Along Track Scanning Radiometer (ATSR-1) data over Central Africa were used to detect areas affected by burning during the 1994-1995 dry season. A method based on temporal spectral profiles for every pixel was developed. When a time trajectory exhibits a sharp fall in the short-wave infrared reflectance with a simultaneous rise in the surface brightness temperature, the pixel is labelled as being affected by burning. The resulting map showed good agreement with contemporaneous airborne video data and Landsat Thematic Mapper (TM) data, which cover a number of the predominant ecosystems in the region. Some 52 per cent of the 1 km2 pixels in the Guinean savannahs and 28 per cent of the pixels in the Sudanian savannas within the study area had been affected by fire during the period from mid-October 1994 to early March 1995. These figures are lower, at the regional scale, than previous estimates over West Africa which had been unduly extrapolated to all African savannahs in a calculation of atmospheric emissions by fires in Africa. When combined with information on biomass loading and vegetation types, such a map is a useful input to estimate greenhouse gas emissions from savannah fires at regional level and to investigate potential impacts of biomass burning on land cover.