Mapping Murundus Fields in the Rio Claro River Basin (MG) Using Google Earth Engine

Wetlands are transition zones between aquatic and terrestrial environments that are home to several animal and plant species, providing several ecosystem services. Murundus fields are typical wetlands of the Brazilian Cerrado, with an important role in local biodiversity and water regime. The present study aimed to detect the murundus fields in the Claro river basin (MG) using Sentinel 2 satellite images. For this, the Google Earth Engine platform was used to implement an algorithm based on machine learning (Random Forests) aiming to map the study area from four distinct combinations of input data. The results indicate that the joint use of all Sentinel-2 bands and the normalized difference water index was the most efficient, presenting 71.95% producer accuracy, 79.42% user accuracy and F-measure of 75.5%. Furthermore, it is observed that the combination of the near-infrared and red-edge bands result in a satisfactory F-measure index (69.39%), indicating that these spectral regions provide important information to detect wetlands. The development of technologies that assist in the detection and monitoring of murundus fields can aid in the preservation of these habitats. Due to its processing in the cloud, GEE has the potential to map wetlands over large areas.