Fire Behavior Prediction Systems

Computer models can be used to predict fire behavior based on differences in fuels, weather, and topography. These models are used to support fire management decisions, as a training tool to improve fire management skills, and can help display and explain fire behavior and fire management strategies to the general public. While these prediction models have inherent limitations and will never replace the knowledge and experience of wildland managers, they can provide tools to estimate fire behavior and possibly save time, money, and lives. Many of these fire behavior prediction systems require descriptions of fuel properties as inputs for calculations- this information is usually provided by fuel models. Below are examples of both current and developing fire behavior computer models.

• BEHAVE Plus Fire Modeling System. This freeware is the most commonly used software tool for predicting fire behavior in wildland fuels. Originally written in FORTRAN in 1984, the most recent version has user-friendly graphical interfaces that estimate wildland fire behavior under various fuel, weather, and topographic situations. It is used for real-time predictions of behavior of wildfires or prescribed fires. The related fireLib, which uses BEHAVE algorithms is for programmers who need a highly optimized API for developing fire behavior growth simulators.
• FARSITE (Fire Area Simulator) is used to predict fire spread in two dimensions on a landscape (as opposed to BEHAVE, a one-dimensional model). FARSITE requires GIS data for model creation. It is designed for use by those familiar with fuels, weather, topography, wildfire situations and requires training for its use.
• EMBYR is a probabilistic model that predicts potential burn patterns of large fires relative to variations in fuel types and weather patterns in an area. Ignitions can occur at random points or specific locations and ignitions from firebrands can be simulated relative to fuel type. EMBYR requires a GIS layer of fuel types based upon age classes and species composition. Fire spread probabilities are specified for three possible fuel moisture conditions; wet, intermediate, or dry. Probabilities are then adjusted using one of three wind speed categories and one of eight wind directions. The output from EMBYR indicates the final burn pattern of one or more potential landscape-scale fires allowing impacts from future fires to be estimated. The principal drawback to EMBYR is the tremendous diversity in forest species composition and age classes at varying fuel moisture conditions on a large scale thereby making probabilistic map predictions difficult.
• DYNAFIRE requires GIS data layers of fuel types, elevation, slope, aspect, and stream channels. Fuel types are standard fuel models as used in Behave Plus. Elevations can be input from DEMs and slope and aspect are computed from the elevation data. Stream channels can be digitized from topographic maps and this information is related to the elevation layer to indicate fire growth directions. Additional DYNAFIRE inputs include temperature, relative humidity, wind direction and speed, and 10-hour, 100-hour, and live fuel moistures. DYNAFIRE output maps show hourly time contours for fire perimeter, fire line intensity, average fire spread rates, and fire spread direction.
• The LANDFIRE project is one of the most massive undertakings in computer based mapping and modeling. It’s a cooperative effort between the USGS and the US Forest Service. Included in this project are 18 years of US daily weather data, 13 layers of GIS data at 30m resolution, and a host of modeling applications and tools (Keane et al. 2003). A prototype has been funded through April 2005 with final applications due at that time.

• NEXUS is an Excel spreadsheet that links surface and crown fire prediction models. It is used to estimate surface, transition, and crown fire behavior, generate site specific indices of torching and crown fire potential, build and test custom surface fuel models, and evaluate alternative treatments for reducing risk of crown fire.

• FireFamily Plus is a fire climatology and occurrence program that combines and replaces the PCFIRDAT, PCSEASON, FIRES, and CLIMATOLOGY programs into a single package with a graphical user interface for the PC.
• FOFEM (a First Order Fire Effects Model) is a computer program for predicting tree mortality, fuel consumption, smoke production, and soil heating caused by prescribed fire or wildfire. Individual tree species and relative abundance are part of the model input.
• Geo-spatially referenced methods have been explored whose end purpose is the provision of real-time, or very near real-time, mapping of wildfires and associated flame fronts. Satellite imagery has been considered and attempted as a tool for mapping and predicting wildfire spread. While the imagery collected is useful for mapping applications that aren’t time sensitive (Quayle et al. 2003), the lag time involved between overhead passes has proven too great to be effectively integrated in any wildfire attack plan. A more promising application is the Phoenix airborne infra-red fire detection and mapping system. Among its capabilities are infra-red, roll-corrected, geo-spatially referenced photographs that can be sent to the field and downloaded on laptops as jpegs in near real-time. It has the capability of automatically detecting 15-20cm hotspots from 3000 meters altitude traveling at 200 knots. It can cover 40.5 square kilometers per minute and its geo-information can be imported into GIS formats and applications (Smith and Chamberlain 2003).