Remote sensing measurements provide a vauable means of determining the
extent of burning areas and of estimating the overall distribution of
pollutant sources (identified from experimental studies) in time and
space. This distribution has to be taken into account in the boundary
conditions of chemistry atmospheric models. Recent methods developed f
or the remote sensing of active fires in tropical or temperated forest
zones, have been found to be completely inadequate for fire detection
on West African savannas. In order to accurately estimate the active
fire distribution in the function of different sorts of West African s
avannas (Sahelian, Sudanian and Guinean) and forests, a multispectral
methodology has been developed based on NOAA/11-AVHRR satellite data,
with the purpose of eliminating as much as possible the problems relat
ed to large surface heterogeneity, confusion and bias, produced by clo
uds, smoke, haze, background emissivities, etc. Unlike other methods,
the results show that the multispectral method, in spite of its select
ivity, provides realistic results, and does not under- or over-estimat
e the number of fires that can be sensed by the satellite. Consequentl
y, this methodology is more appropriate than the simplest ones for a s
ystematic sensing of this phenomenon.