Comparing evaporation from water balance framework and multiple models on a global scale

Terrestrial evaporation (ET) estimates from the water balance framework and large-scale modeling have been widely used in the evaluation and prediction of hydrological regimes. However, each method has its inherent limitations, including the external bias introduced by forcing variables, simplified functional relationships, and unconsidered human modules. A systematic comparison between water balance ET and modeled ET remains unexplored. Here, we quantify and attribute the difference between water balance estimations of ET and model-simulated ET (i.e., DET) on a global scale. We apply an unprecedentedly unique probabilistic ensemble of 84,042 DET estimates (2002–2021) based on all currently available datasets on water balance components. Satisfactory performance is found from the validation of the water balance-derived ET against several benchmarking ET products. We identify the regions with significantly positive DET in South and East Asia, Southern and Northern Africa, and southwestern parts of North America, with a global mean of 7 mm/a (5 % spread range: –2 to 16 mm/a). The patterns are primarily contributed by human water use and reservoir construction. We also report negative DET in the majority of South America, which may be related to human-induced deforestation. In addition, the seasonality of DET reflects the significant role of irrigation in regional ET dynamics. Variance analysis indicates higher uncertainties of DET in humid zones, mainly contributed by precipitation and simulated ET. Our uncertainty-constrained DET estimates have potential implications for assessing global and regional water availability, benchmarking climate and hydrological models, and developing sustainable mitigation and adaptation strategies.