Downscaling Global and Regional Climate Models

Statistical Downscaling (SD) methods were first developed for applications in weather forecasting. Numerous methods are now in operation across the world. Since the end of the 1990s, these methods have been used intensively to develop high spatial and temporal resolution climate change information. Climate change scenarios are mainly based on the results of Atmosphere-Ocean Global Climate Models (AOGCMs) and more recently those of Regional Climate Model (RCM) outputs, which operate at horizontal resolutions of 300-km and 45-km, respectively. More reliable information at much finer scales, utilising the appropriate SD approach, is essential for decision-makers and planners tasked with adaptation to climate change. Impact and adaptation solutions are highly demanding in terms of topographic resolution and the representation of physical processes, and neither AOGCMs or RCMs can currently meet those needs. We will therefore require SD applications. Until now, SD methods have mostly downscaled output from AOGCMs, but there is no reason why SD methods could not be applied to higher resolution models. This paper investigates the reliability of atmospheric input variables when used in the SD process, from both AOGCMs, global and regional reanalysis products, and RCMs. This allows us to evaluate the potential added value from particular single site regression-based SD approaches, by comparison with the use of raw AOGCM and RCM outputs over various areas across Canada. This work also investigates the ability of the SD scheme to reproduce observed trends and variability within the predictand under consideration. New developments within multivariate and multisite SD methods are also suggested through on-going projects and collaboration between Environment Canada and various universities across Canada.