Lógica fuzzy aplicada à modelagem do risco de incêndios florestais no bioma Cerrado, Brasil

Abstract

Wildfires cause environmental and socioeconomic impacts by compromising soil quality, vegetation, and natural resources. In the Cerrado biome, despite the recurrence of fire, the factors associated with large-scale wildfires are still poorly understood. The objective of this study was to develop and validate a wildfire risk (WR) model for the Cerrado biome using fuzzy logic. In Chapter 1, a literature review is presented that aims to introduce key concepts and contextualize the Cerrado in light of current environmental and socioeconomic challenges. In Chapter 2, the study on WR modeling using fuzzy logic and GIS is presented, integrating biotic, abiotic, and anthropogenic factors for the Brazilian Cerrado. The WR model was developed using the fuzzy gamma overlay method. The validation was performed through cluster analysis and the construction of a dendrogram between the WR classes and wildfire hotspot data from the Brazilian National Institute for Space Research (INPE), complemented by Kernel density analysis. The “moderate” and “high” wildfire risk classes represented 68.35% and 28.43% of the biome, respectively. There was a greater concentration of these classes in the centralsouthern and southeastern regions of the Cerrado. The variables land use and land cover, roads, population density, precipitation, mean temperature, and water deficit were identified as the most influential variables in the modeling. Kernel density analysis of fire hotspots recorded by the Brazilian National Institute for Space Research (INPE) revealed critical areas along the agricultural frontier of MATOPIBA, which was also identified as a critical area in the WR model. The dendrogram revealed two main groups: (1) “high” risk, associated with fire risk, and (2) “moderate” risk, associated with wildfire frequency, both derived from data provided by the Brazilian National Institute for Space Research (INPE). Meanwhile, the “low” and “very high” risk classes remained isolated at the adopted cut level, validating the quality of the developed model. In conclusion, the developed model proved effective in predicting and identifying wildfire risk areas in the Cerrado. In addition, it demonstrated potential for application in other regions due to its low processing cost, thereby contributing to the development of wildfire prevention and suppression strategies.