TOWARD THE DEVELOPMENT OF A MULTIDIRECTIONAL VEGETATION INDEX

Vegetation-related information is critical for modeling hydrological processes. Remotely sensed spectral data provide powerful means to characterize vegetation status. However, the non-Lambertian behavior of most surfaces induces a large view/Sun angle dependence. Two approaches are possible to correct such contamination. One approach consists of using directional reflectance models; the other consists of normalizing the Sun/view angle effects directly on vegetation indices that have the advantage of being less sensitive to surface anisotropy than individual reflectances. Ground-based multiple view direction/angle measurements made over a semiarid grassland canopy at the Walnut Gulch experiment watershed (Monsoon '90 experiment) were used to develop and to validate a semiempirical model to normalize the MSAVI (modified soil-adjusted vegetation index) response to a nadir, regardless of view/direction angle. We further advanced this model to account simultaneously for both Sun and view angle variations by introducing a shadow parameterization. The results showed that this model can be used to monitor the vegetation status using a single view/Sun configuration throughout the growing season. We therefore believe that we have taken a further step toward the development of a multidirectional vegetation index.