Land surface temperature (Ts) integrated in contextual evapotranspiration models can be used to derive the water stress level of forests. One difficulty of such models implemented over forests is related to the impact of trees casting shadows on the remotely sensed Ts, which potentially hide the water stress signature. Until now, there has been no method to correct for this effect at the spatial resolution of current (Landsat) thermal sensors. This study investigates the impact of the solar zenith angle as a proxy of trees casting shadows on the water deficit index (WDI), using Landsat-7 and Landsat-8 data over a 21-km 2 area partially covered by a holm oak forest in South-eastern France, for 7 successive summers (2015 to 2021). Results are compared to in situ evaporative fraction (EF) measurements. First, a calibration of the correction of WDI for shadow effects correlates the WDI error (WDI minus 1-EF) with the solar zenith angle. The correction improves the correlation (R) and slope between WDI and 1-EF, from R=0.32 to R=0.62 and slope=0.21 to slope=0.63, for non-corrected and corrected WDI respectively. Second, a calibration of the correction method that does not rely on in situ measurements, evaluates the linear relationship between remotely sensed WDI and the solar zenith angle in dry conditions. The correction improves the correlation and slope between WDI and 1-EF, from R=0.32 to R=0.57 and slope=0.21 to slope=0.50, for non-corrected and corrected WDI respectively.