Varying performance of eight evapotranspiration products with aridity and vegetation greenness across the globe

Abstract

The wide application of the evapotranspiration (ET) products has deepened our understanding of the water, energy and carbon cycles, driving increased interest in regional and global assessments of their performance. However, evaluating ET products at a global scale with varying levels of dryness and vegetation greenness poses challenges due to a relative lack of reference data and potential water imbalance. Here, we evaluated the performance of eight state-of-the-art ET products derived from remote sensing, Land Surface Models, and machine learning methods. Specifically, we assessed their ability to capture ET magnitude, variability, and trend, using 1,381 global watershed water balance ET as a baseline. Furthermore, we created aridity and vegetation categories to investigate performance differences among products under varying environmental conditions. Our results demonstrate that the spatial and temporal performances of the ET products were strongly affected by aridity and vegetation greenness. The poorer performances, such as underestimation of interannual variability and misjudged trend, tend to occur in abundant humidity and vegetation. Our findings emphasize the significance of considering aridity and vegetation greenness into ET product generation, especially in the context of ongoing global warming and greening. Which hopefully will contribute to the directional optimizations and effective applications of ET simulations.