Combined with the daily meteorological data from the Kunming station, daily precipitation and runoff data from the Songhuaba Water Source Area, and the data of land use, and soil, a SWAT hydrological model was built for the water source area. Calibration and validation of SWAT were performed using the monthly runoff data from the Zhonghe Station in the Muyang Basin and the Baiyi Station in the Lengshui Basin. The calibrated and validated SWAT model was used to predict the runoff variation of the Songhuaba Water Source Area under A1F1, B1 and 15 hypothesis scenarios. The results show that the mean annual flow under A1F1 and B1 scenarios respectively decreased by 9.6% and 5.3% in the Muyang Basin while there was big change of the mean annual flow in the Lengshui Basin. In the Muyang Basin and Lengshui Basin, moreover, the mean annual flow will respectively decrease by 9.6% and 1.53% if the temperature increase by 1℃, and the mean annual flow will respectively increase by 22.55% and 13.9% when the precipitation increase by 10%.