Streamflow in the Shiyang River basin is numerically investigated based on the soil and water assessment tool (SWAT). The interpolation precipitation datasets of GSI, multisource satellite and reanalysis precipitation datasets including TRMM, CMDF, CFSR, CHIRPS and PGF are specially applied as the inputs for SWAT model, and the sensitivities of model parameters, as well as streamflow prediction uncertainties, are discussed via the sequential uncertainty fitting procedure (SUFI-2). Results indicate that streamflow simulation can be effectively improved by downscaling the precipitation datasets. The sensitivities of model parameters vary significantly with respect to different precipitation datasets and sub-basins. CN2 (initial SCS runoff curve number for moisture condition II) and SMTMP (base temperature of snow melt) are found to be the most sensitive parameters, which implies that the generations of surface runoff and snowmelt are extremely crucial for streamflow in this basin. Moreover, the uncertainty analysis of streamflow prediction indicates that the performance of simulation can be further improved by parameter optimization. It also demonstrates that the precipitation data from satellite and reanalysis datasets can be applied to streamflow simulation as effective inputs, and the dependences of parameter sensitivities on basin and precipitation dataset are responsible for the variation of simulation performance.