Classificação do uso e cobertura da terra em área tropical costeira

Abstract

The coastal zone is a transition zone between the terrestrial and marine environments. The interactions of variables, such as wind and tides, give high energy to this region, providing dynamism and making it vulnerable to threats. It is home to high-biodiversity ecosystems that provide fundamental ecosystem services, both locally and globally. In addition to these special aspects, with the increasing human occupation in this region, human pressure on ecosystems has also increased. Given this scenario, there is a need to define strategies for their conservation. With a view to territorial management, remote sensing products have been used in environmental studies, allowing a view of natural resources in space and time. Orbital products have been the main resource, particularly with the advent of free availability of LANDSAT mission products. On the other hand, there is a limitation in mapping land use and land cover, due to the presence of clouds, a persistent atmospheric condition in coastal tropical areas. An alternative is the use of SAR images, active sensors that operate in a different range of the electromagnetic spectrum, allowing continuous data acquisition overcoming the presence of cloudiness. Despite the potential of SAR images, this type of product requires investigation as the interpretation of its response is complex. This work evaluated the response of SAR image classification in a coastal tropical area using machine learning algorithms that have allowed high accuracy, such as Random Forest, among others. Despite the inconclusive results, which were limited by the size of the set of samples used and the quality of the segmentation, the potential of the adopted methodology can be observed, which was adapted based on similar studies.