5.1. Events in the Saint-Lawrence River Valley
Even though the warmer climate in the PGW simulation is associated with warmer surface temperatures (2–4°C), some areas show only small changes in annual freezing rain occurrence (cf. Figure 6a). This is the case for YUL, where there is a 1-h increase in median annual freezing rain occurrence from CTRL to the PGW simulation (18 h and 19 h, respectively). Freezing rain events in both simulations and observations were evaluated for stations in the SLRV (YUL and YQB). At both locations, the number of short-duration events tends to be overestimated in the historical simulation compared to the observations, whereas the number of long-duration events is slightly underestimated (Fig. 11a, b). This indicates that simulated freezing rain events are generally shorter than the observed events, leading to the underestimation of freezing rain occurrence frequencies at these locations. At YUL, the median number of annual freezing rain hours is 23 h for the observations and 18 h for CTRL. The median is similar (19 h) in the PGW simulation, despite warmer conditions. Despite similar median values between both simulations, the majority of freezing rain hours in the current climate are during events lasting < 9 h, whereas more events have a duration ≥ 9 h in the future climate. At YQB, the simulated (CTRL) median (22 h) is close to the observed median (23 h). There is a slight increase to 25 h in the PGW simulation, compared to the CTRL simulation.
5.2 Physical mechanisms in the Saint-Lawrence River Valley
Wind channelling in the SLRV contributes to the subfreezing surface temperatures, even in warmer conditions. Near-surface (10-m) wind distributions were generated at YUL (Fig. 12) and YQB (Fig. 13) in the SLRV. At YUL, northeasterly winds are dominant during freezing rain events in both simulations due to wind channeling in the SLRV, especially for long duration events. However, they are slightly more frequent in the PGW simulation (82.3%) than in the current climate (69.5%). In the warmer climate, wind speed is higher during long-duration freezing rain events. Similar conditions occur at YQB (Fig. 13), but stronger northeasterly winds are more frequent under PGW conditions. In summary, stronger northeasterly winds enhance low-level cold-air advection and could explain the small change in the occurrence of annual freezing rain in warmer conditions, at this location.
We also investigated wind speed and direction at 850 hPa, since they contribute to the presence of warm air aloft at these locations (Fig. 14). During freezing rain events, the 850 hPa-wind speeds are generally stronger and blow from the southwest in the CTRL simulation, whereas the PGW produced a broader wind direction distribution, from southwest to southeast (Fig. 14). The stronger winds in the current climate at this location occurred during the short-duration events. The mean wind speed during all events decreased by 3.5 m s− 1, from 20.3 m s− 1 in the CTRL simulation to 16.8 m s− 1 in the PGW simulation. While all events are associated with the same mean wind speed in the historical climate (~ 20 m/s), the short-duration events under PGW conditions are associated with a lower wind speed (15 m/s). Conditions at YQB (Fig. 15) are generally similar to YUL, with some differences in wind speed. In PGW conditions, up to 20% of the short-duration events are associated with 850-hPa easterly winds. The decrease in wind speed is similar to the decrease at YUL for these events, but the wind speed is slightly lower for the long-duration events (~ 3 m/s at YQB and ~ 5 m/s at YUL). Despite stronger 10-m northeasterly winds being produced in the PGW simulation in the SLRV, weaker 850-hPa winds occur during freezing rain events. This suggests that wind channelling in the SLRV will continue to facilitate freezing rain events in the future.
The physical mechanisms that are responsible for wind channeling in the SLRV are identified by studying the distribution of wind directions within and above the valley (Whiteman and Doran 1993; Carrera et al. 2009). Wind direction distributions were generated using a two-dimensional Gaussian kernel density estimation (KDE) with 3-hour wind data at YUL and YQB (Fig. 16). The 10-m winds are used for within-valley distributions and the 925-hPa winds were used for above the valley, because this pressure level is always above the valley walls. In general, there is both downward momentum transport and pressure-driven channelling in the SLRV. During freezing rain events, the channeling is mainly pressure-driven, especially at YQB, but a small component of downward momentum transport is still present at YUL. Therefore, the stronger winds at the surface during freezing rain (cf. Figures 12, 13) are likely caused by a stronger pressure gradient along the valley axis.
A composite of sea level pressure (SLP), 850-hPa geopotential height, and temperature at the onset of freezing rain events was produced at both locations in the SLRV (Fig. 17). For both simulations, the temperature gradient is stronger for long-duration events, which maintained the melting layer aloft for a longer duration/period of time. During freezing rain episodes at YUL, the warm air advection is weaker in the PGW simulation for short-duration events, but stronger for longer events. At YQB, the temperature gradient is stronger in the PGW simulation than in the CTRL simulation, with lower wind speeds. The pressure gradient is stronger at the onset of long-duration events, thus enhancing pressure-driven wind channeling.
5.3 Thermodynamic conditions aloft
The vertical temperature structures in both simulations were compared in Fig. 18. Compared to the current climate, the depth of the melting layer aloft under PGW conditions was on average 196 m and 197 m thicker during freezing rain episodes at YUL and YQB, respectively. In contrast, the refreezing layer remained the same at YUL but increased by 180 m at YQB in the PGW simulation, compared to CTRL. Both layers are generally thicker during long-duration events and thinner during short-duration events at both locations. This difference is more pronounced at YUL, where the mean melting layer is 327 m thicker in the PGW simulation. The maximum temperature of the melting layer aloft generally increased under PGW conditions. For instance, during freezing rain occurrences, the maximum temperature of the melting layer aloft increased by ~ 1°C and ~ 0.6°C at YUL and YQB, respectively. The mean minimum temperature of the refreezing layer in the PGW simulation is ~ 1°C warmer at YUL but remains the same at YQB (Fig. 18h) for both simulations.
In summary, the thicker and warmer melting layers aloft under PGW conditions are consistent with the generally warmer conditions that occur during freezing rain events at these locations. In the future, weaker warm air advection could produce the above-0°C layer aloft, but stronger northeasterly winds within the SLRV are needed to produce and maintain a refreezing layer near the surface.
5.4 Locations of significant changes
A substantial increase in median annual freezing rain occurrences was observed at Rivière-du-Moulin wind farm (RDM) in warmer climate conditions (cf. Figure 6a). Since this location is not in the SLRV, the conditions that lead to freezing rain are not linked to wind channeling. The distribution of event duration (cf. Figure 11c) shows a general increase in the number of freezing rain events under PGW conditions, with three times as many long-duration events (≥ 6 h). The annual number of hours of precipitation at this location is increased by 15%, from 688 h/y in historical climate to 794 h/y in a warmer climate. The annual number of freezing rain hours increases by 35%, from 26 h/y to 35 h/y in the PGW simulation. Therefore, not only do the total precipitation hours increase in warmer conditions, but the freezing rain occurrence increases even more, meaning that the freezing rain to total precipitation ratio is higher in the PGW simulation. Freezing rain associated with a supercooled warm rain process increases by 31%, going from 0.9 h/y in the CTRL simulation to 3.7 h/y for the PGW simulation, but a smaller increase in freezing rain produced by melting solid precipitation of 7.6 h/y occurs (42.2 h/y in the CTRL simulation to 49.8 h/y in the PGW simulation). Therefore, the increase in freezing rain occurrence in the warmer climate at this location is mainly due to an increase in the occurrence of a melting layer aloft (68.2%), as well as the occurrence of supercooled warm rain (31.8%).
Even though the median annual freezing rain hours over northern Maine are similar in both CTRL (21 h) and PGW (22 h) simulations, the annual freezing hours attributed to long-duration events in the PGW simulation increased (cf. Figure 7f and Fig. 8f). For example, the total freezing rain hours is higher in a warmer climate for events that are longer than 10 h (cf. Figure 11d). Also, a significant portion of long-duration events in the PGW simulation are linked to similar synoptic conditions, but are categorized as short-duration events in the CTRL simulation (cf. Figure 10b, f). This is due to the delayed onset of freezing rain within the precipitation event. Compared to CTRL, concurrent precipitation events for the PGW approach last up to 2.4 h longer, but freezing rain lasts more than 2 times longer (5.4 h) during these events. Precipitation events for PGW generally start slightly earlier (1.5 h) and end significantly later (2.8 h), and freezing rain also starts earlier during those events (~ 10.9 h in the CTRL simulation vs ~ 5.5 h in the PGW simulation). Therefore, these precipitation events have slightly longer durations in the PGW simulation, and freezing rain starts much earlier and lasts way longer in the warmer climate. Figure 19 shows the composite temporal evolution of these concurrent events 9 h after the onset of freezing rain. In the PGW simulation, freezing rain starts earlier than in the CTRL simulation and therefore the location is farther northeast of the center of the low-pressure system. The low-pressure system also moves slightly slower in the PGW simulation. Note that even if PGW does not impact the storm tracks, warmer conditions lead to mesoscale changes in storm location and propagation speed. Finally, 850-hPa warm air is present for a longer duration in the PGW simulation compared to CTRL, once the events commence.