Assessment of transpiration in different almond production systems with two-source energy balance models using high-resolution aerial imagery.

The almond production has increased by doubling their hectares under irrigation treatments in Spain. In a context of water scarcity, the estimation of Evapotranspiration (ET) and its components, Transpiration (T) and Evaporation (E), are key variables to monitor and manage the water resources. High-resolution ET can be retrieved from surface energy flux modeling, such as a Two-Source Energy Balance (TSEB) model, using an Unmanned Aerial System (sUAS). sUAS equipped with Thermal and Multispectral cameras allows us to obtain the main parameters required in TSEB. Currently, there are no studies that evaluate the T obtained with TSEB Priestley Taylor (TSEB-PT) and TSEB-2T models in tree-scale almonds under different irrigation treatments (IR) and production systems (PS). In this context, we evaluated the T retrieved with TSEB-PT and TSEB-2T models using Sap Flows sensor in trees with three PS, Open Vase with Minimal Pruning (OVMP), Central Axis (CA) and Hedgerow (HGR), and three levels IR, Full Irrigation (FI), Mild Stressed (MS) and Stressed (SS). Five flights were conducted from March 2021 to July 2021 to analyze the almond growing season with an aircraft equipped with a thermal and multispectral camera. Leaf area index (LAI), stem water potential (Ψ_stem) and Fractional Intercepted Photosynthetically Active Radiation (fIPAR) was also measured concomitant to image acquisition. PS presents significant differences in fractional canopy cover (F_C), tree height (H_C), LAI and Sap Flow transpiration (Tsf). The two TSEB models show a generalized overestimation with a BIAS of 0.99 and 1.22 for TSEB-2T and TSEB-PT respectively. TSEB-PT presented worse statistics and R2 decreases in the more intensive production system. HGR has equal or greater LAI but lower F_C, which would imply an overestimation of canopy temperature (T_C) by the PT method. This is in addition to the difficulty of setting the PT coefficient according to the context of the crop. The overestimation in both models could be associated with an error in Campbell (1998) Radiative Transfer Model used to estimate transmittance, which has an error of 0.14 RMSE and 0.12 BIAS compared with fIPAR. Our results suggest the use of TSEB-2T with high resolution images considering the current available technology that allows us to estimate T_C and T_S separately, especially in intensive or super-intensive almond crops. To improve the T estimation, it is recommended to use in situ PAR measurement to decrease the influence of LAI measurements on the models.