Numerical modeling is increasingly used as a tool for environmental assessment and planning, including for Drinking Water Source Protection in Ontario as outlined in the Government of Ontario's 2006 Clean Water Act. However, modeling procedures are often inadequate in the organizational context and tight budgets. It remains a challenge to adapt these procedures such that they are transparent and efficient for watershed practitioners. This paper reviews and analyzes the application of the event-based approach, as defined in the technical rules to the ‘Clean Water Act’. Four limitations are then identified in a case study using the common procedure. Addressing these requires iterative model refinements, which likely result in cost overruns and undermine sound budget planning. An alternative method is then suggested, demonstrated and validated, which uses numerical modeling for creating a generic diagram. Such a generic diagram offers a transparent decision-making tool for planners, other non-technical employes and land owners.

Introduction

In the scientific community of hydrology, numerical modeling has become the standard method to assess environmental risks to surface water, such as contaminant spills. A number of procedures have been suggested to provide robust modeling procedures (e.g. Crout et al. 2008) and quantify uncertainty (e.g. Walker et al. 2003). However, there exists a lack of guidelines that also take into account procedural barriers within a real-world organizational context of watershed management (Arnold 2013), especially if multiple types of agencies collaborate (public institutions, private-sector consultants, academia). In such a context, only proper matching of modeling procedures with organizational constraints (capacity, funding, knowledge flow, timing) makes modeling a successful tool for governance.

This paper aims to bridge the academic quest for robust methods in water governance and the reality of watershed management tasks in public organizations. Within the context of Ontario's initiative to protect drinking water sources for municipal intakes and the passing of the Clean Water Act (2006), a model-intensive scientific assessment was used to identify risks to source water. Some approaches were straightforward, while the so-called ‘event-based approach’ (EBA) challenged watershed practitioners and only some regions have applied it. In this paper, we review the legal context and the application of the EBA in the first round of the assessment report, which is expected to be updated and extended regularly. The objective of this paper is to work toward a simple, model-derived diagram that supports watershed managers in using public dollars efficiently by reducing modeling expenses. A distance-to-intake/spill-quantity diagram (for a given maximum concentration level) suggests quantity ranges where model-based assessment is unnecessary, and also suggests ranges where only modeling can provide meaningful risk assessment.

Across the province, a sequential modeling approach was commonly applied in the first round of the Drinking Water Source Protection Assessment Report and also attempted by the author. The physical model components are presented below under ‘The physical flow model’. During the subsequent step of spill modeling, several limitations of the model were identified that lead to procedural problems. These are associated with the need to revisit assumptions repeatedly. The resulting iteration cycles of modeling lead to substantial cost increases. These are elaborated below under ‘Spill modeling – a screening approach’. Lastly, a new method is suggested that meets the regulatory requirements of the EBA and is scientifically sound, but also leads to a relatively simple result that avoids iteration cycles of modeling.

This paper makes two academic contributions: it uses the procedural perspective to assess a governance process that is based on numerical modeling. It also offers a new method to apply the EBA within Ontario's Drinking Water Source Protection initiative.

Source water protection in Ontario

Ontario's Clean Water Act (2006) helps communities to protect their municipal drinking water supplies by developing collaborative, locally driven, science-based protection plans. The ‘Act’ establishes a framework for the development and implementation of source protection plans across Ontario.

The scientific basis of source protection planning are watershed-based assessment reports that include a watershed characterization, water budget, municipal long-term water supply strategies, groundwater and/or surface water vulnerability analysis, threats assessment and issues evaluation, and risk assessment for water quality and quantity. This scientific basis forms the basis for a source protection plan that defines a strategy, policies and measures to reduce or eliminate threats to drinking water quality (DWQ).

Under the Clean Water Act (2006), land use activities are identified as significant threats to DWQ in five manners:

  1. If the quality of source water for a drinking water system has already deteriorated or is deteriorating, a DWQ issue is declared and contributing activities are significant DWQ threats.

  2. Land use activities can pose a significant risk to the water source of a drinking water system. Under the threats-based approach, activities that have a high hazard rating and are located within vulnerable areas are considered DWQ threats.

  3. Contamination from historic activities (called conditions).

  4. This paper deals with accident scenarios during extreme weather conditions. Under this EBA, land use activities are marked as a significant DWQ threat if they have the potential to interrupt the safe operation of a water supply, because a spill could be transported to an intake such that DWQ standards are exceeded (Ontario MOE 2009, Part IV.6). The technical rules define an extreme event as either a period of heavy precipitation or winds up to a 100-year storm event; a freshet; or a surface water body exceeding its high water mark (Ontario MOE 2009, Part I.1).

  5. Furthermore, the ‘Clean Water Act’ empowers the Ministry of the Environment (MOE) to designate Great Lake targets and related contaminations would also be considered a significant DWQ threat.

DWQ threats and the EBA

The EBA is reserved to drinking water systems that rely on ‘large’ surface water bodies, which are surface water intakes located in the Great Lakes, and other large lakes, connecting channels, and the Ottawa River as specified in Rule 68 (Ontario MOE 2009). To apply the EBA for assessing activities that may pose a threat to a drinking water system that is supplied by a large surface water body, three conditions must be fulfilled:

  • Condition 1. ‘An investigation must be requested’ by the source protection committee.

  • Condition 2. ‘Contaminant transport’ from the location of the spill to an intake (in accordance with Rule 68).

  • Condition 3. ‘Water quality deteriorates’ at the intake as a result of the spill (in accordance with Rule 130).

Unlike the issue-based approach and the threats-based approach, the application of the EBA is not prescribed in the technical rules and can be applied under the discretion of the Source Protection Committee, but once it was shown that contamination can occur then the activities must be identified as a significant DWQ threat and adequate policies must be developed. To date, the financial burden for a model-based test remains with the Source Protection Authority and Provincial support.

Rules 68 and 130 apply, respectively, to determine the contaminant transport condition and the water quality deterioration condition. Methodologically, rule 68 prescribes ‘modeling or other methods’ and Rule 130 prescribes ‘modeling’, or ‘another method used in accordance with rule 15.1’ (explicit approval by the director at the ministry). These rules grant some discretion to the source protection committee for choosing an adequate method. Notably, the sentence ‘if modeling shows’ leaves many options and choices on how to apply modeling. Furthermore, an intake protection zone 3 (IPZ-3) needs to be delineated, but there is little technical guidance on how to do that.

Implementing the EBA using four steps

A review of current practices for applying the EBA revealed that a great variety of interpretations of the technical rules were adapted by implementing agencies. The IPZ-3 has been interpreted as an area butting against intake protection zones 1 and 2 or as a connecting line between an activity and the intake (e.g. TRSPA 2012). If delineated as an area, the whole watershed was included according to condition 2 (e.g. LSRC 2012, section 8.4.1.3), or only the area that envelopes all activities which were shown to be a significant DWQ threat (e.g. Conservation Halton 2012). Between two and eight extreme event wind scenarios were evaluated for testing flow conditions. Contaminants were mostly modeled as generic conservative tracers, but some areas tested positive and negative buoyancy as well (e.g. LSRC 2012). Spill quantities and spill durations were generally defined using expert opinion. In some cases, the total storage quantity was assumed to be spilled (marina fuel storage). In other cases, historical spills were selected (e.g. tritium spill at Pickering, pipeline rupture at Enbridge). Validation data were available mainly for the Lake Ontario collaborative study, where the 1992 Pickering tritium spill offers a rich data base. In other areas, calibration and validation data were minimal or not available. Potential future activities are only considered in two reports (Nottawasaga Valley). Finally, many regions have postponed the application of the EBA to future assessment reports.

Apart from these differences, all regions have followed four work steps when applying the EBA:

  • Step 1. Prioritize spill scenarios (location, quantity and type of contaminant, duration of spill).

  • Step 2. Select set of storm conditions that reflect a 100-year extreme events.

  • Step 3. Physical modeling using 3D hydrodynamic models and storm conditions.

  • Step 4. Spill modeling using water quality models.

At the technical level, several assessment reports indicate the iterative nature of these work steps, because insights from latter steps may require revisiting earlier steps (e.g. Halton, Hamilton, Lake Simcoe, Nottawasaga). All regions which have intakes where the EBA is applicable have indicated in their reports that further modeling may be necessary in future assessment reports. One reason is that, if new land use activities may be developed, planners cannot assess from the existing information whether or not this activity would fall under source water policies. In other instances, spill modeling may be required if other types of storm scenarios occur or should be tested, and if changes in the risk classification of chemicals (MOE DWQ threat tables) or Ontario's DWQ standards widen the scope of existing policies. Any of such tasks leave the burden of proof with public watershed managers, who must access the knowledge embedded in the existing models and raise financial resources for a refined study.

Overview and modeling approach in this paper

The four common work steps for the EBA were also the basis for this study, which examines the Great Lakes intake of the City of Owen Sound (Figure 1). Corresponding to work step 1, screening of agricultural, industrial and commercial land uses was performed in order to identify potential spills and a total of 20 sites were identified. Contaminants of concern include gasoline and its additives that enhance the octane number and have high toxicity, chemical storage at the Sydenham Heights industrial area, and heavy oil associated with the harbor. Other potential contaminant sources are related to pathogens in storm water, sewage treatment and larger commercial holding tanks. These screening results are of a preliminary nature and remain undisclosed out of privacy considerations. For work steps 2 and 3, an existing physical model was available that was created for delineating a 2-h time of travel. Wind and runoff conditions were updated to reflect 100-year storm statistics (see below under ‘previous modeling work’). Then, spill modeling was performed building on the physical flow model.

Figure 1

Owen Sound harbor: bathymetry and location of drinking water intake and intake protection zone 1 (IPZ-1, dark line) and IPZ-2 (light line).

Figure 1

Owen Sound harbor: bathymetry and location of drinking water intake and intake protection zone 1 (IPZ-1, dark line) and IPZ-2 (light line).

Close modal

During the spill modeling phase, we experienced several limitations of the four work step approach. These are explained using a modeling case study of the Kenny Drain (Step 4a, see below under ‘Model update with 100-year storm conditions’). Then, a generalized screening approach is suggested and compared with other EBA studies that were performed in the Province (Step 4b, see below under ‘Spill modeling and limitations of the four-step modeling process’).

Previous modeling work

The numerical modeling study builds on the delineation of intake protection zone 2 (IPZ-2), which is the 2-h time of travel zone around a surface water intake. This earlier study was undertaken by the engineering firm Baird Coastal Engineering and implemented with the Delft3D hydrodynamic software by DeltaRes.

During this study, two hydrodynamic models were nested into each other. The larger West Georgian Bay Model (WGBM) represents the east shore of Georgian Bay using grid sizes from 5 to 1 km with 10 vertical sigma layers and 2 min time steps. Bathymetric data used to develop the model grids were extracted from 23 different data sets obtained from three sources: Canadian Hydrographic Service Field Sheets, Canadian Hydrographic Service Charts, and National Ocean Service hydrographic surveys from National Oceanic and Atmospheric Administration (NOAA). All depth values were adjusted to Chart Datum, which is equivalent to Low Water Datum at this site (Baird 2010). The WGBM is further embedded into the Lake Huron Operational Forecast System by NOAA, a department of the US government using open, Riemann type boundary conditions for currents and water levels (Baird 2012). All modeling for the present study was completed using mean lake level. Flow data were unavailable to calibrate or validate the WGBM; however, the calibration undertaken for the close-by Port Elgin intake for which some flow measurement data exist (Baird 2010) provides an improved level of confidence, because the same physical parameters were used. For more details pertaining to the model calibration, refer to Baird (2010).

The City of Owen Sound has 20,000 inhabitants and is located at the south end of the sound. A nested model with higher resolution captures the near-shore bathymetry around the intake of Owen Sound (Baird 2007, 2010). The area depicted in Figure 1 is 6.5 km long and covers 16.5 km2, with a maximum water depth of 25 m. The model uses a time step of 15 sec and has three open boundaries: the offshore boundary of the lake with water levels and currents (hydrodynamic conditions), and two rivers. Hydrodynamic conditions were extracted from the WGBM. The flow for the Sydenham River was established as 27 m3/s, corresponding to the 2-year return period flow. The Pottawatomi River is not gauged; flow was estimated based on its drainage area and established at 3 m3/s. The wind conditions for the simulation were extracted from the spatially-varying hourly wind data of the LHOFS (Baird 2007, 2010).

The modeling experiment highlights that flow conditions are strongly dependent on wind direction. Under south-west wind, surface water is blown out of the bay toward the east coast and, to a lesser extent, through near-shore currents of the west coast (Figure 2(a), full arrows). An undercurrent develops that pulls deep water from Georgian Bay through the deeper trench into the sound (dotted arrows). With north winds, surface and near-bottom currents are fairly in line with each other. Water is pressed into the bay from both the east and, with lesser force, the west shore and leaves the bay through a strong undercurrent at the center (Figure 2(b)). With east winds, surface water flows from the north-east into the bay and leaves toward the north-west (Figure 2(c)). At the lake bottom, currents form gyres that cycle within the tail end of the sound, moving water from the intake along the west coast. Based on the forward particle tracking results (Baird 2010), it is possible that contaminants from the Pottawatomi River, the Sydenham River or sewage treatment plant outfall could reach the intake during south or south­-west winds. Together, these diagrams highlight the variability of lake currents as well as their strong dependency on wind directions.

Figure 2

Flow directions from different wind forcing (adapted from IPZ-2 study, Baird 2010). Full arrows represent surface currents, dotted dark arrows deep currents. The dot marks the intake location.

Figure 2

Flow directions from different wind forcing (adapted from IPZ-2 study, Baird 2010). Full arrows represent surface currents, dotted dark arrows deep currents. The dot marks the intake location.

Close modal

Model update with 100-year storm conditions

To apply the event-based approach and delineate the IPZ-3, existing models were updated to reflect 100-year storm conditions. For these conditions, the medium-scale WGBM was updated, providing the boundary conditions for executing the nested Owen Sound model (Baird 2012). Driving forces that affect the potential transport of contaminants from the spill location to intakes in Georgian Bay are: (a) wind as the main force driving the circulation in these water bodies; and (b) tributary flows that transport contaminants from the watershed and upstream to the lake, where they may then be transported to the intakes.

The extreme events were selected based on a joint probability and persistence analysis of wind and tributary flows. Wind data recordings were taken from Wiarton Airport, the closest available station. Data were adjusted to 10 m elevation and converted from over land wind to over water wind after Hsu (1986). Flow station data are available for the Sydenham River in Owen Sound for 1949–2006. A joint probability analysis was undertaken to select modeling scenarios based on a combination of wind storm and tributary flow events (Baird 2012). Wind storms vary in duration from several hours to several days. The tributary flow data show that spring freshets may extend over a number of weeks.

Wind-driven currents are the primary mechanism by which a spill may be transported to an intake in Georgian Bay. At Owen Sound, storms from the NNE and SSW cause the critical conditions (Table 1). Historical wind events were selected from the time series data and the magnitude of the wind speed was scaled to equal the 100-year event (the identified wind represents the predominant direction during the event, though direction varied). For each event, the model was run for 14 days.

Table 1

Wind conditions for Owen Sound model (100-year storm)

NNE WindSSW Wind
Wind storm duration 111 h 43 h 
Wind speed 18.9 m/s 27.3 m/s 
NNE WindSSW Wind
Wind storm duration 111 h 43 h 
Wind speed 18.9 m/s 27.3 m/s 

Spill modeling and limitations of the four-step modeling process

For the model-based application of the EBA, our region initially followed a procedure based on the four work steps that lead to several limitations discussed further below.

For Step 1, a screening was performed where all commercial and industrial land uses were visited and activities determined. Based on this screening, 20 potential spill scenarios were formulated. Many of these are located in the industrial area of Owen Sound, called Sydenham Heights. Drainage is collected in a large storm sewer, the Kenny Drain, which discharges into the bay, at a distance of 800 m from the intake and within IPZ-1. Other potential spills could occur at the marina, the harbor, at a commercial park near the Pottawatomi River, and at the sewage treatment plant. Steps 2 (storm scenarios) and 3 (physical modeling) were described above under ‘The physical flow model’. For Step 4, contaminant spills were added to the flow vector field given by the physical model, each defined as the contaminant load of a spill (kg/s) and spill duration (s). During the modeling process, four limitations were identified that cast doubt on the viability of this approach.

Limitation 1: resolution of physical flow model and localized current regimes

After a typical rainfall event, Kenny Drain carries significant turbid water that is transported through a heavily armored drain, down the Niagara escarpment, and into Owen Sound Bay. The rapid flow velocity in the drain creates a plume with sufficient momentum such that it forms a jet that is carried into deeper waters in the bay and broadens where the water bottom slopes downward. There, the jet dissolves and the plume is carried along by prevalent lake currents.

A modeling study that follows the four-step work flow fails to reproduce such behavior of a plume, because the Kenny drain is below model resolution. If a spill is added at the location of the drain's outlet, then the plume immediately follows the flow vector field along the shoreline south (Figure 3, dark arrow), unlike the momentum-conserving jet that carries contaminants into deeper water (Figure 3, gray arrow). The source of this error is local model resolution. The model could be refined such that this flow is captured adequately, by: (a) increasing the resolution of the grid from its current size, 180 × 80 m, to the width of Kenny Drain, approximately 10 × 10 m; (2) adding part of the drain such that momentum is adequately reproduced; and (3) adding drainage from Sydenham Heights as an additional boundary condition. Furthermore, the time step would have to be increased from 15 s of the existing model to 1 s with the updated grid, the minimum of three restrictions suggested in the Delft3D user manual (DeltaRes 2011, p. 40).

Figure 3

Modeling study and observed turbidity plume at Kenny Drain.

Figure 3

Modeling study and observed turbidity plume at Kenny Drain.

Close modal

These modifications require resources that the watershed manager must provide. If this study is done in-house, these include staff capacity and time, hardware, and software licenses for data processing tools, the physical modeling software, and a software tool for the generation and manipulation of curvilinear grids (the latter costs €3,750). Alternatively, a consulting study can be commissioned to an engineering firm.

Limitation 2: resolution of the chemical model and unrealistic dilution effects

The model creates near-instant dilution of any spill, which can be attributed to the low resolution/large grid size of the water quality model. A contaminant is instantly mixed homogeneously within each grid cell, resulting in an evenly distributed and strongly diluted plume with low contaminant concentration in a large area. In contrast, observations show plumes with localized high concentrations and strong gradients, e.g. along the flanks and the front of the jet. For example, with the existing grid of 1 × 180 × 80 m3, a contaminant inflow of 1 kg/s would increase the contaminant concentration by 0.07 mg/L during each second, while the same inflow into the smaller grid of 1 × 10 × 10 m3 would cause an increase of 10 mg/L. Because EBA Condition 3 tests modeled concentration against a DWQ threshold, model sensitivity to grid size is problematic.

Addressing this limitation would require similar model modifications of the physical model as described under limitation 1, plus mixing parameters of the contaminant transport model.

Limitation 3: selection of wind scenario (direction and timing)

Another spill scenario assesses a gasoline spill within the marina, which is at a distance of 1.6 km from the drinking water intake and located within IPZ-2. Gasoline not only contains saturated and linear hydrocarbons, but also additives with high toxicity. Other regions across the province were especially concerned over the aromatic Benzene (2% in crude oil and up to 0.8% in gasoline; see Rahumathulla 2008). The Ontario Drinking Water Standard for this known carcinogen is 0.005 mg/L.

 Figure 4 shows the superposition of contaminant plumes that result from various wind scenarios, at the time step where the highest concentration at the intake location (triangle) is reached. Within the bay, three directions of contaminant transport and dispersion result from these wind scenarios: two plumes that follow the shallow shoreline at the East and West shore, and one plume that meanders through the center of the bay. The peak concentration of this meandering plume is well above the drinking water threshold of 0.005 mg/L and would identify an activity that can cause such spill as a significant DWQ threat under the EBA. However, under all given storm scenarios, the concentration peak of the plume misses the intake by 200 m and would not be a DWQ threat. Under slightly modified wind conditions, the modeled concentration would most likely exceed DWQ standards at the intake and elevate the spill to be a DWQ threat.

Figure 4

Contaminant plumes from hypothetical gasoline spill inside of marina. Figure shows overlay of plumes created with multiple wind scenarios.

Figure 4

Contaminant plumes from hypothetical gasoline spill inside of marina. Figure shows overlay of plumes created with multiple wind scenarios.

Close modal

To address this shortcoming, wind directions of the hypothetic storm scenario would have to be modified, and physical as well as chemical analysis would have to be repeated with these, until the plume hits the intake. This modification requires access to all modeling data and processing tools, capacity to perform these tasks, hardware, and software resources – by the water manager or an engineering consultant.

Limitation 4: burden of proof, fairness, and practicality

The fourth limitation stems from the obligation of the watershed manager to make fair, equitable and science-based decisions. Current law stipulates the burden of proof lies with the watershed manager, together with the obligation to cover the costs for performing numerical modeling studies. Owing to the iterative nature of the EBA modeling process, resource needs and costs are hard to anticipate. If given resources are used for assessing a fixed number of spill scenarios, the source protection committee must prioritize how limited and insufficient funds are attributed among many activities. Only those activities tested under the EBA may eventually be flagged as ‘significant’ DWQ threats that must comply with policies. Any prioritization decision is not free from political influence and could be perceived as unfair, undermining the credibility and legitimacy of source water protection.

Taking another perspective, decision-makers must decide how much resources from the overall budget can be spent for modeling. To make this modeling scientifically robust, an ever-increasing portion of the budget may be absorbed by modeling without any immediate benefits to the quality of drinking water. An opportunity cost assessment is advisable to compare returns from improved modeling with returns from other potential program measures, such as using this money immediately for supporting landowners to improve their practices on the ground.

In summary, four limitations exemplify how each of the four work steps can fail (see summary in Table 2). Because the technical rules put the burden of proof with the watershed manager, the manager has to come up with the necessary resources as well. Owing to the iterative nature of how the proposed work steps are structured, the manager is not in a position to cap the cost of modeling, without the risk that results are poorly defensible. If a landowner challenges these results (for example at the Ontario Municipal Board), then the defense of the model-based assessment would additionally require access to the modeling knowledge. Because it has been common practice that the modeling work is performed by external consultants in many regions, these litigations require additional consulting contracts by watershed managers and pose further risk for cost hikes.

Table 2

Four limitations to integrated modeling for the EBA

StepLimitationCause
Prioritization determined by resource constraints and political interventions, not by sound science 
Selected storm event(s) may create current pattern and meanders that miss the intake, even if this would happen with different wind directions 
The physical model may be insufficient to represent the flow direction of contaminant transport 
The chemical model may be insufficient to represent concentration peaks 
StepLimitationCause
Prioritization determined by resource constraints and political interventions, not by sound science 
Selected storm event(s) may create current pattern and meanders that miss the intake, even if this would happen with different wind directions 
The physical model may be insufficient to represent the flow direction of contaminant transport 
The chemical model may be insufficient to represent concentration peaks 

Rationale

The four work step method that was commonly applied during the first round of assessment reports was chosen for several reasons: as good scientific practice; to be compliant with the technical rules; to reflect procedural knowledge; and to respond to procedural constraints, such as timelines to meet milestones and funding availability through the Ontario Ministry of the Environment. In this section, an alternative method is presented that is equally based on physical modeling and contaminant transport modeling and in line with Rules 68 and 130. However, it avoids the need for iterative modeling and the results are more generic and far simpler from a planning perspective.

The method that is proposed in the next section addresses all four limitations that were encountered, using the following insights:

  • An infinite number of wind scenarios can exist that reflect the strengths, duration and runoff of a 100-year storm event. Because near-shore currents in the Great Lake are primarily determined by wind direction, an approach is preferable that reflects general flow characteristics rather than particular wind conditions.

  • It is not feasible to predict the exact directions of meandering plumes with a physical model, because of limitations in model resolution and due to implications from the theory of chaos and turbulent flows. Little relevance can thus be given to the predicted propagation of a meandering plume. Instead, only the typical distance of concentration maxima and dilution are relevant.

  • Below-grid concentration peaks and fronts are not resolved by any contaminant transport model. However, especially in locations close to the intake where model resolution may artificially overestimate contaminant dilution, conservative interpretation of results is required.

Final results are applicable to any location and spill (as will be shown later) so that prioritization is not necessary and iterative modeling is avoided.

Methodology

  • Nine locations were chosen in the vicinity of the intake where potential spills may occur. Some of these were identified during screening; others were added to increase representation of characteristic locations. The location, water depth and names of these test spills can be found in Figure 5.

  • The area where water depth resembles the intake depth was identified as the area between the gray and dark contour line (Figure 5). This area is called the ‘sensitive area’. In the example of Owen Sound, the depth was chosen between 11 and 12 m, reflecting the lake depth at the intake and the depth of the intake crib.

  • Eight spills of increasing contaminant quantities were modeled at each location: 36, 100, 360, 1,000, 3,600, 10,000, 36,000, and 100,000 L.

  • The Delft3D hydrodynamic modeling suite was used. In this package, physical and chemical modeling are two separate steps where the physical model computes currents and exports flow speed and direction as a vector field. The chemical model uses these outputs and subsequently computes chemical dispersion, with options to specify buoyancy, decay, sedimentation and re-suspension, particle interactions, evaporation, reactions, and more.

  • Numerical scheme 22 was applied, which uses a second order explicit transport scheme with flux correction after Boris & Book (1973). It avoids the introduction of local extrema by switching between a local explicit and implicit solution method (Delft3D Water Quality User Manual 2014, p. 324).

  • To reflect the lack of calibration data, the simplest parameterization of the contaminant and its transport was chosen. The contaminant was assumed as neutrally buoyant and conservative. Thus, no chemical reactions occur, no sedimentation or evaporation, and no re-suspension from the ground. This is warranted because contaminant maxima were generally reached in less than 12 h. Grids and time steps were chosen identical to the physical model. Otherwise, default parameters were used to specify the chemical transport model.

  • The model was run for each spill and location. Multiple spills were conducted simultaneously to save runtime.

  • Model results were analyzed and the maximum concentration within the water column was determined at any location and time, reducing the model output to a two dimensional time series.

  • The location within the sensitive area was determined where (1) a concentration threshold is reached and that (2) is most distant from the spill location. Several concentration thresholds were used (0.5, 0.1, 0.05, 0.01, and 0.005 mg/L).

Figure 5

Characteristic spill locations in the vicinity of the intake. Note the 11 m (gray) and 12 m (dark) contour lines.

Figure 5

Characteristic spill locations in the vicinity of the intake. Note the 11 m (gray) and 12 m (dark) contour lines.

Close modal

Results

A table was generated that specifies storm event scenario, spill location, spill quantity, and the maximum distance where each concentration threshold was exceeded within the sensitive area. Table 3 shows an extract from this table for a single spill location, the Kenny Drain (the full table can be accessed as an Appendix, available online at http://www.iwaponline.com/wqrjc/050/017.pdf).

Table 3

Maximum distance where concentration threshold within sensitive area is reached during model run (full table provided in the Appendix, online at http://www.iwaponline.com/wqrjc/050/017.pdf)

Distance to concentration threshold [km]
Storm scenarioSize of contaminant spill [kg]0.5 mg/L0.1 mg/L0.05 mg/L0.01 mg/L0.005 mg/L
North-east 36      
100     0.25 
360    0.5 0.9 
1,000   0.1 1.1 1.3 
3,600  0.5 0.9 2.3 2.3 
10,000 0.5 1.7 2.5 2.8 
36,000 1.3 2.6 2.7 4.3 10.3 
100,000 2.6 2.8 2.9 14.3 16.9 
South-west 36      
100    0.2 0.4 
360    0.6 1.3 
1,000   3.1 4.3 
3,600  0.5 2.2 5.2 9.3 
10,000 0.5 2.5 3.2 16.3 19.9 
36,000 3.2 6.2 7.7 23.2 
100,000 4.2 16.3 20.4 
Distance to concentration threshold [km]
Storm scenarioSize of contaminant spill [kg]0.5 mg/L0.1 mg/L0.05 mg/L0.01 mg/L0.005 mg/L
North-east 36      
100     0.25 
360    0.5 0.9 
1,000   0.1 1.1 1.3 
3,600  0.5 0.9 2.3 2.3 
10,000 0.5 1.7 2.5 2.8 
36,000 1.3 2.6 2.7 4.3 10.3 
100,000 2.6 2.8 2.9 14.3 16.9 
South-west 36      
100    0.2 0.4 
360    0.6 1.3 
1,000   3.1 4.3 
3,600  0.5 2.2 5.2 9.3 
10,000 0.5 2.5 3.2 16.3 19.9 
36,000 3.2 6.2 7.7 23.2 
100,000 4.2 16.3 20.4 
Results were further aggregated by taking the minimum/maximum and average distance from all spill locations and storm scenarios and then reducing the data set by two dimensions
formula
1
formula
2

In these equations, is the longest distance D under storm scenario E where the water concentration exceeds a DWQ threshold C within the sensitive area after a hypothetical spill {L, Q} with contaminant quantity Q at location L.

  • Equation (1) describes the ‘minimum’ longest distance for a given spill quantity Q and concentration threshold C, identified among all storm scenarios E and all spill locations L. This is the distance where modeling indicates that drinking water contamination can occur in all modeled scenarios. If a potential spill location with a contaminant quantity Q and a drinking water threshold C of that contaminant is closer to the intake than , then it is ‘almost certain’ that the spill can result in an exceedance of the concentration threshold at the intake. The activity that is closer to the intake than and can result in a spill of quantity Q of contaminant C would thus be considered a significant DWQ threat.

  • Equation (2) describes the ‘maximum’ longest distance where the contaminant plume after a spill Q and a concentration threshold C, among all storm scenarios E and all spill locations L. represents a ‘most conservative’ estimate of the distance between a spill location and the intake such that a DWQ threshold ‘could’ be exceeded during an extreme event. If an activity where a spill {Q, C} may occur is further away then , then it is NOT a significant DWQ threat.

  • To support decision-making, a third equation is defined that looks at a typical distance that a contaminant plume is transported within the harbor area, defined using the arithmetic mean
    formula
    3
  • Equation (3) describes a ‘typical’ distance at which a plume can exceed a concentration threshold C within the sensitive area.

For each concentration threshold C, generic quantity-distance diagrams can be drawn using Equations (1)–(3), plotting lines for distances , , and that, respectively, represent the highest, most probable and the lowest likelihood that a spill will exceed DWQ thresholds. The concentration threshold mg/L is exemplified in Figure 6.

Figure 6

Quantity-Distance diagram for concentration threshold 0.005 mg/L. Dots mark significant DWQ threats identified in other source protection regions for Benzene spills (Ontario drinking water quality standard 0.005 mg/L).

Figure 6

Quantity-Distance diagram for concentration threshold 0.005 mg/L. Dots mark significant DWQ threats identified in other source protection regions for Benzene spills (Ontario drinking water quality standard 0.005 mg/L).

Close modal

Watershed managers who need to classify the threat level of potential contaminant spills can easily determine whether a quantity falls outside of the area enclosed by Equations (1) and (2) by using distance and total spill quantity only. If a potential spill scenario falls above the line of Equation (2) (low risk), then the activity associated with a potential spills is not a significant threat. If the potential spill is below the bottom line defined by Equation (1), then it is very likely a significant drinking water threat under the EBA. In both cases, extensive numerical modeling is not necessary.

The area between these two equations reflects higher uncertainty. We distinguish two cases: the potential spill is below the mean travel contaminant transport distance that exceeds the DWQ standard, , but above the line of Equation (1) that is associated with a ‘high’ likelihood. This potential spill is likely a significant DWQ threat and could thus be identified such that source protection policies apply. Finally, the area below Equation (2) and above Equation (3) marks an area where the likelihood that a contaminant spill may cause a threshold exceedance such that further research is recommended.

For validation, no empirical data on spills are available for the Owen Sound intake on which this study is based. However, several drinking water source protection regions in the Province of Ontario have applied the EBA and identified significant DWQ threats in the vicinity of other intakes with different coastal topography.

Many of these threats are associated with spills of gasoline or other forms of petrochemicals. The contaminant that was chosen as the most relevant additive was Benzene in all regions, for which the Ontario Drinking Water Standard is 0.005 mg/L. From assessment reports across the Province, a total of 57 activities were identified to be ‘significant’ DWQ threats under the EBA due to their potential for large Benzene spills. For these, the assessment reports stated the location of the spill, the distance to the intake, the spill quantity, and for some of these also the concentration of the contaminant at the intake (see Table 4; see also Appendix, available online at http://www.iwaponline.com/wqrjc/050/017.pdf). A total of 57 spill scenarios from other regions were plotted into the diagram (Figure 6).

Table 4

Significant threats under the EBA, as of February 2012 (parameter of concern: Benzene, Ontario Drinking Water Standard 0.005 mg/L)

Intake nameSpill model scenarioDistance from spill [km]Concentration at intake [mg/L]Spill quantity [L]
Ajax Carruthers Creek pipeline break NA 260,000 
Duffins Creek pipeline break NA 
Highland Creek pipeline break 13 NA 
Petticoat Creek pipeline break 9.5 NA 
Rouge River pipeline break 10 NA 
Lynde Creek pipeline break NA 
Oshawa Creek pipeline break 14 0.4 
Bowmanville Bowmanville Creek pipeline break 260,000 
Graham Creek pipeline break 9.2 
Wilmot Creek pipeline break 7.6 3.3 
Newcastle Bowmanville Creek pipeline break 260,000 
Graham Creek pipeline break 1.5 NA 
Wilmot Creek pipeline break NA 
Oshawa Oshawa Creek pipeline break 1.4 260,000 
Whitby Carruthers Creek pipeline break NA 260,000 
Duffins Creek pipeline break NA 
Lynde Creek pipeline break NA 
Oshawa Creek pipeline break 0.32 
Lorne Park 16 Mile Creek pipeline break 10 0.012 260,000 
Credit River pipeline break 2.4 
Bulk storage spill, Oakville facility 14 1.25 
Small (mini-tank) spills 15 min duration  0.0068 
North York petroleum storage 13 0.078 
Lakeview Credit River pipeline break 0.37 260,000 
Humber River pipeline break 8.2 0.30 
Don River pipeline break 14 0.023 
Bulk storage spill, Oakville facility 20 0.5 
North York petroleum storage spill via Humber River 0.31 
R.L. Clark Credit River pipeline break 6.5 0.015 260,000 
Etobicoke Creek pipeline break NA 
Humber River pipeline break 0.079 
Don River pipeline break 15 0.32 
Bulk storage spill, Oakville facility 26 0.014 
North York petroleum storage spill via Humber River 0.55 
Toronto Island (shallow) Humber River pipeline break 0.4 260,000 
Don River pipeline break 
Toronto Island (deep) Humber River pipeline break 0.01 260,000 
Don River pipeline break 11 0.01 
North York petroleum storage spill via Humber River 0.015 
North York petroleum storage spill via Don River 11 0.009 
R. C. Harris Humber River pipeline break 20 0.1 260,000 
 Don River pipeline break 14 0.32  
 Highland Creek pipeline break 15 0.088  
 Rouge River pipeline break 20 0.045  
 Duffins Creek pipeline break 25 0.047  
 North York petroleum storage spill via Humber River 20 0.006  
 North York petroleum storage spill via Don River 14 0.059  
Burlington Intake Fuel handling and storage NA 15,000 
Conveyance oil/petroleum 14 NA 
Burloak Fuel handling and storage NA 15,000 
Conveyance oil/petroleum NA 
Collingwood Marina 1 fuel tank rupture 0.003 700 
Marina 2 fuel tank rupture NA 
Rope Subdivision Fuel spill from marina at south end Little Lake (M10) 1.3 0.06 1,500 
Victoria Harbor Fuel spill from marina at Port McNicoll (M8) 0.01 1,500 
Fuel spill from marina north of Port McNicoll (M6) 1.5 0.02 
Intake nameSpill model scenarioDistance from spill [km]Concentration at intake [mg/L]Spill quantity [L]
Ajax Carruthers Creek pipeline break NA 260,000 
Duffins Creek pipeline break NA 
Highland Creek pipeline break 13 NA 
Petticoat Creek pipeline break 9.5 NA 
Rouge River pipeline break 10 NA 
Lynde Creek pipeline break NA 
Oshawa Creek pipeline break 14 0.4 
Bowmanville Bowmanville Creek pipeline break 260,000 
Graham Creek pipeline break 9.2 
Wilmot Creek pipeline break 7.6 3.3 
Newcastle Bowmanville Creek pipeline break 260,000 
Graham Creek pipeline break 1.5 NA 
Wilmot Creek pipeline break NA 
Oshawa Oshawa Creek pipeline break 1.4 260,000 
Whitby Carruthers Creek pipeline break NA 260,000 
Duffins Creek pipeline break NA 
Lynde Creek pipeline break NA 
Oshawa Creek pipeline break 0.32 
Lorne Park 16 Mile Creek pipeline break 10 0.012 260,000 
Credit River pipeline break 2.4 
Bulk storage spill, Oakville facility 14 1.25 
Small (mini-tank) spills 15 min duration  0.0068 
North York petroleum storage 13 0.078 
Lakeview Credit River pipeline break 0.37 260,000 
Humber River pipeline break 8.2 0.30 
Don River pipeline break 14 0.023 
Bulk storage spill, Oakville facility 20 0.5 
North York petroleum storage spill via Humber River 0.31 
R.L. Clark Credit River pipeline break 6.5 0.015 260,000 
Etobicoke Creek pipeline break NA 
Humber River pipeline break 0.079 
Don River pipeline break 15 0.32 
Bulk storage spill, Oakville facility 26 0.014 
North York petroleum storage spill via Humber River 0.55 
Toronto Island (shallow) Humber River pipeline break 0.4 260,000 
Don River pipeline break 
Toronto Island (deep) Humber River pipeline break 0.01 260,000 
Don River pipeline break 11 0.01 
North York petroleum storage spill via Humber River 0.015 
North York petroleum storage spill via Don River 11 0.009 
R. C. Harris Humber River pipeline break 20 0.1 260,000 
 Don River pipeline break 14 0.32  
 Highland Creek pipeline break 15 0.088  
 Rouge River pipeline break 20 0.045  
 Duffins Creek pipeline break 25 0.047  
 North York petroleum storage spill via Humber River 20 0.006  
 North York petroleum storage spill via Don River 14 0.059  
Burlington Intake Fuel handling and storage NA 15,000 
Conveyance oil/petroleum 14 NA 
Burloak Fuel handling and storage NA 15,000 
Conveyance oil/petroleum NA 
Collingwood Marina 1 fuel tank rupture 0.003 700 
Marina 2 fuel tank rupture NA 
Rope Subdivision Fuel spill from marina at south end Little Lake (M10) 1.3 0.06 1,500 
Victoria Harbor Fuel spill from marina at Port McNicoll (M8) 0.01 1,500 
Fuel spill from marina north of Port McNicoll (M6) 1.5 0.02 

Even though these spill scenarios were identified around intakes with different coastal topography, the diagram derived for Owen Sound reflects the significance level surprisingly well. For example, 46 large pipeline spills were identified by the three regions of the Greater Toronto Area (Credit Valley, Central Lake Ontario, Toronto Region). The 46 spill scenarios were associated with the same pipeline at different locations, and with various intakes. Each spill consists of 260 metric tons of Benzene (assuming 2% of the oil) at a crossing of the pipeline with a creek or river, which drains into Lake Ontario (Figure 6), dots on the far right). All pipeline spill scenarios that were considered ‘significant’ DWQ threats fall beneath the limit (Equation (1)). Except for one smaller spill scenario, all other spill scenarios are also closer to the intake than the ‘average’ plume travel distance marked by Equation (3), which marks ‘typical’ significant threat likelihood. For one single spill that was identified as significant threat in another region using numerical modeling, the method proposed in this paper suggests ‘maybe significant’. Here, the proposed screening method lacks specificity and recommends that the watershed manager must address uncertainty, either by advanced modeling or by applying the precautionary principle.

In summary, this study performed within the Owen Sound harbor and intake area cannot be validated with empirical data. However, the simple model-based diagram predicts all other (available) significant threats that were identified under the EBA very well. Not a single significant threat was miss-characterized by the diagram. Furthermore, only in one spill scenario would the screening method be insufficient and recommend further research. The simple diagram predicts all other modeling study results correctly, which would have saved watershed managers across the province several million dollars invested into modeling.

Unfortunately, data are not available for those spill scenarios that were tested in other regions and did ‘not’ meet the significant threat threshold. Access to these data would add additional confidence in the validity of our model diagram.

This paper summarized how the EBA was applied for the identification of ‘significant’ DWQ threats in Great Lake intakes. While there are considerable methodological differences among regions, there are also commonalities, especially the four basic work steps that lead to an iterative numerical modeling problem which arguably is hard to implement and can easily cause cost explosion. Instead, this paper offers an alternative approach that provides government agencies with a simple tool to assess the risk potential of activities where chemicals are handled that may put DWQ at risk. The only information required is the substance, the location of the spill and its distance to the intake, the DWQ standard for the substance, and the quantity of a potential spill. With the proposed method, the agency is enabled to classify a proposed land use into four categories without additional costs: (1) very likely, (2) likely, (3) maybe, or (4) not likely a significant DWQ threat. In the examples from other regions suggested for validation, 56 out of 57 scenarios would immediately be classified as ‘very likely’ or ‘likely’ significant, while only one scenario remains in the unclear category 3 where further modeling may be advisable.

Several technical improvements of the modeling remain possible and should be addressed in further research. First, even though storm scenarios were chosen based on expert judgment that was informed by past modeling, additional storm and extreme event scenarios can generalize results further. It is expected that the spread between Equations (1) and (2) would widen with more storm scenarios, while the average or ‘typical’ line, which is recommended as the cutoff for the identification of significant DWQ threats, may also change. Second, model validation with flow and contaminant transport measurements, e.g. turbidity events at an intake, would reduce model uncertainty. Because such empirical data tend to be available only for normal weather conditions, the applicability of such a validation exercise under storm event conditions remains unclear. Third, the suggested methodology is not capable of dealing with small spills where model resolution causes artificial dilution of the contaminant plume (limitation 2). Additional modeling with high-resolution models that capture plume characteristics (fronts, vertical profiles) is recommended. Fourth, a larger model area would prevent boundary effects, where the plume reaches the edge of the model grid, especially for low concentration thresholds and large spills. In addition to improvements of the model, comparative analysis of multiple harbors is recommended to determine similarities and differences between their D(Q) diagrams, as a function of local topography, wind exposure and intake depth.

On the other hand, the proposed method offers a simple diagram that supports watershed managers and planners with a clear basis for decisions at minimal costs. Even land owners who investigate the financial prospects of future activities are empowered, because this methodology increases the transparency of when and how drinking water protection policies under the EBA can impact their business. The proposed methodology could be corroborated further using more recent IPZ3 studies, by testing whether the graph predicts the significance level of other potential spills adequately.

The risk level of most combinations of distance D from the intake and potential spill quantity Q of a contaminant C can be immediately determined as ‘significant’ or ‘not significant’ with the proposed methodology. However, it cannot resolve the uncertainty associated with the combinations that fall between Equations (1) and (3) (labeled ‘maybe significant’ in Figure 6). Here, the watershed manager has three options: The precautionary principle can be applied and the activity is deemed ‘significant threat’ unless proven otherwise. Then, the burden of proof may be reversed, such that it becomes the land owner's responsibility to prove that an activity is ‘not’ a significant threat. Alternatively, the watershed manager may deem the activity safe, unless a third party proves it to be a potential threat under the EBA. In this case, the watershed manager can avoid additional costs. The third option is that the watershed manager secures funding to reduce uncertainty further, hoping for clearer results. Regardless, uncertainty remains part of any decision-making based on numerical models and requires human decision-making.

From a watershed management perspective, numerical modeling is increasingly used for decisions that have legal implications and impact the economic decisions of land owners. While such use of scientific tools can improve the robustness of watershed planning, the limitations identified in this paper with respect to the EBA also underline the need to fit modeling processes into the operational context of watershed management organizations. This paper discusses Ontario's model-based, EBA for identifying significant drinking water threats, and derives a simple diagram. A generic approach is suggested that is model-based and meets the requirements laid out in Rules 68 and 130 of the technical rules to Ontario's ‘Clean Water Act’.

We greatly appreciate the encouragement of Don Smith, the review of David Ellingwood, and the financial support of the Province of Ontario.

Baird
2007
Surface Water Source Protection Technical Studies for Saugeen Valley Conservation Authority. Intake Protection Zone Delineation for Georgian Bay Intakes
.
Technical report prepared for Stantec Inc., 26 March 2007 (available through Saugeen Valley Conservation Authority)
.
Baird
2010
In-water IPZ-2 Delineation for Saugeen, Grey Sauble, Northern Bruce Peninsula Phase II Studies
.
Final report prepared for Stantec Inc., 8 December 2010 (available through Saugeen Valley Conservation Authority)
.
Baird
2012
Numerical Modelling in Support of IPZ-3 Delineation. Saugeen, Grey Sauble, Northern Bruce Peninsula Source Protection Region. 9 February 2012 (available through Saugeen Valley Conservation Authority)
.
Boris
J.
Book
D.
1973
Flux corrected transport I. SHASTA, a fluid transport algorithm that works
.
Journal of Computational Physics
11
,
38
96
.
220
.
Clean Water Act 2006 Government of Ontario, S. O.
2006
.
Conservation Halton
2012
Assessment Reports for the Halton Region and Hamilton Region Source Protection Areas
. .
Crout
N.
Kokkonen
T.
Jakeman
A. J.
Norton
J. P.
Newham
L. T. H.
Anderson
R.
Assaf
H.
Croke
B. F. W.
Gaber
N.
Gibbons
J.
Holzworth
D.
Mysiak
J.
Reichl
J.
Seppelt
R.
Wagener
T.
Whitfield
P.
2008
Good Modelling Practice
.
Developments in Integrated Environmental Assessment
3
,
15
31
.
DeltaRes
2011
Delft3D-WAQ. Versatile Water Quality Modelling in 1D, 2D or 3D Systems Including Physical, (Bio) Chemical and Biological Processes. User Manual
.
WL/Delft Hydraulics
,
Delft
,
The Netherlands
,
4.03 edn
.
DeltaRes
2014
D-Water Quality – Water quality and aquatic ecology modelling suite. User Manual for D-Water Quality
.
Version 4.99.34158, 28 May 2014. DeltaRes, Delft, The Netherlands. Retrieved from http://oss.deltares.nl/web/delft3d/manuals on 29 July 2015
.
LSRC
2012
Lakes Simcoe and Couchiching Black Severn Source Protection Area Assessment Report. South Georgian Bay Lake Simcoe Source Protection Region
. .
Ontario MOE
2009
Technical Rules: Assessment Report. Clean Water Act, 2006
.
EBR Posting Number EBRO10-7573, 16 November 2009. Ontario Ministry of the Environment
.
Rahumathulla
R.
2008
Benzene in Canadian Gasoline: Effect of the Benzene in Gasoline Regulations
.
2008 Annual Report. Environment Canada 2008
.
TRSPA
2012
Approved Updated Assessment Report. Toronto and Region Source Protection Area
. .
Walker
W.
Harremoes
P.
Rotmans
J.
der Sluijs
J. V.
Asselt
M. V.
Janssen
P.
von Krauss
M. K.
2003
Defining uncertainty: a conceptual basis for uncertainty management in model-based decision support
.
Integrated Assessment
1
,
5
17
.

Supplementary data