An analysis of the performance of RCMs in simulating current climate over western Canada

The performance of eight National Center for Environmental Prediction (NCEP2) reanalysis-driven regional climate models (RCMs), seven from the North American Regional Climate Change Program (NARCCAP) and one from the Coordinated Regional Downscaling Experiment (CORDEX), in simulating the 1980-2004 climate of western Canada was assessed at a number of spatial and temporal scales. Results indicated that the RCMs were more successful at capturing the seasonal spatial distribution of mean temperature than precipitation and that inaccuracies in the spatial distribution of the summer climate moisture index were likely due to the errors in precipitation distribution and amount. All RCMs performed less well in simulating summer precipitation, most likely due to continued problemswith the simulation of convective precipitation. At the grid box scale, quantile-quantile (q-q) plots for temperature indicated that all RCMs showed very similar distributions to observed but with warm or cold biases, and errors in the simulation of a number of temperature-based extremes indices were related to these biases. For precipitation, q-q plots indicated that most RCMs overestimated precipitation totals, and while tending to follow the observed quantiles at smaller precipitation amounts, they diverged at larger precipitation totals. Performance in simulating the precipitation-based extremes indices depended largely on whether or not a RCM overor under-estimated precipitation totals - with those RCMs simulating too much precipitation underestimating the number of consecutive dry days and dry day persistence, and vice versa. Despite improvements in RCM resolution and parameterisation schemes, this work indicates that the simulation of precipitation in particular is still problematic in western Canada. This implies that scenarios of climate change constructed from RCM output require some form of bias correction to be of most use in impacts studies.