An algorithm for identifying individual fires from the Modis burned area data product is introduced for southern Africa. This algorithm gives the date of burning, size of fire, and location of the centroid for all fires identified over 8 years in Africa south of the equator. The results are compared with other available spatial information on fires, and tested to see whether they give reasonable results in a range of different fire systems in the region. Initial results indicate that humans can have massive impacts on fire size distributions and on the area burned by large fires. This information on fire number and fire size can be used to ask important questions on the effect of increasing human ignitions, and the barriers to fire spread in savanna systems. Keywords: Algorithms, Fires, Environmental factors, Human factors, Image processing, Ignition frequency, Fire size, Fire number.