Simulating low-intensity experimental fires using a coupled-fire/atmosphere behavior model

The capability for physics-based wildfire models to simulate low intensity fires such as those that might be part of prescribed fire events in the southeastern United States has been explored through a series of FIRETEC simulations. This modeling work attempted to simulate the S5 burn of the 2012 Prescribed Fire Combustion and Atmospheric Dynamics Research Experiments (RxCADRE) using field data that was collected during the burn. A variety of lessons were learned through the process of trying to use the on-site atmospheric and fuel measurements to initialize simulations for this low intensity burn.

The low intensity nature of the 2012 RxCADRE S5 experimental burn is typical of many of the prescribed fires in the Southeast. While it might appear that modeling low intensity fires such as this would be straight forward, in reality matching experimental results under these marginal conditions is inherently difficult. The vegetation at the location of the S5 burn was heterogeneous and patchy, with sandy regions interspersed within the plot. The details of the fire-spread patterns were determined by the ability and rate of spread within the mosaic of patches. The surface winds were light and variable, with fluctuations in the wind speed being of the same order of magnitude as the mean wind. These fluctuations are important to fire behavior, but they are not known.

Novel methods were used to develop a reasonable representation of the heterogeneous fuel bed of the S5 plot by combining high-resolution airborne photography, feature- extraction image analysis software and destructive sampling that was performed by the RxCADRE team at the time of the burns. The fuel bed primarily consisted of patches of a variety of herbaceous vegetation, palmetto and woody goldenrod, with typical patch sizes less than 10 m. This process generated a heterogeneous fuel bed resolved on a 2 m FIRETEC grid. The light and variable nature of the fluctuating winds also presented a challenge for modeling as the poorly resolved atmospheric turbulence had a very significant influence on different portions of the evolving fire line. Upstream wind boundary conditions for the computational domain were determined by blending site wind measurements using several different methods. Depending on the blending methodology, the fluctuations in the wind field were different and the fire's response to these fluctuations was different. Simulations initialized with winds from the different methodologies illustrate how details of the fire behavior depend on the local turbulent fluctuations.

The results of this analysis suggest that physics-based fire behavior models such as FIRETEC can be used to simulate fire behavior for fires with characteristics similar to those of prescribed fires in the southeastern United States. However, under these marginal conditions with heterogeneous fuels and light and variable winds, the details of the fire behavior is very difficult to replicate using the data collected during this burn.