Monitoring of post-fire forest scars in Serbia based on satellite Sentinel-2 data

Abstract

This study aims to improving long-term post-fire environment assessment. It proposes a method for monitoring fire impact using Sentinel-2 satellite data by combining spectral and textural features of land cover types inside a post-fire study sites. Specific objectives were to 1) test stability of the burnt area index for Sentinel-2 (BAIS2) for identification of burn in study sites, 2) investigate the optimal feature combination for mapping land covers inside study sites, and 3) assess and analyse dynamic in land covers of study sites. BAIS2 was shown independent on date acquisition of satellite images to distinguish forest burn from other land covers over the analysed May–September vegetation period. Texture of study site improved the classification results. The most accurate classification method for identification of study sites land covers (with 0.84 Kappa coefficient and 0.86 overall accuracy) was based on combination of Sentinel-2 bands, BAIS2, and texture by Fourier transform. Analysis of vegetation recovery within the study sites demonstrated different recovery rates. Natural regeneration of pine was not observed, during three to six years of observations following fire events. The proposed method and findings can support planning of forest management measures needed to effectively restore forest cover.

Keywords

• forest burn
• classification
• recovery
• image texture
• bais2
• spectral analysis