To meet the needs of hydrological models for high-precision precipitation data, with the humid Tunxi Basin and the semi-humid Zijingguan Basin as the study areas, a multi-source precipitation merging algorithm based on external drift kriging was constructed to merge the measured precipitation data at stations and the IMERG satellite precipitation data, generating merged precipitation data with high precision and high spatiotemporal resolution. The WRF-Hydro hydrological model was driven by the merged precipitation data to conduct runoff simulations in the study areas, and its applicability in flood forecasting for small and medium-sized rivers was evaluated. The results show that taking the measured precipitation data at stations as the benchmark, the satellite precipitation data exhibits a serious underestimation of precipitation intensity in the Zijingguan Basin, and the evaluation indicators of the merged precipitation data are better than those of the original IMERG satellite precipitation data. Furthermore, the WRF-Hydro model driven by the merged precipitation data can better simulate the flood process than that driven by the IMERG satellite precipitation data, and the model shows stronger applicability in the humid basin than in the semi-humid basin.