Evaluating multi-source precipitation datasets for hydrological applications in ungauged alpine region of Tibetan Plateau

An alpine watershed in the Tibetan Plateau. Assessing the accuracy of gridded precipitation datasets (e.g., reanalysis, satellite and merged datasets) remains challenging due to insufficient ground-based validation. This study proposes an approach to evaluate the accuracy of gridded datasets over ungauged alpine regions using physically-based hydrological modeling. The spatial and temporal characteristics of multi-source precipitation datasets (i.e., ERA5-Land, TPReanalysis, TPMFD, CMFD and GPM) are first intercompared. The accuracy of five datasets, are then evaluated through the model simulated streamflow using the Weather Research and Forecasting-Hydrological Glacier modelling system (WRF-Hydro/Glacier). All datasets exhibit similar spatial distribution, inter-annual variability and intra-annual distribution; while large discrepancies exist in precipitation amounts. For the watershed averaged annual precipitation, ERA5-Land and TPReanalysis have the largest precipitation amounts (>1450 mm), followed by TPMFD, while GPM and CMFD have the least (<725 mm). For hydrological performance, simulated streamflow driving by ERA5-Land and TPReanalysis match observed streamflow most closely, achieving Nash-Sutcliffe efficiency coefficient (NSE) being above 0.86 for the 2004–2018 period. These two datasets and TPMFD also capture high-flow events well. Conversely, GPM and CMFD considerably underestimate streamflow time series. This study emphasizes caution when drawing conclusions based on sparsely gauged or satellite-derived precipitation datasets in Asian alpine regions and highlights the potential of the high-resolution reanalysis to improve hydrological studies in the Tibetan Plateau. • Physically-based hydrological modeling evaluates gridded precipitation accuracy. • All precipitation datasets show similar patterns, but vary in amounts. • ERA5-Land and TPReanalysis excel in discharge modeling, CMFD and GPM underestimate. • Caution is needed in interpreting evaluations from sparse gauges. • Global and regional reanalysis data show potential for hydrological simulation.