Global Evaluation of Terrestrial Evapotranspiration Trend from Diagnostic Products

Abstract

Evapotranspiration (ET) represents a pivotal process interlinking the water, energy, and carbon cycles within the Earth's environmental systems. In the context of global climate change, the discernible escalation in global ET has been extensively documented since the early 1980s. Nevertheless, considerable uncertainties persist in appraising the trajectory of estimated ET trends, with the magnitude of the global ET trends revealed by individual estimates differing by over an order of magnitude. Here, we present a comprehensive comparison of 11 state-of-the-art global ET products, comparing them with water balance-inferred ET across 69 major global basins, and contrasting them with direct, long-term ET observations from 20 eddy covariance sites. Our findings underscore a generally inadequate performance of existing ET products in replicating global and regional ET trends. A notable revelation is that the majority of ET products falter in correctly identifying the sign of water balance-derived and/or eddy covariance-observed ET trends in over 50% of catchments/flux sites. For catchment/flux sites where the signs of ET trends are accurately identified by the products, there is a prevalent tendency towards underestimating the magnitude of these trends. In addition, we find these ET products generally perform better in estimating ET trends in relatively arid climates and in croplands where the vegetation cover is more uniform. Finally, we elucidate that misclassification of land use types and insufficient representation of human activities, such as irrigation, groundwater extraction, and large-scale water diversion, constitute primary sources of uncertainty in the estimated ET. These insights are poised to advance future data assimilation efforts and foster the development of more reliable ET products at both basin and ecosystem scales, offering decision-makers an informed basis for selecting appropriate ET products.