1. Introduction
The 2015 Pan and Parapan American Games from 10 July to 15 August provided Environment and Climate Change Canada (ECCC) with a unique opportunity to undertake an extensive observation campaign in the Greater Toronto Area (GTA), including a mesoscale network specifically designed to detect and track lake breezes and, in particular, the lake-breeze front (Joe et al. 2018). Additionally, two Doppler lidars (hereinafter referred to as lidars) provided real-time observations of winds. The Canadian Global Environmental Multiscale (GEM) numerical model was run at the horizontal grid spacings of 2.5, 1, and 0.25 km to study its ability to predict lake breezes and urban meteorological conditions (Leroyer et al. 2018).
Lake breezes develop because of the temperature contrast between air over cool lake water and air over the warm land surface (Atkinson 1981; Pielke 1984). The thermal contrast produces a pressure difference between the lake and land that forces cooler air inland off the lake. Figure 2 of Sills et al. (2011) shows an idealized lake breeze circulation. The lake-breeze front develops at the leading edge of the inflow layer. The surface convergence and updraft at the lake-breeze front can generate a narrow band of convective clouds (Lyons 1972). The depth of the inflow layer typically ranges from 100 to 1000 m (Lyons 1972; Keen and Lyons 1978; Curry et al. 2017; Mariani et al. 2018); however, the return flow above the inflow layer can be twice as deep (Lyons 1972).
The GTA is often affected by lake breezes because of its proximity to Lake Ontario. Estoque et al. (1976) investigated the structure and diurnal variations of lake breezes over the southern part of Lake Ontario using both observations and numerical simulations. The passage of the lake-breeze front was marked by a sharp shift in wind direction, decrease in temperature, and increase in relative humidity. Estoque et al. (1976) also showed that the lake-breeze front depth can reach 250 m and the lake breeze can penetrate as far as 30 km inland. Comer and McKendry (1993) extended the work of Estoque et al. (1976) by investigating a wider range of data. They used the lake-breeze index developed by Biggs and Graves (1962) to identify lake breezes. They found that lake breezes developed on over 30% of the days during summer over Lake Ontario and could penetrate as far as 45 km inland. They also suggested that the wind field over Lake Ontario can be influenced significantly by nearby lakes.
In the more recent studies of lake breezes in the GTA, it was found that GTA lake breezes occurred on more than 70% of warm season days (Wentworth et al. 2015; Mariani et al. 2018). Other studies of lake breezes in southern Ontario have shown that lake breezes can penetrate as far inland as 215 km (Sills et al. 2011), initiate thunderstorms (Sills et al. 2002; King et al. 2003), and affect air quality (Hastie et al. 1999; Hayden et al. 2011; Wentworth et al. 2015). Lake breezes have a large influence on the meteorology and climate of coastal cities, particularly in spring and summer, and it is therefore important to forecast lake breezes accurately.
Previous modeling studies of Lake Ontario lake breezes are limited to numerical models with grid spacings of 20 and 10 km (Estoque and Gross 1981; Comer and McKendry 1993). Estoque and Gross (1981) used a primitive equation model (e.g., momentum, thermodynamic, and continuity equations) with variable grid spacings of 20 km (along the x axis of the domain) and 10 km (along the y axis of the domain) and five vertical levels. They compared the simulated lake breeze with observations for one day. Their results showed that the effect of prevailing flows and orography were important in simulating the characteristics of the lake breeze. The comparison of the simulated and observed lake-breeze front showed general agreement. It was suggested that the detailed differences (e.g., lake-breeze location and convergence zone) were due to deficiencies of the model equations, unrealistic initial conditions, and a flat terrain. Comer and McKendry (1993) simulated Lake Ontario breezes using the Colorado State University (CSU) mesoscale model with grid spacings of 40 km for the main domain and 10 km for the nested domain. Simulations with four different gradient wind directions showed generally good agreement with observations. However, the model underestimated the inland penetration of lake breezes. They also showed that the Lake Ontario lake breeze was strongly influenced by the size and shape of the lake as well as the large-scale wind direction.
Sills et al. (2011) identified lake-breeze fronts using GEM 2.5 km simulations over the Great Lakes. The model showed some ability to predict lake breezes successfully. However, the timing and locations of the lake-breeze fronts did not always match the observations in detailed case studies over Lake Erie, Lake St. Clair, and Lake Huron. The Lake Ontario and Toronto regions were not included in their study. Leroyer et al. (2014) studied the sea-breeze events around the urban coastal area of Vancouver using GEM with grid spacings of 2.5, 1, and 0.25 km. Results showed that although GEM 2.5 and 1 km provided accurate near-surface meteorological variables (e.g., temperature, wind speed, and wind direction), the physical processes involved with sea-breeze fronts (e.g., sea-breeze inland penetration, interaction with large-scale flow) were handled better with GEM 0.25 km. Kehler et al. (2016) examined 56 cases of lake breezes over Lake Winnipeg and Lake Manitoba. They showed that GEM 2.5 km correctly simulated 78% and 68% of the Lake Winnipeg and Lake Manitoba lake breeze occurrences, respectively.
During the Pan and Parapan American Games, in addition to ground-based observations, the experimental high-resolution GEM 1 and 0.25 km were run semi-operationally for the first time for the GTA and Lake Ontario to support the weather forecast program and to evaluate the high-resolution GEM forecasts. Mariani et al. (2018) demonstrated that synoptic winds had an important impact on the characteristics of the lake-breeze fronts in the GTA during the Games. Thus, the main objective of this paper is to test the ability of the GEM model to predict Lake Ontario lake breezes under two different synoptic wind regimes, and to determine if increasing the model spatial resolution improves the forecast of lake-breeze characteristics. The ground-based observational network is used to verify the accuracy of predicted temperature, dewpoint temperature, wind speed, and wind direction. The data, model design, and lake-breeze identification methods are presented in section 2. Analysis and discussion of the model forecasts, including their comparison to ground-based observations and characteristics of lake-breeze fronts, are provided in section 3. The conclusions are given in section 4.
2. Data and methodology
a. Doppler lidar data
ECCC’s Halo Doppler lidar provides high-resolution (3 m) radial velocity (wind velocity along the lidar line of sight) measurements by measuring the Doppler shift of the backscattered pulse from aerosols. This allows remote observation of the horizontal and vertical structure of lake-breeze circulations at high resolution (Darby et al. 2002; Tsunematsu et al. 2009; Mariani et al. 2018). During the 2015 Pan and Parapan American Games, two scanning lidars operated in constant elevation [plan position indicator (PPI)], constant azimuth [range height indicator (RHI)], and 90° elevation (vertically staring) modes. The vertical velocity was estimated by measuring radial velocity in vertically staring mode. One of the lidars was deployed at Hanlan’s Point (43°36′44″N, 79°23′19″W) on Toronto Island and operated continuously. The second lidar was mounted on the back of a pickup truck and driven to different locations within the GTA in order to track the lake-breeze front as is transited inland. The maximum range of the lidar measurements varied from 2 to 5 km depending on weather conditions. The lidar measurements conducted at Hanlan’s Point and the Highway 400 King City ONroute service center at 43°53′38″N, 79°33′26″W will be used in this study.
b. Mesonet data
During the 2015 Pan and Parapan American Games, 53 automated stations were added to the existing network to increase the spatial density of surface weather observations. The resulting mesoscale network, or “mesonet,” measured 1-min temperature, dewpoint, barometric pressure, wind speed and direction, and precipitation at locations across the GTA. While some stations were located at the Games’ venues, others were set up along or near transects perpendicular to the lakeshore in order to track the inland penetration of the lake-breeze fronts (Joe et al. 2018). Thirteen “tower” stations measured wind at 10 m AGL and temperature and dewpoint at 1.5 m AGL, except at the North York location, where a shortened tower was installed atop of a low-rise building. The tower stations also measured incoming solar radiation. Twenty all-in-one “compact” stations measured wind and temperature at 2.5 m AGL, while another 20 made measurements from rooftops of mostly one- and two-story buildings. The compact station data were lightly quality controlled to remove out-of-bound values while the tower station data underwent more thorough quality control. Rooftop locations were chosen only when no suitable ground-level site could be found, most often in highly urbanized areas. No attempt was made to quantify or remove errors introduced by the use of rooftop locations. Table 1 provides information about the particular stations used for this study.
Selected surface stations.
c. Doppler radar data
The C-band Doppler radar used in this study was located north of Toronto in King City (43°57′50″N, 79°34′26″W). The radar operated at 5625-MHz frequency with a beamwidth of 0.62°. Data are sampled at 250 m and 0.5° range and azimuth resolution, respectively. These measurements, which are performed on a 10-min cycle, cover the GTA and Lake Ontario (Hudak et al. 2006; Boodoo et al. 2010). Radar “fine lines” are often observed and are due to the presence of insects along the updrafts of lake-breeze fronts and other mesoscale boundaries (Wilson et al. 1994). The radar fine lines can be used along with other observations to track the position of the lake-breeze front (Sills et al. 2011).
d. GEM model data
The GEM atmospheric model was originally developed in the 1990s at ECCC (Côté et al. 1998; Zadra et al. 2008). It is based on a fully implicit temporal solution on staggered vertical and horizontal grids (Girard et al. 2014). A full suite of physical processes is represented in the GEM model (Bélair et al. 2003a,b). The model data used in this study were produced following the configuration established for the Pan and Parapan American Games project; most of the features were similar to those in Leroyer et al. (2014) and Bélair et al. (2018). The output of the Regional Deterministic Prediction System (RDPS; Fillion et al. 2010) with a grid spacing of 10 km was downscaled through nesting to domains with grid spacings of 2.5, 1, and 0.25 km (see Fig. 1). The number of vertical levels of nested domains was 57, with the lowest thermodynamic and momentum atmospheric levels at ~5 and ~10 m AGL, respectively. Furthermore, the time step in the model decreased with decreasing grid spacings to ensure model stability. A summary of the physics schemes, time steps, horizontal grid spacing, and vertical levels is provided in Table 2. To simulate lake-breeze flows, accurate differential heating between the lake and the land is required. Therefore, in addition to previous configurations, surface temperatures for the Great Lakes were prescribed using 2-km hourly output from a coupled ocean–atmosphere forecasting system (Dupont et al. 2012) using the Nucleus for European Modelling of the Ocean (NEMO) for the daily runs. For the remaining water bodies over the model domains, direct output from the 10-km RDPS and analyses based on buoys and satellite data (Brasnett 2008) were used. Turbulent fluxes were calculated for different surface types (Bélair et al. 2003a; Leroyer et al. 2010), and over the water, they were estimated using the aerodynamic roughness length of Charnock (1955). Furthermore, the thermal and humidity roughness length of Deacu et al. (2012) was used since they found an improvement of the fluxes’ simulations over Lake Ontario. The model also used the advanced double-moment microphysics scheme of Milbrandt and Yau (2005). The land surface model of the Interaction between Soil, Biosphere, and Atmosphere (ISBA; Noilhan and Planton 1989; Bélair et al. 2003a,b) and Town Energy Balance (TEB; Masson 2000; Leroyer et al. 2011) represented land surface physical processes over natural and urban land surfaces, respectively.
The GEM domains.
Citation: Journal of Applied Meteorology and Climatology 57, 7; 10.1175/JAMC-D-17-0231.1
RDPS and GEM configurations. Here, “ConSun” is a simple condensation scheme and “MoistTKE” is a turbulent kinetic energy scheme.
e. Lake-breeze identification methods
The analysis approach described in Sills et al. (2011) used mesonet data including temperature, dewpoint, wind speed and wind direction measurements, satellite images from GOES-13, and the C-band radar reflectivity to identify the lake-breeze front. The criteria for the identification of lake-breeze fronts are given in Table 1 of Sills et al. (2011). Briefly, they include
a cumulus cloud line and/or radar fine line quasi parallel to shore and either quasi stationary or moving inland,
an elongated area of converging near-surface winds quasi parallel to shore and either quasi stationary or moving inland, and
a rapid shift in wind direction to onshore as the lake breeze moves inland.
The GEM forecasts of a wind direction shift (at ~10 m AGL), a decrease in temperature (at ~5 m AGL), and an increase in dewpoint (at ~5 m AGL) were used to identify lake breezes at 15-min intervals. Additionally, predicted vertical velocities were analyzed since a vertical velocity maximum could be an indicator of a lake-breeze front (Harris and Kotamarthi 2005; Sills et al. 2011). The vertical velocities at ~120 m AGL were used in order to minimize near-surface effects.
3. Results and discussion
The mesoscale and lidar analyses over the GTA (Mariani et al. 2018; see also appendixes A and B for mesoscale analyses) indicated that the lake-breeze front on 15 July 2015 was slow-moving with limited maximum inland penetration of 6 km under a northerly/north-northeasterly synoptic wind (opposing flow). The front remained inland from the shore for ~10 h before it retreated somewhat then dissipated. In contrast, the lake-breeze front on 9 August 2015 was fast moving, traveling more than 60 km inland within ~5 h under easterly/east-northeasterly synoptic winds (nonopposing flow). The primary purpose of this section is to determine whether the high-resolution GEM model predicted the characteristics and impact of the lake breezes under the two different synoptic flows.
a. Lake-breeze events
High surface pressure dominated the GTA with northerly/north-northeasterly (offshore) synoptic flow on 15 July 2015. The mesoscale analysis showed that the surface wind shifted to south/southwesterly as the lake-breeze front passed the lakeshore at 1508 UTC (Toronto local time +5 h). The lake-breeze front traveled 6 km inland before it began to retreat toward the lakeshore at 2000 UTC. On 9 August, the easterly/east-northeasterly synoptic flow was dominant throughout the day. Mesoscale analyses (Fig. B1) showed that the lake-breeze front developed at the eastern part of the lakeshore at 1400 UTC and extended to the western part of the GTA by 1500 UTC. The lake-breeze front reached its maximum distance of 60 km in the GTA at 2300 UTC.
Figures 2–4 show examples of mesoscale and GEM model output analyses used for identification of lake-breeze fronts on 15 July and 9 August. Figure 2a clearly shows a significant wind shift and quasi-linear convergence line just onshore, meeting two of the lake-breeze front criteria at 2000 UTC 15 July. The wind shift was captured by GEM 0.25 km, which predicted northeasterly/northwesterly winds in Fig. 3b. The predicted vertical velocity plot for 15 July (Fig. 3a) shows that the model generated a narrow updraft zone parallel to the lakeshore coinciding with a wind shift to onshore, a decrease in wind speed (Fig. 3b), a decrease in temperature (Fig. 3c), and an increase in dewpoint (Fig. 3d). The position of the updraft zone was similar to the position of the observed lake-breeze front (magenta line in Fig. 3).
Mesonet analyses at (a) 2000 UTC 15 Jul and (b) 1600 UTC 9 Aug. The meteorological data including wind barbs and the radar reflectivity are shown. The locations of lake-breeze fronts are indicated by the magenta lines. Note that the lake-breeze fronts at the top of (a) are generated by Lake Simcoe and Georgian Bay.
Citation: Journal of Applied Meteorology and Climatology 57, 7; 10.1175/JAMC-D-17-0231.1
Plots of the GEM 0.25 km numerical output for the lake-breeze event at 2000 UTC 15 Jul 2015, showing (a) vertical velocity (m s−1) at ~120 m AGL, (b) horizontal wind speed (m s−1) and direction (°) at ~10 m AGL, (c) temperature (°C), and (d) dewpoint (°C) at ~5 m AGL. The plots cover an area of ~50 × 30 km2. The white and magenta lines represent the GTA lakeshores and the lake-breeze front determined by the mesoscale analyses, respectively. The red line indicates the cross section passing through the selected surface stations in Table 1. Hanlan’s Point (HAN) and the Highway 400 ONroute site (MOB) are the locations of the lidars, and Z2D is the location of the surface station at the lakeshore. The black arrows representing the wind directions are plotted only at every tenth grid point for clarity.
Citation: Journal of Applied Meteorology and Climatology 57, 7; 10.1175/JAMC-D-17-0231.1
As in Fig. 3, but for 1600 UTC 9 Aug 2015.
Citation: Journal of Applied Meteorology and Climatology 57, 7; 10.1175/JAMC-D-17-0231.1
The mesoscale analyses in Fig. 2b identified the lake-breeze front on 9 August at 1600 UTC. Figure 2b shows that there was a clear line of cumulus accompanied by a weaker wind shift at 1600 UTC. However, because of the onshore synoptic-scale flow, gradients along the front were markedly weaker and the front was less well defined in radar imagery than was the case for 15 July. The leading edge of the lake breeze in the GEM 0.25-km model output was not clearly defined in the analysis of vertical velocity (Fig. 4a). However, similar to the 15 July case, the model produced more turbulent boundary layer flow deeper inland (depicted in the upper portion of Figs. 4a,b) and more uniform boundary layer flow close to Lake Ontario. This suggests that the model predicted the suppressing effect of the relatively cool marine air on thermal developments due to air advection from the lake. The model also predicted a decrease in temperature and an increase in dewpoint close to the leading edge of the observed lake-breeze front but with weaker gradients compared to the 15 July case. A further discussion of the lake-breeze front temporal evolution on 15 July and 9 August is given in appendixes A and B.
b. Lake-breeze front characteristics
1) Inland penetration
The inland penetration distance of the lake-breeze front was examined using the interpolation of vertical velocity for 100 points along the shore-A2T cross section (red line in Figs. 3a and 4a). Figures 5 and 6 show the temporal cross section of vertical velocities for 15 July and 9 August, respectively. Since the distance between the shore and A2T along the cross section is 28 km, the values are given to the nearest 0.28 km due to the density of points used along the transect (note that the uncertainty is larger than that, depending on the grid size used). The intersections of the observed lake-breeze fronts (mesoscale analyses) with the cross section were also determined and marked in Figs. 5 and 6.
Vertical velocity (m s−1) along the shore-A2T cross section at ~120 m AGL with (a) GEM 2.5 km, (b) GEM 1 km, and (c) GEM 0.25 km on 15 Jul. The white dots indicate the inland penetration of lake-breeze front as given by mesonet analysis. The location of the cross section in the GTA is shown in Fig. 3a. Note that figures are plotted using different scales to show clearly the updraft zone.
Citation: Journal of Applied Meteorology and Climatology 57, 7; 10.1175/JAMC-D-17-0231.1
As in Fig. 5, but for 9 Aug 2015.
Citation: Journal of Applied Meteorology and Climatology 57, 7; 10.1175/JAMC-D-17-0231.1
The predicted vertical velocity maxima in Fig. 5 clearly illustrates that the updraft zone moved inland slowly on 15 July and retreated to the lakeshore in agreement with mesoscale analyses. However, the predicted updraft zone of vertical velocity maxima with GEM 0.25 km was not continuous (Fig. 5c) since the model tended to resolve smaller structures of updrafts and downdrafts. This was more evident in the 9 August case since the high-resolution model produced more thermals on this day.
On 9 August, GEM 0.25 km produced two different regimes of vertical motions in Fig. 6c; one with smaller updraft structures ahead of the observed lake-breeze front and another with elongated structures behind the observed lake-breeze front. The boundary between the two flow regimes moved inland in the proximity of the observed lake-breeze front. Figure 7 also illustrates the horizontal wind shift of northeasterly (offshore) to southeasterly (onshore) flow, suggesting lake breeze passage even though the updraft zone of the lake-breeze front was not clear (Fig. 6c). It appears that the GEM 0.25 km model produced a weak convergence zone along the leading edge of the lake breeze (due to a lack of opposing wind) in this case and at the same time resolved the larger eddies. This makes it more challenging to locate the lake-breeze front using vertical velocity maxima.
Horizontal velocity (m s−1) along the shore-A2T cross section with GEM 0.25 km on 9 Aug. North is up, and east is to the right. The black arrows representing wind directions are plotted only at every fifth grid point for clarity.
Citation: Journal of Applied Meteorology and Climatology 57, 7; 10.1175/JAMC-D-17-0231.1
The results also showed that the magnitude of vertical velocity increased for GEM 1 and 0.25 km (Figs. 5b,c and 6b,c) compared to GEM 2.5 km (Figs. 5a, 6a), while the width of the updraft zone decreased. The width of the vertical velocity maxima was visually determined as the width of the updraft zone. As a result, GEM 0.25 km produced an updraft zone with a width of less than 2 km on 15 July. Lake-breeze fronts are generally less than 2 km in width (Lyons 1972; Curry et al. 2017). Hence, GEM 0.25 km better represented the lake-breeze width in this case.
The distance traveled by the predicted lake breeze (Fig. 8) was determined by locating the vertical velocity maxima in the updraft zone. Since the updraft zone of the lake-breeze front is not clear in the vertical velocity plots generated by GEM 0.25 km on 9 August, the boundary between the uniform and turbulent flows (Fig. 4b), wind direction changes to onshore (Fig. 7), and gradients of temperature and dewpoint (Figs. 4c,d) were used to visually estimate the lake breeze inland penetration. This method may not be as accurate as locating the vertical velocity maxima (when it is clearly defined), but it can be used to approximate the location of the lake-breeze penetration in this case. The results were compared to the inland penetration of the lake-breeze fronts identified by mesoscale analysis. While the observed lake-breeze front reached its maximum distance from the lakeshore (~6 km) at 2000 UTC 15 July, the predicted lake-breeze fronts with GEM 2.5 and 1 km reached their maximum distance of 2.2 and 3.9 km at 1700 and 2300 UTC, respectively. The predicted lake-breeze front with GEM 0.25 km penetrated to a maximum of 5.6 km at 2200 UTC before it retreated toward the lakeshore. The model tended to underestimate the inland penetration in this case. The mean absolute error (MAE) of the predicted inland penetrations from 1700 to 2300 UTC were 2.3, 2.4, and 0.9 km for GEM 2.5, 1, and 0.25 km, respectively. On 9 August, the model initially underestimated the inland penetrations, but the predicted lake breeze traveled deeper inland than the observed lake breeze after 1 h with GEM 2.5 and 1 km, and after 30 min with GEM 0.25 km (Fig. 8b). The MAE of the predicted lake-breeze penetrations from 1500 to 1700 UTC were 2.5, 1.1, and 2.3 km with GEM 2.5, 1, and 0.25 km, respectively. Overall, the location of the lake-breeze front was predicted more accurately with GEM 0.25 km on 15 July and with GEM 1 km on 9 August.
Locations of the modeled and observed lake-breeze front along the shore-A2T cross section (red line in Figs. 3a and 4a) on (a) 15 Jul and (b) 9 Aug. Note that the predicted lake breeze with GEM 0.25 km has passed by A2T station by 1700 UTC 9 Aug.
Citation: Journal of Applied Meteorology and Climatology 57, 7; 10.1175/JAMC-D-17-0231.1
2) Updraft intensity
The intensity of the lake-breeze updraft was determined by the maximum vertical velocity. Figure 9 shows the vertical profiles of vertical velocities at the Hanlan’s Point site from 1400 UTC 15 July until 0000 UTC 16 July. The positive (updraft) and negative (downdraft) vertical velocities measured by lidar (Fig. 9a) were associated with convective mixing in the atmospheric boundary layer. Lidar measurements exhibited increased updraft intensity at 1423–1431 UTC extending from surface to about 600 m AGL. The maximum vertical velocity of 2.3 m s−1 was measured at 1427 UTC at the altitude of 310 m AGL. Furthermore, the lidar PPI scan over Lake Ontario in Fig. 10a shows onshore winds at 1424. The full evolution of the wind shift as shown by several lidar PPI scans in Mariani et al. (2018) indicates that this was the time of the lake-breeze front passage at Hanlan’s Point. The GEM 2.5, 1, and 0.25 km predicted that the maximum vertical velocity occurred later at 1645, 1600, and 1600 UTC, respectively (Figs. 9b–d). Similar to Fig. 5, by increasing the model resolution, the updraft zone narrowed and the vertical velocities increased in Figs. 9b–d. Maximum vertical velocities of 0.2 and 0.5 m s−1 were predicted with GEM 2.5 and 1 km, respectively. These values are significantly smaller than the maximum vertical velocity observed by the lidar. The GEM 0.25 km predicted a higher maximum vertical velocity of 1.9 m s−1 at 365 m AGL, suggesting that the increase of model resolution improved the representation of the updraft intensity, though it did not improve the accuracy of the updraft timing in this case.
(a) Vertical velocity (m s−1) measured by lidar at Hanlan’s Point from 1400 UTC 15 Jul until 0000 UTC 16 Jul, and the predicted vertical velocities (m s−1) at the nearest grid point to Hanlan’s Point for the same period with (b) GEM 2.5 km, (c) GEM 1 km, and (d) GEM 0.25 km. The white color indicates no measurements. Note that figures are plotted using different scales to clearly show the updraft zone but the scales for (a) and (d) are the same.
Citation: Journal of Applied Meteorology and Climatology 57, 7; 10.1175/JAMC-D-17-0231.1
A snapshot of lidar measurements of radial velocity (m s−1; PPI scan) when the lake-breeze front was passing over (a) Hanlan’s Point at 1424 UTC 15 Jul and (b) the Highway 400 ONroute site at 1827 UTC 9 Aug. Negative (blue) velocities represent winds toward the lidar, indicating onshore flow; positive (red) velocities represent winds away from the lidar (indicating offshore flow at Hanlan’s Point). The black arrow shows the wind direction.
Citation: Journal of Applied Meteorology and Climatology 57, 7; 10.1175/JAMC-D-17-0231.1
The profiles of vertical velocity for the 9 August case at the Highway 400 ONroute site are presented in Figs. 11 and 12. The mobile lidar operated from 1800 to 2100 UTC; its range was limited because of fewer targets (aerosols) on this particular day at this location. A maximum vertical velocity of 3.3 m s−1 was measured at 1819 UTC at an altitude of 230 m AGL (Figs. 11a, 12). Mariani et al. (2018) determined the passage of the lake-breeze front at 1824 UTC at this location on 9 August. Additionally, the PPI scan in Fig. 10b illustrates the onshore winds at 1827 UTC. The predicted vertical velocities with GEM 2.5 km (Fig. 11b) for the period from 1400 to 2100 UTC shows that a maximum vertical velocity of 0.17 m s−1 occurred at 760 m AGL at 1815 UTC. Figures 11c and 11d show that GEM 1 and 0.25 km resolved smaller structures, producing more updrafts compared to GEM 2.5 km. Near to the time of the observed lake-breeze front passage, GEM 1 km predicted the maximum vertical velocity of 0.75 m s−1 at 1800 UTC at 760 AGL (Fig. 11c), and GEM 0.25 km predicted maxima of 2.6 m s−1 at 1715 UTC at 605 m AGL and 2.1 m s−1 at 1815 UTC at 679 m AGL (Fig. 11d). These updraft zones are more vertically extended than the ones predicted earlier (at 1515 and 1615 UTC), which could suggest that they are more likely associated with the lake-breeze front rather than convective rolls. The lack of strong updrafts after 1815 UTC also suggests that the latest updraft could be associated with the lake-breeze front. Results show that the order of magnitude of the lidar maximum vertical velocity for the available measurements (Fig. 12) was more comparable to the 0.25-km GEM prediction of vertical velocity (Fig. 11d). The timing of the maximum vertical velocity did not change significantly for different resolutions of the model.
As in Fig. 9, but at the Highway 400 ONroute site from 1400 until 2100 UTC 9 Aug. The arrow shows the maximum vertical velocity for the available lidar measurements.
Citation: Journal of Applied Meteorology and Climatology 57, 7; 10.1175/JAMC-D-17-0231.1
Lidar measurements of vertical velocity from 1800 to 1830 UTC at the height range from 60 to 240 m AGL at the Highway 400 ONroute site. The maximum vertical velocity occurred at 1819 UTC for the measurements below 240 m AGL.
Citation: Journal of Applied Meteorology and Climatology 57, 7; 10.1175/JAMC-D-17-0231.1
3) Depth
The RHI scans taken at Hanlan’s Point and the Highway 400 ONroute site (Fig. 13) were used to find lake-breeze depths by determining the altitude at which the direction of radial velocity changed from onshore to offshore. The strongest changeover of the radial velocity direction from onshore to offshore occurred at 190 and 900 m above mean sea level, as measured within 100 m (horizontally) from the lidar at Hanlan’s Point and the Highway 400 ONroute site, respectively. Similarly, the modeled lake-breeze depth was estimated by locating the altitude at which the horizontal velocity changed to an offshore wind. Figure 14 shows the observed and predicted lake-breeze depths for 15 July and 9 August. The results in Fig. 14a indicate that the depth increased after the lake-breeze front passage at 1424 UTC 15 July and decreased after the lake breeze dissipated at Hanlan’s Point. The comparisons between GEM output and lidar-measured depth show that the model did not generate any lake-breeze depth at Hanlan’s Point until 1615 UTC because of the late arrival of the late lake-breeze front in the model. The model underestimated the lake-breeze depth on average by 83 and 37 m with GEM 2.5 and 1 km, respectively, and overestimated it by 27 m with GEM 0.25 km from 1630 to 2315 UTC.
A snapshot of lidar measurements of radial velocity (m s−1; RHI scan) when the lake-breeze front was passing over (a) Hanlan’s Point at 1445 UTC 15 Jul and (b) the Highway 400 ONroute site at 1815 UTC 9 Aug. Negative (blue) velocities represent winds toward the lidar, indicating onshore flow; positive (red) velocities represent winds away from the lidar, indicating offshore flow. The direction of the x axis is facing south.
Citation: Journal of Applied Meteorology and Climatology 57, 7; 10.1175/JAMC-D-17-0231.1
Observed and predicted lake-breeze depths using lidar and GEM at intervals of 15 min at (a) Hanlan’s Point on 15 Jul and (b) the Highway 400 ONroute site on 9 Aug. Note that the lake-breeze front arrived at Hanlan’s Point and the Highway 400 ONroute site approximately at 1424 and 1815 UTC, respectively. The modeled depths were estimated at the nearest grid point to the lidar sites.
Citation: Journal of Applied Meteorology and Climatology 57, 7; 10.1175/JAMC-D-17-0231.1
On 9 August, GEM 2.5, 1 and 0.25 km overestimated the depth by 255, 133, and 143 m, respectively, from 1815 to 2045 UTC (Fig. 14b). While the GEM predictions of the lake-breeze depth were generally larger than observations, GEM 0.25 km predicted closer values to the observations prior to 1900 UTC, when it underestimated the depth by 28 m, but the error increased after 1900 UTC. Note that lake-breeze depths on this day were larger than the depths on 15 July, likely due to greater low-level instability in the atmosphere (observed by radiosonde; not shown), which could have encouraged extension of the lake breeze vertical structure (Atkinson 1981).
Overall, GEM 1 and GEM 0.25 km performed better in predicting the lake-breeze depth for the two events. Both the measured and predicted lake-breeze depths were within the ranges (100–1000 m) of previous studies of lake-breeze depth (Lyons 1972; Curry et al. 2017).
c. Lake-breeze front impact
Time series of 1-min observations at selected surface stations (Table 1) are used to examine the accuracy of the predicted temperature drop, dewpoint rise, horizontal wind speed decrease, and the timing of the wind shift to onshore upon arrival of the lake-breeze front. Since the wind shift occurs gradually, for the purpose of this paper, the timing is chosen to best represent the wind shift by matching the timing of the mesoscale analyses. The decrease in temperature and increase in dewpoint were determined within 60 min of the wind shift, since the change in temperature and dewpoint can begin earlier/later than the wind shift. A similar method was used to analyze the model output. The results are presented in Table 3. The decrease in wind speed due to the lake-breeze front is not included in the table since it was only observed at the Z2D and L1B, and at L1F on 9 August. Figures 15 and 16 show the time series of temperature, dewpoint, wind direction, and wind speed at Z2D station on 15 July and at L1F on 9 August.
Temperature drops (T), dewpoint rises (Td), and wind shift timings due to lake-breeze front at selected surface stations. The n/a for Td means highly variable measurements. When the wind shift to onshore was not observed, n/a was recorded for the timing. Zero means no decrease in temperature (or increase in dewpoint) occurred.
Comparisons of observations with the model output at the nearest grid point to Z2D station from 1200 UTC 15 Jul to 0115 UTC 16 Jul 2015: (a) temperature (°C), (b) dewpoint (°C), (c) horizontal wind direction (°), and (d) horizontal wind speed (m s−1). The observed lake-breeze front arrived at 1508 UTC. The temporal resolution of observations and predictions are 1 and 15 min, respectively. The black arrows show approximately the beginning of the lake-breeze front impact on temperature, dewpoint, wind direction, and wind speed.
Citation: Journal of Applied Meteorology and Climatology 57, 7; 10.1175/JAMC-D-17-0231.1
As in Fig. 15, but from 1200 UTC 9 Aug to 0000 UTC 10 Aug at L1F station. The lake-breeze front arrived at 1643 UTC.
Citation: Journal of Applied Meteorology and Climatology 57, 7; 10.1175/JAMC-D-17-0231.1
On 15 July, the temperature dropped 1.3°C and the dewpoint rose 2.1°C at ~1500 UTC at Z2D (Fig. 15). The offshore wind (1°–90° and 270°–360°) also shifted to onshore (90°–270°) at 1508 UTC and the wind speed decreased by ~3 m s−1, indicating that the lake-breeze arrived at the station (Fig. 15). Comparisons between observations and the GEM output show that the model failed to capture the sharpness of the wind direction changes due to diffusive processes (related to advection and turbulence in the atmosphere; Bélair et al. 2018) in the model. The model also predicted a smaller drop in temperature at 1530–1545 UTC (Fig. 15). A maximum temperature decrease of 0.7°C at 1530 UTC and a maximum dewpoint increase of 1.5°C at 1515 UTC were predicted by the model. The ground-based observations also showed that the lake-breeze front reached the L1B site at 1845 UTC (Table 3) and remained quasi stationary until 2030 UTC, causing a temperature drop of 2.9°C, a dewpoint rise of 4.7°C, and a wind speed decrease of ~2.6 m s−1. The lake-breeze front retreated slowly, arriving at the lakeshore at 0000 UTC 16 July. The model predicted a similar pattern, though it could not propagate the front to the L1B station (see Fig. 5). As a result, the model did not predict any wind shift or temperature decrease (except with GEM 0.25 km), but predicted the increase in dewpoint.
The observations at Z2D on 9 August (Table 3) showed a decrease of 1.4°C in temperature, an increase of 1.2°C in dewpoint, a change of wind direction from offshore to onshore at ~1442 UTC, and a wind speed decrease of ~2 m s−1. The model predicted a maximum temperature drop and a maximum dewpoint increase of 0.1° and 0.4°C, respectively, for this station (Table 3). The impact of the lake-breeze front was more significant at some of the stations located deep inland. For example, ~23 km from the lakeshore at the L1F station (Fig. 16), the offshore wind shifted to onshore at ~1643 UTC, indicating lake-breeze front passage at this location which dropped the temperature by 1.5°C at ~1643 UTC and increased the dewpoint by 2.5°C at ~1700 UTC (Fig. 16). However, the model predicted a maximum decrease of 0.3°C in temperature at ~1615 UTC and a maximum increase of 2.9°C in dewpoint at 1515 UTC at this station (Fig. 16). The wind speed observations showed a decrease of ~3.6 m s−1 while GEM 1 and 0.25 km predicted a decrease of 2 m s−1. The GEM 2.5 km model did not produce any decrease in wind speed during the lake breeze passage.
The model consistently underestimated the temperature drop associated with the lake-breeze front for all examined cases in this study. The errors of the predicted temperature drops ranged from 0.6° to 2.9°C and were reduced by up to 33% by increasing the model resolution except at Z2D. The model also underestimated the increase in dewpoint by up to 4.1°C. The predicted lake-breeze front timing (via the wind shift) was late by a maximum of 82 min for stations close to lakeshore and early by a maximum of 135 min for stations located deep inland. The increase of model resolution improved the prediction accuracy of timing at all the stations except at L1B.
d. Near-surface meteorological variables
The predicted temperature, dewpoint, wind direction, and horizontal wind speed were compared to ground-based observations to evaluate the performance of the model from the time the lake-breeze fronts arrived at selected surface stations (Table 1) until the time the lake-breeze circulations ended. Following the approach in Sills et al. (2011), the end time was defined as the last hour that the lake breeze could be seen at the lakeshore. Therefore, the time at which the wind shifted to offshore was considered to be the end time of the lake-breeze circulation. For example, on 15 July, the model was evaluated at Z2D from the arrival time of lake-breeze front at 1508 UTC until the end of the circulation at 0100 UTC 16 July.
Figure 17 shows the mean bias error (MBE) and standard deviation of error (STDE) estimated at 15-min intervals on 15 July and 9 August. The MBE and STDE represent the mean bias and the deviation of errors from the mean bias, respectively. The model data at the first prognostic level (~10 m AGL for wind and ~5 m AGL for temperature and dewpoint) were used to calculate the metrics since this was the nearest level to the altitudes of observations (2.5–10.3 m AGL). In addition, the lake-breeze circulation timing during which the metrics were calculated varied depending on the lake-breeze front arrival time. Therefore, the errors at surface stations cannot be compared directly. Nevertheless, the focus of this section is to obtain a range of errors during the lake-breeze events rather than comparing the results of different surface stations.
The MBE values for (a) temperature (°C), (b) dewpoint (°C), (c) wind direction (°), and (d) wind speed (m s−1) at the nearest grid point to surface stations for the periods of time that surface stations were affected by the lake-breeze circulations on 15 Jul and 9 Aug. The error bars represents the STDE values.
Citation: Journal of Applied Meteorology and Climatology 57, 7; 10.1175/JAMC-D-17-0231.1
The results indicate that the GEM 2.5 km model underestimated temperature by 1.4°–3.6°C in both case studies at all the selected stations. It also overestimated the dewpoint by 0.6°–3°C except at the Z2D (both cases) and A2T stations. The wind direction errors were determined by estimating the difference using the smallest distance on the circle to account for its wrap-up nature (i.e., 359° + 1° = 0°). The wind direction MBEs were high ranging from 9° to 93° and were particularly large at L1B on 15 July since no lake-breeze front was predicted for the L1B station. The predicted wind direction remained offshore during most of the day at this location, leading to large errors during the lake-breeze circulation. The wind speed was overestimated on 15 July and underestimated on 9 August with GEM 2.5 km. The wind speed MBE ranged from 0.1 to 2.2 m s−1 with GEM 2.5 km. The increase of model resolution (grid spacings of 1 and 0.25 km) improved the accuracy of temperature prediction, reducing the MBE by as much as 72%. GEM 1 and 0.25 km mostly underestimated the dewpoint and overestimated the wind speed. The MBEs of dewpoint and wind speed were reduced at some stations (e.g., L1E) by up to 86% with GEM 1 and 0.25 km. The increase of model resolution did not reduce the wind direction MBE significantly, except at L1B on 15 July. The results also show that the wind direction errors had the highest variability (STDE) compared to temperature, dewpoint, and wind speed. This was expected due to the natural variability of wind direction and the inability of numerical models to accurately capture these variabilities, including the timing of wind shifts (Hanna 1994; Harris and Kotamarthi 2005).
The forecast accuracy was determined by estimating root-mean-square errors (RMSEs). The RMSE represents the deviation of forecasts from observations. The RMSE values of temperature, dewpoint, wind direction, and wind speed ranged over 0.6°–3.6°C, 0.7°–3.1°C, 19°–126°, and 0.6–2.4 m s−1, respectively. To find the confidence interval of RMSE values, the bootstrap method (DiCiccio and Efron 1996) was used. The method is based on resampling using replaced data from the given sample. For this work, the errors (forecast minus observation) were resampled 10 000 times. The RMSE of the resampled errors, including the 10% and 90% percentile of the RMSEs distribution, were estimated as shown in Fig. 18. The forecast accuracy of temperature improved significantly at all the selected stations when the model resolution increased (grid spacings of 1 and 0.25 km), leading to a decrease in RMSE by a maximum of 66% and improving of the forecast accuracy of dewpoint, wind direction, and speed by a maximum of 60% at some stations (e.g., L1D).
The RMSE values and corresponding 10% and 90% confidence intervals for (a) temperature (°C), (b) dewpoint (°C), (c) wind direction (°), and (d) wind speed (m s−1) at the nearest grid point to surface stations for the periods of time that surface stations were affected by the lake-breeze circulations on 15 Jul and 9 Aug 2015.
Citation: Journal of Applied Meteorology and Climatology 57, 7; 10.1175/JAMC-D-17-0231.1
4. Conclusions
This study explored the ability of the GEM model to forecast the lake breezes under opposing and nonopposing synoptic flows during the 2015 Pan and Parapan American Games in Toronto. The case studies included the 15 July event where a slow-moving lake-breeze front impacted the GTA lakeshore regions for ~10 h and the 9 August event where a fast-moving lake-breeze front penetrated more than 60 km inland through the GTA in ~6 h. The modeled lake breezes were compared with mesoscale analyses, lidar observations of radial winds, and surface stations observations. The following are our key findings:
The GEM model successfully predicted the lake-breeze fronts for the two lake-breeze events. The wind direction shifts to onshore were captured by the model, as were the decrease in predicted temperature and increase in dewpoint. The model also produced stronger temperature and dewpoint gradients along the lake-breeze front during the opposing synoptic flow on 15 July relative to the 9 August case with a nonopposing synoptic flow; this is in agreement with observations.
The predicted vertical velocity maxima with GEM 2.5 and 1 km clearly showed the lake-breeze frontal zone for the two events, although the GEM 0.25 km did not produce a clear updraft zone associated with lake-breeze front on 9 August. It seems that the model resolved the large eddies in this case while producing a weak convergence zone associated with lake-breeze front. We speculate that the representation of turbulence in the model contributed to this issue.
GEM 0.25 km generated elongated, weak updraft structures in the lake breeze inflow region that were approximately aligned with the onshore surface wind for both 15 July and 9 August. This suggests that the high-resolution model likely captured the suppressing effect of the cooler lake air on the generation of thermals.
Comparisons of the predicted characteristics of lake-breeze fronts, including inland penetration, updraft intensity, depth, and timing with observations showed that GEM 2.5 km predicted the lake-breeze front characteristics with some degree of accuracy during the two events. However, the accuracy improved significantly with GEM 0.25 km for the 15 July case and GEM 1 km for the 9 August case.
The model underestimated the cooling behind the lake-breeze front by up to 2.9°C and underestimated the rise in dewpoint by up to 4.1°C. While the increase of model resolution improved the prediction of the temperature drops at all the selected locations, it improved the prediction of dewpoint increase only at some locations. In addition, the model sometimes failed to capture the sharpness of changes in the wind direction during the passage of the lake-breeze front, possibly due to diffusion processes in the model.
During the lake-breeze circulation the model underestimated the temperature by up to 3.6°C at the selected stations. While GEM 2.5 km overestimated the dewpoint by a maximum 3°C, GEM 1 and 0.25 km underestimated the dewpoint by up to 1.3°C. The GEM 2.5 km model also underestimated the wind speed while the higher-resolution model overestimated it by up to 2.2 m s−1. The biases and variability of errors for wind direction predictions were generally very high, with maxima of 93° and 85° for MBE and STDE, respectively.
During the lake-breeze circulation, the increase of model resolution increased the accuracy of temperature predictions significantly within the 90% percentile at all the selected stations. However, it improved the accuracy of dewpoint, wind speed, and direction predictions at some of the selected stations.
Acknowledgments
The authors thank all of the participants of the PanAm project, including John MacPhee and Reno Sit. Special thanks are given to Ivan Heckman and Dr. Zhipeng Qu for their assistance with data analysis. Thanks are also given to Brian Greaves for his help with the mesonet plots.
APPENDIX A
15 July 2015 Hourly Depictions
The hourly depictions of the positions of lake-breeze fronts between 1600 and 2300 UTC 15 July 2015 are shown in Fig. A1. Positions of lake-breeze fronts are estimated using radar, satellite, and surface station data employing the method of Sills et al. (2011).
Hourly mesonet analyses for 15 Jul 2015.
Citation: Journal of Applied Meteorology and Climatology 57, 7; 10.1175/JAMC-D-17-0231.1
Lake-breeze circulations developed on Lakes Ontario, Simcoe, and Huron (including Georgian Bay) beginning after 1500 UTC under a northerly to north-northeasterly synoptic-scale flow of around ~6 m s−1. Since the synoptic-scale flow was offshore along the north shore of Lake Ontario, the inland progression of that lake-breeze front was limited to within approximately 6 km of the Toronto lakeshore. The front began to retreat between 2200 and 2300 UTC.
Conversely, the synoptic-scale flow resulted in Georgian Bay and Simcoe lake-breeze fronts extending well inland to the southeast through the day. The Georgian Bay front began to interact with the front on the western side of Lake Simcoe near 2000 UTC, and analysis of radar fine lines suggests that, even after the Georgian Bay front passed southeast, the Lake Simcoe front on the west side of the lake could still be detected. This suggests that the Georgian Bay marine air was colder and more dense than the Lake Simcoe marine air and therefore slid beneath the Lake Simcoe air, but it is beyond the scope of this paper to do more than speculate that this was the case. By 2300 UTC, the front on the western side of Lake Simcoe dissipated, and both the front on the eastern side of Lake Simcoe and the Georgian Bay front continued to progress southeastward toward Lake Ontario. Both fronts had likely become detached vortices by this time (see Sills et al. 2011). All fronts had dissipated by 0200 UTC the next day.
APPENDIX B
9 August 2015 Hourly Depictions
The hourly depictions of the positions of lake-breeze fronts between 1400 and 2300 UTC 9 August 2015 are shown in Fig. B1. As in appendix A, positions of lake-breeze fronts are estimated using radar, satellite, and surface station data employing the method of Sills et al. (2011).
As in Fig. A1, but for 9 Aug 2015.
Citation: Journal of Applied Meteorology and Climatology 57, 7; 10.1175/JAMC-D-17-0231.1
Lake-breeze circulations developed on Lakes Ontario, Erie, Simcoe, and Huron (including Georgian Bay) beginning after 1200 UTC under an east to east-northeast synoptic-scale flow of around 3 m s−1. Since the synoptic-scale flow was onshore along the north shore of Lake Ontario, the lake-breeze front penetrated rapidly inland through the day. Initially, the front was not well defined because of the lack of surface convergence, and radar and surface observations offered only subtle clues as to the position of the developing front. Through the day, surface convergence along the front increased somewhat and the position of the front became better defined, producing an elongated line of cumulus and enhanced reflectivity apparent in satellite and radar imagery, respectively, quasi parallel to shore. By 0000 UTC the following day, the northern front had penetrated inland more than 60 km and may have become a detached vortex by this time.
The lake-breeze front was better defined along the south shore of Lake Ontario since the synoptic-scale flow had an offshore component. The front penetrated inland at a slower rate than the north-shore front, but nevertheless reached a maximum inland distance of more than 40 km. All fronts had dissipated by 0100 UTC.
REFERENCES
Atkinson, B. W., 1981: Mesoscale Atmospheric Circulations. Academic Press, 495 pp.
Bélair, S., L.-P. Crevier, J. Mailhot, B. Bilodeau, and Y. Delage, 2003a: Operational implementation of the ISBA land surface scheme in the Canadian regional weather forecast model. Part I: Warm season results. J. Hydrometeor., 4, 352–370, https://doi.org/10.1175/1525-7541(2003)4<352:OIOTIL>2.0.CO;2.
Bélair, S., R. Brown, J. Mailhot, B. Bilodeau, and L.-P. Crevier, 2003b: Operational implementation of the ISBA land surface scheme in the Canadian regional weather forecast model. Part II: Cold season results. J. Hydrometeor., 4, 371–386, https://doi.org/10.1175/1525-7541(2003)4<371:OIOTIL>2.0.CO;2.
Bélair, S., S. Leroyer, N. Seino, L. Spacek, V. Souvanlassy, and D. Paquin-Ricard, 2018: Role and impact of the urban environment in a numerical forecast of an intense summertime precipitation event over Tokyo. J. Meteor. Soc. Japan, 96A, 77–94, https://doi.org/10.2151/jmsj.2018-011.
Biggs, W. G., and M. E. Graves, 1962: A lake breeze index. J. Appl. Meteor., 1, 474–480, https://doi.org/10.1175/1520-0450(1962)001<0474:ALBI>2.0.CO;2.
Boodoo, S., D. Hudak, N. Donaldson, and M. Leduc, 2010: Application of dual-polarization radar melting-layer detection algorithm. J. Appl. Meteor. Climatol., 49, 1779–1793, https://doi.org/10.1175/2010JAMC2421.1.
Brasnett, B., 2008: The impact of satellite retrievals in a global sea-surface-temperature analysis. Quart. J. Roy. Meteor. Soc., 134, 1745–1760, https://doi.org/10.1002/qj.319.
Charnock, H., 1955: Wind stress on a water surface. Quart. J. Roy. Meteor. Soc., 81, 639–640, https://doi.org/10.1002/qj.49708135027.
Comer, N. T., and I. G. McKendry, 1993: Observations and numerical modelling of Lake Ontario breezes. Atmos.–Ocean, 31, 481–499, https://doi.org/10.1080/07055900.1993.9649482.
Côté, J., S. Gravel, A. Méthot, A. Patoine, M. Roch, and A. Staniforth, 1998: The operational CMC–MRB Global Environmental Multiscale (GEM) model. Part I: Design considerations and formulation. Mon. Wea. Rev., 126, 1373–1395, https://doi.org/10.1175/1520-0493(1998)126<1373:TOCMGE>2.0.CO;2.
Curry, M., J. Hanesiak, S. Kehler, D. Sills, and N. M. Taylor, 2017: Ground-based observations of the thermodynamic and kinematic properties of lake-breeze fronts in southern Manitoba, Canada. Bound.-Layer Meteor., 163, 143–159, https://doi.org/10.1007/s10546-016-0214-1.
Darby, L. S., R. M. Banta, and R. A. Pielke Sr., 2002: Comparisons between mesoscale model terrain sensitivity studies and Doppler lidar measurements of the sea breeze at Monterey Bay. Mon. Wea. Rev., 130, 2813–2838, https://doi.org/10.1175/1520-0493(2002)130<2813:CBMMTS>2.0.CO;2.
Deacu, D., V. Fortin, E. Klyszejko, C. Spence, and P. D. Blanken, 2012: Predicting the net basin supply to the Great Lakes with a hydrometeorological model. J. Hydrometeor., 13, 1739–1759, https://doi.org/10.1175/JHM-D-11-0151.1.
DiCiccio, T. J., and B. Efron, 1996: Bootstrap confidence intervals. Stat. Sci., 11, 189–228, https://doi.org/10.1214/ss/1032280214.
Dupont, F., P. Chittibabu, V. Fortin, Y. R. Rao, and Y. Lu, 2012: Assessment of a NEMO-based hydrodynamic modelling system for the Great Lakes. Water Qual. Res. J. Canada, 47, 198–214, https://doi.org/10.2166/wqrjc.2012.014.
Estoque, M. A., and J. M. Gross, 1981: Further studies of a lake breeze. Part II: Theoretical study. Mon. Wea. Rev., 109, 619–634, https://doi.org/10.1175/1520-0493(1981)109<0619:FSOALB>2.0.CO;2.
Estoque, M. A., J. Gross, and H. W. Lai, 1976: A lake breeze over southern Lake Ontario. Mon. Wea. Rev., 104, 386–396, https://doi.org/10.1175/1520-0493(1976)104<0386:ALBOSL>2.0.CO;2.
Fillion, L., and Coauthors, 2010: The Canadian regional data assimilation and forecasting system. Wea. Forecasting, 25, 1645–1669, https://doi.org/10.1175/2010WAF2222401.1.
Girard, C., and Coauthors, 2014: Staggered vertical discretization of the Canadian environmental multiscale (GEM) model using a coordinate of the log-hydrostatic-pressure type. Mon. Wea. Rev., 142, 1183–1196, https://doi.org/10.1175/MWR-D-13-00255.1.
Hanna, S. R., 1994: Mesoscale meteorological model evaluation techniques with emphasis on needs of air quality models. Mesoscale Modeling of the Atmosphere, Meteor. Monogr., No. 47, Amer. Meteor. Soc., 47–62.
Harris, L., and V. R. Kotamarthi, 2005: The characteristics of the Chicago lake breeze and its effects on trace particle transport: Results from an episodic event simulation. J. Appl. Meteor., 44, 1637–1654, https://doi.org/10.1175/JAM2301.1.
Hastie, D. R., and Coauthors, 1999: Observational evidence for the impact of lake breeze circulation on ozone concentrations in southern Ontario. Atmos. Environ., 33, 323–335, https://doi.org/10.1016/S1352-2310(98)00199-X.
Hayden, K. L., and Coauthors, 2011: Aircraft study of the impact of lake-breeze circulations on trace gases and particles during BAQS-Met 2007. Atmos. Chem. Phys., 11, 10 173–10 192, https://doi.org/10.5194/acp-11-10173-2011.
Hudak, D., P. Rodriguez, G. Lee, A. Ryzhkov, F. Fabry, and N. Donaldson, 2006: Winter precipitation studies with a dual polarized C-band radar. Preprints, Fourth European Conf. on Radar in Meteorology and Hydrology, Barcelona, Spain, Servei Meteorologic de Catalunya, 9–12.
Joe, P., and Coauthors, 2018: The Environment Canada Pan and ParaPan American Science Showcase project. Bull. Amer. Meteor. Soc., 99, 921–953, https://doi.org/10.1175/BAMS-D-16-0162.1.
Keen, C. S., and W. A. Lyons, 1978: Lake/land breeze circulations on the western shore of Lake Michigan. J. Appl. Meteor., 17, 1843–1855, https://doi.org/10.1175/1520-0450(1978)017<1843:LBCOTW>2.0.CO;2.
Kehler, S., J. Hanesiak, M. Curry, D. Sills, and N. Taylor, 2016: High Resolution Deterministic Prediction System (HRDPS) simulations of Manitoba lake breezes. Atmos.–Ocean, 54, 93–107, https://doi.org/10.1080/07055900.2015.1137857.
King, P. W. S., M. J. Leduc, D. Sills, N. R. Donaldson, D. R. Hudak, P. Joe, and B. P. Murphy, 2003: Lake breezes in southern Ontario and their relation to tornado climatology. Wea. Forecasting, 18, 795–807, https://doi.org/10.1175/1520-0434(2003)018<0795:LBISOA>2.0.CO;2.
Leroyer, S., J. Mailhot, S. Bélair, A. Lemonsu, and I. B. Strachan, 2010: Modeling the surface energy budget during the thawing period of the 2006 Montreal urban snow experiment. J. Appl. Meteor. Climatol., 49, 68–84, https://doi.org/10.1175/2009JAMC2153.1.
Leroyer, S., S. Bélair, J. Mailhot, and I. B. Strachan, 2011: Microscale numerical prediction over Montreal with the Canadian external urban modeling system. J. Appl. Meteor. Climatol., 50, 2410–2428, https://doi.org/10.1175/JAMC-D-11-013.1.
Leroyer, S., S. Bélair, S. Z. Husain, and J. Mailhot, 2014: Subkilometer numerical weather prediction in an urban coastal area: A case study over the Vancouver metropolitan area. J. Appl. Meteor. Climatol., 53, 1433–1453, https://doi.org/10.1175/JAMC-D-13-0202.1.
Leroyer, S., S. Bélair, L. Spacek, and I. Gultepe, 2018: Modelling of radiation-based thermal stress indicators for urban numerical weather prediction. Urban Climate, 25, 64–81, https://doi.org/10.1016/j.uclim.2018.05.003.
Lyons, W. A., 1972: The climatology and prediction of the Chicago lake breeze. J. Appl. Meteor., 11, 1259–1270, https://doi.org/10.1175/1520-0450(1972)011<1259:TCAPOT>2.0.CO;2.
Mariani, Z., A. Dehghan, P. Joe, and D. Sills, 2018: Observations of lake breeze events during the Toronto 2015 Pan-American Games. Bound.-Layer Meteor., 166, 113–135, https://doi.org/10.1007/s10546-017-0289-3.
Masson, V., 2000: A physically-based scheme for the urban energy budget in atmospheric models. Bound.-Layer Meteor., 94, 357–397, https://doi.org/10.1023/A:1002463829265.
Milbrandt, J. A., and M. K. Yau, 2005: A multimoment bulk microphysics parameterization. Part I: Analysis of the role of the spectral shape parameter. J. Atmos. Sci., 62, 3051–3064, https://doi.org/10.1175/JAS3534.1.
Noilhan, J., and S. Planton, 1989: A simple parameterization of land surface processes for meteorological models. Mon. Wea. Rev., 117, 536–549, https://doi.org/10.1175/1520-0493(1989)117<0536:ASPOLS>2.0.CO;2.
Pielke, R. A., 1984: Mesoscale Meteorological Modeling. Academic Press, 612 pp.
Sills, D., P. Taylor, P. King, W. Hocking, and I. Nichols, 2002: ELBOW 2001—Studying the relationship between lake breezes and severe weather: Project overview and preliminary results. Proc. 21st Conf. on Severe Local Storms, San Antonio, TX, Amer. Meteor. Soc., P12.7, https://ams.confex.com/ams/pdfpapers/46679.pdf.
Sills, D., J. R. Brook, I. Levy, P. A. Makar, J. Zhang, and P. A. Taylor, 2011: Lake breezes in the southern Great Lakes region and their influence during BAQS-Met 2007. Atmos. Chem. Phys., 11, 7955–7973, https://doi.org/10.5194/acp-11-7955-2011.
Tsunematsu, N., H. Iwai, S. Ishii, Y. Murayama, M. Yasui, and K. Mizutani, 2009: The formation of sharp multi-layered wind structure over Tokyo associated with sea-breeze circulation. SOLA, 5, 1–4, https://doi.org/10.2151/sola.2009-001.
Wentworth, G. R., J. G. Murphy, and D. Sills, 2015: Impact of lake breezes on ozone and nitrogen oxides in the Greater Toronto Area. Atmos. Environ., 109, 52–60, https://doi.org/10.1016/j.atmosenv.2015.03.002.
Wilson, J. W., T. M. Weckwerth, J. Vivekanandan, R. M. Wakimoto, and R. W. Russell, 1994: Boundary layer clear-air radar echoes: Origin of echoes and accuracy of derived winds. J. Atmos. Oceanic Technol., 11, 1184–1206, https://doi.org/10.1175/1520-0426(1994)011<1184:BLCARE>2.0.CO;2.
Zadra, A., D. Caya, J. Côté, B. Dugas, C. Jones, R. Laprise, K. Winger, and L.-P. Caron, 2008: The next Canadian regional climate model. Phys. Canada, 64, 75–83.
