Detection of burned forests in Amazonia using the Normalized Burn Ratio (NBR) and Linear Spectral Mixture Model from Landsat 8 images
Year: 2015
Type: article
Abstract: Wildfires represent a major disturbance factor leading to environmental changes with local or regional impacts. In the Amazon, although fire is associated with several land-practices, the long dry season in some regions, especially during extreme droughts, makes vegetation more susceptible to uncontrolled fires. Furthermore,the burn of biomass is a considerable source of atmospheric pollution, including carbon dioxide, a major greenhouse gas. Due to the large geographical extent of fires at regional and global scales, remote sensing approaches became relevant in the last decades. In this paper, we compare two different methodologies of fire detection for Amazon region and evaluate possible spectral confusions these two methods can generate on the analysis. We compared the Normalized Burn Ratio (NBR) and the Linear Spectral Mixture Model data extracted from high resolution satellite Landsat 8 image. Our results indicate that the detection of burned forests areas using the LSMM index performs better over fragmented landscapes, with a index Kappa of0.68 and 0.66 against a index Kappa for NBR of 0.52 and 0.52, in the study sites B and C respectively. Considering areas less fragmented as study site A in this study, both methodologies showed the same Kappa value (0.88). Thus, considering the complexity of Amazonian landscapes, which encompass both high and low fragmentation areas, the LSMM index is likely to perform better in the detection of burnt forests than the NBR index.
Institution: Instituto Nacional de Pesquisas Espaciais
Language: English
Topic: Fire effects on ecosystems
Subfield: Global and Planetary Change
Field: Environmental Science
Domain: Physical Sciences
SDGs: Life on land, Climate action
Open Access status: green
OpenAlex ID: W2186837672