Bayesian Model Averaging for Satellite Precipitation Data Fusion: From Accuracy Estimation to Runoff Simulation

Precipitation plays a vital role in the hydrological cycle, directly affecting water resource management and influencing flood and drought risk prediction. This study proposes a Bayesian Model Averaging (BMA) framework to integrate multiple precipitation datasets. The framework enhances estimation accuracy for hydrological simulations. The BMA framework synthesizes four precipitation products—Climate Hazards Group Infrared Precipitation with Station (CHIRPS), the fifth-generation ECMWF Atmospheric Reanalysis (ERA5), Global Satellite Mapping of Precipitation (GSMaP), and Integrated Multi-satellitE Retrievals (IMERG)—over China’s Ganjiang River Basin from 2008 to 2020. We evaluated the merged dataset’s performance against its constituent datasets and the Multi-Source Weighted-Ensemble Precipitation (MSWEP) at daily, monthly, and seasonal scales. Evaluation metrics included the correlation coefficient (CC), root mean square error (RMSE), and Kling–Gupta efficiency (KGE). The Variable Infiltration Capacity (VIC) hydrological model was further applied to assess how these datasets affect runoff simulations. The results indicate that the BMA-merged dataset substantially improves precipitation estimation accuracy when compared with individual inputs. The merged product achieved optimal daily performance (CC = 0.72, KGE = 0.70) and showed superior seasonal skill, notably reducing biases in autumn and winter. In hydrological applications, the BMA-driven VIC model effectively replicated observed runoff patterns, demonstrating its efficacy for regional long-term predictions. This study highlights BMA’s potential for optimizing hydrological model inputs, providing critical insights for sustainable water management and risk reduction in complex basins.