The research on taking output of CMIP5 mode as input of downscaling to forecast regional climate conducts less, so this article used the CMIP5 RCP4.5 CanESM2 mode with higher accuracy in the scene (moderate greenhouse gas emissions) to forecast the meteorological elements in the future of Yellow River. By using the Yellow River (on the interpretation of stand above) 14 weather stations from 1967 to 2010 of monthly precipitation, temperature and NCEP reanalysis data, it chose the fit, mean relative error and relative standard deviation error as the evaluation index,used stepwise regression algorithm to screen 22 predictors and establish monthly data sequence statistical downscaling model .The model was applied to RCP4.5 under CanESM2 mode of CMIP5, which produced the change scenarios of future climate factors.The results show that simulation effect of the model on precipitation is better than that on temperature.