Central African Forest Cover Revisited: A Multisatellite Analysis

Abstract This article proposes, through a joint analysis of a range of satellite data sets, a regional approach to the assessment of forest cover of Central Africa and a continuously updated information base on which to build a monitoring system. The following landscapes are described in detail: lowland rain forest, swamp forest, secondary formations, forest–savanna mosaic, and plantations. The separability between the vegetation types is thus established for the sensors available at a regional scale (AVHRR, ATSR, ERS-1 SAR) and over a broad range of ecotones. The performances of the different sensors illustrate the complementarity of the presently available remote sensing techniques. A regional vegetation map was produced of a part of the Congo Basin covering about 20 million ha by the combination of the best sensors used in the present study. Each vegetation type is mapped with the most appropriate sensor in terms of spectral behavior and spatial resolution. AVHRR data are used for the distinction between forest and savanna and for overall ecosystem monitoring, ATSR data have been showed appropriate for mapping the secondary forests, while ERS SAR data are reliable for mapping the gallery-forests, the plantations, and the swamp forests. A contingency matrix has been computed between the synthetic vegetation map and the national forest map of Congo-Kinshasa. The overall accuracy of the synthetic map is 74.6%. The main source of difference is related to the confusion between lowland rain forest and swamp forest. The combination of these sensors contributes thus to a new product, the thematic content and spatial detail of which has never been achieved before at the regional level.