You are on page 1of 24

CCPS 1999 UDM paper - 1 - 07-09-04

-------------------------------------------------------------------------------------------------------------
A unified model for jet, heavy and passive dispersion
including droplet rainout and re-evaporation

by

H.W.M. Witlox and A. Holt
Det Norske Veritas, London, UK


Abstract

The Unified Dispersion Model (UDM) models the dispersion following a ground-level or elevated two-phase
unpressurised or pressurised release. It allows for continuous, instantaneous, constant finite-duration, and
general time-varying releases. It includes a unified model for jet, heavy and passive two-phase dispersion
including possible droplet rainout, pool spreading and re-evaporation. It calculates the phase distribution and
cloud temperature using either a non-equilibrium thermodynamics model, a non-reactive equilibrium model,
or an equilibrium model specific for HF (including effects of polymerisation).

As part of the current work each of the modules in the UDM (passive, jet, and heavy dispersion; non-
equilibrium, equilibrium and HF thermodynamics; steady, instantaneous or finite-duration releases; pool
spreading/evaporation on land or water) has been investigated. The fundamental underlying physics has been
considered in conjunction with a literature review, comparison against wind-tunnel experiments, verification
of the numerical solution against analytical solutions (where possible) and sensitivity analyses. In addition
the model has been compared against predictions by third-party models such as the HGSYSTEM dispersion
model HEGADAS and the pool model GASP.

As a result of this work the tuning present in the original UDM model has largely been eliminated, and model
coefficients are obtained directly from established data in the literature which are largely based on wind-tunnel
experiments. The latest UDM version implemented in the DNV software application PHAST 6.0 represents a
significant revision and extension to all parts of the model. This paper includes an overview of the new
model and a summary of the module verification and validation.

The paper also includes a description of the validation of the overall UDM model against large-scale field
experiments. The validation set of experiments includes both continuous and instantaneous releases. These
experiments address complex phenomena such as aerosol discharges and HF thermodynamics, in addition to
relatively simplistic releases. A subset of these experiments include those evaluated as part of the EEC
programme SMEDIS (Scientific Model Evaluation of Dense Gas Dispersion Models).

1. Introduction

This paper describes the new version of the Unified Dispersion Model (UDM) implemented into the DNV
software package PHAST 6.0. The original version of UDM was developed by Woodward and Cook
1,2
as
new technology in the early nineties. The new UDM 6.0 version represents a significant revision and
extension to all parts of the model. This has been carried out in conjunction with a detailed literature review,
verification and validation of the model.

The UDM models the dispersion following a ground-level or elevated two-phase pressurised release. It
effectively consists of the following linked modules (see Figure 1 and Figure 8):

• jet dispersion
• droplet evaporation and rainout, touchdown
• pool spread and vaporisation
• heavy gas dispersion
• passive dispersion

CCPS 1999 UDM paper - 2 - 07-09-04
-------------------------------------------------------------------------------------------------------------
A single form of concentration profile is used to cover all stages of a release. This allows for anything from a
sharp-edged profile in the initial stages of a jet release through to the diffuse Gaussian profile that would be
expected in the final passive stage of spreading.

The UDM includes the effects of droplet vaporisation using a non-equilibrium model (see Figure 8). Rainout
produces a pool which spreads and vaporises. Vapour is added back into the plume and allowance is made
for this additional vapour flow to vary with time. In addition to the non-equilibrium model, UDM also allows
for an equilibrium model and an equilibrium model specific for HF (including effects of polymerisation).

The UDM allows for vertical variation in ambient wind speed, temperature and pressure. Another feature of
the UDM is possible plume lift-off, where a grounded cloud becomes buoyant and rises into the air. Rising
clouds may be constrained to the mixing layer if it is reached.

The UDM allows for continuous, instantaneous, constant finite-duration, and general time-varying releases.

For the original UDM Cook and Woodward adopted a tuning process, where the tuning coefficients were
obtained by the comparison of UDM results against a relatively large set of ‘tuning’ experiments. This type of
tuning has largely been eliminated as part of the current work. The model coefficients have now been obtained
directly from established data in the literature (based on wind-tunnel experiments), rather than doing UDM
simulations and fitting the UDM results to the experimental data.

Sections 2, 3, and 4 of this paper include a description of the theory, verification and module validation of
the UDM dispersion model, the UDM thermodynamics, and the UDM pool spreading/evaporation model.
Section 5 summarises the validation of the overall UDM model against large-scale field experiments. In
Section 6 the major conclusions are summarised. The reader is referred to the UDM Technical Reference
Manual
3
for details not included in the present paper.

2. Dispersion model
2.1 Concentration profile and cloud geometry

Figure 1 shows the movement of the cloud in the downwind direction. The Cartesian co-ordinates x, y, z
correspond to the downwind, cross-wind (lateral horizontal) and vertical directions, respectively; x=0
corresponds to the point of release, y = 0 to the plume centre-line and z = 0 to ground-level. Furthermore s is the
arc length measured along the plume centre, θ the angle between the plume centre-line and the horizontal, and ζ
the distance from the plume centre-line.

The concentration c is given by a similarity profile c = c(x,y,ζ) as suggested by Webber et al.
4



, exp , exp ,
¦
)
¦
`
¹
¦
¹
¦
´
¦

¦
)
¦
`
¹
¦
¹
¦
´
¦

m
y
h
n
z
v h v o
R
y
= (y)
F
R
= ) (
F
with (y)
F
) (
F
(x)
c
= ) y, c(x,
ζ
ζ ζ ζ (continuous)
( 1 )


2 2
) ( , ; y x x r radius horizontal with (r)
F
) (
F
(t)
c
= t) y, c(x,
cld h v o
+ − = ζ ζ , (instantaneous)
( 2 )

In case of steady-state dispersion, Equation ( 1 ) describes exponential decay of the concentration in y and ζ in
terms of the cross-wind and vertical dispersion coefficients R
y
(x), R
z
(x). Empirical correlations are adopted for
the exponents m, n such that the near-field sharp-edged profile (large value for m) develops into a Gaussian
profile in the far field (m=2). As shown in Figure 1a, the plume cross-section is a circle (radius R
y
=R
z
) during
elevated jet dispersion, a truncated circle during touching down, and a semi-ellipse after touching down. The area
of this plume cross-section, A
cld
(x), is obtained by integration of F
v
(ζ)F
h
(y) over y,ζ . As in many other
dispersion models, the cloud is also characterised by an equivalent ‘effective cloud’ [rectangular cross-section
with area A
cld
, effective half-width W
eff
(x), and effective height H
eff
(x)(1+h
d
) with h
d
=0 for grounded plume and
CCPS 1999 UDM paper - 3 - 07-09-04
-------------------------------------------------------------------------------------------------------------
h
d
=1 for elevated plume] with centroid cloud speed u
cld
, and equivalent top-hat concentration equal to the centre-
line concentration c
o
(x).

In the case of instantaneous dispersion, the cloud moves in the downwind direction with cloud centre at time t
denoted by x=x
cld
(t), y=0, z=z
cld
(t). Equation ( 2 ) describes exponential decay of the concentration in x, y, ζ =z-
z
cld
in terms of the horizontal and vertical dispersion coefficients R
y
(t), R
z
(t), with downwind spreading assumed
to be equal to cross-wind spreading (R
x
=R
y
). As shown in Figure 1b, the cloud is a sphere (radius R
y
=R
z
) during
elevated jet dispersion, a truncated sphere during touching down, and a semi-ellipsoid after touching down. The
volume of this cloud, V
cld
, is determined by integration F
v
(ζ)F
h
(r) over x, y, ζ. The cloud is also characterised by
an equivalent ‘effective cloud’ [cylindrical shape with volume V
cld
, effective horizontal radius W
eff
, and effective
height H
eff
(1+h
d
)] with centroid cloud speed u
cld
, and equivalent top-hat concentration equal to the centre-line
concentration c
o
(t).

2.2 Dispersion variables and equations

The model evaluates the plume variables as a function of downwind distance. The plume variables are given in
the table below {u
a
(z
c
) = wind speed at centroid height z
c
, [u
x
,u
z
]=u
cld
[cos θ, sin θ] = cloud speed }

plume variable symbol unit (continuous) unit (instantaneous)
mass of the cloud m
cld
kg/s kg
excess downwind momentum I
x2
= m
cld
[u
x
- u
a
(z
c
)] kg m/s
2
kg m/s
vertical momentum I
z
= m
cld
u
z
kg m/s
2
kg m/s
downwind position x
cld
m m
vertical position z
cld
m m
heat conduction from substrate q
gnd
J/s J
water evaporated from substrate m
wv
gnd

kg/s kg
cross-wind dispersion coefficient R
y
m m

After initialisation at the point of release, these variables are determined by solving a set of ordinary differential
equations forward in the downwind direction (continuous) or time (instantaneous). These equations express
conservation of mass (air entrainment and water added from substrate), conservation of momentum, relation
between cloud speed and cloud position, a heat-transfer relation, a water-vapour transfer relation, and a cross-
wind spreading equation. These equations are described below (see the UDM Technical Reference Manual
3
for
full details).

• Conservation of total cloud mass


ds
dm
E =
ds
m
d
gnd
wv
tot
cld
+ , (continuous)
( 3 )


dt
dm
E =
dt
m
d
gnd
wv
tot
cld
+ , (instantaneous)
( 4 )

The above equation expresses the change in cloud mass as a result of entrainment of air into the cloud and
(in case of dispersion over water) water-vapour transfer from the substrate. The total air entrainment is
E
tot
(kg/m/s for continuous, and kg/s for instantaneous dispersion).

Air entrainment into a plume may be caused by a range of mechanism (see Figure 2). Jet entrainment
and crosswind entrainment are dominant in the near field after a high-pressure continuous release. During
the jet dispersion phase, the centreline velocity decays until either the heavy gas or the passive dispersion
mechanisms become dominant. For a low-energy release, the jet dispersion mode may never be
dominant. A transition is made to passive dispersion if the cloud density is sufficiently close to the
ambient density (for heavy gas dispersion the Richardson number must be sufficiently small), the cloud
speed is sufficiently close to the ambient speed and the contribution of non-passive entrainment is
sufficiently small. The entrainment contributions are calculated as follows:
CCPS 1999 UDM paper - 4 - 07-09-04
-------------------------------------------------------------------------------------------------------------

1. Jet entrainment E
jet
is caused by turbulence resulting from the difference between the jet speed
and the ambient wind speed:

|
u
-
u
| P e = E
a cld a above jet jet
θ ρ cos in kg/m/s (continuous) ( 5 )

|
u
-
u
| S e = E
a cld a above jet jet
θ ρ cos in kg/s (instantaneous) ( 6 )

where ρ
a
is the ambient density, P
above
the cross-wind perimeter (m) of the continuous plume
above the ground (m), and S
above
the instantaneous cloud surface area above the ground (m
2
).
Equation ( 5 ) for continuous dispersion is the Morton-Taylor-Turner formulation
5
, and Equation
( 6 ) is a generalised version for instantaneous dispersion. Assuming a jet of uniform density
(top-hat profile), Ricou and Spalding
6
determined from experiments α
1
= 2π
0.5
e
jet

0.5
tan(β

) =
0.282, where β

=9.1
o
is the empirical value of the asymptotic half-angle of the jet. Ratios
between 1.5 and 2.0 have found to be quoted in the literature for conversion between top-hat
(averaged) concentrations and maximum concentrations. Since the value of the observed
maximum concentrations was approximately 70% larger, we adopt α
1
= 0.282/1.7 = 0.17.

2. Cross-wind entrainment E
cross
in response to the deflection of the plume by the wind:

| u |
P
= E
a above
a
2 cross
θ ρ
α
sin in kg/m/s (continuous) ( 7 )

| u |
S
= E
a above
a
2 cross
θ ρ
α
sin in kg/s (instantaneous) ( 8 )

Equation ( 7 ) for continuous dispersion is from Morton et al.
5
, and Equation ( 8 ) is a generalised
version for instantaneous dispersion. The value for α
2
adopted by Briggs
7
is 0.6 for a top-hat
profile. Again the value of the maximum concentrations is assumed to be 70% larger, and
therefore we adopt α
2
= 0.6/1.7 = 0.35.

3. Passive entrainment is caused by ambient turbulence; it is present both in the near-field (E
pas
nf
)
and the far-field (E
pas
ff
). For continuous elevated dispersion, the near-field passive entrainment
formulation is taken from McFarlane
11
derived from experiments by Disselhorst
8
, while for
instantaneous dispersion a generalised version is adopted. The far-field passive entrainment is
derived from empirical correlations σ
ya
(x) and σ
za
(x) for the Gaussian cross-wind and vertical
dispersion coefficients as a function of x.

4. Heavy-gas entrainment E
hvy
is included for a grounded heavy-gas plume:


, , ) 2 (
)
Ri
(
u
=
u
W
u
R
W
=
E
*
*
top a eff top
y
gnd
hvy
Φ
(
(
¸
(

¸

κ
ρ (continuous)
( 9 )


{ }
dt
dW
u
A u
+
A u
R
W
=
E
eff
side a
top top side side
y
gnd
hvy
γ ρ =
(
(
¸
(

¸

, , (instantaneous)
( 10 )


Here u
top
is the top-surface entrainment velocity, κ = 0.4 the Von Karman constant, u
*
the friction
velocity and Φ(Ri
*
) the entrainment function of the Richardson number Ri
*
. For instantaneous
dispersion, u
side
is the side entrainment velocity, γ the side entrainment coefficient, A
side
the side
area and A
top
the top area [for grounded plume, A
side
= 2πW
eff
H
eff
, and A
top
= πW
eff
2
]. The
entrainment function Φ is taken from Havens and Spicer
9
for Ri
*
<0 and from Britter
10
for Ri
*
>0.
This ensures the best fit for the top-entrainment velocity against a wide range of experimental
data. Finally W
gnd
is the cross-wind radius of the part of the cloud touching the ground. Thus the
term W
gnd
/R
y
in Equations ( 9 ),( 10 ) ensures that heavy-gas entrainment is gradually phased in
CCPS 1999 UDM paper - 5 - 07-09-04
-------------------------------------------------------------------------------------------------------------
during touching down, i.e. from the elevated plume (W
gnd
/R
y
=0) to the ground-level plume
(W
gnd
/R
y
=1).


• Conservation of excess horizontal and vertical component of momentum

The adopted momentum equations (vector notation) are as follows for continuous dispersion [cloud
area A
cld
= m
cld
/ (ρ
cld
u
cld
), where ρ
cld
is the cloud density] or instantaneous dispersion [cloud volume
V
cld
= m
cld
/ ρ
cld
],


( )
(
¸
(

¸


− +
(
¸
(

¸

+
(
(
(
¸
(

¸

(
(
(
(
¸
(

¸

1
0
0
1
cos
0
sin
g A F F =
ds
dI
ds
dI
a cld cld
ground
drag
ground
impact
z
x2
ρ ρ
θ
θ
, (continuous)

( 11 )


( )
(
¸
(

¸


− +
(
¸
(

¸

+
(
(
(
¸
(

¸

(
(
(
(
¸
(

¸

1
0
0
1
cos
0
sin
g V F F =
dt
dI
dt
dI
a cld cld
ground
drag
ground
impact
z
x2
ρ ρ
θ
θ
, (instantaneous)

( 12 )
The terms in the right-hand side represent forces on the plume. They are respectively:

- the ground impact force F
impact
ground
(N/m or N) resulting from plume collision with the ground.
This force is perpendicular to the plume centre line, and is added during touching down only.
- the horizontal ground drag force F
drag
ground
(N/m or N). This force is added after onset of
touchdown only.
- the vertical buoyancy force (N/m or N). This force is proportional to the gravitational acceleration
g (= 9.81 m
2
/s) and the density difference between the plume and the air.
The vertical momentum equation is not used when the cloud is grounded or capped at the mixing
layer (constant plume height). The formulas for the ground drag and ground impact forces are partly
taken from McFarlane
11
for continuous dispersion, and have been generalised for application to
instantaneous dispersion.

• Horizontal and vertical position:


θ θ =
ds
dz
=
ds
dx
cld cld
sin , cos , (continuous)
( 13 )


θ θ
u
=
u
=
dt
dz

u
=
u
=
dt
dx
cld z
cld
cld x
cld
sin , cos , (instantaneous)
( 14 )

• Rate of heat convection from the substrate

The heat convection from the substrate to the cloud is described by the following differential
equation,


W/m in W Q =
ds
q d
gnd gnd
gnd
, ] 2 [ (continuous)
( 15 )


W in S Q =
dt
q d
gnd gnd
gnd
, (instantaneous)
( 16 )

CCPS 1999 UDM paper - 6 - 07-09-04
-------------------------------------------------------------------------------------------------------------
In case of continuous releases, dq
gnd
/ds

(J/m/s) is the heat transferred from the substrate per second
and per unit of downwind direction and W
gnd
is the half-width of the cloud in contact with the
substrate [see Figure 1a]. In case of instantaneous releases q
gnd
is the total heat (J) transferred from the
substrate to the cloud and S
gnd
is the area of the cloud in contact with the substrate [see Figure 1b]. The
heat conduction flux Q
gnd
(W/m
2
) transferred from the substrate (temperature T
gnd
) to the cloud (vapour
temperature T
vap
) is given by


{ }
vap gnd
f
gnd
vap gnd
f
gnd
n
gnd gnd
T T Q
T T Q Q Q
≤ =
> =
,
, , max


( 17 )

where Q
gnd
n
and Q
gnd
f
are the natural and forced convection flux from the substrate to the vapour cloud
(W/m
2
) derived from expressions by McAdams (1954)
12
and Holman (1981)
13
, respectively.

• Water-vapour transfer from the substrate

Water vapour can be transferred from a water surface into the cloud when the vapour temperature of the
cloud is less than that of the water surface. This has been included in the Unified Dispersion Model
following the approach of the Colenbrander and Puttock described by Witlox
14
which relates the rate of
water vapour pick-up to the rate of heat convection from the water surface:


[ ]
vap gnd
a gnd
cld
p
gnd
vap
w
v gnd
w
v
gnd
wv
T T
P T C
ds
dq
)
T
(
P
- )
T
(
P
5
=
ds
dm
> , (continuous)

( 18 )


[ ]
vap gnd
a gnd
cld
p
gnd
vap
w
v gnd
w
v
gnd
wv
T T
P T C
dt
dq
)
T
(
P
- )
T
(
P
5
=
dt
dm
> , (instantaneous)

( 19 )

where P
v
w
is the saturated vapour pressure of water. If T
gnd
< T
vap
or T
gnd
< 0
o
C (substrate is ice) or if
the cloud is passing over dry ground, dm
wv
gnd
/ds = 0 (continuous) or dm
wv
gnd
/dt = 0 (instantaneous).

• Crosswind spreading

In general cross-wind spreading consists of the following three subsequent phases (see Figure 2).

1. Jet spreading. The cloud is assumed to remain circular until the passive transition or until the
spread rate reduces to the heavy-gas spread rate (the latter can only occur after touchdown), i.e.

z y
R R =

2. Heavy-gas spreading. The heavy spread rate is applied until the passive transition. For
instantaneous dispersion it is given by


[ ] { }
(
¸
(

¸

|
¹
|

\
|
Γ
+ =
m
2
+ 1 =
C

h H ) z z - g

C
C
=
dt
R
d
1/2
m
cld
d eff c a cld
m
E
y
,
) 1 ( ( , 0 max
ρ
ρ ρ

and for continuous dispersion by


[ ] { }
(
¸
(

¸

+ Γ =
+ =
m
C

h H ) z z - g

C u
C
=
dx
R
d
m
cld
d eff c a cld
m x
E
y 1
1 ,
) 1 ( ( , 0 max
ρ
ρ ρ

CCPS 1999 UDM paper - 7 - 07-09-04
-------------------------------------------------------------------------------------------------------------

where C
E
= 1.15 is the cross-wind spreading parameter from experiments by Van Ulden
15
, and Γ
is the gamma function. Note that C
m
= W
eff
/R
y
.

3. Passive spreading. After the passive transition the passive spread rate is applied

dx
d
=
dx
dR ya y σ
5 . 0
2 , (continuous)
dx
d
u =
dt
dR ya
x
y σ
5 . 0
2 , (instantaneous releases)

where σ
ya
(x) is the ambient passive cross-wind dispersion coefficient; σ
ya
increases with
averaging time as a result of wind meander.

2.3 Module verification and validation

The UDM Technical Reference Manual
3
includes a detailed description of the UDM verification for the
near-field elevated (jet) dispersion, the ground-level heavy-gas dispersion and the far-field passive
dispersion. This verification can be summarised as follows.

1. Jet and near-field passive dispersion. For an elevated horizontal continuous jet (of air), the UDM
numerical results are shown to be identical to the results obtained by an analytical solution. Very good
agreement has been obtained against both the Pratte and Baines correlation
16
(no ambient turbulence, i.e.
no near-field passive dispersion) and the Briggs correlation
7
(including ambient turbulence, i.e.
including near-field passive dispersion); see Figure 3 for example results.

A sensitivity analysis has been carried out for a given base-case with parameter variations to the release
height, release speed, release angle and transition criterion.

2. Heavy-gas dispersion. The UDM numerical results are shown to be in identical agreement against an
analytical solution for a 2-D isothermal ground-level plume. The UDM has been validated against the
set of three 2-D wind-tunnel experiments of McQuaid
17
(steady-state ground-level dispersion of CO
2
).
Good agreement was obtained for all experiments. Figure 4 includes the results for McQuaid experiment
3 for three types of UDM simulations, i.e.

(i) E
tot
= E
hvy
[inclusion of heavy entrainment only; too conservative assumption]
(ii) E
tot
= E
hvy
+E
jet
[sum of heavy and ‘jet’ entrainment, with ‘jet’ entrainment resulting from difference
between cloud and (lower) ambient speed at point of release]
(iii) E
tot
= max(E
hvy
,E
jet
) [this may be more appropriate than the above assumption since the ‘heavy’ and
‘jet’ entrainment mechanisms are not independent; this ‘conservative’ assumption is adopted for the
final UDM model]

The new formulation has also been validated against the HTAG
18
experiments 139 and 140 (isothermal
ground-level dispersion of Heavier Than Air Gas) and further verified against HGSYSTEM model
results. Figure 5 includes the following results for experiment 140:

(i) UDM run [with imposed HGSYSTEM wind-speed profile and concentration exponent n]
(ii) HEGADAS run adopting standard passive dispersion coefficient σ
ya
without inclusion of turbulence
collapse of heavy-gas spreading [in line with UDM assumptions]
(iii) HEGADAS run adopting experimentally observed σ
ya
without collapse of gravity spreading
(iv) HEGADAS run adopting experimentally observed σ
ya
with collapse of gravity spreading

By comparing the results for (i) and (ii), it is concluded that UDM and HEGADAS predictions are in
close agreement, if similar assumptions are adopted. This is an important verification of the correct
implementation of the model. From the figure it is also inferred, that future implementation into the
CCPS 1999 UDM paper - 8 - 07-09-04
-------------------------------------------------------------------------------------------------------------
UDM of collapse of gravity spreading may be desirable. Note however that this phenonmenon is
important for a subset of heavy-gas dispersion problems only.

3. Far-field passive dispersion. For purely (far-field) passive continuous dispersion, the UDM numerical
results are shown to be in close agreement with the vertical and crosswind dispersion coefficients and
concentrations obtained from the commonly adopted analytical Gaussian passive dispersion formula.
The same agreement has been obtained for the case of purely (far-field) passive instantaneous
dispersion, while assuming along-wind spreading equal to cross-wind spreading in the analytical profile.

A sensitivity analysis has been carried out for a base-case with parameter variations to the release
height, averaging time, surface roughness length, stability class, release rate and wind speed.

2.4 Finite-duration releases
To model finite-duration releases with a uniform release rate, the UDM allows for the quasi-instantaneous
(QI) model or the finite-duration correction (FDC) model.

The QI model models the initial phase as a continuous source (neglect of downwind gravity spreading and
downwind diffusion). When the cloud width becomes ‘large’ with respect to the cloud length, the cloud is
replaced by an ‘equivalent’ circular cloud, and the subsequent phase is modelled as an ‘instantaneous’ circular
cloud. The disadvantage of the QI model is the abrupt transition (sometimes resulting in severe discontinuities,
e.g. erroneous significant increase in maximum concentration). The QI model can be applied with or without the
‘duration adjustment’, where the duration adjustment applies the effect of averaging time because of time-
dependency of the concentrations (for averaging times larger than release duration). The current duration
adjustment over-estimates this effect downwind of the QI transition.
The FDC model is based on the HGSYSTEM formulation derived from that adopted in the SLAB dispersion
model
19,20
. It has a better scientific basis and is derived from an analytical solution of the Gaussian plume
passive-dispersion equations. It takes the effects of downwind diffusion gradually into account including effects
of both turbulent spread and vertical wind shear. A limitation of this model is however that it is strictly speaking
only applicable to ground-level non-pressurised releases without significant rainout. Moreover it produces
predictions of the maximum (centre-line ground-level) concentrations only. The finite-duration correction
includes the effect of averaging time because of time-dependency of the concentrations. The FDC module has
been verified against the HGSYSTEM/SLAB steady-state results, and shown to lead to finite-duration
corrections virtually identical to the latter programs.

3. Thermodynamics model

UDM invokes the thermodynamics module while solving the dispersion equations in the downwind
direction. The module describes the mixing of the released component with moist air, and may take into
account water-vapour and heat transfer from the substrate to the cloud. The module calculates the phase
distribution [component (vapour, liquid), water (vapour, liquid, ice)], vapour and liquid cloud temperature,
and cloud density. Thus separate water (liquid or ice) and component (liquid) aerosols may form.

The liquid component in the aerosol is considered to consist of spherical droplets and additional droplet
equations may be solved to determine the droplet trajectories, droplet mass and droplet temperature. Rainout
of the liquid component occurs if the droplet size is sufficiently large.

The UDM includes the following types of thermodynamic models:

1. Equilibrium model (no reactions). Thermal equilibrium is assumed, which implies that the same
temperature is adopted for all compounds in the cloud (vapour and liquid). The equilibrium model
determines the phase distribution and the mixture temperature. It is based on a simplified version of the
multi-compound algorithm developed by Witlox for use in HGSYSTEM
21
.

The equilibrium model is tested for mixing of propane with moist air at 20C. Ambient humidity,
propane liquid fraction, propane temperature have been varied. The cooling effect because of component
evaporation and the heating effect because of water condensation is shown. The UDM equilibrium-
CCPS 1999 UDM paper - 9 - 07-09-04
-------------------------------------------------------------------------------------------------------------
model predictions are shown in Figure 6 to be very close with HEGADAS predictions for the case of
mixing of propane vapour/liquid (at –43/42C) with 0%/100% humid air.

2. Equilibrium model (HF). The same temperature is adopted for all compounds in the cloud (vapour and
liquid). The model includes the effect of HF polymerisation and fog formation.

The HF thermodynamics model is based on the HGSYSTEM thermodynamics model
22
. Therefore the
UDM thermodynamic predictions are compared against those of HGSYSTEM, and predictions are
shown to be consistent. This also implies that good agreement is obtained against Schotte’s experiment
23

as is shown by Figure 7. A sensitivity analysis is carried out for mixing of HF with moist air, whereby
both humidity and initial liquid mass fraction have been varied.

3. Non-equilibrium model (no reactions). This model allows the temperature of the droplet (liquid
component) to be different of the temperature of the other compounds in the cloud. The non-equilibrium
model determines the phase distribution of the water and the vapour temperature.

A brief assessment of the UDM droplet model has been carried out. In conjunction with the equilibrium
thermodynamics model the droplet model is used to set the droplet trajectories and the point of rainout only.
In conjunction with the non-equilibrium thermodynamics model, it additionally calculates the droplet mass
and the liquid droplet temperature. The initial drop size is taken as the minimum of the droplet size
calculated by mechanical break-up and flashing break-up. Further work needs to be carried out to further
improve the model for flash calculations and droplet correlations. A limited sensitivity analysis has been
carried out in which droplet trajectories etc. have been compared. It is confirmed that for reducing droplet
size the non-equilibrium model converges to the equilibrium model.

4. Pool spreading and vaporisation

If the droplet reaches the ground, rainout occurs, i.e. removal of the liquid component from the cloud. This
produces a liquid pool which spreads and vaporises (see Figure 8). Vapour is added back into the cloud and
allowance is made for this additional vapour flow to vary with time.

Pool spreading/evaporation.

The UDM source term model PVAP calculates the spreading and vapour flow rate from the pool. Different
models are adopted depending whether the spill is on land or water, and whether it is an instantaneous or a
continuous release. The original version of PVAP was developed by Cook and Woodward
24
.

The pool spreads until it reaches a bund or a minimum pool thickness. The pool may either boil or evaporate
while simultaneously spreading. For spills on land, the model takes into account heat conduction from the
ground, ambient convection form the air, radiation and vapour diffusion. These are usually the main
mechanisms for boiling and evaporation. Solution and possible reaction of the liquid in water are also included
for spills on water, these being important for some chemicals. These effects are modelled numerically,
maintaining mass and heat balances for both boiling and evaporating pools. This allows the pool temperature to
vary as heat is either absorbed by the liquid or lost during evaporation.

The PVAP results were compared by David Webber against the SRD/HSE model GASP for a range of
scenarios with the aim of testing the various sub-modules. Thus further work for improvement was
identified. See the UDM Technical Reference Manual
3
for further details and model validation against
experimental data.

Addition of pool vapour back to the cloud.

For a continuous release, the rate of generation of vapour from the spilt liquid is added to the vapour already
in the cloud to give a total flow rate for the combined source. When the release stops there may then be a
period of vapour generated from the liquid pool alone.

In the case of an instantaneous release the vapour produced by the spilt liquid is added back into the cloud,
so long as part of the cloud still covers the point at which the pool was formed by the rained-out liquid. If the
CCPS 1999 UDM paper - 10 - 07-09-04
-------------------------------------------------------------------------------------------------------------
spilt liquid all evaporates while covered by the cloud then all that is produced is an instantaneous cloud. If
the liquid has not all evaporated then once the upwind edge of the cloud has moved past the pool any
remaining liquid is assumed to form a continuous source of vapour.

5. Validation

The above sections described the verification and the validation against wind-tunnel experiments for the
individual UDM modules. This section is concerned with the validation of the overall model against large-
scale field experiments.

The basis and choice of these experiments stems from both the model evaluation carried out by Hanna et
al.
25
and from the ongoing EEC SMEDIS
26
program (Scientific Model Evaluation of Dense Gas Dispersion
Models). Thus seven sets of continous experiments and one set of instantaneous experiments were selected:

1. Prairie Grass (continuous passive dispersion of sulphur dioxide; 10 experiments).
2. Desert Tortoise (continuous elevated two-phase ammonia jet; 4 experiments)
3. FLADIS (continuous elevated two-phase ammonia jet; 3 experiments)
4. EEC (continuous elevated two-phase propane jet; 2 experiments)
5. Goldfish (continuous elevated two-phase HF jet; 3 experiments)
6. Maplin Sands (continuous evaporation of LNG from pool; 4 experiments)
7. Burro (continuous evaporation of LNG from pool; 2 experiments)
8. Thorney Island (instantaneous unpressurised ground-level release of Freon-12; 9 experiments)

Each of the above experimental sets were statistically evaluated to determine the accuracy and precision of
the UDM predictions with the observed data. Formulas adopted by Hanna et al.
25
were calculated to
calculated the geometric mean bias (under or over-prediction of mean) and mean variance (scatter from
observed data) for each validation run. This was carried out for centre-line concentrations, cloud widths, and
(for the SMEDIS experiments) also off centre-line concentrations.

The overall performance of the UDM in predicting both peak centreline concentration and cloud widths was
found to be good for the above experiments.

- The predictions for the neutrally buoyant Prairie Grass experiments (Figure 9) and the aerosol
releases of Desert Tortoise (Figure 10), FLADIS and EEC were found to be very good.
- Good agreement was also obtained for the Goldfish experiments prior to the passive transition
(Figure 11). A more strict passive transition criterion would improve the results in the far field. This
conclusion also applied for the Maplin Sand experiments. Good resuls were obtained for Burro.
- Good agreement was obtained for the instantaneous heavy-gas Thorney Island experiments (Figure
12).

Following an analysis of the experiments, recommendations for future work may centre upon the
enhancement of the heavy spread formulation to include a gravity collapse criteria, improved modelling of
dispersion from a pool and improved transition to passive dispersion. See the UDM Technical Reference
Manual
3
for further details.

6. Conclusions

Each of the modules in the UDM has been investigated and verified in detail in conjunction with a literature
review and a sensitivity analysis. The modules have been corrected where necessary, validated where
possible, and been compared with similar third-party sofware applications. This has been carried out in
extensive detail for the entire basic continuous model [phases of dispersion (passive, jet, heavy), equilibrium
thermodynamics without rainout] and to some extent also the instantaneous model [far-field passive
dispersion, ground-level heavy-gas dispersion]. Following this work the UDM comparison against large-
scale experiments improved considerably, despite the elimination of tuning coefficients.

CCPS 1999 UDM paper - 11 - 07-09-04
-------------------------------------------------------------------------------------------------------------
A detailed assessment and limited corrections have been carried out for the transition to passive, finite-
duration releases, HF thermodynamics, and heat/water transfer from the substrate.

A brief assessment has been carried out for droplet modelling, pool spreading/evaporation, the link between
pool and dispersion model, pressurised instantaneous expansion, lift-off and mixing-layer logic, and time-
dependent releases. A more detailed assessment is to be carried out for these areas.

The key differences resulting from the modelling changes between the old UDM model (PHAST 5.20) and
the new UDM model (PHAST 6.0) are as follows:

- larger concentrations for far-field passive dispersion [corrections to passive dispersion logic]
- lower concentrations for near-field elevated jet dispersion [increased jet and cross-wind entrainment
coefficients, added near-field dispersion]
- larger ground-level heavy-gas spreading (wider clouds) [increased cross-wind spreading parameter]
- often increased water aerosol formation and therefore hotter clouds [new thermodynamics models]

The key advantages of the new UDM model with respect to other typical dispersion models can be
summarised as follows:

(a) a single model for the entire dispersion regime from the point of release to the far-field dispersion
including possible rainout and pool re-evaporation; this eliminates discontinuities and matching
problems

(b) a very extensive verification and validation to ensure that the model shows the correct behaviour and
produces accurate predictions

(c) integration within the user-friendly and well-established DNV consequence-modelling package PHAST
and the risk-analysis package SAFETI. This enables plotting, linking with discharge/fire/explosion
models, toxic/flammable impact and risk calculations.

(d) rigid procedures and control using ISO9001/TickIT quality standards

CCPS 1999 UDM paper - 12 - 07-09-04
-------------------------------------------------------------------------------------------------------------



x
z
TOUCHING
DOWN
ELEVATED
PLUME
z
cld
θ θθ θ
ζ ζζ ζ
y
GROUND-LEVEL
PLUME
circular
cross-section
(R
y
= R
z
)
truncated
cross-section
(R
y
> R
z
)
semi-elliptic
cross-section
(R
y
> R
z
)
s




(a) continuous dispersion


z
y
upper cloud envelope
x
TOUCHING
DOWN
ELEVATED
DISPERSION
GROUND-LEVEL
DISPERSION
spherical cloud
truncated cloud
(circular ground surface)
semi-ellipsoid cloud
(circular ground surface)
ENERGETIC
EXPANSION
PHASE
lower cloud envelope


(b) instantaneous dispersion



Figure 1. UDM cloud geometry


CCPS 1999 UDM paper - 13 - 07-09-04
-------------------------------------------------------------------------------------------------------------
Z
X
JET(elevated)
TRANSITION PASSIVE
(a) elevated jet/plume (no touching down)
circular spread
E
jet
+ E
cross
+ E
pas
nf
phase in passive spread and
entrainment
passive spread (d
σ
ya
/dx )
passive entrainment (E
pas
ff
)


Z
X
HEAVY
(touching down)
TRANSITION PASSIVE
(b) jet/plume becomes passive during touching down
JET
(elevated)
JET
(touching down)
circular spread
Ejet+ Ecross+ Epas
nf phase in passive spread and
entrainment
d
σ
ya/dx
Epas
ff
circular spread
max(Ejet+ Ecross,Ehvy)
+ Epas
nf
heavy spread
max(Ejet+ Ecross,Ehvy)
+ Epas
nf

Z
X
JET
(ground level)
TRANSITION PASSIVE
(c) jet/plume become passive after touch down
JET
(elevated)
JET
(touching down)
HEAVY
(ground level)
circular spread
Ejet+ Ecross+ Epas
nf
phase in passive
spread and
entrainment
d
σ
ya/dx
Epas
ff
circular spread
max(Ejet+ Ecross,Ehvy)
+ Epas
nf
circular spread
max(Ejet,Ehvy)
heavy spread
Ehvy



Figure 2. Phases in UDM cloud dispersion for range of scenarios
The figures indicate for each phase the type of spreading (circular jet, heavy or passive)
and the mechanism of entrainment (E
jet
= jet; E
cross
= cross-wind; E
pas
ff
= near-field
elevated passive, E
hvy
= ground-level heavy, E
pas
ff
= far-field passive). Along the transition
zone the near-field spread/entrainment are phased out and the far-field spread/entrainment
are phased in.
CCPS 1999 UDM paper - 14 - 07-09-04
-------------------------------------------------------------------------------------------------------------
0.1
1
10
0.1 1 10 100
dimensionless plume arclength
d
i
m
e
n
s
i
o
n
l
e
s
s

p
l
u
m
e

r
i
s
e
Pratte and Baines
UDM (no passive)

(a) case of no ambient turbulence; comparison against Pratte and Baines correlation
(jet/ambient velocity ratio = 13)
1
10
0.1 1 10 100
downwind distance (m)
c
e
n
t
r
e
-
l
i
n
e

h
e
i
g
h
t

(
m
)
Briggs formulae
UDM (no passive)
UDM (with near-field passive)


(b) case of ambient turbulence; comparison against Briggs plume rise formula
(jet/ambient velocity ratio = 4.6)

Figure 3. Comparison of UDM results against plume-rise correlations for
vertical neutrally buoyant jet
CCPS 1999 UDM paper - 15 - 07-09-04
-------------------------------------------------------------------------------------------------------------

10
100
1.00E-01 1.00E+00 1.00E+01
downwind distance (m)
g
r
o
u
n
d
l
e
v
e
l

c
o
n
c
e
n
t
r
a
t
i
o
n

(
m
o
l
%
)
UDM max(Jet, Heavy)
McQuaid Experimental data
UDM (Jet included)
UDM (No jet)

(a) ground-level concentration c
ov
(mole %)

0.01
0.1
1.00E-01 1.00E+00 1.00E+01
downwind distance (m)
c
l
o
u
d

d
e
p
t
h

(
m
)
UDM max(Jet, Heavy)
McQuaid Experimental data
UDM (Jet included)
UDM (No jet)
(b) height H
1/2
(m) at which concentration is halved

Figure 4. McQuaid experiment 3; experimental data and UDM predictions
assuming (i) E
tot
= E
hvy
, (ii) E
tot
= E
hvy
+E
jet
, (iii) E
tot
= max(E
hvy
,E
jet
)

CCPS 1999 UDM paper - 16 - 07-09-04
-------------------------------------------------------------------------------------------------------------
0
20
40
60
80
100
120
140
160
180
200
0 80 160 240 320 400 480 560 640
downwind distance (m)
c
l
o
u
d

h
a
l
f
-
w
i
d
t
h

(
m
)
UDM effective half width
HGSYSTEM (experimental
SIGMAY, gravity collapse)
HGSYSTEM (standard SIGMAY,
gravity collapse)
HGSYSTEM (standard SIGMAY,
no gravit y collapse)


(a) effective cloud half-width W
eff


(b) cloud half-width (SMEDIS definition)

0 . 1
1
1 0
10 0
0 8 0 1 60 2 40 32 0 40 0 48 0 56 0 64 0
d o w n w i n d d i s ta n ce ( m )
c
o
n
c
e
n
t
r
a
t
i
o
n

(
m
o
l
e

%
)
U D M ( Z = 0, s t an da r d S I G M A Y )
e xp er i m e nt a l da t a
H G S Y S T E M (e xpe r im en t a l S I G M A Y , gr a vit y c ol l ap s e
H G S Y S T E M (s t a nd ar d S I G M A Y , g ra vi t y c ol l ap s e)
H G S Y S T E M (s t a nd ar d S I G M A Y , n o g ra vi t y c o ll a ps e )

(c) centre-line ground-level concentration
Figure 5. HTAG experiment 140; experimental data and UDM/HEGADAS predictions:
(i) UDM [use HEGADAS wind-speed profile and value for n], (ii) HEGADAS
(standard σ σσ σ
ya
, no gravity collapse), (iii) HEGADAS (standard σ σσ σ
ya
, gravity collapse),
(iv) HEGADAS (experimental σ σσ σ
ya
, gravity collapse)
0
20
40
60
80
100
120
140
160
180
200
0 80 160 240 320 400 480 560 640
downwind di stance (m)
c
l
o
u
d

h
a
l
f
-
w
i
d
t
h

(
m
)
UDM width (SMEDI S definition)
experimental data
HGSYSTEM (experimental SIGMAY, gravity collapse)
HGSYSTEM (standard SIGMAY, gravity collapse)
HGSYSTEM (standard SIMAY, no gravity collapse)
CCPS 1999 UDM paper - 17 - 07-09-04
-------------------------------------------------------------------------------------------------------------
-45
-35
-25
-15
-5
5
15
25
0 20 40 60 80 100
% propane mole fraction
t
e
m
p
e
r
a
t
u
r
e

(
C
)
HEGADAS - 0 % humidity
HEGADAS - 100 % humidity
UDM - 0% humidity
UDM - 100% humidity

(a) mixing of propane vapour (-42C) with air (20C)

-80
-70
-60
-50
-40
-30
-20
-10
0
10
20
0 10 20 30 40 50 60 70 80 90 100
% propane mole fraction
t
e
m
p
e
r
a
t
u
r
e

(
C
)
HEGADAS - 0 % humidity
HEGADAS - 70 % humidity
HEGADAS - 100 % humidity
UDM - 0% humidity
UDM - 70% humidity
UDM - 100% humidity


(b) mixing of propane liquid (-43C) with air (20C)

Figure 6. Temperature predictions by HEGADAS and UDM thermodynamics
equilibrium models for mixing of propane (vapour or liquid) with air
(dry or humid)
CCPS 1999 UDM paper - 18 - 07-09-04
-------------------------------------------------------------------------------------------------------------
273.15
278.15
283.15
288.15
293.15
298.15
303.15
308.15
313.15
0.001 0.01 0.1 1
total HF mole fraction
c
l
o
u
d

t
e
m
p
e
r
a
t
u
r
e
UDM
HEGADAS
experiment

Figure 7. Mixing at 26C of HF vapour with moist air (humidity = 50%);
HEGADAS and UDM thermodynamic test-bed predictions


spreading
evaporating
liquid pool
droplet
trajectory
(flashing) two-phase
discharge from
pipe/vessel
vapour-plume
centre-line
point of rainout


Figure 8. Droplet evaporation, rainout, and pool spreading/evaporation.
CCPS 1999 UDM paper - 19 - 07-09-04
-------------------------------------------------------------------------------------------------------------

(a) mol % concentration



(b) cloud dimensions


Figure 9. UDM validation against Prairie Grass experiment 8 (passive dispersion)
Prairie Grass 8 - Elevated release of Sulfur Dioxide
0.00001
0.0001
0.001
0.01
0.1
1
10
100
1 10 100 1000 10000
downwind distance (m)
c
o
n
c
e
n
t
r
a
t
i
o
n

(
m
o
l

%
)
centre-line
at z =0
experimental data
Prairie Grass 8 - Elevated release of Sulfur Dioxide
0.1
1
10
100
1000
1 10 100 1000 10000
downwind distance (m)
c
l
o
u
d

d
i
m
e
n
s
i
o
n

(
m
)
half width
depth
Experimental width
CCPS 1999 UDM paper - 20 - 07-09-04
-------------------------------------------------------------------------------------------------------------
Desert Tortoise 2 - Ammonia Aerosol
0.001
0.01
0.1
1
10
100
1 10 100 1000 10000
downwind distance (m)
c
o
n
c
e
n
t
r
a
t
i
o
n

(
m
o
l

%
)
centre-line
at z =1
experimental data

(a) mol % concentration

Desert Tortoise 2 - Ammonia Aerosol
0.1
1
10
100
1000
1 10 100 1000 10000
downwind distance (m)
c
l
o
u
d

d
i
m
e
n
s
i
o
n

(
m
)
half width
depth
Experimental width

(b) cloud dimensions

Figure 10. UDM validation against Desert Tortoise 2 (two-phase ammonia jet)
CCPS 1999 UDM paper - 21 - 07-09-04
-------------------------------------------------------------------------------------------------------------


(a) maximum concentration versus downwind distance

GoldFish3 - Including heat transfer
-2.00E+01
-1.00E+01
0.00E+00
1.00E+01
2.00E+01
3.00E+01
4.00E+01
1.00E+00 1.00E+01 1.00E+02 1.00E+03 1.00E+04
downwind distance (m)
t
e
m
p
e
r
a
t
u
r
e

(
K
)
experiment
HFPLUME
HEGADAS
UDM Module test


(b) temperature versus downwind distance


Figure 11. Goldfish 3 simulations for HFPLUME/HEGADAS (including heat
transfer) and UDM
GoldFish3 - Elevated HF aerosol
1.00E-03
1.00E-02
1.00E-01
1.00E+00
1.00E+01
1.00E+02
1 10 100 1000 10000
downwind distance (m)
c
o
n
c
e
n
t
r
a
t
i
o
n

(
m
o
l

%
)
experiment
HFPLUME
HEGADAS
HEGADAS (no heat transfer)
HEGADAS (exp. SY)
UDM module test
CCPS 1999 UDM paper - 22 - 07-09-04
-------------------------------------------------------------------------------------------------------------
Thorney Island 8 - Instantaneous release of Nitrogen and Freon
0.0001
0.001
0.01
0.1
1
10
100
1 10 100 1000 10000
downwind distance (m)
c
o
n
c
e
n
t
r
a
t
i
o
n

(
m
o
l

%
)
centre-line
at z =0.4
experimental data



Figure 12. UDM validation against Thorney Island 8 (instantaneous heavy-gas
dispersion)


Nomenclature

A
cld
cross sectional area of continuous cloud, m
2

A
side
effective side area of instantaneous plume, m
2

A
top
effective top area of instantaneous plume, m
2

c concentration, kg of component /m
3

c
o
centre-line concentration, kg of component / m
3

E
cross
cross-wind entrainment rate, kg/s or kg/m/s
E
hvy
dense gas entrainment rate, kg/s or kg/m/s
E
jet
jet (high-momentum) entrainment rate, kg/s or kg/m/s
E
pas
nf
near-field passive dispersion entrainment rate, kg/s or kg/m/s
E
pas
ff
far-field passive dispersion entrainment rate, kg/s or kg/m/s
E
tot
total dispersion entrainment rate, kg/s or kg/m/s
F
drag
air
airborne drag force, N/m or N
F
drag
ground
ground drag force, N/m or N
F
h
(x) horizontal distribution function for concentration (-)
F
v
(ζ) vertical distribution function for concentration (-)
H
eff
effective height of cloud after full touchdown, m [height prior to full touchdown = H
eff
(1+h
d
)]
h
d
fraction of bottom half of cloud which is above the ground [h
d
=0,1

for grounded,elevated cloud]
I
x2
horizontal plume momentum in excess to ambient momentum [I
x2
= I
x
– m
cld
u
a
], kg m/s or kg/m/s
2

I
z
vertical component of plume momentum, kg m/s or kg/m/s
2

m exponent of horizontal distribution function for concentration (-)
m
cld
mass in plume (instantaneous release, kg) or mass rate in plume (continuous release, kg/s)
m
wv
gnd
water-vapour added from the substrate, kg or kg/s
n exponent of vertical distribution function for concentration (-)
P
above
perimeter length of jet, m
P
v
w
(T) saturated vapour pressure as function of temperature T (K) for water, Pa
CCPS 1999 UDM paper - 23 - 07-09-04
-------------------------------------------------------------------------------------------------------------
q
gnd
heat transfer rate from ground to cloud, J or J/s
R
y
term in cross-wind concentration profile, m [R
y
= R
y
(x) = 2
1/2
σ
y
(x)]
R
z
term in vertical concentration profile, m [R
z
= R
z
(x) 2
1/2
σ
z
(x)]
Ri
*
layer Richardson number, (-)
s arclength along centre-line of the plume, m
S
gnd
footprint area for instantaneous plume, m
2

t time since onset of release, s
T
a
ambient temperature, K
T
gnd
substrate temperature, K
T
vap
temperature of vapour phase of the cloud, K
u
*
friction velocity for cloud, m/s
u
a
ambient wind-speed, u
a
= u
a
(z), m/s
u
cld
total cloud speed, m/s
u
ref
value of ambient windspeed u
a
at reference height z = z
ref
, m/s
u
side
entrainment velocity through sides of plume, m/s
u
top
entrainment velocity through top of plume, m/s
u
x
,u
z
horiontal and vertial component of cloud speed u
cld
, m/s
V
cld
volume of cloud, m
3

W
eff
effective half width of plume, m
W
gnd
footprint half-width for continuous plume, m
x horizontal downwind distance, m
x
cld
horizontal downwind position of cloud, m
y crosswind distance, m
z vertical height, m
z
cld
cloud centre-line height above ground, m
z
o
surface roughness length, m
z
c
height above ground of cloud centroid, m
z
ref
reference height, m

Greek letters
α
1
, α
2
jet- and cross-wind entrainment coeficient , (-)
γ side-entrainment coefficient for instantaneous heavy-gas dispersion, (-)
Γ Gamma function (-)
θ angle to horizontal of plume, rad; θ = 0 corresponds to a horizontal plume (in downwind x-
direction), while θ = π/2 corresponds to a vertical upwards plume (in z-direction)
ζ distance perpendical to plume centre-line, m
κ Von Karman constant, κ = 0.4 (-)
ρ
cld
density of plume, kg/m
3

ρ
a
density of ambient air, kg/m
3

σ
ya
standard empirical correlation for passive crosswind dispersion coefficient, m
σ
za
standard empirical correlation for vertical crosswind dispersion coefficient, m


Acknowledgements

The authors gratefully acknowledge the feedback from Torstein K. Fanneløp, David M. Webber, John L.
Woodward, and Staale Selmer-Olsen, who reviewed the UDM Technical Reference Manual, on which the
present paper is based.


References

1
Woodward, J.L, Cook, J., and Papadourakis, A., “Modelling and validation of a dispersing aerosol jet”,
Journal of hazardous materials 44, pp. 185-207 (1995)
2
Woodward, J.L, and Papadourakis, A., “Reassessment and reevaluation of rainout and drop size correlation
for an aerosol jet”, Journal of hazardous materials 44, pp. 209-230 (1995)
3
Witlox, H.W.M., and Holt, A., “Unified Dispersion Model – Technical Reference Manual”, UDM Version
6.0, June 1999, Det Norske Veritas, London (1999)
CCPS 1999 UDM paper - 24 - 07-09-04
-------------------------------------------------------------------------------------------------------------

4
Webber, D.M., S.J. Jones, G.A. Tickle and T. Wren, “A model of a dispersing dense gas cloud and the
computer implementation D*R*I*F*T. I: Near instantaneous release.” SRD Report SRD/HSE R586 April
1992, and “. . . II: Steady continuous releases.” SRD Report SRD/HSE R587 April 1992
5
Morton, B.R., Taylor, G.I., and Turner, J.S.,”Turbulent gravitational convection from maintained and
instantaneous sources”, Proc. R.Soc., Series A, 234, 1 (1956)
6
Ricou, F.P. and Spalding, D.B., “Measurements of entrainment by aximsymmetrical turbulent jets”, J. Fluid
Mech. 11 (1), pp. 21-31 (1961)
7
Briggs, G.A., “Plume rise and buoyancy effects”, Ch. 8 (pp. 327-366) in Randerson, D. (ed.), “Atmospheric
Science and Power Production”, Technical Information Center, US Department of Energy, Report
DOE/TIC-27601 (1984)
8
Disselhorst, J.H.M., “The incorporation of atmospheric turbulence in the KSLA Plumepath program”, Shell
Internationale Research Maatschappij B.V., The Hague, AMGR.84.059 (1984)
9
Havens, J., and Spicer, T.O.,“Development of an atmospheric dispersion model for heavier-than-air
mixtures” (DEGADIS model), Vols. 1- 3, University of Arkansas (1985)
10
Britter, R.E., “A review of mixing experiments relevant to dense gas dispersion”, IMA Conference on
stably stratified flow and dense gase dispersion “, April 1986, Chester, Oxford University Press (1988)
11
McFarlane, K., “Development of plume and jet release models”, International Conference and workshop
on modelling and mitigating the accidental releases of hazardous materials, AIChE, CCPS, New Orleans,
LA, May 20-24, pp. 657-688 (1991)
12
McAdams, W.H., “Heat Transmission”, McGraw-Hill (1954)
13
Holman, J.I., “Heat transfer”, 5
th
Ed., McGrawHill (1981)
14
Witlox, H.W.M., “Technical description of the heavy-gas-dispersion program HEGADAS”, Report
TNER.93.032, Thornton Research Centre, Shell Research, Chester, England (1993)
15
Van Ulden, A.P., “A new bulk model for dense gas dispersion: two-dimensional spread in still air, in
“Atmospheric dispersion of heavy gases and small particles” (Ooms, G. and Tennekes, H., eds.), pp. 419-
440, Springer-Verlag, Berlin (1984)
16
Pratte, B.D., and Baines, W.D., “Profiles for the round turbulent jet in a cross flow”, Proc. ASCE, HY
(Journal of the Hydraulics Division), Vol. 6, pp. 53-64 (1967)
17
McQuaid, J., “Some experiments on the structure of stably stratified shear flows”, Technical Paper P21,
Safety in Mines Research Establishment, Sheffield, UK (1976)
18
Petersen, R.L., and Ratcliff, M.A., “Effect of homogeneous and heterogeneous surface roughness on
HTAG dispersion”, CPP Incorporated, Colorado. Contract for API, Draft Report CPP-87-0417 (1988)
19
Witlox, H.W.M., McFarlane, K., Rees, F.J., and Puttock, J.S., “Development and validation of
atmospheric dispersion models for ideal gases and hydrogen fluoride”, Part II: HGSYSTEM program user’s
manual, Report RKER.90.016, Thornton Research Centre, Shell Research, Chester, England (1990)
20
Ermak, D.L., “Unpublished notes on downwind spreading formulation in finite-duration release version of
SLAB”, Lawrence Livermore National Laboratorym, California (1986)
21
Witlox, H.W.M, Two-phase thermodynamics model for mixing of a non-reactive multi-compound
pollutant with moist air, Report TNER.93.022, Thornton Research Centre, Shell Research, Chester, England
(1993)
22
Witlox, H.W.M., “Thermodynamics model for mixing of moist air with pollutant consisting of HF, ideal
gas and water”, Report TNER.93.021, Thornton Research Centre, Shell Research, Chester, England (1993)
23
Schotte, W., “Fog formation of hydrogen fluoride in air”, Ind. Eng. Chem. Res. 26, pp. 300-306 (1987);
see also Schotte, W., “Thermodynamic model for HF formation”, Letter from Schotte to Soczek, E.I. Du
Pont de Nemours & Company, Du Pont Experimental Station, Engineering Department, Wilmington,
Delawere 19898, 31 August 1988
24
Cook, J. and Woodward, J.L., “A new integrated model for pool spreading, evaporation and solution on
land and water”, International Conference and Exhibition on Safety, Health and loss prevention in the Oil,
Chemical and Process Industries, Singapore, February 15-19 (1993)
25
Hanna, S.R., D.G.Strimaitis, and J.C.Chang, “Hazard response modelling uncertainty (A quantitative
method)”, Sigma Research Corp. report, Westford, MA for the API (1991)
26
Daish, N.C., Britter, R.E., Linden, P.F., Jagger. S.F., and Carissimo, B., “SMEDIS: Scientific Model
Evaluation techniques applied to dense gas dispersion models in complex situations”, International
Conference and Workshop on Modelling the Consequences of Accidental Releases of Hazardous Materials,
CCPS, San Francisco, California, September 28 – October 1 (1999)

CCPS 1999 UDM paper -207-09-04 ------------------------------------------------------------------------------------------------------------A single form of concentration profile is used to cover all stages of a release. This allows for anything from a sharp-edged profile in the initial stages of a jet release through to the diffuse Gaussian profile that would be expected in the final passive stage of spreading. The UDM includes the effects of droplet vaporisation using a non-equilibrium model (see Figure 8). Rainout produces a pool which spreads and vaporises. Vapour is added back into the plume and allowance is made for this additional vapour flow to vary with time. In addition to the non-equilibrium model, UDM also allows for an equilibrium model and an equilibrium model specific for HF (including effects of polymerisation). The UDM allows for vertical variation in ambient wind speed, temperature and pressure. Another feature of the UDM is possible plume lift-off, where a grounded cloud becomes buoyant and rises into the air. Rising clouds may be constrained to the mixing layer if it is reached. The UDM allows for continuous, instantaneous, constant finite-duration, and general time-varying releases. For the original UDM Cook and Woodward adopted a tuning process, where the tuning coefficients were obtained by the comparison of UDM results against a relatively large set of ‘tuning’ experiments. This type of tuning has largely been eliminated as part of the current work. The model coefficients have now been obtained directly from established data in the literature (based on wind-tunnel experiments), rather than doing UDM simulations and fitting the UDM results to the experimental data. Sections 2, 3, and 4 of this paper include a description of the theory, verification and module validation of the UDM dispersion model, the UDM thermodynamics, and the UDM pool spreading/evaporation model. Section 5 summarises the validation of the overall UDM model against large-scale field experiments. In Section 6 the major conclusions are summarised. The reader is referred to the UDM Technical Reference Manual3 for details not included in the present paper.

2. Dispersion model
2.1 Concentration profile and cloud geometry

Figure 1 shows the movement of the cloud in the downwind direction. The Cartesian co-ordinates x, y, z correspond to the downwind, cross-wind (lateral horizontal) and vertical directions, respectively; x=0 corresponds to the point of release, y = 0 to the plume centre-line and z = 0 to ground-level. Furthermore s is the arc length measured along the plume centre, θ the angle between the plume centre-line and the horizontal, and ζ the distance from the plume centre-line. The concentration c is given by a similarity profile c = c(x,y,ζ) as suggested by Webber et al.4

 ζ  c(x, y,ζ ) = co (x) F v ( ζ ) F h (y), with F v ( ζ ) = exp−  Rz 

n

   y , F h (y) = exp−   Ry  

m

 , (continuous)  

(1)

c(x, y,ζ ; t) = co (t) F v ( ζ ) F h (r) , with horizontal radius r = ( x − x cld ) 2 + y 2 ,

(instantaneous)

(2)

In case of steady-state dispersion, Equation ( 1 ) describes exponential decay of the concentration in y and ζ in terms of the cross-wind and vertical dispersion coefficients Ry(x), Rz(x). Empirical correlations are adopted for the exponents m, n such that the near-field sharp-edged profile (large value for m) develops into a Gaussian profile in the far field (m=2). As shown in Figure 1a, the plume cross-section is a circle (radius Ry=Rz) during elevated jet dispersion, a truncated circle during touching down, and a semi-ellipse after touching down. The area of this plume cross-section, Acld(x), is obtained by integration of Fv(ζ)Fh(y) over y,ζ . As in many other dispersion models, the cloud is also characterised by an equivalent ‘effective cloud’ [rectangular cross-section with area Acld, effective half-width Weff(x), and effective height Heff(x)(1+hd) with hd=0 for grounded plume and

with downwind spreading assumed to be equal to cross-wind spreading (Rx=Ry). relation between cloud speed and cloud position. Vcld. The cloud is also characterised by an equivalent ‘effective cloud’ [cylindrical shape with volume Vcld. and equivalent top-hat concentration equal to the centre-line concentration co(t).CCPS 1999 UDM paper -307-09-04 ------------------------------------------------------------------------------------------------------------hd=1 for elevated plume] with centroid cloud speed ucld. • Conservation of total cloud mass gnd dm wv d mcld = E tot + ds ds gnd dm wv d mcld = E tot + dt dt (3) . Air entrainment into a plume may be caused by a range of mechanism (see Figure 2). These equations express conservation of mass (air entrainment and water added from substrate). effective horizontal radius Weff. For a low-energy release. During the jet dispersion phase. Rz(t). In the case of instantaneous dispersion. Jet entrainment and crosswind entrainment are dominant in the near field after a high-pressure continuous release. The plume variables are given in the table below {ua(zc) = wind speed at centroid height zc.ua(zc)] Iz = mcld uz xcld zcld qgnd mwvgnd Ry unit (continuous) kg/s kg m/s2 kg m/s2 m m J/s kg/s m unit (instantaneous) kg kg m/s kg m/s m m J kg m After initialisation at the point of release. a truncated sphere during touching down. and kg/s for instantaneous dispersion). and equivalent top-hat concentration equal to the centreline concentration co(x). ζ =zzcld in terms of the horizontal and vertical dispersion coefficients Ry(t). z=zcld(t). y. the cloud speed is sufficiently close to the ambient speed and the contribution of non-passive entrainment is sufficiently small. The volume of this cloud. the cloud is a sphere (radius Ry=Rz) during elevated jet dispersion. the cloud moves in the downwind direction with cloud centre at time t denoted by x=xcld(t). is determined by integration Fv(ζ)Fh(r) over x.uz]=ucld[cos θ. and a cross3 wind spreading equation. a heat-transfer relation. the centreline velocity decays until either the heavy gas or the passive dispersion mechanisms become dominant. 2. y=0. The total air entrainment is Etot (kg/m/s for continuous. and effective height Heff(1+hd)] with centroid cloud speed ucld. (continuous) (4) . (instantaneous) The above equation expresses the change in cloud mass as a result of entrainment of air into the cloud and (in case of dispersion over water) water-vapour transfer from the substrate. sin θ] = cloud speed } plume variable mass of the cloud excess downwind momentum vertical momentum downwind position vertical position heat conduction from substrate water evaporated from substrate cross-wind dispersion coefficient symbol mcld Ix2 = mcld[ux . conservation of momentum. A transition is made to passive dispersion if the cloud density is sufficiently close to the ambient density (for heavy gas dispersion the Richardson number must be sufficiently small). these variables are determined by solving a set of ordinary differential equations forward in the downwind direction (continuous) or time (instantaneous). The entrainment contributions are calculated as follows: . the jet dispersion mode may never be dominant. [ux. ζ. These equations are described below (see the UDM Technical Reference Manual for full details).2 Dispersion variables and equations The model evaluates the plume variables as a function of downwind distance. a water-vapour transfer relation. and a semi-ellipsoid after touching down. As shown in Figure 1b. y. Equation ( 2 ) describes exponential decay of the concentration in x.

Ricou and Spalding6 determined from experiments α1 = 2π0.6 for a top-hat profile. u* the friction velocity and Φ(Ri*) the entrainment function of the Richardson number Ri*. uside is the side entrainment velocity. Jet entrainment Ejet is caused by turbulence resulting from the difference between the jet speed and the ambient wind speed: E jet = e jet Pabove ρ a | u cld .u a cos θ | E jet = e jet S above ρ a | u cld . Assuming a jet of uniform density (top-hat profile). Equation ( 5 ) for continuous dispersion is the Morton-Taylor-Turner formulation5. while for instantaneous dispersion a generalised version is adopted.5 and 2.282/1. and Equation ( 8 ) is a generalised version for instantaneous dispersion. Aside the side area and Atop the top area [for grounded plume. Since the value of the observed maximum concentrations was approximately 70% larger.( 10 ) ensures that heavy-gas entrainment is gradually phased in .5. where β∞=9.1o is the empirical value of the asymptotic half-angle of the jet.0 have found to be quoted in the literature for conversion between top-hat (averaged) concentrations and maximum concentrations. and Equation ( 6 ) is a generalised version for instantaneous dispersion.4 the Von Karman constant.282. it is present both in the near-field (Epasnf) and the far-field (Epasff). Pabove the cross-wind perimeter (m) of the continuous plume above the ground (m). we adopt α1 = 0.6/1. Heavy-gas entrainment Ehvy is included for a grounded heavy-gas plume: 4. The entrainment function Φ is taken from Havens and Spicer9 for Ri*<0 and from Britter10 for Ri*>0. Cross-wind entrainment Ecross in response to the deflection of the plume by the wind: E cross = α 2 ρ a P above | u a sin θ | E cross = α 2 ρ a S above | u a sin θ | in kg/m/s in kg/s (continuous) (7) (8) (instantaneous) Equation ( 7 ) for continuous dispersion is from Morton et al.17. (continuous) W gnd   {u side Aside + u top Atop} ρ a .CCPS 1999 UDM paper -407-09-04 ------------------------------------------------------------------------------------------------------------1. Passive entrainment is caused by ambient turbulence. and therefore we adopt α2 = 0. 3. For continuous elevated dispersion. The far-field passive entrainment is derived from empirical correlations σya(x) and σza(x) for the Gaussian cross-wind and vertical dispersion coefficients as a function of x. Aside = 2πWeffHeff. The value for α2 adopted by Briggs7 is 0.7 = 0. κ = 0.35.u a cos θ | in in kg/m/s kg/s (continuous) (instantaneous) (5) (6) where ρa is the ambient density. Again the value of the maximum concentrations is assumed to be 70% larger.5ejet=π0. and Sabove the instantaneous cloud surface area above the ground (m2). W gnd E hvy =   Ry    u top (2Weff ) ρ a . γ the side entrainment coefficient. Finally Wgnd is the cross-wind radius of the part of the cloud touching the ground.7 = 0. (instantaneous) Here utop is the top-surface entrainment velocity. E hvy =   Ry    u side = γ dWeff dt ( 10 ) . 2.5tan(β∞) = 0. and Atop = πWeff2]. Ratios between 1. For instantaneous dispersion. Thus the term Wgnd/Ry in Equations ( 9 ). This ensures the best fit for the top-entrainment velocity against a wide range of experimental data.   u top = κ u* Φ ( Ri* ) (9) . the near-field passive entrainment 11 formulation is taken from McFarlane derived from experiments by Disselhorst8.

and is added during touching down only.CCPS 1999 UDM paper -507-09-04 ------------------------------------------------------------------------------------------------------------during touching down.81 m2/s) and the density difference between the plume and the air.e. d q gnd ds d q gnd dt ( 15 ) = Q gnd [2W gnd ] . ds (continuous) ( 13 ) dz cld = u z = u cld sin θ . dt • dz cld = sin θ . the vertical buoyancy force (N/m or N). where ρcld is the cloud density] or instantaneous dispersion [cloud volume Vcld = mcld / ρcld]. the horizontal ground drag force Fdragground (N/m or N). (instantaneous)     ( 12 ) The terms in the right-hand side represent forces on the plume. from the elevated plume (Wgnd/Ry=0) to the ground-level plume (Wgnd/Ry=1).  dI x2   ds    =  dI z   ds    ground Fimpact − sin θ  0  + F ground drag   cos θ    1  0  0 + Acld (ρ cld − ρ a )g − 1 . This force is perpendicular to the plume centre line. • Horizontal and vertical position: dxcld = cos θ . • Conservation of excess horizontal and vertical component of momentum The adopted momentum equations (vector notation) are as follows for continuous dispersion [cloud area Acld = mcld / (ρclducld).     (continuous) ( 11 )  dI x2   dt  ground   = Fimpact dI z    dt    − sin θ  0  + F ground drag   cos θ    1  0  0 + Vcld (ρ cld − ρ a )g − 1 . They are respectively: the ground impact force Fimpactground (N/m or N) resulting from plume collision with the ground. This force is added after onset of touchdown only. and have been generalised for application to instantaneous dispersion. i. in W (instantaneous) . ds dx cld = u x = u cld cos θ . in W/m (continuous) ( 16 ) = Q gnd S gnd . (instantaneous) dt ( 14 ) Rate of heat convection from the substrate The heat convection from the substrate to the cloud is described by the following differential equation. The formulas for the ground drag and ground impact forces are partly taken from McFarlane11 for continuous dispersion. The vertical momentum equation is not used when the cloud is grounded or capped at the mixing layer (constant plume height). This force is proportional to the gravitational acceleration g (= 9.

e. In case of instantaneous releases qgnd is the total heat (J) transferred from the substrate to the cloud and Sgnd is the area of the cloud in contact with the substrate [see Figure 1b]. i. The heavy spread rate is applied until the passive transition. This has been included in the Unified Dispersion Model following the approach of the Colenbrander and Puttock described by Witlox14 which relates the rate of water vapour pick-up to the rate of heat convection from the water surface: gnd dm wv 5 P v w ( T gnd ) . R y = Rz 2. Jet spreading. instantaneous dispersion it is given by For d Ry CE = dt Cm g { max[0. T gnd > Tvap (instantaneous) ( 19 ) dt T gnd P a where Pvw is the saturated vapour pressure of water. C m = Γ 1 +   m . C m = Γ  1 +   m    1/2 and for continuous dispersion by d Ry C = E dx ux Cm g { max[0.ρ a ( z = z c )] } H eff (1 + hd ) ρ cld   2  . The cloud is assumed to remain circular until the passive transition or until the spread rate reduces to the heavy-gas spread rate (the latter can only occur after touchdown). Heavy-gas spreading. where Qgndn and Qgndf are the natural and forced convection flux from the substrate to the vapour cloud (W/m2) derived from expressions by McAdams (1954)12 and Holman (1981)13. The heat conduction flux Qgnd (W/m2) transferred from the substrate (temperature Tgnd) to the cloud (vapour temperature Tvap) is given by Q gnd = max Q gnd n . 1. respectively. ρ cld . T gnd > Tvap T gnd ≤ Tvap ( 17 ) . Q gnd = Q gnd f { f }.P v w ( T vap ) = Cp cld [ ] dq gnd dt . If Tgnd < Tvap or Tgnd < 0oC (substrate is ice) or if the cloud is passing over dry ground. • Water-vapour transfer from the substrate Water vapour can be transferred from a water surface into the cloud when the vapour temperature of the cloud is less than that of the water surface. • Crosswind spreading In general cross-wind spreading consists of the following three subsequent phases (see Figure 2). ρ cld . dqgnd/ds (J/m/s) is the heat transferred from the substrate per second and per unit of downwind direction and Wgnd is the half-width of the cloud in contact with the substrate [see Figure 1a].CCPS 1999 UDM paper -607-09-04 ------------------------------------------------------------------------------------------------------------In case of continuous releases. T gnd > Tvap (continuous) ( 18 ) ds T gnd P a gnd dm wv 5 P v w ( T gnd ) .ρ a ( z = z c )] } H eff (1 + hd ) ρ cld  1 .P v w ( T vap ) = Cp cld [ ] dq gnd ds . dmwvgnd/ds = 0 (continuous) or dmwvgnd/dt = 0 (instantaneous).

e.5 dt dx where σya(x) is the ambient passive cross-wind dispersion coefficient. no near-field passive dispersion) and the Briggs correlation7 (including ambient turbulence. the ground-level heavy-gas dispersion and the far-field passive dispersion.e. with ‘jet’ entrainment resulting from difference between cloud and (lower) ambient speed at point of release] (iii) Etot = max(Ehvy. this ‘conservative’ assumption is adopted for the final UDM model] The new formulation has also been validated against the HTAG18 experiments 139 and 140 (isothermal ground-level dispersion of Heavier Than Air Gas) and further verified against HGSYSTEM model results. Figure 5 includes the following results for experiment 140: (i) UDM run [with imposed HGSYSTEM wind-speed profile and concentration exponent n] (ii) HEGADAS run adopting standard passive dispersion coefficient σya without inclusion of turbulence collapse of heavy-gas spreading [in line with UDM assumptions] (iii) HEGADAS run adopting experimentally observed σya without collapse of gravity spreading (iv) HEGADAS run adopting experimentally observed σya with collapse of gravity spreading By comparing the results for (i) and (ii). 3. From the figure it is also inferred. it is concluded that UDM and HEGADAS predictions are in close agreement. i.3 Module verification and validation 3 The UDM Technical Reference Manual includes a detailed description of the UDM verification for the near-field elevated (jet) dispersion. the UDM numerical results are shown to be identical to the results obtained by an analytical solution. For an elevated horizontal continuous jet (of air). that future implementation into the . too conservative assumption] (ii) Etot = Ehvy+Ejet [sum of heavy and ‘jet’ entrainment. and Γ is the gamma function.15 is the cross-wind spreading parameter from experiments by Van Ulden15.CCPS 1999 UDM paper -707-09-04 ------------------------------------------------------------------------------------------------------------where CE = 1. Good agreement was obtained for all experiments. if similar assumptions are adopted. Passive spreading. Jet and near-field passive dispersion. see Figure 3 for example results. σya increases with averaging time as a result of wind meander. 2. 2. i.5 . Very good agreement has been obtained against both the Pratte and Baines correlation16 (no ambient turbulence.Ejet) [this may be more appropriate than the above assumption since the ‘heavy’ and ‘jet’ entrainment mechanisms are not independent.e. including near-field passive dispersion). Heavy-gas dispersion. i. (continuous) dx dx d σ ya dR y . After the passive transition the passive spread rate is applied d σ ya dR y = 2 0. 1. Note that Cm = Weff/Ry. Figure 4 includes the results for McQuaid experiment 3 for three types of UDM simulations. (i) Etot = Ehvy [inclusion of heavy entrainment only. release angle and transition criterion. release speed. This is an important verification of the correct implementation of the model. This verification can be summarised as follows. A sensitivity analysis has been carried out for a given base-case with parameter variations to the release height. The UDM numerical results are shown to be in identical agreement against an analytical solution for a 2-D isothermal ground-level plume. (instantaneous releases) = u x 2 0. The UDM has been validated against the set of three 2-D wind-tunnel experiments of McQuaid17 (steady-state ground-level dispersion of CO2).

the UDM allows for the quasi-instantaneous (QI) model or the finite-duration correction (FDC) model. propane liquid fraction. The UDM equilibrium- . and the subsequent phase is modelled as an ‘instantaneous’ circular cloud. Note however that this phenonmenon is important for a subset of heavy-gas dispersion problems only. A limitation of this model is however that it is strictly speaking only applicable to ground-level non-pressurised releases without significant rainout. The equilibrium model determines the phase distribution and the mixture temperature. and cloud density. The finite-duration correction includes the effect of averaging time because of time-dependency of the concentrations. Ambient humidity. averaging time. Thermal equilibrium is assumed. the cloud is replaced by an ‘equivalent’ circular cloud. The QI model can be applied with or without the ‘duration adjustment’. Thermodynamics model UDM invokes the thermodynamics module while solving the dispersion equations in the downwind direction. while assuming along-wind spreading equal to cross-wind spreading in the analytical profile. The UDM includes the following types of thermodynamic models: 1. liquid). vapour and liquid cloud temperature. 3. The same agreement has been obtained for the case of purely (far-field) passive instantaneous dispersion. e.g. For purely (far-field) passive continuous dispersion. The cooling effect because of component evaporation and the heating effect because of water condensation is shown. water (vapour. It takes the effects of downwind diffusion gradually into account including effects of both turbulent spread and vertical wind shear. release rate and wind speed. The FDC module has been verified against the HGSYSTEM/SLAB steady-state results.4 Finite-duration releases To model finite-duration releases with a uniform release rate. The module describes the mixing of the released component with moist air. The FDC model is based on the HGSYSTEM formulation derived from that adopted in the SLAB dispersion model19. The liquid component in the aerosol is considered to consist of spherical droplets and additional droplet equations may be solved to determine the droplet trajectories. A sensitivity analysis has been carried out for a base-case with parameter variations to the release height.CCPS 1999 UDM paper -807-09-04 ------------------------------------------------------------------------------------------------------------UDM of collapse of gravity spreading may be desirable. droplet mass and droplet temperature. The QI model models the initial phase as a continuous source (neglect of downwind gravity spreading and downwind diffusion). Far-field passive dispersion. the UDM numerical results are shown to be in close agreement with the vertical and crosswind dispersion coefficients and concentrations obtained from the commonly adopted analytical Gaussian passive dispersion formula. Rainout of the liquid component occurs if the droplet size is sufficiently large. where the duration adjustment applies the effect of averaging time because of timedependency of the concentrations (for averaging times larger than release duration). which implies that the same temperature is adopted for all compounds in the cloud (vapour and liquid). erroneous significant increase in maximum concentration). and may take into account water-vapour and heat transfer from the substrate to the cloud. The disadvantage of the QI model is the abrupt transition (sometimes resulting in severe discontinuities. propane temperature have been varied. stability class. ice)]. The equilibrium model is tested for mixing of propane with moist air at 20C. 3. surface roughness length. When the cloud width becomes ‘large’ with respect to the cloud length. and shown to lead to finite-duration corrections virtually identical to the latter programs. Moreover it produces predictions of the maximum (centre-line ground-level) concentrations only. The current duration adjustment over-estimates this effect downwind of the QI transition. It is based on a simplified version of the multi-compound algorithm developed by Witlox for use in HGSYSTEM21. 2. Thus separate water (liquid or ice) and component (liquid) aerosols may form. The module calculates the phase distribution [component (vapour. It has a better scientific basis and is derived from an analytical solution of the Gaussian plume passive-dispersion equations.20. liquid. Equilibrium model (no reactions).

The initial drop size is taken as the minimum of the droplet size calculated by mechanical break-up and flashing break-up. The pool spreads until it reaches a bund or a minimum pool thickness. A brief assessment of the UDM droplet model has been carried out. If the . The PVAP results were compared by David Webber against the SRD/HSE model GASP for a range of scenarios with the aim of testing the various sub-modules. The model includes the effect of HF polymerisation and fog formation. these being important for some chemicals. It is confirmed that for reducing droplet size the non-equilibrium model converges to the equilibrium model. A limited sensitivity analysis has been carried out in which droplet trajectories etc. maintaining mass and heat balances for both boiling and evaporating pools. In the case of an instantaneous release the vapour produced by the spilt liquid is added back into the cloud. rainout occurs. Therefore the UDM thermodynamic predictions are compared against those of HGSYSTEM. and whether it is an instantaneous or a continuous release. Non-equilibrium model (no reactions). Vapour is added back into the cloud and allowance is made for this additional vapour flow to vary with time. and predictions are shown to be consistent. In conjunction with the equilibrium thermodynamics model the droplet model is used to set the droplet trajectories and the point of rainout only. Pool spreading and vaporisation If the droplet reaches the ground. it additionally calculates the droplet mass and the liquid droplet temperature. The UDM source term model PVAP calculates the spreading and vapour flow rate from the pool. the model takes into account heat conduction from the ground. Pool spreading/evaporation. so long as part of the cloud still covers the point at which the pool was formed by the rained-out liquid. the rate of generation of vapour from the spilt liquid is added to the vapour already in the cloud to give a total flow rate for the combined source. 2.e. have been compared. Addition of pool vapour back to the cloud. radiation and vapour diffusion. The same temperature is adopted for all compounds in the cloud (vapour and liquid). See the UDM Technical Reference Manual3 for further details and model validation against experimental data. i. When the release stops there may then be a period of vapour generated from the liquid pool alone. Thus further work for improvement was identified. 4. The original version of PVAP was developed by Cook and Woodward24.CCPS 1999 UDM paper -907-09-04 ------------------------------------------------------------------------------------------------------------model predictions are shown in Figure 6 to be very close with HEGADAS predictions for the case of mixing of propane vapour/liquid (at –43/42C) with 0%/100% humid air. ambient convection form the air. Different models are adopted depending whether the spill is on land or water. Further work needs to be carried out to further improve the model for flash calculations and droplet correlations. The HF thermodynamics model is based on the HGSYSTEM thermodynamics model22. removal of the liquid component from the cloud. The non-equilibrium model determines the phase distribution of the water and the vapour temperature. This allows the pool temperature to vary as heat is either absorbed by the liquid or lost during evaporation. These are usually the main mechanisms for boiling and evaporation. In conjunction with the non-equilibrium thermodynamics model. This also implies that good agreement is obtained against Schotte’s experiment23 as is shown by Figure 7. This model allows the temperature of the droplet (liquid component) to be different of the temperature of the other compounds in the cloud. A sensitivity analysis is carried out for mixing of HF with moist air. For spills on land. These effects are modelled numerically. Solution and possible reaction of the liquid in water are also included for spills on water. whereby both humidity and initial liquid mass fraction have been varied. Equilibrium model (HF). The pool may either boil or evaporate while simultaneously spreading. 3. For a continuous release. This produces a liquid pool which spreads and vaporises (see Figure 8).

were calculated to calculated the geometric mean bias (under or over-prediction of mean) and mean variance (scatter from observed data) for each validation run. 10 experiments). 5.10 07-09-04 ------------------------------------------------------------------------------------------------------------spilt liquid all evaporates while covered by the cloud then all that is produced is an instantaneous cloud. 2 experiments) Thorney Island (instantaneous unpressurised ground-level release of Freon-12. Good agreement was also obtained for the Goldfish experiments prior to the passive transition (Figure 11). recommendations for future work may centre upon the enhancement of the heavy spread formulation to include a gravity collapse criteria. validated where possible. This has been carried out in extensive detail for the entire basic continuous model [phases of dispersion (passive. If the liquid has not all evaporated then once the upwind edge of the cloud has moved past the pool any remaining liquid is assumed to form a continuous source of vapour. 2 experiments) Goldfish (continuous elevated two-phase HF jet. The predictions for the neutrally buoyant Prairie Grass experiments (Figure 9) and the aerosol releases of Desert Tortoise (Figure 10). 4 experiments) FLADIS (continuous elevated two-phase ammonia jet. Formulas adopted by Hanna et al. The modules have been corrected where necessary. The overall performance of the UDM in predicting both peak centreline concentration and cloud widths was found to be good for the above experiments. 3 experiments) EEC (continuous elevated two-phase propane jet. 3. 2. 4 experiments) Burro (continuous evaporation of LNG from pool. This section is concerned with the validation of the overall model against largescale field experiments. 5. and been compared with similar third-party sofware applications. 6. Conclusions Each of the modules in the UDM has been investigated and verified in detail in conjunction with a literature review and a sensitivity analysis. 3 experiments) Maplin Sands (continuous evaporation of LNG from pool. Good resuls were obtained for Burro. 4. 9 experiments) Each of the above experimental sets were statistically evaluated to determine the accuracy and precision of 25 the UDM predictions with the observed data. cloud widths. Validation The above sections described the verification and the validation against wind-tunnel experiments for the individual UDM modules.25 and from the ongoing EEC SMEDIS26 program (Scientific Model Evaluation of Dense Gas Dispersion Models). This was carried out for centre-line concentrations. FLADIS and EEC were found to be very good. . and (for the SMEDIS experiments) also off centre-line concentrations. 7. See the UDM Technical Reference 3 Manual for further details. Thus seven sets of continous experiments and one set of instantaneous experiments were selected: 1. The basis and choice of these experiments stems from both the model evaluation carried out by Hanna et al. A more strict passive transition criterion would improve the results in the far field. 8.CCPS 1999 UDM paper . jet. Prairie Grass (continuous passive dispersion of sulphur dioxide. Desert Tortoise (continuous elevated two-phase ammonia jet. Following this work the UDM comparison against largescale experiments improved considerably. ground-level heavy-gas dispersion]. improved modelling of dispersion from a pool and improved transition to passive dispersion. heavy). - Following an analysis of the experiments. Good agreement was obtained for the instantaneous heavy-gas Thorney Island experiments (Figure 12). equilibrium thermodynamics without rainout] and to some extent also the instantaneous model [far-field passive dispersion. despite the elimination of tuning coefficients. This conclusion also applied for the Maplin Sand experiments. 6.

A more detailed assessment is to be carried out for these areas. this eliminates discontinuities and matching problems (b) a very extensive verification and validation to ensure that the model shows the correct behaviour and produces accurate predictions (c) integration within the user-friendly and well-established DNV consequence-modelling package PHAST and the risk-analysis package SAFETI. pressurised instantaneous expansion. A brief assessment has been carried out for droplet modelling. lift-off and mixing-layer logic. The key differences resulting from the modelling changes between the old UDM model (PHAST 5. This enables plotting.0) are as follows: larger concentrations for far-field passive dispersion [corrections to passive dispersion logic] lower concentrations for near-field elevated jet dispersion [increased jet and cross-wind entrainment coefficients. added near-field dispersion] larger ground-level heavy-gas spreading (wider clouds) [increased cross-wind spreading parameter] often increased water aerosol formation and therefore hotter clouds [new thermodynamics models] The key advantages of the new UDM model with respect to other typical dispersion models can be summarised as follows: (a) a single model for the entire dispersion regime from the point of release to the far-field dispersion including possible rainout and pool re-evaporation. HF thermodynamics. linking with discharge/fire/explosion models. pool spreading/evaporation. finiteduration releases. and timedependent releases.CCPS 1999 UDM paper . and heat/water transfer from the substrate. toxic/flammable impact and risk calculations. the link between pool and dispersion model.20) and the new UDM model (PHAST 6. (d) rigid procedures and control using ISO9001/TickIT quality standards .11 07-09-04 ------------------------------------------------------------------------------------------------------------A detailed assessment and limited corrections have been carried out for the transition to passive.

12 07-09-04 ------------------------------------------------------------------------------------------------------------- z ζ circular cross-section (Ry= R z) truncated cross-section (Ry > R z) θ semi-elliptic cross-section (R y > Rz) zcld s x ELEVATED PLU ME y TOU CHIN G DOWN GROUN D-LEVEL PLU ME (a) continuous dispersion z spherical cloud truncated cloud (circular ground surface) EN ERGETIC EXPAN SION PHASE semi-ellip soid cloud (circular ground surface) up per cloud envelop e lower cloud envelope x GROUN D-LEVEL DISPERSION y ELEVATED DISPERSION TOUCHING DOWN (b) instantaneous dispersion Figure 1.CCPS 1999 UDM paper . UDM cloud geometry .

Epasff = near-field elevated passive. Ehvy = ground-level heavy.CCPS 1999 UDM paper . heavy or passive) and the mechanism of entrainment (Ejet = jet.Ehvy) + Epasnf circular spread max(Ejet. Along the transition zone the near-field spread/entrainment are phased out and the far-field spread/entrainment are phased in. Epasff = far-field passive). Phases in UDM cloud dispersion for range of scenarios The figures indicate for each phase the type of spreading (circular jet.Ehvy) + Epasnf heavy spread max(Ejet+ Ecross.Ehvy) + Epasnf phase in passive spread and entrainment d σya/dx Epasff X (b) jet/plume becomes passive during touching down JET (elevated) Z circular spread nf Ejet+ Ecross+ Epas circular spread max(Ejet+ Ecross.13 07-09-04 ------------------------------------------------------------------------------------------------------------Z JET (elevated) circular spread E jet+ E cross+ E pasnf TRANSITION phase in passive spread and entrainment PASSIVE passive spread (d σ ya/dx ) passive entrainment (E pasff) X (a) elevated jet/plume (no touching down) JET (elevated) Z circular spread nf Ejet+ Ecross+ Epas JET (touching down) HEAVY (touching down) TRANSITION PASSIVE circular spread max(Ejet+ Ecross. Ecross = cross-wind. .Ehvy) heavy spread E hvy phase in passive spread and entrainment d σya/dx Epasff JET (touching down) JET (ground level) HEAVY (ground level) TRANSITION PASSIVE X (c) jet/plume become passive after touch down Figure 2.

comparison against Briggs plume rise formula (jet/ambient velocity ratio = 4.1 1 10 dimensionless plume arclength 100 (a) case of no ambient turbulence.CCPS 1999 UDM paper . comparison against Pratte and Baines correlation (jet/ambient velocity ratio = 13) 10 Briggs formulae UDM (no passive) centre-line height (m) UDM (with near-field passive) 1 0.14 07-09-04 ------------------------------------------------------------------------------------------------------------- 10 Pratte and Baines UDM (no passive) dimensionless plume rise 1 0.1 0.6) Figure 3.1 1 downwind distance (m) 10 100 (b) case of ambient turbulence. Comparison of UDM results against plume-rise correlations for vertical neutrally buoyant jet .

(iii) Etot = max(Ehvy.00E-01 1.00E+01 (a) ground-level concentration cov (mole %) 0. Heavy) McQuaid Experimental data UDM (Jet included) UDM (No jet) 0.00E+00 downwind distance (m) 1. McQuaid experiment 3.00E-01 1.15 07-09-04 ------------------------------------------------------------------------------------------------------------- 100 groundlevel concentration (mol%) UDM max(Jet.00E+00 downwind distance (m) 1.CCPS 1999 UDM paper . (ii) Etot = Ehvy+Ejet.00E+01 (b) height H1/2 (m) at which concentration is halved Figure 4. experimental data and UDM predictions assuming (i) Etot = Ehvy. Heavy) McQuaid Experimental data UDM (Jet included) UDM (No jet) 10 1.01 1.Ejet) .1 cloud depth (m) UDM max(Jet.

gravity collapse) . gravity collapse). (ii) HEGADAS (standard σya . (iii) HEGADAS (standard σya . (iv) HEGADAS (experimental σya . n o g ra vity c o lla p s e ) (a) effective cloud half-width Weff U D M w id th (S M E D IS d e fin itio n ) 200 e xp e rim e n ta l d a ta 180 H G S Y S T E M (e xp e rim e n ta l S IG M A Y . experimental data and UDM/HEGADAS predictions: (i) UDM [use HEGADAS wind-speed profile and value for n]. g ra v ity c o lla p s e ) concentration (mole %) 10 H G S Y S T E M (s ta n d a r d S IG M A Y .16 07-09-04 ------------------------------------------------------------------------------------------------------------U D M e ffe c tive h a lf w id th 200 180 160 cloud half-width (m) 140 120 100 80 60 40 20 0 0 80 160 240 320 d o w n w in d d is ta n c e (m ) 400 480 560 640 H G S Y S T E M (e xp e rim e n ta l S IG M A Y . s ta n d a r d S IG M A Y ) e x p e r im e n ta l d a ta H G S Y S T E M (e x p e r im e n ta l S IG M A Y . g ra vity c o lla p s e ) H G S Y S T E M (s ta n d a rd S IM A Y . g ra vity c o lla p s e ) H G S Y S T E M (s ta n d a rd S IG M A Y . n o g ra v ity c o lla p s e ) 1 0 . HTAG experiment 140.CCPS 1999 UDM paper .1 0 80 160 240 320 d o w n w in d d is ta n c e ( m ) 400 480 560 640 (c) centre-line ground-level concentration Figure 5. n o g ra vity c o lla p s e ) (b) cloud half-width (SMEDIS definition) 100 U D M ( Z = 0 . g ra v ity c o lla p s e ) H G S Y S T E M (s ta n d a rd S IG M A Y . g r a v it y c o lla p s e H G S Y S T E M (s ta n d a r d S IG M A Y . no gravity collapse). g ra vity c o lla p s e ) 160 cloud half-width (m) 140 120 100 80 60 40 20 0 0 80 160 240 320 d o w n w in d d is ta n c e (m ) 400 480 560 640 H G S Y S T E M (s ta n d a rd S IG M A Y .

100% humidity temperature (C) 0 -5 20 40 60 80 100 -15 -25 -35 -45 % propane mole fraction (a) mixing of propane vapour (-42C) with air (20C) 20 10 0 0 -10 HEGADAS .100% humidity 10 20 30 40 50 60 70 80 90 100 temperature (C) -20 -30 -40 -50 -60 -70 -80 % propane mole fraction (b) mixing of propane liquid (-43C) with air (20C) Figure 6.70 % humidity HEGADAS .CCPS 1999 UDM paper . Temperature predictions by HEGADAS and UDM thermodynamics equilibrium models for mixing of propane (vapour or liquid) with air (dry or humid) .70% humidity UDM .100 % humidity UDM .17 07-09-04 ------------------------------------------------------------------------------------------------------------- 25 HEGADAS .0 % humidity HEGADAS .0% humidity 5 UDM .0% humidity UDM .0 % humidity 15 HEGADAS .100 % humidity UDM .

001 0. Droplet evaporation. . HEGADAS and UDM thermodynamic test-bed predictions (flashing) two-phase discharge from pipe/vessel vapour-plume centre-line droplet trajectory point of rainout spreading evaporating liquid pool Figure 8.15 283.15 303. rainout. Mixing at 26C of HF vapour with moist air (humidity = 50%).15 288.01 total HF mole fraction 0.18 07-09-04 ------------------------------------------------------------------------------------------------------------- 313.CCPS 1999 UDM paper .15 278.15 UDM HEGADAS experiment 293.1 1 Figure 7.15 273.15 cloud temperature 298. and pool spreading/evaporation.15 308.15 0.

1 0. UDM validation against Prairie Grass experiment 8 (passive dispersion) .CCPS 1999 UDM paper .00001 1 10 100 downwind distance (m) 1000 10000 centre-line at z =0 experimental data (a) mol % concentration Prairie Grass 8 .1 1 10 100 downwind distance (m) 1000 10000 (b) cloud dimensions Figure 9.0001 0.01 0.001 0.19 07-09-04 ------------------------------------------------------------------------------------------------------------Prairie Grass 8 .Elevated release of Sulfur Dioxide 1000 100 cloud dimension (m) half width 10 depth Experimental width 1 0.Elevated release of Sulfur Dioxide 100 10 concentration (mol %) 1 0.

001 1 10 100 downwind distance (m) 1000 10000 (a) mol % concentration Desert Tortoise 2 .01 0.20 07-09-04 ------------------------------------------------------------------------------------------------------------Desert Tortoise 2 .1 0.CCPS 1999 UDM paper .1 1 10 100 downwind distance (m) 1000 10000 (b) cloud dimensions Figure 10.Ammonia Aerosol 1000 100 cloud dimension (m) half width 10 depth Experimental width 1 0.Ammonia Aerosol 100 10 concentration (mol %) 1 centre-line at z =1 experimental data 0. UDM validation against Desert Tortoise 2 (two-phase ammonia jet) .

Elevated HF aerosol 1.00E+00 experiment HFPLUME HEGADAS HEGADAS (no heat transfer) 1.00E+01 HFPLUME HEGADAS UDM Module test 0.00E-03 1 10 100 1000 10000 downwind distance (m) (a) maximum concentration versus downwind distance GoldFish3 .00E+02 1.00E-01 HEGADAS (exp. SY) UDM module test 1.00E+01 2.00E+02 1.00E+00 1.00E-02 1.00E+01 1.21 07-09-04 ------------------------------------------------------------------------------------------------------------- GoldFish3 .00E+01 temperature (K) experiment 1.00E+01 1.00E+01 3.00E+03 1.00E+01 -2.00E+00 -1. Goldfish 3 simulations for HFPLUME/HEGADAS (including heat transfer) and UDM .Including heat transfer 4.00E+04 downwind distance (m) (b) temperature versus downwind distance Figure 11.00E+01 concentration (mol %) 1.CCPS 1999 UDM paper .

01 0.001 0. m2 effective top area of instantaneous plume. kg m/s or kg/m/s2 vertical component of plume momentum. kg/s or kg/m/s total dispersion entrainment rate.0001 1 10 100 downwind distance (m) 1000 10000 Figure 12.22 07-09-04 ------------------------------------------------------------------------------------------------------------Thorney Island 8 . m2 effective side area of instantaneous plume. kg) or mass rate in plume (continuous release. UDM validation against Thorney Island 8 (instantaneous heavy-gas dispersion) Nomenclature Acld Aside Atop c co Ecross Ehvy Ejet Epasnf Epasff Etot Fdragair Fdragground Fh(x) Fv(ζ) Heff hd Ix2 Iz m mcld mwvgnd n Pabove Pvw(T) cross sectional area of continuous cloud. kg/s) water-vapour added from the substrate. N/m or N horizontal distribution function for concentration (-) vertical distribution function for concentration (-) effective height of cloud after full touchdown. Pa . kg/s or kg/m/s dense gas entrainment rate. kg/s or kg/m/s near-field passive dispersion entrainment rate.4 experimental data 0. m [height prior to full touchdown = Heff(1+hd)] fraction of bottom half of cloud which is above the ground [hd=0. m saturated vapour pressure as function of temperature T (K) for water. kg or kg/s exponent of vertical distribution function for concentration (-) perimeter length of jet. kg/s or kg/m/s far-field passive dispersion entrainment rate.Instantaneous release of Nitrogen and Freon 100 10 concentration (mol %) 1 centre-line 0. kg of component / m3 cross-wind entrainment rate. kg/s or kg/m/s airborne drag force. kg/s or kg/m/s jet (high-momentum) entrainment rate.1 at z =0. kg of component /m3 centre-line concentration.1 for grounded.elevated cloud] horizontal plume momentum in excess to ambient momentum [Ix2= Ix – mcldua]. m2 concentration.CCPS 1999 UDM paper . N/m or N ground drag force. kg m/s or kg/m/s2 exponent of horizontal distribution function for concentration (-) mass in plume (instantaneous release.

m/s volume of cloud. John L. pp. Det Norske Veritas. m surface roughness length.CCPS 1999 UDM paper . 185-207 (1995) 2 Woodward. m/s ambient wind-speed. κ = 0.23 07-09-04 ------------------------------------------------------------------------------------------------------------qgnd Ry Rz Ri* s Sgnd t Ta Tgnd Tvap u* ua ucld uref uside utop ux.. m κ Von Karman constant.W. α2 jet. m Greek letters α1. s ambient temperature. m crosswind distance. 209-230 (1995) 3 Witlox. David M. J or J/s term in cross-wind concentration profile. and Staale Selmer-Olsen. m footprint area for instantaneous plume.. K friction velocity for cloud. m height above ground of cloud centroid. Journal of hazardous materials 44. A. m/s total cloud speed. m σza standard empirical correlation for vertical crosswind dispersion coefficient. m horizontal downwind position of cloud. kg/m3 density of ambient air. K substrate temperature.. June 1999. ua = ua(z). and Holt. (-) arclength along centre-line of the plume. A. m Acknowledgements The authors gratefully acknowledge the feedback from Torstein K. (-) γ side-entrainment coefficient for instantaneous heavy-gas dispersion. m2 time since onset of release. Woodward.0. who reviewed the UDM Technical Reference Manual. rad. while θ = π/2 corresponds to a vertical upwards plume (in z-direction) ζ distance perpendical to plume centre-line. m/s entrainment velocity through sides of plume. m vertical height. K temperature of vapour phase of the cloud. m reference height. Journal of hazardous materials 44. J. m [Rz = Rz(x) 21/2σz(x)] layer Richardson number. m3 effective half width of plume.. H. θ = 0 corresponds to a horizontal plume (in downwind xdirection).. m [Ry = Ry(x) = 21/2σy(x)] term in vertical concentration profile. UDM Version 6. “Reassessment and reevaluation of rainout and drop size correlation for an aerosol jet”. Cook. References Woodward. Fanneløp. (-) Γ Gamma function (-) θ angle to horizontal of plume. m cloud centre-line height above ground. J. m horizontal downwind distance.uz Vcld Weff Wgnd x xcld y z zcld zo zc zref heat transfer rate from ground to cloud.4 (-) ρcld density of plume. Webber. “Modelling and validation of a dispersing aerosol jet”. kg/m3 ρa σya standard empirical correlation for passive crosswind dispersion coefficient. and Papadourakis. m footprint half-width for continuous plume. on which the present paper is based. and Papadourakis. m/s value of ambient windspeed ua at reference height z = zref. pp.L.L. m/s entrainment velocity through top of plume. m/s horiontal and vertial component of cloud speed ucld. J. “Unified Dispersion Model – Technical Reference Manual”.and cross-wind entrainment coeficient .M. London (1999) 1 . A.

I. Technical Paper P21.B.. and Turner. Ch.022. and Baines..84. Series A.M. Shell Internationale Research Maatschappij B.. W..E.. England (1993) 23 Schotte. D. “A review of mixing experiments relevant to dense gas dispersion”.I. International Conference and Workshop on Modelling the Consequences of Accidental Releases of Hazardous Materials. California (1986) 21 Witlox. “Unpublished notes on downwind spreading formulation in finite-duration release version of SLAB”. in “Atmospheric dispersion of heavy gases and small particles” (Ooms.).. Letter from Schotte to Soczek.S. K. S. 6..A.M. Ind.Chang. . “Thermodynamic model for HF formation”. Contract for API.. 21-31 (1961) 7 Briggs. Singapore. Engineering Department. “Heat transfer”. Report DOE/TIC-27601 (1984) 8 Disselhorst.”Turbulent gravitational convection from maintained and instantaneous sources”. D. and Woodward.. February 15-19 (1993) 25 Hanna. Rees. Tickle and T. “Development of plume and jet release models”.A.. May 20-24. Eng. J.. and Tennekes. Jones. California. G. Proc.R. B. New Orleans. Thornton Research Centre. University of Arkansas (1985) 10 Britter..H. 327-366) in Randerson.L. G. J.W. IMA Conference on stably stratified flow and dense gase dispersion “. 1 (1956) 6 Ricou.” SRD Report SRD/HSE R586 April 1992. Britter.. see also Schotte. England (1990) 20 Ermak.021. G. England (1993) 22 Witlox. AMGR. “SMEDIS: Scientific Model Evaluation techniques applied to dense gas dispersion models in complex situations”. Wilmington. Report RKER.P. “Heat Transmission”. Thornton Research Centre. “The incorporation of atmospheric turbulence in the KSLA Plumepath program”. ASCE.I.3.93. 53-64 (1967) 17 McQuaid.E.. CCPS. . E.. “A new bulk model for dense gas dispersion: two-dimensional spread in still air. “Development and validation of atmospheric dispersion models for ideal gases and hydrogen fluoride”.D. M.93. K. Springer-Verlag.M. “Some experiments on the structure of stably stratified shear flows”. 26. April 1986. H. (ed. and Carissimo. “Effect of homogeneous and heterogeneous surface roughness on HTAG dispersion”.O. pp.L. 419440.. R. A. Delawere 19898. Chester. England (1993) 15 Van Ulden. “Hazard response modelling uncertainty (A quantitative method)”. Oxford University Press (1988) 11 McFarlane. US Department of Energy. San Francisco. Res.J. T. CCPS. and Spalding. ideal gas and water”. P. D. I: Near instantaneous release.Strimaitis. Lawrence Livermore National Laboratorym. D. and “.... J.. UK (1976) 18 Petersen.. LA.059 (1984) 9 Havens. Du Pont de Nemours & Company... Proc. J. Jagger.. Chester.016. Report TNER. “Profiles for the round turbulent jet in a cross flow”. “Thermodynamics model for mixing of moist air with pollutant consisting of HF.. “Technical description of the heavy-gas-dispersion program HEGADAS”. H.24 07-09-04 ------------------------------------------------------------------------------------------------------------Webber. International Conference and workshop on modelling and mitigating the accidental releases of hazardous materials. “Measurements of entrainment by aximsymmetrical turbulent jets”. R. H.. Health and loss prevention in the Oil.M. Report TNER.. International Conference and Exhibition on Safety.L. pp. Du Pont Experimental Station.. 11 (1). Sheffield. H.H. “A new integrated model for pool spreading. J. Chester. Chem.. Linden. “Plume rise and buoyancy effects”.M.. McGraw-Hill (1954) 13 Holman. 8 (pp. F.D. R.W. Part II: HGSYSTEM program user’s manual.A. and Ratcliff. 5th Ed. F.R. Shell Research.90. “Atmospheric Science and Power Production”. 300-306 (1987)... The Hague. Draft Report CPP-87-0417 (1988) 19 Witlox. eds..93.F. Thornton Research Centre. Taylor. “Fog formation of hydrogen fluoride in air”. N. Report TNER. evaporation and solution on land and water”. MA for the API (1991) 26 Daish. McGrawHill (1981) 14 Witlox.F. W. Chester. Technical Information Center.G. B. and Puttock. Berlin (1984) 16 Pratte. McFarlane. S. J. Vols. AIChE.CCPS 1999 UDM paper .032. pp. J. Safety in Mines Research Establishment. D. report.. J.. W.” SRD Report SRD/HSE R587 April 1992 5 Morton. 234. 31 August 1988 24 Cook.. Sigma Research Corp. HY (Journal of the Hydraulics Division). Chester.C. Two-phase thermodynamics model for mixing of a non-reactive multi-compound pollutant with moist air.). II: Steady continuous releases. Vol. H. R.Soc. September 28 – October 1 (1999) 4 . W.. Wren. and J. Shell Research. S. pp. Shell Research.W.“Development of an atmospheric dispersion model for heavier-than-air mixtures” (DEGADIS model). Thornton Research Centre.M. G.. Chemical and Process Industries. “A model of a dispersing dense gas cloud and the computer implementation D*R*I*F*T. Colorado. 1.J.S.P.W. CPP Incorporated.V. Westford.C. Fluid Mech. 657-688 (1991) 12 McAdams. pp. J.. B. Shell Research. and Spicer.