Simple class-set based vegetation classification of a South Pacific volcanic island (Moorea Island, French Polynesia) using both airSAR and >ASTERdata

Abstract

This paper addresses the vegetation mapping and land use of Opunohu Valley (Moorea Island, French Polynesia) using JPL-AirSAR and MASTER (MODIS/ASTER simulator) images. We first define an original set of classes based on the relative canopy-height of vegetation, out of a well-suited RGB SAR-composite image that visually discriminates our vegetation classes. An interesting pineapple fields class (an important economic resource in Moorea island) proves to discriminate particularly from the height-related Low Vegetation class. Two supervised maximum likelihood classification maps have been processed on both the AirSAR and the MASTER images, using aerial photographs as a ground truth training set. The vegetal species included in each class as well as the classification results are discussed. Comparison of the MASTER and AirSAR based classification results leads us to propose a fusion of AirSAR and MASTER classification maps keeping the best of both worlds in order to improve the overall accuracy of the AirSAR classification.