Detecção de hotspots de desmatamento na cidade de Canavieiras/BА, por meio da modelagem Risk Terrain Modelling (RTM), como uma estratégia para a implantação do policiamento preditivo

Abstract

This study investigates the detection of deforestation hotspots in the municipality of Canavieiras, Bahia, using Risk Terrain Modeling (RTM). The theoretical framework examines concepts such as overt policing, predictive policing, environmental criminology and criminal hotspot mapping techniques. It highlights the application of RTM to identify deforestation hotspots, adapting principles of environmental criminology to the context of forest degradation, in order to optimize monitoring and enforcement operations, contributing to environmental conservation. The study also sought to identify the risk factors that contribute to deforestation. The methodology adopted in this study consisted of dividing the study area into grids and using geospatial data related to deforestation risk factors, selected on the basis of specialized literature. These factors included slope, pasture areas, roads, among others. The data was used as input variables for the RTM. To identify the statistically significant risk factors associated with deforestation between 2019 and 2022, a binomial regression analysis was carried out. The results revealed the recurrence of deforestation risk factors, such as areas with Rural Environmental Registration (CAR), pastures and land with gentle slopes. It was observed that more than half of the deforestation alerts occurred in flat areas, and that more than 94% of the deforested area was concentrated on small plots of land, showing a positive asymmetry in the alert data, reflecting a pattern often observed in criminal occurrences. With regard to the classification of grid cells, the areas designated as high or very high risk represented a lower percentage than those classified as medium or low risk, which corroborates findings in the literature on crime. In addition, it was found that deforestation hotspots are predominantly located on the edges and clearings of forest fragments. This shows that deforestation hotspots do not occur randomly or uniformly, but follow predictable patterns, similar to common crimes. In addition, the effectiveness of the RTM model was assessed using predictive evaluation metrics. The main results indicated an accuracy rate of 67% in predicting deforestation alerts, with greater accuracy in 2020 and 2021. The Recapture Rate Index (RRI) showed that, although the overall accuracy was moderate, the year 2020 showed superior performance, with an RRI above 1. It was observed that RTM modeling proved to be robust and effective in predicting deforestation hotspots, indicating that predictive policing can be adapted for environmental conservation. This study presents a promising tool for the more efficient allocation of enforcement resources, which can contribute significantly to the protection of forest remnants in Canavieiras