Modelling the spreading of large-scale wildland fires

TL;DR

This paper presents a hybrid stochastic-deterministic model for simulating large-scale wildland fire spread, validated with real fire data and sensitivity analysis of key parameters.

Contribution

It introduces a novel hybrid model combining network and semi-physical approaches for fire spread simulation, validated with real-world fire data.

Findings

Model predictions align well with experimental data

The model accurately predicts fire spread rate, area, and shape

Sensitivity analysis identifies key parameters influencing fire spread

Abstract

The objective of the present study is twofold. First, the last developments and validation results of a hybrid model designed to simulate fire patterns in heterogeneous landscapes are presented. The model combines the features of a stochastic small-world network model with those of a deterministic semi-physical model of the interaction between burning and non-burning cells that strongly depends on local conditions of wind, topography, and vegetation. Radiation and convection from the flaming zone, and radiative heat loss to the ambient are considered in the preheating process of unburned cells. Second, the model is applied to an Australian grassland fire experiment as well as to a real fire that took place in Corsica in 2009. Predictions compare favorably to experiments in terms of rate of spread, area and shape of the burn. Finally, the sensitivity of the model outcomes (here the rate of spread) to six input parameters is studied using a two-level full factorial design.