pattern resembling that of bacterial re-growth. To reduce sulfide production by 80%, a dosing interval of 4.5 days is required. When methane control is the primary goal, lower FNA dosage (e.g. 0.09 mg-N/L) and shorter exposure time (e.g. 6 h) can completely suppress methane production. Also, a long dosing interval (e.g. 20 days) would be adequate to achieve 90% control efficiency. The intermittent FNA dosing strategy developed in this study is more cost-effective than other chemicals commonly used by water industry. However, field tests are required to verify the effectiveness and benefits of the proposed dosing strategy.
Acknowledgments The authors acknowledge the Sewer Corrosion and Odour Research (SCORe) Project LP0882016 funded by an Australian Research Council Industry Linkage Project Grant and supported financially and in-kind by key members of the
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 6 3 e6 1 7 2
Australian water industry and acknowledge the other University Research Partners (for more details see: www. score.org.au). Guangming Jiang is grateful to the scholarships: Endeavour International Postgraduate Research Scholarship (IPRS) and University of Queensland International Living Allowance Scholarship (UQILAS).
Appendix A. Supplementary information Supplementary data associated with this article can be found in the online version at doi:10.1016/j.watres.2011.09.009.
references
Anthonisen, A.C., Loehr, R.C., Prakasam, T.B.S., Srinath, E.G., 1976. Inhibition of nitrification by ammonia and nitrous acid. Journal Water Pollution Control Federation 48 (5), 835e852. Bertra´n de Lis, F., Saracevic, E., Matsche´, N., 2007. Control of sulphide problems in pressure sewers. In: Novatech 2007. Sixth International Conference on Sustainable Techniques and Strategies in Urban Water Management, Lyon, France. Betlach, M.R., Tiedje, J.M., 1981. Kinetic explanation for accumulation of nitrite, nitric-oxide, and nitrous-oxide during bacterial denitrification. Applied and Environmental Microbiology 42 (6), 1074e1084. Einsiedl, F., 2009. Effect of NO 2 on Stable Isotope Fractionation during Bacterial Sulfate Reduction. Environmental Science & Technology 43 (1), 82e87. doi:10.1021/es801592t. Foley, J., Yuan, Z., Lant, P., 2009. Dissolved methane in rising main sewer systems: field measurements and simple model development for estimating greenhouse gas emissions. Water Science and Technology 60 (11), 2963e2971. Greene, E.A., Brunelle, V., Jenneman, G.E., Voordouw, G., 2006. Synergistic inhibition of microbial sulfide production by combinations of the metabolic inhibitor nitrite and biocides. Applied and Environmental Microbiology 72 (12), 7897e7901. doi:10.1128/aem.01526-06. Guisasola, A., de Haas, D., Keller, J., Yuan, Z., 2008. Methane formation in sewer systems. Water Research 42 (6e7), 1421e1430. Guisasola, A., Sharma, K.R., Keller, J., Yuan, Z.Q., 2009. Development of a model for assessing methane formation in rising main sewers. Water Research 43 (11), 2874e2884. doi:10. 1016/j.watres.2009.03.040. Gutierrez, O., Mohanakrishnan, J., Sharma, K.R., Meyer, R.L., Keller, J., Yuan, Z., 2008. Evaluation of oxygen injection as a means of controlling sulfide production in a sewer system. Water Research 42 (17), 4549e4561. Gutierrez, O., Park, D., Sharma, K.R., Yuan, Z., 2009. Effects of long-term pH elevation on the sulfate-reducing and methanogenic activities of anaerobic sewer biofilms. Water Research 43 (9), 2549e2557. Hvitved-Jacobsen, T., 2002. Sewer Processes: Microbial and Chemical Process Engineering of Sewer Networks. CRC Press, Boca Raton, London, New York, Washington, D.C. IPCC, 2006. Prepared by the National Greenhouse Gas Inventories Programme. In: Eggleston, H.S., Buendia, L., Miwa, K., Ngara, T., Tanabe, K. (Eds.), 2006 IPCC Guidelines for National Greenhouse Gas Inventories. IGES, Japan. Jenneman, G.E., McInerney, M.J., Knapp, R.M., 1986. Effect of nitrate on biogenic sulfide production. Applied and Environmental Microbiology 51 (6), 1205e1211.
6171
Jiang, G., Gutierrez, O., Sharma, K.R., Yuan, Z., 2010. Effects of nitrite concentration and exposure time on sulfide and methane production in sewer systems. Water Research 44 (14), 4241e4251. Jiang, G., Gutierrez, O., Yuan, Z., 2011. The strong biocidal effect of free nitrous acid on anaerobic sewer biofilms. Water Research 45 (12), 3735e3743. Keller-Lehmann, B., Corrie, S., Ravn, R., Yuan, Z., Keller, J., 2006. Preservation and simultaneous analysis of relevant soluble sulfur species in sewage samples. In: Proceedings of the Second International IWA Conference on Sewer Operation and Maintenance, Vienna, Austria. Melbourne and Metropolitan Board of Works (MMBW), 1989. Hydrogen Sulfide Control Manual: Septicity, Corrosion, and Odour Control in Sewerage Systems. Technological Standing Committee on Hydrogen Sulfide Corrosion in Sewerage, MMBW. Mohanakrishnan, J., Gutierrez, O., Meyer, R.L., Yuan, Z., 2008. Nitrite effectively inhibits sulfide and methane production in a laboratory scale sewer reactor. Water Research 42 (14), 3961e3971. Mohanakrishnan, J., Gutierrez, O., Sharma, K.R., Guisasola, A., Werner, U., Meyer, R.L., Keller, J., Yuan, Z., 2009. Impact of nitrate addition on biofilm properties and activities in rising main sewers. Water Research 43 (17), 4225e4237. Nielsen, A.H., Hvitved-Jacobsen, T., Vollertsen, J., 2005. Kinetics and stoichiometry of sulfide oxidation by sewer biofilms. Water Research 39 (17), 4119e4125. Okabe, S., Ito, T., Satoh, H., Watanabe, Y., 2003a. Effect of nitrite and nitrate on biogenic sulfide production in sewer biofilms determined by the use of microelectrodes. Water Science and Technology 47 (11), 281e288. Okabe, S., Santegoeds, C.M., De Beer, D., 2003b. Effect of nitrite and nitrate on in situ sulfide production in an activated sludge immobilized agar gel film as determined by use of microelectrodes. Biotechnology and Bioengineering 81 (5), 570e577. doi:10.1002/bit.10495. Pages, R.G., Gutierrez, O., Rootsey, R., Yuan, Z. Chemical dosing for sulfide control in Australia: an industry survey. Water Research, submitted for publication. Rodriguez-Gomez, L.E., Delgado, S., Alvarez, M., Elmaleh, S., 2005. Inhibition of sulfide generation in a reclaimed wastewater pipe by nitrate dosage and denitrification kinetics. Water Environment Research 77 (2), 193e198. Saracevic, E., Bertran de Lis, F., Matsche, N., 2006. Odour and corrosion problems in pressure sewers. In: Second IWA Conference on Sewer Operation and Maintenance, 26e28 October, Vienna. Spencer, A.U., Nolland, S.S., Gottlieb, L.J., 2006. Bathtub fire: an extraordinary burn injury. Journal of Burn Care & Research 27 (1), 97e98. Sutka, R.L., Ostrom, N.E., Ostrom, P.H., Breznak, J.A., Gandhi, H., Pitt, A.J., Li, F., 2006. Distinguishing nitrous oxide production from nitrification and denitrification on the basis of isotopomer abundances. Applied and Environmental Microbiology 72 (1), 638e644. doi:10.1128/aem.72.1.638-644.2006. Tomar, M., Abdullah, T.H.A., 1994. Evaluation of chemicals to control the generation of malodorous hydrogen sulfide in waste water. Water Research 28 (12), 2545e2552. US EPA, 1991. Hydrogen Sulphide Corrosion in Wastewater Collection and Treatment System. US EPA. US EPA, 1992. Detection, Control, and Correction of Hydrogen Sulfide Corrosion in Existing Wastewater System. Office of Wastewater Enforcement and Compliance, Washington, DC, 20460 pp. Weon, S.-Y., Lee, C.-W., Lee, S.-I., Koopman, B., 2002. Nitrite inhibition of aerobic growth of Acinetobacter sp. Water Research 36 (18), 4471e4476.
6172
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 6 3 e6 1 7 2
Water Environment Research Foundation (WERF), 2007. In: Apgar, D., Witherspoon, J., Easter, C., Bassrai, S., Dillon, C., Torres, E., Bowker, R.P.G., Corsi, R., Davidson, S., Wolstenholme, P., Forbes, B., Quigley, C., Ward, M., Joyce, J., Morton, R., Weiss, J., Stuetz, R. (Eds.), Minimization of Odor and Corrosion in Collection Systems Phase 1. WERF, London, UK. World Health Organization (WHO), 2003. Hydrogen Sulfide: Human Health Aspects (Concise International Chemical Assessment Document: 53). WHO, Geneva.
Yang, W., Vollertsen, J., Hvitved-Jacobsen, T., 2005. Anoxic sulfide oxidation in wastewater of sewer networks. Water Science and Technology 52 (3), 191e199. Zhang, L., De Schryver, P., De Gusseme, B., De Muynck, W., Boon, N., Verstraete, W., 2008. Chemical and biological technologies for hydrogen sulfide emission control in sewer systems: a review. Water Research 42 (1-2), 1e12. Zhang, L., Keller, J., Yuan, Z., 2009. Inhibition of sulfate-reducing and methanogenic activities of anaerobic sewer biofilms by ferric iron dosing. Water Research 43 (17), 4123e4132.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 7 3 e6 1 8 0
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Photocatalytic oxidation of DBP precursors using UV with suspended and fixed TiO2 Fraser C. Kent a,*, Krysta R. Montreuil a, Ryan M. Brookman a, Robbie Sanderson b, Jeff R. Dahn b, Graham A. Gagnon a a b
Department of Civil and Resource Engineering, Dalhousie University, Halifax, NS B3H 3J5, Canada Department of Physics, Dalhousie University, Halifax, NS B3H 3J5, Canada
article info
abstract
Article history:
French River water (Nova Scotia, Canada) was separated into six different natural organic
Received 24 May 2011
matter (NOM) fractions, including hydrophobic acids, bases and neutrals and hydrophilic
Received in revised form
acids, bases and neutrals. The raw water, as well as each of the NOM fractions were
31 August 2011
analysed for disinfection by-product (DBP) formation potential before and after advanced
Accepted 6 September 2011
oxidation with UV/TiO2 to determine the efficacy of this treatment for the removal of DBP
Available online 14 September 2011
precursors. The UV/TiO2 treatment was carried out with a nanostructured thin film (NSTF), coated with TiO2 which is compared with the use of a TiO2 suspension. For the raw river
Keywords:
water, removals of total trihalomethane formation potential (TTHMFP) and total haloacetic
Disinfection by-products
acid formation potential (THAA9FP) were found to be approximately 20% and 90%,
Advanced oxidation processes
respectively, with 50 mJ/cm2 UV exposure and 1 mg/L TiO2. For the fractionated samples,
TiO2
approximately 75% of both trihalomethane (THM) and haloacetic acid (HAA) precursors
UV
were found to be associated with the hydrophobic acid fraction. For this individual fraction
NOM fractionation
the same UV/TiO2 treatments exhibited approximately 20e25% removal of both TTHMFP and THAA9FP, suggesting that the fractionation process may have affected the treatability of HAA precursors or may have altered the results of the oxidation processes. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
The formation of disinfection by-products (DBPs) has undergone significant scrutiny in recent years due to the carcinogenic potential associated with these substances (Bull, 1983). The USEPA has implemented the Stage 2 Disinfectant and Disinfection Byproducts Rule with maximum contaminant levels of 80 mg/L for total trihalomethanes (TTHM) and 60 mg/L for total haloacetic acids (THAA) in order to control the formation of these compounds within drinking water
treatment systems (USEPA, 2006). This is especially important for surface water sources whose natural organic matter (NOM) levels can be high. The removal of NOM from source waters can reduce the formation of DBPs within disinfection processes (Minear and Amy, 1996). This can be achieved through coagulation processes; however, coagulation develops a residual product that requires treatment prior to disposal (Bourgeois et al., 2004). By contrast, advanced oxidation processes (AOPs) oxidize NOM without the formation of waste residuals.
* Corresponding author. Tel.: þ1 223 71 279722. E-mail addresses: [email protected] (F.C. Kent), [email protected] (K.R. Montreuil), [email protected] (R.M. Brookman), [email protected] (R. Sanderson), [email protected] (J.R. Dahn), [email protected] (G.A. Gagnon). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.09.013
6174
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 7 3 e6 1 8 0
The use of advanced oxidation processes (AOPs) for removal of DBP precursors has undergone investigation including processes such as UV/H2O2 (Toor and Mohseni, 2007), Ozone/UV (Chin and Berube, 2005) and the PhotoFenton process (Moncayo-Lasso et al., 2008). More recently the use of UV/TiO2 for DBP precursor removal has been evaluated; however there is still a lack of understanding regarding the mechanisms by which DBP precursors are removed through these processes (Gerrity et al., 2009; Liu et al., 2008; Philippe et al., 2010). Titanium dioxide has been employed as a catalyst for UV oxidation and has been the focus of many studies, some of which are summarized by Blanco-Galvez et al. (2007). This advanced oxidation process (AOP) has strong potential for removing NOM compounds that are otherwise difficult to degrade. One advantage of using TiO2 with UV is that it is not consumed during the formation of hydroxyl radicals and can be retained for repeated use (Malato et al., 2009). A disadvantage, however, is the cost associated with separation of TiO2 particles from suspensions used in UV/TiO2 oxidation processes. This can be overcome by attaching TiO2 particles to surfaces, thereby eliminating the need for downstream particle separation (Choi et al., 2007). The practical application of this technology for removal of DBP precursors could involve a separate UV/TiO2 process upstream of chlorination processes. This would ensure all remaining DBP precursors present in the raw water as well as those formed in other unit operations (e.g. coagulation, flocculation, sedimentation, filtration) are subjected to this advance oxidation treatment. Liu et al. (2008) divided raw water DOC into four organic fractions and each fraction was characterized before and after UV/TiO2 oxidation. They found that UV with 100 mg/L suspended TiO2 could substantially reduce the total trihalomethane formation potential (TTHMFP) of the surface waters used in their study. In contrast, Liu et al. (2008) found that the total haloacetic acid formation potential (THAA5FP) was increased after short exposures and ultimately remained unchanged from its original raw water value. Gerrity et al. (2009) compared an UV/TiO2 pilot-scale system with enhanced coagulation processes for DBP precursor reductions. These researchers found that UV/TiO2 could dramatically decrease TTHMFP. At limited exposure levels, however, they found increases in TTHMFP. These studies show the need for a better understanding of the degradation pathways of NOM compounds for the removal of DBP precursors before this technology can be applied in real systems. The objectives of the present work were to investigate these DBP degradation pathways by examining surface water after limited UV/TiO2 treatment and to evaluate an immobilized TiO2 configuration. Surface water was divided into six DOC fractions according to a technique described by Leenheer (1981). Each fraction was exposed to three different UV/TiO2 oxidation treatments. The UV exposures were constant for each test while different TiO2 treatments were used (1 mg/L and 10 mg/L). Low TiO2 concentrations, such as those used in the present work, encourage the formation of intermediates and minimize the impact of adsorption of organics onto TiO2 surfaces as observed by Philippe et al. (2010). In addition, the use of fixed TiO2 particles sputtered onto surfaces is compared with TiO2 suspensions to assess
the potential of fixed TiO2 within UV/TiO2 processes for DBP precursor removal.
2.
Materials & methods
Samples of French River water were analyzed for DOC, pH, turbidity, UV254 and selected ions. These samples were also tested for DBP formation potential as described below. The raw water was separated into six NOM fractions and each fraction was analyzed for DBP formation potential. The raw and fractionated samples were then exposed to three different UV/TiO2 treatments and analyzed again for DBP formation potential to evaluate the DBP degradation potential of each fraction.
2.1.
Water quality methods
Dissolved organic carbon (DOC) samples were analyzed using a Shimadzu TOC-Vcph Total Organic Carbon Analyzer. The samples were filtered through a 0.45 mm filter paper (ColeeParmer Nylon Membranes) and placed in 50 mL headspace free vials before acidifying the samples below a pH of 2 with phosphoric acid. Temperature and pH were measured using an Accumet Excel XL50. The pH probe was calibrated daily using standard buffer solutions (pH 2, 4, 7, 10) from Fisher Scientific. The temperature probe was also calibrated daily using a mercury thermometer. All ions were analyzed using a Metrohm 761 Compact ion chromatograph (Fisher Scientific, USA).
2.2.
NOM fractionation methods
Using the fractionation procedure developed by Leenheer (1981) and later modified by Marhaba et al. (2003), the dissolved organic matter from the raw and filtered water was separated into six organic fractions; hydrophobic acid (HON), base (HOB) and neutral (HON) and hydrophilic acid (HIA), base (HIB) and neutral (HIN). DAX-8 resins, procured from SUPELCO, were used to absorb the hydrophobic organic fractions. Diaion WA-10 and AG-MP 50 resins, procured from Bio-Rad Laboratories, were used to absorb the HIA and HIB fractions respectively. New DAX-8 resins were passed through a 500 mm sieve to remove large resins. The resins were then stored in 0.1 N NaOH for 24 h before sequential 24 h cleanings with hexane and acetone using a soxhlet extractor (Leenheer, 1981). The clean resins were then packed into 2.5 120 cm Kontes Chromaflex chromatography columns and further prepared by passing methanol, 0.1 NaOH, 0.1 HCl and Milli-Q through each column (Leenheer, 1981). Resin quantities in each column were determined using the resin absorption quantities calculated by Leenheer (1981). Additional resin cleaning and preparation instructions for Diaion WA10 and AG-MP 50 resins can be found in Leenheer (1981). Before passing the sample through each column, the conductivity and absorbance (UV254) of the Milli-Q effluent were determined to ensure a conductivity of <10 mS/cm and an absorbance <0.001 cm1. Additionally, DOC samples were collected to determine the DOC of the resin bleed prior to
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 7 3 e6 1 8 0
6175
Fig. 1 e Schematic for the resin fraction procedure described by Marhaba et al. (2003).
passing the water sample through. Fig. 1 describes the experimental procedure used in this study. The left-hand axis describes the pH to which the sample was adjusted prior to passing it through the columns. The right hand axis describes the eluent used to desorb the desired organic material from the resins. The horizontal axis describes the order in which water was passed through each of the five columns.
2.3.
DBP formation potential methods
Trihalomethane and haloacetic acid total formation potentials (THMfp and HAAfp) were analyzed for each of the six organic fractions extracted at each sampling event. THM and HAA samples were prepared following Standard Methods (5710) with minor modifications (APHA, 1995). Samples were buffered to a pH of 8 with borate and stored for 24 h after dosing with chlorine. Chlorine dosing under uniform conditions varied between samples such that the residual chlorine, after a 24 h incubation period, was 1 0.4 mg/L (Summers et al., 1996). THM and HAA samples were further prepared for gas chromatography analysis using liquideliquid extraction with pentane and tert-butyl ether (MTBE) respectively. Gas chromatography using a Varian CP-3800 GC and a Varian
CP-8400 auto-sampler, coupled with an electron capture detector (GC-ECD) were used for the detection of THMs and HAAs according to the US EPA Methods 551.1 and 552.2 (Hodgeson and Cohen, 1990). GC measurements were analyzed using a Hewlett Packard 5890 Series II e Plus GC equipped with a DB-5 column and a DB-1701 column. Four THMs were measured; chloroform, dichlorobromomethane, dibromochloro-methane and bromoform. Nine HAAs were measured; chloroacetic acid, dichloroacetic acid, trichloroacetic acid, bromoacetic acid, dibromoacetic acid, tribromoacetic acid, bromochloroacetic acid bromodichloroacetic acid and chlorodibromoacetic acid.
2.4.
TiO2 surface coating
Debe et al. (1994) describe the production and characterization of nanostructured thin film (NSTF) catalyst supports developed by 3 M Company (St. Paul, MN) for use in the fuel cell industry. The NSTF, shown in Fig. 2, consists of 3e5 billion crystalline organic pigment support whiskers per cm2 attached to an advanced polyimide carrier sheet that can be coated with any desired material (Bonakdarpour et al., 2008; Dahn et al., 2002; Debe, 2003; Debe and Drube, 1995; Debe
Fig. 2 e Nano-structured thin film support coated with 300 nm of TiO2 at two different magnifications.
6176
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 7 3 e6 1 8 0
and Poirier, 1994). For this study, NSTF was coated with TiO2 by magnetron sputter deposition as described by Dahn et al. (2002) using two TiO2 targets and constant mask openings. The depositions produced catalyst-coated whisker films with a planar loading equivalent of 300 nm or 0.125 mg TiO2 cm2 (Fig. 2).
2.5.
TiO2eUV methodology
Three UV/TiO2 treatments were applied, including one with suspended TiO2 and two with TiO2 fixed on a NSTF as described above. All sample volumes were 250 mL with TiO2 quantities adjusted according the targeted concentrations. For suspended TiO2 experiments, a stock solution of Degussa P25 anatase titanium (IV) oxide nanopowder (99.7% trace metals basis, Sigma Aldrich, St. Louis, MO) was mixed with Milli-Q water and continuously stirred prior to being spiked into the sample and exposed to UV light. Samples were irradiated using a low pressure UV lamp emitting at 253.7 nm in a collimated beam (Trojan Technologies, London, ON) by following the standardized testing protocol for collimated beam disinfection (Kuo et al., 2003). Fluence values were achieved using a radiometer (International Light Technologies 1400A, Peabody, MA) in accordance with the methods described by Bolton and Linden (2003) which take into account water, divergence, reflection and Petri factors associated with collimated beam experiments. Exposure durations were adjusted based on the measured fluence values to give the targeted exposure of 50 mJ/cm2. For fixed TiO2 treatments NSTF coupons of 2 cm2 and 20 cm2 were used for the 1 mg/L and 10 mg/L experiments, respectively. Based on the NSTF TiO2 sputtered density (0.125 mg/cm2), these coupon sizes provided the effective concentrations employed in the experiments.
3.
Results & discussion
3.1.
Water quality results
The raw water quality (Table 1) of the French River is typical of other river waters in the region. As with most river water
sources there is a high degree of fluctuation in quality resulting from rainfall events, however; these data are representative of the water quality over the period of sampling. The DOC concentrations for the raw water and for the three samples exposed to different UV/TiO2 treatments are shown in Fig. 3. The removal of DOC for all three treatments was less than 15%. This indicates that the UV/TiO2 processes did not result in complete mineralization of the DOC present and that most of the DOC compounds found in the raw water were degraded into intermediates. Blanco et al. (2007) highlights the weaknesses of using DOC for assessing the removal of compounds with AOPs and reinforces the need to monitor specific compounds in order to evaluate their removals. Given the relatively low concentration of TiO2 applied in this study, the results provide a good opportunity to understand the nature of intermediates and their potential for DBP formation. It should be noted that the intermediate organic compounds formed may have an effect on the stability of the DOC present. For example the biodegradability of the treated DOC may be improved resulting in higher assimilable organic compound (AOC) levels.
3.2.
DBP formation potential results
The primary goal of this work was to examine the removal of DBP precursors resulting from the various UV/TiO2 treatments employed. Fig. 4 shows the TTHMFP and THAA9FP values for the raw water and how this formation potential changed after the three different UV/TiO2 treatments were applied. As shown in Fig. 4, none of the three UV/TiO2 treatments exhibited a large effect on the formation potential of THM compounds. Less than 20% removal was found for all treatments. Other work using higher TiO2 doses have shown a high degree of removal of THM precursor compounds (Liu et al., 2008; Philippe et al., 2010). The lack of removal of TTHMFP with the treatments that were employed in the present study emphasizes the inability of UV alone to achieve the high THM removals that have been observed in these other studies. Furthermore, Philippe (2010), who used 1000 mg/L of TiO2 in their study, identified a high degree of removal of THM precursors in the absence of UV light with TiO2, indicating that a significant degree of absorption influenced the THM precursor removals.
Table 1 e French River water quality summary. Analyte pH UV254 Alkalinity DOC SUVA Turbidity Sodium Magnesium Calcium Chloride Nitrite Nitrate Phosphate
Units cm1 mg/L (as CaCO3) mg/L L/mg DOC-m NTU mg/L mg/L mg/L mg/L mg/L mg/L mg/L
Value 6.7 0.096 <5 5.34 1.8 0.92 4.1 0.74 4.8 5.0 0.14 0.38 2.0
Fig. 3 e DOC concentration of raw treated samples.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 7 3 e6 1 8 0
Fig. 4 e Disinfection by-product formation potential of raw and treated waters.
In contrast, the THAA9FP was reduced with UV/TiO2 treatment; this is especially noteworthy for the treatment with suspended TiO2 which demonstrated a removal of more than 88% despite the low concentration of TiO2 (1 mg/L). In a similar study carried out by Liu et al. (2008), using 100 mg/L TiO2, very little HAA removal was found. They observed that, after 30 min of exposure, the specific THAA5FP concentration actually increased suggesting that intermediates formed in the UV/TiO2 oxidation process contributed to the THAAFP. The relatively low TiO2 dose (1 mg/L) used in the present study may have provided some degradation of HAA precursors while preventing the formation of intermediates that contribute to THAAFP such as those found by Liu et al. (2008). Overall, the TTHMFP and THAA9FP results indicate a high degree of removal of THAA precursors and very little removal of TTHM precursors. Both of these observations contradict the findings reported in the study by Liu et al. (2008) where high TTHMFP removal and low THAAFP removal were observed. It
6177
should be noted that differences in the raw water DOC composition in the present work and the work of others may have contributed significantly to the differences found in TTHMFP and THAA9FP removal rates. The UV/TiO2 process can lead to the formation of intermediate organic matter depending on the TiO2 dose. Thus, a better understanding of the degradation pathways may lead to effective treatment options for the removal of DBP precursors using UV/TiO2 advanced oxidation technology. The individual THM and HAA species are shown in Figs. 5 and 6, respectively, for the raw water and after each of the three treatments. Of the four species analysed, as shown in Fig. 5, almost all of the TTHMFP was associated with chloroform, which had an approximate formation potential of 400 mg/ L. The formation potentials of dichlorobromomethane and dibromochloro-methane were less than 25 mg/L while the formation potential of bromoform was below the detection limit. Similar to Figs. 4 and5 shows that the different UV/TiO2 treatments had very little effect on chloroform formation potential. Fig. 6 shows that of the nine HAA species analysed, almost all of the THAA9FP was associated with di- and trichloroacetic acids with values of approximately 220 mg/L and 130 mg/L, respectively. The treatment with UV and 1 mg/L suspended TiO2 resulted in a removal of approximately 90% of formation potential for both of these species. This is a very interesting result given the low concentration of TiO2 compared with previous work where 100 times the concentration of TiO2 was used and the THAA5FP remained unchanged (Liu et al., 2008). The results for the UV/TiO2 treatments with 1 mg/L and 10 mg/L sputtered TiO2 showed much less removal with values closer to approximately 30% for dichloroacetic acid and 16% for trichloroacetic acid. These results suggest that sputtered TiO2 is a less efficient catalyst configuration for the degradation of these compounds than suspended TiO2. It is possible that the lower removal found for HAA precursors with sputtered TiO2 was associated with mass transfer limitations. The 1 mg/L TiO2 suspension had a higher surface area of contact with the sample than the flat sputtered coupons used in these tests since the particles in the suspension were
Fig. 5 e Formation potential of different THM compounds for raw and UV/TiO2 treated waters.
6178
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 7 3 e6 1 8 0
Fig. 6 e Formation potential of different HAA compounds for raw and UV/TiO2 treated samples.
spread throughout the sample. Although the sputtered TiO2 had a very high surface area due to the NSTF surface morphology, compounds within the sample would first have to migrate to the NSTF surface in order to be exposed to the oxidants. These oxidants are generally short-lived and would not migrate far from the NSTF coupon. Figs. 5 and 6 also show that the treatments with 10 mg/L sputtered TiO2 did not show a higher degree of removal compared to the same treatment with 1 mg/L sputtered TiO2. This suggests that the amount of available TiO2 catalyst on the surface was not limiting the oxidation process and strengthens the hypothesis that the oxidation process was limited by mass transfer of DBP precursors to the NSTF surface. The experiments in this study were not carried out in mixed systems; future work should examine the effect of mixing to evaluate the mass flux at the NSTF surface.
Fig. 7 e Total THM formation potential of different NOM fractions for raw and treated samples.
3.3.
NOM fractionation results
Fractionation of raw water samples was performed and each fraction was analysed for TTHMFP and THAA9FP. The six NOM fractions were also measured for formation potential of these DBPs after treatment with UV/TiO2 (1 mg/L suspended and 1 mg/L sputtered TiO2). The results for this analysis are given in Figs. 7 and 8, respectively. As shown in Fig. 7, the results suggest that the majority (75%) of TTHM formation potential came from the HOA fraction with more than 265 mg/L associated with this fraction. This suggests that only a relatively small percentage (17%) of the TTHM compounds were formed from the HIN fraction and even fewer (less than 5%) being attributed to any other fraction. Treatment of these fractions resulted in an overall removal of approximately 30% of TTHMFP for both the suspended and fixed UV/TiO2 treatments.
Fig. 8 e Total HAA9 formation potential of different NOM fractions for raw and treated samples.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 7 3 e6 1 8 0
The vast majority (74%) of THAA9FP was attributed to the HOA fraction (Fig. 8). Approximately 10% of the THAA9FP was associated with both the HIA fraction and the HIN fraction and less than 5% was attributed to the other fractions. The sum of the THAA9FP for all fractions was 153 mg/L, indicating that approximately 40% of the DOC in the raw water was recovered in the fractionation process. The UV/TiO2 treatments of the fractionated samples led to an overall removal of 33% of the THAA9FP for the treatment with suspended TiO2 and 29% for the treatment with fixed TiO2. As discussed, for both types of DBPs, most of the DBP formation potential was attributed to the HOA fraction (approximately 75%). This is similar to the results found in other work (Liu et al., 2008) for this fraction; which is made up of humic and fulvic acids. For both THM and HAA precursors approximately 20e25% of the HOA fraction was removed with the application of 1 mg/L TiO2. The sum of the DBPFP from the six fractions was lower than the DBPFP of the raw water for both THM and HAA compounds suggesting that some DOC was not recovered during the fractionation process. Croue et al. (2000) reported a DOC recovery between 60 and 80% using XAD-8 and XAD-4 ion exchange resins. While the overall recovery of DOC using DAX-8, AG-MP 50 and WA-10 ion exchange resins was not reported, it is reasonable to expect some loss. In the present work only 40% of THAA9FP was recovered, somewhat lower than the value found by Croue et al. (2000). This may be due to differences in the specific resins used, or alternatively due to differences in synergistic DBPFP caused by interaction between multiple NOM fractions. When these fractions are separated, the resulting DBPFP may be reduced. Differences in the raw water may also have impacted the recovery of DOC, such as the presence of more organic compounds that bind strongly to resins. The THAA9FP in the raw water was reduced by 90% with suspended TiO2 treatment. For the fractionated results, however, the overall THAA9FP removal was approximately 33%. This suggests that the HAA precursors in the fractionated samples were more difficult to degrade with the applied processes than the HAA precursors in the raw water (on a percentage basis). It is possible that the DOC that was not recovered during the fractionation process was associated with HAA precursors that are more easily removed. This could explain why fractionating the DOC resulted in lower percent removals of HAA precursors. However, there was less THAA9FP in the treated raw water (44 mg/L) than there was in the sum of the treated fractions (102 mg/L). If the difference in removal rates was caused only by a loss of easily degraded DBP precursors in the fractionation process, the final concentration of THAA9FP in the treated water for the sum of the fractions would be less than or equal to the concentration found for the treated raw water. These differences in HAA precursor removal rates can only be explained by a change in the nature of the DOC after fractionation, either by the addition of DOC that leached from the resins, or changes in the original organic compounds as a result of the fractionation process. This is explained by the fact that compounds within a particular fraction are formed from smaller aliphatic and amphiphillic molecules (Croue et al., 2000). For future work in this area, it may be better to perform fractionation following
6179
UV/TiO2 treatment rather than to provide UV/TiO2 treatment following fractionation. In bench-scale experiments such as in the present work, it is not feasible to fractionate TiO2 treated samples due to the low sample volumes used. This suggests that the use of alternative fractionation processes or pilotscale work may be more appropriate for future investigations of this nature.
4.
Conclusions
The removal of DBP precursors from surface water through treatment with UV/TiO2 was investigated. A relatively low TiO2 concentration was employed to examine the DBP formation potential of intermediates formed given limited catalyst quantities. The following conclusions have been made: The UV/TiO2 treatments used were not sufficient for complete removal of DOC, but were able to reduce DBP formation potential of THMs to a limited degree (20%) and HAAs to a large degree (90%) The fixed TiO2 configuration was less effective than the suspended TiO2 at removal of DBPs For the fixed TiO2 configuration, the rate of removal was not governed by the surface area of the NSTF coupon The vast majority of all DBPs formed were in the form of chloroform, dichloroacetic acid and trichloroacetic acid The vast majority of all DBP precursors were associated with the hydrophobic acid (HOA) fraction of the DOC present Removal of THAA9FP from the fractionated water was much lower than the removal found for the raw water The findings of this work and other similar studies highlight the complexity of the degradation pathways involved in DBP precursor degradation using UV/TiO2 and the need for further work in this area.
Acknowledgments The authors would like to acknowledge the valuable contributions of Heather Daurie as well a Trojan Technologies for providing the UV equipment. The authors also thank Mark Debe of 3M for supplying the NSTF used in this project. Funding support was provided from the NSERC/Halifax Water Industrial Research Chair program.
references
APHA, 1995. Standard Methods for the Examination of Water and Wastewater, nineteenth ed. American Water Works Association and the Water Environment Federation, USA. Blanco-Galvez, J., Fernandez-Ibanez, P., Malato-Rodriguez, S., 2007. Solar photo catalytic detoxification and disinfection of water: recent overview. J. Sol. Energ. -T. ASME 129 (1), 4e15. Bolton, J., Linden, K., 2003. Standardization of methods for fluence (UV dose) determination in bench-scale UV experiments. J. Environ. Eng. 129 (3), 209e215.
6180
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 7 3 e6 1 8 0
Bonakdarpour, A., Dahn, T.R., Atanasoski, R.T., Debe, M.K., Dahn, J.R., 2008. H2O2 release during oxygen reduction reaction on Pt nanoparticles. Electrochem. Solid St. 11 (11), B208eB211. Bourgeois, J.C., Walsh, M.E., Gagnon, G.A., 2004. Comparison of process options for treatment of water treatment residual streams. J. Environ. Eng. Sci. 3 (6), 408e416. Bull, R.J., 1983. Health Risks of Drinking Water Disinfectants and Disinfection By-products. Pacific Grove, CA, USA, pp. 1401e1415. Chin, A., Berube, P.R., 2005. Removal of disinfection by-product precursors with ozone-UV advanced oxidation process. Water Res. 39 (10), 2136e2144. Choi, H., Stathatos, E., Dionysiou, D.D., 2007. Photocatalytic TiO2 films and membranes for the development of efficient wastewater treatment and reuse systems. Desalination 202 (1e3), 199e206. Croue, J.P., Korshin, G.V., Benjamin, M., 2000. Characterization of Natural Organic Matter in Drinking Water. AWWA Research Foundation and AWWA Association. Dahn, J.R., Trussler, S., Hatchard, T.D., Bonakdarpour, A., MuellerNeuhaus, J.R., Hewitt, K.C., Fleischauer, M., 2002. Economical sputtering system to produce large-size composition-spread libraries having linear and orthogonal stoichiometry variations. Chem. Mater. 14 (8), 3519e3523. Debe, M.K., 2003. In: Vielstich, W., Gasteiger, H.A., Lamm, A. (Eds.), Handbook of Fuel Cells - Fundamentals, Technology and Applications. John Wiley & Sons, USA. Debe, M.K., Drube, A.R., 1995. Structural characteristics of a uniquely nanostructured organic thin film. J. Vac. Sci. Technol. B. Microelectron. Nanometer Struct. 13 (3), 1236e1241. Debe, M.K., Poirier, R.J., 1994. Postdeposition growth of a uniquely nanostructured organic film by vacuum annealing. J. Vac. Sci. Technol. A, Vac. Surf. Films 12 (4), 2017e2022. Gerrity, D., Mayer, B., Ryu, H., Crittenden, J., Abbaszadegan, M., 2009. A comparison of pilot-scale photocatalysis and enhanced coagulation for disinfection byproduct mitigation. Water Res. 43 (6), 1597e1610. Hodgeson, J.W., Cohen, A.L., 1990. Determination of Chlorination Disinfection Byproducts, Chlorinated Solvents, and
Halogenated Pesticides/herbicides in Drinking Water by Liquid-liquid Extraction and Gas Chromatography with Electron Capture Detection. Environmental Protection Agency, Cincinnati, USA. Kuo, J., Chen, C.L., Nellor, M., 2003. Standardized collimated beam testing protocol for water/wastewater ultraviolet disinfection. J. Environ. Eng. 129 (8), 774e779. Leenheer, J.A., 1981. Comprehensive approach to preparative isolation and fractionation of dissolved organic carbon from natural waters and wastewaters. Environ. Sci. Technol. 15 (5), 578e587. Liu, S., Lim, M., Fabris, R., Chow, C., Drikas, M., Amal, R., 2008. TiO2 photocatalysis of natural organic matter in surface water: impact on trihalomethane and haloacetic acid formation potential. Environ. Sci. Technol. 42 (16), 6218e6223. Malato, S., Fernandez-Ibanez, P., Maldonado, M.I., Blanco, J., Gernjak, W., 2009. Decontamination and disinfection of water by solar photocatalysis: recent overview and trends. Catal. Today 147 (1), 1e59. Marhaba, T.F., Pu, Y., Bengraine, K., 2003. Modified dissolved organic matter fractionation technique for natural water. J. Hazard. Mater. 101 (1), 43e53. Minear, R.A., Amy, G.L., 1996. Water disinfection and natural organic matter: history and overview. ACS Symp. Ser. 649, 1e9. Moncayo-Lasso, A., Pulgarin, C., Benitez, N., 2008. Degradation of DBPs’ precursors in river water before and after slow sand filtration by photo-Fenton process at pH 5 in a solar CPC reactor. Water Res. 42 (15), 4125e4132. Philippe, K.K., Hans, C., MacAdam, J., Jefferson, B., Hart, J., Parsons, S.A., 2010. Photocatalytic oxidation of natural organic matter surrogates and the impact on trihalomethane formation potential. Chemosphere 81 (11), 1509e1516. Summers, R.S., Hooper, S.M., Shukairy, H.M., Solarik, G., Owen, D., 1996. Assessing DBP yield: uniform formation conditions. J. AWWA 88 (6), 80e93. Toor, R., Mohseni, M., 2007. UV-H2O2 based AOP and its integration with biological activated carbon treatment for DBP reduction in drinking water. Chemosphere 66 (11), 2087e2095. USEPA, 2006. National primary drinking water regulations: stage 2 disinfectants and disinfection byproducts rule; Final rule. 40 CFR Parts 9, 141 and 142. Fed. Regist. 71 (2), 388.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 8 1 e6 1 8 8
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
The impact of pH on floc structure characteristic of polyferric chloride in a low DOC and high alkalinity surface water treatment Baichuan Cao, Baoyu Gao*, Xin Liu, Mengmeng Wang, Zhonglian Yang, Qinyan Yue School of Environmental Science and Engineering, Shandong University, No. 27 Shanda Nan Road, Jinan 250100, PR China
article info
abstract
Article history:
The adjustment of pH is an important way to enhance removal efficiency in coagulation
Received 16 May 2011
units, and in this process, the floc size, strength and structure can be changed, influencing
Received in revised form
the subsequent solid/liquid separation effect. In this study, an inorganic polymer coagulant,
5 September 2011
polyferric chloride (PFC) was used in a low dissolved organic carbon (DOC) and high alkalinity
Accepted 7 September 2011
surface water treatment. The influence of coagulation pH on removal efficiency, floc growth,
Available online 16 September 2011
strength, re-growth capability and fractal dimension was examined. The optimum dosage was predetermined as 0.150 mmol/L, and excellent particle and organic matter removal
Keywords:
appeared in the pH range of 5.50e5.75. The structure characteristics of flocs formed under
Polyferric chloride
four pH conditions were investigated through the analysis of floc size, effect of shear and
Surface water treatment
particle scattering properties by a laser scattering instrument. The results indicated that
pH
flocs formed at neutral pH condition gave the largest floc size and the highest growth rate.
Floc growth
During the coagulation period, the fractal dimension of floc aggregates increased in the first
Floc breakage and re-growth
minutes and then decreased and larger flocs generally had smaller fractal dimensions. The
Fractal dimension
floc strength, which was assessed by the relationship of floc diameter and velocity gradient, decreased with the increase of coagulation pH. Flocs formed at pH 4.00 had better recovery capability when exposed to lower shear forces, while flocs formed at neutral and alkaline conditions had better performance under higher shear forces. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
Coagulation is one of the most important processes used for the removal of fine particles and natural organic matter (NOM) in drinking water treatment (Hu et al., 2006). Inorganic coagulants, such as aluminum and ferric salts, are widely used to remove the pollutants by charge neutralization, adsorption, entrapment and complexation (Matilainen et al., 2010; Zhao et al., 2011). Aluminum coagulants are reported to have potential hazard to human health (Gauthier et al., 2000). Therefore, ferric coagulants, especially polyferric coagulants, have been more widely
used recently. It has been found that ferric coagulants are more efficient for NOM removal and less sensitive to low temperatures than aluminum coagulants (Edzwald and Tobiason, 1999). Moreover, the flocs formed by ferric salts are generally larger than those formed by aluminum (Jarvis et al., 2005b; Zhao et al., 2011). Polyferric chloride (PFC), a kind of polyferric coagulant, has received more attention recently (Zhan et al., 2010). The physical properties of flocs, such as floc size, strength and compaction, may significantly affect the efficiency of solid/ liquid separation (Boller and Blaser, 1998). In the coagulation and separation units and transfer weirs, regions of high shear
* Corresponding author. Tel.: þ86 531 88364832. E-mail addresses: [email protected] (B. Cao), [email protected] (B. Gao), [email protected] (X. Liu), [email protected] (M. Wang), [email protected] (Z. Yang), [email protected] (Q. Yue). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.09.019
6182
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 8 1 e6 1 8 8
are very common, which may result in the breakage of flocs and lower removal efficiency (Jarvis et al., 2005a). The strength of flocs depends on floc structure, floc formation mechanisms and the interparticle bonds (Yu et al., 2009). Floc strength can be measured by applying an increasing shear rate to the preformed flocs and relating the energy dissipation to the broken floc size. And the empirical relationship has been reported as (Parker et al., 1972): logd1 ¼ logC glogG
(1)
where d1 is the median floc diameter after breakage, mm; C is the floc strength coefficient; g is the stable floc exponent; and G is the average velocity gradient in the system, s1. The broken flocs are capable to reaggregate, which may improve the subsequent solid/liquid separation efficiency (Yukselen and Gregory, 2004). The re-growth capability is usually evaluated by floc recovery factor (Rf), which is generally defined as follows (Yu et al., 2010; Zhao et al., 2011): Rf ¼
d2 d1 100% d0 d1
(2)
where d0 is the average floc size of the steady phase before breakage, d1 is the floc size after floc breakage, and d2 is the floc size after reaggregation to a new steady phase. Fractal geometry, which was founded by Mandelbrot in 1975, has been widely used to describe the structure of particles, such as latex, aluminum oxide, ceramic materials and floc aggregates (Gregory, 1998; Jarvis et al., 2005b; Schaefer, 1989; Waite et al., 2001). The aggregates structure can be simply described by a parameter Df, fractal dimension, which was defined as the exponent of relationship of mass (M ) and size (L) (Rieker et al., 2000): MfLDf
(3)
Similar relationships can be obtained regarding the volume and sedimentation velocity to particle size (Jiang and Logan, 1991). As having been reported in the previous studies (Bridgeman et al., 2009), flocs with open structure have low Df values, whereas, high Df values generally indicate more compact structures. Approximation of floc fractal dimension can be achieved by a number of different techniques, such as image analysis (Chakraborti et al., 2000), settling experiments (Jarvis et al., 2006) and small angle laser light scattering (SALLS) (Jarvis et al., 2005b; Lin et al., 2008; Wei et al., 2009). The SALLS technique has been successfully used for the determination of various suspension systems over a wide range of particle size, such as hematite particles (5e13 mm) (Lee et al., 2005), salthumic flocs (100e200 mm) (Wei et al., 2009), salt-kaolin flocs (60e650 mm) (Li et al., 2006) and ferric precipitate (1000 mm) (Jarvis et al., 2005b). In the enhanced coagulation process, the adjustment of pH is an important way to improve turbidity and NOM removal efficiency (Gregory and Carlson, 2003). Previous studies have given the optimum pH range 5.0e6.5 for aluminum coagulants and 4.5e6.0 for ferric coagulants (Matilainen et al., 2010). The influence of pH on coagulation effect has been well investigated (Gregor et al., 1997), and the mechanisms of NOM removal at different coagulation pH conditions have been widely accepted (Matilainen et al., 2010). Gregory and Carlson
(2003) investigated the floc formation kinetics of aluminum in the pH range of 6.0e7.5 via a photometric dispersion analyzer (PDA). Xu et al. (2010) studied the impact of pH on properties of Al13-HA flocs by a laser diffraction instrument. Wang et al. (2011) investigated the influence of pH on fractal dimensions of PFC-HA flocs based on the image analysis technique. However, the combined investigation of floc size, strength, regrowth and fractal dimension based on the SALLS technique has not been well developed. Furthermore, the recent works investigated the formation, strength and fractal dimension of flocs in modeling water system, and the application of SALLS technique has not been fully developed in actual surface water system. In this study, the PFC coagulant was used to treat a low dissolved organic carbon (DOC) and high alkalinity surface water, and the investigation of pH influence on coagulation behavior, floc formation, strength, re-growth and fractal structure, is considered to be significant for the enhanced coagulation process in water treatment plants.
2.
Materials and methods
2.1.
Materials and raw water
All reagents used were of analytical grade. Deionized water was used to prepare all solutions. The raw water investigated in the study was sampled from Loukou section of the Yellow River. The raw water was allowed to settle for 24 h, and the supernatant was then withdrawn by siphon from upper-middle of the container carefully and stored in refrigerator for subsequent experiments. The settled water is generally defined as a low DOC and high alkalinity water. In this study, the properties of the supernatant were: temperature ¼ 20.2e24.5 C, pH ¼ 8.15e8.39, turbidity ¼ 8.42e15.5 NTU, UV absorbance at 254 nm wavelength (UV254) was 0.062e0.070 cm1, DOC ¼ 1.86e3.71 mg/L, alkalinity ¼ 245.1e262.7 mg CaCO3/L.
2.2.
Preparation of coagulant
The PFC used in this experiment was prepared with FeCl3$6H2O and Na2CO3 as raw materials. Firstly, a certain concentration of FeCl3$6H2O was dissolved in deionized water, then Na2CO3 powder was gradually added to FeCl3 solutions with stirring at room temperature to reach the final [OH-]/[Fe] ratio (B) of 0.5 (Wei et al., 2009; Zhan et al., 2010). The solution was stirred until foam disappeared and became transparent. Then, Na2HPO4$12H2O was added to the solution as a stabilizer ([Na2HPO4]/[Fe] ¼ 0.08). The concentration of Fe was about 5% (w/w) in the target PFC solution. The dosages of PFC were calculated as mmol/L of Fe during coagulation experiments.
2.3.
Jar test
Standard jar tests were conducted on a program-controlled jar test apparatus (ZR4-6, Zhongrun Water Industry Technology Development Co. Ltd., China). The test involved 1 min rapid mixture at 200 rpm, a 15 min 40 rpm coagulation stage, and a 15 min settlement period. The relationship of velocity
6183
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 8 1 e6 1 8 8
where P0 is the impeller power number; N is the impeller speed, rps; D is the impeller diameter, m, m is viscosity, Pa/s; V is vessel volume, m3;. The mean velocity gradients of the two stirring stages were 101.7 s1 and 11.7 s1, respectively. Sample was collected from about 2 cm below the surface and portion was filtered through a 0.45 mm membrane to measure the UV absorbance at 254 nm wavelength (UV254) using a JH-752 UV/VIS spectrophotometer (Jinghua Scientific Instrument Co. Ltd., Shanghai, China), and DOC using a TOCVPH analyzer (Shimadzu Co., Japan). The residual turbidity was measured via a 2100P turbidimeter (Hach Co., US). About 10 mL of sample was taken immediately for the measurement of zeta potential by a Zetasizer 3000 HSa (Malvern Instruments, UK) after the 1 min rapid mixture (Hu et al., 2006). NaOH and HCl solutions with concentrations of 0.1 and 0.01 mol/L were added into water samples at first to adjust the sample pH.
2.4. Dynamic determination of floc size and fractal dimension The flocs were formed by performing a series of jar tests. A program-controlled jar test apparatus (ZR4-2, Zhongrun Water Industry Technology Development Co. Ltd., China) was used with 50 40 mm flat paddle impellers with cylindrical jars containing 1 L samples. After the slow stir phase the effect of increasing shear was investigated by increasing the shear force on the jar tester for a further 15 min. Floc breakage was performed at stirring speeds of 50, 75, 100, 150, and 200 rpm. The calculated respective mean G values of each shear speed were 15.9, 27.4, 40.0, 69.1, and 101.7 s1. After the breakage period, slow stirring at 40 rpm (G ¼ 11.7 s1) was introduced for another 15 min to allow the flocs to reaggregate. A laser diffraction instrument Mastersizer 2000 (Malvern Instruments, UK) was used to measure dynamic floc size during the whole experiment. Previous researches have reported the determination of aggregate mass fractal dimension by using Mastersizer 2000 as a small angle laser light scattering (SALLS) (Jarvis et al., 2005b; Lin et al., 2008; Wei et al., 2009). In the Mastersizer instrument, a laser light beam with the wavelength of 632.8 nm passes through the suspension. There are 52 photosensitive detectors in the Mastersizer 2000 for receiving the scatter light within the angle of 0.01e40.6 . The total scattered light intensity I, the scattering vector Q, and the aggregate fractal dimension Df followed a power law, which is shown as (Rieker et al., 2000; Waite et al., 2001): IfQ Df
(5)
The scattering vector Q is the difference between the incident and scattered wave vectors of the radiation beam in the medium (Lin et al., 2008; Waite et al., 2001), which is given by: Q¼
4pnsinðq=2Þ l
(6)
3.
Results and discussion
3.1.
Influence of pH on coagulation effect
Firstly, coagulation optimization tests were conducted to ascertain the optimum dosage for particles and organic matter removal under the raw water pH conditions. The results showed that, when the dosage was 0.060 mmol/L and above, PFC achieved excellent turbidity removal, and the residual turbidity was lower than 1 NTU. For organic matter removal by PFC, the removal efficiency increased with the increase of dosage, and excellent removal appeared when the dosages were 0.150e0.180 mmol/L. Therefore, the 0.150 mmol/L coagulant dosage was chosen for the subsequent experiments. The sample pH was adjusted by HCl and NaOH solutions, and the coagulation experiments were conducted on the jar tester with PFC dosage of 0.150 mmol/L. The variation of residual turbidity and DOC removal with increasing pH is shown in Fig. 1. It can be seen that the residual turbidity decreased as the pH increased from 4.00 to 6.25, and increased slightly as pH continually increased. In acidic conditions, the hydrolysis of PFC may be restrained, and the charge neutralization dominated the coagulation mechanism (Zhan et al., 2010). Therefore, the residual turbidity decreased with the decrease of positive charge in colloid surface. In alkaline conditions, the excellent turbidity removal may be attributed to the adsorption and physical entrapment of PFC hydrolysis products.
4.5
6 50
4.0
Residual turibidity Removal of DOC Zeta potential
3.5
4 2
3.0 0
2.5 -2
2.0 -4
1.5 1.0
-6
0.5
-8
45 40 35 30 25 20
Removal of DOC (%)
(4)
Zate potential (mV)
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi P0 N3 D5 G¼ m,V
where, n, l and q are the refractive index of the medium, the laser light wavelength in vacuum, and the scattering angle, respectively. The relationship shown in equation (5) indicates the determination of Df, which can be given by the slope of log I versus log Q by fitting a straight line. This relationship is valid only when the length scale of the analysis is much larger than the size of primary particles and much smaller than the size of floc aggregates (Jarvis et al., 2005b; Waite et al., 2001). For salt-NOM floc systems, the primary particles of NOM are heterogeneous and smaller than the detection limit of Mastersizer 2000 (20 nm). Therefore the limitation of primary particle size was neglected in the previous studies (Jarvis et al., 2005b).
Residual turbidity (NTU)
gradient (G) to the stirring speed can be calculated as following (Bridgeman et al., 2008):
15 10 5
4
5
6
7
8
9
pH
Fig. 1 e Influence of pH on coagulation effect at dosage of 0.150 mmol/L.
6184
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 8 1 e6 1 8 8
3.2.
Influence of dosage and pH on floc size and structure
The growth profiles of flocs during 0e16 min at raw water pH conditions and with PFC dosages of 0.015, 0.045, 0.090 and 0.150 mmol/L are shown in Fig. 2, where the floc size is represented by median equivalent diameter (d0.5). The result showed that the flocs grew rapidly once the coagulant was dosed to the raw water until a steady-state floc size distribution was reached. The floc growth rate can be expressed by the steady-state floc sizes (d0) and the growth time needed to reach the steady-state (Wang et al., 2009). The floc growth rates were low when the dosage was low (0.015 and 0.045 mmol/L), and did not reach the steady state during the 16 min coagulation period. With increasing dosage, the time needed to reach the steady-state became shorter and the floc size was larger. When the PFC dosage was 0.090 mmol/ L, for example, it took 11.5 min to reach the steady flocs, with
2.1 0.015 mmol/L 0.045 mmol/L 0.090 mmol/L 0.150 mmol/L
2.0
F ractal d imensio n
For organic matter removal, PFC showed a peak shape with a highest DOC removal efficiency (47.34%) at pH 5.50, where the zeta potential is þ2.47 mV. The removal of UV254 was similar to that of DOC, and the best removal (58.80%) was obtained when the pH was 5.50 (not shown in Fig. 1). It has been widely accepted that the organic matter removal mechanism involved a combination of charge neutralization, complexation of organic matters and metal ions, adsorption and entrapment (Matilainen et al., 2010; Zhao et al., 2011). Under acidic coagulation conditions, charge neutralization of ferric species and complexation between ferric species and organic compounds are the main coagulation mechanisms (Cao et al., 2010). Therefore, excellent organic matter removal was obtained in the pH region of 5.50e6.00, with the zeta potential window between 0.53 mV and þ2.47 mV. When the coagulation pH is neutral and alkaline, the organic matter and coagulant are well hydrolyzed. The organic matters tend to be adsorbed and entrapped by the PFC hydrolysates and the floc aggregates (Gregor et al., 1997). However, the existence of OHinterferes the adsorption and reaction between organic matter and coagulant hydrolysis products, which results in lower DOC removal efficiency.
1.9
1.8
1.7
1.6
0
2
4
6
8
10
12
14
16
18
Times (min)
Fig. 3 e Variation of floc fractal dimension with increasing dosage during coagulation processes (raw water pH conditions).
an average size of 696.5 mm and a growth rate of 60.56 mm/min. In comparison, when the dosage of PFC was increased to 0.150 mmol/L, the floc growth time, d0.5 and growth rate were 6 min, 746.6 mm and 124.4 mm/min, respectively. It has been reported that the growth of flocs is related to the balance between the formation and breakage of flocs (Biggs and Lant, 2000; Parker et al., 1972). The formation of flocs depends on the particles collision rate and collision efficiency factor, which decrease with the reduction of particle number in the system. However, the influence of floc breakage increases with the increase of floc size. At PFC dosage of 0.090 and 0.150 mmol/L, the flocculation of microflocs dominates the formation of flocs. The rapid increase of floc size resulting increasing breakage rate and a steady state was reached soon. When coagulant dosage was lower (0.015 and 0.045 mmol/L), the effects of charge neutralization and enmeshment were weaker. The rate of flocs growth was rather slow, however, the breakage of flocs was restricted by the small floc size. Therefore, the floc size did not reach a steady state during the experimental period.
900
0.015 mmol/L 0.045 mmol/L 0.090 mmol/L 0.150 mmol/L
800 700
800
600
Floc size d0.5 (µm)
600
Floc size d0.5 (µm)
pH=4.00 pH=5.50 pH=7.00 pH=9.00
700
500 400 300 200
500 400 300 200
100 100
0 0
-100 -2
0
2
4
6
8
10
12
14
16
Time (mins)
Fig. 2 e Increase in floc size with increasing PFC dosage under raw water pH condition.
0
2
4
6
8
10
12
14
16
18
Time (mins)
Fig. 4 e Development of floc diameters during coagulation period at different pH conditions (Dosage, 0.150 mmol/L).
6185
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 8 1 e6 1 8 8
Table 1 e Floc sizes and growth rates of PFC at different pH conditions (Dosage, 0.150 mmol/L).
d0 (mm) growth time (min) average growth rate (mm/min) Df zeta potential (mV)
4.00
5.50
7.00
9.00
669.6 14.5 47.18
622.8 13.0 47.91
724.6 9.0 80.51
611.8 8.5 71.98
1.99 þ4.31
2.50 þ2.47
1.96 5.33
2.10 7.85
The floc structure during the 16 min coagulation period at different PFC dosages was investigated in terms of fractal dimension (Df). The fractal dimension was derived from the relationship of scattered light intensity I and the scattering vector Q as shown in equation (2). In this study, the Q was restricted to the range of 2.76 104e2.99 103 nm1 to allow the power law to be applicable for the Df calculation, which gave an upper aggregate particles size region of 3.6 mm, and it was much smaller than the final floc aggregates’ median diameters (33e748 mm). The variation of Df during the 16 min coagulation process at increasing PFC dosage is shown in Fig. 3. The Df values increased in the first minutes and then decreased till the end of the slow stirring phase. The variation of fractal dimension could be explained as follows: when coagulant was added to the raw water, the negative charge on colloid particles and organic matters was neutralized by the positively charged ferric species, and the physical process was completed very quickly. Then, the destabilized particles began to interact with each other and aggregate to form compact microflocs with increasing Df values. When the coagulation proceeded, the formation of flocs was dominated by the adsorption, entrapment of coagulant hydrolyzate and the bridging of microflocs resulting in incompact structures which were indicated by the
Floc sizes d0.5 ( µm)
800 700
a
800
15.9
600 500
27.4
400
40.0
300
69.1
200
101.7
100
Floc sizes d0.5 ( µm)
pH
decrease in Df values. The decrease in fractal dimension as bridging increased for activated sludge flocs has been reported (Wu et al., 2002), where similar technique was used. As shown in Fig. 3, at higher PFC dosage (0.150 mmol/L), the decrease in Df values appeared at 1 min, and a similar decrease was observed at 12 min for the dosage of 0.015 mmol/L. The final fractal dimension of 0.015 mmol/L appeared to be anomalous, as other Df values increased with the increase of dosage. It could be attributed to the incomplete growth at the low dosage, where the formation of incompact flocs by bridging did not function completely. The influence of coagulation pH on floc growth at PFC dosage of 0.150 mmol/L was investigated and the results are shown in Fig. 4. The result showed that pH played a major role in floc formation process. In short, by increasing the coagulation pH, the growth time decreased and steady-state diameters of flocs increased, which indicated an increase in floc growth rate. When the coagulation condition was acidic, the formation of flocs was much slower, and about 47 mm/min average floc growth rate was obtained at both 4.00 and 5.50 pH conditions (Table 1). In the first 4 min after PFC was dosed, flocs grew a bit rapidly at alkaline condition than neutral condition. Afterwards, flocs in neutral suspension kept growing to achieve the largest medium diameter, 724.6 mm, while the aggregation of flocs in alkaline condition slowed down and the d0.5 remained 611.8 mm, which was the smallest in the experimental pH conditions. Coagulants gave larger growth rate, and similar or even smaller final floc sizes at alkaline conditions (Cao et al., 2010; Gregory and Carlson, 2003). This phenomenon should be explained by the formation mechanism and the structure of flocs themselves. On one hand, the charge neutralization dominated the coagulation process of PFC at lower pH, and the positively charged colloidal particles (Fig. 1, Table 1) collided with each other and aggregated to form microflocs. The bridging of microflocs came late because of the slow hydrolysis of PFC. For example, an inflection point appeared at 4.0 min for floc growth at pH 4.00
0
700
b
600
15.9
500
27.4
400
40.0
300
69.1
200
101.7
100 0
0
5
10
15
20
25
30
35
0
5
10
Time (mins)
800
c
800
20
25
30
35
d
700
600
15.9
500
27.4
400
40.0
300
69.1
200
101.7
100 0
Floc sizes d0.5 ( µm)
700
Floc sizes d0.5 ( µm)
15
Time (mins)
600
15.9
500
27.4
400
40.0
300
69.1
200
101.7
100 0
0
5
10
15
20
Time (mins)
25
30
35
0
5
10
15
20
25
30
35
Time (mins)
Fig. 5 e Floc breakage profiles at different average velocity gradients (sL1) for each pH conditions (a, pH [ 4.00; b, pH [ 5.50; c, pH [ 7.00; d, pH [ 9.00; Dosage, 0.150 mmol/L; shear time, 15 min).
6186
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 8 1 e6 1 8 8
(Fig. 4), which divided the initial formation of microflocs (growth rate of 8.18 mm/min) and the bridging of microflocs (growth rate of 58.67 mm/min). At higher pH conditions, the hydrolysis tendency of PFC was much larger, and the main coagulation mechanisms were adsorption and sweep flocculation. In this situation, particles which were still negatively charged and stable were adsorbed by the coagulant hydrolysis product to form the flocs. On the other hand, in the formation of flocs, floc growth is restrained by floc breakage, and the increase of floc size is a balance of floc formation and floc breakage (Biggs and Lant, 2000). Normally, flocs formed by adsorption and sweep, which contain large porosity, are easily broken by surface erosion during the slow stirring phase. A similar decrease was observed for PAC-NOM flocs at larger dosage, where the mechanism was considered as sweep coagulation (Lin et al., 2008). The fractal dimensions of the final flocs formed at the experimental pH conditions were calculated, which are shown in Table 1. The Df values of the aggregates were in the region of 1.96e2.50, which was comparable with previously reported Df values of ferric-NOM flocs (Jarvis et al., 2005b). Generally, the larger the floc median diameter, the smaller the fractal dimension. This phenomenon could be explained by that the final floc aggregates contained numerous colloid particles and microflocs which were bridged by coagulant hydrolyzates. Therefore, the larger flocs were made up by more particles, which generated more irregular and incompact structures with lower Df values. There was an exception at pH 9.00, which gave the smallest d0 of 611.8 mm and the second largest Df value of 2.10. The degree of PFC hydrolyzation increased with the increasing of solution pH (Wang et al., 2011). At alkaline conditions, bridging and entrapment dominated the coagulation mechanism. Therefore, the formed flocs were less compact.
3.3.
Influence of pH on floc strength
After the slow stir phase, the preformed flocs were exposed to a series of increased shear force for another 15 min. The variation of d0.5 is shown in Fig. 5. The result showed that d0.5
900 800 700 600
Floc sizes d0.5 (µm)
500 400
Table 2 e Floc strength coefficient (C ) and stable floc exponent (g) for flocs formed at increasing pH conditions (Dosage, 0.150 mmol/L). pH
4.00
5.50
7.00
9.00
log C g R2
3.781 0.759 0.977
3.652 0.753 0.997
3.643 0.714 0.999
3.477 0.674 0.995
had a drop at the moment when the shear was applied, and then a gentle decrease followed. The drop degree was positively correlated with the applied shear force. The floc strength can be indicated by the rate at which a floc size decays under exposure to shear, and the relationship is shown as equation (1). The d0.5 of broken flocs after 15 min increased shear was plotted against the velocity gradient on a logelog scale (Fig. 6). The fitted floc strength coefficients (C ) and stable floc exponents (g) are shown in Table 2. The strength of flocs can be indicated by the value of log C, which is the intercept of the fitted line. Generally, for a given formation condition (shear rate), larger log C indicates stronger flocs (Bache, 2004). The value of g provides information how the floc size vary when exposed to increasing shear rate. A larger g means that the flocs are more prone to be broken into smaller sizes with increasing shear velocity gradients (Jarvis et al., 2005a). The log C values of the flocs decreased with the increase of coagulation pH, which meant that the flocs formed under acidic conditions were much stronger. However, the values of g decreased with the increase of coagulation pH. That is to say, the stronger flocs, formed at acidic conditions, were easily to be broken into smaller fragments under increasing shear force. Table 3 shows the average size and fractal dimension of broken flocs after exposure to weaker shear force (G ¼ 27.4 s1) and stronger shear force (G ¼ 101.7 s1). The d0.5 sizes of the flocs, which were exposed to 27.4 s1 shear, distributed widely in the range of 332e535 mm, compared to a narrow region of 132e163 mm at 101.7 s1 shear. When G ¼ 27.4 s1, the Df values increased slightly (except for flocs formed at pH 4.00). Increasing the shear force resulted in a further increase of Df values with the region of 2.33e2.58 at G ¼ 101.7 s1. Some researches have reported a similar increase of fractal dimension for ferric precipitate and latex floc (Jarvis et al., 2005b), and the change in compaction has been attributed to that the breakage of flocs occurred at the weak points and rearranged into more stable structures (Selomulya et al., 2001).
300
200
Table 3 e Mean diameters and fractal dimensions of broken flocs under velocity gradients of 27.4 sL1 and 101.7 sL1 (Dosage, 0.150 mmol/L).
pH =4.00 pH =5.50 pH =7.00
G ¼ 27.4 s1
pH =9.00 100 20
40
60
80
100
120
-1
Average velocity gradient (s )
Fig. 6 e Floc strength plots for flocs formed at different pH conditions (Dosage, 0.150 mmol/L; shear time, 15 min).
pH pH pH pH
¼ ¼ ¼ ¼
4.00 5.50 7.00 9.00
G ¼ 101.7 s1
d2 (mm)
Df
d2 (mm)
Df
534.7 366.4 409.5 332.0
1.91 2.55 1.98 2.17
163.2 130.8 157.1 131.6
2.33 2.58 2.42 2.37
6187
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 8 1 e6 1 8 8
900
Breakage -1 G=27.4 s
Re-growth -1 G=11.5 s
800
700
700
600
600
500 400 300
pH=4.00 pH=5.50 pH=7.00 pH=9.00
200 100
Floc sizes d0.5 (µm)
Floc sizes d0.5 (µm)
800
900
Growth -1 G=11.5 s
Growth -1 G=11.5 s
Breakage Re-growth -1 G=101.7 s G=11.5 s-1
pH=4.00 pH=5.50 pH=7.00 pH=9.00
500 400 300 200 100
0
0 0
10
20
30
40
50
0
Time (mins)
10
20
30
40
50
Time (mins)
Fig. 7 e Floc growth, breakage and re-growth profiles with different pH conditions (The average velocity gradients of shear phase are 27.4 s-1 and 101.7 s-1).
3.4.
Influence of pH on floc re-growth capability
A slow stirring with average velocity gradient of 11.5 s1 was reintroduced for another 15 min following the breakage phase to allow the broken flocs to reaggregate. The recovery of floc sizes was recorded by the Mastersizer 2000. Two typical profiles of floc breakage and re-growth are shown in Fig. 7. The results showed that, once the slow stirring was introduced, some reaggregation of broken flocs would occur. During the 15 min re-growth period, the floc size reached a new steady-state, but the d0.5 of recovered flocs could not reach the value as it occurred before the breakage. The recovery factors of flocs formed at different pH conditions were calculated by equations (3) and listed in Table 4. The Rf values showed that, flocs exposed to gentle shear force (G ¼ 15.9 s1) had poor re-growth capability, and flocs formed at acidic and neutral conditions even gave negative recovery factors. This maybe attributed to their large floc size (Fig. 2, Table 1). Generally, the reaggregation of flocs dominates the variation of floc size when the high shear is removed, then the influence of breakage increases until a new steady state is reached. The flocs formed at acidic and neutral conditions were larger than those formed at pH 9.00. The breakage was relatively heavy at gentle shear (G ¼ 15.9 s1) and dominated over the aggregation. Therefore, the floc size decreased continually, regardless of the reintroduction of slow stirring. A similar decrease in floc size in a long
Table 4 e Recovery factors of PFC flocs formed under different pH conditions (Dosage, 0.150 mmol/L).
G G G G G
1
¼ 15.9 s ¼ 27.4 s1 ¼ 40.0 s1 ¼ 69.1 s1 ¼ 101.7 s1
pH ¼ 4.00
pH ¼ 5.50
pH ¼ 7.00
pH ¼ 9.00
144.8 43.12 40.22 30.10 22.92
111.0 23.03 24.97 25.03 24.83
43.64 22.18 19.53 30.62 32.03
4.09 19.26 24.72 30.21 28.51
continuing slow stirring phase has been reported by Lin et al. (2008). Under other shear conditions, Rf of flocs formed at pH 4.00 decreased as the shear force increased. In other words, for pH 4.00 flocs, the increase of shear force resulted in an increase of Df and a decrease of Rf. It is similar with the current understanding that the compaction restricts the re-growth of broken flocs. At neutral and alkaline conditions, flocs gave larger Rf values when exposed to higher shear forces, and Rf of flocs formed at optimum pH conditions was less affected by shear force. When the shear force was 27.4 s1, flocs formed at pH 4.00 had the best re-growth capability with Rf value of 43.12%, and the Rf values decreased with the increase of pH. When the shear force increased to 101.7 s1, flocs formed at neutral condition had better re-growth capability than those formed at acidic and alkaline conditions.
4.
Conclusions
A series of experimental procedures were conducted to investigate the impact of pH on floc size, fractal dimension, strength and recovery capability in treating a low concentration of organic matter surface water using polyferric chloride coagulant. The optimum dosage was predetermined as 0.150 mmol/L. The optimum pH range was 5.50e5.75, where lower residual turbidity and the best organic removal were achieved. The results indicated that flocs formed at neutral pH condition gave the largest floc size and the highest growth rate. During the coagulation period, the fractal dimension of flocs aggregates increased in the first minutes and then decreased, and larger flocs generally had smaller fractal dimensions. The floc strength, which was assessed by the relationship of floc diameter and velocity gradient, was in the order: pH 4.00 > pH 5.50 > pH 7.00 > pH 9.00, however, flocs formed at low coagulation pH are easily to be broken into smaller
6188
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 8 1 e6 1 8 8
fragments under increasing shear force. The recovery factors were dependent on coagulation pH and the applied velocity gradient.
Acknowledgements This work was supported by the Sub-projects of the National Water Pollution Control and Government Key Special Project in the Eleventh Five-year Plan Period (No. 2008ZX07422-00302), the National Natural Science Foundation of China (No. 21077066), the Scientific Technology Research and Development Program of Shandong, China (No. 2010GZX20605), and the Natural Science Foundation of Shandong Province China (No. ZR2010BM014).
references
Bache, D.H., 2004. Floc rupture and turbulence: a framework for analysis. Chem. Eng. Sci. 59, 2521e2534. Biggs, C.A., Lant, P.A., 2000. Activated sludge flocculation: on-line determination of floc size and the effect of shear. Water Res. 34, 2542e2550. Boller, M., Blaser, S., 1998. Particles under stress. Water Sci. Technol. 37 (10), 9e29. Bridgeman, J., Jefferson, B., Parsons, S.A., 2008. Assessing floc strength using CFD to improve organics removal. Chem. Eng. Res. Des. 86, 941e950. Bridgeman, J., Jefferson, B., Parsons, S.A., 2009. Computational fluid dynamics modelling of flocculation in water treatment: a review. Eng. Appl. Comp. Fluid 3 (2), 220e241. Cao, B.C., Gao, B.Y., Xu, C.H., Fu, Y., Liu, X., 2010. Effects of pH on coagulation behavior and floc properties in Yellow River water treatment using ferric based coagulants. Chin. Sci. Bull. 55 (14), 1382e1387. Chakraborti, R.K., Atkinson, J.F., Van Benschoten, J.E., 2000. Characterization of alum floc by image analysis. Environ. Sci. Technol. 34 (18), 3969e3976. Edzwald, J.K., Tobiason, J.E., 1999. Enhanced coagulation: US requirements and a broader view. Water Sci. Technol. 40 (9), 63e70. Gauthier, E., Fortier, I., Courchesne, F., Pepin, P., Mortimer, J., 2000. Aluminum forms in drinking water and risk of Alzheimer’s disease. Environ. Res. 84, 234e246. Gregor, J.E., Nokes, C., Fenton, J.E., 1997. Optimising natural organic matter removal from low turbidity waters by controlled pH adjustment of aluminium coagulation. Water Res. 31 (12), 2949e2958. Gregory, J., 1998. The role of floc density in solid-liquid separation. Filtr. Separate. 35 (4), 367e371. Gregory, D., Carlson, K., 2003. Relationship of pH and floc formation kinetics to granular media filtration performance. Environ. Sci. Tchnol. 37, 1398e1403. Hu, C.Z., Liu, H.J., Qu, J.H., Wang, D.S., Ru, J., 2006. Coagulation behavior of aluminum salts in eutrophic water, significance of Al13 species and pH control. Environ. Sci. Technol. 40, 325e331. Jarvis, P., Jefferson, B., Gregory, J., Parsons, S.A., 2005a. A review of floc strength and breakage. Water Res. 39, 3121e3137. Jarvis, P., Jefferson, B., Parsons, S.A., 2005b. Breakage, regrowth, and fractal natural organic matter flocs. Environ. Sci. Tchnol. 39 (7), 2307e2314.
Jarvis, P., Jefferson, B., Parsons, S.A., 2006. Floc structural characteristic using conventional coagulation for a high doc, low alkalinity surface water source. Water Res. 40, 2727e2737. Jiang, Q., Logan, B.E., 1991. Fractal dimensions of aggregates determined from steady- state size distributions. Environ. Sci. Technol. 25 (12), 2031e2038. Lee, S.Y., Anthony, A., Fane, G., Waite, T.D., 2005. Impact of natural organic matter on floc size and structure effects in membrane filtration. Environ. Sci. Technol. 39 (17), 6477e6486. Li, T., Zhu, Z., Wang, D.S., Yao, C.H., Tang, H.X., 2006. Characterization of floc size, strength and structure under various coagulation mechanisms. Powder Technol. 168, 104e110. Lin, J.L., Huang, C., Chin, C.J.M., Pan, J.R., 2008. Coagulation dynamics of fractal flocs induced by enmeshment and electrostatic patch mechanisms. Water Res. 42, 4457e4466. Matilainen, A., Vepsa¨la¨inen, M., Sillanpa¨a¨, M., 2010. Natural organic matter removal by coagulation during drinking water treatment: a review. Adv. Colloid Interface. Sci. 159, 189e197. Parker, D.S., Kaufman, W.J., Jenkins, D., 1972. Floc breakup in turbulent flocculation processes. J. Sanit. Eng. Div.: Proc. Am. Soc. Civ. Eng. SA1, 79e99. Rieker, T.P., Hindermann-Bischoff, M., Ehrburger-Dolle, F., 2000. Small-angle X-ray scattering study of the morphology of carbon black mass fractal aggregates in polymeric composites. Langmuir 16, 5588e5592. Schaefer, D.W., 1989. Polymers, fractals, and ceramic materials. Science 243, 1023e1027. Selomulya, C., Amal, R., Bushell, G., Waite, T.D., 2001. Evidence of shear rate dependence on restructuring and breakup of latex aggregates. J. Colloid Interf. Sci. 236, 67e77. Waite, T.D., Cleaver, J.K., Beattie, J.K., 2001. Aggregation kinetics and fractal structure of g-alumina assemblages. J. Colloid Interf. Sci. 241, 333e339. Wang, Y., Gao, B.Y., Xu, X.M., Xu, W.Y., Xu, G.Y., 2009. Characterization of floc size, strength and structure in various aluminum coagulants treatment. J. Colloid Interf. Sci. 332, 354e359. Wang, Y.L., Feng, J., Dentel, S.K., Lu, J., Shi, B.Y., Wang, D.S., 2011. Effect of polyferric chloride (PFC) doses and pH on the fractal characteristics of PFC-HA flocs. Colloids Surf. A: Physicochem. Eng. Aspects 379, 51e61. Wei, J.C., Gao, B.Y., Yue, Q.Y., Wang, Y., Li, W.W., Zhu, X.B., 2009. Comparison of coagulation behavior and floc structure characteristic of different polyferric-cationic polymer dualcoagulants in humic acid solution. Water Res. 43, 724e732. Wu, R.M., Lee, D.J., Waite, T.D., Guan, J., 2002. Multilevel structure of sludge flocs. J. Colloid Interf. Sci. 252, 383e392. Xu, W.Y., Gao, B.Y., Yue, Q.Y., Wang, Y., 2010. Effect of shear force and solution pH on flocs breakage and re-growth formed by nano-Al13 polymer. Water Res. 44, 1893e1899. Yu, W.Z., Li, G.B., Xu, Y.P., Yang, X., 2009. Breakage and re-growth of flocs formed by alum and PACl. Powder Technol. 189, 439e443. Yu, W.Z., Gregory, J., Campos, L., 2010. Breakage and regrowth of Al-humic flocs - effect of additional coagulant dosage. Environ. Sci. Technol. 44 (16), 6371e6376. Yukselen, M.A., Gregory, J., 2004. The reversibility of floc breakage. Int. J. Miner. Process. 73 (2e4), 251e259. Zhan, X., Gao, B.Y., Yue, Q.Y., Wang, Y., Cao, B.C., 2010. Coagulation behavior of polyferric chloride for removing NOM from surface water with low concentration of organic matter and its effect on chlorine decay model. Sep Purif. Technol. 75, 61e68. Zhao, Y.X., Gao, B.Y., Shon, H.K., Cao, B.C., Kim, J.-H., 2011. Coagulation characteristics of titanium (Ti) salt coagulant compared with aluminum (Al) and iron (Fe) salts. J. Hazard. Mater. 185, 1536e1542.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 8 9 e6 1 9 4
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Sulfate radical-advanced oxidation process (SR-AOP) for simultaneous removal of refractory organic contaminants and ammonia in landfill leachate Yang Deng*, Casey M. Ezyske Department of Earth and Environmental Studies, Montclair State University, Mallory Hall 252, 1 Normal Ave, Montclair, NJ 07043, USA
article info
abstract
Article history:
Typically, a mature landfill leachate contains high levels of non-biodegradable organics
Received 30 April 2011
and ammonia nitrogen. Simultaneous removal of the both persistent leachate pollutants is
Received in revised form
a significant challenge. This paper reports the first scientific study to apply a sulfate radical
3 September 2011
(SO$ 4 ) e based advanced oxidation process (SR-AOP) to treat a mature leachate, with an
Accepted 7 September 2011
emphasis of concurrent removal of refractory organics and ammonia. In this study, all the
Available online 21 September 2011
experiments were run in a batch reactor with temperature control. In the thermal persulfate oxidation (TPO) process, persulfate (S2O28 ) was activated by heat to produce
Keywords:
o powerful oxidants, SO$ 4 (E ¼ 2.6 V). Three factors affecting the removal efficiencies of
Sulfate radical
chemical oxygen demand (COD) and ammonia nitrogen were investigated, including initial
Persulfate
solution pH (3e8.3), temperature (27e50 C), and chemical dose (S2 O2 8 :12COD0 ¼ 0.25e2.0).
Heat
Typically, acidic pH (3e4), higher temperature, and higher dose favored the removal of COD
Landfill leachate
and ammonia. At S2 O2 8 :12COD0 ¼ 2 and 50 C, the COD removal rates were 79% and 91% at
Refractory organics
pH 8.3 (no pH adjustment) and 4, respectively; and the ammonia nitrogen removal reached
Ammonia
100% at pH 8.3 or 4. SR-AOP appears to be more advantageous over hydroxyl radical (OH∙)based advanced oxidation processes (HR-AOPs) because OH∙ almost does not oxidize ammonia. Furthermore, compared with Fenton treatment of the same batch leachate sample, the TPO could achieve a higher COD removal at an identical chemical dose. For example, COD removal was 40% at H2O2:2.125COD0 ¼ 2 during Fenton treatment (pH 3), but 91% at S2 O2 8 :12COD0 ¼ 2 during TPO (pH 4). These findings demonstrate that SR-AOP is a promising landfill leachate treatment method. Published by Elsevier Ltd.
1.
Introduction
A primary environmental concern of municipal solid waste (MSW) landfills is incessant generation of landfill leachate. Landfill leachate is a high strength wastewater with a variety of organic wastes and inorganic species, exhibiting acute and chronic toxicity. Refractory organics and ammonia are two major contaminants in leachate (Kjeldsen et al., 2002). As
MSW decomposes, leachate is categorized with young leachate (typically 2 yrs) and mature leachate (typically 5 yrs) (Deng, 2007). Although organic content dramatically declines with time, the major organic fraction in a mature leachate gradually becomes refractory. Ammonia in leachate (typically 500e2000 mg/L NH3eN) is also problematic. Without a decreasing trend over time, ammonia is of the most concern in leachate in a long run (Kjeldsen et al., 2002). In practice,
* Corresponding author. Tel.: þ1 973 655 6678; fax: þ1 973 655 4072. E-mail addresses: [email protected], [email protected] (Y. Deng). 0043-1354/$ e see front matter Published by Elsevier Ltd. doi:10.1016/j.watres.2011.09.015
6190
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 8 9 e6 1 9 4
discharge to publicly owned treatment works (POTW) is a common leachate management option in the US. Apart from occasional disturbance and termination of POTW operation (Berry and Lin, 1997; Englehardt et al., 2006), the co-treatment cannot provide an ultimate solution to persistent leachate pollutants (e.g. refractory organics) except dilution of leachate with much more sewage. Furthermore, this option is not feasible in many cases due to the lack of sewer lines or capacity for additional loading at nearby sewage treatment plants, and the unwillingness of communities to permit connection for other reasons (Spengel and Dzombak, 1991). Over the past decades, landfill leachate has remained one of the leading groundwater and soil contamination sources in the world. For example, New Jersey (NJ) has 146 Superfund sites, ranking No. 1 among the 50 states and Washington D.C. in the US. Forty-five of them were contaminated by landfill leachate, accounting for 31% of total NJ Superfund sites (Ezyske, 2010). Therefore, an appropriate treatment is vital to control leachate pollution. Among various leachate treatment methods, chemical oxidation is favorable due to chemical destruction of, rather than phase transfer of, toxic and persistent leachate chemicals. The leachate oxidation technologies that have been widely studied include electrochemical oxidation (Deng and Englehardt, 2007), and hydroxyl radical-based advanced oxidation processes (HR-AOPs) such as Fenton oxidation (Deng and Englehardt, 2006; Mohajeri et al., 2010; Zhang et al., 2005, 2006), ultrasonic irradiation (Gonze et al., 2003), ultraviolet (UV) irradiation/H2O2 (Shu et al., 2006), O3/H2O2 (Wang et al., 2004), and ozonation (Wu et al., 2004). However, they only remove one of the two primary pollutants (e.g. HR-AOPs), are limited by unavailability of large-scale devices (e.g. ultrasonic irradiation), or have too high operational costs (e.g. electrochemical oxidation and ozonation). Thus, development of a technically and economically feasible leachate treatment method is strategically important. Persulfate (S2 O2 8 ) is the newest oxidant used in in-situ chemical oxidation (ISCO) for groundwater and soil cleanup (Huling and Pivetz, 2006). S2 O2 8 itself is a strong oxidant with a standard oxidation potential (Eo) of 2.01 V (Eq. (1)), comparable to O3 (2.07 V) (Kolthoff and Stenger, 1947). 2 S2 O2 8 þ 2e/2SO4
(1)
Furthermore, once activated by heat (Eq. (2)), metal, elevated pH, or UV irradiation, S2 O2 8 can form more powerful o , E ¼ 2.6 V) to initiate sulfate radicalsulfate radicals (SO$ 4 based advanced oxidation processes (SR-AOPs) (House, 1962; Kolthoff and Miller, 1951). The thermal persulfate oxidation (TPO) using heat activation is as follows. D
$ S2 O2 8 / 2SO4
ðEa ¼ 33:5 kcal=molÞ
2.
Experimental
Landfill leachate was collected from the Cabo Rojo Landfill (Cabo Rojo, Puerto Rico, US). The collected leachate was stored in a zero headspace plastic bottle in refrigerator at 4 C until use. The average composition of the landfill leachate was as follows: pH 8.3; alkalinity, 4885 mg/L CaCO3; initial COD (COD0), 1254 mg/L; ammonia nitrogen, 2000 mg/L; Cl-, 2 1270 mg/L; NO 3 , 0 mg/L; and SO4 , 0 mg/L. All chemicals were at least analytical grade, except as noted. If needed, the initial solution pH was adjusted with concentrated nitric acid (HNO3, 70%, FishChemical, Fair Lawn, NJ, US). In a typical run, 20 mL leachate was dispensed to a 60 mL serum vial installed in a water bath shaker (Precision, reciprocal shaking bath model 50) at a controlled temperature of 27, 40, or 50 C. Oxidation was initiated by addition of certain amount of sodium persulfate (Na2S2O8, Fisher Scientific UK, Bishop Meadow Road, Loughborough). Once the reactions were started, the shaker was turned on at 150 rpm to thoroughly stir the solution. Persulfate was qualitatively measured as follows once every day to determine whether all the persulfate was decomposed. One half milliliter of uniform sample was collected and mixed with 20 mL NaI and starch solution. If the solution became blue, persulfate still remained; otherwise, persulfate was completely decomposed. Once persulfate was depleted, the residual samples in the reactors will be collected for COD and ammonia analysis. COD was measured colorimetrically (20e1500 mg/L, HACH, Loveland, CO, USA). And ammonia was quantified using the colorimetric method according to the Standard Methods (APHA, AWWA, and WPCF, 1992). All experiments were run in duplicate. The error bars in the figures represent one standard deviation.
(2)
Subsequently, SO$ 4 may initiate production of other intermediate highly reactive oxygen species (ROS) such as hydroxyl radicals (OH$) (Eq. (3)) (Huie et al., 1991). SO$ 4 þ H2 O/OH$þHSO4
a broad spectrum of organic pollutants, such as trichloroethylene (TCE), 1,1,1-trichloroethane (TCE) (Liang et al., 2003, 2004), trichloroethane (TCA) (Liang et al., 2003), MTBE (Huang et al., 2002), and diphenylamine (Li et al., 2009). Moreover, a low cost of persulfate salt (Na2S2O8 < $2.65/kg, Huling and Pivetz, 2006) facilitates SR-AOP for potential application in wastewater treatment. With a strong oxidative capacity and a low cost, SR-AOP may be a promising leachate treatment method. To our knowledge, however, there has never been any report regarding persulfate treatment of landfill leachate. The objective of this study is to evaluate performance of thermal persulfate oxidation, as an alternative for the POTW treatment, for landfill leachate treatment with an emphasis of simultaneous removal of refractory organics and ammonia.
(3)
These ROS can initiate a series of radical propagation and termination chain reactions where organics are partially and even fully decomposed (Berlin, 1986). Recent studies have demonstrated the ability of surface radicals to rapidly degrade
3.
Results
3.1.
Effect of initial pH
The effects of initial pH on the removal of COD and ammonia by persulfate oxidation at room temperature (27 C) are shown in Figs. 1 and 2, respectively. Since the equivalent weight ratio 2 of S2 O2 8 to O2 is 12, the mass ratio of S2 O8 to the product of 12 2 and initial COD (S2 O8 :12COD0) was used to indicate the S2 O2 8 dose in this study. For example, at S2 O2 8 :12COD0 ¼ 1, the
6191
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 8 9 e6 1 9 4
50%
a COD Removal
COD Removal
45% 40% 35% 30% 25% 20%
100% 90% 80% 70% 60% 50% 40%
27 degree 40 degree 50 degree
30% 20%
10%
15%
0%
10% 2
3
4
5
6
7
8
0
9
0.5
Fig. 1 e Eeffect of initial pH on COD removal (Conditions: S2 O2L 8 :12COD0 [ 0.5; and temperature [ 27 C).
b
2
2.5
2
2.5
100%
COD Removal
80% 70% 60% 50% 40% 30%
27 degree 40 degree 50 degree
20%
10% 0% 0
0.5
1
1.5
S2O82-:12COD0 Fig. 3 e COD removal vs. persulfate dose at different temepratures and initial pH: (a) pH 8.3; (b) pH 4.
Effects of persulfate dose and temperture
Effects of persulfate dose (S2 O2 8 :12COD0 ¼ 0.1e2) and temperature (27e50 C) on COD removal at pH 8.3 and 4 are shown in Fig. 3(a) and (b), respectively. In general, high persulfate and high temperature favored COD removal. At pH 8.3, COD removal efficiencies of 5%, 9%, and 11% were achieved at a persulfate dose of 0.1 at 27, 40, and 50 C, respectively (Fig. 3 (a)). The COD removal efficiencies at 27, 40, and 50 C were significantly increased to 25%, 34%, and 55%, respectively, with the increase of persulfate dose from 0.1 to 0.5, and continued to go up to 49%, 65%, and 79%, respectively, when the persulfate was further raised to 2. Of note, the temperature positively exhibited an enhancement effect in COD
20% 18%
Ammonia Nitrogen Removal
1.5
90%
added S2 O2 8 theoretically just oxidizes all the COD, though an actual COD removal is most likely below 100% due to the consumption of persulfate by competing species. As shown in Fig. 1, the maximum COD removal of 37% was achieved at pH 3 and 4 at S2O2 8 :12COD0 ¼ 0.5. As the pH increased from 4 to 8.3 (original), the COD removal almost linearly decreased to 25%. Concurrently, the ammonia nitrogen removal at different pH exhibited a similar pattern (Fig. 2). The ammonia nitrogen removal peaked at 18% at pH 3, and almost linearly dropped down to 10% with the increasing pH to 8.3.
3.2.
1
S2O82-:12COD0
pH
16% 14% 12% 10% 8% 6% 4% 2% 0% 2
3
4
5
6
7
pH Fig. 2 e Eeffect of initial pH on ammonia removal (Conditions: S2 O2L 8 :12COD0 [ 0.5; and temperature [ 27 C).
8
9
reduction within the entire range of 27e50 C, especially at the 2 S2 O2 8 :12COD0 above 0.25. For example, at S2 O8 :12COD0 ¼ 2, the COD removal was increased by 32% and 21% from 27 to 40 C and from 40 to 50 C, respectively. At pH 4, COD removal efficiencies at 27, 40, and 50 C were 8%, 39%, and 39% at the S2 O2 8 :12COD0 ¼ 0.1, respectively, as shown in Fig. 3(b). With the increasing persulfate dose from 0.1 to 0.5, these COD reduction rates went up to 37%, 62%, and 67%, respectively. When the persulfate dose reached 2, COD removal efficiencies at 27, 40, and 50 C were correspondingly increased to 62%, 83%, and 91%. Interestingly, the enhancement effect of temperature was significant at 27e40 C, but almost marginal at 40e50 C, especially at a persulfate dose of 0.1e0.25. For example, at any particular persulfate dose (0.1e2), the augment in the COD reduction ranged within 21e42% when the temperature went up from 27 to 40 C, but the COD reduction was increased by 0e8% with the increasing temperature from 40 to 50 C. Furthermore, any COD removal efficiency at pH 4 (Fig. 3(b)) was greater than the corresponding COD removal at pH 8.3 (Fig. 3(a)) under identical experimental conditions, again demonstrating that acidic condition favored persuflate oxidation of leachate organics. Effects of persulfate dose (S2 O2 8 :12COD0 ¼ 0.1e2) and temperature (27e50 C) on ammonia nitrogen reduction at pH 8.3 and 4 are shown in Fig. 4(a) and (b), respectively. Similarly, high persulfate dose and high temperature enhanced removal of ammonia nitrogen. At any particular temperature, the ammonia nitrogen removal appeared to be linearly increased with the increasing persulfate dose. At pH 8.3, when the
6192
Ammonia Nitrogen Removal
a
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 8 9 e6 1 9 4
100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0%
27 degree 40 degree 50 degree 0
0.5
1
1.5
2
2.5
S2O82-:12COD0
b
100%
Ammonia Nitrogen Removal
90% 80% 70% 60% 50% 40% 30%
27 degree 40 degree 50 degree
20%
10% 0%
0
0.5
1
1.5
2
2.5
S2O82-:12COD0 Fig. 4 e Ammonia nitrogen removal vs. persulfate dose at different temepratures and initial pH: (a) pH 8.3; (b) pH 4.
persulfate dose was added from 0.1 to 2, the ammonia nitrogen removal went up from 0%, 15%, and 23%, to 86%, 91%, and 100% at 27, 40, and 50 C, correspondingly (Fig. 4 (a)). The ammonia nitrogen removal was increased from 0 to 92%, from 20 to 96%, and from 25 to 100% at 27, 40, and 50 C, respectively, at pH 4. Of note, any ammonia nitrogen removal at pH 4 was higher than the corresponding ammonia removal at pH 8.3 under the identical operational conditions.
4.
Discussion
To our knowledge, this study represented the first scientific effort to apply persulfate for leachate treatment. Our results demonstrated that SR-based AOP could adequately and simultaneously remove refractory organics and ammonia from landfill leachate. For the COD removal, TPO showed a much higher removal capacity than persulfate oxidation (at 27 C), as shown in Fig. 3(a) and (b), suggesting that sulfate radicals (the dominant TPO oxidizing species) more readily degraded the leachate organic molecules than persulfate (the principal oxidant in persulfate oxidation). However, the detailed mechanisms of sulfate radical attack of leachate organic matters are not well understood. Dissolved organic matters in leachate are a mixture of molecules with very complex chemical structures and uncertain chemical formula, of which the major fraction in a mature leachate exhibit many characteristics similar to humic substances (Deng, 2009). Consequently, the pathways of organic
degradation are difficultly determined. For the ammonia removal, sulfate radicals produced during TPO were able to effectively remove ammonia, like COD. Sulfate radicals, as a strong oxidizing agent, most likely gain electrons from the reductive ammonia nitrogen (NH3 and NHþ 4 ), so that the nitrogen of ammonia is oxidized to a higher valence state (e.g. N2). Furthermore, persulfate significantly removed ammonia only at a high dose (S2 O2 8 :12COD0 > 0.5), showing that persulfate might be a potential oxidant for ammonia under certain conditions. Generally, the solution pH affects persulfate oxidation in to produce two aspects. First, elevated pH activates S2 O2 8 enhancing the oxidation capacity. Second, pH more SO$ 4 decrease can reduce the alkalinity level, thus lessening the ROS scavenging degree of CO2 3 and HCO3 and increasing the oxidation efficiency. Although the two factors could simultaneously occur, the second factor appeared to prevail in this study, as demonstrated in Figs. 1 and 2. Once pH was below 4.5, alkalinity was completely eliminated. Therefore, the maximum COD removal and ammonia nitrogen reduction were observed at pH 3e4 (Figs. 1 and 2). Temperature is another critical factor to control the oxidation. Typically, an augment in temperature boosted the oxidation efficiencies of COD and ammonia, because $ heat-activated S2 O2 8 produced SO4 , a stronger oxidant (Eq. (2)). Our results highlighted that an appropriate temperature might be observed for oxidation of COD or ammonia at a specific pH, above which the increase of the removal efficiency was almost marginal. As shown in Fig. 3 (b), once the temperature was beyond 40 C at pH 3, the improvement in the COD removal was slight. On the other hand, the impact of temperature in decay rate of persulfate needs to be considered. Although the detailed study in the persulfate decay kinetics was not in the scope of this study, the persulfate decay seemed to be greatly controlled by temperature. For example, the half life of persulfate was less than 1 day at 50 C, but in the order of days at 27 C. The difference will significantly influence the reactor design (e.g. sizing reactor) in practice. To maintain a high treatment efficiency, heat was required during TPO. In fact, it is ubiquitous that fresh landfill leachate has a high temperature due to the heat release from anaerobic digestion of solid wastes within landfills. Another potential energy source is biogas (e.g. CH4) produced during landfilling, which may be utilized to sustainably support TPO for leachate treatment. Of interest, persulfate oxidation shares a few common characteristics with hydroxyl radical e based classical Fenton oxidation, which has been intensively studied over the past decade, with regards toward leachate treatment. First, the both show a high removal in COD. Our previous study (Deng et al., in press) of Fenton treatment on the same batch of leachate sample used in this study exhibited a similar COD removal efficiency, suggesting that the major refractory leachate organic molecules (mostly humic-like substances) are amenable to both OH∙ in Fenton treatment and SO$ 4 in TPO. Second, the most favorable pH (3e4) observed in this study for the TPO treatment of leachate falls within the typical optimal pH range in Fenton treatment of landfill leachate (Deng and Englehardt, 2006). A major reason may be that an
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 8 9 e6 1 9 4
acidic condition eliminates alkalinity and thus removes the $ scavenging of CO2 3 and HCO3 for OH∙ and SO4 . Third, persulfate oxidation and Fenton treatment both lead to an increase of total dissolved solids (TDS) in treated leachate (data not shown here), due to the addition of high concentrations of salts. However, unlike Fenton oxidation, persulfate oxidation exhibits several distinctive features in leachate treatment. First, ammonia is oxidized accompanied with COD reduction during TPO or persulfate oxidation, suggesting that ammonia or S2 O2 is susceptible to oxidation due to SO$ 4 8 . On the contrary, in this study, Fenton oxidation was also used for treatment of the same batch of leachate, but the ammonia removal was nearly zero (data not shown here). A few studies have reported that OH∙ has a very low rate constant with ammonia, so that Fenton oxidation is almost ineffective for ammonia removal (Deng, 2009; Zhang et al., 2005). Second, persulfate oxidation does not produce undesirable sludge that is inevitably formed and needs to be properly disposed of during Fenton treatment. Third, oxidation during TPO is the only mechanism for COD reduction, while oxidation and coagulation both contributed to organics removal in Fenton treatment (Deng, 2007). Fourth, to achieve a high oxidation efficiency, TPO requires heat to activate persulfate to produce stronger sulfate radicals. However, Fenton oxidation can effectively proceed at room temperature. Finally, Fenton oxidation of leachate is typically completed within a few hours, but persulfate oxidation usually needs a much longer time.
5.
Conclusion
This study demonstrates that sulfate radical-based AOP is an effective chemical oxidation process in simultaneous reduction of COD and ammonia from a mature landfill leachate. Solution pH, temperature, and chemical dose all influence the oxidation efficiency. Under certain conditions, >90% of COD and ammonia could be effectively removed in this study. The findings allow SR-AOP to be advantageous over most HR-AOPs because the latter utilize OH∙ that cannot oxidize ammonia. Moreover, persulfate is a low-cost chemical. Therefore, SRAOP may be an alternative for HR-AOPs in leachate treatment. This paper reports the first study in application of SRAOP for leachate treatment. Other persulfate inactivation methods such as the use of metal ions, as well as characterization of organic molecules subsequent to sulfate radical oxidation of leachate organic matters, will be further studied in the near future.
Acknowledgments This study was partially supported from the new faculty startup fund from College of Science and Mathematics (CSAM) at Montclair State University. Special thanks to Mr. Edualberto Rosario-Muniz, and Ms. Margarita Otero Diaz for their help in experiments. We are grateful to City of Cabo Rojo (Puerto Rico) for collection of the landfill leachate sample.
6193
references
APHA, AWWA, WPCF, 1992. Standard Methods for the Examination of Water and Wastewater, eighth ed. American Public Health Association, American Water Works Association, Water Pollution Control Federation, Washington DC, USA. Berlin, A.A., 1986. Kinetics of radical-chain decomposition of persulfate in aqueous solutions of organic compounds. Kinetics and Catalysis 27, 34e39. Berry, R.C., and Lin, K.C., 1997. Experimental studies on the cotreatment of landfill leachate and sewage in Fredericton, N.B., Canada. Proceedings of International Conference on Water Pollution, Lake Bled, Slovenia. Deng, Y., 2007. Physical and oxidative removal of organics during Fenton treatment of mature municipal landfill leachate. Journal of Hazardous Materials 146 (1e2), 334e340. Deng, Y., Rosario Muniz, E., and Ma, X., in press. "Effects of inorganic anions on Fenton oxidation of landfill leachate." Waste Management and Research. doi: 10.1177/ 0734242X10378185. Deng, Y., Englehardt, J.D., 2006. Treatment of landfill leachate by the Fenton process. Water Research 40 (20), 3683e3694. Deng, Y., Englehardt, J.D., 2007. Electrochemical oxidation for landfill leachate treatment. Waste Management 27 (3), 380e388. Deng, Y., 2009. Advanced Oxidation Processes (AOPs) for reduction of organic pollutants in landfill leachate: a review. International Journal of Environment and Waste Management 4 (3/4), 366e384. Englehardt, J.D., Deng, Y., Meeroff, D., Legrenzi, Y., Mognol, J., Polar, J., 2006. Options for Managing Municipal Landfill Leachate: Year 1 Development of Iron-Mediated Treatment Processes. Florida Center for Solid and Hazardous Waste Management, Gainesville, FL, USA. Ezyske, C.M., 2010. Landfills in New Jersey. Project Report. Montclair State University, NJ, USA. Gonze, E., Commenges, N., Gonthier, Y., Bernis, A., 2003. High frequency ultrasound as a pre- or a post-oxidation for paper mill wastewaters and landfill leachate treatment. Chemical Engineering Journal 92 (1e3), 215e225. House, D.A., 1962. Kinetics and mechanism of oxidations by peroxydisulfate. Chemical Reviews 62, 185e203. Huang, K.C., Couttenye, R.A., Hoag, G.E., 2002. Kinetics of heatassisted persulfate oxidation of methyl tert-butyl ether (MTBE). Chemosphere 49 (4), 413e420. Huie, R.E., Clifton, C.L., Neta, P., 1991. Electron transfer reaction rates and equilibria of the carbonate and sulfate radical anions. Radiation Physics and Chemistry 38, 477e481. Huling, S.G., Pivetz, B.E., 2006. In-situ Chemical Oxidation (EPA/ 600/R-06/072). US EPA. Kjeldsen, P., Barlaz, M.A., Rooker, A.P., Baun, A., Ledin, A., Christensen, T.H., 2002. Present and long-term composition of MSW landfill leachate: a review. Critical Reviews in Environmental Science and Technology 32 (4), 297e336. Kolthoff, I.M., Miller, J.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, 3055e3059. Kolthoff, I.M., Stenger, V.A., 1947. Volumetric Analysis, Second Revised Ed Titration Methods: Acid-base, Precipitation, and Complex Reactions, vol. II. Interscience Publishers, Inc., New York. Li, S.X., Wei, D., Mak, N.K., Cai, Z., Xu, X.R., Li, H.B., Jiang, Y., 2009. Degradation of diphenylamine by persulfate: performance optimization, kinetics and mechanism. Journal of Hazardous Materials 164 (1), 26e31.
6194
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 8 9 e6 1 9 4
Liang, C., Bruell, C.J., Marley, M.C., Sperry, K.L., 2003. Thermally activated persulfate oxidation of trichloroethylene (TCE) and 1,1,1-trichloroethane (TCA) in aqueous systems and soil slurries. Soil and Sediment Contamination 12 (2), 207e228. Liang, C., 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 per sulfate-thiosulfate redox couple. Chemosphere 55 (9), 1213e1223. Mohajeri, S., Aziz, H.A., Isa, M.H., Bashir, M.J.K., Mohajeri, L., Adlan, M.N., 2010. Influence of Fenton reagent oxidation on mineralization and decolorization of municipal landfill leachate. Journal of Environmental Science and Health, Part A: Toxic/Hazardous Substances and Environmental Engineering 45 (6), 692e698. Shu, H.-Y., Fan, H.-J., Chang, M.-C., Hsieh, W.-P., 2006. Treatment of MSW landfill leachate by a thin gap annular UV/H2O2 photoreactor with multi-UV lamps. Journal of Hazardous Materials 129 (1e3), 73e79.
Spengel, D.B., Dzombak, D.A., 1991. Treatment of landfill leachate with rotating biological contactors - bench-scale experiments. Research Journal of the Water Pollution Control Federation 63 (7), 971e981. Wang, F., Gamal El-Din, M., Smith, D.W., 2004. Oxidation of aged raw landfill leachate with O3 only and O3/H2O2: treatment efficiency and molecular size distribution analysis. Ozone: Science and Engineering 26 (3), 287e298. Wu, J.J., Wu, C.-C., Ma, H.-W., Chang, C.-C., 2004. Treatment of landfill leachate by ozone-based advanced oxidation processes. Chemosphere 4 (7), 997e1003. Zhang, H., Choi, H.J., Huang, C.P., 2005. Landfill leachate treatment by Fenton’s reagent. The variation of leachate characteristics. Fresenius Environmental Bulletin 14 (12b), 1178e1183. Zhang, H., Choi, H.J., Huang, C.P., 2006. Effects of important reaction conditions on landfill leachate treatment by Fenton process. Fresenius Environmental Bulletin 15 (1), 43e47.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 9 5 e6 2 0 6
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Size and structure evolution of kaolineAl(OH)3 flocs in the electroflocculation process: A study using static light scattering T. Harif*, A. Adin Soil &Water Sciences Department, Robert H. Smith Faculty of Agricultural, Food and Environmental Quality Sciences, The Hebrew University of Jerusalem, POB 12, Rehovot 71600, Israel
article info
abstract
Article history:
Electroflocculation (EF) is gaining recognition as an alternative process to conventional
Received 14 June 2011
coagulation/flocculation. The electrical current applied in EF that generates the active
Received in revised form
coagulant species creates a unique chemical/physical environment in which competing
9 September 2011
redox reactions occur, primarily water electrolysis. This causes a transient rise in pH, due
Accepted 10 September 2011
to cathodic formation of hydroxyl ions, which, in turn, causes a continuous shift in
Available online 17 September 2011
coagulation/flocculation mechanisms throughout the process. This highly impacts the formation of a sweep floc regime that relies on precipitation of metal hydroxide and its
Keywords:
growth into floc. The size and structural evolution of kaolineAl(OH)3 flocs was examined
Electroflocculation
using static light scattering techniques, in aim of elucidating kinetic aspects of the process.
Coagulation
An EF cell was operated in batch mode and comprised of two concentric electrodes e
Aluminum
a stainless steel cathode (inner electrode) and an aluminum anode (outer electrode). The
Floc
cell was run at constant current between 0.042 A and 0.22 A, and analyses performed at
Structure
pre-determined time intervals. The results demonstrate that EF is able to generate a range
Scattering exponent
of flocs, exhibiting different growth rates and structural characteristics, depending on the conditions of operation. Growth patterns were sigmoidal and a linear correlation between growth rate and current applied was observed. The dependency of growth rate on current can be related to initial pH and aluminum dosing, with a stronger dependency apparent for initial optimal sweep floc regime. All flocs exhibited a fragile nature and undergo compaction and structural fluctuations during growth. This is the first time size and structural evolution of flocs formed in the EF process is reported. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
Electroflocculation (EF, also termed “electrocoagulation”) can be considered an alternative to conventional coagulation/ flocculation processes, although substantial differences between the two exist. In EF active coagulant species are generated in situ by electrolytic oxidation of an appropriate
anode material, thus differing from the conventional process in which chemical coagulants such as metal salts or polymers and polyelectrolytes are used. The use of electrolysis creates a unique coagulation and flocculation environment as the dosing regime is additive over time, negative counter-ions are not introduced, and competing redox reactions occur simultaneously, primarily water electrolysis. With regard to
* Corresponding author. Tel.: þ972 54 6490227. E-mail addresses: [email protected], [email protected] (T. Harif), [email protected] (A. Adin). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.09.027
6196
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 9 5 e6 2 0 6
coagulation/flocculation mechanisms, the applied current in fact has a two-fold affect as it both governs the dissolution of metal coagulant into the system, and also causes a transient rise in pH. Cathodic hydroxyl ion (OH) formation has been used to explain this phenomenon, although there is still some uncertainty regarding the cause and alternative mechanisms have been offered (Chen et al., 2000; Lakshmanan et al., 2009). Coagulation mechanisms depend primarily on factors such as pH and coagulant dosage, which govern speciation of the active mononuclear species (Amirtharajah and O’Melia, 1990; Letterman et al., 1999). In EF the coagulation mechanisms shift throughout the process, depending on the speciation at each given time, until the current is stopped, enabling the pH to stabilize and the species to reach equilibrium. These conditions have a profound impact on the generation of a stable sweep floc regime which is of utmost importance in particle removal in water treatment and relies on precipitation of metal hydroxide precursors and their growth into floc. Research into EF, although not extensive, has primarily examined colloidal and organic matter removal (Vik et al., 1984; Matteson et al., 1995; Holt et al., 2002; Yildiz et al., 2008; Sun et al., 2009) and explored a wide range of applications such as urban and industrial wastewater treatment (Do and Chen, 1994; Pouet and Grasmick, 1995; Belongia et al., 1999; Mollah et al., 2001, 2004; Adin and Vescan, 2002; Emamjomeh and Sivakumar, 2009). These studies have focused on the applicability of the EF process, however, few have looked into the mechanisms of coagulation and flocculation in EF and subsequent floc growth, that are directly affected by the unique conditions the process generates (Harif and Adin, 2007). This study utilizes static light scattering techniques to measure the size and structural evolution of kaolinealuminum hydroxide (kaolineAl(OH)3) flocs formed in EF, in aim of understanding floc formation and evolution patterns that can shed light on governing coagulation and flocculation mechanisms typical of EF.
1.1.
EF technology and coagulants
In its simplest form, an EF reactor may be made up of an electrolytic cell containing one anode and one cathode. The anode metals most commonly used are aluminum or iron because when electrochemically oxidized they produce the most commonly used ionic coagulants, Alþ3 and Feþ3 (or Feþ2) respectively. The redox reactions that occur with aluminum anodes are described by: ð Þ Cathode : 2H2 O þ 2e /H2ðgÞ þ 2OH ðaqÞ
(1)
ð þ Þ Anode : 2H2 O/O2ðgÞ þ 4Hþ ðaqÞ þ 4e
(2)
þ3
ð þ Þ Anode : AlðsÞ /AlðaqÞ þ 3e
(3) þ3
Following anodic dissolution of aluminum ion (Al ), hydrolysis of aluminum takes place spontaneously, leading to the formation of an hydration shell. The initial hydration shell is polarized because of the high charge on the aluminum ion, leading to water molecule replacement by OH ions (Duan and
Gregory, 2003). The pH in solution dictates to what degree the water molecules are replaced, depending on OH availability, thus various aluminum hydroxide mononuclear species can form (Eqs. (4)e(7)): þ3
þ2
AlðaqÞ þ H2 O/AlOHðaqÞ þ Hþ ðaqÞ
(4)
þ2
þ AlOHðaqÞ þ H2 O/AlðOHÞþ 2 ðaqÞ þ HðaqÞ
(5)
þ AlðOHÞþ 2 ðaqÞ þ H2 O/AlðOHÞ3ðsÞ þHðaqÞ
(6)
þ AlðOHÞ3ðsÞ þH2 O/AlðOHÞ 4 ðaqÞ þ HðaqÞ
(7)
At a sufficient concentration of OH , Al(OH)3 will form, which has zero charge and is capable of condensing and forming the solid phase. The initial products of Al(OH)3 precipitation (hereafter termed “Al(OH)3 precursors”) are amorphous, possessing a gelatinous texture, with no definite ordered structure, however with ageing will evolve into crystalline structures. In the context of coagulation, the amorphous precipitate is relevant. The amount of Al(OH)3 precursors formed depends on the solubility boundary that denotes the thermodynamic equilibrium between the solvated mononuclear aluminum species and solid Al(OH)3 at a given pH. The minimum solubility, 0.03 mg/l Alþ3, occurs at pH 6.3, with solubility increasing as the solution becomes more acidic or alkaline. Although dimeric, trimeric and polynuclear hydrolysis products of Alþ3 can form, these can often be ignored, especially in dilute solutions, and may not affect the overall speciation (Duan and Gregory, 2003). As such, this paper will assume that mononuclear hydrolyzed species adequately predict Al(OH)3 precipitation.
1.2.
Flocculation, floc size and structure
The classical expression by von Smoluchowski (1917) has formed the core of almost all subsequent research into flocculation modeling. The model he developed shows the rate of irreversible aggregation of flocs containing i- and jnumber of particles, to form aggregates with m-particles, where m ¼ i þ j: dnm =dt ¼ 1=2
i¼m1 X i¼1;j¼mi
aij bij ni nj nm
i¼N X
aim bim ni
(8)
i¼1
nij is the number concentration, and aij and bij, are the collision efficiency and frequency, respectively. The basic simplifying assumptions for this equation (such as every collision is successful (aij ¼ 1), monodispersity, and both particle and floc are spherical) rendering its relevance to practical systems limited. As such, modifications have been added while taking into account the presence of short-ranged forces, aij < 1, polydispersity, and floc structure (Jiang and Logan, 1991; Kusters et al., 1997). Several phases of floc growth occur during flocculation. Initially, particle growth is dominant, in which particles combine by coagulation and their size increases rapidly. As flocculation continues, the flocs form large, porous, and open structures that are more susceptible to fragmentation by fluid shear (Tambo, 1991). After a characteristic time,
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 9 5 e6 2 0 6
a steady state is reached between aggregation and fragmentation, characterized by an aggregate size distribution that does not change with time and is unique to each system (Spicer and Pratsinis, 1996). Aggregate structure has been found to impact the kinetics of the flocculation process, because more open structures possess larger collision profiles (Kusters et al., 1997). Additionally, porous aggregates are fragmented by fluid shear stresses more rapidly than compact mass equivalent particles (Potanin, 1991; Flesch et al., 1999), leading to higher restructuring propensities (Oles, 1992). The structures of a range of aggregate types can be characterized using fractal mathematics (Mandelbrot, 1987), in which the expression of fractal dimension gives an indication of aggregate compactness. The mass fractal dimension, Df, describes the relationship between the characteristic length, l, of an aggregate and its mass M: MflDf
(9)
Df can have values between 1 (a line of particles) and 3 (a compact sphere of particles). The measurement of floc structure can be conducted using techniques such as static light scattering (Guan et al., 1998; Lo and Waite, 2000; Jarvis et al., 2005) and image analysis (Tence et al., 1986; Clark and Flora, 1991; Chakraborti et al., 2003). The former will be used in this research, and is considered advantageous due to the non-interfering nature of the technique, however, assumptions concerning data interpretation can be limiting (Thill et al., 2000).
1.3.
Light scattering techniques
Scattering patterns of particles are specific for different particle sizes, and are influenced by the shape, homogeneity, the relative refractive index of particles to medium, and by the wavelength of the incoming radiation. These scattering intensities can be detected by a specific detector array and are translated into a measure of the particle size, according to available approximations that were developed for particles of different sizes and optical properties (Sorensen, 1997). The scattering patterns are related to the scattering angle in a way which can also be used to obtain information on the structure of the aggregates. The vector sum of the scattered waves from each of the component scatterers, q, is related to the scattering angle, q, by (Sorensen, 1997): q ¼ ð4pn=lÞ sinðq=2Þ
(10)
where l is the in vacuo wavelength of the incident beam, and n is the refractive index of the medium. The inverse of this variable (q1) indicates the length scale of the scattering experiment. The angular scattering intensity, I(q), of a fractal aggregate with no multiple scattering is the product of the form factor, P(q), which represents the scattered intensity function from a single primary particle, and the structure factor, S(q), that describes additional scattered intensity due to the spatial correlation between the particles in the aggregate: IðqÞfSðqÞPðqÞ
(11)
6197
The form factor, P(q), is constant at small q, so that I(q) solely depends on the aggregate structure at large length scales (q1 [ rs, primary particle/scatterer radius). The structure factor, S(q), is constant at large q, and therefore I(q) depends on the primary particles at the low length scale of the aggregate. In the range of fractal geometry, S(q) is a power law function of the form (Thill et al., 2000): SðqÞfqDf
(12)
For: rs q1 x, where x is the distance above which the mass distribution inside the aggregate cannot be considered fractal. The above equation can be used for structural measurements if applying the RaleigheGanseDebye (RGD) approximation, which assumes that elementary units (primary particles) within the scattering body (aggregate) all scatter independently. The approximation is valid for non-absorbing particles if both of the following criteria are met: jm 1j 1
(13a)
ð4pn=lÞLjm 1j 1
(13b)
where m is the relative refractive index of the scatterers (primary particles) and L is the length of the scattering body (the diameter of the primary particles). Df is acquired from the slope of the logarithmic plot of I(q) versus q. However, in reality the measured scattering intensity is contributed to by all the aggregates in the scattering volume, and not only a single aggregate composed of equalsized particles as approximated by the scattering theories. Albeit in a dispersion all aggregates contribute differently to S(q) due to variation in their sizes (and therefore also in the validity limit, x), thus the global slope of the structure factor is not strictly the fractal dimension. In light of this, the absolute slope of the logarithmic plot of I(q) versus q will be referred to as the scattering exponent (SE), as it may not necessarily represent the real mass fractal dimension of the aggregates. The SE, does, however, still portray structural properties, and higher SE values will indicate more compact aggregates e and vice versa (Guan et al., 1998; Waite, 1999). The system presented in this study represents a complex water system, thus the light scattering results presented for a mixture of Al(OH)3 and kaolin should be interpreted with caution. However, for relatively narrow distributions, polydispersity effects are expected to be insignificant (Lawler, 1997; Bushell and Amal, 1998).
2.
Materials and methods
2.1.
Colloidal suspension
1.2 g of kaolin (AlSi2O5(OH)4, Aldrich Chemical Company Inc. USA) was suspended in 20 l of distilled water (60 mg/l final concentration), dispersed and homogenized using an Ultraturax 2000 (Ganke and Kunkel, GMBH). 1.66 g of NaHCO3was added (final concentration 83 mg/l), the pH was corrected to 5
6198
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 9 5 e6 2 0 6
and 6.5 with NaOH or H2SO4, and conductivity was increased to 1 mS/cm with NaNO3.
2.2.
EF cell
An EF batch unit was designed (Fig. 1). It consisted of a plexyglass cap, which could be fitted onto a 1 l chemical glass and to which the electrodes were attached. The inner electrode, the cathode, was made from stainless steel and concentric in form (H ¼ 10 cm, D ¼ 2.5 cm). It was fitted onto the arm of the magnetic stirrer, and kept stationary throughout all the experiments. The outer electrode, the anode, was made from aluminum and could be used in two sizes, for applying different current densities: a concentric electrode (H ¼ 10 cm, D ¼ 9.5 cm) and two separate electrodes, with a total effective area of 1/4 of the former electrode. The electrodes were connected to a DC external power source. Mixing conditions within the cell were achieved with the magnetic stirrer and a constant gentle mixing speed of 145 rpm was used for all experiments.
2.3.
Floc size and structure analysis
X X ðVi di Þ= ðVi Þ i
(14)
i
+ -
a
b
2.4.
c d
Fig. 1 e Components of the EF batch cell: a. Full aluminum anode (213.3 cm2); b. 2 aluminum anodes, 1/4 total area of full anode (53.33 cm2); c. stainless steel cathode fitted onto magnetic stirrer arm and attached to plexyglass cap; d. complete EF apparatus fixed onto a 1 l chemical glass and placed onto a magnetic stirrer.
Image analysis
Image analysis was used in conjunction with scattering measurements as a complementary analysis to ascertain floc properties over time. The aggregates resulting from the process were photographed using a digital camera (model DP11, Olympus, Japan) which was mounted onto a Stereoscope (model SZX1, Olympus, Japan). All photographs were taken at 3.0 mega pixel resolution and magnification 50.
2.5.
Size distributions and structural information of kaolineAl(OH)3 flocs were determined as a function of time using a Malvern Mastersizer Microplus (Worchestershire, UK), which ascertains size by analysis of forward scattered light. The size distribution data given by the instrument covers the size range of 50 nm to 500 mm. A HeeNe laser light (l ¼ 633 nm) was passed through a 2.0-mm-width measurement cell in which the sample flowed. The beam was converged by a 300 mm focusing lens and 42 detectors enabling the collection of light scattered from 0.03 to 46.4 . The plane of polarization of the laser beam is parallel to the detector axis (vertical). Size distribution information was obtained using supplied software which uses Mie theory to develop a scattering pattern that matches the scattering pattern of the sample being measured. Information on distribution size is presented in this paper as the volume mean diameter: DðV; 0:5Þ ¼
where Vi is the relative volume in size class i with mean class diameter di. Information on floc structure was obtained by measuring the intensity of light (I) at all detectors and plotting log I versus log q (Eq. (10)). Information regarding the angles of the detectors and intensity correction data, based on the geometric configuration of the detectors, was supplied by Malvern Instruments.
z Potential
The z potential of the suspension was measured using a Malvern Zetamaster S (Worchestershire, UK). Each result was an average of three readings. It is an indirect measurement of the charge on particles, and its value determines the extent of the electrostatic forces of repulsion between charged particles which change with the addition of a coagulant.
2.6.
Experimental procedure
The EF apparatus was fixed onto a 1 l chemical glass containing 800 ml of kaolin suspension, the electrodes submerged in the suspension. The employed initial pH values were chosen to cover the pH range in which Al(OH)3 precipitation would occur and floc evolution could be measurable (5 and 6.5). A current was applied for pre-determined time spans (3, 6 and 10 min) at the end of which samples were analyzed (for pH, size distribution, z potential and image analysis). The complete dose of aluminum was achieved at 10 min, for each applied current, after which additional aluminum was not further introduced into solution. At this stage samples underwent continuous mixing and measurements were taken at two more time intervals: 15 and 20 minutes. For all size distribution measurements, the Malvern Microplus was operated at a gentle pump speed of 400 rpm, so to minimize disruption of floc structure. The EF cell was operated in galvanostatic mode: the current was set and the potential found its own value dependent on the system’s overall resistance. This ensured coagulant production at a pre-determined rate, defined by Faraday’s law. The currents used for the experiments were 0.042 A, 0.11 A and 0.22 A, yielding aluminum doses of 2.43 mg/l, 6.48 mg/l and 12.96 mg/l respectively. These doses are equivalent to aluminum content in 30 mg/l, 80 mg/l and 160 mg/l commercial alum (8.1% of the total molecule). Current densities were altered by changing the anode size. These are summarized in Table 1.
6199
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 9 5 e6 2 0 6
Table 1 e Current densities used for experiments. Current applied 0.042 A 0.11 A 0.22 A
Current density (full anode)
Current density (1/4 anode)
0.197 mA/cm2 0.516 mA/cm2 1.031 mA/cm2
0.788 mA/cm2 2.062 mA/cm2 4.125 mA/cm2
3.
Results and discussion
3.1.
Size distributions
The initial stages of floc growth in the EF process were examined. In EF the application of current does not only promote gradual anodic dissolution of aluminum, but also simultaneous cathodic formation of OH ions that cause a transient rise in pH. This results in speciation transition between the aluminum mononuclear species throughout the process, until cessation of current. Aluminum supersaturation occurs when the aluminum concentration reaches the critical saturation point, above which Al(OH)3 nucleation and precipitation manifests. The degree of Al(OH)3 precursor formation will depend primarily on aluminum dose and pH (Sposito, 1996). Hence, during the initial stages of EF, while the current is still employed, Al(OH)3 precursor mass is continuously being added to the system, as function of current and pH. Fig. 2 shows the evolution of size distributions over time for currents of 0.042 A and 0.22 A (maximum anode area), at initial pH 5 and 6.5. The size distribution graphs show
a
14 0 min (kaolin)
the evolution of size over time, with modal diameters at 15 min reaching above 200 mm, the spread reaching the upper detection limit of the instrument. For 0.042 A, the growth appears similar for the initial 3 min of the process, at both pH values, while after 6 minutes, at pH 6.5 it accelerates, resulting in a wider distribution spread. This trend also occurs when using 0.22 A. The initial 3 min show similar size distributions after which at pH 6.5 an accelerated growth rate is observed. The initial 3 min appear to be an induction period, in which the Al(OH)3 precursors are precipitating. The rate of solid Al(OH)3 formation depends on the precursor concentration, that initially is low, hence the low growth rate. At 3e6 min a polymerization reaction is initiated, resulting in sufficient precipitate sizes and enhanced aggregation as the process progresses. This stage is considerably shorter when using 0.22 A at pH 6.5, as within the first 3 min sizes as large as 100 mm are already observed. The wide distribution spread indicates simultaneous formation of smaller particle aside growth into larger floc. For 0.042 A at pH 5 the size distribution narrows after 15 min, whereas at pH 6.5 after 10 min. For 0.22 A, this occurs at pH 5 and 6.5, after 10 and 6 min respectively. Al(OH)3 precursors are continuously being formed until the current is ceased, at 10 min. Their presence in solution, as colloidal matter, depends on contact opportunities. A larger concentration of precursors would increase the number of collisions, leading to formation of larger aggregates and subsequently providing additional contact opportunities for the submicron precursors (Chowdhury et al., 1991). Initial pH 6.5 provides optimal sweep floc conditions, resulting in effective precipitation of Al(OH)3 precursors. The optimal
c 12 0 min (kaolin)
12
10
3 min
10 6 min 8
10 min
6
15 min
4
volume %
volume %
3 min
2
6 min
8
10 min 6 15 min 4 2
0 0.1
1
10
100
1000
0 0.1
size (µm)
1
10
100
1000
size (µm)
d 16
12 0 min (kaolin)
volume %
10
3 min
3 min 6 min
8
10 min
6
15 min 4
0 min (kaolin)
14 12
volume %
b
6 min
10
10 min
8
15 min
6 4
2
2
0 0.1
1
10
size (µm)
100
1000
0 0.1
1
10
100
1000
size (µm)
Fig. 2 e Evolution of size distributions over time for 0.042 A and 0.22 A at pH 5 and 6.5. a. 0.042 A, pH 5; b. 0.042 A, pH 6.5; c. 0.22 A, pH 5; d. 0.22 A, pH 6.5.
6200
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 9 5 e6 2 0 6
conditions for precipitation at this pH result in an effective polymerization reaction, leading to a more rapid growth rate. This also explains why for the higher employed current a significantly shorter induction period is observed. The increased dosage of aluminum drives the reaction toward enhanced precipitation of Al(OH)3 precursors, which, in turn, accelerates the growth rate. Although pH 5 is not considered an optimal sweep floc regime (Amirtharajah and O’Melia, 1990), the gradual addition of OH ions into solution coupled with the continuous aluminum dosing drives the equilibrium toward Al(OH)3 precursor formation and substantial growth is observed, even so, at a reduced rate compared to pH 6.5. At pH 5 the positive soluble aluminum mononuclear species govern the solution and adsorption/charge destabilization is the predominant coagulation mechanism (Duan and Gregory, 2003). However, in EF, the pH rise over time drives the speciation into sweep floc regime and conditions become favorable for Al(OH)3 precipitation. This transition leads to pH stabilization within 10 min (Fig. 3) due to excess OH removal from solution in the form of solid Al(OH)3. In contrast, at pH 6.5, the pH increases rapidly at the later stages compared to the initial stages. The rise in pH toward a value of 7.0 causes transition out of sweep floc regime as speciation shifts toward soluble negative mononuclear species. In these conditions removal of excess OH is limited, because the Al(OH)3 precipitation is less favorable, leading to the sharper rise in pH. Despite the increase in pH and shift out of optimal precipitation conditions at later stages, the growth at pH 6.5 is more rapid than at pH 5. The impact of optimal precipitation conditions at the initial stages of the process is apparently of great importance in dictating the overall growth rate. Growth of particles into larger flocs is not only dependent on precursor numbers, but also on their ability to form stable bonds. Surface charge also plays a role in determining floc evolution (Nowostawska et al., 2005) which is measured in the form of z potential e indicating the magnitude of electrostatic repulsive forces in the suspension (Fig. 4). The initial values of the kaolin suspension are 30 and 34 mV, which are the z potentials of the kaolin suspension at initial pH 5 and 6.5 respectively. Within 3 min of EF, for both pH values, charge reversal occurs, the z potential acquiring positive values. This is indicative of positively charged hydrolyzed species formation following aluminum anodic dissolution. In the initial stages, at pH 5, positive aluminum mononuclear species in solution are dominant, and the z potential of the suspension attains
a more positive value compared to pH 6.5. Stabilization occurs at 3, 6 and 10 minutes for 0.22 A, 0.11 A and 0.042 A respectively, as conditions become favorable for Al(OH)3 precipitation. The maximum pH values reached in all runs originating from initial pH 5 retain a positive charge on the Al(OH)3 precipitate because its iso-electric point (i.e.p) occurs between pH 8 and 9 (Sposito, 1996). At initial pH 6.5 the process commences in optimal sweep floc conditions and proceeds toward negative soluble aluminum mononuclear species formation. The z potential in these conditions acquires a less positive value than at pH 5, due to the diminished positive mononuclear speciation. Al(OH)3 surface charge is still positive, but nearer its i.e.p. As anodic dissolution proceeds, a small drop in the z potential is measured for all currents after 6 min. At this stage, speciation toward the negative mononuclear species begins to impact, indicated by the sharper rise in pH (Fig. 3), as OH removal from solution in the form of Al(OH)3 precipitate is reduced.
3.2.
Floc growth stages
Several stages of floc growth occur during flocculation. Initially, precursor nucleation is dominant, after which they combine into aggregates and their size increases rapidly. As flocculation continues, flocs form, exhibiting large, porous and open structures that are more susceptible to fragmentation by fluid shear (Tambo, 1991). As a result, the final floc size distribution is a balance between particle growth and breakage (Spicer and Pratsinis, 1996). Fig. 5 shows the evolution of the volume mean diameter with time, for 0.042 A and 0.22 A with different current densities and pH values. The graphs demonstrate that in fact the current density does not affect the growth pattern of the flocs, within the ranges used in this experiment. Hence, all results now will refer to experimental data obtained using full anode area. Within the experimental conditions, the total dose of aluminum dictates the growth pattern, and factors such as current density and voltage (which is dependent on current density) do not impact floc development. All graphs exhibit sigmoidal behavior and three development stages. The first stage is the induction of Al(OH)3 precursors which is reflected by limited growth. An aluminum and OH concentration limit is required to initiate a polymerization reaction that would lead to the formation of adequate precursor numbers. Stage two exhibits enhanced growth of an exponential type (as precursor numbers reach concentration numbers that result in effective collisions and subsequent bond formation), and in pH 6.5
pH 5 6.2
7.5
0.042A
6
0.042A 7.3
0.11A 0.22A
5.6 5.4
0.11A
7.1 pH
pH
5.8
0.22A
6.9 6.7
5.2
6.5
5 0
3
6 10 time (min)
15
0
3
6
10
15
time (min)
Fig. 3 e Changes in pH for each time interval, for 0.042 A, 0.11 A, 0.22 A, at initial pH 5 and 6.5.
6201
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 9 5 e6 2 0 6
0.042 A, pH 6.5
30
0.11 A, pH 6.5
ζ potential (mV)
20
0.22 A, pH 6.5
10
0.042 A, pH 5 0.11 A, pH 5
0 0
5
10
15
20
-10
0.22 A, pH 5
-20 -30 -40 time (min)
Fig. 4 e z potential changes over time for 0.042 A, 0.11 A, 0.22 A at pH 5 and 6.5.
stage three the flocculation rate diminishes because of aggregate fragmentation. These stages have been observed elsewhere (Kusters et al., 1997; Flesch et al., 1999). For both currents stage two commences at similar times (after 3 min), after which enhanced growth rates are observed for 0.22 A, more so at pH 6.5. The transition into steady state (stage three) occurs when the flocs have reached a volume mean diameter of approximately 200 mm, a size where disruption is caused by hydrodynamic stresses leading to fragmentation. The time of transition depends on the growth rate. For 0.042 A at pH 5, this transition occurs at 15 min, while at pH 6.5 at an earlier time of 10 min. For 0.22 A, at pH 5 and 6.5, these times are 10 and 6 min respectively. Additionally, the volume mean diameter obtained at steady state using 0.22 A at pH 6.5 is larger than for all other conditions (above 250 mm). Employing a higher
current (higher aluminum concentration over time), in optimal sweep floc conditions, promotes both an increased floc growth rate and increased floc strength, which manifests itself a lower fragmentation propensity and larger final floc size. This is in agreement with other studies (Francois, 1988), which found an increase both in floc growth rate and collision efficiency with increased alum concentration during kaolin flocculation. Although in EF at initial pH 5, the coagulation mechanism shifts away from adsorption/charge destabilization and into sweep floc regime as the process progresses, this apparently has less impact on the growth rates, than when starting out in optimal sweep floc regime, where initial nucleation and growth are maximized. In the growth stages (induction and exponential), before transition to a steady state due to fragmentation, the connection between the volume mean diameter and time can be written as: DðV; 0:5Þ ¼ Kebt
(15)
where K is a fitting parameter and b is the growth factor, dependent on the unique flocculation conditions. By fitting the initial growth curves with a simple exponential regression, b can be calculated. Table 2 summarizes the various growth factors obtained for various currents and pH values. Fig. 6 shows a good degree of linear correlation between the calculated growth factor and current, for each initial pH. Thus, higher currents produce larger growth factors. In fact, EF at pH 6.5 yields not only larger growth factors, but also a more significant growth dependency on current e which is stronger
0.042A, pH 5
0.22A, pH 5
300
300 0.197mA/cm^2
250
1.03mA/cm^2
200
D(v,0.5)
D(v,0.5)
4.12mA/cm^2
250
0.78mA/cm^2
150
200 150
100
100
50
50
0
0 0
5
10
15
20
25
0
5
time (min)
15
20
25
time (min) 0.22A, pH 6.5
0.042A, pH 6.5 300
300
4.12mA/cm^2
0.197mA/cm^2
250
250
1.03mA/cm^2
0.78mA/cm^2 200
D(v, 0.5)
D(v, 0.5)
10
150
200 150
100
100
50
50 0
0 0
5
10
15
time (min)
20
25
0
5
10
15
20
25
time (min)
Fig. 5 e Evolution of the volume mean diameter for various current densities and pH values.
6202
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 9 5 e6 2 0 6
Table 2 e Growth factors calculated from initial growth curves. Initial pH b (0.042 A) 5.0 6.5
0.236 0.297
R2
b (0.11 A)
R2
b (0.22 A)
R2
0.98 0.98
0.303 0.329
0.92 0.94
0.381 0.528
0.99 0.99
by approximately a factor of 1.7, compared to the dependency observed at pH 5. These differences are most likely connected to Al(OH)3 precursor formation in the primary stages of the process, where at initial pH 6.5 Al(OH)3 precipitation in the induction phase is more pronounced compared to initial pH 5. However, initial precursor formation is not necessarily the sole reason for the differences observed. Collision efficiency, dictated by attractive/repulsive surface forces also plays a significant role in aggregation and growth into larger floc. Indeed, the electrostatic repulsive barrier at pH 6.5, was lower than at pH 5, reflected by a less positive z potential. The growth profile is also a function of the collision frequency, with larger collision profiles (a more open structure), enhancing the collision frequency (Kusters et al., 1997; Flesch et al., 1999). Different alum doses can generate different floc structures (Tambo and Watanabe, 1979), thus the current applied in EF should also have a similar effect. However, in EF, the current not only dictates the coagulant dose, but also creates a dynamic physical/chemical environment in which additional by-products are formed that directly affect the coagulation/flocculation mechanisms. Cathodic hydroxyl formation causes a rise in pH over time, shifting the speciation of mononuclear species toward sweep floc regime or out of sweep floc regime, depending on the initial pH. Gaseous hydrolysis products (oxygen and hydrogen) are simultaneously produced, hence micro-bubbles are also being introduced in solution. Bubble nucleation rates increases with increasing currents (Shahjahan Kaisar Alam Sarkar et al., 2010) and may contribute as a “sticking” factor toward effective precursor aggregation and subsequent floc growth. An electrical field exists, however one can assume that in orthokinetic coagulation, the effect of such a field on particle/ colloidal transport is negligible, because bulk fluid motion dominates. All the above demonstrate the intricate environment EF generates. As such, one must consider all these factors when investigating evolution of floc structure. Understanding floc structural evolution is a crucial factor in elucidating floc size evolution. 0.6 pH 5
0.5
y = 1.3466x + 0.2177 R² = 0.9339
pH 6.5
b (arb)
0.4 0.3
y = 0.8045x + 0.2069 R² = 0.9916
0.2 0.1 0 0
0.05
0.1
0.15
0.2
0.25
current (A)
Fig. 6 e Dependency of growth factor on current at pH 5 and 6.5.
3.3.
Structural evolution
Flocs undergo a series of processes including aggregation, fragmentation and, in some cases, structural rearrangement. The evolution of floc structure throughout the flocculation process depends on various factors such as floc size, formation and break-up rates, bonding forces and hydrodynamic forces (Selomulya et al., 2003). The change in floc structure can be quantified by monitoring the variation in the fractal dimension, or the scattering exponent (SE). As mentioned previously, the SE will be used in this research to define the structural properties of the flocs. For obtaining structural data, the RGD approximation requirement must be met (Eqs. (13a) and (13b)). Precipitation of Al(OH)3 precursors and their growth has been studied (Li et al., 2005). It has been shown that detectable precursors/ nuclei appear in the range of 100 nm. For amorphous Al(OH)3 precursor formation, a refractive index of 1.59 can be used, based on the refractive index of other Al(OH)3 solid forms (Li et al., 2005), thereby meeting the RGD approximation requirement: m ¼ nAlðOHÞ3 =nwater ¼ 1:59=1:33 ¼ 1:195: and ð4pnwater =llaser ÞLjm 1j ¼ ð4p1:33=633Þ 100 0:195 ¼ 0:5 Fig. 7 shows typical scattering graphs of flocs at different time intervals. The graphs obtained for 0.22 A, at both pH values, show that transition into an angle independent scattering region at low q is hardly apparent. A mild transition does occur at the initial stages of the process, at 3 min, after which the graphs exhibit a dependency on q, even in the low range. This is presumably a result of the large size of the flocs obtained for 0.22 A, after 6 min and above, at both pH values. For 0.042 A, the transition is more apparent, and occurs for most time intervals, at both pH values, with the exception of pH 6.5 in which the transition occurs after 10 and 15 min, most likely due to the size of the flocs. For all graphs, one decade of linearity is observed for: 1.4*104 < q < 1.6*103. This region will be used to calculate the SE, from the slope of the scattering plots. The results demonstrate that for all currents and pH values compaction occurs during the process, resulting in a final SE, at 15 min, which is lower than that obtained during the initial 3 min. In some conditions, fluctuations in the structural evolution are observed, thus providing some indication of the extreme fragile nature of the flocs. For 0.042 A, the flocs formed at pH 6.5 are more compact than those formed at pH 5. However, for 0.22 A, the differences between the two pH values are not as distinct. Comparing 0.22 A to 0.042 A, at pH 6.5 it is evident that the higher current (aluminum dose) produced less compact flocs, throughout all stages of the process. The larger growth profile, reflected by more porous flocs, also contributed to the growth rate which is enhanced at 0.22 A. A larger collision profile leads to a larger collision frequency, hence the increased growth rate. Despite early fragmentation when employing 0.22 A, compared to 0.042 A (Fig. 5), the flocs still maintain a more open structure, and reach a larger size by 15 min. This suggests that at higher currents stronger bonds exist between the precursors comprising the floc, resulting in a floc strength that is able to
6203
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 9 5 e6 2 0 6
0.042A, pH 6.5
0.042A, pH 5
100000
3 min 6 min 10 min 15 min
10000
100 10
3 min 6 min 10 min 15 min
10000 1000 100
I (arb)
I (arb)
1000
100000
10
1
1
0.1
0.1
0.01 0.00001
0.0001
q
0.001
0.01
0.01
0.1
0.00001
0.0001
(nm-1)
0.001
q
0.22A, pH 6.5
0.1
0.22A, pH 5
100000
100000
6 min 10 min
1000
15 min
100 10
3 min
10000
6 min 10 min
1000
I (arb)
3 min 10000
I (arb)
0.01
(nm-1)
10
1
1
0.1
0.1
0.01
15 min
100
0.01
0.00001
0.0001
q
0.001
0.01
0.1
0.00001
(nm-1)
0.0001
0.001
q
0.01
0.1
(nm-1)
Fig. 7 e Conventional static light scattering plots (log (I ) vs. log (q)) for Al(OH)3 flocs forming as a function of time.
withstand shear forces far better than that obtained with lower currents. Table 3 shows structural characteristics of flocs, portrayed by the SE, at different operating conditions. This could be a result of the extent of transition out of favorable precipitation conditions. For 0.22 A, compared to 0.042 A, this transition is sharper, with the final pH reaching 7.4 as opposed to 7.0 for 0.042 A. This affects the surface charge of the particles, as a higher pH reduces the surface charge of the Al(OH)3 precipitate, lowering it toward its isoelectric point. A lower surface charge decreases electrostatic repulsive forces, resulting in a higher collision efficiency and a stronger floc (stronger bonds). Moreover, with higher currents, the generation of gaseous products is increased leading to larger amounts of micro-bubbles. These are most likely incorporated in the floc, and may also serve as additional “glue”, thus assisting in the formation of a more porous structure. Fig. 8 illustrates these differences visually. For 0.042 A at pH 6.5, the flocs are well defined, and compact in structure, whereas 0.22 A yields structures which are extremely tenuous displaying a porosity which seems even
transparent in some regions. At pH 5, differences in the SE obtained at different currents are not appreciable, and exhibit a rather open structure, as indicated in Fig. 9. At this pH, the transition into more favorable Al(OH)3 precipitation conditions occurs at the later stages of the process (as the pH and aluminum dose increase over time), thus the overall precipitate mass introduced into the system is less than at pH 6.5. In this case, incorporation of the micro-bubbles into flocs containing less mass would probably result in a more open structure. In contrast to pH 6.5, the flocs also exhibit less strength, and despite their large collision profile do not reach the size obtained at pH 6.5 when using high currents. At pH 5, the surface charge of the Al(OH)3 precursors is more positive (further from the i.e.p) compared to that obtained at pH 6.5, indicating that a stronger repulsive electrostatic barrier exists. The above data shows that size and structure evolution of flocs in EF of a colloidal suspension stems from a delicate balance between mass introduced (precursor numbers) at the initial stages of the process and the amount of gaseous products (in the form of micro-bubbles) in solution e both are
Table 3 e Scattering exponents obtained for various currents and pH values. Higher values indicate more compact structures and lower values more porous structures. Time (min)
3 6 10 15
SE, pH 6.5
SE, pH 5
0.042 A
0.11 A
0.22 A
0.042 A
0.11 A
0.22 A
1.83 0.04 1.93 0.01 1.89 0.01 1.93 0.01
1.78 0.02 1.95 0.01 1.88 0.02 1.93 0.01
1.77 0.02 1.78 0.01 1.81 0.02 1.86 0.01
1.75 0.06 1.78 0.02 1.88 0.01 1.87 0.01
1.86 0.02 1.81 0.02 1.90 0.01 1.87 0.01
1.78 0.03 1.84 0.02 1.87 0.02 1.87 0.02
6204
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 9 5 e6 2 0 6
Fig. 8 e Floc structures obtained at pH 6.5 for 0.042 A and 0.22 A e 6 and 10 min.
current dependent. The initial precursor numbers are also a function of pH, which creates, in conjunction with the current, conditions favorable for Al(OH)3 super-saturation and precipitation. However, these also affect the surface charge of the flocs forming, which in turn affects the strength of the bonds between the precursor particles and ultimately the collision efficiency and size of the floc. All these parameters, which are intertwined, are summarized in a conceptual model (Fig. 10) which can serve as a prediction tool for floc structure and size evolution, with reference to the specific conditions studied here (EF of a model kaolin suspension). At pH 5 less mass is generated as conditions do not produce an optimal sweep floc regime. The repulsive electrostatic barrier at this pH is stronger for all currents, therefore the collision efficiency (a) is lower. The floc structure depends on
the ratio of micro-bubbles/mass, as more open structures evolve, as the ratio increases. For lower currents this ratio is maintained (a low current results both in low mass and less micro-bubbles), but the reduced mass will result in flocs exhibiting a lower collision efficiency (b) that evolve slowly e but still portray a porous structure. For higher currents, the collision frequency is higher as more mass is generated along with hydrolysis products, resulting in a more rapidly evolving porous structure. At pH 6.5 more mass is generated as the conditions are indeed optimal for Al(OH)3 precipitation. The repulsive surface forces are weaker, reflected by a z potential which is nearer to the i.e.p of Al(OH)3 e hence, the collision efficiency is higher. For lower currents, the micro-bubble/mass ratio is smaller, and coupled with the reduced mass will result in
Fig. 9 e Floc structures obtained at pH 5, for various currents at 10 min.
6205
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 9 5 e6 2 0 6
applied current
high
low
sweep floc regime optimal
lower surface charge higher mass higher bubble generation higher α higher β rapidly evolving more stable porous flocs
not optimal
higher surface charge lower mass higher bubble generation lower α higher β rapidly evolving less stable porous flocs
sweep floc regime not optimal
higher surface charge lower mass lower bubble generation lower α lower β slowly evolving less stable porous flocs
optimal
lower surface charge higher mass lower bubble generation higher α lower β slowly evolving more stable compact flocs
Fig. 10 e Conceptual model predicting floc evolution e rate and structure.
a more compact floc which evolves slowly. For higher currents, micro-bubble generation is increased, resulting in a more open structure, but exhibiting strength. These flocs have both a high collision frequency and high collision efficiency e hence, the increased growth dependency on current and increased growth rate.
This versatility has clear implications to many processes of importance in water and wastewater treatment as flocs can be “tailored” for specific processes, depending on their requirements.
Acknowledgments 4.
Conclusions
EF consists of synergistic processes which create a dynamic environment in which floc formation and growth occur. The system’s response to current and pH change will dictate the kinetics of floc formation and growth in EF, and subsequently floc structure. The system studied serves as a model for natural occurring systems containing colloidal matter, and displayed kinetic behavior which can be connected to specific system characteristics, derived from operational parameters. Floc growth was found to be independent of current density, and only dependent on absolute current values (within the values examined). All evolution patterns, for all conditions, included three typical stages: induction, exponential growth and fragmentation, as found in conventional flocculation. Higher growth rates were obtained for higher currents. A higher dependency of growth rate on current occurred at pH 6.5, in optimal sweep floc regime. This appears to stem from the stronger linkage between particles at this pH, coupled with increased mass and micro-bubble generation for higher currents e resulting in larger collision efficiencies and enhanced growth rate. The flocs undergo compaction and in some cases structural fluctuation occurring during growth. This reflects somewhat their fragile nature. EF can generate range of flocs, exhibiting different structural characteristics, depending on the conditions of operation.
The work was partially supported by BMBF Germany and Israeli Ministry of Science. We gratefully acknowledge Dr. Rivka Amit and Mr. Yoav Nahamias from the Geological Survey of Israel for assistance with the Malvern Mastersizer Microplus, and other instrumentation.
references
Adin, A., Vescan, N., 2002. Electroflocculation for particle destabilization and aggregation for municipal water and wastewater treatment. Proceedings of the American Chemistry Society 42 (2), 537e541. Amirtharajah, A., O’Melia, C.R., 1990. Coagulation processes: destabilization, mixing and flocculation. In: Water Quality and Treatment. McGraw-Hill, New York. Belongia, B.M., Haworth, P.D., Baygents, J.C., Raghavan, S., 1999. Treatment of alumina and silica chemical mechanical polishing waste by electrodecantation and electrocoagulation. Journal of the Electrochemical Society 146 (11), 4124e4130. Bushell, G., Amal, R., 1998. Fractal aggregates of polydisperse particles. Journal of Colloid and Interface Science 205, 459e469. Chakraborti, R.K., Gardner, K.H., Atkinson, J.F., Van Benschoten, J. E., 2003. Changes in fractal dimension during aggregation. Water Research 37, 873e883. Chen, G., Chen, X., Yue, P.L., 2000. Electrocoagulation and electroflotation of restaurant wastewater. Journal of Environmental Engineering 126 (9), 858e863. Chowdhury, Z.K., Amy, G.L., Bales, R.C., 1991. Coagulation of submicron colloids in water treatment by incorporation into
6206
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 1 9 5 e6 2 0 6
aluminum hydroxide floc. Journal of Environmental Science and Technology 25, 1766e1773. Clark, M.M., Flora, J.R.V., 1991. Floc restructuring in varied turbulent mixing. Journal of Colloid and Interface Science 147, 407e421. Do, J.S., Chen, M.L., 1994. Decolorization of dye-containing solutions by electrocoagulation. Journal of Applied Electrochemistry 24, 785e790. Duan, J., Gregory, J., 2003. Coagulation by hydrolyzing salts. Advances in Colloid and Interface Science 100e102, 475e502. Emamjomeh, M.M., Sivakumar, M., 2009. Review of pollutants removed by electrocoagulation and electrocoagulation/ flotation process. Journal of Environmental Management 90, 1663e1679. Flesch, J.C., Spicer, P.T., Pratsinis, S.E., 1999. Laminar and turbulent shear-induced flocculation of fractal aggregates. AIChE Journal 45 (5), 1114e1124. Francois, R.J., 1988. Growth kinetics of hydroxide flocs. Journal of American Water Works Association 80, 92e96. Guan, J., Waite, T.D., Amal, R., 1998. Rapid structure characterization of bacterial aggregates. Environmental Science and Technology 32, 3735e3742. Harif, T., Adin, A., 2007. Characteristics of aggregates formed by electroflocculation of a colloidal suspension. Water Research 41, 2951e2961. Holt, P.K., Barton, G.W., Wark, M., Mitchell, C.A., 2002. A quantitative comparison between chemical dosing and electrocoagulation. Colloids and Surfaces A: Physicochemical Engineering Aspects 211, 233e248. Jarvis, P., Jefferson, B., Parsons, S.A., 2005. Measuring of floc structural characteristics. Reviews in Environmental Science and Biotechnology 4 (1e2), 1e18. Jiang, Q., Logan, B.E., 1991. Fractal dimensions of aggregates determined from steady state size distributions. Environmental Science and Technology 25, 2031e2038. Kusters, K.A., Wijers, J.G., Thoenes, D., 1997. Aggregation kinetics of small particles in agitated vessels. Chemical Engineering Science 52 (1), 107e121. Lakshmanan, D., Clifford, D.A., Samanta, G., 2009. Ferrous and ferric ion generation during iron electrocoagulation. Environmental Science and Technology 43, 3853e3859. Lawler, D.F., 1997. Particle size distributions in treatment processes: theory and practice. Water Science and Technology 36 (4), 15e23. Letterman, R.D., Amirtharajah, A., O’Melia, C.R., 1999. Coagulation and flocculation. In: Letterman, R.D. (Ed.), AWWA “Water Quality and Treatment”. Li, H., Addai-Mensah, J., Thomas, J.C., Gerson, A.R., 2005. The influence of Al(III) supersaturation and NaOH concentration on the rate of crystallization of Al(OH)3 precursor particles from sodium aluminate solutions. Journal of Colloid and Interface Science 286, 511e519. Lo, B., Waite, T.D., 2000. Structure of hydrous ferric oxide aggregates. Journal of Colloid and Interface Science 222, 83e89. Mandelbrot, B.B., 1987. The Fractal Geometry of Nature. Freeman, New York. Matteson, M.J., Dobson, R.L., Glenn Jr., R.W., Kukunoor, N.S., Waits III, W.H., Clayfield, E.J., 1995. Electrocoagulation and separation of aqueous suspensions of ultrafine particles. Colloids and Surfaces A: Physicochemical Engineering Aspects 104, 101e109.
Mollah, M.Y.A., Schennech, R., Parga, J.R., Cocke, D.L., 2001. Electrocoagulation (EC) e science and applications. Journal of Hazardous Materials B84, 29e41. Mollah, M.Y.A., Morkovsky, P., Gomes, J.A.G., Kesmez, M., Parga, J., Cocke, D.L., 2004. Fundamentals, present and future perspectives of electrocoagulation. Journal of Hazardous Materials B114, 199e210. Nowostawska, U., Sander, S.G., McGrath, K.M., Hunter, K.A., 2005. Effects of coagulants on the surface forces of colloidal alumina under water treatment conditions. Colloids and Surfaces A: Physicochemical Engineering Aspects 266, 214e222. Oles, V., 1992. Shear-induced aggregation and breakup of polystyrene latex particles. Journal of Colloid and Interface Science 154, 351e358. Potanin, A.A., 1991. On the mechanism of aggregation and breakup of polystyrene. Journal of Colloid and Interface Science 145, 140e157. Pouet, M.F., Grasmick, A., 1995. Urban wastewater treatment by electrocoagulation and flotation. Water Science and Technology 31 (3e4), 275e283. Selomulya, C., Bushell, G., Amal, R., Waite, T.D., 2003. Understanding the role of restructuring in flocculation: the application of a population balance model. Chemical Engineering Science 58, 327e338. Shahjahan Kaisar Alam Sarkar, Md., Evans, G.M., Donne, S.W., 2010. Bubble size measurement in electroflotation. Minerals Engineering 23, 1058e1065. Smoluchowski, M., 1917. Versuch einer mathematischen theorie der koagulations e kinetik kolloides losungen. Zeitschrift fu¨r Physikalische Chemie 92, 129e168. Sorensen, C.M., 1997. In: Birdi, K.S. (Ed.), Surface and Colloid Chemistry. CRC Press, Boca Raton, FL, pp. 533e558 (chapter 13). Spicer, P.T., Pratsinis, S.E., 1996. Shear-induced flocculation: the evolution of floc structure and the shape of the size distribution at steady-state. Water Research 30 (5), 1049e1056. Sposito, G., 1996. The Environmental Chemistry of Aluminum, second ed. CRC Press Ltd., Florida. Sun, L., Miznikov, E., Wang, L., Adin, A., 2009. Nickel removal from wastewater by electroflocculation-filtration by hybridization. Desalination 249, 832e836. Tambo, N., 1991. Basic concepts and innovative turn of coagulation/flocculation. Water Supply 9, 1e10. Tambo, N., Watanabe, Y., 1979. Physical characteristics of flocs I: the floc density and aluminum floc. Water Research 13, 409e419. Tence, M., Chevalier, J.P., Jullien, R., 1986. On the measurement of the fractal dimension of aggregated particles by electronmicroscopy experimental method, corrections and comparison with numerical models. Journal De Physique 47 (11), 1989e1998. Thill, A., Lambert, S., Moustier, S., Ginestet, P., Audic, J.M., Bottero, J.Y., 2000. Structural interpretations of static light scattering patterns of aggregates II. Experimental study. Journal of Colloid and Interface Science 228, 386e392. Vik, E.A., Carlson, D.A., Eikum, A.S., Gjessing, E.T., 1984. Electrocoagulation of potable water. Water Research 18 (11), 1355e1360. Waite, T.D., 1999. Measurement and implications of floc structure in water and wastewater treatment. Colloids and Surfaces A 151, 27e41. Yildiz, Y.S., Koparal, A.S., Keskinler, B., 2008. Effect of initial pH and supporting electrolyte on the treatment of water containing high concentration of humic substances by electrocoagulation. Chemical Engineering Journal 138, 63e72.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 0 7 e6 2 1 6
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Dynamic and distribution of ammonia-oxidizing bacteria communities during sludge granulation in an anaerobiceaerobic sequencing batch reactor Zhang Bin a,b, Chen Zhe a,b, Qiu Zhigang a,b, Jin Min a,b, Chen Zhiqiang a,b, Chen Zhaoli a,b, Li Junwen a,b, Wang Xuan c,*, Wang Jingfeng a,b,** a
Institute of Hygiene and Environmental Medicine, Academy of Military Medical Sciences, Tianjin 300050, PR China Tianjin Key Laboratory of Risk Assessment and Control for Environment and Food Safety, Tianjin 300050, PR China c Tianjin Key Laboratory of Hollow Fiber Membrane Material and Membrane Process, Institute of Biological and Chemical Engineering, Tianjin Polytechnical University, Tianjin 300160, PR China b
article info
abstract
Article history:
The structure dynamic of ammonia-oxidizing bacteria (AOB) community and the
Received 30 June 2011
distribution of AOB and nitrite-oxidizing bacteria (NOB) in granular sludge from an
Received in revised form
anaerobiceaerobic sequencing batch reactor (SBR) were investigated. A combination of
10 September 2011
process studies, molecular biotechniques and microscale techniques were employed to
Accepted 10 September 2011
identify and characterize these organisms. The AOB community structure in granules was
Available online 17 September 2011
substantially different from that of the initial pattern of the inoculants sludge. Along with granules formation, the AOB diversity declined due to the selection pressure imposed by
Keywords:
process conditions. Denaturing gradient gel electrophoresis (DGGE) and sequencing results
Ammonia-oxidizing bacteria
demonstrated that most of Nitrosomonas in the inoculating sludge were remained because
Granular sludge
of their ability to rapidly adapt to the settlingewashing out action. Furthermore, DGGE
Community development
analysis revealed that larger granules benefit more AOB species surviving in the reactor. In
Granule size
the SBR were various size granules coexisted, granule diameter affected the distribution
Nitrifying bacteria distribution
range of AOB and NOB. Small and medium granules (d < 0.6 mm) cannot restrict oxygen
Phylogenetic diversity
mass transfer in all spaces of the sludge. Larger granules (d > 0.9 mm) can result in smaller aerobic volume fraction and inhibition of NOB growth. All these observations provide support to future studies on the mechanisms responsible for the AOB in granules systems. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
At sufficiently high levels, ammonia in aquatic environments can be toxic to aquatic life and can contribute to eutrophication. Accordingly, biodegradation and elimination of ammonia in wastewater are the primary functions of the
wastewater treatment process. Nitrification, the conversion of ammonia to nitrate via nitrite, is an important way to remove ammonia nitrogen. It is a two-step process catalyzed by ammonia-oxidizing and nitrite-oxidizing bacteria (AOB and NOB). Aerobic ammonia-oxidation is often the first, ratelimiting step of nitrification; however, it is essential for the
* Corresponding author. ** Corresponding author. Institute of Hygiene and Environmental Medicine, Academy of Military Medical Sciences, Tianjin 300050, PR China. Tel.: +86 22 84655498; fax: +86 22 23328809. E-mail addresses: [email protected] (W. Xuan), [email protected] (W. Jingfeng). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.09.026
6208
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 0 7 e6 2 1 6
removal of ammonia from the wastewater (Prosser and Nicol, 2008). Comparative analyses of 16S rRNA sequences have revealed that most AOB in activated sludge are phylogenetically closely related to the clade of b-Proteobacteria (Kowalchuk and Stephen, 2001). However, a number of studies have suggested that there are physiological and ecological differences between different AOB genera and lineages, and that environmental factors such as process parameter, dissolved oxygen, salinity, pH, and concentrations of free ammonia can impact certain species of AOB (Erguder et al., 2008; Kim et al., 2006; Koops and Pommerening-Ro¨ser, 2001; Kowalchuk and Stephen, 2001; Shi et al., 2010). Therefore, the physiological activity and abundance of AOB in wastewater processing is critical in the design and operation of waste treatment systems. For this reason, a better understanding of the ecology and microbiology of AOB in wastewater treatment systems is necessary to enhance treatment performance. Recently, several developed techniques have served as valuable tools for the characterization of microbial diversity in biological wastewater treatment systems (Li et al., 2008; Yin and Xu, 2009). Currently, the application of molecular biotechniques can provide clarification of the ammoniaoxidizing community in detail (Haseborg et al., 2010; Tawan et al., 2005; Vlaeminck et al., 2010). In recent years, the aerobic granular sludge process has become an attractive alternative to conventional processes for wastewater treatment mainly due to its cell immobilization strategy (de Bruin et al., 2004; Liu et al., 2009; Schwarzenbeck et al., 2005; Schwarzenbeck et al., 2004a,b; Xavier et al., 2007). Granules have a more tightly compact structure (Li et al., 2008; Liu and Tay, 2008; Wang et al., 2004) and rapid settling velocity (Kong et al., 2009; Lemaire et al., 2008). Therefore, granular sludge systems have a higher mixed liquid suspended sludge (MLSS) concentration and longer solid retention times (SRT) than conventional activated sludge systems. Longer SRT can provide enough time for the growth of organisms that require a long generation time (e.g., AOB). Some studies have indicated that nitrifying granules can be cultivated with ammonia-rich inorganic wastewater and the diameter of granules was small (Shi et al., 2010; Tsuneda et al., 2003). Other researchers reported that larger granules have been developed with the synthetic organic wastewater in sequencing batch reactors (SBRs) (Li et al., 2008; Liu and Tay, 2008). The diverse populations of microorganisms that coexist in granules remove the chemical oxygen demand (COD), nitrogen and phosphate (de Kreuk et al., 2005). However, for larger granules with a particle diameter greater than 0.6 mm, an outer aerobic shell and an inner anaerobic zone coexist because of restricted oxygen diffusion to the granule core. These properties of granular sludge suggest that the inner environment of granules is unfavorable to AOB growth. Some research has shown that particle size and density induced the different distribution and dominance of AOB, NOB and anammox (Winkler et al., 2011b). Although a number of studies have been conducted to assess the ecology and microbiology of AOB in wastewater treatment systems, the information on the dynamics, distribution, and quantification of AOB communities during sludge granulation is still limited up to now. To address these concerns, the main objective of the present work was to investigate the population dynamics of AOB communities during the development of seeding flocs
into granules, and the distribution of AOB and NOB in different size granules from an anaerobiceaerobic SBR. A combination of process studies, molecular biotechniques and microscale techniques were employed to identify and characterize these organisms. Based on these approaches, we demonstrate the differences in both AOB community evolution and composition of the flocs and granules co-existing in the SBR and further elucidate the relationship between distribution of nitrifying bacteria and granule size. It is expected that the work would be useful to better understand the mechanisms responsible for the AOB in granules and apply them for optimal control and management strategies of granulation systems.
2.
Material and methods
2.1.
Reactor set-up and operation
The granules were cultivated in a lab-scale SBR with an effective volume of 4 L. The effective diameter and height of the reactor was 10 cm and 51 cm, respectively. The hydraulic retention time was set at 8 h. Activated sludge from a full-scale sewage treatment plant (Jizhuangzi Sewage Treatment Works, Tianjin, China) was used as the seed sludge for the reactor at an initial sludge concentration of 3876 mg L1 in MLSS. The reactor was operated on 6-h cycles, consisting of 2-min influent feeding, 90min anaerobic phase (mixing), 240-min aeration phase and 5min effluent discharge periods. The sludge settling time was reduced gradually from 10 to 5 min after 80 SBR cycles in 20 days, and only particles with a settling velocity higher than 4.5 m h1 were retained in the reactor. The composition of the influent media were NaAc (450 mg L1), NH4Cl (100 mg L1), (NH4)2SO4 (10 mg L1), KH2PO4 (20 mg L1), MgSO4$7H2O (50 mg L1), KCl (20 mg L1), CaCl2 (20 mg L1), FeSO4$7H2O (1 mg L1), pH 7.0e7.5, and 0.1 mL L1 trace element solution (Li et al., 2007). Analytical methods-The total organic carbon (TOC), NHþ 4 eN, NO 2 eN, NO3 eN, total nitrogen (TN), total phosphate (TP) concentration, mixed liquid suspended solids (MLSS) concentration, and sludge volume index at 10 min (SVI10) were measured regularly according to the standard methods (APHA-AWWA-WEF, 2005). Sludge size distribution was determined by the sieving method (Laguna et al., 1999). Screening was performed with four stainless steel sieves of 5 cm diameter having respective mesh openings of 0.9, 0.6, 0.45, and 0.2 mm. A 100 mL volume of sludge from the reactor was sampled with a calibrated cylinder and then deposited on the 0.9 mm mesh sieve. The sample was subsequently washed with distilled water and particles less than 0.9 mm in diameter passed through this sieve to the sieves with smaller openings. The washing procedure was repeated several times to separate the granules. The granules collected on the different screens were recovered by backwashing with distilled water. Each fraction was collected in a different beaker and filtered on quantitative filter paper to determine the total suspended solid (TSS). Once the amount of total suspended solid (TSS) retained on each sieve was acquired, it was reasonable to determine for each class of size (<0.2, [0.2e0.45], [0.45e0.6], [0.6e0.9], >0.9 mm) the percentage of the total weight that they represent.
6209
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 0 7 e6 2 1 6
2.2.
DNA extraction and nested PCReDGGE
2.3.
The sludge from approximately 8 mg of MLSS was transferred into a 1.5-mL Eppendorf tube and then centrifuged at 14,000 g for 10 min. The supernatant was removed, and the pellet was added to 1 mL of sodium phosphate buffer solution and aseptically mixed with a sterilized pestle in order to detach granules. Genomic DNA was extracted from the pellets using E.Z.N.A. Soil DNA kit (D5625-01, Omega Bio-tek Inc., USA). To amplify ammonia-oxidizer specific 16S rRNA for denaturing gradient gel electrophoresis (DGGE), a nested PCR approach was performed as described previously (Zhang et al., 2010). 30 ml of nested PCR amplicons (with 5 ml 6 loading buffer) were loaded and separated by DGGE on polyacrylamide gels (8%, 37.5:1 acrylamideebisacrylamide) with a linear gradient of 35%e55% denaturant (100% denaturant ¼ 7 M urea plus 40% formamide). The gel was run for 6.5 h at 140 V in 1 TAE buffer (40 mM Tris-acetate, 20 mM sodium acetate, 1 mM Na2EDTA, pH 7.4) maintained at 60 C (DCode Universal Mutation Detection System, Bio-Rad, Hercules, CA, USA). After electrophoresis, silver-staining and development of the gels were performed as described by Sanguinetti et al. (1994). These were followed by air-drying and scanning with a gel imaging analysis system (Image Quant350, GE Inc., USA). The gel images were analyzed with the software Quantity One, version 4.31(Bio-rad). Dice index (Cs) of pair wise community similarity was calculated to evaluate the similarity of the AOB community among DGGE lanes (LaPara et al., 2002). This index ranges from 0% (no common band) to 100% (identical band patterns) with the assistance of Quantity One. The Shannon diversity index (H ) was used to measure the microbial diversity that takes into account the richness and proportion of each species in a population. H was calculated P ni n log i , where ni/N using the following equation: H ¼ N N is the proportion of community made up by species i (brightness of the band i/total brightness of all bands in the lane). Dendrograms relating band pattern similarities were automatically calculated without band weighting (consideration of band density) by the unweighted pair group method with arithmetic mean (UPGMA) algorithms in the Quantity One software. Prominent DGGE bands were excised and dissolved in 30 mL Milli-Q water overnight, at 4 C. DNA was recovered from the gel by freezeethawing thrice. Cloning and sequencing of the target DNA fragments were conducted following the established method (Zhang et al., 2010).
Distribution of nitrifying bacteria
Three classes of size ([0.2e0.45], [0.45e0.6], >0.9 mm) were chosen on day 180 for FISH analysis in order to investigate the spatial distribution characteristics of AOB and NOB in granules. 2 mg sludge samples were fixed in 4% paraformaldehyde solution for 16e24 h at 4 C and then washed twice with sodium phosphate buffer; the samples were dehydrated in 50%, 80% and 100% ethanol for 10 min each. Ethanol in the granules was then completely replaced by xylene by serial immersion in ethanol-xylene solutions of 3:1, 1:1, and 1:3 by volume and finally in 100% xylene, for 10 min periods at room temperature. Subsequently, the granules were embedded in paraffin (m.p. 56e58 C) by serial immersion in 1: 1 xylene-paraffin for 30 min at 60 C, followed by 100% paraffin. After solidification in paraffin, 8-mm-thick sections were prepared and placed on gelatin-coated microscopic slides. Paraffin was removed by immersing the slide in xylene and ethanol for 30 min each, followed by air-drying of the slides. The three oligonucleotide probes were used for hybridization (Downing and Nerenberg, 2008): FITC-labeled Nso190, which targets the majority of AOB; TRITC-labeled NIT3, which targets Nitrobacter sp.; TRITC-labeled NSR1156, which targets Nitrospira sp. All probe sequences, their hybridization conditions, and washing conditions are given in Table 1. Oligonucleotides were synthesized and fluorescently labeled with fluorochomes by Takara, Inc. (Dalian, China). Hybridizations were performed at 46 C for 2 h with a hybridization buffer (0.9 M NaCl, formamide at the percentage shown in Table 1, 20 mM Tris/HCl, pH 8.0, 0.01% SDS) containing each labeled probe (5 ng mL1). After hybridization, unbound oligonucleotides were removed by a stringent washing step at 48 C for 15 min in washing buffer containing the same components as the hybridization buffer except for the probes. For detection of all DNA, 4, 6-diamidino-2-phenylindole (DAPI) was diluted with methanol to a final concentration of 1 ng mL1. Cover the slides with DAPIemethanol and incubate for 15 min at 37 C. The slides were subsequently washed once with methanol, rinsed briefly with ddH2O and immediately air-dried. Vectashield (Vector Laboratories) was used to prevent photo bleaching. The hybridization images were captured using a confocal laser scanning microscope (CLSM, Zeiss 710). A total of 10 images were captured for each probe at each class of size. The representative images were selected and final image evaluation was done in Adobe PhotoShop.
Table 1 e Oligonucleotide probes used for ecology analysis in different size granules. Probe
Probe sequence (50 e30 )
Target sitea (rRNA positions)
%FAb
NaCl (mM)c
NSO190 NIT3d NSR1156
CGATCCCCTGCTTTTCTCC CCTGTGCTCCATGCTCCG CCCGTTCTCCTGGGCAGT
16S (190e208) 16S (1156e1173) 16S (1035e1048)
55 40 30
20 122 56
a b c d
Escherichia coli numbering. Percentage of formamide in the hybridization buffer. Millimolar concentration of sodium chloride in washing buffer. Used with an equimolar amount of unlabeled competitor oligonucleotide cNIT3.
6210
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 0 7 e6 2 1 6
SVI10
140
MLSS
10000 9000
120
6000 5000
60
4000
40
3000
-1
80
MLSS (mg.L )
7000
-1
SVI10 (mL.g )
8000 100
2000 20
1000
0
0 1
25
41
59
73
84
94
104
115
125
135
147
160
172
188
Time (d) Fig. 1 e Change in biomass content and SVI10 during whole operation. SVI, sludge volume index; MLSS, mixed liquid suspended solids.
3.
Results
3.1.
SBR performance and granule characteristics
During the startup period, the reactor removed TOC and NHþ 4N efficiently. 98% of NHþ 4 -N and 100% of TOC were removed from the influent by day 3 and day 5 respectively (Figs. S2, S3, Supporting information). Removal of TN and TP were lower during this period (Figs. S3, S4, Supporting information), though the removal of TP gradually improved to 100% removal by day 33 (Fig. S4, Supporting information). To determine the sludge volume index of granular sludge, a settling time of 10 min was chosen instead of 30 min, because granular sludge has a similar SVI after 60 min and after 5 min of settling (Schwarzenbeck et al., 2004b). The SVI10 of the inoculating sludge was 108.2 mL g1. The changing patterns of MLSS and SVI10 in the continuous operation of the SBR are illustrated in Fig. 1. The sludge settleability increased markedly during the set-up period. Fig. 2 reflects the slow and
gradual process of sludge granulation, i.e., from flocculent sludge to granules.
3.2. DGGE analysis: AOB communities structure changes during sludge granulation The results of nested PCR were shown in Fig. S1. The wellresolved DGGE bands were obtained at the representative points throughout the GSBR operation and the patterns revealed that the structure of the AOB communities was dynamic during sludge granulation and stabilization (Fig. 3). The community structure at the end of experiment was different from that of the initial pattern of the seed sludge. The AOB communities on day 1 showed 40% similarity only to that at the end of the GSBR operation (Table S1, Supporting information), indicating the considerable difference of AOB communities structures between inoculated sludge and granular sludge. Biodiversity based on the DGGE patterns was analyzed by calculating the Shannon diversity index H as
Fig. 2 e Variation in granule size distribution in the sludge during operation. d, particle diameter; TSS, total suspended solids.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 0 7 e6 2 1 6
6211
Fig. 3 e DGGE profile of the AOB communities in the SBR during the sludge granulation process (lane labels along the top show the sampling time (days) from startup of the bioreactor). The major bands were labeled with the numbers (bands 1e15).
shown in Fig.S5. In the phase of sludge inoculation (before day 38), H decreased remarkably (from 0.94 to 0.75) due to the absence of some species in the reactor. Though several dominant species (bands2, 7, 10, 11) in the inoculating sludge were preserved, many bands disappeared or weakened (bands 3, 4, 6, 8, 13, 14, 15). After day 45, the diversity index tended to be stable and showed small fluctuation (from 0.72 to 0.82).
Banding pattern similarity was analyzed by applying UPGMA (Fig. 4) algorithms. The UPGMA analysis showed three groups with intragroup similarity at approximately 67%e78% and intergroup similarity at 44e62%. Generally, the clustering followed the time course; and the algorithms showed a closer clustering of groups II and III. In the analysis, group I was associated with sludge inoculation and washout, group II with
Fig. 4 e UPGMA analysis dendrograms of AOB community DGGE banding patterns, showing schematics of banding patterns. Roman numerals indicate major clusters.
6212
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 0 7 e6 2 1 6
startup sludge granulation and decreasing SVI10, and group III with a stable system and excellent biomass settleability. In Fig. 3, the locations of the predominant bands were excised from the gel. DNA in these bands were reamplified, cloned and sequenced. The comparative analysis of these partial 16S rRNA sequences (Table 2 and Fig. S6) revealed the phylogenetic affiliation of 13 sequences retrieved. The majority of the bacteria in seed sludge grouped with members of Nitrosomonas and Nitrosospira. Along with sludge granulation, most of Nitrosomonas (Bands 2, 5, 7, 9, 10, 11) were remained or eventually became dominant in GSBR; however, all of Nitrosospira (Bands 6, 13, 15) were gradually eliminated from the reactor.
3.3. Distribution of AOB and NOB in different sized granules FISH was performed on the granule sections mainly to determine the location of AOB and NOB within the different size classes of granules, and the images were not further analyzed for quantification of cell counts. As shown in Fig. 6, in small granules (0.2 mm < d < 0.45 mm), AOB located mainly in the outer part of granular space, whereas NOB were detected only in the core of granules. In medium granules (0.45 mm < d < 0.6 mm), AOB distributed evenly throughout the whole granular space, whereas NOB still existed in the inner part. In the larger granules (d > 0.9 mm), AOB and NOB were mostly located in the surface area of the granules, and moreover, NOB became rare.
4.
Discussion
4.1. Relationship between granule formation and reactor performance After day 32, the SVI10 stabilized at 20e35 mL g1, which is very low compared to the values measured for activated sludge (100e150 mL g1). However, the size distribution of the granules measured on day 32 (Fig. 2) indicated that only 22% of the biomass was made of granular sludge with diameter larger
than 0.2 mm. These results suggest that sludge settleability increased prior to granule formation and was not affected by different particle sizes in the sludge during the GSBR operation. It was observed, however, that the diameter of the granules fluctuated over longer durations. The large granules tended to destabilize due to endogenous respiration, and broke into smaller granules that could seed the formation of large granules again. Pochana and Keller reported that physically broken sludge flocs contribute to lower denitrification rates, due to their reduced anoxic zone (Pochana and Keller, 1999). Therefore, TN removal efficiency raises fluctuantly throughout the experiment. Some previous research had demonstrated that bigger, more dense granules favored the enrichment of PAO (Winkler et al., 2011a). Hence, after day 77, removal efficiency of TP was higher and relatively stable because the granules mass fraction was over 90% and more larger granules formed.
4.2. Relationship between AOB communities dynamic and sludge granulation For granule formation, a short settling time was set, and only particles with a settling velocity higher than 4.5 m h1 were retained in the reactor. Moreover, as shown in Fig. 1, the variation in SVI10 was greater before day 41 (from 108.2 mL g1e34.1 mL g1). During this phase, large amounts of biomass could not survive in the reactor. A clear shift in populations was evident, with 58% similarity between days 8 and 18 (Table S1). In the SBR system fed with acetate-based synthetic wastewater, heterotrophic bacteria can produce much larger amounts of extracellular polysaccharides than autotrophic bacteria (Tsuneda et al., 2003). Some researchers found that microorganisms in high shear environments adhered by extracellular polymeric substances (EPS) to resist the damage of suspended cells by environmental forces (Trinet et al., 1991). Additionally, it had been proved that the dominant heterotrophic species in the inoculating sludge were preserved throughout the process in our previous research (Zhang et al., 2011). It is well known that AOB are chemoautotrophic and slow-growing; accordingly, numerous AOB
Table 2 e Species identification of selected DGGE bands (the bands are labeled in Fig. 3).a Bands 2 3 5 6 7 8 9 10 11 12 13 14 15
% Identity
Closest relatives
Origin
100 95 97 96 97 98 93 94 95 98 97 96 98
Uncultured Nitrosomonas sp. (AM418441) Uncultured Nitrosomonas sp. (AY543667) Nitrosomonas sp. Is32 (AJ621027) Uncultured Nitrosospira sp. (GQ255610) Uncultured Nitrosomonas sp. (GQ227644) Uncultured Nitrosomonas sp. (DQ857301) Uncultured Nitrosomonas sp. (DQ887679) Nitrosomonas sp. NM 51 (AF272424) Uncultured bacterium (AB176866) Uncultured bacterium (GQ390307) Uncultured Nitrosospira sp. (GQ255609) Uncultured beta proteobacterium (DQ676335) Uncultured Nitrosospira sp. (GQ325303)
Freshwater submerged macrophytes Biological aerated filter Freshwater sediment Constructed wetland treated with swine wastewater Freshwater Activated sludge Overland flow systems for treatment of landfill leachates Wastewater treatment plants Sewage activated sludge system Enriched nitrifying activated sludge Constructed wetland treated with swine wastewater Suboxic freshwater-pond sediment MBR treated municipal wastewater
a The partial sequences of 16S rRNA genes obtained in the current study have been submitted to the GenBank database under accession numbers GU066569eGU06678 and GU189062eGU189065.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 0 7 e6 2 1 6
6213
Fig. 5 e DGGE profile of the AOB communities in different size of granules (lane labels along the top show the range of particle diameter (d, mm)). Values along the bottom indicate the Shannon diversity index (H ). Bands labeled with the numbers were consistent with the bands in Fig. 3.
populations that cannot become big and dense enough to settle fast were washed out from the system. As a result, the variation in AOB was remarkable in the period of sludge inoculation, and the diversity index of population decreased rapidly. After day 45, AOB communities’ structure became stable due to the improvement of sludge settleability and the retention of more biomass. These results suggest that the short settling time (selection pressure) apparently stressed the biomass, leading to a violent dynamic of AOB communities. Further, these results suggest that certain populations may have been responsible for the operational success of the GSBR and were able to persist despite the large fluctuations in population similarity. This bacterial population instability,
coupled with a generally acceptable bioreactor performance, is congruent with the results obtained from a membrane bioreactor (MBR) for graywater treatment (Stamper et al., 2003). Nitrosomonaselike and Nitrosospiraelike populations are the dominant AOB populations in wastewater treatment systems (Kowalchuk and Stephen, 2001). A few previous studies revealed that the predominant populations in AOB communities are different in various wastewater treatment processes (Tawan et al., 2005; Thomas et al., 2010). Some researchers found that the community was dominated by AOB from the genus Nitrosospira in MBRs (Zhang et al., 2010), whereas Nitrosomonas sp. is the predominant population in biofilter sludge (Yin and Xu, 2009). In the current study,
6214
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 0 7 e6 2 1 6
Fig. 6 e Micrographs of FISH performed on three size classes of granule sections. DAPI stain micrographs (A, D, G); AOB appear as green fluorescence (B, E, H), and NOB appear as red fluorescence (C, F, I). Bar [ 100 mm in (A)e(C) and (G)e(I). d, particle diameter. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
sequence analysis revealed that selection pressure evidently effect on the survival of Nitrosospira in granular sludge. Almost all of Nitrosospira were washed out initially and had no chance to evolve with the environmental changes. However, some members of Nitrosomonas sp. have been shown to produce more amounts of EPS than Nitrosospira, especially under limited ammonia conditions (Stehr et al., 1995); and this feature has also been observed for other members of the same
lineage. Accordingly, these EPS are helpful to communicate cells with each other and granulate sludge (Adav et al., 2008). Therefore, most of Nitrosomonas could adapt to this challenge (to become big and dense enough to settle fast) and were retained in the reactor. At the end of reactor operation (day 180), granules with different particle size were sieved. The effects of variation in granules size on the composition of the AOB communities
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 0 7 e6 2 1 6
were investigated. As shown in Fig. 5, AOB communities structures in different size of granules were varied. Although several predominant bands (bands 2, 5, 11) were present in all samples, only bands 3 and 6 appeared in the granules with diameters larger than 0.6 mm. Additionally, bands 7 and 10 were intense in the granules larger than 0.45 mm. According to Table 2, it can be clearly indicated that Nitrosospira could be retained merely in the granules larger than 0.6 mm. Therefore, Nitrosospira was not present at a high level in Fig. 3 due to the lower proportion of larger granules (d > 0.6 mm) in TSS along with reactor operation. DGGE analysis also revealed that larger granules had a greater microbial diversity than smaller ones. This result also demonstrates that more organisms can survive in larger granules as a result of more space, which can provide the suitable environment for the growth of microbes (Fig. 6).
4.3. Effect of variance in particle size on the distribution of AOB and NOB in granules Although an influence of granule size has been observed in experiments and simulations for simultaneous N- and P-removal (de Kreuk et al., 2007), the effect of granule size on the distribution of different biomass species need be revealed further with the assistance of visible experimental results, especially in the same granular sludge reactors. Related studies on the diversity of bacterial communities in granular sludge often focus on the distribution of important functional bacteria populations in single-size granules (Matsumoto et al., 2010). In the present study, different size granules were sieved, and the distribution patterns of AOB and NOB were explored. In the nitrification processes considered, AOB and NOB compete for space and oxygen in the granules (Volcke et al., 2010). Since ammonium oxidizers have a higher NOB oxygen affinity (KAOB O2 < KO2 ) and accumulate more rapidly in the reactor than nitrite oxidizers (Volcke et al., 2010), NOB are located just below the layer of AOB, where still some oxygen is present and allows ready access to the nitrite produced. In smaller granules, the location boundaries of the both biomass species were distinct due to the limited existence space provided by granules for both microorganism’s growth. AOB exist outside of the granules where oxygen and ammonia are present. Medium granules can provide broader space for microbe multiplying; accordingly, AOB spread out in the whole granules. This result also confirms that oxygen could penetrate deep into the granule’s core without restriction when particle diameter is less than 0.6 mm. Some mathematic model also supposed that NOBs are favored to grow in smaller granules because of the higher fractional aerobic volume (Volcke et al., 2010). As shown in the results of the batch experiments (Zhang et al., 2011), nitrite accumulation temporarily occurred, accompanied by the more large granules (d > 0.9 mm) forming. This phenomenon can be attributed to the increased ammonium surface load associated with larger granules and smaller aerobic volume fraction, resulting in outcompetes of NOB. It also suggests that the core areas of large granules (d > 0.9 mm) could provide anoxic environment for the growth of anaerobic denitrificans (such as Tb. denitrificans or Tb. thioparus in Fig. S7, Supporting information). As shown in Fig. 2 and Fig. S3, the removal efficiency of total nitrogen increased with formation of larger granules.
5.
6215
Conclusions
The variation in AOB communities’ structure was remarkable during sludge inoculation, and the diversity index of population decreased rapidly. Most of Nitrosomonas in the inoculating sludge were retained because of their capability to rapidly adapt to the settlingewashing out action. DGGE analysis also revealed that larger granules had greater AOB diversity than that of smaller ones. Oxygen penetration was not restricted in the granules of less than 0.6 mm particle diameter. However, the larger granules (d > 0.9 mm) can result in the smaller aerobic volume fraction and inhibition of NOB growth. Henceforth, further studies on controlling and optimizing distribution of granule size could be beneficial to the nitrogen removal and expansive application of granular sludge technology.
Acknowledgments This work was supported by grants from the National Natural Science Foundation of China (No. 51108456, 50908227) and the National High Technology Research and Development Program of China (No. 2009AA06Z312).
Appendix. Supplementary data Supplementary data associated with this article can be found in online version at doi:10.1016/j.watres.2011.09.026.
references
Adav, S.S., Lee, D.J., Show, K.Y., 2008. Aerobic granular sludge: recent advances. Biotechnology Advances 26, 411e423. APHA-AWWA-WEF, 2005. Standard Methods for the Examination of Water and Wastewater, first ed. American Public Health Association/American Water Works Association/Water Environment Federation, Washington, DC. de Bruin, L.M., de Kreuk, M., van der Roest, H.F., Uijterlinde, C., van Loosdrecht, M.C.M., 2004. Aerobic granular sludge technology: an alternative to activated sludge? Water Science and Technology 49, 1e7. de Kreuk, M., Heijnen, J.J., van Loosdrecht, M.C.M., 2005. Simultaneous COD, nitrogen, and phosphate removal by aerobic granular sludge. Biotechnology and Bioengineering 90, 761e769. de Kreuk, M., Picioreanu, C., Hosseini, M., Xavier, J.B., van Loosdrecht, M.C.M., 2007. Kinetic model of a granular sludge SBR: influences on nutrient removal. Biotechnology and Bioengineering 97, 801e815. Downing, L.S., Nerenberg, R., 2008. Total nitrogen removal in a hybrid, membrane-aerated activated sludge process. Water Research 42, 3697e3708. Erguder, T.H., Boon, N., Vlaeminck, S.E., Verstraete, W., 2008. Partial nitrification achieved by pulse sulfide doses in a sequential batch reactor. Environmental Science and Technology 42, 8715e8720.
6216
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 0 7 e6 2 1 6
Haseborg, E., Zamora, T.M., Fro¨hlich, J., Frimmel, F.H., 2010. Nitrifying microorganisms in fixed-bed biofilm reactors fed with different nitrite and ammonia concentrations. Bioresource Technology 101, 1701e1706. Kim, D.J., Lee, D.I., Keller, J., 2006. Effect of temperature and free ammonia on nitrification and nitrite accumulation in landfill leachate and analysis of its nitrifying bacterial community by FISH. Bioresource Technology 97, 459e468. Kong, Y., Liu, Y.Q., Tay, J.H., Wong, F.S., Zhu, J., 2009. Aerobic granulation in sequencing batch reactors with different reactor height/diameter ratios. Enzyme and Microbial Technology 45, 379e383. Koops, H.P., Pommerening-Ro¨ser, A., 2001. Distribution and ecophysiology of the nitrifying bacteria emphasizing cultured species. FEMS Microbiology Ecology 37, 1e9. Kowalchuk, G.A., Stephen, J.R., 2001. Ammonia-oxidizing bacteria: a model for molecular microbial ecology. Annual Review of Microbiology 55, 485e529. Laguna, A., Ouattara, A., Gonzalez, R.O., Baron, O., Fama´, G., El Mamouni, R., Guiot, S., Monroy, O., Macarie, H., 1999. A simple and low cost technique for determining the granulometry of upflow anaerobic sludge blanket reactor sludge. Water Science and Technology 40, 1e8. LaPara, T.M., Nakatsu, C.H., Pantea, L.M., Alleman, J.E., 2002. Stability of the bacterial communities supported by a seven stage biological process treating pharmaceutical wastewater as revealed by PCReDGGE. Water Research 36, 638e646. Lemaire, R., Yuan, Z.G., Blackall, L.L., Crocetti, G.R., 2008. Microbial distribution of Accumulibacter spp. and Competibacter spp. in aerobic granules from a lab-scale biological nutrient removal system. Environmental Microbiology 10, 354e363. Li, A., Yang, S., Li, X., Gu, J., 2008. Microbial population dynamics during aerobic sludge granulation at different organic loading rates. Water Research 42, 3552e3560. Li, T., Liu, J., Bai, R., Ohandja, D.G., Wong, F.S., 2007. Biodegradation of organonitriles by adapted activated sludge consortium with acetonitrile-degrading microorganisms. Water Research 41, 3465e3473. Liu, X.W., Sheng, G.P., Yu, H.Q., 2009. Physicochemical characteristics of microbial granules. Biotechnology Advances 27, 1061e1070. Liu, Y.Q., Tay, J.H., 2008. Influence of starvation time on formation and stability of aerobic granules in sequencing batch reactors. Bioresource Technology 99, 980e985. Matsumoto, S., Katoku, M., Saeki, G., Terada, A., Aoi, Y., Tsuneda, S., Picioreanu, C., van Loosdrecht, M.C.M., 2010. Microbial community structure in autotrophic nitrifying granules characterized by experimental and simulation analyses. Environmental Microbiology 12, 192e206. Pochana, K., Keller, J., 1999. Study of factors affecting simultaneous nitrification and denitrification (SND). Water Science and Technology 39, 61e68. Prosser, J.I., Nicol, G.W., 2008. Relative contributions of archaea and bacteria to aerobic ammonia oxidation in the environment. Environmental Microbiology 10, 2931e2941. Sanguinetti, C.J., Dias, N.E., Simpson, A.J., 1994. Rapid silver staining and recovery of PCR products separated on polyacrylamide gels. Biotechniques 17, 914e921. Schwarzenbeck, N., Borges, J.M., Wilderer, P.A., 2005. Treatment of dairy effluents in an aerobic granular sludge sequencing batch reactor. Applied Microbiology and Biotechnology 66, 711e718. Schwarzenbeck, N., Erley, R., Mc Swain, B.S., Wilderer, P.A., Irvine, R.L., 2004a. Treatment of malting wastewater in a granular sludge sequencing batch reactor (SBR). Acta Hydrochimica et Hydrobiologica 32, 16e24.
Schwarzenbeck, N., Erley, R., Wilderer, P.A., 2004b. Aerobic granular sludge in an SBR-system treating wastewater rich in particulate matter. Water Science and Technology 49, 41e46. Shi, X.Y., Sheng, G.P., Li, X.Y., Yu, H.Q., 2010. Operation of a sequencing batch reactor for cultivating autotrophic nitrifying granules. Bioresource Technology 101, 2960e2964. Stamper, D.M., Walch, M., Jacobs, R.N., 2003. Bacterial population changes in a membrane bioreactor for graywater treatment monitored by denaturing gradient gel electrophoretic analysis of 16S rRNA gene fragments. Applied and Environmental Microbiology 69, 852e860. Stehr, G., Zo¨rner, S., Bo¨ttcher, B., Koops, H.P., 1995. Exopolymers: an ecological characteristic of a floc-attached, ammoniaoxidizing bacterium. Microbial Ecology 30, 115e126. Tawan, L., Yuko, S., Futoshi, K., Osami, Y., 2005. Communities of ammonia-oxidizing bacteria in activated sludge of various sewage treatment plants in Tokyo. FEMS Microbiology Ecology 54, 205e217. Thomas, F.D., Matias, B.V., Anthony, D.S., Ariel, A.S., Aprel, Q.E., 2010. Characterization of a microbial community capable of nitrification at cold temperature. Bioresource Technology 101, 491e500. Trinet, F., Heim, R., Amar, D., Chang, H.T., Rittmann, B.E., 1991. Study of biofilm and fluidization of bioparticles in a threephase fluidized-bed reactor. Water Science and Technology 23, 1347e1354. Tsuneda, S., Nagano, T., Hoshino, T., Ejiri, Y., Noda, N., Hirata, A., 2003. Characterization of nitrifying granules produced in an aerobic upflow fluidized bed reactor. Water Research 37, 4965e4973. Vlaeminck, S.E., Terada, A., Smets, B.F., Clippeleir, H.D., Schaubroeck, T., Bolca, S., Demeestere, L., Mast, J., Boon, N., Carballa, M., Verstraete, W., 2010. Aggregate size and architecture determine microbial activity balance for onestage partial nitritation and anammox. Applied and Environmental Microbiology 76, 900e909. Volcke, E.I.P., Picioreanu, C., de Baets, B., van Loosdrecht, M.C.M., 2010. Effect of granule size on autotrophic nitrogen removal in a granular sludge reactor. Environmental Technology 31, 1271e1280. Wang, Q., Du, G.C., Chen, J., 2004. Aerobic granular sludge cultivated under the selective pressure as a driving force. Process Biochemistry 39, 557e563. Winkler, M.K.H., Bassin, J.P., Kleerebezem, R., de Bruin, L.M.M., van den Brand, T.P.H., van Loosdrecht, M.C.M., 2011a. Selective sludge removal in a segregated aerobic granular biomass system as a strategy to control PAO-GAO competition at high temperatures. Water Research 45, 3291e3299. Winkler, M.K.H., Kleerebezem, R., Gijs Kuenen, J., Yang, J., van Loosdrecht, M.C.M., 2011b. Segregation of biomass in cyclic anaerobic/aerobic granular sludge allows the enrichment of anaerobic ammonium oxidizing bacteria at low temperatures. Environmental Science and Technology 45, 7330e7337. Xavier, J.B., de Kreuk, M., Picioreanu, C., van Loosdrecht, M.C.M., 2007. Multi-scale individual-based model of microbial and bioconversion dynamics in aerobic granular sludge. Environmental Science and Technology 41, 6410e6417. Yin, J., Xu, W., 2009. Ammonia biofiltration and community analysis of ammonia-oxidizing bacteria in biofilters. Bioresource Technology 100, 3869e3876. Zhang, B., Ji, M., Qiu, Z.G., Liu, H.N., Wang, J.F., Li, J.W., 2011. Microbial population dynamics during sludge granulation in an anaerobic-aerobic biological phosphorus removal system. Bioresource Technology 102, 2474e2480. Zhang, B., Sun, B.S., Ji, M., Liu, H.N., Liu, X.H., 2010. Quantification and comparison of ammonia-oxidizing bacterial communities in MBRs treating various types of wastewater. Bioresource Technology 101, 3054e3059.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 1 7 e6 2 2 6
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Altering the characteristics of a leaf litter-derived humic substance by adsorptive fractionation versus simulated solar irradiation Jin Hur a,**, Ka-Young Jung a, Mark A. Schlautman b,* a b
Department of Environment and Energy, Sejong University, Seoul 143-747, South Korea Department of Environmental Engineering and Earth Sciences, Clemson University, 342 Computer Court, Anderson, SC 29625-6510, USA
article info
abstract
Article history:
Changes in the characteristics of a leaf litter-derived humic substance (LLHS) that resulted
Received 15 March 2011
from its adsorption onto kaolinite or exposure to simulated solar irradiation were tracked
Received in revised form
using selected spectroscopic descriptors, apparent weight-average molecular weight
5 September 2011
(MWw) and pyrene binding. Heterogeneity within the original bulk LLHS was confirmed by
Accepted 10 September 2011
a range of different characteristics obtained from ultrafiltration-based size fractions. In
Available online 21 September 2011
general, trends of some changing LLHS characteristics were similar for the adsorption and irradiation processes when tracked against percent carbon removal. For example, the
Keywords:
overall values of specific ultraviolet absorbance (SUVA), MWw, and humification index
Adsorption
(HIX) all decreased with increasing irradiation time and with increasing concentration of
Photodegradation
mineral adsorbent in the respective experiments, indicating that both processes resulted in
Ultrafiltration
less aromatic and smaller-sized LLHS components remaining in solution. In addition, both
Size exclusion chromatography
the adsorption and irradiation experiments resulted in enrichment of the relative distri-
Fluorescence
bution of protein-like fluorescence (PLF), implying the PLF-related components had low
UVevis absorbance
affinities for phototransformation and mineral surface adsorption. Despite these apparently similar overall trends in LLHS characteristics caused by the adsorption and irradiation processes, closer examination revealed considerable differences in how the two processes altered the original material. Net production of intermediate-sized constituents was observed only with the irradiation experiments. In addition, residual LLHS resulting from the adsorptive fractionation experiments exhibited consistently higher pyrene binding versus the irradiated LLHS despite having comparable MWw values. Changes in LLHS characteristics due to adsorption by kaolinite were likely caused by physical mechanisms (primarily hydrophobic interactions between LLHS components and the kaolinite surface) whereas the irradiation-induced changes appear to have been governed by the combined effects of several alteration mechanisms, including the transformation of more condensed aromatic structures to less aromatic constituents, conformational changes resulting from selective photooxidation, and the photochemical disruption of intramolecular charge-transfer interactions. ª 2011 Elsevier Ltd. All rights reserved.
* Corresponding author. Tel.: þ1 864 656 4059; fax: þ1 864 656 0672. ** Corresponding author. Tel.: þ82 2 3408 3826; fax: þ82 2 3408 4320. E-mail addresses: [email protected] (J. Hur), [email protected] (M.A. Schlautman). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.09.023
6218
1.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 1 7 e6 2 2 6
Introduction
Humic substances (HS) are heterogeneous mixtures of organic compounds with different molecular sizes and functional groups. In natural waters, HS constitute the major portion of terrestrially-derived dissolved organic matter (DOM) (Thurman, 1985). The presence of HS in aquatic ecosystems affects the transport, bioavailability, toxicity, and ultimate fate of hydrophobic organic contaminants (HOCs). The extent of HOC binding to HS is generally represented by organic carbon-normalized partition coefficient (Koc). Selected physicochemical properties of HS are known to be critical factors in determining Koc values. For example, it has been reported that higher Koc values are associated with higher molecular weight HS and with more aromatic and/or aliphatic carbon content (Chin et al., 1997; Chefetz et al., 2000; Hur and Schlautman, 2003a). Recently, several parameters derived from fluorescence spectroscopy such as a humification index (HIX) have shown positive correlations with the extent of HOC partitioning to HS (Sun et al., 2007; Hur and Kim, 2009). Terrestrial HS are often considered to be biologically recalcitrant, but other geochemical and photochemical processes may substantially alter their properties and functionalities (Steinburg, 2003). Consequently, the concentrations and characteristics of HS typically encountered in natural waters are likely the outcome of not only physical mixing of different individual HS sources but also their modification by various natural processes such as photodegradation, adsorption onto mineral surfaces, and humification from labile organic substances (Meier et al., 2004; Pullin et al., 2004a; Asakawa et al., 2007). Simultaneous occurrence of multiple natural processes is common in nature and thus the subsequent alternation of HS may become more complicated due to its inherent structural and chemical heterogeneity. For example, preferential adsorption of higher MW fractions onto mineral surfaces tends to be greater for more heterogeneous HS (Hur and Schlautman, 2003b). In addition, a recent study suggests that HS composed of lower MW acids and more neutral compounds is less vulnerable to photochemical reactions (Liu et al., 2010). Environmental effects of photooxidation of HS may differ by its source and composition. For example, microbial utilization of HS enhanced by photodegradation was more evident for terrestrial HS which contained more aromatic structures than it was for algal derived HS (Obernosterer and Benner, 2004). Adsorption by minerals and exposure to solar irradiation are two natural processes that can lead to substantial changes in the properties of terrestrial HS. Previous studies have shown that adsorption is a key process governing seasonal variations of HS in forest organic soil horizons whereas solar irradiation becomes more important after the HS is leached from terrestrial environments and enters adjacent streams (Qualls and Richardson, 2003; Asakawa et al., 2007). Under some circumstances in aquatic environments, however, such as freshwaters and estuaries with high loads of suspended sediment, the contribution of mineral adsorption for altering HS characteristics may become comparable to that of solar irradiation (Shank et al., 2005). It is intriguing that these two very different processes might possibly lead to changes in HS that appear similar, for example, when the resultant HS
materials are characterized by correlations between MW and the Koc values. It has been reported that average MW values of residual dissolved HS and its subsequent ability to bind HOCs both tend to decrease upon contact with mineral surfaces (Hur and Schlautman, 2003b, 2004). However, similar decreases in MW and HOC partitioning have been reported for irradiated HS (Lou and Xie, 2006; Lou et al., 2006). For both processes, positive correlations existed between the extent of HOC binding and apparent MW value (Hur and Schlautman, 2004; Lou et al., 2006). Therefore, such similar HS responses to the different processes may make it difficult to distinguish which process was the primary driver for HS alterations. Thus far, little effort has been made to rigorously compare the changes to HS caused by different individual natural processes. Furthermore, it remains questionable whether the correlations among the characteristics of HS resulting from the different processes are, in fact, similar. The objectives of this study were to: 1) investigate the heterogeneity within a leaf litter-derived HS (LLHS) based on its different size fractions, 2) examine the changes in selected operational descriptors of HS brought about by mineral adsorption versus simulated solar irradiation, and 3) compare correlations between selected HS characteristics and the extent of HOC binding by HS as affected by the two different processes. Leaf litter is a major source of DOM in streams via leaching from particulate matter and/or through fall from forested areas (Meyer et al., 1998).
2.
Experimental section
2.1. Preparation of leaf litter-derived HS and size fractionation by ultrafiltration (UF) Fallen leaves were collected from several locations in the upstream Han River basin in Korea to represent typical allochthonous HS in forested areas (Hur, 2011). The dominant types of the leaves were Quercus (white oak) species and Robinia pseudoacacia Linnaeus (black locust). The partially decomposed leaves were air-dried and shredded to smaller sizes. Water soluble extracts were prepared by mixing the shredded leaves with distilled, deionized water (DDW) at a solid-to-solution mass ratio of 1:10 for 24 h. The extracts were filtered through a precleaned 0.2-mm pore size membrane filter (cellulose acetate, Advantec) to remove particulate matter and microorganisms, acidified to pH 2 with concentrated HCl, and then passed through a DAX-8 resin (Supelco, SigmaeAldrich) column. The HS fraction retained on the resin was subsequently eluted with 0.1 N NaOH and further purified by passing through a cation-exchange resin (Dowex 50WX8-100, Sigma). The final ionic strength of the LLHS solution was adjusted to 0.1 M by adding an appropriate amount of NaCl. Size fractionation of the LLHS was performed using an Amicon batch stirred cell (250 mL capacity) and ultrafiltration (UF) membranes with four different molecular weight cutoffs (YM1, YM10, TM30, YM100) to obtain five different size fractions (<1 K, 1e10 K, 10e30 K, 30e100 K, >100 kDa). Additional details for the UF procedure have been previously reported (Hur and Kim, 2009). Loss of HS during the UF procedure
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 1 7 e6 2 2 6
amounted to less than 5% based on the mass balance of organic carbon.
2.2.
Mineral adsorption experiments
Commercially available kaolinite (SigmaeAldrich) was used as a representative mineral for this study (Hur and Schlautman, 2003b). Adsorption experiments were conducted by mixing different dosages of the adsorbent (20, 40, 60, 150, 250 g/L) with a constant initial LLHS solution (56 mg C/L) so that a wide range of residual concentrations and thus removal percentages could be obtained. Solution pH was adjusted to 6.0 with HCl or NaOH. All samples were then placed on a reciprocating shaker at room temperature (21 2 C) to equilibrate for 72 h. This contact time was selected to enable the adsorptive fractionation kinetics to reach a pseudo-equilibrium state (Hur and Schlautman, 2003b). The mineral particles were then separated from solution by centrifugation at 5000 rpm for 30 min and aliquots of the supernatant taken for analysis. Typical adsorption results are shown in Fig. 1. Percent removal of HS was calculated by mass balance based on the initial and final organic carbon concentrations. No significant system losses were observed based on control samples that contained the same LLHS solution without kaolinite.
2.3.
Simulated solar irradiation experiments
A diluted LLHS solution (23 mg C/L) was prepared from the original stock solution by adding 0.1 M NaCl. The same ionic
Residual DOC (mg C/L)
a
60 50 40 30 20 10 0 0
50
100
150
200
250
300
Kaolinite concentration (g/L)
Residual DOC (mg C/L)
b
25 20 15 10 5 0 0
2
4
6
8
10
12
14
Irradiation time (days) Fig. 1 e Effects of increasing kaolinite concentration (a) and increasing irradiation time (b) on the DOC concentration of residual LLHS.
6219
strength was used as described in the previous adsorption experiments to compare the results under the same solution composition. The kinetic rate of a photochemical reaction tends to increase at a higher ionic strength due to changing solvent cage effects (Clark et al., 2007). The diluted solution was then transferred to 18 quartz tubes (70 mL, 3 cm diameter) that were sealed with Teflon caps. The tubes were placed vertically at a distance of 10 cm from an array of three UVA340 lamps (Q-Panel). Output from the lamp has a spectral shape similar to natural sunlight from 295 to 365 nm (Helms et al., 2008) with an intensity (10 W/m2) equivalent to sunlight at noon during the summer in the Han River basin (137 290 N). The sunlight intensity was calculated over the wavelengths ranging from 320 nm to 400 nm. Prior to irradiation, all samples were bubbled with pure oxygen. Two or three tubes were randomly chosen and removed for analyses at the selected irradiation times (0.4, 1, 2, 3, 6, and 12 days). Sample pH was monitored throughout the experiment and decreased from 6.5 to 6.2 over the 12 days. Typical irradiation results are shown in Fig. 1.
2.4.
Analytical methods
Samples from UF, mineral adsorption, and irradiation experiments were filtered through a 0.2-mm pore size membrane filter and adjusted to pH 6 prior to analyses. Concentrations of HS in all samples were determined by measuring dissolved organic carbon (DOC) on acidified, air-sparged samples using a Shimadzu V-CPH analyzer. The relative precision of DOC measurements was <3% based on repeated measurements. Ultravioletevisible absorption spectra were measured at 1-nm increments over the wavelength range 200e600 nm with a spectrophotometer (Evolution 60, Thermo Scientific). The same absorption spectra were also used in applying innerfilter corrections for pyrene binding (Gauthier et al., 1986) as described below. The SUVA values of the samples were determined by dividing 100-fold of the UV absorbance at 254 nm by the DOC concentration. Synchronous fluorescence spectra of samples adjusted to pH 3 were recorded with a luminescence spectrometer (PerkineElmer LS-50B) with excitation and emission slits both set at 10 nm. The excitation wavelengths ranged from 250 to 600 nm and emission was recorded at a constant offset (Dl) of 30 nm. All samples were diluted to less than 4 mg C L1, and an innerfilter correction was used to account for absorption of incident and emitted light by HS (Hur et al., 2009). To limit second-order Raleigh scattering, a 290-nm cutoff filter was used for all samples. The fluorescence response to a LLHS control solution (0.1 M NaCl solution without the kaolinite contact) was subtracted from the spectrum of each sample. Finally, fluorescence intensities of all samples were normalized to units of quinine sulfate equivalents (QSE) based on fluorescence measured from a series of diluted quinine sulfate dehydrate solutions in 0.05 M sulfuric acid at the excitation/emission wavelengths of 350/ 450 nm (Chen et al., 2007). For this study, three relative fluorescence regions were assigned to protein-like (% PLF), fulviclike (% FLF), and humic-like (% HLF) fluorescence, each of which corresponds to the relative percentage of fluorescence intensity at the wavelengths of 250e300 nm, 300e380 nm, and 380e600 nm, respectively (Hur, 2011).
6220
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 1 7 e6 2 2 6
Size exclusion chromatography (SEC) was used to determine apparent MWw values of all HS samples following the methodology reported by Hur and Schlautman (2003a). Briefly, a high performance liquid chromatograph (HPLC, Waters model 590) with a UV detector (Waters 486 detector) and a Protein-Pak 125 column for size separation was used. All standards and samples were measured at a detection wavelength of 254 nm. The flow rate of the mobile phase was maintained at 1 mL/min. The analytical conditions, the choice of the standards, and calculation of MW values are further described in Hur and Schlautman (2003a). To estimate pyrene Koc values, the modified fluorescence quenching (FQ) technique reported by Hur and Schlautman (2003a) was used due to the limited sample concentrations and volumes available. Pyrene binding was evaluated at a constant DOC concentration of 12 mg C/L for all samples. Further details are provided in Hur and Schlautman (2003a) and Hur et al. (2009).
3.
Results and discussion
3.1. Comparison of operational descriptors for different LLHS UF fractions SUVA values ranged from 1.15 to 3.88 L/mg C-m among the different UF fractions of LLHS (Table 1). Relative to the bulk SUVA value of 3.13, this range was 63% to þ24% and reflected the heterogeneous distribution of UV-absorbing moieties including aromatic groups within the bulk material. Although UV-absorbing structures exhibited a generally increasing trend with size overall for the UF fractions, the correlation between SUVA and MWw was not statistically significant (r ¼ 0.688, p ¼ 0.199). The highest SUVA value was found for the 30e100 K fraction rather than the largest (>100 kDa) UF fraction. The observed distribution pattern of SUVA values with size of LLHS fractions is similar to HS extracted from sediments and composts (Hur and Kim, 2009; Canellas et al., 2010). Fluorescence index (FI), the ratio of fluorescence emission intensities at 450 and 500 nm (F450/F500) for an excitation wavelength of 370 nm, has been utilized previously to distinguish between allochthonous and autochthonous HS (McKnight et al., 2001). For the LLHS UF fractions, FI values
ranged from 1.26 to 1.57 (Table 1). Relative to the bulk FI value of 1.40, this range was only 10% to þ12% and thus indicates that the distribution of fluorescent moieties excited at 370 nm was more uniform with size than was the aromatic moieties tracked by SUVA. Interestingly, there was a significant negative correlation (r ¼ 0.924, p ¼ 0.025) between FI and SUVA values for the UF fractions of LLHS. The humification index (HIX) has been defined as the relative ratio of summed fluorescence intensities over two different wavelength ranges (300e345 nm and 435e480 nm) when using an excitation wavelength of 254 nm (Zsolnay et al., 1999). It has been reported that a higher HIX value is associated with the presence of more aromatic condensed structures and/or an increased degree of conjugation in unsaturated aliphatic chains (Fuentes et al., 2006). For the different UF fractions of LLHS, there was a positive correlation between HIX and SUVA values but it was not statistically significant (r ¼ 0.744, p ¼ 0.149). From Table 1, it can be seen that the distribution of HIX among the different UF fractions does not consistently match the SUVA trends (e.g., highest HIX value not observed with the 30e100 K fraction having the highest SUVA value). The lack of a better correlation between HIX and SUVA likely indicates the heterogeneous distribution of UV-absorbing moieties that are not fluorescent, such as unsaturated aliphatic carbon structures. LLHS constituents contributing to the % PLF were preferentially enriched in the two smallest UF fractions (i.e., <1 K and 1e10 K) and again in the largest UF fraction (i.e., >100 K) (Table 1). Because of this distribution among the UF fractions, there was a strong correlation (r ¼ 0.932, p ¼ 0.021) between % PLF and FI. The nonlinear distribution of % PLF with size may originate from two different sources of the PLF-related LLHS structures. For example, Maie et al. (2007) observed two separate PLF-like peaks of mangrove-derived DOM from a SEC chromatogram using fluorescence detection. Because the PLF peak intensities did not have a strong positive correlation with organic nitrogen concentrations, Maie et al. proposed that low MW-associated PLF originated from the presence of tanninlike polyphenolic moieties within the DOM and not from aromatic amino acid-like compounds. Consistent with the SUVA measurements, the highest % HLF value was found for the 30e100 K UF fraction. Overall, % HLF tracked very well with SUVA measurements across the UF fractions (r ¼ 0.971, p ¼ 0.006).
Table 1 e Spectroscopic descriptors, weight-average molecular weight, and pyrene Koc values of the original bulk LLHS and its UF size fractions. LLHS Original >100 K 30e100 K 10e30 K 1e10 K <1 K
Distribution (%)
SUVA254a (L/mg C-m)
100.0 27.0 19.4 16.2 17.0 20.5
3.13 (0.01) 3.63 (0.07) 3.88 (0.08) 3.78 (0.10) 2.41 (0.07) 1.15 (0.03)
Fluores. Indexa (FI) 1.40 1.41 1.26 1.34 1.46 1.57
(0.04) (0.04) (0.04) (0.04) (0.04) (0.04)
HIXa
MWw
5.68 (0.16) 6.51 (0.18) 9.90 (0.28) 15.84 (0.45) 8.41 (0.23) 2.12 (0.06)
2127 4557 2846 1445 725 474
%PLFa 28.4 13.7 0.1 0.1 21.5 61.2
(0.8) (0.4) (0.0) (0.0) (0.6) (1.7)
%HLFa 34.6 46.1 64.4 58.4 33.3 11.3
(1.0) (1.3) (1.8) (1.7) (0.9) (0.3)
Kocb (mL/g C) (103) 22.9 32.4 31.1 40.1 13.6 5.9
(0.4) (2.0) (1.1) (1.7) (1.8) (0.1)
a Numbers in parentheses are standard errors based on propagating corresponding value uncertainties. Uncertainties in UV absorption and fluorescence were based on the precision (<2%) for each spectrometer. b Numbers in parentheses are standard errors based on triplicate sample measurements.
6221
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 1 7 e6 2 2 6
As expected from the heterogeneous character of LLHS, pyrene partitioning also showed an appreciable variation among the LLHS UF fractions. The highest Koc value observed was with the intermediate-sized fraction 10e30 K and was approximately 7 times higher than that of the smallest UF fraction. The correlation between Koc and MWw among the LLHS UF fractions was not statistically significant (r ¼ 0.614, p ¼ 0.271). However, the trend between Koc and SUVA values was statistically significant (r ¼ 0.950, p ¼ 0.013). The wide range of Koc values among the UF fractions implies that HOC binding by the LLHS can change greatly depending on any fractionation or selective removal processes.
3.2. Changes in operational descriptors of LLHS due to mineral adsorption Adsorption of LLHS with increasing concentrations of kaolinite resulted in removals of DOC up to 63% (Fig. 1a). Electrostatic attraction, hydrophobic interactions and ligand exchange have been suggested as major controlling factors for HS adsorption to minerals (Hur and Schlautman, 2003b). For LLHS, however, electrostatic attraction does not appear to be a predominant mechanism because the equilibrium pH (6.0) was much higher than the point of zero charge (PZC) of kaolinite (w4.2) (Hur and Schlautman, 2003b). In addition, adsorption on kaolinite via ligand exchange is expected to occur only near the edges of the basal plane where surface hydroxyl groups are concentrated (Murphy et al., 1990). Therefore, hydrophobic interactions most likely are responsible for adsorption of LLHS in the present study, whereby larger and/or more hydrophobic constituents
2000
3.0
1500
2.5
1000
2.0
500
b 1.7
1.5
MWw
50
1.6 40 %PLF
3.5
Fluorescence index (FI)
2500
SUVA (L/mg C-m)
MW w (g/mol as PSS)
a
tend to be preferentially adsorbed by the mineral (Hur and Schlautman, 2003b, 2004). Hydrophobic interactions also have been suggested as a major adsorption process for HS on natural adsorbents such as minerals and biosolids at neutral pH conditions (Esparza-Soto and Westerhoff, 2003). It is also possible that some contribution to the adsorption of LLHS resulted from cation bridging, brought about by the high concentration (0.1 M) of NaCl used in the experiments. Cation bridging is a particularly important contributor to HS adsorption for mineral surfaces (e.g., kaolinite) bearing charged siloxane ditrigonal cavities and monovalent exchangeable cations (Schlautman and Morgan, 1994). In the present study MWw values of the LLHS remaining in solution after adsorption exhibited a consistent decreasing trend with increasing DOC removal, varying from 2127 to 696 Da (Fig. 2a). This result indicates that the higher MW constituents within the bulk LLHS were preferentially adsorbed by kaolinite. SEC chromatograms confirmed the preferential adsorptive removal of higher MW components (Fig. 3a). In addition, as the percent DOC removal increased, the SEC chromatograms revealed that progressively more of the lower MW constituents were removed from solution. These findings are consistent with adsorption being driven by hydrophobic interactions. When available mineral surfaces are a limiting factor (i.e., smaller DOC removal), competition among the different sized components within the bulk LLHS becomes stronger and the larger constituents are preferentially adsorbed. SUVA values of the residual LLHS in solution decreased with increasing percent DOC removal (Fig. 2a). For relatively low DOC removal ranges (i.e., <45%), however, the changes in SUVA were
1.5 30 1.4 FI %PLF
SUVA 0
1.0 0
10
20
30
40
50
60
1.3
70
20 0
10
20
Percent carbon removal (%)
c
30
40
50
60
70
Percent carbon removal (%)
8
d
30
2.0
3.0
4.0
HIX
5.0
6.0
7.0
HIX
20 4 10 2
Koc x10-3 (mL/g C)
6
K oc x10-3 (mL/g C)
30
20
10 MWw
HIX
HIX
Koc 0
0 0
10
20
30
40
50
Percent carbon removal (%)
60
70
0 500
1000
1500
2000
MW w (mol/g as PSS)
Fig. 2 e Changes in the characteristics of residual LLHS as a function of percent DOC removal due to adsorption by kaolinite (pH 6.0 and 0.1 M NaCl). (a) MWw and SUVA values. (b) Fluorescence index (FI) and % PLF values. (c) HIX and pyrene Koc values. (d) Trends of pyrene Koc with MWw and HIX.
6222
a
100
Removal of UV-absorbing components (%)
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 1 7 e6 2 2 6
80 60 40 20 Increasing carbon removal by adsorption
0 -20 2.0
2.5
3.0
3.5
4.0
4.5
b
100
Removal of UV-absorbing components (%)
Log MW
80 60 40 20 Increasing irrradiation times
0 -20 2.0
2.5
3.0
3.5
4.0
4.5
misinterpret an allochthonous HS undergoing appreciable adsorptive fractionation as being an autochthonous HS. Individual relationships between the parameters previously described and the Koc values of the residual LLHS after adsorptive fractionation were all statistically significant ( p < 0.005) (data not shown). Positive correlations of Koc with SUVA, HIX and MWw (all r > 0.85) suggest that the LLHS components associated with humified UV-absorbing structures and larger sizes are primarily involved in pyrene binding. In contrast, increasing % PLF and FI are negatively correlated (both r < 0.85) with pyrene partitioning (i.e., % PLF versus Koc values and FI versus Koc values for adsorptive fractionated LLHS). These general correlation trends between the various operational descriptors and Koc values were similar to those trends observed with the UF fractions, suggesting that the contributions of the particular HS spectroscopic and chromatographic characteristics to enhancing or reducing HOC binding do not fundamentally change upon contact with the kaolinite surface. This observation is consistent with our previous report that showed similar pyrene KoceMWw regression lines for a commercial HS undergoing mineral surface adsorptive fractionation and UF fractionation (Hur and Schlautman, 2004).
3.3. Changes in operational descriptors of LLHS due to simulated solar irradiation
Log MW
Fig. 3 e Removal of UV254-absorbing components as a function of log MW for the LLHS samples from kaolinite adsorption experiments (a) and from simulated solar irradiation experiments (b). The percent UV removal was calculated by difference using the initial LLHS SEC detection signal and the signal for the residual LLHS in solution.
less dramatic. For example, only a 13% difference was observed between SUVA values at 43% DOC removal and the original bulk LLHS value. Such small differences in SUVA over this range of low DOC removal percentages can be explained by the preferential adsorption of higher MW constituents of LLHS having similar SUVA values. For example, the three largest UF fractions of LLHS had relatively similar SUVA values, ranging only from 3.63 to 3.88 L/mg C-m (Table 1). Increasing LLHS adsorption to kaolinite resulted in increasing FI and % PLF values while HIX values decreased (Fig. 2b and c). As expected from the UF results discussed earlier, the changes in FI and % PLF exhibited similar patterns whereas the HIX changes mirrored those for SUVA values. For all four descriptors, the changes became more pronounced at the higher percentages of DOC removal (Fig. 2a, b and c). The increase in % PLF indicates that fulvic- and humic-like fluorescent components tend to be associated more with the higher MW and/or more hydrophobic constituents that were preferentially adsorbed on kaolinite (Maie et al., 2007; Hur and Kim, 2009). Fig. 2b indicates that caution must be exercised when using FI to distinguish between allochthonous and autochthonous HS. Based on the increasing FI values with percent DOC removal, one can see that it is possible to
Except for FI and % PLF, general trends in the changes of selected descriptors of LLHS with irradiation time were relatively similar to those caused by adsorptive fractionation. For example, irradiation over the 12 day period resulted in considerable reductions in SUVA and MWw values (Fig. 4a). Although loss of DOC reached only w30% over this time period, SUVA and MWw decreased w60% and w70%, respectively. The decreasing SUVA values likely indicate oxidative cleavage of aromatic carbon structure during irradiation, possibly induced through direct photolysis (Xie et al., 2004; Sulzberger and Durisch-Kaiser, 2009; Thorn et al., 2010) and/ or by indirect pathways utilizing reactive oxygen species (Lou and Xie, 2006). Major constituents of the products resulting from photo-irradiation often include non UV-absorbing species such as low MW organic acids, alcohols, aldehydes, and inorganic carbon (Pullin et al., 2004b; Vidali et al., 2010). In contrast, aliphatic carbon structures within HS are reported to be resistant to photodegradation although they can react with photochemically generated hydroxyl radicals (Thorn et al., 2010). The large reduction in MWw resulting from photoirradiation may be attributed to preferential photooxidation of the larger LLHS constituents enriched with UV-absorbing moieties (i.e., higher SUVA values) and/or the reactions between LLHS molecules and the reactive oxygen species produced in the oxygen-saturated solution (Lou and Xie, 2006). Similar results of decreasing MW with photo-irradiation of HS have been reported by other investigators (Lou and Xie, 2006; Vidali et al., 2010). The SEC chromatograms revealed that removal percentages were higher for the larger components and that with increasing irradiation time removal efficiencies increased for the intermediate MW components (Fig. 3b). For example, at the shortest irradiation time (0.4 day), more than
6223
3.0 2000 2.5 2.0 1000 1.5
MWw
b 1.5
40
30
1.4
1.3 FI
SUVA
%PLF
0
1.0 0
10
20
30
1.2
40
20 0
5
10
Percent carbon removal (%)
c
%PLF
3.5
SUVA (L/mg C-m)
MW w (g/mol as PSS)
a 3000
Fluorescence index (FI)
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 1 7 e6 2 2 6
15
20
25
30
35
Percent carbon removal (%)
6
d
30
2.0
3.0
4.0
HIX
5.0
6.0
7.0
2
10
Koc x10-3 (mL/g C)
20
HIX
4
Koc x10-3 (mL/g C)
30
20
10 MWw
HIX
HIX
Koc 0
0 0
5
10
15
20
25
30
35
Percent carbon removal (%)
0 0
500
1000
1500
2000
2500
3000
MW w (mol/g as PSS)
Fig. 4 e Changes in the characteristics of residual LLHS as a function of percent DOC removal due to simulated solar irradiation (pH 6.2e6.5 and 0.1 M NaCl). (a) MWw and SUVA values. (b) Fluorescence index (FI) and % PLF values. (c) HIX and pyrene Koc values. (d) Trends of pyrene Koc with MWw and HIX.
80% removal of the detection signal was observed only for the LLHS constituents larger than MW w15.8 kDa. By 12 days, however, this equivalent level of removal extended down to MW w1.3 kDa. At several irradiation times the SEC chromatograms also revealed (net) production of intermediateand smaller-sized molecules that could absorb energy at 254 nm and thus were detected by our SEC system (Fig. 3b). This observation may be due to partial oxidation of higher MW LLHS constituents and the subsequent production of the lower MW materials, perhaps with fewer condensed aromatic structures (Lou and Xie, 2006; Carvalho et al., 2008). Except for the 12 day samples (i.e., DOC removal of w30%), FI values showed a generally decreasing trend with irradiation time and DOC removal although there was appreciable scatter in the data (Fig. 4b). Reduction in FI values upon photoirradiation has been reported previously for a surface water DOM (Brooks et al., 2007). Conversely, the % PLF values initially increased dramatically and then remained more or less stable thereafter (Fig. 4b). The latter observation indicates that fulvicand humic-like fluorescent moieties were initially degraded to a greater extent than the PLF-related structures. However, as the irradiation progressed fluorescence intensities diminished across the entire wavelength spectrum (data not shown). Previous study results on the photobleaching of particular fluorescent structures and/or wavelengths have been mixed and appear to depend on the particular DOM type studied. For example, more photobleaching of humic-like versus proteinlike fluorophores was reported for a surface water DOM (Stedmon et al., 2007) whereas the opposite trend was observed
for a DOM that contained substantial tannin-like properties (Carvalho et al., 2008; Shank et al., 2010). The energy spectrum of the photo-irradiation source may also affect the wavelength range of photobleached fluorophores because photolysis occurs primarily for molecular bonds having the same energy as the irradiated light (Del Vecchio and Blough, 2002, 2004). HIX values exhibited an initial rapid decrease followed by less change with increasing irradiation time and percent DOC removal (Fig. 4c). One possible explanation is that these changes in HIX result from more photobleaching of the condensed polyaromatic structures associated with longer wavelengths. However, this would not be consistent with the previous observation of little change in MWw upon initial irradiation nor with the general decreasing MWw trend with longer irradiation time (Fig. 4a). An alternative explanation may be that the initial reduction in HIX is initiated by the destruction of intramolecular donoreacceptor (DeA) interactions, which are more likely to be present in the higher MW components. For example, Del Vecchio and Blough (2004) proposed that fluorescence emission at longer wavelengths (>350 nm) are based on a continuum of coupled states resulting from intramolecular chargeetransfer interactions between polyhydroxylated aromatic donors and quinone-like oxidized acceptors. Previous studies have concluded that the photo-induced variations of fluorescence-based HS descriptors may result from the combined effects of preferential photolysis of molecular bonds having the same energy as the incident light (Del Vecchio and Blough, 2002, 2004), conformational changes
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 1 7 e6 2 2 6
3.4. Comparison of operational descriptors and pyrene binding trends with SUVA and MWw for adsorptive fractionated versus photo-irradiated LLHS In general, responses of the operational descriptors of LLHS to the adsorption and the irradiation processes were similar. For both cases, SUVA, MWw, HIX, and Koc values of LLHS all decreased with percent carbon removal, indicating that more aromatic and larger-sized LLHS components with higher affinity for pyrene were preferentially removed by both processes. In contrast, % PLF tends to increase with increasing irradiation time and increasing concentration of mineral adsorbent, suggesting that PLF-related components were relatively enriched in the remaining solutions after both processes. Values of SUVA have been shown to be positively correlated with HOC partitioning for a variety of DOMs (Akkanen et al., 2004), as well as to fractionated constituents from a single DOM or even microbiologically-altered DOM (Hur and Kim, 2009; Hur et al., 2009). In this study as well, Koc values exhibited positive trends with SUVA for both the adsorptive fractionated and photo-irradiated LLHS, although the correlation was less strong with the irradiated samples due to more data scatter (Fig. 5a). Interestingly, there appears to be no significant difference between trend lines of Koc vs. SUVA for the two different processes. Koc values exhibited positive trends as well with MWw for both the adsorptive fractionated and photo-irradiated LLHS (Fig. 5b). In contrast with SUVA, there was much less data scatter for the irradiated samples and thus the correlation coefficient was much higher. Also, unlike SUVA there was a clear distinction between the trend lines of Koc vs. MWw for
a K oc (L/mg C)
in aggregate fluorophores (Patel-Sorrentino et al., 2004) and possible formation of relatively simple structures with other fluorescence characteristics (Rodrı´guez-Zu´n˜iga et al., 2008). The relatively small variation of HIX after a certain period of irradiation may be attributed to the combined effects of the initial substantial destruction of intramolecular chargetransfer interactions and the direct photolysis of fluorescent compounds related to relatively short wavelengths. Similar to SUVA and MWw, Koc values decreased continuously with irradiation time and increased DOC removal (Fig. 4c). For the 12-day samples, no pyrene quenching was observed and thus the Koc was not measurable. This was somewhat unexpected because the corresponding SUVA and MWw values of those samples were higher than the smallest UF fraction of LLHS which did exhibit measureable, albeit low, pyrene binding. Our results suggest that extended photo-irradiation periods transform HS and induce formation of altered moieties/structures less capable of binding HOCs. In other words, the hydrophilic nature of the bulk HS may be intensified by production of oxygen-containing moieties (Pullin et al., 2004b; Xie et al., 2004), and/or substitution of oxygen into aliphatic chains (Carvalho et al., 2008). In addition, selective destruction of intramolecular DeA interactions may lead to the reduction of hydrophobic cavities within the bulk HS available for pyrene partitioning (Hur and Schlautman, 2003a; Del Vecchio and Blough, 2004; Polubesova et al., 2007; Jung et al., 2010).
30 Adsorptive fractionated (r=0.97, p<0.001) Photo-irradiated (r=0.69, p=0.008)
20
10
0 1.5
2.0
2.5
3.0
3.5
2000
2500
SUVA (L/mg C-m)
b Koc (L/mg C)
6224
30 Adsorptive fractionated (r=0.98, p<0.001) Photo-irradiated (r=0.93, p<0.001)
20
10
0 500
1000
1500
MWw (g/mol as PSS)
Fig. 5 e (a) Trends between pyrene Koc and SUVA for adsorptive fractionated and photo-irradiated LLHS. (b) Trends between pyrene Koc and MW for adsorptive fractionated and photo-irradiated LLHS. In both figures, the dashed and solid lines designate observed linear trends for the LLHS samples undergoing adsorptive fractionation and photo-irradiation, respectively.
the two different processes. Therefore, unless one knew which process was dominant, predicting Koc values based on MWw would be problematic. Under a comparable range of the MWw values, more pyrene binding was observed for the residual LLHS after contact with kaolinite. As discussed above, this may be due to the creation of less favorable environments for HOCs upon photo-irradiation despite the fact that large molecules with aromatic structures are still present in solution. Xie et al. (2004) have shown that the DOC-normalized carboxyl content of DOM increased upon UV irradiation in the presence of oxygen. Production of such oxygen-containing moieties is expected to render the remaining HS less hydrophobic, thereby reducing pyrene partitioning. Conformational changes in LLHS resulting from photo-irradiation may have also contributed to the lower Koc values (Patel-Sorrentino et al., 2004).
4.
Conclusions
Comparison of the changes in LLHS characteristics resulting from fractionation upon adsorptive removal by kaolinite versus alteration by simulated solar irradiation provides insight into the fate and distribution of terrestrial HS in aquatic environments. Based on the findings of this study, we conclude:
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 1 7 e6 2 2 6
Certain operational descriptors of LLHS properties respond similarly to the two different processes. Values of SUVA, MWw, % PLF and HIX tend to decrease with increasing percent DOC removal whether caused by increasing irradiation times or higher concentrations of kaolinite. Both processes lead to preferential removal of the more humified and condensed aromatic structures from the original bulk LLHS and higher relative fractions of PLF-related structures. Despite these observed similarities in the LLHS characteristics, inherent differences in the two processes result in the opposing trends in the changes in FI values, net production of intermediate size UV-absorbing constituents during photo-irradiation but not adsorption, and higher pyrene binding for adsorptive fractionated LLHS samples despite comparable MWw ranges. Alteration of LLHS by photo-irradiation appears to produce more acidic and hydrophilic constituents, conformational changes by selective destruction, and/or disruption of intramolecular charge-transfer interactions, all of which lead to less pyrene binding.
Acknowledgments This work was primarily supported by the National Research Foundation of Korea Grant funded by the Korean government (No. 2011-0026553). Additional support was provided by the USDA Cooperative State Research, Education, and Extension Service (USDA-CSREES) under project number SC-1700395. Any opinions, findings, conclusions or recommendations expressed in this article are solely those of the authors and do not necessarily reflect the views of the two funding agencies.
references
Akkanen, J., Vogt, R.D., Kukkonen, J.V.K., 2004. Essential characteristics of natural dissolved organic matter affecting the sorption of hydrophobic organic contaminants. Aquatic Sciences 66, 171e177. Asakawa, D., Mochizuki, H., Yanagi, Y., Eujitake, N., 2007. Characterization of hydrophobic acid fractions in water-soluble organic matter in dystric cambisol and in a stream in a small forested watershed: seasonal and vertical variations in chemical properties. Soil Science and Plant Nutrition 53, 551e561. Brooks, M.L., Meyer, J.S., McKnight, D.M., 2007. Photooxidation of wetland and riverine dissolved organic matter: altered copper complexation and organic composition. Hydrobiologia 579, 95e113. Canellas, L.P., Piccolo, A., Dobbss, L.B., Spaccini, R., Olivares, F.L., Zandonadi, D.B., Fac¸anha, A.R., 2010. Chemical composition and bioactivity properties of size-fractions separated from a vermicompost humic acid. Chemosphere 78, 457e466. Carvalho, S.I.M., Otero, M., Duarte, A.C., Santos, E.B.H., 2008. Spectroscopic changes on fulvic acids from a kraft pulp mill effluent caused by sun irradiation. Chemosphere 73, 1845e1852. Chin, Y.P., Aiken, G.R., Danielsen, K.M., 1997. Binding of pyrene to aquatic and commercial humic substances: the role of molecular weight and aromaticity. Environmental Science & Technology 31, 1630e1635.
6225
Chefetz, B., Deshmukh, A.P., Hatcher, P.G., Guthrie, E.A., 2000. Pyrene sorption by natural organic matter. Environmental Science & Technology 34, 2925e2930. Chen, Z., Hu, C., Conmy, R.N., Muller-Karger, F., Swarzenski, P., 2007. Colored dissolved organic matter in Tampa Bay, Florida. Marine Chemistry 104, 98e109. Clark, C.D., De Bruyn, W.J., Ting, J., Scholle, W., 2007. Solution medium effects on the photochemical degradation of pyrene in water. Journal of Photochemistry and Photobiology A: Chemistry 186, 342e348. Del Vecchio, R., Blough, N.V., 2002. Photobleaching of chromophoric dissolved organic matter in natural waters: kinetics and modeling. Marine Chemistry 78, 231e253. Del Vecchio, R., Blough, N.V., 2004. On the origin of the optical properties of humic substances. Environmental Science & Technology 38, 3885e3891. Esparza-Soto, M., Westerhoff, P., 2003. Biosorption of humic and fulvic acids to live activated sludge biomass. Water Research 37, 2301e2310. Fuentes, M., Gonzalez-Gaitano, G., Ma Garcia-Mina, J., 2006. The usefulness of UVevisible and fluorescence spectroscopies to study the chemical nature of humic substances from soils and composts. Organic Geochemistry 37, 1949e1959. Gauthier, T.M., Shane, E.C., Guerin, W.F., Seitz, W.R., Grant, C.L., 1986. Fluorescence quenching method for determining equilibrium constants for polycyclic aromatic hydrocarbons binding to dissolved humic materials. Environmental Science & Technology 20, 1162e1166. Helms, J.R., Stubbins, A., Ritchie, J.D., Minor, E.C., Kieber, D.J., Mopper, K., 2008. Absorption spectral slopes and slope ratios as indicators of molecular weight, source, and photobleaching of chromophoric dissolved organic matter. Limnology and Oceanography 53 (3), 955e969. Hur, J., Schlautman, M.A., 2003a. Using selected operational descriptors to examine the heterogeneity within a bulk humic substance. Environmental Science & Technology 37, 880e887. Hur, J., Schlautman, M.A., 2003b. Molecular weight fractionation of humic substances by adsorption onto minerals. Journal of Colloid and Interface Science 264, 313e321. Hur, J., Schlautman, M.A., 2004. Influence of humic substance adsorptive fractionation on pyrene partitioning to dissolved and mineral-associated humic substances. Environmental Science & Technology 38, 5871e5877. Hur, J., Kim, G., 2009. Comparison of the heterogeneity within bulk sediment humic substances from a stream and reservoir via selected operational descriptors. Chemosphere 75, 483e490. Hur, J., Park, M.-H., Schlautman, M.A., 2009. Microbial transformation of dissolved leaf litter organic matter and its effects on selected organic matter operational descriptors. Environmental Science & Technology 43, 2315e2321. Hur, J., 2011. Microbial changes in selected operational descriptors of dissolved organic matters from various sources in a watershed. Water, Air & Soil Pollution 215, 465e475. Jung, A.-V., Frochot, C., Villieras, F., Lartiges, B.S., Parant, S., Viriot, M.-L., Bersillon, J.-L., 2010. Interaction of pyrene fluoroprobe with natural and synthetic humic substances: Examining the local molecular organization from photophysical and interfacial processes. Chemosphere 80, 228e234. Liu, S., Lim, M., Fabris, R., Chow, C.W.K., Drikas, M., Korsin, G., Amal, R., 2010. Multi-wavelength spectroscopic and chromatography study on the photocatalytic oxidation of natural organic matter. Water Research 44, 2525e2532. Lou, T., Xie, H., 2006. Photochemical alternation of the molecular weight of dissolved organic matter. Chemosphere 65, 2333e2342. Lou, T., Xie, H.X., Chen, G.H., Gagne, J.P., 2006. Effects of photodegradation of dissolved organic matter on the binding of benzo(a)pyrene. Chemosphere 64, 1204e1211.
6226
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 1 7 e6 2 2 6
Maie, N., Scully, N.M., Pisani, O., Jaffe´, R., 2007. Composition of a protein-like fluorophore of dissolved organic matter in coastal wetland and estuarine ecosystems. Water Research 41, 563e570. McKnight, D.M., Boyer, E.W., Westerhoff, P.K., Doran, P.T., Kulbe, T., Andersen, D.T., 2001. Spectrofluorometric characterization of dissolved organic matter for indication of precursor organic material and aromaticity. Limnology and Oceanography 46, 38e48. Meier, M., Chin, Y.-P., Maurice, P., 2004. Variations in the composition and adsorption behavior of dissolved organic matter at a small, forested watershed. Biogeochemistry 67, 39e56. Meyer, J.L., Wallace, J.B., Eggert, S.L., 1998. Leaf litter as a source of dissolved organic carbon in streams. Ecosystems 1, 240e249. Murphy, E.M., Zachara, J.M., Smith, S.C., 1990. Influence of mineral-bound humic substances on the sorption of hydrophobic organic compounds. Environmental Science & Technology 24, 1507e1516. Obernosterer, I., Benner, R., 2004. Competition between biological and photochemical processes in the mineralization of dissolved organic carbon. Limnology and Oceanography 49, 117e124. Patel-, Sorrentino, Mounier, S., Lucas, Y., Benaim, J.Y., 2004. Effects of UV-visible irradiation on natural organic matter from the Amazon basin. Science of the Total Environment 321, 231e239. Polubesova, T., Sherman-Nakache, M., Chefetz, B., 2007. Binding of pyrene to hydrophobic fractions of dissolved organic matter: effect of polyvalent metal complexation. Environmental Science & Technology 41, 5389e5394. Pullin, M.J., Progess, C.A., Maurice, P.A., 2004a. Effects of photoirradiation on the adsorption of dissolved organic matter to goethite. Geochimica et Cosmochimica 68 (18), 3643e3656. Pullin, M.J., Bertilsson, S., Goldstone, J.V., Voelker, B.M., 2004b. Effects of sunlight and hydroxyl radical on dissolved organic matter: bacterial growth efficiency and production of carboxylic acids and other substances. Limnology and Oceanography 49 (6), 2011e2022. Qualls, R.G., Richardson, C.J., 2003. Factors controlling concentration, export, and decomposition of dissolved organic nutrients in the everglades of Florida. Biogeochemistry 62, 197e229. Rodrı´guez-Zu´n˜iga, U.F., Miloria, D.M.B.P., Da Silva, W.T.L., MartinNeto, L., Oliveira, L.C., Rocha, J.C., 2008. Changes in optical properties caused by UV-irradiation of aquatic humic
substances from the Amazon River basin: seasonal variability evaluation. Environmental Science & Technology 42, 1948e1953. Schlautman, M.A., Morgan, J.J., 1994. Adsorption of aquatic humic substances on colloidal-size aluminum oxide particles: influence of solution chemistry. Geochimica Cosmochimica Acta 58, 4293e4303. Shank, G.C., Zepp, R.G., Whitehead, R.F., Moran, M.A., 2005. Variations in the spectral properties of freshwater and estuarine CDOM caused by partitioning onto river and estuarine sediments. Estuarine. Coastal and Shelf Science 65, 289e301. Shank, G.C., Zepp, R.G., Vahatalo, A., Lee, R., Bartels, E., 2010. Photobleaching kinetics of chromophoric dissolved organic matter derived from mangrove leaf litter and floating Sargassum colonies. Marine Chemistry 119, 162e171. Stedmon, C.A., Markager, S., Tranvik, L., Kronberg, L., Sla¨tis, T., Martinsen, W., 2007. Photochemical production of ammonium and transformation of dissolved organic matter in the Baltic Sea. Marine Chemistry 104, 227e240. Steinburg, C.E., 2003. Ecology of Humic Substances in Freshwaters. Springer, New York. Sulzberger, B., Durisch-Kaiser, E., 2009. Chemical characterization of dissolved organic matter (DOM): a prerequisite for understanding UV-induced changes of DOM absorption properties and bioavailability. Aquatic Sciences 71, 104e126. Sun, W.L., Ni, J.R., Xu, N., Sun, L.Y., 2007. Fluorescence of sediment humic substance and its effect on the sorption of selected endocrine disruptors. Chemosphere 66, 700e707. Thorn, K.A., Younger, S.J., Cox, L.G., 2010. Order of functionality loss during photodegradation of aquatic humic substances. Journal of Environmental Quality 39, 1416e1428. Thurman, E.M., 1985. Organic Geochemistry of Natural Waters. Martinus Nijhoff/Junk Publisher, Nordrecht, The Netherlands. Vidali, R., Remoundaki, E., Tsezos, M., 2010. Humic acids copper binding following their photochemical alteration by simulated solar light. Aquatic Geochemistry 16, 207e218. Xie, H., Zafiriou, O.C., Cai, W.-J., Zepp, R.G., Wang, Y., 2004. Photooxidation and its effects on the carboxyl content of dissolved organic matter in two coastal rivers in the Southeastern United States. Environmental Science & Technology 38, 4113e4119. Zsolnay, A., Baigar, E., Jimenez, M., Steinweg, B., Saccomandi, F., 1999. Differentiating with fluorescence spectroscopy the sources of dissolved organic matter in soils subjected to drying. Chemosphere 38, 45e50.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 2 7 e6 2 3 9
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Bacterial, viral and turbidity removal by intermittent slow sand filtration for household use in developing countries: Experimental investigation and modeling Marion W. Jenkins a,*, Sangam K. Tiwari b, Jeannie Darby a a b
Department of Civil & Environmental Engineering, University of California, Davis, One Shields Ave., Davis, CA 95616, USA Trussell Technologies, Inc., 232 North Lake Avenue, Suite 300, Pasadena, CA 91101, USA
article info
abstract
Article history:
A two-factor three-block experimental design was developed to permit rigorous evaluation
Received 17 March 2011
and modeling of the main effects and interactions of sand size (d10 of 0.17 and 0.52 mm)
Received in revised form
and hydraulic head (10, 20, and 30 cm) on removal of fecal coliform (FC) bacteria, MS2
9 September 2011
bacteriophage virus, and turbidity, under two batch operating modes (‘long’ and ‘short’) in
Accepted 10 September 2011
intermittent slow sand filters (ISSFs). Long operation involved an overnight pause time
Available online 22 September 2011
between feeding of two successive 20 L batches (16 h average batch residence time (RT)). Short operation involved no pause between two 20 L batch feeds (5 h average batch RT).
Keywords:
Conditions tested were representative of those encountered in developing country field
Point-of-use
settings. Over a ten week period, the 18 experimental filters were fed river water
Drinking water treatment
augmented with wastewater (influent turbidity of 5.4e58.6 NTU) and maintained with the
Fecal coliform bacteria
wet harrowing method. Linear mixed modeling allowed systematic estimates of the
MS2 bacteriophage
independent marginal effects of each independent variable on each performance outcome
Biosand filter
of interest while controlling for the effects of variations in a batch’s actual residence time,
Linear mixed models
days since maintenance, and influent turbidity. This is the first study in which simulta-
Factorial design experiment
neous measurement of bacteria, viruses and turbidity removal at the batch level over an
Residence time
extended duration has been undertaken with a large number of replicate units to permit
Influent turbidity
rigorous modeling of ISSF performance variability within and across a range of likely filter design configurations and operating conditions. On average, the experimental filters removed 1.40 log fecal coliform CFU (SD 0.40 log, N ¼ 249), 0.54 log MS2 PFU (SD 0.42 log, N ¼ 245) and 89.0 percent turbidity (SD 6.9 percent, N ¼ 263). Effluent turbidity averaged 1.24 NTU (SD 0.53 NTU, N ¼ 263) and always remained below 3 NTU. Under the best performing design configuration and operating mode (fine sand, 10 cm head, long operation, initial HLR of 0.01e0.03 m/h), mean 1.82 log removal of bacteria (98.5%) and mean 0.94 log removal of MS2 viruses (88.5%) were achieved. Results point to new recommendations regarding filter design, manufacture, and operation for implementing ISSFs in local settings in developing countries. Sand size emerged as a critical design factor on performance. A single layer of river sand used in this investigation demonstrated removals comparable to those reported for 2 layers of crushed sand. Pause time and increased residence time each emerged as highly beneficial for improving removal performance on all four outcomes. A relatively large and significant negative effect of influent turbidity on MS2 viral removal in the ISSF was measured in
* Corresponding author. Tel.: þ1 530 754 6427; fax: þ1 530 752 7872. E-mail address: [email protected] (M.W. Jenkins). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.09.022
6228
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 2 7 e6 2 3 9
parallel with a much smaller weaker positive effect of influent turbidity on FC bacterial removal. Disturbance of the schmutzdecke by wet harrowing showed no effect on virus removal and a modest reductive effect on the bacterial and turbidity removal as measured 7 days or more after the disturbance. For existing coarse sand ISSFs, this research indicates that a reduction in batch feed volume, effectively reducing the operating head and increasing the pore:batch volume ratio, could improve their removal performance by increasing batch residence time. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
Access to improved drinking water is unavailable to an estimated 884 million people in the world, most of who live in rural, dispersed, and often remote communities in developing countries (WHO/UNICEF, 2010). Diarrhea and other waterborne diseases from exposure to microbial pathogens in unsafe water constitute a major threat to health in these settings. The World Health Organization recommends pointof-use household water treatment (POU) as an intervention to address the need, drawing on appropriate low-cost technologies (Sobsey, 2002; WHO, 2007). A recent assessment of POU options in developing countries identified intermittently operated slow sand filtration (ISSF), commonly referred to as the BioSand filter (BSF), among the most promising (Sobsey et al., 2008). The BSF was adapted for household use from traditional slow sand filtration (SSF) and is designed to treat 20e60 L/day in a batch-like gravity flow operating mode (Buzunis, 1995; Manz, 2004) under close to plug flow hydraulics (Elliott et al., 2008). ISSF containers have typically been designed to accept about 20 L at a time at a maximum head of 17e29 cm, which continuously declines until filtration is complete. Ideally, the batch remains within the filter until the next batch is added, however, this retention depends greatly on a filter design that ensures at least a 1:1 volume ratio of sand pore space to batch feed and efficient plug flow hydraulics. Assuming a batch mostly remains within the filter until the next feed, the time from the start of one 20 L batch feed to the start of the next batch feed is defined in this study as the batch residence time. In limited controlled laboratory testing of the original Davnor BioSand Water Filter (D-BSF), the following improvements in the microbial quality of water have been reported: bacterial removal for fecal coliform or Escherichia coli ranging from 63% up to 99% (2 log10) with averages of 94% and 96% (Buzunis, 1995; Stauber et al., 2006); viral removal ranging from 0 to 0.75 log10 measured using MS2 and PRD-1 bacteriophage surrogates, and 1.14 log10 of echovirus 12 (Elliott et al., 2008); protozoan removals of greater than 5 log10 for Giardia lamblia cysts (6e16 mm diameter) and 99.98% for Cryptosporidium oocysts (4e7 mm diameter) (Palmateer et al., 1999). The BSF has several advantages as a POU technology in low income developing country rural settings where improved water supplies are often difficult and costly to develop, operate or maintain. Using a concrete or plastic container with a typical sand column of 45e50 cm, the simple yet robust design of BSF units allows construction with local materials and skills found anywhere in the world, making it affordable
(US $20e30/unit), accessible and durable (Duke et al., 2006; Fewster et al., 2004). There are no recurring costs and operation and maintenance requirements can be performed by the household. Relative to other options, for example, solar and chemical disinfection, ceramic filtration, and flocculants, the BSF’s high flow rate and ability to tolerate turbid surface water provide added advantages. An estimated 140,000 locally constructed BSF units were in operation in over 24 countries by 2007, largely through the efforts of decentralized small-scale development organizations (Clasen, 2009). Field designs and local construction methods in developing countries often result in BSFs that differ from the original DBSF design specifications. A single layer of local river sand of variable size (characterized by effective size, d10, and uniformity coefficient, UC) is often used as the filtration media instead of the D-BSF’s two different size layers of crushed sand (Manz, 2004). ISSF containers used in field projects are generally made of concrete, and can vary in their maximum hydraulic head, sand column depth, and headspace volume to a greater or lesser degree from the original plastic D-BSF container specifications. Variations and less than ideal performance in field testing have been reported for BSFs, ranging from negative up to 100 percent bacterial removal (Duke et al., 2006; Earwaker, 2006; Fewster et al. 2004; Kaiser et al. 2002; Stauber et al., 2006; Wiesent-Brandsma et al. 2004) and 39e91 percent for turbidity reductions (Duke et al., 2006; Earwaker, 2006; Jenkins et al., 2009; Stauber et al., 2006; Wiesent-Brandsma et al. 2004). Difficult logistics in developing countries necessitate collecting BSF effluent and influent grab samples for field evaluations simultaneously during a single house visit, limiting comparability and usefulness of field-reported removal efficiencies. Influent water quality in settings where BSFs are typically installed can vary from batch to batch as households switch sources and source water quality varies naturally from day to day. A switch, for example, from a turbid surface source to a less turbid rain feed can lead to erroneously low or even negative removal measurements based on simultaneous influenteeffluent (flush-pore) water sampling (Earwaker, 2006). Systematic scientific investigation of the effects of variations in BSF design, construction and operation on performance across multiple outcomes of concern, including bacterial, viral, and turbidity removal, is absent in the literature. Several recent evaluations have pointed to the absence of and the need for rigorous investigations to support optimization of ISSF design (Elliott et al., 2008; Kubare and Haarhoff, 2010). Operating conditions are another likely
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 2 7 e6 2 3 9
important influence on performance. Baumgartner et al. (2007) demonstrated that residence time and dosing volume significantly affected total coliform removal in the D-BSF. Elliott et al. (2008) observed that feed volumes greater than 50 percent of the filter pore volume for the D-BSF tended to show decreased incremental removal efficiencies for E. coli and bacteriophages. Application of slow sand filters for household use has spread rapidly across the globe in recent years, creating a need for sound scientific understanding of mechanisms and factors controlling ISSF microbial removal. This includes understanding of how performance is affected by variations in design, construction materials, sand characteristics, and household operation and maintenance practices. Such knowledge would provide a rational basis to inform development of design standards, quality control measures, and guidelines for local construction and operation to maximize ISSF performance in a local setting. In this paper we report on experimental research undertaken to systematically investigate and measure the effects of ISSF design and operating factors on its ability to simultaneously remove bacteria, viruses and turbidity. A factorial design experiment was developed to permit rigorous evaluation and modeling of the main effects and interactions of sand size and hydraulic head on ISSF removal of fecal coliform bacteria, MS2 bacteriophage virus, and turbidity, under two batch operating modes.
2.
Materials and methods
2.1.
Filter design
A diagram of the experimental ISSF is shown in Fig 1. The container was constructed from 12-inch polyvinyl chloride (PVC) irrigation pipe (30.5 cm diameter). Each filter consisted of a 5 cm rock layer at the base, followed by 5 cm of gravel, and 60 cm of a single layer of one of the two experimental sands. A single sand layer is commonly used in developing countries to save on costs. Two and a half cm of water were maintained above the sand at all times, ensuring saturated conditions. This configuration was selected to contain approximately 20 L of water within the sand column pore space and in the headspace at 30 cm above the static water level. Fig 1 shows a simple constant head controller (CHC) feed bottle above the filter constructed from a five gallon carboy. The CHC was required for filter configurations with less than 30 cm head to passively feed a 20 L batch without exceeding the filter’s design head.
2.2.
Factorial experimental design
A two-factor three-block experimental design was selected (Montgomery, 2005). Each block consisted of the same six filter configurations of interest (Table 1). The fine sand size (d10 of 0.17 mm) was selected to represent the recommended lower range for typical slow sand filters (Huisman and Wood, 1974). The coarse sand size (d10 of 0.52 mm) was selected to represent a worst-case scenario for ISSF in places that have only coarse sand readily available. Naturally occurring river sand
6229
Fig. 1 e Experimental intermittent slow sand filter.
was used, as it is the most commonly available and affordable sand in developing community settings. The coarse and fine experimental sands were derived from ASTM concrete and utility river sand, respectively (Granite Construction Company, Sacramento, CA). A minimum hydraulic head of 10 cm was selected so that, when coupled with the fine sand, it produced a hydraulic loading rate (HLR) sufficient for minimum household daily drinking water needs. The maximum head of 30 cm represents one commonly used BSF container design (www. biosandfilter.org). BSF households typically operate their filter under a range of modes, treating from one to three 20 L batches per day, resulting in wide variation in batch residence time (RT). In this research, a short and a long batch RT operation were examined. The short RT operating mode represents the shortest possible batch residence time (experimental average 5.1 h; variable with filter configuration) and is approximately equal to the start-to-start time of two successive 20 L batches fed to the filter with little or no pause between feeds. The long RT operating mode represents the longest possible residence time (experimental average 15.6 h; less than theoretical 24 h due to varying daily feed times) that would result from one 20 L batch per day operation.
2.3.
Filter operation
Each filter was fed a standard batch of 20 L of the influent water mixture per day for 10 weeks, except during weekly testing. Testing involved feeding three 20 L test batches over two days, as explained with this example. Batch I was started
6230
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 2 7 e6 2 3 9
Table 1 e Experimental design filter configurations. Blocka 1
2
3
Unit
Factor 1: effective grain size (d10 mm/UCb)
Factor 2: nominal head (cm)
Initial hydraulic loading ratec (m/h)
1 2 3 4 5 6
0.17/2.4 0.52/2.1 0.17/2.4 0.52/2.1 0.17/2.4 0.52/2.1
10
0.01 0.11 0.06 0.31 0.13 0.4
7 8 9 10 11 12
0.17/2.4 0.52/2.1 0.17/2.4 0.52/2.1 0.17/2.4 0.52/2.1
13 14 15 16 17 18
0.17/2.4 0.52/2.1 0.17/2.4 0.52/2.1 0.17/2.4 0.52/2.1
20 30
10 20 30
10 20 30
0.03 0.1 0.07 0.24 0.1 0.39 0.02 0.08 0.07 0.25 0.11 0.41
a Block 1 was conducted from April 4 to June 18, 2006 and blocks 2 and 3 were conducted simultaneously from July 7 to September 17, 2006. b d10 is the sand size diameter for which 10% of the sand mass is smaller and 90% is larger in size; UC is the uniformity coefficient of the sand measured as the ratio of the d60 to d10 grain size. c Determined from time to collect one liter of filtered water at the beginning of each batch feed.
at the same time (e.g., 3 pm) in all six filters within a block on test day 1. Infiltration of batch I in a block of filters finished at varying times on day 1. The next day (test day 2), batch II was started in the same six filters at the same time (e.g., noon). Batch II finished infiltrating in filter A of the block at 4 pm. Upon complete infiltration of batch II in filter A (equal to complete exit of batch I from filter A), batch III was started in filter A. Infiltration of batch III in filter A finished at 8 pm, equal to the time batch II fully exited filter A and could be tested. Batch I is the long RT batch which experiences an overnight pause time in the filter pore space. Batch II is the short RT batch which is flushed out of the filter as soon as it has finished infiltrating. In the example, the long batch I and short batch II RTs for filter A are 21 h (difference of start times of batches II and I) and 4 h (difference in start times of batches III and II), respectively. At a 30 cm nominal head above the static water level, the headspace of the experimental filter held approximately 20 L, thus a 20 L batch was poured directly onto the diffuser plate at the start of each batch feed for the 30 cm head filter configurations. For the 10 and 20 cm nominal head configurations, the CHC was filled with 20 L and inverted above the filter at the start of a batch with its narrow mouth opening set at the prescribed height above the static water level so as to maintain the supernatant head at the filter’s nominal head until the CHC was empty. After controlled release of all 20 L of water from the CHC into the headspace at the nominal head, the head of the remaining portion of the batch declined steadily until filtration was complete. Official testing began in week 3, allowing an initial 2-week maturation period for the biological zone to establish within the sand. The filters were maintained by the wet harrowing method, a gentle rubbing of the top 2 cm of sand followed by
decanting of the resulting suspension of clogging material. After maintenance, the filter was allowed to mature for one week before resumption of sampling measurements. Filters in the first block were maintained when their flow rate became too slow to filter a 20 L batch in 24 h, whereas, filters in blocks 2 and 3 were cleaned when their flow rates reached 50 percent of their initial value, resulting in more frequent filter maintenance.
2.4.
Influent water
The influent water quality was designed to roughly simulate a typical surface water source used in a developing country. Influent water fed to the filters throughout the study was 95 percent untreated Sacramento River water augmented with 5 percent raw wastewater from the University of California, Davis Wastewater Treatment Plant (UCD WWTP). The wastewater had an average BOD of 200 mg/L, fecal coliform concentration of 2 million CFU/100 mL, and ammonia-N of 10 mg/L. The mixture was spiked every day with MS2 coliphage (ATCC 15597-B1) due to a low background concentration. The MS2 coliphage was prepared using Standard Methods 9224 C (APHA, 2005). Raw river water was collected weekly throughout the study. Raw sewage was collected every other day, except for sampling days when fresh raw sewage was collected and used. Maximum, minimum, and mean values of the influent water characteristics within each block and overall are shown in Table 2. Influent turbidities and MS2 coliphage concentrations were significantly higher in block 1 than in blocks 2 and 3. Sacramento River water at the West Sacramento intake was considerably more turbid during the spring run-off months from April to June, when block 1 was conducted, than in the
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 2 7 e6 2 3 9
6231
Table 2 e Filter influent watera quality characteristics. Block
Water quality parameter Log fecal coliform (CFU/100 mL)
Log MS2 coliphage (PFU/100 mL)
Turbidity NTU
1. Mean SD (N )b MineMax
5.18 0.48 (96) 4.40e5.99
6.77 1.21 (96)c 4.94e8.05
24.19 12.4 (96)c 12.93e58.57
2. Mean SD (N ) MineMax
5.18 0.48 (114) 3.94e5.84
4.96 0.53 (114) 4.04e6.27
12.99 7.9 (126) 5.51e37.60
3. Mean SD (N ) MineMax
5.16 0.49 (120) 4.12e5.84
4.95 0.51 (120) 4.15e6.37
11.89 5.4 (132) 5.36e26.27
Combined Mean SD (N ) MineMax
5.17 0.48 (330) 3.94e5.99
5.48 1.14 (330) 4.04e8.05
15.62 10.1 (354) 5.36e58.57
a Includes filter measurements during initial and maintenance weeks. b N refers to the number of sample measurements. c Block 1 influent water quality is significantly different from quality of Blocks 2 and 3 influent, at p < 0.001, for MS2 coliphage concentration and turbidity.
dry season summer months of JulyeSeptember, when blocks 2 and 3 were conducted. Unobserved seasonal variation in chemical, physical and microbiological characteristics of Sacramento River water between block 1 and blocks 2/3 is also possible.
2.5.
Sampling and measurements
Following the initial 2-week start-up, experimental measurements were conducted weekly on each filter for a long and short residence time test batch as described above. Influent and effluent water samples for each long and short test batch were collected. Effluent samples were collected from the 20 L composite effluent volume upon exit of the test batch. Influent samples were collected from the 120 L influent batch prepared separately for each test batch feed for each block of 6 filter units. Samples were analyzed for fecal coliform bacteria, MS2 coliphage virus, and turbidity. Fecal coliform was enumerated using Standard Method 9222D with M-FC medium as specified therein (21st edition, APHA, 2005). MS2 coliphage was enumerated as per Standard Method 9224D (APHA, 2005) with E. coli (ATCC 15597) as the host and no antibiotics. Turbidity was measured using a turbidimeter (Model 2100AN, Hach Company, Loveland, CO). The hydraulic loading rate (HLR) was determined from the time to collect one liter of filtered water at the beginning of each batch feed. On each test day and for each test batch and filter, several covariates of interest were measured and recorded. These included influent water and room temperature, date and time of start of each influent test batch feed and time of exit of test batch effluent, and date of each filter maintenance event.
2.6.
Analysis and modeling
The research experiment was designed to identify statistically significant independent effects on ISSF batch removal performance caused by differences in effective sand size, nominal head, and residence time operation as well as the interactions among them. The significance and size of the main factor effects were estimated using linear mixed modeling (LMM),
controlling for repeated sampling of a filter unit, random block differences, and covariate effects on performance variability (Faraway, 2006; Verbeke, 2000). Accounting for repeated filter measurement in LMM analysis controls for possible correlation (statistical non-independence) of measurements from a given filter unit. Setting block as a random effect controls for the possibility of unobserved systematic differences in filter set-up and operating characteristics between blocks, such as sand batch differences, seasonal variation of influent water characteristics, and maintenance schedules (Verbeke, 2000). Covariates of interest included in the analysis were: a) deviation of the measured residence time of a long or short RT sample batch from the long or short RT operation group average, b) days since filter maintenance, and c) influent turbidity, with the latter included only in fecal coliform and MS2 removal performance models. Temperature was unnecessary to include as it remained uniform throughout the controlled experiment. Test batch measurements within seven days of a maintenance event were excluded from performance results and analyses. Results were analyzed using SPSS statistical software (SPSS Inc., Chicago, Illinois). Four dependent variable outcomes were modeled: log10 fecal coliform removal, log10 MS2 coliphage removal, percent turbidity reduction, and effluent turbidity in the measured long and short 20 L test batches, across the 18 experimental filter units. First, 2-factor LMM analysis was undertaken to examine the effects of grain size (2 levels) and nominal head (3 levels) separately for short and long RT batch operation. Then, batch operation mode (2 levels) was added as a third factor in a three-factor LMM of all long and short batch measurements combined, comprising from 245 to 263 performance data points for each outcome of interest. Missing covariate values for batch residence time deviation and days since maintenance were replaced by group averages. Only statistically significant interaction terms, at the 0.10 level, were retained in the final model. The main factor and covariate effects and their marginal means from LMM indicate level of significance of each factor or covariate on filter performance and the mean effect size of a change in a specified factor level, or a unit increase in a covariate, adjusted for repeated filter sampling
6232
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 2 7 e6 2 3 9
and random block effects. Pairwise comparisons of mean performance effect size of each nominal head level were made using the Tukey method, which adjusts significance for multiple comparisons. Model appropriateness was assessed using the Levene-style test for equal variance of the residuals and graphical analysis of residuals. Normality was assessed using normal probability plots and the ShapiroeWilks test. No violations of the LMM assumptions were found.
3.
Results
3.1.
Filter characteristics and experimental conditions
Average porosity of the sand column for the 18 filter units was 0.448 0.022. The uniformity coefficient of the fine and coarse experimental sands was 2.4 and 2.1, respectively. The initial HLR of each unit varied from a low of 0.01 m/h (fine sand, 10 cm head, Block 1) to a high of 0.41 m/h (coarse sand, 30 cm head, Block 3) (Table 1). Average influent water and room temperature across all blocks were 24.2 and 24.3 C, respectively. Influent water pH ranged between 6.7 and 7. Influent water MS2 coliphage and turbidity characteristics during block 1 were significantly different from blocks 2 and 3 (Table 2). Influent turbidity varied from a low of 5.36 NTU to a high of 58.57 NTU. Table 3 presents the range of covariate values within each block and overall during the experiment. On average, both the long and short batch residence times were longer in block 1 than in blocks 2 or 3. The two-tailed t-test for the long batch residence time difference is significant (at the 0.05 level) between blocks 1 and 2 ( p ¼ 0.020), but not between blocks 1 and 3 ( p ¼ 0.31) or blocks 2 and 3 ( p ¼ 0.154). The short batch residence time difference between blocks 1 and 2 ( p ¼ 0.003) and 1 and 3 ( p ¼ 0.003) is also significant, but not between blocks 2 and 3 ( p ¼ 0.72). A less frequent maintenance schedule applied during block 1 is the most apparent reason for the higher block 1 long and short batch residence times but could also be the result of influent water quality differences or small unobserved differences in the sand characteristics or packing of block 1 compared to blocks 2 and 3. Days since last maintenance is lowest for block 2 and highest for block 3,
although this difference is not significant ( p ¼ 0.075; 2-tailed ttest). Inclusion of model covariates for the deviation of a batch’s actual residence time from the long or short group average (across all blocks), for days since maintenance, and for influent turbidity where relevant, allow explicit examination of the independent effects of these operational differences on filter performance, separated from the main factor effects. In particular, controlling for residence time variation of a particular batch of water separates configuration-related variations in residence time under a given operation mode, for example those attributable to sand size or nominal head configuration differences, from the main effect of the pause between batch feeds that arises under long RT operating mode.
3.2.
Overall performance
Filter performance averaged across the six different configurations is shown in Table 4 for each block and combined across all blocks. On average, the experimental filters removed 1.40 log fecal coliform CFU (SD 0.40 log, N ¼ 249), 0.54 log MS2 PFU (SD 0.42 log, N ¼ 245) and 89.0 percent turbidity (SD 6.9 percent, N ¼ 263). Effluent turbidity averaged 1.24 NTU (SD 0.53 NTU, N ¼ 263) and always remained below 3 NTU. Filter performance on all four outcomes was better under long than under short operation. Fecal coliform removal was higher and MS2 removal was lower in block 1 compared to blocks 2 and 3, under both long and short operation. Turbidity removal was higher in block 1 compared to blocks 2 and 3 under long operation. Fecal coliform removal ranged from a high of 3.19 log (99.94%) (fine, 10 cm, block 1, week 3, long) to a low of 0.50 log (68.4%)(coarse, 30 cm, block 3, week 3, short). MS2 removal ranged from a high 1.55 log (97.2%) (fine, 30 cm, block 3, week 8, long) to a low of 0.32 log (109% increase)(coarse, 20 cm, block 3, week 8.2, long). Highest and lowest turbidity removals were 98.9 percent (coarse, 10 cm, block 1, week 7, long) and 62.8 percent (fine, 20 cm, block 3, week 3, short), respectively.
Table 3 e Filter operating characteristics. Block
Operating characteristic Days since last filter maintenance or start-up
Hours residence time e long batch
Hours residence time e short batch
1. Mean SD (N )a MineMax
24.1 15.8 (61) 8e58
16.6 4.0 (31) 12.33e23.75
7.9 6.1 (30) 2.17e21.75
2. Mean SD (N ) MineMax
22.8 14.5 (86) 8e67
14.6 3.0 (40) 12.17e24.62
4.2 1.4 (46) 2.27e6.83
3. Mean SD (N ) MineMax
26.9 16.9 (98) 8e67
15.7 4.1 (46) 12.17e26.92
4.3 1.4 (52) 2.32e6.83
Combined Mean SD (N ) MineMax
24.8 15.9 (245) 8e67
15.6 3.8 (117) 12.17e26.92
5.1 3.5 (128) 2.17e21.75
a N refers to the number of valid values measured during a block’s run.
6233
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 2 7 e6 2 3 9
Table 4 e Mean performancea across blocks of all filter configurations combined. Block
FC removal, Log10
MS2 removal, Log10
Turbidity removal, %
Effluent turbidity, NTU
Long batch outcomes 1. Mean SD (N )a MineMax
1.65 0.56 (32) 0.81e3.19
0.47 0.24 (31) 0.04e1.08
95.4 2.6 (32) 90.1e98.9
1.03 0.48 (32) 0.50e2.03
2. Mean SD (N ) MineMax
1.45 0.26 (41) 0.96e1.94
0.77 0.35 (40) 0.21e1.30
90.8 6.1 (44) 69.6e97.7
0.93 0.41 (44) 0.26e1.94
3. Mean SD (N ) MineMax
1.46 0.40 (46) 0.52e2.29
0.84 0.39 (46) 0.32 to 1.55
89.4 6.0 (50) 67.2e98.9
1.09 0.40 (50) 0.45e1.88
Combined blocks long batch Mean SD (N )
1.59 0.42 (119)
0.72 0.37 (117)
91.4 5.9 (126)
1.02 0.43 (126)
Short batch outcomes 1. Mean SD (N )a MineMax
1.60 0.41 (31) 0.85e2.53
0.15 0.19 (30) 0.16e0.64
89.5 6.2 (31) 77.1e96.9
1.45 0.61 (31) 0.55e2.59
2. Mean SD (N ) MineMax
1.23 0.24 (47) 0.72e1.85
0.42 0.42 (46) 0.31e1.21
88.5 5.4 (50) 75.1e96.6
1.30 0.46 (50) 0.58e2.71
3. Mean SD (N ) MineMax
1.18 0.31 (52) 0.50e1.78
0.46 0.40 (52) 0.22e1.13
83.5 7.4 (56) 62.8e93.8
1.56 0.53 (56) 0.55e2.79
Combined blocks short batch Mean SD (N )
1.30 0.36 (130)
0.38 0.39 (128)
86.7 7.0 (137)
1.44 0.53 (137)
All batches All blocks Mean SD (N )
1.40 0.40 (249)
0.54 0.42 (245)
89.0 6.9 (263)
1.24 0.53 (263)
a Based on measurements taken after 14 days of filter maturation and 7 days since filter maintenance.
3.3.
Modeling results
LMM multivariate modeling results for the 2-factor long batch operation model, the 2-factor short batch operation model, and the combined 3-factor model are shown in Tables 5 and 6. They provide systematic estimates of the independent marginal effect of a change in the sand grain size, hydraulic head, and batch operation (combined 3-factor model) on each performance outcome of interest: bacteria removal, viral removal, turbidity removal and effluent turbidity based on our selected indicators organisms and measures, while controlling for the effects of variations in observed operating characteristics of interest, namely, a batch’s actual residence time, days since maintenance, and influent turbidity. Table 5 presents the significance levels of the factors and covariates of each model for each outcome. Table 6 lists the marginal effect size of each factor and covariate or interaction term for effects with a significance level of p < 0.10 in the models shown in Table 5.
3.3.1. Significant factors and covariates affecting filter performance Under long batch operation, all five conditions: sand size, head, residence time, maintenance, and influent turbidity significantly ( p < 0.05) and independently affected filter performance for one or more of the four outcomes of concern (see Table 5: 2-factor long batch model, bold text cells). Residence time (deviation) was the most consistently and highly significant variable, followed by sand size. In addition to their independent main effects, sand size and residence time produced a significant interaction effect on bacterial removal under long operation ( p ¼ 0.004). Influent turbidity, included
in models of bacterial and viral removal, produced highly significant effects on viral removal, and weak effects on bacterial removal. Head was highly significant for bacterial removal, weakly significant for effluent turbidity, and insignificant for viral and turbidity removal under long operation. The significant factors and covariates of short batch performance differed from those for long batch performance (see Table 5: 2-factor short batch model, bold text cells). Residence time significantly affected bacterial removal and turbidity performance but not viral removal under short batch operation. Sand size was significant for short batch bacterial and viral removal but not for either of the two turbidity performance outcomes. Head effects on short batch operation were less significant for bacterial removal and more significant for viral and turbidity removal, compared to long batch operation. Maintenance effects were similarly important for both turbidity performance outcomes and bacterial removal under short and long operation. Lastly, influent turbidity was highly significant for viral removal and also significant for bacterial removal. No significant interactions were observed under short operation among the variables. The above patterns of factor and covariate significance are confirmed in the 3-factor model which examined the full data set and included batch (long versus short RT operation) as a third factor (Table 5). Batch operation and actual residence time both had highly significant independent effects on all four filter performance outcomes, and interacted to produce significant additional interaction effects on bacterial removal ( p < 0.001). Head changes were significant or near significant for all four outcomes, independent of batch or residence time. Sand size was significant or near significant for bacterial and viral removal, but not for turbidity performance (neither
6234
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 2 7 e6 2 3 9
Table 5 e Significance of fixed factor and covariate effects on filter performance ( bold [ p < 0.05; italics [ 0.05 < p < 0.10). Linear mixed model
a
Factor (F) or covariate (C)
ISSF performance outcome Log10 FC removal
Log10 MS2 removal
Turbidity % removal
Effluent turbidity
p-value
p-value
p-value
p-value
1
2
3
4
2-factor analysis (long batch)
(F) Sand (F) Head (C) Residence time (h)e (C) Days since maint.f (C) Influent turbidity
<0.001 <0.001 <0.001 0.188 0.396
0.691 0.646 <0.001 n/id <0.001
0.020 0.318 <0.001 0.037 NAc
0.019 0.105 0.167 0.090 NAc
2-factor analysis (short batch)
(F) Sand (F) Head (C) Residence time (h) (C) Days since maint. (C) Influent turbidity
0.011 0.108 <0.001 0.280 0.014
0.010 0.065 0.565 n/id <0.001
0.757 0.085 0.001 <0.001 NAc
0.370 0.184 0.004 0.126 NAc
3-factor analysisb (combined batches)
(F) Sand (F) Head (F) Batch (C) Residence time (h) (C) Days since maint. (C) Influent turbidity
<0.001 <0.001 <0.001 <0.001 0.121 0.011
0.071 0.048 <0.001 0.009 n/id <0.001
0.748 0.153 <0.001 <0.001 <0.001 NAc
0.172 0.200 <0.001 0.001 0.014 NAc
a For Log FC removal 2-factor long batch analysis, sand and residence time interaction was significant ( p ¼ 0.004). b In Log MS2 removal 3-factor analysis, sand and batch interaction was nearly significant ( p < 0.099) indicating that 0.17 mm sand provides relatively greater MS2 viral removal under short batch operation than it does under long batch operation. In Log FC removal 3-factor analysis, batch and residence time interaction was significant ( p < 0.001). c NA: not applicable to this performance outcome. d n/i: not included in final model due to lack of any bi-variate relationship. e Actual residence time adjustment in plus or minus hours, relative to the mean batch residence time for long or short batch. f Includes measurements for filters after 14 days since start-up or at least 7 days since maintenance.
removal nor effluent). Sand size and batch interacted to produce nearly significant additional effects on viral removal ( p < 0.099). Maintenance significantly affected both turbidity performance outcomes independent of other conditions, while influent turbidity had a significant independent effect on both viral and bacterial removal in the combined model. The following sections review the magnitude of the significant factor and covariate direct and interaction effects on ISSF performance shown in Table 6.
3.3.2.
Bacterial removal
Fine sand (d10 ¼ 0.17) increased bacterial removal (as measured by fecal coliform) by 0.16 log under short operation and 0.30 log under long operation, compared to coarse sand (d10 ¼ 0.52). Reducing head under long operation removed an additional 0.17 log at 20 cm and 0.29 log at 10 cm. Irrespective of operating mode, fine sand increased removal on average by 0.18 log independent of other conditions, while head reductions of 10 cm and 20 cm increased removal on average by 0.10 log and 0.16 log, respectively. Switching from short to long RT batch operation increased bacterial removal by 0.29 log on average, irrespective of configuration or operating conditions. Increasing residence time further increased bacterial removal by an estimated 0.050 and 0.063 log per hour, under long and short operation, respectively. The benefit from another hour of residence time occurred irrespective of sand size under short
operation but was largely limited to the fine sand configuration under long operation. Each additional NTU of influent turbidity increased bacterial removal by approximately 0.0035 log, irrespective of batch, configuration, or other conditions. Days since maintenance was positively associated with bacterial removal however the effect was not significant ( p ¼ 0.18).
3.3.3.
Viral removal
Operation mode alone, and through interaction with sand size, had the greatest impact on viral removal. A change from short to long RT operation increased viral removal by 0.36 log on average, irrespective of filter configuration. When changing to long operation in a fine sand filter, an additional 0.31 log was removed over and above the average change. An even larger add-on increase in removal of 0.41 log occurred when changing to long operation in a coarse sand filter, indicating a notably larger marginal gain in viral removal for long over short operation in coarse sand filters. Beyond operation mode and sand size-operation mode interaction effects, each additional hour of residence time further increased removal, independent of other conditions, by an average 0.012 log, with this effect more pronounced under long operation (0.025 log per hour). Reducing head or sand size each produced limited viral removal improvements, compared to operation mode. Head reduction from 30 to 10 cm produced a significant
6235
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 2 7 e6 2 3 9
Table 6 e Effect on ISSF performance of significant factor level differences and covariates. Performance outcome
Log FC removal
Significant factors & covariates ( p < 0.10 from Table 5)
(F) Sand (F) Head (F) Batch (C) Residence time (F*C) Sand & Res. time interaction (F*C) Batch and Res. Time interaction (C) Influent turbidity
Factor level paired comparison or covariate increment Better
Worse
Significant performance difference (for comparison or covariate p <0.10) 2-factor analysis
2-factor analysis
3-factor analysis
Short batch
Long batch
All batches
1
2
3
0.17 mm 10 cm 20 cm Long
0.52 mm 30 cm 30 cm Short 1 additional hour 1 additional hour e 0.17 mm sand 1 additional hour e 0.52 mm sand 1 additional hour e Long batch 1 additional hour e Short batch 1 additional NTU
0.16 NS NS n/a 0.064 NS NS n/a n/a 0.0039
0.30 0.29 0.17 n/a n/a 0.053 NS n/a n/a NS
0.18 0.16 0.10 0.29 n/a NS NS 0.050 0.063 0.0035
Log MS2 removal
(F) Sand (F) Head (F) Batch (F*F) Sand and batch interaction (C) Residence time (C) Influent Turbidity
0.17 mm 0.52 mm 10 cm 30 cm Long Short 0.17 mm and long 0.17 mm and short 0.52 mm and long 0.52 mm and short 1 additional hour 1 additional NTU
0.10 0.094b n/a n/a n/a NS 0.019
NS NS n/a n/a n/a 0.025 0.017
0.053a 0.082 0.36 0.31 0.41 0.012 0.017
Turbidity % removal
(F) Sand (F) Head (F) Batch (C) Residence time (C) Days since maint.
0.17 mm 10 cm Long
0.52 mm 30 cm Short 1 additional hour 1 additional day
NS 4.26c n/a 0.49 0.15
1.67 NS n/a 0.24 0.049
NS NS 3.85 0.46 0.11
Effluent turbidity (NTU)
(F) Sand (F) Batch (C) Residence time (C) Days since maint.
0.17 mm Long
0.52 mm Short 1 additional hour 1 additional day
NS n/a .034 NS
0.21 n/a NS .0038d
NS 0.40 0.026 0.0048
a b c d
Sand paired difference Bonfernoi significance ¼ 0.071. Head paired difference Bonferoni significance ¼ 0.080. Head paired difference Bonferoni significance ¼ 0.105. Days since maintenance marginal effect significance ¼ 0.090.
increase of 0.082 log, on average, in viral removal, independent of other conditions, with this association stronger under short RT operation. Fine sand produced an average increase of 0.053 log in viral removal over coarse sand, independent of other conditions, also more pronounced and larger for short RT operation. Viral removal improvements from head and sand size reductions, while positive, are considerably smaller than those achieved for bacterial removal. Influent turbidity had a consistently strong negative effect on viral removal: each additional NTU decreased viral removal by an average 0.017 log, independent of other conditions. Influent turbidity had a slightly greater negative effect on viral removal under short RT operation (0.019 log per NTU). The negative impact of influent turbidity on viral removal (0.017 log/NTU) is nearly 5 times the magnitude of the positive impact observed for bacterial removal (þ0.0035 log/NTU).
3.3.4.
Turbidity performance
Longer contact time, reflected in batch operation mode and residence time, produced the only consistent improvements in turbidity performance across the models, for both removal
and effluent quality. Changing from short to long RT operation increased turbidity removal by 3.85 percentage points and decreased effluent turbidity by 0.40 NTU, independent of filter configuration or other conditions. Each additional hour of residence time increased removal by 0.46 percentage points and decreased effluent turbidity by 0.026 NTU. The use of fine sand, under long RT operation only, caused a small detrimental impact on turbidity performance compared to the use of coarse sand, decreasing removal by 1.67 percentage points and increasing effluent turbidity by 0.21 NTU, on average. Reducing head from 30 cm to 10 cm, under short operation, produced a marginally significant ( p ¼ 0.105) improvement in removal of 4.26 percentage points but this did not translate into a significant improvement in effluent turbidity. Impacts of sand size and head on turbidity removal and effluent performance were no longer significant when controlling for batch and residence time deviations in the combined all-batch model. Each additional day since the last maintenance resulted in 0.11 percentage points more removal of influent turbidity and a 0.005 NTU reduction in effluent turbidity, on average.
6236
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 2 7 e6 2 3 9
4.
Discussion
4.1.
General overview
Relative to earlier work, several new and important aspects of ISSF performance are addressed in this study that allow for a more thorough evaluation and understanding of performance and removal mechanisms. This is the first study in which simultaneous measurement of bacteria, viruses and turbidity removal at the batch level over an extended duration has been undertaken with a large number of replicate units to permit rigorous modeling of ISSF performance variability within and across a range of likely BSF design configurations and operating conditions. Published laboratory evaluations of the BSF to date have involved testing of 1 or 2 filter units for limited durations and have not reported on maintenance or its effects. This study included 18 filter units operated daily and maintained and sampled over ten weeks. Over the course of filter testing, significant seasonal variation in natural surface water turbidity was encountered and routine wet harrowing of the filters was required to maintain filter operation, conditions typically representative of real-life fieldscale operation. Simultaneous measurement of turbidity removal in this study has revealed interactions between turbidity and microbiological removal in the ISSF which have not been reported and which have important implications for applications that treat surface waters with varying or high turbidity. The experimental filter in this study differs in several ways from the D-BSF design used to date in published controlled laboratory studies of the BSF. The D-BSF design calls for a 40 cm column of crushed sand, comprised of an upper layer of effective size and uniformity coefficient of 0.15 mm and 1, respectively, and a lower sand layer with effective size and uniformity coefficient of 0.35 mm and 1, respectively (Manz, 2004). This sand specification is often costly to obtain or produce locally in developing communities. In this study, a 60 cm column composed of a single layer of river sand manually processed to meet specified sand size and uniformity characteristics was used to better reflect typical community sand sources and processing capability in developing countries. The slightly deeper sand column was chosen to maintain a minimum 1:1 ratio of filter pore space to 20 L batch charge volume, and also differs from the D-BSF’s reported pore space to charge volume ratio of 0.9:1 (Stauber et al., 2006). Across the 18 filter units and 10-week test duration, including both long and short operation but excluding the first 2 weeks of start-up and first seven days after maintenance, on average the ISSF was capable of removing 96% of bacteria (1.4 log fecal coliform), 71% of viruses (0.54 log MS2), 89% of turbidity and produced effluent always below 3 NTU turbidity in each 20 L sampled batch of treated water (Table 4). Results reported from testing of the D-BSF (Buzunis, 1995; Elliott et al., 2008; Stauber et al., 2006) in which no maintenance was performed, fall within the range of our testing results, for the most part. In contrast to Elliott et al. (2008), however, we found viral shedding (negative viral removal) to be a real concern with ISSF performance. More importantly, the modeling analysis sheds light on those
design and operating conditions that are most responsible for improved ISSF performance, including uncontrollable operating conditions, at both filter and batch scale, and provides insight into mechanisms of removal that have the greatest effects on ISSF performance for each of the four outcomes measured. This study confirms that removal of bacteria and viruses in ISSFs is significantly lower than in SSFs. Even under the best performing design configuration and operating mode in this study (fine sand, 10 cm head, long operation, initial HLR of 0.01e0.03 m/h) which achieved mean 1.82 log removal of fecal coliform bacteria (98.5%) and mean 0.94 log removal of MS2 viruses (88.5%), these rates are 0.5e1.0 log lower than typically obtained by SSF (Hijnen et al., 2004). Viral shedding, which we found to be a problem with the ISSF, has also been observed in SSF challenge testing (Anderson et al., 2009). Anderson et al. report MS2 removal from 0.2 to 2.2 log for a range of typical SSF design and operating conditions. Recent work indicates iron amendments that enhance viral adsorption offer a potential solution for increasing viral removal in the BSF (Bradley et al., 2011).
4.2.
Contact time
Contact time appears to be one of the most critical factors for adequate removal of bacteria, turbidity and, particularly, viruses, in the ISSF. Both Table 4 and the LMM results highlight and quantify the independent and significant effects of long RT compared to short RT operation for all outcomes studied, irrespective of design configuration or other condition. On average, long RT operation can be expected to increase removal of bacteria by 0.29 log, viruses by 0.67e0.77 log (depending on sand size), turbidity by 3.85 percentage points, and reduce effluent turbidity by 0.40 NTU, relative to short RT operation. Additional benefits, beyond the gain from long RT operation to all four outcomes, accrue for each additional hour of contact time within the sand column beyond the operating average (15.6 h for long, 5.1 h for short) within the range studied (up to 27 h). If we consider the BSF as a plug flow batch reactor, having a pause time between feeds (long RT operation) appears important. The biological and physio-chemical processes within the sand column need sufficient time to clear pore spaces and biofilm adsorption sites loaded with contaminants, before the next batch is added. Indeed, for ISSF viral removal the most important controllable parameter of those tested appears to be batch contact time within the sand column reactor which is consistent with adsorption and attachment being the dominant mechanisms for successful viral removal in SSF (Anderson et al., 2009). The role of contact time in viral removal was also found by Elliott et al. (2011).
4.3. Importance of design factors and design and operation interactions The main mechanisms that are responsible for removal of bacteria, viruses and turbidity in SSF, namely mechanical straining, adsorption, attachment, and biological activity
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 2 7 e6 2 3 9
including predation, are also expected to operate in ISSF. In theory, smaller sand size slows down the rate of infiltration, reduces the size of pore space passages, and supports a larger biofilm surface area. Thus a smaller media would be expected to improve ISSF removal performance by increasing settling and straining, increasing contact (residence) time, and providing more attachment sites. Reduced ISSF nominal head, resulting in a lower hydraulic loading rate, reduces a batch’s maximum and average infiltration rate which allows for greater headspace settling, increases contact time, and reduces biofilm shear forces at the beginning of batch infiltration. This research helps quantify the extent to which these different potential benefits from reduced sand size and head, within feasible operating ranges for the ISSF, translate into significant improvements in removal performance. These potential improvements may merit changes in BSF design alone, or in conjunction with operating changes. The modeling analysis indicates that using fine rather than coarse sand size (d10 0.17 mm vs. d10 0.52 mm) yields a significant and meaningful increase in ISSF removal of bacteria (0.16e0.30 log), independent of the residence time effect from the sand size change, with less consistent effects observed for viruses (0e0.10 log). The combination of fine sand with long RT operation can be expected to produce a mean increase in removal of 0.63 log bacteria and 0.41 log viruses over mean levels achieved with the combination of coarse sand with short RT operation, irrespective of head or other conditions. Changing to a lower head, without changing the batch volume (i.e., keeping the pore:batch ratio constant at 1:1 in this case), provides relatively small mean improvements in treated water quality for bacteria and viruses. This suggests that the enhanced supernatant settling and reduced velocity shear effects of a lower head achieve relatively little on their own (i.e., independent of the increased residence time effects of reduced head which were controlled for separately in the LLM modeling). No benefit of sand size or head reduction on turbidity performance, separate from the increased residence time benefits, was observed.
4.4. Important and opposing effects of influent turbidity on viral and bacterial removal
the near-neutral pH of the influent water, causing repulsion by the negatively charged sand (Schijven and Hassanizadeh, 2000). Moreover, the bacteria were more likely to be particle associated due to their origin in the combined river water/ wastewater influent whereas the MS2 coliphage were cultured in the laboratory and thus not indigenous. The observed negative effect of turbidity on viral removal is a concern for ISSF performance in developing countries where this technology is needed the most because many BSF-target households tend to be more rural and depend on surface water sources that may experience high levels of seasonal or year round turbidity.
4.5.
Limited effects of maintenance on performance
This is the first time maintenance effects on ISSF performance have been evaluated systematically. The maintenance technique used in the experiments was wet harrowing which disturbs no more than the top 2 cm of the sand column. Consistent with SSF performance measured by Hijnen et al. (2004), we found no effect from disturbance of the schmutzdecke by wet harrowing on virus removal and a modest reductive effect on bacterial and turbidity removal seven days or more after a maintenance event. Lack of a maintenance effect on viral removal is consistent with adsorption and attachment within the whole sand column being the main removal mechanism for viruses and contact time the most important parameter affecting viral performance. Viruses are too small to be caught by the schmutzdecke sand surface mat so its disturbance by wet harrowing would be expected to have little or no impact on the retention of viruses. However, its disturbance would be expected to impact retention of larger sized bacteria and turbidity particles at the sand surface. Indeed, despite restricting performance results and removal values in the analysis to 7 or more days since a maintenance event, and replacement of missing values for days since maintenance with its average value, we found a small negative effect of filter maintenance on bacteria and turbidity performance.
4.6. This study uncovered an important and relatively large and significant negative effect of influent turbidity on viral removal in the ISSF, estimated as an average reduction in removal of 0.017e0.019 log for each additional NTU of influent turbidity. To our knowledge, this is the first time the relationship between influent turbidity and viral removal, whether for SSF or ISSF, has been systematically identified and measured. At the same time, we found an opposite positive effect, albeit much smaller, of influent turbidity on ISSF bacterial removal in which 1 NTU of influent turbidity enhanced bacterial removal by an estimated 0.0035 log, on average. These opposing effects are plausible if influent turbidity particles competed effectively with the MS2 coliphage for adsorption sites whereas those same particles provided additional sites for bacteria removal. Although surface charges were not measured in this research and virus adsorption in soil is a very complicated phenomenon, it is likely that MS2 coliphage had a strong net-negative charge at
6237
Other sources of variability in ISSF performance
Examination of the experimental data in Table 4 indicates significant variability in performance between blocks. Some of the block-level variation can be explained by observed differences in influent turbidity and maintenance practices as discussed above, however, there may be other important unobserved influent water quality physical and chemical characteristics, separately or in conjunction with turbidity, such as hardness, alkalinity or pH, that may account for block performance differences. Small unobserved differences in the characteristics of unprocessed river sand and the manual sand processing and packing of filters between block 1 and blocks 2 and 3, as suggested by the lower block 1 initial HLR values for fine sand may also have contributed to block-level variations in overall performance. These additional parameters that may affect filter performance variability should be fully considered and processes developed to address them within BSF implementation and construction projects,
6238
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 2 7 e6 2 3 9
especially those involving on-going large-scale filter production and distribution activities.
4.7. Implications for developing country design, local manufacturing, and operation Results of this research point to a number of recommendations for implementing BSFs in local settings in developing countries. A single layer of manually processed natural river sand, used in this experiment, appears to produce comparable levels of removal performance as the more expensive 2 layers of crushed manufactured sand recommended by the D-BSF design. If treated BSF water is to be consumed, fine sand (d10 in the 0.15 mm range) should be used instead of coarse sand (d10 in the 0.35e0.50 mm) to maximize bacterial and viral removal. Sand size has emerged as the most important design factor on performance in this study. Controlling sand size to achieve close to SSF fine sand characteristics (0.15 mm d10 and under 3 UC) and maintaining quality control of filter packing during installation to achieve the low hydraulic loading rates required for sufficient residence time may be the most important aspects of filter design and production for maximizing performance on all four outcomes of concern. Pause time and increased residence time have emerged as highly beneficial for improving removal performance on all four outcomes. Pause time may be particularly important for avoiding viral shedding by allowing sufficient time for adsorbed and attached viral particles to decay or be consumed before addition of the next batch. Thus, for a batch of treated ISSF water for drinking purposes, the residence time should be maximized in daily use routines by allowing for a pause time between feeds used for drinking. This implies instructing users to use either the morning batch for drinking if the filter idle time is longest at night, or the evening batch if the idle time is longest during the day, in households which require more than one 20 L batch a day. Given overall poor viral removal, a follow-up disinfection step is recommended. This may be especially necessary when deploying the BSF as a household technology to treat higher turbidity waters in settings where viral pathogens are a significant cause of water-borne illness. Simple options are likely to be highly efficacious in destroying residual bacteria and viruses in ISSF effluent. Modifications to current BSF containers to treat a 20 L batch under the low hydraulic loading rates associated with reduced heads tested in this study could involve costly and complex design changes. Table 1 illustrates how the same reduction in head produces a greater absolute reduction in initial hydraulic loading rate in the coarse sand experimental filters compared to the fine sand experimental filters. One very simple way to improve removal performance for existing coarse sand BioSand filters would be to reduce the batch feed volume from the container design volume, effectively reducing the operating head, increasing the pore:batch volume ratio, and in turn increasing residence time and reducing the likelihood of breakthrough. While benefits would likely accrue for fine sand as well, given the smaller magnitude decrease in initial HLR from a lower head, the marginal gain in performance would likely be smaller.
5.
Conclusions
Across the 18 filter units and 10-week test duration, including both long and short residence time operation, the ISSF was capable of removing 96% of bacteria (1.4 log fecal coliform), 71% of viruses (0.54 log MS2 bacteriophage), 89% of turbidity and produced effluent below 3 NTU turbidity in each 20 L sampled batch of treated water. A single layer of manually processed natural river sand achieved performance comparable to that obtained with more expensive 2 layers of crushed manufactured sand recommended in current BSF field design. Sand size emerged as the most important design factor in this study. Controlling sand size in filter production and maintaining quality control of filter packing during installation to achieve the low hydraulic loading rates required for sufficient residence time is recommended to enhance performance on all four outcomes. A reduction in sand size increased removal of indicator bacteria on average by 0.16e0.30 log, independent of the residence time effect of the sand size change, with a less consistent viral indicator removal improvement of 0e0.10 log. On average, long residence time operation in which a pause occurs between batch feeds can be expected to increase removal of bacteria by 0.29 log, viruses by 0.67e0.77 log, turbidity by 3.85 percentage points, and reduce effluent turbidity by 0.40 NTU, compared to short residence time operation in which no pause occurs. Pause time may be particularly important for avoiding viral shedding by allowing sufficient time for adsorbed and attached viral particles to decay or be consumed before addition of the next batch. The best design and operation combines fine sand with long residence time operation, delivering marginal increases in mean removal of 0.63 log bacteria and 0.72 log viruses compared to that achieved with coarse sand and short residence time operation, irrespective of head or other conditions. Additional benefits, beyond the mean gain from long relative to short residence time (pause vs. flush) operation, accrue to all four outcomes for each additional hour of contact time in the sand column beyond the operating average, within the range studied (up to 27 h). Each NTU increase of influent turbidity reduced MS2 viral removal in the ISSF by 0.017e0.019 log per NTU. Turbidity’s negative effect on viral removal is a concern for ISSF performance in developing countries and may necessitate a follow-on disinfection step in settings with high levels of seasonal or year round turbidity. Disturbance of the schmutzdecke by wet harrowing had no measurable effect on virus removal and only a modest reductive effect on bacterial and turbidity removal, based on measurements taken 7 days or more after the disturbance. However, maintenance reduces residence time which can be expected to reduce performance on all outcomes.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 2 7 e6 2 3 9
Acknowledgments Funding for this research was provided by USAID through the Global Livestock Collaborative Research Program at UC Davis with additional support provided by a National Science Foundation grant and the Gerald T. Orlob Professorship at UC Davis. Granite Construction of Sacramento, CA donated the river sand. Bill Fleenor and Bill Sluis provided invaluable guidance with filter construction and experimental set-up and operation.
references
Anderson, W.B., DeLoyde, J.L., Van Dyke, M.I., Huck, P.M., 2009. Influence of design and operating conditions on the removal of MS2 bacteriophage by pilot-scale multistage slow sand filtration. Journal of Water Supply Research and TechnologyAqua 58, 450e462. APHA, 2005. Standard Methods for the Examination of Water and Wastewater, 21th ed. American Public Health Association, Washington, DC. Baumgartner, J., Murcott, S., Ezzati, M., 2007. Reconsidering ‘appropriate technology’: the effects of operating conditions on the bacterial removal performance of two household drinkingwater filter systems. Environmental Research Letters 2. Bradley, I., Straub, A., Maraccini, P., Markazi, S., Nguyen, T.H., 2011. Iron oxide amended biosand filters for virus removal. Water Research 45, 4501e4510. Buzunis, B., 1995. Intermittently Operation Slow Sand Filtration: A New Water Treatment Process. Department of Civil Engineering. University of Calgary, Calgary, Alberta, Canada, p. 235. Clasen, T., 2009. Scaling Up Household Water Treatment Among Low-Income Populations. WHO/HSE/WSH/09.02. World Health Organization, Geneva. Duke, W., Nordin, N., Baker, D., Mazumder, A., 2006. The use and performance of BioSand filters in the Artibonite Valley of Haiti: a field study of 107 households. Rural and Remote Health 6. Earwaker, P., 2006. Evaluation of Household BioSand Filters in Ethiopia. Cranfield University, Silsoe, Bedfordshire, UK. Elliott, M.A., Stauber, C.E., Koksal, F., DiGiano, F.A., Sobsey, M.D., 2008. Reductions of E-coli, echovirus type 12 and bacteriophages in an intermittently operated household-scale slow sand filter. Water Research 42, 2662e2670. Elliott, M.A., DiGiano, F.A., Sobsey, M.D., 2011. Virus attenuation by microbial mechanisms during the idle time of a household slow sand filter. Water Research 45, 4092e4102. Faraway, J.J., 2006. Extending the Linear Model with R: Generalized Linear, Mixed Effects and Nonparametric Regression Models. Chapman & Hall/CRC, Boca Raton, FL. Fewster, E., Mol, A., Wiesent-Brandsma, C., 2004. The long term sustainability of household bio-sand filtration. In: PeopleCentered Approaches to Water and Environmental Sanitation: 30th WEDC International Conference, Vientiane, Lao, pp. 558e561.
6239
Hijnen, W.A.M., Schijven, J.F., Bonne, P., Visser, A., Medema, G.J., 2004. Elimination of viruses, bacteria and protozoan oocysts by slow sand filtration. Water Science and Technology 50, 147e154. Huisman, L., Wood, W.E., 1974. Slow Sand Filtration. World Health Organization, Geneva. Jenkins, M.W., Tiwari, S., Darby, J., Nyakash, N., Saenyi, W., Langenbach, K., 2009. The BioSand Filter for Improved Drinking Water Quality in High Risk Communities in the Njoro Watershed, Kenya. Research Brief 09-06-SUMAWA. Global Livestock Collaborative Research Support Program. University of California, Davis. Kaiser, N., Liang, K., Maertens, M., Snider, R., 2002. BioSand Household Water Filter Evaluation 2001: A Comprehensive Evaluation of the Samaritan’s Purse BioSand Filter (BSF) Projects in Kenya, Mozambique, Cambodia, Vietnam, Honduras, and Nicaragua. Samaritan’s Purse Canada, Calgary, AB. Kubare, M., Haarhoff, J., 2010. Rational design of domestic biosand filters. Journal of Water Supply Research and Technology-Aqua 59, 1e15. Manz, D.H., 2004. New horizons for slow sand filtration. In: The Eleventh Canadian National Conference and Second Policy Forum on Drinking Water and the Biennial Conference of the Federal-Provincial-Territorial Committee on Drinking Water, Promoting Public Health Through Safe Drinking Water, April 3e6, Calgary, Alberta, Canada, pp. 682e692. Montgomery, G.C., 2005. Design and Analysis of Experiments, fifth ed. John Wiley & Sons, New York. Palmateer, G., Manz, D., Jurkovic, A., McInnis, R., Unger, S., Kwan, K.K., Dutka, B.J., 1999. Toxicant and parasite challenge of Manz intermittent slow sand filter. Environmental Toxicology 14, 217e225. Schijven, J.F., Hassanizadeh, S.M., 2000. Virus removal by soil passage at field scale and groundwater protection of sandy aquifers. Water Science & Technology 46, 123e129. Sobsey, M.D., 2002. Managing Water in the Home: Accelerated Health Gains from Improved Water Supply. World Health Organization, Geneva. Sobsey, M.D., Stauber, C.E., Casanova, L.M., Brown, J.M., Elliott, M. A., 2008. Point of use household drinking water filtration: a practical, effective solution for providing sustained access to safe drinking water in the developing world. Environmental Science & Technology 42, 4261e4267. Stauber, C.E., Elliott, M.A., Koksal, F., Ortiz, G.M., DiGiano, F.A., Sobsey, M.D., 2006. Characterisation of the biosand filter for E. coli reductions from household drinking water under controlled laboratory and field use conditions. Water Science and Technology 54, 1e7. Verbeke, G., 2000. Linear Mixed Model for Longitudinal Data. Springer, New York. WHO, 2007. Combating Waterborne Disease at the Household Level. World Health Organization, Geneva. WHO/UNICEF, 2010. Progress on Sanitation and Drinking-water: 2010 Update. WHO/UNICEF Joint Monitoring Programme for Water Supply and Sanitation. World Health Organization, Geneva. Wiesent-Brandsma, C., Fewster, E., Mol, A., 2004. Medair Kenya BioSand Filter Project Evaluation e Interpretation of Results. Medair, Ecublen.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 4 0 e6 2 4 8
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Fast and precise method for HPLCesize exclusion chromatography with UV and TOC (NDIR) detection: Importance of multiple detectors to evaluate the characteristics of dissolved organic matter Nobuyuki Kawasaki a,*, Kazuo Matsushige a, Kazuhiro Komatsu a, Ayato Kohzu a, Fumiko Watanabe Nara b, Fumikazu Ogishi c, Masahito Yahata c, Hirohisa Mikami c, Takeshi Goto c, Akio Imai a a
National Institute for Environmental Studies, 16-2 Onogawa, Tsukuba, Ibaraki 305-8506, Japan Graduate School of Science, Tohoku University, 6-3 Aramaki Aza Aoba, Aoba-ku, Sendai 980-8578, Japan c Shimadzu Corporation, 1 Nishinokyo Kuwabaracho, Nakagyou-ku, Kyoto 604-8511, Japan b
article info
abstract
Article history:
A new type of high-performance liquid chromatography (HPLC)esize exclusion chroma-
Received 14 April 2011
tography (SEC) system with ultraviolet (UV) absorbance detection and non-dispersive
Received in revised form
infrared (NDIR) detection of total organic carbon is described. The introduction of an
8 September 2011
online degassing tube and a low-volume HPLC column helped to reduce the analytical time
Accepted 10 September 2011
and increase the sensitivity of the SEC system. This study is the first in which linear
Available online 17 September 2011
calibration curves (R2 > 0.99) were obtained for both UV absorbance and NDIR data for polystyrene sulfonate standards, which are the most suitable standards for molecular size
Keywords:
analysis of aquatic humic substances as well as dissolved organic matter (DOM). Using the
HPLC
calibration curves, the molecular size distribution of DOM in water collected from Lake
SEC
Kasumigaura and in pore water from lake sediments was estimated. Most of the DOM had
Apparent MW
a molecular weight less than 4000 Daltons (Da), and the amount of low-molecular-weight
DOM size
DOM (w2000 Da) with low UV absorbance increased with depth in the sediment pore water. This result shows the importance of combining quantitative analysis by NDIR detection with qualitative analysis by UV detection to determine the chemical and physical properties of DOM. The possible sources and reactivity of DOM in Lake Kasumigaura and its sediment pore water are also discussed. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
Dissolved organic matter (DOM) in natural aquatic systems is one of the largest active carbon reservoirs along with land plants and atmospheric carbon dioxide (Hedges, 1992). Increased attention to global carbon cycles in the last few decades has led
to increased interest in DOM in natural aquatic systems because temperature increases and increases in the frequency of severe droughts are likely to increase DOM concentrations steadily in river waters (Worrall et al., 2004). DOM in natural aquatic systems is regarded as a source of organic pollution, as an energy source for the microbe-based aquatic food web, as
* Corresponding author. Tel.: þ81 29 886 0970; fax: þ81 29 886 0938. E-mail address: [email protected] (N. Kawasaki). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.09.021
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 4 0 e6 2 4 8
a factor in the cycling of trace elements, and as an influence on the biological activity of phytoplankton and bacteria (Salonen et al., 1992). DOM affects the color of drinking water and leads to disinfection-byproduct formation (Amy et al., 1987) and membrane fouling during water treatment (Fonseca et al., 2007). Determining DOM sources and the factors regulating its production and degradation is a key to understanding both the carbon cycle and water quality in aquatic environments. The chemical and physical properties of DOM, such as composition and molecular size, appear to affect its reactivity and degradability. For example, in phytoplankton, more than 80% of organic matter is composed of amino acids, sugars, and lipids, whereas these components make up <10% of deep ocean DOM (Benner, 2002). DOM freshly produced by phytoplankton is easily degraded, whereas DOM collected in the deep ocean is highly refractory (Barber, 1968; Hansell and Carlson, 1998). The molecular size of DOM is also an important determinant of DOM reactivity. In the last decade or so, our understanding of the size and reactivity of DOM has been completely revised (Amon and Benner, 1996). For example, small DOM used to be considered reactive, and large DOM to be refractory, but larger molecules are currently regarded as much more reactive than smaller molecules. In addition, owing to advancements in analytical techniques, we now know that aquatic humic substances, once thought to be over 100,000 Da, are actually smaller than 1000 Da (Chin et al., 1994). The most widely used method for estimating the molecular size of DOM is size exclusion chromatography (SEC) with ultraviolet absorbance (UVA) detection. This method is relatively easy to determine the size of DOM, but since it only relies on UV absorbance, obtained results only show the qualitative properties of DOM because different chemical structures in organic carbon give different responses. Lowmolecular-weight organic acids, which are recognized as a main cause of membrane fouling (Speth et al., 1998), may not be detected by this system. Therefore, quantitative analyses of DOC are important to thoroughly understand the molecular size distribution of DOM. Huber and Frimmel (1991, 1996) developed a system that combines high-performance size exclusion chromatography with UV absorbance (HPSECeUVA) detection with a highly sensitive online organic carbon detector. Because this system allows detection of all organic carbon, the size of DOM can be estimated both qualitatively and quantitatively. Her et al. (2002, 2003) further improved the HPSECeUVA system with organic carbon detection. Now, the system has been developed with a combination of different carbon detectors. Besides the use of NDIR (Speth et al., 1998; Allpike et al., 2005) and conductivity (Allpike et al., 2007), Warton et al. (2008) designed to use a mass spectrometer for quantitative organic carbon detection, which would enable to measure carbon isotopes. Chin et al. (1994) first examined the combination of columns and mobile phases. They concluded that a mobile phase with ionic strength equivalent to 0.1 M NaCl and a pH 6.8 and a silica column would give the most optimal condition that the apparent molecular size by PSS was the closest to that of humic substance. However, for HPLCeUVA with organic carbon detector systems, the use of NaCl would cause the production of chlorine gas, harmful product, during the oxidation processes. Chlorine gas also causes destruction of
6241
NDIR detectors. Therefore, the use of NaCl is not suitable for HPLCeUVA with organic carbon detector systems. The use of phosphate buffer appears to be the best choice for mobile phase. The column used by Chin et al. (1994) was a silica-based column (Waters Protein-Pak 125). The silica-based columns are usually a very powerful SEC column to separate molecules. Therefore, they have been used for HPLCeSEC system with UVA detection for a long time. However, they may release organic matter from the column. The internal structure of the columns is packed with silica beads coated on organic silanol groups. These silanol groups are gradually peeled off from the silica beads, causing the contaminants for organic carbon detection. Allpike et al. (2007) reported that the carbon concentrations from the silica-based columns such as Tosoh TSKgel G3000SWXL were over the limit of their detector. The recent HPLCeSEC systems with UVA and organic carbon detection comprise of a preparatory column packed with Toyo Pearl resin as well as phosphate buffer (Huber and Frimmel, 1991) and numerous publications. The Toyo Pearl resin is known as clean resin with almost no carbon released. However, each Toyo Pearl resin is relatively large (20e40 mm) compared to silica beads (w5 mm) Therefore, the current system requires a relatively large sample volume (2 mL) and a long analytical time (>1 h). Our objectives in the current study were to develop a highly sensitive HPSEC system with both UV and non-dispersive infrared (NDIR) total organic carbon (TOC) detectors, a relatively low sample volume, and a fast analytical time and to determine possible source(s) of DOM at different sizes.
2.
Materials and methods
2.1.
Sample collections and preparation
Water and sediment samples were collected from Lake Kasumigaura (Fig. 1), a shallow (maximum depth ¼ 7 m) and highly eutrophic lake in Japan, during the monthly sampling carried out as part of the GEMS/Water Trend Monitoring Program of the National Institute for Environmental Studies (NIES) (GEMS/Water Website). Water samples were collected in a 1-L precombusted glass bottle (450 C, 4 h) with a 2-m column sampler on July 17, 2007. The samples were immediately cooled in an ice cooler. Sediment cores were collected with a Plexiglas core sampler (4-cm inner diameter, Rigosha, Tokyo) with a sharpened lower rim. The liner of the Plexiglas core was capped on top and bottom with a white rubber stopper (Unistopper, Tokyo) and kept in the dark at in situ temperature. After collection, all samples were immediately processed at the NIES Lake Kasumigaura Water Research Station. The water samples were gently filtered through precombusted Whatman GF/F filters (450 C, 4 h) using a vacuum pump and were then filtered through a 0.45-mm cartridge filter. Sediment pore water samples were obtained by sectioning of the sediments in a simple glovebox under continuous nitrogen gas flushing and centrifugation of the sectioned sediments at low temperature; cores were sectioned at 1e5-cm intervals with a stainless steel blade, and the sections were packed into 300-
6242
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 4 0 e6 2 4 8
Fig. 1 e Location of Lake Kasumigaura sampling site.
mL polycarbonate centrifuge tubes, purged with nitrogen gas, and centrifuged at 3500 rpm (2278 g) for 30 min at 4 C. The supernatants were carefully removed with a Teflon-coated syringe and filtered through a precombusted GF/F filter and then through a 0.45-mm cartridge filter. All samples were stored at 4 C or frozen at 30 C until analysis.
2.2.
Analytical setup
Molecular weights were determined with a homemade prototype high-performance liquid chromatography (HPLC)eSEC system with a UVA detector (SPD-6A Shimadzu) and an NDIR TOC detector (taken from a Shimadzu TOC-500), as shown in Fig. 2. The wavelength of the UV detector was set at 260 nm. UVA at 260 nm was selected because the maximum UVA of humic substances occurs at a wavelength of 200e290 nm (Buffle et al., 1982; Nippon-Kagaku-kai, 1977), and the presence of nitrate and borate does not affect the measurement of UVA at 260 nm (Nippon-Kagaku-kai, 1977). Only a minor difference of UVA between 254 nm and 260 nm was observed (data not shown). A mixture of disodium hydrogen phosphate (Na2HPO4$12H2O) and monobasic dihydrate (NaH2PO4$2H2O) was adjusted to pH 6.8 and used as a mobile phase. The concentration of phosphate and ionic strength of the mobile phase was 0.02 mol/L and 0.0335 M, respectively. This concentration
HPLC unit
Mobile phase
SEC column
H3PO4
Sample injection
of the mobile phase was chosen based upon closeness of the sample peaks and apparent molecular weight of UVA between this system and the standard method by Chin et al. (1994) after testing various concentrations of phosphate buffer (data not shown). The flow rate of the mobile phase and the sample volume were set at 0.5 mL/min, and 100 mL, respectively. The sample was introduced into an SEC column (Tosoh TSKgel G3000SWXL) for separation based on molecular size. After the eluent was passed through the UVA detector, the mobile phase was acidified to
Persulfate
UV oxidation unit NDIR detector
UV detector Vacuum pump Degassing unit
Waste
Fig. 2 e A schematic diagram of HPLCeSEC system used in this study.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 4 0 e6 2 4 8
Nitrogen gas was introduced to the inside of the cylinder to eliminate oxygen, which would have reacted with UV light to produce ozone. The cylinder was then covered with an aluminum box (6 cm 6 cm 60 cm). Before the oxidation unit, nitrogen gas was also introduced at a rate of 40 mL/min to help the eluent form a thin layer along the inside of the quartz tube so that DOC was effectively oxidized to CO2 during UV irradiation. The produced CO2 was introduced to the NDIR detector after all the liquid eluent was removed by means of a gaseliquid separator and then an electric cooler. The amount of organic carbon was then detected by the NDIR detector as the amount of produced CO2. For baseline noise for TOC or DOC measurements, two baselines were measured. Each baseline was divided by every minute. In every minute, the lowest and highest values were picked up. The difference between the highest and lowest values was averaged and determined as noise. Different concentrations of potassium phthalate (0.5, 1, 2, 3, 5 and 10 mg C/L) were prepared and measured to obtain height and area. Based upon these data, the minimum concentrations of DOC measurable for this system were determined. To examine oxidation ability of the SEC system, 5 different compounds (glucose, potassium phthalate, glutamic acid, urea, and thiourea) were prepared based upon relative ease of oxidation according to Chen and Wangersky (1993). Each compound was adjusted to be 10 mg C/L and measured for peak height and area. DOC quantitative measurements for field samples were conducted as non-purgeable DOC with a Shimadzu TOC-5000 total organic carbon analyzer equipped with a Pt catalyst on quartz wool. At least three measurements were made for each sample, and analytical precision was typically less than 1%. Potassium hydrogen phthalate (Kanto Chemical Co., Tokyo) was used as a standard.
2.3.
6243
the mobile phase. The silica-based columns are packed with silica beads whose surface is coated by organic compounds such as silanol groups. These compounds interact with water, causing breakup from the silica beads and are released to the mobile phase. The concentrations of organic carbon from columns were examined. DOC samples were collected from before and after the mobile phase passed through the column. The concentrations of DOC in the mobile phase were 0.62 (0.02) mg C/L (n ¼ 3) and 3.12 (0.06) mg C/L (n ¼ 3) before and after the column, respectively. These concentrations were comfortably within the range of normal NDIR detectors. The release of DOC from the column appeared to come from contamination in column, degradation of organic compounds coated on the silica beads or both. Since the columns we used were washed by the mobile phase at least 100 times as much as the volume of the column, any contaminants and excess release of organic carbon from the column appeared to be washed out. The release of organic matter from the column appeared to be constant after sufficient rinsing. The baselines of UVA and NDIR are shown in Fig. 3. The UV baseline is almost flat (Fig. 3A), while some noise was observed for the TOC baseline (Fig. 3B). To examine the noise, baselines are divided by every minute. In every minute, the lowest and highest values are identified and averaged. The average difference between the highest and lowest value is 0.39 (0.17) mV (n ¼ 35) and 0.43 (0.21) mV (n ¼ 35) for baselines 1 and 2, respectively, showing the average noise ranges w0.4 mV. This indicates that a TOC peak higher than 4 mV would be considered detectable if a signal to noise ratio for quantitative analyses is assumed to be 10. The peak heights and areas of different concentrations of potassium phthalate are shown in Table 1. The results indicated the height of 4 mV was equivalent to w0.2 mg C/L. This
Standards
Sodium polystyrene sulfonate (PSS) was used as a standard for molecular weight analysis. The MW range of the PSS was 1800e35,000 Da. Blue dextran was used as the standard for the exclusion limit (MW ¼ 2,000,000 Da), and acetone was used as the standard for the permeability limit (MW ¼ 58 Da). For oxidation efficiency test, five different compounds: potassium phthalate, glucose, glutamic acid, urea, and thiourea were also prepared. The concentrations of all standards were adjusted to be 5 mg C/L, which is similar to the DOC concentration in Lake Kasumigaura. All the standards and samples were spiked with a concentrated mixture of Na2HPO4$12H2O and NaH2PO4$2H2O (20) to adjust the ionic strength to that of the mobile phase.
3.
Results and discussion
3.1.
New SEC system
The purpose of establishing this SEC system was to provide the fast and precise analyses of both UVA and TOC detection. As pointed out by Allpike et al. (2007), silica-based columns including Tosoh TSKgel G3000SWXL release organic carbon into
Fig. 3 e Chromatograms of baselines. A) UVA and B) NDIR response.
6244
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 4 0 e6 2 4 8
Table 1 e Peak heights, areas area over height (A/H ) of different concentrations of potassium phthalate for TOC detection and their slopes and R2. Phthalate concentrations (mg C/L)
Area
Height (mV)
Area/height (A/H )
0.5 1 2 3 5 10
860 1801 3441 5082 8269 16,105
10.2 21.2 40.5 59.4 96.6 183.3
84.6 85.1 85.0 85.6 85.7 87.8
Slope (area/mg C) R2
1599 0.999
19.6 0.999
accounts for only w5% of DOC concentrations of Lake Kasumigaura. Therefore, sensitivity of carbon detection for our system would be sufficient for at least eutrophic lakes such as Lake Kasumigaura. The degassing tube was introduced to increase the speed and reliability of the SEC system for detection of TOC. The effectiveness of the tube at removing dissolved gas was tested at various concentrations of inorganic carbon (Fig. 4). The 7-mlong tube effectively removed inorganic carbon at concentrations 100 mg C/L. The inorganic carbon concentration in natural aquatic environments usually ranges from 5 to 20 mg C/L (Raven and Samuelsson, 1988; Surif and Raven, 1989). Therefore, our degassing tube would be suitable for removing inorganic carbon in normal aquatic conditions and it appears to have the same ability as commercially available degassing modules such as a Sievers TOC degassing module. Another goal of this study was to reduce the injection volume. According to Her et al. (2003), a sample volume of 2 mL is sufficient to detect 1 mg/L DOC with a 20 250 mm preparatory gel filtration column packed with TSK-50S. The fact that the peaks for our standards were sharp and high enough to be detected (Fig. 5) at a sample concentration of 5 mg C/L and an injection volume of 100 mL indicates that our system may be at least 4 times as sensitive as the system used by Her et al. This increased sensitivity was expected when using an analytical-grade column, as opposed to a preparatory column. The main way to increase the sensitivity of SEC systems is to reduce both the size of resin particles in the
Fig. 4 e Chromatogram of dissolved inorganic carbon at 100, 150, and 200 mg C/L.
Fig. 5 e Chromatogram of glucose, glutamic acid, potassium hydrogen phthalate, urea, and thiourea at 10 mg C/L.
column and the system volume, although other factors such as the sensitivity of the detectors are also important. The particle size of the resin in the TSK-50S column (Toyopearl HW-50; superfine grade, 30 mm) is much larger than that used for the SEC column used in this study (5 mm), and the resulting increase in surface area was the likely cause of the increased sensitivity. The system volume is the inside volume of the system through which the sample travels, including the lines, detectors, and columns. The volume of the SEC column is the largest contributor to the system volume, and the inside volume of the column used by Her et al. was about 6 times that of the analytical column used in our study (7.8 300 mm). The use of the low-volume SEC column greatly reduced the analytical time to only 35 min, compared to over 1 h for the system used by Her et al. (2003). A longer analytical time results in greater diffusion of the sample; and thus a shorter analytical time can be expected to lead to improve sensitivity. In our UV oxidation unit, the mobile phase was UVirradiated as a mixture of liquid and gas. Because the large, 25-W UV lamp produced a lot of heat, a large cooling unit was required for the gaseliquid separation. Using a lower-wattage UV lamp might be advantageous because lamp length decreases with wattage. Thus, the total volume of the quartz tube would be smaller, leading to reduction in the entire volume of the oxidation unit and, consequently, to a reduction in band broadening. The smallest UV lamp that retains sufficient oxidation ability should be used. Compared to high-temperature combustion oxidation (HTCO), UV and wet oxidation may underestimate DOC because some types of organic compounds resist oxidation. Chen and Wangersky (1993) tested various organic compounds and found that whereas the HTCO method almost completely oxidizes all compounds; neither UV nor wet oxidation oxidizes compounds such as urea and thiourea. Therefore, we tested the oxidation effectiveness of our SEC system with potassium phthalate, glucose, glutamic acid, urea, and thiourea. Comparison of the peak areas showed that the error was within 5% (Fig. 4). According to Chen and Wangersky (1993), the UV oxidation method cannot degrade all types of organic compounds. However, our results showed that UV oxidation combined with
6245
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 4 0 e6 2 4 8
chemical oxidation using persulfate degraded all the organic compounds they tested. It may also be possible to use a lowerwattage UV lamp while retaining the same oxidation ability. These points need to be further tested to reduce the sample volume and increase the sensitivity of the SEC system.
3.2.
TOC UVA
Standards and calibration curves
Our SEC system detected all the standards (Fig. 6). As expected, blue dextran, which has the highest molecular weight (MW), 2,000,000 Da, eluted first and was followed by the 35,000-, 18,000-, 8000-, 5400-, and 1800-Da PSS and acetone (MW ¼ 58). The calibration curves of apparent molecular weight (AMW) of PSS were estimated using retention time of each peak top as an x-axis and common logarithm of molecular weight as a y-axis. The calibration curves of UVA and TOC were calculated as follows: AMWUVA ¼ 10ð2:4010
3 Time
ðsÞþ6:53Þ
;
3 AMWTOC ¼ 10ð2:3510 Time ðsÞþ7:32Þ ;
where AMWUVA and AMWTOC were apparent molecular weight of UVA and TOC, respectively (Fig. 7). The slopes of both UVA and TOC were very linear (R2 > 0.998). Other researchers have used polyethylene glycols (PEGs) as standards because PSS can interact with some columns (Her et al.,
Fig. 7 e Calibration curves of standards for UVA detection (-) and NDIR detection (A). Both curves were linear for compounds with MWs between 1800 and 35,000 (R2 > 0.997).
2002). Chin et al. (1994) examined the relationship between the MWs of standards and samples and found that the molecular configuration of PSS was closer to that of aquatic humic substances, a major constituent of DOM in natural environments, than that of PEGs. Therefore, it is important to use PSS as a standard to estimate the molecular size of DOM. To our knowledge, ours is the first study to show highly linear calibration curves for both UV and NDIR DOC detection data in an SEC system with PSS as a standard.
A Acetone 1800 5400 8000 18000 35000 100000 2000000
B Acetone 1800 5400 8000 18000 35000 100000 2000000
Fig. 6 e Chromatograms of standards at 5 mg C/L: (A) UV absorbance detection and (B) NDIR detection.
6246
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 4 0 e6 2 4 8
3.3. Molecular size distribution and properties of DOM in Lake Kasumigaura and sediment pore water Using the calibration curves described in the previous section, we estimated the molecular size distribution of DOM in Lake Kasumigaura and its sediment pore water (Fig. 8). In all samples, five peaks were observed for UVA, while three peaks for TOC. The AMWUVA of peak top was estimated to be >35,000, 4500, 3700, 2800, and 2000 Da (Fig. 8A), and for NDIR DOC detection, the AMWTOC of peak top was observed to be >35,000, 3800, and 2000 Da (Fig. 8B). Comparison of the UVA- and NDIRdetected peaks indicates both similarities and differences. Both detectors exhibited a small peak for AMW > 35,000 Da. This observation is consistent with a recent finding that the AMW of most DOM is <3000 Da (Chin and Gschwend, 1991; Conte and Piccolo, 1999). The difference of DOC >35,000 Da between UVA and TOC was also found. While the area of UVA in the >35,000 Da peak accounted for only w1% of total area of UVA, the area of TOC in the >35,000 Da peak accounted for w15% in lake DOM. This indicates that DOM > 35,000 Da could come from microbial biopolymers which consist of low UVA DOM. Many phytoplankton are known to release exopolymers (Decho, 1990), and these compounds could be a main source of high-molecular-weight DOM in Lake Kasumigaura. This result also shows the importance of multiple detectors since only the
Fig. 8 e Chromatograms of samples from Lake Kasumigaura and its sediment pore water at different depths: (A) UV absorbance detection and (B) NDIR detection.
Table 2 e Concentrations of DOC in Lake Kasumigaura and its pore waters. Sample Lake Kasumigaura Pore water (0e1 cm) Pore water (1e2 cm) Pore water (2e4 cm) Pore water (4e6 cm) Pore water (6e8 cm) Pore water (8e10 cm) Pore water (10e15 cm)
DOC concentration (mg C/L) 4.3 5.3 5.4 5.6 5.5 5.6 5.6 6.5
combination of UVA and TOC detectors could determine the characteristics and possible sources of DOM > 35,000 Da. For UVA detection, the peak shapes were mostly similar for the various samples, and the dominant peak had an AMW between 3700 and 4500 Da. However, the size distribution for NDIR detection was different. The abundance for AMW 3800 Da was similar for all the samples, but that for AMW 2000 Da increased with sediment depth. Because the UVA peak for AMW 2000 Da was relatively small and constant at different depths, this small DOM appeared to be composed of organic compounds having low UVA. TOC measurements of the sediment pore water and lake water showed that the DOC concentration also increased with depth (Table 2). This result suggests that a low-molecular-weight DOM with low UVA accumulated in the pore water. Organic acids include humic substances that possess high UV absorptivity, whereas the UV absorptivity of hydrophilic organic acids is usually low (Imai et al., 2001; Swietlik and Sikorska, 2006). The pore water collected from sediments in Lake Kasumigaura has previously been examined by National Institute for Environmental Studies (NIES); the ratio of hydrophilic organic acids to humic substances augmented from 0.52 to 1.74 as the sediment depth was increased from 0e2 cm to 10e12 cm (Imai; person. commun.). The source of the increased DOM in the pore water might be particulate organic matter in sediment; DOM could be released from the organic matter as intermediate products of suboxic or anoxic degradation. Under anoxic conditions, the mineralization of organic matter occurs by a sequence of processes starting with hydrolysis, proceeding to fermentation, and ending either with respiration to produce carbon dioxide or with production of methane (Fenchel and Findlay, 1995). Bru¨chert and Arnosti (2003) examined each step using a novel multicell continuous flow assembly and found that hydrolysis of organic matter is much faster than the other steps and that, as a result, by-products accumulate in pore water. The gradual increase of DOM with depth may be the result of different oxidation conditions at different depths. For example, redox conditions determine which electron acceptors are used by heterotrophic bacteria to oxidize organic matter: oxygen is used under aerobic conditions: nitrate, manganese, and iron are used under anoxic conditions; and sulfate and CO2 are used under anaerobic conditions. The energy yield decreases in going from aerobic to anaerobic conditions, which indicates that the degradation of by-products would slow down as the amount of available oxygen decreased. Thus, the slowed degradation would result in a gradual increase of DOM in pore water. Although the reactivity as well as chemical property of the accumulated DOM
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 4 0 e6 2 4 8
is largely unknown, DOM from the sediments could be the source of refractory and hydrophilic low-molecular-weight DOM (w500 Da) in shallow eutrophic lakes such as Lake Kasumigaura. Microbial activity is likely to control the production and concentration of DOM in sediment pore water.
4.
Conclusion
A new type of size exclusion chromatography system with both UVA and total organic carbon (TOC) detectors has been developed to provide important information about molecular size distributions of dissolved organic matter qualitatively and quantitatively. Our newly developed provided reasonably precise analyses of size distribution of DOM with a relatively short time (w35 min) and low sample volume (100 mL) as compared to other similar systems, which usually need a long analytical time (>90 min) and large sample volume (w2 mL). Our system uses a combination of Shimadzu HPLC system and prototype units of UV light reactor and NDIR detector (these prototype units of our system will be on the market by Shimadzu Co.). It can be widely applied for different aquatic environments such as lake, river, groundwater, and soil water as well as water treatment facilities. In Lake Kasumigaura, DOM > 35,000 Da could mainly come from microbial biopolymers because of very low UVA. In sediment pore water, the increase in DOM concentrations with depth resulted in the increase of DOM with a size ranging w2000 Da. These results showed that the combination of UVA and TOC detectors could be a powerful tool to determine the characteristics and source(s) of DOM in aquatic environments.
Acknowledgments This work was supported by Environmental Technology Research Fund from Ministry of the Environment, Government of Japan (2006e2007) and by Grant-in-Aid for Scientific Research (No. 21241008) from the Japan Society for the Promotion of Science. Sampling was supported by the GEMS/ Water Trend Monitoring Project at Lake Kasumigaura. We thank the members of the project for their cooperation.
references
Allpike, B.P., Heitz, A., Joll, C.A., Kagi, R.I., 2005. Size exclusion chromatography to characterize DOC removal in drinking water treatment. Environmental Science and Technology 39 (7), 2334e2342. Allpike, B.P., Heitz, A., Joll, C.A., Kagi, R.I., 2007. A new organic carbon detector for size exclusion chromatography. Journal of Chromatography A 1157 (1e2), 472e476. Amon, R.M.W., Benner, R., 1996. Bacterial utilization of different size classes of dissolved organic matter. Limnology and Oceanography 41 (1), 41e51. Amy, G.L., Collins, M.R., Kuo, C.J., King, P.H., 1987. Comparing gelpermeation chromatography and ultrafiltration for the molecular-weight characterization of aquatic organic-matter. Journal of American Water Works Association 79 (1), 43e49.
6247
Barber, R.T., 1968. Dissolved organic carbon from deep waters resists microbial oxidation. Nature 220 (5164), 274. Benner, R., 2002. Chemical composition and reactivity. In: Hansell, D.A., Carlson, C.A. (Eds.), Biogeochemistry of Marine Dissolved Organic Matter. Academic Press, London, pp. 59e90. Bru¨chert, V., Arnosti, C., 2003. Anaerobic carbon transformation: experimental studies with flow-through cells. Marine Chemistry 80 (2e3), 171e183. Buffle, J., Delandoety, P., Zumstein, J., Haerdi, W., 1982. Analysis and characterization of natural organic matters in freshwaters. I. Study of analytical techniques. Schweizerische Zeitschrift fur Hydrologie-Swiss Journal of Hydrology 44, 325e362. Chen, W., Wangersky, P.J., 1993. A high-temperature catalyticoxidation method for the determination of marine dissolved organic-carbon and its comparison with the UV photooxidation method. Marine Chemistry 42 (2), 95e106. Chin, Y.P., Gschwend, P.M., 1991. The abundance, distribution, and configuration of porewater organic colloids in recent sediments. Geochimica et Cosmochimica Acta 55 (5), 1309e1317. Chin, Y.P., Aiken, G., O’Loughlin, E., 1994. Molecular-weight, polydispersity, and spectroscopic properties of aquatic humic substances. Environmental Science and Technology 28 (11), 1853e1858. Conte, P., Piccolo, A., 1999. High pressure size exclusion chromatography (HPSEC) of humic substances: molecular sizes, analytical parameters, and column performance. Chemosphere 38 (3), 517e528. Decho, A.W., 1990. Microbial exopolymer secretions in ocean environments e their role(s) in food webs and marine processes. Oceanography and Marine Biology 28, 75e153. Fenchel, T.M., Findlay, B.J., 1995. Ecology and Evolution in Anoxic Worlds. Oxford University Press, Oxford, U.K. Fonseca, A.C., Summers, R.S., Greenberg, A.R., Hernandez, M.T., 2007. Extra-cellular polysaccharides, soluble microbial products, and natural organic matter impact on nanofiltration membranes flux decline. Environmental Science and Technology 41 (7), 2491e2497. GEMS/Water Website, http://www.gems-water.org/index.html. Hansell, D.A., Carlson, C.A., 1998. Deep-ocean gradients in the concentration of dissolved organic carbon. Nature 395 (6699), 263e266. Hedges, J.I., 1992. Global biogeochemical cycles e progress and problems. Marine Chemistry 39 (1e3), 67e93. Her, N., Amy, G., Foss, D., Cho, J., Yoon, Y., Kosenka, P., 2002. Optimization of method for detecting and characterizing NOM by HPLCesize exclusion chromatography with UV and on-line DOC detection. Environmental Science and Technology 36 (5), 1069e1076. Her, N., Amy, G., McKnight, D., Sohn, J., Yoon, Y., 2003. Characterization of DOM as a function of MW by fluorescence EEM and HPLCeSEC using UVA, DOC, and fluorescence detection. Water Research 37 (17), 4295e4303. Huber, S.A., Frimmel, F.H., 1991. Flow-injection analysis of organic and inorganic carbon in the low-ppb range. Analytical Chemistry 63 (19), 2122e2130. Huber, S.A., Frimmel, F.H., 1996. Gelchromatographie mit Kohlenstoffdetektion (LC-OCD): Ein rasches und aussagekra¨ftiges Verfahren zur Charakterisierung hydrophiler organischer Wasserinhaltsstoffe. Vom Wasser 86, 277e290. Imai, A., Fukushima, T., Matsushige, K., Kim, Y.H., 2001. Fractionation and characterization of dissolved organic matter in a shallow eutrophic lake, its inflowing rivers, and other organic matter sources. Water Research 35 (17), 4019e4028.
6248
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 4 0 e6 2 4 8
Nippon-Kagaku-kai, 1977. Structure of organic compounds II. In Shin-Jikken-Kagaku-Kouza 13, Maruzen, Tokyo (in Japanese). Raven, J.A., Samuelsson, G., 1988. Ecophysiology of Fucus vesiculosus L. close to its northern limit in the Gulf of Bothnia. Botanica Marina 31 (5), 399e410. Salonen, K., Kairesalo, T., Jones, R.I. (Eds.), 1992. Dissolved Organic Matter in Lacustrine Ecosystems: Energy Source and System Regulator (Developments in Hydrobiology). Kluwer Academic Publishers, Dordrecht [Reprinted from Hydrobiologia vol. 229]. Speth, T.F., Summers, S.R., Gusses, A.M., 1998. Nanofiltration foulants from a treated surface water. Environmental Science and Technology 32 (22), 3612e3617. Surif, M.B., Raven, J.A., 1989. Exogenous inorganic carbon-sources for photosynthesis in seawater by members of the Fucales and
the Laminariales (Phaeophyta) e ecological and taxonomic implications. Oecologia 78 (1), 97e105. Swietlik, J., Sikorska, E., 2006. Characterization of natural organic matter fractions by high pressure size-exclusion chromatography, specific UV absorbance and total luminescence spectroscopy. Polish Journal of Environmental Studies 15 (1), 145e153. Warton, B., Heitz, A., Allpike, B.P., Kagi, R.I., 2008. Size-exclusion chromatography with organic carbon detection using a mass spectrometer. Journal of Chromatography A 1207 (1e2), 186e189. Worrall, F., Harriman, R., Evans, C.D., Watts, C.D., Adamson, J., Neal, C., Tipping, E.D., Burt, T., Grieve, I., Monteith, D., Naden, P.S., Nisbet, T., Reynolds, B., Stevens, P., 2004. Trends in dissolved organic carbon in UK rivers and lakes. Biogeochemistry 70 (3), 369e402.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 4 9 e6 2 5 8
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Enantiospecific fate of ibuprofen, ketoprofen and naproxen in a laboratory-scale membrane bioreactor N.H. Hashim a,b, L.D. Nghiem c, R.M. Stuetz a, S.J. Khan a,* a
UNSW Water Research Centre, School of Civil and Environmental Engineering, University of New South Wales, NSW 2052, Australia University of Tun Hussein Onn Malaysia, Malaysia c Strategic Water Infrastructure Laboratory, School of Civil Mining and Environmental Engineering, The University of Wollongong, NSW 2522, Australia b
article info
abstract
Article history:
The enantiospecific fate of three common pharmaceuticals was monitored in a laboratory-
Received 10 June 2011
scale membrane bioreactor (MBR). The MBR was operated with a hydraulic retention time
Received in revised form
of 24 h and a mixed liquor suspended solids concentration of 8.6e10 g/L. Standard solu-
8 September 2011
tions of ibuprofen, ketoprofen and naproxen were dosed into the synthetic feed of the MBR.
Accepted 10 September 2011
Influent and permeate samples were then collected for enantiospecific analysis. The
Available online 17 September 2011
individual (R)- and (S )-enantiomers of the three pharmaceuticals were derivatised using a chiral derivatizing agent to form pairs of diastereomers, which could then be separated
Keywords:
and analysed by gas chromatographyetandem mass spectrometry (GCeMS/MS). Accurate
Chiral analysis
quantitation of individual enantiomers was undertaken by an isotope dilution process. By
Enantiomers
comparing the total concentration (as the sum of the two enantiomers) in the MBR influent
Non-steroidal anti-inflammatory
and permeate, ibuprofen, ketoprofen and naproxen concentrations were observed to have
drugs (NSAIDs)
been reduced as much as 99%, 43% and 68%, respectively. Furthermore, evidence of
Profens
enantioselective biodegradation was observed for all three pharmaceuticals. (S )-Ibuprofen
Gas chromatographyetandem mass
was shown to be preferentially degraded compared to (R)-ibuprofen with an average
spectrometry
decrease in enantiomeric fraction (EF ) from 0.52 to 0.39. In contrast, (R)-ketoprofen was preferentially degraded compared to (S )-ketoprofen with a relatively minor increase in EF from 0.52 to 0.63. The use of a relatively pure enantiomeric solution of (S )-naproxen resulted in a significant change in EF from 0.99 to 0.65. However, this experiment consistently revealed significantly increased concentrations of (R)-naproxen during MBR treatment. It is hypothesised that the source of this (R)-naproxen was the enantiomeric inversion of (S )-naproxen. Such enantiomeric inversion of chiral pharmaceuticals during wastewater treatment processes has not previously been reported. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
The presence of pharmaceutically active compounds in the environment has been an increasingly active area of investigation throughout the last decade (Halling-Sorensen et al.,
1998; Heberer, 2002; Jones et al., 2005; Khetan and Collins, 2007; Ziylan and Ince, 2011). Many of the pharmaceuticals found in the aquatic environment are chiral chemicals; possessing at least one element of asymmetry and leading to the existence of two or more stereoisomers called enantiomers.
* Corresponding author. Tel.: þ61 2 93855070; fax: þ61 2 93138624. E-mail addresses: [email protected] (N.H. Hashim), [email protected] (L.D. Nghiem), [email protected] (R.M. Stuetz), [email protected] (S.J. Khan). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.09.020
6250
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 4 9 e6 2 5 8
Some pharmaceuticals are dispensed and consumed as mixtures of enantiomers, while others are dispensed as relatively pure single enantiomers (Murakami, 2007). Enantiomers commonly differ in their biological and pharmacological activity as a result of their stereo-selective interaction with enzymes or other chiral molecules (Caldwell, 1995). Consequently, the relative enantiomeric composition of some chiral pharmaceuticals changes during human metabolism (de Oliveira et al., 2005). Similarly, changes in relative enantiomeric composition have been observed during biological treatment processes in wastewater treatment plants (WWTPs) (Fono and Sedlak, 2005; KasprzykHordern et al., 2010; MacLeod et al., 2007; MacLeod and Wong, 2010; Matamoros et al., 2009; Nikolai et al., 2006). Enantiomers often have very different toxicity compared to each other (Stanley et al., 2006). In some cases, this may be an important consideration for environmental risk assessment since actual ecotoxicity implications will be highly dependent on the precise enantiomeric composition. Many international studies have reported on the effectiveness of wastewater treatment processes for removing various chiral pharmaceuticals (Joss et al., 2005; Kim et al., 2007; Rosal et al., 2010). However, the vast majority of these studies have been undertaken using achiral analytical methods, which do not differentiate between enantiomeric pairs. Although it is comparatively rarely reported, enantiospecific analysis of pharmaceuticals in wastewater and environmental samples has several important potential applications. For example, a number of authors have observed that the enantiomeric composition of specific chemicals has the potential to be a useful marker of biologically mediated degradation (Hashim et al., 2010; Kasprzyk-Hordern, 2010; Wong, 2006). In contrast to the alternative approach of measuring absolute concentrations before and after biological degradation, attention to the relative abundance of enantiomers has the advantages of not being affected by sample volume, short-term concentration fluctuations or variable extraction recoveries of different chemicals. Abiotic treatment processes such as physical settling, UV irradiation and membrane filtration are generally assumed to affect both enantiomers in equal proportion (Wong, 2006). Accordingly, microbial processes are presumed to be fully responsible for observed enantioselective degradation. Quantitative description of relative compositions of (R)and (S )-enantiomers is usually described as an enantiomeric ratio (ER) or an enantiomeric fraction (EF ) (Harner et al., 2000). ER has most commonly been expressed as the ratio of (þ)-enantiomer over the ()-enantiomer. For some chromatographic separation studies, for which the enantiomeric conformation is unknown, ER has been defined as the ratio of the first eluting enantiomer over the second eluting enantiomer. EF is conventionally described as the concentration fraction of the (þ)-enantiomer contributing to the total concentration of racemic mixture (Harner et al., 2000; Hashim et al., 2010). Enantiospecific monitoring of the fate of pharmaceuticals during biological wastewater treatment and environmental residence has revealed that the pharmaceuticals ibuprofen, naproxen and propranolol do exhibit enantioselective
degradation (Buser et al., 1999; Fono and Sedlak, 2005; MacLeod et al., 2007; Matamoros et al., 2009; Nikolai et al., 2006). On the other hand, non-enantioselective degradation has been observed for numerous other chiral pharmaceuticals including atenolol, fluoxetine, metoprolol, nadolol, sotalol, citalopram, salbutamol, 4-methylenedioxymethamphetamine and venlafaxine (Kasprzyk-Hordern et al., 2010; MacLeod et al., 2007; Nikolai et al., 2006). In all reported enantiospecific studies of the fate of chiral pharmaceuticals during wastewater treatment, observed changes in ER or EF have been interpreted as having been caused solely by enantioselective biodegradation. The possibility of changes in ER or EF having been (partially) caused by interconversion of one enantiomer to the other has not previously been seriously considered. In order to more precisely investigate the enantiospecific fate of pharmaceuticals, the experiments described in this study were undertaken under stable controlled conditions in a laboratory-scale membrane bioreactor (MBR). The pharmaceuticals ibuprofen, naproxen and ketoprofen, were introduced to the feed of the MBR in precisely known enantiomeric composition. In particular, the use of enantiomerically pure (S )-naproxen enabled the observation of the appearance of (R)-naproxen during the biological treatment process.
2.
Materials and method
2.1.
Chemicals and consumables
Racemic ibuprofen, racemic ketoprofen, enantiomerically pure (S )-naproxen (99%), (R)-1-phenylethylamine (PEA) (99.5%), triethylamine (TEA) and ethyl chloroformate (ECF) were purchased from SigmaeAldrich (St. Louis, MO, USA). Racemic (a-methyl-D3)-ibuprofen (D3-ibuprofen), (a-methylD3)-naproxen (D3-naproxen) and (a-methyl-D3)-ketoprofen (D3-ketoprofen) were purchased from CDN Isotopes Inc., Canada. HPLC grade acetonitrile and methanol were purchased from Ajax Finechem (Tarron Point, NSW, Australia). Analytical grade ethyl acetate was purchased from Fisher Scientific, Australia. Kimble culture tubes (13 mm I.D. 100 mm) were purchased from Biolab (Clayton, Vic. Australia). Two sizes of Oasis hydrophilicelipophilic balance (HLB) solid phase extraction (SPE) cartridges (6 cc, 500 mg and 1 cc, 30 mg) were purchased from Waters (Rydalmere, NSW, Australia). Whatman filter papers (0.75 mm) were purchased from Millipore, Australia. The Oasis HLB sorbent is an achiral macroporous copolymer of two monomers, the lipophilic divinylbenzene and the hydrophilic N-vinylpyrrolidone.
2.2.
Laboratory-scale MBR system
The laboratory-scale MBR system and the MBR experimental protocol used in this study have been previously described in detail (Alturki et al., 2010; Tadkaew et al., 2011, 2010). This MBR consisted of a glass reactor with an active volume of 9 L, a continuous mixer, two air pumps, a pressure sensor, and influent and permeate pumps. Two ZeeWeed-1 (ZW-1) submerged hollow fibre ultrafiltration membrane modules supplied by Zenon Environmental (Ontario, Canada) were
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 4 9 e6 2 5 8
used in this apparatus. This membrane has a nominal pore size of 0.04 mm. Each module had an effective membrane surface area of 0.047 m2. An electrical magnetic air pump (Heilea, model ACO 012) with a maximum air flow rate of 150 L/min was used to aerate the MBR system via a diffuser located at the bottom of the reactor. Dissolved oxygen concentration in the reactor was monitored daily and kept constant at 2 1 mg/L by controlling the aeration flow rate. Another small air pump was also used to provide a constant air flow rate through the membrane module to reduce fouling and cake formation. Transmembrane pressure was continuously monitored using a high resolution pressure sensor (0.1 kPa) which was connected to a personal computer for data recording. The temperature of the reactor was kept constant using a chiller/heater (Neslab RTE 7) equipped with a stainless steel heat exchanging coil. The personal computer also operated the permeate peristaltic pump on a 14 min suction and 1 min rest cycle to provide relaxation time to the membrane modules. The flow rate of the influent pump was matched to that of the permeate pump to maintain a constant reactor volume. The continuous mixer ensured homogeneous mixing of the liquor and prevented the settling of biomass.
2.3.
MBR experimental protocol
Synthetic wastewater was used in this study to simulate municipal sewage. The concentrated synthetic wastewater was prepared and stored in a refrigerator at 4 C. It was then diluted with Milli-Q 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). The reactor was seeded with activated sludge from the Wollongong sewage treatment plant (NSW, Australia). After the initial start-up process which lasted about two 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 permeate flux of 4.3 L/m2 h (or 6.7 mL/min). The MBR reactor temperature was kept constant at 20.0 0.1 C. Performance of the MBR system with regard to basic water quality parameters was then monitored for an extended period of more than 6 months. The measured parameters included total organic carbon removal (98.5e99.2%), total nitrogen removal (66e97%), mixed liquor suspended solids concentration (8.6e10 g/L), effluent total organic carbon (3.0e7.5 mg/L), mixed liquor pH (7.4e7.6), effluent conductivity (540e577 mS/cm) and effluent turbidity (<0.2). Once stable operation had been achieved, samples of the synthetic feed, mixed liquor and MBR permeate were collected and subjected to achiral analysis of ibuprofen, ketoprofen and naproxen using a high performance liquid chromatographyetandem mass spectrometry (HPLCeMS/MS) method described elsewhere (Alturki et al., 2010). None of the three analytes were present in any of these samples above the analytical detection limit of 5 ng/L. The three pharmaceuticals were then continuously introduced to the feed solution to achieve a constant (non-enantiospecific) concentration of approximately 2 mg L1 of each. The chemical analysis of the influent samples confirmed the accuracy and consistency of this dosing process throughout the duration of the
6251
experiment. The feed solution was kept in a stainless steel reservoir at air-conditioned room temperature (20 2 C). Duplicate feed and permeate samples (500 mL each) were collected once per week over four weeks indicated by S1, S2, S3 and S4.
2.4.
Sample extraction and chiral derivatisation
Enantiospecific analysis was undertaken according to a previously described and validated method (Hashim and Khan, 2011). However, a few minor modifications were made to the extraction method in order to accommodate the coanalysis of a larger number of trace chemicals by an additional non-enantioselective analytical method (data available elsewhere (Alturki et al., 2010)). Sample extraction was performed immediately after collection on 500 mg HLB SPE cartridges that had been preconditioned with tert-butyl methyl ether (MTBE) (5 mL), methanol (5 mL) and deionised water (5 mL). Samples were then spiked with a standard solution of isotopically labelled versions of each analyte (50 ng) for isotope dilution quantitation. These isotopically labelled standards were racemic D3-ibuprofen, D3-ketoprofen and D3-naproxen. All samples were then loaded onto the SPE cartridges under vacuum and maintained with a constant flow rate at 15 mL/min, rinsed with 5 mL reagent water, and finally dried under a gentle flow of nitrogen gas for 30 min. Analytes were eluted from the SPE cartridges with 5 mL of methanol followed by 5 mL of 1/9 (v/v) methanol/MTBE into Kimble culture tubes. The resulting extracts were concentrated using vacuum-assisted evaporation to approximately 100 mL. The extracts were brought to a final volume of 1 mL with methanol. Non-enantioselective analysis of these methanol samples was then undertaken by HPLCeMS/MS for the concurrent study (Alturki et al., 2010). In preparation for chiral derivatisation, the methanol solutions were again evaporated dryness and the samples reconstituted in acetonitrile (300 mL). Samples were sonicated at 30 C for an hour in order to dissolve the dried materials on the bottom of the culture tubes. All samples were then subjected to chiral derivatisation as previously described (Hashim and Khan, 2011). This derivatisation process is based on the conversion of the three chiral pharmaceuticals to amide diastereomers, which can subsequently be separated by achiral gas chromatography. Triethylamine (50 mM, 30 mL) and ethyl chloroformate (60 mM, 40 mL) were added to all samples and this mixture was sonicated for 2 min. Subsequently, (R)-1-phenylethylamine (0.5 M, 40 mL) was added and sonication was repeated for a further 2 min. Finally, sulphuric acid (100 mL, 0.1 M) and ultrapure water (3 mL) were added to stop the reaction. A second SPE procedure was used to extract the derivatised pharmaceuticals from the derivatisation reaction solution prior to GCeMS/MS analysis. This SPE process was undertaken on Oasis HLB cartridges (1 mL, 10 mg) that had been preconditioned with ethyl acetate (1 mL), methanol (1 mL) and ultrapure water adjusted to pH 9.5 (1 mL). The 3 mL aqueous solutions from the derivatisation step were passed through the SPE cartridges under gravity, without the assistance of any applied pressure or vacuum. The cartridges were then rinsed twice with ultrapure water (1 mL) adjusted to pH 9.5. The SPE
6252
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 4 9 e6 2 5 8
cartridges then were dried under a gentle flow of nitrogen for 30 min. The derivatised pharmaceuticals were eluted from the cartridges with ethyl acetate (1 mL) directly to 2 mL GC autosampler vials prior to GCeMS/MS analysis.
2.5.
GCeMS/MS analysis
The separation of the amide diastereomers was performed on an Agilent 7890A gas chromatograph (GC) equipped with an Agilent 7693 autosampler and split/splitless injector. Identification of separated analytes was accomplished by an Agilent 7000B triple quadrupole tandem mass spectrometer (MS/MS). The method used for GCeMS/MS analysis has been previously described (Hashim and Khan, 2011). Briefly, the enantioseparations of analytes were performed on a HP5-MS fused silica capillary column (30 m 0.25 mm I.D. 0.25 mm film thickness) with 0.8 mL/min helium flow. The injector, interface and
source temperature were 270 C, 260 C and 280 C, respectively. 1 mL samples were injected in splitless mode with a purge delay of 1 min. GC oven temperatures were programmed initially at 120 C for 1 min then increased by 40 C min1 to 240 C and finally by 5 C min1 to 300 C and maintained for 4 min. Total run time was 18 min per sample. Mass spectrometric ionisation was undertaken in electron ionisation (EI) mode with an EI voltage of 70 eV. The triple quadrupole MS detector was operated in multiple reaction monitoring (MRM) mode with the gain set to 100 for all analytes. Three MRM transitions were monitored for all pharmaceuticals and their isotope-labelled internal standards as shown in Table 1. Concentrations of (R)- and (S )-enantiomers of pharmaceuticals were determined from 6-point standard calibration curves prepared with standard quantities of (S )-naproxen (1, 10, 50, 100, 500 and 1000 ng) and racemic ibuprofen and
Table 1 e Molecular structures of analytes, their isotope labelled standards and optimal analyte-dependent parameters for tandem mass spectrometry. Analyte
Molecular structure of derivatised 2-APAs
Precursor ion (amu)
Product ions (amu) and collision energy (eV) shown in parentheses
(R)-Ibuprofen-(R)-1-PEA (S )-Ibuprofen-(R)-1-PEA 309.2
161.2(10)
119.1(20)
105.0(35)
312.0
164.0(15)
122.1(30)
105.0(30)
333.2
185.1(10)
171.1(35)
105.0(35)
336.0
188.1(15)
171.1(30)
105.0(30)
357.3
210.2(5)
120.1(10)
105.0(25)
360.0
213.2(5)
120.1(15)
105.0(30)
(R)-D3-Ibuprofen-(R)-1-PEA (S )-D3-Ibuprofen-(R)-1-PEA
(R)-Naproxen-(R)-1-PEA (S )-Naproxen-(R)-1-PEA
CH3
H3C
O
CH3
O
(R)-D3-Naproxen-(R)-1-PEA (S )-D3-Naproxen-(R)-1-PEA
CD3 H N
*
H3C
(R)-Ketoprofen-(R)-1-PEA (S )-Ketoprofen-(R)-1-PEA
H N
*
O
CH3
O
O
CH3 *
H N
O
(R)-D3-Ketoprofen-(R)-1-PEA (S )-D3-Ketoprofen-(R)-1-PEA
O
CD3 H N
∗
O
*Chiral centre of the 2-APA analyte. PEA: Phenylethylamide.
CH3
CH3
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 4 9 e6 2 5 8
ketoprofen (2, 20, 100, 200, 1000 and 2000 ng). In all cases, these calibration solutions also contained 50 ng of racemic D3ibuprofen, D3-ketoprofen and D3-naproxen for isotope dilution quantification. All calibration standards underwent derivatisation followed by extraction from the derivatisation solution by the ‘SPE 2’ process. Final eluent volumes were approximately 1 mL in ethyl acetate. The isotope dilution quantification process accounts for variable derivatisation performance, extraction recovery and final eluent volume as previously described (Hashim and Khan, 2011). Statistical analysis has previously been reported to confirm the veracity of using a calibration curves developed for the (S )-enantiomers to quantify the (R)-enantiomers of these analytes (Hashim and Khan, 2011). This approach was required since no commercially available source of (R)-naproxen could be identified. All calibration curves showed good linearity with regression coefficients for all three (S )-pharmaceuticals and two (R)-pharmaceuticals above 0.999 in all sample batches. Method detection limits (MDLs) for the enantiomeric analysis of these chemicals in synthetic MBR permeates were established and previously reported for (S )-ibuprofen (0.7 ng/L), (S )-naproxen (0.7 ng/L) and (S )-ketoprofen (2.2 ng/L) (Hashim and Khan, 2011). The EF for ibuprofen, naproxen and ketoprofen were calculated, based on the mean values of duplicate samples, according to Equation (1) (Harner et al., 2000). EF ¼
3.
½ðSÞ enantiomer ½ðSÞ enantiomer þ ½ðRÞ enantiomer
(1)
Results and discussion
Concentrations of each of the (R)- and (S )-enantiomers were determined for both MBR influent and permeate at each sampling event. Total concentrations of ibuprofen, naproxen and ketoprofen were calculated as the sum of both (R)- and (S )-enantiomers.
6253
In all sampling events, total concentration of ibuprofen was found to be around 1.7e1.9 mg/L in the MBR influent and reduced more than 99% to less than 9 ng/L in the MBR permeate. Total concentrations of naproxen were observed to have been reduced by about 38e47% from the influent concentrations of 2.0e2.4 mg/L, while total ketoprofen was reduced by around 67e70% from influent concentrations of 1.4e1.7 mg/L. Comparable removal efficiencies for ibuprofen (97%), naproxen (40%) and ketoprofen (71%) have previously been reported from non-enantioselective investigations involving the same laboratory-scale MBR system (Alturki et al., 2010; Tadkaew et al., 2011). Analytical reproducibility was confirmed to be very good with measured concentrations of 47 of the 48 duplicate sample pairs (3 profens 2 enantiomers 2 sample types 4 sample events) showing less than 10% variability from their mean value. Only one duplicate pair ((S )-ibuprofen in MBR permeate from sampling event S4) exhibited slightly greater variability of 13% from their mean value. Specific enantiomeric compositions of (R)- and (S )ibuprofen in the MBR influent were determined to have an EF ¼ 0.50e0.54 as presented in Fig. 1. This reflects the racemic mixture of the analytical standard used to spike ibuprofen to the reactor. The ibuprofen EF was observed to be significantly reduced during MBR treatment, with average permeate EF values ranging from 0.31 to 0.44. Furthermore, very high removal efficiencies (>99%) for both (R)- and (S )-enantiomers were achieved for ibuprofen. A notable observation is that the ibuprofen EF in MBR permeates appeared to increase somewhat during the first three weekly sampling periods, S1eS3 (Fig. 1). The significance of this observation is currently unknown. However the possibility of varying rates of metabolism and production of the ibuprofen enantiomers, as a consequence of microbial adaptation, has previously been suggested (Hanlon et al., 1994). (S )-Ibuprofen has previously been reported to be present in enantiomeric excess in raw wastewaters since ibuprofen is known to be enantioselectively metabolised with the (S )enantiomer being excreted in greater concentration than the
Fig. 1 e Enantiospecific concentrations (ng/L) and enantiomeric fraction (EF ) of ibuprofen from four sampling events. Error bars show the upper and lower values from duplicate samples.
6254
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 4 9 e6 2 5 8
Fig. 2 e Enantiospecific concentrations (ng/L) and enantiomeric fraction (EF ) of ketoprofen from four sampling events. Error bars show the upper and lower values from duplicate samples.
(R)-enantiomer (Nicoll-Griffith et al., 1988). The initial excess of the (S )-enantiomer would normally facilitate observations of EF change during treatment processes. Although racemic ibuprofen was used in the feed solution of the current experiment, the reduced EF during treatment could still be observed. Previous studies from real wastewater treatment processes have reported influent ibuprofen EF in the range of 0.84e0.95 to 0.47e0.67 in the effluent (Buser et al., 1999) and 0.73e0.90 in influent to 0.60e0.76 in effluent (Matamoros et al., 2009). This enantioselective degradation appears to be specific to aerobic biodegradation conditions and was not observed under anaerobic conditions (Matamoros et al., 2009). For ketoprofen, a relatively minor increase in EF was observed from MBR influent to permeate in this study. This change from EF ¼ 0.51e0.53 in influent to EF ¼ 0.60e0.66 in the
permeate is presented in Fig. 2. As much as 74e78% of (R)ketoprofen and 58e64% of (S )-ketoprofen were removed from the initial influent solutions. No previously published data was identified for an enantioselective degradation study of ketoprofen in full-scale wastewater treatment processes. However, the relatively minor change in ketoprofen EF during the current experiments suggest that aerobic biodegradation of ketoprofen may not be as highly enantioselective as ibuprofen. Unlike ibuprofen and ketoprofen, naproxen was spiked as an almost pure solution of (S )-naproxen with only a very small trace of (R)-naproxen in the analytical standard (EF ¼ 0.995). Accordingly, an EF value of 0.99 was observed in the influent. This was significantly reduced to EF ¼ 0.65e0.66 in the permeate as shown in Fig. 3.
Fig. 3 e Enantiospecific concentrations (ng/L) and enantiomeric fraction (EF ) of naproxen from four sampling events. Error bars show the upper and lower values from duplicate samples.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 4 9 e6 2 5 8
The most surprising observation to emerge from the EF data during the enantiodegradation of naproxen was the increase in absolute concentration of (R)-naproxen in the permeate compared to the influent for all four sampling events. As an example, two MRM chromatograms of (R)- and (S )-naproxen in MBR influent and permeate are shown in Fig. 4. Among all of the samples collected, the concentration of (R)-naproxen was found to be 25e33 times higher in the MBR permeate than in the feed. This measured increase was significantly greater than the variability detected among duplicate samples, and was observed consistently for all samples taken. Conversely, (S )-naproxen was removed by 59e65%, leading to a total naproxen removal of 38e47%. The observed increase in (R)-naproxen concentration is consistent with observations from a closed bottle biodegradation test undertaken as a preliminary investigation to the
6255
current study (Hashim et al., 2011). In that investigation, the concentration of (R)-naproxen was observed to have been increased approximately five-fold during a 10 day incubation experiment. These observations support the hypothesis that (R)-naproxen may be produced as a result of enantiomeric interconversion of (S )-naproxen. To the authors’ knowledge, such enantiomeric interconversion has not been previously reported in laboratory or environmental biodegradation studies. Naproxen is known to be metabolised in the human body to o-desmethyl-naproxen, which is excreted in urine either unchanged or conjugated with glucoronic acid (Aresta et al., 2005). No other human metabolite or product of microbial degradation of naproxen during wastewater treatment has been reported (Quintana et al., 2005; Selke et al., 2010). It has been reported that o-desmethyl-naproxen is present in
Fig. 4 e Example of MRM chromatogram of naproxen in MBR (a) influent and (b) permeate.
6256
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 4 9 e6 2 5 8
effluents and surface waters with an excess of the (S )-enantiomer (Selke et al., 2010). Accordingly, deconjugation and transformation of conjugated o-desmethyl-naproxen appears to be an unlikely alternative source of the increased (R)-naproxen concentrations observed in this experiment. A number of studies have shown that 2-APAs undergo in vivo chiral inversion during mammalian metabolism, as previously reviewed by Wsol et al. (2004) and Ali et al. (2007). The most intensively studied is ibuprofen, which has been reported to undergo unidirectional (R) to (S ) inversion during metabolism by many mammalian species including humans (Hao et al., 2005). However, it is notable that some evidence has also been reported for a small degree of ibuprofen inversion in the reverse direction ((S ) to (R)) (Shirley et al., 1994). Ketoprofen inversion (R) to (S ) has been reported in many species, and while the reverse inversion is rare, it has also been observed (Wsol et al., 2004). Mammalian chiral inversion of naproxen is generally not considered to be significant with only very minor (R) to (S ) inversion having been reported (Iwakawa et al., 1991). Some species of fungi have also been shown to induce metabolic chiral inversions of 2-APAs including ibuprofen (Hanlon et al., 1994; Hutt et al., 1993). Mechanistic investigations of fungal chiral inversion suggest that the biochemical mechanism is an enzymatic process closely related, but not identical to that observed in mammalian studies (Thomason et al., 1997). Accordingly, most studies have observed the most common (R) to (S ) inversion direction. More recently, a type of actinomycetes bacterium was shown to produce enzymes, which invert the chirality of 2-APAs from the (S ) to the (R)-configuration (Kato et al., 2002; Mitsukura et al., 2002). Investigations into the mechanism of this bacterial inversion are, as yet, inconclusive, but multiple enzymes are believed to play a role (Kato et al., 2003, 2004). Evidence presented suggests that the chiral inversion of the (S )-enantiomers proceeds via enzyme-catalysed formation of an ‘activated’ coenzyme A derivative, followed by epimerisation to yield the (R)-enantiomer and, finally hydrolysis of the (R)-acyl-coenzyme A ester. The exploitation of such enzymatic processes for the commercial synthesis of enantiomerically purified substances is of growing interest in the field of synthetic organic chemistry (Gruber et al., 2006). It appears possible that a similar bacterially mediated transformation may have been responsible for the appearance of (R)-naproxen in the current study. The formation of (R)-naproxen from (S )-naproxen was previously suspected after (R)-naproxen was observed in wastewater effluents from Hamburg, Germany, but only (S )naproxen could be detected in surface waters in Karachi, Pakistan, where no sewage treatment processes exist (Selke et al., 2010). Transformation of the deuterated standards used for accurate quantitation in the analytical method was also considered as a potential source of (R)-naproxen. However, these standards are tri-deuterated at the b-methyl group of each of the 2-APAs, which has been shown to be completely retained during mammalian (Baillie et al., 1989), fungal (Thomason et al., 1997) and bacterially (Kato et al., 2002) induced chiral inversion experiments. Furthermore, (R)-naproxen has never been observed in any of the many control
blanks that we have run and to which D3-naproxen has been added prior to extraction. Naproxen EF has previously been nominated as a potentially good tracer to indicate the effectiveness of the drug removal during both anaerobic and aerobic biodegradation conditions (Matamoros et al., 2009). The authors of that study reported a useful correlation between effluent naproxen EF and overall removal efficiency of naproxen. However, according to that correlation, the effluent EF observed in the current study (0.65e0.66) would suggest a removal efficiency of almost 100%. The actual removal efficiency was, in fact, less than 50%. This indicates that the relationship between permeate naproxen EF and naproxen biodegradation may not be as robust as previously suggested. Potential sources of the variability between systems may include variable influent EF values and the enantiomeric interconversion proposed from this study. A potentially more robust measure may be the change in EF observed from influent to effluent samples.
4.
Conclusion
Laboratory-scale MBR experiments were undertaken to observe potential enantioselective biodegradation of three pharmaceuticals. Measurable enantioselective biodegradation was observed for all three pharmaceuticals. For ibuprofen, (S )-ibuprofen was preferentially degraded compared to (R)-ibuprofen with calculated average EF shown to decrease from 0.52 to 0.39 after the MBR treatment. A comparatively minor increase in EF for ketoprofen was observed during MBR treatment, with average EF 0.52 in the influent and 0.63 in the permeate. The EF of naproxen was found to decrease from 0.99 to 0.65 after MBR treatment. However, in this case, the change in EF was observed to be partly due to an increase in (R)-naproxen concentration during treatment. It is hypothesised that this increase in (R)naproxen occurred as a result of enantiomeric interconversion of (S )-naproxen during MBR treatment. Such enantiomeric interconversion has not been clearly identified in previous studies of wastewater treatment of chiral pharmaceuticals. The results of this study suggest that monitoring the enantiomeric fraction of some carefully selected pharmaceuticals in municipal wastewaters may be an effective means of gaining insights to the overall treatment performance of some biological wastewater treatment processes.
Acknowledgement This work was supported by the Australian Research Council (ARC) Discovery Project DP0772864. The authors thank the Ministry of Higher Education, Malaysia for sponsorship of Nor H. Hashim as well as Dr James McDonald from UNSW for laboratory and technical support with the undertaking of this work. The authors also thank Dr Nichanan Tadkaew and Mr Bob Rowlan from the University of Wollongong for their laboratory assistance.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 4 9 e6 2 5 8
references
Ali, I., Gupta, V.K., Aboul-Enein, H.Y., Singh, P., Sharma, B., 2007. Role of racemization in optically active drugs development. Chirality 19, 453e463. Alturki, A.A., Tadkaew, N., McDonald, J.A., Khan, S.J., Price, W.E., Nghiem, L.D., 2010. Combining MBR and NF/RO membrane filtration for the removal of trace organics in indirect potable water reuse applications. Journal of Membrane Science 365, 206e215. Aresta, A., Palmisano, F., Zambonin, C.G., 2005. Determination of naproxen in human urine by solid-phase microextraction coupled to liquid chromatography. Journal of Pharmaceutical and Biomedical Analysis 39 (3e4), 643e647. Baillie, T.A., Adams, W.J., Kaiser, D.G., Olanoff, L.S., Halstead, G. W., Harpootlian, H., Van Giessen, G.J., 1989. Mechanistic studies of the metabolic chiral inversion of (R)-ibuprofen in humans. Journal of Pharmacology and Experimental Therapeutics 249 (2), 517e523. Buser, H.-R., Poiger, T., Mu¨ller, M.D., 1999. Occurrence and environmental behaviour of the chiral pharmaceutical drug ibuprofen in surface waters and in wastewater. Environmental Science and Technology 33, 2529e2535. Caldwell, J., 1995. Stereochemical determinants of the nature and consequences of drug metabolism. Journal of Chromatography A 694 (1), 39e48. de Oliveira, A.R.M., Cesarino, E.J., Bonato, P.S., 2005. Solidphase microextraction and chiral HPLC analysis of ibuprofen in urine. Journal of Chromatography B 818 (2), 285e291. Fono, L.J., Sedlak, D.L., 2005. Use of the chiral pharmaceutical propranolol to identify sewage discharges into surface waters. Environmental Science and Technology 39 (23), 9244e9252. Gruber, C.C., Lavandera, I., Faber, K., Kroutil, W.A.S.C., 2006. From a racemate to a single enantiomer: deracemization by stereoinversion. Advanced Synthesis and Catalysis 348 (14), 1789e1805. Halling-Sorensen, B., Nielsen, S.N., Lanzky, P.F., Ingerslev, F., Lutzhoft, H.C.H., Jorgensen, S.E., 1998. Occurrence, fate and effects of pharmaceutical substances in the environment e a review. Chemosphere 36 (2), 357e393. Hanlon, G.W., Kooloobandi, A., Hutt, A.J., 1994. Microbial metabolism of 2-arylpropionic acids: effect of environment on the metabolism of ibuprofen by Verticillium lecanii. The Journal of Applied Bacteriology 76 (5), 442e447. Hao, H.P., Wang, G.J., Sun, J.G., 2005. Enantioselective pharmacokinetics of ibuprofen and involved mechanisms. Drug Metabolism Reviews 37 (1), 215e234. Harner, T., Wiberg, K., Norstrom, R., 2000. Enantiomer fractions are preferred to enantiomer ratios for describing chiral signatures in environmental analysis. Environmental Science and Technology 34, 218. Hashim, N.H., Khan, S.J., 2011. Enantioselective analysis of ibuprofen, ketoprofen and naproxen in wastewater and environmental water samples. Journal of Chromatography A 1218, 4746e4754. Hashim, N.H., Shafie, S., Khan, S.J., 2010. Enantiomeric fraction as an indicator of pharmaceutical biotransformation during wastewater treatment and in the environment e a review. Environmental Technology 31 (12), 1349e1370. Hashim, N.H., Shafie, S., Khan, S.J., 2011. Enantioselective degradation of ibuprofen, ketoprofen and naproxen in laboratory scale biodegradation study, 7th IWA specialist conference on Assessment and Control of Micropollutants/ Hazardous Substances in Water, Sydney, NSW, Australia, 11e13 July 2011.
6257
Heberer, T., 2002. Occurrence, fate, and removal of pharmaceutical residues in the aquatic environment: a review of recent research data. Toxicology Letters 131 (1e2), 5e17. Hutt, A.J., Kooloobandi, A., Hanlon, G.W., 1993. Microbial metabolism of 2-arylpropionic acids e chiral inversion of ibuprofen and 2-phenylpropionic acid. Chirality 5 (8), 596e601. Iwakawa, S., Spahn, H., Benet, L.Z., Lin, E.T., 1991. Stereoselective disposition of carprofen, flunoxaprofen and naproxen in rats. Drug Metabolism and Disposition 19 (5), 853e857. Jones, O.A.H., Voulvoulis, N., Lester, J.N., 2005. Human pharmaceuticals in wastewater treatment processes. Critical Reviews in Environmental Science and Technology 35 (4), 401e427. Joss, A., Keller, E., Alder, A.C., Gobel, A., McArdell, C.S., Ternes, T., Siegrist, H., 2005. Removal of pharmaceuticals and fragrances in biological wastewater treatment. Water Research 39, 3139e3152. Kasprzyk-Hordern, B., 2010. Pharmacologically active compounds in the environment and their chirality. Chemical Society Reviews, 4466e4503. Kasprzyk-Hordern, B., Kondakal, V.V.R., Baker, D.R., 2010. Enantiomeric analysis of drugs of abuse in wastewater by chiral liquid chromatography coupled with tandem mass spectrometry. Journal of Chromatography A 1217, 4575e4586. Kato, D., Mitsuda, S., Ohta, H., 2002. Microbial deracemization of alpha-substituted carboxylic acids. Organic Letters 4 (3), 371e373. Kato, D., Mitsuda, S., Ohta, H., 2003. Microbial deracemization of alpha-substituted carboxylic acids: substrate specificity and mechanistic investigation. The Journal of Organic Chemistry 68 (19), 7234e7242. Kato, D., Miyamoto, K., Ohta, H., 2004. Microbial deracemization of alpha-substituted carboxylic acids: control of the reaction path. Tetrahedron Asymmetry 15 (18), 2965e2973. Khetan, S.K., Collins, T.J., 2007. Human pharmaceuticals in the aquatic environment: a challenge to green chemistry. Chemical Reviews 107 (6), 2319e2364. Kim, S.D., Cho, J., Kim, I.S., Vanderford, J.B., Snyder, S.A., 2007. Occurrence and removal of pharmaceuticals and endocrine disruptors in South Korean surface, drinking and waste waters. Water Research 41, 1013e1021. MacLeod, S.L., Sudhir, P., Wong, C.S., 2007. Stereoisomer analysis of wastewater-derived [beta]-blockers, selective serotonin reuptake inhibitors, and salbutamol by high-performance liquid chromatography-tandem mass spectrometry. Journal of Chromatography A 1170 (1e2), 23e33. MacLeod, S.L., Wong, C.S., 2010. Loadings, trends, comparisons, and fate of achiral and chiral pharmaceuticals in wastewaters from urban tertiary and rural aerated lagoon treatments. Water Research 44, 533e544. Matamoros, V., Hijosa, M., Bayona, J.M., 2009. Assessment of the pharmaceutical active compounds removal in wastewater treatment systems at enantiomeric level. Ibuprofen and naproxen. Chemosphere 75 (2), 200e205. Mitsukura, K., Yoshida, T., Nagasawa, T., 2002. Synthesis of (R)-2phenylpropanoic acid from its racemate through an isomerase-involving reaction by Nocardia diaphanozonaria. Biotechnology Letters 24 (19), 1615e1621. Murakami, H., 2007. From racemates to single enantiomers-chiral synthetic drugs over the last 20 years. Topics in Current Chemistry 269, 273e299. Nicoll-Griffith, D.A., Inaba, T., Tang, B.K., Kalow, W., 1988. Method to determine the enantiomers of ibuprofen from human urine by high-performance liquid chromatography. Journal of Chromatography B 428, 103e112. Nikolai, L.N., McClure, E.L., MacLeod, S.L., Wong, C.S., 2006. Stereoisomer quantification of the beta-blocker drugs atenolol, metoprolol, and propranolol in wastewaters by chiral
6258
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 4 9 e6 2 5 8
high-performance liquid chromatography-tandem mass spectrometry. Journal of Chromatography A 1131 (1e2), 103e109. 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 (12), 2654e2664. Rosal, R., Rodriguez, A., Perdigon-Melon, J.A., Petre, A., GarciaCalvo, E., Gomez, M.J., Aguera, A., Fernandez-Alba, A.R., 2010. Occurrence of emerging pollutants in urban wastewater and their removal through biological treatment followed by ozonation. Water Research 44, 578e588. Selke, S., Scheurell, M., Shah, M.R., Hu¨hnerfuss, H., 2010. Identification and enantioselective gas chromatographic mass-spectrometric separation of O-desmethylnaproxen, the main metabolite of the drug naproxen, as a new environmental contaminant. Journal of Chromatography A 1217 (3), 419e423. Shirley, M.A., Guan, X.M., Kaiser, D.G., Halstead, G.W., Baillie, T.A. , 1994. Taurine conjugation of ibuprofen in human and in rat liver in vitro. Relationship to metabolic chiral inversion. Journal of Pharmacology and Experimental Therapeutics 269 (3), 1166e1175. Stanley, J.K., Ramirez, A.J., Mottaleb, M., Chambliss, C.K., Brooks, B.W., 2006. Enantiospecific toxicity of the beta-blocker
propranolol to Daphnia magna and Pimephales promelas. Environmental Toxicology and Chemistry 25 (7), 1780e1786. Tadkaew, N., Hai, F.I., McDonald, J.A., Khan, S.J., Nghiem, L.D., 2011. Removal of trace organics by MBR treatment: the role of molecular properties. Water Research 45, 2439e2451. 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. Thomason, M.J., RhysWilliams, W., Hung, Y.F., Baker, J.A., Hanlon, G.W., Lloyd, A.W., 1997. A mechanistic investigation of the microbial chiral inversion of 2-phenylpropionic acid by Verticillium lecanii. Chirality 9 (3), 254e260. Wong, C.S., 2006. Environmental fate processes and biochemical transformations of chiral emerging organic pollutants. Analytical and Bioanalytical Chemistry 386, 544e558. Wsol, V., Skalova, L., Szotakova, B., 2004. Chiral inversion of drugs: coincidence or principle? Current Drug Metabolism 5 (6), 517e533. Ziylan, A., Ince, N.H., 2011. The occurrence and fate of antiinflammatory and analgesic pharmaceuticals in sewage and fresh water: treatability by conventional and nonconventional processes. Journal of Hazardous Materials 187, 24e36.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 5 9 e6 2 6 6
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Occurrence, distribution and fluxes of benzotriazoles along the German large river basins into the North Sea Hendrik Wolschke, Zhiyong Xie*, Axel Mo¨ller, Renate Sturm, Ralf Ebinghaus Helmholtz-Zentrum Geesthacht, Centre for Materials and Coastal Research, Institute of Coastal Research, Department for Environmental Chemistry, Max-Planck-Strasse 1, 21502 Geesthacht, Germany
article info
abstract
Article history:
Benzotriazole (BT) and tolyltriazole (TT) are high production volume chemicals which are
Received 3 July 2011
used in various industrial and household applications. In this study, the distribution of
Received in revised form
benzotriazoles in the estuaries of different rivers of central Europe and in the North Sea
7 September 2011
was analyzed by solid-phase extraction (SPE) and liquid chromatography-mass spec-
Accepted 12 September 2011
trometry (HPLC-MS/MS). BT as well as TT was detected in all water samples. The
Available online 17 September 2011
concentrations for total benzotriazoles (BTs) ranged from 1.7 to 40 ng/L in the North Sea in costal areas. Concentrations in rivers are from 200 to 1250 ng/L, respectively. The mass flux
Keywords:
of total benzotriazoles from the major rivers of central Europe into the North Sea was
Benzotriazole
calculated to 78 t/a, dominated by the Rhine with an individual flux of 57 t/a of BTs. The
Tolyltriazole
analysis of the distribution profile in the North Sea showed that the decrease of the
River water
concentration was mostly caused by dilution and that the benzotriazoles are poorly
North Sea
degradable in the North Sea. This paper presents the first report of benzotriazoles in the
HPLC-MS/MS
marine environment. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
1-H-benzotriazole (BT) and 1H-methyl-benzotriazole (tolyltriazole, TT, used as a technical mixture of 4- and 5-TT) are a class of high production volume chemicals (HPVC) with broad applications in various industrial processes as well as in households. They show metal complexing properties and are used as anticorrosive additives and flame retardants in aircraft de-icers and anti-ice fluids (ADAFs) (Gruden et al., 2001), in cooling and hydraulic fluids and for silver protection in dishwashing agents (Ort et al., 2005; Janna et al., 2011). The content of BTs in ADAFs varies between 0.2 and 1.7% (WIPO, 2002). The annual production of BTs in the United States was reported to be in the range of 9000 t per year (Hem et al., 2003). The content of Benzotriazoles in Dishwashing agents is reported with 0.09 mg per wash to 27.8 mg (Janna et al., 2011).
BTs are compounds with a low vapour pressure, high water solubility and a high polarity (logKow 1.44 for BT and 1.71 for TT, respectively) (US EPA, 2010). Moreover, they are quite persistent against biological and photochemical degradation processes in the aquatic environment (Hart et al., 2004). Thus, they can be characterized as mobile in aquatic environment. They are classified as toxic to aquatic organisms as they can cause adverse long-term effects on the aquatic environments (Cancilla et al., 1998, 2003; Pillard et al., 2001; Hem et al., 2003). In the environment benzotriazoles were firstly reported in subsurface water close to airports as potential source in concentrations of 128 mg/L for BT, 17 mg/L for 4-TT and 198 mg/L for total TT (Cancilla et al., 1998). Because of the broad applications, BT and TT have high loads in waste waters treatment plants (WWTPs). The mean concentrations in untreated wastewaters of WWTPs of Berlin were 12 mg/L,
* Corresponding author. Tel.: þ49 4152 87 2330; fax: þ49 4152 87 2332. E-mail address: [email protected] (Z. Xie). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.09.028
6260
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 5 9 e6 2 6 6
1.3 mg/L and 2.1 mg/L for BT, 4-TT and 5-TT, respectively (Weiss et al., 2006). The elimination efficiency in WWTPs shows wide variations. In a study of 24 WWTPs of the Glatt Valley, Switzerland elimination rates were reported between 13 and 62% for BT and between 23 and 74% for the TT-isomers (Voutsa et al., 2006) while Weiss et al. (2006) determined removals of 37% for BT and 11% for 5-TT but no significant removal for the 4-TT-isomer. In sewage effluents of the UK were concentrations up to 3.6 mg/L for BT and 5.7 mg/L determined (Janna et al., 2011). Several studies showed that the BT and TT are widely distributed in surface waters (Giger et al., 2006; Reemtsma et al., 2006, 2010; Voutsa et al., 2006). Loos et al. (2009) detected BT and TT in 94% and 81% of 122 river water samples distributed over the European Union (EU) with mean concentrations of 493 ng/L and 617 ng/L for BT and TT, respectively. In an EU-wide study on groundwater, BT and TT were detected in mean concentrations of 1 and 4 ng/L, respectively (Loos et al., 2010a). Only one flux calculation is available for the river Danube into the Black Sea (Loos et al., 2010b). However, there are no data about concentration and distribution of BTs in the marine environment. The objective of the present study was to investigate the occurrence and distribution of benzotriazoles in the German Bight and its major affluxes with respect to their spatial distribution as well as seasonal variations. Furthermore, the mass flux of BTs from the main rivers as sources of BTs into the North Sea was investigated. This study presents the first data on BTs in the marine environment and improves the understanding of the fate and the ecosystem risk assessment of BTs.
2.
Experimental
2.1.
Chemicals
1-H-benzotriazole (BT; 99%), 5-methyl-benzotriazole (5tolyltriazole, 5-TT; 98%) and the deuterated 1H-Benzotriazoled4 solution were purchased from SigmaeAldrich (Steinheim, Germany). A tolyltriazole-isomer mixture was purchased from Dr. Ehrenstorfer (Augsburg, Germany). Acetone (Nanograde) and methanol (Picograde) were purchased from LGC Promochem (Wesel, Germany). Formic acid and hydrochloric acid were purchased from Merck (Darmstadt, Germany). Pure water was obtained from a Milli-Q system (Millipore, Billericia, MA, USA).
2.2.
2.3.
Sample preparation
Solid phase extraction (SPE) was used for compound extraction and enrichment from the water samples. The unique river samples were filtered through glass fibre filters (GF-C, Whatman) while seawater samples were not filtered. SPE was performed with a self-designed glass based setup to prevent contaminations from plastic tubings. 500e700 mL surface water were acidified with hydrochloric acid to pH 2 and spiked with 20 ng of the internal standard 1-H-benzotriazole-d4. Oasis HLB cartridges (500 mg, 60 mm, Waters, Milfort, USA) were conditioned with 15 mL acetone, followed by 15 mL methanol and 5 mL acidified Milli-Q-Water. The water samples were passed through the pre-conditioned cartridges at a flow rate of 1e2 mL min1 and afterwards washed with 5 mL acidified Milli-Q-Water. The cartridges were dried under vacuum for 5 min. The analytes were eluted with 15 mL methanol. The extracts were Roti-evaporated to 5 mL and further concentrated to a volume of 150 mL under a gentle stream of preheated nitrogen. The particulate phase on the filters was extracted by ultrasonic with 20 mL methanol two times for 15 min. The evaporation procedures were the same as described for the cartridges.
2.4.
Instrumental analysis
The extracts were analyzed using high performance liquid chromatography-tandem mass-spectrometry (HPLC-MS/MS) in electrospray positive ionization mode. A HP 1100 HPLC by Agilent Technologies was used for separation equipped with a Phenomenex Synergi Hydro RP 80A column and a suitable guard column (Phenomenex Synergi 2 l Hydro RP Mercury 20 2 mm). The injection volume was 10 mL. The mobile phases were 1% formic acid as ionization aid in Millipore water and methanol, respectively. The operating flow was 200 mL/min in gradient mode, starting with 90% of water as mobile phase
Sample collection
Surface water samples (water depth <0.5 m) were taken by stainless steel bucked in pre-cleaned 1 L glass bottles on board of the research vessel Ludwig Prandtl along the rivers Elbe and Weser during four campaigns in March, May, August and October 2010 with sampling intervals of 10 km. In addition, 15 spot samples were taken from the shore of the rivers Elbe, Weser, Ems, the RhineeMeuse delta and the river Scheldt in August 2010. North Sea samples were taken on board of the research vessel Heincke during three campaigns in March, July and September 2010. The sampling stations are shown in Fig. 1. Additional information on sampling are included in the Supplementary material.
Fig. 1 e Map showing the sampling positions in the investigated rivers and the North Sea.
6261
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 5 9 e6 2 6 6
decreasing to 0% over 25 min, then kept constant for the next 5 min. Detection and quantification of the Benzotriazoles were carried out by an API 3000 tripleequadrupole mass spectrometer (Applied Biosystems/MDS SCIEX) with an electrospray ionization interface in positive ionization mode. The mass spectrometer was operated in a multiple reaction monitoring mode (MRM) with a dwell time of 15 ms. The ion spray voltage was set to 5000 V and the temperature of the source block was 400 C. For identification, the mass transitions 120.1e65.2 for BT, 134.1e77.0 for TT and 124.2e68.9 for the internal standard BT-d4 were monitored. The chromatographical separation of the tolyltriazole isomers was not possible in this study, so the isomers were as sum considered.
2.5.
Quality control
Quality control and quality assurance included the use of a mass labelled internal standard, the determination of recovery rates, of method and instrumental blanks as well as breakthrough and reproducibility experiments. Method blanks were prepared using 500 mL pre-cleaned Millipore water at every set of 8 samples together with the samples. Blank levels were below the method quantification limits
(MQL). The total analytical error is calculated by 7.6% over the method. Recovery rates of the mass labelled internal standard 1HBenzotriazole-d4 in river water samples ranged between 53 11% for the winter season samples and 40 6% for the summer season. The recovery for the seawater samples was 69 10%. The breakthrough of BTs was found to be negligible with <1%. The recovery displacements were caused by matrix suppressions and losses during the solvent evaporation. The MQLs were calculated based on a signal-to-noise ratio of ten. The resulting MQLs are 1.2 ng/L for BT and 0.4 ng/L for TT, respectively.
3.
Results and discussion
3.1.
Occurrence in the rivers Elbe and Weser
Benzotriazole and tolyltriazole were detected in the dissolved phase of all investigated surface water samples. The individual concentrations of the four sampling campaigns along the two rivers are shown in Table 1. In the Elbe the concentrations ranged from 24 ng/L and 21 ng/L in the outer estuary of Elbe up to 304 ng/L and 322 ng/L in the harbour of Hamburg for BT and TT, respectively. The concentrations of BT and TT
Table 1 e Concentrations of benzotriazoles in ng/L in Elbe (E) und Weser (W). (e) [ not sampled; station names are stream kilometres; NE1eNE3 are positions in the North Sea; exact positions are included in the Supplementary data. Station
E570 E579 E589 E599 E609 E619 E624 E629 E639 E649 E659 E669 E679 E689 E699 E709 E719 NE1 NE2 NE3 W2 W12 W22 W32 W42 W52 W62 W72 W90 W110
March
May
August
October
BT [ng/L]
TT [ng/L]
BT [ng/L]
TT [ng/L]
BT [ng/L]
TT [ng/L]
BT [ng/L]
TT [ng/L]
142 155 139 140 118 142 134 157 179 135 156 139 137 114 103 115 97 101 48 25 125 138 128 117 114 134 113 109 70 31
191 281 270 279 229 213 195 239 323 243 250 231 235 223 176 228 180 203 74 37 208 190 193 146 175 233 195 160 128 53
139 142 136 138 134 e 136 152 141 139 128 134 127 121 115 102 e 75 51 24 215 219 215 207 199 188 156 106 69 39
231 268 159 246 209 e 234 231 177 123 109 130 133 103 126 81 e 114 85 34 454 325 517 426 381 395 211 189 158 65
e e e e 193 209 228 254 268 304 257 212 180 155 144 110 e e e e e e e e e e e e e e
e e e e 142 214 173 196 195 222 194 195 186 125 113 122 e e e e e e e e e e e e e e
e e e e 93 92 98 91 120 164 163 162 162 160 148 120 105 92 57 29 169 186 168 179 198 203 182 107 63 28
e e e e 64 49 71 72 101 138 154 169 146 152 119 84 93 73 41 21 136 192 159 180 155 221 232 133 54 21
6262
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 5 9 e6 2 6 6
in the Weser ranged from 28 ng/L in the seawater influenced area up to 219 ng/L (BT) and from 21 ng/L to up to 454 ng/L (TT), respectively. The maximum concentration was found in the City of Bremen. The observed concentrations are comparable to Reemtsma et al. (2010), who reported concentrations of 480 ng/L for BT and 570 ng/L for the TT-isomers, respectively, in the Elbe downstream of the city of Hamburg (km 638) in January 2006. In addition to the dissolved phase, the particulate phase of three samples taken in the turbidity maximum of the Elbe estuary (around km 689) was analyzed. The concentrations of BT and TT in the particulate phase were found to be less than 1% of the dissolved phase or even below the MQL. Based on these results, the filters of other stations were not analyzed.
3.2.
Mass fluxes of benzotriazoles in the Elbe
For an improved comparison of the occurrence of benzotriazoles among the different campaigns, riverine mass fluxes of BTs were calculated. They consider the water discharge variations caused by rain events and dry periods. Mass fluxes for the Elbe were calculated based on the sum concentrations of BT and TT and the daily water discharges with a correction factor for seawater influence. The calculation based on the following Eq. (1), 0
1
C B c C Qday f ¼B @ Ssample A 1 Ssea
(1)
where f is the mass flux [mg/s], c is the measured concentration [ng/L], Ssample means salinity at sampling point in PSU (Practical Salinity Units), Ssea means salinity in seawater (35 PSU) and Qday is the water discharge [m3/s] on the sampling day. The water discharge data are supplied from the FGG Elbe. The salinity of the river Elbe is negligible (<1 PSU) compared to the Salinity in the North Sea. Hence, the salinity of the River was in the calculation not considered. The results are shown in Fig. 2. The mean mass fluxes were 450 mg/s, 194 mg/s, 232 mg/s and 474 mg/s in March, May, August and October 2010, respectively. The high flux in March, which was
constant along the entire sampling course along the river, was probably caused by the first spring discharge after a strong winter season with high usage of benzotriazoles containing de-icing agents during winter. In March the median concentrations were very similar to the concentrations in May and August. The variations of the mass flux are caused by different water discharges, a constant flux level is not observed. An elevated input in the area of the city of Hamburg was observed in August and especially in October. On this input based the high averages in October. A possible reason for this increase could be a discontinuous emission by a WWTP or another source located in the harbour area of Hamburg. In contrast, no influence of the city of Hamburg was observed in March and May. In October, to the outer estuary the high mass flows disappear. Probably there leave the sampling transect a wastewater plume, which is responsible for the high contamination downstream from Hamburg. In all cruises slightly decreasing mass fluxes were observed downstream of the city of Hamburg. This could be caused by dilution with fresh waters of inflows of tributary streams and surface run off. Based on the study in Section 3.6 degradation processes are implausible. The exchange with other compartments like air or sediment seems not to be an important process for BTs, due to their physicalechemical properties.
3.3. Occurrence of benzotriazoles in the tributary rivers of the North Sea Spot samples taken in the estuaries of the tributary rivers were investigated. The concentrations of BTs in the investigated affluxes of the North Sea are shown in Table 2. The concentrations, which were not influenced by seawater,
Table 2 e Concentrations of benzotriazoles of the North Sea feeder rivers in ng/L; exact positions are included in the Supplementary data. River Elbe Elbe Weser Weser Ems Ems Rhine e IJssel Rhine e North Sea Canal Rhine e New Waterway Rhine e Hollands Diep Rhine e GER/NED border Waal Meuse Scheldt
Fig. 2 e Calculated benzotriazole mass fluxes [mg/s] of the river Elbe.
Scheldt
Sample station
BT TT Salinity [ng/L] [ng/L] [PSU]
E679 e Glu¨ckstadt E 712 e Cuxhaven W 90 e Bremerhaven W2 e Hemelingen Ems2 e Emden Ems1 e Herbrum R1 e Kampen R2 e Ijmuiden
172 55 105 278 146 291 369 155
232 77 126 357 208 312 332 184
0.6 20.1 19.8 0.5 10.7 0.6 0.0 10.0
R3 e Maassluis
246
431
1.1
R4 e Bovensluis
241
428
0.1
R5 e Lobith
262
378
0.0
R6 e Andelst R7 e Megen S2 e Schaar van Ouden Doel S1 e Schelle
233 397 180
342 417 374
0.0 0.1 16.7
390
1114
3.6
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 5 9 e6 2 6 6
ranged from 172 ng/L for BT and 232 ng/L for TT in the river Elbe, and up to 390 ng/L (BT) and 1114 ng/L for TT in the river Scheldt. Especially upstream of the city of Antwerp, the river Scheldt is in comparison to other large rivers with 1500 ng/L for benzotriazoles highly contaminated. In the Rhine and its anthropogenic modified delta the concentrations ranged from 155 to 369 ng/L for BT and 184e431 ng/L for TT. In the rivers Ems and Weser, concentrations were within 278e291 ng/L for BT and 312e357 ng/L for TT, which are relatively comparable to those in the Elbe and Rhine. Obviously the concentrations of BTs decreased towards the mouth from Ems-1 to Ems-2 and from Weser-W2 to Weser-W90, even under consideration of the seawater influence. It is assumable that this is due to dilution processes from a diffuse inflow of the rural surrounding areas and probably degradation processes occurring. In comparison to other studies for BTs in surface waters, the results in this study showed relatively similar concentrations. In an European survey on more than 100 rivers from 27 countries Loos et al. (2009) found median concentrations of 226 ng/L for BT and 140 ng/L for TT, respectively. In the river Main, which is influenced by discharge of the Frankfurt airport, median concentrations were 132 ng/L for BT and 162 ng/L for TT, respectively. Reemtsma et al. (2010) reported increasing concentrations along the rivers Elbe (by means of 5 samples) and Rhine (by means of 4 samples) over distances of 650 and 700 km, respectively. In the present study their results were affirmed. This increasing of concentration could be caused by an industrial input of the high industrial density along estuaries due to the logistical connection from rivers and the sea and a accumulation along the rivers. This indicates a bad degradation of the investigated compounds. Especially noticeable in the present study is that the concentration of TT in relation to BT is elevated. In studies in Switzerland (Giger et al., 2006; Voutsa et al., 2006) as well as in the European survey study on polar organic contaminations (Loos et al., 2009) higher median concentrations for BT than TT are reported. In the study on the Rhine and Elbe (Reemtsma et al., 2010), a decrease of the BT/TT ratio along the rivers was observed. Since 5-TT is known to be more degradable than BT and 4-TT (Weiss et al., 2006), this effect is likely not related to degradation processes. A possible explanation could be different usages of benzotriazoles in the upper reaches regions of the rivers than in the lower reaches regions.
3.4.
Mass fluxes of European rivers into the North Sea
The mass flux estimations are based on one sampling campaign in August 2010 only. Mass fluxes were calculated by Eq. (1) with the modification that the average annual water discharges were used. The mass fluxes are shown in Fig. 3. The overall total input from all investigated rivers into the North Sea was calculated with 30 t/a for BT and 47 t/a for TT, respectively. The dominate source is the Rhine with his 4 main inflows into the North Sea (North Sea Canal, New Waterway Canal, Ijssel and Hollands Diep/Haringvliet) with approximately 22 t/a for BT and 35 t/a for TT, respectively. This accounted for 72% of the discharge of the investigated rivers.
6263
Fig. 3 e Estimated mass fluxes in t/a of European rivers into the North Sea; Triangles show the water discharge of the river.
Seasonal variations are not considered in this estimation. In a previous study, Giger et al. (2006) showed no significant seasonal trends in the Rhine. The seasonal variations of the mass flux of BTs from the Elbe to the North Sea estimated in this study are correlated to the water discharge. For more specified conclusions of the seasonal trends are detail analyses with smaller time intervals required.
3.5.
Occurrence in the North Sea
BT and TT were detected above the MQL in all seawater samples from the German Bight. The results are shown in Fig. 4. In the coastal area concentrations up to 21 ng/L and 37 ng/L were found for BT and TT, respectively. In the open sea in approximately 300 km distance to the coast, concentrations of BT and TT were 1.4 ng/L and 1.1 ng/L, respectively, which are just above the MQL. The main current conditions in the German Bright are dominated on an eastern current the “Continental Coastal Water”. This transported water masses from the Dutch Coast into the German Bright and turned than to the north (Turrell, 1992). The benzotriazoles distribution in the North Sea indicates an input from the rivers Elbe and Weser from the southeeast into the German Bright. The high contamination in front of the East Frisian Islands can be traced back to the input of the rivers Rhine and Scheldt. The contamination in the area of the North Frisian Islands is based on the input over the Rhine and Scheldt as well as from the Elbe and Weser. The samples from the open sea were mainly influenced by Atlantic Ocean water streamed along the Scottish Coast as central and south North Sea waters into this area (Turrell, 1992). A slight seasonal variation was observed. In comparison to the summer in spring high concentrations were detected in the coastal area, with a strong decreasing trend towards the open sea. In summer the concentrations in the coastal area were slightly lower, but interestingly, the open sea stations showed higher concentrations in summer than in spring. This
6264
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 5 9 e6 2 6 6
Fig. 4 e Distribution of benzotriazoles in the North Sea in ng/L.
is against the theory that the highest photo- and biodegradation is in summer. In autumn the trend followed the summer pattern. The contamination of Benzotriazoles in the North Sea is relatively high in comparison to other organic pollutants. For example, concentrations of polyfluoroalkyl compounds (PFCs) such as perfluorooctanoic acid (PFOA) and perfluorooctane sulfonate (PFOS) are one order of magnitude lower (Ahrens et al., 2009), concentrations of the endocrine disruptor Triclosan are three orders of magnitude lower as benzotriazoles (Xie et al., 2008). The sum concentrations of organophosphate flame retardants (OPE) showed comparable concentrations to single concentrations of BT and TT (Andresen et al., 2007).
3.6.
Fig. 5 displays the dilution factor (DF) at each sampling point with seawater influence (salinity > 1 PSU) of the March and September/October sampling in the Elbe and the North Sea in comparison to the normalized concentration (NC) for BT and TT. The mostly linear observed relationship signifies that the decrease of the concentration is mainly due to dilution with seawater. Only for TT in autumn are, a slight degradation can be ascertained in autumn, because the most points are under the line. This can be explained by higher degradability of the 5-TT isomer than BT and the 4-TT isomer (Weiss et al., 2006). In both graphs the sampling station a point E719 for TT is
Study of dilution and persistency
To investigate if decreasing concentrations are based on degradation or merely on dilution, a comparison with salinity is useful. The correlation of dilution versus salinity was assessed with utilization of the Eqs. (2) and (3).
Dilution Factor (DF) ¼ salinity sampling point/salinity seawater
(2)
Normalized concentration (NC) ¼ concentration sampling point/concentration freshwater (3) The salinity of seawater was assumed with 35 PSU. The freshwater concentration was set to the last sampling point in the estuary with a salinity <1 PSU.
Fig. 5 e Correlation of dilution factor and normalized concentrations for benzotriazole and tolyltriazole in the river Elbe and the North Sea.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 5 9 e6 2 6 6
located in the front of the harbour of Cuxhaven and has assumedly a local source of tolyltriazole. Based on this analysis, the seasonal variations which are described in Section 3.5 were caused by seawater currents and dilution processes rather than degradation.
4.
Conclusions
Benzotriazoles are poorly degradable polar organic pollutants which are present in the anthropogenic water cycle. This study has shown that approximately 80 t per year of these chemicals are discharged via the investigated rivers into the North Sea, mainly via the river Rhine. Further it is shown that the decrease of the concentration in the German Bight is mostly attributed to dilution. On the one hand the toxicity of these substances is reported to be moderate, the bioaccumulation potential is low and the detected concentration are two magnitudes lower as the chronic predicted no effect concentrations. But on the other hand the benzotriazoles are observed concentrations of benzotriazoles in the aquatic environment rank among the highest in the group of polar organic pollutants in the water cycle. Thus the widespread distribution in European lakes, groundwater, rivers and the North Sea in addition to the shown persistence in aquatic samples has to be noted. Further studies have to focus on the occurrence and degradation of the benzotriazoles in the marine water cycle, especially separate investigation of the tolyltriazole isomers.
Acknowledgement We acknowledge the Alfred-Wegener-Institute for Polar and Marine Research (AWI), Bremerhaven, Germany, for the possibility of taking part in the expedition cruises. We are grateful to the captains and the crews of RV Heincke and RV L. Prandtl.
Appendix. Supplementary data Supplementary data related to this article can be found online at doi:10.1016/j.watres.2011.09.028.
references
Ahrens, L., Plassmann, M., Xie, Z.Y., Ebinghaus, R., 2009. Determination of polyfluoroalkyl compounds in water and suspended particulate matter in the river Elbe and North Sea, Germany. Frontiers of Environmental Science & Engineering in China 3 (2), 152e170. Andresen, J.A., Muir, D., Ueno, D., Darling, C., Theobald, N., Bester, K., 2007. Emerging pollutants in the North Sea in comparison to Lake Ontario, Canada, data. Environmental Toxicology and Chemistry 26 (6), 1081e1089. Cancilla, D.A., Baird, J.C., Rosa, R., 2003. Detection of aircraft deicing additives in groundwater and soil samples from
6265
fairchild air force base, a small to moderate user of deicing fluids. Bulletin of Environmental Contamination and Toxicology 70 (5), 868e875. Cancilla, D.A., Martinez, J., Van Aggelen, G.C., 1998. Detection of aircraft deicing/antiicing fluid additives in a perched water monitoring well at an international airport. Environmental Science & Technology 32 (23), 3834e3835. FGG Elbe, Flussgebietsgemeinschaft Elbe, www.fgg-elbe.de. Giger, W., Schaffner, C., Kohler, H.-P.E., 2006. Benzotriazole and tolyltriazole as aquatic contaminants. 1. Input and occurrence in rivers and lakes. Environmental Science & Technology 40 (23), 7186e7192. Gruden, C.L., Dow, S.M., Hernandez, M.T., 2001. Fate and toxicity of aircraft deicing fluid additives through anaerobic digestion. Water Environment Research 73 (1), 72e79. Hart, D.S., Davis, L.C., Erickson, L.E., Callender, T.M., 2004. Sorption and partitioning parameters of benzotriazole compounds. Microchemical Journal 77 (1), 9e17. Hem, L.J., Hartnik, T., Roseth, R., Breedyleld, G.D., 2003. Photochemical degradation of benzotriazole. Journal of Environmental Science and Health Part A-Toxic/Hazardous Substances & Environmental Engineering 38 (3), 471e481. Janna, H., Scrimshaw, M.D., Williams, R.J., Churchley, J., Sumpter, J.P., 2011. From dishwasher to tap? Xenobiotic substances benzotriazole and tolyltriazole in the enviroment. Environmental Science & Technology 45, 3858e3864. Loos, R., Gawlik, B.M., Locoro, G., Rimaviciute, E., Contini, S., Bidoglio, G., 2009. EU-wide survey of polar organic persistent pollutants in European river waters. Environmental Pollution 157 (2), 561e568. Loos, R., Locoro, G., Comero, S., Contini, S., Schwesig, D., Werres, F., Balsaa, P., Gans, O., Weiss, S., Blaha, L., Bolchi, M., Gawlik, B.M., 2010a. Pan-European survey on the occurrence of selected polar organic persistent pollutants in ground water. Water Research 44 (14), 4115e4126. Loos, R., Locoro, G., Contini, S., 2010b. Occurrence of polar organic contaminants in the dissolved water phase of the Danube River and its major tributaries using SPE-LCMSMS analysis. Water Research 44 (7), 2325e2335. Ort, C., Schaffner, C., Giger, W., Gujer, W., 2005. Modeling stochastic load variations in sewer systems. Water Science and Technology 52 (5), 113e122. Pillard, D.A., Cornell, J.S., Dufresne, D.L., Hernandez, M.T., 2001. Toxicity of benzotriazole and benzotriazole derivatives to three aquatic species. Water Research 35 (2), 557e560. Reemtsma, T., Miehe, U., Duennbier, U., Jekel, M., 2010. Polar pollutants in municipal wastewater and the water cycle: occurrence and removal of benzotriazoles. Water Research 44 (2), 596e604. Reemtsma, T., Weiss, S., Mueller, J., Petrovic, M., Gonzalez, S., Barcelo, D., Ventura, F., Knepper, T.P., 2006. Polar pollutants entry into the water cycle by municipal wastewater: a European perspective. Environmental Science & Technology 40 (17), 5451e5458. Turrell, W.R., 1992. New hypotheses concerning the circulation of the northern North Sea and its relation to North Sea fish stock recruitment. ICES Journal of Marine Science: Journal du Conseil 49 (1), 107e123. US EPA, 2010. Estimation Programs Interface Suite for Microsoft Windows, V4.0. United States Environmental Protection Agency, Washington, DC, USA. Voutsa, D., Hartmann, P., Schaffner, C., Giger, W., 2006. Benzotriazoles, alkylphenols and bisphenol A in municipal wastewaters and in the Glatt River, Switzerland. Environmental Science and Pollution Research 13 (5), 333e341. Weiss, S., Jakobs, J., Reemtsma, T., 2006. Discharge of three benzotriazole corrosion inhibitors with municipal wastewater
6266
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 5 9 e6 2 6 6
and improvements by membrane bioreactor treatment and ozonation. Environmental Science & Technology 40 (23), 7193e7199. WIPO, 2002. Environmentally Compatible Defrosting and Antifreeze Agents for Aeroplanes. The World Intellectual Property Organization (online). http://www.wipo.int/pctdb/
en/wo.jsp?WO¼2002099004&IA¼EP2002005584&DISPLAY ¼DESC (accessed 10.10.10.). Xie, Z.Y., Ebinghaus, R., Floser, G., Caba, A., Ruck, W., 2008. Occurrence and distribution of triclosan in the German Bight (North Sea). Environmental Pollution 156 (3), 1190e1195.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 6 7 e6 2 7 6
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Influence of wastewater treatment process and the population size on human virus profiles in wastewater Joanne Hewitt a,c,*, Margaret Leonard b,1, Gail E. Greening a, Gillian D. Lewis c a
Institute of Environmental Science & Research Ltd, Kenepuru Science Centre, PO Box 50-348, Porirua, New Zealand Institute of Environmental Science & Research Ltd, Christchurch Science Centre, PO Box 29-181, Christchurch, New Zealand c School of Biological Sciences, University of Auckland, Auckland, New Zealand b
article info
abstract
Article history:
Human adenovirus (AdV and AdV species F), enterovirus (EV) and norovirus (NoV)
Received 1 May 2011
concentrations entering wastewater treatment plants (WWTP) serving different-sized
Received in revised form
communities, and effectiveness of different treatment processes in reducing concentra-
27 August 2011
tions were established. Data was combined to create a characteristic and unique descriptor
Accepted 13 September 2011
of the individual viral composition and termed as the sample virus profile.
Available online 19 September 2011
Virus profiles were generally independent of population size and treatment process (moving bed biofilm reactors, activated sludge, waste stabilisation ponds). AdV and EV
Keywords:
concentrations in wastewater were more variable in small (<4000) and medium-sized
Virus removal
(10,000e64,000) WWTP than in large-sized (>130,000 inhabitants) plants. AdV and EV
Wastewater
concentrations were detected in influent of most WWTP (AdV range 1.00e4.08 log10
Norovirus
infectious units (IU)/L, 3.25e8.62 log10 genome copies/L; EV range 0.7e3.52 log10 plaque
Adenovirus
forming units (PFU)/L; 2.84e6.67 log10 genome copies/L) with a reduced median concen-
Enterovirus
tration in effluent (AdV range 0.70e3.26 log10 IU/L, 2.97e6.95 log10 genome copies/L; EV range 0.7e2.15 log10PFU/L, 1.54e5.28 log10 genome copies/L). Highest culturable AdV and EV concentrations in effluent were from a medium-sized WWTP. NoV was sporadic in all WWTP with GI and GII concentrations being similar in influent (range 2.11e4.64 and 2.19e5.46 log10 genome copies/L) as in effluent (range 2.18e5.06 and 2.88e5.46 log10 genome copies/L). Effective management of WWTP requires recognition that virus concentration in influent will vary e particularly in small and medium plants. Irrespective of treatment type, culturable viruses and NoV are likely to be present in non-disinfected effluent, with associated human health risks dependent on concentration and receiving water usage. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
Enteric viruses transmitted by the faecaleoral route are excreted by infected people in high concentrations in faeces, and are potentially present in large numbers in domestic
wastewater (Farthing, 1989; Wadell, 1984). Effective treatment of such wastewater to reduce the infectious virus load is important to minimise public health risks in receiving water (Hafliger et al., 2000; Hewitt et al., 2007; Le Guyader et al., 2006). The potential risk to human health depends on the types of
* Corresponding author. Institute of Environmental Science & Research Ltd, Kenepuru Science Centre, PO Box 50-348, Porirua, New Zealand. Tel.: þ64 4 914 0690; fax: þ64 4 914 0770. E-mail address: [email protected] (J. Hewitt). 1 Present address. Christchurch Polytechnic, PO Box 540, Christchurch 8140, New Zealand. 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.09.029
6268
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 6 7 e6 2 7 6
viruses present, the types of viruses that survive treatment, their viability and concentration and the use of the receiving water, such as for drinking water, recreational activity or other use. Human adenoviruses (AdV), enteroviruses (EV) and noroviruses (NoV) are commonly found in influent and effluent wastewater (da Silva et al., 2007; Haramoto et al., 2007; Irving and Smith, 1981; Krikelis et al., 1985; Laverick et al., 2004; Metcalf et al., 1995; Nordgren et al., 2009) and are important etiological agents of many human illnesses (Green, 2007; Pallansch and Roos, 2007; Wold and Horwitz, 2007). Human AdV and EV outbreaks have occurred because of recreational water becoming contaminated with human faecal waste (Sinclair et al., 2009). Outbreaks of NoV have been frequently associated with contact with either contaminated water or shellfish (Hewitt et al., 2007; Le Guyader et al., 2006; Lees, 2000; Simmons et al., 2007; Sinclair et al., 2009). Most data on virus occurrence and concentration in wastewater and removal efficiency are sourced from WWTP that generally serve cities or large communities with populations greater than 100,000 inhabitants. There is little published data on the removal of enteric viruses by waste stabilisation ponds (WSP) that serve small populations (da Silva et al., 2007; Lewis et al., 1986; Nordgren et al., 2009) mainly due to the expense and availability of virus monitoring data in wastewater or treated effluents. In this study, the presence and quantity of three human enteric viruses, AdV (including specifically human AdV species F (AdV-F), types 40 and 41), EV and NoV genogroups I and II (GI and GII) were determined in influent and effluent wastewater from ten New Zealand WWTP. The WWTP studied represented those serving different-sized communities, geographical locations and a range of wastewater treatment processes. Real-time (reverse transcription, RT)-quantitative PCR (qPCR) and cell culture assays, where applicable, were used for virus detection and quantitation. The data from each virus analysis was combined for each sample to create a characteristic and unique descriptor of the individual viral composition and was termed the ‘virus profile’. Each virus profile included the occurrence and concentration for each
virus type tested and for each method used for the analysis of that virus. Virus profiles were compared and similarities visualised between samples using non-metric multi-dimensional scaling (MDS) and analysis of similarity (ANOSIM). The aim of the study was to obtain an understanding of virus prevalence in the influent of WWTP serving various population sizes, the persistence of enteric viruses through treatment and the influence of treatment type on the virus profiles of the effluent.
2.
Materials and methods
2.1.
WWTP and samples
Primary screened influent and treated effluent wastewater samples (1 L) were collected from ten community WWTP (AeJ). Table 1 shows the WWTP treatment process, wastewater source and the population size served by each plant. The selected WWTP served large (130,000e1,000,000), medium (10,000e64,000) and small (<1100e4000) communities (LWWTP, M-WWTP, S-WWTP respectively) and were located across New Zealand. The wastewater treatment processes included moving bed biofilm reactor (MBBR) and activated sludge (AS) plants that served large populations, to WSP that served small populations. WWTP D, F, H, I and J processed mainly domestic wastewater whilst WWTP A, B, C, E and G processed domestic wastewater and significant amounts of industrial influent, including animal abattoir wastes. Three samples were taken from each WWTP, one on each of three sampling windows during the New Zealand summer (24 Nov to 12 Dec 2003, 19 Jan to 14 Feb 2004, and 8 Mar 04 to 2 Apr 2004). The timing of sampling was defined by criteria for a wider study investigating a range of analytes in wastewater. Influent and effluent wastewater samples were temporally separate, even when collected on the same day, owing to the time the wastewater took to move through the WWTP. Effluent samples were collected prior to any mechanical or chemical disinfection processes.
Table 1 e Wastewater treatment plants (WWTP) location, treatment process, predominate wastewater source and population size. WWTP A B C D E F G H I J
Treatment processa
Wastewater source
Moving bed biofilm reactor Trickling filter, activated sludge, waste stabilisation pond Activated sludge, nitrogen removal Activated sludge, nitrogen removal Activated sludge Moving bed biofilm reactor Waste stabilisation pond, aerator Waste stabilisation pond Waste stabilisation pond Waste stabilisation pond, wetland
Domestic and significant industrial Domestic and significant industrial
Large Large
Domestic and significant industrial Mainly domestic Domestic and significant industrial Mainly domestic Domestic and significant industrial Mainly domestic Mainly domestic Mainly domestic
Large Medium Medium Medium Medium Small Small Small
Approximate population sizeb
a WWTP A, C, E and F also include UV as part of the treatment. Samples for this study were taken prior to UV disinfection. b Population range served by the wastewater treatment plants; large 130,000e1,000,000; medium 10,000e64,000; small <1100e4000 people.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 6 7 e6 2 7 6
2.2.
Virus controls
Human AdV type 2 (American Type Tissue Culture Collection (ATTC) #VR-846, Manassas, VA), human AdV type 41 (ATCC # VR-930), and Sabin poliovirus type 2 (PV2) LSc ab (monovalent Pfizer vaccine strain) were used as positive controls for the AdV qPCR and EV RT-qPCR assay respectively. PV2 was also used for seeding experiments and as a positive control for the EV culture assay. For NoV GI and GII RT-qPCR positive controls, 10% (wt/vol) faecal suspensions from NoV positive samples were prepared, clarified by the addition of chloroform (10%, wt/vol) and then centrifuged at 13,000 g for 10 min.
2.3.
Sample processing
Within 24 h of collection, viruses from 1 L wastewater samples were concentrated to 10 mL (modified from Green and Lewis, 1995; Greening et al., 2002; Lewis and Metcalf, 1988; Wait and Sobsey, 1983). Samples (1 L) were first centrifuged at 2400 g for 20 min at 4 C and the supernatant held at 4 C for later use. Sodium nitrate (2 M) in 3% (v/w) beef extract (pH 5.5) was added to the resulting pellet and pH adjusted to 5.5. Viruses were eluted for 1 h at 4 C with continuous mixing on a horizontal shaker at 180 rpm, and centrifuged at 10,000 g for 20 min. The resulting supernatant was added to the first supernatant and pH adjusted to 7e7.2. Viruses were then concentrated by adding polyethylene glycol (PEG) 6000 and NaCl to give final concentrations of 10% (w/v) and 2% (w/v) respectively. Samples were mixed at room temperature to dissolve the components, then placed on a horizontal shaker (120 rpm) for a minimum of 2 h at 4 C and then centrifuged at 10,000 g for 25 min at 4 C. The supernatant was discarded and the pellet resuspended in 10 mL phosphate buffered saline (pH 7.4), so that 1 mL concentrate was equivalent to 0.1 L wastewater. The pH was adjusted to 8.0, sonicated in an Ultrasonic Cleaner (Model FX10, Unisonics Pty Ltd., Sydney, Australia) for 2 min and viruses left to elute from the pellet for 1 h at room temperature with occasional vortexing. Samples were sonicated for another 2 min and centrifuged at 10,000 g for 20 min. Chloroform (10 mL) was added, tubes inverted every 5 min for 15 min and then centrifuged at 1200 g for 5 min. The chloroform phase was discarded and 10 mL chloroform added to the aqueous and interfacial phases, mixed and then centrifuged at 1200 g for 5 min. Penicillinestreptomycin (P/S) (Gibco, Invitrogen Corporation, Carlsbad, CA) and amphotericin B (Gibco) were added. Samples were stored at 80 C, if not analysed within 24 h. Viral nucleic acid was extracted from the concentrates (200 mL) using the High Pure Viral Nucleic Acid kit (Roche Molecular Biochemicals Ltd, Mannheim, Germany) as per manufacturer’s instructions and eluted in 50 mL elution buffer. Poliovirus-negative wastewater was seeded with 2.0, 3.0 and 4.0 log10 plaque forming units (PFU)/L PV2 in triplicate and recovered using L20B cells that are very selective for PV growth.
2.4.
Virus culture assays
Cells were propagated in medium 199 (Gibco) (A549 and BGM) or Dulbecco’s Modified Eagle Medium High Glucose (DMEM)
6269
(293) (Gibco) supplemented with 10% (v/v) foetal calf serum (FCS), and P/S (Gibco). Human AdV type 2, human AdV type 41 and PV2 stocks were prepared from infected A549, 293 and BGM cell monolayers respectively by freezeethaw lysis followed by sonication for 3 min in an Ultrasonic Cleaner (Unisonics Pty Ltd). Lysates were extracted by the addition of an equal volume of chloroform and clarified by centrifugation at 400 g for 5 min. Preparations were aliquoted and stored at 80 C. Nucleic acid extracts from each virus stock were used to generate DNA plasmids. Quantitation of culturable AdV was determined by observing cytopathic effect (CPE) in A549 cells following sample inoculation. A maximum of 2 mL concentrate (equivalent to 0.2 L sample) was inoculated by adding 10 mL per well in 96 well plates and applying the most probable number (MPN) method after 28 day incubation. Guanidine hydrochloride (100 mg/mL) was used to suppress EV growth in the AdV assay as previously described (Hurst et al., 1988). Confirmation and species typing of AdV was done by extracting DNA from the selected cell lysates of each sample, followed by AdV specific PCR (Allard et al., 2001) and DNA sequencing. Human AdV type 2 was used as both extraction and PCR-positive control. Quantitation of culturable EV was determined by the agar cell suspension assay using BGM cells (Greening et al., 2002) with a maximum of 2 mL concentrate, equivalent to 0.2 L wastewater, inoculated (0.2 mL 10 plates). PV2 (10 PFU) was used as a positive control in the EV culture assay. Titres were expressed as log10 MPN infectious units (IU)/L for AdV and log10 PFU/L for EV. The theoretical limit of detection per sample based on the amount of sample tested was therefore 0.7 log10 IU/L for culturable AdV and 0.7 log10 PFU/L for culturable EV.
2.5.
RT-qPCR and qPCR
For the generic human AdV and specific human AdV-F qPCR assays, the PCR was carried out using qPCR Supermix-UDG (Invitrogen) with either 0.6 mM (generic AdV) or 0.2 mM (human AdV-F) of each primer and 0.25 mM (generic HAdV) or 0.2 mM (HAdV-F) of each probe (Heim et al., 2003; Wolf et al., 2010). For EV RT-qPCR, the Platinum Quantitative RT-PCR ThermoScript One-Step System (Invitrogen), 0.6 mM primers, 0.25 mM probe and PCR conditions were used as previously described (Donaldson et al., 2002). For NoV, the RT step was carried out using SuperScript III-First Strand Synthesis System for RT-PCR (Invitrogen) with 0.1 mM of each reverse primer (Wolf et al., 2010), followed by qPCR assay performed as a multiplex assay using qPCR Supermix-UDG (Invitrogen), with 0.4 mM of each primer and 0.2 mM of each probe (modified from Wolf et al., 2010). Primers and probes were verified in our laboratory for the detection of the selected viruses in wastewater (data not shown) and were considered appropriate for the purposes of this study. Armored RNANorwalk Virus GI was used to monitor potential RT-qPCR inhibition as previously described (Hewitt et al., 2007). Where (RT)-qPCR inhibition was suspected, the nucleic acid was diluted 1/5 and the sample retested. All qPCR assays were carried out using either a Rotor-Gene 3000 (generic AdV, EV) or 6000 (NoV GI and GII, human AdV-F) real-time rotary analyzer (Corbett Life Science, Sydney, Australia). All samples
6270
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 6 7 e6 2 7 6
were analysed at least in duplicate. Quantification of viruses (genome copies) was determined by comparing the cycle threshold (Ct) value against the appropriate standard curve generated from a dilution series ranging between 107 and 10 genome copies per reaction of the appropriate DNA plasmid prepared from PCR products as previously described (Wolf et al., 2010). Titres were expressed as log10 genome copies/L wastewater. Using the standard curve, the lowest concentration detected was 1.5 log10 genome copies/L in the PCR assays. In each (RT)-qPCR, virus positive extraction controls, DNA plasmid controls, and nuclease-free water controls were used.
2.6.
effluent wastewater samples, population size and between the three categorised groups of WWTP treatment types e MBBR, AS and WSP (Table 1). One way ANOSIM (Clarke and Gorley, 2006) was performed on the resemblance data to determine the statistical significances. Using PRIMER, a test statistic R was generated that ranged from 1 to þ1 depending on the similarity between samples within a group compared to the similarity between samples from a different group. The relationship between AdV and human AdV-F concentrations determined by PCR was also determined for 30 influent samples using Microsoft Excel and the correlation coefficient (r) calculated to determine the degree of correlation.
Data analysis
Virus quantities as determined by culture and (RT)-qPCR were log10 transformed prior to statistical analysis. For S-WWTP, MWWTP and L-WWTP, box plots were used to show the median concentrations (log10/L) with the 25the75th percentile values. The whiskers extended to the most extreme data point no more than 1.5 times the interquartile (25the75th) range from the box. For representation on the box plots, samples with log10/L values greater than 1.5 times the interquartile range from the box were shown as individual points (stars). For data sets where all samples were positive, in addition to the median, mean concentrations with a standard deviation were also calculated. To analyse the multivariate data and summarise (ordinate) patterns of virus presence and concentration between samples, MDS and ANOSIM were used to assess similarities and dissimilarities according to WWTP size, wastewater type (influent or effluent) and treatment process. MDS was chosen as it creates a spatial representation of multivariate sample data which does not assume linear relationships between variables (concentrations), and hence is robust and suitable for comparison of data descriptions comprised of multiple components of variable magnitude. Log10 transformed culture and PCR concentration data for each sample were placed in an Excel spreadsheet (Microsoft, Redmond, WA) to give a matrix consisting of one row per dataset (AdV PCR, AdV culture, EV PCR, EV culture, NoV GI and GII) with one column per sample (n ¼ 60). The data was then exported into PRIMER v6 software (Primer-E Ltd., Plymouth, United Kingdom) and the Manhattan distance between samples calculated for each dataset. MDS plots were created from each resemblance matrix, and were presented as two-dimensional figures to display distances between different virus profiles. Plots were generated to represent the relationship between influent and
3.
Results
3.1.
Culturable viruses
Prevalence (% samples positive) and concentration (log10/L) of AdV and EV detected by cell culture are summarised in Table 2 and Fig. 1 respectively. Seeding and recovery experiments showed at least 2 log10/L could be recovered by culture following the virus concentration method, which equated to a recovery efficiency of at least 10%. Most influent samples (26/30, 87%) contained either culturable AdV and/or EV with concentrations ranging from 1.00 to 4.08 log10 IU/L and 0.7e3.52 log10 PFU/L respectively. MWWTP G and S-WWTP H and I were negative for both culturable AdV and/or EV. The highest median (Fig. 1a, b), and mean culturable AdV and EV concentrations in influent were observed in the three L-WWTP (AdV, mean 2.61 0.45 log10 IU/ L; EV, mean 2.06 1.05 log10 PFU/L). However, virus concentrations for individual WWTP were not necessarily dependent on the population size that the WWTP served. Culturable AdV and EV concentrations in influent from S-WWTP and MWWTP could be as high as in L-WWTP influent, but with a lower detection frequency (Table 2, Fig. 1). For example, the highest concentration of culturable AdV and EV detected in a single influent sample occurred in S-WWTP J (3.92 log10 PFU/ L) and M-WWTP D (3.52 log10 PFU/L) respectively. In effluent, culturable AdV and/or EV were detected in 19/ 30 (63%) samples with the highest concentrations from MWWTP F (AdV, 3.26 log10 IU/L; EV, 2.15 log10 PFU/L). The median concentration, as grouped by population size, in effluent samples was consistently lower (AdV range 0.70e3.26 log10 IU/L; EV range 0.7e2.15 log10 PFU/L) than in the
Table 2 e Prevalence of culturable adenovirus and enterovirus in wastewater treatment plants (WWTP) AeJ. Virus
Wastewater type
No. positive/total samples, % positive samples WWTP A, B, C (large)
WWTP D, E, F, G (medium)
WWTP H, I, J (small)
Total
Adenovirus
Influent Effluent
9/9, 100% 3/9, 33%
10/12, 83% 6/12, 50%
5/9, 55% 3/9, 33%
24/30, 80% 12/30, 40%
Enterovirus
Influent Effluent
8/9, 89% 7/9, 78%
7/12, 58% 4/12, 33%
3/9, 33% 1/9, 11%
18/30, 60% 12/30, 40%
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 6 7 e6 2 7 6
6271
Fig. 1 e Box plots showing quantitation (log10/L) of (a) adenovirus (AdV) by culture, (b) enterovirus (EV) by culture, (c) AdV by PCR, (d) EV by PCR, (e) norovirus (NoV) GI by PCR, and (f) NoV GII by PCR, in influent (Inf) and effluent (Eff) samples from wastewater treatment plants serving small, medium and large populations. Box plots show median concentrations (log10/L) with the 25the75th percentile values. The whiskers extend to the most extreme data point no more than 1.5 times the interquartile (25the75th) range from the box.
influent samples, with the L-WWTP showing a larger median decrease than the smaller WWTP (Fig. 1a and b). The 11 effluent samples negative for both AdV and EV were collected from M-WWTP D, E and G and S-WWTP H, I and J, but only two WWTP (M-WWTP G and S-WWTP H) were negative on all three sampling occasions. Further studies showed that culturable AdV detected using A549 cells in both influent and effluent was mainly from human AdV species AeE (data not shown).
3.2.
Virus detection and quantitation by PCR
Prevalence (% samples positive) and concentration (log10 genome copies/L) of AdV, EV, NoV GI and NoV GII detected by
real-time qPCR are summarised in Table 3 and Fig. 1 respectively. In influent, most samples contained AdV and EV, with AdV present in higher concentrations (range 3.25e8.62, mean 6.28 0.34 log10 genome copies/L) than EV (range 2.84e6.67, mean 5.30 0.60 log10 genome copies/L) (Fig. 1c and d). PCR quantitation showed that total AdV and human AdV-F concentrations were similar, indicating that the majority of AdV were AdV-F. Fig. 2 shows the total human AdV and AdVF concentrations (genome copies/L) and correlation between them (R2 ¼ 0.9001) in influent samples. Whilst the highest AdV concentration (8.62 log10 genome copies/L) in influent detected by PCR was in S-WWTP H, the mean AdV concentration by PCR in positive influent samples was significantly
6272
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 6 7 e6 2 7 6
Table 3 e Prevalence of adenovirus, enterovirus, and norovirus GI and GII, as determined by real-time qPCR for wastewater treatment plants (WWTP) serving large, medium and small populations. Virus
Wastewater type
No. positive/total samples, % positive samples WWTP A, B, C (large)
WWTP D, E, F, G (medium)
WWTP H, I, J (small)
Total
Adenovirusa
Influent Effluent
9/9, 100% 9/9, 100%
10/12, 83% 11/12a, 92%
8/9, 89% 7/9a, 89%
27/30, 90% 27/30, 90%
Enterovirus
Influent Effluent
9/9, 100% 7/9, 78%
12/12, 100% 7/12, 58%
9/9 100% 5/9, 55%
30/30, 100% 19/30, 63%
Norovirus GI
Influent Effluent
5/9, 56% 6/9, 67%
7/12, 58% 5/12, 42%
3/9, 33% 6/9, 67%
15/30, 50% 17/30, 57%
Norovirus GII
Influent Effluent
6/9, 67% 6/9, 67%
5/12, 42% 6/12, 50%
2/9, 22% 5/9, 55%
13/30, 43% 17/30, 57%
a Three samples negative by generic adenovirus PCR assay (Heim et al., 2003) were positive by the human adenovirus species F PCR assay (Wolf et al., 2010).
human AdV-F (log10 genome copies/L)
higher in L-WWTP than in M-WWTP and S-WWTP ( p < 0.05). The lowest variation in AdV and EV concentrations as determined by comparison of the range of PCR concentrations was observed in L-WWTP (Fig. 1a and b). The frequency of NoV GI and GII in influent was less, particularly in the SWWTP, than AdV and EV, with generally lower concentrations (NoV GI and GII range 2.11e4.64 and 2.19e5.46 log10 genome copies/L respectively (Fig. 1e and f)). The highest median NoV concentrations in influent were from samples collected from L-WWTP. In influent, 10/30 (33%) samples were positive for both NoV GI and GII, with 4/30 (13%) samples positive for NoV GI only and 5/30 (17%) samples positive for NoV GII only. Eleven influent samples (37%) were negative for both NoV GI and GII. In effluent, median AdV and EV concentrations (2.97e6.95 and 1.54e5.28 log10 genome copies/L respectively) of each sized WWTP were lower than the influent samples (Fig. 1c and d). For all WWTP sizes, median AdV concentrations in effluent were greater than EV. High AdV concentrations (up to 6.95 log10 genome copies/L) were frequently detected including from SWWTP. In effluent, AdV was more commonly detected than EV and NoV. Effluents from all WWTP were positive by either the 10
y = 1.0801x - 0.1593 R2 = 0.9001
generic AdV or human AdV-F PCR assay (data not shown). In contrast to AdV and EV, the prevalence and median NoV GI and GII concentrations in PCR-positive effluent samples (range 2.18e5.06 and 2.88e5.46 log10 genome copies/L respectively), were not generally lower than in influent samples (Table 3, Fig. 1e and f). Effluent from M-WWTP E was negative for both NoV GI and GII on all three sampling occasions. More samples were positive for both NoV GI and GII in effluent (16/30) than in influent (10/30) with fewer effluent samples positive for GI only (1/30) and GII only (1/30) than the influent. Twelve effluent samples (40%) were negative for both NoV GI and GII. As expected, overall virus prevalence and concentrations detected by PCR were greater than by culture. The mean ( SD) AdV concentration by PCR was 3.6 1.2 log10 and 3.9 0.9 log10 greater than culture for influent and effluent samples respectively. For EV, the values were 3.0 0.9 log10 and 1.4 1.4 log10. It should be noted that some RT-PCR inhibition was observed in the one-step EV RT-qPCR assay and so RNA from all EV PCR negative samples was diluted (1/5) to reduce inhibition effects. An additional six samples were identified as EV positive using diluted RNA; four from effluent and two from influent samples. For AdV and EV, there were three influent and two effluent samples respectively that were positive by culture, but repeatedly negative by PCR despite diluting the sample.
3.3.
8
6
4
2
0 0
2
4
6
8
10
all human AdV species (log10 genome copies /L)
Fig. 2 e Concentration (log10 genome copies/L) and correlation of adenovirus (AdV) and human AdV species F (AdV-F) in influent wastewater samples (n [ 30).
Virus profiles
Figs. 3e5 show MDS plots of virus profiles generated from all PCR and culture quantitative data from the influent and effluent wastewater samples. Virus profiles with the highest similarity are shown by ordinate points plotted nearest to each other, whilst those far apart represent virus profiles that are less similar. Statistical analysis (ANOSIM) of influent and effluent sample groups showed that virus profiles of influent samples were more similar to each other (i.e. within the group) than to the effluent samples (i.e. between the groups) (ANOSIM R ¼ 0.238, where 0 is highly similar (no difference) and 1 is highly dissimilar). The similarities between the virus profiles of influent and effluent are also observed in Fig. 3 as two clusters showing a degree of separation. This dissimilarity
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 6 7 e6 2 7 6
Fig. 3 e Non-metric multi-dimensional scaling plot showing virus profile comparison of influent (solid symbols) and effluent (non-solid symbols) wastewater samples from each wastewater treatment plant on three occasions. Samples grouping closer together show a more similar virus profile than those that are further apart.
between the influent and effluent sample groups is most likely due to the overall reductions in median AdV and EV concentrations from all WWTP (Fig. 1). MDS analysis showed no strong differentiation within influent or effluent virus profiles of the different-sized WWTP, with virus profiles of influent samples appearing independent of the population size that the WWTP served (Fig. 4). However, virus profiles of L-WWTP influent samples showed greater similarity to each other, as shown by a clustering in the centre of the plot, than the virus profiles of M-WWTP and S-WWTP that are further apart (Fig. 4). This is a possible reflection of the smaller variation between AdV and EV concentrations in the samples taken from the L-WWTP. MDS analysis of the virus
Fig. 4 e Non-metric multi-dimensional scaling plot showing virus profile comparison of influent samples from wastewater treatment plants serving large (:), medium (D) and small (,) populations. Samples grouping closer together (:) show a more similar virus profile than those that are further apart.
6273
Fig. 5 e Non-metric multi-dimensional scaling plot showing virus profile of effluent samples collected from moving bed biofilm reactor (MBBR D), activated sludge (AS 3) and waste stabilisation pond (WSP C) wastewater treatment plants. Virus profiles show no clear similarity to each other.
profiles of the effluent showed no strong differentiation between the three groups of treatment processes (MMBR, AS and WSP) as shown by the dispersal of virus profiles of effluent samples (albeit distributed in two main groups with no identified commonality) (Fig. 5).
4.
Discussion
An important function of wastewater treatment is to remove pathogens efficiently from wastewater before discharge or disposal to minimise public health risks associated with contact of infectious pathogens via contaminated water, shellfish or aerosols. In this study, we showed that AdV, EV and NoV were present in wastewater from large (>100,000), medium (10,000e64,000), and small (<4000) populations in New Zealand. We demonstrated that although wastewater treatment processes removed EV and AdV, prevalence of NoV was similar in both influent and effluent. Both culturable AdV and EV were often found in wastewater effluent at concentrations similar to those reported elsewhere (Haramoto et al., 2007; Krikelis et al., 1985; Lodder and de Roda Husman, 2005). The range of AdV and EV concentrations by PCR in influent (3.25e8.62 and 2.84e6.67 log10 genome copies/L respectively) was also similar to other studies (Bofill-Mas et al., 2006; Katayama et al., 2008; Schvoerer et al., 2001). The high incidence and concentration of AdV by PCR in effluent samples demonstrated in this and other studies (Haramoto et al., 2007; Pina et al., 1998) may be due either to the high stability of the dominant human AdV species F (type 40 and 41) through the various WWTP processes, or the environmental resistance of AdV (Nwachuku et al., 2005). Occasionally, samples were positive by culture but negative by PCR. This discrepancy is probably due to a much smaller quantity of material tested in the PCR than in the
6274
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 6 7 e6 2 7 6
culture assays. In the culture assay, the equivalent of 200 mL sample is assayed per virus compared with an equivalent of 1 mL per PCR. Following a 1/5 dilution, even less sample was tested resulting in a potential further reduction in sensitivity. Undetected (RT)-qPCR inhibition may also cause false negative results. Demonstration of NoV infectivity is not yet possible because of NoV culture limitations (Duizer et al., 2004). From this study, assuming that NoV has similar survival properties and abundance as AdV and EV, we deduce that a proportion of NoV detected in effluents could be infectious due to the concurrent presence of culturable enteric viruses, and the ratio between culturable enteric viruses and PCR concentrations (1.4e3.9 log10 difference). Overall occurrence of NoV GI and GII was sporadic in both influent and effluent samples from all WWTP. In effluent, although variable, NoV concentrations were in the same range as previous reports (Laverick et al., 2004; Sima et al., 2011) but concentrations in influent were generally lower than the 6e8 log10 genome copies/L reported in overseas studies (Laverick et al., 2004; Lodder and de Roda Husman, 2005; Nordgren et al., 2009; Pusch et al., 2005; Sima et al., 2011; van den Berg et al., 2005). One possible reason why NoV concentrations were frequently low and sporadic in influent wastewater may be due to lower NoV prevalence in the community than AdV and EV during the study period. Generally NoV infection is associated with peaks in winter (Mounts et al., 2000), but there is no clear seasonal peak for NoV infection in New Zealand. A similar number of outbreaks are reported by the Norovirus Reference Laboratory in the summer months as during other seasons of the year (ESR, 2009). Therefore, we would not expect that the timing of the study would result in a low NoV prevalence. Whilst NoV surveillance data system does not capture all outbreaks and cases in the community, during JanuaryeMarch 2004, only 12 NoV outbreaks were recorded by the New Zealand Public Health Services, which was the same number of outbreaks as reported in the previous quarter. Most outbreaks were caused by NoV GII (ESR, 2004a,b). Similar to our finding for NoV, da Silva et al. (2007) reported that NoV (GI) occurrence in four WWTP studied was ‘erratic’. We found that NoV GI and GII prevalence were similar whereas most studies report that either NoV GII is more prevalent than NoV GI (da Silva et al., 2007; Katayama et al., 2008; Lodder and de Roda Husman, 2005) or vice versa (Nordgren et al., 2009). Data from other studies conducted in our laboratory have showed that no seasonal variation in the prevalence or quantitation of EV or AdV in influent wastewater is observed (data not shown). As the overall incidence and concentration of NoV in influent and effluent were similar, it may also suggest that NoV was more stable than AdV and EV through the various WWTP processes. As reported elsewhere (Sima et al., 2011), the NoV concentration in effluent was seemingly often unrelated to the influent, and so the efficiency of removal could not be determined or compared directly to AdV or EV. The influent virus profiles were generally independent of the WWTP size, indicating a similar epidemiological distribution of viruses in NZ. The virus profiles in effluent samples were independent of the treatment process. da Silva et al. (2007) found NoV concentrations were similar in the WWTP
serving different-sized populations. In contrast, our data showed less variance in virus prevalence and concentrations in large WWTP due to the population size. Large communities are more likely to have virus infections occurring almost continuously in the community, whereas in small communities, infection is likely to be sporadic and may affect a larger proportion of the community. The virus prevalence in large communities may also be more consistent due to comparative homogeneity of the wastewater from extended infrastructure, and better comparative mixing as wastewater travels long distances into the WWTP. The individual virus concentrations were unrelated to the size of population contributing to the treatment plant. For example, an extremely high AdV concentration was observed on one occasion in the influent of a small WWTP (8.6 log10 genome copies/L, WWTP H). This may be due to the occurrence of a possible family/small community outbreak prior to sampling, resulting in up to 1011AdV/gram faeces being excreted in the wastewater (Wadell, 1984). Such an unpredictable increase in virus concentrations would be highly significant in a small WWTP. Human AdV species AeE, mainly responsible for nonenteric illnesses, were predominately isolated using A549 cells, that are not optimal for growing human AdV-F strains. However, PCR data indicated that AdV-F was the predominant human AdV species present in influent and effluent samples at concentrations up to 8.6 log10 genome copies/L. The prevalence of human AdV-F (a common cause of gastroenteritis in children) (Uhnoo et al., 1986) in water has been previously observed (Dong et al., 2010; Haramoto et al., 2007). Further work using 293 cells showed that culturable human AdV-F was present in several influent and effluent samples (unpublished data). It is likely that culturable AdV concentrations were underestimated in this study, due to the use of A549 cells, which would influence the accuracy of gastro-enteric health risk assessments. If AdV-F survives the WWTP process, then viable viruses are likely to persist in fresh and seawaters (Enriquez et al., 1995) with the potential for transmission by the waterborne route. The estimated log10 reduction of viruses through WWTP can be theoretically determined by comparing overall virus presence in pre and post treated wastewater. For accuracy, sampling should ensure that samples are as representative as possible and also account for the time required for wastewater to pass through the plant. These issues can be problematic, as it is difficult to determine the actual time wastewater spends in a WWTP especially in a WSP treatment plant (Berg, 1973). Although wastewater is theoretically retained for weeks under normal circumstances, this may be only days if short circuiting occurs where flow bypasses most of the pond. Use of tracer dyes is the only practical method to determine the hydraulic retention time but this was outside the scope of the project. For these reasons, in this study, the median influent concentration for each population category was calculated and compared with the median effluent concentration for each population size. Although we demonstrated that the median AdV and EV concentrations were reduced through the treatment processes independent of treatment types, median NoV GI and GII concentrations (0e4 log10/L) in the effluent were often greater than those in
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 6 7 e6 2 7 6
influent (0e3 log10/L). Other studies have reported NoV reductions ranging from 0.0 to 3.6 log10 (Kuo et al., 2010; Lodder and de Roda Husman, 2005; Nordgren et al., 2009; Ottoson et al., 2006; Pusch et al., 2005; van den Berg et al., 2005). One probable reason for the lack of reduction observed for NoV may be due to its sporadic occurrence. Even over a 24 h period, this could result in variable NoV concentrations entering a WWTP, as has been previously described for other enteric viruses (Berg, 1973). In summary, this study presents virus profiles for different WWTP serving a range of populations and using different treatment processes. While culturable AdV and EV were detected frequently in effluent samples from several WWTP, the concentrations were generally lower than the influent samples. Typical influent concentrations of culturable viruses were approximately 2e3 log10/L compared to 1e2 log10/L in the effluent. The frequent presence of NoV in the effluent along with that of culturable enteric viruses is of particular concern because of the low infectious dose needed to cause NoV disease, although it is not possible to determine the infectivity of NoV. Over the three-month study period, we showed that virus presence and concentration were generally independent of the size of population served by the WWTP and, for effluent, the type of WWTP process. However, a narrower range of AdV and EV concentrations in both influent and effluent samples was observed in the larger plants compared with the smaller plants. Smaller plants were characterised by generally variable virus concentrations with sporadic spikes in both the influent and effluent. These spikes in virus concentration need to be taken into account when assessing the treatment requirements for small communities. The potential health effects of these discharges on small receiving waterways used for recreation and food gathering could be significant.
5.
Conclusions
Effective management of wastewater treatment plants requires recognition that human virus concentrations in both influent and effluent wastewater are generally independent of the WWTP size and treatment process, and that virus concentrations in influent will vary e particularly in small and medium treatment plants. Irrespective of treatment type, human viruses are likely to be present in non-disinfected effluent, with associated human health risks from virus-contaminated water. The extent of the risk may be dependent on viability, concentration, and the potential virus transmission routes of the receiving water usage.
Acknowledgements This work was funded by the New Zealand Ministry for the Environment. We appreciate the assistance of Beverly Horn and Kelly Roberts, and thank Andrew Ball, Richard Lardner, Je´re´mie Langlet and Stephen On for their critical review of the manuscript.
6275
references
Allard, A., Albinsson, B., Wadell, G., 2001. Rapid typing of human adenoviruses by a general PCR combined with restriction endonuclease analysis. Journal of Clinical Microbiology 39 (2), 498e505. Berg, G., 1973. Removal of viruses from sewage, effluents and waters. 2. Present and future trends. Bullettin of the World Health Organisation 49 (5), 461e469. Bofill-Mas, S., Albinana-Gimenez, N., Clemente-Casares, P., Hundesa, A., Rodriguez-Manzano, J., Allard, A., Calvo, M., Girones, R., 2006. Quantification and stability of human adenoviruses and polyomavirus JCPyV in wastewater matrices. Applied and Environmental Microbiology 72 (12), 7894e7896. Clarke, K.R., Gorley, R.N., 2006. PRIMER v6: User Manual/Tutorial. PRIMER-E, Plymouth, United Kingdom. da Silva, A.K., Le Saux, J.C., Parnaudeau, S., Pommepuy, M., Elimelech, M., Le Guyader, F.S., 2007. Evaluation of removal of noroviruses during wastewater treatment, using real-time reverse transcription-PCR: different behaviors of genogroups I and II. Applied and Environmental Microbiology 73 (24), 7891e7897. Donaldson, K.A., Griffin, D.W., Paul, J.H., 2002. Detection, quantitation and identification of enteroviruses from surface waters and sponge tissue from the Florida Keys using realtime RT-PCR. Water Research 36 (10), 2505e2514. Dong, Y., Kim, J., Lewis, G.D., 2010. Evaluation of methodology for detection of human adenoviruses in wastewater, drinking water, stream water and recreational waters. Journal of Applied Microbiology 108 (3), 800e809. Duizer, E., Schwab, K.J., Neill, F.H., Atmar, R.L., Koopmans, M.P., Estes, M.K., 2004. Laboratory efforts to cultivate noroviruses. Journal of General Virology 85 (Pt 1), 79e87. Enriquez, C.E., Hurst, C.J., Gerba, C.P., 1995. Survival of the enteric adenoviruses 40 and 41 in tap, sea, and waste water. Water Research 29 (11), 2548e2553. ESR, 2004a. New Zealand Public Health Surveillance Report. Volume 2, Issue 1. http://www.surv.esr.cri.nz/PDF_ surveillance/NZPHSR/2004/NZPHSR2004Summer.pdf (accessed 27.02.11). ESR, 2004b. New Zealand Public Health Surveillance Report. Volume 2, Issue 2. http://www.surv.esr.cri.nz/PDF_surveillance/ NZPHSR/2004/NZPHSR2004June.pdf (accessed 07.02.11). ESR, 2009. Annual Summary of Outbreaks in New Zealand, 2009. http://www.surv.esr.cri.nz/PDF_surveillance/AnnualRpt/ AnnualOutbreak/2009OutbreakRpt.pdf (accessed 23.08.11). Farthing, M.J.G., 1989. Viruses and the gut. In: Proceedings of the Ninth BSG. SK&F International Workshop 1988. Smith Kline and French, Welwyn Garden City, Hertforshire, UK. Green, D.H., Lewis, G.D., 1995. Enzymatic amplification of enteric viruses from wastewaters. Water Science and Technology 31 (5e6), 329e336. Green, K.Y., 2007. Caliciviridae: the noroviruses. In: Knipe, D.M., Howley, P.M. (Eds.), Fields Virology. Lippincott Williams & Wilkins, Philadelphia, PA, pp. 949e979. Greening, G.E., Hewitt, J., Lewis, G.D., 2002. Evaluation of integrated cell culture-PCR (C-PCR) for virological analysis of environmental samples. Journal of Applied Microbiology 93 (5), 745e750. Hafliger, D., Hubner, P., Luthy, J., 2000. Outbreak of viral gastroenteritis due to sewage-contaminated drinking water. International Journal of Food Microbiology 54 (1e2), 123e126. Haramoto, E., Katayama, H., Oguma, K., Ohgaki, S., 2007. Quantitative analysis of human enteric adenoviruses in aquatic environments. Journal of Applied Microbiology 103 (6), 2153e2159.
6276
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 6 7 e6 2 7 6
Heim, A., Ebnet, C., Harste, G., Pring-Akerblom, P., 2003. Rapid and quantitative detection of human adenovirus DNA by real-time PCR. Journal of Medical Virology 70 (2), 228e239. Hewitt, J., Bell, D., Simmons, G.C., Rivera-Aban, M., Wolf, S., Greening, G.E., 2007. Gastroenteritis outbreak caused by waterborne norovirus at a New Zealand ski resort. Applied and Environmental Microbiology 73 (24), 7853e7857. Hurst, C.J., Benton, W.H., McClellan, K.A., 1988. Suppression of viral replication by guanidine: a comparison of human adenoviruses and enteroviruses. Journal of Virological Methods 22 (1), 1e11. Irving, L.G., Smith, F.A., 1981. One-year survey of enteroviruses, adenoviruses, and reoviruses isolated from effluent at an activated-sludge purification plant. Applied and Environmental Microbiology 41 (1), 51e59. 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. Krikelis, V., Spyrou, N., Markoulatos, P., Serie, C., 1985. Seasonal distribution of enteroviruses and adenoviruses in domestic sewage. Canadian Journal of Microbiology 31 (1), 24e25. Kuo, D.H., Simmons, F.J., Blair, S., Hart, E., Rose, J.B., Xagoraraki, I., 2010. Assessment of human adenovirus removal in a full-scale membrane bioreactor treating municipal wastewater. Water Research 44 (5), 1520e1530. Laverick, M.A., Wyn-Jones, A.P., Carter, M.J., 2004. Quantitative RT-PCR for the enumeration of noroviruses (Norwalk-like viruses) in water and sewage. Letters in Applied Microbiology 39 (2), 127e136. Le Guyader, F.S., Bon, F., DeMedici, D., Parnaudeau, S., Bertone, A., Crudeli, S., Doyle, A., Zidane, M., Suffredini, E., Kohli, E., Maddalo, F., Monini, M., Gallay, A., Pommepuy, M., Pothier, P., Ruggeri, F.M., 2006. Detection of multiple noroviruses associated with an international gastroenteritis outbreak linked to oyster consumption. Journal of Clinical Microbiology 44 (11), 3878e3882. Lees, D., 2000. Viruses and bivalve shellfish. International Journal of Food Microbiology 59 (1e2), 81e116. Lewis, G.D., Austin, F.J., Loutit, M.W., Sharples, K., 1986. Enterovirus removal from sewage e the effectiveness of four different treatment plants. Water Research 20, 1291e1297. Lewis, G.D., Metcalf, T.G., 1988. Polyethylene glycol precipitation for recovery of pathogenic viruses, including hepatitis A virus and human rotavirus, from oyster, water, and sediment samples. Applied and Environmental Microbiology 54 (8), 1983e1988. Lodder, W.J., de Roda Husman, A.M., 2005. Presence of noroviruses and other enteric viruses in sewage and surface waters in The Netherlands. Applied and Environmental Microbiology 71 (3), 1453e1461. Metcalf, T.G., Melnick, J.L., Estes, M.K., 1995. Environmental virology: from detection of virus in sewage and water by isolation to identification by molecular biology e a trip of over 50 years. Annual Review of Microbiology 49, 461e487. Mounts, A.W., Ando, T., Koopmans, M., Bresee, J.S., Noel, J., Glass, R.I., 2000. Cold weather seasonality of gastroenteritis associated with Norwalk-like viruses. The Journal of Infectious Diseases 181 (Suppl. 2), S284eS287.
Nordgren, J., Matussek, A., Mattsson, A., Svensson, L., Lindgren, P. E., 2009. Prevalence of norovirus and factors influencing virus concentrations during one year in a full-scale wastewater treatment plant. Water Research 43 (4), 1117e1125. Nwachuku, N., Gerba, C.P., Oswald, A., Mashadi, F.D., 2005. Comparative inactivation of adenovirus serotypes by UV light disinfection. Applied and Environmental Microbiology 71 (9), 5633e5636. Ottoson, J., Hansen, A., Westrell, T., Johansen, K., Norder, H., Stenstrom, T.A., 2006. Removal of noro- and enteroviruses, Giardia cysts, Cryptosporidium oocysts, and fecal indicators at four secondary wastewater treatment plants in Sweden. Water Environment Research 78 (8), 828e834. Pallansch, M., Roos, R., 2007. Enteroviruses: polioviruses, coxsackieviruses, echoviruses and newer enteroviruses. In: Knipe, D.M., Howley, P.M. (Eds.), Fields Virology. Lippincott Williams & Wilkins., Philadelphia, PA, pp. 839e893. Pina, S., Puig, M., Lucena, F., Jofre, J., Girones, R., 1998. Viral pollution in the environment and in shellfish: human adenovirus detection by PCR as an index of human viruses. Applied and Environmental Microbiology 64 (9), 3376e3382. Pusch, D., Oh, D.Y., Wolf, S., Dumke, R., Schroter-Bobsin, U., Hohne, M., Roske, I., Schreier, E., 2005. Detection of enteric viruses and bacterial indicators in German environmental waters. Archives of Virology 150 (5), 929e947. Schvoerer, E., Ventura, M., Dubos, O., Cazaux, G., Serceau, R., Gournier, N., Dubois, V., Caminade, P., Fleury, H.J., Lafon, M.E., 2001. Qualitative and quantitative molecular detection of enteroviruses in water from bathing areas and from a sewage treatment plant. Research in Microbiology 152 (2), 179e186. Sima, L.C., Schaeffer, J., Le Saux, J.C., Parnaudeau, S., Elimelech, M., Le Guyader, F.S., 2011. Calicivirus removal in a membrane bioreactor wastewater treatment plant. Applied and Environmental Microbiology 77 (15), 5170e5177. Simmons, G., Garbutt, C., Hewitt, J., Greening, G., 2007. A New Zealand outbreak of norovirus gastroenteritis linked to the consumption of imported raw Korean oysters. The New Zealand Medical Journal 120 (1264), U2773. Sinclair, R.G., Jones, E.L., Gerba, C.P., 2009. Viruses in recreational water-borne disease outbreaks: a review. Journal of Applied Microbiology 107 (6), 1769e1780. Uhnoo, I., Wadell, G., Svensson, L., Olding-Stenkvist, E., Ekwall, E., Molby, R., 1986. Aetiology and epidemiology of acute gastroenteritis in Swedish children. The Journal of Infection 13 (1), 73. van den Berg, H., Lodder, W., van der Poel, W., Vennema, H., de Roda Husman, A.M., 2005. Genetic diversity of noroviruses in raw and treated sewage water. Research in Microbiology 156 (4), 532e540. Wadell, G., 1984. Molecular epidemiology of human adenoviruses. Current Topics in Microbiology and Immunology 110, 191e220. Wait, D.A., Sobsey, M.D., 1983. Method for recovery of enteric viruses from estuarine sediments with chaotropic agents. Applied and Environmental Microbiology 46 (2), 379e385. Wold, W.S.M., Horwitz, M.S., 2007. Adenoviruses. In: Knipe, D.M., Howley, P.M. (Eds.), Fields Virology. Lippincott Williams & Wilkins, Philadelphia, PA, pp. 2395e2436. Wolf, S., Hewitt, J., Greening, G.E., 2010. Viral multiplex quantitative PCR assays for tracking sources of fecal contamination. Applied and Environmental Microbiology 76 (5), 1388e1394.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 7 7 e6 2 8 8
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
A coupled model tree (MT) genetic algorithm (GA) scheme for biofouling assessment in pipelines Tamar Opher 1, Avi Ostfeld* Faculty of Civil and Environmental Engineering, Technion e Israel Institute of Technology, Haifa 32000, Israel
article info
abstract
Article history:
A computerized learning algorithm was developed for assessing the extent of biofouling
Received 9 February 2011
formations on the inner surfaces of water supply pipelines. Four identical pipeline
Received in revised form
experimental systems with four different types of inlet waters were set up as part of a large
15 September 2011
cooperative project between academia and industry in Israel on biofouling modeling,
Accepted 19 September 2011
prediction, and prevention in pipeline systems. Samples were taken periodically for
Available online 24 September 2011
hydraulic, chemical, and biological analyses. Biofilm sampling was done using Robbins devices, carrying stainless steel coupons. An MTeGA, a hybrid model combining model
Keywords:
trees (MTs) and genetic algorithms (GAs) in which the sampled input data are selected by
Biofouling
the proposed methodology, was developed. The method outcome is a set of empirical
Data mining
linear rules which form a model tree, iteratively optimized by a GA and verified using the
Pipeline
dataset resulting from the empirical field studies. Good correlations were achieved
Model trees
between modeled and observed cell coverage area within the biofilm. Sensitivity analysis
Genetic algorithms
was conducted by testing the model’s response to changes in: (1) the biofilm measure used
Evolutionary optimization
as output (target) variable; (2) variability of GA parameters; and (3) input attributes. The proposed methodology provides a new tool for biofouling assessment in pipelines. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
The undesired deposition and growth of microorganisms in the form of biofilms is referred to as biofouling. Biofouling may occur in practically any water system under a wide range of conditions. It may occur on the surface of membranes, within conduction pipelines, or on reservoir walls, even in oligotrophic environments or those containing disinfectants. In many cases the development of a biofilm may be cause for concern for public health, as well as for the proper function of the system in question, due to clogging or corrosion. Though the direct and indirect cost of biofouling has not been accurately determined, it is generally accepted that it is significant. The extent of biofouling is very difficult to monitor, thus
models for estimation or prediction of current or future states of biofilm formation are a necessary means for successful operation of water systems.
2.
Literature review
2.1.
Biofilms
A typical biofilm is constructed of microorganisms embedded in a matrix of microbial origin, consisting of extracellular polymeric substances (EPS). Bacteria producing EPS seem to be important to initial biofilm formation. These can cause the formation of patchy biofilms on the steel surface of the pipe,
* Corresponding author. Tel.: þ972 4 8292782; fax: þ972 4 8228898. E-mail addresses: [email protected] (T. Opher), [email protected] (A. Ostfeld). 1 Tel.: þ972 4 8293311; fax: þ972 4 8228898. 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.09.037
6278
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 7 7 e6 2 8 8
initiate the corrosion process by contributing to redox reactions involving metals (Cloete, 2003) and facilitate the settlement of other bacteria. The EPS comprise mainly of polysaccharides and proteins and their main functions are binding cells together in the biofilm, protecting them against hostile conditions, retaining water, and accumulating nutrients (Cloete, 2003). Pathogen survival in treated water has indeed been associated with biofilms growing at the waterepipe material interface, where a protective layer of polysaccharides can maintain microorganisms in a viable state, mainly due to reduced efficiency of chlorine action (Simo˜es et al., 2006). The amount of biofilm in a given system after a certain period of time depends on biofilm accumulation, which has been defined as the balance between bacterial attachment from the planktonic phase, bacterial growth within the biofilm and biofilm detachment from the surface (Stoodley et al., 1998). When that balance is null, the biofilm is said to have reached a steady-state. The final amount of biofilm in that state, which can be assessed by cell counts or biofilm mass, is directly related to the biofilm formation potential of that system.
2.2.
Modeling biofouling
Monitoring biofilm formation in a functioning system is nearly impossible. With the aim of overcoming the limitations of actual monitoring many models have been developed which attempt to predict or assess the extent of biofouling. Most of these models refer to membrane, rather than pipeline biofouling. Though these two may be equally harmful, the mechanisms involved in their formation are different. Although much progress has been made during the last couple of decades, the mechanisms by which microorganisms become attached to solid surfaces are still not well understood, thus creating difficulties in establishing reliable physical models for biofouling predictions. The processes leading to biofilm development are extremely complex, constituting chemical, physical, and biological processes interacting on different spatial and temporal scales. The extent of pipeline biofouling, the structure of the biofilm and the level of cellular activity within it depends on a large variety of factors, such as: predominant carbon source, level of biocides, hydrodynamic conditions, nutrient levels and type of pipe material. Interactions between some of these factors further complicate the task of predicting biofilm growth (Camper et al., 1999; Stoodley et al., 1999). Different aspects of membrane biofouling have been modeled in recent years. A few examples are: EPS production rates (Ahn et al., 2006; Kim et al., 2006), membrane permeability (Meng et al., 2005; Pellegrino et al., 2005) and attempts at modeling the specific mechanisms responsible for the decrease in membrane permeability, such as pore blockage, pore constriction and cake filtration (Ho and Zydney, 2000; Bolton et al., 2006; Duclos-Orsello et al., 2006). Models for biofouling of pipe surfaces are scarce and due to the complexity of the processes involved, always focus on a small selection of affecting factors and/or derived effects (Towler et al., 2007; Smith et al., 2008; Men et al., 2009).
3.
Methods
3.1.
Experimental setup
Data for modeling was acquired from four study sites set up in Israel within the framework of a large-scale cooperative research project of biofouling, involving groups from academia and industry. All four experimental systems were similarly set up at different water treatment plants which supplied their feed waters, thus each system was fed a different type of water, as detailed in Table 1. Water was fed into the system in a manner resembling flow in an actual water supply system, with pressure mostly between 2 and 3 bars. A gross filter at the entrance to the system removed large particles from the water. Pressure and flow were monitored at the entrance to the Robbins device (Ruseska et al., 1982), which served for biofilm sampling (Fig. 1). Experiment durations were up to 42 days, during which weekly samples of the feed water and biofilm were taken without interrupting continuous flow through the pipeline. The samples were analyzed for a set of hydraulic, chemical, and bacteriological parameters (Table 2).
3.2.
Robbins device
A Robbins device (Ruseska et al., 1982) consists of a central pipeline installed inline in the test system with a series of openings along its sides. In each such opening a removable stud is placed, fitted with a surface of choice (coupon) that comes in contact with the water flowing in the pipe. The device allows for several samples to be taken simultaneously as well as sampling more than a single time point in the development of the biofilm. The biofilm is sampled by extracting a stud from the device and removing the coupon fitted on it.
3.3.
Laboratory and field methods
Chemical water parameters were measured in a certified laboratory in Israel according to the standard methods for the examination of water & wastewater (Eaton et al., 2005).
Table 1 e Types and characteristics of the feed waters at the study sites in Israel and the number of samples in the dataset originating from each site. Study site
Feed water
Main characteristics
Number of data rows
Ktziot
Ground water
14
Shafdan
Secondary wastewater effluent Treated surface water Desalinated sea water
Rich with sulfides and ferrites Rich with nutrients and organic matter “Soft” water, low salinity “Hard” water, high salinity
21
Genosar Palmakhim
Total
47
18
100
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 7 7 e6 2 8 8
6279
tree is constructed with all training cases being predicted by the tree leaves (i.e., each leave is a linear regression model which predicts continuous values for the numerical attributes). The tree is then pruned bottom-up and transformed into a set of ifethen rules which simplify the tree structure, and thus improves its ability to classify new instances. The predictive ability of the tree is measured using a correlation coefficient for the training and validation datasets. The correlation coefficient equals one in case of a complete fit between measured values and model tree predictions.
3.6.
Fig. 1 e The biofouling experimental setup at the Shafdan site.
Microbial community composition of the water was determined by 16S RNA gene clone libraries. Relative abundance (in percentage) of each phylogenetic group in the community was calculated based on about 90 clones for each sampling time and location. Parameters measured on site, such as pressure, flow, temperature, turbidity etc. were taken using standard equipment or analysis kits.
3.4.
Data-driven modeling
Data-driven modeling (DDM) is a generic name for machine learning (ML) methods which make use of a set of real-life data for inductive inference regarding the observed phenomena. DDMs are capable of identifying complex nonlinear relationships between input and output datasets. On the basis of the discovered relationships within data of historical observations, inductive inference is then made for new input data. Such computerized modeling techniques are extremely useful in cases where estimation is sought, yet quantifiable physical understanding is incomplete. The enormous computational resources available nowadays make it possible to exploit vast datasets, a fact that increases estimation accuracy. Datadriven modeling can thus be considered as an approach to modeling that focuses on using computational intelligence (particularly ML) methods in building models that would complement or replace the “knowledge-driven” models describing physical behavior. The role of a modeler is to tune the DDM methods to a particular application area by screening and preprocessing the input data and appropriately selecting the decision attributes.
3.5.
Model trees (MT)
MTs are a generalization of Decision Trees (DT) which are widely used in solving classification problems and more specifically very common in data mining applications. Whereas DTs handle qualitative or discrete-value attributes only, MTs deal with continuous values. A model tree is a datadriven algorithm (Quinlan, 1992) built of a rule-based predictive structure using a top-down induction approach. The tree is fitted to a training dataset by splitting the data into homogeneous subsets based on the data attributes. Thereafter, the
Genetic algorithms (GA)
GAs are heuristic search procedures based on the mechanisms of genetics and Darwin’s natural selection principles, combining an artificial survival of the fittest with genetic operators abstracted from nature (Holland, 1975). GAs differ from other search techniques in that they search among a population of points and use probabilistic rather than deterministic transition rules. As a result, GAs search more globally (Wang, 1997; Haupt and Haupt, 1998). An initial random population of genomes within the search space is generated. Each genome represents a possible solution to the search/optimization problem and is represented by a string of values (genes), one per each search variable. Survival of the fittest is accomplished by evaluating each genome’s fitness through an appropriate objective function and a biased random selection procedure of individuals for “reproduction”, where higher rated genomes are more likely to be selected. Generation of a new population is achieved by means of crossover (partial exchange of information between pairs of strings) and mutation (a random change in a random location within the string). The fittest individuals are transferred unchanged to the next generation, an approach known as “elitism”. Every new generation of genomes is expected to be more closely concentrated in the vicinity of the optimal solution. The process is repeated until a convergence criterion is met or a pre-set maximum number of generations reached. GA input parameters include: population size, number of generations, range limits of each gene, crossover and mutation rates and a fitness function for genome evaluation. In this application GA is implemented using Roulette Wheel selection algorithm (Goldberg, 1989) for pair selection. Each chromosome pair undergoes random multiple-point crossover, resulting in an arbitrary mix of the data rows from the two data matrices comprising the parent genomes. After pairing and crossover each child genome is subject to mutations at 4 random sites with a probability of 0.8 each, i.e. a maximum of 4 mutations per genome. The principle of elitism is accomplished by passing 10% of the individuals unchanged to the next generation. Elite genomes may also be selected for pairing and procreation.
3.7.
The proposed hybrid MTeGA methodology
The hybrid approach takes advantage of the MT’s strength in solving classification problems and applies the proven capabilities of GA in optimization to polish its performance.
6280
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 7 7 e6 2 8 8
In each iteration the Model Tree’s performance is used as the GA’s objective function. Thus optimization is guided by the accuracy of the MT model’s predictions and improves it continuously. This approach has been applied with considerable success in several other fields of hydraulics and
water quality (Preis and Ostfeld, 2008; Opher and Friedler, 2009a,b; Opher et al., 2009). In the context of the work presented here, each genome in the GA application represents a hypothetical set of input and output data in the form of the actual experimental results
Table 2 e Input and output parameters list. Category (number of parameters) Inputs Time (1)
Parameter name
Description
Units
Minimum
Maximum
Age
Biofouling age
days
2
77
Hydraulics (2)
Pressure Qr
Pressure in pipe Flow through pipe
bar m3/hr
1 0
5 1
Water chemistry (50)
Ag Al Ammonia AOC As Ba Be Br Ca Cd Ch Cl Co Cr Cu DS DO DOC Fe Ha K Li Mg Mn Mo Na Ni Nitrate Nitrite Nk Pb pHFD pHLB Phosphate Se Si Sn Sr SS10 SS55 Sulfate Sulfide TDS T TEP TOC Tp TUFD TURB Zn
Argentum Aluminum NH3 Assimilated organic carbon Arsenic Barium Beryllium Bromine Calcium Cadmium Chlorophyll Chlorine Cobalt Chromium Copper Dissolved solids Dissolved oxygen Dissolved organic compounds Iron Hardness Potassium Lithium Magnesium Manganese Molybdenum Sodium Nickel NO 3 NO 2 Nitrogen kal Lead pH field pH lab PO3 4 Selenium Silica Tin Strontium Suspended solids at 105 C Suspended solids at 550 C SO2 4 S2 Total dissolved solids Temperature Transparent exopolymer particles Total organic compounds Total phosphorus as P Turbidity field Turbidity Zinc
mg/L mg/L mg/L mg/L mg/L mg/L mg/L mg/L mg/L mg/L mgChl/L mg/L mg/L mg/L mg/L mg/L mg/L mg/L mg/L mg/L mg/L mg/L mg/L mg/L mg/L mg/L mg/L mg/L mg/L mg/L mg/L e e mg/L mg/L mg/L mg/L mg/L mg/L mg/L mg/L mg/L mg/L C mgGX/L mg/L mg/L NTU NTU mg/L
1 1 0 70 1 0 0 1 44 0 0 218 2 2 2 677 2 0 0 243 8 0 20 2 2 125 2 0 0 0 1 4 8 0 1 3 0 1 0 0 59 0 626 14 132 0 0 0 0 3
10 514 12 323 8 82 0 79 448 2 58 22,548 8 8 73 677 9 13 485 7142 560 216 1473 73 60 13,150 35 7 12 19 8 9 8 3 8 3 13 11,155 71 26 59 3 43,250 38 4394 18 3 33 7 228
6281
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 7 7 e6 2 8 8
Table 2 e (continued ) Category (number of parameters) Water bacteriology (27)
Outputs Biofouling structure (5)
Parameter name
Description
Units
Acidithiobacillus Acidobacteria Actinobacteria Alphaproteobacteria Bacteriodetes Betaproteobacteria Chlorobi Chloroflexi Cyanobacteria Deferribacteres Deinococcus-Thermus Deltaproteobacteria GN09 OD1 OP11 TM6 TM7 Epsilonproteobacteria Fibrobacters Firmicutes Gammaproteobacteria Halothiobacillus Planctomycetes Thiomonas Verrucomicrobia WS6 Other bacteria
Measured bacteria
% of total population
CB CCA EPS-B
Cell biovolume Cell coverage area Extracellular polymeric substances-biovolume Extracellular polymeric substances-coverage area Total biovolume
mm3 % mm3
EPS-CA TB
(each data row in Table 2 serves as a gene). The initial population consists of a pre-set number of such genomes, each randomly generated within a range defined by boundaries which are derived from the statistical characteristics of the experimental dataset. The actual experimental data collected from all four study sites serves for evaluation of the MT models constructed during the optimization process. In every generation the GA module calls the MT module for each of the genomes in the current population. The MT module constructs a model using the synthetic training data coded by the genome and passes back to the GA module this model’s fitness score, which is the accuracy of its correlation with the evaluation data, in the form of a Pearson correlation coefficient (R) which is a measure of the strength of a linear dependency between two variables (see mathematical definition at any statistical handbook, e.g., Bruning and Kintz, 1997). A minimum average square error (ASE) is sought as a second priority to R (i.e. of two genomes carrying an identical R score, the one with the smaller ASE is considered as having better fitness; the prioritization of R over ASE can be considered as ‘lexicographic preferences’). As the GA population advances toward the objective function optimum (i.e.
% mm3
Minimum
Maximum
15 0 1 1 0 1 0 1 0 0 0 0 0 0 0 0 0 0 0 0 1 26 0 11 0 0 0
58 0 73 40 64 34 2 6 14 0 1 5 0 18 9 1 29 5 6 10 84 63 4 25 2 3 18
2014 3 3009
360,000 48 470,000
4
55
4614
690,000
maximum R value with a minimum average square error), the corresponding MT constructed becomes more accurate in predicting the target value of the evaluation data. The coupled MTeGA model was written in Matlab, incorporating the commercial Cubist M5 model tree protocol (Rulequest-Research, 2007) as the core of the MT module. A schematic description of the proposed methodology is given in Fig. 2. The program was run on a 2.80 GHz PC with 2.00 GB of RAM, taking about 15 min to complete a single cycle of 50 GA generations, as described below.
3.8.
Base run (BR)
All MTeGA runs were carried out with constant GA parameters and consisted of 90 genes per genome, 40 genomes per population, 50 generations per execution and the fitness function described above. Input and output variables varied for the BR and sensitivity analysis, as outlined below. 24 repetitions of MTeGA were run with CCA (biofilm cell coverage area) as target attribute, input data constraints of 3 times the observed maximum for each variable and 22 input variables. Model inputs were limited by categorization in the
6282
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 7 7 e6 2 8 8
Fig. 2 e Methodology flowchart. following manner: age (i.e. time elapsed from previous coupon replacement), one hydraulic variable out of the two available (pressure and flow), 15 of the 50 available chemistry measures and 5 bacteriological inputs of the existing 27 (see categories and data availability in Table 2 and Fig. 3). At the initiation of each MTeGA execution input variables were randomly selected according to the determined category limitations. Of the resulting 24 model trees, each making use of a different set of classification attributes, the one with the best predicted vs. observed performance was chosen as the BR for further reference.
3.9.
Sensitivity analyses
To assess the model’s sensitivity 4 additional sets of MTeGA runs were performed, a single parameter varied in each (Table
3), excluding sensitivity analysis 4 (SA4), as will be further elaborated below. The model’s response to each change was examined to determine the extent to which each of the factors affects the modeling process. SA1 tests MTeGA’s sensitivity to the target attribute, by looking at its performance for cell biovolume (CB) (SA1a), and EPS coverage area (EPS-CA) (SA1b), while keeping all other parameters constant. SA2 examines the extent to which the input variables’ variability (dictated by the data constraints for the genomes generated within the GA) affects the MTeGA outcome. In SA2a variability was decreased by 30%, by allowing a maximum of twice the measured maximum per variable. In SA2b it was increased by 30%, by setting the multiplication factor to 4. In SA3, the effect of restricting input attributes by categorization was put to a test by comparing the BR outcome to that of a run with 22 uncategorized inputs,
Fig. 3 e Schematics of the database structure.
6283
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 7 7 e6 2 8 8
Table 3 e Base run and sensitivity analyses modeling parameters. Run type
BR SA1a SA1b SA2a SA2b SA3 SA4
Number of selected input attributes (total available quantity) Age (1)
Hydraulics (2)
Water chemistry (50)
Water bacteriology (27)
1 1 1 1 1 22 1
1 1 1 1 1
15 15 15 15 15
5 5 5 5 5
1
15
5
Target output attribute
Input data constraint factor
CCA CB EPS-CA CCA CCA CCA CCA
3 3 3 2 4 3 3
BR ¼ base run; SA1a ¼ sensitivity analysis 1a; CCA ¼ Cell coverage area; CB ¼ Cell biovolume; EPS-CA ¼ Extracellular polymeric substancescoverage area.
while keeping all other factors unchanged. Finally in SA4 the dataset was partitioned into w50% for fitness evaluation and w50% for results verification. The SA4 partition is aimed at demonstrating the model results on a non explored verification fraction of the dataset. SA4 is not intended (as of the small number of data points) to present the outcome of a training and validation dataset evaluation of a data-driven modeling framework. The challenges of utilizing few number of data points in a data-driven modeling scheme are further discussed at the conclusions section.
4.
of input attributes per category (e.g. hydraulics, chemistry, water bacteriology) were defined, as well as the target attribute, which may have been any one of the 5 available biofilm measures (Table 2). The specified number of input variables was then randomly selected from each category and rows containing missing data were omitted, thus significantly reducing the size of the useful dataset. The resulting dataset associated with each specific MTeGA consists of 11e24 rows, depending on the variables selected.
5.
Results
5.1.
Base run (BR)
Data
The full dataset consists of 167 rows (Fig. 3), each a result of an individual sampling event. Due to technical problems at the study sites some of the data rows did not contain enough data for modeling purposes and were left out. Data rows representing the initial “time 0” of each experiment were not used since no data describing the biofilm, serving as output to the model, is available at that point in time. In the 100 data rows remaining, there still are numerous blank cells due to incomplete laboratory analyses of the water samples (Fig. 3). The available field and laboratory data served as evaluation data (ED) for model testing and scoring during the GA optimization process. Prior to each MTeGA run, a certain number
Table 4 summarizes the results of the base run and the sensitivity analyses. Cell coverage area was chosen as the target attribute for the base run of this modeling experiment, giving a very accurate estimation of the 11 available observations, with R ¼ 0.98 and a relative average error (RAE) of 7% (relative to the average observed CCA) (Fig. 4). Of the 22 input attributes only 6 were chosen by the algorithm as effective classifiers for the model tree: age, aluminum, cadmium, manganese, turbidity and the presence of the bacteria Acidithiobacillus in the feed water (Fig. 4). Fig. 5 describes the Pearson correlation coefficient
Table 4 e Results summary. Run type BR SA1a SA1b SA2a SA2b SA3 SA4
Target output attribute
Pearson correlation coefficient
Relative average errora
Nb
CCA CB EPS-CA CCA CCA CCA CCA
0.98 0.81 0.92 0.99 0.97 0.98 0.81
0.07 0.60 0.27 0.12 0.11 0.13 0.51
11 24 15 11 14 11 6c
BR ¼ base run; SA1a ¼ sensitivity analysis 1a; CCA ¼ Cell coverage area; CB ¼ Cell biovolume; EPS-CA ¼ Extracellular polymeric substancescoverage area. a Average of the absolute error divided by the observed average. b Number of data points. c Verification set.
6284
Cell coverage area (%)
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 7 7 e6 2 8 8
40 30
Legend
20
Measured Modeled
10 0
1
2
3
4
5 6 7 Data point
8
9
10
11
Model tree (see parameters list and units at Table 2): CCA = 12.81 + 0.168 Age – 0.224 Mn + 7.7 Cd + 0.13 TUFD + 0.0126 Al + 0.241 Acidithiobacillus
Fig. 4 e Base run (BR) results.
30
R
0.98
25
0.97 0.96
20
0.95 0.94
15
0.93 10
0.92 0.91
5
ASE
0.9 0.89
0 1
11
21
31
Average square error (ASE) (%)2
Pearson correlation coefficient (R) (-)
0.99
41
Generation number
Fig. 5 e Pearson correlation coefficient (R) and average square error (ASE) convergence for the base run (BR).
fairly accurately estimate the observed, with R ¼ 0.92 and RAE ¼ 27% (Fig. 7).
(R) and the average square error (ASE) convergence for the BR. Similar convergence behavior was observed at all other model runs.
Sensitivity analysis 2 e data constraints
5.3. 5.2.
Sensitivity analysis 1 e target variable When running MTeGA for the output CCA, neither narrowing nor widening input attributes’ variability, resulted in significant changes in model performance. In SA2a (data range narrowed by 30%) the final MT’s predicted vs. observed correlation is 0.99, with RAE ¼ 12% (Fig. 8). In SA2b, where input data distribution was 30% wider, R ¼ 0.97 and RAE ¼ 11% (Fig. 9).
Cell biovolume (μm 3)
Models were generated for four additional target variables which are potential indicators of the extent of biofouling: cell biovolume (CB), EPS biovolume (EPS-B), EPS coverage area (EPS-CA) and total biovolume (TB). Of these four, EPS-B and TB resulted in model trees with poor performance in estimating the observed biofilm measures (results not shown). The MTeGA executions for CB and EPS-CA resulted in good fitting models in terms of correlation between predicted and observed data for the sets of field data available. The CB model (SA1a) displayed inferior performance of R ¼ 0.81 and an RAE of 60% (Fig. 6), where the modeled values of EPS-CA (SA1b)
Sensitivity analysis 3 e input category restriction
5.4.
When running MTeGA with a free choice of 22 inputs from the complete set of 80 available variables, due to the enormous
1.E+05 8.E+04 Legend
6.E+04
Measured Modeled
4.E+04 2.E+04 0.E+00 1
3
5
7
9
11
13
15
17
19
21
23
Data point Model tree (see parameters list and units at Table 2): CB = 52663 – 99129 Qr – 9.8 Ha + 4003 Sn – 1956 TOC + 1494 Bacteriodetes + 2872 Thiomonas
Fig. 6 e Sensitivity analysis 1a (SA1a) results.
6285
Extracellular polymeric substances-coverage area (%)
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 7 7 e6 2 8 8
50 40 Legend
30
Measured Modeled
20 10 0
1
3
5
7
9
11
13
15
Data point Model tree (see parameters list and units at Table 2): EPS-CA = 14.737 + 0.193 Age - 0.209 Mn + 1.48 pHFD – 0.7 Nk + 4 Cd + 0.384 Acidithiobacillus
Cell coverage area (%)
Fig. 7 e Sensitivity analysis 1b (SA1b) results.
40 30
Legend Measured Modeled
20 10 0 1
2
3
4
5
6
7
8
9
10
11
Data point Model tree (see parameters list and units at Table 2): If Mn ≤ 112.4 Then CCA = 13.54 + 1.22 Ag + 0.022 Al – 0.301 Mn + 0.88 Nitrate + 0.261 Acidithiobacillus If Mn > 112.4 Then CCA = 76.76 + 0.18 Ag + 0.0036 Al + 0.14 Nitrate + 0.03 Acidithiobacillus
Fig. 8 e Sensitivity analysis 2a (SA2a) results.
number of possible combinations (340), a larger number of repeated executions were necessary. Of the 120 MTs constructed few were of good calibre and one matched the quality of performance displayed by the BR model (Fig. 10). This model has an R of 0.98 and an RAE of 13%, which is slightly higher than that of the BR. The combination of 22 inputs, which were randomly chosen from the full list of variables in the construction of this model, included the variable age, one hydraulic, 11 chemical, and 9 bacteriological inputs.
5.5.
verification. The 7 fitness evaluation points are utilized, as in all other runs, for assessing the model trees fitness (i.e., its ability to predict biofouling), with which the best model tree is eventually selected. Once the best model tree is chosen, its performance is measured on the 6 points of the verification set. This yields a predicted vs. observed correlation of 0.81, and an RAE of 51% (Table 4). The 0.81 and 51% figures are lower than most of the other runs yet they are still within the range of expected outcomes of the proposed model. Of the subgroup of 6 and 7 decision attributes actually used in the BR and in SA3 respectively, two are in common to both: the variables age and Acidithiobacillus. Between the chemical input variables there is no overlapping. While in the BR 4 chemical variables (3 metals þ turbidity) are included in the
Sensitivity analysis 4 e verification
Cell-coverage area (%)
In SA4 (Fig. 11) the dataset is partitioned into w50% (7 points) for fitness evaluation and w50% (6 points) for results
40 30
Legend
20
Measured Modeled
10 0 1
3
5
7
9
11
13
Data point Model tree (see parameters list and units at Table 2): CCA = 3.095 – 0.132 Age + 2.79 Pressure + 1.2 Nitrate + 0.00144 Sr + 0.175 Acidithiobacillus
Fig. 9 e Sensitivity analysis 2b (SA2b) results.
6286
Cell coverage area (%)
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 7 7 e6 2 8 8
40 30
Legend Measured Modeled
20 10 0 1
2
3
4
5
6
7
8
9
10
11
Data point Model tree (see parameters list and units at Table 2): CCA = 7.58 + 0.24 Age + 5 Qr + 0.14 Ni – 0.55 Nk + 0.341 Acidithiobacillus + + 0.18 Betaproteobacteria – 0.63 Other bacteria
Cell coverage area (%)
Fig. 10 e Sensitivity analysis 3 (SA3) results.
35 30
Fitness
Verification
25
Legend
20
Measured Model
15 10 5 0 1
2
3
4
5
6
7
8
9
10
11
12
13
Data point Model tree (see parameters list and units at Table 2): CCA = 2.29 + 2.86 Pressure + 0.192 Br - 7.6 Cd + 0.023 Ammonia
Fig. 11 e Sensitivity analysis 4 (SA4) results.
model rules, in the SA3 model only two variables of the chemistry category are used (nickel and nitrogen kal). In SA4 one hydraulics group (pressure) and three chemical variables are selected.
6.
Discussion
Results show that the MTeGA approach to modeling biofilm formation in water conduction pipes may succeed where other modeling methods have failed. Though simple statistical dependencies between biofilm characteristics and system qualities are not apparent, the algorithm was able to produce rather accurate model trees that estimate certain measures of biofouling. The choice of target variable seems to be critical to the success of the modeling process, as shown by the dramatic drop in performance for other outputs, in sensitivity analysis 1. Biofilm cell coverage area is the one biofouling measure of the five tested in this study which gave the best outcomes in terms of modeled vs. observed correlation. Working with this specific variable as output, the MTeGA process shows considerable robustness, as shown by sensitivity analysis 2, which displayed very little effect of changes in training data variability. The outcome of SA3 in which no limitations were forced on input selection from the long list of variables available, supports the assumption that the modeling potential of the MTeGA methodology was fully utilized, as its performance did not exceed that achieved by the base run. The SA4 outcome gives rise to the model ability to perform its
task as evident from the model results on a non explored verification portion of the dataset. Apparently, the presence of the phylogenetic group of Acidithiobacillus bacteria is a valuable indicator of the state of a biofilm, as it appears in four CCA models and in 5 of the total 7 models created. This empirical observation coincides with experimental and theory evidence on the significance of the Acidithiobacillus bacteria on biofouling formations (Sand and Gehrke, 2006; Jameson et al., 2010). Four of the five CCA models (SA2a, SA2b, SA3, and SA4) include some form of nitrogen as an input attribute. The concentrations of nutrients in the water are generally a good indication of the presence of bacteria, though in chlorinated water, as are some of the feed waters, with levels of Cl as high as 22,000 mg/L in Palmakhim and 2000 mg/L in Ktziot (Table 2), this is not the case. Yet surprisingly, Cl is not included as an input attribute in any of our MTeGA models. This may hint that biofilm structure provides such good defense to its constituents, that even a few bacteria cells surviving the disinfectant in the water and settling on the pipe surface, may be sufficient to initiate a thriving population. Time of exposure to feed water, expressed in the variable named Age was incorporated in 3 of the five models for CCA, indicating that this is a major factor for biofouling predictions yet it might be suppressed by other hydraulic, chemical, or biological parameters of the supplied water. The problem of chronological order of events remains to be dealt with, as the input attributes in each entry of the dataset relate to the date of biofilm sampling. Thus the outcome of this exercise is a model which estimates the current state of
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 7 7 e6 2 8 8
biofouling, rather than its future growth. As this exercise was intended as a feasibility test and preliminary exploration of MTeGA in the context of biofouling, no manipulations on input data were performed. After having proved applicability of this approach for biofouling, we plan to alter training and validation data to construct MTeGA models which will attempt to predict the state of the biofilm a certain amount of time in advance, hence having greater value as an aid for system planning or operation.
7.
Conclusions
The developed methodology of MTeGA seems to serve the goal of biofilm estimation quite well under the right circumstances and data considerations. More work should be done to narrow down the list of effective input attributes, to enable sampling fewer parameters, while still achieving good modeling. The identity of the target attribute has a significant effect on the outcome of the models’ accuracy of predictions. Cell coverage area was found to be a good predictive measure of the state of the biofilm, while other measures tested did not yield useful models. Additional biofilm characteristics should be tested as target variables, as there may be others which would provide better results. The main limitation of the proposed methodology is the lack of proper validation as of the small available database. This observation addresses a major challenge research topic of developing data-driven models in cases where data is short. Such modeling efforts may include expansion of data points using hybrid physical-data driven modeling schemes and/or the utilization of statistical methodologies for capturing the database multivariate statistical properties.
Acknowledgments This work was supported by the Israeli Ministry of Industry and Trade under project MAGNET Biofouling Consortium.
references
Ahn, Y.T., Choi, Y.K., Jeong, H.S., Chae, S.R., Shin, H.S., 2006. Modeling of extracellular polymeric substances and soluble microbial products production in a submerged membrane bioreactor at various SRTs. Water Science and Technology 53 (7), 209e216. Bolton, G., LaCasse, D., Kuriyel, R., 2006. Combined models of membrane fouling: development and application to microfiltration and ultrafiltration of biological fluids. Journal of Membrane Science 277, 75e84. Bruning, J.L., Kintz, B.L., 1997. Computational Handbook of Statistics. Longman Inc., 361 p. Camper, A., Burr, M., Ellis, B., Butterfield, P., Abernathy, C., 1999. Development and structure of drinking water biofilms and techniques for their study. Journal of Applied Microbiology 85 (S1), 1Se12S.
6287
Cloete, T.E., 2003. Biofouling control in industrial water systems: what we know and what we need to know. Materials and Corrosion 54, 520e526. Duclos-Orsello, C., Li, W., Ho, C.C., 2006. A three mechanism model to describe fouling of microfiltration membranes. Journal of Membrane Science 280, 856e866. Eaton, A.D., Clesceri, L.S., Rice, E.W., Greenberg, A.E., Franson, M. A.H. (Eds.), 2005. Standard Methods for the Examination of Water & Wastewater, Centennial ed.Standard Methods for the Examination of Water and Wastewater American Public Health Association. Goldberg, D.E., 1989. Genetic Algorithms in Search, Optimization, and Machine Learning. Addison-Wesley, New York, 412 p. Haupt, R.L., Haupt, S.E., 1998. Practical Genetic Algorithms. John Wiley & Sons, Inc., 177 p. Ho, C.C., Zydney, A.L., 2000. A combined pore blockage and cake filtration model for protein fouling during microfiltration. Journal of Colloid and Interface Science 232, 389e399. Holland, J.H., 1975. Adaptations in Natural and Artificial Systems. The University of Michigan Press, Ann Arbor. Jameson, E., Rowe, O.F., Hallberg, K.B., Johnson, D.B., 2010. Sulfidogenesis and selective precipitation of metals at low pH mediated by Acidithiobacillus spp. and acidophilic sulfatereducing bacteria. Hydrometallurgy 104, 488e493. Kim, A.S., Chen, H., Yuan, R., 2006. EPS biofouling in membrane filtration: an analytic modeling study. Journal of Colloid and Interface Science 303, 243e249. Men, H., Li, Y., Tian, F., 2009. Microbial growth model and stability analysis. In: First International Workshop on Education Technology and Computer Science, 2009, vol. 3. IEEE, Wuhan, Hubei, pp. 29e33. Meng, F., Zhang, H., Li, Y., Zhang, X., Yang, F., 2005. Application of fractal permeation model to investigate membrane fouling in membrane bioreactor. Journal of Membrane Science 262, 107e116. Opher, T., Friedler, E., 2009a. Development of a Data Driven Model for Estimating Pollutant Levels in Highway Stormwater Runoff. VDM Verlag Dr. Muller, Saarbrucken, Germany, 141 p. Opher, T., Friedler, E., 2009b. A preliminary coupled MTeGA model for the prediction of highway runoff quality. Science of the Total Environment 407, 4490e4496. Opher, T., Ostfeld, A., Friedler, E., 2009. Modeling highway runoff pollutant levels using a data driven model. Water Science & Technology 60 (1), 19e28. Pellegrino, J., Wright, S., Ranvill, J., Amy, G., 2005. Predicting membrane flux decline from complex mixtures using flowfield flow fractionation measurements and semi-empirical theory. Water Science and Technology 51 (6e7), 85e92. Preis, A., Ostfeld, A., 2008. A coupled model tree e genetic algorithm scheme for flow and water quality predictions in watersheds. Journal of Hydrology 349, 364e375. Quinlan, J.R., 1992. Learning with continuous classes. In: 5th Australian Joint Conference on Artificial Intelligence. World Scientific, Singapore. Rulequest-Research, 2007. Data Mining with Cubist. http://www. rulequest.com/cubist-info.html (accessed 27.02.06). Ruseska, I., Robbins, J., Lashen, E.S., Costerton, J.W., 1982. Biocide testing against corrosion-causing oilfield bacteria helps control plugging. Oil Gas Journal 25, 253e264. Sand, W., Gehrke, T., 2006. Extracellular polymeric substances mediate bioleaching/biocorrosion via interfacial processes involving iron (III) ions and acidophilic bacteria. Research in Microbiology 157 (1), 49e56. Simo˜es, L.C., Azevedo, N., Pacheco, A., Keevil, C.W., Vieira, M.J., 2006. Drinking water biofilm assessment of total and culturable bacteria under different operating conditions. Biofouling 22 (2), 91e99.
6288
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 6 2 7 7 e6 2 8 8
Smith, P., Roy, S., Swailes, D., Maxwell, S., Page, D., Lawson, J., 2008. A Predictive Model for Microbial Induced Corrosion (MIC) in Sub-sea Production Pipelines: Part 1dAbiotic Corrosion Model. Chevron Upstream Europe. Society of Petroleum Engineers, Aberdeen, UK. Stoodley, P., Dodds, I., Boyle, J.D., Lappin-Scott, H.M., 1998. Influence of hydrodynamics and nutrients on biofilm structure. Journal of Applied Microbiology 85 (S1), 19Se28S.
Stoodley, P., Dodds, I., Boyle, J.D., Lappin-Scott, H.M., 1999. Influence of hydrodynamics and nutrients on biofilm structure. Journal of Applied Microbiology 85 (S1), 19Se28S. Towler, B.W., Cunningham, A., Stoodley, P., McKittrick, L., 2007. A model of fluidebiofilm interaction using a Burger material law. Biotechnology and Bioengineering 96 (2), 259e271. Wang, Q.J., 1997. Using genetic algorithms to optimise model parameters. Environmental Modelling & Software 12 (1), 27e34.