Revisão Sistemática de Literatura sobre o Uso de Sensoriamento Remoto e Algoritmos Baseados em Machine Learning no Mapeamento de Florestas Tropicais Secas

Tropical dry forests are a forest ecosystem notably marked by climatic seasonality and predominance of tree species that present marked deciduousness during the dry season and great perenniality during the wet season. These environments are home to a rich diversity of fauna and flora, playing a crucial role in the quality of life of millions of people by providing essential ecosystem services. Due to their occurrence in fertile soils, there have been increasing records of anthropic pressures on their forest remnants for conversion to agriculture. For this reason, the demand for environmental monitoring actions has been growing. In this sense, the use of Machine Learning (ML) methods applied to remote sensing of tropical dry forests has stood out in recent literature. Thus, the present study proposed, through a systematic literature review, a quantitative evaluation of the main remote sensing and ML algorithms used in recent years, so that it was possible to obtain an overview of current methods, as well as promising topics for future research. As main results, it was found that the main approaches have prioritized the use of the Random Forest algorithm and data from optical sensors. By carrying out this review, it was observed that opportunities for future research lie in the analysis of the integrated use of SAR and optical sensors, as well as in comparative evaluation between different ML algorithms, so that new methods can be evaluated for progress in the monitoring and conservation of dry tropical forests.