Neglecting land–atmosphere feedbacks overestimates climate-driven increases in evapotranspiration

Accurate projections of evapotranspiration (ET) are crucial for understanding hydrological responses to climate warming, which remain highly uncertain because of complex land–atmosphere interactions. Here we develop a theoretical framework to disentangle these interactions, achieving highly consistent ET projections between offline and coupled models. Our findings show that previous estimates of climate-driven ET increases have been exaggerated, primarily due to a substantial overestimation of atmospheric evaporative demand. Notably, the atmospheric conditions often assumed to drive ET are, in fact, responses to ET changes induced by soil moisture and vegetation dynamics. Neglecting these land–atmosphere feedbacks has led to a 25–39% overestimation of climate-driven global ET increases and a 77–121% exaggeration of the negative contribution from land surface changes. These biases have caused large discrepancies in hydrological projections and attributions between offline and coupled models, underscoring the importance of accurately representing land–atmosphere interactions to improve the reliability and consistency of future hydrological projections. How evapotranspiration changes with warming is not well understood. Here the authors show that when often-neglected land–atmosphere feedbacks are considered, evapotranspiration increases less than currently projected by offline models.