The objectives of this project were to evaluate the effects of spatial scaling in computational fluid dynamics (CFD) based fire behavior models as relevant to changes in coarsening fuels inputs coupled with changes in wind speed and fuel moisture. Decision making for prescribed fire planning is a challenge with the current portfolio of fire behavior modeling as characterization of fire behavior is generalized, coarse in scale, and has high uncertainty in estimates of smoke production that is critical to burn planning on Department of Defense (DoD) lands. The approach described in this limited scope project leverages a new quick solving\cellular automata model, QUIC-FIRE, to begin testing assumptions of scale that could have impacts on how DoD managers utilize this emerging tool. This was a “high-risk” project because there has been limited development of quantifiable high resolution fuels data until recently, the model at time of the proposal was still in the development stage, and the inclusion of highly realistic fuelbed simulations could improve characterization of intrinsic fuel properties (e.g. surface area to volume ratio and bulk density) that play critical roles in how CFD models propagate fire across landscapes.
Improvements in the characterization of surface fuelbeds through the development of new methods of analyzing terrestrial and airborne laser scanning systems have advanced the discipline of fuels science. The project team employed two modes of data from these systems, describing a new novel approach of fuels characterization using terrestrial laser scanning (TLS) estimates of porosity and surface area. The project team also employed methods to use TLS–based fuels estimates to predict landscape scale fuels from airborne lidar scanning (ALS). These data were used to test scaling effects of fuels, wind, and fuel moisture within the QUIC-FIRE model at two test sites, Pebble Hill Plantation in south Georgia and Eglin Air Force Base in Florida, USA. Multiple and ensemble simulations were compared between scales and environmental factors to evaluate optimal scales of fuels inputs. Using diversity indices from information theory, this study sought to identify thresholds where no additional information is lost as scale coarsens. Information loss is evaluated as a function of both inputs and fire behavior outputs.
Generation of fuels inputs from TLS and ALS data sources demonstrate that robust characterization of fine-scale fuel variables as bulk density can be integrated as inputs to CFD models. The aggregation of these fuels input and subsequent modeling simulations describe information loss that indicates specific ranges of fuel scale that may be determined as the maximum threshold of scale for fuel characteristics. Scales that exceeded 20 meter voxel scale indicated that most or all of the information had been lost as compared to finer scales suggesting that there is limited advantage to using coarse scale data if understanding the impact of higher resolution fire model outputs as rate and shape of spread, energy release, and consumption are important metrics. The project team also investigated the ability to improve and automate the characterization of three-dimensional fuelbeds using high fidelity plant and fuel particle models with an improved method to distribute pine litter across landscape that also includes model estimates of critical intrinsic fuel properties.
This limited scope project identified several technical considerations and improvements for deriving, scaling, and assessing CFD modeled fire dynamics and effects. The short duration of the project necessitated using existing “realistic” fuelbeds from TLS and ALS data, however there was uncertainty regarding fire spread as function of the model or fuels characterization. The project team also identified that more work needs to be conducted in characterizing the fire energy outputs using spatial entropy as opposed to only a diversity index to better understand the links between input scale and expected outputs. The ability to describe fuels through the simulated fuelbeds method unveiled multiple limitations for implementation based on complexity and resolution of the models. Future research in this domain should focus on the trade-offs between resolution and efficiency of distribution and characterization of these fuel elements across large spatial domains.
Improvements of CFD-based fire behavior models requires robust and scalable fuels inputs across DoD lands for better planning and monitoring of prescribed fire. Remotely sensed data is the critical link characterizing fuels across DoD landscapes and the scales described in this project are useful to determining levels of data scale that will meet the needs of managers as they start employing CFD models in their strategic portfolio. The results of this project also have impact in regard to sources of data that need to be available to managers and researchers alike. Generalized fuels representation from national repositories are two-dimensional and treat the surface fuels in homogenous categories, thus some level of finer-scale fuels are important to scale to larger landscape scale datasets.