Assessment of a burned forest area using satellite images: A case of the çandir forest fire

This study aims to identify and evaluate forest areas affected by a wildfire that occurred in 2021 within the boundaries of the Çandır Forest Subdistrict in the Isparta region, using remote sensing and geographic information systems (GIS). In the study, spectral indices NDVI (Normalized Difference Vegetation Index) and NBR (Normalized Burn Ratio) derived from pre- and post-fire Landsat-8 and Sentinel-2 satellite images were used. To quantitatively determine the impact of the fire, dNDVI and dNBR images, representing the differences between pre- and post-fire NDVI and NBR indices, were generated. A thresholding method was applied to these difference images to identify the burned forest areas. Using a supervised classification method on Landsat-8 and Sentinel-2 data, thematic maps were produced with the Maximum Likelihood algorithm, and the burned areas were also detected based on the classified images. The highest classification accuracies were achieved using Sentinel-2 satellite data, with 92% accuracy (kappa 0.83) for dNBR and 88% accuracy (kappa 0.77) for dNDVI. For Landsat-8 data, classification accuracies for dNBR and dNDVI were approximately 85% (kappa  0.7). In the classified images, overall accuracy values of 88% (kappa 0.85) for Sentinel-2 and 85% (kappa 0.81) for Landsat-8 were obtained. Based on the results, the average burned area was calculated to be approximately 760 hectares. In conclusion, dNDVI and dNBR images derived from Sentinel-2 and Landsat-8 satellite data provided effective and reliable results in identifying and evaluating burned forest areas. It is recommended that remote sensing data and methods, along with GIS technologies, be used effectively in wildfire management.