Physics-based coupled fire-atmosphere models are based on approximations to the governing equations of fluid dynamics, combustion, and the thermal degradation of solid fuel. They require significantly more computational resources than the most commonly used fire spread models, which are semi-empirical or empirical. However, there are a number of fire behaviour problems, of increasing relevance, that are outside the scope of empirical and semi-empirical models. Examples are wildland-urban interface fires, assessing how well fuel treatments work to reduce the intensity of wildland fires, and investigating the mechanisms and conditions underlying blow-up fires and fire spread through heterogeneous fuels. These problems are not amenable to repeatable full-scale field studies. Suitably validated coupled atmosphere-fire models are one way to address these problems. This paper describes the development of a three-dimensional, fully transient, physics-based computer simulation approach for modelling fire spread through surface fuels. Grassland fires were simulated and compared to findings from Australian experiments. Predictions of the head fire spread rate for a range of ambient wind speeds and ignition line-fire lengths compared favourably to experiments. In addition, two specific experimental cases were simulated in order to evaluate how well the model predicts the development of the entire fire perimeter.

Wildfires that spread into wildland–urban interface (WUI) communities present significant challenges on several fronts. In the United States, the WUI accounts for a significant portion of wildland fire suppression and wildland fuel treatment costs. Methods to reduce structure losses are focussed on fuel treatments in either wildland fuels or residential fuels. There is a need for a well-characterised, systematic testing of these approaches across a range of community and structure types and fire conditions. Laboratory experiments, field measurements and fire behaviour models can be used to better determine the exposure conditions faced by communities and structures. The outcome of such an effort would be proven fuel treatment techniques for wildland and residential fuels, risk assessment strategies, economic cost analysis models, and test methods with representative exposure conditions for fire-resistant building designs and materials.

No abstract

A series of real-scale fire experiments were performed to determine the size and mass distribution of firebrands generated from Douglas-fir (Pseudotsuga menziesii) trees. The experiments were performed in the Large Fire Laboratory at the National Institute of Standards and Technology. The Douglas-fir trees used for the experiments ranged in total height from 2.6 to 5.2 m and the tree moisture content was varied. An array of pans filled with water was used to collect the firebrands that were generated from the burning trees. This ensured that firebrands would be quenched as soon as they made contact with the pans. The firebrands were subsequently dried and the sizes were measured using callipers and the dry mass was determined using a precision balance. For all experiments performed, the firebrands were cylindrical in shape. The average firebrand size measured from the 2.6-m Douglas-fir trees was 3 mm in diameter, 40 mm in length. The average firebrand size measured for the 5.2-m Douglas-fir trees was 4 mm in diameter with a length of 53 mm. The mass distribution of firebrands produced from two different tree sizes under similar tree moisture levels was similar. The only noticeable difference occurred in the largest mass class. Firebrands with masses up to 3.5 g to 3.7 g were observed for the larger tree height used (5.2 m). The surface area of the firebrands scaled with firebrand weight.

No abstract

An experimental approach has been developed to quantify the characteristics and flux of firebrands during a management-scale wildfire in a pine-dominated ecosystem. By characterizing the local fire behavior and measuring the temporal and spatial variation in firebrand collection, the flux of firebrands has been related to the fire behavior for the first time. This linkage is seen as the first step in risk mitigation at the wildland urban interface (WUI). Data analyses allowed the evaluation of firebrand flux with respect to observed fire intensities for this ecosystem. Typical firebrand fluxes of 0.824-1.361 pcs.m -2 .s -1 were observed for fire intensities ranging between 7.35±3.48 MW.m -1 to 12.59±5.87 MW.m -1 . The experimental approach is shown to provide consistent experimental data, with small variations within the firebrand collection area. Particle size distributions show that small particles of area 0.75-5×10 -5 m 2 are the most abundant (0.6-1 pcs.m -2 .s -1 ), with the total flux of particles >5 ×10 -5 m 2 equal to 0.2 to 0.3 pcs.m -2 .s -1 . The experimental method and the data gathered show substantial promise for future investigation and quantification of firebrand generation and consequently a better description of the firebrand risk at the WUI.

No abstract

Modeling the behavior of crown fires is challenging due to the complex set of coupled processes that drive the characteristics of a spreading wildfire and the large range of spatial and temporal scales over which these processes occur. Detailed physics-based modeling approaches such as FIRETEC and the Wildland Urban Interface Fire Dynamics Simulator (WFDS) simulate fire behavior using computational fluid dynamics based methods to numerically solve the three-dimensional, time dependent, model equations that govern, to some approximation, the component physical processes and their interactions that drive fire behavior. Both of these models have had limited evaluation and have not been assessed for predicting crown fire behavior. In this paper, we utilized a published set of field-scale measured crown fire rate of spread (ROS) data to provide a coarse assessment of crown fire ROS predictions from previously published studies that have utilized WFDS or FIRETEC. Overall, 86% of all simulated ROS values using WFDS or FIRETEC fell within the 95% prediction interval of the empirical data, which was above the goal of 75% for dynamic ecological modeling. However, scarcity of available empirical data is a bottleneck for further assessment of model performance.