Two-source energy balance schemes exploiting land surface temperature and soil moisture for continuous vineyard water use estimation

Accurate knowledge of evapotranspiration (ET) and its partitioned components, transpiration (T) and evaporation (E), is essential for improving irrigation water monitoring and management, especially in semi-arid areas. In this study, we conduct an intercomparison between the FEST-2-EWB and the TSEB models over a Californian vineyard. TSEB solves for ET using remotely sensed land surface temperature (LST) and leaf area index as key inputs, computing energy fluxes instantaneously at acquisition time and has been extensively tested over California vineyards. FEST-2-EWB employs an innovative time-continuous formulation that computes LST internally and uses it for model calibration, while simulating soil moisture and the surface energy fluxes. Thus, FEST-2-EWB does not require LST derived from satellite imagery at each timestep. Both models are compared to local eddy covariance (EC) measurements and to modeled soil and canopy temperatures and fluxes. The latent heat flux over the growing season is reproduced with mean biases of + 15 Wm-2 (TSEB) and + 46 Wm-2 (FEST-2-EWB). The estimated hourly T from both models is compared to partitioned ET fluxes from EC data obtained from three different methods (flux-variance similarity, FVS; modified relaxed eddy accumulation, MREA; conditional eddy covariance, CEC) that partition ET based on the correlation between CO2 and water vapor exchange at soil surface and plant canopy. Fairly similar RMSE values for T were found for FEST-2-EWB (68 Wm-2) and TSEB (56 Wm-2). Relative transpiration (T/ET) showed similar seasonal evolutions across both models with a modeled vine T that agreed with both the MREA and CEC estimates. However, the two models differed on the variation of T/ET within each day, with TSEB showing a morning peak and a gradual decline and FEST-2-EWB displaying a stable daytime trend, like those from the partitioning methods.