Coupling computer-vision house fuel load estimation with fire spread simulation in the Wildland-Urban interface

Abstract

Wildfires directly threaten human lives and properties in the wildland-urban interface (WUI). The quantity of combustible materials within houses is critical, yet variability in house fuel load remains unaccounted in current WUI fire modelling. This study integrates House Fuel Load (HFL) into the Streamlined Wildland–Urban Interface Fire Tracing (SWUIFT) simulation framework to examine its influence on WUI fire spread. HFL assessments were derived from computer-vision assessment of structural attributes and combined with standardized indoor fuel load values to calculate fully developed fire durations of ignited structures. We compared a baseline SWUIFT runs (fixed fire-duration configuration) with HFL-coupled cases under two pHRR scenarios of 100 kW/m2 and 150 kW/m2 in simulations of the 2023 Lahaina fire. Results indicate that although varied HFL modifies the theoretical heat release of individual structures, the WUI fire spread patterns and ignition statistics remain largely consistent. Sensitivity analysis further indicates that reducing burning time below 60 min substantially inhibits fire spread, whereas durations beyond 100 min have limited additional influence. Overall, the HFL-based method provides reasonable estimates of residential burning time, and the corresponding simulated fire spread paths align closely with the real event, supporting the applicability of the approach.