Utilização do algoritmo de árvore de decisão para mapeamento do uso e cobertura da terra da área de contribuição de uma barragem localizada na Bacia Hidrográfica do São Francisco

Abstract

The aim of this study is to map the land use and cover of the contribution area of the Bico da Pedra dam through the decision tree algorithm. For this, the SRTM elevation model and two images (summer and winter) from the Landsat 8 satellite OLI / TIRS sensor, year 2017 were acquired. Then, the NDVI and SAVI indices were calculated, and then a set of eleven variables in the decision tree and in a land use and land cover map with nine classes: cerrado, montane deciduous forest, fields, water, pasture / exposed soil, crops, rocky outcrop, tailings dam and urban. The most expressive class verified in the study area was the vegetation corresponding to almost 80%. In addition, the presence of crops close to the source of the Gorutuba river was verified. The class of pasture / exposed soil presented in almost half of its samples as mistaken. The accuracy of the classification was satisfactory and the Kappa Index excellent. Therefore, the results presented allowed to acquire information on land use and cover capable of assisting in the management and planning of water and environmental resources in the area of contribution of the Bico da Pedra dam.