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

Abstract

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 problems with 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 over‐ or 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.

Projections of climatic extremes in a data poor transboundary river basin of India and Pakistan

“…It is apparent that the magnitude of improvement of statistics is large for the rainfall compared to maximum and minimum temperature. The inadequate model physics or coarse resolution of the driving GCM led to the overestimation of rainfall for raw RCMs (Tripathi and Dominguez, 2013;Olsson et al, 2014;Barrow and Sauchyn, 2017;Rajbhandari et al, 2017) especially over rugged topography (north east) and caused large differences in statistics for raw RCMs and range for correction. The narrow range of improvement for maximum and minimum temperature is because of high functional relationship between local and large-scale driving parameters.…”

Projecting Canadian Prairie Runoff for 2041–2070 with North American Regional Climate Change Assessment Program (NARCCAP) Data

“…The higher resolution of RCMs, vs. that of GCMs, offers greater topographic complexity and allows finer scale atmospheric dynamics to be simulated, thereby providing a more adequate method of producing the information needed for regional impact studies (Poitras et al. ; Barrow and Sauchyn ). We used nine nested RCM/GCM combinations (Table ).…”

Uncertainty in climate projections and time of emergence of climate signals in the western Canadian Prairies

“…Analyses of model biases relative to historical observations and reanalysis data are relatively common, (see, e.g., Chapter 9 of the IPCC Fifth Assessment Report [IPCC AR5; Flato et al, ] and a recent paper on the capacity of regional climate models to simulate the climate of western Canada [Barrow and Sauchyn, ]). Less common is research which partitions the variance among model projections between signal, the simulated response to greenhouse gas forcing, and noise, the natural internal variability.…”