Increasing the concentration of greenhouse gases has contributed to significant changes in hydrological cycle. To study the effect of climate change on precipitation and temperature, general circulation models and downscaling are the best methods for estimating these effects. In this research, the capability of linear and data-driven method for measuring the magnitude of rainfall and temperature at three synoptic stations of Birjand, Tabas and Qaen were investigated. The indexes of estimating the coefficient of determination, the efficiency of Nash-Sutcliff and the mean absolute error were used. The large-scale data in this study relates to the CanESM2 Canadian Fifth Report. The results of downscaling of precipitation by three models at three stations in Birjand, Tabas and Qaen showed that the NARX model had a weak in the average daily rainfall at the Qaen Station in the low rainfall months of the year. In the rainy months, SDSM and NARX models have been more capable at Birjand and Tabas stations. The SDSM model showed over estimation at Birjand station in the first six months of the year and NAR and NARX showed under estimation for maximum temperature. The results showed that both SDSM and NARX downscaling methods have a more acceptable performance than the NAR over the whole year. The predicted results for the near future horizons of 2021-49 showed that, while decreasing the amount of annual precipitation, the rainfall time also changes to the end of winter and early spring. The maximum and minimum temperatures will increase on this horizon.