Sensitivity of the Statistical DownScaling Model (SDSM) to Reanalysis Products

Numerous general circulation models (GCMs) have been designed by climate centres to predict future climate. An outstanding issue with the use of GCM output for local applications is the coarse spatial resolution. To produce accurate daily predictions of future climate variables at the regional scale, the Statistical DownScaling Model (SDSM) is a commonly used downscaling technique. The SDSM statistically identifies relationships between large-scale predictors (i.e., GCM) and local-scale predictands, using a multiple linear regression model. Reanalyses, such as those produced by the National Centers for Environmental Prediction/National Center for Atmospheric Research (NCEP/NCAR) and the European Centre for Medium-range Weather Forecasts (ECMWF), are important components for the structuring of the SDSM as they supply the predictor values for the calibration and validation of the model. It is well known that the reanalysis products contain biases which may subsequently affect the development of downscaling scenarios when used with the SDSM. In this paper, separate downscaled precipitation and temperature scenarios were generated using the SDSM with the calibrations and validations derived from two different reanalyses for a climate station in southern Ontario. From these comparisons, we have identified statistically significant differences between the two time series. Therefore, it is clear that choice of the reanalysis used to calibrate the SDSM can significantly affect the downscaled scenario over a region evaluated in southern Ontario.