Sources and Persistence of Fecal Coliform Bacteria in a Rural Watershed 33–47 R.C

vii

WATER QUALITY RESEARCH JOURNAL OF CANADA

Volume 38, No. 1, 2003

Theme Issue — Drinking Water: From Source to Tap

Foreword 1 K.L.E. KAISER

Overview — Drinking Water Quality and Sustainability 3–13 S.B. WATSON AND J. LAWRENCE

Spatial and Temporal Distribution of Fecal Indicator Bacteria 15–32 within the Oldman River Basin of Southern Alberta, Canada R. HYLAND, J. BYRNE, B. SELINGER, T. GRAHAM, J. THOMAS, I. TOWNSHEND AND V. G ANNON

Sources and Persistence of Fecal Coliform Bacteria in a Rural Watershed 33–47 R.C. JAMIESON, R.J. GORDON, S.C. TATTRIE AND G.W. STRATTON

Portrait of Drinking Water Quality in Small Quebec Municipal Utilities 49–76 H.D. COULIBALY AND M.J. RODRIGUEZ

Conceptual Model for Cryptosporidium Transport in Watersheds 77–113 C.H. PARK AND P.M. HUCK

Actinomycetes in the Elbow River Basin, Alberta, Canada 115–125 B. ZAITLIN, S.B. WATSON, J. DIXON AND D. STEEL

New Microcystin Concerns in the Lower Great Lakes 127–140 T.P. MURPHY, K. IRVINE, J. GUO, J. DAVIES, H. MURKIN, M. CHARLTON AND S.B. WATSON

Mass Appearance of Cyanobacterium Planktothrix rubescens 141–152 in Lake Piaseczno, Poland D. KRUPA AND K. CZERNAS´

Characterization of Particles in Slow Sand Filtration at 153–168 North Caribou Water Treatment Plant B. GORCZYCA AND D. LONDON

Sulfate Removal from Water 169–182 A. DARBI, T. VIRARAGHAVAN, Y.-C. JIN, L. BRAUL AND D. CORKAL

continued on next page A Kinetic Model for Autotrophic Denitrification using Sulphur: 183–192 Limestone Reactors A. DARBI AND T. VIRARAGHAVAN

Removal of Arsenic from Groundwater using Crystalline Hydrous 193–210 Ferric Oxide (CHFO) B.R. MANNA, S. DEY, S. DEBNATH AND U.C. GHOSH

Philip H. Jones Student Award—18th Canadian Eastern Symposium on ix Water Quality

Introductory Remarks

Drinking Water Quality and Sustainability

SUSAN B. WATSON AND JOHN LAWRENCE

National Water Research Institute, 867 Lakeshore Road, Burlington, Ontario L7R 4A6

Introduction

Drinking water has received considerable attention recently. However, the marked difference between escalating demand and shrink- ing supplies illustrates that our progress towards sustainable management of this vital resource has been less than adequate. Over the past few decades, global drinking water consumption has outpaced population growth (Fig. 1), while misuse and mismanagement have resulted in a rapid and widespread decline in source-water quality and supply. Nevertheless, the growing movement among governments, scientists, and private and public sectors to evaluate current practices, adopt more stringent quality control, and develop integrated and proactive resource management is encouraging. Canada is a key player in the global balance between water supply and consumption. This country faces many of the same drinking water concerns as the rest of the world. Canada has ~9% of the total available freshwater in

Fig. 1. Past and predicted global population growth and municipal water consumption (source: Shiklomanov 1999). 4 WATSON AND LAWRENCE the world, almost half the lakes, several major river basins (e.g., Yukon, Mackenzie, St. Lawrence) and a huge diversity of climate, catchments and source waters. With a wealth of scientific and technical resources, this country has the opportunity to contribute to the development of an integrated national and international sustainable approach to this issue. We refer readers to a number of excellent reviews on the state of global freshwater resources (e.g., Bau et al. 1991; Postel et al. 1996; Ongley 1997; Gleik 1999, 2001; Shiklomanov 1999; WHO/UNICEF 2000; Vörösmarty et al. 2000; Jackson et al. 2001; Pringle 2001; Somlyódy et al. 2001). Global water supply and demand estimates differ somewhat among studies (see Gleik 2001 for an in-depth comparison), but all of these authors predict that with the current population and water use patterns, growth in consumption is predicted to continue, with the largest increase in Asia. Among the more industrialized continents, North America will remain as the top consumer (Fig. 2). Overall, three fundamental points are clear. First, the current approach to the management and consumption of freshwater supplies is small-scale and exploitative. Second, this approach to water is unsustain- able because of the profound impacts of human growth and activity on this resource. Third, the overriding factors that will continue to determine

Fig. 2. Trends in total freshwater consumption by continent; inset shows patterns on a smaller scale, with global and Asian consumption removed. (Redrawn data from the State Hydrological Institute, St. Petersburg, Russia, I. Shiklomanov, http://espejo.unesco.org.uy/part’3/_read’me.html). INTRODUCTORY REMARKS 5

the future of freshwater resources are human population growth and distribution, and economic development. This theme issue of the journal is devoted to drinking water, from source to tap. Many of the papers in this issue were first presented at the 37th Central Canadian Symposium on Water Pollution Research co-sponsored by the Canadian Association on Water Quality and the National Water Research Institute. The articles fall into two main groups, source water quality and management and water treatment. They represent examples of the broad range of issues related to sustainable water management. In this introductory review, we examine global and Canadian drinking water concerns, and Canada’s contribution to current research in the field. We identify some of the key areas in drinking water treatment and management still critically in need of attention. Above all, we under- score the principle that a sustainable drinking water supply requires management at source, treatment plant, tap, and beyond.

Drinking Water Demand and Supply: Growth, Demographics and Development

Freshwater supplies are neither isolated nor infinite, but part of a single resource continuum shaped by large- and small-scale processes. Although this concept of ‘hydrological connectivity’ has gained formal recognition in the scientific literature (Pringle 2001), it has not been adopted widely by water managers for several reasons (Somlyódy et al. 2001). A narrow, exploitative approach usually has been more immediately prof- itable; supplies are drawn from catchments that are partitioned among multiple stakeholders and regulatory agencies; and vital data on catchment-scale hydrologic processes are rarely available (Pringle 2001; this issue: Park and Huck 2003). Freshwater supplies are influenced by temporal and geographical variation in natural and anthropogenic factors. In Canada, raw water levels of major ions, nutrients and metals (notably iron and arsenic), and labile and non-labile dissolved organic compounds (DOC) vary significantly across region and with season. Surface waters show an extensive diversity across Canada, ranging from ultra-oligotrophic glacial headwaters to eutrophic/hypereutrophic saline lakes and dugouts. Groundwater sources, which collectively supply over 26% of Canadians, have estimated residence times from ~2 days to >1000 years, affecting their accumulation of total dissolved solids (~25–2 x 105 mg/L) and contaminants (this issue: Darbi et al. 2003; Darbi and Viraraghavan 2003; Manna et al. 2003; A. Crowe pers. comm.). These differences in supply characteristics are important, because domestic water consumption is highly dependent on both the local availability of water, and population size (UNESCO 2000). In addition, aes- thetics (taste and odour) play a significant role, especially in developed 6 WATSON AND LAWRENCE countries where there is both a growing awareness of drinking water safety issues, and access to alternative supplies (Watson In press; WHO 1999). Human activity affects the water resource continuum at many levels. Climate warming is changing spatial and temporal patterns in freshwater supply (Schindler 2001; Vörösmarty et al. 2000). Long-range atmospheric transport is resulting in the movement of persistent organic pollutants (POPs) across catchment basins and aquatic ecosystems (Bro-Rasmussen 1996; Blais et al. 2001). Surface water deployment and impoundment modify watershed and source-water characteristics, and generate significant evaporative losses (Day and Quinn 1992). Eutrophication and pollutants change aquatic systems towards decreased stability, water quality and food web diversity, and more outbreaks of invasive and noxious species (Mills et al. 1993; Paerl 1997; Revenga et al. 1998; this issue: Krupa and Czernas 2003; Murphy et al. 2003). Population growth and demographics also have substantial impacts on water supply and quality. Almost two billion people (30%) now face water scarcity, and with current growth, it is estimated that by 2025 water stress will affect two thirds of the human population (Vörösmarty et al. 2000; UNESCO 2000). The current migration from rural areas to urban centres is creating nodes of high freshwater demand and intense pollutant stress. Politics at local, regional and national levels generate conflicting water-related issues (Gleik 2001). In Canada, for example, a number of major trans-boundary catchment basins and rivers experience fragment- ed and changeable source-water use (Wolf et al. 1999). Because Canada has a small population (ca. 31 million) and large geographic area (>9 million km2), demographics are particularly important in freshwater consumption patterns. The population has developed with a highly uneven distribution. An estimated 65% of the total population lives in 25 central metropolitan areas such as seen in Southern Ontario (<1% total land area of Canada; Statistics Canada 2001), with the associated stress on local environment and water resources. Minimal watershed management has allowed widespread degradation of ground- and surface waters by industry, agriculture and domestic sources (e.g., Brownlee et al. 1993; Chen et al. 1998; Peel and Campbell 1999; Maule and Fonstad 2000; Schindler 2000; Fox 2001; Ritter et al. 2002; this issue: Gorczyca and London 2003; Hyland et al. 2003; Zaitlin et al. 2003). In spite of these current and predicted shortages of drinking water, agriculture and industry are by far the largest users of global freshwater resources (ca. 67% and 20%, respectively). Furthermore, this striking imbalance in resource partitioning is predicted to increase (Postel et al. 1996; Gleik 2001; Wallace 2000; Fig. 3). Agriculture and industry also incur the greatest losses through evaporation from reservoirs and irrigation, and contribute significantly to the degradation and pollution of source and supplies. Consumption patterns differ at national and regional levels, and are continually shifting as these two sectors become more centralized and intense. In Canada, industry is presently the largest overall user (~80% use; Fig. 4), particularly in regions with extensive hydro- and ther- INTRODUCTORY REMARKS 7

Fig. 3. Past and predicted global water withdrawal by sector: agriculture (white series), industry (grey series) and municipalities (black series) (source: Shiklomanov 1999).

mal-electric developments. In areas such as the prairies, agriculture (i.e., irrigation) is the major consumer. Differences in surface versus ground water availability and characteristics modify regional consumption; groundwater provides from 1% to 100% of the total freshwater supply in individual provinces (Nunavut and P.E.I., respectively; Environment Canada 1990; Somlyódy et al. 2001). These consumption patterns exacer- bate the severe drought already manifest in some areas such as the prairies, where recent climate change and heavy irrigation are depleting freshwater supplies (Schindler 2001).

Fig. 4. Canada 2000: water withdrawal by agriculture (white series), industry (grey series) and municipalities (black series); redrawn from Gleik (2001). 8 WATSON AND LAWRENCE

Drinking Water Safety: Source, Treatment and Management

Safe drinking water has been taken for granted in much of the developed world. However, safe water has been widely accessible only in the last hunderd years after the introduction of chemical disinfectants. Subsequently, the delivery of safe drinking water has become increasingly dependent on the widespread use of disinfection treatment, and the primary role of source water integrity has been largely overlooked (Watson et al. submitted for publication). This treatment-based approach has left consumers highly vulnerable to process failure, as recently demonstrated in Canada by outbreaks of serious waterborne disease linked to fecal waste (e.g., CRCWQT 2002; O’Connor 2002). Following these outbreaks, several major investigations identified serious deficiencies in the regulation, management and delivery of drinking water. These studies found major disparities in water treatment, supply and accessibility among different provinces and communities, and resulted in widespread assessment and revision of the water industry’s standards, compliance and accountability. Importantly, the reports also recognized the vital importance of source protection, and of a need to adopt a “multi-barrier approach” to drinking water safety, based on vigi- lance at all critical process levels between source and consumer (Fig. 5). Such an approach clearly begins with source protection, followed by a systematic development, installation and evaluation of treatment technologies for contaminant removal and monitoring. There is also a need to address the regional imbalance in drinking water treatment capacity and finished water quality. Much of this imbalance is related to regional differences in raw water quality, and technical and socio-

Fig. 5. The Canadian Multi-Barrier Approach to safe drinking water (redrawn from Craig and Shortreed 2002). INTRODUCTORY REMARKS 9 economic resources (Postel et al. 1996; Shiklomanov 1999). In Canada, First Nations and small communities in particular often have inadequate water (and sewage) treatment, few economic resources and a shortage of trained personnel. As a result, many of these communities also have a high frequency of boiled water advisories (BWAs) issued in response to irregulari- ties in disinfection, turbidity or coliform levels (Peterson et al. 2002). On the other hand, recent BWAs issued to cities such as St. John’s, Charlottown, Winnipeg, and Vancouver reflect the overall susceptibility of both large and small population bases to source water and treatment unpredictability.

Towards the Future: Drinking Water Research in Canada

A series of working groups recently identified 13 key threats to Canadian drinking water sources (Environment Canada 2001; Fig. 6). These threats are recognized at three levels: i) ultimate sources (e.g., agricultural, industrial and municipal disturbance), ii) proximal contaminants (e.g., nutrients, toxins, endocrine disruptors, pathogens, etc.), and iii) environmental modifiers (e.g., climate change). Not only are many of these factors closely linked, but they are inevitably coupled with watershed and aquatic ecosystem processes. Canada is involved in active research at all three of these levels. This country has long made significant contributions to research into terrestrial and aquatic hydrology and ecology (e.g., Vollenweider 1970; Schindler 1978). This remains an active field of research, which now has an increased emphasis on applied issues such as drinking water. At the same

Fig. 6. Threats to drinking water (from Environment Canada 2001). 10 WATSON AND LAWRENCE time, a large body of research is directed towards source-water contaminants and their treatment. An additional significant focus centres around socio-economic issues such as health risk assessment, risk management and risk communication. However, if Canadians are to be assured of a reliable supply of safe drinking water well into the future, the following concerns also need to be addressed: i) The scope, severity and potential impacts of emerging contaminant threats on human health. In particular, cost-effective techniques for their detection, viability determination and monitoring are unavail- able to most facilities (Ruecker et al. 2002; this issue: Hyland et al. 2003; Jamieson et al. 2003). The health effects of low-level exposure to such contaminants and to contaminant mixes are largely unknown. ii) A systematic means of evaluating and optimizing removal efficiency of contaminants by emerging treatment technologies (e.g., alternative coagulants, membranes, biological filtration, UV disinfection and process combinations). Currently, chlorination DBPs are studied vigorously; a similar investigation of the by-products and residuals of other treatment practices and process combinations is needed. iii) Water recycling and reuse. These are fundamental approaches to water shortages in need of development and evaluation, which must overcome the major problems associated with water re-use, notably dissolved salts, metals, nutrients and possibly other trace contaminants. iv) Innovative and integrated treatment approaches for small rural and Northern communities. There is a critical need for economical systems that are robust, and that can be customized to specific source waters (Corkal 2002; this issue: Coulibaly and Rodriguez 2003). Many of these communities are facing more stringent guidelines, yet are unable to meet existing levels of drinking water acceptability. v) A publicly accessible database of recommended practices, and access to appropriate expertise for small communities and individual stakeholders.

Summary and Conclusions

Drinking water is part of a global water cycle, which is a function of many abiotic and biotic processes. This cycle is increasingly impacted by human activities, many not necessarily directly related to water consumption and disposal. Projected changes in climate, and urban and industrial development threaten drinking water integrity and sustainability. Populations, consumption and source degradation are increasing as if there were no limits to the supply of clean water. The development of a long-term approach to sustainable water use requires national and international collaboration. At the same time, the causes and mitigation of source-water degradation need to be addressed. Canada faces many of the challenges seen across the world, but also has a unique advantage in resources, research and catchment basin diversity. These offer natural lab- INTRODUCTORY REMARKS 11 oratories in which to design, test and implement innovative approaches to a safe and sustainable supply.

Acknowledgement

We are very grateful for the tireless assistance of Kristin Alward in the compilation of this theme issue of the journal.

References

Bau J, Lobo Ferreira JP, Henriques J, de Oliveira Raposo J. 1991. Integrated approaches to water pollution problems. Elsevier Science Publishers, London. Blais JM, Schindler DW, Muir D, Sharp M, Donald D, Lafreniere M, Braekevelt E, Strachan W. 2001. Melting glaciers: a major source of persistent organochlo- rines to subalpine Bow Lake in Banff National Park, Canada. Ambio 30:410–415. Bro-Rasmussen F. 1996. Contamination by persistent chemicals in food chain and human health. Sci. Tot. Environ. 188:S45–S60. Brownlee B, Macinnis G, Norton L. 1993. Chlorinated anisoles and veratroles in a Canadian river receiving bleached kraft pulp mill effluent: identification, distribution, and olfactory evaluation. Environ. Sci. Technol. 27:2450–2455. Chen Z, Huang G, Chakma A. 1998. Integrated environmental risk assessment for petroleum-contaminated sites: a North American case study. Water Sci. Technol. 38:131–138. Corkal D. 2002. Rural water safety from the source to the on-farm tap. Drinking water safety: a total quality management approach. International conference on water and health. Ottawa, Sept. 2002. Coulibaly HD, Rodriguez MJ. 2003. Portrait of drinking water quality in small Quebec municipal utilities. Water Qual. Res. J. Canada 38(1):49–76. Craig L, Shortreed J. 2002. A total quality approach to drinking water. In International Conference on Water and Health. Ottawa. CRCWQT. 2002. First report of the Walkerton Inquiry. Health ctream: information analysis for water and health professionals. March 2002. Prahan, VIC. Australia. Crowe A. National Water Research Institute, Environment Canada, Burlington, Ontario, pers. comm. Darbi A, Viraraghavan T. 2003. A kinetic model for autotrophic denitrification using sulfur:limestone reactors. Water Qual. Res. J. Canada 38(1):183–192. Darbi A, Viraraghavan T, Jin Y-C, Braul L, Corkal D. 2003. Sulfate removal from water. Water Qual. Res. J. Canada 38(1):169–182. Day J, Quinn F. 1992. Water diversions and export: learning from the Canadian experience. Dept. Geography Publ. Series no. 36. University of Waterloo. 215 p. Environment Canada. 1990. Groundwater: nature’s hidden treasure. Fact Sheet 5. Ottawa, Ont. Environment Canada. 2001. Threats to sources of drinking water and aquatic ecosystem health in Canada. National Water Research Institute, Burlington, Ontario. NWRI Scientific Assessment Report Series No. 1. 72 p. Fox GA. 2001. Effects of endocrine disrupting chemicals on wildlife in Canada: past, present and future. Water Qual. Res. J. Canada 36:233–251. 12 WATSON AND LAWRENCE

Gleik PH. 1999. The world’s water 1998–1999: biennial report on freshwater resources. Island Press. 307 p. Gleik PH. 2001. The world’s water 2000–2001: biennial report on freshwater resources. Island Press. 315 p. Gorczyca B, London D. 2003. Characterization of particles in slow sand filtration at North Caribou water treatment plant. Water Qual. Res. J. Canada 38(1): 153–168. Hyland R, Byrne J, Selinger B, Graham T, Thomas J, Townshend I, Gannon V. 2003. Spatial and temporal distribution of fecal indicator bacteria in southern Alberta, Canada. Water Qual. Res. J. Canada 38(1):15–32. Jackson RB, Carpenter SR, Dahm CN, McKnight DM, Naiman RJ, Postel SL, Running SW. 2001. Water in a changing world. Ecol. Appl. 11:1027–1045. Jamieson RC, Gordon RJ, Tattrie SC, Stratton GW. 2003. Sources and persistence of fecal coliform bacteria in a rural watershed. Water Qual. Res. J. Canada 38(1):33–47. Krupa D, Czernas K. 2003. Mass appearance of the cyanobacterium Planktothrix rubescens in Lake Piaseczno, Poland. Water Qual. Res. J. Canada 38(1):141–152. Manna BR, Dey S, Debnath S, Ghosh UC. 2003. Removal of arsenic from groundwater using crystalline hydrous ferric oxide (CHFO). Water Qual. Res. J. Canada 38(1):193–210. Maule C, Fonstad T. 2000. Impacts of cattle penning on groundwater quality beneath feedlots. Can. Agric. Eng. 42:87–93. Mills E, Leach J, Carlton J, Secor C. 1993. Exotic species in the Great Lakes: a his- tory of biotic crises and anthropogenic introductions. J. Great Lakes Res. 19:1–54. Murphy T, Irvine K, Guo J, Davies J, Murkin H, Charlton M, Watson SB. 2003. New microcystin concerns in the lower Great Lakes. Water Qual. Res. J. Canada 38(1):127–140. O’Connor D. 2002. Part two. Report of the Walkerton Inquiry: a strategy for safe water. Queens Printer, Toronto, Ont. 499 p. Ongley E. 1997. Guidelines on water and sustainable development: principles and policy. UN Water Resources Series No. 77. 136 p. Paerl HW. 1997. Coastal eutrophication and harmful algal blooms: importance of atmospheric deposition and groundwater as “new” nitrogen and other nutrient sources. Limnol. Oceanogr. 42:1154–1165. Park CL, Huck PM. 2003. A conceptual model for Cryptosporidium transport in watersheds. Water Qual. Res. J. Canada 38(1):77–113. Peel M, Campbell W. 1999. Selection of clc, cba, and fcb chlorobenzoate-catabolic genotypes from groundwater and surface waters adjacent to the Hyde Park, Niagara Falls, chemical landfill. Appl. Environ. Microbiol. 65:1627–1635. Peterson H, Watson SB, Ruecker N, Braithwaite SL. 2002. Raw and treated water for rural areas in Saskatchewan. Presented at the 37th Central Canadian Symposium on Water Quality Research, Burlington, Ont. Feb. 2002. Postel SL, Daily GC, Ehrlich PR. 1996. Human appropriation of renewable fresh water. Science 271:785–788. Pringle CM. 2001. Hydrologic connectivity and the management of biological reserves: a global perspective. Ecol. Appl. 11:981–998. Ruecker N, Fout G, Peterson H, Watson SB, Lawrence J, Appleyard G, Christofi INTRODUCTORY REMARKS 13

N. 2002. The use of RT–PCR for the detection of enteric viruses in prairie surface drinking water supplies. Presented at the 37th Central Canadian Symposium on Water Quality Research, Burlington, Ont. Feb. 2002. Revenga C, Murray S, Abromavitz J, Hammond A. 1998. Watersheds of the world: ecological value and vulnerability. World Resources Institute and Worldwatch Institute, Washington D.C. Ritter L, Solomon K, Sibley P, Hall K, Keen P, Mattu G, Linton B. 2002. Sources, pathways, and relative risks of contaminants in surface water and groundwater: a perspective prepared for the Walkerton Inquiry. J. Toxicol. Environ. Health A 65:1–142. Schindler DW. 1978. Factors regulating phytoplankton production and standing crop in the world’s freshwaters. Limnol. Oceanogr. 23:478–486. Schindler DW. 2000. Aquatic problems caused by human activities in Banff National Park, Alberta, Canada. Ambio 29:401–407. Schindler DW. 2001. The cumulative effects of climate warming and other human stresses on Canadian freshwaters in the new millennium. Can. J. Fish. Aquat. Sci. 58:18–29. Shiklomanov IA. 1999. World water resources: modern assessment and outlook for the 21st Century (Summary of World Water Resources at the Beginning of the 21st Century, prepared in the framework of the IHP UNESCO). Federal Service of Russia for Hydrometeorology & Environment Monitoring, State Hydrological Institute, St. Petersburg. Somlyódy L, Yates D, Varis O. 2001. Challenges to freshwater management. Ecohydrol. Hydrobiol. 1:65–95. Statistics Canada. 2001. CANSIM II, Table 052–0001. UNESCO. 2000. Water use in the world: present situation/future needs. (www.unesco.org/science/waterday2000/water_use_in_the_world.htm). Vollenweider RA. 1970. Scientific fundamentals of the eutrophication of lakes and flowing waters, with particular reference to nitrogen and phosphorus as factors in eutrophication. O.E.C.D., Paris, France. Vörösmarty CJ, Green P, Salisbury J, Lammers RB. 2000. Global water resources: vulnerability from climate change and population growth. Science 289: 284–288. Wallace J. 2000. Increasing agricultural water use efficiency to meet future food production. Agric. Ecosyst. Environ. 82:105–119. Watson SB. Cyanobacterial and eukaryotic algal odour compounds: signals or side-products? A review of their biological activity. Phycologia, In press. Watson SB, Satchwill T, Wiltse K. Submitted for publication. Wolf AT, Natharius JA, Danielson JJ, Ward BS, Pender JK. 1999. International river basins of the world. Water resources development. Vol. 15, No. 4, p. 387–427. WHO. 1999. Bottled drinking water. Fact Sheet No. 256. Geneva. WHO/UNICEF. 2000. Global water supply and sanitation assessment 2000 report. UNICEF, New York. Zaitlin B, Watson SB, Dixon J, Steel D. 2003. Actinomycetes in the Elbow River Basin, Alberta, Canada. Water Qual. Res. J. Canada 38(1):115–125. Water Qual. Res. J. Canada, 2003 Volume 38, No. 1, 15–32 Copyright © 2003, CAWQ

Spatial and Temporal Distribution of Fecal Indicator Bacteria within the Oldman River Basin of Southern Alberta, Canada

ROMNEY HYLAND,1 JAMES BYRNE,2* BRENT SELINGER,3 THOMAS GRAHAM,1 JAMES THOMAS,3 IVAN TOWNSHEND2 AND VICTOR GANNON1

1Health Canada, Lethbridge, Alberta, Canada 2Department of Geography and 3Department of Biological Sciences, University of Lethbridge, Alberta, Canada

Fecal coliform (FC) and Escherichia coli (EC) concentrations in the Oldman River and its tributaries, and in irrigation canals in southern Alberta, Canada, were monitored during 1998, 1999 and 2000. High FC and EC counts were found in drainages from agricultural lands in all years and in the Oldman River downstream of the City of Lethbridge wastewater treatment plant (WTP) during 1998. A significant decrease in the FC and EC concentrations downstream from the Lethbridge WTP was observed in 1999, after an upgrade to the WTP, which included a UV disinfection system. Spikes in FC and EC concentrations were observed in surface waters following heavy rainfall events. It is possible that a decrease in precipitation was responsible for decreases in the FC and EC concentrations observed in the Oldman River over the three years of the study. The increase in FC and EC counts in the tributaries and irrigation canals during this same period presumably reflects greater waste inputs from agricultural lands.

Key words: fecal contamination, fecal coliforms, Escherichia coli, river systems, southern Alberta

Introduction

It has been estimated that waterborne diseases kill more than 5 million people annually (Hunter et al. 2002). The microbial pathogens responsible for most of these deaths originate from human and animal feces. These wastes contaminate water through a variety of means including direct discharge, surface runoff and seepage through groundwater. Fortunately, simple water treatment procedures such as chlorination inactivate the majority of these pathogens. Therefore, most cases of waterborne disease result from the consumption of inadequately treated water. However, it must also be noted that certain of these pathogens, such as the protozoan parasites Cryptosporidium and Giardia, are very resistant to chlorine (Carpenter et al. 1999; Korich et al. 1990). In Canada, there have been a number of recent disease outbreaks associated with contaminated drinking water (Glover et al. 1992; Johnstone et

* Corresponding author; [email protected] 16 HYLAND ET AL. al. 1991; Langille et al. 1992; Stirling et al. 2001). In Walkerton, Ontario, in May of 2000, an estimated 2300 individuals were affected and 6 deaths occurred following the accidental consumption of untreated drinking water which contained Escherichia coli O157:H7 and Campylobacter jejuni (Bruce- Grey-Owen Sound Health Unit 2000; Health Canada 2000). In this outbreak, surface water flow from a nearby cattle farm is thought to have contaminated a well which contributed to the town water supply. This and other incidents have raised concerns about the risks to human health of living in communities with intensive animal agriculture. In Alberta, the southern portion of the province is reported to have the highest rates of gastrointestinal illness (Waters et al. 1994). This region is characterized by very low annual precipitation rates, an extensive crop irrigation system and a bourgeoning animal agriculture industry. Much of the field crop production is used as feed for animal agriculture and there are numerous large cattle feedlot operations as well as facilities for raising other domestic animal and poultry species. Manure waste from these intensive livestock operations is spread on neighbouring agricultural lands. In addition, there are several towns and villages and the City of Lethbridge that discharge treated wastewater into the Oldman River. While detailed epidemiological studies have not been conducted, there is a suspicion that waterborne transmission of pathogens to humans and among animal species may be responsible for the increased incidence of human enteric disease in the region. Fecal coliforms (FC) and E. coli (EC) are widely used as indicators of fecal contamination in water (Hagedorn et al. 1999; Harwood et al. 2000; Hunter et al. 2000; Kaspar et al. 1990; Parveen et al. 1999) and have been shown to be positively correlated with the presence of pathogenic enteric bacteria (Hunter et al. 2000; Rice and Johnson 2000). This investigation considers the spatio-temporal distribution of fecal coliforms and E. coli in the Oldman River basin of southern Alberta, Canada, over a three-year period, from April of 1998 to December of 2000.

Materials and Methods

Study Area The Oldman River is a tributary of the South Saskatchewan River, which passes through the southern portion of the province of Alberta (Fig. 1). The Oldman River flows eastward from headwaters in the Rocky Mountains through foothills and prairie. The watershed is 22,641 km2 and has a population of approximately 200,000 living in rural farms and villages and in the City of Lethbridge (ORBWQI 2000). The basin contains two major irrigation districts; the Lethbridge Northern Irrigation District (LNID), lying north of the Oldman River main trunk, and the Saint Mary River Irrigation Project, lying south of the main trunk. The LNID contains the greatest density of livestock operations in the basin, with up to 426 head of cattle/km2, the highest livestock FECAL BACTERIA IN SOUTHERN ALBERTA RIVERS 17

Fig. 1. Study area in the Oldman River basin, southern Alberta, Canada.

density in southern Alberta (Fent, Unpublished data). The geographic information system (GIS) ArcView 3.2 was used to divide the Oldman River basin into 27 sub-basins; two of the largest sub-basins contained sampling sites used in the study: the upper and the lower Oldman sub- basins (Fig. 2). The upper Oldman sub-basin is located upstream of the City of Lethbridge and the lower sub-basin is downstream of the city. The lower sub-basin has 140 cattle/km2, whereas the upper Oldman has 57 cattle/km2.

Water Sample Collection and Site Locations Water samples were taken upstream from foothold using a pole, which reached approximately 30 cm below the surface of the water; samples were collected in sterile 250-mL plastic bottles, placed on ice and transported to the laboratory to be tested before the end of the next day. Water samples were collected regularly (weekly, bi-weekly, tri-weekly, or monthly) by staff of Alberta Environment at 30 sites in the Oldman River basin, fourteen of which were on the Oldman River trunk (Fig. 2). The remaining sixteen sites were either on tributaries of the Oldman River (sites 1, 3, 4, 7, 8, 10, 11, 14, and 22) or irrigation drainage canals (sites 16, 19, 21, 23, 24, 26, and 28). Twenty-two of the sites lie within the upper and lower Oldman River sub-basins. Each sub-basin has eleven sampling sites, and eight sites (1, 2, 3, 4, 8, 11, 14, and 23) are not within either of these sub-basins (Fig. 2). Sites 7 (Pincher Creek), 8 (Beaver Creek), 10 (Willow Creek), 16 (Six Mile Coulee), 19 (Piyami Drain), 21 (Battersea 18 HYLAND ET AL.

Fig. 2. Water quality sampling sites within the Oldman River basin during 1998, 1999 and 2000. Site numbers and names are listed below.

1) Crowsnest River above Coleman 16) Six Mile Coulee Spillway 2) Oldman River near Olin Creek 17) Oldman River at Hwy 3 Bridge 3) Crowsnest River u/s of Connelly 18) Oldman River s/w of Diamond City 4) Castle River at Recreation Area 19) Piyami Drain 5) Oldman River d/s of Bottom Release 20) Oldman River d/s of Picture Butte 6) Oldman River near Brocket 21) Battersea Drain 7) Pincher Creek at Hwy 3 22) Little Bow River 8) Beaver Creek at Hwy 785 23) Bountiful Coulee 9) Oldman River at Ft. MacLeod 24) Drain T-2 10) Willow Creek at Hwy 811 25) Oldman River above Taber 11) Belly River u/s of confluence 26) Expanse Coulee with OMR 27) Oldman River at Hwy 36 12) Oldman River near Monarch 28) Drain T11 13) Oldman River d/s of Belly River 29) Oldman River near Purple 14) St. Mary River Springs 15) Oldman River at Popson Park 30) Oldman River at the mouth

Drain), and 22 (Little Bow River) carry water to the Oldman River from tilled land; sites 7, 8, and 10 from pasture land; site 16 from land with poultry operations; and sites 19, 21, and 22 from land with cattle feedlot operations. Sites 15 (Oldman River at Popson Park), 16 (Six Mile Coulee), 17 (Oldman River at the Hwy 3 bridge) and 18 (Oldman River at Diamond FECAL BACTERIA IN SOUTHERN ALBERTA RIVERS 19

City) occur within a 23-km stretch of the Oldman River within the City of Lethbridge (Fig. 2). Site 16 is located in a small stream that enters the Oldman River between sites 15 and 17. A LNID return canal flows into the river directly upstream of both site 16 and of site 17; two LNID canals also flow into the river upstream of site 18. Site 18 is also located immediately downstream of the City of Lethbridge wastewater treatment plant (WTP) and site 17 is immediately upstream of the WTP.

Water Sample Analysis FC and EC were enumerated by a membrane filtration technique at the Provincial Laboratory of Public Health, Foothills Hospital, Calgary, Alberta. Counts represent the total number of bacteria (as colony-forming units, CFU) per 100 mL.

Spatial and Temporal Distribution of FC and EC FC and EC counts above 200 CFU/100 mL, the Canadian Recreational Water Quality Guideline (CRWQG) (Health Canada 1992), were mapped in ArcView GIS (version 3.2 ESRI, Redlands, Calif.) in Ten Transverse Mercator (TTM) projection. Land variables, such as livestock distribution, human populations, WTPs and irrigation canal outflow points were imported into ArcView as themes and mapped along with the FC and EC data. Spatial data for southern Alberta was obtained from L. Fent (Government of Alberta, Sustainable Resource Development, Edmonton, Alberta). Rainfall data from the Connelly Creek weather station was obtained from L. LaFlamme (Alberta Environment Climate Services, Edmonton, Alberta).

Results

Major Trends in the Study Area Twenty-four of the thirty sites (80%) had at least one sample date during the three years of study where FC levels were above 200 CFU/100 mL (Fig. 3, 4). Sites 1 to 5 (the sites furthest upstream and closest to the Rocky Mountains) and 9 (Oldman River at Fort MacLeod) did not have any samples with FC concentrations over 200 CFU/100 mL (Fig. 2, 3). Upper Oldman River sub-basin sites 1 to 6, site 9 (Oldman River at Fort MacLeod), site 13 (Oldman River downstream of the Belly River), and site 17 (Oldman River at Hwy 3 Bridge, Lethbridge) also did not have any EC levels above 200 CFU/100 mL from 1998 to 2000 (Fig. 3). Seven sites (7, 8, 18, 19, 20, 21, and 22) had FC concentrations above 200 CFU/100 mL in 25% or more of the samples taken over the three-year study period (Fig. 3). All but two (site 18, downstream from the City of Lethbridge WTP and site 20, downstream from Picture Butte) are from waters entering the Oldman River from agricultural lands. Sites 18 (downstream of Lethbridge WTP) and 19 (Piyami Drain) had the highest 20 HYLAND ET AL.

Fig. 3. Percentage of samples with fecal coliform and E. coli concentrations above 200 CFU/100 mL at all sites in the study area. number of water samples with FC concentrations over 200 CFU/100 mL, 50% (18/36) and 56% (15/27), respectively (Fig. 3). Sites 18 and 19 also had the highest number of water samples with EC concentrations over 200 CFU/100 mL (Fig. 3). Twenty-five percent of samples (11/44 samples) at site 18 and 48% of samples (13/27 samples) at site 19 were over 200 CFU/100 mL for EC. In 1998 and 1999, the largest number of study sites with FC concentrations over 200 CFU/100 mL occurred in June, with 60% in 1998 and 41% in 1999 (Fig. 5). In 2000, June, July and August had a relatively even distribution of sites with samples above the 200 CFU/100 mL limit (45% in June of 2000, 44% in July 2000 and 40% in August 2000; Fig. 5). Overall, in 1998, 24% of all basin samples had FC concentrations over 200 CFU/100 mL. In contrast, 1999 and 2000 had only 10% and 11%, respectively, of all samples above this level (Table 1).

Agricultural Lands All sixteen of the sites in waters entering the Oldman River from agricultural lands show FC and EC peaks in June, July and August. This is best illustrated by three agriculture sites: 7 (Pincher Creek at Hwy 3); 19 (Piyami Drain); and 22 (Little Bow River) (Fig. 6). Site 7 (Pincher Creek at Hwy 3), peaks occurred on August 4 in 1998 (380 CFU/100 mL) and on August 3 in 1999 (650 CFU/100 mL); no data are available for 2000 at this site. Site 19 summer peaks occurred on July 8 in 1998 (610 CFU/100 mL); July 20 in 1999 (870 CFU/100 mL); and on August 2 in 2000 (1100 CFU/100 mL). Summer peaks occurred on July 7, 1998 FECAL BACTERIA IN SOUTHERN ALBERTA RIVERS 21

Fig. 4. Water sampling sites with FC concentrations over 200 CFU/100 mL in: a) 1998; b) 1999 and c) 2000. Cattle density in the study area is shown by green gradations, from 0–10 cattle/km2 (lightest shade) to 142–426 cattle/km2 (darkest shade). 22 HYLAND ET AL.

Fig. 5. Temporal distribution of fecal coliform and E. coli counts in: a) 1998, b) 1999 and c) 2000. Counts are presented as percent of bacterial counts above 200 CFU/100 mL per total number of water samples taken during the month. FECAL BACTERIA IN SOUTHERN ALBERTA RIVERS 23

Table 1. Percentage of water samples with FC counts above 200 CFU/100 mL and percentage of sites with FC counts over 1000 CFU/100 mL in the study region (all sites), in the lower Oldman River sub-basin (LB) and in the upper Oldman River sub-basin (UB)a

% Samples over 200 % Sites over 1000 CFU/100 mL CFU/100 mL

Year all sites LB UB all sites LB UB

1998 24 35 23 23 40 18 1999 10 18 6 23 40 9 2000 11 16 7 13 27 9 1998–2000 15 24 12 40 88 18

a Values are given as: (# samples >200 CFU/100 mL)/(total # samples) * 100%; or as (# sites with concentrations of 1000 CFU/100 mL)/(total # sites) * 100%.

(1100 CFU/100 mL); August 12, 1999 (460 CFU/100 mL) and June 19, 2000 (210 CFU/100 mL) at site 22. The lower Oldman sub-basin is located in an area of higher cattle density than is the upper Oldman (Fig. 4); as noted, 57 head of cattle/km2 were located in the upper Oldman sub-basin, while 140 head/km2 were located in the lower Oldman sub-basin. During the three-year sampling period, all ten of the sampling sites (100%) within the lower Oldman River sub-basin had FC concentrations above 200 CFU/100 mL at least once, and 8 (sites 19, 20, 21, 22, 24, 25, 26 and 28) of the 10 sites (80%) had FC concentrations at, or over, 1000 CFU/100 mL one or more times (Table 1). Twenty-four percent of all samples taken in the lower Oldman sub-basin were above 200 CFU/100 mL; 35% in 1998, 18% in 1999 and 16% in 2000 (Table 1). In contrast, only nine of the eleven sites in the upper Oldman sub- basin were over 200 CFU/100 mL one or more times over the three-year period. Only two sites (15 and 16) of the eleven sites (18%) in the upper Oldman River sub-basin had FC concentrations at or over 1000 CFU/100 mL one or more times during the three years and 12% of the total samples taken in the upper Oldman sub-basin were above 200 CFU/100 mL; 23% in 1998, 6% in 1999 and 7% in 2000 (Table 1).

Precipitation The study area experienced more precipitation during 1998 than in 1999 or 2000 (Table 2, Fig. 7). May was the peak month of precipitation in 1998 and 1999. In May of 1998, 16.5 mm of rain fell, whereas in 1999, only 7.0 mm fell, less than half of that seen in 1998. In 2000, the total precipita- 24 HYLAND ET AL.

Fig. 6. Temporal diagrams of FC and EC concentrations, expressed as number of bacteria, from April 16, 1998, to October 3, 2000. All sites received runoff predominantly from agricultural lands: a) site 7, Pincher Creek; b) site 19, Piyami Drain; and c) site 22, Little Bow River. FECAL BACTERIA IN SOUTHERN ALBERTA RIVERS 25

Table 2. Precipitation and fecal coliforms at sites in the Oldman River basina

Year Irrigation drains Tributaries Oldman River Total rainfall

1998 36% 32% 34% 38 mm 1999 25% 23% 8% 21 mm 2000 46% 35% 0% 13 mm

a Values are given as percentage of samples between May and September over 200 CFU/100 mL at sites located in irrigation canals; on tributaries to the Oldman River; and in the Oldman River. Total rainfall is given as mm of rain fall- en between May and September of 1998, 1999 and 2000. tion for May was only 2.3 mm, and for June 5.3 mm. The percentage of samples with FC over 200 CFU/100 mL increased with increased precipitation only in sample sites located on the Oldman River (Table 2). Thirty- four percent of samples were over 200 CFU/100 mL between May and September 1998 (a high precipitation year), 8% in 1999 and 0% in 2000 (low precipitation years; Table 2). In contrast, sample sites located in irrigation drains and in tributaries of the Oldman River did not display a clear relationship between FC concentration and precipitation, as high percentages of samples over 200 CFU/100 mL were found in 2000, despite a low total rainfall in this year. High levels of precipitation were recorded between June 12 and 16, 1998 (14 mm) and May 30 and June 2, 1999 (15 mm) (Fig. 8). Six of fourteen sites (sites 16, 19, 20, 21, 22, and 26) sampled on June 16, 1998, had a spike in FC concentrations, with an overall average FC concentration of 1000 CFU/100 mL across the fourteen sites (Fig. 8). Seven of thirteen sites (sites 16, 25, 26, 27, 28, 29 and 30) sampled on June 2, 1999, had a spike in FC concentrations, with an average FC concentration of 2000 CFU/100 mL across the thirteen sites (Fig. 8). All sites with spikes in FC concentrations after rainfall events were in the Oldman River and in irrigation canals and tributaries located in the lower Oldman sub-basin.

The City of Lethbridge The City of Lethbridge lies on the Oldman River between sites 15 and 18 (Fig. 3). Site 15 is directly upstream of the city, site 16 is on a small stream entering the Oldman River within the city, site 17 is directly upstream of the City of Lethbridge WTP and site 18 is directly downstream of the WTP. In January of 1999, the City of Lethbridge finished upgrades to its WTP, including a UV irradiation system. Water entering the city at site 15 (Fig. 9a) had a low occurrence (10.8%, or 4 of 37 samples) of FC concentrations above 200 CFU/100 mL. Site 17 (Fig. 9c) also had a low number of samples with FC concentrations over 200 CFU/100 mL 26 HYLAND ET AL.

Fig. 7. Connelly Creek total monthly precipitation, expressed in mm, over the sampling period: a) 1998, b) 1999 and c) 2000. FECAL BACTERIA IN SOUTHERN ALBERTA RIVERS 27

Fig. 8. Spike fecal coliform and E. coli counts associated with rain events in: a) 1998 and b) 1999. Bacterial counts are expressed as average counts over the sample period; precipitation represents an average, one and four days prior to the rain event.

(2.8%, or 1 of 36), despite a high concentration of these bacteria at site 16 (Six Mile Coulee, a small stream entering the river between sites 15 and 17). At site 16 (Fig. 9b), FC levels were above 200 CFU/100 mL in 22% of all samples (12 out of 55 samples), and EC levels were above 200 CFU/100 mL in 11% of all samples (6 out of 55 samples). Not surprisingly, site 17 FC levels were also lower than those at site 18 (Fig. 9d) (2.8% versus 50% 28 HYLAND ET AL. the concentrations at sites located within the City of Lethbridge: a) Oldman River E. coli mporal diagrams of fecal coliform and Te Popson Park; b) Six Mile Coulee; and c) Oldman River upstream of the wastewater treatment plant; d) Oldman River downstream of plant; d) Oldman River downstream of the wastewater treatment Popson Park; b) Six Mile Coulee; and c) Oldman River upstream plant. wastewater treatment Fig. 9. FECAL BACTERIA IN SOUTHERN ALBERTA RIVERS 29 of all samples were above 200 CFU/100 mL). Fourteen of the eighteen (78%) samples at site 18 with FC concentrations above 200 CFU/100 mL occurred in 1998 before the WTP upgrade; and four of the eighteen (22%) occurred in 1999 after the WTP upgrade. No data are available for 2000.

Discussion

While most sites had both FC and EC concentrations above the CRWQG limit for recreational water (200 CFU/100 mL) at some time during the three-year study, fewer sites in the upper Oldman River sub-basin had high indicator bacteria counts than sites in the lower Oldman River sub-basin. Waters with the lowest FC and EC counts were nearest the river source waters in the Rocky Mountains, and those with the highest were downstream of areas with substantial agricultural activity. Doran and Linn (1979) found FC counts 5 to 10 times higher in runoff from grazed lands than in ungrazed lands. High FC levels at agricultural sites may be due to a number of factors such as domestic or wild animals defecating in or near the surface waters, runoff from manure applied to lands adjacent to surface waters and liquid effluent entering these waters from animal rearing operations entering these waters (Crowther et al. 2002; Tian et al. 2002). Unfortunately, it was not possible to differentiate among, and quantify these sources of contamination in the present study. In general, the highest FC and EC counts were from sites located near agricultural land during the summer months (compare Fig. 6 and 9), particularly after rainfall events (Fig. 8). Others have also noted high levels of coliform bacteria in water during the summer (Geldreich 1996; Hunter and McDonald 1991; Tian et al. 2002). This may be attributed to a number of factors, such as increased livestock distribution and numbers, and increased agricultural activity, including manure application to crops (Hunter et al. 2002). Spikes in FC counts following rain storms have also been observed by other investigators. McDonald et al. (1982), reported that coliform bacteria concentrations increased 10- to 30-fold following artificial release in water catchments; and Kistemann et al. (2002) found a 20- to 100-fold increase in fecal indicator bacteria concentrations transported to reservoirs during heavy rainfall. The high levels of fecal indicator bacteria found in the river water in our study was most likely due to rainfall runoff from agricultural lands and churning of the river waters, with resuspension of bacteria settled on the river bottom (McDonald et al. 1982). Total precipitation was higher in 1998 than it was in 1999 or in 2000 (Fig. 7a, b, c; Table 2). As would be expected, during the latter two years there was also a decrease in FC for sites on the Oldman River. Interestingly, no corresponding drops in FC counts with annual precipitation were observed in irrigation canals or tributaries of the Oldman River during this time (Table 2). It is possible that irrigation canals are not as affected by rainfall and runoff as is the natural water system. A possible explanation could be that in dry periods there would be an increase in irrigation, during rain a decrease in irrigation. The net effect may be to keep surface runoff into 30 HYLAND ET AL. irrigation canals relatively constant in wet and dry years. Increased manure application or other inputs could also be responsible for increasing indicator bacteria counts in irrigation canals in certain years. In dry years, it is also possible that cattle would be given more access and would frequent tributary water more often on pasture, causing an increase in FC concentrations from direct introduction of feces into the water. Similar circumstances are thought to have been responsible for a waterborne outbreak of E. coli O157:H7 in Swaziland in 1992 (Effler et al. 2002). In this case, a drought in the region is thought to have been responsible for cattle frequenting and defecating in streambeds. An outbreak of disease associated with this enteric pathogen occurred after a sudden torrential rain carried bovine feces containing this pathogen into the water. As stated above, fecal indicator bacteria populations normally peak in the spring and summer months if the source of fecal pollution is livestock or agriculture runoff (Geldreich 1996; Hunter and McDonald 1991). At site 18 (downstream of the City of Lethbridge WTP), a continuous high level of fecal indicator bacteria was found throughout 1998, with a moderate peak in the summer months. This suggests that a constant source of contamination, such as human waste, was responsible for the majority of this contamination (Brezonik and Stadelmann 2002). In January of 1999, the City of Lethbridge completed an upgrade to their WTP. This upgrade consisted of discharge pipe repairs and system updates, including the addition of a UV disinfection system (City of Lethbridge 1999). The WTP upgrade appears to have substantially reduced concentrations of these indicator organisms.

Acknowledgements

This study was funded by Health Canada and the University of Lethbridge in association with the Oldman River Basin Water Quality Initiative which provided both water sampling and analysis support. Water samples for this study were collected by staff of Alberta Environment (Water Management Division). Members of the research team are participating members of the Oldman River Basin Water Quality Initiative and the Canadian Water Network, National Centre of Excellence.

References

Brezonick PL, Stadelmann TH. 2002. Analysis and predictive models of stormwater runoff volumes, loads and pollutant concentrations from watersheds in the Twin Cities metropolitan area, Minnesota, USA. Water Res. 36:1743–1757. Bruce-Grey-Owen Sound Health Unit. 2002. Waterborne outbreak of gastroenteritis associated with a contaminated municipal water supply, Walkerton, Ontario May-June 2000. Can. Commun. Dis. Rep. 26:20. Carpenter C, Fayer R, Trout J, Beach MJ. 1999. Chlorine disinfection of recreational water for Cryptosporidium parvum. Emerg. Infect. Dis. 5:579–584. FECAL BACTERIA IN SOUTHERN ALBERTA RIVERS 31

City of Lethbridge. 1999. Council Minutes of April 19th, 1999. Clerk’s Office, Lethbridge Alberta. Crowther J, Kay D, Wyer MD. 2002. Fecal-indicator concentrations in waters draining lowland pastoral catchments in the UK: relationships with land use and farming practices. Water Res. 36:1725–1734. Doran JW, Linn DM. 1979. Bacteriological quality of runoff water from pasture- land. App. Environ. Microbiol. 37:985–991. Effler P, Isaacson M, Arntzen L, Heenan R, Canter P, Barrett T, Lee L, Mambo C, Levine W, Zaidi A, Griffin PM. 2002. Factors contributing to the emergence of Escherichia coli O157 in Africa. Emerg. Infect. Dis. 7:812–819. Fent L. Government of Alberta, Sustainable Resource Development, Edmonton, Alberta. Unpublished data. Geldreich EE. 1996. Pathogenic agents in freshwater resources. Hydrol. Proc. 10:315–333. Glover D, Ross R, Lugsdin J. 1992. Gastroenteritis outbreak at an industrial camp- British Columbia. Can. Commun. Dis. Rep. 18:66–68. Hagedorn C, Robinson SL, Filtz JR, Grubbs SM, Angier TA, Reneau RB. 1999. Determining sources of fecal pollution in a rural Virginia watershed with antibiotic resistance patterns in fecal streptococci. Appl. Environ. Microbiol. 65:5522–5531. Harwood VJ, Whitlock J, Withington V. 2000. Classification of antibiotic resistance patterns of indicator bacteria by discriminant analysis: use in predicting the source of fecal contamination in subtropical waters. Appl. Environ. Microbiol. 66:3698–3704. Health Canada. 1992. Guidelines for Canadian recreational water quality. (http://www.hc-sc.gc.ca/ehp/ehd/catalogue/bch_pubs/recreational_ water.htm). Health Canada. 2000. Center for Infectious Disease Prevention and Control Division of Enteric, Foodborne and Waterborne Diseases: waterborne outbreak in Walkerton. May 2000. Conclusions. Health Canada, Ottawa. Hunter C, Perkins J, Tranter J, Hardwick P. 2000. Fecal bacteria in the waters of an upland area in Derbyshire, England: the influence of agricultural land use. J. Environ. Qual. 29:1253–1261. Hunter C, MacDonald A. 1991. Seasonal changes in the sanitary bacterial quality of water draining a small upland catchments in the Yorkshire Dales. Water Res. 25:447–453. Hunter PR, Colford JM, Lechevallier MW, Binder S, Berger PS. 2002. Waterborne diseases. Emerg. Infect. Dis. 7S:544–545. Johnstone T, Marchenski M, Hines W, Davis J, Eng S, Gibson G, Peck SH. 1991. Outbreak of gastroenteritis associated with drinking well water–British Columbia. Can. Commun. Dis. Rep. 17:111–112. Kaspar CW, Burgess JL, Knight IT, Colwell RR. 1990. Antibiotic resistance index- ing of Escherichia coli to identify sources of fecal contamination in water. Can. J. Microbiol. 36:891–894. 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. Envrion. Microbiol. 36:2188–2197. Korich DG, Mead HR, Madore MS, Sinclair NA, Sterling CR. 1990. Effects of ozone, chlorine dioxide, chlorine, and monochloramine on Cryptosporidium parvum oocyst viability. Appl. Environ. Microbiol. 56:1423–1428. 32 HYLAND ET AL.

Langille DB, Leblanc JC, Durkee MK. 1992. Gastrointestinal illness associated with drinking well water delivered by tank truck during a period of drought—Nova Scotia. Can. Commun. Dis. Rep. 18:108–112. McDonald A, Kay D, Jenkins A. 1982. Generation of fecal and total coliform surges by stream flow manipulation in the absence of normal hydrometeo- rological stimuli. Appl. Environ. Microbiol. 44:292–300. ORBWQI. 2000. Oldman river basin water quality initiative summary report 2000. Oldman River Basin Water Quality Initiative, Lethbridge, Alberta. Parveen S, Portier KM, Robinson K, Edmiston L, Tamplin ML. 1999. Discriminant analysis of ribotype profiles of Escherichia coli for differentiat- ing human and nonhuman sources of fecal pollution. Appl. Environ. Microbiol. 65:3142–3147. Rice EW, Johnson CH. 2000. Survival of E. coli 0157:H7 in dairy cattle drinking water. J. Dairy Sci. 83:2021–2023. Stirling R, Aramini J, Ellis A, Lim G, Meyers R, Fleury M, Werker D. 2001. Waterborne Cryptosporidium outbreak, North Battleford, Saskatchewan, spring 2001. Can. Commun. Dis. Rep. 27:185–192. Tian YQ, Gong P, Radke JD, Scarborough J. 2002. Spatial and temporal modeling of microbial contaminants on grazing farmlands. J. Environ. Qual. 31:860–869. Waters JR, Sharp JC, Dev VJ. 1994. Infection caused by Escherichia coli O157:H7 in Alberta, Canada, and in Scotland: a five-year review, 1987–1991. Clin. Infect. Dis. 19:834–843. Water Qual. Res. J. Canada, 2003 Volume 38, No. 1, 33–47 Copyright © 2003, CAWQ

Sources and Persistence of Fecal Coliform Bacteria in a Rural Watershed

ROB C. JAMIESON,1* ROBERT J. GORDON,2 STEVEN C. TATTRIE4 AND GLENN W. STRATTON3

1School of Engineering, University of Guelph, Guelph, Ontario, Canada N1G 2W1 2Department of Engineering and 3Department of Environmental Sciences, Nova Scotia Agricultural College, P.O. Box 550, Truro, Nova Scotia, Canada B2N 5E3 4Resource Stewardship Division, Nova Scotia Department of Agriculture and Fisheries, Truro, Nova Scotia, Canada B2N 5E3

Water quality within the Thomas Brook watershed, which is a small catchment located in the headwaters of the Cornwallis River drainage basin, was assessed through an integrated monitoring program. The Thomas Brook watershed is approximately 1000 ha and is characterized by both agricultural and residential land uses. Fecal coliform concentrations and stream flow were monitored at several points throughout the watershed during an eight-month period (May to December, 2001). Thomas Brook was seriously degraded with respect to microbial water quality. Fecal coliform levels frequently exceeded recreational water quality guidelines. At the watershed outlet, 94% of the collected samples exceeded the recreational water quality guideline during low flow conditions. Substantial bacterial loading was observed along stream reaches impacted by livestock operations. Bacterial loading was also observed along a stream reach that was not impacted by agricultural activities. A dense clustering of residences, using on-site septic systems, was the suspected source. Results from this study indicate the presence of a reservoir of fecal microorganisms within the stream sediments. The release of fecal microorganisms from the stream sediments to the water column during both low and high flow conditions could be a major source of bacterial loading.

Key words: fecal coliform, sources, watershed, water quality

Introduction

The entry of pathogenic microorganisms into drinking, irrigation and recreational water sources poses a risk to human health. Drinking water boil orders and beach closures, due to elevated levels of fecal pollution indicators, have become a common occurrence in many regions of North America. Livestock agriculture is often cited as one of the major sources of pathogenic microorganisms in surface and ground water systems. As a result, there has been an increasing need for the development of farm best management practices (BMPs) which minimize the risk of

* Corresponding author; [email protected] 34 JAMIESON ET AL. water contamination, especially with respect to human pathogens. The efficacy of these BMPs must be thoroughly evaluated in terms of improv- ing water quality at the watershed scale. Pathogens associated with agricultural waste that are of concern include virulent strains of Escherichia coli, Salmonella, Giardia, Campylobacter, Shigella and Cryptosporidium parvum (Landry and Wolfe 1999). Difficulties and expenses involved in the testing for specific pathogens have generally led to the use of indicator organisms of enteric origin to estimate the persistence and fate of enteric bacteria in the environment (Crane et al. 1981). Fecal coliforms (FC) are the most commonly used indicator organisms. Fecal coliforms are identified by their ability to produce gas from lactose at 44.5°C (APHA 1998). Escherichia coli is the most common FC and although most E. coli strains are non-pathogenic, some strains, such as E. coli O157:H7, pose a serious health risk to humans. Enterohemorrhagic E. coli O157:H7 was first identified as a human pathogen in 1982 (Kudva et al. 1998) and has been found in asymptomatic cattle, sheep, swine, deer, dogs, horses and fowl (Mubiru et al. 2000). To begin assessing the issue of microbial pollution we must first identify the various pollution sources. In a rural mixed land use watershed, sources could consist of both point (sewage treatment plants, domestic septic systems, manure storages and animal defecation directly into streams) and non-point (sediments, manure and sludge application) sources. Secondly, transport mechanisms must be identified. Possible mechanisms by which bacteria can enter a surface water system include: surface runoff, direct discharges, groundwater flow, and tile drainage systems. Numerous studies have revealed the presence of fecal indicator organisms and pathogens in both farmed and non-agricultural watersheds. Doran and Linn (1979) monitored surface runoff from grazed and ungrazed pasture land over a three-year period in Nebraska. Fecal coliform counts were 5 to10 times greater in grazed areas, however total coliform (TC) counts differed little. Fecal coliform counts from ungrazed pas- tures also commonly exceeded water quality standards. Patni et al. (1985) observed that FC bacteria were almost always present in runoff from non- manured cropland, presumably due to non-livestock sources. Fecal indicators were common in the surface waters of agricultural areas in southern Finland (Niemi and Niemi 1991), generally exceeding 100 colony forming units (cfu) per 100 mL and occasionally exceeding 1000 cfu/100 mL. Fecal indicators were also found in 50% of water samples from non- agricultural or pristine watersheds, sometimes exceeding 100 cfu/100 mL. Howell et al. (1995) monitored FC and fecal streptococci (FS) in streams, wells and springs of two agricultural watersheds in Kentucky. All sites that were monitored yielded samples that exceeded primary contact standards (>200 cfu/100 mL). Streams exceeded primary contact standards between 87 and 100% of the time. Springs and wells exceeded primary contact standards between 28 and 74% of the time. FECAL COLIFORMS IN RURAL WATERSHEDS 35

Until recently, it was thought that the survival of fecally derived microorganisms was limited in the natural environment. However, it has recently been demonstrated that fecal bacteria can survive for weeks, and sometime months, within surface water systems. Temperature has a significant influence on bacterial survival with several researchers noting that survival is enhanced under cooler conditions (Davenport et al. 1976; Barcina et al. 1986; Flint 1987). Extended survival patterns have especially been noted for bacteria that have attached to sediment particles and settled to the bottom of streams and lakes (Burton et al. 1987). The survival of fecal bacteria in sediments is primarily attributed to the availability of soluble organics (Davies et al. 1995; Marino and Gannon 1991) and to increased protection from predatory protozoans (Enzinger and Cooper 1976). Several studies have shown that concentrations of indicator organisms are typically higher in sediment as opposed to the overly- ing water column in both marine and freshwater systems (Burton et al. 1987; Sherer et al. 1992; Gary and Adams 1985). It has therefore been pos- tulated that enteric bacteria can survive longer, and possibly grow, within stream bottom sediments. The objective of this research was to quantify the presence of fecal coliform bacteria in the surface waters of a rural watershed and to attempt to determine the primary sources of fecal pollution within rural watersheds.

Materials and Methods

A small sub-catchment within the Cornwallis River watershed was selected for this study. The Cornwallis River watershed is approximately 26,000 ha in size and is located within the Annapolis Valley of Nova Scotia. The Annapolis Valley is the most intensively farmed region of the province. The Cornwallis River discharges into the Minas Basin. Thomas Brook is a tributary of the Cornwallis River. The stream originates on North Mountain and discharges into the Cornwallis River, just north of the town of Berwick. The stream network consists of two upper branches which join to form one channel approximately one-third of the distance through the watershed. Thomas Brook is a small stream, rarely greater than 2 m in width. The main channel of the stream network is 5800 m in length. The average slope of the stream channel is 3.5%. Channel grades in the upper third of the watershed are much higher (9%) than in the lower portion of the watershed (0.5–1.3%). The Thomas Brook watershed area is approximately 1000 ha. Land use within the lower two-thirds of the watershed consists primarily of pasture and cropland. Crops grown within the watershed include corn, strawberries and grains. A variety of soil types exist within the watershed. The areal distribution of soil type, and their respective properties, are provided in Table 1. In general, the soils are a reddish brown sandy loam (Cann et al. 1965). There are several potential microbial pollution sources within the watershed including three livestock operations (1 dairy and 2 beef). The 36 JAMIESON ET AL.

Table 1. Soil conditions within the Thomas Brook watershed

% of Soil Series Watershed Location Description

Pelton 40 Upland Dark reddish-brown fine sandy loam over yellowish loam, well drained Rossway 10 Upland Dark brown friable silt loam over reddish-brown loam, well drained Eroded 10 Upland Eroded or eroding land, steep forested slopes, variably drained Somerset 13 Lowland Dark reddish-brown loamy sand over red sandy loam, well drained Kingsport 8 Lowland Dark reddish-brown friable sand over red sandy loam, imperfectly drained Stewiacke 4 Lowland Reddish-brown loamy sand over red sandy loam, moderately drained Debert 4 Lowland Dark brown friable sandy loam over reddish sandy loam, imperfectly drained Woodville 4 Lowland Dark brown friable sandy loam over yellowish sandy loam, well drained Millar 4 Lowland Very dark grey sand over gleyed grey sand, poorly drained Lawrencetown 3 Lowland Dark grey clay loam over brown clay loam, very poorly drained

locations of the three farms are shown on Fig. 1. The dairy operation consists of approximately 300 animals. Runoff from manure storages, or from cropland that has received manure, could be a microbial pollution source. Allowing cattle to have direct access to the stream could also result in pollutant loading. On-site septic systems are also a potential source of pollution within the watershed. Residences which are within 50 m of the brook are shown on Fig. 1. There is a dense clustering of domestic residences along the upper right branch of the stream network (Fig. 1). Water quality monitoring stations were established at five locations throughout the watershed. The stations were designated ST1 to ST5 and are shown on Fig. 1. ST1 is located just downgradient from a headwater spring and was intended to represent background water quality. ST1A was not a permanent monitoring station. It was on an intermittent stream that was sampled twice during rainfall events to verify it possessed similar bacterial water quality to ST1. ST2 was located downstream from the dairy farm, while ST3 was located downstream from the cluster of hous- FECAL COLIFORMS IN RURAL WATERSHEDS 37

Fig. 1. Locations of monitoring stations and sources of fecal pollution. es on the right upper branch. ST4 was located approximately two-thirds of the way through the watershed, after the two upper branches had joined. ST5 is located at the outlet of the watershed, just before the stream meets the Cornwallis River. At ST2, ST4, and ST5 continuous streamflow measurements were obtained. Water depths at ST2, ST4, and ST5 were measured using KPSI Series 169 pressure transducers (Pressure Systems, Inc., Hampton, Va.) which were placed approximately 5 cm above the stream bed. Pressure transducer measurements were recorded every 60 s by a Campbell Scientific CR10X datalogger (CSI, Logan, Utah) and then averaged hourly. Water depth measurements were manually performed at ST1 and ST3 when samples were collected. Water depth measurements were converted to flow estimates by constructing a stage-discharge relation for the stream at each of the monitoring stations. The area- velocity method was used to calculate flow rates at each location under varying flow conditions (Linsley et al. 1982). Rainfall was also measured at ST5 using a Campbell Scientific TE525M tipping bucket rain gauge (CSI, Logan, Utah). Grab samples were collected and analyzed for FC on a weekly basis during dry periods. Samples were collected more frequently during storm events. Samples were collected mid-channel, 3 to 5 cm below the water 38 JAMIESON ET AL. surface at all sampling locations in sterile 250-mL plastic bottles. Bacterial colonies were estimated using the multiple tube fermentation technique (Method 9221: APHA 1998). This technique yields a most probable number (MPN) of FC organisms contained within 100 mL of sampled water. Fecal coliform densities within stream sediments were also enumerated on one occasion. Stream sediments near each of the monitoring stations were collected with a small, sterile metal scoop. In the lab, the sediments were homogenized and a 10-g sample of each was removed and used to aseptically prepare serial 1/10, 1/100, and 1/100 dilutions in sterile water. The dilutions were thoroughly mixed and analyzed for FC by the multiple tube fermentation technique (Method 9221: APHA 1998), as modified by the U.S. EPA (1999) for sludges and other solid samples to provide a MPN FC per g fresh weight of sediment. Water quality within the watershed was also monitored for several other parameters. At each sampling station water samples were collected with ISCO 6700 autosamplers (ISCO, Inc., Lincoln, Nebr.). The autosamplers were programmed to collect a 200-mL sample every 6 h and to composite every four samples (one composite 800-mL sample every 24 h). When the samples were collected (typically every 6 d) three consecutive 800-mL samples were combined to form one composite sample for every three-day period. These samples were transported in a cooler to an accredited analytical laboratory for analysis. Analysis of the composite samples included: total P (TP), soluble reactive P (SRP), total kjedhal N (TKN), ammonia-N (NH3-N), nitrate-N (NO3-N), sulfate (SO4), pH, and electrical conductivity (EC).

Results

Hydrology Table 2 presents monthly rainfall totals recorded within the watershed. Also provided in Table 2 are the Climate Normals values for Kentville, N.S. The 2001 growing season was characterized by drought conditions. During the period of June through October, rainfall in 2001 was 53% of the normal. Average mean daily flows and maximum mean daily flows at each of the monitoring stations are presented in Table 3. Average daily flows are also presented graphically in Fig. 2. Peak flows within the watershed occurred during the month of May (Fig. 2) due to high rainfall in combination with snowmelt runoff. After the first week of June, baseflow conditions existed for the majority of the summer and early fall. Considerable stormflow did not occur until the first week of November. The average mean daily flow rate at the watershed outlet was 0.76 m3/s, with maximum mean daily flow values reaching 10 m3/s in early May. As illustrated in Table 3, flow recorded at ST4 was approximately 50% of the flow recorded at ST5, the outlet of the watershed. It appears that groundwater inputs in the lower third of the watershed constitute a FECAL COLIFORMS IN RURAL WATERSHEDS 39

Fig. 2. Fecal coliform concentrations and average daily flows. 40 JAMIESON ET AL.

Table 2. Monthly rainfall totals recorded within the watershed during the study period

Monthly rainfall (mm)

Month 2001 Normala

May 137.7 88 June 59.2 81.6 July 46.1 81.9 August 15.6 90.7 September 71.7 86.6 October 43.8 102.1 November 97 107.7 December 30.2 74

a Climate Normals for Kentville, N.S. (1951–1996) (Environment Canada 1998). large portion of the streamflow, especially during low flow conditions. Flow rates measured at ST1 (headwater) averaged 0.005 m3/s. Maximum flow rates observed at ST1 were approximately 0.01 m3/s during early May. Flow was not present at ST1 from mid-June to early November.

General Water Quality A summary of the water chemistry observed at each of the monitoring stations is provided in Table 4. Reported are the arithmetic mean concentrations (AM), flow-weighted mean concentrations (FWM), and the maximum concentrations for each primary water quality parameter which was measured during this study. Throughout the stream system, pH levels ranged from neutral to slightly alkaline. Surface water appears

Table 3. Average mean daily flows and maximum mean daily flows at each monitoring station during the study period (May to December, 2001)

Average mean Maximum mean Station daily flow (m3/s) daily flow (m3/s)

ST1 0.005 0.01 ST2 0.16 3 ST3 0.22 5 ST4 0.37 8.2 ST5 0.76 10 FECAL COLIFORMS IN RURAL WATERSHEDS 41 EC 4 -N SO 3 -N NH 3 TP SRP NO 0.0550.26 0.05 0.24 0.54 0.75 0.02 0.5 8.5 179 7.07 — 218 — b a MaxAMMax 0.18AMMax 0.3 0.56 0.07AMMax 0.055 0.14 0.29 0.63AM 1.62 0.09 0.055 0.42 0.11 0.9 2.53 0.07 0.05 0.08 0.57 0.15 0.8 0.24 1.17 0.07 27.7 2.25 0.14 4.05 0.55 247 512 2.4 7.44 100 0.16 0.44 200 7.65 8.12 0.17 413 136 7.61 193 7.94 1050 597 115 7.74 558 8.18 811 7.84 520 678 499 Max 0.17 0.1 4.56 0.43 165 8.29 632 Summary of water chemistry observed at each of the monitoring stations during the study period (May to December, 2001) Summary of water chemistry observed at each the monitoring stations during study period (May to December, FWM; Flow weighted mean. AM; Arithmetic mean. a b able 4. ST2ST3 FWM ST4 FWMST5 FWM 0.056 FWM 0.08 0.057 0.07 0.07 0.58 0.07 1.95 0.21 2.34 0.17 57 0.17 93 — 80 — — — — — T StationST1 AM (mg/L) (mg/L) (mg/L) (mg/L) (mg/L) pH µmhos 42 JAMIESON ET AL. to become more alkaline moving downstream through the watershed system. With respect to TP concentrations, both the FWM and the AM at the watershed outlet (ST5) were 0.07 mg/L. The highest average TP concentrations were observed at ST2, just below the dairy farm, where a FWM of 0.26 mg/L was recorded. Mean NO3-N concentrations were low in the upper reaches of the watershed (ST2 and ST3). However, stream NO3-N levels increased moving downstream from ST3 to ST5, with a maximum FWM of 2.34 mg/L observed at the watershed outlet (ST5). Flow weighted mean values of NH3-N were relatively low throughout the watershed. The highest concentrations were observed at ST2.

Fecal Coliforms Fecal coliform concentrations observed during the study period at ST2 to ST5 are presented in Fig. 2. The Canadian FC water quality criteria for recreational activities (200 MPN FC/100 mL: CCME 1999) is also shown in Fig. 2 as a horizontal dashed line. Fecal coliforms were usually not detected in samples collected at ST1, and are therefore not presented in Fig. 2. However, one sample collected at ST1 during the first major flow event in November possessed a FC concentration of 900 MPN FC/100 mL. Fecal coliform concentrations collected after this time were <5 MPN FC/100 mL and FC was not detected at ST1 after December 1. Wildlife sources could have contributed to the formation of a sediment FC reservoir during the dry summer and fall, which could have been subsequently flushed during the first major flow event. Fecal coliforms were not detected in the two samples collected at ST1A. High FC concentrations occurred concurrently with stormflow peaks in May at all monitoring stations (Fig. 2). Fecal coliform levels then decreased substantially as stormflows receded. Concentrations at all stations then proceeded to rise again and tended to remain above the recreational water quality guideline for the remainder of the summer and early fall. When substantial stormflows occurred again in early November, FC concentrations decreased throughout the watershed. Concentrations remained below the recreational water quality guideline until the end of the study period (December 10), except at ST4 where FC concentrations rose again after a December 1 storm event. This could indicate a source of FC somewhere between ST2 and ST4. Another possibility is that FC loads generated in the upper part of the watershed were deposited along the stream bed upstream from ST4, where the channel slope starts to decrease. An examination of Fig. 2 tends to indicate that ST2 is the most impaired with respect to bacterial water quality. Concentrations of FC observed in this study are similar to other water quality studies conducted in agricultural watersheds (Niemi and Niemi 1991; Howell et al. 1995). Fecal coliform data were also assessed in terms of exceeding applicable water quality guidelines. The two primary uses of surface water within the Cornwallis watershed are irrigation and recreation. The percentage of samples which exceeded Canadian guidelines (CCME 1999) for FECAL COLIFORMS IN RURAL WATERSHEDS 43 these two purposes (100 MPN FC/100 mL for irrigation, 200 MPN FC/100 mL for recreation) at ST2 to ST5 are presented in Table 5. The data set has also been separated into samples which were collected during wet, or stormflow, periods and samples which were collected during dry, or baseflow, periods. The study period was divided into wet and dry periods by manually interpolating each storm hydrograph’s recession to baseflow conditions. At the watershed outlet (ST5) 87% of the samples exceeded the irrigation water quality guideline and 65% exceeded the recreational water quality guideline. Exceedances were more common during baseflow, or dry, conditions. At ST5, 94% of the samples collected during low flow conditions exceeded 200 MPN FC/100 mL. The highest percentage of exceedances were observed at ST2. During baseflow conditions, all samples collected at ST2 possessed FC concentrations exceeding 200 MPN FC/100 mL. It appears that the reach upstream of ST2 is a major contrib- utor of fecal organisms during low flow conditions. Bacterial water quality was also worse at ST3 during low flow conditions, with 67% of dry weather samples exceeding 200 MPN FC/100 mL. There was no active agricultural activity upstream of ST3 during the study period. A number of residences, serviced by on-site septic systems, could be the source of FC to the Thomas Brook upstream of ST3. The dairy farm located upstream from ST2 is a possible source of FC within that reach.

Table 5. Percentage (%) of fecal coliform samples which exceeded the water quality standards

% of Samples exceeding standarda Total Wet Dry Station Standard (n = 31) (n = 13) (n = 18)

ST2 100 MPN FC/100 mLb 90 77 100 200 MPN FC/100 mLc 84 62 100 ST3 100 MPN FC/100 mL 58 39 72 200 MPN FC/100 mL 52 31 67 ST4 100 MPN FC/100 mL 84 62 100 200 MPN FC/100 mL 65 54 72 ST5 100 MPN FC/100 mL 87 77 94 200 MPN FC/100 mL 65 54 72

aResults are reported for the total sample set (Total), samples collected during wet periods (Wet), and for samples collected during dry periods (Dry). bIrrigation water quality standard. cRecreational water quality standard. 44 JAMIESON ET AL.

Concentrations of FC were combined with streamflow data to determine the load of bacteria at ST2 to ST5 (Table 6). The calculated bacteria loads presented in Table 6 should only be considered approximate estimates, however, they provide useful information on the relative contribution of FC from various parts of the watershed. It was determined that approximately 75% of the bacteria load at the watershed outlet occurred during stormflow conditions. This would be expected as high rainfall would generate several bacteria transport mechanisms, including surface runoff and tile drain flow. Saturated soils may also cause septic systems to malfunction. Another significant source could be FC reservoirs in the stream bottom sediments. Turbulent flow could resuspend fecal organisms into the water column. Although the dry weather load represents only 25% of the total load, it is still substantial enough to impair the stream because of decreased dilution. In an attempt to identify the major sources of bacterial pollution under different flow conditions, the FC loads at ST2, ST3, and ST4 were expressed as a percent of the watershed outlet (ST5) load (Table 6). The FC loads observed at ST2, ST3, and ST4 were simply divided by the load observed at ST5. Please note that the values presented in the second part of Table 6 will not sum to 100%. The load observed at ST4 is a combination of loads observed at ST2 and ST3. These values illustrate the cumulative loading of FC moving down through the watershed. These values can only be considered approximate, as they do not account for resettle- ment and die-off of bacteria as they move through the stream network. Overall, the FC load at ST2 represented 45% of the outlet mass load, increasing to 57% during low flow conditions. In comparison, the FC mass load at ST3 represents only 20% of the outlet load during low flow conditions. The FC load at ST4 represents approximately two-thirds of the

Table 6. Fecal coliform loads

Fecal coliform load expressed Fecal coliform load (MPN FC)a as a % of outlet loadb

Station Total Wet Dry Total Wet Dry

ST2 4.9 x 1013 3.2 x 1013 1.7 x 1013 45 40 57 ST3 3.2 x 1013 2.6 x 1013 6.1 x 1012 29 33 20 ST4 6.4 x 1013 4.4 x 1013 2.0 x 1013 58 55 67 ST5 1.1 x 1014 8.0 x 1013 3.0 x 1013 ———

aResults are reported for the total study period (Total), wet periods (Wet), and the dry periods (Dry). bThese values were calculated by dividing the FC load observed at ST2, ST3, and ST4 by the FC load observed at ST5. FECAL COLIFORMS IN RURAL WATERSHEDS 45

FC load observed at the watershed outlet. Therefore, approximately one- third of the total watershed FC load is generated in the lower reach of the watershed, between ST4 and ST5. Concentrations of FC in stream sediments throughout the watershed are presented in Fig. 3. Large numbers of FC were enumerated in sediment samples collected at ST1 to ST4. The FC levels at ST5 were at least an order of magnitude lower than at the other stations. The highest concentrations were observed at ST2, where 58,500 MPN FC/g sediment were enumerated. There appears to be a substantial reservoir of fecal bacteria within the stream sediments of Thomas Brook. This reservoir could be acting as a major source of bacteria during both low and high flow conditions. A stream sediment bacterial reservoir has been noted in several other surface water quality investigations (Burton et al. 1987; Seyfried and Harris 1990; Sherer et al. 1992; Crabill et al. 1999). Allowing cattle to enter streams could disturb sediments and release bacteria into the water column (Sherer et al. 1988). Usage of the river for recreational purposes, such as canoeing, could also result in sediment resuspension. If the bacteria remain adsorbed to the sediment particles, they will eventually settle back to the stream bed. The turbulence, however, could cause the bacteria to desorb off the sediment, where it could continue to migrate downstream. As previously mentioned, stream bottom sediments can provide an environment which is conducive to the extended survival, and possibly growth, of enteric microorganisms. This means that bacteria that are transported into streams during spring runoff events could effectively degrade in-stream water quality for several months. This phenomena makes it extremely difficult to determine the source of bacterial pollution through routine monitoring programs. It is likely that only a few precipitation events during the course of a year transport significant quantities of bacteria to the stream system. Controls must be implemented to minimize bacterial transport during these key events. If controls fail during even one rainfall event, surface water quality could be degraded for several months.

Fig. 3. Concentrations of fecal coliforms in stream sediments. 46 JAMIESON ET AL.

Conclusions

Within this watershed, the primary surface water pollutant was fecal bacteria. Levels of FC frequently exceeded recreational water quality guidelines. Exceedances were more common during dry weather, or low flow conditions. This could pose a concern as the use of surface waters for recreation and irrigation activities frequently occurs during the drier summer months, when bacteria concentrations are high. Substantial FC loading was observed along stream reaches impacted by livestock operations. Substantial FC loading was also observed along a stream reach which was not impacted by agricultural activities. A dense clustering of residences with domestic on-site septic systems is the suspected source. Results from this study indicate the presence of a reservoir of fecal microorganisms within the stream sediments. The release of fecal microorganisms from the stream sediments to the water column during both low and high flow conditions could be a major source of bacterial loading. The possible extended survival of fecal microorganisms within stream sediments requires further investigation. It should be determined if this phenomena is the primary cause of microbial water quality degradation during low flow conditions.

Acknowledgements

This project was funded through the Nova Scotia Department of Agriculture and Fisheries AWARDS program and the Agri-Futures Nova Scotia Adaptation Council (Agriculture and Agri-Food Canada). Several individuals were involved in the field and reporting activities associated with this project: Lindsay Carter (Nova Scotia Department of Agriculture and Fisheries), Bruce Curry (Nova Scotia Department of Agriculture and Fisheries) and Gary Patterson (Agriculture and Agri-Food Canada).

References

APHA (American Public Health Association). 1998. Standard methods for the examination of water and wastewater, (20th edition). Washington, D.C. Barcina I, Arana I, Iriberri I, Egea L. 1986. Factors affecting the survival of E. coli in a river. Hydrobiol. 141:249–253. Burton G, Gunnison D, Lanza G. 1987. Survival of pathogenic bacteria in various freshwater sediments. Appl. Environ. Microbiol. 53:633–638. Cann D, MacDougall J, Hilchey J. 1965. Soil survey of Kings County, Nova Scotia. Report No. 15. Nova Scotia Soil Survey. CCME (Canadian Council of Ministers of the Environment). 1999. Canadian environmental quality guidelines. Crabill C, Donald R, Snelling J, Foust R, Southam G. 1999. The impact of sediment fecal coliform reservoirs on seasonal water quality in Oak Creek, Arizona. Water Res. 33:2163–2171. Crane SR, Westerman PW, Overcash MR. 1981. Die-off of fecal indicator organisms following land application of poultry manure. J. Environ. Qual. 9:531–537. FECAL COLIFORMS IN RURAL WATERSHEDS 47

Davies C, Long J, Donald M, Ashbolt N. 1995. Survival of fecal microorganisms in marine and freshwater sediments. Appl. Environ. Microbiol. 61:1888–1896. Davenport C, Sparrow E, Gordon R. 1976. Fecal indicator bacteria persistence under natural conditions in an ice-covered river. Appl. Environ. Microbiol. 32:527–536. Doran JW, Linn DM. 1979. Bacteriological quality of runoff water from pasture- land. Appl. Environ. Microbiol. 37:985–991. Environment Canada. 1998. Canadian climate normals. http://www. cmc.ec.gc.ca/climate/normals/NSH002.htm. Accessed January 31, 2002. Enzinger R, Cooper R. 1976. Role of bacteria and protozoa in the removal of Escherichia coli from estuarine waters. Appl. Environ. Microbiol. 31:758–763. Flint K. 1987. The long-term survival of Escherichia coli in river water. J. Appl. Bact. 63:261–270. Gary H, Adams J. 1985. Indicator bacteria in water and stream sediments near the snowy range in southern Wyoming. Water Air Soil Pollut. 25:133–144. Howell JM, Coyne MS, Cornelius PL. 1995. Fecal bacteria in agricultural waters of the bluegrass region of Kentucky. J. Environ. Qual. 24:411–419. Kudva IT, Blanch K, Hovde CJ. 1998. Analysis of Escherichia coli O157:H7 in ovine or bovine manure and manure slurry. Appl. Environ. Microbiol. 64:3166–3174. Landry MS, Wolfe ML. 1999. Fecal bacteria contamination of surface waters associated with land application of animal waste. Paper No. 99-4024. Toronto, Ont., ASAE. Linsley R, Kohler M, Paulhus J. 1982. Hydrology for engineers. McGraw Hill, Inc. New York. Marino R, Gannon J. 1991. Survival of fecal coliforms and fecal streptococci in storm drain sediments. Water Res. 25:1089–1098. Mubiru DN, Coyne MS, Grove JH. 2000. Mortality of Escherichia coli O157:H7 in two soils with different physical and chemical properties. J. Environ. Qual. 29:1821–1825. Niemi RM, Niemi JS. 1991. Bacterial pollution of waters in pristine and agricultural lands. J. Environ. Qual. 20:620–627. Patni NK, Toxepeus HR, Jui PY. 1985. Bacterial quality of runoff from manured and non-manured cropland. TRANS ASAE 28:1871–1877. Seyfried P, Harris E. 1990. Bacteriological water quality study examining the impact of sediment and survival times in the Humber River and Black Creek. R.A.C. Project No. 198G. Sherer B, Miner R, Moore J, Buckhouse J. 1988. Resuspending organisms from a rangeland stream bottom. TRANS ASAE 31:1217–1222. Sherer B, Miner R, Moore J, Buckhouse J. 1992. Indicator bacterial survival in stream sediments. J. Environ. Qual. 21:591–595. U.S. EPA. 1999. Control of pathogens and vector attraction in sewage sludge - appendix F: sample preparation for fecal coliform tests and Salmonella sp. analysis. U.S. EPA Office of Research and Development, Washington, D.C., Doc # EPA/625/R-92/013, p. 137–140. Water Qual. Res. J. Canada, 2003 Volume 38, No. 1, 49–76 Copyright © 2003, CAWQ

Portrait of Drinking Water Quality in Small Quebec Municipal Utilities

HOUSSEINI D. COULIBALY AND MANUEL J. RODRIGUEZ*

Département d’aménagement, Université Laval, 1624 Pavillon Savard, Québec City, QC G1K 7P4

This study of small Quebec municipal drinking water utilities (i.e., those serving 10,000 or fewer people) focuses on a portrait of microbiological water quality (based on total and fecal coliform data) and distribution system management strategies. It also addresses relationships between some important water quality and operational parameters and management strategies, as well as total or fecal coliform occurrences. Along with descriptive analyses, statistical means tests (Student t-tests) were performed to identify significant differences between utilities with high coliform occurrence and utilities with low coliform occurrence according to chlorine dose, distribution system flushings, pipe age, main breakage, and some environmental factors. Even though many interesting trends have been noted, only a few resulted in statistically significant differences. For surface water utilities using chlorination alone, the mean difference of annual system flushings proved statistically significant. In addition, some agricultural land-use indicators within the municipal territory appeared significantly correlated with coliform occurrences.

Key words: drinking water, water quality, small utilities, coliform occurrences, distribution system, Quebec

Introduction

In North America, small drinking water utilities (serving 10,000 or fewer people) have more difficulty than larger utilities in complying with the increasingly stringent regulations on distributed water quality. Indeed, small utilities have generally limited financial and technical resources allowing them to remove contaminants from raw water, to adequately operate the treatment and to implement management strategies to monitor and control water quality in the distribution systems (Gouvernement du Québec 1997; AWWA 2000). Among the important challenges for managers of small drinking water utilities are the necessi- ty of simultaneously ensuring adequate microorganism inactivation in the plant and control in the distribution system, and minimizing the formation of disinfection by-products (DBPs) that are potentially carcino- genic, such as trihalomethanes (THMs) (Levallois 1997). When surface waters are utilized as raw water and chlorine is used as the principal disinfectant, such challenges become more considerable. In the U.S., small

* Corresponding author; [email protected] 50 COULIBALY AND RODRIGUEZ water utilities using either surface or ground water will, in the coming months or years, have to comply with a number of new National Primary Drinking Water Regulations such as stage 1 of the Disinfectants/DBP rule (for residual disinfectant, maximum DBP levels and required treatment for organic carbon removal), the long-term 1 Enhanced Surface Water Treatment Rule (requirements for contact time [CT] calculations and filter monitoring/performance based on turbidity) and the Groundwater Rule (requirement of sanitary surveys, raw water monitoring and treatment- based monitoring results) (U.S. EPA 1998a,b; U.S. EPA 2000). In Canada, the federal government has elaborated guidelines for drinking water quality that are not mandatory, but that can be used by provinces to promulgate regulations for utilities within their territory (Santé Canada 1996). Before 2000, only two Canadian provinces, Alberta and Quebec (Gouvernement du Québec 1984), had promulgated mandatory regulations. Following the water contamination event in the small community of Walkerton (Ontario) during the summer of 2000, in which seven people died and more than 2000 were taken ill, some other Canadian provinces (British Columbia, Nova Scotia, Ontario) published new regulations or updated existing ones. This was the case of the Province of Quebec, where, in June 2001, the government published new Drinking Water Quality Regulations (DWQR), of which the application was mandatory for all utilities supplying water to more than 20 people (Gouvernement du Québec 2001). The 2001 DWQR constitutes a considerable update of the first regulations promulgated in 1984. Small utilities in Quebec, especially those using surface water, are particularly concerned by the 2001 DWQR, mainly because new requirements or stringent standards are considered for turbidity of clear water, microorganism inactivation (virus, Giardia and Cryptosporidium), bacterial monitoring, minimum levels of residual chlorine and maximum annual average THM levels in the distribution systems.

Small Utilities in Quebec

In the Province of Quebec, there are about 1000 municipal utilities that serve between 51 and 10,000 people. According to the Quebec Ministry of Environment (QME), small utilities are known to have difficulties distributing water of good quality (Gouvernement du Québec 1997). Among small municipal utilities in Quebec, about 350 supply chlorinated water with no previous physicochemical treatment to about 900,000 people (Gouvernement du Québec 1997). The relatively high concentrations of natural organic matter (NOM) in most lakes, and the microbial pollution and high turbidity in southern Quebec streams, in particular associated with agricultural drainage, will make compliance with the 2001 DWQR apparently very difficult for most of these utilities. In accordance with the quality of the source water, particularly with the turbidity levels, most of these utilities will probably have to modify their water treatment strategy in the coming years. SMALL DRINKING WATER UTILITIES 51

Even if the 2001 DWQR are currently in application, and although small utilities will be especially impacted by these regulations in terms of infrastructure and operation updating requirements, there is currently very little knowledge about the characteristics of these types of utilities. There is little knowledge about their current state in terms of raw and distributed water quality, treatment and disinfection practices, infrastructure and strategies to maintain the quality of water within the distribution system. In other words, at present there is no portrait of small utilities in Quebec allowing identification of their state and problems. The aim of this paper is to establish a portrait of small drinking water utilities in Quebec using a combination of data developed by the authors and public data from diverse sources. Special emphasis will be placed on relating the state of the water quality (especially microbiological) with existing management practices. This portrait will allow identification of priorities and challenges for the upcoming years.

Sources of Data To establish a portrait of Quebec’s small utilities, three main sources of data were used. The first is a database managed by the QME, whose implementation is based on the routine reports carried out by water utilities to comply with 1984 DWQR (database 1). Database 1 contains mainly information on microbiological water quality for small utilities. The second source of data was generated by the QME from sampling programs aiming at getting information about, among other things, organic substances in the distribution system (database 2). For the purpose of this research, database 2 is comprised of THM data for small utilities. The third source of data is a database developed by the authors following a questionnaire-based survey directed to small utilities in Quebec (database 3) in order to obtain information about utility characteristics (treatment, operations, management, maintenance, etc.). The fourth database is comprised of information on manure production in Quebec municipalities. Other published and unpublished data is also being considered in order to compare characteristics with those of other utilities.

Database 1 According to the 1984 DWQR, drinking water utilities in Quebec have had to test their drinking water for specific parameters and send a report of results to the QME. The frequency for parameter monitoring depends principally on the utility size (i.e., the population served). For small utilities, a very low monitoring frequency was required for most inorganic and organic water quality parameters (in general, one or two samples per year). These parameters include turbidity and residual chlorine, while total THMs were not required to be monitored even if a 350-µg/L maximum contaminant level existed. However, for microbiological quality control, from 1 to 10 samples per month (half of them taken at distribution system extremity) were required to be analyzed (for 52 COULIBALY AND RODRIGUEZ utilities serving 201 to 10,000 people) in order to comply with regulations concerning total and fecal coliforms. This was the only mandatory microbiological parameter according to 1984 DWQR (in June 2002, one year after the promulgation of the 2001 DWQR, higher monitoring frequen- cies and additional microbiological parameters became mandatory). According to this, data about total and fecal coliform tests currently constitute the only historical information on microbiological water quality based on utility compliance reports to QME. Consequently, database 1 consists of this kind of data for a three-year period, 1997, 1998 and 1999. By considering this period, it is possible to represent the recent trends and take into account variations from year to year. Database 1 includes information on 927 utilities. The latter are municipal utilities that serve from 201 to 10,000 people and which transmitted data on bacteriological control (i.e., fecal and total coliforms) to QME from 1997 to 1999. This choice was based on the fact that in accordance with the Quebec 1984 drinking water regulations (Gouvernement du Québec 1984), utilities serving from 51 to 200 people had to take only two samples per year for bacteriological analyses.

Database 2 Since 1985, the QME has carried out sampling campaigns in selected Quebec water utilities in order to investigate, among other things, the occurrence of organic compounds (called the Surveillance Program) (Gouvernement du Québec 1997). In this program, special attention has been focused on THM presence, particularly in vulnerable utilities, mean- ing those using surface waters with moderate or high organic carbon content, during the summer period (generally April to October). In general, within the Surveillance Program, samples for THMs were collected following chlorination and/or within the distribution system (about 1.5 km from the plant). Information generated from the Surveillance Program was used by the QME, for example, to evaluate the technical and economical feasibility of updating the 350-µg/L THM standard included in the 1984 DWQR (Riopel 1992; Rousseau 1993; Tremblay and Trinh-Viet 1995). This THM standard was used only as a guideline for utilities, since monitoring requirements were not stipulated until the publication of the 2001 DWQR. Consequently, the information on THM occurrence provided by such sampling programs constitutes the only data available on a historical basis for Quebec water utilities. For the purpose of this research, database 2 consequently consists of THM data from the small utilities which took part at least once (all the year or only in summer) in the QME Surveillance Program during 1997, 1998 and 1999.

Database 3 Databases 1 and 2 provide information about two key parameters characterizing water quality in distribution systems, coliform counts and THM. Because there is no database containing information on characteristics for small water utilities in Quebec, it was decided to conduct a ques- SMALL DRINKING WATER UTILITIES 53 tionnaire-based survey specifically for utilities serving from 201 to 10,000 people. Utilities selected for the survey were part of those included in database 1. The survey completed in early 2000 was based on a questionnaire sent by mail to the principal manager/operator of each utility, asking for information about various issues. These issues include general characteristics (type of water source, population served, number of municipalities served, flow rates, etc.), water treatment procedures, disinfection issues, the quality of treated and distributed water, distribution system infrastructure and strategies to maintain water quality throughout the distribution system. To validate the questionnaire (test), fifteen utilities were pre- selected at random and asked to respond. Some minor adjustments were made based on their responses and comments. The questionnaire was then sent to about 25% of the 927 above-mentioned utilities (precisely, to 247 utilities). In total, 114 small utilities returned the completed questionnaire, resulting in a response rate of about 46%. For specific questions within the questionnaire the response rate varied considerably.

Database 4 In many municipalities in Quebec, agricultural pollution is of great concern. In order to control this threat to the environment and regulate the issue, the QME developed a database comprising data related to manure production in each municipality. For the purpose of the present study, a subsidiary database was built from data of municipalities corresponding to the 114 responding utilities.

Other published data Results obtained from databases 1, 2 and 3 will allow the establishment of a portrait of small utilities in Quebec. Some particular characteristics of this portrait will in addition be compared to characteristics of small and larger utilities of the U.S. Basically, comparisons will be carried out using information collected from the Water Utility Database (now known as ‘Water:\Stats,’ AWWA 1998), which is a survey conducted in 1996 essentially among large and medium sized utilities in the U.S. (i.e., those serving 10,000 or more people) and from the results of the disinfection practices survey for U.S. small utilities (serving 10,000 or fewer people) conducted in 1998 (AWWA 2000).

Portrait of Small Utilities

The portrait of small Quebec utilities was established based on all of this information. The portrait comprises mainly the state of the microbiological water quality (database 1), the state of the physicochemical water quality (databases 2 and 3), an overview of management strategies which may influence water quality in the distribution system (database 3), the relationships existing between microbiological quality and some management strategies (databases 1 and 3), and the relationships between microbiological quality and agricultural (environmental) factors (databases 1 and 4). 54 COULIBALY AND RODRIGUEZ

General Characteristics and Physicochemical Water Quality Most of the utilities where staff completed and returned the survey questionnaire operated very small utilities: 37% served from 201 to 1000; 57% served from 1001 to 5000 and only 6% served from 5001 to 10,000 people. Indeed, the survey response rate for utilities serving from 5001 to 10,000 was found to be considerably lower than the response rate for the rest of the utilities (only 30.4%, while those serving from 201 to 1000 and from 1001 to 5000 people had a response rate of 45.2% and 49.6%, respectively). The majority of the surveyed utilities use surface water (i.e., water from lakes, rivers and streams, or groundwater directly influenced by surface water) (Table 1). The average served flow rate of distributed water was found to be about 2600 m3 per day according to 58 utilities who provided this information. Concerning the parameters of water quality, the response rate was relatively low (Table 2). This is understandable, considering that in 1984 DWQR, requirements for monitoring physicochemical parameters were weak. Thus, only data for parameters for which the response rate is above 15% are presented (Fig. 1). From the utilities providing quality parameter data, about 55% indicated the turbidity of their raw water to be lower than 1 nephelemetric turbidity unit (NTU) and about 28% to be higher than 5 NTU (Fig. 1a). All groundwater utilities providing data (except two) have indicated turbidity levels lower than 1 NTU, whereas about one-third of surface water utilities indicated average raw water turbidity higher than 5 NTU. It is important to mention that these turbidity values are average values, and that there may be large differences between the average and maximum values encountered. These maximum turbidity values, which are not documented in this paper, are often the main source of problems. Only one-third of utilities indicated the true colour to be lower than 15 true

Table 1. Water source, treatment, and disinfectant type for the surveyed utilities

Number of Response rate Characteristics respondents (n) (out of 114 utilities)

Source water Surface water 70 61.4 Groundwater 44 38.6 Treatment No treatment 28 24.6 Chlorination alone 71 62.3 Treatment 15 13.1 Disinfectant None 28 24.6 Cl2 41 36.0 NaClO 42 36.8 Cl2 and NaClO 3 2.6 SMALL DRINKING WATER UTILITIES 55

Table 2. Water quality and operational parameters for the surveyed utilities

Number of Response rate respondents (out of Parameters (n) 114 utilities)

Turbidity raw water winter 32 28.1 summer 30 26.3 treated water winter 28 24.6 summer 27 23.7 Colour raw water winter 20 17.5 summer 20 17.5 treated water winter 18 15.8 summer 18 15.8 Chlorine dose winter 51 44.7 (operational) summer 51 44.7 Free chlorine treated water winter 24 21.1 residual summer 24 21.1 distributed water winter 19 16.7 summer 21 18.4

colour units (TCU), whereas about the same proportion indicated true colour higher than 50 TCU (Fig. 1b). Distribution of turbidity and colour values indicated by respondents was significantly higher in summer than winter. A reasonable explanation for this is the presence of snow/ice layers in southern Quebec surface watersheds during about four months of winter, which naturally protect sources of water from watershed runoff. Another possible explanation is that because of high summer water temperatures, biological activity, planktonic in particular, is much higher. Very few surveyed utilities indicated the use of a physicochemical treatment to remove turbidity, colour and organic carbon (Table 1). Practically all those utilities (with one exception) indicated the use of surface water as a water source. Only 20% of utilities using surface water apply other treatment before chlorination (flocculation, settling, filtration). This is extremely different from the results obtained by others in the U.S., where 94% of surveyed small utilities indicated the use of at least filtration before disinfection (AWWA 2000), which is a direct consequence of the U.S. National Primary Drinking Water Regulations (U.S. EPA 1989). A more surprising result is that about one-fourth of small Quebec utilities do not apply any treatment or disinfection (even for residual disinfectant maintenance) before delivering water into the distribution system. However, practically all of these utilities use groundwater as a raw water source (only one utility indicated using surface water without any treatment). All utilities which disinfect water before distribution use chlorine-based dis- 56 COULIBALY AND RODRIGUEZ

(20) 15 (31) 150

12 120 (32) 9 90 (21)

6 60 (18) (18) Turbidity, NTU Turbidity, (27) True color, TCU 30 3 (28)

0 0 winter summer winter summer winter summer winter summer Raw water Treated water ab Raw water Treated water

5 1.5 (24) (24)

(21) 4 (51) 1.2

(51) 3 0.9

dose, mg/L dose, (19) 2 0.6

1 0.3 Free chlorine, mg/L chlorine, Free Chlorine

0 0 winter Summer winter summer winter summer

c Season d Treated water Distributed water

Fig. 1. Distribution of water quality parameters among responding utilities: a) turbidity, b) colour, c) chlorine dose, and d) free chlorine residual. In brackets: number of utilities; lower (–): C10; upper (–): C90; (x): mean. infectants, the same proportion with gas chlorine and with calcium/ sodium hypochlorite. The use of hypochlorite as chlorine-based disinfectant appeared a little more widespread in small Quebec utilities (in 50% of utilities with disinfection) than in small U.S. utilities (34%) (AWWA 2000). Applied chlorine doses before water distribution appeared to be higher in summer than in winter (Fig. 1c). This is not consistent with the fact that chlorine efficacy for microorganism inactivation is higher in warm waters than in cold. However, these data appear realistic, considering that chlorine is the unique disinfectant applied in the surveyed utilities, that is, it is utilized simultaneously for ensuring inactivation and to maintain residual chlorine levels in the distribution system. For the latter purpose, higher chlorine doses are generally applied in summer to coun- terbalance the rapid decay of residual chlorine associated with higher water temperatures. These results are comparable to those of research programs undertaken by the authors with medium and large drinking water utilities of Quebec (Milot et al. 2000; Rodriguez et al. 2000; Rodriguez and Sérodes 2001). It was also observed that average dose levels for small Quebec utilities were higher among utilities not using treatment (1.44 mg/L on average) than among those using treatment (1.12 mg/L on average). The higher doses were indicated by utilities that chlorinate surface waters without any previous physicochemical treatment (on average, 1.66 mg/L). Concerning the physicochemical quality in treated water (before distribution), more than 80% of responding utilities indicated producing drinking water with less than 1 NTU and about 50% with less than SMALL DRINKING WATER UTILITIES 57

0.5 NTU, which is the standard of the Quebec 2001 DWQR (Fig. 1a). A surprising result is that no responding utility indicated producing treated water with turbidity levels higher than 5 NTU, which was the maximum acceptable level in the 1984 DWQR. Average values for true colour indicated by respondents were lower than 5 TCU in more than 75% of the utilities (Fig. 1b). Also, according to the respondents, levels of both turbidity and colour in the distribution system were comparable to levels of these parameters in treated water. It was observed that utilities using treatment (all with surface water, except for one) had the best quality according to both parameters (average values for turbidity and colour in treated water equal 0.2 NTU and 1.2 TCU, respectively), followed by groundwater utilities with or without treatment (average values for turbidity and colour in treated water equal 0.7 NTU and 4.8 TCU, respectively). Utilities applying direct chlorination to surface water had the lowest physicochemical water quality (average values for turbidity and colour in treated water equal 0.7 NTU and 19.3 TCU, respectively). In the case of free residual chlorine in the water leaving the plant (following chlorination), practically all utilities reported more than 0.2 mg/L (which was the standard included in the 1984 DWQR), with higher values in summer than in winter (Fig. 1d). Moreover, all utilities reported having detectable levels for this parameter at the extremity of the distribution system: about 90% of the surveyed utilities reported levels for this parameter as being above 0.1 mg/L, whereas more than half reported values above 0.2 mg/L. These reported values appear higher than expected (especially in summer, when chlorine demand is high), considering that all utilities reported maximum residence time of water higher than 12 hours. Considering the fact that a new standard for THMs is included in the Quebec 2001 DWQR (80 µg/L based on a quarterly annual average of samples taken at the extremity of the distribution system), it was considered appropriate to create a portrait of concentrations of this parameter in small Quebec utilities. Because only about 2% of responding utilities provided data about THMs, the portrait for these parameters was made on the basis of information included in database 2 (Fig. 2a,b). According to results, in more than 30% of the utilities THM levels in the distribution system (that is, 1.5 km from the plant) were below 50 µg/L, and in more than 55% below 80 µg/L. Considering that all available data were generated from samples taken between April and October, it is probable that the annual average concentration of THMs for these utilities are in reality further below the THM values shown in Fig. 2. This allows us to infer that, if for such utilities sample location represents the extremity of the distribution system, the majority of utilities would comply with the 2001 DWQR. According to Fig. 2, utilities more susceptible to not complying with the THM standard are those that directly chlorinate surface waters (without any previous treatment). This is understandable, considering that THM precursors contained in raw waters (natural organic matter) are not removed by a physicochemical treatment in these utilities (mean TOC values are 3.22, 3.81 and 3.10 mg/L for groundwater plus chlorination 58 COULIBALY AND RODRIGUEZ

200 200 (n = 64) (n = 79) (121) (10) (n = 15) 150 (26) 150

100 100

50 50 Total THMs, ug/l THMs, Total Total THMs, ug/l THMs, Total

0 0 GW+chlorination SW+chlorination SW+treatment 201-1000 1001-5000 5001-10000 a b System size (population served)

Fig. 2. Average total THM concentrations according to: a) source water and treatment; and b) utility size. In brackets: number of utilities; lower (–): C10; upper (–): C90; (x): mean. GW denotes Groundwater; SW denotes Surface Water. alone, surface water plus chlorination alone and surface water plus treatment, respectively).

Portrait of the Microbiological Water Quality To establish the portrait of microbiological water quality of the treated and distributed water of small Quebec utilities, two indicators were built up using the information concerning total coliform (TC) and fecal coliform (FC) counts of database 1. To distinguish water samples considered microbiologically contaminated from those not contaminated, TC and FC data were initially converted in a dummy variable, indicating negative samples for TC (less than 10 organisms/100 mL, the maximum that does not systematically represent a violation of the Quebec DWQR) and positive samples for TC (more than 10 organisms/100 mL). As for FC, any count different from zero was considered positive. From this new variable, two indicators were created. The first is called coliform episode and indicates one or a set of coliform-positive samples occurring in a given distribution system during the three-year period (1997–1999), separated by at least 15 days from any other coliform-positive sample in the same system. Such a criterion allows us to consider as a unique episode a number of positive samples occurring in a short period of time, and that have probably been associated with the same cause. This criterion also allows us not to consider as independent episodes the number of positive samples encountered following the intensive sampling program that generally follows the detection of a first positive sample for TC or FC (sampling carried out in the days following the laboratory results). The second indicator is called problematic utility and designates a utility that registered one or more coliform episodes in at least two of the three reference years. Consequently, utilities that registered no coliform episode, or had episodes in only one of the above-mentioned three years, were called non- problematic utilities. This indicator distinguishes utilities with recurrent coliform occurrences from those with rare or no such occurrences. Using data about total and fecal coliform in database 1 and the two indicators described above, a portrait of water quality was established SMALL DRINKING WATER UTILITIES 59

(Table 3). Judging by data in Table 3, it would be hard to say that the respondents’ sample is representative of the population. However, since emphasis, as stated above, is put on relating the microbiological water quality with the existing management practices, the responding sample representativeness may be of less concern. The central issue is to look for factors that may explain coliform appearances (i.e., episodes) in studied distribution systems. Moreover, even assuming that the survey sample

Table 3. Portrait of the coliform appearances in investigated small utilities (1997 through 1999)

Target Responding utilities utilities (n = 927) (n = 114)

Percent out of concerned utilities total — 12 Percent out of total population served — 14 Average population served 1807 2062 Percent of coliform-positive samples 200–1000a 2.7 4.0 in summer (April through September) 1001–5000 1.3 2.0 5001–10,000 0.8 0.9 Percent of coliform-positive samples 200–1000 1.3 1.1 in winter (October through March) 1001–5000 0.4 0.5 5001–10,000 0.2 0.3 Percent of utilities with at least 200–1000 55 79 1 coliform episode 1001–5000 50 66 5001–10,000 42 57b Average number of coliform episodes 200–1000 2.6 3.2 for utilities with at least 1 episode 1001–5000 2.4 2.6 5001–10,000 2.2 3.7 Percent of problematic utilities out 200–1000 27 55 of total number of utilities 1001–5000 21 34 5001–10,000 16 57b Percent of problematic utilities among 200–1000 50 70 those with at least 1 coliform episode 1001–5000 42 51 5001–10,000 37 100b Average number of episodes for 200–1000 3.8 3.9 problematic utilities 1001–5000 3.9 3.8 5001–10,000 3.9 3.7

aPopulation served. bSuch abnormally high values are due to a very small total for this group (7 utilities only, compared to 83 among the 927). 60 COULIBALY AND RODRIGUEZ

(n = 247) is representative of the population of utilities (n = 927), it would be impossible to ensure that the respondents’ sample be representative, since one could have no control on the ultimate decision of a surveyed utility manager to respond or not. Information in Table 3 shows that even though the average percentage of coliform-positive samples appears relatively low (about 1%), a high number of utilities have experienced microbiological water quality problems. According to the period under study, half of small Quebec utilities have experienced one or more coliform episodes. Among these utilities, the average number of episodes was about 2.4, whereas one-fifth of utilities experienced more than three episodes (only 1% experienced all of them in one year). About 25% of all are problematic utilities, that is, having experienced recurrent microbiological problems in the distributed water. It is also observed that the portrait for microbiological water quality varies considerably according to the utility size. Indeed, more than 2% of all water samples collected in very small utilities (serving between 201 and 1000 people) during the period under investigation were found coliform-positive, this percentage being significantly higher during summer periods. However, no significant differences were observed between samples taken in the distribution system extremities in comparison with those taken in other locations (data not shown in Table 3), which is a surprising result considering that it is well known that extremities constitute favorable locations for biofilm development and locations at which levels of residual chlorine are the lowest. Differences between utilities according to their size are also observable when examining both indicators, coliform episodes and problematic utilities, but such differences appeared less important (in terms of relative value) than in the case when only the percentage of coliform-positive samples are examined. Such a result means that in very small utilities, a single coliform episode is represented by a higher number of positive TC or FC samples than in larger utilities. This suggests that in larger utilities (especially those serving between 5000 and 10,000 persons), coliform episodes are relatively short, probably related to the shorter time required in these utilities (having generally more human and technical resources) for identifying the source of microorganisms and the more rapid and efficient measures taken to resolve the problem. Differences in microbiological water quality according to the utility size appear directly related to the source of water and the treatment process applied (Table 4). Indeed, utilities using physicochemical treatment before chlorination, which are those that serve larger populations on average, have experienced significantly fewer problems of microbiological water quality in the distribution system than utilities using chlorination alone (from surface or ground water sources) or utilities that do not use treatment at all. Utilities that encountered the most important and frequent difficulties with microbiological water quality in the distribution system are those that directly chlorinate surface waters. One can notice from the analysis made earlier that the same group of utilities (which represent one- third of small Quebec utilities) are those that also have the highest values SMALL DRINKING WATER UTILITIES 61

Table 4. Distribution of population served and coliform episodes according to water source and treatment types (for responding utilities)

Type of treatment

No treatmenta Chlorination alone Treatmentb

Groundwater Surface water Groundwater Surface water n = 27 n = 55 n = 16 n = 14

Percent out of total 24 48 14 12 number of respondents Percent out of 17 46 17 19 population served by responding utilities Average population 1478 1965 2504 3242 served Percent of utilities 59 82 75 43 with at least 1 coliform episode Average number 2.6 3.2 2.6 2.7 of episodes for utilities with at least 1 episode Percent of 33 53 37 29 problematic utilities Average number 3.2 4.2 3.5 3.5 of episodes for problematic utilities

aThere was only one utility which used surface water and no treatment; that case was ignored. bSo was the sole utility that used groundwater and treatment. of THMs. Finally, this group is also the one that encompasses the highest percentage of utilities that experienced at least one coliform episode, the highest percentage of problematic utilities, and the highest average number of episodes. Generally speaking, the coliform occurrences appear more recurrent for utilities supplied by surface sources, confirming the high vulnerability to microbial intrusion for such utilities. 62 COULIBALY AND RODRIGUEZ

Strategies for Maintaining Microbiological Water Quality in the Distribution System Small utilities were also asked during the survey for information about characteristics of their infrastructure and the routine and long-term strategies to manage water quality in the distribution system. Issues investigated, such as rechlorination and pipe characteristics and maintenance (age, material, break rate, corrosion strategies and pipe cleaning strategies), can directly or indirectly affect the water quality within the distribution system. Table 5 presents the information obtained for some of these issues.

Rechlorination practices Residual chlorine is recognized to be an indicator of water quality in a distribution system, particularly because it can reduce the risk of microbial regrowth (Sobsey et al. 1993; Sérodes et al. 1998; Haas 1999; LeChevallier 1999). It is noteworthy however that the issue of maintaining a residual is not clear cut, and has generated some controversy in recent years. In this respect, a number of authors consider that the neces- sity of chlorine residual maintenance is arguable due to its poor efficacy to inactivate waterborne pathogens in drinking water distribution systems (Payment 1999; van der Kooij et al. 1999). Because chlorine reacts with organic and inorganic compounds when added to water in the plant before distribution, residual chlorine levels can rapidly decay and even disappear at extremities, especially for utilities with extensive distribution systems and long retention times (Kirmeyer et al. 1993; Reiber 1993). Rechlorination of water within the distribution system may counterbal- ance initial chlorine decay. According to small Quebec utility respondents, only a small percentage of utilities (about 10%), particularly the larger ones (in terms of both population served and pipeline length), use rechlorination facilities within the distribution system to maintain sufficient residual chlorine levels. However, it was found that the average residual chlorine for small responding utilities using rechlorination is practically the same in winter and significantly lower in summer than average residual for utilities not practicing rechlorination. It is interesting to observe that almost all responding utilities that rechlorinate are surface water utilities that do not use any treatment before chlorination. This is probably due to the fact that the chlorine demand following the dose application is higher for those utilities because of the lower quality of water. Thus, to compensate for high initial chlorine demand, rechlorination generally appears to be a good strategy to ensure minimal levels of residual chlorine and minimize the probability of bacterial regrowth.

Pipe characteristics The issue of water main assessment and associated research needs is well documented (AWWA 1994; Rajani and McDonald 1995; Kitaura and Miyajima 1996; Makar 2000; Rajani et al. 2000). Aging distribution system SMALL DRINKING WATER UTILITIES 63 Maximum Mean 90 C 50 C Percentiles 10 PVCothers 100 100 0.00 0.00 0.00 0.00 20.0 0.00 75.0 30.9 100 100 27.8 9.52 Survey responses for specific distribution system characteristics (pipe age, pipe material, main breaks, and system flushings) for specific distribution system characteristics (pipe age, pipe material, main breaks, Survey responses able 5. Pipe age, yr of total pipe materialPercent Cast-ironNumber of main breaks/km/yr 100 0.00 95 104 2.20 0.01 75.0 2.00 100 0.07 20.0 0.22 100 33.5 62.8 0.62 60.0 1.50 100 0.29 36.2 T CharacteristicsNumber of flushings per yr Respondents (n) Minimum C 104 1.00 1.00 2.00 3.00 12 1.93 64 COULIBALY AND RODRIGUEZ pipes, in particular those made of iron-based material, can cause water quality deterioration within the distribution system, especially through corrosion. In addition to favoring precipitation of metal ions, which can cause coloured water, pipe corrosion may favor the formation of tubercles within which a biological film can form or cause breaks in the main, both aspects being favorable conditions for deterioration of microbiological water quality (LeChevallier et al. 1990). Distribution system pipes of the responding small utilities appeared, surprisingly, relatively older in comparison to medium and large utilities in Quebec. Indeed, an average of about 57% of pipes of small Quebec utilities are, according to respondents, 35 years old or less, compared to an average of 65% of medium and large Quebec utilities (Villeneuve and Hamel 1998; Fougères et al. 1998), and 24% of responding utility pipes are 20 years or less, compared to 34% for medium and large utilities. However, only a minority (about 30%) of responding small Quebec utilities acknowledged that their pipes suffered from corrosion problems, and only a third of those utilities implemented corrosion control strategies (generally by ensuring a relatively high pH by adding calcium or phosphate). This result appeared surprising, considering that on average, 63% of the distribution system pipes are made of cast iron (on average, 28% made of PVC). Concerning the infrastructure of the distribution system, small Quebec utilities reported an average rate of breaks which can be considered acceptable according to McDonald et al. (1994), who judged that a main break rate can be considered abnormally high when it exceeds 40/100 km/year (78% have had this many or less). However, it is observed that only half the utilities reported a break rate that is lower than 25/100 km/year, which is the average for distribution systems of Ontario towns, according to the Ontario Sewer and Watermain Contractors Association (CMHC 1992). The average break rate indicated by responding utilities (about 29/100 km/year) is also more than double the one for U.S. towns’ distribution systems, that is, about 13/100 km/year (AWWA 1994). Results indicate that the average main break rate for responding utilities more than 50 years old (30/100 km/year) is slightly lower than the one for those with ages ranging from 31 through 50 years (32/100 km/year). As expected, the younger utilities (30 years old or less) experienced many fewer main breaks (27/100 km/year). The average for all utilities more than 30 years old is also about 32/100 km/year. Surprisingly, the average main break rate for utilities which suffer from corrosion problems is slightly lower than for those not experiencing such problems: 29/100 km/year and 30/100 km/year, respectively. Moreover, the mean age for utilities experiencing corrosion (about 41 years) is higher than the one for utilities without corrosion (about 35 years). Utilities applying corrosion control strategies had significantly fewer breaks (24/100 km/year) than those which have not developed such strategies (33/100 km/year), but the former are younger than the latter (mean ages of 39 and 44 years, respectively). So, it seems that all of this is tied to pipe age, hence the importance of an adequate pipe replacement policy. Besides, according to Villeneuve et al. (1998), only about 1% of the total pipe mileage of Quebec utilities is SMALL DRINKING WATER UTILITIES 65 replaced every year. This replacement rate may appear too low, judging by the above-mentioned (observed) breakage rates. One important strategy for maintenance of water quality in distribution systems is to carry out periodic flushing in order to eliminate deposits in the pipe wall internal surface. Flushing is considered an efficient strategy, particularly to take out biofilm and corrosion tubercles which both favor microbiological deterioration within the distribution system (Antoun et al. 1999; Duranceau et al. 1999). All small Quebec utilities reported flushing the distribution system at least once each year and more than half reported at least two flushings. Most of the utilities carrying out only one flushing usually made it later in summer or in fall. According to Antoun et al. (1999), this may be a good management strategy, because it ensures pipeline cleaning after the period within which biofilm development is most proliferous. This similarity appears surprising, but encouraging, considering that generally speaking, larger utilities possess higher financial capacities for maintenance of infrastructure. However, it was also observed that very small Quebec water utilities (those serving less than 1000 persons) carry out as many flushings as larger ones (on average two per year).

Relationships between Management Strategies and Microbiological Water Quality The portrait of microbiological water quality was also investigated in accordance with the management strategies mentioned earlier. It was developed principally by combining the information contained in databases 1 and 3. Tables 6A to D present the results concerning these analyses. Particular attention was paid to the more vulnerable utilities, mean- ing those which directly chlorinate surface waters. According to results, utilities not having water quality problems generally apply lower chlorine doses during both winter and summer (Table 6A). These results appear surprising, because it is expected that higher chlorine doses ensure higher microbial inactivation and higher free chlorine residual concentration and, thus, greater protection from microbiological degradation of water quality in the distribution system. These results suggest that in small utilities where recurrent microbiological problems exist, managers use higher chlorine doses as a corrective measure, but this strategy does not necessarily prevent or reduce these problems. Certainly, increasing the applied chlorine dose does not necessarily ensure an increase of residual chlorine in every location of the distribution system, and thus does not necessarily ensure an improvement of microbiological water quality, since many other factors can be related to coliform regrowth in drinking water (LeChevallier et al. 1996). According to Table 6B, utilities with recurrent water quality problems practice fewer flushings of their distribution system on average than those which do not have these problems. Even if the median for the number of annual flushings is similar for utilities with and without recurrent problems, it appears that utilities which make two or more flushings per year have better results within a perspective of microbiological water 66 COULIBALY AND RODRIGUEZ P 0.92 0.62 0.94 0.94 0.80 1.50 0.50 1.34 of dose dose Surface water utilities using 5 5 n 18 1.35 1.90 18 1.15 1.38 a P 0.76 0.70 0.22 0.43 (N = 114) (N = 55) of dose dose Median Mean Median Mean All responding utilitiesAll responding chlorination alone n 1 episode 37 1.50 1.74 25 2.00 2.01 1 episode 37 1.00 1.22 24 1.50 1.41 ≥ ≥ Observed relationship between the chlorine dose (mg/L) and the utility microbiological status between the chlorine dose (mg/L) and utility microbiological Observed relationship ; Significance level of the means test. P a Utilities utilitiesNon-problematic 32 1.20 1.56 Utilities utilitiesNon-problematic utilitiesProblematic Utilities with no episode 33 18 1.00 14 1.11 1.25 0.90 1.21 1.23 11 1.50 1.41 Utilities with no episode 14 0.55 0.95 Problematic utilitiesProblematic 19 1.50 1.67 12 2.25 1.95 inter able 6A. Summer T W SMALL DRINKING WATER UTILITIES 67 P 0.16 0.005 2.00 3.78 flushings flushings Surface water utilities using 9 n 24 2.00 2.58 a P 0.29 0.14 (N = 114) (N = 55) flushings flushings All responding utilitiesAll responding chlorination alone Median of Mean of Median of Mean of n Observed relationship between distribution system flushings and the utility microbiological status between distribution system flushings and the utility microbiological Observed relationship 1 episode 74 2.00 1.82 43 2.00 1.84 ≥ ; Significance level of the means test. P a able 6B. T utilitiesProblematic 45 2.00 1.69 28 2.00 1.82 Utilities with no episodeUtilities 30 2.00 2.20 Non-problematic utilitiesNon-problematic 59 2.00 2.12 68 COULIBALY AND RODRIGUEZ quality than those which make only one. This trend was much stronger and statistically significant for utilities which directly chlorinate surface waters. The results in Table 6B suggest that flushing generally has real, positive impacts on distribution system water quality. As mentioned earlier, it is well known that aging distribution systems may favor corrosion and biofilm development in the pipe wall, thereby possibly affecting water quality. However, according to Table 6C, no significant differences in microbiological water quality were observed in small Quebec utilities according to the age of the distribution system, even if the age variations of the utilities under study are important (as presented earlier in Table 5). A possible explanation for this is that the average age of the distribution system pipes is not necessarily representative of the entire utility (i.e., very large pipe age variations can exist in a single utility), because it is very probable that some pipes have never been replaced, while others could have been replaced very recently. However, no information about pipe replacement rate was available from database 3. Finally, even if pipe breaks are known to be a possible source of microbial intrusion in distribution systems, no significant differences of the annual breakage rate were observed between utilities having water quality problems and those not having them (Table 6D). However, a surprising result is observed for utilities that directly chlorinate surface waters. Among these utilities, those not having microbiological problems at all (i.e., any episode at all) have significantly higher pipe breakage rates (for both mean and median values) than those that do have quality problems. In addition, the average pipe breakage rate in these utilities appeared higher than the maximum acceptable (as reported by McDonald 1994). Many possible explana- tions may be put forward to explain this apparently illogical result. First, it appears that extreme breakage statistics are more frequent among utilities experiencing 29 breaks/100 km/year (the overall average value) or fewer. Second, the relative weight of utilities practicing chlorination alone (which were found to be more often problematic than all others) is bigger among this same group. Third, the age, type and corrodibility of pipe material may also be involved; for instance, for utilities having less than 50% of their pipelines made of PVC, the average number of main breaks is much higher than that for utilities with more than 50% PVC (32 breaks/ 100 km/year and 22 breaks/100 km/year, respectively).

Relationships between Agricultural Land Use and Microbiological Water Quality As part of Quebec’s recent regulations about agricultural pollution, and in order to control cattle breeding expansion in locations where agriculture is already too intensive, all municipalities of the province have been designated a manure status (as specified in data received from QME). Such a status is a function of the intensity of agricultural pressure on their territory (soils). This factor is measured by the annual balance of phosphorus in terms of kilograms of phosphorous (P2O5) per hectare (ha). It considers the total SMALL DRINKING WATER UTILITIES 69 P 0.20 0.39 of age age Surface water utilities using n 1025 22.0 25.0 27.2 32.5 a P 0.65 0.37 (N = 114) (N = 55) of age age Median Mean Median Mean All responding utilitiesAll responding chlorination alone n Observed relationship between distribution pipe age and the utility microbiological status between distribution pipe age and the utility microbiological Observed relationship 1 episode 72 34.0 36.8 42 32.0 36.6 ≥ ; Significance level of the means test. P a able 6C. T utilitiesProblematic 44 35.0 38.0 27 34.0 36.9 Utilities with no episodeUtilities 32 31.5 35.0 Non-problematic utilitiesNon-problematic 60 30.0 34.9 70 COULIBALY AND RODRIGUEZ P 0.06 0.33 eaks breaks 0.24 0.43 r Surface water utilities using 8 nb 22 0.17 0.31 a P 0.82 0.67 (N = 114) (N = 55) eaks breaks r All responding utilitiesAll responding chlorination alone Median of Mean of Median of Mean of nb Observed relationship between distribution main breaks and the utility microbiological status and the utility microbiological between distribution main breaks Observed relationship 1 episode 67 0.25 0.29 39 0.17 0.23 ≥ ; Significance level of the means test. P a able 6D. T utilitiesProblematic 41 0.22 0.28 25 0.17 0.23 Utilities with no episodeUtilities 28 0.20 0.31 Non-problematic utilitiesNon-problematic 54 0.23 0.30 SMALL DRINKING WATER UTILITIES 71 manure production within the municipality, the nutrient requirements of the crops and the cultivated area. When the annual balance is more than 20 kg P2O5/ha/year, the authorities consider the corresponding municipality as being in manure surplus. However, for a number of municipalities, even a zero annual balance is considered an administrative surplus, because they are situated in watersheds with already significant phosphorus excess. Even if such an annual balance was not calculated based on watershed limits but rather on municipal limits, it can be used as an indicator of the susceptibility of surface waters to be contaminated by surface or subsurface runoff. For the purpose of this study, information about the manure status had been considered under four variables in order to associate it with water quality in small utilities. These variables are: zone with/without manure production, zone administratively/not administratively in a surplus situation, annual manure balance less or equal to/more than 0 kg P2O5/ha, and surplus of phosphorus less/equal to or more than 20 kg P2O5/ha/year. The impact of each of these factors on microbiological water quality is analyzed in Table 7. The results indicate that on the whole, utilities located in zones with high agricultural pressure experienced more water quality problems (related to total or fecal coliforms). The impact of agricultural pressure on water quality appeared significant for the more vulnerable utilities, that is, those chlori- nating surface water without any previous treatment. Indeed, two of the four manure-related variables, (i.e., “administratively in phosphorus surplus” and “phosphorus annual balance”) were found to be significantly correlated with the variable “number of coliform episodes.” This suggests that future controlling of cattle breeding and piggery expansion may have a considerable effect on small vulnerable utilities.

Multivariate Analyses In order to evaluate interactions between variables or potential col- lective impacts of the studied management strategies on microbiological water quality, multivariate analyses were carried out. Three variables: “problematic/non-problematic,” “episodes/no episode,” and “number of episodes” had to be explained. Because the first two are dichotomous, a binary stepwise logistic regression analysis was performed to search for factors explaining them. For the continuous variable (“number of episodes”), a linear regression analysis was used. First, analyses were carried out for all responding utilities, and then for respondents using surface water and chlorination alone. When all respondents are considered, the only variable that exhibits a significant relationship with the three specified dependent variables is the treatment type. This is obvious, and needs no particular explanation. So, no multivariate model emerges for the whole set of respondents. As for surface water utilities using chlorination alone, one model comes out and indicates that 33% of the explained variance related to the dichotomous variable “problematic/non-problematic” is tied to variables “phosphorus annual balance” and “phosphorus surplus more than 20 kg P2O5/ha/year,” with 72 COULIBALY AND RODRIGUEZ P 1 utilities, episode ≥ Surface water utilities using episode,% % number n a P (N = 114) (N = 55) All responding utilitiesAll responding chlorination alone 1 utilities, episode 1 utilities, ≥ Utilities Problematic Average Average Utilities Problematic Average Utilities Problematic episode,% % number n 0 kg P2O5/ha/yr 83 71 41 1.96 39 79 46 2.26 ≤ 20 kg P2O5/ha/yr 16 69 37 2.06 0.984 6 100 67 3.83 0.208 Observed relationship between agricultural pressure indicators and microbiological characteristics of utilities indicators and microbiological between agricultural pressure Observed relationship ≥ ; Significance level of the means test. P a able 7. T Surplus <20 kg P2O5/ha/yr 98 70 44 2.05 49 80 51 2.43 Zone with manure production production Zone with manure productionZone without manure Zone administratively in surplusZone administratively not in surplus 93 21 balance >0 kg P2O5/ha/yrManure 91 23 balance Manure 72 62 31 71 65 68 44 38 43 43 2.13 48 1.71 1.97 0.427 2.39 2.29 0.471 43 0.501 12 9 84 16 46 100 75 87 78 53 78 50 2.65 69 48 0.691 4.67 2.33 0.020 3.37 2.17 0.141 Surplus SMALL DRINKING WATER UTILITIES 73 the model being significant at the 1% (0.01) level (logistic regression analysis: χ2 = 11.9; R2 = 0.33; P = 0.003). This suggests that the fact that a surface water utility with chlorination alone is either problematic or non-problematic with regard to microbiological quality (total or fecal coliforms) is relatively easy to explain by the agricultural land use of the municipality where the utility is located. Such an indication seems easily explicable, since it is well known that cattle feces and piggery effluents contain great quantities of bacteria and parasites that may eventually find their way into water springs by means of agricultural runoff or infiltration into groundwater.

Conclusions

This research has documented some important characteristics of small Quebec drinking water utilities. First of all, one notes that even though all of these utilities are called small utilities, and are supposed to have very comparable financial and technical resources, the quality of their distributed water may vary considerably. Actually, three groups of utilities emerged during this study: first, utilities which never experienced problems with microbiological water quality during the reference three-year period (1997 through 1999); second, utilities that occasionally encountered difficulties complying with the Quebec 1984 DWQR relating to fecal or total coliforms; and, third, utilities which very often violated the mentioned DWQR. The first two groups can be considered as distributing relatively safe water to their customers. The last group obviously consists of utilities that have major problems. Most of the latter are utilities that directly chlorinate surface waters without any other treatment. These problematic utilities may need to acquire a treatment facility, especially considering the new and much more stringent DWQR promulgated by the Quebec government in June 2001. These utilities, unable to comply with coliform standards, will now have to cope with parasites, viruses, and monitoring of trihalomethanes, to name a few. It is hard to believe that small problematic utilities will overcome such obstacles, without managing, at least, to acquire a filtration facility in order to reduce NOM content of their distributed water. In any case, they will have to apply filtration in a relatively near future, since new DWQR (that came into force in June 2002, except for a few recently amended clauses including filtration, the effective date for the latter being postponed until June 2005 for utilities serving fewer than 50,000 people, and until June 2007 for those serving 50,000 or more people) make filtration practically inevitable for all Quebec surface water utilities. Concerning infrastructure and water quality maintenance, small utilities appeared to be aging, compared to medium and large ones. This may be attributable to the fact that most medium and large utilities pertain to numerous relatively young suburbs that grew all around big Quebec metropolitan areas like Montreal or Quebec City, some 40 to 50 years ago. Among distribution water quality management strategies analyzed, some 74 COULIBALY AND RODRIGUEZ interesting trends were noted when comparing mean values for utilities with no episode to those with episodes on the one hand, and for problematic and non-problematic utilities, on the other. However, very few of these trends were confirmed by results of bivariate or multivariate analyses (possibly due to the very discrete nature of microbial dissemination in distribution systems). Apart from treatment-related variables, only the manure-related variables exhibit some statistical impact. This may not be surprising, considering that many of the responding utilities are located in zones under high agricultural pressure. In terms of strict public health concern, it must be underlined that the so-called problematic utilities are not necessarily serving water bearing more of a health threat than the water served by the non-problematic ones. In fact, most of reported episodes concern total coliforms, which may tell more about the general salubrity of the distribution system than about real health hazards. Moreover, databases used for this study did not include data on parasites like Giardia lamblia and Cryptosporidium parvum, nor on viruses or other waterborne pathogens. These microorganisms are of great concern, since they have been tied to waterborne disease outbreaks in the U.S. and elsewhere. The only reason these parameters were not included in this study is that there is an almost total lack of data about them in small Quebec utilities. The fact that data came from different sources has led to different data considerations, which, to some extent, hindered this study. This situation may render difficult a comparison of these results to those of other studies. Despite these limitations, this study has the advantage of trying to create an overall portrait of microbiological and physicochemical water quality in small Quebec utilities, and trying to establish and explain relationships between the portrayed quality and some management practices or environmental factors (manure). This may be interesting for those who want to know more about the specificity of small utilities and the challenges they face, for instance, from a regulatory point of view.

Acknowledgements We are thankful to the Canadian Agency for International Development (CAID) for providing financial support through the Canadian French-speaking-world Fellowship Program. Likewise, we are grateful to the Quebec Ministry of Environment, and personally to Alain Riopel, Donald Ellis and Yolaine Blais, for providing historical water quality and agricultural land use data. Finally, we wish to express our grati- tude to the Centre de recherche en aménagement et développement (CRAD, Université Laval) for financial and logistic support.

References

Antoun EN, Dyksen JE, Hiltebrand DJ. 1999. Unidirectional flushing–a powerful tool. J. Am. Water Works Assoc. 91:62–71. AWWA. 1994. An assessment of water distribution systems and associated research needs. American Water Works Association, Denver, Colorado. SMALL DRINKING WATER UTILITIES 75

AWWA. 1998. Water:\stats: the water utility database. American Water Works Association, Denver, Colorado. AWWA. 2000. Disinfection at small systems. AWWA Water Quality Division Disinfection Systems Committee report. J. Am. Water Works Assoc. 92:24–31. CMHC. 1992. Urban infrastructure in Canada. Canada Mortgage and Housing Corporation, Ottawa. Duranceau SJ, Poole J, Foster JV. 1999. Wet-pipe fire sprinklers and water quality. J. Am. Water Works Assoc. 91:78–90. Fougères D, Gaudreau M, Hamel PJ, Poitras C, Sénécal G, Trépanier M, Vachon N, Veillette R. 1998. Évaluation des besoins des municipalités québécoises en réfection et construction d’infrastructures d’eaux. INRS-Urbanisation, Montréal. 266 p. Gouvernement du Québec. 1984. Règlement sur l’eau potable. Éditeur officiel du Québec, Québec. 7 p. Gouvernement du Québec. 1997. L’eau potable au Québec, un second bilan de sa qualité : 1989–1994. Ministère de l’Environnement et de la Faune, Québec. 36 p. Gouvernement du Québec. 2001. Règlement sur la qualité de l’eau potable. Ministère de l’Environnement, Québec. 19 p. Haas CN. 1999. Benefits of employing a disinfection residual. J. Water Supply: Res. Tech. Aqua 48:11–15. LeChevallier MW. 1999. The case for maintaining a disinfectant residual. J. Am. Water Works Assoc. 91:86–94. LeChevallier MW, Schulz WH, Lee RG. 1990. Bacterial nutrients in drinking water, p. 143–201. In American Water Works Association Research Foundation (ed.), Assessing and controlling bacterial regrowth in distribution systems. American Water Works Association Research Foundation, Denver, Colorado. LeChevallier MW, Welch NJ, Smith DB. 1996. Full-scale studies of factors related to coliform regrowth in drinking water. Appl. Environ. Microbiol. 62:2201–2211. Levallois P. 1997. Qualité de l’eau potable et trihalométhanes. Bulletin d’Information en Santé Environnementale (BISE) 8:1–4. Kirmeyer GJ, Foust GW, Pierson GL, Simmler JJ, LeChevallier MW. 1993. Optimizing chloramine treatment. American Water Works Association Research Foundation and American Water Works Association, Denver, Colorado. Kitaura M, Miyajima M. 1996. Damage to water supply pipelines. Soils and Foundations 36:325–333. Makar JM. 2000. A preliminary analysis of failures in grey cast iron water pipes. Institute for Research in Construction, National Research Council of Canada. 17 p. McDonald S, Daigle L, Félio G. 1994. Réseaux d’aqueduc et systèmes d’égouts. Laboratoire d’infrastructures, Institut de recherche en construction, Conseil national de la recherche du Canada, version préliminaire. Milot J, Rodriguez MJ, Sérodes J. 2000. Modeling the susceptibility of drinking water utilities to form high concentrations of trihalomethanes. J. Environ. Manage. 60:155–171. Payment P. 1999. Poor efficacy of residual chlorine disinfectant in drinking water to inactivate waterborne pathogens in distribution systems. Can. J. Microbiol. 45:709–715. Rajani BB, Makar JM, McDonald SE, Zhan C, Kuraoka S, Jen C-K, Viens M. 2000. Investigation of grey cast iron water mains to develop a methodology for estimating service life. American Water Works Association Research Foundation and American Water Works Association, Denver, Colorado. 266 p. 76 COULIBALY AND RODRIGUEZ

Rajani B, McDonald S. 1995. Water main break data for different pipe materials for 1992 and 1993. Report No. A-7019.1, National Research Council, Ottawa, Canada. Reiber S. 1993. Chloramine effects on distribution system materials. American Water Works Association Research Foundation and American Water Works Association, Denver, Colorado. Riopel A. 1992. Les trihalométhanes dans les petits systèmes de distribution au Québec: campagnes d’échantillonnage de 1987 et 1988. Direction des écosys- tèmes urbains, Ministère de l’Environnement, Gouvernement du Québec. 21 p. Rodriguez MJ, Sérodes J-B. 2001. Spatial and temporal evolution of trihalomethanes in three water distribution systems. Water Res. 35:1572–1586. Rodriguez MJ, Sérodes J-B, Morin M. 2000. Estimation of water utility compliance with trihalomethane regulations using a modelling approach. J. Water Supply: Res. Technol. Aqua 49:57–73. Rousseau H. 1993. Suivi des concentrations de THM dans huit (8) réseaux de distribution d’eau potable au Québec. Division des eaux de consommation, Direction des écosystèmes urbains, Ministère de l’Environnement et de la Faune, Gouvernement du Québec. 54 p. Santé Canada. 1996. Recommandations pour la qualité de l’eau potable au Canada. Sixième édition. Édition du Groupe Communication Canada, Ottawa. 102 p. Sérodes JB, Rodriguez MJ, Poncton A. 1998. Development and on-site evaluation of a decision-making tool for chlorine disinfection dose and residual control. Presented at the 8th National Conference on Drinking Water, Canadian Water and Wastewater Association (CWWA). Sobsey MD, Dufour PA, Gerba CP, LeChevallier MW, Payment P. 1993. Using a conceptual framework for assessing risks to health from microbes in drinking water. J. Am. Water Works Assoc. 85:44–48. Tremblay H, Trinh-Viet H. 1995. Réseaux municipaux visés par le règlement sur l’eau potable susceptibles de présenter une concentration moyenne annuelle de THM supérieure à 100 µg/l : estimation des coûts de réalisation des ouvrages. Service de l’assainissement des eaux et du traitement des eaux de consommation, Ministère de l’Environnement et de la Faune, Gouvernement du Québec. 31 p. U.S. EPA. 1989. National primary drinking water regulations: filtration, disinfection, turbidity, Giardia lamblia, viruses, Legionella, and heterotrophic bacteria. Final rule. Fed. Reg. 54:124:27486. U.S. EPA. 1998a. National primary drinking water regulations: disinfectants and disinfection by-products. Final rule. Fed. Reg. 63:241:69389. U.S. EPA. 1998b. National primary drinking water regulations: interim enhanced surface water treatment rule. Fed. Reg., 63:241:69477. U.S. EPA. 2000. National primary drinking water regulations: groundwater rule. Proposed rule. Fed. Reg. 65:91:30194. van der Kooij D, van Lieverloo JHM, Schellart J, Hiemstra P. 1999. Maintaining quality without a disinfectant residual. J. Am. Water Works Assoc. 91:55–64. Villeneuve JP, Hamel PJ. 1998. Synthèse des rapports INRS-Urbanisation et INRS-Eau sur les besoins des municipalités québécoises en réfection et construction d’infrastructures d’eaux. INRS-Urbanisation, Montréal. 50 p. Water Qual. Res. J. Canada, 2003 Volume 38, No. 1, 77–113 Copyright © 2003, CAWQ

A Conceptual Model for Cryptosporidium Transport in Watersheds

CHAN HEE PARK AND PETER M. HUCK*

Department of Civil Engineering, University of Waterloo, 200 University Avenue West, Waterloo, Ontario N2L 3G1

This paper describes a conceptual model to estimate Cryptosporidium parvum oocyst transport from source to water treatment plant intake. The intent of the model is ultimately to be able to predict oocyst concentrations at an intake to an order-of-magnitude level. The transport and fate mechanisms included are: oocyst detachment from waste or soil, generation of runoff, overland transport, reservoir and in-stream transport, and oocyst die-off. The model is formulated in finite difference form, and deals with both non-point sources from manure- applied areas, and point sources from wastewater treatment plants. An important contribution of this work is the recognition that the settling rates of free and floc- or particle-associated oocysts can be considerably different. This has important implications for their transport. A finite difference scheme was developed for five sections of a hypothetical watershed: a point source, a lake or reservoir (which can be modelled as either a continuous stirred tank reactor or an ideal rectangular setting tank), the section of stream channel from the outlet of the lake or reservoir to the confluence with another stream, a tributary with a non-point source, and the stream section from the confluence to a water treatment plant intake. The stream confluence is handled with a simple mass and flow balance. It would be very expensive to collect the necessary data to test the model. Because an appropriate data set was not available, the model was tested by means of a sensitivity analysis for the hypothetical watershed, using reasonable parameter settings for the base case. The major contribution of the model is in defining the mechanisms involved in oocyst transport within a watershed. It gives important insights into the significance of various factors, provides a basis for data collection, and identifies areas where experimental investigations are required to avoid the need for simplifying assumptions. At its current state of development, the model cannot be used to provide quantitative predictions, but defines a base from which further detailed modelling can be developed to aid in decision- making for pathogen control. Using the framework that this model provides, contributions from other sources of Cryptosporidium oocysts such as domestic animals and combined sewage overflows could also be modelled.

Key words: Cryptosporidium, oocysts, oocyst transport, modelling, drinking water, point source, non-point source, sensitivity analysis

* Corresponding author; [email protected] 78 PARK AND HUCK

Introduction

Cryptosporidium parvum has been implicated in a number of waterborne outbreaks of gastroenteritis. The most notorious outbreak was in Milwaukee, in 1993. More than 400,000 people were affected (MacKenzie et al. 1994), and it was the largest outbreak of waterborne disease ever recorded in the U.S.A. In some outbreaks, inadequacies in the water treatment process have been blamed for passage of oocysts into the water supply. Cryptosporidium parvum exists in the environment in the form of an oocyst. These oocysts are commonly detected in surface waters, whether the waters are pristine/near pristine or affected by human activity. Some surveys have found more than 70% of the waters sampled to be oocyst-positive (LeChevallier et al. 1991; Rose et al. 1991; Lisle and Rose 1995). Major oocyst sources include animal husbandry operations and sewage treatment effluents. Studies have shown that the concentration of Cryptosporidium oocysts is related to hydrologic events (Stewart et al. 1997; Atherholt et al. 1998) such as rainfall and runoff, and to watershed characteristics (Hansen and Ongerth 1991; Kelley et al. 1995; Ong et al. 1996). Therefore, an understanding of source terms and factors affecting oocyst transport is required to estimate raw water oocyst concentrations at drinking water intakes. In terms of transport, hydrologic events and watershed and watercourse characteristics are important. Because of the difficulty in inactivating Cryptosporidium with con- ventional disinfectants, enhanced treatment is often required to eliminate this pathogen effectively. The extent of treatment required is determined by the levels of pathogens in the raw water. In this regard, it is not only the average concentration that is important, but also the expected peak values. Knowing the significance of parameters affecting Cryptosporidium oocyst transport and developing models to estimate the concentrations at a drinking water intake dynamically are important in providing public health protection. To date, little work has been done to describe oocyst transport both to and within watercourses, although various models describing sediment transport, hydrologic processes and contaminant transport (chemical or other types of microorganisms) exist. This paper describes the development of a conceptual model for Cryptosporidium oocyst transport in a watershed. The long-term goal is to apply the model to specific watersheds in order to predict approximate peak oocyst concentration ranges to be expected at given water treatment plant intakes. This would require further model refinement and experimental work to quantify certain assumed model parameters and provide calibration data. Such a model likely could only predict oocyst concentrations to an order-of-magnitude level. However, predictions at this level are what is required for the design of water treatment facilities, because treatment requirements are related to order-of-magnitude oocyst concentrations in the raw water. MODEL FOR CRYPTOSPORIDIUM TRANSPORT IN WATERSHEDS 79

The specific objectives of this paper are: i) to describe the role of hydrologic parameters such as rainfall, runoff, and lake residence time with respect to oocyst transport in a watershed; ii) to establish an oocyst- specific conceptual model that could also be used to describe transport of other pathogens; iii) to present the finite difference scheme for the model; and iv) to test the model using a sensitivity analysis. A sensitivity analysis using a hypothetical watershed was the approach used for testing the model because an appropriate data set was not available. Such a data set would cost on the order of several hundred thousand dollars to obtain. Although the focus of modelling is Cryptosporidium parvum, the more general term Cryptosporidium is used in this paper.

Background Physical Behaviour of Cryptosporidium Oocysts Cryptosporidium oocysts may be considered as organic particles which have their own physical properties. Oocysts have a spherical shape, a diameter of 4 to 6 µm and a density of 1.05 g/cm3 (Medema et al. 1998). According to Drozd and Schwartzbrod (1996), oocysts do not demonstrate marked hydrophobic properties. These authors observed a zeta potential close to –25mV at pH 6 to 6.5 in deionized water, which is a little higher than –10 mV measured in natural water (Gregory 1994). Animal waste or oocysts on manure-applied areas can be thought of as a form of physical mixture of oocysts and waste or soil. Because of the surface properties of oocysts, their detachment from waste to water may be likely during runoff. Oocysts can exist in water in two states, a) detached, or b) attached to other suspended particles. It is important to consider their transport in a watershed separately, as this affects processes such as coagulation, flocculation, and sedimentation. Medema et al. (1998) compared experimentally determined sedimentation velocities of free and attached oocysts with calculated velocities (Stokes’ law) based on oocyst size and density, and the density and viscosity of the sedimentation medium. The theoretically calculated sedimentation rates showed good agreement with the experimentally observed rates. The initial apparent sedimentation velocity of free oocysts in Hanks balanced salt solution at 23°C was 0.35 µm/s (1.26 mm/h), which is extremely low. Considering a hypothetical example of runoff lasting one hour and having a depth of about 1 cm, it is reasonable to expect significant sedimentation of free oocysts would likely occur in lakes and reservoirs. Consequently, free oocysts could be transported quite a long way. Medema et al. (1998) also showed that oocysts in the surface water environment may be attached to other particles, which will affect their sedimentation rate. These authors mixed oocysts with settled secondary effluent (biologically treated sewage) and found that oocysts readily attached to the (biological) particles in the effluent. Thirty percent of the oocysts 80 PARK AND HUCK attached during the first minutes of mixing, and this fraction increased to approximately 75% after 24 h. The sedimentation velocity of oocysts attached to secondary effluent particles increased with particle size and was determined by the sedimentation rate of the effluent particles.

Existing Non-point Source Contaminant Models The Agricultural Nonpoint Source Pollution Model (AGNPS) was developed to compare the effects of different watershed pollution control management practices (Young et al. 1986). AGNPS simulates sediment and nutrient loadings from agricultural watersheds for single storm events or for continuous data input. Watersheds in the model are discretized into a series of square cells, for which homogeneous characteristic parameters are assigned. AGNPS is partitioned into two submodels. The erosion portion of the model provides estimates of upland erosion, channel erosion, and sediment yield. The model uses the Modified Universal Soil Loss Equation (Williams 1975) for soil erosion calculations and distributes predicted erosion into five particle size categories: sand, silt, clay, small aggregates, and large aggregates. The pollutant transport portion of AGNPS addresses only soluble pollutants, and loads are determined using relationships between chemical concentrations, sediment yield, and runoff volume (Young et al. 1986). The hydrology is calculated by the Soil Conservation Service (SCS) runoff curve number method (SCS 1972). The Chemicals, Runoff, and Erosion from Agricultural Management Systems (CREAMS) model was developed to aid in the assessment of agricultural best management practices for pollution control (Knisel 1980). CREAMS is a continuous simulation model, requiring continuous precipitation data and monthly values of air temperature and solar radiation. Soil and crop type data are also provided as inputs. In order to assess best management practices, the user can simulate various management activities, such as aerial spraying or ground application of pesticides, animal waste management, tillage operations, or terracing (Knisel 1980). CREAMS calculates runoff volume, peak flow, infiltration, evapo- transpiration, soil water content, and percolation on a daily basis. Daily erosion and sediment yield are also estimated and average concentrations of sediment-associated and solute chemicals are calculated for the runoff, sediment, and percolating water (Knisel 1980). CREAMS also uses the SCS curve number method. The erosion component of the model considers the basic processes of soil detachment (Universal Soil Loss Equation, USLE [Wischmeier and Smith 1978]), transport, and deposition. The ANSWERS (Areal Nonpoint Source Watershed Environment Response Simulation) is a distributed parameter, event-based model for predicting the hydrologic and erosion response of agricultural watersheds (Beasley and Huggins 1981). The watershed is divided into uni- form square elements. Within each element the model simulates processes of interception, infiltration, surface storage, surface flow, sediment detachment (USLE) and transport. It is primarily a runoff and sediment MODEL FOR CRYPTOSPORIDIUM TRANSPORT IN WATERSHEDS 81 model; the nutrient simulation is based on a simple correlation between chemical concentrations, sediment yield, and runoff volume. For four separate rainfall events, the simulated (ANSWERS) hydrologic responses were found to correlate closely with actual hydrograph responses in the watershed. Predicted sediment loads from ANSWERS, however, were significantly and consistently less than actual measured loads (Engel et al. 1993). Those authors showed that rough estimates for ANSWERS input parameters were sufficient for the prediction of hydrologic response, but not for predicting sediment loads. The Soil Water and Assessment Tool (SWAT; Arnold et al. 1993) was developed as an extension to the Simulator for Water Resources in Rural Basins (SWRRB; Williams et al. 1985). SWAT is a continuous spatially distributed watershed model operating on a daily time step. It simulates runoff, sediment, nutrient, and pesticide movement through a watershed and aids in assessing water supplies and non-point source pollution in large basins (Arnold et al. 1993). SWAT was one of the non-point source pollution water quality models assessed in a comparison conducted by Engel et al. (1993). SWAT estimates for total runoff and nutrient and sediment loads were less accurate than the ANSWERS simulated values. The Enhanced Stream Water Quality Model (QUAL2E), although not a non-point source model, is used for modelling stream water quality. It is applicable to well-mixed dendritic streams and simulates the major reac- tions of nutrient cycles, algal production, benthic and carbonaceous demand, atmospheric re-aeration and their effects on the dissolved oxygen balance. It can predict up to 15 water quality constituent concentrations. It is intended as a water quality planning tool for developing total maximum daily loads and can also be used in conjunction with field sampling for identifying the magnitude and quality characteristics of non- point sources. By operating the model dynamically, the user can study diurnal dissolved oxygen variations and algal growth, but the effects of dynamic forcing functions, such as headwater flows or point source loads, cannot be modelled. QUAL2EU is an enhancement allowing users to per- form three types of uncertainty analyses: sensitivity analysis, first order error analysis, and Monte Carlo simulation (Brown and Barnwell 1987). QUAL2E, which can be operated either as a steady-state or as a dynamic model, assumes that the major transport mechanisms, advection and dispersion, are significant only along the main direction of flow (lon- gitudinal axis of the stream or canal). It allows for multiple waste discharges, withdrawals, tributary flows, and incremental inflow and outflow. It also has the capability to compute required dilution flows for flow augmentation to meet any pre-specified dissolved oxygen level.

Limitations of Existing Models Saunders and Maidment (1996) performed a GIS assessment of non- point source pollution in the San Antonio–Nueces coastal basin in the U.S.A. They evaluated four constituents: phosphorus, nitrogen, cadmium, 82 PARK AND HUCK and fecal coliforms. Predicted concentrations for phosphorus and nitrogen from the simulated point sources matched closely with average observed concentrations in the basin. Predicted fecal coliform concentrations, however, did not match well with average observed values. This may be because this model was not intended for fecal coliforms. Watershed level modelling is complex and involves various parameters. Most of the existing models predict common processes such as runoff estimation, soil detachment and water channel transport. Although these processes also affect Cryptosporidium oocyst transport in a watershed, existing models were developed to deal with sediments, nutrients and pesticides. There has been no model developed specifically for pathogens such as Cryptosporidium oocysts in a watershed. Existing models deal with sediments as a whole, which is insufficient for modelling oocyst transport. With these models, it is not possible to distinguish oocysts from other transported sediment particles of the same size. Although some existing models do incorporate bacteria, these models do not include aspects that are important regarding Cryptosporidium, such as the distinction between free and floc-associated oocysts. It is therefore important to consider oocyst transport at the conceptual level, because there are some fundamental differences from the transport processes for other contaminants in a watershed.

Approach

Because of the difficulties identified in applying existing models to describe Cryptosporidium oocyst transport in a watershed, a conceptual modelling approach based on first principles was adopted. The conceptual model assumes oocysts to behave as particles, but which exhibit die-off. Although to some extent the relationships are similar to those in sediment transport models such as AGNPS and ANSWERS, as noted previously those models deal with sediments as a whole, and do not allow oocysts to be distinguished from other transported sediment particles of the same size. The following subsections describe the various mechanisms included in the conceptual model.

Transfer from Solids to Water In overland flow, oocysts can be transported either as free-floating, or as floc-incorporated or particle-associated entities. The ratio of free-floating to floc-incorporated oocysts in runoff is a function of several factors. Manure-incorporated oocysts are not likely to be chemically bonded to the other fecal material, but rather are physically attached or trapped. Being of a size between that of clay and silt and having nearly neutral buoyancy (using textural classifications presented by Brady [1991]), oocysts would be entrained easily and suspended in runoff water. However, they may have a high affinity for mineral and organic fractions MODEL FOR CRYPTOSPORIDIUM TRANSPORT IN WATERSHEDS 83

of the soil, effectively changing their diameter and mass for transport purposes. The threshold of kinetic energy needed to entrain and transport a particle in flowing water is a function of the particle’s effective diameter and mass. Oocysts may be present in waste deposited on the surface directly from animal droppings, or from manure applied on the top of the soil. Oocysts may also be in the soil, because some manure is incorporated into the soil when it is applied. In this case, oocysts could penetrate deeper into the soil due to rainfall and become less likely to be transported into surface runoff. Because of an oocyst’s relatively large size compared to bacteria, it is much more difficult for oocysts to move through soil or into groundwater than bacteria unless there are fractures in the subsurface or extensive tile drainage. Thus, Couillard and Li (1993) showed that applying manure in the soil can improve water quality, compared to surface application. For modelling purposes, it is considered reasonable to neglect transport through the subsurface (unless there is fracturing or extensive tile drainage), for oocysts applied in the soil.

Quantification of Detachment Runoff begins to generate shear forces upon oocysts in the soil when rainfall rate exceeds the infiltration rate. This shear force can differ from that produced by the slope of the terrain. Bacterial detachment is commonly modelled as a kinetic process, and solely related to the sorbed concentration of bacteria (Bales et al. 1991; Harvey and Garabedian 1991; Lindqvist et al. 1994). Experimental observations, however, have demonstrated that detachment also is a function of bacterial residence time and is not solely dependent upon attached concentration (Escher 1986, as cited by Escher and Characklis 1990). For the model developed in this paper, oocyst detachment is assumed to be analogous to bacterial detachment. Bacterial detachment as a function of time has been described by a single-term exponential decay model (Escher 1986). Using a similar approach, oocyst detachment can be described as follows:

N -wt = e t < t N i 0 (1)

N=const.t ≥ ti

where N is the number of oocysts remaining in the waste at runoff time t (T), N0 is the initial number of oocysts in the waste at the time of initial runoff, and w is an exponential desorption rate constant (T-1). N is set to a constant after a specified time, ti (T), in order to provide for a fraction of oocysts that may be considered irreversibly attached after that time. Higher values of w increase the rate of detachment. 84 PARK AND HUCK

A two-rate desorption model describes the phenomenon that a large number of oocysts are quickly released initially, followed by a slower release. The model is given as:

N -kt -k t = Ae 1 +(1- A)e 2 (2) N 0

-1 where k1 and k2 are the fast and slow desorption rate coefficients (T ), respectively, and A is a weighting factor. For high values of A, the shift in domination from fast desorption to slow desorption occurs at a high value of N/N0. For lower values of A, this shift occurs at a lower value of N/N0.

Settling Rate As noted previously, the effective settling rate of an oocyst is influenced strongly by whether it is present in a free state or floc-associated. According to Droppo and Ongley (1993), the average floc size of the Grand River in Ontario is 9.1 µm. Even though the characteristics of overland flow are different from those of the Grand River, in the absence of better information, this diameter will be used here for floc-associated oocysts. For purposes of the present work the exact floc size does not have a large impact on the results; the fraction of free versus floc-associated oocysts is far more important.

The settling of dilute suspensions in water follows Stokes’ law, provided the flow is laminar and minimal interaction between particles is assumed. Stokes’ equation is given by: g V = (r - r)d2 (3) s 18m s

-1 where Vs is settling velocity (LT ), g is acceleration due to gravity -2 -1 –1 (LT ), µ is dynamic viscosity (ML T ), ρ s is mass density of the particle (ML-3), ρ is mass density of water (ML-3), d is diameter of the particle (L), and where L is length and M is mass. Assuming that most overland flow is laminar (Harry 1970), Stokes’ law can be used to estimate oocyst settling rates in this flow. The calculated oocyst settling velocity therefore would be 1.5 to 3.5 mm/h for a size range 4 to 6 µm and density of 1.05 g/cm3 (size and density values are taken from Medema et al. [1998]). In contrast, for a size of 9.1 µm (the Grand River average floc size) and assumed density of 1.5 g/cm3, the floc settling rate would be 77 mm/h (the density 1.5 is based on a sediment density of 2.6 g/cm3 [clay] and an oocyst density of 1.05 g/cm3. A more accurate floc density would not affect the point under discussion). The settling velocity of floc-associated oocysts thus estimated by Stokes’ law is more then an order of magnitude higher than that of free oocysts. This MODEL FOR CRYPTOSPORIDIUM TRANSPORT IN WATERSHEDS 85

is due to both the larger diameter and higher density of the floc-associated oocysts. Therefore, based on theoretical calculations, more floc-incorporated oocysts would settle out than free-floating oocysts, because of the generally shallow depth of runoff. Although scouring of settled oocysts will occur, it is possible that several storms may be required for floc-incorporated oocysts to reach a watercourse. Depending on the die-off rate of oocysts, only a small fraction of floc-associated oocysts may be viable if they reach a water treatment plant as a result of a number of storm events. This would be particularly true if these storm events are infrequent. However, floc-associated oocysts near a water body should be considered in calculations, especially when a high number of such oocysts may be present.

Overland Transport If the rate of oocyst transfer from solids to runoff water is obtained or calculated by the equations presented earlier, the flow velocity is determined by Manning’s equation, and the number of free-floating and floc- associated oocysts present on areas near a water body are estimated, the oocyst transportation rate can be written as:

N free = (C free )Q (4)

N floc = (C floc)Q (5)

Nt = (+C free)Q (C floc)Q (6)

where Nfree is number of free-floating oocysts transferred per unit -1 time (T ), Cfree is concentration of free-floating oocysts (number of free- -3 floating oocysts/volume of runoff water) (L ), Nfloc is number of floc- -1 associated oocysts transferred per unit time (T ), Cfloc is concentration of floc-associated oocysts (number of particulate-associated oocysts/volume of runoff water) (L-3), Q is overland flow rate (volume of runoff 3 -1 water per unit time) (L T ), and Nt is total number of oocysts transferred per unit time (T-1).

Water Channel Transport Water channel transport is based on the advection-dispersion equation. Therefore, the governing equation describing water channel transport of oocysts (including a die-off function) is:

∂C ∂2C ∂C = D – ν – kC + S(x, t) (7) ∂t ∂x2 ∂x 86 PARK AND HUCK

where C is the concentration of oocysts (number of oocysts/unit volume) (L-3), D is the dispersion coefficient for the oocysts (L2T-1), v is the flow velocity (LT-1), k is the die-off coefficient (T-1), and S is the source function (L-3T-1), which can be a point or non-point source in each stream. Walker et al. (1998) reported that a contamination event represents a short-duration pulse of oocysts in raw water supply systems that is likely to be related directly to a hydrologic event. Since oocyst transport in a watershed is driven largely by hydrologic events, the transport must be represented by a dynamic (transient) state model. Therefore the left-hand side of the equation is not zero.

Die-off (Viability) Many pathogenic microorganisms that occur in animal wastes survive well in soils but are quickly inactivated in natural waters (Maas et al. 1987). Sattar et al. (1999) investigated various factors affecting inactivation. Investigations of the survival of C. parvum in soil and water show the influence of the type and intensity of stresses encountered in each environment. Under some conditions, loss of viability can be very slow. Robertson et al. (1992) did experiments involving oocyst immersion in river water, which indicated that half the initial number of oocysts would remain infective after ~30 days (estimated by assuming first-order decay). Walker et al. (1998) summarized oocyst die-off for incubation experiments in feces. Jenkins et al. (1997) also did experiments and gathered information about inactivation of C. parvum oocysts stored in fecal pools or water at various temperatures. It is very difficult to apply a die-off function to each stage that oocysts go through as a result of hydrologic events. Aside from site-specific sampling and analysis, there is no way to know the viability of oocysts on a given field. However, information on oocyst viability in both feces and natural water can be used to model oocyst die-off, and assumptions can be made regarding the viability of oocysts found on a field. By assuming that oocyst die-off is a first-order process, oocyst transport in a water channel can be modelled by adding a first-order decay term to the advection-dispersion equation. Because the decay rate constant would be dependent on factors such as temperature, at a later stage of model development a formal means of including such factors should be incorporated.

Oocyst Behaviour in Lakes Lakes and reservoirs are natural coagulation and sedimentation basins. With hydraulic residence times often ranging from months to years, considerable sedimentation can occur. Natural aggregation increases particles sizes and thus settling velocities, accelerates particle removal to bottom sediments, and decreases the concentrations of particles in the water column (Weilenmann et al. 1989). Two important fac- MODEL FOR CRYPTOSPORIDIUM TRANSPORT IN WATERSHEDS 87

tors affecting sedimentation in a lake or reservoir are: (i) the hydraulic loading or overflow rate and (ii) the colloidal stability or sticking factor of the particles. Two simple models that can be used to approximate the behaviour of oocysts in lakes and reservoirs are the continuous stirred tank reactor (CSTR) model and the ideal rectangular settling tank (IRST) model.

Continuous stirred tank reactor A CSTR model by nature implies mixing. It is useful in approximat- ing the bulk water behaviour of small lakes and reservoirs, especially where the inflow and outflow are relatively large in relation to the volume. Oocysts in a lake or reservoir will settle relative to the bulk water at a velocity that can be calculated by Stokes’ law. Even though there is mixing in a lake, it is a relatively stationary water body in comparison to a river. It is reasonable to assume that there is a laminar flow, unless wind or other disturbances are present. Therefore, the settling of oocysts as a primary loss mechanism in a lake should be incorporated into the CSTR model. The laminar flow assumption may imply a slight departure from the perfect mixing assumed for an ideal CSTR. The governing mass balance equation for a lake modelled as a CSTR, including a settling term, is given by:

dClake in " = Q C -Q C - v A C - k"C (8) Lake dt in out lake s s lake lake

3 3 -1 where " Llake is the volume of lake (L ), Qin is the inflow (L T ), Qout is 3 -1 the outflow (L T ) (Qin and Qout are assumed to be equal), vs is the settling -1 2 velocity of the oocysts (LT ), As is the area of the lake (L ), k is the first- order oocyst die-off coefficient (T-1), Cin is the inflow oocyst concentration -3 -3 (L ), and Clake is the oocyst concentration (L ), both in the outflow and in the lake itself. Assuming complete mixing, equation 8 can be rewritten as: dC Q +lC = C in (9) dt "

where C is the concentration of oocysts in the lake (and in the outflow), Q is the input and output flow, Cin is the input concentration of oocysts, and λ can be derived (Chapra 1997) from equation 8.

Q + vsAs+ k" l = (10) "

Equation 9 shows that the output oocyst concentration from a lake is a function of the input concentration. The complete mixing assumption of the CSTR model implies that, regardless of lake size or retention time, oocysts will start to appear in the outflow as soon as they enter the lake. 88 PARK AND HUCK

Since it is a transient state model, the CSTR model is preferable to the IRST model discussed below for modelling dynamically changing oocyst inputs into a lake. The oocyst residence time (τ) is affected by die-off and settling in addition to the rate of flow. Chapra (1997) defined a contaminant residence time as: 1 t = (11) l

Ideal rectangular settling tank The theory of an ideal rectangular settling tank (e.g., Reynolds 1982) is well known with respect to discrete settling. In theory, one can -1 select a terminal (settling) velocity, Vs (LT ), and design the settling basin such that all particles with a terminal velocity greater than Vs will be removed. The overflow rate is given by: Q V=s (12) As

3 -1 where Q represents both inflow and outflow (L T ), and As is the surface area of the basin (L2). The overflow rate, VS, is equivalent to the settling velocity of the smallest particle that is 100% removed. If the settling velocity, Voocyst, is less than Vs, the fraction removed, Roocyst, is equal to Voocyst/Vs. Thus,

Voocyst R=oocyst (13) Vs

As an initial approximation, oocyst sedimentation in a lake or reservoir can be estimated using an IRST approach, by using the settling velocities of free and floc-associated oocysts, and the dimensions and inflow/outflow (Q) of the lake. An IRST lake model would also contain an oocyst die-off term.

Comparison of CSTR and IRST approaches For the same constant input concentration, the outlet concentration of a CSTR at steady state would be the same as for an IRST, neglecting minor differences in die-off between the two approaches. Due to the mixing assumptions in the CSTR model, it is difficult to use it to represent a lake with a long retention time. However, for a small lake or reservoir, it is likely that the CSTR model would better explain the dynamics than would the IRST model. The IRST model would give a higher outlet concentration following a residence time corresponding to that for an input to move through the lake in plug flow. Therefore, it is likely appropriate to use the IRST approach to provide a worst case estimate. MODEL FOR CRYPTOSPORIDIUM TRANSPORT IN WATERSHEDS 89

The Conceptual Cryptosporidium Transport Model

Overview The watershed-level conceptual Cryptosporidium transport model (Fig. 1) includes the following mechanisms: oocyst detachment, runoff, overland transport, lake transport, water channel transport and stream confluence (the meeting of two streams). By estimating the impact on oocyst concentration of each mechanism or step, the model helps to determine their relative importance. Following a description of the sample watershed, this section addresses the sequential transport of oocysts initi- ated by a storm event: runoff assessment, overland velocity calculation, point and non-point source calculation, discharge time control of a non- point source in overland transport, and finally travel distance control of a non-point source in overland transport. The equations used in the model are given in Table 1. The assumptions that are made because no data are available are discussed and are summarized in Table 2. In the future, further development of the model would allow elimination of some of the simplifying assumptions that are currently required.

Sample or Hypothetical Watershed In order to proceed with the model, a sample or hypothetical watershed was necessary since there were no suitable data available from a real watershed for use in this research. The hypothetical watershed includes both a point source and a non-point source of oocysts, a lake and a stream confluence (Fig. 2). The point source is the discharge from a wastewater treatment plant and the non-point source is an area to which manure has been applied. For the non-point source a rainfall event and the ensuing runoff are the causes of oocyst delivery to a water channel. Then, the water flows through the channel to the intake of the water treatment plant. A lake is located in one of the water channels to illustrate its effect on oocyst concentrations. There is also a stream confluence. Detailed

Fig. 1. Transport steps or mechanisms included in the conceptual model. 90 PARK AND HUCK

Table 1. Equations used in the conceptual model

Mechanism Equation Remarks

g 2 Settling rate V = (r - r)d Stoke’s law s 18m s

Runoff calculation SCS Curve Procedure SCS

1 2 1 Overland transportV= R 3S 2 Manning’s equation n

N -wt = e , t < ti N 0 Single exponential function N=const. t ≥ ti Detachment

N -kt -k t = Ae 1 +(1- A)e 2 Two-term exponential N 0 model

10% detachment Simple assumption

W S-(x,t)=d(x- a)[H (t)- H (t t )] A disch arg e Point source Source calculation cross W S-(x,t)= [H (x-a)-H (x-b)][H (t)- H (t t )] discharge Non-point source A cross

Oocyst fluxW=(tt) Q (t)C () Mass balance

∂ N ∂2N ∂N Water channel = D – ν – kN + S(x, t) The advection transport ∂t ∂x2 ∂x dispersion equation

N=Q + N Q N Q Advective mass Stream confluence 3 3 4 4 5 5 Q=3 + Q 4 Q 5 balance

dC " lake = Q C in - Q C - v A C - k"C Lake dt in lake out lake s s lake lake CSTR Lake transport Vs R=removal IRST Vo MODEL FOR CRYPTOSPORIDIUM TRANSPORT IN WATERSHEDS 91

Table 2. Model assumptions

• Overland flow and the flow discharged by a point source do not affect the flow in the water channel. • There is no groundwater flow into the water channel. • There is no evaporation in the watershed. • The water channel cross-sectional area where a point or non-point source discharges is completely mixed (this assumption is required because the model is one-dimensional). • For a lake or reservoir the usual assumptions for a CSTR or IRST are applied, depending on which sub-model is chosen. • Two types of oocysts (free and floc-associated) exist in a watershed. • Both types of oocysts are assumed to behave as discrete particles during transport, including in lakes or reservoirs. Thus, there is no further aggregation or flocculation during oocyst transport. • The physical properties of oocysts do not change during transport. • At time zero the oocysts from all sources are assumed viable. • Overland flow is assumed to be laminar in all cases. • Oocyst transport in overland flow is assumed to be dominated by advection. • Overland transport begins with a one-time detachment of oocysts on a field. Consequently, the next runoff event does not entrain oocysts remaining on the top of the soil, i.e., those that have been only transported part way to the watercourse. Future incorporation of a re-entrainment function represents an area for model refinement. However the fact that considerable oocyst die-off could occur prior to the next runoff event would reduce the importance of re-entrainment. • Settling starting from the half depth of runoff is assumed for oocyst transport in overland flow. • Oocyst transport to a watercourse from non-point sources is due to overland flow (i.e., subsurface transport is neglected). In the absence of karst geology or extensive tile drainage, this is a reasonable assumption. • The hydraulic regimes in the lake and runoff are assumed to be favorable for discrete settling. • There is no oocyst removal by settling in the stream or river channel, because of the relatively short travel time, turbulent flow and relatively large depth.

modelling of the confluence itself is very complicated. However, since the model is designed to give the relative importance of the various parameters involved, a simple mass balance approach is used. This adequately represents concentrations downstream of the confluence once complete mixing of the two streams has occurred. The hypothetical watershed was created using a simplified part of the Grand River watershed, in Ontario, Canada. Therefore the scale, as well as the geometry of the lake and stream are similar to those in the 92 PARK AND HUCK

Fig. 2. The hypothetical watershed.

Grand River watershed. In Fig. 2, WWT represents a wastewater treatment plant and the components are not represented at actual scale.

Runoff Assessment The conceptual model employs the most commonly used method to estimate runoff, the U.S. Soil Conservation Service (SCS) curve number procedure (SCS 1972). Basically, the SCS method takes into account the infiltration capacity of a given type of soil, land use, and the soil water conditions at the start of a rainfall event (antecedent soil water condition).

Overland Velocity Calculation When rainfall generates runoff, overland flow velocity needs to be estimated to understand how and when oocysts on a field reach a water channel. The conceptual model uses Manning’s equation, which describes the effect of flow resistance caused by channel roughness:

1 2 1 V= R 3S 2 (14) n

where V is the average flow velocity in metres per second, n is Manning’s roughness coefficient, R is the hydraulic radius in metres, and S is the dimensionless energy slope. The hydraulic radius is defined as the flow cross-sectional area divided by the wetted perimeter. For a wide channel (width >20 times the depth), which would apply for overland flow, the hydraulic radius is approximately the same as the flow depth. MODEL FOR CRYPTOSPORIDIUM TRANSPORT IN WATERSHEDS 93

Source Calculation As mentioned previously, there are two types of source functions for oocysts in the sample watershed: point and non-point. A wastewater treatment plant discharge is used here as an example of a point source; overland runoff and areas to which manure has been applied as non- point sources. In the conceptual model, a point source is represented by a dirac delta function while a non-point source is represented by a line load function.

Discharge Time Control for a Non-point Source in Overland Transport Overland flow velocity is one of the important factors used to estimate non-point source discharge time. Since overland transport is assumed to be caused mainly by the advective term, the velocity at which oocysts are transported to the water channel can be estimated by the overland flow velocity. An important variable is the duration of runoff, which is not simple to calculate, since it is a function of land use, soil type, storm event, etc. For the sample watershed case, the duration of runoff is assumed to be one hour. (A model sensitivity analysis is provided later in this paper.) The distance of a manure-applied area from the water channel affects the discharge time of that non-point source. For example, if it takes 50 min for oocysts to reach the water channel, then the oocyst discharge time is 10 min; i.e., the total runoff time (60 min) minus the oocyst travel time (50 min). Currently, the conceptual model starts with the oocyst concentration existing on the top of the manure-applied area at the beginning of runoff. At this stage of development the model does not include a function to describe oocysts reaching the manure-applied area as a result of their detachment into the runoff water somewhere up-gradient. The time during which runoff water passes over the manure-applied area can be shorter than the discharge time, if the length of the manure- applied area in the direction of flow is small. Suppose, continuing the example above, that it only takes the runoff three minutes to pass through the manure-applied area, then, there is only a three-minute discharge of oocysts, even though the difference between the runoff time and the travel time would allow 10 min. Fig. 3 illustrates the two different cases for what determines the non-point source discharge time. As a check, the conceptual model calculates the actual non-point source discharge time and validates it by ensuring that the number of oocysts arriving at the intake of the water treatment plant is not greater than the number discharged into the water channel.

Travel Distance Control for a Non-point Source in Overland Transport As discussed previously, either free or floc- or particle-associated oocysts can be present in water. The ratio of free to floc-associated oocysts can be a function of various factors. The ratio used in this research (1:3) is taken from Medema et al. (1998), in the absence of other information. In 94 PARK AND HUCK

Fig. 3. Cases for determining non-point source discharge time. practice, this ratio will depend on soil type, the concentration of particulate organic matter, and different environmental conditions such as temperature, pH, suspended solids concentration, etc. Experimental determination of this ratio under a range of conditions would be important for detailed modelling. Settling velocity is one of the most important physical properties of oocysts with respect to transport. Settling rate eventually limits both free and floc-associated oocyst travel distances. Because of the difference in settling rates, the travel distances of both free and floc-associated oocysts during runoff are different (Fig. 4) and should be calculated dynamically in each storm event. In the present model, it is assumed that oocysts start settling from the half-depth of runoff after detachment from a field. The estimated travel distances of both types of oocysts are taken into account in determining a line load of non-point oocysts to a water channel.

Fig. 4. The difference in travel distance for free and floc-associated oocysts. MODEL FOR CRYPTOSPORIDIUM TRANSPORT IN WATERSHEDS 95

Mathematical Description of the Model

Hypothetical Watershed Due to the difficulty involved in identifying the non-point source function for oocyst generation in some cases and the fact that no analytical solution exists for some stages of oocyst transport, incorporating all mechanisms of the conceptual model into a simulation program is the most con- venient way of investigating the relative importance of the parameters involved in oocyst transport. Several numerical techniques, such as finite difference or finite element methods, could be used. In this research, the finite difference form of the model was developed. Further details concerning the model, including the source code, can be found in Park (1999). As discussed previously, the transport of free and floc-associated oocysts in some situations is different because of the difference in their settling velocities. Because of this difference, free and floc-associated oocyst transport modelling is handled separately in the following parts of the overall transport sequence: overland transport, lake sedimentation, and water channel transport upstream of a lake. In these components, oocyst settling velocities are important. For transport purposes, the hypothetical watershed is divided into five sections: a point source upstream of a lake, the lake or reservoir itself, a stream channel from the outlet of the lake to a stream confluence, a tributary having a non-point source, and the stream section between the confluence and a water treatment plant intake. The criteria for defining the five sections are the different mechanisms involved in each section and relate mainly to water channel transport. Figure 5 shows the finite difference scheme for the five sections of the hypothetical watershed. The parameter settings used for the various sections in the finite difference scheme are given in Table 3. The mathematical details of the finite difference schemes for each of the sections, including the appropriate boundary conditions, are given in Appendix A. Comments related to the mathematical form of the model and the finite difference scheme for each section are provided below.

Point Source Upstream of Lake (Section 1)

Free and floc-associated oocyst modelling are performed indepen- dently for this section because of the lake downstream and the importance of sedimentation in the lake. In this first section, the initial condition is set to zero in the river. In other words, it is assumed that there are no oocysts in the water before a rainfall event. Setting the initial condition to zero is standard practice for this type of modelling and simplifies the work. Because in reality there is a continuous point source discharge (the wastewater treatment plant) in Section 1, the early modelled oocyst concentrations in this section will be inaccurate, except for the case when a WWTP is being initially brought on stream. 96 PARK AND HUCK

Fig. 5. Finite difference scheme for the five sections modelled of the hypothetical watershed.

Lake (Section 2) As discussed previously, oocyst transport in a lake can be modelled to a first approximation by considering the lake either as a continuous stirred tank reactor (CSTR) or an ideal rectangular setting tank (IRST). Implementation of the numerical solutions for both a CSTR and an IRST is discussed in Appendix A. Because of the importance of the settling velocity for oocyst behaviour in a lake, free and floc-associated oocysts should be modelled separately. The governing equation for a lake modelled as a CSTR was given previously (equation 8). The CSTR model is easily solved for certain simple loads. For the case when the loading is time dependent a simple numerical approach is adopted. Euler’s method is the simplest numerical method for solving ordinary differential equations. Appendix A shows the application of Euler’s method for solving the completely mixed lake model (Chapra 1997). The CSTR model can be solved with this method, even though the upstream oocyst source function may be very dynamic. This is one of the reasons for using a numerical approach for the model, because oocyst loadings may be difficult to represent accurately as pulse, step, linear, exponential, or sinusoidal functions. MODEL FOR CRYPTOSPORIDIUM TRANSPORT IN WATERSHEDS 97

Table 3. Parameter settings for the finite difference methoda

Parameter Valueb Notes

Length (m) 15,000 21,000 in Section 4 No. cells (nn) 10,000 Flow velocity (m/s) 1 Die-off coefficient (d-1) 0.004 Dispersion coefficient (cm2/s) 1000 Distance step (∆ x) (m) 1.5 2.1 in Section 4 Time step (∆ t) (s) 1

aNotes: The point source in Section 1 (cell 2) discharges continuously; the non-point source in Section 4 (cell 1429 is the first discharge cell) discharges inter- mittently as discussed in the text, and the WTP intake is located 10,000 m downstream from the stream confluence. bThe same parameter setting is used in Sections 1, 3, 4 and 5 except as noted.

An IRST model assumes ideal behaviour, as its name implies. The assumptions required for an ideal setting tank are discussed elsewhere (e.g., Reynolds 1982). The removal rates of both free and floc-associated oocysts are calculated by equation 13. Thus, the concentration of oocysts leaving the lake is calculated from the oocyst input concentration multiplied by (1-Roocyst). If the lake is dynamic with respect to parameters such as flow, oocyst concentration, etc., the IRST model will be poor at representing the system because of the limitations imposed by its assumptions.

Between Lake and Stream Confluence (Section 3) The third section of the hypothetical watershed is similar to the first section, the difference being that there is no source term in the third section. Thus, the only oocyst source is the oocyst concentration leaving the lake. However, in this section, there is no need for separate modelling of free and floc-associated oocysts, because the transport mechanisms are not closely related to oocyst settling velocity.

Tributary with Non-point Source (Section 4) In Section 4, there is a runoff-generated non-point source of oocysts that is converted to a line source at the water channel, as discussed by Park (1999). The difference in water channel transport between this section and Section 1 is that the source is non-point, and the source discharge is not continuous. Non-point source discharge duration is related to fac- 98 PARK AND HUCK tors that affect overland oocyst transport, as described by Park (1999), and discussed later in this paper.

Stream Confluence A stream confluence would be difficult to model in detail. In particular, if the flow in one stream is considerably greater than in the other, the mixing pattern would be very complicated. In the current model, the stream confluence is handled using a simple mass balance. As discussed previously, this approach adequately represents concentrations downstream of the confluence once complete mixing of the two streams has occurred. For the model in its present form, complete mixing of the two streams is assumed to occur instantaneously.

Between Stream Confluence and WTP (Section 5) In this section only in-stream transport of oocysts occurs. As in Section 2, there is no source of oocysts in Section 5, however there is a water treatment plant intake. The oocyst concentration at the intake is the object of the modelling.

Model Sensitivity Analysis

At its present state of development, the model cannot be used to obtain quantitative results. However, it can be used to provide an indication of the relative importance of various factors for oocyst transport. This section addresses this by means of a sensitivity analysis.

Approach Since a data set for the model was not available, required values were either obtained from the literature or assigned reasonable values in cases where no data were available or where a number of different values could be used, such as the location of the manure-applied area. It is important that the values assumed for the base setting are reasonable, because in a complex model one factor may be critical when other factors are within a certain range. The assumed factors were included in the sensitivity analysis to evaluate their importance. Table 4 shows the base settings for the parameters. Although the hypothetical watershed contains both point and non- point sources, their relative impact on the oocyst concentration at the drinking water intake in Section 5 depends on the assumed sizes of the two sources. Therefore, two sensitivity analyses are performed: a point source analysis (involving Sections 1, 2, 3 and 5 of the hypothetical watershed) and a non-point source analysis (involving Sections 4 and 5). Due to the different units of the various model parameters, their relative impact on the state variable (oocyst concentration) can only be established using a normalized sensitivity analysis. The basic equation for nor- malization is given by MODEL FOR CRYPTOSPORIDIUM TRANSPORT IN WATERSHEDS 99

Sn = ∂φ . αi (15) ∂αi φ

where Sn is the normalized sensitivity coefficient, φ is the state vari- ∂φ able, and αi is the model parameter being investigated. For the term, ∂αi a first-backwards approximation or the central difference form can be used. In this analysis, the central difference form was used.

The central difference form is given by:

∂φ φ(α + ∆α ) – φ(α – ∆α ) ≈ i i i i (16) ∂αi 2∆αi

where ∆αi is the difference in a perturbed parameter.

A positive sign indicates that an increase in αi will result in an increase in φ (the state variable). A ∆ value of 1% and was used for the analysis. Park (1999) also used a ∆ value of 2%, and found that the trends were the same as at 1%.

Results Point source oocysts The results for the normalized sensitivity analysis for point source oocysts are shown in Fig. 6. It will be recalled that in the hypothetical watershed a lake is located downstream of the point source. From Fig. 6, it can be seen that the most important factors are the concentration of oocysts in the point source discharge (Opoint1), and the human population contributing to that discharge (popu1). As would be expected, both of these factors are of equal importance, because each has the same impact on numbers of oocysts discharged. For the base settings of the parameters used in this sensitivity analysis, an equal percentage change in either the oocyst die-off rate (dieoff1) or the lake retention time (retime1), has a much lower impact on oocyst concentration.

Non-point source oocysts In the model, non-point source oocysts are generated by runoff. Therefore, the parameters examined in the sensitivity analysis are related to the original source of the oocysts, the detachment process and runoff itself. The results (Fig. 7) show that the most critical factor is the amount of rainfall. The next section discusses the results of the sensitivity analysis for point and non-point sources in further detail. 100 PARK AND HUCK

Table 4. Base settings used in the model sensitivity analysis.

Parameter valuea Comments

Point source Oocyst concentration 5 oocyst/L Based on Rose (1988) Population 80,000 people Approximate population of City of Waterloo Pipe diameter 1.5 m Assumed Advection – dispersion equation D (dispersion coefficient) 1000 cm2/s Contaminant transport in surface water k (decay rate coefficient) 0.012 oocysts/day Robertson et al. (1992) River velocity 1 m/s High flow in the Grand River Water channel Cross-sectional area 200 m2 Assumed of channel Ratio of free to floc- 1:3 Medema et al. (1998) associated oocysts Detachment and any loss 10% Assumed Non-point source Number of oocysts 4.3 × 1010 Zhang’s (1999) estimation on a field based on Hansen and Ongerth (1991) Runoff Precipitation 1 inch (25 mm) Assumed AMC II Assumed Land use Pasture or range, Assumed good condition CN (curve number) 80 Calculated or read off from the graph – SCS (1972) Manning’s equation n (roughness coefficient) 0.03 Taken from land use Slope 0.02 Assumed Runoff time 1 hour Assumed Overland flow Size of manure- 500 m × 500 m Assumed applied area Distance from water 100 m Assumed channel

a Values obtained from the literature or assumed. MODEL FOR CRYPTOSPORIDIUM TRANSPORT IN WATERSHEDS 101

Fig. 6. Normalized sensitivity coefficients for point source parameters (1% perturbation). Opoint1 is the concentration of oocysts in the point source discharge; popu1 is the human population contributing to that discharge; dieoff1 is oocyst die-off rate and retime1 is the lake retention time.

Discussion An important aspect of the conceptual model is the division of oocysts into free and floc-associated. This leads to a difference in settling velocities, as discussed previously, which has implications for the overland transports of oocysts. The approach to transport by runoff in the current model is different from that of Walker et al. (1999). In their fate and transport model for Cryptosporidium, they assumed that oocysts and Normalized Sensitivity Coefficient

Fig. 7. Normalized sensitivity coefficients for non-point source parameters (1% perturbation). Amount of rainfall is rainfall1; number of oocysts on a field is Onon1; detachment ratio to runoff is Rde1; river velocity is river_v1; land slope is slope1; and distance to watercourse is distance1. 102 PARK AND HUCK manure moved together. The model described in the present paper leads to greater oocyst transmission to a watercourse because free oocysts detached into water can move along with overland flow. This means that runoff water with very low turbidity can carry oocysts to a watercourse. This transmission can occur because of the relatively small size of oocysts and the fact that their density is close to that of water (1.05 g/cm3 [Medema et al. 1998] versus an average of about 2.7 g/cm3 for clay sediment). The importance of the factors related to overland transport is an additional reason why the sensitivity analysis for point and non-point source oocysts needs to be performed separately.

Point source oocysts The order of importance of factors for point source oocyst transport as estimated by normalized sensitivity analysis (Fig. 6) was as follows:

Oocyst concentration = population >> lake retention time = die-off

For a point source oocyst discharge (i.e., a wastewater treatment plant), the number of oocysts discharged per unit time is proportional to both the human population served by the treatment plant and to the oocyst concentration (number per unit volume) in the discharge. Thus, the normalized sensitivity coefficients of the two factors are the same. This result would be essentially the same if the point source were a combined sewer outfall. Since the population served by a WWTP cannot normally be reduced, the sensitivity analysis suggests, for situations where the parameter values are similar to the base settings used herein, that point source contributions to oocyst levels at a water treatment plant intake could be most effectively mitigated by maximizing the removal or inactivation of oocysts in upstream wastewater treatment plants. This would normally imply a wastewater disinfection step capable of oocyst inactivation, and would be an argument for the use of a technology such as UV instead of chlorination for wastewater disinfection. Normally, the retention time of a lake may vary from several days to more than a year. In the sample watershed, the retention time of the lake is 4.3 days. Oocyst removal by sedimentation was estimated at 5.3% and total oocyst removal, which includes die-off during the residence time, was 10.1%. If the length and width of the lake are each increased to four times the original, the residence time becomes about 70 days. In this case, 57% of the incoming oocysts would be removed by sedimentation and the total oocyst removal with die-off becomes 81%, which represents a dramatic increase. Even though lake residence time and the die-off rate were not the most significant factors for point source oocyst transport based on the normalized sensitivity analysis, lakes or reservoirs with a long residence time should act as a barrier that substantially reduces numbers of oocysts arriving at downstream water treatment plants. Walker et al. (1999) performed Cryptosporidium modelling for the New York City water supply system, which is dominated by several large MODEL FOR CRYPTOSPORIDIUM TRANSPORT IN WATERSHEDS 103 reservoirs with residence times on the order of many months to years. Before the present conceptual model could be applied to such watersheds, the submodels for oocyst fate and transport in lakes would need to be more fully developed because of the dominant role that such water bodies play in systems such as the one for New York City.

Non-point source oocysts Figure 7 ranked the relative importance of various factors affecting in- stream oocyst concentrations derived from the non-point source as follows:

Amount of rainfall (rainfall1) > number of oocysts on a field (Onon1) = detachment ratio to runoff (Rde1) = river velocity (river_v1) > land slope (slope1) > distance to watercourse (distance1).

This order could vary somewhat if different base settings of the parameters were used. For example, for the given storm event, no oocysts would reach the watercourse from the manure-applied area if its distance from the watercourse were increased by more than 30%. Because the travel distance of free and floc-associated oocysts in overland flow is different, only free oocysts can reach the watercourse when the distance from the manure-applied area is above a certain value. Thus, the distance parameter functions as a switch. It turns on or off the oocyst discharge (free, or both types or neither) depending on the distance for a given intensity of storm and given physical conditions such as land slope. Figure 7 indicates that the most critical factor for non-point source oocyst transport is the amount of rainfall, which is of course beyond human control. The second most important factors are the number of oocysts on a field, the detachment ratio and the river velocity. Controlling the river velocity is normally not practical. Similarly, land slope is essentially a noncontrollable factor. However, oocyst generation from a field can be controlled by modifying farming practices. For example, the detachment ratio used assumes that oocysts are exposed on top of the soil. By applying manure in the soil, this detachment ratio can be dramatically reduced. Applying manure in areas away from a watercourse is also a way to minimize the number of non-point source oocysts reaching that watercourse. Theoretically, runoff can be reduced by applying manure to areas with different antecedent land use, but this may be difficult in practice.

Summary

This paper describes the formulation of a model for Cryptosporidium oocyst transport in watersheds. The ultimate objective of the modelling is to provide order-of-magnitude estimates of oocyst concentrations at the intake of a water treatment plant. Estimates to this level of accuracy are useful and appropriate for treatment plant design. The model includes the 104 PARK AND HUCK generation of oocysts from both point and non-point sources, the overland transport of oocysts from a non-point source to a watercourse, in- stream oocyst transport and transport through a lake or reservoir. The model also incorporates the confluence of two streams, and includes a function for oocyst die-off. The finite difference scheme of the model was developed for five sections in a hypothetical watershed: 1. A point source (WWTP discharge); 2. A lake or reservoir, which can be modelled either as a continuous stirred tank reactor, or as an ideal rectangular sampling tank; 3. The section of stream channel leading from the outlet of the lake or reservoir to the confluence with another stream; 4. A tributary stream with a non-point source; and 5.The stream section from the confluence to a water treatment plant intake. The criteria for defining each of the five sections were the different transport mechanisms involved. The stream confluence is handled by a simple mass and flow balance. A key element and contribution of the model is the division of oocysts into two categories: free and floc-associated. This distinction is important with regard to overland transport. Data to test the model are not currently available. Therefore the relative importance of the various factors was tested for the hypothetical watershed. Reasonable base settings for each parameter were used and a normalized sensitivity analysis (in which each factor was varied by the same percentage amount) was performed. Separate sensitivity analyses were performed for the point and non-point sources. For the point source, the most important factors were the concentration of oocysts in the discharge and population contributing to that discharge. For the non-point source, the most important factor was the amount of rainfall. There were differences in importance among the other factors, however their relative importance is likely to be situation-specific. A practical conclusion that can be drawn from the sensitivity analysis is that wastewater disinfection capable of effectively inactivating oocysts may be important. The major contribution of the model is in the definition of the mechanisms involved in oocyst transport within a watershed. At its present stage of development, the model cannot be used to obtain quantitative results. However, it does give insights into the relative importance of various factors. It can also be applied to other pathogens such as Giardia. It provides a framework for data collection and identifies areas where experimental investigations are required to avoid or minimize the need for simplifying assumptions. The model provides a base for further modelling. Ultimately, further more detailed modelling could be expected to help development of management practices to minimize the contributions of both point and non-point sources to pathogen loadings at drinking water intakes. MODEL FOR CRYPTOSPORIDIUM TRANSPORT IN WATERSHEDS 105

Acknowledgements

We acknowledge the financial support of the Natural Sciences and Engineering Research Council of Canada and the assistance of Angela Semple, Ada Chavez and Sarah Dorner in completion of the manuscript.

References

Arnold JG, Engel BA, Srinivasan R. 1993. A continuous time, grid cell watershed model. In Proceedings of the application of advanced information technologies: effective management of natural resources conference. American Society of Agricultural Engineers, St. Joseph, Mich. Atherholt TB, LeChevallier MW, Norton WD, Rosen JS. 1998. Effect of rainfall on Giardia and Cryptosporidium. J. AWWA 90(9):66–88. Bales RC, Hinkle SR, Kroeger TW, Stocking K. 1991. Bacteria-phase adsorption during transport through porous media: chemical perturbations and reversibility. Environ. Sci. Technol. 25:2088–2095. Beasley DB, Huggins LF. 1981. ANSWERS user’s manual. EPA–905/9-82-001, U.S. Environmental Protection Agency, Region V, Chicago, Illinois. Brady NC. 1991. The nature and property of soils. Macmillan, New York. Brown LC, Barnwell TO Jr. 1987. The enhanced stream water quality models QUAL2E and QUAL2E-UNCAS: documentation and users manual. U.S. Environmental Protection Agency, Athens, Georgia, Report EPA/600/3- 87/007. Chapra SC. 1997. Surface water-quality modelling. McGraw-Hill series in water resources and environmental engineering, p. 121–123. Couillard D, Li JF. 1993. Assessment of manure-application effects upon the runoff water quality by algal assays and chemical analyses. Environ. Pollut. 80:273–279. Droppo IG, Ongley ED. 1994. Flocculation of suspended sediment in rivers of southeastern Canada. Water Res. 28:1799–1809. Drozd C, Schwartzbrod J. 1996. Hydrophobic and electrostatic cell surface properties of Cryptosporidium parvum. Appl. Environ. Microbiol. 62:1227–1232. Engel BA, Srinivasan R, Arnold J, Rewerts C, Brown SJ. 1993. Nonpoint source (NPS) pollution modelling using models integrated with geographic information systems (GIS), Wat. Sci. & Technol. In Olem H (ed.), Diffuse pollution. Pergamon Press Ltd., Oxford, UK. Escher AR. 1986. Bacterial colonization of a smooth surface, an analysis with image analyzer. Ph.D. thesis, Montana State University, Bozeman. Escher AR, Characklis WG. 1990. Modelling the initial events in biofilm accumulation in biofilms, p. 445–486. Characklis WG, Marshall FC (ed.). John Wiley, New York. Gregory J. 1994. Cryptosporidium in water: treatment and monitoring methods. Filtrat. Separat. 31(3):283–289. Hansen JS, Ongerth JE. 1991. Effects of time and watershed characteristics on the concentration of Cryptosporidium oocysts in river water. Appl. Environ. Microbiol. 57:2790–2795. 106 PARK AND HUCK

Harry GW. 1970. The effect of raindrop impact and surface roughness on sheet flow. Research Report No. 34, University of Illinois Water Resources Center. Harvey RW, Garabedian SP. 1991. Use of colloid filtration theory in modelling movement of bacteria through a contaminated sandy aquifer. Environ. Sci. Technol. 23:178–185. Jenkins MB, Anguish LJ, Bowman DD, Walker MJ, Ghiorse WC. 1997. Assessment of a dye-permeability assay for determination of inactivation rates of Cryptosporidium parvum oocysts. Appl. Environ. Microbiol. 63: 3844–3850. Kelley MB, Warrier PK, Brokaw JK, Barrett KL, Komisar SJ. 1995. A study of two U.S. Army installation drinking water sources and treatment systems for the removal of Giardia and Cryptosporidium, p. 2197–2230. In Proc. AWWA Water Quality Technology Conference. Knisel W. 1980. CREAMS a field scale model for chemicals, runoff, and erosion from agricultural management systems. U.S. Department of Agriculture, Conservation research report No. 26. 640 p. LeChevallier MW, Norton WD, Lee RG. 1991. Occurrence of Giardia and Cryptosporidium sp. in surface water supplies. Appl. Environ. Microbiol. 57: 2610–2616. Lindqvist R, Cho JS, Enfield CG. 1994. A kinetic model for cell density dependent bacterial transport in porous media. Water Resour. Res. 30:3291–3299. Lisle JT, Rose JB. 1995. Cryptosporidium contamination of water in the USA and UK: a mini-review. J. Water SRT Aqua 44(3):103–117. Maas RP, Smolen MD, Jamieson CA, Weinberg AC. 1987. Setting priorities: the key to nonpoint source control. Office of Water Regulations and Standards, U.S. EPA, Washington, D.C. 50 p. MacKenzie WR, Hoxie NJ, Proctor ME, Gradus MS, Blair KA, Peterson DE, Kazmierczak JJ, Addiss DG, Fox KR, Rose JB, Davis JP. 1994. A massive outbreak in Milwaukee of Cryptosporidium infection transmitted through the public water supply. N. Eng. J. Med. 331(3):161–167. Medema GJ, Schets FM, Teunis PFM, Havelaar AH. 1998. Sedimentation of free and attached Cryptosporidium oocysts and Giardia cysts in water. Appl. Environ. Microbiol. 64:4460–4466. Ong C, Moorehead W, Ross A, Isaac-Renton J. 1996. Studies of Giardia spp. and Cryptosporidium spp. in two adjacent watersheds. Appl. Environ. Microbiol. 62:2798–2805. Park CH. 1999. Conceptual modelling of the transport of Cryptosporidium oocysts in a watershed. Master’s thesis, University of Waterloo. Reynolds TD. 1982. Unit operations and processes in environmental engineering. Wadsworth, Inc., p. 70–80. Robertson LJ, Campbell AT, Smith HV. 1992. Survival of Cryptosporidium parvum oocysts under various environmental stressors. Appl. Environ. Microbiol. 58:3494–3500. Rose JB. 1988. Occurrence and significance of Cryptosporidium in water. J. AWWA 80(2):53–58. Rose JB, Gerba CP, Jakubowski W. 1991. Survey of potable water supplies for Cryptosporidium and Giardia. Environ. Sci. Technol. 25:1393–1400. MODEL FOR CRYPTOSPORIDIUM TRANSPORT IN WATERSHEDS 107

Sattar SA, Chauret C, Springthorpe VS, Battigelli D, Abbaszadegan M, LeChevallier M. 1999. Giardia cyst and Crypotosporidium oocyst survival in watersheds and factors affecting inactivation. Report No. 90761. American Water Works Association (AWWA) and AWWA Research Foundation, Denver, Colo. Saunders WK, Maidment DR. 1996. A GIS assessment of nonpoint source pollution in the San Antonio – Nueces Coastal basin. Center for Research in Water Resources University of Texas online report 96–1, http://www.crwr. utexas.edu/repots/1996/rpt96-1.shtml (Oct. 8, 2002). Soil Conservation Service (SCS). 1972. National engineering handbook, Section 4, Hydrology. Stewart MH, Ferguson DM, DeLeon R, Tayor WD. 1997. Monitoring program to determine pathogen occurrence in relationship to storm events and watershed conditions. In Proc. AWWA Water Quality Technology Conference (WQTC), CD–ROM paper EB1-1, Denver, Colo. Walker FR Jr, Stedinger JR. 1999. Fate and transport model of Cryptosporidium. J. Environ. Eng. 125(4):325–333. Walker MJ, Montemagno CA, Jenkins MB. 1998. Source water assessment and nonpoint sources of acutely toxic contaminants: a review of research related to survival and transport of Cryptosporidium parvum. Water Resour. Res. 34: 3383–3392. Weilenmann U, O’Melia CR, Stumm W. 1989. Particle transport in lakes: models and measurements. Limnol. Oceanogr. 34:1–18. William HP, Teukolsky SA, Press WH, Flannery BP. 1992. Numerical recipes in C: the art of scientific computing. Cambridge, University Press. Williams JR. 1975. Sediment-yield prediction with universal soil loss equation using runoff energy factor. In Present and prospective technology for predicting sediment yields and sources. U.S. Department of Agriculture, ARS-S-40. Wischmeier WH, Smith DD. 1978. Predicting rainfall erosion loss – a guide to conservation planning. Agriculture Handbook No. 537. Washington D.C. USDA-ARS. Young RA, Onstad CA, Bosch DD, Anderson WP. 1986. Agricultural nonpoint source pollution model: a watershed analysis tool, agriculture research service. U.S. Department of Agriculture, Morris, Minn. Zhang X. 1999. Development of a framework for defining source terms of Cryptosporidium oocysts in watersheds. Master’s thesis, University of Waterloo. 108 PARK AND HUCK

Appendix A The Finite Difference Scheme of the Model

Water Channel Transport The advection-dispersion equation, which forms the basis for modelling transport in the water channel, is a second-order differential equation, requiring two boundary conditions and one initial condition for solution. Equation A1 (equivalent to equation 7 in the main text) is the classical advection-dispersion equation, including a die-off term:

∂N ∂2N ∂N = D – ν – kN + S(x, t) 0< x < L (A1) ∂t ∂x2 ∂x

where N is the oocyst concentration (oocyst/m3), D is the dispersion coefficient (m2/s), v is the river velocity (m/s), k is a first order decay coefficient (s-1), and S is a source function. Figure A1 shows the basic finite difference scheme for the water channel, using central block difference. The initial condition is N(x,0) = 0 and the two boundary conditions are N(0,t) = 0 at x = 0 and ∂N = 0 at x = L. The first is a type I boundary condition, while the sec- ∂x ond is a type II boundary condition. The latter boundary condition cannot be proven but is considered reasonable under the circumstances. The finite difference method for pointwise expressions is given by:

∂N N k+1 – N k ≈ i i , where k = time step (A2) ∂t ∆t

∂2N N – 2N + N ≈ i+1 i i–1 (A3) ∂x2 ∆x2

∂N N – N ≈ i+1 i–1 (A4) ∂x2∆x

Fig. A1. The basic finite difference scheme for a water channel. MODEL FOR CRYPTOSPORIDIUM TRANSPORT IN WATERSHEDS 109

Therefore, the solution for the advection-dispersion equation (equation A1) can be determined using the fully implicit method, given by:

(A5)

Rearranging equation A5 gives:

(A6)

The boundary condition N(0,t) = 0 at x = 0 (type I) can be replaced by ∂N N k+1 = 0 , and= 0 at x = L (type II) can be replaced by: 1 ∂x

(A7)

Therefore, the row entry for block n is: k k+1 k+1 N n k (a++ a )N +a N = S (A8) 1 3 n-1 2 n Dt i

To solve this row entry for a block, a tridiagonal matrix is generated by:

(A9) 110 PARK AND HUCK

The Thomas algorithm written in C (William et al. 1992) was used (Park 1999) to solve this tridiagonal matrix in each water channel.

Point Source Upstream of Lake (Section 1) Figure A2 shows Section 1 as a source term with boundary conditions. ∂N The actual mass flux at x = L is given by νN – D at x = L. Because ∂X the mass flux due to advection (vN) is much greater than that due to dis- ∂N persion, the approximation = 0 is justified. ∂x Lake (Section 2) The following equations show the application of Euler’s method for solving the completely mixed lake model (Chapra 1997):

dN W (t) = -lN (A10) dt V

W (t) whereS(=t) (A11) V

Q vs andl = + k+ (A12) V H lake

For a numerical solution, difference approximations can be used to express derivatives in arithmetic terms. For example, using a forward difference, approximation of the first derivative of N with respect to t is given by:

dNi DN Ni+1 - N i @ = (A13) dt Dt ti+1 -ti

Fig. A2. The finite difference scheme for Section 1. ∂N Boundary conditions: N(0,t)=0 at x = 0 and = 0 at x = L. ∂x MODEL FOR CRYPTOSPORIDIUM TRANSPORT IN WATERSHEDS 111

where Ni and Ni+1 are concentrations at a present and a future time ti and ti+1, respectively. Substituting equation A13 into equation A10 yields:

N i+1 - N i W (t) = - lN i (A14) ti+1 -ti V

which can be solved for:

(A15)

Between Lake and Stream Confluence (Section 3) The finite difference scheme used for this section is shown in Fig. A3.

Tributary with Non-point Source (Section 4) The finite difference scheme for this section is given in Fig. A4. The non-point source has been converted to a line source at the water channel.

Fig. A3. The finite difference scheme for Section 3.

Boundary conditions: ∂N N(0,t)=N0 (the concentration exiting the lake) at x = 0 and = 0 at x = L. ∂x

Fig. A4. The finite difference scheme for Section 4.

Boundary conditions: ∂N N(0,t)=0 at x = 0 and = 0 at x = L. ∂x 112 PARK AND HUCK

Stream Confluence Based on the advective mass balance (Fig. A5), oocyst flux at the stream confluence is given by:

N3Q3 + N4Q4 = N5Q5 (A16)

and water flow is given by:

Q3 + Q4 = Q5 (A17)

where N3 is the oocyst concentration in the last cell of Section 3, Q3 is the flow in Section 3, N4 is the oocyst concentration in the last cell of Section 4, Q4 is the flow in Section 4, N5 is the oocyst concentration in the first cell of Section 5, and Q5 is the flow in Section 5. Considering oocyst flux at the stream confluence, one boundary condition for Section 5 is of type III based on mass balance, and is given by:

(A18)

where N is the concentration in the first cell of Section 5. The sign of the right hand side is determined by running the model and checking the mass balance. For example, if equation A18 gives negative values of oocyst concentration in the fifth section, the negative sign should be switched to a positive sign. In this way, the boundary condition at the stream confluence can be determined.

Fig. A5. Mass balance for stream confluence. MODEL FOR CRYPTOSPORIDIUM TRANSPORT IN WATERSHEDS 113

Between Stream Confluence and WTP (Section 5) One of the boundary conditions for Section 5 was discussed immediately above with respect to the stream confluence. The other boundary condition is type II as used previously. The oocyst concentration at the water treatment plant intake in Section 5 is the object of the modelling. Details of the finite difference scheme for Section 5 are shown in Fig. A6.

Fig. A6. The finite difference scheme for Section 5.

Actinomycetes in the Elbow River Basin, Alberta, Canada

BERYL ZAITLIN,1* SUSAN B. WATSON,2 JAMIESON DIXON3 AND DEBORAH STEEL4

1Department of Biological Science, University of Calgary, Calgary, Alberta T2N 1N4 2Aquatic Ecosystem Management Research Branch, National Water Research Institute, P.O. Box 5050, Burlington, Ontario L7R 4A6 3City of Calgary Waterworks and Wastewater, Laboratory Services, Watershed Protection, P.O. Box 2100, Stn. M., Calgary, Alberta T2P 2M5 4RR #1, Hubbards, Nova Scotia BOJ 1T0

Actinomycetes can produce significant amounts of the earthy-muddy odour compounds geosmin and 2-methylisoborneol (MIB). These filamentous bacteria are found in both terrestrial and aquatic environments, and are particularly abundant in soil. They can enter freshwater systems via terrestrial runoff and subsequently cause taste and odour outbreaks in drinking water. Since it is well known that actinomycete growth and odour production is modified strongly by environmental factors such as moisture and nutrient levels, we hypothesized that watershed and stream characteristics should influence the potential odour impact of soil runoff on surface water. In this study, 1) the relationship between actinomycete abundance and characteristics such as stream discharge, turbidity and Escherichia coli levels was investigated, and 2) actinomycetes from contrasting terrestrial sources were examined for differences in their geosmin and MIB production. Actinomycetes and stream characteristics were sampled from the Elbow River, an important drinking water source for the City of Calgary (Alberta, Canada), and three tributary streams. Actinomycetes from forested regions and agricultural land were tested for taste and odour compound production. Actinomycete levels in streams were found to correlate closely with E. coli levels and to a lesser extent with turbidity, suggesting that actinomycetes are particularly abundant in runoff from terrestrial sources with fecal contamination. Most of the 18 actinomycete isolates tested were able to produce geosmin and/or MIB regardless of their terrestrial sources, suggesting that taste and odour outbreaks due to actinomycetes may be more influenced by differences in abundance than differences in source.

Key words: actinomycetes, land use, runoff, geosmin, 2-methylisoborneol

Introduction

Actinomycetes are gram-positive filamentous bacteria that are abundant in soils (Goodfellow and Williams 1983), but are often found in freshwater (Cross 1981; Wnorowski 1992). Some actinomycete species may be resident in freshwater environments (Roach and Silvey 1958; Willoughby

* Corresponding author; [email protected] 116 ZAITLIN ET AL.

1974; Zaitlin et al. In press), but large numbers of these taxa also enter freshwater from land with soil runoff (Persson 1980; Niemi et al. 1982). Actinomycetes from terrestrial environments and those isolated from aquatic environments are often capable of producing highly potent volatile odourous compounds, particularly geosmin (trans-1,10-dimethyl-trans- 9-decalol) and MIB (2-methylisoborneol) (Gerber 1979; Izaguirre 1992; Juttner 1995; Stahl and Parkin 1996). These compounds are responsible for a significant number of taste and odour outbreaks in drinking water supplies (Izaguirre 1992; Juttner 1995; Sugiura and Nakano 2000). Cyanobacteria, which are also major geosmin/MIB producers, have usually been considered responsible for most of these outbreaks (Izaguirre 1992; Juttner 1995), but in some cases, actinomycetes have been implicated (Jensen et al. 1994; Sugiura and Nakano 2000). Actinomycetes tend to be most abundant in soils that are warm, dry and slightly acidic (Goodfellow and Williams 1983). Their abundance and diversity also appears to be influenced by the levels and type of organic matter in agricultural soils (Weyman-Kaczmarkowa and Pedziwilk 1996) or by the composition and depth of forest soil (Davies and Williams 1970). In vitro culture work has shown that production of geosmin/MIB can vary considerably among actinomycete species and with growth conditions (Sivonen 1982; Aoyama 1990). In surface waters, actinomycetes often are found associated with sediment (Johnston and Cross 1976; Cross 1981). However, the type of soil or land use that contributes significant numbers of actinomycetes to streams has not been examined. This research focused on the relationship of actinomycete abundance in freshwater streams to major stream and watershed characteristics. The hypothesis was that the numbers of actinomycetes found in these streams would be correlated with turbidity, streamflow, nutrients, and differences in watershed development. Actinomycetes from different land-use areas were also tested for their taste and odour compound production patterns. In order to test these ideas, 3 streams were selected, draining into the Elbow River (Alta., Canada) from catchments with: i) undisturbed forest, or ii) agricultural/pasture land with some urban development. The river downstream of the 3 stream inflows was sampled, and actinomycetes were isolated from agricultural and forest soil, and their in vitro odour production was characterized. These data were used to assess whether there is any relationship between the numbers and/or the odour production potential of actinomycetes entering streams and associated catchment basin characteristics, which clearly has significant implications for surface water odour levels.

Materials and Methods

Sampling Sites The Elbow River is a major source of drinking water supplies for the City of Calgary, serving the water needs of approximately half the current ACTINOMYCETES IN THE ELBOW RIVER BASIN, ALBERTA 117

population of over 900,000. The Elbow River originates at Elbow Lake in the Front Range of the Canadian Rocky Mountains of southwestern Alberta (50°37’20”N; 115°00’15”W), draining a watershed of 1220 km2 (Fig. 1). The river extends from a largely forested headwater region in Kananaskis Country through alpine, sub-alpine, boreal foothills, and aspen parkland ecoregions, to a predominantly agricultural mid-region of improved pasture with dispersed cattle grazing and accompanying for- age crop production from the hamlet of Bragg Creek to the City of Calgary corporate limits, and thereafter through the city under the influence of the urban environment. In the southwest quadrant of the city, the river is impounded to form the Glenmore Reservoir from which the Glenmore Water Treatment Plant receives its drinking water supply. Several kilome- tres downstream, the Elbow River joins the Bow River. Dominant vegetation in the upper Elbow River drainage basin includes trembling aspen (Populus tremuloides Michx.), balsam poplar (P. balsamifera L.), lodgepole pine (Pinus contorta Loud.) and white spruce (Picea glauca Moench). Soils in the Elbow watershed are primarily black chernozemics, orthic gray luvisols, eutric brunisols, and coarse loam over- lying glaciofluvial gravels (Mitchell and Prepas 1990). Land use in the upper Elbow watershed in Kananaskis Country is centred primarily on recreation, including camping, hiking, mountain

Fig. 1. The Elbow River watershed of southwestern Alberta, Canada, showing the location of the mainstem river site (Weaselhead) and three tributary sites (Prairie Creek, Pirmez Creek, and Springbank Creek). 118 ZAITLIN ET AL. biking, equestrian, and some limited off-road vehicle activity. Logging, oil and gas production, and cattle grazing leases are also present. The hamlet of Bragg Creek is the only municipality in the Elbow River watershed upstream of Calgary; however, country residential estate and acreage lot development is increasing throughout the area. Within the City of Calgary, two storm sewer outfalls draining urban residential catchments flow to the Elbow River above the Weaselhead site; several others drain directly to the Glenmore Reservoir. One mainstem site and three tributary sites of the Elbow River were selected for this study. The mainstem site is located in the Weaselhead Natural Environment Area of North Glenmore Park, immediately upstream of Glenmore Reservoir (50°59’31.1”N; 114°08’51.7”W). Prairie Creek, the uppermost of the tributary sites, drains a small and largely forested watershed of 43 km2. Prairie Creek was sampled near the mouth at 50°52’00.9”N; 114°47’20.3”W. Pirmez Creek, the smallest of the sampled tributary watersheds at 2.5 km2, arises from a small spring in a ranch yard, and flows through primarily pasture land, the grounds of a private residential retreat, and a forested zone near the confluence with the Elbow River. Pirmez Creek was sampled at about mid-reach at 51°02’26.0”N; 114°25’11.6”W. Springbank Creek drains a moderately sized watershed of 32 km2 of mixed agricultural, country residential estate developments, and the small estate/acreage community of Springbank. Springbank Creek was sampled near the mouth at 51°02’06.9”N; 114°19’12.5”W. Sampling for all sites was done during a period of no detectable odours. In the second phase of the work, actinomycetes selected for geosmin/MIB analysis were taken from forest soil in the Kananaskis region, and from agricultural soils in south-central Alberta, at the Lacombe Research Station, Lacombe, Alberta.

Water Sampling and Analysis Samples were taken at weekly intervals during June and July in 1999. At each site, subsurface water samples were taken from the centre of the stream at mid-depth by direct grab, returned within 2 h to the laboratory and stored at 4°C prior to analysis. Samples for chemical analysis were collected by filling clean, pre-rinsed 1-L polyethylene bottles. Samples for bacterial analysis were taken by filling sterile 250-mL polyethylene bacteriological bottles. Stream temperatures were measured on site using a dig- ital thermometer. Stream discharge (Q) was calculated for the tributaries by measuring stream velocity, v, (Global Flow Probe), and depth, d, at set intervals across the stream width, w, as the product of v.d.w (in m3/s). Elbow River discharge at the Weaselhead was estimated as mean daily Glenmore Reservoir inflows based on a reservoir stage-discharge relationship developed for the dam. The calculation considers reservoir level, plant raw intake pump rates, and monitored downstream flows (City of Calgary Waterworks, unpublished). ACTINOMYCETES IN THE ELBOW RIVER BASIN, ALBERTA 119

Turbidity, total phosphorus (TP), sulphate (SO4), nitrate (NO3), and nitrite (NO2) were measured in the laboratory according to Standard Methods (APHA 1998). Turbidity was measured using a Hach 2100N® turbidimeter. TP was analyzed after persulphate digestion using the auto- mated stannous chloride method with a Technicon AAII® continuous flow analyzer. SO4, NO3 and NO2 were measured by ion chromatography using a Dionex DX-120 Ion Chromatograph®. Total coliform and E. coli were measured using a Colilert® Quantitray 2000® system (Edberg et al. 1991; IDEXX Laboratories Inc.), enumerated using the MPN (Most Probable Number) method (APHA 1998). For actinomycete determination, three replicate bottles were shaken vigorously, and then 0.1 mL per bottle was plated on chitin agar with cycloheximide (0.005 g L-1), a medium semiselective for Streptomycetes (Hsu and Lockwood 1975). Plates were incubated under ambient laboratory conditions (cf. 23°C) and counted after 10 days to allow for development of slow-growing forms.

Terrestrial Sampling Samples were taken from the F, H, and A horizons at five locations in the Kananaskis region in July 2001 and from four fields used for barley and other grain production at the Lacombe Research Station, Lacombe, Alberta, on 8 May 1998. Isolations were made by diluting 1 g of soil in sterile distilled water to 0.1 mg L-1 concentration, and plating the diluted soil on chitin agar amended with cycloheximide. Actinomycete colonies were serially plated on Czapek’s agar (CZ) and nutrient agar until free of contamination. The ten most common actinomycetes from the Kananaskis region (D. Jayasingh, University of Calgary, Calgary, Alta., pers. comm.) and common actinomycetes from Lacombe were selected for further study. Plates were incubated under ambient laboratory conditions (~23°C). Eighteen of these terrestrial actinomycete isolates were tested for geosmin/MIB production using a headspace microextraction GC-MS pro- tocol (HSPME/GC-MS; Watson et al. 2000) that has a detection limit of 2 ng L-1, which is below the human threshold odour concentrations of 10 to 30 ng L-1 for geosmin and MIB (cf. Young et al. 1996). The routine replic- ability of the measurements was ± ca. 10 to 12%. At 14 days growth, plates were sampled by removing approximately 1 cm2 of agar with the associated biomass, which was then added to 25 mL distilled water in a septum- capped glass vial containing a stir bar and 6 g of pre-baked NaCl. -1 Biphenyl-D10 was added at 200 ng L as an internal standard. The vials were sealed with a septum screw cap, stirred and extracted for 30 min using a polydimethylsiloxane/divinylbenzene (PDMS/DVB, 65 µm) SPME fibre (Supelco; Sigma-Aldrich Canada), inserted through the septum and extended into the headspace. Samples were measured using a Hewlett Packard 6890 GC coupled to a Hewlett Packard 5972® mass selective detector. GC-MS analyses were carried out after 1 min desorption in splitless mode. Samples were run in full scan mode and geosmin and MIB 120 ZAITLIN ET AL. were identified and quantified using retention times and mass spectra derived from analytical standards (Sigma-Aldrich Canada).

Data Analysis Total actinomycete, total coliform and E. coli counts, turbidity, flow rate, temperature, total phosphorus, sulphate, and nitrate were analyzed for linear relationships by scatterplot analysis, then a simple regression was done to determine if actinomycete counts could be predicted by the other parameters. Analysis was done using the SPSS software package (SPSS 11.0, SPSS Inc., Chicago, Ill.).

Results

Over the duration of the experiment, the Elbow River had the highest and lowest single-day actinomycete counts, with a mean count of 256 colony-forming units (CFU mL-1) on 5 July and a count of zero on 7 June. Overall maxima, minima and average abundances are shown in Table 1. Lognormal transformed actinomycete colony counts closely tracked ln E. coli counts (Fig. 2) and total coliform counts (figure not shown). Regression analysis indicated highly significant correlations between lognormal transformed levels of actinomycete and E. coli (n = 15, P < 0.001), and between lognormal transformed actinomycete and total coliform counts (n = 15; P = 0.005). Significant correlations were found between lognormal transformed actinomycete counts and turbidity (n = 14, P=0.1). No significant correlations were found between actinomycetes and all other measured parameters (stream discharge, temperature, TP, SO4, NO3, and NO2). Taken individually, there was a negative correlation between lognormal transformed actinomycete counts and turbidity in Pirmez creek (n = 3, P = 0.007), but not in the Elbow River (n = 3, P=0.145). Prairie Creek and Springbank Creek could not be analyzed due

Table 1. Mean actinomycete colony counts (CFU/mL) for three tributaries of the Elbow River (Pirmez Creek, Prairie Creek and Springbank Creek), and the Elbow River at the Weaselhead Natural Environment Area

Average Highest mean Lowest mean actinomycete actinomycete actinomycete counts in count in count in Site CFU/mL CFU/mL (date) CFU/mL (date)

Prairie Creek 48.4 137 (5 July 1999) 10 (7 June 1999) Pirmez Creek 4.5 37 (5 June 1999) 0.3 (7 June 1999) Springbank Creek 84.1 230 (5 July 1999) 13 (21 June 1999) Elbow River 84.1 257 (5 July 1999) 0 (7 June 1999) ACTINOMYCETES IN THE ELBOW RIVER BASIN, ALBERTA 121

Fig. 2. Lognormal densities of actinomycetes (CFU/mL) in relation to Escherichia coli (CFU/100 mL) in four water systems.

to the small sample size. Increased turbidity in all cases during the study was due to precipitation and subsequent runoff. Actinomycete counts, E. coli counts, total coliform counts and TP peaked or were elevated in Springbank Creek, Pirmez Creek and the Elbow River on 5 or 6 July, but only actinomycete counts peaked on this date in Prairie Creek (Table 2). There were no significant differences between overall actinomycete totals at any of the stream sites; however, the 5/6 July sampling date had significantly more actinomycetes (as lognormal transformed data) than the other three sampling dates (7–8 June, 21–22 June, 19–20 July). Other stream characteristics also did not differ significantly between the sites, except the Elbow River site had significantly higher SO4 levels (n = 15, P = 0.016). The majority of the 18 forest and agricultural soil actinomycete isolates tested produced geosmin and/or MIB (Fig. 3) but there were no differences in odour patterns between actinomycetes from any of the different soil types.

Discussion

In this study, the pristine creeks Prairie Creek and Pirmez Creek both had low levels of E. coli and lower levels of actinomycetes than Springbank Creek and the Elbow River. However, actinomycetes decreased with increasing turbidity in Pirmez Creek, suggesting they were diluted out with increased water input in this pristine creek. There was a large peak in actinomycete levels that correlated with turbidity but not E. coli levels in 122 ZAITLIN ET AL. 2 NO 3 NO 4 /s) (°C) Turbidity (mg/L) (mg/L) (mg/L) (mg/L) 3 l coliform Stream Tota E. coli total count (CFU/ (CFU/ discharge Temperature TP SO Actinomycete Actinomycetes, total coliform, Escherichia coli, and stream parameters for three tributaries of the Elbow River (Pirmez Creek, parameters for three Actinomycetes, total coliform, Escherichia coli, and stream 6/7 36/8 1 206/7 96/7 1 613 0 0.43 72 25 4.7 2420 8 0.28 0.01 6.5 517 3.5 0.002 10.4 0.85 10.7 0 0.003 10.5 0.93 0 0 0.079 3.94 0 0 0.008 0.03 63.15 0.108 0 0 ER; Elbow River. PC; Pirmez Creek. PRC; Prairie Creek. SC; Springbank Creek. a b c d a b d c able 2. PRC T Area Natural Environment and the Elbow River at Weaselhead and Springbank Creek), Prairie Creek Site Date (CFU/0.1 mL) 100 mL) 100 mL) (m ER 7/23 19 67 2420 41.27 9.3 120.00 0.111 30.77 0.079 0 PRCPRC 6/21PRC 7/5PC 7/19PC 3PC 6/22 41PC 11 7/6SC 7/20SCSC 17 5 6/21SC 20 11 31 7/5ER 1 7/19ERER 4 649 6/21 4 69 422 34 12 7/5 613 12 0.48 5 22 0.35 504 77 1.84 51 345 31 261 6.8 12 24,192 980 7.8 228 0.28 9.63 11.24 2909 23.80 0.17 0.39 0.01 7.9 172 5.2 2420 5.6 0.53 0.003 0.002 9.2 11.61 42.24 7.38 0.40 8.41 11.2 0.006 0.17 0.228 0.229 8.36 0.002 0.045 1.52 8.6 0.288 0 0.79 0.94 0.81 0.023 0 0.170 0 0 401.00 0.84 39.99 0.058 0 7.13 0.108 0.019 0 0.026 0 0 46.62 0 38.65 0.025 0.03 0 0.082 0 0 0 ACTINOMYCETES IN THE ELBOW RIVER BASIN, ALBERTA 123

Fig. 3. Production of geosmin and MIB by 18 actinomycete isolates from forest and agricultural soil.

July in Prairie Creek. Prairie Creek contrasted to Springbank Creek, which drains a watershed of equivalent size. In Springbank Creek, E. coli and actinomycete levels had higher maxima at times of increased turbidity, and similar values to the other creeks at times of low turbidity. The Elbow River at Weaselhead, which receives inputs from the three creeks in this study, other creeks and two large storm sewers, also showed increased actinomycetes and E. coli levels at times of increased turbidity, and lower levels at times of low turbidity. This suggests that the actinomycetes were coming from similar areas as the E. coli, i.e., terrestrial sources associated with animal manure. This is similar to the findings of Al-Diwany and Cross (1978) who also reported a close correlation between levels of Streptomyces, Rhodococcus coprophilus and fecal streptococci in river water. The majority of the actinomycetes isolated from both forest and agricultural soil were capable of producing geosmin and/or MIB, similar to actinomycetes isolated from sediment in the Lake Ontario basin (Zaitlin et al. In press) and actinomycetes isolated from Lake Kasumigaura, Japan (Sugiura and Nakano 2000). This suggests that if conditions were favorable for geosmin/MIB production and a heavy sediment load entered streams from forest or agricultural soil, significant taste and odour outbreaks could occur. The type of sediment entering the water could have a major effect on the magnitude of the odour outbreak. Odour is a function of the number of odour-producing cells and the amount of odour produced per cell. This in turn is a function of both the species producing the odour and the conditions under which the odour is produced. This research indicated that both forest- and agricultural-origin actinomycetes are capable of producing odour. Furthermore, as fecal coliform bacteria are found in low concentrations in soils that do not receive manure inputs (Faust 1982), the close correlation in this study between E. coli and actinomycete counts suggests that the primary source of actinomycetes in these streams is not forest soil but soil associated with animal manure. The prevalence of actinomycetes in land used for animal pasture has never been studied. If odour-producing actinomycetes are highly preva- lent in pasture soils, runoff from pasture soils may be more significant in 124 ZAITLIN ET AL. taste and odour outbreaks than runoff from other areas. Clearly, this is a significant area of future investigation.

Acknowledgements

We gratefully acknowledge the exceptional support of the City of Calgary Waterworks and Wastewater Laboratory (Glenmore), in particular Dave Johnson of the watershed group for field sampling and chemical analyses, and Alison Strilchuck and Paul Mayberry of the Microbiology group for coliform bacteriological analysis. We would like to thank Patricia Rathwell for technical assistance, and the lab of Dennis Parkinson at the University of Calgary for use of the facilities.

References

Al-Diwany LJ, Cross T. 1978. Ecological studies on Nocardioforms and other actinomycetes in aquatic habitats, p. 153–160. In Mordarski M, Kurylowicz W and Jeljaszewicz J (ed.), Proc. Int. Symp. on Nocardia and Streptomycetes. Gustav Fischer Verlag, Stuttgart. Aoyama K. 1990. Studies on the earthy-musty odours in natural water (IV). Mechanism of earthy-musty odour production of actinomycetes. J. Appl. Bacteriol. 68:405–410. APHA. 1998. Standard methods for the examination of water and wastewater. 20th Edition. Clesceri LS, Greenberg AE and Eaton AD (ed.). American Public Health Association, American Water Works Association, and Water Environment Federation. Washington, D.C. Cross T. 1981. Aquatic actinomycetes: a critical survey of the occurrence, growth and role of actinomycetes in aquatic habitats. J. Appl. Bacteriol. 50:397–423. Davies FL, Williams ST. 1970. Studies on the ecology of actinomycetes in soil. I. The occurrence and distribution of actinomycetes in a pine forest soil. Soil Biol. Biochem. 2:227–238. Edberg SC, Ludwig F, Smith DB. 1991. The Colilert® system for total coliforms and Escherichia coli. AWWA Research Foundation and American Water Works Association. Denver, Colo. Faust MA. 1982. Relationship between land use practices and fecal bacteria in soils. J. Environ. Qual. 11:141–146. Gerber N. 1979. Volatile substances from actinomycetes: their role in the odor pollution of water. CRC Crit. Rev. Microbiol. 7:191–214. Goodfellow M, Williams ST. 1983. Ecology of actinomycetes. Ann. Rev. Microbiol. 37:189–216. Hsu SC, Lockwood JL. 1975. Powdered chitin agar as a selective medium for enumeration of actinomycetes in water and soil. Appl. Microbiol. 29:422–426. IDEXX Laboratories Inc. One IDEXX Drive. Westbrook, Me. 04092. Izaguirre G. 1992. A copper-tolerant Phormidium species from Lake Matthews, California, that produces 2-methylisoborneol and geosmin. Water Sci. Technol. 25:217–223. Jayasingh D. University of Calgary, Calgary, Alta., pers. comm. Jensen SE, Anders CL, Goatcher LJ, Perley T, Kenefick S, Hrudey SE. 1994. Actinomycetes as a factor in odour problems affecting drinking water from the North Saskatchewan River. Water Res. 28:1393–1401. ACTINOMYCETES IN THE ELBOW RIVER BASIN, ALBERTA 125

Juttner F. 1995. Physiology and biochemistry of odourous compounds from freshwater cyanobacteria and algae. Water Sci. Technol. 31:69–78. Johnston DW, Cross T. 1976. Actinomycetes in lake muds: dormant spores or metabolically active mycelium? Freshwater Biol. 6:465–470. Niemi RM, Knuth S, Lundström K. 1982. Actinomycetes and fungi in surface waters and potable water. Appl. Environ. Microbiol. 43:378–388. Mitchell PA, Prepas EE (eds.). 1990. Atlas of Alberta lakes. The University of Alberta Press. 690 p. Persson PE. 1980. Muddy odour in fish from hypertrophic waters. Dev. Hydrobiol. 2:203–208. Roach AW, Silvey JKG. 1958. The morphology and life cycle of fresh-water actinomycetes. Trans. Am. Microsc. Soc. 77:36–47. Sivonen K. 1982. Factors influencing odour production by actinomycetes. Hydrobiologia 86:165–170. Stahl PD, Parkin TB. 1996. Microbial production of volatile organic compounds in soil microcosms. Soil Sci. Soc. Am. J. 60:821–828. Sugiura N, Nakano K. 2000. Causative microorganisms for musty odor occurrence in the eutrophic Lake Kasumigaura. Hydrobiologia 434:145–150. Watson SB, Brownlee B, Satchwill T. 2000. Quantitative analysis of trace levels of geosmin and MIB in source and drinking water using headspace SPME. Water Res. 34:2818–2828. Weyman-Kaczmarkowa W, Pedziwilk Z. 1996. Humidity conditions and the occurrence of actinomycetes and their fungistatic forms in soils of contrasting texture. Acta Microbiologica Polonica 45:285–290. Willoughby LG. 1974. The activity of Streptomyces phage-virus in fresh water. Hydrobiologia 49:215–228. Wnorowski AU. 1992. Tastes and odors in the aquatic environment: a review. Water SA 18:203–213. Young WF, Horth H, Crane R, Ogden T, Arnott M. 1996. Taste and odour threshold concentrations of potential potable water contaminants. Water Res. 30:331–340. Zaitlin B, Watson SB, Ridal J, Satchwill T, Parkinson D. Actinomycetes in Lake Ontario: habitats and geosmin and 2-methylisoborneol production. J. Amer. Water Works Assoc. In press. Water Qual. Res. J. Canada, 2003 Volume 38, No. 1, 127–140 Copyright © 2003, CAWQ

New Microcystin Concerns in the Lower Great Lakes

TOM P. M URPHY,1* KIM IRVINE,2 JAY GUO,1 JOHN DAVIES,1,3 HENRY MURKIN,3 MURRAY CHARLTON1 AND SUSAN B. WATSON1

1National Water Research Institute, 867 Lakeshore Road, Burlington, Ontario L7R 4A6 2State University College at Buffalo, 1300 Elmwood Avenue, Buffalo, New York, 14222, U.S.A. 3Institute for Wetland and Waterfowl Research, P.O. Box 1160, Stonewall, Manitoba R0C 2Z0

Water samples were collected in the summer of 2001 for microcystin analysis, nutrients and algal enumeration from Hamilton Harbour (Lake Ontario), Wendt Beach (Lake Erie) and Presque Isle (Lake Erie). Microcystin concentrations varied largely and were present at acute toxicity levels only in some wind- concentrated scums of blue-green algae (>90% Microcystis, primarily M. botrys, M. viridis and some M. wesenbergii) in Hamilton Harbour. In Hamilton Harbour, microcystin-RR was the main microcystin with microcystin-YR and -LR also present. The two samples of August 17 and September 7, taken during the peak of the cyanobacterial bloom, contained 60 and 400 µg/L, respectively. A few dying birds were seen in the Hamilton scums. The concentrations of microcystins at the Lake Erie sites were less than 1 µg/L, yet dead birds were common. The major limitation with this approach is that current analysis (ELISA and HPLC) methods are unable to measure covalently bound microcystins, the form that is assimilated into the food chain.

Key words: microcystin, Microcystis, botulism, algal toxins

Introduction

Toxic algal blooms have been occurring more frequently throughout the world (Hallegraeff 1993). The recent deaths of 55 people in Brazil from microcystins may be the most alarming occurrence, but human health has been affected by algal toxins for some time (Falconer 1999). In freshwater, toxin-producing Cylindrospermopsis is spreading from tropical lakes north (Chapman and Schelske 1997) and has reached Indiana (St. Amand 2002). In addition, toxic algal blooms and avian botulism appear linked. Reports of dead birds or fish and blue-green algae are common (Yoo et al. 1995). Toxic blue-green algae in the prairies of North America are associated with outbreaks of avian botulism that have resulted in millions of dead birds (Yoo et al. 1995; Murphy et al. 2000). The association is not well understood and what may apply to type C botulism may not be relevant to type E botulism. Type E botulism is a major recent problem in lakes Erie

* Corresponding author; [email protected] 128 MURPHY ET AL. and Huron, and has spread to Lake Ontario. This event is of significant concern to the public. Botulism in fish taken from Lake Michigan killed 7 people in the 1960s. It led to a study in 1966 showing that 57% of fish in Green Bay, Michigan, had spores of type E Clostridium (Brand 2001). In Lake Erie the algal toxin of concern is microcystin (Brittain et al. 2000). There is both a general need to understand the causes and ecology of toxic algal blooms, and specific needs to clarify the effect of imposed changes such as those caused by nutrients, introduced species and other uncertainties such as climate change. Many changes are taking place globally that may be responsible for the increase in toxic algae. Overall changes in the food web structure imply changing relationships between algal blooms and surrounding environments (Lehman et al. 2000). Nitrogen is increasing globally (Sutton 2002). Charlton et al. (1999) observed that nitrogen concentrations in Lake Erie have increased two- to six-fold since 1970. Some of this link between enhanced nitrogen supply and toxic algae is clear, as it increases both the potential for toxic algal outbreaks and their production of cyanotoxins. Blue-green algae like Microcystis do not fix significant amounts of nitrogen and are clearly stimulated by enhanced dissolved inorganic nitrogen availability (Murphy and Brownlee 1981; Hyenstrard et al. 1998). Furthermore, microcystin production has been shown to be positively correlated to nitrogen availability (Sivonen 1990). Global changes in UV light also are known to stimulate microcystin production (Wulff 2001). This latter change has particular relevance to the Great Lakes, where water filtration by introduced species like zebra mussels (Dreissena polymorpha) and quagga mussels (Dreissena bugensis) has dramatically increased light penetration (Horgan and Mills 1997). Microcystis species are enhanced in this environment for many reasons, notably their ability to resist grazing by these and other herbivores, and to minimize the adverse effects of UV irradiation, which is particularly severe towards surface scums. Microcystin is broken down by UV light (Tsuji et al. 1995) and one of its functions may be to protect the cell from UV damage. The purpose of this paper was to measure algal toxins and compare their presence to variables that could influence the botulism outbreak. The analytical limitations of botulism analysis and resource restrictions precluded precise hypothesis testing of the linkages between microcystins and botulism.

Methods

Sample Collection A Hydrolab® was used in Lake Erie to measure temperature, pH, conductivity, dissolved oxygen, and turbidity. Initially, samples were collected from six sites at both Presque Isle and Wendt Beach but later were reduced to five at Presque Isle and three at Wendt Beach (Fig. 1). All samples were collected with a Van Dorn bottle at 1.0-m depth. Water depths MICROCYSTINS IN LOWER GREAT LAKES 129 were between 0.8 and 3.8 m. Secchi depths varied from 1.6 to 3.8 m. Samples were placed in a cooler for processing within 24 h at the Canada Centre for Inland Waters (CCIW). The following analyses were submitted to the National Laboratory for Environmental Testing (NLET), Burlington, Ontario: total phosphorus (filtered and unfiltered), dissolved inorganic carbon, and dissolved organic carbon. Metals and alkalinity were also performed by NLET for the first sampling date only (NLET 1994). Total phosphorus was measured on unfiltered water samples and water samples that had been filtered through 0.45-µm cellulose acetate filter paper. Both filtered and unfiltered samples were placed into 100-mL square glass bottles, preserved with 1 mL of 30% sulphuric acid and refrigerated. For metal analysis, water samples were filtered through 0.45- µm cellulose acetate filter paper, the samples were then placed into 500- mL square plastic bottles, preserved with 1 mL of concentrated nitric acid and refrigerated. For dissolved inorganic and organic carbon analysis, water was filtered through 0.45-µm cellulose acetate filter paper, placed in 125-mL round glass bottles and refrigerated. Suspended solids were prepared by filtering water through preweighed glass fibre filters (GF/F). Ammonia samples were filtered (GF/F), and placed into 250-mL plastic bottles. The samples were preserved with 10 mL of phenol and refrigerated prior to analysis. The analysis took place within 18 h and used the col-

Fig. 1. Map of sampling sites. 130 MURPHY ET AL.

orimetric phenate method 4500-NH3 D (APHA 1989). Absorbance was measured on a LKB Biochrom UV spectrophotometer at 640 nm. Sulphate and nitrate anions were filtered through a GF/F filter and later analyzed on a Dionex model 2010i ion chromatograph. Samples for algal enumeration were transferred to 125-mL glass bottles and preserved with 2.5 mL of Lugol’s solution and refrigerated prior to identification of blue-green algae with the Utermöhl technique. Water samples for chlorophyll analysis were filtered through GF/F filters in subdued (yellow) light and frozen for later analysis according to Standard Method 10200 H-Chlorophyll method (APHA 1989). Particles for algal toxin analysis were filtered through GF/C filters (approximately 50 mg dry algae), placed in petri dishes and frozen. Most samples were processed by ELISA analysis but a subset was evaluated with HPLC. Samples were sonicated with a cell disruptor by Heat Systems® with a microtip for 6 min with 5 mL of methanol. After a further 45 min of extraction, samples were centrifuged for 40 min on 6 IEC coni- cal benchtop centrifuge at 2000 rpm. The extracted supernatant was removed using Pasteur pipette and placed into 4-mL glass vials with Teflon caps. ELISA analysis was conducted on a Biorad® 3550 Microplate reader with a Envirologix® microcystin plate kit. Filters for HPLC analysis were extracted using 4 mL of 75% methanol in test tubes, sonicated 5 min in bath sonicator, repeated 3 times, letting extraction solvent sit 1 h each time. The extracts were pooled together in a clean tube and filtered through Aerodisk® filters. The extracts were then evaporated to 0 volume using N-Vap nitrogen evaporator. Next, the samples were reconstituted with 1 mL of methanol. The solution was then placed in Waters autosampler vials. The HPLC mobile phase was MeOH:0.05 m KH2PO4 60:40, pH 3.0. The HPLC system consisted of dual Waters 510 pumps, WISP® autoinjector, and a Kratos® UV detector set at 238 nm. The column was a 5-?m 4.6 x 150 mm Xterra MS C18 (Waters) connected to a µBondapak guard insert. Injection volumes were 30 µL per standard and sample. Target compounds (microcystin-LR, microcystin- RR) were identified by retention time comparison of standards and sample spiking. Concentrations were determined by comparing peak areas against a calibration curve. In Hamilton Harbour, most samples were collected north of CCIW near the wind surfing launch area; these samples were surface samples. For the midharbour samples (Station 1001), a Van Dorn bottle was used to collect samples at depths. Sample processing was similar to that used for the Lake Erie samples.

Results

In late August and September of 2001, Hamilton Harbour developed a conspicuous cyanobacterial bloom, which concentrated at the surface on calm days (Watson et al. Submitted). The bloom was composed of a number of blue-green taxa, dominated by several species of Microcystis MICROCYSTINS IN LOWER GREAT LAKES 131

(M. viridis, M. botrys and M. wesenbergii); Aphanizomenon flos-aquae also was abundant, particularly towards the end of the summer. The concentrations of microcystin in the wind bloom scums in Hamilton Harbour were as high as 400 µg/L (Table 1). The blooms were blown by the west wind to the eastern end of the harbour. This is an important recreational area so human contact is a potential concern. It is also the area where major wildlife restoration projects have resulted in establishment of large bird colonies, but fortunately the birds were matured by the time the blooms developed. The movement of the surface scums was visually obvious. They tended to move into the shipping canal, into Lake Ontario

Table 1. Hamilton Harbour: total microcystins (MC), summer 2001

MC MC Date Samplea µg/L µg/g wet wt.

SESTON Aug 9/01 B 1 m 0.17 46.1 Aug 9/01 B 3 0,19 33.1 Aug 9/01 B 5 M 0.01 2.4 Aug 9/01 B 7 M 0.01 1.3 Aug 9/01 B 19 M 0.0 0.3 Aug 14/01 D 0.02 8.6 Aug 17/01 C N/Ac 641.7 Aug 14/01 D 0.06 25.4 Aug 14/01 C N/A 110.1 Aug 14/01 E N/A 32.1 Aug 17/01 C 33.1 429.9 Aug 24/01 B 0.25 53.2 Aug 24/01 A 5.03 188.2 Sept 4/01 B 0.96 267.3 Sept 4/01 C 0.47 164.6 Sept 7/01 Cb 238.8 656.7 Sept 7/01 C 202.2 Sept 12/01 Bb 0.98 244.6 Sept 18/01 Bb 0.24 84.3 Oct 3/01 Bb 0.0 0.0 GOBIE Oct 3/01 Liver 0.001 CCIW dock Muscle 0.001

aUnless marked otherwise, all samples are from 1 m. bHPLC (all other analyses are via ELISA). cN/A; Not available. 132 MURPHY ET AL. and potentially near the drinking water intakes. On one occasion during the peak of the cyanobacterial bloom, the depth profile of microcystins was evaluated and the results confirmed the visual impression; the toxins were only in the surface water. The limited number of samples prevents statistical analysis but visual observations indicate that most of the time, much of the harbour had low concentrations of microcystins in the water column. The toxic scums were obvious and could be avoided. There were very few dead birds and no dead fish observed in Hamilton Harbour. It is possible that dead animals were not detected but it is more likely that animals avoided the toxic scums. Shortly after the worst of the bloom, several cormorants were found dead under local hydro lines but no analysis of the carcasses was done. Live gobies that were sampled at this time had very large livers, which might indicate microcystin stress. The livers had about 1.4 µg/g of microcystins by ELISA analysis. The concentrations of microcystins in Lake Erie were much lower (Table 2). The chlorophyll a content was never higher than 11 µg/L (Table 3). Unlike Hamilton Harbour, there were many dead gulls, carp, freshwater drum and other fish observed during our sampling. The nutrient concentration of the Lake Erie sites was much lower than seen in Hamilton Harbour (Table 4 and 5). The metal concentration at the Lake Erie sites, especially iron, was low (Table 5). There were no unusual temperature or oxygen measurements at the Lake Erie sites (Table 6); both variables are known to be associated with fish and bird kills.

Discussion

The potential for risk to humans was minimal but more analysis is war- ranted. In part, the scum was readily visible and avoidable. The harbour was posted to warn people of the risks of contact and not to eat the fish. Complications include difficulty in measuring each algal toxin (Chorus 2002) and the potential for unusual human exposure. For example, there had been some concerns about workers around the cooling water intakes of the steel mills but since they draw deeper water, they had no exposure. There is no doubt that microcystins are toxic to fish (Erickson et al. 1989; Andersen et al. 1993), but more often concentrations in lakes are not acutely lethal (Carbis et al. 1996). The latter situation seems to apply to our sampling at two sites in Lake Erie in 2000. Our results are biased in that we did not measure covalently bonded microcystins and in animal tissue this is the major form (Tencalla and Dietrich 1997; Williams et al. 1997a, b). We did measure microcystins (concentrations usually <1 µg/L, but up to 200 µg/g) in a free form at Presque Isle. These concentrations may not stress fish, but could influence their behaviour and would impair development (Bury et al. 1996; Baganz et al. 1998; Wiegand et al. 1999, respectively). There is less information on the toxicity of microcystins to birds. Certainly, birds have been observed dying in the presence of algal toxins (Yoo et al. 1995; Murphy et al. 2000; Matsunaga 1999). One laboratory study with quails observed an MICROCYSTINS IN LOWER GREAT LAKES 133

Table 2. Lake Erie: total microcystins (MC), summer 2001

MC MC Date Samplea µg/L µg/g wet wt.

June 28/01 PIb 1 0.024 N/Ad June 28/01 PI 2 0.035 N/A June 28/01 PI 3 <0.000 N/A June 28/01 PI 4 <0.000 N/A June 28/01 PI 5 0.000 N/A June 27/01 WBc 1 <0.000 N/A June 27/01 WB 2 <0.000 N/A June 27/01 WB 3 <0.000 N/A June 27/01 WB 4 <0.000 N/A June 27/01 WB 5 0.000 N/A Aug 9/01 PI 1 0.142 0.3 Aug 9/01 PI 2 0.302 0.5 Aug 9/01 PI 3 0.016 0.5 Aug 9/01 PI 4 0.009 24.9 Aug 9/01 PI 5 0.008 59.8 Aug 8/01 WB1 0.001 3.3 Aug 8/01 WB2 0.006 1.0 Aug 8/01 WB4 0.002 2.7 Aug 29/01 PI1 0.371 43.8 Aug 29/01 PI2 0.407 77.8 Aug 29/01 PI3 0.200 62.5 Aug 29/01 PI4 0.028 21.0 Aug 29/01 PI5 0.031 13.4 Aug 28/01 WB1 0.009 16.8 Aug 28/01 WB2 0.008 14.5 Aug 28/01 WB4 0.009 12.9

aUnless marked otherwise, all samples are from 1 m. bPI; Presque Isle. cWB; Wendt Beach at Sturgeon Point. dN/A; Not available.

LC50 of 256 µg/kg with microcystins (Takahashi and Kaya 1993). However, when microcystins were fed to mallard ducklings orally, they were not toxic, although when injected, they were (Wobeser and Murphy 1999). A similar observation was made with zebra fish, where crude isolates were much more toxic than pure microcystins (Oberemm et al. 1997). Many reac- tions can explain this discrepancy, including the possible protection of microcystins by being cellular, complexed or covalently bound to proteins. If microcystins were responsible for some of the dead fish and birds at our 134 MURPHY ET AL.

Table 3. Lake Erie nutrient, chlorophyll (chla) and chemical data, 2001

a b c d Site Date TP-UF TP-F DOC DIC Chla NH4-N NO3-N SO4 µg/L µg/L mg/L mg/L µg/L µg/L µg/L mg/L

1Presque 27-Jun 111 9 4 16.1 N/De 14 0 32.4 2Presque 27-Jun 39 4 3.5 20.1 1.4 17 153 35.5 3Presque 27-Jun 7 2 2.8 21.4 N/D 22 238 30.5 4Presque 27-Jun 7 2 3.1 21.9 N/D 18 171 21.4 5Presque 27-Jun 7 2 3 21.7 0.3 12 223 29.3

1Wendt 28-Jun 9 4 2.9 21.7 N/D 2 183 32.9 2Wendt 28-Jun 8 4 2.8 22 N/D 3 199 28.4 3Wendt 28-Jun 8 3 2.8 21.6 2.8 8 186 30.4 4Wendt 28-Jun 8 6 2.8 21.4 11.7 10 202 28.8 5Wendt 28-Jun 7 3 2.8 21.6 0.4 18 203 29.7

1Presque 08-Aug 20 8 3.7 19.5 3.8 3 35 26.6 2Presque 08-Aug 23 8 3.2 20.8 4.4 6 18 28.2 3Presque 08-Aug 14 5 2.5 23.2 0.9 6 147 25.1 4Presque 08-Aug 11 4 2.3 22.4 0.6 33 139 25.5 5Presque 08-Aug 11 6 2.5 23.6 0.9 25 112 25.6

1Wendt 09-Aug 15 6 2.7 23.2 2.1 20 122 25.6 2Wendt 09-Aug 38 6 2.8 24.7 9.9 120 195 27.6 3Wendt 09-Aug 4Wendt 09-Aug 38 5 2.7 23.3 N/D 28 147 27.4 5Wendt 09-Aug

1Presque 29-Aug 36 9 3.6 19.4 5.0 9 32 22.7 2Presque 29-Aug 26 7 3.2 21 8.1 5 84 26.9 3Presque 29-Aug 22 7 2.8 22.5 6.5 5 219 30.7 4Presque 29-Aug 8 4 2.5 22.1 1.0 6 146 22.6 5Presque 29-Aug 8 5 2.4 22.2 0.7 16 123 23.3

1Wendt 30-Aug 7 3 2.8 22.7 0.2 N/D 125 24.8 2Wendt 30-Aug 6 3 2.6 22.8 0.4 N/D 92 23.4 3Wendt 30-Aug N/D 4Wendt 30-Aug 6 4 2.7 22.1 0.4 3 83 23.7 5Wendt 30-Aug N/D

aTP-UF; Total phosphorus, unfiltered. bTP-F; Total phosphorus, filtered. cDOC; Dissolved organic carbon. dDIC; Dissolved inorganic carbon. eND; Not detected. MICROCYSTINS IN LOWER GREAT LAKES 135

Table 4. Average nutrients in Hamilton Harbour, 2001a

Seasonal 2001 June July August September average

TP-UF b 34 38 38 30 35 NH4-N 0.57 0.19 0.12 0.08 0.24 NO2+3 2.15 1.98 1.74 1.62 1.87

aValues represent averages of 1-, 3- and 5-m samples (µg/L). bTP-UF; Total phosphorus, unfiltered.

sites, we would have to assume one or more of the following: bioaccumulation of toxins, chronic effects enhancing a disease or other toxin, movement of toxin from the western basin of the lake, or storm-generated resuspension of a pulse of toxin. For Lake Erie, the pathway of assimilation with animals like zebra mussels seems important. Zebra mussels appear to promote Microcystis blooms in Lake Erie, Saginaw Bay, Lake Huron (Vanderploeg et al. 1998), Hamilton Harbour and the Bay of Quinte (Lake Ontario). Bioaccumulation of microcystins into mussels and snails has been observed in Canada, Finland and The Netherlands (Eriksson et al. 1989; Zurawell and Prepas 1995; Burger et al. 1997; Watanabe et al. 1997). Zebra mussels may promote Microcystis blooms, although they do not directly ingest the algae; bioaccumulation of microcystin in zebra mussels does not seem likely. However, it is believed that they concentrate and eject Microcystis in pseudofeces, thus making it available to other benthic inhabitants, or other inhabitants in the water column when resuspension takes place. Studies are needed to evaluate concentrations of microcystin in pseudofeces. A similar pathway for the death of birds in Monterey Bay, California, has been proposed (Fritz et al. 1992). At that site, the diatom Pseudonitzschia australis produced the toxin domoic acid. It was assimilated into anchovies, which then killed large numbers of pelicans and cormorants. It is interesting that outbreaks of avian botulism at Salton Sea, California, are apparently related to blue- green algal blooms and dead fish (Sturm 1998). At Lake Erie, it is fish-eat- ing birds that die during fish kills. A small study cannot resolve the reasons behind the recent outbreaks of botulism in Lake Erie or the initiation of toxic algae in Hamilton Harbour. However, comparisons of the two sites allow formulation of hypotheses or at least interesting speculations. The increased nitrogen in Lake Erie may contribute to the growth of toxic Microcystis but Hamilton Harbour has had much higher concentrations of nitrogen (Charlton and Le Sage 1996) so nitrogen does seem to explain the changes, at least in Hamilton Harbour. Mussels and gobies are found in both Hamilton Harbour and Lake Erie. The lag of a few years in Hamilton Harbour in the 136 MURPHY ET AL. a Lake Erie, 2001: early summer metal data (mg/L) Al, Pb, and Ag were undetectable; Cd, Cr, Li, V, Mo, and Ni were at near detection. Mo, and Ni were Li, V, undetectable; Cd, Cr, Ag were Al, Pb, and a able 5. T Site1 Presque2 Presque Jun 273 Presque Date Jun 274 Presque 0.0824 Jun 275 Presque 0.0824 Ba Jun 27 0.00261 Wendt 0.0487 <0.0002 Jun 27 0.0032 Wendt 0.0452 <0.0002 0.002 Be3 Wendt 0.0493 <0.0002 0.005 Jun 28 0.001 0.014 Wendt <0.001 <0.0002 Jun 28 0.029 0.0055 Wendt 0.0662 Cr Jun 28 0.005 0.002 0.007 0.0511 0.0036 0.002 Jun 28 0.0033 0.008 0.002 0.0494 0.0016 Jun 28 0.0005 Cu 0.134 0.0006 0.004 0.0512 <0.0002 0.0005 0.004 0.152 0.001 <0.001 0.0509 <0.0002 0.003 0.004 0.164 0.0025 0.156 Fe <0.001 <0.0002 0.003 0.002 0.005 0.006 0.012 25.6 <0.001 0.004 0.159 0.001 0.006 32.5 0.005 0.0044 Mn 0.009 35.8 0.008 0.005 33.9 8.2 0.0017 0.0008 0.15 0.003 8.6 0.0007 34.3 Sr 0.151 9 0.152 15.8 8.6 0.0006 0.008 0.152 15.5 0.005 0.006 8.7 0.152 Zn 0.003 32.9 1.2 9.4 9.9 33.5 33.2 0.004 1.4 33.1 9.8 Ca 8.3 1.3 1.3 33.1 8.3 8.4 8.5 1.3 Mg 9.3 8.4 9.3 9.5 9.6 Na 1.4 9.5 1.3 1.3 1.3 K 1.3 MICROCYSTINS IN LOWER GREAT LAKES 137 initiation of a dense algal bloom may reflect the time required for the algal biomass to accumulate or could be related to differing hydrometeorolog- ical and ecosystem processes. The growth of Microcystis in earlier years may have gone unnoticed. Plus, the overwintering of algae on sediments may require time for zebra mussel pseudofaeces to cover sediments toxic in sulphides. Another difference between the sites that needs more analysis is the availability of iron. Iron concentrations were low at Lake Erie but likely were higher in Hamilton Harbour where for the last few years advanced sewage treatment has used much more iron to remove phosphorus. Iron has long been known to stimulate blue-green algae (Murphy et al. 1976; Hyenstrand et al. 2000). Furthermore, the timing of the current spread of botulism outbreaks is not understood. The kill has moved progressively eastward. It appears that it is not moved simply by water flow, as the outbreaks have bypassed Lake St. Clair. If introduced species are the focus of these toxic outbreaks, and somehow linked to microcystins, then the low biomass of Microcystis in Lake Erie raises questions. It is certainly possible for microcystins to be bioaccumulated (Kotak et al. 1996; Williams et al. 1997a,b; Thostrup and Christoffersen 1999). Bioaccumulation of microcystins has been proposed to explain sea-pen disease in aquaculture (Andersen et al. 1993). The weakness in this argument is that it assumes that the bioaccumulated microcystin is toxic. As noted above, the toxin is present covalently bound to phosphatase but the toxicity of the complex has not been reported. Few labs have the ability to measure covalently bound microcystins and its stability after ingestion is unknown. There are similar uncertainties associated with the link between microcystins and Clostridium. The microcystin dose required to kill via oral ingestion is much higher than via injection but the potential for in vivo growth of Clostridium should also be considered.

Public Consultation The American Water Association published an excellent review of the protocols that should be coordinated between governments, medical staff and pubic media so that the public both understands the risks and takes appropriate precautions (Yoo et al. 1995). More recently this was modified by Backer (2002). In the Great Lakes, options for monitoring might include remote sensing with appropriate validation. It is not clear how some agencies that have undergone extensive downsizing can respond to the challenge of monitoring and reporting to the relevant agencies such as water treatment facilities. Water treatment can remove algal toxins but activated carbon is required and there are concerns about how chlorination is done (Lambert et al. 1996; Tsuji et al. 1997). In situations like Hamilton Harbour where the toxic algae grow in the harbour and the water intakes are in the lake, techniques to prevent the algal scums from leaving the harbour might provide further protection for the drinking water. 138 MURPHY ET AL.

Acknowledgements

Ines Guerrero and Tanya Parr conducted many of the laboratory analyses. John Freidhoff was the captain of our vessel from SUNY. Mary Perrelli and Erica Somogye from SUNY assisted with sampling.

References

Andersen RJ, Luu HA, Chen DZX, Holmes CFB, Kent ML, LeBlanc M, Taylor FJRM, Williams DE. 1993. Chemical and biological evidence links microcystins to salmon ‘netpen liver disease’. Toxicon 31(10):1315–1323. APHA. 1989. Standard Methods for the Examination of Water and Wastewater. 17th ed. Backer LC. 2002. Cyanobacterial harmful algal blooms (cyanohabs): developing a public health response. Lake and Reserv. Manage. 18(1):20–31. Baganz D, Staaks G, Steinberg C. 1998. Impact of the cyanobacteria toxin, microcystin-LR on the behavior of zebra fish, Danio rerio. Water Res. 32:948–952. Brand C. 2001. Review of botulism. Presentation at Avian Botulism Workshop, Erie, Pennsylvania, Jan 24–25. Co-sponsored by Pennsylvania and New York Sea Grant. http://www.pserie.psu.edu/seagrant/extension/Avian_ Botulism_Workshop.pdf Brittain SM, Wang J, Babcock-Jackson L, Carmichael WW, Rinehart KL, Culver DA. 2000. Isolation and characterization of microcystins, cyclic heptapeptide hepatotoxins from a Lake Erie stain of Microcystis aeruginosa. J. Great Lakes Res. 26(3):241–249. Burger T, Bruning C, Collombon M, Ibelings BW, Reeders HH. 1997. Ecologishce effecten van cyanobacterietoxines. Aquasense 1090 HC Amsterdam. Bury NR, Eddy FB, Codd GA. 1996. Stress response of brown trout, Salmo trutta L., to the cyanobacterium Microcystis aeruginosa. Environ Tox. Water Qual. 11:187–193. Carbis CR, Mitchell GF, Anderson JW, McCauley I. 1996. The effects of microcystins on the serum biochemistry of carp, Cyprinus carpio L., when the toxins are administered by gavage, immersion and intraperitoneal routes. J. Fish Disease 19:151–159. Chapman AD, Schelske CL. 1997. Recent appearance of Cylindrospermopsis (Cyanobacteria) in five hypertrophic Florida lakes. J. Phycology 33:219–226. Charlton MN, LeSage R. 1996. Water quality trends in Hamilton Harbour: 1987–95. Water Qual. Res. J. Canada 31:473–484. Charlton MN, LeSage R, Milne JE. 1999. Lake Erie in transition: the 1990’s. In Munawar M, Edsall T, Munawar IF (ed.), State of Lake Erie (SOLE) – Past, Present and Future, p. 97–123. Bachhuys Publishers, Leiden, The Netherlands. Chorus I. 2002. Cyanobacterial toxin research and its application in Germany: a review of the current status. Environ. Toxicol. 17:358–360. Eriksson JE, Meriluoto JAO, Lindholm T. 1989. Accumulation of a peptide toxin from the cyanobacterium Oscillatoria agardhii in the freshwater mussel Anadonta cygnea. Hydrobiologia 183:211–216. Falconer IR. 1999. An overview of problems caused by toxic blue-green algae (cyanobacteria) in drinking and recreational water. Environ. Toxicol 14:5–12. Fritz L, Quilliam MA, Wright JLC, Beal AM, Work TM. 1992. An outbreak of domoic acid poisoning attributed to the pennate diatom Pseudonitzschia australis. J. Phycol. 28:439–442. MICROCYSTINS IN LOWER GREAT LAKES 139

Hallegraeff GM. 1993. A review of harmful algal blooms and their apparent global increase. Phycologia 32(2):79–99. Horgan MJ, Mills EL. 1997. Clearance rates and filtering activity of zebra mussel (Dreissena polymorpha): implications for freshwater lakes. Can. J. Fish. Aquat. Sci. 54:249–255. Hyenstrand P, Nyvall P, Pettersson A, Blomqvist P. 1998. Regulation of non-nitrogen-fixing cyanobacteria by inorganic nitrogen sources — experiments from Lake Erken. In Forsberg C, Petterson K (ed.), Lake Erken — 50 years of lim- nological research. Archiv fuer Hydrobiologie 51:29–40. Hyenstrand P, Rydin E, Gunnerhed M. 2000. Response of pelagic cyanobacteria to iron additions – enclosure experiments from Lake Erken. J. Plankton Res. 22:1113–1126. Kotak BG, Zurawell RW, Prepas EE, Holmes CFB. 1996. Microcystin-LR concentration in aquatic food web compartments from lakes of varying trophic status. Can. J. Fish. Aquat. Sci. 53:1974–1985. Lambert TW, Holmes CFB, Hrudey SE. 1996. Adsorption of microcystin-LR by activated carbon and removal in full scale water treatment. Water Res. 30:1411–1422. Lehman JT, Brooks AS, Zastrow JC. 2000. Water ecology, p. 43–49. In Sousunis PJ and Bisanz JM (ed.), Preparing for a climate change: a summary by the Great Lakes Regional Assessment Group. U.S. Environmental Protection Agency, Washington, D.C. Matsunaga H, Harada KI, Senma M, Ito Y, Yasuda N, Ushida S, Kimura Y. 1999. Possible cause of unnatural mass death of wild birds in a pond in Nishinomiya, Japan: sudden appearance of toxic cyanobacteria. Natural Toxins 7:81–84. Murphy TP, Brownlee BG. 1981. Blue-green algal ammonia uptake in hypertrophic prairie lakes. Can. J. Fish. Aquat. Sci. 38:1040–1044. Murphy TP, Lean DRS, Nalewajko C. 1976. Blue-green algae: their excretion of iron selective chelators enables them to dominate other algae. Science l92:900–902. Murphy TP, Lawson A, Nalewajko C, Murkin H, Ross L, Oguma K, McIntyre T. 2000. Algal toxins – initiators of avian botulism? Environ. Tox. 15:558–567. Nagata S, Soutome H, Tsutsumi T, Hasegawa A, Sekijima M, Sugamata K. NYS Department of Environmental Conservation, January 2002, Botulism Type E in Lake Erie, internal report. NLET. 1994. Manual of analytical methods. Volume I and II, National Laboratory for Environmental Testing, Environment Canada. Oberemm A, Fastner J, Steinberg CEW. 1997. Effects of microcystin-LR and cyanobacterial crude extracts on embryo-larval development of zebra fish (Danio rerio). Water Res. 31:2918–2921. Sivonen K. 1990. Effects of light, temperature, nitrate, orthophosphate and bacteria on growth of and hepatotoxin production by Oscillatoria aghardii strains. Appl. Environ. Microbiol. 56:2658–2666. St. Amand A. 2002. Cylindrospermopsis: an invasive toxic alga. Lakeline 22(1): 36–37. Sturm K. 1998. Avian botulism outbreaks in Salton Sea, California – a significant risk to brown pelicans. Presentation to the Binational Avian Botulism Task Force, Madison, Wisconsin, April 30, 1998. Sutton MA. 2002. Introduction: fluxes and impacts of atmospheric ammonia on national, landscape and farm scales. Environ. Pollut. 119:7–8. Takahashi S, Kaya K. 1993. Quail spleen is enlarged by microcystin-RR as a blue- green algal hepatotoxin. Nat. Toxins 1(5):283–285. 140 MURPHY ET AL.

Tencalla F, Dietrich D. 1997. Biochemical characterization of microcystin toxicity in rainbow trout (Oncorhynchus mykiss). Toxicon 35(4):585–595. Thostrup L, Christoffersen K. 1999. Accumulation of microcystin in Daphnia magna feeding on toxic Microcystis. Arch. Hydrobiol. 145:447–467. Tsuji K, Watanuki T, Kondo F, Watanabe MF, Suzuki S, Nakazawa H, Suzuki M, Uchida H, Harada K. 1995. Stability of microcystins from cyanobacteria – II. Effect of UV light on decomposition and isomerization. Toxicon 33(12): 1619–1631. Tsuji K, Watanuki T, Kondo F, Watanabe MF, Nakazawa H, Suzuki M, Uchida H, Harada K. 1997. Stability of microcystins from cyanobacteria – IV. Effect of chlorination on decomposition. Toxicon. 35:1033–1041. Vanderploeg H, Johengen T, Carmichael W, Liebig J, Nalepa T, Stickler J. 1998. How zebra mussels promote Microcystis bloom: experience gained from Saginaw Bay (Lake Huron) and Lake Erie. International Association of Theoretical and applied Limnology XXVII Congress, Dublin, abstract. Watanabe MF, Park HD, Kondo F, Harada K. 1997. Identification and estimation of microcystins in freshwater mussels. Natural toxins 5:31–35. Watson SB, Kling H, Charlton MN, Murphy T. Toxic algal outbreaks in remedi- ated surface waters: Hamilton Harbour revisited. Submitted. Wiegand C, Pflugmacher S, Obermann A, Meems N, Beattie KA, Steinberg CEW, Codd GA. 1999. Uptake and effects of microcystin-LR on detoxifica- tion enzymes of early life stages of the zebra fish (Danio rerio). Environ. Toxicol. 14:89–95. Williams DE, Craig M, Dawe SC, Kent ML, Holmes CFB, Andersen RJ. 1997a. Evidence for a covalently bound form of microcystin-LR in salmon liver and dungeness crab larvae. Chem. Res. Toxicol. 10:463–469. Williams DE, Dawe SC, Kent ML, Andersen RJ, Craig M, Holmes CFB. 1997b. Bioaccumulation and clearance of microcystins from salt water mussels, Mytilus edulis, and in vivo evidence for covalently bound microcystins in mussel tissues. Toxicon 35:1617–1625. Wobeser G, Murphy TP. 1999. Results of exposing Mallard duckling to microcystin-LR. Report to the binational working group on avian botulism. 4 p. Wulff A. 2001. Is there a coupling between increased UVB radiation and toxic algal blooms. SIL 2001 XXVIII Congress, Melbourne, Australia, abstract. Yoo RS, Carmichael WW, Hoehn RC, Hrudey SE. 1995. Cyanobacterial (blue-green algal) toxins: a resource guide. American Water Works Association. 230 p. Zurawell R, Prepas EE. 1995. Occurrence and accumulation of microcystin-LR in freshwater pulmonate snails. Lake Reserv. Manage. 11(2):206. Water Qual. Res. J. Canada, 2003 Volume 38, No. 1, 141–152 Copyright © 2003, CAWQ

Mass Appearance of Cyanobacterium Planktothrix rubescens in Lake Piaseczno, Poland

DANUTA KRUPA AND KRZYSZTOF CZERNA´S*

Department of General Ecology, University of Agriculture, Akademicka 15, 20-950 Lublin, Poland

In 1989, Lake Piaseczno, Poland, exhibited a mass appearance of Planktothrix rubescens. During this time the pelagic and littoral areas exhibited significant increases in areal primary production (400 and 41 mg C m-2 h-1, respectively), chlorophyll a (100 and 6.9 mg m-2, respectively) and assimilation number (4 and 5.9 mg C m-2 h-1/mg chla m-2, respectively). After the water bloom subsided, a reduction of dissolved oxygen concentration (down to 1.5 mg L-1) and high water temperature (10.2°C) in the offshore bottom zone was observed. While from 1991 to 1996, the primary production, chlorophyll a concentration and assimilation number values were decreasing, they were significantly higher than the values reported in 1986, prior to the mass cyanobacteria appearance. An indirect correlation with ion levels indicated that the outbreak of the cyanobacteria was linked with inflow of nutrients from the catchment area. The dramatic changes in the range and variability of the phytoplankton density indicate that the recent eutrophication of the lake has had profound effects on the structure and productivity of the aquatic community.

Key words: Lake Piaseczno, cyanobacterial bloom, Planktothrix rubescens, primary production, chlorophyll a, assimilation number, eutrophication

Introduction

Eighty years ago, Lake Piaseczno located in the ´L˛eczna-W´l odawa Lake District, Eastern Poland, was still considered an oligotrophic lake (Lityn´ski 1919). Nearly sixty years later, Brz˛ek et al. (1975) defined it as α- mesotrophic. In the years that followed, the extensive anthropogenic use of the lake and the transformations taking place in its immediate vicinity intensified the process of degradation. The lake’s catchment area of 284.9 ha underwent significant changes. From 1976 to 1984, the area of forests increased by 11%, peat bogs and meadows by 22%, and recreational areas and roads by as much as 88%. The area of arable land and orchards decreased considerably by 73% (Furtak and Turczyn´ ski 1998). The farms located on the sandy soils by the eastern lakeshore were replaced by summer houses without sewage systems. Around the lake, 210 summer houses were built prior to 1986. Their numbers increased to 300 in 1990 and 570 in 1996. The simultaneous lowering of the water level resulted in reduction of the lake area by 6.2%, exposing the blanket peat bog adjacent to it.

* Corresponding author; [email protected] 142 KRUPA AND CZERNAS´

Draining and subsequent mineralization of the peat bog intensified the influx of nutrients into the lake water and increased its trophic state (Wojciechowski and Górniak 1990). Research focussing on the phytoplankton of Lake Piaseczno conducted since the late 1960s indicated that until the mid-1980s, filamentous cyanobacteria were only occasionally found in the lake. They were represented by only a few species, and their density was low. Cyanobacteria from Chroococcales order were much more abundant in the lake’s phytoplankton. In the first half of the 1970s, during springtime, the phytoplankton was often dominated or subdomi- nated by Microcystis incerta Lemm. and Aphanothece clathrata W. et G.S. West (Wojciechowska and Krupa 1980). Until as late as 1986, no Planktothrix rubescens (D.C. ex Gom.) Anagn. et Kom. (previously known as Oscillatoria rubescens D.C. Gom.) was reported in the lake. The objective of this paper was to determine the process of degradation of the water quality of Lake Piaseczno resulting from the intensifica- tion of anthropogenic factors, as demonstrated by the appearance and domination of phytoplankton by Planktothrix rubescens.

Materials and Methods

This paper presents the results of research undertaken in the spring, summer and autumn of 1986, 1989, 1991, 1993 and 1996 (samples collected 4 to 6 times each year) in Lake Piaseczno (51°23’N, 23°02’E, 170.6 m a.s.l., area of 68.47 ha, maximum depth 38.8 m) in the ´L˛eczna-W´l odawa Lake District of Poland. Water samples for physico-chemical analysis were taken at the same time that phytoplankton was sampled. The physico-chemical characteristics measured in the pelagic and littoral zones included Secchi disc transparency, temperature, pH, conductivity and oxygen concentration (Winkler method). Dissolved nutrients were also measured in samples from both sites, after filtration with a Filtrak 390 unit (nominal pore size 0.45 µm). These were then analyzed for NH4-N (Nessler method), NO3-N (using phenoldisulfonic acid), PO4-P (using ammonium molybdate; Golterman 1971), and K (measured photometri- cally). Ionic concentration was also measured in the groundwater of the piezometer well located 10 m from the lake shoreline. The biological investigation included quantitative and qualitative structure of phytoplankton, gross primary production and chlorophyll a concentration of the plankton. The samples for phytoplankton determinations were drawn in the pelagic area at the maximum depth, then separately for each thermal layer (on 22 dates), and in the shallow littoral area with water depth of 0.75 m (on 28 dates). Live material was used for taxonomic identification. The quantitative composition of phytoplankton was determined in samples fixed with Lugol’s solution and then with 4% formalin and glyc- erin, with an inverted microscope according to Utermöhl’s method (Vollenweider 1969). The gross primary production of phytoplankton was estimated by the oxygen method in light and dark bottles during four- hour exposures at the sampling sites (in the pelagic zone at depths of 0, P. RUBESCENS IN LAKE PIASECZNO, POLAND 143

2.5, 5.0, 7.5, 10.0 and 12.5 m, and in the shallow littoral area at a depth of 0.75 m). The chlorophyll a concentration was measured from an ethanol extract of the filtrate homogenized together with the GF/C filter, using a Specol Zeiss 11 spectrophotometer at 665 and 750 nm (Nusch 1980).

Results

Over the entire period of research, the lake demonstrated the attrib- utes of dimictic lakes, with a well-developed epilimnion, metalimnion and hypolimnion. Until 1989 and in 1996, water temperature in the offshore bottom zone ranged from 5.0 to 7.4°C. In the summers of 1989 and 1993, a considerable temperature increase was recorded at the bottom (on 1 June 1989, the temperature was 11°C, and on 15 July 1993, 10.2°C). Conductivity and pH of the pelagic and littoral water did not show any significant changes and over the entire period of research it ranged from pH 7.1 to 7.5, and for conductivity, from 85 to 104 µS cm-1. Water transparency, measured at the maximum depth, ranged from 2.5 m in the autumn of 1989 to 6.2 m in the spring of 1989 and 1993. The average annual values were 5 m (Fig.1). Oxygen concentration at the individual sampling depths was always high. For example, on 24 September 1986, at the depth of 1 m, it was 9.5 mg L-1, and at the bottom, 6.37 mg L-1. In September 1989, for the first time, a significant reduction in oxygen concentration was recorded at the depth of 12.5 m, with the bottom zone concentration of only 1.5 mg L-1. A similar situation occurred again in 1991 and 1996 when the oxygen content of bottom-zone water was only 0.76 and 1.02 mg L-1, respectively (Fig. 2).

Fig. 1. Annual minimum, maximum and mean values of Secchi transparency in the pelagic zone of Lake Piaseczno. 144 KRUPA AND CZERNAS´

Fig. 2. Oxygen concentration in offshore bottom zone of Lake Piaseczno.

The concentration of NH4-N in the pelagic and littoral water ranged from 0.1 to 0.26 mg L-1, except for 1996, when it reached the highest value -1 of 0.31 mg L in the pelagic zone. The NO3-N concentration at both sampling sites did not vary greatly and ranged from 0.05 to 0.06 mg L-1. The -1 highest concentration of 0.12 mg L NO3-N was recorded in the pelagic zone in 1993. The concentration of PO4-P increased over the entire research period from 0.03 mg L-1 in 1986 and 1989, to 0.13 mg L-1 (in the pelagic zone) and 0.12 mg L-1 (in the littoral) in 1996. The concentration of potassium varied only slightly, with the highest value of 3.64 mg L-1 recorded both for littoral and open water in 1989. The groundwater from the piezometer, its chemical composition the result of the water infiltrat- ing from the surface and lake water (Misztal and Smal 1991), had many times higher concentrations of all ions tested (Table 1). The analysis of phytoplankton density in the selected years of the decade (1986–1996) demonstrates the rapid algal increase both in the littoral and pelagic areas in 1989 (Fig. 3). This was characterized by the considerable increase of the cyanobacteria density resulting from mass appearance of Planktothrix rubescens. The presence of this species in Lake Piaseczno was reported for the first time in 1986. From 1986 to 1989, the average annual density of cyanobacteria increased 3.5-fold in the littoral area, and in the pelagic area by 4.5-fold (Fig. 4). In the summer of 1989, the high density of P. rubescens continued both in the littoral and the pelagic regions (Fig. 5 and 6). In the littoral zone, the highest density was recorded in March and April. It ranged from 2.1 million to 2.7 x 106 L-1, thus accounting for 85% to 92.9% of the total phytoplankton density. In the pelagic zone, the maximum density of P. rubescens was reported in late April in the epilimnion and metalimnion, 1.3 and 1.5 x 106 filaments L-1, P. RUBESCENS IN LAKE PIASECZNO, POLAND 145 ) -1 1986 1989 1993 1996 8.22 2.85 2.85 13.75 3.64 3.64 3.24 2.64 2.55 3.32 2 2 water water water water water water water water water water water water Ground LittoralPelagic Ground Littoral Pelagic Ground Ground Littoral Pelagic Ground Littoral Pelagic Mean concentration of N, P, and K in the littoral and pelagic zones of Lake Piaseczno and in the piezometer groundwater (mg L and K in the littoral pelagic zones of Lake Piaseczno piezometer groundwater Mean concentration of N, P, -N 1.41 0.1 0.1 3.8 0.26 0.26 1.33 0.14 0.16 1.27 0.14 0.31 -N 2.82 0.05 0.05 6.56 0.05 0.05 0.064 0.029 0.12 0.13 0.045 0.06 -P 0.315 0.03 0.03 1.22 0.03 0.03 0.08 0.06 0.13 0.145 0.13 0.12 4 3 4 able 1. NH T NO PO K 146 KRUPA AND CZERNAS´

Fig. 3. Mean annual phytoplankton density in the littoral and pelagic zones of Lake Piaseczno. respectively, or 82.3% and 92.6% in the total phytoplankton density. At the end of spring, the density of the species decreased considerably in the littoral, while in the pelagic area, it moved to the cold water of the hypolimnion, to reach the level of 1.55 x 106 filaments L-1 (95.4% of total density). In the summer and autumn, P. rubescens nearly completely left the littoral, while in the lake’s pelagic hypolimnion it reached the density of 2.0 x 106 filaments L-1, or 83.4% of total density of the phytoplankton. In the following years, the density of P. rubescens was very low and it was not

Fig. 4. Mean annual cyanobacteria density in the littoral and pelagic zones of Lake Piaseczno. P. RUBESCENS IN LAKE PIASECZNO, POLAND 147

Fig. 5. Density of Planktothrix rubescens in the littoral of Lake Piaseczno. until the autumn of 1996 that it increased in the littoral region to 1.84 x 105 filaments L-1, and in the pelagic epilimnion to 1.91 x 105 filaments L-1. The values reflecting the primary production measured in the littoral and the pelagic zones are presented as annual averages for the individual depths and seasons (Fig. 7). In the littoral area, they were significantly lower than in the pelagic zone, however, over the entire period of research, similar trends were seen. The highest primary production values were recorded in 1989. They were over four times higher than the primary

Fig. 6. Density of Planktothrix rubescens in the pelagic zones of Lake Piaseczno. 148 KRUPA AND CZERNAS´

Fig. 7. Gross primary production in the littoral and pelagic zones of Lake Piaseczno. production in 1986, and totalled 41 mg C m-2 h-1 in the littoral and 400 mg C m-2 h-1 in the pelagic zones. The chlorophyll a concentration in the phytoplankton was measured, as in the case of primary production, at the same sampling sites, and presented as annual averages (Fig. 8). The lowest areal chlorophyll a concentration values were reported both for the littoral and the open water zones in 1986, 4.7 and 58 mg m-2, respectively. The highest value for the littoral zone in 1996 was 7.88 mg m-2 and for the pelagic zone in

Fig. 8. Chlorophyll a concentration in the littoral and pelagic zones of Lake Piaseczno. P. RUBESCENS IN LAKE PIASECZNO, POLAND 149

1989 was 100.02 mg m-2. The assimilation number was also calculated (ratio of primary production in mg C m-2 h-1 and chlorophyll a concentration in mg m-2, Ichimura 1968), as an indicator of community productivity. Prior to 1989, it did not exceed 2.0 both for the littoral and the pelagic areas, which indicated a relatively low fertility of the lakewater. In the following years, the values went up, to reach littoral zone levels of 5.94 and 6.21, and planktonic levels of 4.0 and 3.79 in 1989 and 1993, respectively.

Discussion

During the earlier long-term research of Lake Piaseczno phytoplankton, mass appearance of cyanobacteria was never recorded. Even at the time when it dominated the phytoplankton of the lake (this applies to the Chroococcales order), its density and biomass did not reach levels typical of water bloom. In the latter half of the 1980s, the density of filamentous cyanobacteria, Oscillatoria sp., Limnothrix redeckei (Van Goor) M.-E. Meffert (previously known as Oscillatoria redeckei Van Goor), and Planktothrix rubescens, increased considerably (Wojciechowska and Krupa 1992). Sosnowska (1974) reported the presence of P. rubescens in lakes with a varying degree of trophic nature, from oligotrophy to eutrophy. This cyanobacterium dominates the phytoplankton of mesotrophic Lake Zürich forming a metalimnetic layer during the summer thermal stratifi- cation (Walsby 2001). Lang and Reymond (1993) reported it for eutrophic Swiss lakes. The appearance of this species as a regular phytoplankton component in Lake Maggiore and the increase in its density was recognized as a sign of the lake’s eutrophication (Ravera and Vollenweider 1968). Planktothrix rubescens blooms can have harmful effects for animals

Fig. 9. Assimilation number (mg C m-2 h-2/mg Chla m-2) for the littoral and pelagic zone in Lake Piaseczno. 150 KRUPA AND CZERNAS´ and human health. Some of the strains of this cyanobacterium produce the hepatotoxic microcystins (Fastner et al. 1995, 1998) The mass appearance of P. rubescens in Lake Piaseczno was pre- ceded by many years of adverse impact upon the lake: intensified recreation (increase in the number of tourists and pollutant input to the lake water) and groundwater lowering. From 1981 to 1996 the levels of the groundwater table and lake water lowered by 1.3 and 1.6 m, respectively (Michalczyk 1998). It was related to poor rainfall from 1981 to 1988. The mean value of precipitation for this period reached 450 mm and was significantly lower than the average from 1974 to 1980, which was 600 mm (Kaszewski et al. 2001). Exploitation of groundwater from wells used for recreation and nearby hard-coal mines, as well as the formation of shallow depressions in the area of the mines contributed further to the lowering of the water level. The lowering of the water level also caused some draining and mineralization of the peat bog adjacent to the lake. These phenomena intensified in the latter half of the 1980s. While from 1976 to 1984, the area of the lake decreased by 6.2%, from 1984 to 1992 it shrank by as much as 13.5%, thus resulting in further exposure of the littoral (Furtak and Turczyn´ ski 1998). The rapid increase in the density of P. rubescens was accompanied by the increase of other biological parameters, such as primary gross production, chlorophyll a concentration and reduced oxygen concentration in the offshore bottom zone of the lake. During the earlier long-term research (Lecewicz et al. 1973; Radwan and Kowalczyk 1979; Radwan et al. 1987; Wojciechowski et al. 1995), until as late as 1989, no oxygen deficit in the offshore bottom zone was recorded. There is some indirect evidence that the outbreak of P. rubescens was better correlated with the groundwater concentrations of NO3-N (r = 0.75; p< 0.05; N = 16) and K (r = 0.69; p < 0.05; N = 16), which both showed the highest levels in 1989, than those levels seen in the lake water (Table 1). These correlations appeared to be stronger than with other nutrients (PO4- P: r = 0.35, p < 0.05, N = 16; NH4-N: r = 0.47, p < 0.05, N = 16). Earlier research conducted in the lake demonstrated a correlation between the primary production and chlorophyll a concentration, on one hand, and the chemical composition of groundwater, on the other, as well as the absence of such correlation with the composition of lake water (Czernas´ 2001). Perhaps the determination of the chemical composition of groundwater draining to the lake from its catchment area could enable better projection of water degradation processes than the commonly used determination of the chemical composition of the lake’s water. The abate- ment of the P. rubescens did not result in a reduction of the primary production and chlorophyll a concentration levels to those recorded before the bloom, and while the quantity of oxygen in the offshore bottom zone went up compared to the pre-1989 levels, it continued at a low level. In years following the bloom, Rhodomonas pusilla Bachmann (Javorn.), Chromulina sp., Coenococcus planctonicus Conrad and Chlorella vulgaris P. RUBESCENS IN LAKE PIASECZNO, POLAND 151

Beij. were predominant, with maximal shares in the total phytoplankton density of 68.7%, 56.2%, 51.5% and 21%, respectively. These dominant nannoplankters were responsible for the continued high productivity levels from 1991 to 1993.

Acknowledgements

We wish to thank Modest Misztal for providing us with the results of ion concentration determinations conducted by the chemical laboratory of the Institute of Soil Science and Natural Environmental Development of the University of Agriculture in Lublin.

References

Brz˛ek G, Kowalczyk C, Lecewicz W, Radwan S, Wojciechowska W, Wojciechowski I. 1975. Influence of abiotic environmental factors on plankton in lakes of different trophy. Pol. Arch. Hydrobiol. 22:123–139. Czernas´ K. 2001. Productivity of the psammic algal communities in the near shore zone of the mesotrophic Lake Piaseczno (Eastern Poland). Water Qual. Res. J. Canada 36:537–564. Fastner J, Flieger I, Neumann U. 1998. Optimized extraction of microcystins from field samples–a comparison of different solvents and procedures. Water Res. 32:3177–3181. Fastner J, Heinze R, Chorus I. 1995. Microcystin-content, hepatotoxicity and cyto- toxicity of cyanobacteria in some German water bodies. Water Sci. Tech. 32:162–170. Furtak T, Turczyn´ski M. 1998. Changes in catchement area land development as exemplified by Lake Piaseczno, p. 145–148. In Harasimiuk M, Michalczyk Z, Turczyn´ski M (ed.), Lakes of ´L˛eczna-W´l odawa Lake District. Natural Science Monograph (In Polish). Publ. Univ. Mariae Curie-Sk´l odowska. Golterman HL. 1971. Methods for chemical analysis of fresh waters. Oxford- Edinburgh, Blackwell, IBP Handbook No. 3. Ichimura S. 1968. Phytoplankton photosynthesis. Algae, man and the enviroment. Syracuse. Kaszewski BM, Michalczyk Z, Siwek K. 2001. Climatic conditions of water cir- culation, p. 23–33. In Michalczyk Z (ed.), Springs of Lublin Upland and Roztocze (In Polish). Publ. Univ. Mariae Curie Sklodowska. Lang C, Reymond O. 1993. Trends in phytoplanktonic and zoobenthic communities after the decrease of phosphorous concentration in Lake Joux. Rev. Suisse Zool. 100, 4:907–912. Lecewicz W, Soko´l owska W, Wojciechowski I. 1973. The changes of winter phytoplankton in relation to the light climate in the lakes with various trophy. Ekol. pol. 21:193–207. Lityn´ski A. 1919. Preliminary reports on investigations in the ´L˛eczna-W´l odawa Lakeland in July and August 1919 (In Polish). Przegl. Ryb. 1:173–181. Michalczyk Z. 1998. Water relations of the ´L˛eczna-W´l odawa Lake District, p. 55–71. In Harasimiuk M, Michalczyk Z, Turczyn´ ski M (ed.), Lakes of ´L˛eczna-W´l odawa Lake District. Natural Science Monograph (In Polish). Publ. Univ. Mariae Curie-Sk´lodowska. 152 KRUPA AND CZERNAS´

Misztal M, Smal H. 1991. The study of nitrogen and phosphorus concentrations in shallow ground waters of the catchment areas of lakes Piaseczno and G´l˛ebokie (´L˛eczna-W´l odawa Lake District, South-East Poland) (In Polish with English summary). Studia Os´r. Dok. Fizjogr. Polish Academy of Sciences, Kraków 19:209–218. Nusch AE. 1980. Comparison of different methods for chlorophyll and phaeopig- ment determination. Arch. Hydrobiol. Beih. Ergebn. Limnol. 14:143. Radwan S, Kornijów R, Kowalik W, Jarzynowa B, Zwolski W, Kowalczyk Cz, Popio´lek B. 1987. Ecological and fishery characteristics of lakes situated in the future Western Polesie National Park. Ann. Univ. Mariae Curie Sk´lodowska, sectio C, 42:163–183. Radwan S, Kowalczyk Cz. 1979. Chemical compounds of three trophically different lakes of the ´L˛eczna-W´l odawa Lakeland (In Polish with English summary). Ann. Univ. Mariae Curie-Sk´lodowska, sectio C, 34:229–260. Ravera O, Vollenweider RA. 1968. Oscillatoria rubescens D.C. as an indicator of Lago Maggiore eutrophication. Schweiz. Ztschr. Hydrol. 30:374–380. Sosnowska J. 1974. Plankton communities of three Masurian Lakes and chlorophyll content in their phytoplankton (In Polish with English summary). Monogr. Bot. 42:3–150. Vollenweider RA. 1969. A manual on methods for measuring primary production in aquatic environments. Oxford-Edinburgh, Blackwell (IBP Handbook No. 12). Walsby AE. 2001. Determining the photosynthetic productivity of a stratified phytoplankton population. Aquat. Sci. 63:18–43. Wojciechowska W, Krupa D. 1980. Changes in numbers and biomass of phytoplankton in a-mesotrophic Lake Piaseczno in the years 1971–1972 and 1975–1976. Ekol. pol. 28:231–243. Wojciechowska W, Krupa D. 1992. Many years’ and seasonal changes in phytoplankton of lakes of Polesie National Park and its protection zone. Ekol. pol. 40:317–332. Wojciechowski I, Czernas´ K, Krupa D. 1995. The biotic values and conditions of lakes in The Polesie National Park and its protection zone, p. 38–45. In Radwan S (ed.), Protection of freshwater ecosystems in the Polesie National Park and its protection zone (In Polish). Lublin, TWWP:38–45. Wojciechowski I, Górniak A. 1990. Influence of the brown humic and fulvic acids originating from nearby peat bogs on phytoplankton activity in the littoral of two lakes in Mid-Eastern Poland. Verh. Internat.Verein. Limnol. 24:295–297. Water Qual. Res. J. Canada, 2003 Volume 38, No. 1, 153–168 Copyright © 2003, CAWQ

Characterization of Particles in Slow Sand Filtration at North Caribou Water Treatment Plant

BEATA GORCZYCA* AND DAVID LONDON

Department of Civil Engineering, University of Manitoba, 342 Engineering Building, 15 Gillson Street, Winnipeg, Manitoba R3T 5V6

Microscopic analysis of particles in water can indicate the size of filter media required, and can be used to monitor filter performance. This study investigated a malfunctioning slow sand filter in a water treatment facility on a First Nations community in Northern Ontario. There has been a boil-water advisory in the community due to high turbidity in the drinking water since the plant was put into operation. Also, the slow sand filters in the plant clog frequently resulting in outflow volumes significantly below the plant design capacity. On-line particle counts and microscopic analysis of water were conducted in the plant at various points before and after the slow sand filters. Microscopic analysis of the slow sand filter effluent showed that the high water turbidity was due to an increased concentration of particles smaller than 2 µm in size. This observation could not have been made with the particle counter data alone, as it is not capable of measuring particles of that size. Visual inspection of microscopic images of these small particles indicated that they were being washed out from the filter media, and did not originate from the raw water. Significant numbers of large filter-clogging diatoms (up to 50 µm in size) were identified under the microscope; yet, the particle counter did not report particles larger than 15 µm in raw water supply. Turbidimeters and particle counters were found to be unable to identify these diatoms, due to the transparent nature of the cell walls of these microorganisms. It is likely that most of these diatoms originated from sediment accumulated in the raw water intake pipe.

Key words: slow sand filtration, particle counter, microscope analysis

Introduction

Slow sand filtration is a simple technology requiring no knowledge of coagulation chemistry, and is quite attractive for small installations treating high-quality surface waters. Although slow sand filters have been used in water treatment for over a century, they are recently gaining a lot of new interest. This is primarily because of their effectiveness in controlling a wide variety of microorganisms including Giardia cysts (Table 1). Source water quality is one of the key factors in determining whether slow sand filtration should be selected. Water quality guidelines for slow sand filtration indicate that waters acceptable for this treatment should have turbidity less than 10 NTU and chlorophyll a concentration (repre-

* Corresponding author; [email protected] 154 GORCZYCA AND LONDON

Table 1. Water quality guidelines for selection of slow sand filtration (SSF) (American Water Works Research Foundation 1991)

Parameter Removal experiences Guidelines and comments

True colour (TCU) 25% 5–10 Total organic carbon 25% None (removal depends on biofilm in filter) Turbidity (NTU) 27–39% <10 NTU Coliform bacteria 2-log to 4-log None (removal determined through pilot plant studies) Giardia cysts 3-log to 4-log None (removal determined through pilot plant studies) Algae <5 mg/m3 (surface mat)

senting algal content of surface water) of less than 5 mg/m3 (Table 1). The deficiency of raw water turbidity as an indicator of filter run length has been emphasized many times (American Water Works Research Foundation 1991). Lower turbidity does not necessarily mean longer filter run lengths and higher turbidity does not necessarily mean shorter filter run lengths. This is because turbidity gives no indication as to type, concentration, size, shape or refractive index of particles suspended in water. Turbidity is just an expression of the optical property that causes light to be scattered and absorbed by organic and inorganic matter, plankton and other microscopic organisms suspended in water (American Public Health Association 1998). This study analyzed a water treatment plant in North Caribou, a Northern Ontario community of 800 people. The plant flow schematic is shown in Fig. 1. The raw water is provided from Round Lake through a

Fig. 1. North Caribou Water Treatment Plant flow schematic. PARTICILES IN SLOW SAND FILTRATION 155

390-m long intake pipe with the diameter of 300 mm (Fig 2). The water in the intake pipe flows at a very slow rate mainly due to the size and the length of the intake pipe as well as the fact that the water is supplied to the plant on an intermittent basis. The plant consists of two parallel slow sand filters (SSF) followed by two granular activated carbon filters (GAC). Finally, the water is disinfected with sodium hypochlorite. The two slow sand filters in the plant each have a surface area of about 48 m2 (5.50 x 8.65 m). The filter media consists of a 1-m deep layer of sand with an effective grain size of 0.2 to 0.3 mm, and a 0.5-m deep layer of support gravel. The two GAC filters contain about a 1-m deep layer of activated carbon media with a particle size ranging from 0.4 to 1.7 mm. The slow sand filters were designed to operate at a 7.3 L/s flow rate, with the filter run length ranging from 3 to 6 months. Shortly after plant commissioning in December 2000, the flow from the filters had significantly decreased from 7.3 to about 2 L/s; also, the filter runs were only 2 to 3 weeks long, suggesting that the slow sand filters were clogging prematurely. The quality of raw water in Round Lake is presented in Table 2. The raw water meets the Canadian Drinking Water Guidelines with the excep-

Fig. 2. Water intake structure (Neegan Burnside Engineering and Environmental Ltd.). 156 GORCZYCA AND LONDON

Table 2. Round Lake raw water quality—major parameters (The Burnside Group of Companies 1999)

Parameters Valuea pH 7.6–7.98 Colour (TCU) 10–30 Turbidity (NTU) 0.92–3.8 Hardness (mg/L as CaCO3) 45.2–51.7 Alkalinity (mg/L as CaCO3)44 Nitrates (mg/L) 0.1–0.47 Total organic carbon (mg/L) 3–12 Total dissolved solids (mg/L) 56

a Data based on measurements on 5 samples taken in 1999, 1987, and 1996. tion of turbidity and colour (Table 3). Table 4 shows turbidity readings of raw water, slow sand filtrate and finished water during the summer of 2001. The raw water turbidity is measured continuously at the raw water pumps’ discharge headers by a HACH turbidimeter (model 1720D). The treated water turbidity is also measured continuously on the water effluent from the GAC filters. Slow sand filter effluent turbidity was measured with a portable turbidimeter (HACH, Model 2100C). Data collected during the summer of 2001 indicated that the slow sand filter performance was inconsistent with regard to the raw water turbidity and colour removal. Higher turbidity in the outflow, relative to the inflow of the slow sand filter was reported (Table 4). The rising turbidity of the post filter water was the main reason for the Boil Water Advisory issued by Health Canada on September 24, 2001. The objective of this work was to study the origin of particles causing filter clogging and high water turbidity in the plant.

Materials and Methods

On December 18 to 19, 2001, the authors traveled to North Caribou, Ontario, the site of the treatment plant in question. A number of water sam-

Table 3. Canadian drinking water guidelines

Canadian drinking water guidelines Guidelines (Health Canada 1996)

Colour ≤15 Turbidity 1.0 PARTICILES IN SLOW SAND FILTRATION 157

Table 4. Water quality summaries in North Caribou May to September 2001

Parametersa Turbidity (NTU) Colour (ACU) pH

Raw water May 0.63 44 7.8 June 0.87 43 7.8 September 1.06 34 7.8 Slow sand filter (SSF) effluent SSF1 SSF2 SSF1 SSF2 SSF1 SSF2 May 0.83 1.24 June 2.68 0.96 36 8.2 September 1.71 1.08 35 31 7.8 7.9 Finished water May 0.47 27 7.7 June 0.52 23 7.8 September 0.93 20 7.6

a Averages based on about 30 samples. ples were taken during the two-day period: raw water, slow sand filter (SSF) outflow, GAC filter outflow, and water leaving the plant to the distribution system. For each sample, turbidity measurements, particle counts and microscopic analysis were conducted. In addition, microscopic analysis of water washed out from slow sand filter media taken from the bag originally supplied by a contractor was conducted. A sample of raw water was sent for algal identification and enumeration (Korzeniowski 2001).

Particle Counts An on-line particle counter (model PCX by Met One) is designed to measure particles of diameters ranging from 2 to 500 µm. The counter was connected to the raw water supply, slow sand filter effluent (SSF1), GAC filter effluent and distribution water line. The numbers of particles in the following size ranges (µm) were measured: 2.0 to 3.0, 3.1 to 5.0, 5.1 to 7.0, 7.1 to 10.0, and 10.1 to 15.0. Initially, there was an interval of particles with sizes 15.0 µm and larger, but the counter consistently returned an insignificant number of particles larger than 15 µm (less than 10 particles per 100 mL).

Microscope Analysis

Sample preparation Microscope analysis usually requires only a few drops of water sample. Since the particle concentration in the analyzed water was low, it was 158 GORCZYCA AND LONDON difficult to have a representative number of particles in a few drops of water volume. To increase the concentration of particles in the sample, the following procedure was applied. A large clean container was filled with about 20 L of water sample, covered, and set aside for 12 h. During that time most of the particles causing water turbidity would have settled to the bottom of the container. Following the settling period, the water at the top of the containers was decanted, leaving about 1 L of sediment concentrate at the bottom. The concentrate sample was centrifuged at 4000 rpm for 10 min and the water at the top of the vials decanted, leaving about 1 mL of concentrated sample at the bottom. A drop of this “concentrated” water sample was placed on each microscope slide. A microscope slide was also prepared from a filter media sample taken from an original bag as supplied by the contractor. The sample was mixed with tap water (Fig. 3) and left aside for about 10 min for the coars- er sand to settle to the bottom of the container. A drop of the suspension above the settled sand was placed on a microscope slide. Once dry, the slides were sealed with a slide epoxy (Permount) for sample preservation.

Slide Analysis The slides were analyzed under a microscope using three different magnifications: 100x, 400x, and 1000x. It was found that the 100x and 1000x objectives were the most useful; the 100x objective provided data in

Fig. 3. Water washed out from the original filter media. PARTICILES IN SLOW SAND FILTRATION 159 the range of the larger diameter particles, while the 1000x objective provided data in the range of the smallest particles. The microscope was connected to a computer running the Bioquant Classic 95 (R & M Biometrics, Inc.) image analysis program. Cross-sectional areas of particles present in the water were determined according to the procedures described elsewhere (Gorczyca and Ganczarczyk 1996). Equivalent diameters were calculated as the diameters of a circle having the same cross-sectional area as the particle image:

Equivalent diameter = (4* Cross-sectional Area/π)0.5 (1)

Results and Discussion

Comparison of Particle Counter Data with Microscopic Analysis The total numbers of particles measured by the particle counter as well as their average sizes are shown in Table 5. Although the particle counts indicated a gradual decrease in the total number of particles, the water turbidity increased during the treatment process (Table 5). Particle size distributions as measured by the particle counter are shown in Fig. 4. The number of particles in the size range 2 to 3 µm increased from 45% of the total particle count in the raw water to 70% in the slow sand filter effluent. The concentration of these small particles remained high throughout the treatment process. This result suggests that the high water turbidity is caused by an increase in the number of 2- to 3-µm sized particles. The particle counter, however, did not clearly indicate whether these small particles originated from the raw water or from the filter media. Since the source of turbidity could not be identified based on the particle counter data, microscopy analysis of the water was conducted.

Table 5. Particle counter data (December 17–19, 2001)

Total number of particles in 100 mL Average size of Water turbidity Water sample of samplea particles (µm) (NTU)

Raw water 2300 3.7 0.51 Slow sand filter 1200 3.2 0.74 effluent (SSF1) GAC effluent 680 3.0 0.68 (both units running) Distribution water 380 3.0 0.55

a Averages are based on on-line measurements performed every 2 min for 12 h. 160 GORCZYCA AND LONDON

The cross-sectional area of about 100 particles from each sample was measured using the Bioquant 95 image analysis system, and equivalent diameters were calculated using equation 1. Microscope analysis indicated presence of significant numbers of particles smaller than 2 µm and larger than 15 µm, which is outside of the 2- to 15-µm size range reported by the particle counter. This is the main cause of the discrepancy in the particle size obtained from the particle counter and a microscope (Fig. 4, 5, 6, and Table 6). Only when water samples were analyzed at the highest magnification (1000x) and the particles outside of the 2- to 15-µm size range were excluded from microscopy data, the sizes reported by the microscope and particle counter comparable (row 1 and 2 in Table 6). The discrepancy in the microscope data collected at different magnifications is due to the fact that the higher power magnification provides a better rep- resentation of the small particle distribution, as it is able to identify the majority of small particles, whereas the lower magnification can only identify the largest of particles. Microscopic size distributions showed a significant increase in the number of 0- to 2-µm particles during the treatment process (Fig. 5); but the particle counter used in this study could not recognize these particles. Moreover, microscope analysis allowed for observation of the morphology of these small particles. Fig. 7a shows a photograph of small particles

Fig. 4. Size distribution of particles in the size range 0 to 15 µm as measured by the particle counter. PARTICILES IN SLOW SAND FILTRATION 161

Fig. 5. Size distribution of particles in the size range 0 to 15 µm as determined by microscopy (magnification 1000x).

(0- to 2-µm size) identified on the microscope slide prepared from the slow sand filter effluent water (SSF1). Fig. 7b shows particles identified in the water washed out from the filter media sample originally supplied by a contractor (preparation of this microscope slide has been described in Methods and Materials section). Visual inspection of morphology and sizes of particles shown in Fig. 7a and b shows that the particles causing high turbidity in the filter effluent are similar to the particles washed out from the original filter media. Full size range distributions of the microscopic data presented in Fig. 6 show a significant number of particles that are larger than 15 µm. These particles have not been reported in the particle counter data, thus, there is a large discrepancy in the size reported by the particle counter and the microscopic analysis (Table 6). The images of these large particles as seen on the microscope slides are shown in Fig. 8. The majority of the large particles in raw water supply have been identified as diatoms. Several filter- clogging genera were identified from the Round Lake raw water, notably Cymbella, Fragilaria, and Navicula (Table 7). These diatoms were not found in SSF effluent water (Fig. 7a). Therefore, it appears that the high water turbidity in the SSF effluent is caused by particles originating in the filter media rather than in the raw water supply. Such observations could not have been made with the particle counter data alone. 162 GORCZYCA AND LONDON

Fig. 6. Size distribution of particles in the size range 0 to 50 µm as determined by microscopy (magnification 100x and 1000x).

Table 6. Average size of particles—comparison of particle counter and microscope data

Water washed Method of Raw SSF effluent GAC Distribution out from measurement water (SSF1) outflow water filter media

Particle counter 3.70 3.01 3.16 2.98 N.A. Microscope 3.87 3.87 3.85 3.10 N.A. (2–15 µm size range, 1000x ) Microscope 10.54 8.97 7.80 7.71 N.A. (2–15 µm size range 100x) Microscope 1.70 1.36 1.53 1.46 1.19 (0–50 µm size range 1000x) Microscope 21.33 14.82 11.38 10.92 N.A. (0–50 µm size range 100x) PARTICILES IN SLOW SAND FILTRATION 163

The diatoms identified in the sample from Round Lake have a glass- like cell wall made of silicon dioxide. Because of the transparent cell walls, the turbidity meters as well as particle counter did not detect the high diatom concentration in the water. The Met One PCX is one of a number of particle counters that is based on a light-blocking principle. A light

Fig. 7. Images of particles causing high turbidity in slow sand filter effluent — a) microscopic image of particles identified in the slow sand filter effluent, and b) microscopic image of particles found in the water washed out from the sample of slow sand filter media supplied by a contractor. 164 GORCZYCA AND LONDON source is aimed through a narrow stream of the sample water, and is focused on a light detector (Fig. 9). As a particle passes through the light source (in the case of the PCX, a laser is used), the detector recognizes a break in the light, and with the known flow of water, this pulse can be

Fig. 8. Microscope images of filter clogging diatoms identified in raw water (magnifications 100x and 1000x). PARTICILES IN SLOW SAND FILTRATION 165

Table 7. Algal species and densities identified from Round Lake water sample (enumerated by J.K. Engineering, Calgary, Canada)

Algae Density (units/L)

CHLOROPHYTA Cosmarium sp. 1443 Sphaerocystis schroeteri Chodat 1443 Scenedesmus acutus Meyen 1443 Tetraedron minimum (A.Braun) Hansgrid 1443 Ulothrix sp. 1443 DIATOMS Cymbella affinis Kuetzing 5773 Navicula cryptotenella Lange-Bertalot. 1443 Navicula sp. 1443 Fragilaria tenera (W.Smith) Lange Bertalot 14433 Gomphonema parvulum Kuetzing 12,989 Cocconeis placentula Ehrenberg 7216 CYANOBACTERIA Snowella lacustris (Chodat) Romarek et Hindak 1443 Limnothrix sp. 5773 CHRYSOPHYCEAE Ochromonas sp. 2886 OTHERS Algal cysts 105,361

converted into a particle size. Fig. 10 shows a single diatom, a drop of water and a coin (quarter). Because a large proportion of the light will be transmitted through the water drop, the particle counter will recognize the water drop to be much smaller than the quarter. Using the same principle, the size of a diatom shown in Fig. 10 would be underestimated, if recorded at all. This particular source of error in particle counter analysis has not been identified elsewhere (Broadwell 2001). The raw water supply for North Caribou water treatment plant contained about 100,000 algal units/L (Table 7), which were recognized poorly by the particle counter. Based on the estimated biovolume of the predominant algal taxa, it was calculated that the raw water contains about 170 mg/m3 of algal biomass (H. Kling, pers. comm.). Assuming average chlorophyll a concentration of 1.5% (American Public Health Association 1998) the algal concentration is significantly lower than the water quality guidelines for slow sand filtration (Table 1) and indicates very oligotrophic conditions. The majority of algae identified in the samples from Round Lake were diatoms. These taxa often have a low chlorophyll a content per unit 166 GORCZYCA AND LONDON

Fig. 9. Met One Particle Counter operating principles. cell volume, which can vary depending on species, and environmental conditions. Because chlorophyll a concentration is not always well correlated with the biomass of algae, particularly diatoms (Tolstoy 1979; Rijkeboer et al. 1992), the application of chlorophyll a concentration guideline for selection of slow sand filtration (Table 1), particularly these devised for algal surface mats, is questionable for the biomass identified in Round Lake water (H. Kling, pers. comm.). It is also important to note that most of the species identified in Round Lake are not planktonic (Table 7); also, the images of the diatoms show no presence of internal structures, which indicates that these were are not alive. As discussed earlier, the water flow in the intake pipe is slow and allows for material to settle in the pipe. It is therefore possible that the majority of diatoms clogging the filters in the plant did not originate from the open water, but from resus- pended sediment or sloughed material growing close to, or on the intake structure (S. Watson, pers. comm.).

Conclusions

1. Microscopic analysis of water and filter media in the North Caribou water treatment plant allowed for identification of the sources of high water turbidity. 2. Turbidimeters and the particle counter used in this study were unable to identify filter-clogging diatoms, due to the transparent nature of the cell walls of these microorganisms. Only microscope analysis of the water allowed for identification of these organisms. 3. Chlorophyll a water quality guidelines for selection of slow sand filtration may not be applicable to many waters. PARTICILES IN SLOW SAND FILTRATION 167

Fig. 10. Size of transparent particles as determined by the particle counter. Particle Counter Reports for a) size of drop of water < size of quarter , because drop of water blocks less light than a quarter. Particle Counter Reports for b) size of a diatom particle < the actual size of a diatom.

Acknowledgements

Tad Smoter (Windigo Tribal Council) and Jan Korzeniowski (J.K. Engineering) generously provided technical information regarding the North Caribou water treatment plant. Ron Oger (Danron Mechanical) conducted particle counts.

References

American Public Health Association. 1998. Standard methods for the examination of water and wastewater. A.W.W.A., Water Environment Federation, Washington, American Public Health Association. American Water Works Research Foundation. 1991. Manual of design for slow sand filtration. Denver, Colo., AWWA Research Foundation. Broadwell M. 2001. Light blocking sensors, p. 8–9. In A practical guide to particle counting. Lewis Publishers, Washington, D.C. 168 GORCZYCA AND LONDON

Gorczyca B, Ganczarczyk J. 1996. Image analysis of alum coagulated mineral suspensions. Environ. Technol. 17:1361–1369. Health Canada. 1996. Guidelines for Canadian drinking water quality, Government of Canada. Kling H. 2002. Algal Taxonomy and Ecology Inc., Winnipeg, Canada (personal communication). Korzeniowski J. 2001. J.K. Engineering, Calgary, Canada (personal communication) Rijkeboer M, Dekloet WA, Gons HJ. 1992. Interspecific variation in pigmentation– implications for production estimates for shallow eutrophic lakes using an incubator. Hydrobiol. 238:197–202. The Burnside Group of Companies. 1999. Design brief water supply North Caribou First Nation. Winnipeg, Burnside Engineering Western. A Division of R.J. Burnside and Associates Limited. Tolstoy A. 1979. Chlorophyll a in relation to phytoplankton volume in some Swedish lakes. Arch. Hydrobiol. 85:133–151. Watson S. 2002. Environment Canada, National Water Research Institute, Burlington, Ontario (personal communication). Water Qual. Res. J. Canada, 2003 Volume 38, No. 1, 169–182 Copyright © 2003, CAWQ

Sulfate Removal from Water

ASHREF DARBI,1 THIRUVENKATACHARI VIRARAGHAVAN,1* YEE-CHUNG JIN,1 LARRY BRAUL2 AND DARRELL CORKAL3

1Faculty of Engineering, University of Regina, Regina, Saskatchewan S4S 0A2 2Agriculture and Agri-Food Canada, Regina, Saskatchewan 3Agriculture and Agri-Food Canada, Saskatoon, Saskatchewan

Sulfate occurs naturally in groundwater. Concerns regarding the health effects from sulfate in drinking water have been raised because of reports that diarrhea may be associated with water that contains high levels of sulfate. In the livestock production industry, there is a concern that high levels of sulfate in water can adversely affect productivity. Different methods can be used to remove sulfate from water. Proven technologies are ion-exchange, nanofiltration, reverse osmosis, and electrodialysis. A few earlier studies have shown that the use of bentonite/kaolinite for sulfate removal has produced encouraging results. Experimental work was undertaken to examine in detail the feasibility of such processes. Laboratory studies using bentonite showed poor or no removal in the case of high sulfate water. Ion exchange and nanofiltration were found to be very effective in removing sulfate. Ion exchange is likely to be more reliable than nanofiltration because of the sensitivity of the nanofiltration process to total dissolved solids and biofouling.

Key words: sulfate removal, drinking water, bentonite, ion exchange, nanofiltration

Introduction

Sulfate occurs naturally in groundwater. Sulfate ions present in water in high concentrations may cause temporary and acute effects on humans and animals, including diarrhea. The United States Environmental Protection Agency (U.S. EPA) has proposed a maximum allowable concentration of 500 mg/L for sulfate in drinking water in order to avoid any health concern regarding human consumption. A secondary maximum allowable concentration for sulfate has been set at 250 mg/L (U.S. EPA 1994). It is understood that approximately 30% of groundwater in Saskatchewan exceeds a sulfate concentration of 1000 mg/L (Shaheen and Sketchell 1998), the maximum objective level for livestock watering set by the Saskatchewan Environment and Resource Management. In some cases, sulfate concentrations are reported to be as high as 3000 mg/L. It is believed that the removal of sulfates from drinking water will lead to a healthier livestock and a more productive herd. On the other hand, sulfate

* Corresponding author; [email protected] 170 DARBI ET AL. is a necessary constituent in the bodies of humans and other animals. In humans, serum sulfate levels range from 24 to 36.5 mg/L. Sulfate is involved in many biochemical activities including the production of chon- droitin sulfate and sulfation of exogenous chemicals. Three cases from Saskatchewan were reported with infants experiencing gastroenteritis with diarrhea and dehydration upon ingesting water that had high levels of sulfate (650–1150 mg/L) (Chien et al. 1968). Diarrhea subsided in all infants when different water sources with lower sulfate concentrations were used (Backer 2000). Members of an expert workshop on sulfate concluded that there was not enough scientific evidence to support a regulation creating a Maximum Contaminant Level (MCL) for sulfate in drinking water (Backer et al. 2001). In the livestock production industry, there is a concern that high levels of sulfate in drinking water can adversely affect productivity. A study conducted by Veenhuizen et al. (1992) on nursery pigs given drinking water containing sodium or magnesium sulfate at 600, 1200, and 1800 mg/L of sulfate for 28 days, showed that pigs drinking high sulfate water had a higher frequency of non-pathogenic diarrhea than the controls (i.e., pigs drinking water with 54 mg/L naturally occurring sulfate) (Backer 2000). It also showed that sulfate levels in excess of 500 mg/L can cause laxative effects on young animals, with cattle becoming more resistant within several weeks (J. Cory, Prairie Farm Rehabilitation Administration, Regina, Sask., pers. comm.). Levels of sulfate greater than 300 to 600 mg/L can cause chronic diarrhea, electrolyte imbalance, and possible death. Different treatment technologies have been investigated for sulfate removal. Proven technologies for the removal of sulfate from drinking water include ion exchange, nanofiltration, reverse osmosis and electrodialysis (Table 1). In addition, several studies have been reported on the use of bentonite/kaolinite adsorption for sulfate removal. While such

Table 1. Different sulfate removal technologies (Marhaba and Washington 1997)

Treatment technology Description

Reverse osmosis/nanofiltration Water is forced under pressure through a porous membrane designed to remove ions from the water Ion exchange Inorganics are removed by passing water over cation and anion exchangers, replacing cations and anions with H+ or Cl- or OH- Electrodialysis Direct current is applied across a body of water separated into vertical layers alterna- tively permeable to cations and anions SULFATE REMOVAL FROM WATER 171

studies have reported mixed results, an optimized system based on bentonite adsorption would have a significant economic advantage over the other technologies. A study conducted by Rao and Sridharan (1984) on the adsorption of sulfate by kaolinite found that sulfate was adsorbed at positive and neu- - tral sites with the displacement of OH2 and OH groups. Adsorption of sulfate occurred significantly at positive sites at low concentrations whereas on increasing the solution concentration, the proportion of sulfate adsorption at the neutral site increased. The level of positive charge on the clay surface apparently governs the form of surface bonding. At low anion saturation, sulfate was adsorbed on kaolinite as a divalent ion. At a higher solution concentration, the surface favours the adsorption of the monovalent ions and thus sulfates formed are in both monodentate and bidentate complexes. The objectives of this study were to examine in detail the feasibility of using bentonite to remove sulfate from groundwater; and to compare sulfate removal using bentonite with ion exchange and nanofiltration processes.

Materials and Methods

The removal of sulfate from water was investigated using various treatment methods. Removal of sulfate through adsorption was examined using different concentrations of bentonite. Three types of test water (tap water spiked with sulfate, and groundwater from Leroy and Swift Current in Saskatchewan) were used in the experiments. Sulfate-spiked tap water was prepared by weighing a known amount of calcium sulfate or magnesium sulfate and adding this amount to the tap water to obtain the desired sulfate concentration. Final sulfate concentrations were measured by Dionex before starting the experiments. Groundwater samples from Leroy and Swift Current were collected by the Prairie Farm Rehabilitation Administration (PFRA) in February 2001, transported to the University of Regina in an open van and stored in a refrigerator at 4°C. All adsorption experiments with bentonite were repeated three times using tap water spiked with sulfate. In addition, ion exchange and nanofiltration techniques were investigated to examine the removal capa- bilities through these systems. Ion exchange and nanofiltration experiments were repeated three times for each water type used. Duplicate samples were collected for each analysis. Average values were used in data analysis. The characteristics of groundwater samples from Leroy and Swift Current provided by PFRA are shown in Table 2.

Adsorption of Sulfate to Bentonite Sulfate adsorption studies were conducted under different bentonite concentrations. Bentonite was obtained from Canadian Clay Products, Wilcox, Saskatchewan. Different concentrations of sulfate 172 DARBI ET AL.

Table 2. Characteristics of water samples

Hardness - - -- Cl NO3 -N SO4 (mg/L Conductivity TDS Sample pH (mg/L) (mg/L) (mg/L) as CaCO3)µS/cm (mg/L)

Tap water 7.5 18 0 185–200 232 534 230 Leroy 7.2 295 15 2280 2735 — 3750 (groundwater) Swift Current 7.7 112 18 3665 3700 6200 — (groundwater)

were used in the study. Calcium sulfate was dissolved using tap water. Batch isotherm studies with bentonite concentrations of 400, 500, 600, 700, 800, 900, and 1000 mg/L were performed at 23 ± 1°C. Bentonite was weighed and placed in 250-mL Erlenmeyer flasks. The flasks were then filled to 100 mL with calcium sulfate solution and covered with parafilm wax. The mixtures were placed in a gyratory shaker, at 200 rpm, for 15, 30, 45, 60, 75, 90, 115, 120, 180, and 240 minutes. pH values were measured prior to, and at the end of the contact period, using a Hanna model 1024 pH meter. At the end of each contact time, the bottle reactor was removed and 40 mL of the mixture was decanted into a centrifuge tube for analysis. The mixture was centrifuged for 15 minutes at 6000 rpm to separate the bentonite from the solution. The supernatant was diluted 1:10 with distilled water and analyzed for sulfate using a Dionex Ion Chromatograph.

Sulfate Removal by Ion Exchange Column experiments were conducted to examine sulfate removal by an anion exchange resin. A high capacity, type 2 ionic resin (ASB 2) was used for all column experiments (Sybron Chemicals Inc., Birmingham, New Jersey). Bead size distribution of the ionic resin ranged between 0.3 to 1.2 mm with a particle density of approximately 1.11 g/mL. Total 3 exchange capacity, as CaCO3, was 1.4 eq/L or 30.6 kg/ft . Water content of the ASB 2 resin was between 38 and 45%. The column apparatus consisted of an acrylic column, 90 cm long and 1 cm internal diameter (ID) with two butyl rubber stoppers used as end caps. The resin was packed to a height of 80 cm. The end caps were machined with a small hole, 0.64 cm in diameter, to allow for influent and effluent discharge. A glass filter was placed above the bottom end cap to prevent resins from leaving the column. Silicon tubing (Nalgene) was used to connect the input reservoir to the column. The effective volume packing was measured gravimetrically for each individual packing. Tap SULFATE REMOVAL FROM WATER 173 water spiked with sulfate and Leroy and Swift Current groundwater were used in ion exchange column experiments. The column was flooded with 20 L of tap water containing 1000 mg/L of sulfate, using a peristaltic pump (Cole Parmer). Magnesium sulfate was used instead of calcium sulfate because a high concentration of calcium sulfate will cause a high-turbidity solution. High-turbidity water will form scale on the membrane and affect the capacity of the resin. Saturation of the column was achieved by a down-flow gradient of 75 mL/min, which maintained a constant head of 1 cm above the anion resin. The effluent was collected in scintillation vials after the first 60 min and every 30 min thereafter for 4 h, when breakthrough was observed.

Nanofiltration Nanofiltration is a pressure-driven membrane process with performance characteristics between reverse osmosis and ultrafiltration. The theoretical pore size of the membrane is 1 nm. A nanofiltration unit was obtained from Water Group with Filmtec 2.5” nanofiltration elements, model NF 90-1812-HF. The unit was designed for home use and small production. Spiked tap water using magnesium sulfate with different initial concentrations was used at two different pressures of 40 and 80 psi. Initial concentrations varying from 916 to 5363 mg/L were used in the runs. Sampling was conducted during the initial, intermediate, and final stages of the experimental procedure. Groundwater samples from Leroy and Swift Current were also examined using nanofiltration.

Results and Discussion

Sulfate Adsorption to Bentonite Figure 1 illustrates the concentration of sulfate using different bentonite concentrations of 400, 500, 600, 700, 800, and 900 mg/L. There was no significant reduction in sulfate concentration at these bentonite concentrations. Bentonite did not appear to possess any adsorption capacity for sulfate. The results showed that the change in sulfate concentration was marginal during the four-hour study period. Table 3 also shows that during a two-hour study, bentonite leached sulfate into water. Commercial-and laboratory-grade bentonites were used in these experiments. The commercial-grade bentonite leached more sulfate than the laboratory-grade bentonite. Since the commercial-grade bentonite used in the studies would have leached some sulfate into the water, it is likely that any marginal adsorption by bentonite may be offset by this leaching.

Ion Exchange Figure 2 illustrates the concentration of sulfate and regeneration by ion exchange using spiked tap water with sulfate concentration of 174 DARBI ET AL.

Fig. 1. Average sulfate concentrations in the effluent using different concentrations of bentonite (400, 500, 600, 700, 800, and 900 mg/L) (N = 3).

1000 mg/L. Effluent samples were taken every 30 min until breakthrough occurred. Breakthrough is defined as the concentration of sulfate passing through the column when the absorbent has been saturated with time. The breakthrough of sulfate occurred after 150 min. Sodium chloride at a concentration of 5% was used with tap water to regenerate the column.

Table 3. Sulfate leaching from lab and commercial bentonite

Initial sulfate, Commercial bentonite Laboratory-grade mg/L sulfate, mg/L bentonite sulfate, mg/L

25 166 122 25 139 121 SULFATE REMOVAL FROM WATER 175

Regeneration was complete after 50 min. Within the 50-min regeneration period, the sulfate concentration decreased from 8000 mg/L to <100 mg/L sulfate. Figure 3 illustrates the concentration of sulfate and regeneration using ion exchange with an initial sulfate concentration of 2000 mg/L. Figure 4 compares the sulfate concentration in Leroy and Swift Current samples using the ion exchange system. Initial sulfate concentrations at Leroy and Swift Current were 2280 and 3665 mg/L, respectively. The breakthrough time of sulfate for both samples occurred at approximately 120 min and the removal rates were greater than 90% for both samples. ASB 2 exhibited the capacity to remove sulfate from both waters. The column removed 21 and 33 g of sulfate from the Leroy and Swift Current water samples, respectively. Regeneration of the column required 180 g of NaCl. The concentration of the regenerant was 40 g/L. Figure 5 shows the regeneration time for Leroy and Swift Current samples. The regeneration cycle required 60 min with an inflow of 75 mL/min.

Fig. 2. Average sulfate concentration in the effluent using 1000 mg/L sulfate in tap water (ion exchange and regeneration) (N = 3). 176 DARBI ET AL.

Fig. 3. Average sulfate concentration in the effluent using 2000 mg/L sulfate in tap water (ion exchange and regeneration) (N = 3).

Nanofiltration Two different pressures of 40 and 80 psi were applied. Low, medium and high sulfate concentrations were used in the runs (see Table 4). It can be seen from the table that the sulfate removal efficiency increased with an increase in applied pressure. Salt rejection also increased with applied pressure. In both situations, the reduction of sulfate was effective using the nanofiltration system. Under high applied pressure of 80 psi, the removal efficiency was greater when compared to 40 psi. Table 5 shows high sulfate removal for both Leroy and Swift Current samples. It was found that the amount of salt rejection was higher at 80 psi than at 40 psi. The results shown in Table 5 confirm the excellent rejection of sulfate with an average rejection of 93%. Table 6 compares the percentage removal of sulfate by the three technologies. Nanofiltration, while exhibiting slightly higher removal efficiencies, possesses several disadvantages, including lowered removal of sulfates under high TDS in water and membrane fouling from heavy metals such SULFATE REMOVAL FROM WATER 177 as iron, and bacteria (American Water Works Association 1990). The most restrictive factor in nanofiltration is scaling by CaSO4. Acidification is required to overcome mineral scaling on membranes (American Water Works Association 1990). Further studies are required to examine the effect of raw water quality parameters on sulfate removal in the context of water supplies studied.

Conclusions and Recommendations

Ion exchange and nanofiltration are the two best available technologies for sulfate removal, and are also proven technologies used for desali- nation of seawater and brackish water (American Water Works Association 1990). This study showed that ion exchange is the recommended option in removing sulfate from water. Although high amounts of salt are required to regenerate the column, ion exchange appears to be the most beneficial compared to nanofiltration. More detailed studies are

Fig. 4. Average sulfate concentration in the effluent using Leroy and Swift Current samples (ion exchange) (N = 3). 178 DARBI ET AL.

Fig. 5. Regeneration of high capacity, type 2 ionic resin (ASB 2) using sodium chloride. needed to examine sulfate adsorption by bentonite or kaolinite. In this study sulfate concentration appeared to increase in all the experiments. These results are in contrast to studies by other investigators, who claimed success with bentonite especially when sulfate concentrations were low (less than 50 mg/L). SULFATE REMOVAL FROM WATER 179 /L /L /L /L 4 4 4 4 1245 3716 4597 5205 a a a a /L mg SO /L mg SO /L mg SO /L mg SO 4 4 4 4 /L mg SO /L mg SO /L mg SO 4 4 4 /L mg SO 4 /L 916 mg SO /L 2810 mg SO /L 3600 mg SO /L 4095 mg SO 4 4 4 4 2172 Conc. Average 3187 Conc. Average 67 ± 11 4016 Conc. Average 97 ± 7 5883 164 ± Conc. Average 21 144 ± 13 concentrations under 40 and 80 psi 4 a a a a /L mg SO /L mg SO /L mg SO /L mg SO 4 4 4 4 Pressure 40 psiPressure 80 psi Pressure /L mg SO /L mg SO /L mg SO /L mg SO 4 4 4 4 Nanofiltration runs with varying SO Nanofiltration runs Standard deviation. Standard a verage Conc. 93 ± 4 verage Conc. 199 ± 22 verage Conc. 400 ± 13 verage Conc. 638 ± 23 able 4. A A A A T 1650 mg SO Initial Conc.2643 mg SO Conc. Prod. Initial Conc.3600 mg SO Brine Conc. Conc. Prod. Initial Conc. Initial Conc.5363 mg SO Brine Conc. Salt Rejection (%) Conc. Prod. Initial Conc. Conc. Prod. 88 Brine Conc. Brine Conc. Initial Conc. Conc. Prod. Brine Conc. Conc. Prod. Salt Rejection (%) Brine Conc. 96 Initial Conc. Conc. Prod. Brine Conc. Initial Conc. Conc. Prod. Brine Conc. Salt Rejection (%) 94Salt Rejection (%) 92Salt Rejection (%) 89 Salt Rejection (%) Salt Rejection (%) 93 Salt Rejection (%) 97 95 180 DARBI ET AL. /L /L 4 4 2687 5132 a a /L mg SO /L mg SO 4 4 Leroy Swift Current /L/L mg SO mg SO 4 4 3835 Conc. Average 4251 199 ± 31 Conc. Average 127 ± 25 a a /L mg SO /L mg SO 4 4 mg SO mg SO Prod. Conc. Prod. Brine Conc. Conc. Prod. Brine Conc. Conc. Prod. Brine Conc. Conc. Prod. Brine Conc. Pressure 40 psiPressure 40 psiPressure 80 psi Pressure 80 psi Pressure Nanofiltration runs using Leroy and Swift Current samples under 40 and 80 psi and Swift Current using Leroy Nanofiltration runs Standard deviation, initial concentration is 2280 and 3665 mg/LStandard samples, respectively. and Swift Current sulfate for Leroy a verage Conc. 229 ± 1 verage Conc. 296 ± 42 able 5. A A T Salt Rejection (%) 92 Salt Rejection (%) 97 Salt Rejection (%) 90 Salt Rejection (%) 91 SULFATE REMOVAL FROM WATER 181 91% 97% Nil a b 95% 94% 97% ≅ . . 90% –– –– 4 4 ≅ Ion-exchange Nanofiltration at 80 psi Bentonite adsorption 90% ≅ Percentage removal comparison of the three technologies comparison of the three removal Percentage At concentration of 3600 mg SO At concentration of 2810 mg SO b a ater Spiked tap Swift Spiked tap Swift Spiked tap able 6. T usedSulfate temoval water Leroy Current water Leroy Current water W 182 DARBI ET AL.

References

American Water Works Association. 1990. Water quality and treatment: a handbook of community water supply. 4th Edition. Backer LC. 2000. Assessing the acute gastrointestinal effects of ingesting naturally occurring high levels of sulfate in drinking water. Crit. Rev. Clinic. Lab. Sci. 37:389–400. Backer LC, Esteban E, Rubin CH, Kieszak S, Mcgeehin MA. 2001. Assessing acute diarrhea from sulphate in drinking water. J. Am. Water Works Assoc. 93:76–84. Chien L, Robertson H, Gerrard JW. 1968. Infantile gastroenteritis due to water with high sulphate content. Can. Med. Assoc. J. 99:102–104. Cory J. Water quality and cattle performance. Range Management Division, Prairie Farm Rehabilitation Administration, Regina, Saskatchewan. Pers. comm. Marhaba TF, Washington MB. 1997. Sulfate removal from drinking water. In Proceedings CSCE/ASCE Environmental Engineering conference. Edmonton, Alberta, Canada. Rao SM, Sridharan A. 1984. Mechanism of sulfate adsorption by kaolinite. Clays Clay Miner. 32:414–418. Shaheen N, Sketchell J. 1998. Groundwater chemistry program pilot project. Sask Water, Moose Jaw, Saskatchewan. U.S. EPA. 1994. Sulfate. Proposed Rule, Federal Register, 59:243:65578 (Dec. 20, 1994). Veenhuizen MF, Shurson GC, Kohler EM. 1992. Effect of concentration and source on nursery pig performance and health. J. Am. Vet. Med. Ass. 201:1203–1208. Water Qual. Res. J. Canada, 2003 Volume 38, No. 1, 183–192 Copyright © 2003, CAWQ

A Kinetic Model for Autotrophic Denitrification using Sulphur:Limestone Reactors

ASHREF DARBI AND THIRUVENKATACHARI VIRARAGHAVAN*

Faculty of Engineering, University of Regina, Regina, Saskatchewan, Canada

The kinetics of autotrophic denitrification of groundwater by Thiobacillus denitrificans in a sulfur:limestone upflow reactor was examined in order to predict effluent concentrations. Experiments were performed using water containing 60 and 90 mg NO3—N/L and sulfur and limestone with average particle size of 3.5 mm. Results clearly showed that nitrate was completely removed from 60 and 90 mg NO3—N/L influent concentrations. The results showed that the autotrophic denitrification rates in sulfur:limestone reactors can be described by half-order kinetics. The half-order reaction rate constants for the entire media were estimated at 1.34 and 1.54 mg1/2/L1/2 h for influent concentrations of 60 and 90 mg NO3—N/L, respectively.

Key words: autotrophic denitrification, nitrate removal, Thiobacillus denitrificans, sulfur, limestone, drinking water, kinetics

Introduction

Numerous water agencies face problems of increasing concentrations of nitrate in groundwater. The main reason for increasing nitrate concentrations in groundwater is the extensive application of artificial fertilizers and animal manure in agriculture. The contamination of groundwater by excessive concentrations of nitrate is a significant public health problem. Infant methemoglobinemia is known to occur when nitrate-nitrogen in drinking water exceeds 10 mg/L (Packham 1992). The Canadian drinking water standard for nitrate is 10 mg/L as nitrate-nitrogen (Guidelines for Canadian Drinking Water Quality 1996). A wide range of physico-chemical processes such as ion exchange, reverse osmosis and electrodialysis, biological denitrification and chemical denitrification, are currently in use or under development for the removal of nitrates from drinking water (Kapoor and Viraraghavan 1997). Methanol is often added for nitrate removal by heterotrophic denitrification since the availability of organic carbon is limited in drinking water (Hoek et al. 1987). As an alternative, autotrophic denitrification using Thiobacillus denitrificans can reduce nitrate to nitrogen gas while oxidizing elemental sulfur to sulfate. Limestone is used to maintain the pH and provide an inorganic carbon

* Corresponding author; [email protected] 184 DARBI AND VIRARAGHAVAN source for the bacteria. The reaction proceeds as follows (Batchelor and Lawrence 1978; Schippers and Kruithof 1987):

- + 55S + 50NO3 + 38H2O + 20CO2 + 4NH 4 → 2- + (1) 4C5H7O2N + 25N2 + 55SO 4 + 64H

Further practical application of autotrophic denitrification requires a kinetic model. Data on concentration profiles available from experiments made by Sutton (1973) at 20°C showed that the data fitted the half-order reaction and gave volume removal rates in the range of 0.08 to 0.14 mg1/2/L1/2 min and surface removal rates between 1.3 and 2.2 x 10-3 mg1/2/L1/2 min for 13-mm ceramic intalox saddles. The scatter of the data at other temperatures made conclusions on the order of reaction difficult (Harremoës 1976). Harremoës and Riemer (1975) examined a large number of profiles with different hydraulic loadings on a pilot-scale downflow filter with a 3- to 4-mm gravel medium. When the removal rate per unit volume was plotted against the nitrate concentration, they found that a half-order reaction fitted the data well. Batchelor and Lawrence (1987) developed a kinetic model to describe continuous flow reactors with elemental sulfur biomass slurries. LeCloirec (1985) developed a mathematical model for denitrification by Thiobacillus denitrificans in a sulfur:limestone (S:L) reactor. However, none of these models are easily applicable to predict nitrate profiles and effluent nitrate concentrations in the upflow sulfur:limestone reactor. Koenig and Liu (2001) conducted experiments using autotrophic denitrification of synthetic wastewater by Thiobacillus denitrificans in upflow sulfur packed-bed reactors. The experimental results showed that autotrophic denitrification by the biofilm in the reactors can be described by a half-order kinetic model. A batch study made by Koenig and Liu (2002) showed that in the case of autotrophic denitrification with sufficient alkalinity (initial alkalinity of 680 mg/L as CaCO3), nitrate concentration decreased linearly with time and autotrophic denitrification followed first-order kinetics. The objectives of this study were to examine the use of a suitable kinetic model for upflow sulfur:limestone reactors, and to determine the kinetic constants for reactors with different nitrate-nitrogen concentrations and compare the results with other studies.

Materials and Methods

Thiobacillus denitrificans Culture Thiobacillus denitrificans (ATCC 23642) was grown in a medium as described by Lampe and Zhang (1996). The composition of the medium was 6 g/L Na2S2O3.5H2O, 3 g/L KNO3, 1.5 g/L NaHCO3, 1.5 g/L Na2HPO4, 0.3 g/L KH2PO4, 0.4 g/L MgSO4.7H2O and 1 mL/L trace nutrient solution. The composition of the trace nutrient solution was SULPHUR: LIMESTONE REACTORS 185

56.25 mg/L K2HPO4, 5.74 mg/L NH4Cl, 1 mg/L MgCl2.6H2O, 1 mg/L MnSO4.H2O, 1 mg/L CaCl2 and 1 mg/L FeCl2.6H2O. The stock culture was inoculated into 1 L of medium, flushed with nitrogen and incubated at room temperature for 7 to 14 d.

Column Studies A(2:1) sulfur:limestone (mass/mass) reactor (Fig. 1) was used in the column study. This same reactor had a nitrate removal rate of approximately 100% at the end of one year of operation. The column was under continuous operation with a flow rate of 2 mL/min. The results showed that the nitrate removal efficiencies were almost exactly the same over a period of six months, indicating that the reactor had reached steady-state conditions. Earlier detailed column studies have shown that 2:1 (mass/mass) is the optimum S:L ratio; under a hydraulic retention time (HRT) of 13 h, nitrate concentration of 60 mg NO3—N/L was reduced to less than 5 mg NO3—N/L (Darbi et al., Unpublished). Tap water was used to prepare two different initial nitrate concentrations (60 and 90 mg NO3—N/L); the sulfur and limestone particles both had a mean particle size of 3.5 mm. Characteristics of the tap water are shown in Table 1. The reactor was fed continuously in the upflow mode by peristaltic pumps. All experiments were conducted at room temperature of 21 ± 1°C. Samples were collected for analysis from the influent, the three sampling ports and the effluent.

Fig. 1. Upflow fixed-bed column reactor. 186 DARBI AND VIRARAGHAVAN

Table 1. Tap water characteristics

Parameters Tap water pH 7.5 Conductivity, µS/cm 534 NO3—N, mg/L 0 NO2—N, mg/L 0 Cl-, mg/L 18 — SO4 , mg/L 185–200 Hardness, mg/L as CaCO3 232 Alkalinity, mg/L as CaCO3 124 TDS, mg/L 230

Specific surface areas of sulfur and particles were determined using a Flowsorb II 2300 manufactured by Micromeritics Instrument Corporation, Georgia, U.S.A. Single point surface area measurements were carried out using a gas mixture of 29.0 M % nitrogen and 71.0 M % helium. Liquid nitrogen was used to set the temperature for adsorption of nitrogen gas by the samples. Specific gravity and porosity of the media were determined to be 2.03 and 0.42, respectively, using methods described in American Society for Testing and Materials (ASTM 1991).

Kinetics As the Monod saturation constant (Ks) for autotrophic denitrification was reported to be low (Claus and Kutzner 1985; Batchelor and Lawrence 1987), the intrinsic reaction inside the biofilm can be taken as zero-order. Since the penetration of substrate in the pores of biofilm is less than fully effective, a zero-order reaction in the biofilm becomes a half-order reaction at the surface of biofilm (Koenig and Liu 2001).

Assuming the filter as a plug flow reactor: ∂C A = – rv (2) ∂YQ

where C is concentration, Y is distance from the entrance, Q is flow rate, A is cross-sectional area, and rv is removal rate per unit volume of the filter.

A partially efficient biofilm will result in a half-order reaction:

1/2 rv = k1/2 v C (3) SULPHUR: LIMESTONE REACTORS 187

where k1/2 v is the half-order reaction constant per unit volume of the reactor.

∂C – = – kC1/2 (4) ∂t

(5)

1/2 1/2 1 Ce =C 0 - k t (6) 2 1/2v H

where Ce is effluent concentration, C0 is influent concentration, and tH is the empty bed residence time = AH/Q (H = height of reactor). 1/2 The profile should be a straight line in a plot of C versus tH. The half-order reaction rate per unit volume can be calculated from the slope of the line. The specific surface area of the filter media is ω which was estimated by multiplying the specific surface area of the particles by the factor (1-ε) where ε is the porosity of the reactor.

k1/2 v = ω k1/2 a (7)

Where k1/2 a is the half-order surface reaction rate of the biofilm and ω the specific surface area of the reactor media.

Results and Discussion

Column Studies The column contained sulfur and limestone media with a ratio of 2/1 mass/mass and all kinetic constants calculated were for all the media. The limestone provided buffering capacity and was the major inorganic carbon source for the S:L system (Flere and Zhang 1999). Figure 2 shows clearly that nitrate was completely removed under both initial concentrations of 60 and 90 mg NO3—N/L. The figure shows the reduction of nitrate concentration as a function of residence time. A plot of square root of nitrate concentration versus empty bed residence times shows a — straight line. Plotting √C in mg1/2/L and T in h gives the following 1/2 1/2 dimensions for the volume rate k1/2v = mg /L h. Figure 3 provides a plot between the square root of nitrate concentration and HRT using the data from Fig. 2. The relationship was found to be linear, with a correlation coefficient of nearly 1.0. Using equation 7 the half-order reaction rate constants for reactor media can be estimated. Table 2 shows the k1/2a to be 0.018 and 0.020 mg1/2/dm1/2 h for influent concentrations of 60 and 90 mg NO3—N/L, respectively. 188 DARBI AND VIRARAGHAVAN

Fig. 2. Nitrate concentration profile through the – column for 60 and 90 mg NO3 –N/L.

Fig. 3. Square root of nitrate concentration versus – HRT for 60 and 90 mg NO3 –N/L. SULPHUR: LIMESTONE REACTORS 189 fur. fur. hT°C 1/2 1/2a /dm k 1/2 mg 3 /dm ω 2 hdm a 1:2 /L 1/2v 1/2 value of 2.17 on the basis a 100% sulfur column. 1/2v 1.2–16 0.40 1.12–1.34 26.47 0.043–0.050 20–25 5.6–11.2 0.40 1.49–2.04 42.86 0.035–0.047 20–25 S S1 ype of Medium k T Comparison of half-order reaction rate constants for autotrophic denitrification rate constants for autotrophic reaction Comparison of half-order S:L to a S:L ratio of 2:1 by mass would correspond ratio by volume of about 2.5:1, i.e., 71% the media consists sul value of 1.45 for the mixed column would therefore correspond to a k correspond value of 1.45 for the mixed column would therefore A a 1/2v able 2. Ak T Refer.Koenig and Liu (1997) medium size, mm S Porosity 2.8–5.6 mg 0.40 2.36–3.52 81.82 0.029–0.043 20–25 This studyThis studyKoenig and Liu (2001) S S+L S 2.8–5.6 2.38–4.76 2.38–4.76 0.421 0.40 0.421 1.34–1.54 2.94–3.6 1.89–2.17 76.6 81.82 76.6 0.018–0.020 0.036–0.044 0.025–0.028 21±1 20–25 21±1 190 DARBI AND VIRARAGHAVAN

The plots shown in Fig. 3 clearly demonstrate the applicability of the half-order reaction rate model. Table 2 also shows the data from other sources in the literature. The results from this study are comparable to literature values where these rates are defined in terms of surface. In this study, the reactor contained sulfur and limestone with a ratio of 2:1 (mass/mass); in other studies, the medium was only sulfur. The concentrations used in this study were 60 and 90 mg NO3—N/L, while in the Koenig and Liu (2001) studies, the concentrations were different and the porosity in this study was slightly different. Consequently, different half-order reaction rate constants were obtained. The theoretical requirement of alkalinity for complete denitrification based on equation 1 is 427 mg/L as CaCO3. The autotrophic denitrification with sufficient alkalinity (680 mg/L as CaCO3) followed first-order kinetics (Liu and Koenig 2002). When the initial alkalinity decreased to 217 and 113 mg/L as CaCO3 the slopes of nitrate profiles became lower indicating a gradual decrease in the denitrification rates (Liu and Koenig 2002). In these experiments, the result showed that the alkalinity in the raw water was 145 mg/L as CaCO3 and increased to 196 mg/L as CaCO3 after it passed through the bioreactor. These levels of alkalinity were insufficient to allow complete denitrification in the system. Based on calculations using data from other sources in the literature with filter sizes ranging from laboratory to full scale, media size from 3 to 28 mm, porosity from 0.4 to 0.9, with different shapes and with temperature from 6 to 27°C, surface rates varied from 1 to 10 x 10-3 mg1/2/dm1/2 min. This indicates that all filters functioned on the same principle: zero- order reaction in pores that are only partly effective, resulting in overall half-order reaction kinetics (Harremoës 1976). Based on the analysis of the data with zero-, half- and first-order reaction kinetics, it was also found that the half-order reaction rate fitted the data well and the correlation coefficient was higher in this case than with the other reaction orders (Table 3).

Conclusions

The half-order reaction model can be applied to describe a sulfur:limestone autotrophic denitrification system. Nitrate concentra-

Table 3. Correlation coefficient of the reaction rates

Zero order Half order First order

– Column 60 mg NO3 –N 0.81 0.93 0.92 – Column 90 mg NO3 –N 0.87 0.96 0.75 SULPHUR: LIMESTONE REACTORS 191 tions of autotrophic denitrification using sulfur:limestone reactor can be estimated using a half-order reaction equation:

C 1/2 = C 1/2 – 1– k tH e 0 2 1/2 v

where k1/2v for the entire media in the column study is 1.34 and 1.54 1/2 1/2 mg /L h for 60 and 90 mg NO3—N/L, respectively. The half-order reaction rate constant depends on the specific surface area of the reactor 1/2 1/2 media. k1/2a is 0.018 and 0.20 mg /dm h for influent concentrations of 60 and 90 mg NO3—N/L, respectively, for the column reactor.

References

American Society for Testing and Materials (ASTM). 1991. West Conshohocken, Pennsylvania, USA. Batchelor B, Lawrence AW. 1978. Autotrophic denitrification using elemental sulfur. JWPCF 50:1986–2001. Claus G, Kutzner HJ. 1985. Physiology and kinetics of autotrophic denitrification by Thiobacillus denitrificans. Appl. Microbiol. Biotechnol. 22:283–288. Darbi A, Viraraghavan T, Butler R, Corkal D. Unpublished. Flere JM, Zhang TC. 1999. Nitrate removal with sulfur-limestone autotrophic denitrification processes. J. Environ. Eng. 125:721–729. Guidelines for Canadian Drinking Water Quality. 1996. Prepared by Federal- Provincial subcomittee on drinking water of the Federal-Provincial advisory committee on environmental and Occupational Health. Authority of Minister of National Health and Welfare. 6th ed. Harremoës P, Riemer M. 1975. Report on pilot experiments on down-flow filter denitrification. Report from Department of Sanitary Engineering, Technical University of Denmark, Lyngby. Harremoës P. 1976. The significance of pore diffusion to filter denitrification. J. Water Pollut. Control Fed. 48:377–388. Hoek JP, van der Klapwijk A. 1987. Nitrate removal from groundwater. Water Res. 21:989–997. Lampe DG, Zhang TC. 1996. Evaluation of sulfur-based autotrophic denitrification. Proceeding of the HSRC/WERC Joint Conference on Environment. Albuquerque, N.M., May 21–23. p. 444–458. Kapoor A, Viraraghavan T. 1997. Nitrate removal from drinking water-review. J. Environ. Eng. 123: 371–380. Koenig A, Liu LH. 1997. Autotrophic denitrification of nitrified leachate in sulphur packed-bed reactors, p. 283–292. In S. Margherita di Pula (ed.), Proceedings SARDINIA’97 Sixth International Landfill Symposium, Cagliari, Italy, 13–17 October. Koenig A, Liu LH. 2001. Kinetic model of autotrophic denitrification in sulphur packed-bed reactors. Water Res. 35:1969–1978. Koenig A, Liu LH. 2002. Use of limestone for pH control in autotrophic denitrification: batch experiments. Process Biochem. 37:885–893. 192 DARBI AND VIRARAGHAVAN

LeCloirec P. 1985. Mathematical model of denitrification on sulfur-calcium carbonate filters. Chem. Eng. J. 3:B9–B18. Packham R. 1992. Public health and regulatory aspects of inorganic nitrogen compounds in drinking water. Water Supply 10:1–6. Schippers CJ, Kruithof CJ. 1987. Removal of nitrate by slow sulfur/limestone filtration. Aqua 5:274–280. Sutton PM. 1973. Continuous biological denitrification of wastewater. M. Eng. Thesis, McMaster University, Hamilton, Ontario, Canada. Water Qual. Res. J. Canada, 2003 Volume 38, No. 1, 193–210 Copyright © 2003, CAWQ

Removal of Arsenic from Groundwater using Crystalline Hydrous Ferric Oxide (CHFO)

BISWA RANJAN MANNA, SOUMEN DEY, SUSHANTA DEBNATH AND UDAY CHAND GHOSH*

Department of Chemistry, Presidency College, Kolkata - 700073, West Bengal, India

Synthesis of crystalline hydrous ferric oxide (CHFO), a modified iron-based adsorbent, and its arsenic sorption behaviour have been reported. Here, the effects of pH with variation of arsenic concentrations, contact time, pre-drying of CHFO, competition of some other anions and regeneration of arsenic- saturated CHFO are conducted by batch method. The sorption of As(V) is highly dependent on the concentration and pH of the experimental system, while that for As(III) is pH insensitive. As(III) is found to require less contact time to attain equilibrium than that of arsenic(V). Pre-drying of CHFO in the temperature range of 200 to 300°C is found to be effective in removing both As(III) and As(V). Adsorption kinetics follow the first-order Lagergren model. The equilibrium data conform to the Langmuir isotherm. Evaluated Langmuir constants and equilibrium parameter (RL) indicate that CHFO is a better As(III) adsorbent under experimental conditions. Sulphate, phosphate and bicarbonate compete poorly with As(III) sorption. A field test using CHFO-packed fixed-bed column is reported. Effluent water bed volumes of 14,000, 11,000 and 9000 BV (arsenic ≤ 0.01 mg L-1) were obtained in the first, second and third cycle of operation from a groundwater sample (arsenic content: 320–400 µg L-1). Regeneration of the exhausted column was achieved with up to 80 to 85% efficiency using 3 BV of 5 M NaOH solution recycled through the column five times. Arsenic was recovered as As2S3 from the regenerates, to avoid recycling of arsenic into the environment.

Key words: crystalline ferric oxide, arsenic, removal, groundwater, regeneration, recovery

Introduction

In India as in many developing countries in the world, groundwater is the main source of drinking water in rural and semi-urban areas. During the last two decades, it has been reported that groundwater in parts of India and Bangladesh have arsenic contamination above the permissible level (0.01 mg L-1) (Dave 1996) and potability of such water may pose a risk to public health. Arsenic in groundwater is a problem in many countries around the world. Arsenic contamination above 0.01 mg L-1 in natural water has been reported (Smedly and Kinniburgh 2002) in Argentina, Arizona, Bangladesh, Red River Delta in Vietnam,

* Corresponding author; [email protected] 194 MANNA ET AL.

Inner Mongolia, Mexico, Northern Chile, Taiwan, Thailand, Tulare Basin, San Joaquin Valley in California and Southern Carson Desert in Nevada. The scale of the problem in terms of population exposed to high arsenic contamination in the groundwater is very high in the Bengal Basin, where more than forty million people are consuming water with excessive arsenic. Methods for reducing arsenic contamination from water to below the permissible level are based on the principles of coprecipitation-filtration, adsorption, ion exchange and reverse osmosis. Of these, adsorption is thought to be the simplest approach. Use of numerous adsorbents, such as activated carbon (Huang and Fu 1984), aluminium oxide (Rajakovic and Mitravic 1992; Gulledge and O’Conor 1973; Anderson et al. 1976), activated alumina (Rosenblum and Clifford 1984; Clifford and Lin 1995), ferrihydrite (Raven et al. 1998; Jain et al. 1999), amorphous iron hydroxide (Gulledge and O’Conor 1973; Pierce and Moore 1982; Wilkie and Hering 1976), granular ferric hydroxide (Driehaus et al. 1998), goethite (Manning et al. 1998; Matis et al. 1997, 1999), manganese oxides (Driehaus et al. 1995; Oscarson et al. 1983), geological materials (Xu et al. 1988; Bowell 1994), silica gel impregnated with ferric hydroxide (Joshida et al. 1976), basic yttrium carbonate (Wassay et al. 1996), lanthanum compounds (Tokunaga et al. 1997), hydrated zirconium oxide (Manna et al. 1999), hydrotalcite (Manju and Anirudhan 2000) and lanthanum impregnated saw dust carbon (Raji and Anirudhan 1999) have been reported. Aluminium oxide and granular ferric hydroxide have been widely used in West Bengal (India). A recent medical report (World Health Organization 1997) circulated a cautionary notice on the use of aluminium-based compounds for water treatment. Preferred use of granular ferric hydroxide to reduce groundwater arsenic poses a problem because of colloidal adsorbent particles forming in the back- wash water when regenerating packed columns. A study was undertaken to improve the quality of iron-based adsorbent. In this communication, the synthesis of crystalline hydrous ferric oxide (CHFO), a modified adsorbent, is reported together with arsenic sorption characteristics and the regeneration of arsenic saturated CHFO. Finally, recovery of arsenic from arsenic-rich regenerate as water-insolu- ble As(III) sulfide is also incorporated to avoid recycling of arsenic into the environment.

Materials and Methods

Reagents Sodium meta-arsenite and disodium hydrogen arsenate used for preparation of standard stock solutions (1000 mg L-1) were of A.R. (BDH) grade. Silver diethyldithiocarbamate (SDDC) used for arsenic determination spectrophotometrically was of G.R. (E. Merck) grade. All other chemicals were of reagent grade. REMOVAL OF ARSENIC FROM WATER BY MODIFIED ADSORBENT 195

Instruments An Elico pH meter (model LI-127) was used for pH measurement. Arsenic was measured using a spectrophotometer (Hitachi, model U-3210) and hydride generator atomic absorption spectrophotometer (Perkin Elmer-3100). Iron and manganese were also determined by the atomic absorption spectrophotometer. A Nephelo turbidity meter-131 (Systronics) was used to measure turbidity.

Synthesis of Adsorbent Crystalline hydrous ferric oxide (CHFO) was prepared by hydrolyz- ing 0.1 M FeCl3 in 0.01 M HCl with 0.1 M NH3 solution. Addition of the ammonia solution to the ferric chloride was continued until the pH of the supernatant liquid was between 4.0 and 5.0. The brown ferric hydroxide precipitate was aged with the mother liquor for five days. Then, the supernatant liquid was decanted carefully and the precipitate was washed with water until acid free. After filtering the precipitate, it was dried at 40°C in an air oven.

Sorption Experiments Batch sorption determinations were conducted by mechanical agitation at room temperature (30 ± 2°C). Fifty-mL aliquots of varying concentrations of As(III) and As(V) solutions were each treated with 0.2 g of CHFO in 100-mL polythene containers. The pH of the test solutions as well as that of the adsorbent suspension was adjusted as required, using dilute HCl and/or NaOH solution as necessary. Adsorbent particle size was in the range of 0.14 to 0.29 mm for all the experiments. After equili- bration had been reached, the solutions were filtered and the filtrates analyzed for arsenic.

Regeneration of Adsorbent Arsenic-rich CHFO (arsenic content: 66.6 mg/g), was regenerated in a batch reactor at room temperature (30 ± 2°C). The arsenic-rich adsorbent (0.1 g) was equilibrated for an hour with 50-mL aliquots of 1.0 M regenerating solutions of different composition. Regenerating agents tested included KOH, NaOH, Na2CO3, NH3 and NaCl. The best results were obtained using either KOH or NaOH. NaOH being cheaper, was further studied to optimize the concentration and contact time.

Recovery of Arsenic Arsenic-rich regenerate obtained above was acidified with HCl followed by boiling with NaHSO3 until evolution of SO2 ceased. This solution was placed in a vessel fitted with an outlet immersed into 0.1 M NaOH for trapping H2S, and adjusted to 8.0 to 9.0 M with HCl and saturated with H2S. The resulting As(III) sulfide precipitate was filtered 196 MANNA ET AL. through a previously weighed sintered glass crucible, washed with carbon disulfide and alcohol and dried at 110°C to constant weight. The residual H2S gas in the filtrate was oxidized to sulfate by boiling with NaOCl before disposal.

Results and Discussion

X-ray diffraction pattern showed that the synthesized hydrous ferric oxide was crystalline in nature. Thermal gravimetric analysis (Fig. 1) of the compound showed a rapid initial weight loss of 16.50% on drying below 110°C. This was attributed to a loss of physically adsorbed water from the hydrated synthesized product. This was also confirmed from a sharp endogenous differential thermal analysis, DTA, peak at 76.4°C (Fig. 2). A broad exogenous DTA peak between 110 and 360°C and a second-step weight loss of 9.89% in TG analysis are assumed to be due to a chemical change resulting from polymerization between iron atoms through oxo bridges. It was reported (Kraus 1935) earlier that the aging and drying of ferric hydroxide led to an increase in polymerization of up to 40 to 50 Fe atoms in a chain structure resulting in the formation of a crystalline product. Another slow-step weight loss of 1.48% above 360°C is assumed to be due to dehydration with a phase transition, which is confirmed from a sharp exogenous DTA peak between 380 and 499°C.

Fig. 1. Thermal gravimetric analysis of CHFO. REMOVAL OF ARSENIC FROM WATER BY MODIFIED ADSORBENT 197

Fig. 2. Differential thermal analysis of CHFO.

Physico-chemical characteristics of CHFO measured included: iron content (55.85%), moisture content (18.34%), bulk density (1.25 g/cm3), particle size (0.14–0.29 mm) and adsorption capacity (66–68 g As(III)/kg and 55–58 g As(V)/kg at pH 6.0).

Effect of pH The data in Table 1 show that adsorption affinity of CHFO for As(III) and As(V) from their respective solutions depends upon concentration and pH. For arsenic concentrations ranging from 10 to 20 mg L-1, the percentage adsorption of both species onto CHFO was found to be nearly the same up to the initial pH of 6.0. At pH >6.0, As(V) adsorption decreased while As(III) adsorption decreased at pH >9.0. However, when the arsenic concentration was 25 mg L-1, As(V) adsorption percentage decreased at pH >5.0. At 50 mg L-1 concentration, the decrease in As(V) adsorption onto CHFO was found to occur at pH >3.0, while adsorption of As(III) remains nearly constant up to pH 9.0. Thus, it appears that As(V) adsorption is both concentration and pH dependent, while that for As(III) is independent of concentration and pH for pH ≤9. This is presumably due to rapid initial As(V) sorption at higher concentrations on to CHFO which leads to a sharp increase in pH of the solution. As(V) sorption onto CHFO was found to take place via an anion-exchange process in acidic solution. At higher As(V) concentrations, initial rapid uptake onto CHFO at low 198 MANNA ET AL. Arsenic adsorption percentage 10 15 20 25 50 100 Arsenic sorption percentage on to CHFO as a function of pH and concentration Arsenic sorption percentage ; Initial arsenic concentration. Values are average of three experimental results. average of three are ; Initial arsenic concentration. Values 0 → C a a 0 able 1. ↓ T C (mg/L) pH2.03.04.0 As(III)5.0 As(V)6.0 As(III)7.0 98.55 As(V)8.0 98.95 99.999.0 98.89 99.95 As(III)10.0 99.15 99.80 As(V) 98.65 98.80 99.50 98.89 99.85 98.88 99.25 As(III) 99.36 99.65 99.05 As(V) 97.10 98.85 99.15 98.95 98.25 85.75 98.29 96.85 98.89 As(III) 98.50 78.85 99.90 97.83 98.56 67.65 As(V) 98.75 98.99 98.05 95.58 97.86 98.85 97.95 96.65 86.45 As(III) 97.96 95.65 98.50 97.42 77.89 As(V) 99.79 97.65 98.62 65.05 98.10 98.65 97.60 95.05 98.60 98.10 98.01 93.11 85.95 98.93 94.21 97.75 95.91 75.90 99.70 98.73 95.85 63.93 97.56 96.23 98.05 89.99 99.49 86.68 98.56 87.37 75.65 98.48 93.78 69.36 91.07 66.08 99.50 97.61 67.14 58.95 94.34 97.34 98.24 49.83 94.74 60.00 98.40 45.61 94.27 92.22 46.05 28.52 95.14 44.39 20.75 94.94 24.08 95.14 13.42 85.61 12.75 10.87 REMOVAL OF ARSENIC FROM WATER BY MODIFIED ADSORBENT 199

pH (i.e., pH 3.0–6.0) leads to a sharp rise in pH of the system due to lib- eration of OH-. This increase in pH was more pronounced at higher As(V) concentrations. - 2- As(V) exists as H2AsO4 and HAsO4 in solution as dominating I species at pH between 2.0 and 7.0, and 7.0 and 11.0, respectively (pKa , II III pKa and pKa of H3AsO4 are 2.19, 6.94 and 11.5, respectively). As(III) - I exists as As(OH)3 up to pH 9.0 and AsO(OH)2 at pH >9.0 [pKa of As(OH)3 = 9.2] in solution. The adsorbent, CHFO was reported (Amphlett 1964; De and Sen 1978) to function as an anion-exchanger at pH ≤6.0 and a cation-exchanger at pH ≥7.0 with a transition of exchange property in between the two pH values. Hence, at pH ≤6.0, As(V) adsorption onto CHFO can be attributed to anion-exchange (equation 1) as well as electrostatic type of attraction (equation 2).

- + FeOOH + H2AsO4 + H i FeOH2AsO4 + H2O (1)

- + + - FeOOH + H2AsO4 + H i FeOOH2 - - - H2AsO4 (2)

Both adsorption equilibria (equations 1 and 2) are consistent with the increase in final pH of the equilibrated solution. At pH >6.0, As(V) adsorption is thought to take place by the non-specific (van der Waals type) columbic forces.

- + 2- FeOOH + H2AsO4 i FeOOH2 - - - HAsO4 (3)

2- + 3- FeOOH+ HAsO4 i FeOOH2 - - - AsO4 (4)

- + - - + - FeOOH + H2AsO4 + Na + OH i FeOO Na - - - H2AsO4 + H2O (5)

2- + - - + 2- FeOOH + HAsO4 + Na + OH i FeOO Na - - - HAsO4 + H2O (6)

Here, the adsorption phenomena (equations 5 and 6) explain the decrease in final pH of the solution. On the other hand, As(III) adsorption takes place presumably due to ion-dipole type of interactions.

+ + δ- FeOOH + H + As(OH)3 i FeOOH2 - - - O — As(OH)2 (7)  Hδ+ (2.0 ≤ pH ≤ 6.0)

+ - + δ- + FeOOH + Na + As(OH)3 i FeOO Na - - - O — As(OH)2 + H (8)  Hδ+ (6.0 ≤ pH ≤ 9.0)

Increase in pH leads to small increases in As(III) adsorption due to the decrease in surface positive charge density on CHFO particles resulting from a decrease in electrostatic hindrances. The decrease in final pH of the equilibrated solution strongly supports the mechanistic path (equation 8). 200 MANNA ET AL.

At pH >9.0, As(III) adsorption takes place according to equilibrium equation 9 which is consistent with a decrease in pH of the final solution:

- + - - + - FeOOH + OAs(OH)2 + Na + OH i FeOO Na - - - OAs(OH)2 + H2O (9)

The decrease in adsorption of anionic arsenic species on to CHFO is due to an increase in competing OH- for adsorption sites with increasing pH.

Effect of Contact Time The effect of contact time between a fixed amount of CHFO and varying concentrations of both As(III) and As(V) was examined at pH 6.0 in a batch experiment. Here, the time of agitation was varied from one to eight hours. It was found that the percentage adsorption for As(III) was more than that for As(V). For a 50-mg L-1 solution of inorganic arsenic, percentage adsorption for As(III) was 92 ± 1.2 and 97.0 ± 0.8 while that for As(V) was 19.5 ± 1.5 and 70.5 ± 1.0, after 1.0 and 4.0 hours of agitation, respectively. This indicates that CHFO requires less contact time for As(III) removal from the contaminated water than for As(V) removal. Although the percentage adsorption of inorganic arsenic decreased with an increase in concentration, the equilibrium time for maximum adsorption was independent on their concentrations and differed for As(III) and As(V). Minimum contact time for maximum As(III) and As(V) adsorption was 3.0 and 5.0 hours of agitation, respectively.

Effect of Drying Temperature The effect of drying temperature on the adsorption of As(III) and As(V) between 25 and 500°C was studied using 50-mg L-1 solutions at pH 6. Results showed that the percentage uptake for As(III) remained nearly constant (~96.0 to 98.5), while that for As(V) increased from ~67.0 to 95.0 for drying temperature up to 300°C. Drying temperature above 300°C resulted in a decrease in percentage adsorption for both species. Percentage adsorption for As(III) decreased from ~98.5 to ~80.0 while that for As(V) decreased from ~95.0 to ~54.0, when the drying temperature of CHFO increased from 300 to 500°C. The sharp exogenous DTA peak (Fig. 2) at 380 to 450°C, indicating phase transfer of CHFO, is consistent with the decrease in adsorption percentage with increase in drying temperature above 300°C. The increase in As(V) sorption onto pre-dried adsorbent with increasing drying temperature from 25 to 300°C is attributed to an increase in active sites with porosity due to the loss of inter-lattice hydrogen-bonded (physically adsorbed) water molecules. Consequently, it is suggested that the optimum drying temperature for adsorption of inorganic arsenic species from natural water samples is 200 to 300°C. Use of pre-dried CHFO at a temperature between 200 and 300°C was also found to decrease the possibility of incremental iron solubilization from the adsorbent medium even after long contact periods. REMOVAL OF ARSENIC FROM WATER BY MODIFIED ADSORBENT 201

Adsorption Dynamics In the present investigation, two approaches have been used for explaining the adsorption dynamics: (a) kinetic modelling using Lagergren equation and, (b) pore diffusion rate constant.

Kinetic modelling The kinetic study for As(III) adsorption onto CHFO was conducted at pH 6.0 by varying sorbate concentration and temperature of solution. The adsorption rate constants were calculated using the model proposed by Lagergren first-order rate process:

log10(qe - q ) = log10qe - (Kad/2.303)t

-1 where qe and q (both in mg g ) are the adsorption capacity of As(III) at equilibrium and at time t, and Kad is the rate constant. The straight line plots of log10(qe - q) versus time (t) at different initial As(III) concentration (Fig. 4) and temperature (Fig. 3) show the validi- ty of the above model and suggest that As(III) sorption obeys first- order kinetics and that Kad is independent of initial concentration within the range studied. The rate constant, Kad for the adsorbent and for some other reported data is given in Table 2. The increase in Kad value with increase in temperature indicates endogenous sorption of As(III) onto CHFO. The present values of Kad are found to be comparable to those reported for other adsorbents(Manju and Anirudhan 2000; Raji and Anirudhan 1999).

Pore diffusion coefficient Pore diffusion coefficient (D) for intraparticle transport of As(III) has been calculated using the following equation, assuming spherical geometry for the adsorbent.

2 t1/2 = 0.03 r0 /D

where t1/2 is time for half adsorption, and r0 is the mean radius of particle used for the study. The calculated values for D were found to be 46.045 x 10-10, 55.25 x 10-10 and 63.76 x 10-10 cm2/sec for initial As(III) concentrations of 50, 75 and 100 mg L-1, respectively. These values are in the range of 10-9 to 10-12 demonstrating that the adsorption phenomenon is controlled by pore diffusion (Poots et al. 1976).

Adsorption Isotherm The equilibrium data on adsorption of As(III) and As(V) were analyzed separately using the Langmuir adsorption model.

1/qe = 1/(θobCe) + 1/θo 202 MANNA ET AL.

Fig. 3. Kinetic study on As(III) adsorption at different temperatures.

where qe is amount of As(III) or As(V) adsorbed per unit weight of -1 the adsorbent (mg g ), Ce is the equilibrium concentration of As(III) or -1 As(V) in solution (mg L ), θo and b are Langmuir constants related to the sorption capacity (mg g-1) and energy of sorption (L mg-1), respectively. The initial concentration was 50 mg/L. The linear plots of 1/qe versus 1/Ce (Fig. 5) suggest the applicability of the above model for the system, indicating monolayer coverage of sorbate at the outer surface of the adsorbent. The values for θo and b were calculated from the plots and are given in Table 3 together with compar- ative data for some other adsorbents. Comparison of adsorption constants for both As(III) and As(V) sorption onto CHFO with other adsorbents (determined and literature data) indicates that CHFO has good prospects for removing arsenic from contaminated water. The essential characteristics of a Langmuir isotherm may be expressed in terms of dimensionless equilibrium parameter, RL, using the following equation:

Fig. 4. Kinetic study on As(III) adsorption at different concentrations. REMOVAL OF ARSENIC FROM WATER BY MODIFIED ADSORBENT 203 -1 b LISDC Co = 100 mg L -2 -1 a 3 ) x 10 -1 (min ad Co = 100 mg L K -1 CHFO HTCO = 50 mg L c b LISDC -2 a 3 ) x 10 -1 (min ad K CHFO HTCO pH = 6 pH = 8.5 pH = 12 Temperature pH = 6 pH = 8.5 pH = 12 )(°C)Co -1 Kad values for As(III) sorption compared with some reported data with some reported As(III) sorption compared Kad values for Raji and Anirudhan (1999). Anirudhan Raji and Co; arsenic concentration. Manju and Anirudhan (2000). Anirudhan Manju and a b c able 2. T *Co (mg L 5075100 1.16 1.15 1.12 — 1.244 1.233 3.83 — 3.79 30 50 70 1.14 1.21 1.28 1.249 1.486 — 3.76 3.89 — 204 MANNA ET AL.

RL = 1/(1 + bC0)

-1 where C0 is initial concentration of sorbates (mg L ). The values of RL lie between 0 and 1 (Table 3) at different initial sorbate concentrations showing favourable sorption of As(III) and As(V) onto CHFO.

Effect of Other Anions - - - - - The competing effect of other anions such as Cl , NO3 , F , OAc , HCO3 -2 -3 , SO4 and PO4 on As(III) adsorption was investigated at an initial pH of 6.0 by batch method. It was found that the percentage adsorption of As(III) decreased from (98.0 ± 1.0) to (92.5 ± 1.5) with increase in initial concentration ratio (ICR) of phosphate to As(III) in solution from 0 to 1.0. Similar results were also found when ICR of sulfate or bicarbonate to As(III) was varied from 0 to 16. Other anions showed no remarkable effect when ICR was varied from 0 to 10. These results indicate that phosphate has a somewhat greater competing effect with As(III) for sorption sites than that for sulfate or bicarbonate. However, it is apparent that phosphate, sulfate and bicarbonate will all interfere with As(III) removal from groundwater.

Regeneration of Arsenic-Rich CHFO Regeneration studies of As(III)-rich CHFO (As content: 66.6 mg g-1) conducted by the authors showed that 1.0 M solution of either NaOH or KOH desorbed ~60 ± 1% of initial arsenic content while some other regenerating agents desorbed <30%. Regeneration of the adsorbent was tested by varying the concentration of NaOH from 1.0 to 10.0 M. It was found that 5.0 M and 10.0 M NaOH desorbed 78 ± 2% and 83 ± 3% of initial arsenic content, respectively. Increasing the contact time to 24 h did not

Fig. 5. Langmuir adsorption isotherm of arsenic. REMOVAL OF ARSENIC FROM WATER BY MODIFIED ADSORBENT 205 As(III) As(V) value (CHFO) L ) -1 (mg L o Reference C Raji and Anirudhan (1999)Anirudhan Raji and 37.5 0.0631 0.6106 Manju and Anirudhan (2000)Anirudhan Manju and 50.0 0.0418 0.5405 ) -1 )b (Lmg )b -1 (mg g o θ Langmuir constants R As(V) 25.00As(V) 0.017 12.74 0.4215 As(III) 240.73As(III) 0.0105 208.69 0.0083 -1 -1 -1 -1 c Langmuir constants for arsenic sorption compared with reported data with reported Langmuir constants for arsenic sorption compared Co; arsenic concentration. c Particle sizes used were not reported. Particle sizes used were Dosage used per 50 mL solution. : 0.2 g a b c b = 50 mg L a able 3. T AdsorbentHZOParticle size: 0.14-0.29 mm Adsorbate Co = 50 mg L LISDC Hydrotalcite Dose: 2 g As(III) 21.74 0.3116 Manna et al. (1999) 25.0 0.0117 0.7017 Dose: 2 g Dose Dose: 0.2 g Co = 100 mg L Co = 100 mg L CHFOParticle size: 0.14–0.29 mmCo As(III) 33.33 0.396 work Present 12.5 0.168 0.8247 206 MANNA ET AL. improve the percentage regenerated. The 15 to 20% arsenic which is not desorbed even under extreme conditions may be due to the chemisorption or fouling of the adsorbent. This implies that reused CHFO will have 15 to 20% less arsenic removal capacity than the fresh product.

Recovery of Arsenic Arsenic was recovered from the arsenic-rich regenerate as As(III) sulfide by the method described in the above section ‘Recovery of arsenic.’ Quantitative determination showed that 99 ± 0.5% of the arsenic content was recovered. The remaining mother liquor was analyzed for arsenic and found to contain <0.05 mg L-1 arsenic. This is consistent with the very low solubility product of isolated solid sulfide (Ksp -63 of As2S3 ~10 ). Thus, the solution obtained after recovery of arsenic can be discharged safely onto surface soil. This avoids the recycling of arsenic back into the environment.

Field Experiment

A glass column (internal diameter: 2.0 cm) was packed with CHFO (grain size: 0.3–0.5 mm) to a height of 16 cm and connected to a tube-well in a residential house at Barasat (Bengal Basin, West Bengal, India). The tube-well water (arsenic content: 320–400 µg L-1) was passed through the bed at a flow rate of ~60 BV/h (contact time: ~18 sec). The arsenic in the effluent was determined at specified bed volumes (Fig. 6). Fourteen thousand BV of water with arsenic ≤0.01 mg L-1 were obtained from the system. Some other water quality parameters were also determined includ- - - -3 -2 ing pH, hardness, turbidity, Fe, Mn, Cl , NO3 , PO4 , SO4 , and TDS (Table 4). It was found that the CHFO-packed column not only decreased arsenic

Fig. 6. Removal of arsenic by fixed bed CHFO. REMOVAL OF ARSENIC FROM WATER BY MODIFIED ADSORBENT 207 TDS -3 4 PO -2 4 SO - 3 NO - (ppm) (NTU) (ppb) (ppb) (ppb) (ppm) (ppm) (ppm) (ppb) (ppm) Different water quality parameters of influents and effluents at different run lengths run at different water quality parameters of influents and effluents Different ↓ → able 4. 1000 7.49 264 1.11 8.86 44 2.0 14.37 3.15 21.78 2.2 442 100020003000 7.394000 7.415000 7.366000 260 7.497000 281 7.538000 272 7.519000 3.05 270 7.5910000 1.99 262 7.461 1.51 265 7.6012000 0.32 1.32 7.55 25613000 1.25 1.56 25914000 0.93 1.40 7.56 272 290 0.81 269 1.12 7.51 220 2.50 1.23 7.58 170 2.20 261 1.37 190 12 1.19 3.70 266 120 11 5.23 271 9.0 140 1.25 7.40 8.20 22.11 7.0 1.16 80 23.74 7.0 1.29 62 21.19 8.77 5.0 49 4.77 20.74 66 9.2 6.0 4.51 18.98 9.78 3.93 8.0 19.79 51 29.89 3.91 5.5 27.97 3.0 20.12 4.11 28 29.21 19 17.65 3.97 5.0 31.56 3.5 18.11 3.79 9.0 16.52 25.18 2.6 7.1 3.11 22.05 1.9 5.2 465 2.93 11.25 20.76 2.95 3.6 420 21.98 9.33 461 2.9 10.15 18.66 3.08 472 20.42 3.1 465 3.9 3.01 2.95 455 19.37 2.7 1.8 449 17.18 18.75 470 1.7 462 469 2.5 3.3 449 455 447 T BV pH Hardness Turbidity As Fe Mn Cl 15000 7.61 259 1.18 9.96 37 2.0 12.31 3.11 17.98 4.8 452 Raw water 7.31–7.59Effluent 312–362water 5.55–6.75 320–420 1490–2160 50–66 12.98–21.75 5.32–7.75 26.5–38.1 13–22 578–680 208 MANNA ET AL. but also resulted in substantial reduction of the other parameters in the influent water including iron. When arsenic in the effluent water reached 0.01 mg L-1, the column was regenerated using 3 BV of 5.0 M NaOH solution with a down-flow rate of 3 BV/h. The regenerating solution was recycled five times through the column. This removed 80 to 85% of the adsorbed arsenic from the column. Finally, the column was washed with de-ionized water and then with 0.01 M HCl and water, alternately until the pH of the effluent was between 6.0 and 7.0. The regenerated column was used again and a further 11,000 BV of water was obtained with arsenic concentration <0.01 mg L-1 (Fig. 6). Regenerating the column as above, it was again used for a third cycle. This time 9,000 BV of water of potable standard was collected (Fig. 6). Results indicate that the arsenic removal efficiency of the CHFO column decreases by 15 to 20% after each cycle of use. This decrease is consistent with the extent of regeneration as mentioned earlier.

Conclusions

(i) Absolute adsorption capacity of CHFO for arsenic (III) is somewhat greater than that for arsenic(V) at pH 6.0. (ii) Arsenic(III) adsorption is greater than arsenic(V) adsorption at pH >6.0 in the range of arsenic concentration 10 to 20 mg L-1 while at concentration ≥25 mg L-1, similar results are obtained at pH <6.0. (iii) The adsorption process follows a first-order Lagergren kinetic model and the data fit the Langmuir isotherm. (iv) Phosphate, sulfate and bicarbonate interfere negatively with the adsorption of As(III) onto CHFO. (v) Regeneration of arsenic-rich CHFO is tedious but 80 to 85% regeneration is possible by 5.0 to 10.0 M of NaOH solution. (vi) The field test reveals 15 to 20% less arsenic removal efficiency after each column regeneration. (vii) Recovery of 99.0 ± 0.5% arsenic is possible as arsenic(III) sulfide from the regenerates. This avoids the disposal of the hazardous waste.

Acknowledgements

We are grateful to U.G.C. (New Delhi) for financial support and fellowship to one of the authors. We are also grateful to the Head, Department of Chemistry, Presidency College, Kolkata, India, for providing facilities.

References

Adamson AW. 1977. Physical chemistry of surfaces. 3rd Edition. Interscience Publishers, New York. Amphlett CB. 1964. Inorganic ion exchanger (Monograph-2). Elsevier. p. 84–133. Anderson MA, Ferguson JF, Gravis J. 1976. Arsenate adsorption on amorphous aluminium hydroxide. J. Colloid Interface Sci. 54:391–399. REMOVAL OF ARSENIC FROM WATER BY MODIFIED ADSORBENT 209

Bowell RJ. 1994. Sorption of arsenic by natural iron oxides and oxyhydroxides in soil. Appl. Geochem. 9:279–286. Clifford D, Lin CC. Feb 1995. Ion exchange, activated alumina and membrane process for arsenic removal from groundwater. In Proceedings of the 45th Annual Engineering Conf., University of Kansas. Dave JM. 1996. Arsenic contamination of drinking water in Bangladesh. SEA/EH/500. New Delhi, India, South East Asia Region, World Health Organization. De AK, Sen AK. 1978. Review – synthetic inorganic ion exchangers. Sep. Sci. Technol. 13:517–540. Driehaus W, Jeckel M, Hildbrandt U. 1998. Granular ferric hydroxide – a new adsorbent for the removal of arsenic from natural water. J. Water SRT-Aqua 47:30–35. Driehaus W, Seith R, Jeckel M. 1995. Oxidation of arsenic(III) with manganese oxides in water treatment. Water Res. 29:297–305. Gulledge JH, O’Conor JT. 1973. Removal of arsenic(V) from water by adsorption on aluminium and ferric hydroxides. J. Am. Water Works Assoc. 65:548–553. Huang CP, Fu PLK. 1984. Treatment of arsenic(V) contaminated water by activated carbon process. J. Water Pollut. Control Fed. 56:233–240. Jain A, Raven KP, Loeppert RH. 1999. Arsenite and arsenate adsorption on ferrihydrite: surface charge reduction and net OH- release stoichiometry. Environ. Sci. Technol. 33:1179–1184. Joshida I, Kobayshi H, Veno K. 1976. Selective adsorption of arsenic ions on silica gel impregnated with ferric hydroxides. Anal. Lett. 9:1125–1129. Kraus A. 1935. A review. Kolloid Z 72:18–28. Manju GN, Anirudhan TS. 2000. Treatment of arsenic(III) containing wastewater by adsorption on hydrotalcite. Ind. J. Environ. Health 42:1–8. Manna BR, Bhat SC, Dasgupta M, Ghosh UC. 1999. Studies on removal of arsenic from water using hydrated zirconium oxide. J. Chem. Environ. Res. 8:51–56. Manning BA, Fendrof SE, Goldberg S. 1998. Surface structures and stability of arsenic(III) on goethite: spectroscopic evidence for inner-sphere complexes. Environ. Sci. Technol. 32:2383–2388. Matis KA, Zouboulis AI, Malamas FB, Afono MDR, Hudson MJ. 1997. Floatation removal of arsenic(V) onto goethite. Environ. Pollut. 97:239–245. Matis KA, Zouboulis AI, Zamboulis D, Voltadorou AV. 1999. Sorption of arsenic(V) by goethite particles and study of their flocculation. Water Air Soil Pollut. 111:297–316. Oscarson DW, Huang PM, Liow WK, Hammer UT. 1983. Kinetics of oxidation of arsenite by various manganese dioxides. Soil Sci. Soc. Am. J. 47:644–648. Pierce ML, Moore CB. 1982. Adsorption of arsenic on amorphous iron hydroxide. Water Res. 16:1247–1253. Poots VJP, Mckay G, Healy JJ. 1976. Removal of acid dye from effluent using naturally occuring adsorbents – II Peat. Water Res. 10:1067. Rajakovic LV, Mitravic MH. 1992. Arsenic removal from water by chemisorption filter. Environ. Pollut. 75:279–287. Raji C, Anirudhan TS. 1999. Sorption characteristics of As(III) on surface modified saw dust carbon. Ind. J. Environ. Health 41:184–193. Raven KP, Jain A, Loeppert RH. 1998. Arsenite and arsenate adsorption on ferrihydrite: kinetics, equilibrium and adsorption envelops. Environ. Sci. Technol. 32:344–349. 210 MANNA ET AL.

Rosenblum E, Clifford D. 1984. The equilibrium arsenic sorption capacity of activated alumina. A report. U.S. EPA–600 32–83–107. Smedly PL, Kinniburgh G. 2002. A review of the source, behaviour and distribution of arsenic in natural water. Appl. Geochem. 17:517–568. Tokunaga S, Wasay SA, Park SW. 1997. Removal of arsenic(V) ion from aqueous solution by lanthanum compounds. Water Sci. Tech. 35:71–78. Wassay SA, Haron MJ, Uchivmi K, Tokunaga S. 1996. Removal of arsenite and arsenate ions from aqueous solutions by basic yttrium carbonate. Water Res. 30:1143–1149. Wilkie JA, Hering JG. 1976. Adsorption of arsenic onto hydrous ferric oxide: effects of adsorbate/adsorbent ratios and co-occurring solutes. Colloid Surf. A 107:97–110. World Health Organization. 1997. Aluminium. Environmental Health Criteria 194. Geneva. 282 p. Xu H, Allard B, Grimvall A. 1988. Influence of pH and organic substances on the adsorption of arsenic(V) on geological materials. Water Air Soil Pollut. 40:293–305.