Estimation of actual evapotranspiration using remote sensing techniques and energy balance models in the lower Seyhan plain of the Eastern Mediterranean Region of Türkiye

This study provides a comprehensive analysis of estimating actual evapotranspiration (ETa) using surface energy balance (SEB) models in the Lower Seyhan Plain (LSP) in Türkiye, encompassing both non-stressed areas (Akarsu Irrigation District, i.e., AID) and stressed areas near saline lagoons of the LSP. It evaluates the applicability of SEB models (METRIC, SEBAL, EEFlux-METRIC, and GeeSEBAL) and the Kc NDVI method for computing ETa over a large irrigation scheme in four consecutive water years (2020, 2021, 2022, and 2023). Landsat data, data sets obtained from field campaigns, in-situ climatic observations from local meteorological stations within the study area, and global meteorological data were utilized to estimate ETa by SEB models and the Kc NDVI method. In this study, the results of the crop classification in the AID and the ground truth data acquired from the stressed areas were used to estimate ETa using SEB models, to evaluate the accuracy of EEFlux-METRIC and GeeSEBAL, and to investigate the relationship between NDVI and LAI for selected crops. The R-METRIC model and LandMOD ET Mapper toolbox in MATLAB facilitated robust ETa estimation. Furthermore, spatio-temporal variations for different crop types were shown on a large scale. SEB models were validated against the ETo FAO model and ETc method, focusing on metrics such as R², RMSE, MAE, and agreement index (d) to assess model performance. The findings demonstrate strong agreement between ETa estimates from SEB models and ETo-FAO, ETc methods, with the METRIC model showing slightly better performance than the SEBAL algorithm. The Kc NDVI method demonstrated a stronger correlation with the ETc method during the mid-growth stages of crops, with a Pearson correlation coefficient (r) ranging from 0.67 to 0.99 and RMSE values between 1.30 mm day⁻¹ and 0.21 mm day⁻¹, compared to the initial and harvesting stages in the AID throughout the water years 2020-2023. The proposed equations (ETa-METRIC-ETo, ETa-EEFlux-ETo, ETa SEBAL-ETo, Kc NDVI, LAI-NDVI) have the potential to apply and generalize to different crop types in semi-arid regions worldwide. Keywords: Evapotranspiration, Remote sensing (RS), Surface energy balance models, Artificial neural network (ANN), Lower Seyhan Plain (LSP)