Multidirectional remotely sensed optical and thermal images acquired within a row cotton crop in Montpellier (France) were used to test the opportunities and limitations of an existing water stress index, the Water Deficit Index (WDI, based on the trapezoid approach). The WDI was applied with multidirectional crop surface temperatures (Ts) and reflectance data acquired on a row-cotton crop with different water and cover conditions from 11 different view angles in the east/west plane. This data set allowed a biophysical analysis of this index both inside and outside its validity domain, initially limited in terms of Ts measurements in a [ - 20°, + 20°] view angles interval around nadir. Results showed that the WDI was a robust approach, since its calculation is based on the relationship between crop cover and Ts However, it yielded some directional errors in the case of sparse crops even in its validity domain where the relative variation of WDI between oblique angles and nadir could reach 14% (and more than 40% for larger view angles). The same degree of variability was observed between WDI values estimated on a same plot at two different times in a given day from a nadir observation. In a large range of crop heterogeneity, hourly sunlit soil fraction presented a stronger influence on Ts than the total soil fraction. However, by adapting the view angle to daytime measurements and crop structure, it seemed possible to overcome sunlit soil effects. These experimental results were tested and extrapolated using a 3D crop energy balance model (Thermo). It allowed simulations of directional Ts measurements according to various sun/sensor angular configurations, crop structure, and water status characteristics. This confirmed the limitations of the trapezoid method both within and outside its validity domain. Moreover, Thermo allowed the computation of a "directional" WDI accounting for angular and hourly sunlit soil effects variability on Ts. The interest of adapting the