Abstract ID: 3.11430 | Accepted as Talk | Talk/Oral | TBA | TBA

Precipitation in mountainous regions remains one of the largest uncertainties in climate and hydrological modeling, particularly due to sparse or absent in-situ observations at high altitudes. This observational gap restricts our ability to accurately quantify precipitation variability, impacting hydrological assessments. To address this, gridded precipitation datasets—including satellite-based products, global and regional reanalyses, and merged datasets—are widely used. However, their reliability in complex alpine environments remains uncertain. This study evaluates the accuracy of five precipitation datasets—ERA5-Land (global reanalysis), TPReanalysis (regional reanalysis), TPMFD and CMFD (merged datasets), and GPM (satellite-based)—in an ungauged alpine watershed of the Tibetan Plateau. Using a physically-based glacier-hydrological modeling approach (WRF-Hydro/Glacier), we assess their ability to reproduce spatial, inter-annual, and extreme precipitation characteristics and their hydrological impacts. Results show that all five precipitation datasets perform similarly in terms of spatial distribution, inter-annual variability and intra-annual distribution, while large discrepancies exist in precipitation amounts, where the watershed-averaged mean annual precipitation varying between 718 and 1593 mm. Generally, the ERA5-Land and TPReanalysis have the largest precipitation amounts, followed by the TPMFD, while the GPM and CMFD have the least precipitation amounts. In terms of the hydrological modeling, ERA5-Land and TPReanalysis simulated runoff show the best agreement with the observed streamflow with the Nash efficiency coefficient (NSE) value above 0.86 for the 2004-2018 period. In particular, the high flow is reasonably captured by those two datasets, as well as the TPMFD. In contrast, the GPM and CMFD considerably underestimate the observed runoff, especially for the high flow. Findings highlight the need for caution when using satellite-derived precipitation datasets as benchmarks in alpine regions with sparse in-situ observations. Furthermore, our results suggest that high-resolution global and regional reanalysis offer significant potential for improving hydrological assessments in data-scarce mountain environments. This study underscores the critical role of physically-based modeling in evaluating precipitation uncertainties and bridging the gap between observation and simulation in complex hydroclimates.