AOGCMs predict the future climate based on an increasing Green House Gases (GHG) scenario. The temporal resolution of those models is suitable but, their spatial resolution is very coarse and the output of those models can’t be used in earth sciences such as hydrology, water resources and soil conservation. For this reason, downscaling of their outputs is necessary. At present, Statistical downscaling models mostly used in different sciences and a lot of research was carried out. Those models have different accuracy and their accuracy is depended on geographical and climatic conditions.

This research was carried out for accuracy evaluation of LARS-WG on different morpho-climatic conditions in northwest of Iran. Northwest of Iran has complex topography and climate due to intrusion of different rain bearing weather systems to the region. Firstly, daily climate data (precipitation, maximum and minimum temperature) of 7 synoptic stations was collected and their time series were created. Geographic data and climate variable were used in LARS-WG Model for each station, and then the model was calibrated and validated. The error of model was calculated for each climate variable by MAE method. The morpho-climatic data was extracted for synoptic stations and correlated with error of model for each climate variable. The results showed that the error in precipitation has significant relation with distance to grid center, whereas the error in maximum temperature is related to elevation of stations.