A CLOUD-BASED SAR-OPTICAL DATA APPROACH FOR LAND USE AND LAND COVER CLASSIFICATION IN A TROPICAL AREA OF BRAZIL

Understanding and measuring potential environmental impacts at local, regional, and global scales heavily relies on Land Use and Land Cover (LULC) data. Unfortunately, in tropical regions with frequent cloud cover, obtaining this information can be challenging, often leading to unavailability or outdated datasets. To address this issue, our study proposes a multisensor approach, combining SAR and optical data, to deliver accurate LULC information for Roraima, a tropical state in Brazil. By implementing our methodology within the Google Earth Engine, we were able to process large datasets with fewer power requirements compared to local processing methods. The results were promising, with an Overall Accuracy of 89.94%, and Producers' Accuracy (PA) consistently around 90% for all ten classes. Notably, this approach successfully identified critical classes in the region, including perennial crops and conversion areas. These findings contribute to a more comprehensive understanding of land use dynamics in Roraima, assisting in environmental monitoring and decision-making processes.