WETLAND CLASSIFICATION USING MACHINE LEARNING MODELS IN THE BRAZILIAN PANTANAL

The Pantanal is the largest tropical wetland in the world and covers the countries of Bolivia, Brazil, and Paraguay. Some regions within the Pantanal are part of the Convention on Wetlands or are recognized as Biosphere Reserves and World Heritage Sites by UNESCO. This work's objective is to classify surface water bodies into wetlands using machine learning techniques, through clustering techniques. Initially, the (Normalized Difference Vegetation Index) was used from a mosaic of images covering the study area from the Sentinel 2-A satellite processed on the Google Earth Engine platform. From this index, a threshold is established empirically, and the water bodies are segmented. The clustering technique is then applied to the morphological characteristics of each segmented object. The results obtained show that it is possible to categorize water bodies with unsupervised learning techniques.