Daniel Farkat Alves Fontes, a graduate student representing INPE, successfully presented his master's proposal titled "Comparative Approach for Burn Scar Detection in the Pantanal Using Sentinel-2 Imagery and Machine Learning Techniques." The presentation addressed the increasing occurrence of wildfires in the Pantanal, a Brazilian biome recognized as a UNESCO World Natural Heritage site, particularly during the dry season.

Despite its ecological importance, there is a lack of studies that use machine learning techniques to automate the detection of changes caused by burn scars in the Pantanal. This master's dissertation proposes a comparative approach using remote sensing imagery from the Sentinel-2 satellite, combining spectral bands and indices to test different artificial intelligence techniques for detecting changes caused by burn scars. Established methods from the literature, such as Random Forest and Siamese U-Net, were tested, achieving over 90% accuracy, recall, and F1-score on a pair of scenes containing a large wildfire with few clouds.

The results demonstrate the potential of these techniques for automatic identification of burned areas and highlight the feasibility of creating a dedicated dataset for the Pantanal. This dataset could contribute to the development of specific protocols to assist public agencies in monitoring and mitigating environmental disasters in this biome.

The defense was chaired by Dr. Valdivino Alexandre de Santiago Júnior, with the guidance of Dr. Gilberto Ribeiro de Queiroz and Dr. Fabiano Morelli, all from INPE, and external member Dr. Rogério Galante Negri from UNESP. The panel provided invaluable feedback, which contributed to the success of the defense.