Preliminary evaluation of remote sensing evapotranspiration models for field-scale agricultural water management in arid central Iran

Accurate estimation of actual evapotranspiration (ET a ) is crucial for effective water resource management and optimizing agricultural yields. While satellite-based surface energy balance ET a models are widely adopted, their field-scale accuracy in under-researched regions, such as Iran, remains a critical knowledge gap. This study assesses five prominent models—PySEBAL, PyMETRIC, SSEBop, PyTSEB, and ETLook (from FAO’s WaPOR v.2, L1 product)—for daily ET a estimation over an alfalfa field in the arid central part of Iran. Models were adjusted for the field using in situ weather data and Landsat-8 images, and validated against the scintillometer data. Results showed SSEBop provided the most accurate ET a estimates (KGE = 0.83), closely followed by PyMETRIC, TSEB, and PySEBAL (KGEs ≥ 0.73). Conversely, ETLook performed poorly and failed to capture spatial ET a variations. A significant performance enhancement was achieved (RMSE= 0.34 mm day⁻¹ and KGE= 0.90) by an ensemble mean of models. We further demonstrate that two-source ET a models do not inherently outperform one-source models, likely due to greater parameter uncertainty. We emphasize the importance of considering irrigation, harvest, and oasis effects for accurate model application. All evaluated models, except ETLook, were found to meet the recommended accuracies for on-farm irrigation management. This study sheds light on the selection of sophisticated field-scale ET a models for agricultural water management, while considering the dynamism of irrigation and harvest. Our findings provide critical insights for the operational application of remote sensing ET a models and promoting smart agriculture in arid agricultural settings. • Five local-sclae remote sensing ET a models were validated and compared for agricultural water management applications in Iran. • The mean ensembled ET a had the best performance, followed by SSEBop, PyMETRIC, TSEB, and PySEBAL. • Irrigation and the oasis effect significantly affect the ET a models’ performance. • Two-source ET a models, like TSEB, did not necessarily result in higher accuracy compared with one-source models. • Except for ETLook, all high-resolution models, and the ensemble mean ET a, met the accuracy requirements for on-farm irrigation management.