View Zenith Angle Effects on the Forest Information Content of Three Spectral Indices

The effects of view zenith angle on the relationship between three spectral vegetation indices (NDVI, SAVI, GEMI) and forest biophysical characteristics is examined. The work uses both simulated reflectance data, derived from a Monte Carlo ray-tracing model, and multiangle Advanced Solid-state Array Spectrometer (ASAS) data. Factorial analysis of variance was applied to the simulated reflectance data set in order to quantify the view angle dependence of the relative sensitivity of an index to independent variations in forest cover, crown leaf-area-index (LAI), stand structure, and background reflectance. Overall, it was found that off-nadir viewing improved the performance of the indices for discriminating cover and LAI, when compared with nadir viewing. The potential of the indices for discriminating forest cover was low either because of strong sensitivity to the other stand characteristics, or because of nonlinear relationships with cover. At low covers, GEMI performed best in the forward-scatter direction, while GEMI had poor sensitivity in the nadir case. The potential for discriminating crown LAI was good; both SAVI and GEMI were found to be sensitive to crown LAI in the backscatter direction. In the plane perpendicular to the solar principal plane, off-nadir viewing gave improved sensitivity to cover (NDVI) and to LAI (GEMI). The results indicate that Sun–sensor geometry may be exploited to improve discrimination of forest characteristics using spectral indices, and further work is needed to test these findings with satellite data. The results are discussed in the context of off-nadir viewing sensors such as the ATSR-2.