Scientists believe that rising concentration of greenhouse gases in the atmosphere will cause changes to our climate. Global Climate Models (GCMs) have been developed to estimate this change in the future. However the coarse resolution of GCMs makes it impossible to use them in the local impact studies. Spatial resolutions of Most GCMs are generally greater than 2.0° for both latitude and longitude. Therefore, there is a need to downscale the prediction of GCMs to local and regional scales. In this study, a regression based statistical downscaling method for estimating the daily precipitation and temperature (maximum and minimum) over north of Iran is evaluated. Statistical downscaling was performed using ASD (Automated regression-based Statistical Downscaling) model and the predictors were selected with backward stepwise regression. We have used the outputs from the second generation Canadian Coupled Global Climate Model (CGCM2) to test this approach and compare results with observed temperature and precipitation from five meteorological stations in north of Iran. The NCEP (National Center of Environmental Prediction) reanalysis data over the 1961-1975 period was used for calibration of the regression model. Also climate change trend is estimated with the Sen’s slope. The results indicate the difficulty of downscaling precipitation and high performance of the model to predict temperature. Observed mean temperature has increased over 1961-2004 and the rate of increasing is much higher in minimum temperature than in maximum temperature.