Modeling susceptibility to forest fires in the Central Corridor of the Atlantic Forest using the frequency ratio method

Abstract

Fire is one of the main disturbances of tropical forests. Understanding the spatial and temporal dynamics of forest fires is of fundamental importance for the conservation of tropical forests. We used a frequency ratio model to identify those areas most susceptible to forest fires in the Central Corridor of the Atlantic Forest, from 2001 to 2019. We used data from the burned area of MODIS MCD64A1 to create the dependent variable grouped as climatic, topographic and human and landscape variables. The receiver operating characteristic (ROC) curve and area under the curve (AUC) were used to assess the model's performance. Land use and relief orientation were the most and least important variables in the model, respectively. The model showed good AUC values, ranging from 0.72 to 0.96, with an average of 0.81 for the study period. The average distribution of susceptibility classes was low (19.62 %), medium (24.45 %) and high (20.55 %). The northwestern region of the CAFC was the one that presented the greatest susceptibility to the occurrence of forest fires. The frequency ratio proved to be a good model for mapping areas susceptible to forest fires in an area of the Atlantic Forest.

Keywords: Deforestation; Predictive models; Remote sensing; Susceptibility and forest fires; Tropical forests.