This research project looked into the ability of FIRETEC, a physics-based wildland fire model coupled to a computational fluid dynamics model, to simulate fire behavior for various prescribed fire ignition patterns in longleaf pine sandhill fuel beds. Unlike FIRETEC and other physics-based models, currently used fire behavior models were designed to model spread from a single ignition point and are inadequate for predicting the complex influences of atmosphere, forest structure, and self-generating fire processes on wildland fire behavior. The project is a direct result of intense collaboration between fire scientists, fire modelers and fire managers with a goal to co-produce research that is directly and immediately relevant to fire managers by using FIRETEC to explore managers’ specific “burning questions”. 

The performance objectives of the project were to (a) further validate FIRETEC’s ability to simulate representative coupled fire/atmosphere behavior, (b) simulate prescribed fire behavior and phenomena in longleaf pine fuels on Eglin Air Force Base (AFB) and, (c) use modeling results to enhance knowledge, skills, and abilities of fire practitioners nationally. The project was designed and managed using a co-production approach to demonstrate the potential to leverage the modeling power of a next generation fire spread model, FIRETEC, to improve wildland fire managers’ knowledge base, situational awareness and prescribed fire outcomes, particularly as related to fire behavior dynamics associated with various fuel types, atmospheric conditions, and complex firing patterns by providing powerful visual training tools.

Technology Description

FIRETEC is a physics-based, three-dimensional computer code designed to capture what is a constantly changing, interactive relationship between wildland fire and its environment. During a FIRETEC wildland fire simulation, heat is produced, fuel (vegetation) and oxygen are reduced, and the surrounding air becomes buoyant, causing hot air to rise quickly above the fire and draw cooler air in near the base of the fire. Buoyant rise and related indrafts are mechanisms through which local combustion affects other areas of fires by changing the larger-scale flow patterns. Heat is exchanged between vegetation and gases as air moves through plants by convective heat exchange, and thermal radiation is emitted and absorbed by hot gases and vegetation. Moreover, vegetation imposes aerodynamic drag on airflow in relation to the bulk properties of the fuel bed. 


To accurately represent such interactive fire processes, FIRETEC combines physics models that represent combustion, heat transfer, aerodynamic drag and turbulence with a computational fluid-dynamics model, HIGRAD, which represents airflow and its adjustments to terrain, vegetative obstructions and the fire itself. FIRETEC simulates the dynamic processes that occur within a fire and the way those processes influence each other by solving a set of coupled partial differential equations for the conservation of mass, momentum, energy, chemical species, and turbulence. These equations describe the evolution in time and the variations in space of many physical properties that influence, or are influenced by a fire, e.g., gas and vegetation temperatures, wind speed and direction, kinetic energy in the form of turbulence, oxygen concentrations, masses and characteristics of remaining vegetation, and other variables. These physical properties are computed as functions of time at millions of locations distributed in a three-dimensional terrain-following mesh that follows the simulated landscape.

Demonstration Results

FIRETEC modeling results from this project are visually impressive, have proven very useful for enhancing knowledge of fire managers and provide a significant step forward in FIRETEC validation efforts as related to complex prescribed fire simulations. In validating FIRETEC’s ability to capture realistic coupled fire/atmosphere behavior, model performance success was based on the ability of FIRETEC to simulate fire-induced winds as well as radiative fluxes within one standard deviation of field values measured during Prescribed Fire Combustion Atmospheric Dynamics Research Experiment (RxCADRE) campaigns at Eglin AFB. In the end, data from the 2012 RxCADRE S5 burn, which was approximately 2 hectares in size, were used to assess this performance objective. Several direct and secondary conclusions were reached during this demonstration. First, FIRETEC spread patterns were appropriate in direction and magnitude. In addition, the demonstration found that FIRETEC spread patterns for RxCADRE S5 plot burns were heavily influenced by the details of the prescribed boundary conditions. Unfortunately, validation of the coupled fire/atmosphere model using standard statistical validation methodology for the RxCADRE 2012 S5 burn might be impossible because there are too many degrees of freedom in the specification of the environment. Some of the secondary lessons learned during this demonstration include that sparse fuels, high humidity and light winds produced marginal burning conditions in which small changes in environment can have large changes in fire behavior, leading to the need for more detailed data. Furthermore, predictability of fire behavior depends on the relative magnitude of the ambient spread drivers compared to the fluctuations in these drivers. This demonstration supported the need for an increase in the use of models in the design of future fire experiments in order to improve understanding of the data adequacy.

Implementation Issues

Standard statistical evaluation of fire behavior model performance based on replication is not plausible because no two fires are identical, i.e. they cannot be replicated. Even when ignition patterns, weather forecast, and plot layouts are similar for operational burns, differences in timing, strength, and directions of wind gusts, fuel arrangement, time of day, time of year, drought index and numerous other factors will produce different fire behavior and fire effects. These same confounding factors preclude the use of standard statistical validation based on multiple replicates for fire models when comparing modeled outputs to actual fires. It was also discovered during attempts to simulate and compare RxCADRE small plot experimental burns that data gaps due to RxCADRE sampling design would make true model validation infeasible. Accordingly, other than the direct comparison of FIRETEC outputs to the RxCADRE S5 experimental burn, the focus of the FIRETEC simulations in this project was to explore the trends in phenomenology associated with various prescribed fire practices, not to predict exact metrics for the various scenarios.

  • Demonstration,

  • Fire,