FUZZY LOGIC BASED BURNED SEVERITY CLASSIFICATION AND MAPPING WITH LANDSAT-8 DATA

Abstract. Wildfire has a strong effect on both land use and land cover so that every year, thousands of hectares of forests, farms, and urban infrastructure are destroyed. Mapping and estimation of damages are crucial for planning and decision making. The aim of this study is classification and mapping burn severity using multi-temporal Landsat data and well-known burn severity indices including Normalized Burn Ratio (NBR) and Burned Area Index (BAI) calculated for pre- and post- Landsat 8 images. Subtracted images such as dNBR (Difference Normalized Burn Ratio), RBR (Relativized Burn Ratio) and dBAI (Difference Burned Area Index) were produced on the bases of indices as classification input. Among classification methods, the fuzzy supervised classification was utilized with three classes. The result shows the strong performance of the Fuzzy Logic system (FLS) in the detection of the area with the limited number of training data so that the average accuracy of the classes is 85%; plus, from the human logic perspective, the result was meaningful so that recognition of the features and changes visually were understandable.