Elaboração do mapa de uso e ocupação do solo de Pirambu/SE a partir da classificação supervisionada de imagem LANDSAT

Abstract

This work presents the preparation of the land use and occupation map of Pirambu/SE, using the method of supervised image classification from the LANDSAT08 satellite from the year 2020. This type of classification relies on the differences in spectral behaviors of the targets and the algorithm's capability to distinguish these differences. The materials and methods used were as follows: literature research/review; creation of geographic database; field surveys; pre-processing of satellite images; satellite image classification and classification validation. In the supervised classification, using the Minimum Euclidean Distance (MinDist) and Spectral Angle algorithms, the following classes were initially defined by the analyst: Water; Dune/Sand; Interdune; Vegetation; moist soil; Urbanization; Agriculture. The results of the supervised classification for land use and occupation based on the comparison of the map with field samples showed an accuracy of 58.32% for the MinDist algorithm and 55.88% for the Spectral Angle algorithm. Validation of the thematic maps with high-resolution image analysis resulted in an overall accuracy of 55% for the MinDist algorithm and 61% for the Spectral Angle algorithm. The Kappa index calculated for the 2 algorithms resulted in a moderate agreement. The thematic maps pointed to an area of almost 50% made up of agriculture and around 15% of urbanization, while natural elements such as: water, dunes and sandy strips, interdune and vegetation correspond to around 35% of the municipality's area. Even if possible inaccuracies in the analyzes are considered and the areas of agriculture and urbanization are overestimated, the values are considered high, as can be attested in the field. Therefore, subsequent temporal analyzes are recommended for environmental monitoring of the municipality.