WUI-CAST: Real-Time Forecast Updating for Partially Observed Wildland-Urban Interface Fires

Abstract

Wildland-urban interface (WUI) fires couple landscape spread, structural fire development, and wind-driven ember transport. Forecasting building risk requires models that incorporate emerging observations and update predictions as fires evolve. This study presents WUI-CAST, a framework integrating established fire-propagation representations with sequential Bayesian state estimation to support real-time forecast updating under partial observation. A Dynamic Bayesian Network represents nonlocal ember-driven ignition and local structural propagation, while Bayesian filtering updates building-state probabilities using sparse reports. A retrospective Camp Fire application provides a proof-of-concept evaluation. Open-loop simulations captured key features of the wildfire-to-community transition. In two 30-minute forecasts evaluated against subsequently reported buildings before their assimilation, closed-loop updating increased Recall@10% from 29.03% to 83.87% and from 0.00% to 43.75%, with improved conditional spatial logarithmic scores in both windows. Regional and local diagnostics reveal how assimilation redistributes risk beyond directly observed buildings. Controlled experiments demonstrate gains across the tested observation conditions and filtering implementations. A forecast-update cycle averaged 42 s for a 5-min simulation interval. Together, these findings demonstrate improved spatial risk prioritization from sparse observations and a computationally feasible approach to real-time WUI forecast updating.