Anfossi, D. et al. (2003) Plume Rise.

Chapter 6 of AIR QUALITY

MODELING - Theories, Methodologies, Computational Techniques,
and Available Databases and Software. Vol. I - Fundamentals (P.
Zannetti, Editor). Published by The EnviroComp Institute
([Link] and the Air & Waste Management
Association ([Link]

Chapter 6

Plume Rise
Domenico Anfossi (1), Elisa Canepa (2) and Han van Dop (3)
(1)
CNR Istituto di Scienze dell'Atmosfera e del Clima, Corso Fiume 4, I-10133
Torino, Italy
anfossi@[Link]
(2)
INFM (National Institute for the Physics of Matter), Department of Physics -
University of Genova, Via Dodecaneso 33, I-16146 Genova, Italy
canepae@[Link]
(3)
Institute for Marine and Atmospheric Research Utrecht, P.O. box 80.005, 3508
TA Utrecht, The Netherlands
dop@[Link]

Abstract: Plume rise determination is one of the main processes encountered in air pollution
modeling. Therefore, the most commonly used methods for introducing plume rise in dispersion
models are presented. They encompass simple but robust and documented semi empirical
formulations, easy to be implemented in operative models, and advanced plume rise models.
Then, the problem of how to account for plume rise in Lagrangian dispersion particle models is
addressed. Finally, special situations of plume rise, like the occurrence of an elevated inversion, or
the presence of building and/or stacks features interacting with the plume, are investigated.

Key Words: buoyant plumes, jet plumes, ambient turbulence, self-induced turbulence, dispersion
modeling, effective plume heights, stability conditions.

 EnviroComp Institute and Air & Waste Management Association 103

104 Air Quality Modeling – Vol. I

List of Symbols

A dimensional constant in Equation (57) [L3/4 M-1/4 T-1/4]

Ai drift coefficient for velocity in Equations (97) [L T-2]
Ap dimensionless constant in Equation (87)
As stack outlet area [L2]
A1 dimensional constant in Equation (36) [T6/5 L-6/5]
A2 dimensional constant in Equation (36) [T6/5 L-3/5]
A3 dimensional constant in Equation (37) [T15/8 L-3/2]
A4 dimensional constant in Equation (37) [T6/5 L-3/5]
Aϑ drift coefficient for potential temperature in Equations (97)
[K T-1]
a drift coefficient in Equation (77) [L T-2]
B plume particle buoyancy [L T-2]
B0 initial plume buoyancy [L T-2]
bi acceleration of air displaced through an inversion [L T2]
bij diffusion coefficient for velocity in Equations (97) [L T-3/2]
bp lower edge of the plume [L]
bϑ diffusion coefficient for potential temperature in Equations
(97) [K T-1/2]
C parameter in Equation (138) [L-1/3 T]
C0 dimensionless constant in Equation (77)
cB dimensional constant in Equations (92) [L2 T-5]
cp specific heat at constant pressure [L2 T-2 K-1]
cw dimensional constant in Equations (92) [L2 T-3]
c1 dimensionless constant in Equations (91)
c2 dimensionless constant in Equations (91)
D dimensionless parameter in Equation (139)
D0 dimensional constant in Equation (57) [L]
D1 dimensionless constant in Equation (57)
D2 dimensional constant in Equation (57) [T K]
d spacing between adjacent stacks [L]
ds internal diameter of the stack outlet [L]
d1 dimensionless constant in Equation (70)
d2 dimensionless constant in Equation (70)
d3 dimensionless constant in Equation (71)
d4 dimensionless constant in Equation (72)
dW random increment in Equation (77)
dωB random increments for buoyancy in Equations (84)
[T1/2]
dωj random increments for velocity in Equations (97) [T1/2]
dωw random increments for velocity in Equations (84) [T1/2]
dω ϑ random increments for potential temperature in
Equations (84) [T1/2]
6 Plume Rise 105

E turbulent kinetic energy [L2 T-2]

En dimensionless enhancement factor
f dimensionless stack tip downwash correction factor
f′ dimensionless fraction of the plume trapped below inversion
fb atmospheric turbulence buffet frequency [T-1]
Fb buoyancy flux parameter [L4 T-3]
Fbi buoyancy flux of i-th particle [L4 T-3]
Fe plume buoyancy flux at the end of bending-over phase [L4 T-3]
Fj buoyancy of the j-th stack [L4 T-3]
Fm momentum flux parameter [L4 T-2]
Fm* dimensionless momentum flux
Fmi momentum flux of i-th particle [L4 T-3]
Fr dimensionless Froude number
F* dimensionless buoyancy flux
G plume volume in Equations (79) [L3 T-1]
Gs G value at stack outlet [L3 T-1]
g acceleration due to gravity [L T-2]
kv dimensionless added mass
H upward surface sensible heat flux times g (c p ρ aϑ ) [L2 T-3]
H* dimensional parameter in Equation (57) [L]
Hb building height [L]
Hi merging point height in Equation (138) [L]
Hj height of the j-th stack in Equation (138) [L]
H max highest stack in Equation (139) [L]
H min lowest stack in Equation (139) [L]
h mixing height [L]
h′ inversion height with respect to stack top [L]
he effective stack height [L]
ht height of the base of atmospheric thermal discontinuity [L]
Lb buoyancy length scale [L]
Le effective length [L]
Lm momentum length scale [L]
Me plume momentum flux at the end of bending-over phase [L4 T-2]
N Brunt-Väisälä frequency [T-1]
N' modified Brunt-Väisälä frequency [T-1]
n dimensionless number of stacks
Pb dimensionless buoyancy flux in Equation (127)
Ps dimensionless buoyancy flux in Equation (136)
Qf total heat release rate of a flare [M L2 T-3]
Qh stack effluent heat emission rate [M L2 T-3]
Qm stack effluent mass emission rate [M T-1]
R plume radius [L]
106 Air Quality Modeling – Vol. I

Re plume radius at end of bending-over phase [L]

Rs dimensionless initial plume radius
R0 dilution radius [L]
ra radius of air eddies [L]
rs internal radius (or equivalent radius) of the stack outlet [L]
3
S wa third moment of ambient Eulerian PDF [L3 T-3]
S dimensionless parameter in Equation (137)
s stability parameter [T-2]
sc distance along the centerline [L]
Ta ambient temperature [K]
Ta0 ambient temperature at stack outlet height [K]
Te Eulerian time scale [T]
Tm Lagrangian time scale in Equation (90) [T]
Ts0 temperature of stack effluent at stack outlet [K]
Tp Lagrangian time scale of the plume [T]
t travel time [T]
tp upper edge of the plume [L]
t0 initial turbulent timescale of the plume particle [T]
Ua wind speed that can vary with height [L T-1]
Ui particle velocity component (i = 1, 2, 3) [L T-1]
Usc velocity along the centerline [L T-1]
Uz(t) mean vertical wind component [L T-1]
u mean uniform horizontal wind speed in Equation (59) [L T-1]
up horizontal particle velocity [L T-1]
u′(t) turbulent velocity fluctuation in Equation (77) [L T-1]
u* friction velocity [L T-1]
u0 mean wind speed at the stack outlet height [L T-1]
u5 wind speed at the 1.5 zs [L T-1]
va′ r.m.s. velocity of an air eddy with respect to the plume [L T-1]
vs0 effluent emission speed at stack outlet [L T-1]
va relative velocity of two particles [L T-1]
ve entrainment velocity [L T-1]
W vertical velocity of the plume [L T-1]
wb buoyancy contribution to the vertical velocity [L T-1]
wm momentum contribution to the vertical velocity [L T-1]
wp vertical velocity of the particle [L T-1]
w′(t) turbulent velocity fluctuation in Equation (66) [L T-1]
w* convective velocity scale [L T-1]
Xi particle position component (i = 1, 2, 3) [L]
X* dimensionless downwind distance
Xz plume particle’s vertical position [L]
x downwind distance from stack [L]
x* function of downwind distance from Equation (56) [L]
6 Plume Rise 107

x*c critical distance defined by Equation (36) [L]

xf downwind distance of maximum plume rise [L]
xT length defined by Equation (56) [L]
y lateral space coordinate or radial distance from axis [L]
z elevation a.g.l. [L]
′
zeq equilibrium height with respect to the stack top [L]
zs stack height [L]
α dimensionless constant in Equations (79)
αB dimensionless constant in Equations (92)
αv dimensionless constant in Section (4.3.3)
αw dimensionless constant in Equations (92)
β dimensionless classic entrainment parameter
β1 dimensionless neutral entrainment parameter in Equation (116)
β2 dimensionless stable entrainment parameter in Equation (117)
βj dimensionless jet entrainment parameter in Equation (116)
γ dimensionless constant in Equations (79)
γ1 dimensionless constant in Equation (80)
γ2 dimensionless constant in Equation (82)
∆H min maximum single plume rise from lowest stack in Equation
(139) [L]
∆H N final rise for merged plumes [L]
∆h final plume rise [L]
∆h(t ) plume rise as a function of travel time [L]
∆h( x ) plume rise as a function of distance downwind of stack [L]
∆h′ final plume rise corrected for the stack tip downwash [L]
∆h′′ actual plume rise in Equation (123) [L]
∆hd plume rise of a building downwashed plume [L]
∆hi thickness of the inversion layer [L]
∆h1 plume rise from a single stack in multiple sources [L]
∆hmax maximum plume rise [L]
∆Tc critical temperature difference [K]
∆T0 temperature difference between air and plume at the stack
outlet [K]
∆t time step [T]
∆u difference in the horizontal velocity between the plume and
the ambient environment [L T-1]
∆z vertical increment [L]
∆ϑ* dimensional parameter in Equation (57) [K]
∆ϑi potential temperature jump of the inversion [K]
∆ϑm maximum excess temperature [K]
∆ϑ100 potential temperature variation over 100 m in Section 2.7.1 [K]
δ dimensionless parameter in Equation (133)
108 Air Quality Modeling – Vol. I

∂ϑ ∂ z vertical gradient of absolute potential temperature [K L-1]

εa ambient rate of dissipation of turbulent kinetic energy [L2 T-3]
εB dissipation rate of buoyancy of the plume particle [L2 T-5]
εm dissipation rate in Equation (96) [L2 T-3]
εp rate of dissipation of turbulent kinetic energy in the plume
[L2 T-3]
εw dissipation rate of velocity of the plume particle [L2 T-3]
λ dimensionless parameter in Equation (132)
ηeq dimensionless parameter in Equation (133)
φ angle between the horizontal and the centerline [deg]
µa air molecular weight [M mol-1]
µs emission molecular weight [M mol-1]
ρa density of ambient atmosphere [M L-3]
ρa0 ambient density at stack outlet height [M L-3]
ρs density of effluent [M L-3]
ρs0 density of effluent at stack outlet [M L-3]
σy0 enhanced horizontal dispersion coefficient [L]
σz plume width [L]
σz0 enhanced vertical dispersion coefficient [L]
σ up2 longitudinal velocity variance due to the plume rise [L2 T-2]
σ vp2 crosswind velocity variance due to the plume rise [L2 T-2]
σ wa
2
second moment of ambient Eulerian PDF [L2 T-2]
σ w2 vertical wind velocity variance [L2 T-2]
σ wp
2
vertical velocity variance due to the plume rise [L2 T-2]
ϑa ambient potential temperature [K]
ϑp potential temperature of the plume particle [K]
τ dimensionless travel time

1 Introduction

The behavior of a chimney plume in the atmosphere is a rather complex process,

which is influenced by the emission characteristics, the nearby terrain features,
the actual wind profiles, stratification (vertical gradient of potential temperature)
and turbulence. Basically, plumes emitted into the atmosphere rise under the
action of their initial momentum and buoyancy (if they possess a temperature
which is greater than the ambient temperature). For power plants and other
moderate-to-large industrial sources, the major contribution to the rise is from the
heat flux. For example, a modern power plant typically discharges ≈ 100 MW of
heat from its stack. These are called buoyant plumes. In such conditions plumes
can rise for hundreds of meters. Initial momentum can be important for smaller
6 Plume Rise 109

sources, with little or no buoyancy, such as those typically found in light

manufacturing. Plumes from these sources are referred to as jet plumes.

A jet plume, moving through the ambient atmosphere, experiences a shear force
at its perimeter, where momentum is transferred from the jet to the surrounding
air. This causes an increase of the plume diameter and a decrease of its velocity.
This phenomenon is known as entrainment. In a buoyant plume, air is entrained in
the same way as in a jet and the buoyancy forces help maintain the motion of the
plume as it transfers momentum to the surrounding air. For this reason, buoyant
plumes generally rise higher than jet plumes. The entrained ambient air mixes
with the plume air, thus diluting the plume components before they reach ground
level and, in the case of buoyant plumes, decreasing the average temperature
difference between air and plume. In a calm atmosphere, plumes rise almost
vertically, whereas in windy situations they bend over. In this case, the velocity of
any plume parcel is the vector composition of horizontal wind velocity and
vertical plume velocity in the first stage and then approaches the horizontal wind
velocity.

The motion of bent over plumes can be schematically divided into three phases
(Slawson and Csanady, 1967; 1971): an initial phase, in which the self-generated
turbulence, due to the action of their mechanical and thermal energy, prevails; an
intermediate phase, where the ambient turbulence in the inertial sub-range is
important; a final phase, in which the main mechanism is the mixing due to the
large atmospheric energy containing eddies (see Figure 1).

Figure 1. Two possibilities of unstable plume behavior (adapted from Slawson

and Csanady, 1971). [Reprinted with permission from Cambridge University
Press]
110 Air Quality Modeling – Vol. I

Effective plume height he (elevation of plume centerline relative to ground level)

results from the sum of stack height zs and plume rise ∆h (Figure 2)

he = z s + ∆h (1)

Thus, a correct estimation of buoyant plume rise is one of the basic requirements for
the determination of ground level concentrations of airborne pollutant emitted by
industrial stacks. In fact maximum ground level concentration is roughly inversely
proportional to the square of the final height he. For this reason, in many

virtual
source plume
centerline
∆h

he

zs

Figure 2. The plume rise: schematic representation.

simple dispersion models, stack gases are assumed to be emitted from a virtual
source located at a height he (see Figure 2) along the vertical above the stack.

The description of plume rise is based on the fluid dynamic equations, namely on
the mass, momentum and energy conservation equations. A complete, exhaustive
theory is not yet available. Therefore some simplifying assumptions need to be
made. These will give rise to simplified models that can just take into account the
main variables of the examined case.

Plume rise formulae can be ranked either empirical or theoretical, but the
distinction is not so clear:
• the empirical formulae are based almost exclusively on experimental data
both for their numerical parameters and for their functional form
• the theoretical formulae, in spite of including some parameters with an
experimental origin, have a functional form based on the solution of
equations expressing laws of mass, momentum and energy conservation
6 Plume Rise 111

Some formulae provide the plume rise as a function of the distance, but most of
them provide a constant value (final plume rise) that the plume reaches at a large
downwind distance. These formulae contain height depending atmospheric
variables normally specified at the stack outlet height.

Several studies and review works have provided semi-empirical formulae for
evaluating ∆h (e.g., Holland, 1953; Brummage, 1966; Bringfelt, 1969; Fay et al.,
1970; Carpenter et al., 1971; Briggs, 1975; Strom, 1976; Hanna et al., 1982; and
many others); others have provided more complex and comprehensive
descriptions of several physical interactions between the plume and the ambient
air (e.g., Golay, 1982; Netterville, 1990). Relevant and exhaustive review papers
on the plume rise subject can be found in the literature, such as, for instance,
Briggs (1975) and Weil (1988). In this chapter, we will utilize a great deal of
material from these reviews.

For many specific applications, literature supplies functional forms and

empirically determined parameters, but such models may provide wildly
inaccurate results, if they are used beyond the context where they have been
obtained. In uncertain cases, Briggs (1975) recommends to use, in the application,
the formula that provides the minimum plume rise; this result is “the most
conservative”, since it gives rise to the maximum values of concentration
expected at the ground, thus limiting the risk of a possible underestimation. It is
hard to specify clearly the accuracy of plume rise formulae: some discordance up
to 25% between the observed and the expected value are not unusual.

This chapter, which is concerned with plume rise from continuous releases,
focuses on:
• semi-empirical formulations
• advanced plume rise models
• particle models for plume rise
• special cases (like building downwash, penetration of elevated inversion,
multiple source, flare stacks, fires and so on)

The semi-empirical formulations, expressed as analytical relationships, have a

functional form obtained from the solution of mass, momentum and energy
conservation equations in simplified conditions (such as steady conditions,
uniform wind and stability) and their numerical parameters are generally deduced
from experimental data. These are the plume rise estimations mostly used in
regulatory model applications.

In the advanced plume rise models the conservation equations are numerically
integrated, thus giving practical solutions for varying winds and thermal structure.
Due to present days computer capabilities, these models may also be used for
regulatory applications. However it cannot be automatically accepted that these
fully 3D models always yield results better than simpler models due to the
112 Air Quality Modeling – Vol. I

difficulty, in some applications to real cases, of getting the needed input data with
the necessary time and space resolution.

The particle models for plume rise are relatively new methods, not yet widely
used for regulatory purposes. We think that it is important to present and discuss
them in some detail since the Lagrangian approach is a more natural way of
describing the dispersion process (Sawford, 1985). Furthermore these are
probably the methods of the future and allow a high resolution, particularly the
small time behavior of plume dispersion (Nguyen et al., 1997)

The section on special cases covers many aspects of the plume rise phenomenon
that are of practical importance in many applications.

Our discussion of plume rise addresses fundamental aspects and major problems,
but it is not exhaustive. We intend neither to make any ranking of the models
presented in the next sections, nor to recommend which model is the best for a
specific application, because we want to avoid any subjective judgment which
may be also influenced by the particular national regulatory laws. We wish to
present a review of updated and validated existing techniques that can be used by
modelers according to their specific needs. Even if no guidance is given whether a
reader should use one of the models out of those suggested for a specific problem,
the general rule might be to preferably use those formulas, if any, that are
validated, recommended or suggested by National Environmental Protection
Agencies. Being used and tested by hundreds of users, these models will, at least,
guarantee that the major bugs and/or uncertainties were identified and amended
and unrealistic results avoided.

2 Semi-Empirical Formulations

2.1 Governing Equations

As above anticipated, the differential equations expressing the conservation of the

total fluxes of mass, momentum, and energy through a plume cross section (e.g.,
Morton et al., 1956; Briggs 1975; Weil, 1988) are the basis of all the analytical
plume rise models. These equations are closed using the entrainment assumption
(Morton et al., 1956), which prescribes that the entrainment velocity, i.e. the rate
at which ambient air is entrained into the plume, is proportional to the mean local
rise velocity. It may be worth mentioning that Priestley (1956) provided an
alternative entrainment assumption, based on energy arguments, that gives the
same basic plume rise results as Morton et al. (1956).

The following simplifications are made: the plume rises in a steady, horizontal
wind of constant direction and variable with height speed Ua(z); stratification, if
present, is constant with the height; plume cross section is circular with radius R;
plume properties (mass, velocity, temperature) have a “top hat” distribution (that
6 Plume Rise 113

is to say in each section the cited quantities are constant inside the plume and null
outside); the plume pressure is the same as in the local environment; the density
differences are sufficiently small to allow making the Boussinesq approximation;
since the efflux volume quickly mixes with a large volume of ambient air, the
effluent has the same molecular weight and specific heat as air.

Figure 3. Schematic and nomenclature for plume in a crosswind (adapted

from Weil, 1988). [Reprinted with permission from American Meteorological
Society].

In the case of a crosswind, conservation of mass, horizontal momentum, vertical

momentum and energy are given respectively by (Weil, 1988)

d
dsc
( )
U sc R 2 = 2 RβW (2)

d
ds c
( )
U sc R 2 ∆u = − R 2W
dU a
dz
(3)

d
( ) (ρ − ρ s )
U sc R 2W = gR 2 a (4)
dsc ρs

dFb
= sWR 2 (5)
ds c
114 Air Quality Modeling – Vol. I

where (see also Figure 3): sc and Usc are the distance and velocity along the
centerline, β is the dimensionless entrainment parameter, W is the plume vertical
velocity, ∆u is the difference in the horizontal velocity between the plume and the
ambient environment, g is the acceleration due to gravity, ρa and ρs are the air and
plume density, respectively, Fb is the buoyancy flux defined by

Fb = U sc R 2 g
(ρ a − ρ s ) (6)
ρa

and s is the stability parameter defined by

g ∂ ϑa
s= (7)
ϑa ∂ z

ϑ a being the potential temperature of the air and ∂ϑ a ∂ z its vertical gradient.
The plume trajectory, ∆h( x ) , can be obtained from the above equations and from
the following relationships (kinematic conditions)

d∆h W dx U a + ∆u
= , = (8)
ds c U sc dsc U sc

For sake of completeness, and for reference to previous and/or related work on
the plume rise, the following remarks may be important.

Fb is related to the heat emission rate Qh by the following relationship

Fb = gQh π c p ρ a Ta (9)

where Ta = temperature of the air;

cp = specific heat of air at constant pressure.

Qh is given by the following equation

Qh = Qm c p ( Ts 0 − Ta 0 ) (10)

where Qm = effluent mass emission rate;

Ts0 = absolute temperature of effluent at stack outlet;
Ta0 = absolute temperature of ambient atmosphere at stack outlet height.

Qm may be expressed in terms of other variables as follows

6 Plume Rise 115

Qm = ρ s 0 As vs 0 (11)

where ρ s 0 = mass density of effluent at stack outlet;

As = stack outlet area;
vs0 = effluent emission speed at stack outlet.

Using Equation (11) and the following equality

ρ s 0 (Ts 0 − Ta 0 ) = Ta 0 ( ρ a 0 − ρ s 0 ) (12)

Qh can be expressed in terms of the mass density of ambient air at stack outlet
height, ρ a0, and of the pollutants, ρ s 0 , as follows

Qh = As vs 0Ta 0 c p ( ρ a 0 − ρ s 0 ) (13)

In the plume rise formulae for jet plumes, where the initial momentum plays the
major role in the rising process, Fb is substituted by the momentum flux Fm, given
by the following equation

As ρ s 0
Fm = vs20 (14)
π ρa0

Notice that the buoyancy and momentum fluxes - Equations (6) and (14) - by
convention are divided by π . This convention derives (Briggs, 1984) from the
assumption of round top hat profile of all the plume quantities in the early plume
rise studies. This assumption leads to the presence of π on both sides of the flux
conservation equations.

In some plume rise formulations (e.g., Hewett et al., 1971), the buoyancy flux is
defined by Fb = U sc R 2 g ( ρ a − ρ s ) ρ s . According to Briggs (1972), this
definition is equivalent to assuming that the buoyant force acts on a fluid of
density ρ s . However, the density of the fluid driven by the buoyant force is better
approximated by ρ a , since a turbulent plume is made up mostly of entrained fluid
(Briggs, 1984).

In some formulae, the Brunt-Väisälä frequency, N, defined as

N =+ s (15)

is used instead of the stability parameter s - see Equation (7). In stable conditions,
the Brunt-Väisälä frequency is the natural frequency of oscillation of a fluid
particle if perturbed from its equilibrium position; for plumes, N −1 is the time
116 Air Quality Modeling – Vol. I

scale for the depletion of the buoyancy flux and for the maximum rise in a stable
environment (in the atmosphere a typical value of N −1 is 1 min).

The formulae that are going to be presented in the following sections are mostly,
but not exclusively, derived from Briggs (1969, 1972, 1975, 1984). Briggs
formulae, together with some new result recently appeared in the literature due to
other authors, have been incorporated into most of the U.S. EPA models
([Link] These formulae represent a reasonable
compromise between accuracy and simplicity, even though, according to many
(e.g., Henderson-Sellers and Allen, 1985), they may tend to overestimate the
plume rise at large downwind distances. Note that some authors (Strom, 1976;
Hanna, 1994) suggest, in the absence of particular expressions derived for
specific problems, using Briggs formulae.

It is worthwhile pointing out that Manins (1985) summarized evidence (from

large fires, volcanic eruptions and clouds from thermonuclear explosions)
showing that the vertical plume rise equation derived in the next section -
Equation (45) - holds for over four orders of magnitude variation in rise height.

2.2 Plume Rise in the Transitional Phase

By solving the system of equations presented in Section 2.1, simple analytical

expressions, easy to use in dispersion models, can be achieved.

2.2.1 Neutral and Unstable Case

Let us firstly consider the rise of a bent over plume in neutral conditions (s = 0)
and uniform wind (no shear) and neglect ambient turbulence. At some distance
from the source, plume can be considered as nearly horizontal and the following
approximations can be made: U sc ≅ U a = const and W << Ua. From Equation (2)
it follows that the plume radius grows linearly with height. In these stability
conditions, Equation (5) implies that the buoyancy flux Fb is conserved. Thus, Fb
is expressed by means of its value at the stack outlet, in terms of, respectively,
temperatures and densities

As (Ts 0 − Ta 0 )
Fb = g vs 0 (16)
π Ts 0
As ( ρ a 0 − ρ s 0 )
Fb = g vs 0 (17)
π ρ a0

Similarly, Fm is expressed by
6 Plume Rise 117

Ta 0
Fm = v s20 rs2 (18)
Ts 0

ρ s0
Fm = v s20 rs2 (19)
ρ a0

where rs is the internal radius of the stack outlet.

Note that, usually, in Equations (16) and (17), instead of As, we find rs (or, for a
non-circular stack having an area As, the equivalent radius given by rs = As π ).
It may be worth pointing out that for the emissions whose molecular weight, µ s,
differs considerably from the air molecular weight, µ a, Hanna et al. (1982)
suggests, in relation to Equation (16), replacing Ts0 with Ts 0 µ s and Ta0 with
Ta 0 µ a .

The resulting equation of the plume centerline trajectory ∆h(t ) is (Briggs, 1975)

13
 3 F 3 Fb 2 
∆h(t ) =  2 m t + t  (20)
 β u0 2 β 2 u0 

Equation (20) takes into account both the buoyancy and initial vertical momentum
contributions. In the very initial stage, the momentum dominates and the plume
rise is described by

13
 3 F 
∆h(t ) =  2 m t  (21)
 β u0 

whereas, when t is larger than 2 Fm Fb (about 10 s for many sources, Briggs

1975) the buoyancy dominates and Equation (20) reduces to

13
 3 F 
∆h(t ) =  2 b  t 2 3 (22)
 2 β u0 

Expressed as a function of downwind distance, Equations (20), (21) and (22)

become

13
 3 Fm x 3 Fb x 2 
∆h( x ) =  2 2 + 
 (23)
 β u0 2 β 2 u 03 
118 Air Quality Modeling – Vol. I

13
 3 
∆h( x) =  2  Fm1 3 x1 3u0−2 3 (24)
β 

and

13
 3 
∆h( x) =  2  Fb1 3 x 2 3u0−1 (25)
 2β 

In order to use the above equations in practical models, the value of β, the
dimensionless entrainment parameter, must be empirically established. In the bent
over buoyant plumes (φ → 0, where φ is the angle between the horizontal and the
centerline) β = 0.6, whereas for vertical buoyant plumes (φ → 900) β = 0.11
(Briggs, 1975; Hoult and Weil, 1972). For jet plumes (Briggs, 1975, 1984)

u0
β = 0.4 + 1.2 (26)
vs 0

In particular, Equation (25) becomes

∆h( x ) = 1.6 Fb1 3 x 2 3 u0−1 (27)

This equation is widely known as the “two-thirds” law. It was confirmed by a large
amount of experimental work (see Briggs, 1975 for a comprehensive summary). Figure 4
is an example of the quality of the agreement found in the literature. Consequently, most
practical models use the “two-thirds” law to describe the plume rise in the transitional
phase under neutral and unstable conditions. However some models - see, for instance,
AERMOD (U.S. EPA, 1998) or Weil et al. (1997) - use the complete Equation (20). In
this case they use the value β = 0.6 for the momentum term too.

By defining the momentum length scale Lm and the buoyancy length scale Lb as

Fm Fb
Lm = and Lb = (28)
u0 u 03
6 Plume Rise 119

Equations (23 - 25) become

Figure 4. Observed trajectories of buoyancy-dominated plumes compared with the

“two-thirds” law (from Weil, 1988). [Reprinted with permission from
American Meteorological Society ].

13
∆h( x )  3  Lm
2
 x 3  x 
2

= 2   +    (29)
 β  Lb 2 β
2
Lb  Lb  Lb  

13
∆h( x )  3 x 
=  2  (30)
Lm  β Lm 

13
∆h( x )  3 x2 
=  
2 
(31)
 2 β Lb 
2
Lb

Lm and Lb allow an alternative criterion to establish in which range of downwind

distances the buoyancy or the momentum dominates the plume rise: momentum
dominates for x << L2m Lb whereas buoyancy dominates for x >> L2m Lb .
120 Air Quality Modeling – Vol. I

2.2.2 Stable Case

In case of a stable atmosphere (s = const, s ≠ 0), again neglecting ambient

turbulence, and considering uniform wind and the bent-over phase, integration of
the conservation equations leads to the following expression (Briggs, 1975)

1 1
 3(1 + k )   '  x  
3
3
 ' x  
∆h( x ) =  2 v   N Fm sin N  + Fb 1 − cos N '   (32)
 β u0 s    u 0    u 0  

1
for x ≤ πu 0 N , where N = s (1 + k v ) . The term, 1 + k v , accounts for the so-
1 2
' ' 2

called “added mass” (Briggs, 1972, 1975, 1984; Weil, 1988, 1994). This added
mass takes into account the momentum of the ambient air displaced by the rising
plume. Consequently, the effective plume radius is larger than the visible plume
radius. Many models did not consider this aspect that can explain the difference
found in the value of the entrainment parameter from the measurements of plume
rise in different stability conditions. Concerning the numerical value of the added
mass, Briggs (1975) and Weil (1994) suggest 1 + k v = 2.25.

When the atmosphere is stable, ambient turbulence is very low and a plume levels
off where its density difference with respect to ambient air approaches zero. For
distances greater than πu0 N ' a plume, in principle, overshoots its equilibrium
height and displays a quickly damped oscillation. This was experimentally
verified in some occasions (Briggs, 1975). However, often plumes approach an
asymptotic height with no overshoot at all (Briggs, 1984). In this range of
distances a plume drifts downwind with a very small increase in thickness, due to
its mixing with stable, almost non turbulent, air.

In the two asymptotic cases in which the momentum dominates ( Fb << N ' Fm ) or
the buoyancy dominates ( Fb >> N ' Fm ), the above equation reduces to,
respectively

1
  ' x  3

 3(1 + k v ) 2 Fm sin N
1

  u 0  
∆h( x ) = (33)
 1 
 β 2 u0 s 2 
 
6 Plume Rise 121

1
   ' x  
3

 3(1 + k v )Fb 1 − cos N  

   u 0  
∆h( x ) =   (34)
 β u0 s
2

 
 

2
 x  x x 1
Notice that, by approximating cos N '  as 1 −  N '  for << ' in
 u0   u0  u0 N
Equation (32), the “two third law” - Equation (27) - is recovered.

2.3 Formulae for the Final Height of Buoyant Plumes

Stable stratification is the only condition in which a plume levels off and, consequently,
the definition of a final height is correct. Since in neutral and unstable conditions the
buoyancy flux Fb is conserved, plumes cannot, in principle, level off. However,
ambient turbulence significantly affects the buoyant plume growth. Another
limitation to the continuous rise of the plume is the presence, above the mixing
height, of a capping inversion (see Section 5.2).

The importance of assessing correct ways to determine the plume “final height”
derives from the wide use of Gaussian models in dispersion calculations. These
dispersion models disregard the transitional phase and assume that a plume is
emitted by a virtual height (see Figure 2) located at a final effective height he,
given by the sum of stack height zs and plume final rise ∆h - see Equation (1).

2.3.1 Neutral and Unstable Case

For neutral or unstable conditions, Briggs (1969) suggested using Equation (27)
up to x = xc* , and the following equation

 2 16 x 11  x 
2
 
−2

∆h( x ) = 1.6 F u b
1 3 −1
0 (x )
* 23
c  + *
+  * 
4x
 1 + *  (35)
 5 25 xc 5  xc   5 xc 

for x > xc* . xc* is a critical distance representing the downwind distance at which
ambient turbulence begins to dominate the entrainment process, which can be
expressed either by

xc* = A1 Fb2 5 z s3 5 z s < 305m

xc* = A2 Fb2 5 z s > 305m (36)

or
122 Air Quality Modeling – Vol. I

xc* = A3 Fb5 8 Fb < 55 m 4 s −3

xc* = A4 Fb2 5 Fb > 55 m 4 s −3 (37)

in which the values of the four dimensional constants are the

following: A1 = 2.16 s 6 5 m −6 5 , A2 = 67.0 s 6 5 m −3 5 , A3 = 49.0 s 15 8 m −3 2 , and
A4 = 119.0 s 6 5 m −3 5 .

In the case of fossil fuel plants with Qh of 20 MW or more, xc* can be satisfactory
approximated by the following equation

xc* = 10 z s (38)

Subsequently, Briggs (1975) made a distinction between neutral and unstable

conditions accounting for the effects of ambient turbulence on the plume rise.
While self-generated turbulence affects the entrainment process near the source,
ambient turbulence (with both small and large scale eddies) becomes important
further downwind. Small scale eddies (with typical length scale ≤ R ), are
responsible for the increase of the plume growth rate beyond that given by self-
induced turbulence. The breakup model (Briggs, 1984; Weil, 1988), assumes that
plume rise finishes when ambient turbulence “breaks up” the self-generated
structure of the plume, causing a vigorous mixing and, consequently, plume
gradually loses buoyancy and momentum and eventually level off. Thus, this
process leads to an asymptotic rise. According to Briggs, the plume breakup
occurs when the ambient rate of dissipation of turbulent kinetic energy, εa,
exceeds the one of the plume εp. Large scale eddies (updrafts and downdrafts in
the CBL) may transport plume segments up and down, thereby dispersing the
plume by vertical meandering and pushing some of them to the surface. When this
happens, the time averaged ground level concentration is more dependent on how
many times, during the averaging period, the plume touches the ground than on
the height of the asymptotic rise. As a consequence, in the CBL case the surface
sensible heat flux, which plays the major role in the development of updrafts and
downdrafts is assumed to be the leading parameter.

Therefore, for neutral conditions, in which the rise is limited by the mechanical
ambient turbulence, Briggs proposed the iterative formula

35
 F 
∆h = 1.2  b 2  (z s + ∆h )2 5 (39)
 u0 u* 

where u* is the friction velocity. For unstable conditions, in which the

termination of the rise is due to the breakup by plume-scale, Briggs (1975, 1984)
proposed
6 Plume Rise 123

35
F 
∆h = 3.0  b  H −2 5 (40)
 u0 

where H is the upward surface sensible heat flux times g (c p ρ aϑ ) . This equation
may be also written as

∆h = 3.0 F* 3 5 h (41)

where

(
F* = Fb u0 w*2 h ) (42)

is the dimensionless buoyancy flux, w* is the convective velocity scale and h is

the mixing height. Notice that Hurley and Physick (1993) derived an expression
similar to (41), but with the constant equal to 2 instead of 3. Examining the Willis
and Deardorff (1983) CBL water tank experiment, they also found that a value of
2 gave better agreement than the value 3. Best et al. (1990, reference from Hurley
and Physick, 1993), found 2.3 by fitting field ground level concentration (g.l.c.)
data. It is difficult to decide which constant value is to be preferred due to the lack
of direct measurements of the final height in this conditions and to the large
scatter in the indirect methods (g.l.c.).

2.3.2 Stable Case

Let us consider Equation (32). This equation has its maximum, ∆hmax , for
x f = πu 0 N ' . Considering that for most hot plumes the effect of the initial
momentum can be neglected, and that the leveling off or equilibrium height is
observed to occur at about 5/6 ∆hmax , the following expression for the final height
in stable and windy conditions (u0 > 1 ms −1 ) is obtained

∆h = 2.8 ( Fb u 0 s )1 3 (43)

However, Briggs (1975; 1984) on the analysis of many field and laboratory
observations, found that a slightly different numerical coefficient, 2.6, yielded the
best fit to the observation. Consequently, Briggs recommended that the most
accurate estimate of the plume final height in stable conditions is given by

∆h = 2.6 ( Fb u0 s )1 3 (44)
124 Air Quality Modeling – Vol. I

For stable and calm conditions (u0 < 1 ms −1 ), in which a plume rises nearly
vertically, Briggs (1969), on the basis of previous work of Morton et al. (1956)
and the examination of many field observations, proposed

∆h = 5.0 Fb1 4 s −3 8 (45)

Subsequently, Briggs (1984) proposed for the same conditions

∆h = 5.3 Fb1 4 s −3 8 − 6 rs (46)

2.4 Formulae for Jet Plumes in the Transitional Phase

Also these formulae, as the ones for buoyant plumes, have a semi-empirical
origin.

The formulae for the transitional phase of jet rise, in neutral/unstable or stable
conditions, were already introduced, namely Equations (24) and (33), in which
the entrainment parameter β was defined by Equation (26).

2.5 Formulae for the Final height of Jet Plumes

For neutral conditions Briggs (1969) previously suggested

3v s d s
∆h = (47)
u0

where ds is the internal diameter of the stack outlet, and later (1975, 1984)
suggested

12
0.9  Fm 
∆h =   (48)
β  u0 u* 

For unstable conditions Briggs (1975) suggested

37
F 
∆h = 2.3  m  H −1 7 (49)
 u0 

and, subsequently, Briggs (1984) suggested for the same conditions

37
1.3  F 
∆h = 6 7  m  h1 7 (50)
β  u0 w* 
Also in Equations (48) and (50) β is defined by Equation (26).
6 Plume Rise 125

Briggs (1969, 1975) suggested for stable and windy conditions ( u0 > 1 ms −1 )

∆h = 1.5 ( Fm u0−1s −1 2 )1 3 (51)

This is obtained from considering that Equation (33) attains its maximum ∆hmax
for x f = πu0 (2 N ' ) and that the equilibrium height occurs at about 2/3 ∆hmax .

For stable and calm conditions ( u0 < 1 ms −1 ), Briggs (1969, 1975) suggested, on
the basis of a few observations, the following relationship

∆h = 4.0 ( Fm s −1 )1 4 (52)

Taking into account that observations on the rise of jet plumes in stable
conditions are very sparse, Briggs suggests considering the above formulae as
tentative.

2.6 Buoyant Plumes or Jet Plumes

While for jets plumes or for highly buoyant plumes it is clear which type of plume
rise formulae is to be used in practical dispersion applications, for cases in which
∆T0 = Ts 0 − Ta0 (where ∆T0 is the temperature difference between emission and
ambient air at the stack mouth) is greater than zero but not very high, whether the
plume rise is dominated by momentum or by buoyancy must be determined. Two
methods able to solve this problems are presented. The first one - see, for instance,
AERMOD (U.S. EPA, 1998) - consists in using Equation (23) or (32) for the
transitional phase in neutral/unstable or stable case, respectively. These cited
equations include both contributions. The final plume height is calculated
according to the methods resumed in Section 2.3.

The second one is based on the U.S. EPA models PTPLU (Pierce et al., 1982),
SCREEN3 (U.S. EPA, 1995a), and ISC3 model (U.S. EPA, 1995b). In this method
a critical temperature difference ∆Tc is defined. If ∆T0 > ∆Tc the plume has to be
treated as buoyant; otherwise the plume has to be treated as a jet. ∆Tc is defined
as:
• in stable atmosphere

∆Tc = 0.19 vs 0Ta 0 s 1 2 g −1 (53)

126 Air Quality Modeling – Vol. I

• in neutral or unstable atmosphere

0.29 v s103Ts 0 d s−2 3 g −1 if Fb < 55 m 4 s -3

∆Tc =  (54)
0.056 v s203Ts 0 d s−1 3 g −1 if Fb ≥ 55 m 4 s -3

The units of the two dimensional coefficients in Equation (54) are m 4 3 s −5 3 and
m 2 3 s −4 3 respectively. Equation (53) is obtained by equating Equations (44) and
(51) and solving for ∆T0 . Similarly, Equations (54) are obtained by inserting
Equations (37) into (27) and then equating Equation (47).

2.7 Moore's and Netterville's Models for the Plume Rise of Buoyant
Plumes

The formulae for the buoyant plume rise presented in the previous part of this
Section form a complete and validated set of equations, widely used by the
modeler community. However we briefly introduce also the Moore and
Netterville models because we think that these two models are important for
scientific and/or historical reason. As such, they may be of interest for the reader.
Moreover, Netterville's parameterization of ambient turbulence is used later in the
chapter (see Sections 3.2 and 4.4).

2.7.1 The Moore Model

Briggs models are, as we have seen above, derived by assuming that a plume is
continuous, accounts for the crosswind and vertical spread and disregards the
along wind spread and its diameter increases as it rises and travels downwind. For
this reason this kind of approach is called “two-dimensional”. Instead Moore,
(1974) points out that the dilution of an hot smoke plume is a three-dimensional
phenomenon, because the plume, rather than rising as a continuous cone, breaks
up into a discrete series of puffs which tend to recombine and merge into each
other as the plume travels downwind, so that the number of puffs per unit length
of plume decreases with downwind distance. The problem becomes three-
dimensional because the along wind spread must be considered as well. One of
the main interest in the Moore's model is in the recognition that observations of
stack plumes sometimes reveal some three-dimensional features (Ooms, 1972)
either due to its dynamic (formation of two counter-rotating vortices as it leaves
the stack which may cause the plume bifurcation, split of the plume in lumps) or
to terrain characteristics in case of low emissions.

Moore model is a generalized one that can be applied in a large variety of

meteorological situations both during the transitional and final stage of rise
without switching to various different expressions.
6 Plume Rise 127

The basic difference between the two-dimensional and the three-dimensional

approach is that in the former the plume rise is proportional to Fb1 3 , that is to
Qh1 3 , see Equations (9) and (27), whereas in the latter it results proportional to
Qh1 4 , namely

∆h( x) = AQh1 4 x*3 4 u1−.51 (55)

where: u1.5 is the wind speed at the height 1.5 z s ; x* = x for short distances and
x* = xT for large distances. These two asymptotic values are connected by a
smooth curve possessing the correct asymptotic and near field limiting forms:

xT x
x* = (56)
xT2 + x 2

and x T is given by the following expression

( D0 + D1H *)u1.5
xT = (57)
∆ϑ *
( D0 + D1H *)2 2 + u12.5
D2

A, D0 and D2 are dimensional constant, whereas D1 is dimensionless. Their values

were estimated by Moore to be: A = 2.4 MW −1 4 m 5 4 s −1 for z s > 120 m while, for
∂ϑ ∂ z
z s < 120 m , A = 2.4 for very stable conditions (i.e. for 2
> 2.5 K s 2 m −3 ),
u1.5
otherwise A = 2.4(0.16 + 0.007 z s ) ; D0 = 1920 m, D1 = 19.2, and D2 = 120 ms.
H * and ∆ϑ * are parameters related, respectively, to the following two
assumptions: 1) the atmospheric turbulence effects on the first steps of the plume
rise evolution are dependent on the height for low sources ( z s < 120 m ), but
independent of the height for high sources ( z s > 120 m ); this assumption is
parameterized setting: H * = z s if z s < 120 m , H* = 120 m if z s > 120 m ; 2) the
atmosphere is assumed to be stably stratified, even in convective conditions; since
∆ϑ100 is the variation in potential temperature per each 100 m of height increase,
this assumption is parameterized setting: ∆ϑ * = 0.08 K if ∆ϑ100 < 0.08K ,
∆ϑ* = ∆ϑ100 if ∆ϑ100 > 0.08K .

Moore claimed that his model is applicable when the difference in temperature
between effluent and air ranges between 80 and 150 K, the effluent emission
velocity vs0 does not considerably overtake the value of 30 m s −1 and in the
following conditions: x > 400 m ; 30m < z s < 230m ; 10 MW < Qh < 150 MW .
128 Air Quality Modeling – Vol. I

An explicit expression of the final height is obtained by inserting x* = xT 2 in

Equation (55).

2.7.2 The Netterville Model

Netterville (1990) gave a detailed description of the entrainment process, which

he called the “two-way model”. This is based on an understanding of several
quantitative aspects of turbulent mixing within free shear layers and on the
availability of more detailed ambient turbulence data from remote sensors, like
SODARS and RASS.

The two-way entrainment model predicts that the turbulent atmosphere must
entrain plume material just as the plume must entrain the atmosphere. Figure 5
(from Netterville, 1990) illustrates what is meant with two-way entrainment. It
shows a turbulent plume of radius R that rises at relative speed W through an
atmosphere containing turbulent eddies of length scale ra and relative root-mean-
square velocity v ′a . The plume cross-section is assumed to have a ‘spongy’
internal structure caused by atmospheric turbulence eddies, in transit through the
plume, that form transient holes in the surrounding matrix of turbulent plume
material. Similarly, also the ambient air eddies become spongy due to penetration
by the plume’s internal turbulent eddies.

The entrainment process is split into three processes: direct entrainment (a

process by which plume eddies, due the self-generated turbulence, capture
ambient air masses), indirect entrainment (ambient air eddies in the plume that are
in turn penetrated by eddies of the internal plume) and extrainment (transfer of
plume mass from the plume itself to ambient air due to the turbulent eddies that
enter the plume and carry off plume mass).
6 Plume Rise 129

Figure 5. Schematic of plume/atmospheric interaction (from Netterville,

1990). [Reprinted with permission from Elsevier Science]

Netterville plume rise model is based on this description of entrainment process,

and on the solution of the mass, momentum and energy conservation equations.
The same simplifying assumptions on the plume shape and atmospheric
conditions as in the Briggs models are made, thus obtaining the following
scheme.

In stable atmosphere ( N 2 > 0)

 3    N  f  N 
∆h(τ ) =  2  Fe + f b M e +  NM e sin τ  − b cos τ  +
(
 β u f b + s
2
)     f b  N  f b 
(58)
13
 N  f  N    R 
3
 Re
− Fe cos τ  + b sin τ   exp(− τ ) +  e   −
  f b  N  f b     β   β

where τ = t f b is the dimensionless plume travel time given by the product of the
plume travel time, t, and the atmospheric turbulence buffet frequency, fb. This last
is defined as

f b = 2 βσ u (uτ E ) (59)
130 Air Quality Modeling – Vol. I

in which τ E is the Eulerian integral time scale of atmospheric turbulence and σ u

is the standard deviation of longitudinal horizontal wind.

In neutral atmosphere ( N 2 = 0)

 3
∆h(τ ) =  2 2 [Fe + f b M e − { f b M e + Fe (τ + 1)}exp(− τ )] +
 β fb u
13 (60)
R 
3
 Re
+  e   −
β   β

In unstable atmosphere ( N 2 < 0, N ≡ + N 2 )

 3    N  f  N 
∆h(τ ) =  2 2  Fe + f b M e − NFm sinh τ  − b cosh τ  +
(
 β u f b + s )     fb  N  f b 
(61)
13
  N  fb  N     R 3  R
+ Fe cosh τ  + sinh τ   exp(− τ ) +  e   − e
  fb  N  f b     β   β

As t → ∞, all three solutions asymptotically approach the same functional form

for final rise

3 13
 3  F + f M R   Re
∆h =  2  e 2 b e +  e   − (62)
 β u  f b + s β ( )   β

In the above equations, Fe, Me, and Re are initial values of plume buoyancy,
momentum and radius at the end of bending-over phase. Djurfors (1983) has
shown that they are given by
12
 v T 
Re ≅ rs  2 s 0 a  (63)
 u Ts 0 

v T 
M e = Fm  s 0 a  (64)
 u Ts 0 

Fe = Fb (65)

For stable conditions (s > 0) the final rise is always finite. For neutral conditions
(s = 0) the final rise is finite only if the atmosphere is turbulent (fb > 0). For
unstable conditions (s < 0) the final rise is finite or infinite depending on the sign
6 Plume Rise 131

( )
of f b2 + s , i.e. on whether thermal instability or atmospheric turbulence
dominates plume motion.

The Netterville plume rise model is consistent with the ‘decay constant’ approach
of Djurfors (1977), which recognized that the mathematical form of leveling-off
behavior was one in which the vertical distance between the plume centerline and
its final height would decrease exponentially with time.

The validation of this model is based on one data set of LIDAR measurements.

3 Advanced Plume Rise Models

3.1 Introduction

The semi-empirical formulations presented in the previous section have shown,

on several occasions, a great degree of uncertainty. This is partly due to the
simplifications introduced in such formulations. Advanced methods, based on the
numerical integration of a set of differential equations expressing the conservation
equations and on revised entrainment assumptions, have been proposed. They
account explicitly sufficient transport mechanisms to be of general use,
particularly in the cases that are too complicated to be modeled by simple
analytical models. They provide, at least in principle, a better physical
representation of the two basic phenomena related to plume rise: the grow of the
plume centerline and the entrainment of ambient air into the plume and its
consequent horizontal and vertical spreading. They also allow dealing with
complex atmospheric conditions. However, they require more computational
resources and more detailed input data.

Among the advanced models we may distinguish: integral models (they use
spatially integrated forms of the fluid motion equations), differential models (they
integrate on Eulerian grids Reynolds-averaged flow conservation equations) and
large eddy simulation (LES) models. In all these models, the system of equations
must be closed by a proper number of assumptions and closure hypothesis.
Essentially they are empirical, but are based on physical reasoning and/or
observations.

The first two categories, that are not so computationally costly (particularly for
nowadays computers) are not only of scientific interest, but may also be useful
tools in air pollution modeling, since they are able to deal with any kind of stack
plume (jet, dense or buoyant plume) and complex atmospheric structures.

We would like to stress the importance of initialization in numerically solving the

plume rise equations. Stack geometries and plume exit temperatures and
132 Air Quality Modeling – Vol. I

velocities may vary over wide ranges. Consequently, large errors in the plume rise
estimations can be made if the initial conditions are not correctly formulated.

Many advanced models have been developed. It is hopeless to review them all,
therefore only some of them are briefly presented below. We would also mention
that a number of the most recent advanced plume rise models are based on
Lagrangian particle techniques. These last models for plume rise are discussed in
Section 4.

3.2 Integral Models

The models developed by Schatzmann (1979), Ooms and Mahieu (1981),

Glendening et al. (1984), Chiang and Sill (1985), Gangoiti et al. (1997) and
Janicke and Janicke (2001) are considered here. They give an overview of these
kind of models developed during the last 20 years. There are some characteristics
common to all these models. Since the trajectory of a plume (jet, dense or
buoyant) in windy conditions is not a straight line, they generally use the natural
coordinate (or curvilinear) system that moves and rotates as it follows the plume
centerline trajectory, rather than the Cartesian coordinate system. They do not use
the common Boussinesq approximation thus allowing the treatment of plumes
with greatly different density from that of ambient air. The plume is assumed to
exhibit local similarity, i.e., the shapes of the radial profiles of excess velocity,
temperature and concentration do not change downstream. The profiles of plume
velocity, temperature, and density are assumed to be of Gaussian (Schatzmann,
1979; Ooms and Mahieu, 1981; Chiang and Sill, 1985), “top hat” (Glendening et
al., 1984) or exponential (Gangoiti et al., 1997) shape for mathematical
simplicity. Models do not use different parameterizations for each phase
(buoyancy dominate, intermediate and turbulence dominated) of the plume
trajectory. Additional assumptions are steady state conditions for both plume and
environment, zero environmental vertical velocity and absence of stack
downwash effects (see Section 5.1), which is appropriate for plumes with large
buoyancy, and exit velocity. In most models it is assumed that the mean excess
and turbulent quantities plume are axisymmetric and, consequently, that the three-
dimensionality of the plume motion can be ignored. Although it is recognized that
two counter rotating vortices are formed at the stack mouth exit and that the
plume may break into distinct puffs (that may also merge downwind), these
effects are neglected since they are assumed to be incorporated in some way in
the entrainment formulations. All the models were tested against laboratory and
field data.

Basically, the main difference among the various models lies in the modeling of
entrainment, i.e. the rate of mixing of ambient air into the plume. Other
characteristics that make different the models are the inclusion in some of them of
pollutant dispersion besides the path and spread of the plume or the capability of
some models to treat arbitrary atmospheric structures, whereas the others should
divide the atmosphere in a certain number of layers with different constant
6 Plume Rise 133

atmospheric properties. Only a few models accounts for the plume rise
modifications due to the condensation of plume water vapor.
The plume rise model of Schatzmann (1979) assumes that the wind velocity is
constant in value and direction and that the atmosphere is stratified with a
constant density gradient. It includes seven equations for the following seven
unknowns: centerline excess velocity, temperature and concentration, centerline
density defect, jet radius, angle of inclination and local rate of entrainment. A
rather complex entrainment function is used which is based on the local
densimetric Froude number, the plume radius, the macroscale of the energy-
containing eddies, angle of the plume trajectory, free-stream velocity, centerline
excess velocity and five empirical constant. The tests of the model performances
against many observations including jets, buoyant and dense plumes, gave
reasonable agreement. This model, however, fails to account for the inertia of
“effective mass” outside the plume, seems to contain an unrealistic drag term, and
shows problems in the mass conservation equation (Briggs, personal
communication to Zannetti, from Zannetti, 1990).

Ooms and Mahieu (1981) proposed a model able to calculate both the path of the
plume in a windy atmosphere and the ground level concentration. Such model is a
development of the method for the calculation of the plume path, therefore of the
plume rise as well, presented in Ooms (1972) and Ooms et al. (1974). The model
contains eight equations: two equations relate the Cartesian coordinate to the
curvilinear coordinates; six equations describe the entrainment, conservation of
mass (pollutant), momentum in the x-direction and in the z-direction and energy;
the last equation expresses the assumed atmospheric linear stratification. The
description of the entrainment and the drag force, is based on the theoretical work
of Abraham (1970) and Loh-Nien Fan (1967). The rate of entrainment of air into
the plume due to atmospheric turbulence depends on the eddy energy dissipation
ε. For neutral conditions a relation for ε due to Briggs (1969) is used and for the
other stability conditions data from Kaimal et al. (1976) are considered. Cross-
sections of the plume are assumed to be ellipses. Moreover, this model takes into
account the first part of the plume, known as the zone of flow establishment, and
also the turbulence and stratification of the atmosphere so that the influence of the
different stability on the plume path can, in principle, be studied. The simulated
ground level concentrations were compared with those obtained by a classical
Gaussian plume model (using Briggs formulae – see Section 2 – for plume rise
and Singer-Smith, 1966, sigma curves). The agreement was good in neutral and
unstable conditions, while large differences were found in stable conditions.

The Ooms and Mahieu model is used in the ADMS model (e.g., Carruthers et al.,
1999).

The plume rise model proposed by Glendening et al. (1984) is able to treat
arbitrary complex atmospheric structures also when there are large vertical
variations in atmospheric stability or wind velocity (conditions particularly
common for near shoreline power plants). The model consists of a set of eight
134 Air Quality Modeling – Vol. I

ordinary differential equations (conservation of mass, energy, horizontal x and y

momentum, vertical momentum, plus three relationships between curvilinear and
Cartesian coordinates) and three equations (gas equation and the definitions of
virtual temperature and virtual potential temperature). Entrainment is
parameterized according to Hewett et al. (1971). Profiles of temperature and wind
are needed to run the model. The accuracy of the model prediction, verified
against field observations, were found satisfactory and superior to those from a
standard plume rise formula - Equation (44).

The Glendening et al. plume rise model, as modified by Hurley and Manins
(1995), is used in the models LADM (e.g., Physick, 1996) and TAPM (e.g.,
Hurley et al., 2001; see also Section 4.3.2).

The model developed by Chiang and Sill (1985) is applicable to all stability
conditions but only to simple atmospheric structures (that can be expressed by
analytical relationships). The governing equations express the conservation of
mass, momentum in the direction oriented along the plume path and to the normal
to it, thermal energy and tracer concentration. Two relationships between natural
and Cartesian coordinates are also used and the system is closed with an
entrainment model. Basically, the authors' interest was to develop new
entrainment models. Thus they proposed different entrainment models for
different turbulent mixing mechanisms (such as shear, buoyancy, or ambient
turbulence). Then, these authors proposed that, when the turbulent mixing is due
to the contemporary action of different mechanisms, the total entrainment rate is
the linear combination of the various rates derived individually from growth rate
models, i.e. a superposition approach.

The agreement between predicted plume trajectories, velocities and dilution rates
and the observed ones was satisfactory.

Gangoiti et al. (1997) presented a three-dimensional plume rise model for tall
stacks capable of dealing with complex atmospheric profiles. Ambient turbulence
is assumed to be homogeneous and isotropic, the plume is considered to exit from
the stack as a mixture of dry combustion gas, water vapor and liquid water. Dry
air and ambient water vapor, but not liquid water, are then entrained during the
plume motion. Thus the model allows for condensation and/or re-evaporation
within the plume. Condensation in a moist atmosphere increases buoyancy
through release of latent heat while evaporation of droplets absorbs latent heat
from the plume, which consequently loses buoyancy. The classical
parameterization for entrainment of air into plume due to the self-generated
turbulence has been completed with entrainment-extrainment processes in
turbulent winds. This is based on the model of turbulent mass transfer between
plume and environment proposed by Netterville (1990, see Section 2.7). A set of
equations describing in great details the balance of mass, momentum and energy
in the plume constitutes the model. This can be used also to predict plume
penetration into elevated inversion layers but can provide only qualitative
6 Plume Rise 135

estimates of the fraction of plume material that penetrates into them (see Section
5.2).

These authors compared the performance of their numerical model with a set of
simpler models widely used in regulatory applications for plume rise calculation.
Plume condensation has been found to be a major cause of underestimation in
those simpler models, while wind shear causes systematic overestimation in
stably stratified atmospheres. The assumed power law similarity profiles for the
plume temperature and velocity gave better results in light winds (< 1.5 ms −1 )
than the "top hat" profiles.

Also the PLURIS model (Janicke and Janicke, 2001) can be applied to situations
with arbitrary three-dimensional wind fields and to both dry and moist plumes.
Arbitrary directions of the source exit can be considered. Unlike models based on
a similarity profile description, it is not necessary to make assumptions about the
structure or symmetry of the plume cross-section or about the zone of flow
establishment near the source exit. The similarity profiles enter into the model
only via the definition of the liquid water content and affect mainly the prediction
of the visible plume boundary. In the absence of condensation, the model is
independent of any similarity profile assumptions. The model consists of 8
differential equations for mass, x, y, z-momentum, enthalpy, velocity fluctuations,
total water content, and concentration. In addition, there are three differential
equations for the three Cartesian coordinates of the plume axis. In the special case
of a bent-over plume the model can be solved analytically. The model was
validated by a direct comparison with various plume rise measurements obtained
by means of water tank, wind tunnel, and field experiments. The model is
presently implemented and used in combination with the Lagrangian dispersion
model LASAT (Janicke, 1983).

3.3 Differential Models

Golay (1982) proposed a differential entrainment model. It is able to simulate

bent-over plumes in complex vertical atmospheric structures by numerically
integrating the conservation equations of mass, momentum, heat, water vapor,
liquid water, and the two equations for the turbulent kinetic energy and eddy
viscosity in a form presented by Stuhmiller (1974). It uses a mixed Eulerian-
Lagrangian reference system. A two-dimensional Eulerian computational mesh
translates downwind at a plume mass averaged wind speed.

Data from field study of airborne SO2 plume and for ground level SO2
concentration were used to test the model performances. The model simulations
resulted in better agreement with observations than those obtained by standard
analytical formulations.

The major limitation of Golay’s approach is the detailed meteorological

information that is required; i.e., the vertical profiles of wind speed, virtual
136 Air Quality Modeling – Vol. I

potential temperature, relative humidity, turbulent kinetic energy, and turbulent

viscosity.

3.4 LES Models

Probably the most promising technique for the simulation of buoyant plumes in
unstable conditions, at least from a theoretical viewpoint, is the Large Eddy
Simulation (LES). These model simulations allow studying in great details the
contribution to the plume motion caused by convective turbulence and that caused
by plume buoyancy. Nieuwstadt and de Valk (1987) applied such a model to a
line source, in which buoyancy was added by increasing the temperature of the
source with respect to the ambient temperature. Then they solved the equation for
the concentration conservation simultaneously with the other LES equations.
Further work in this direction was performed by van Haren and Nieuwstadt
(1989), who obtained reasonable agreement between the output of their LES,
which however considered only a modest plume buoyancy, and the field
experiments of Carras and Williams (1984). It was found that the fraction of the
plume motion caused by plume buoyancy does not seem to obey the “two-thirds”
law. Plume buoyancy strongly affects the contribution of ambient turbulence to
the mean plume height. Nieuwstadt (1992a) showed that the two contributions
(internal buoyancy and ambient turbulence) cannot simply be calculated
independently but that they interact. Thus ambient convection influences the
plume rise (large eddies modify the entrainment) and vice versa (the interaction
ambient turbulence – plume motion depends on plume rise which transports the
plume to different PBL heights).

Zhang and Ghoniem (1993, 1994 a,b) developed a computational model based on
the Lagrangian interpretation of the dynamics of buoyancy-driven flows that uses
the vortex element and transport element methods to solve the governing
equations. The solution they have constructed causes the model to be considered
as a LES model, since the governing equations describe the effects on the plume
motions of the large scales and the small scales are modeled phenomenologically
(Zhang and Ghoniem, 1993). They faced problems of increasing complexity in
three subsequent papers: firstly they considered a neutral atmosphere with small
scale turbulence in a horizontal uniform wind (Zhang and Ghoniem, 1993), then
considered a linearly stratified atmosphere (Zhang and Ghoniem, 1994 a) and,
finally, a linearly stratified atmosphere capped by an inversion layer (Zhang and
Ghoniem, 1994 b). The following results may be important not only from a
theoretical point of view but also for their practical implications. In neutral
atmosphere it was found that the plume cross-section is kidney-shaped and that
the initial shape of the cross-section (that can be circular or elliptical) has some
effects on the plume trajectory. In the case of a circular plume the "two-third" law
is closely followed. The entrainment is dominated by large scale engulfment
which is inhomogeneous and non-isotropic. In the second case, linearly stratified
atmosphere, it was found that the entrainment constant β (estimated equal to
0.49) mainly affects the equilibrium height, whereas the added mass constant kv
6 Plume Rise 137

(see Section 2.2.2 – estimated equal to 0.7) influences the downwind distance
where this equilibrium height is reached. In the third case, the interaction of the
plume with an inversion layer (i.e.: partial or total or null plume penetration - see
also Section 5.2) has been studied. In particular it was found that when the plume
bumps against an inversion layer, internal gravity waves are generated along the
layer, radiating the energy of the plume and reducing its penetration capacity.

4 Particle Models for Plume Rise

4.1 Introduction

In Eulerian and Gaussian models, the final plume height ∆h is generally

computed by means of simple analytical expressions (like those presented in
Section 2 of this chapter) and inserted in the model as an input parameter. On the
contrary the inclusion of plume rise in Lagrangian Stochastic Models (LSM, see
also Chapter 11) can be done dynamically, i.e. each particle, at each time step,
can be acted upon by local wind speed and direction, ambient stability and
turbulence (both the self-generated and ambient ones). Therefore it is possible to
obtain a degree of resolution and accuracy not obtainable with other simulation
techniques. Furthermore, the interaction of a plume with a capping inversion layer
can be simulated in a rather "natural" way. However the correct incorporation of
plume rise in LSM is still an open problem, since it is needed to simulate the
entrainment phenomenon, that is the exchange processes between the plume
particles and the turbulent environment must be described. Since entrainment
acts, primarily, at the edge of a plume, the position, velocity and buoyancy of the
other particles should be also taken into account.

A completely satisfying approach, based on fundamental particle behavior, is not

yet available. Nevertheless many formulations have been proposed in the
literature to practically solve the problem, with a different degree of
approximation, allowing the plume rise calculation in LSM. They try to achieve a
good compromise among computational requirements, physical consistency and
reliability of the numerical results. Indeed, most of them proved to give reliable
results when compared to laboratory and/or field data. In the following these
approaches will be presented. They include: empirical methods; semi-empirical
methods, in which the plume rise is computed by numerically integrating, at each
time step, the conservation equations - see Equations (2 - 5) - and the plume
spread is calculated by the Langevin equation for the vertical velocity; theoretical
models, in which an attempt is made of directly simulating the rise of buoyant
plumes in a Lagrangian framework.

4.2 Empirical Methods

The first attempt to include plume rise into LSMs, taking into account the vertical
variation of wind and stability, was done by Zannetti and Al-Madani (1984). Let
138 Air Quality Modeling – Vol. I

us recall (see Chapter 11) that in LSMs, the vertical particle positions Xz is
generally computed, at each time step ∆t, as follows

X z (t + ∆t ) = X z (t ) + [U z (t ) + w′(t )]∆t (66)

where U z (t ) represents the mean vertical wind component (generally equal to

zero in flat terrain) and w′(t ) refers to the ambient turbulent term (random
forcing) which is computed from a stochastic equation for the velocity
fluctuation. The idea, is to add an additional vertical velocity accounting for the
buoyant rise, wb , to Equation (66), thus obtaining

X z (t + ∆t ) = X z (t ) + [U z (t ) + w′(t ) + wb ]∆t (67)

They expressed this extra-velocity by time differentiating an empirical analytical

plume rise equation for the transitional phase (TVA formula – Strom, 1976). The
same plume rise contribution is given to all the particles provided they are at the
same height and have the same age. The plume spreads in the vertical as a
consequence of the ambient turbulence only. This last appears in the stochastic
equation for the vertical velocity. This method can correctly simulate the
ensemble averaged plume mean height (provided the used analytical formula is
correct), but the vertical spread, particularly in convective conditions, is likely to
be underestimated. The authors presented some numerical examples showing how
the method works and its ability to give qualitatively reasonable results. However
they did not indicate how to compute the time at which the plume rise stops
contributing to the vertical particle motion. They also suggested a possible
alternative method that could better simulate the vertical spread, but without
developing it. It consisted in tagging each particle with a random buoyancy of
"suitable intensity".

Cogan (1985) was the first to try to model the entrainment process, even if on an
empirical basis. The plume is divided into layers of constant thickness and, within
each layer, it is separated into an inner region (containing the particles included
within the center of mass ± one standard deviation) and an outer region. In the
inner region the temperature of each particle is computed as a function of its
distance from the center of mass, whereas in the outer region the particle
temperature is reduced by a preset amount. This last depends on the chosen value
of the entrainment constant.

Shimanuki and Nomura (1991) tried to numerically simulate the instantaneous

images of chimney plumes under convective conditions. Their technique is based
on single Lagrangian particle trajectories, whose velocity fluctuations are
spatially correlated. The spatial auto-correlation function is prescribed in a
completely empiric way and all the trajectories within a single cell assume the
same value of the spatial auto-correlation function. The buoyancy effect is
roughly accounted for by assigning a given initial vertical velocity to each
6 Plume Rise 139

particle. The air stability does not affect directly the particle motion but is taken
into consideration in the computation of turbulence scales.

The Zannetti and Al-Madani (1984) suggestion was applied by Anfossi et al.
(1993) for buoyant plumes and by Anfossi (2000) for jet plumes. To each i-th
particle a normally distributed buoyancy flux Fbi is assigned at the stack exit,
fixing the mean value equal to the mean buoyancy flux Fb and the standard
deviation equal to Fb 3 (this value, 1/3, was empirically fixed requiring that the
plume radius near the source was approximately equal to 0.6 ∆h( z ) ). In the case
of a jet plume, Fmi , Fm and Fm 3 are used. Instead of computing wb by means of
an empirical analytical plume rise equation for the transitional phase only, they
assumed that the plume centerline grows according to a plume rise formula
describing both the transitional and final phases and different stability conditions.
Thus, a simple algebraic expression giving a smooth curve and possessing the
correct asymptotic and near field limiting forms was used. This interpolation
curve was built following the Moore's suggestion - see Equation (56). For
buoyant plumes, the interpolation curve has the following expression (Anfossi,
1985)

(
∆h(t ) = 2.6 Fb t 2 U a ) (t s + 4.3)
13 2 −1 3
(68)

and was obtained considering the “two-thirds” law - Equation (27) - for the
transitional phase both in neutral/unstable and stable conditions (Arya, 1999) and
Equations (44) for the final rise in stable conditions. For neutral/unstable
conditions (s = 0) the plume final height is fixed according to Equation (38). The
low wind speed conditions are dealt with inserting a minimum wind speed (0.3
ms1). wb is computed as follows

∆z [∆h(U a , s , t + ∆t ) − ∆h(U a , s , t )]
wb = = (69)
∆t ∆t

The model simulations were validated against DIAL measurements of a Thermal

Power Plant plume in complex terrain (Anfossi et al., 1993). Predicted plume
centerline height and horizontal and vertical plume width satisfactorily compared
to the observed ones.

This method is used in the 3-D Lagrangian Stochastic Model SPRAY (Tinarelli et
al., 2000; Finardi et al., 2001), which is also used (in some complex terrain cases)
for regulatory purposes in Italy.

Equation (68) was also used by Graziani et al. (1997) for their LSM simulation of
the dispersion of the volcanic emission from Vulcano Island.
140 Air Quality Modeling – Vol. I

height (m)

downwind distance (m)

Figure 6. See the text (adapted from Anfossi et al., 1993). [Reprinted with
permission from Elsevier Science].

Figure 6 shows a simulation result of this approach in non-homogeneous vertical

meteorological conditions, in which simpler analytical approaches cannot be
used. It refers to a typical fog situation in the Po Valley (Northern Italy): the fog
layer extends from ground level to 250 m, an inversion layer with a large
temperature increase (8 °C) lies between 250 and 400 m and the superior layer is
nearly isothermal. Nearly calm conditions prevail ( u0 ≅ 1 ms −1 ). Each point in
Figure 6 represents a model particle position. One can see that the rise of the
plume is stopped by the inversion layer, that the pollutant reaches the ground
level very close to the stack (fumigation) and that a partial inversion penetration
occurs.

The case of jet plume is similarly treated. The starting points are the Briggs'
formulae for jet plumes, Equations (24), (48), (51), and (52). Equation (24) works
in the transitional phase, where the other equations are valid for the final stage of
rise both in neutral and stable windy or calm conditions. In this case, Equation
(68) becomes:

• for neutral conditions

6 Plume Rise 141

−1 6
 d 6 U 4t 2 
1 3 −2 3 1 3
∆h( x) = d F u
1 m 0 t  1  a + 1 (70)
 d 2  Fm 

0.9
were: d1 = 2.3 and d 2 = ;
 U 
u*1 2  0.4 + 1.2 a 
 vs 

• for stable and windy conditions (u > 1 m s −1 )

−1 6

− 2 3 1 3  d1 
6

∆h( x) = d1Fm U a t   s 2t 2 + 1
13
(71)
 d 3  

were: d3 = 1.5;
• for stable and calm conditions (u < 1 m s −1 )

−1 6
 d 6 Fm1 2 s 3 2t 2 
−2 3 1 3
∆h( x) = d1Fm1 3U a t  1  2
+ 1 (72)
 d 4  Ua 

were: d4 = 4.

The vertical velocity, wm , is computed as in Equation (69) and the plume rise
calculation is stopped when the difference between ∆h( x) and the corresponding
asymptotic final value is less than a chosen small value.

Also Souto et al. (2001) estimated the rise of buoyant plumes according to
Equation (67). The extra velocity due to the buoyancy effects, wb, was estimated
as follows
1.35 F 1 x 0.58
wb = 3 (73)
∆t u
in unstable conditions and

1
2.04  0.86 F (1 − cos( Nx / U a ) 3  3
wb =  + rs  (74)
∆t  N 2U a 

in stable conditions. The numerical coefficients of these equations and the x

exponent in Equation (73) were obtained by best fit of field observations.
Equation (74) was proposed by Zhang and Ghoniem (1994a). This plume rise
142 Air Quality Modeling – Vol. I

calculation was inserted in two operational models, a LSM and an adaptive puff
model.

4.3 Semi-Empirical Methods

Some interesting methods for incorporating buoyancy effects in LSM, were

proposed by Luhar and Britter (1992), Hurley and Physick (1993), Hurley (1999,
2000) - these two methods are inserted and operative in the CSIRO models
LADM and TAPM - and Weil (1994). Instead of assuming valid an analytical
formula (or an interpolation formula), these methods compute the mean plume
rise by directly solving the energy, mass and momentum conservation equations.
This procedure is time consuming. However, also the Langevin equation,
computing the velocity fluctuations, is already numerically solved. Therefore the
increase in calculating time, with present days computer technology, is likely not
to be a real problem. Different schemes are used to compute the plume spread and
the interaction of ambient turbulence with the plume.

4.3.1 The Luhar and Britter (1992) Method

The LSM of Luhar and Britter (1992) accounts for the effects of source buoyancy
on plume dispersion in the CBL by including the mean plume rise and the
additional dispersion due to plume’s self-generated turbulence. To incorporate the
mean plume rise they added a new acceleration term in the Langevin equation for
the vertical velocity component of their previously developed LSM (Luhar and
Britter, 1992) for the dispersion of passive plumes in the CBL (consequently, the
model is one-dimensional). The new acceleration term was based on the
following expression

g
(ρ a − ρ (t ) ) = −U dW
(75)
ρa
a
dx

where ρ(t) is the plume density at travel time t (= x/Ua). In the model, it was
assumed that in the CBL the effects of ambient stability and wind shear on plume
dispersion and rise could be neglected since the potential temperature and
horizontal wind do not change appreciably with the height. This assumption
allowed using the conservation equations in a simplified form, valid in the neutral
boundary layer. The solution obtained from these equations for the mean vertical
velocity W for three phases of plume development was used in Equation (75). The
three phases correspond to the three different entrainment relationships of
Slawson and Csanady (1971). During the initial phase, the intermediate phase,
and the final phase, plume’s self-generated turbulence, inertial sub-range
turbulence, and energy containing eddies, respectively, govern plume rise. The
expression for W in the initial phase of the rise, for example, is
6 Plume Rise 143

1
 4  3 1 −1
W =  2  Fb 3 x 3 (76)
 9β 

which corresponds to the well-known ‘two-thirds’ behavior of plume rise. Luhar

and Britter (1992) emphasized the importance of including the ambient turbulence
effects in calculating W through the use of the entrainment relationships of the
last two phases. Effects of plume’s initial momentum on plume rise were also
included.

To account for the additional dispersion due to the self-generated turbulence, the
model assumed that the mass, and hence the computer particles and their velocity
and acceleration, have a Gaussian distribution about the plume centerline. Thus
the model uses two random numbers drawn from a Gaussian distribution: one for
the random component of the acceleration due to the ambient turbulence and the
other for the random acceleration due to buoyancy.

The model simulations of crosswind-integrated concentrations from a few

laboratory and field studies reported in the literature appeared to be satisfactory.

4.3.2 The CSIRO Methods

Instead of accounting for the effects of buoyancy on plume dispersion simply by

introducing an extra term into the random walk equation for displacement, Hurley
and Physick (1993) compute the vertical velocity of each particle as the sum of
the plume velocity due to the buoyancy and initial momentum effect and a
stochastic perturbation due to the combined effects of the self generated and
ambient turbulence. However, the problem of simulating the entrainment process
is not solved since the classical entrainment assumption - the plume radius grows
linearly with height - is imposed in both the deterministic and stochastic parts.

The deterministic vertical velocity is obtained by numerically integrating, at each

time step, the basic conservation equations (see Section 2) in which the standard
assumptions (Boussinesq approximation, top-hat profile and bent over plume) are
made. The stochastic wind components are calculated by the Langevin equations
for the velocity (Thomson, 1987; see also Chapter 11), namely

dui' = a dt + C0 ε a dW (77)

where i = 1,2 ,3 , ui' are the Lagrangian velocity fluctuations, a depends on the
Eulerian probability density function (PDF) of the turbulent velocity and is
determined from the Fokker-Planck equation, C0 is a numerical constant, dW is
a random term, normally distributed (mean 0 and variance dt ). For the
simulation of plume rise in the CBL, the Gaussian form of the PDF is assumed for
the two horizontal components, while a skewed distribution, obtained by a linear
144 Air Quality Modeling – Vol. I

combination of three Gaussian functions, is used for the vertical component.

These three functions represent the contribution to the turbulence due to the
updrafts, downdrafts and plume self-generation, respectively. In the LSM for
passive tracers based on the Langevin equation, the coefficients of the Gaussian
functions are obtained by equating the zeroth through third moments of the
Eulerian ambient PDF ( 0 , σ wa
2 3
, S wa ). In this case, the first three moments of the
resulting PDF are equated to 0 , σ wa
2
+ σ wp
2 3
, S wa , where σ wp is the velocity
2

variance due to the plume rise effects. The three plume velocity variances are
defined, on the basis of the above mentioned classical entrainment assumption, as
follows

σ up2 = σ vp2 = (βw p )2

(78)
σ = (β w p 2 )
2 2
wp

where wp is the particle vertical velocity. The form of the PDF of the vertical
velocity fluctuations for simulations in stable conditions is assumed to be
Gaussian with the variance and the eddy dissipation rates being calculated as the
sum of ambient plus plume rise induced components.

Plume rise computation is terminated either when ε p ≤ ε a (convective conditions)

or when the buoyancy of a particle becomes less than or equal to zero (stable
conditions). The plume penetration of the inversion layer capping the mixing
height is simulated letting the plume particles to overcome the mixing height if
they do not have yet satisfied the termination condition. When the plume rise
calculation is stopped, particles are reflected at the mixing height.

The authors advise of some numerical problems in the first couple of time steps (1
(
s) after release due, very likely, to the height dependence of ε p = 1.5 w 3p ∆h(x ) )
on the rise height, which tends to diverge for ∆h ∝ 0 . The problem was partially
solved by imposing, near the stack mouth, ε p ≤ σ w2 (C0 ∆t ) .

Comparisons with CBL water tank dispersion experiments (Willis and Deardorff,
1987), characterized by different values of the dimensionless buoyancy flux F* -
see Equation (42) - were shown. Predicted final plume height ∆h and plume
entrapment above the CBL were found to be in reasonable agreement with the
observed ones, even if a slight overestimation of ∆h in three out of four
experiments, causing peak entrainment values higher than observed, was found.
This could be corrected by changing the value of the entrainment parameter β
from 0.6 to 0.7. The overall distribution of concentration compared quite well.
6 Plume Rise 145

An alternative approach to including plume-rise induced turbulence in a

Lagrangian approach, which avoids some of the very-near-source numerical
problems above mentioned is contained in the model TAPM (Hurley, 1999,
2000). The equations of conservation of plume volume, buoyancy and momentum
flux, G, Fb and Fm, are written in this model - see, for comparison, Equations (2 -
5)

dG
= 2 RαW 2 + βU aW + γu p E 2 
1

dt  

dFb F  1 
= −s m  Ua +W  (79)
dt u p  2.25 

dFm
= Fb
dt

Ta
where G= u p R2 , Fm = GW , Fb is defined as in Equation (6),
Ts
u p = U a2 + W 2 is the plume velocity, E is the turbulent kinetic energy, a = 0.1 ,
β = 0.6 and γ = 0.1 are the vertical plume, bent-over plume and ambient
turbulence entrainment constants, respectively. The initial conditions for Fb and
Fm are the same as in Equations (16) and (18),

Gs = Fm vs 0 and Rs = vs 0 u02 + vs20 .

Equations (79) are based on the model proposed by Glendening et al. (1984), as
simplified by Hurley and Manins (1995). Tests on these equations, performed by
the authors, showed that they performed as good as the full original ones and
collapsed to the Briggs form for a bent-over Boussinesq plume, and to the Briggs
vertical plume model equations for calm conditions.

4.3.3 The Weil (1994) Method

This method was designed to deal with CBL dispersion of weakly to moderately
buoyant plumes ( F* ≤ 0.1 ). Also in this method, an extra acceleration term is
added in the Langevin equation in order to account for plume rise. This is
obtained by numerically solving the conservation Equations (2 - 5) in which the
entrainment assumption appearing in Equation (2) is modified to account for the
ambient and self-generated turbulence. Consequently, Equation (2) becomes

d
dt
( )
U sc R 2 = 2 βRU sc w p + 2γ 1 RU sc ve (80)
146 Air Quality Modeling – Vol. I

where the identity (d dt = U sc d d sc ) was used, ve is the entrainment velocity

and γ 1 an empirical parameter. The first term on the r.h.s of Equation (80)
describes the entrainment due to the plume-generated turbulence, whereas the
second term accounts for the ambient turbulence. For a jet plume ve = va where

va = (2ε a R )
1
3 (81)

is the relative velocity of two particles separated by a distance R. Notice that in

this approach the radius R is defined as “the region enclosing all of the buoyant
fluid”. For a buoyant plume, having assumed U sc = U a , ve is given by

va
ve = (82)
1 + γ 2 Fb U a v a2 R ( )
where γ 2 is another empirical parameter. Defining the dimensionless distance X*
and momentum flux Fm* as

w* x Fm
X* = and Fm* = (83)
U ah U a w* h 2

Weil (1994) obtained that ambient turbulence becomes effective at a non-

3 2
dimensional distance X a = 1.6 α v 5 F* 5
for a buoyant plume and
9 2
X a = 0.6 α v 7 Fm*7 for a jet, having defined α v = ve wp . Fitting his model to the
Willis and Deardorff (1983) observations, this author determined the following
values for the unknown parameters: γ 1 = 0.4 , γ 2 = 1 and α = 0.49 or 1.9 for
buoyant and jet plumes, respectively. Notice that α = 0.49 implies that ambient
turbulence starts to dominate when the entrainment velocity is about half the
plume velocity ( ve ≈ 0.5 w p ).

When the plume reaches the temperature inversion height h (capping the CBL)
and penetration occurs, the vertical plume velocity becomes zero (actually it
oscillates about zero) but some plume segments may be brought into the CBL by
negative ambient velocities. The buoyant acceleration, due to the fact that the
potential temperature is greater than below h, provides a positive velocity tending
to keep the plume aloft and the ambient velocity (calculated by the Langevin
equation) varies randomly and therefore may be either positive or negative.

Weil (1994) also found that, despite the differences in the models, the mean fields
computed with this model are very similar to those produced by the Luhar and
Britter (1992) model.
6 Plume Rise 147

4.4 Theoretical Models

4.4.1 The Van Dop (1992) Particle Model for Buoyant Plume Rise

A buoyant plume is defined as the volume, which contains a mixture of ambient

and originally released, buoyant fluid. The envelope of the plume is the
(imaginary) and in a way, arbitrary boundary of this volume. A fraction of the
original buoyant fluid separates from the plume and becomes so remote that it is
no longer considered to be part of it. On the other hand the volume of the plume
expands due to turbulent intrusions of ambient air resulting in an increasing
ambient fraction and consequently, a gradual loss of plume temperature and
vertical acceleration.

A Lagrangian ‘plume particle’ can now be defined as a small entity, which

possesses the mean characteristics (velocity, temperature) of the plume.
Stochastic fluctuations, directly related to the turbulent intensity within the
plume, determine the rate of growth of the plume width and are superimposed on
the mean characteristics. Ultimately, due to the entrainment and extrainment
processes, the plume (particle) dynamics must converge to the environmental
dynamics. Hence, the equation of motion in the vertical dimension for a buoyant
plume particle can be formulated as

W
dW = − dt + Bdt + ε 1w/ 2 dω w (t )
Tp

B
dB = − dt − NW 2 + ε 1B/ 2 dω B (t ) (84)
Tp

dX z = W dt

(see Van Dop, 1992). Here, W is the plume particle vertical velocity and B is the
plume particle buoyancy, defined by

g
B= (ϑ − ϑa )
Ta p
(85)

The assumption that the buoyancy of each individual fluid particle is defined by
the difference between the particle's temperature and the ambient temperature is a
crude one, and in fact the proper buoyancy should be related to the full
surrounding temperature field, which includes the temperatures of the other
buoyant fluid particles. However, this inclusion would lead to a set of (coupled)
Langevin equations for each individual fluid particle, and the attractiveness of the
Lagrangian approach would be lost. The Lagrangian time scale of the plume is
148 Air Quality Modeling – Vol. I

assumed to be equal for velocity and buoyancy and is denoted by Tp. The
dissipation of velocity and buoyancy is denoted by εw and εB, respectively. dωw(t)
and dωB(t) are random increments in the Lagrangian equations for velocity and
buoyancy respectively. The stratification of the environment is given by N - see
Equation (15). Xz is the plume particle’s vertical position.

The usual assumptions for the random terms are

dω w (t ) = 0

dω B (t ) = 0

dω w (t ′)dω w (t ) = δ tt′ dt (86)

dω B (t ′)dω B (t ) = δ tt′ dt

dω B (t ′)dω w (t ) = 0

The last assumption implies that stochastic velocity and buoyancy changes are
uncorrelated on the (small) Kolmogorov scales, which may be questionable, but is
perhaps not very important for the present consideration. It should be noted that
for the mean plume rise, X z , the Equations (84) reduce to a deterministic set -
due to the properties listed in Equations (86), and thus X z does not depend on
the dissipation terms.

In order to simulate a power law behavior for plume rise, which, in the early stage
in a calm neutral environment, is confirmed by experimental evidence (see for
example Turner, 1973), the Lagrangian time scale must be proportional to t.
Assuming Tp = Ap (t + to), it is retrieved the similarity solution (Csanady, 1973),
provided that

Ap =3/4 (87)
and
t 0 = (2rs / 3βB0 )
1/ 2
(88)
The relation with the 2/3 law is imposed by the choice of Ap = 3/4. The initial
plume radius and buoyancy are denoted by rs and B0, respectively, and β is the
plume entrainment constant (~ 0.6). Through the definitions of the plume particle
buoyancy, B, and the heat output of the source, Qh, to can be related to the
buoyancy flux parameter Fb - see Equation (9) – by
6 Plume Rise 149

9 3 4 1 F
B0 t 0 = 2 b (89)
4 β u

With this choice of parameters the Langevin formulation can be forced to

correspond asymptotically ( t → ∞ ) to the classical plume rise formulae.

Plume rise in a turbulent environment was addressed in detail by Netterville

(1990) (see Section 2.7.2), who introduced an additional turbulent exchange
mechanism, ‘extrainment’, generated by the ambient turbulence. A logical
consequence of his theory for the Lagrangian framework is that if the plume
turbulence dominates, the turbulent time scale of the plume, Tp should be applied,
whereas if the environmental turbulence dominates, the ambient time scale, Te ,
should be used. This view is reflected in an modified expression for the time
scale

1 1 1
= + (90)
Tm T p Te

The Lagrangian formulation using Equation (90) was compared with Netterville's
expression for the mean plume rise and though the Lagrangian formulation results
in somewhat lower values, it has the same leveling off behavior in the final stage.

The Lagrangian equations provide also for an independent evaluation of the

plume variance or plume width. This requires, however, explicit expressions for
the dissipation, ε w and εB in Equations (84). Van Dop (1992) suggests to use the
actual particle velocity and buoyancy to parameterize the dissipation and assumes

ε w = c1W 2 / T p
(91)
ε B = c2 B 2 / T p

where c1 and c2 are constants O(1). Numerical solutions for the plume width, σz,
( 2
defined as X z − X z
2
)
12
, were obtained, but do not agree with the similarity
prediction, σ z ∝ t 2 / 3 .

Alternatively, the dissipation may be parameterized as

−α
 t + t0  w
ε w = cw  
 t0 
(92)
−α B
 t + t0 
ε B = cB  
 t0 
150 Air Quality Modeling – Vol. I

Choosing αw > 5/3 and αB > 11/3, and for example equal to 2 and 4, respectively,
it can be proven numerically and (in the neutral case, N = 0) also analytically that
the plume width converges to the similarity prediction

2/3
 t + t0 
σ z = (3 c t
B 0 )
5 1/ 2
  (93)
 t0 

and the ratio σ z X z is given by

1/ 2
σz  3c t 
=  B2 0  (94)
X z  B0 

Csanady (1973) suggests (pp. 176-195) that this ratio is approximately 1/3. From
this a value for the coefficient cB of

B02
cB = (95)
27t 0

is inferred. Finally, evaluating the plume width in a turbulent environment

requires also that the ambient turbulent dissipation rate is considered. As in the
case for the Lagrangian time scale - see Equation (90) - Van Dop (1992) suggests
the parameterization

ε p + (T p / Te )2 ε a
εm = (96)
1 + (T p / Te )
2

where εp is given by W2 / Tp.

The algorithm can without much difficulty be extended to an non-homogeneous

and non-stationary ambient turbulence by including height dependent
formulations for ambient time scales and dissipation rates and using the
appropriate version of the Langevin equations in these conditions. Arbitrary
stratification, including a CBL with a capping inversion can be accounted for by
introducing a height dependent N.
Though the method contains a number of heuristic elements, the Lagrangian
formulation is transparent and computationally straightforward. It is consistent
with the classical formulations for plume rise in a calm environment (see Turner,
1973; Briggs, 1969; or Csanady, 1973), but also accommodates more recent
Eulerian formulations in a turbulent environment (Netterville, 1990; Nieuwstadt,
1992a,b). This makes it attractive for various practical applications. Yamada
(2000) included this algorithm in a modeling system (HOTMAC-RAPTAD) and
6 Plume Rise 151

examined its performance. He concluded that the overall performance was ‘as
least as good as those of the ‘better’ models reported by Hanna et al. (1993)’.

A drawback of the Lagrangian method is that in order to remove the statistical

noise, a large number of flow realizations should be evaluated.

4.4.2 Buoyant Plume Rise Described by a Lagrangian Turbulence Model

The research interest in buoyant plume rise is driven by the theoretical aspects of
the simulation of the turbulent mixing of fluids with different temperatures.
Similarity theory provides parameterizations for the mean plume height and width
(Csanady, 1973, Briggs, 1975) if the influence of the ambient turbulence can be
neglected, i.e., if the turbulence is generated only by the plume. This applies to
the initial stage of plume rise, and for emissions into neutrally stratified ambient
flows with a negligible turbulence. However, for practical plume rise calculations,
models are required that:
• are computationally not too expensive
• can be applied to both stages of the buoyant plume rise and different
ambient conditions
• permit the assessment of fluctuations, i.e., provide also plume statistics

The attempt to derive directly such models leads within the Eulerian framework
to Reynolds-averaged Navier-Stokes (RANS) equations, and within the
Lagrangian framework to Lagrangian particle models. RANS equation methods
(Weil, 1988, Netterville, 1990, Gangoiti et al., 1997) apply parameterizations for
terms that are related to the turbulent mixing of the plume and the ambient flow.

By means of Lagrangian methods both the mean plume behavior and the plume
statistics can be described in accordance with constraints of the similarity theory
and observations. This was demonstrated by van Dop (1992) (see Section 4.4.1)
in a first systematic analysis of the description of buoyant plume rise by
Lagrangian methods. Alternative methods are described by Anfossi et al. (1993)
(see Section 4.2), where also a review can be found on earlier work (see Zannetti
and Al-Madani, 1984, Cogan, 1985) to describe buoyant plume rise by means of
Lagrangian methods. Lagrangian particle models simulate the plume dynamics,
but they require knowledge about the flow field that has to be provided by
Eulerian models, or has to be approximated.

Lagrangian turbulence models (LTM) give a full description of both the motion
and properties of plume and of the ambient flow. In particular, these Lagrangian
equations are constructed consistent with the Eulerian RANS equations. In
analogy to direct numerical simulation (DNS) or large eddy simulation (LES)
(Nieuwstadt and de Valk, 1987, Nieuwstadt, 1992a,b, Zhang and Ghoniem, 1993,
1994a,b; see Section 3.4), LTM resolves mixing for high-Reynolds number flows
avoiding the high computational costs of DNS or LES. We shall derive a buoyant
152 Air Quality Modeling – Vol. I

plume rise model from an LTM, which can be used for regulatory applications
and satisfies the constraints (i), (ii) and (iii) considered above.

The Lagrangian description of fluid motion (i.e., of plume- and ambient-air

particles) requires Lagrangian equations that are consistent with the Navier-
Stokes equations. Two methods are used to date which provide for this
consistency: first, the derivation of stochastic Lagrangian equations that are
consistent with RANS equations up to second-order (van Dop et al., 1985,
Sawford, 1986, Pope, 1994, Heinz, 1997, 1998), and second, the derivation of
these equations consistent with an Eulerian velocity PDF (Thomson, 1987). The
first approach, which is applied here, requires closure assumptions for the
pressure redistribution and dissipation terms in the RANS equations of second-
order. Closure assumptions for the pressure redistribution and dissipation terms in
the RANS equations are known and relatively well-investigated for buoyant flows
(see, e.g., Craft et al., 1996).

Details about the derivation of stochastic Lagrangian equations for buoyant

turbulence can be found elsewhere (Heinz, 1997, 1998; Heinz and Van Dop,
1999). Here we present a summary.

The change of particle position dXi (i = 1, 2, 3), velocity dUi and potential
temperature dϑp is described by a set of linear stochastic differential equations:

dX i = Ui dt
dU i = Ai dt + bijdω j (97)
dϑ p = Aθ dt + bθ dω θ

where

Ai = ai + Gij (U j − u j )+ Gi (ϑ p − θ )
Aθ = aθ + Gθj (U j − u j )+ Gθ (ϑ p − θ ) (98)

The (Eulerian) ensemble average is denoted by overbars and summation over

repeated subscripts is assumed. Deterministic changes of the particle velocity are
described through the first terms on the right-hand side of Equation (97) with the
unknown coefficients ai ,aθ ,Gi ,Gij ,Gθ and Gθj . The second terms describe the
stochastic force caused by the small-scale turbulence and contains the additional
unknown coefficients bij and bθ. The properties of dωj and dωθ, random
increments for velocity and potential temperature (~ buoyancy) - see Equation
(84) - are defined in Section (4.4.1).
6 Plume Rise 153

Equations (97) can be transformed into a Fokker-Planck equation for the one-
point joint velocity-temperature PDF of the flow P(u,ϑ, x, t) (Gardiner, 1983;
Risken, 1984)

∂ Ai P ∂ Aθ P ∂ Bij P ∂ 2 Bθ P
2
∂ P ∂ ui P
+ =− − + + (99)
∂t ∂ xi ∂ ui ∂θ ∂ ui∂ u j ∂θ2

with

1 1
Bij = bik bkj ≡ C 0 ε q δ ij
2 2
(100)
1
2
( )
2 1
B θ = bθ ≡ C1 ε θ
2

The viscous dissipation and potential temperature dissipation are denoted by εq

and εθ, respectively. C 0 and C1 are constants whose value will be determined
later.

From Equation (99) arbitrary moments of velocity and potential temperature can
be obtained. In this way we are able to derive a set of equations which are similar
to the RANS equations. We can summarize the latter in a suitable approximation
up to second order as

d ui ∂ u i′u k′ 1 ∂p g ∂ 2 ui
+ =− + (θ − ϑa )δ i 3 + ν
dt ∂ xk ρ a ∂ xi T ∂x k ∂x k

dθ ∂ u ′kθ ′ ∂ 2θ
+ =κ
dt ∂x k ∂x k ∂xk

d ui′θ ′ ∂ ui′u k′θ ′ ∂u ∂θ g 2 p ′ ∂θ ′

+ + u k′θ ′ i + ui′u ′k = θ ′ δ i3 + (101)
dt ∂x k ∂x k ∂x k T ρ a ∂xi

dθ ′2 ∂ ui′θ ′ 2 ∂θ
+ + 2uk′θ ′ = −2 εθ
dt ∂x k ∂x k

d ui′u ′j ∂ ui′u ′j u k′ ∂ ui ∂uj

+ + u ′j u k′ + ui′u ′k =
dt ∂x k ∂x k ∂x k
p ′  ∂ui′ ∂u ′j 
=
g
T
(
ui′θ ′δ j 3 + u ′jθ ′δ i 3 + ) +
ρ a  ∂x j ∂xi
 − 2ε q


154 Air Quality Modeling – Vol. I

Here, T is a boundary-layer reference temperature, and ϑa is the mean ambient

potential temperature,ν is the molecular viscosity and κ the conductivity. In order
to be able to solve Equations (101) we have to make a number of closure
assumptions

∂u ′p ∂u ′q 1 ui′ 2 (102)
2ν = δ
∂xk ∂xk 3 τ pq

where τ is a dissipation time scale, which obeys

dτ 2τ  ∂ ul 
= Cε 2 − 1 − (Cε 1 − 1) ⋅ 2 − u k u l + βg u 3θ  (103)
dt q  ∂x k 

where Cε1 and Cε2 are constants equal to 1.56 and 1.9, respectively, and q is twice
the turbulent kinetic energy.

k q2
εϑ = 4 θ ′2 εq = (104)
2τ 2τ

 ∂u ′ ∂u ′j   ∂u 
p ′ i +  = − k1  u i′u ′j − 1 q 2δ ij  + k 2 q 2  ∂u i + j  (105)
 ∂x  2τ   ∂x j ∂xi 
 j ∂xi  3   

∂θ ′ k
p′ = − 3 u i′θ ′ (106)
∂x j 2τ

In the closure assumptions appear a number of closure constants ki (i=1-4), which

will be determined later.

It can be shown (cf. Heinz, 1997; Heinz and van Dop, 1999) that first and second
moment equations derived from Equation (99) can be written similar to Equations
(101), provided that the Lagrangian constants appearing in Equations (97) obey
the following relationships
6 Plume Rise 155

2 k 3 − k1
Glθ = 0 Gθ = −
4τ

g k1
Gi = δ i3 Gil = − δ il
T 4τ (107)

k1 − 2
C0 = C1 = 2k 3 − 2k 4 − k1
3

An additional consistency requirement is that k2 = 0. Hence we may reformulate

the Lagrangian Equations (97) now as

dX z = W dt

k1 C0 q 2
dW = − W dt + B dt + dω w
4τ 2τ

(108)
2k3 − k1 C1 ( g / T ) 2θ 2
dB = − Bdt − N 2W dt + dω θ
4τ 2τ

dτ 2τ  ∂ ul 
= Cε 2 − 1 − (Cε 1 − 1) ⋅  − u k ul + βg u 3θ 
dt u i′ 2  ∂x k 

where N is the Brunt- Väisälä frequency and B = (g / T )(ϑ p − θ ) . The constants

Cε1 and Cε2 are related by Cε1 = 1 + (Cε 2 − 1)/1.6 (Heinz,1998). Note that for the
evaluation of the Eulerian moments appearing in Equation (108) it is still required
to solve Equation (101) numerically.

In order to determine the remaining constants in Equation (97) we have compared

the Lagrangian predictions of plume rise with Eulerian approaches (Weil, 1988;
Netterville, 1990) in a still environment. This yields

3k1  k3 1 
Cε 2 = 1 + ⋅ − (109)
8  k1 2 

and

1/ 4 −1 / 4 −3 / 4
 3πβ 2   k3 1   1  k 3 1 
k1 = 4   −  1 −  −  (110)
 2   k1 2   2  k1 2 
156 Air Quality Modeling – Vol. I

Similarity theory also provides algebraic scaling laws for the variances. Applying
these we obtain

3
Cε 2 = 1 + ⋅ k4 (111)
8

Equations (109), (110), and (111) relate four unknown constants,

Cε 2 ,k1 ,k3 and k4 . A value for Cε2 was found from a comparison of our model
results with the LES data of Zhang and Ghoniem (1994a) in stably stratified flow,
yielding Cε2 = 1.286. All constants in Equations (108) are now determined.

The full model was evaluated against experimental data collected by Erbrink
(1994). The effects of ambient wind shear and stability were considered by
solving the parameterized equations for q2,θ 2 ,u1u3 and u3θ , see Equations (101).
Details are given in Heinz and Van Dop (1999).

Figure 7(a) shows a scatter plot of measured plume heights versus the
corresponding modeled plume heights in neutral to slightly stable conditions. The
figure shows that the agreement between the observed plume heights and our
predictions is very good. This means in particular that the model predictions do
not only agree with the two-thirds power law, but also estimate correctly the
leveling-off of the plume due to ambient stability.

In Figure 7(b) the plume radii are compared. The agreement is still fair, though
the scatter has increased. It gives some support for the observation that the 2/3
similarity prediction also holds for the spreading of the plume.
6 Plume Rise 157

Figure 7. (a) Scatter plot of measured versus modeled normalized mean

particle heights, X z / B0τ 02 . B0 is the initial buoyancy and τ0 is defined by

τ 0 = (π 1 2 rs B0 )
12
. (b) the same comparison but now for the plume
2
X z − X z / B0τ 02 . Ambient stability conditions during
2
widths defined as
the measurements varied from slightly stable to neutral.

5 Special Cases

All the plume rise estimates discussed in the previous four sections apply to
effluents coming out from elevated and isolated sources, without accounting for
possible effects of nearby buildings or stacks or of the presence of inversion
layers. However, in practice, these situations (and others, like the case of ground
level fires, of flares and of the presence of scrubber) need to be investigated since
they may affect the plume rise and, consequently, modify the plume trajectory
and the ground level concentration distribution. Generally these special topics are
treated with “ad hoc” formulations. In this review we will briefly consider the
following cases: downwash parameterization, penetration of elevated inversions,
plume rise from multiple sources, plume rise from flare stacks, plume rise from
fires, plume rise from stacks with scrubber. This review is not exhaustive and will
give some examples of possible modeling solutions that are, in general, inserted
in regulatory models.

5.1 Downwash Parameterization

In the plume rise computation, a special care must paid to the possible occurrence
of downwash effects. These can be classified as:
• stack tip downwash, a possible drag of the effluent in the wake downwind
the stack due to the presence of the stack itself
• building downwash, effluent emitted from a stack near a building and
brought downward by the flow of air over and around the building
158 Air Quality Modeling – Vol. I

• Stack tip and building downwash cause a decrease in the plume rise
because of two concomitant phenomena
• the drag of the effluent in the stack and/or building wake; since this wake
extends below the stack outlet, this drag causes the plume to decrease its
height
• the increase of the entrainment with ambient air (causing a consequent
decrease of buoyancy) due to the wake turbulence

In both cases, the reduced plume rise has the effects of increasing the ground
level concentration. In particular, these last may be very high immediately
downwind the stack or building if the plume is completely trapped in their wake.

Both stack tip downwash (e.g., Briggs, 1973; Bjorklund and Bowers, 1982;
Overcamp, 2001) and building downwash (e.g., Briggs, 1973; Huber and Snyder,
1982; Schulman and Hanna, 1986) are to be considered when the ratio vs u is
small. In the first case, the buoyancy amount has to be accounted for as well.

The procedures for the correction of the final plume rise, presented in the
remainder of this section, do not provide any information about the plume
trajectory near the stack outlet. They turn out useful in case one is interested in
the prediction of pollutant concentrations in some areas that are at least a few
hundreds of meters away from their source.

5.1.1 Stack Tip Downwash

The generally accepted practical rule (Briggs, 1969 and 1973) is that stack
downwash will occur if the ratio of effluent speed, vs0, to wind speed, u0, is less
than about 1.5. However Briggs (1969) also suggested that this rule may be
relaxed for highly buoyant plume (emitted by modern fossil-fuel power plants and
larger industrial stacks). However, stack tip downwash is still an important
problem for neutrally buoyant effluents or small industrial emissions.
Furthermore, Overcamp (2001) stressed that it is a very important problem in
simulating plumes in wind tunnels and towing tanks.

Bjorklund and Bowers (1982) proposed the following expression for the final
plume rise corrected for the stack tip downwash, ∆h′

∆h′ = f ∆h (112)

where f is a dimensionless parameter calculated with the following procedure:

6 Plume Rise 159

first, compute the Froude number of the effluent, Fr, defined as

12
 vs20 
Fr =   (113)
 2 g A π (T − T ) T 
 s s0 a0 a0 

• if Fr2 < 3 then f = 1 (no correction);

• if Fr2 ≥ 3 :
if vs > 1.5 u0 then f = 1 (no correction);
if u < vs ≤ 1.5 u0 then f = 3(vs 0 − u0 ) vs 0 ;
if vs ≤ u0 then f = 0 (no plume rise).

Snyder and Lawson (1991) modeled the downwash of neutrally buoyant effluent
on the immediate lee side of a circular stack in a wind tunnel. They addressed the
study to neutrally buoyant plumes solely because, as discussed in the paper, it
appears not to be possible to perform the same study for buoyant plumes in a
small-scale laboratory. They simulated both sub-critical (Reynolds numbers
below the critical Reynolds number, ≅ 2 × 105) and supercritical (Reynolds
numbers above the critical Reynolds number) turbulent flow. Sub-critical
Reynolds numbers are typically attained by small-diameter stacks in relatively
light winds; supercritical ones are attained by large-diameter stacks in strong
winds (supercritical regimes are typical of the majority of full-scale stacks). The
downwash characteristics differ markedly in the two regimes. For example,
Snyder and Lawson (1991) found that downwash is much more serious in the sub-
critical case than in the supercritical one. Furthermore, in the sub-critical regime,
downwash begins when the ratio of effluent speed, vs0, to wind speed, u0, is less
than about 1.5; while in the supercritical regimes, downwash begins when such
ratio is less than about 1.1. Empirical expressions are provided for vertical plume
widths in the sub- and supercritical regimes, for lateral plume widths in the
supercritical flow regime (not measured in sub-critical regime), and for plume
centroids in the supercritical regime – the centroids in the sub-critical regime are
too complex to be fitted by simple expressions.

Overcamp (2001) studied the range of conditions that may lead to downwash in
designing simulation of buoyant plumes in wind tunnels and towing tanks. He
made a comparison between data on the occurrence of downwash from ten sub-
critical model studies and the theory proposed by Tatom (1986) - reference from
Overcamp (2001). The Tatom’s theory predicts that downwash does not occur if
the following implicit relationship is satisfied
160 Air Quality Modeling – Vol. I

(3C D )2 3 − 1
R′ ≥ (114)
 
1 4B 
πA +
F 2
 r ( 2
4 + B2  )
where:

1
 ρ  2 vs0  B2  R′
R ′ =  s 0  ; A =1− exp −  ; B = (115)
 ρ a0  u0  2  β + αR ′

and CD is the drag coefficient. All the ten independent experiments that Overcamp
considered were characterized by R ′ < 2 . He found that there was good
agreement of Tatom’s theory with the occurrence of downwash.

5.1.2 Building Downwash

We describe in some detail the method for taking into account the building
downwash proposed by Schulman-Scire (Schulman and Scire, 1980; Scire and
Schulman, 1980; Schulman and Hanna, 1986), because this method is
implemented in both the ISC3 (U.S. EPA, 1995b) and AERMOD (U.S. EPA, 1998)
models and is inserted in the CALPUFF code (Scire et al, 1999;
[Link] as well. Then, more recently developed
building downwash parameterizations will be also presented.

The Schulman-Scire method incorporates the effects of building downwash both

on reducing plume rise and on enhancing dispersion parameters.
• σy0 ≤ σz0
neutral-unstable conditions

the plume rise, ∆hd(x), of a downwashed plume is the real solution of the cubic
equation

 3 R ∆h 3R 2  3F x 3F x 2
∆hd3 +  0 d + 20  ∆hd = 2m 2 + b2 3 (116)
 β1 β1  β j u 0 2 β1 u 0

where β1 is the neutral entrainment parameter (∼ 0.6), βj is the jet entrainment

coefficient ( β j = 1 3 + u0 vs 0 ), R0 = 2 σ z 0 is the dilution radius, and σy0, σz0 are
the horizontal and vertical dispersion coefficients, respectively, at a downwind
distance of 3Hb (Hb = building height);
6 Plume Rise 161

- final stable plume rise

 3 R ∆h 3R 2  3F 6F
∆hd3 +  0 d + 20  ∆hd = 2 m 1 2 + 2 b (117)
 β2 β2  β j u0 s β 2 u0 s

where β2 is the stable entrainment parameter (∼ 0.36). Transitional plume rise

during stable conditions is computed with Equation (116) until the final plume
height predicted by Equation (117) is obtained.

• σy0 > σz0

It is necessary to account for the elongated shape of the plume caused by

horizontal mixing of the plume in the building wake; the plume can be
represented as a finite line source:

- neutral-unstable conditions

the plume rise, ∆hd(x), for a line source of length Le is

 3L   3R ∆h 6 R L 3R 2 
∆hd3 +  e  ∆hd2 +  0 d + 0 2e + 20  ∆hd =
 π β1   β1 π β1 β1 
(118)
3F x 3 F x 2
= 2m 2 + b2 3
β j u 0 2 β1 u 0

- final stable plume rise

 3L   3R ∆h 6 R L 3R 2 
∆hd3 +  e  ∆hd2 +  0 d + 0 2e + 20  ∆hd =
π β 2   β2 π β2 β2 
(119)
3F 6F
= 2 m1 2 + 2 b
β j u0 s β 2 u0 s

The effective line length is Le = 2π (σ y 0 − σ z 0 ) if σy0 > σz0; otherwise Le = 0 and

Equations (118) and (119) reduce to Equations (116) and (117).

The enhanced dispersion coefficients, σy0 and σz0, vary with stack height,
momentum rise, and building dimensions. As σy0 and σz0 approach zero (e.g.,
building downwash effects become negligible), Equations (116) to (119)
approach the unmodified Briggs (1975) equations. The effect of R0 and Le is
always to lower the plume height, thereby tending to increase the predicted
ground-level concentration.
162 Air Quality Modeling – Vol. I

Hanna et al. (1998) developed a model to describe the lift-off of ground-based

buoyant plumes using wind tunnel observations. Special emphasis was given to
the development of simple empirical lift-off equations for buoyant plumes, which
are trapped in building wakes. The model was developed using wind tunnel
observations of plumes for which buoyancy was conserved, but the authors also
proposed to use it for plumes whose buoyancy flux varies with distance
(phenomenon that can occur due to the presence of aerosols, chemical reactions,
and evaporation and condensation processes). Hanna et al. (1998) suggested that
the effects of plume lift-off can be accounted for by multiplying the calculated
ground-level concentration in the absence of lift-off by an exponential term
depending on buoyancy flux. For buoyant plumes trapped in building wakes, the
empirical formula that is proposed combines the exponential term with four
additional terms related to the spread of plumes in building wakes. Such lift-off
formula is incorporated in the HGSYSTEM/UF6 hazardous gas dispersion code
(Hanna and Chang, 1997).

The ADMS code (e.g., see Carruthers et al., 1999) includes a module for building
effects based on the model of Hunt and Robins (1982). This module computes the
dispersion of pollution from sources near isolated large buildings or closely spaced
blocks. The model is able to deal with the influence on turbulent and mean velocity
field of an extensive downstream wake. A simplified flow field is defined, based on a
well mixed cavity (or recirculating flow region) and a downstream momentum wake.
It takes into account the source position and allows for complete or partial
entrainment into the recirculating flow region. Within the recirculating flow region
concentrations are uniformly calculated. For partially entrained emissions, the
entrained and non-entrained components form a two-plume structure downwind.
Alternative spread parameters describe dispersion inside and outside the downstream
wake.

Flowe and Kumar (2000) showed that a three-dimensional turbulent kinetic

energy/dissipation (k-ε) numerical model, FLUENT, can be used as a tool for
modeling air flow past a building and stack geometry, and the recirculation cavities
associated with wide buildings, and to develop parameterizations useful to air quality
modeling needs. These modeling capabilities were proved through the comparison
with experimental wind tunnel data generated for several ratios of building width to
building heights. Then, the flow field was examined to determine the length of the
recirculation cavity as a function of the ratio of building width to building height
both in front of and in the rear of the building. The height and length of the front
recirculation cavity were parameterized as a function of the ratio of building width to
building height. This is a novelty as far as regulatory models are concerned.

Schulman et al. (2000) proposed the Gaussian dispersion model PRIME for plume
rise and building downwash. The plume trajectory within the modified fields
downwind of the building is estimated using the Zhang and Ghoniem (1993 – see
Section 3.4) numerical plume rise model. Such model is based on a numerical
solution of the mass, energy and momentum conservation laws. It allows arbitrary
6 Plume Rise 163

ambient temperature stratification, uni-directional wind shear, and initial plume size.
A cavity module calculates the fraction of plume mass captured by and recirculated
within the near wake. The captured mass is re-emitted to the far wake as a volume
source and added to the uncaptured plume contribution to obtain the far wake
concentrations. The PRIME model is implemented within the ISC3 code (Schulman
et al., 1997), but it can be implemented in other refined or screening air quality
models

5.2 Penetration of Elevated Inversions

Elevated inversions can be divided in thin and thick inversions according to their
depth: when the plume cross section is greater than the inversion layer thickness,
we have the case of the thin inversion, whereas when the entire plume cross
section is contained in the inversion layer, we have the case of a thick inversion.

Plume buoyancy is often large enough to allow plumes to fully or partially

penetrate an elevated temperature inversion layer (see Figure 8). Plume material
will penetrate an inversion if the temperature excess of a part of the plume at a
given height exceeds the temperature change through the layer at the same height.
This typically happens during daytime, where the CBL is generally capped by
stable air. In the case of a thin inversion, the potential temperature jump ∆ϑi is the
important parameter, whereas in the case of a deep inversion layer the potential
temperature gradient, ∂ϑi ∂z is the characteristic quantity. Consequently, the fate
of the plume depends upon these parameters and on the inversion base height
(Zannetti, 1990; Weil, 1988). A plume, which is able to completely penetrate the
inversion, makes little or no ground level concentration contribution. On the
contrary a plume trapped below the inversion can easily be diffused towards the
ground bringing about consistent ground-level concentrations (fumigation).

Most of present applied dispersion models (Weil, 1988) only distinguish between
complete penetration and no penetration. However many studies (see, for
instance: Manins, 1979 or Thompson et al., 2000) have shown that the situation is
not so simple and more detailed methods are needed. In particular, Manins (1979)
and Zannetti (1990) concluded that a complete plume penetration is almost
impossible since, upon reaching the inversion, there will be a part of the plume
having insufficient buoyancy for further rise. This was also qualitatively shown
by LSM simulations (Zannetti et Al Madani, 1983 and 1984). Thus, it is
important to know the fraction of the plume that is trapped.

The simple method provided by Turner (1985) for discriminating between these
two cases in presented in Section 5.2.1. Then, we review other penetration models
for bent-over plume: first, for a thin inversion, the Briggs (1975), Manins (1979),
and Weil (1988) models, see Section 5.2.2; then, for a thick one, the Briggs
(1984), and Berkowicz et al. (1986) models, see Section 5.2.3.
164 Air Quality Modeling – Vol. I

Figure 8. Schematic of the interaction of a buoyant plume and an elevated

inversion layer (from Manins, 1979). [Reprinted with permission from
Pergamon Press.]

5.2.1 The Turner Procedure

Turner (1985) worked out a pragmatic method using a modification to Briggs

(1975, 1984) formulae for computing the final buoyant plume rise by layers and
the possible (partial or total) penetration of the plume above the atmospheric
thermal discontinuities (such as, typically, the mixing height).

Turner method considers that the plume, during its rise, may meet atmospheric
layers of different wind speed and stability. To use this method, one must know or
estimate the values of temperature and wind speed close to the stack, on at least
two different levels. One at a height between ground level and the stack outlet
height, the other at an elevation higher than that reached by the upper edge of the
plume at the end of its rise (given by he + 0.5∆h ). Obviously, to be able to make
optimum use of Turner method, one ought to know the values of temperature and
wind speed at numerous intermediate levels, as well as at the previous two levels.
Furthermore it is also assumed that the mixing height and the rate of change of
potential temperature with height above the mixing height are available.

This procedure for computing the final plume rise consists of the following steps.

1) Calculation of the stack tip downwash factor (f) through Bjorklund and
Bowers’ model (see Section 5.1.1);
• if f = 0
6 Plume Rise 165

Turner method provides a null final plume rise and an effective emission (see
item 4 below) equal to the emission at the stack.
• if f > 0 we can go onto the following steps;

2) Calculation of the final plume rise keeping into account the plume transit
through some atmospheric layers having different characteristics of temperature,
wind speed and stability.

We start computing the final plume rise, by using a modification to Briggs (1975,
1984) formulae and the meteorological parameters of the atmospheric layer that
includes the stack outlet. If either the obtained final plume rise (in neutral or
unstable conditions) or the upper edge of the plume (in stable conditions) does not
overtake the top of the layer containing the stack outlet, the calculated final plume
rise is the results of the first part of Turner method. If neither of them does, we
have to compute the residual buoyancy, and use it to repeat the computation
procedure of the final plume rise of the next layer; the new plume rise has the
same fate as the previous one. The process goes on from layer to layer till the
result will be obtained.

3) The final plume rise computed through the procedure illustrated at point 2) is
adjusted by the stack tip downwash factor (see Section 5.1.1).

4) To calculate the penetration of the plume above atmospheric thermal

discontinuities, this procedure assumes that a fraction f ′ ( 0 ≤ f ′ ≤ 1 ) of the total
emission, Q, remains trapped below the base of the thermal discontinuity, placed
at the height ht, and affects the concentration measured by receptors below this
height. The product f ′ Q is known as the “effective emission”.

As far as f ′ is concerned two options are possible:

a) f ′ = 1 ; this means that this option disregards the penetration of the plume
above the thermal discontinuity, but takes only into account the modification
made to the plume rise;

b) f ′ ≠ 1 and the modified plume rise are calculated with the method discussed
below.

Option b) can be chosen only if ht is greater than the distance of receptors from
ground level. With this method, three possibilities are considered, depending on
the values taken by the parameters

t p = he + 0.5∆h (120)

and
166 Air Quality Modeling – Vol. I

b p = he − 0.5∆h (121)

assumed by Turner to represent the upper and the lower edge of the plume. These
three possibilities are:
• if t p ≤ ht , no penetration of the plume above the thermal discontinuity is
assumed, i. e. f ′ = 1;
• if b p ≥ ht , the entire plume is assumed to penetrate above the thermal
discontinuity, i. e. f ′ = 0 ; no concentration is measured by receptors;
• in the intermediate case, i. e. b p < ht < t p , Turner proposes

ht − b p
0< f′= <1 (122)
∆h

and

1+ f ′
∆h′′ = ∆h (123)
2

as the actual plume rise, instead of ∆h .

This plume rise/partial penetration technique provides a computationally simple

solution for engineering calculations. However, the problem of correctly
modeling partial penetration is still wide open.

One can observe that, since this method requires a detailed knowledge of the
various atmospheric layers crossed by the plume (e.g., coming from vertical
profile observations), it would be similarly simpler to solve directly the
conservation equations as mentioned in Section 3. However, this method was
recalled here since it is incorporated in some dispersion models, like, for instance,
in the TUPOS model (Turner et al., 1986), in the PTSRCE preprocessor program
of UAM-V (U.S. EPA12), and in the SAFE_AIR package (Canepa et al., 20003) as
a user option.

5.2.2 Thin Inversion

For a vertical plume, Briggs (1975) predicts that a thin inversion layer can be
completely penetrated if the mean temperature excess of the plume at height h′ (h′
= ht – zs) exceeds the temperature jump ∆ϑi. Defining bi = ( g ϑ ) ∆ϑi , complete
penetration occurs if

1
[Link]
2
[Link]
3
[Link]
6 Plume Rise 167

Fb > 0.019 bi3 2 (h′)

52
(124)

for a buoyant plume, and

Fm > 0.25 bi (h′)
3
(125)

for a jet. This last equation is based upon experimental results by Vadot (1965).

For bent over buoyant plumes the finite depth of the plume cannot be neglected
and, consequently, partial penetration is more likely than complete penetration.
The Briggs (1975) model considers the plume buoyancy depletion during the
inversion traverse. Defining an equilibrium height with respect to the top of the
stack, z′eq, where its buoyancy flux is equal to zero and assuming the plume cross-
section to be rectangular with a depth equal to the rise ∆h, and a width equal to
0.5 ∆h, z′eq is found to be (Briggs, 1975; Weil, 1988):

′
z eq 2
= (1 + 9π Pb )1 2 (126)
h′ 3

where the dimensionless buoyancy flux Pb is given by

Fb
Pb = (127)
U a bi h ′ 2

The percentage of plume trapped by the inversion and thus diffused downwards is

h′
f ′ = 1 − Pb = − 0.5 (128)
′
z eq

From which the following simple criteria derive:

2
′ < h′ , no penetration (f′ = 1);
• z eq
3

2
• h ′ < z eq
′ < 2 h ′ , partial penetration - f′ is given by Equation (128);
3

• ′ > 2 h ′ , complete penetration (f′ = 0).

z eq

For bent over jets, substantial inversion penetration may be assumed (Briggs,
1975) when

Fm > 2.2 β 2u0bi1 2 (h′)

52
(129)
168 Air Quality Modeling – Vol. I

The Manins (1979) model is based on the assumption that the density or
temperature is normally distributed in a bent over buoyant plume when its
centerline reaches the inversion. As in the previous Briggs (1975) model, the
inversion is idealized as a jump of zero thickness. Defining ∆ϑm to be the
maximum excess temperature, Manins also assumed that penetration starts when
∆ϑm = ∆ϑi, and found that this happens if

Pb = 0.08 (130)

He suggested that partial penetration would occur, for the part of the plume with
∆ϑ > ∆ϑi, when Pb > 0.08. Accounting for the effects of the momentum overshoot
of the plume and the re-entrainment back into the plume of material trapped
within the inversion, the above condition leads to the following expression for the
fraction of the plume trapped in the inversion layer

0.08
f′= − (Pb − 0.08 ) (131)
Pb

Weil (1988) compared laboratory observations (see Figure 9) of inversion

penetration by Manins (1979) and Richards (1963) and noted that Manins model,
Equation (131), fits reasonably well his data but overestimates most of Richards
data while Briggs model, Equations (126) and (128), overestimates part of the
observed f ′ . Weil argues that these differences between models and observations
can possibly be due to some different configurations in the experimental
conditions between the two experiments and, in particular, of the ratio ∆hi h′ ,
where ∆hi is the finite thickness of the inversion layer. As a consequence, Weil
(1988) considers the effect of ∆hi on the plume penetration capacity and of a
different temperature distribution. He found that the fraction of the plume below
the inversion top, h′ + ∆hi, is

f ′ = 1−
1
π
[cos −1
λ − λ (1 − λ2 )
12
] (132)

where
1 + δ − η eq
λ= (133)
β ′η eq

′
z eq ∆hi
with η eq = , and δ = .
h′ h′
6 Plume Rise 169

Figure 9. Models and laboratory measurements of the fraction of a plume

trapped by an elevated inversion as a function of the dimensionless
buoyancy flux P (adapted from Weil, 1988). [Reprinted with permission
from American Meteorological Society]

5.2.3 Thick Inversion

In this case the reference inversion height h′, to calculate penetration and trapping
probabilities of occurrence, is the height of the inversion base. Briggs (1975,
1984) considered the simple case in which s is constant with height. He also
assumed that plume equilibrium height is given by Equation (44), namely

13
 F 
′ = 2.6  b 
zeq (134)
 su0 

in which s is computed from dϑi/dz. To estimate the fraction of the plume trapped
f’ =1 - Pb, he used Equation (128), obtaining
2
′ < h′ , no penetration (f′ = 1);
• z eq
3

2
• h ′ < z eq
′ < 2 h ′ , partial penetration - f′ is given by Equation (128);
3

• ′ > 2 h′ , complete penetration (f′ = 0).

zeq

Briggs’ (1984) model gives conservative estimates since the plume initially rises
in an atmosphere with s = 0, in which it should not experience any buoyancy
depletion.
170 Air Quality Modeling – Vol. I

Berkowicz et al. (1986) considered this aspect. By assuming that the process of
buoyancy reduction initiates only when the upper boundary of the plume arrives
′ h′
at h′ , they proposed the following equation for the ratio zeq

h′
′
z eq
[
= 2.63 Ps + (2 3) ]
3 13
(135)
where

Fb
Ps = (136)
U a N i2 h ′ 3

5.3 Plume Rise from Multiple Sources

When several stacks are located close to each other, the resulting plume rise is
different from that of a single stack. Plumes coming from the various stacks
generally merge during the rise stage thus causing enhanced rise due to reduced
ambient air entrainment and increased buoyancy. Consequently, ground level
concentration is reduced. This enhanced plume rise was observed both in
laboratory experiments and in the field (Manins et al., 1992). In general the
enhancement is greater in the case of flow parallel to the stacks than in the normal
flow and, in both cases, the plume rise exceeds that of a single plume (Anfossi,
1985). Overcamp and Ku (1988) also confirmed that enhancement is a function of
the angle between the direction of the wind and the line of stacks, finding that the
rise is larger when the angle is small. In the same way, also plumes coming from
cooling towers (Bornoff and Mokhtarzadeh-Dehghan, 2001) or from multiple
fires (Trelles et al., 1999b) may merge and experience enhanced rise.

Briggs (1975) provided a semi empirical formulation for determining the plume
rise in the case of stacks of equal height and buoyancy flux. He defined the
enhancement factor, En, as the ratio of the plume rise from n stacks to that of one
stack, whose expression is the following

n+S 3
1
 n −1  2

En =   , S = 1  (137)
 1+ S   n 3 ∆h / d 
 1 

where d is the spacing between the stacks and ∆h1 is the plume rise from a single
stack.

Anfossi et al. (1978) developed and tested (Anfossi et al. 1979; Anfossi, 1982,
1985; Sandroni et al., 1981) a virtual stack concept that allows two or more stacks
of different buoyancy and heights to be merged. Their model for the plume rise
from multiple sources is expressed by the following equation
6 Plume Rise 171

{ ]}
1

[
= H i + C ∑ F j 3 − (H i − H j ) / C
N 1 3
3
∆H N
(138)
j =1

in which

∆H min − (H max − H min )

H i = H max + (139)
1 + [∆H min − (H max − H min )] / D

is the merging point height, ∆H min is the maximum single plume rise from lowest
stack H min , C = ∆H min / Fmin3 and D = (n − 1)d . In the case of stacks of equal
1

height and buoyancy flux, Equations (138) and (139) reduce to

1 + n 3 (∆h1 / D )
1

En = (140)
1 + (∆h1 / D )

Anfossi (1985) and Manins et al. (1992) demonstrated that Equations (137) and
(140) give almost equal results.

5.4 Plume Rise from Flare Stacks

A flare stack is a vent gas stack with a small pilot flame at the stack exit.
Combustible vent gases flowing from the stack exit are ignited by the pilot flame
and burned in the open atmosphere just above the stack exit. The hot, combusted
gas plume then rises and disperses in the atmosphere just as does any hot, buoyant
plume. Flare stacks are widely used in industrial plants; in particular, flare stacks
are an essential safety requirement in hydrocarbon processing facilities.

By means of direct observation, Leahey and Davies (1984) showed that the
entrainment of ambient air into the flare plume is similar to what found in stack
plumes and that the flare plumes rise according to the "two-third" law.

The SCREEN3 model (U.S. EPA, 1995a) deals with flare. Buoyancy flux for flare
release is estimated from

Fb = 1.66 10 −5 Q f (141)

where Qf is the total heat release rate of the flare (cal s-1). This formula - see
Equation (9) - was proposed by Briggs (1969). The value of the constant was
derived fixing Ta = 293 K, ρa = 1205 g m-3, cp = 0.24 cal g-1 K-1, and assuming the
following relationship between Qf and the sensible heat release rate Qh: Qh = 0.45
Qf. The sensible heat rate is based on the assumption that 55 % of the total heat
released is lost due to radiation (Leahey and Davies, 1984). The buoyancy flux
172 Air Quality Modeling – Vol. I

for flares is calculated in SCREEN by assuming effective stack parameters of vs0

= 20 m s-1, Ts0 = 1273K, and solving for an effective stack diameter, ds = 9.88 10-4
(Qh)0.5.

5.5 Plume Rise from Fires

Environmental consequences of large fires are of interest since the rise and
transport of combustion products can distribute potentially hazardous materials
over a wide area (McGrattan et al., 1996). Plume rise simulation from fires is not
straightforward. For example, the life cycles of a forest fire includes an initial
developing stage with large increases in heat generation and pollutant emissions
rates, followed by a stage of decreasing values. Therefore, the source parameters
of a forest fire are usually not constant. The magnitude of the variation in heat
generation and emission rates may be two orders of magnitude over the course of
burn (Scire et al, 1999).

Various models dealing with plume rise from fires are available in the literature.
For instance Manins (1985) considered plumes from fires from thermonuclear
explosions (direct bomb fires, incineration of the immediate blast area and
injection from fires which spread from the blast area). The prediction of fire
plume-rise was based on the Boussinesq buoyant plume model of Morton et al.
(1956) since this was shown by Turner (1973) and Briggs (1975) to give good
results for small to large heat sources when the ambient wind is light (see also
Section 2.1).

Recently McGrattan et al. (1996) presented a LES model of smoke plumes

generated by large outdoor pool fires transported by a uniform ambient wind. This
model was extended by Trelles et al. (1999a) to deal with the problem of large-
scale fire plumes in the presence of winds, which vary, in the vertical direction. A
further extension performed by Trelles et al. (1999b) investigated multiple fire
plumes. In fact large scale fire scenarios commonly involve multiple combustion
sources: the class of problems considered excludes fires large enough to alter the
prevailing atmosphere, but it allows for fires sufficiently strong to interact with
each other and to have local atmospheric influence.

Also some regulatory computer codes include the treatment of plume rise from
fires. For example, the FIREPLUME code (Brown et al., 1999) and the CALPUFF
code (Scire et al, 1999; [Link] that are briefly
described below.

FIREPLUME (Brown et al., 1999) is able to simulate atmospheric dispersion and

air quality impacts from fires. FIREPLUME deals with plume rise by means of
the MCLDM Lagrangian particle model (Brown et al, 1996). The framework for
treating source buoyancy closely follows from the “two-thirds” law, which is
applicable in cases where the buoyant source has low initial momentum. Fires
clearly fall into this category (Weil, 1982). Although the “two-thirds” law is
6 Plume Rise 173

primarily used for stack emissions, its extension to fire buoyancy is

straightforward. The plume rise relationships are incorporated into MCLDM to
provide a mean vertical velocity for the individual particles. The vertical
dispersion from a variety of buoyant release scenarios can be evaluated, from
intensely buoyant sources typical in actively burning forest fires to very low
buoyancy sources, such as, in the residual stages of smoldering biomass.

In CALPUFF (Scire et al, 1999; [Link] the area

source plume rise model is formulated to calculate the rise of buoyant plumes
resulting from forest fires, the burning of leaking oil, and other type of buoyant
area sources. The model is designed to be applicable to the following conditions:
1) all types of ambient temperature stratifications; 2) all types of wind
stratifications (wind shear is important because the forest fire plume starts at
ground where there is a zone of large velocity gradients in the vertical); 3) any
size of finite emission source; 4) include the effects of plume radiative heat loss;
and 5) Boussinesq approximation is not assumed.

5.6 Plume Rise from Stacks with Scrubber

Desulfurization techniques have often been adopted for either the combustibles
(e.g., coal cleaning) or the flue gas (scrubbers). The latter technique seems by far
the most cost effective for SO2 emission reduction. Most flue gas desulfurization
devices employ a wet scrubbing technique in which a Ca(OH)2 solution is used
for partial removal of SO2.

Plumes from stacks with scrubbers are frequently modeled using the same
techniques as the other plumes. Schatzmann and Policastro (1984) reviewed the
problem of evaluating ∆h for stacks with scrubbers, concluding that “the
significant moisture content of the scrubbed plume upon exit leads to important
thermodynamic effects during plume rise that are unaccounted for in the usual dry
plume rise theories”.

Plume rise models for wet plumes (e.g., cooling tower plumes) have been
developed by Hanna (1972), Weil (1974), and Wigley and Slawson (1975). Even
these formulations, however, are inappropriate for scrubbed plumes, according to
Schatzmann and Policastro (1984), because of the simplifications they adopt.
Sutherland and Spangler (1980) compared observed plume rise heights for
scrubbed and unscrubbed plumes and evaluated the performance of several plume
rise formulations. They found that simple plume rise formulae are questionable
even for dry plumes, while moisture effects in scrubbed plumes increase the
plume buoyancy and almost compensate for the loss of plume rise due to the
temperature decrease induced by the scrubbing system. Plume rise of moist
plumes was reviewed by Briggs (1984).

Schatzmann and Policastro (1984) recommend integral-type models for scrubbed

plumes, with the additional requirement of avoiding some common
174 Air Quality Modeling – Vol. I

simplifications such as the linearization of the equation of state, first-order

approximations in the calculation of the local saturation deficit, and the
Boussinesq approximation.

References

Abraham, G. (1970): Round buoyant jet in cross-flow. Paper presented at the 5th International
Conf. Water Pollut. Res., San Francisco, CA.

Anfossi, D., G. Bonino, F. Bossa, and R. Richiardone (1978): Plume rise from multiple sources: a
new model. Atmos. Environ., 12:1821-1826.

Anfossi, D., G. Bonino, and R. Richiardone (1979): An application of plume rise model for
multiple sources to the cooling tower plumes of John E. Amos Power Plant. Il Nuovo Cimento,
2C,N.4:488-498.

Anfossi, D. (1982): Plume rise measurements at Turbigo. Atmos. Environ., 16:2565-2574.

Anfossi, D. (1985): Analysis of plume rise data from five TVA Steam Plants. J. Clim. App. Met.,
24:1225-1236.

Anfossi, D., E. Ferrero, G. Brusasca, A. Marzorati, and G. Tinarelli (1993): A simple Way of
Computing Buoyant Plume Rise in Lagrangian Stochastic Dispersion Models, Atmos. Environ.,
27A:1443-1451.

Anfossi, D. (2000): Jet plume rise description in Lagrangian Stochastic Models. ICGF Rapp. Int.
No 414/2000, 9 pp.

Arya, P. S., (1999) Air pollution meteorology and dispersion. Oxford University Press, New
York. [Link]

Berkowicz, R., H.R. Olesen, and U. Torp (1986): The Danish Gaussian air pollution model
(OML): description, test and sensitivity analysis in view of regulatory applications. Air Pollution
Modeling and Its Application V. C. De Wispelaere, F.A. Schiermeier, and N.V. Gillani Editors.,
Plenum, New York, 453-481.

Bjorklund, J.R., and J.F. Bowers (1982): User’s instruction for the SHORTZ and LONGZ
computer programs, Volumes I and II. EPA Document EPA-903/9-82-004A and B, U.S. EPA,
Middle Atlantic Region III, Philadelphia, Pennsylvania, USA.

Bornoff R.B. and Mokhtarzadeh-Dehghan M.R. (2001) A numerical study of interacting buoyant
cooling-tower plumes. Atmos. Environ., 35, 589-598

Briggs, G.A. (1969): Plume rise. AEC Crit. Rev. Ser., TID-25075. Springfield (VA), pp. 80.

Briggs, G.A. (1972): Chimney plumes in neutral and stable surroundings. Atmos. Environ. 6:507-
510.

Briggs, G.A. (1973): Diffusion estimation for small emission. In Atmospheric Turbulence and
Diffusion Laboratory 1973 Annual Report. (NOAA/ATDL, Oak Ridge, TN, publication no.
ATDL-106, pp 739-747
6 Plume Rise 175

Briggs, G.A. (1975): Plume rise predictions. In: Lectures on air pollution and environmental
impact analyses, D. Haugen (editor). Workshop proceedings, Boston, Massachusetts, September
29 – October 3, pp. 59-111, American Meteorological Society, Boston, Massachusetts, USA.

Briggs, G.A. (1984): Plume rise and buoyancy effects. Technical Report Atmospheric Science and
Power Production, DOE/TIC-27601, edited by D. Randerson, U.S. Department of Energy,
Washington, D.C.

Bringfelt, B. (1969): A study of buoyant chimney plumes in neutral and stable atmospheres. Atmos.
Environ., 3:609-623.

Brown, D.F., W. E. Dunn, A.J. Policastro, and D. Malony (1996): FIREPLUME model for plume
dispersion from fires: applications to UF6 cylinder fires, ANL/EAD-PM69.

Brown, D.F., W.E. Dunn, M.A. Lazaro, and A.J. Policastro (1999): The FIREPLUME model: tool for
eventual application to prescribed burns and wildland fires. The joint fire science conference and
workshop. Boise,. Idaho, June 15-17, 1999.

Brummage, K.G. (1966): The calculation of atmospheric dispersion from a stack. Stichting
CONCAWE, The Hague, The Netherlands.

Canepa, E., F. Modesti, and C.F. Ratto (2000): Evaluation of the SAFE_AIR code against air
pollution field and laboratory experiments. Atmos. Environ., 34(28):4805-4818.

Carpenter, S.B., T.L. Montgimery, J.M. Leavitt, W.C. Colbaugh, and F.W. Thomas (1971): Principal
plume dispersion models: TVA power plants. J. Air Pollut. Control Ass., 21:491-495.

Carras, J.N., and D.J. Williams (1984): Experimental studies of plume dispersion in convective
conditions – 1. Atmos. Environ., 18:135-144.

Carruthers, D.J., A.M. Mckeown, D.J. Hall, and S. Porter (1999): Validation of ADMS against Wind
Tunnel Data of Dispersion from Chemical Warehouse Fires. Atmos. Environ., 33:1937–1953.

Chiang H.C. and Sill B.L. (1985): Entrainment models and their application to jets in a turbulent
cross flow. Atmos. Environ., 19:1425-1438

Cogan, J.L. (1985): Monte Carlo simulation of a buoyant dispersion. Atmos. Environ., 19:867-
878.

Craft, T.J., N.Z. Ince, and B.E. Launder (1996): Recent Developments in Second-Moment Closure
for Buoyancy-Affected Flows. Dynamics of Atmos. and Oceans, 23:99-114.

Csanady, G.T. (1973): Turbulent Diffusion in the Environment. Geophysics and Astrophysics
Monographs, Vol. 3. R. Reidel Publishing Company, Dordrecht-Holland, Boston-USA.

Djurfors, S.G. (1977): On the rise of buoyant plumes in turbulent environments. Syncrude Canada Ltd.
Professional Paper 1977-4.

Djurfors, S.G. (1983): Discussion of effective source flux parameters for use in analytical plume rise
models. Atmos. Environ., 17:197-198.

Erbrink, H.J. (1994): Plume Rise in Different Atmospheres: A Practical Scheme and some
Comparisons with LIDAR Measurements. Atmos. Environ., 28:3625-3636.
176 Air Quality Modeling – Vol. I

Fay, J.A., M.P. Escudier, and D.P. Hoult (1970): A correlation of field observations of plume rise. J.
Air Pollut. Control Assoc., 20:391-397.

Finardi, S., G. Tinarelli, A. Nanni, D. Anfossi, E. Ferrero, S. Trini Castelli (2001): In situ
diagnostic or nested prognostic meteorological models to drive dispersion simulations in complex
area: a comparison in a real application. Air Pollution Modelling and its Applications XIV, S.E.
Gryning and F.A. Schiermeier eds., Kluwer Academic / Plenum Press, New York.,641-649

Flowe, A.C., and A. Kumar (2000): Analysis of velocity fields and dispersive cavity parameters as
a function of building with to building height ratio using a 3-D computer model for squat
buildings. Journal of Wind Engineering and Industrial Aerodynamics, 86:87-122.

Gangoiti, G., J. Sancho, G. Ibarra, L. Alonso, J.A. García, M. Navazo, N. Durana, and J.L. Ilardia
(1997): Rise of Moist Plumes from Tall Stacks in Turbulent and Stratified Atmospheres. Atmos.
Environ., 31A:253-269.

Gardiner, C.W. (1983): Handbook of Statistical Methods. Springer Verlag, Berlin, Heidelberg,
New York, Tokyo.

Glendening, J.W., J.A. Businger, and R.J. Farber (1984): Improving plume rise prediction accuracy for
stable atmospheres with complex vertical structure. J. Air Pollut. Control Ass., 34:1128-1133.

Golay, M.W. (1982): Numerical modeling of buoyant plumes in a turbulent stratified atmosphere.
Atmos. Environ., 16:2373-2381.

Graziani G., Martilli A., Pareschi M.T. and Valenza M. (1997) Atmospheric dispersion of natural
gases at Vulcano island. Journal of Volcanology and Geothermal Research, 75, 283-308

Hanna, S.R. (1972): Rise and condensation of large cooling tower plumes. J. Appl. Meteor.,
11:793-799.

Hanna, S.R., G.A. Briggs, and R.P. Jr. Hosker, (1982): Handbook on atmospheric diffusion.
Prepared for the Office of Health and Environmental Research, Office of Energy Research, U.S.
Department of Energy, DOE/TIC 11223, J.S. Smith, Publication Editor.

Hanna, S.R., Chang J.C. Strimaitis D.G. (1993): Hazardous gas model evaluation with field
observations. Atmos. Environ., 27A:2265-2285.

Hanna, S.R. (1994): Lecture on Air quality modeling over short distances. College on atmospheric
boundary layer and air pollution modelling. ICTP Trieste, 16 May-3 June 1994.

Hanna, S.R., and J.C. Chang (1997): HGSYSTEM/UF6 model enhancements for plume rise and
dispersion around buildings, lift-off of buoyant plumes, and robustness of numerical solver.
K/SUB/93-XJ947/R, OSTI, P.O. Box 62, Oak Ridge, TN 37831.

Hanna, S.R., G.A. Briggs, and J.C. Chang (1998): Lift-off of ground-based buoyant plumes.
Journal of Hazardous Materials, 59:123-130.

Heinz, S. (1997): Nonlinear Lagrangian Equations for Turbulent Motion and Buoyancy in
Inhomogeneous Flows. Phys. Fluids, 9:703-716.

Heinz, S. (1998): Time Scales of Stratified Turbulent Flows and Relations between Second-Order
Closure Parameters and Flow Numbers. Phys. Fluids, 10:958-973.
6 Plume Rise 177

Heinz, S., and H. van Dop (1999): Buoyant plume rise described by a Lagrangian turbulence
model. Atmos. Environ., 33:2031-2043.

Henderson-Sellers, B., and S.E. Allen (1985): Verification of the plume rise/dispersion model
USPR: plume rise for single stack emissions. Ecological Modelling, 30:209-277.

Hewett, T.A., J.A. Fay, and D.P. Hoult (1971) Laboratory experiments of smokestack plumes in a
stable atmosphere. Atmos., Environ., 5:767-789.

Holland, J.Z. (1953): Meteorology survey of the Oak Ridge area. ORO-99. U.S. At. Energy
Comm., Oak Ridge, Tennessee.

Hoult, D.P., and J.C. Weil (1972): A turbulent plume in a laminar crossflow. Atmos. Environ.,
6:513-531.

Huber, A.H., and W.H. Snyder (1982): Wind tunnel investigation of the effects of a rectangular-
shaped building on the dispersion of effluents from short adjacent stacks. Atmos. Environ.,
17:2837-2848.

Hunt, J.C.R., and A.G. Robins (1982): A model for assessing dispersion of plumes from sources
in the vicinity of cuboid shaped buildings. Proceedings of the EUROMECH Conference on
Surface Mounted Bluff Bodies in Turbulent Boundary Layers, Lisbon.

Hurley, P.J., and W. Physick (1993): Lagrangian particle modelling of buoyant point sources:
plume rise and entrapment under convective conditions, Atmos. Environ., 27A: 1579-1584.

Hurley, P.J., and P.C. Manins (1995): Plume rise and enhanced dispersion in LADM. ECRU
Technical Note No.4, CSIRO Division of Atmospheric Research, Australia.

Hurley, P.J. (1999): The Air Pollution Model (TAPM) Version 1: Technical Description and
Examples. CSIRO Atmospheric Research Technical Paper No. 43.

Hurley, P.J. (2000): TAPM Technical Report. TAPM web page

[Link]

Hurley, P.J., A. Blockley, and K. Rayner (2001): Verification of a prognostic meteorological and
air pollution model for year-long predictions in the Kwinana industrial region of Western
Australia. Atmos. Environ., 35:1871–1880.

Janicke, L. (1983): Particle simulation of inhomogeneous turbulent diffusion. In: Weber, B. (Ed.),
Air Pollution Modeling and its Application II. Plenum Press, New York, 527-535.
[Link]

Janicke, U., and L. Janicke (2001): A three-dimensional plume rise model for dry and wet plumes.
Atmos. Environ., 35:877-890.

Kaimal, J.C., Wyngaard, J.C., Haugen, D.A., Cote', O.R., Izumi, Y., Caughey, S.J. and Readings,
C.J. (1976). Turbulence structure in the convective boundary layer. J. Atmos. Sci., 33, 2152-2169.

Leahey, D.M., and M.J.E. Davies (1984): Observations of plume rise from sour gas flares. Atmos.
Environ., 18:917-922.

Loh-Nien Fan (1967): Turbulent buoyant jets into stratified or flowing ambient fluids. Technical
Report No. KH-R-15, California Institute of Technology, Pasadena, CA.
178 Air Quality Modeling – Vol. I

Luhar, A.K., and R.E. Britter (1992): Random walk modelling of buoyant-plume dispersion in the
convective boundary layer. Atmos. Environ., 26A:1283-1298.

Manins, P.C. (1979): Partial penetration of an elevated inversion layer by chimney plumes. Atmos.
Environ., 13:733-741.

Manins, P.C. (1985): Cloud heights and stratospheric injections resulting from a thermonuclear
war. Atmos. Environ., 19, 1245–1255

Manins, P.C., Carras J.N. and Williams D.L. (1992): Plume rise from multiple stacks. Clean Air,
26, 65-68.

McGrattan K.B., Baum H.R. and Rehm R.G. (1996) Numerical simulation of smoke plumes from
large oil fire. Atmos. Environ., 30:4125-4136

Moore, D.J. (1974): A comparison of the trajectories of rising buoyant plumes with
theoretical/empirical models. Atmos. Environ., 8:441-457.

Morton, B.R., G.I. Taylor, and J.S. Turner (1956): Turbulent gravitational convection from
maintained and instantaneous sources. Proc. Roy. Soc. London, A234:1-23.

Netterville, D.D.J. (1990): Plume rise, entrainment and dispersion in turbulent winds. Atmos.
Environ., 24:1061-1081.

Nguyen K.C., Noonan J.A., Galbally I.E. and W.L. Physick (1997) Predictions of plume
dispersion in complex terrain: Eulerian versus Lagrangian Models. Atmos. Environ., 31:947-958.

Nieuwstadt, F.T., and J.P. de Valk (1987): A large-eddy simulation of buoyant and non-buoyant
plume dispersion in the atmospheric boundary layer. Atmos. Environ., 21:2573-2587.

Nieuwstadt, F.T.M. (1992a): A Large-Eddy Simulation of a Line Source in a Convective

Atmospheric Boundary Layer - I. Dispersion Characteristics. Atmos. Environ., 26A:485-495.

Nieuwstadt, F.T.M. (1992b): A Large-Eddy Simulation of a Line Source in a Convective

Atmospheric Boundary Layer - II. Dynamics of a Buoyant Line Source. Atmos. Environ.,
26A:497-503.

Ooms, G. (1972): A new method for the calculation of the plume path of gases emitted by a stack.
Atmos. Environ., 6: 899–909.

Ooms, G., A.P. Mahieu, and F. Zelis (1974): The plume path of vent gases heavier than air. First
International Symposium on Loss Prevention and Safety Promotion in the Process Industries, The
Hague, The Netherlands.

Ooms, G., and A.P. Mahieu (1981): A comparison between a plume path model and a virtual
point source model for a stack plume. Applied Scientific Research, 36: 339–356.

Overcamp T.J and Ku T. (1988): Plume rise from two or more adjacent stacks. Atmos. Environ.,
22: 625–637.

Overcamp T.J (2001): A review of the conditions leading to downwash in physical modeling
experiments. Atmos. Environ., 35: 3503–3508.
6 Plume Rise 179

Physick, W.L. (1996) Photochemical smog studies in Australian cities. In: Urban air pollution,
volume 2. H. Power, and N. Moussiopoulos (editors). Southhampton: Computational Mechanics
Publications. p. 141–184.

Pierce, T.E., D.B. Turner, J.A. Catalano, and F.V. Hale (1982): PTPLU – A single source
gaussian dispersion algorithm. User’s Guide. Technical Report PB83-211235, U.S. Environmental
Protection Agency, Environmental Sciences Research Laboratory, Research Triangle Park, NC.

Pope, S.B. (1994): On the Relationship between Stochastic Lagrangian Models of Turbulence and
Second-Moment Closures. Phys. Fluids, 6:973-985.

Priestley, C.H.B. (1956): A working theory of the bent-over plume of hot gas. Quart. J. Roy. Met.
Soc., 82, 165–176

Richards J.M. (1963): The penetration of interfaces by cylindrical thermals. Quart. J. Roy. Met.
Soc., 89, 254-264

Risken, H. (1984): The Fokker-Planck Equation. Springer Verlag, Berlin, Heidelberg, New York.

Sandroni, S., P. Bacci, and D. Anfossi (1981): Aircraft observations of plume emitted from
elevated sources. Atmos. Environ., 15:95-100

Sawford, B.L. (1985): Lagrangian statistical simulation of concentration mean and fluctuating
fields. J. Climate Appl. Met., 24, 1152-1166

Sawford, B.L. (1986): Generalized Random Forcing in Random-Walk Turbulent Dispersion

Models. Phys. Fluids, 29:3582-3585.

Schatzmann, M. (1979): An integral model of plume rise. Atmos. Environ., 13:721-731.

Schatzmann, M., and A.J. Policastro (1984): An advanced integral model for cooling tower plume
dispersion. Atmos. Environ., 18:663-674.

Schulman, L.L., and J.S. Scire (1980): Buoyant line point source (BLP) dispersion model user’s
guide. Document P-7304-B, Environmental Research & Technology, Inc., Concord, MA.

Schulman, L.L., and S.R. Hanna (1986): Evaluation of downwash modifications to the Industrial
Source Complex Model. J. Air Pollut. Control Assoc., 36:258-264.

Schulman, L.L., D.G. Strimaitis, and J.S. Scire (1997): Addendum to ISC3 user’s guide. The
PRIME plume rise and building downwash model. Electric Power Research Institute.
[Link]

Schulman, L.L., D.G. Strimaitis, and J.S. Scire (2000): Development and evaluation of the
PRIME plume rise and building downwash model. J. Air Waste Manag. Assoc., 50:378-390.

Scire, J.S., and L.L. Schulmann (1980): Modeling plume rise from low-level buoyant line and point
sources. Proceedings Second Point Conference on Applications of Air Pollution Meteorology, 24-28
March 1980, New Orleans, LA, 133-139.

Scire J.S., D.G. Strimaitis, and R.J. Yamartino (1999): A User’s Guide for the CALPUFF
Dispersion Model (Version 5.0). Technical Report. Earth Tech. Inc., 196 Baker Avenue, 01742
Concord, MA.
180 Air Quality Modeling – Vol. I

Shimanuki, A., and Y. Nomura (1991): Numerical simulation on instantaneous images of the
smoke released from a chimney. J. Met. Soc. Japan, 69:187-196.

Singer I.A. and Smith M.E. (1966): Atmospheric dispersion at Brookhaven National Laboratory. Int. J.
Air Water Poll., 10, 125

Slawson, P.R., and G.T. Csanady (1967): On the mean path of buoyant, bent-over chimney plumes. J.
Fluid Mech., 28:311-322.

Slawson, P.R., and G.T. Csanady (1971): The effect of atmospheric conditions on plume rise. J.
Fluid Mech., 47:33-49.

Snyder, W.H., and R.E. Lawson Jr. (1991): Fluid modeling simulation of stack-tip downwash for
neutrally buoyant plumes. Atmos. Environ., 12:2837-2850.

Souto M.J., J.A. Souto, V. Perez-Munuzuri, J.J. Casares, and J.L. Bermudez (2001): A comparison of
operational Lagrangian particle and adaptive puff models for plume dispersion forecasting. Atmos.
Environ., 35:2349-2360.

Strom, G.H. (1976): Transport and diffusion of stack effluents. In: Stern, A.C. (editor) Air
pollution, Vol.I. Academic Press, New York, USA.

Stuhmiller, J. (1974): Development and validation of a two-variable turbulence model. Science

Applications, Inc., Report SAI-74-509-LJ, La Jolla, California.

Sutherland, V.C., and T.C. Spangler (1980): Comparison of calculated and observed plume rise
heights for scrubbed and nonscrubbed buoyant plumes. Preprints, Second Joint Conference on
Applications of Air Pollution Meteorology, New Orleans, American Meteorological Society, pp.
129-132.

Thomson, D.J. (1987): Criteria for the Selection of Stochastic Models of Particle Trajectories in
Turbulent Flows. J. Fluid Mech., 180:529-556.

Thompson, R.S., Snyder W.H. and Weil J.C. (2000): Laboratort simulation of the rise of buoyant
thermals created by open detonation. J. Fluid. Mech., 417: 127-156

Tinarelli, G., D. Anfossi, M. Bider, E. Ferrero, S. Trini Castelli (2000): A new high performance
version of the Lagrangian particle dispersion model SPRAY, some case studies. Air Pollution
Modelling and its Applications XIII, S.E. Gryning and E. Batchvarova eds., Kluwer Academic /
Plenum Press, New York, 499-507. [Link]

Trelles J., K.B. McGrattan, H.R. Baum (1999a): Smoke transport by sheared winds. Combust.
Theory Modelling, 3:323-341.

Trelles J., K.B., McGrattan, H.R. Baum (1999b): Smoke dispersion from multiple fire plumes.
AIAA Journal, 37:1588-1601.

Turner, J.S. (1973): Buoyancy effects in fluid. Cambridge University Press, London.

Turner, D.B. (1985): Proposed pragmatic methods for estimating plume rise and plume
penetration through atmospheric layers. Atmos. Environ., 7:1215-1218.

Turner D.B., T. Chico and J.A. Catalano (1986) TUPOS - A multiple source gaussian dispersion
algorithm using on site turbulence data. EPA Publication No. EPA - 600/8 - 86/010 U.S.
Environmental Protection Agency, Research Triangle Park, North Carolina
6 Plume Rise 181

U.S. Environmental Protection Agency (1995a): SCREEN3 Model User’s Guide. Technical
Report EPA-454/B-95-004. Research Triangle Park, NC.
[Link]

U.S. Environmental Protection Agency (1995b): User’s Guide for the Industrial Source Complex
(ISC3) dispersion models. Volume II – description of model algorithms. Technical Report A-
454/B-95-003b. Research Triangle Park, NC. [Link]

U.S. Environmental Protection Agency (1998): Revised draft user’s guide for the AMS/EPA
regulatory model – AERMOD. Draft 11/10/98. [Link]

Vadot L. (1965): Study of diffusion of smoke plumes into the atmosphere (in French): Centre
Interprofessionel Technique d’Etudes de la Pollution Atmospheric, Paris

van Dop, H., F.T.M. Nieuwstadt, and J.C.R. Hunt (1985): Random Walk Models for Particle
Displacements in Inhomogeneous Unsteady Turbulent Flows. Phys. Fluids, 28:1639-1653.

van Dop, H. (1992): Buoyant plume rise in a Lagrangian Framework. Atmos. Environ.,
26A:1335-1346.

van Haren, L., and F.T. Nieuwstadt (1989): The behaviour of passive and buoyant plumes in a
convective boundary layer, as simulated with a large-eddy model. J. Appl. Meteor., 28:818-832.

Yamada (2000) Numerical simulations of airflows and tracer transport in the southwestern United
States. J. Appl. Meteor., 39: 399-411.

Weil, J.C. (1974): The rise of moist, buoyant plumes. J. Appl. Meteor., 13:435-443.

Weil, J.C. (1982): Source buoyancy effects in boundary layer diffusion. Workshop on the
parameterization of mixed layer diffusion. R. Cionco (Ed.), Physical Science Laboratory, New
Mexico State University, Las Cruces NM.

Weil, J.C. (1988): Plume Rise. In: Lectures on Air Pollution Modeling. A. Venkatram and J.C.
Wyngaard, eds. American Meteorological Society, Boston, 119-166.

Weil J.C. (1994) A hybrid Lagrangian dispersion model for elevated sources in the convective
boundary layer. Atmos. Environ., 28:3433-3448.

Weil J.C., L.A. Corio, and R.P. Brower (1997). APDF dispersion model for buoyant plumes in
the convective boundary layer. J. Appl. Meteoror., 36, 982-1003.

Wigley, T.M., and P.R. Slawson (1975): The effect of atmospheric conditions on the length of
visible cooling tower plumes. Atmos. Environ., 9:437-445.

Willis G. and J. Deardorff (1983): On plume rise within a convective boundary layer. Atmos.
Environ., 17:2435-2447.

Willis G. and J. Deardorff (1987). Buoyant plume dispersion and entrapment in and above a
laboratory mixed layer, Atmos. Environ., 21:1725-1735.

Zannetti, P., and N. Al-Madani (1983): Numerical simulations of Lagrangian particle diffusion by
Monte-Carlo techniques. VIth World Congress on Air Quality (IUAPPA), Paris, France, May
1983.
182 Air Quality Modeling – Vol. I

Zannetti, P., and N. Al-Madani (1984): Simulation of transformation, buoyancy and removal
processes by Lagrangian particle methods. Proceedings of the 14th International Technical
Meeting on Air Pollution Modelling and its Application (ed. Ch. de Wispelaere) Plenum Press,
New York, 733-744.

Zannetti, P. (1990): Air pollution modeling: theories, computational methods and available
software. Computational Mechanics Publications. Van Nostrand Reinhold, New York.

Zhang, X., and A.F.A. Ghoniem (1993): Computational Model for the Rise and Dispersion of
Wind-Blown, Buoyancy Driven Plumes - I. Neutrally Stratified Atmosphere. Atmos. Environ.,
27A:2295-2311.

Zhang, X., and A.F.A. Ghoniem (1994a): Computational Model for the Rise and Dispersion of
Wind-Blown, Buoyancy Driven Plumes - II. Linearly Stratified Atmosphere. Atmos. Environ.,
28:3005-3018.

Zhang, X., and A.F.A. Ghoniem (1994b): Computational Model for the Rise and Dispersion of
Wind-Blown, Buoyancy Driven Plumes - III. Penetration of Atmospheric Inversion. Atmos.
Environ., 28:3019-3032.