Benchmarking multi-model evapotranspiration in the Yangtze River Basin using GNSS/GRACE fusion and anthropogenic disturbance quantification

Study region: The Yangtze River Basin (YRB), China, a humid subtropical basin extending from the Tibetan Plateau to the East China Sea, supporting over 400 million people and facing frequent drought-flood extremes. Study focus: This research develops an integrated geodesy-hydrology framework to benchmark evapotranspiration (ET) in intensively managed basins. We reconstructed terrestrial water storage changes (TWSC) from 2006 to 2024 via frequency-domain joint inversion of GNSS and GRACE data using Slepian basis functions, eliminating empirical smoothing. Basin-scale ET was derived by incorporating anthropogenic fluxes (e.g., inter-basin water transfers and consumptive use) into a revised water budget equation. The geodetically constrained ET dataset was used to evaluate three global models: GLDAS, ERA5-Land, and GLEAM. New hydrological insights for the region: The joint inversion reduced TWSC spatial discrepancies by approximately 40 % in critical sub-basins such as the Jinsha River, significantly improving mass change localization. Neglecting anthropogenic disturbances led to a 6.4 % ET underestimation in the Hanjiang Basin, exceeding model uncertainties and highlighting the necessity of quantifying human interventions. Among the evaluated models, GLEAM exhibited the highest accuracy (RMSE = 1.9 mm/month) due to its physical representation of vegetation-water coupling. This study provides a transferable methodology for delineating human-natural interactions in humid basins under hydrological extremes, with direct implications for water resources management in the YRB and similar regions. • Novel GNSS/GRACE fusion via Slepian functions reconstructs TWSC with > 90 % accuracy. • Inter-basin water transfer omission causes 6.4 % ET underestimation in Hanjiang Basin. • GLEAM is optimal for humid basins (RMSE 1.9 mm/month) due to vegetation-water coupling.