It is difficult to make reasonable predictions of climate change on regional scale by Global Climate Models (GCM) due to low spatial resolution.Now downscaling method has been widely used to make up for these defects of GCM.The temperature changes in the future in Yangtze River middle and lower reaches were predicted based on the temperature data in January and July from 1980 to 2011 statistic treatments with downscaling method combining stepwise linear regression (SLR) and principal component analysis (PCA).A monthly statistical downscaling model was formulated based on gridded data of ERA_interim from ECMWF reanalysis data and observed data,then apply it to the CMIP5 data to generate the series of temperature changes in future in the middle and lower reaches of the Yangtze River.The results showed that:(1) the simulated January and July temperature with statistical downscaling method was in good agreement with the observed temperature;(2) By the end of the 21st century,temperatures in January and July will both increase by 2-3℃ and the latter warm more intense under the different scenarios,and the increment of temperature of July will greater than that of January.