An Extended Cellular Automaton Framework for Wildfire Spread Simulation with Dual-Scale Spatial Representation and First-Arrival-Time Propagation

Fixed-time-step updating and cell-level states in cellular automaton (CA) models can limit representation of directional travel times and within-cell fire-front progression. This study extends a CA-based wildfire spread framework by combining first-arrival-time propagation with dual-scale spatial representation. Environmental attributes remain on 30 m macro-cells, while finer state points record arrival times; a priority queue orders updates. The framework was evaluated using the 2020 Muli wildfire in China. With 30 m state-point spacing, first-arrival-time propagation increased intersection over union (IoU) from 0.5563 to 0.6129, reduced the 95th-percentile Hausdorff distance (HD95) from 3695.0 to 2722.9 m, and decreased relative burned-area error from 21.79% to 17.80% compared with fixed-time-step updating. At 15 m spacing, the framework represented partially burned macro-cells and achieved an IoU of 0.6133 and an HD95 of 2718.6 m, but runtime increased from 41.6 to 189.8 s. Further reduction to 10 m did not improve IoU or HD95. These findings show that temporal updating primarily improved final burned-area agreement, whereas spatial refinement provided within-cell state detail at higher computational cost. This trade-off may inform state-point spacing choices in resource-constrained simulations, pending validation across additional fires.