Precipitation is a key input for hydrological modeling, and high-resolution, accurate data are essential for flood forecasting and water resource management. This study presents a Hybrid Downscaling and Multi-source Precipitation Fusion (HDMPF) framework to improve the spatial resolution and accuracy of precipitation estimates and enhance simulations of extreme precipitation and hydrological responses. HDMPF combines a Radial Basis Function network and Random Forest for downscaling, and applies Bayesian Model Averaging to fuse multiple satellite precipitation products. The fused dataset was used to drive the Grid-Xin’anjiang model for extreme flood simulations. The results show that HDMPF significantly improves spatiotemporal precipitation accuracy, increasing the KGE to 0.90–0.95 and reducing the RMSE to below 0.3 mm/h. The framework accurately reproduces precipitation cores, peak intensities, flood peaks, timing, and multi-peak hydrographs, demonstrating strong potential for improving basin-scale modeling and flood early warning.
Chao et al. (Fri,) studied this question.