You are on page 1of 24

MATHEMATICAL MODELLING OF THE HEAVY GAS DISPERSION

M. Markiewicz


The Faculty of Environmental Engineering,
Warsaw University of Technology
Mathematical Modelling of the Heavy Gas Dispersion

Models and Techniques for Health and Environmental Hazard Assessment and Management

280
CONTENTS
1. Introduction............................................................................................................................................... 281
2. Dispersion of heavy gas clouds................................................................................................................. 281
3. Classification of heavy gas dispersion models......................................................................................... 283
3.1. General remarks.................................................................................................................................. 283
3.2. Empirical models ................................................................................................................................ 283
3.3. Engineering models............................................................................................................................. 284
3.3.1. Box models for instantaneous releases ........................................................................................ 285
3.3.2. Uniform or Gaussian plume models ............................................................................................ 287
3.3.3. The generalised plume models .................................................................................................... 288
3.3.4. The integral jet models................................................................................................................ 290
3.3.5. Shallow layer models .................................................................................................................. 293
3.4. Research models.................................................................................................................................. 294
4. Practical application of heavy gas dispersion models ............................................................................ 297
5. Evaluation protocol of heavy gas dispersion models.............................................................................. 297
6. Future directions of the model development........................................................................................... 298
7. Summary.................................................................................................................................................... 298
References.......................................................................................................................................................... 299
Mathematical Modelling of the Heavy Gas Dispersion

Models and Techniques for Health and Environmental Hazard Assessment and Management
281

1. Introduction

Large quantities of hazardous toxic substances are produced, stored or transported in a modern industry. Many of
these gases form clouds heavier then air when accidentally released to the atmosphere. The gases may have a
density greater than that of air for several reasons. These include the high molecular weight (for example
chlorine), low release temperature (for example liquefied natural gas), high storage pressure (for example a
failure of a containment of ammonia and subsequent formation of aerosol) or chemical reactions of the released
substance with water vapour in the atmosphere (the polymerisation of hydrogen fluoride).

The heavy gas clouds have the negative buoyancy. It affects and modifies their behaviour in relation to positively
buoyant or neutrally pollution clouds. The differences include the additional gravity driven flow, wind shear at
the heavy gas cloud interfaces, turbulence dumping, inertia of the released material. Special models are
developed to describe the heavy gas clouds dispersion in the atmospheric air. They are called heavy gas
dispersion models or dense gas dispersion models.

In this publication mathematical modelling of heavy gas clouds is described. The dispersion of heavy gases in the
atmosphere is presented in chapter 2. A classification of heavy gas dispersion models is given in chapter 3. In
chapter 4 an application of models is listed. A need for future work is discussed in chapter 5. In chapter 6 an
evaluation protocol for heavy gas dispersion models is characterised.

2. Dispersion of heavy gas clouds

The releases of heavy gases differ from conventional pollutant releases. The reasons are as follows (Britter, 1989,
1998). The volumes of released gas may be large due to the storage of material in a liquid phase. There is a big
variety of types of emission sources and the specification of emission parameters is usually difficult due to their
time dependence and very diversified source geometry. The releases may consist of different component mixtures
of gases and liquids. There may be heat and/or mass transfer with the ground surface and the ambient air. The
phase changes of the released material during the cloud formation typically take place.

The emission sources can be divided into classes based on different criteria. The release duration, release height,
characteristics of the released material, characteristics of storage conditions and the source scenario are used as
the criteria the most often. With regard to the release duration instantaneous and continuous sources are
distinguished. With regard to the release height elevated and ground level releases are specified. With regard to
the phase of released emissions there are gas phase, liquid phase and two-phase flows. With regard to the release
composition of gases one-component and multi-component releases are known. With regard to the storage
conditions pressurised and non-pressurised releases (typically refrigerated) are specified. There are low velocity
jets, momentum jets, evaporation from the pools or explosion with regard to the source scenario.

To determine a source emission rate or volume of released emissions different input data depending on a source
type are needed. In general physical and chemical properties of the released material, geometry of a source,
characteristics of the ground surface and mitigation measures should be supplied. The estimation of the source
emission rate or the volume of released emissions is realised in an emission model.

The density difference between the released heavy gas and the atmospheric air results in specific physical
processes influencing the heavy gas dispersion in the atmosphere. Major physical processes specific for
negatively buoyant clouds include the gravitational velocity field (the gravitational slumping), wind shear at the
cloud interfaces, turbulence dumping and inertia of the released material (Britter, 1989).

The gravitational velocity field is produced due to horizontal density gradients. It is an additional transport
mechanism to the ambient wind field. It results in heavy gas clouds with increased in horizontal and reduced in
vertical dimensions in comparison to passive pollution clouds.

The gravitational flow causes the shear at the ground cloud interface and at the air cloud interface. It may result
in intermingling of a heavy gas cloud with the surrounding air and eventually in the turbulence generation. This
mechanism becomes important when the wind velocity is small and the self generated velocity large.


Mathematical Modelling of the Heavy Gas Dispersion

Models and Techniques for Health and Environmental Hazard Assessment and Management

282
The stable stratification of heavy gas clouds is connected with vertical density differences. The heavy gas clouds
have negative vertical density gradients. This results in dumping the turbulence and turbulent mixing in the
clouds and in the wind flow over them.

The inertia of the released material directly depends upon the material density. The inertia of heavy gas clouds
depends upon the density difference between the released gas and the ambient air. If this density difference
vanishes the dispersion of heavy gas clouds approaches this of passive clouds.

It is obvious that the density difference between the released material and the atmospheric air is not the only
parameter in delineating the cloud behaviour as the heavy gas or passive dispersion. In addition the volume or
volume flux of the released substance, the parameter characterising the ambient flow such as the wind velocity
and the parameter characterising the source dimension are needed. Combinations of these variables allow to
define source Richardson numbers for continuous and instantaneous releases. The Richardson number Ri
o

represents a ratio of the potential energy due to density excess inside a heavy gas cloud to the kinetic energy due
to the ambient turbulence.

In heavy gas dispersion studies slightly different forms of Ri
o
are suggested in different publications. The most
known are those of Hanna and Drivas (1989), Britter and McQuid (1985). Definitions of the source Richardson
numbers for continuous and instantaneous releases proposed by Hanna and Drivas (1989) are
3
* c
co
a
a co
o
u D
V ) ( g
Ri


= . (2.1)
3
* c
io
a
co
o
u D
V ) ( g
Ri


= , (2.2)
where:
g - the gravitational constant,

co
- the initial cloud density,

a
- the atmospheric air density,
V
co
- the initial volume flow rate for the continuous plume,
D
c
- the specific length scale, (for grounded releases it is equal to the initial cloud width, for elevated releases the
initial diameter of the rapture close to the source or the height of the release further downwind when the cloud
touches the ground are used),
u
* -
the friction velocity,
V
io
- the initial volume of the instantaneous puff.

Britter and McQuid (1988) use slightly different forms of the source Richardson numbers for continuous and
instantaneous releases. They are as follows
3
h c
co
a
a co
o
u D
V ) ( g
Ri


= , (2.3)
2 2
h
c
io
a
a co
o
u D
V ) ( g
Ri


= , (2.4)
where u
h
is the wind velocity at the reference height h.

If the value of Ri
o
is greater then some critical number then the cloud motion is dominated by the internal
buoyancy and the cloud dispersion is considered as the heavy gas dispersion. If the value of Ri
o
is smaller than
some critical number then the cloud motion is dominated by the ambient turbulence and the heavy gas effects are
relatively unimportant. However the critical value of Ri
o
should not be thought of as an absolute number which
separates the heavy gas dispersion from the passive gas dispersion. It should be rather interpreted as an
approximation that only provides a guidance about where these two effects may be roughly separated. In reality
the heavy gas effects gradually become important with the increase of Ri
o
and specific physical processes in
heavy gas clouds are depended on a specific source configuration.

Hanna and Drivas (1989) suggest the critical value of Ri
o
about 50 for grounded releases and assume as the first
approximation the same value for Ri
o
for elevated releases. Britter and McQuid (1985) suggest different values
of the critical source Richardson numbers for continuous and instantaneous releases. In the first case it is 0,15
3

and in the second 0,04.
Mathematical Modelling of the Heavy Gas Dispersion

Models and Techniques for Health and Environmental Hazard Assessment and Management
283

The heavy gas cloud evolution can be divided into several phases depending on a the dispersion mechanism
(figure 2.1). These include, with increasing distance from a source, the source emission phase, initial acceleration
and dilution phase, internal buoyancy dominance phase, transition phase, passive dispersion phase (the ambient
turbulence dominance phase). It is important to notice that in a specific emission release scenario some of the
distinguished regions do not occur. However a sequence of the phases never changes. Evaporation from a pool
serves as an example. In this emission release scenario the cloud slumping regime is omitted.



Figure 2.1. An illustration of different phases in the dispersion of heavy gas clouds.

It has to be notice that heavy gas clouds during their transport and mixing with the surrounding air change their
thermodynamic and chemical properties. The thermodynamic phenomena may include the latent heat exchanges
due to phase changes (evaporation and condensation) in the cloud, the heat exchanges with the surface or due to
the entrainment of air (natural and forced convection) to the cloud, the heat release or absorption due to chemical
reactions.

These processes are included to different degrees in heavy gas dispersion models. Simple models do not account
for any of these processes. They are derived assuming that the clouds do not exchange heat with the environment
and do not generate heat internally through latent processes or chemical reactions. There is an abundance of
models dealing with one or two thermodynamic phenomena. In advanced models the thermodynamic and
chemical processes are covered in thermodynamic and chemical modules separately from calculations of
transport and mixing of the clouds. The thermodynamic behaviour of heavy gas releases of multi-component
mixtures is very complicated and calculations involved can be very cumbersome.

3. Classification of heavy gas dispersion models

3.1. General remarks

The mathematical heavy gas dispersion models can be classified using different criteria. The mathematical
principles, emission source type and model complexity are used as the criteria the most often (Britter, 1989,
Hanna and Drivas, 1996, MEG Report, 1994, Borysiewicz et al., 2000, Borysiewicz and Markowski, 2000,
Markiewicz, 2003, 2004). Based on this last criterion heavy gas dispersion models are divided into three groups.
The models are known as phenomenological (empirical) models, intermediate (engineering) models and
computational fluid dynamic (research) models. The intermediate models include box models for instantaneous
releases, uniform or Gaussian plume models, generalised plume models, integral plume models, shallow layer
models.

3.2. Empirical models

In the empirical models the description of the dispersion of heavy gas clouds is based on series of nomograms or
simple correlations. They are constructed using results of measurements from many field and laboratory
experiments. In these experiments releases from ground level instantaneous and continuous sources were studied.
They concerned the grassy flat terrain, neutral atmospheric stability or slightly unstable conditions. Measured
concentrations were averaged over the period of time of duration from 3 to 10 minutes. To derive the basic
relationships the influence of the atmospheric stability state, surface roughness and averaging time was neglected
Mathematical Modelling of the Heavy Gas Dispersion

Models and Techniques for Health and Environmental Hazard Assessment and Management

284
as the available data did not show the strong influence of these parameters on the dense gas dispersion. The
centreline concentration of the heavy gas is calculated from different relationships for instantaneous and
continuous releases in terms of the gravity constant, density difference between the gas and the ambient air,
release volume or release flow rate, ambient wind velocity.

These models are useful in a routine application as screening models. Examples of these models are described in
the Workbook on the dispersion of dense gases (Britter and McQuid, 1988) and German VDI Guidelines VDI
3783 (1990).

In the Britter and McQuid (1988) model the centreline ground level concentration (C
m
) is evaluated in relation to
the initial concentration (C
o
) as a function of two dimensionless variables: the dimensionless distance and the
source Richardson number. Equations for continuous and instantaneous releases have the forms
( )
|
|

\
|
=
2 5
2 1
2 1 /
h a
/
co a co
/
h co
c
o
m
u
V g
,
) u / V (
x
f
C
C


, (3.1)
( )
|
|

\
|
=
2
3 1
3 1
h a
/
io a co
/
io
i
o
m
u
V g
,
) V (
x
f
C
C


, (3.2)
where:
x - the downwind distance from the source,
V
co
- the release flow rate,
u
h
- the wind velocity at the height h,

co
- the initial density of the released gas,

a
- the density of air,
g - the gravitational constant,
V
io
- the release volume.

The nomograms for both types of releases are given in figure 3.1. The dashed boxes in these figures marked as
the full-scale data region indicate ranges of data covered by field observations. The vertical line marked as the
passive limit indicates the transition from the passive to the stable stratified dispersion regime.

Figure 3.1. Nomograms to calculate a ratio of the mean concentration (C
m
) to the initial concentration (C
o
) in the
heavy gas cloud for continuous and instantaneous releases in the Britter and McQuid model (1989).

3.3. Engineering models

The engineering models can be divided into five groups (MEG Report, 1994): box models for instantaneous
releases, uniform or Gaussian plume models, generalised plume models, integral plume models and shallow layer
X
Vio
1/3
FULL-SCALE
DATA REGION
C /C m 0
0-001
0-002
0-005
0-01
0-05
0-02
0-1
P
A
S
S
I
V
E

L
I
M
I
TInstantaneous releases
2 x 10
0
10
1
10
2
10 10
10
-1 0 1
g( - ) o a
Vio
uref
1/3
1/2
2
a
g( - ) o a
Vco
uref
1/2
2/5
5/2
a
FULL-SCALE
DATA REGION
C /C m 0
0-002
0-005
0-01
0-05
0-02
0-1
10
10
10
2
1
3
10 10 10
-1 0
continuous releases
P
A
S
S
I
V
E

L
I
M
I
T
X
Vco

Mathematical Modelling of the Heavy Gas Dispersion



Models and Techniques for Health and Environmental Hazard Assessment and Management
285
models. These last ones are the most complex models among the engineering models. They differ from other
models of this group in the ability to treat topography effects and in the dimensionality.

3.3.1. Box models for instantaneous releases

The box models for instantaneous releases are used to describe grounded puffs of heavy gases. In these models it
is assumed that the pollution puff forms a uniform cylinder (figure 3.2). The main puff variables such as its
radius, mass and enthalpy are obtained by numerical integration of basic ordinary differential equations with
respect to time. The basic equations represent the puff horizontal spreading, mass and energy conservation. Other
puff variables are calculated from additional equations. The horizontal spreading influencing the cloud radius is
assessed using the gravitational front velocity. The exchange of mass between the puff and the atmospheric air
taking place through the cylinder top and edge is described by entrance velocities. The values of these parameters
depend on the turbulence intensity, difference of densities between the puff and the environment, transport
velocity of the puff. The puff heating due to its contact with the ground and air is introduced straightforwardly.
The puff spreads in still air or is moved downwind. The puff transport velocity is calculated based on the
entrained momentum or on the wind velocity. The concentration averaged over the box volume is calculated as a
ratio of mass of the substance released to the box volume. Assuming empirical similarity profiles in vertical and
horizontal directions the concentration variation in the box can be later reintroduced.


Figure 3.2. A scheme of a box model for heavy gas releases, where H, R are the height and radius of a puff, u
c
is
the puff transport velocity, u
a
is the wind velocity, u
t
, u
e
are the top and edge entrainment velocities of air, u
f
is
the gravitational frontal velocity.

The basic equations in a box model include
the puff horizontal spreading equation
f
u
dt
dR
= , (3.3)
the entrained mass equation
( ) ( )
e a t a
a
u RH u R
dt
dM
2
2
+ = , (3.4)
the energy equation
( ) ( )
2
R E h h
dt
dM
dt
dh
M
f c a
a c
+ = , (3.5)
where:
Mathematical Modelling of the Heavy Gas Dispersion

Models and Techniques for Health and Environmental Hazard Assessment and Management

286
R, H, M, h
c
- the puff radius, height, mass and enthalpy per unit mass respectively,

a
, M
a
, h
a
- the atmospheric air density, entrained mass, and the enthalpy per unit mass,
u
f
- the gravitational frontal velocity,
u
e
, u
t
- the entrainment velocities for the edge and top of the puff respectively,
E
f
- the heat flux transferred to the puff per unit area of the underlying surface.

The equations form a set of three coupled linear differential equations which may be solved numerically for the
independent variables R, M
a
and h.

The gravitational velocity u
f
is usually calculated from the gravity current formula
2 1
1
/
a
a c
f
H ) ( g
c u
|
|

\
|
=


, (3.6)
where
c
- the puff density.

The edge entrainment velocity u
e
is typically scaled with u
f
and given by
f e
u c u
2
= , (3.7)
where c
2
- the constant of the value 0,6<c
2
<0,9.

For the top entrainment velocity generally the correlations of the following form are used
|
|
|
|
|

\
|

+
=
2
3
1
* a
a p
* t
u
H ) ( g
c
u u

, (3.8)
where:
- the von Karman constant equal to 0,4.
c
3
- the constant which value can be deduced from laboratory experiments, for example Britter (1978) suggests a
value of 0,125.

To determine when the transition to passive dispersion occurs it is enough if one of the following criteria is
fulfilled. The difference of density is less then the specified value. The growth of the cloud radius is smaller then
the specific value.

The centre of the puff position equation is
c
u
dt
dx
= , (3.9)
where u
c
- the transport velocity of the puff.

Box models based on this simple approach satisfactorily reproduce many aspects of field and laboratory
experiments. Examples of box models include the GASTAR model (Britter, 1990), the HEGABOX model in the
HGSYSTEM computer package (Witlox, 1994), the IIT Heavy Gas Model I (Mohan et al., 1995), the DENZ-
EDF model (Kaiser and Walker, 1978) and many other models described in works of van Ulden (1987), Fay and
Zemba (1985), Eidsvik (1980), Fay and Ranck (1982), Delvosalle et al. (1993), Cleaver (1994), Kunsch and
Fannelop (1995), Kunsch and Webber (2000), Kumar and coworkers (2003), Webber (1993), Nielsen (1994).

In general in most of box models the flat uniform terrain is assumed. Only in some of these models the dispersion
on slopes (Webber, 2000, Kumar et al., 2003, Webber, 1993, Nielsen, 1994, Britter, 1990) or over the fences and
obstacles (Cleaver, 1994) is described.

Some of the models of this group have been adopted to continuous grounded releases of heavy gases. Then the
plume is simulated by a series of puffs of a rectangular shape.

Mathematical Modelling of the Heavy Gas Dispersion

Models and Techniques for Health and Environmental Hazard Assessment and Management
287
3.3.2. Uniform or Gaussian plume models

The uniform or Gaussian plume models are used for continuous steady state grounded releases of heavy gases.
They are developed in a similar manner as box models for grounded instantaneous releases. All basic phenomena
associated with heave gas releases such as the horizontal spreading, the mass exchange between the plume and
the surrounding air, the plume heating are described by ordinary differential equations similar to the equations in
the box models. However here these equations are integrated with respect to the downwind distance. The main
variables are the average mass flux through the plume cross section, enthalpy flux, plume width. The plume
moves downwind with the wind velocity. The plume cross section is assumed to be a rectangle (figure 3.3). The
uniform or Gaussian profile is adjusted to this shape for concentrations. In uniform profile models the
concentration averaged over a plume cross section is calculated as a ratio of the mass flow rate of the substance
to the volume flow rate of the plume. In case of Gaussian profile models the concentration is allowed to vary in
the cross section and is calculated directly from a Gaussian formula. The dispersion parameters
y
and
z
before
transition are given by b/2,14 and H/2,14 respectively. The passive plume dispersion in the far field is described
by the same Gaussian formula using the dispersion coefficients for passive releases.



Figure 3.3. A scheme of a plume model for heavy gas releases, where E is the emission rate from a source, H, b
are the height and width of a plume cross section respectively, u
c
, u
a
are the plume and wind velocities
respectively, u
f
is the horizontal spreading velocity, u
t
, u
e
are the top and edge entrainment of air velocities
(Kunsch and Fannelop, 1995).

The set of the basic equations in uniform or Gaussian plume models includes
the balance equation of the entrained air mass
( ) ( ) [ ]
t c e c a
a
u bu u Hu
dt
M d
+ = 2
&
, (3.10a)
( ) [ ]
t e a
a
bu u H
dx
M d
+ = 2
&
, (3.10b)
the gravity slumping equation
f
u
dt
db
2 = , (3.11a)
u
u
dx
db
f
1
2 = , (3.11b)
the energy equation
Mathematical Modelling of the Heavy Gas Dispersion

Models and Techniques for Health and Environmental Hazard Assessment and Management

288
( ) b u E h h
dt
M d
dt
dh
M
c f c a
a c
+ =
&
&
, (3.12a)
( ) b E h h
dx
M d
dx
dh
M
f c a
a c
+ =
&
&
, (3.12b)
where:
b, H, M
&
, h
c
- the plume width, height, mass flux, enthalpy per unit mass respectively,
a
M
&
,
a
, h
a
- the entrained air mass flux, density and enthalpy per unit mass respectively,
u
f
- the horizontal spreading velocity,
u
e
, u
t
- the entrained velocities through the edge and top of the plume respectively,
u
c
- the transport velocity of the plume,
E
f
- the heat flux transferred to the plume per unit area of the underlying surface.

The edge and the top entrainment velocities are typically calculated from the relationships given by (3.7) and
(3.8) respectively. Also the horizontal spreading is given by (3.6). The transition between the heavy gas and
passive dispersion is evaluated using the tests for instantaneous releases.
The plume cross section position equation is
c
u
dt
dx
= , (3.13)
where u
c
- the transport plume velocity.

The models described in works of Delvosalle (1993), Fay and Zemba (1986) are examples of uniform plume
models. The IIT heavy gas model (Mohan et al., 1994) is an example of a Gaussian plume model.

3.3.3. The generalised plume models

The generalised plume models can be considered as an extension of uniform or Gaussian plume models in the
sense that the spatial variation of concentrations and other parameters in a plume cross section do not need to
follow the Gaussian or uniform profile. Similarity profiles determined empirically are used to describe them. This
allows to model some of physical processes more realistically. The concentration is expressed in terms of the
centreline ground level concentration, vertical and horizontal dispersion parameters and plume width. These
quantities are determined from a number of basic equations describing the heavy gas mass conservation, air
entrainment, horizontal crosswind gravity spreading and crosswind diffusion. In the far field the models make a
smooth transition to the passive gas dispersion.

The HAGADAS model included into the HGSYSTEM computer package (Witlox, 1994) and the DEGADIS
model (Spicer and Havens, 1984) are examples of generalised plume models. The release of the gas is at a
rectangular ground level source. In both models a plume is represented as being composed of horizontally
homogenous cross sections with Gaussian profile edges in the horizontal direction and the exponential profile in
the vertical direction. The average transport velocity in the plume is determined adopting a power law profile for
the wind velocity. The geometry of the plume cross section with the similarity profiles defining the decay of the
concentration and its effective characteristics are shown in figure 3.4.

Effective plume parameters which include the average velocity (u
eff
), effective half width (B
eff
), effective height
(H
eff
) are given by
y
,
eff
S b dy ) z , y , x ( c
) z , , x ( c
B
5 0
0
2
1
0
1

+ = = , (3.14)
z eff
S dz ) z , y , c ( c
) , y , x ( c
H
|
|

\
|
= =


1 1
0
1
0
, (3.15)
( ) [ ]
( )


|
|

\
|

+
=
|
|

\
|
=

a
z
h eff
h
S
u
/
/
ucdz cdc u
1
1
0
1
0
, (3.16)

where:
Mathematical Modelling of the Heavy Gas Dispersion

Models and Techniques for Health and Environmental Hazard Assessment and Management
289
c(x,y,z) - the concentration of pollutant,
x,y,z - the Cartesian coordinates of the coordinate system located at the middle of the pool in which the OX, OY,
OZ axis correspond to the downwind, crosswind and vertical directions respectively,
b - the half width of the middle part of the cross wind concentration profile,
S
y
, S
z
- the horizontal and vertical dispersion coefficients,
u
h
- the wind velocity measured at a height h
a
,
- the parameter in the wind profile, =1+,
- the Gamma function.




Figure 3.4. A scheme of a generalised plume model for heavy gas releases, where E is the emission rate from a
source, H
eff
, B
eff
are the effective height and effective half width of the plume respectively, u
eff
, u
a
are the effective
plume and wind velocity respectively, u
f
is the horizontal spreading velocity, u
t
, u
e
are the top and edge
entrainment of air velocities, b is the crosswind half width along which the ground level concentration is equal to
c
A
, S
y
is the crosswind dispersion coefficient defining the Gaussian decay in concentration at longer crosswind
distance, S
z
is the vertical dispersion coefficient defining the vertical decay (Witlox, 1994)

The set of the basic equations include
the gas mass conservation equation
( ) [ ]
eff eff eff A z y
,
A
a
h
u H B c S S b c
h
/ u
cudydz E 2
2
1 1 2
1 5 0
0
= |

\
|
+
+
= =
+



, (3.17)
the vertical entrained of the air mass equation
( )
t
h a a
z
eff
eff
u
u h h
S
B
dx
d
B

+
=
(
(

|
|

\
|
+
1
2
1
1
, (3.18)
the horizontal cross wind gravity spreading equation
( )
5 0
4
0
1
,
c
a
eff f
eff
z
gH c u
dt
dB
)
`

=
= =

, (3.19)
the horizontal crosswind diffusion equation
( )
eff
e
y
y
y
B k
dx
dS
S 2 = , (3.20)
where:
Mathematical Modelling of the Heavy Gas Dispersion

Models and Techniques for Health and Environmental Hazard Assessment and Management

290
E - the rate of pollution emission,
u
t
- the top entrainment velocity,
u
h
- the wind velocity measured at a height h
a
,
c
A
- the centreline ground level concentration of pollution,
u
f
- the gravity spreading velocity in the horizontal direction,

c
- the cloud density,

a
- the air density,
g is the gravitational constant,
c
4
- the spreading coefficient,
k
y
e
- a function which relate the crosswind diffusion coefficient to the wind velocity.

The top entrainment velocity u
t
is given by
( )
*
*
t
Ri
u
u

= , 3.21)
where:
- the von Karman constant,
u
*
- the friction velocity,
- the empirical function dependent on the bulk Richardson number Ri
*

( ) ) /( Ri , ) Ri (
,
* *
+ + = 1 8 0 1
5 0
for 0 >
*
Ri , (3.22)
( ) ) /( Ri , ) Ri (
,
*
+ =

1 6 0 1
5 0
for 0 <
*
Ri , (3.23)
( )
2
0
*
eff
a
eff a c
*
u
H
) z (
) H z ( g
Ri
=
=
=


, (3.24)

The set of ordinary differential equations is solved simultaneously to predict the basic solution variables S
z
, S
y
,
c
A
, b provided that closure relations are given.

The concentration c is calculated from the formulas
( )

) x ( S
z
exp c z , y , x c
z
A
for b y < , (3.25)

(
(

) x ( S
z
) x ( S
) x ( b y
exp c ) z , y , x ( c
z y
A
2
for b y > , (3.26)

In the process of updating of the model two additional basic equations can be added to take into account the heat
transfer and water vapour transfer from the substrate into the cloud. This has been realised in the HEGADAS
model (Witlox, 1986) of the fourth version.

3.3.4. The integral jet models

The integral jet models are used to describe continuous, elevated releases of heavy gases. They are based on the
integration of conservation equations of mass, species, downwind and crosswind momentum and energy averaged
over a jet cross section. These equations directly predict jet variables such as the concentration, jet velocity,
radius, enthalpy. In steady state integral models the jet variables are evaluated as a function of the downwind
distance. In more general time dependent models these variables are evaluated as a function of the downwind
distance and time. A jet cross section is assumed to be a circle, ellipse or rectangle. The uniform, Gaussian or
similarity profile is used to describe the space variability of jet variables in a cross section. First the profile is
used to average the jet variables over the jet cross section and to simplify the equations. In the next stage the
profile is used to introduce the spatial variability of the plume variables. A type of the emission source is the
basic difference in the identification of integral models in relation to one dimensional shallow layer models used
for grounded releases. The elevated release may remain the passive elevated jet in the far field or may touch the
ground still being sufficiently dense. The gravity, drag force of the ambient flow and momentum of the entrained
Mathematical Modelling of the Heavy Gas Dispersion

Models and Techniques for Health and Environmental Hazard Assessment and Management
291
air influence the jet path. The entrainment rate in these models is different from that in the grounded release
models. It mainly depends on a velocity shear between the elevated jet and the surrounding air.

Examples of steady state integral plume models include the HMP model (Hoot, Meroney and Peterka, 1973), the
Ooms model (1974), the Khan and Abassi model (1999), the AEROPLUME and HFPLUME models in the
HGSYSTEM package (Witlox and McFarlane, 1994) and the models described by Epstein and coworkers
(1990), Woodward and coworkers (1991, 1995), Tickle and coworkers (1992), Muralidhar and coworkers
(1995). The CLOUD model can follow the steady state or transient conditions (Banerjee et al., 1996). An up to
date review of two phase jet dispersion models is described by Bricard and Friedel (1998).



Figure 3.5. A scheme of an integral jet model for heavy gas releases, where s is the curve linear coordinate of the
jet centreline, is the angle between the jet centreline and the horizontal axis, R is the jet radius, u
c
, u
a
are the jet
and wind velocities, u
e
is the entrainment velocity of air (Khan and Abbasi, 1994).

The set of the basic equations in integral models for steady state jet releases assuming that a jet centreline
remains in the wind gravity takes the form (figure 3.5) (Bricard and Friendel, 1998)
the mixture mass balance equation
E dA u
ds
d
A
c c
=

, (3.27)
the contaminant mass balance equation

=
A
c
dA cu
ds
d
0 , (3.28)
the mixture enthalpy balance equation

=
A
a c c c
Eh dA h u
ds
d
, (3.29)
the mixture momentum balance equation along the OX axis

x a c c
F Eu dA cos u
ds
d
+ =


2
, (3.30)
the mixture momentum balance equation along the OZ axis

=
z c c
F dA sin u
ds
d

2
, (3.31)
the position equations
cos
ds
dx
= , (3.32)
Mathematical Modelling of the Heavy Gas Dispersion

Models and Techniques for Health and Environmental Hazard Assessment and Management

292
sin
dx
dz
= , (3.33)
where:
s - the curve linear coordinate of the jet centreline,
- the angle between the jet centreline and the horizontal axis,
x,y,z - the Cartesian coordiantes such that wind and gravity are oriented along the OX and OZ direction
respectively,
A - the jet cross section area,

c
, u
c
, h - the jet density, velocity and specific enthalpy respectively,
c - the contaminant mass concentration in the jet,
E - the entrainment function,
u
a
, h
a -
the velocity and specific enthalpy of the ambient air respectively,
F
x
, F
y
- the forces which act on the jet in the horizontal and vertical directions respectively.

A system of the ordinary differential equations is solved for u
c
, c, A, , h
c
, x, z (mean values when appropriate)
provided that closure relations are given. For the ground level clouds formed by elevated jets which touch the
ground additional transfer terms with the ground could be introduced in the momentum equations. For releases
which are not in the wind gravity plane additional momentum equation in the OY direction is introduced
(Pattison et al., 1998). The entrainment function E is an important item in the integral jet models because it
controls a rate of dilution of the entrainment. The function is composed of several terms corresponding to
different mechanisms. The jet turbulence, cross-flow perpendicular to the jet axis, atmospheric turbulence are
usually taken into account.

The test to determine when an elevated jet changes into a ground level cloud covers the following criteria. The
lower jet boundary reaches the ground. The jet velocity is close to the ambient wind velocity. The total
entrainment of air to the jet is sufficiently close to the entrainment as a result of vertical mixing at the top of the
cloud. The test to determine when the elevated heavy gas dispersion changes into elevated passive dispersion
includes the following criteria. The jet velocity is close to the ambient wind velocity. The total entrainment of air
to the jet is close to the entrainment corresponding to the atmospheric turbulence.

In the HMP model (Hoot et al., 1973) which is one of the very first models of this type it is assumed that the
distributions of the variables within the jet of the circular shape are uniform at a given distance from the source.
It is also assumed that the specific heats of the jet and air are equal. The model equations are solved analytically
for an upward-pointing dense jet leading to the analytical expressions.
The maximum initial rise (h) is given by
( )
3 1
2
3 1 3 1
2
32 1
2
/
a co o
co o
/
a
co
/
a
o
o
g R
w
u
w
,
R
h
|
|

\
|

|
|

\
|
|
|

\
|
=

, (3.34)
where:
w
o
, R
o
,
co
- the initial velocity, radius and density of the jet,
u
a
- the wind velocity,
g - the gravitational acceleration,

a
- the air density.

The downwind distance at which the centreline of the jet strikes the ground (x
g
) is

( ) [ ]
( ) [ ] { }
2 1
3
2 1
3
3
2 1 2
2
56 0
2 2
/
a co o o a a
/
s
o
a co o
co a o
o
g
gw R u
h
h
R
h
,
g R
u w
R
x

(
(

\
|

+
|
|

\
|
+

=
, (3.35)
where h
s
- the stack height.

The ratios of the maximum concentration (C
m
) to the initial mass concentration (C
o
) at the downwind point where
the maximum initial rise occur and at the point where the centreline strikes the ground are as follows
Mathematical Modelling of the Heavy Gas Dispersion

Models and Techniques for Health and Environmental Hazard Assessment and Management
293
85 1
2
688 1
,
o a
o
o
m
R
h
u
w
,
C
C

|
|

\
|
|
|

\
|
= , (3.36)
95 1
2
2
43 2
.
o
s
a
o
o
m
R
h h
u
w
,
C
C

|
|

\
| +
|
|

\
|
= , (3.37)

3.3.5. Shallow layer models

The one or two dimensional shallow layer models are used for grounded releases. They are based on partial
differential equations describing the principles of conservation of mass, species, momentum and energy averaged
over a cloud depth. This kind of averaging is convenient due to the cloud geometry. Its vertical dimension is
small compared to horizontal dimensions. The pollutant cloud behaviour is described using the variables
changing in one or two dimensions in space and in time. The cloud top is difficult to define in reality and a
vertical concentration distribution is used to define it. Approximations of the shallow layer theory are adapted.
They state that the pressure distribution is hydrostatic within the main body of the cloud and the dispensation is
made only for special processes at the leading edge. The exchange of mass between the cloud and the
atmospheric air is described by the entrainment velocity. The complex topography is introduced quit easily by
adding some terms to the momentum equations. This is an advantage in comparison to the simpler models which
in general are not suitable for complex topography.

Shallow layer models allow for a realistic description of the behaviour of heavy gas clouds. The time to run them
is somewhere between the time to run intermediate models and fluid dynamics models.

The SLAB model, the DISPLAY 1 model are examples of one dimensional shallow layer models. The TWODEE
model (Hankin and Britter, 1999) and the DISPLAY 2 model (Wurtzs et al., 1999) are examples of two
dimensional shallow layer models.

The set of basic equations in one dimensional shallow layer model such as the set from the SLAB model includes
the mass conservation equation
( ) ( )
s s s e a e a
c c c
B W B w H v
x
BH u
t
BH


+ + =

, (3.38)
the vapour conservation equation
( )
BH
B W
H
w
B
v
x
u
t
c
s s s
c
e a
c
e a
c


+ =

1
, (3.39)
the momentum equations
( ) [ ]
H H
u B W
) u u (
H
w
B
v
x
BH
BH
g
x
u
u
t
u
c
x
c
c s s s
c a
e e
c
a a c
c
c
c
c

+ +

2
2
, (3.40)
( )
H BH
v B W
v
H
w
B
v
B
gH
x
v
u
t
v
c
y
c
g s s s
g
e e
c
a
a c
g
c
g

2
|

\
|
+ =

, (3.41)
the energy equation
p c p c
c s s s pn s
c a
e e
p c
pa a
c
c
c
Hc
Q
BH c
) T T ( B W c
) T T (
H
w
B
v
c
c
x
T
u
t
T

+ |

\
|
+ =

, (3.42)
the rate equation for the cloud half-width
e g
v v
x
B
u
t
B
+ =

, (3.43)
where:
H, B - the height and half width of the perpendicular plume cross section,
Mathematical Modelling of the Heavy Gas Dispersion

Models and Techniques for Health and Environmental Hazard Assessment and Management

294

c
, , u
c
, v
g
,

T
c

,
c
p
- the layer averaged cloud density, mass fraction, velocity in the wind direction, cross wind
cloud velocity at the side edges, temperature and specific heat respectively,
w
e
, v
e
- the vertical and horizontal entrainment rate respectively,

a
, T
a
, u
a
, c
pa
- the density, temperature, velocity and specific heat of the atmospheric air,

s
, W
s
, B
s
- the density, velocity, half width of the source respectively,

x
,
y
, Q - the momentum fluxes in the OX and OY directions and the heat flux respectively.

The vertical entrainment rate (w
e
) and the horizontal entrainment rate (v
e
) are given by
( )
( ) ( )H g
u ,
) Ri (
u
w
a s
* a
a s a a
c a *
,
e

=
3 5 0
5 2
, (3.44)
( )
e e
w
B
H
, v |

\
|
=
2
8 1 , (3.45)
where:
- the von Karman constant,
u
*
- the friction velocity,
g - the gravitational constant,

c
,
a
,
s
- the cloud, atmospheric air and source densities respectively,
B, H - the half width and height of the cloud respectively.

The function dependent on the ambient Richardson number Ri
a
is given by
( )
4 1
16 1
/
a a a
Ri ) Ri (

= for 0
a
Ri , (3.46)
a a a
Ri ) Ri ( + =1 for 0 >
a
Ri , (3.47)
L
z
Ri
a
= , (3.48)
where:
z - the height above the ground,
L - the Monin Obukchov length.

The six coupled non-linear partial differential equations of the SLAB model are solved using the finite element
collocation methods based on piecewise polynomials for the spatial discretization technique and standard implicit
methods for the time integration.

3.4. Research models

The research models are the three dimensional models, in which a full set of partial differential equations
depended on time and the three space coordinates describing the principles of conservation of mass, momentum,
energy and substance are solved. These models can be applied to any emission scenario, terrain or meteorological
conditions. The description of physical processes of the heavy gas dispersion is detail and complete. Some of
these models include a concentration fluctuation model.

In the models the turbulence description is usually based on the K theory closure or k- closure. However the k-l
closure, k- closure, SST closure and SSG closure have been also tested. To simplify the model equations the
hydrostatic, anelastic or Boussinesq approximations are introduced. The methods of solution of equations include
the finite difference schemes, finite element methods or finite volume methods.

The FEM3 model (Ermak et al., 1982), the MARIAH model, the MDPG model (Bellasio and Tamponi, 1993),
the HEAVYGAS model (Deaves, 1985), the MERCURE- GL model (Dujim N.J., 1994), the ANDREA model
(Wurtz et al., 1996) and the models described by Sklavounos and Rigas (2004), Pereira and Chen (1995) and
Burman (1998) are the examples of research models.

The set of the basic equations in K-theory turbulence closure model such as the set in the FEM3 model includes
the mass continuity equation
Mathematical Modelling of the Heavy Gas Dispersion

Models and Techniques for Health and Environmental Hazard Assessment and Management
295
( )
t
U
c
c c

=

, (3.49)
the motion equations
( )
( ) ( ) ( )G U K p p U U
t
U
a c c
m
c a c c c
c c

+ + = +

, (3.50)
the mass transport equation of the pollutant vapour
( )
( )
( )
ph
v c
v c
C
c
c
v c c
v c
t
C
C K C U
t
C
|

\
|

+ = +

1
, (3.51)
the mass transport equation of the liquid
ph
v c
l c
C
c
c
l c c
l c
t
) C (
) C K ( C U
t
) C (
|

\
|

+ = +

1
, (3.52)
the energy conservation equation
( ) ( )
( )
ph
v c
p
c
v c
C
p
pa pn
p c
p c
c
t
C
c
L
C K
c
c c
K c
c
U
t
|

\
|

+ = +


1
, (3.53)
where:

c
,
a
- the density of the cloud and the air respectively,
U
c
- the cloud velocity vector,
p, p
a
- the pressure in the cloud and the pressure if the has been no cloud,
G - the gravitational vector,
C
v
, C
l
- the concentration of the hazardous gas (vapour) in the plume and the concentration of the hazardous
liquid in the cloud respectively,
- the potential temperature deviation from the adiabatic atmosphere,
L
c
- the heat of vaporisation of the liquid in the cloud,
c
p
, c
pn
, c
pa
- the specific heats of the cloud mixture, hazardous gas and the air respectively,
K
m
. K
c
, K

-
the specific turbulent coefficient vectors (each K vector has the two horizontal components K
x
, K
y
and
the vertical component K
z
).

The approximation of the state equation, equation defining the potential temperature and equations defining the
K turbulent coefficients are added to close the system of equations.
The state equation has the form
|
|

\
|

+
= =
a
v
g
v *
*
c
M
C
M
C
T R
p
T R
pM
1
, (3.54)
where:
M, M
a
, M
g
- the molecular weights of the cloud mixture, air and vapour respectively,
R
*
- the universal constant,
T - the absolute temperature.

The vertical and horizontal turbulent coefficients are
( ) ( ) [ ]
( ) Ri
l w z u
K
,
* *
z

5 0
2 2
+
= , (3.55)
) Ri (
z u c
K
*
y

5
= , (3.56)
where:
u
*
, w
*
- the friction and convective velocities respectively,
l - the local length scale of the cloud (it is equal approximately to the cloud depth),
z - the height above the ground,
c
5
- the constant,
- the empirical function dependent on the ambient Richardson number R
ia

Mathematical Modelling of the Heavy Gas Dispersion

Models and Techniques for Health and Environmental Hazard Assessment and Management

296
( ) Ri Ri 5 1+ = for Ri0, (3.57)
( ) ( )
4 1
16 1
/
Ri Ri

= for momentum Ri<0, (3.58)
( ) ( )
2 1
16 1
/
Ri Ri

= for energy and species for Ri<0, (3.59)
2 2 2 2
2
05 0
* * c
a c
* *
a *
w u
gl
,
w u
Ri u
Ri
+

+
+
=


, (3.60)
[ ] ( )
( )
|
|

\
|

+
=
3
01 0
z / u
m / K , z / T
T
g
Ri
h
a
a
a
, (3.61)

The value of the horizontal turbulent coefficient K
x
is proportional to the K
z
value.

The main governing equations in the FEM3 model along with those for the submodels are solved numerically
with appropriate initial and boundary conditions. Equations are spatially discretised by the finite element method
in conjunction with the Galerkin method of weighted residuals. The time integration scheme is basically the
explicit forward Euler method except for pressure which must be computed implicitly. The turbulence closure
model based on K theory is also introduced in the MARIAH model and the MDPG model.

The k- turbulence closure belongs to the family of eddy viscosity closures. It introduces two new variables into
the system of conservation equations. These variables are the turbulence kinetic energy (k) and the turbulence
dissipation rate (). These variables are directly calculated from the differential transport equations for the
turbulence kinetic energy and turbulence dissipation rate. It is assumed that the turbulence viscosity is a function
of the turbulence kinetic energy and the turbulence dissipation rate. This closure type has proven to be stable and
numerically robust having the well established predictive capability. It is introduced in the following models the
HEAVYGAS model (Deaves, 1985), the MERCURE model, the model of Pereira and Chen (1995), the model of
Burman (1998).

The k-l turbulence closure is also a kind of the eddy viscosity closure. The turbulent viscosity is assumed to be a
function of the turbulent kinetic energy (k) and a turbulent length scale (l). The turbulent kinetic energy is found
by solving a conservation equation where one of the most important terms is the turbulent energy dissipation. The
dissipation is assumed to be a function of the turbulence kinetic energy and the length scale. This type of closure
is introduced in the ANDREA model. It is interesting to notice that the conservation equations are formulated in
this model in terms of the finite volume method in space and are fully implicit in time. They are solved by means
of the method based on the SIMPLER algorithm. This method involves a transformation of the total mixture mass
equation into a full pressure equation. The resulting linear equations are solved iteratively by a Gauss-Seidel
point iteration method (Wurtz, 1996).

In the modelling system of Rigas and Sklavounos (2004) three turbulence eddy viscosity closure models (the k-
closure, k- closure, k- with a limiter) and one Reynold stress model (SSG) can be used. In the k- closure the
turbulence viscosity is linked to the turbulent kinetic energy and the turbulent frequency (). These two variables
are determined from the turbulence kinetic energy conservation equation and the turbulent frequency
conservation equation. The k- with a limiter closure has been developed to remove a deficiency of both the k-
and k- turbulence closures which do not account for the transport of the turbulent shear stress resulting in the
over prediction of the eddy viscosity. It is based on the same equations as the k- closure but in addition a limiter
to the formulation of the eddy viscosity is introduced. The Reynold stress model is based on transport equations
for all components of the Reynolds stress tensor and the dissipation rate. These turbulence models are generally
flexible but the increased number of transport equations leads to a higher degree of complexity, reduced
robustness and increased computational costs. In the modelling system of Rigas and Sklavounos the finite volume
method is used to discretise the governing equations. The governing partial differential equations are integrated
over all the control volumes using the Gauss divergence theorem to convert volume integrals to surface integrals
and the integral equations are converted to a system of algebraic equations. They are solved iteratively at nodal
points inside each cell aiming at minimisation of the residuals until the prescribed convergence criteria are
satisfied.

The intercomparison results of the MERCURE and ANDREA models show that although a detailed analysis of
the entire set of model output reveals some differences between the two model results they both agree fairly well
with the available observations for all the different comparison performed. This indicate in particular that both
Mathematical Modelling of the Heavy Gas Dispersion

Models and Techniques for Health and Environmental Hazard Assessment and Management
297
the k- and k-l turbulence closure are capable, as implemented in the MERCURE and ANDREA models to
reproduce the heavy cloud behaviour sufficiently well to simulate realistic concentration fields (Dujin, 1995).

The results of the simulations carried out with the modelling system of Sklavounos and Rigas (2004) using the
four turbulent closure modules show good agreement compared with the experimental data. The k- and k- with
a limiter models show improved robustness. The SSG model entailed increase CPU time without significant
enhancement of accuracy of results. The SSG, k- and k- with the limiter models appear to overestimate
maximal concentrations recorded in the trials, whereas the k- model underestimates them.

It is expected that the group of scientific models will be extended in the near future by the development of the
Large scale eddy simulation models suitable for the heavy gases dispersion.

4. Practical application of heavy gas dispersion models

The intermediate and empirical models are used in the West European countries and in the USA in routine
calculations. They are important components of emergency response systems and valuable tool for environmental
impact assessment and risk assessment. Input data needed by theses models are easy to obtain, their idea is
simple and computer costs are low or reasonable.

The fluid dynamic (research) models are usually used as a research tool to get to know better the heavy gas
properties. However they are very attractive their usage is limited due to big computational costs and the input
data requirements.

5. Evaluation protocol of heavy gas dispersion models

The evaluation protocol for heavy gas dispersion model has been produced by the Heavy Gas Dispersion Expert
Group (MEG, 1994). The evaluation is divided into 6 steps. They include the model description (the origin, type,
documentation, etc.), database description (the reference, type, release conditions etc.), scientific assessment (the
completeness of the description of the physical and chemical phenomenon, assumptions made, use of model
constants, solution techniques), user oriented assessment (the user friendliness, guidance and computer
requirements), verification of the code (the software errors), validation of the model by comparing model
predictions with experimental observations.

With respect to the statistical validation of the model any specific parameters have been proposed. These
developed by Hanna and coworkers (1991) are used the most often. The data used in the model evaluation arise
from laboratory and field experiments. The list of experiments on heavy gas dispersion included in the
REDIPHEM database is presented in tab. 5.1. The REDIPHEM database contains easily accessible data. It is the
extension of the Modeller Data Archive of Hanna (1991) as it contains the full time series of meteorological and
concentration data. The Modeller Data Archive contains only time averaged meteorological parameters and
maximum concentration and plume width as a function of distance. The REDIPHEM database focuses on data
from experiments funded by the EC and on some data from USA (Nielsen et al., 1997). The well known Thorney
Island data set is lacking because the original data are not transferred to the needed form. The data from the
extensive experimental project coordinated by the US Petroleum Environmental research Forum (Hanna and
Steinberg, 2001, Snyder, 2001, Robins et al., 2001, Havens et al., 2001, Briggs et al., 2001, Hanna and Chang,
2001) and the data from other projects carried out in EC lately (Donat and Schatzman, 1999, Ayrault et al., 1998,
Davies and Hall, 1996, Papaspyros et al., 1995, Billeter and Fannelop, 1996) are not included in the REDIPHEM
data base for the time being. The data sets from the REDIPHEM database are listed in table 5.1.

Table 5.1 Data sets included in the REDIPHEM database (MEG report, 1994).

Name of the experiment Location/
Source
Phenomena involved References
Burro experiment China Lake CA.,
1980
US Lawrance
Livermore
Laboratory
Atmospheric releases of LNG
(liquid), Spill on water pond, Fast
evaporation
Koopman,
1982
Mathematical Modelling of the Heavy Gas Dispersion

Models and Techniques for Health and Environmental Hazard Assessment and Management

298
Coyote experiment China Lake CA.,
1981
US Lawrance
Livermore
Laboratory
Atmospheric releases of LNG
(liquid), Spill on water pond, Fast
evaporation Cloud ignition
Goldwire et al.,
1981
Desert Tortoise experiment Nevada 1983 US
Lawrance
Livermore
Laboratory
Atmospheric releases Flashing
Ammonia jet, Dense to neutral
clouds, Non-obstructed, Flat
terrain
Goldwire et al.,
1985
Eagle experiment Nevada 1983 US
Lawrance
Livermore
Laboratory
Atmospheric releases Evaporation
of N
2
O
4
, Chemical transformation
to NO
2

,
Flat terrain
Koopman et al.,
1984
FLADIS field experiments Landskrona 1993-
1994
Riso, HydroCare,
FOA, CBDE
Ammonia atmospheric releases,
Flashing jet, Dense to neutral
cloud, Non-obstructed
Nielsen et al.,
1994, Nielsen,
1996


EC Major Technological Hazards-
project BA and FLADIS wind
tunnel experiments
Hamburg
University
SF
6
, SF
6
and air, Continuous and
instantaneous, Flat, various
obstacles, slopes
Schatzmann et
al. (1991,
1993)
CEC MTH project BA field
experiments
Lathen 1988-1989,
TUV, Riso, FOA
Propane, Flashing jet/ cyclone,
Atmospheric releases, Dense
cloud, Flat terrain, Obstructions
(fences)
Heinrich and
Scherwinski,
1990
MT-TNO project BA wind tunnel
experiments
Apeldoorn The
Netherlands 1988-
1990 TNO
Continuous plume, Instantaneous
releases (Thorney Isl. No 17),
Fence
Van Oort and
Builties, 1991
STEP project FLADIS wind
tunnel experiments
Apeldoorn The
Netherlands 1992
TNO
Continuous plume, Many sensor,
locations SF
6
gas
Duijm (1994)
CEC MTH project BA wind
tunnel experiments
Stevenage UK
1988-1991 Warren
Springs Laboratory
Instantaneous isothermal release of
Thorney Island type in 1:100
model scale Variable bulk release
stability Solid fence/crenellated
fence/ no fence Variable fence
height 50-100 repetitions of each
case
Hall et al.,
1991


6. Future directions of the model development

Despite the fact that common work of many institutions mainly from the Western European countries and the
USA has resulted in big advances in the area of heavy gas dispersion during the last decade there are still some
problems, which need the solution or improvement. This includes the modelling of the heavy gas dispersion
under low wind conditions, in stable or unstable conditions, within the chemical complex and in complex terrain.
Special attention should be given to the modelling the multi-component and multi-phase heavy gas releases
(Britter, 1998).

7. Summary

There are many different types of heavy gas dispersion models developed. They include empirical models,
intermediate models and fluid dynamic models. The empirical and intermediate models are important
components of emergency response systems and valuable tools for environmental impact assessment and risk
Mathematical Modelling of the Heavy Gas Dispersion

Models and Techniques for Health and Environmental Hazard Assessment and Management
299
assessment. The fluid dynamics models are usually used as a research tool to get to know better the heavy gas
properties.

The presented classification of heavy gas dispersion models is an idealisation. Some models described in the
literature seem not to fit to any of the categories. There are also computer packages incorporating several gas
dispersion models and then each of models has to be treated independently in this classification.

Despite a big progress within the last decade taking place abroad in the area of the heavy gas dispersion
modelling there is a need for further development of the modelling methods. They should be able to better
describe the more realistic emission scenarios.

References
Ayrault M. (1994) Effects of obstacles on the dispersion of continuous and instantaneous dense gas releases.
STEP-FLADIS report, Lab de Mecanique des fluides et dacoustique, URA 263 CNRS-ECL, BP163, 69131
Ecully, France.
Ayrault M. et al. (1998) Negative buoyancy effects on the dispersion of continuous gas plumes downwind solid
obstacles. Journal of Hazardous Materials, 57, 79-103.
Bartzis J.G. (1991) ANDREA-HF: a three dimensional finite volume code for vapour cloud dispersion in
complex terrain. CEC IRC Ispra Report EUR 13580 EN, Ispra, Italy.
Borysiewicz M., Furtek A., Potempski S. (2000) Guidance on risk assessment methods related to hazardous
installations. Institute of Atomic Energy, Otwock (in Polish).
Borysiewicz M., Markowski A. (2000) Fundamentals of the gas dispersion modelling in the environment.
Technical University of Lodz, Lodz (in Polish).
Bricard P, Friedel L. (1998) Two-phase jet dispersion. Journal of Hazardous Materials, 59, 287-310.
Bellasio R., Tamponi M. (1993) MDPG: a new Eulerian 3D unsteady state model for heavy gas dispersion.
Atmospheric Environment, 28, 1633-1643.
Benerjee S, Martini R., Pattison M.J. (1996) CLOUD: A vapour-aerosol dispersion model accounting for plume
3d motion and heat and mass transfer between phases. Journal of Hazardous Materials, 46, 231-240.
Billeter L., Fannelop T.K. (1996) Concentration measurements in dense isothermal gas clouds with different
starting conditions. Atmospheric Environment, 31, 755-771.
Briggs G.A. et al. (2001) Dense gas vertical diffusion over rough surfaces: results of wind tunnel studies.
Atmospheric Environment, 35, 2265-2284.
Britter R.E. et al. (2001) The passive limit of dispersion in a turbulent boundary layer. Atmospheric Environment,
35.
Britter R.E. (1990) GASTAR user manual, CERC Cambridge, UK.
Britter R.E. (1989) Atmospheric dispersion of dense gases. Annual Review of Fluid Mechanics, 21, 317-344.
Britter R. (1998) Recent research on the dispersion of hazardous materials. Report, University of Cambridge, EC
EUR 18198 EN, Brussels, Belgium.
Britter R.E. (1994) Model evaluation. Materials from the Summer school, Trieste, Italy.
Britter R.E. (1988) A review of some experiments relevant to dense gas dispersion. In : Stably stratified flow and
dense gas dispersion.(ed. Puttock J.S.), Clarendon. Oxford.
Britter R. E., McQuaid J. (1988) Workbook on the dispersion of dense gases. HSE Contract Research Report No
17/1988, Sheffield, UK.
Burman J. (1998) An evaluation of topographic effects on neutral and heavy gas dispersion with CFD model.
Journal of Wind Engineering and Industrial Aerodynamics, 74-76, 515-325.
Builtjes P.H. (1992) Research on continuous and instantaneous heavy gas releases TNO-ME report 92-135, TNO,
Appeldorn, Netherlands.
Mathematical Modelling of the Heavy Gas Dispersion

Models and Techniques for Health and Environmental Hazard Assessment and Management

300
Cleaver R.P. et al. (1994) Further development of a model for dense gas dispersion over real terrain. Journal of
Hazardous Materials, 40, 85-108.
Davies J.K.W., Hall D.J. (1996) An analysis of some replicated wind tunnel experiments on instantaneously
released heavy gas clouds dispersing over solid and cranellated fences. Journal of Hazardous Materials, 49, 311-
328.
Deaves D. M. (1985) Application of advanced turbulence models in determining the structure and dispersion of
heavy gas clouds. Atmospheric Dispersion of Heavy Gases and Small Particles. IUTAM Symposium (eds. G.
Ooms, H. Tennekes) Springer Verlag, Berlin.
Delvosalle C. et al. (1993) Major industrial hazards: The SEVEX project: source terms and dispersion
calculation in complex terrain.
Donat J., Schatzmann M. (1999) Wind tunnel experiments of single phase heavy gas jets released under various
angles into turbulent cross flows. Journal of Wind Engineering and Industrial Aerodynamics, 83, 361-370.
Duijm N.J. (1994) Research on the dispersion of two phase flashing releases- FLADIS. Fladis final report R94-
451, TNO Environmental and Energy Research, Appeldorn, Netherlands.
Eidsvik K.J. (1980) A model for heavy gas dispersion in the atmosphere. Atmospheric Environment, 14, 769-
777.
Epstain M. et al. (1990) A model of dilution of a forced two phase chemical plume in a horizontal wind. Journal
of Loss Prevention Process Industry, 3, 280-290.
Ermak D.L., Chan S.T., Morgan D.L., Morris L.K. (1982) A comparison of dense gas dispersion model
simulations with Burro Series LNG test results. Journal of Hazardous Materials, 129-160.
Fay J.A., Ranck D.A. (1982) Comparison of experiments on dense gas cloud dispersion. Atmospheric
Environment, 17, 239-248.
Fay J.A., Zemba S.G. (1985) Dispersion of initially compact dense gas clouds. Atmospheric Environment, 19,
1257-1261.
Fay J.A., Zemba S.G. (1986) Integral model of dense gas plume dispersion. Atmospheric Environment, 20, 1347-
1354.
Goldwire H.C. et al. (1981) Coyote series data report, LLNL/NWC, LNG spill tests, Lawrence Livermore
National Laboratories, Livermore, California.
Goldwire H.C. et al. (1985) desert Tortoise second data report: 1983 pressurised ammonia spills. UCID-20562,
LLNL/Livermore, CA., USA.
Hall D.J. et al. (1991) repeat variability in instantaneous released heavy gas clouds- some wind tunnel
experiments. Report No. LR 804, (PA), Warren Spring Laboratory, UK.
Hankin R.K.S., Britter R.E. (1999) TWODEE: the Health and Safety Laboratorys shallow layer model for heavy
gas dispersion. Part I: Mathematical basis and physical assumptions. Journal of Hazardous Materials, 211-226.
Hankin R.K.S., Britter R.E. (1999) TWODEE: the Health and safety Laboratorys shallow layer model for heavy
gas dispersion. Part II: Model validation. Journal of Hazardous Materials.
Hanna S.R., Drivas P.J. (1989) Guidelines for use of vapour cloud dispersion models. Center for Chemical
Process Safety, Institute of Chemical Engineers, New York.
Hanna S.R. et al. (1991) Hazard response modelling uncertainty. Evaluation of commonly used hazardous gas
dispersion models. Final report, Sigma Research, Westford, MA, USA.
Hanna S.R. et al. (1993) hazardous gas model evaluation with field observations. Atmospheric Environment,
27A, 15, 2265-2285.
Hanna S.R., Steinberg K.W. (2001) Overview of Petroleum Environmental Research Forum (PERF) dense gas
dispersion modelling project. Atmospheric Environment, 35, 2223-2229.
Hanna S.R., Chang J.C. (2001) Use of the Kit Fox data to analyse dense gas dispersion modelling issue.
Atmospheric Environment, 35, 2231-2242.
Havens J. et al. (2001) Wind tunnel study of air entrainment into two dimensional dense gas plumes at the
Chemical Hazards research Centre. Atmospheric Environment, 35, 2305-2317.
Mathematical Modelling of the Heavy Gas Dispersion

Models and Techniques for Health and Environmental Hazard Assessment and Management
301
Heavy Gas Dispersion Expert Group, Model Evaluation Group Final Report, 1999, Brussels, Belgium.
Heinrich M., Scherwinski R, (1990) research on propane releases under realistic conditions. Determination of gas
concentrations considering obstacles, Report 123U1000780, TUV report,TUV, Germany.
Hoot T.G. et al. (1973) Wind tunnel tests of negatively buoyant plumes. Technical report, Colorado State
University, Fort Collins, CO, USA.
Kaiser G.D., Walker B.C. (1978) Releases of adhydrous ammonia from pressurised containers the importance
of denser than air mixtures. Atmospheric Environment, 12, 2289-2300.
Khan F.I., Abassi S.A. (1999) Modelling and simulation of heavy gas dispersion on the basis of modifications in
plume path theory. Journal of Hazardous Materials, A80, 15-30.
Koopman R.P. (1982) Burro series data report LLNL/NWC, LNG Spill Tests, UICD 19075, Lawrence
Livermore National Laboratories, Livermore, CA.,USA.
Koopman R.P. et al. (1984) Eagle series data report:1983 nitrogen tetroxide spills, UCID-200063, Lawrance
Livermore National Laboratory, Livermore, CA., USA.
Kunsch J.P., Fannelop T.K. (1995) Unsteady heat transfer effects on the spreading and dilution of dense cold
clouds. Journal of Hazardous Materials, 169-193.
Kunsch J.P., Webber D.M. (2000) Simple box model for dense gas dispersion in straight sloping channel.
Journal of Hazardous Material, A75, 29-46.
Kumar A. et al. (2003) Study of the spread of a cold instantaneous heavy gas release with surface heat transfer
and variable entrainment. Journal of Hazardous Materials, B101, 157-177.
Markiewicz M. (2003) A proposition of the methodology of the atmospheric air contamination releases of
heavy gases. Part I and II. Report, Warsaw University of Technology, Warsaw (in Polish).
Markiewicz M. (2004) The fundamentals of air pollution dispersion modelling. Warsaw University of
Technology Office, Warsaw, (in Polish).
Markiewicz M. (2005) Perspectives and application of heavy gas dispersion models in Poland. Monographs of
the Polish Academy of Science, 890-896 (in Polish).
McQuaid (1985, 1987) Special issues of the Journal of Hazardous Materials on Thorney Island experiment.
Mohan M, Panwar T.S. , Singh P.M. (1995) Development of dense gas dispersion model for emergency
preparedness, Atmospheric Environment, 29, 2075-2087.
Muralidhar R. et al. (1995) A two-phase release model for quantifying risk reduction for modified HF alkylation
catalysts. Journal of Hazardous Materials, 44, 141-183.
Nielsen M. (1994) Design of the FLADIS field experiments with dispersion of liquefied ammonia. Draft Report,
Riso National Laboratory, Roskilde, Denmark.
Nielsen M. (1994) Comment on: A model of the motion of a heavy gas cloud released on a uniform slope.
Journal of Hazardous Materials, 48, 251-258.
Nielsen M. (1996) Surface concentrations in the FLADIS field experiments. Report No Riso- I- 995 (EN), Riso
National Laboratory, Roskilde, Denmark.
Ooms et al. (1974) the plume path of vent gases heavier then air. Proceedings of the First International
Symposium on Loss Prevention and Safety Promotion in the Process Industries. (ed. C.H. Birschman), Elsevier
Press.
Papaspyros J.N.E. (1996) Wind tunnel simulation of ammonia gas release transport process. Journal of
Hazardous Materials., 241-252.
Pereira J.C.F., Chen X.Q. (1995) Numerical calculations of unsteady heavy gas dispersion. Journal of Hazardous
Materials, 46, 253-272.
Robins A. et al. (2001) A wind study of dense gas dispersion in a stable boundary layer over a rough surface.
Atmospheric Environment, 35, 2253-2263.
Robins A. et al. (2001) A wind tunnel study of dense gas dispersion in a neutral boundary layer over a rough
surface. Atmospheric Environment, 35, 2243-2252.
Mathematical Modelling of the Heavy Gas Dispersion

Models and Techniques for Health and Environmental Hazard Assessment and Management

302
Schatzmann M. et al. (1991) research on continuous and instantaneous heavy gas clouds. Contribution to project
EV 4T-0021-D to the final report of the joint CEC project. University of Hamburg, Hamburg, Germany.
Schatzmann M. et al. (1993) Experiments with heavy gas jets in laminar and turbulent cross flows. Atmospheric
Environment, 27, 1105-1116.
Sklavounos S., Rigas F. (2004) Validation of turbulence models in heavy gas dispersion over obstacles. Journal
of Hazardous Materials, A108, 9-20.
Snyder W.H. Wind-tunnel study of entrainment in two-dimensional dense gas plumes at the EPAs fluid
modelling facility, Atmospheric Environment, 35, 2285-2304.
Spicer T.O., Havens J.A. (1985) Modelling the Phase I Thorney Island experiments. In: Heavy gas dispersion
trials at Thorney Island. McQuaid J., Proceedings of Symposium held at University of Sheffield, G.B. , 3-5 April,
1984, Elsevier, Amsterdam, 1985.
Tickle G.A. et al. (1992) An integral model for an elevated two phase jet in a cross flow. In: Heat Transfer,
AICHE Symposium Series No.288, 88, 350-356.
Van Ulden A.P. (1987) The heavy gas mixing process in still air at Thorney Island and in the laboratory. Journal
of Hazardous Materials, 16, 411-425.
Van Oort H., Builtjes P.J.H. (1991) Research on continuous and instantaneous heavy gas clouds. MT-TNO wind
tunnel experiments, TNO report 91-026, TNO Division of Technology for Society, TNO, Apeldoorn,
Netherlands.
VDI Guidelines (1990) VDI 3783 Part II Environmental Meteorology, Dispersion of heavy gases (in German).
Webber D.M. (1993) A model of the motion of a heavy gas cloud released on a uniform slope. Journal of
Hazardous Materials, 33, 101-122.
Wurtz J., et al. (1996) A dense vapour dispersion code package for applications in the chemical and process
industry. Journal of Hazardous Materials, 273-284.
Wurtz J. (1993) Shallow layer simulations of dispersion of propane releases with and without momentum in an
obstructed terrain. Journal of Hazardous Materials.
Witlox H.W.M, McFarlane K (1994) Interfacing dispersion models in the HGSYSTEM hazard-assessment
package. Atmospheric Environment, 28, 2947-2962.
Witlox H.W.M. (1994) The HEGADAS model for ground level heavy gas dispersion. Part II. Time dependent
model. Atmospheric Environment, 2933-2946.
Witlox H.W.M. (1994) The HEGADAS model for grounded level heavy gas dispersion. Part I. Steady state
model. Atmospheric Environment, 28, 2917-2932.
Woodward J.L., Papadourakis A. (1995) Modelling of droplet size, evaporation and rainout in a dispersing
aerosol jet. In: Proceedings of the International Conference and Exhibition on Safety, Health and Loss Prevention
in the Oil, Chemical and Process Industries, Singapore.

You might also like