The monthly generalized model of Xin'an River in three-conponents was established to simulate the rainfall runoff. Where genetic algorithm was used to calibrate and optimize parameters by the historical data of precipitation, evaporation and runoff in the period of 1968 to 2004, then the process of monthly flow of Duliu River upper Basin was predicted by driving the hydrological model from using GCM downscaling data under the various scenarios of RCP. Furthermore, the scenarios of the future climate change and the simulation of hydrogragh in this region were analyzed by using linear regression methods. Results indicate that in the period of 2020 to 2099 the annual precipitation will be presenting a steady tendency, the mean annual temperature and evaporation will show a trend of fluctuation in growth while the locals decline in a short period of time. The total runoff amount in this region will be lower than the historical average level, the interannual fluctuation presents a decreasing tendency and the annual distribution is not symmetrical and mainly distributed from May to October.