Lake Maggiore - Italy, 3-6 April 2000 Application of SPOT 4-VEGETATION data for mapping the forest-cover of MadagascarMayaux
Philippe, Gond Valéry and Bartholomé Etienne The objective of this poster is to demonstrate the possibility of updating forest-cover maps in ecosystems affected by rapid changes using VEGETATION data in a limited time. In this poster, an efficient technique for cloud decontamination of the ten-day composites is presented. A 36 composites were used in this study covering the period from October 98 to September 99. These composite images were still too contaminated by clouds and haze to allow for direct classification. Monthly images were produced in order to reduce the remaining clouds. Two different criteria of second stage compositing were tested: the maximum NDVI and the minimum SWIR. A forest-cover classification was derived from the 12 monthly composite images. The 36-band image was classified into 40 clusters using the "Isodata" unsupervised method. The algorithm based on the minimum SWIR was selected because it produced more spatially homogeneous monthly composites. Then the monthly profiles are interpreted in terms of vegetation phenology. The class labeling was done based on available field knowledge, ancillary information and visual analysis. The accuracy of the resulting map was assessed by comparison with Landsat classifications interpreted by local experts over three sites. Compared to previous similar exercises with AVHRR, VEGETATION offered 4 main comparative advantages:
|
 |
|
|