Published works

Operational models that are used to predict fire behaviour can be implemented easily and rapidly. However, the operational models are only truly valid in the range of experimental conditions used to build the model. This leads to a number of difficulties when using the existing operational models to predict real-world wildfires. Physics-based modelling, that is simulating the fire behaviour from the basic equations of atmospheric fluid flow, combustion, and thermal degradation of fuel materials offers considerable insight into the dynamics of wildfire. However, physics-based simulations are computationally intensive and, at present, can only be applied to small, idealised cases. Nevertheless, the aim of this project is to use physics based models to gain insight into wildfire behaviour and use that insight to improve the current operational models for fire behaviour prediction. With this in mind, a number of investigations are underway.

The effect of a tree canopy on the near surface wind speed is investigated using Large Eddy Simulation with a view to modelling the wind reduction factor due to the canopy. The wind reduction factor is used in operational fire prediction models such as the McArthur model, to account for the reduction in wind velocity due to a tree canopy. A set of full three-dimensional simulations over idealised rectangular canopies, where the length and leaf area density of the canopy are varied, were conducted. The flow over the canopy is characterised and the potential effects on fire spread of complicated flow structures that develop at the leading and trailing edges are assessed. The simulated wind speed in the fully-developed canopy flow and the wind speed far from canopy region is used to assess the constant wind-reduction factor modelling approach.

The physics of firebrand, or ember, transport is not well understood. The distance and dispersion of the firebrands depends greatly on the turbulent fluid flow which transports the firebrands. The physics-based modelling of firebrand transport is at a preliminary stage. We seek to validate a Lagrangian particle approach for firebrand transport modelling by comparing the results from an experimental firebrand generator with simulations of same scenario. Three particle shapes, cubical, cylindrical, and disc shaped particles, representing idealised firebrands have been studied. Qualitative features of the landing distribution of the firebrands have been identified. Simulations are in progress to compare with the experimental results. So far the results are encouraging. In a related study, the thermos-kinetic properties of firebrand materials, such as bark, twigs, and leaves have been measured. This data will then be used for simulations of burning firebrands. 

Fires in grasslands are prevalent in Australia, and are relatively simple to model computationally due to the uniform fuel and flat simple terrain. In the present study, the CSIRO grassland experiments are used as validation cases for the physics-based simulations.  A parametric study has been conducted where the background windspeed and the grass height have been varied independently.  The rate-of-spread was found to be linear with windspeed in the parameter range considered. Two simulations were conducted at different heights and correspondingly different bulk density, representing grass which had been cut and left on the ground.  The fire in the taller grassland was found to have a higher rate-of-spread. Three simulations were conducted where the bulk density was kept constant as height varied. In these simulations no systematic dependence on grass height was observed.

Direct Numerical Simulation (DNS) is a numerical technique to faithfully study fluid flows by resolving all the turbulent motions instead of resorting to modelling small scale turbulence. DNS provides great insight into the physics of flows but a limited to highly idealised and numerically tractable geometries such as channels. Nonetheless, such simulations can be used to gain insight into flows which have relevance to wildfire modelling. Three flows are studied: pressure driven flow over sinusoidal roughness, mixed convective flow (flow driven by both a temperature difference and a pressure gradient), and a buoyant line plume in a confined region.

In the future, the DNS work will contribute to improved turbulence and near-wall modelling used in the physics-based wildfire models. The results of simulations from physics-based wildfire models will in turn improve operational models. We eventually aim to produce models for the wind-reduction factor, improve knowledge of firespread in grasslands, and provide a model of firebrand transport, all of which can be used operationally.