High-resolution actual evapotranspiration in almond orchards by integrating physiological and TSEB models

Spatial and temporal high-resolution thermal infrared (TIR) remote sensing systems are emerging as valuable tools for monitoring crop water use and stress. Thermal cameras onboard small uncrewed aerial systems (UAS) provide high spatial and temporal resolution TIR. Reliable high-resolution TIR imagery is a crucial input for remote sensing-based land surface energy balance models in mapping instantaneous actual evapotranspiration (ET a ). Therefore, evaluating UAS-acquired TIR temperature correction and ET a upscaling method is needed to improve the accuracy of instantaneous and daily ET a , respectively. In this study, we applied a novel physiological-surface energy balance-based TIR temperature correction, a key input for the two-source energy balance (TSEB) model, in combination with a daily ET a upscaling method, to estimate high-resolution ET a at almond orchards in California. A coupled physiological model was developed and compared to in-situ physiological measurements of net CO 2 assimilation rate (A n ), stomatal conductance (g s ), and leaf temperature (T l ). The TSEB Priestley-Taylor (TSEB-PT) and contextual approach (TSEB-2T) models were applied to UAS imagery and proximal tower data, and their performance was compared with eddy covariance (EC) measurements. For ET a to be useful for irrigation scheduling, upscaling methods based on solar radiation (R s ) and TSEB time series of hourly modeled LE (latent heat flux) were evaluated using EC measurements. Results showed that the coupled physiological and surface energy balance model predicted T l in good agreement with observed canopy temperature (T c ), with the highest R 2 (0.92–0.98) and the lowest RMSE (0.84–1.6 °C). Applying physiological-surface energy balance-based TIR temperature correction to UAS-acquired TIR imagery improved LE estimates, reducing the MAE by 18.3%. The upscaling methods performed best when using instantaneous LE values between 11:00 and 13:00, with RMSE from 0.02 to 0.65 mm day −1 based on EC measurements. The TSEB-2T UAS-CMT and R s upscaling method outperformed the other scenarios, with RMSE of 0.66 mm day −1 . This study demonstrates the potential to enhance the accuracy of high-spatial-resolution ET a estimates in cropping systems, supporting precision irrigation.