Avaliação de técnica semi-supervisionada de análise de vetor comprimido (C2VA) em imagens de satélites para detecção de mudanças de uso e cobertura da terra

Abstract

Remote sensing image databases and Geographical Information System have the potential to act as accurate tools for environmental monitoring. Carajas Mountains are an important mineral deposit in Brazil and as environmental laws protect a great portion of this region, they have been at the core of conflicts involving human and nature interests. The biggest mining project in Brazil is active in this region (Carajas project) and this analysis aims at identifying the environmental impact caused directly or indirectly by this activity using state of the art methods. This study collects information of land-use and land-coverage from an area larger than 111, 000 km2 including five municipalities, aiming at observing the landscape intervention from a big scale perspective. Therefore, employing the resultant products of the multi-spectral approach called compressed change vector analysis, this work analyses both the environmental changes in five municipalities of the Carajas Mountains: Parauapebas, Canaa dos Carajas, Ourilandia do Norte, Sao Felix do Xingu and Tucuma. From the point of view of the applicability of the proposed tools, it was noticed that the combination of the compressed change vector analysis with the fuzzy clustering is sufficiently efficient for the detection and analysis of changes in land use and land coverage. The quantitative and qualitative evaluation metrics show good accuracy compared to the manually obtained reference images. The results show that in general the vegetative coverage was replaced by pasture lands, which in turn were replaced by urban occupations. The comparison with official statistics and by visual inspection indicate good accuracy of the present study in the estimation of vegetative cover. The presence of environmentally protected areas has prevented the increase of deforestation in the mountains, in which the observed change rates were at least 15% lower than non-protected regions.