Remote sensing of burned areas in southern Africa. Contribution to reduction of uncertainty in fire incidence estimates

Doutoramento em Engenharia Florestal - Instituto Superior de Agronomia The broad geographical range and temporal dynamics of the areas affected by biomass burning require the use of remote sensing for quantitative assessment of the process. Due to the complex factors that interact in the tropical regions, the “Southern Africa Regional Science Initiative” (SAFARI 2000) was developed to investigate the role of biomass burning on the savannas of southern Africa, its impact over the atmosphere, climate and ecology. This was also an opportunity to explore the adequacy of the new MODIS sensor, for burnt area mapping and to produce accurate estimates of other parameters involved in the evaluation of the emission estimates from vegetation fires. Under this framework, two campaigns were developed, in Zambia and Mozambique, to evaluate the spectral and spatial ability of MODIS to detect burnt areas and improve the current burnt area extent and biomass burnt estimates at the regional level. In this thesis we showed that MODIS data has the potential to improve these estimates. Factors such as climate, vegetation and anthropological are also determinants of fire incidence, and understand the pyrogeography is important to incorporate this information on burnt area classification algorithms and biomass burning models. We found non-stationary fire environment relationships in Africa.