Wildfire Simulation Software

Charles Erwin
Simple Wildfire Simulator from NOVA
 http://www.pbs.org/wgbh/nova/fire/simulation.html
(requires Flash)
 Wildfire Simulator is a simple computer simulation that
predicts the behavior of fire in a wildland environment.
 Not meant for research, only to demonstrate some basic
ideas about wildfire simulation.
 Programming for this feature is derived from FARSITE.

CS 521: Computational Science 2

 EMBYR: “Ecological Model for Burning
the Yellowstone Region”
 Created by William W. Hargrove and Robert H. Gardner
 Designed to simulate wildfires, the subsequent pattern of
vegetation, and then the next generation of burn
patterns.
 While the EMBYR model parameters could be adjusted
to reproduce a particular historical wildfire exactly, it is
more important to reproduce any wildfire relatively well
on average.
 EMBYR can generate "Risk Maps", which are
constructed from many replications of a single simulated
fire. Cells which burned in many of the replications are
colored black, while cells which burned in only a few
simulations are colored white, with gray levels in
intermediate cases.
CS 521: Computational Science 3
EMBYR Fire Model
 The fire model, EMBYR, depicts the landscape as a grid in
which the dimension of each cell is 50 m (2500 m2).
 Diffusive Spread: Fire spreads from each ignited cell to
any of eight unburned neighbors (the four adjacent cells
and four diagonal cells) as an independent stochastic event
with probability I, where I may range from 0 to 1.
 Each cell burns for a single time step of variable length, and
the fire goes out if new sites are not ignited at each time
step.
 Theoretical studies have demonstrated that if I is less then
a critical value, ic fires are unlikely to propagate across the
landscape

CS 521: Computational Science 4

EMBYR Fire Model (cont)
 They estimated ic by performing 50 simulations for each
value of I (0.245 ≤ I ≤ 0.252 in increments of 0.001) on
a 300x300 grid.
 The proportion of simulations with fires reaching the top
edge of the map after the entire bottom edge was ignited
was 38% for I = 0.250 and 60% for I = 0.251. Since ic is
the threshold at which 50% of the fires reach the
opposite edge of the map, these results indicated that
ic
lies between 0.250 and 0.251.

CS 521: Computational Science 5

EMBYR Fire Model (cont)
 Simulating multiple fuel classes: EMBYR explicitly
simulates multiple classes of fuel by varying the
probability of fire spread as a function of fuel type.
 The fuel classes considered are four successional
stages of lodgepole pine forest, nonforested regions
such as meadows, and nonflammable areas such as
rock, roads, and water.
 Derived probabilities on fire spreading between
different types of fuel.

CS 521: Computational Science 6

EMBYR Fire Model (cont)
 Variation in fuel moisture: EMBYR uses a standard fire
danger measure known as percent 1000-h time-lagged fuel
moisture.
 In this measure, an assumption is made about how long
fuel of a particular diameter would take to soak to the
core, or to dry out once soaked.
 Current internal moisture in fuels of that diameter is
modeled with appropriately time-lagged ambient
atmospheric humidity.
 Obviously, if fuels are sufficiently wet, fires do not occur.

CS 521: Computational Science 7

EMBYR Fire Model (cont)
 Simulating the effects of wind: Three classes of wind
speeds (WS), measured at a standard height of 6.1 m
(20 ft) above the surface, are considered:
 WS 0, with speeds ranging from 0 to 3.1 kph (5 mph)
 WS 1, moderate winds ranging from 3.1 to 21.7 kph (5–35 mph)
 WS 2, strong wind with speeds greater than 21.7 kph
 For each of the three wind speed classes, a bias value b
is used to modify the probability of spread to each
neighboring cell.

CS 521: Computational Science 8

EMBYR Fire Model (cont)

CS 521: Computational Science 9

EMBYR Fire Model (cont)
 Simulating the effects of firebrands: EMBYR
simulates a second mechanism of fire spread — the
production of firebrands which are carried aloft in the
rising convection column, and then drift and fall on
remote sites.
 The ‘spotting’ effect of firebrands is simulated by
permitting each burning site to generate a fixed number
of firebrands as a function of fuel type.

CS 521: Computational Science 10

Simulation: homogeneous landscapes

Area burned (in cells) in a 500x500 cell homogeneous fuel class landscape with a
single fixed ignition as a function of the probability of fire spread, I, to the eight
surrounding neighbors where (a) fire is allowed to propagate by adjacent spread
only (no firebrands), and (b) fire is allowed to propagate by adjacent spread and by
firebrands. The simulation was ended before fire could reach the edge of the map.
Means and standard deviations are shown for five replications.
CS 521: Computational Science 11
Simulation: Using actual Landscapes

CS 521: Computational Science 12

Simulation: Using actual Landscapes
The cumulative
frequency of risk of fires
of increasing size for
four alternative weather
conditions of (from left
to right) (a)
Scenario 1: ‘moist’ with
‘strong’ winds; (b)
Scenario 2: ‘dry’
weather with ‘moderate’
winds; (c) Scenario 3:
‘very dry’ weather with
‘moderate’ winds; and
(d) Scenario 4: ‘very dry’
weather with ‘strong’
winds
CS 521: Computational Science 13
Examples of EMBYR In action

CS 521: Computational Science 14

FARSITE: “Fire Area Simulator”
 Two Dimensional model of fire behaviour and growth
simulator.
 A simple ellipse fit observed fire growth data as well as other
shapes. Regardless of the correct shape (if a single one
exists), the eccentricity of the fire is known to increase with
increasing windspeed or slope steepness or both.
 Cellular Model: Simulate fire growth as a discrete process of
ignitions across a regularly spaced landscape grid.
 In general, cellular models have had diminishing success in
reproducing the expected twodimensional shapes and growth
patterns as environmental conditions become more
heterogeneous.

CS 521: Computational Science 15

FARSITE Model
 Problems with cellular models are avoided by the vector
or wave approach to fire growth modeling (Huygens’
principle).
 The fire front is propagated as a continuously expanding
fire polygon at specified timesteps.
 Essentially the inverse of the cellular method, the fire
polygon is defined by a series of two-dimensional
vertices (points with X,Y coordinates). The number of
vertices increases as the fire grows over time (polygon
expands). The expansion of the fire polygon is
determined by computing the spread rate and direction
from each vertex and multiplying by the duration of the
timestep.
CS 521: Computational Science 16
FARSITE Model (cont)
 The reliance on an assumed fire shape, in this case an
ellipse, is necessary because the spread rate of only the
heading portion of a fire is predicted by the present fire
spread model. Fire spread in all other directions is
inferred from the forward spread rate using the
mathematical properties of the ellipse.
 There are still many problems in accurately simulating
fire with this approach, different methods, however, will
probably be of little consequence to the practical
application of a fire growth model until the greater
uncertainties are resolved as to how wind, slope, and
fuels affect fire shapes.

CS 521: Computational Science 17

FARSITE Model (cont)

CS 521: Computational Science 18

FARSITE Model: Richards’ Equations
Xs, Ys The orientation of the vertex on the
fire front in terms of component
differentials.

θ
The direction of maximum fire
spread rate.

The shape of an elliptical fire

determined from the conditions
a, b, c local to that vertex in terms of
dimensions.

CS 521: Computational Science 19

FARSITE Model (cont)

CS 521: Computational Science 20

 FARSITE: Transformations for Sloping
Terrain
 Richards’ equations were originally developed for flat
terrain.
 On flat terrain, a horizontal coordinate system remains
unchanged when projected onto the ground surface. This
is not the case for sloping terrain.
 This means that the inputs to equations [1] and [2] must
be transformed from the horizontal to the surface plane,
and outputs must be transformed from the surface plane
back to the horizontal plane.

CS 521: Computational Science 21

FARSITE Model (cont)

CS 521: Computational Science 22

FARSITE Model (cont)

CS 521: Computational Science 23

FARSITE Model (cont)

CS 521: Computational Science 24

FARSITE Model (cont)
 Other models used include the Van Wagner crown fire
model, and Albini’s spotting model.
 For input, FARSITE uses GIS raster data in lieu of vector
data. For fuel moisture, BEHAVE and NFDRS equations
are used.

CS 521: Computational Science 25

FARSITE

Raster Landscape input

layers required from the
GIS for FARSITE
simulation.

CS 521: Computational Science 26

FARSITE

Animation of a FARSITE
v4.0x simulation in a 3D
window.

CS 521: Computational Science 27

FARSITE

Screen shot of a
FARSITE v4.00
simulation
utilizing the post-
frontal
combustion
model.

CS 521: Computational Science 28

References
 http://www.pbs.org/wgbh/nova/fire/simulation.html
 http://fire.org
 Finney, Mark A. 1998. FARSITE: Fire Area Simulator-model
development and evaluation. Res. Pap. RMRS-RP-4, Ogden, UT:
U.S. Department of Agriculture, Forest Service, Rocky Mountain
Research Station
 http://research.esd.ornl.gov/~hnw/embyr/
 Hargrove, W.W., R.H. Gardner, M.G. Turner, W.H. Romme, and
D.G. Despain. 2000. Simulating fire patterns in heterogeneous
landscapes. Ecological Modelling 135(2-3):243-263

CS 521: Computational Science 29