Research Information System
Turkish Journal of Remote Sensing and GIS, vol.7, no.1, pp.145-165, 2026 (TRDizin)
In this study, burned forest areas were identified using bi-temporal Sentinel-2 imagery acquired before and after the fire, through the application of spectral indices such as the Normalized Burn Ratio (NBR) and the Normalized Difference Vegetation Index (NDVI). Specifically, the forest fire that occurred in the Manavgat district of Antalya Province, Türkiye, on July 27, 2021, was analyzed based on NBR and NDVI values derived from Sentinel-2 satellite data. The Normalized Burn Ratio (NBR) variants NBR1 and NBR2, and NDVI indexes were evaluated comparatively. To reduce confusion between agricultural areas and burnt forest regions we excluded agricultural fields from processing through using the CORINE land cover database. The achieved results showed that the NDVI-based detection provided the highest overall accuracy (OA) of 97.1% and a Kappa of 0.950, while NBR2 and NBR1 resulted in OA values of 96.8% and 96.3%, respectively. The high NDVI performance is primarily attributed to the mixed forest–cropland mosaic structure of the study area, which enhances spectral contrast between burned and unburned surfaces. Masking agricultural areas with the existing CORINE database reduced the false positives and hence improved detection reliability for all indexes. The results demonstrated that the used methodology has high effectiveness in quantitatively mapping fire damages and supporting rehabilitation planning.