Development of Fire-Spread Model for Fire-Following Earthquake Risk Assessment in Areas with Insufficient Fire-related Data

Fire-following earthquakes (FFEs) are among the most damaging secondary hazards that can potentially occur after earthquakes. Existing risk models underuse local buildings and spatial factors, and rarely account for fire data scarcity and imbalance, which limits their accuracy and transferability. Herein, we present a regression-based fire-spread model that integrates geographic information system (GIS)-derived building density, a mix of fire-resistive construction types, and a spatial layout into a static FFE risk assessment framework for data-limited cities. Using 6,604 scenario-based dynamic simulations, we calibrated the model with the relationship between covering volume fraction (CVF) and burn-loss ratio. This approach achieved a predictive accuracy of ≥ 80% against simulation outputs and > 70% improvement over empirical baselines. Applied to 29 districts in Pohang, the model was validated through a static analysis-based assessment, confirming its predictive accuracy and utility.