Central African forest cover revisited: A multisatellite analysis

This article proposes, through a joint analysis of a range of satellite dat a sets, a regional approach to the assessment of forest cover of Central Af rica and a continuously updated information base on which to build a monito ring system. The following landscapes are described in detail: lowland rain forest, swamp forest, secondary formations, forest-savanna mosaic, and pla ntations. 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 te chniques. A regional vegetation map was produced of a part of the Congo Bas in 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 appropr iate sensor in terns of spectral behavior and spatial resolution. AVHRR dat a are used for the distinction between forest and savanna and for overall e cosystem 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 source of difference is related to the c onfusion between lowland rain forest and swamp forest. The combination of t hese sensors contributes thus to a new product, the thematic content and sp atial detail of which has never been achieved before at the regional level. (C) Elsevier Science Inc., 2000.