Techniques for spatio-temporal analysis of vegetation fires in the tropical belt of Africa

Biomass burning of forests and savannas is a phenomenon of continental or even global proportions, capable of causing large scale environmental changes. Satellite space observations, in particular from NOAA-AVHRR GAC data, are the only source of information allowing documentation of burning patterns at regional and continental scale and over long periods of time. This paper presents some techniques, such as clustering and rose- diagram, useful in the spatio-temporal analysis of satellite derived fire maps to characterize the evolution of spatial patterns of vegetation fires at regional scale. An automatic clustering approach is presented which enables us to describe and parametrize spatial distribution of fire patterns at different scales. The problem of geographical distribution of vegetation fires with respect to some location of interest, point or lines, is also considered and presented. In particular rose-diagrams are used to relate fire patterns to some reference point, as experimental sites of tropospheric chemistry measurements. Different temporal data-sets in the tropical belt of Africa, covering both Northern and Southern Hemisphere dry seasons, using these techniques were analyzed and show very promising results when compared with data from rain chemistry studies at different sampling sites in the equatorial forest.