عنوان مقاله [English]
نویسنده [English]چکیده [English]
In the recent years, many downscaling techniques have been developed for projection of station-scale meteorological variables from large-scale atmospheric variables by general circulation models (GCMs). In this study, the performance of three downscaling methods, including support vector machines (SVM), decision tree (M5) and K-nearest neighbor (KNN) methods of data mining models, were compared in downscaling precipitation simulated by NCEP general circulation models in the Kermanshah station. Simulation was performed between daily precipitation of Kermanshah station and NCEP model output parameters during the period of 1961 to 1991, and its result was tested during the period 1992-2001. The results of this study showed that the mean and standard deviation of output of the data mining models are less than the observed data and these models can't predict extreme values well. However, the KNN method gives better results than other considered methods.