D2.4-2 Simulations of fire impact on tree foliage: 2D results

fingersfieldMécanique

22 févr. 2014 (il y a 3 années et 1 mois)

141 vue(s)





Project no. FP6
-
018505

Project Acronym FIRE PARADOX

Project Title FIRE PARADOX:

An Innovative Approach of Integrated Wildland Fire
Management Regulating the Wildfire Problem by the Wise Use of Fire: Solving the Fire
Paradox



Instrument Integrated Pr
oject (IP)

Thematic Priority Sustainable development, global change and ecosystems



D2.4
-
2
Simulations of fire impact on tree foliage: 2D results



Due date of deliverable: Month 18

Actual submission date: Month 20



Start date of project: 1
st

March 2006



Duration: 48months



Organisation name of lead contractor for this deliverable: INRA


Revision (1000)


Project co
-
funded by the European Commission within the Sixth Framework Programme
(2002
-
2006)

Dissemination Level

PU

Public


PP

Restricted to other

programme participants (including the Commission Services)


RE

Restricted to a group specified by the consortium (including the Commission Services)

X

CO

Confidential, only for members of the consortium (including the Commission Services)




D2.4
-
2
-
18
-
1000
-
1





Page
2

of 2
6


Table of co
ntents:


1. Introduction

................................
................................
................................
......

3

2. Background

................................
................................
................................
.......

3

3. Methods

................................
................................
................................
............

4

4. No canopy simulations (no wind)

................................
................................
......

11

5. Influence of tree canopy (no wind)

................................
................................
...

13

6 Simulations with a wind

................................
................................
....................

18

7 Needs for reference data

................................
................................
..................

23

8. Conclusions

................................
................................
................................
.....

24

References

................................
................................
................................
..........

25



D2.4
-
2
-
18
-
1000
-
1





Page
3

of 2
6


Simulations of fire impact on tree foliage: 2D results.

J
-
L Dupuy and V. Konovalov


1
.

Introduction

The aim of the WP2.4 workpackage (activity 3) is to study numerically crown scorch and
crown f
ire ignition as the effects of a fire line spreading through surface fuel under a tree
canopy. Here we report a preliminary study performed with the FIRESTAR 2D model. The
objective was to assess the usual assumptions made when one uses the Van Wagner crit
eria,
based on plume theory, to estimate crown scorch or crown ignition. The Van Wagner criteria
indeed are simple predictive models for crown scorch height or crown fire initiation
occurrence. For this purpose we simulated the fire line by heat source put

on the ground and
mainly investigated the temperature field. As a first step we tested the sensitivity of the
predictions to the mode of heat input and to the selected parameters of the
k
-


turbulence
model used in FIRESTAR. As a second step we ran comput
ations of thermal plumes with no
-
wind and with no canopy, for first comparison to plume theory. The influence of crown
existence to the temperature field above the heat source and, so way, to the crown scorch
and fire ignition conditions, was then investi
gated. In the last part we show and discuss some
results obtained in the presence of a wind.


2
.

Background

Following Thomas (1963,

1964), Van Wagner (1973) derived from the plume theory a
formula relating the crown scorch height with the linear fire front

intensity



2 3
11.61
60
s
a
I
h
t

(1)

where
h
s

(m) is the crown scorch height,
I

(kW/m) is the linear fire front intensity,
t
a

(°C)

is
the ambient temperature. The numerical coefficient was derived from experimental data.
Crown scorch was considered to a
ppear at 60°C, as it is usual. Recently, Michaletz and
Johnson (2006) showed that this threshold of temperature is actually well adapted for the
prediction of tree foliage necrosis, but that higher thresholds are to be used for vegetative
buds. This is due

to the lower surface
-
to
-
volume ratio of vegetative buds as compared with
the one of needles, which leads to higher response time to a heat flux. Here we will consider
that the fuel elements are in thermal equilibrium with the gaseous phase, which thus is
well
supported for foliage (except during water evaporation process).

A basic assumption of the plume theory is to consider points far enough from the heat
source. It also means that the main mechanism of heat transfer is convection, since heat
-
conduction
or radiation should only be significant close to the source. Van Wagner criterion
also assumes that the plume structure is not significantly affected by the presence of a
canopy.


D2.4
-
2
-
18
-
1000
-
1





Page
4

of 2
6


Van Wagner (1973) also investigated the effect of weak wind on crown scorch.
A weak wind
means that plume structure is not destroyed with wind, but just inclined (Thomas 1964). A
simple correction (see appendix) gives the formula including the effect of wind





7 6
3 1 2
3.94
0.107 60
s
a
I
h
I U t

 

(2)

where
u

(m/s) is the wind velocity.

Finally Va
n Wagner (1977) derived a formula for the determination of the minimum surface
fire intensity that could start a crown fire through ignition of the lowest crown fuel elements
located at height
h
above the ground:

I
0
= (C
h

Q
)
3/2

(3)


where
Q

is the heat of

ignition of the fuel which depends on the ignition temperature
(usually 600 K) and the fuel moisture content.

Again the above formula was derived from the plume theory following the work of Thomas
(1963).

Other models have been proposed more recently for
crown fire initiation, which are also
based on plume theory (Cruz 2006). These models were derived from a more sophisticated
formulation of the plume theory than the Van Wagner criteria (Mercer and Weber 1994), the
so
-
called integral models of plume rise,
and they could be applied to the prediction of crown
scorch as well. They predict the trajectory of a plume with wind and different characteristics
like temperature and velocity on plume axis and plume lateral extent (plume width). They
require solving a s
ystem of ordinary differential equations. But basically they may suffer or
not from the same limitations as plume theory or Van Wagner criteria.


3.

Methods

To investigate crown scorch and fire ignition, the FIRESTAR 2D model was used. FIRESTAR is
a physic
ally
-
based model for wildfire behaviour and as such it allows to simulate the
propagation of a fire through a vegetation layer in the presence of ambient wind or not
(Morvan and Dupuy 2004). An up
-
dated description of FIRESTAR 2D is available in
deliverabl
e D2.4
-
1 entitled «

Simulations of wildfire behavior: 2D results

»
.

For the present
study we rather implemented a heat source, which was a rectangle area located in the
middle of the domain and put on the ground. The heat source in fact represents a steady

fire
line of given intensity (power in kW/m). As illustrations we show the fields of temperature
and of turbulence kinetic energy (Fig. 1) above the heat source as computed by FIRESTAR,
there was no tree canopy in this simulation of a thermal plume.


D2.4
-
2
-
18
-
1000
-
1





Page
5

of 2
6



Fi
g. 1

Gas temperature field (left) and turbule
nt kinetic energy field (right).


The mesh size in FIRESTAR (Cartesian mesh) may be refined in the area of interest, namely
the heat source and the plume here. The size of mesh was
0.2 x 0.1 m

at burner location
.
The numerical domain was 100 m long and 50 m high.
With no wind, we used Neuman
boundary conditions for gas velocity at all boundaries. For lateral boundaries, we used
Dirichlet conditions for all other variables including the pressure. That means we sup
pose
atmospheric pressure far from the heat source. For the upper boundary, we used Neuman
boundary conditions for all variables except for pressure.

3
.1 Selection of heat source mode

We tested three different modes for the simulation of the heat source. I
n the two first ways,
we injected a mass of CO in the heat source area, which then reacted with oxygen and
released heat. Two models of combustion rate were tested for the simulated burner: the first
one was based on the Eddy Dissipation Concept (EDC, Magn
ussen and Hjartager 1976); the
second one assumed that either CO or oxygen was fully consumed by combustion in a time
step and so the rate of combustion was limited by the amount of CO or oxygen according to
the stochiometric requirement (‘Mixed is burned’

model, MIB). The injection of CO was
adjusted to get the desired burner power (i.e. fireline intensity). In the third way, we directly
injected heat through an artificial heat source term in the equation for energy conservation of
the gaseous phase. Figur
e 2 shows the vertical temperature profiles we got using the
different heat source modes (power 400 kW/m, burner 2.4 m x 0.6 m). In all cases a
constant profile was achieved quickly. Figure 3 shows comparisons of these profiles after 5 s
and 15 s respectiv
ely. Although some differences appear at the first time, they all converge
towards the same profile for height above 1 m after 15 s.




D2.4
-
2
-
18
-
1000
-
1





Page
6

of 2
6







Fig. 2

Vertical profiles of gas temperature obtained with the three mo
des of heat source
(power 400kW
/m)
.


D2.4
-
2
-
18
-
1000
-
1





Page
7

of 2
6



Fig
. 3

Comparison of the three modes of heat source after 5 s (left) and 15s (right)
.

When the burner was big enough, the temperature inside of the burner was not high enough
to create significant loss of energy by radiation. Such an energy loss would have ch
anged the
amount of energy available for convection motion. To avoid any possibility of energy loss due
to radiation, we used a direct heat input, since the difference between vertical temperature
profiles were not significant with the different heat sourc
e modes beyond some small
elevation (a few meters) above the ground.

The heat source size can influence the temperature field above the burner and has some
value for the condition of crown fire ignition. We show the vertical profiles of temperature
(Fig. 4
) and vertical velocity (Fig. 5) above a heat source of different sizes, for two different
powers (100 kW/m and 400 kW/m). When the height of the heat source was changed, the
temperature profile was only affected within the vertical extent of the burner. T
his had no
consequence in the thermal plume above. Some little changes may appear in the velocity
profile (Fig. 5). When the horizontal size of the burner was changed, some difference
appeared above the heat source too, but is not significant beyond a few
meters.



D2.4
-
2
-
18
-
1000
-
1





Page
8

of 2
6



In reality the fireline intensity (i.e. heat source power) and the dimensions of the pyrolysis
zone (i.e. heat source size) are usually not independent variables. In the following the size of
the heat source was relate
d to the intensity according to the following formula

2
3
2000
H
I


(4)

where
H

is the heat source height and the heat source width was taken equal to 2
H
.

We derived this formula from empirical knowledge of fire front characteristics of surface f
ires.


Fig. 4

Vertical profiles of gas temperature obtained with three different sizes of heat source.
Powers (left) 100 kW/m, (right) 400 kW/m. Here the colour bars are graphical images
of the heaters of different sizes.



D2.4
-
2
-
18
-
1000
-
1





Page
9

of 2
6



Fig. 5

Vertical profiles of
gas vertical velocity obtained with three different sizes of heat
source. Powers (left) 100 kW/m, (right) 400 kW/m.


3
.2 Selection of turbulence model parameters

FIRESTAR turbulence model is a
k
-


model (
k
sets for turbulent kinetic energy and


for the
di
ssipation rate of this energy). Different
k
-


models are implemented in FIRESTAR (standard,
RNG, low Reynolds). Van Maele and Merci (2006) considered different more or less
sophisticated variants of the standard
k
-


model and test them against thermal plum
e data.
Sophistications concerned in particular the production of turbulence by buoyancy (simple
gradient diffusion approximation following Markatos
et al.

1982 or generalized gradient
diffusion following Daly and Harlow 1970) and the realizability constra
int (leading to the so
-
called realizable model, Shih
et al
. 1995). It appeared that no version of the
k
-


model gave
fully satisfactory results with respect to experimental data on thermal plumes, although
refinements may improve some predictions. Due to

we are in a preliminary 2D study of the
effect of some parameters on crown scorch or crown fire initiation, we selected the standard
model, but with two sets of parameters: the standard one and the Nam and Bill parameters
(1993). In fact Nam and Bill (199
3) adjusted the parameters of the standard
k
-


model so
that predictions verify measurements in thermal plumes. They mainly changed the value of
the C


parameter from 0.09 to 0.18, which is the multiplicative constant for the turbulent
diffusion coefficien
t expression. This has the effect of doubling the turbulence diffusion
coefficient for a given set of values for
k
and


.



D2.4
-
2
-
18
-
1000
-
1





Page
10

of 2
6







D2.4
-
2
-
18
-
1000
-
1





Page
11

of 2
6



Fig. 6

Vertical profiles of gas temperature rise as predicted by the FIRESTAR 2D model with
two sets of turbulence parameter
s (standard, Nam and Bill) and by the plume theory
(powers 100, 400 and 1600 kW/m).

The predictions of FIRESTAR 2D obtained with the standard and the Nam and Bill sets of
parameters were compared with the plume theory predictions. The plume theory predicti
ons
were computed with the Van Wagner numerical constant in (1). These comparisons are
shown in Fig. 6 for three different powers in logarithmic scales.

The predictions of FIRESTAR 2D with the two set of turbulence parameters were close
together and beyond

some threshold value of height, they showed the same trend as plume
theory. The threshold of height increased with power (about 3, 10 and 20 m respectively for
100, 400 and 1600 kW/m with the Nam and Bill set of parameters). This observation
illustrates t
he fact that the plume theory should only apply ‘far enough’ from the heat source.

The set of standard parameters gave results closer to the plume theory for low powers, the
set of Nam and Bill parameters gave results closer to the plume theory for high po
wers.

For the rest of the study, we selected the standard set of parameters.


4
.

No canopy simulations (no wind)

First we consider the case without any crown and compare the results of FIRESTAR
simulations (‘virtual crown’) with predictions of Van Wagner f
ormula for crown scorch
(Fig.

7). In both models crown scorch was considered to occur when gas temperature
reaches 60°C. The ambient air temperature here is 300

K or 27°C. Such ambient temperature
is used for all simulation here and below. One can see the
excellent agreement in crown
scorch predictions. This agreement may be explained through inspection of Fig. 6. As the

D2.4
-
2
-
18
-
1000
-
1





Page
12

of 2
6


power of the heat source is increased, we expect that crown scorch height increases.
According to Fig. 6, there is some threshold of power

beyond which the FIRESTAR
predictions of vertical temperature profile are very close to the predictions from Van Wagner
formula. The fact that crown scorch heights were in agreement means that the
corresponding intensities were beyond the above threshold.



Fig. 7

Crown scorch height as predicted by FIRESTAR in absence of tree canopy (virtual
crown) and by the plume theory (Van Wagner 1973).


Fig. 8 shows the minimum intensity necessary to get ignition of canopy fuel elements at a
given height above the g
round. In these virtual crown simulations, we considered that
ignition occurred as soon as the gaseous phase reached the ignition temperature of fuel (600
K). A more accurate method should consider the duration of heating and the heat required to
evaporate

the water of canopy fuel elements.



D2.4
-
2
-
18
-
1000
-
1





Page
13

of 2
6



Fig. 8

Critical intensity for crown ignition. The fitted curve is in agreement with plume theory
trend (power to the 2/3 of intensity).

The predictions of FIRESTAR were fitted to a power law of the fire intensity as s
hown Fig. 8
leading to the following formula

a
i
T
I
h


600
7
.
18
3
2
, (5)

where
T
a

(K) is

the a
m
bient absolute temperature related with the temperature
t
a

(°C) as
T
a

= 273.15 +
t
a

.

It is worth noting that again the exponent is the same as from the therma
l plume theory.


5
.

Influence of tree canopy (no wind)

We defined a tree canopy model based on data reported by Mitsopoulos and Dimitrakopoulos
(2007), who measured characteristics of a variety of Aleppo pine (
Pinus halepensis
) forests in
Greece. Fig. 9 sh
ows the vertical profile of crown bulk density we used for the foliage
(needles). This profile extends over 10 m height, from 3 m to 13 m in Fig. 9. The maximum
density was 0.16 kg/m
3

and the average value was about 0.08 kg/m
3
. We added a second
family of
fuel representing the smallest twigs (0
-
6 mm diameter) with a maximum density
half the one of the needles and the same vertical profile. The tree canopy model was given
the label ‘light crown’.


D2.4
-
2
-
18
-
1000
-
1





Page
14

of 2
6


We also defined a very dense tree canopy by setting the maximu
m value of needles density
to 0.8 kg/m3 (five times denser than the the light crown) and it was given the label ‘dense
crown’ (we ignored twigs in this case). The dense crown must be viewed as a limiting case,
not as a realistic canopy.

We used a surface
-
t
o
-
volume ratio of 10000 m
-
1

and a material density of 800 kg/m
3

for
needles (data measured on
Pinus halepensis
, INRA).

Crown bulk density profile (needles)
0
2
4
6
8
10
12
14
16
18
0
0,05
0,1
0,15
0,2
Bulk density (kg/m3)
Height (m)

Fig.9

Crown bulk density profile (canopy model for Aleppo pine)


The addition of vegetation in FIRESTAR 2D results in solving a set of
mass and energy
equations for the solid phases and including new terms in momentum and turbulence
transport equations of the gaseous phase.

In the present simulations the canopy fuel elements gained energy from the gaseous phase
through convection and thei
r temperature increased. A corresponding heat loss term was of
course present in the gaseous phase energy equation. For the fine fuels we considered here
the solid phase temperature was not significantly different from the gaseous phase
temperature due to
the response time of fuel particles was small (order of 1 s)

; in addition
radiation was not considered in these computations and so the canopy elements were not
heated by radiation.

The mechanical interaction between the gas flow and the vegetation was re
ndered through a
drag force term in the momentum equations and additional production and dissipation terms
in the
k

and


. These last terms were treated like in Green (1992).
Although this modelling of air flow
-
vegetation interaction w
as developed for wind over plant
canopies, we applied it also to the no
-
wind (no ambient wind) present simulations.


D2.4
-
2
-
18
-
1000
-
1





Page
15

of 2
6


As a first step, we considered canopies of crown base height ranging between 3 and 20 m
and we looked for the minimum intensity that causes

scorch at the crown base. The results
are plotted in Fig. 10. For a given intensity, we can see that the crown scorch height
increases very slightly in presence of a canopy at height intensities, with respect to the no
canopy case (virtual crown). But ove
rall the effect is negligible.



Fig. 10

Minimum intensity necessary to get scorch at crown base height level.

Fig.11 shows scorch height as computed with two different levels of density, in canopy
crowns existing between 3 m and 13 m above the ground. We

also plotted the plume theory
prediction and the prediction of FIRESTAR with no crown (‘virtual crown’) for comparison.
Obviously the presence of the canopy increased the crown scorch height with respect to the
virtual canopy or plume theory predictions.

Furthermore the presence of the canopy had
more effect for a dense canopy.



D2.4
-
2
-
18
-
1000
-
1





Page
16

of 2
6




Fig.
11

Crown scorch height as function of heat source intensity for crowns ranging between
3 and 13 m.

We show the turbulent kinetic energy fields (Fig.12) and the vertical ve
locity fields (Fig. 13)
as computed by FIRESTAR respectively with no crown (left) and with a light crown (right)
starting at 3.4 m above the ground (Fig. 12). Even for a light crown the foliage enhanced the
production of turbulence (Fig.
12
) and decreased
the vertical velocity (Fig. 13). As a result
the width of the thermal plume was increased due to the presence of the canopy. At a given
level above the ground, the addition of a canopy was expected to produce more lateral (or
horizontal) diffusion of heat
and conversely to give a lower maximum temperature on the
plume axis (heat source axis). In fact examination of the corresponding temperature fields
(Fig. 14) revealed that the lateral diffusion of heat was actually increased but that the
maximum temperatu
re at a given level above the ground (on the axis) was also increased
and not decreased. This surprising behaviour eventually explains why the addition of a
canopy in the simulations increased the crown scorch height as computed with FIRESTAR.

An obvious b
ut bad reason for the above behaviour would be that mass or energy
conservation was not satisfied. But we started the present study with a verification of the
conservation principles. In fact we verified them when a heat source was added (with no
canopy) a
nd, even, we did some modifications to the solution of equations for chemical
species conservation due to we observed that the mass ba
lance was not fully respected.
The
modifications corrected the problem and besides, in the present computations, the equat
ions
for chemical species were not solved since we used a heat input to the energy equation and
not the burner based on CO injection.


D2.4
-
2
-
18
-
1000
-
1





Page
17

of 2
6



Fig.

12

Turbulent kinetic energy fields above the heat source (power 178 kW/m). (left) with
no canopy (right) with a l
ight canopy.


Fig.

13

Vertical velocity fields above the heat source (power 178 kW/m). (left) with no
canopy (right) with a light canopy.


Fig.
14

Gas temperature fields above the heat source (power 178 kW/m). (left) with no
canopy (right) with a light

canopy.


D2.4
-
2
-
18
-
1000
-
1





Page
18

of 2
6



Fig.

15

Horizontal velocity fields above the heat source (power 178 kW/m). (left) with no
canopy (right) with a light canopy.

It is worth noting that the field of horizontal component of velocity was significantly modified
by the presence of th
e canopy (Fig.15). Under the canopy two symmetrical areas are
characterized by higher horizontal components of velocity with

respect to the no canopy
case;
they results from a downward flow approaching the ground level. This means that
entrainment of fresh

air into the plume was probably enhanced under the canopy and also in
the first meters of the canopy. Above 5 m and within the canopy, we observed the reversed
effect, due to drag: the entrainment of fresh air was reduced. This could explain that
temperat
ure did not decrease within the canopy, but rather increased, with respect to the no
canopy case.

Finally we observed that the presence of the canopy induced almost no significant change in
the temperature fields under the canopy layer. This is why there w
as almost no canopy effect
in Fig. 10 since crown scorch was tracked at crown base height level. This observation also
means that we do not expect an effect of the presence of the canopy on the minimal
intensity necessary to get ignition of the crown base.


6
.

Simulations with a wind

The initial objective of the present study was also to run computations of thermal plumes in
the presence of an ambient wind to determine how the wind affects the predictions of crown
ignition or crown scorch height, with respe
ct to the no
-
wind case. For example, Van Wagner
(1973) derived a simple trigonometric correction to the no
-
wind model of crown scorch: he
assumed that the plume is just tilted of a constant angle from vertical direction, under very
low wind speed condition
s (less than 1 m/s). For upper wind speeds, the question remains
open. Furthermore the addition of a tree canopy would make the situation much more
complex since the turbulent wind field is strongly influenced by the presence of trees.



D2.4
-
2
-
18
-
1000
-
1





Page
19

of 2
6


We observed the fol
lowing behaviors when we ran thermal plume computations in the
presence of an ambient wind.

Even with a small wind speed (1 m/s at 10 m height), the simulated plume was more and
more tilted from vertical as simulated real time elapsed. At the end the plume

was
characterized by an unstable flow and, in average, the plume was flattened on the ground
over downwind distances of one or a few tens of meters (Fig
.
16). This was observed for
powers of 100, 400 and 800 kW/m.


Fig
.

16

Fields of gas temperature and v
elocity vector at time 235 s. The power of the burner
was 400 kW/m and the wind speed was 1 m/s at 10 m height. The full domain was
100 m long and 50 m high. The label 5 m/s here on the figure and below (fig. 17,

18) is a graphical image of a velocity vect
or of 5 m/s value.

Fig. 17 shows a zoom of Fig. 16 on the hottest part

of the plume (i.e. the flame).

It can be
seen that the flame is horizontally oriented over several meters. Although little amount of
data is available on two
-
dimensional firelines, this

behaviour had not been reported for such
low wind speeds to our knowledge. To support this assertion one can refer to
Nelson and
Adkins (1986)

who ran experimental wind
-
driven fires in laboratory conditions or to
Sekine
(1997)

who reported outdoor crib fi
res driven by wind. In this last case, the plume showed a
straight and constant trajectory inclined of 35° from the vertical direction for ambient wind
speed ranging between 1 and 5 m/s, which
contrast

with the unstable behaviour of the
plume as simulated
with FIRESTAR (see also below).

As discussed above, the behaviour observed in FIRESTAR simulations was not reported in the
literature for wind
-
driven firelines in woody fuels, but it is not necessarily unrealistic. Indeed
as reported by Sinai and Owens (19
95) the attachment of the plume basis to the ground in
the down wind area is also known as a Coanda phenomenon. Sinai and Owens (1995)
simulated pool fires with cross
-
wind using a Computational Fluid Dynamics 3D model (CFDS
-

D2.4
-
2
-
18
-
1000
-
1





Page
20

of 2
6


FLOW3D code) and their simulatio
ns predicted the attachment of the flame over a significant
length in the downwind direction. Comparison with an experimental square pool fire showed
that the predicted plume was more inclined than the observed one. In addition it seems that
the predicted
attachment length was higher than the observed one, although the conclusion
of the author was not so clear on this point. The experimental square pool fire had a 54000
kW/m power (total power divided by square side) and the wind speed was 4,5 m/s at 10 m
h
eight. These conditions are very different from the conditions we tested with FIRESTAR, but
reminding that fire intensity scales with the cube of wind speed in correlations for flame
angle (see Nelson and Adkins 1986), we can see that the two situations ar
e still comparable.

To explain the behaviour observed in FIRESTAR simulations, we also assumed that it might
be due to the two
-
dimensional assumption of FIRESTAR. In a two
-
dimensional configuration,
the wind cannot flow across the thermal plume, whereas so
me qualitative measurements
have shown the opposite result (Beer 1991). Based on similarity analysis and mass
conservation arguments, Raupach (1990) showed that the wind should usually cross over a
thermal plume from a forest fire, except in fires of very
strong intensity. Recently three
-
dimensional computations of grassland fires using HIGRAD/FIRETEC have shown that above
some threshold of wind speed, the wind crossed over the fire front and the thermal plume
(Figure 2 in Cunningham and Linn 2007).



Fig.

17

Fields of gas temperature and velocity vector at time 195 s (zoom of Fig.

16). The
power of the heat source was 400 kW/m and the wind speed was 1 m/s at 10 m
height. The heat source was between 4

8.8 m and 51.2 m along X direction and was
0.6 m height.

The full domain was 100 m long and 50 m high.



D2.4
-
2
-
18
-
1000
-
1





Page
21

of 2
6


Even if the two
-
dimensional hypothesis may not fully explain the results we got in the
presence of a wind, the above considerations show the limitations of a two
-
dimensional
approach when one considers the in
teraction of the wind with a fire front or a thermal plume.

In FIRESTAR computations of a thermal plume with a wind, the formation of a large vortex
on the downwind side of the thermal plume was also observed systematically (Fig. 16, Fig.
18b). This vorte
x also ensured recirculation of the flow in the downwind area of the domain
and it ‘travelled’ towards the downwind boundary (right boundary, the wind blowing from the
left). The occurrence of vortices did not lead to a very unstable trajectory of the ther
mal
plume for power below or equal to 800 kW/m (a mean trajectory can be defined). At a power
of 1600 kW/m, the thermal plume was in contrast ver
y unstable, as shown in Fig. 18
: the
plume traject
o
ry had very large oscillations and it would not make sense t
o determine a
mean tr
a
jectory and a mean vertical profile of temperature as we did in the no
-
wind cases.



D2.4
-
2
-
18
-
1000
-
1





Page
22

of 2
6






Fig. 18

Fields of gas temperature and velocity vector at times (up) 185 s, (middle) 210 s,
(bottom) 235 s. The power of the burner was 1600 kW
/m and the wind speed was 1
m/s at 10 m height. The full domain was 100 m long and 50 m high.



D2.4
-
2
-
18
-
1000
-
1





Page
23

of 2
6


For the above reasons, we preferred to continue the study with a second model, operating in
a three
-
dimensional configuration, to establish the conclusions
.

For
this, we will use the
HIGRAD/FIRETEC 3D model of fire behavior (Linn
et al.

2002). Besides, in this model, cyclic
boundary conditions are also implemented. This kind of conditions is necessary to get realistic
turbulent wind profiles over tree canopies (Sh
aw and Patton 2003, Shen and Leclerc 1997,
see also deliverable D2.2
-
2). A specific new deliverable has been planned to report this future
work.


7
.

Needs for reference data

The
needs for experimental data become

apparent from the simulations of the effect
s of
canopy or wind, on the thermal plume. We observed interesting behaviours but with no
experimental support to assess the relevance of these predictions. Experimental works are
beyond the scope of the present workpackage, they are rather conducted in th
e WP2.3. But
we can identify needs from the above work.

The plume theory is well established for free plumes above a circular heat source or a line
heat source and, as we already mentioned, it should apply some distance above the heat
source. In fact the v
ertical development of the plume may be modified depending on the
heat source characteristics or the finite size of any real line heat source, so that it is
interesting to verify where the theory applies well. Such a work has been done in the frame
of WP2.
3 in laboratory conditions. We measured temperatures in thermal plumes from line
fires in no
-
wind and no
-
slope conditions at INRA facilities and we found that the plume
theory applied very well above the flames (at least for temperatures below 500 K) and w
ith a
proportionality constant very close to that of Van Wagner (1973).

But the main need of data is to be collected in field experiments to test
both the effect of the
presence of a canopy and

the effect of the wind, on a thermal plume. Little data are av
ailable
in both cases (see above discussion on the wind case). In such experiments, vertical profiles
of temperature at least should be measured (a vertical profile is enough in a no wind case
since the plume is expected to be vertical). The same experimen
ts could be carried out with
some artificialized fire line respectively in the open and under a tree canopy, in absence of
wind. This would give some evidence of the effects of vegetation elements on the thermal
plume. The wind case is much more complicate
d to instrument since the plume orientation
and structure should be very affected by ambient air flow. In addition, in presence of a tree
canopy, the mean and fluctuating parts of the air flow are strongly modified due to
vegetation drag and average behavi
our must be defined over very large time periods (several
tens of minutes) (Finnigan 2000). An alternative experiment would be to observe fire
damages on trees (crown scorch, bark carbonization) and to run the full fire model (i.e. not
only a heater) on th
e observed situations. This is a necessary final validation of the model,
but it does not say where the model might fail (thermal conditions about tree elements or
local heat transfer to these elements) if any significant departure from observation is foun
d.



D2.4
-
2
-
18
-
1000
-
1





Page
24

of 2
6


8
.
Conclusions

This preliminary study of fire impact on tree foliage was carried out with the FIRESTAR 2D
model, simulating a thermal plume from constant heat source. This configuration makes the
comparison with plume theory predictions easy.

First we
ran simulations with no wind and no canopy. When FIRESTAR was used with no
canopy to compute crown scorch heights, there was excellent agreement with the predictions
of plume theory (Van Wagner criterion for crown scorch). For crown fire initiation, FIRES
TAR
simulations were used to derive a simple relation between the critical intensity leading to
crown ignition and the crown base height. This relation was also consistent with plume
theory outcomes.

Then we ran no wind simulations but with a canopy. Accor
ding to FIRESTAR simulations, the
addition of a canopy was found to increase the crown scorch height. This means that
temperature was higher with a canopy than with no canopy, at a given level above the
ground within the canopy. This result has not been fu
lly understood to date, but we plan to
examine it further. Thus, we rather conclude that the flow was significantly affected by the
presence of the canopy and that the effect was more important for a dense canopy.

Finally we ran simulations with an ambient

wind but with no canopy. According to FIRESTAR
simulations, the addition of a weak ambient wind (1 m/s) completely modified the structure
of the plume, which was not unexpected, but also led the plume base (i.e. the flame in a real
fire) to be bent over t
he ground beyond several meters. The observed behavior of the plume
which was also very uns
table in some cases

led us to investigate the interaction between the
wind and the thermal plume (and latter the canopy) with a second model (HIGRAD/FIRETEC)
to esta
blish the conclusions.

References

Beer T. (1991) The interaction of wind and fire.
Meteorology

54
,287
-
308.

Cruz M.G., Butler B.W., Alexander M.E., Forthofer J.M., Wakimoto R.H. (2006) Predicting the
ignition of crown fuels above a spreading surface fire.
Part I: model idealization.
International
Journal of Wildland Fire

15
, 47
-
60.

Cunningham P., Linn R.R. (2007) Numerical simulations of grass fires using
a coupled
atmosphere
-
fire model
: dynamics of fire spread.
Journal of Geophysical Research

112
,
D05108.

Daly B.J.,Harlow F.H. (1970) Transport equations in turbulence.
Physics of Fluids

13, 2634
-
49.

Finnigan J (2000) Turbulence in plant canopies.
Annual review of fluid mechanics

32
, 519
-
571.

Green S.R. (1992) Modelling turbulence air flow in a stand of widel
y
-
spaced trees. Phoenics,
Journal of Computational Fluid Dynamics

and Applications
5
, 294
-
312.

Linn RR, Reisner J, Colman JJ, Winterkamp J (2002) Studying wildfire behavior using
FIRETEC.
International Journal of Wildland Fire

11
, 233
-
246.

Magnussen BF, Hj
ertager B (1976) On mathematical modeling of turbulent combustion with
special emphasis on soot formation and combustion. In ‘Proceedings of the 16th Symposium
(International) on Combustion’. pp. 719
-
729. (The Combustion Institute: Pittsburgh).

Markatos N.
C., Malin M.R., Cox G. (1982) Mathematical modeling of buoyancy
-
induced smoke
flow in enclosures.
International Journal of Heat and Mass Transfer

25
,

63
-
75.

Mercer G.N., Weber R.O. (1994) Plume above line fires in a cross wind.
International Journal
of W
ildland Fire

4
(4),201
-
207.

Michaletz S.T. and E.A. Johnson (2006) A heat transfer model of crown scorch in forest fires.
Canadian Journal of Forest Research

36
:2839
-
2851.

Mitsoploulos I.D., Dimitrakopoulos A.P. (2007) Canopy fuel characateristics and pote
ntial
crown fire behaviour in Aleppo pine (
Pinus halepensis

Mill.) forests.
Annals of Forest Science

64
, 287
-
299.

Morvan D.

Dupuy J.L. (2004) Modeling the propagation of a wildfire through a Mediterranean
shrub using a multiphase formulation.
Combustion an
d Flame

138
,

199
-
210.


D2.4
-
2
-
18
-
1000
-
1





Page
26

of 2
6


Nam S., Bill R.G (1993). Numerical simulation of thermal plumes.
Fire Safety Journal

21
:231
-
256.

Nelson R.M., Adkins C.W. (1986) Flame characteristics of wind
-
driven surface fires.
Canadian
Journal of Forest Research

16
,

1293
-
1300.

Ra
upach M.R. (1990) Similarity analysis of the interaction of bushfire plumes with ambient
winds.
Mathematical and Computerl Modelling

12
,113
-
121.

Sekine T. (1997) Behaviour of wind
-
blown crib fires. Thirteenth meeting of the UJNR panel
on fire research and
safety, March 13
-
20, 1996. (Ed.) kelly Ann Beall, Building and Fire
Research Laboratory, National Institutes of Standards and Technology, Gaithersburg (MD),
US. NISTIR 6030, 8p.

Shaw R.H., Patton E.G. (2003) Canopy element influences on resolved and subgri
d scale
energy within a large
-
eddy simulation.
Agricultural and Forest Meteorology

115
, 5
-
17.

Shen S., M. Y. Leclerc (1997) Modelling the turbulence structure in the canopy layer.
Agricultural and Forest Meteorology

87
, 3
-
25.

Shih T.H., Liou W.W., Shabbir
A., Yang Z., Zhu J. (1995) A new
k
-


eddy viscosity model for
free buoyant plumes.
Computational Fluids

24
,227
-
38.

Sinai Y.L., Owens M.P. (1995) Validation of CFD modelling of unconfined pool fires with
cross
-
wind: flame geometry.
Fire Safety Journal

24
, 1
-
34.

Thomas, P.H. 1963. The size of flames from natural fires. In ‘Proceedings of the Ninth
International Symposium on Combustion’. Academic Press Inc. New York, N.Y. 884
-
859.

Thomas, P.H. 1964. The effect of wind on plumes from a line heat source. Joint F
ire Res.
Organ., Fire Res. Stn., Boreham Wood, Engl. Fire Res., Note 572.

Van Maele K., Merci B. (2006) Application of two buoyancy
-
modified
k
-


models to different
types of buoyant plumes.
Fire Safety Journal

41
,122
-
138.

Van Wagner C.E. (1973) Height of c
rown scorch in forest fires. Canadian Journal of Forest
Research
3
(3), 373
-
378.

Van Wagner C.E. (1977) Conditions for the start and spread of crown fire.
Canadian Journal
of Forest Research

7
, 23
-
33.