To study the impact of the accuracy of precipitation inputs on flood simulation, the Dapoling Basin was selected as a small-scale representative watershed. The improvement effects of three fusion algorithms, namely, inverse distance weight (IDW), optimal interpolation (OI) and Kalman filter (KF), on the satellite precipitation products of CMORPH were evaluated, and the flood simulation capabilities of different precipitation data sources were analyzed based on the Xin'anjiang model. The results show that compared with the original data, the OI method improves CMORPH most significantly, with the median correlation coefficient raised to 0.783, and the median mean absolute error and root mean square error reduced by 49.2% and 36.6%, respectively. The accuracy of IDW method is second, and the improvement of KF method is limited. The flood simulation driven by using OI fused precipitation data is the most effective, with the mean Nash Sutcliffe efficiency coefficient improved by about 32.8% compared with the ground data, of which the simulation of post2000 flood field is improved by 57.1%, which confirms that the multi source fused precipitation data can effectively improve the flood simulation accuracy of the Dapoling Basin.