Evaluating the precise grapevine water stress detection using unmanned aerial vehicles and evapotranspiration-based metrics

Precise irrigation management requires accurate knowledge of crop water demand to adequately optimize water use efficiency, especially relevant in arid and semi-arid regions. While unoccupied aerial vehicles (UAV) have shown great promise to improve the water management for crops such as vineyards, there still remains large uncertainties to accurately quantify vegetation water requirements, especially through physically-based methods. Notably, thermal remote sensing has been shown to be a promising tool to evaluate water stress at different scales, most commonly through the Crop Water Stress Index (CWSI). This work aimed to evaluate the potential of a UAV payload to estimate evapotranspiration (ET) and alternative ET-based crop water stress indices to better monitor and detect irrigation requirements in vineyards. As a case study, three irrigation treatments within a vineyard were implemented to impose weekly crop coefficient (Kc) of 0.2 (extreme deficit irrigation), 0.4 (typical deficit irrigation) and 0.8 (over-irrigated) of reference ET. Both the original Priestley-Taylor initialized two-source energy balance model (TSEB-PT) and the dual temperature TSEB (TSEB-2T), which takes advantage of high-resolution imagery to discriminate canopy and soil temperatures, were implemented to estimate ET. In a first step, both ET models were evaluated at the footprint level using an eddy covariance (EC) tower, with modelled fluxes comparing well against the EC measurements. Secondly, in-situ physiological measurements at vine level, such as stomatal conductance (g st ), leaf (Ψ leaf ) and stem (Ψ stem ) water potential, were collected simultaneously to UAV overpasses as plant proxies of water stress. Different variants of the CWSI and alternative metrics that take advantage of the partitioned ET from TSEB, such as Crop Transpiration Stress Index (CTSI) and the Crop Stomatal Stress Index (CSSI), were also evaluated to test their statistical relationship against these in-situ physiological indicators using the Spearman correlation coefficient ( ρ ). Both TSEB-PT and TSEB-2T CWSI related similarly to in-situ measurements (Ψ leaf : ρ ~ 0.4; Ψ stem : ρ ~ 0.55). On the other hand, stress indicators using canopy fluxes (i.e. CTSI and CSSI) were much more effective when using TSEB-2 T (Ψ leaf : ρ = 0.45; Ψ stem : ρ = 0.62) compared to TSEB-PT (Ψ leaf : ρ = 0.18; Ψ stem : ρ = 0.49), revealing important differences in the ET partitioning between model variants. Thes