REMOTE SENSING ROLE IN ENVIRONMENTAL STRESS ANALYSIS: ЕАST SERBIA WILDFIRES CASE STUDY (2007–2017)

Abstract

Wildfire has been one of the most dangerous environmental stressors nowadays. It is an important disturbance where ecosystem biomass is burned and where organisms are damaged or killed by fire. Therefore, the detecting and monitoring of this stressor are of great importance. During last decades, extensive forest fires have spread in Southern Europe, and they are registered in Serbia as well. During year 2007, several significant fires were registered in Stara Planina and Svrljiške Planine Mountains. The aims of this study were to detect land cover changes for the studied site from 2007–2017, to focus on monitoring the area affected by the wildfire, and to analyse the environment response to stressor. The study area is situated in East Serbia, partially covering the Mountains Stara Planina (western part) and Svrljiške planine (eastern part). The remote sensing techniques were used in the analysis and main satellite data were obtained via USGS Earth Explorer application. Six different classes were selected: Water, Forest, Pastures, Artificial area, Agriculture, and Bare soil. Results showed significant changes in two classes, Forest, and Pastures — the forest spread for more than 20% at the expense of pasture, which decreased more than 23%.

Keywords

• machine learning
• random forest
• change detection
• normalized burn ratio (NBR) index