Systematic overestimation of evapotranspiration over irrigated areas by an offline land surface model

Abstract

Abstract. Offline Land Surface Models (LSMs) are essential for a wide range of applications, including water resource management and agricultural planning. A critical variable in these models is evapotranspiration, but its value is easily biased in irrigated areas. In fact, irrigation fundamentally alters local atmospheric conditions – cooling and humidifying the air and reducing wind speeds – factors that contribute to reducing evapotranspiration rates. This phenomenon is called “atmospheric feedback”, but is often missing or poorly represented in offline LSMs because most of the atmospheric forcings used, such as reanalyses and climate model outputs, overlook the atmospheric effect of irrigation. This leads to a tendency for offline LSMs to overestimate evapotranspiration rates over irrigated areas. In this study, the atmospheric effects of irrigation are quantified using data from the Land surface Interactions with the Atmosphere over the Iberian Semi-arid Environment (LIAISE) project field campaign. The various surface processes that influence the dynamics of evapotranspiration in response to the atmospheric feedback are then systematically investigated. The results confirm the importance of considering the atmospheric feedback in the Interactions Soil Biosphere Atmosphere (ISBA) LSM over irrigated areas in many configurations. For well irrigated crops, the average overestimation of evapotranspiration is about 25 %. Conversely, for water-stressed crops, this overestimation is negligible because of the delay in stomatal closure caused by the atmospheric feedback mechanisms, providing a compensatory effect which mitigates the overestimation. These findings highlight the need for improved representation of irrigation-related atmospheric feedback in the atmospheric forcings used as upper boundary conditions in LSMs to improve the accuracy of evapotranspiration estimates in agricultural or hydrological contexts.