Precipitation is the most active factor influencing basin water cycle. The inhomogeneous spatial and temporal distribution of precipitation has great influence on basin runoff. Therefore, the space-time variability of precipitation should be fully taken into consideration when the process of basin water cycle is simulated. This paper introduces a spatial interpolation method of daily precipitation in large-scale basin. According to the method, the reference rain gauges are selected with correlativity. If there exists the reference rain gauges significant in correlativity, the Reverse Distance Square (RDS) method is adopted to interpolate the precipitation. If not, the Thiessen polygon method is adopted. The interpolation result is compared with the result of the Countrywide Water Resources Planning in the progress. The difference is not too much which shows that the interpolation method presented in this paper is credible. Because rain gauges are sparely distributed in large-scale basin, the precision of generated precipitation of every unit is not so good if the spatial interpolation method is adopted. A downscaling method of daily precipitation whose value is more than 10mm is presented. The relation model of precipitation intensity and duration is developed which can generate hourly precipitation from daily data. According to the characteristics of precipitation in different parts of the Yellow River Basin, the basin is divided into five districts, The parameters of the model are calibrated and the model is validated with measured data in respective districts. The method has been applied in hydrological simulation in the Yellow River and is proved to be satisfying by comparingthe generated data with the measured data.