Normalized Radar Burn Ratio: A Case Study for Burned Area Mapping in Mediterranean Forests

Abstract

This research introduces the normalized radar burn ratio (NRBR), an index designed to enhance burned area detection using Sentinel-1 C-band radar imagery. The research utilizes postfire to prefire ratios of VV and VH backscatter coefficient to compute the NRBR, thus optimizing the contrast between the burned and unburned areas. The 2017 wildfires in Portugal were used to validate the methodology. Using the U-Net architecture, the NRBR-based model outperforms previous ratio-based indices in metrics, such as overall accuracy (OA), omission error (OE), and intersection over union, among other metrics. Additionally, high correlations ( $r \gt 0.7$ ) between NRBR and the optical indices NDVI (postfire) and dNBR were observed. This approach has promising implications for improving burned area mapping, particularly for periods with cloud cover or occlusion from fire smoke.