1
A conceptual framework for ranking crown fire potential
2
in wildland fuelbeds
3
4
Mark D. Schaaf1,4, David V. Sandberg2, Cynthia L. Riccardi3, and Maarten D. Schreuder1
5
1
6
Air Sciences Inc.
7
421 SW 6th Avenue, Suite 1400
8
Portland, OR 97204, USA.
9
2
10
11
USDA Forest Service, Pacific Northwest Research Station
3200 SW Jefferson Way, Corvallis OR 97331, USA.
12
13
3
Pacific Wildland Fire Science Laboratory, USDA Forest Service, Pacific Northwest
14
Research Station
15
400 N. 34th St., Suite 201
16
Seattle WA 98103-8600, USA.
17
18
19
4
Corresponding author. Telephone: +1 503-525-9394 ext. 11; fax: +1 503-525-9412;
email: mschaaf@airsci.com
1
1
Abstract: This paper presents a conceptual framework for ranking the crown fire potential
2
of wildland fuelbeds with vegetative canopies. This approach extends the work by Van
3
Wagner (1977) and Rothermel (1991), and introduces several new physical concepts to the
4
modeling of crown fire behaviour derived from the reformulated Rothermel (1972) surface
5
fire modeling concepts proposed by Sandberg et al. (this issue b). This framework forms the
6
basis for calculating the crown fire potentials of Fuel Characteristic Classification System
7
(FCCS) fuelbeds (Ottmar and others, this issue). Two new crown fire potentials are
8
proposed: the Torching Potential (TP) and the Active Crown Potential (AP). A systematic
9
comparison of TP and AP against field observations and CFIS model outputs produced
10
encouraging results, suggesting that the FCCS framework might be a useful tool for fire
11
managers to consider when ranking the potential for crown fires or evaluating the relative
12
behavior of crown fires in forest canopies.
2
1
Introduction
2
The Fuel Characteristic Classification System (FCCS; Ottmar and others, this issue)
3
offers the capacity to describe the physical characteristics of any wildland fuelbed no matter
4
how complex, and the capacity to compare one fuelbed to another. FCCS enables the user to
5
assess the absolute and relative effects of fuelbed differences due to natural events, fuel
6
management practices, or the passage of time. The differences can be expressed as native
7
physical differences, such as changes in loadings and arrangements of fuelbed components;
8
or as changes in the potential fire behaviour and effects, such as fire behaviour or fuel
9
consumption (Sandberg and others, this issue b). Most problematic, because of the lack of a
10
broadly applicable and physics-based crown fire model, is the comparison of crown fire
11
initiation and crown fire spread among the various fuelbeds. FCCS does not require specific
12
prediction of crown fire behaviour across the full range of fire environments, but does require
13
a relative ranking of crown fire potential over the full range of wildland fuelbed
14
characteristics.
15
The past 40 years or so of fire research and observations have produced a significant
16
body of literature on crown fires. Van Wagner’s (1964, 1968) papers on experimental crown
17
fires in red pine plantations may be considered the start of the modern era of crown fire
18
research. Subsequent studies ranged from observations of the characteristics of intense,
19
rapidly moving wildfires, to descriptions of fire types (e.g., Van Wagner 1977, Rothermel
20
1991), to a heuristic key for rating crown fire potential (e.g., Fahnestock 1970), to the
21
development of various mathematical models for predicting crown fire behaviour (e.g.,
22
Kilgore and Sando 1975; Scott and Reinhardt 2001; Van Wagner 1977, 1989, 1993). The
3
1
identification of dependent crown fire thresholds by Scott and Reinhardt (2001) based on
2
stylized fuelbeds and Rothermel’s (1972) surface flame length predictions have proven
3
useful to managers within the limitations of current knowledge. A series of observational
4
experiments was completed during the International Crown Fire Modeling Experiment
5
(ICFME) in Canada (Stocks et al. 2004a). Additional refinements and conjectures into crown
6
fire modeling have been advanced by Butler et al. (2004b); Cruz et al. 2003, 2004, 2006a,b,c;
7
Alexander et al. 2006; and Alexander and Cruz 2006.
8
9
Collectively, these studies have shown that the potential for crown fire occurrence does
not depend on any single element of the fuel complex or on any single element in the fire
10
weather environment. Rather, crown fires result from various combinations of factors in the
11
fuel, weather, and topography of the fuelbed. Important factors include: surface fire intensity,
12
canopy closure, crown density, the presence of ladder fuels, height to base of the combustible
13
crown, crown foliar moisture content, and wind speed. This is the foundation for the Fuel
14
Characteristic Classification System (FCCS) crown fire equations. FCCS crown fire
15
potentials (Sandberg et al. 2001; Sandberg et al. this issue a) are based on an updated semi-
16
empirical model that describes crown fire initiation and propagation in vegetative canopies. It
17
is based on the work by Van Wagner (1977) and Rothermel (1991), but contains additional
18
physical concepts for modeling crown fire behaviour derived from the reformulated
19
Rothermel (1972) surface fire modeling concepts proposed by Sandberg et al. (this issue b).
20
This modeling framework is conceptual in nature. It has not yet been comprehensively tested
21
against independent data sets. Its use is currently limited to assessing the crown fire potential
22
of the FCCS Fuelbeds. Additional refinement and verification are needed before the FCCS
23
crown fire model can be considered for wider application.
4
1
The FCCS crown fire modeling framework ranks the relative potential for crown fire
2
initiation and spread of natural fuelbeds based on a set of actual and inferred characteristics.
3
It draws upon published models results from crown fire experiments by others, personal
4
observations of crown fires, and conversations with fire managers. This model is intended to
5
objectively assess, on a relative scale, the probability of experiencing torching or active
6
crown fire spread in any FCCS fuelbed. Currently applied crown fire models (Van Wagner
7
1977; Scott and Reinhardt 2001; Cruz et al. 2006a,b,c; Alexander et al. 2006; Alexander and
8
Cruz 2006) are largely empirically based and appropriate only when applied to the range of
9
stand structures and fire behaviours observed. While they can be very useful in those cases,
10
they do not provide the broad conceptual framework or applicability necessary to compare
11
the crown fire potential within families of dissimilar fuelbeds.
12
This paper makes extensive use of symbols in the various equations that are presented. A
13
comprehensive List of Symbols is presented in Appendix 1.
14
Background
15
Wildland fires are categorized generally in terms of three types: ground fires, surface
16
fires, and crown fires (Peterson et al. 2005). Ground fires are fires that burn fuels in the
17
ground only; for example, duff, roots, and buried decomposing material. Ground fires are
18
characterized by low spread rates but can nevertheless cause considerable injury to live trees
19
and shrubs (Scott and Reinhardt 2001). Surface fires are those that burn in the layer
20
immediately above the surface and consume fuels such as needles, grass, shrubs, and dead-
21
and-down woody debris (Scott and Reinhardt 2001; Peterson et al. 2005). Crown fires occur
22
in forested ecosystems, and may involve the entire fuel complex from just above the mineral
23
soil to the tops of the trees. These fires are of special interest to managers because they
5
1
represent the upper end of observed wildland fire intensities and are characterized by high
2
rates of spread, long flame lengths, and high energy release rates (Butler et al. 2004a).
3
Some of the earliest published work on crown fire models focused on indices of crown
4
fire behaviour. Fahnestock (1970) developed a heuristic key of crown fire potential that
5
incorporated various stand characteristics such as canopy closure, crown density, and the
6
presence or absence of ladder fuels. Similarly, Kilgore and Sando (1975) expressed crown
7
fire potential in giant sequoia (Sequoiadendron giganteum) as a function of the mean height
8
to the canopy base, crown fuel weight, presence or absence of ladder fuels, crown volume
9
ratio, and the vertical profile of the canopy fuel packing ratio. Neither approach modeled the
10
physical mechanisms that control crown fire initiation and spread.
11
Van Wagner (1977) advanced the physical understanding of crown fires by presenting
12
theory and observations on the factors that govern the start and spread of crown fires. He
13
hypothesized that crown fires are initiated when the convective heating from a surface fire
14
drives off the moisture in the crown fuels, raising the fuel elements to ignition temperature.
15
Van Wagner (1977) defined the critical fireline intensity (that is, the minimum fireline
16
intensity) required for crowning by rearranging a relationship developed by Thomas (1963)
17
that linked fire intensity (as defined by Byram 1959) with the maximum temperature in the
18
convective plume above the fire. The resulting equation expressed the critical fireline
19
intensity as a function of the canopy base height, the foliar moisture content of canopy fuels,
20
and a proportionality constant defined by Van Wagner (1977) as “an empirical constant of
21
complex dimensions.” The value of the proportionality constant was estimated to be 0.01
22
based on several assumptions appropriate for red pine (Pinus resinosa) plantation stands in
23
the lake region of southern Canada and northern United States, including: an estimated
6
1
minimum surface intensity of 2500 kW m-1, 6-m crown base height, and foliar moisture
2
content of 100%.
3
The Van Wagner (1977) model has several limitations (outlined in Cruz et al. 2004): (1)
4
the original formulation by Yih (1953) relates the heat transfer to the maximum temperature
5
in the heat plume and not, more appropriately, to the total heat production found by
6
integrating the time-temperature profile, (2) the model relies exclusively on convection
7
theory, ignoring the contribution of upward radiant heat fluxes in heating the fuel elements to
8
ignition temperature, (3) the model does not properly reflect the influence of wind flow in
9
tilting the heat plume and entraining air into the plume (Mercer and Weber 1994), and (4) the
10
proportionality constant is not universal and should vary with changes in the structural
11
characteristics of different fuelbed complexes (e.g., Alexander 1998, Mercer and Weber
12
2001, Cruz et al. 2004). Despite these limitations, the Van Wagner (1977) model is used in
13
whole or in part in several North American systems used to predict crown fire initiation (Van
14
Wagner 1989; Forestry Canada Fire Danger Group 1992; Finney 1998, 2004; Scott and
15
Reinhardt 2001).
16
Scott and Reinhardt (2001) provided an example of a new modeling approach based on
17
earlier work by Rothermel and Van Wagner. They combined the Rothermel equations (1972,
18
1991) and the Van Wagner equations (1977) into two crown-fire initiation indices: a torching
19
index, and a crowning index. The Torching Index is the 6.1-m wind speed at which crown
20
fire is expected to initiate based on Rothermel’s (1972) surface fire model and Van Wagner’s
21
(1977) crown fire initiation criteria. The Crowning Index is the 6.1-m wind speed at which
22
active crowning is possible based on Rothermel’s (1991) crown fire spread rate model and
23
Van Wagner’s (1977) criterion for active crown fire spread. Both indices are used to
7
1
ordinate different forest stands by their relative susceptibility to crown fire and to compare
2
the effectiveness of crown fire mitigation treatments.
3
Combining and refining the Van Wagner (1977) and Xanthopoulos (1990) approaches,
4
Alexander (1998) developed an algorithm to predict the onset of crown fire. His approach
5
used an estimate of the convective plume angle based on the Taylor (1961) and Thomas
6
(1964) relationship between plume angle and fireline intensity and wind speed, and an
7
estimate of the temperature increase above the ambient temperature at the base of the crown
8
using a modification of Byram’s (1959) fireline intensity equation. This approach possessed
9
several limitations that are common to other models, including the use of Byram’s (1959)
10
fireline intensity, which is not necessarily a good descriptor of the surface heat fluxes
11
reaching the base of the crown; the inadequacy of current methods for predicting residence
12
times as a function of fuelbed properties and fuel availability for flaming; and the need to
13
parameterize the several constants for different fuelbed complexes with different structural
14
characteristics.
15
Building on earlier work by Cruz (1999), Cruz et al. (2003b, 2004) developed a
16
probabilistic model, based on logistic regression techniques, for the prediction of crown fire
17
occurrence (i.e., initiation) based on several fire environment and fire behaviour variables
18
normally available to support fire-management decisions. These variables included: 10-m
19
open wind speed, fuel strata gap, surface fuel consumption (a surrogate for fireline intensity),
20
and estimated fine-fuel moisture content. The strength of this model is that it predicts the
21
probability of a crown fire rather than the dichotomous “crown/no crown” results from Van
22
Wagner (1977) and Alexander (1998). Its greatest limitation is a lack of physical reasoning.
23
For example, the model does not directly account for the heat energy released by a surface
8
1
fire but instead uses surface fuel consumption as a surrogate. Other limitations include: bias
2
in the data set used to condition the model (e.g., in 60 percent of the situations the fuel strata
3
gaps were less than 3 m), inability to consider the influence of crown bulk density on crown
4
fire spread beyond the fuel strata gap, and inability to account for point-source fires or
5
prescribed-fire ignition patterns (e.g., perimeter mass-ignitions) that do not approximate
6
those of free-burning wildfires. These models are most applicable to live conifer forests on
7
level terrain. Moreover, the majority of fuelbeds used to develop the model were conifer
8
stands found in Canada and the northern United States. It is not known whether this model
9
can be accurately applied to other forest types and regions.
10
Cruz et al. (2006a,b,c) described the development and testing of a semi-physical model
11
for predicting the temperature and ignition of canopy fuels above a spreading surface fire,
12
titled the Crown Fuel Ignition Model (CFIM). CFIM uses a set of physical equations, along
13
with empirically based submodels, to define the heat source, buoyant plume dynamics, and
14
radiative and convective energy transfer to the fuel elements at the base of the canopy layer.
15
Simulation testing showed that the fuel strata gap and moisture content of fine dead surface
16
fuels were the dominant variables controlling the fuel-particle temperature rise and
17
subsequent crown fuel ignition. Flame-available fuel loading and 10-m open wind speed
18
were of lesser importance. Foliar moisture content and canopy surface-area-to-volume ratio
19
showed the least effect. The authors make no claims about the efficacy or performance of
20
the models. All that is said is the test results were considered to be “within the range of
21
predictions” produced by Van Wagner (1977), Alexander (1998), and Cruz et al. (2004).
22
Alexander et al. (2006) reported on the development and testing of the Crown Fire
23
Initiation and Spread (CFIS) model, a crown-fire modeling system that incorporates elements
9
1
of CFIM (Cruz et al. 2006a,b,c) in addition to several other models designed to simulate
2
various aspects of crown fire behaviour. In determining whether a crown fire is active or
3
passive, CFIS uses the criterion for active crowning (suggested by Van Wagner 1977), which
4
is a function of the canopy bulk density and the predicted active crown fire rate of spread.
5
The active crown fire rate of spread was represented by a non-linear regression equation with
6
independent variables of 10-m open wind speed, estimated fine-fuel moisture content of dead
7
surface fuels, and canopy bulk density. CFIS provides the user with a means of evaluating
8
the impacts of proposed fuel treatments on the potential crown fire initiation and spread. It is
9
not known whether CFIS is superior to other models for predicting the initiation and spread
10
of crown fires. CFIS is considered most appropriate for free-burning fires that have reach
11
pseudo steady state in live, boreal or boreal-like conifer forests on level terrain.
12
Alexander and Cruz (2006) evaluated the predictive capacity of the Cruz et al. (2005)
13
crown fire spread rate models against an independent set of wildfire observations from the
14
United States and Canada. The Cruz et al. (2005) model integrates the effects of 10 m open
15
wind speed, canopy bulk density, and estimated fine-fuel moisture content to yield estimates
16
of the crown rate of spread for both passive and active crown fires after the fire type has been
17
determined using Van Wagner’s (1977, 1993) criterion for active crowning. The model
18
performed reasonably well against a large sample set of 57 North American wildfires,
19
producing fewer under predictions than the Rothermel (1991) spread model, which was
20
included in the evaluation. However, critical information on canopy bulk density was
21
missing on all but two of the 43 Canadian wildfires, and five of the 14 American wildfires.
22
Instead the missing values were assigned to one of several broad classes based on forest type.
10
1
Canopy bulk density is an important variable in determining the crown fire rate of spread and
2
the use of inferred values diminishes the value of this evaluation.
3
Scott (2006) compared the relative behaviour of the surface and crown fire rates of spread
4
contained in three fire management applications: FlamMap (Stratton 2004), NEXUS (Scott
5
1999), and CFIS (Alexander et al. 2006). FlamMap and NEXUS are largely based on
6
Rothermel (1991) and Van Wagner (1993), although they differ in their implementation.
7
Scott (2006) found that the three models predicted considerably different crown fire rates of
8
spread, especially in the transitions from surface to passive and from passive to active crown
9
fire (Scott 2006). However, the relative, ordinal, ranking of crown fire behaviour was similar
10
for the three systems (Scott 2006). FCCS provides the same kind of relative ranking of
11
potential crown fire behaviour for all of the 216 FCCS fuels beds (Ottmar and others, this
12
issue).
13
All of these methods used to assess crown fire potential have limitations. Most of these
14
methods are semi physical (or semi empirical) in nature, combining physical equations of
15
heat generation and transfer and fuelbed characteristics with one or more empirically based
16
calibration constants. Many of the physical characteristics of the fuelbed, of fuel
17
consumption, and of fire conditions are difficult to obtain or missing from previously
18
reported studies. This makes it challenging to populate the physical and semi-empirical
19
models with appropriate parameters. In most cases, model development has been based on a
20
limited set of actual fires or in a limited geographic area.
21
The conceptual framework offered in this paper provides decision makers with a tool for
22
objectively assessing, on a relative scale, the probability of experiencing torching or active
23
crown fire spread in any FCCS fuelbed. Although it suffers from some of the same
11
1
limitations as earlier models, its strength is in providing a comprehensive consideration of the
2
factors that are commonly used by fire behaviour analysts to assess crown fire potential.
3
Some factors that are considered in the FCC conceptual framework, such as canopy
4
continuity and ladder fuel abundance, are found in few or none of the other models. By
5
advancing this transparent quantitative framework, which includes several heuristic
6
algorithms, we provide a more robust consideration of factors that can be evaluated against
7
future observations and experience by users.
8
9
10
Model description
The FCCS crown fire potential equations are based on a two-layer system consisting of a
11
surface layer and a canopy layer. The surface layer is composed of fuel elements in the
12
woody, nonwoody (e.g., grasses and forbs), shrub, and litter-lichen-moss fuelbed strata. The
13
canopy layer, in turn divided into three sublayers (understory, mid story, overstory), is
14
composed of woody and nonwoody (e.g., foliage) fuel components. A vertical gap separates
15
the combustible fuels in the canopy layer from those in the surface layer. For purposes of
16
estimating the propagating flux ratio, the canopy is assumed to be composed of uniformly
17
distributed thermally thin, short, cylindrical fuel elements. We have assumed that the
18
unburned fuels ahead of the canopy fire heated by a combination of radiant heat transfer and
19
forced convection under laminar flow conditions over the canopy elements (Reynolds
20
numbers 50 to 400). In this simplified formulation, terrain effects are ignored (i.e., flat
21
terrain) and wind speed does not explicitly aid in heat transfer other than by changing the
22
crown-to-crown flame transmissivity within canopies with less than 100% cover.
12
1
Basic model
2
The general form of the FCCS crown fire potential (CFP) equation is:
3
4
[1]
CFP = max (TP, AP )
5
6
where TP is the torching potential and AP is the active crown fire potential. Both are
7
dimensionless, ranging in value from zero to 10. TP is the potential for a surface fire to
8
spread into the canopy as single tree or multiple tree torching. If TP is greater than one, then
9
torching is possible. TP is defined as the scaled crown fire initiation term, IC (dimensionless,
10
range zero to 10):
11
12
[2]
TP = cTP ⋅ I C
13
14
Here, cTP is a scaling function to limit TP within the range of zero to 10. Fuelbeds with IC
15
values greater than ten are assigned a TP of ten. Fuelbeds with IC values less than ten are
16
scaled from zero to ten.
17
18
AP is the potential for a surface fire to spread into the canopy as an active crown fire. If
19
AP is greater than zero, than active crown fire spread is possible. AP is computed as the
20
scaled product of four terms:
21
22
[3]
AP = c AP ⋅ I C ⋅ FC ⋅ RC
23
13
1
Here, IC is the crown fire initiation term, FC is the crown-to-crown flame transmissivity term
2
(dimensionless, range zero to one), and RC is the crown fire spread-rate term (m min-1, range
3
one to >100 m min-1). cAP is a scaling function that limits AP to a range of zero to 10
4
(dimensionless). Fuelbeds with the product of IC .FC .RC greater than 10 are assigned an AP
5
of 10, and fuelbeds with the product of IC .FC .RC less than 10 are scaled from zero to 10.
The IC, FC, and RC terms are developed below.
6
7
Crown fire initiation term (IC)
The crown fire initiation term originates from an early definition of fire intensity, IB (kJ
8
9
B
m-1 s-1 or kW m-1), from Byram (1959):
10
11
[4]
IB =
H Wf R
60
12
13
where H is the heat yield of the fuel (kJ kg-1), Wf is the weight of the fuel consumed in the
14
flaming front (kg m-2), R is the forward rate of spread of the fire (m min-1), and 60 is a
15
conversion factor that reduces IB from units of kJ m-1 min-1 to units of kJ m-1 s-1 (kW m-1).
16
Andrews and Rothermel (1982) defined heat per unit area, H’ (kJ m-2), equivalent to HWf in
17
Eq. 4:
B
18
19
[5]
H ' = H W f = I R tR
20
21
where IR is reaction intensity (kJ m-2 min-1), and tR is residence time in minutes. Combining
22
Eqs. 4 and 5 above yields:
14
1
2
[6]
IB =
I R tR R
60
3
4
To define the concept of a critical fireline intensity, Van Wagner (1977) rearranged a
5
relationship (developed by Thomas 1963) linking fire intensity as defined by Byram (1959)
6
(see Eq. 4) with the maximum temperature in the convection plume above the fire. The
7
critical fireline intensity, I’ (kJ m-1 s-1 or kW m-1) is the minimum fireline intensity at which
8
crowning is initiated (after Van Wagner 1977, 1989):
9
10
[7]
⎛ CBH (460 + 25.9 FMC ) ⎞
I'=⎜
⎟
100
⎝
⎠
3
2
11
12
where CBH is canopy base height (m), FMC is foliar moisture content of canopy fuels (%),
13
and 100 is an empirical constant. One aspect of crown fire modeling that would greatly
14
benefit from additional field research is the critical bulk density or minimum fuel load
15
required to define the CBH.
16
As noted earlier, the basis for Van Wagner’s (1977) model is that a crown fire occurs
17
when the convective heat supplied by a surface fire drives off the moisture in the crown fuels
18
and raises them to ignition temperature. However, recent studies using statistically based
19
models have shown that the fine-fuel moisture content of dead woody fuels on the surface
20
has a much greater effect on the ignition of crown fuels than the canopy foliar moisture
21
content (Cruz 1999; Cruz et al. 2004; Cruz et al. 2006c). The explanation given in Cruz et al.
22
(2006c) is that the heating of canopy fuel elements is continuous and prolonged, and that the
15
1
heat energy required to evaporate the foliar moisture content is small compared with the total
2
heat energy released from high intensity surface fires. Foliar moisture content could still be
3
an important variable under conditions of low surface fireline intensity and high foliar
4
moisture content, or in cases where the heating of the canopy fuel elements is less effective
5
due to a high stand density or large fuel strata gap. Several recent papers have acknowledged
6
the importance of foliar moisture content in determining crown fire behaviour (Xanthopolous
7
and Wakimoto 1993, Butler et al. 2004b).
8
Following Van Wagner (1977), crown fire initiation is expected whenever the surface
9
fireline intensity exceeds the critical fireline intensity; that is, when IB/I’ >1. The ratio of Eq.
10
B
6 and 7 yields:
11
12
[8]
IB
16.667 I R t R R
=
I ' [CBH (460 + 25.9 FMC ) ]3 2
13
14
Eq. 8, expressed in terms of FCCS variables, with the residence time set proportional to the
15
inverse of the surface potential reaction velocity, defines the FCCS crown fire initiation term,
16
I C:
17
18
[9]
IC =
(
16.667 I R. FCCS (1/ Γ ' ) RFCCS .S
)
⎡ gap
(460 + 25.9 FMC ) ⎤
⎢⎣
⎥⎦
ladder
3
2
19
20
where Γ' is the potential reaction velocity (min-1) of surface fuels from Rothermel (1972),
21
RFCCS.S is the surface fire spread rate from FCCS (m min-1), gap is the physical separation
16
1
distance (m) between the top of the surface fuel layer and the bottom of the combustible
2
canopy layer, and ladder is an heuristically assigned value representing the presence and type
3
of ladder fuels sufficient to act as a vertical carrier of fire to the canopy base (default ladder
4
=1, meaning no ladder fuel). These terms are as defined in Sandberg et al. (this issue b).
5
While the traditional Van Wagner (1977) equation (Eq. 5), bases the calculation of I ' on
6
canopy base height (see denominator of Eq. 7), the FCCS I ' uses the vertical gap between the
7
top of the surface fuelbed layer and the bottom of the combustible canopy layer, which may
8
be adjusted by a factor related to the abundance of combustible ladder fuels. The validity of
9
this modification will be evaluated in future model validation efforts. Note that Cruz (1999),
10
Cruz et al. (2004), and Cruz et al. (2006a) use the same measure of fuel layer separation
11
(which they term fuel strata gap), which can also be modified for ladder fuels.
12
IC is evaluated along a continuum ranging from zero to infinity. The higher the IC value,
13
the greater is the potential for initiating a crown fire. This is the same equation set used in
14
Scott and Reinhardt (2001) except that they structured the equations in a manner that
15
established midflame wind speed as the principal variable, whereas we have structured the
16
equations to evaluate the initiation potential across a range of fuelbeds with different surface
17
reaction intensities, and rates of spread at a variable benchmark wind speed (default
18
midflame wind speed is 107 m min-1, or ~6.4 km h-1).
19
The influences of reaction efficiency, gap/ladder distance, and foliar moisture content on
20
IC are displayed in Figs. 1a,b,c. These formulations are all based on the assumption that
21
energy from the surface fire that initiates crowning is attributable to the propagating surface
22
flame, although it is well accepted that torching and active crown fire initiation also may
23
occur after flame front passage from heat supplied by residual surface flames or from
17
1
radiating smoldering fires. We expect to present a more complete description of surface-to-
2
canopy heat transfer in the future.
3
Crown-to-crown flame transmissivity term (FC)
4
The FCCS crown-to-crown flame transmissivity term (FC) is a dimensionless measure of
5
the capacity of the canopy to transfer flames through the canopy based on leaf area index
6
(LAI), wind speed, and horizontal continuity of tree crowns. The higher the wind speed, the
7
higher the effective horizontal continuity of tree crowns and the higher the crown-to-crown
8
transmissivity. And the higher the transmissivity, the greater is the potential to sustain an
9
active crown fire. Torching is not affected by the horizontal continuity of tree crowns.
10
This new conceptual term is proposed as a replacement for the model originally proposed
11
by Van Wagner (1977), which determines whether an active crown fire will occur by
12
comparing the estimated crown fire spread rate with a critical spread rate required to sustain
13
an active crown fire. Although practical and widely used, Van Wagner’s (1977) model
14
assumes that the canopy is horizontally uniform and continuous. It does not explicitly
15
account for spacing between tree crowns nor does it consider the effect of increasing mid-
16
canopy wind speed in reducing effective spacing. Moreover, application of the Van Wagner
17
(1977) model relies on an estimate of crown-fire spread rate based on a limited correlation
18
developed by Rothermel (1991). However, this new, more physically intuitive FC term is less
19
supported by observations than the approach developed by Van Wagner (1977) and
20
Rothermel (1991). Additional testing of this modeling concept is planned.
21
The FCCS crown-to-crown flame transmissivity term, FC, is defined as follows:
22
18
1
[10]
for LAI < TLAI ⎫
⎧ 0,
⎪⎪
⎪⎪
0.3
FC = ⎨ max 0, ([COV × WAF ] − TCOV )
⎬
, for LAI ≥ TLAI ⎪
⎪
0.3
([100 × WAF ] − TCOV )
⎪⎩
⎪⎭
)
(
2
3
where LAI is leaf area index (m2 m-2), TLAI is threshold LAI for active crowning (m2 m-2),
4
COV is total percentage cover of tree crowns (i.e., percentage canopy cover, or percentage
5
ground area covered by tree crowns) (dimensionless), WAF is a canopy wind speed
6
adjustment factor (dimensionless), and TCOV is threshold percentage canopy cover
7
(dimensionless) required to propagate an active crown fire when WAF = 1 (TCOV = 40). This
8
formulation assumes that a relatively continuous canopy is required for efficient crown-to-
9
crown heat transfer. Whether this is true or not could be the subject of additional field
10
research. In the interim, we propose that the influence of cover and wind speed on TC is as
11
described by Eq. 10 and illustrated in Fig. 2.
12
13
Threshold leaf area index (TLAI)
14
TLAI may be estimated using Van Wagner’s (1977) empirical relationship that describes
15
the interaction between canopy bulk density (CBD; kg m-3) and minimum spread rate needed
16
to sustain an active crown fire (RAC; m min-1):
17
18
[11a]
CBD =
0.05
RAC
19
19
1
The numerator in this equation is an empirically derived critical mass flow rate (units of
2
kg m-2 s-1). Alexander (1988) used a critical mass flow rate of 3.0 kg m-2 min-1 in order to
3
express the RAC in units of m min-1:
4
5
[11b]
CBD =
3.0
RAC
6
7
8
Rothermel (1991) derived the following empirical relationship for the active-crowning
rate of spread, Ractive (m min-1):
9
10
[12]
Ractive = 3.34 ( R10 ) 40% WRF
11
12
where (R10)40% WRF is the spread rate (m min-1) predicted by Rothermel (1972) using fuel
13
characteristics for Fire Behaviour Prediction System model 10 and a midflame wind speed
14
set at 40 percent of the 6.1-meter wind speed. Rothermel (1991) did not address the
15
application of his crown fire spread model, other than to say that it would be used under
16
“severe burning conditions” conducive to crown fire. Replacing the critical rate of spread in
17
Eq. 11b with the active-crowning rate of spread in Eq. 12 allows us to fit a relationship
18
between the critical CBD and surface rate of spread for model 10:
19
20
[13]
CBDcritical =
3.0
3.34 ( R10 ) 40% WRF
21
20
1
The TLAI required to propagate an active crown fire may be defined in terms of CBDcritical,
2
surface-area-to-volume ratio of foliage elements (σ, m2 m-3), and particle density (ρp; kg m-3):
3
[14a]
CBD =
Mass of canopy fuel elements
Volume of canopy
[14b]
CBD =
(Volume of canopy fuel elements ) ( ρ p )
( Area of ground )( Dc )
[14c]
CBD =
( Surface area of canopy fuel elements ) ( ρ p )
( Area of ground )( Dc ) (σ )
10
[14d]
CBD =
11
where ρp is the particle density (kg m-3), σ (m2 m-3) is the surface-area-to-volume ratio of
12
foliage elements, and DC is the mean canopy depth (m). The TLAI is the LAI that is
13
required to produce the critical CBD. Thus:
4
5
6
7
8
9
LAI ρ p
Dc σ
14
15
[14e]
CBDcritical =
TLAI ρ p
Dc σ
16
21
1
where DC is the mean canopy depth (m). Substituting Eq. 13 into Eq. 14e yields the following
2
for the TLAI required to sustain an active crown fire:
3
4
[15]
TLAI =
3 σ DC
3.34 ( R10 ) 40% WRF ρ p
5
6
The BEHAVE algorithm (Andrews 1986) was used to compute (R10)40% WFR as a function
7
of the 6.1-meter open wind speed (U6.1; m min-1) for a 1-hr surface fuel moisture content of 9
8
percent. Eq.15 was then used to compute TLAI/DC, assuming a foliage particle density ρp of
9
400 kg m-3 and surface area-to-volume ratios σ of 6,562 and 8,202 m2 m-3 (FCCS default
10
values) for coniferous trees and broadleaf shrub species, respectively. The resulting general
11
equation is:
12
13
[16]
−3
TLAI
= A e −2.6 x10 U 6.1
DC
14
15
where A is 10.8 and 8.6 for coniferous trees and flammable broad-leaved shrub species,
16
respectively.
17
18
Canopy wind speed adjustment factor (WAF)
19
The wind speed adjustment factor increases (decreases) the effective canopy cover for
20
wind speeds that are higher (lower) than the FCCS benchmark wind speed of 107 m min-1.
22
1
The effective canopy cover is the product of the actual cover percent COV and the wind
2
speed adjustment factor WAF. The canopy wind speed adjustment factor, WAF, is given by:
3
4
[17]
WAF =
U/
U 2 +W 2
UB /
U B2 + W 2
5
6
where U is horizontal midflame wind speed (m min-1), W is an assumed vertical,
7
convectively driven wind speed of 1,590 m min-1 (95.4 km h-1), and UB is the FCCS
8
benchmark horizontal midflame wind speed of 107 m min-1 (6.4 km h-1). The W constant was
9
based on the average maximum vertical wind speed above 16 experimental crown fires in
B
10
black spruce reported by Clark et al. (1999; Table 2). If the mid-flame wind speed is higher
11
(lower) than the benchmark wind speed, then effective canopy cover will be higher (lower)
12
by the WAF shown in Fig 3.
13
The use of FC to assess the potential for propagating an active crown fire is departs from
14
the conventional approach comparing actual crown rate of spread with RAC (Eqs. 11 and 12).
15
Unlike the combined Van Wagner (1977) and Rothermel (1991) approach, FC takes into
16
account horizontal continuity of tree crowns as well as threshold density (as measured by
17
threshold LAI) when assessing the potential for propagating an active crown fire.
18
19
20
21
Crown fire rate of spread term (RC)
This section presents a new physically-based mathematical approach for estimating the
crown fire spread rate using the reformulated Rothermel’s (1972) surface fire spread rate
23
1
adapted to vegetative canopies. Sandberg et al. (this issue b) proposed the following
2
reformulation of Rothermel’s (1972) spread rate (m min-1) for surface fires:
3
4
[18] RFCCS =
IR.FCCS ξ (1 + φW.FCCS )
⎡⎣FAI ζ I ρp Qig / δ ⎤⎦
+ ⎡⎣θ ρb Qig ⎤⎦
+ ⎡⎣θ ρb Qig ⎤⎦ + ⎡⎣θ ρb Qig ηΔ'.llm ⎤⎦
woody
non-woody
shrub
llm
5
6
where IR.FCCS is the reaction intensity (kJ m-2 min-1), ξ is propagating flux ratio (the
7
proportion of IR.FCCS transferred ahead of the surface fire to the unburned fuels
8
(dimensionless) 1+φW.FCCS is the acceleration factor for wind (dimensionless), FAI is the
9
fuel-area index of surface fuelbed categories (m2 m-2), ζI is the ignition thickness heated on
10
the surfaces of thermally thick fuel elements (m), ρp is particle density of the woody fuels
11
(400 kg m-3), ρb is bulk density of the fuels (kg m-3), Qig is the heat of pre-ignition (i.e., the
12
amount of energy required to raise the fuel temperature to ignition, a function of fuel
13
moisture content, kJ kg-1), δ is the depth of the woody fuelbed stratum (m), θ is the planform
14
area proportional to the coverage for the individual non-woody, shrub, and litter-lichen-moss
15
strata (m2 m-2), and η∆’.llm is the absorption efficiency of the litter-lichen-moss stratum
16
(assumed equal to the reaction efficiency, dimensionless). The numerator of this equation
17
represents the surface-fuel heat source acting to accelerate the fire spread. The denominator
18
contains the individual heat sinks acting to retard the fire spread. These terms are all as
19
defined in Sandberg et al. (this issue b).
20
For active crown fires, the combined reaction intensity from the flaming combustion of
21
the surface and canopy fuels should result in greater forward heating of the fuels and greater
22
spread rates. In equation form, this is expressed as:
24
1
2
[19]
IR.FCCS = IR.FCCS.S + IR.FCCS.C
3
4
where IR.FCCS.S is the reaction intensity of surface fuels (defined in Sandberg et al. this issue
5
b) and IR.FCCS.C is the reaction intensity of canopy fuels. Eq. 18 may be further adapted to
6
crown fires (designated by subscript c) by eliminating the shrub and litter-lichen-moss sink
7
terms in the denominator, by replacing the FAI for the nonwoody fuels with LAI, by
8
replacing ξ with a relationship appropriate for canopies, by replacing δ with the mean crown
9
depth (DC, m); and by replacing Rothermel’s (1972) wind speed coefficient, 1 + φW . FCCS , with
10
the dimensionless wind speed coefficient, U/UB, introduced in Eq. 17. The result is a spread
11
rate equation for canopies that is influenced by both the surface and canopy heat sources and
12
heat sinks:
13
14
[20]
15
RFCCS.C =
(IR.FCCS.S + IR.FCCS.C ) ξC (U / UB )
2
+
∑⎡⎣FAI ζ I ρp Qig / DC ⎤⎦canopy
woody
i=1
2
∑⎡⎣θ LAI ζ
i =1
I
ρp Qig / DC ⎤⎦canopy + surface sink terms, Eq. [17]
non-woody
16
17
18
This spread rate equation is designed to accommodate a variety of vegetative canopies. The
19
index i represents the overstory and understory canopy subcategories. The IR.FCC.C, ξC, and
20
DC terms are described in more detail below. The calculation of ζI, ρb, Qig, and θ terms are as
21
described in Sandberg et al. (this issue b).
22
25
1
2
3
Canopy reaction intensity
The reformulated Rothermel (1972) reaction intensity, IR.FCCS.S, for the surface fuelbed is
(Sandberg et al. this issue b):
4
3
5
[21]
I R.FCCS .S = ηΔ '.S ∑ (Γ'max.FCCS ϒR ρP hηM .FCCS ηK )i + ⎡⎣ηΔ '.llm Γ'max.FCCS ϒR ρP hηM .FCCS ηK ⎤⎦
i =1
llm
6
7
where ηΔ’.S is the surface-fuel reaction efficiency (dimensionless), Γ'max.FCCS is the maximum
8
surface reaction velocity for fuels of mean surface-area-to-volume ratio at the optimum
9
packing ratio (min-1), γR is the reaction volume of fuels involved in the reaction zone (m3 m-2,
10
volume density of fuels that contribute energy forward to unburned fuels), ρp is the oven-dry
11
particle density (kg m-3), h is the fuel low heat content (kJ kg-1), ηM.FCCS is the moisture
12
damping coefficient (dimensionless, acts to reduce the reaction velocity), ηK is mineral
13
damping coefficient (dimensionless, assumed to be 0.42, consistent with 1% silica-free ash
14
content), and ηΔ’.llm is the reaction efficiency for the litter-lichen-moss stratum
15
(dimensionless).
16
This equation may be adapted to crown fires by replacing the surface-fuel reaction
17
efficiency, ηΔ’.S, with the canopy reaction efficiency, ηΔ’.C, and by eliminating the litter-
18
lichen-moss term:
19
20
[22]
I R.FCCS .C = ηΔ '.C
2
∑ (Γ
i =1
'
max. FCCS
ϒR ρP h ηM .FCCS .C ηK )i
21
26
1
Here, the index i represents the overstory and understory canopy subcategories. The ηΔ’.C
2
and ηM.FCCS.C terms are described in greater detail below. The γR, ρp, h, Γ’max.FCCS, and ηK
3
terms are as defined above as well as in Sandberg et al. (this issue b).
4
5
6
Canopy reaction efficiency
Not all tree canopies burn with the same efficiency because of differing crown
7
morphologies and densities, among other factors. Fahnestock (1970) attempted to capture
8
this in a heuristic equation that reduced crown fire potential by as much as 40 percent based
9
on crown density. Although Rothermel (1972) did not examine tree canopies when he
10
developed his surface fire spread model, we have applied his concept of an efficiency term
11
based on the relative packing ratio to the rating of the crown fire hazard. Rothermel (1972)
12
observed that the optimum packing ratio in surface fuels shifted to a lower value in the
13
presence of wind, and we have assumed that the same will occur in canopies. Rothermel’s
14
estimate of optimum density, corrected for convective flow in the canopy, provides a basis
15
for estimating the reaction efficiency in the canopy.
16
The canopy reaction efficiency term, ηΔ ' c , is the ratio of the reaction velocity over the
17
potential reaction velocity of canopy fuels (range 0 to 1). Reaction velocity is a dynamic
18
variable that quantifies the completeness and rate of the fuel consumption. It is strongly
19
influenced by the fuelbed density, particle size, moisture content, and mineral content
20
(Rothermel 1972), as expressed by the η∆’.C, ρp, ηM.FCCS.C, and ηK terms in Eq. 21.
21
For canopy layers, the canopy reaction efficiency, η∆’.C, is defined as:
22
27
1
[23]
η Δ ' C = ⎡ Δ 'C ⋅ exp {1- Δ C' }⎤
⎢⎣
0.2
⎥⎦
2
3
where Δ C' is the relative depth of the canopy (dimensionless) (Fig. 4), and the exponent 0.2
4
expresses the response of the reaction efficiency to changes in relative crown depth. The
5
canopy reaction efficiency is analogous to the surface fuelbed reaction efficiency, ηΔ’.S
6
(Sandberg et al. this issue b), but less sensitive to deviations from the optimum air-to-fuel
7
particle ratio. The concept of relative depth, Δ C' , replaces Rothermel’s (1972) relative
8
packing ratio, as described in Sandberg et al. (this issue b). The two would be numerically
9
identical except for changes made in estimating the effective heating number, ε. The relative
10
canopy depth is defined as the optimum canopy depth, δopC (m), divided by the mean canopy
11
depth, DC (m):
12
13
[24]
Δ C' =
δ opC
DC
14
15
The mean canopy depth, DC, is the difference between the mean total canopy height and the
16
mean height to the base of the combustible crowns, or the canopy base height, CBH.
17
18
Moisture damping coefficient
19
The canopy foliar moisture content damping term in the FCCS, ηM.FCCS’ is based on Van
20
Wagner’s (1977) relationship of foliar moisture to canopy ignition (dimensionless) (Fig. 5):
21
28
1
[25]
η M.FCCS
⎛
⎞
m
=⎜
⎟
⎝ max(m, FMC ) ⎠
0.61
2
3
where m is the reference foliar moisture content at maximum foliar flammability
4
(dimensionless), and FMC is the mean live foliar moisture content of the canopy layer in
5
question (upper, mid, and under story). The exponent 0.61 approximates Van Wagner’s
6
(1977) influence of foliar moisture on canopy flammability (dimensionless).
7
8
Flammability is assumed to be constant below some reference (minimum) foliar moisture
content m. For canopies, the default value of m is 75 percent.
9
10
11
Propagating flux ratio
The propagating flux ratio represents the proportion of reaction intensity transferred to
12
the unburned fuel by radiative and convective processes. The total propagating flux ratio in
13
Eq. 20 is assumed to be the sum of the propagating flux ratios for radiation and convection:
14
15
[26]
ξC = ξ radiation + ξ convection
16
17
In some instances, radiant energy transfer dominates the process of transferring heat energy
18
from the fire to the unburned crown fuels ahead of the fire; for example, when fires spread in
19
still air or against the wind (Butler et al. 2004a). Other times, convective energy transfer
20
clearly dominates when fires burns upslope or under very high winds, both of which reduce
21
the flame angle relative to the fuel surface. Both mechanisms are important for evaluating the
22
rate of forward heating in fires.
29
1
The fraction of radiant energy transferred to the unburned crown fuels may be
2
approximated using a modification of Beer’s Law, which expresses the amount of radiant
3
energy absorbed ahead of the fire by the unburned fuel elements:
4
5
[27]
⎛ I rad ⎞
−k x
⎟ =1− e
⎝ I0 ⎠
ξ radiation = 1 − ⎜
6
7
where ξradiation is the propagating flux ratio at distance x (m) from the radiant source, Irad is
8
the transmitted radiant intensity passing through a spheroid (kJ m-2 s-1), I0 is the initial radiant
9
intensity (kJ m-2 s-1), and k is the rate of attenuation through the canopy media, also termed
10
11
12
optical depth or opacity (m-1).
The rate of attenuation of radiant energy from fires in vegetative canopies was given by
Butler et al. (2004b) as:
13
14
[28]
k=
σ Wmass
4 δk ρ p
15
16
where σ is the fuel surface area-to-volume ratio (m2 m-3); Wmass is the dry mass of fuel per
17
unit planar area (kg m-2), δk is the fuel array depth (this is the same as DC in canopies; m), and
18
ρp is the oven-dry mass density of the fuel particles (kg m-3). Expressed in terms of LAI and
19
DC, Eq. 28 becomes:
20
21
[29]
k=
LAI
4 DC
30
1
2
Substituting Eq. 29 into Eq. 27, and replacing distance x with some characteristic distance for
x%
3
radiative transfer in canopies yields:
4
5
[30]
ξ radiation = 1 − e
⎛ LAI ⎞
−⎜
⎟ x%
⎝ 4 DC ⎠
6
7
The denser the canopy (higher LAI) for a given canopy depth, the greater the absorption of
8
the radiant energy by the unburned fuel elements and the greater the propagating flux ratio.
9
Conversely, the greater the canopy depth for a given LAI, the lower the density of the stand
10
11
and the lower the propagating flux ratio.
Butler et al. (2004a) found that radiant energy transfer between the flame and fuels can
12
occur over distances as great as 60 m in the upper canopy, and 20 m in the portion of canopy
13
with the highest bulk density. A characteristic distance x% of 1 meter was chosen, yielding a
14
radiation propagating flux ratio of 0.56 percent for FCCS Fuelbed 52 (Pseudotsuga
15
menzeisii/Pinus ponderosa).
16
The propagation flux ratio based on convective heating of canopy fuels, ξconvection
17
(dimensionless), is defined as the ratio of the convective heat flux (heat energy per unit time)
18
absorbed by the canopy fuels, specifically the foliage, divided by the total heat flux in the
19
corresponding air volume:
20
21
[31]
ξ convection =
C forced LAI
E
22
31
1
where Cforced is the forced-convection heat flux absorbed per unit area of foliage
2
(W m-2 foliage; by definition, 1 W = 1 J s-1), LAI is the leaf-area index (m2 foliage m-2
3
ground), and E is the total heat flux in the canopy air per unit of ground area (W m-2 ground).
4
Cforced is defined as (Monteith and Unsworth 1990, pg. 122):
5
6
[32]
C forced =
kC (Tair − T foliage ) Nu
d
7
8
where k is the thermal conductivity of air (W m-1 K-1), Tair and Tfoliage are the temperatures of
9
the heated air and foliage surface (K), respectively, Nu is the Nusselt number
10
(dimensionless), and d is a characteristic dimension of the fuel element. The Nusselt number
11
is a function of the Reynolds number. Based on a relationship reported for forced flow in
12
arrays of pine needles (Cruz et al. 2006b), the Nusselt number was defined as:
13
14
[33]
Nu ≈ 0.1417 Re0.6503
15
16
17
The total heat flux in the canopy air column, E (W m-2 ground), is calculated from
physical principles:
18
19
[34]
hair ρ air DC Γ '
E =
60
20
21
where hair is the heat content of air (J kg-1), ρair is the density of air (kg m-3), DC is the canopy
22
depth (m), Γ’ is the potential reaction velocity (min-1), and 60 converts from min-1 to s-1. The
32
1
potential reaction velocity is the reaction velocity (defined as the ratio of the reaction zone
2
efficiency to the reaction time) times the mineral and moisture damping coefficients
3
(Rothermel 1972, Burgan 1987). The potential reaction velocity may also be considered a
4
measure of the fraction of canopy fuels consumed (i.e., the fraction of fuel consumed scaled
5
by the inverse product of the reaction time and the mineral and moisture damping
6
coefficients). The greater the fraction of canopy fuels consumed, the greater the heat flux
7
within the air column.
8
9
Convection propagating flux ratios ranged from 0.1 to 0.35 for Re values of 50 to 400,
respectively, for FCCS Fuelbed 52 (Pseudotsuga menziesii/Pinus ponderosa).
10
Model evaluation
11
Sensitivity Analysis
12
The FCCS crown fire potentials were evaluated using three approaches: a sensitivity
13
analysis, a comparison of predicted crown fire rates of spread with observed data, and a
14
comparison of the FCCS crown fire potentials with results obtained from the CFIS model
15
(Alexander et al. 2006). The sensitivity analysis examined the relationship between TP
16
and AP and two important environmental variables: estimated fine fuel moisture content
17
(EFFM), and wind speed. FCCS results for a range of EFFM values (0 to 12.5%) were
18
obtained by translating the effects of EFFM on the parameters that drive the FCCS
19
surface fire potentials (described in Sandberg et al., this issue b). These effects include
20
the nonwoody and shrub moisture damping coefficients. Each of these moisture scenarios
21
resulted in different surface fire characteristics, and through the formulations presented in
22
this paper, resulted in different crown fire characteristics. In FCCS, the torching
23
potential, TP, was influenced by EFFM even though EFFM is not a normal input to the
33
1
model. The sensitivity analysis based on the FCCS conifer fuelbeds (n=86) showed that
2
TP decreased with increasing EFFM (Fig. 6), as evidenced by a shift in the frequency of
3
high TP values to lower TP values with increasing EFFM (Fig. 6). Similarly, the active
4
crowning potential, AP, decreased with increasing EFFM (Fig. 7).
5
The sensitivity of AP to variations in the wind speed was tested for 21 natural FCCS
6
conifer fuelbeds in Alaska and the Northeastern United States. The 10 m open wind
7
speeds ranged from 5 to 50 km h-1. AP increased with increasing wind speed (Fig. 8). At
8
a wind speed less than 5 km h-1, the AP values were zero for all 21 fuelbeds (i.e., no
9
crowning). At 10 km h-1, half of the fuelbeds had AP values greater than zero, and above
10
50 km h-1, more than 90% of the models were predicted to crown (Fig. 8).
11
Comparison with Field Observations
12
The FCCS-predicted crown fire rates of spread, RC, were compared against the rates
13
of spread observed in 15 black spruce stands reported by Alexander et al. (2006, Table
14
A1). These field observations were compared to the fire potential results for FCCS
15
Fuelbed 87, a black spruce-feather moss stand. Black spruce/feather moss was chosen
16
from among the 216 available FCCS fuelbeds because it yielded the closest agreement
17
with the CBD reported in Alexander et al. (2006): 0.27 kg m-3 for FCCS, and 0.20 kg m-3
18
for the observations. Input parameters to FCCS consisted of the moisture content, set to
19
the value neared to the observed EFFM, and wind speed, set to the actual observed value.
20
Linear regression of log10-transformed RC values (needed to adjust for high-leverage
21
points) produced a coefficient of determination (R2) of 0.48 (Fig. 9). This comparison
22
also indicates that the FCCS-predicted rates of spread are typically lower than observed
23
values. This is consistent with other comparative studies showing that the crown fire
34
1
models that utilize the Rothermel (1991, 1972) equations in some form tend to produce
2
relatively low predicted crown fire rates of spread (for example, Cruz et al. 2005, Scott,
3
2006). However, since FCCS is intended to provide relative crown fire potentials, an
4
additional regression was performed based on the relative ranks of the RC values. This
5
comparison showed that higher observed RC values tended to coincide with higher
6
predicted RC values (R2=0.44; EF=0.33). EF is defined as the modeling efficiency, which
7
is similar to R2 except that the latter provides an indication of the fit to a 1:1 line whereas
8
the former provides an indication of the fit to a linear regression line (Mayer and Butler
9
1993, Alexander et al. 2006). These results suggest that the predicted rates of spread by
10
FCCS track the observed values quite well, especially based on their relative ranking.
11
Comparison with CFIS Predicted Fire Behaviour
12
The FCCS crown fire potentials were then compared against the CFIS model (Cruz et al.
13
2006b,c). In FCCS, the potential for a surface fire to crown is expressed as TP, and in CFIS,
14
as the “probability of crowning.” These parameters were regressed against each other using
15
21 FCCS conifer fuelbeds in Alaska and the northeastern United States, at a wind speed of 15
16
km h-1 (10 m open wind speed) and over EFFM values ranging from 0 to 12.5% (N=105).
17
As described in the previous section, the comparison was based on the relative ranks of the
18
TP and CFIS crown fire probabilities, not the actual values. The linear regression of the
19
ranked values indicated that TP tracks well with the CFIS-based crown fire probabilities,
20
yielding an R2 value of 0.43 and an EF value of 0.35 (data not shown).
21
Finally, the FCCS predicted TP, AP, and RC values were summarized as a function of the
22
fire type predicted by the CFIS model. The CFIS fire types consist of: surface fire, torching
23
fire, or active crown fire (Cruz 2006b,c; Alexander et al. 2006). All of the FCCS conifer
35
1
models were used to test for differences in TP (EFFM ranging from 0 to 12.5%, wind speed
2
fixed at 15 km h-1), and a subset of 21 conifer stands (see earlier) was used to test for
3
differences in AP and RC (EFFM fixed at 9.4%, wind speed 5 to 60 km h-1). The underlying
4
question of this test was: if one considers CFIS as an independent determination of the
5
potential fire type (a discrete classification), then how well do the FCCS crown fire potentials
6
(continuous variables) reflect these distinct fire types? The results based on this subset of
7
FCCS fuelbeds indicate that the potential CFIS fire types have statistically significant
8
distributions of the FCCS crown fire potentials, and that these differences are consistent with
9
the expected trends (Fig. 10). Surface fires are characterized by low TP, AP, and RC values.
10
Torching fires are characterized by high, intermediate, and low values for TP, AP, and RC,
11
respectively, reflecting the limited ability of the canopy to carry an active crown fire (Fig.
12
10). Finally, active crown fires were characterized by high TP, AP, and RC values (Fig. 10).
13
Differences between these fire types in terms of the FCCS crown fire potentials (Fig.10)
14
were statistically significant based on Analysis of Variance (using log10-transformed values
15
to obtain an approximately normal data distribution with a Bonferroni post-hoc test): for TP –
16
F2,427=78.6, P < 0.001, Surface < both Torching and Active); for AP – F2,249=63.8, P < 0.001,
17
Surface < Torching < Active); for RC – F2,249=115.5, P < 0.001, both Surface and Torching <
18
Active).
19
20
21
Summary
Anticipating the occurrence of crown fires in natural fuelbeds is critical for effective fire
22
management and planning. This paper presents a conceptual framework for ranking the
23
crown fire potential of different wildland fuelbeds with vegetative canopies. We have
36
1
extended the work by Van Wagner (1977) and Rothermel (1991) and introduced several new
2
physical concepts to the modeling of crown fire behaviour derived from the reformulated
3
Rothermel (1972) surface fire behaviour equations described in Sandberg et al. (this issue b).
4
The FCCS framework for computing the crown fire potential of wildland fuelbeds
5
systematically and objectively considers the many variables that are commonly used by fire
6
behaviour analysts to assess crown fire potential. Some key parameters, such as canopy
7
continuity and ladder fuel abundance, are ignored in most other fire management systems.
8
By advancing a transparent quantitative framework, even one containing heuristic
9
knowledge-based algorithms, we hope to provide a robust framework to evaluate crown fire
10
11
behaviour against future models and field observations.
Evaluation of the FCCS crown fire potentials shows that TP and AP respond in a
12
consistent manner with changes in key environmental parameters such as fine fuel moisture
13
content and wind speed, and that the relative ranking of the predicted crown fire spread rates,
14
RC, track quite well with observations. Moreover, the predicted FCCS crown fire potentials
15
accurately reflect the differences in fire type that were predicted by an independent crown
16
fire model, CFIS (Alexander et al. 2006). These results are encouraging, and suggest that the
17
FCCS crown fire potentials can serve as a useful tool for fire managers to consider when
18
evaluating the potential for crown fires, or the relative behavior of active crown fires, in
19
forest canopies.
20
21
References
22
Alexander, M.E. 1988. Help with making crown fire hazard assessments. In Protecting
23
people and homes from wildfire in the Interior West: proceedings of the symposium and
37
1
workshop; 1987 October 6-8; Missoula, MT. Compiled by W.C. Fischer and S.F. Arno.
2
USDA For. Serv. Gen. Tech. Rep. INT-251. pp. 147-156.
3
4
5
6
7
Alexander, M. E. 1998. Crown fire thresholds in exotic pine plantations of Australasia.
Ph.D. Thesis. Australian National University, Canberra, Australia: 228.
Alexander, M.E., and Cruz, M.G. 2006. Evaluating a model for predicting active crown fire
rate of spread using wildfire observations. Can J. For. Res. 36: 3015-3028.
Alexander, M.E., Cruz, M.G., and Lopes, A.M.G. 2006. CFIS: A software tool for
8
simulating crown fire initiation and spread. In: Viegas, D.X. (Ed.), Proceedings of V
9
International Conference on Forest Fire Research. Millpress Sci. Publ., Rotterdam,
10
11
12
13
14
15
16
17
Netherlands. CD-ROM.
Andrews, P.L. 1986. BEHAVE: Fire behavior prediction and fuelbeding system--BURN
subsystem, Part 1. USDA For. Serv. Gen. Tech. Rep. INT-194.
Andrews, P.L., and Rothermel, R.C. 1982. Charts for interpreting wildland fire behavior
characteristics. USDA For. Serv. Gen. Tech. Rep. INT-194.
Burgen, R.E. 1987. Concepts and interpreted examples in advanced fuelbeding. USDA For.
Serv. Gen. Tech. Rep. INT-238; Intermountain Research Station, Missoula, MT: 40.
Butler, B.W., Cohen, J., Latham, D.J., Schuette, R.D., Sopko, P., Shannon, K.S., Jimenez, D.,
18
and Bradshaw, L.S. 2004a. Measurements of radiant emissive power and temperatures in
19
crown fires. Can. J. For. Res. 34: 1577-1587.
20
21
22
23
Butler, B.W., Finney, M.A., Andrews, P.L., and Albini, F.A. 2004b. A radiation-driven
model for crown fire spread. Can. J. For. Res. 34: 1588-1599.
Byram, G.M. 1959. Combustion of forest fuels. In: ‘Forest fire: control and use.’ (Ed. K.P.
Davis). pp. 61-89. McGraw-Hill, New York.
38
1
Clark, T.L., Radke, L, Coen, J., and Middleton, D. 1999. Analysis of small-scale convective
2
dynamics in a crown fire using infrared video camera imagery. J. Appl. Met., 38: 1401-
3
1420 (October 1999)
4
5
6
Cruz, M.G. 1999. Modeling the initiation and spread of crown fires. M.S. thesis. Univ.
Montana, Missoula.
Cruz, M. G., Alexander, M. E., and Wakimoto, R. H. 2003a. Definition of a fire behavior
7
model evaluation protocol: A case study application to crown fire behavior models. In:
8
USDA Forest Service, Fire, Fuel Treatments, and Ecological Restoration: Conference
9
proceedings. USDA Forest Service Proceedings RMRS-P-29. pp. 49-67.
10
11
12
13
14
Cruz, M.G., Alexander, M.E., and Wakimoto, R.H. 2003b. Assessing the probability of
crown fire initiation based on fire danger indices. For. Chron. 79, 976-983.
Cruz, M.G., Alexander, M.E., and Wakimoto. 2004. Modeling the likelihood of crown fire
occurrence in conifer forest stands. For. Sci. 50, 640-658.
Cruz, M.G., Alexander, M.E., and Wakimoto, R.H. 2005. Development and testing of models
15
for predicting crown fire rate of spread in conifer forest stands. Can. J. For. Res. 35,
16
1626-1639.
17
Cruz, M.G., Butler, B.W., Alexander, M.E, and D.X. Viegas. 2006a. Development and
18
evaluation of a semi-physical crown fire initiation model. In: Viegas, D.X. (Ed.),
19
Proceedings of V International Conference on Forest Fire Research. Millpress Sci. Publ.,
20
Rotterdam, Netherlands. CD-ROM.
21
Cruz, M.G., Butler, B.W., Alexander, M.E, Forthofer, J.M., and Wakimoto, R.H. 2006b.
22
Predicting the ignition of crown fuels above a spreading surface fire. Part I: model
23
idealization. Int. J. Wildl. Fire. 15: 47-60.
39
1
Cruz, M.G., Butler, B.W., and Alexander, M.E. 2006c. Predicting the ignition of crown fuels
2
above a spreading surface fire. Part II: model evaluation. Int. J. Wildl. Fire. 15: 61-72.
3
4
5
6
7
8
9
Fahnestock, G.R. 1970. Two keys for appraising forest fuels. USDA For. Serv. Res. Pap.
PNW-99.
Finney, M.A. 1998. FARSITE: Fire Area Simulator - model development and evaluation.
USDA For. Serv. Res. Paper RMRS-RP-4.
Finney, M.A. 2004. FARSITE: Fire Area Simulator - model development and evaluation.
USDA For. Serv. Res. Paper RMRS-RP-4. [Revised]
Forestry Canada Fire Danger Group. 1992. Development and structure of the Canadian
10
Forest Fire Behavior Prediction System. Forestry Canada Fire Danger Group. Inf. Rep.
11
ST-X-3: 63.
12
13
Kilgore, B.M., and Sando, R.W. 1975. Crown-fire potential in a sequoia forest after
prescribed burning. For. Sci. 21: 83-87.
14
Mayer, D.G. and D.G. Butler. 1993. Statistical validation. Ecol. Model. 68: 21-23.
15
Mercer, G.N. and R.O. Weber. 1994. Plumes above line fires in a cross wind. Int. J. Wild.
16
Fire. 4:201-207.
17
Mercer, G.N., and R.O. Weber. 2001. Fire plumes. Pp. 225-255 In Forest fires: behavior and
18
ecological effects, Johnson, E.A. and K.Miyanishi (eds.) Academy Press, San Diego.
19
20
Monteith, J.L., and M.H. Unsworth. 1990. Principals of environmental physics. 2nd ed.
Edward Arnold, London, U.K.
21
Ottmar, R.D., Sandberg, D.V., Riccardi, C.L., and Prichard, S.J. The Fuel Characteristic
22
Classification System (FCCS) – a system to build, characterize, and classify fuels for
23
resource planning. Can. J. For. Res. This issue.
40
1
Peterson, D.L., Johnson, M.C., Agee, J.K., Jain, T.B., McKenzie, D., and E.D Reinhardt.
2
2005. Forest structure and fire hazard in dry forests of the western United States. USDA
3
Forest Service Gen. Tech. Rep. PNW-GTR-628.
4
5
6
7
8
9
10
Rothermel, R.C. 1972. A mathematical model for predicting fire spread in wildland fuels.
USDA For. Serv. Res. Paper INT-115.
Rothermel, R.C. 1991. Predicting behavior and size of crown fires in the Northern Rocky
Mountains. USDA For. Serv. Res. Pap. INT-438.
Sandberg, D.V., Ottmar, R.D., and Cushon, G.H. 2001. Characterizing fuels in the 21st
century. Int. J. Wildl. Fire 10: 381-387.
Sandberg, D.V., Riccardi, C.L., and Schaaf, M.D. a. Fire potential rating for wildland
11
fuelbeds using the Fuel Characteristic Classification System. Can. J. For. Res. this issue.
12
Sandberg, D.V., Riccardi, C.L., and Schaaf, M.D. b. Reformulation of Rothermel's wildland
13
fire behavior model for heterogeneous fuelbeds. Can. J. For. Res. This issue.
14
Scott, J.H. 1999. A system for assessing crown fire hazard. Fire Manage. Notes. 59: 20-24.
15
Scott, J.H. 2006. Comparison of crown fire modeling systems used in three fire management
16
17
18
19
20
applications. USDA For. Serv. Res. Pap. RMRS-RP-58.
Scott, J.H., and Reinhardt, E.D. 2001. Assessing crown fire potential by linking models of
surface and crown fire behavior. USDA For. Serv. Res. Pap. RMRS-RP-29.
Stocks, B.J., Alexander, M.E., and Lanoville, R.A. 2004a. Overview of the International
Crown Fire Modelling Experiment (ICFME). Can. J. For. Res. 34: 1543-1547.
21
Stocks, B.J., Alexander, M.E., Wotton, B.M., Stefner, C.N., Flannigan, M.D., Taylor, S.W.,
22
Lavoie, N., Mason, J.A., Hartley, G.R., Maffey, M.E., Dalrymple, G.N., Blake, T.W.,
41
1
Cruz, M.G., and Lanoville, R.A. 2004b. Crown fire behaviour in a northern jack pine-
2
black spruce forest. Can. J. For. Res. 34: 1548-1560.
3
4
5
Stratton, R.D. 2004. Assessing the effectiveness of landscape fuel treatments on fire growth
and behavior. J. For. 102: 32-40.
Taylor, S.W. 1961. Fire under the influence of natural convection. Proceedings, Use of
6
Models in Fire Research, Nat. Acad. Sci. Nat. Res. Counc., Washington, D.C.: 10-31.
7
Thomas, P.H. 1963. Size of flames from natural fires. In Ninth (International) symposium on
8
9
combustion. The Combustion Institute, Pittsburgh, PA. pp. 844-859.
Thomas, P.H. 1964. The effect of wind on plumes from a line heat source. Dep. Sci. Indust.
10
Res. and Fire Off. Comm., Joint Fir. Res. Org., Fire Res. Stn., England. Fire Res. Note
11
No. 576: 36 pp.
12
13
14
15
16
17
18
Van Wagner, C.E. 1964. History of a small crown fire. Forestry Chronicle 40: 202-205,
208-209.
Van Wagner C.E. 1968. Fire behaviour mechanisms in a red pine plantation: Field and
laboratory evidence. Can. Dep. For. Rural Develop. For. Branch Dep. Publ. 1229.
Van Wagner, C.E. 1977. Conditions for the start and spread of crown fire. Can. J. For. Res.
7: 23-34.
Van Wagner, C.E. 1989. Prediction of crown fire in conifer stands. In Proceedings of the
19
10th Conference on Fire and Forest Meteorology. Edited by D.C. MacIver, H. Auld, and
20
R. Whitewood. Forestry Canada and Environment Canada, Ottawa, Ont. pp. 207-212.
21
Van Wagner, C.E. 1993. Prediction of crown fire behavior in two stands of jack pine. Can. J.
22
For. Res. 23: 442-449.
42
1
2
3
Xanthopoulos, G. 1990. Development of a wildland crown fire initiation model. Ph.D. thesis.
Univ. Montana, Missoula.
Xanthopolous, G. and Wakimoto, R.H. 1993. A time to ignition-temperature-moisture
4
relationship for branches of three western conifers. Can. J. For. Res. 23:253-258.
5
Yih, C.S. 1953. Free convection due to boundary sources. In: Proceedings of the first
6
symposium on the use of models in geophysics. Washington, DC: U.S. Government
7
Printing Office: 117-133.
8
43
1
Appendix 1 - List of Symbols
2
cAP
Scaling function to limit AP within the range of zero to 10
3
cTP
Scaling function to limit TP within the range of zero to 10
4
d
Characteristic spatial dimension of a fuel element (m)
5
gap
Physical separation between the top of the surface fuel layer and the bottom of
6
the canopy fuel layer (m)
7
h
Fuel low-heat content (kJ kg-1)
8
hair
Heat content of air (J kg-1)
9
k
Opacity or optical depth (m-1)
10
kC
Thermal conductivity of air (W m-1 K-1)
11
m
Reference foliar moisture content (dimensionless)
12
tR
Residence time of flaming fire front (minutes)
13
AP
FCCS Active Crown Fire Potential (dimensionless, scale 0 to 10)
14
CBD
Canopy bulk density (kg m-3)
15
CBDcritical
Critical canopy bulk density required for active crown fire spread (kg m-3)
16
CBH
Canopy base height (m)
17
Cforced
Forced-convection heat flux absorbed by foliage (W m-2)
18
CFP
FCCS Crown Fire Potential (dimensionless, scale 0 to 10)
19
COV
Total cover of tree crowns (% of ground area covered by tree crowns)
20
DC
Mean canopy depth, or mean difference between stand height and canopy base
21
(m)
22
E
Total heat flux in the canopy air space per unit of ground area (W m-2 ground)
23
FC
FCCS crown-to-crown flame transmissivity term (dimensionless)
44
1
FAI
Fuel-area index of surface fuelbed categories (m2 m-2)
2
FCCS
Fuel Characteristic Classification System
3
FMC
Foliar moisture content of canopy fuels (percent)
4
H
Fuel heat yield (kJ kg-1)
5
H’
Fuel heat yield per unit area (kJ m-2)
6
IB
Byram’s fireline intensity (kJ m-1 s-1 or kW m-1)
7
I’
Critical fireline intensity defined by Van Wagner (1977)(kJ m-1 s-1 or kW m-1)
8
IC
FCCS Crown fire initiation term (units)
9
IR
Rothermel’s (1972) reaction intensity (kJ m-2 min-1)
10
Irad
Transmitted radiant intensity passing through a spheroid (kJ m-2 s-1)
11
IR.FCCS
FCCS combined reaction intensity of surface and crown fuels (kJ m-2 min-1)
12
IR.FCCS.S
FCCS reaction intensity of surface fuels (kJ m-2 min-1)
13
IR.FCCS.C
FCCS Reaction intensity of canopy fuels (kJ m-2 min-1)
14
I0
Initial radiant intensity (kJ m-2 s-1)
15
ladder
FCCS Heuristically assigned value representing abundance of canopy ladder
16
fuels (default=1, representing no ladder fuel).
17
LAI
Leaf-area index (m2 m-2)
18
Nu
Nusselt number (dimensionless)
19
Qig
Heat of pre-ignition (J kg-1)
20
R
Fire rate of surface fire spread (m min-1)
21
RAC
Minimum spread rate required to sustain an active crown fire (m min-1)
22
Ractive
Active crowning rate of spread (m min-1)
23
RC
FCCS crown fire spread-rate term (m min-1)
45
1
Re
Reynolds number (dimensionlesss)
2
RFCCS.S
FCCS surface fire rate of spread (m min-1)
3
(R10)40% WRF
Surface rate of fire spread (m min-1) predicted by Rothermel (1972) using fuel
4
characteristics for Fire Behaviour Prediction System model 10 and a midflame
5
wind speed set at 40 percent of the 6.1-meter wind speed.
6
Tair
Temperature of the heated air (K)
7
TP
FCCS Torching Potential (dimensionless, scale 0 to 10)
8
TCOV
FCCS threshold % canopy cover required for active crown fire
9
(dimensionless)
10
Tfoliage
Temperature of the foliage surface (K)
11
TLAI
FCCS threshold LAI required for efficient crown-to-crown flame transfer (m2
m-2)
12
Horizontal midflame wind speed (m min-1)
13
U
14
UB
FCCS benchmark horizontal midflame wind speed (m min-1, set to 107)
15
U6.1
Wind speed measured in the open at 6.1-meter height above ground (m min-1)
16
W
FCCS vertical, convectively driven wind speed (m min-1; assumed 268)
17
WAF
FCCS canopy wind speed adjustment factor (dimensionless)
18
Wf
Weight of fuel consumed in the flaming fire front (kg m-2)
19
Wmass
Dry mass of fuel per unit planar area (kg m-2)
20
γR
Reaction volume of fuels involved in the reaction zone (m3 m-2)
21
δ
Depth of the woody fuelbed stratum (m)
22
δk
Fuel array depth used in calculating radiant propagating flux term (same as DC
23
B
in canopies; m)
46
1
δopC
FCCS optimum canopy depth (m)
2
ζI
FCCS ignition thickness heated on surfaces of thermally thick fuel elements
3
4
(m)
ηK
5
Mineral damping coefficient (dimensionless, assumed to be 0.42 based on 1percent silica-free ash)
6
ηM.FCCS
FCCS moisture damping coefficient (dimensionless)
7
η∆’.llm
FCCS absorption efficiency of the litter-lichen-moss stratum (dimensionless)
8
ηΔ’.C
FCCS canopy reaction efficiency (dimensionless)
9
ηΔ’.S
FCCS surface-fuel reaction efficiency (dimensionless)
10
θ
Planform area (m2 m-2, non-woody, shrub, and litter-lichen-moss)
11
ξ
Propagating flux ratio for surface fuelbed (dimensionless)
12
ξconvection
FCCS propagating flux ratio for convective heat transfer in the canopy
(dimensionless)
13
14
ξradiation
FCCS propagating flux ratio for radiant heat transfer in the canopy
(dimensionless)
15
16
ξC
FCCS combined propagating flux ratio in the canopy (dimensionless)
17
ρair
Density of air (kg m-3)
18
ρb
Bulk density of fuel elements (kg m-3)
19
ρp
Oven-dry mass density of the fuel particles (kg m-3; assumed 400 kg m-3)
20
σ
Surface-area-to-volume ratio of fuel elements (m2 m-3)
21
1+φW.FCCS
Acceleration factor for wind (dimensionless)
22
Δ C'
FCCS relative canopy fuelbed depth; defined as the optimum canopy
23
depth, δopC, divided by the mean canopy depth, DC (dimensionless)
47
1
Γ'
Potential reaction velocity (min-1)
2
Γ'max.FCCS
FCCS maximum reaction velocity at the optimum packing ratio (min-1)
48
1
Fig. 1(a). Reduction in crown fire initiation with increasing gap over ladder distance for
2
FCCS Fuelbed 52 (Pseudotsuga menziesii-Pinus ponderosa)
3
49
1
Fig. 1(b). Reduction in crown fire initiation potential with increasing canopy reaction
2
efficiency for FCCS Fuelbed 52 (Pseudotsuga menziesii-Pinus ponderosa). A higher
3
reaction efficiency translates to a higher reaction velocity and a shorter residence time,
4
hence less time for heating of canopy fuels.
5
50
1
Fig. 1(c). Reduction in crown fire initiation with increasing foliar moisture content
2
(percent) for FCCS Fuelbed 52 (Pseudotsuga menziesii-Pinus ponderosa)
3
51
1
Fig. 2. Effect of canopy cover and WAF on crown-to-crown transmission of fire (Eq. 9).
2
Lines A, B, C, and D represent WAF values of 0.75, 1.0, 1.5, and 3.0, respectively.
3
Values of WAF above one decrease the threshold canopy cover (default 40 percent)
4
required for crown-to-crown transmission, and values of WAF below one increase the
5
threshold canopy cover required for crown-to-crown transmission. At any single canopy
6
cover, higher WAF values yield higher rates of crown-to-crown transmission. The
7
greatest effect of wind speed occurs at low canopy covers.
8
52
1
Fig.3. Effect of midflame wind speed on WAF from Eq. 17. WAF is the multiplier used
2
to adjust the canopy cover (percent) when computing the crown-to-crown flame
3
transmission term (FC).
4
53
1
Fig. 4. Effect of increasing relative canopy depth (a measure of foliage density) on the
2
canopy reaction efficiency, η∆'C, from Eq. 23. This curve reflects the relatively weak
3
influence of less-than-optimum foliage density (relative canopy depths greater than one)
4
on canopy flammability.
5
54
1
Fig. 5. Effect of canopy foliar moisture content (percent) on the moisture damping
2
coefficient and canopy flammability (Eq. 25). The shape of this curve approximates the
3
influence of foliar moisture content on canopy flammability in Van Wagner (1977).
4
55
1
Fig. 6. Distribution of Torching Potential (TP) as a function of estimated fine fuel
2
moisture (EFFM) based on 86 FCCS conifer fuelbeds. Classes shown are zero (black), 0
3
to 1 (dark gray), 1 to 6 (open), and 6 to 10 (light gray), respectively.
4
56
1
Fig. 7. Distribution of Active Crowning Potential (AP) as a function of estimated fine
2
fuel moisture (EFFM) based on 86 FCCS conifer fuelbeds. Classes shown are zero
3
(black), 0 to 1 (dark gray), 1 to 6 (open), and 6 to 10 (light gray), respectively.
4
57
1
Fig. 8. Distribution of Active Crowning Potential (AP) as a function of 10 m open wind
2
speed based on 21 FCCS conifer fuelbeds in Alaska and the northeastern United States.
3
Classes shown are zero (black), 0 to 1 (dark gray), 1 to 6 (open), and 6 to 10 (light gray),
4
respectively.
5
58
1
Fig. 9. Predicted versus observed crown fire rate of spread, log10-tranformed (log10
2
[RC]). Observations are based on 15 crown fires in black spruce stands reported by
3
Alexander et al. (2006). Predictions were obtained based on FCCS Fuelbed 87, a black
4
spruce/feather moss stand, using the reported EFFM and wind speed values. (R2 = 0.48)
5
59
1
Fig. 10. FCCS crown fire potentials—TP and AP—and crown fire rate of spread, RC, by
2
fire type. The TP comparison is based on 86 FCCS conifer fuelbeds. The AP and RC
3
comparison is based on 21 FCCS conifer fuelbeds in Alaska and the northeastern United
4
States. Variable input parameters were EFFM for TP comparisons (N=430) and wind
5
speed for AP and RC comparisons (N=252). Fire type was calculated from CFIS (Cruz et
6
al 2006b,c; Alexander et al. 2006) using the FCCS fuelbed characteristics and
7
environmental inputs, EFFM, and wind speed. Fire types consisted of surface (black
8
bar), torching (gray bar), and active crown fire (open bar). Differences between fire type
9
were statistically different for each of the potentials (P<0.001; zero values were set to a
10
value of 10-2 or log10=-2). Error bars represent the 95 percent confidence interval.
11
60