Multi-scale accuracy assessment of meteorological satellite precipitation products in the dry-hot valley of Jinsha River

The dry-hot valley of Jinsha River. This study focuses on the dry-hot valley of Jinsha River, using the daily precipitation observation data of two decades as a reference. To evaluate the spatial and temporal performance of five satellite and reanalysis precipitation products (IMERG, TRMM 3B42, PERSIANN-CDR, CHIRPS, and ERA5), seven statistical and three classification indicators were used. Voronoi partitioning and K-Means clustering were employed to enhance the analysis. The findings reveal: (1) On an annual scale, IMERG showed the best long-term performance, while caution was needed when using PERSIANN-CDR and ERA5. (2) The dry season was the most reliable period for precipitation estimates, with low false alarm rates and high hit rates. However, heavy rainfall events tended to be overestimated. (3) Monthly accuracy was more stable from November to April, while it decreased and became more variable from May to October. IMERG successfully captured the spatial and temporal distribution of monthly precipitation, while ERA5 showed strong fluctuations and significant overestimation. Additionally, precipitation products performed better in the 1200–1600 m elevation range, with IMERG and TRMM 3B42 excelling across all elevations, while PERSIANN-CDR and ERA5 performed poorly. (4) The IMERG product demonstrated strong agreement with observed daily precipitation during the rainy season, whereas ERA5 and CHIRPS exhibited significant overestimation. Overall, IMERG demonstrated excellent adaptability, providing valuable soil erosion precipitation data for the dry-hot valley. • The applicability of five precipitation products in dry-hot valley was studied. • Multi-scale and environmental features are taken into accounts. • ANUSPLIN interpolation, Voronoi division and spatial correlation analysis were used. • IMERG has better precipitation characterization ability than other products. • The gap of precipitation data in dry-hot valley of Jinsha River was filled.