A Semiempirical Downscaling Approach for Predicting Regional Temperature Impacts Associated with Climatic Change

Abstract

A statistical downscaling approach is developed for generating regional temperature change predictions from GCM results. The approach utilizes GCM free atmosphere output and surface observations in a framework conceptually similar to the model output statistics approach common in the forecasting community. The appropriateness of this approach is demonstrated through a comparison of GCM and observed free atmosphere variables. Seasonal downscaling models are presented for eight sites within four community climate model (CCM) grid cells in the United States. The majority of these models are capable of explaining more than 90% of the variance in the temperature time series. The results indicate a wide range of differences between downscaled climate change predictions and grid cell?level CCM predictions.