Mapeamento das áreas de garimpo em terras indígenas Munduruku utilizando modelo baseado em conhecimento e integração multisensores

Abstract

The consequences of mining activity in the Amazon, which is responsible for a third of Brazilian mineral extraction, cause a transformation of the landscape and nature, as well as several social problems. Considering this problem, this work proposes a knowledge-based classification model that best corresponds to the identification and mapping of mining areas, particularly in indigenous lands, and, consequently, verifies the expansion of these areas in the years 2010, 2016, 2018 and 2021. The Munduruku Indigenous Lands are among the lands that most suffer from the pressures of illegal mineral exploration and speculation by large mining companies. For the temporal analysis, images from the satellites Landsat 5, Landsat 8 and CBERS 04A were used, thus characterizing a multisensor study. As part of the modeling, the supervised classification algorithms Random Forest, Support Vector Machine and Geographic Object-Based Image Analysis were tested, and spectral indices such as Normalized Difference Vegetation Index (NDVI) and Normalized Difference Water Index (NDWI). The results found in this work point to a considerable growth in the mining areas. Compared to the year 2000, in 2010 there were no signs of new areas of mining activity, in 2016, an area of more than 700 transformed hectares was mapped, which corresponds to a growth of more than 15 times the area of 2010 From 2016 to 2018, there was a growth of 66%. From 2018 to 2021, the growth was 525%. Considering from 2016 to 2021, the growth was 936.6%, that is, in five years the areas containing mining activity in the region had an increase of almost ten times.