Fire Safety Journal 48 (2012) 1–10

Contents lists available at SciVerse ScienceDirect

Fire Safety Journal

journal homepage: www.elsevier.com/locate/firesaf

Modeling of pool fires in cold regions

Gaffar Keshavarz n, Faisal Khan, Kelly Hawboldt
Process Engineering, Faculty of Engineering and Applied Science, Memorial University, St. John’s, Nfld., Canada A1B 3x5

a r t i c l e i n f o abstract

Article history: Fires and especially pool fires are among the most frequent accidents in process facilities. Flame
Received 17 June 2011 impingement and thermal radiation are the main hazardous characteristics of pool fires. Pool fires have
Received in revised form been the subject of numerous modeling and experimental studies covering a variety of areas such as
4 November 2011
fire and flame structure, emissive power, temperature distribution and fire characteristics. The effects
Accepted 14 November 2011
Available online 27 December 2011
of environmental parameters such as wind velocity, humidity and water/ice droplets in the air have not
been studied extensively. Further, the effect of surrounding surface reflectivity has not been studied.
Keywords: This issue is very important for cold regions like the Arctic, where outdoor surfaces are covered with
Absorption snow and ice for several months of the year. Furthermore, there is no comprehensive fire consequence
Arctic
modeling tool that includes pool fire development, environmental characteristics effects and thermal
Pool fire
radiation. This study proposes a new comprehensive model for steady state and fully developed pool
Surface reflectivity
fires. This new model takes into account the effects of environmental variables such as temperature, the
presence of droplets and surface reflectivity on thermal radiation and subsequently on the fire
consequence assessment.
& 2011 Elsevier Ltd. All rights reserved.

1. Introduction radiation and duration are the most important parameters in

Among different types of accidental fires like jet fires and flash
fires, a pool fire is the most frequent accident [1–3]. In storage
facilities, the design and implementation of surrounding systems
that collect and drain accidentally released liquid hydrocarbons is
a usual practice [4]. Ignition of the collected and drained liquid
results in a pool fire. A pool fire is defined as a turbulence
diffusion flame controlled by buoyancy forces [5–7]. Here the
combustion phenomenon is characterized by low momentum
diffusion flames. A pool fire is divided into three different regions:
flame base (persistent zone), intermittent zone and finally plume
zone [6–8]. The persistent region is rich in flammable vapors and
the flow is laminar, and as a result it maintains its shape and
structure. The intermittent zone is characterized by fluctuating
and turbulent flame and flow. In this zone, fuel vapors are
consumed completely and due to turbulence, the temperature
and radiation are higher compared to the persistent zone [9]. In
the plume zone no flaming reaction occurs. Turbulent flow and
smoke characterize the zone. Due to the presence of smoke and
soot, the radiation level is lower.
The behavior of a pool fire is tightly bound to the pool size
[10]. The size of pool controls the height of different zones,
thermal radiation and temperature; therefore, extrapolation of
these parameters may result in large errors. Temperature, thermal
n
Corresponding author.
E-mail address: ghaffarkeshavarz@gmail.com (G. Keshavarz).
consequence assessment of a pool fire [11]. Depending on these
parameters, the consequences of fire may be immediate (e.g. the
personnel are exposed to radiation) or delayed (e.g. heating up
the structures and domino effects). The hazards associated with
pool fires are related to thermal radiation [12]. Thermal radiation
depends on the type of fuel, soot yield and flame temperatures
[13]. Numerous studies have been conducted to investigate the
radiation phenomena of pool fires. Modak [14] presented a
theoretical study on thermal radiation of horizontal and axisym-
metric pool fires. Orloff [15] introduced a simple model to
calculate the radiation of pool fires by simplifying non-homo-
genous and non-isothermal fires to be equivalent to isothermal
and homogenous fires. Hamins et al. [16] conducted a set of
experiments on different fuels to develop and modify methodol-
ogies for thermal radiation measurements. Later, Rew et al. [6]
proposed a semi-empirical correlation for thermal radiation of
hydrocarbon pool fires. Chun et al. [12] have conducted both
experiments and CFD simulation to study radiation of pool fires.
Jensen et al. [13] and Krishnamoorthy [17] studied different
approaches to radiation modeling, effects of soot and smoke
effects on the estimated radiation level. However, environmental
variables like surface reflectivity or the presence of water/ice
droplets were not included in these studies.
Kim et al. [18] experimentally investigated variables such as
the effects of direct-downward water/ice spray on the burn-
ing rate, behavior and extinction of pool fires. They have
reported that small water/ice droplets are ineffective and may
inversely increase the burning rate. Later studies suggested that

0379-7112/$ - see front matter & 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.firesaf.2011.11.003
2 G. Keshavarz et al. / Fire Safety Journal 48 (2012) 1–10
Nomenclature
A area cross section (m2)
Cp specific heat (kJ/kg K)
D diameter (m)
E energy (kJ)
Lc height of the combustion zone (m)
M mass (kg)
P momentum (kg m/s)
T temperature (K)
Ta ambient temperature (K)
Z height (m)
Z0 height of imaginary source (m)
di distance (m)
f mass ratio (product/fuel)
g gravity of Earth (9.8 m/s2)
hc fuel heating value (kJ/kg)
hv heat of evaporation (kJ/kg)
m_ fuel burning rate (kg/s)
m0 fuel burning rate per volume (kg/s m3)
q released energy (kJ)
direct-downward spray is not the optimum direction and the
efficiency of extinction increases as spray direction goes toward
horizontal [19]. Chen et al. [20] studied the effect of the initial fuel
temperature on the burning rate and burning states of pool fires.
Their experimental study shows that the duration of steady
burning decreases as the initial temperature of the fuel increases.
Furthermore, if the initial fuel temperature reaches the boiling
temperature, there would not be a steady burning period.
Ravigururajan [21] has proposed a method to calculate the effect
of water/ice droplets on thermal radiation attenuation.
Despite numerous works investigating pool fires, the effect of
environmental variables has not been comprehensively studied.
In most of the studies only a single environmental parameter has
been considered. The combination of these environmental variables
will affect the consequences of any pool fire, especially in the Arctic
region, as these parameters coexist and affect the behavior and
thermal radiation of a pool fire. In this study a new model is
developed that takes into account environmental parameters,
including temperature, wind, the presence of droplets and surface
reflectivity. The Fay model [11] was used as the base for the model.
This model has the advantage of a continuous temperature distribu-
tion. Furthermore, it also includes the effect of water/ice droplets
and surface reflectivity. Therefore, the proposed model accounts for
all environmental variables related to the Arctic.
2. Pool fire modeling
The flame of pool fires is typically nonuniform in temperature
and species concentration distribution in terms of three-dimen-
sional calculations; therefore, modeling of pool fires is challen-
ging. Pool fire models are generally classified into two groups; the
simplest model assumes thermal radiation controls fuel evapora-
tion. These models do not account for combustion and air
entrainment in their calculations; rather, they calculate radiation
based on flame temperature, size and shape. Orloff [15] presented
such a model to calculate the radiation of pool fires. In contrast to
simple models, fundamental models prescribe the fire behavior
based on air entrainment, mixing, combustion, flow and plume
rising. Fay [11] has recently proposed such a model for a wide
variety of pool sizes and wind velocities.
q00 thermal radiation intensity (kW/m2)
rf reflection index
Greek symbols
a dimensionless correction constant
g dimensionless radiation constant
Z dimensionless correction constant
r density (kg/m3)
ra ambient air density (kg/m3)
jc combustion zone equivalent ratio
o vertical velocity (m/s)
Superscripts
‘‘ . ’’ average value
Subscripts
c combustion zone
p plume zone
r reflection
2.1. Fay model
This model is a two zone pool fire model that describes
flame properties, combustion and plume zones. The combustion
zone begins from the base of the fire and goes up to the end
of the visible flame. In this zone the evaporated fuel vapor reacts
with air in stoichiometric proportions and forms combustion
products. When fuel evaporates from the pool surface, it enters
a circulation region. Fuel vapor flows radially toward the edges
of the pool fire and then moves upward and inward towards
the flame tilt in the center line. A portion of fuel vapor moves
downward towards the liquid surface and hence the circulation
is completed. This fuel circulation provides fuel vapor for the flame
surface. Fuel vapor and air diffuse, respectively, outward and inward
and intersect each other on the flame surface, where reaction occurs
and heat is released. The released heat results in a temperature
increase and density reduction, and subsequently the upward
diffusion and bulk movement of products. The low density and high
temperature products enter the plume zone. The plume zone is
above the combustion zone and rises until temperature significantly
decreases and produced gases dilute.
Temperature and fuel concentration distribution are functions of
the mixing process. If mixing is diffusion driven and laminar, oxygen
and fuel vapor diffuse and reaction occurs. For this case, the flame
surface is thin and the temperature is almost equal to the adiabatic
flame temperature. Since the phenomenon is laminar, the changes
of temperature and combustion products concentration are sudden,
where both temperature and combustion products’ concentration
increase rapidly in a thin layer. In contrast to laminar and diffusive
flames, the rapid mixing ensures the temperature change is smooth
due to turbulence. Hence the flame surface temperature is lower
and flame surface thickness is greater compared to the laminar
regime. Details and equations can be found in Fay’s work [11].
As Fig. 1 illustrates, the Fay model is a simple, quick and direct
approach for pool fire modeling. This model requires fuel proper-
ties like heating value and heat of evaporation, stoichiometric
ratios, pool size, wind and the environment’s temperature.
Although this model has been verified and calibrated against
the results of field tests and has acceptable accuracy, there are
some limitations that, if overcome, will result in a better and
more comprehensive model.
 G. Keshavarz et al. / Fire Safety Journal 48 (2012) 1–10 3

written as
Fuel type and pool size
qz ¼ C p T z o_rAc ð5Þ

where ȯ is average vertical velocity at z and Tg is the average

Mass burning rate
temperature. Using Eqs. (1)–(5), the temperature profile in the
combustion zone can be determined. As the temperature in the
Fuel Froude number combustion zone changes, Fay’s model cannot be used to predict
thermal radiation. Furthermore, this model is limited to the
combustion zone, and the contribution of the plume zone to the
Wind velocity effect radiation is not taken into account. In this model variables such as
mass, momentum and temperature are discontinuous at Z¼Lc. In
addition, parameters like temperature and mass flux are incor-
Height and cross area of com bustion zone
rectly modeled near the combustion zone. These parameters
become accurate when ZbLc and in this situation the tempera-
ture is too low and radiation is negligible. To maintain continuity
Temperature, mass ux and velocity of the combustion zone
and accuracy of variables in the plume zone, we may rewrite mass
and momentum equations as
Temperature, mass ux and velocity of the plume zone " #0:33
243 Zp ra 2 Ep ap 4
Mp ¼ M c ðLc Þ þ ðzLc Þ1:66 ðz ZLc Þ ð6Þ
2500 CpTa
Thermal radiation to receiver
 0:4
5Zp Ep
Pp ¼ P c ðLc Þ þ pffiffiffiffiffiffi ðM p M c ðLc ÞÞ0:8 ð7Þ
Fig. 1. Steps that must be followed in the Fay model. 4ap C p T a ra

The other equations can simply be written as

2.2. Revised model
Pp
oUp ¼ ð8Þ
Mp
In the Fay model, temperature is constant over the height of
the combustion zone up to Lc. This does not reflect reality, where
Ep
temperature at the edge of the pool base is close to air tempera- T Up T a ¼ ð9Þ
Mp C p
ture and the center of pool base is close to the fuel boiling
temperature. At the edge of the pool base, temperature is
Mp
controlled by the entrainment flow and at the center, temperature Ap ¼ ð10Þ
is controlled by fuel evaporation. In other words, this model will
ra oUp
overpredict the temperature near the ground and subsequently The variable temperature in the combustion zone and rewrit-
overpredicts the thermal radiation from this section of the fire. ing of the plume zone equations imply that we need new
Neglected heat losses are the main cause of overprediction. equations to calculate thermal radiation from the plume zone to
Heat losses are important in thermally thin layers like the base of a target. Conservation of energy for a segment can be written as
pool fire. To take into account these losses of thermal energy,
Z A
energy conservation equations are required to be rewritten. To
q00 DA ¼ q00 9di dA, A ¼ 2p di2 ð11Þ
obtain the temperature profile for the combustion zone, we have 0
assumed that the pool fire is axisymmetric and that the fuel
burning rate per unit volume is the same over the entire where q00 is the emissive power at a segment of outer boundary
combustion zone. Assuming an axisymmetric pool fire reduces with the area DA and q00 9di is the received thermal radiation at
the accuracy especially in the presence of cross-wind. Assuming distance di from this segment. Eq. (11) can be rewritten as
that fuel burning rate is constant for the entire combustion zone Z A !!
n :di 00 !!
affects the accuracy of solution. However, such an assumption is sT 4 DA ¼ q 9di dA ð n :di Þ 40 ð12Þ
required in order to close the problem and obtain unique results. 0 di
This burning rate per unit volume can be obtained as ! !
where n is the normal of DA and di is the radiation direction
Z Lc Z Lc vector. Eq. (12) assures that the maximum of thermal radiation is
2 0:5 1
V¼ Ac dz ¼ ac Dz dz ¼ ac 0:5 DLc 2 ð1Þ in the direction of DA normal. Therefore we can say that for any
0 0 3 3
distance from the DA segment, q00 9di is constant for all the
D2 =4 3 D directions and the difference in received thermal radiation is
m0 ¼ ¼ ð2Þ !!
V 4 a0:5 2 controlled by n :di =di. Therefore q00 9di for any distance can be
c Lc
calculated as (see the proof in Appendix A)
where m0 is the fuel burning rate per volume (kg/m3 s) and the
other parameters are as outlined above. The energy conservation sT 4 DA
q00 9di ¼ 2
ð13Þ
equation is written as 4di
X
qc þ qz þ qz þ dz ¼ 0 ð3Þ Therefore the thermal radiation that the target receives from
the DA segment is
where qc, qz and qz þ dz are combustion released energy, inlet/
outlet energy at z and inlet/outlet energy at zþ dz: sT DA 4 DA ! !
q00 9DA ¼ ð n :di Þ ð14Þ
0
qc ¼ m V g hc , V g ¼ wA ð4Þ 4di3
where w is the width of the grid, and it can be shown that it will Finally, thermal radiation from the outer boundary can be
be canceled from the equation. Vertical heat transfer can be obtained by integrating Eq. (14) over the entire combustion and
4 G. Keshavarz et al. / Fire Safety Journal 48 (2012) 1–10

plume zones: coefficient kj is

Z !! kj ¼ kdj Ad þ ð1Ad Þka ð19Þ
Ab
T DA 4 ð n :di Þ
q00 ¼ s dA ð15Þ where kdj, ka and Ad are the absorption coefficients of droplets
0 4di3
related to jth wave band, pure air absorption coefficient and the
where Ab is the boundary of the pool fire. Now we can further relative area of droplets, respectively. This relative area is directly
expand the proposed model to include the effect of surface related to md and is inversely related to droplet diameter. If s2 is
emissivity on thermal radiation. Surface emissivity is important, mathematically equal to 0, then the relationship between Ad, md
especially in regions like the Arctic, where the reflection index of and m becomes (see Appendix B for proof)
the ground is very high due to the presence of snow and ice
3md
covered surfaces. These surfaces have a higher reflection index Ad ¼ ð20Þ
4rd m
and subsequently magnify the thermal radiation that the target
receives. The assumption is that radiation index of the ground is where rd is particle density (kg/m3). For the cases in which s2 is
constant and does not change. This assumption is completely not zero, one needs to integrate this area as
valid for far distances (depending on pool fire size). However, it Z
 
3md ru cdf ðr þdrÞcdf ðrÞ rm
should be mentioned that at very close distance, reflection index Ad ¼ dr, cdf ðrÞ ¼ 0:5 1þ erf pffiffiffiffiffiffiffiffiffi
4rd rd r þ dr=2 2s2
may reduce slightly due to phase changing. For each radiative
ð21Þ
section corresponding to the target location, there is one specific
angle that results in thermal radiation reflection. Therefore we where cdf is cumulative distribution function and rd and ru
have extended Eq. (15) as determine the interval of integration. These values are symme-
0 1 trical with respect to m and the size of interval depends on s2. For
Z Aob 4 ! ! ! ! the jth band, the thermal energy after passing a distance x in air is
00 T @ n :di n : dir A
q ¼s þ rf dA ð16Þ
0 4 di3 dir 3 given as
2hc2 1
where dir is the traveling distance vector of a reflected beam and ej ðx,TÞ ¼   expðkj xÞ ð22Þ
lj 5 exp ðhcÞ=ðlj k TÞ 1
U
rf is reflectivity index of the surface; dir is always greater then di
particularly if the target is close to the pool fire. Another where lj is the average wavelength of the jth band. Hence, the
important issue that has not been considered in previous studies total thermal energy that passes through the media (air and
is the absorption coefficient of air, which has been taken as zero droplets) is
in the above equations. This is valid for most cases since the air Nw
X Nw
X 2hc2 1
absorption coefficient is very low, 5  10  6 m  1 [22]. However, eðx,TÞ ¼ Ej ðxÞ ¼   expðkj xÞ ð23Þ
lj 5 exp ðhcÞ=ðlj k TÞ 1
U
the presence of water/ice droplets in air can significantly increase j¼1 j¼1

the absorption coefficient of air and subsequently reduce the
intensity of radiation. Temperature, wavelength, size and mass
per volume of the droplets are the main parameters that affect the
absorption coefficient [23,24]. As the emission from water/ice
droplets is much smaller than the energy that they absorb [25],
thermal radiation from the droplets has not been taken into
calculations.
The absorption coefficient and reflection index of ice and
water/ice droplets for a variety of temperatures and wavelengths
have been provided in Refs. [24,26–28]. The absorption coeffi-
cients of ice and water/ice droplets vary significantly with
wavelength. On the other hand, wavelength also determines
radiation energy:
 between the proposed method and the Fay model is explained
2hc2 1 2hc2 hc
el ¼    exp 
l5 exp ðhcÞ=ðlkU TÞ 1 l5 lkU T
where h, c and kU are Planck’s constant (6.62  10  34 J s), the
speed of light (3  108 m/s) and Boltzmann’s constant
(1.38  10  23 J/K), respectively. A differentiation of this equation
(17) determines which wavelengths contribute more to energy
transfer:
hc 1 2897
lmax ¼ , lmax ðmmÞ ¼
5kU T T
In the present model, the interval of thermal radiation wave-
length, 10  8–10  3 m, is divided into Nw wave bands and attenua-
tion is calculated for each band. The widths of these bands are not
equal and reduce as wavelength becomes closer to lmax. For each
wavelength band, one needs to calculate the absorption coeffi-
cient of air. Assuming that mass per volume of droplets in the air
is md and that droplets have a Gaussian distribution with mean
and variance values m and s2, respectively, for any wave band like
the jth with average wavelength lj, the corresponding absorption
This equation can be substituted in Eq. (16) to obtain the final
form of the thermal radiation equation:
0 1 0 1
Z Aob !! ! !
1 n :di n : di r
q00 ¼ eðdi,TÞ@ 3 A þrf  eðdir ,TÞ@ A dA ð24Þ
4 0 di dir 3
So far a model has been developed that is capable of including
the environmental parameters effects on the behavior of pool
fires. Fig. 2 presents the sequence of steps for the proposed model.
The main environmental parameters for the Arctic region are
snow/ice covered surfaces and droplets. In the next section we
will see how these two parameters can affect the thermal
radiation intensity from a pool fire. The quantitative comparison
through a pool fire modeling case study. Here, we use a pool fire
ð17Þ
of gasoline with 1.4 m diameter to illustrate these models. The
assumption is that there is no wind.
3. Results and discussion
3.1. Results
ð18Þ The Fay model was revised and extended to include the effect
of environmental parameters like surface reflectivity and the
presence of water/ice particles. As mentioned by Fay [11], his
model is discontinuous at Z¼Lc. An example of this discontinuity
has been shown in Fig. 3 for the purpose of illustration. The main
reason for this discontinuity is that vertical energy transport due
to mass movement has not been considered in the combustion
zone and subsequently this zone has a constant temperature. The
initial temperature of the plume zone depends on the coefficients;
particularly Zp Fig. 3 shows temperature in the plume zone for
three different Zp. The major issue of using the Fay model is that
 G. Keshavarz et al. / Fire Safety Journal 48 (2012) 1–10 5

950
Fuel type and pool size Fay model
900 New model

850
Mass burning rate
800

Temperature (K)
Fuel Froude number 750

700
Wind velocity effect
650

600
Modified temperature of the combustion zone
550

500
Temperature, mass flux and velocity of the plume zone
450
0 0.5 1 1.5 2 2.5 3 3.5
Height (m)
Surface reflectivity and droplets effect
Fig. 4. Temperature distribution for different models: Fay model (1) and new
model (2).
Thermal radiation to receiver
55
(1)
Fig. 2. Solution steps of the proposed model. Thermal Radiation (KW/Square meter) 50 (2)
45
1000
40
(1)
900 (2) 35
(3)
30
800
25
Temperature (K)

20
700
15
600 10

500
400
300
0 0.5 1 1.5 2 2.5
Height (m)
Fig. 3. Temperature distribution for the Fay model; very small Zp (1), small Zp
(2) and medium Zp (3). Temperature is discontinuous at the intersection of zones.
there is no specific value or method to obtain Zp and, as Fig. 3
shows, temperature highly depends on this parameter.
To avoid these issues of discontinuity in the temperature and
not having the exact value of Zp, the temperature profile in the
combustion zone has been modified (Fig. 4). Hence the value of Zp
is calculated based on continuity of mass and energy at the
intersection of the combustion zone and the plume zone. The
new distribution of temperature is the result of considering heat
transfer between different heights in the combustion zone.
As temperature is continuous, the initial temperature of the
plume zone is set equal to the temperature at the end of
combustion zone.
In Fig. 4 the integrals of temperature curves for both models,
the Fay model and the present model, are the same and the total
released energy is the same for both cases. However in the
present model, the temperature is low near the pool surface and
it increases with height as the released energy rate increases and
energy is transferred upward due to hot gases movement.
5
1.5 2 2.5 3 3.5 4 4.5 5
Radial Distance (m)
Fig. 5. Comparison of thermal radiation predicted by Fay model (1) and the
proposed model (2) at height 1.0 m.
3 3.5
Furthermore, the radiation equation that has been used in the
Fay model uses an average temperature and ultimately under-
predicts the radiation flux. Thermal radiation fluxes for both
models have been compared in Fig. 5. The difference is higher
near the pool fire and as the radial distance increases this
difference is reduced.
Another major limitation in the Fay model is not considering
radiation from the plume zone as it cannot account for radiation
where the temperature profile is not constant. Neglecting the
plume zone might be reasonable when the height of the combus-
tion zone is large or when the target is near the ground. However,
for the cases where the target is high, radiation from the plume
zone cannot be neglected. The thermal radiation flux of the plume
zone that has been calculated by the proposed model is shown in
Fig. 6. This figure clearly illustrates that the plume zone
contributes more at higher heights, although the total radiation
from this zone might be small compared to that of the combus-
tion zone. Fig. 7 demonstrates the percentage of plume zone
contribution compared to total radiation flux that a target
receives. This figure shows that plume zone contribution for
higher heights is more important compared to lower heights.
Therefore, neglecting radiation from the plume zone results in
severe underestimating of radiation flux at large heights.
6 G. Keshavarz et al. / Fire Safety Journal 48 (2012) 1–10

Difference in Thermal Radiation (KW/Square meter)

5 9
(1)
4.5 8 (2)
0. (3)
5
4 7
1
3.5 6

3 5
Height (m)

2.5 0.5 4
1.5
2

3
2
2.5

2
1.5
3

1
1
0.5

0
1 .5

0.5 0 2 4 6 8 10 12
2

Radial Distance (m)

2 3 4 5 6 7 8
Radial Distance (m) Fig. 8. Difference (proposed model  Fay model) in the received thermal radiation
at height 1.0 m; surface with reflectivity index 0.9 (1), reflectivity index 0.7 (2) and
2
Fig. 6. Contribution of plume zone in thermal radiation flux (kW/m ). reflectivity index 0.5 (3).

5 underestimates radiation. Furthermore, surface reflectivity makes

14
18

16

4.5 a small contribution to thermal radiation for very close distances

20

or far away from the pool fire. The reason for this small
4
contribution is that at short distances the reflection has a very
3.5 large angle and the second part of Eq. (16) becomes negligible.
Finally the difference reduces as radial distance increases.
14
Height (m)

3
12 We expect that the contribution of reflection reduces as the
18

16

2.5
12
2
14
12
1.5
12
1
0.5
10
2 3 4 5 6 7
Radial Distance (m)
Fig. 7. Precentage of plume zone contribution in the total radiation flux.
Fig. 7 also illustrates that as the radial distance increases, the
influence of the plume zone decreases and the combustion zone
dominates the thermal radiation flux. At large radial distances,
even for large heights, the radiation from the plume zone
becomes negligible compared to the combustion zone. In this
case, the radiation angle and the distance to the target become
almost the same for both the plume and combustion zones.
3.2. Surface reflectivity
The plume zone’s share of thermal radiation is small compared
to that of the combustion zone and neglecting it results in
underestimation of thermal hazards. Therefore, plume zone
radiation has been considered in all subsequent sections. After
correcting the temperature distribution and including the plume
zone in thermal radiation, we may investigate the effect of surface
reflectivity on thermal radiation. The Fay model does not account
for this parameter or we can say that the surface reflectivity is
0 for this model. For a snow and ice covered surface, the surface
reflectivity may be as high as 0.93.
In Fig. 8, the difference in thermal radiation flux predicted
by the Fay model and the proposed model is shown and
illustrates that as surface reflectivity increases, the Fay model
height increases. The reason for this behavior is that both the
reflection angle and distance increase with height. However, this
is not always the case and near the surface at approximate height
of 0.4–0.7 m, as shown in Fig. 9, the reflection contribution
increases slightly. This behavior is due to radiation from the top
of the combustion zone, from the bottom of the plume zone and
reflected thermal radiation. Except for the slight growth in the
reflected thermal radiation, the reflection contribution reduces as
8
height increases. A comparison between Fig. 10a and b shows that
the thermal radiation distribution changes significantly with
reflection.
The proposed model was further extended to include the effect
of water/ice droplets. The introduction of droplets into air reduces
the transparency of air, increases the extinction coefficient of air
and subsequently results in higher attenuation of thermal radia-
tion flux. The attenuation depends on parameters such as pool fire
temperature and wavelength, and droplet kind, size, distribution
and mass concentration in air. Here the effect of droplet size,
distribution, mass concentration and droplet kind (water or ice)
has been examined.
Fig. 11 presents the thermal radiation flux for several cases. In
this figure, variance is considered zero (all droplets have the same
size in each case). The mass concentration for Cases 2–4 is the
same. As shown in Fig. 11, large droplets have a negligible effect
on thermal radiation (comparing Cases 1 and 2). However, as the
droplet size reduces the attenuation increases, and for small
droplets the attenuation of radiation is rapid (Case 4).
Fig. 12 demonstrates the effect of droplets’ size distribution on
thermal radiation. As the variance increases, thermal radiation
reduces more due to the presence of smaller particles in air. It
must be noted that as the mean size of droplets reduces, variance
becomes more important.
The relationship between the attenuation coefficient and the
mean size of droplets is observed to be an inverse relationship.
Eq. (21) indicates a direct relation between mass concentration
and the attenuation coefficient. Fig. 13 shows that as mass
concentration increases, thermal radiation decreases.
 G. Keshavarz et al. / Fire Safety Journal 48 (2012) 1–10 7

5 0.
5

0.5
4.5 1
1
4

3.5
1
Height (m) 3 2

1
2.5
2

2 2
4

1
1.5

1
10

4
12

0.5
2
2 3 4 5 6 7 8 9 10 11 12
Radial DIstance (m)

Fig. 9. Reflection contribution in the thermal radiation when surface reflects the radiation (reflectivity index is 0.9).

5 5
5

5
2

2
4.5 4.5
5

4 4 10
10
3.5 3.5
Height (m)

Height (m)

5
3 3
2

2.5 2.5
10

10
5

2 2
1.5 1.5

5
1 1
2

50
50
100

100

10
10

0.5 0.5
5

2 3 4 5 6 7 8 9 10 11 12 2 3 4 5 6 7 8 9 10 11 12
Radial Distance (m) Radial Distance (m)

Fig. 10. Thermal radiation flux when (a) surface reflectivity index is 0.0 and (b) surface reflectivity is 0.9.

60 35
(1) (1)
Thermal Radiation (KW/Square meter)

(2) (2)
Thermal Radiation (KW/Square meter)

30 (3)
50 (3)
(4) (4)
25
40
20

30
15

20 10

5
10

0
0 2 2.5 3 3.5 4 4.5 5
2 2.5 3 3.5 4 4.5 5 Radial Distance (m)
Radial Distance (m)
Fig. 12. Effect of droplet diameter distribution on thermal radiation extinction; for
Fig. 11. Attenuation of thermal radiation flux for different droplets sizes: no all cases the mean diameter is 5m and distribution is Gaussian: s2 ¼ 0.0 (1) (all
droplet in air (1), very large droplets (D ¼ 400m) (2), large droplets (D ¼100m) droplets have the same diameter), s2 ¼1.0 (2), s2 ¼ 2.0 (3) and s2 ¼ 5.0 (4).
(3) and medium droplets (D ¼ 20m) (4).

Fig. 14 shows the impact of water droplets on thermal radia- increases gradually with the radial distance until reaching a peak
tion. Case 1 refers to the presence of water droplets in air while in and then decreases. Over long distances the difference reduces
Case 2 there are no water droplets. At each height, the difference since the thermal radiation of both cases decreases significantly.
8 G. Keshavarz et al. / Fire Safety Journal 48 (2012) 1–10

60 4

-4
(1)
Thermal Radiation (KW/Square meter)

(2)
50 (3) 3.5
(4) -6
(5) 3
40 -6
2.5

Height (m)

-4

-6
30
2
20
1.5
-2 -4
10 0 -6
1

0 0.5
2 2.5 3 3.5 4 4.5 5 8 6 4 2 0 -2 -4 -6
Radial Distance (m) 2 2.5 3 3.5 4 4.5 5
Radial Distance (m)
Fig. 13. Effect of droplet mass concentration on the attenuation of thermal
radiation: md ¼5 g/m3 (1), md ¼ 10 g/m3 (2), md ¼ 20 g/m3 (3), md ¼40 g/m3
Fig. 15. Difference in received thermal radiation (Case 1–Case 2), md ¼10 g/m3,
(4) and md ¼ 80 g/m3 (5).
D¼ 50m and ref. ¼0.9 for Case 1.

4 -5
-5 -5.5 45
.5
-4

-4 (1)
5 -5.5
3.5 -5. -6
40 (2)

Thermal Radiation (KW/Square meter)

-6
-6 -6.5 -6.5 35
3
-5

-7
30
5

-7
Height (m)

2.5
-5.

-7.5
25
5
-6.

.5
-7

-7
-6

2 20
-8
15
-7.

1.5
-8

10
1
-6.5

-7.5

5
-8
-7
-6

0.5 0
2 2.5 3 3.5 4 4.5 5 2 2.5 3 3.5 4 4.5 5
Radial Distance (m) Radial Distance (m)

Fig. 14. Difference in received thermal radiation (Case 1–Case 2), md ¼ 10 g/m3, Fig. 16. Effect of different type of droplets on the thermal radiation extinction:
D ¼50m. For this case the surface reflexivity has not been considered. water droplets (1) and ice droplets (2).

might be significantly different as the absorption coefficient of

3.3. Arctic region conditions water droplets highly depends on the wavelength.
The effect of ice droplets versus water droplets has also been
Figs. 8–14 indicate that surface reflectivity and the presence modeled and compared in Fig. 16. The mean value, variance and
of water droplets have significant and opposite effects on mass concentration are the same for both ice and water droplets
the thermal radiation flux. While surface reflectivity amplifies in this figure. For the specified pool fire with its specific tem-
thermal radiation, the presence of water droplets results in a perature distribution, ice droplets result in less thermal radiation
faster attenuation of radiation. Therefore, it is expected that the reduction compared to water droplets. As such, surface reflectiv-
interaction between these two parameters results in a more ity becomes the dominant affecting parameter up to a larger
complicated pattern. Fig. 15 demonstrates an example of this radial distance and height as shown in Fig. 17. In this figure, all
interaction. This figure shows the difference between Cases 1 and the parameters are the same as in Fig. 15 except the absorption
2. Case 1 refers to Arctic condition where a high surface reflectiv- coefficient, which is the absorption coefficient of ice. In general
ity and water droplets coexist. In contrast, Case 2 is for a scenario we can say that for the Arctic compared to a more temperate
where surface reflectivity is low (0.1) and there are no droplets region, thermal radiation is higher for short distances and is lower
in air. for large distances.
As shown in Fig. 15, the thermal radiation flux is higher in Case
1 for short distances and low heights. The distance and the height
at which thermal radiation of Case 1 is higher depend on the pool 4. Conclusion
fire diameter and fuel type. For short distances and low heights,
reflection compensates for the effect of water droplets because of A new model is proposed that is a modified and extended
the small reflection angle. For larger heights, the reflection angle version of Fay’s pool fire model. The Fay model has been revised
is large and subsequently water droplets become the dominant as it has several limitations: constant temperature for the
parameter. Note that these results also depend on the wave- combustion zone, discontinuity of variables at the intersection
lengths, which depend on the pool fire temperature. Hence, for a of the combustion and plume zones and neglecting the plume
different temperature distribution of the pool fire, the results zone in radiation calculations. Furthermore, the model was
 G. Keshavarz et al. / Fire Safety Journal 48 (2012) 1–10 9

where y and j are azimuthal and polar angles, respectively. The

-4
3.5 dA in this integral is equal to

-6 dA ¼ di sinðdyÞdi sinðdjÞ ¼ di2 sinðdyÞsinðdjÞ ðA:3Þ

3
-6
On the other hand, we know that
2.5
Height (m)

-2

lim sinðxÞ ¼ x ðA:4Þ

x-0
-4
2
Using Eq. (A.4) in Eq. (A.3) and substituting in Eq. (A.2) results in
Z A Z p=2 Z p=2

-6
1.5 -2
0 cosðyÞcosðjÞdA ¼ di2 cosðyÞcosðjÞdy dj ¼ 4di2
2 0 ðp=2Þ ðp=2Þ
1 -4
ðA:5Þ
4

0.5 Substituting Eq. (A.5) in Eq. (A.1) results in

-6
6

-2

8
2

1.5 2 2.5 3 3.5 4 4.5 sT 4 DA

q00 9di ¼ ðA:6Þ
Radial Distance (m) 4di2
Fig. 17. Difference in received thermal radiation (Case 1–Case 2), md ¼ 10 g/m3,
D ¼50m and ref. ¼ 0.9 for Case 1.

Appendix B
extended to account for surface reflectivity and the presence of
water/ice particles in air. As a result, the proposed model is For a Gaussian distribution, the Probability Density Function
effective in a region like the Arctic, where surface reflectivity and (PFD) can be shown to be
!
the presence of droplets are the main environmental parameters. 1 ðxmÞ2
The proposed model was expanded to account for radiation f ðxÞ ¼ pffiffiffiffiffiffiffiffiffiffiffiffi exp  ðB:1Þ
2ps2 2ps2
from the plume zone. Although the thermal radiation from this
zone is small compared to the combustion zone, the plume zone where m presents the mean value and s2 determines the width of
still has an important role, especially for large heights. The model distribution. ‘‘s2 equals 0’’ mathematically means that all the
also includes thermal radiation reflection from the surrounding droplets have the same diameter m. Then we have
surfaces. The model evaluates the presence of water/ice droplets
4prd 3
in air. All these modifications/extensions are important for fire md ¼ rd V d ¼ N m ðB:2Þ
3
risk assessment in cold regions.
The results show that thermal radiation increases significantly where Vd is the occupied volume by droplets and N is the number
near the surface due to reflection. However, droplets will have the of droplets per unit volume. Finally
opposite effect. For the Arctic region, where surface and droplets 3md
act simultaneously, thermal radiation is expected to be higher at Ad ¼ Npm2 ¼ ðB:3Þ
4rd m
short distances and less for large distances. However, these
distances are related to the diameter and height of the pool fire,
fuel type, surface reflectivity index, droplet size and distribution
and the mass concentration of droplets. Results show that there References
could be a considerable difference in the thermal radiation profile
based on the type of droplets. [1] M. Munoz, E. Planas, F. Ferrero, J. Casal, Predicting the emissive power of
hydrocarbon pool fires, J. Hazard. Mater. 144 (2007) 725–729.
[2] E.P. Cuchi, J. Casal, Flame temperature distribution in a pool fire, J. Hazard.
Mater. 62 (1998) 231–241.
Acknowledgment [3] M. Munoz, J. Arnaldos, J. Casal, E. Planas, Analysis of the geometric and
radiative characteristics of hydrocarbon pool fires, Combust. Flame 139
(2004) 263–277.
The authors gratefully acknowledge the financial support [4] K.S. Mudan, Thermal radiation hazards from hydrocarbon pool fires, Prog.
provided by the Natural Science and Engineering Research Coun- Energy Combust. Sci. 10 (1984) 59–80.
[5] P. Joulain, The behavior of pool fires: state of art and new insights, in:
cil (NSERC), Canada, and American Bureau of Shipping (ABS) to
Proceedings of the Symposium on Combustion, 1998, vol. 27, pp. 2691–2706.
conduct this research. [6] P.J. Rew, W.G. Hulbert, D.M. Deaves, Modeling of thermal radiation from
external hydrocarbon pool fires, Trans. IChemE 75 (1997) 81–89.
[7] P. Joulain, Convective and radiative transport in pool and wall fires: 20 years
of research in Poitiers, Fire Saf. J. 26 (1996) 99–149.
Appendix A [8] B.J. McCaffrey, Momentum implications for buoyant diffusion flames,
Combust. Flame 52 (1983) 149–167.
[9] S.J. Fischer, B.H. Duparc, W.L. Grosshandler, The structure and radiation of an
ethanol pool fire, Combust. Flame 70 (1987) 291–306.
[10] P.K. Raj, Large hydrocarbon fuel pool fires: physical characteristics and
Z A !!
n :di 00 !! thermal emission variation with height, J. Hazard. Mater. 140 (2007)
sT 4 DA ¼ q 9di dA ð n :di Þ 40 ðA:1Þ 280–292.
0 di
[11] J.A. Fay, Model of large pool fires, J. Hazard. Mater. 136 (2006) 219–232.
q00 9di is constant and we need to calculate the right side integral. If [12] H. Chun, K.D. Wehrstedt, I. Vela, A. Schonbucher, Thermal radiation of di-tert-
butyl peroxide pool fires: experimental investigation and CFD simulation,
we transform this integral to spherical coordinates, this integral J. Hazard. Mater. 167 (2009) 105–113.
can be written as [13] K.A. Jensen, J.F. Ripoll, A.A. Wary, D. Joseph, M.E. Hafi, On various modeling
approaches to radiative heat transfer in pool fires, Combust. Flame 148
Z A !! Z A (2007) 263–279.
n :di
dA ¼ cosðyÞcosðjÞ dA ðA:2Þ [14] A.T. Modak, Thermal radiation from pool fires, Combust. Flame 29 (1977)
0 di 0 177–192.
10 G. Keshavarz et al. / Fire Safety Journal 48 (2012) 1–10
[15] L. Orloff, Simplified radiation modeling of pool fires, in: Proceedings of the
Symposium on Combustion, 1981, vol. 18, pp. 549–561.
[16] A. Hamins, M. Klassen, J. Gorge, T. Kashiwagi, Estimate of flame radiation via
a single location measurement in liquid pool fires, Combust. Flame 86 (1991)
223–228.
[17] G. Krishnamoorthy, A comparison of gray and non-gray modeling approaches
to radiative transfer in pool fire simulations, J. Hazard. Mater. 182 (2010)
570–580.
[18] M.B. Kim, Y.J. Jang, J.K. Kim, Burning rate of a pool fire with downward-
directed sprays, Fire Saf. J. 27 (1996) 37–48.
[19] J.C. Chang, C.M. Lin, S.L. Huang, Experimental study on the extinction of liquid
pool fire by water droplet streams and sprays, Fire Saf. J. 42 (2007) 295–309.
[20] B. Chen, S.X. Lu, C.H. Li, Q.S. Kang, V. Lecoustre, Initial fuel temperature effects
on burning rate of pool fires, J. Hazard. Mater. 188 (2011) 369–373.
[21] T.S. Ravigururajan, M.R. Beltran, A model for attenuation of fire radiation
through water droplets, Fire Saf. J. 15 (1989) 171–181.
[22] E. Gramsch, L. Catalan, I. Ormeno, G. Palma, Traffic and seasonal dependence
of the light absorption coefficient in Santiago De Chile, Appl. Opt. 39 (2000)
4895–4901.
[23] C. Doolan, The Effect of Water Mist and Water Spray on Radiative Heat
Transfer for Stored Ordnance, DSTO-TN-0501, Defense Science and Technol-
ogy Organization, 2003.
[24] W.M. Irvine, Infrared optical properties of water and ice spheres, Icarus 8
(1968) 324–360.
[25] S. Hostikka, K. McGrattan, Numerical modeling of radiative heat transfer in
water sprays, Fire Saf. J. 41 (2006) 76–86.
[26] V.M. Zolotarev, B.A. Mikhilov, L.L. Alperovich, S.I. Popov, Dispersion and
absorption of liquid water in the infrared and radio regions of the spectrum,
Opt. Spectrosc. 27 (1969) 430–432.
[27] G.M. Hale, M.R. Querry, Optical constants of water in the 200 nm–200 mm
wavelength region, Appl. Opt. 12 (1973) 555–563.
[28] P. Schiebener, J. Straub, J.M.H.L. Sengers, J.S. Gallagher, Refractive index of
water and steam as functions of wavelength, temperature and density, J.
Phys. Chem. Ref. Data 19 (1990) 677–717.