Avaliação de queimadas por sensoriamento remoto no Parque Nacional de Itatiaia

Abstract

The occurrence of forest fires has substantially increased over the past few years across the world with patterns of fire activity being controlled by the climate and driven by climate changes. In Brazil, the context is no different with an increase in the number of fires as well as in the extension of burned areas, negatively influencing the conservation of the Brazilian biomes especially in protected areas. In this sense, the present thesis was developed at the Itatiaia National Park and it aimed at analyzing a prescribed fire using both RGB and multispectral imaging obtained by UAS. In general, results show that the automatic classification method using RGB and multispectral images is accurate in discriminating the vegetation both pre- and post-fire (Article 1). An evaluation of the use of medium-spatial-resolution satellite imagery was also carried out to identify burned areas in the interior and in the buffer zone of the park and tested as an alternative for the refinement of data from the Fire Occurrence Records (FORs). Results indicate that these images (Landsat-8 and Sentinel-2) may be used to make the detection of fires more accurate in protected areas (Article 2). Finally, the severity of the fire and how long the change remained (fire scars in the landscape) were examined in forest fires and prescribed fires and it was concluded that the last ones generate a higher percentage of areas with low to moderate severity whereas forest fires have larger areas burned with moderate to high severity; moreover, this analysis allowed us to infer an estimated time for the recovery of the burned vegetation in each area studied (Article 3). Thus, by obtaining and improving these data, it is possible to better understand the ecological role of the fire and it serves as a subsidy to managers of conservation units for a better planning of actions related to an integrated management of fire.