A Geographical Information System based methodology was developed in the framework of global survey of tropical vegetation using satellite data; it was tested on the forests of Sumatra with time series of NOAA AVHRR GAC and LAC data of 1982, 1988 and 1989. After some geometrical registration and radiometric optimization, these data are processed through a mono-date approach for minimizing the strong and highly variable atmospheric effects. Final results are derived from superimposing single-date classification outputs. Due to the low resolution of satellite data and the small size of local systems, adequate reliability for sampling and classification processes require the introduction of geographical information (topography, soils, bio-climates); iterative interactive procedures are used for that approach. Elevation is especially useful due to the altitudinal zonation of natural vegetation. Different classification rules based on satellite and geographical information are tested and compared for assessing the potential of NOAA AVHRR data for small-scale survey of tropical forests. Finally, the vegetation maps of Sumatra in 1982 and 1988 are obtained, and the on going deforestation rate is estimated.
展开▼
