Evaluation of evapotranspiration estimates for two river basins on the Tibetan Plateau by a water balance method

Evapotranspiration (E) at regional or basin scale is difficult to estimate. This study estimates E with a water balance method for the upper Yellow River and Yangtze River basins on the Tibetan Plateau, where in situ data accessibility is especially insufficient. Results indicate that annual terrestrial water storage change in the two basins is negligible, and basin-scale E can be reliably estimated by the difference between precipitation and runoff. Thus, four E products from Zhang—(Zhang_E), MODIS (MODIS_E), Japanese 25-year reanalysis product (JRA_E), and the newly published Global Land Data Assimilation System with Noah Land Surface Model-2 (GLDAS_E)—are evaluated against E estimated by the water balance method. GLDAS_E and Zhang_E had the best performance for the upper Yellow River basin and Yangtze River basin, respectively, with relatively small underestimation. Further analysis showed that the underestimation of GLDAS_E was mainly caused by its negative bias for precipitation, whereas the overestimation of JRA_E was due to overestimation of downward shortwave radiation. MODIS_E greatly overestimated E in both basins, which was also caused by high downward shortwave radiation flux inputs from the Global Modeling and Assimilation Office. Thus, more accurate forcing data for these products should be a future focus, since they can improve E estimates, at least for the Tibetan Plateau.