UC Berkeley

Wildland fire behavior has been studied and modeled for decades, but structure ignition and loss in the wildland-urban interface (WUI) remain poorly quantified. Recent large-loss fires in California and elsewhere show that current risk assessment methods focus on landscape fuels and topography while treating buildings and communities too crudely, failing to resolve how construction, defensible space, and urban form control ignition and spread. There is a particular need for risk analysis methods that connect modeled fire behavior metrics to structure loss and that scale to large regions. This work combines three components to advance methods toward addressing that gap. First, it develops fragility functions that relate reconstructed wildfire hazard metrics (such as flame length, ember exposure, and compound exposure indices) to observed structure damage, including the influence of building codes, separation distances, and vegetation clearance. Second, it constructs a framework to characterize urban fuels, assigning building fuel models that encode structure fire performance and combustibility from structure- and parcel-scale attributes. Third, it introduces FireDX, a transparent data pipeline for structure-level dataset generation compatible with downstream fire modeling and risk analysis that integrates open geospatial datasets, building fuel model assignments, and fire spread simulations to generate exposure and damage probability estimates and to evaluate relative effects of mitigation. Applications include a statewide analysis of structures across recent large fires in California and a hindcast of the 2017 Tubbs Fire. The analyses show that structure fragility depends on hazard intensity, building characteristics, and neighborhood morphology. Hindcast simulations demonstrate that including building fuel models and urban form metrics improves predictive skill relative to wildland-only spread models, and mitigation scenarios quantify how changes in fuels and ignition resistance translate into relative reductions in structure loss. Together, these results provide a framework for fire modeling and spatially explicit risk quantification and highlight the importance of building-level data and open-source tools for community planning, mitigation, and insurance reform.