Evaluation of six latest precipitation datasets for extreme precipitation estimates and hydrological application across various climate regions in China

The evaluation of gridded precipitation datasets is crucial for enhancing precipitation accuracy and supporting its applications. This study comprehensively evaluated the performances of six widely used long-term precipitation datasets in capturing extreme precipitation and streamflow over China using two hydrological models. These datasets include one satellite-reanalysis-gauge dataset (MSWEP V2), two gauged-based datasets (GPCC and CPC), and three reanalysis datasets (NECP-2, MERRA-2, and ERA5). The evaluation was performed at a daily timescale for the period 1982–2020. Compared with the rain gauge observations, GPCC provides the best performance in extreme precipitation estimation, followed by MSWEPV2, CPC, and MERRA-2. All precipitation datasets tend to underestimate annual maximum 1-day precipitation (Rx1) and annual maximum consecutive 5-day precipitation (RX5), while they overestimate the extremely wet days (R95p) in dry northwestern China and underestimate it in wet southeastern China. Integrating gauge data into gridded precipitation datasets enhances the accuracy of extreme precipitation measurements. For streamflow simulation, GPCC shows the best performances across most catchments regarding hydrological calibration score (Kling–Gupta efficiency, KGE), except in arid northwestern China, where MSWEP V2 performed best. The ability of precipitation datasets to capture extreme streamflow is associated with considerable uncertainties, depending on the hydrological model used, and no single dataset consistently outperforms others. Besides, the influence of hydrological model selection in streamflow simulations is more significant in dry and high-latitude mountainous regions than in wet and low-latitude regions. This study provides significant insights into the reliability of the latest precipitation datasets and their applications in hydrological modeling, which is expected to serve as a reference for utilizing these datasets.