Assessing fire severity in Turkey’s forest ecosystems using spectral indices from satellite images

Abstract

Fire severity classifications determine fire damage and regeneration potential in post-fire areas for effective implementation of restoration applications. Since fire damage varies according to vegetation and fire characteristics, regional assessment of fire severity is crucial. The objectives of this study were: (1) to test the performance of different satellite imagery and spectral indices, and two field—measured severity indices, CBI (Composite Burn Index) and GeoCBI (Geometrically structured Composite Burn Index) to assess fire severity; (2) to calculate classification thresholds for spectral indices that performed best in the study areas; and (3) to generate fire severity maps that could be used to determine the ecological impact of forest fires. Five large fires in Pinus brutia (Turkish pine) and Pinus nigra subsp. pallasiana var. pallasiana (Anatolian black pine)—dominated forests during 2020 and 2021 were selected as study sites. The results show that GeoCBI provided more reliable estimates of field—measured fire severity than CBI. While Sentinel-2 and Landsat-8/OLI images performed similarly well, MODIS performed poorly. Fire severity classification thresholds were determined for Sentinel-2 based RdNBR, dNBR, dSAVI, dNDVI, and dNDMI and Landsat-8/OLI based dNBR, dNDVI, and dSAVI. Among several spectral indices, the highest accuracy for fire severity classification was found for Sentinel-2 based RdNBR (72.1%) and Landsat-8/OLI based dNBR (69.2%). The results can be used to assess and map fire severity in forest ecosystems similar to those in this study.

Keywords

Remote sensing / Forest fire / Fire severity / Spectral indices / Composite burn index