Detection of Forest Removal in PRODES Mata Atlântica: Discussion on the Transition from Visual to Semiautomatic Interpretation

The Atlantic Forest is a global hotspot rich in biodiversity, but highly threatened by deforestation. The study addresses the PRODES Atlantic Forest monitoring system and its remote sensing techniques, as well as the challenges with the adoption of semi-automatic classification algorithms to process time series of images. We highlight the benefits of transitioning from Landsat series to high spatial resolution Sentinel-2 images, and the combination of Sentinel-2 and Sentinel-1 data to improve visualization in some areas where the landscape is impaired due to clouds. We reviewed existing approaches in the literature for semi-automated deforestation detection, including optical data fusion and SAR, discussing the need to improve the monitoring methodology. We emphasize considering local and seasonal factors to accurately detect the removal of the natural vegetation in the Atlantic Forest and we recommend further testing of algorithms based on time series images. Even though it reveals significant patterns, validation still heavily relies on visual interpretation.