Evaluating dNBR and CBI for Fire Severity Estimation in Fennoscandian Boreal Surface Fires

Abstract

Fire severity—the magnitude of ecological change induced by fire—plays a critical role in shaping post-fire recovery dynamics, carbon cycling, and forest management decisions. In Fennoscandian boreal forests, however, accurate severity estimation using remote sensing remains challenging due to relatively small burn areas and canopy cover obscuring surface-level fire impacts.This thesis aims to investigate the accuracy of the differenced Normalized Burn Ratio (dNBR), a widely used satellite-derived spectral index, and the Composite Burn Index (CBI), a field-based ground truthing index, in estimating fire severity following a low- to moderate-intensity surface fire in Fennoscandia. Using a prescribed burn in Leppävirta, Finland as a case study, I compare dNBR to a study-specific, custom index, the Data-Derived Severity Index (DDSI), by me developed from pre-existing field measurements—tree mortality, O-horizon (including litter and humus layers) thickness loss, and mass loss. This enables direct evaluation of how accurately dNBR reflects fire effects on both canopy and soil strata. Additionally, I also evaluate the level of alignment between dNBR and CBI, and the performance of CBI as a ground-based reference for fire severity assessment in this region.Results reveal that dNBR underperforms in estimating fire severity under Fennoscandian conditions. Additionally, dNBR does not exhibit statistically significant correlation with O-horizon level fire effects, likely due to canopy occlusion. Results also reveal that dNBR-derived severity classifications do not align with CBI-derived ones, with CBI demonstrating stronger overall agreement with field-observed fire effects. These findings suggest that dNBR–originally calibrated for the high-intensity, stand-replacing North American fire regime–does not accurately reflect fire severity in surface fire ecosystems where combustion primarily affects the organic soil layers. They also highlight the ecosystem-specific nature of the dNBR-CBI relationship and indicate that, while CBI more reliably captures overall fire impact patterns, it presents a structural bias toward canopy-level indicators. This indicates its possible unsuitability as a ground-truthing index for fire severity in Fennoscandian surface-fire regimes.Together, these results highlight the limitations of vegetation-focused severity indices and the need for region-specific, stratified severity assessment tools that integrate both spectral and field-based data.