State of the Knowledge and Practice: Assessing the Adequacy of Urban Nutrient Load Measurements

Supporting Document 3: Whitepaper

Prepared for Canadian Water Network by Duncan Boyd

March 2018

2

CONTENTS FOREWORD...................................................................................................................................... 3

ACKNOWLEDGEMENTS ................................................................................................................... 4

1. BACKGROUND AND OBJECTIVES ................................................................................................. 5

2. REVIEW AND CATEGORIZATION OF GENERAL MANAGEMENT OBJECTIVES ............................... 6

2.1 Loadings from Watersheds .................................................................................................. 8

2.2 Critical Source Areas (CSAs) and Land Use Effects ............................................................. 12

2.3 Best Management Practices (BMPs), Low Impact Development (LID), and Infrastructure

Improvement ............................................................................................................................ 14

3. REVIEW OF MONITORING/MODELING APPROACHES AND ASSESSMENT OF RELEVANT

LIMITATIONS OR DEFICIENCIES ..................................................................................................... 16

3.1 Loadings from Watersheds ................................................................................................ 19

3.2 Critical Source Areas (CSAs) and Land Use Effects ............................................................. 27

3.3 Best Management Practices (BMPs), Low Impact Development (LID), and Infrastructure

Improvement ............................................................................................................................ 33

4. CONCLUSIONS AND RECOMMENDATIONS FOR DEVELOPMENT OF AN URBAN NPS NUTRIENT

LOAD WATERSHED DECISION FRAMEWORK ................................................................................. 39

5. REFERENCES............................................................................................................................... 42

APPENDIX A: EPA RISK ASSESSMENT GUIDANCE FOR SUPERFUND: VOLUME III - PART A,

PROCESS FOR CONDUCTING PROBABILISTIC RISK ASSESSMENT .................................................. 56

3

FOREWORD

Water managers and municipal, provincial and federal policy “decision-makers” are the intended audience for this assessment, therefore it is important to clarify that this exercise is about improving the application of science to decision-making rather than a comprehensive assessment of all monitoring and modeling activities related to assessment of urban water quality and estimation of urban nonpoint source nutrient loads. The contents have been restricted to a small number of representative examples that illustrate the concepts being discussed. This review is intended to inform decision-makers: those tasked with making decisions regarding the development and implementation of policies and/or regulatory instruments, as well as those responsible for determining the allocation of resources for flow management and pollution abatement infrastructure. The ultimate goal of this review exercise is to provide guidance on the optimum allocation of resources through examination of decision-making information requirements and the implications of uncertainty associated with water quality and nutrient load assessments. A previous draft of this whitepaper was reviewed by an expert Advisory Group and was used as background material at Canadian Water Network’s workshop held in Toronto on February 6 and 7, 2018. This updated version incorporates insights generated at the workshop and is intended to inform the development of a framework to help guide future actions and decisions by watershed management agencies and organizations.

4

ACKNOWLEDGEMENTS

Canadian Water Network would like to acknowledge the important contributions made by advisory group members and additional experts who participated in the February 2018 workshop, and thank them for sharing their time and expertise. The information in this whitepaper is based on views expressed by the respective contributors and does not necessarily represent the views of their employers. ADVISORY GROUP MEMBERS Duncan Boyd, D.A. Boyd Associates (Advisory Group Chair)

Jonathan Bastien, Hamilton Conservation Authority

Nandita Basu, University of Waterloo

Sandra Cooke, Grand River Conservation Authority

Mohamed Mohamed, Ontario Ministry of Environment and Climate Change

Frank Quarisa, Toronto Water

Christopher Wellen, Ryerson University

Ram Yerubandi, Environment and Climate Change Canada

OTHER CONTRIBUTORS Bernadette Conant, Canadian Water Network

Jiri Marsalek, Consultant

Korice Moir, Project Coordinator

Bill Snodgrass, Toronto Water

This project was undertaken with the financial support of:

Ce projet a été realisé avec l’appui financier de:

5

1. BACKGROUND AND OBJECTIVES Lake and watershed management objectives comprise a wide range of outcomes associated with the protection and restoration of aquatic habitat and water quality. Designing policies and establishing best management practices (BMPs) for controlling and reducing nutrient inputs to water bodies from urban nonpoint sources is one such objective typically driven by a need to mitigate problems associated with nutrient-enriched receiving waters. This will generally require an assessment of receiving water quality as well as some investigation of nutrient loads from potential sources. Evaluating how well current approaches to estimating nutrient loads meet management decision needs requires clear management objectives that describe how the results are to be used. Well-defined objectives will make it possible to assess the degree to which uncertainty in the estimation of nutrient loadings compromises or constrains decision making and such an exercise will help improve the alignment between nutrient loading study objectives and methods. It will also clarify under what circumstances current practices may be inadequate to support management decisions. There is an extensive and varied literature on the subject of nonpoint source (NPS) nutrient loads, but it is apparent by surveying the peer reviewed literature that over time there has been greater emphasis on agricultural and rural land uses than urban landscapes1. Although there are many similar challenges regarding data collection and interpretation associated with loading estimate methodologies, there are also several differences. For example, it has long been documented that urban landscapes tend to exhibit a different (flashier) hydrological response to wet weather events due to their typically greater proportion of impervious surfaces (Leopold 1968, Hutchinson et al. 2011, Lamera et al. 2014). Another difference is the tendency for rural and agricultural landscapes to exhibit a summer minimum in nutrient exports due to the summer growing season (Mohamed 2012) which is generally not as apparent in urban landscapes (Long et al. 2014, Wellen et al. 2014b). There is also ongoing discussion as to whether there may be a difference in the proportion of P- load that is bioavailable driven by differences in the particulate and soluble fractions and variability in the bioavailable fractions of particulate P (i.e. surface adsorbed P, iron-bound P, and extractable organic P). Agriculture increases dissolved P concentrations as a result of livestock waste and fertilizer application to cropland (e.g., Howarth et al. 1996) whereas urbanization tends to increase dissolved P concentrations in runoff waters because of discharge of sewage and septic effluents, as well as lawn fertilizer application (Hatt et al. 2004, O’Brien and Wehr 2010). A Danish study by Egemose and Jensen (2009) found that one-third of the P-load occurred as particulate P independent of catchment type. For urban runoff, 62% of particulate P was bioavailable with the corresponding value

1

A quick keyword search on ScienceDirect for "urban non-point source" + “nutrients” yields 35 hits dating back to

1980 compared with 463 for "agricultural non-point source" + “nutrients” over a similar period.

6

for agricultural tributaries being 76%. For both types of catchments more than 70% of total P (particulate P and dissolved P) was bioavailable. The following review addresses the relative paucity of information on how well current scientific knowledge and practice related to assessment of urban water quality and NPS nutrient loads meet management and policy decision making needs by:

(a) Categorizing potential management objectives for nutrient assessment and loading studies;

(b) Reviewing monitoring/modeling approaches associated with these different management objective categories and

(c) Providing recommendations for development of an urban NPS nutrient load watershed management decision-making framework.

2. REVIEW AND CATEGORIZATION OF GENERAL MANAGEMENT OBJECTIVES Concerns with respect to nutrient status in any water body are generally driven by an interest in excessive or insufficient primary productivity. Cultural eutrophication resulting in excessive productivity has typically been the principal nutrient-related concern of watershed management agencies due to the alteration or disruption of aquatic habitat and other water uses. Issues include increased production of phytoplankton leading to diminished water clarity and/or oxygen depletion, production of nuisance algae (e.g. benthic filamentous algae such as Cladophora), or a shift to cyanobacteria (blue-green algae) species capable of producing toxins (harmful algal blooms, or HABs) or taste and odour related compounds. In the most general sense, then, the fundamental decision driver for assessment of nutrient status and loads has been mitigation of the undesirable effects of eutrophication. Although phosphorus has been identified as the overall principal nutrient of concern in most temperate freshwater systems (Schindler 1974, 2012) it is worth noting that it is seldom the only component of water quality assessments undertaken by watershed management agencies and researchers. Many monitoring programs associated with characterizing nutrient status typically include some measure of total and soluble phosphorus/nitrogen/carbon species and chlorophyll. Additionally, watershed management concerns about water quality in urban settings are generally not limited to eutrophication Issues and issues associated with drinking water, aesthetics, recreational use and the restoration and protection of aquatic habitat may all be important to decision makers responsible for water quality in rivers and streams. Corresponding monitoring programs typically include other general chemistry measurements such as temperature, pH, alkalinity, BOD, suspended/dissolved solids and turbidity (Long et al. 2014, 2015, Boyd et al. 1999, Mohamed 2012) and many studies also include some microbiological assessment to identify sources of potentially pathogenic organisms (Boyd et al. 1999, Mohamed 2012). Additional water quality concerns in urban environments can also be

7

related to such contaminants as chloride, oil and grease, heavy metals, PAHs, and other in-use or legacy persistent organic pollutants (POPs) such as PCBs, and pesticides (Boyd et al. 1999). Another point worth noting is that for municipalities and Conservation Authorities concerns related to management of tributary flow have generally been an overarching priority that predates a focus on nutrient enrichment and receiving water eutrophication. In Ontario for example, the Conservation Authorities Act was legislated by the provincial government in 1946 to create the means to manage erosion and flood risks at a watershed scale. Following Hurricane Hazel in 1954 this Act was amended to provide Conservation Authorities (CA) with a mandate to manage flood risk through enforcement of zoning restrictions on flood plain development. Significant long-term investments have been made by municipalities to address issues associated with urban hydrology and stormwater management and the need to increase resiliency to storms and extreme weather associated with climate change has been identified as a pan-Canadian priority (Canadian Municipal Water Consortium 2015). Toronto's “Wet Weather Flow Master Plan” and a 25-Year Implementation Plan was adopted back in 2003 and represents a good example of the municipal interest in reducing the adverse impacts of wet weather flow (D’Andrea et al. 2004; Toronto Water 2017). In Hamilton and Toronto infrastructure projects to intercept and store CSO flows to allow subsequent conveyance and treatment at Wastewater Treatment Plants have been undertaken to reduce or prevent the discharge of combined stormwater and sanitary sewage into Lake Ontario, Toronto Harbour, Hamilton Harbour and the Cootes Paradise Marsh. These have been driven by an interest in improving overall water quality and diminishing public health risks associated with exposure to pathogens rather than a specific interest in solely reducing nutrient loads. Although management of eutrophication is frequently only one of numerous watershed management agency concerns, a review of the scale, scope and approaches taken to urban NPS phosphorus (P) loadings, and an assessment of the adequacy of current knowledge and practices will be of value when considering management of other water quality issues. Monitoring and modeling activities corresponding to the wide-ranging interest in managing flow and water quality issues by watershed management agencies can support management of nutrient and eutrophication issues where warranted since any measurement of flow into and from urban tributaries can be used to estimate nutrient loads when combined with a suitable range of nutrient concentration measurements. Eutrophication issues can range from a synoptic, regional scale down to a localized embayment or tributary reach. Regardless of the geographic scale, it will be convenient to partition eutrophication study projects and investigations according to their general management objectives which will usually fall into one or more of the following categories:

8

(a) Estimating loadings from watersheds to receiving water bodies; (b) Identifying critical source areas (CSAs) and assessing land use effects on

tributary water quality; and (c) Assessing and/or predicting the effectiveness of policies, best management

practices (BMPs), low impact development (LID), or infrastructure improvement projects.

Although these categories are by no means mutually exclusive, they tend to operate at different (if overlapping) scales and have differing objectives and data requirements which will generally require differing survey designs and loading estimation methodologies.

2.1 Loadings from Watersheds

This is by far the most common category for nutrient loading studies. Across Canada there are numerous examples of eutrophication management concerns on a regional scale involving measurement of, and target setting for, nutrient loadings. Typical needs associated with this category of investigation would include not only estimation of loads delivered from the watershed but also an interest in the subsequent fate and transport including assessment of bioavailability and use in receiving water eutrophication modelling or mass balance exercises. A cross section of these could include Lake Winnipeg, Lake of the Woods, Lake Erie, Lake Simcoe and Bay of Quinte (Lake Ontario). Eutrophication issues in Hamilton Harbour and Cootes Paradise marsh (Lake Ontario) and Lac St. Charles (Quebec) cover a smaller geographic scale but are similar in the sense that the problem is manifested in a receiving water body with limited assimilative capacity downstream from multiple point and nonpoint sources. In all these cases, the primary focus on assessment of urban NPS nutrient status and loadings will be the export from one or more watersheds into the water body of interest and hence on assessing conditions at watershed mouths. From this perspective, tributaries are considered as sources of nutrients analogous to industrial or municipal point sources despite being categorically different in that they represent the integrated influence of upstream inputs and land uses rather than the output from a treated wastewater stream. Lake Winnipeg

A regional scale effort is being made to assess and manage the effects of human activities throughout the Lake Winnipeg watershed that have resulted in eutrophication effects such as large HABs in the north basin of the lake (Environment Canada and Manitoba Water Stewardship 2011) capable of producing neurotoxins (Bishop et al. 2017). The watershed of 953,000 square kilometres is very large relative to lake’s surface area with a ratio of about 40:1 which is much larger than for Lake Simcoe or the Great Lakes (<5:1).

9

This large ratio raises the potential for NPS loadings of nutrients, contaminants, and sediments derived from human activities to exceed the lake’s assimilative capacity (Lake Winnipeg Stewardship Board 2006) given its mean depth of 12 m and relatively small volume. Urban and agricultural land uses are the predominant nutrient sources in the watershed. Although 6.6 million people live in the Lake Winnipeg watershed with about 80 per cent of the population in major urban centres (Edmonton, Calgary, Saskatoon, Regina, Brandon, Winnipeg, Grand Forks and Fargo, North Dakota) the population is distributed widely throughout the extremely large watershed and the lowest population densities are located in the sub-watersheds immediately bordering Lake Winnipeg. This means that although transfers downstream through tributaries have implications for Lake Winnipeg, urban NPS P loads will tend to have more immediate local effects in these urban centres. The highest P export has been observed in river reaches characterized by urban and residential development attributed to nutrient loading from wastewater treatment facilities as well as an unquantified contribution urban NPS P loads. Lake of the Woods

The Lake of the Woods (a large, trans-boundary water body located at the junction of the provinces of Ontario and Manitoba, and the U.S. state of Minnesota) is another area where local and national concerns with eutrophication, HABs and declining water quality have led to the development of a multilateral effort to set P and achieve loading targets. External P loads to this water body are dominated by tributary inputs from the Rainy River, however urban land uses account for less than 0.2% of the watershed so effects of urban or cottage development will only manifest themselves on an extremely local scale (Clark and Sellers 2014). A range of P concentration targets have been proposed for tributaries in the watershed although quantitative allocations of nutrient loads are not yet finalized. Lake Erie

The Great Lakes Water Quality Agreement 2012 Nutrient Annex (Annex 4) outlines binational commitments to set P concentration and loading targets for Lake Erie, develop programs to reduce P loads from urban, rural, industrial and agricultural sources and identify priority watersheds that contribute to local algae development. There is a clear, primary focus on Lake Erie in this binational context driven by the chronic occurrence of HABs in the Lake Erie Western Basin and hypoxia in the Central Basin. Urban NPS loads are a minor contributor for Lake Erie on a regional scale since there is only 3.3% urban land use on the Canadian portion of the drainage basin (OMNRF 2015) and 11% on the U.S. side (Myers et al. 2000). The urban NPS influence will be higher on a more local scale on the U.S. side of the border since the U.S. Cities of Detroit and Toledo contribute directly to the Western Basin and Cleveland contributes directly to the Central Basin. There is no equivalent situation in Canada for Lake Erie where the largest urban centres (Guelph, Kitchener-Waterloo, Brantford, London) are far upstream in agriculturally-dominated watersheds. The Western Basin of Lake Ontario is the most obvious example of a situation

10

where a large urban population lives adjacent to the lake. Urban land use in the Golden Horseshoe approaches 20% (OMNRF 2015) however Lake Ontario’s much greater depth and assimilative capacity mean that there are no HAB or hypoxia issues analogous to the Lake Erie situation. Despite having achieved the binational TP water quality concentration target of 10 ug/l, both the Eastern Basin of Lake Erie and the Western Basin of Lake Ontario have experienced a resurgence in issues associated with “nuisance” benthic algae, principally but not exclusively Cladophora, as the result of benthification following colonization by invasive dreissenid mussels (Hecky et al. 2004, Higgins et al. 2005, Higgins et al. 2012). The complex nature of the interaction between P inputs to the nearshore zone and the subsequent cycling in areas having high density mussel beds means that there are currently no quantitative P load reduction targets for these areas. However, a need to limit further inputs as the result of increased population growth and urbanization could certainly drive an interest in measuring and managing urban NPS nutrient loads. Lake Simcoe

Lake Simcoe represents another regional scale multi-lateral effort to measure and reduce P loads to manage eutrophication and achieve desired dissolved oxygen targets in support of the cold-water fishery. The Lake Simcoe Protection Plan is being implemented by all levels of government and has set P load reduction targets for point sources, as well as urban and rural NPS (Lake Simcoe Science Advisory Committee 2008). As with Lake Erie, urban land use accounts for a relatively small proportion of the watershed, in this case about 6% (Lake Simcoe Science Advisory Committee 2012) but urban sources and urbanization have been identified as stressors and the relative contribution of direct urban runoff to the lake has been estimated at from 5% to 11% (Winter et al. 2007). As with western Lake Erie and Lake Ontario, urban NPS P loads can be expected to have a relatively larger influence on a more local scale. For example, the City of Barrie is situated at the west end of the lake's western arm (Kempenfelt Bay) and the City of Orillia is situated at its northern end and, if present, localized eutrophication issues could amplify the interest in achieving NPS P load reductions in these areas. Bay of Quinte

The Bay of Quinte in Lake Ontario has been identified as a Canadian Area of Concern (AOC) to the International Joint Commission (IJC) and is developing a Remedial Action Plan (RAP) with support from all levels of government and input from public advisory groups. The primary eutrophication issues are perennial HABs and excessive algal biomass. Phosphorus concentration targets have been set for the Upper Bay and major tributaries and a P reduction strategy has been adopted that is to be based on best available science and modeling studies, and best available technology for wastewater and stormwater treatment. P load reduction targets have been set that include 20% for agricultural land

11

use, 60% for municipal and industrial wastewater, and 50% for urban NPS (Bay of Quinte Remedial Action Plan 2016). Hamilton Harbour

Hamilton Harbour has also been identified as a Canadian AOC to the IJC and is also developing a RAP with support from all levels of government and industry along with input from public advisory groups. Occasional HABs, limited water clarity and chronic hypolimnetic anoxia have led to P load reduction targets as a means of achieving reduced receiving water P concentrations. These reductions have been allocated across point and nonpoint sources and significant load reductions have already been achieved compared with historical values through improvements at municipal Wastewater Treatment Plants (WWTPs). A further significant reduction will be achieved within the next five years from the largest WWTP (Hamilton) at which point the relative significance of NPS P loads from the four large tributaries that drain into the harbour will increase and be a significant driver of fluctuations in water quality and trophic status (Bay Area Restoration Council 2014). An intensive, event-based tributary monitoring program was conducted from 2010 to 2012 which led to an updated assessment of P loads from both urban and rural watersheds (Long et al. 2014, 2015). Lac Saint-Charles

Lac Saint-Charles is the drinking water reservoir for Québec City and has experienced accelerated nutrient enrichment due to increased human activities in its drainage basin. Previous studies have indicated that although eutrophic conditions are being mitigated by a rapid flushing rate, in late summer bottom waters were anoxic and P enriched, and surface phytoplankton concentrations were high (Tremblay et al. 2001). Harmful cyanobacterial blooms were first recorded in autumn 2006 and have become a chronic concern due to the potential for algal toxins to adversely affect the drinking water supply (Rolland et al. 2013). Summary

The regional scale nature of the eutrophication concerns in Lake Winnipeg, Lake of the Woods, Lake Erie, Lake Simcoe, Bay of Quinte, Hamilton Harbour and Lac Saint-Charles, have all created an interest in P loading estimates from various point and nonpoint sources. Load reduction targets and allocations are currently in place for Lake Erie, Lake Simcoe, Bay of Quinte and Hamilton Harbour but are still being developed elsewhere. Other than Lake of the Woods, urban NPS P loads have been identified as potentially meriting management in all these regions however in all cases other than western Lake Ontario, the urban footprint is relatively small. This means that the benefits of urban NPS P management will tend to be most pronounced on a local scale and within watersheds. This is an important consideration when developing and communicating the objectives for monitoring and modeling activities related to urban NPS P loads.

12

2.2 Critical Source Areas (CSAs) and Land Use Effects

Assessment of nutrient status and loadings within urbanized watersheds will typically be driven by an interest in identifying potentially manageable nutrient sources that contribute to watershed export or an interest in amelioration of a eutrophication issue at a local scale. It is evident that tracking down and controlling nutrient sources within a watershed will ultimately be critical to reducing watershed loadings as well as resolving within-watershed eutrophication problems. However, it is also evident that this represents a significantly greater challenge than estimating exports from the mouth of large watersheds. The increasing progression of smaller sub-watersheds dictates that identification of nutrient contributions within large watersheds requires an increasing sampling effort from the mouth to head-water areas and for this reason studies to identify critical source areas of P in urban watersheds are not as prevalent or well documented as those seeking to estimate watershed loads. Although conceptually similar, the distinction between assessment of land use effects and identification of Critical Source Areas (CSAs) is really a matter of scale. There does not appear to be any specific scale-based working definition for a CSA. The term is widely used in rural and agricultural landscapes to refer to high-contribution regions or sub-watersheds but has also been applied to mapping at a sub-field scale for managing legacy P in agricultural catchments (Thomas et al. 2016). It may be most useful to consider a CSA as a high priority area that has been identified at a sufficiently fine scale to allow implementation of a specific remedial action or policy. The scales could vary widely ranging from extremely local re-engineering of stormwater drainage from an individual property to management of runoff from large areas under construction. This purely functional definition could also be applied to the large-scale implementation of a street cleaning and catch basin maintenance program throughout entire regions of a municipality, although this will be of limited practical value unless a distinction is being made between areas and application of on-the-ground actions. Ideally, identification of localized CSAs that are contributing excessive nutrient loads would be accomplished by sampling nutrient exports from all constituent sub-watersheds and this would form the basis of subsequent urban stormwater quantity and quality management. The Grand River Water Management Plan identifies the relative yield from nine major sub-watersheds (Holeton 2013) but such an approach entails a sampling effort that is not always achievable due to resource limitations. More commonly, limited sampling results are used to generate export coefficients (typically in rural areas or other relatively large homogenous areas), and/or event-mean concentrations (EMCs, typically on a smaller scale in urban areas) and then combined with distributed hydrologic models or watershed models to estimate contributions from catchments or local “hydrologic response units” (Driver and Tasker 1990, LeBoutillier et al. 2000, Niraula et al. 2012). The consensus from a series of workshops undertaken in 2014 to gain a better understanding of urban P loads and develop an approach for inventorying urban and non-

13

farm rural nonpoint source P loads to Lake Erie was that there are not a lot of data on the concentration of P in raw stormwater (McElmurry 2014). Historically, characterization of urban stormwater runoff and loadings to the Great Lakes was included as part of the International Joint Commission (IJC) Pollution from Land Use Activities Reference Group (PLUARG) studies over the period 1972 – 1979 (e.g. Marsalek 1978). Although the work was undertaken to estimate contaminant loads to the Great Lakes, results were based on sampling in urban test catchments in southern Ontario. The U.S. Environmental Protection Agency (U.S. EPA) Nationwide Urban Runoff Program (NURP) ran from 1978 to 1983 and generated abundant data on nutrients and other pollutants in urban stormwater (U.S. EPA 1983). The program objectives were to characterize stormwater quality across a range of land use types, determine the extent to which urban runoff contributes to water quality problems across the U.S. and assess the effectiveness of pollution control management practices. It should be noted that the urban catchments studied under the NURP program did not assess the various uses to the same degree as PLUARG and the data were collected prior to implementation of any BMP and LID practices. Since these historical data reflect conditions without any effective runoff pollution controls they can be assumed to represent “worst case” conditions. More recent examples of studies undertaken specifically to identify urban CSAs include work in Madison Wisconsin assessing runoff water quality from a range of urban land uses (Bannerman et al. 1993) and a follow-up study (Waschbusch et al. 1999) in two residential sub-watersheds. In the early 1990s the City of Toronto undertook studies to characterize urban stormwater quality by sampling in waterfront storm sewers and combined sewers (Paul Theil Associates and Beak Consultants 1995, Maunder et al. 1995). This was done originally to estimate the relative significance of direct contaminant loadings to Lake Ontario, rather than identify CSAs, but the data still represent the most comprehensive assessment of urban land use effects on urban stormwater quality in Ontario and have since been combined with stormwater hydraulic models to guide decision making regarding implementation of the Wet Weather Flow Master plan (D'Andrea et al. 2004). Recent U.S. data are available through the National Stormwater Quality Database (NSQD) which was initiated in 2001, through a U.S. EPA grant to the University of Alabama and the Center for Watershed to collect and evaluate stormwater data from a portion of the NPDES (National Pollutant Discharge Elimination System) MS4 (municipal separate storm sewer system) stormwater permit holders2. In 2008, the NSQD was updated with additional data with ongoing support from the U.S. EPA. These stormwater quality data and site descriptions were collected and reviewed to describe the characteristics of national stormwater quality, to provide guidance for future sampling needs, and to enhance local stormwater management activities in areas having limited data. Monitoring data have been collected over nearly a ten-year period from more than 200 municipalities throughout the U.S.2.

2 http://bmpdatabase.org/nsqd.html

14

Other examples of published studies characterizing nutrient concentrations in urban stormwater runoff include: (a) prediction of pollution loads from an urban residential catchment Saskatoon (LeBoutillier et al. 2000); (b) stormwater quality measurement to investigate the relationships between water quality and urban form in six Gold Coast, Australia catchments (Goonetilleke et al. 2005); (c) assessment of stormwater from five catchments in the city of Poznań, Poland (Baralkiewicz et al. 2014); (d) assessment of contaminants in road and roof runoff in Handan, China (Wei et al. 2013); and (e) investigation of N and P transport from residential neighborhood lawns in Orange County, California (Toor et al. 2017). It should be noted that these international sources of urban runoff nutrient data often reflect land features that differ considerably from those in Canada and the Great Lakes region of the U.S. and hence they may be of secondary interest. Collectively, these studies provide a representative range of data describing the effects of various types of urban land use on nutrient related water quality. These data can be used to derive export coefficients for use in mass balance calculations (e.g. Hobbie et al. 2017) or hydrologic and watershed models (e.g. LeBoutillier et al. 2000, White et al. 2009, Ye et al. 2012, Niraula et al. 2012, Wellen et al. 2014a, b, c). Such models may be sufficient to inform general stormwater management goals and policies, or they may be used to focus subsequent investigative efforts to refine identification of CSAs for local abatement. Depending on the model study objectives there may be a need for specific local assessment of nutrient concentrations and loads.

2.3 Best Management Practices (BMPs), Low Impact Development (LID), and Infrastructure Improvement

The development of municipal growth plans necessitates prediction and assessment of urbanization and climate change effects on water quality and flows in watersheds as part of managing population growth and infrastructure planning. As noted in the Stormwater Management Planning and Design Manual (OMOE 2003):

Criteria for the protection of water quantity, water quality, habitat, and biota are established to help achieve the goals set for the watershed. Strategies to manage human activities within the watershed are developed to meet protection criteria. A stormwater management strategy may include protection of natural areas, design of communities to reduce stormwater generation, and pollution prevention programs, as well as the stormwater management practices.

Over the past 20 years or so “low impact development (LID)”, “ecologically sustainable growth” and “sustainable drainage systems” have become common themes for municipal planning. These tend to be geared to source controls and small catchments and hence

15

differ from structural end-of-pipe (EOP) controls used to service numerous lots or whole subdivisions (e.g. wet ponds, artificial wetlands) (OMOE 2003). The primary focus tends to be on flooding and flow management related to urbanization, population growth and climate change (e.g. Toronto Water 2017, Alves et al. 2016, Kaspersen and Halsnaes 2017) however benefits to water quality have also been recognized (e.g. IJC 2004, Goonetilleke et al. 2005, Dietz 2007). The International BMP Database3 is a useful source of information but since much of the urban landscape in Ontario was developed prior to the introduction of BMP/LID into urban planning, these older areas will convey higher runoff and nutrient loads than the most recent areas.

A large-scale example of applying an urban and land-cover change transition model (SLEUTH, USGS 2003) to provide a regional predictive assessment of future development and the potential impacts of different regional management scenarios was undertaken in the Baltimore-Washington area of the Chesapeake Bay watershed (Jantz et al. 2004, Duan et al. 2012). Miller and Hutchins (2017) provide a more recent example of large scale projections regarding the impacts of urbanization and climate change on urban flooding and urban water quality in the United Kingdom.

Assessing the effectiveness of low impact development (LID) policy initiatives, BMPs, stewardship projects and infrastructure improvements on nutrient concentrations and loads is an obvious priority if these have been linked to eutrophication issues. There are numerous examples of such assessments. Results from a Canadian Water Network project investigating innovative stormwater management in Canada have been published as a 2009 Special Issue of the Water Quality Research Journal of Canada Vol 44 (1): 1-110.

Clearly, any assessment of BMP effectiveness would best be accomplished with a local survey specifically designed to maximize the likelihood of detecting change attributable to the action(s) of interest and consequently typical studies compare mass balances of inflows and outflows at the assessed facility. While this is the best means of assessing the efficacy of a specific facility, it cannot determine whether the beneficial effect propagates all the way to the receiving waters, or if it is even measurable in the receiving waters. The percentage of implementation of BMP/LID in the catchment is of obvious importance since discernable effects in receiving waters will depend on the cumulative benefits of multiple undertakings (Pennino et al. 2016). In many cases, however, inferences regarding the effectiveness of LID and BMPs have relied upon analysis of existing water quality data to discern long-term concentration and loading trends at tributary mouths or other long-term monitoring stations (e.g. Van Meter and Basu 2017, Meals et al. 2010, Basu et al. 2010, Hirsch et al. 2010, Mohamed 2012, Zhang et al. 2016). This type of after-the-fact analysis can provide useful insights especially in assessing the cumulative impacts of widespread changes in land use or management practices. This approach is not so well suited to evaluating the benefits of

3 http://bmpdatabase.org/publications.html

16

specific local improvements since, notwithstanding issues of statistical power, uncontrolled confounding factors will increase as the distance between the action(s) of interest and the monitoring location increases. This approach can also be problematic in that comprehensive data regarding changes in land uses and management practices are often unavailable and inferring causality without this information can lead to development of model structures that make implicit positive assumptions about the efficacy of existing management practices.

Construction of urban stormwater ponds or wetlands to manage flow and water quality has been a common management practice undertaken in a wide range of jurisdictions (OMOE 2003) and there are numerous studies evaluating their effectiveness. Typically, this has involved comparison of inflow and outflow concentrations and loads at specific locations (e.g. Sychterz et al. 2014, Mulroy 2010, SWAMP 2005, Mayer et al. 1996) and/or the development of sedimentation and retention models to assist with design (e.g. Walker 1987; OMOE 2003). Other stormwater management practices that have been implemented and evaluated include detention tanks, stormwater infiltration systems such as swales or perforated pipe (e.g. SWAMP 2005, Sychterz et al. 2014), bioretention and green roofs (Linden and Stone 2009, Van Seters et al. 2009, Lamera et al. 2014).

Studies into the effect of limiting the application of lawn fertilizers containing phosphorus in Minnesota (Minnesota Department of Agriculture 2007) and Ann Arbor, Michigan (Lehman et al. 2009, 2011) are examples of measures specifically targeting urban NPS P load reductions. Although the Minnesota study was inconclusive, the Ann Arbor project was able to demonstrate P reductions in receiving waters even though the changes could not be solely attributed to the fertilizer ordinance since it was enacted along with a public education campaign designed to reduce discharges into storm drains. Generally, answering the questions regarding the efficacy of specific control measures such as limiting the use of P lawn fertilizers with statistical significance is very difficult and requires large data sets.

3. REVIEW OF MONITORING/MODELING APPROACHES AND ASSESSMENT OF RELEVANT LIMITATIONS OR DEFICIENCIES The fundamental basis for assessing the adequacy of current NPS nutrient load estimation knowledge and practice will be a comparison of study results with the management decision information requirements. It is axiomatic, but worth re-stating, that good monitoring design begins with clear monitoring objectives since these will dictate the data requirements and corresponding sampling approaches. In simple terms then, an assessment of information adequacy will represent a verdict on whether, or how well, study objectives were met. Management information requirements will vary according to the tolerance for uncertainty by those charged with the decision-making responsibility and will, therefore,

17

contain an element of subjectivity. It is also true that management decisions will not depend solely on scientific input. Even quantitative, benefit-cost analyses incorporate socio-economic factors that require the translation of scientifically measured or predicted “effects” into “costs” or “benefits”. Such a translation of scientific observations or cause-effect linkages into a policy or management decision-making process implicitly incorporates value judgements as to what is “acceptable” or “unacceptable”. These elements of subjectivity and the corresponding difficulty in reaching consensus on scientific, social and economic factors when deciding when and how to intervene to alleviate observed or predicted adverse environmental effects are by no means limited to eutrophication issues. There have been various efforts to provide guidance on science-based decision making including identification of guiding principles (Mills et al. 2001) and advocacy for application of a standardized Science Consistency Review (Guldin et al. 2003). A traditional scientific perspective might be summed up as “there are no easy answers” contrasted with a decision-makers' plea “don't bring us problems, bring us solutions”. There is general agreement that uncertainty analysis is an important component of good scientific practice and that uncertainties associated with scientific information should be acknowledged and documented as should the relevant management risks and consequences (e.g. Pappenberger and Beven 2006, Guldin et al. 2003). In practice, however, uncertainty analysis may not be well described and is sometimes absent altogether in reported results from hydrology, hydraulics and water resources related work (Pappenberger and Beven 2006, Wellen et al. 2015). There may be several reasons for this reluctance to acknowledge uncertainty such as concerns that policy makers and the public will not be comfortable with results that include uncertainty distributions, or that uncertainty analysis cannot easily be incorporated into the decision-making process (Pappenberger and Beven 2006). It is also frequently the case that there are simply insufficient data available to permit a rigorous assessment of annual loading estimate accuracy since an ideal assessment would involve a comparison of estimated annual loads versus “true” loads which are usually only available for those relatively few days or episodes that have been sampled rather than the entire year. For sampled days, however, there are several approaches that can be employed to compare model predictions with measured loads (e.g. Kobayashi and Salam 2000, Gauch et al. 2003, Pineiro et al. 2008) so in principle there is always some potential to quantify this relationship. The field of Ecological Risk Assessment (ERA) has well established protocols for evaluating the information needs of decision makers in the face of uncertainty (CCME 1996, U.S. EPA 2001) and these may be helpful in this context. A key feature of an ERA approach to information gathering and decision making is the distinction between “Assessment Endpoints” and “Measurement Endpoints”. Assessment endpoints indicate why measurements are being taken and hence should be established prior to the selection of the corresponding measurement endpoints which are the measurable environmental characteristic of the assessment endpoint (Suter 1991). As noted by Suter (1991) there

18

are many monitoring programs that have identified measurement endpoints, but that have not clearly defined the assessment endpoints. This distinction is important in the current context because it forces monitoring program objectives to be framed first as management questions to be answered that can then be translated into an appropriate environmental measurement. Another feature of an ERA approach is the progression from a Screening Level Assessment, to a Preliminary Quantitative Assessment, and ultimately to a Detailed Quantitative Assessment. This data-based approach to decision making is intended to increase the confidence of risk management decision makers by diminishing uncertainty with each successive assessment. This iterative approach (see Appendix A for schematic) generally incorporates Scientific-Management Decision Points (SMDPs) to document whether a particular level of assessment provides sufficient information to make a decision or whether there is a need to progress to the next level (U.S. EPA 2001). A key feature of this step is that it incorporates “Decision Rules” that have been determined in advance so that there is clarity and consensus on the sufficiency, or insufficiency, of existing information before embarking on more detailed monitoring and modeling. This approach can also address uncertainty by combining “multiple lines of evidence” that collectively reinforce a conclusion despite a lack of certainty within any individual line of evidence. This approach is integral to the Canada-Ontario Decision-Making Framework for Assessment of Great Lakes Contaminated Sediment (Chapman et al. 2007) and is incorporated into the Guidelines for Identifying, Assessing and Managing Contaminated Sediments in Ontario: An Integrated Approach (Fletcher et al. 2008). A SMDP following a screening level analysis that demonstrates negligible risk, even with considerable uncertainty, can support a decision that no further analysis is necessary provided the uncertainty error is translated into conservative assumptions. In the context of managing eutrophication this might incorporate documentation that no HABs have been observed in a receiving water body and that there are no future scenarios where HAB conditions can be predicted to occur. In this hypothetical situation, there would be no need for a detailed, quantitative assessment of all nutrient loading scenarios if a screening level scenario incorporating highly conservative loading assumptions (e.g. using upper limits of estimated nutrient export coefficients) failed to predict receiving water conditions necessary to trigger a HAB. The primary benefit of this kind of systematic SMDP discussion is that it forces an explicit evaluation of outcome risks associated with making a decision in the face of uncertainty. It can accommodate best professional judgement and multiple lines of evidence as well incorporate the concept of “continuous improvement” (engineering) or “adaptive management” (resource management). Such a discussion can and should identify intrinsic uncertainties that cannot be substantially reduced despite the expenditure of additional monitoring and modeling resources to improve data quality. There will be little value in pursuing increasingly sophisticated and resource-intensive nutrient loading or CSA export coefficient estimates that reduce some sources of uncertainty in the presence of other

19

large uncertainties that cannot be reduced. A review of intrinsic uncertainty for NPS nutrient assessment and loading activities by management objective category will help manage the expectations of management decision makers as well as assist those undertaking such assessments maximize the effectiveness of allocated resources. In effect, this analysis replaces “there are no easy answers” with “this is the best you can expect”.

3.1 Loadings from Watersheds

Eutrophication related management decisions associated with nutrient load estimates from watersheds will generally be interested in: (a) identifying the largest contributing watersheds from a mass balance perspective, and (b) flagging watersheds with the highest export coefficients or unit-area loads. The former is an important first step in establishing a P budget that estimates both external sources and internal re-cycling while the latter will be of the greatest importance when pursuing management options for achieving load reduction targets. In both cases it is important to keep in mind that such loading estimates are a means to an end (i.e. accounting for observed excess TP concentrations in a receiving water body) rather than an end in itself. There are numerous potential sources of uncertainty and bias for watershed P loading estimates that can be reduced through an increased sampling effort and these have been well described (e.g. Richards 1998, Fox et al. 2005, Zamyadi et al. 2007, Dressing et al. 2016, Lee et al. 2017). A common problem confronting those attempting to use long-term, historical data sources such as the Ontario Provincial Water Quality Monitoring Network (PWQMN) is the low sampling frequency which is generally limited to collection of monthly samples or less and is generally biased towards low-flow events. Although when pooled these data can provide insights into long term change or regional differences in water quality, they are not well suited to estimation of loads due to their failure to capture the significant influence of low-frequency, high-flow events. Some of the most obvious possible monitoring approaches to reducing uncertainty and bias in load estimation include: increasing the frequency of event-based sampling to characterize more wet weather events; sampling through individual wet weather events to capture hysteresis effects; derivation of event-mean concentrations using flow-weighted rather than time-weighted samples; and use of hourly flows rather than daily averages. The value of the corresponding improvement in loading estimate accuracy will, however, depend on the management decisions that they are intended to inform. This would be an important part of a SMDP discussion. Estimates of NPS P loads have been estimated and summarized for Lake Erie (Dolan and Chapra 2012, Maccoux et al. 2016), Lake Simcoe (Winter et al. 2007, Lake Simcoe Science Advisory Committee 2012) and Hamilton Harbour (Long et al. 2015). The following discussion focuses on the Lake Erie and Hamilton Harbour cases which have incorporated receiving water eutrophication modeling (Annex 4 Objectives and Targets Task Team 2015, Gudimov et al. 2010, Ramin et al. 2011) and event-based, high frequency tributary

20

monitoring (Maccoux et al. 2016, Long et al. 2015). These examples have been well documented and a review of management objectives and study findings for these areas can be used to illustrate the potential benefits of a SMDP approach to making decisions regarding the allocation of future scientific resources. Lake Erie

In Lake Erie, initial P load reduction targets were established in 1978 to drive the implementation of basin wide secondary treatment at municipal WWTPs. A coarse-scale, whole-lake model was employed that functionally back-calculated loading targets from desired P concentrations assuming the lake was a well-mixed reactor (IJC 1978). The statistically based Vollenweider model (relating P loads to P concentrations and P concentrations to chlorophyll concentrations) was modified by Chapra to include a dynamic P loss component and by Ditoro to predict secondary parameters of interest such as chlorophyll and dissolved oxygen (Vallentyne and Thomas 1978). The resulting Lake Erie target of 11,000 metric tonnes per annum (MTA) was initially achieved as early as 1981 following implementation of secondary treatment at municipal WWTPs within the drainage basin and, over the period 1982–2002, the target was achieved roughly half the time (Dolan and Chapra 2012). Their analysis showed that the variability in the load resulted from basin hydrology with loads exceeding the target occurring in years with relatively high precipitation and runoff. More recently, estimation of 2003 to 2013 annual total P loads to Lake Erie by Maccoux et al. (2016) also showed significant fluctuations over this period with results varying from a low of 5,839 MTA in 2010 to a high of 11,946 MTA in 2011, with an average of 9,125 MTA. This same analysis also determined that behind this high interannual variability there was no significant positive change over the 26 years prior to 2013. As noted by Dolan and Chapra (2012), precipitation and flow were the primary factors in this extreme year-on-year variability. Precipitation totals for London, Ontario4 and Cleveland, Ohio5 were 931 mm and 901mm respectively in 2010 compared with 1,165 mm and 1,659 in 2011. The average daily flow for the Maumee River in Ohio6 (the largest river flowing into the Lake Erie Western Basin) went from 154 m3 s-1 in 2010 to 211 m3 s-1 in 2011 with a corresponding increase from 50 m3 s-1 to 106 m3 s-1 for the Thames River (at Thamesville) in Ontario.7 The original 1978 targets and models were re-evaluated and updated following the re-emergence of HABs in the Western Basin in the mid 1990s and continued hypoxia in the Central Basin. This updated approach to target setting was necessitated by the failure of the original models to predict recent conditions in the Western and Central Basins even

4

The Weather Network https://www.theweathernetwork.com/weather/historical-weather/caon0383 5

Current Results weather and science facts https://www.currentresults.com/Yearly-

Weather/USA/OH/Cleveland/ recent-annual-cleveland-temperature-precipitation.php 6

USGS Water Surface-Water Annual Statistics for the Nation 7

Water Survey of Canada - HYDAT

21

when the original 11,000 MTA load target was met and a recognition that nearshore processes were not well represented by whole-lake conditions particularly following colonization by dreissenid mussels and a recent relative increase in soluble P loads. Several updated models were applied to predict the production of Western Basin cyanobacteria biomass as well as the areal extent of Central Basin hypoxia in response to P loads (Annex 4 Objectives and Targets Task Team 2015).

Figure 1. (from Annex 4 Objectives and Targets Task Team 2015) Cyanobacteria bloom size (peak 30-day average biomass) predicted by the Stumpf et al. (2012) model in the Western Basin as a function of spring Maumee River TP load. The solid line represents mean predictions, while dashed lines represent 95% prediction intervals. The green horizontal dotted line indicates the threshold for “severe” blooms.

Figure 2. (from Annex 4 Objectives and Targets Task Team 2015) August-September average extent of the hypoxic area predicted by different models in the Central Basin as a function of annual TP loads to the Western and Central Basin. The green horizontal dotted line indicates a suggested threshold of 2000 km2. (Nine Box model not used for target setting).

22

It is worth noting that the 95% prediction ranges for total phosphorus (TP) loads estimates corresponding to the desired thresholds of both cyanobacteria biomass and hypoxia extent are +/- 30% to 40% (see Figure 1, 2). This is not an indictment of the model performance, which is entirely representative of this type of analysis, but it is worth keeping in mind when considering the need for additional precision and accuracy in loading estimates. The loading target update review also recognized the need for an additional target related to nuisance benthic algae problems in the Eastern Basin, but the Annex 4 Task Team observed that the current understanding of processes determining the growth and transport of Cladophora was not yet sufficient to accomplish this. The revised load reduction targets derived from the cyanobacteria biomass and hypoxia extent models were framed separately for Western Basin HABs and Central Basin hypoxia. For the former, a 40% reduction from the 2008 spring loads of TP and DRP loads from the Maumee River in Ohio as well as other priority Western Basin tributaries was recommended. For the latter, an annual target load of 6,000 MTA of TP to the Central Basin (a 40% reduction in loadings from 2008 levels) was recommended. In presenting the revised loading targets the Annex 4 Task Team encouraged adoption of a carefully-designed adaptive management process to track the response of the system, evaluate the effectiveness of management efforts, and update management recommendations as more was learned about the processes underlying the system response (Annex 4 Objectives and Targets Task Team 2015). The Annex 4 Task Team also observed that this would require a monitoring program capable of tracking loading trends over time but specifically recommended that flow-weighted mean concentrations 8 at tributary mouths should be used as a benchmark to track progress in load reductions. These targets and recommendations provide a clear, primary objective (assessment endpoint) for tributary NPS nutrient assessment and loading studies. Dolan and Chapra (2012) and Maccoux et al. (2016) acknowledge that whereas calculation of statistical error (i.e. the variance of the input data and the associated error in calculating the total load) is relatively straightforward, the overall accuracy of the estimates is more difficult to determine as it is related mainly to the adequacy of the underlying sampling programs (Dolan et al. 1981). Maccoux et al. (2016) do not include any explicit analysis of the model goodness-of-fit for the various methods used at different tributaries so it is not possible to compare the relative error of the underlying models used with each other or with the interannual variability in flow. However, Lee et al. (2016) undertook a comparison of loading estimate methodologies using random sub-samples drawn from temporally-dense data sets from 14 watersheds across the U.S. and demonstrated that for the Beale Ratio estimator method (see Dolan et al. 1981, Richards 1998 for summary of the methodology) used by Maccoux et al. (2016) at the sparsely sampled U.S. and Canadian tributaries, the relative error in predicted loads was 54%. The high frequency sampling (n > 365) undertaken by Heidelberg University at key large U.S. tributaries would, of course, have achieved far better results than this.

8

This is dimensionally equivalent to the load:flow ratio and hence normalizes loads for variations in flow

23

Maccoux et al. (2016) compiled and updated annual loading estimates to Lake Erie for TP up to 2013 incorporating updated data previously reported by Dolan and Chapra (2012), Baker et al. (2014) and Dolan and McGunagle (2005). They also reported recent (2009-2013) soluble reactive P (SRP) for the first time. The information has been interpreted with respect to the distribution of loads by basin and by source category, and the authors recommend that this form the basis for more detailed analysis and program and policy development most notably development of U.S. and Canadian Domestic Action Plans to make progress toward the Annex 4 Task Team load reduction targets. In addition to the high interannual variability previously noted, Maccoux et al. (2016) note that long-term (1967–2013) declining trends for TP loads are driven by both early and more recent declines in point source loadings and that recent data show no temporal trends in SRP loads. They note an increase in nonpoint source loads for TP since 1999 driven by increasing tributary flows. They calculate flow-weighted mean concentrations (FWMCs) and observe no significant trend over the same period with most of the individual tributaries (24/28) showing no recent change in TP FWMCs. The authors acknowledge that additional analysis will need to be conducted involving additional monitoring and modeling. They flag a need to account for usage of phosphorus by in-stream processes that may occur downstream of monitoring locations and for possible sedimentation or other in-stream processes in the case of indirect point sources discharges when apportioning the total loads into source categories and calculating unit area loads. They observe that their annual load estimates do not consider that discharges of nonpoint sources are delivered episodically following runoff events while point source discharges are relatively continuous and that the annual estimates do not lend themselves readily to calculations of seasonal loads that are required for tracking of progress with respect to the Western Basin spring TP and SRP loads. Their recommendations emphasize the need for improved monitoring at all intermediate and large Lake Erie tributaries with sufficient frequency and duration to provide the statistical power to calculate loads and detect trends. They acknowledge that their reliance on infrequently collected data from tributary water quality programs for some of the individual States and the Province of Ontario has been insufficient for the robust computation of loads and has likely led to an underestimate of certain U.S. tributaries and most of the Canadian tributary nutrient contributions to Lake Erie (Maccoux et al. 2016). As previously noted, the predominant sources of bias result from a lack of high-frequency, event-based water quality sampling data leading to under estimates of actual loads (Richards 1998, Fox et al. 2005, Dressing et al. 2016). Although Maccoux et al. (2016) do not quantify this potential underestimation, Richards (1998) has demonstrated the potential for the stratified Beale Ratio estimator approach used by Maccoux et al. (2016) to underestimate loads by more than 60% when thinning daily data to weekly data. This is consistent with the comparison of methodologies undertaken by Lee et al. (2016) in which application of the Beale Ratio estimator to low flow data resulted in

24

underestimation of 40%. This extreme low bias is even more likely to be the case for the quasi-monthly data available from the Ontario Provincial Water Quality Monitoring Network (PWQMN). While this concern justifiably underpins the recommendation for high frequency sampling of intermediate and large tributaries to track annual loads, in the case of Lake Erie Western and Central Basin the need to address the underestimation of annual loads from Canadian tributaries must be considered in the context of their relative contribution, particularly when it comes to urban NPS inputs. Maccoux et al. (2016) estimate that Canadian monitored and unmonitored tributaries account for only about 14% of the total P load to Lake Erie over the period 2003 to 2013 and even if this errs substantially on the low side, as previously noted, urban land use accounts for less than 4% of the Canadian drainage area (OMNRF 2015). It should also be noted that the largest Canadian tributary input to western and central Lake Erie is the Thames River where event-based sampling has already been undertaken by ECCC and where this bias is not likely to be as pronounced. Prior to allocating additional resources to high frequency sampling at more tributaries it would be worth discussing exactly how the management objectives and actions based on annual loading estimates might change if more accurate results were available for urban portions of Canadian tributaries. The high degree of interannual variability in flow, as well as the intrinsic uncertainty associated with receiving water cyanobacteria biomass and hypoxia extent models, need to be considered as part of this discussion which would function as Scientific-Management Decision Point (SMDP). Ideally, this would be accompanied by the formulation of a Decision Rule that would justify either the sufficiency of existing information or the need for additional, specific monitoring capable of providing more reliable information. Interestingly, Maccoux et al. (2016) also flag the need for much additional work to understand how best to accomplish loading reductions. More specifically, the need to investigate the relationships between various land uses and water quality, the relative contributions of various land uses to nutrient loadings, and practices that will achieve nutrient reductions. These recommendations align perfectly with the second and third study categories mentioned in the previous section of this whitepaper. If the development of a Canadian Action Plan to achieve load reductions is a high priority for management decision-makers, it would be well worth discussing the relative merits of allocating resources to improving estimates of annual mass loadings from large and intermediate tributaries versus increasing support for investigation of CSAs and evaluating the effectiveness of BMPs to achieve load reductions. As noted, interannual variability will remain high due to varying precipitation and tracking FWMCs at the mouths of large tributaries cannot be an effective means of identifying CSAs or understanding the relationship between land use and water quality due to the variety of land uses that are inevitable present in large watersheds. From this perspective, data

25

gathering in small watersheds with relatively homogenous land uses would provide an easier means of quantifying the relationship between land use and water quality. Such a discussion would also represent a SMDP regarding the sufficiency of existing information to drive management decision making with respect to identification of CSAs and the effectiveness of management practices to achieve loadings. It seems probable that this SMDP would strongly support the recommendations of Maccoux et al. (2016) for additional work in these areas. Notwithstanding the apparent need for substantially more study within watersheds, the fact that urban land uses represent less than 4% of the Lake Erie drainage basin on the Canadian side (OMNRF 2015) means that their corresponding NPS loads of total P will be negligible when considering assessment endpoints on the scale of the Lake Erie Western and Central Basins. The SMDP and corresponding Decision Rule will, therefore, most probably suggest a focus on rural and agricultural CSAs and BMPs for achieving Canadian Lake Erie “Action Plan” management objectives at this scale. This does not, however, preclude the formulation of locally-based management objectives within Canadian watersheds that may indeed flag the potential for improved nutrient-related water quality affected by urban landscapes. These could either be used as supplementary assessment endpoints for Canada's Annex 4 commitments or they could be undertaken as part of an entirely separate process in partnership with municipalities or Conservation Authorities. Hamilton Harbour

A statistical examination of the correlation between annual P loads and receiving water P concentrations may be sufficient for preliminary load reduction target setting. In this case the primary data quality requirement for loading estimates is that they be derived in a consistent manner since the correlation depends on the relative change in loading from year to year rather than its accuracy. This approach also requires that there be a substantial range in annual loads corresponding to a wide range of receiving water concentrations to establish significant correlation. The Hamilton Harbour RAP used such an approach to set initial water quality and loading targets of 34 ug l-1 and 330 kg d-1 (O'Connor 2003) however, once loads levelled off following optimization upgrades to municipal WWTP inputs the relative error in loading corresponding to the desired final water quality target was too great to support a reliable regression-based prediction of the final loading target. The original “final” target for total P of 17 ug l-1 for 13 consecutive weeks at the centre station corresponding to a total load of 142 kg d-1 was critically reviewed and updated (Labencki 2010) and expressed probabilistically to better accommodate natural variability in water quality resulting from natural factors such as the exchange with Lake Ontario and variation in tributary loads attributable to wet and dry years.9 The resulting final water 9 Concerns about the influence of colonization by dreissenid mussels on water clarity and near-surface TP

concentrations were determined not to be a significant factor within the harbour due to its morphology and substrate which is not conducive to those densities observed in open Lake Ontario.

26

quality target for total P was for 15 of 17 epilimnetic integrated samples analyzed weekly at the centre station from June to September to meet or be better than 20 ug l-1 and the original final loading target of 142 kg d-1 was flagged for ongoing assessment by more closely investigating the role of nonpoint sources from tributaries. A more deterministic, process-based eutrophication modelling approach was developed to evaluate the potential to achieve these final receiving water quality and fine tune the related loading targets (Gudimov et al. 2010). This biogeochemical model attempted to describe the different ecological mechanisms modulating the eutrophication problems in Hamilton Harbour and was used to evaluate the critical nutrient loads in the Harbour based on acceptable probabilities of compliance with different water quality criteria (e.g., chlorophyll a, total phosphorus). This model suggested that the water quality goals for TP and chlorophyll a would likely be met when the Hamilton Harbour RAP phosphorus loading target at the level of 142 kg d-1 was achieved. Additional analysis incorporating a Bayesian calibration of several process-based models was subsequently undertaken (Ramin et al. 2011) and this confirmed the potential for the total P target to be met but cautioned that the summer chlorophyll a target of 10 ug l-1 may not be consistently achieved. This modeling included goodness-of-fit error analysis of model predictions that demonstrated a relative error of approximately 10% for epilimnetic total P but more than 30% for phosphate and greater than 40% for total zooplankton biomass. It concluded that the anticipated structural shifts of the zooplankton community would determine the restoration rate of the Harbour but noted that the complexity of this process made this hard to predict and that, in addition to the traditional “bottom up” approach of controlling external nutrient inputs, “top down” impacts of changes in food web structure should also be considered. These studies supported the need for a more quantitative assessment of nutrient loads from tributaries to provide a more realistic assessment of future conditions following additional, significant load reductions from the Hamilton Woodward Avenue WWTP. Additional event-based tributary sampling was undertaken over the period July 2010 to May 2012 at the four major inputs to the harbour: Red Hill Creek, Indian Creek, Grindstone Creek and the Desjardins Canal (which drains the Cootes Paradise marsh which receives flow from the Dundas WWTP, Spencer Creek, and several small local tributaries). Results were used to estimate annual P loads based on a highly statistically significant log flow versus log concentration regression model (r2 > 0.60). Results showed that previous tributary loading estimates based on limited sampling in the late 1980s had under-predicted total P loads associated with high-flow events and over-predicted loads associated with baseflow (Long et al. 2015). The analysis also showed considerable interannual variability in estimated tributary P loads associated with wet versus dry years ranging from a high of 98.5 kg d-1 in 2011 to a low of 30.2 kg d-1 in 2012. The Long et al. (2015) analysis also included a critical evaluation of the statistical model used. This is very helpful and is not always a feature of loading estimate studies. The error analysis associated with the regression model showed excellent agreement between

27

predicted and measured loads for the 80 to 90 sampled days over the two-year sampling period (r2 > 0.80). It also indicated a RMSE relative error of from 53% to 88% depending on the tributary when comparing predicted loads with measured loads. This underlying error in the regression model appears high but is representative of modelled loads using regression-based methodologies (Richards 1998). As noted, this type of error analysis is frequently not included as part of loading estimate studies which generally attach confidence intervals for estimated annual average loads based on the variability in predicted daily or monthly loads rather than the underlying statistical model. The three-fold variation in estimated tributary P loads solely due to variation in flow, as well as the relative error associated in soluble P and zooplankton biomass predicted by the receiving eutrophication model needs to be kept in mind when considering the adequacy of these tributary loading estimates. Long et al. (2015) note that the use of the flow versus concentration regression-based method allows the estimation of loads where only flow data are available and at finer time scales than annual or seasonal estimates. These will allow refinement of the receiving water eutrophication model to consider the response of the harbour at time scales that may be more appropriate. In this case the current loading estimates from the heavily urbanized watersheds were deemed entirely sufficient to support a management decision to allocate future resources to identifying and ameliorating sources within watersheds through additional distributed source modeling and related water quality monitoring. The need for additional, high-frequency event-based sampling to estimate mass loads from tributaries into the harbour was flagged as being unnecessary until after implementation of remedial actions and BMPs within the watersheds. Although no formal Decision Rule was applied in this case, the outcome demonstrated that the uncertainty associated with nutrient loading estimates from tributaries with a high proportion of urban land use was acceptable to management decision makers.

3.2 Critical Source Areas (CSAs) and Land Use Effects

As previously noted, assessment of land use effects and identification of CSAs are conceptually similar but differ in scale. For purposes of this discussion it will be useful to consider CSAs as functioning at a scale where specific local remedial actions, BMPs or policies might be implemented. Land use effect assessments generally operate at a coarser scale and will be of value in determining and validating nutrient export coefficients to extrapolate results from monitored areas to unmonitored areas when undertaking source modeling to estimate watershed and sub-watershed loads. They may also represent a step towards identification of CSAs by helping focus local scale investigations.

28

Land Use Effects

Monitoring and loading studies that provide insights into the effects of urban land use on water quality have been undertaken in Toronto (Boyd et al. 1999) and Hamilton (Long et. al. 2014, 2015). The Toronto tributary monitoring project undertook sampling in six tributaries including the Don River and Highland Creek with watershed areas of 376 km-2 and 110 km-2 having 67% and 81% Industrial, Commercial, Institutional and Residential land uses respectively. The Don River was sampled over the period November 1990 to March 1992 and Highland Creek from July 1991 to September 1992. The corresponding dry weather and wet weather P concentrations for the Don River were 0.160 mg l-1 +/- 0.020 mg l-1 (95% confidence interval) and 0.280 mg l-1 +/- 0.080 mg l-1 (the influence of the North Toronto WWTP discharge into the Don River is highly evident in the dry weather results). Results for Highland Creek were 0.020 mg l-1 +/-0.010 mg l-1 and 0.180 mg l-1 +/- 0.170 mg l-1 for dry weather and wet weather respectively. The study used the stratified Beale Ratio Estimator method (see Dolan et al. 1981, Richards 1998 for a summary of the methodology) to estimate dry weather, wet weather, and annual daily average TP loads. Corresponding estimates of the error associated with calculated loads were also expressed as Root Mean Square Error (RMSE). The report cautioned that extreme conditions such as large storms or lower than average spring flows resulting from below average snow accumulation (as was experienced in 1992) will have influenced the estimates of contaminant mass loadings and that results should be considered representative of relative loadings among tributaries rather than long-term average conditions. The Don River calculations comprised seven dry weather samples and 31 wet weather samples and yielded a unit-area TP load of ~100 kg km-2 yr-1 with a relative RMSE error of approximately 9%. The Highland Creek result was ~150 kg km-2 yr-1 with a relative RMSE error of about 21% based on five dry weather samples and 12 wet weather samples. These results were central to the subsequent modeling undertaken as part of developing Toronto’s Wet Weather Flow Master Plan (D’Andrea et al. 2004). The Hamilton Harbour tributary study included sampling at Red Hill Creek and Indian Creek with watershed areas of 65 km-2 and 23 km-2 having 81% and 72% urban and urban green space land uses respectively. Over the period July 2010 to May 2012, 21 and 23 baseflow samples were collected at Red Hill Creek and Indian Creek along with 71 and 74 wet weather samples. As with the Toronto study, the sampling period covered a range of winter conditions with a cold, snowy winter in 2010/2011 and an extremely warm, relatively snow-free winter in 2011/2012. Dry weather median TP concentrations were 0.024 mg l-1 for Red Hill Creek and 0.033 mg l-1 for Indian Creek with corresponding wet weather median concentrations of 0.134 mg l-1 and 0.126 mg l-1. As previously described, this study used a log flow - log TP concentration regression method to estimate concentrations for un-sampled days based on flow and these were then summed to produce seasonal and annual total and daily average loads. The RMSE error associated with the regression-based loading methodology was higher than for the

29

Beale Ratio Estimator approach used in Toronto which is consistent with observations by Lee et al. (2016) and Richards (1998) for calculations that incorporate high flow sampling. The estimated TP unit-area load for Red Hill Creek averaged over 2010 and 2011 was 126 kg km-2 yr-1 with the corresponding result for Indian Creek being 113 kg km-2 yr-1. If these results for relatively highly urbanized watersheds in Toronto and Hamilton are normalized to account for variability in the relative proportions of urban land uses, the TP unit-area load results converge somewhat as follows: Don River 149 kg km-2 yr-1, Highland Creek 185 kg km-2 yr-1, Red Hill Creek 156 kg km-2 yr-1 and Indian Creek 157 kg km-2 yr-1. It is interesting, if challenging, to compare these results for southern Ontario with those from other studies given the differing approaches to characterizing urban land uses. A common metric is to use GIS mapping tools to estimate the relative proportion of “impervious” land use and compare this with other categories such as “agricultural” and “forest. Literature surveys of studies deriving export coefficients for various land uses such as those by Reckhow et al. (1980), Rast and Lee (1983), McFarland and Hauck (2001), Lin et al. (2004) and Jeje (2006) summarize a range of urban land use categories including “urban”, “mostly urban”, “urban residential”, “urban residential – low density”, “urban residential – high density”, “urban developed”, “urban commercial”, “urban commercial – high density”, “urban average”, “lawns/golf courses”, “roads”, “logging roads”, “stormwater”, “stormwater commercial” and “stormwater residential”. To further complicate matters, GIS analysis of imagery uses categories such as “transportation”, “built up area impervious” and “built up area pervious” (OMNRF 2015). Not surprisingly studies assessing nutrient loads from this wide spectrum of urban land uses have yielded a considerable range of TP export coefficients from as low as 19 kg km-2 yr-1 for “lawns” where infiltration can occur to 350 kg km-2 yr-1 for “urban commercial” and “roads” where the high proportion of impervious surface results in much greater runoff. These ranges will reflect the variation in the available contribution of TP from the varying land uses but will also incorporate variability from controllable factors such as survey design and uncontrollable factors such as weather conditions. Survey design would include such factors as the type of sampling (e.g. grab sampling, time-integrated sampling, flow-weighted sampling) and the timing and duration of sampling (event sampling through intense summer storms or spring freshet). Weather conditions could result in variation of event and cumulative precipitation as well as the differing responses associated with pervious land uses (lawns, golf courses) depending on antecedent moisture conditions. The use of average export coefficients can at least partially resolve many of these issues by producing loading estimates that represent “typical” weather conditions over a range of urban land use types and average results for larger urban areas with mixed sub-categories do indeed converge in the range of 100 to 200 kg km-2 yr-1. Estimates of urban NPS P loads for Lake Simcoe, for example, used values of 132 kg km-2 yr-1 for residential areas and 182 kg km-2 yr-1 for commercial areas (Winter et al. 2007) and these values are consistent with those generated by the Toronto and Hamilton tributary assessments summarized above.

30

If the primary objective of a monitoring study is to estimate annual average P loads from relatively large urban areas with mixed urban land uses, then a SMDP discussion might well determine that a screening level assessment based on existing export coefficient data would be sufficient since the probable outcome for an average water year in terms of precipitation and streamflow will deliver results of between 100 to 200 kg km-2 yr-1 depending on the proportion of other land uses in the watershed. The fact that studies undertaken in Toronto and Hamilton area tributaries approximately 20 years apart yielded comparable results that were also consistent with literature values suggests that the findings are genuinely typical and could be applied to a screening level assessment in any similar context. This SMDP discussion could determine that the allocation of future monitoring and modeling resources might be more effectively used to narrow the search for CSAs within the urban landscape that are delivering atypically elevated unit-area loads such as the urbanizing fringe where a high proportion of the land is under construction. In watersheds where remedial measures and BMPs have already been implemented, the SMDP discussion could determine that the most worthwhile use of resources would be to design a monitoring study with the specific objective of evaluating their effectiveness. CSA Identification

Unlike assessments of larger urban areas where the effects of different subcategories of urban land uses will usually average out to generate similar export coefficients, identification of CSAs will be undertaken at finer scales and will be much more likely to yield highly variable and event specific results. The Wisconsin Department of Natural Resources (WDNR) and the U.S. Geological Survey (USGS) undertook a study in 1991 to characterize nutrient and sediment contributions from source areas located in residential, commercial and industrial land uses in Madison Wisconsin across two small urban residential drainage basins (Bannerman et al. 1993). Assessed source areas included streets, lawns, roofs, driveways, sidewalks and parking lots. The study objective was to identify CSAs to guide targeted and more effective BMPs for solids, TP and metals. Approximately seven runoff samples were collected at more than 40 sampling locations yielding such highly variable results that geometric mean concentrations were calculated for each source category. These were combined with an urban nonpoint-source model (SLAMM, Pitt and Voorhees 1989) to estimate loads and these were compared with samples from storm sewers. The study found a wide range in the percentage of the contaminant loads contributed by each source area for each land use category. Runoff from streets in the residential land use usually had the largest contaminant loads but TP loads differed somewhat from this trend because runoff from lawns and driveways had loads similar to runoff from the collector streets. Although runoff volumes from lawns were small, phosphorus loads were relatively large due to higher concentrations. Runoff from residential roofs had the smallest contaminant loads, although the percentage contribution of metals was similar to that from lawn runoff.

31

The authors defined a CSA by totalling the fewest source areas that collectively contributed > 75% of the contaminant load from a land use and found that: streets were generally a CSA in every land use; the majority of runoff loads for many contaminants were from streets in residential and commercial land uses; parking lots were another critical source area for commercial and industrial land uses; and phosphorus loads in runoff from lawns can become critical for significant runoff events. Although by this definition paved areas are flagged as a CSA, there are limited management options beyond those associated with street-cleaning and catch basin maintenance which would typically be part of a general BMP approach. In this context, the CSA concept may prove more useful when examining the potential to prevent drainage from adjacent, unpaved areas from reaching paved surfaces as part of a larger scale assessment that includes both. A significant finding from Bannerman et al. (1993) was that targeting BMPs to 14% of the residential area and 40% of the industrial area could reduce contaminant loads by up to 75%. Consideration of the various sources of error found that the sum of the modeled source-area loads was always larger than the loads measured at storm sewer outfalls. This was attributed to source-area samplers collecting primarily first-flush samples with larger concentrations than outfall composite samples. The authors considered that although the method over-estimated loads it met the objectives of determining the relative significance of various critical source areas and was sufficient to make meaningful management recommendations. The study acknowledged the need to use geometric means due to the high degree of variability in sampled results but did not provide any measure of variance or central tendency (presumably due to the low sampling frequency), so no statistical significance could be attached to the comparisons among source areas or land uses. A more intensive follow up study focusing on suspended solids and phosphorus and incorporating a much higher sampling frequency was undertaken in the Madison Wisconsin area by WDNR and USGS in 1994 and 1995 to update the 1991 work (Waschbusch et al. 1999). Phosphorus data were collected from five source areas-streets, lawns, roofs, driveways, and parking lots within two drainage basins (Monroe Basin and Harper Basin) from urban residential and commercial areas to estimate loads from each source area. The resulting data were used to calibrate an updated version of the SLAMM source model (version 6.3) to determine which source areas were contributing the most phosphorus within these basins. A total of 25 runoff events were monitored at each basin. Runoff samples were collected from May to November of 1994 in the Monroe basin and from June to November of 1995 for the Harper Basin. A third basin (Lakeland) was also monitored for lawn runoff in 1995. Sample results appeared to have a log-normal distribution and consequently were used to compute geometric means with corresponding coefficients of variation.

32

Results between basins and source areas were highly variable. For lawn runoff, the difference between geometric mean phosphorus concentrations of 0.790 mg l-1 (CoV 62%) from 1994 in the Monroe Basin to 1.610 mg l-1 (CoV 112%) from 1995 in the Harper Basin was greater than a factor of 2 although data collected from the Lakeland Basin in 1995 were similar to the Harper Basin. Lawn runoff had the highest total and dissolved phosphorus concentrations in both the Monroe and Harper Basins and patterns in geometric mean concentrations between source areas within each basin were similar although their magnitudes were very different. The geometric mean concentration of 0.400 mg l-1 phosphorus (CoV 124%) for low-traffic streets in the Monroe Basin was almost twice the geometric mean of 0.240 mg l-1 (CoV 75%) observed at the Harper Basin. Some of the same lawns in the Monroe Basin were monitored for phosphorus concentrations in the previous 1991 Madison study (Bannerman et al. 1993). Both the dissolved and total phosphorus geometric means calculated for that study were more than three times higher than those in 1994 so not only did phosphorus concentrations vary highly between the Monroe and Harper Basins, the Monroe results did not agree well with previous studies (although this may have been an artifact of the low sampling frequency in 1991). The geometric mean of the combined data collected at Monroe and Harper Basins was used for the SLAMM modeling phase of the study to provide a more general characterization of source areas. Results of the modeling showed that streets contributed most of the suspended-solids loads at both Monroe and Harper Basins, generating 81% and 73%, respectively with lawns contributing slightly more than 10% of the solids loads at both basins. For phosphorus, however, lawns in the Harper Basin generated more than two-thirds of the phosphorus loads, whereas phosphorus in the Monroe Basin was more evenly distributed between lawns and streets. These differing load distributions resulted from much higher phosphorus concentrations, especially for lawns, in the Harper Basin. The results illustrate several important points for consideration when undertaking sampling to identify CSAs for phosphorus in urban landscapes. As noted by White et al. (2009) it is much easier to identify CSAs than to evaluate their contribution to the total pollutant load. The collection of water quality data at a sufficient resolution and scale is problematic since even in small watersheds the level of instrumentation and monitoring duration necessary to quantify pollutant loads from CSAs will be extremely resource intensive. Although it will generally be prohibitively expensive to generate this type of fine scale data over an entire watershed, this will not generally be the point since the extremely high degree of variability in results such as those from Waschbusch et al. (1999) demonstrates that monitoring at this relatively fine scale is not well suited to estimation of larger scale loads without pooling of results. It can, however, successfully discriminate relative differences among sources and it is this ability to assess relative differences between such sources as low-traffic roads, high-traffic roads, lawns, roofs, driveways and sidewalks that will be of the greatest value from a management perspective since this can support decisions for targeted BMPs or policies.

33

With this in mind, an important consideration for CSA related sampling and modeling will be to ensure that identical sampling methods are employed across all types of source area so that the relative differences in water quality and estimated loads can be ascribed to differences among source areas rather than differences in sampling. A SMDP discussion following examination of literature data such as those from Madison Wisconsin might need to consider whether the observed relative differences among source areas in other jurisdictions are sufficiently consistent and robust to proceed with implementing management strategies or policies or whether these findings would first require local validation. Such a discussion might also determine that allocating monitoring and modeling resources to identify streets as CSAs would be of limited value given the limited management options other than normal cleaning and catch basin maintenance.

3.3 Best Management Practices (BMPs), Low Impact Development (LID), and Infrastructure Improvement

Assessing the effectiveness of existing remedial measures and BMPs will clearly be of considerable interest to management decision makers both to confirm that the past allocation of resources was successful and worthwhile and to provide insights into the most effective strategies for future consideration. Such assessments will also support the prediction and comparison of potential BMPs and policies for new development. As previously described, there have been many assessments of Low Impact Development (LID), BMPs and end-of-pipe (EOP) controls. The Stormwater Management Planning and Design Manual (OMOE 2003) used modeling as a design tool for SWM ponds to achieve various levels of removal efficiencies and these were supported by the SWAMP program which evaluated nine stormwater management facilities in the Greater Toronto Area between 1995 and 2002 (SWAMP 2005). The Sustainable Technologies Evaluation Program (STEP) water component is an ongoing partnership between Toronto and Region Conservation, Credit Valley Conservation and Lake Simcoe Region Conservation Authority that seeks to support broader implementation of sustainable technologies and practices. Studies have been undertaken to evaluate the effectiveness of engineered infrastructure for stormwater management such as retention ponds/wetlands, bioretention, detention tanks, green roofs, stormwater infiltration systems (e.g. swale/perforated pipe and permeable pavement), specific facilities for construction site water clarification and erosion control as well as practices such as rainwater harvesting and residential stormwater landscaping (Pennino et al. 2016). Although these programs are designed to report the efficacy of various BMP and LID measures (with some uncertainties), it is generally unrealistic to extrapolate data from small, single facilities to an entire catchment. As mentioned previously, until BMP and LID measures have been implemented over a substantial proportion of a catchment area the cumulative effects will be insufficient to discern positive change against a highly variable (noisy) set of background conditions except in the immediate vicinity of a particular facility. In principle it is possible to undertake a statistical power analysis to estimate the

34

sampling effort required to detect the change associated with a range of BMP and LID uptake within a specified catchment. A similar analysis for Lake Erie (Betanzo et al. 2015) suggests that even if the degree of change were large (i.e. 40% reduction in TP load) ten years of monthly sampling data would be required to detect this change in small or large watersheds. For a more realistic TP load reduction of 10% their case study found that more than 40 years of monthly TP data would be needed to detect a change at a given monitoring site with statistical significance because the natural variation that occurs in streamflow and water quality from year to year obscures this small magnitude of change (Betanzo et al. 2015). This somewhat chastening conclusion should not be considered a reason for inaction, but rather a reminder that it is unrealistic to expect the benefits of BMP and LID measures to be easily observed at a watershed scale except over the long term following improvements over a sizable proportion of the observed catchment. Success will need to be measured on a local scale. Typically, evaluation of end-of-pipe (EOP) retention systems such as ponds, wetlands and tanks designed to service multiple lots or entire subdivisions involves comparison of influent and effluent streams as well as analysis of retained sediment. The heavy influence of influent concentrations on removal efficiencies has led several researchers to caution against using percent removals to characterize BMP performance since they can be mischaracterized as ineffective simply because cleaner influent has resulted in low percent removal. Effluent concentrations provide less biased and more direct characterization of BMP effectiveness (SWAMP 2005). It should be noted that in most cases stormwater management (SWM) facilities and infrastructure are designed to manage flow as well as to provide some level of water quality treatment for a wide range of contaminants associated with suspended solids rather than just nutrients. Detention pond design criteria based on predictive empirical models and observation have been available for many decades (see for example Walker 1987, OMOE 2003) and incorporate factors such as the watershed area captured so as the maximize retention time and removal efficiency. In Walker’s review of the design and performance of ponds in the St. Paul’s area of Minnesota removal efficiencies for P ranged from 47% to 68%. Brabec et al. (2002) reviewed the literature and noted a wide range of detention pond effectiveness for phosphorus removal (from 14% to 64%) as well as a general tendency for the effectiveness of BMPs such as riparian buffers to become limited with an increase in the proportion of impervious cover. Walker (1987) noted the importance of operational factors required to maintain depth and hydraulic residence time and suggested that since construction costs are typically less than post operational dredging costs that the design incorporate excess capacity to allow for sediment accumulation and the gradual loss of capacity. More recently, Mulroy (2010) assessed the performance of two SWM pond designs (conventional and extended detention) in Waterloo, Ontario and quantified the concentration of TP, soluble reactive phosphorus (SRP) and suspended solids (TSS) at the inlet(s) and outlet of each pond. The trap efficiency for each pond was characterized for

35

30 baseflow and 10 stormflow events and results showed that the conventional design pond retained 65% of influent TP and 36% SRP and that both ponds were sinks for TSS during baseflow periods with 61% and 48% of SS retained for the extended detention and conventional ponds respectively. Retention during storm events was higher compared to baseflow periods with the extended detention pond retaining 11%, 40% and 66% of TP, SRP and TSS, and the conventional pond having a trap efficiency of 90%, 93% and 97% for TP, SRP and TSS. A notable finding was that the extended detention design pond was actually a source of TP and SRP during baseflow periods with trap efficiencies of -4% and -61% respectively. Although dry weather loads are very small relative to wet weather events and hence SWM facilities achieve overall positive removal, this last finding does highlight the need for further research into P speciation and cycling at the sediment-water interface to further improve trap efficiencies through improved pond design. Other LID initiatives such as green roofs have also been assessed. Linden and Stone (2009) undertook a study to assess the utility of vegetated roofs as a source control BMP for stormwater in Waterloo, Ontario. They determined that the vegetated roof had an average stormwater retention rate of 47.6% compared with 4.7% in the control roof and that the mean storage capacity of the vegetated roof was over ten times greater than for the control roof. Although the roof proved to be an effective means of flow reduction it did not significantly improve water quality and was a source of TP and SRP in comparison with the control roof presumably due to the organic content present in the vegetated roof growth medium and fertilizer application. Van Seters et al. (2009) undertook a three-year study to evaluate the quantity and quality of runoff from an extensive green roof on a multistory building in Toronto. Analyses of eleven commercially available green roof growing media were also undertaken to help identify the potential influence that the growing media may have on runoff chemistry. The survey design provided sufficient statistical power for all comparisons of green roof and conventional roof results. Continuous precipitation and runoff data demonstrated that the green roof discharged 63% less runoff than a neighbouring conventional roof with runoff volumes from the green roof averaging 42% less than the conventional roof in April and November, and between 70 and 93% less during the summer months. Water samples were collected from both roofs during 21 rain events in 2003 and 2004 showed reduced loads of many contaminants from the green roof. Total P was an exception with concentrations in green roof runoff being significantly higher than the conventional roof although P concentrations fell significantly after the first year of monitoring suggesting that it was being leached from the media. The study flagged the importance of engineering green roof media to minimize leaching of nutrients and other contaminants while maintaining their ability to support plant growth. Study results such as those summarized above have demonstrated the potential for SWM ponds to achieve substantial reductions in concentrations of suspended solids and nutrients such as TP, but a consistent theme is that the efficacy of retention facilities is highly dependent on proper operation and maintenance. The finding that ponds can

36

function as a source of P during baseflow conditions due to physical, chemical and biological cycling of P is also of considerable significance. Results of green roof studies are also instructive in that they demonstrate the clear benefits of green roof facilities for flow management while recognizing that any corresponding reduction in nutrient loading will be diminished somewhat by the potential increase in nutrient concentrations due to leaching from the roofs. Consideration of allocating resources to future studies related to assessing the effectiveness of SWM infrastructure would benefit from a SMDP discussion to clarify objectives and assessment end points. For example, if the objective is to assess the ongoing performance of a specific SWM pond facility to determine the need for maintenance dredging, routine monitoring of physical characteristics such as depth and volume at all SWM facilities as part of ongoing infrastructure maintenance may generally be sufficient to support a policy that incorporates periodic solids removal without the need for detailed study. If the goal is to assess the optimal combination of SWM infrastructure as part of new development or retrofitting of existing development the SMDP discussion regarding future monitoring and modeling would benefit from existing guidance documents (e.g. OMOE 2003; ICF Marbek 2012) as well as the results from past programs such SWAMP (2005) and the ongoing STEP program. Engineering criteria will always be important but in many cases, pragmatism may dictate overall choices due to limitations associated with existing development or opportunities associated with green field development. Assessing the effectiveness of policy initiatives such as limiting the application of lawn fertilizers containing phosphorus or upgraded street cleaning and catch basin maintenance will function at a larger scale than for assessment of specific infrastructure projects. Studies undertaken in Minnesota (Minnesota Department of Agriculture 2007) and Ann Arbor, Michigan (Lehman et al. 2009, 2011) have shown mixed results. The Minnesota Phosphorus Lawn Fertilizer law was enacted in 2002 and amended in 2004 and prohibited the use of P fertilizer unless new turf is being established or soil testing demonstrated a need. The Minnesota Department of Agriculture was able to demonstrate a 48% reduction in P applied as fertilizer but found it challenging to document improvements in water quality due to extremely high variability in the runoff data in the years following fertilizer restrictions and the high number of P runoff sources that needed observation. Unsurprisingly, much of the variability in water quality was driven by year-to-year variation in the amount and timing of precipitation. They also observed that the effects of the fertilizer restrictions were difficult to isolate from other sources such as construction site runoff, grass clippings, animal waste, tree leaves and seeds. They noted the potential for a lag time in the response due to the release of P from lake sediments and concluded that many years of watershed monitoring data may be required to discern statistically significant trends. The issue of insufficient statistical power to discern water quality trends attributable to BMP or policy implementation is extremely common and, in many cases, the best solution is to narrow the scope of the

37

monitoring to a finer temporal and spatial scale so that the connection between the abatement action and the receiving water effect is more immediate. Lehman et al. (2009, 2011) set out to assess improvements in water quality following implementation of a municipal ordinance limiting the application of lawn fertilizers containing phosphorus. Their field site was part of the Huron River catchment in southeastern Michigan. They employed a statistical survey design that held Type I error reasonably low while seeking a sufficient level of statistical power to detect environmental changes if they occurred. Their survey design calculations indicated that weekly sampling could reasonably be expected to detect a 25% change in DP and TP after two to three years. Water quality data collected weekly from May to September 2008–2010 were indeed able to demonstrate statistically significant reductions in total phosphorus (TP), dissolved phosphorus (DP) for the portion of the drainage basin affected by the ordinance. No reductions were seen at an upstream control river site not affected by the ordinance and the effect was confirmed by comparison with other water quality indicators that did not change systematically at experimental sites. The data were compared with a multiyear historical dataset that preceded the ordinance and mean P concentrations were lower than those of the reference period in 43 of 45 comparisons. A statistical model that incorporated temporal effects, stream flow, and variations at the upstream control site detected highly significant effects of pre- or post-ordinance period for P, but typically no significant effects for other water quality indicators. The average reductions compared to reference conditions from June to September at three sites affected by the ordinance ranged from 24 to 52% for SRP, 23 to 35% for DP, and 11 to 23% for TP. These findings provide valuable insights and guidance for those contemplating embarking on a monitoring and modeling study to demonstrate the effectiveness of a policy implementation such as this ordinance banning the use of P fertilizer. The primary message is that the ability to detect change of this kind requires a survey explicitly designed with sufficient statistical power along with a large degree of change within the boundaries of the assessed area. This Michigan experience provides useful case study that would be a useful information source for a SMDP discussion regarding the deployment of resources to demonstrate the effectiveness of any similar policy. A final observation that needs to be made is that although BMPs are, by definition, the “best” practices selected to reduce runoff and pollution within a particular catchment area, they differ from engineered treatment facilities such as WWTPs with performance metrics that must be guaranteed to meet regulatory requirements. Optimizing design is important to achieve the best results possible for BMPs and LID so that they can be expected to achieve positive change or at least minimize negative change following development of natural lands. Realistically, however, their performance cannot be predicted and guaranteed over a wide range of conditions through time in the same way

38

as a WWTP facility. This means that although assessment of performance will provide useful feedback regarding the efficacy of a particular practice at a particular site under a range of conditions, it will generally be unrealistic to set precise performance metrics analogous to those associated with WWTPs. It is also worth considering that despite the intrinsic uncertainty of BMP and LID measures on receiving water quality, they will generally deliver multiple benefits rather than just achieving a reduction in nutrient loads. This means that although it will be difficult to quantify the TP load reduction benefits of allocating resources to urban stormwater management projects, the list of additional co-benefits can make a compelling case for such allocations despite this uncertainty. The most obvious of these co-benefits would include flow management (and the related benefits associated with flood and erosion control) as well as improved water quality through reduction in inputs of suspended solids (water clarity) and other substances with potentially toxic effects such as ammonia, nitrate, Biochemical Oxygen Demand (BOD), pathogens, heavy metals, PAHs, in-use pesticides and PCBs. For example, the City of Toronto Wet Weather Flow Master Plan was designed to address the objectives of the City’s Environmental Plan regarding water quality improvements and make progress towards the water quality improvement objectives of the Toronto and Region Remedial Action Plan (D’Andrea et al. 2004). Overall benefits projected through the implementation of the Master Plan were substantial across the City and included:

Swimmable waterfront beaches;

Elimination of combined sewer overflows - in compliance with Ministry of the Environment requirements;

Elimination of dry weather discharges (cross connections);

Basement flooding protection;

Protection of City’s infrastructure from stream erosion;

Restoration of degraded local streams and improved stream water quality;

Reduction of algae growth along the waterfront and in streams; and

Restoration of aquatic habitat. In this case, the reduced eutrophication benefit is only one of eight categories used to justify the implementation of an ambitious (> $1 billion) 25-year plan incorporating a combination of source, conveyance and end-of-pipe controls. In effect, these co-benefits become part of a multiple-lines-of-evidence approach to evaluating the costs and benefits of urban NPS controls. It is also worth noting that although these multiple benefits will improve water quality conditions in the downstream receiving water body (Lake Ontario in the case of Toronto), these benefits accrue within the watersheds themselves and will therefore yield human health and ecological benefits over a larger geographical scale than those resulting from nutrient reductions at WWTPs discharging directly to the downstream water body.

39

4. CONCLUSIONS AND RECOMMENDATIONS FOR DEVELOPMENT OF AN URBAN NPS NUTRIENT LOAD WATERSHED DECISION FRAMEWORK The preceding survey of studies associated with nutrient loadings from tributaries, identifying land use effects and CSAs, and evaluation of policies, BMPs and LID measures have demonstrated a general relationship between uncertainty and the scale of assessment. In general, the least uncertainty is observed at the watershed scale where the fine-scale differences tend to average out and where high-frequency, event-based sampling can yield loading estimates with statistical error terms that are less than those associated with receiving water eutrophication models. Of course, generating urban NPS nutrient loading estimates sufficient to meet the requirement of a receiving water model is only the first step. In order to act on this information and rectify receiving water eutrophication issues by achieving loading reduction targets it is necessary to find locally controllable sources within watersheds. Typically, the greatest uncertainty in estimating nutrient loading and export is observed at the smallest scales associated with evaluating the effectiveness of local stormwater management measures and facilities due to the intrinsic spatial and temporal variability in responses to wet weather at this scale. The highest degree of certainty is associated with identifying the need for action at a large scale while identifying urban NPS over-contributing areas and evaluating the effectiveness of remedial measures is associated with the greatest uncertainty. Although it is apparent that much of this fine-scale uncertainty is intrinsic, using the co-benefits of urban stormwater management controls as multiple-lines-of-evidence when examining costs and benefits suggests that it is possible to develop plans despite this uncertainty (e.g. as reflected in the Toronto Wet Weather Flow Master Plan). Several themes emerge from this survey of current practices and the February 2018 CWN Workshop:

i. Although currently available tools for estimating urban nutrient loads from watersheds to eutrophication-sensitive receiving water bodies may not have been applied in all relevant problem areas, they are generally sufficient to confirm the need for development of action plans to achieve loading reductions;

ii. Existing data on the effects of urban land use on tributary water quality are generally sufficient to provide screening level estimates of the potential changes associated with urban development in rural or natural landscapes although such estimates may not have been generated for all areas of potential interest;

iii. Current models for assessing stormwater flow in urban catchments can provide reasonable predictions for various infrastructure improvement scenarios although there may be relevant problem areas where they have not yet been used;

iv. Although the effectiveness of BMPs and LID practices cannot be predicted and guaranteed over a wide range of conditions through time, consideration of their multiple co-benefits (i.e. in addition to P load reductions) and application of a “one

40

water” approach (i.e. drinking water, wastewater and stormwater) will provide multiple lines-of-evidence to support their implementation; and

v. Assistance from senior levels of government with the implementation of urban stormwater control projects, BMPs and LID in watersheds would yield the greatest progress towards reduction of urban NPS sources of TP as well as achieving a wide range of other water quality and quantity benefits both within watersheds and receiving water bodies.

These themes suggest that it would be helpful to develop a standardized, iterative decision-making process to guide water managers and decision-makers that incorporates the lessons learned to date in various jurisdictions. More specifically, this process should follow a progression from:

1. Confirming a receiving water quality problem (e.g. eutrophication);

2. Confirming the relative significance of urban nonpoint sources (i.e. compared with point sources and rural/agricultural NPS);

3. Identifying high priority or Critical Source Areas for urban NPS

nutrient/contaminant inputs;

4. Evaluating and selecting appropriate policies, BMPs, infrastructure improvement to manage urban runoff and ameliorate urban NPS nutrient/contaminant inputs; and

5. Confirming that implemented solutions are achieving acceptable results.

The iterative inclusion of Scientific-Management Decision Point (SMDP) discussions between steps similar to those described in the U.S. EPA Superfund guidance (U.S. EPA 2001; see Appendix A for schematic outline) would clarify and confirm management objectives and ensure consensus prior to proceeding with management actions or additional study. The discussion would consider the adequacy of existing information (e.g. output from a Screening Level Assessment) and whether management decision-makers require additional information from a more Detailed, Quantitative Assessment generating more locally relevant data before proceeding. The discussion would also compel all interested parties to consider the probable reduction in uncertainty that more detailed studies would yield and whether this would be sufficient to justify the additional work. A key benefit of these discussions will be to clarify objectives and document the management-related information needs to achieve them. Such an iterative approach will intrinsically incorporate adaptive management into the decision-making process and will provide a strong, logical basis for evaluating the benefits of additional information gathering, and the refinement of existing monitoring and modeling tools, at any step from problem definition to implementation of solutions.

41

In many situations the relative significance of urban NPS nutrient loads on receiving water bodies may be small relative to those from rural and agricultural landscapes. A regional assessment of land use would provide useful insights as to where the benefits of urban stormwater management will deliver the greatest benefit to downstream receiving water bodies. For example, as previously noted, urban land use accounts for less than 4% of the Canadian portion of the Lake Erie drainage basin but represents almost 20% of the area draining into Western Lake Ontario (OMNRF 2015). The high degree of urbanization in the zone from Oshawa to Niagara-on-the-Lake means that the combination of urban NPS and WWTP inputs will have a greater influence here than anywhere else in the Canadian Great Lakes Basin. A Screening Level Assessment based on land-use based export coefficients to estimate relative contributions from municipal WWTPs and urban NPS in this region would be of considerable value in determining the path forward once an assessment of conditions and processes in Lake Ontario has yielded some direction regarding optimum nutrient loads. It is possible that the complexity of nutrient dynamics involving colonization by invasive mussels and growth of benthic algae will constrain the calculation of nutrient loading targets, at least in the short and medium term. In this context, therefore, the decision-making framework should also explore the multiple, local-scale co-benefits (i.e. those in addition to reductions in P loads and concentrations) within watersheds to establish the merits of urban stormwater remedial actions, as well as implementation of BMPs and LID for new urban development. Ultimately, the most effective management decisions will consider all benefits within watersheds, rather than just those that accrue to downstream receiving waterbodies as the result of P load reductions.

42

5. REFERENCES

Alves, A., Sanchez, A., Gersonius, B., & Vojinovic, Z. (2016). A model-based framework

for selection and development of multi-functional and adaptive strategies to cope with

urban floods. Procedia Engineering, 154, 877-884. doi:10.1016/j.proeng.2016.07.463

Annex 4 Objectives and Targets Task Team (2015). Recommended phosphorus loading

targets for Lake Erie. Final report to the Nutrients Annex Subcommittee May 11, 2015.

Retrieved from: https://www.epa.gov/sites/production/files/2015-

06/documents/report-recommended-phosphorus-loading-targets-lake-erie-201505.pdf

Baker, D. B., Confesor, R., Ewing, D. E., Johnson, L. T., Kramer, J. W., & Merryfield, B. J.

(2014). Phosphorus loading to Lake Erie from the Maumee, Sandusky and Cuyahoga

rivers: The importance of bioavailability. Journal of Great Lakes Research, 40(3), 502-

517. doi:10.1016/j.jglr.2014.05.001

Bannerman, R. T., Owens, D. W., Dodds, R. B., & Hornewer, N. J. (1993). Sources of

pollutants in Wisconsin stormwater. Water Sci. Technol, 28(3-5), 241-259.

Baralkiewicz, D., Chudzińska, M., Szpakowska, B., Świerk, D., Goldyn, R. & Dondajewska,

R. (2014). Storm water contamination and its effect on the quality of urban surface

waters. Environ Monit Assess, 186(10), 6789-6803. doi: 10.1007/s10661-014-3889-0

Basu, N. B., Destouni, G., Jawitz, J. W., Thompson, S. E., Loukinova, N.V., Darracq, A.,

Zanardo, S., Yaeger, M., Sivapalan, M., Rinaldo, A. & Rao, P. S. C. (2010). Nutrient loads

exported from managed catchments reveal emergent biogeochemical stationarity.

Geophysical Research Letters, 37(23). doi:10.1029/2010GL045168

Bay Area Restoration Council (2014). Muddied waters - The ongoing challenge of

sediment and phosphorus for Hamilton Harbour remediation. 2014 Towards Safe

Harbour Report. Retrieved from:

http://hamiltonharbour.ca/resources/documents/TSHReport,2014.pdf

Bay of Quinte Remedial Action Plan (2016). Impaired beneficial use #8 and phosphorus

reduction strategy for Bay of Quinte. Bay of Quinte Nutrient Monitoring Workshop, Ajax

Convention Centre, June 13, 2016.

43

Betanzo, E. A., Choquette, A. F., Reckhow, K. H., Hayes, L., Hagen, E. R., Argue, D. M., &

Cangelosi, A. A. (2015). Water data to answer urgent water policy questions: Monitoring

design, available data and filling data gaps for determining the effectiveness of

agricultural management practices for reducing tributary nutrient loads to Lake Erie.

Northeast-Midwest Institute Report, 169 p. Retrieved from: http://www.nemw.org doi:

10.13140/RG.2.1.1102.5684

Bishop, S. L., Kerkovius, J. K., Menard, F., & Murch, S. J. (2017). N-β-Methylamino-L-

alanine and its naturally occurring isomers in cyanobacterial blooms in Lake Winnipeg.

Neurotox Res, 33(1), 133-142. doi:10.1007/s12640-017-9820-z

Boyd, D., D’Andrea, M., and Anderton, R. (1999). Assessment of six tributary discharges

to the Toronto area waterfront. Volume 1: Project synopsis and selected results. Volume

2: Technical appendix and data summary. Reports prepared for The Toronto and Region

Remedial Action Plan.

Brabec, E., Schulte, S., & Richards, P. L. (2002). Impervious surfaces and water quality: A

review of current literature and its implications for watershed planning. Journal of

Planning Literature, 16(4), 499-514. doi:10.1177/088541202400903563

Canadian Council of Ministers of the Environment (CCME) (1996). A framework for

ecological risk assessment: General guidance. The National Contaminated Sites

Remediation Program. ISBN 0-662-24246-3. Retrieved from:

https://www.ccme.ca/files/Resources/csm/pn_1195_e.pdf

Canadian Municipal Water Consortium (2015). Canadian municipal water priorities

report. Towards sustainable and resilient water management. Canadian Water Network.

Retrieved from: http://www.cwn-rce.ca/focus-areas/blue-cities/canadian-municipal-

water-priorities-report/

Chapman, P., Anderson, J., Boyd, D., Grapentine, L., deBarros, C., Santiago, R., Fletcher,

T., Zarull, M., Inch, P., Richman, L., Abernethy, S. & Welsh, P. (2007). Canada-Ontario

decision-making framework for assessment of Great Lakes contaminated sediment.

Environment Canada and Ontario Ministry of the Environment. ISBN 978-0-662-46148-7.

Retrieved from: http://publications.gc.ca/collections/collection_2010/ec/En164-14-

2007-eng.pdf

Clark, B. J. & Sellers, T. J. (Editors) (2014). Rainy-Lake of the Woods: State of the basin

report. 2nd ed. Lake of the Woods Water Sustainability Foundation. Retrieved from:

http://ijc.org/files/publications/R-LoW_SOBR_2014_web-res.compressed.pdf

44

D’Andrea, M., Snodgrass, W. J., & Chessie, P. D. (2004). Development of a wet weather

flow management master plan for the City of Toronto. Water Quality Research Journal

of Canada, 39(4), 417-431.

Dietz, M. E. (2007). Low impact development practices: A review of current research and

recommendations for future directions. Water Air Soil Pollut, 186(1-4), 351-363.

doi:10.1007/s11270-007-9484-z

Dolan, D. M., Yui, A. K., & Geist, R. D. (1981). Evaluation of river load estimation

methods for total phosphorus. Journal of Great Lakes Research, 7(3), 207-214.

doi:10.1016/S0380-1330(81)72047-1

Dolan, D. M. & Chapra, S. C. (2012). Great Lakes total phosphorus revisited: 1. Loading

analysis and update (1994-2008). Journal of Great Lakes Research, 38(4), 730-740.

doi:10.1016/j.jglr.2012.10.001

Dolan, D. M. & McGunagle, K. P. (2005). Lake Erie total phosphorus loading analysis and

update: 1996-2002. Journal of Great Lakes Research, 31 (Supplement 2), 11-22.

doi:10.1016/S0380-1330(05)70301-4

Dressing, S.A., Meals, D.W., Harcum, J.B., Spooner, J., Stribling, J.B., Richards, R.P.,

Millard, C.J., Lanberg, S.A., & O’Donnell, J.G. (2016). Monitoring and evaluating nonpoint

source watershed projects. Contract to U.S. Environmental Protection Agency by Tetra

Tech, Inc. Retrieved from: https://www.epa.gov/nps/monitoring-and-evaluating-

nonpoint-source-watershed-projects

Driver, N. E. & Tasker, G. D. (1990). Techniques for estimation of storm-runoff loads,

volumes, and selected constituent concentrations in urban watersheds in the United

States. U.S. Geological Survey Water-Supply Paper No. 2363. Retrieved from:

https://pubs.usgs.gov/wsp/wsp2363/pdf/wsp_2363.pdf

Duan, S., Kaushal, S. S., Groffman, P. M., Band, L. E., & Belt, K. T. (2012). Phosphorus

export across an urban to rural gradient in the Chesapeake Bay watershed. Journal of

Geophysical Research, 117(G1). doi:10.1029/2011JG001782.

Egemose, S. & Jensen, H.S. (2009). Phosphorus forms in urban and agricultural runoff:

Implications for management of Danish Lake Nordborg. Lake and Reservoir

Management, 25(4), 410-418. doi:10.1080/07438140903413228

45

Environment Canada & Manitoba Water Stewardship (2011). State of Lake Winnipeg:

1999 to 2007. Manitoba Watershed Stewardship Division, Water Science and

Management. Retrieved from:

https://gov.mb.ca/waterstewardship/water_quality/state_lk_winnipeg_report/pdf/stat

e_of_lake_winnipeg_rpt_technical_low_resolution.pdf

Fletcher, R., Welsh, P., & Fletcher, T. (2008). Guidelines for identifying, assessing and

managing contaminated sediments in Ontario: An integrated approach. Ontario Ministry

of the Environment, PIBS 6658e 2008 Queen’s Printer for Ontario. Retrieved from:

https://www.ontario.ca/document/guidelines-identifying-assessing-and-managing-

contaminated-sediments-ontario

Fox, D. R., Etchells, T. & Tan, K. S. (2005). Protocols for the optimal measurement of

nutrient loads. A Report to West Gippsland Catchment Management Authority. Australian

Centre for Environmetrics. Retrieved from: http://www.loveourlakes.net.au/wp-

content/uploads/2015/05/Protocols_for_the_optimum_measurement_of_Nutrient_Load

s.pdf

Gauch, H. G., Hwang, J. T. G., & Fick, G. W. (2003). Model evaluation by comparison of

model-based predictions and measured values. Agronomy Journal, 95(6), 1442-1446.

doi:10.2134/agronj2003.1442

Goonetilleke, A., Thomas, E., Ginn, S., & Gilbert, D. (2005). Understanding the role of

land use in urban stormwater quality management. Journal of Environmental

Management, 74(1), 31-42. doi:10.1016/j.jenvman.2004.08.006

Gudimov, A., Ramin, M., Stremilov, S., & Arhonditsis, G. B. (2010). Eutrophication risk

assessment in Hamilton Harbour: System analysis and evaluation of nutrient loading

scenarios. Journal of Great Lakes Research, 36(3), 520-539.

doi:10.1016/j.jglr.2010.04.001

Guldin, J. M., Cawrse, D., Graham, R., Hemstrom, M., Joyce, L., Kessler, S., McNair, R.,

Peterson, G., C.G. Shaw, C. G., Stine, P., Twery, M., & Walter, J. (2003). The science

consistency review: A tool to evaluate the use of scientific information in land

management decision making. Publication FS-772. United States Department of

Agriculture, Forest Service, Washington Office. Retrieved from:

https://www.srs.fs.usda.gov/pubs/misc/fs_772.pdf

46

Hamilton Harbour Remedial Action Plan (RAP) (1992). Goals, options and

recommendations – Stage 2 Report. Remedial Action Plan for Hamilton Harbour, ISBN 0-

7778-0533-2. Retrieved from:

http://hamiltonharbour.ca/resources/documents/HamiltonHarbourRAPStageTwoSumm

aryReport,1992.pdf

Hatt, B.E., Fletcher, T.D., Walsh, C.J., Taylor, S.L. (2004). The influence of urban density

and drainage infrastructure on the concentrations and loads of pollutants in small

streams. Environ Manage, 34, 112-124. doi:10.1007/s00267-004-0221-8

Hecky, R. E., Smith, R. E. H., Barton, D. R., Guildford, S. J., Taylor, W. D., Charlton, M. N.,

& Howell, T. (2004). The nearshore phosphorus shunt: A consequence of ecosystem

engineering by dreissenids in the Laurentian Great Lakes. Can. J. Fish. Aquat. Sci., 61,

1285-1293. doi:10.1139/F04-065.

Higgins, S. N., Howell, E. T., Hecky, R. E., Guildford, S. J. & Smith, R. E. (2005). The wall of

green: The status of Cladophora glomerata on the northern shores of Lake Erie’s eastern

basin, 1995–2002. J. Great Lakes Res., 31, 547-563. doi:10.1016/S0380-1330(05)70283-5

Higgins, S. N., Pennuto, C. M. Howell, E. T., Lewis, T. & Makarewicz, J. C. (2012). Urban

influences on Cladophora blooms in Lake Ontario. J Great Lakes Res., 38(Supplement 4),

116-123. doi:10.1016/j.jglr.2011.11.017

Hirsch, R. M., Moyer, D. L., & Archfield, S. A. (2010). Weighted regressions on time,

discharge, and season (WRTDS), with an application to Chesapeake Bay River inputs.

Journal of the American Water Resources Association, 46(5), 857-880.

doi:10.1111/j.1752-1688.2010.00482.x

Hobbie, S. E., Finlay, J. C. Janke, B. D. Nidzgorski, D. A., Millet, D. B., & Baker, L. A. (2017).

Contrasting nitrogen and phosphorus budgets in urban watersheds and implications for

managing urban water pollution. Proc Natl. Acad. Sci. USA, 114(16), 4177-4182.

doi:10.1073/pnas.1618536114

Holeton, C. (2013). Sources of nutrients and sediments in the Grand River watershed.

Grand River watershed water management plan. Prepared by the Water Quality

Working Group. Grand River Conservation Authority, Cambridge, ON. Retrieved from:

https://www.grandriver.ca/en/our-

watershed/resources/Documents/WMP/Water_WMP_Report_NutrientSources.pdf

47

Howarth, R.W., Billen, G., Swaney, D., Townsend, A., Jaworski, N., Lajtha, K., Downing,

J.A., Elmgren, R., Caraco, N., & Jordan. T. (1996). Regional nitrogen budgets and riverine

N and P fluxes for the drainages to the North Atlantic Ocean: Natural and human

influences. Biogeochemistry, 35, 75-139.

Hutchinson, S. L., Keane, T., Christianson, R. D., Skabeland, L., Moored, T. L., Greene, A.

M., & Kingery-Page, K. (2011). Management practices for the amelioration of urban

stormwater. Procedia Environmental Sciences 9, 83-89.

doi:10.1016/j.proenv.2011.11.014

ICF Marbek (2012). Low impact development discussion paper. Retrieved from:

http://www.sustainabletechnologies.ca/wp/wp-content/uploads/2014/09/LID-

Discussion-Paper_Nov-2012.pdf

International Joint Commission (IJC) (1978). Great Lakes water quality agreement of

1978. Agreement, with annexes and terms of reference, between the United States and

Canada signed at Ottawa November 22, 1978. International Joint Commission United

States and Canada. Retrieved from: www.ijc.org/files/tinymce/uploaded/GLWQA_e.pdf

International Joint Commission (IJC) (2004). The impact of urban development on water

Quality. 12th Biennial report on Great Lakes water quality. Retrieved from:

http://www.ijc.org/php/publications/html/12br/english/report/index.html

Jantz, C. A., Goetz, S. J., & Shelley, M. K. (2004). Using the SLEUTH urban growth model

to simulate the impacts of future policy scenarios on urban land use in the Baltimore -

Washington metropolitan area. Environment and Planning B Planning and Design, 31(2),

251-271. doi:10.1068/b2983

Jeje, Y. (2006). Export coefficients for total phosphorus, total nitrogen and total

suspended solids in the southern Alberta region: A review of literature. Alberta

Environment Pub. No. T-863 ISBN: 0-7785-4617-9.

Kobayashi, K. & Salam, M. U. (2000). Comparing simulated and measured values using

mean squared deviation and its components. Agronomy Journal, 92, 345-352.

doi:10.1007/s100870050043

Labencki, T. (2010). Hamilton Harbour RAP water quality goals and targets interim part

2 review (draft). Report to Water Quality Technical Team.

48

Lake Simcoe Science Advisory Committee (2008). A compendium of reports prepared by

the Lake Simcoe Science Advisory Committee: Report 2 - Stressors affecting the

ecological health of Lake Simcoe and its watershed. Report to the Minister of the

Environment. Queen’s Printer for Ontario. Retrieved from:

https://archive.org/stream/stdprod078249.ome/stdprod078249_djvu.txt

Lake Simcoe Science Advisory Committee (2012). Recommendations for the ecological

health and sustainable future of Lake Simcoe and its watershed. Report to the Minister

of the Environment. Queen’s Printer for Ontario.

Lake Winnipeg Stewardship Board (2006). Reducing nutrient loading to Lake Winnipeg

and its watershed, our collective responsibility and commitment to action. Report to the

Minister of Water Stewardship. Retrieved from:

https://www.gov.mb.ca/waterstewardship/water_quality/lake_winnipeg/lwsb2007-

12_final_rpt.pdf

Lamera, C., Becciu, G., Rulli, M. C. & Rosso, R. (2014). Green roofs effects on the urban

water cycle components. Procedia Engineering 70, 988-997.

doi:10.1016/j.proeng.2014.02.110

LeBoutillier, D. W., Kells, J. A., & Putz, G. J. (2000). Prediction of pollutant load in

stormwater runoff from an urban residential area. Canadian Water Resources Journal,

25(4), 343-359. doi:10.4296/cwrj2504343

Lee C. J., Hirsch, R. M., Schwarz, G. E., Holtschlag, D. J., Preston, S. D., Crawford, C. G., &

Vecchia, A. V. (2016). An evaluation of methods for estimating decadal stream loads.

Journal of Hydrology 542, 185-203. doi:10.1016/j.jhydrol.2016.08.059

Lee, C. J., Murphy, J. C., Crawford, C. G., & Deacon, J. R. (2017). Methods for computing

water-quality loads at sites in the U.S. Geological Survey National Water Quality Network.

Open-File Report 2017-1120. U.S. Department of the Interior and U.S. Geological Survey.

Retrieved from: https://pubs.usgs.gov/of/2017/1120/ofr20171120.pdf

Lehman, J. T., Bell, D.W., & McDonald, K. E. (2009). Reduced river phosphorus following

implementation of a lawn fertilizer ordinance. Lake and Reservoir Management, 25,

307-312.

49

Lehman, J. T., Bell, D.W., Doubek, J.P., & McDonald, K.E. (2011). Reduced additions to

river phosphorus for three years following implementation of a lawn fertilizer

ordinance. Lake and Reservoir Management, 27, 390-397.

doi:10.1080/07438141.2011.629769

Leopold, L. B. (1968). Hydrology for urban land planning - A guidebook on the hydrologic

effects of urban land use. Geological Survey Circular, 554. Retrieved from:

https://pubs.usgs.gov/circ/1968/0554/report.pdf

Lin, J. P. (2004). Review of published export coefficient and event mean concentration

(EMC) data. Wetlands Regulatory Assistance Program, ERDC TN-WRAP-04-3.

Linden, K.V. & Stone, M. (2009). Treatment performance of an extensive vegetated roof

in Waterloo, Ontario. Water Quality Research Journal of Canada, 44(1), 26-32. (Special

Issue: Innovative Approaches to Stormwater Management in Canada).

Long, T., Wellen, C, Arhonditsis, G. & Boyd, D. (2014). Evaluation of stormwater and

snowmelt inputs, land use and seasonality on nutrient dynamics in the watersheds of

Hamilton Harbour, Ontario, Canada. Journal of Great Lakes Research 40(4), 964-979.

doi: 10.1016/j.jglr.2014.09.0171

Long, T., Wellen, C., Arhonditsis, G., Boyd, D., Mohamed, M., & O'Connor, K. (2015).

Estimation of tributary total phosphorus loads to Hamilton Harbour, Ontario, Canada,

using a series of regression equations. Journal of Great Lakes Research, 41(3), 780-793.

doi:10.1016/j.jglr.2015.04.001

Maccoux, M. J., Dove, A., Backus, S. M., & Dolan, D.M. (2016). Total and soluble reactive

phosphorus loadings to Lake Erie: A detailed accounting by year, basin, country, and

tributary. Journal of Great Lakes Research, 42(6), 1151-1165.

doi:10.1016/j.jglr.2016.08.005

Marsalek, J. (1978). Pollution due to urban runoff: Unit loads and abatement measures.

Report to the Pollution from Land Use Activities Reference Group. Retrieved from:

http://agrienvarchive.ca/download/PLUARG_pollution_urban_runoff.pdf

Maunder, D. Whyte, R., & D’Andrea, M. (1995). Metropolitan Toronto waterfront wet

weather outfall study – Phase II City of Toronto. Report for Metropolitan Toronto and

Region Remedial Action Plan. Queen's Printer for Ontario. Retrieved from:

https://archive.org/stream/METROPOLITANTORO00MAUN08179.ome/METROPOLITAN

TORO00MAUN08179_djvu.txt

50

Mayer, T., Marsalek, J., & Delos Reyes, E. (1996). Nutrients and metal contaminants

status of urban stormwater ponds, lake and reservoir management, 12(3), 348-363.

doi:10.1080/07438149609354276

McElmurry, S. P. (2014). Inventorying urban phosphorus loads. A report to the Bi-

National Annex 4 Municipal and Rural Non-Farm Task Group. Retrieved from:

http://abca.iwebsmart.net/downloads/McElmurry-2014.pdf

McFarland, A. M. S. & Hauck, L. M. (2001). Determining nutrient export coefficients &

source loading uncertainty using in-stream monitoring data. Journal of the Amer. Water

Res. Assoc., 37(1), 223-236. doi:10.1111/j.1752-1688.2001.tb05488.x

Meals, D. W., Dressing, S. A. & Davenport, T. E. (2010). Lag time in water quality

response to best management practices: A review. J. Environ. Qual., 39, 85-96,

doi:10.2134/jeq2009.0108.

Miller, J. D. & Hutchins, M. (2017). The impacts of urbanisation and climate change on

urban flooding and urban water quality: A review of the evidence concerning the United

Kingdom. Journal of Hydrology: Regional Studies, 12, 345-36.

doi:10.1016/j.ejrh.2017.06.006

Mills, T. J., Quigley, T. M. & Everest, F. J. (2001). Science-based natural resource

management decisions: What are they? Renewable Resources Journal, 19(2), 10-15,

Summer 2001.

Minnesota Department of Agriculture (2007). Report to the Minnesota Legislature:

Effectiveness of the Minnesota phosphorus lawn fertilizer law. Pesticide and Fertilizer

Management Division.

Mohamed, M. N. (2012). Water quality of 15 streams in agricultural watersheds of

Southwestern Ontario 2004-2009. Seasonal patterns, regional comparisons, and the

influence of land use. Ontario Ministry of the Environment, Environmental Monitoring

and Reporting Branch, Queen’s Printer for Ontario, 2012 PIBS 8613e.

https://dr6j45jk9xcmk.cloudfront.net/documents/1180/98-15-streams-in-agricultural-

watersheds-en.pdf

Mulroy, K. (2010). Assessing the performance of two stormwater management ponds in

Waterloo, Ontario. M.Sc. thesis, UWSpace.

https://uwspace.uwaterloo.ca/bitstream/handle/10012/5057/Mulroy_Kathleen.pdf?se

quence=1&isAllowed=y

51

Myers, D. N., Thomas, M. A., Frey, J. W., Rheaume, S. J., & Button, D. T. (2000). Water

quality in the Lake Erie-Lake Saint Clair Drainages; Michigan, Ohio, Indiana, New York,

and Pennsylvania, 1996-98. U.S. Geological Survey Circular 1203, 35 p. Retrieved from:

https://pubs.usgs.gov/circ/circ1203/

Niraula, R., Kalin, L., Wang, R., & Srivastava, P. (2012). Determining nutrient and

sediment critical source areas with SWAT. Transactions of the ASABE, 55(1), 137-147.

O‘Brien, P.J. & Wehr, J.D. (2010). Periphyton biomass and ecological stoichiometry in

streams within an urban to rural land-use gradient. Hydrobiologia. 657, 89-105. doi:

10.1007/s10750-009-9984-5

O'Connor, K. (2003). Hamilton harbour remedial action plan process and derivation of

water quality targets. Report prepared for the Hamilton Harbour Remedial Action Plan

(HHRAP). Unpublished report prepared by the HHRAP Office February 2003.

Ontario Ministry of the Environment (OMOE) (2003). Stormwater management planning

and design manual. ISBN 0-7794-2969-9

Ontario Ministry of Natural Resources and Forestry (OMNRF) (2015). Southern Ontario

and resources information system (SOLRIS) Version 2.0: Data specifications. Science and

Research Branch, Natural Resources Information Section, SOLRIS Technical Team,

January 2015. ISBN PDF 978-1-4606-5238-1.

Pappenberger, F. & Beven, K. J. (2006). Ignorance is bliss: Or seven reasons not to use

uncertainty analysis, Water Resour. Res., 42(5), W05302. doi:10.1029/2005WR004820

Paul Theil Associates and Beak Consultants (1995). Metropolitan Toronto waterfront wet

weather outfall study - Phase I. Report for Metropolitan Toronto and Region Remedial

Action Plan. Queen's Printer for Ontario.

Pennino, M. J., McDonald, R. I., & Jaffe, P. R. (2016). Watershed-scale impacts of

stormwater green infrastructure on hydrology, nutrient fluxes, and combined sewer

overflows in the mid-Atlantic region. Science of the Total Environment, 565, 1044-1053.

doi:10.1016/j.scitotenv.2016.05.101

Pineiro, G., Perelman, S., Guerschman, J. P., & Paruelo, J. M. (2008). How to evaluate

models: Observed vs. predicted or predicted vs. observed? Ecological Modelling 216(3-

4), 316-322. doi:10.1016/j.ecolmodel.2008.05.006

52

Pitt, R. E. & Voorhees, J. (1989). Source load and management model: An urban

nonpoint source water quality model, Volumes I - III, Wisconsin Department of Natural

Resources, PUBL-WR-218-89.

Preston, S. D., Bierman, V. J., & Silliman, S. E. (1989). An evaluation of methods for the

estimation of tributary mass loads. Water Resources Research, 25(6), 1379-1389.

doi:10.1029/WR025i006p01379

Ramin, M., Stremilov, S., Labencki, T., Gudimov, A., Boyd, D., & Arhonditsis, G.B. (2011).

Integration of numerical modeling and Bayesian analysis for setting water quality

criteria in Hamilton Harbour, Ontario, Canada. Environmental Modelling and Software,

26(4), 337-353. doi:10.1016/j.envsoft.2010.08.006

Rast, W. & Lee, G. F. (1983). Nutrient loading estimates for lakes. Journal of

Environmental Engineering, 109(2). © ASCE, ISSN 0773-9372/83/0002-0502

Reckhow, K.H., Beaulac, M. N., & Simpson, J. T. (1980). Modeling phosphorous loading

and lake response under uncertainty: A manual and compilation of export coefficients.

EPA 440/5-80-011 U.S.E.P.A 214 pp. Retrieved from:

http://www.academia.edu/2606491/Modeling_phosphorus_loading_and_lake_respons

e_under_uncertainty_A_manual_and_compilation_of_export_coefficients

Richards, R. P. (1998). Estimation of pollutant loads in rivers and streams: A guidance

document for NPS programs. Prepared for U.S. Environmental Protection Agency (Grant

X998397-01-0). Retrieved from: http://abca.iwebsmart.net/downloads/Richards-

1998.pdf

Rolland, D. C., Bourget, S., Warren, A., Laurion, I., & Vincent, W. F. (2013). Extreme

variability of cyanobacterial blooms in an urban drinking water supply. Journal of

Plankton Research, 35(4), 744-758. doi:10.1093/plankt/fbt042

Schindler, D. W. (1974). Eutrophication and recovery in experimental lakes: Implications

for lake management. Science, 184(4139), 897-899. doi:10.1126/science.184.4139.897

Schindler, D. W. (2012). The dilemma of controlling cultural eutrophication of lakes.

Proc. R. Soc. B, 279(1746), 4322-4333. doi:10.1098/rspb.2012.1032

Stormwater Assessment Monitoring and Performance (SWAMP) Program (2005).

Synthesis of monitoring studies conducted under the stormwater assessment monitoring

and performance program. Report prepared for Great Lakes Sustainability Fund of the

53

Government of Canada, Toronto and Region Conservation Authority, Municipal Engineers

Association of Ontario, and Ontario Ministry of the Environment. Retrieved from:

https://sustainabletechnologies.ca/app/uploads/2013/01/Final_SWAMP_Synthesis.pdf

Stumpf, R. P., Wynne, T. T., Baker, D. B., & Fahnenstiel, G. L. (2012). Interannual

variability of cyanobacterial blooms in Lake Erie. PLoS ONE 7(8), e42444.

doi:10.1371/journal.pone.0042444

Suter, G.W. II. (1991). Ecological endpoints. In Ecological assessment of hazardous waste

sites: A field and laboratory reference, W. Warren-Hicks, B.R. Parkhurst, and S.S. Baker,

eds., pp. 2.1-2.28. Prepared by Kilkelly Environmental Associates for the U.S.

Environmental Protection Agency. EPA/600/3-89/013. Retrieved from:

https://rais.ornl.gov/documents/ECOASR.PDF

Sychterz, A., Schill, T. & Verspagen, B. (2014). Monitoring performance of low impact

development measures implemented at the Conestoga College South Campus. Journal

of Water Management Modeling C371. doi:10.14796/JWMM.C371

Thomas, I.A., Mellander, P.E., Murphy, P.N.C., Fenton, O., Shine, O., Djodjic, F., Dunlop,

& P. Jordan P. (2016). A sub-field scale critical source area index for legacy phosphorus

management using high resolution data. Agriculture, Ecosystems & Environment,

233(Oct 2016), 238-252. doi:10.1016/j.agee.2016.09.012

Toor, G. S., Occhipinti, M. L., Yang, Y. Y., Majcherek, T., Haver, D., & Oki, L. (2017).

Managing urban runoff in residential neighborhoods: Nitrogen and phosphorus in lawn

irrigation driven runoff. PLoS One, 12(6): e0179151. doi:10.1371/journal.pone.0179151

Toronto Water (2017). Wet weather flow master plan implementation status update.

Report PW21.6 to Public Works and Infrastructure Committee, April 24, 2017. Retrieved

from: https://www.toronto.ca/legdocs/mmis/2017/pw/bgrd/backgroundfile-

103216.pdf

Tremblay, R., Légaré, S., Pienitz, R., Vincent, W. F., & Hall, R. I. (2001). Paleolimnological

analysis of changes in the trophic status of Lake Saint-Charles, a drinking water reservoir

for the Québec urban community. Journal of Water Science, 14(4), 14(4), 489-510

[article in French].

U.S. Environmental Protection Agency (EPA) (1983). Results of the nationwide urban

runoff program. Volume I - Final Report. Water Planning Division WH-554. Retrieved

from: https://www3.epa.gov/npdes/pubs/sw_nurp_vol_1_finalreport.pdf

54

U.S. Environmental Protection Agency (EPA) (2001). Risk assessment guidance for

superfund (RAGS): Volume III - Part A, Process for conducting probabilistic risk

assessment. Office of Emergency and Remedial Response, U.S. Environmental

Protection Agency, Washington, DC 20460, EPA 540-R-02-002. Retrieved from:

https://www.epa.gov/risk/risk-assessment-guidance-superfund-rags-volume-iii-part

U.S. Geological Survey (USGS) (2003). Project gigalopolis: Urban and land cover

modeling. Retrieved from: http://www.ncgia.ucsb.edu/projects/gig/

Vallentyne, J. R. & Thomas, N. A. (1978). Fifth year review of Canada-United States Great

Lakes water quality agreement. Report of International Joint Commission (IJC) Task

Group 3: A Technical Group to Review Phosphorus Loadings. International Joint

Commission Digital Archive. Retrieved from: https://scholar.uwindsor.ca/ijcarchive/170

Van Meter, K. J. & Basu, N. B. (2017). Time lags in watershed-scale nutrient transport:

An exploration of dominant controls. Environ. Res. Lett., 12, 084017. doi:10.1088/1748-

9326/aa7bf4

Van Seters, T., Rocha, L., Smith, D., & MacMillan, G. (2009). Evaluation of green roofs for

runoff retention, runoff quality, and leachability. Water Quality Research Journal of

Canada 44(1), 33-47. (Special Issue: Innovative Approaches to Stormwater Management

in Canada).

Walker, W. W. Jr. (1987). Phosphorus removal by urban runoff detention basins. Lake

and Reservoir Management, 3(1), 314-326. doi:10.1080/07438148709354787

Waschbusch, R. J., Selbig, W. R., & Bannerman, R. T. (1999). Sources of phosphorus in

stormwater and street dirt from two urban residential basins in Madison, Wisconsin,

1994-95. USGS Water-Resources Investigations Report 99-4021. Retrieved from:

https://pubs.er.usgs.gov/publication/wri994021

Wei, Z., Simin, L., & Fengbing, T. (2013). Characterization of urban runoff pollution

between dissolved and particulate phases. The Scientific World Journal, Volume 13,

Article ID 964737, doi:10.1155/2013/964737

Wellen, C., Arhonditsis, G. B., Labencki, T., & Boyd, D. (2014a). Application of the

SPARROW model in watersheds with limited information: A Bayesian assessment of the

model uncertainty and the value of additional monitoring. Hydrol. Process., 28(3), 1260-

1283. doi:10.1002/hyp.9614

55

Wellen, C., Arhonditsis, G. B., Long, T., & Boyd, D. (2014b). Accommodating

environmental thresholds and extreme events in hydrological models: A Bayesian

approach. Journal of Great Lakes Research, 40(Supplement 3), 102-116.

doi:10.1016/j.jglr.2014.04.002

Wellen, C., Arhonditsis, G.B. Long, T., & Boyd, D. (2014c). Quantifying the uncertainty of

nonpoint source attribution in distributed water quality models: A Bayesian assessment

of SWAT’s sediment export predictions. Journal of Hydrology, 519, 3353-3368.

Wellen, C., Kamran-Disfani, A-R., & Arhonditsis, G. B. (2015). Evaluation of the current

state of distributed watershed nutrient water quality modeling. Environ. Sci. Technol.

49(6), 3278-3290. doi:10.1021/es5049557

Winter, J. G., Eimers, M. C., Dillon, P. J., Scott, L. D., Scheider, W. A., & Willox, C. C.

(2007). Phosphorus inputs to Lake Simcoe from 1990 to 2003: Declines in tributary loads

and observations on lake water quality. Journal of Great Lakes Research 33, 381–396.

doi:10.3394/0380-1330(2007)33[381:PITLSF]2.0.CO;2

Ye, S., Covino, T. P., Sivapalan, M., Basu, N. B., Li, H.-Y., & Wang, S.-W. (2012). Dissolved

nutrient retention dynamics in river networks: A modeling investigation of transient

flows and scale effects. Water Resources Research, 48, W00J17.

doi:10.1029/2011WR010508

Zamyadi, A., Gallichand, J., & Duchemin, M. (2007). Comparison of methods for

estimating sediment and nitrogen loads from a small agricultural watershed. Canadian

Biosystems Engineering, 49, 1.27-1.36

Zhang, Q., Ball, W. P., & Moyer, D. L. (2016). Decadal-scale export of nitrogen,

phosphorus, and sediment from the Susquehanna River basin, USA: Analysis and

synthesis of temporal and spatial patterns. Science of the Total Environment, 563-564,

1016-1029. doi:10.1016/j.scitotenv.2016.03.104

56

APPENDIX A: EPA RISK ASSESSMENT GUIDANCE FOR SUPERFUND: VOLUME III - PART A, PROCESS FOR CONDUCTING PROBABILISTIC RISK ASSESSMENT

PRA – Probabilistic Risk Assessment

SMDP – Scientific Management Decision Point

RI/FS - Remedial Investigation/Feasibility Study

Source: U.S. EPA (2001).