Evaluation of the performance of various vegetation indices to retrieve vegetation cover from AVHRR data

Publisher Website
Google Scholar
Abstract
A variety of vegetation indices have been used to assess the state and monitor the evolution of the terrestrial biosphere. Early indices were easy to compute but very sensitive to soil and atmospheric effects. Modified indices with a reduced sensitivity to soil brightness changes were then proposed. More recently, new indices have been designed to be less affected by either atmospheric or soil conditions, or both. In this paper, we propose and demonstrate an objective method to evaluate, through model simulation studies, the performance of a representative sample of three such indices (NDVI, SAVI and GEMI) with respect to their capability to retrieve the fractional vegetation cover and the leaf area index from AVHRR optical data. The proposed performance criterion is based on the concept of signal to noise ratio, where the signal is defined to capture the sensitivity of the index to the desired information, and the noise is designed to measure the sensitivity of this index to undesirable perturbations. It is found that no single index is optimal under all conditions, but that improved indices (GEMI and SAVI) are generally much better than the NDVI for assessing the fractional vegetation cover. Reliable estimates of the leaf area index appear very difficult to obtain when this parameter exceeds about two because of the insensitivity of the red reflectance to multiple leaf layers.