Abstract

On March 24, 2017, a fuel spill from a partially submerged barge in Picton Bay contaminated the source water for the drinking water supply of the local township. Immediately after the spill, management decisions regarding the water intake plant operations were made based on contaminant observations and projected wind conditions. From a management perspective, it is essential to understand all the dynamical forcing for a system to direct the best decision-making but, unfortunately, there are no historical observations of currents in Picton Bay or any in-depth numerical modelling studies that have established the circulation patterns or hydrodynamics of the bay. This paper presents observations of surface speeds and drifter pathways collected using Lagrangian drifters and compares the observations to the velocity field estimates from a wind forced three-dimensional hydrodynamic model. Surface drifters were deployed from July to September and moved southwest into the bay during each deployment with almost no tendency to turn and drift out of the bay. Model simulations indicated that currents in the bay are sensitive to small-scale local winds and that a boundary current exists that connects the spill site to the area of the water intake pipes in wind conditions that are to the southwest or southeast.

INTRODUCTION

In the spring of 2017, a fuel spill from a partially submerged barge put the drinking water supply for two towns in southern Ontario at risk of contamination. Picton Bay, at the entrance to the Bay of Quinte, is used as the source for drinking water for Picton and Bloomfield, Ontario. The bay is situated on the southwestern side of a wide triangular passage where water from the Bay of Quinte mixes with that of Lake Ontario (Freeman & Prinsenberg 1986; Minns et al. 1986) (Figure 1). There is a shipping port located at the entrance to Picton Bay that provides a terminal for import and export of materials including road salts, farm fertilizers and steel products and at which a barge partially sank on March 24, 2017. To meet the policy set out by Ontario's Clean Water Act, intake Protection Zones (IPZ) have been established in Picton Bay which identify vulnerable areas around the water supply intake pipes (Quinte Region Source Protection Committee 2014). The shipping terminal is outside of IPZ-2, which is defined by a region having a minimum of 2 hours of travel time for any contaminant to the water intake pipes. To define this region, conservative estimates of maximum surface contaminant flow speeds of about 20 cm/s were used. The sinking of the barge released gasoline into the bay and post-spill water samples were taken by the Ontario Ministry of Environment and Climate Change. Anecdotal reports showed remnants of the fuel spill had crossed the bay within days of the spill, and were evident north of the spill site. Municipal management decisions were made based on contaminant observations and projected wind conditions, and Picton water intake plant operations were shut down 5 days after the spill following reports of an oil-like smell being evident in the water arriving from the intake pipes. The municipal managers relied on forecasted wind directions to estimate the potential movement of the fuel, however it is unclear whether wind conditions are the primary driving force for the bay's dynamics. This situation highlights the larger problem of the lack of support for managers of smaller municipalities and the resulting sparse in situ observations of water movement and transport pathways in important bodies of water. The circulation within Picton Bay is affected by wind and solar forcing, water level changes at the entrance to the bay, runoff, and boundary currents at the shoreline, yet there are no in situ current moorings in the bay, nor are there historical observations of surface currents by which an environmental assessment could be made for that spill or any potential future fuel spills or seepage events from the terminal.

Figure 1

Picton Bay is a small embayment at the mouth of the Bay of Quinte where Quinte water mixes with Lake Ontario water via the Glenora Gap. There is a shipping terminal (marked with an x) at the north end of Picton Bay. Small inset shows the z-shaped Bay of Quinte with Picton and Bloomfield (marked with a dot) and middle inset shows the bay in relation to Longreach and the Glenora Gap. Observed drifter pathways for all drifter releases are shown (release locations marked with a diamond). Locations of the 25 May 2017 fuel spill and drinking water intake pipes for the Picton water treatment plant are marked with a plus.

Figure 1

Picton Bay is a small embayment at the mouth of the Bay of Quinte where Quinte water mixes with Lake Ontario water via the Glenora Gap. There is a shipping terminal (marked with an x) at the north end of Picton Bay. Small inset shows the z-shaped Bay of Quinte with Picton and Bloomfield (marked with a dot) and middle inset shows the bay in relation to Longreach and the Glenora Gap. Observed drifter pathways for all drifter releases are shown (release locations marked with a diamond). Locations of the 25 May 2017 fuel spill and drinking water intake pipes for the Picton water treatment plant are marked with a plus.

The Bay of Quinte is a 60 km long z-shaped embayment that is flushed largely by the inflow of water from the Trent River located at its head. Due to its classification as a UN Area of Concern in 1986, the temperature, oxygen and nutrient loads of the bay have been continuously monitored for more than 20 years. Much of the ensuing research and policy decisions have focused on the mitigation of frequently occurring eutrophication events and, as a result, phosphorus loading in the bay has been significantly reduced. Nevertheless, the bay continues to be at risk from several factors including increased land use and climate change (e.g. Minns et al. 2011b; Arhonditsis et al. 2016; Croft-White et al. 2017). Maximum surface water temperatures in the bay have increased at rates consistent with climate warming rates and the increased heat is being retained above the thermocline which, as a consequence, suppresses vertical mixing (Minns et al. 2011a). The upper Bay of Quinte is sufficiently shallow such that the water temperature profile is unstratified throughout the year, however the lower bay water in Picton Bay and around the Glenora Gap (Figure 1) is known to stratify as a result of the seasonal surface heatflux forcing. The average depth of Picton Bay is about 8 m and it is stratified in the early summer, but summer heating forces the thermocline to the bottom and deeper than the sill at Glenora (Freeman & Prinsenberg 1986; Minns et al. 2011a).

Lake Ontario, into which the Bay of Quinte empties, is also stratified in summer and upwelling occurs along the north shore of Lake Ontario when the wind is from the west (e.g. Miners et al. 2002; Rao et al. 2003). The small size of Picton Bay, with its long axis aligned in almost a north-south orientation, prevents large scale geostrophic upwelling and downwelling events but suggests that local wind events may be important in driving circulation. Local diurnal winds are known to affect the temperature response of small coastal embayments and a wind blowing to the south-southwest would have a fetch of at least 10 km (Walter et al. 2017). Given these factors, it is clear the hydrography of Picton Bay varies temporally and can have a baroclinic response.

Notwithstanding the number of studies of the ecosystem dynamics of the Bay of Quinte, there are few studies that provide water current estimates with good spatial resolution and that also include variability at a monthly or seasonal timescale. Freeman & Prinsenberg (1986) provide current observations collected in the summers of 1976 and 1978 for the Glenora Gap. They observed that Lake Ontario waters frequently backflow through the gap at depth, with currents up to about 10 cm/s. Using a steady-state two-layer flow model, they linked these currents to the thermocline tilt through the gap that sets up as a result of summer heating. Backflow at 10 m depth is strong in the early summer period at a time when surface flows driven by upstream river runoff are strongly out of the bay. The two-layer flow can reverse if the thermocline continues to deepen past the sill depth at Glenora. Freeman & Prinsenberg (1986) also noted a second flow feature at the Glenora Gap governed by free oscillations set up in the channel east of the gap. These oscillations were in evidence as a 5.2 h periodic flow reversal in the current meter records. The observations taken at the Glenora Gap are currently the only available velocity dataset for the Bay of Quinte, and it is unclear whether the flows at the gap are representative of the flow regime in Picton Bay itself.

Oveisy et al. (2015) provided the first high resolution simulation of the Bay of Quinte using the Estuary and Lake Computer Model (ELCOM). The model was forced with meteorological data from 2004 and was used to simulate the hydrography to provide estimates including averaged velocities and flushing times that are pertinent to the understanding of nutrient loading. Model velocity maps and simulated particle tracks show that the mean current tends to flow in a connected stream from Longreach to the Gap bypassing Picton Bay and the model estimated seasonally averaged currents to be about 0.5 cm/s within the bay. Shore et al. (2016) simulated the ice-free periods in the Bay of Quinte for the years 1979–2006 using the Finite Volume Community Model (FVCOM). The model was used to produce seasonal averages of the surface flow for the bay and produced similar flow speeds to the Oveisy et al. (2015) results. The Freeman & Prinsenberg (1986) instantaneous flow observations are predictably larger compared to the averaged results from either model. Until now, no direct observations of water speeds or transport pathways have been made within Picton Bay and the resolution of both models within the bay is inadequate to provide realistic estimates.

The shipping terminal receives and stores road salts which could provide a potential point source for chlorides in Picton Bay. The introduction of these contaminants into the bay via rainfall events could significantly reduce the quality of the drinking water as chlorides are persistent and can be difficult to treat (McCullough et al. 2012; Dailey et al. 2014; St-Hilaire et al. 2015). Environment and Climate Change Canada (ECCC) recommends identifying salt vulnerable areas and creating appropriate management plans, especially for source water areas like Picton Bay (Stone & Marsalek 2011; Environment and Climate Change Canada 2012; Betts et al. 2014).

Given the occurrence of the fuel spill and the presence of the salt piles at the entrance to Picton Bay, it would be beneficial for the implementation of management policy to have knowledge of the local circulation dynamics based on observational data. This would also provide context for the local behavior of any chlorides or other contaminants present in the water. The speeds presented in this study add to the small number of observed velocities in the Bay of Quinte and are the first in situ observations for Picton Bay. The previous modelling efforts had such course resolution in Picton Bay as to effectively ignore it. A new FVCOM model focused specifically on Picton Bay was implemented and forced with hourly winds from Point Petre. Point Petre is the nearest weather station that provides observed wind data with sufficient coverage, however, it is located 20 km to the south and is situated in terrain unlike the somewhat sheltered terrain of Picton Bay. The observed drifter speeds and tracks are compared to simulated particle pathways derived from the velocity fields. Results provide a better understanding of the circulation in the bay and around the water intake pipes which could lead to better management decision-making tools or identify sites for future water intake pipes if water supply demands increase.

METHODS

This study describes observations of surface currents in Picton Bay that were collected by Lagrangian surface drifters in the summer of 2017 and comparative results of a wind forced numerical model. Lagrangian drifters move with the currents and have long been used to collect observations of circulation in a variety of water environments (Davis 1991; Lumpkin et al. 2017). In the summer of 2017, surface drifters designed to be suitable for the low current speeds and minimal wave action found in the Bay of Quinte were deployed in Picton Bay. Surface drifters were used that were low-cost, re-useable and had a small enough footprint to not obstruct the heavy boat traffic in the bay. The drifters sampled the water speeds down to a depth of about 1.5 m and were outfitted with satellite tracked GPS units and position data were recorded every 2.5 minutes. The drifters used a drogue with four vanes similar to a CODE drifter and had minimal wind slip (Davis 1985; Niiler & Paduan 1995).

The surface drifters were deployed in two wind conditions: to the east, representative of the average westerly wind state for southern Ontario; or aligned with the long axis of Picton Bay. Hourly wind data from Point Petre, the closest ECCC monitored station, were analyzed for the deployments. Almost half of the drifters were released near the fuel spill site to sample the circulation in that location while the rest were released to sample other parts of the bay. Because the drifters updated their position every few minutes via satellite, the drifters were deployed and left unmonitored to be retrieved at a later time, generally when they became trapped along the shoreline. Surface drifters were also released at the Glenora Gap site on the same day to determine the flow speed and direction of the Bay of Quinte outflow (Table 1). Drifter location data were converted to speed using a three-point centered difference scheme and quality controlled via the removal of data having atypical accelerations. Using these methods, the velocity estimates are expected to be accurate to about 1 cm/s (De Dominicis et al. 2012).

Table 1

Release date, ID, location and duration of the release of the individual drifters shown in Figure 1 

Date released (2017) ID Release location (lat/lon) Deployment duration (hours) Average deployment wind speed (m/s) 
July 11 Jul 11 44.0376, –77.1257 2.9 2.7 
July 25 Jul 25A 44.0378, –77.1269 24.1 2.9 
Jul 25B 44.0385, –77.1262 10.2 4.2 
July 31 Jul 31 44.0361, –77.1274 45.4 3.9 
August 8 Aug 8 44.0268, –77.1274 5.3 5.8 
September 16 Sep 16A 44.0345, –77.1275 1.9 2.9 
Sep 16B 44.0335, –77.1259 2.6 2.9 
Sep 16C 44.0328, –77.1248 2.4 2.9 
Sep 16D 44.0316, –77.1227 1.8 2.9 
Sep 16E 44.0321, –77.1235 2.6 2.9 
Date released (2017) ID Release location (lat/lon) Deployment duration (hours) Average deployment wind speed (m/s) 
July 11 Jul 11 44.0376, –77.1257 2.9 2.7 
July 25 Jul 25A 44.0378, –77.1269 24.1 2.9 
Jul 25B 44.0385, –77.1262 10.2 4.2 
July 31 Jul 31 44.0361, –77.1274 45.4 3.9 
August 8 Aug 8 44.0268, –77.1274 5.3 5.8 
September 16 Sep 16A 44.0345, –77.1275 1.9 2.9 
Sep 16B 44.0335, –77.1259 2.6 2.9 
Sep 16C 44.0328, –77.1248 2.4 2.9 
Sep 16D 44.0316, –77.1227 1.8 2.9 
Sep 16E 44.0321, –77.1235 2.6 2.9 

Average wind speeds are calculated using Environment and Climate Change Canada station at Point Petre averaged over the duration of the deployment.

A numerical model (FVCOM) was then used to simulate the flow in Picton Bay during the drifter deployments. The FVCOM model uses an unstructured mesh to minimize computational costs and is effective in domains that have complex geometries. The model prognostically solves the 3D primitive equations where mass and momentum fluxes are conserved on unit control volumes (Chen et al. 2007). It has been successfully applied to a variety of large and small scale domains and small embayments (e.g. Anderson et al. 2010; Ma et al. 2012; Bai et al. 2013; Xue et al. 2015). The Picton Bay model domain was simulated using bathymetric data by the National Geophysical Data Center (1999) with an open boundary across the northern end of the bay. The location of this open boundary allowed sensitivity tests to varying boundary inflows to be investigated. Barotropic runs of constant temperature (15 °C) and salinity (0 ppt) were simulated for each of the drifter deployments days using hourly winds from the ECCC weather station at Point Petre. The model used 11 uniformly spaced vertical sigma levels. The 3D current fields were then saved every 5 minutes and used to subsequently compute the resulting particle pathways. Particle positions were solved iteratively forward in time from their initial position using a 5 minute time step. The particles are constrained to stay at the surface, have no density, and are used to simulate Lagrangian pathways in the current field.

RESULTS AND DISCUSSION

Lagrangian surface drifters were released in Picton Bay after the fuel spill on March 24, 2017, four near the fuel spill site, one in the middle of the bay further to the south, and five in a line cutting across the bay (Figure 1). Position data were saved approximately every 2.5 minutes with individual deployments lasting between 2 hours and 2 days (Table 1).

Surface speeds for all of the drifter deployments range from about 0.5 to 11 cm/s (Figure 2). The average speeds and variability were observed to change monthly. Speeds mid-summer from late-July through August are smaller, consistent with a reduction in summer wind speeds. The highest speeds were in the spring and fall (July 11 and September 16) averaging above 9 cm/s. The longest deployment, 45.4 hours on July 31, showed very little variability in the observed speed 2 ± 1 cm/s whereas 1 week later (August 8), speeds had the largest variability of about 3 cm/s in a 5.3 hour period. Two days illustrate the spatial differences that are present in the current field. The two deployments that took place on July 25 had about a 4 cm/s difference in mean speeds with the faster speeds being observed farther from shore and on September 16, average speeds for four of the drifters were consistently around 6–7 cm/s but the westernmost drifter had significantly higher average speed of around 9.3 cm/s. The average observed speeds are all consistent with generalized averages from previous model studies as well as the expected reduction of speeds in the mid-summer (Oveisy et al. 2015; Shore et al. 2016).

Figure 2

Statistical values for all speeds (cm/s) measured by each drifter.

Figure 2

Statistical values for all speeds (cm/s) measured by each drifter.

All the observed drifter pathways are shown in Figure 1 relative to the fuel spill site (marked with an +). Wind roses for each deployment are provided in Figure 3 and it can be seen that the drifters were released in one of two wind conditions, either to the east or aligned with the long axis of Picton Bay. In addition, each individual speed observation (available as frequently as every 2.5 minutes) is shown for each drifter (Figure 4). The two drifters released on July 25 were in a wind field that was directed along-shore, while the other drifter releases were in wind fields generally to the east. However, regardless of the direction of the observed wind field at Point Petre, all drifters tended to move to the southwest away from their release site (marked with symbols). Many of the drifters (on July 11 and 25, August 8 and September 16) moved to the southwest and became trapped at the shoreline, however the July 31 deployment, having the slowest average speed and the longest deployment, followed the shoreline down to the region of the water intake pipes. The September 16 deployment released drifters in a line cutting across the width of the bay to capture any spatial differences in the central part of the bay. These drifters remained deployed for 4 hours and tracked to the southwest with very little dispersion and no evidence of recirculation. The general behavior seen for all the drifters for all five deployments over the five month period was to travel southward into Picton Bay both when winds were directed along the axis or to the east.

Figure 3

Wind roses (m/s) for ECCC wind station Point Petre for the date and duration of each deployment (Table 1). Direction specifies direction toward which the wind blows.

Figure 3

Wind roses (m/s) for ECCC wind station Point Petre for the date and duration of each deployment (Table 1). Direction specifies direction toward which the wind blows.

Figure 4

Individual observations of drifter speed (cm/s) shown at each sampled location along the drifter track for each deployment. Distances (m) are relative to the fuel spill site (marked with an x and located at the point (0,0)).

Figure 4

Individual observations of drifter speed (cm/s) shown at each sampled location along the drifter track for each deployment. Distances (m) are relative to the fuel spill site (marked with an x and located at the point (0,0)).

The two drifters released on July 25 were the only drifters released in a wind field not directed toward the east. In this case, the wind was originally about 4.2 m/s to the southwest but then slackened and switched after 13 hours to be about 2.9 m/s to the northeast. Neither drifter travelled farther than about 1 km from its release point near the shipping terminal and while the Jul 25A drifter experienced very small speeds, the Jul 25B drifter, which was further off shore, was consistently faster by almost 4 cm/s. Both these drifters and the July 11 drifter appeared to be caught in a boundary flow and became trapped on the shoreline southwest of the terminal. The average July 11 winds were to the east-northeast at about 2.7 m/s, while the drifter moved to the southwest at speeds between about 8 and 10 cm/s. Note that the GPS signal was frequently lost for this drifter (and Jul 25B) and data are missing, especially near the end of the deployment.

The behavior of the drifter on July 31 is also indicative of a possible boundary flow along the western side of the bay. Even though the average winds at Point Petre were 3.9 m/s to the east-northeast and the drifter initially moved almost 500 m northeast in 3.5 hours, it eventually turned back to its release point and continued to the southwest for over 40 hours.

The drifters deployed in the middle of the bay on August 8 and September 16 indicate a further complexity of the observed flow field. All drifters moved toward the southwest though the wind field observed at Point Petre was generally to the east. Both deployment times were short, less than 6 hours on August 8 (2 hours on September 16) and yet both showed high variability, up to 4 cm/s along the particle track on August 8 and between the different drifters on September 16.

In summary, all drifters ended up travelling toward the head of the bay, with almost no tendency to recirculate and seemingly not with the prevailing wind direction as measured at Point Petre. This drifter movement, which is not in a direction consistent with wind forcing and the temporal variation of the drifter speeds along the track, indicates that the currents in the bay are more likely influenced by local topographically steered winds or changes in forcing at the mouth of the bay. Additionally, no oscillatory behavior was observed similar to the 5.2 hour behavior at Glenora Gap.

To examine the drifter pathways in the bay, FVCOM was used to estimate drifter pathways for each of the five deployments. Point Petre wind observations were used to force the model for two cases: one using the observed wind speeds and directions and one with the observed speeds but with a wind direction constantly to the southwest (parallel to the western shoreline). FVCOM current fields were output and simulated particle tracks were then computed using the release locations and path duration for each drifter (Table 1). Figure 5 shows the comparative results for both wind forced cases.

Figure 5

Comparison of the observed drifter tracks to model simulations in two wind conditions: wind speeds and directions as observed at Point Petre and wind speeds as observed at Point Petre directed to the southwest parallel to the western shoreline. Results of two wind conditions versus observed paths for (a) and (b) July 25 and (c) and (d) all other deployments.

Figure 5

Comparison of the observed drifter tracks to model simulations in two wind conditions: wind speeds and directions as observed at Point Petre and wind speeds as observed at Point Petre directed to the southwest parallel to the western shoreline. Results of two wind conditions versus observed paths for (a) and (b) July 25 and (c) and (d) all other deployments.

Results for the drifters released on July 25 show almost no difference between the real case and the case with the constant wind direction to the southwest because winds on those days were already aligned southwest-northeast and drifters became trapped along the shoreline in both simulated wind field cases (Figure 5(a) and 5(b)).

Drifter simulations using the observed wind forcing for all other deployments (Figure 5(c)) show that the drifters do not travel to the east as expected given the Point Petre wind directions. When wind directions are replaced with a wind to the southwest, the model predicts pathways very similar to the observed paths. The modeled paths using the observed wind directions predict that all drifters travel eastward across the bay, with the July 31 and August 8 drifters turning to the north close to the shoreline. Because the July 31 deployment was the longest, the modeled drifter had time to leave the bay entirely. In comparison, when enforcing a strict southwest wind direction, the modeled paths closely match the observed paths for all the drifters. The July 31 and August 8 modeled paths travel farthest into the bay before eventually becoming trapped on the shoreline near the water intake pipes. The September 16 modeled drifter paths show improvement as well with the directions of the drifters being well simulated though the modeled drifters pathways are shorter. Sensitivity tests showed a southwest wind at 120°, i.e. parallel to the shoreline near the terminal, best reproduced the pathway trajectories.

To examine the surface velocity field, the model was forced with a constant wind speed of 4 m/s with winds parallel to the shoreline at ± 120° chosen based on the results of the drifter pathway projections and in the cross-shore direction (at ± 30°). Average Point Petre wind speeds for April–September are about 4 m/s. Average surface currents for the four simulated wind conditions show the presence of boundary currents that can move material from the terminal toward the southern end of the bay (Figure 6). When winds are to the southwest, boundary currents exist on both the eastern and western sides of the bay and extend along the entire shoreline. There is also some recirculation predicted in the southern end of the bay. If the winds are to the southeast, currents generally move across the bay but the boundary current on the western shore in the lower part of the bay still exists. Only when winds turn to the northwest do the boundary currents reverse and flow out of the bay on both the eastern and western shorelines. Northeast wind conditions predict flow to the eastern side with a slight southward turning at the eastern shoreline. These simulations show that the boundary flows can be up to 2 cm/s faster than speeds in the middle of the bay on any given day.

Figure 6

Picton Bay surface currents simulated under four wind conditions: 4 m/s at (a) –120° (to the southwest), (b) –30° (to the southeast), (c) 120° (to the northwest), (d) 30° (to the northeast).

Figure 6

Picton Bay surface currents simulated under four wind conditions: 4 m/s at (a) –120° (to the southwest), (b) –30° (to the southeast), (c) 120° (to the northwest), (d) 30° (to the northeast).

The velocity fields generated by these uniform along- and across-shore wind fields were used to predict particle pathway spreading away from the shipping terminal. Particles were released in a semi-circular line surrounding the area of the terminal and tracked for 48 hours (Figure 7). Results show the particles disperse strongly when trapped in the boundary current on the western side when winds are to the southwest. After 12 hours, particles extend into the lower part of the bay, reaching the southern end in 2 days. When winds are to the northeast or southeast, the particles spread across the bay and reach the eastern shoreline. Winds to the northeast tend to show more dispersion of the particles than winds to the southeast consistent with the divergence of the currents in the middle of the bay. The overall movement of the particles to the east is consistent with the anecdotal evidence of fuel residue being found on the eastern shore soon after the spill, with the model predicting that would happen within 2 days. Winds to the northwest tend to keep particles trapped to the western shore until they move out of the bay.

Figure 7

Snapshots of particle positions for particles released in a semi-circular shape near the shipping terminal four wind conditions: 4 m/s at (a) –120° (to the southwest), (b) 30° (to the northeast), (c) –30° (to the southeast), (d) 120° (to the northwest).

Figure 7

Snapshots of particle positions for particles released in a semi-circular shape near the shipping terminal four wind conditions: 4 m/s at (a) –120° (to the southwest), (b) 30° (to the northeast), (c) –30° (to the southeast), (d) 120° (to the northwest).

CONCLUSIONS

The Bay of Quinte is an embayment in Lake Ontario that is used by five different municipalities to provide drinking water for over 70,000 Ontarians. Picton Bay, a part of the Bay of Quinte, is used to supply drinking water to one of these municipalities. Unfortunately, a fuel spill occurred at a shipping terminal at the mouth of Picton Bay in March 2017, and this highlighted a need for observations of the local water circulation and an improved understanding of the bay's hydrodynamics. In particular, the primary driving forces for flows in the bay are not completely understood. In fact, there is often a sparsity of in situ data and models for these water systems have their own unique issues, that result in not being able to predict the fate of contaminants such as phosphorus, salts and microplastics. Observations of drifter data collected under different wind conditions at various locations in the bay in the summer of 2017 are used to provide ground truth data of the circulation and current speeds. Furthermore, a 3D hydrodynamic numerical model was used to model the dynamics of the bay on the drifter deployment days to provide an overview of the common flow features of the bay and predictions of the movement of particles away from the spill site in different wind conditions.

The fate of the Lagrangian surface drifters deployed during the months July–September 2017 showed that all the drifters ended up travelling southwest into the bay regardless of the wind forcing as measured at Point Petre. This suggested that wind was not the primary forcing mechanism for the surface flow or that winds at Point Petre were not representative of the local winds. Numerical models using wind speeds taken from the Point Petre station indicate that a redirection of the wind field to align with the local shorelines could account for the observed movement of the drifters. It is possible that the steep cliffs that run along the western and eastern shorelines of the bay act to topographically steer local winds to be more closely aligned to the shoreline.

The shipping terminal where the fuel spill occurred is located approximately 2 km from the water intake pipes. Based on the average observed surface flow speeds, a surface contaminant moving at 6–10 cm/s could arrive at the surface above the pipes' location in about 6–9 hours if it moved in a straight-line direction. Simulated particle paths show that material could reach the water intake pipe locations in 24–48 hours if it moved within the boundary flow along the western shore. Road salts are stored at the shipping terminal and thus the terminal could be a potential point source for salt to enter into the bay. The action of chloride in cold water conditions is still not well known and a study in this region would have to consider the diffusion and transport of this material as well as the underlying 3D hydrodynamics which varies seasonally (Perera et al. 2010; Mayer et al. 2011). It is likely that ice would be an important dynamic acting within Picton Bay that would affect any potential chloride contamination.

Future data collection using these Lagrangian drifters should be directed toward sampling the deeper flow structure and sampling under more wind conditions that are directed north-south. Further, as the drifter data indicated that local winds may be significantly different from the winds at Point Petre, it would be beneficial to monitor the local wind field during future drifter deployments. The goals of this study were to provide observations of surface flow speed in Picton Bay and to provide a better understanding of the circulation to aid in the development of science-based management policies and the results of this study indicate that the flow in Picton Bay has a complex spatial structure with boundary flows influenced by local wind conditions. This study highlights the need for managers to be able to have access to in situ observations and understand the complexity of the water environments that are used as drinking water sources.

ACKNOWLEDGEMENTS

This project was supported by a grant from the Royal Military College of Canada Academic Research Program which had no other involvement in the writing or submission of this manuscript.

REFERENCES

REFERENCES
Anderson
E. J.
,
Schwab
D. J.
&
Lang
G.A.
2010
Real-time hydraulic and hydrodynamic model of the St. Clair River, Lake St. Clair, Detroit River System
.
J. Hydraul. Eng
.
136
(
8
),
507
518
.
Arhonditsis
G. B.
,
Kim
D-K.
,
Shimoda
Y.
,
Zhang
W.
,
Watson
S.
,
Mugalingam
S.
,
Dittrich
M.
,
Geater
K.
,
McClure
C.
,
Keene
B.
,
Morley
A.
,
Richards
A.
,
Long
T.
,
Rao
Y. R.
&
Kalinauskas
R.
2016
Integration of best management practices in the Bay of Quinte watershed with the phosphorus dynamics in the receiving waterbody: what do the models predict?
Aquat. Ecosyst. Health Manage
.
19
(
1
),
1
18
.
Bai
X.
,
Wang
J.
,
Schwab
D. J.
,
Yang
Y.
,
Luo
L.
,
Leshkevich
G. A.
&
Liu
S.
2013
Modeling 1993–2008 climatology of seasonal general circulation and thermal structure in the Great Lakes using FVCOM
.
Ocean Model.
65
,
40
63
.
Betts
A. R.
,
Gharabaghi
B.
&
McBean
E. A.
2014
Salt vulnerability assessment methodology for urban streams
.
J. Hydrol.
517
,
877
888
.
Chen
C.
,
Huang
H.
,
Beardsley
R. C.
,
Liu
H.
,
Zu
Q.
&
Cowles
G.
2007
A finite volume numerical approach for coastal ocean circulation studies: comparisons with finite difference models
.
J. Geophys. Res.
112
,
C03018
.
Croft-White
M. V.
,
Cvetkovic
M.
,
Rokitnicki-Wojcik
D.
,
Midwood
J. D.
&
Grabas
G. P.
2017
A shoreline divided: twelve-year water quality and land cover trends in Lake Ontario coastal wetlands
.
J. Great Lakes. Res.
43
(
6
),
1005
1015
.
Dailey
K. R.
,
Welch
K. A.
&
Lyons
W. B.
2014
Evaluating the influence of road salt on water quality of Ohio rivers over time
.
Applied Geochemistry
47
,
25
35
.
Davis
R. E.
1991
Lagrangian ocean studies
.
Annu. Rev. Fluid Mech.
23
,
43
64
.
De Dominicis
M.
,
Leuzzi
G.
,
Monti
P.
,
Pinardi
N.
&
Poulain
P.-M.
2012
Eddy diffusivity derived from drifter data for dispersion model applications
.
Ocean Dynamics
62
,
1381
1398
.
Environment and Climate Change Canada
2012
Five-year Review of Progress: Code of Practice for the Environmental Management of Road Salts
.
Minister of the Environment, Government of Canada
,
Ottawa
, p.
95
.
Freeman
N. G.
,
Prinsenberg
S. J.
1986
Exchange flows in the Adolphus Reach/North Channel
. In:
Project Quinte: Point-Source Phosphorus Control and Ecosystem Response in the Bay of Quinte, Lake Ontario
(
Minns
C. K.
,
Hurley
D. A.
&
Nicholls
K. H.
, eds),
Department of Fisheries and Oceans
,
Ottawa
, pp.
27
39
.
Can. Spec. Pub I. Fish. Aquat. Sci. 86, p. 270
.
Lumpkin
R.
,
Ozgokmen
T.
&
Centurioni
L.
2017
Advances in the application of surface drifters
.
Annu. Rev. Mar. Sci.
9
,
59
81
.
Ma
Z. M.
,
Han
G. Q.
&
deYoung
B.
2012
Modelling temperature, currents and stratification in Placentia Bay
.
Atmosphere-Ocean
50
(
3
),
244
260
.
Mayer
T.
,
Rochfort
Q.
,
Marsalek
J.
,
Parrott
J.
,
Servos
M.
,
Baker
M.
,
McInnis
R.
,
Jurkovic
A.
&
Scott
I.
2011
Environmental characterization of surface runoff from three highway sites in Southern Ontario, Canada: 1. Chemistry
.
Water Qual. Res. J. Can.
46
(
2
),
110
120
.
McCullough
G. K.
,
Page
S. J.
,
Hesslein
R. H.
,
Stainton
M. P.
,
Kling
H. J.
,
Salki
A. G.
&
Barber
D. G.
2012
Hydrological forcing of a recent trophic surge in Lake Winnipeg
.
J. Great Lakes Res.
38
,
95
105
.
Miners
K. C.
,
Chiocchio
F.
,
Yerubandi
Y. R.
,
Pal
B.
&
Murthy
C. R.
2002
Physical Processes in Western Lake Ontario Affecting Sustainable Water Use
.
NWRI Contribution No. 02-176
.
Minns
C. K.
,
Owen
G. E.
,
Johnson
M. G.
1986
Nutrient loads and budgets in the Bay of Quinte, Lake Ontario
. In:
Project Quinte: Point-Source Phosphorus Control and Ecosystem Response in the Bay of Quinte, Lake Ontario
(
Minns
C. K.
,
Hurley
D. A.
&
Nicholls
D. A.
, eds),
Department of Fisheries and Oceans
,
Ottawa
, pp.
59
76
.
Can. Spec. Publ. Fish. Aquat. Sci. 86, p. 270
.
Minns
C. K.
,
Moore
J. E.
,
Doka
S. E.
&
St John
M. A.
2011a
Temporal trends and spatial patterns in the temperature and oxygen regimes in the Bay of Quinte, Lake Ontario, 1972–2008
.
Aquat. Ecosyst. Health Manage.
14
(
1
),
9
20
.
Minns
C. K.
,
Munawar
M.
,
Koops
M. A.
&
Millard
E. S.
2011b
Long-term ecosystem studies in the Bay of Quinte, Lake Ontario, 1972–2008: a prospectus
.
Aquat. Ecosyst. Health Manage.
14
(
1
),
3
8
.
National Geophysical Data Center
1999
Bathymetry of Lake Ontario. National Geophysical Data Center - NOAA, Boulder, CO
.
Retrieved from http://dx.doi.org/10.7289/V56H4FBH (Web. 16 September 2009)
.
Niiler
P. P.
&
Paduan
J. D.
1995
Wind-driven motions in the northeast Pacific as measured by Lagrangian drifters
.
J. Phys. Oceanogr.
25
,
2819
2830
.
Oveisy
A.
,
Boegman
L.
&
Rao
Y. R.
2015
A model of the three-dimensional hydrodynamics, transport and flushing in the Bay of Quinte
.
J. Great Lakes Res.
41
(
2
),
536
548
.
Perera
N.
,
Gharabaghi
B.
,
Noehammer
P.
&
Kilgour
B.
2010
Road salt application in Highland Creek watershed, Toronto, Ontario – chloride mass balance
.
Water Qual. Res. J. Can.
45
(
4
),
451
461
.
Rao
Y. R.
,
Skafel
M. G.
,
Howell
T.
&
Murthy
R. C.
2003
Physical processes controlling taste and odor episodes in Lake Ontario drinking water
.
J. Great Lakes. Res.
29
(
1
),
70
78
.
Quinte Region Source Protection Committee
2014
Approved Quinte Region Source Protection Plan
. .
Shore
J. A.
,
Valipour
R.
&
Blukcacz-Richards
E. A.
2016
Twenty-eight years of hydrodynamic variability in the Bay of Quinte (ice-free periods of 1979–2006)
.
J. Great Lakes Res.
42
,
985
996
.
St-Hilaire
A.
,
Duchesne
S.
&
Rousseau
A. N.
2015
Floods and water quality in Canada: a review of the interactions with urbanization, agriculture and forestry
.
Can. Water Resour. J.
41
(
1–2
),
273
287
.
Stone
M.
&
Marsalek
J.
2011
Adoption of best practices for the environmental management of road salt in Ontario
.
Water Qual. Res. J. Can.
46
(
2
)
174
182
.
Walter
R. K.
,
Reid
E. C.
,
Davis
K. A.
,
Armenta
K. J.
,
Merhoff
K.
&
Nidzieko
N. J.
2017
Local diurnal wind-driven variability and upwelling in a small coastal embayment
.
J. Geophys. Res. Oceans
122
,
955
972
.