Assessing the impacts of climate change on water resources is often based on simulations from global climate models (GCMs) that have been downscaled. Downscaling of GCMs, to improve representation over a limited region, can be done either by use of a regional climate model (RCM) or by statistical downscaling (SD) of the GCMs. SD techniques have been widely used in the previous studies for local impact assessment of climate change and is based on the fundamental assumption that regional climate is conditioned by the local physiographic characteristics as well as the large scale atmospheric state. Based on this assumption, large scale climate fields are related to local variables through a statistical model in which GCM simulations are used as input for the large scale atmospheric variables (or “predictors”) to downscale the local climate variables with the use of observed meteorological data. Choosing the proper predictors to calibrate SD models is one of the major challenges which can significantly affect the accuracy of the results. In this study, a new approach has been developed in which genetic-based clustering technique is used for automatic selection of predictors. A2 scenario derived by HadCM3 model has been used to determine the variations in local temperature over Tehran, capital city of Iran in the calibrating period of 1961-1995 and the results for the period of 1996-2005 are compared with the observed data in the meteorological station in the study area to validate the proposed methodology. The results of this study have shown that the proposed method for automatic selection of predictors can significantly reduce the needed time and costs of modeling efforts compared to the traditional way of choosing the predictors based on trial and error. Keywords: Climate Change, GCM, Genetic-based Clustering, Statistical Downscaling

Assessing the impacts of climate change on water resources is often based on simulations from global climate models (GCMs) that have been downscaled. Downscaling of GCMs, to improve representation over a limited region, can be done either by use of a regional climate model (RCM) or by statistical do...