The exploitation of natural resources is of concern because economic growth results in negative impacts on environmental balance. This study analyzed the spatio-temporal changes in land cover and land use (LULC) in the Araranguá River Watershed (ARW), southern of Santa Catarina state, south Brazil, in the period of 2016-2023. Images from the Sentinel-2A satellite were used, the RGB, NIR and SWIR 1 bands were selected and the EVI2, MNDWI, NDBI indices were applied, which resulted in the selection of eight LULC classes. The orbital images were classified using programming routines in Google Earth Engine (GEE) and validation was performed by obtaining data generated by the platform. The overall accuracy was 93% for both years assessed. The Native Forest class was the most representative and increased by 1.62% in the last seven years. The Built Area class grew the most, and Pasture/Herbaceous Vegetation class decreased by 5.6%. The results revealed slight changes in the landscape, with areas with native forests being maintained and urban expansion occurring. These data can help public policy makers and decision makers to manage the basin territory with a bias towards the conservation and preservation of natural resources.
Keywords: environmental degradation; machine learning; decision trees.