Fire severity shows limited dependence on fuel structure under adverse fire weather conditions: a case study of two extreme wildfire events

Abstract Background Extreme wildfire events (EWEs) are becoming increasingly frequent in many biomes worldwide. Although the drivers of fire severity are generally well understood, current knowledge in the context of EWEs remains limited. In this study, we focused on two pyro-convective wildfires, classified as extreme based on current definitions, that occurred in northwest Spain during 2022. The differenced Normalized Burn Ratio-Enhanced Vegetation Index (dNBR-EVI) computed from Sentinel-2 multispectral imagery was used as a proxy for fire severity. Among other datasets, we leveraged ecologically relevant LiDAR-derived metrics capturing fuel density and leaf area density distribution across different strata, along with high-resolution meteorological data (vapor pressure deficit; VPD) as a proxy for dead fine fuel moisture content, to investigate the relative importance of fuel type and structure, fire weather, and topography in determining fire severity in EWEs through a Random Forest (RF) modeling approach. Results The RF model explained the spatial variability of fire severity with high accuracy (R 2 = 0.81). Among the predictors, LiDAR-based metrics indicative of surface, ladder, and canopy fuel density, as well as VPD, were identified as the most influential drivers of fire severity in the two EWEs analyzed. Coniferous forests exhibited the highest fire severity, associated with structurally homogeneous fuel loads across multiple strata that may promote surface and crown fire spread. In contrast, broadleaf-dominated forests and mosaic landscapes with agricultural areas displayed lower fire severity, highlighting their potential to moderate fire behavior. Under extremely dry conditions (high VPD), fire severity showed little sensitivity to variations in fuel parameters, indicating that fire weather may override bottom-up controls under this circumstance. Conversely, under low to moderate VPD values, high fire severity was observed when very hazardous fuel accumulation and arrangement were present. Topographic factors, although slightly contributing to severity, proved relatively less important in the two EWEs. Conclusions Our results emphasize the need for targeted fuel management in coniferous forests and shrublands to mitigate the ecological and socioeconomic impacts of intensifying wildfire regimes. However, strategies aimed at removing fuel load may not be effective under extreme fire weather conditions in the context of EWEs.