Dense gauge observations-based evaluation of gridded precipitation in southwest China and its implication for future data development

Abstract

Southwest China, characterized by complex topography and pronounced land surface heterogeneity, demands high-quality precipitation data for land surface processes monitoring and modeling. However, the quality of existing gridded precipitation products in this region remains unclear. Using high-density meteorological station observations from 525 national stations and 13,713 regional stations covering 2016–2020, this study comprehensively evaluates eight mainstream gridded precipitation products, including three satellite-based (GSMaP-Gauge, IMERG-Final, GSMaP-Mvk), two model-based (ERA5, ERA5-Land), and three merged (MSWEP, CMFD, GLDAS) products. Results suggest that: (1) none of the products possess satisfactory performance (KGE < 0) over the entire region, and the uncertainty in precipitation estimation—reflected by inter-product discrepancies—increases significantly (from ∼150 mm/y to ∼400 mm/y) with terrain complexity (from ∼25 m to ∼1000 m of sub-grid heterogeneity); (2) all products can reasonably capture seasonal variability but fail to characterize extremely heavy and light rainfall events, and model-based products exhibit relatively better precipitation detection capability (e.g., with POD >0.90 and CSI > 0.50 for ERA5 and ERA5-Land); (3) while most products exhibit relatively favorable consistency with observations from national stations, their performance is remarkably degraded over regional stations (e.g., monthly KGE values decrease from >0.50 to <0 for all products). Additional analyses reveal that model-based products (e.g., ERA5 and ERA5-Land) can better depict the spatial distribution of precipitation in complex and sparsely monitored regions, and the incorporation of gauge observations can further mitigate abovementioned uncertainties (e.g., with Bias <60 mm/y and RMSE <345 mm/y for GSMaP-Gauge and MSWEP).