Discrimination of tropical vegetation types using SPOT multispectral data

Publisher Website
Google Scholar
Abstract
Vegetation types were discriminated using SPOT multispectral data on Miti'aro, a tropical oceanic island in the Cook Islands, Polynesia. Vegetation categories included undisturbed and disturbed forest on limestone, scrub, marsh, and other forest vegetation (including secondary upland forest and agroforestry). Most category pairs had high separability as measured by Jeffries‐Matusita distance and Euclidean distance for training site data. However, there was some class overlap as illustrated by unsuperaised clustering and assigning spectral clusters to vegetation classes using a reference map. Cloud cover was a problem encountered in optical imaging of this maritime tropical study area.