Modelagem dos motivadores para ocorrência de queimadas na mesorregião Norte Goiano

Abstract

Fire is an essential natural element in many ecosystems, as it shapes vegetation and species diversification, especially in savanna areas. However, human interference has significantly altered the natural fire regime, increasing its frequency and intensity, resulting in negative environmental, social, and economic impacts. Therefore, the aim of this research is to investigate the dynamics of burned areas, explore the accumulated knowledge on fire studies in the Cerrado, and examine the causes of wildfires in the Northern Goiás Mesoregion, Goiás. The methodology involved conducting a literature review, from both bibliometric and systematic perspectives, to assess the accumulated knowledge on the topic. To model the drivers associated with environmental, economic, and social factors of fire, Machine Learning models, Random Forest, and eXtreme Gradient Boosting were employed. The main results revealed that 79% of the mapped articles were published in the last 10 years, and that researchers Carlos Dacamara and Renata Libonati stand out with the highest H-index and its variants (g-index and m-index). The collaboration network highlights the prominent roles of researchers Renata Libonati and Yosio Edemir Shimabukuro in two prominent clusters. From a systematic perspective, research mainly focuses on the impact of fire on water resources, detection of burned areas, vegetation resilience, fire regimes and spatial patterns, fire emissions, and integrated fire management. In the modeling of the drivers, the best performance, with the lowest RMSE (125.4) and highest R² (0.7411), was achieved by the model with three variables: maximum temperature, mountainous terrain, and savanna formation, meaning the independent variables explain 74% of the model. However, it is important to consider the ecological interactions between variables, as this is crucial for understanding the dynamics of burned areas in the Cerrado. Keywords: Cerrado; Machine Learning; Fire and Wildfires.