The calculation formula of annual runoff change was derived based on Budyko hypothesis. The spatial distribution laws of annual mean potential evapotranspiration,precipitation,and aridity index as well as sensitivity coefficient over Yangtze River basin were analyzed. The outputs of global climate model BCC-CSM1-1under RCP4. 5 emission scenario was combined with LS-SVM statistical downscaling method to predict future temperature,precipitation and runoff changes in Yangtze River sub-basins. The observed hydrological data series of Wujiang and Hanjiang subbasins were used to verify Budyko formula for estimating relative runoff change. The results show that precipitation is a dominant factor affecting runoff change; the relative runoff change values are different in each sub-basin and the largest change value is around 10%; The relative runoff changes increase in the north bank and decrease in the south bank of Yangtze River basin during three future periods,i. e. ,2020s (2010-2039) ,2050s (2040-2069) and 2080s(2070-2099) . The proposed methodology provides a new approach for runoff prediction in large scale basin and un-gauged region.