Published works

It is crucial for emergency and disaster management organisations to predict of the rate of spread and intensity of bushfires for operational planning, community warnings and the deployment of their resources. Currently, this is achieved by simulation using simplified operational models that have the useful attribute of providing results on time scales commensurate with those required by emergency managers. However, when Cruz & Alexander [1] reviewed the performance of the operational fire models used by fire and emergency service analysts on seven vegetation types found in Australia, they found that on an average most of the fire models have an error of 20–80% in estimating the rate of fire spread. These differences in prediction are due to the assumptions and limitations of these models. Therefore, it is essential that these simplified operational tools be refined so that they can better predict fire behaviour. Additionally, a more physically based firebrand model needs to be included in operational models to predict firebrand distribution and subsequent spotting, which lead to an increased rate of fire spread (ROS). Currently, no such model exists. With an increased population in the rural–urban interface (or wildland–urban interface, WUI), it is also important to understand the vulnerability of houses from radiant heat and firebrand flux in order to minimise such vulnerability.

In this project, we tested two established reliable physics-based models: Fire Dynamics Simulator (FDS) and FIRESTAR3D to simulate bushfire scenarios in three broad areas:

(1)  sub-canopy wind flow,

(2)  firebrand transport, and

(3)  propagation of grass and forest fires.

We have made significant inroads into providing usable outputs as well understanding various aspects of bushfire behaviour. The following are particular highlights:

• Modelling of wind flow through tree canopies. We found that a simple mathematical model gives a similar wind profile to that given by the FDS in a homogeneous tree canopy on flat ground under neutral atmospheric conditions. We went on to implement this simple model (known as the Harman and Finnigan model) to generate a dynamic[1] wind reduction factor (WRF) in the fire spread model SPARK (developed by CSIRO Data61). Using six real-fire case studies, we demonstrated that a dynamic WRF generally provides better fire propagation prediction than a fixed WRF. Some flow simulations in heterogeneous tree canopies show interesting flow features that not only affect fire propagation but can also affect firebrand transport.
• Modelling of transport of firebrands (cubical, cylindrical and disc-shaped particles, representing idealised firebrands) and their landing distribution with view to developing models of ember attack for use in future iterations of Australian Standard AS3959. We extended the simulation work to study the transport of ember particles across realistic forest edges. The results show that a lower sphericity of the firebrand can increase lateral dispersion.
• Simulations of rate of grassfire propagation as functions of wind, grass height, slope, initial wind field with a view to improving Australian Standard AS3959. The foundations of rigorous simulation have been laid in a number of our journal publications [2, 3] and we are now exploring various parametric studies. In one such study, we evaluated heat load from incoming grassfire on structures in relation to AS3959.

This project was also established to create a capability and capacity in Australia to conduct research and understand physical-based wildfire modelling approaches. There are several international groups developing these models, and it is imperative that Australia can interact and work alongside these researchers to translate the findings to the Australian context.

Overall, we have achieved our goal of obtaining greater insight into bushfire physics and we are now utilising those insights to parameterise various phenomena for operational models.

This report explains these issues in more detail.