Analysis of change detection techniques for disaster mapping using remote sensing data

Abstract

In recent years Brazil has suffered a series of environmental disasters that impaired its society, economy, and biodiversity in immeasurable levels. The use of remote sensing data rises as a fundamental source when analyzing environmental disaster impacts. Among several methods proposed in the literature, change detection techniques have shown potential use on environmental disaster investigations. In this context, this study presents an analysis with different change detection techniques applied to map the landscape modifications caused by recent environmental disasters in Brazil. The methods discussed in this paper are restricted to the unsupervised approach, which does not require a priori information about the data. Three applications were carried regarding the collapse that occurred in Mariana and Brumadinho dams, the state of Minas Gerais, and forest burning in the state of Mato Grosso. Images acquired by Landsat-8 and Sentinel-1 and 2 satellites were used in these studies. The results evidenced that such techniques, as long as parameterized and using favorable attributes, they can provide accurate maps regarding the changes from disaster events.