REBAI: Development and validation of a novel indicator for burned area detection using Sentinel-2 images

With global climate change intensifying wildfire frequency, precise monitoring of burned area dynamics is critical for assessing forest carbon cycles and informing ecological restoration efforts. However, conventional methods often struggle to accurately delineate small-scale burned areas or those with subtle spectral differences from unburned vegetation. To address this gap, we introduce the novel Red-Edge Burned Area Index (REBAI) using Sentinel-2 data, leveraging differential spectral responses in red-edge, red, and short-wave infrared (SWIR) bands between burned and unburned forest. We validated REBAI against 10 spectral bands and 13 common indices using two separability metrics and feature importance analysis. Its detection accuracy was then assessed by applying 16 adaptive thresholding algorithms. Comprehensive validation across 12 diverse global forest study areas confirmed REBAI’s superior performance, particularly for small-scale and subtle spectral burns, achieving a Jeffries-Matusita (JM) distance > 1.9 and a mean Common Isolation Values (CIV) proportion of 0.001, ranking highest in feature importance. Detection accuracy is high and stable, with metrics tightly clustered around their mean overall accuracy (96.60 % ± 3.58 %), precision (92.97 % ± 11.27 %), recall (87.12 % ± 8.07 %), and F1 score (0.8917 ± 0.0702), significantly outperforming common indices (e.g., NDVI, NBR). Furthermore, preliminary evidence suggests that REBAI may possess broad applicability across various vegetation types, with an exploratory analysis on grassland burns yielding an average F1 score of 0.91. Its accuracy could potentially be enhanced through future adaptations such as bi-temporal analysis or machine learning integration. In summary, we propose REBAI, a robust, physically-based burned area detection index, offering strong support for post-fire damage assessment and ecosystem monitoring.