The Coupled Atmosphere-Wildland
Fire Environment Model (CAWFE)
for Wildland fire Modeling
The NCAR-USDA Forest Service Coupled Atmosphere – Wildland Fire - Environment (CAWFE) Model contains two parts: a numerical weather prediction model and a fire behavior model that simulates the growth of a wildfire in response to weather, fuel conditions, and terrain (Clark et al. 2004; Coen 2005a). These are two-way coupled to constantly exchange information so that heat and water vapor fluxes from the fire alter the atmospheric state, notably producing fire winds, as the evolving atmospheric state and changes in humidity (including effects from the fire) simultaneously affect fire behavior, notably how fast and in what direction the fire propagates.
a. Atmospheric Model
The meteorological model is a three-dimensional non-hydrostatic numerical model based on the Navier-Stokes equations of motion, a thermodynamic equation, and conservation of mass equations using the anelastic approximation. Vertically stretched terrain-following coordinates allow the user to simulate in detail the airflow over complex terrain. The numerical weather prediction model is capable of modeling fine scale atmospheric flows (horizontal resolution of 10s-100s of m) in very steep terrain where the inclination may exceed 30o. Forecasted changes in the larger-scale atmospheric environment are used to initialize the outer of several nested domains and update lateral boundary conditions. Two-way interactive nested grids capture the outer forcing domain scale of the synoptic-scale environment while allowing the user to telescope down to tens of meters near the fireline through horizontal and vertical grid refinement. Weather processes such as the production of cloud droplets, rain, and ice are parameterized using standard cloud microphysical treatments.
b. Fire Model
Because the modeling system is trying to represent fires at spatial and temporal scales too coarse to simulate combustion, the wildland fire component of the model is based on semi-empirical relationships.
· One algorithm defines the burning region within each grid cell. At the surface, each atmospheric grid cell is further divided into two-dimensional fuel cells with fuel physical characteristics and fuel loads specified by the user, the defaults being those of the 13 standard fuel models (Anderson, 1982). Four points within each cell (called tracers) make up a quadrilateral that contains the burning region of the cell. Together with neighboring cells, these define the fire front, which is the interface between burning and unburned fuel. A local contour advection scheme assures consistency along the fireline.
· Another algorithm relates fire properties such as rate of spread to local wind, terrain slope, and fuel characteristics through the Rothermel (1972) surface-fire algorithms. Fire spread rates are calculated locally along the fire front as a function of fuels, wind speed and direction from the atmospheric model (which includes the effects of the fire), and terrain slope.
· A third algorithm implements a canopy fire model that calculates the energy used to heat and dry any canopy above a surface fire. The canopy is ignited if the residual heat flux, after heating and drying the canopy, exceeds a specified threshold value. Any canopy fire is assumed to remain collocated with the surface fire.
· A fourth algorithm treats the post-frontal heat release rate (Albini, 1994) which characterizes how the fire consumes fuels of different sizes with time after ignition, distinguishing between rapidly consumed grasses and slowly burned logs.
· An emission factor calculates how much of the consumed fuel mass is released into the atmosphere as particulate matter. The simulated smoke plume and 3-D winds of the weather model transport the smoke.
· A simple radiation treatment distributes the sensible and latent heat and smoke into the lowest atmospheric grid levels. The e-folding height over which the heat is distributed is specified by the user – typically 10 m for grass fires and 50 m for crown fires, based on analysis of fire observations (Clements et al., 2007; Coen et al., 2004).
The fire behavior is coupled to the atmospheric model: low level winds drive the spread of the surface fire, which releases sensible heat, latent heat, and smoke into the lower atmosphere, in turn feeding back to affect the winds directing the fire. Although this influence is most dramatic near the fire, model simulations show this influence can change the wind speed by several kilometers per hour even kilometers from the fire (Coen, 2005).
PURPOSE:
(1) Studies using CAWFE have shown that complex interactions between a fire and the atmosphere are behind even the most basic fire fundamentals (Coen, 2011), including the elliptical shape of the fire itself and the formation of the rapidly moving, most intense burning head, flanking regions where the wind blows parallel to the interface, and backing region that creeps against the wind. In contrast to FARSITE, which forces the evolving shape of the fire to follow an ellipse, the interplay of the fire with the atmosphere in CAWFE allows this often-observed behavior, to evolve from the physics (Figure 1).
Thus, although simpler tools such as BEHAVE+ and FARSITE can in some circumstances, with daily calibration of fuel loads, produce acceptable estimates of anticipated rates of spread, much more detailed understanding/insights are now possible with coupled weather-fire models. Because the predicted rate of spread for a given input wind is fixed for tools such as BEHAVE+ or FARSITE, it is not possible to capture these feedbacks with incremental improvements; models such as CAWFE are fundamentally different because the forces on the air created by the fire change the winds that in turn direct the direction and rate of spread of the fire. This becomes even more important when simulating the unfolding of a large fire event, where more complex results of this coupling become apparent.
Figure 1. Heat produced by the fire (more intense colors are hotter), smoke (misty purple field), and surface winds (longer arrows indicate stronger winds, the arrow indicates direction). In this simulation, a fire began as a line in winds that were all coming at 3 m/s from the left, but which created a fire with a head, flanks, and backing region, and shaped the winds in the fire vicinity to be moving rapidly forward at the fire head, parallel to the flanks, and weak in the backing region. (Animation at http://www.mmm.ucar.edu/people/coen/files/newpage_f.html ). The winds along the flank carry small perturbations, which grow into fire whirls, along to the head of the fire where they may hook together and shoot forward in bursts.
(2) Numerical weather prediction models represent the three-dimensional wind structure in mountainous terrain. These flows depend on many simultaneous factors, including the steepness, height, and width of terrain, the wind speed and direction and how this changes with height, the temperature profile in the air (the atmospheric stability), heat released by the formation or evaporation of precipitation and clouds, the solar heating of slopes, upwind terrain features, and the roughness of the surface. The past 30 yrs of atmospheric science research has established that no simpler model than a full weather model can consider all these factors together and predict which of many types of flow might occur at one particular time and location. In mountainous terrain, weather information from a weather station is extremely unlikely to be representative of the winds driving the fire.
A striking example of the limitations of current approaches is the 2007 Esperanza Fire in Riverside County, California, which ignited on the upwind edge of the San Jacinto mountains during dry, windy Santa Ana conditions. CAWFE simulations show the simulated weather conditions, fire growth, and smoke production and transport. This work is the first to simulate simultaneously the evolving meteorological flow, fire behavior, and fire-induced flow for a landscape-scale naturally evolving fire. The simulation captures how strong upper-altitude winds from the east-northeast were brought down to the surface in terrain-generated atmospheric waves, driving the fire and smoke to the west-southwest. (Figure 2). Local surface weather stations (RAWS) were located in Banning Pass and caused FARSITE to predict a slow spread due west.
Other features characteristic of this fire are brought out in the CAWFE simulation - the splitting of the fire, the fire drawing itself up canyons along the flanks, the most intense burning often being along the flanks rather than at the leading edge (the "head") of the fire. These can only be reproduced in models that allow the fire to modify the winds. For infrared imagery of this fire from U.S.D.A. Forest Service research aircraft, see http://fireimaging.com/ .
Figure 2. View: towards the south. Cabazon, CA, is in the foreground. Animations are at http://www.mmm.ucar.edu/people/coen/files/newpage_m.html . The misty field represents smoke, colored by concentration - higher concentrations are more opaque (linearly with concentration) and darker. The colors identifying the burning parts of the fire are inspired by the radiant temperature color bar at fireimaging.com - brighter colors like yellow reflect higher surface fire sensible heat fluxes. Darker browns are lower fluxes. The other field on the surface is the 'fuel load remaining' - where the fire has passed, the surface appears dark brown. (The spots of dark brown ahead of the fire reflect places where the fuel was light grass and the load was small, not burned out.) The boxiness to the fireline shows the atmospheric grid sizes, onto which the fire fluxes on the fire fuel cell scale (5x5 within each atmospheric cell) have been summed.
Figure 3. Similar to Figure 2, but the arrows show the winds at 1500 m above sea level, near the surface. The length and direction of the vectors show the strength and direction of the horizontal winds. The colors superimposed on the arrows represent the vertical velocity: white is 0, warm colors (yellow to orange to red) reveal air traveling up, cold colors (green increasing to blue increasing to violet) represent air traveling downward. Over the fire area, the arrows are greenish, which is interpreted as a slight downdraft at this elevation. Some areas are mostly white, which means there is no up or down component there. A few vectors near lower right and sometimes over Cabazon Peak (center left) show upward motion. The location and strength of the upward and downward air motions over the fire vary with time. These waves are set off by being forced over the San Bernardino mountain range upwind, and then, encountering the San Jacintos, are complicated by smaller-scale terrain features like Cabazon Peak, shear-generated motions, and convection generated by the fire.
(3) Just how strong is the effect of the fire on its surrounding environment and why is it important?
The modification of the winds by the fire is the cause of virtually all phenomena that create the individual character of large event - splitting of fire fronts, draws of flanks up canyon perpendicular to overall fire spread, how fires drawn themselves together (ex.: deliberately set fires may either be drawn into large wildfires or turn into wildfires themselves), and in the extreme, the generation of fire whirls and blowups, where the firestorm-like connection/grip/bond between the increasing fire intensity and the atmosphere tightens, such that the ‘fire creates its own weather’.
This simulation shows several hours in the early period of the Big Elk Fire, a 2200 ha Colorado wildfire. Fire behavior was extreme reflecting the extremely dry conditions throughout Colorado (including the lowest fuel moistures ever recorded in the area). Initial spread was rapid, moving up a south slope of ponderosa pine mixed with Douglas fir with crowning and torching into high density thin-stemmed lodgepole pine at upper elevations. This case represents a relatively simple scenario, with no large-scale weather features - winds were driven primarily by solar heating of mountain slopes, producing weak afternoon upslope conditions during the active fire periods, and by the fire-induced winds themselves.
Figure 4. The orange lines outline topography contours, increasing toward Kenney Mountain, in the center. The red field shows where the air was warmed at least 10 degrees by the heat released from the fire. The misty white field represent smoke, with denser areas representing higher concentrations. The wind speed is shown by the length of the arrows (longer arrows indicate stronger winds) and direction near the surface.
The magnitude of the effect of the fire on the winds is shown in Figure 5.
Figure 5. The effect of the fire on the winds, shown by the difference between a simulation and one in which the fire could not modify the winds. The perspective is the same as Figure 4, but looking straight down. The strength and direction of the fire’s change of the wind is shown both in the arrows and in the contours. The effect of the fire-atmosphere interactions is that the fire draws air into the base of the plume at its leading edge, creating strong winds over the leading edge of the fire that, as in Figure 1, increase the spread of the fire.
Two important points: (1) The magnitude of the fire’s effects are to change the winds by 10-12 m s-1 near the fire, but they may also changes the winds in the fire’s environment by several m s-1 even 5 km from the fire. This is how one fire may alter the winds affecting a nearby fire. (2) The fire’s effects on the winds make them unsteady, which may cause wind shifts that are a further hazard. Again, these effects are only seen in models that capture the fire’s ability to alter the winds in its environment.
EXAMPLES OF USE:
·
CAWFE has been used
on case
studies of large wildfires (Ex. The Big Elk Fire in Colorado, Coen
2005a; The
Esperanza Fire, Coen and Riggan (2010), and the 2012 High Park Fire) to
aid understanding of these
complicated events. See the examples.
· CAWFE has also been tested as a real-time forecasting tool (Coen 2005b) and can currently be configured to run in real time on a single processor of a desktop computer. Many possible usage scenarios exist: strategic vs. tactical, wildland fire use planning, anticipating prescribed fire burning windows, testing “What-If ?” fuel mitigation scenarios) should be tested and examined has not yet occurred; the current capabilities are appropriate for testing and feedback in conjunction with a practical setting.
ACKNOWLEDGEMENTS:
This
material is
based upon work supported by the National Science Foundation under
Grants No.
0324910, 0421498, and 0835598. The National Center for Atmospheric
Research is
sponsored by the National Science Foundation. Any opinions, findings,
and
conclusions or recommendations expressed in this material are those of
the
authors and do not necessarily reflect the views of the National
Science
Foundation. No endorsement by the
U.S. Department of Agriculture is implied. The model was
developed with support from the USDA Forest Service Riverside Fire
Laboratory and Missoula Fire Laboratory and contributions from
individuals at NCAR, the USDA Forest Service Riverside Fire Laboratory
(now Pacific Southwest Research Station), and The Australian Bureau of
Meteorology.
REFERENCES
Albini, F.A.: PROGRAM BURNUP, 1994: A simulation model of the burning of large woody natural fuels. Final Report on Research Grant INT-92754-GR by U.S.F.S. to Montana State Univ., Mechanical Engineering Dept.
Anderson, H.E., 1982:Aids to determining fuel models for estimating fire behavior. General Technical Report INT-122, U.S. Dept. of Agriculture, Forest Service, Intermountain Forest and Range Expt. Station, Ogden, UT. 26 p.
Clark, T.L., J. Coen, D. Latham: Description of a coupled atmosphere-fire model, 2004: Intl. J. Wildland Fire, 13, 49–64.
Clements, C. B., S. Zhong, S. Goodrick, J. Li, B. E. Potter, X. Bian, W. E. Heilman, J. J. Charney, R. Perna, M. Jang, D. Lee, M. Patel, S. Street, G. Aumann, 2007: Observing The Dynamics Of Wildland Grass Fires: FireFlux: A Field Validation Experiment. Bulletin Am. Meteor. Soc. 88(9), 1369-1382.
Coen, J. L., 2011. Some new basics of fire behavior. Fire Management Today. 71(1), 37-42.
Coen, J. L. 2005a Simulation of the Big Elk Fire using coupled atmosphere-fire modeling. Intl. J. Wildland Fire, 14, 49–59.
Coen, J. L., 2005b, Applications of coupled atmosphere-fire modeling: Prototype demonstration of real-time modeling of fire behavior. Amer. Meteor. Soc. Joint 6th Symp. on Fire & Forest Meteor./Interior West Fire Council Conf. 25-27 October. Canmore, AB, Canada. CD-ROM, Paper 8.1
Coen, J. L., S. Mahalingam, and J. W. Daily, 2004: Infrared imagery of crown-fire dynamics during FROSTFIRE. J. Appl. Meteor. 43,1241-1259.
Coen, J. L. and P. J. Riggan, 2010: A landscape-scale wildland fire study using a coupled weather-wildland fire model and airborne remote sensing. Proceedings of 3rd Fire Behavior and Fuels Conference, October 25-29, 2010, Spokane, Washington, USA. Published by the International Association of Wildland Fire, Birmingham, Alabama, USA. CD-ROM. 12 pp.
Rothermel, R.C. 1972. A mathematical model for predicting fire spread in wildland fuels. Research Paper INT-115, United States Department of Agriculture, Forest Service, Intermountain Forest and Range Experiment Station, Ogden, UT. 42 p.
Keywords: wildfire models, fire behavior, forest fires, fire model, wildland fire model