Prediction of local extreme climatic events, particularly days of heavy rainfall and high temperature which can lead to flash flooding and droughts respectively, is an important study in the present day context. These extreme climatic events can cause devastative effects in agriculture, water infiltration, soil, public hygiene, industry, economy of the country etc. In the future, the rainfall in the region of Karkheh basin (Iran) is expected to be less, but there may be more intense rainfall events. The temperature and rainfall are important parameters for water planning and management. Hence, the climatic change impact studies on different systems such as water resources can lead to more optimal water resource management and planning. On a large scale, General Circulation Models (GCMs) are able to simulate reliably the most important means features of the global climate. The main disadvantage of GCMs is low spatial resolution (2.50). Their results are representatives on a large scale, but not on a regional or even a local scale. So these models should be downscaled to study at station scale. For statistical downscaling, 26 predictors are used, but before downscaling the best predictor should be selected. The present paper compares different methods and highlights the steps followed to select the best predictor in Karkheh basin (Iran).