Abstract
This study presents a high-resolution (1 km-daily) gridded dataset of potential evapotranspiration (PET) for 2004–2022 across Europe and the Mediterranean. PET estimates are derived from a combination of ground-based meteorological data and remote sensing products. Four PET models are tested, ranging from simple approaches, i.e. the temperature-based Hamon and Priestley-Taylor (PT) models, to more advanced formulations, including the calibration-free Priestley-Taylor (mPT) model and the Penman-Monteith FAO model, both of which explicitly account for aerodynamic and radiative influences.
Evaluation against 38 FLUXNET sites and triple collocation analysis with satellite-based PET products demonstrates that FAO and mPT outperform other models across more than 80% of the study domain, with higher accuracy in grasslands, croplands, forests, shrublands, and wetlands. These models also exhibit consistent performance across climate zones, particularly excelling in arid steppe and temperate regions.
At the river basin scale, daily crop coefficients are incorporated to refine PET models based on crop growth phases, and these PET estimates are integrated into a hydrological digital twin model of the Adige River basin, a heavily human-influenced watershed in northern Italy. Simulations using FAO and mPT better reproduced key components of the water cycle. Modelled actual evapotranspiration (AET) exhibits a high correlation (0.78) and low RMSE (1.15 mm. day−1) compared to ground-based observations from two FLUXNET sites within the basin.
This high-resolution PET dataset represents a valuable resource for water resources management and regional-scale agricultural applications across Europe and the Mediterranean.