The Regional Atmospheric Modeling System (RAMS) is used to assess a realistic estimate of emissions by savanna fires from NOAA advanced very high resolution radiometer (AVHRR) satellite imagery during the Experiment for Regional Sources and Sinks of Oxidants (EXPRESSO) campaign. The daily amounts of gases released by biomass burning depend on burnt biomaterial, the mass of which is estimated from two independent methods: classification of vegetation formations (CV) or net primary productivity (NPP). The area of the burnt zone is derived from satellite imagery. The percent of burnt biomass or combustion factor (a) is determined by an adjustment procedure using two model runs and assumed to be 40%, a value consistent with the biomass wetness. The simulated CO redistribution is compared with aircraft measurements (tracks and profiles). Two events of 4 and 5 days are examined, where flights above forest and savanna zones have been performed. In this study we show that the main mechanisms of the chemical species spatial redistribution originates from mesoscale features. The RAMS results are in rather good agreement with the aircraft measurements (vertical profiles or horizontal legs) over forest and savanna. The difference between the CO simulated fields obtained from CV or NPP is weak (about 20% or 30%), and it is difficult to discriminate the best method because this difference is generally inferior to the gap between model and observations. One may consider this study as a preliminary approach to improve the estimate of released gases by biomass burning with the help of mesoscale modeling and the knowledge of the real sources.