Radar Burn Ratio for fire severity estimation at canopy level: An example for temperate forests

Fires affect wide areas and their effects can be successfully estimated from a range of remote sensing sensors, with synthetic aperture radars (SAR) being of particular interest due to their sensitivity to forest vertical structure, global availability and independence of cloud cover or solar elevation. Previous studies have demonstrated the sensitivity to fire effects of L-band SAR sensors using post-fire datasets and empirical modeling. This study proposed an innovative method for estimating fire severity by combining pre- and post-fire SAR datasets within a change detection framework to compute a novel index, the Radar Burn Ratio (RBR). More importantly, a standardized RBR was developed and tested over seven temperate forest types located on three continents with above ground biomass values ranging from 30 to over 500 t ha− 1. RBR standardization allowed for common thresholds to be defined and subsequently used for estimating the Composite Burn Index (CBI, a measure of fire impact) without the need for a priori information (i.e., in situ data) on local post-fire conditions. The estimation accuracy of the standardized RBR was compared to locally-calibrated empirical models based on field CBI data. The results showed similar estimation errors and a strong agreement with the reference in situ data (i.e., Cohen's weighted kappa > 0.61). The RBR index most sensitive to fire severity was based on the cross-polarized channel applied under dry environmental conditions. Under wet conditions the estimation accuracy was considerably lower. The methods proposed in this study are particularly valuable for rapid fire severity assessments at regional to global scales, requiring only that RBR thresholds be calibrated for a range of environments and that CBI scores be related to fuel consumption for each forest type.