Evaluating performances of spectral indices for burned area mapping using object-based image analysis

Determining post-fire information is crucial for post-fire management activities and rehabilitation treatments. The use of robust and advanced approaches is needed to determine fire severity and thoroughly analyze post-fire rehabilitation period. Object-based image analysis (OBIA) is a powerful approach that has been successfully applied in many research problems in remote sensing arena. However, its use in forest fire and related studies including fire severity and burned area estimation is quite limited. This study was carried out in Antalya's Tasaǧil district (Turkey) where according to Directorate of Forestry reports one of the largest wildfires in the Turkey occurred in 2008. The objectives of the present work are (i) to investigate the performance of object based analysis for burned area mapping; (ii) to compare the performances of widely-used burned area related spectral indices in identifying burned, slightly burned, water and non-burned areas from each other, and (iii) to delineate the boundaries of burned area. In this context, spectral indices of Normalized Burn Ratio (NBR), Normalized Vegetation Index (NDVI), Burned Area Index (BAI) derived from the satellite image were employed in analyses. Multiresolution segmentation and fuzzy membership function classifier were applied to the combinations of the selected indices (NDVI, BAI-NBR, NDVI-NBR) to discriminate burned, slightly-burned and non-burned areas from each other. Results showed that all combinations constructed in this study produced satisfactory results in terms of classification accuracy. However, the highest accuracy (98.37%) was achieved by NDVI-NBR index combination whilst the lowest accuracy (94.59%) was achieved when only the NDVI index was employed in OBIA process. It is hoped that with this work a contribution will be made for the government agencies to delineate fire perimeter and determine risk of wildfire for post-fire damage management.