Jonathan Gula and W. Richard Peltier Department of Physics, University of Toronto, Toronto, ON

Transcription

1 Generated using V3.0 of the official AMS L A TEX template Dynamical Downscaling over the Great Lakes Basin of North America using the WRF Regional Climate Model, Part 2: Changes in daily and extreme temperature and precipitation indices. Jonathan Gula and W. Richard Peltier Department of Physics, University of Toronto, Toronto, ON Corresponding author address: Jonathan Gula, Department of Physics, University of Toronto, 60 St George St., Toronto, ON, M5S 1A7. gula@atmosp.physics.utoronto.ca 1

2 ABSTRACT This paper is the second part of the work on dynamical downscaling over the Great Lakes Basin of North America using the WRF Regional Climate Model. Part 1, discussing the skill of the models in reproducing late 20 th century temperature and precipitation and interactions between lakes and climate, was presented in the companion paper Gula and Peltier (2011). Here, we investigate the regional climate changes to be expected over the Great Lakes Basin of North America during the next century following different emission scenario. Large freshwater systems, such as the Great Lakes, play a key role in determining the climate of their basins and adjacent regions through the exchange of heat and moisture with the atmosphere. Even systems as extensive as the Great Lakes are unresolved in coarse resolution global climate simulations but may be accurately captured in finer-mesh regional simulations by dynamical downscaling. Historical ( ) and future ( and ) conditions are simulated using the Weather Research and Forecasting model (WRF) forced by CCSM3 global simulations. A two-step nesting procedure is employed for the purpose of downscaling, in which the first nested WRF model is of North American continental scale at 30 km resolution, whereas the innermost domain at 10 km resolution covers the Great Lakes Basin and the Canadian Province of Ontario. As the WRF model does not currently have an explicit lake component, simulations are performed using output from the freshwater lake model FLake (Mironov, D. V., 2008, COSMO Technical Report, No. 11, Deutscher Wetterdienst, Offenbach am Main, Germany) forced by atmospheric fields from the global simulations. Changes in surface temperature and ice cover for the Great Lakes under future atmospheric conditions are discussed. The trends in temperature and precipitation for the future regional climate as simulated by WRF are discussed in view of these effects, one emphasis being upon the impact of greenhouses gas induced climate change on winter snowfall in the lee of the lakes. 1

3 1. Introduction Recent studies of the Great Lakes region and surronding areas have shown that significant changes in temperature and precipitation occured during the twentieth century, and further changes are likely to occur in response to increases in CO2 and greenhouses gases emissions during the twenty-first century. The Great Lakes, as the largest single sources of freshwater in the world and with over 35 million inhabitants living within their watershed, are vital to the economies of both US and Canada. It is well-known that the Great Lakes play a key role on the climate of their region. Much of this influence is due to differences in the heat capacities between the water and land surfaces area, and to the large source of moisture the lakes provide to the lower atmosphere. In addition, changes in terrain height adjacent to the lakes further modify climate variables. Previous studies (e.g. Scott and Huff (1996) and references therein) suggest that major effects of the lakes are to moderate maximum and minimum temperatures of the region in all seasons, to increase cloud cover and precipitation over and downwind of the lakes during winter and to decrease convective clouds and rainfall over the lakes in summer because of the greater atmospheric stability imparted by the relatively cooler water. The role of ice cover is of particular importance as it is known to have a great effect on both regional climate and weather events (see Brown and Dugay (2010) and references therein). The presence of ice cover on lakes strongly modifies the interaction with the atmosphere, which makes timings of lakes freezeup and ice break-up crucial for the simulations. Mishra et al. (2011), which studied small inland lakes in the Great Lakes region during past century, stated that a significant increase ( C/decade) in air temperature resulted in a significant change ( days/decade) in lake ice freeze-up and breakup dates. Projected climate warming will likely have severe impacts on lake ice phenology, with the potential of further increasing the role of lakes (Brown and Dugay (2010)). A number of studies have been investigated responses to climate which may include: a shift in spring snowmelt peak (Cayan et al., 2001), changes in thermal stratification in lakes (O Reilly et al., 2003), and a reduction in lake ice cover (Austin et Colman, 2007, 2008; Desau et al., 2009). During the past few years, there have been numerous studies of extreme temperature and precipitation indices for various regions of the globe. Overall, the findings have suggested a significant decrease in the number of days with extreme cold temperatures, an increase in the number of days with extreme warm temperature and some detectable increase in the number of extreme wet days. Such analysis in Southern Canada (Vincent and Mekis (2006)) similarly showed fewer cold nights, cold days and frost days, and conversely more warm nights, warm days and summer days across the country. Analysis of stalagmites from Spring Valley Caverns in Minnesota, south-west of the Great Lakes, shows that these extreme flood events have increased from the last half of the nineteenth century (Dasgupta et al. (2010)). Increased moisture availability in the Midwestern region, due to rise in temperature from global warming could lead to an increase in the occurrence of extreme rainfall events. While global simulations indicate large-scale patterns of change associated with natural and anthropogenic climate forcing, they cannot capture the effects of localized mountain ranges, complex landwater interactions, or regional variations in land use. A major question is then whether such mesoscale geography will significantly alter the local temperature and precipitation trends under climate change. Because extreme events tend to be local and are strongly influenced by topography, regional climate models predictions of extremes are expected to be substantially better than those from GCM s. For many resource allocation decisions, information is required at very small scales. To address these shortcomings, we are performing new century-scale climate simulations at a high resolution of 10 km over the Great Lakes Basin of North America using the Weather Research and Forecasting (WRF) model 2

4 as an RCM and the National Center for Atmospheric Research Community Climate System Model 3.0 (CCSM3) as a GCM. Both models are being run at the SciNet Supercomputer facility, on its IBM Power 6 platform. The paper is organized as follows. The regional climate model, the lake parameterization and the experimental configuration are briefly recalled in section 2. Projections for the two periods and under the two different emissions scenarios SRES A1B and SRES A2 are given in section 3. Changes in surface temperatures and ice cover for the Great Lakes under future atmospheric conditions are described. The trends in temperature and precipitation, and the differences in extreme temperature and precipitation events are discussed in view of these effects. Finally, discussion and conclusions are presented in section Model description and experimental setup The methodology used throughout this paper is the same than the methodology presented in Gula and Peltier (2011). a. The global model of climate system evolution: CCSM3 The global GCM to be employed for the purpose of the analyses to follow is the National Center for Atmospheric Research Community Climate System Model 3.0 (CCSM3). This model is a coupled climate model for simulation of the evolution of Earth s climate system and was featured prominently in the IPCC Fourth Assessment Report (IPCC, 2007). It is composed of four separate models describing the coupling between the Earth s atmosphere, oceans, land surface and sea-ice, which are individual modules that interact with one-another via a central coupler component. We employ the version of the model at T85 resolution for the atmosphere and land which operates on a 256x128 regular longitude/latitude global horizontal grid (giving a 1.4 degree resolution) with 26 levels in the vertical. Output from an experiment run with observed 20 th century greenhouse gasses was used to force the historical WRF simulations. The global simulations used in this study for future climate projections were forced with the Special Report on Emissions Scenarios (SRES) A2 and A1B emissions scenario (Nakicenovic and Swart (2000)). b. The regional climate model: WRF The regional meteorological model being employed in this ongoing project is the Weather Research and Forecasting (WRF) model with the Advanced Research WRF (ARW) dynamic core, version 3.3 (Skamarock et al. (2007)). WRF is a next-generation, limited area, non-hydrostatic model, with terrain following eta-coordinate mesoscale modeling system designed to serve both operational forecasting and atmospheric research needs. We choose WRF because it is being developed and studied by a broad community of government and university researchers; however, there have only been a limited number of studies applying WRF for regional climate applications. The simulations to be discussed herein have been performed with the 3.3 version of the model with the following physical options: the WRF Single-Moment 6-class (WSM6) microphysical parameterization (Hong et al. (2004)), the CAM3 short-wave and long-wave radiation scheme, similar to the scheme used in CCSM, and which allows for aerosols and trace gases, the Yonsei University (YSU) planetary boundary layer (PBL) scheme (Noh et al. (2003)), and the Noah land surface model (Chen and Dudhia (2001)). The CLWRF 3.1 (Fita et al. (2009)) which is a set of modifications to the WRF code 3

5 (a) (b) Fig. 1. (a) The map of North-America showing the WRF 30-km horizontal resolution domain over North-America (domain 1) and the 10-km horizontal resolution domain over Ontario and the Great Lakes Basin (domain 2). (b) Lake depth for the Ontario and Great Lake Basin from the Global Lake Dataset (Kourzeneva (2009)). Data from Buoy locations in three of the lakes, shown in (b) as white markers, are employed to compare model output to observations. Small white dots in (b) show the spatial distribution of the Canadian rehabilitated precipitation dataset stations. employed to perform more flexible regional climate simulations has been incorporated in the 3.3 version used here. The two main capabilities added to the model then include a more flexible GHG scenario implementation and output of mean and extreme statistics of surface variables. We use WRF with multiple nesting, configuring the innermost domain to cover Ontario and the Great Lakes Basin and the outermost domain to cover the whole of Canada and most of North America. The model outer domain is centered at 50 N and 100 W with dimensions of horizontal grid points with spacing of 30 km (Figure 1). The Lambert conformal conic projection is used as the model horizontal coordinates with the standard parallel at 50 N. The domain boundaries are located mostly over flat land points or ocean points to avoid vertical interpolation problems due to the differences in topography between the forcing data and WRF. In the vertical, we use 28 terrain-following eta levels. The horizontal grid size for the innermost domain is 10 km and is nested to the outermost domain. The model provides outputs every 6h for domain 1 and 3h for domain 2. Initial and boundary conditions for the large-scale atmospheric fields, sea-ice and sea surface temperature (SST), as well as initial soil parameters (soil water, moisture and temperature) are provided by the CCSM output every 6h. The domain specified lateral boundary is composed of a 1-point specified zone and a 9-point relaxation zone. Boundary conditions at the specified zone are determined entirely by temporal interpolation from the 6-hourly CCSM data. Lateral boundary conditions at the relaxation zone are nudged toward the CCSM data following the method of Davies and Turner (1977), with higher nudging coefficients for grid points which are closer to the specified zone. The SST, sea ice and green vegetation fraction are updated every 6h during the simulations. In the Noah land surface 4

6 model (LSM), however, the 3-m-deep soil temperature is prescribed as a constant climatological value, which does not allow the soil column to realistically respond to climate forcing, the program has been modified to use soil temperatures at the lower boundary from the CCSM 6h-output. The method of Newtonian relaxation or nudging, as first introduced by Charney et al. (1969), relaxes the model state toward the observed state by adding, to one or more of the prognostic equations, artificial tendency terms based on the difference between the two states. The model solution can be nudged toward either gridded analyses (analysis nudging) or individual observations (observational nudging) during the period of time surrounding the observations. Here we apply analysis nudging to the downscaling simulations to investigate the effect of assimilating the large-scale driving fields throughout the integration. In WRF the analysis nudging can be applied to different fields (potential temperature, velocity and/or humidity) in the whole domain or only above a given vertical layer, and for any nudging coefficients. We then conducted different experiments applying full 3-D nudging throughout the whole atmospheric column for one or more variables, or turning it off for the lowest vertical model layers. If the regional model is constrained too strongly to the GCM fields, however, there is the possibility that the benefit of using the higher-resolution regional model will not be realized. What is needed is a delicate balance between the amount of constraint given to the model and freedom to simulate its own mesoscale features. A new feature of WRF is the use of spectral nudging (Miguez-Macho et al. (2004)), which is a form of nudging where only the lowest wave number (largest scales) are used to nudge the interior of the domain. The simulations presented in the following have been realized using spectral nudging for the outer domain. The inner domain is not nudged, allowing the mesoscale model to freely develop atmospheric structures at finer spatial scale. Thus, this approach maintains the objective of downscaling, which is to generate mesoscale meteorological details consistent with the large-scale state simulated by the global model. c. Parameterization of the lakes The WRF model does not currently have an explicit lake component, lake ice and lake surface temperature need to be prescribed in the model. The CCSM3 simulation provides all fields necessary for the driving of the regional model, but is unable to provide accurate lake ice cover and lake surface water temperature, as its resolution is too coarse to produce realistic fields. An additional lake parameterization is needed to fill the gap between the driving CCSM3 data and the regional model. The FLake model is based on a two-layer parametric representation of the evolving temperature profile, with the mixed layer at the surface, and the thermocline extending from the lake bottom to the base of the mixed layer (see Mironov (2008) for a detailed description of the model). The lake thermocline is described using the concept of self-similarity of the thermal structure of the water column, which originates from observations of oceanic mixed layer dynamics (Kitaigorodskii and Miropolsky (1970)). The same parametric concept is applied to the temperature structure of the bottom sediment layer and of the ice and snow layers. A system of prognostic ordinary differential equations is solved for the time-dependent quantities which are the mixed-layer temperature and depth, the lake ice thickness and temperature and the temperature and depth of the layer of bottom sediments penetrated by the annual thermal wave. Convective entrainment, wind-driven mixing and volumetric solar radiation absorption are included in the formulation of the mixed-layer depth equation. The two-layer water temperature parameterization limits the ability of the FLake model to simulate very deep lakes, and in such cases it is suggested by model developers to use a virtual bottom at 50 to 60 meters with the bottom-sediment module switched off, instead of actual lake depth. A snow 5

7 module is present in the FLake model but has yet to be sufficiently tested. A correction of the ice albedo, taking into account the influence of the snow cover, is applied instead. The exchanges with the atmosphere are mainly determined by air temperature, precipitation, wind and radiation, while heat storage in the lake is also determined by morphometry. Lake morphometry is a determinant of ice cover as this affects the wind fetch, water circulation and temperature, as well as heat storage (Jeffreys and Morris (2007)). The most important morphological aspect of a lake is the depth (Korhonen (2006)), since this will determine the amount of heat storage in the water and hence the time needed for the lake to cool and ultimately freeze. For the purpose of the following analyses we have employed data from the Global Lake Dataset (Kourzeneva (2009)) in the two WRF domains. Lake depths for Ontario and the Great Lake Basin at 10km resolution are also shown in figure 1. 6

8 Fig. 2. Greenhouse gases emissions for A2 (red) and A1B (blue) emissions scenario (left panel). Projected surface-air temperature (10-year running average) as simulated by CCSM3 for A2 (red) and A1B (blue) emissions scenario (right panel) averaged, from top to bottom, over the world, over North- America (domain 1 in Figure 1) and over Ontario (domain 2 in Figure 1). 3. Future climate projections a. Emission scenarios The SRES-A2 scenario coresponds to a very heterogeneous world which results in continuously increasing population and anthropogenic emissions throughout the 21 st century. The SRES-A1B scenario describes a future world of very rapid economic growth, global population that peaks in mid-century and declines thereafter, and the rapid introduction of new and more efficient technologies with a balanced emphasis on all energy sources. Anthropogenic emissions of CO2, CH4, N2O corresponding to the two scenarios are shown Figure 2, left panel. We have produced CCSM3 6-hourly output fields from 1960 to 2100 using these two scenarios at T85 resolution. Projected changes in surface air temperature are shown Figure 2, right panel, in red for the A2 scenario and in blue for scenario A1B. Following prescribed anthropogenic emissions, the increase in mean temperature is slightly larger for scenario A1B compared to A2 during the first half of the 21 st century. Anthropogenic emissions for scenario A1B reach a peak around , after which the warming starts to slow down. Projected changes in surface air temperature are quite similar during the period for both scenario (see Table 1), but become divergent afterwards. During the period, an average 3.5 C increase in air temperature is projected for the two domains under scenario A1B, while a 5.0 C increase in air temperature is projected for scenario A2. 7

9 Table 1. Projected changes in surface air temperature A1B A A1B A World Domain Domain

10 b. Lake ice and temperature for the 21 st century future scenarios Simulations using the Flake model were performed for the period using the results from the CCSM simulations under IPCC SRES A2 and A1B scenarios. Figure 3 show the annual average ice duration and surface lake temperature for the , and periods for the A1B scenario. Figure 4 show the annual average ice duration and surface lake temperature for the , and periods for the A2 scenario. Following the warming of air temperature predicted by the CCSM simulations (Figure 2), the surface lake temperature rises and the percent of ice cover is greatly diminished. Under scenario A1B, projected changes in surface water temperature averaged over the domain of Figure 3 are 2.3 C during and 3.1 C during , which translate to a reduction in lake ice cover period of 27.7 days and 40.0 days, respectively. Under scenario A2, projected changes in surface water temperature averaged over the domain of Figure 4 are 2.6 C during and 4.7 C during , which translate to a reduction in lake ice cover period of 29.0 days and 56 days, respectively. Projected changes in break-up and freeze-up dates are highly variable depending on the location, due to the great differences in the morphometry of the Great Lakes. The largest changes are projected for the deepest parts of all lakes. Projected changes in surface lake temperature show a slightly larger increase for the northern part of the domain than the southern, related to the latitudinal effect of air temperature and the larger decrease in the number of days with ice cover. During the period, under scenario A2, the largest increase in mean annual surface lake temperature (5.5 C) is projected in the northern part of Lake Superior while the lowest increase (4.3 C) is projected in the southern parts of Lake Michigan and lake Erie. Projected changes in water surface temperature show the largest increase in summer in the deepest part of the lakes. Porojected changes go up to 6 C for Lake Superior during under scenario A2, which is 2 C higher than the air temperature increase projected by the global CCSM simulation above Lake Superior in summer during the same period. In shallower parts changes follow more closely projected changes in air temperature. As observed in recent studies (Austin and Colman (2007)), the declining ice cover is causing the onset of the positively stratified season to occur earlier and contribute significantly to the increase of summer temperature. These effects are amplified in very large lakes such as Lake Superior due to the larger heat storage and greater ice decline, while shallower lakes undergo a rapid lake-atmosphere equilibration. It should be noted that Austin and Colman (2007) observed an even greater rate of change (about 0.11 C/yr) for summer water temperature in Lake Superior during the period than projected here. c. Projected mean temperature changes Changes in the seasonal and annual means and annual cycle of daily maximum and minimum temperature and precipitation from the historical period to the end of the 21st century are presented here. Ten year time slices were chosen in order to strike a balance between a long enough time period to sample interannual variability, such as El Nino/La Nina Southern Oscillation (ENSO) variability, but short enough to isolate a time period for which global climate change would essentially be static (Salathé et al. (2008)). The projected mean changes in surface-air temperature for late 21 st century ( and means) relative to late 20 th century temperature ( ) are shown figure 5 for the A1B and A2 scenario. The spatial pattern and the magnitude of the projected temperature change are quite similar for both scenario during the period. In winter, the largest projected changes are in the northern part of Ontario, with changes as large as 5 C in the Hudson Bay region and in the vicinity 9

11 (a) (b) (c) Fig. 3. Average ice duration in days and surface lake temperature for the Great Lakes simulated by Flake driven by CCSM3 atmospheric fields, (a) for the historical period, (b) for the period and (c) and for the period following scenario SRES-A1B. 10

12 (a) (b) (c) Fig. 4. Average ice duration in days and surface lake temperature for the Great Lakes simulated by Flake driven by CCSM3 atmospheric fields, (a) for the historical period, (b) for the period and (c) and for the period following scenario SRES-A2. 11

13 of the Lake Superior and 3-4 C in the northern part of Lake Michigan and Lake Huron. The projected summer-time temperatures are far more modest, with the southern part of Ontario experiencing the largest changes (order 2-3 C). Projected changes for the period following the A1B scenario are quite similar than for the period. The spatial pattern is identical and the magnitude larger by 1 C. The projected changes are much larger for the A2 scenario with changes as large as 8-9 C in the northern part of Ontario during winter and 5-6 C in the southern part of Ontario during summer. d. Projected changes in extreme temperatures In addition to exploring projected changes in mean monthly temperatures, it is also of interest to consider projected changes in daily temperature extremes. Histograms of daily mean, maximum and minimum temperatures in the months January and July over the decades , and for scenario A1B and A2 are shown Figure 6 and 7, respectively. For both scenario the shift in distributions is clearest in July, where the distributions shift to the right such that the peaks are quite distinct between the late 20 th century and the late 21 st century distributions. The changes in minimum, mean and maximum daily temperature all show a similar shift, which is about 3 C under scenario A1B and 5 C under scenario A2. The effect is still discernible in the January temperatures, but no so pronounced as for July. Besides, changes are also seen in the shapes of the distributions, with a different behaviour between January and July. In January the distributions become thinner for the late 21 st century, with a small increase of the distribution toward higher temperature but a large decrease for very cold temperatures. In July the distributions widen and the tail toward very hot temperatures is largely increased, especially for the maximum temperature under scenario A2. In general, all histograms show a shift towards significantly mode hotter days over Ontario and the Great Lake Basin by the end of the 21 st century. Another approach is to consider how often the daily minimum and maximum temperature exceeds various thresholds. A set of eight temperatures indices was selected for this study. The definitions are presented in Table 2. The indices describe cold events (frost days, cold days, cold nights), warm events (summer days, warm days, warm nights), and provide a measure of the temperatures variability (diurnal temperature range, standard deviation of the daily mean temperature). They are calculated on an annual basis. Some are based on a fixed threshold (e.g. frost days, summer days), and their impact is eay to understand, however, they are not applicable to every region (e.g. summer days are rare in Northern Ontario). Others are based on thresholds defined as percentiles calculated at each grid point (e.g. cold nights, warm days), which can be used to facilitate comparison between different climatic regions. The diurnal tenperature range is the annual average difference between daily maximum and minimum temperatures. The standard deviation is calculated using the daily mean temperature departures from the daily average for the period. The indices based on percentiles represent the annual number of days above the 90 th or below the 10 th percentile level. The percentile values are calculated from the reference period. Thery are obtained for every day of the calendar year and do not necessarily extreme hot days in the summer or extreme cold days in the winter. The percentiles are computed using 115 values: a five-day window centered on each calendar year day for The spatial pattern and magnitude of the changes for these indices are shown for the periods and for the two scenario relative to the period in Figure 8. The results show that the number of cold events is projected to significantly decrease while the number of warm events is projected to significantly increase. The numbers of frost days, cold days and cold nights show a similar 12

14 Fig. 5. Changes in surface air temperature for the and periods relative to the historical period ( ) over Ontario for annual, winter and summer precipitation from WRF simulation for scenario A1B and A2. 13

15 (a) T min (January) (b) T min (July) (c) T mean (January) (d) T mean (July) (e) T max (January) (f) T max (July) Fig. 6. Histograms showing Mean, Maximum and Minimum daily temperature for all model grid points over the Great Lakes Basin (domain of figure 3) in January (left) and july (right) for the periods (black), (blue) and (red). 21 st century results are from the A1B scenario. Histrograms are each binned into 100 bins. Temperature are in degrees Celcius. 14

16 (a) T min (January) (b) T min (July) (c) T mean (January) (d) T mean (July) (e) T max (January) (f) T max (July) Fig. 7. Histograms showing Mean, Maximum and Minimum daily temperature for all model grid points over the Great Lakes Basin (domain of figure 3) in January (left) and july (right) for the periods (black), (blue) and (red). 21 st century results are from the A2 scenario. Histrograms are each binned into 100 bins. Temperature are in degrees Celcius. 15

17 Table 2. Definitions of the eight temperature indices used in the study. Tmax, Tmin and Tmean are daily maximum, minimum and mean temperatures respectively. Definitions Units Temperature Indices Frost days Number of days with T min < 0 C days Cold days Number of days with T max < 10 th percentile days Cold nights Number of days with T min < 10 th percentile days Summer days Number of days with T max > 25 C days Warm days Number of days with T max > 90 th percentile days Warm nights Number of days with T min > 90 th percentile days Diurnal temperature range Mean of the difference between Tmax and Tmin K Standard deviation of Tmean Std dev. of daily mean temperature from Tmean normal K decrease of 20 to 30 days for the northern part of the domain and 10 to 20 days in the southern part in under scenario A1B. During the period, the decrease in the number of cold days and cold nights reaches 20 to 30 days for the most of the domain except for Hudson Bay and the Great Lakes surface where the decrease is 30 to 40 days. The decrease in frost days is larger and reaches 30 to 40 days for the northern part of the domain. Under scenario A2 these changes are larger, the period beeing equivalent to the changes projected for the period following A1B scenario. Projected changes for the period under scenario A2 are even greater with 30 to 40 fewer cold days and cold nights and up 40 to 50 fewer frost days in the eastern half of the domain. The number of summer days is projected to increase by 20 to 30 days in under scenario A1B with local increases of 30 to 40 days in the south-east part of the domain, east of Lake Erie. During under scenario A1B and during under scenario A2, the number of summer days is projected to increase by 30 to 40 days in a large part of the domain surrounding the Great Lakes. During following scenario A2 this change increases to 50 days and more, especially east of Lake Erie and Ontario. The greatest changes are seen for the number of warm days and warm nights, with an increase by 20 to 40 days globally for the period under scenario A1B and up to 80 to 100 days in the south east of the domain and in the area surrounding the Great Lakes during under scenario A2. The diurnal temperature change do not indicate any strong change, with a small decrease in the nothern part of the domain and a small increase in the southern part. The standard deviation of the daily mean temperature is projected to increase with mean temperature. e. Projected precipitation changes Precipitation rate and intensity can change due to a large number of processes, including changes in water vapor content of the atmosphere, changes in cloud cover and type, and atmospheric lapse rates and stability profiles (Trenberth et al. (2003)). In general, most of North-America is projected to see an increase in precipitation under increasing greenhouse gas concentrations. Results from the AR4 climate simulations indicate that as surface temperature increases, total atmospheric water vapor increases and precipitation increases by roughly 2.2% per degree K (Held and Soden (2006)). However, precipitation changes are highly non-uniform spatially, and do not display a simple relationship with water-vapor change. For CCSM and WRF downscaled results, under both scenario, the amount of annual mean 16

18 (a) Frost Days (b) Cold Days (a) Summer Days (b) Warm Days (a) Warm nights (b) Diurnal Temperature range Fig. 8. Projected changes in temperature indices for and compared to for scenario A1B and A2. Units are in days except for the Diurnal Temperature range in C. 17

19 precipitation is projected to show a significant increase over most of Canada, Alaska, and the eastern seaboard (not shown). The projections also indicate a smaller increase in precipitation over North- California, parts of Washington state and British-Columbia, and even a decrease in the southwestern US states (Arizona, New-Mexico, Texas), which is consistent with the results seen in earlier results from multi-model ensemble means (e.g. Dai (2010), Meehl et al. (2007)). However, precipitation changes are projected to be far smaller over most of the interior of the land, and results from various models under different scenario show the greatest uncertainty in regions such as the Great Lakes Basin. Projected changes in total precipitation, rainfall and snowfall over Ontario and the Great Lakes Basin are shown in Figure 9 for the and periods compared to for the two emission scenarios. In , both scenario show a small increase (0-10%) in total precipitation in the northern part of the domain, close to Hudson Bay, corresponding to a significant increase in rainfall (20-30%) and a slight decrease in snowfall (10-20&). Projected changes are more contrasted and show less amplitude in the rest of the domain, especially in the vicinity of the Great Lakes. Projected snowfall show an overall decrease in most parts of the domain, but on local areas the amount of snowfall is predicted to stay the same or at least undergo a smaller decrease. Projected changes become more marked in the period. For both scenario rainfall show an overall increase by 10-20% in the southern part of the domain, which is up to 40% in the northern areas, and a larger decrease in snowfall, as large as 40-50% less in the southern part of the domain under A2 scenario. As with temperature, there are strong contrasts in the seasonal response of precipitation. Figures 10 and 11 show the annual cycles of temperature, heat fluxes, rainfall, snowfall and difference between precipitation and evaporation (P-E) averaged over the Great Lakes Basin. Projected changes in rainfall for under scenario A1B (Figure 10, left) show a large increase in late fall, early winter and spring, while the late winter shows little change, and the summer is projected to experience a decrease in rainfall. A reduction in snowfall is projected most of the winter except for a small increase visible in January. During these changes become more marked with a larger increase in rainfall and a decrease in snowfall at all times. Projected changes for under scenario A2 (Figure 11, left), show some differences with the previous scenario due to a a slightly lower temperature change in early winter and an earlier warming in late winter and spring. The decrease in snowfall is lower during early winter but larger during late winter, with subsequent increasing rainfall during the same period. Projected changes during are even more contrasted with a larger increase of rainfall during fall, winter and spring, a decrease of rainfall during summer and a larger decrease in snowfall at all times due to a large decrease in the number of frost days. The difference between precipitation and evaporation (P-E), shown in figures 10 and 11, is also projected to increase during fall, early winter and spring, but to show a larger decrease in summer for all scenarios. A small increase of evaporation is projected during winter, which is consistent with the warming trend in temperature and the decreasing trend in ice coverage of the lake surfaces, but is balanced by the increase in precipitation during the same period. During early spring and late fall, the increase in precipitation is also larger than the increase in evaporation. However due to an earlier annual cycle of precipitation for all scenarios, and a shift of the precipitation peak towards earlier dates, a larger difference appear from June to September with a negative P-E. Li et al. (2010) investigated the moisture budget in the Great Lakes region through reanalysis of the period and found that the evaporation over the region is insufficient to account for the total precipitation. Additional moisture is supplied by the meridional transport from the south, low-level jets over the Great Plains, which overcomes a net loss in moisture due to transport in the zonal direction by the westerlies. In both CCSM and WRF downscaled simulations, an analysis of the large scale 18

20 Fig. 9. Changes in precipitation, rainfall and snowfall (in %) for the and periods relative to the historical period ( ) over Ontario under scenario A1B and A2. 19

21 Fig. 10. Mean annual cycle for (left) and (right) periods (red dashed line) following A1B scenario and the period (blue dashed line) for the air surface temperature, sensible heat flux, latent heat flux, precipitation and snowfall spatially averaged over the Great Lakes Basin on a 7-day running average. circulation shows an increase in southwesterly winds over the Great Lakes Basin during late fall and a decrease in westerlies during winter which may account for the excess moisture during these periods. However no clear trends were observed during spring and summer. Li et al. (2010) also observed an increasing trend in evaporation for all seasons, an increasing trend in moisture convergence during spring and fall, and an increasing trend in precipitation only in the cold season with no significative changes during summer and fall. Another analysis of historical fall precipitation over the Great Lakes during the twentieth century (Grover and Sousounis (2002)) found an increasing precipitation trend and concluded that it was associated with an enhanced upper-level subtropical jet, an increase in moisture and low-level baroclinicity, and stronger low-level southerly flow. All these features contributed to an increase in southern cyclones and an increase in the frequency and intensity of the precipitation from various frontal types. As seen in Gula and Peltier (2011), snowfall would typically decrease with increasing temperature but may locally increase, especially in the areas referred to as snowbelts, downwind of the lakes, due to 20

22 Fig. 11. Mean annual cycle for (left) and (right) periods (red dashed line) following A2 scenario and the period (blue dashed line) for the air surface temperature, sensible heat flux, latent heat flux, precipitation and snowfall spatially averaged over the Great Lakes Basin on a 7-day running average. 21

23 the diminution of ice cover, earlier warming of the lakes and enhanced interactions with the atmosphere. Figures 12 and 13 show the annual cycles of ice cover, temperature gradient, heat fluxes and boundarylayer height averaged over the surface of the lakes. Annual cycles of precipitation and snowfall averaged over the snowbelt of the five lakes are also plotted Figures 12 and 13. For all periods, the longer ice-free period and the larger ice-free lake surface greatly amplify the sensible and latent heat fluxes from the surface of the lakes to the atmosphere, with an increase of the temperature gradient between the surface of the lakes and the atmosphere and subsequently the height of the boundary-layer above the lakes, which are all favorable conditions for lake-effect rain and snow. As expected, the total amount of snowfall (Figures 12 and 13) in the snowbelts only, as compared to the amount of snowfall averaged over the whole Great Lakes Basin (Figures 10 and 11), is larger by 30-40%. The amount of rainfall in the snowbelts is also larger during fall and winter but slightly lower in summer. During , under scenario A1B, projected changes in rainfall and snowfall in the snowbelt are similar to the projected changes for the whole area, except for the larger amplitude. During under scenario A2 and during under scenario A1B, which get larger temperature and ice cover changes, snowfall is projected to increase in the snowbelt during the January and February months, while a small decrease was projected for the whole area. The increase in temperature is compensated by the reduced extent and the earlier break-up of ice which enhances the lake-effect. During under scenario A2, as the temperature gets warmer, the lake-effect is still enhanced but mostly in the form of rainfall as freezing conditions are not ensure long enough. Figures 14 and 15 show changes in ice cover, temperature, heat fluxes, snowfall and rainfall for January and February. A large increase is seen for the sensible and latent heat fluxes especially over the part of the lakes which used to be entirely covered by ice but which remain more often ice free during the and periods. Changes are smaller for lake Ontario and the deepest parts of lake Huron and Michigan which were already seldom covered by ice during the historical period (see Fig. 3). f. Projected changes in high precipitation events A set of precipitation indices was selected for this study. The definitions are presented in Table 3. The days with rain and snow provide an indication of the number of precipitation and rain events. The highest five-day precipitation amount, the very wet days and the heavy precipitation days describe some extreme features of precipitation. For very wet days, the 95 th percentile reference value was obtained from all non-zero total precipitation events for It is often preferable to use indices based on percentile threshold rather than fixed threshold in Canada due to the huge scale differences in total precipitation between the different regions. The spatial pattern and magnitude of the changes for these indices are shown for the periods and for the two scenario relative to the period in Figure 16. Projected changes in the number of precipitation events usually follow the same trends than those previously shown for the annual total rainfall or snowfall (Figure 9). However there are some local differences, such as the vicinity of Lake Ontario, where the total amount of rainfall is projected to increase by 0-10% in under scenario A2, while the number of days with rain is projected to decrease by more than 6 days during the same period. Such discrepancies are also seen in the northern part of the domain, close to Hudson Bay, where snowfall are projected to slightly increase by the end of the 21 st century but where the number of days with snow is projected to decrease by days (compared to a mean historical period of days of snow per year in this region). There is a mixture of increasing and 22

24 Fig. 12. Mean annual cycle for (left) and (right) periods (red dashed line) following A1B scenario and the period (blue dashed line) for the lake-ice cover (in percent of the lake surface), air surface temperature, temperature difference between surface and 850mb isobar, sensible heat flux, latent heat flux and boundary-layer height averaged over the lakes surface; and for the rain and snowfall averaged over the snowbelt areas (filled areas in figure 14) 23

25 Fig. 13. Mean annual cycle for (left) and (right) periods (red dashed line) following A2 scenario and the period (blue dashed line) for the lake-ice cover (in percent of the lake surface), air surface temperature, temperature difference between surface and 850mb isobar, sensible heat flux, latent heat flux and boundary-layer height averaged over the lakes surface; and for the rain and snowfall averaged over the snowbelt areas (filled areas in figure 15) 24

26 A1B A1B Fig. 14. Mean changes for January and February for the period (left) and period (right) under A1B scenario, compared to the period for lake-ice cover, temperature at 2m, sensible heat flux, latent heat flux, total precipitation and snowfall. The filled areas correspond to the snowbelt location computed from historical snowfall (see Gula and Peltier (2011)) 25

27 A A Fig. 15. Mean changes for January and February for the period (left) and period (right) under A2 scenario, compared to the period for lake-ice cover, temperature at 2m, sensible heat flux, latent heat flux, total precipitation and snowfall. The filled areas correspond to the snowbelt location computed from historical snowfall (see Gula and Peltier (2011)) 26

28 Table 3. Definitions of precipitation indices used in the study. P and R denote total precipitation and rain, respectively. Definitions Units Precipitation Indices Days with rain Number of days with rain days Days with snow Number of days with snow days Highest 5-day precipitation amount Maximum precipitation sum for 5-day interval mm Very wet days (> 95 th percentile) Number of days with precipitation > 95 th percentile days Heavy P days (> 20 mm) Number of days with precipitation > 20mm days Heavy snow days (> 5 cm) Number of days with snow > 5cm days decreasing trends in the highest five-day precipitation (Figure 9), but a significant increase is projected for both scenario at the end of the 21 st century in the region of Lake Erie where a 30-40% increase is projected. Projected changes in high precipitation events (> 95 th percentile) show a greater consensus with a global increase at all periods all over the domain. The greatest changes are seen north of the domain with up to a 10 days increase under A2 scenario in An increase of 2 to 6 days of very wet days is projected on the main part of the Great Lakes Basin. Projected changes in heavy snow days show an overall decrease on most part of the domains with locally different signal in the northwest of the domain, east of Hudson Bay and south of Lake Superior, due partially to the changes in lake-ice and sea-ice cover. Figure 17 shows histograms (log-linear) of daily and 6-hourly precipitation for the whole domain over the decades (black), (blue) and (red). It is of interest to note that while there is almost no net change for precipitation rates < 10mm/day, at higher rates there are more regions experiencing high precipitation than today. 27

29 (a) Days with Rain (b) Days with Snow (c) Highest 5-day P (d) Very wet days (> 95 th percentile) (e) Very heavy P days (> 20 mm) (f) Heavy snow days (> 5 cm) Fig. 16. Projected changes in precipitation indices for and compared to for scenario A1B and A2. Units are days except for the highest 5-day precipitation amount which are in % 28

30 (a) A1B - Daily precipitation (b) A1B - 6h precipitation (c) A2 - Daily precipitation (d) A2-6h precipitation Fig. 17. Histograms with y-axis on logarithmic scale showing the daily (left) and 6h (right) precipitation over Ontario and the Great Lakes Basin. The period is shown in red, the period is shown in blue and the period is shown in black. 29

31 4. Conclusions This study has served to illustrate the fine-scale patterns of temperature and precipitation change over Ontario and the Great Lake Basin projected by the dynamical downscaling method using a regional climate model with 10-km grid spacing. The simulations were performed with the WRF regional climate model forced by a CCSM3 global climate model simulation with the use of the freshwater lake model Flake to account for the influence of the Great Lakes. Two future time periods ( and ) were simulated under SRES A2 and SRES A1B emissions scenario. Following the global warming predicted by global CCSM3 simulations for both scenarios, a comparable increase in water lake temperatures and air temperature is projected on average. Differences between the global model and the regional model appear in the vicinity of the lakes and suggest a locally larger warming. The moderating effect of the lakes in the summer season is compensated by the deeper parts of the Great Lakes seeing an earlier start of the stratified season which result in an increase of water temperature in excess of regional atmospheric warming. The number of warm events is projected to significantly increase while the number of cold events is projected to decrease with the greatest change towards extreme warm days. 30