General Circulation Models (GCMs) have been identified as a suitable tool for studying climate change. But these models simulate climatic parameters in the large-scale which has poor performance in the simulation of processes such as rain fall-run off. Therefore, several of downscaling methods were developed. This research is presented downscaling model based on k-nearest neighbor (K-NN) non-parametric method. The model is used to simulate daily precipitation data in Ahvaz station for the next period (2015-2044) under climate change scenarios based on out puts of three General Circulation Models, including HADCM3, NCARPCM and CSIROMK3.5. The results indicate that the model has a high capacity for down scaling data. It is predicted that the frequency of storm is increased with high intensity on future period in Ahvaz station while dry spells will be prolonged.