Disentangling sources of errors in simulated wheat evapotranspiration

Accurate simulation of crop evapotranspiration (ET) is critical for water management and for understanding the impacts of climate variability and change. Yet evidence on model skill in simulating ET remains limited, and the sources of error remain poorly understood. Daily simulations from 17 wheat crop models were evaluated against field observations from two sites in semi-arid and Mediterranean environments under water-limited and irrigated conditions. Absolute ET errors were larger under irrigation, but this contrast was substantially attenuated after normalization by observed ET. The larger absolute errors under irrigation were mainly associated with a greater lack-of-correlation component rather than stronger systematic bias, indicating greater day-to-day mismatch between simulations and observations. Differences among ET- demand approaches also tended to be more evident under irrigation, with Priestley–Taylor-based models tending to produce larger errors, although the broad method classes also differed in other model components. Mixed-effects attribution showed that the conditional associations between ET error and climatic factors differed between the two water-supply datasets. Under irrigation, ET simulation error was associated with vapour pressure deficit (VPD), wind speed, radiation, and the component of maximum temperature not linearly explained by VPD, whereas under water-limited conditions meteorological attribution was weaker and observed soil–water status provided additional information. Root-zone soil moisture provided greater model support than shallow or deep soil moisture and modified the association between VPD and ET error. These findings identify atmospheric-demand representation and root-zone water availability as priorities for improving the evaluation and development of crop ET simulations.