A new method is described for calculating the amount of biomass burned, its type and location, and the time of burning. Active fires in 1989 were detected using daily advanced very high resolution radiometer (AVHRR) satellite imagery. The fire count was calibrated to area burned using a stratified sample of multitemporal multispectral (MSS) imagery. The calibration factor is strongly dependent on mean individual fire area, which is in turn strongly related to cumulative normalized difference vegetation index (NDVI). The best available vegetation maps for southern hemisphere Africa were combined and reclassified into functional vegetation types with a similar fire ecology. The fuel load was calculated in each 0.5¿¿0.5¿ grid square using a production model specific for each vegetation type, driven by monthly rainfall data. Multiyear fuel accumulation, herbivory, and decay were accounted for. Combustion completeness was modeled as a function of fuel mass and fuel type, established from field-collected data. The method was compared to the conventional procedure for calculating biomass burned, based on classification. The estimated amount of biomass burned in vegetation fires in southern hemisphere Africa annually is 90--264 Tg dry matter (DM) by the new modeling method and 247--2719 Tg DM by the conventional classification method. The modeling method is conservative since it does not include burning due to forest clearing or the burning of agricultural waste or domestic biomass fuels, but it is believed to be more realistic than the classification method and provides space-and-time-resolved output. The bulk of the burning occurs between June and September, with a peak in August. Half of the burning takes place in the broad-leaved, low-nutrient-status savannas which dominate the zone between 5¿ and 18 ¿S. ¿ American Geophysical Union 1996