Based on precipitation data of the Qinghai Lake Basin from 2010 to 2020, the ANUSPLIN interpolation method was used to invert the spatiotemporal distribution of precipitation in the Qinghai Lake Basin.A digital watershed model of the Qinghai Lake Basin was constructed by combining hydrological, meteorological, and earth remote sensing interpretation data, and the model parameters were calibrated and verified.The model was applied to explore the impact of underlying surface changes on watershed hydrological processes, analyze the evolution laws and driving mechanisms of water cycle in the Qinghai Lake Basin, and conduct sensitivity analysis on key parameters.The results show that during the calibration and validation periods, the NashSutcliffe efficiency coefficients of the Buhahekou Station are 0.72 and 0.67, respectively, while those of the Gangcha Station are 0.81 and 0.84, indicating that the digital watershed model is suitable for simulating longterm spatiotemporal sequences of precipitation in the Qinghai Lake Basin.The vertical and the horizontal infiltration rates of soil mainly affect the peak of runoff, with the horizontal infiltration rate of surface soil being the most sensitive.The soil porosity and initial water content mainly affect the base flow of runoff, with the initial water content of the middle soil being the most sensitive.