4TH International Congress on Technology - Engineering & SCIENCE - Kuala Lumpur - Malaysia (2017-08-05)

Real-time Simulation Of Wildland Fire Spre
A raster-based model for simulating fire spread through vegetative fuels is presented and evaluated using observational data from different fires. In this model, called SWIFFT, the combustion process is driven by unsteady energy conservation within the fuel stratum and detailed heat transfer mechanisms: radiation from the flaming zone and embers, surface and internal convection, and radiation loss. The local effects of wind, topography, and vegetation are included. To address the challenge of real-time fire spread simulations, the model is also extended in two ways. First, the Monte Carlo method is used in conjunction with a genetic algorithm to create a database of radiation view factors from the flame to the fuel surface for a wide variety of flame properties and environment conditions. Second, the front-tracking method, which consists in tracking the fire front by a moving separate grid of lower dimension than the fixed Digital Elevation Model (DEM) grid, is used. As fire spread, fuel items are generated “on the fly” in the very few grid pixels of the DEM attached to the fire front, namely all the active DEM pixels where fuel elements are heated or burning and their nearest neighboring pixels. The SWIFFT model is applied to three different fire scenarios: an arson Mediterranean fire that occurred in Favone in Corsica in 2009 [1]; a grassland fire experiment in Australia [2], [3]; and a prescribed burning in Thailand (present study). It is found that for such fire scenarios, including convection-driven or radiation-driven fires, the SWIFFT model is capable of capturing the trends observed in experiments and provides a good approximation of the observed rate of spread, and the area and shape of the burn, with reduced computational resources (Figs. 1 and 2). Differences between predictions and data are observed that can be attributable to possible errors in the estimation of model parameters and spatial heterogeneity in the fuel load distribution. It is also found that the location of the head fire is fairly well predicted, whereas SWIFFT over predicts the spread rate of fire on the flanks. Areas for future model development and evaluation include: 1) to perform a sensitivity analysis in order to evaluate the impact of parameter variations on fire propagation and thus to identify the most influential parameters; 2) to increase the raster resolution of flank fires; and 3) to extend the validation effort to a larger set of experiments and to a wider range of fuel parameters in order to better assess SWIFFT capabilities. Another way to improve the wildfire spread predictions would be to combine observational data with SWIFFT predictions through data assimilation [4].
Keywords: network model; real-time simulation; validation; wildfires
Porterie Bernard, Pizzo Yannick, Billaud Yann